Introduction
Thin elastic filaments immersed in very viscous fluids play a central role in many biological processes, including mitotic spindle positioning during cell division [55 ] , cytoskeletal mechanics [44 ] , sperm motility [32 , 11 ] , and ciliary transport [13 ] . To capture the effective interaction between a thin semiflexible filament and the surrounding viscous fluid, it is natural to seek a lower dimensional representation of immersed filament dynamics. This is the basic tenet of slender body theories, which play an important role in many computational models.
Given an immersed filament with finite radius ϵ > 0 \epsilon> 0 ϵ > 0 and centerline position X ( s , t ) : R / Z × [ 0 , T ] → R 3 \mathbf{X}(s,t):\mathbb{R}/\mathbb{Z}\times[0,T]\to\mathbb{R}^{3} X ( s , t ) : R / Z × [ 0 , T ] → R 3 , the fundamental aim is to capture the filament evolution using an equation of the form
∂ X ∂ t = − L ϵ ( X ) [ ∂ s 4 X − ∂ s ( τ ∂ s X ) ] , ∣ ∂ s X ∣ = 1. (1) \frac{\partial\mathbf{X}}{\partial t}=-\mathcal{L}_{\epsilon}(\mathbf{X})\left[\partial_s^4\mathbf{X}-\partial_s(\tau\partial_s\mathbf{X})\right],\qquad|\partial_s\mathbf{X}|=1. \tag*{(1)} ∂ t ∂ X = − L ϵ ( X ) [ ∂ s 4 X − ∂ s ( τ ∂ s X ) ] , ∣ ∂ s X ∣ = 1. (1)
Here L ϵ ( X ) \mathcal{L}_{\epsilon}(\mathbf{X}) L ϵ ( X ) is an operator mapping force data defined along a 1D curve to a 3D velocity field governing the position of the filament centerline. This operator must also incorporate the effects of the surrounding 3D fluid on the filament evolution. The forcing terms − ∂ s 4 X + ∂ s ( τ ∂ s X ) -\partial_s^4\mathbf{X}+\partial_s(\tau\partial_s\mathbf{X}) − ∂ s 4 X + ∂ s ( τ ∂ s X ) come from Euler-Bernoulli beam theory and model a simple elastic response along the filament, with τ ( s , t ) \tau(s,t) τ ( s , t ) , the filament tension, serving as a Lagrange multiplier to enforce the inextensibility constraint ∣ ∂ s X ∣ = 1 |\partial_s\mathbf{X}|=1 ∣ ∂ s X ∣ = 1 . The key feature of this type of formulation (1) is that the filament dynamics are 1D. A major difficulty is coupling these 1D dynamics and the surrounding 3D fluid via the operator L ϵ \mathcal{L}_{\epsilon} L ϵ in both a physically and a mathematically reasonable way.
A simple and mathematically reasonable choice for this coupling operator is local slender body theory , L ϵ , l o c S B ( X ) = c ∣ log ϵ ∣ ( I + ∂ s X ⊗ ∂ s X ) \mathcal{L}_{\epsilon,\mathrm{loc}}^{\mathrm{SB}}(\mathbf{X})=c|\log\epsilon|(\mathbf{I}+\partial_s\mathbf{X}\otimes\partial_s\mathbf{X}) L ϵ , loc SB ( X ) = c ∣ log ϵ ∣ ( I + ∂ s X ⊗ ∂ s X ) [31 , 18 , 6 , 32 , 70 , 10 , 11 ] . For this very simple choice of coupling, the analogue of (1) is a fourth order parabolic equation for the filament position coupled with a 1D elliptic equation for the filament tension τ \tau τ . In addition to establishing well-posedness in this setting, the author and Mori [51 , 58 ] prove that the filament can generate net forward motion (i.e. swim) with a suitable choice of time-periodic forcing. However, L ϵ , l o c S B \mathcal{L}_{\epsilon,\mathrm{loc}}^{\mathrm{SB}} L ϵ , loc SB incorporates only a very simplified description of the surrounding fluid physics on the filament evolution.
To better capture the physics of the fluid-structure interaction, a natural attempt would be to use nonlocal slender body theory L ϵ , n l o c S B ( X ) \mathcal{L}_{\epsilon,\mathrm{nloc}}^{\mathrm{SB}}(\mathbf{X}) L ϵ , nloc SB ( X ) [36 , 34 , 69 ] , which, in addition to the terms from L ϵ , l o c S B \mathcal{L}_{\epsilon,\mathrm{loc}}^{\mathrm{SB}} L ϵ , loc SB , includes a 1D integral operator incorporating nonlocal interactions along the filament. This is a popular choice for computational models [41 , 64 , 68 , 40 , 17 , 4 , 60 , 43 , 45 ] because it is more physically realistic than L ϵ , l o c S B \mathcal{L}_{\epsilon,\mathrm{loc}}^{\mathrm{SB}} L ϵ , loc SB and the evolution (1) is still relatively straightforward to compute numerically. However, the classical L ϵ , n l o c S B \mathcal{L}_{\epsilon,\mathrm{nloc}}^{\mathrm{SB}} L ϵ , nloc SB operator is known to have issues at high wavenumbers ∼ 1 / ϵ \sim1/\epsilon ∼ 1/ ϵ leading to nonsensical mapping properties [30 , 50 , 69 , 63 ] , even in analytic regularity. This issue does not typically affect numerics, as these high frequencies are truncated by discretization. Even so, from a PDE perspective, L ϵ , n l o c S B \mathcal{L}_{\epsilon,\mathrm{nloc}}^{\mathrm{SB}} L ϵ , nloc SB cannot give rise to a well-posed evolution equation (1).
Thus, a central question is how to incorporate more of the fluid physics into the operator L ϵ \mathcal{L}_{\epsilon} L ϵ while still yielding a well-posed PDE evolution (1). In [52 , 53 ] , the author along with Mori and Spirn propose a novel candidate for the coupling operator L ϵ \mathcal{L}_{\epsilon} L ϵ in the form of an angle-averaged Neumann-to-Dirichlet operator for the 3D Stokes equations. This notion of slender body Neumann-to-Dirichlet (NtD) map is defined as the solution to a natural type of boundary value problem for the Stokes equations in the exterior of the thin filament (see equation (11)). It therefore incorporates much more of the fluid physics than local slender body theory but also has more reasonable mapping properties than nonlocal slender body theory. In [52 , 53 ] , we establish well-posedness for the static slender body boundary value problem in a natural energy space and moreover show that the solution is close in a suitable sense to nonlocal slender body theory. We have since extended this notion of slender body NtD map in various directions [49 , 57 , 50 , 46 , 59 ] , but all analysis has thus far been in the static setting.
Here we give a more complete answer to the central question above by establishing local-in-time well-posedness for a simplified version of (1) (see equation (20) ) using the slender body NtD map L ϵ \mathcal{L}_{\epsilon} L ϵ . In our simplified dynamics, the filament is allowed to change length over time, although not necessarily in a physically realistic way. This allows us to develop the theory for the main part of the evolution (1) without also needing to solve for the filament tension τ \tau τ . Developing this well-posedness theory requires a detailed understanding of the mapping properties of the map L ϵ \mathcal{L}_{\epsilon} L ϵ . The bulk of this paper is therefore devoted to proving Theorem 1.5, a technically demanding decomposition of L ϵ \mathcal{L}_{\epsilon} L ϵ into a main part plus lower order remainders which are smoother or small with respect to ϵ \epsilon ϵ .
Using a layer potential argument, we show that the principal behavior of L ϵ \mathcal{L}_{\epsilon} L ϵ is that of the corresponding operator about a straight cylinder with periodic boundary conditions, for which we can derive an explicit symbol. Using this explicit symbol, we show that the toy extensible version of (1) is third-order parabolic and locally well-posed. This provides a proof-of-concept for using L ϵ \mathcal{L}_{\epsilon} L ϵ to solve a dynamic free boundary problem, and represents an important step toward establishing well-posedness for the full evolution (1) and a more complete mathematical justification of slender body theories of filament dynamics in Stokes flow.
Our extraction of the straight operator as the leading order behavior of L ϵ \mathcal{L}_{\epsilon} L ϵ is a type of small-scale decomposition analogous to similar Dirichlet-to-Neumann operator decompositions used in other interfacial fluids problems. These include Hele-Shaw and water waves problems [1 , 3 , 7 , 33 , 39 ] , the Muskat problem [2 , 22 , 56 ] , and the Peskin problem for the evolution of a 1D filament in 2D Stokes flow [12 , 14 , 25 , 27 , 42 , 38 , 54 , 66 , 67 , 26 ] . In many of these other interfacial fluids problems, however, the principal part of the operator is simpler to state and analyze. In contrast, the slender body NtD operator L ϵ \mathcal{L}_{\epsilon} L ϵ requires a cylindrical geometry to be well-defined, and even for the straight cylinder, the symbol, while explicit, is not a simple expression. Considering a curved, closed filament as a perturbation of the straight filament and showing that the local effects due to curvature are lower order has also been used as a strategy to study vortex filament solutions of the Ginzburg-Landau, Navier-Stokes, and Euler equations [8 , 19 , 20 ] . We will also frequently refer to the author’s earlier work [59 ] deriving an analogous operator decomposition in the much simpler setting of the Laplace equation.
Geometry
We begin by defining the filament geometry at a fixed instant in time (for example, the initial configuration of the fiber). The geometric quantities described here will play a major role in the eventual extraction of the principal behavior for the slender body NtD map (Theorem 1.5). At this fixed time, we let X : T : = R / Z → R 3 X:\mathbb{T}:=\mathbb{R}/\mathbb{Z}\to\mathbb{R}^{3} X : T := R / Z → R 3 denote the arclength parametrization of a closed, unit-length curve Γ 0 \Gamma_{0} Γ 0 in R 3 \mathbb{R}^{3} R 3 (see Figure 1 ). We require that the curve belongs to C 2 , β ( T ) C^{2,\beta}(\mathbb{T}) C 2 , β ( T ) , 0 < β < 1 0<\beta<1 0 < β < 1 , and satisfies a non-self-intersection condition
Figure 1. A typical curved filament Σ ϵ \Sigma_\epsilon Σ ϵ as in (7) in comparison with the straight filament C ϵ C_\epsilon C ϵ given by (21) .
inf s ≠ s ′ ∣ X ( s ) − X ( s ′ ) ∣ ∣ s − s ′ ∣ ≥ c Γ (2) \inf_{s\ne s'}\frac{\lvert X(s)-X(s')\rvert}{\lvert s-s'\rvert}\ge c_{\Gamma} \tag*{(2)} s = s ′ inf ∣ s − s ′ ∣ ∣ X ( s ) − X ( s ′ )∣ ≥ c Γ (2)
for some constant c Γ > 0 c_{\Gamma}>0 c Γ > 0 . Taking e t ( s ) = d X d s ( s ) e_t(s)=\frac{dX}{ds}(s) e t ( s ) = d s d X ( s ) to be the unit tangent vector to the curve at s s s , points sufficiently close to Γ 0 \Gamma_{0} Γ 0 may be described by a C 1 , β C^{1,\beta} C 1 , β orthonormal frame ( e t ( s ) , e n 1 ( s ) , e n 2 ( s ) ) (e_t(s),e_{n_1}(s),e_{n_2}(s)) ( e t ( s ) , e n 1 ( s ) , e n 2 ( s )) satisfying the ODEs
d d s ( e t ( s ) e n 1 ( s ) e n 2 ( s ) ) = ( 0 κ 1 ( s ) κ 2 ( s ) − κ 1 ( s ) 0 κ 3 − κ 2 ( s ) − κ 3 0 ) ( e t ( s ) e n 1 ( s ) e n 2 ( s ) ) . (3) \frac{d}{ds}
\begin{pmatrix}
e_t(s)\\
e_{n_1}(s)\\
e_{n_2}(s)
\end{pmatrix}
=
\begin{pmatrix}
0 & \kappa_1(s) & \kappa_2(s)\\
-\kappa_1(s) & 0 & \kappa_3\\
-\kappa_2(s) & -\kappa_3 & 0
\end{pmatrix}
\begin{pmatrix}
e_t(s)\\
e_{n_1}(s)\\
e_{n_2}(s)
\end{pmatrix}. \tag*{(3)} d s d e t ( s ) e n 1 ( s ) e n 2 ( s ) = 0 − κ 1 ( s ) − κ 2 ( s ) κ 1 ( s ) 0 − κ 3 κ 2 ( s ) κ 3 0 e t ( s ) e n 1 ( s ) e n 2 ( s ) . (3)
The coefficients κ 1 \kappa_1 κ 1 and κ 2 \kappa_2 κ 2 are related to the shape of the curve via
κ 1 2 ( s ) + κ 2 2 ( s ) = κ 2 ( s ) , \kappa_1^2(s) + \kappa_2^2(s) = \kappa^2(s), κ 1 2 ( s ) + κ 2 2 ( s ) = κ 2 ( s ) ,
where κ 2 ( s ) = ∣ X s s ∣ 2 \kappa^2(s) = \lvert\mathbf{X}_{ss}\rvert^2 κ 2 ( s ) = ∣ X ss ∣ 2 is the squared curvature of X ( s ) X(s) X ( s ) . The coefficient κ 3 \kappa_3 κ 3 may be taken to be a constant satisfying ∣ κ 3 ∣ ≤ π \lvert\kappa_3\rvert\le\pi ∣ κ 3 ∣ ≤ π (see [52 Lemma 1.1] ). Throughout, we will use the notation
κ ∗ : = sup s ∈ T ∣ κ ( s ) ∣ , κ ∗ , β : = sup s ≠ s ′ s , s ′ ∈ T ∣ X s s ( s ) − X s s ( s ′ ) ∣ ∣ s − s ′ ∣ β + κ ∗ . (4) \kappa_* := \sup_{s\in\mathbb{T}} \lvert\kappa(s)\rvert,\qquad\kappa_{*,\beta} := \sup_{\substack{s\ne s'\\s,s'\in\mathbb{T}}} \frac{\lvert\mathbf{X}_{ss}(s)-\mathbf{X}_{ss}(s')\rvert}{\lvert s-s'\rvert^\beta}+\kappa_* . \tag*{(4)} κ ∗ := s ∈ T sup ∣ κ ( s )∣ , κ ∗ , β := s = s ′ s , s ′ ∈ T sup ∣ s − s ′ ∣ β ∣ X ss ( s ) − X ss ( s ′ )∣ + κ ∗ . (4)
Note that, using (3) , the coefficients κ ℓ \kappa_\ell κ ℓ , ℓ = 1 , 2 \ell=1,2 ℓ = 1 , 2 , satisfy
∥ κ ℓ ∥ C 0 , β ( T ) = ∥ X s s ⋅ e n ℓ ∥ C 0 , β ≤ κ ∗ , β . \lVert\kappa_\ell\rVert_{C^{0,\beta}(\mathbb{T})}=\lVert\mathbf{X}_{ss}\cdot\mathbf{e}_{n_\ell}\rVert_{C^{0,\beta}}\le\kappa_{*,\beta}. ∥ κ ℓ ∥ C 0 , β ( T ) = ∥ X ss ⋅ e n ℓ ∥ C 0 , β ≤ κ ∗ , β .
Given (3) , it will also be convenient to define the following cylindrical basis vectors:
e r ( s , θ ) = cos θ e n 1 ( s ) + sin θ e n 2 ( s ) , e θ ( s , θ ) = − sin θ e n 1 ( s ) + cos θ e n 2 ( s ) . (5) \mathbf{e}_r(s,\theta)=\cos\theta\mathbf{e}_{n_1}(s)+\sin\theta\mathbf{e}_{n_2}(s),\qquad\mathbf{e}_\theta(s,\theta)=-\sin\theta\mathbf{e}_{n_1}(s)+\cos\theta\mathbf{e}_{n_2}(s). \tag*{(5)} e r ( s , θ ) = cos θ e n 1 ( s ) + sin θ e n 2 ( s ) , e θ ( s , θ ) = − sin θ e n 1 ( s ) + cos θ e n 2 ( s ) . (5)
Now, for Γ 0 \Gamma_0 Γ 0 as above, there exists a maximal radius
0 < r ∗ = r ∗ ( c Γ , κ ∗ ) ≤ min { 1 2 κ ∗ , c Γ 2 } . (6) 0<r_*=r_*(c_{\Gamma},\kappa_*)\le\min\left\{\frac{1}{2\kappa_*},\frac{c_{\Gamma}}{2}\right\}. \tag*{(6)} 0 < r ∗ = r ∗ ( c Γ , κ ∗ ) ≤ min { 2 κ ∗ 1 , 2 c Γ } . (6)
such that we may uniquely parameterize points x x x satisfying dist ( x , Γ 0 ) < r ∗ \operatorname{dist}(x,\Gamma_0)<r_* dist ( x , Γ 0 ) < r ∗ as
x = X ( s ) + r e r ( s , θ ) , 0 ≤ r < r ∗ . x=\mathbf{X}(s)+r\mathbf{e}_r(s,\theta),\qquad0\le r<r_*. x = X ( s ) + r e r ( s , θ ) , 0 ≤ r < r ∗ .
For any 0 < ϵ < r ∗ 4 0<\epsilon<\frac{r_*}{4} 0 < ϵ < 4 r ∗ , we may define a slender filament with uniform radius ϵ \epsilon ϵ as
Σ ϵ : = { x ∈ R 3 : x = X ( s ) + r e r ( s , θ ) , s ∈ T , r < ϵ , θ ∈ 2 π T } . (7) \Sigma_\epsilon:=\left\{x\in\mathbb{R}^3:x=\mathbf{X}(s)+r\mathbf{e}_r(s,\theta),\ s\in\mathbb{T},\ r<\epsilon,\ \theta\in2\pi\mathbb{T}\right\}. \tag*{(7)} Σ ϵ := { x ∈ R 3 : x = X ( s ) + r e r ( s , θ ) , s ∈ T , r < ϵ , θ ∈ 2 π T } . (7)
The filament surface Γ ϵ = ∂ Σ ϵ \Gamma_\epsilon=\partial\Sigma_\epsilon Γ ϵ = ∂ Σ ϵ is given by
Γ ϵ : = { x ∈ R 3 : x = X ( s ) + ϵ e r ( s , θ ) , s ∈ T , θ ∈ 2 π T } , \Gamma_\epsilon:=\left\{x\in\mathbb{R}^3:x=\mathbf{X}(s)+\epsilon\mathbf{e}_r(s,\theta),\ s\in\mathbb{T},\ \theta\in2\pi\mathbb{T}\right\}, Γ ϵ := { x ∈ R 3 : x = X ( s ) + ϵ e r ( s , θ ) , s ∈ T , θ ∈ 2 π T } ,
with surface element d S ϵ = J ϵ ( s , θ ) d θ d s dS_\epsilon=J_\epsilon(s,\theta)\,d\theta ds d S ϵ = J ϵ ( s , θ ) d θ d s where J ϵ ( s , θ ) J_\epsilon(s,\theta) J ϵ ( s , θ ) is given by
J ϵ ( s , θ ) = ϵ ( 1 − ϵ κ ^ ( s , θ ) ) , κ ^ ( s , θ ) : = κ 1 ( s ) cos θ + κ 2 ( s ) sin θ . (8) J_\epsilon(s,\theta)=\epsilon\left(1-\epsilon\widehat{\kappa}(s,\theta)\right),\qquad\widehat{\kappa}(s,\theta):=\kappa_1(s)\cos\theta+\kappa_2(s)\sin\theta. \tag*{(8)} J ϵ ( s , θ ) = ϵ ( 1 − ϵ κ ( s , θ ) ) , κ ( s , θ ) := κ 1 ( s ) cos θ + κ 2 ( s ) sin θ . (8)
Let Σ ϵ ( a ) \Sigma_{\epsilon}^{(a)} Σ ϵ ( a ) and Σ ϵ ( b ) \Sigma_{\epsilon}^{(b)} Σ ϵ ( b ) be two closed filaments as in (7) with centerlines parameterized by X ( a ) ( s ) \mathbf{X}^{(a)}(s) X ( a ) ( s ) and X ( b ) ( s ) \mathbf{X}^{(b)}(s) X ( b ) ( s ) , respectively. We will be concerned with pairs of curves that are “close” in the following sense. Choose the basepoints X ( a ) ( 0 ) \mathbf{X}^{(a)}(0) X ( a ) ( 0 ) , X ( b ) ( 0 ) \mathbf{X}^{(b)}(0) X ( b ) ( 0 ) to be the nearest points on the two curves, i.e.
min s 1 , s 2 ∈ T ∣ X ( a ) ( s 1 ) − X ( b ) ( s 2 ) ∣ . \min_{s_1,s_2\in\mathbb{T}}\left|\mathbf{X}^{(a)}(s_1)-\mathbf{X}^{(b)}(s_2)\right|. s 1 , s 2 ∈ T min X ( a ) ( s 1 ) − X ( b ) ( s 2 ) .
Note that there may be multiple pairs of nearest points, in which case we may then select one such pair arbitrarily. Given this choice of basepoint, we consider pairs of curves satisfying
∥ X ( a ) − X ( b ) ∥ C 2 , β ( T ) = δ (9) \left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\beta}(\mathbb{T})}=\delta \tag*{(9)} X ( a ) − X ( b ) C 2 , β ( T ) = δ (9)
for δ ≥ 0 \delta\geq0 δ ≥ 0 sufficiently small. For two such nearby curves, the corresponding orthonormal frames may also be chosen to be close. In particular, from (9) we have that e t ( a ) = d X ( a ) d s \mathbf{e}_{t}^{(a)}=\frac{d\mathbf{X}^{(a)}}{ds} e t ( a ) = d s d X ( a ) and e t ( b ) = d X ( b ) d s \mathbf{e}_{t}^{(b)}=\frac{d\mathbf{X}^{(b)}}{ds} e t ( b ) = d s d X ( b ) satisfy
∥ e t ( a ) − e t ( b ) ∥ C 1 , β ( T ) ≤ δ , \left\|\mathbf{e}_{t}^{(a)}-\mathbf{e}_{t}^{(b)}\right\|_{C^{1,\beta}(\mathbb{T})}\leq\delta, e t ( a ) − e t ( b ) C 1 , β ( T ) ≤ δ ,
and there exists a systematic choice of normal vectors along each curve such that the frames ( e t ( a ) , e n 1 ( a ) , e n 2 ( a ) ) (\mathbf{e}_{t}^{(a)},\mathbf{e}_{n_1}^{(a)},\mathbf{e}_{n_2}^{(a)}) ( e t ( a ) , e n 1 ( a ) , e n 2 ( a ) ) , ( e t ( b ) , e n 1 ( b ) , e n 2 ( b ) ) (\mathbf{e}_{t}^{(b)},\mathbf{e}_{n_1}^{(b)},\mathbf{e}_{n_2}^{(b)}) ( e t ( b ) , e n 1 ( b ) , e n 2 ( b ) ) both satisfy (3) and the following lemma holds.
Lemma 1.1 (Nearby parameterizations for nearby curves). Given two closed filaments with centerlines X ( a ) ( s ) \mathbf{X}^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) \mathbf{X}^{(b)}(s) X ( b ) ( s ) satisfying (9) , there exists a choice of orthonormal frames ( e t ( a ) , e n 1 ( a ) , e n 2 ( a ) ) (\mathbf{e}_{t}^{(a)},\mathbf{e}_{n_1}^{(a)},\mathbf{e}_{n_2}^{(a)}) ( e t ( a ) , e n 1 ( a ) , e n 2 ( a ) ) and ( e t ( b ) , e n 1 ( b ) , e n 2 ( b ) ) (\mathbf{e}_{t}^{(b)},\mathbf{e}_{n_1}^{(b)},\mathbf{e}_{n_2}^{(b)}) ( e t ( b ) , e n 1 ( b ) , e n 2 ( b ) ) , respectively, satisfying (3) such that
∑ ℓ = 1 3 ∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ C 0 , β ( T ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , β ) ∥ X ( a ) − X ( b ) ∥ C 2 , β ( T ) . (10) \sum_{\ell=1}^{3}\left\|\kappa_{\ell}^{(a)}-\kappa_{\ell}^{(b)}\right\|_{C^{0,\beta}(\mathbb{T})}\leq c(\kappa_{*}^{(a)},\kappa_{*}^{(b)},\beta)\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\beta}(\mathbb{T})}. \tag*{(10)} ℓ = 1 ∑ 3 κ ℓ ( a ) − κ ℓ ( b ) C 0 , β ( T ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , β ) X ( a ) − X ( b ) C 2 , β ( T ) . (10)
The proof of Lemma 1.1 appears in Appendix A. Given two sufficiently nearby filaments, we will always choose a parameterization satisfying the estimate (10).
Slender body Neumann-to-Dirichlet map
To study an evolution equation of the form (1), we must first define the map L ϵ \mathcal{L}_{\epsilon} L ϵ being used to map force data defined along a 1D curve to a 3D velocity field governing the motion of the filament. Here we define L ϵ \mathcal{L}_{\epsilon} L ϵ as the solution to the following slender body boundary value problem , first introduced by the author along with Mori and Spirn in [52 ] . Given force data f ( s ) \mathbf{f}(s) f ( s ) , s ∈ T s\in\mathbb{T} s ∈ T , we consider ( u , p ) (\mathbf{u},p) ( u , p ) satisfying
− Δ u + ∇ p = 0 , div u = 0 in Ω ϵ = R 3 ∖ Σ ϵ ‾ , -\Delta\mathbf{u}+\nabla p=0,\qquad\operatorname{div}\mathbf{u}=0\qquad\text{in }\Omega_{\epsilon}=\mathbb{R}^{3}\setminus\overline{\Sigma_{\epsilon}}, − Δ u + ∇ p = 0 , div u = 0 in Ω ϵ = R 3 ∖ Σ ϵ ,
∫ 0 2 π ( σ n ) J ϵ ( s , θ ) d θ = f ( s ) on Γ ϵ , (11) \int_{0}^{2\pi}(\boldsymbol{\sigma}\mathbf{n})\mathcal{J}_{\epsilon}(s,\theta)\,d\theta=\mathbf{f}(s)\qquad\text{on }\Gamma_{\epsilon}, \tag*{(11)} ∫ 0 2 π ( σ n ) J ϵ ( s , θ ) d θ = f ( s ) on Γ ϵ , (11)
u ∣ Γ ϵ = v ( s ) , unknown but independent of θ , \left.\mathbf{u}\right|_{\Gamma_{\epsilon}}=\mathbf{v}(s),\qquad\text{unknown but independent of }\theta, u ∣ Γ ϵ = v ( s ) , unknown but independent of θ ,
and ∣ u ∣ → 0 |\mathbf{u}|\to0 ∣ u ∣ → 0 as ∣ x ∣ → ∞ |\mathbf{x}|\to\infty ∣ x ∣ → ∞ . Here σ = σ [ u ] = 1 2 ( ∇ u + ∇ u T ) − p I \boldsymbol{\sigma}=\boldsymbol{\sigma}[\mathbf{u}]=\frac{1}{2}(\nabla\mathbf{u}+\nabla\mathbf{u}^{\mathsf{T}})-p\mathbf{I} σ = σ [ u ] = 2 1 ( ∇ u + ∇ u T ) − p I is the viscous stress tensor and n ( x ) \mathbf{n}(\mathbf{x}) n ( x ) is the unit normal vector at x ∈ Γ ϵ \mathbf{x}\in\Gamma_{\epsilon} x ∈ Γ ϵ . The force data f ( s ) \mathbf{f}(s) f ( s ) is understood as the total surface stress σ [ u ] n \boldsymbol{\sigma}[\mathbf{u}]\mathbf{n} σ [ u ] n over each cross section of the filament, weighted by the fiber surface area via the Jacobian factor (8). The corresponding Dirichlet boundary value v ( s ) \mathbf{v}(s) v ( s ) along the filament surface is a priori unknown but constrained to be constant over each cross section of the fiber.
In [52 , 53 , 49 ] , the slender body BVP is shown to be well posed in a natural energy space (see section 6 for further details). Moreover, given f ∈ C 1 ( T ) \mathbf{f}\in C^{1}(\mathbb{T}) f ∈ C 1 ( T ) , the solution u ( x ) \mathbf{u}(\mathbf{x}) u ( x ) to (11) and its Dirichlet boundary value v ( s ) \mathbf{v}(s) v ( s ) are shown to be close in a quantitative sense to the corresponding nonlocal slender body theory expressions. In particular,
∥ v − L ϵ , n l o c S B [ f ] ∥ L 2 ( T ) ≤ c ( κ ∗ , c Γ ) ϵ ∣ log ϵ ∣ 3 / 2 ∥ f ∥ C 1 ( T ) . \left\|\mathbf{v}-\mathcal{L}_{\epsilon,\mathrm{nloc}}^{\mathrm{SB}}[\mathbf{f}]\right\|_{L^{2}(\mathbb{T})}\leq c(\kappa_{*},c_{\Gamma})\epsilon|\log\epsilon|^{3/2}\|\mathbf{f}\|_{C^{1}(\mathbb{T})}. v − L ϵ , nloc SB [ f ] L 2 ( T ) ≤ c ( κ ∗ , c Γ ) ϵ ∣ log ϵ ∣ 3/2 ∥ f ∥ C 1 ( T ) .
We define the Stokes slender body Neumann-to-Dirichlet (NtD) map L ϵ \mathcal{L}_{\epsilon} L ϵ as the operator
L ϵ : f ( s ) ↦ v ( s ) (12) \mathcal{L}_{\epsilon}: \boldsymbol{f}(s) \mapsto\boldsymbol{v}(s) \tag*{(12)} L ϵ : f ( s ) ↦ v ( s ) (12)
mapping the total force per cross section f ( s ) \boldsymbol{f}(s) f ( s ) to the θ \theta θ -independent velocity v ( s ) \boldsymbol{v}(s) v ( s ) along Γ ϵ \Gamma_{\epsilon} Γ ϵ by solving the boundary value problem (11). Our analysis hinges on a detailed understanding of this map. Given the nonlocal nature of the boundary conditions in (11), our analysis of L ϵ \mathcal{L}_{\epsilon} L ϵ will rely heavily on the fact that the inverse map, the slender body Dirichlet-to-Neumann (DtN) map ,
L ϵ − 1 : v ( s ) ↦ f ( s ) , (13) \mathcal{L}_{\epsilon}^{-1}: \boldsymbol{v}(s) \mapsto\boldsymbol{f}(s), \tag*{(13)} L ϵ − 1 : v ( s ) ↦ f ( s ) , (13)
is much simpler to study. In particular, L ϵ − 1 \mathcal{L}_{\epsilon}^{-1} L ϵ − 1 is just the usual Stokes Dirichlet-to-Neumann map along a surface using θ \theta θ -independent Dirichlet data, with an additional step of angle-averaging the resulting Neumann data.
To study the maps L ϵ \mathcal{L}_{\epsilon} L ϵ and L ϵ − 1 \mathcal{L}_{\epsilon}^{-1} L ϵ − 1 in detail, we will work in the setting of Hölder spaces along the filament surface Γ ϵ \Gamma_{\epsilon} Γ ϵ , as this will be a natural setting for analyzing the boundary integral formulation of (11) considered in section 1.2.4. For a vector-valued function h \boldsymbol{h} h defined along Γ ϵ \Gamma_{\epsilon} Γ ϵ , we denote
∥ h ∥ L ∞ ( Γ ϵ ) = ess sup x ∈ Γ ϵ ∣ h ( x ) ∣ , ∣ h ∣ C 0 , α ( Γ ϵ ) = sup x ≠ x ′ ∈ Γ ϵ ∣ h ( x ) − h ( x ′ ) ∣ ∣ x − x ′ ∣ α , 0 < α < 1 , (14) \|\boldsymbol{h}\|_{L^{\infty}(\Gamma_{\epsilon})}=\underset{\boldsymbol{x}\in\Gamma_{\epsilon}}{\operatorname{ess\,sup}}|\boldsymbol{h}(\boldsymbol{x})|,\qquad
|\boldsymbol{h}|_{C^{0,\alpha}(\Gamma_{\epsilon})}=\sup_{\boldsymbol{x}\ne\boldsymbol{x}'\in\Gamma_{\epsilon}}\frac{|\boldsymbol{h}(\boldsymbol{x})-\boldsymbol{h}(\boldsymbol{x}')|}{|\boldsymbol{x}-\boldsymbol{x}'|^{\alpha}},\qquad0<\alpha<1, \tag*{(14)} ∥ h ∥ L ∞ ( Γ ϵ ) = x ∈ Γ ϵ ess sup ∣ h ( x ) ∣ , ∣ h ∣ C 0 , α ( Γ ϵ ) = x = x ′ ∈ Γ ϵ sup ∣ x − x ′ ∣ α ∣ h ( x ) − h ( x ′ ) ∣ , 0 < α < 1 , (14)
and define the Hölder spaces C k , α ( Γ ϵ ) C^{k,\alpha}(\Gamma_{\epsilon}) C k , α ( Γ ϵ ) as
C k , α ( Γ ϵ ) = { h : ∥ h ∥ C k , α ( Γ ϵ ) < ∞ } , ∥ h ∥ C k , α ( Γ ϵ ) = ∑ ℓ = 0 k ∥ ∇ ℓ h ∥ L ∞ ( Γ ϵ ) + ∣ ∇ k h ∣ C 0 , α ( Γ ϵ ) . (15) C^{k,\alpha}(\Gamma_{\epsilon})=\{\boldsymbol{h}:\|\boldsymbol{h}\|_{C^{k,\alpha}(\Gamma_{\epsilon})}<\infty\},\qquad
\|\boldsymbol{h}\|_{C^{k,\alpha}(\Gamma_{\epsilon})}=\sum_{\ell=0}^{k}\|\nabla^{\ell}\boldsymbol{h}\|_{L^{\infty}(\Gamma_{\epsilon})}+|\nabla^{k}\boldsymbol{h}|_{C^{0,\alpha}(\Gamma_{\epsilon})}. \tag*{(15)} C k , α ( Γ ϵ ) = { h : ∥ h ∥ C k , α ( Γ ϵ ) < ∞ } , ∥ h ∥ C k , α ( Γ ϵ ) = ℓ = 0 ∑ k ∥ ∇ ℓ h ∥ L ∞ ( Γ ϵ ) + ∣ ∇ k h ∣ C 0 , α ( Γ ϵ ) . (15)
1.2.1. Evolution problem and theorem statement. Given the map L ϵ \mathcal{L}_{\epsilon} L ϵ defined by solving the slender body boundary value problem (11), we consider a simplified version of (1) without the inextensibility constraint. This simplifies many aspects of the evolution, as we avoid the need to simultaneously solve a tension determination problem for τ ( s , t ) \tau(s,t) τ ( s , t ) along with the evolution equation for X \boldsymbol{X} X [38 , 51 , 58 ] . This allows us to focus on the behavior of L ϵ \mathcal{L}_{\epsilon} L ϵ as the main difficulty in this paper, and develop the theory for the main part of the evolution (1), namely, L ϵ \mathcal{L}_{\epsilon} L ϵ applied to a fourth derivative of the filament centerline. The tension determination problem with L ϵ \mathcal{L}_{\epsilon} L ϵ will be considered in future work.
However, we now have to consider a filament that is possibly growing or shrinking in length over time in a non-uniform way. To treat this possible filament length change in the simplest way, we choose to artificially introduce some rescalings such that the main operator always acts on a unit length curve. We emphasize that this choice of dynamics has the same main term as (1).
For an extensible filament, we will consider the evolution of X σ ( σ , t ) \boldsymbol{X}^{\sigma}(\sigma,t) X σ ( σ , t ) where 0 ≤ σ ≤ 1 0\leq\sigma\leq1 0 ≤ σ ≤ 1 is now a material coordinate rather than an arclength coordinate. Let
λ ( t ) = ∫ 0 1 ∣ ∂ σ X σ ( σ , t ) ∣ d σ (16) \lambda(t)=\int_{0}^{1}|\partial_{\sigma}\boldsymbol{X}^{\sigma}(\sigma,t)|\,d\sigma \tag*{(16)} λ ( t ) = ∫ 0 1 ∣ ∂ σ X σ ( σ , t ) ∣ d σ (16)
denote the length of the filament at time t t t . We will require the filament to have length 1 at the initial time, with ∣ ∂ σ ( X σ ) i n ∣ = 1 |\partial_{\sigma}(\boldsymbol{X}^{\sigma})^{\mathrm{in}}|=1 ∣ ∂ σ ( X σ ) in ∣ = 1 at each σ \sigma σ .
The NtD operator L ϵ \mathcal{L}_{\epsilon} L ϵ depends on two arguments: a curve in arclength parameterization and a function of the arclength parameter. In (11), the problem is introduced for a unit length curve, but may be readily extended to more general curves. By making the change of variables x ~ = λ x \widetilde{\boldsymbol{x}}=\lambda\boldsymbol{x} x = λ x in (11), we can see how L ϵ \mathcal{L}_{\epsilon} L ϵ changes under uniform rescaling. For now, we suppress the time dependence in our notation. Let s ~ = λ s \widetilde{s}=\lambda s s = λ s for 0 ≤ s ≤ 1 0\leq s\leq1 0 ≤ s ≤ 1 and denote v ~ ( s ~ ) = v ( s ) \widetilde{\boldsymbol{v}}(\widetilde{s})=\boldsymbol{v}(s) v ( s ) = v ( s ) , f ~ ( s ~ ) = f ( s ) \widetilde{\boldsymbol{f}}(\widetilde{s})=\boldsymbol{f}(s) f ( s ) = f ( s ) , and X ~ ( s ~ ) = X ( s ) \widetilde{\boldsymbol{X}}(\widetilde{s})=\boldsymbol{X}(s) X ( s ) = X ( s ) . We then have
L λ ϵ ( X ~ ) [ f ~ ] ( s ~ ) = λ − 1 v ~ ( s ~ ) = λ − 1 v ( s ) = λ − 1 L ϵ ( X ) [ f ] ( s ) . (17) \mathcal{L}_{\lambda\epsilon}(\widetilde{\boldsymbol{X}})[\widetilde{\boldsymbol{f}}](\widetilde{s})
=\lambda^{-1}\widetilde{\boldsymbol{v}}(\widetilde{s})
=\lambda^{-1}\boldsymbol{v}(s)
=\lambda^{-1}\mathcal{L}_{\epsilon}(\boldsymbol{X})[\boldsymbol{f}](s). \tag*{(17)} L λ ϵ ( X ) [ f ] ( s ) = λ − 1 v ( s ) = λ − 1 v ( s ) = λ − 1 L ϵ ( X ) [ f ] ( s ) . (17)
We will consider the evolution
∂ X σ ∂ t ( σ , t ) = L λ ϵ [ X ~ ] [ f ~ ( s ~ , t ) ] ( λ σ ) , (18) \frac{\partial X^\sigma}{\partial t}(\sigma,t)=\mathcal{L}_{\lambda\epsilon}[\widetilde{X}][\widetilde{f}(\widetilde{s},t)](\lambda\sigma), \tag*{(18)} ∂ t ∂ X σ ( σ , t ) = L λ ϵ [ X ] [ f ( s , t )] ( λσ ) , (18)
where X σ X^\sigma X σ parameterizes the curve of interest and X ~ \widetilde{X} X is the same curve parameterized by arclength s ~ \widetilde{s} s . Here we will use the forcing term
f ~ ( s ~ , t ) = − ( ∂ σ 4 X σ ) ∣ σ = s ~ / λ , (19) \widetilde{f}(\widetilde{s},t)=-\left(\partial_\sigma^4X^\sigma\right)\bigg|_{\sigma=\widetilde{s}/\lambda}, \tag*{(19)} f ( s , t ) = − ( ∂ σ 4 X σ ) σ = s / λ , (19)
i.e., we take the fourth derivative with respect to the material parameter and evaluate at the rescaled arclength parameter 0 ≤ s ~ / λ ( t ) ≤ 1 0\leq\widetilde{s}/\lambda(t)\leq1 0 ≤ s / λ ( t ) ≤ 1 , where the rescaling occurs around a fixed basepoint X ( 0 , t ) X(0,t) X ( 0 , t ) . Note that the material points and rescaled arclength parameter do not necessarily match up precisely after the initial time. An additional nonphysical feature is that for convenience we will take a fixed ratio ϵ \epsilon ϵ between the filament radius and length, rather than having a fixed filament radius.
Putting (17) , (18) , and (19) together, we obtain the curve evolution
∂ X σ ∂ t ( σ , t ) = − λ ( t ) − 1 L ϵ ( R ~ 1 [ X σ ] ) [ ∂ σ 4 X σ ] , X σ ( σ , 0 ) = ( X σ ) i n ( σ ) , ∣ ∂ σ ( X σ ) i n ∣ = 1 , λ ( t ) = ∫ 0 1 ∣ ∂ σ X σ ∣ d σ . (20) \begin{aligned}
\frac{\partial X^\sigma}{\partial t}(\sigma,t)&=-\lambda(t)^{-1}\mathcal{L}_{\epsilon}(\widetilde{R}_1[X^\sigma])[\partial_\sigma^4X^\sigma],\\
X^\sigma(\sigma,0)&=(X^\sigma)^{\mathrm{in}}(\sigma),\qquad\left|\partial_\sigma(X^\sigma)^{\mathrm{in}}\right|=1,\\
\lambda(t)&=\int_0^1\left|\partial_\sigma X^\sigma\right|\,d\sigma. \tag*{(20)}
\end{aligned} ∂ t ∂ X σ ( σ , t ) X σ ( σ , 0 ) λ ( t ) = − λ ( t ) − 1 L ϵ ( R 1 [ X σ ]) [ ∂ σ 4 X σ ] , = ( X σ ) in ( σ ) , ∂ σ ( X σ ) in = 1 , = ∫ 0 1 ∣ ∂ σ X σ ∣ d σ . (20)
Here we use R ~ 1 [ X σ ] ( s ) \widetilde{R}_1[X^\sigma](s) R 1 [ X σ ] ( s ) to denote the arclength parameterization of the uniform rescaling of the curve X σ X^\sigma X σ to length 1. We show the following.
Theorem 1.2 (Local well-posedness for filament evolution). Given ( X σ ) i n ∈ C 4 , α ( T ) (X^\sigma)^{\mathrm{in}}\in C^{4,\alpha}(\mathbb{T}) ( X σ ) in ∈ C 4 , α ( T ) satisfying (2), there exists ϵ 0 > 0 \epsilon_0>0 ϵ 0 > 0 depending only on ∥ ( X σ ) i n ∥ C 4 , α ( T ) \|(X^\sigma)^{\mathrm{in}}\|_{C^{4,\alpha}(\mathbb{T})} ∥ ( X σ ) in ∥ C 4 , α ( T ) and c Γ c_\Gamma c Γ such that for 0 < ϵ < ϵ 0 0<\epsilon<\epsilon_0 0 < ϵ < ϵ 0 , there exists a time T > 0 T>0 T > 0 depending on ϵ \epsilon ϵ , ∥ ( X σ ) i n ∥ C 4 , α ( T ) \|(X^\sigma)^{\mathrm{in}}\|_{C^{4,\alpha}(\mathbb{T})} ∥ ( X σ ) in ∥ C 4 , α ( T ) , and c Γ c_\Gamma c Γ such that the evolution (20) admits a unique solution X σ ( σ , t ) ∈ L ∞ ( 0 , T ; C 4 , α ( T ) ) X^\sigma(\sigma,t)\in L^\infty(0,T;C^{4,\alpha}(\mathbb{T})) X σ ( σ , t ) ∈ L ∞ ( 0 , T ; C 4 , α ( T )) .
The proof of Theorem 1.2 in section 2.2 is relatively straightforward by design. It serves mainly as a proof-of-concept for the use of the slender body NtD operator L ϵ \mathcal{L}_\epsilon L ϵ for a dynamic curve evolution problem. The main content of this paper is a detailed decomposition of L ϵ \mathcal{L}_\epsilon L ϵ that we will introduce in the next sections and summarize in Theorem 1.5 . The decomposition of L ϵ = L ϵ ( X ) \mathcal{L}_\epsilon=\mathcal{L}_\epsilon(X) L ϵ = L ϵ ( X ) will be presented for a unit length centerline curve X ( s ) X(s) X ( s ) in arclength parameterization. We will typically suppress the curve dependence of the operator L ϵ \mathcal{L}_\epsilon L ϵ in our notation. The material parameter σ \sigma σ and curve X σ X^\sigma X σ will only be used to define the evolution (20) and prove Theorem 1.2 ; everything else will be in terms of X ( s ) X(s) X ( s ) .
Slender body NtD map for the straight cylinder
The proof of Theorem 1.2 relies heavily on the fact that we can derive very detailed information about the slender body NtD and its inverse, the slender body DtN, for the simple geometry of a straight cylinder with periodic boundary conditions. We define
C ϵ : = { x ∈ R 2 × T : x = s e z + ϵ e r ( θ ) , s ∈ T , θ ∈ 2 π T } . (21) C_\epsilon:=\left\{x\in\mathbb{R}^2\times\mathbb{T}:x=se_z+\epsilon e_r(\theta),\ s\in\mathbb{T},\ \theta\in2\pi\mathbb{T}\right\}. \tag*{(21)} C ϵ := { x ∈ R 2 × T : x = s e z + ϵ e r ( θ ) , s ∈ T , θ ∈ 2 π T } . (21)
Let L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ and L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 denote the slender body NtD and DtN maps along C ϵ C_\epsilon C ϵ . Due to symmetry, the behavior of L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ and L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 in this simple geometry decouples into directions tangential (e z e_z e z ) and normal (e x , e y e_x,e_y e x , e y ) to the filament centerline. The eigenfunctions of L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ and L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 in each direction are then just e 2 π i k s e z e^{2\pi iks}e_z e 2 π ik s e z , e 2 π i k s e x e^{2\pi iks}e_x e 2 π ik s e x , e 2 π i k s e y e^{2\pi iks}e_y e 2 π ik s e y , and we may consider the eigenvalues m ϵ , t − 1 ( k ) m_{\epsilon,t}^{-1}(k) m ϵ , t − 1 ( k ) and m ϵ , n − 1 ( k ) m_{\epsilon,n}^{-1}(k) m ϵ , n − 1 ( k ) of the slender body DtN L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 satisfying
L ‾ ϵ − 1 [ e 2 π i k s e z ] = m ϵ , t − 1 ( k ) e 2 π i k s e z , L ‾ ϵ − 1 [ e 2 π i k s e x ] = m ϵ , n − 1 ( k ) e 2 π i k s e x . (22) \overline{\mathcal{L}}_\epsilon^{-1}\left[e^{2\pi iks}e_z\right] = m_{\epsilon,t}^{-1}(k)e^{2\pi iks}e_z,\qquad\overline{\mathcal{L}}_\epsilon^{-1}\left[e^{2\pi iks}e_x\right] = m_{\epsilon,n}^{-1}(k)e^{2\pi iks}e_x . \tag*{(22)} L ϵ − 1 [ e 2 π ik s e z ] = m ϵ , t − 1 ( k ) e 2 π ik s e z , L ϵ − 1 [ e 2 π ik s e x ] = m ϵ , n − 1 ( k ) e 2 π ik s e x . (22)
Note that by symmetry the e y e_y e y direction is identical to the e x e_x e x direction. In [50 ] , the following explicit expressions are calculated for m ϵ , t − 1 ( k ) m_{\epsilon,t}^{-1}(k) m ϵ , t − 1 ( k ) and m ϵ , n − 1 ( k ) m_{\epsilon,n}^{-1}(k) m ϵ , n − 1 ( k ) .
Proposition 1.3 (Slender body DtN spectrum). The eigenvalues m ϵ , t − 1 ( k ) m_{\epsilon,t}^{-1}(k) m ϵ , t − 1 ( k ) , m ϵ , n − 1 ( k ) m_{\epsilon,n}^{-1}(k) m ϵ , n − 1 ( k ) of the straight slender body DtN map L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 satisfying (22) are given by
m ϵ , t − 1 ( k ) = 8 π 2 ϵ ∣ k ∣ K 1 2 2 K 0 K 1 + 2 π ϵ ∣ k ∣ ( K 0 2 − K 1 2 ) (23) m_{\epsilon,t}^{-1}(k) = \frac{8\pi^2\epsilon|k|K_1^2}{2K_0K_1 + 2\pi\epsilon|k|\left(K_0^2-K_1^2\right)} \tag*{(23)} m ϵ , t − 1 ( k ) = 2 K 0 K 1 + 2 π ϵ ∣ k ∣ ( K 0 2 − K 1 2 ) 8 π 2 ϵ ∣ k ∣ K 1 2 (23) m ϵ , n − 1 ( k ) = 4 π 2 ϵ ∣ k ∣ 4 K 1 2 K 2 + 2 π ϵ ∣ k ∣ K 1 ( K 1 2 − K 0 K 2 ) 2 K 0 K 1 K 2 + 2 π ϵ ∣ k ∣ ( K 1 2 ( K 0 + K 2 ) − 2 K 0 2 K 2 ) (24) m_{\epsilon,n}^{-1}(k) = 4\pi^2\epsilon|k|\frac{4K_1^2K_2 + 2\pi\epsilon|k|K_1\left(K_1^2-K_0K_2\right)}{2K_0K_1K_2 + 2\pi\epsilon|k|\left(K_1^2\left(K_0+K_2\right)-2K_0^2K_2\right)} \tag*{(24)} m ϵ , n − 1 ( k ) = 4 π 2 ϵ ∣ k ∣ 2 K 0 K 1 K 2 + 2 π ϵ ∣ k ∣ ( K 1 2 ( K 0 + K 2 ) − 2 K 0 2 K 2 ) 4 K 1 2 K 2 + 2 π ϵ ∣ k ∣ K 1 ( K 1 2 − K 0 K 2 ) (24) where each K j = K j ( 2 π ϵ ∣ k ∣ ) K_j=K_j(2\pi\epsilon|k|) K j = K j ( 2 π ϵ ∣ k ∣ ) , j = 0 , 1 , 2 j=0,1,2 j = 0 , 1 , 2 , is a j t h j^{\mathrm{th}} j th order modified Bessel function of the second kind.
In [50 ] , both m ϵ , t − 1 ( k ) m_{\epsilon,t}^{-1}(k) m ϵ , t − 1 ( k ) and m ϵ , n − 1 ( k ) m_{\epsilon,n}^{-1}(k) m ϵ , n − 1 ( k ) are shown to grow like ϵ ∣ k ∣ \epsilon|k| ϵ ∣ k ∣ as ∣ k ∣ → ∞ |k|\to\infty ∣ k ∣ → ∞ ; in particular, L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 maps the Sobolev space H m H^m H m to H m − 1 ( T ) H^{m-1}(\mathbb{T}) H m − 1 ( T ) . In addition, at low wavenumbers ∣ k ∣ ≲ 1 ϵ |k|\lesssim\frac{1}{\epsilon} ∣ k ∣ ≲ ϵ 1 , both m ϵ , t − 1 ( k ) m_{\epsilon,t}^{-1}(k) m ϵ , t − 1 ( k ) and m ϵ , n − 1 ( k ) m_{\epsilon,n}^{-1}(k) m ϵ , n − 1 ( k ) scale as ∣ log ( ϵ ∣ k ∣ ) ∣ − 1 |\log(\epsilon|k|)|^{-1} ∣ log ( ϵ ∣ k ∣ ) ∣ − 1 (see Figure 2 ). Here we will be working in Hölder spaces and thus will require a more detailed characterization of the behavior of m ϵ , t − 1 ( k ) m_{\epsilon,t}^{-1}(k) m ϵ , t − 1 ( k ) and m ϵ , n − 1 ( k ) m_{\epsilon,n}^{-1}(k) m ϵ , n − 1 ( k ) . We use the explicit expressions (23) and (24) to show the following.
Figure 2. Plot of the eigenvalue expressions in directions tangential (left) and normal (right) to the straight filament C ϵ C_\epsilon C ϵ . Here the horizontal axis is z = 2 π ϵ ∣ k ∣ z = 2\pi\epsilon|k| z = 2 π ϵ ∣ k ∣ . In both plots, the solid blue curves (“Direct”) are given by the expressions in Proposition 1.3 while the dotted curves (“BI”) are the boundary integral expressions (60) and (61), respectively.
Lemma 1.4 (Mapping properties of SB NtD and DtN along C ϵ C_\epsilon C ϵ ). Let L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ denote the slender body NtD map (12) along the straight filament C ϵ C_\epsilon C ϵ . Given total force data f ( s ) ∈ C 0 , α ( T ) \boldsymbol{f}(s)\in C^{0,\alpha}(\mathbb{T}) f ( s ) ∈ C 0 , α ( T ) along C ϵ C_\epsilon C ϵ with ∫ T f ( s ) d s = 0 \int_{\mathbb{T}}\boldsymbol{f}(s)\,ds=0 ∫ T f ( s ) d s = 0 , we have
∥ L ‾ ϵ [ f ] ∥ C 1 , α ( T ) ≤ c ( ∣ log ϵ ∣ ∥ f ∥ L ∞ ( T ) + ϵ − 1 ∣ f ∣ C ˙ 0 , α ( T ) ) . (25) \left\|\overline{\mathcal{L}}_\epsilon[\boldsymbol{f}]\right\|_{C^{1,\alpha}(\mathbb{T})}\leq c\left(|\log\epsilon|\left\|\boldsymbol{f}\right\|_{L^\infty(\mathbb{T})}+\epsilon^{-1}\left|\boldsymbol{f}\right|_{\dot{C}^{0,\alpha}(\mathbb{T})}\right). \tag*{(25)} L ϵ [ f ] C 1 , α ( T ) ≤ c ( ∣ log ϵ ∣ ∥ f ∥ L ∞ ( T ) + ϵ − 1 ∣ f ∣ C ˙ 0 , α ( T ) ) . (25) Letting L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 denote the SB DtN map (13) along C ϵ C_\epsilon C ϵ , given a θ \theta θ -independent velocity field v ( s ) ∈ C 1 , α ( T ) \boldsymbol{v}(s)\in C^{1,\alpha}(\mathbb{T}) v ( s ) ∈ C 1 , α ( T ) along C ϵ C_\epsilon C ϵ , we have
∥ L ‾ ϵ − 1 [ v ] ∥ C 0 , α ( T ) ≤ c ∣ log ϵ ∣ − 1 ∥ v ∥ C 1 , α ( T ) . (26) \left\|\overline{\mathcal{L}}_\epsilon^{-1}[\boldsymbol{v}]\right\|_{C^{0,\alpha}(\mathbb{T})}\leq c|\log\epsilon|^{-1}\left\|\boldsymbol{v}\right\|_{C^{1,\alpha}(\mathbb{T})}. \tag*{(26)} L ϵ − 1 [ v ] C 0 , α ( T ) ≤ c ∣ log ϵ ∣ − 1 ∥ v ∥ C 1 , α ( T ) . (26)
The proof of Lemma 1.4 appears in section 3.4.
Decomposition of the slender body DtN and NtD maps
The proof of Theorem 1.2 relies on a detailed decomposition of the slender body DtN and NtD maps for arbitrarily curved filaments. In particular, given a curved filament, we extract the straight operators L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 , L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ as the leading order behavior and show that the effects of curvature are lower order: either small in ϵ \epsilon ϵ or more regular. Given the explicit expressions for the behavior of L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 and L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ , this decomposition will allow us to go on to prove local well-posedness for the evolution (20).
In order to accomplish this, we will need to define a way of identifying vector valued functions defined on curvilinear basis vectors ( e t ( s ) , e n 1 ( s ) , e n 2 ( s ) ) (e_t(s),e_{n_1}(s),e_{n_2}(s)) ( e t ( s ) , e n 1 ( s ) , e n 2 ( s )) along Γ ϵ \Gamma_\epsilon Γ ϵ with vector valued functions defined on ( e z , e x , e y ) (e_z,e_x,e_y) ( e z , e x , e y ) along C ϵ C_\epsilon C ϵ . We define a map Φ \Phi Φ identifying tangential and normal components of vectors along C ϵ C_\epsilon C ϵ with tangential and normal components of vectors along Γ ϵ \Gamma_\epsilon Γ ϵ . In particular, given a vector-valued function g \boldsymbol{g} g along the straight filament C ϵ C_\epsilon C ϵ and vector-valued h \boldsymbol{h} h about the curved filament Γ ϵ \Gamma_\epsilon Γ ϵ , we identify the tangential and normal components in the straight versus curved setting via
Φ g = ( g ⋅ e z ) e t + ( g ⋅ e x ) e n 1 + ( g ⋅ e y ) e n 2 \Phi\mathbf{g}=(\mathbf{g}\cdot\mathbf{e}_z)\mathbf{e}_t+(\mathbf{g}\cdot\mathbf{e}_x)\mathbf{e}_{n_1}+(\mathbf{g}\cdot\mathbf{e}_y)\mathbf{e}_{n_2} Φ g = ( g ⋅ e z ) e t + ( g ⋅ e x ) e n 1 + ( g ⋅ e y ) e n 2
Φ − 1 h = ( h ⋅ e t ) e z + ( h ⋅ e n 1 ) e x + ( h ⋅ e n 2 ) e y . (27) \Phi^{-1}\mathbf{h}=(\mathbf{h}\cdot\mathbf{e}_t)\mathbf{e}_z+(\mathbf{h}\cdot\mathbf{e}_{n_1})\mathbf{e}_x+(\mathbf{h}\cdot\mathbf{e}_{n_2})\mathbf{e}_y. \tag*{(27)} Φ − 1 h = ( h ⋅ e t ) e z + ( h ⋅ e n 1 ) e x + ( h ⋅ e n 2 ) e y . (27)
For any h \mathbf{h} h defined along Γ ϵ \Gamma_\epsilon Γ ϵ , we define
h 0 Φ : = h − Φ ∫ T ( Φ − 1 h ) d s ; (28) \mathbf{h}_0^\Phi:=\mathbf{h}-\Phi\int_{\mathbb{T}}(\Phi^{-1}\mathbf{h})\,ds; \tag*{(28)} h 0 Φ := h − Φ ∫ T ( Φ − 1 h ) d s ; (28)
in particular, Φ − 1 h 0 Φ \Phi^{-1}\mathbf{h}_0^\Phi Φ − 1 h 0 Φ is mean-zero in s s s along the straight filament C ϵ C_\epsilon C ϵ .
Given the definition (27) of Φ \Phi Φ , we show the following.
Theorem 1.5 (Decomposition of SB DtN and NtD). Let 0 < α < γ < β < 1 0<\alpha<\gamma<\beta<1 0 < α < γ < β < 1 and consider a closed filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) X(s)\in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Let v ( s ) ∈ C 1 , α ( T ) v(s)\in C^{1,\alpha}(\mathbb{T}) v ( s ) ∈ C 1 , α ( T ) be θ \theta θ -independent Dirichlet data along Γ ϵ \Gamma_\epsilon Γ ϵ . There exists ϵ d = ϵ d ( c Γ , κ ∗ ) > 0 \epsilon_d=\epsilon_d(c_\Gamma,\kappa_*)>0 ϵ d = ϵ d ( c Γ , κ ∗ ) > 0 such that for 0 < ϵ ≤ ϵ d 0<\epsilon\leq\epsilon_d 0 < ϵ ≤ ϵ d , the slender body Dirichlet-to-Neumann operator L ϵ − 1 \mathcal{L}_\epsilon^{-1} L ϵ − 1 as defined in (13) may be decomposed as
L ϵ − 1 [ v ] ( s ) = Φ L ‾ ϵ − 1 [ Φ − 1 v ( s ) ] + R d , ϵ [ v ( s ) ] + R d , + [ v ( s ) ] (29) \mathcal{L}_\epsilon^{-1}[v](s)=\Phi\overline{\mathcal{L}}_\epsilon^{-1}[\Phi^{-1}v(s)]+\mathcal{R}_{d,\epsilon}[v(s)]+\mathcal{R}_{d,+}[v(s)] \tag*{(29)} L ϵ − 1 [ v ] ( s ) = Φ L ϵ − 1 [ Φ − 1 v ( s )] + R d , ϵ [ v ( s )] + R d , + [ v ( s )] (29) where the remainder terms satisfy
∥ R d , ϵ [ v ] ∥ C 0 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) \left\|\mathcal{R}_{d,\epsilon}[v]\right\|_{C^{0,\alpha}(\mathbb{T})}\leq c(\kappa_{*,\alpha^+},c_\Gamma)\epsilon^{2-\alpha^+}\|v\|_{C^{1,\alpha}(\mathbb{T})} ∥ R d , ϵ [ v ] ∥ C 0 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) ∥ R d , + [ v ] ∥ C 0 , γ ( T ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) (30) \left\|\mathcal{R}_{d,+}[v]\right\|_{C^{0,\gamma}(\mathbb{T})}\leq c(\epsilon,\kappa_{*,\gamma^+},c_\Gamma)\|v\|_{C^{1,\alpha}(\mathbb{T})} \tag*{(30)} ∥ R d , + [ v ] ∥ C 0 , γ ( T ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) (30) for any α + ∈ ( α , β ] \alpha^+\in(\alpha,\beta] α + ∈ ( α , β ] and γ + ∈ ( γ , β ] \gamma^+\in(\gamma,\beta] γ + ∈ ( γ , β ] . Furthermore, given two nearby filaments with centerlines X ( a ) ( s ) X^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) X^{(b)}(s) X ( b ) ( s ) satisfying Lemma 1.1, the differences between their corresponding remainder terms R d , ϵ ( a ) − R d , ϵ ( b ) \mathcal{R}_{d,\epsilon}^{(a)}-\mathcal{R}_{d,\epsilon}^{(b)} R d , ϵ ( a ) − R d , ϵ ( b ) and R d , + ( a ) − R d , + ( b ) \mathcal{R}_{d,+}^{(a)}-\mathcal{R}_{d,+}^{(b)} R d , + ( a ) − R d , + ( b ) satisfy
∥ ( R d , ϵ ( a ) − R d , ϵ ( b ) ) [ v ] ∥ C 0 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 2 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ v ∥ C 1 , α ( T ) \left\|\left(\mathcal{R}_{d,\epsilon}^{(a)}-\mathcal{R}_{d,\epsilon}^{(b)}\right)[v]\right\|_{C^{0,\alpha}(\mathbb{T})}\leq c(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_\Gamma)\epsilon^{2-\alpha^+}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha^+}(\mathbb{T})}\|v\|_{C^{1,\alpha}(\mathbb{T})} ( R d , ϵ ( a ) − R d , ϵ ( b ) ) [ v ] C 0 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 2 − α + X ( a ) − X ( b ) C 2 , α + ( T ) ∥ v ∥ C 1 , α ( T ) ∥ ( R d , + ( a ) − R d , + ( b ) ) [ v ] ∥ C 0 , γ ( T ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ( T ) ∥ v ∥ C 1 , α ( T ) . (31) \left\|\left(\mathcal{R}_{d,+}^{(a)}-\mathcal{R}_{d,+}^{(b)}\right)[v]\right\|_{C^{0,\gamma}(\mathbb{T})}\leq c(\epsilon,\kappa_{*,\gamma^+}^{(a)},\kappa_{*,\gamma^+}^{(b)},c_\Gamma)\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma^+}(\mathbb{T})}\|v\|_{C^{1,\alpha}(\mathbb{T})}. \tag*{(31)} ( R d , + ( a ) − R d , + ( b ) ) [ v ] C 0 , γ ( T ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) X ( a ) − X ( b ) C 2 , γ + ( T ) ∥ v ∥ C 1 , α ( T ) . (31) For a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 3 , α ( T ) X(s)\in C^{3,\alpha}(\mathbb{T}) X ( s ) ∈ C 3 , α ( T ) , let f ( s ) ∈ C 0 , α ( T ) f(s)\in C^{0,\alpha}(\mathbb{T}) f ( s ) ∈ C 0 , α ( T ) be slender body force data. There exists ϵ n = ϵ n ( c Γ , κ ∗ , γ ) > 0 \epsilon_n=\epsilon_n(c_\Gamma,\kappa_{*,\gamma})>0 ϵ n = ϵ n ( c Γ , κ ∗ , γ ) > 0 such that for 0 < ϵ ≤ ϵ n 0<\epsilon\leq\epsilon_n 0 < ϵ ≤ ϵ n , the slender body Neumann-to-Dirichlet operator L ϵ \mathcal{L}_\epsilon L ϵ given by (12) may be similarly decomposed as
L ϵ [ f ] ( s ) = ( I + R n , ϵ ) [ Φ L ‾ ϵ [ Φ − 1 f 0 Φ ( s ) ] ] + R n , + [ f ( s ) ] (32) \mathcal{L}_\epsilon[f](s)=(\mathbf{I}+\mathcal{R}_{n,\epsilon})\left[\Phi\overline{\mathcal{L}}_\epsilon\left[\Phi^{-1}f_0^\Phi(s)\right]\right]+\mathcal{R}_{n,+}[f(s)] \tag*{(32)} L ϵ [ f ] ( s ) = ( I + R n , ϵ ) [ Φ L ϵ [ Φ − 1 f 0 Φ ( s ) ] ] + R n , + [ f ( s )] (32) where the remainder terms satisfy
∥ R n , ϵ [ g ] ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ g ∥ C 1 , α ( T ) \left\|\mathcal{R}_{n,\epsilon}[g]\right\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\kappa_{*,\alpha^+},c_\Gamma)\frac{\epsilon^{1-\alpha^+}}{|\log\epsilon|}\|g\|_{C^{1,\alpha}(\mathbb{T})} ∥ R n , ϵ [ g ] ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ∣ log ϵ ∣ ϵ 1 − α + ∥ g ∥ C 1 , α ( T ) ∥ R n , + [ f ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) ∥ f ∥ C 0 , α ( T ) (33) \left\|\mathcal{R}_{n,+}[f]\right\|_{C^{1,\gamma}(\mathbb{T})}\leq c(\epsilon,\|X_{ss}\|_{C^{1,\alpha}},c_\Gamma)\|f\|_{C^{0,\alpha}(\mathbb{T})} \tag*{(33)} ∥ R n , + [ f ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) ∥ f ∥ C 0 , α ( T ) (33) for any α + ∈ ( α , β ] \alpha^+\in(\alpha,\beta] α + ∈ ( α , β ] . Given two nearby filaments with centerlines X ( a ) ( s ) X^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) X^{(b)}(s) X ( b ) ( s ) in C 3 , α C^{3,\alpha} C 3 , α and satisfying Lemma 1.1, we also have
∥ ( R n , ϵ ( a ) − R n , ϵ ( b ) ) [ g ] ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ g ∥ C 1 , α ( T ) \left\|\left(\mathcal{R}_{n,\epsilon}^{(a)}-\mathcal{R}_{n,\epsilon}^{(b)}\right)[g]\right\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_\Gamma)\frac{\epsilon^{1-\alpha^+}}{|\log\epsilon|}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha^+}}\|g\|_{C^{1,\alpha}(\mathbb{T})} ( R n , ϵ ( a ) − R n , ϵ ( b ) ) [ g ] C 1 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ∣ log ϵ ∣ ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ g ∥ C 1 , α ( T ) ∥ ( R n , + ( a ) − R n , + ( b ) ) [ f ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , ∥ X s s ( a ) ∥ C 1 , α , ∥ X s s ( b ) ∥ C 1 , α , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ f ∥ C 0 , α ( T ) . (34) \left\|\left(\mathcal{R}_{n,+}^{(a)}-\mathcal{R}_{n,+}^{(b)}\right)[f]\right\|_{C^{1,\gamma}(\mathbb{T})}\leq c(\epsilon,\|X_{ss}^{(a)}\|_{C^{1,\alpha}},\|X_{ss}^{(b)}\|_{C^{1,\alpha}},c_\Gamma)\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha^+}}\|f\|_{C^{0,\alpha}(\mathbb{T})}. \tag*{(34)} ( R n , + ( a ) − R n , + ( b ) ) [ f ] C 1 , γ ( T ) ≤ c ( ϵ , ∥ X ss ( a ) ∥ C 1 , α , ∥ X ss ( b ) ∥ C 1 , α , c Γ ) X ( a ) − X ( b ) C 2 , α + ∥ f ∥ C 0 , α ( T ) . (34)
For both L ϵ − 1 \mathcal{L}_\epsilon^{-1} L ϵ − 1 and L ϵ \mathcal{L}_\epsilon L ϵ , the ϵ \epsilon ϵ -dependence is explicit in the bound for the remainder terms R j , ϵ \mathcal{R}_{j,\epsilon} R j , ϵ but not for the smoother remainders R j , + \mathcal{R}_{j,+} R j , + . Although only the decomposition (32) of L ϵ \mathcal{L}_\epsilon L ϵ is required to show Theorem 1.2 , the proof of this representation is dependent on the decomposition (29) of L ϵ − 1 \mathcal{L}_{\epsilon}^{-1} L ϵ − 1 , which is simpler to study and forms the bulk of this paper. We thus state both results together.
Given force data f ( s ) \boldsymbol{f}(s) f ( s ) involving multiple derivatives in s s s , as in (1), we may refine the decomposition (32) of L ϵ \mathcal{L}_{\epsilon} L ϵ as follows.
Corollary 1.6 (SB NtD decomposition involving derivatives). For 0 < α < γ < 1 0 < \alpha< \gamma< 1 0 < α < γ < 1 , given X ( s ) ∈ C 3 , γ ( T ) \boldsymbol{X}(s) \in C^{3,\gamma}(\mathbb{T}) X ( s ) ∈ C 3 , γ ( T ) and slender body force data of the form f ( s ) = ∂ s 4 F \boldsymbol{f}(s) = \partial_s^4 \boldsymbol{F} f ( s ) = ∂ s 4 F for some F ∈ C 4 , α ( T ) \boldsymbol{F} \in C^{4,\alpha}(\mathbb{T}) F ∈ C 4 , α ( T ) , the decomposition (32) of L ϵ \mathcal{L}_{\epsilon} L ϵ simplifies to
L ϵ [ ∂ s 4 F ] = ( I + R n , ϵ ) [ L ϵ [ ∂ s 4 F ] ] + R ~ n , + [ ∂ s 4 F ] , \mathcal{L}_{\epsilon}[\partial_s^4\boldsymbol{F}] = (\mathbf{I} + \mathcal{R}_{n,\epsilon})[\mathcal{L}_{\epsilon}[\partial_s^4\boldsymbol{F}]] + \widetilde{\mathcal{R}}_{n,+}[\partial_s^4\boldsymbol{F}], L ϵ [ ∂ s 4 F ] = ( I + R n , ϵ ) [ L ϵ [ ∂ s 4 F ]] + R n , + [ ∂ s 4 F ] , ∥ R n , ϵ [ L ϵ [ ∂ s 4 F ] ] ] ∥ C 1 , α ( T ) ≤ c ( ∥ X ∥ C 2 , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ L ϵ [ ∂ s 4 F ] ∥ C 1 , α ( T ) , (35) \left\|\mathcal{R}_{n,\epsilon}\left[\mathcal{L}_{\epsilon}[\partial_s^4\boldsymbol{F}]]\right]\right\|_{C^{1,\alpha}(\mathbb{T})} \leq c\left(\|\boldsymbol{X}\|_{C^{2,\alpha_+}},c_{\Gamma}\right)\epsilon^{1-\alpha_+}|\log\epsilon|\left\|\mathcal{L}_{\epsilon}[\partial_s^4\boldsymbol{F}]\right\|_{C^{1,\alpha}(\mathbb{T})}, \tag*{(35)} R n , ϵ [ L ϵ [ ∂ s 4 F ]] ] C 1 , α ( T ) ≤ c ( ∥ X ∥ C 2 , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ L ϵ [ ∂ s 4 F ] C 1 , α ( T ) , (35) ∥ R ~ n , + [ ∂ s 4 F ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , ∥ X ∥ C 3 , γ , c Γ ) ∥ F ∥ C 4 , α ( T ) , \left\|\widetilde{\mathcal{R}}_{n,+}[\partial_s^4\boldsymbol{F}]\right\|_{C^{1,\gamma}(\mathbb{T})} \leq c\left(\epsilon,\|\boldsymbol{X}\|_{C^{3,\gamma}},c_{\Gamma}\right)\|\boldsymbol{F}\|_{C^{4,\alpha}(\mathbb{T})}, R n , + [ ∂ s 4 F ] C 1 , γ ( T ) ≤ c ( ϵ , ∥ X ∥ C 3 , γ , c Γ ) ∥ F ∥ C 4 , α ( T ) , where α < α + < 1 \alpha< \alpha_+ < 1 α < α + < 1 . Furthermore, given two nearby filaments with centerlines X ( a ) ( s ) \boldsymbol{X}^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) \boldsymbol{X}^{(b)}(s) X ( b ) ( s ) in C 3 , γ C^{3,\gamma} C 3 , γ satisfying Lemma 1.1, we have
∥ ( R n , ϵ ( a ) − R n , ϵ ( b ) ) [ L ϵ [ ∂ s 4 F ] ] ∥ C 1 , α ( T ) ≤ c ( ∥ X ( a ) ∥ C 2 , α + , ∥ X ( b ) ∥ C 2 , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ L ϵ [ ∂ s 4 F ] ∥ C 1 , α ( T ) , (36) \left\|(\mathcal{R}_{n,\epsilon}^{(a)}-\mathcal{R}_{n,\epsilon}^{(b)})[\mathcal{L}_{\epsilon}[\partial_s^4\boldsymbol{F}]]\right\|_{C^{1,\alpha}(\mathbb{T})}
\leq c\left(\|\boldsymbol{X}^{(a)}\|_{C^{2,\alpha_+}},\|\boldsymbol{X}^{(b)}\|_{C^{2,\alpha_+}},c_{\Gamma}\right)\epsilon^{1-\alpha_+}|\log\epsilon|\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{2,\alpha_+}(\mathbb{T})}\left\|\mathcal{L}_{\epsilon}[\partial_s^4\boldsymbol{F}]\right\|_{C^{1,\alpha}(\mathbb{T})}, \tag*{(36)} ( R n , ϵ ( a ) − R n , ϵ ( b ) ) [ L ϵ [ ∂ s 4 F ]] C 1 , α ( T ) ≤ c ( ∥ X ( a ) ∥ C 2 , α + , ∥ X ( b ) ∥ C 2 , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ X ( a ) − X ( b ) C 2 , α + ( T ) L ϵ [ ∂ s 4 F ] C 1 , α ( T ) , (36) ∥ ( R ~ n , + ( a ) − R ~ n , + ( b ) ) [ ∂ s 4 F ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , ∥ X ( a ) ∥ C 3 , γ , ∥ X ( b ) ∥ C 3 , γ , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 3 , γ ( T ) ∥ F ∥ C 4 , α ( T ) . \left\|(\widetilde{\mathcal{R}}_{n,+}^{(a)}-\widetilde{\mathcal{R}}_{n,+}^{(b)})[\partial_s^4\boldsymbol{F}]\right\|_{C^{1,\gamma}(\mathbb{T})}
\leq c\left(\epsilon,\|\boldsymbol{X}^{(a)}\|_{C^{3,\gamma}},\|\boldsymbol{X}^{(b)}\|_{C^{3,\gamma}},c_{\Gamma}\right)\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{3,\gamma}(\mathbb{T})}\|\boldsymbol{F}\|_{C^{4,\alpha}(\mathbb{T})}. ( R n , + ( a ) − R n , + ( b ) ) [ ∂ s 4 F ] C 1 , γ ( T ) ≤ c ( ϵ , ∥ X ( a ) ∥ C 3 , γ , ∥ X ( b ) ∥ C 3 , γ , c Γ ) X ( a ) − X ( b ) C 3 , γ ( T ) ∥ F ∥ C 4 , α ( T ) .
The proof of Theorem 1.5 relies on a series of key lemmas, which we will state first before proving Theorem 1.5 in section 2.1. The proof of Corollary 1.6 will appear immediately after the proof of Theorem 1.5 . The proofs of each of these key lemmas then form the remainder of the paper.
Boundary integral formulation of slender body DtN map
Motivated by our analysis in the Laplace setting [59 ] , the strategy for understanding the slender body NtD map will rely on an explicit boundary integral representation for the inverse map, the slender body DtN map, for which we can more easily obtain estimates.
We use G ( x , x ′ ) \boldsymbol{G}(x,x') G ( x , x ′ ) to denote the Stokeslet, the free-space Green’s function for the Stokes equations, given by the matrix
G ( x , x ′ ) = 1 8 π ( I ∣ x − x ′ ∣ + ( x − x ′ ) ⊗ ( x − x ′ ) ∣ x − x ′ ∣ 3 ) . (37) \boldsymbol{G}(x,x') = \frac{1}{8\pi}\left(\frac{\mathbf{I}}{|x-x'|}+\frac{(x-x')\otimes(x-x')}{|x-x'|^3}\right). \tag*{(37)} G ( x , x ′ ) = 8 π 1 ( ∣ x − x ′ ∣ I + ∣ x − x ′ ∣ 3 ( x − x ′ ) ⊗ ( x − x ′ ) ) . (37)
For x ∈ Γ ϵ x \in\Gamma_{\epsilon} x ∈ Γ ϵ , let n ( x ) \boldsymbol{n}(x) n ( x ) denote the unit normal vector to Γ ϵ \Gamma_{\epsilon} Γ ϵ pointing out from the slender body Σ ϵ \Sigma_{\epsilon} Σ ϵ , into the fluid domain Ω ϵ \Omega_{\epsilon} Ω ϵ . We additionally define the stresslet kernel1
K D ( x , x ′ ; n ( x ′ ) ) = 3 4 π ( x − x ′ ) ⊗ ( x − x ′ ) ∣ x − x ′ ∣ 5 ( x − x ′ ) ⋅ n ( x ′ ) . (38) K_D(x,x';\boldsymbol{n}(x')) = \frac{3}{4\pi}\frac{(x-x')\otimes(x-x')}{|x-x'|^5}(x-x')\cdot\boldsymbol{n}(x'). \tag*{(38)} K D ( x , x ′ ; n ( x ′ )) = 4 π 3 ∣ x − x ′ ∣ 5 ( x − x ′ ) ⊗ ( x − x ′ ) ( x − x ′ ) ⋅ n ( x ′ ) . (38)
Given a continuous φ : Γ ϵ → R 3 \boldsymbol{\varphi}:\Gamma_{\epsilon}\to\mathbb{R}^3 φ : Γ ϵ → R 3 , for x ∈ Ω ϵ x\in\Omega_{\epsilon} x ∈ Ω ϵ , we define the single layer operator S \mathcal{S} S and double layer operator D \mathcal{D} D as
S [ φ ] ( x ) = ∫ Γ ϵ G ( x , x ′ ) φ ( x ′ ) d S x ′ , D [ φ ] ( x ) = ∫ Γ ϵ K D ( x , x ′ ; n ( x ′ ) ) φ ( x ′ ) d S x ′ . (39) \mathcal{S}[\boldsymbol{\varphi}](x)=\int_{\Gamma_{\epsilon}}\boldsymbol{G}(x,x')\boldsymbol{\varphi}(x')\,dS_{x'},\qquad\mathcal{D}[\boldsymbol{\varphi}](x)=\int_{\Gamma_{\epsilon}}K_D(x,x';\boldsymbol{n}(x'))\boldsymbol{\varphi}(x')\,dS_{x'}. \tag*{(39)} S [ φ ] ( x ) = ∫ Γ ϵ G ( x , x ′ ) φ ( x ′ ) d S x ′ , D [ φ ] ( x ) = ∫ Γ ϵ K D ( x , x ′ ; n ( x ′ )) φ ( x ′ ) d S x ′ . (39)
As in the Laplace setting [37 , 65 , 23 ] , the single layer operator is continuous up to Γ ϵ \Gamma_{\epsilon} Γ ϵ , while the double layer operator satisfies the exterior jump relation (see [62 eq. (2.3.12)] ):
lim h → 0 D [ φ ] ( x + h n ( x ) ) = D [ φ ] ( x ) + 1 2 φ ( x ) . (40) \lim_{h\to0}D[\varphi](x+h n(x))=D[\varphi](x)+\frac{1}{2}\varphi(x). \tag*{(40)} h → 0 lim D [ φ ] ( x + hn ( x )) = D [ φ ] ( x ) + 2 1 φ ( x ) . (40)
Following [62 eq. (2.3.10)] , the Stokes flow u ( x ) \boldsymbol{u}(x) u ( x ) at a point x ∈ Ω ϵ x\in\Omega_{\epsilon} x ∈ Ω ϵ may be represented by the boundary integral expression
u ( x ) = S [ w ] ( x ) + D [ v ] ( x ) (41) \boldsymbol{u}(x)=S[\boldsymbol{w}](x)+D[\boldsymbol{v}](x) \tag*{(41)} u ( x ) = S [ w ] ( x ) + D [ v ] ( x ) (41)
where, for x ∈ Γ ϵ x\in\Gamma_{\epsilon} x ∈ Γ ϵ ,
w ( x ) = − σ [ u ] n ( x ) ∣ Γ ϵ , \boldsymbol{w}(x)=-\boldsymbol{\sigma}[\boldsymbol{u}]\boldsymbol{n}(x)\big|_{\Gamma_{\epsilon}}, w ( x ) = − σ [ u ] n ( x ) Γ ϵ ,
v ( x ) = u ∣ Γ ϵ . (42) \boldsymbol{v}(x)=\boldsymbol{u}\big|_{\Gamma_{\epsilon}}. \tag*{(42)} v ( x ) = u Γ ϵ . (42)
The negative sign in w \boldsymbol{w} w arises due to the convention that the normal vector n ( x ) \boldsymbol{n}(x) n ( x ) points into the fluid domain Ω ϵ \Omega_{\epsilon} Ω ϵ , making − σ [ u ] n ( x ) -\boldsymbol{\sigma}[\boldsymbol{u}]\boldsymbol{n}(x) − σ [ u ] n ( x ) the Neumann boundary value for u \boldsymbol{u} u . Using (40) in (41), on the filament surface Γ ϵ \Gamma_{\epsilon} Γ ϵ , we obtain a relation between the Dirichlet (v \boldsymbol{v} v ) and Neumann (w \boldsymbol{w} w ) boundary values for u \boldsymbol{u} u :
1 2 v ( x ) − D [ v ] ( x ) = S [ w ] ( x ) , x ∈ Γ ϵ . (43) \frac{1}{2}\boldsymbol{v}(x)-D[\boldsymbol{v}](x)=S[\boldsymbol{w}](x),\qquad x\in\Gamma_{\epsilon}. \tag*{(43)} 2 1 v ( x ) − D [ v ] ( x ) = S [ w ] ( x ) , x ∈ Γ ϵ . (43)
Using (43), we may thus relate the slender body Dirichlet data v ( s ) \boldsymbol{v}(s) v ( s ) and angle-averaged Neumann data f ( s ) \boldsymbol{f}(s) f ( s ) from (11) as
1 2 v ( s ) − D [ v ( s ) ] = S [ w ( s , θ ) ] \frac{1}{2}\boldsymbol{v}(s)-D[\boldsymbol{v}(s)]=S[\boldsymbol{w}(s,\theta)] 2 1 v ( s ) − D [ v ( s )] = S [ w ( s , θ )]
∫ 0 2 π w ( s , θ ) J ϵ ( s , θ ) d θ = f ( s ) (44) \int_{0}^{2\pi}\boldsymbol{w}(s,\theta)J_{\epsilon}(s,\theta)\,d\theta=\boldsymbol{f}(s) \tag*{(44)} ∫ 0 2 π w ( s , θ ) J ϵ ( s , θ ) d θ = f ( s ) (44)
As in the Laplace setting [59 ] , we consider w ( s , θ ) = w ( x ( s , θ ) ) \boldsymbol{w}(s,\theta)=\boldsymbol{w}(x(s,\theta)) w ( s , θ ) = w ( x ( s , θ )) as a function of the parameters s s s and θ \theta θ along the surface Γ ϵ \Gamma_{\epsilon} Γ ϵ .
Accordingly, to better compare across different filament centerline geometries, we use the following characterization of the ∣ ⋅ ∣ C ˙ 0 , α |\cdot|_{\dot{C}^{0,\alpha}} ∣ ⋅ ∣ C ˙ 0 , α seminorm along Γ ϵ \Gamma_{\epsilon} Γ ϵ :
∣ h ∣ C ˙ 0 , α ( Γ ϵ ) = sup ( s − s ′ ) 2 + ( θ − θ ′ ) 2 ≠ 0 ∣ h ( s , θ ) − h ( s ′ , θ ′ ) ∣ ( s − s ′ ) 2 + ϵ 2 ( θ − θ ′ ) 2 α . (45) |h|_{\dot{C}^{0,\alpha}(\Gamma_{\epsilon})}
=
\sup_{\substack{(s-s')^{2}+(\theta-\theta')^{2}\ne0}}
\frac{|h(s,\theta)-h(s',\theta')|}
{\sqrt{(s-s')^{2}+\epsilon^{2}(\theta-\theta')^{2}}^{\alpha}}. \tag*{(45)} ∣ h ∣ C ˙ 0 , α ( Γ ϵ ) = ( s − s ′ ) 2 + ( θ − θ ′ ) 2 = 0 sup ( s − s ′ ) 2 + ϵ 2 ( θ − θ ′ ) 2 α ∣ h ( s , θ ) − h ( s ′ , θ ′ ) ∣ . (45)
The seminorm (45) may be seen to be an equivalent characterization of C ˙ 0 , α ( Γ ϵ ) \dot{C}^{0,\alpha}(\Gamma_{\epsilon}) C ˙ 0 , α ( Γ ϵ ) to (15) (see [59 Lemma 3.1] ). In addition, it will be useful to define the following spaces measuring additional Hölder regularity in s s s only:
∥ h ∥ C ˙ s 0 , α ( Γ ϵ ) : = ∥ h ∥ L ∞ ( Γ ϵ ) + sup θ ∈ [ 0 , 2 π ) ( sup s ≠ s ′ ∣ h ( s , θ ) − h ( s ′ , θ ) ∣ ∣ s − s ′ ∣ α ) \|h\|_{\dot{C}^{0,\alpha}_{s}(\Gamma_{\epsilon})}
:=\|h\|_{L^{\infty}(\Gamma_{\epsilon})}
+\sup_{\theta\in[0,2\pi)}
\left(\sup_{s\ne s'}\frac{|h(s,\theta)-h(s',\theta)|}{|s-s'|^{\alpha}}\right) ∥ h ∥ C ˙ s 0 , α ( Γ ϵ ) := ∥ h ∥ L ∞ ( Γ ϵ ) + θ ∈ [ 0 , 2 π ) sup ( s = s ′ sup ∣ s − s ′ ∣ α ∣ h ( s , θ ) − h ( s ′ , θ ) ∣ )
∥ h ∥ C ˙ s m , α ( Γ ϵ ) : = ∑ ℓ = 0 m ∥ ∂ s ℓ h ∥ L ∞ ( Γ ϵ ) + sup θ ∈ [ 0 , 2 π ) ( sup s ≠ s ′ ∣ ∂ s m h ( s , θ ) − ∂ s m h ( s ′ , θ ) ∣ ∣ s − s ′ ∣ α ) . (46) \|h\|_{\dot{C}^{m,\alpha}_{s}(\Gamma_{\epsilon})}
:=\sum_{\ell=0}^{m}\|\partial_{s}^{\ell}h\|_{L^{\infty}(\Gamma_{\epsilon})}
+\sup_{\theta\in[0,2\pi)}
\left(\sup_{s\ne s'}\frac{|\partial_{s}^{m}h(s,\theta)-\partial_{s}^{m}h(s',\theta)|}{|s-s'|^{\alpha}}\right). \tag*{(46)} ∥ h ∥ C ˙ s m , α ( Γ ϵ ) := ℓ = 0 ∑ m ∥ ∂ s ℓ h ∥ L ∞ ( Γ ϵ ) + θ ∈ [ 0 , 2 π ) sup ( s = s ′ sup ∣ s − s ′ ∣ α ∣ ∂ s m h ( s , θ ) − ∂ s m h ( s ′ , θ ) ∣ ) . (46)
Note that if h ∈ C m , α ( Γ ϵ ) h\in C^{m,\alpha}(\Gamma_{\epsilon}) h ∈ C m , α ( Γ ϵ ) is a function of s s s only, then both (45) and (46) reduce to
∣ h ( s ) ∣ C ˙ 0 , α ( Γ ϵ ) = ∣ h ( s ) ∣ C ˙ s 0 , α ( Γ ϵ ) = ∣ h ( s ) ∣ C ˙ 0 , α ( T ) = sup s ≠ s ′ ∣ h ( s ) − h ( s ′ ) ∣ ∣ s − s ′ ∣ α . |h(s)|_{\dot{C}^{0,\alpha}(\Gamma_{\epsilon})}
=
|h(s)|_{\dot{C}^{0,\alpha}_{s}(\Gamma_{\epsilon})}
=
|h(s)|_{\dot{C}^{0,\alpha}(\mathbb{T})}
=
\sup_{s\ne s'}\frac{|h(s)-h(s')|}{|s-s'|^{\alpha}}. ∣ h ( s ) ∣ C ˙ 0 , α ( Γ ϵ ) = ∣ h ( s ) ∣ C ˙ s 0 , α ( Γ ϵ ) = ∣ h ( s ) ∣ C ˙ 0 , α ( T ) = s = s ′ sup ∣ s − s ′ ∣ α ∣ h ( s ) − h ( s ′ ) ∣ .
Single and double layer operators about C ϵ C_{\epsilon} C ϵ
In the Laplace setting [59 ] , we were able to calculate explicit Fourier multiplier expressions for the corresponding single and double layer potentials about the straight cylinder C ϵ C_{\epsilon} C ϵ . These expressions were then applied as maps between θ \theta θ -independent scalar-valued functions on Γ ϵ \Gamma_{\epsilon} Γ ϵ .
In the Stokes setting, we again may calculate explicit Fourier multiplier expressions for the single and double layer operators about the straight filament C ϵ C_{\epsilon} C ϵ , which we will denote by S ‾ \overline{S} S and D ˉ \bar{D} D ˉ , respectively. However, the situation is far more complicated. In particular, the multipliers are now matrix-valued expressions that must be applied between vector-valued functions living on cylindrical basis vectors e r \mathbf{e}_r e r , e θ \mathbf{e}_\theta e θ , e z \mathbf{e}_z e z .
Furthermore, the θ \theta θ -independent functions on C ϵ C_\epsilon C ϵ are represented not just by zero Fourier modes in θ \theta θ , but also by one modes: given a θ \theta θ -independent vector-valued function f ( s ) \mathbf{f}(s) f ( s ) along the straight cylinder C ϵ C_\epsilon C ϵ , for s ∈ T s \in\mathbb{T} s ∈ T we may write
f ( s ) = f z ( s ) e z + f x ( s ) e x + f y ( s ) e y = f z ( s ) e z + f x ( s ) ( cos θ e r − sin θ e θ ) + f y ( s ) ( sin θ e r + cos θ e θ ) . \mathbf{f}(s) = f^z(s)\mathbf{e}_z + f^x(s)\mathbf{e}_x + f^y(s)\mathbf{e}_y
= f^z(s)\mathbf{e}_z + f^x(s)(\cos\theta\mathbf{e}_r-\sin\theta\mathbf{e}_\theta) + f^y(s)(\sin\theta\mathbf{e}_r+\cos\theta\mathbf{e}_\theta). f ( s ) = f z ( s ) e z + f x ( s ) e x + f y ( s ) e y = f z ( s ) e z + f x ( s ) ( cos θ e r − sin θ e θ ) + f y ( s ) ( sin θ e r + cos θ e θ ) .
Along C ϵ C_\epsilon C ϵ , we may treat the tangential (e z \mathbf{e}_z e z ) and normal (e x \mathbf{e}_x e x and e y \mathbf{e}_y e y ) directions along the filament separately. Expanding each of f z ( s ) f^z(s) f z ( s ) , f x ( s ) f^x(s) f x ( s ) , and f y ( s ) f^y(s) f y ( s ) as a Fourier series in s s s , we show that the double and single layer operators D ˉ \bar{D} D ˉ and S ˉ \bar{S} S ˉ act in the following way.
Lemma 1.7 (Single and double layer multipliers about C ϵ C_\epsilon C ϵ ). When applied to θ \theta θ -independent data in the tangential direction along the straight cylinder C ϵ C_\epsilon C ϵ , the double layer operator D ˉ \bar{D} D ˉ is given by
D ˉ [ e 2 π i k s e z ] = Q t E ( 2 π ϵ ∣ k ∣ ) e 2 π i k s e z + i Q t F ( 2 π ϵ ∣ k ∣ ) e 2 π i k s e r , \bar{D}\left[e^{2\pi i k s}\mathbf{e}_z\right] = Q_{\mathrm{tE}}(2\pi\epsilon|k|)e^{2\pi i k s}\mathbf{e}_z + iQ_{\mathrm{tF}}(2\pi\epsilon|k|)e^{2\pi i k s}\mathbf{e}_r, D ˉ [ e 2 π ik s e z ] = Q tE ( 2 π ϵ ∣ k ∣ ) e 2 π ik s e z + i Q tF ( 2 π ϵ ∣ k ∣ ) e 2 π ik s e r , Q t E ( z ) = z 2 ( I 0 K 0 − I 1 K 1 ) − z 2 ( I 0 K 1 − I 1 K 0 ) , Q_{\mathrm{tE}}(z) = z^2(I_0K_0-I_1K_1)-\frac{z}{2}(I_0K_1-I_1K_0), Q tE ( z ) = z 2 ( I 0 K 0 − I 1 K 1 ) − 2 z ( I 0 K 1 − I 1 K 0 ) , Q t F ( z ) = − z 2 ( I 1 K 0 − I 0 K 1 ) − z I 1 K 1 . (47) Q_{\mathrm{tF}}(z) = -z^2(I_1K_0-I_0K_1)-zI_1K_1. \tag*{(47)} Q tF ( z ) = − z 2 ( I 1 K 0 − I 0 K 1 ) − z I 1 K 1 . (47) Here and throughout, each I j = I j ( z ) I_j=I_j(z) I j = I j ( z ) and K j = K j ( z ) K_j=K_j(z) K j = K j ( z ) are modified Bessel functions of the first and second kind, respectively. Given the form (47) of the double layer operator, let
F z = ( f z z e z + f z r e r ) e 2 π i k s . (48) F_z=(f_z^z\mathbf{e}_z+f_z^r\mathbf{e}_r)e^{2\pi i k s}. \tag*{(48)} F z = ( f z z e z + f z r e r ) e 2 π ik s . (48) When applied to data of the form (48) the single layer operator S ˉ \bar{S} S ˉ is invertible and satisfies
S ˉ − 1 [ F z ] = ϵ − 1 M S , t − 1 ( k ) F z , \bar{S}^{-1}[F_z]=\epsilon^{-1}M_{\mathrm{S,t}}^{-1}(k)F_z, S ˉ − 1 [ F z ] = ϵ − 1 M S , t − 1 ( k ) F z , M S , t − 1 ( k ) = m ϵ , t A ( k ) e z ⊗ e z + i m ϵ , t B ( k ) ( e r ⊗ e z − e z ⊗ e r ) + m ϵ , t C ( k ) e r ⊗ e r . (49) M_{\mathrm{S,t}}^{-1}(k)=m_{\epsilon,\mathrm{tA}}(k)\mathbf{e}_z\otimes\mathbf{e}_z+i\,m_{\epsilon,\mathrm{tB}}(k)(\mathbf{e}_r\otimes\mathbf{e}_z-\mathbf{e}_z\otimes\mathbf{e}_r)+m_{\epsilon,\mathrm{tC}}(k)\mathbf{e}_r\otimes\mathbf{e}_r. \tag*{(49)} M S , t − 1 ( k ) = m ϵ , tA ( k ) e z ⊗ e z + i m ϵ , tB ( k ) ( e r ⊗ e z − e z ⊗ e r ) + m ϵ , tC ( k ) e r ⊗ e r . (49) The components of the matrix-valued multiplier M S , t − 1 ( k ) M_{\mathrm{S,t}}^{-1}(k) M S , t − 1 ( k ) are given by
m ϵ , t j ( k ) = Q t j ( 2 π ϵ ∣ k ∣ ) Q t D ( 2 π ϵ ∣ k ∣ ) , j = A , B , C , m_{\epsilon,\mathrm{t}j}(k)=\frac{Q_{\mathrm{t}j}(2\pi\epsilon|k|)}{Q_{\mathrm{tD}}(2\pi\epsilon|k|)},\qquad j=\mathrm{A},\mathrm{B},\mathrm{C}, m ϵ , t j ( k ) = Q tD ( 2 π ϵ ∣ k ∣ ) Q t j ( 2 π ϵ ∣ k ∣ ) , j = A , B , C , Q t A ( z ) = 1 I 0 K 0 ( 1 + z 2 ( K 0 K 1 − I 0 I 1 ) ) , Q t C ( z ) = 1 I 1 K 1 ( 1 + z 2 ( I 1 I 0 − K 1 K 0 ) ) , Q_{\mathrm{tA}}(z)=\frac{1}{I_0K_0}\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)\right),\qquad Q_{\mathrm{tC}}(z)=\frac{1}{I_1K_1}\left(1+\frac{z}{2}\left(\frac{I_1}{I_0}-\frac{K_1}{K_0}\right)\right), Q tA ( z ) = I 0 K 0 1 ( 1 + 2 z ( K 1 K 0 − I 1 I 0 ) ) , Q tC ( z ) = I 1 K 1 1 ( 1 + 2 z ( I 0 I 1 − K 0 K 1 ) ) , Q t B ( z ) = z 2 ( 1 I 1 K 1 − 1 I 0 K 0 ) , Q t D ( z ) = ( 1 + z 2 ( I 1 I 0 − I 0 I 1 ) ) ( 1 + z 2 ( K 0 K 1 − K 1 K 0 ) ) . (50) Q_{\mathrm{tB}}(z)=\frac{z}{2}\left(\frac{1}{I_1K_1}-\frac{1}{I_0K_0}\right),\qquad Q_{\mathrm{tD}}(z)=\left(1+\frac{z}{2}\left(\frac{I_1}{I_0}-\frac{I_0}{I_1}\right)\right)\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)\right). \tag*{(50)} Q tB ( z ) = 2 z ( I 1 K 1 1 − I 0 K 0 1 ) , Q tD ( z ) = ( 1 + 2 z ( I 0 I 1 − I 1 I 0 ) ) ( 1 + 2 z ( K 1 K 0 − K 0 K 1 ) ) . (50) When applied to θ \theta θ -independent data in the normal direction to the straight cylinder C ϵ C_\epsilon C ϵ , the double layer operator D ˉ \bar{D} D ˉ is given by
D ˉ [ e 2 π i k s e x ] = D ˉ [ e 2 π i k s ( cos θ e r − sin θ e θ ) ] \bar{D}\left[e^{2\pi i k s}\mathbf{e}_x\right]=\bar{D}\left[e^{2\pi i k s}(\cos\theta\mathbf{e}_r-\sin\theta\mathbf{e}_\theta)\right] D ˉ [ e 2 π ik s e x ] = D ˉ [ e 2 π ik s ( cos θ e r − sin θ e θ ) ] = ( Q n N ( 2 π ϵ ∣ k ∣ ) cos θ e r − Q n O ( 2 π ϵ ∣ k ∣ ) sin θ e θ + i Q n P ( 2 π ϵ ∣ k ∣ ) cos θ e z ) e 2 π i k s , =\left(Q_{\mathrm{nN}}(2\pi\epsilon|k|)\cos\theta\mathbf{e}_r-Q_{\mathrm{nO}}(2\pi\epsilon|k|)\sin\theta\mathbf{e}_\theta+iQ_{\mathrm{nP}}(2\pi\epsilon|k|)\cos\theta\mathbf{e}_z\right)e^{2\pi i k s}, = ( Q nN ( 2 π ϵ ∣ k ∣ ) cos θ e r − Q nO ( 2 π ϵ ∣ k ∣ ) sin θ e θ + i Q nP ( 2 π ϵ ∣ k ∣ ) cos θ e z ) e 2 π ik s , D ˉ [ e 2 π i k s e y ] = D ˉ [ e 2 π i k s ( sin θ e r + cos θ e θ ) ] \bar{D}\left[e^{2\pi i k s}\mathbf{e}_y\right]=\bar{D}\left[e^{2\pi i k s}(\sin\theta\mathbf{e}_r+\cos\theta\mathbf{e}_\theta)\right] D ˉ [ e 2 π ik s e y ] = D ˉ [ e 2 π ik s ( sin θ e r + cos θ e θ ) ] = ( Q n N ( 2 π ϵ ∣ k ∣ ) sin θ e r + Q n O ( 2 π ϵ ∣ k ∣ ) cos θ e θ + i Q n P ( 2 π ϵ ∣ k ∣ ) sin θ e z ) e 2 π i k s ; (51) =\left(Q_{\mathrm{nN}}(2\pi\epsilon|k|)\sin\theta\mathbf{e}_r+Q_{\mathrm{nO}}(2\pi\epsilon|k|)\cos\theta\mathbf{e}_\theta+iQ_{\mathrm{nP}}(2\pi\epsilon|k|)\sin\theta\mathbf{e}_z\right)e^{2\pi i k s}; \tag*{(51)} = ( Q nN ( 2 π ϵ ∣ k ∣ ) sin θ e r + Q nO ( 2 π ϵ ∣ k ∣ ) cos θ e θ + i Q nP ( 2 π ϵ ∣ k ∣ ) sin θ e z ) e 2 π ik s ; (51) Q n N ( z ) = z 2 ( I 1 K 1 − I 0 K 0 ) + 3 2 z ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1 , Q_{\mathrm{nN}}(z)=z^2(I_1K_1-I_0K_0)+\frac{3}{2}z(I_1K_0-I_0K_1)+2I_1K_1, Q nN ( z ) = z 2 ( I 1 K 1 − I 0 K 0 ) + 2 3 z ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1 , Q n O ( z ) = − z 2 ( I 1 K 0 − I 0 K 1 ) − 2 I 1 K 1 , Q n P ( z ) = z 2 ( I 0 K 1 − I 1 K 0 ) − z I 1 K 1 . Q_{\mathrm{nO}}(z)=-\frac{z}{2}(I_1K_0-I_0K_1)-2I_1K_1,\qquad Q_{\mathrm{nP}}(z)=z^2(I_0K_1-I_1K_0)-zI_1K_1. Q nO ( z ) = − 2 z ( I 1 K 0 − I 0 K 1 ) − 2 I 1 K 1 , Q nP ( z ) = z 2 ( I 0 K 1 − I 1 K 0 ) − z I 1 K 1 .
Given the form of the expression (51), let
F x ( s , θ ) = ( f η x cos θ e r + f θ x sin θ e θ + f z x cos θ e z ) e 2 π i k s F y ( s , θ ) = ( f η y sin θ e r − f θ y cos θ e θ + f z y sin θ e z ) e 2 π i k s . (52) \begin{aligned}
F_x(s,\theta)&=\left(f_\eta^x\cos\theta e_r+f_\theta^x\sin\theta e_\theta+f_z^x\cos\theta e_z\right)e^{2\pi iks}\\
F_y(s,\theta)&=\left(f_\eta^y\sin\theta e_r-f_\theta^y\cos\theta e_\theta+f_z^y\sin\theta e_z\right)e^{2\pi iks}. \tag*{(52)}
\end{aligned} F x ( s , θ ) F y ( s , θ ) = ( f η x cos θ e r + f θ x sin θ e θ + f z x cos θ e z ) e 2 π ik s = ( f η y sin θ e r − f θ y cos θ e θ + f z y sin θ e z ) e 2 π ik s . (52)
When applied to data of the form (52), the single layer operator S ˉ \bar{S} S ˉ is again invertible and satisfies
S ˉ − 1 [ F x ] = ϵ − 1 M S , n − 1 ( k ) F x , S ˉ − 1 [ F y ] = ϵ − 1 M S , n − 1 ( k ) F y , (53) \bar{S}^{-1}[F_x]=\epsilon^{-1}M_{S,n}^{-1}(k)F_x,\qquad\bar{S}^{-1}[F_y]=\epsilon^{-1}M_{S,n}^{-1}(k)F_y, \tag*{(53)} S ˉ − 1 [ F x ] = ϵ − 1 M S , n − 1 ( k ) F x , S ˉ − 1 [ F y ] = ϵ − 1 M S , n − 1 ( k ) F y , (53)
where M S , n − 1 M_{S,n}^{-1} M S , n − 1 is a matrix-valued multiplier of the form
M S , n − 1 ( k ) = m ϵ , n A ( k ) e r ⊗ e r + m ϵ , n B ( k ) ( e r ⊗ e θ + e θ ⊗ e r ) + m ϵ , n D ( k ) e θ ⊗ e θ + i m ϵ , n C ( k ) ( e r ⊗ e z − e z ⊗ e r ) + i m ϵ , n E ( k ) ( e z ⊗ e θ − e θ ⊗ e z ) + m ϵ , n F ( k ) e z ⊗ e z . (54) \begin{aligned}
M_{S,n}^{-1}(k)={}&m_{\epsilon,nA}(k)e_r\otimes e_r+m_{\epsilon,nB}(k)(e_r\otimes e_\theta+e_\theta\otimes e_r)+m_{\epsilon,nD}(k)e_\theta\otimes e_\theta\\
&+im_{\epsilon,nC}(k)(e_r\otimes e_z-e_z\otimes e_r)+im_{\epsilon,nE}(k)(e_z\otimes e_\theta-e_\theta\otimes e_z)+m_{\epsilon,nF}(k)e_z\otimes e_z. \tag*{(54)}
\end{aligned} M S , n − 1 ( k ) = m ϵ , n A ( k ) e r ⊗ e r + m ϵ , n B ( k ) ( e r ⊗ e θ + e θ ⊗ e r ) + m ϵ , n D ( k ) e θ ⊗ e θ + i m ϵ , n C ( k ) ( e r ⊗ e z − e z ⊗ e r ) + i m ϵ , n E ( k ) ( e z ⊗ e θ − e θ ⊗ e z ) + m ϵ , n F ( k ) e z ⊗ e z . (54)
The components of M S , n − 1 ( k ) M_{S,n}^{-1}(k) M S , n − 1 ( k ) are given explicitly by
m ϵ , n j ( k ) = Q n j ( 2 π ϵ ∣ k ∣ ) Q n G ( 2 π ϵ ∣ k ∣ ) , j = A , B , C , D , E , F , m_{\epsilon,nj}(k)=\frac{Q_{nj}(2\pi\epsilon|k|)}{Q_{nG}(2\pi\epsilon|k|)},\qquad j=\mathrm{A,B,C,D,E,F}, m ϵ , nj ( k ) = Q n G ( 2 π ϵ ∣ k ∣ ) Q nj ( 2 π ϵ ∣ k ∣ ) , j = A , B , C , D , E , F ,
Q n A ( z ) = 1 I 1 K 1 ( 4 + 1 z 2 I 1 I 0 K 1 K 0 + 2 z ( K 1 K 0 − I 1 I 0 ) − 2 z ( I 0 I 1 − K 0 K 1 ) − 2 ( I 0 I 1 K 1 K 0 + I 1 I 0 K 0 K 1 ) ) , Q_{nA}(z)=\frac{1}{I_1K_1}\left(4+\frac{1}{z^2}\frac{I_1}{I_0}\frac{K_1}{K_0}+\frac{2}{z}\left(\frac{K_1}{K_0}-\frac{I_1}{I_0}\right)-2z\left(\frac{I_0}{I_1}-\frac{K_0}{K_1}\right)-2\left(\frac{I_0}{I_1}\frac{K_1}{K_0}+\frac{I_1}{I_0}\frac{K_0}{K_1}\right)\right), Q n A ( z ) = I 1 K 1 1 ( 4 + z 2 1 I 0 I 1 K 0 K 1 + z 2 ( K 0 K 1 − I 0 I 1 ) − 2 z ( I 1 I 0 − K 1 K 0 ) − 2 ( I 1 I 0 K 0 K 1 + I 0 I 1 K 1 K 0 ) ) ,
Q n B ( z ) = 1 I 1 K 1 ( 2 K 1 K 0 ( I 0 I 1 − I 1 I 0 ) + 2 ( I 1 I 0 K 0 K 1 − 1 ) + 2 z ( I 1 I 0 − K 1 K 0 ) − 4 z 2 I 1 K 1 I 0 K 0 ) , Q_{nB}(z)=\frac{1}{I_1K_1}\left(2\frac{K_1}{K_0}\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)+2\left(\frac{I_1}{I_0}\frac{K_0}{K_1}-1\right)+\frac{2}{z}\left(\frac{I_1}{I_0}-\frac{K_1}{K_0}\right)-\frac{4}{z^2}\frac{I_1K_1}{I_0K_0}\right), Q n B ( z ) = I 1 K 1 1 ( 2 K 0 K 1 ( I 1 I 0 − I 0 I 1 ) + 2 ( I 0 I 1 K 1 K 0 − 1 ) + z 2 ( I 0 I 1 − K 0 K 1 ) − z 2 4 I 0 K 0 I 1 K 1 ) ,
Q n C ( z ) = 2 I 1 K 1 ( 2 z I 1 I 0 K 1 K 0 + 2 ( I 0 I 1 − K 0 K 1 ) + ( I 1 I 0 − K 1 K 0 ) + z I 0 I 1 K 0 K 1 − z − 4 z ) , Q_{nC}(z)=\frac{2}{I_1K_1}\left(\frac{2}{z}\frac{I_1}{I_0}\frac{K_1}{K_0}+2\left(\frac{I_0}{I_1}-\frac{K_0}{K_1}\right)+\left(\frac{I_1}{I_0}-\frac{K_1}{K_0}\right)+z\frac{I_0}{I_1}\frac{K_0}{K_1}-z-\frac{4}{z}\right), Q n C ( z ) = I 1 K 1 2 ( z 2 I 0 I 1 K 0 K 1 + 2 ( I 1 I 0 − K 1 K 0 ) + ( I 0 I 1 − K 0 K 1 ) + z I 1 I 0 K 1 K 0 − z − z 4 ) ,
\begin{aligned}
Q_{nD}(z)={}&\frac{1}{I_1K_1}\left(\frac{4}{z^2}\frac{I_1}{I_0}\frac{K_1}{K_0}+\frac{2}{z}\left(\frac{K_1}{K_0}-\frac{I_1}{I_0}\right)+2\left(2\frac{I_1}{I_0}\frac{K_1}{K_0}-\frac{I_1}{I_0}\frac{K_0}{K_1}-\frac{I_0}{I_1}\frac{K_1}{K_0}\right)\\
&+z^2\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)\right). \tag*{(55)}
\end{aligned}
Q n E ( z ) = 1 I 1 K 1 ( 4 z I 1 I 0 K 1 K 0 − 4 z + 2 ( I 0 I 1 − K 0 K 1 ) + z ( I 0 I 1 − I 1 I 0 ) ( K 0 K 1 − K 1 K 0 ) ) , Q_{nE}(z)=\frac{1}{I_1K_1}\left(\frac{4}{z}\frac{I_1}{I_0}\frac{K_1}{K_0}-\frac{4}{z}+2\left(\frac{I_0}{I_1}-\frac{K_0}{K_1}\right)+z\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)\right), Q n E ( z ) = I 1 K 1 1 ( z 4 I 0 I 1 K 0 K 1 − z 4 + 2 ( I 1 I 0 − K 1 K 0 ) + z ( I 1 I 0 − I 0 I 1 ) ( K 1 K 0 − K 0 K 1 ) ) ,
Q n F ( z ) = 1 I 1 K 1 ( 12 z 2 + 6 z ( K 0 K 1 − I 0 I 1 + K 1 K 0 − I 1 I 0 ) − 3 ( I 0 I 1 ( K 0 K 1 + K 1 K 0 ) + I 1 K 0 I 0 K 1 ) + ( I 1 I 0 K 1 K 0 + 8 ) + 2 z ( K 0 K 1 − I 0 I 1 ) ) , \begin{aligned}
Q_{nF}(z)={}&\frac{1}{I_1K_1}\left(\frac{12}{z^2}+\frac{6}{z}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}+\frac{K_1}{K_0}-\frac{I_1}{I_0}\right)-3\left(\frac{I_0}{I_1}\left(\frac{K_0}{K_1}+\frac{K_1}{K_0}\right)+\frac{I_1K_0}{I_0K_1}\right)\right.\\
&\left.+\left(\frac{I_1}{I_0}\frac{K_1}{K_0}+8\right)+2z\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)\right),
\end{aligned} Q n F ( z ) = I 1 K 1 1 ( z 2 12 + z 6 ( K 1 K 0 − I 1 I 0 + K 0 K 1 − I 0 I 1 ) − 3 ( I 1 I 0 ( K 1 K 0 + K 0 K 1 ) + I 0 K 1 I 1 K 0 ) + ( I 0 I 1 K 0 K 1 + 8 ) + 2 z ( K 1 K 0 − I 1 I 0 ) ) ,
Q n G ( z ) = ( 2 z + ( 1 − z I 0 I 1 ) ( I 0 I 1 − I 1 I 0 ) ) ( 2 z − ( 1 + z K 0 K 1 ) ( K 0 K 1 − K 1 K 0 ) ) . Q_{nG}(z)=\left(\frac{2}{z}+\left(1-z\frac{I_0}{I_1}\right)\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)\right)\left(\frac{2}{z}-\left(1+z\frac{K_0}{K_1}\right)\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)\right). Q n G ( z ) = ( z 2 + ( 1 − z I 1 I 0 ) ( I 1 I 0 − I 0 I 1 ) ) ( z 2 − ( 1 + z K 1 K 0 ) ( K 1 K 0 − K 0 K 1 ) ) .
The calculation of the explicit symbols in Lemma 1.7 appears in Appendix B. For data of the form F z F_z F z (48) or F x , F y F_x,F_y F x , F y (52), instead of working with the full inverse single layer operator S ˉ − 1 \bar{S}^{-1} S ˉ − 1 , it will be convenient to define the following angle-averaged inverse single layer along C ϵ C_\epsilon C ϵ :
A ˉ ϵ [ F j ] : = ∫ 0 2 π S ˉ − 1 [ F j ] ϵ d θ . (56) \bar{A}_\epsilon[F_j]:=\int_0^{2\pi}\bar{S}^{-1}[F_j]\epsilon\,d\theta. \tag*{(56)} A ˉ ϵ [ F j ] := ∫ 0 2 π S ˉ − 1 [ F j ] ϵ d θ . (56)
Using that
∫ 0 2 π e z d θ = 2 π e z , ∫ 0 2 π e r d θ = ∫ 0 2 π e θ d θ = ∫ 0 2 π cos θ e z d θ = 0 , \int_0^{2\pi}e_z\,d\theta=2\pi e_z,\qquad\int_0^{2\pi}e_r\,d\theta=\int_0^{2\pi}e_\theta\,d\theta=\int_0^{2\pi}\cos\theta e_z\,d\theta=0, ∫ 0 2 π e z d θ = 2 π e z , ∫ 0 2 π e r d θ = ∫ 0 2 π e θ d θ = ∫ 0 2 π cos θ e z d θ = 0 ,
∫ 0 2 π cos θ e r d θ = − ∫ 0 2 π sin θ e θ d θ = π e x , ∫ 0 2 π sin θ e r d θ = ∫ 0 2 π cos θ e θ d θ = π e y , \int_0^{2\pi}\cos\theta e_r\,d\theta=-\int_0^{2\pi}\sin\theta e_\theta\,d\theta=\pi e_x,\qquad\int_0^{2\pi}\sin\theta e_r\,d\theta=\int_0^{2\pi}\cos\theta e_\theta\,d\theta=\pi e_y, ∫ 0 2 π cos θ e r d θ = − ∫ 0 2 π sin θ e θ d θ = π e x , ∫ 0 2 π sin θ e r d θ = ∫ 0 2 π cos θ e θ d θ = π e y ,
we immediately obtain the following corollary to Lemma 1.7.
Corollary 1.8 (Angle-averaged single layer). *When applied to functions F μ F_\mu F μ of the form (48) or (52) , the angle-averaged inverse single layer operator A ‾ ϵ \overline{A}_\epsilon A ϵ given by (56) satisfies
A ‾ ϵ [ F μ ] = π ( m ϵ , r μ ( k ) f r μ + m ϵ , θ μ ( k ) f θ μ + m ϵ , z μ ( k ) f z μ ) e 2 π i k s e μ , μ = x , y , z , (57) \overline{A}_\epsilon[F_\mu] = \pi\left(m^\mu_{\epsilon,r}(k)f^\mu_r + m^\mu_{\epsilon,\theta}(k)f^\mu_\theta+ m^\mu_{\epsilon,z}(k)f^\mu_z\right)e^{2\pi i k s}e_\mu,\qquad\mu=x,y,z, \tag*{(57)} A ϵ [ F μ ] = π ( m ϵ , r μ ( k ) f r μ + m ϵ , θ μ ( k ) f θ μ + m ϵ , z μ ( k ) f z μ ) e 2 π ik s e μ , μ = x , y , z , (57) where the tangential multipliers m ϵ , ℓ z m^z_{\epsilon,\ell} m ϵ , ℓ z are given by
m ϵ , z z ( k ) = 2 m ϵ , t A ( k ) , m ϵ , θ z ( k ) = 0 , m ϵ , r z ( k ) = − 2 i m ϵ , t B ( k ) ; (58) m^z_{\epsilon,z}(k)=2m_{\epsilon,tA}(k),\qquad m^z_{\epsilon,\theta}(k)=0,\qquad m^z_{\epsilon,r}(k)=-2i\,m_{\epsilon,tB}(k); \tag*{(58)} m ϵ , z z ( k ) = 2 m ϵ , t A ( k ) , m ϵ , θ z ( k ) = 0 , m ϵ , r z ( k ) = − 2 i m ϵ , tB ( k ) ; (58) and the normal direction multipliers m ϵ , ℓ x = m ϵ , ℓ y m^x_{\epsilon,\ell}=m^y_{\epsilon,\ell} m ϵ , ℓ x = m ϵ , ℓ y are given by
m ϵ , r x ( k ) = m ϵ , r y ( k ) = m ϵ , n A ( k ) − m ϵ , n B ( k ) , m ϵ , θ x ( k ) = m ϵ , θ y ( k ) = m ϵ , n B ( k ) − m ϵ , n D ( k ) , m ϵ , z x ( k ) = m ϵ , z y ( k ) = i ( m ϵ , n C ( k ) + m ϵ , n E ( k ) ) . (59) \begin{aligned}
m^x_{\epsilon,r}(k)&=m^y_{\epsilon,r}(k)=m_{\epsilon,nA}(k)-m_{\epsilon,nB}(k),\\
m^x_{\epsilon,\theta}(k)&=m^y_{\epsilon,\theta}(k)=m_{\epsilon,nB}(k)-m_{\epsilon,nD}(k),\\
m^x_{\epsilon,z}(k)&=m^y_{\epsilon,z}(k)=i\left(m_{\epsilon,nC}(k)+m_{\epsilon,nE}(k)\right). \tag*{(59)}
\end{aligned} m ϵ , r x ( k ) m ϵ , θ x ( k ) m ϵ , z x ( k ) = m ϵ , r y ( k ) = m ϵ , n A ( k ) − m ϵ , n B ( k ) , = m ϵ , θ y ( k ) = m ϵ , n B ( k ) − m ϵ , n D ( k ) , = m ϵ , z y ( k ) = i ( m ϵ , n C ( k ) + m ϵ , n E ( k ) ) . (59)
It is ultimately the angle-averaged operator A ‾ ϵ \overline{A}_\epsilon A ϵ for which we will need detailed information. Before we turn to the mapping properties of A ‾ ϵ \overline{A}_\epsilon A ϵ , we may perform a sanity check to demonstrate that we indeed have the correct expressions for the single and double layer multipliers on C ϵ C_\epsilon C ϵ . In particular, we can verify that the double and single layer expressions of Lemma 1.7 combine to yield the multipliers m ϵ , t − 1 m^{-1}_{\epsilon,t} m ϵ , t − 1 and m ϵ , n − 1 m^{-1}_{\epsilon,n} m ϵ , n − 1 given by (23) and (24) . For the tangential direction, we may combine (47) and (58) to obtain
L ‾ ϵ − 1 [ e 2 π i k s e z ] = A ‾ ϵ [ ( 1 2 I − D ‾ ) [ e 2 π i k s e z ] ] = 2 π ( m ϵ , t A ( k ) Q t E ( 2 π ϵ ∣ k ∣ ) + m ϵ , t B ( k ) Q t F ( 2 π ϵ ∣ k ∣ ) ) e 2 π i k s e z . (60) \begin{aligned}
\overline{\mathcal{L}}_\epsilon^{-1}\left[e^{2\pi i k s}e_z\right]=\overline{A}_\epsilon\left[\left(\tfrac12 I-\overline{D}\right)\left[e^{2\pi i k s}e_z\right]\right] \\
&=2\pi\left(m_{\epsilon,tA}(k)Q_{tE}(2\pi\epsilon|k|)+m_{\epsilon,tB}(k)Q_{tF}(2\pi\epsilon|k|)\right)e^{2\pi i k s}e_z. \tag*{(60)}
\end{aligned} L ϵ − 1 [ e 2 π ik s e z ] = A ϵ [ ( 2 1 I − D ) [ e 2 π ik s e z ] ] = 2 π ( m ϵ , t A ( k ) Q tE ( 2 π ϵ ∣ k ∣ ) + m ϵ , tB ( k ) Q tF ( 2 π ϵ ∣ k ∣ ) ) e 2 π ik s e z . (60)
For the normal direction, we may combine (51) and (59) to obtain
L ‾ ϵ − 1 [ e 2 π i k s e x ] = A ‾ ϵ [ ( 1 2 I − D ‾ ) [ e 2 π i k s e x ] ] \overline{\mathcal{L}}_\epsilon^{-1}\left[e^{2\pi i k s}e_x\right]=\overline{A}_\epsilon\left[\left(\tfrac12 I-\overline{D}\right)\left[e^{2\pi i k s}e_x\right]\right] L ϵ − 1 [ e 2 π ik s e x ] = A ϵ [ ( 2 1 I − D ) [ e 2 π ik s e x ] ]
= π ( ( m ϵ , n A ( k ) − m ϵ , n B ( k ) ) Q n N ( 2 π ϵ ∣ k ∣ ) − ( m ϵ , n B ( k ) − m ϵ , n D ( k ) ) Q n O ( 2 π ϵ ∣ k ∣ ) − ( m ϵ , n C ( k ) + m ϵ , n E ( k ) ) Q n P ( 2 π ϵ ∣ k ∣ ) ) e 2 π i k s e x . (61) \begin{aligned}
={}&\pi\Bigl(\left(m_{\epsilon,nA}(k)-m_{\epsilon,nB}(k)\right)Q_{nN}(2\pi\epsilon|k|)-\left(m_{\epsilon,nB}(k)-m_{\epsilon,nD}(k)\right)Q_{nO}(2\pi\epsilon|k|)\\
&\qquad-\left(m_{\epsilon,nC}(k)+m_{\epsilon,nE}(k)\right)Q_{nP}(2\pi\epsilon|k|)\Bigr)e^{2\pi i k s}e_x. \tag*{(61)}
\end{aligned} = π ( ( m ϵ , n A ( k ) − m ϵ , n B ( k ) ) Q n N ( 2 π ϵ ∣ k ∣ ) − ( m ϵ , n B ( k ) − m ϵ , n D ( k ) ) Q n O ( 2 π ϵ ∣ k ∣ ) − ( m ϵ , n C ( k ) + m ϵ , n E ( k ) ) Q n P ( 2 π ϵ ∣ k ∣ ) ) e 2 π ik s e x . (61)
The e y e_y e y direction is identical. Here each of the m ϵ , j m_{\epsilon,j} m ϵ , j and Q j Q_j Q j are as in Lemma 1.7. Just by looking at the components of the expressions (60) and (61) , it is not clear that these are the same as the multipliers m ϵ , t − 1 ( k ) m^{-1}_{\epsilon,t}(k) m ϵ , t − 1 ( k ) and m ϵ , n − 1 ( k ) m^{-1}_{\epsilon,n}(k) m ϵ , n − 1 ( k ) of Proposition 1.3 . However, by plotting (see Figure 2 ), we can verify that (60) and (61) are in fact the same as (23) and (24) from Proposition 1.3 .
The important takeaway from the explicit expressions in Lemma 1.7 is that we can then directly calculate the mapping properties of the angle-averaged operator A ‾ ϵ \overline{A}_\epsilon A ϵ when applied to functions consisting of zero- and one-modes in θ \theta θ . This is in direct analogy to the Laplace setting [59 ] , but the calculation here is much more involved.
For a general vector-valued function g ( s , θ ) \mathbf{g}(s,\theta) g ( s , θ ) defined along the straight cylinder C ϵ C_\epsilon C ϵ , we write
g ( s , θ ) = ∑ j ∈ { r , θ , z } ( g 0 j ( s ) + ∑ ℓ = 0 ∞ g c , ℓ j ( s ) cos ( ℓ θ ) + g s , ℓ j ( s ) sin ( ℓ θ ) ) e j . \mathbf{g}(s,\theta)=\sum_{j\in\{r,\theta,z\}}\left(g^j_0(s)+\sum_{\ell=0}^{\infty}g^j_{c,\ell}(s)\cos(\ell\theta)+g^j_{s,\ell}(s)\sin(\ell\theta)\right)e_j. g ( s , θ ) = j ∈ { r , θ , z } ∑ ( g 0 j ( s ) + ℓ = 0 ∑ ∞ g c , ℓ j ( s ) cos ( ℓ θ ) + g s , ℓ j ( s ) sin ( ℓ θ ) ) e j .
We define the projection P 01 P_{01} P 01 onto zero- and one-modes in θ \theta θ by
P 01 g = g 0 z e z + g 0 r e r + g c , 1 r cos θ e r + g s , 1 θ sin θ e θ + g c , 1 z cos θ e z + g s , 1 r sin θ e r + g c , 1 θ cos θ e θ + g s , 1 z sin θ e z . (62) P_{01}\mathbf{g}=g^z_0e_z+g^r_0e_r+g^r_{c,1}\cos\theta e_r+g^\theta_{s,1}\sin\theta e_\theta+g^z_{c,1}\cos\theta e_z+g^r_{s,1}\sin\theta e_r+g^\theta_{c,1}\cos\theta e_\theta+g^z_{s,1}\sin\theta e_z. \tag*{(62)} P 01 g = g 0 z e z + g 0 r e r + g c , 1 r cos θ e r + g s , 1 θ sin θ e θ + g c , 1 z cos θ e z + g s , 1 r sin θ e r + g c , 1 θ cos θ e θ + g s , 1 z sin θ e z . (62)
Lemma 1.9. Let A ‾ ϵ \overline{\mathcal{A}}_\epsilon A ϵ be the angle-averaged inverse single layer operator about C ϵ C_\epsilon C ϵ , given by (56). Given g ∈ C s 1 , α ( Γ ϵ ) \boldsymbol{g} \in C_s^{1,\alpha}(\Gamma_\epsilon) g ∈ C s 1 , α ( Γ ϵ ) and the projection operator (62), we may decompose A ‾ ϵ P 01 \overline{\mathcal{A}}_\epsilon\mathbb{P}_{01} A ϵ P 01 as
A ‾ ϵ [ P 01 g ] = A ‾ ϵ ϵ [ g ] + A ‾ ϵ + [ g ] \overline{\mathcal{A}}_\epsilon[\mathbb{P}_{01}\boldsymbol{g}] = \overline{\mathcal{A}}_\epsilon^{\epsilon}[\boldsymbol{g}] + \overline{\mathcal{A}}_\epsilon^{+}[\boldsymbol{g}] A ϵ [ P 01 g ] = A ϵ ϵ [ g ] + A ϵ + [ g ] where
∥ A ‾ ϵ ϵ [ g ] ∥ C 0 , α ( T ) ≤ c ∥ g ∥ C s 0 , α ( Γ ϵ ) + c ϵ ∣ g ∣ C s 1 , α ( Γ ϵ ) , \left\|\overline{\mathcal{A}}_\epsilon^{\epsilon}[\boldsymbol{g}]\right\|_{C^{0,\alpha}(\mathbb{T})} \leq c\|\boldsymbol{g}\|_{C_s^{0,\alpha}(\Gamma_\epsilon)} + c\epsilon\lvert\boldsymbol{g}\rvert_{C_s^{1,\alpha}(\Gamma_\epsilon)}, A ϵ ϵ [ g ] C 0 , α ( T ) ≤ c ∥ g ∥ C s 0 , α ( Γ ϵ ) + cϵ ∣ g ∣ C s 1 , α ( Γ ϵ ) , ∥ A ‾ ϵ + [ g ] ∥ C 1 , α ( T ) ≤ c ϵ − 1 ∥ g ∥ C s 1 , α ( Γ ϵ ) . (63) \left\|\overline{\mathcal{A}}_\epsilon^{+}[\boldsymbol{g}]\right\|_{C^{1,\alpha}(\mathbb{T})} \leq c\epsilon^{-1}\|\boldsymbol{g}\|_{C_s^{1,\alpha}(\Gamma_\epsilon)}. \tag*{(63)} A ϵ + [ g ] C 1 , α ( T ) ≤ c ϵ − 1 ∥ g ∥ C s 1 , α ( Γ ϵ ) . (63) The proof of Lemma 1.9 appears in section 3.4.
Decomposition of SB DtN
Given a curved filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) \boldsymbol{X}(s) \in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) , we consider the slender body Dirichlet-to-Neumann map L ϵ − 1 \mathcal{L}_\epsilon^{-1} L ϵ − 1 along Γ ϵ \Gamma_\epsilon Γ ϵ and aim to extract the corresponding map L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 about the straight filament C ϵ C_\epsilon C ϵ as the leading order behavior.
Recalling the definition (27) of the operator Φ \Phi Φ identifying the normal/tangential directions along the curved and straight filaments, we begin by decomposing the boundary integral representation (44) as
( 1 2 I − D ‾ ) [ Φ − 1 v ( s ) ] − R D [ v ( s ) ] = S ‾ [ Φ − 1 w 0 Φ ] + R S [ w 0 Φ ] + Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] (64) \left(\frac{1}{2}\mathbb{I}-\overline{\mathcal{D}}\right)[\Phi^{-1}\boldsymbol{v}(s)]-\mathcal{R}_{D}[\boldsymbol{v}(s)] = \overline{\mathcal{S}}[\Phi^{-1}\boldsymbol{w}_{0}^{\Phi}] + \mathcal{R}_{S}[\boldsymbol{w}_{0}^{\Phi}] + \Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}\boldsymbol{w}(s,\theta)\,ds\right] \tag*{(64)} ( 2 1 I − D ) [ Φ − 1 v ( s )] − R D [ v ( s )] = S [ Φ − 1 w 0 Φ ] + R S [ w 0 Φ ] + Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] (64)
where w 0 Φ ( s , θ ) \boldsymbol{w}_{0}^{\Phi}(s,\theta) w 0 Φ ( s , θ ) is as in (28) and we define
R D [ v ( s ) ] : = Φ − 1 D [ v ( s ) ] − D ‾ [ Φ − 1 v ( s ) ] , \mathcal{R}_{D}[\boldsymbol{v}(s)] := \Phi^{-1}\mathcal{D}[\boldsymbol{v}(s)]-\overline{\mathcal{D}}[\Phi^{-1}\boldsymbol{v}(s)], R D [ v ( s )] := Φ − 1 D [ v ( s )] − D [ Φ − 1 v ( s )] ,
R S [ w 0 Φ ( s , θ ) ] : = Φ − 1 S [ w 0 Φ ( s , θ ) ] − S ‾ [ Φ − 1 w 0 Φ ( s , θ ) ] . (65) \mathcal{R}_{S}[\boldsymbol{w}_{0}^{\Phi}(s,\theta)] := \Phi^{-1}\mathcal{S}[\boldsymbol{w}_{0}^{\Phi}(s,\theta)]-\overline{\mathcal{S}}[\Phi^{-1}\boldsymbol{w}_{0}^{\Phi}(s,\theta)]. \tag*{(65)} R S [ w 0 Φ ( s , θ )] := Φ − 1 S [ w 0 Φ ( s , θ )] − S [ Φ − 1 w 0 Φ ( s , θ )] . (65)
We next would like to use Lemma 1.7 to apply S ‾ − 1 \overline{\mathcal{S}}^{-1} S − 1 to (64), but in order to do so we must first apply the projection operator P 01 \mathbb{P}_{01} P 01 (62) onto zero- and one-modes in θ \theta θ . Note that P 01 \mathbb{P}_{01} P 01 commutes with the straight single and double layer operators S ‾ \overline{\mathcal{S}} S and D ‾ \overline{\mathcal{D}} D about C ϵ C_\epsilon C ϵ . In particular, applying $\mathbb{P}_{01} to (64) , we have
( 1 2 I − D ‾ ) [ Φ − 1 v ( s ) ] − P 01 R D [ v ( s ) ] = S ‾ [ P 01 Φ − 1 w 0 Φ ] + P 01 R S [ w 0 Φ ] + P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] . \left(\frac{1}{2}\mathbf{I}-\overline{D}\right)[\Phi^{-1}v(s)]-\mathbb{P}_{01}\mathcal{R}_{D}[v(s)]=\overline{\mathcal{S}}\left[\mathbb{P}_{01}\Phi^{-1}w_{0}^{\Phi}\right]+\mathbb{P}_{01}\mathcal{R}_{S}\left[w_{0}^{\Phi}\right]+\mathbb{P}_{01}\Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w(s,\theta)\,\mathrm{d}s\right]. ( 2 1 I − D ) [ Φ − 1 v ( s )] − P 01 R D [ v ( s )] = S [ P 01 Φ − 1 w 0 Φ ] + P 01 R S [ w 0 Φ ] + P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] .
We may now apply S ‾ − 1 \overline{\mathcal{S}}^{-1} S − 1 to the above equation, multiply by ϵ \epsilon ϵ , and integrate in θ \theta θ to obtain
A ‾ ϵ [ ( 1 2 I − D ‾ ) [ Φ − 1 v ( s ) ] ] = ∫ 0 2 π P 01 ( Φ − 1 w 0 Φ ) ϵ d θ + A ‾ ϵ [ P 01 R D [ v ] ] + A ‾ ϵ [ P 01 R S [ w 0 Φ ] ] \overline{\mathcal{A}}_{\epsilon}\left[\left(\frac{1}{2}\mathbf{I}-\overline{D}\right)[\Phi^{-1}v(s)]\right]=\int_{0}^{2\pi}\mathbb{P}_{01}\left(\Phi^{-1}w_{0}^{\Phi}\right)\epsilon\,\mathrm{d}\theta+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{D}[v]\right]+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{S}\left[w_{0}^{\Phi}\right]\right] A ϵ [ ( 2 1 I − D ) [ Φ − 1 v ( s )] ] = ∫ 0 2 π P 01 ( Φ − 1 w 0 Φ ) ϵ d θ + A ϵ [ P 01 R D [ v ] ] + A ϵ [ P 01 R S [ w 0 Φ ] ]
+ A ‾ ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] = Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ − ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s + A ‾ ϵ [ P 01 R D [ v ] ] \begin{aligned}
\qquad+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w(s,\theta)\,\mathrm{d}s\right]\right] \\
&=\Phi^{-1}\int_{0}^{2\pi}w(s,\theta)\epsilon\,\mathrm{d}\theta-\int_{\mathbb{T}}\Phi^{-1}\int_{0}^{2\pi}w(s,\theta)\epsilon\,\mathrm{d}\theta\mathrm{d}s+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{D}[v]\right]
\end{aligned} + A ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] = Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ − ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s + A ϵ [ P 01 R D [ v ] ]
+ A ‾ ϵ [ P 01 R S [ w 0 Φ ] ] + A ‾ ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] = ( Φ − 1 f ) ( s ) + Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ 2 κ ^ d θ − ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s \begin{aligned}
\qquad+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{S}\left[w_{0}^{\Phi}\right]\right]+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w(s,\theta)\,\mathrm{d}s\right]\right] \\
&=(\Phi^{-1}f)(s)+\Phi^{-1}\int_{0}^{2\pi}w(s,\theta)\epsilon^{2}\widehat{\kappa}\,\mathrm{d}\theta-\int_{\mathbb{T}}\Phi^{-1}\int_{0}^{2\pi}w(s,\theta)\epsilon\,\mathrm{d}\theta\mathrm{d}s
\end{aligned} + A ϵ [ P 01 R S [ w 0 Φ ] ] + A ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] = ( Φ − 1 f ) ( s ) + Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ 2 κ d θ − ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s
+ A ‾ ϵ [ P 01 R D [ v ] ] + A ‾ ϵ [ P 01 R S [ w 0 Φ ] ] + A ‾ ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] . \qquad+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{D}[v]\right]+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{S}\left[w_{0}^{\Phi}\right]\right]+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w(s,\theta)\,\mathrm{d}s\right]\right]. + A ϵ [ P 01 R D [ v ] ] + A ϵ [ P 01 R S [ w 0 Φ ] ] + A ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] .
Here we have used the expression for f ( s ) f(s) f ( s ) from (44) . The left hand side is exactly L ‾ ϵ − 1 [ Φ − 1 v ( s ) ] \overline{\mathcal{L}}_{\epsilon}^{-1}[\Phi^{-1}v(s)] L ϵ − 1 [ Φ − 1 v ( s )] , and using that f ( s ) = L ϵ − 1 [ v ( s ) ] f(s)=\mathcal{L}_{\epsilon}^{-1}[v(s)] f ( s ) = L ϵ − 1 [ v ( s )] , we thus obtain the following decomposition of the slender body DtN map:
L ϵ − 1 [ v ( s ) ] = Φ L ‾ ϵ − 1 [ Φ − 1 v ( s ) ] − ∫ 0 2 π w ( s , θ ) ϵ 2 κ ^ d θ + Φ ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s (66) \mathcal{L}_{\epsilon}^{-1}[v(s)]=\Phi\overline{\mathcal{L}}_{\epsilon}^{-1}[\Phi^{-1}v(s)]-\int_{0}^{2\pi}w(s,\theta)\epsilon^{2}\widehat{\kappa}\,\mathrm{d}\theta+\Phi\int_{\mathbb{T}}\Phi^{-1}\int_{0}^{2\pi}w(s,\theta)\epsilon\,\mathrm{d}\theta\mathrm{d}s \tag*{(66)} L ϵ − 1 [ v ( s )] = Φ L ϵ − 1 [ Φ − 1 v ( s )] − ∫ 0 2 π w ( s , θ ) ϵ 2 κ d θ + Φ ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s (66)
− Φ A ‾ ϵ [ P 01 R D [ v ] ] − Φ A ‾ ϵ [ P 01 R S [ w 0 Φ ] ] − Φ A ‾ ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] . \qquad-\Phi\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{D}[v]\right]-\Phi\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{S}\left[w_{0}^{\Phi}\right]\right]-\Phi\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w(s,\theta)\,\mathrm{d}s\right]\right]. − Φ A ϵ [ P 01 R D [ v ] ] − Φ A ϵ [ P 01 R S [ w 0 Φ ] ] − Φ A ϵ [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w ( s , θ ) d s ] ] .
We will rely on this representation of L ϵ − 1 \mathcal{L}_{\epsilon}^{-1} L ϵ − 1 to prove Theorem 1.5 . In particular, we show that the five remainder terms appearing in (66) are lower order with respect to regularity or size in ϵ \epsilon ϵ . These terms may each be bounded using a combination of Lemma 1.9 and the following series of key lemmas.
The first two lemmas give the mapping properties of the single and double layer remainders R S \mathcal{R}_{S} R S and R D \mathcal{R}_{D} R D , respectively.
Lemma 1.10 (Single layer remainder). Let 0 < α < β < 1 0<\alpha<\beta<1 0 < α < β < 1 and consider a filament Σ ϵ \Sigma_{\epsilon} Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) X(s)\in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Let the single layer remainder R S \mathcal{R}_{S} R S be as defined in (65) , and let the map Φ \Phi Φ be as in (27) . Given φ ∈ C 0 , α ( Γ ϵ ) \varphi\in C^{0,\alpha}(\Gamma_{\epsilon}) φ ∈ C 0 , α ( Γ ϵ ) with ∫ T Φ − 1 φ ( s , θ ) d s = 0 \int_{\mathbb{T}}\Phi^{-1}\varphi(s,\theta)\,\mathrm{d}s=0 ∫ T Φ − 1 φ ( s , θ ) d s = 0 , R S \mathcal{R}_{S} R S may be decomposed as
R S [ φ ] = R S , ϵ [ φ ] + R S , + [ φ ] \mathcal{R}_{S}[\varphi]=\mathcal{R}_{S,\epsilon}[\varphi]+\mathcal{R}_{S,+}[\varphi] R S [ φ ] = R S , ϵ [ φ ] + R S , + [ φ ] where
∥ R S , ϵ [ φ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ , ∣ R S , ϵ [ φ ] ∣ C ˙ 1 , α + ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α \|\mathcal{R}_{S,\epsilon}[\varphi]\|_{C^{0,\alpha}}\leq c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{2-\alpha}\|\varphi\|_{L^{\infty}},\qquad\left|\mathcal{R}_{S,\epsilon}[\varphi]\right|_{\dot{C}^{1,\alpha^{+}}}\leq c(\kappa_{*},\alpha_{+},c_{\Gamma})\epsilon^{1-\alpha^{+}}\|\varphi\|_{C^{0,\alpha}} ∥ R S , ϵ [ φ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ , ∣ R S , ϵ [ φ ] ∣ C ˙ 1 , α + ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α ∥ R S , + [ φ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ , ∣ R S , + [ φ ] ∣ C ˙ 1 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ C 0 , α (67) \|\mathcal{R}_{S,+}[\varphi]\|_{C^{0,\alpha}}\leq c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{1-\alpha}\|\varphi\|_{L^{\infty}},\qquad\left|\mathcal{R}_{S,+}[\varphi]\right|_{\dot{C}^{1,\beta}}\leq c(\kappa_{*},\beta,c_{\Gamma})\epsilon^{1-2\beta}\|\varphi\|_{C^{0,\alpha}} \tag*{(67)} ∥ R S , + [ φ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ , ∣ R S , + [ φ ] ∣ C ˙ 1 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ C 0 , α (67) for any α + ∈ ( α , β ] \alpha^{+}\in(\alpha,\beta] α + ∈ ( α , β ] . Furthermore, given two nearby filaments with centerlines X ( a ) ( s ) X^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) X^{(b)}(s) X ( b ) ( s ) satisfying Lemma 1.1 , let φ 0 Φ ( a ) = φ − ∫ T ( Φ ( a ) ) − 1 φ ( s , θ ) d s \varphi_{0}^{\Phi^{(a)}}=\varphi-\int_{\mathbb{T}}(\Phi^{(a)})^{-1}\varphi(s,\theta)\,\mathrm{d}s φ 0 Φ ( a ) = φ − ∫ T ( Φ ( a ) ) − 1 φ ( s , θ ) d s and φ 0 Φ ( b ) = φ − ∫ T ( Φ ( b ) ) − 1 φ ( s , θ ) d s \varphi_{0}^{\Phi^{(b)}}=\varphi-\int_{\mathbb{T}}(\Phi^{(b)})^{-1}\varphi(s,\theta)\,\mathrm{d}s φ 0 Φ ( b ) = φ − ∫ T ( Φ ( b ) ) − 1 φ ( s , θ ) d s . The difference between the corresponding single layer remainders R S ( a ) \mathcal{R}_{S}^{(a)} R S ( a ) and R S ( b ) \mathcal{R}_{S}^{(b)} R S ( b ) may be decomposed
as above, and the components satisfy
∥ R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ \left\|\mathcal{R}_{S,\epsilon}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,\epsilon}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{C^{0,\alpha}} \le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_\Gamma)\epsilon^{2-\alpha}\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty} R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞
∥ R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] ∥ C ˙ 1 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α \left\|\mathcal{R}_{S,\epsilon}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,\epsilon}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{\dot{C}^{1,\alpha}} \le c(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_\Gamma)\epsilon^{1-\alpha^+}\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{2,\alpha^+}}\|\varphi\|_{C^{0,\alpha}} R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] C ˙ 1 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ φ ∥ C 0 , α
∥ R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ \left\|\mathcal{R}_{S,+}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,+}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{C^{0,\alpha}} \le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_\Gamma)\epsilon^{1-\alpha}\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty} R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞
∥ R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] ∥ C ˙ 1 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ C 0 , α . (68) \left\|\mathcal{R}_{S,+}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,+}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{\dot{C}^{1,\beta}} \le c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)},c_\Gamma)\epsilon^{1-2\beta}\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{2,\beta}}\|\varphi\|_{C^{0,\alpha}} . \tag*{(68)} R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] C ˙ 1 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β X ( a ) − X ( b ) C 2 , β ∥ φ ∥ C 0 , α . (68)
The proof of Lemma 1.10 appears in § 4.2 .
Lemma 1.11 (Double layer remainder). Let 0 < γ < β < 1 0 < \gamma< \beta< 1 0 < γ < β < 1 and consider a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) \boldsymbol{X}(s) \in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Let the double layer remainder R D \mathcal{R}_D R D be as defined in (65) . Given ψ ∈ C 0 , γ ( Γ ϵ ) \psi\in C^{0,\gamma}(\Gamma_\epsilon) ψ ∈ C 0 , γ ( Γ ϵ ) , we have that R D \mathcal{R}_D R D satisfies
∥ R D [ ψ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ (69) \|\mathcal{R}_D[\psi]\|_{C^{1,\gamma}} \le c(\kappa_{*,\gamma^+},c_\Gamma)\epsilon^{-\gamma^+}\|\psi\|_{C^{0,\gamma}} \tag*{(69)} ∥ R D [ ψ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ (69) for any γ + ∈ ( γ , β ] \gamma^+ \in(\gamma,\beta] γ + ∈ ( γ , β ] . Furthermore, given two nearby filaments with centerlines X ( a ) ( s ) \boldsymbol{X}^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) \boldsymbol{X}^{(b)}(s) X ( b ) ( s ) satisfying Lemma 1.1, the difference between their corresponding double layer remainders R D ( a ) \mathcal{R}_D^{(a)} R D ( a ) and R D ( b ) \mathcal{R}_D^{(b)} R D ( b ) satisfies
∥ R D ( a ) [ ψ ] − R D ( b ) [ ψ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (70) \|\mathcal{R}_D^{(a)}[\psi]-\mathcal{R}_D^{(b)}[\psi]\|_{C^{1,\gamma}} \le c(\kappa_{*,\gamma^+}^{(a)},\kappa_{*,\gamma^+}^{(b)},c_\Gamma)\epsilon^{-\gamma^+}\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\gamma^+}}\|\psi\|_{C^{0,\gamma}} . \tag*{(70)} ∥ R D ( a ) [ ψ ] − R D ( b ) [ ψ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (70)
The proof of Lemma 1.11 appears in § 4.3 . In addition to Lemmas 1.10 and 1.11, we will require a bound for the single layer operator applied to constant-in-s s s functions along Γ ϵ \Gamma_\epsilon Γ ϵ . We show the following.
Lemma 1.12 (Single layer applied to constant-in-s s s ). Let 0 < α < β < 1 0 < \alpha< \beta< 1 0 < α < β < 1 and consider a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) \boldsymbol{X}(s) \in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Given h ‾ ( θ ) ∈ C 0 , α ( 2 π T ) \overline{\boldsymbol{h}}(\theta) \in C^{0,\alpha}(2\pi\mathbb{T}) h ( θ ) ∈ C 0 , α ( 2 π T ) of the form h ‾ = h ‾ 1 ( θ ) e z + h ‾ 2 ( θ ) e x + h ‾ 3 ( θ ) e y \overline{\boldsymbol{h}}=\overline{h}_1(\theta)\boldsymbol{e}_z+\overline{h}_2(\theta)\boldsymbol{e}_x+\overline{h}_3(\theta)\boldsymbol{e}_y h = h 1 ( θ ) e z + h 2 ( θ ) e x + h 3 ( θ ) e y , and recalling the map Φ \Phi Φ (27) , we may decompose
S [ Φ h ‾ ( θ ) ] = H ϵ [ Φ h ‾ ( θ ) ] + H + [ Φ h ‾ ( θ ) ] S[\Phi\overline{\boldsymbol{h}}(\theta)]=\mathcal{H}_\epsilon[\Phi\overline{\boldsymbol{h}}(\theta)]+\mathcal{H}_+[\Phi\overline{\boldsymbol{h}}(\theta)] S [ Φ h ( θ )] = H ϵ [ Φ h ( θ )] + H + [ Φ h ( θ )] where
∥ H ϵ [ Φ h ‾ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ‾ ∥ L ∞ , ∥ H ϵ [ Φ h ‾ ] ∥ C ˙ s 1 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ‾ ∥ C 0 , α \|\mathcal{H}_\epsilon[\Phi\overline{\boldsymbol{h}}]\|_{C^{0,\alpha}}\le c(\kappa_{*,\alpha},c_\Gamma)\epsilon^{2-\alpha}\|\overline{\boldsymbol{h}}\|_{L^\infty},\qquad\|\mathcal{H}_\epsilon[\Phi\overline{\boldsymbol{h}}]\|_{\dot{C}^{1,\alpha}_s}\le c(\kappa_{*,\alpha^+},c_\Gamma)\epsilon^{1-\alpha^+}\|\overline{\boldsymbol{h}}\|_{C^{0,\alpha}} ∥ H ϵ [ Φ h ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ∥ L ∞ , ∥ H ϵ [ Φ h ] ∥ C ˙ s 1 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ∥ C 0 , α ∥ H + [ Φ h ‾ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ h ‾ ∥ L ∞ , ∥ H + [ Φ h ‾ ] ∥ C ˙ s 1 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ‾ ∥ C 0 , α (71) \|\mathcal{H}_+[\Phi\overline{\boldsymbol{h}}]\|_{C^{0,\alpha}}\le c(\kappa_{*,\alpha},c_\Gamma)\epsilon^{1-\alpha}\|\overline{\boldsymbol{h}}\|_{L^\infty},\qquad\|\mathcal{H}_+[\Phi\overline{\boldsymbol{h}}]\|_{\dot{C}^{1,\beta}_s}\le c(\kappa_{*,\beta},c_\Gamma)\epsilon^{1-2\beta}\|\overline{\boldsymbol{h}}\|_{C^{0,\alpha}} \tag*{(71)} ∥ H + [ Φ h ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ h ∥ L ∞ , ∥ H + [ Φ h ] ∥ C ˙ s 1 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ∥ C 0 , α (71) for any α + ∈ ( α , β ] \alpha^+ \in(\alpha,\beta] α + ∈ ( α , β ] . In addition, given two nearby curves X ( a ) , X ( b ) \boldsymbol{X}^{(a)},\boldsymbol{X}^{(b)} X ( a ) , X ( b ) satisfying Lemma 1.1, the corresponding differences H ϵ ( a ) − H ϵ ( b ) \mathcal{H}_\epsilon^{(a)}-\mathcal{H}_\epsilon^{(b)} H ϵ ( a ) − H ϵ ( b ) and H + ( a ) − H + ( b ) \mathcal{H}_+^{(a)}-\mathcal{H}_+^{(b)} H + ( a ) − H + ( b ) satisfy
∥ H ϵ ( a ) [ Φ ( a ) h ‾ ] − H ϵ ( b ) [ Φ ( b ) h ‾ ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ‾ ∥ L ∞ \|\mathcal{H}_\epsilon^{(a)}[\Phi^{(a)}\overline{\boldsymbol{h}}]-\mathcal{H}_\epsilon^{(b)}[\Phi^{(b)}\overline{\boldsymbol{h}}]\|_{C^{0,\alpha}}\le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_\Gamma)\epsilon^{2-\alpha}\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha}}\|\overline{\boldsymbol{h}}\|_{L^\infty} ∥ H ϵ ( a ) [ Φ ( a ) h ] − H ϵ ( b ) [ Φ ( b ) h ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ∥ L ∞ ∥ H ϵ ( a ) [ Φ ( a ) h ‾ ] − H ϵ ( b ) [ Φ ( b ) h ‾ ] ∥ C ˙ s 1 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ h ‾ ∥ C 0 , α \|\mathcal{H}_\epsilon^{(a)}[\Phi^{(a)}\overline{\boldsymbol{h}}]-\mathcal{H}_\epsilon^{(b)}[\Phi^{(b)}\overline{\boldsymbol{h}}]\|_{\dot{C}^{1,\alpha}_s}\le c(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_\Gamma)\epsilon^{1-\alpha^+}\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha^+}}\|\overline{\boldsymbol{h}}\|_{C^{0,\alpha}} ∥ H ϵ ( a ) [ Φ ( a ) h ] − H ϵ ( b ) [ Φ ( b ) h ] ∥ C ˙ s 1 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ h ∥ C 0 , α ∥ H + ( a ) [ Φ ( a ) h ‾ ] − H + ( b ) [ Φ ( b ) h ‾ ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ‾ ∥ L ∞ \|\mathcal{H}_+^{(a)}[\Phi^{(a)}\overline{\boldsymbol{h}}]-\mathcal{H}_+^{(b)}[\Phi^{(b)}\overline{\boldsymbol{h}}]\|_{C^{0,\alpha}}\le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_\Gamma)\epsilon^{1-\alpha}\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha}}\|\overline{\boldsymbol{h}}\|_{L^\infty} ∥ H + ( a ) [ Φ ( a ) h ] − H + ( b ) [ Φ ( b ) h ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ∥ L ∞ ∥ H + ( a ) [ Φ ( a ) h ‾ ] − H + ( b ) [ Φ ( b ) h ‾ ] ∥ C ˙ s 1 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ h ‾ ∥ C 0 , α . (72) \|\mathcal{H}_+^{(a)}[\Phi^{(a)}\overline{\boldsymbol{h}}]-\mathcal{H}_+^{(b)}[\Phi^{(b)}\overline{\boldsymbol{h}}]\|_{\dot{C}^{1,\beta}_s}\le c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)},c_\Gamma)\epsilon^{1-2\beta}\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\beta}}\|\overline{\boldsymbol{h}}\|_{C^{0,\alpha}} . \tag*{(72)} ∥ H + ( a ) [ Φ ( a ) h ] − H + ( b ) [ Φ ( b ) h ] ∥ C ˙ s 1 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ h ∥ C 0 , α . (72)
The proof of Lemma 1.12 appears in § 4.4 .
The next lemma concerns bounds for the full Neumann data w ( s , θ ) \boldsymbol{w}(s,\theta) w ( s , θ ) given by (42) and appearing in the decomposition (66) .
Lemma 1.13 (Bounds for w ( s , θ ) w(s,\theta) w ( s , θ ) ). Let 0 < α < γ < β < 1 0<\alpha<\gamma<\beta<1 0 < α < γ < β < 1 and consider a slender body Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) \mathbf{X}(s)\in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Given θ \theta θ -independent Dirichlet data v ( s ) ∈ C 1 , α ( T ) \mathbf{v}(s)\in C^{1,\alpha}(\mathbb{T}) v ( s ) ∈ C 1 , α ( T ) along the filament surface Γ ϵ \Gamma_\epsilon Γ ϵ , let u ( x ) \mathbf{u}(\mathbf{x}) u ( x ) , x ∈ Ω ϵ \mathbf{x}\in\Omega_\epsilon x ∈ Ω ϵ , denote the corresponding solution to the Stokes equations in Ω ϵ \Omega_\epsilon Ω ϵ , and let w ( s , θ ) \mathbf{w}(s,\theta) w ( s , θ ) denote the corresponding Neumann boundary value w = − σ [ u ] n ∣ Γ ϵ \mathbf{w}=-\boldsymbol{\sigma}[\mathbf{u}]\mathbf{n}|_{\Gamma_\epsilon} w = − σ [ u ] n ∣ Γ ϵ . Then w \mathbf{w} w may be decomposed as
w ( s , θ ) = w ϵ ( s , θ ) + w + ( s , θ ) \mathbf{w}(s,\theta)=\mathbf{w}_\epsilon(s,\theta)+\mathbf{w}_+(s,\theta) w ( s , θ ) = w ϵ ( s , θ ) + w + ( s , θ ) where
∥ w ϵ ∥ C 0 , α ( Γ ϵ ) ≤ c ( κ ∗ , c Γ ) ∥ v ∥ C 1 , α ( T ) ∥ w + ∥ C 0 , γ ( Γ ϵ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ ( T ) (73) \begin{aligned}
\|\mathbf{w}_\epsilon\|_{C^{0,\alpha}(\Gamma_\epsilon)}&\leq c(\kappa_*,c_\Gamma)\|\mathbf{v}\|_{C^{1,\alpha}(\mathbb{T})}\\
\|\mathbf{w}_+\|_{C^{0,\gamma}(\Gamma_\epsilon)}&\leq c(\epsilon,\kappa_{*,\gamma_+},c_\Gamma)\|\mathbf{v}\|_{C^{0,\gamma}(\mathbb{T})} \tag*{(73)}
\end{aligned} ∥ w ϵ ∥ C 0 , α ( Γ ϵ ) ∥ w + ∥ C 0 , γ ( Γ ϵ ) ≤ c ( κ ∗ , c Γ ) ∥ v ∥ C 1 , α ( T ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ ( T ) (73) for any γ + ∈ ( γ , β ] \gamma_+\in(\gamma,\beta] γ + ∈ ( γ , β ] . In addition, given two nearby filaments Σ ϵ ( a ) \Sigma_\epsilon^{(a)} Σ ϵ ( a ) , Σ ϵ ( b ) \Sigma_\epsilon^{(b)} Σ ϵ ( b ) with centerlines X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) satisfying Lemma 1.1, we have
∥ w ϵ ( a ) − w ϵ ( b ) ∥ C 0 , α ( Γ ϵ ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ( T ) ∥ v ∥ C 1 , α ( T ) ∥ w + ( a ) − w + ( b ) ∥ C 0 , γ ( Γ ϵ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ( T ) ∥ v ∥ C 0 , γ ( T ) , (74) \begin{aligned}
\|\mathbf{w}_\epsilon^{(a)}-\mathbf{w}_\epsilon^{(b)}\|_{C^{0,\alpha}(\Gamma_\epsilon)}
&\leq c(\kappa_*^{(a)},\kappa_*^{(b)},c_\Gamma)\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2(\mathbb{T})}\|\mathbf{v}\|_{C^{1,\alpha}(\mathbb{T})}\\
\|\mathbf{w}_+^{(a)}-\mathbf{w}_+^{(b)}\|_{C^{0,\gamma}(\Gamma_\epsilon)}
&\leq c(\epsilon,\kappa_{*,\gamma_+}^{(a)},\kappa_{*,\gamma_+}^{(b)},c_\Gamma)\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma_+}(\mathbb{T})}\|\mathbf{v}\|_{C^{0,\gamma}(\mathbb{T})}, \tag*{(74)}
\end{aligned} ∥ w ϵ ( a ) − w ϵ ( b ) ∥ C 0 , α ( Γ ϵ ) ∥ w + ( a ) − w + ( b ) ∥ C 0 , γ ( Γ ϵ ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ( T ) ∥ v ∥ C 1 , α ( T ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ( T ) ∥ v ∥ C 0 , γ ( T ) , (74) where w ( a ) \mathbf{w}^{(a)} w ( a ) , w ( b ) \mathbf{w}^{(b)} w ( b ) denote the Neumann data along filaments Σ ϵ ( a ) \Sigma_\epsilon^{(a)} Σ ϵ ( a ) and Σ ϵ ( b ) \Sigma_\epsilon^{(b)} Σ ϵ ( b ) , respectively.
In both (73) and (74), the ϵ \epsilon ϵ -dependence is explicit in the bounds for w ϵ \mathbf{w}_\epsilon w ϵ but not for w + \mathbf{w}_+ w + . The proof of Lemma 1.13 is given in section 5.
The final key lemma concerns Hölder space estimates for the slender body NtD map, which will be used in inverting the representation (66) of L ϵ − 1 \mathcal{L}_\epsilon^{-1} L ϵ − 1 to obtain a decomposition of L ϵ \mathcal{L}_\epsilon L ϵ instead.
Lemma 1.14 (Slender body NtD in Hölder spaces). For 0 < α < α + < 1 0<\alpha<\alpha_+<1 0 < α < α + < 1 , given a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 3 , α ( T ) \mathbf{X}(s)\in C^{3,\alpha}(\mathbb{T}) X ( s ) ∈ C 3 , α ( T ) and slender body force data f ( s ) ∈ C 0 , α ( T ) \mathbf{f}(s)\in C^{0,\alpha}(\mathbb{T}) f ( s ) ∈ C 0 , α ( T ) , the slender body NtD map L ϵ [ f ] \mathcal{L}_\epsilon[\mathbf{f}] L ϵ [ f ] (12) satisfies
∥ L ϵ [ f ] ∥ C 1 , α ( T ) ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) ∥ f ∥ C 0 , α ( T ) . (75) \|\mathcal{L}_\epsilon[\mathbf{f}]\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha}},c_\Gamma)\|\mathbf{f}\|_{C^{0,\alpha}(\mathbb{T})}. \tag*{(75)} ∥ L ϵ [ f ] ∥ C 1 , α ( T ) ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) ∥ f ∥ C 0 , α ( T ) . (75) In addition, given two nearby filaments with centerlines X ( a ) ( s ) \mathbf{X}^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) \mathbf{X}^{(b)}(s) X ( b ) ( s ) satisfying Lemma 1.1, the difference between the corresponding NtD maps ( L ϵ ( a ) − L ϵ ( b ) ) [ f ] (\mathcal{L}_\epsilon^{(a)}-\mathcal{L}_\epsilon^{(b)})[\mathbf{f}] ( L ϵ ( a ) − L ϵ ( b ) ) [ f ] satisfies
∥ ( L ϵ ( a ) − L ϵ ( b ) ) [ f ] ∥ C 1 , α ( T ) ≤ c ( ϵ , ∥ X s s ( a ) ∥ C 1 , α , ∥ X s s ( b ) ∥ C 1 , α , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ f ∥ C 0 , α ( T ) . (76) \|(\mathcal{L}_\epsilon^{(a)}-\mathcal{L}_\epsilon^{(b)})[\mathbf{f}]\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\epsilon,\|\mathbf{X}_{ss}^{(a)}\|_{C^{1,\alpha}},\|\mathbf{X}_{ss}^{(b)}\|_{C^{1,\alpha}},c_\Gamma)\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\alpha_+}(\mathbb{T})}\|\mathbf{f}\|_{C^{0,\alpha}(\mathbb{T})}. \tag*{(76)} ∥ ( L ϵ ( a ) − L ϵ ( b ) ) [ f ] ∥ C 1 , α ( T ) ≤ c ( ϵ , ∥ X ss ( a ) ∥ C 1 , α , ∥ X ss ( b ) ∥ C 1 , α , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ f ∥ C 0 , α ( T ) . (76) Note that additional C 3 , α C^{3,\alpha} C 3 , α regularity is needed on the filament centerline to show Lemma 1.14. Furthermore, the ϵ \epsilon ϵ -dependence in the bounds (75) and (76) is not explicit. The proof of Lemma 1.14 appears in section 6.
Proofs of Theorems 1.5 and 1.2 Given the decomposition (66) of L ϵ − 1 \mathcal{L}_\epsilon^{-1} L ϵ − 1 along with Lemmas 1.9, 1.10, 1.11, 1.12, 1.13, and 1.14, we may proceed immediately to the proof of Theorem 1.5 in section 2.1. Given Theorem 1.5 and Lemma 1.4, we may then prove Theorem 1.2 for the simple evolution (20) in section 2.2. The remainder of this paper is devoted to the proofs of each of the above key lemmas.
Proof of Theorem 1.5: operator decomposition
We begin by stating some properties of the map Φ \Phi Φ (27) used to identify the tangential and normal directions along curved Γ ϵ \Gamma_\epsilon Γ ϵ with the tangential and normal directions along the straight filament C ϵ C_\epsilon C ϵ . Note that the map Φ \Phi Φ commutes with derivatives in θ \theta θ along Γ ϵ \Gamma_\epsilon Γ ϵ but does not commute with derivatives in s s s . In particular, we have
∂ s ( Φ g ) = Φ ( ∂ s g ) − g ⋅ ( κ 1 e x + κ 2 e y ) e t + g ⋅ ( κ 1 e z − κ 3 e y ) e n 1 + g ⋅ ( κ 2 e z + κ 3 e x ) e n 2 , ∂ s ( Φ − 1 h ) = Φ − 1 ( ∂ s h ) + h ⋅ ( κ 1 e n 1 + κ 2 e n 2 ) e z + h ⋅ ( κ 3 e n 2 − κ 1 e t ) e x − h ⋅ ( κ 3 e n 1 + κ 2 e t ) e y . (77) \begin{aligned}
\partial_s(\Phi g) &= \Phi(\partial_s g)-g\cdot(\kappa_1e_x+\kappa_2e_y)e_t+g\cdot(\kappa_1e_z-\kappa_3e_y)e_{n_1}\\
&\quad+g\cdot(\kappa_2e_z+\kappa_3e_x)e_{n_2},\\
\partial_s(\Phi^{-1}h) &= \Phi^{-1}(\partial_s h)+h\cdot(\kappa_1e_{n_1}+\kappa_2e_{n_2})e_z+h\cdot(\kappa_3e_{n_2}-\kappa_1e_t)e_x\\
&\quad-h\cdot(\kappa_3e_{n_1}+\kappa_2e_t)e_y. \tag*{(77)}
\end{aligned} ∂ s ( Φ g ) ∂ s ( Φ − 1 h ) = Φ ( ∂ s g ) − g ⋅ ( κ 1 e x + κ 2 e y ) e t + g ⋅ ( κ 1 e z − κ 3 e y ) e n 1 + g ⋅ ( κ 2 e z + κ 3 e x ) e n 2 , = Φ − 1 ( ∂ s h ) + h ⋅ ( κ 1 e n 1 + κ 2 e n 2 ) e z + h ⋅ ( κ 3 e n 2 − κ 1 e t ) e x − h ⋅ ( κ 3 e n 1 + κ 2 e t ) e y . (77)
From (77) , we can see that
∥ [ ∂ s , Φ ] g ∥ L ∞ ≤ c ( κ ∗ ) ∥ g ∥ L ∞ , ∥ [ ∂ s , Φ ] g ∥ C 0 , α ≤ c ( κ ∗ , α ) ∥ g ∥ C 0 , α , ∥ [ ∂ s , Φ − 1 ] h ∥ L ∞ ≤ c ( κ ∗ ) ∥ h ∥ L ∞ , ∥ [ ∂ s , Φ − 1 ] h ∥ C 0 , α ≤ c ( κ ∗ , α ) ∥ h ∥ C 0 , α . (78) \begin{aligned}
\left\|[\partial_s,\Phi]g\right\|_{L^\infty}&\le c(\kappa_*)\left\|g\right\|_{L^\infty},&
\left\|[\partial_s,\Phi]g\right\|_{C^{0,\alpha}}&\le c(\kappa_{*,\alpha})\left\|g\right\|_{C^{0,\alpha}},\\
\left\|[\partial_s,\Phi^{-1}]h\right\|_{L^\infty}&\le c(\kappa_*)\left\|h\right\|_{L^\infty},&
\left\|[\partial_s,\Phi^{-1}]h\right\|_{C^{0,\alpha}}&\le c(\kappa_{*,\alpha})\left\|h\right\|_{C^{0,\alpha}}. \tag*{(78)}
\end{aligned} ∥ [ ∂ s , Φ ] g ∥ L ∞ [ ∂ s , Φ − 1 ] h L ∞ ≤ c ( κ ∗ ) ∥ g ∥ L ∞ , ≤ c ( κ ∗ ) ∥ h ∥ L ∞ , ∥ [ ∂ s , Φ ] g ∥ C 0 , α [ ∂ s , Φ − 1 ] h C 0 , α ≤ c ( κ ∗ , α ) ∥ g ∥ C 0 , α , ≤ c ( κ ∗ , α ) ∥ h ∥ C 0 , α . (78)
Furthermore, given two nearby filaments with centerlines X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1 , we note the following form of the difference between the corresponding maps Φ ( a ) \Phi^{(a)} Φ ( a ) and Φ ( b ) \Phi^{(b)} Φ ( b ) :
Φ ( a ) g − Φ ( b ) g = ( g ⋅ e z ) ( e t ( a ) − e t ( b ) ) + ( g ⋅ e x ) ( e n 1 ( a ) − e n 1 ( b ) ) + ( g ⋅ e y ) ( e n 2 ( a ) − e n 2 ( b ) ) \Phi^{(a)}g-\Phi^{(b)}g=(g\cdot e_z)(e_t^{(a)}-e_t^{(b)})+(g\cdot e_x)(e_{n_1}^{(a)}-e_{n_1}^{(b)})+(g\cdot e_y)(e_{n_2}^{(a)}-e_{n_2}^{(b)}) Φ ( a ) g − Φ ( b ) g = ( g ⋅ e z ) ( e t ( a ) − e t ( b ) ) + ( g ⋅ e x ) ( e n 1 ( a ) − e n 1 ( b ) ) + ( g ⋅ e y ) ( e n 2 ( a ) − e n 2 ( b ) )
( Φ ( a ) ) − 1 h − ( Φ ( b ) ) − 1 h = h ⋅ ( e t ( a ) − e t ( b ) ) e z + h ⋅ ( e n 1 ( a ) − e n 1 ( b ) ) e x + h ⋅ ( e n 2 ( a ) − e n 2 ( b ) ) e y . (79) (\Phi^{(a)})^{-1}h-(\Phi^{(b)})^{-1}h=h\cdot(e_t^{(a)}-e_t^{(b)})e_z+h\cdot(e_{n_1}^{(a)}-e_{n_1}^{(b)})e_x+h\cdot(e_{n_2}^{(a)}-e_{n_2}^{(b)})e_y. \tag*{(79)} ( Φ ( a ) ) − 1 h − ( Φ ( b ) ) − 1 h = h ⋅ ( e t ( a ) − e t ( b ) ) e z + h ⋅ ( e n 1 ( a ) − e n 1 ( b ) ) e x + h ⋅ ( e n 2 ( a ) − e n 2 ( b ) ) e y . (79)
Recalling the definition (28) of h 0 Φ h_0^\Phi h 0 Φ , we note that the subtracted mean differs slightly for different curves. However, using the definition (27) of Φ \Phi Φ , we may write
( Φ ( a ) ) − 1 h 0 Φ ( a ) − ( Φ ( b ) ) − 1 h 0 Φ ( b ) = ( Φ ( a ) ) − 1 h − ( Φ ( b ) ) − 1 h − e z ∫ T h ⋅ ( e t ( a ) − e t ( b ) ) d s − e x ∫ T h ⋅ ( e n 1 ( a ) − e n 1 ( b ) ) d s − e y ∫ T h ⋅ ( e n 2 ( a ) − e n 2 ( b ) ) d s . (80) \begin{aligned}
(\Phi^{(a)})^{-1}h_0^{\Phi^{(a)}}-(\Phi^{(b)})^{-1}h_0^{\Phi^{(b)}}={}&(\Phi^{(a)})^{-1}h-(\Phi^{(b)})^{-1}h-e_z\int_{\mathbb{T}}h\cdot(e_t^{(a)}-e_t^{(b)})\,ds\\
&-e_x\int_{\mathbb{T}}h\cdot(e_{n_1}^{(a)}-e_{n_1}^{(b)})\,ds-e_y\int_{\mathbb{T}}h\cdot(e_{n_2}^{(a)}-e_{n_2}^{(b)})\,ds. \tag*{(80)}
\end{aligned} ( Φ ( a ) ) − 1 h 0 Φ ( a ) − ( Φ ( b ) ) − 1 h 0 Φ ( b ) = ( Φ ( a ) ) − 1 h − ( Φ ( b ) ) − 1 h − e z ∫ T h ⋅ ( e t ( a ) − e t ( b ) ) d s − e x ∫ T h ⋅ ( e n 1 ( a ) − e n 1 ( b ) ) d s − e y ∫ T h ⋅ ( e n 2 ( a ) − e n 2 ( b ) ) d s . (80)
Combining (79) and (80) , we thus have
∥ ( Φ ( a ) ) − 1 h 0 Φ ( a ) − ( Φ ( b ) ) − 1 h 0 Φ ( b ) ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ∥ X ( a ) − X ( b ) ∥ C 2 ∥ h ∥ L ∞ \left\|(\Phi^{(a)})^{-1}h_0^{\Phi^{(a)}}-(\Phi^{(b)})^{-1}h_0^{\Phi^{(b)}}\right\|_{L^\infty}\le c(\kappa_*^{(a)},\kappa_*^{(b)})\left\|X^{(a)}-X^{(b)}\right\|_{C^2}\left\|h\right\|_{L^\infty} ( Φ ( a ) ) − 1 h 0 Φ ( a ) − ( Φ ( b ) ) − 1 h 0 Φ ( b ) L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) X ( a ) − X ( b ) C 2 ∥ h ∥ L ∞
∥ ( Φ ( a ) ) − 1 h 0 Φ ( a ) − ( Φ ( b ) ) − 1 h 0 Φ ( b ) ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) ) ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ∥ C 0 , α . (81) \left\|(\Phi^{(a)})^{-1}h_0^{\Phi^{(a)}}-(\Phi^{(b)})^{-1}h_0^{\Phi^{(b)}}\right\|_{C^{0,\alpha}}\le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)})\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha}}\left\|h\right\|_{C^{0,\alpha}}. \tag*{(81)} ( Φ ( a ) ) − 1 h 0 Φ ( a ) − ( Φ ( b ) ) − 1 h 0 Φ ( b ) C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) ) X ( a ) − X ( b ) C 2 , α ∥ h ∥ C 0 , α . (81)
In addition, from (77) , we obtain
∥ [ ∂ s , ( Φ ( a ) ) − 1 ] h 0 Φ ( a ) − [ ∂ s , ( Φ ( b ) ) − 1 ] h 0 Φ ( b ) ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ∥ X ( a ) − X ( b ) ∥ C 2 ∥ h ∥ L ∞ \left\|[\partial_s,(\Phi^{(a)})^{-1}]h_0^{\Phi^{(a)}}-[\partial_s,(\Phi^{(b)})^{-1}]h_0^{\Phi^{(b)}}\right\|_{L^\infty}\le c(\kappa_*^{(a)},\kappa_*^{(b)})\left\|X^{(a)}-X^{(b)}\right\|_{C^2}\left\|h\right\|_{L^\infty} [ ∂ s , ( Φ ( a ) ) − 1 ] h 0 Φ ( a ) − [ ∂ s , ( Φ ( b ) ) − 1 ] h 0 Φ ( b ) L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) X ( a ) − X ( b ) C 2 ∥ h ∥ L ∞
∥ [ ∂ s , ( Φ ( a ) ) − 1 ] h 0 Φ ( a ) − [ ∂ s , ( Φ ( b ) ) − 1 ] h 0 Φ ( b ) ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) ) ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ∥ C 0 , α . (82) \left\|[\partial_s,(\Phi^{(a)})^{-1}]h_0^{\Phi^{(a)}}-[\partial_s,(\Phi^{(b)})^{-1}]h_0^{\Phi^{(b)}}\right\|_{C^{0,\alpha}}\le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)})\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha}}\left\|h\right\|_{C^{0,\alpha}}. \tag*{(82)} [ ∂ s , ( Φ ( a ) ) − 1 ] h 0 Φ ( a ) − [ ∂ s , ( Φ ( b ) ) − 1 ] h 0 Φ ( b ) C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) ) X ( a ) − X ( b ) C 2 , α ∥ h ∥ C 0 , α . (82)
We may now proceed with the decomposition (29) of L ϵ − 1 \mathcal{L}_\epsilon^{-1} L ϵ − 1 . Recalling the representation (66) of L ϵ − 1 \mathcal{L}_\epsilon^{-1} L ϵ − 1 , we first separate the remainder terms into terms which are small in ϵ \epsilon ϵ and terms which are smoother. Using the decompositions A ϵ ∓ \mathcal{A}_\epsilon^{\mp} A ϵ ∓ , R S , ϵ / + \mathcal{R}_{S,\epsilon/\!+} R S , ϵ / + , H ϵ / + \mathcal{H}_{\epsilon/\!+} H ϵ / + and w ϵ / + w_{\epsilon/\!+} w ϵ / + from Lemmas 1.9 , 1.10 , 1.12 , and 1.13 , respectively, we define the terms R d , ϵ \mathcal{R}_{d,\epsilon} R d , ϵ and R d , + \mathcal{R}_{d,+} R d , + in (29) as
R d , ϵ [ v ( s ) ] = − ∫ 0 2 π w ϵ ( s , θ ) ϵ κ ^ 2 d θ b − Φ A ‾ ϵ − [ P 01 R S , ϵ [ ( w ϵ ) Φ ] ] − Φ A ‾ ϵ − [ P 01 Φ − 1 H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] ] \mathcal{R}_{d,\epsilon}[v(s)]=-\int_0^{2\pi}w_\epsilon(s,\theta)\,\epsilon\widehat{\kappa}^2\,d\theta\,b-\overline{\Phi\mathcal{A}}_\epsilon^{\, -}\left[\mathbb{P}_{01}\mathcal{R}_{S,\epsilon}[(w_\epsilon)^\Phi]\right]-\overline{\Phi\mathcal{A}}_\epsilon^{\, -}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{H}_\epsilon\left[\Phi\left[\int_{\mathbb{T}}\Phi^{-1}w_\epsilon\,ds\right]\right]\right] R d , ϵ [ v ( s )] = − ∫ 0 2 π w ϵ ( s , θ ) ϵ κ 2 d θ b − Φ A ϵ − [ P 01 R S , ϵ [( w ϵ ) Φ ] ] − Φ A ϵ − [ P 01 Φ − 1 H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] ]
R d , + [ v ( s ) ] = − ∫ 0 2 π w + ( s , θ ) ϵ κ ^ 2 d θ b + Φ ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s − Φ A ‾ ϵ − [ P 01 R D [ v ] ] − Φ A ‾ ϵ + [ P 01 R S [ w 0 Φ ] ] − Φ A ‾ ϵ − [ P 01 R S , + [ w 0 Φ ] ] − Φ A ‾ ϵ − [ P 01 R S , ϵ [ ( w + ) 0 Φ ] ] − Φ A ‾ ϵ + [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w d s ] ] − Φ A ‾ ϵ − [ P 01 Φ − 1 H + [ Φ ∫ T Φ − 1 w d s ] ] . (83) \begin{aligned}
\mathcal{R}_{d,+}[v(s)]={}&-\int_0^{2\pi}w_+(s,\theta)\,\epsilon\widehat{\kappa}^2\,d\theta\,b+\Phi\int_{\mathbb{T}}\Phi^{-1}\int_0^{2\pi}w(s,\theta)\,\epsilon\,d\theta\,ds-\overline{\Phi\mathcal{A}}_\epsilon^{\, -}\left[\mathbb{P}_{01}\mathcal{R}_D[v]\right]\\
&-\overline{\Phi\mathcal{A}}_\epsilon^{\, +}\left[\mathbb{P}_{01}\mathcal{R}_S[w_0^\Phi]\right]-\overline{\Phi\mathcal{A}}_\epsilon^{\, -}\left[\mathbb{P}_{01}\mathcal{R}_{S,+}[w_0^\Phi]\right]-\overline{\Phi\mathcal{A}}_\epsilon^{\, -}\left[\mathbb{P}_{01}\mathcal{R}_{S,\epsilon}[(w_+)_{0}^\Phi]\right]\\
&-\overline{\Phi\mathcal{A}}_\epsilon^{\, +}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w\,ds\right]\right]-\overline{\Phi\mathcal{A}}_\epsilon^{\, -}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{H}_+\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w\,ds\right]\right]. \tag*{(83)}
\end{aligned} R d , + [ v ( s )] = − ∫ 0 2 π w + ( s , θ ) ϵ κ 2 d θ b + Φ ∫ T Φ − 1 ∫ 0 2 π w ( s , θ ) ϵ d θ d s − Φ A ϵ − [ P 01 R D [ v ] ] − Φ A ϵ + [ P 01 R S [ w 0 Φ ] ] − Φ A ϵ − [ P 01 R S , + [ w 0 Φ ] ] − Φ A ϵ − [ P 01 R S , ϵ [( w + ) 0 Φ ] ] − Φ A ϵ + [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w d s ] ] − Φ A ϵ − [ P 01 Φ − 1 H + [ Φ ∫ T Φ − 1 w d s ] ] . (83)
− Φ ‾ A ϵ E ‾ [ P 01 Φ − 1 H ϵ [ Φ [ ∫ T Φ − 1 w + d s ] ] ] . -\overline{\Phi}\mathcal{A}_{\epsilon}^{\overline{E}}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{H}_{\epsilon}\left[\Phi\left[\int_{\mathbb{T}}\Phi^{-1}w_{+}\,ds\right]\right]\right]. − Φ A ϵ E [ P 01 Φ − 1 H ϵ [ Φ [ ∫ T Φ − 1 w + d s ] ] ] .
We first bound R d , ϵ \mathcal{R}_{d,\epsilon} R d , ϵ . Using Lemmas 1.9, 1.10, and 1.13, we make note of the following estimates for the second term.
∥ Φ ‾ A ϵ E ‾ [ P 01 R s , ϵ [ ( w ϵ ) 0 Φ ] ] ∥ C 0 , α ≤ c ( κ ∗ ) ( ∥ R s , ϵ [ ( w ϵ ) 0 Φ ] ∥ C s 0 , α + ϵ ∣ R s , ϵ [ ( w ϵ ) 0 Φ ] ∣ C s 1 , α ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ w ϵ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) , \begin{aligned}
\left\|\overline{\Phi}\mathcal{A}_{\epsilon}^{\overline{E}}\left[\mathbb{P}_{01}\mathcal{R}_{s,\epsilon}\left[(w_{\epsilon})_{0}^{\Phi}\right]\right]\right\|_{C_{0,\alpha}}
\le c(\kappa_{*})\left(\left\|\mathcal{R}_{s,\epsilon}\left[(w_{\epsilon})_{0}^{\Phi}\right]\right\|_{C_{s}^{0,\alpha}}+\epsilon\left|\mathcal{R}_{s,\epsilon}\left[(w_{\epsilon})_{0}^{\Phi}\right]\right|_{C_{s}^{1,\alpha}}\right) \\
&\le c(\kappa_{*,\alpha_{+},c_{\Gamma}})\epsilon^{2-\alpha_{+}}\|w_{\epsilon}\|_{C_{0,\alpha}}
\le c(\kappa_{*,\alpha_{+},c_{\Gamma}})\epsilon^{2-\alpha_{+}}\|v\|_{C^{1,\alpha}(\mathbb{T})},
\end{aligned} Φ A ϵ E [ P 01 R s , ϵ [ ( w ϵ ) 0 Φ ] ] C 0 , α ≤ c ( κ ∗ ) ( R s , ϵ [ ( w ϵ ) 0 Φ ] C s 0 , α + ϵ R s , ϵ [ ( w ϵ ) 0 Φ ] C s 1 , α ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ w ϵ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) ,
∥ Φ ( a ) A ‾ ϵ E ‾ [ P 01 R s , ϵ ( a ) [ ( w ϵ ( a ) ) 0 Φ ( a ) ] ] − Φ ( b ) A ‾ ϵ E ‾ [ P 01 R s , ϵ ( b ) [ ( w ϵ ( b ) ) 0 Φ ( b ) ] ] ∥ C 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∥ X ( a ) − X ( b ) ∥ C 2 ( ∥ R s , ϵ ( a ) [ ( w ϵ ( a ) ) 0 Φ ( a ) ] ∥ C s 0 , α + ϵ ∣ R s , ϵ ( a ) [ ( w ϵ ( a ) ) 0 Φ ( a ) ] ∣ C s 1 , α ) \begin{aligned}
\left\|\Phi^{(a)}\overline{\mathcal{A}}_{\epsilon}^{\overline{E}}\left[\mathbb{P}_{01}\mathcal{R}_{s,\epsilon}^{(a)}\left[(w_{\epsilon}^{(a)})_{0}^{\Phi^{(a)}}\right]\right]
-\Phi^{(b)}\overline{\mathcal{A}}_{\epsilon}^{\overline{E}}\left[\mathbb{P}_{01}\mathcal{R}_{s,\epsilon}^{(b)}\left[(w_{\epsilon}^{(b)})_{0}^{\Phi^{(b)}}\right]\right]\right\|_{C_{0,\alpha}} \\
&\le c(\kappa_{*}^{(a)},\kappa_{*}^{(b)})\left(\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2}} \left(\|\mathcal{R}_{s,\epsilon}^{(a)}[(w_{\epsilon}^{(a)})_{0}^{\Phi^{(a)}}]\|_{C_{s}^{0,\alpha}}+\epsilon|\mathcal{R}_{s,\epsilon}^{(a)}[(w_{\epsilon}^{(a)})_{0}^{\Phi^{(a)}}]|_{C_{s}^{1,\alpha}}\right)
\right.
\end{aligned} Φ ( a ) A ϵ E [ P 01 R s , ϵ ( a ) [ ( w ϵ ( a ) ) 0 Φ ( a ) ] ] − Φ ( b ) A ϵ E [ P 01 R s , ϵ ( b ) [ ( w ϵ ( b ) ) 0 Φ ( b ) ] ] C 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∥ X ( a ) − X ( b ) ∥ C 2 ( ∥ R s , ϵ ( a ) [( w ϵ ( a ) ) 0 Φ ( a ) ] ∥ C s 0 , α + ϵ ∣ R s , ϵ ( a ) [( w ϵ ( a ) ) 0 Φ ( a ) ] ∣ C s 1 , α )
+ ∥ R s , ϵ ( a ) [ ( w ϵ ( a ) ) 0 Φ ( a ) ] − R s , ϵ ( b ) [ ( w ϵ ( b ) ) 0 Φ ( b ) ] ∥ C s 0 , α + ϵ ∣ R s , ϵ ( a ) [ ( w ϵ ( a ) ) 0 Φ ( a ) ] − R s , ϵ ( b ) [ ( w ϵ ( b ) ) 0 Φ ( b ) ] ∣ C s 1 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ( ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ w ϵ ( a ) ∥ C 0 , α + ∥ w ϵ ( a ) − w ϵ ( b ) ∥ C 0 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ v ∥ C 1 , α ( T ) . \begin{aligned}
\left.
+\|\mathcal{R}_{s,\epsilon}^{(a)}[(w_{\epsilon}^{(a)})_{0}^{\Phi^{(a)}}]-\mathcal{R}_{s,\epsilon}^{(b)}[(w_{\epsilon}^{(b)})_{0}^{\Phi^{(b)}}]\|_{C_{s}^{0,\alpha}}
+\epsilon|\mathcal{R}_{s,\epsilon}^{(a)}[(w_{\epsilon}^{(a)})_{0}^{\Phi^{(a)}}]-\mathcal{R}_{s,\epsilon}^{(b)}[(w_{\epsilon}^{(b)})_{0}^{\Phi^{(b)}}]|_{C_{s}^{1,\alpha}}\right) \\
&\le c(\kappa_{*}^{(a)},\kappa_{*}^{(b)},c_{\Gamma})\epsilon^{2-\alpha_{+}}\left(\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\alpha_{+}}}+\|w_{\epsilon}^{(a)}\|_{C_{0,\alpha}}+\|w_{\epsilon}^{(a)}-w_{\epsilon}^{(b)}\|_{C_{0,\alpha}}\right) \\
&\le c(\kappa_{*}^{(a)},\kappa_{*}^{(b)},c_{\Gamma})\epsilon^{2-\alpha_{+}}\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\alpha_{+}}}+\|v\|_{C^{1,\alpha}(\mathbb{T})}.
\end{aligned} + ∥ R s , ϵ ( a ) [( w ϵ ( a ) ) 0 Φ ( a ) ] − R s , ϵ ( b ) [( w ϵ ( b ) ) 0 Φ ( b ) ] ∥ C s 0 , α + ϵ ∣ R s , ϵ ( a ) [( w ϵ ( a ) ) 0 Φ ( a ) ] − R s , ϵ ( b ) [( w ϵ ( b ) ) 0 Φ ( b ) ] ∣ C s 1 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ( ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ w ϵ ( a ) ∥ C 0 , α + ∥ w ϵ ( a ) − w ϵ ( b ) ∥ C 0 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ v ∥ C 1 , α ( T ) .
For the third term of R d , ϵ \mathcal{R}_{d,\epsilon} R d , ϵ , we may use Lemmas 1.9, 1.12, and 1.13 to estimate
∥ Φ A ‾ ϵ E ‾ [ P 01 Φ − 1 H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] ] ∥ C 0 , α ≤ c ( κ ∗ ) ( ∥ H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] ∥ C s 0 , α + ϵ ∣ H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] ∣ C s 1 , α ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ w ϵ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) , \begin{aligned}
\left\|\Phi\overline{\mathcal{A}}_{\epsilon}^{\overline{E}}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{H}_{\epsilon}\left[\Phi\left[\int_{\mathbb{T}}\Phi^{-1}w_{\epsilon}\,ds\right]\right]\right]\right\|_{C^{0,\alpha}} \\
&\le c(\kappa_{*})\left(\left\|\mathcal{H}_{\epsilon}\left[\Phi\left[\int_{\mathbb{T}}\Phi^{-1}w_{\epsilon}\,ds\right]\right]\right\|_{C_{s}^{0,\alpha}}+\epsilon\left|\mathcal{H}_{\epsilon}\left[\Phi\left[\int_{\mathbb{T}}\Phi^{-1}w_{\epsilon}\,ds\right]\right]\right|_{C_{s}^{1,\alpha}}\right) \\
&\le c(\kappa_{*,\alpha_{+},c_{\Gamma}})\epsilon^{2-\alpha_{+}}\|w_{\epsilon}\|_{C^{0,\alpha}}
\le c(\kappa_{*,\alpha_{+},c_{\Gamma}})\epsilon^{2-\alpha_{+}}\|v\|_{C^{1,\alpha}(\mathbb{T})},
\end{aligned} Φ A ϵ E [ P 01 Φ − 1 H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] ] C 0 , α ≤ c ( κ ∗ ) ( H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] C s 0 , α + ϵ H ϵ [ Φ [ ∫ T Φ − 1 w ϵ d s ] ] C s 1 , α ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ w ϵ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) ,
∥ ( Φ ( a ) − Φ ( b ) ) A ‾ ϵ E ‾ [ P 01 ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] ] \left\|(\Phi^{(a)}-\Phi^{(b)})\overline{\mathcal{A}}_{\epsilon}^{\overline{E}}\left[\mathbb{P}_{01}(\Phi^{(a)})^{-1}\mathcal{H}_{\epsilon}^{(a)}\left[\Phi^{(a)}\left[\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_{\epsilon}^{(a)}\,ds\right]\right]\right]\right. ( Φ ( a ) − Φ ( b ) ) A ϵ E [ P 01 ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] ]
+ Φ ( b ) A ‾ ϵ E ‾ [ P 01 ( ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] − ( Φ ( b ) ) − 1 H ϵ ( b ) [ Φ ( b ) [ ∫ T ( Φ ( b ) ) − 1 w ϵ ( b ) d s ] ] ) ] ∥ C 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] ∥ C s 0 , α \begin{aligned}
\left.
+\Phi^{(b)}\overline{\mathcal{A}}_{\epsilon}^{\overline{E}}\left[\mathbb{P}_{01}\left((\Phi^{(a)})^{-1}\mathcal{H}_{\epsilon}^{(a)}\left[\Phi^{(a)}\left[\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_{\epsilon}^{(a)}\,ds\right]\right]-(\Phi^{(b)})^{-1}\mathcal{H}_{\epsilon}^{(b)}\left[\Phi^{(b)}\left[\int_{\mathbb{T}}(\Phi^{(b)})^{-1}w_{\epsilon}^{(b)}\,ds\right]\right]\right)\right]\right\|_{C^{0,\alpha}} \\
&\le c(\kappa_{*}^{(a)},\kappa_{*}^{(b)})\left(\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\alpha}}\left\|\mathcal{H}_{\epsilon}^{(a)}\left[\Phi^{(a)}\left[\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_{\epsilon}^{(a)}\,ds\right]\right]\right\|_{C_{s}^{0,\alpha}}
\right.
\end{aligned} + Φ ( b ) A ϵ E [ P 01 ( ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] − ( Φ ( b ) ) − 1 H ϵ ( b ) [ Φ ( b ) [ ∫ T ( Φ ( b ) ) − 1 w ϵ ( b ) d s ] ] ) ] C 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∥ X ( a ) − X ( b ) ∥ C 2 , α H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] C s 0 , α
+ ∥ X ( a ) − X ( b ) ∥ C 2 , α ϵ ∣ H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] ∣ C s 1 , α + ∥ ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] − ( Φ ( b ) ) − 1 H ϵ ( b ) [ Φ ( b ) [ ∫ T ( Φ ( b ) ) − 1 w ϵ ( b ) d s ] ] ∥ C s 0 , α \begin{aligned}
\left.
+\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\alpha}}\epsilon\left|\mathcal{H}_{\epsilon}^{(a)}\left[\Phi^{(a)}\left[\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_{\epsilon}^{(a)}\,ds\right]\right]\right|_{C_{s}^{1,\alpha}}
\right. \\
&+\|(\Phi^{(a)})^{-1}\mathcal{H}_{\epsilon}^{(a)}[\Phi^{(a)}[\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_{\epsilon}^{(a)}\,ds]]-(\Phi^{(b)})^{-1}\mathcal{H}_{\epsilon}^{(b)}[\Phi^{(b)}[\int_{\mathbb{T}}(\Phi^{(b)})^{-1}w_{\epsilon}^{(b)}\,ds]]\|_{C_{s}^{0,\alpha}}
\end{aligned} + ∥ X ( a ) − X ( b ) ∥ C 2 , α ϵ H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] C s 1 , α + ∥ ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ]] − ( Φ ( b ) ) − 1 H ϵ ( b ) [ Φ ( b ) [ ∫ T ( Φ ( b ) ) − 1 w ϵ ( b ) d s ]] ∥ C s 0 , α
+ ϵ ∣ ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] − ( Φ ( b ) ) − 1 H ϵ ( b ) [ Φ ( b ) [ ∫ T ( Φ ( b ) ) − 1 w ϵ ( b ) d s ] ] ∣ C s 1 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ( ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ w ϵ ( a ) ∥ C 0 , α + ∥ w ϵ ( a ) − w ϵ ( b ) ∥ C 0 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ( ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ v ∥ C 1 , α ( T ) ) . \begin{aligned}
\left.
+\epsilon\left|(\Phi^{(a)})^{-1}\mathcal{H}_{\epsilon}^{(a)}\left[\Phi^{(a)}\left[\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_{\epsilon}^{(a)}\,ds\right]\right]-(\Phi^{(b)})^{-1}\mathcal{H}_{\epsilon}^{(b)}\left[\Phi^{(b)}\left[\int_{\mathbb{T}}(\Phi^{(b)})^{-1}w_{\epsilon}^{(b)}\,ds\right]\right]\right|_{C_{s}^{1,\alpha}}\right) \\
&\le c(\kappa_{*}^{(a)},\kappa_{*}^{(b)},c_{\Gamma})\epsilon^{2-\alpha_{+}}\left(\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\alpha_{+}}}+\|w_{\epsilon}^{(a)}\|_{C^{0,\alpha}}+\|w_{\epsilon}^{(a)}-w_{\epsilon}^{(b)}\|_{C^{0,\alpha}}\right) \\
&\le c(\kappa_{*}^{(a)},\kappa_{*}^{(b)},c_{\Gamma})\epsilon^{2-\alpha_{+}}\left(\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\alpha_{+}}}+\|v\|_{C^{1,\alpha}(\mathbb{T})}\right).
\end{aligned} + ϵ ( Φ ( a ) ) − 1 H ϵ ( a ) [ Φ ( a ) [ ∫ T ( Φ ( a ) ) − 1 w ϵ ( a ) d s ] ] − ( Φ ( b ) ) − 1 H ϵ ( b ) [ Φ ( b ) [ ∫ T ( Φ ( b ) ) − 1 w ϵ ( b ) d s ] ] C s 1 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ( ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ w ϵ ( a ) ∥ C 0 , α + ∥ w ϵ ( a ) − w ϵ ( b ) ∥ C 0 , α ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ 2 − α + ( ∥ X ( a ) − X ( b ) ∥ C 2 , α + + ∥ v ∥ C 1 , α ( T ) ) .
Altogether, we may bound R d , ϵ \mathcal{R}_{d,\epsilon} R d , ϵ as
∥ R d , ϵ [ v ] ∥ C 0 , α ( T ) ≤ c ( κ ∗ , α ) ϵ 2 − α ∥ w ϵ ∥ C 0 , α + c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) , \begin{aligned}
\|\mathcal{R}_{d,\epsilon}[v]\|_{C^{0,\alpha}(\mathbb{T})}
\le c(\kappa_{*,\alpha})\epsilon^{2-\alpha}\|w_{\epsilon}\|_{C^{0,\alpha}}+c(\kappa_{*,\alpha_{+},c_{\Gamma}})\epsilon^{2-\alpha_{+}}\|v\|_{C^{1,\alpha}(\mathbb{T})} \\
&\le c(\kappa_{*,\alpha_{+},c_{\Gamma}})\epsilon^{2-\alpha_{+}}\|v\|_{C^{1,\alpha}(\mathbb{T})},
\end{aligned} ∥ R d , ϵ [ v ] ∥ C 0 , α ( T ) ≤ c ( κ ∗ , α ) ϵ 2 − α ∥ w ϵ ∥ C 0 , α + c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 2 − α + ∥ v ∥ C 1 , α ( T ) ,
∥ ( R d , ϵ ( a ) − R d , ϵ ( b ) ) [ v ] ∥ C 0 , α ( T ) ≤ ϵ 2 ∥ ∫ 0 2 π ( w ϵ ( a ) κ ^ ( a ) − w ϵ ( b ) κ ^ ( b ) ) d θ ∥ C 0 , α + c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 2 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ v ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 2 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ v ∥ C 1 , α ( T ) . \left\|\left(\mathcal{R}_{d,\epsilon}^{(a)}-\mathcal{R}_{d,\epsilon}^{(b)}\right)[v]\right\|_{C^{0,\alpha}(\mathbb{T})}
\leq\epsilon^2\left\|\int_{0}^{2\pi}\left(w_{\epsilon}^{(a)}\hat{\kappa}^{(a)}-w_{\epsilon}^{(b)}\hat{\kappa}^{(b)}\right)\,d\theta\right\|_{C^{0,\alpha}}
+c\left(\kappa_{*,\alpha+}^{(a)},\kappa_{*,\alpha+}^{(b)},c_{\Gamma}\right)\epsilon^{2-\alpha+}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha+}}\|v\|_{C^{1,\alpha}(\mathbb{T})}
\leq c\left(\kappa_{*,\alpha+}^{(a)},\kappa_{*,\alpha+}^{(b)},c_{\Gamma}\right)\epsilon^{2-\alpha+}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha+}}\|v\|_{C^{1,\alpha}(\mathbb{T})}. ( R d , ϵ ( a ) − R d , ϵ ( b ) ) [ v ] C 0 , α ( T ) ≤ ϵ 2 ∫ 0 2 π ( w ϵ ( a ) κ ^ ( a ) − w ϵ ( b ) κ ^ ( b ) ) d θ C 0 , α + c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 2 − α + X ( a ) − X ( b ) C 2 , α + ∥ v ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 2 − α + X ( a ) − X ( b ) C 2 , α + ∥ v ∥ C 1 , α ( T ) .
We next consider R d , + \mathcal{R}_{d,+} R d , + . For 0 < α < γ < β < 1 0<\alpha<\gamma<\beta<1 0 < α < γ < β < 1 , we may use Lemmas 1.9, 1.11, and 1.13 to estimate the third term involving the double layer remainder as
∥ Φ A ‾ ϵ [ P 01 R D [ v ] ] ∥ C 0 , γ ≤ c ( κ ∗ ) ϵ − 1 ∥ R D [ v ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ∥ v ∥ C 0 , γ ( T ) , \left\|\Phi\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{D}[v]\right]\right\|_{C^{0,\gamma}}
\leq c(\kappa_{*})\epsilon^{-1}\|\mathcal{R}_{D}[v]\|_{C^{1,\gamma}}
\leq c(\kappa_{*,\gamma+},c_{\Gamma})\epsilon^{-1-\gamma}\|v\|_{C^{0,\gamma}(\mathbb{T})}, Φ A ϵ [ P 01 R D [ v ] ] C 0 , γ ≤ c ( κ ∗ ) ϵ − 1 ∥ R D [ v ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ∥ v ∥ C 0 , γ ( T ) ,
∥ Φ ( a ) A ‾ ϵ [ P 01 R D ( a ) [ v ] ] − Φ ( b ) A ‾ ϵ [ P 01 R D ( b ) [ v ] ] ∥ C 0 , γ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ − 1 ( ∥ X ( a ) − X ( b ) ∥ C 2 ∥ R D ( a ) [ v ] ∥ C s 1 , γ + ∥ ( R D ( a ) − R D ( b ) ) [ v ] ∥ C s 1 , γ ) ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ ( T ) . \left\|\Phi^{(a)}\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{D}^{(a)}[v]\right]-\Phi^{(b)}\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\mathcal{R}_{D}^{(b)}[v]\right]\right\|_{C^{0,\gamma}}
\leq c\left(\kappa_{*}^{(a)},\kappa_{*}^{(b)}\right)\epsilon^{-1}\left(\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2}}\|\mathcal{R}_{D}^{(a)}[v]\|_{C_{s}^{1,\gamma}}+\|(\mathcal{R}_{D}^{(a)}-\mathcal{R}_{D}^{(b)})[v]\|_{C_{s}^{1,\gamma}}\right)
\leq c\left(\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma}\right)\epsilon^{-1-\gamma+}\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma+}}\|v\|_{C^{0,\gamma}(\mathbb{T})}. Φ ( a ) A ϵ [ P 01 R D ( a ) [ v ] ] − Φ ( b ) A ϵ [ P 01 R D ( b ) [ v ] ] C 0 , γ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ − 1 ( ∥ X ( a ) − X ( b ) ∥ C 2 ∥ R D ( a ) [ v ] ∥ C s 1 , γ + ∥ ( R D ( a ) − R D ( b ) ) [ v ] ∥ C s 1 , γ ) ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ ( T ) .
Letting Φ M 1 = Φ ( A ‾ ϵ + [ P 01 R S [ w 0 Φ ] ] + A ‾ ϵ + [ P 01 R S , + [ w 0 Φ ] ] + A ‾ ϵ [ P 01 R S , ϵ [ ( w + ) 0 Φ ] ] ) \Phi\mathcal{M}_{1}=\Phi\left(\overline{\mathcal{A}}_{\epsilon}^{+}[\mathbb{P}_{01}\mathcal{R}_{S}[w_{0}^{\Phi}]]+\overline{\mathcal{A}}_{\epsilon}^{+}[\mathbb{P}_{01}\mathcal{R}_{S,+}[w_{0}^{\Phi}]]+\overline{\mathcal{A}}_{\epsilon}[\mathbb{P}_{01}\mathcal{R}_{S,\epsilon}[(w_{+})_{0}^{\Phi}]]\right) Φ M 1 = Φ ( A ϵ + [ P 01 R S [ w 0 Φ ]] + A ϵ + [ P 01 R S , + [ w 0 Φ ]] + A ϵ [ P 01 R S , ϵ [( w + ) 0 Φ ]] ) denote the middle three terms of R d , + \mathcal{R}_{d,+} R d , + , we may use Lemmas 1.9, 1.10, and 1.13 to estimate
∥ Φ M 1 ∥ C 0 , γ ≤ c ( κ ∗ ) ( ϵ − 1 ∥ R S [ w 0 Φ ] ∥ C s 1 , α + ∥ R S , + [ w 0 Φ ] ∥ C s 1 , γ + ∥ R S , ϵ [ ( w + ) 0 Φ ] ∥ C s 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ( ∥ w ∥ C 0 , α + ∥ w + ∥ C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) , \|\Phi\mathcal{M}_{1}\|_{C^{0,\gamma}}
\leq c(\kappa_{*})\left(\epsilon^{-1}\|\mathcal{R}_{S}[w_{0}^{\Phi}]\|_{C_{s}^{1,\alpha}}+\|\mathcal{R}_{S,+}[w_{0}^{\Phi}]\|_{C_{s}^{1,\gamma}}+\|\mathcal{R}_{S,\epsilon}[(w_{+})_{0}^{\Phi}]\|_{C_{s}^{1,\gamma}}\right)
\leq c(\epsilon,\kappa_{*,\gamma+},c_{\Gamma})\left(\|w\|_{C^{0,\alpha}}+\|w_{+}\|_{C^{0,\gamma}}\right)
\leq c(\epsilon,\kappa_{*,\gamma+},c_{\Gamma})\|v\|_{C^{1,\alpha}(\mathbb{T})}, ∥Φ M 1 ∥ C 0 , γ ≤ c ( κ ∗ ) ( ϵ − 1 ∥ R S [ w 0 Φ ] ∥ C s 1 , α + ∥ R S , + [ w 0 Φ ] ∥ C s 1 , γ + ∥ R S , ϵ [( w + ) 0 Φ ] ∥ C s 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ( ∥ w ∥ C 0 , α + ∥ w + ∥ C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) ,
∥ Φ ( a ) M 1 ( a ) − Φ ( b ) M 1 ( b ) ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ( ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ( ∥ w 0 ( a ) ∥ C 0 , α + ∥ w + ( a ) ∥ C 0 , γ ) + ∥ w 0 ( a ) − w 0 ( b ) ∥ C 0 , α + ∥ w + ( a ) − w + ( b ) ∥ C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 1 , α ( T ) . \left\|\Phi^{(a)}\mathcal{M}_{1}^{(a)}-\Phi^{(b)}\mathcal{M}_{1}^{(b)}\right\|_{C^{0,\gamma}}
\leq c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\left(\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma+}}\left(\|w_{0}^{(a)}\|_{C^{0,\alpha}}+\|w_{+}^{(a)}\|_{C^{0,\gamma}}\right)+\|w_{0}^{(a)}-w_{0}^{(b)}\|_{C^{0,\alpha}}+\|w_{+}^{(a)}-w_{+}^{(b)}\|_{C^{0,\gamma}}\right)
\leq c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma+}}\|v\|_{C^{1,\alpha}(\mathbb{T})}. Φ ( a ) M 1 ( a ) − Φ ( b ) M 1 ( b ) C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ( ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ( ∥ w 0 ( a ) ∥ C 0 , α + ∥ w + ( a ) ∥ C 0 , γ ) + ∥ w 0 ( a ) − w 0 ( b ) ∥ C 0 , α + ∥ w + ( a ) − w + ( b ) ∥ C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 1 , α ( T ) .
Finally, letting Φ M 2 = Φ ( A ‾ ϵ + [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w d s ] ] + A ‾ ϵ [ P 01 Φ − 1 H + [ Φ ∫ T Φ − 1 w d s ] ] + A ‾ ϵ [ P 01 Φ − 1 H ϵ [ Φ ∫ T Φ − 1 w + d s ] ] ) \Phi\mathcal{M}_{2}=\Phi\left(\overline{\mathcal{A}}_{\epsilon}^{+}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w\,ds\right]\right]+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{H}_{+}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w\,ds\right]\right]+\overline{\mathcal{A}}_{\epsilon}\left[\mathbb{P}_{01}\Phi^{-1}\mathcal{H}_{\epsilon}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w_{+}\,ds\right]\right]\right) Φ M 2 = Φ ( A ϵ + [ P 01 Φ − 1 S [ Φ ∫ T Φ − 1 w d s ] ] + A ϵ [ P 01 Φ − 1 H + [ Φ ∫ T Φ − 1 w d s ] ] + A ϵ [ P 01 Φ − 1 H ϵ [ Φ ∫ T Φ − 1 w + d s ] ] ) denote the final three terms of R d , + \mathcal{R}_{d,+} R d , + , we may use Lemmas 1.9, 1.12, and 1.13 to obtain
∥ Φ M 2 ∥ C 0 , γ ≤ c ( κ ∗ ) ( ∥ S [ Φ ∫ T Φ − 1 w d s ] ∥ C s 1 , α + ∥ H + [ Φ ∫ T Φ − 1 w d s ] ∥ C s 1 , γ + ∥ H ϵ [ Φ ∫ T Φ − 1 w + d s ] ∥ C s 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ( ∥ w ∥ C 0 , α + ∥ w + ∥ C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) , \|\Phi\mathcal{M}_{2}\|_{C^{0,\gamma}}
\leq c(\kappa_{*})\left(\left\|\mathcal{S}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w\,ds\right]\right\|_{C_{s}^{1,\alpha}}+\left\|\mathcal{H}_{+}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w\,ds\right]\right\|_{C_{s}^{1,\gamma}}+\left\|\mathcal{H}_{\epsilon}\left[\Phi\int_{\mathbb{T}}\Phi^{-1}w_{+}\,ds\right]\right\|_{C_{s}^{1,\gamma}}\right)
\leq c(\epsilon,\kappa_{*,\gamma+},c_{\Gamma})\left(\|w\|_{C^{0,\alpha}}+\|w_{+}\|_{C^{0,\gamma}}\right)
\leq c(\epsilon,\kappa_{*,\gamma+},c_{\Gamma})\|v\|_{C^{1,\alpha}(\mathbb{T})}, ∥Φ M 2 ∥ C 0 , γ ≤ c ( κ ∗ ) ( S [ Φ ∫ T Φ − 1 w d s ] C s 1 , α + H + [ Φ ∫ T Φ − 1 w d s ] C s 1 , γ + H ϵ [ Φ ∫ T Φ − 1 w + d s ] C s 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ( ∥ w ∥ C 0 , α + ∥ w + ∥ C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) ,
∥ Φ ( a ) M 2 ( a ) − Φ ( b ) M 2 ( b ) ∥ C 0 , γ \left\|\Phi^{(a)}\mathcal{M}_{2}^{(a)}-\Phi^{(b)}\mathcal{M}_{2}^{(b)}\right\|_{C^{0,\gamma}} Φ ( a ) M 2 ( a ) − Φ ( b ) M 2 ( b ) C 0 , γ
≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∥ X ( a ) − X ( b ) ∥ C 2 ( ∥ S ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] ∥ C s 1 , α + ∥ H + ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] ∥ C s 1 , γ + ∥ H ϵ ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w + ( a ) d s ] ∥ C s 1 , γ ) + ∥ S ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] − S ( b ) [ Φ ( b ) ∫ T ( Φ ( b ) ) − 1 w ( b ) d s ] ∥ C s 1 , α + ∥ H + ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] − H + ( b ) [ Φ ( b ) ∫ T ( Φ ( b ) ) − 1 w ( b ) d s ] ∥ C s 1 , γ + ∥ H ϵ ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w + ( a ) d s ] − H ϵ ( b ) [ Φ ( b ) ∫ T ( Φ ( b ) ) − 1 w + ( b ) d s ] ∥ C s 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ( ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ( ∥ w ( a ) ∥ C 0 , α + ∥ w + ( a ) ∥ C 0 , γ ) + ∥ w ( a ) − w ( b ) ∥ C 0 , α + ∥ w + ( a ) − w + ( b ) ∥ C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 1 , α ( T ) . \begin{aligned}
&\le c(\kappa_*^{(a)},\kappa_*^{(b)})\Bigg(
\left\|X^{(a)}-X^{(b)}\right\|_{C^2}
\Bigg(
\left\|\mathcal{S}^{(a)}\left[\Phi^{(a)}\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w^{(a)}\,ds\right]\right\|_{C_s^{1,\alpha}}\\
&\qquad+\left\|\mathcal{H}_+^{(a)}\left[\Phi^{(a)}\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w^{(a)}\,ds\right]\right\|_{C_s^{1,\gamma}}
+\left\|\mathcal{H}_\epsilon^{(a)}\left[\Phi^{(a)}\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_+^{(a)}\,ds\right]\right\|_{C_s^{1,\gamma}}
\Bigg)\\
&\qquad+\left\|\mathcal{S}^{(a)}\left[\Phi^{(a)}\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w^{(a)}\,ds\right]
-\mathcal{S}^{(b)}\left[\Phi^{(b)}\int_{\mathbb{T}}(\Phi^{(b)})^{-1}w^{(b)}\,ds\right]\right\|_{C_s^{1,\alpha}}\\
&\qquad+\left\|\mathcal{H}_+^{(a)}\left[\Phi^{(a)}\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w^{(a)}\,ds\right]
-\mathcal{H}_+^{(b)}\left[\Phi^{(b)}\int_{\mathbb{T}}(\Phi^{(b)})^{-1}w^{(b)}\,ds\right]\right\|_{C_s^{1,\gamma}}\\
&\qquad+\left\|\mathcal{H}_\epsilon^{(a)}\left[\Phi^{(a)}\int_{\mathbb{T}}(\Phi^{(a)})^{-1}w_+^{(a)}\,ds\right]
-\mathcal{H}_\epsilon^{(b)}\left[\Phi^{(b)}\int_{\mathbb{T}}(\Phi^{(b)})^{-1}w_+^{(b)}\,ds\right]\right\|_{C_s^{1,\gamma}}
\Bigg)\\
&\le c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_\Gamma)
\Bigg(
\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma+}}
\left(\left\|w^{(a)}\right\|_{C^{0,\alpha}}+\left\|w_+^{(a)}\right\|_{C^{0,\gamma}}\right)\\
&\qquad+\left\|w^{(a)}-w^{(b)}\right\|_{C^{0,\alpha}}
+\left\|w_+^{(a)}-w_+^{(b)}\right\|_{C^{0,\gamma}}
\Bigg)\\
&\le c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_\Gamma)
\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma+}}\left\|v\right\|_{C^{1,\alpha}(\mathbb{T})}.
\end{aligned} ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( X ( a ) − X ( b ) C 2 ( S ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] C s 1 , α + H + ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] C s 1 , γ + H ϵ ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w + ( a ) d s ] C s 1 , γ ) + S ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] − S ( b ) [ Φ ( b ) ∫ T ( Φ ( b ) ) − 1 w ( b ) d s ] C s 1 , α + H + ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w ( a ) d s ] − H + ( b ) [ Φ ( b ) ∫ T ( Φ ( b ) ) − 1 w ( b ) d s ] C s 1 , γ + H ϵ ( a ) [ Φ ( a ) ∫ T ( Φ ( a ) ) − 1 w + ( a ) d s ] − H ϵ ( b ) [ Φ ( b ) ∫ T ( Φ ( b ) ) − 1 w + ( b ) d s ] C s 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ( X ( a ) − X ( b ) C 2 , γ + ( w ( a ) C 0 , α + w + ( a ) C 0 , γ ) + w ( a ) − w ( b ) C 0 , α + w + ( a ) − w + ( b ) C 0 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) X ( a ) − X ( b ) C 2 , γ + ∥ v ∥ C 1 , α ( T ) .
Altogether, we may estimate R d , + \mathcal{R}_{d,+} R d , + as
∥ R d , + [ v ] ∥ C 0 , γ ( T ) ≤ c ( κ ∗ , γ ) ϵ 2 − γ ∥ w + ∥ C 0 , γ + ϵ ∥ w ∥ L ∞ + c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) , \begin{aligned}
\left\|\mathcal{R}_{d,+}[v]\right\|_{C^{0,\gamma}(\mathbb{T})}
\le c(\kappa_*,\gamma)\epsilon^{2-\gamma}\left\|w_+\right\|_{C^{0,\gamma}}+\epsilon\left\|w\right\|_{L^\infty}
+c(\epsilon,\kappa_{*,\gamma+},c_\Gamma)\left\|v\right\|_{C^{1,\alpha}(\mathbb{T})} \\
&\le c(\epsilon,\kappa_{*,\gamma+},c_\Gamma)\left\|v\right\|_{C^{1,\alpha}(\mathbb{T})},
\end{aligned} ∥ R d , + [ v ] ∥ C 0 , γ ( T ) ≤ c ( κ ∗ , γ ) ϵ 2 − γ ∥ w + ∥ C 0 , γ + ϵ ∥ w ∥ L ∞ + c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 1 , α ( T ) ,
∥ ( R d , + ( a ) − R d , + ( b ) ) [ v ] ∥ C 0 , γ ( T ) ≤ ϵ 2 ∥ w + ( a ) ( κ ^ ( a ) − κ ^ ( b ) ) ∥ C 0 , γ + ϵ 2 ∥ κ ^ ( b ) ( w + ( a ) − w + ( b ) ) ∥ C 0 , γ + ∥ ( Φ ( a ) ∫ T ( Φ ( a ) ) − 1 − Φ ( b ) ∫ T ( Φ ( b ) ) − 1 ) ∫ 0 2 π w ( a ) ϵ d θ d s ∥ C 0 , γ + c ( κ ∗ , γ ( b ) ) ϵ ∥ w ( a ) − w ( b ) ∥ L ∞ + c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 1 , α ( T ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 1 , α ( T ) . (84) \begin{aligned}
\left\|\left(\mathcal{R}_{d,+}^{(a)}-\mathcal{R}_{d,+}^{(b)}\right)[v]\right\|_{C^{0,\gamma}(\mathbb{T})}
&\le\epsilon^2\left\|w_+^{(a)}(\widehat{\kappa}^{(a)}-\widehat{\kappa}^{(b)})\right\|_{C^{0,\gamma}}
+\epsilon^2\left\|\widehat{\kappa}^{(b)}(w_+^{(a)}-w_+^{(b)})\right\|_{C^{0,\gamma}}\\
&\quad+\left\|\left(\Phi^{(a)}\int_{\mathbb{T}}(\Phi^{(a)})^{-1}
-\Phi^{(b)}\int_{\mathbb{T}}(\Phi^{(b)})^{-1}\right)
\int_0^{2\pi}w^{(a)}\epsilon\,d\theta ds\right\|_{C^{0,\gamma}}\\
&\quad+c(\kappa_{*,\gamma}^{(b)})\epsilon\left\|w^{(a)}-w^{(b)}\right\|_{L^\infty}\\
&\quad+c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_\Gamma)
\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma+}}\left\|v\right\|_{C^{1,\alpha}(\mathbb{T})}\\
&\le c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_\Gamma)
\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma+}}\left\|v\right\|_{C^{1,\alpha}(\mathbb{T})}. \tag*{(84)}
\end{aligned} ( R d , + ( a ) − R d , + ( b ) ) [ v ] C 0 , γ ( T ) ≤ ϵ 2 w + ( a ) ( κ ( a ) − κ ( b ) ) C 0 , γ + ϵ 2 κ ( b ) ( w + ( a ) − w + ( b ) ) C 0 , γ + ( Φ ( a ) ∫ T ( Φ ( a ) ) − 1 − Φ ( b ) ∫ T ( Φ ( b ) ) − 1 ) ∫ 0 2 π w ( a ) ϵ d θ d s C 0 , γ + c ( κ ∗ , γ ( b ) ) ϵ w ( a ) − w ( b ) L ∞ + c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) X ( a ) − X ( b ) C 2 , γ + ∥ v ∥ C 1 , α ( T ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) X ( a ) − X ( b ) C 2 , γ + ∥ v ∥ C 1 , α ( T ) . (84)
In total we obtain the estimates (30) and (31) for the slender body DtN map decomposition (29).
We next turn to the decomposition (32) of the slender body NtD operator L ϵ \mathcal{L}_\epsilon L ϵ . We will rely on the DtN decomposition (29) to prove the subsequent bounds (33) and (34). In particular, given X ( s ) X(s) X ( s ) and corresponding map Φ \Phi Φ , for f ( s ) ∈ C 0 , α ( T ) f(s)\in C^{0,\alpha}(\mathbb{T}) f ( s ) ∈ C 0 , α ( T ) and f 0 Φ ( s ) = f ( s ) − Φ ∫ T Φ − 1 f ( s ) d s f_0^\Phi(s)=f(s)-\Phi\int_{\mathbb{T}}\Phi^{-1}f(s)\,ds f 0 Φ ( s ) = f ( s ) − Φ ∫ T Φ − 1 f ( s ) d s , we may use (29) to write
Φ − 1 v ( s ) = ( I + L ‾ ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ‾ ϵ [ Φ − 1 f 0 Φ ( s ) ] + ( I + L ‾ ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ‾ ϵ [ ( R d , + [ v ( s ) ] ) 0 Φ ] = ( I + L ‾ ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ‾ ϵ [ Φ − 1 f 0 Φ ( s ) ] + ( I + L ‾ ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ‾ ϵ [ ( R d , + [ L ϵ [ f ( s ) ] ] ) 0 Φ ] . \begin{aligned}
\Phi^{-1}v(s)=\left(I+\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)\Phi^{-1}\right)^{-1}\overline{\mathcal{L}}_\epsilon[\Phi^{-1}f_0^\Phi(s)]
+\left(I+\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)\Phi^{-1}\right)^{-1}\overline{\mathcal{L}}_\epsilon\left[(\mathcal{R}_{d,+}[v(s)])_0^\Phi\right] \\
&=\left(I+\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)\Phi^{-1}\right)^{-1}\overline{\mathcal{L}}_\epsilon[\Phi^{-1}f_0^\Phi(s)]
+\left(I+\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)\Phi^{-1}\right)^{-1}\overline{\mathcal{L}}_\epsilon\left[(\mathcal{R}_{d,+}[\mathcal{L}_\epsilon[f(s)]])_0^\Phi\right].
\end{aligned} Φ − 1 v ( s ) = ( I + L ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ϵ [ Φ − 1 f 0 Φ ( s )] + ( I + L ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ϵ [ ( R d , + [ v ( s )] ) 0 Φ ] = ( I + L ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ϵ [ Φ − 1 f 0 Φ ( s )] + ( I + L ϵ ( R d , ϵ Φ ) Φ − 1 ) − 1 L ϵ [ ( R d , + [ L ϵ [ f ( s )]] ) 0 Φ ] .
Now, using Lemma 1.4 and the estimate (30) for R d , ϵ \mathcal{R}_{d,\epsilon} R d , ϵ , for any g ∈ C 1 , α ( T ) g\in C^{1,\alpha}(\mathbb{T}) g ∈ C 1 , α ( T ) we have
∥ L ‾ ϵ ( R d , ϵ [ Φ g ] ) Φ ∥ C 1 , α ( T ) ≤ c ∣ log ϵ ∣ ( ∥ R d , ϵ [ Φ g ] ∥ L ∞ ( T ) + ϵ − 1 ∥ R d , ϵ [ Φ g ] ∥ C ˙ 0 , α ( T ) ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ Φ g ∥ C 1 , α ( T ) . \begin{aligned}
\left\|\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}[\Phi g])^\Phi\right\|_{C^{1,\alpha}(\mathbb{T})}
\le c\lvert\log\epsilon\rvert\left(\left\|\mathcal{R}_{d,\epsilon}[\Phi g]\right\|_{L^\infty(\mathbb{T})}
+\epsilon^{-1}\left\|\mathcal{R}_{d,\epsilon}[\Phi g]\right\|_{\dot C^{0,\alpha}(\mathbb{T})}\right) \\
&\le c(\kappa_{*,\alpha+},c_\Gamma)\epsilon^{1-\alpha+}\lvert\log\epsilon\rvert\left\|\Phi g\right\|_{C^{1,\alpha}(\mathbb{T})}.
\end{aligned} L ϵ ( R d , ϵ [ Φ g ] ) Φ C 1 , α ( T ) ≤ c ∣ log ϵ ∣ ( ∥ R d , ϵ [ Φ g ] ∥ L ∞ ( T ) + ϵ − 1 ∥ R d , ϵ [ Φ g ] ∥ C ˙ 0 , α ( T ) ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ Φ g ∥ C 1 , α ( T ) .
Thus for ϵ \epsilon ϵ sufficiently small, we may expand ( I + L ‾ ϵ ( R d , ϵ Φ ) 0 Φ ) − 1 (I+\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)_0^\Phi)^{-1} ( I + L ϵ ( R d , ϵ Φ ) 0 Φ ) − 1 as a Neumann series:
( I + L ‾ ϵ ( R d , ϵ Φ ) 0 Φ ) − 1 = I + ∑ j = 1 ∞ ( L ‾ ϵ ( R d , ϵ Φ ) 0 Φ ) j = : I + R n , ϵ , (85) (I+\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)_0^\Phi)^{-1}=I+\sum_{j=1}^{\infty}\left(\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)_0^\Phi\right)^j=:I+\mathcal{R}_{n,\epsilon}, \tag*{(85)} ( I + L ϵ ( R d , ϵ Φ ) 0 Φ ) − 1 = I + j = 1 ∑ ∞ ( L ϵ ( R d , ϵ Φ ) 0 Φ ) j =: I + R n , ϵ , (85)
with
∥ R n , ϵ [ g ] ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ g ∥ C 1 , α ( T ) . (86) \|\mathcal{R}_{n,\epsilon}[g]\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\kappa_{*},\alpha_{+},c_{\Gamma})\epsilon^{1-\alpha_{+}}|\log\epsilon|\|g\|_{C^{1,\alpha}(\mathbb{T})}. \tag*{(86)} ∥ R n , ϵ [ g ] ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∣ log ϵ ∣∥ g ∥ C 1 , α ( T ) . (86)
We thus obtain the first part of the decomposition (32) . Furthermore, using the Lipschitz estimate (31) for R d , ϵ \mathcal{R}_{d,\epsilon} R d , ϵ , we also have
∥ ( R n , ϵ ( a ) − R n , ϵ ( b ) ) [ g ] ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∣ log ϵ ∣ ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ g ∥ C 1 , α ( T ) . (87) \|(\mathcal{R}_{n,\epsilon}^{(a)}-\mathcal{R}_{n,\epsilon}^{(b)})[g]\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\kappa_{*,\alpha_{+}}^{(a)},\kappa_{*,\alpha_{+}}^{(b)},c_{\Gamma})\epsilon^{1-\alpha_{+}}|\log\epsilon|\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha_{+}}(\mathbb{T})}\|g\|_{C^{1,\alpha}(\mathbb{T})}. \tag*{(87)} ∥ ( R n , ϵ ( a ) − R n , ϵ ( b ) ) [ g ] ∥ C 1 , α ( T ) ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∣ log ϵ ∣∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ g ∥ C 1 , α ( T ) . (87)
In addition, defining
R n , + : = ( I + L ‾ ϵ ( R d , ϵ Φ ) 0 Φ ) − 1 L ‾ ϵ [ ( R d , + [ L ϵ [ ⋅ ] ] ) 0 Φ ] = ( I + R n 0 , ϵ ) L ‾ ϵ [ ( R d , + [ L ϵ [ ⋅ ] ] ) 0 Φ ] , \begin{aligned}
\mathcal{R}_{n,+}:=(I+\overline{\mathcal{L}}_\epsilon(\mathcal{R}_{d,\epsilon}\Phi)_0^\Phi)^{-1}\overline{\mathcal{L}}_\epsilon[(\mathcal{R}_{d,+}[\mathcal{L}_\epsilon[\cdot]])_0^\Phi] \\
&=(I+\mathcal{R}_{n0,\epsilon})\overline{\mathcal{L}}_\epsilon[(\mathcal{R}_{d,+}[\mathcal{L}_\epsilon[\cdot]])_0^\Phi],
\end{aligned} R n , + := ( I + L ϵ ( R d , ϵ Φ ) 0 Φ ) − 1 L ϵ [( R d , + [ L ϵ [ ⋅ ]] ) 0 Φ ] = ( I + R n 0 , ϵ ) L ϵ [( R d , + [ L ϵ [ ⋅ ]] ) 0 Φ ] ,
we may use (86) , Lemma 1.4 , (83), and Lemma 1.14 to estimate
∥ R n , + [ f ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ ( R d , + [ L ϵ [ f ] ] ) 0 Φ ∥ C 1 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ R d , + [ L ϵ [ f ] ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ L ϵ [ f ] ∥ C 1 , α ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) ∥ f ∥ C 0 , α ( T ) . \begin{aligned}
\|\mathcal{R}_{n,+}[f]\|_{C^{1,\gamma}(\mathbb{T})}\leq c(\epsilon,\kappa_{*,\gamma_{+}},c_{\Gamma})\|(\mathcal{R}_{d,+}[\mathcal{L}_\epsilon[f]])_0^\Phi\|_{C^{1,\gamma}}\leq c(\epsilon,\kappa_{*,\gamma_{+}},c_{\Gamma})\|\mathcal{R}_{d,+}[\mathcal{L}_\epsilon[f]]\|_{C^{0,\gamma}} \\
&\leq c(\epsilon,\kappa_{*,\gamma_{+}},c_{\Gamma})\|\mathcal{L}_\epsilon[f]\|_{C^{1,\alpha}}\leq c(\epsilon,\|\boldsymbol{X}_{ss}\|_{C^{1,\alpha}},c_{\Gamma})\|f\|_{C^{0,\alpha}(\mathbb{T})}.
\end{aligned} ∥ R n , + [ f ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ ( R d , + [ L ϵ [ f ]] ) 0 Φ ∥ C 1 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ R d , + [ L ϵ [ f ]] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ L ϵ [ f ] ∥ C 1 , α ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) ∥ f ∥ C 0 , α ( T ) .
Using the Lipschitz estimates (87) , (31) , and (76) for R n , ϵ \mathcal{R}_{n,\epsilon} R n , ϵ , R d , + \mathcal{R}_{d,+} R d , + , and L ϵ \mathcal{L}_\epsilon L ϵ , we obtain
∥ ( R n , + ( a ) − R n , + ( b ) ) [ f ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , ∥ X s s ( a ) ∥ C 1 , α , ∥ X s s ( b ) ∥ C 1 , α , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ f ∥ C 0 , α ( T ) . \|(\mathcal{R}_{n,+}^{(a)}-\mathcal{R}_{n,+}^{(b)})[f]\|_{C^{1,\gamma}(\mathbb{T})}\leq c(\epsilon,\|\boldsymbol{X}_{ss}^{(a)}\|_{C^{1,\alpha}},\|\boldsymbol{X}_{ss}^{(b)}\|_{C^{1,\alpha}},c_{\Gamma})\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha_{+}}(\mathbb{T})}\|f\|_{C^{0,\alpha}(\mathbb{T})}. ∥ ( R n , + ( a ) − R n , + ( b ) ) [ f ] ∥ C 1 , γ ( T ) ≤ c ( ϵ , ∥ X ss ( a ) ∥ C 1 , α , ∥ X ss ( b ) ∥ C 1 , α , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( T ) ∥ f ∥ C 0 , α ( T ) .
In total, we obtain the decomposition (32) and estimates (33) , (34) for the slender body NtD map L ϵ \mathcal{L}_\epsilon L ϵ . □
Proof of Corollary 1.6. Given slender body force data of the form f ( s ) = ∂ s 4 F \boldsymbol{f}(s)=\partial_s^4F f ( s ) = ∂ s 4 F , we may first use the commutator bounds (78) for Φ \Phi Φ to write
Φ − 1 ∂ s 4 F = ∂ s 2 ( Φ − 1 ∂ s 2 F ) + R X , 1 , \Phi^{-1}\partial_s^4F=\partial_s^2(\Phi^{-1}\partial_s^2F)+\mathcal{R}_{X,1}, Φ − 1 ∂ s 4 F = ∂ s 2 ( Φ − 1 ∂ s 2 F ) + R X , 1 ,
∥ R X , 1 ∥ C 0 , α = ∥ ∂ s ( [ ∂ s , Φ − 1 ] ∂ s 2 F ) + [ ∂ s , Φ − 1 ] ∂ s 3 F ∥ C 0 , α ≤ c ( ∥ X s s ∥ C 1 , α ) ∥ F ∥ C 3 , α . \|\mathcal{R}_{X,1}\|_{C^{0,\alpha}}=\|\partial_s([\partial_s,\Phi^{-1}]\partial_s^2F)+[\partial_s,\Phi^{-1}]\partial_s^3F\|_{C^{0,\alpha}}\leq c(\|\boldsymbol{X}_{ss}\|_{C^{1,\alpha}})\|F\|_{C^{3,\alpha}}. ∥ R X , 1 ∥ C 0 , α = ∥ ∂ s ([ ∂ s , Φ − 1 ] ∂ s 2 F ) + [ ∂ s , Φ − 1 ] ∂ s 3 F ∥ C 0 , α ≤ c ( ∥ X ss ∥ C 1 , α ) ∥ F ∥ C 3 , α .
For two nearby filaments with centerlines X ( a ) \boldsymbol{X}^{(a)} X ( a ) and X ( b ) \boldsymbol{X}^{(b)} X ( b ) satisfying Lemma 1.1 , using the commutator estimates (82) , we additionally have
∥ R X , 1 ( a ) − R X , 1 ( b ) ∥ C 0 , α ≤ c ( ∥ X ( a ) ∥ C 3 , α , ∥ X ( b ) ∥ C 3 , α ) ∥ X ( a ) − X ( b ) ∥ C 3 , α ∥ F ∥ C 3 , α . \|\mathcal{R}_{X,1}^{(a)}-\mathcal{R}_{X,1}^{(b)}\|_{C^{0,\alpha}}\leq c(\|\boldsymbol{X}^{(a)}\|_{C^{3,\alpha}},\|\boldsymbol{X}^{(b)}\|_{C^{3,\alpha}})\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{3,\alpha}}\|F\|_{C^{3,\alpha}}. ∥ R X , 1 ( a ) − R X , 1 ( b ) ∥ C 0 , α ≤ c ( ∥ X ( a ) ∥ C 3 , α , ∥ X ( b ) ∥ C 3 , α ) ∥ X ( a ) − X ( b ) ∥ C 3 , α ∥ F ∥ C 3 , α .
We next note the following commutator estimate for L ‾ ϵ ∂ s 2 \overline{\mathcal{L}}_\epsilon\partial_s^2 L ϵ ∂ s 2 .
Proposition 2.1 (Commutator estimate for L ‾ ϵ ∂ s 2 \overline{\mathcal{L}}_\epsilon\partial_s^2 L ϵ ∂ s 2 ). Given a matrix-valued function A ∈ C n + 1 , α ( T ) \boldsymbol{A}\in C^{n+1,\alpha}(\mathbb{T}) A ∈ C n + 1 , α ( T ) and a vector-valued function g ∈ C n , α ( T ) \boldsymbol{g}\in C^{n,\alpha}(\mathbb{T}) g ∈ C n , α ( T ) for 0 ≤ n ∈ Z 0\leq n\in\mathbb{Z} 0 ≤ n ∈ Z , the commutator [ L ‾ ϵ ∂ s 2 , A ] g [\overline{\mathcal{L}}_\epsilon\partial_s^2,\boldsymbol{A}]\boldsymbol{g} [ L ϵ ∂ s 2 , A ] g satisfies
∥ [ L ‾ ϵ ∂ s 2 , A ] g ∥ C n , α ( T ) ≤ c ϵ − 1 ∣ log ϵ ∣ ∥ A ∥ C n + 1 , α ∥ g ∥ C n , α ( T ) . (88) \|[\overline{\mathcal{L}}_\epsilon\partial_s^2,\boldsymbol{A}]\boldsymbol{g}\|_{C^{n,\alpha}(\mathbb{T})}\leq c\epsilon^{-1}|\log\epsilon|\|\boldsymbol{A}\|_{C^{n+1,\alpha}}\|\boldsymbol{g}\|_{C^{n,\alpha}(\mathbb{T})}. \tag*{(88)} ∥ [ L ϵ ∂ s 2 , A ] g ∥ C n , α ( T ) ≤ c ϵ − 1 ∣ log ϵ ∣∥ A ∥ C n + 1 , α ∥ g ∥ C n , α ( T ) . (88)
The proof of Proposition 2.1 appears in Appendix C. Applying Proposition 2.1 in the case A = Φ − 1 = e z ⊗ e t + e x ⊗ e n 1 + e y ⊗ e n 2 \boldsymbol{A}=\Phi^{-1}=\boldsymbol{e}_z\otimes\boldsymbol{e}_t+\boldsymbol{e}_x\otimes\boldsymbol{e}_{n_1}+\boldsymbol{e}_y\otimes\boldsymbol{e}_{n_2} A = Φ − 1 = e z ⊗ e t + e x ⊗ e n 1 + e y ⊗ e n 2 and n = 1 n=1 n = 1 , we have
( L ‾ ϵ ∂ s 2 ) [ Φ − 1 ∂ s 2 F ] = Φ − 1 L ‾ ϵ ∂ s 4 F + R X , 2 , (\overline{\mathcal{L}}_\epsilon\partial_s^2)[\Phi^{-1}\partial_s^2F]=\Phi^{-1}\overline{\mathcal{L}}_\epsilon\partial_s^4F+\mathcal{R}_{X,2}, ( L ϵ ∂ s 2 ) [ Φ − 1 ∂ s 2 F ] = Φ − 1 L ϵ ∂ s 4 F + R X , 2 ,
∥ R X , 2 ∥ C 1 , γ = ∥ [ L ‾ ϵ ∂ s 2 , Φ − 1 ] ∂ s 2 F ∥ C 1 , γ ≤ c ( ϵ , ∥ X ∥ C 3 , γ ) ∥ F ∥ C 3 , γ . \|\mathcal{R}_{X,2}\|_{C^{1,\gamma}}=\|[\overline{\mathcal{L}}_\epsilon\partial_s^2,\Phi^{-1}]\partial_s^2F\|_{C^{1,\gamma}}\leq c(\epsilon,\|\boldsymbol{X}\|_{C^{3,\gamma}})\|F\|_{C^{3,\gamma}}. ∥ R X , 2 ∥ C 1 , γ = ∥ [ L ϵ ∂ s 2 , Φ − 1 ] ∂ s 2 F ∥ C 1 , γ ≤ c ( ϵ , ∥ X ∥ C 3 , γ ) ∥ F ∥ C 3 , γ .
Furthermore, for two nearby curves X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) , we have
∥ R X , 2 ( a ) − R X , 2 ( b ) ∥ C 1 , γ ≤ c ( ϵ , ∥ X ( a ) ∥ C 3 , γ , ∥ X ( b ) ∥ C 3 , γ ) ∥ X ( a ) − X ( b ) ∥ C 3 , γ ∥ F ∥ C 3 , γ . \left\|\mathcal{R}_{X,2}^{(a)}-\mathcal{R}_{X,2}^{(b)}\right\|_{C^{1,\gamma}}\leq c\left(\epsilon,\left\|\mathbf{X}^{(a)}\right\|_{C^{3,\gamma}},\left\|\mathbf{X}^{(b)}\right\|_{C^{3,\gamma}}\right)\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{3,\gamma}}\|F\|_{C^{3,\gamma}}. R X , 2 ( a ) − R X , 2 ( b ) C 1 , γ ≤ c ( ϵ , X ( a ) C 3 , γ , X ( b ) C 3 , γ ) X ( a ) − X ( b ) C 3 , γ ∥ F ∥ C 3 , γ .
In total, we may rewrite
Φ ˉ L ϵ [ Φ − 1 [ ∂ s 4 F ] 0 Φ ] = L ˉ ϵ [ ∂ s 4 F ] + Φ L ˉ ϵ ( R X , 1 ) 0 Φ + Φ R X , 2 . (89) \bar{\Phi}\mathcal{L}_{\epsilon}\left[\Phi^{-1}\left[\partial_s^4F\right]^\Phi_0\right]=\bar{\mathcal{L}}_{\epsilon}\left[\partial_s^4F\right]+\Phi\bar{\mathcal{L}}_{\epsilon}\left(\mathcal{R}_{X,1}\right)^\Phi_0+\Phi\mathcal{R}_{X,2}. \tag*{(89)} Φ ˉ L ϵ [ Φ − 1 [ ∂ s 4 F ] 0 Φ ] = L ˉ ϵ [ ∂ s 4 F ] + Φ L ˉ ϵ ( R X , 1 ) 0 Φ + Φ R X , 2 . (89)
Defining R ~ n , + = R n , + + ( I + R n , ϵ ) Φ L ˉ ϵ ( R X , 1 ) 0 Φ + ( I + R n , ϵ ) Φ R X , 2 \widetilde{\mathcal{R}}_{n,+}=\mathcal{R}_{n,+}+(\mathbf{I}+\mathcal{R}_{n,\epsilon})\Phi\bar{\mathcal{L}}_{\epsilon}\left(\mathcal{R}_{X,1}\right)^\Phi_0+(\mathbf{I}+\mathcal{R}_{n,\epsilon})\Phi\mathcal{R}_{X,2} R n , + = R n , + + ( I + R n , ϵ ) Φ L ˉ ϵ ( R X , 1 ) 0 Φ + ( I + R n , ϵ ) Φ R X , 2 , we obtain Corollary 1.6. □
Proof of Theorem 1.2: local well-posedness
We return to the notation X σ ( σ , t ) \mathbf{X}^{\sigma}(\sigma,t) X σ ( σ , t ) of section 1.2.1 for an evolving filament that is possibly changing length over time. Using Corollary 1.6, the evolution (20) may be written as
∂ X σ ∂ t ( σ , t ) = − λ − 1 ( ( I + R n , ϵ ) L ˉ ϵ [ ∂ σ 4 X ] + R ~ n , + [ ∂ σ 4 X ] ) , \frac{\partial\mathbf{X}^{\sigma}}{\partial t}(\sigma,t)=-\lambda^{-1}\left((\mathbf{I}+\mathcal{R}_{n,\epsilon})\bar{\mathcal{L}}_{\epsilon}\left[\partial_{\sigma}^{4}\mathbf{X}\right]+\widetilde{\mathcal{R}}_{n,+}\left[\partial_{\sigma}^{4}\mathbf{X}\right]\right), ∂ t ∂ X σ ( σ , t ) = − λ − 1 ( ( I + R n , ϵ ) L ˉ ϵ [ ∂ σ 4 X ] + R n , + [ ∂ σ 4 X ] ) ,
X σ ( σ , 0 ) = ( X σ ) i n ( σ ) , ∣ ∂ σ ( X σ ) i n ∣ = 1 , \mathbf{X}^{\sigma}(\sigma,0)=\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}(\sigma),\qquad\left|\partial_{\sigma}\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}\right|=1, X σ ( σ , 0 ) = ( X σ ) in ( σ ) , ∂ σ ( X σ ) in = 1 ,
λ ( t ) = ∫ 0 1 ∣ ∂ σ X σ ∣ d σ . \lambda(t)=\int_0^1\left|\partial_{\sigma}\mathbf{X}^{\sigma}\right|\,d\sigma. λ ( t ) = ∫ 0 1 ∣ ∂ σ X σ ∣ d σ .
We will consider the evolution of the difference Y ( σ , t ) = X σ ( σ , t ) − ( X σ ) i n ( σ ) \mathbf{Y}(\sigma,t)=\mathbf{X}^{\sigma}(\sigma,t)-\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}(\sigma) Y ( σ , t ) = X σ ( σ , t ) − ( X σ ) in ( σ ) , which satisfies
∂ Y ∂ t ( σ , t ) = − λ − 1 ( ( I + R n , ϵ ) L ˉ ϵ [ ∂ σ 4 ( Y + ( X σ ) i n ) ] + R ~ n , + [ ∂ σ 4 ( Y + ( X σ ) i n ) ] ) . (90) \frac{\partial\mathbf{Y}}{\partial t}(\sigma,t)=-\lambda^{-1}\left((\mathbf{I}+\mathcal{R}_{n,\epsilon})\bar{\mathcal{L}}_{\epsilon}\left[\partial_{\sigma}^{4}\left(\mathbf{Y}+\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}\right)\right]+\widetilde{\mathcal{R}}_{n,+}\left[\partial_{\sigma}^{4}\left(\mathbf{Y}+\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}\right)\right]\right). \tag*{(90)} ∂ t ∂ Y ( σ , t ) = − λ − 1 ( ( I + R n , ϵ ) L ˉ ϵ [ ∂ σ 4 ( Y + ( X σ ) in ) ] + R n , + [ ∂ σ 4 ( Y + ( X σ ) in ) ] ) . (90)
Using Duhamel’s formula, we may write Y ( σ , t ) \mathbf{Y}(\sigma,t) Y ( σ , t ) as
Y ( σ , t ) = e − ∫ 0 t λ − 1 ( τ ) d τ L ˉ ϵ ∂ σ 4 ( X σ ) i n − ( X σ ) i n \mathbf{Y}(\sigma,t)=e^{-\int_0^t\lambda^{-1}(\tau)d\tau\,\bar{\mathcal{L}}_{\epsilon}\partial_{\sigma}^{4}}\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}-\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}} Y ( σ , t ) = e − ∫ 0 t λ − 1 ( τ ) d τ L ˉ ϵ ∂ σ 4 ( X σ ) in − ( X σ ) in
− ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ˉ ϵ ∂ σ 4 λ − 1 ( R ϵ [ L ˉ ϵ ∂ σ 4 ( Y + ( X σ ) i n ) ] + R ~ + [ ∂ σ 4 ( Y + ( X σ ) i n ) ] ) d t ′ . (91) \qquad-\int_0^t e^{-\int_{t'}^t\lambda^{-1}(\tau)d\tau\,\bar{\mathcal{L}}_{\epsilon}\partial_{\sigma}^{4}}\lambda^{-1}\left(\mathcal{R}_{\epsilon}\left[\bar{\mathcal{L}}_{\epsilon}\partial_{\sigma}^{4}\left(\mathbf{Y}+\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}\right)\right]+\widetilde{\mathcal{R}}_{+}\left[\partial_{\sigma}^{4}\left(\mathbf{Y}+\left(\mathbf{X}^{\sigma}\right)^{\mathrm{in}}\right)\right]\right)\,dt'. \tag*{(91)} − ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ˉ ϵ ∂ σ 4 λ − 1 ( R ϵ [ L ˉ ϵ ∂ σ 4 ( Y + ( X σ ) in ) ] + R + [ ∂ σ 4 ( Y + ( X σ ) in ) ] ) d t ′ . (91)
We will rely on the following two lemmas to estimate Y \mathbf{Y} Y given by (90).
Lemma 2.2 (Semigroup estimates). Let n , m n,m n , m be nonnegative integers and α , γ ∈ ( 0 , 1 ) \alpha,\gamma\in(0,1) α , γ ∈ ( 0 , 1 ) with n + α ≥ m + γ n+\alpha\geq m+\gamma n + α ≥ m + γ . For V ( s ) ∈ C m , γ ( T ) V(s)\in C^{m,\gamma}(\mathbb{T}) V ( s ) ∈ C m , γ ( T ) , we have
∥ e − t L ˉ ϵ ∂ s 4 V ∥ C n , α ( T ) ≤ c t − n + α − ( m + γ ) 3 ∥ V ∥ C m , γ ( T ) . (92) \left\|e^{-t\bar{\mathcal{L}}_{\epsilon}\partial_s^4}V\right\|_{C^{n,\alpha}(\mathbb{T})}\leq ct^{-\frac{n+\alpha-(m+\gamma)}{3}}\|V\|_{C^{m,\gamma}(\mathbb{T})}. \tag*{(92)} e − t L ˉ ϵ ∂ s 4 V C n , α ( T ) ≤ c t − 3 n + α − ( m + γ ) ∥ V ∥ C m , γ ( T ) . (92) In each case, c c c is an absolute constant bounded independent of ϵ \epsilon ϵ as ϵ → 0 \epsilon\to0 ϵ → 0 .
For the next lemma, given T > 0 T>0 T > 0 and 0 ≤ k ∈ Z 0\leq k\in\mathbb{Z} 0 ≤ k ∈ Z , we define the function spaces
Y k = { V ∈ L ∞ ( 0 , T ; C k , α ( T ) ) : ∥ V ∥ Y k < ∞ } , ∥ ⋅ ∥ Y k = ess sup t ∈ [ 0 , T ] ∥ ⋅ ∥ C k , α ( T ) . (93) \mathcal{Y}_k=\left\{V\in L^\infty(0,T;C^{k,\alpha}(\mathbb{T})):\|V\|_{\mathcal{Y}_k}<\infty\right\},\qquad\|\cdot\|_{\mathcal{Y}_k}=\operatorname*{ess\,sup}_{t\in[0,T]}\|\cdot\|_{C^{k,\alpha}(\mathbb{T})}. \tag*{(93)} Y k = { V ∈ L ∞ ( 0 , T ; C k , α ( T )) : ∥ V ∥ Y k < ∞ } , ∥ ⋅ ∥ Y k = t ∈ [ 0 , T ] ess sup ∥ ⋅ ∥ C k , α ( T ) . (93)
We have the following.
Lemma 2.3 (Maximal regularity bound). Given g ∈ Y 1 g\in\mathcal{Y}_1 g ∈ Y 1 , let
h ( s , t ) = ∫ 0 t e − ( t − t ′ ) L ˉ ϵ ∂ s 4 g ( s , t ′ ) d t ′ . h(s,t)=\int_0^t e^{-(t-t')\bar{\mathcal{L}}_{\epsilon}\partial_s^4}g(s,t')\,dt'. h ( s , t ) = ∫ 0 t e − ( t − t ′ ) L ˉ ϵ ∂ s 4 g ( s , t ′ ) d t ′ . Then, for some absolute constant c c c ,
∥ h ∥ Y 4 ≤ ( ϵ + T 1 / 4 ∣ log ϵ ∣ − 3 / 4 + T ) c ∥ g ∥ Y 1 . (94) \|h\|_{\mathcal{Y}_4}\leq\left(\epsilon+T^{1/4}|\log\epsilon|^{-3/4}+T\right)c\|g\|_{\mathcal{Y}_1}. \tag*{(94)} ∥ h ∥ Y 4 ≤ ( ϵ + T 1/4 ∣ log ϵ ∣ − 3/4 + T ) c ∥ g ∥ Y 1 . (94)
The proofs of Lemmas 2.2 and (2.3) appear in section 3.5.
Now, to track the non-self-intersection property (2), we will define the quantity
∣ X σ ( ⋅ , t ) ∣ ⋆ : = inf σ ≠ σ ′ ∣ X σ ( σ , t ) − X σ ( σ ′ , t ) ∣ ∣ σ − σ ′ ∣ . \lvert X^\sigma(\cdot,t)\rvert_\star:= \inf_{\sigma\ne\sigma'} \frac{\lvert X^\sigma(\sigma,t)-X^\sigma(\sigma',t)\rvert}{\lvert\sigma-\sigma'\rvert}. ∣ X σ ( ⋅ , t ) ∣ ⋆ := σ = σ ′ inf ∣ σ − σ ′ ∣ ∣ X σ ( σ , t ) − X σ ( σ ′ , t )∣ .
Using the definition (92) of the spaces Y k \mathcal{Y}_k Y k , we consider the closed ball
B M , 1 2 ( T ) = { Y ∈ Y 4 : ∥ Y ∥ Y 4 ≤ M , ∥ Y ∥ Y 1 ≤ min { 1 2 ∣ ( X σ ) i n ∣ ⋆ , 1 2 } } . \mathcal{B}_{M,\frac{1}{2}}(T)=\left\{Y\in\mathcal{Y}_4:\lVert Y\rVert_{\mathcal{Y}_4}\le M,\ \lVert Y\rVert_{\mathcal{Y}_1}\le\min\left\{\frac{1}{2}\lvert(X^\sigma)^{\mathrm{in}}\rvert_\star,\frac{1}{2}\right\}\right\}. B M , 2 1 ( T ) = { Y ∈ Y 4 : ∥ Y ∥ Y 4 ≤ M , ∥ Y ∥ Y 1 ≤ min { 2 1 ∣( X σ ) in ∣ ⋆ , 2 1 } } .
The Y 1 \mathcal{Y}_1 Y 1 bound will be used to propagate a non-self-intersection condition and to consider only situations where the filament length remains close to its initial unit length. Note that this bound ensures that
ess sup t ∈ [ 0 , T ] ∣ X σ ( ⋅ , t ) ∣ ⋆ ≥ ∣ ( X σ ) i n ∣ ⋆ − ∥ Y ∥ Y 1 ≥ 1 2 ∣ ( X σ ) i n ∣ ⋆ . \mathop{\operatorname{ess\,sup}}_{t\in[0,T]}\lvert X^\sigma(\cdot,t)\rvert_\star\ge\lvert(X^\sigma)^{\mathrm{in}}\rvert_\star-\lVert Y\rVert_{\mathcal{Y}_1}\ge\frac{1}{2}\lvert(X^\sigma)^{\mathrm{in}}\rvert_\star. ess sup t ∈ [ 0 , T ] ∣ X σ ( ⋅ , t ) ∣ ⋆ ≥ ∣( X σ ) in ∣ ⋆ − ∥ Y ∥ Y 1 ≥ 2 1 ∣( X σ ) in ∣ ⋆ .
Furthermore, if Y Y Y satisfies ∥ Y ∥ Y 1 ≤ 1 2 \lVert Y\rVert_{\mathcal{Y}_1}\le\frac{1}{2} ∥ Y ∥ Y 1 ≤ 2 1 , then, using that ∣ ∂ σ ( X σ ) i n ∣ = 1 \lvert\partial_\sigma(X^\sigma)^{\mathrm{in}}\rvert=1 ∣ ∂ σ ( X σ ) in ∣ = 1 and
λ = ∫ 0 1 ∣ ∂ σ Y + ∂ σ ( X σ ) i n ∣ d σ , (95) \lambda=\int_0^1\lvert\partial_\sigma Y+\partial_\sigma(X^\sigma)^{\mathrm{in}}\rvert\,d\sigma, \tag*{(95)} λ = ∫ 0 1 ∣ ∂ σ Y + ∂ σ ( X σ ) in ∣ d σ , (95)
we have 1 2 ≤ λ ≤ 3 2 \frac{1}{2}\le\lambda\le\frac{3}{2} 2 1 ≤ λ ≤ 2 3 and, in particular,
2 3 ( t − t ′ ) ≤ ∫ t ′ t λ − 1 ( τ ) d τ ≤ 2 ( t − t ′ ) . (96) \frac{2}{3}(t-t')\le\int_{t'}^t\lambda^{-1}(\tau)\,d\tau\le2(t-t'). \tag*{(96)} 3 2 ( t − t ′ ) ≤ ∫ t ′ t λ − 1 ( τ ) d τ ≤ 2 ( t − t ′ ) . (96)
Now, for Y ∈ B M , 1 2 ( T ) Y\in\mathcal{B}_{M,\frac{1}{2}}(T) Y ∈ B M , 2 1 ( T ) , let Λ [ Y ] \Lambda[Y] Λ [ Y ] denote the map
Λ [ Y ] = e − ∫ 0 t λ − 1 ( τ ) d τ L ε ∂ σ 4 ( X σ ) i n − ( X σ ) i n − ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ε ∂ σ 4 λ − 1 ( R ε [ L ε ∂ σ 4 ( Y + ( X σ ) i n ) ] + R ~ + [ ∂ σ 4 ( Y + ( X σ ) i n ) ] ) d t ′ . (97) \begin{aligned}
\Lambda[Y]={}&e^{-\int_0^t\lambda^{-1}(\tau)\,d\tau\,\mathcal{L}_\varepsilon\partial_\sigma^4(X^\sigma)^{\mathrm{in}}-(X^\sigma)^{\mathrm{in}}}\\
&-\int_0^t e^{-\int_{t'}^t\lambda^{-1}(\tau)\,d\tau\,\mathcal{L}_\varepsilon\partial_\sigma^4}\lambda^{-1}\left(\mathcal{R}_\varepsilon\left[\mathcal{L}_\varepsilon\partial_\sigma^4(Y+(X^\sigma)^{\mathrm{in}})\right]+\widetilde{\mathcal{R}}_+\left[\partial_\sigma^4(Y+(X^\sigma)^{\mathrm{in}})\right]\right)\,dt'. \tag*{(97)}
\end{aligned} Λ [ Y ] = e − ∫ 0 t λ − 1 ( τ ) d τ L ε ∂ σ 4 ( X σ ) in − ( X σ ) in − ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ε ∂ σ 4 λ − 1 ( R ε [ L ε ∂ σ 4 ( Y + ( X σ ) in ) ] + R + [ ∂ σ 4 ( Y + ( X σ ) in ) ] ) d t ′ . (97)
We show that the map Λ \Lambda Λ takes B M , 1 2 ( T ) \mathcal{B}_{M,\frac{1}{2}}(T) B M , 2 1 ( T ) to itself for some choice of ε , T > 0 \varepsilon,T>0 ε , T > 0 .
We first estimate the Y 4 \mathcal{Y}_4 Y 4 contribution from the initial condition. Note that by (95) and Lemma 2.2, we have
∥ e − ∫ 0 t λ − 1 ( τ ) d τ L ε ∂ σ 4 ( X σ ) i n − ( X σ ) i n ∥ C 4 , α ( T ) ≤ c ∥ ( X σ ) i n ∥ C 4 , α ( T ) . (98) \left\lVert e^{-\int_0^t\lambda^{-1}(\tau)\,d\tau\,\mathcal{L}_\varepsilon\partial_\sigma^4}(X^\sigma)^{\mathrm{in}}-(X^\sigma)^{\mathrm{in}}\right\rVert_{C^{4,\alpha}(T)}
\le c\lVert(X^\sigma)^{\mathrm{in}}\rVert_{C^{4,\alpha}(T)}. \tag*{(98)} e − ∫ 0 t λ − 1 ( τ ) d τ L ε ∂ σ 4 ( X σ ) in − ( X σ ) in C 4 , α ( T ) ≤ c ∥( X σ ) in ∥ C 4 , α ( T ) . (98)
for some absolute constant c c c . We will take M = 3 c ∥ ( X σ ) i n ∥ C 4 , α ( T ) M=3c\lVert(X^\sigma)^{\mathrm{in}}\rVert_{C^{4,\alpha}(T)} M = 3 c ∥( X σ ) in ∥ C 4 , α ( T ) .
Next, using (95), Lemma 2.2 and Corollary 1.6, we may estimate the smooth remainder term involving R ~ + \widetilde{\mathcal{R}}_+ R + as
∥ ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ε ∂ σ 4 λ − 1 R ~ + [ ∂ σ 4 ( Y + ( X σ ) i n ) ] d t ′ ∥ C 4 , α ( T ) ≤ ∫ 0 t c ( t − t ′ ) − ( 1 − γ − α 3 ) ∥ R ~ + [ ∂ σ 4 ( Y + ( X σ ) i n ) ] ∥ C 1 , γ ( T ) d t ′ ≤ ∫ 0 t ( t − t ′ ) − ( 1 − γ − α 3 ) c ( ε , ∥ Y + ( X σ ) i n ∥ C 3 , γ , ∣ ( X σ ) i n ∣ ⋆ ) ( ∥ Y ∥ C 4 , α ( T ) + ∥ ( X σ ) i n ∥ C 4 , α ( T ) ) d t ′ ≤ t γ − α 3 c ( ε , ∥ Y ∥ Y 4 , ∣ ( X σ ) i n ∣ ⋆ , ∥ ( X σ ) i n ∥ Y 4 ) ( ∥ Y ∥ Y 4 + ∥ ( X σ ) i n ∥ Y 4 ) ≤ t γ − α 3 c ( ε , M , ∣ ( X σ ) i n ∣ ⋆ ) M . \begin{aligned}
\left\lVert\int_0^t e^{-\int_{t'}^t\lambda^{-1}(\tau)\,d\tau\,\mathcal{L}_\varepsilon\partial_\sigma^4}\lambda^{-1}\widetilde{\mathcal{R}}_+\left[\partial_\sigma^4(Y+(X^\sigma)^{\mathrm{in}})\right]\,dt'\right\rVert_{C^{4,\alpha}(T)}
&\le\int_0^t c(t-t')^{-(1-\frac{\gamma-\alpha}{3})}\left\lVert\widetilde{\mathcal{R}}_+\left[\partial_\sigma^4(Y+(X^\sigma)^{\mathrm{in}})\right]\right\rVert_{C^{1,\gamma}(T)}\,dt'\\
&\le\int_0^t(t-t')^{-(1-\frac{\gamma-\alpha}{3})}c\left(\varepsilon,\lVert Y+(X^\sigma)^{\mathrm{in}}\rVert_{C^{3,\gamma}},\lvert(X^\sigma)^{\mathrm{in}}\rvert_\star\right)\left(\lVert Y\rVert_{C^{4,\alpha}(T)}+\lVert(X^\sigma)^{\mathrm{in}}\rVert_{C^{4,\alpha}(T)}\right)\,dt'\\
&\le t^{\frac{\gamma-\alpha}{3}}c\left(\varepsilon,\lVert Y\rVert_{\mathcal{Y}_4},\lvert(X^\sigma)^{\mathrm{in}}\rvert_\star,\lVert(X^\sigma)^{\mathrm{in}}\rVert_{\mathcal{Y}_4}\right)\left(\lVert Y\rVert_{\mathcal{Y}_4}+\lVert(X^\sigma)^{\mathrm{in}}\rVert_{\mathcal{Y}_4}\right)\\
&\le t^{\frac{\gamma-\alpha}{3}}c\left(\varepsilon,M,\lvert(X^\sigma)^{\mathrm{in}}\rvert_\star\right)M.
\end{aligned} ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ε ∂ σ 4 λ − 1 R + [ ∂ σ 4 ( Y + ( X σ ) in ) ] d t ′ C 4 , α ( T ) ≤ ∫ 0 t c ( t − t ′ ) − ( 1 − 3 γ − α ) R + [ ∂ σ 4 ( Y + ( X σ ) in ) ] C 1 , γ ( T ) d t ′ ≤ ∫ 0 t ( t − t ′ ) − ( 1 − 3 γ − α ) c ( ε , ∥ Y + ( X σ ) in ∥ C 3 , γ , ∣( X σ ) in ∣ ⋆ ) ( ∥ Y ∥ C 4 , α ( T ) + ∥( X σ ) in ∥ C 4 , α ( T ) ) d t ′ ≤ t 3 γ − α c ( ε , ∥ Y ∥ Y 4 , ∣( X σ ) in ∣ ⋆ , ∥( X σ ) in ∥ Y 4 ) ( ∥ Y ∥ Y 4 + ∥( X σ ) in ∥ Y 4 ) ≤ t 3 γ − α c ( ε , M , ∣( X σ ) in ∣ ⋆ ) M .
In addition, we may estimate the smooth remainder term in Y 1 \mathcal{Y}_1 Y 1 using
∥ ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ‾ ϵ ∂ σ 4 λ − 1 R ϵ [ ∂ σ 4 ( Y + ( X σ ) i n ) ] d t ′ ∥ C 1 , α ( T ) ≤ c ∫ 0 t ∥ R ϵ [ ∂ σ 4 ( Y + ( X σ ) i n ) ] ∥ C 1 , α ( T ) d t ′ ≤ t c ( ϵ , ∥ Y ∥ Y 4 , ∥ ( X σ ) i n ∥ Y 4 , ∣ ( X σ ) i n ∣ ⋆ ) ( ∥ Y ∥ Y 4 + ∥ ( X σ ) i n ∥ Y 4 ) ≤ t c ( ϵ , M , ∣ ( X σ ) i n ∣ ⋆ ) M . \begin{aligned}
\left\|\int_0^t e^{-\int_{t'}^t \lambda^{-1}(\tau)\,d\tau}\,\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\lambda^{-1}\mathcal{R}_{\epsilon}\left[\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right]dt'\right\|_{C^{1,\alpha}(\mathbb{T})}
\leq c\int_0^t\left\|\mathcal{R}_{\epsilon}\left[\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right]\right\|_{C^{1,\alpha}(\mathbb{T})}\,dt' \\
&\leq t\,c\left(\epsilon,\|\mathbf{Y}\|_{\mathcal{Y}_4},\left\|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{\mathcal{Y}_4},\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)\left(\|\mathbf{Y}\|_{\mathcal{Y}_4}+\left\|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{\mathcal{Y}_4}\right)\leq t\,c\left(\epsilon,M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)M.
\end{aligned} ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ϵ ∂ σ 4 λ − 1 R ϵ [ ∂ σ 4 ( Y + ( X σ ) in ) ] d t ′ C 1 , α ( T ) ≤ c ∫ 0 t R ϵ [ ∂ σ 4 ( Y + ( X σ ) in ) ] C 1 , α ( T ) d t ′ ≤ t c ( ϵ , ∥ Y ∥ Y 4 , ( X σ ) in Y 4 , ( X σ ) in ⋆ ) ( ∥ Y ∥ Y 4 + ( X σ ) in Y 4 ) ≤ t c ( ϵ , M , ( X σ ) in ⋆ ) M .
Furthermore, for the small-in-ϵ \epsilon ϵ remainder term, we may use (95) and Lemma 2.3 along with Corollary 1.6 and Lemma 1.4 to estimate
∥ ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ‾ ϵ ∂ σ 4 λ − 1 R ϵ [ L ‾ ϵ ∂ σ 4 ( Y + ( X σ ) i n ) ] d t ′ ∥ Y 4 ≤ ( ϵ + T 1 / 4 ∣ log ϵ ∣ − 3 / 4 + T ) c ∥ R ϵ [ L ‾ ϵ ∂ σ 4 ( Y + ( X σ ) i n ) ] ∥ Y 1 ≤ ϵ 1 − α + ∣ log ϵ ∣ ( ϵ + T 1 / 4 ∣ log ϵ ∣ − 3 / 4 + T ) c ∥ Y + ( X σ ) i n ∥ Y 4 ∥ ( X σ ) i n ∥ ⋆ ∥ L ‾ ϵ ∂ σ 4 ( Y + ( X σ ) i n ) ∥ Y 1 ≤ ϵ 1 − α + ∣ log ϵ ∣ ( ϵ + T 1 / 4 ∣ log ϵ ∣ − 3 / 4 + T ) c ( M , ∣ ( X σ ) i n ∣ ⋆ ) ( ϵ − 1 ∣ log ϵ ∣ M ) ≤ ϵ 1 − α + ∣ log ϵ ∣ 2 c ( M , ∣ ( X σ ) i n ∣ ⋆ ) M + ( T 1 / 4 + T ) c ( ϵ , M , ∣ ( X σ ) i n ∣ ⋆ ) M . \begin{aligned}
\left\|\int_0^t e^{-\int_{t'}^t\lambda^{-1}(\tau)\,d\tau}\,\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\lambda^{-1}\mathcal{R}_{\epsilon}\left[\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right]dt'\right\|_{\mathcal{Y}_4}
\leq\left(\epsilon+T^{1/4}|\log\epsilon|^{-3/4}+T\right)c\left\|\mathcal{R}_{\epsilon}\left[\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right]\right\|_{\mathcal{Y}_1} \\
&\leq\epsilon^{1-\alpha^+}|\log\epsilon|\left(\epsilon+T^{1/4}|\log\epsilon|^{-3/4}+T\right)c\left\|\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{\mathcal{Y}_4}\left\|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{\star}\left\|\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right\|_{\mathcal{Y}_1} \\
&\leq\epsilon^{1-\alpha^+}|\log\epsilon|\left(\epsilon+T^{1/4}|\log\epsilon|^{-3/4}+T\right)c\left(M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)\left(\epsilon^{-1}|\log\epsilon|M\right) \\
&\leq\epsilon^{1-\alpha^+}|\log\epsilon|^2c\left(M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)M+\left(T^{1/4}+T\right)c\left(\epsilon,M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)M.
\end{aligned} ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ϵ ∂ σ 4 λ − 1 R ϵ [ L ϵ ∂ σ 4 ( Y + ( X σ ) in ) ] d t ′ Y 4 ≤ ( ϵ + T 1/4 ∣ log ϵ ∣ − 3/4 + T ) c R ϵ [ L ϵ ∂ σ 4 ( Y + ( X σ ) in ) ] Y 1 ≤ ϵ 1 − α + ∣ log ϵ ∣ ( ϵ + T 1/4 ∣ log ϵ ∣ − 3/4 + T ) c Y + ( X σ ) in Y 4 ( X σ ) in ⋆ L ϵ ∂ σ 4 ( Y + ( X σ ) in ) Y 1 ≤ ϵ 1 − α + ∣ log ϵ ∣ ( ϵ + T 1/4 ∣ log ϵ ∣ − 3/4 + T ) c ( M , ( X σ ) in ⋆ ) ( ϵ − 1 ∣ log ϵ ∣ M ) ≤ ϵ 1 − α + ∣ log ϵ ∣ 2 c ( M , ( X σ ) in ⋆ ) M + ( T 1/4 + T ) c ( ϵ , M , ( X σ ) in ⋆ ) M .
In addition, by (95) and Lemma 2.2, we have the C 1 , α ( T ) C^{1,\alpha}(\mathbb{T}) C 1 , α ( T ) bound
∥ ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ‾ ϵ ∂ σ 4 λ − 1 R ϵ [ L ‾ ϵ ∂ σ 4 ( Y + ( X σ ) i n ) ] d t ′ ∥ C 1 , α ( T ) ≤ c ∫ 0 t ∥ R ϵ [ L ‾ ϵ ∂ σ 4 ( Y + ( X σ ) i n ) ] ∥ C 1 , α ( T ) d t ′ ≤ t c ( ϵ , ∥ Y ∥ Y 4 , ∥ ( X σ ) i n ∥ Y 4 , ∣ ( X σ ) i n ∣ ⋆ ) ∥ L ‾ ϵ ∂ σ 4 ( Y + ( X σ ) i n ) ∥ Y 1 ≤ t c ( ϵ , M , ∣ ( X σ ) i n ∣ ⋆ ) M . \begin{aligned}
\left\|\int_0^t e^{-\int_{t'}^t\lambda^{-1}(\tau)\,d\tau}\,\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\lambda^{-1}\mathcal{R}_{\epsilon}\left[\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right]dt'\right\|_{C^{1,\alpha}(\mathbb{T})}
\leq c\int_0^t\left\|\mathcal{R}_{\epsilon}\left[\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right]\right\|_{C^{1,\alpha}(\mathbb{T})}\,dt' \\
&\leq t\,c\left(\epsilon,\|\mathbf{Y}\|_{\mathcal{Y}_4},\left\|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{\mathcal{Y}_4},\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)\left\|\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\left(\mathbf{Y}+\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right)\right\|_{\mathcal{Y}_1}\leq t\,c\left(\epsilon,M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)M.
\end{aligned} ∫ 0 t e − ∫ t ′ t λ − 1 ( τ ) d τ L ϵ ∂ σ 4 λ − 1 R ϵ [ L ϵ ∂ σ 4 ( Y + ( X σ ) in ) ] d t ′ C 1 , α ( T ) ≤ c ∫ 0 t R ϵ [ L ϵ ∂ σ 4 ( Y + ( X σ ) in ) ] C 1 , α ( T ) d t ′ ≤ t c ( ϵ , ∥ Y ∥ Y 4 , ( X σ ) in Y 4 , ( X σ ) in ⋆ ) L ϵ ∂ σ 4 ( Y + ( X σ ) in ) Y 1 ≤ t c ( ϵ , M , ( X σ ) in ⋆ ) M .
Finally, to estimate the Y 1 \mathcal{Y}_1 Y 1 contribution from the initial condition, we consider the equation
∂ Z ∂ t ( σ , t ) = − λ ( t ) − 1 L ϵ ∂ σ 4 Z , Z i n = ( X σ ) i n ( σ ) . \frac{\partial Z}{\partial t}(\sigma,t)=-\lambda(t)^{-1}\mathcal{L}_{\epsilon}\partial_\sigma^4 Z,\qquad\mathbf{Z}^{\mathrm{in}}=\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}(\sigma). ∂ t ∂ Z ( σ , t ) = − λ ( t ) − 1 L ϵ ∂ σ 4 Z , Z in = ( X σ ) in ( σ ) .
Using the mapping properties of L ‾ ϵ \overline{\mathcal{L}}_{\epsilon} L ϵ (Lemma 1.4), at each t t t we have
∥ ∂ t Z ∥ C 1 , α ( T ) ≤ c ( ϵ ) ∥ Z ∥ C 4 , α ( T ) ≤ c ( ϵ ) ∥ ( X σ ) i n ∥ C 4 , α ( T ) . \|\partial_t Z\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\epsilon)\|\mathbf{Z}\|_{C^{4,\alpha}(\mathbb{T})}\leq c(\epsilon)\left\|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{C^{4,\alpha}(\mathbb{T})}. ∥ ∂ t Z ∥ C 1 , α ( T ) ≤ c ( ϵ ) ∥ Z ∥ C 4 , α ( T ) ≤ c ( ϵ ) ( X σ ) in C 4 , α ( T ) .
Thus
∥ Z − ( X σ ) i n ∥ C 1 , α ( T ) = ∥ ∫ 0 t ∂ t Z d t ′ ∥ C 1 , α ( T ) ≤ c ( ϵ ) t ∥ ( X σ ) i n ∥ C 4 , α ( T ) ; \left\|\mathbf{Z}-\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{C^{1,\alpha}(\mathbb{T})}=\left\|\int_0^t\partial_t Z\,dt'\right\|_{C^{1,\alpha}(\mathbb{T})}\leq c(\epsilon)t\left\|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{C^{4,\alpha}(\mathbb{T})}; Z − ( X σ ) in C 1 , α ( T ) = ∫ 0 t ∂ t Z d t ′ C 1 , α ( T ) ≤ c ( ϵ ) t ( X σ ) in C 4 , α ( T ) ;
in particular, we have
∥ e − ∫ 0 t λ − 1 ( τ ) d τ L ‾ ϵ ∂ σ 4 ( X σ ) i n − ( X σ ) i n ∥ Y 1 ≤ c ( ϵ ) T ∥ ( X σ ) i n ∥ C 4 , α ( T ) ≤ c ( ϵ ) T M . (99) \left\|e^{-\int_0^t\lambda^{-1}(\tau)\,d\tau}\overline{\mathcal{L}}_{\epsilon}\partial_\sigma^4\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}-\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{\mathcal{Y}_1}\leq c(\epsilon)T\left\|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right\|_{C^{4,\alpha}(\mathbb{T})}\leq c(\epsilon)TM. \tag*{(99)} e − ∫ 0 t λ − 1 ( τ ) d τ L ϵ ∂ σ 4 ( X σ ) in − ( X σ ) in Y 1 ≤ c ( ϵ ) T ( X σ ) in C 4 , α ( T ) ≤ c ( ϵ ) T M . (99)
Altogether, we have
∥ Λ [ Y ] ∥ Y 4 ≤ ( 1 3 + ( T γ − α 3 + T 1 / 4 + T ) c ( ϵ , M , ∣ ( X σ ) i n ∣ ⋆ ) + ϵ 1 − α + ∣ log ϵ ∣ 2 c ( M , ∣ ( X σ ) i n ∣ ⋆ ) ) M , \|\Lambda[\mathbf{Y}]\|_{\mathcal{Y}_4}\leq\left(\frac{1}{3}+\left(T^{\frac{\gamma-\alpha}{3}}+T^{1/4}+T\right)c\left(\epsilon,M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)+\epsilon^{1-\alpha^+}|\log\epsilon|^2c\left(M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)\right)M, ∥Λ [ Y ] ∥ Y 4 ≤ ( 3 1 + ( T 3 γ − α + T 1/4 + T ) c ( ϵ , M , ( X σ ) in ⋆ ) + ϵ 1 − α + ∣ log ϵ ∣ 2 c ( M , ( X σ ) in ⋆ ) ) M ,
∥ Λ [ Y ] ∥ Y 1 ≤ T c ( ϵ , M , ∣ ( X σ ) i n ∣ ⋆ ) M . \|\Lambda[\mathbf{Y}]\|_{\mathcal{Y}_1}\leq Tc\left(\epsilon,M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)M. ∥Λ [ Y ] ∥ Y 1 ≤ T c ( ϵ , M , ( X σ ) in ⋆ ) M .
We first take ϵ \epsilon ϵ small enough that
ϵ 1 − α + ∣ log ϵ ∣ 2 c ( M , ∣ ( X σ ) i n ∣ ⋆ ) ≤ 1 3 \epsilon^{1-\alpha^+}|\log\epsilon|^2c\left(M,\left|\left(\mathbf{X}^\sigma\right)^{\mathrm{in}}\right|_{\star}\right)\leq\frac{1}{3} ϵ 1 − α + ∣ log ϵ ∣ 2 c ( M , ( X σ ) in ⋆ ) ≤ 3 1
and
ϵ ≤ min { 1 2 r ⋆ ( ∣ X i n ∣ ⋆ 2 , M ) , ϵ n ( ∣ X i n ∣ ⋆ 2 , M ) } . (100) \epsilon\leq\min\left\{\frac{1}{2}r_{\star}\left(\frac{\left|\mathbf{X}^{\mathrm{in}}\right|_{\star}}{2},M\right),\epsilon_n\left(\frac{\left|\mathbf{X}^{\mathrm{in}}\right|_{\star}}{2},M\right)\right\}. \tag*{(100)} ϵ ≤ min { 2 1 r ⋆ ( 2 X in ⋆ , M ) , ϵ n ( 2 X in ⋆ , M ) } . (100)
where r ∗ r_{*} r ∗ is as in (6) and ϵ n \epsilon_n ϵ n as in Theorem 1.5 . Then, taking T T T small enough that
( T γ − α 3 + T 1 / 4 + T ) c ( ϵ , M , ∣ ( X σ ) i n ∣ ∗ ) ≤ 1 3 and \left(T^{\frac{\gamma-\alpha}{3}}+T^{1/4}+T\right)c(\epsilon,M,\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*})\leq\frac{1}{3}\quad\text{and} ( T 3 γ − α + T 1/4 + T ) c ( ϵ , M , ∣( X σ ) in ∣ ∗ ) ≤ 3 1 and
T c ( ϵ , M , ∣ ( X σ ) i n ∣ ∗ ) ≤ min { 1 2 ∣ ( X σ ) i n ∣ ∗ , 1 2 } , (101) Tc(\epsilon,M,\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*})\leq\min\left\{\frac{1}{2}\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*},\frac{1}{2}\right\}, \tag*{(101)} T c ( ϵ , M , ∣( X σ ) in ∣ ∗ ) ≤ min { 2 1 ∣( X σ ) in ∣ ∗ , 2 1 } , (101)
we obtain
∥ Λ [ Y ] ∥ Y 4 ≤ M , ∥ Λ [ Y ] ∥ Y 1 ≤ min { 1 2 ∣ ( X σ ) i n ∣ ∗ , 1 2 } . \lVert\Lambda[Y]\rVert_{\mathcal{Y}_{4}}\leq M,\qquad\lVert\Lambda[Y]\rVert_{\mathcal{Y}_{1}}\leq\min\left\{\frac{1}{2}\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*},\frac{1}{2}\right\}. ∥ Λ [ Y ] ∥ Y 4 ≤ M , ∥ Λ [ Y ] ∥ Y 1 ≤ min { 2 1 ∣( X σ ) in ∣ ∗ , 2 1 } .
We next show that the map Λ [ Y ] \Lambda[Y] Λ [ Y ] admits a unique fixed point on B M , 1 2 ( T ) \mathcal{B}_{M,\frac{1}{2}}(T) B M , 2 1 ( T ) . We now consider two nearby curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) , and define Y ( a ) = X ( a ) − ( X σ ) i n Y^{(a)}=X^{(a)}-(X^{\sigma})^{\mathrm{in}} Y ( a ) = X ( a ) − ( X σ ) in , Y ( b ) = X ( b ) − ( X σ ) i n Y^{(b)}=X^{(b)}-(X^{\sigma})^{\mathrm{in}} Y ( b ) = X ( b ) − ( X σ ) in . Note that the lengths λ ( a ) \lambda^{(a)} λ ( a ) and λ ( b ) \lambda^{(b)} λ ( b ) of curve (a) and (b), respectively, satisfy
λ ( a ) − λ ( b ) = ∫ 0 1 ( ∣ ∂ σ Y ( a ) + ∂ σ ( X σ ) i n ∣ − ∣ ∂ σ Y ( b ) + ∂ σ ( X σ ) i n ∣ ) d σ = ∫ 0 1 ∂ σ ( Y ( a ) − Y ( b ) ) ⋅ ( ∂ σ Y ( a ) + ∂ σ Y ( b ) + 2 ∂ σ ( X σ ) i n ) ∣ ∂ σ Y ( a ) + ∂ σ ( X σ ) i n ∣ + ∣ ∂ σ Y ( b ) + ∂ σ ( X σ ) i n ∣ d σ ; \begin{aligned}
\lambda^{(a)}-\lambda^{(b)}=\int_{0}^{1}\left(\left\lvert\partial_{\sigma}Y^{(a)}+\partial_{\sigma}(X^{\sigma})^{\mathrm{in}}\right\rvert-\left\lvert\partial_{\sigma}Y^{(b)}+\partial_{\sigma}(X^{\sigma})^{\mathrm{in}}\right\rvert\right)\,d\sigma\\
&=\int_{0}^{1}\frac{\partial_{\sigma}(Y^{(a)}-Y^{(b)})\cdot\left(\partial_{\sigma}Y^{(a)}+\partial_{\sigma}Y^{(b)}+2\partial_{\sigma}(X^{\sigma})^{\mathrm{in}}\right)}{\left\lvert\partial_{\sigma}Y^{(a)}+\partial_{\sigma}(X^{\sigma})^{\mathrm{in}}\right\rvert+\left\lvert\partial_{\sigma}Y^{(b)}+\partial_{\sigma}(X^{\sigma})^{\mathrm{in}}\right\rvert}\,d\sigma;
\end{aligned} λ ( a ) − λ ( b ) = ∫ 0 1 ( ∂ σ Y ( a ) + ∂ σ ( X σ ) in − ∂ σ Y ( b ) + ∂ σ ( X σ ) in ) d σ = ∫ 0 1 ∂ σ Y ( a ) + ∂ σ ( X σ ) in + ∂ σ Y ( b ) + ∂ σ ( X σ ) in ∂ σ ( Y ( a ) − Y ( b ) ) ⋅ ( ∂ σ Y ( a ) + ∂ σ Y ( b ) + 2 ∂ σ ( X σ ) in ) d σ ;
in particular, in B M , 1 2 ( T ) \mathcal{B}_{M,\frac{1}{2}}(T) B M , 2 1 ( T ) , we have
∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) ≤ ∣ λ ( a ) − λ ( b ) ∣ ≤ 3 ∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) , \lVert Y^{(a)}-Y^{(b)}\rVert_{C^{1,\alpha}(\mathbb{T})}\leq\lvert\lambda^{(a)}-\lambda^{(b)}\rvert\leq3\lVert Y^{(a)}-Y^{(b)}\rVert_{C^{1,\alpha}(\mathbb{T})}, ∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) ≤ ∣ λ ( a ) − λ ( b ) ∣ ≤ 3 ∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) ,
4 9 ∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) ≤ ∣ 1 λ ( a ) − 1 λ ( b ) ∣ ≤ 12 ∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) . (102) \frac{4}{9}\lVert Y^{(a)}-Y^{(b)}\rVert_{C^{1,\alpha}(\mathbb{T})}\leq\left\lvert\frac{1}{\lambda^{(a)}}-\frac{1}{\lambda^{(b)}}\right\rvert\leq12\lVert Y^{(a)}-Y^{(b)}\rVert_{C^{1,\alpha}(\mathbb{T})}. \tag*{(102)} 9 4 ∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) ≤ λ ( a ) 1 − λ ( b ) 1 ≤ 12 ∥ Y ( a ) − Y ( b ) ∥ C 1 , α ( T ) . (102)
Let
W + ( j ) = ( λ ( j ) ) − 1 R + ( j ) [ ∂ σ 4 ( Y ( j ) + ( X σ ) i n ) ] W_{+}^{(j)}=(\lambda^{(j)})^{-1}\mathcal{R}^{(j)}_{+}\left[\partial_{\sigma}^{4}(Y^{(j)}+(X^{\sigma})^{\mathrm{in}})\right] W + ( j ) = ( λ ( j ) ) − 1 R + ( j ) [ ∂ σ 4 ( Y ( j ) + ( X σ ) in ) ]
W ϵ ( j ) = ( λ ( j ) ) − 1 R ϵ ( j ) [ L ϵ ∂ σ 4 ( Y ( j ) + ( X σ ) i n ) ] , j = a , b . W_{\epsilon}^{(j)}=(\lambda^{(j)})^{-1}\mathcal{R}^{(j)}_{\epsilon}\left[\mathcal{L}_{\epsilon}\partial_{\sigma}^{4}(Y^{(j)}+(X^{\sigma})^{\mathrm{in}})\right],\qquad j=a,b. W ϵ ( j ) = ( λ ( j ) ) − 1 R ϵ ( j ) [ L ϵ ∂ σ 4 ( Y ( j ) + ( X σ ) in ) ] , j = a , b .
Using (101) and (36), we may estimate
∥ W + ( a ) − W + ( b ) ∥ C 1 , γ ≤ c ( ϵ , M , ∣ ( X σ ) i n ∣ ∗ ) ( ∥ Y ( a ) − Y ( b ) ∥ C 1 , α + ∥ Y ( a ) − Y ( b ) ∥ C 2 , β ∥ Y ( a ) + ( X σ ) i n ∥ C 0 , α + ∥ Y ( a ) − Y ( b ) ∥ C 4 , α ) ≤ c ( ϵ , M , ∣ ( X σ ) i n ∣ ∗ ) ∥ Y ( a ) − Y ( b ) ∥ Y 4 . (103) \lVert W_{+}^{(a)}-W_{+}^{(b)}\rVert_{C^{1,\gamma}}\leq c(\epsilon,M,\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*})\left(\lVert Y^{(a)}-Y^{(b)}\rVert_{C^{1,\alpha}}+\lVert Y^{(a)}-Y^{(b)}\rVert_{C^{2,\beta}}\lVert Y^{(a)}+(X^{\sigma})^{\mathrm{in}}\rVert_{C^{0,\alpha}}+\lVert Y^{(a)}-Y^{(b)}\rVert_{C^{4,\alpha}}\right)\leq c(\epsilon,M,\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*})\lVert Y^{(a)}-Y^{(b)}\rVert_{\mathcal{Y}_{4}}. \tag*{(103)} ∥ W + ( a ) − W + ( b ) ∥ C 1 , γ ≤ c ( ϵ , M , ∣( X σ ) in ∣ ∗ ) ( ∥ Y ( a ) − Y ( b ) ∥ C 1 , α + ∥ Y ( a ) − Y ( b ) ∥ C 2 , β ∥ Y ( a ) + ( X σ ) in ∥ C 0 , α + ∥ Y ( a ) − Y ( b ) ∥ C 4 , α ) ≤ c ( ϵ , M , ∣( X σ ) in ∣ ∗ ) ∥ Y ( a ) − Y ( b ) ∥ Y 4 . (103)
Furthermore, using (101), (36), and Lemma 1.4, we have
∥ W ϵ ( a ) − W ϵ ( b ) ∥ C 1 , α ≤ ϵ 1 − α + † ∣ log ϵ ∣ c ( M , ∣ ( X σ ) i n ∣ ∗ ) ( ∥ Y ( a ) − Y ( b ) ∥ C 2 , β ∥ L ϵ ∂ σ 4 ( Y ( a ) + ( X σ ) i n ) ∥ C 1 , α + ∥ L ϵ ∂ σ 4 ( Y ( a ) − Y ( b ) ) ∥ C 1 , α ) ≤ ϵ − α + † ∣ log ϵ ∣ 2 c ( M , ∣ ( X σ ) i n ∣ ∗ ) ∥ Y ( a ) − Y ( b ) ∥ Y 4 . (104) \begin{aligned}
\lVert W_{\epsilon}^{(a)}-W_{\epsilon}^{(b)}\rVert_{C^{1,\alpha}}\leq\epsilon^{1-\alpha+\dagger}\lvert\log\epsilon\rvert c(M,\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*})\left(\lVert Y^{(a)}-Y^{(b)}\rVert_{C^{2,\beta}}\lVert\mathcal{L}_{\epsilon}\partial_{\sigma}^{4}(Y^{(a)}+(X^{\sigma})^{\mathrm{in}})\rVert_{C^{1,\alpha}}+\lVert\mathcal{L}_{\epsilon}\partial_{\sigma}^{4}(Y^{(a)}-Y^{(b)})\rVert_{C^{1,\alpha}}\right) \\
&\leq\epsilon^{-\alpha+\dagger}\lvert\log\epsilon\rvert^{2}c(M,\lvert(X^{\sigma})^{\mathrm{in}}\rvert_{*})\lVert Y^{(a)}-Y^{(b)}\rVert_{\mathcal{Y}_{4}}. \tag*{(104)}
\end{aligned} ∥ W ϵ ( a ) − W ϵ ( b ) ∥ C 1 , α ≤ ϵ 1 − α +† ∣ log ϵ ∣ c ( M , ∣( X σ ) in ∣ ∗ ) ( ∥ Y ( a ) − Y ( b ) ∥ C 2 , β ∥ L ϵ ∂ σ 4 ( Y ( a ) + ( X σ ) in ) ∥ C 1 , α + ∥ L ϵ ∂ σ 4 ( Y ( a ) − Y ( b ) ) ∥ C 1 , α ) ≤ ϵ − α +† ∣ log ϵ ∣ 2 c ( M , ∣( X σ ) in ∣ ∗ ) ∥ Y ( a ) − Y ( b ) ∥ Y 4 . (104)
Now, integrating (101) in time, we have
4 9 ( t − t ′ ) ∥ Y ( a ) − Y ( b ) ∥ Y 1 ≤ ∫ t ′ t ( 1 λ ( a ) ( τ ) − 1 λ ( b ) ( τ ) ) d τ ≤ 12 ( t − t ′ ) ∥ Y ( a ) − Y ( b ) ∥ Y 1 . (105) \frac{4}{9}(t-t')\lVert Y^{(a)}-Y^{(b)}\rVert_{\mathcal{Y}_{1}}\leq\int_{t'}^{t}\left(\frac{1}{\lambda^{(a)}(\tau)}-\frac{1}{\lambda^{(b)}(\tau)}\right)\,d\tau\leq12(t-t')\lVert Y^{(a)}-Y^{(b)}\rVert_{\mathcal{Y}_{1}}. \tag*{(105)} 9 4 ( t − t ′ ) ∥ Y ( a ) − Y ( b ) ∥ Y 1 ≤ ∫ t ′ t ( λ ( a ) ( τ ) 1 − λ ( b ) ( τ ) 1 ) d τ ≤ 12 ( t − t ′ ) ∥ Y ( a ) − Y ( b ) ∥ Y 1 . (105)
In particular, combining (104), (102), and Lemma 2.2, we may estimate
∥ ∫ 0 t ( e − ∫ t ′ t ( λ ( a ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W + ( a ) − e − ∫ t ′ t ( λ ( b ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W + ( b ) ) d t ′ ∥ C 4 , α ( T ) ≤ ∫ 0 t c ( t − t ′ ) − ( 1 − γ − α 3 ) ( ∥ Y ( a ) − Y ( b ) ∥ Y 1 ∥ W + ( a ) ∥ C 1 , γ ( T ) + ∥ W + ( a ) − W + ( b ) ∥ C 1 , γ ( T ) ) d t ′ ≤ t γ − α 3 c ( ϵ , M , ∥ ( X σ ) i n ∥ ⋆ ) ∥ Y ( a ) − Y ( b ) ∥ Y 4 . \begin{aligned}
\left\|\int_0^t\left(e^{-\int_{t'}^t(\lambda^{(a)})^{-1}(\tau)\,d\tau}\mathcal{L}_\epsilon\partial_\sigma^4\mathbf{W}_+^{(a)}-e^{-\int_{t'}^t(\lambda^{(b)})^{-1}(\tau)\,d\tau}\mathcal{L}_\epsilon\partial_\sigma^4\mathbf{W}_+^{(b)}\right)\,dt'\right\|_{C^{4,\alpha}(\mathbb{T})} \\
&\leq\int_0^t c(t-t')^{-(1-\frac{\gamma-\alpha}{3})}\left(\left\|\mathbf{Y}^{(a)}-\mathbf{Y}^{(b)}\right\|_{\mathcal{Y}_1}\left\|\mathbf{W}_+^{(a)}\right\|_{C^{1,\gamma}(\mathbb{T})}+\left\|\mathbf{W}_+^{(a)}-\mathbf{W}_+^{(b)}\right\|_{C^{1,\gamma}(\mathbb{T})}\right)\,dt' \\
&\leq t^{\frac{\gamma-\alpha}{3}}c(\epsilon,M,\left\|(\mathbf{X}^\sigma)^{\mathrm{in}}\right\|_{\star})\left\|\mathbf{Y}^{(a)}-\mathbf{Y}^{(b)}\right\|_{\mathcal{Y}_4}.
\end{aligned} ∫ 0 t ( e − ∫ t ′ t ( λ ( a ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W + ( a ) − e − ∫ t ′ t ( λ ( b ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W + ( b ) ) d t ′ C 4 , α ( T ) ≤ ∫ 0 t c ( t − t ′ ) − ( 1 − 3 γ − α ) ( Y ( a ) − Y ( b ) Y 1 W + ( a ) C 1 , γ ( T ) + W + ( a ) − W + ( b ) C 1 , γ ( T ) ) d t ′ ≤ t 3 γ − α c ( ϵ , M , ( X σ ) in ⋆ ) Y ( a ) − Y ( b ) Y 4 .
Furthermore, combining (104), (103), and Lemma 2.3, we have
∥ ∫ 0 t ( e − ∫ t ′ t ( λ ( a ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W ϵ ( a ) − e − ∫ t ′ t ( λ ( b ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W ϵ ( b ) ) d t ′ ∥ Y 4 ≤ ( ϵ + T 1 / 4 ∣ log ϵ ∣ − 3 / 4 + T ) c ( ∥ Y ( a ) − Y ( b ) ∥ Y 1 ∥ W ϵ ( a ) ∥ Y 1 + ∥ W ϵ ( a ) − W ϵ ( b ) ∥ Y 1 ) ≤ ϵ − α t ∣ log ϵ ∣ 2 ( ϵ + T 1 / 4 ∣ log ϵ ∣ − 3 / 4 + T ) c ( M , ∥ ( X σ ) i n ∥ ⋆ ) ∥ Y ( a ) − Y ( b ) ∥ Y 4 . \begin{aligned}
\left\|\int_0^t\left(e^{-\int_{t'}^t(\lambda^{(a)})^{-1}(\tau)\,d\tau}\mathcal{L}_\epsilon\partial_\sigma^4\mathbf{W}_\epsilon^{(a)}-e^{-\int_{t'}^t(\lambda^{(b)})^{-1}(\tau)\,d\tau}\mathcal{L}_\epsilon\partial_\sigma^4\mathbf{W}_\epsilon^{(b)}\right)\,dt'\right\|_{\mathcal{Y}_4} \\
&\leq\left(\epsilon+T^{1/4}|\log\epsilon|^{-3/4}+T\right)c\left(\left\|\mathbf{Y}^{(a)}-\mathbf{Y}^{(b)}\right\|_{\mathcal{Y}_1}\left\|\mathbf{W}_\epsilon^{(a)}\right\|_{\mathcal{Y}_1}+\left\|\mathbf{W}_\epsilon^{(a)}-\mathbf{W}_\epsilon^{(b)}\right\|_{\mathcal{Y}_1}\right) \\
&\leq\epsilon^{-\alpha t}|\log\epsilon|^2\left(\epsilon+T^{1/4}|\log\epsilon|^{-3/4}+T\right)c(M,\left\|(\mathbf{X}^\sigma)^{\mathrm{in}}\right\|_{\star})\left\|\mathbf{Y}^{(a)}-\mathbf{Y}^{(b)}\right\|_{\mathcal{Y}_4}.
\end{aligned} ∫ 0 t ( e − ∫ t ′ t ( λ ( a ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W ϵ ( a ) − e − ∫ t ′ t ( λ ( b ) ) − 1 ( τ ) d τ L ϵ ∂ σ 4 W ϵ ( b ) ) d t ′ Y 4 ≤ ( ϵ + T 1/4 ∣ log ϵ ∣ − 3/4 + T ) c ( Y ( a ) − Y ( b ) Y 1 W ϵ ( a ) Y 1 + W ϵ ( a ) − W ϵ ( b ) Y 1 ) ≤ ϵ − α t ∣ log ϵ ∣ 2 ( ϵ + T 1/4 ∣ log ϵ ∣ − 3/4 + T ) c ( M , ( X σ ) in ⋆ ) Y ( a ) − Y ( b ) Y 4 .
Altogether, we obtain
∥ Λ [ Y ( a ) ] − Λ [ Y ( b ) ] ∥ Y 4 ≤ ( ( T γ − α 3 + T 1 / 4 + T ) c ( ϵ , M , ∥ ( X σ ) i n ∥ ⋆ ) + ϵ 1 − α t ∣ log ϵ ∣ 2 c ( M , ∥ ( X σ ) i n ∥ ⋆ ) ) ∥ Y ( a ) − Y ( b ) ∥ Y 4 . \left\|\Lambda[\mathbf{Y}^{(a)}]-\Lambda[\mathbf{Y}^{(b)}]\right\|_{\mathcal{Y}_4}
\leq\left(\left(T^{\frac{\gamma-\alpha}{3}}+T^{1/4}+T\right)c(\epsilon,M,\left\|(\mathbf{X}^\sigma)^{\mathrm{in}}\right\|_{\star})+\epsilon^{1-\alpha t}|\log\epsilon|^2c(M,\left\|(\mathbf{X}^\sigma)^{\mathrm{in}}\right\|_{\star})\right)\left\|\mathbf{Y}^{(a)}-\mathbf{Y}^{(b)}\right\|_{\mathcal{Y}_4}. Λ [ Y ( a ) ] − Λ [ Y ( b ) ] Y 4 ≤ ( ( T 3 γ − α + T 1/4 + T ) c ( ϵ , M , ( X σ ) in ⋆ ) + ϵ 1 − α t ∣ log ϵ ∣ 2 c ( M , ( X σ ) in ⋆ ) ) Y ( a ) − Y ( b ) Y 4 .
Taking ϵ \epsilon ϵ small enough that ϵ 1 − α t ∣ log ϵ ∣ 2 c ( M , ∥ ( X σ ) i n ∥ ⋆ ) ≤ 1 3 \epsilon^{1-\alpha t}|\log\epsilon|^2c(M,\left\|(\mathbf{X}^\sigma)^{\mathrm{in}}\right\|_{\star})\leq\frac{1}{3} ϵ 1 − α t ∣ log ϵ ∣ 2 c ( M , ( X σ ) in ⋆ ) ≤ 3 1 in addition to (99), and then taking T T T small enough that ( T γ − α 3 + T 1 / 4 + T ) c ( ϵ , M , ∥ ( X σ ) i n ∥ ⋆ ) ≤ 1 3 \left(T^{\frac{\gamma-\alpha}{3}}+T^{1/4}+T\right)c(\epsilon,M,\left\|(\mathbf{X}^\sigma)^{\mathrm{in}}\right\|_{\star})\leq\frac{1}{3} ( T 3 γ − α + T 1/4 + T ) c ( ϵ , M , ( X σ ) in ⋆ ) ≤ 3 1 in addition to (100), we have that the map Λ \Lambda Λ admits a unique fixed point in B M , 1 2 ( T ) \mathcal{B}_{M,\frac{1}{2}}(T) B M , 2 1 ( T ) . □
Mapping properties about the straight cylinder
This section is devoted to the proofs of Lemmas 1.4 and 1.9 regarding the mapping properties of the slender body NtD and DtN operators L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ and L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 and the angle-averaged single layer operator A ‾ ϵ \overline{\mathcal{A}}_\epsilon A ϵ about the straight filament C ϵ \mathcal{C}_\epsilon C ϵ .
Preliminaries
As in the Laplace setting [59 section 2] , we will consider the Hölder spaces C k , α ( T ) C^{k,\alpha}(\mathbb{T}) C k , α ( T ) and C s k , α ( Γ ϵ ) C_s^{k,\alpha}(\Gamma_\epsilon) C s k , α ( Γ ϵ ) as subsets of C k , α ( R ) C^{k,\alpha}(\mathbb{R}) C k , α ( R ) and C s k , α ( R × 2 π T ) C_s^{k,\alpha}(\mathbb{R}\times2\pi\mathbb{T}) C s k , α ( R × 2 π T ) and will rely on a Besov space characterization of C k , α ( R ) C^{k,\alpha}(\mathbb{R}) C k , α ( R ) and C s k , α ( R × 2 π T ) C_s^{k,\alpha}(\mathbb{R}\times2\pi\mathbb{T}) C s k , α ( R × 2 π T ) .
Following [5 Chapter 2.2] , we consider a smooth cutoff function ϕ ( ξ ) \phi(\xi) ϕ ( ξ ) supported on the annulus { 3 4 < ∣ ξ ∣ < 2 } \left\{\frac{3}{4}<|\xi|<2\right\} { 4 3 < ∣ ξ ∣ < 2 } and satisfying ∑ j ∈ Z ϕ ( 2 − j ξ ) = 1 \sum_{j\in\mathbb{Z}}\phi(2^{-j}\xi)=1 ∑ j ∈ Z ϕ ( 2 − j ξ ) = 1 for ξ ≠ 0 \xi\ne0 ξ = 0 . We define
ϕ j ( ξ ) = ϕ ( 2 − j ξ ) (106) \phi_j(\xi)=\phi(2^{-j}\xi) \tag*{(106)} ϕ j ( ξ ) = ϕ ( 2 − j ξ ) (106)
and note that
supp ( ϕ j ) ∩ supp ( ϕ k ) = ∅ , ∣ j − k ∣ ≥ 2. \operatorname{supp}(\phi_j)\cap\operatorname{supp}(\phi_k)=\varnothing,\qquad|j-k|\geq2. supp ( ϕ j ) ∩ supp ( ϕ k ) = ∅ , ∣ j − k ∣ ≥ 2.
For a function g ( s , θ ) \mathbf{g}(s,\theta) g ( s , θ ) , we will use F [ g ] \mathcal{F}[\mathbf{g}] F [ g ] to denote the Fourier transform in s s s only:
F [ g ] ( ξ , θ ) = ∫ − ∞ ∞ g ( s , θ ) e − 2 π i ξ s d s . \mathcal{F}[\mathbf{g}](\xi,\theta)=\int_{-\infty}^{\infty}\mathbf{g}(s,\theta)e^{-2\pi i\xi s}\,ds. F [ g ] ( ξ , θ ) = ∫ − ∞ ∞ g ( s , θ ) e − 2 π i ξ s d s .
We define the Littlewood-Paley projection in s s s , P j g P_jg P j g , by
F [ P j g ] ( ξ ) = ϕ j ( ξ ) F [ g ] ( ξ , θ ) ; (107) \mathcal{F}[P_jg](\xi)=\phi_j(\xi)\mathcal{F}[g](\xi,\theta); \tag*{(107)} F [ P j g ] ( ξ ) = ϕ j ( ξ ) F [ g ] ( ξ , θ ) ; (107)
i.e. the projection of g g g onto frequencies in the s s s -variable which are supported within annulus j j j . In addition, we will use the notation
P ≤ j ∗ : = ∑ j ≤ j ∗ P j . P_{\leq j_*}:=\sum_{j\leq j_*}P_j. P ≤ j ∗ := j ≤ j ∗ ∑ P j .
Given ν ∈ R \nu\in\mathbb{R} ν ∈ R , we define the seminorm
∣ g ∣ ( B ˙ ∞ , ∞ ν ) s = sup j ∈ Z 2 j ν ∥ P j g ∥ L ∞ . |g|_{(\dot{B}^{\nu}_{\infty,\infty})_s}=\sup_{j\in\mathbb{Z}}2^{j\nu}\|P_jg\|_{L^\infty}. ∣ g ∣ ( B ˙ ∞ , ∞ ν ) s = j ∈ Z sup 2 j ν ∥ P j g ∥ L ∞ .
Note that if g g g is a function of s s s only, then this is just the Besov seminorm ∣ ⋅ ∣ B ˙ ∞ , ∞ ν ( R ) |\cdot|_{\dot{B}^{\nu}_{\infty,\infty}(\mathbb{R})} ∣ ⋅ ∣ B ˙ ∞ , ∞ ν ( R ) . We define the full norm
∥ g ∥ ( B ∞ , ∞ ν ) s = sup ( ∥ P ≤ 0 g ∥ L ∞ , sup j > 0 2 j ν ∥ P j g ∥ L ∞ ) . (108) \|g\|_{(B^{\nu}_{\infty,\infty})_s}=\sup\left(\|P_{\leq0}g\|_{L^\infty},\sup_{j>0}2^{j\nu}\|P_jg\|_{L^\infty}\right). \tag*{(108)} ∥ g ∥ ( B ∞ , ∞ ν ) s = sup ( ∥ P ≤ 0 g ∥ L ∞ , j > 0 sup 2 j ν ∥ P j g ∥ L ∞ ) . (108)
Letting ⌊ ν ⌋ \lfloor\nu\rfloor ⌊ ν ⌋ denote the integer part of ν \nu ν , it may be shown that the ( B ∞ , ∞ ν ) s (B^\nu_{\infty,\infty})_s ( B ∞ , ∞ ν ) s norm (107) is equivalent to the C s ⌊ ν ⌋ , ν − ⌊ ν ⌋ C_s^{\lfloor\nu\rfloor,\nu-\lfloor\nu\rfloor} C s ⌊ ν ⌋ , ν − ⌊ ν ⌋ Hölder norm (46). Furthermore, if g g g is a function of s s s only, then ∥ g ∥ ( B ∞ , ∞ ν ) s = ∥ g ∥ B ∞ , ∞ ν ( R ) \|g\|_{(B^\nu_{\infty,\infty})_s}=\|g\|_{B^\nu_{\infty,\infty}(\mathbb{R})} ∥ g ∥ ( B ∞ , ∞ ν ) s = ∥ g ∥ B ∞ , ∞ ν ( R ) is equivalent to ∥ g ∥ C ⌊ ν ⌋ , ν − ⌊ ν ⌋ ( R ) \|g\|_{C^{\lfloor\nu\rfloor,\nu-\lfloor\nu\rfloor}(\mathbb{R})} ∥ g ∥ C ⌊ ν ⌋ , ν − ⌊ ν ⌋ ( R ) .
Given a matrix-valued Fourier multiplier M ^ ( ξ ) \widehat{M}(\xi) M ( ξ ) , we denote
T M ^ g = F − 1 [ M ^ F [ g ] ] , P j T M ^ g = F − 1 [ ϕ j M ^ F [ g ] ] . (109) T_{\widehat{M}}g=\mathcal{F}^{-1}[\widehat{M}\mathcal{F}[g]],\qquad P_jT_{\widehat{M}}g=\mathcal{F}^{-1}[\phi_j\widehat{M}\mathcal{F}[g]]. \tag*{(109)} T M g = F − 1 [ M F [ g ]] , P j T M g = F − 1 [ ϕ j M F [ g ]] . (109)
For ϕ j \phi_j ϕ j as in (105), we may define M j = F − 1 [ ϕ j M ^ ] M_j=\mathcal{F}^{-1}[\phi_j\widehat{M}] M j = F − 1 [ ϕ j M ] so that we may write P j T M ^ g = M j ∗ g P_jT_{\widehat{M}}g=M_j*g P j T M g = M j ∗ g , where the convolution is with respect to s s s only:
M j ∗ g = ∫ R M j ( s − s ′ ) g ( s ′ , θ ) d s ′ . M_j*g=\int_{\mathbb{R}}M_j(s-s')g(s',\theta)\,ds'. M j ∗ g = ∫ R M j ( s − s ′ ) g ( s ′ , θ ) d s ′ .
By Young’s inequality for convolutions, we may estimate
∥ P j T M ^ g ∥ L ∞ = ∥ M j ∗ g ∥ L ∞ ≤ ∥ M j ∥ L 1 ∥ g ∥ L ∞ . \|P_jT_{\widehat{M}}g\|_{L^\infty}=\|M_j*g\|_{L^\infty}\leq\|M_j\|_{L^1}\|g\|_{L^\infty}. ∥ P j T M g ∥ L ∞ = ∥ M j ∗ g ∥ L ∞ ≤ ∥ M j ∥ L 1 ∥ g ∥ L ∞ .
Using the definition (107) of the ( B ∞ , ∞ ν ) s (B^\nu_{\infty,\infty})_s ( B ∞ , ∞ ν ) s norm and its equivalence with C s ⌊ ν ⌋ , ν − ⌊ ν ⌋ C_s^{\lfloor\nu\rfloor,\nu-\lfloor\nu\rfloor} C s ⌊ ν ⌋ , ν − ⌊ ν ⌋ , we may thus obtain the mapping properties of an operator T M ^ g T_{\widehat{M}}g T M g by obtaining L 1 L^1 L 1 bounds for the function M j M_j M j in physical space. One way to prove such bounds is via the following lemma, which is just the matrix-valued version of [59 Lemma 2.1] .
Lemma 3.1 (Physical space L 1 L^1 L 1 bounds for multipliers). For fixed j ∈ Z j\in\mathbb{Z} j ∈ Z , let M ^ j ( ξ ) ∈ C ∞ ( R ) \widehat{M}_j(\xi)\in C^\infty(\mathbb{R}) M j ( ξ ) ∈ C ∞ ( R ) be a matrix-valued Fourier multiplier supported in the annulus 2 j − 1 < ∣ ξ ∣ < 2 j + 1 2^{j-1}<|\xi|<2^{j+1} 2 j − 1 < ∣ ξ ∣ < 2 j + 1 and satisfying
∣ M ^ j ∣ ≤ A , ∣ ∂ ξ 2 M ^ j ∣ ≤ B (110) |\widehat{M}_j|\leq A,\qquad|\partial_\xi^2\widehat{M}_j|\leq B \tag*{(110)} ∣ M j ∣ ≤ A , ∣ ∂ ξ 2 M j ∣ ≤ B (110) for some numbers A A A , B B B . Then M j = F − 1 [ M ^ j ] M_j=\mathcal{F}^{-1}[\widehat{M}_j] M j = F − 1 [ M j ] satisfies the L 1 L^1 L 1 bound
∥ M j ∥ L 1 ( R ) ≲ 2 j A B . (111) \|M_j\|_{L^1(\mathbb{R})}\lesssim2^j\sqrt{AB}. \tag*{(111)} ∥ M j ∥ L 1 ( R ) ≲ 2 j A B . (111)
The strategy for showing the mapping properties in Lemmas 1.4 and 1.9 will thus be to prove L ∞ L^\infty L ∞ bounds for the zeroth and second derivatives of the components of the Fourier multipliers corresponding to L ‾ ϵ \overline{\mathcal{L}}_\epsilon L ϵ , L ‾ ϵ − 1 \overline{\mathcal{L}}_\epsilon^{-1} L ϵ − 1 , and A ‾ ϵ \overline{\mathcal{A}}_\epsilon A ϵ .
Bounds for SB DtN and NtD multipliers
We begin by considering the tangential and normal direction multipliers m e , t − 1 m_{\mathrm{e},t}^{-1} m e , t − 1 and m e , n − 1 m_{\mathrm{e},n}^{-1} m e , n − 1 corresponding to L e − 1 \mathcal{L}_{\mathrm{e}}^{-1} L e − 1 , given by Proposition 1.3. Given the form of these multipliers, it will be essential to have bounds for ratios of second-kind modified Bessel functions K 0 K_0 K 0 and K 1 K_1 K 1 as well as their derivatives. We will make use of the following proposition.
Proposition 3.2. Let K 0 ( z ) K_0(z) K 0 ( z ) and K 1 ( z ) K_1(z) K 1 ( z ) denote zeroth and first order modified Bessel functions of the second kind. The following bounds hold:
0 < K 1 ( z ) K 0 ( z ) − 1 − 1 2 z + 1 8 z 2 ≤ 1 8 z 3 , z ≥ 1 (112) 0 < \frac{K_1(z)}{K_0(z)} - 1 - \frac{1}{2z} + \frac{1}{8z^2} \leq\frac{1}{8z^3}, \qquad z \geq1 \tag*{(112)} 0 < K 0 ( z ) K 1 ( z ) − 1 − 2 z 1 + 8 z 2 1 ≤ 8 z 3 1 , z ≥ 1 (112) − 4 11 z 3 ≤ K 0 ( z ) K 1 ( z ) − 1 + 1 2 z − 3 8 z 2 < 0 , z ≥ 1 (113) -\frac{4}{11z^3} \leq\frac{K_0(z)}{K_1(z)} - 1 + \frac{1}{2z} - \frac{3}{8z^2} < 0, \qquad z \geq1 \tag*{(113)} − 11 z 3 4 ≤ K 1 ( z ) K 0 ( z ) − 1 + 2 z 1 − 8 z 2 3 < 0 , z ≥ 1 (113) c z ∣ log ( z / 3 ) ∣ < K 1 ( z ) K 0 ( z ) < c z ∣ log ( z / 2 ) ∣ , 0 < z < 1 (114) \frac{c}{z\lvert\log(z/3)\rvert} < \frac{K_1(z)}{K_0(z)} < \frac{c}{z\lvert\log(z/2)\rvert}, \qquad0 < z < 1 \tag*{(114)} z ∣ log ( z /3 )∣ c < K 0 ( z ) K 1 ( z ) < z ∣ log ( z /2 )∣ c , 0 < z < 1 (114) 0 ≤ K 1 K 0 − K 0 K 1 ≤ 1 z , all z > 0. (115) 0 \leq\frac{K_1}{K_0} - \frac{K_0}{K_1} \leq\frac{1}{z}, \qquad\text{all } z > 0. \tag*{(115)} 0 ≤ K 0 K 1 − K 1 K 0 ≤ z 1 , all z > 0. (115)
A proof of the quantitative bounds (111) and (112) for z ≥ 1 z \geq1 z ≥ 1 appears below. These quantitative bounds will be useful for bounding the complicated functions (23), (24) of the ratios K 1 K 0 \frac{K_1}{K_0} K 0 K 1 and K 0 K 1 \frac{K_0}{K_1} K 1 K 0 at intermediate z ∼ 1 z \sim1 z ∼ 1 . For small 0 < z < 1 0 < z < 1 0 < z < 1 , such quantitative bounds are not necessary, and (113) follows from the well-known small-z z z asymptotics of K 0 ( z ) K_0(z) K 0 ( z ) and K 1 ( z ) K_1(z) K 1 ( z ) [21 Chapter 10] . The general bound (114) is due to estimate (77) in [50 ] .
Noting that the first and second derivatives of the ratios in Proposition 3.2 are given by
( K 1 K 0 ) ′ = K 1 2 K 0 2 − K 1 z K 0 − 1 , ( K 1 K 0 ) ′ ′ = 2 K 1 K 0 ( K 1 K 0 ) ′ + 1 z 2 K 1 K 0 − 1 z ( K 1 K 0 ) ′ , ( K 0 K 1 ) ′ = K 0 2 K 1 2 + K 0 z K 1 − 1 , ( K 0 K 1 ) ′ ′ = 2 K 0 K 1 ( K 0 K 1 ) ′ − 1 z 2 K 0 K 1 + 1 z ( K 0 K 1 ) ′ , \begin{aligned}
\left(\frac{K_1}{K_0}\right)' &= \frac{K_1^2}{K_0^2} - \frac{K_1}{zK_0} - 1, &
\left(\frac{K_1}{K_0}\right)'' &= 2\frac{K_1}{K_0}\left(\frac{K_1}{K_0}\right)' + \frac{1}{z^2}\frac{K_1}{K_0} - \frac{1}{z}\left(\frac{K_1}{K_0}\right)', \\
\left(\frac{K_0}{K_1}\right)' &= \frac{K_0^2}{K_1^2} + \frac{K_0}{zK_1} - 1, &
\left(\frac{K_0}{K_1}\right)'' &= 2\frac{K_0}{K_1}\left(\frac{K_0}{K_1}\right)' - \frac{1}{z^2}\frac{K_0}{K_1} + \frac{1}{z}\left(\frac{K_0}{K_1}\right)' ,
\end{aligned} ( K 0 K 1 ) ′ ( K 1 K 0 ) ′ = K 0 2 K 1 2 − z K 0 K 1 − 1 , = K 1 2 K 0 2 + z K 1 K 0 − 1 , ( K 0 K 1 ) ′′ ( K 1 K 0 ) ′′ = 2 K 0 K 1 ( K 0 K 1 ) ′ + z 2 1 K 0 K 1 − z 1 ( K 0 K 1 ) ′ , = 2 K 1 K 0 ( K 1 K 0 ) ′ − z 2 1 K 1 K 0 + z 1 ( K 1 K 0 ) ′ ,
as an immediate consequence of Proposition 3.2, we obtain the following corollary.
Corollary 3.3. For z ≥ 1 z \geq1 z ≥ 1 , the following bounds hold for derivatives of ratios of K 1 ( z ) K_1(z) K 1 ( z ) and K 0 ( z ) K_0(z) K 0 ( z ) :
∣ ( K 1 K 0 ) ′ + 1 2 z 2 ∣ ≤ 3 4 z 3 , ∣ ( K 1 K 0 ) ′ ′ ∣ ≤ 9 2 z 3 , (116) \left|\left(\frac{K_1}{K_0}\right)' + \frac{1}{2z^2}\right| \leq\frac{3}{4z^3}, \qquad\left|\left(\frac{K_1}{K_0}\right)''\right| \leq\frac{9}{2z^3}, \tag*{(116)} ( K 0 K 1 ) ′ + 2 z 2 1 ≤ 4 z 3 3 , ( K 0 K 1 ) ′′ ≤ 2 z 3 9 , (116) ∣ ( K 0 K 1 ) ′ − 1 2 z 2 ∣ ≤ 2 z 3 , ∣ ( K 0 K 1 ) ′ ′ ∣ ≤ 25 2 z 3 . \left|\left(\frac{K_0}{K_1}\right)' - \frac{1}{2z^2}\right| \leq\frac{2}{z^3}, \qquad\left|\left(\frac{K_0}{K_1}\right)''\right| \leq\frac{25}{2z^3}. ( K 1 K 0 ) ′ − 2 z 2 1 ≤ z 3 2 , ( K 1 K 0 ) ′′ ≤ 2 z 3 25 . For 0 < z < 1 0 < z < 1 0 < z < 1 , we have the bounds
∣ ( K 1 K 0 ) ′ ∣ ≤ c z 2 ∣ log ( z / 2 ) ∣ , ∣ ( K 1 K 0 ) ′ ′ ∣ ≤ c z 3 ∣ log ( z / 2 ) ∣ , (117) \left|\left(\frac{K_1}{K_0}\right)'\right| \leq\frac{c}{z^2\lvert\log(z/2)\rvert}, \qquad\left|\left(\frac{K_1}{K_0}\right)''\right| \leq\frac{c}{z^3\lvert\log(z/2)\rvert}, \tag*{(117)} ( K 0 K 1 ) ′ ≤ z 2 ∣ log ( z /2 )∣ c , ( K 0 K 1 ) ′′ ≤ z 3 ∣ log ( z /2 )∣ c , (117) ∣ ( K 0 K 1 ) ′ ∣ ≤ c ∣ log ( z / 3 ) ∣ , ∣ ( K 0 K 1 ) ′ ′ ∣ ≤ c z ∣ log ( z / 3 ) ∣ . \left|\left(\frac{K_0}{K_1}\right)'\right| \leq c\lvert\log(z/3)\rvert, \qquad\left|\left(\frac{K_0}{K_1}\right)''\right| \leq\frac{c}{z}\lvert\log(z/3)\rvert. ( K 1 K 0 ) ′ ≤ c ∣ log ( z /3 )∣ , ( K 1 K 0 ) ′′ ≤ z c ∣ log ( z /3 )∣ .
Proof of Proposition 3.2. estimates (111) and (112). We first outline a general strategy for obtaining bounds of this form. Consider a function B ( z ) B(z) B ( z ) satisfying the ODE
B ′ ( z ) = F ( z , B ( z ) ) , B ( z 0 ) = B 0 (118) B'(z) = F(z,B(z)), \qquad B(z_0) = B_0 \tag*{(118)} B ′ ( z ) = F ( z , B ( z )) , B ( z 0 ) = B 0 (118)
for some F F F . To obtain an upper bound on B B B , we seek b ( z ) b(z) b ( z ) satisfying
b ′ ( z ) < F ( z , b ( z ) ) for z ≥ z 0 ≥ 0 ∃ z 1 > z 0 such that b ( z ) > B ( z ) for z ≥ z 1 . (119) \begin{aligned}
b'(z) &< F(z,b(z)) && \text{for } z \ge z_0 \ge0 \\
\exists z_1 &> z_0 \text{ such that } b(z)>B(z) && \text{for } z \ge z_1 . \tag*{(119)}
\end{aligned} b ′ ( z ) ∃ z 1 < F ( z , b ( z )) > z 0 such that b ( z ) > B ( z ) for z ≥ z 0 ≥ 0 for z ≥ z 1 . (119)
Then b ( z ) > B ( z ) b(z)>B(z) b ( z ) > B ( z ) for all z > z 0 z>z_0 z > z 0 . Similarly, to obtain a lower bound for B B B , we seek h ( z ) h(z) h ( z ) satisfying
h ′ ( z ) > F ( z , h ( z ) ) for z ≥ z 0 ≥ 0 ∃ z 1 > z 0 such that h ( z ) < B ( z ) for z ≥ z 1 . (120) \begin{aligned}
h'(z) &> F(z,h(z)) && \text{for } z \ge z_0 \ge0 \\
\exists z_1 &> z_0 \text{ such that } h(z)<B(z) && \text{for } z \ge z_1 . \tag*{(120)}
\end{aligned} h ′ ( z ) ∃ z 1 > F ( z , h ( z )) > z 0 such that h ( z ) < B ( z ) for z ≥ z 0 ≥ 0 for z ≥ z 1 . (120)
Then h ( z ) < B ( z ) h(z)<B(z) h ( z ) < B ( z ) for all z > z 0 z>z_0 z > z 0 .
Consider the function B K ( z ) = z K 1 ( z ) K 0 ( z ) − z − 1 2 + 1 8 z B_K(z)=z\frac{K_1(z)}{K_0(z)}-z-\frac{1}{2}+\frac{1}{8z} B K ( z ) = z K 0 ( z ) K 1 ( z ) − z − 2 1 + 8 z 1 . It can be seen that B K B_K B K satisfies the ODE
B K ′ ( z ) = 1 z ( B K 2 + B K − 1 4 z B K + 1 64 z 2 − 1 4 z ) + 2 B K , B K ( 1 ) = K 1 ( 1 ) K 0 ( 1 ) − 11 8 ≈ 0.055. (121) B_K'(z)=\frac{1}{z}\left(B_K^2+B_K-\frac{1}{4z}B_K+\frac{1}{64z^2}-\frac{1}{4z}\right)+2B_K,\qquad B_K(1)=\frac{K_1(1)}{K_0(1)}-\frac{11}{8}\approx0.055 . \tag*{(121)} B K ′ ( z ) = z 1 ( B K 2 + B K − 4 z 1 B K + 64 z 2 1 − 4 z 1 ) + 2 B K , B K ( 1 ) = K 0 ( 1 ) K 1 ( 1 ) − 8 11 ≈ 0.055. (121)
Let b ( z ) = 1 8 z 2 b(z)=\frac{1}{8z^2} b ( z ) = 8 z 2 1 . We have that
b ′ ( z ) − 1 z ( b 2 + b − 1 4 z b + 1 64 z 2 − 1 4 z ) − 2 b = − 1 64 z 5 + 1 32 z 4 − 25 64 z 3 < 0 b'(z)-\frac{1}{z}\left(b^2+b-\frac{1}{4z}b+\frac{1}{64z^2}-\frac{1}{4z}\right)-2b=-\frac{1}{64z^5}+\frac{1}{32z^4}-\frac{25}{64z^3}<0 b ′ ( z ) − z 1 ( b 2 + b − 4 z 1 b + 64 z 2 1 − 4 z 1 ) − 2 b = − 64 z 5 1 + 32 z 4 1 − 64 z 3 25 < 0
for z ≥ 1 z\ge1 z ≥ 1 . Using the large-z z z asymptotics of K 0 ( z ) K_0(z) K 0 ( z ) and K 1 ( z ) K_1(z) K 1 ( z ) [21 ] , we have
z K 1 ( z ) K 0 ( z ) = z + 1 2 − 1 8 z + 1 8 z 2 − 25 128 z 3 + O ( 1 z 4 ) , z → ∞ . (122) z\frac{K_1(z)}{K_0(z)}=z+\frac{1}{2}-\frac{1}{8z}+\frac{1}{8z^2}-\frac{25}{128z^3}+O\left(\frac{1}{z^4}\right),\qquad z\to\infty. \tag*{(122)} z K 0 ( z ) K 1 ( z ) = z + 2 1 − 8 z 1 + 8 z 2 1 − 128 z 3 25 + O ( z 4 1 ) , z → ∞. (122)
In particular, for sufficiently large z z z , we have B K ( z ) < b ( z ) B_K(z)<b(z) B K ( z ) < b ( z ) . Thus by (119) , B K ( z ) < b ( z ) B_K(z)<b(z) B K ( z ) < b ( z ) for all z ≥ 1 z\ge1 z ≥ 1 .
Now let h ( z ) = 0 h(z)=0 h ( z ) = 0 . We have
h ′ ( z ) − 1 z ( h 2 + h − 1 4 z h + 1 64 z 2 − 1 4 z ) − 2 h = 1 4 z 2 − 1 64 z 3 > 0 h'(z)-\frac{1}{z}\left(h^2+h-\frac{1}{4z}h+\frac{1}{64z^2}-\frac{1}{4z}\right)-2h=\frac{1}{4z^2}-\frac{1}{64z^3}>0 h ′ ( z ) − z 1 ( h 2 + h − 4 z 1 h + 64 z 2 1 − 4 z 1 ) − 2 h = 4 z 2 1 − 64 z 3 1 > 0
for z ≥ 1 z\ge1 z ≥ 1 . Again by the large-z z z asymptotics (122) for B K B_K B K , we have that for sufficiently large z z z , B K > 0 B_K>0 B K > 0 . By (120) , we thus have B K > 0 B_K>0 B K > 0 for all z ≥ 1 z\ge1 z ≥ 1 .
Note that using (112) we immediately obtain the bound (113) . In particular, we have
1 − 1 2 z + 3 8 z 2 − 4 11 z 3 ≤ 1 1 + 1 2 z − 1 8 z 2 + 1 8 z 3 ≤ K 0 K 1 ≤ 1 1 + 1 2 z − 1 8 z 2 ≤ 1 − 1 2 z + 3 8 z 2 . 1-\frac{1}{2z}+\frac{3}{8z^2}-\frac{4}{11z^3}\le\frac{1}{1+\frac{1}{2z}-\frac{1}{8z^2}+\frac{1}{8z^3}}\le\frac{K_0}{K_1}\le\frac{1}{1+\frac{1}{2z}-\frac{1}{8z^2}}\le1-\frac{1}{2z}+\frac{3}{8z^2}. 1 − 2 z 1 + 8 z 2 3 − 11 z 3 4 ≤ 1 + 2 z 1 − 8 z 2 1 + 8 z 3 1 1 ≤ K 1 K 0 ≤ 1 + 2 z 1 − 8 z 2 1 1 ≤ 1 − 2 z 1 + 8 z 2 3 .
We may use the bounds of Proposition 3.2 and Corollary 3.3 to obtain the following estimates for the tangential direction multiplier m ϵ , t − 1 m_{\epsilon,t}^{-1} m ϵ , t − 1 given by Proposition 1.3.
Lemma 3.4 (Bounds for m ϵ , t − 1 m_{\epsilon,t}^{-1} m ϵ , t − 1 and m ϵ , t m_{\epsilon,t} m ϵ , t ). Let m ϵ , t − 1 ( ξ ) m_{\epsilon,t}^{-1}(\xi) m ϵ , t − 1 ( ξ ) be as in Proposition 1.3. For ℓ = 0 , 1 , 2 \ell=0,1,2 ℓ = 0 , 1 , 2 , the following bounds hold:
∣ ∂ ξ ℓ m ϵ , t − 1 ( ξ ) ∣ ≤ { c ϵ ∣ ξ ∣ 1 − ℓ , ∣ ξ ∣ ≥ 1 2 π ϵ , c ∣ log ϵ ∣ − 1 ∣ ξ ∣ − ℓ , ∣ ξ ∣ < 1 2 π ϵ . (123) \left|\partial_\xi^\ell m_{\epsilon,t}^{-1}(\xi)\right|\le
\begin{cases}
c\epsilon|\xi|^{1-\ell}, & |\xi|\ge\frac{1}{2\pi\epsilon},\\
c|\log\epsilon|^{-1}|\xi|^{-\ell}, & |\xi|<\frac{1}{2\pi\epsilon}.
\end{cases} \tag*{(123)} ∂ ξ ℓ m ϵ , t − 1 ( ξ ) ≤ { cϵ ∣ ξ ∣ 1 − ℓ , c ∣ log ϵ ∣ − 1 ∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 2 π ϵ 1 , ∣ ξ ∣ < 2 π ϵ 1 . (123) Furthermore, we have that m ϵ , t ( ξ ) = 1 m ϵ , t − 1 ( ξ ) m_{\epsilon,t}(\xi)=\frac{1}{m_{\epsilon,t}^{-1}(\xi)} m ϵ , t ( ξ ) = m ϵ , t − 1 ( ξ ) 1 satisfies
∣ ∂ ξ ℓ m ϵ , t ( ξ ) ∣ ≤ { c ϵ − 1 ∣ ξ ∣ − ℓ − 1 , ∣ ξ ∣ ≥ 1 2 π ϵ , c ∣ log ϵ ∣ ∣ ξ ∣ − ℓ , ∣ ξ ∣ < 1 2 π ϵ . (124) \left|\partial_\xi^\ell m_{\epsilon,t}(\xi)\right|\le
\begin{cases}
c\epsilon^{-1}|\xi|^{-\ell-1}, & |\xi|\ge\frac{1}{2\pi\epsilon},\\
c|\log\epsilon||\xi|^{-\ell}, & |\xi|<\frac{1}{2\pi\epsilon}.
\end{cases} \tag*{(124)} ∂ ξ ℓ m ϵ , t ( ξ ) ≤ { c ϵ − 1 ∣ ξ ∣ − ℓ − 1 , c ∣ log ϵ ∣∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 2 π ϵ 1 , ∣ ξ ∣ < 2 π ϵ 1 . (124)
Proof of Lemma 3.4. Consider the function
m t − 1 ( z ) = m t 1 ( z ) m t 2 ( z ) , m t 1 ( z ) : = K 1 ( z ) z K 0 ( z ) , m t 2 ( z ) : = 1 + z 2 ( K 0 ( z ) K 1 ( z ) − K 1 ( z ) K 0 ( z ) ) (125) \begin{aligned}
m_t^{-1}(z) &= \frac{m_{t1}(z)}{m_{t2}(z)},\\
m_{t1}(z) &:= \frac{K_1(z)}{zK_0(z)},\qquad m_{t2}(z) := 1+\frac{z}{2}\left(\frac{K_0(z)}{K_1(z)}-\frac{K_1(z)}{K_0(z)}\right) \tag*{(125)}
\end{aligned} m t − 1 ( z ) m t 1 ( z ) = m t 2 ( z ) m t 1 ( z ) , := z K 0 ( z ) K 1 ( z ) , m t 2 ( z ) := 1 + 2 z ( K 1 ( z ) K 0 ( z ) − K 0 ( z ) K 1 ( z ) ) (125)
and note that m ϵ , t − 1 ( ξ ) = 2 π m t − 1 ( 2 π ϵ ∣ ξ ∣ ) m_{\epsilon,t}^{-1}(\xi)=2\pi m_t^{-1}(2\pi\epsilon|\xi|) m ϵ , t − 1 ( ξ ) = 2 π m t − 1 ( 2 π ϵ ∣ ξ ∣ ) for m ϵ , t ( ξ ) m_{\epsilon,t}(\xi) m ϵ , t ( ξ ) as in Proposition 1.3. We will additionally consider
m t ( z ) = 1 m t − 1 ( z ) = m t 2 ( z ) m t 1 ( z ) . (126) m_t(z)=\frac{1}{m_t^{-1}(z)}=\frac{m_{t2}(z)}{m_{t1}(z)}. \tag*{(126)} m t ( z ) = m t − 1 ( z ) 1 = m t 1 ( z ) m t 2 ( z ) . (126)
The first and second derivatives of m t − 1 ( z ) m_t^{-1}(z) m t − 1 ( z ) and m t ( z ) m_t(z) m t ( z ) satisfy
( m t − 1 ) ′ ( z ) = m t 1 ′ m t 2 − m t 1 m t 2 ′ m t 2 2 , ( m t − 1 ) ′ ′ ( z ) = m t 2 m t 1 ′ ′ − m t 1 m t 2 ′ ′ m t 2 2 + 2 m t 2 ′ m t 2 ( m t − 1 ) ′ , m t ′ ( z ) = m t 2 ′ m t 1 − m t 2 m t 1 ′ m t 1 2 , m t ′ ′ ( z ) = m t 1 m t 2 ′ ′ − m t 2 m t 1 ′ ′ m t 1 2 + 2 m t 1 ′ m t 1 m t ′ . (127) \begin{aligned}
(m_t^{-1})'(z) &= \frac{m_{t1}'m_{t2}-m_{t1}m_{t2}'}{m_{t2}^2},&
(m_t^{-1})''(z) &= \frac{m_{t2}m_{t1}''-m_{t1}m_{t2}''}{m_{t2}^2}+2\frac{m_{t2}'}{m_{t2}}(m_t^{-1})',\\
m_t'(z) &= \frac{m_{t2}'m_{t1}-m_{t2}m_{t1}'}{m_{t1}^2},&
m_t''(z) &= \frac{m_{t1}m_{t2}''-m_{t2}m_{t1}''}{m_{t1}^2}+2\frac{m_{t1}'}{m_{t1}}m_t'. \tag*{(127)}
\end{aligned} ( m t − 1 ) ′ ( z ) m t ′ ( z ) = m t 2 2 m t 1 ′ m t 2 − m t 1 m t 2 ′ , = m t 1 2 m t 2 ′ m t 1 − m t 2 m t 1 ′ , ( m t − 1 ) ′′ ( z ) m t ′′ ( z ) = m t 2 2 m t 2 m t 1 ′′ − m t 1 m t 2 ′′ + 2 m t 2 m t 2 ′ ( m t − 1 ) ′ , = m t 1 2 m t 1 m t 2 ′′ − m t 2 m t 1 ′′ + 2 m t 1 m t 1 ′ m t ′ . (127)
From Proposition 3.2, for z ≥ 1 z\ge1 z ≥ 1 we have
∣ m t 1 − z − 1 2 + 1 8 z ∣ ≤ 1 8 z 2 , ∣ m t 2 − 1 2 − 1 8 z ∣ ≤ 5 16 z 2 . (128) \left|m_{t1}-z-\frac{1}{2}+\frac{1}{8z}\right|\le\frac{1}{8z^2},\qquad\left|m_{t2}-\frac{1}{2}-\frac{1}{8z}\right|\le\frac{5}{16z^2}. \tag*{(128)} m t 1 − z − 2 1 + 8 z 1 ≤ 8 z 2 1 , m t 2 − 2 1 − 8 z 1 ≤ 16 z 2 5 . (128)
Furthermore, using Corollary 3.3, for z ≥ 1 z\ge1 z ≥ 1 the first and second derivatives of m t A m_{tA} m t A and m t B m_{tB} m tB satisfy
∣ m t 1 ′ − 1 ∣ = ∣ K 1 K 0 + z ( K 1 K 0 ) ′ − 1 ∣ ≤ 1 z 2 , ∣ m t 2 ′ ∣ = ∣ 1 2 ( K 0 K 1 − K 1 K 0 ) + z 2 ( K 0 K 1 − K 1 K 0 ) ′ ∣ ≤ 7 4 z 2 , ∣ m t 1 ′ ′ ∣ = ∣ 2 ( K 1 K 0 ) ′ + z ( K 1 K 0 ) ′ ′ ∣ ≤ 7 z 2 , ∣ m t 2 ′ ′ ∣ = ∣ ( K 0 K 1 − K 1 K 0 ) ′ + z 2 ( K 0 K 1 − K 1 K 0 ) ′ ′ ∣ ≤ 13 z 2 . (129) \begin{aligned}
|m_{t1}'-1|&=\left|\frac{K_1}{K_0}+z\left(\frac{K_1}{K_0}\right)'-1\right|\le\frac{1}{z^2},\\
|m_{t2}'|&=\left|\frac{1}{2}\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)'\right|\le\frac{7}{4z^2},\\
|m_{t1}''|&=\left|2\left(\frac{K_1}{K_0}\right)'+z\left(\frac{K_1}{K_0}\right)''\right|\le\frac{7}{z^2},\\
|m_{t2}''|&=\left|\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)'+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)''\right|\le\frac{13}{z^2}. \tag*{(129)}
\end{aligned} ∣ m t 1 ′ − 1∣ ∣ m t 2 ′ ∣ ∣ m t 1 ′′ ∣ ∣ m t 2 ′′ ∣ = K 0 K 1 + z ( K 0 K 1 ) ′ − 1 ≤ z 2 1 , = 2 1 ( K 1 K 0 − K 0 K 1 ) + 2 z ( K 1 K 0 − K 0 K 1 ) ′ ≤ 4 z 2 7 , = 2 ( K 0 K 1 ) ′ + z ( K 0 K 1 ) ′′ ≤ z 2 7 , = ( K 1 K 0 − K 0 K 1 ) ′ + 2 z ( K 1 K 0 − K 0 K 1 ) ′′ ≤ z 2 13 . (129)
Using (127) and (128) in the expressions (124), (125), and (126), we obtain the following bounds for m t − 1 m_t^{-1} m t − 1 and m t m_t m t when z ≥ 1 z\ge1 z ≥ 1 :
∣ m t − 1 ( z ) ∣ ≤ c z , ∣ ( m t − 1 ) ′ ( z ) ∣ ≤ c , ∣ ( m t − 1 ) ′ ′ ( z ) ∣ ≤ c z , ∣ m t ( z ) ∣ ≤ c z , ∣ m t ′ ( z ) ∣ ≤ c z 2 , ∣ m t ′ ′ ( z ) ∣ ≤ c z 3 . (130) \begin{aligned}
|m_t^{-1}(z)|&\le cz,& |(m_t^{-1})'(z)|&\le c,& |(m_t^{-1})''(z)|&\le\frac{c}{z},\\
|m_t(z)|&\le\frac{c}{z},& |m_t'(z)|&\le\frac{c}{z^2},& |m_t''(z)|&\le\frac{c}{z^3}. \tag*{(130)}
\end{aligned} ∣ m t − 1 ( z ) ∣ ∣ m t ( z ) ∣ ≤ cz , ≤ z c , ∣ ( m t − 1 ) ′ ( z ) ∣ ∣ m t ′ ( z ) ∣ ≤ c , ≤ z 2 c , ∣ ( m t − 1 ) ′′ ( z ) ∣ ∣ m t ′′ ( z ) ∣ ≤ z c , ≤ z 3 c . (130)
For 0 < z < 1 0<z<1 0 < z < 1 , we may use the bounds (113) and (114) to obtain
c ∣ log ( z / 3 ) ∣ ≤ m t 1 ( z ) ≤ c ∣ log ( z / 2 ) ∣ , 1 2 ≤ m t 2 ( z ) ≤ 1 , ∣ m t 1 ′ ( z ) ∣ ≤ c z ∣ log ( z / 2 ) ∣ , ∣ m t 2 ′ ( z ) ∣ ≤ c z , ∣ m t 1 ′ ′ ( z ) ∣ ≤ c z 2 ∣ log ( z / 2 ) ∣ , ∣ m t 2 ′ ′ ( z ) ∣ ≤ c z 2 . (131) \begin{aligned}
\frac{c}{|\log(z/3)|}&\le m_{t1}(z)\le\frac{c}{|\log(z/2)|},& \frac{1}{2}&\le m_{t2}(z)\le1,\\
|m_{t1}'(z)|&\le\frac{c}{z|\log(z/2)|},& |m_{t2}'(z)|&\le\frac{c}{z},\\
|m_{t1}''(z)|&\le\frac{c}{z^2|\log(z/2)|},& |m_{t2}''(z)|&\le\frac{c}{z^2}. \tag*{(131)}
\end{aligned} ∣ log ( z /3 ) ∣ c ∣ m t 1 ′ ( z ) ∣ ∣ m t 1 ′′ ( z ) ∣ ≤ m t 1 ( z ) ≤ ∣ log ( z /2 ) ∣ c , ≤ z ∣ log ( z /2 ) ∣ c , ≤ z 2 ∣ log ( z /2 ) ∣ c , 2 1 ∣ m t 2 ′ ( z ) ∣ ∣ m t 2 ′′ ( z ) ∣ ≤ m t 2 ( z ) ≤ 1 , ≤ z c , ≤ z 2 c . (131)
Using (130) in the expressions (124), (125), and (126), for 0 < z < 1 0<z<1 0 < z < 1 we obtain the bounds
∣ m t − 1 ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ , ∣ ( m t − 1 ) ′ ( z ) ∣ ≤ c z ∣ log ( z / 2 ) ∣ , ∣ ( m t − 1 ) ′ ′ ( z ) ∣ ≤ c z 2 ∣ log ( z / 2 ) ∣ , ∣ m t ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ , ∣ m t ′ ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ z , ∣ m t ′ ′ ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ z 2 . (132) \begin{aligned}
|m_{\mathrm{t}}^{-1}(z)|&\le\frac{c}{|\log(z/2)|},&
|(m_{\mathrm{t}}^{-1})'(z)|&\le\frac{c}{z|\log(z/2)|},&
|(m_{\mathrm{t}}^{-1})''(z)|&\le\frac{c}{z^2|\log(z/2)|},\\
|m_{\mathrm{t}}(z)|&\le c|\log(z/2)|,&
|m_{\mathrm{t}}'(z)|&\le\frac{c|\log(z/2)|}{z},&
|m_{\mathrm{t}}''(z)|&\le\frac{c|\log(z/2)|}{z^2}. \tag*{(132)}
\end{aligned} ∣ m t − 1 ( z ) ∣ ∣ m t ( z ) ∣ ≤ ∣ log ( z /2 ) ∣ c , ≤ c ∣ log ( z /2 ) ∣ , ∣ ( m t − 1 ) ′ ( z ) ∣ ∣ m t ′ ( z ) ∣ ≤ z ∣ log ( z /2 ) ∣ c , ≤ z c ∣ log ( z /2 ) ∣ , ∣ ( m t − 1 ) ′′ ( z ) ∣ ∣ m t ′′ ( z ) ∣ ≤ z 2 ∣ log ( z /2 ) ∣ c , ≤ z 2 c ∣ log ( z /2 ) ∣ . (132)
Combining the z ≥ 1 z\ge1 z ≥ 1 bounds (129) with the 0 < z < 1 0<z<1 0 < z < 1 bounds (131) and using that m ϵ , t − 1 ( ξ ) = 2 π m t − 1 ( 2 π ϵ ∣ ξ ∣ ) m_{\epsilon,\mathrm{t}}^{-1}(\xi)=2\pi m_{\mathrm{t}}^{-1}(2\pi\epsilon|\xi|) m ϵ , t − 1 ( ξ ) = 2 π m t − 1 ( 2 π ϵ ∣ ξ ∣ ) , we obtain Lemma 3.4. □
We next consider the normal direction multiplier m ϵ , n − 1 m_{\epsilon,\mathrm{n}}^{-1} m ϵ , n − 1 given by Proposition 1.3 and use the bounds of Proposition 3.2 and Corollary 3.3 to obtain the following lemma.
Lemma 3.5 (Bounds for m ϵ , n − 1 m_{\epsilon,\mathrm{n}}^{-1} m ϵ , n − 1 and m ϵ , n m_{\epsilon,\mathrm{n}} m ϵ , n ). Let m ϵ , n − 1 ( ξ ) m_{\epsilon,\mathrm{n}}^{-1}(\xi) m ϵ , n − 1 ( ξ ) be as in Proposition 1.3. For ℓ = 0 , 1 , 2 \ell=0,1,2 ℓ = 0 , 1 , 2 , the following bounds hold:
∣ ∂ ξ ℓ m ϵ , n − 1 ( ξ ) ∣ ≤ { c ϵ ∣ ξ ∣ 1 − ℓ , ∣ ξ ∣ ≥ 1 2 π ϵ , c ∣ log ϵ ∣ − 1 ∣ ξ ∣ − ℓ , ∣ ξ ∣ < 1 2 π ϵ . (133) \left|\partial_\xi^\ell m_{\epsilon,\mathrm{n}}^{-1}(\xi)\right|
\le
\begin{cases}
c\epsilon|\xi|^{1-\ell}, & |\xi|\ge\frac{1}{2\pi\epsilon},\\
c|\log\epsilon|^{-1}|\xi|^{-\ell}, & |\xi|<\frac{1}{2\pi\epsilon}.
\end{cases} \tag*{(133)} ∂ ξ ℓ m ϵ , n − 1 ( ξ ) ≤ { cϵ ∣ ξ ∣ 1 − ℓ , c ∣ log ϵ ∣ − 1 ∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 2 π ϵ 1 , ∣ ξ ∣ < 2 π ϵ 1 . (133) Furthermore, we have that m ϵ , n ( ξ ) = 1 m ϵ , n − 1 ( ξ ) m_{\epsilon,\mathrm{n}}(\xi)=\frac{1}{m_{\epsilon,\mathrm{n}}^{-1}(\xi)} m ϵ , n ( ξ ) = m ϵ , n − 1 ( ξ ) 1 satisfies
∣ ∂ ξ ℓ m ϵ , n ( ξ ) ∣ ≤ { c ϵ − 1 ∣ ξ ∣ − ℓ − 1 , ∣ ξ ∣ ≥ 1 2 π ϵ , c ∣ log ϵ ∣ ∣ ξ ∣ − ℓ , ∣ ξ ∣ < 1 2 π ϵ . (134) \left|\partial_\xi^\ell m_{\epsilon,\mathrm{n}}(\xi)\right|
\le
\begin{cases}
c\epsilon^{-1}|\xi|^{-\ell-1}, & |\xi|\ge\frac{1}{2\pi\epsilon},\\
c|\log\epsilon||\xi|^{-\ell}, & |\xi|<\frac{1}{2\pi\epsilon}.
\end{cases} \tag*{(134)} ∂ ξ ℓ m ϵ , n ( ξ ) ≤ { c ϵ − 1 ∣ ξ ∣ − ℓ − 1 , c ∣ log ϵ ∣∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 2 π ϵ 1 , ∣ ξ ∣ < 2 π ϵ 1 . (134)
Proof of Lemma 3.5. We consider the function
m n − 1 ( z ) = m n 1 ( z ) m n 2 ( z ) , m_{\mathrm{n}}^{-1}(z)=\frac{m_{\mathrm{n1}}(z)}{m_{\mathrm{n2}}(z)}, m n − 1 ( z ) = m n2 ( z ) m n1 ( z ) ,
m n 1 ( z ) = 2 + 8 z K 1 ( z ) K 0 ( z ) + z ( K 1 ( z ) K 0 ( z ) − K 0 ( z ) K 1 ( z ) ) , m n 2 ( z ) = 4 z 2 + 2 ( 1 − K 2 2 ( z ) K 1 2 ( z ) ) + 2 z ( K 1 ( z ) K 0 ( z ) − K 0 ( z ) K 1 ( z ) ) . (135) \begin{aligned}
m_{\mathrm{n1}}(z)&=2+\frac{8}{z}\frac{K_1(z)}{K_0(z)}+z\left(\frac{K_1(z)}{K_0(z)}-\frac{K_0(z)}{K_1(z)}\right),\\
m_{\mathrm{n2}}(z)&=\frac{4}{z^2}+2\left(1-\frac{K_2^2(z)}{K_1^2(z)}\right)+\frac{2}{z}\left(\frac{K_1(z)}{K_0(z)}-\frac{K_0(z)}{K_1(z)}\right). \tag*{(135)}
\end{aligned} m n1 ( z ) m n2 ( z ) = 2 + z 8 K 0 ( z ) K 1 ( z ) + z ( K 0 ( z ) K 1 ( z ) − K 1 ( z ) K 0 ( z ) ) , = z 2 4 + 2 ( 1 − K 1 2 ( z ) K 2 2 ( z ) ) + z 2 ( K 0 ( z ) K 1 ( z ) − K 1 ( z ) K 0 ( z ) ) . (135)
Using the Bessel function identity
K 2 ( z ) = K 0 ( z ) + 2 z K 1 ( z ) . (136) K_2(z)=K_0(z)+\frac{2}{z}K_1(z). \tag*{(136)} K 2 ( z ) = K 0 ( z ) + z 2 K 1 ( z ) . (136)
it can be seen that in fact m ϵ , n − 1 ( ξ ) = 2 π m n − 1 ( 2 π ϵ ∣ ξ ∣ ) m_{\epsilon,\mathrm{n}}^{-1}(\xi)=2\pi m_{\mathrm{n}}^{-1}(2\pi\epsilon|\xi|) m ϵ , n − 1 ( ξ ) = 2 π m n − 1 ( 2 π ϵ ∣ ξ ∣ ) where m n − 1 ( ξ ) m_{\mathrm{n}}^{-1}(\xi) m n − 1 ( ξ ) is given by Proposition 1.3. In addition, we define
m n ( z ) = 1 m n − 1 ( z ) = m n 2 ( z ) m n 1 ( z ) . (137) m_{\mathrm{n}}(z)=\frac{1}{m_{\mathrm{n}}^{-1}(z)}=\frac{m_{\mathrm{n2}}(z)}{m_{\mathrm{n1}}(z)}. \tag*{(137)} m n ( z ) = m n − 1 ( z ) 1 = m n1 ( z ) m n2 ( z ) . (137)
The first and second derivatives of m n − 1 m_{\mathrm{n}}^{-1} m n − 1 and m n m_{\mathrm{n}} m n then satisfy
( m n − 1 ) ′ ( z ) = m n 1 ′ ( z ) m n 2 ( z ) − m n 1 ( z ) m n 2 ′ ( z ) m n 2 2 ( z ) , ( m n − 1 ) ′ ′ ( z ) = m n 2 ( z ) m n 1 ′ ′ ( z ) − m n 1 ( z ) m n 2 ′ ′ ( z ) m n 2 2 ( z ) + 2 m n 2 ′ ( z ) m n 2 ( z ) ( m n − 1 ) ′ ( z ) , m n ′ ( z ) = m n 2 ′ ( z ) m n 1 ( z ) − m n 2 ( z ) m n 1 ′ ( z ) m n 1 2 ( z ) , m n ′ ′ ( z ) = m n 1 ( z ) m n 2 ′ ′ ( z ) − m n 2 ( z ) m n 1 ′ ′ ( z ) m n 1 2 ( z ) + 2 m n 1 ′ ( z ) m n 1 ( z ) m n ′ ( z ) . (138) \begin{aligned}
(m_{\mathrm{n}}^{-1})'(z)&=\frac{m_{\mathrm{n1}}'(z)m_{\mathrm{n2}}(z)-m_{\mathrm{n1}}(z)m_{\mathrm{n2}}'(z)}{m_{\mathrm{n2}}^2(z)},&
(m_{\mathrm{n}}^{-1})''(z)&=\frac{m_{\mathrm{n2}}(z)m_{\mathrm{n1}}''(z)-m_{\mathrm{n1}}(z)m_{\mathrm{n2}}''(z)}{m_{\mathrm{n2}}^2(z)}+2\frac{m_{\mathrm{n2}}'(z)}{m_{\mathrm{n2}}(z)}(m_{\mathrm{n}}^{-1})'(z),\\
m_{\mathrm{n}}'(z)&=\frac{m_{\mathrm{n2}}'(z)m_{\mathrm{n1}}(z)-m_{\mathrm{n2}}(z)m_{\mathrm{n1}}'(z)}{m_{\mathrm{n1}}^2(z)},&
m_{\mathrm{n}}''(z)&=\frac{m_{\mathrm{n1}}(z)m_{\mathrm{n2}}''(z)-m_{\mathrm{n2}}(z)m_{\mathrm{n1}}''(z)}{m_{\mathrm{n1}}^2(z)}+2\frac{m_{\mathrm{n1}}'(z)}{m_{\mathrm{n1}}(z)}m_{\mathrm{n}}'(z). \tag*{(138)}
\end{aligned} ( m n − 1 ) ′ ( z ) m n ′ ( z ) = m n2 2 ( z ) m n1 ′ ( z ) m n2 ( z ) − m n1 ( z ) m n2 ′ ( z ) , = m n1 2 ( z ) m n2 ′ ( z ) m n1 ( z ) − m n2 ( z ) m n1 ′ ( z ) , ( m n − 1 ) ′′ ( z ) m n ′′ ( z ) = m n2 2 ( z ) m n2 ( z ) m n1 ′′ ( z ) − m n1 ( z ) m n2 ′′ ( z ) + 2 m n2 ( z ) m n2 ′ ( z ) ( m n − 1 ) ′ ( z ) , = m n1 2 ( z ) m n1 ( z ) m n2 ′′ ( z ) − m n2 ( z ) m n1 ′′ ( z ) + 2 m n1 ( z ) m n1 ′ ( z ) m n ′ ( z ) . (138)
Using Proposition 3.2 and Corollary 3.3, for z ≥ 1 z \ge1 z ≥ 1 we may obtain the following bounds for the components m n A m_{n\mathrm{A}} m n A and m n B m_{n\mathrm{B}} m n B :
∣ m n 1 − 3 − 15 2 z ∣ ≤ ∣ 8 z ( 1 + 1 2 z − 1 8 z 2 + 1 8 z 3 ) − 8 z + ∣ z ( 1 z − 1 2 z 2 + 5 8 z 3 ) − 1 + 1 2 z ∣ ≤ 7 z 2 |m_{n1}-3-\frac{15}{2z}|\leq|\frac{8}{z}(1+\frac{1}{2z}-\frac{1}{8z^{2}}+\frac{1}{8z^{3}})-\frac{8}{z}+|z(\frac{1}{z}-\frac{1}{2z^{2}}+\frac{5}{8z^{3}})-1+\frac{1}{2z}|\leq\frac{7}{z^{2}} ∣ m n 1 − 3 − 2 z 15 ∣ ≤ ∣ z 8 ( 1 + 2 z 1 − 8 z 2 1 + 8 z 3 1 ) − z 8 + ∣ z ( z 1 − 2 z 2 1 + 8 z 3 5 ) − 1 + 2 z 1 ∣ ≤ z 2 7
∣ m n 2 − 2 z − 4 z 2 ∣ ≤ ∣ ( 1 z − 3 4 z 2 ) ( 2 − 1 2 z + 3 8 z 2 ) − 2 z + 2 z 2 ∣ + 1 z 3 + 5 4 z 4 ≤ 3 z 3 \left|m_{n2}-\frac{2}{z}-\frac{4}{z^{2}}\right|\leq\left|\left(\frac{1}{z}-\frac{3}{4z^{2}}\right)\left(2-\frac{1}{2z}+\frac{3}{8z^{2}}\right)-\frac{2}{z}+\frac{2}{z^{2}}\right|+\frac{1}{z^{3}}+\frac{5}{4z^{4}}\leq\frac{3}{z^{3}} m n 2 − z 2 − z 2 4 ≤ ( z 1 − 4 z 2 3 ) ( 2 − 2 z 1 + 8 z 2 3 ) − z 2 + z 2 2 + z 3 1 + 4 z 4 5 ≤ z 3 3
∣ m n 1 ′ ∣ = ∣ − 8 z 2 K 1 K 0 + 8 z ( K 1 K 0 ) ′ + ( K 1 K 0 − K 0 K 1 ) + z ( K 1 K 0 − K 0 K 1 ) ′ ∣ ≤ 32 z \left|m'_{n1}\right|=\left|-\frac{8}{z^{2}}\frac{K_{1}}{K_{0}}+\frac{8}{z}\left(\frac{K_{1}}{K_{0}}\right)'+\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)+z\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)'\right|\leq\frac{32}{z} ∣ m n 1 ′ ∣ = − z 2 8 K 0 K 1 + z 8 ( K 0 K 1 ) ′ + ( K 0 K 1 − K 1 K 0 ) + z ( K 0 K 1 − K 1 K 0 ) ′ ≤ z 32
∣ m n 2 ′ ∣ = ∣ − 8 z 3 − 4 K 0 K 1 ( K 0 K 1 ) ′ − 2 z 2 ( K 1 K 0 − K 0 K 1 ) + 2 z ( K 1 K 0 − K 0 K 1 ) ′ ∣ ≤ 47 z 2 (139) \left|m'_{n2}\right|=\left|-\frac{8}{z^{3}}-4\frac{K_{0}}{K_{1}}\left(\frac{K_{0}}{K_{1}}\right)'-\frac{2}{z^{2}}\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)+\frac{2}{z}\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)'\right|\leq\frac{47}{z^{2}} \tag*{(139)} ∣ m n 2 ′ ∣ = − z 3 8 − 4 K 1 K 0 ( K 1 K 0 ) ′ − z 2 2 ( K 0 K 1 − K 1 K 0 ) + z 2 ( K 0 K 1 − K 1 K 0 ) ′ ≤ z 2 47 (139)
∣ m n 1 ′ ′ ∣ = ∣ 16 z 3 K 1 K 0 − 16 z 2 ( K 1 K 0 ) ′ + 8 z ( K 1 K 0 ) ′ ′ + 2 ( K 1 K 0 − K 0 K 1 ) ′ + z ( K 1 K 0 − K 0 K 1 ) ′ ′ ∣ ≤ 103 z 2 \left|m''_{n1}\right|=\left|\frac{16}{z^{3}}\frac{K_{1}}{K_{0}}-\frac{16}{z^{2}}\left(\frac{K_{1}}{K_{0}}\right)'+\frac{8}{z}\left(\frac{K_{1}}{K_{0}}\right)''+2\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)'+z\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)''\right|\leq\frac{103}{z^{2}} ∣ m n 1 ′′ ∣ = z 3 16 K 0 K 1 − z 2 16 ( K 0 K 1 ) ′ + z 8 ( K 0 K 1 ) ′′ + 2 ( K 0 K 1 − K 1 K 0 ) ′ + z ( K 0 K 1 − K 1 K 0 ) ′′ ≤ z 2 103
∣ m n 2 ′ ′ ∣ = ∣ 24 z 4 − 4 ( K 0 K 1 ) ′ 2 − 4 K 0 K 1 ( K 0 K 1 ) ′ ′ + 4 z 3 ( K 1 K 0 − K 0 K 1 ) − 4 z 2 ( K 1 K 0 − K 0 K 1 ) ′ + 2 z ( K 1 K 0 − K 0 K 1 ) ′ ∣ ≤ 222 z 3 . \left|m''_{n2}\right|=\left|\frac{24}{z^{4}}-4\left(\frac{K_{0}}{K_{1}}\right)'^{2}-4\frac{K_{0}}{K_{1}}\left(\frac{K_{0}}{K_{1}}\right)''+\frac{4}{z^{3}}\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)-\frac{4}{z^{2}}\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)'+\frac{2}{z}\left(\frac{K_{1}}{K_{0}}-\frac{K_{0}}{K_{1}}\right)'\right|\leq\frac{222}{z^{3}}. ∣ m n 2 ′′ ∣ = z 4 24 − 4 ( K 1 K 0 ) ′ 2 − 4 K 1 K 0 ( K 1 K 0 ) ′′ + z 3 4 ( K 0 K 1 − K 1 K 0 ) − z 2 4 ( K 0 K 1 − K 1 K 0 ) ′ + z 2 ( K 0 K 1 − K 1 K 0 ) ′ ≤ z 3 222 .
In particular, using (138) in (137), for z ≥ 1 z\geq1 z ≥ 1 we have
∣ m n − 1 ( z ) ∣ ≤ c z , ∣ ( m n − 1 ) ′ ( z ) ∣ ≤ c , ∣ ( m n − 1 ) ′ ′ ( z ) ∣ ≤ c z \left|m_{n}^{-1}(z)\right|\leq cz,\qquad\left|\left(m_{n}^{-1}\right)'(z)\right|\leq c,\qquad\left|\left(m_{n}^{-1}\right)''(z)\right|\leq\frac{c}{z} m n − 1 ( z ) ≤ cz , ( m n − 1 ) ′ ( z ) ≤ c , ( m n − 1 ) ′′ ( z ) ≤ z c
∣ m n ( z ) ∣ ≤ c z , ∣ m n ′ ( z ) ∣ ≤ c z 2 , ∣ m n ′ ′ ( z ) ∣ ≤ c z 3 . (140) \left|m_{n}(z)\right|\leq\frac{c}{z},\qquad\left|m'_{n}(z)\right|\leq\frac{c}{z^{2}},\qquad\left|m''_{n}(z)\right|\leq\frac{c}{z^{3}}. \tag*{(140)} ∣ m n ( z ) ∣ ≤ z c , ∣ m n ′ ( z ) ∣ ≤ z 2 c , ∣ m n ′′ ( z ) ∣ ≤ z 3 c . (140)
For 0 < z < 1 0<z<1 0 < z < 1 , using Proposition 3.2, Corollary 3.3, and the forms of the derivatives in (138), we may calculate
c z 2 ∣ log ( z / 3 ) ∣ ≤ ∣ m n 1 ∣ ≤ c z 2 ∣ log ( z / 2 ) ∣ , 4 z 2 ≤ ∣ m n 2 ∣ ≤ 8 z 2 , \frac{c}{z^{2}\left|\log(z/3)\right|}\leq\left|m_{n1}\right|\leq\frac{c}{z^{2}\left|\log(z/2)\right|},\qquad\frac{4}{z^{2}}\leq\left|m_{n2}\right|\leq\frac{8}{z^{2}}, z 2 ∣ log ( z /3 ) ∣ c ≤ ∣ m n 1 ∣ ≤ z 2 ∣ log ( z /2 ) ∣ c , z 2 4 ≤ ∣ m n 2 ∣ ≤ z 2 8 ,
∣ m n 1 ′ ∣ ≤ c z 3 ∣ log ( z / 2 ) ∣ , ∣ m n 2 ′ ∣ ≤ c z 3 , \left|m'_{n1}\right|\leq\frac{c}{z^{3}\left|\log(z/2)\right|},\qquad\left|m'_{n2}\right|\leq\frac{c}{z^{3}}, ∣ m n 1 ′ ∣ ≤ z 3 ∣ log ( z /2 ) ∣ c , ∣ m n 2 ′ ∣ ≤ z 3 c ,
∣ m n 1 ′ ′ ∣ ≤ c z 4 ∣ log ( z / 2 ) ∣ , ∣ m n 2 ′ ′ ∣ ≤ c z 4 . (141) \left|m''_{n1}\right|\leq\frac{c}{z^{4}\left|\log(z/2)\right|},\qquad\left|m''_{n2}\right|\leq\frac{c}{z^{4}}. \tag*{(141)} ∣ m n 1 ′′ ∣ ≤ z 4 ∣ log ( z /2 ) ∣ c , ∣ m n 2 ′′ ∣ ≤ z 4 c . (141)
Using (140) with the forms of m n − 1 ( z ) m_{n}^{-1}(z) m n − 1 ( z ) , m n ( z ) m_{n}(z) m n ( z ) , and their derivatives (137), we obtain
∣ m n − 1 ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ , ∣ ( m n − 1 ) ′ ( z ) ∣ ≤ c z ∣ log ( z / 2 ) ∣ , ∣ ( m n − 1 ) ′ ′ ( z ) ∣ ≤ c z 2 ∣ log ( z / 2 ) ∣ \left|m_{n}^{-1}(z)\right|\leq\frac{c}{\left|\log(z/2)\right|},\qquad\left|\left(m_{n}^{-1}\right)'(z)\right|\leq\frac{c}{z\left|\log(z/2)\right|},\qquad\left|\left(m_{n}^{-1}\right)''(z)\right|\leq\frac{c}{z^{2}\left|\log(z/2)\right|} m n − 1 ( z ) ≤ ∣ log ( z /2 ) ∣ c , ( m n − 1 ) ′ ( z ) ≤ z ∣ log ( z /2 ) ∣ c , ( m n − 1 ) ′′ ( z ) ≤ z 2 ∣ log ( z /2 ) ∣ c
∣ m n ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ , ∣ m n ′ ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ z , ∣ m n ′ ′ ( z ) ∣ ≤ c ∣ log ( z / 2 ) ∣ z 2 . (142) \left|m_{n}(z)\right|\leq c\left|\log(z/2)\right|,\qquad\left|m'_{n}(z)\right|\leq\frac{c\left|\log(z/2)\right|}{z},\qquad\left|m''_{n}(z)\right|\leq\frac{c\left|\log(z/2)\right|}{z^{2}}. \tag*{(142)} ∣ m n ( z ) ∣ ≤ c ∣ log ( z /2 ) ∣ , ∣ m n ′ ( z ) ∣ ≤ z c ∣ log ( z /2 ) ∣ , ∣ m n ′′ ( z ) ∣ ≤ z 2 c ∣ log ( z /2 ) ∣ . (142)
Again combining the z ≥ 1 z\geq1 z ≥ 1 bounds (139) with the 0 < z < 1 0<z<1 0 < z < 1 bounds (141) and using that m e , n − 1 ( ξ ) = 2 π m n − 1 ( 2 π ϵ ∣ ξ ∣ ) m_{e,n}^{-1}(\xi)=2\pi m_{n}^{-1}(2\pi\epsilon|\xi|) m e , n − 1 ( ξ ) = 2 π m n − 1 ( 2 π ϵ ∣ ξ ∣ ) , we obtain Lemma 3.5.
Bounds for single layer multipliers
We now turn to the angle-averaged inverse single layer operator A e \mathcal{A}_{e} A e , defined in (56), and its corresponding multipliers, given by Corollary 1.8. To bound these multipliers, in addition to Proposition 3.2 and Corollary 3.3, we will need bounds on ratios of first-kind modified Bessel functions I 0 I_{0} I 0 and I 1 I_{1} I 1 .
We first make note of a useful identity that is a consequence of a more general property of the Wronskian between first and second kind modified Bessel functions [21 ] . In particular, first- and second-order modified Bessel functions of the first and second kind may be related via the following:
K 1 ( z ) I 0 ( z ) + K 0 ( z ) I 1 ( z ) = 1 z , z > 0. (143) K_1(z)I_0(z)+K_0(z)I_1(z)=\frac{1}{z},\qquad z>0. \tag*{(143)} K 1 ( z ) I 0 ( z ) + K 0 ( z ) I 1 ( z ) = z 1 , z > 0. (143)
In addition, we have the following proposition.
Proposition 3.6. Let I 0 ( z ) I_0(z) I 0 ( z ) and I 1 ( z ) I_1(z) I 1 ( z ) denote zeroth and first order modified Bessel functions of the first kind. The following bounds each hold for some c > 0 c>0 c > 0 :
∣ I 1 I 0 − 1 + 1 2 z + 1 8 z 2 ∣ ≤ c z 3 , z ≥ 1 (144) \left|\frac{I_1}{I_0}-1+\frac{1}{2z}+\frac{1}{8z^2}\right|\leq\frac{c}{z^3},\qquad z\geq1 \tag*{(144)} I 0 I 1 − 1 + 2 z 1 + 8 z 2 1 ≤ z 3 c , z ≥ 1 (144) ∣ I 0 I 1 − 1 − 1 2 z − 3 8 z 2 ∣ ≤ c z 3 , z ≥ 1 (145) \left|\frac{I_0}{I_1}-1-\frac{1}{2z}-\frac{3}{8z^2}\right|\leq\frac{c}{z^3},\qquad z\geq1 \tag*{(145)} I 1 I 0 − 1 − 2 z 1 − 8 z 2 3 ≤ z 3 c , z ≥ 1 (145) ∣ I 1 I 0 − z 2 ∣ ≤ c z 3 , 0 < z < 1 (146) \left|\frac{I_1}{I_0}-\frac{z}{2}\right|\leq cz^3,\qquad0<z<1 \tag*{(146)} I 0 I 1 − 2 z ≤ c z 3 , 0 < z < 1 (146) ∣ I 0 I 1 − 2 z − z 4 ∣ ≤ c z 3 , 0 < z < 1. (147) \left|\frac{I_0}{I_1}-\frac{2}{z}-\frac{z}{4}\right|\leq cz^3,\qquad0<z<1. \tag*{(147)} I 1 I 0 − z 2 − 4 z ≤ c z 3 , 0 < z < 1. (147)
Each of the bounds in Proposition 3.6 may be seen immediately using the large and small-z z z asymptotics of I 0 ( z ) I_0(z) I 0 ( z ) and I 1 ( z ) I_1(z) I 1 ( z ) ; in particular, from [21 Chap. 10] , we have
I 0 ( z ) ∼ e z 2 π z ( 1 + 1 8 z + 9 128 z 2 + 75 124 z 3 + ⋯ ) , I_0(z)\sim\frac{e^z}{\sqrt{2\pi z}}\left(1+\frac{1}{8z}+\frac{9}{128z^2}+\frac{75}{124z^3}+\cdots\right), I 0 ( z ) ∼ 2 π z e z ( 1 + 8 z 1 + 128 z 2 9 + 124 z 3 75 + ⋯ ) ,
I 1 ( z ) ∼ e z 2 π z ( 1 − 3 8 z − 15 128 z 2 − 105 1025 z 3 − ⋯ ) , z → ∞ ; (148) I_1(z)\sim\frac{e^z}{\sqrt{2\pi z}}\left(1-\frac{3}{8z}-\frac{15}{128z^2}-\frac{105}{1025z^3}-\cdots\right),\qquad z\to\infty; \tag*{(148)} I 1 ( z ) ∼ 2 π z e z ( 1 − 8 z 3 − 128 z 2 15 − 1025 z 3 105 − ⋯ ) , z → ∞ ; (148)
I 0 ( z ) ∼ 1 + z 2 4 + z 4 64 + ⋯ , I 1 ( z ) ∼ z 2 + z 3 16 + ⋯ , z → 0. I_0(z)\sim1+\frac{z^2}{4}+\frac{z^4}{64}+\cdots,\qquad I_1(z)\sim\frac{z}{2}+\frac{z^3}{16}+\cdots,\qquad z\to0. I 0 ( z ) ∼ 1 + 4 z 2 + 64 z 4 + ⋯ , I 1 ( z ) ∼ 2 z + 16 z 3 + ⋯ , z → 0.
Note that the bounds (143) and (144) are stated with an unknown constant c c c , unlike in Proposition 3.2 . Because of this, we will need an additional refinement in order to prove all of the bounds we will eventually need (see Proposition 3.8 ). Before stating and proving this refinement, however, we note an immediate corollary of Proposition 3.6 .
Calculating the first and second derivatives of the quantities in Proposition 3.6 yields
( I 1 I 0 ) ′ = − I 1 2 I 0 2 − I 1 z I 0 + 1 , ( I 1 I 0 ) ′ ′ = − 2 I 1 I 0 ( I 1 I 0 ) ′ + 1 z 2 I 1 I 0 − 1 z ( I 1 I 0 ) ′ \left(\frac{I_1}{I_0}\right)'=-\frac{I_1^2}{I_0^2}-\frac{I_1}{zI_0}+1,\qquad
\left(\frac{I_1}{I_0}\right)''=-2\frac{I_1}{I_0}\left(\frac{I_1}{I_0}\right)'+\frac{1}{z^2}\frac{I_1}{I_0}-\frac{1}{z}\left(\frac{I_1}{I_0}\right)' ( I 0 I 1 ) ′ = − I 0 2 I 1 2 − z I 0 I 1 + 1 , ( I 0 I 1 ) ′′ = − 2 I 0 I 1 ( I 0 I 1 ) ′ + z 2 1 I 0 I 1 − z 1 ( I 0 I 1 ) ′
( I 0 I 1 ) ′ = − I 0 2 I 1 2 + I 0 z I 1 + 1 , ( I 0 I 1 ) ′ ′ = − 2 I 0 I 1 ( I 0 I 1 ) ′ + 1 z 2 I 0 I 1 + 1 z ( I 0 I 1 ) ′ , \left(\frac{I_0}{I_1}\right)'=-\frac{I_0^2}{I_1^2}+\frac{I_0}{zI_1}+1,\qquad
\left(\frac{I_0}{I_1}\right)''=-2\frac{I_0}{I_1}\left(\frac{I_0}{I_1}\right)'+\frac{1}{z^2}\frac{I_0}{I_1}+\frac{1}{z}\left(\frac{I_0}{I_1}\right)', ( I 1 I 0 ) ′ = − I 1 2 I 0 2 + z I 1 I 0 + 1 , ( I 1 I 0 ) ′′ = − 2 I 1 I 0 ( I 1 I 0 ) ′ + z 2 1 I 1 I 0 + z 1 ( I 1 I 0 ) ′ ,
from which we may again immediately obtain the following.
Corollary 3.7. For z ≥ 1 z\geq1 z ≥ 1 , the following bounds hold for derivatives of ratios of I 1 ( z ) I_1(z) I 1 ( z ) and I 0 ( z ) I_0(z) I 0 ( z ) :
∣ ( I 1 I 0 ) ′ − 1 2 z 2 ∣ ≤ c z 3 , ∣ ( I 1 I 0 ) ′ ′ ∣ ≤ c z 3 \left|\left(\frac{I_1}{I_0}\right)'-\frac{1}{2z^2}\right|\leq\frac{c}{z^3},\qquad\left|\left(\frac{I_1}{I_0}\right)''\right|\leq\frac{c}{z^3} ( I 0 I 1 ) ′ − 2 z 2 1 ≤ z 3 c , ( I 0 I 1 ) ′′ ≤ z 3 c ∣ ( I 0 I 1 ) ′ + 1 2 z 2 ∣ ≤ c z 3 , ∣ ( I 0 I 1 ) ′ ′ ∣ ≤ c z 3 . (149) \left|\left(\frac{I_0}{I_1}\right)'+\frac{1}{2z^2}\right|\leq\frac{c}{z^3},\qquad\left|\left(\frac{I_0}{I_1}\right)''\right|\leq\frac{c}{z^3}. \tag*{(149)} ( I 1 I 0 ) ′ + 2 z 2 1 ≤ z 3 c , ( I 1 I 0 ) ′′ ≤ z 3 c . (149) For 0 < z < 1 0<z<1 0 < z < 1 , we have the bounds
∣ ( I 1 I 0 ) ′ − 1 2 ∣ ≤ c z 2 , ∣ ( I 1 I 0 ) ′ ′ ∣ ≤ c z \left|\left(\frac{I_1}{I_0}\right)'-\frac{1}{2}\right|\leq cz^2,\qquad\left|\left(\frac{I_1}{I_0}\right)''\right|\leq cz ( I 0 I 1 ) ′ − 2 1 ≤ c z 2 , ( I 0 I 1 ) ′′ ≤ cz ∣ ( I 0 I 1 ) ′ + 2 z 2 − 1 4 ∣ ≤ c z 2 , ∣ ( I 0 I 1 ) ′ ′ − 4 z 3 ∣ ≤ c z . (150) \left|\left(\frac{I_0}{I_1}\right)'+\frac{2}{z^2}-\frac{1}{4}\right|\leq cz^2,\qquad\left|\left(\frac{I_0}{I_1}\right)''-\frac{4}{z^3}\right|\leq cz. \tag*{(150)} ( I 1 I 0 ) ′ + z 2 2 − 4 1 ≤ c z 2 , ( I 1 I 0 ) ′′ − z 3 4 ≤ cz . (150)
Since we do not have precise values for the constants c c c appearing in (144) and (145) , we will additionally need the following refinement. Using the small-z z z asymptotics (146) and (147) , note that we may choose 0 < δ < 1 4 0<\delta<\frac{1}{4} 0 < δ < 4 1 small enough that
∣ I 1 ( z ) I 0 ( z ) − z 2 ∣ ≤ z 2 8 and ∣ I 0 ( z ) I 1 ( z ) − 2 z − z 4 ∣ ≤ z 2 8 , 0 < z ≤ δ . (151) \left|\frac{I_1(z)}{I_0(z)}-\frac{z}{2}\right|\leq\frac{z^2}{8}
\quad\text{and}\quad
\left|\frac{I_0(z)}{I_1(z)}-\frac{2}{z}-\frac{z}{4}\right|\leq\frac{z^2}{8},
\qquad0<z\leq\delta. \tag*{(151)} I 0 ( z ) I 1 ( z ) − 2 z ≤ 8 z 2 and I 1 ( z ) I 0 ( z ) − z 2 − 4 z ≤ 8 z 2 , 0 < z ≤ δ . (151)
With this choice of δ \delta δ , we may show the following proposition.
Proposition 3.8. Let δ \delta δ be as in (151) . For z ≥ δ z\geq\delta z ≥ δ , the following bound holds between the difference of ratios of first kind modified Bessel functions:
1 z ≤ I 0 ( z ) I 1 ( z ) − I 1 ( z ) I 0 ( z ) ≤ c δ z , c δ < 2. (152) \frac{1}{z}\leq\frac{I_0(z)}{I_1(z)}-\frac{I_1(z)}{I_0(z)}\leq\frac{c_\delta}{z},
\qquad c_\delta<2. \tag*{(152)} z 1 ≤ I 1 ( z ) I 0 ( z ) − I 0 ( z ) I 1 ( z ) ≤ z c δ , c δ < 2. (152) In addition, for z ≥ δ z\geq\delta z ≥ δ , we may show that the following function is bounded away from zero:
2 z + ( 1 − z I 0 ( z ) I 1 ( z ) ) ( I 0 ( z ) I 1 ( z ) − I 1 ( z ) I 0 ( z ) ) < − c (153) \frac{2}{z}+\left(1-z\frac{I_0(z)}{I_1(z)}\right)\left(\frac{I_0(z)}{I_1(z)}-\frac{I_1(z)}{I_0(z)}\right)<-c \tag*{(153)} z 2 + ( 1 − z I 1 ( z ) I 0 ( z ) ) ( I 1 ( z ) I 0 ( z ) − I 0 ( z ) I 1 ( z ) ) < − c (153) for some c > 0 c>0 c > 0 .
Proof. We consider the function
B ( z ) : = z ( I 0 ( z ) I 1 ( z ) − I 1 ( z ) I 0 ( z ) ) , (154) B(z):=z\left(\frac{I_0(z)}{I_1(z)}-\frac{I_1(z)}{I_0(z)}\right), \tag*{(154)} B ( z ) := z ( I 1 ( z ) I 0 ( z ) − I 0 ( z ) I 1 ( z ) ) , (154)
which satisfies the ODE
B ′ ( z ) = − B ( z ) ( I 0 ( z ) I 1 ( z ) + I 1 ( z ) I 0 ( z ) ) + 2 I 0 ( z ) I 1 ( z ) B'(z)=-B(z)\left(\frac{I_0(z)}{I_1(z)}+\frac{I_1(z)}{I_0(z)}\right)+2\frac{I_0(z)}{I_1(z)} B ′ ( z ) = − B ( z ) ( I 1 ( z ) I 0 ( z ) + I 0 ( z ) I 1 ( z ) ) + 2 I 1 ( z ) I 0 ( z )
with initial condition B ( δ ) = δ I 0 ( δ ) I 1 ( δ ) − I 1 ( δ ) I 0 ( δ ) B(\delta)=\delta\frac{I_0(\delta)}{I_1(\delta)}-\frac{I_1(\delta)}{I_0(\delta)} B ( δ ) = δ I 1 ( δ ) I 0 ( δ ) − I 0 ( δ ) I 1 ( δ ) . Using the bounds (151) , we have that
2 − δ 2 4 ( 1 + δ ) ≤ B ( δ ) ≤ 2 − δ 2 4 ( 1 − δ ) . (155) 2-\frac{\delta^2}{4}(1+\delta)\leq B(\delta)\leq2-\frac{\delta^2}{4}(1-\delta). \tag*{(155)} 2 − 4 δ 2 ( 1 + δ ) ≤ B ( δ ) ≤ 2 − 4 δ 2 ( 1 − δ ) . (155)
Noting that I 0 ( z ) > I 1 ( z ) I_0(z)>I_1(z) I 0 ( z ) > I 1 ( z ) for all z z z [21 ] , we have
B ′ ( z ) ≤ ( − B ( z ) + 2 ) I 0 ( z ) I 1 ( z ) , B'(z)\leq(-B(z)+2)\frac{I_0(z)}{I_1(z)}, B ′ ( z ) ≤ ( − B ( z ) + 2 ) I 1 ( z ) I 0 ( z ) ,
B ′ ( z ) ≥ ( − B ( z ) + 1 ) ( I 0 ( z ) I 1 ( z ) + I 1 ( z ) I 0 ( z ) ) . B'(z)\geq(-B(z)+1)\left(\frac{I_0(z)}{I_1(z)}+\frac{I_1(z)}{I_0(z)}\right). B ′ ( z ) ≥ ( − B ( z ) + 1 ) ( I 1 ( z ) I 0 ( z ) + I 0 ( z ) I 1 ( z ) ) .
Let b ( z ) b(z) b ( z ) be the solution to the ODE
b ′ ( z ) = ( − b ( z ) + 2 ) I 0 ( z ) I 1 ( z ) , b ( δ ) = B ( δ ) . b'(z)=(-b(z)+2)\frac{I_0(z)}{I_1(z)},\qquad b(\delta)=B(\delta). b ′ ( z ) = ( − b ( z ) + 2 ) I 1 ( z ) I 0 ( z ) , b ( δ ) = B ( δ ) .
Defining s = log ( z I 1 ( z ) ) s=\log(zI_1(z)) s = log ( z I 1 ( z )) , we have
d b d s = − b ( s ) + 2 , b ( s ( δ ) ) = B ( δ ) , \frac{db}{ds}=-b(s)+2,\qquad b(s(\delta))=B(\delta), d s d b = − b ( s ) + 2 , b ( s ( δ )) = B ( δ ) ,
which we can integrate directly to yield b ( s ) = 2 + ( B ( δ ) − 2 ) e − ( s − s ( δ ) ) b(s)=2+(B(\delta)-2)e^{-(s-s(\delta))} b ( s ) = 2 + ( B ( δ ) − 2 ) e − ( s − s ( δ )) , or
b ( s ( z ) ) = 2 + ( B ( δ ) − 2 ) δ I 1 ( δ ) z I 1 ( z ) . (156) b(s(z))=2+(B(\delta)-2)\frac{\delta I_1(\delta)}{zI_1(z)}. \tag*{(156)} b ( s ( z )) = 2 + ( B ( δ ) − 2 ) z I 1 ( z ) δ I 1 ( δ ) . (156)
Using the large-z z z asymptotics (144) and (145) of I 1 ( z ) I 0 ( z ) \frac{I_1(z)}{I_0(z)} I 0 ( z ) I 1 ( z ) and I 0 ( z ) I 1 ( z ) \frac{I_0(z)}{I_1(z)} I 1 ( z ) I 0 ( z ) , we note that there exists η > 1 \eta>1 η > 1 such that
B ( z ) ≤ 3 2 , z ≥ η . (157) B(z)\leq\frac{3}{2},\qquad z\geq\eta. \tag*{(157)} B ( z ) ≤ 2 3 , z ≥ η . (157)
Meanwhile, b ( s ( z ) ) → 2 b(s(z)) \to2 b ( s ( z )) → 2 as z → ∞ z \to\infty z → ∞ , so for z z z sufficiently large we have b ( s ( z ) ) > B ( z ) b(s(z)) > B(z) b ( s ( z )) > B ( z ) . Thus by (119) , we have B ( z ) ≤ b ( s ( z ) ) B(z) \le b(s(z)) B ( z ) ≤ b ( s ( z )) for z ≥ δ z \ge\delta z ≥ δ . In particular, for δ ≤ z ≤ η \delta\le z \le\eta δ ≤ z ≤ η , we have
B ( z ) ≤ b ( s ( η ) ) ≤ 2 − δ 8 η δ I 1 ( δ ) I 1 ( η ) < 2. (158) B(z) \le b(s(\eta)) \le2-\frac{\delta}{8\eta}\frac{\delta I_1(\delta)}{I_1(\eta)} < 2. \tag*{(158)} B ( z ) ≤ b ( s ( η )) ≤ 2 − 8 η δ I 1 ( η ) δ I 1 ( δ ) < 2. (158)
Combining (158) and (157) , we obtain the upper bound of (152) .
Similarly, let h ( z ) h(z) h ( z ) be the solution to the ODE
h ′ ( z ) = ( − h ( z ) + 1 ) ( I 0 ( z ) I 1 ( z ) + I 1 ( z ) I 0 ( z ) ) , h ( δ ) = B ( δ ) . h'(z)=(-h(z)+1)\left(\frac{I_0(z)}{I_1(z)}+\frac{I_1(z)}{I_0(z)}\right), \qquad h(\delta)=B(\delta). h ′ ( z ) = ( − h ( z ) + 1 ) ( I 1 ( z ) I 0 ( z ) + I 0 ( z ) I 1 ( z ) ) , h ( δ ) = B ( δ ) .
Defining t = log ( z I 1 ( z ) I 0 ( z ) ) t=\log(zI_1(z)I_0(z)) t = log ( z I 1 ( z ) I 0 ( z )) , we have
d h d t = − h ( t ) + 1 , h ( t ( δ ) ) = B ( δ ) . \frac{dh}{dt}=-h(t)+1,\qquad h(t(\delta))=B(\delta). d t d h = − h ( t ) + 1 , h ( t ( δ )) = B ( δ ) .
This can again be directly integrated to yield h ( t ) = 1 + ( B ( δ ) − 1 ) e − ( t − t ( δ ) ) h(t)=1+(B(\delta)-1)e^{-(t-t(\delta))} h ( t ) = 1 + ( B ( δ ) − 1 ) e − ( t − t ( δ )) , or
h ( t ( z ) ) = 1 + ( B ( δ ) − 1 ) δ I 0 ( δ ) I 1 ( δ ) z I 0 ( z ) I 1 ( z ) . (159) h(t(z))=1+(B(\delta)-1)\frac{\delta I_0(\delta)I_1(\delta)}{zI_0(z)I_1(z)}. \tag*{(159)} h ( t ( z )) = 1 + ( B ( δ ) − 1 ) z I 0 ( z ) I 1 ( z ) δ I 0 ( δ ) I 1 ( δ ) . (159)
Since both I 1 ( z ) I_1(z) I 1 ( z ) and I 0 ( z ) I_0(z) I 0 ( z ) grow exponentially in z z z [21 ] , for sufficiently large z z z , we have h ( t ( z ) ) < B ( z ) h(t(z))<B(z) h ( t ( z )) < B ( z ) . Thus by (120) , we have
B ( z ) ≥ h ( t ( z ) ) ≥ 1 , z ≥ δ . (160) B(z)\ge h(t(z))\ge1,\qquad z\ge\delta. \tag*{(160)} B ( z ) ≥ h ( t ( z )) ≥ 1 , z ≥ δ . (160)
This yields the lower bound of (152) .
To show the second bound (153) , we let
G ( z ) = 2 z + ( 1 − z I 0 ( z ) I 1 ( z ) ) ( I 0 ( z ) I 1 ( z ) − I 1 ( z ) I 0 ( z ) ) . G(z)=\frac{2}{z}+\left(1-z\frac{I_0(z)}{I_1(z)}\right)\left(\frac{I_0(z)}{I_1(z)}-\frac{I_1(z)}{I_0(z)}\right). G ( z ) = z 2 + ( 1 − z I 1 ( z ) I 0 ( z ) ) ( I 1 ( z ) I 0 ( z ) − I 0 ( z ) I 1 ( z ) ) .
For z ≥ 3 z\ge3 z ≥ 3 , since I 0 ( z ) I 1 ( z ) > 1 \frac{I_0(z)}{I_1(z)}>1 I 1 ( z ) I 0 ( z ) > 1 for all z z z , we may use (152) to immediately obtain
G ( z ) ≤ 2 z + ( 1 − z I 0 ( z ) I 1 ( z ) ) 1 z = 3 z − I 0 ( z ) I 1 ( z ) ≤ − c < 0. (161) G(z)\le\frac{2}{z}+\left(1-z\frac{I_0(z)}{I_1(z)}\right)\frac{1}{z}=\frac{3}{z}-\frac{I_0(z)}{I_1(z)}\le-c<0. \tag*{(161)} G ( z ) ≤ z 2 + ( 1 − z I 1 ( z ) I 0 ( z ) ) z 1 = z 3 − I 1 ( z ) I 0 ( z ) ≤ − c < 0. (161)
For δ ≤ z ≤ 3 \delta\le z\le3 δ ≤ z ≤ 3 , it suffices to show that H ( z ) = z G ( z ) H(z)=zG(z) H ( z ) = z G ( z ) satisfies H ( z ) ≤ − c H(z)\le-c H ( z ) ≤ − c for some c > 0 c>0 c > 0 . Writing
H ( z ) = 2 + ( 1 − z I 0 ( z ) I 1 ( z ) ) B ( z ) H(z)=2+\left(1-z\frac{I_0(z)}{I_1(z)}\right)B(z) H ( z ) = 2 + ( 1 − z I 1 ( z ) I 0 ( z ) ) B ( z )
for B ( z ) B(z) B ( z ) as in (154) , we may calculate that H ( z ) H(z) H ( z ) satisfies
H ′ ( z ) = ( − H ( z ) + B ( z ) ( 2 − B ( z ) 2 ) − 2 B ( z ) − 1 ) B ( z ) − 1 B ( z ) ( 2 I 0 ( z ) I 1 ( z ) + I 1 ( z ) I 0 ( z ) ) − I 1 ( z ) I 0 ( z ) ( 1 − B ( z ) 2 ) − z ( B ( z ) − 1 ) . H'(z)=\left(-H(z)+\frac{B(z)(2-\frac{B(z)}{2})-2}{B(z)-1}\right)\frac{B(z)-1}{B(z)}\left(2\frac{I_0(z)}{I_1(z)}+\frac{I_1(z)}{I_0(z)}\right)-\frac{I_1(z)}{I_0(z)}\left(1-\frac{B(z)}{2}\right)-z(B(z)-1). H ′ ( z ) = ( − H ( z ) + B ( z ) − 1 B ( z ) ( 2 − 2 B ( z ) ) − 2 ) B ( z ) B ( z ) − 1 ( 2 I 1 ( z ) I 0 ( z ) + I 0 ( z ) I 1 ( z ) ) − I 0 ( z ) I 1 ( z ) ( 1 − 2 B ( z ) ) − z ( B ( z ) − 1 ) .
Noting that B ( z ) ≤ c δ < 2 B(z)\le c_\delta<2 B ( z ) ≤ c δ < 2 implies B ( z ) ( 2 − B ( z ) 2 ) ≤ c < 2 B(z)(2-\frac{B(z)}{2})\le c<2 B ( z ) ( 2 − 2 B ( z ) ) ≤ c < 2 , we have B ( z ) ( 2 − B ( z ) 2 ) − 2 B ( z ) − 1 ≤ − c B \frac{B(z)(2-\frac{B(z)}{2})-2}{B(z)-1}\le-c_B B ( z ) − 1 B ( z ) ( 2 − 2 B ( z ) ) − 2 ≤ − c B for some constant c B > 0 c_B>0 c B > 0 . In particular,
H ′ ( z ) ≤ ( − H ( z ) − c B ) B ( z ) − 1 B ( z ) ( 2 I 0 ( z ) I 1 ( z ) + I 1 ( z ) I 0 ( z ) ) . H'(z)\le(-H(z)-c_B)\frac{B(z)-1}{B(z)}\left(2\frac{I_0(z)}{I_1(z)}+\frac{I_1(z)}{I_0(z)}\right). H ′ ( z ) ≤ ( − H ( z ) − c B ) B ( z ) B ( z ) − 1 ( 2 I 1 ( z ) I 0 ( z ) + I 0 ( z ) I 1 ( z ) ) .
Now, using the bounds (151) , for 0 < z ≤ δ 0<z\le\delta 0 < z ≤ δ , we have
∣ H ( z ) + z 2 4 ∣ ≤ 3 z 3 4 . \left|H(z)+\frac{z^2}{4}\right|\le\frac{3z^3}{4}. H ( z ) + 4 z 2 ≤ 4 3 z 3 .
In particular, H ( δ ) ≤ − δ 2 16 < 0 H(\delta)\le-\frac{\delta^2}{16}<0 H ( δ ) ≤ − 16 δ 2 < 0 , since δ ≤ 1 4 \delta\le\frac14 δ ≤ 4 1 . Letting s ( z ) s(z) s ( z ) be such that
s ′ ( z ) = B ( z ) − 1 B ( z ) ( 2 I 0 ( z ) I 1 ( z ) + I 1 ( z ) I 0 ( z ) ) > 0 s'(z)=\frac{B(z)-1}{B(z)}\left(2\frac{I_0(z)}{I_1(z)}+\frac{I_1(z)}{I_0(z)}\right)>0 s ′ ( z ) = B ( z ) B ( z ) − 1 ( 2 I 1 ( z ) I 0 ( z ) + I 0 ( z ) I 1 ( z ) ) > 0
for all z z z , we define w w w to be the solution to the ODE
d w d s = − w ( s ) − c B , w ( s ( δ ) ) = H ( δ ) . \frac{dw}{ds}=-w(s)-c_B,\qquad w(s(\delta))=H(\delta). d s d w = − w ( s ) − c B , w ( s ( δ )) = H ( δ ) .
We may solve for w w w explicitly as w ( s ) = − c B + ( c B + H ( δ ) ) e − s − s ( δ ) w(s)=-c_B+(c_B+H(\delta))e^{-s-s(\delta)} w ( s ) = − c B + ( c B + H ( δ )) e − s − s ( δ ) . Using the large-z z z asymptotics (144) and (145) of I 1 I 0 \frac{I_1}{I_0} I 0 I 1 and I 0 I 1 \frac{I_0}{I_1} I 1 I 0 , along with (119) , we thus have H ( z ) ≤ w ( s ( z ) ) ≤ − c < 0 H(z)\le w(s(z))\le-c<0 H ( z ) ≤ w ( s ( z )) ≤ − c < 0 for all z ≥ δ z\ge\delta z ≥ δ ; in particular, for δ ≤ z ≤ 3 \delta\le z\le3 δ ≤ z ≤ 3 . Combined with the bound (161) for z ≥ 3 z\ge3 z ≥ 3 , we obtain the bound (153) . □
Using Propositions 3.6 , 3.8 , and Corollary 3.7, we may obtain bounds for the tangential multipliers m ϵ , z z m^z_{\epsilon,z} m ϵ , z z and m ϵ , r z m^z_{\epsilon,r} m ϵ , r z for the angle-averaged single layer operator A ‾ ϵ \overline{A}_\epsilon A ϵ , defined in Corollary 1.8 .
Lemma 3.9 (Bounds for A ‾ ϵ \overline{A}_\epsilon A ϵ multipliers, tangential direction). Consider the multipliers m ϵ , z z ( ξ ) m^z_{\epsilon,z}(\xi) m ϵ , z z ( ξ ) and m ϵ , r z ( ξ ) m^z_{\epsilon,r}(\xi) m ϵ , r z ( ξ ) given by (58) in Corollary 1.8 . For ℓ = 0 , 1 , 2 \ell=0,1,2 ℓ = 0 , 1 , 2 , these multipliers satisfy:
∣ ∂ ϵ ℓ m ϵ , z z ( ξ ) ∣ ≤ { c ϵ ∣ ξ ∣ 1 − ℓ , ∣ ξ ∣ ≥ 1 2 π ϵ , c ∣ ξ ∣ − ℓ , ∣ ξ ∣ < 1 2 π ϵ , \left|\partial_\epsilon^\ell m^z_{\epsilon,z}(\xi)\right|
\le
\begin{cases}
c\epsilon|\xi|^{1-\ell}, & |\xi|\ge\frac{1}{2\pi\epsilon},\\
c|\xi|^{-\ell}, & |\xi|<\frac{1}{2\pi\epsilon},
\end{cases} ∂ ϵ ℓ m ϵ , z z ( ξ ) ≤ { cϵ ∣ ξ ∣ 1 − ℓ , c ∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 2 π ϵ 1 , ∣ ξ ∣ < 2 π ϵ 1 , ∣ ∂ ϵ ℓ m ϵ , r z ( ξ ) ∣ ≤ { c , ∣ ξ ∣ ≥ 1 2 π ϵ , ℓ = 0 , c ∣ ξ ∣ − 1 , ∣ ξ ∣ ≥ 1 2 π ϵ , ℓ = 1 , 2 , c ϵ − 1 ∣ ξ ∣ − 1 − ℓ , ∣ ξ ∣ < 1 2 π ϵ . (162) \left|\partial_\epsilon^\ell m^z_{\epsilon,r}(\xi)\right|
\le
\begin{cases}
c, & |\xi|\ge\frac{1}{2\pi\epsilon},\ \ell=0,\\
c|\xi|^{-1}, & |\xi|\ge\frac{1}{2\pi\epsilon},\ \ell=1,2,\\
c\epsilon^{-1}|\xi|^{-1-\ell}, & |\xi|<\frac{1}{2\pi\epsilon}.
\end{cases} \tag*{(162)} ∂ ϵ ℓ m ϵ , r z ( ξ ) ≤ ⎩ ⎨ ⎧ c , c ∣ ξ ∣ − 1 , c ϵ − 1 ∣ ξ ∣ − 1 − ℓ , ∣ ξ ∣ ≥ 2 π ϵ 1 , ℓ = 0 , ∣ ξ ∣ ≥ 2 π ϵ 1 , ℓ = 1 , 2 , ∣ ξ ∣ < 2 π ϵ 1 . (162)
Proof. We begin by defining
m t A ( z ) = Q t A ( z ) Q t D ( z ) , m t B ( z ) = Q t B ( z ) Q t D ( z ) . m_{tA}(z)=\frac{Q_{tA}(z)}{Q_{tD}(z)},\qquad m_{tB}(z)=\frac{Q_{tB}(z)}{Q_{tD}(z)}. m t A ( z ) = Q t D ( z ) Q t A ( z ) , m tB ( z ) = Q t D ( z ) Q tB ( z ) .
where Q t j Q_{tj} Q t j are as in (50) of Lemma 1.7. Note that m t A ( 2 π ϵ ∣ ξ ∣ ) = 1 2 m ϵ , z z ( ξ ) m_{tA}(2\pi\epsilon|\xi|)=\frac12m^z_{\epsilon,z}(\xi) m t A ( 2 π ϵ ∣ ξ ∣ ) = 2 1 m ϵ , z z ( ξ ) and m t B ( 2 π ϵ ∣ ξ ∣ ) = 1 2 m ϵ , r z ( ξ ) m_{tB}(2\pi\epsilon|\xi|)=\frac12m^z_{\epsilon,r}(\xi) m tB ( 2 π ϵ ∣ ξ ∣ ) = 2 1 m ϵ , r z ( ξ ) . For both j = A , B j=A,B j = A , B , we have
m t j ′ ( z ) = Q t j ′ Q t D − Q t j Q t D ′ Q t D 2 , m t j ′ ′ ( z ) = Q t D Q t j ′ ′ − Q t j Q t D ′ ′ Q t D 2 + 2 Q t D ′ Q t D m t j ′ . (163) m'_{tj}(z)=\frac{Q'_{tj}Q_{tD}-Q_{tj}Q'_{tD}}{Q_{tD}^2},\qquad
m''_{tj}(z)=\frac{Q_{tD}Q''_{tj}-Q_{tj}Q''_{tD}}{Q_{tD}^2}+2\frac{Q'_{tD}}{Q_{tD}}m'_{tj}. \tag*{(163)} m t j ′ ( z ) = Q t D 2 Q t j ′ Q t D − Q t j Q t D ′ , m t j ′′ ( z ) = Q t D 2 Q t D Q t j ′′ − Q t j Q t D ′′ + 2 Q t D Q t D ′ m t j ′ . (163)
It thus remains to bound each of Q t A Q_{tA} Q t A , Q t B Q_{tB} Q tB , Q t D Q_{tD} Q t D , and their derivatives.
We begin with Q t D Q_{tD} Q t D . Using Proposition 3.8 and Proposition 3.2 , we may obtain a lower bound for Q t D ( z ) Q_{tD}(z) Q t D ( z ) when z ≥ 1 z\ge1 z ≥ 1 : in particular,
Q t D ( z ) = ( 1 + z 2 ( I 1 I 0 − I 0 I 1 ) ) ( 1 + z 2 ( K 0 K 1 − K 1 K 0 ) ) ≥ c > 0 , z ≥ 1. (164) Q_{tD}(z)=\left(1+\frac{z}{2}\left(\frac{I_1}{I_0}-\frac{I_0}{I_1}\right)\right)\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)\right)\ge c>0,\qquad z\ge1. \tag*{(164)} Q t D ( z ) = ( 1 + 2 z ( I 0 I 1 − I 1 I 0 ) ) ( 1 + 2 z ( K 1 K 0 − K 0 K 1 ) ) ≥ c > 0 , z ≥ 1. (164)
Furthermore, using Propositions 3.2 and 3.6 , for z ≥ 1 z \ge1 z ≥ 1 we have
∣ Q t D ( z ) − 1 4 ∣ ≤ c z 2 , \left|Q_{\mathrm{tD}}(z)-\frac{1}{4}\right|\le\frac{c}{z^2}, Q tD ( z ) − 4 1 ≤ z 2 c ,
∣ Q t D ′ ( z ) ∣ = ∣ 1 2 ( I 1 I 0 − I 0 I 1 ) + z 2 ( I 1 I 0 − I 0 I 1 ) ′ ) ( 1 + z 2 ( K 0 K 1 − K 1 K 0 ) ) + ( 1 + z 2 ( I 1 I 0 − I 0 I 1 ) ) ( 1 2 ( K 0 K 1 − K 1 K 0 ) + z 2 ( K 0 K 1 − K 1 K 0 ) ′ ) ∣ ≤ c z 2 (165) |Q_{\mathrm{tD}}'(z)|=|\frac{1}{2}(\frac{I_1}{I_0}-\frac{I_0}{I_1})+\frac{z}{2}(\frac{I_1}{I_0}-\frac{I_0}{I_1})')(1+\frac{z}{2}(\frac{K_0}{K_1}-\frac{K_1}{K_0}))
+(1+\frac{z}{2}(\frac{I_1}{I_0}-\frac{I_0}{I_1}))(\frac{1}{2}(\frac{K_0}{K_1}-\frac{K_1}{K_0})+\frac{z}{2}(\frac{K_0}{K_1}-\frac{K_1}{K_0})')|\le\frac{c}{z^2} \tag*{(165)} ∣ Q tD ′ ( z ) ∣ = ∣ 2 1 ( I 0 I 1 − I 1 I 0 ) + 2 z ( I 0 I 1 − I 1 I 0 ) ′ ) ( 1 + 2 z ( K 1 K 0 − K 0 K 1 )) + ( 1 + 2 z ( I 0 I 1 − I 1 I 0 )) ( 2 1 ( K 1 K 0 − K 0 K 1 ) + 2 z ( K 1 K 0 − K 0 K 1 ) ′ ) ∣ ≤ z 2 c (165)
∣ Q t D ′ ′ ( z ) ∣ = ∣ ( I 1 I 0 − I 0 I 1 ) ′ + z 2 ( I 1 I 0 − I 0 I 1 ) ′ ′ ) ( 1 + z 2 ( K 0 K 1 − K 1 K 0 ) ) + ( I 1 I 0 − I 0 I 1 + z ( I 1 I 0 − I 0 I 1 ) ′ ) ( 1 2 ( K 0 K 1 − K 1 K 0 ) + z 2 ( K 0 K 1 − K 1 K 0 ) ′ ) + ( 1 + z 2 ( I 1 I 0 − I 0 I 1 ) ) ( ( K 0 K 1 − K 1 K 0 ) ′ + z 2 ( K 0 K 1 − K 1 K 0 ) ′ ′ ) ∣ ≤ c z 2 . |Q_{\mathrm{tD}}''(z)|=|(\frac{I_1}{I_0}-\frac{I_0}{I_1})'+\frac{z}{2}(\frac{I_1}{I_0}-\frac{I_0}{I_1})'')(1+\frac{z}{2}(\frac{K_0}{K_1}-\frac{K_1}{K_0}))
+(\frac{I_1}{I_0}-\frac{I_0}{I_1}+z(\frac{I_1}{I_0}-\frac{I_0}{I_1})')(\frac{1}{2}(\frac{K_0}{K_1}-\frac{K_1}{K_0})+\frac{z}{2}(\frac{K_0}{K_1}-\frac{K_1}{K_0})')
+(1+\frac{z}{2}(\frac{I_1}{I_0}-\frac{I_0}{I_1}))((\frac{K_0}{K_1}-\frac{K_1}{K_0})'+\frac{z}{2}(\frac{K_0}{K_1}-\frac{K_1}{K_0})'')|\le\frac{c}{z^2}. ∣ Q tD ′′ ( z ) ∣ = ∣ ( I 0 I 1 − I 1 I 0 ) ′ + 2 z ( I 0 I 1 − I 1 I 0 ) ′′ ) ( 1 + 2 z ( K 1 K 0 − K 0 K 1 )) + ( I 0 I 1 − I 1 I 0 + z ( I 0 I 1 − I 1 I 0 ) ′ ) ( 2 1 ( K 1 K 0 − K 0 K 1 ) + 2 z ( K 1 K 0 − K 0 K 1 ) ′ ) + ( 1 + 2 z ( I 0 I 1 − I 1 I 0 )) (( K 1 K 0 − K 0 K 1 ) ′ + 2 z ( K 1 K 0 − K 0 K 1 ) ′′ ) ∣ ≤ z 2 c .
In addition, for 0 < z < 1 0<z<1 0 < z < 1 , using cancellations in the asymptotics (147) , (150) for I 0 I 1 \frac{I_0}{I_1} I 1 I 0 , we have
z 2 c ≤ ∣ Q t D ( z ) ∣ ≤ c z 2 , ∣ Q t D ′ ( z ) ∣ ≤ c z , ∣ Q t D ′ ′ ( z ) ∣ ≤ c . (166) \frac{z^2}{c}\le\left|Q_{\mathrm{tD}}(z)\right|\le cz^2,\qquad\left|Q_{\mathrm{tD}}'(z)\right|\le cz,\qquad\left|Q_{\mathrm{tD}}''(z)\right|\le c. \tag*{(166)} c z 2 ≤ ∣ Q tD ( z ) ∣ ≤ c z 2 , ∣ Q tD ′ ( z ) ∣ ≤ cz , ∣ Q tD ′′ ( z ) ∣ ≤ c . (166)
For Q t A ( z ) Q_{\mathrm{tA}}(z) Q tA ( z ) and Q t B ( z ) Q_{\mathrm{tB}}(z) Q tB ( z ) , we first note that, using the identity (143) , we may write
1 I 0 K 0 = z ( I 1 I 0 + K 1 K 0 ) , 1 I 1 K 1 = z ( I 0 I 1 + K 0 K 1 ) . (167) \frac{1}{I_0K_0}=z\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right),\qquad\frac{1}{I_1K_1}=z\left(\frac{I_0}{I_1}+\frac{K_0}{K_1}\right). \tag*{(167)} I 0 K 0 1 = z ( I 0 I 1 + K 0 K 1 ) , I 1 K 1 1 = z ( I 1 I 0 + K 1 K 0 ) . (167)
For z ≥ 1 z\ge1 z ≥ 1 , we may then use Propositions 3.6 and 3.2 to obtain
∣ Q t A ( z ) − z ∣ = ∣ z ( I 1 I 0 + K 1 K 0 ) ( 1 + z 2 ( K 0 K 1 − I 0 I 1 ) ) − z ∣ ≤ c z \left|Q_{\mathrm{tA}}(z)-z\right|=\left|z\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)\right)-z\right|\le\frac{c}{z} ∣ Q tA ( z ) − z ∣ = z ( I 0 I 1 + K 0 K 1 ) ( 1 + 2 z ( K 1 K 0 − I 1 I 0 ) ) − z ≤ z c
∣ Q t A ′ ( z ) − 1 ∣ = ∣ ( I 1 I 0 + K 1 K 0 ) ( 1 + z 2 ( K 0 K 1 − I 0 I 1 ) ) + z ( I 1 I 0 + K 1 K 0 ) ′ ( 1 + z 2 ( K 0 K 1 − I 0 I 1 ) ) + z ( I 1 I 0 + K 1 K 0 ) ( 1 2 ( K 0 K 1 − I 0 I 1 ) + z 2 ( K 0 K 1 − I 0 I 1 ) ′ ) − 1 ∣ ≤ c z \left|Q_{\mathrm{tA}}'(z)-1\right|=\left|\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)\right)+z\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)'\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)\right)
+z\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)\left(\frac{1}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)'\right)-1\right|\le\frac{c}{z} ∣ Q tA ′ ( z ) − 1 ∣ = ( I 0 I 1 + K 0 K 1 ) ( 1 + 2 z ( K 1 K 0 − I 1 I 0 ) ) + z ( I 0 I 1 + K 0 K 1 ) ′ ( 1 + 2 z ( K 1 K 0 − I 1 I 0 ) ) + z ( I 0 I 1 + K 0 K 1 ) ( 2 1 ( K 1 K 0 − I 1 I 0 ) + 2 z ( K 1 K 0 − I 1 I 0 ) ′ ) − 1 ≤ z c
∣ Q t A ′ ′ ( z ) ∣ = ∣ 2 ( I 1 I 0 + K 1 K 0 ) ′ ( 1 + z 2 ( K 0 K 1 − I 0 I 1 ) ) + 2 ( I 1 I 0 + K 1 K 0 ) ( 1 2 ( K 0 K 1 − I 0 I 1 ) + z 2 ( K 0 K 1 − I 0 I 1 ) ′ ) + z ( I 1 I 0 + K 1 K 0 ) ′ ′ ( 1 + z 2 ( K 0 K 1 − I 0 I 1 ) ) + 2 z ( I 1 I 0 + K 1 K 0 ) ′ ( 1 2 ( K 0 K 1 − I 0 I 1 ) + z 2 ( K 0 K 1 − I 0 I 1 ) ′ ) + z ( I 1 I 0 + K 1 K 0 ) ( ( K 0 K 1 − I 0 I 1 ) ′ + z 2 ( K 0 K 1 − I 0 I 1 ) ′ ′ ) ∣ ≤ c z . (168) \left|Q_{\mathrm{tA}}''(z)\right|=\left|2\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)'\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)\right)+2\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)\left(\frac{1}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)'\right)
+z\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)''\left(1+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)\right)+2z\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)'\left(\frac{1}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)'\right)
+z\left(\frac{I_1}{I_0}+\frac{K_1}{K_0}\right)\left(\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)'+\frac{z}{2}\left(\frac{K_0}{K_1}-\frac{I_0}{I_1}\right)''\right)\right|\le\frac{c}{z}. \tag*{(168)} ∣ Q tA ′′ ( z ) ∣ = 2 ( I 0 I 1 + K 0 K 1 ) ′ ( 1 + 2 z ( K 1 K 0 − I 1 I 0 ) ) + 2 ( I 0 I 1 + K 0 K 1 ) ( 2 1 ( K 1 K 0 − I 1 I 0 ) + 2 z ( K 1 K 0 − I 1 I 0 ) ′ ) + z ( I 0 I 1 + K 0 K 1 ) ′′ ( 1 + 2 z ( K 1 K 0 − I 1 I 0 ) ) + 2 z ( I 0 I 1 + K 0 K 1 ) ′ ( 2 1 ( K 1 K 0 − I 1 I 0 ) + 2 z ( K 1 K 0 − I 1 I 0 ) ′ ) + z ( I 0 I 1 + K 0 K 1 ) ( ( K 1 K 0 − I 1 I 0 ) ′ + 2 z ( K 1 K 0 − I 1 I 0 ) ′′ ) ≤ z c . (168)
Similarly, for 0 < z < 1 0<z<1 0 < z < 1 , using the small-z z z asymptotics of Proposition 3.6 and the bounds of Proposition 3.2 , we have
∣ Q t A ( z ) ∣ ≤ c z 2 , ∣ Q t A ′ ( z ) ∣ ≤ c z , ∣ Q t A ′ ′ ( z ) ∣ ≤ c . (169) \left|Q_{\mathrm{tA}}(z)\right|\le cz^2,\qquad\left|Q_{\mathrm{tA}}'(z)\right|\le cz,\qquad\left|Q_{\mathrm{tA}}''(z)\right|\le c. \tag*{(169)} ∣ Q tA ( z ) ∣ ≤ c z 2 , ∣ Q tA ′ ( z ) ∣ ≤ cz , ∣ Q tA ′′ ( z ) ∣ ≤ c . (169)
For Q t B Q_{\mathrm{tB}} Q tB , again using (167) and Propositions 3.6 and 3.2, for z ≥ 1 z \ge1 z ≥ 1 we have
∣ Q t B ( z ) − 1 2 ∣ = ∣ z 2 2 ( I 0 I 1 − I 1 I 0 + K 0 K 1 − K 1 K 0 ) − 1 2 ∣ ≤ c z , ∣ Q t B ′ ( z ) ∣ = ∣ z ( I 0 I 1 − I 1 I 0 + K 0 K 1 − K 1 K 0 ) + z 2 2 ( I 0 I 1 − I 1 I 0 + K 0 K 1 − K 1 K 0 ) ′ ∣ ≤ c z , ∣ Q t B ′ ′ ( z ) ∣ = ∣ ( I 0 I 1 − I 1 I 0 + K 0 K 1 − K 1 K 0 ) + 2 z ( I 0 I 1 − I 1 I 0 + K 0 K 1 − K 1 K 0 ) ′ + z 2 2 ( I 0 I 1 − I 1 I 0 + K 0 K 1 − K 1 K 0 ) ′ ′ ∣ ≤ c z . (170) \begin{aligned}
\left|Q_{\mathrm{tB}}(z)-\frac{1}{2}\right|&=\left|\frac{z^2}{2}\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}+\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)-\frac{1}{2}\right|\leq\frac{c}{z},\\
\left|Q'_{\mathrm{tB}}(z)\right|&=\left|z\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}+\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)+\frac{z^2}{2}\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}+\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)'\right|\leq\frac{c}{z},\\
\left|Q''_{\mathrm{tB}}(z)\right|&=\left|\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}+\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)+2z\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}+\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)' \right.\\
&\qquad\left. +\frac{z^2}{2}\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}+\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)''\right|\leq\frac{c}{z}. \tag*{(170)}
\end{aligned} Q tB ( z ) − 2 1 ∣ Q tB ′ ( z ) ∣ ∣ Q tB ′′ ( z ) ∣ = 2 z 2 ( I 1 I 0 − I 0 I 1 + K 1 K 0 − K 0 K 1 ) − 2 1 ≤ z c , = z ( I 1 I 0 − I 0 I 1 + K 1 K 0 − K 0 K 1 ) + 2 z 2 ( I 1 I 0 − I 0 I 1 + K 1 K 0 − K 0 K 1 ) ′ ≤ z c , = ( I 1 I 0 − I 0 I 1 + K 1 K 0 − K 0 K 1 ) + 2 z ( I 1 I 0 − I 0 I 1 + K 1 K 0 − K 0 K 1 ) ′ + 2 z 2 ( I 1 I 0 − I 0 I 1 + K 1 K 0 − K 0 K 1 ) ′′ ≤ z c . (170)
For 0 < z < 1 0<z<1 0 < z < 1 , we have
∣ Q t B ( z ) ∣ ≤ c z , ∣ Q t B ′ ( z ) ∣ ≤ c , ∣ Q t B ′ ′ ( z ) ∣ ≤ c z . (171) \left|Q_{\mathrm{tB}}(z)\right|\leq cz,\qquad\left|Q'_{\mathrm{tB}}(z)\right|\leq c,\qquad\left|Q''_{\mathrm{tB}}(z)\right|\leq\frac{c}{z}. \tag*{(171)} ∣ Q tB ( z ) ∣ ≤ cz , ∣ Q tB ′ ( z ) ∣ ≤ c , ∣ Q tB ′′ ( z ) ∣ ≤ z c . (171)
Using the lower and upper bounds (164) and (165) for Q t D Q_{\mathrm{tD}} Q tD , along with the bounds (168) and (170) for Q t A Q_{\mathrm{tA}} Q tA and Q t B Q_{\mathrm{tB}} Q tB , for z ≥ 1 z\geq1 z ≥ 1 the formulas (163) yield
∣ m t A ( z ) ∣ ≤ c z , ∣ m t A ′ ( z ) ∣ ≤ c , ∣ m t A ′ ′ ( z ) ∣ ≤ c z ; \left|m_{\mathrm{tA}}(z)\right|\leq cz,\qquad\left|m'_{\mathrm{tA}}(z)\right|\leq c,\qquad\left|m''_{\mathrm{tA}}(z)\right|\leq\frac{c}{z}; ∣ m tA ( z ) ∣ ≤ cz , ∣ m tA ′ ( z ) ∣ ≤ c , ∣ m tA ′′ ( z ) ∣ ≤ z c ;
∣ m t B ( z ) ∣ ≤ c , ∣ m t B ′ ( z ) ∣ ≤ c z , ∣ m t B ′ ′ ( z ) ∣ ≤ c z . \left|m_{\mathrm{tB}}(z)\right|\leq c,\qquad\left|m'_{\mathrm{tB}}(z)\right|\leq\frac{c}{z},\qquad\left|m''_{\mathrm{tB}}(z)\right|\leq\frac{c}{z}. ∣ m tB ( z ) ∣ ≤ c , ∣ m tB ′ ( z ) ∣ ≤ z c , ∣ m tB ′′ ( z ) ∣ ≤ z c .
Similarly, using (166) with each of (169) , (171) in the formulas (163) , for 0 < z < 1 0<z<1 0 < z < 1 we have
∣ m t A ( z ) ∣ ≤ c , ∣ m t A ′ ( z ) ∣ ≤ c z , ∣ m t A ′ ′ ( z ) ∣ ≤ c z 2 ; \left|m_{\mathrm{tA}}(z)\right|\leq c,\qquad\left|m'_{\mathrm{tA}}(z)\right|\leq\frac{c}{z},\qquad\left|m''_{\mathrm{tA}}(z)\right|\leq\frac{c}{z^2}; ∣ m tA ( z ) ∣ ≤ c , ∣ m tA ′ ( z ) ∣ ≤ z c , ∣ m tA ′′ ( z ) ∣ ≤ z 2 c ;
∣ m t B ( z ) ∣ ≤ c z , ∣ m t B ′ ( z ) ∣ ≤ c z 2 , ∣ m t B ′ ′ ( z ) ∣ ≤ c z 3 . \left|m_{\mathrm{tB}}(z)\right|\leq\frac{c}{z},\qquad\left|m'_{\mathrm{tB}}(z)\right|\leq\frac{c}{z^2},\qquad\left|m''_{\mathrm{tB}}(z)\right|\leq\frac{c}{z^3}. ∣ m tB ( z ) ∣ ≤ z c , ∣ m tB ′ ( z ) ∣ ≤ z 2 c , ∣ m tB ′′ ( z ) ∣ ≤ z 3 c .
Using that m t A ( 2 π ∣ ξ ∣ ) = 1 2 m ε , z x ( ξ ) m_{\mathrm{tA}}(2\pi\lvert\xi\rvert)=\frac{1}{2}m_{\varepsilon,z}^{x}(\xi) m tA ( 2 π ∣ ξ ∣) = 2 1 m ε , z x ( ξ ) and m t B ( 2 π ∣ ξ ∣ ) = i 2 m ε , r ( ξ ) m_{\mathrm{tB}}(2\pi\lvert\xi\rvert)=\frac{i}{2}m_{\varepsilon,r}(\xi) m tB ( 2 π ∣ ξ ∣) = 2 i m ε , r ( ξ ) , we obtain Lemma 3.9. □
We next consider the normal direction multipliers m ε , r x m_{\varepsilon,r}^{x} m ε , r x , m ε , θ x m_{\varepsilon,\theta}^{x} m ε , θ x , m ε , z x m_{\varepsilon,z}^{x} m ε , z x for the angle-averaged single layer operator A ‾ ε \overline{\mathcal{A}}_{\varepsilon} A ε , defined in Corollary 1.8. Using Propositions 3.2, 3.6, 3.8, and Corollaries 3.3 and 3.7, we may obtain the following bounds for the normal direction multipliers and their derivatives.
Lemma 3.10 (Bounds for A ‾ ε \overline{\mathcal{A}}_{\varepsilon} A ε multipliers, normal direction). Consider the multipliers m ε , r x ( ξ ) m_{\varepsilon,r}^{x}(\xi) m ε , r x ( ξ ) , m ε , θ x ( ξ ) m_{\varepsilon,\theta}^{x}(\xi) m ε , θ x ( ξ ) , and m ε , z x ( ξ ) m_{\varepsilon,z}^{x}(\xi) m ε , z x ( ξ ) given by (59) in Corollary 1.8. For ℓ = 0 , 1 , 2 \ell=0,1,2 ℓ = 0 , 1 , 2 , these multipliers satisfy:
∣ ∂ ξ ℓ m ε , r x ( ξ ) ∣ ≤ { c ε ∣ ξ ∣ 1 − ℓ , ∣ ξ ∣ ≥ 1 2 π ε , c ∣ ξ ∣ − ℓ , ∣ ξ ∣ < 1 2 π ε , \left|\partial_{\xi}^{\ell}m_{\varepsilon,r}^{x}(\xi)\right|\leq
\begin{cases}
c\varepsilon|\xi|^{1-\ell}, & |\xi|\geq\frac{1}{2\pi\varepsilon},\\
c|\xi|^{-\ell}, & |\xi|<\frac{1}{2\pi\varepsilon},
\end{cases} ∂ ξ ℓ m ε , r x ( ξ ) ≤ { c ε ∣ ξ ∣ 1 − ℓ , c ∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 2 π ε 1 , ∣ ξ ∣ < 2 π ε 1 , ∣ ∂ ξ ℓ m ε , θ x ( ξ ) ∣ ≤ { c ε ∣ ξ ∣ 1 − ℓ , ∣ ξ ∣ ≥ 1 2 π ε , c ∣ ξ ∣ − ℓ , ∣ ξ ∣ < 1 2 π ε , \left|\partial_{\xi}^{\ell}m_{\varepsilon,\theta}^{x}(\xi)\right|\leq
\begin{cases}
c\varepsilon|\xi|^{1-\ell}, & |\xi|\geq\frac{1}{2\pi\varepsilon},\\
c|\xi|^{-\ell}, & |\xi|<\frac{1}{2\pi\varepsilon},
\end{cases} ∂ ξ ℓ m ε , θ x ( ξ ) ≤ { c ε ∣ ξ ∣ 1 − ℓ , c ∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 2 π ε 1 , ∣ ξ ∣ < 2 π ε 1 , ∣ ∂ ξ ℓ m ε , z x ( ξ ) ∣ ≤ { c ∣ ξ ∣ − ℓ , ∣ ξ ∣ ≥ 1 2 π ε , c ε ∣ log ε ∣ ∣ ξ ∣ 1 − ℓ , ∣ ξ ∣ < 1 2 π ε . (172) \left|\partial_{\xi}^{\ell}m_{\varepsilon,z}^{x}(\xi)\right|\leq
\begin{cases}
c|\xi|^{-\ell}, & |\xi|\geq\frac{1}{2\pi\varepsilon},\\
c\varepsilon|\log\varepsilon||\xi|^{1-\ell}, & |\xi|<\frac{1}{2\pi\varepsilon}.
\end{cases} \tag*{(172)} ∂ ξ ℓ m ε , z x ( ξ ) ≤ { c ∣ ξ ∣ − ℓ , c ε ∣ log ε ∣∣ ξ ∣ 1 − ℓ , ∣ ξ ∣ ≥ 2 π ε 1 , ∣ ξ ∣ < 2 π ε 1 . (172)
Proof. We begin by defining
m r x ( z ) = Q n A ( z ) − Q n B ( z ) Q n G ( z ) , m θ x ( z ) = Q n B ( z ) − Q n D ( z ) Q n G ( z ) , m z x ( z ) = Q n C ( z ) + Q n E ( z ) Q n G ( z ) . (173) m_{r}^{x}(z)=\frac{Q_{\mathrm{nA}}(z)-Q_{\mathrm{nB}}(z)}{Q_{\mathrm{nG}}(z)},\qquad
m_{\theta}^{x}(z)=\frac{Q_{\mathrm{nB}}(z)-Q_{\mathrm{nD}}(z)}{Q_{\mathrm{nG}}(z)},\qquad
m_{z}^{x}(z)=\frac{Q_{\mathrm{nC}}(z)+Q_{\mathrm{nE}}(z)}{Q_{\mathrm{nG}}(z)}. \tag*{(173)} m r x ( z ) = Q nG ( z ) Q nA ( z ) − Q nB ( z ) , m θ x ( z ) = Q nG ( z ) Q nB ( z ) − Q nD ( z ) , m z x ( z ) = Q nG ( z ) Q nC ( z ) + Q nE ( z ) . (173)
where the functions Q n j ( z ) Q_{\mathrm{n}j}(z) Q n j ( z ) are as in (55). Note that m ε , r x ( ξ ) = m r x ( 2 π ε ∣ ξ ∣ ) m_{\varepsilon,r}^{x}(\xi)=m_{r}^{x}(2\pi\varepsilon|\xi|) m ε , r x ( ξ ) = m r x ( 2 π ε ∣ ξ ∣ ) , m ε , θ x ( ξ ) = m θ x ( 2 π ε ∣ ξ ∣ ) m_{\varepsilon,\theta}^{x}(\xi)=m_{\theta}^{x}(2\pi\varepsilon|\xi|) m ε , θ x ( ξ ) = m θ x ( 2 π ε ∣ ξ ∣ ) , and m ε , z x ( ξ ) = i m z x ( 2 π ε ∣ ξ ∣ ) m_{\varepsilon,z}^{x}(\xi)=i\,m_{z}^{x}(2\pi\varepsilon|\xi|) m ε , z x ( ξ ) = i m z x ( 2 π ε ∣ ξ ∣ ) for the expressions (59) in Corollary 1.8. The first and second derivatives of m r x m^x_r m r x are given by
( m r x ) ′ ( z ) = ( Q n A ′ − Q n B ′ ) Q n G − ( Q n A − Q n B ) Q n G ′ Q n G 2 ( m r x ) ′ ′ ( z ) = ( Q n A ′ ′ − Q n B ′ ′ ) Q n G − ( Q n A − Q n B ) Q n G ′ ′ Q n G 2 + 2 Q n G ′ Q n G ( Q n A ′ − Q n B ′ ) , (174) \begin{aligned}
(m^x_r)'(z)&=\frac{(Q'_{\mathrm{nA}}-Q'_{\mathrm{nB}})Q_{\mathrm{nG}}-(Q_{\mathrm{nA}}-Q_{\mathrm{nB}})Q'_{\mathrm{nG}}}{Q_{\mathrm{nG}}^2}\\
(m^x_r)''(z)&=\frac{(Q''_{\mathrm{nA}}-Q''_{\mathrm{nB}})Q_{\mathrm{nG}}-(Q_{\mathrm{nA}}-Q_{\mathrm{nB}})Q''_{\mathrm{nG}}}{Q_{\mathrm{nG}}^2}+2\frac{Q'_{\mathrm{nG}}}{Q_{\mathrm{nG}}}(Q'_{\mathrm{nA}}-Q'_{\mathrm{nB}}), \tag*{(174)}
\end{aligned} ( m r x ) ′ ( z ) ( m r x ) ′′ ( z ) = Q nG 2 ( Q nA ′ − Q nB ′ ) Q nG − ( Q nA − Q nB ) Q nG ′ = Q nG 2 ( Q nA ′′ − Q nB ′′ ) Q nG − ( Q nA − Q nB ) Q nG ′′ + 2 Q nG Q nG ′ ( Q nA ′ − Q nB ′ ) , (174)
and the first and second derivatives of m θ x m^x_\theta m θ x and m z x m^x_z m z x may be written analogously. It thus remains to obtain bounds for the numerators and denominator of each multiplier.
We begin with bounds for Q n G Q_{\mathrm{nG}} Q nG . Using Proposition 3.8 and Proposition 3.2, we may obtain the following lower bound for ∣ Q n G ( z ) ∣ |Q_{\mathrm{nG}}(z)| ∣ Q nG ( z ) ∣ when z ≥ 1 z \ge1 z ≥ 1 :
∣ Q n G ( z ) ∣ = ∣ ( 2 z + ( 1 − z I 0 I 1 ) ( I 0 I 1 − I 1 I 0 ) ) ( 2 z − ( 1 + z K 0 K 1 ) ( K 0 K 1 − K 1 K 0 ) ) ∣ ≥ c > 0. (175) \left|Q_{\mathrm{nG}}(z)\right|=\left|\left(\frac{2}{z}+\left(1-z\frac{I_0}{I_1}\right)\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)\right)\left(\frac{2}{z}-\left(1+z\frac{K_0}{K_1}\right)\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)\right)\right|\ge c>0. \tag*{(175)} ∣ Q nG ( z ) ∣ = ( z 2 + ( 1 − z I 1 I 0 ) ( I 1 I 0 − I 0 I 1 ) ) ( z 2 − ( 1 + z K 1 K 0 ) ( K 1 K 0 − K 0 K 1 ) ) ≥ c > 0. (175)
In addition, by Propositions 3.6 and 3.2, for z ≥ 1 z \ge1 z ≥ 1 we have the upper bounds
∣ Q n G ( z ) ∣ ≤ c \left|Q_{\mathrm{nG}}(z)\right|\le c ∣ Q nG ( z ) ∣ ≤ c
∣ Q n G ′ ( z ) ∣ = ∣ ( 2 z 2 + ( I 0 I 1 + z ( I 0 I 1 ) ′ ) ( I 0 I 1 − I 1 I 0 ) − ( 1 − z I 0 I 1 ) ( I 0 I 1 − I 1 I 0 ) ′ ) × ( 2 z − ( 1 + z K 0 K 1 ) ( K 0 K 1 − K 1 K 0 ) ) + ( 2 z + ( 1 − z I 0 I 1 ) ( I 0 I 1 − I 1 I 0 ) ) × ( 2 z 2 + ( K 0 K 1 + z ( K 0 K 1 ) ′ ) ( K 0 K 1 − K 1 K 0 ) + ( 1 + z K 0 K 1 ) ( K 0 K 1 − K 1 K 0 ) ′ ) ∣ ≤ c z \begin{aligned}
\left|Q'_{\mathrm{nG}}(z)\right|={}&\left|\left(\frac{2}{z^2}+\left(\frac{I_0}{I_1}+z\left(\frac{I_0}{I_1}\right)'\right)\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)-\left(1-z\frac{I_0}{I_1}\right)\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)'\right)\right.\\
&\quad\times\left(\frac{2}{z}-\left(1+z\frac{K_0}{K_1}\right)\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)\right)\\
&\quad+\left(\frac{2}{z}+\left(1-z\frac{I_0}{I_1}\right)\left(\frac{I_0}{I_1}-\frac{I_1}{I_0}\right)\right)\\
&\quad\left.\times\left(\frac{2}{z^2}+\left(\frac{K_0}{K_1}+z\left(\frac{K_0}{K_1}\right)'\right)\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)+\left(1+z\frac{K_0}{K_1}\right)\left(\frac{K_0}{K_1}-\frac{K_1}{K_0}\right)'\right)\right|\le\frac{c}{z}
\end{aligned} ∣ Q nG ′ ( z ) ∣ = ( z 2 2 + ( I 1 I 0 + z ( I 1 I 0 ) ′ ) ( I 1 I 0 − I 0 I 1 ) − ( 1 − z I 1 I 0 ) ( I 1 I 0 − I 0 I 1 ) ′ ) × ( z 2 − ( 1 + z K 1 K 0 ) ( K 1 K 0 − K 0 K 1 ) ) + ( z 2 + ( 1 − z I 1 I 0 ) ( I 1 I 0 − I 0 I 1 ) ) × ( z 2 2 + ( K 1 K 0 + z ( K 1 K 0 ) ′ ) ( K 1 K 0 − K 0 K 1 ) + ( 1 + z K 1 K 0 ) ( K 1 K 0 − K 0 K 1 ) ′ ) ≤ z c
∣ Q n G ′ ′ ( z ) ∣ = ∣ ( − 4 z 3 + 2 ( I 0 I 1 ) ′ + ( I 0 I 1 ) ′ ′ ) ( I 0 I 1 − I 1 I 0 ) + 2 ( I 0 I 1 + z ( I 0 I 1 ) ′ ) ( I 0 I 1 − I 1 I 0 ) ′ − ( 1 − z I 0 I 1 ) ( I 0 I 1 − I 1 I 0 ) ′ ′ ) ( 2 z − ( 1 + z K 0 K 1 ) ( K 0 K 1 − K 1 K 0 ) ) + ( 2 z + ( 1 − z I 0 I 1 ) ( I 0 I 1 − I 1 I 0 ) ) ( − 4 z 3 + 2 ( K 0 K 1 ) ′ + ( K 0 K 1 ) ′ ′ ) ( K 0 K 1 − K 1 K 0 ) + 2 ( K 0 K 1 + z ( K 0 K 1 ) ′ ) ( K 0 K 1 − K 1 K 0 ) ′ + ( 1 + z K 0 K 1 ) ( K 0 K 1 − K 1 K 0 ) ′ ′ + 2 ( 2 z 2 + ( I 0 I 1 + z ( I 0 I 1 ) ′ ) ( I 0 I 1 − I 1 I 0 ) − ( 1 − z I 0 I 1 ) ( I 0 I 1 − I 1 I 0 ) ′ ) × ( 2 z 2 + ( K 0 K 1 + z ( K 0 K 1 ) ′ ) ( K 0 K 1 − K 1 K 0 ) + ( 1 + z K 0 K 1 ) ( K 0 K 1 − K 1 K 0 ) ′ ) ∣ ≤ c z 2 . (176) \begin{aligned}
|Q''_{\mathrm{nG}}(z)|={}&|(-\frac{4}{z^3}+2(\frac{I_0}{I_1})'+(\frac{I_0}{I_1})'')(\frac{I_0}{I_1}-\frac{I_1}{I_0})+2(\frac{I_0}{I_1}+z(\frac{I_0}{I_1})')(\frac{I_0}{I_1}-\frac{I_1}{I_0})'\\
&\quad-(1-z\frac{I_0}{I_1})(\frac{I_0}{I_1}-\frac{I_1}{I_0})'')(\frac{2}{z}-(1+z\frac{K_0}{K_1})(\frac{K_0}{K_1}-\frac{K_1}{K_0}))\\
&\quad+(\frac{2}{z}+(1-z\frac{I_0}{I_1})(\frac{I_0}{I_1}-\frac{I_1}{I_0}))(-\frac{4}{z^3}+2(\frac{K_0}{K_1})'+(\frac{K_0}{K_1})'')(\frac{K_0}{K_1}-\frac{K_1}{K_0})\\
&\quad+2(\frac{K_0}{K_1}+z(\frac{K_0}{K_1})')(\frac{K_0}{K_1}-\frac{K_1}{K_0})'+(1+z\frac{K_0}{K_1})(\frac{K_0}{K_1}-\frac{K_1}{K_0})''\\
&\quad+2(\frac{2}{z^2}+(\frac{I_0}{I_1}+z(\frac{I_0}{I_1})')(\frac{I_0}{I_1}-\frac{I_1}{I_0})-(1-z\frac{I_0}{I_1})(\frac{I_0}{I_1}-\frac{I_1}{I_0})')\\
&\quad\times(\frac{2}{z^2}+(\frac{K_0}{K_1}+z(\frac{K_0}{K_1})')(\frac{K_0}{K_1}-\frac{K_1}{K_0})+(1+z\frac{K_0}{K_1})(\frac{K_0}{K_1}-\frac{K_1}{K_0})')|\le\frac{c}{z^2}. \tag*{(176)}
\end{aligned} ∣ Q nG ′′ ( z ) ∣ = ∣ ( − z 3 4 + 2 ( I 1 I 0 ) ′ + ( I 1 I 0 ) ′′ ) ( I 1 I 0 − I 0 I 1 ) + 2 ( I 1 I 0 + z ( I 1 I 0 ) ′ ) ( I 1 I 0 − I 0 I 1 ) ′ − ( 1 − z I 1 I 0 ) ( I 1 I 0 − I 0 I 1 ) ′′ ) ( z 2 − ( 1 + z K 1 K 0 ) ( K 1 K 0 − K 0 K 1 )) + ( z 2 + ( 1 − z I 1 I 0 ) ( I 1 I 0 − I 0 I 1 )) ( − z 3 4 + 2 ( K 1 K 0 ) ′ + ( K 1 K 0 ) ′′ ) ( K 1 K 0 − K 0 K 1 ) + 2 ( K 1 K 0 + z ( K 1 K 0 ) ′ ) ( K 1 K 0 − K 0 K 1 ) ′ + ( 1 + z K 1 K 0 ) ( K 1 K 0 − K 0 K 1 ) ′′ + 2 ( z 2 2 + ( I 1 I 0 + z ( I 1 I 0 ) ′ ) ( I 1 I 0 − I 0 I 1 ) − ( 1 − z I 1 I 0 ) ( I 1 I 0 − I 0 I 1 ) ′ ) × ( z 2 2 + ( K 1 K 0 + z ( K 1 K 0 ) ′ ) ( K 1 K 0 − K 0 K 1 ) + ( 1 + z K 1 K 0 ) ( K 1 K 0 − K 0 K 1 ) ′ ) ∣ ≤ z 2 c . (176)
For 0 < z < 1 0<z<1 0 < z < 1 , we may use cancellations in the small-z z z asymptotics (146) for I 0 I 1 \frac{I_0}{I_1} I 1 I 0 along with the small-z z z bounds of Propositions 3.2 and 3.6 to obtain
1 c ≤ ∣ Q n G ( z ) ∣ ≤ c , ∣ Q n G ′ ( z ) ∣ ≤ c z , ∣ Q n G ′ ′ ( z ) ∣ ≤ c z 2 , 0 < z < 1. (177) \frac{1}{c}\le|Q_{\mathrm{nG}}(z)|\le c,\qquad|Q'_{\mathrm{nG}}(z)|\le\frac{c}{z},\qquad|Q''_{\mathrm{nG}}(z)|\le\frac{c}{z^2},\qquad0<z<1. \tag*{(177)} c 1 ≤ ∣ Q nG ( z ) ∣ ≤ c , ∣ Q nG ′ ( z ) ∣ ≤ z c , ∣ Q nG ′′ ( z ) ∣ ≤ z 2 c , 0 < z < 1. (177)
We next consider the numerator of m r x ( z ) m^x_r(z) m r x ( z ) in (172). Using the form (55) of Q n A Q_{\mathrm{nA}} Q nA and Q n B Q_{\mathrm{nB}} Q nB , and using (166) to rewrite 1 I 1 K 1 \frac{1}{I_1K_1} I 1 K 1 1 , for z ≥ 1 z\ge1 z ≥ 1 we may use Propositions 3.2, 3.6 and Corollaries 3.3, 3.7 to obtain:
∣ Q n A − Q n B ∣ = ∣ z ( I 0 I 1 + K 0 K 1 ) ( 6 + ( 8 z 2 + 2 ) I 1 I 0 K 1 K 0 + 4 z ( K 1 K 0 − I 1 I 0 ) − 2 z ( I 0 I 1 − K 0 K 1 ) − 4 ( I 0 I 1 K 1 K 0 + I 1 I 0 K 0 K 1 ) ) ∣ ≤ c z , (178) \left|Q_{n\mathrm{A}}-Q_{n\mathrm{B}}\right|=\left|z\left(\frac{I_{0}}{I_{1}}+\frac{K_{0}}{K_{1}}\right)\left(6+\left(\frac{8}{z^{2}}+2\right)\frac{I_{1}}{I_{0}}\frac{K_{1}}{K_{0}}+\frac{4}{z}\left(\frac{K_{1}}{K_{0}}-\frac{I_{1}}{I_{0}}\right)-2z\left(\frac{I_{0}}{I_{1}}-\frac{K_{0}}{K_{1}}\right)-4\left(\frac{I_{0}}{I_{1}}\frac{K_{1}}{K_{0}}+\frac{I_{1}}{I_{0}}\frac{K_{0}}{K_{1}}\right)\right)\right|\leq cz, \tag*{(178)} ∣ Q n A − Q n B ∣ = z ( I 1 I 0 + K 1 K 0 ) ( 6 + ( z 2 8 + 2 ) I 0 I 1 K 0 K 1 + z 4 ( K 0 K 1 − I 0 I 1 ) − 2 z ( I 1 I 0 − K 1 K 0 ) − 4 ( I 1 I 0 K 0 K 1 + I 0 I 1 K 1 K 0 ) ) ≤ cz , (178)
∣ Q n A ′ − Q n B ′ ∣ ≤ c , ∣ Q n A ′ ′ − Q n B ′ ′ ∣ ≤ c z . \left|Q'_{n\mathrm{A}}-Q'_{n\mathrm{B}}\right|\leq c,\qquad\left|Q''_{n\mathrm{A}}-Q''_{n\mathrm{B}}\right|\leq\frac{c}{z}. ∣ Q n A ′ − Q n B ′ ∣ ≤ c , ∣ Q n A ′′ − Q n B ′′ ∣ ≤ z c .
We have suppressed the full expressions for the first and second derivatives above, but note that they can easily be computed from the first expression.
For 0 < z < 1 0<z<1 0 < z < 1 , we may use cancellations in the small-z z z asymptotics (147) , (146) of I 0 I 1 \frac{I_{0}}{I_{1}} I 1 I 0 and I 1 I 0 \frac{I_{1}}{I_{0}} I 0 I 1 along with Proposition (3.2) to obtain
∣ Q n A − Q n B ∣ ≤ c , ∣ Q n A ′ − Q n B ′ ∣ ≤ c z , ∣ Q n A ′ ′ − Q n B ′ ′ ∣ ≤ c z 2 . (179) \left|Q_{n\mathrm{A}}-Q_{n\mathrm{B}}\right|\leq c,\qquad\left|Q'_{n\mathrm{A}}-Q'_{n\mathrm{B}}\right|\leq\frac{c}{z},\qquad\left|Q''_{n\mathrm{A}}-Q''_{n\mathrm{B}}\right|\leq\frac{c}{z^{2}}. \tag*{(179)} ∣ Q n A − Q n B ∣ ≤ c , ∣ Q n A ′ − Q n B ′ ∣ ≤ z c , ∣ Q n A ′′ − Q n B ′′ ∣ ≤ z 2 c . (179)
Combining (175) , (176) , (177) , (178) , and (179) with the definition (173) of m r x ( z ) m_{r}^{x}(z) m r x ( z ) and the expression (174) of its derivatives, we obtain
∣ m r x ( z ) ∣ ≤ { c z , z ≥ 1 c , 0 < z < 1 , ∣ ( m r x ) ′ ( z ) ∣ ≤ { c , z ≥ 1 c z , 0 < z < 1 , \left|m_{r}^{x}(z)\right|\leq
\begin{cases}
cz, & z\geq1\\
c, & 0<z<1,
\end{cases}
\qquad
\left|(m_{r}^{x})'(z)\right|\leq
\begin{cases}
c, & z\geq1\\
\frac{c}{z}, & 0<z<1,
\end{cases} ∣ m r x ( z ) ∣ ≤ { cz , c , z ≥ 1 0 < z < 1 , ∣ ( m r x ) ′ ( z ) ∣ ≤ { c , z c , z ≥ 1 0 < z < 1 ,
∣ ( m r x ) ′ ′ ( z ) ∣ ≤ { c z , z ≥ 1 c z 2 , 0 < z < 1. (180) \left|(m_{r}^{x})''(z)\right|\leq
\begin{cases}
\frac{c}{z}, & z\geq1\\
\frac{c}{z^{2}}, & 0<z<1.
\end{cases} \tag*{(180)} ∣ ( m r x ) ′′ ( z ) ∣ ≤ { z c , z 2 c , z ≥ 1 0 < z < 1. (180)
We now turn to the numerator of m θ x ( z ) m_{\theta}^{x}(z) m θ x ( z ) from (173) . We recall the form (55) of Q n B Q_{n\mathrm{B}} Q n B and Q n D Q_{n\mathrm{D}} Q n D and again use (167) to rewrite 1 I 1 K 1 \frac{1}{I_{1}K_{1}} I 1 K 1 1 . By Propositions (3.2), (3.6) and Corollaries (3.3), (3.7), for z ≥ 1 z\geq1 z ≥ 1 we have
∣ Q n B − Q n D ∣ = ∣ z ( I 0 I 1 + K 0 K 1 ) ( 4 K 1 K 0 I 0 I 1 − 6 K 1 K 0 I 1 I 0 + 4 I 1 I 0 K 0 K 1 − 2 + 4 z ( I 1 I 0 − K 1 K 0 ) − 8 z 2 I 1 I 0 K 1 K 0 − z 2 ( I 0 I 1 − I 1 I 0 ) ( K 0 K 1 − K 1 K 0 ) ) ∣ ≤ c z , (181) \left|Q_{n\mathrm{B}}-Q_{n\mathrm{D}}\right|=\left|z\left(\frac{I_{0}}{I_{1}}+\frac{K_{0}}{K_{1}}\right)\left(4\frac{K_{1}}{K_{0}}\frac{I_{0}}{I_{1}}-6\frac{K_{1}}{K_{0}}\frac{I_{1}}{I_{0}}+4\frac{I_{1}}{I_{0}}\frac{K_{0}}{K_{1}}-2+\frac{4}{z}\left(\frac{I_{1}}{I_{0}}-\frac{K_{1}}{K_{0}}\right)-\frac{8}{z^{2}}\frac{I_{1}}{I_{0}}\frac{K_{1}}{K_{0}}-z^{2}\left(\frac{I_{0}}{I_{1}}-\frac{I_{1}}{I_{0}}\right)\left(\frac{K_{0}}{K_{1}}-\frac{K_{1}}{K_{0}}\right)\right)\right|\leq cz, \tag*{(181)} ∣ Q n B − Q n D ∣ = z ( I 1 I 0 + K 1 K 0 ) ( 4 K 0 K 1 I 1 I 0 − 6 K 0 K 1 I 0 I 1 + 4 I 0 I 1 K 1 K 0 − 2 + z 4 ( I 0 I 1 − K 0 K 1 ) − z 2 8 I 0 I 1 K 0 K 1 − z 2 ( I 1 I 0 − I 0 I 1 ) ( K 1 K 0 − K 0 K 1 ) ) ≤ cz , (181)
∣ Q n B ′ − Q n D ′ ∣ ≤ c , ∣ Q n B ′ ′ − Q n D ′ ′ ∣ ≤ c z . \left|Q'_{n\mathrm{B}}-Q'_{n\mathrm{D}}\right|\leq c,\qquad\left|Q''_{n\mathrm{B}}-Q''_{n\mathrm{D}}\right|\leq\frac{c}{z}. ∣ Q n B ′ − Q n D ′ ∣ ≤ c , ∣ Q n B ′′ − Q n D ′′ ∣ ≤ z c .
Again we have suppressed the full expressions for the first and second derivatives, but again they may be easily computed from the first expression. For 0 < z < 1 0<z<1 0 < z < 1 , again using Proposition (3.2) and cancellations in the small-z z z asymptotics (147) , (146) of I 0 I 1 \frac{I_{0}}{I_{1}} I 1 I 0 and I 1 I 0 \frac{I_{1}}{I_{0}} I 0 I 1 , we have
∣ Q n B − Q n D ∣ ≤ c , ∣ Q n B ′ − Q n D ′ ∣ ≤ c z , ∣ Q n B ′ ′ − Q n D ′ ′ ∣ ≤ c z 2 . (182) \left|Q_{n\mathrm{B}}-Q_{n\mathrm{D}}\right|\leq c,\qquad\left|Q'_{n\mathrm{B}}-Q'_{n\mathrm{D}}\right|\leq\frac{c}{z},\qquad\left|Q''_{n\mathrm{B}}-Q''_{n\mathrm{D}}\right|\leq\frac{c}{z^{2}}. \tag*{(182)} ∣ Q n B − Q n D ∣ ≤ c , ∣ Q n B ′ − Q n D ′ ∣ ≤ z c , ∣ Q n B ′′ − Q n D ′′ ∣ ≤ z 2 c . (182)
Combining (175) , (176) , (177) , (181) , and (182) with the definition (173) of m θ x ( z ) m_{\theta}^{x}(z) m θ x ( z ) and the analogous expression (174) of its derivatives, we obtain
∣ m θ x ( z ) ∣ ≤ { c z , z ≥ 1 c , 0 < z < 1 , ∣ ( m θ x ) ′ ( z ) ∣ ≤ { c , z ≥ 1 c z , 0 < z < 1 , \left|m_{\theta}^{x}(z)\right|\leq
\begin{cases}
cz, & z\geq1\\
c, & 0<z<1,
\end{cases}
\qquad
\left|(m_{\theta}^{x})'(z)\right|\leq
\begin{cases}
c, & z\geq1\\
\frac{c}{z}, & 0<z<1,
\end{cases} ∣ m θ x ( z ) ∣ ≤ { cz , c , z ≥ 1 0 < z < 1 , ∣ ( m θ x ) ′ ( z ) ∣ ≤ { c , z c , z ≥ 1 0 < z < 1 ,
∣ ( m θ x ) ′ ′ ( z ) ∣ ≤ { c z , z ≥ 1 c z 2 , 0 < z < 1. (183) \left|(m_{\theta}^{x})''(z)\right|\leq
\begin{cases}
\frac{c}{z}, & z\geq1\\
\frac{c}{z^{2}}, & 0<z<1.
\end{cases} \tag*{(183)} ∣ ( m θ x ) ′′ ( z ) ∣ ≤ { z c , z 2 c , z ≥ 1 0 < z < 1. (183)
Finally, we consider the numerator of m z x ( z ) m_z^x(z) m z x ( z ) from (173) . Using the form (55) of Q n C Q_{\mathrm{nC}} Q nC and Q n E Q_{\mathrm{nE}} Q nE along with the identity (167) , for z ≥ 1 z \ge1 z ≥ 1 we have
∣ Q n C + Q n E ∣ = ∣ z ( I 0 I 1 + K 0 K 1 ) ( 8 z I 1 I 0 K 1 K 0 + 6 ( I 0 I 1 − K 0 K 1 ) + 2 ( I 1 I 0 − K 1 K 0 ) + 2 z ( I 0 I 1 K 0 K 1 − 1 ) |Q_{\mathrm{nC}}+Q_{\mathrm{nE}}|=|z(\frac{I_0}{I_1}+\frac{K_0}{K_1})(\frac{8}{z}\frac{I_1}{I_0}\frac{K_1}{K_0}+6(\frac{I_0}{I_1}-\frac{K_0}{K_1})+2(\frac{I_1}{I_0}-\frac{K_1}{K_0})+2z(\frac{I_0}{I_1}\frac{K_0}{K_1}-1) ∣ Q nC + Q nE ∣ = ∣ z ( I 1 I 0 + K 1 K 0 ) ( z 8 I 0 I 1 K 0 K 1 + 6 ( I 1 I 0 − K 1 K 0 ) + 2 ( I 0 I 1 − K 0 K 1 ) + 2 z ( I 1 I 0 K 1 K 0 − 1 )
− 12 z + z ( I 0 I 1 − I 1 I 0 ) ( K 0 K 1 − K 1 K 0 ) ) ∣ ≤ c , (184) \qquad-\frac{12}{z}+z(\frac{I_0}{I_1}-\frac{I_1}{I_0})(\frac{K_0}{K_1}-\frac{K_1}{K_0}))| \le c, \tag*{(184)} − z 12 + z ( I 1 I 0 − I 0 I 1 ) ( K 1 K 0 − K 0 K 1 )) ∣ ≤ c , (184)
∣ Q n C ′ + Q n E ′ ∣ ≤ c z , ∣ Q n C ′ ′ + Q n E ′ ′ ∣ ≤ c z 2 . \left|Q'_{\mathrm{nC}}+Q'_{\mathrm{nE}}\right|\le\frac{c}{z},\qquad\left|Q''_{\mathrm{nC}}+Q''_{\mathrm{nE}}\right|\le\frac{c}{z^2}. ∣ Q nC ′ + Q nE ′ ∣ ≤ z c , ∣ Q nC ′′ + Q nE ′′ ∣ ≤ z 2 c .
Again we suppress the full expressions for the first and second derivatives, as they are straightforward to compute. For 0 < z < 1 0<z<1 0 < z < 1 , we again use Proposition 3.2 along with cancellations in the small-z z z asymptotics (147) , (146) of I 0 I 1 \frac{I_0}{I_1} I 1 I 0 and I 1 I 0 \frac{I_1}{I_0} I 0 I 1 to obtain
∣ Q n C + Q n E ∣ ≤ c z ∣ log ( z 2 ) ∣ , ∣ Q n C ′ + Q n E ′ ∣ ≤ c ∣ log ( z 2 ) ∣ , ∣ Q n C ′ ′ + Q n E ′ ′ ∣ ≤ c ∣ log ( z 2 ) ∣ z . (185) \left|Q_{\mathrm{nC}}+Q_{\mathrm{nE}}\right|\le cz\left|\log\left(\frac{z}{2}\right)\right|,\qquad\left|Q'_{\mathrm{nC}}+Q'_{\mathrm{nE}}\right|\le c\left|\log\left(\frac{z}{2}\right)\right|,\qquad\left|Q''_{\mathrm{nC}}+Q''_{\mathrm{nE}}\right|\le c\frac{\left|\log\left(\frac{z}{2}\right)\right|}{z}. \tag*{(185)} ∣ Q nC + Q nE ∣ ≤ cz log ( 2 z ) , ∣ Q nC ′ + Q nE ′ ∣ ≤ c log ( 2 z ) , ∣ Q nC ′′ + Q nE ′′ ∣ ≤ c z log ( 2 z ) . (185)
Combining (175) , (176) , (177) , (184) , and (185) with the definition (173) of m z x ( z ) m_z^x(z) m z x ( z ) and the analogous expression (174) of its derivatives, we obtain
∣ m z x ( z ) ∣ ≤ { c , z ≥ 1 , c z ∣ log ( z 2 ) ∣ , 0 < z < 1 , ∣ ( m z x ) ′ ( z ) ∣ ≤ { c z , z ≥ 1 , c ∣ log ( z 2 ) ∣ , 0 < z < 1 , \left|m_z^x(z)\right|\le
\begin{cases}
c, & z\ge1,\\
cz\left|\log\left(\frac{z}{2}\right)\right|, & 0<z<1,
\end{cases}
\qquad
\left|(m_z^x)'(z)\right|\le
\begin{cases}
\frac{c}{z}, & z\ge1,\\
c\left|\log\left(\frac{z}{2}\right)\right|, & 0<z<1,
\end{cases} ∣ m z x ( z ) ∣ ≤ { c , cz log ( 2 z ) , z ≥ 1 , 0 < z < 1 , ∣ ( m z x ) ′ ( z ) ∣ ≤ { z c , c log ( 2 z ) , z ≥ 1 , 0 < z < 1 ,
∣ ( m z x ) ′ ′ ( z ) ∣ ≤ { c z 2 , z ≥ 1 , c ∣ log ( z / 2 ) ∣ z , 0 < z < 1. (186) \left|(m_z^x)''(z)\right|\le
\begin{cases}
\frac{c}{z^2}, & z\ge1,\\
c\frac{\left|\log(z/2)\right|}{z}, & 0<z<1.
\end{cases} \tag*{(186)} ∣ ( m z x ) ′′ ( z ) ∣ ≤ { z 2 c , c z ∣ l o g ( z /2 ) ∣ , z ≥ 1 , 0 < z < 1. (186)
Using the bounds (180) , (183) , and (186) along with the fact that m ϵ , r x ( ξ ) = m r x ( 2 π ϵ ∣ ξ ∣ ) m_{\epsilon,r}^x(\xi)=m_r^x(2\pi\epsilon|\xi|) m ϵ , r x ( ξ ) = m r x ( 2 π ϵ ∣ ξ ∣ ) , m ϵ , θ x ( ξ ) = m θ x ( 2 π ϵ ∣ ξ ∣ ) m_{\epsilon,\theta}^x(\xi)=m_\theta^x(2\pi\epsilon|\xi|) m ϵ , θ x ( ξ ) = m θ x ( 2 π ϵ ∣ ξ ∣ ) , and m ϵ , z x ( ξ ) = i m z x ( 2 π ϵ ∣ ξ ∣ ) m_{\epsilon,z}^x(\xi)=i m_z^x(2\pi\epsilon|\xi|) m ϵ , z x ( ξ ) = i m z x ( 2 π ϵ ∣ ξ ∣ ) , we obtain Lemma 3.10 . □
Proofs of Lemmas 1.4 and 1.9
Given the multiplier bounds in Lemmas 3.4, 3.5, 3.9 , 3.10 , we may now use Lemma 3.1 to prove Lemmas 1.4 and 1.9 . We begin with Lemma 1.4 .
Proof of Lemma 1.4
Recalling the definitions (23) and (24) of m ϵ , t m_{\epsilon,t} m ϵ , t and m ϵ , η m_{\epsilon,\eta} m ϵ , η , we define the matrix-valued Fourier multipliers
M ^ ϵ , j ( 1 ) ( ξ ) = m ϵ , t ( ξ ) e z ⊗ e z + m ϵ , η ( ξ ) ( e x ⊗ e x + e y ⊗ e y ) , M ^ ϵ , j ( 1 ) ( ξ ) = ϕ j ( ξ ) M ^ ϵ ( 1 ) ( ξ ) , \widehat{M}_{\epsilon,j}^{(1)}(\xi)=m_{\epsilon,t}(\xi)e_z\otimes e_z+m_{\epsilon,\eta}(\xi)(e_x\otimes e_x+e_y\otimes e_y),\qquad\widehat{M}_{\epsilon,j}^{(1)}(\xi)=\phi_j(\xi)\widehat{M}_{\epsilon}^{(1)}(\xi), M ϵ , j ( 1 ) ( ξ ) = m ϵ , t ( ξ ) e z ⊗ e z + m ϵ , η ( ξ ) ( e x ⊗ e x + e y ⊗ e y ) , M ϵ , j ( 1 ) ( ξ ) = ϕ j ( ξ ) M ϵ ( 1 ) ( ξ ) ,
M ^ ϵ ( 2 ) ( ξ ) = m ϵ , t − 1 ( ξ ) e z ⊗ e z + m ϵ , η − 1 ( ξ ) ( e x ⊗ e x + e y ⊗ e y ) , M ^ ϵ , j ( 2 ) ( ξ ) = ϕ j ( ξ ) M ^ ϵ ( 2 ) ( ξ ) . (187) \widehat{M}_{\epsilon}^{(2)}(\xi)=m_{\epsilon,t}^{-1}(\xi)e_z\otimes e_z+m_{\epsilon,\eta}^{-1}(\xi)(e_x\otimes e_x+e_y\otimes e_y),\qquad\widehat{M}_{\epsilon,j}^{(2)}(\xi)=\phi_j(\xi)\widehat{M}_{\epsilon}^{(2)}(\xi). \tag*{(187)} M ϵ ( 2 ) ( ξ ) = m ϵ , t − 1 ( ξ ) e z ⊗ e z + m ϵ , η − 1 ( ξ ) ( e x ⊗ e x + e y ⊗ e y ) , M ϵ , j ( 2 ) ( ξ ) = ϕ j ( ξ ) M ϵ ( 2 ) ( ξ ) . (187)
Here ϕ j ( ξ ) \phi_j(\xi) ϕ j ( ξ ) is as in (106) . Letting j ϵ = ∣ log ( 2 π ϵ ) ∣ log ( 2 ) j_\epsilon=\frac{\left|\log(2\pi\epsilon)\right|}{\log(2)} j ϵ = l o g ( 2 ) ∣ l o g ( 2 π ϵ ) ∣ , by Lemmas 3.4 and 3.5, we have
∥ ∂ ξ ℓ M ^ ϵ , j ( 1 ) ∥ L ∞ ≤ { c ϵ − 1 2 − j ( ℓ + 1 ) , j ≥ j ϵ , c ∣ log ϵ ∣ 2 − j ℓ , j < j ϵ , ℓ = 0 , 1 , 2 ; \left\|\partial_\xi^\ell\widehat{M}_{\epsilon,j}^{(1)}\right\|_{L^\infty}\le
\begin{cases}
c\epsilon^{-1}2^{-j(\ell+1)}, & j\ge j_\epsilon,\\
c|\log\epsilon|2^{-j\ell}, & j<j_\epsilon,
\end{cases}
\qquad\ell=0,1,2; ∂ ξ ℓ M ϵ , j ( 1 ) L ∞ ≤ { c ϵ − 1 2 − j ( ℓ + 1 ) , c ∣ log ϵ ∣ 2 − j ℓ , j ≥ j ϵ , j < j ϵ , ℓ = 0 , 1 , 2 ;
∥ ∂ ξ ℓ M ^ ϵ , j ( 2 ) ∥ L ∞ ≤ { c ϵ 2 j ( 1 − ℓ ) , j ≥ j ϵ , c ∣ log ϵ ∣ − 1 2 − j ℓ , j < j ϵ , ℓ = 0 , 1 , 2. \left\|\partial_\xi^\ell\widehat{M}_{\epsilon,j}^{(2)}\right\|_{L^\infty}\le
\begin{cases}
c\epsilon2^{j(1-\ell)}, & j\ge j_\epsilon,\\
c|\log\epsilon|^{-1}2^{-j\ell}, & j<j_\epsilon,
\end{cases}
\qquad\ell=0,1,2. ∂ ξ ℓ M ϵ , j ( 2 ) L ∞ ≤ { cϵ 2 j ( 1 − ℓ ) , c ∣ log ϵ ∣ − 1 2 − j ℓ , j ≥ j ϵ , j < j ϵ , ℓ = 0 , 1 , 2.
We may then use Lemma 3.1 to obtain physical space estimates for M ϵ , j ( 1 ) M_{\epsilon,j}^{(1)} M ϵ , j ( 1 ) and M ϵ , j ( 2 ) M_{\epsilon,j}^{(2)} M ϵ , j ( 2 ) :
∥ M ϵ , j ( 1 ) ∥ L 1 ≤ { c ϵ − 1 2 − j , j ≥ j ϵ , c ∣ log ϵ ∣ , j < j ϵ , ∥ M ϵ , j ( 2 ) ∥ L 1 ≤ { c ϵ 2 j , j ≥ j ϵ , c ∣ log ϵ ∣ − 1 , j < j ϵ . \left\|M_{\epsilon,j}^{(1)}\right\|_{L^1}\le
\begin{cases}
c\epsilon^{-1}2^{-j}, & j\ge j_\epsilon,\\
c|\log\epsilon|, & j<j_\epsilon,
\end{cases}
\qquad
\left\|M_{\epsilon,j}^{(2)}\right\|_{L^1}\le
\begin{cases}
c\epsilon2^j, & j\ge j_\epsilon,\\
c|\log\epsilon|^{-1}, & j<j_\epsilon.
\end{cases} M ϵ , j ( 1 ) L 1 ≤ { c ϵ − 1 2 − j , c ∣ log ϵ ∣ , j ≥ j ϵ , j < j ϵ , M ϵ , j ( 2 ) L 1 ≤ { cϵ 2 j , c ∣ log ϵ ∣ − 1 , j ≥ j ϵ , j < j ϵ .
Given h ∈ C 0 , α ( T ) h \in C^{0,\alpha}(\mathbb{T}) h ∈ C 0 , α ( T ) with ∫ T h ( s ) d s = 0 \int_{\mathbb{T}} h(s)\,\mathrm{d}s=0 ∫ T h ( s ) d s = 0 and g ∈ C 1 , α ( T ) g \in C^{1,\alpha}(\mathbb{T}) g ∈ C 1 , α ( T ) , we may use the dyadic partition of unity (105) to write
L ~ ϵ [ h ] = T ~ M ϵ ( 1 ) h = ∑ j P j T ~ M ϵ ( 1 ) h , L ~ ϵ − 1 [ g ] = T ~ M ϵ ( 2 ) g = ∑ j P j T ~ M ϵ ( 2 ) g . \widetilde{\mathcal{L}}_{\epsilon}[h]=\widetilde{T}_{M_{\epsilon}^{(1)}}h=\sum_j P_j\widetilde{T}_{M_{\epsilon}^{(1)}}h,\qquad\widetilde{\mathcal{L}}_{\epsilon}^{-1}[g]=\widetilde{T}_{M_{\epsilon}^{(2)}}g=\sum_j P_j\widetilde{T}_{M_{\epsilon}^{(2)}}g. L ϵ [ h ] = T M ϵ ( 1 ) h = j ∑ P j T M ϵ ( 1 ) h , L ϵ − 1 [ g ] = T M ϵ ( 2 ) g = j ∑ P j T M ϵ ( 2 ) g .
On T \mathbb{T} T , we have P ≤ 0 T ~ M ϵ ( 1 ) h = P 0 T ~ M ϵ ( 1 ) h P_{\leq0}\widetilde{T}_{M_{\epsilon}^{(1)}}h=P_0\widetilde{T}_{M_{\epsilon}^{(1)}}h P ≤ 0 T M ϵ ( 1 ) h = P 0 T M ϵ ( 1 ) h and P ≤ 0 T ~ M ϵ ( 2 ) g = P 0 T ~ M ϵ ( 2 ) g P_{\leq0}\widetilde{T}_{M_{\epsilon}^{(2)}}g=P_0\widetilde{T}_{M_{\epsilon}^{(2)}}g P ≤ 0 T M ϵ ( 2 ) g = P 0 T M ϵ ( 2 ) g . We may then estimate
∥ P ≤ 0 T ~ M ϵ ( 1 ) h ∥ L ∞ = ∥ M ϵ , 0 ( 1 ) ∗ h ∥ L ∞ ≤ ∥ M ϵ , 0 ( 1 ) ∥ L 1 ∥ h ∥ L ∞ ≤ c ∣ log ϵ ∣ ∥ h ∥ L ∞ , ∥ P ≤ 0 T ~ M ϵ ( 2 ) g ∥ L ∞ = ∥ M ϵ , 0 ( 2 ) ∗ g ∥ L ∞ ≤ ∥ M ϵ , 0 ( 2 ) ∥ L 1 ∥ g ∥ L ∞ ≤ c ∣ log ϵ ∣ − 1 ∥ g ∥ L ∞ . (188) \begin{aligned}
\left\|P_{\leq0}\widetilde{T}_{M_{\epsilon}^{(1)}}h\right\|_{L^\infty}
&=\left\|M_{\epsilon,0}^{(1)}*h\right\|_{L^\infty}
\leq\left\|M_{\epsilon,0}^{(1)}\right\|_{L^1}\|h\|_{L^\infty}
\leq c|\log\epsilon|\|h\|_{L^\infty},\\
\left\|P_{\leq0}\widetilde{T}_{M_{\epsilon}^{(2)}}g\right\|_{L^\infty}
&=\left\|M_{\epsilon,0}^{(2)}*g\right\|_{L^\infty}
\leq\left\|M_{\epsilon,0}^{(2)}\right\|_{L^1}\|g\|_{L^\infty}
\leq c|\log\epsilon|^{-1}\|g\|_{L^\infty}. \tag*{(188)}
\end{aligned} P ≤ 0 T M ϵ ( 1 ) h L ∞ P ≤ 0 T M ϵ ( 2 ) g L ∞ = M ϵ , 0 ( 1 ) ∗ h L ∞ ≤ M ϵ , 0 ( 1 ) L 1 ∥ h ∥ L ∞ ≤ c ∣ log ϵ ∣∥ h ∥ L ∞ , = M ϵ , 0 ( 2 ) ∗ g L ∞ ≤ M ϵ , 0 ( 2 ) L 1 ∥ g ∥ L ∞ ≤ c ∣ log ϵ ∣ − 1 ∥ g ∥ L ∞ . (188)
For j > 0 j>0 j > 0 we introduce the notation P ~ j = P j − 1 + P j + P j + 1 \widetilde{P}_j=P_{j-1}+P_j+P_{j+1} P j = P j − 1 + P j + P j + 1 and note that P j T ~ M ϵ ( 1 ) h = P j T ~ M ϵ ( 1 ) P ~ j h P_j\widetilde{T}_{M_{\epsilon}^{(1)}}h=P_j\widetilde{T}_{M_{\epsilon}^{(1)}}\widetilde{P}_jh P j T M ϵ ( 1 ) h = P j T M ϵ ( 1 ) P j h and P j T ~ M ϵ ( 2 ) g = P j T ~ M ϵ ( 2 ) P ~ j g P_j\widetilde{T}_{M_{\epsilon}^{(2)}}g=P_j\widetilde{T}_{M_{\epsilon}^{(2)}}\widetilde{P}_jg P j T M ϵ ( 2 ) g = P j T M ϵ ( 2 ) P j g . For low frequencies 0 < j < j ϵ 0<j<j_\epsilon 0 < j < j ϵ , we estimate
sup 0 < j < j ϵ 2 j ( 1 + α ) ∥ P j T ~ M ϵ ( 1 ) h ∥ L ∞ = sup 0 < j < j ϵ 2 j ( 1 + α ) ∥ P j T ~ M ϵ ( 1 ) P ~ j h ∥ L ∞ = sup 0 < j < j ϵ 2 j ( 1 + α ) ∥ M ϵ , j ( 1 ) ∗ ( P ~ j h ) ∥ L ∞ ≤ 2 j ϵ sup 0 < j < j ϵ ∥ M ϵ , j ( 1 ) ∥ L 1 2 j α ∥ P ~ j h ∥ L ∞ ≤ c ϵ − 1 ∣ log ϵ ∣ ∣ h ∣ B ˙ ∞ , ∞ α , sup 0 < j < j ϵ 2 j α ∥ P j T ~ M ϵ ( 2 ) g ∥ L ∞ = sup 0 < j < j ϵ 2 j α ∥ P j T ~ M ϵ ( 2 ) P ~ j g ∥ L ∞ = sup 0 < j < j ϵ 2 j α ∥ M ϵ , j ( 2 ) ∗ ( P ~ j g ) ∥ L ∞ ≤ sup 0 < j < j ϵ ∥ M ϵ , j ( 2 ) ∥ L 1 2 j α ∥ P ~ j g ∥ L ∞ ≤ c ∣ log ϵ ∣ − 1 ∣ g ∣ B ˙ ∞ , ∞ α . (189) \begin{aligned}
\sup_{0<j<j_\epsilon}2^{j(1+\alpha)}
\left\|P_j\widetilde{T}_{M_{\epsilon}^{(1)}}h\right\|_{L^\infty}
&=\sup_{0<j<j_\epsilon}2^{j(1+\alpha)}
\left\|P_j\widetilde{T}_{M_{\epsilon}^{(1)}}\widetilde{P}_jh\right\|_{L^\infty}\\
&=\sup_{0<j<j_\epsilon}2^{j(1+\alpha)}
\left\|M_{\epsilon,j}^{(1)}*(\widetilde{P}_jh)\right\|_{L^\infty}\\
&\leq2^{j_\epsilon}\sup_{0<j<j_\epsilon}
\left\|M_{\epsilon,j}^{(1)}\right\|_{L^1}2^{j\alpha}\left\|\widetilde{P}_jh\right\|_{L^\infty}
\leq c\epsilon^{-1}|\log\epsilon||h|_{\dot{B}^{\alpha}_{\infty,\infty}},
\\
\sup_{0<j<j_\epsilon}2^{j\alpha}
\left\|P_j\widetilde{T}_{M_{\epsilon}^{(2)}}g\right\|_{L^\infty}
&=\sup_{0<j<j_\epsilon}2^{j\alpha}
\left\|P_j\widetilde{T}_{M_{\epsilon}^{(2)}}\widetilde{P}_jg\right\|_{L^\infty}\\
&=\sup_{0<j<j_\epsilon}2^{j\alpha}
\left\|M_{\epsilon,j}^{(2)}*(\widetilde{P}_jg)\right\|_{L^\infty}\\
&\leq\sup_{0<j<j_\epsilon}
\left\|M_{\epsilon,j}^{(2)}\right\|_{L^1}2^{j\alpha}\left\|\widetilde{P}_jg\right\|_{L^\infty}
\leq c|\log\epsilon|^{-1}|g|_{\dot{B}^{\alpha}_{\infty,\infty}}. \tag*{(189)}
\end{aligned} 0 < j < j ϵ sup 2 j ( 1 + α ) P j T M ϵ ( 1 ) h L ∞ 0 < j < j ϵ sup 2 j α P j T M ϵ ( 2 ) g L ∞ = 0 < j < j ϵ sup 2 j ( 1 + α ) P j T M ϵ ( 1 ) P j h L ∞ = 0 < j < j ϵ sup 2 j ( 1 + α ) M ϵ , j ( 1 ) ∗ ( P j h ) L ∞ ≤ 2 j ϵ 0 < j < j ϵ sup M ϵ , j ( 1 ) L 1 2 j α P j h L ∞ ≤ c ϵ − 1 ∣ log ϵ ∣∣ h ∣ B ˙ ∞ , ∞ α , = 0 < j < j ϵ sup 2 j α P j T M ϵ ( 2 ) P j g L ∞ = 0 < j < j ϵ sup 2 j α M ϵ , j ( 2 ) ∗ ( P j g ) L ∞ ≤ 0 < j < j ϵ sup M ϵ , j ( 2 ) L 1 2 j α P j g L ∞ ≤ c ∣ log ϵ ∣ − 1 ∣ g ∣ B ˙ ∞ , ∞ α . (189)
For high frequencies j ≥ j ϵ j\geq j_\epsilon j ≥ j ϵ , we have
sup j ≥ j ϵ 2 j ( 1 + α ) ∥ P j T ~ M ϵ ( 1 ) h ∥ L ∞ ≤ sup j ≥ j ϵ 2 j ( 1 + α ) ∥ M ϵ , j ( 1 ) ∥ L 1 ∥ P ~ j h ∥ L ∞ ≤ c ϵ − 1 sup j ≥ j ϵ 2 j α ∥ P ~ j h ∥ L ∞ ≤ c ϵ − 1 ∣ h ∣ B ˙ ∞ , ∞ α , sup j ≥ j ϵ 2 j α ∥ P j T ~ M ϵ ( 2 ) g ∥ L ∞ ≤ sup j ≥ j ϵ 2 j α ∥ M ϵ , j ( 2 ) ∥ L 1 ∥ P ~ j g ∥ L ∞ ≤ c ϵ sup j ≥ j ϵ 2 j ( 1 + α ) ∥ P ~ j g ∥ L ∞ ≤ c ϵ ∣ g ∣ B ˙ ∞ , ∞ 1 + α . (190) \begin{aligned}
\sup_{j\geq j_\epsilon}2^{j(1+\alpha)}
\left\|P_j\widetilde{T}_{M_{\epsilon}^{(1)}}h\right\|_{L^\infty}
&\leq\sup_{j\geq j_\epsilon}2^{j(1+\alpha)}
\left\|M_{\epsilon,j}^{(1)}\right\|_{L^1}\left\|\widetilde{P}_jh\right\|_{L^\infty}\\
&\leq c\epsilon^{-1}\sup_{j\geq j_\epsilon}2^{j\alpha}\left\|\widetilde{P}_jh\right\|_{L^\infty}
\leq c\epsilon^{-1}|h|_{\dot{B}^{\alpha}_{\infty,\infty}},\\
\sup_{j\geq j_\epsilon}2^{j\alpha}
\left\|P_j\widetilde{T}_{M_{\epsilon}^{(2)}}g\right\|_{L^\infty}
&\leq\sup_{j\geq j_\epsilon}2^{j\alpha}
\left\|M_{\epsilon,j}^{(2)}\right\|_{L^1}\left\|\widetilde{P}_jg\right\|_{L^\infty}\\
&\leq c\epsilon\sup_{j\geq j_\epsilon}2^{j(1+\alpha)}
\left\|\widetilde{P}_jg\right\|_{L^\infty}
\leq c\epsilon|g|_{\dot{B}^{1+\alpha}_{\infty,\infty}}. \tag*{(190)}
\end{aligned} j ≥ j ϵ sup 2 j ( 1 + α ) P j T M ϵ ( 1 ) h L ∞ j ≥ j ϵ sup 2 j α P j T M ϵ ( 2 ) g L ∞ ≤ j ≥ j ϵ sup 2 j ( 1 + α ) M ϵ , j ( 1 ) L 1 P j h L ∞ ≤ c ϵ − 1 j ≥ j ϵ sup 2 j α P j h L ∞ ≤ c ϵ − 1 ∣ h ∣ B ˙ ∞ , ∞ α , ≤ j ≥ j ϵ sup 2 j α M ϵ , j ( 2 ) L 1 P j g L ∞ ≤ cϵ j ≥ j ϵ sup 2 j ( 1 + α ) P j g L ∞ ≤ cϵ ∣ g ∣ B ˙ ∞ , ∞ 1 + α . (190)
Combining the estimates (188) , (189) , and (190) and using the norm definition (107), we obtain Lemma 1.4. □
3.4.2. Proof of Lemma 1.9. The mapping properties of the operator A ~ ϵ P 01 \widetilde{\mathcal{A}}_\epsilon\mathbb{P}_{01} A ϵ P 01 may be shown by similar means to the previous section. Recall the definition (105) of ϕ j \phi_j ϕ j and the forms of the five distinct multipliers m ϵ , z z m_{\epsilon,z}^{z} m ϵ , z z , m ϵ , r z m_{\epsilon,r}^{z} m ϵ , r z , m ϵ , z x m_{\epsilon,z}^{x} m ϵ , z x , m ϵ , r x m_{\epsilon,r}^{x} m ϵ , r x , m ϵ , θ x m_{\epsilon,\theta}^{x} m ϵ , θ x appearing in Corollary 1.8 as the components of A ~ ϵ P 01 \widetilde{\mathcal{A}}_\epsilon\mathbb{P}_{01} A ϵ P 01 . We define the scalar-valued Fourier multipliers
M ^ ϵ , z , j z ( ξ ) = ϕ j ( ξ ) m ϵ , z z ( ξ ) , M ^ ϵ , r , j z ( ξ ) = ϕ j ( ξ ) m ϵ , r z ( ξ ) , \widehat{M}_{\epsilon,z,j}^{z}(\xi)=\phi_j(\xi)m_{\epsilon,z}^{z}(\xi),\qquad
\widehat{M}_{\epsilon,r,j}^{z}(\xi)=\phi_j(\xi)m_{\epsilon,r}^{z}(\xi), M ϵ , z , j z ( ξ ) = ϕ j ( ξ ) m ϵ , z z ( ξ ) , M ϵ , r , j z ( ξ ) = ϕ j ( ξ ) m ϵ , r z ( ξ ) ,
M ^ ϵ , z , j x ( ξ ) = ϕ j ( ξ ) m ϵ , z x ( ξ ) , M ^ ϵ , r , j x ( ξ ) = ϕ j ( ξ ) m ϵ , r x ( ξ ) , M ^ ϵ , θ , j x ( ξ ) = ϕ j ( ξ ) m ϵ , θ x ( ξ ) . \widehat{M}_{\epsilon,z,j}^{x}(\xi)=\phi_j(\xi)m_{\epsilon,z}^{x}(\xi),\qquad
\widehat{M}_{\epsilon,r,j}^{x}(\xi)=\phi_j(\xi)m_{\epsilon,r}^{x}(\xi),\qquad
\widehat{M}_{\epsilon,\theta,j}^{x}(\xi)=\phi_j(\xi)m_{\epsilon,\theta}^{x}(\xi). M ϵ , z , j x ( ξ ) = ϕ j ( ξ ) m ϵ , z x ( ξ ) , M ϵ , r , j x ( ξ ) = ϕ j ( ξ ) m ϵ , r x ( ξ ) , M ϵ , θ , j x ( ξ ) = ϕ j ( ξ ) m ϵ , θ x ( ξ ) .
As in the proof of Lemma 1.4, we define j ϵ = ∣ log ( 2 π ϵ ) ∣ log ( 2 ) j_{\epsilon}=\frac{|\log(2\pi\epsilon)|}{\log(2)} j ϵ = l o g ( 2 ) ∣ l o g ( 2 π ϵ ) ∣ . Using Lemmas 3.9 and 3.10, we have that three of the multipliers M ^ ϵ , z , j z \widehat{M}^{z}_{\epsilon,z,j} M ϵ , z , j z , M ^ ϵ , r , j x \widehat{M}^{x}_{\epsilon,r,j} M ϵ , r , j x , and M ^ ϵ , θ , j x \widehat{M}^{x}_{\epsilon,\theta,j} M ϵ , θ , j x each satisfy the same set of estimates, namely
∥ ∂ ϵ ℓ M ^ ϵ , z , j z ∥ L ∞ , ∥ ∂ ϵ ℓ M ^ ϵ , r , j x ∥ L ∞ , ∥ ∂ ϵ ℓ M ^ ϵ , θ , j x ∥ L ∞ ≤ { c ϵ 2 j ( 1 − ℓ ) , j ≥ j ϵ c 2 − j ℓ , j < j ϵ , ℓ = 0 , 1 , 2. \left\|\partial_{\epsilon}^{\ell}\widehat{M}^{z}_{\epsilon,z,j}\right\|_{L^{\infty}},\left\|\partial_{\epsilon}^{\ell}\widehat{M}^{x}_{\epsilon,r,j}\right\|_{L^{\infty}},\left\|\partial_{\epsilon}^{\ell}\widehat{M}^{x}_{\epsilon,\theta,j}\right\|_{L^{\infty}}\leq
\begin{cases}
c\epsilon2^{j(1-\ell)}, & j\geq j_{\epsilon}\\
c2^{-j\ell}, & j<j_{\epsilon},
\end{cases}
\qquad\ell=0,1,2. ∂ ϵ ℓ M ϵ , z , j z L ∞ , ∂ ϵ ℓ M ϵ , r , j x L ∞ , ∂ ϵ ℓ M ϵ , θ , j x L ∞ ≤ { cϵ 2 j ( 1 − ℓ ) , c 2 − j ℓ , j ≥ j ϵ j < j ϵ , ℓ = 0 , 1 , 2.
These multipliers will comprise A ‾ ϵ ϵ \overline{\mathcal{A}}_{\epsilon}^{\epsilon} A ϵ ϵ , the leading order behavior (with respect to regularity) of the angle-averaged single layer operator A ‾ ϵ P 01 \overline{\mathcal{A}}_{\epsilon}\mathbb{P}_{01} A ϵ P 01 . The remaining multipliers M ^ ϵ , r , j z \widehat{M}^{z}_{\epsilon,r,j} M ϵ , r , j z and M ^ ϵ , z , j x \widehat{M}^{x}_{\epsilon,z,j} M ϵ , z , j x satisfy a different set of estimates:
∥ ∂ ϵ ℓ M ^ ϵ , r , j z ∥ L ∞ ≤ { c , j ≥ j ϵ , ℓ = 0 c 2 − j , j ≥ j ϵ , ℓ = 1 , 2 c ϵ − 1 2 − j ( 1 + ℓ ) , j < j ϵ , ℓ = 0 , 1 , 2 , \left\|\partial_{\epsilon}^{\ell}\widehat{M}^{z}_{\epsilon,r,j}\right\|_{L^{\infty}}\leq
\begin{cases}
c, & j\geq j_{\epsilon},\ \ell=0\\
c2^{-j}, & j\geq j_{\epsilon},\ \ell=1,2\\
c\epsilon^{-1}2^{-j(1+\ell)}, & j<j_{\epsilon},\ \ell=0,1,2,
\end{cases} ∂ ϵ ℓ M ϵ , r , j z L ∞ ≤ ⎩ ⎨ ⎧ c , c 2 − j , c ϵ − 1 2 − j ( 1 + ℓ ) , j ≥ j ϵ , ℓ = 0 j ≥ j ϵ , ℓ = 1 , 2 j < j ϵ , ℓ = 0 , 1 , 2 ,
∥ ∂ ϵ ℓ M ^ ϵ , z , j x ∥ L ∞ ≤ { c 2 − j ℓ , j ≥ j ϵ c ϵ ∣ log ϵ ∣ 2 j ( 1 − ℓ ) , j < j ϵ . \left\|\partial_{\epsilon}^{\ell}\widehat{M}^{x}_{\epsilon,z,j}\right\|_{L^{\infty}}\leq
\begin{cases}
c2^{-j\ell}, & j\geq j_{\epsilon}\\
c\epsilon|\log\epsilon|2^{j(1-\ell)}, & j<j_{\epsilon}.
\end{cases} ∂ ϵ ℓ M ϵ , z , j x L ∞ ≤ { c 2 − j ℓ , cϵ ∣ log ϵ ∣ 2 j ( 1 − ℓ ) , j ≥ j ϵ j < j ϵ .
These multipliers will make up the operator A ‾ ϵ + \overline{\mathcal{A}}_{\epsilon}^{+} A ϵ + , which has ‘bad’ ϵ \epsilon ϵ -dependence but is smoother.
Using the scalar-valued version of Lemma 3.1 (see [59 Lemma 2.1] ), we may obtain the following physical space estimates for the inverse Fourier transforms of each M ^ ϵ , j \widehat{M}_{\epsilon,j} M ϵ , j :
∥ M ϵ , z , j z ∥ L 1 , ∥ M ϵ , r , j x ∥ L 1 , ∥ M ϵ , θ , j x ∥ L 1 ≤ { c ϵ 2 j , j ≥ j ϵ c , j < j ϵ , \left\|M^{z}_{\epsilon,z,j}\right\|_{L^{1}},\left\|M^{x}_{\epsilon,r,j}\right\|_{L^{1}},\left\|M^{x}_{\epsilon,\theta,j}\right\|_{L^{1}}\leq
\begin{cases}
c\epsilon2^{j}, & j\geq j_{\epsilon}\\
c, & j<j_{\epsilon},
\end{cases} M ϵ , z , j z L 1 , M ϵ , r , j x L 1 , M ϵ , θ , j x L 1 ≤ { cϵ 2 j , c , j ≥ j ϵ j < j ϵ ,
∥ M ϵ , r , j z ∥ L 1 ≤ { c 2 − j / 2 , j ≥ j ϵ c ϵ − 1 2 − j , j < j ϵ , \left\|M^{z}_{\epsilon,r,j}\right\|_{L^{1}}\leq
\begin{cases}
c2^{-j/2}, & j\geq j_{\epsilon}\\
c\epsilon^{-1}2^{-j}, & j<j_{\epsilon},
\end{cases} M ϵ , r , j z L 1 ≤ { c 2 − j /2 , c ϵ − 1 2 − j , j ≥ j ϵ j < j ϵ ,
∥ M ϵ , z , j x ∥ L 1 ≤ { c , j ≥ j ϵ c ϵ ∣ log ϵ ∣ 2 j , j < j ϵ . (191) \left\|M^{x}_{\epsilon,z,j}\right\|_{L^{1}}\leq
\begin{cases}
c, & j\geq j_{\epsilon}\\
c\epsilon|\log\epsilon|2^{j}, & j<j_{\epsilon}.
\end{cases} \tag*{(191)} M ϵ , z , j x L 1 ≤ { c , cϵ ∣ log ϵ ∣ 2 j , j ≥ j ϵ j < j ϵ . (191)
It will now be useful to further decompose the projection operator P 01 \mathbb{P}_{01} P 01 as follows. Given g ( s , θ ) g(s,\theta) g ( s , θ ) defined along C ϵ \mathcal{C}_{\epsilon} C ϵ , we may consider P 01 g \mathbb{P}_{01}g P 01 g as in (62) and we may further define the following projections:
P 0 z g = g 0 z e z , P 0 r g = g 0 r e r , \mathbb{P}_{0}^{z}g=g_{0}^{z}e_{z},\qquad\mathbb{P}_{0}^{r}g=g_{0}^{r}e_{r}, P 0 z g = g 0 z e z , P 0 r g = g 0 r e r ,
P 1 z g = g c , 1 z cos θ e z + g s , 1 z sin θ e z , \mathbb{P}_{1}^{z}g=g_{c,1}^{z}\cos\theta e_{z}+g_{s,1}^{z}\sin\theta e_{z}, P 1 z g = g c , 1 z cos θ e z + g s , 1 z sin θ e z ,
P 1 r g = g c , 1 r cos θ e r + g s , 1 r sin θ e r , \mathbb{P}_{1}^{r}g=g_{c,1}^{r}\cos\theta e_{r}+g_{s,1}^{r}\sin\theta e_{r}, P 1 r g = g c , 1 r cos θ e r + g s , 1 r sin θ e r ,
P 1 θ g = g s , 1 θ sin θ e θ + g c , 1 θ cos θ e θ . \mathbb{P}_{1}^{\theta}g=g_{s,1}^{\theta}\sin\theta e_{\theta}+g_{c,1}^{\theta}\cos\theta e_{\theta}. P 1 θ g = g s , 1 θ sin θ e θ + g c , 1 θ cos θ e θ .
We then decompose A ‾ ϵ P 01 \overline{\mathcal{A}}_{\epsilon}\mathbb{P}_{01} A ϵ P 01 as A ‾ ϵ P 01 = A ‾ ϵ ϵ + A ‾ ϵ + \overline{\mathcal{A}}_{\epsilon}\mathbb{P}_{01}=\overline{\mathcal{A}}_{\epsilon}^{\epsilon}+\overline{\mathcal{A}}_{\epsilon}^{+} A ϵ P 01 = A ϵ ϵ + A ϵ + where
A ‾ ϵ ϵ = A ‾ ϵ ( P 0 z + P 1 r + P 1 θ ) , A ‾ ϵ + = A ‾ ϵ ( P 0 r + P 1 z ) . \overline{\mathcal{A}}_{\epsilon}^{\epsilon}=\overline{\mathcal{A}}_{\epsilon}\left(\mathbb{P}_{0}^{z}+\mathbb{P}_{1}^{r}+\mathbb{P}_{1}^{\theta}\right),\qquad
\overline{\mathcal{A}}_{\epsilon}^{+}=\overline{\mathcal{A}}_{\epsilon}\left(\mathbb{P}_{0}^{r}+\mathbb{P}_{1}^{z}\right). A ϵ ϵ = A ϵ ( P 0 z + P 1 r + P 1 θ ) , A ϵ + = A ϵ ( P 0 r + P 1 z ) .
For g ∈ C s 1 , α ( Γ ϵ ) g \in C_s^{1,\alpha}(\Gamma_\epsilon) g ∈ C s 1 , α ( Γ ϵ ) , we use the dyadic partition of unity (105) to write
A ‾ ϵ g = T M ϵ , z , j z P 0 z g + T M ϵ , r , j x P 1 r g + T M ϵ , θ , j x P 1 θ g = ∑ j P j ( T M ϵ , z , j z P 0 z g + T M ϵ , r , j x P 1 r g + T M ϵ , θ , j x P 1 θ g ) , \begin{aligned}
\overline{\mathcal{A}}_\epsilon g=T M_{\epsilon,z,j}^{z}\mathbb{P}_0^z g+T M_{\epsilon,r,j}^{x}\mathbb{P}_1^r g+T M_{\epsilon,\theta,j}^{x}\mathbb{P}_1^\theta g \\
&=\sum_j P_j\left(T M_{\epsilon,z,j}^{z}\mathbb{P}_0^z g+T M_{\epsilon,r,j}^{x}\mathbb{P}_1^r g+T M_{\epsilon,\theta,j}^{x}\mathbb{P}_1^\theta g\right),
\end{aligned} A ϵ g = T M ϵ , z , j z P 0 z g + T M ϵ , r , j x P 1 r g + T M ϵ , θ , j x P 1 θ g = j ∑ P j ( T M ϵ , z , j z P 0 z g + T M ϵ , r , j x P 1 r g + T M ϵ , θ , j x P 1 θ g ) ,
A ‾ ϵ + g = T M ϵ , r , j x P 0 r g + T M ϵ , z , j x P 1 z g = ∑ j P j ( T M ϵ , r , j x P 0 r g + T M ϵ , z , j x P 1 z g ) . (192) \begin{aligned}
\overline{\mathcal{A}}_\epsilon^+g=T M_{\epsilon,r,j}^{x}\mathbb{P}_0^r g+T M_{\epsilon,z,j}^{x}\mathbb{P}_1^z g \\
&=\sum_j P_j\left(T M_{\epsilon,r,j}^{x}\mathbb{P}_0^r g+T M_{\epsilon,z,j}^{x}\mathbb{P}_1^z g\right). \tag*{(192)}
\end{aligned} A ϵ + g = T M ϵ , r , j x P 0 r g + T M ϵ , z , j x P 1 z g = j ∑ P j ( T M ϵ , r , j x P 0 r g + T M ϵ , z , j x P 1 z g ) . (192)
We may now obtain bounds for A ‾ ϵ g \overline{\mathcal{A}}_\epsilon g A ϵ g and A ‾ ϵ + g \overline{\mathcal{A}}_\epsilon^+g A ϵ + g . For each multiplier M M M , we have P ≤ 0 T M g = P 0 T M g P_{\leq0}TMg=P_0TMg P ≤ 0 T M g = P 0 T M g , so using the L 1 L^1 L 1 bounds (190) we may estimate each of the components of A ‾ ϵ \overline{\mathcal{A}}_\epsilon A ϵ as
∥ P ≤ 0 T M ϵ , z , j z P 0 z g ∥ L ∞ = ∥ M ϵ , z , 0 z ∗ ( P 0 z g ) ∥ L ∞ ≤ c ∥ M ϵ , z , 0 z ∥ L 1 ∥ g ∥ L ∞ ≤ c ∥ g ∥ L ∞ , \left\|P_{\leq0}TM_{\epsilon,z,j}^{z}\mathbb{P}_0^z g\right\|_{L^\infty}
=\left\|M_{\epsilon,z,0}^{z}*(\mathbb{P}_0^z g)\right\|_{L^\infty}
\leq c\left\|M_{\epsilon,z,0}^{z}\right\|_{L^1}\|g\|_{L^\infty}
\leq c\|g\|_{L^\infty}, P ≤ 0 T M ϵ , z , j z P 0 z g L ∞ = M ϵ , z , 0 z ∗ ( P 0 z g ) L ∞ ≤ c M ϵ , z , 0 z L 1 ∥ g ∥ L ∞ ≤ c ∥ g ∥ L ∞ ,
∥ P ≤ 0 T M ϵ , r , j x P 1 r g ∥ L ∞ ≤ c ∥ g ∥ L ∞ , ∥ P ≤ 0 T M ϵ , θ , j x P 1 θ g ∥ L ∞ ≤ c ∥ g ∥ L ∞ . \left\|P_{\leq0}TM_{\epsilon,r,j}^{x}\mathbb{P}_1^r g\right\|_{L^\infty}
\leq c\|g\|_{L^\infty},
\qquad
\left\|P_{\leq0}TM_{\epsilon,\theta,j}^{x}\mathbb{P}_1^\theta g\right\|_{L^\infty}
\leq c\|g\|_{L^\infty}. P ≤ 0 T M ϵ , r , j x P 1 r g L ∞ ≤ c ∥ g ∥ L ∞ , P ≤ 0 T M ϵ , θ , j x P 1 θ g L ∞ ≤ c ∥ g ∥ L ∞ .
Here and throughout, the convolution ∗ * ∗ is with respect to s s s only. The components of A ‾ ϵ + \overline{\mathcal{A}}_\epsilon^+ A ϵ + may be similarly estimated as
∥ P ≤ 0 T M ϵ , r , j z P 0 r g ∥ L ∞ = ∥ M ϵ , r , 0 z ∗ ( P 0 r g ) ∥ L ∞ ≤ c ∥ M ϵ , r , 0 z ∥ L 1 ∥ g ∥ L ∞ ≤ c ϵ − 1 ∥ g ∥ L ∞ , \left\|P_{\leq0}TM_{\epsilon,r,j}^{z}\mathbb{P}_0^r g\right\|_{L^\infty}
=\left\|M_{\epsilon,r,0}^{z}*(\mathbb{P}_0^r g)\right\|_{L^\infty}
\leq c\left\|M_{\epsilon,r,0}^{z}\right\|_{L^1}\|g\|_{L^\infty}
\leq c\epsilon^{-1}\|g\|_{L^\infty}, P ≤ 0 T M ϵ , r , j z P 0 r g L ∞ = M ϵ , r , 0 z ∗ ( P 0 r g ) L ∞ ≤ c M ϵ , r , 0 z L 1 ∥ g ∥ L ∞ ≤ c ϵ − 1 ∥ g ∥ L ∞ ,
∥ P ≤ 0 T M ϵ , z , j x P 1 z g ∥ L ∞ = ∥ M ϵ , z , 0 x ∗ ( P 1 z g ) ∥ L ∞ ≤ c ∥ M ϵ , z , 0 x ∥ L 1 ∥ g ∥ L ∞ ≤ c ϵ ∣ log ϵ ∣ ∥ g ∥ L ∞ . \left\|P_{\leq0}TM_{\epsilon,z,j}^{x}\mathbb{P}_1^z g\right\|_{L^\infty}
=\left\|M_{\epsilon,z,0}^{x}*(\mathbb{P}_1^z g)\right\|_{L^\infty}
\leq c\left\|M_{\epsilon,z,0}^{x}\right\|_{L^1}\|g\|_{L^\infty}
\leq c\epsilon|\log\epsilon|\|g\|_{L^\infty}. P ≤ 0 T M ϵ , z , j x P 1 z g L ∞ = M ϵ , z , 0 x ∗ ( P 1 z g ) L ∞ ≤ c M ϵ , z , 0 x L 1 ∥ g ∥ L ∞ ≤ cϵ ∣ log ϵ ∣∥ g ∥ L ∞ .
We next bound the low frequencies 0 < j < j ϵ 0<j<j_\epsilon 0 < j < j ϵ . As in the proof of Lemma 1.4, we will use the notation P ~ j g = ( P j − 1 + P j + P j + 1 ) g \widetilde{P}_jg=(P_{j-1}+P_j+P_{j+1})g P j g = ( P j − 1 + P j + P j + 1 ) g and again note that P j T m g = P j T m P ~ j g P_jT_mg=P_jT_m\widetilde{P}_jg P j T m g = P j T m P j g .
Using (190), we may estimate the first component of A ‾ ϵ \overline{\mathcal{A}}_\epsilon A ϵ as
sup 0 < j < j ϵ 2 j α ∥ P j T M ϵ , z , j z P 0 z g ∥ L ∞ ≤ sup 0 < j < j ϵ 2 j α ∥ P j T M ϵ , z , j z P ~ j ( P 0 z g ) ∥ L ∞ = sup 0 < j < j ϵ 2 j α ∥ M ϵ , z , j z ∗ ( P ~ j P 0 z g ) ∥ L ∞ ≤ sup 0 < j < j ϵ 2 j α ∥ M ϵ , z , j z ∥ L 1 ∥ P ~ j P 0 z g ∥ L ∞ ≤ c sup 0 < j < j ϵ 2 j α ∥ P ~ j P 0 z g ∥ L ∞ ≤ c ∣ g ∣ ( B ˙ ∞ , ∞ α ) s . \begin{aligned}
\sup_{0<j<j_\epsilon}2^{j\alpha}\left\|P_jTM_{\epsilon,z,j}^{z}\mathbb{P}_0^z g\right\|_{L^\infty}
\leq
\sup_{0<j<j_\epsilon}2^{j\alpha}\left\|P_jTM_{\epsilon,z,j}^{z}\widetilde{P}_j(\mathbb{P}_0^z g)\right\|_{L^\infty} \\
&=
\sup_{0<j<j_\epsilon}2^{j\alpha}\left\|M_{\epsilon,z,j}^{z}*(\widetilde{P}_j\mathbb{P}_0^z g)\right\|_{L^\infty} \\
&\leq
\sup_{0<j<j_\epsilon}2^{j\alpha}\left\|M_{\epsilon,z,j}^{z}\right\|_{L^1}\left\|\widetilde{P}_j\mathbb{P}_0^z g\right\|_{L^\infty} \\
&\leq c\sup_{0<j<j_\epsilon}2^{j\alpha}\left\|\widetilde{P}_j\mathbb{P}_0^z g\right\|_{L^\infty}
\leq c|g|_{(\dot{B}^{\alpha}_{\infty,\infty})_s}.
\end{aligned} 0 < j < j ϵ sup 2 j α P j T M ϵ , z , j z P 0 z g L ∞ ≤ 0 < j < j ϵ sup 2 j α P j T M ϵ , z , j z P j ( P 0 z g ) L ∞ = 0 < j < j ϵ sup 2 j α M ϵ , z , j z ∗ ( P j P 0 z g ) L ∞ ≤ 0 < j < j ϵ sup 2 j α M ϵ , z , j z L 1 P j P 0 z g L ∞ ≤ c 0 < j < j ϵ sup 2 j α P j P 0 z g L ∞ ≤ c ∣ g ∣ ( B ˙ ∞ , ∞ α ) s .
Likewise, the remaining two terms of A ‾ ϵ \overline{\mathcal{A}}_\epsilon A ϵ satisfy
sup 0 < j < j ϵ 2 j α ∥ P j T M ϵ , r , j x P 1 r g ∥ L ∞ ≤ c ∣ g ∣ ( B ˙ ∞ , ∞ α ) s , sup 0 < j < j ϵ 2 j α ∥ P j T M ϵ , θ , j x P 1 θ g ∥ L ∞ ≤ c ∣ g ∣ ( B ˙ ∞ , ∞ α ) s . \sup_{0<j<j_\epsilon}2^{j\alpha}\left\|P_jTM_{\epsilon,r,j}^{x}\mathbb{P}_1^r g\right\|_{L^\infty}
\leq c|g|_{(\dot{B}^{\alpha}_{\infty,\infty})_s},
\qquad
\sup_{0<j<j_\epsilon}2^{j\alpha}\left\|P_jTM_{\epsilon,\theta,j}^{x}\mathbb{P}_1^\theta g\right\|_{L^\infty}
\leq c|g|_{(\dot{B}^{\alpha}_{\infty,\infty})_s}. 0 < j < j ϵ sup 2 j α P j T M ϵ , r , j x P 1 r g L ∞ ≤ c ∣ g ∣ ( B ˙ ∞ , ∞ α ) s , 0 < j < j ϵ sup 2 j α P j T M ϵ , θ , j x P 1 θ g L ∞ ≤ c ∣ g ∣ ( B ˙ ∞ , ∞ α ) s .
In addition, the two components of A ‾ ϵ + \overline{\mathcal{A}}_\epsilon^+ A ϵ + satisfy
sup 0 < j < j ϵ 2 j ( 3 / 2 + α ) ∥ P j T M ϵ , r , j z P 0 r g ∥ L ∞ ≤ sup 0 < j < j ϵ 2 j ( 3 / 2 + α ) ∥ M ϵ , r , j z ∥ L 1 ∥ P ~ j P 0 r g ∥ L ∞ ≤ c ϵ − 1 sup 0 < j < j ϵ 2 j ( 1 / 2 + α ) ∥ P ~ j P 0 r g ∥ L ∞ ≤ c ϵ − 1 ∣ g ∣ ( B ˙ ∞ , ∞ 1 / 2 + α ) s , \begin{aligned}
\sup_{0<j<j_\epsilon}2^{j(3/2+\alpha)}\left\|P_jTM_{\epsilon,r,j}^{z}\mathbb{P}_0^r g\right\|_{L^\infty}
\leq
\sup_{0<j<j_\epsilon}2^{j(3/2+\alpha)}\left\|M_{\epsilon,r,j}^{z}\right\|_{L^1}\left\|\widetilde{P}_j\mathbb{P}_0^r g\right\|_{L^\infty} \\
&\leq c\epsilon^{-1}\sup_{0<j<j_\epsilon}2^{j(1/2+\alpha)}\left\|\widetilde{P}_j\mathbb{P}_0^r g\right\|_{L^\infty}
\leq c\epsilon^{-1}|g|_{(\dot{B}^{1/2+\alpha}_{\infty,\infty})_s},
\end{aligned} 0 < j < j ϵ sup 2 j ( 3/2 + α ) P j T M ϵ , r , j z P 0 r g L ∞ ≤ 0 < j < j ϵ sup 2 j ( 3/2 + α ) M ϵ , r , j z L 1 P j P 0 r g L ∞ ≤ c ϵ − 1 0 < j < j ϵ sup 2 j ( 1/2 + α ) P j P 0 r g L ∞ ≤ c ϵ − 1 ∣ g ∣ ( B ˙ ∞ , ∞ 1/2 + α ) s ,
sup 0 < j < j ϵ 2 j ( 1 + α ) ∥ P j T M ϵ , z , j x P 1 z g ∥ L ∞ ≤ sup 0 < j < j ϵ 2 j ( 1 + α ) ∥ M ϵ , z , j x ∥ L 1 ∥ P ~ j P 1 z g ∥ L ∞ ≤ c ϵ ∣ log ϵ ∣ sup 0 < j < j ϵ 2 j ( 2 + α ) ∥ P ~ j P 1 z g ∥ L ∞ ≤ c ϵ 2 j ϵ ∣ log ϵ ∣ sup 0 < j < j ϵ 2 j ( 1 + α ) ∥ P ~ j P j z g ∥ L ∞ ≤ c ∣ log ϵ ∣ ∥ g ∥ B ˙ ∞ , ∞ 1 + α . \begin{aligned}
\sup_{0<j<j_\epsilon}2^{j(1+\alpha)}\left\|P_jTM_{\epsilon,z,j}^{x}\mathbb{P}_1^z g\right\|_{L^\infty}
\leq
\sup_{0<j<j_\epsilon}2^{j(1+\alpha)}\left\|M_{\epsilon,z,j}^{x}\right\|_{L^1}\left\|\widetilde{P}_j\mathbb{P}_1^z g\right\|_{L^\infty} \\
&\leq c\epsilon|\log\epsilon|\sup_{0<j<j_\epsilon}2^{j(2+\alpha)}\left\|\widetilde{P}_j\mathbb{P}_1^z g\right\|_{L^\infty} \\
&\le c\epsilon2^{j_\epsilon}|\log\epsilon|\sup_{0<j<j_\epsilon}2^{j(1+\alpha)}\left\|\widetilde P_j\mathbb{P}_j^z g\right\|_{L^\infty}\le c|\log\epsilon|\|g\|_{\dot B^{1+\alpha}_{\infty,\infty}}.
\end{aligned} 0 < j < j ϵ sup 2 j ( 1 + α ) P j T M ϵ , z , j x P 1 z g L ∞ ≤ 0 < j < j ϵ sup 2 j ( 1 + α ) M ϵ , z , j x L 1 P j P 1 z g L ∞ ≤ cϵ ∣ log ϵ ∣ 0 < j < j ϵ sup 2 j ( 2 + α ) P j P 1 z g L ∞ ≤ cϵ 2 j ϵ ∣ log ϵ ∣ 0 < j < j ϵ sup 2 j ( 1 + α ) P j P j z g L ∞ ≤ c ∣ log ϵ ∣∥ g ∥ B ˙ ∞ , ∞ 1 + α .
Here we have use that 2 j ϵ = 1 2 π ϵ 2^{j_\epsilon}=\frac{1}{2\pi\epsilon} 2 j ϵ = 2 π ϵ 1 .
Finally, for high frequencies j ≥ j ϵ j\ge j_\epsilon j ≥ j ϵ , we may use the L 1 L^1 L 1 estimates (191) to show that the components of A ‾ ϵ − \overline A_\epsilon^- A ϵ − each satisfy
sup j ≥ j ϵ 2 j α ∥ P j T M ϵ , z , j z P j z g ∥ L ∞ ≤ sup j ≥ j ϵ 2 j α ∥ P j T M ϵ , z , j z P ~ j ( P z g ) ∥ L ∞ = sup j ≥ j ϵ 2 j α ∥ M ϵ , z , j z ∗ ( P ~ j P z g ) ∥ L ∞ ≤ sup j ≥ j ϵ 2 j α ∥ M ϵ , z , j z ∥ L 1 ∥ P ~ j P j z g ∥ L ∞ ≤ c ϵ sup j ≥ j ϵ 2 j ( 1 + α ) ∥ P ~ j P j z g ∥ L ∞ ≤ c ϵ ∥ g ∥ B ˙ ∞ , ∞ 1 + α , \begin{aligned}
\sup_{j\ge j_\epsilon}2^{j\alpha}\left\|P_jT M_{\epsilon,z,j}^z\mathbb{P}_j^z g\right\|_{L^\infty}
\le\sup_{j\ge j_\epsilon}2^{j\alpha}\left\|P_jT M_{\epsilon,z,j}^z\widetilde P_j(\mathbb{P}^z g)\right\|_{L^\infty} \\
&= \sup_{j\ge j_\epsilon}2^{j\alpha}\left\|M_{\epsilon,z,j}^z*(\widetilde P_j\mathbb{P}^z g)\right\|_{L^\infty} \\
&\le\sup_{j\ge j_\epsilon}2^{j\alpha}\left\|M_{\epsilon,z,j}^z\right\|_{L^1}\left\|\widetilde P_j\mathbb{P}_j^z g\right\|_{L^\infty} \\
&\le c\epsilon\sup_{j\ge j_\epsilon}2^{j(1+\alpha)}\left\|\widetilde P_j\mathbb{P}_j^z g\right\|_{L^\infty}
\le c\epsilon\|g\|_{\dot B^{1+\alpha}_{\infty,\infty}},
\end{aligned} j ≥ j ϵ sup 2 j α P j T M ϵ , z , j z P j z g L ∞ ≤ j ≥ j ϵ sup 2 j α P j T M ϵ , z , j z P j ( P z g ) L ∞ = j ≥ j ϵ sup 2 j α M ϵ , z , j z ∗ ( P j P z g ) L ∞ ≤ j ≥ j ϵ sup 2 j α M ϵ , z , j z L 1 P j P j z g L ∞ ≤ cϵ j ≥ j ϵ sup 2 j ( 1 + α ) P j P j z g L ∞ ≤ cϵ ∥ g ∥ B ˙ ∞ , ∞ 1 + α ,
sup j ≥ j ϵ 2 j α ∥ P j T M ϵ , r , j x P j r g ∥ L ∞ ≤ c ϵ ∥ g ∥ B ˙ ∞ , ∞ 1 + α , sup j ≥ j ϵ 2 j α ∥ P j T M ϵ , θ , j x P j θ g ∥ L ∞ ≤ c ϵ ∥ g ∥ B ˙ ∞ , ∞ 1 + α . \sup_{j\ge j_\epsilon}2^{j\alpha}\left\|P_jT M_{\epsilon,r,j}^x\mathbb{P}_j^r g\right\|_{L^\infty}
\le c\epsilon\|g\|_{\dot B^{1+\alpha}_{\infty,\infty}},
\qquad
\sup_{j\ge j_\epsilon}2^{j\alpha}\left\|P_jT M_{\epsilon,\theta,j}^x\mathbb{P}_j^\theta g\right\|_{L^\infty}
\le c\epsilon\|g\|_{\dot B^{1+\alpha}_{\infty,\infty}}. j ≥ j ϵ sup 2 j α P j T M ϵ , r , j x P j r g L ∞ ≤ cϵ ∥ g ∥ B ˙ ∞ , ∞ 1 + α , j ≥ j ϵ sup 2 j α P j T M ϵ , θ , j x P j θ g L ∞ ≤ cϵ ∥ g ∥ B ˙ ∞ , ∞ 1 + α .
For A ‾ ϵ + \overline A_\epsilon^+ A ϵ + , we have
sup j ≥ j ϵ 2 j ( 3 / 2 + α ) ∥ P j T M ϵ , r , j x P j r g ∥ L ∞ ≤ sup j ≥ j ϵ 2 j ( 3 / 2 + α ) ∥ M ϵ , r , j x ∥ L 1 ∥ P ~ j P j r g ∥ L ∞ ≤ c sup j ≥ j ϵ 2 j ( 1 + α ) ∥ P ~ j P j r g ∥ L ∞ ≤ c ∥ g ∥ B ˙ ∞ , ∞ 1 + α , \begin{aligned}
\sup_{j\ge j_\epsilon}2^{j(3/2+\alpha)}\left\|P_jT M_{\epsilon,r,j}^x\mathbb{P}_j^r g\right\|_{L^\infty}
\le\sup_{j\ge j_\epsilon}2^{j(3/2+\alpha)}\left\|M_{\epsilon,r,j}^x\right\|_{L^1}\left\|\widetilde P_j\mathbb{P}_j^r g\right\|_{L^\infty} \\
&\le c\sup_{j\ge j_\epsilon}2^{j(1+\alpha)}\left\|\widetilde P_j\mathbb{P}_j^r g\right\|_{L^\infty}
\le c\|g\|_{\dot B^{1+\alpha}_{\infty,\infty}},
\end{aligned} j ≥ j ϵ sup 2 j ( 3/2 + α ) P j T M ϵ , r , j x P j r g L ∞ ≤ j ≥ j ϵ sup 2 j ( 3/2 + α ) M ϵ , r , j x L 1 P j P j r g L ∞ ≤ c j ≥ j ϵ sup 2 j ( 1 + α ) P j P j r g L ∞ ≤ c ∥ g ∥ B ˙ ∞ , ∞ 1 + α ,
sup j ≥ j ϵ 2 j ( 1 + α ) ∥ P j T M ϵ , z , j x P j z g ∥ L ∞ ≤ sup j ≥ j ϵ 2 j ( 1 + α ) ∥ M ϵ , z , j x ∥ L 1 ∥ P ~ j P j z g ∥ L ∞ ≤ c sup j ≥ j ϵ 2 j ( 1 + α ) ∥ P ~ j P j z g ∥ L ∞ ≤ c ∥ g ∥ B ˙ ∞ , ∞ 1 + α . \begin{aligned}
\sup_{j\ge j_\epsilon}2^{j(1+\alpha)}\left\|P_jT M_{\epsilon,z,j}^x\mathbb{P}_j^z g\right\|_{L^\infty}
\le\sup_{j\ge j_\epsilon}2^{j(1+\alpha)}\left\|M_{\epsilon,z,j}^x\right\|_{L^1}\left\|\widetilde P_j\mathbb{P}_j^z g\right\|_{L^\infty} \\
&\le c\sup_{j\ge j_\epsilon}2^{j(1+\alpha)}\left\|\widetilde P_j\mathbb{P}_j^z g\right\|_{L^\infty}
\le c\|g\|_{\dot B^{1+\alpha}_{\infty,\infty}}.
\end{aligned} j ≥ j ϵ sup 2 j ( 1 + α ) P j T M ϵ , z , j x P j z g L ∞ ≤ j ≥ j ϵ sup 2 j ( 1 + α ) M ϵ , z , j x L 1 P j P j z g L ∞ ≤ c j ≥ j ϵ sup 2 j ( 1 + α ) P j P j z g L ∞ ≤ c ∥ g ∥ B ˙ ∞ , ∞ 1 + α .
Returning to the definitions (192) of A ‾ ϵ − \overline A_\epsilon^- A ϵ − and A ‾ ϵ + \overline A_\epsilon^+ A ϵ + , we may combine the above estimates and use the Besov characterization (108) of C s 0 , α C_s^{0,\alpha} C s 0 , α to obtain
∥ A ‾ ϵ − g ∥ C 0 0 , α ( T ) ≤ ∥ T M ϵ , z , j x P j z g ∥ C 0 0 , α ( T ) + ∥ T M ϵ , r , j x P j r g ∥ C 0 0 , α ( T ) + ∥ T M ϵ , θ , j x P j θ g ∥ C 0 0 , α ( T ) ≤ c ∥ g ∥ C s 0 , α ( C ϵ ) + c ϵ ∥ g ∥ C s 1 , α ( C ϵ ) , \begin{aligned}
\left\|\overline A_\epsilon^-g\right\|_{C^{0,\alpha}_0(\mathbb{T})}
\le\left\|T M_{\epsilon,z,j}^x\mathbb{P}_j^z g\right\|_{C^{0,\alpha}_0(\mathbb{T})}
+\left\|T M_{\epsilon,r,j}^x\mathbb{P}_j^r g\right\|_{C^{0,\alpha}_0(\mathbb{T})}
+\left\|T M_{\epsilon,\theta,j}^x\mathbb{P}_j^\theta g\right\|_{C^{0,\alpha}_0(\mathbb{T})} \\
&\le c\|g\|_{C_s^{0,\alpha}(C_\epsilon)}+c\epsilon\|g\|_{C_s^{1,\alpha}(C_\epsilon)},
\end{aligned} A ϵ − g C 0 0 , α ( T ) ≤ T M ϵ , z , j x P j z g C 0 0 , α ( T ) + T M ϵ , r , j x P j r g C 0 0 , α ( T ) + T M ϵ , θ , j x P j θ g C 0 0 , α ( T ) ≤ c ∥ g ∥ C s 0 , α ( C ϵ ) + cϵ ∥ g ∥ C s 1 , α ( C ϵ ) ,
∥ A ‾ ϵ + g ∥ C 0 1 , α ( T ) ≤ ∥ T M ϵ , r , j x P j r g ∥ C 0 1 , α ( T ) + ∥ T M ϵ , z , j x P j z g ∥ C 0 1 , α ( T ) ≤ c ϵ − 1 ∥ g ∥ C s 1 , α ( C ϵ ) . \left\|\overline A_\epsilon^+g\right\|_{C^{1,\alpha}_0(\mathbb{T})}
\le\left\|T M_{\epsilon,r,j}^x\mathbb{P}_j^r g\right\|_{C^{1,\alpha}_0(\mathbb{T})}
+\left\|T M_{\epsilon,z,j}^x\mathbb{P}_j^z g\right\|_{C^{1,\alpha}_0(\mathbb{T})}
\le c\epsilon^{-1}\|g\|_{C_s^{1,\alpha}(C_\epsilon)}. A ϵ + g C 0 1 , α ( T ) ≤ T M ϵ , r , j x P j r g C 0 1 , α ( T ) + T M ϵ , z , j x P j z g C 0 1 , α ( T ) ≤ c ϵ − 1 ∥ g ∥ C s 1 , α ( C ϵ ) .
Proofs of semigroup lemmas
We may additionally use the Littlewood–Paley decomposition (107) to show the estimates (92) and (94) for the semigroup e − t L ~ ϵ ∂ s 4 e^{-t\widetilde L_\epsilon\partial_s^4} e − t L ϵ ∂ s 4 stated in Lemmas 2.2 and 2.3.
Proof of Lemma 2.2. We will require bounds for the Littlewood–Paley projection P j e − t L ~ ϵ ∂ s 4 V = e − t L ~ ϵ ∂ s 4 P j V P_je^{-t\widetilde L_\epsilon\partial_s^4}V=e^{-t\widetilde L_\epsilon\partial_s^4}P_jV P j e − t L ϵ ∂ s 4 V = e − t L ϵ ∂ s 4 P j V for each j j j . Using the notation of section 3.4.1, we first recall the definition of the matrix-valued Fourier multiplier
M ^ ϵ ( 1 ) ( k ) = m ϵ , t ( k ) e z ⊗ e z + m ϵ , n ( k ) ( e x ⊗ e x + e y ⊗ e y ) . \widehat M_\epsilon^{(1)}(k)=m_{\epsilon,t}(k)e_z\otimes e_z+m_{\epsilon,n}(k)(e_x\otimes e_x+e_y\otimes e_y). M ϵ ( 1 ) ( k ) = m ϵ , t ( k ) e z ⊗ e z + m ϵ , n ( k ) ( e x ⊗ e x + e y ⊗ e y ) .
At each time t t t , we may write the semigroup e − t L ‾ ϵ ∂ s 4 e^{-t\overline{L}_{\epsilon}\partial_s^4} e − t L ϵ ∂ s 4 as a Fourier multiplier in s s s :
F [ e − t L ‾ ϵ ∂ s 4 V ] ( k ) = e − t k 4 M ^ ϵ ( 1 ) ( k ) F [ V ] ( k ) . \mathcal{F}\left[e^{-t\overline{L}_{\epsilon}\partial_s^4}V\right](k)=e^{-t k^4\widehat{M}^{(1)}_{\epsilon}(k)}\mathcal{F}[V](k). F [ e − t L ϵ ∂ s 4 V ] ( k ) = e − t k 4 M ϵ ( 1 ) ( k ) F [ V ] ( k ) .
Recalling the definition (105) of ϕ j \phi_j ϕ j , at each t t t we consider
L ^ ϵ , j ( ξ , t ) = e − t ξ 4 M ^ ϵ , j ( ξ ) , M ^ ϵ , j = ( ϕ j − 1 ( ξ ) + ϕ j ( ξ ) + ϕ j + 1 ( ξ ) ) M ^ ϵ ( 1 ) ( ξ ) . \widehat{L}_{\epsilon,j}(\xi,t)=e^{-t\xi^4\widehat{M}_{\epsilon,j}(\xi)},\qquad\widehat{M}_{\epsilon,j}=\left(\phi_{j-1}(\xi)+\phi_j(\xi)+\phi_{j+1}(\xi)\right)\widehat{M}^{(1)}_{\epsilon}(\xi). L ϵ , j ( ξ , t ) = e − t ξ 4 M ϵ , j ( ξ ) , M ϵ , j = ( ϕ j − 1 ( ξ ) + ϕ j ( ξ ) + ϕ j + 1 ( ξ ) ) M ϵ ( 1 ) ( ξ ) .
Using the growth bounds of Lemmas 3.4 and 3.5, for j ϵ = ∣ log ( 2 π ϵ ) ∣ log ( 2 ) j_\epsilon=\frac{|\log(2\pi\epsilon)|}{\log(2)} j ϵ = l o g ( 2 ) ∣ l o g ( 2 π ϵ ) ∣ , we have
∣ L ^ ϵ , j ∣ ≤ { e − c t ϵ − 1 2 3 j , j ≥ j ϵ , e − c t ∣ log ϵ ∣ 2 2 4 j , j < j ϵ . \left|\widehat{L}_{\epsilon,j}\right|\le
\begin{cases}
e^{-ct\epsilon^{-1}2^{3j}}, & j\ge j_\epsilon,\\
e^{-ct|\log\epsilon|^2 2^{4j}}, & j<j_\epsilon.
\end{cases} L ϵ , j ≤ { e − c t ϵ − 1 2 3 j , e − c t ∣ l o g ϵ ∣ 2 2 4 j , j ≥ j ϵ , j < j ϵ .
∣ ∂ ξ 2 L ^ ϵ , j ∣ ≤ { c ( t ϵ − 1 2 j + ( t ϵ − 1 2 2 j ) 2 ) e − c t ϵ − 1 2 3 j , j ≥ j ϵ , c ( t ∣ log ϵ ∣ 2 2 j + ( t ∣ log ϵ ∣ 2 3 j ) 2 ) e − c t ∣ log ϵ ∣ 2 2 4 j , j < j ϵ . \left|\partial_\xi^2\widehat{L}_{\epsilon,j}\right|\le
\begin{cases}
c\left(t\epsilon^{-1}2^j+\left(t\epsilon^{-1}2^{2j}\right)^2\right)e^{-ct\epsilon^{-1}2^{3j}}, & j\ge j_\epsilon,\\
c\left(t|\log\epsilon|2^{2j}+\left(t|\log\epsilon|2^{3j}\right)^2\right)e^{-ct|\log\epsilon|^2 2^{4j}}, & j<j_\epsilon.
\end{cases} ∂ ξ 2 L ϵ , j ≤ ⎩ ⎨ ⎧ c ( t ϵ − 1 2 j + ( t ϵ − 1 2 2 j ) 2 ) e − c t ϵ − 1 2 3 j , c ( t ∣ log ϵ ∣ 2 2 j + ( t ∣ log ϵ ∣ 2 3 j ) 2 ) e − c t ∣ l o g ϵ ∣ 2 2 4 j , j ≥ j ϵ , j < j ϵ .
Using Lemma 3.1, we thus obtain
∥ L ϵ , j ( ⋅ , t ) ∥ L 1 ≤ { c ( t ϵ − 1 2 3 j ( 1 + t ϵ − 1 2 3 j ) ) 1 / 2 e − c t ϵ − 1 2 3 j , j ≥ j ϵ , c ( t ∣ log ϵ ∣ 2 2 4 j ( 1 + t ∣ log ϵ ∣ 2 2 4 j ) ) 1 / 2 e − c t ∣ log ϵ ∣ 2 2 4 j , j < j ϵ . \left\|L_{\epsilon,j}(\cdot,t)\right\|_{L^1}\le
\begin{cases}
c\left(t\epsilon^{-1}2^{3j}\left(1+t\epsilon^{-1}2^{3j}\right)\right)^{1/2}e^{-ct\epsilon^{-1}2^{3j}}, & j\ge j_\epsilon,\\
c\left(t|\log\epsilon|^2 2^{4j}\left(1+t|\log\epsilon|^2 2^{4j}\right)\right)^{1/2}e^{-ct|\log\epsilon|^2 2^{4j}}, & j<j_\epsilon.
\end{cases} ∥ L ϵ , j ( ⋅ , t ) ∥ L 1 ≤ { c ( t ϵ − 1 2 3 j ( 1 + t ϵ − 1 2 3 j ) ) 1/2 e − c t ϵ − 1 2 3 j , c ( t ∣ log ϵ ∣ 2 2 4 j ( 1 + t ∣ log ϵ ∣ 2 2 4 j ) ) 1/2 e − c t ∣ l o g ϵ ∣ 2 2 4 j , j ≥ j ϵ , j < j ϵ .
For V ∈ C m , γ ( T ) V\in C^{m,\gamma}(\mathbb{T}) V ∈ C m , γ ( T ) , we use the definitions (108), (106) of T M ^ T_{\widehat{M}} T M and P j P_j P j to write
e − t L ‾ ϵ ∂ s 4 V = ∑ j T L ^ ϵ , j P j V . e^{-t\overline{L}_{\epsilon}\partial_s^4}V=\sum_j T_{\widehat{L}_{\epsilon,j}}P_jV. e − t L ϵ ∂ s 4 V = j ∑ T L ϵ , j P j V .
We will let ω = n + α \omega=n+\alpha ω = n + α and ν = m + γ \nu=m+\gamma ν = m + γ and use the characterization (107) of C n , α ( T ) C^{n,\alpha}(\mathbb{T}) C n , α ( T ) and C m , γ ( T ) C^{m,\gamma}(\mathbb{T}) C m , γ ( T ) . Noting that P ≤ 0 = P 0 P_{\leq0}=P_0 P ≤ 0 = P 0 on T \mathbb{T} T , we may first estimate the lowest part of the norm (107) as
∥ T L ^ ϵ , 0 P ≤ 0 V ∥ L ∞ = ∥ L ϵ , 0 ∗ ( P 0 V ) ∥ L ∞ ≤ ∥ L ϵ , 0 ∥ L 1 ∥ V ∥ L ∞ ≤ c ( t ∣ log ϵ ∣ ( 1 + t ∣ log ϵ ∣ ) ) 1 / 2 e − c t ∣ log ϵ ∣ ∥ V ∥ L ∞ ≤ c ∥ V ∥ L ∞ . (193) \begin{aligned}
\left\|T_{\widehat{L}_{\epsilon,0}}P_{\leq0}V\right\|_{L^\infty}
=\left\|L_{\epsilon,0}*(P_0V)\right\|_{L^\infty}
\leq\left\|L_{\epsilon,0}\right\|_{L^1}\left\|V\right\|_{L^\infty} \\
&\leq c\left(t|\log\epsilon|\left(1+t|\log\epsilon|\right)\right)^{1/2}e^{-ct|\log\epsilon|}\left\|V\right\|_{L^\infty}
\leq c\left\|V\right\|_{L^\infty}. \tag*{(193)}
\end{aligned} T L ϵ , 0 P ≤ 0 V L ∞ = ∥ L ϵ , 0 ∗ ( P 0 V ) ∥ L ∞ ≤ ∥ L ϵ , 0 ∥ L 1 ∥ V ∥ L ∞ ≤ c ( t ∣ log ϵ ∣ ( 1 + t ∣ log ϵ ∣ ) ) 1/2 e − c t ∣ l o g ϵ ∣ ∥ V ∥ L ∞ ≤ c ∥ V ∥ L ∞ . (193)
For the low frequencies 0 < j < j ϵ 0<j<j_\epsilon 0 < j < j ϵ , we have
sup 0 < j < j ϵ 2 j ω ∥ T L ^ ϵ , j P j V ∥ L ∞ = sup 0 < j < j ϵ 2 j ω ∥ L ϵ , j ∗ ( P j V ) ∥ L ∞ ≤ sup 0 < j < j ϵ 2 j ( ω − ν ) ∥ L ϵ , j ∥ L 1 2 j ν ∥ P j V ∥ L ∞ ≤ sup 0 < j < j ϵ 2 j ( ω − ν ) ( t ∣ log ϵ ∣ 2 2 4 j ( 1 + t ∣ log ϵ ∣ 2 2 4 j ) ) 1 / 2 e − c t ∣ log ϵ ∣ 2 2 4 j 2 j ν ∥ P j V ∥ L ∞ ≤ c t − ω − ν 4 ∣ log ϵ ∣ − ω − ν 4 ∣ V ∣ B ˙ ∞ , ∞ ν . (194) \begin{aligned}
\sup_{0<j<j_\epsilon}2^{j\omega}\left\|T_{\widehat{L}_{\epsilon,j}}P_jV\right\|_{L^\infty}
=\sup_{0<j<j_\epsilon}2^{j\omega}\left\|L_{\epsilon,j}*(P_jV)\right\|_{L^\infty} \\
&\leq\sup_{0<j<j_\epsilon}2^{j(\omega-\nu)}\left\|L_{\epsilon,j}\right\|_{L^1}2^{j\nu}\left\|P_jV\right\|_{L^\infty} \\
&\leq\sup_{0<j<j_\epsilon}2^{j(\omega-\nu)}
\left(t|\log\epsilon|^2 2^{4j}\left(1+t|\log\epsilon|^2 2^{4j}\right)\right)^{1/2}
e^{-ct|\log\epsilon|^2 2^{4j}}2^{j\nu}\left\|P_jV\right\|_{L^\infty} \\
&\leq ct^{-\frac{\omega-\nu}{4}}|\log\epsilon|^{-\frac{\omega-\nu}{4}}\left|V\right|_{\dot{B}^{\nu}_{\infty,\infty}}. \tag*{(194)}
\end{aligned} 0 < j < j ϵ sup 2 j ω T L ϵ , j P j V L ∞ = 0 < j < j ϵ sup 2 j ω ∥ L ϵ , j ∗ ( P j V ) ∥ L ∞ ≤ 0 < j < j ϵ sup 2 j ( ω − ν ) ∥ L ϵ , j ∥ L 1 2 j ν ∥ P j V ∥ L ∞ ≤ 0 < j < j ϵ sup 2 j ( ω − ν ) ( t ∣ log ϵ ∣ 2 2 4 j ( 1 + t ∣ log ϵ ∣ 2 2 4 j ) ) 1/2 e − c t ∣ l o g ϵ ∣ 2 2 4 j 2 j ν ∥ P j V ∥ L ∞ ≤ c t − 4 ω − ν ∣ log ϵ ∣ − 4 ω − ν ∣ V ∣ B ˙ ∞ , ∞ ν . (194)
For the high frequencies j ≥ j ϵ j\geq j_\epsilon j ≥ j ϵ , we may similarly estimate
sup j ≥ j ϵ 2 j ω ∥ T L ^ ϵ , j P j V ∥ L ∞ ≤ sup j ≥ j ϵ 2 j ω ∥ L ϵ , j ∥ L 1 ∥ P j V ∥ L ∞ ≤ sup j ≥ j ϵ 2 j ( ω − ν ) ( t ϵ − 1 2 3 j ( 1 + t ϵ − 1 2 3 j ) ) 1 / 2 e − c t ϵ − 1 2 3 j 2 j ν ∥ P j V ∥ L ∞ ≤ c t − ω − ν 3 ϵ ω − ν 3 ∣ V ∣ B ˙ ∞ , ∞ ν . (195) \begin{aligned}
\sup_{j\geq j_\epsilon}2^{j\omega}\left\|T_{\widehat{L}_{\epsilon,j}}P_jV\right\|_{L^\infty}
\leq\sup_{j\geq j_\epsilon}2^{j\omega}\left\|L_{\epsilon,j}\right\|_{L^1}\left\|P_jV\right\|_{L^\infty} \\
&\leq\sup_{j\geq j_\epsilon}2^{j(\omega-\nu)}
\left(t\epsilon^{-1}2^{3j}\left(1+t\epsilon^{-1}2^{3j}\right)\right)^{1/2}
e^{-ct\epsilon^{-1}2^{3j}}2^{j\nu}\left\|P_jV\right\|_{L^\infty} \\
&\leq ct^{-\frac{\omega-\nu}{3}}\epsilon^{\frac{\omega-\nu}{3}}\left|V\right|_{\dot{B}^{\nu}_{\infty,\infty}}. \tag*{(195)}
\end{aligned} j ≥ j ϵ sup 2 j ω T L ϵ , j P j V L ∞ ≤ j ≥ j ϵ sup 2 j ω ∥ L ϵ , j ∥ L 1 ∥ P j V ∥ L ∞ ≤ j ≥ j ϵ sup 2 j ( ω − ν ) ( t ϵ − 1 2 3 j ( 1 + t ϵ − 1 2 3 j ) ) 1/2 e − c t ϵ − 1 2 3 j 2 j ν ∥ P j V ∥ L ∞ ≤ c t − 3 ω − ν ϵ 3 ω − ν ∣ V ∣ B ˙ ∞ , ∞ ν . (195)
Combining the estimates (192), (193), and (194) with the characterization (107) of C n , α C^{n,\alpha} C n , α and C m , γ C^{m,\gamma} C m , γ , we obtain Lemma 2.2. □
We next show Lemma 2.3 regarding a maximal regularity estimate for the semigroup e − t L ‾ ϵ ∂ s 4 e^{-t\overline{L}_{\epsilon}\partial_s^4} e − t L ϵ ∂ s 4 . Proof of Lemma 2.3 . We first note that, using the characterization (108) of C k , α ( T ) C^{k,\alpha}(\mathbb{T}) C k , α ( T ) and the definition (93) of the spaces Y k Y_k Y k , we may write ∥ g ∥ Y 1 \|g\|_{Y_1} ∥ g ∥ Y 1 as
∥ g ∥ Y 1 = ess sup t ∈ [ 0 , T ] ( ∥ P ≤ 0 g ( ⋅ , t ) ∥ L ∞ , sup j > 0 2 j ( 1 + α ) ∥ P j g ( ⋅ , t ) ∥ L ∞ ) = sup ( ess sup t ∈ [ 0 , T ] ∥ P ≤ 0 g ( ⋅ , t ) ∥ L ∞ , sup j > 0 2 j ( 1 + α ) ess sup t ∈ [ 0 , T ] ∥ P j g ( ⋅ , t ) ∥ L ∞ ) . \begin{aligned}
\|g\|_{Y_1}
&=\operatorname*{ess\,sup}_{t\in[0,T]}\left(\|P_{\leq0}g(\cdot,t)\|_{L^\infty},\sup_{j>0}2^{j(1+\alpha)}\|P_jg(\cdot,t)\|_{L^\infty}\right)\\
&=\sup\left(\operatorname*{ess\,sup}_{t\in[0,T]}\|P_{\leq0}g(\cdot,t)\|_{L^\infty},\sup_{j>0}2^{j(1+\alpha)}\operatorname*{ess\,sup}_{t\in[0,T]}\|P_jg(\cdot,t)\|_{L^\infty}\right).
\end{aligned} ∥ g ∥ Y 1 = t ∈ [ 0 , T ] ess sup ( ∥ P ≤ 0 g ( ⋅ , t ) ∥ L ∞ , j > 0 sup 2 j ( 1 + α ) ∥ P j g ( ⋅ , t ) ∥ L ∞ ) = sup ( t ∈ [ 0 , T ] ess sup ∥ P ≤ 0 g ( ⋅ , t ) ∥ L ∞ , j > 0 sup 2 j ( 1 + α ) t ∈ [ 0 , T ] ess sup ∥ P j g ( ⋅ , t ) ∥ L ∞ ) .
Again taking j ϵ = ∣ log ( 2 π ϵ ) ∣ log ( 2 ) j_\epsilon=\frac{|\log(2\pi\epsilon)|}{\log(2)} j ϵ = l o g ( 2 ) ∣ l o g ( 2 π ϵ ) ∣ and using Lemmas 3.4 and 3.5, for j ≥ j ϵ j\geq j_\epsilon j ≥ j ϵ we have
ess sup t ∈ [ 0 , T ] 2 j ( 4 + α ) ∥ P j h ( ⋅ , t ) ∥ L ∞ ( T ) = ess sup t ∈ [ 0 , T ] 2 j ( 4 + α ) ∥ ∫ 0 t e − ( t − t ′ ) L ϵ ‾ ∂ s 4 ∂ s 4 P j g ( ⋅ , t ′ ) d t ′ ∥ L ∞ ( T ) ≤ c ess sup t ∈ [ 0 , T ] 2 j ( 4 + α ) ∫ 0 t e − c ( t − t ′ ) ϵ − 1 2 3 j ∥ P j g ( ⋅ , t ′ ) ∥ L ∞ ( T ) d t ′ ≤ c ( ess sup t ∈ [ 0 , T ] 2 3 j ∫ 0 t e − c ( t − t ′ ) ϵ − 1 2 3 j d t ′ ) ( ess sup t ∈ [ 0 , T ] 2 j ( 1 + α ) ∥ P j g ( ⋅ , t ) ∥ L ∞ ( T ) ) ≤ c ϵ ∥ g ∥ Y 1 . \begin{aligned}
\operatorname*{ess\,sup}_{t\in[0,T]}2^{j(4+\alpha)}\|P_jh(\cdot,t)\|_{L^\infty(\mathbb{T})}
&=\operatorname*{ess\,sup}_{t\in[0,T]}2^{j(4+\alpha)}
\left\|\int_0^t e^{-(t-t')\overline{L_\epsilon}\partial_s^4}\partial_s^4P_jg(\cdot,t')\,dt'\right\|_{L^\infty(\mathbb{T})}\\
&\leq c\operatorname*{ess\,sup}_{t\in[0,T]}2^{j(4+\alpha)}
\int_0^t e^{-c(t-t')\epsilon^{-1}2^{3j}}\|P_jg(\cdot,t')\|_{L^\infty(\mathbb{T})}\,dt'\\
&\leq c\left(\operatorname*{ess\,sup}_{t\in[0,T]}2^{3j}\int_0^t e^{-c(t-t')\epsilon^{-1}2^{3j}}\,dt'\right)
\left(\operatorname*{ess\,sup}_{t\in[0,T]}2^{j(1+\alpha)}\|P_jg(\cdot,t)\|_{L^\infty(\mathbb{T})}\right)\\
&\leq c\epsilon\|g\|_{Y_1}.
\end{aligned} t ∈ [ 0 , T ] ess sup 2 j ( 4 + α ) ∥ P j h ( ⋅ , t ) ∥ L ∞ ( T ) = t ∈ [ 0 , T ] ess sup 2 j ( 4 + α ) ∫ 0 t e − ( t − t ′ ) L ϵ ∂ s 4 ∂ s 4 P j g ( ⋅ , t ′ ) d t ′ L ∞ ( T ) ≤ c t ∈ [ 0 , T ] ess sup 2 j ( 4 + α ) ∫ 0 t e − c ( t − t ′ ) ϵ − 1 2 3 j ∥ P j g ( ⋅ , t ′ ) ∥ L ∞ ( T ) d t ′ ≤ c ( t ∈ [ 0 , T ] ess sup 2 3 j ∫ 0 t e − c ( t − t ′ ) ϵ − 1 2 3 j d t ′ ) ( t ∈ [ 0 , T ] ess sup 2 j ( 1 + α ) ∥ P j g ( ⋅ , t ) ∥ L ∞ ( T ) ) ≤ cϵ ∥ g ∥ Y 1 .
In the first line we have used that P j P_j P j commutes with the semigroup e − t L ϵ ‾ ∂ s 4 e^{-t\overline{L_\epsilon}\partial_s^4} e − t L ϵ ∂ s 4 .
For low modes with 0 < j < j ϵ 0<j<j_\epsilon 0 < j < j ϵ , we have
ess sup t ∈ [ 0 , T ] 2 j ( 4 + α ) ∥ P j h ( ⋅ , t ) ∥ L ∞ ( T ) ≤ c ess sup t ∈ [ 0 , T ] 2 j ( 4 + α ) ∫ 0 t e − c ( t − t ′ ) ∣ log ϵ ∣ 2 4 j ∥ P j g ( ⋅ , t ′ ) ∥ L ∞ ( T ) d t ′ ≤ c ( ess sup t ∈ [ 0 , T ] 2 3 j ∫ 0 t e − c ( t − t ′ ) ∣ log ϵ ∣ 2 4 j d t ′ ) ( ess sup t ∈ [ 0 , T ] 2 j ( 1 + α ) ∥ P j g ( ⋅ , t ) ∥ L ∞ ( T ) ) ≤ c ∣ log ϵ ∣ − 3 / 4 ( ess sup t ∈ [ 0 , T ] ∫ 0 t ( t − t ′ ) − 3 / 4 d t ′ ) ∥ g ∥ Y 1 ≤ c T 1 / 4 ∣ log ϵ ∣ − 3 / 4 ∥ g ∥ Y 1 . \begin{aligned}
\operatorname*{ess\,sup}_{t\in[0,T]}2^{j(4+\alpha)}\|P_jh(\cdot,t)\|_{L^\infty(\mathbb{T})}
&\leq c\operatorname*{ess\,sup}_{t\in[0,T]}2^{j(4+\alpha)}
\int_0^t e^{-c(t-t')|\log\epsilon|\,2^{4j}}\|P_jg(\cdot,t')\|_{L^\infty(\mathbb{T})}\,dt'\\
&\leq c\left(\operatorname*{ess\,sup}_{t\in[0,T]}2^{3j}\int_0^t e^{-c(t-t')|\log\epsilon|\,2^{4j}}\,dt'\right)
\left(\operatorname*{ess\,sup}_{t\in[0,T]}2^{j(1+\alpha)}\|P_jg(\cdot,t)\|_{L^\infty(\mathbb{T})}\right)\\
&\leq c|\log\epsilon|^{-3/4}\left(\operatorname*{ess\,sup}_{t\in[0,T]}\int_0^t(t-t')^{-3/4}\,dt'\right)\|g\|_{Y_1}\\
&\leq cT^{1/4}|\log\epsilon|^{-3/4}\|g\|_{Y_1}.
\end{aligned} t ∈ [ 0 , T ] ess sup 2 j ( 4 + α ) ∥ P j h ( ⋅ , t ) ∥ L ∞ ( T ) ≤ c t ∈ [ 0 , T ] ess sup 2 j ( 4 + α ) ∫ 0 t e − c ( t − t ′ ) ∣ l o g ϵ ∣ 2 4 j ∥ P j g ( ⋅ , t ′ ) ∥ L ∞ ( T ) d t ′ ≤ c ( t ∈ [ 0 , T ] ess sup 2 3 j ∫ 0 t e − c ( t − t ′ ) ∣ l o g ϵ ∣ 2 4 j d t ′ ) ( t ∈ [ 0 , T ] ess sup 2 j ( 1 + α ) ∥ P j g ( ⋅ , t ) ∥ L ∞ ( T ) ) ≤ c ∣ log ϵ ∣ − 3/4 ( t ∈ [ 0 , T ] ess sup ∫ 0 t ( t − t ′ ) − 3/4 d t ′ ) ∥ g ∥ Y 1 ≤ c T 1/4 ∣ log ϵ ∣ − 3/4 ∥ g ∥ Y 1 .
Finally, for the zero-mode projection P ≤ 0 = P 0 P_{\leq0}=P_0 P ≤ 0 = P 0 , we have
ess sup t ∈ [ 0 , T ] ∥ P 0 h ( ⋅ , t ) ∥ L ∞ ( T ) ≤ c ess sup t ∈ [ 0 , T ] ∫ 0 t ∥ P 0 g ( ⋅ , t ′ ) ∥ L ∞ ( T ) d t ′ ≤ c T ∥ g ∥ Y 1 . \operatorname*{ess\,sup}_{t\in[0,T]}\|P_0h(\cdot,t)\|_{L^\infty(\mathbb{T})}
\leq c\operatorname*{ess\,sup}_{t\in[0,T]}\int_0^t\|P_0g(\cdot,t')\|_{L^\infty(\mathbb{T})}\,dt'
\leq cT\|g\|_{Y_1}. t ∈ [ 0 , T ] ess sup ∥ P 0 h ( ⋅ , t ) ∥ L ∞ ( T ) ≤ c t ∈ [ 0 , T ] ess sup ∫ 0 t ∥ P 0 g ( ⋅ , t ′ ) ∥ L ∞ ( T ) d t ′ ≤ c T ∥ g ∥ Y 1 .
In total, we obtain Lemma 2.3 . □
Bounds for single and double layer remainders
Setup and tools
This section is devoted to the proofs of Lemmas 1.10 , 1.11 , and 1.12 regarding the mapping properties of the single and double layer remainder terms. We first require a more detailed understanding of the kernels (37) and (38) along Γ ϵ \Gamma_\epsilon Γ ϵ as functions of the surface parameters s s s and θ \theta θ .
Throughout, we will be proving bounds for a single filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) X(s) X ( s ) , as well as corresponding Lipschitz estimates for two nearby filaments Σ ϵ ( a ) \Sigma_\epsilon^{(a)} Σ ϵ ( a ) and Σ ϵ ( b ) \Sigma_\epsilon^{(b)} Σ ϵ ( b ) with centerlines X ( a ) ( s ) X^{(a)}(s) X ( a ) ( s ) and X ( b ) ( s ) X^{(b)}(s) X ( b ) ( s ) satisfying Lemma 1.1 . When comparing two nearby filaments, we will use the superscripts ( a ) (a) ( a ) and ( b ) (b) ( b ) to denote quantities belonging to filament Σ ϵ ( a ) \Sigma_\epsilon^{(a)} Σ ϵ ( a ) and Σ ϵ ( b ) \Sigma_\epsilon^{(b)} Σ ϵ ( b ) , respectively.
Given the single and double layer operators for an arbitrarily curved filament, we will be extracting the corresponding single and double layer operators about the straight filament C ϵ C_\epsilon C ϵ . We will continue to denote quantities defined along C ϵ C_\epsilon C ϵ using overline notation. For x ‾ , x ‾ ′ ∈ C ϵ \overline{x},\overline{x}' \in C_\epsilon x , x ′ ∈ C ϵ , we define
R ‾ : = x ‾ − x ‾ ′ = ( s − s ′ ) e z + ϵ ( e r ( θ ) − e r ( θ ′ ) ) = ( s − s ′ ) e z + 2 ϵ sin ( θ − θ ′ 2 ) e θ ( θ + θ ′ 2 ) . (196) \overline{R} := \overline{x} - \overline{x}' = (s-s')e_z + \epsilon(e_r(\theta)-e_r(\theta')) = (s-s')e_z + 2\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)e_\theta\left(\frac{\theta+\theta'}{2}\right). \tag*{(196)} R := x − x ′ = ( s − s ′ ) e z + ϵ ( e r ( θ ) − e r ( θ ′ )) = ( s − s ′ ) e z + 2 ϵ sin ( 2 θ − θ ′ ) e θ ( 2 θ + θ ′ ) . (196)
Defining s ^ : = s − s ′ \widehat{s}:=s-s' s := s − s ′ and θ ^ : = θ − θ ′ \widehat{\theta}:=\theta-\theta' θ := θ − θ ′ , we may work in terms of s , s ^ , θ , θ ^ s,\widehat{s},\theta,\widehat{\theta} s , s , θ , θ rather than s , s ′ , θ , θ ′ s,s',\theta,\theta' s , s ′ , θ , θ ′ , and write
R ‾ = s ^ e z + 2 ϵ sin ( θ ^ 2 ) e θ ( θ − θ ^ 2 ) , s ^ = s − s ′ , θ ^ = θ − θ ′ . (197) \overline{R} = \widehat{s}e_z + 2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\left(\theta-\frac{\widehat{\theta}}{2}\right), \qquad\widehat{s}=s-s',\ \widehat{\theta}=\theta-\theta'. \tag*{(197)} R = s e z + 2 ϵ sin ( 2 θ ) e θ ( θ − 2 θ ) , s = s − s ′ , θ = θ − θ ′ . (197)
Noting that n ‾ ( x ‾ ) = e r ( θ ) \overline{n}(\overline{x})=e_r(\theta) n ( x ) = e r ( θ ) and n ‾ ( x ‾ ′ ) = e r ( θ − θ ^ ) \overline{n}(\overline{x}')=e_r(\theta-\widehat{\theta}) n ( x ′ ) = e r ( θ − θ ) , we may write
∣ R ‾ ( s ^ , θ ^ ) ∣ 2 = s ^ 2 + 4 ϵ 2 sin 2 ( θ ^ 2 ) , \left|\overline{R}(\widehat{s},\widehat{\theta})\right|^2=\widehat{s}^2+4\epsilon^2\sin^2\left(\frac{\widehat{\theta}}{2}\right), R ( s , θ ) 2 = s 2 + 4 ϵ 2 sin 2 ( 2 θ ) ,
R ‾ ⋅ n ‾ = − R ‾ ⋅ n ‾ ′ = 2 ϵ sin 2 ( θ ^ 2 ) . \overline{R}\cdot\overline{n}=-\overline{R}\cdot\overline{n}'=2\epsilon\sin^2\left(\frac{\widehat{\theta}}{2}\right). R ⋅ n = − R ⋅ n ′ = 2 ϵ sin 2 ( 2 θ ) .
Along a curved filament Σ ϵ \Sigma_\epsilon Σ ϵ , we similarly consider
as a function of ( s , s ^ , θ , θ ^ ) (s,\widehat{s},\theta,\widehat{\theta}) ( s , s , θ , θ ) . Using the orthonormal frame ODEs (3) , we note the following expansions with respect to arclength s s s :
X ( s ) − X ( s − s ^ ) = s ^ e t ( s ) − s ^ 2 Q t ( s , s ^ ) e t ( s ) − e t ( s − s ^ ) = s ^ Q t ( s , s ^ ) e n 1 ( s ) − e n 1 ( s − s ^ ) = s ^ ( − κ 1 ( s ) e t ( s ) + κ 3 e n 2 ( s ) ) + s ^ 2 Q n 1 ( s , s ^ ) e n 2 ( s ) − e n 2 ( s − s ^ ) = s ^ ( − κ 2 ( s ) e t ( s ) − κ 3 e n 1 ( s ) ) + s ^ 2 Q n 2 ( s , s ^ ) . (198) \begin{aligned}
X(s)-X(s-\widehat{s}) &= \widehat{s}e_t(s)-\widehat{s}^{2}Q_t(s,\widehat{s})\\
e_t(s)-e_t(s-\widehat{s}) &= \widehat{s}Q_t(s,\widehat{s})\\
e_{n_1}(s)-e_{n_1}(s-\widehat{s}) &= \widehat{s}\left(-\kappa_1(s)e_t(s)+\kappa_3e_{n_2}(s)\right)+\widehat{s}^{2}Q_{n_1}(s,\widehat{s})\\
e_{n_2}(s)-e_{n_2}(s-\widehat{s}) &= \widehat{s}\left(-\kappa_2(s)e_t(s)-\kappa_3e_{n_1}(s)\right)+\widehat{s}^{2}Q_{n_2}(s,\widehat{s}). \tag*{(198)}
\end{aligned} X ( s ) − X ( s − s ) e t ( s ) − e t ( s − s ) e n 1 ( s ) − e n 1 ( s − s ) e n 2 ( s ) − e n 2 ( s − s ) = s e t ( s ) − s 2 Q t ( s , s ) = s Q t ( s , s ) = s ( − κ 1 ( s ) e t ( s ) + κ 3 e n 2 ( s ) ) + s 2 Q n 1 ( s , s ) = s ( − κ 2 ( s ) e t ( s ) − κ 3 e n 1 ( s ) ) + s 2 Q n 2 ( s , s ) . (198)
Given two filaments with centerlines X ( a ) , X ( b ) X^{(a)},X^{(b)} X ( a ) , X ( b ) in C 2 , β ( T ) C^{2,\beta}(\mathbb{T}) C 2 , β ( T ) , each of the remainder terms Q i Q_i Q i in (198) satisfies
∥ Q i ( a ) ⋅ e j ( a ) − Q i ( b ) ⋅ e j ( b ) ∥ Y ≤ c ∑ ℓ = 1 3 ∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ Y , i , j ∈ { t , n 1 , n 2 } . (199) \left\|Q_i^{(a)}\cdot e_j^{(a)}-Q_i^{(b)}\cdot e_j^{(b)}\right\|_Y\leq c\sum_{\ell=1}^{3}\left\|\kappa_\ell^{(a)}-\kappa_\ell^{(b)}\right\|_Y,\qquad i,j\in\{t,n_1,n_2\}. \tag*{(199)} Q i ( a ) ⋅ e j ( a ) − Q i ( b ) ⋅ e j ( b ) Y ≤ c ℓ = 1 ∑ 3 κ ℓ ( a ) − κ ℓ ( b ) Y , i , j ∈ { t , n 1 , n 2 } . (199)
where the norms ∥ ⋅ ∥ Y \|\cdot\|_Y ∥ ⋅ ∥ Y are defined for functions of both ( s , θ ) (s,\theta) ( s , θ ) and ( s ^ , θ ^ ) ∈ T × 2 π T (\widehat{s},\widehat{\theta})\in\mathbb{T}\times2\pi\mathbb{T} ( s , θ ) ∈ T × 2 π T by any of the following:
∥ Q ∥ L ∞ : = e s s s u p s , θ , s ^ , θ ^ ∣ Q ( s , θ , s ^ , θ ^ ) ∣ , ∥ Q ∥ C 1 0 , β : = e s s s u p s ^ , θ ^ ∥ Q ( ⋅ , ⋅ , s ^ , θ ^ ) ∥ C 0 , β , ∥ Q ∥ C 2 0 , β : = e s s s u p s , θ ∥ Q ( s , θ , ⋅ , ⋅ ) ∥ C 0 , β . (200) \begin{aligned}
\|Q\|_{L^\infty} &:= \mathop{\mathrm{ess\,sup}}_{s,\theta,\widehat{s},\widehat{\theta}}\left|Q(s,\theta,\widehat{s},\widehat{\theta})\right|,\qquad
\|Q\|_{C_1^{0,\beta}} := \mathop{\mathrm{ess\,sup}}_{\widehat{s},\widehat{\theta}}\left\|Q(\cdot,\cdot,\widehat{s},\widehat{\theta})\right\|_{C^{0,\beta}},\\
\|Q\|_{C_2^{0,\beta}} &:= \mathop{\mathrm{ess\,sup}}_{s,\theta}\left\|Q(s,\theta,\cdot,\cdot)\right\|_{C^{0,\beta}}. \tag*{(200)}
\end{aligned} ∥ Q ∥ L ∞ ∥ Q ∥ C 2 0 , β := ess sup s , θ , s , θ Q ( s , θ , s , θ ) , ∥ Q ∥ C 1 0 , β := ess sup s , θ Q ( ⋅ , ⋅ , s , θ ) C 0 , β , := ess sup s , θ ∥ Q ( s , θ , ⋅ , ⋅ ) ∥ C 0 , β . (200)
Note that for Q = Q ( s , θ ) Q=Q(s,\theta) Q = Q ( s , θ ) we have ∥ Q ∥ C 1 0 , β = ∥ Q ∥ C 0 , β \|Q\|_{C_1^{0,\beta}}=\|Q\|_{C^{0,\beta}} ∥ Q ∥ C 1 0 , β = ∥ Q ∥ C 0 , β , and for Q = Q ( s ^ , θ ^ ) Q=Q(\widehat{s},\widehat{\theta}) Q = Q ( s , θ ) we have ∥ Q ∥ C 2 0 , β = ∥ Q ∥ C 0 , β \|Q\|_{C_2^{0,\beta}}=\|Q\|_{C^{0,\beta}} ∥ Q ∥ C 2 0 , β = ∥ Q ∥ C 0 , β . For some combinations of Q i ⋅ e j Q_i\cdot e_j Q i ⋅ e j , we note the following more refined expansions:
Q t ( s , s ^ ) ⋅ e t ( s ) = s ^ Q ^ t ( s , s ^ ) , Q n j ( s , s ^ ) ⋅ e t ( s ) + Q t ( s , s ^ ) ⋅ e n j ( s ) = s ^ Q ^ t n j ( s , s ^ ) , j = 1 , 2 ; ∥ Q i ( a ) − Q i ( b ) ∥ Y ≤ c ∑ ℓ = 1 3 ∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ Y , i ∈ { t , t n 1 , t n 2 } . (201) \begin{aligned}
Q_t(s,\widehat{s})\cdot e_t(s) &= \widehat{s}\widehat{Q}_t(s,\widehat{s}),\\
Q_{n_j}(s,\widehat{s})\cdot e_t(s)+Q_t(s,\widehat{s})\cdot e_{n_j}(s) &= \widehat{s}\widehat{Q}_{tn_j}(s,\widehat{s}),\qquad j=1,2;\\
\left\|Q_i^{(a)}-Q_i^{(b)}\right\|_Y &\leq c\sum_{\ell=1}^{3}\left\|\kappa_\ell^{(a)}-\kappa_\ell^{(b)}\right\|_Y,\qquad i\in\{t,tn_1,tn_2\}. \tag*{(201)}
\end{aligned} Q t ( s , s ) ⋅ e t ( s ) Q n j ( s , s ) ⋅ e t ( s ) + Q t ( s , s ) ⋅ e n j ( s ) Q i ( a ) − Q i ( b ) Y = s Q t ( s , s ) , = s Q t n j ( s , s ) , j = 1 , 2 ; ≤ c ℓ = 1 ∑ 3 κ ℓ ( a ) − κ ℓ ( b ) Y , i ∈ { t , t n 1 , t n 2 } . (201)
Using (198) and the definition (5) of e r ( s , θ ) e_r(s,\theta) e r ( s , θ ) and e θ ( s , θ ) e_\theta(s,\theta) e θ ( s , θ ) , we may write
e r ( s , θ ) − e r ( s − s ^ , θ − θ ^ ) = s ^ Q r ( s , s ^ , θ ^ ) , e θ ( s , θ ) − e θ ( s − s ^ , θ − θ ^ ) = s ^ Q θ ( s , s ^ , θ ^ ) . (202) e_r(s,\theta)-e_r(s-\widehat{s},\theta-\widehat{\theta})=\widehat{s}Q_r(s,\widehat{s},\widehat{\theta}),\qquad e_\theta(s,\theta)-e_\theta(s-\widehat{s},\theta-\widehat{\theta})=\widehat{s}Q_\theta(s,\widehat{s},\widehat{\theta}). \tag*{(202)} e r ( s , θ ) − e r ( s − s , θ − θ ) = s Q r ( s , s , θ ) , e θ ( s , θ ) − e θ ( s − s , θ − θ ) = s Q θ ( s , s , θ ) . (202)
where, using (199) and (201) , Q r Q_r Q r and Q θ Q_\theta Q θ satisfy
e t ( s ) ⋅ Q r ( s , s ^ , θ ) = − κ ^ ( s , θ ) + s ^ Q 0 , 1 ( s , θ , s ^ ) e t ( s ) ⋅ Q θ ( s , θ , s ^ ) + Q t ( s , s ^ ) ⋅ e θ ( s , θ ) = s ^ Q 0 , 2 ( s , θ , s ^ ) e θ ( s , θ − θ ^ 2 ) ⋅ Q r ( s , s ^ , θ − θ ^ ) = κ 3 cos ( θ ^ 2 ) + s ^ Q 0 , 3 ( s , θ , s ^ , θ ^ ) ∣ Q r ( s , θ , s ^ ) ∣ 2 = κ ^ ( s , θ ) 2 + κ 3 2 + s ^ Q 0 , 4 ( s , θ , s ^ ) e r ( s , θ ) ⋅ Q r ( s , θ , s ^ ) = s ^ Q 0 , 5 ( s , θ , s ^ ) . (203) \begin{aligned}
e_t(s)\cdot Q_r(s,\widehat{s},\theta) &= -\widehat{\kappa}(s,\theta)+\widehat{s}Q_{0,1}(s,\theta,\widehat{s})\\
e_t(s)\cdot Q_\theta(s,\theta,\widehat{s})+Q_t(s,\widehat{s})\cdot e_\theta(s,\theta) &= \widehat{s}Q_{0,2}(s,\theta,\widehat{s})\\
e_\theta\left(s,\theta-\frac{\widehat{\theta}}{2}\right)\cdot Q_r(s,\widehat{s},\theta-\widehat{\theta}) &= \kappa_3\cos\left(\frac{\widehat{\theta}}{2}\right)+\widehat{s}Q_{0,3}(s,\theta,\widehat{s},\widehat{\theta})\\
\lvert Q_r(s,\theta,\widehat{s})\rvert^2 &= \widehat{\kappa}(s,\theta)^2+\kappa_3^2+\widehat{s}Q_{0,4}(s,\theta,\widehat{s})\\
e_r(s,\theta)\cdot Q_r(s,\theta,\widehat{s}) &= \widehat{s}Q_{0,5}(s,\theta,\widehat{s}). \tag*{(203)}
\end{aligned} e t ( s ) ⋅ Q r ( s , s , θ ) e t ( s ) ⋅ Q θ ( s , θ , s ) + Q t ( s , s ) ⋅ e θ ( s , θ ) e θ ( s , θ − 2 θ ) ⋅ Q r ( s , s , θ − θ ) ∣ Q r ( s , θ , s ) ∣ 2 e r ( s , θ ) ⋅ Q r ( s , θ , s ) = − κ ( s , θ ) + s Q 0 , 1 ( s , θ , s ) = s Q 0 , 2 ( s , θ , s ) = κ 3 cos ( 2 θ ) + s Q 0 , 3 ( s , θ , s , θ ) = κ ( s , θ ) 2 + κ 3 2 + s Q 0 , 4 ( s , θ , s ) = s Q 0 , 5 ( s , θ , s ) . (203)
For each Q 0 , j Q_{0,j} Q 0 , j above, using the norms ∥ ⋅ ∥ Y \lVert\cdot\rVert_Y ∥ ⋅ ∥ Y defined in (200) , we have
∥ Q 0 , j ( a ) − Q 0 , j ( b ) ∥ Y ≤ c ϵ − β ∑ ℓ = 1 3 ∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ Y . (204) \left\lVert Q_{0,j}^{(a)}-Q_{0,j}^{(b)}\right\rVert_Y\leq c\epsilon^{-\beta}\sum_{\ell=1}^{3}\left\lVert\kappa_\ell^{(a)}-\kappa_\ell^{(b)}\right\rVert_Y. \tag*{(204)} Q 0 , j ( a ) − Q 0 , j ( b ) Y ≤ c ϵ − β ℓ = 1 ∑ 3 κ ℓ ( a ) − κ ℓ ( b ) Y . (204)
Here the ϵ − β \epsilon^{-\beta} ϵ − β arises due to the θ \theta θ -dependence of these terms along Γ ϵ \Gamma_\epsilon Γ ϵ , and we have β = 0 \beta=0 β = 0 for the Y = L ∞ Y=L^\infty Y = L ∞ estimate. In addition, we note that the following identities hold along Γ ϵ \Gamma_\epsilon Γ ϵ :
e r ( s , θ ) − e r ( s , θ − θ ^ ) = 2 sin ( θ ^ 2 ) e θ ( s , θ − θ ^ 2 ) , e θ ( s , θ − θ ^ 2 ) ⋅ e r ( s , θ ) = sin ( θ ^ 2 ) , Q r ( s , θ , s ^ ) − Q r ( s , θ − θ ^ , s ^ ) = 2 sin ( θ ^ 2 ) Q θ ( s , θ − θ ^ 2 , s ^ ) . (205) \begin{aligned}
e_r(s,\theta)-e_r\left(s,\theta-\widehat{\theta}\right)&=2\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\left(s,\theta-\frac{\widehat{\theta}}{2}\right),\\
e_\theta\left(s,\theta-\frac{\widehat{\theta}}{2}\right)\cdot e_r(s,\theta)&=\sin\left(\frac{\widehat{\theta}}{2}\right),\\
Q_r(s,\theta,\widehat{s})-Q_r(s,\theta-\widehat{\theta},\widehat{s})&=2\sin\left(\frac{\widehat{\theta}}{2}\right)Q_\theta\left(s,\theta-\frac{\widehat{\theta}}{2},\widehat{s}\right). \tag*{(205)}
\end{aligned} e r ( s , θ ) − e r ( s , θ − θ ) e θ ( s , θ − 2 θ ) ⋅ e r ( s , θ ) Q r ( s , θ , s ) − Q r ( s , θ − θ , s ) = 2 sin ( 2 θ ) e θ ( s , θ − 2 θ ) , = sin ( 2 θ ) , = 2 sin ( 2 θ ) Q θ ( s , θ − 2 θ , s ) . (205)
Using the above expansions, we may write (see [59 Section 3] ):
R ( s , θ , s ^ , θ ^ ) = X ( s ) − X ( s − s ^ ) + ϵ ( e r ( s , θ ) − e r ( s − s ^ , θ − θ ^ ) ) = s ^ e t ( s ) + 2 ϵ sin ( θ ^ 2 ) e θ ( s , θ − θ ^ 2 ) − s ^ 2 Q t ( s , s ^ ) + ϵ s ^ Q r ( s , s ^ , θ − θ ^ ) . (206) \begin{aligned}
R(s,\theta,\widehat{s},\widehat{\theta})&=X(s)-X(s-\widehat{s})+\epsilon\left(e_r(s,\theta)-e_r(s-\widehat{s},\theta-\widehat{\theta})\right)\\
&=\widehat{s}e_t(s)+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\left(s,\theta-\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q_t(s,\widehat{s})+\epsilon\widehat{s}Q_r(s,\widehat{s},\theta-\widehat{\theta}). \tag*{(206)}
\end{aligned} R ( s , θ , s , θ ) = X ( s ) − X ( s − s ) + ϵ ( e r ( s , θ ) − e r ( s − s , θ − θ ) ) = s e t ( s ) + 2 ϵ sin ( 2 θ ) e θ ( s , θ − 2 θ ) − s 2 Q t ( s , s ) + ϵ s Q r ( s , s , θ − θ ) . (206)
Recalling the form (197) of R ‾ \overline{R} R , we have
∣ R ∣ 2 = ∣ R ‾ ∣ 2 + ϵ s ^ 2 Q R , 0 ( s , θ ) + 2 κ 3 ϵ 2 s ^ sin ( θ ^ ) + s ^ 4 Q R , 1 ( s , s ^ ) + ϵ s ^ 3 Q R , 2 ( s , θ , s ^ , θ ^ ) + ϵ 2 s ^ 2 sin ( θ ^ 2 ) Q R , 3 ( s , θ , s ^ , θ ^ ) , (207) \lvert R\rvert^2=\lvert\overline{R}\rvert^2+\epsilon\widehat{s}^{2}Q_{R,0}(s,\theta)+2\kappa_3\epsilon^{2}\widehat{s}\sin(\widehat{\theta})+\widehat{s}^{4}Q_{R,1}(s,\widehat{s})+\epsilon\widehat{s}^{3}Q_{R,2}(s,\theta,\widehat{s},\widehat{\theta})+\epsilon^{2}\widehat{s}^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)Q_{R,3}(s,\theta,\widehat{s},\widehat{\theta}), \tag*{(207)} ∣ R ∣ 2 = ∣ R ∣ 2 + ϵ s 2 Q R , 0 ( s , θ ) + 2 κ 3 ϵ 2 s sin ( θ ) + s 4 Q R , 1 ( s , s ) + ϵ s 3 Q R , 2 ( s , θ , s , θ ) + ϵ 2 s 2 sin ( 2 θ ) Q R , 3 ( s , θ , s , θ ) , (207)
where Q R , j Q_{R,j} Q R , j are given by
Q R , 0 ( s , θ ) = − 2 κ ^ ( s , θ ) + ϵ κ ^ ( s , θ ) 2 + ϵ κ 3 2 Q R , 1 ( s , s ^ ) = 3 Q t ( s , s ^ ) + ∣ Q t ( s , s ^ ) ∣ 2 Q R , 2 ( s , θ , s ^ , θ ^ ) = 2 Q 0 , 1 ( s , s ^ , θ ) + 4 sin ( θ ^ 2 ) Q 0 , 2 ( s , s ^ , θ − θ ^ 2 ) + 2 Q t ( s , s ^ ) ⋅ Q r ( s , s ^ , θ − θ ^ ) + ϵ Q 0 , 4 ( s , s ^ , θ ) Q R , 3 ( s , θ , s ^ , θ ^ ) = 4 Q 0 , 3 ( s , θ , s ^ , θ ^ ) − 4 Q θ ( s , s ^ , θ − θ ^ 2 ) ⋅ Q r ( s , s ^ , θ ) + 4 sin ( θ ^ 2 ) ∣ Q θ ( s , s ^ , θ − θ ^ 2 ) ∣ 2 . (208) \begin{aligned}
Q_{R,0}(s,\theta)&=-2\widehat{\kappa}(s,\theta)+\epsilon\widehat{\kappa}(s,\theta)^2+\epsilon\kappa_3^2\\
Q_{R,1}(s,\widehat{s})&=3Q_t(s,\widehat{s})+\lvert Q_t(s,\widehat{s})\rvert^2\\
Q_{R,2}(s,\theta,\widehat{s},\widehat{\theta})&=2Q_{0,1}(s,\widehat{s},\theta)+4\sin\left(\frac{\widehat{\theta}}{2}\right)Q_{0,2}\left(s,\widehat{s},\theta-\frac{\widehat{\theta}}{2}\right)\\
&\quad+2Q_t(s,\widehat{s})\cdot Q_r(s,\widehat{s},\theta-\widehat{\theta})+\epsilon Q_{0,4}(s,\widehat{s},\theta)\\
Q_{R,3}(s,\theta,\widehat{s},\widehat{\theta})&=4Q_{0,3}(s,\theta,\widehat{s},\widehat{\theta})-4Q_\theta\left(s,\widehat{s},\theta-\frac{\widehat{\theta}}{2}\right)\cdot Q_r(s,\widehat{s},\theta)+4\sin\left(\frac{\widehat{\theta}}{2}\right)\left\lvert Q_\theta\left(s,\widehat{s},\theta-\frac{\widehat{\theta}}{2}\right)\right\rvert^2. \tag*{(208)}
\end{aligned} Q R , 0 ( s , θ ) Q R , 1 ( s , s ) Q R , 2 ( s , θ , s , θ ) Q R , 3 ( s , θ , s , θ ) = − 2 κ ( s , θ ) + ϵ κ ( s , θ ) 2 + ϵ κ 3 2 = 3 Q t ( s , s ) + ∣ Q t ( s , s ) ∣ 2 = 2 Q 0 , 1 ( s , s , θ ) + 4 sin ( 2 θ ) Q 0 , 2 ( s , s , θ − 2 θ ) + 2 Q t ( s , s ) ⋅ Q r ( s , s , θ − θ ) + ϵ Q 0 , 4 ( s , s , θ ) = 4 Q 0 , 3 ( s , θ , s , θ ) − 4 Q θ ( s , s , θ − 2 θ ) ⋅ Q r ( s , s , θ ) + 4 sin ( 2 θ ) Q θ ( s , s , θ − 2 θ ) 2 . (208)
Here we emphasize that Q R , 0 ( s , θ ) Q_{R,0}(s,\theta) Q R , 0 ( s , θ ) does not depend on s ^ \widehat{s} s and θ ^ \widehat{\theta} θ , as this will be used in the proof of Lemma 5.4 later on. Using the definition (200) of the norms ∥ ⋅ ∥ Y \lVert\cdot\rVert_Y ∥ ⋅ ∥ Y and the bounds (204) , the remaining Q R , j Q_{R,j} Q R , j can be seen to satisfy
∥ Q R , 1 ( a ) − Q R , 1 ( b ) ∥ Y ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ∑ ℓ = 1 3 ∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ Y , \left\lVert Q_{R,1}^{(a)}-Q_{R,1}^{(b)}\right\rVert_Y\leq c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)})\sum_{\ell=1}^{3}\left\lVert\kappa_\ell^{(a)}-\kappa_\ell^{(b)}\right\rVert_Y, Q R , 1 ( a ) − Q R , 1 ( b ) Y ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ℓ = 1 ∑ 3 κ ℓ ( a ) − κ ℓ ( b ) Y ,
∥ Q R , j ( a ) − Q R , j ( b ) ∥ Y ≤ ϵ − β c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ∑ ℓ = 1 3 ∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ Y , j = 2 , 3. \left\lVert Q_{R,j}^{(a)}-Q_{R,j}^{(b)}\right\rVert_Y\leq\epsilon^{-\beta}c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)})\sum_{\ell=1}^{3}\left\lVert\kappa_\ell^{(a)}-\kappa_\ell^{(b)}\right\rVert_Y,\qquad j=2,3. Q R , j ( a ) − Q R , j ( b ) Y ≤ ϵ − β c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ℓ = 1 ∑ 3 κ ℓ ( a ) − κ ℓ ( b ) Y , j = 2 , 3.
For two filaments with centerlines X ( a ) X^{(a)} X ( a ) , X ( b ) X^{(b)} X ( b ) in C 2 , β ( T ) C^{2,\beta}(\mathbb{T}) C 2 , β ( T ) , we have
R ( a ) − R ( b ) = s ^ ( e t ( a ) ( s ) − e t ( b ) ( s ) ) + 2 ϵ sin ( θ ^ 2 ) ( e θ ( a ) ( s , θ − θ ^ 2 ) − e θ ( b ) ( s , θ − θ ^ 2 ) ) − s ^ 2 ( Q t ( a ) ( s , s ^ ) − Q t ( b ) ( s , s ^ ) ) + ϵ s ^ ( Q r ( a ) ( s , s ^ , θ − θ ^ ) − Q r ( b ) ( s , s ^ , θ − θ ^ ) ) . (209) R^{(a)}-R^{(b)}=\hat{s}(e_t^{(a)}(s)-e_t^{(b)}(s))+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)(e_\theta^{(a)}(s,\theta-\frac{\hat{\theta}}{2})-e_\theta^{(b)}(s,\theta-\frac{\hat{\theta}}{2}))
-\hat{s}^{2}(Q_t^{(a)}(s,\hat{s})-Q_t^{(b)}(s,\hat{s}))+\epsilon\hat{s}(Q_r^{(a)}(s,\hat{s},\theta-\hat{\theta})-Q_r^{(b)}(s,\hat{s},\theta-\hat{\theta})) . \tag*{(209)} R ( a ) − R ( b ) = s ^ ( e t ( a ) ( s ) − e t ( b ) ( s )) + 2 ϵ sin ( 2 θ ^ ) ( e θ ( a ) ( s , θ − 2 θ ^ ) − e θ ( b ) ( s , θ − 2 θ ^ )) − s ^ 2 ( Q t ( a ) ( s , s ^ ) − Q t ( b ) ( s , s ^ )) + ϵ s ^ ( Q r ( a ) ( s , s ^ , θ − θ ^ ) − Q r ( b ) ( s , s ^ , θ − θ ^ )) . (209)
Furthermore, we may write the differences
1 ∣ R ∣ − 1 ∣ R ‾ ∣ = ϵ s ^ 2 Q R , 0 + ϵ 2 s ^ sin ( θ ^ ) κ 3 + s ^ 4 Q R , 1 + ϵ 3 s ^ Q R , 2 + ϵ 2 s ^ 2 sin ( θ ^ 2 ) Q R , 3 ∣ R ∣ ∣ R ‾ ∣ ( ∣ R ∣ + ∣ R ‾ ∣ ) \frac{1}{|R|}-\frac{1}{|\overline{R}|}=
\frac{\epsilon\hat{s}^{2}Q_{R,0}+\epsilon^{2}\hat{s}\sin(\hat{\theta})\kappa_{3}+\hat{s}^{4}Q_{R,1}+\epsilon^{3}\hat{s}Q_{R,2}+\epsilon^{2}\hat{s}^{2}\sin\left(\frac{\hat{\theta}}{2}\right)Q_{R,3}}
{|R||\overline{R}|(|R|+|\overline{R}|)} ∣ R ∣ 1 − ∣ R ∣ 1 = ∣ R ∣∣ R ∣ ( ∣ R ∣ + ∣ R ∣ ) ϵ s ^ 2 Q R , 0 + ϵ 2 s ^ sin ( θ ^ ) κ 3 + s ^ 4 Q R , 1 + ϵ 3 s ^ Q R , 2 + ϵ 2 s ^ 2 sin ( 2 θ ^ ) Q R , 3
1 ∣ R ( a ) ∣ − 1 ∣ R ( b ) ∣ = ϵ s ^ 2 ( Q R , 0 ( a ) − Q R , 0 ( b ) ) + ϵ 2 s ^ sin ( θ ^ ) ( κ 3 ( a ) − κ 3 ( b ) ) ∣ R ( a ) ∣ ∣ R ( b ) ∣ ( ∣ R ( a ) ∣ + ∣ R ( b ) ∣ ) + s ^ 4 ( Q R , 1 ( a ) − Q R , 1 ( b ) ) + ϵ 3 s ^ ( Q R , 2 ( a ) − Q R , 2 ( b ) ) + ϵ 2 s ^ 2 sin ( θ ^ 2 ) ( Q R , 3 ( a ) − Q R , 3 ( b ) ) ∣ R ( a ) ∣ ∣ R ( b ) ∣ ( ∣ R ( a ) ∣ + ∣ R ( b ) ∣ ) . (210) \frac{1}{|R^{(a)}|}-\frac{1}{|R^{(b)}|}=
\frac{\epsilon\hat{s}^{2}(Q_{R,0}^{(a)}-Q_{R,0}^{(b)})+\epsilon^{2}\hat{s}\sin(\hat{\theta})(\kappa_{3}^{(a)}-\kappa_{3}^{(b)})}
{|R^{(a)}||R^{(b)}|(|R^{(a)}|+|R^{(b)}|)}
+
\frac{\hat{s}^{4}(Q_{R,1}^{(a)}-Q_{R,1}^{(b)})+\epsilon^{3}\hat{s}(Q_{R,2}^{(a)}-Q_{R,2}^{(b)})+\epsilon^{2}\hat{s}^{2}\sin\left(\frac{\hat{\theta}}{2}\right)(Q_{R,3}^{(a)}-Q_{R,3}^{(b)})}
{|R^{(a)}||R^{(b)}|(|R^{(a)}|+|R^{(b)}|)} . \tag*{(210)} ∣ R ( a ) ∣ 1 − ∣ R ( b ) ∣ 1 = ∣ R ( a ) ∣∣ R ( b ) ∣ ( ∣ R ( a ) ∣ + ∣ R ( b ) ∣ ) ϵ s ^ 2 ( Q R , 0 ( a ) − Q R , 0 ( b ) ) + ϵ 2 s ^ sin ( θ ^ ) ( κ 3 ( a ) − κ 3 ( b ) ) + ∣ R ( a ) ∣∣ R ( b ) ∣ ( ∣ R ( a ) ∣ + ∣ R ( b ) ∣ ) s ^ 4 ( Q R , 1 ( a ) − Q R , 1 ( b ) ) + ϵ 3 s ^ ( Q R , 2 ( a ) − Q R , 2 ( b ) ) + ϵ 2 s ^ 2 sin ( 2 θ ^ ) ( Q R , 3 ( a ) − Q R , 3 ( b ) ) . (210)
We may use (210) in expanding differences of higher powers k ∈ N k\in\mathbb{N} k ∈ N as
1 ∣ R ∣ k − 1 ∣ R ‾ ∣ k = ( 1 ∣ R ∣ − 1 ∣ R ‾ ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R ∣ ℓ ∣ R ‾ ∣ k − 1 − ℓ , \frac{1}{|R|^{k}}-\frac{1}{|\overline{R}|^{k}}=\left(\frac{1}{|R|}-\frac{1}{|\overline{R}|}\right)\sum_{\ell=0}^{k-1}\frac{1}{|R|^{\ell}|\overline{R}|^{k-1-\ell}}, ∣ R ∣ k 1 − ∣ R ∣ k 1 = ( ∣ R ∣ 1 − ∣ R ∣ 1 ) ℓ = 0 ∑ k − 1 ∣ R ∣ ℓ ∣ R ∣ k − 1 − ℓ 1 ,
1 ∣ R ( a ) ∣ k − 1 ∣ R ( b ) ∣ k = ( 1 ∣ R ( a ) ∣ − 1 ∣ R ( b ) ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R ( a ) ∣ ℓ ∣ R ( b ) ∣ k − 1 − ℓ . (211) \frac{1}{|R^{(a)}|^{k}}-\frac{1}{|R^{(b)}|^{k}}=\left(\frac{1}{|R^{(a)}|}-\frac{1}{|R^{(b)}|}\right)\sum_{\ell=0}^{k-1}\frac{1}{|R^{(a)}|^{\ell}|R^{(b)}|^{k-1-\ell}} . \tag*{(211)} ∣ R ( a ) ∣ k 1 − ∣ R ( b ) ∣ k 1 = ( ∣ R ( a ) ∣ 1 − ∣ R ( b ) ∣ 1 ) ℓ = 0 ∑ k − 1 ∣ R ( a ) ∣ ℓ ∣ R ( b ) ∣ k − 1 − ℓ 1 . (211)
In addition, we note the following expansions for R R R in the direction of the normal vectors n ( x ′ ) n(x') n ( x ′ ) and n ( x ) n(x) n ( x ) . Noting that n ′ : = n ( x ′ ) = e r ( s − s ^ , θ − θ ^ ) n':=n(x')=e_r(s-\hat{s},\theta-\hat{\theta}) n ′ := n ( x ′ ) = e r ( s − s ^ , θ − θ ^ ) and n = n ( x ) = e r ( s , θ ) n=n(x)=e_r(s,\theta) n = n ( x ) = e r ( s , θ ) , we have
R ⋅ n ′ = − 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ′ ( s , s ^ , θ − θ ^ ) = R ‾ ⋅ n ′ ‾ + s ^ 2 Q R n ′ , R\cdot n'=-2\epsilon\sin^{2}\left(\frac{\hat{\theta}}{2}\right)+\hat{s}^{2}Q_{Rn'}(s,\hat{s},\theta-\hat{\theta})=\overline{R}\cdot\overline{n'}+\hat{s}^{2}Q_{Rn'}, R ⋅ n ′ = − 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ′ ( s , s ^ , θ − θ ^ ) = R ⋅ n ′ + s ^ 2 Q R n ′ ,
R ⋅ n = 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ( s , s ^ , θ − θ ^ ) = R ‾ ⋅ n ‾ + s ^ 2 Q R n . (212) R\cdot n=2\epsilon\sin^{2}\left(\frac{\hat{\theta}}{2}\right)+\hat{s}^{2}Q_{Rn}(s,\hat{s},\theta-\hat{\theta})=\overline{R}\cdot\overline{n}+\hat{s}^{2}Q_{Rn}. \tag*{(212)} R ⋅ n = 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ( s , s ^ , θ − θ ^ ) = R ⋅ n + s ^ 2 Q R n . (212)
Furthermore, using (203) , we have
R ⋅ e t ( s ) = s ^ + ϵ s ^ κ ( s , θ ) + s ^ 2 Q R t ( s , s ^ , θ , θ ^ ) = R ‾ ⋅ e z + ϵ s ^ κ ( s , θ ) + s ^ 2 Q R t R\cdot e_t(s)=\hat{s}+\epsilon\hat{s}\kappa(s,\theta)+\hat{s}^{2}Q_{Rt}(s,\hat{s},\theta,\hat{\theta})=\overline{R}\cdot e_z+\epsilon\hat{s}\kappa(s,\theta)+\hat{s}^{2}Q_{Rt} R ⋅ e t ( s ) = s ^ + ϵ s ^ κ ( s , θ ) + s ^ 2 Q R t ( s , s ^ , θ , θ ^ ) = R ⋅ e z + ϵ s ^ κ ( s , θ ) + s ^ 2 Q R t
R ⋅ e θ ( s , θ ) = ϵ sin ( θ ^ ) + ϵ s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ( s , s , θ ^ , θ ^ ) = R ‾ ⋅ e θ ( θ ) + ϵ s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ \begin{aligned}
R\cdot e_\theta(s,\theta)=\epsilon\sin(\hat{\theta})+\epsilon\hat{s}\kappa_3\cos(\theta-\hat{\theta})+\hat{s}^{2}Q_{R\theta}(s,s,\hat{\theta},\hat{\theta}) \\
&=\overline{R}\cdot e_\theta(\theta)+\epsilon\hat{s}\kappa_3\cos(\theta-\hat{\theta})+\hat{s}^{2}Q_{R\theta}
\end{aligned} R ⋅ e θ ( s , θ ) = ϵ sin ( θ ^ ) + ϵ s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ( s , s , θ ^ , θ ^ ) = R ⋅ e θ ( θ ) + ϵ s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ
R ⋅ φ = s ^ e t ( s ) ⋅ φ + 2 ϵ sin ( θ ^ 2 ) e θ ( s , θ − θ ^ 2 ) ⋅ φ − s ^ 2 Q t ⋅ φ + ϵ s ^ Q r ⋅ φ = R ‾ ⋅ ( Φ − 1 φ ) − s ^ 2 Q t ⋅ φ + ϵ s ^ Q r ⋅ φ . (213) \begin{aligned}
R\cdot\varphi=\hat{s}e_t(s)\cdot\varphi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta(s,\theta-\frac{\hat{\theta}}{2})\cdot\varphi-\hat{s}^{2}Q_t\cdot\varphi+\epsilon\hat{s}Q_r\cdot\varphi\\
&=\overline{R}\cdot(\Phi^{-1}\varphi)-\hat{s}^{2}Q_t\cdot\varphi+\epsilon\hat{s}Q_r\cdot\varphi. \tag*{(213)}
\end{aligned} R ⋅ φ = s ^ e t ( s ) ⋅ φ + 2 ϵ sin ( 2 θ ^ ) e θ ( s , θ − 2 θ ^ ) ⋅ φ − s ^ 2 Q t ⋅ φ + ϵ s ^ Q r ⋅ φ = R ⋅ ( Φ − 1 φ ) − s ^ 2 Q t ⋅ φ + ϵ s ^ Q r ⋅ φ . (213)
Here, given two curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) in C 2 , β C^{2,\beta} C 2 , β , each of the remainder terms appearing in (212) and (213) satisfies an estimate of the form
∥ Q ( a ) − Q ( b ) ∥ Y ≤ ϵ − β c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ∑ ℓ = 1 3 ∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ Y (214) \left\|Q^{(a)}-Q^{(b)}\right\|_{Y}\leq\epsilon^{-\beta}c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)})\sum_{\ell=1}^{3}\left\|\kappa_{\ell}^{(a)}-\kappa_{\ell}^{(b)}\right\|_{Y} \tag*{(214)} Q ( a ) − Q ( b ) Y ≤ ϵ − β c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ℓ = 1 ∑ 3 κ ℓ ( a ) − κ ℓ ( b ) Y (214)
where the norms ∥ ⋅ ∥ Y \|\cdot\|_{Y} ∥ ⋅ ∥ Y are given by (200) and β = 0 \beta=0 β = 0 for the L ∞ L^{\infty} L ∞ bound. Here and throughout, we note that if the two curves b o l d s y m b o l X ( a ) boldsymbol{X}^{(a)} b o l d sy mb o l X ( a ) and b o l d s y m b o l X ( b ) boldsymbol{X}^{(b)} b o l d sy mb o l X ( b ) are sufficiently close in the sense of Lemma 1.1, then the bounds of the form (214) imply that
∥ Q ( a ) − Q ( b ) ∥ Y ≤ ϵ − β c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ∥ X ( a ) − X ( b ) ∥ C 2 , β ( T ) . (215) \left\|Q^{(a)}-Q^{(b)}\right\|_{Y}\leq\epsilon^{-\beta}c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)})\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}(\mathbb{T})}. \tag*{(215)} Q ( a ) − Q ( b ) Y ≤ ϵ − β c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) X ( a ) − X ( b ) C 2 , β ( T ) . (215)
Given all of the above expansions, we now state a series of lemmas that will be used to estimate the actual boundary integral expressions appearing in the decomposition (66) and to obtain Lemmas 1.10, 1.11, and 1.12. Each of the following (Lemmas 4.1 - 4.4) comes directly from [59 section 3] and will thus be stated here without proof.
Lemma 4.1 (Relating R R R and R ‾ \overline{R} R ). Given R ‾ \overline{R} R , R R R as in (196), (205), respectively, the following bounds hold for ε \varepsilon ε sufficiently small:
∣ ∣ R ∣ − ∣ R ‾ ∣ ∣ ≤ κ ∗ 2 s ^ 2 + c ( κ ∗ ) ε ∣ s ^ ∣ (216) \left||R|-\left|\overline{R}\right|\right| \leq\frac{\kappa_*}{2}\widehat{s}^{2}+c(\kappa_*)\varepsilon|\widehat{s}| \tag*{(216)} ∣ R ∣ − R ≤ 2 κ ∗ s 2 + c ( κ ∗ ) ε ∣ s ∣ (216) ∣ R ∣ ≥ c ( κ ∗ , c Γ ) ∣ R ‾ ∣ . (217) |R| \geq c(\kappa_*,c_{\Gamma})\left|\overline{R}\right|. \tag*{(217)} ∣ R ∣ ≥ c ( κ ∗ , c Γ ) R . (217)
In addition to Lemma 4.1, we have the following series of integral estimates.
Lemma 4.2 (Integral bounds for R R R and R ‾ \overline{R} R ). Given R R R , R ‾ \overline{R} R as in (205), (196), respectively, and given 0 ≤ α < 1 0 \leq\alpha< 1 0 ≤ α < 1 , let k k k be an integer satisfying k − α < 2 k-\alpha< 2 k − α < 2 . We have
∫ − 1 / 2 1 / 2 ∫ − π π 1 ∣ R ∣ k − α ε d θ d s ^ ≤ c ( κ ∗ , c Γ ) ∫ − 1 / 2 1 / 2 ∫ − π π 1 ∣ R ‾ ∣ k − α ε d θ d s ^ ≤ { c ε 2 − k + α , 1 < k − α < 2 , c ε , k − α ≤ 1. (218) \int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{|R|^{k-\alpha}}\varepsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}
\leq c(\kappa_*,c_{\Gamma})
\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{|\overline{R}|^{k-\alpha}}\varepsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}
\leq
\begin{cases}
c\varepsilon^{2-k+\alpha}, & 1<k-\alpha<2,\\
c\varepsilon, & k-\alpha\leq1.
\end{cases} \tag*{(218)} ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ k − α 1 ε d θ d s ≤ c ( κ ∗ , c Γ ) ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ k − α 1 ε d θ d s ≤ { c ε 2 − k + α , c ε , 1 < k − α < 2 , k − α ≤ 1. (218)
Lemma 4.3 (Bound for integrands with odd powers). Let R ‾ \overline{R} R , R R R be as in (196), (205) and let ℓ , k , n , m \ell,k,n,m ℓ , k , n , m be nonnegative integers with ℓ + k + 2 = n + m \ell+k+2=n+m ℓ + k + 2 = n + m and ℓ + k \ell+k ℓ + k odd. Given g , φ ∈ C 0 , α ( Γ ε ) g,\varphi\in C^{0,\alpha}(\Gamma_{\varepsilon}) g , φ ∈ C 0 , α ( Γ ε ) , we have
∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π s ^ ℓ ( ε sin ( θ ^ / 2 ) ) k g ( s ^ , θ ^ ) ∣ R ‾ ∣ m ∣ R ∣ n φ ( s − s ^ , θ − θ ^ ) ε d θ d s ^ ∣ ≤ c ( κ ∗ , c Γ ) ε α ∥ g ∥ C 0 , α ∥ φ ∥ C 0 , α . (219) \left|\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
\frac{\widehat{s}^{\ell}(\varepsilon\sin(\widehat{\theta}/2))^{k}g(\widehat{s},\widehat{\theta})}
{|\overline{R}|^{m}|R|^{n}}
\varphi(s-\widehat{s},\theta-\widehat{\theta})\varepsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}\right|
\leq c(\kappa_*,c_{\Gamma})\varepsilon^{\alpha}
\|g\|_{C^{0,\alpha}}\|\varphi\|_{C^{0,\alpha}}. \tag*{(219)} p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ m ∣ R ∣ n s ℓ ( ε sin ( θ /2 ) ) k g ( s , θ ) φ ( s − s , θ − θ ) ε d θ d s ≤ c ( κ ∗ , c Γ ) ε α ∥ g ∥ C 0 , α ∥ φ ∥ C 0 , α . (219)
Here p.v. is the Cauchy principal value, given along Γ ε \Gamma_{\varepsilon} Γ ε by
p.v. ∫ − 1 / 2 1 / 2 ∫ − π π h ( s ^ , θ ^ ) ε d θ d s ^ = lim δ → 0 + ( ∫ − 1 / 2 − δ + ∫ δ 1 / 2 ) ( ∫ − π − 2 π δ + ∫ 2 π δ π ) h ( s ^ , θ ^ ) ε d θ d s ^ . \operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}h(\widehat{s},\widehat{\theta})\varepsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}
=
\lim_{\delta\to0^+}
\left(\int_{-1/2}^{-\delta}+\int_{\delta}^{1/2}\right)
\left(\int_{-\pi}^{-2\pi\delta}+\int_{2\pi\delta}^{\pi}\right)
h(\widehat{s},\widehat{\theta})\varepsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}. p.v. ∫ − 1/2 1/2 ∫ − π π h ( s , θ ) ε d θ d s = δ → 0 + lim ( ∫ − 1/2 − δ + ∫ δ 1/2 ) ( ∫ − π − 2 π δ + ∫ 2 π δ π ) h ( s , θ ) ε d θ d s .
Finally, we state the following bound for the C ˙ 0 , α \dot{C}^{0,\alpha} C ˙ 0 , α seminorm of integral operators along Γ ε \Gamma_{\varepsilon} Γ ε .
Lemma 4.4 (Estimating ∣ ⋅ ∣ C ˙ 0 , α |\cdot|_{\dot{C}^{0,\alpha}} ∣ ⋅ ∣ C ˙ 0 , α seminorms). Let ℓ , k , m , n \ell,k,m,n ℓ , k , m , n be nonnegative integers satisfying either
ℓ + k = m + n − 1 \ell+k=m+n-1 ℓ + k = m + n − 1 or
ℓ + k = m + n − 2 \ell+k=m+n-2 ℓ + k = m + n − 2 and ℓ + k \ell+k ℓ + k is odd.
Given g ( s , θ , s − s ^ , θ − θ ^ ) ∈ C 0 , α + ( Γ ε × Γ ε ) g(s,\theta,s-\widehat{s},\theta-\widehat{\theta})\in C^{0,\alpha^+}(\Gamma_{\varepsilon}\times\Gamma_{\varepsilon}) g ( s , θ , s − s , θ − θ ) ∈ C 0 , α + ( Γ ε × Γ ε ) for any 0 < α < α + < 1 0<\alpha<\alpha^+<1 0 < α < α + < 1 , for R ‾ \overline{R} R and R R R as in (196), (205), let K ℓ k m n ( s , θ , s ^ , θ ^ ) K_{\ell kmn}(s,\theta,\widehat{s},\widehat{\theta}) K ℓ k mn ( s , θ , s , θ ) denote
K ℓ k m n ( s , θ , s ^ , θ ^ ) : = s ^ ℓ ( ε sin ( θ ^ / 2 ) ) k g ( s , θ , s − s ^ , θ − θ ^ ) ∣ R ‾ ( s , θ , s ^ , θ ^ ) ∣ m ∣ R ( s , θ , s ^ , θ ^ ) ∣ n . (220) K_{\ell kmn}(s,\theta,\widehat{s},\widehat{\theta})
:=
\frac{\widehat{s}^{\ell}(\varepsilon\sin(\widehat{\theta}/2))^{k}g(s,\theta,s-\widehat{s},\theta-\widehat{\theta})}
{|\overline{R}(s,\theta,\widehat{s},\widehat{\theta})|^{m}|R(s,\theta,\widehat{s},\widehat{\theta})|^{n}}. \tag*{(220)} K ℓ k mn ( s , θ , s , θ ) := ∣ R ( s , θ , s , θ ) ∣ m ∣ R ( s , θ , s , θ ) ∣ n s ℓ ( ε sin ( θ /2 ) ) k g ( s , θ , s − s , θ − θ ) . (220) Suppose φ ( s , θ ) ∈ C 0 , α ( Γ ε ) \varphi(s,\theta)\in C^{0,\alpha}(\Gamma_{\varepsilon}) φ ( s , θ ) ∈ C 0 , α ( Γ ε ) . Then for any ( s 0 , θ 0 ) ∈ [ − 1 / 2 , 1 / 2 ] × [ − π , π ] (s_0,\theta_0)\in[-1/2,1/2]\times[-\pi,\pi] ( s 0 , θ 0 ) ∈ [ − 1/2 , 1/2 ] × [ − π , π ] with s 0 2 + θ 0 2 ≠ 0 s_0^2+\theta_0^2\ne0 s 0 2 + θ 0 2 = 0 we have
∣ p.v. ∫ − s 0 − 1 / 2 − s 0 + 1 / 2 ∫ − θ 0 − π − θ 0 + π K ℓ k m n ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) φ ( s − s ^ , θ − θ ^ ) ε d θ d s ^ − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π K ℓ k m n ( s , θ , s ^ , θ ^ ) φ ( s − s ^ , θ − θ ^ ) ε d θ d s ^ ∣ ≤ { c ( κ ∗ , c Γ ) ε 1 − α ∥ g ∥ C 1 0 , α ∥ φ ∥ L ∞ s 0 2 + ε 2 θ 0 2 α , in case (1) , c ( κ ∗ , c Γ ) ( ∥ g ∥ C 1 0 , α + + ∥ g ∥ C 2 0 , α ) ∥ φ ∥ C 0 , α s 0 2 + ε 2 θ 0 2 α , in case (2) . (221) \begin{aligned}
&\left|
\operatorname{p.v.}\int_{-s_0-1/2}^{-s_0+1/2}
\int_{-\theta_0-\pi}^{-\theta_0+\pi}
K_{\ell kmn}(s_0+s,\theta_0+\theta,s_0+\widehat{s},\theta_0+\widehat{\theta})
\varphi(s-\widehat{s},\theta-\widehat{\theta})
\varepsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}\right.\\
&\left.\qquad-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
K_{\ell kmn}(s,\theta,\widehat{s},\widehat{\theta})
\varphi(s-\widehat{s},\theta-\widehat{\theta})
\varepsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}
\right|\\
&\leq
\begin{cases}
c(\kappa_*,c_{\Gamma})\varepsilon^{1-\alpha}\|g\|_{C_1^{0,\alpha}}\|\varphi\|_{L^\infty}\sqrt{s_0^2+\varepsilon^2\theta_0^2}^{\alpha},
& \text{in case (1)},\\
c(\kappa_*,c_{\Gamma})\left(\|g\|_{C_1^{0,\alpha^+}}+\|g\|_{C_2^{0,\alpha}}\right)\|\varphi\|_{C^{0,\alpha}}\sqrt{s_0^2+\varepsilon^2\theta_0^2}^{\alpha},
& \text{in case (2)}.
\end{cases} \tag*{(221)}
\end{aligned} p.v. ∫ − s 0 − 1/2 − s 0 + 1/2 ∫ − θ 0 − π − θ 0 + π K ℓ k mn ( s 0 + s , θ 0 + θ , s 0 + s , θ 0 + θ ) φ ( s − s , θ − θ ) ε d θ d s − p.v. ∫ − 1/2 1/2 ∫ − π π K ℓ k mn ( s , θ , s , θ ) φ ( s − s , θ − θ ) ε d θ d s ≤ ⎩ ⎨ ⎧ c ( κ ∗ , c Γ ) ε 1 − α ∥ g ∥ C 1 0 , α ∥ φ ∥ L ∞ s 0 2 + ε 2 θ 0 2 α , c ( κ ∗ , c Γ ) ( ∥ g ∥ C 1 0 , α + + ∥ g ∥ C 2 0 , α ) ∥ φ ∥ C 0 , α s 0 2 + ε 2 θ 0 2 α , in case (1) , in case (2) . (221) where ∥ ⋅ ∥ C 1 0 , α \|\cdot\|_{C_1^{0,\alpha}} ∥ ⋅ ∥ C 1 0 , α and ∥ ⋅ ∥ C 2 0 , α \|\cdot\|_{C_2^{0,\alpha}} ∥ ⋅ ∥ C 2 0 , α are as in (199).
Single layer remainder terms
Given the tools of section 4.1, we may proceed to the proof of Lemma 1.10 regarding the mapping properties of the single layer remainder R S \mathcal{R}_{S} R S , defined in (65) . We restate the lemma here for convenience.
Lemma 4.5 (Single layer remainder). Let 0 < α < β 0<\alpha<\beta 0 < α < β and consider a filament Σ ϵ \Sigma_{\epsilon} Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) X(s)\in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Let the single layer remainder R S \mathcal{R}_{S} R S be as defined in (65) , and let the map Φ \Phi Φ be as in (27) . Given φ ∈ C 0 , α ( Γ ϵ ) \varphi\in C^{0,\alpha}(\Gamma_{\epsilon}) φ ∈ C 0 , α ( Γ ϵ ) with ∫ T Φ − 1 φ ( s , θ ) d s = 0 \int_{\mathbb{T}}\Phi^{-1}\varphi(s,\theta)\,\mathrm{d}s=0 ∫ T Φ − 1 φ ( s , θ ) d s = 0 , R S \mathcal{R}_{S} R S may be decomposed as
R S [ φ ] = R S , ϵ [ φ ] + R S , + [ φ ] \mathcal{R}_{S}[\varphi]=\mathcal{R}_{S,\epsilon}[\varphi]+\mathcal{R}_{S,+}[\varphi] R S [ φ ] = R S , ϵ [ φ ] + R S , + [ φ ] where
∥ R S , ϵ [ φ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ , ∣ R S , ϵ [ φ ] ∣ C ˚ ϵ 1 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α ∥ R S , + [ φ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ , ∣ R S , + [ φ ] ∣ C ˚ ϵ 1 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ C 0 , α . (222) \begin{aligned}
\|\mathcal{R}_{S,\epsilon}[\varphi]\|_{C^{0,\alpha}}&\leq c(\kappa_{*,\alpha},c_{\Gamma})\epsilon^{2-\alpha}\|\varphi\|_{L^\infty},&
\left|\mathcal{R}_{S,\epsilon}[\varphi]\right|_{\mathring{C}^{1,\alpha}_{\epsilon}}&\leq c(\kappa_{*,\alpha_+},c_{\Gamma})\epsilon^{1-\alpha_+}\|\varphi\|_{C^{0,\alpha}}\\
\|\mathcal{R}_{S,+}[\varphi]\|_{C^{0,\alpha}}&\leq c(\kappa_{*,\alpha},c_{\Gamma})\epsilon^{1-\alpha}\|\varphi\|_{L^\infty},&
\left|\mathcal{R}_{S,+}[\varphi]\right|_{\mathring{C}^{1,\beta}_{\epsilon}}&\leq c(\kappa_{*,\beta},c_{\Gamma})\epsilon^{1-2\beta}\|\varphi\|_{C^{0,\alpha}}. \tag*{(222)}
\end{aligned} ∥ R S , ϵ [ φ ] ∥ C 0 , α ∥ R S , + [ φ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ , ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ , ∣ R S , ϵ [ φ ] ∣ C ˚ ϵ 1 , α ∣ R S , + [ φ ] ∣ C ˚ ϵ 1 , β ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ C 0 , α . (222) Furthermore, given two nearby filaments with centerlines X ( a ) ( s ) X^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) X^{(b)}(s) X ( b ) ( s ) satisfying Lemma 1.1 , let φ 0 Φ ( a ) = φ − ∫ T ( Φ ( a ) ) − 1 φ ( s , θ ) d s \varphi^{\Phi^{(a)}}_{0}=\varphi-\int_{\mathbb{T}}(\Phi^{(a)})^{-1}\varphi(s,\theta)\,\mathrm{d}s φ 0 Φ ( a ) = φ − ∫ T ( Φ ( a ) ) − 1 φ ( s , θ ) d s and φ 0 Φ ( b ) = φ − ∫ T ( Φ ( b ) ) − 1 φ ( s , θ ) d s \varphi^{\Phi^{(b)}}_{0}=\varphi-\int_{\mathbb{T}}(\Phi^{(b)})^{-1}\varphi(s,\theta)\,\mathrm{d}s φ 0 Φ ( b ) = φ − ∫ T ( Φ ( b ) ) − 1 φ ( s , θ ) d s . The difference between the corresponding single layer remainders R S ( a ) \mathcal{R}^{(a)}_{S} R S ( a ) and R S ( b ) \mathcal{R}^{(b)}_{S} R S ( b ) may be decomposed as above, and the components satisfy
∥ R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ ∣ R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] ∣ C ϵ 1 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α ∥ R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ ∣ R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] ∣ C ϵ 1 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ C 0 , α . (223) \begin{aligned}
\left\|\mathcal{R}^{(a)}_{S,\epsilon}[\varphi^{\Phi^{(a)}}_{0}]-\mathcal{R}^{(b)}_{S,\epsilon}[\varphi^{\Phi^{(b)}}_{0}]\right\|_{C^{0,\alpha}}
&\leq c(\kappa^{(a)}_{*,\alpha},\kappa^{(b)}_{*,\alpha},c_{\Gamma})\epsilon^{2-\alpha}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty}\\
\left|\mathcal{R}^{(a)}_{S,\epsilon}[\varphi^{\Phi^{(a)}}_{0}]-\mathcal{R}^{(b)}_{S,\epsilon}[\varphi^{\Phi^{(b)}}_{0}]\right|_{C^{1,\alpha}_{\epsilon}}
&\leq c(\kappa^{(a)}_{*,\alpha_+},\kappa^{(b)}_{*,\alpha_+},c_{\Gamma})\epsilon^{1-\alpha_+}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha_+}}\|\varphi\|_{C^{0,\alpha}}\\
\left\|\mathcal{R}^{(a)}_{S,+}[\varphi^{\Phi^{(a)}}_{0}]-\mathcal{R}^{(b)}_{S,+}[\varphi^{\Phi^{(b)}}_{0}]\right\|_{C^{0,\alpha}}
&\leq c(\kappa^{(a)}_{*,\alpha},\kappa^{(b)}_{*,\alpha},c_{\Gamma})\epsilon^{1-\alpha}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty}\\
\left|\mathcal{R}^{(a)}_{S,+}[\varphi^{\Phi^{(a)}}_{0}]-\mathcal{R}^{(b)}_{S,+}[\varphi^{\Phi^{(b)}}_{0}]\right|_{C^{1,\beta}_{\epsilon}}
&\leq c(\kappa^{(a)}_{*,\beta},\kappa^{(b)}_{*,\beta},c_{\Gamma})\epsilon^{1-2\beta}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}}\|\varphi\|_{C^{0,\alpha}}. \tag*{(223)}
\end{aligned} R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] C 0 , α R S , ϵ ( a ) [ φ 0 Φ ( a ) ] − R S , ϵ ( b ) [ φ 0 Φ ( b ) ] C ϵ 1 , α R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] C 0 , α R S , + ( a ) [ φ 0 Φ ( a ) ] − R S , + ( b ) [ φ 0 Φ ( b ) ] C ϵ 1 , β ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞ ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞ ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β X ( a ) − X ( b ) C 2 , β ∥ φ ∥ C 0 , α . (223)
Proof of Lemma 1.10. Recall the definition of R S \mathcal{R}_{S} R S from (65) and note that by assumption φ ( s , θ ) = φ 0 Φ ( s , θ ) \varphi(s,\theta)=\varphi^{\Phi}_{0}(s,\theta) φ ( s , θ ) = φ 0 Φ ( s , θ ) since we consider ∫ T Φ − 1 φ ( s , θ ) d s = 0 \int_{\mathbb{T}}\Phi^{-1}\varphi(s,\theta)\,\mathrm{d}s=0 ∫ T Φ − 1 φ ( s , θ ) d s = 0 . Rewriting each integrand in terms of the surface parameters s , s ^ , θ , θ ^ s,\widehat{s},\theta,\widehat{\theta} s , s , θ , θ , we may decompose R S \mathcal{R}_{S} R S as follows:
R S [ φ ( s , θ ) ] = Φ − 1 S [ φ ( s , θ ) ] − S ‾ [ Φ − 1 φ ( s , θ ) ] = R S , 0 [ φ ] + R S , 1 [ φ ] + R S , 2 [ φ ] + R S , 3 [ φ ] , R S , 0 [ φ ] = − ( ∫ − ∞ − 1 / 2 + ∫ 1 / 2 ∞ ) ∫ − π π G ‾ ( s , s ^ , θ , θ ^ ) Φ − 1 φ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ , R S , 1 [ φ ] = 1 8 π ( Φ − 1 ∫ − 1 / 2 1 / 2 ∫ − π π 1 ∣ R ‾ ∣ φ ( s − s ^ , θ − θ ^ ) ϵ d s ^ d θ ^ − ∫ − 1 / 2 1 / 2 ∫ − π π 1 ∣ R ‾ ∣ ( Φ − 1 φ ) ( s − s ^ , θ − θ ^ ) ϵ d s ^ d θ ^ ) , R S , 2 [ φ ] = 1 8 π ( Φ − 1 ∫ − 1 / 2 1 / 2 ∫ − π π R ‾ ⊗ R ‾ ∣ R ‾ ∣ 3 φ ( s − s ^ , θ − θ ^ ) ϵ d s ^ d θ ^ − ∫ − 1 / 2 1 / 2 ∫ − π π R ‾ ⊗ R ‾ ∣ R ‾ ∣ 3 ( Φ − 1 φ ) ( s − s ^ , θ − θ ^ ) ϵ d s ^ d θ ^ ) , R S , 3 [ φ ] = − Φ − 1 ∫ − 1 / 2 1 / 2 ∫ − π π G ( s , s ^ , θ , θ ^ ) φ ( s − s ^ , θ − θ ^ ) ϵ 2 κ ^ d θ ^ d s ^ . (224) \begin{aligned}
\mathcal{R}_{S}[\varphi(s,\theta)]&=\Phi^{-1}S[\varphi(s,\theta)]-\overline{S}[\Phi^{-1}\varphi(s,\theta)]\\
&=\mathcal{R}_{S,0}[\varphi]+\mathcal{R}_{S,1}[\varphi]+\mathcal{R}_{S,2}[\varphi]+\mathcal{R}_{S,3}[\varphi],\\
\mathcal{R}_{S,0}[\varphi]&=-\left(\int_{-\infty}^{-1/2}+\int_{1/2}^{\infty}\right)\int_{-\pi}^{\pi}\overline{\mathcal{G}}(s,\widehat{s},\theta,\widehat{\theta})\Phi^{-1}\varphi(s-\widehat{s},\theta-\widehat{\theta})\epsilon\,\mathrm{d}\widehat{\theta}\mathrm{d}\widehat{s},\\
\mathcal{R}_{S,1}[\varphi]&=\frac{1}{8\pi}\left(\Phi^{-1}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{|\overline{R}|}\varphi(s-\widehat{s},\theta-\widehat{\theta})\epsilon\,\mathrm{d}\widehat{s}\mathrm{d}\widehat{\theta}\right.\\
&\qquad\left.-\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{|\overline{R}|}(\Phi^{-1}\varphi)(s-\widehat{s},\theta-\widehat{\theta})\epsilon\,\mathrm{d}\widehat{s}\mathrm{d}\widehat{\theta}\right),\\
\mathcal{R}_{S,2}[\varphi]&=\frac{1}{8\pi}\left(\Phi^{-1}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\overline{R}\otimes\overline{R}}{|\overline{R}|^{3}}\varphi(s-\widehat{s},\theta-\widehat{\theta})\epsilon\,\mathrm{d}\widehat{s}\mathrm{d}\widehat{\theta}\right.\\
&\qquad\left.-\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\overline{R}\otimes\overline{R}}{|\overline{R}|^{3}}(\Phi^{-1}\varphi)(s-\widehat{s},\theta-\widehat{\theta})\epsilon\,\mathrm{d}\widehat{s}\mathrm{d}\widehat{\theta}\right),\\
\mathcal{R}_{S,3}[\varphi]&=-\Phi^{-1}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\mathcal{G}(s,\widehat{s},\theta,\widehat{\theta})\varphi(s-\widehat{s},\theta-\widehat{\theta})\epsilon^{2}\widehat{\kappa}\,\mathrm{d}\widehat{\theta}\mathrm{d}\widehat{s}. \tag*{(224)}
\end{aligned} R S [ φ ( s , θ )] R S , 0 [ φ ] R S , 1 [ φ ] R S , 2 [ φ ] R S , 3 [ φ ] = Φ − 1 S [ φ ( s , θ )] − S [ Φ − 1 φ ( s , θ )] = R S , 0 [ φ ] + R S , 1 [ φ ] + R S , 2 [ φ ] + R S , 3 [ φ ] , = − ( ∫ − ∞ − 1/2 + ∫ 1/2 ∞ ) ∫ − π π G ( s , s , θ , θ ) Φ − 1 φ ( s − s , θ − θ ) ϵ d θ d s , = 8 π 1 ( Φ − 1 ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 1 φ ( s − s , θ − θ ) ϵ d s d θ − ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 1 ( Φ − 1 φ ) ( s − s , θ − θ ) ϵ d s d θ ) , = 8 π 1 ( Φ − 1 ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 3 R ⊗ R φ ( s − s , θ − θ ) ϵ d s d θ − ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 3 R ⊗ R ( Φ − 1 φ ) ( s − s , θ − θ ) ϵ d s d θ ) , = − Φ − 1 ∫ − 1/2 1/2 ∫ − π π G ( s , s , θ , θ ) φ ( s − s , θ − θ ) ϵ 2 κ d θ d s . (224)
We begin with bounds for the smooth remainder R S , 0 \mathcal{R}_{S,0} R S , 0 away from the singularity at ( s ^ , θ ^ ) = ( 0 , 0 ) (\hat{s},\hat{\theta})=(0,0) ( s ^ , θ ^ ) = ( 0 , 0 ) . Since ∫ T Φ − 1 φ ( s , θ ) d s = 0 \int_{\mathbb{T}}\Phi^{-1}\varphi(s,\theta)\,ds=0 ∫ T Φ − 1 φ ( s , θ ) d s = 0 , we may write Φ − 1 φ ( s , θ ) = ∂ s φ ‾ ( s , θ ) \Phi^{-1}\varphi(s,\theta)=\partial_s\overline{\varphi}(s,\theta) Φ − 1 φ ( s , θ ) = ∂ s φ ( s , θ ) for some φ ‾ \overline{\varphi} φ with φ ‾ ( 0 , θ ) = 0 \overline{\varphi}(0,\theta)=0 φ ( 0 , θ ) = 0 . Using ∂ s φ ‾ ( s , θ ) \partial_s\overline{\varphi}(s,\theta) ∂ s φ ( s , θ ) in the expression for R S , 0 \mathcal{R}_{S,0} R S , 0 , we may then integrate by parts in s ^ \hat{s} s ^ to obtain
R S , 0 [ φ ] = − ( ∫ − ∞ − 1 / 2 + ∫ 1 / 2 + ∞ ) ∫ − π π ( ∂ s G ‾ ) φ ‾ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ + ∫ − π π G ‾ φ ‾ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ ∣ s ^ = − 1 / 2 s ^ = 1 / 2 . \mathcal{R}_{S,0}[\varphi]=-\left(\int_{-\infty}^{-1/2}+\int_{1/2}^{+\infty}\right)\int_{-\pi}^{\pi}(\partial_s\overline{\mathcal{G}})\overline{\varphi}(s-\hat{s},\theta-\hat{\theta})\,\epsilon d\hat{\theta}d\hat{s}
+\int_{-\pi}^{\pi}\overline{\mathcal{G}}\,\overline{\varphi}(s-\hat{s},\theta-\hat{\theta})\,\epsilon d\hat{\theta}\bigg|_{\hat{s}=-1/2}^{\hat{s}=1/2}. R S , 0 [ φ ] = − ( ∫ − ∞ − 1/2 + ∫ 1/2 + ∞ ) ∫ − π π ( ∂ s G ) φ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ + ∫ − π π G φ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ s ^ = − 1/2 s ^ = 1/2 .
Using the form (37) of the Stokeslet kernel G \mathcal{G} G , we have
∣ R S , 0 ∣ ≤ c ϵ ∥ φ ∥ L ∞ ∫ 1 / 2 ∞ 1 s ^ 2 d s ^ + c ϵ ∥ φ ∥ L ∞ ≤ c ϵ ∥ φ ∥ L ∞ = c ϵ ∥ ∫ 0 s ( Φ − 1 φ ) d s ′ ∥ L ∞ ≤ c ϵ ∥ φ ∥ L ∞ . |\mathcal{R}_{S,0}|\le c\epsilon\|\varphi\|_{L^\infty}\int_{1/2}^{\infty}\frac{1}{\hat{s}^{2}}\,d\hat{s}+c\epsilon\|\varphi\|_{L^\infty}\le c\epsilon\|\varphi\|_{L^\infty}=c\epsilon\left\|\int_{0}^{s}(\Phi^{-1}\varphi)\,ds'\right\|_{L^\infty}\le c\epsilon\|\varphi\|_{L^\infty}. ∣ R S , 0 ∣ ≤ cϵ ∥ φ ∥ L ∞ ∫ 1/2 ∞ s ^ 2 1 d s ^ + cϵ ∥ φ ∥ L ∞ ≤ cϵ ∥ φ ∥ L ∞ = cϵ ∫ 0 s ( Φ − 1 φ ) d s ′ L ∞ ≤ cϵ ∥ φ ∥ L ∞ .
Furthermore, we may estimate
∥ ∂ s 2 R S , 0 ∥ L ∞ ≤ c ϵ ∥ φ ∥ L ∞ ∫ 1 / 2 ∞ ( 1 ∣ s ^ ∣ 3 + ϵ ∣ s ^ ∣ 4 + ϵ 2 ∣ s ^ ∣ 5 ) d s ^ ≤ c ϵ ∥ φ ∥ L ∞ . \|\partial_s^2\mathcal{R}_{S,0}\|_{L^\infty}\le c\epsilon\|\varphi\|_{L^\infty}\int_{1/2}^{\infty}\left(\frac{1}{|\hat{s}|^3}+\frac{\epsilon}{|\hat{s}|^4}+\frac{\epsilon^2}{|\hat{s}|^5}\right)d\hat{s}\le c\epsilon\|\varphi\|_{L^\infty}. ∥ ∂ s 2 R S , 0 ∥ L ∞ ≤ cϵ ∥ φ ∥ L ∞ ∫ 1/2 ∞ ( ∣ s ^ ∣ 3 1 + ∣ s ^ ∣ 4 ϵ + ∣ s ^ ∣ 5 ϵ 2 ) d s ^ ≤ cϵ ∥ φ ∥ L ∞ .
Finally, given two nearby curves X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) satisfying Lemma 1.1, the corresponding remainder pieces R S , 0 ( a ) \mathcal{R}_{S,0}^{(a)} R S , 0 ( a ) and R S , 0 ( b ) \mathcal{R}_{S,0}^{(b)} R S , 0 ( b ) are nearly identical except for slight differences in the s s s -means ∫ T ( Φ ( a ) ) − 1 φ d s \int_{\mathbb{T}}(\Phi^{(a)})^{-1}\varphi\,ds ∫ T ( Φ ( a ) ) − 1 φ d s and ∫ T ( Φ ( b ) ) − 1 φ d s \int_{\mathbb{T}}(\Phi^{(b)})^{-1}\varphi\,ds ∫ T ( Φ ( b ) ) − 1 φ d s . We may write
∣ R S , 0 ( a ) [ φ 0 Φ ( a ) ] − R S , 0 ( b ) [ φ 0 Φ ( b ) ] ∣ = ∣ ( ∫ − ∞ − 1 / 2 + ∫ 1 / 2 ∞ ) ∫ − π π G ‾ ( ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) ) ϵ d θ ^ d s ^ ∣ ≤ ∣ ( ∫ − ∞ − 1 / 2 + ∫ 1 / 2 ∞ ) ∫ − π π ( ∂ s G ‾ ) ( φ ‾ ( a ) − φ ‾ ( b ) ) ϵ d θ ^ d s ^ ∣ + ∣ ∫ − π π G ‾ ( φ ‾ ( a ) − φ ‾ ( b ) ) ϵ d θ ^ ∣ s ^ = − 1 / 2 s ^ = 1 / 2 ∣ ≤ c ϵ ∥ φ ‾ ( a ) − φ ‾ ( b ) ∥ L ∞ ≤ c ϵ ∥ ∫ 0 s ( ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) ) d s ′ ∥ L ∞ ≤ c ϵ ∥ ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ , \begin{aligned}
\left|\mathcal{R}_{S,0}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,0}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right| \\
&=\left|\left(\int_{-\infty}^{-1/2}+\int_{1/2}^{\infty}\right)\int_{-\pi}^{\pi}\overline{\mathcal{G}}\left((\Phi^{(a)})^{-1}\varphi_0^{\Phi^{(a)}}-(\Phi^{(b)})^{-1}\varphi_0^{\Phi^{(b)}}\right)\epsilon d\hat{\theta}d\hat{s}\right| \\
&\le\left|\left(\int_{-\infty}^{-1/2}+\int_{1/2}^{\infty}\right)\int_{-\pi}^{\pi}(\partial_s\overline{\mathcal{G}})\left(\overline{\varphi}^{(a)}-\overline{\varphi}^{(b)}\right)\epsilon d\hat{\theta}d\hat{s}\right|
+\left|\int_{-\pi}^{\pi}\overline{\mathcal{G}}\left(\overline{\varphi}^{(a)}-\overline{\varphi}^{(b)}\right)\epsilon d\hat{\theta}\bigg|_{\hat{s}=-1/2}^{\hat{s}=1/2}\right| \\
&\le c\epsilon\|\overline{\varphi}^{(a)}-\overline{\varphi}^{(b)}\|_{L^\infty}
\le c\epsilon\left\|\int_{0}^{s}\left((\Phi^{(a)})^{-1}\varphi_0^{\Phi^{(a)}}-(\Phi^{(b)})^{-1}\varphi_0^{\Phi^{(b)}}\right)\,ds'\right\|_{L^\infty} \\
&\le c\epsilon\left\|(\Phi^{(a)})^{-1}\varphi_0^{\Phi^{(a)}}-(\Phi^{(b)})^{-1}\varphi_0^{\Phi^{(b)}}\right\|_{L^\infty}
\le c(\kappa_*^{(a)},\kappa_*^{(b)})\epsilon\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|\varphi\|_{L^\infty},
\end{aligned} R S , 0 ( a ) [ φ 0 Φ ( a ) ] − R S , 0 ( b ) [ φ 0 Φ ( b ) ] = ( ∫ − ∞ − 1/2 + ∫ 1/2 ∞ ) ∫ − π π G ( ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) ) ϵ d θ ^ d s ^ ≤ ( ∫ − ∞ − 1/2 + ∫ 1/2 ∞ ) ∫ − π π ( ∂ s G ) ( φ ( a ) − φ ( b ) ) ϵ d θ ^ d s ^ + ∫ − π π G ( φ ( a ) − φ ( b ) ) ϵ d θ ^ s ^ = − 1/2 s ^ = 1/2 ≤ cϵ ∥ φ ( a ) − φ ( b ) ∥ L ∞ ≤ cϵ ∫ 0 s ( ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) ) d s ′ L ∞ ≤ cϵ ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ ,
by the bound (81) . By a similar calculation, we additionally obtain
∥ ∂ s 2 R S , 0 ( a ) [ φ 0 Φ ( a ) ] − ∂ s 2 R S , 0 ( b ) [ φ 0 Φ ( b ) ] ∥ L ∞ ≤ c ϵ ∥ ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ . \begin{aligned}
\left\|\partial_s^2\mathcal{R}_{S,0}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\partial_s^2\mathcal{R}_{S,0}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{L^\infty}
\le c\epsilon\left\|(\Phi^{(a)})^{-1}\varphi_0^{\Phi^{(a)}}-(\Phi^{(b)})^{-1}\varphi_0^{\Phi^{(b)}}\right\|_{L^\infty} \\
&\le c(\kappa_*^{(a)},\kappa_*^{(b)})\epsilon\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|\varphi\|_{L^\infty}.
\end{aligned} ∂ s 2 R S , 0 ( a ) [ φ 0 Φ ( a ) ] − ∂ s 2 R S , 0 ( b ) [ φ 0 Φ ( b ) ] L ∞ ≤ cϵ ( Φ ( a ) ) − 1 φ 0 Φ ( a ) − ( Φ ( b ) ) − 1 φ 0 Φ ( b ) L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ .
In summary, the remainder term R S , 0 \mathcal{R}_{S,0} R S , 0 satisfies the estimates
∥ R S , 0 [ φ ] ∥ C 2 ≤ c ϵ ∥ φ ∥ L ∞ \|\mathcal{R}_{S,0}[\varphi]\|_{C^2}\le c\epsilon\|\varphi\|_{L^\infty} ∥ R S , 0 [ φ ] ∥ C 2 ≤ cϵ ∥ φ ∥ L ∞
∥ R S , 0 ( a ) [ φ 0 Φ ( a ) ] − R S , 0 ( b ) [ φ 0 Φ ( b ) ] ∥ C 2 ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ . (225) \left\|\mathcal{R}_{S,0}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,0}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{C^2}
\le c(\kappa_*^{(a)},\kappa_*^{(b)})\epsilon\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|\varphi\|_{L^\infty}. \tag*{(225)} R S , 0 ( a ) [ φ 0 Φ ( a ) ] − R S , 0 ( b ) [ φ 0 Φ ( b ) ] C 2 ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ . (225)
We now turn to the remainders R S , 1 \mathcal{R}_{S,1} R S , 1 and R S , 2 \mathcal{R}_{S,2} R S , 2 , which both contain some components which are small in ϵ \epsilon ϵ and some components which are more regular. To separate these components, we define an auxiliary function
R t ( s , s ^ , θ , θ ^ ) = s ^ e t ( s ) + ϵ ( e r ( s , θ ) − e r ( s − s ^ , θ − θ ^ ) ) , (226) R_t(s,\hat{s},\theta,\hat{\theta})=\hat{s}e_t(s)+\epsilon(e_r(s,\theta)-e_r(s-\hat{s},\theta-\hat{\theta})), \tag*{(226)} R t ( s , s ^ , θ , θ ^ ) = s ^ e t ( s ) + ϵ ( e r ( s , θ ) − e r ( s − s ^ , θ − θ ^ )) , (226)
as in [59 Section 3.2] , and note that
∣ R t ∣ 2 = ∣ R ‾ ∣ 2 + ϵ s ^ 2 Q S , 1 + ϵ ^ 2 s ^ sin ( θ ^ 2 ) Q S , 2 ∣ R ∣ 2 = ∣ R t ∣ 2 + s ^ 3 Q S , 3 + ϵ s ^ 2 sin ( θ ^ 2 ) Q S , 4 , (227) \begin{aligned}
|R_t|^2 &= |\overline{R}|^2+\epsilon\widehat{s}^2Q_{S,1}+\widehat{\epsilon}^2\widehat{s}\sin\left(\frac{\widehat{\theta}}{2}\right)Q_{S,2}\\
|R|^2 &= |R_t|^2+\widehat{s}^3Q_{S,3}+\epsilon\widehat{s}^2\sin\left(\frac{\widehat{\theta}}{2}\right)Q_{S,4}, \tag*{(227)}
\end{aligned} ∣ R t ∣ 2 ∣ R ∣ 2 = ∣ R ∣ 2 + ϵ s 2 Q S , 1 + ϵ 2 s sin ( 2 θ ) Q S , 2 = ∣ R t ∣ 2 + s 3 Q S , 3 + ϵ s 2 sin ( 2 θ ) Q S , 4 , (227)
where the terms Q S , j ( s , θ , s ^ , θ ^ ) Q_{S,j}(s,\theta,\widehat{s},\widehat{\theta}) Q S , j ( s , θ , s , θ ) are given by
Q S , 1 = − 2 κ ^ ( s , θ ) + ϵ κ ^ ( s , θ ) 2 + ϵ κ 3 2 − 4 sin ( θ ^ 2 ) e t ( s ) ⋅ Q θ ( s , s ^ , θ − θ ^ / 2 ) + 2 s ^ Q 0 , 1 ( s , s ^ , θ ) + ϵ s ^ Q 0 , 4 ( s , s ^ , θ ) Q S , 2 = 4 κ 3 cos ( θ ^ 2 ) + 4 s ^ Q 0 , 3 ( s , θ , s ^ , θ ^ ) − 4 s ^ Q θ ⋅ Q r ( s , s ^ , θ ) + 4 s ^ sin ( θ ^ 2 ) ∣ Q θ ∣ 2 Q S , 3 = 2 e t ⋅ Q t + 2 ϵ Q t ( s , s ^ ) ⋅ Q r ( s , s ^ , θ − θ ^ ) + s ^ ∣ Q t ∣ 2 Q S , 4 = − 4 Q t ( s , s ^ ) ⋅ e θ ( s , θ − θ ^ / 2 ) \begin{aligned}
Q_{S,1}={}&-2\widehat{\kappa}(s,\theta)+\epsilon\widehat{\kappa}(s,\theta)^2+\epsilon\kappa_3^2-4\sin\left(\frac{\widehat{\theta}}{2}\right)e_t(s)\cdot Q_\theta(s,\widehat{s},\theta-\widehat{\theta}/2)\\
&+2\widehat{s}Q_{0,1}(s,\widehat{s},\theta)+\epsilon\widehat{s}Q_{0,4}(s,\widehat{s},\theta)\\
Q_{S,2}={}&4\kappa_3\cos\left(\frac{\widehat{\theta}}{2}\right)+4\widehat{s}Q_{0,3}(s,\theta,\widehat{s},\widehat{\theta})-4\widehat{s}Q_\theta\cdot Q_r(s,\widehat{s},\theta)+4\widehat{s}\sin\left(\frac{\widehat{\theta}}{2}\right)|Q_\theta|^2\\
Q_{S,3}={}&2e_t\cdot Q_t+2\epsilon Q_t(s,\widehat{s})\cdot Q_r(s,\widehat{s},\theta-\widehat{\theta})+\widehat{s}|Q_t|^2\\
Q_{S,4}={}&-4Q_t(s,\widehat{s})\cdot e_\theta(s,\theta-\widehat{\theta}/2)
\end{aligned} Q S , 1 = Q S , 2 = Q S , 3 = Q S , 4 = − 2 κ ( s , θ ) + ϵ κ ( s , θ ) 2 + ϵ κ 3 2 − 4 sin ( 2 θ ) e t ( s ) ⋅ Q θ ( s , s , θ − θ /2 ) + 2 s Q 0 , 1 ( s , s , θ ) + ϵ s Q 0 , 4 ( s , s , θ ) 4 κ 3 cos ( 2 θ ) + 4 s Q 0 , 3 ( s , θ , s , θ ) − 4 s Q θ ⋅ Q r ( s , s , θ ) + 4 s sin ( 2 θ ) ∣ Q θ ∣ 2 2 e t ⋅ Q t + 2 ϵ Q t ( s , s ) ⋅ Q r ( s , s , θ − θ ) + s ∣ Q t ∣ 2 − 4 Q t ( s , s ) ⋅ e θ ( s , θ − θ /2 )
for Q 0 , j Q_{0,j} Q 0 , j as in (203) . In addition, each Q S , j Q_{S,j} Q S , j satisfies an estimate of the form (214) . We emphasize that the difference between R t R_t R t and R R R is higher order in s ^ \widehat{s} s and the difference between R t R_t R t and R ‾ \overline{R} R is higher order in ϵ \epsilon ϵ . Furthermore, R t R_t R t may be shown to satisfy the same estimates as R R R in each of Lemmas 4.1, 4.2, 4.3, and 4.4 (see [59 ] ).
For integers k ≥ 1 k\geq1 k ≥ 1 , we additionally note the following expansions:
1 ∣ R ∣ k − 1 ∣ R t ∣ k = s ^ 3 Q S , 3 + ϵ s ^ 2 sin ( θ ^ 2 ) Q S , 4 ∣ R ∣ ∣ R t ∣ ( ∣ R ∣ + ∣ R t ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R ∣ ℓ ∣ R t ∣ k − 1 − ℓ 1 ∣ R t ∣ k − 1 ∣ R ‾ ∣ k = ϵ s ^ 2 Q S , 1 + ϵ 2 s ^ sin ( θ ^ 2 ) Q S , 2 ∣ R t ∣ ∣ R ‾ ∣ ( ∣ R t ∣ + ∣ R ‾ ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R t ∣ ℓ ∣ R ‾ ∣ k − 1 − ℓ 1 ∣ R t ( a ) ∣ k − 1 ∣ R t ( b ) ∣ k = ϵ s ^ 2 ( Q S , 1 ( a ) − Q S , 1 ( b ) ) + ϵ 2 s ^ sin ( θ ^ 2 ) ( Q S , 2 ( a ) − Q S , 2 ( b ) ) ∣ R t ( a ) ∣ ∣ R t ( b ) ∣ ( ∣ R t ( a ) ∣ + ∣ R t ( b ) ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R t ( a ) ∣ ℓ ∣ R t ( b ) ∣ k − 1 − ℓ . (228) \begin{aligned}
\frac{1}{|R|^k}-\frac{1}{|R_t|^k}
&=\frac{\widehat{s}^3Q_{S,3}+\epsilon\widehat{s}^2\sin\left(\frac{\widehat{\theta}}{2}\right)Q_{S,4}}{|R||R_t|(|R|+|R_t|)}
\sum_{\ell=0}^{k-1}\frac{1}{|R|^\ell|R_t|^{k-1-\ell}}\\
\frac{1}{|R_t|^k}-\frac{1}{|\overline{R}|^k}
&=\frac{\epsilon\widehat{s}^2Q_{S,1}+\epsilon^2\widehat{s}\sin\left(\frac{\widehat{\theta}}{2}\right)Q_{S,2}}{|R_t||\overline{R}|(|R_t|+|\overline{R}|)}
\sum_{\ell=0}^{k-1}\frac{1}{|R_t|^\ell|\overline{R}|^{k-1-\ell}}\\
\frac{1}{|R_t^{(a)}|^k}-\frac{1}{|R_t^{(b)}|^k}
&=\frac{\epsilon\widehat{s}^2(Q_{S,1}^{(a)}-Q_{S,1}^{(b)})+\epsilon^2\widehat{s}\sin\left(\frac{\widehat{\theta}}{2}\right)(Q_{S,2}^{(a)}-Q_{S,2}^{(b)})}{|R_t^{(a)}||R_t^{(b)}|(|R_t^{(a)}|+|R_t^{(b)}|)}
\sum_{\ell=0}^{k-1}\frac{1}{|R_t^{(a)}|^\ell|R_t^{(b)}|^{k-1-\ell}}. \tag*{(228)}
\end{aligned} ∣ R ∣ k 1 − ∣ R t ∣ k 1 ∣ R t ∣ k 1 − ∣ R ∣ k 1 ∣ R t ( a ) ∣ k 1 − ∣ R t ( b ) ∣ k 1 = ∣ R ∣∣ R t ∣ ( ∣ R ∣ + ∣ R t ∣ ) s 3 Q S , 3 + ϵ s 2 sin ( 2 θ ) Q S , 4 ℓ = 0 ∑ k − 1 ∣ R ∣ ℓ ∣ R t ∣ k − 1 − ℓ 1 = ∣ R t ∣∣ R ∣ ( ∣ R t ∣ + ∣ R ∣ ) ϵ s 2 Q S , 1 + ϵ 2 s sin ( 2 θ ) Q S , 2 ℓ = 0 ∑ k − 1 ∣ R t ∣ ℓ ∣ R ∣ k − 1 − ℓ 1 = ∣ R t ( a ) ∣∣ R t ( b ) ∣ ( ∣ R t ( a ) ∣ + ∣ R t ( b ) ∣ ) ϵ s 2 ( Q S , 1 ( a ) − Q S , 1 ( b ) ) + ϵ 2 s sin ( 2 θ ) ( Q S , 2 ( a ) − Q S , 2 ( b ) ) ℓ = 0 ∑ k − 1 ∣ R t ( a ) ∣ ℓ ∣ R t ( b ) ∣ k − 1 − ℓ 1 . (228)
Here, as usual, the superscripts ( a ) (a) ( a ) and ( b ) (b) ( b ) are used to denote quantities corresponding to two different but nearby filament centerlines X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) , respectively, satisfying Lemma 1.1 .
Using R t R_t R t , we further subdivide R S , 1 R_{S,1} R S , 1 as
R S , 1 [ ϕ ] = 1 8 π ( R S , 1 , 1 [ ϕ ] + R S , 1 , 2 [ ϕ ] ) , R_{S,1}[\phi]=\frac{1}{8\pi}\left(R_{S,1,1}[\phi]+R_{S,1,2}[\phi]\right), R S , 1 [ ϕ ] = 8 π 1 ( R S , 1 , 1 [ ϕ ] + R S , 1 , 2 [ ϕ ] ) ,
R S , 1 , 1 [ ϕ ] = Φ − 1 ∫ − 1 / 2 1 / 2 ∫ − π π ( 1 ∣ R ‾ ∣ − 1 ∣ R t ∣ ) ϕ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ , R_{S,1,1}[\phi]=\Phi^{-1}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{1}{|\overline{R}|}-\frac{1}{|R_t|}\right)\phi(s-\widehat{s},\theta-\widehat{\theta})\,\epsilon\,d\widehat{\theta}\,d\widehat{s}, R S , 1 , 1 [ ϕ ] = Φ − 1 ∫ − 1/2 1/2 ∫ − π π ( ∣ R ∣ 1 − ∣ R t ∣ 1 ) ϕ ( s − s , θ − θ ) ϵ d θ d s ,
e ‾ j ⋅ R S , 1 , 2 [ ϕ ] = ∫ − 1 / 2 1 / 2 ∫ − π π ( 1 ∣ R t ∣ − 1 ∣ R ‾ ∣ ) e ‾ j ⋅ ϕ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ . \overline{e}_j\cdot R_{S,1,2}[\phi]=\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{1}{|R_t|}-\frac{1}{|\overline{R}|}\right)\overline{e}_j\cdot\phi(s-\widehat{s},\theta-\widehat{\theta})\,\epsilon\,d\widehat{\theta}\,d\widehat{s}. e j ⋅ R S , 1 , 2 [ ϕ ] = ∫ − 1/2 1/2 ∫ − π π ( ∣ R t ∣ 1 − ∣ R ∣ 1 ) e j ⋅ ϕ ( s − s , θ − θ ) ϵ d θ d s .
Here we take the subscript j ∈ { t , r , θ } j\in\{t,r,\theta\} j ∈ { t , r , θ } and use e ‾ j \overline{e}_j e j to denote the corresponding basis vector with respect to the straight frame about C ϵ C_\epsilon C ϵ . Note that in this notation, e ‾ t = e z \overline{e}_t=e_z e t = e z . We additionally subdivide R S , 2 R_{S,2} R S , 2 as
R S , 2 [ ϕ ] = 1 8 π ( R S , 2 , 1 [ ϕ ] + R S , 2 , 2 [ ϕ ] ) , R_{S,2}[\phi]=\frac{1}{8\pi}\left(R_{S,2,1}[\phi]+R_{S,2,2}[\phi]\right), R S , 2 [ ϕ ] = 8 π 1 ( R S , 2 , 1 [ ϕ ] + R S , 2 , 2 [ ϕ ] ) ,
e ‾ j ⋅ R S , 2 , 1 [ ϕ ] = ∫ − 1 / 2 1 / 2 ∫ − π π ( ( e ‾ j ⋅ R ) ( R ⋅ ϕ ) ∣ R ∣ 3 − ( e ‾ j ⋅ R t ) ( R t ⋅ ϕ ) ∣ R t ∣ 3 ) ϵ d θ ^ d s ^ \overline{e}_j\cdot R_{S,2,1}[\phi]=\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{(\overline{e}_j\cdot R)(R\cdot\phi)}{|R|^3}-\frac{(\overline{e}_j\cdot R_t)(R_t\cdot\phi)}{|R_t|^3}\right)\epsilon\,d\widehat{\theta}\,d\widehat{s} e j ⋅ R S , 2 , 1 [ ϕ ] = ∫ − 1/2 1/2 ∫ − π π ( ∣ R ∣ 3 ( e j ⋅ R ) ( R ⋅ ϕ ) − ∣ R t ∣ 3 ( e j ⋅ R t ) ( R t ⋅ ϕ ) ) ϵ d θ d s
e ‾ j ⋅ R S , 2 , 2 [ φ ] = ∫ − 1 / 2 1 / 2 ∫ − π π ( ( e ‾ j ⋅ R t ) ( R t ⋅ φ ) ∣ R t ∣ 3 − ( e ‾ j ⋅ R ‾ ) ( R ‾ ⋅ ( Φ − 1 φ ) ) ∣ R ‾ ∣ 3 ) ϵ d θ ^ d s ^ . \overline{e}_j \cdot\mathcal{R}_{S,2,2}[\varphi] = \int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{(\overline{e}_j\cdot\mathbf{R}_t)(\mathbf{R}_t\cdot\varphi)}{|\mathbf{R}_t|^3}-\frac{(\overline{e}_j\cdot\overline{\mathbf{R}})(\overline{\mathbf{R}}\cdot(\Phi^{-1}\varphi))}{|\overline{\mathbf{R}}|^3}\right)\epsilon\,d\widehat{\theta}d\widehat{s}. e j ⋅ R S , 2 , 2 [ φ ] = ∫ − 1/2 1/2 ∫ − π π ( ∣ R t ∣ 3 ( e j ⋅ R t ) ( R t ⋅ φ ) − ∣ R ∣ 3 ( e j ⋅ R ) ( R ⋅ ( Φ − 1 φ )) ) ϵ d θ d s .
Using the above expressions along with the forms of R ‾ \overline{\mathbf{R}} R (196), R \mathbf{R} R (205), R t \mathbf{R}_t R t (225), and the differences (209), (210), and (227), by Lemma 4.2 we have
∣ R S , 1 , 1 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π ϵ d θ ^ d s ^ ≤ c ( κ ∗ ) ϵ ∥ φ ∥ L ∞ |\mathcal{R}_{S,1,1}[\varphi]|\le c(\kappa_*)\|\varphi\|_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\epsilon\,d\widehat{\theta}d\widehat{s}\le c(\kappa_*)\epsilon\|\varphi\|_{L^\infty} ∣ R S , 1 , 1 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ϵ d θ d s ≤ c ( κ ∗ ) ϵ ∥ φ ∥ L ∞
∣ R S , 1 , 2 [ φ ] ∣ ≤ c ( κ ∗ , c Γ ) ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π ϵ ∣ R ‾ ∣ d θ ^ d s ^ ≤ c ( κ ∗ , c Γ ) ϵ 2 ∥ φ ∥ L ∞ |\mathcal{R}_{S,1,2}[\varphi]|\le c(\kappa_*,c_\Gamma)\|\varphi\|_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\epsilon}{|\overline{\mathbf{R}}|}\,d\widehat{\theta}d\widehat{s}\le c(\kappa_*,c_\Gamma)\epsilon^2\|\varphi\|_{L^\infty} ∣ R S , 1 , 2 [ φ ] ∣ ≤ c ( κ ∗ , c Γ ) ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ ϵ d θ d s ≤ c ( κ ∗ , c Γ ) ϵ 2 ∥ φ ∥ L ∞
∣ R S , 2 , 1 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π ϵ d θ ^ d s ^ ≤ c ( κ ∗ ) ϵ ∥ φ ∥ L ∞ (229) |\mathcal{R}_{S,2,1}[\varphi]|\le c(\kappa_*)\|\varphi\|_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\epsilon\,d\widehat{\theta}d\widehat{s}\le c(\kappa_*)\epsilon\|\varphi\|_{L^\infty} \tag*{(229)} ∣ R S , 2 , 1 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ϵ d θ d s ≤ c ( κ ∗ ) ϵ ∥ φ ∥ L ∞ (229)
∣ R S , 2 , 2 [ φ ] ∣ ≤ c ( κ ∗ , c Γ ) ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π ϵ ∣ R ‾ ∣ d θ ^ d s ^ ≤ c ( κ ∗ , c Γ ) ϵ 2 ∥ φ ∥ L ∞ |\mathcal{R}_{S,2,2}[\varphi]|\le c(\kappa_*,c_\Gamma)\|\varphi\|_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\epsilon}{|\overline{\mathbf{R}}|}\,d\widehat{\theta}d\widehat{s}\le c(\kappa_*,c_\Gamma)\epsilon^2\|\varphi\|_{L^\infty} ∣ R S , 2 , 2 [ φ ] ∣ ≤ c ( κ ∗ , c Γ ) ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ ϵ d θ d s ≤ c ( κ ∗ , c Γ ) ϵ 2 ∥ φ ∥ L ∞
∣ R S , 3 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π 1 ∣ R ‾ ∣ ϵ 2 d θ ^ d s ^ ≤ c ( κ ∗ , c Γ ) ϵ 2 ∥ φ ∥ L ∞ . |\mathcal{R}_{S,3}[\varphi]|\le c(\kappa_*)\|\varphi\|_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{|\overline{\mathbf{R}}|}\epsilon^2\,d\widehat{\theta}d\widehat{s}\le c(\kappa_*,c_\Gamma)\epsilon^2\|\varphi\|_{L^\infty}. ∣ R S , 3 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 1 ϵ 2 d θ d s ≤ c ( κ ∗ , c Γ ) ϵ 2 ∥ φ ∥ L ∞ .
In addition, using case (1) of Lemma 4.4, we may estimate
∣ R S , 1 , 1 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ |\mathcal{R}_{S,1,1}[\varphi]|_{\dot{C}^{0,\alpha}}\le c(\kappa_*,\alpha,c_\Gamma)\epsilon^{1-\alpha}\|\varphi\|_{L^\infty} ∣ R S , 1 , 1 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞
∣ R S , 1 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ |\mathcal{R}_{S,1,2}[\varphi]|_{\dot{C}^{0,\alpha}}\le c(\kappa_*,\alpha,c_\Gamma)\epsilon^{2-\alpha}\|\varphi\|_{L^\infty} ∣ R S , 1 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞
∣ R S , 2 , 1 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ |\mathcal{R}_{S,2,1}[\varphi]|_{\dot{C}^{0,\alpha}}\le c(\kappa_*,\alpha,c_\Gamma)\epsilon^{1-\alpha}\|\varphi\|_{L^\infty} ∣ R S , 2 , 1 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞
∣ R S , 2 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ |\mathcal{R}_{S,2,2}[\varphi]|_{\dot{C}^{0,\alpha}}\le c(\kappa_*,\alpha,c_\Gamma)\epsilon^{2-\alpha}\|\varphi\|_{L^\infty} ∣ R S , 2 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞
∣ R S , 3 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ . (230) |\mathcal{R}_{S,3}[\varphi]|_{\dot{C}^{0,\alpha}}\le c(\kappa_*,\alpha,c_\Gamma)\epsilon^{2-\alpha}\|\varphi\|_{L^\infty}. \tag*{(230)} ∣ R S , 3 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ φ ∥ L ∞ . (230)
Using that the remainder coefficients in the expansions (205), (226), (226), and (227) each satisfy a Lipschitz bound of the form (213), and using the bounds (81), given two curves X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) satisfying Lemma 1.1, the forms of the above kernels also yield
∥ R S , 1 , 1 ( a ) [ φ 0 Φ ( a ) ] − R S , 1 , 1 ( b ) [ φ 0 Φ ( b ) ] ∥ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ \left\|\mathcal{R}_{S,1,1}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,1,1}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{\dot{C}^{0,\alpha}}\le c(\kappa_*^{(a)},\kappa_*^{(b)},\alpha,c_\Gamma)\epsilon^{1-\alpha}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty} R S , 1 , 1 ( a ) [ φ 0 Φ ( a ) ] − R S , 1 , 1 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 1 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞
∥ R S , 1 , 2 ( a ) [ φ 0 Φ ( a ) ] − R S , 1 , 2 ( b ) [ φ 0 Φ ( b ) ] ∥ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ \left\|\mathcal{R}_{S,1,2}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,1,2}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{\dot{C}^{0,\alpha}}\le c(\kappa_*^{(a)},\kappa_*^{(b)},\alpha,c_\Gamma)\epsilon^{2-\alpha}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty} R S , 1 , 2 ( a ) [ φ 0 Φ ( a ) ] − R S , 1 , 2 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞
∥ R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] ∥ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ \left\|\mathcal{R}_{S,2,1}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,2,1}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{\dot{C}^{0,\alpha}}\le c(\kappa_*^{(a)},\kappa_*^{(b)},\alpha,c_\Gamma)\epsilon^{1-\alpha}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty} R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 1 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞
∥ R S , 2 , 2 ( a ) [ φ 0 Φ ( a ) ] − R S , 2 , 2 ( b ) [ φ 0 Φ ( b ) ] ∥ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ \left\|\mathcal{R}_{S,2,2}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,2,2}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{\dot{C}^{0,\alpha}}\le c(\kappa_*^{(a)},\kappa_*^{(b)},\alpha,c_\Gamma)\epsilon^{2-\alpha}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty} R S , 2 , 2 ( a ) [ φ 0 Φ ( a ) ] − R S , 2 , 2 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞
∥ R S , 3 ( a ) [ φ 0 Φ ( a ) ] − R S , 3 ( b ) [ φ 0 Φ ( b ) ] ∥ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ L ∞ . (231) \left\|\mathcal{R}_{S,3}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\mathcal{R}_{S,3}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\|_{\dot{C}^{0,\alpha}}\le c(\kappa_*^{(a)},\kappa_*^{(b)},\alpha,c_\Gamma)\epsilon^{2-\alpha}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha}}\|\varphi\|_{L^\infty}. \tag*{(231)} R S , 3 ( a ) [ φ 0 Φ ( a ) ] − R S , 3 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ L ∞ . (231)
We now turn to bounds for the derivatives of R S , i \mathcal{R}_{S,i} R S , i , i = 1 , 2 , 3 i=1,2,3 i = 1 , 2 , 3 , along the surface Γ ϵ \Gamma_\epsilon Γ ϵ . We begin with s s s -derivatives and note the expansions
∂ s R = ( 1 − ϵ κ ^ b ( s , θ ) ) e t ( s ) + ϵ κ 3 e θ ( s , θ ) \partial_s\mathbf{R}=(1-\epsilon\widehat{\kappa}b(s,\theta))\mathbf{e}_t(s)+\epsilon\kappa_3\mathbf{e}_\theta(s,\theta) ∂ s R = ( 1 − ϵ κ b ( s , θ )) e t ( s ) + ϵ κ 3 e θ ( s , θ )
∂ s R t = ( 1 − ϵ κ ^ b ( s , θ ) ) e t ( s ) + s ^ ( κ 1 e n 1 ( s ) + κ 2 e n 2 ( s ) ) + ϵ κ 3 e θ ( s , θ ) \partial_s\mathbf{R}_t=(1-\epsilon\widehat{\kappa}b(s,\theta))\mathbf{e}_t(s)+\widehat{s}(\kappa_1\mathbf{e}_{n_1}(s)+\kappa_2\mathbf{e}_{n_2}(s))+\epsilon\kappa_3\mathbf{e}_\theta(s,\theta) ∂ s R t = ( 1 − ϵ κ b ( s , θ )) e t ( s ) + s ( κ 1 e n 1 ( s ) + κ 2 e n 2 ( s )) + ϵ κ 3 e θ ( s , θ )
∂ s R ⋅ R = s ^ + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) + ϵ s ^ Q R s 1 + s ^ 2 Q R s 2 \partial_s\mathbf{R}\cdot\mathbf{R}=\widehat{s}+2\epsilon^2\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})+\epsilon\widehat{s}Q_{\mathbf{R}s1}+\widehat{s}^{\,2}Q_{\mathbf{R}s2} ∂ s R ⋅ R = s + 2 ϵ 2 sin ( 2 θ ) sin ( θ ) + ϵ s Q R s 1 + s 2 Q R s 2
∂ s R ⋅ n ′ = ∂ s R ⋅ e r ( s − s ^ , θ − θ ^ ) = s ^ κ ^ b ( s , θ ) − κ 3 ϵ sin ( θ ^ ) + ϵ s ^ Q R s 3 \partial_s\mathbf{R}\cdot\mathbf{n}'=\partial_s\mathbf{R}\cdot\mathbf{e}_r(s-\widehat{s},\theta-\widehat{\theta})=\widehat{s}\widehat{\kappa}b(s,\theta)-\kappa_3\epsilon\sin(\widehat{\theta})+\epsilon\widehat{s}Q_{\mathbf{R}s3} ∂ s R ⋅ n ′ = ∂ s R ⋅ e r ( s − s , θ − θ ) = s κ b ( s , θ ) − κ 3 ϵ sin ( θ ) + ϵ s Q R s 3
∂ s R ⋅ n = ∂ s R ⋅ e r ( s , θ ) = 0 , ∂ s R ⋅ e t ( s ) = ( 1 − ϵ κ ^ b ( s , θ ) ) , ∂ s R ⋅ e θ ( s , θ ) = ϵ κ 3 . (232) \partial_s\mathbf{R}\cdot\mathbf{n}=\partial_s\mathbf{R}\cdot\mathbf{e}_r(s,\theta)=0,\qquad
\partial_s\mathbf{R}\cdot\mathbf{e}_t(s)=(1-\epsilon\widehat{\kappa}b(s,\theta)),\qquad
\partial_s\mathbf{R}\cdot\mathbf{e}_\theta(s,\theta)=\epsilon\kappa_3. \tag*{(232)} ∂ s R ⋅ n = ∂ s R ⋅ e r ( s , θ ) = 0 , ∂ s R ⋅ e t ( s ) = ( 1 − ϵ κ b ( s , θ )) , ∂ s R ⋅ e θ ( s , θ ) = ϵ κ 3 . (232)
Again the remainders Q R S j Q_{R_{Sj}} Q R S j each satisfy estimates of the form (214) . We may then write
∂ s R t ⋅ R t = ∂ s R ‾ ⋅ R ‾ + ϵ s ^ Q S , 5 ( s , s ^ , θ ^ , θ ^ ) − ϵ 2 κ 3 sin ( θ ^ ) = ∂ s R ‾ ⋅ R ‾ + s ^ 2 Q S , 6 ( s , s ^ , θ ^ , θ ^ ) , (233) \partial_s\mathbf{R}_t\cdot\mathbf{R}_t=\partial_s\overline{\mathbf{R}}\cdot\overline{\mathbf{R}}+\epsilon\hat{s}Q_{S,5}(s,\hat{s},\hat{\theta},\hat{\theta})-\epsilon^2\kappa_3\sin(\hat{\theta})
=\partial_s\overline{\mathbf{R}}\cdot\overline{\mathbf{R}}+\hat{s}^2Q_{S,6}(s,\hat{s},\hat{\theta},\hat{\theta}), \tag*{(233)} ∂ s R t ⋅ R t = ∂ s R ⋅ R + ϵ s ^ Q S , 5 ( s , s ^ , θ ^ , θ ^ ) − ϵ 2 κ 3 sin ( θ ^ ) = ∂ s R ⋅ R + s ^ 2 Q S , 6 ( s , s ^ , θ ^ , θ ^ ) , (233)
where Q S , 5 Q_{S,5} Q S , 5 and Q S , 6 Q_{S,6} Q S , 6 also satisfy (214) .
We begin by estimating ∂ s R S , 1 \partial_s\mathcal{R}_{S,1} ∂ s R S , 1 , which we again consider in two pieces as ∂ s R S , 1 , 1 + ∂ s R S , 1 , 2 \partial_s\mathcal{R}_{S,1,1}+\partial_s\mathcal{R}_{S,1,2} ∂ s R S , 1 , 1 + ∂ s R S , 1 , 2 . Note that we will treat the commutator term [ ∂ s , Φ − 1 ] [\partial_s,\Phi^{-1}] [ ∂ s , Φ − 1 ] as part of the smoother remainder ∂ s R S , 1 , 1 \partial_s\mathcal{R}_{S,1,1} ∂ s R S , 1 , 1 . For each j = t , r , θ j=t,r,\theta j = t , r , θ , we may write
e ˉ j ⋅ ∂ s R S , 1 , 1 = p.v. ∫ − 1 / 2 1 / 2 ∫ − π π S 1 , 1 ϵ d θ ^ d s ^ + e ˉ j 8 π [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π 1 ∣ R ‾ ∣ φ ϵ d θ ^ d s ^ , \bar{e}_j\cdot\partial_s\mathcal{R}_{S,1,1}=\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}S_{1,1}\,\epsilon\,d\hat{\theta}d\hat{s}+\frac{\bar{e}_j}{8\pi}[\partial_s,\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{\lvert\overline{\mathbf{R}}\rvert}\varphi\,\epsilon\,d\hat{\theta}d\hat{s}, e ˉ j ⋅ ∂ s R S , 1 , 1 = p.v. ∫ − 1/2 1/2 ∫ − π π S 1 , 1 ϵ d θ ^ d s ^ + 8 π e ˉ j [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 1 φ ϵ d θ ^ d s ^ ,
S 1 , 1 = ( ∂ s R t ⋅ R t − ∂ s R ‾ ⋅ R ‾ ∣ R ∣ 3 + ( 1 ∣ R t ∣ 3 − 1 ∣ R ‾ ∣ 3 ) ( ∂ s R t ⋅ R t ) ) e ˉ j ⋅ φ = ( s ^ 2 Q S , 6 ∣ R ∣ 3 + ( 1 ∣ R t ∣ 3 − 1 ∣ R ‾ ∣ 3 ) ( s ^ + ϵ s ^ Q S , 5 − ϵ 2 κ 3 sin ( θ ^ ) ) ) e ˉ j ⋅ φ . \begin{aligned}
S_{1,1}=\left(\frac{\partial_s\mathbf{R}_t\cdot\mathbf{R}_t-\partial_s\overline{\mathbf{R}}\cdot\overline{\mathbf{R}}}{\lvert\mathbf{R}\rvert^3}+\left(\frac{1}{\lvert\mathbf{R}_t\rvert^3}-\frac{1}{\lvert\overline{\mathbf{R}}\rvert^3}\right)(\partial_s\mathbf{R}_t\cdot\mathbf{R}_t)\right)\bar{e}_j\cdot\varphi\\
&=\left(\frac{\hat{s}^2Q_{S,6}}{\lvert\mathbf{R}\rvert^3}+\left(\frac{1}{\lvert\mathbf{R}_t\rvert^3}-\frac{1}{\lvert\overline{\mathbf{R}}\rvert^3}\right)(\hat{s}+\epsilon\hat{s}Q_{S,5}-\epsilon^2\kappa_3\sin(\hat{\theta}))\right)\bar{e}_j\cdot\varphi.
\end{aligned} S 1 , 1 = ( ∣ R ∣ 3 ∂ s R t ⋅ R t − ∂ s R ⋅ R + ( ∣ R t ∣ 3 1 − ∣ R ∣ 3 1 ) ( ∂ s R t ⋅ R t ) ) e ˉ j ⋅ φ = ( ∣ R ∣ 3 s ^ 2 Q S , 6 + ( ∣ R t ∣ 3 1 − ∣ R ∣ 3 1 ) ( s ^ + ϵ s ^ Q S , 5 − ϵ 2 κ 3 sin ( θ ^ )) ) e ˉ j ⋅ φ .
Using (228) and Lemma 4.2 along with the commutator estimate (78) , we have
∣ ∂ s R S , 1 , 1 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π 1 ∣ R ‾ ∣ ϵ d θ ^ d s ^ ≤ c ( κ ∗ , c Γ ) ϵ ∥ φ ∥ L ∞ . \left\lvert\partial_s\mathcal{R}_{S,1,1}[\varphi]\right\rvert\leq c(\kappa_*)\lVert\varphi\rVert_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{\lvert\overline{\mathbf{R}}\rvert}\epsilon\,d\hat{\theta}d\hat{s}\leq c(\kappa_*,c_\Gamma)\epsilon\lVert\varphi\rVert_{L^\infty}. ∣ ∂ s R S , 1 , 1 [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 1 ϵ d θ ^ d s ^ ≤ c ( κ ∗ , c Γ ) ϵ ∥ φ ∥ L ∞ .
In addition, for α < β < 1 \alpha<\beta<1 α < β < 1 , by case (1) of Lemma 4.4 we have
∣ ∂ s R S , 1 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( max i ∥ Q i ∥ C ℓ 0 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ . (234) \left\lvert\partial_s\mathcal{R}_{S,1,1}[\varphi]\right\rvert_{\dot{C}^{0,\beta}}\leq c(\kappa_*,c_\Gamma)\epsilon^{1-\beta}\left(\max_i\lVert Q_i\rVert_{C^{0,\beta}_\ell}\right)\lVert\varphi\rVert_{L^\infty}\leq c(\kappa_*,\beta,c_\Gamma)\epsilon^{1-2\beta}\lVert\varphi\rVert_{L^\infty}. \tag*{(234)} ∣ ∂ s R S , 1 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( i max ∥ Q i ∥ C ℓ 0 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ . (234)
Here the Q i Q_i Q i include the remainder coefficients Q S , 3 Q_{S,3} Q S , 3 , Q S , 4 Q_{S,4} Q S , 4 , Q S , 5 Q_{S,5} Q S , 5 , and Q S , 6 Q_{S,6} Q S , 6 . Furthermore, given two filaments with centerlines X ( a ) \mathbf{X}^{(a)} X ( a ) , X ( b ) \mathbf{X}^{(b)} X ( b ) as in Lemma 1.1, we have
S 1 , 1 ( a ) − S 1 , 1 ( b ) = ( s ^ 2 ( Q S , 6 ( a ) − Q S , 6 ( b ) ) ∣ R ( a ) ∣ 3 + ( 1 ∣ R ( a ) ∣ 3 − 1 ∣ R ( b ) ∣ 3 ) s ^ 2 Q S , 6 ( b ) ) + ( 1 ∣ R t ( a ) ∣ 3 − 1 ∣ R ( a ) ∣ 3 ) ( ϵ s ^ ( Q S , 5 ( a ) − Q S , 5 ( b ) ) − ϵ 2 ( κ 3 ( a ) − κ 3 ( b ) ) sin ( θ ^ ) ) + ( 1 ∣ R t ( a ) ∣ 3 − 1 ∣ R ( a ) ∣ 3 − 1 ∣ R t ( b ) ∣ 3 + 1 ∣ R ( b ) ∣ 3 ) ( s ^ + ϵ s ^ Q S , 5 ( b ) − ϵ 2 κ 3 ( b ) sin ( θ ^ ) ) e ˉ j ( a ) ⋅ φ 0 Φ ( a ) + ( s ^ 2 Q S , 6 ( b ) ∣ R ( b ) ∣ 3 + ( 1 ∣ R t ( b ) ∣ 3 − 1 ∣ R ( b ) ∣ 3 ) ( s ^ + ϵ s ^ Q S , 5 ( b ) − ϵ 2 κ 3 ( b ) sin ( θ ^ ) ) ) ( e ˉ j ( a ) ⋅ φ 0 Φ ( a ) − e ˉ j ( b ) ⋅ φ 0 Φ ( b ) ) . \begin{aligned}
S_{1,1}^{(a)}-S_{1,1}^{(b)}
={}&\left(\frac{\hat{s}^2(Q_{S,6}^{(a)}-Q_{S,6}^{(b)})}{\lvert\mathbf{R}^{(a)}\rvert^3}+\left(\frac{1}{\lvert\mathbf{R}^{(a)}\rvert^3}-\frac{1}{\lvert\mathbf{R}^{(b)}\rvert^3}\right)\hat{s}^2Q_{S,6}^{(b)}\right)\\
&+\left(\frac{1}{\lvert\mathbf{R}_t^{(a)}\rvert^3}-\frac{1}{\lvert\mathbf{R}^{(a)}\rvert^3}\right)\left(\epsilon\hat{s}(Q_{S,5}^{(a)}-Q_{S,5}^{(b)})-\epsilon^2(\kappa_3^{(a)}-\kappa_3^{(b)})\sin(\hat{\theta})\right)\\
&+\left(\frac{1}{\lvert\mathbf{R}_t^{(a)}\rvert^3}-\frac{1}{\lvert\mathbf{R}^{(a)}\rvert^3}-\frac{1}{\lvert\mathbf{R}_t^{(b)}\rvert^3}+\frac{1}{\lvert\mathbf{R}^{(b)}\rvert^3}\right)\left(\hat{s}+\epsilon\hat{s}Q_{S,5}^{(b)}-\epsilon^2\kappa_3^{(b)}\sin(\hat{\theta})\right)\bar{e}_j^{(a)}\cdot\varphi_0^{\Phi^{(a)}}\\
&+\left(\frac{\hat{s}^2Q_{S,6}^{(b)}}{\lvert\mathbf{R}^{(b)}\rvert^3}+\left(\frac{1}{\lvert\mathbf{R}_t^{(b)}\rvert^3}-\frac{1}{\lvert\mathbf{R}^{(b)}\rvert^3}\right)\left(\hat{s}+\epsilon\hat{s}Q_{S,5}^{(b)}-\epsilon^2\kappa_3^{(b)}\sin(\hat{\theta})\right)\right)\left(\bar{e}_j^{(a)}\cdot\varphi_0^{\Phi^{(a)}}-\bar{e}_j^{(b)}\cdot\varphi_0^{\Phi^{(b)}}\right).
\end{aligned} S 1 , 1 ( a ) − S 1 , 1 ( b ) = ( ∣ R ( a ) ∣ 3 s ^ 2 ( Q S , 6 ( a ) − Q S , 6 ( b ) ) + ( ∣ R ( a ) ∣ 3 1 − ∣ R ( b ) ∣ 3 1 ) s ^ 2 Q S , 6 ( b ) ) + ( ∣ R t ( a ) ∣ 3 1 − ∣ R ( a ) ∣ 3 1 ) ( ϵ s ^ ( Q S , 5 ( a ) − Q S , 5 ( b ) ) − ϵ 2 ( κ 3 ( a ) − κ 3 ( b ) ) sin ( θ ^ ) ) + ( ∣ R t ( a ) ∣ 3 1 − ∣ R ( a ) ∣ 3 1 − ∣ R t ( b ) ∣ 3 1 + ∣ R ( b ) ∣ 3 1 ) ( s ^ + ϵ s ^ Q S , 5 ( b ) − ϵ 2 κ 3 ( b ) sin ( θ ^ ) ) e ˉ j ( a ) ⋅ φ 0 Φ ( a ) + ( ∣ R ( b ) ∣ 3 s ^ 2 Q S , 6 ( b ) + ( ∣ R t ( b ) ∣ 3 1 − ∣ R ( b ) ∣ 3 1 ) ( s ^ + ϵ s ^ Q S , 5 ( b ) − ϵ 2 κ 3 ( b ) sin ( θ ^ ) ) ) ( e ˉ j ( a ) ⋅ φ 0 Φ ( a ) − e ˉ j ( b ) ⋅ φ 0 Φ ( b ) ) .
Since each remainder coefficient satisfies the Lipschitz estimate (214) , using (210) , (228) , and Lemma 4.4, case (1), along with the bounds (81) and (82) , we have
∣ ∂ s R S , 1 , 1 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 1 , 1 ( b ) [ φ 0 Φ ( b ) ] ∣ C ˙ 0 , β ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , β , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ L ∞ . (235) \left\lvert\partial_s\mathcal{R}_{S,1,1}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\partial_s\mathcal{R}_{S,1,1}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right\rvert_{\dot{C}^{0,\beta}}\leq c(\kappa_*^{(a)},\kappa_*^{(b)},\beta,c_\Gamma)\epsilon^{1-2\beta}\left\lVert\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\rVert_{C^{2,\beta}}\lVert\varphi\rVert_{L^\infty}. \tag*{(235)} ∂ s R S , 1 , 1 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 1 , 1 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , β ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , β , c Γ ) ϵ 1 − 2 β X ( a ) − X ( b ) C 2 , β ∥ φ ∥ L ∞ . (235)
Similarly, for j = t , r , θ j=t,r,\theta j = t , r , θ , we write
e ˉ j ⋅ ∂ s R S , 1 , 2 = p.v. ∫ − 1 / 2 1 / 2 ∫ − π π S 1 , 2 ϵ d θ ^ d s ^ , \bar{e}_j\cdot\partial_s\mathcal{R}_{S,1,2}=\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}S_{1,2}\,\epsilon\,d\hat{\theta}d\hat{s}, e ˉ j ⋅ ∂ s R S , 1 , 2 = p.v. ∫ − 1/2 1/2 ∫ − π π S 1 , 2 ϵ d θ ^ d s ^ ,
S 1 , 2 = ( ∂ s R ‾ ⋅ R ‾ − ∂ s R t ⋅ R t ∣ R t ∣ 3 + ( 1 ∣ R ‾ ∣ 3 − 1 ∣ R t ∣ 3 ) ∂ s R ‾ ⋅ R ‾ ) e ˉ j ⋅ φ = ( − ϵ s ^ Q S , 5 − ϵ 2 κ 3 sin ( θ ^ ) ∣ R ∣ 3 + s ^ ( 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 ) ) e ‾ j ⋅ φ . \begin{aligned}
S_{1,2}=\left(\frac{\partial_s\overline{\mathbf{R}}\cdot\overline{\mathbf{R}}-\partial_s\mathbf{R}_t\cdot\mathbf{R}_t}{\lvert\mathbf{R}_t\rvert^3}+\left(\frac{1}{\lvert\overline{\mathbf{R}}\rvert^3}-\frac{1}{\lvert\mathbf{R}_t\rvert^3}\right)\partial_s\overline{\mathbf{R}}\cdot\overline{\mathbf{R}}\right)\bar{e}_j\cdot\varphi\\
&=\left(-\frac{\epsilon\widehat{s}Q_{S,5}-\epsilon^{2}\kappa_{3}\sin(\widehat{\theta})}{|R|^{3}}+\widehat{s}\left(\frac{1}{|R|^{3}}-\frac{1}{|R_{t}|^{3}}\right)\right)\overline{e}_{j}\cdot\varphi.
\end{aligned} S 1 , 2 = ( ∣ R t ∣ 3 ∂ s R ⋅ R − ∂ s R t ⋅ R t + ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) ∂ s R ⋅ R ) e ˉ j ⋅ φ = ( − ∣ R ∣ 3 ϵ s Q S , 5 − ϵ 2 κ 3 sin ( θ ) + s ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) ) e j ⋅ φ .
Using (228) and Lemmas 4.3 and 4.2 , we have
∣ ∂ s R S , 1 , 2 [ φ ] ∣ ≤ c ( κ ∗ , c Γ ) ϵ 1 + α ( max i ∥ Q i ∥ C 2 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ ∥ φ ∥ C 0 , α , |\partial_{s}\mathcal{R}_{S,1,2}[\varphi]|\leq c(\kappa_{*},c_{\Gamma})\epsilon^{1+\alpha}\left(\max_{i}\|Q_{i}\|_{C_{2}^{0,\alpha}}\right)\|\varphi\|_{C^{0,\alpha}}\leq c(\kappa_{*,\alpha},c_{\Gamma})\epsilon\|\varphi\|_{C^{0,\alpha}}, ∣ ∂ s R S , 1 , 2 [ φ ] ∣ ≤ c ( κ ∗ , c Γ ) ϵ 1 + α ( i max ∥ Q i ∥ C 2 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ ∥ φ ∥ C 0 , α ,
where the Q i Q_i Q i include Q S , 1 Q_{S,1} Q S , 1 , Q S , 2 Q_{S,2} Q S , 2 , and Q S , 5 Q_{S,5} Q S , 5 . In addition, using case (2) of Lemma 4.4 along with (81) and (82) , we have
∣ ∂ s R S , 1 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ max i ( ∥ Q i ∥ C 0 0 , α + + ∥ Q i ∥ C 0 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α , \begin{aligned}
|\partial_{s}\mathcal{R}_{S,1,2}[\varphi]|_{\dot{C}^{0,\alpha}}\leq c(\kappa_{*},c_{\Gamma})\epsilon\max_{i}\left(\|Q_i\|_{C_{0}^{0,\alpha_{+}}}+\|Q_i\|_{C_{0}^{0,\alpha}}\right)\|\varphi\|_{C^{0,\alpha}} \\
&\leq c(\kappa_{*,\alpha_{+}},c_{\Gamma})\epsilon^{1-\alpha_{+}}\|\varphi\|_{C^{0,\alpha}},
\end{aligned} ∣ ∂ s R S , 1 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ i max ( ∥ Q i ∥ C 0 0 , α + + ∥ Q i ∥ C 0 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α ,
∣ ∂ s R S , 1 , 2 [ φ 0 ( a ) ] − ∂ s R S , 1 , 2 [ φ 0 ( b ) ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α . (236) \left|\partial_{s}\mathcal{R}_{S,1,2}\left[\varphi_{0}^{(a)}\right]-\partial_{s}\mathcal{R}_{S,1,2}\left[\varphi_{0}^{(b)}\right]\right|_{\dot{C}^{0,\alpha}}
\leq c(\kappa_{*,\alpha_{+}}^{(a)},\kappa_{*,\alpha_{+}}^{(b)},c_{\Gamma})\epsilon^{1-\alpha_{+}}\|X^{(a)}-X^{(b)}\|_{C^{2,\alpha_{+}}}\|\varphi\|_{C^{0,\alpha}} . \tag*{(236)} ∂ s R S , 1 , 2 [ φ 0 ( a ) ] − ∂ s R S , 1 , 2 [ φ 0 ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α . (236)
We next consider ∂ s R S , 2 \partial_{s}\mathcal{R}_{S,2} ∂ s R S , 2 , which we again treat in two parts as ∂ s R S , 2 , 1 \partial_{s}\mathcal{R}_{S,2,1} ∂ s R S , 2 , 1 and ∂ s R S , 2 , 2 \partial_{s}\mathcal{R}_{S,2,2} ∂ s R S , 2 , 2 . Again, we will treat the commutator term [ ∂ s , Φ − 1 ] [\partial_{s},\Phi^{-1}] [ ∂ s , Φ − 1 ] as part of the smoother remainder ∂ s R S , 2 , 1 \partial_{s}\mathcal{R}_{S,2,1} ∂ s R S , 2 , 1 . For each j ∈ { t , r , θ } j\in\{t,r,\theta\} j ∈ { t , r , θ } , we expand the components of ∂ s R S , 2 , 1 \partial_{s}\mathcal{R}_{S,2,1} ∂ s R S , 2 , 1 as
e ‾ j ⋅ ∂ s R S , 2 , 1 = p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( S 2 , 1 , j + S 2 , 2 , j + S 2 , 3 , j ) ϵ d θ d s ^ + e ‾ j 8 π ⋅ [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π R ( R ⋅ φ ) ∣ R ∣ 3 ϵ d θ d s ^ , \overline{e}_{j}\cdot\partial_{s}\mathcal{R}_{S,2,1}
=\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}(S_{2,1,j}+S_{2,2,j}+S_{2,3,j})\epsilon\,d\theta\,d\widehat{s}
+\frac{\overline{e}_{j}}{8\pi}\cdot[\partial_{s},\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{R(R\cdot\varphi)}{|R|^{3}}\epsilon\,d\theta\,d\widehat{s}, e j ⋅ ∂ s R S , 2 , 1 = p.v. ∫ − 1/2 1/2 ∫ − π π ( S 2 , 1 , j + S 2 , 2 , j + S 2 , 3 , j ) ϵ d θ d s + 8 π e j ⋅ [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 3 R ( R ⋅ φ ) ϵ d θ d s ,
S 2 , 1 , j = ( e ‾ j ⋅ ∂ s R ) ( R ⋅ φ ) ∣ R ∣ 3 − ( e ‾ j ⋅ ∂ s R t ) ( R t ⋅ φ ) ∣ R t ∣ 3 , S_{2,1,j}=\frac{(\overline{e}_{j}\cdot\partial_{s}R)(R\cdot\varphi)}{|R|^{3}}-\frac{(\overline{e}_{j}\cdot\partial_{s}R_{t})(R_{t}\cdot\varphi)}{|R_{t}|^{3}}, S 2 , 1 , j = ∣ R ∣ 3 ( e j ⋅ ∂ s R ) ( R ⋅ φ ) − ∣ R t ∣ 3 ( e j ⋅ ∂ s R t ) ( R t ⋅ φ ) ,
S 2 , 2 , j = ( e ‾ j ⋅ R ) ( ∂ s R ⋅ φ ) ∣ R ∣ 3 − ( e ‾ j ⋅ R t ) ( ∂ s R t ⋅ φ ) ∣ R t ∣ 3 , S_{2,2,j}=\frac{(\overline{e}_{j}\cdot R)(\partial_{s}R\cdot\varphi)}{|R|^{3}}-\frac{(\overline{e}_{j}\cdot R_{t})(\partial_{s}R_{t}\cdot\varphi)}{|R_{t}|^{3}}, S 2 , 2 , j = ∣ R ∣ 3 ( e j ⋅ R ) ( ∂ s R ⋅ φ ) − ∣ R t ∣ 3 ( e j ⋅ R t ) ( ∂ s R t ⋅ φ ) ,
S 2 , 3 , j = − 3 ( e ‾ j ⋅ R ) ( R ⋅ φ ) ( ∂ s R ⋅ R ) ∣ R ∣ 5 + 3 ( e ‾ j ⋅ R t ) ( R t ⋅ φ ) ( ∂ s R t ⋅ R t ) ∣ R t ∣ 5 . S_{2,3,j}=-3\frac{(\overline{e}_{j}\cdot R)(R\cdot\varphi)(\partial_{s}R\cdot R)}{|R|^{5}}+3\frac{(\overline{e}_{j}\cdot R_{t})(R_{t}\cdot\varphi)(\partial_{s}R_{t}\cdot R_{t})}{|R_{t}|^{5}}. S 2 , 3 , j = − 3 ∣ R ∣ 5 ( e j ⋅ R ) ( R ⋅ φ ) ( ∂ s R ⋅ R ) + 3 ∣ R t ∣ 5 ( e j ⋅ R t ) ( R t ⋅ φ ) ( ∂ s R t ⋅ R t ) .
Using the expansions (203) , (206) , (232) , and (233) , we may further expand the tangential direction integrand as
S 2 , 1 , t = − s ^ 2 ( 1 − ϵ κ ^ ) ( Q t ⋅ φ ) ∣ R ∣ 3 + ( 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 ) ( 1 − ϵ κ ^ ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ S_{2,1,t}=-\frac{\widehat{s}^{2}(1-\epsilon\widehat{\kappa})(Q_{t}\cdot\varphi)}{|R|^{3}}+\left(\frac{1}{|R|^{3}}-\frac{1}{|R_{t}|^{3}}\right)(1-\epsilon\widehat{\kappa})(\widehat{s}e_{t}+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}+\epsilon\widehat{s}Q_{r})\cdot\varphi S 2 , 1 , t = − ∣ R ∣ 3 s 2 ( 1 − ϵ κ ) ( Q t ⋅ φ ) + ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) ( 1 − ϵ κ ) ( s e t + 2 ϵ sin ( 2 θ ) e θ + ϵ s Q r ) ⋅ φ
S 2 , 2 , t = − s ^ 3 Q t ( ( 1 − ϵ κ ^ ) e t + ϵ κ 3 e θ ) ⋅ φ ∣ R ∣ 3 − s ^ 2 ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ( 1 + ϵ Q r ⋅ e t ) ∣ R ∣ 3 + ( 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 ) s ^ ( 1 + ϵ Q r ⋅ e t ) ( ( 1 − ϵ κ ^ ) e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ \begin{aligned}
S_{2,2,t}=-\frac{\widehat{s}^{3}Q_{t}((1-\epsilon\widehat{\kappa})e_{t}+\epsilon\kappa_{3}e_{\theta})\cdot\varphi}{|R|^{3}}-\frac{\widehat{s}^{2}(\kappa_{1}e_{n_{1}}+\kappa_{2}e_{n_{2}})\cdot\varphi(1+\epsilon Q_{r}\cdot e_{t})}{|R|^{3}} \\
&+\left(\frac{1}{|R|^{3}}-\frac{1}{|R_{t}|^{3}}\right)\widehat{s}(1+\epsilon Q_{r}\cdot e_{t})((1-\epsilon\widehat{\kappa})e_{t}+\widehat{s}(\kappa_{1}e_{n_{1}}+\kappa_{2}e_{n_{2}})+\epsilon\kappa_{3}e_{\theta})\cdot\varphi
\end{aligned} S 2 , 2 , t = − ∣ R ∣ 3 s 3 Q t (( 1 − ϵ κ ) e t + ϵ κ 3 e θ ) ⋅ φ − ∣ R ∣ 3 s 2 ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ( 1 + ϵ Q r ⋅ e t ) + ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) s ( 1 + ϵ Q r ⋅ e t ) (( 1 − ϵ κ ) e t + s ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ
S 2 , 3 , t = 3 s ^ 3 Q S , 6 ( 1 + ϵ Q r ⋅ e t ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ∣ R ∣ 5 + 3 s ^ 3 [ Q t ( s ^ Q ^ t + 2 ϵ sin ( θ ^ 2 ) e θ ) + ( 1 + ϵ Q r ⋅ e t ) Q t ] ⋅ φ ( s ^ Q ^ R s 1 + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 5 \begin{aligned}
S_{2,3,t}=\frac{3\widehat{s}^{3}Q_{S,6}(1+\epsilon Q_{r}\cdot e_{t})(\widehat{s}e_{t}+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}+\epsilon\widehat{s}Q_{r})\cdot\varphi}{|R|^{5}} \\
&+\frac{3\widehat{s}^{3}[Q_{t}(\widehat{s}\widehat{Q}_{t}+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta})+(1+\epsilon Q_{r}\cdot e_{t})Q_{t}]\cdot\varphi(\widehat{s}\widehat{Q}_{Rs1}+2\epsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta}))}{|R|^{5}}
\end{aligned} S 2 , 3 , t = ∣ R ∣ 5 3 s 3 Q S , 6 ( 1 + ϵ Q r ⋅ e t ) ( s e t + 2 ϵ sin ( 2 θ ) e θ + ϵ s Q r ) ⋅ φ + ∣ R ∣ 5 3 s 3 [ Q t ( s Q t + 2 ϵ sin ( 2 θ ) e θ ) + ( 1 + ϵ Q r ⋅ e t ) Q t ] ⋅ φ ( s Q R s 1 + 2 ϵ 2 sin ( 2 θ ) sin ( θ ))
− 3 ( 1 ∣ R ∣ 5 − 1 ∣ R t ∣ 5 ) s ^ ( 1 + ϵ Q r ⋅ e t ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ( s ^ Q ^ R s 2 + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) , -3\left(\frac{1}{|R|^{5}}-\frac{1}{|R_{t}|^{5}}\right)\widehat{s}(1+\epsilon Q_{r}\cdot e_{t})(\widehat{s}e_{t}+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}+\epsilon\widehat{s}Q_{r})\cdot\varphi(\widehat{s}\widehat{Q}_{Rs2}+2\epsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})), − 3 ( ∣ R ∣ 5 1 − ∣ R t ∣ 5 1 ) s ( 1 + ϵ Q r ⋅ e t ) ( s e t + 2 ϵ sin ( 2 θ ) e θ + ϵ s Q r ) ⋅ φ ( s Q R s 2 + 2 ϵ 2 sin ( 2 θ ) sin ( θ )) ,
where we define Q ~ t = e t − s ^ Q t + ϵ Q r \widetilde{Q}_t=e_t-\hat{s}Q_t+\epsilon Q_r Q t = e t − s ^ Q t + ϵ Q r , Q ~ R s 1 = 1 + ϵ Q R s 1 + s ^ Q R s 2 \widetilde{Q}_{Rs1}=1+\epsilon Q_{Rs1}+\hat{s}Q_{Rs2} Q R s 1 = 1 + ϵ Q R s 1 + s ^ Q R s 2 , and Q ~ R s 2 = 1 + ϵ Q R s 1 + s ^ ( Q R s 2 − Q S , 6 ) \widetilde{Q}_{Rs2}=1+\epsilon Q_{Rs1}+\hat{s}(Q_{Rs2}-Q_{S,6}) Q R s 2 = 1 + ϵ Q R s 1 + s ^ ( Q R s 2 − Q S , 6 ) . Using the expansion (227) for 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 \frac{1}{|R|^3}-\frac{1}{|R_t|^3} ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 and 1 ∣ R ∣ 5 − 1 ∣ R t ∣ 5 \frac{1}{|R|^5}-\frac{1}{|R_t|^5} ∣ R ∣ 5 1 − ∣ R t ∣ 5 1 , along with the commutator bound (78), we may apply Lemma 4.4, case (1), to obtain, for α < β < 1 \alpha<\beta<1 α < β < 1 ,
∣ e z ⋅ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( max i ∥ Q i ∥ C 1 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ . \left|e_z\cdot\partial_s\mathcal{R}_{S,2,1}[\varphi]\right|_{\dot{C}^{0,\beta}}\leq c(\kappa_*,c_\Gamma)\epsilon^{1-\beta}\left(\max_i\|Q_i\|_{C^1,\beta}\right)\|\varphi\|_{L^\infty}\leq c(\kappa_{*,\beta},c_\Gamma)\epsilon^{1-2\beta}\|\varphi\|_{L^\infty}. ∣ e z ⋅ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( i max ∥ Q i ∥ C 1 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ .
Here Q i Q_i Q i is used as a catch-all notation for each of the coefficients appearing in the expressions above, including κ ^ ( s , θ ) \hat{\kappa}(s,\theta) κ ^ ( s , θ ) , κ j ( s ) \kappa_j(s) κ j ( s ) , Q t Q_t Q t , Q ^ t \widehat{Q}_t Q t , Q r Q_r Q r , and additional Q Q Q remainders. In addition, using (213), (210), (227), (81), and (82), by Lemma 4.4, we have
∣ e z ⋅ ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − e z ⋅ ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ L ∞ . \left|e_z\cdot\partial_s\mathcal{R}_{S,2,1}^{(a)}[\varphi_0^{\Phi^{(a)}}]-e_z\cdot\partial_s\mathcal{R}_{S,2,1}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right|_{\dot{C}^{0,\beta}}\leq c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)},c_\Gamma)\epsilon^{1-2\beta}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}}\|\varphi\|_{L^\infty}. e z ⋅ ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − e z ⋅ ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β X ( a ) − X ( b ) C 2 , β ∥ φ ∥ L ∞ .
We next expand the e r e_r e r -direction integrand of the first integral in ∂ s R S , 2 , 1 \partial_s\mathcal{R}_{S,2,1} ∂ s R S , 2 , 1 using (202), (205), (231), and (232). We have
S 2 , 1 , r = − − s ^ κ ^ ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ + s ^ 2 Q t ) ⋅ φ ∣ R ∣ 3 + ( 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 ) s ^ κ ^ ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ , S 2 , 2 , r = − s ^ 2 ( e r ⋅ Q t ) ( ( 1 − ϵ κ ^ ) e t + ϵ κ 3 e θ ) ⋅ φ ∣ R ∣ 3 − s ^ ( 2 ϵ sin 2 ( θ ^ 2 ) + ϵ s ^ 2 Q 0 , 5 ) ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ∣ R ∣ 3 + ( 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 ) ( 2 ϵ sin 2 ( θ ^ 2 ) + ϵ s ^ 2 Q 0 , 5 ) ( ( 1 − ϵ κ ^ ) e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ , S 2 , 3 , r = 3 s ^ 3 κ ^ Q S , 6 ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ∣ R ∣ 5 + 3 s ^ 2 ( e r ⋅ Q t ) ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ ) + ( 2 ϵ sin 2 ( θ ^ 2 ) + ϵ s ^ 2 Q 0 , 5 ) Q t ⋅ φ ( s ^ Q ~ R s 1 + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 5 − 3 ( 1 ∣ R ∣ 5 − 1 ∣ R t ∣ 5 ) ( 2 ϵ sin 2 ( θ ^ 2 ) + ϵ s ^ 2 Q 0 , 5 ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ( s ^ Q ~ R s 2 + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) . \begin{aligned}
S_{2,1,r}={}&-\frac{-\hat{s}\hat{\kappa}(\hat{s}\widetilde{Q}_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\hat{s}^2Q_t)\cdot\varphi}{|R|^3}
+\left(\frac{1}{|R|^3}-\frac{1}{|R_t|^3}\right)\hat{s}\hat{\kappa}(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi,\\
S_{2,2,r}={}&-\frac{\hat{s}^2(e_r\cdot Q_t)((1-\epsilon\hat{\kappa})e_t+\epsilon\kappa_3e_\theta)\cdot\varphi}{|R|^3}
-\frac{\hat{s}(2\epsilon\sin^2(\frac{\hat{\theta}}{2})+\epsilon\hat{s}^2Q_{0,5})(\kappa_1e_{n_1}+\kappa_2e_{n_2})\cdot\varphi}{|R|^3}\\
&+\left(\frac{1}{|R|^3}-\frac{1}{|R_t|^3}\right)(2\epsilon\sin^2(\frac{\hat{\theta}}{2})+\epsilon\hat{s}^2Q_{0,5})((1-\epsilon\hat{\kappa})e_t+\hat{s}(\kappa_1e_{n_1}+\kappa_2e_{n_2})+\epsilon\kappa_3e_\theta)\cdot\varphi,\\
S_{2,3,r}={}&\frac{3\hat{s}^3\hat{\kappa}Q_{S,6}(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi}{|R|^5}\\
&+\frac{3\hat{s}^2(e_r\cdot Q_t)(\hat{s}\widetilde{Q}_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta)+(2\epsilon\sin^2(\frac{\hat{\theta}}{2})+\epsilon\hat{s}^2Q_{0,5})Q_t\cdot\varphi(\hat{s}\widetilde{Q}_{Rs1}+2\epsilon^2\sin(\frac{\hat{\theta}}{2})\sin(\hat{\theta}))}{|R|^5}\\
&-3\left(\frac{1}{|R|^5}-\frac{1}{|R_t|^5}\right)(2\epsilon\sin^2(\frac{\hat{\theta}}{2})+\epsilon\hat{s}^2Q_{0,5})(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi(\hat{s}\widetilde{Q}_{Rs2}+2\epsilon^2\sin(\frac{\hat{\theta}}{2})\sin(\hat{\theta})).
\end{aligned} S 2 , 1 , r = S 2 , 2 , r = S 2 , 3 , r = − ∣ R ∣ 3 − s ^ κ ^ ( s ^ Q t + 2 ϵ sin ( 2 θ ^ ) e θ + s ^ 2 Q t ) ⋅ φ + ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) s ^ κ ^ ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ , − ∣ R ∣ 3 s ^ 2 ( e r ⋅ Q t ) (( 1 − ϵ κ ^ ) e t + ϵ κ 3 e θ ) ⋅ φ − ∣ R ∣ 3 s ^ ( 2 ϵ sin 2 ( 2 θ ^ ) + ϵ s ^ 2 Q 0 , 5 ) ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ + ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) ( 2 ϵ sin 2 ( 2 θ ^ ) + ϵ s ^ 2 Q 0 , 5 ) (( 1 − ϵ κ ^ ) e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ , ∣ R ∣ 5 3 s ^ 3 κ ^ Q S , 6 ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ + ∣ R ∣ 5 3 s ^ 2 ( e r ⋅ Q t ) ( s ^ Q t + 2 ϵ sin ( 2 θ ^ ) e θ ) + ( 2 ϵ sin 2 ( 2 θ ^ ) + ϵ s ^ 2 Q 0 , 5 ) Q t ⋅ φ ( s ^ Q R s 1 + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) − 3 ( ∣ R ∣ 5 1 − ∣ R t ∣ 5 1 ) ( 2 ϵ sin 2 ( 2 θ ^ ) + ϵ s ^ 2 Q 0 , 5 ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ ( s ^ Q R s 2 + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) .
where Q ~ t \widetilde{Q}_t Q t , Q ~ R s 1 \widetilde{Q}_{Rs1} Q R s 1 , and Q ~ R s 2 \widetilde{Q}_{Rs2} Q R s 2 are as defined above. We again use the expansion (227) for 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 \frac{1}{|R|^3}-\frac{1}{|R_t|^3} ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 and 1 ∣ R ∣ 5 − 1 ∣ R t ∣ 5 \frac{1}{|R|^5}-\frac{1}{|R_t|^5} ∣ R ∣ 5 1 − ∣ R t ∣ 5 1 , along with the commutator estimate (78), and apply case (1) of Lemma 4.4 to obtain, for α < β < 1 \alpha<\beta<1 α < β < 1 ,
∣ e ˉ r ⋅ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( max i ∥ Q i ∥ C 1 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ , \left|\bar{e}_r\cdot\partial_s\mathcal{R}_{S,2,1}[\varphi]\right|_{\dot{C}^{0,\beta}}\leq c(\kappa_*,c_\Gamma)\epsilon^{1-\beta}\left(\max_i\|Q_i\|_{C^{1,\beta}}\right)\|\varphi\|_{L^\infty}\leq c(\kappa_{*,\beta},c_\Gamma)\epsilon^{1-2\beta}\|\varphi\|_{L^\infty}, ∣ e ˉ r ⋅ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( i max ∥ Q i ∥ C 1 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ ,
∣ e ˉ r ⋅ ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − e ˉ r ⋅ ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ L ∞ . \left|\bar{e}_r\cdot\partial_s\mathcal{R}_{S,2,1}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\bar{e}_r\cdot\partial_s\mathcal{R}_{S,2,1}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right|_{\dot{C}^{0,\beta}}\leq c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)},c_\Gamma)\epsilon^{1-2\beta}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}}\|\varphi\|_{L^\infty}. e ˉ r ⋅ ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − e ˉ r ⋅ ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β X ( a ) − X ( b ) C 2 , β ∥ φ ∥ L ∞ .
Finally, we expand the e θ e_\theta e θ -direction integrand of the first integral of ∂ s R S , 2 , 1 \partial_s\mathcal{R}_{S,2,1} ∂ s R S , 2 , 1 using (202), (205), (231), and (232). We have
S 2 , 1 , θ = − s ^ ∂ θ ^ κ ^ ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ ∣ R ∣ 3 − s ^ 2 ( Q t ⋅ φ ) ( ϵ κ 3 + s ^ ∂ θ ^ κ ^ ) ∣ R ∣ 3 + ( 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 ) ( ϵ κ 3 + s ^ ∂ θ ^ κ ^ ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ , S 2 , 2 , θ = − s ^ 2 ( e θ ⋅ Q t ) ( ( 1 − ϵ κ ^ ) e t + ϵ κ 3 e θ ) ⋅ φ ∣ R ∣ 3 − s ^ ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ( ϵ sin ( θ ^ ) + ϵ s ^ Q r ⋅ e θ ) ∣ R ∣ 3 + ( 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 ) ( ϵ sin ( θ ^ ) + ϵ s ^ Q r ⋅ e θ ) ( ( 1 − ϵ κ ^ ) e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ \begin{aligned}
S_{2,1,\theta}={}&-\frac{\hat{s}\partial_{\hat{\theta}}\hat{\kappa}(\hat{s}\widetilde{Q}_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta)\cdot\varphi}{|R|^3}
-\frac{\hat{s}^2(Q_t\cdot\varphi)(\epsilon\kappa_3+\hat{s}\partial_{\hat{\theta}}\hat{\kappa})}{|R|^3}\\
&+\left(\frac{1}{|R|^3}-\frac{1}{|R_t|^3}\right)(\epsilon\kappa_3+\hat{s}\partial_{\hat{\theta}}\hat{\kappa})(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi,\\
S_{2,2,\theta}={}&-\frac{\hat{s}^2(e_\theta\cdot Q_t)((1-\epsilon\hat{\kappa})e_t+\epsilon\kappa_3e_\theta)\cdot\varphi}{|R|^3}
-\frac{\hat{s}(\kappa_1e_{n_1}+\kappa_2e_{n_2})\cdot\varphi(\epsilon\sin(\hat{\theta})+\epsilon\hat{s}Q_r\cdot e_\theta)}{|R|^3}\\
&+\left(\frac{1}{|R|^3}-\frac{1}{|R_t|^3}\right)(\epsilon\sin(\hat{\theta})+\epsilon\hat{s}Q_r\cdot e_\theta)((1-\epsilon\hat{\kappa})e_t+\hat{s}(\kappa_1e_{n_1}+\kappa_2e_{n_2})+\epsilon\kappa_3e_\theta)\cdot\varphi
\end{aligned} S 2 , 1 , θ = S 2 , 2 , θ = − ∣ R ∣ 3 s ^ ∂ θ ^ κ ^ ( s ^ Q t + 2 ϵ sin ( 2 θ ^ ) e θ ) ⋅ φ − ∣ R ∣ 3 s ^ 2 ( Q t ⋅ φ ) ( ϵ κ 3 + s ^ ∂ θ ^ κ ^ ) + ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) ( ϵ κ 3 + s ^ ∂ θ ^ κ ^ ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ , − ∣ R ∣ 3 s ^ 2 ( e θ ⋅ Q t ) (( 1 − ϵ κ ^ ) e t + ϵ κ 3 e θ ) ⋅ φ − ∣ R ∣ 3 s ^ ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ( ϵ sin ( θ ^ ) + ϵ s ^ Q r ⋅ e θ ) + ( ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 ) ( ϵ sin ( θ ^ ) + ϵ s ^ Q r ⋅ e θ ) (( 1 − ϵ κ ^ ) e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ
S 2 , 3 , θ = 3 s ^ 2 Q S , 6 ( ϵ sin ( θ ^ 2 ) + ϵ s ^ Q r ⋅ e θ ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ∣ R ∣ 5 + 3 s ^ 2 [ e θ ⋅ Q t ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ ) + ϵ ( sin ( θ ^ ) + s ^ Q r ⋅ e θ ) Q t ] ⋅ φ ( s ^ Q ~ R s 1 + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 5 − 3 ( 1 ∣ R ∣ 5 − 1 ∣ R t ∣ 5 ) ϵ ( sin ( θ ^ ) + s ^ Q r ⋅ e θ ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ( s ^ Q ~ R s 2 + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) , S_{2,3,\theta} = \frac{3\hat{s}^{2}Q_{S,6}(\epsilon\sin(\frac{\hat{\theta}}{2})+\epsilon\hat{s}Q_r\cdot e_\theta)(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi}{|R|^{5}}
+ \frac{3\hat{s}^{2}\left[e_\theta\cdot Q_t(\hat{s}\widetilde{Q}_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta)+\epsilon(\sin(\hat{\theta})+\hat{s}Q_r\cdot e_\theta)Q_t\right]\cdot\varphi(\hat{s}\widetilde{Q}_{Rs1}+2\epsilon^{2}\sin(\frac{\hat{\theta}}{2})\sin(\hat{\theta}))}{|R|^{5}}
-3\left(\frac{1}{|R|^{5}}-\frac{1}{|R_t|^{5}}\right)\epsilon(\sin(\hat{\theta})+\hat{s}Q_r\cdot e_\theta)(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi(\hat{s}\widetilde{Q}_{Rs2}+2\epsilon^{2}\sin(\frac{\hat{\theta}}{2})\sin(\hat{\theta})), S 2 , 3 , θ = ∣ R ∣ 5 3 s ^ 2 Q S , 6 ( ϵ sin ( 2 θ ^ ) + ϵ s ^ Q r ⋅ e θ ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ + ∣ R ∣ 5 3 s ^ 2 [ e θ ⋅ Q t ( s ^ Q t + 2 ϵ sin ( 2 θ ^ ) e θ ) + ϵ ( sin ( θ ^ ) + s ^ Q r ⋅ e θ ) Q t ] ⋅ φ ( s ^ Q R s 1 + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) − 3 ( ∣ R ∣ 5 1 − ∣ R t ∣ 5 1 ) ϵ ( sin ( θ ^ ) + s ^ Q r ⋅ e θ ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ ( s ^ Q R s 2 + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) ,
with Q ~ t \widetilde{Q}_t Q t , Q ~ R s 1 \widetilde{Q}_{Rs1} Q R s 1 , and Q ~ R s 2 \widetilde{Q}_{Rs2} Q R s 2 as before. Again expanding 1 ∣ R ∣ 3 − 1 ∣ R t ∣ 3 \frac{1}{|R|^{3}}-\frac{1}{|R_t|^{3}} ∣ R ∣ 3 1 − ∣ R t ∣ 3 1 and 1 ∣ R ∣ 5 − 1 ∣ R t ∣ 5 \frac{1}{|R|^{5}}-\frac{1}{|R_t|^{5}} ∣ R ∣ 5 1 − ∣ R t ∣ 5 1 using (227), and using the commutator estimate (78), by Lemma 4.4, case (1), we have, for α < β < 1 \alpha<\beta<1 α < β < 1 ,
∣ e θ ‾ ⋅ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( max i ∥ Q i ∥ C 1 0 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ , \left|\overline{e_\theta}\cdot\partial_s\mathcal{R}_{S,2,1}[\varphi]\right|_{\dot{C}^{0,\beta}}\le c(\kappa_*,c_\Gamma)\epsilon^{1-\beta}\left(\max_i\|Q_i\|_{C_1^{0,\beta}}\right)\|\varphi\|_{L^\infty}\le c(\kappa_*,c_\Gamma)\epsilon^{1-2\beta}\|\varphi\|_{L^\infty}, ∣ e θ ⋅ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − β ( i max ∥ Q i ∥ C 1 0 , β ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ ,
∣ e θ ‾ ⋅ ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − e θ ‾ ⋅ ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] ∣ C ˙ 0 , β ≤ c ( κ ∗ ( a , β ) , κ ∗ ( b , β ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ L ∞ . \left|\overline{e_\theta}\cdot\partial_s\mathcal{R}_{S,2,1}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\overline{e_\theta}\cdot\partial_s\mathcal{R}_{S,2,1}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right|_{\dot{C}^{0,\beta}}\le c(\kappa_*^{(a,\beta)},\kappa_*^{(b,\beta)},c_\Gamma)\epsilon^{1-2\beta}\|X^{(a)}-X^{(b)}\|_{C^{2,\beta}}\|\varphi\|_{L^\infty}. e θ ⋅ ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − e θ ⋅ ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , β ≤ c ( κ ∗ ( a , β ) , κ ∗ ( b , β ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ L ∞ .
In total, for α < β < 1 \alpha<\beta<1 α < β < 1 , we have
∣ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞ \left|\partial_s\mathcal{R}_{S,2,1}[\varphi]\right|_{\dot{C}^{0,\beta}}\le c(\kappa_*,c_\Gamma)\epsilon^{1-2\beta}\|\varphi\|_{L^\infty} ∣ ∂ s R S , 2 , 1 [ φ ] ∣ C ˙ 0 , β ≤ c ( κ ∗ , c Γ ) ϵ 1 − 2 β ∥ φ ∥ L ∞
∣ ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] ∣ C ˙ 0 , β ≤ c ( κ ∗ ( a , β ) , κ ∗ ( b , β ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ L ∞ . (237) \left|\partial_s\mathcal{R}_{S,2,1}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\partial_s\mathcal{R}_{S,2,1}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right|_{\dot{C}^{0,\beta}}\le c(\kappa_*^{(a,\beta)},\kappa_*^{(b,\beta)},c_\Gamma)\epsilon^{1-2\beta}\|X^{(a)}-X^{(b)}\|_{C^{2,\beta}}\|\varphi\|_{L^\infty}. \tag*{(237)} ∂ s R S , 2 , 1 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 2 , 1 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , β ≤ c ( κ ∗ ( a , β ) , κ ∗ ( b , β ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ φ ∥ L ∞ . (237)
We now turn to ∂ s R S , 2 , 2 \partial_s\mathcal{R}_{S,2,2} ∂ s R S , 2 , 2 . For j ∈ { t , r , θ } j\in\{t,r,\theta\} j ∈ { t , r , θ } , we again expand the components of ∂ s R S , 2 , 2 \partial_s\mathcal{R}_{S,2,2} ∂ s R S , 2 , 2 as
e j ‾ ⋅ ∂ s R S , 2 , 2 = p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( S 2 , 4 , j + S 2 , 5 , j + S 2 , 6 , j ) ϵ d θ ^ d s ^ , \overline{e_j}\cdot\partial_s\mathcal{R}_{S,2,2}=\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}(S_{2,4,j}+S_{2,5,j}+S_{2,6,j})\epsilon\,d\hat{\theta}\,d\hat{s}, e j ⋅ ∂ s R S , 2 , 2 = p.v. ∫ − 1/2 1/2 ∫ − π π ( S 2 , 4 , j + S 2 , 5 , j + S 2 , 6 , j ) ϵ d θ ^ d s ^ ,
S 2 , 4 , j = ( e j ‾ ⋅ ∂ s R t ) ( R t ⋅ φ ) ∣ R t ∣ 3 − ( e j ‾ ⋅ ∂ s R ‾ ) ( R ‾ ⋅ ( Φ − 1 φ ) ) ∣ R ‾ ∣ 3 S_{2,4,j}=\frac{(\overline{e_j}\cdot\partial_sR_t)(R_t\cdot\varphi)}{|R_t|^{3}}-\frac{(\overline{e_j}\cdot\partial_s\overline{R})(\overline{R}\cdot(\Phi^{-1}\varphi))}{|\overline{R}|^{3}} S 2 , 4 , j = ∣ R t ∣ 3 ( e j ⋅ ∂ s R t ) ( R t ⋅ φ ) − ∣ R ∣ 3 ( e j ⋅ ∂ s R ) ( R ⋅ ( Φ − 1 φ ))
S 2 , 5 , j = ( e j ‾ ⋅ R t ) ( ∂ s R t ⋅ φ ) ∣ R t ∣ 3 − ( e j ‾ ⋅ R ‾ ) ( ∂ s R ‾ ⋅ ( Φ − 1 φ ) ) ∣ R ‾ ∣ 3 S_{2,5,j}=\frac{(\overline{e_j}\cdot R_t)(\partial_sR_t\cdot\varphi)}{|R_t|^{3}}-\frac{(\overline{e_j}\cdot\overline{R})(\partial_s\overline{R}\cdot(\Phi^{-1}\varphi))}{|\overline{R}|^{3}} S 2 , 5 , j = ∣ R t ∣ 3 ( e j ⋅ R t ) ( ∂ s R t ⋅ φ ) − ∣ R ∣ 3 ( e j ⋅ R ) ( ∂ s R ⋅ ( Φ − 1 φ ))
S 2 , 6 , j = − 3 ( e j ‾ ⋅ R t ) ( R t ⋅ φ ) ( ∂ s R t ⋅ R t ) ∣ R t ∣ 5 + 3 ( e j ‾ ⋅ R ‾ ) ( R ‾ ⋅ ( Φ − 1 φ ) ) ( ∂ s R ‾ ⋅ R ‾ ) ∣ R ‾ ∣ 5 . S_{2,6,j}=-3\frac{(\overline{e_j}\cdot R_t)(R_t\cdot\varphi)(\partial_sR_t\cdot R_t)}{|R_t|^{5}}+3\frac{(\overline{e_j}\cdot\overline{R})(\overline{R}\cdot(\Phi^{-1}\varphi))(\partial_s\overline{R}\cdot\overline{R})}{|\overline{R}|^{5}}. S 2 , 6 , j = − 3 ∣ R t ∣ 5 ( e j ⋅ R t ) ( R t ⋅ φ ) ( ∂ s R t ⋅ R t ) + 3 ∣ R ∣ 5 ( e j ⋅ R ) ( R ⋅ ( Φ − 1 φ )) ( ∂ s R ⋅ R ) .
Similar to before, we use (196), (202), and (232) to expand the tangential direction integrand as
S 2 , 4 , t = ϵ κ ^ ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ + ϵ s ^ ( Q r ⋅ φ ) ∣ R t ∣ 3 + ( 1 ∣ R t ∣ 3 − 1 ∣ R ‾ ∣ 3 ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ S_{2,4,t}=\frac{\epsilon\hat{\kappa}(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi+\epsilon\hat{s}(Q_r\cdot\varphi)}{|R_t|^{3}}
+\left(\frac{1}{|R_t|^{3}}-\frac{1}{|\overline{R}|^{3}}\right)(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta)\cdot\varphi S 2 , 4 , t = ∣ R t ∣ 3 ϵ κ ^ ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ + ϵ s ^ ( Q r ⋅ φ ) + ( ∣ R t ∣ 3 1 − ∣ R ∣ 3 1 ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ ) ⋅ φ
S 2 , 5 , t = ϵ s ^ ( Q r ⋅ e t ) ( e t ⋅ φ ) ∣ R t ∣ 3 + s ^ ( 1 + ϵ Q r ⋅ e t ) ( − ϵ κ ^ e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ ∣ R t ∣ 3 + ( 1 ∣ R t ∣ 3 − 1 ∣ R ‾ ∣ 3 ) s ^ ( e t ⋅ φ ) S_{2,5,t}=\frac{\epsilon\hat{s}(Q_r\cdot e_t)(e_t\cdot\varphi)}{|R_t|^{3}}+\frac{\hat{s}(1+\epsilon Q_r\cdot e_t)(-\epsilon\hat{\kappa}e_t+\hat{s}(\kappa_1e_{n_1}+\kappa_2e_{n_2})+\epsilon\kappa_3e_\theta)\cdot\varphi}{|R_t|^{3}}
+\left(\frac{1}{|R_t|^{3}}-\frac{1}{|\overline{R}|^{3}}\right)\hat{s}(e_t\cdot\varphi) S 2 , 5 , t = ∣ R t ∣ 3 ϵ s ^ ( Q r ⋅ e t ) ( e t ⋅ φ ) + ∣ R t ∣ 3 s ^ ( 1 + ϵ Q r ⋅ e t ) ( − ϵ κ ^ e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ + ( ∣ R t ∣ 3 1 − ∣ R ∣ 3 1 ) s ^ ( e t ⋅ φ )
S 2 , 6 , t = − 3 ϵ s ^ ( Q r ⋅ e t ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ( s ^ + ϵ s ^ Q S , 5 − ϵ 2 κ 3 sin ( θ ^ ) ) ∣ R t ∣ 5 − 3 ϵ s ^ [ s ^ ( s ^ + ϵ s ^ Q S , 5 − ϵ 2 κ 3 sin ( θ ^ ) ) Q r + ( s ^ Q S , 5 − ϵ κ 3 sin ( θ ^ ) ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ ) ] ⋅ φ ∣ R t ∣ 5 S_{2,6,t}=-\frac{3\epsilon\hat{s}(Q_r\cdot e_t)(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta+\epsilon\hat{s}Q_r)\cdot\varphi(\hat{s}+\epsilon\hat{s}Q_{S,5}-\epsilon^{2}\kappa_3\sin(\hat{\theta}))}{|R_t|^{5}}
-\frac{3\epsilon\hat{s}\left[\hat{s}(\hat{s}+\epsilon\hat{s}Q_{S,5}-\epsilon^{2}\kappa_3\sin(\hat{\theta}))Q_r+(\hat{s}Q_{S,5}-\epsilon\kappa_3\sin(\hat{\theta}))(\hat{s}e_t+2\epsilon\sin(\frac{\hat{\theta}}{2})e_\theta)\right]\cdot\varphi}{|R_t|^{5}} S 2 , 6 , t = − ∣ R t ∣ 5 3 ϵ s ^ ( Q r ⋅ e t ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ ( s ^ + ϵ s ^ Q S , 5 − ϵ 2 κ 3 sin ( θ ^ )) − ∣ R t ∣ 5 3 ϵ s ^ [ s ^ ( s ^ + ϵ s ^ Q S , 5 − ϵ 2 κ 3 sin ( θ ^ )) Q r + ( s ^ Q S , 5 − ϵ κ 3 sin ( θ ^ )) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ ) ] ⋅ φ
− 3 ( 1 ∣ R t ∣ 5 − 1 ∣ R ∣ 5 ) s ^ 2 ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ . -3\left(\frac{1}{\lvert R_t\rvert^5}-\frac{1}{\lvert R\rvert^5}\right)\hat{s}^{2}\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta\right)\cdot\varphi. − 3 ( ∣ R t ∣ 5 1 − ∣ R ∣ 5 1 ) s ^ 2 ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ ) ⋅ φ .
Here most terms appear with an additional factor of ϵ \epsilon ϵ compared to those appearing in ∂ s R S , 2 , 1 \partial_s\mathcal{R}_{S,2,1} ∂ s R S , 2 , 1 , including the terms involving 1 ∣ R t ∣ 3 − 1 ∣ R ∣ 3 \frac{1}{\lvert R_t\rvert^3}-\frac{1}{\lvert R\rvert^3} ∣ R t ∣ 3 1 − ∣ R ∣ 3 1 and 1 ∣ R t ∣ 5 − 1 ∣ R ∣ 5 \frac{1}{\lvert R_t\rvert^5}-\frac{1}{\lvert R\rvert^5} ∣ R t ∣ 5 1 − ∣ R ∣ 5 1 , by (228) . The second term in S 2 , 5 , t S_{2,5,t} S 2 , 5 , t does not carry an additional factor of ϵ \epsilon ϵ in all components. In particular, there is a term of the form s ^ 2 ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ∣ R t ∣ 3 \frac{\hat{s}^{2}(\kappa_1e_{n_1}+\kappa_2e_{n_2})\cdot\varphi}{\lvert R_t\rvert^3} ∣ R t ∣ 3 s ^ 2 ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ . We emphasize that this term instead includes an additional factor of s ^ \hat{s} s ^ in the numerator, and that κ 1 e n 1 + κ 2 e n 2 \kappa_1e_{n_1}+\kappa_2e_{n_2} κ 1 e n 1 + κ 2 e n 2 depends only on s s s and not θ \theta θ . By case (1) of Lemma 4.4, this term satisfies
∣ ∫ − 1 / 2 1 / 2 ∫ − π π s ^ 2 ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ∣ R t ∣ 3 ϵ d θ d s ^ ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ 1 − α ∥ X s s ∥ C 0 , α ( T ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ . \begin{aligned}
\left\lvert\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\hat{s}^{2}(\kappa_1e_{n_1}+\kappa_2e_{n_2})\cdot\varphi}{\lvert R_t\rvert^3}\epsilon\,\mathrm{d}\theta\mathrm{d}\hat{s}\right\rvert_{\dot{C}^{0,\alpha}}
\le c(\kappa_*,c_\Gamma)\epsilon^{1-\alpha}\lVert X_{ss}\rVert_{C^{0,\alpha}(T)}\lVert\varphi\rVert_{L^\infty} \\
&\le c(\kappa_{*,\alpha},c_\Gamma)\epsilon^{1-\alpha}\lVert\varphi\rVert_{L^\infty}.
\end{aligned} ∫ − 1/2 1/2 ∫ − π π ∣ R t ∣ 3 s ^ 2 ( κ 1 e n 1 + κ 2 e n 2 ) ⋅ φ ϵ d θ d s ^ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ 1 − α ∥ X ss ∥ C 0 , α ( T ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ .
In particular, since κ 1 e n 1 + κ 2 e n 2 \kappa_1e_{n_1}+\kappa_2e_{n_2} κ 1 e n 1 + κ 2 e n 2 depends only on s s s , we do not lose an additional factor of ϵ − α \epsilon^{-\alpha} ϵ − α . The remaining expressions making up e z ⋅ ∂ s R S , 2 , 2 e_z\cdot\partial_s\mathcal{R}_{S,2,2} e z ⋅ ∂ s R S , 2 , 2 may be estimated using case (2) of Lemma 4.4. In total, we obtain
∣ e z ⋅ ∂ s R S , 2 , 2 ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ + c ( κ ∗ , c Γ ) ϵ max i ( ∥ Q i ∥ C 0 , α + + ∥ Q i ∥ C 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α . \begin{aligned}
\left\lvert e_z\cdot\partial_s\mathcal{R}_{S,2,2}\right\rvert_{\dot{C}^{0,\alpha}}
\le c(\kappa_{*,\alpha},c_\Gamma)\epsilon^{1-\alpha}\lVert\varphi\rVert_{L^\infty}
+c(\kappa_*,c_\Gamma)\epsilon\max_i\left(\lVert Q_i\rVert_{C^{0,\alpha_+} }+\lVert Q_i\rVert_{C^{0,\alpha}}\right)\lVert\varphi\rVert_{C^{0,\alpha}} \\
&\le c(\kappa_{*,\alpha_+},c_\Gamma)\epsilon^{1-\alpha_+}\lVert\varphi\rVert_{C^{0,\alpha}}.
\end{aligned} ∣ e z ⋅ ∂ s R S , 2 , 2 ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ + c ( κ ∗ , c Γ ) ϵ i max ( ∥ Q i ∥ C 0 , α + + ∥ Q i ∥ C 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α .
By similar arguments and estimate (81) , for two filaments with centerlines X ( a ) X^{(a)} X ( a ) , X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1, we may additionally obtain the Lipschitz estimate
∣ e z ⋅ ∂ s R S , 2 , 2 ( a ) − e z ⋅ ∂ s R S , 2 , 2 ( b ) ∣ C 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α . \left\lvert e_z\cdot\partial_s\mathcal{R}^{(a)}_{S,2,2}-e_z\cdot\partial_s\mathcal{R}^{(b)}_{S,2,2}\right\rvert_{C^{0,\alpha}}
\le c(\kappa^{(a)}_{*,\alpha_+},\kappa^{(b)}_{*,\alpha_+},c_\Gamma)\epsilon^{1-\alpha_+}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^{2,\alpha_+}}\lVert\varphi\rVert_{C^{0,\alpha}}. e z ⋅ ∂ s R S , 2 , 2 ( a ) − e z ⋅ ∂ s R S , 2 , 2 ( b ) C 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ φ ∥ C 0 , α .
We next consider ∂ s R S , 2 , 2 \partial_s\mathcal{R}_{S,2,2} ∂ s R S , 2 , 2 in directions normal to the filament centerline. It will be convenient to consider S 2 , 4 , n 1 S_{2,4,n_1} S 2 , 4 , n 1 and S 2 , 4 , n 2 S_{2,4,n_2} S 2 , 4 , n 2 (in the directions e n 1 e_{n_1} e n 1 and e n 2 e_{n_2} e n 2 ) instead of S 2 , 4 , r S_{2,4,r} S 2 , 4 , r and S 2 , 4 , θ S_{2,4,\theta} S 2 , 4 , θ to avoid introducing artificial θ \theta θ -dependence resulting in a loss of ϵ − α \epsilon^{-\alpha} ϵ − α . The expressions S 2 , 4 , n 1 S_{2,4,n_1} S 2 , 4 , n 1 and S 2 , 4 , n 2 S_{2,4,n_2} S 2 , 4 , n 2 are given by
S 2 , 4 , n 1 = ( s ^ κ 1 ( s ) − ϵ κ 3 sin ( θ ) ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ∣ R t ∣ 3 S_{2,4,n_1}=
\frac{\left(\hat{s}\kappa_1(s)-\epsilon\kappa_3\sin(\theta)\right)\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)\cdot\varphi}{\lvert R_t\rvert^3} S 2 , 4 , n 1 = ∣ R t ∣ 3 ( s ^ κ 1 ( s ) − ϵ κ 3 sin ( θ ) ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ
S 2 , 4 , n 2 = ( s ^ κ 2 ( s ) + ϵ κ 3 cos ( θ ) ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⋅ φ ∣ R t ∣ 3 . S_{2,4,n_2}=
\frac{\left(\hat{s}\kappa_2(s)+\epsilon\kappa_3\cos(\theta)\right)\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)\cdot\varphi}{\lvert R_t\rvert^3}. S 2 , 4 , n 2 = ∣ R t ∣ 3 ( s ^ κ 2 ( s ) + ϵ κ 3 cos ( θ ) ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⋅ φ .
Both expressions contain terms proportional to s ^ 2 ∣ R t ∣ 3 \frac{\hat{s}^2}{\lvert R_t\rvert^3} ∣ R t ∣ 3 s ^ 2 , s ^ ϵ sin ( θ ^ / 2 ) ∣ R t ∣ 3 \frac{\hat{s}\epsilon\sin(\hat{\theta}/2)}{\lvert R_t\rvert^3} ∣ R t ∣ 3 s ^ ϵ s i n ( θ ^ /2 ) , and terms which are higher order in ϵ \epsilon ϵ . For the terms of the first form, using case (1) of Lemma 4.4, we have
∣ ∫ − 1 / 2 1 / 2 ∫ − π π s ^ 2 κ i ( s ) e t ⋅ φ ∣ R t ∣ 3 ϵ d θ d s ^ ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ 1 − α ∥ X s s ∥ C 0 , α ( T ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ , \begin{aligned}
\left\lvert\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\hat{s}^{2}\kappa_i(s)e_t\cdot\varphi}{\lvert R_t\rvert^3}\epsilon\,\mathrm{d}\theta\mathrm{d}\hat{s}\right\rvert_{\dot{C}^{0,\alpha}}
\le c(\kappa_*,c_\Gamma)\epsilon^{1-\alpha}\lVert X_{ss}\rVert_{C^{0,\alpha}(T)}\lVert\varphi\rVert_{L^\infty} \\
&\le c(\kappa_{*,\alpha},c_\Gamma)\epsilon^{1-\alpha}\lVert\varphi\rVert_{L^\infty},
\end{aligned} ∫ − 1/2 1/2 ∫ − π π ∣ R t ∣ 3 s ^ 2 κ i ( s ) e t ⋅ φ ϵ d θ d s ^ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ 1 − α ∥ X ss ∥ C 0 , α ( T ) ∥ φ ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ φ ∥ L ∞ ,
since κ i ( s ) e t ( s ) \kappa_i(s)e_t(s) κ i ( s ) e t ( s ) , i = 1 , 2 i=1,2 i = 1 , 2 , depends only on s s s and not θ \theta θ or θ ^ \hat{\theta} θ ^ . In particular, we do not lose an additional factor of ϵ − α \epsilon^{-\alpha} ϵ − α from estimating these terms in C 0 , α C^{0,\alpha} C 0 , α . For terms proportional to s ^ ϵ sin ( θ ^ / 2 ) ∣ R t ∣ 3 \frac{\hat{s}\epsilon\sin(\hat{\theta}/2)}{\lvert R_t\rvert^3} ∣ R t ∣ 3 s ^ ϵ s i n ( θ ^ /2 ) , we treat sin ( θ ^ / 2 ) \sin(\hat{\theta}/2) sin ( θ ^ /2 ) as a coefficient, i.e. let Q ( s , s ^ , θ , θ ^ ) = 2 κ i ( s ) sin ( θ ^ / 2 ) e θ ( s , θ ) Q(s,\hat{s},\theta,\hat{\theta})=2\kappa_i(s)\sin(\hat{\theta}/2)e_\theta(s,\theta) Q ( s , s ^ , θ , θ ^ ) = 2 κ i ( s ) sin ( θ ^ /2 ) e θ ( s , θ ) , and use case (2) of Lemma 4.4. 4.4 to estimate
∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ϵ S ^ Q ⋅ φ ∣ R t ∣ 3 ϵ d θ d s ^ ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ ( ∥ Q ∥ C ˙ 1 0 , α + ∥ Q ∥ C ˙ 2 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α . (238) \begin{aligned}
\left|\left.\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\epsilon\widehat{S}Q\cdot\varphi}{\lvert R_t\rvert^3}\,\epsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}\right|_{\dot{C}^{0,\alpha}}\right. \le c(\kappa_{*},c_{\Gamma})\epsilon\left(\lVert Q\rVert_{\dot{C}^{0,\alpha}_1}+\lVert Q\rVert_{\dot{C}^{0,\alpha}_2}\right)\lVert\varphi\rVert_{C^{0,\alpha}} \\
&\le c(\kappa_{*},\alpha_{+},c_{\Gamma})\epsilon^{1-\alpha_{+}}\lVert\varphi\rVert_{C^{0,\alpha}} . \tag*{(238)}
\end{aligned} p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R t ∣ 3 ϵ S Q ⋅ φ ϵ d θ d s C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ϵ ( ∥ Q ∥ C ˙ 1 0 , α + ∥ Q ∥ C ˙ 2 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α . (238)
The remaining terms may be treated using case (2) of Lemma 4.4, yielding, for j ∈ { n 1 , n 2 } j\in\{n_1,n_2\} j ∈ { n 1 , n 2 } ,
∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π S 2 , 4 , j ϵ d θ d s ^ ∣ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α . \left|\left.\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}S_{2,4,j}\,\epsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}\right|_{C^{0,\alpha}}\right. \le c(\kappa_{*},\alpha_{+},c_{\Gamma})\epsilon^{1-\alpha_{+}}\lVert\varphi\rVert_{C^{0,\alpha}} . p.v. ∫ − 1/2 1/2 ∫ − π π S 2 , 4 , j ϵ d θ d s C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α .
In addition, using (213), (210), (227), and (81), by Lemma 4.4, case (2), we obtain
∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( S 2 , 4 , j ( a ) − S 2 , 4 , j ( b ) ) ϵ d θ d s ^ ∣ C 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ C 0 , α . \left|\left.\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(S_{2,4,j}^{(a)}-S_{2,4,j}^{(b)}\right)\epsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}\right|_{C^{0,\alpha}}\right. \le c(\kappa_{*,\alpha_{+}}^{(a)},\kappa_{*,\alpha_{+}}^{(b)},c_{\Gamma})\epsilon^{1-\alpha_{+}}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\alpha}}\lVert\varphi\rVert_{C^{0,\alpha}} . p.v. ∫ − 1/2 1/2 ∫ − π π ( S 2 , 4 , j ( a ) − S 2 , 4 , j ( b ) ) ϵ d θ d s C 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ C 0 , α .
We next use (196), (202), and (232) to expand the remaining terms in the e r e_r e r - and e θ e_\theta e θ -direction integrands as
S 2 , 5 , r = ϵ S ^ Q 0 , 5 ( e t ⋅ φ ) ∣ R t ∣ 3 + ( ϵ S ^ 2 Q 0 , 5 + 2 ϵ sin 2 ( θ ^ 2 ) ) ( − ϵ κ ^ e t + s ^ ( κ 1 ( s ) e n 1 ( s ) + κ 2 ( s ) e n 2 ( s ) ) + ϵ κ 3 e θ ) ⋅ φ ∣ R t ∣ 3 + ( 1 ∣ R t ∣ 3 − 1 ∣ R ‾ ∣ 3 ) 2 ϵ sin 2 ( θ ^ 2 ) ( e t ⋅ φ ) S_{2,5,r}=\frac{\epsilon\widehat{S}Q_{0,5}(e_t\cdot\varphi)}{\lvert R_t\rvert^3}+\frac{\left(\epsilon\widehat{S}^2Q_{0,5}+2\epsilon\sin^2\left(\frac{\widehat{\theta}}{2}\right)\right)\left(-\epsilon\widehat{\kappa}e_t+\widehat{s}\left(\kappa_1(s)e_{n_1}(s)+\kappa_2(s)e_{n_2}(s)\right)+\epsilon\kappa_3e_\theta\right)\cdot\varphi}{\lvert R_t\rvert^3}
+\left(\frac{1}{\lvert R_t\rvert^3}-\frac{1}{\lvert\overline{R}\rvert^3}\right)2\epsilon\sin^2\left(\frac{\widehat{\theta}}{2}\right)(e_t\cdot\varphi) S 2 , 5 , r = ∣ R t ∣ 3 ϵ S Q 0 , 5 ( e t ⋅ φ ) + ∣ R t ∣ 3 ( ϵ S 2 Q 0 , 5 + 2 ϵ sin 2 ( 2 θ ) ) ( − ϵ κ e t + s ( κ 1 ( s ) e n 1 ( s ) + κ 2 ( s ) e n 2 ( s ) ) + ϵ κ 3 e θ ) ⋅ φ + ( ∣ R t ∣ 3 1 − ∣ R ∣ 3 1 ) 2 ϵ sin 2 ( 2 θ ) ( e t ⋅ φ )
S 2 , 6 , r = − 3 ϵ s ^ [ s ^ Q 0 , 5 ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) + 2 ϵ sin 2 ( θ ^ 2 ) Q r ] ⋅ φ ( s ^ + ϵ s ^ Q 5 , 5 − ϵ 2 κ 3 sin ( θ ^ ) ) ∣ R t ∣ 5 − 6 ϵ 2 sin 2 ( θ ^ 2 ) ( s ^ Q 5 , 5 − ϵ κ 3 sin ( θ ^ ) ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ ∣ R t ∣ 5 − ( 1 ∣ R t ∣ 5 − 1 ∣ R ‾ ∣ 5 ) 6 s ^ ϵ sin 2 ( θ ^ 2 ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ S_{2,6,r}=-\frac{3\epsilon\widehat{s}\left[\widehat{s}Q_{0,5}\left(\widehat{s}e_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta+\epsilon\widehat{s}Q_r\right)+2\epsilon\sin^2\left(\frac{\widehat{\theta}}{2}\right)Q_r\right]\cdot\varphi\left(\widehat{s}+\epsilon\widehat{s}Q_{5,5}-\epsilon^2\kappa_3\sin(\widehat{\theta})\right)}{\lvert R_t\rvert^5}
-\frac{6\epsilon^2\sin^2\left(\frac{\widehat{\theta}}{2}\right)\left(\widehat{s}Q_{5,5}-\epsilon\kappa_3\sin(\widehat{\theta})\right)\left(\widehat{s}e_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\cdot\varphi}{\lvert R_t\rvert^5}
- \left(\frac{1}{\lvert R_t\rvert^5}-\frac{1}{\lvert\overline{R}\rvert^5}\right)6\widehat{s}\epsilon\sin^2\left(\frac{\widehat{\theta}}{2}\right)\left(\widehat{s}e_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\cdot\varphi S 2 , 6 , r = − ∣ R t ∣ 5 3 ϵ s [ s Q 0 , 5 ( s e t + 2 ϵ sin ( 2 θ ) e θ + ϵ s Q r ) + 2 ϵ sin 2 ( 2 θ ) Q r ] ⋅ φ ( s + ϵ s Q 5 , 5 − ϵ 2 κ 3 sin ( θ ) ) − ∣ R t ∣ 5 6 ϵ 2 sin 2 ( 2 θ ) ( s Q 5 , 5 − ϵ κ 3 sin ( θ ) ) ( s e t + 2 ϵ sin ( 2 θ ) e θ ) ⋅ φ − ( ∣ R t ∣ 5 1 − ∣ R ∣ 5 1 ) 6 s ϵ sin 2 ( 2 θ ) ( s e t + 2 ϵ sin ( 2 θ ) e θ ) ⋅ φ
S 2 , 5 , θ = ϵ s ^ ( Q r ⋅ e θ ) e t ⋅ φ ∣ R t ∣ 3 + ϵ ( s ^ ( Q r ⋅ e θ ) + sin ( θ ^ ) ) ( − ϵ κ ^ e t + s ^ ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ ∣ R t ∣ 3 + ( 1 ∣ R t ∣ 3 − 1 ∣ R ‾ ∣ 3 ) ϵ sin ( θ ^ ) ( e t ⋅ φ ) S_{2,5,\theta}=\frac{\epsilon\widehat{s}(Q_r\cdot e_\theta)e_t\cdot\varphi}{\lvert R_t\rvert^3}+\frac{\epsilon\left(\widehat{s}(Q_r\cdot e_\theta)+\sin(\widehat{\theta})\right)\left(-\epsilon\widehat{\kappa}e_t+\widehat{s}\left(\kappa_1e_{n_1}+\kappa_2e_{n_2}\right)+\epsilon\kappa_3e_\theta\right)\cdot\varphi}{\lvert R_t\rvert^3}
+\left(\frac{1}{\lvert R_t\rvert^3}-\frac{1}{\lvert\overline{R}\rvert^3}\right)\epsilon\sin(\widehat{\theta})(e_t\cdot\varphi) S 2 , 5 , θ = ∣ R t ∣ 3 ϵ s ( Q r ⋅ e θ ) e t ⋅ φ + ∣ R t ∣ 3 ϵ ( s ( Q r ⋅ e θ ) + sin ( θ ) ) ( − ϵ κ e t + s ( κ 1 e n 1 + κ 2 e n 2 ) + ϵ κ 3 e θ ) ⋅ φ + ( ∣ R t ∣ 3 1 − ∣ R ∣ 3 1 ) ϵ sin ( θ ) ( e t ⋅ φ )
S 2 , 6 , θ = − 3 ϵ s ^ [ ( Q r ⋅ e θ ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) + ϵ sin ( θ ^ ) Q r ] ⋅ φ ( s ^ + ϵ s ^ Q 5 , 5 − ϵ 2 κ 3 sin ( θ ^ ) ) ∣ R t ∣ 5 − 3 ϵ 2 sin ( θ ^ ) ( s ^ Q 5 , 5 − ϵ κ 3 sin ( θ ^ ) ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ ∣ R t ∣ 5 − 3 ( 1 ∣ R t ∣ 5 − 1 ∣ R ‾ ∣ 5 ) ϵ sin ( θ ^ ) s ^ ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ . S_{2,6,\theta}=-\frac{3\epsilon\widehat{s}\left[(Q_r\cdot e_\theta)\left(\widehat{s}e_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta+\epsilon\widehat{s}Q_r\right)+\epsilon\sin(\widehat{\theta})Q_r\right]\cdot\varphi\left(\widehat{s}+\epsilon\widehat{s}Q_{5,5}-\epsilon^2\kappa_3\sin(\widehat{\theta})\right)}{\lvert R_t\rvert^5}
-\frac{3\epsilon^2\sin(\widehat{\theta})\left(\widehat{s}Q_{5,5}-\epsilon\kappa_3\sin(\widehat{\theta})\right)\left(\widehat{s}e_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\cdot\varphi}{\lvert R_t\rvert^5}
-3\left(\frac{1}{\lvert R_t\rvert^5}-\frac{1}{\lvert\overline{R}\rvert^5}\right)\epsilon\sin(\widehat{\theta})\widehat{s}\left(\widehat{s}e_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\cdot\varphi. S 2 , 6 , θ = − ∣ R t ∣ 5 3 ϵ s [ ( Q r ⋅ e θ ) ( s e t + 2 ϵ sin ( 2 θ ) e θ + ϵ s Q r ) + ϵ sin ( θ ) Q r ] ⋅ φ ( s + ϵ s Q 5 , 5 − ϵ 2 κ 3 sin ( θ ) ) − ∣ R t ∣ 5 3 ϵ 2 sin ( θ ) ( s Q 5 , 5 − ϵ κ 3 sin ( θ ) ) ( s e t + 2 ϵ sin ( 2 θ ) e θ ) ⋅ φ − 3 ( ∣ R t ∣ 5 1 − ∣ R ∣ 5 1 ) ϵ sin ( θ ) s ( s e t + 2 ϵ sin ( 2 θ ) e θ ) ⋅ φ .
The second and third terms in S 2 , 5 , r S_{2,5,r} S 2 , 5 , r and S 2 , 6 , r S_{2,6,r} S 2 , 6 , r both contain components that must be treated as in (237) to avoid loss of ϵ − α \epsilon^{-\alpha} ϵ − α . In total, using case (2) of Lemma 4.4, for both j = r j=r j = r and j = θ j=\theta j = θ we obtain the bound
∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( S 2 , 5 , j + S 2 , 6 , j ) ϵ d θ d s ^ ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α . \left|\left.\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(S_{2,5,j}+S_{2,6,j}\right)\epsilon\,\mathrm{d}\theta\mathrm{d}\widehat{s}\right|_{\dot{C}^{0,\alpha}}\right. \le c(\kappa_{*,\alpha_{+}},c_{\Gamma})\epsilon^{1-\alpha_{+}}\lVert\varphi\rVert_{C^{0,\alpha}} . p.v. ∫ − 1/2 1/2 ∫ − π π ( S 2 , 5 , j + S 2 , 6 , j ) ϵ d θ d s C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α .
∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( S 2 , 5 , j ( a ) − S 2 , 5 , j ( b ) + S 2 , 6 , j ( a ) − S 2 , 6 , j ( b ) ) ϵ d θ d b ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α . \left|\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(S_{2,5,j}^{(a)}-S_{2,5,j}^{(b)}+S_{2,6,j}^{(a)}-S_{2,6,j}^{(b)}\right)\epsilon\,\mathrm{d}\theta\mathrm{d}b\right|_{\dot{C}^{0,\alpha}}
\leq c\left(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_{\Gamma}\right)\epsilon^{1-\alpha^+}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha^+}}\|\varphi\|_{C^{0,\alpha}}. p.v. ∫ − 1/2 1/2 ∫ − π π ( S 2 , 5 , j ( a ) − S 2 , 5 , j ( b ) + S 2 , 6 , j ( a ) − S 2 , 6 , j ( b ) ) ϵ d θ d b C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ φ ∥ C 0 , α .
Altogether, for α + \alpha^+ α + satisfying α < α + < 1 \alpha<\alpha^+<1 α < α + < 1 , we have
∣ ∂ s R S , 2 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α , \left|\partial_s\mathcal{R}_{S,2,2}[\varphi]\right|_{\dot{C}^{0,\alpha}}\leq c(\kappa_{*,\alpha^+},c_{\Gamma})\epsilon^{1-\alpha^+}\|\varphi\|_{C^{0,\alpha}}, ∣ ∂ s R S , 2 , 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α ,
∣ ∂ s R S , 2 , 2 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 2 , 2 ( b ) [ φ 0 Φ ( b ) ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α . (239) \left|\partial_s\mathcal{R}_{S,2,2}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\partial_s\mathcal{R}_{S,2,2}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right|_{\dot{C}^{0,\alpha}}\leq c\left(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_{\Gamma}\right)\epsilon^{1-\alpha^+}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha^+}}\|\varphi\|_{C^{0,\alpha}}. \tag*{(239)} ∂ s R S , 2 , 2 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 2 , 2 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ φ ∥ C 0 , α . (239)
Finally, we derive bounds for ∂ s R S , 3 \partial_s\mathcal{R}_{S,3} ∂ s R S , 3 , given by
∂ s R S , 3 [ φ ] = − Φ − 1 p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( ∂ s G ) φ ϵ 2 κ ^ d θ d s ^ − [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π G φ ϵ 2 κ ^ d θ d s ^ . \partial_s\mathcal{R}_{S,3}[\varphi]=-\Phi^{-1}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}(\partial_s\mathcal{G})\varphi\,\epsilon^2\widehat{\kappa}\,\mathrm{d}\theta\mathrm{d}\widehat{s}-[\partial_s,\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\mathcal{G}\varphi\,\epsilon^2\widehat{\kappa}\,\mathrm{d}\theta\mathrm{d}\widehat{s}. ∂ s R S , 3 [ φ ] = − Φ − 1 p.v. ∫ − 1/2 1/2 ∫ − π π ( ∂ s G ) φ ϵ 2 κ d θ d s − [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π G φ ϵ 2 κ d θ d s .
We may write the integrand of the first term of ∂ s R S , 3 \partial_s\mathcal{R}_{S,3} ∂ s R S , 3 as
( ∂ s G ) φ ϵ 2 κ ^ = ( − ( ∂ s R ⋅ R ) φ ∣ R ∣ 3 + ∂ s R ( R ⋅ φ ) ∣ R ∣ 3 + R ( ∂ s R ⋅ φ ) ∣ R ∣ 3 − 3 R ( R ⋅ φ ) ( ∂ s R ⋅ R ) ∣ R ∣ 5 ) ϵ 2 κ ^ = ( − ( s ^ ( 1 − ϵ κ ^ ) e t ⋅ Q ~ t + 2 s ^ ϵ 2 κ 3 e θ ⋅ Q ~ t + ϵ 2 κ 3 sin ( θ ^ ) ) φ ∣ R ∣ 3 \begin{aligned}
(\partial_s\mathcal{G})\varphi\,\epsilon^2\widehat{\kappa}=\left(-\frac{(\partial_sR\cdot R)\varphi}{\lvert R\rvert^3}+\frac{\partial_sR(R\cdot\varphi)}{\lvert R\rvert^3}+\frac{R(\partial_sR\cdot\varphi)}{\lvert R\rvert^3}-3\frac{R(R\cdot\varphi)(\partial_sR\cdot R)}{\lvert R\rvert^5}\right)\epsilon^2\widehat{\kappa} \\
&=\left(-\frac{\left(\widehat{s}(1-\epsilon\widehat{\kappa})e_t\cdot\widetilde{Q}_t+2\widehat{s}\epsilon^2\kappa_3e_\theta\cdot\widetilde{Q}_t+\epsilon^2\kappa_3\sin(\widehat{\theta})\right)\varphi}{\lvert R\rvert^3}
\right.
\end{aligned} ( ∂ s G ) φ ϵ 2 κ = ( − ∣ R ∣ 3 ( ∂ s R ⋅ R ) φ + ∣ R ∣ 3 ∂ s R ( R ⋅ φ ) + ∣ R ∣ 3 R ( ∂ s R ⋅ φ ) − 3 ∣ R ∣ 5 R ( R ⋅ φ ) ( ∂ s R ⋅ R ) ) ϵ 2 κ = − ∣ R ∣ 3 ( s ( 1 − ϵ κ ) e t ⋅ Q t + 2 s ϵ 2 κ 3 e θ ⋅ Q t + ϵ 2 κ 3 sin ( θ ) ) φ
+ ( ( 1 − ϵ κ ^ ) e t + ϵ κ 3 e θ ) ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ ∣ R ∣ 3 + ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ ) ( ( 1 − ϵ κ ^ ) e t + ϵ κ 3 e θ ) ⋅ φ ∣ R ∣ 3 \left.
+\frac{\left((1-\epsilon\widehat{\kappa})e_t+\epsilon\kappa_3e_\theta\right)\left(\widehat{s}\widetilde{Q}_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\cdot\varphi}{\lvert R\rvert^3}
+\frac{\left(\widehat{s}\widetilde{Q}_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\left((1-\epsilon\widehat{\kappa})e_t+\epsilon\kappa_3e_\theta\right)\cdot\varphi}{\lvert R\rvert^3}
\right. + ∣ R ∣ 3 ( ( 1 − ϵ κ ) e t + ϵ κ 3 e θ ) ( s Q t + 2 ϵ sin ( 2 θ ) e θ ) ⋅ φ + ∣ R ∣ 3 ( s Q t + 2 ϵ sin ( 2 θ ) e θ ) ( ( 1 − ϵ κ ) e t + ϵ κ 3 e θ ) ⋅ φ
− 3 ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ ) ( s ^ Q ~ t + 2 ϵ sin ( θ ^ 2 ) e θ ) ⋅ φ ( s ^ ( 1 − ϵ κ ^ ) e t ⋅ Q ~ t + 2 s ^ ϵ 2 κ 3 e θ ⋅ Q ~ t + ϵ 2 κ 3 sin ( θ ^ ) ) ∣ R ∣ 5 ) ϵ 2 κ ^ . \left.
-3\frac{\left(\widehat{s}\widetilde{Q}_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\left(\widehat{s}\widetilde{Q}_t+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\right)\cdot\varphi\left(\widehat{s}(1-\epsilon\widehat{\kappa})e_t\cdot\widetilde{Q}_t+2\widehat{s}\epsilon^2\kappa_3e_\theta\cdot\widetilde{Q}_t+\epsilon^2\kappa_3\sin(\widehat{\theta})\right)}{\lvert R\rvert^5}\right)\epsilon^2\widehat{\kappa}. − 3 ∣ R ∣ 5 ( s Q t + 2 ϵ sin ( 2 θ ) e θ ) ( s Q t + 2 ϵ sin ( 2 θ ) e θ ) ⋅ φ ( s ( 1 − ϵ κ ) e t ⋅ Q t + 2 s ϵ 2 κ 3 e θ ⋅ Q t + ϵ 2 κ 3 sin ( θ ) ) ϵ 2 κ .
Noting the additional factor of ϵ \epsilon ϵ due to (8), by Lemma 4.4, case (2), we then have
∣ Φ − 1 p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( ∂ s G ) φ ϵ 2 κ ^ d θ d s ^ ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α . \left|\Phi^{-1}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}(\partial_s\mathcal{G})\varphi\,\epsilon^2\widehat{\kappa}\,\mathrm{d}\theta\mathrm{d}\widehat{s}\right|_{\dot{C}^{0,\alpha}}\leq c(\kappa_{*,\alpha^+},c_{\Gamma})\epsilon^{1-\alpha^+}\|\varphi\|_{C^{0,\alpha}}. Φ − 1 p.v. ∫ − 1/2 1/2 ∫ − π π ( ∂ s G ) φ ϵ 2 κ d θ d s C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α .
Using the commutator bound (78) to estimate the second term of ∂ s R S , 3 \partial_s\mathcal{R}_{S,3} ∂ s R S , 3 , we then have
∣ ∂ s R S , 3 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α , \left|\partial_s\mathcal{R}_{S,3}[\varphi]\right|_{\dot{C}^{0,\alpha}}\leq c(\kappa_{*,\alpha^+},c_{\Gamma})\epsilon^{1-\alpha^+}\|\varphi\|_{C^{0,\alpha}}, ∣ ∂ s R S , 3 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ φ ∥ C 0 , α ,
∣ ∂ s R S , 3 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 3 ( b ) [ φ 0 Φ ( b ) ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 0 , α . (240) \left|\partial_s\mathcal{R}_{S,3}^{(a)}[\varphi_0^{\Phi^{(a)}}]-\partial_s\mathcal{R}_{S,3}^{(b)}[\varphi_0^{\Phi^{(b)}}]\right|_{\dot{C}^{0,\alpha}}\leq c\left(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_{\Gamma}\right)\epsilon^{1-\alpha^+}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha^+}}\|\varphi\|_{C^{0,\alpha}}. \tag*{(240)} ∂ s R S , 3 ( a ) [ φ 0 Φ ( a ) ] − ∂ s R S , 3 ( b ) [ φ 0 Φ ( b ) ] C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ φ ∥ C 0 , α . (240)
We may now put everything together. Define
R S , + [ φ ] = R S , 0 [ φ ] + R S , 1 , 1 [ φ ] + R S , 2 , 1 [ φ ] , \mathcal{R}_{S,+}[\varphi]=\mathcal{R}_{S,0}[\varphi]+\mathcal{R}_{S,1,1}[\varphi]+\mathcal{R}_{S,2,1}[\varphi], R S , + [ φ ] = R S , 0 [ φ ] + R S , 1 , 1 [ φ ] + R S , 2 , 1 [ φ ] ,
R S , ϵ [ φ ] = R S , 1 , 2 [ φ ] + R S , 2 , 2 [ φ ] + R S , 3 [ φ ] . \mathcal{R}_{S,\epsilon}[\varphi]=\mathcal{R}_{S,1,2}[\varphi]+\mathcal{R}_{S,2,2}[\varphi]+\mathcal{R}_{S,3}[\varphi]. R S , ϵ [ φ ] = R S , 1 , 2 [ φ ] + R S , 2 , 2 [ φ ] + R S , 3 [ φ ] .
Then, combining the bounds (225) , (229) , (230) , (231) , (234) , (235) , (237) , we obtain the estimates (67) and (68) for R S , + \mathcal{R}_{S,+} R S , + . Additionally, combining the bounds (229) , (230) , (231) , (236) , (239) , and (240) , we obtain the estimates (67) and (68) for R S , ϵ \mathcal{R}_{S,\epsilon} R S , ϵ . In total, we obtain Lemma 1.10. □ \square □
Double layer remainder terms
We next prove Lemma 1.11 regarding the mapping properties of the double layer remainder R D \mathcal{R}_D R D , defined in (65) . We again restate the lemma here for convenience.
Lemma 4.6 (Double layer remainder). Let 0 < γ < β < 1 0 < \gamma< \beta< 1 0 < γ < β < 1 and consider a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) \mathbf{X}(s) \in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Let the double layer remainder R D \mathcal{R}_{D} R D be as defined in (65) . Given ψ ∈ C 0 , γ ( Γ ϵ ) \psi\in C^{0,\gamma}(\Gamma_\epsilon) ψ ∈ C 0 , γ ( Γ ϵ ) , we have that R D \mathcal{R}_{D} R D satisfies
∥ R D [ ψ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ (241) \left\|\mathcal{R}_{D}[\psi]\right\|_{C^{1,\gamma}} \leq c(\kappa_{*,\gamma^{+}},c_{\Gamma})\epsilon^{-\gamma^{+}}\|\psi\|_{C^{0,\gamma}} \tag*{(241)} ∥ R D [ ψ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ (241) for any γ + ∈ ( γ , β ] \gamma^{+} \in(\gamma,\beta] γ + ∈ ( γ , β ] . Furthermore, given two nearby filaments with centerlines X ( a ) ( s ) \mathbf{X}^{(a)}(s) X ( a ) ( s ) , X ( b ) ( s ) \mathbf{X}^{(b)}(s) X ( b ) ( s ) satisfying Lemma 1.1 , the difference between their corresponding double layer remainders R D ( a ) \mathcal{R}_{D}^{(a)} R D ( a ) and R D ( b ) \mathcal{R}_{D}^{(b)} R D ( b ) satisfies
∥ R D ( a ) [ ψ ] − R D ( b ) [ ψ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (242) \left\|\mathcal{R}_{D}^{(a)}[\psi]-\mathcal{R}_{D}^{(b)}[\psi]\right\|_{C^{1,\gamma}} \leq c(\kappa_{*,\gamma^{+}}^{(a)},\kappa_{*,\gamma^{+}}^{(b)},c_{\Gamma})\epsilon^{-\gamma^{+}}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\gamma^{+}}}\|\psi\|_{C^{0,\gamma}}. \tag*{(242)} R D ( a ) [ ψ ] − R D ( b ) [ ψ ] C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + X ( a ) − X ( b ) C 2 , γ + ∥ ψ ∥ C 0 , γ . (242)
Proof of Lemma 1.11 . Note that although we do not actually need an estimate for the full C 1 , γ ( Γ ϵ ) C^{1,\gamma}(\Gamma_\epsilon) C 1 , γ ( Γ ϵ ) norm of R D \mathcal{R}_{D} R D , as C s 1 , γ C^{1,\gamma}_{s} C s 1 , γ will suffice, we show the full bound here because the double layer estimates will be used later on in bounding the full Neumann data w \mathbf{w} w (Lemma 1.13 ).
We begin by introducing some notation. Given a vector-valued density ψ ∈ C 0 , γ ( Γ ϵ ) \psi\in C^{0,\gamma}(\Gamma_\epsilon) ψ ∈ C 0 , γ ( Γ ϵ ) and matrix-valued kernel H \mathbf{H} H defined along Γ ϵ \Gamma_\epsilon Γ ϵ , we denote
W ( ψ ; H ) = Φ − 1 p.v. ∫ − 1 / 2 1 / 2 ∫ − π π H ( s , s ^ , θ , θ ^ ) ψ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π H ‾ ( s , s ^ , θ , θ ^ ) ( Φ − 1 ψ ) ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ , (243) \begin{aligned}
W(\psi;\mathbf{H})={}&\Phi^{-1}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\mathbf{H}(s,\hat{s},\theta,\hat{\theta})\psi(s-\hat{s},\theta-\hat{\theta})\epsilon\,\mathrm{d}\hat{\theta}\,\mathrm{d}\hat{s}\\
&-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\overline{\mathbf{H}}(s,\hat{s},\theta,\hat{\theta})(\Phi^{-1}\psi)(s-\hat{s},\theta-\hat{\theta})\epsilon\,\mathrm{d}\hat{\theta}\,\mathrm{d}\hat{s}, \tag*{(243)}
\end{aligned} W ( ψ ; H ) = Φ − 1 p.v. ∫ − 1/2 1/2 ∫ − π π H ( s , s ^ , θ , θ ^ ) ψ ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ − p.v. ∫ − 1/2 1/2 ∫ − π π H ( s , s ^ , θ , θ ^ ) ( Φ − 1 ψ ) ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ , (243)
where H ‾ \overline{\mathbf{H}} H is the corresponding matrix kernel defined along the straight filament ∂ C ϵ \partial C_\epsilon ∂ C ϵ . Noting that W ( ψ ; H ) W(\psi;\mathbf{H}) W ( ψ ; H ) lives on straight basis vectors, we use W j W_j W j to denote the scalar-valued kernel resulting from dotting W ( ψ ; H ) W(\psi;\mathbf{H}) W ( ψ ; H ) with the straight basis vector e ‾ j \overline{\mathbf{e}}_j e j , j ∈ { t , r , θ } j \in\{t,r,\theta\} j ∈ { t , r , θ } :
W ( ψ ; H ) ⋅ e ‾ j = p.v. ∫ − 1 / 2 1 / 2 ∫ − π π W j ( ψ ; H ) ϵ d θ ^ d s ^ ; W j ( ψ ; H ) : = ( H ψ ) ⋅ e j ( s ) − e ‾ j ⋅ ( H Φ − 1 ψ ) . (244) \begin{aligned}
W(\psi;\mathbf{H})\cdot\overline{\mathbf{e}}_j&=\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}W_j(\psi;\mathbf{H})\epsilon\,\mathrm{d}\hat{\theta}\,\mathrm{d}\hat{s};\\
W_j(\psi;\mathbf{H})&:=(\mathbf{H}\psi)\cdot\mathbf{e}_j(s)-\overline{\mathbf{e}}_j\cdot(\mathbf{H}\Phi^{-1}\psi). \tag*{(244)}
\end{aligned} W ( ψ ; H ) ⋅ e j W j ( ψ ; H ) = p.v. ∫ − 1/2 1/2 ∫ − π π W j ( ψ ; H ) ϵ d θ ^ d s ^ ; := ( H ψ ) ⋅ e j ( s ) − e j ⋅ ( H Φ − 1 ψ ) . (244)
We again note that e ‾ t = e z \overline{\mathbf{e}}_t=\mathbf{e}_z e t = e z , but here we will use the overline notation for convenience.
Recalling the definition of R D \mathcal{R}_{D} R D from (65) and rewriting the integrands in terms of ( s , s ^ , θ , θ ^ ) (s,\hat{s},\theta,\hat{\theta}) ( s , s ^ , θ , θ ^ ) , we may write
R D [ ψ ( s , θ ) ] = Φ − 1 D [ ψ ( s , θ ) ] − D ‾ [ ( Φ − 1 ψ ) ( s , θ ) ] = R D , 0 [ ψ ] + R D , 1 [ ψ ] + R D , 2 [ ψ ] , R D , 0 [ ψ ] = − ( ∫ − ∞ − 1 / 2 + ∫ 1 / 2 ∞ ) ∫ − π π K ‾ D ( s , s ^ , θ , θ ^ ) ( Φ − 1 ψ ) ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ , R D , 1 [ ψ ] = W ( ψ ; K D ) , R D , 2 [ ψ ] = − Φ − 1 ∫ − 1 / 2 1 / 2 ∫ − π π K D ( s , s ^ , θ , θ ^ ) ψ ( s − s ^ , θ − θ ^ ) ϵ 2 κ ^ d θ ^ d s ^ . \begin{aligned}
\mathcal{R}_{D}[\psi(s,\theta)]&=\Phi^{-1}\mathcal{D}[\psi(s,\theta)]-\overline{\mathcal{D}}[(\Phi^{-1}\psi)(s,\theta)]\\
&=\mathcal{R}_{D,0}[\psi]+\mathcal{R}_{D,1}[\psi]+\mathcal{R}_{D,2}[\psi],\\
\mathcal{R}_{D,0}[\psi]&=-\left(\int_{-\infty}^{-1/2}+\int_{1/2}^{\infty}\right)\int_{-\pi}^{\pi}\overline{\mathbf{K}}_{D}(s,\hat{s},\theta,\hat{\theta})(\Phi^{-1}\psi)(s-\hat{s},\theta-\hat{\theta})\epsilon\,\mathrm{d}\hat{\theta}\,\mathrm{d}\hat{s},\\
\mathcal{R}_{D,1}[\psi]&=W(\psi;\mathbf{K}_{D}),\\
\mathcal{R}_{D,2}[\psi]&=-\Phi^{-1}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\mathbf{K}_{D}(s,\hat{s},\theta,\hat{\theta})\psi(s-\hat{s},\theta-\hat{\theta})\epsilon^{2}\hat{\kappa}\,\mathrm{d}\hat{\theta}\,\mathrm{d}\hat{s}.
\end{aligned} R D [ ψ ( s , θ )] R D , 0 [ ψ ] R D , 1 [ ψ ] R D , 2 [ ψ ] = Φ − 1 D [ ψ ( s , θ )] − D [( Φ − 1 ψ ) ( s , θ )] = R D , 0 [ ψ ] + R D , 1 [ ψ ] + R D , 2 [ ψ ] , = − ( ∫ − ∞ − 1/2 + ∫ 1/2 ∞ ) ∫ − π π K D ( s , s ^ , θ , θ ^ ) ( Φ − 1 ψ ) ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ , = W ( ψ ; K D ) , = − Φ − 1 ∫ − 1/2 1/2 ∫ − π π K D ( s , s ^ , θ , θ ^ ) ψ ( s − s ^ , θ − θ ^ ) ϵ 2 κ ^ d θ ^ d s ^ .
Here the form of R D , 1 \mathcal{R}_{D,1} R D , 1 is as defined in (243) and the terms R D , 0 \mathcal{R}_{D,0} R D , 0 and R D , 2 \mathcal{R}_{D,2} R D , 2 arise due to the integration limits in the straight setting and the Jacobian factor (8) , respectively.
As we did for the single layer, we begin with bounds for the smooth remainder R D , 0 \mathcal{R}_{D,0} R D , 0 away from the singularity at ( s ^ , θ ^ ) = ( 0 , 0 ) (\hat{s},\hat{\theta})=(0,0) ( s ^ , θ ^ ) = ( 0 , 0 ) . Using the form (38) of the straight kernel K ‾ D \overline{\mathbf{K}}_{D} K D and the expression (196) for R ‾ \overline{R} R , we have
∥ R D , 0 [ ψ ] ∥ C 2 ≤ c ϵ ∥ ψ ∥ L ∞ ∫ 1 / 2 ∞ ( 1 ∣ s ‾ ∣ 2 + 1 ∣ s ‾ ∣ 3 + 1 ∣ s ‾ ∣ 4 ) d s ≤ c ϵ ∥ ψ ∥ L ∞ . (245) \|\mathcal{R}_{D,0}[\psi]\|_{C^2} \le c\epsilon\|\psi\|_{L^\infty}\int_{1/2}^{\infty}\left(\frac{1}{|\overline{s}|^2}+\frac{1}{|\overline{s}|^3}+\frac{1}{|\overline{s}|^4}\right)\,\mathrm{d}s \le c\epsilon\|\psi\|_{L^\infty}. \tag*{(245)} ∥ R D , 0 [ ψ ] ∥ C 2 ≤ cϵ ∥ ψ ∥ L ∞ ∫ 1/2 ∞ ( ∣ s ∣ 2 1 + ∣ s ∣ 3 1 + ∣ s ∣ 4 1 ) d s ≤ cϵ ∥ ψ ∥ L ∞ . (245)
In addition, given two nearby curves X ( a ) \mathbf{X}^{(a)} X ( a ) , X ( b ) \mathbf{X}^{(b)} X ( b ) satisfying Lemma 1.1, the corresponding remainders R D , 0 ( a ) \mathcal{R}_{D,0}^{(a)} R D , 0 ( a ) and R D , 0 ( b ) \mathcal{R}_{D,0}^{(b)} R D , 0 ( b ) are nearly identical except for the map Φ \Phi Φ identifying the tangential and normal components of ψ \psi ψ with straight basis vectors. In particular, we may estimate
∥ R D , 0 ( a ) [ ψ ] − R D , 0 ( b ) [ ψ ] ∥ C 2 = ∥ ( ∫ − ∞ − 1 / 2 + ∫ 1 / 2 ∞ ) ∫ − π π K ‾ D ( ( Φ ( a ) ) − 1 ψ − ( Φ ( b ) ) − 1 ψ ) ϵ d θ ^ d s ^ ∥ C 2 ≤ c ϵ ∥ ( Φ ( a ) ) − 1 ψ − ( Φ ( b ) ) − 1 ψ ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ ψ ∥ L ∞ , (246) \begin{aligned}
\|\mathcal{R}_{D,0}^{(a)}[\psi]-\mathcal{R}_{D,0}^{(b)}[\psi]\|_{C^2}
&=\left\|\left(\int_{-\infty}^{-1/2}+\int_{1/2}^{\infty}\right)\int_{-\pi}^{\pi}\overline{K}_D\left((\Phi^{(a)})^{-1}\psi-(\Phi^{(b)})^{-1}\psi\right)\epsilon\mathrm{d}\widehat{\theta}\mathrm{d}\widehat{s}\right\|_{C^2}\\
&\le c\epsilon\|(\Phi^{(a)})^{-1}\psi-(\Phi^{(b)})^{-1}\psi\|_{L^\infty}\\
&\le c(\kappa_*^{(a)},\kappa_*^{(b)})\epsilon\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|\psi\|_{L^\infty}, \tag*{(246)}
\end{aligned} ∥ R D , 0 ( a ) [ ψ ] − R D , 0 ( b ) [ ψ ] ∥ C 2 = ( ∫ − ∞ − 1/2 + ∫ 1/2 ∞ ) ∫ − π π K D ( ( Φ ( a ) ) − 1 ψ − ( Φ ( b ) ) − 1 ψ ) ϵ d θ d s C 2 ≤ cϵ ∥ ( Φ ( a ) ) − 1 ψ − ( Φ ( b ) ) − 1 ψ ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ ψ ∥ L ∞ , (246)
by (79).
For the remainders R D , 1 \mathcal{R}_{D,1} R D , 1 and R D , 2 \mathcal{R}_{D,2} R D , 2 , we start with L ∞ L^\infty L ∞ bounds. Using the form (38) of K D K_D K D , we may write the integrands W j W_j W j given by (243) as
W j ( ψ ; K D ) = 3 4 π ( e j ( s ) ⋅ R ∣ R ∣ 5 ( R ⋅ n ′ ) ( R ⋅ ψ − R ‾ ⋅ ( Φ − 1 ψ ) ) + e j ( s ) ⋅ R ∣ R ∣ 5 ( R ⋅ n ′ − R ‾ ⋅ n ‾ ′ ) R ‾ ⋅ ( Φ − 1 ψ ) + ( e j ( s ) ⋅ R ) − ( e ‾ j ⋅ R ‾ ) ∣ R ∣ 5 ( R ‾ ⋅ n ‾ ′ ) R ‾ ⋅ ( Φ − 1 ψ ) + ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) ( e ‾ j ⋅ R ‾ ) ( R ‾ ⋅ n ‾ ′ ) R ‾ ⋅ ( Φ − 1 ψ ) ) . \begin{aligned}
W_j(\psi;K_D)={}&\frac{3}{4\pi}\left(\frac{\mathbf{e}_j(s)\cdot\mathbf{R}}{|\mathbf{R}|^5}(\mathbf{R}\cdot\mathbf{n}'\!)(\mathbf{R}\cdot\psi-\overline{\mathbf{R}}\cdot(\Phi^{-1}\psi))\right.\\
&+\frac{\mathbf{e}_j(s)\cdot\mathbf{R}}{|\mathbf{R}|^5}(\mathbf{R}\cdot\mathbf{n}'-\overline{\mathbf{R}}\cdot\overline{\mathbf{n}}')\overline{\mathbf{R}}\cdot(\Phi^{-1}\psi)\\
&\left.+\frac{(\mathbf{e}_j(s)\cdot\mathbf{R})-(\overline{\mathbf{e}}_j\cdot\overline{\mathbf{R}})}{|\mathbf{R}|^5}(\overline{\mathbf{R}}\cdot\overline{\mathbf{n}}')\overline{\mathbf{R}}\cdot(\Phi^{-1}\psi)\right.\\
&\left.+\left(\frac{1}{|\mathbf{R}|^5}-\frac{1}{|\overline{\mathbf{R}}|^5}\right)(\overline{\mathbf{e}}_j\cdot\overline{\mathbf{R}})(\overline{\mathbf{R}}\cdot\overline{\mathbf{n}}')\overline{\mathbf{R}}\cdot(\Phi^{-1}\psi)\right).
\end{aligned} W j ( ψ ; K D ) = 4 π 3 ( ∣ R ∣ 5 e j ( s ) ⋅ R ( R ⋅ n ′ ) ( R ⋅ ψ − R ⋅ ( Φ − 1 ψ )) + ∣ R ∣ 5 e j ( s ) ⋅ R ( R ⋅ n ′ − R ⋅ n ′ ) R ⋅ ( Φ − 1 ψ ) + ∣ R ∣ 5 ( e j ( s ) ⋅ R ) − ( e j ⋅ R ) ( R ⋅ n ′ ) R ⋅ ( Φ − 1 ψ ) + ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) ( e j ⋅ R ) ( R ⋅ n ′ ) R ⋅ ( Φ − 1 ψ ) ) .
Using the expansions (211) and (212) along with (210) and Lemma 4.1, for each j = t , r , θ j=t,r,\theta j = t , r , θ , we may bound
∣ W j ( ψ ; K D ) ∣ ≤ c ( κ ∗ , c Γ ) ∥ ψ ∥ L ∞ 1 ∣ R ‾ ∣ . |W_j(\psi;K_D)|\le c(\kappa_*,c_\Gamma)\|\psi\|_{L^\infty}\frac{1}{|\overline{\mathbf{R}}|}. ∣ W j ( ψ ; K D ) ∣ ≤ c ( κ ∗ , c Γ ) ∥ ψ ∥ L ∞ ∣ R ∣ 1 .
Using the above integrands in the definition (243) of W ( ψ ; K D ) = R D , 1 W(\psi;K_D)=\mathcal{R}_{D,1} W ( ψ ; K D ) = R D , 1 , by Lemma 4.2 we have
∥ R D , 1 [ ψ ] ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ϵ ∥ ψ ∥ L ∞ . (247) \|\mathcal{R}_{D,1}[\psi]\|_{L^\infty}\le c(\kappa_*,c_\Gamma)\epsilon\|\psi\|_{L^\infty}. \tag*{(247)} ∥ R D , 1 [ ψ ] ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ϵ ∥ ψ ∥ L ∞ . (247)
In addition, for two filaments with centerlines X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) satisfying Lemma 1.1, we may use the above expansion of W j ( ψ ; K D ) W_j(\psi;K_D) W j ( ψ ; K D ) along with (210), (211), (212), and the bounds (213) for each remainder term Q Q Q appearing in the expansion to obtain
∥ R D , 1 ( a ) [ ψ ] − R D , 1 ( b ) [ ψ ] ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ ψ ∥ L ∞ , (248) \|\mathcal{R}_{D,1}^{(a)}[\psi]-\mathcal{R}_{D,1}^{(b)}[\psi]\|_{L^\infty}\le c(\kappa_*^{(a)},\kappa_*^{(b)},c_\Gamma)\epsilon\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|\psi\|_{L^\infty}, \tag*{(248)} ∥ R D , 1 ( a ) [ ψ ] − R D , 1 ( b ) [ ψ ] ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ ψ ∥ L ∞ , (248)
by Lemma 4.2. Furthermore, using the form (38) of K D K_D K D and (211), we have
∥ R D , 2 [ ψ ] ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ∥ ϕ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π ϵ − 1 ∣ R ‾ ∣ 2 ϵ 2 d θ ^ d s ^ ≤ ϵ c ( κ ∗ , c Γ ) ∥ ψ ∥ L ∞ , (249) \|\mathcal{R}_{D,2}[\psi]\|_{L^\infty}\le c(\kappa_*,c_\Gamma)\|\phi\|_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\epsilon^{-1}}{|\overline{\mathbf{R}}|^2}\epsilon^2\mathrm{d}\widehat{\theta}\mathrm{d}\widehat{s}\le\epsilon c(\kappa_*,c_\Gamma)\|\psi\|_{L^\infty}, \tag*{(249)} ∥ R D , 2 [ ψ ] ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ∥ ϕ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 2 ϵ − 1 ϵ 2 d θ d s ≤ ϵc ( κ ∗ , c Γ ) ∥ ψ ∥ L ∞ , (249)
∥ R D , 2 ( a ) [ ψ ] − R D , 2 ( b ) [ ψ ] ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ ψ ∥ L ∞ . \|\mathcal{R}_{D,2}^{(a)}[\psi]-\mathcal{R}_{D,2}^{(b)}[\psi]\|_{L^\infty}\le c(\kappa_*^{(a)},\kappa_*^{(b)},c_\Gamma)\epsilon\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|\psi\|_{L^\infty}. ∥ R D , 2 ( a ) [ ψ ] − R D , 2 ( b ) [ ψ ] ∥ L ∞ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ ∥ X ( a ) − X ( b ) ∥ C 2 ∥ ψ ∥ L ∞ .
We now turn to C 0 , γ C^{0,\gamma} C 0 , γ estimates for the components 1 ϵ ∂ θ R D , 1 \frac{1}{\epsilon}\partial_{\theta}\mathcal{R}_{D,1} ϵ 1 ∂ θ R D , 1 , 1 ϵ ∂ θ R D , 2 \frac{1}{\epsilon}\partial_{\theta}\mathcal{R}_{D,2} ϵ 1 ∂ θ R D , 2 , ∂ s R D , 1 \partial_{s}\mathcal{R}_{D,1} ∂ s R D , 1 , ∂ s R D , 2 \partial_{s}\mathcal{R}_{D,2} ∂ s R D , 2 of the gradient of R D \mathcal{R}_{D} R D along the surface Γ ϵ \Gamma_{\epsilon} Γ ϵ . We note the following expansions:
1 ϵ ∂ θ R = 1 ϵ ∂ θ R t = e θ ( s , θ ) \frac{1}{\epsilon}\partial_{\theta}\mathbf{R}=\frac{1}{\epsilon}\partial_{\theta}\mathbf{R}_{t}=\mathbf{e}_{\theta}(s,\theta) ϵ 1 ∂ θ R = ϵ 1 ∂ θ R t = e θ ( s , θ )
1 ϵ ∂ θ R ⋅ R = ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 \frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{R}=\epsilon\sin(\hat{\theta})+\hat{s}^{2}Q_{R\theta1}+\epsilon\hat{s}Q_{R\theta2} ϵ 1 ∂ θ R ⋅ R = ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2
1 ϵ ∂ θ R ⋅ n ′ = 1 ϵ ∂ θ R ⋅ e r ( s − s ^ , θ − θ ^ ) = − sin ( θ ^ ) + s ^ Q R θ 3 \frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{n}'=\frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{e}_{r}(s-\hat{s},\theta-\hat{\theta})=-\sin(\hat{\theta})+\hat{s}Q_{R\theta3} ϵ 1 ∂ θ R ⋅ n ′ = ϵ 1 ∂ θ R ⋅ e r ( s − s ^ , θ − θ ^ ) = − sin ( θ ^ ) + s ^ Q R θ 3
1 ϵ ∂ θ R ⋅ n = 1 ϵ ∂ θ R ⋅ e r ( s , θ ) = 0 , 1 ϵ ∂ θ R ⋅ e t ( s ) = 0 , 1 ϵ ∂ θ R ⋅ e θ ( s , θ ) = 1. (250) \frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{n}=\frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{e}_{r}(s,\theta)=0,\qquad\frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{e}_{t}(s)=0,\qquad\frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{e}_{\theta}(s,\theta)=1. \tag*{(250)} ϵ 1 ∂ θ R ⋅ n = ϵ 1 ∂ θ R ⋅ e r ( s , θ ) = 0 , ϵ 1 ∂ θ R ⋅ e t ( s ) = 0 , ϵ 1 ∂ θ R ⋅ e θ ( s , θ ) = 1. (250)
Here the terms Q R θ j Q_{R\theta j} Q R θ j each satisfy bounds of the form (214) .
We start with the θ \theta θ -derivatives 1 ϵ ∂ θ R D , 1 \frac{1}{\epsilon}\partial_{\theta}\mathcal{R}_{D,1} ϵ 1 ∂ θ R D , 1 and 1 ϵ ∂ θ R D , 2 \frac{1}{\epsilon}\partial_{\theta}\mathcal{R}_{D,2} ϵ 1 ∂ θ R D , 2 , which may be written as follows:
1 ϵ ∂ θ R D , 1 [ ψ ] = W [ ψ ; 1 ϵ ∂ θ K D ] \frac{1}{\epsilon}\partial_{\theta}\mathcal{R}_{D,1}[\psi]=W[\psi;\frac{1}{\epsilon}\partial_{\theta}K_{D}] ϵ 1 ∂ θ R D , 1 [ ψ ] = W [ ψ ; ϵ 1 ∂ θ K D ]
1 ϵ ∂ θ R D , 2 [ ψ ] = − Φ − 1 p.v. ∫ − 1 / 2 1 / 2 ∫ − π π 1 ϵ ∂ θ K D ( s , s ^ , θ , θ ^ ) ψ ( s − s ^ , θ − θ ^ ) ϵ 2 κ ^ d θ ^ d s ^ . \frac{1}{\epsilon}\partial_{\theta}\mathcal{R}_{D,2}[\psi]=-\Phi^{-1}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1}{\epsilon}\partial_{\theta}K_{D}(s,\hat{s},\theta,\hat{\theta})\psi(s-\hat{s},\theta-\hat{\theta})\epsilon^{2}\hat{\kappa}\,d\hat{\theta}\,d\hat{s}. ϵ 1 ∂ θ R D , 2 [ ψ ] = − Φ − 1 p.v. ∫ − 1/2 1/2 ∫ − π π ϵ 1 ∂ θ K D ( s , s ^ , θ , θ ^ ) ψ ( s − s ^ , θ − θ ^ ) ϵ 2 κ ^ d θ ^ d s ^ .
We begin by estimating 1 ϵ ∂ θ R D , 1 [ ψ ] \frac{1}{\epsilon}\partial_{\theta}\mathcal{R}_{D,1}[\psi] ϵ 1 ∂ θ R D , 1 [ ψ ] . Using the definition (244) of the integrands W j W_j W j , we may write W t ( ψ ; 1 ϵ ∂ θ K D ) W_t(\psi;\frac{1}{\epsilon}\partial_{\theta}K_D) W t ( ψ ; ϵ 1 ∂ θ K D ) as
W t ( ψ ; 1 ϵ ∂ θ K D ) = 3 4 π ( W t , 1 + W t , 2 + W t , 3 ) , W_t(\psi;\frac{1}{\epsilon}\partial_{\theta}K_D)=\frac{3}{4\pi}(W_{t,1}+W_{t,2}+W_{t,3}), W t ( ψ ; ϵ 1 ∂ θ K D ) = 4 π 3 ( W t , 1 + W t , 2 + W t , 3 ) ,
W t , 1 = ( ( R ⋅ e t ) ( R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e z ) ( R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 ) ( e θ ⋅ ψ ) W_{t,1}=\left(\frac{(\mathbf{R}\cdot\mathbf{e}_t)(\mathbf{R}\cdot\mathbf{n}')}{|\mathbf{R}|^5}-\frac{(\overline{\mathbf{R}}\cdot\mathbf{e}_z)(\overline{\mathbf{R}}\cdot\overline{\mathbf{n}}')}{|\overline{\mathbf{R}}|^5}\right)(\mathbf{e}_{\theta}\cdot\psi) W t , 1 = ( ∣ R ∣ 5 ( R ⋅ e t ) ( R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e z ) ( R ⋅ n ′ ) ) ( e θ ⋅ ψ )
W t , 2 = ( R ⋅ e t ) ( R ⋅ ψ ) ( 1 ϵ ∂ θ R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e z ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( 1 ϵ ∂ θ R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 W_{t,2}=\frac{(\mathbf{R}\cdot\mathbf{e}_t)(\mathbf{R}\cdot\psi)(\frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{n}')}{|\mathbf{R}|^5}-\frac{(\overline{\mathbf{R}}\cdot\mathbf{e}_z)(\overline{\mathbf{R}}\cdot(\Phi^{-1}\psi))(\frac{1}{\epsilon}\partial_{\theta}\overline{\mathbf{R}}\cdot\overline{\mathbf{n}}')}{|\overline{\mathbf{R}}|^5} W t , 2 = ∣ R ∣ 5 ( R ⋅ e t ) ( R ⋅ ψ ) ( ϵ 1 ∂ θ R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e z ) ( R ⋅ ( Φ − 1 ψ )) ( ϵ 1 ∂ θ R ⋅ n ′ )
W t , 3 = − 5 ( R ⋅ e t ) ( R ⋅ ψ ) ( R ⋅ n ′ ) ( 1 ϵ ∂ θ R ⋅ R ) ∣ R ∣ 7 + 5 ( R ‾ ⋅ e z ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ‾ ′ ) ( 1 ϵ ∂ θ R ‾ ⋅ R ‾ ) ∣ R ‾ ∣ 7 . W_{t,3}=-\frac{5(\mathbf{R}\cdot\mathbf{e}_t)(\mathbf{R}\cdot\psi)(\mathbf{R}\cdot\mathbf{n}')(\frac{1}{\epsilon}\partial_{\theta}\mathbf{R}\cdot\mathbf{R})}{|\mathbf{R}|^7}+\frac{5(\overline{\mathbf{R}}\cdot\mathbf{e}_z)(\overline{\mathbf{R}}\cdot(\Phi^{-1}\psi))(\overline{\mathbf{R}}\cdot\overline{\mathbf{n}}')(\frac{1}{\epsilon}\partial_{\theta}\overline{\mathbf{R}}\cdot\overline{\mathbf{R}})}{|\overline{\mathbf{R}}|^7}. W t , 3 = − ∣ R ∣ 7 5 ( R ⋅ e t ) ( R ⋅ ψ ) ( R ⋅ n ′ ) ( ϵ 1 ∂ θ R ⋅ R ) + ∣ R ∣ 7 5 ( R ⋅ e z ) ( R ⋅ ( Φ − 1 ψ )) ( R ⋅ n ′ ) ( ϵ 1 ∂ θ R ⋅ R ) .
Using the expansions (212) , (213) , and (250) , we may write the components W t , j W_{t,j} W t , j in more detail:
W t , 1 = ( ( ϵ s ^ κ ^ + s ^ 2 Q R t ) ( − 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ′ ) + s ^ 3 Q R n ′ ∣ R ∣ 5 − ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) 2 ϵ s ^ sin 2 ( θ ^ 2 ) ) ( e θ ⋅ ψ ) W_{t,1}=\left(\frac{(\epsilon\hat{s}\hat{\kappa}+\hat{s}^{2}Q_{Rt})(-2\epsilon\sin^{2}(\frac{\hat{\theta}}{2})+\hat{s}^{2}Q_{Rn'})+\hat{s}^{3}Q_{Rn'}}{|\mathbf{R}|^{5}}-\left(\frac{1}{|\mathbf{R}|^{5}}-\frac{1}{|\overline{\mathbf{R}}|^{5}}\right)2\epsilon\hat{s}\sin^{2}(\frac{\hat{\theta}}{2})\right)(\mathbf{e}_{\theta}\cdot\psi) W t , 1 = ( ∣ R ∣ 5 ( ϵ s ^ κ ^ + s ^ 2 Q R t ) ( − 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ′ ) + s ^ 3 Q R n ′ − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) 2 ϵ s ^ sin 2 ( 2 θ ^ ) ) ( e θ ⋅ ψ )
W t , 2 = ( ϵ s ^ κ ^ + s ^ 2 Q R t ) ( − sin ( θ ^ ) + s ^ Q R θ 3 ) + s ^ 2 Q R θ 3 ∣ R ∣ 5 ( s ^ Q t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) + s ^ sin ( θ ^ ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ∣ R ∣ 5 − ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) s ^ sin ( θ ^ ) ( e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) W_{t,2}=\frac{(\epsilon\hat{s}\hat{\kappa}+\hat{s}^{2}Q_{Rt})(-\sin(\hat{\theta})+\hat{s}Q_{R\theta3})+\hat{s}^{2}Q_{R\theta3}}{|\mathbf{R}|^{5}}\left(\hat{s}Q_t\cdot\psi+2\epsilon\sin(\frac{\hat{\theta}}{2})\mathbf{e}_{\theta}\cdot\psi\right)
+\frac{\hat{s}\sin(\hat{\theta})(\hat{s}^{2}Q_t\cdot\psi-\epsilon\hat{s}Q_r\cdot\psi)}{|\mathbf{R}|^{5}}
-\left(\frac{1}{|\mathbf{R}|^{5}}-\frac{1}{|\overline{\mathbf{R}}|^{5}}\right)\hat{s}\sin(\hat{\theta})\left(\mathbf{e}_t\cdot\psi+2\epsilon\sin(\frac{\hat{\theta}}{2})\mathbf{e}_{\theta}\cdot\psi\right) W t , 2 = ∣ R ∣ 5 ( ϵ s ^ κ ^ + s ^ 2 Q R t ) ( − sin ( θ ^ ) + s ^ Q R θ 3 ) + s ^ 2 Q R θ 3 ( s ^ Q t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) + ∣ R ∣ 5 s ^ sin ( θ ^ ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) s ^ sin ( θ ^ ) ( e t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ )
W t , 3 = 10 ϵ s ^ sin 2 ( θ ^ 2 ) ( e t ⋅ φ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ φ ) ( s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 − 5 s ^ 3 Q R n ′ ( s ^ e t ⋅ φ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ φ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 + 5 s ^ ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( − 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 + 5 ( ϵ s ^ κ ^ + s ^ 2 Q R t ) ( s ^ Q t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 − ( 1 ∣ R ∣ 7 − 1 ∣ R ‾ ∣ 7 ) 10 ϵ 2 s ^ sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) , \begin{aligned}
W_{t,3}={}&\frac{10\epsilon\hat{s}\sin^{2}(\frac{\hat{\theta}}{2})\left(\mathbf{e}_t\cdot\varphi+2\epsilon\sin(\frac{\hat{\theta}}{2})\mathbf{e}_{\theta}\cdot\varphi\right)(\hat{s}^{2}Q_{R\theta1}+\epsilon\hat{s}Q_{R\theta2})}{|\mathbf{R}|^{7}}\\
&-\frac{5\hat{s}^{3}Q_{Rn'}\left(\hat{s}\mathbf{e}_t\cdot\varphi+2\epsilon\sin(\frac{\hat{\theta}}{2})\mathbf{e}_{\theta}\cdot\varphi\right)(\epsilon\sin(\hat{\theta})+\hat{s}^{2}Q_{R\theta1}+\epsilon\hat{s}Q_{R\theta2})}{|\mathbf{R}|^{7}}\\
&+\frac{5\hat{s}(\hat{s}^{2}Q_t\cdot\psi-\epsilon\hat{s}Q_r\cdot\psi)(-2\epsilon\sin^{2}(\frac{\hat{\theta}}{2})+\hat{s}^{2}Q_{Rn'})(\epsilon\sin(\hat{\theta})+\hat{s}^{2}Q_{R\theta1}+\epsilon\hat{s}Q_{R\theta2})}{|\mathbf{R}|^{7}}\\
&+\frac{5(\epsilon\hat{s}\hat{\kappa}+\hat{s}^{2}Q_{Rt})(\hat{s}Q_t\cdot\psi+2\epsilon\sin(\frac{\hat{\theta}}{2})\mathbf{e}_{\theta}\cdot\psi)(2\epsilon\sin^{2}(\frac{\hat{\theta}}{2})-\hat{s}^{2}Q_{Rn'})(\epsilon\sin(\hat{\theta})+\hat{s}^{2}Q_{R\theta1}+\epsilon\hat{s}Q_{R\theta2})}{|\mathbf{R}|^{7}}\\
&-\left(\frac{1}{|\mathbf{R}|^{7}}-\frac{1}{|\overline{\mathbf{R}}|^{7}}\right)10\epsilon^{2}\hat{s}\sin^{2}(\frac{\hat{\theta}}{2})\sin(\hat{\theta})\left(\hat{s}\mathbf{e}_t\cdot\psi+2\epsilon\sin(\frac{\hat{\theta}}{2})\mathbf{e}_{\theta}\cdot\psi\right),
\end{aligned} W t , 3 = ∣ R ∣ 7 10 ϵ s ^ sin 2 ( 2 θ ^ ) ( e t ⋅ φ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ φ ) ( s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) − ∣ R ∣ 7 5 s ^ 3 Q R n ′ ( s ^ e t ⋅ φ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ φ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) + ∣ R ∣ 7 5 s ^ ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( − 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) + ∣ R ∣ 7 5 ( ϵ s ^ κ ^ + s ^ 2 Q R t ) ( s ^ Q t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ^ ) − s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) − ( ∣ R ∣ 7 1 − ∣ R ∣ 7 1 ) 10 ϵ 2 s ^ sin 2 ( 2 θ ^ ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) ,
where Q ~ t = e t − s ^ Q t + ϵ Q r \widetilde{Q}_{t}=e_{t}-\widehat{s}Q_{t}+\epsilon Q_{r} Q t = e t − s Q t + ϵ Q r . Using the expansion (210) , (211) for 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 \frac{1}{|R|^{5}}-\frac{1}{|\overline{R}|^{5}} ∣ R ∣ 5 1 − ∣ R ∣ 5 1 and 1 ∣ R ∣ 7 − 1 ∣ R ‾ ∣ 7 \frac{1}{|R|^{7}}-\frac{1}{|\overline{R}|^{7}} ∣ R ∣ 7 1 − ∣ R ∣ 7 1 , we may apply Lemma 4.4 to obtain
∣ W ( ψ ; 1 ϵ ∂ θ K D ) ⋅ e z ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) max j ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ . (251) \begin{aligned}
\left|W\left(\psi;\frac{1}{\epsilon}\partial_{\theta}K_{D}\right)\cdot e_{z}\right|_{\dot{C}^{0,\gamma}}\leq c(\kappa_{*},c_{\Gamma})\max_{j}\left(\|Q_{j}\|_{C_{1}^{0,\gamma_{+}}}+\|Q_{j}\|_{C_{2}^{0,\gamma}}\right)\|\psi\|_{C^{0,\gamma}} \\
&\leq c(\kappa_{*},\gamma_{+},c_{\Gamma})\epsilon^{-\gamma_{+}}\|\psi\|_{C^{0,\gamma}}. \tag*{(251)}
\end{aligned} W ( ψ ; ϵ 1 ∂ θ K D ) ⋅ e z C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) j max ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ . (251)
Here the max j \max_{j} max j is taken over all forms of remainder terms appearing in the above expansions. We also note that terms of the form, for example, s ^ ϵ ϵ sin 2 ( θ ^ 2 ) Q ∣ R ∣ \frac{\widehat{s}^{\epsilon}\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)Q}{|R|} ∣ R ∣ s ϵ ϵ s i n 2 ( 2 θ ) Q may be treated as s ^ ϵ ϵ sin ( θ ^ 2 ) Q ~ ∣ R ∣ \frac{\widehat{s}^{\epsilon}\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\widetilde{Q}}{|R|} ∣ R ∣ s ϵ ϵ s i n ( 2 θ ) Q , Q ~ = sin ( θ ^ 2 ) Q \widetilde{Q}=\sin\left(\frac{\widehat{\theta}}{2}\right)Q Q = sin ( 2 θ ) Q , so as not to lose a factor of ϵ − γ \epsilon^{-\gamma} ϵ − γ from using case (1) of Lemma 4.4 .
Furthermore, using the above expansions of W t , j W_{t,j} W t , j and the Lipschitz estimates (214) for each remainder Q j Q_{j} Q j , we additionally note that, given two curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1 , we have
∣ W ( ψ ; 1 ϵ ∂ θ ( K D ( a ) − K D ( b ) ) ) ⋅ e z ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (252) \left|W\left(\psi;\frac{1}{\epsilon}\partial_{\theta}\left(K_{D}^{(a)}-K_{D}^{(b)}\right)\right)\cdot e_{z}\right|_{\dot{C}^{0,\gamma}}\leq c\left(\kappa_{*,\gamma_{+}}^{(a)},\kappa_{*,\gamma_{+}}^{(b)},c_{\Gamma}\right)\epsilon^{-\gamma_{+}}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma_{+}}}\|\psi\|_{C^{0,\gamma}}. \tag*{(252)} W ( ψ ; ϵ 1 ∂ θ ( K D ( a ) − K D ( b ) ) ) ⋅ e z C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + X ( a ) − X ( b ) C 2 , γ + ∥ ψ ∥ C 0 , γ . (252)
We next consider the integrand W r ( ψ ; 1 ϵ ∂ θ K D ) W_{r}\left(\psi;\frac{1}{\epsilon}\partial_{\theta}K_{D}\right) W r ( ψ ; ϵ 1 ∂ θ K D ) as defined in (244) . We may write
W r ( ψ ; 1 ϵ ∂ θ K D ) = 3 4 π ( W r , 1 + W r , 2 + W r , 3 ) , W_{r}\left(\psi;\frac{1}{\epsilon}\partial_{\theta}K_{D}\right)=\frac{3}{4\pi}\left(W_{r,1}+W_{r,2}+W_{r,3}\right), W r ( ψ ; ϵ 1 ∂ θ K D ) = 4 π 3 ( W r , 1 + W r , 2 + W r , 3 ) ,
W r , 1 = ( ( R ⋅ e r ) R ⋅ n ′ ∣ R ∣ 5 − ( R ‾ ⋅ e ‾ r ) R ‾ ⋅ n ‾ ′ ∣ R ‾ ∣ 5 ) ( e θ ⋅ ψ ) , W_{r,1}=\left(\frac{(R\cdot e_{r})R\cdot n'}{|R|^{5}}-\frac{(\overline{R}\cdot\overline{e}_{r})\overline{R}\cdot\overline{n}'}{|\overline{R}|^{5}}\right)(e_{\theta}\cdot\psi), W r , 1 = ( ∣ R ∣ 5 ( R ⋅ e r ) R ⋅ n ′ − ∣ R ∣ 5 ( R ⋅ e r ) R ⋅ n ′ ) ( e θ ⋅ ψ ) ,
W r , 2 = ( R ⋅ e r ) ( R ⋅ ψ ) ( 1 ϵ ∂ θ R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e ‾ r ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( 1 ϵ ∂ θ R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 , W_{r,2}=\frac{(R\cdot e_{r})(R\cdot\psi)\left(\frac{1}{\epsilon}\partial_{\theta}R\cdot n'\right)}{|R|^{5}}-\frac{(\overline{R}\cdot\overline{e}_{r})(\overline{R}\cdot(\Phi^{-1}\psi))\left(\frac{1}{\epsilon}\partial_{\theta}\overline{R}\cdot\overline{n}'\right)}{|\overline{R}|^{5}}, W r , 2 = ∣ R ∣ 5 ( R ⋅ e r ) ( R ⋅ ψ ) ( ϵ 1 ∂ θ R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e r ) ( R ⋅ ( Φ − 1 ψ )) ( ϵ 1 ∂ θ R ⋅ n ′ ) ,
W r , 3 = − 5 ( R ⋅ e r ) ( R ⋅ ψ ) ( R ⋅ n ′ ) ( 1 ϵ ∂ θ R ⋅ R ) ∣ R ∣ 7 + 5 ( R ‾ ⋅ e ‾ r ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ‾ ′ ) ( 1 ϵ ∂ θ R ‾ ⋅ R ‾ ) ∣ R ‾ ∣ 7 . W_{r,3}=-\frac{5(R\cdot e_{r})(R\cdot\psi)(R\cdot n')\left(\frac{1}{\epsilon}\partial_{\theta}R\cdot R\right)}{|R|^{7}}+\frac{5(\overline{R}\cdot\overline{e}_{r})(\overline{R}\cdot(\Phi^{-1}\psi))(\overline{R}\cdot\overline{n}')\left(\frac{1}{\epsilon}\partial_{\theta}\overline{R}\cdot\overline{R}\right)}{|\overline{R}|^{7}}. W r , 3 = − ∣ R ∣ 7 5 ( R ⋅ e r ) ( R ⋅ ψ ) ( R ⋅ n ′ ) ( ϵ 1 ∂ θ R ⋅ R ) + ∣ R ∣ 7 5 ( R ⋅ e r ) ( R ⋅ ( Φ − 1 ψ )) ( R ⋅ n ′ ) ( ϵ 1 ∂ θ R ⋅ R ) .
Again by (212) , (213) , and (250) , we may further expand each W r , j W_{r,j} W r , j as
W r , 1 = ( − 2 ϵ ^ s ^ 2 sin 2 ( θ ^ 2 ) ( Q R n + Q R n ′ ) + s ^ 4 Q R n Q R n ′ ∣ R ∣ 5 − 4 ϵ 2 sin 4 ( θ ^ 2 ) ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) ) ( e θ ⋅ ψ ) , W_{r,1}=\left(\frac{-2\widehat{\epsilon}\widehat{s}^{2}\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(Q_{Rn}+Q_{Rn'})+\widehat{s}^{4}Q_{Rn}Q_{Rn'}}{|R|^{5}}-4\epsilon^{2}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)\left(\frac{1}{|R|^{5}}-\frac{1}{|\overline{R}|^{5}}\right)\right)(e_{\theta}\cdot\psi), W r , 1 = ∣ R ∣ 5 − 2 ϵ s 2 sin 2 ( 2 θ ) ( Q R n + Q R n ′ ) + s 4 Q R n Q R n ′ − 4 ϵ 2 sin 4 ( 2 θ ) ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) ( e θ ⋅ ψ ) ,
W r , 2 = s ^ 2 Q R n ( s ^ Q ~ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( − sin ( θ ^ ) + s ^ Q R θ 3 ) ∣ R ∣ 5 + 2 ϵ sin 2 ( θ ^ 2 ) [ ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( sin ( θ ^ ) − s ^ Q R θ 3 ) + s ^ Q R θ 3 ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ] ∣ R ∣ 5 − ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) 2 ϵ sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) . \begin{aligned}
W_{r,2}={}&\frac{\widehat{s}^{2}Q_{Rn}\left(\widehat{s}\widetilde{Q}_{t}\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\left(-\sin(\widehat{\theta})+\widehat{s}Q_{R\theta3}\right)}{|R|^{5}}\\
&+\frac{2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\left[\left(\widehat{s}^{2}Q_{t}\cdot\psi-\epsilon\widehat{s}Q_{r}\cdot\psi\right)\left(\sin(\widehat{\theta})-\widehat{s}Q_{R\theta3}\right)+\widehat{s}Q_{R\theta3}\left(\widehat{s}e_{t}\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\right]}{|R|^{5}}\\
&-\left(\frac{1}{|R|^{5}}-\frac{1}{|\overline{R}|^{5}}\right)2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})\left(\widehat{s}e_{t}\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right).
\end{aligned} W r , 2 = ∣ R ∣ 5 s 2 Q R n ( s Q t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( − sin ( θ ) + s Q R θ 3 ) + ∣ R ∣ 5 2 ϵ sin 2 ( 2 θ ) [ ( s 2 Q t ⋅ ψ − ϵ s Q r ⋅ ψ ) ( sin ( θ ) − s Q R θ 3 ) + s Q R θ 3 ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ] − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) 2 ϵ sin 2 ( 2 θ ) sin ( θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) .
W r , 3 = 5 s ^ 2 Q R n ( s ^ Q ~ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 − 10 ϵ sin 2 ( θ ^ 2 ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 − 10 ϵ sin 2 ( θ ^ 2 ) s ^ 2 Q R n ′ ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 + 20 ϵ 2 sin 4 ( θ ^ 2 ) ( s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ∣ 7 . + ( 1 ∣ R ∣ 7 − 1 ∣ R ‾ ∣ 7 ) 20 ε 3 sin 4 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) . \begin{aligned}
W_{r,3}={}&\frac{5\widehat{s}^{2}Q_{Rn}\left(\widehat{s}\widetilde{Q}_{t}\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\left(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q_{Rn'}\right)\left(\epsilon\sin(\widehat{\theta})+\widehat{s}^{2}Q_{R\theta1}+\epsilon\widehat{s}Q_{R\theta2}\right)}{|R|^{7}}\\
&-\frac{10\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\left(\widehat{s}^{2}Q_{t}\cdot\psi-\epsilon\widehat{s}Q_{r}\cdot\psi\right)\left(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q_{Rn'}\right)\left(\epsilon\sin(\widehat{\theta})+\widehat{s}^{2}Q_{R\theta1}+\epsilon\widehat{s}Q_{R\theta2}\right)}{|R|^{7}}\\
&-\frac{10\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\widehat{s}^{2}Q_{Rn'}\left(\widehat{s}e_{t}\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\left(\epsilon\sin(\widehat{\theta})+\widehat{s}^{2}Q_{R\theta1}+\epsilon\widehat{s}Q_{R\theta2}\right)}{|R|^{7}}\\
&+\frac{20\epsilon^{2}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)\left(\widehat{s}^{2}Q_{R\theta1}+\epsilon\widehat{s}Q_{R\theta2}\right)\left(\widehat{s}e_{t}\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)}{|R|^{7}}. \\
&+\left(\frac{1}{\lvert R\rvert^{7}}-\frac{1}{\lvert\overline{R}\rvert^{7}}\right)20\varepsilon^{3}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})\left(\widehat{s}e_t\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right).
\end{aligned} W r , 3 = ∣ R ∣ 7 5 s 2 Q R n ( s Q t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( ϵ sin ( θ ) + s 2 Q R θ 1 + ϵ s Q R θ 2 ) − ∣ R ∣ 7 10 ϵ sin 2 ( 2 θ ) ( s 2 Q t ⋅ ψ − ϵ s Q r ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( ϵ sin ( θ ) + s 2 Q R θ 1 + ϵ s Q R θ 2 ) − ∣ R ∣ 7 10 ϵ sin 2 ( 2 θ ) s 2 Q R n ′ ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( ϵ sin ( θ ) + s 2 Q R θ 1 + ϵ s Q R θ 2 ) + ∣ R ∣ 7 20 ϵ 2 sin 4 ( 2 θ ) ( s 2 Q R θ 1 + ϵ s Q R θ 2 ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) . + ( ∣ R ∣ 7 1 − ∣ R ∣ 7 1 ) 20 ε 3 sin 4 ( 2 θ ) sin ( θ ) ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) .
Here again Q ~ t = e t − s ^ Q t + ε Q r \widetilde{Q}_t=e_t-\widehat{s}Q_t+\varepsilon Q_r Q t = e t − s Q t + ε Q r , and we may use the expansion (210) , (211) for 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 \frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\overline{R}\rvert^{5}} ∣ R ∣ 5 1 − ∣ R ∣ 5 1 and 1 ∣ R ∣ 7 − 1 ∣ R ‾ ∣ 7 \frac{1}{\lvert R\rvert^{7}}-\frac{1}{\lvert\overline{R}\rvert^{7}} ∣ R ∣ 7 1 − ∣ R ∣ 7 1 . We may then apply Lemma 4.4 to obtain
∣ W ( ψ ; 1 ε ∂ θ K D ) ⋅ e ‾ r ∣ C 0 , γ ≤ c ( κ ∗ , c Γ ) max j ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ε − γ + ∥ ψ ∥ C 0 , γ , (253) \begin{aligned}
\left\lvert W\left(\psi;\frac{1}{\varepsilon}\partial_{\theta}K_D\right)\cdot\overline{e}_r\right\rvert_{C^0,\gamma}\leq c(\kappa_*,c_{\Gamma})\max_j\left(\lVert Q_j\rVert_{C_1^{0,\gamma_+}}+\lVert Q_j\rVert_{C_2^{0,\gamma}}\right)\lVert\psi\rVert_{C^{0,\gamma}} \\
&\leq c(\kappa_{*,\gamma_+},c_{\Gamma})\varepsilon^{-\gamma_+}\lVert\psi\rVert_{C^{0,\gamma}}, \tag*{(253)}
\end{aligned} W ( ψ ; ε 1 ∂ θ K D ) ⋅ e r C 0 , γ ≤ c ( κ ∗ , c Γ ) j max ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ε − γ + ∥ ψ ∥ C 0 , γ , (253)
∣ W ( ψ ; 1 ε ∂ θ ( K D ( a ) − K D ( b ) ) ) ⋅ e ‾ r ∣ C 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ε − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . \left\lvert W\left(\psi;\frac{1}{\varepsilon}\partial_{\theta}\left(K_D^{(a)}-K_D^{(b)}\right)\right)\cdot\overline{e}_r\right\rvert_{C^{0,\gamma}}\leq c\left(\kappa_{*,\gamma_+}^{(a)},\kappa_{*,\gamma_+}^{(b)},c_{\Gamma}\right)\varepsilon^{-\gamma_+}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^{2,\gamma_+}}\lVert\psi\rVert_{C^{0,\gamma}}. W ( ψ ; ε 1 ∂ θ ( K D ( a ) − K D ( b ) ) ) ⋅ e r C 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ε − γ + X ( a ) − X ( b ) C 2 , γ + ∥ ψ ∥ C 0 , γ .
Where necessary, we again may exchange some regularity for a factor of ε γ \varepsilon^\gamma ε γ by treating terms of the form, e.g., s ^ 4 ε sin 2 ( θ ^ 2 ) Q ^ ∣ R ∣ \frac{\widehat{s}^{4}\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\widehat{Q}}{\lvert R\rvert} ∣ R ∣ s 4 ε s i n 2 ( 2 θ ) Q as s ^ 4 ε sin ( θ ^ 2 ) Q ^ ∣ R ∣ \frac{\widehat{s}^{4}\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\widehat{Q}}{\lvert R\rvert} ∣ R ∣ s 4 ε s i n ( 2 θ ) Q with Q ~ = sin ( θ ^ 2 ) Q \widetilde{Q}=\sin\left(\frac{\widehat{\theta}}{2}\right)Q Q = sin ( 2 θ ) Q . Then case (2) of Lemma 4.4 yields the bound (252).
Finally we turn to the integrand W θ ( ψ ; 1 ε ∂ θ K D ) W_{\theta}\left(\psi;\frac{1}{\varepsilon}\partial_{\theta}K_D\right) W θ ( ψ ; ε 1 ∂ θ K D ) as defined in (244) . We may write
W θ ( ψ ; 1 ε ∂ θ K D ) = 3 4 π ( W θ , 1 + W θ , 2 + W θ , 3 + W θ , 4 ) , W_{\theta}\left(\psi;\frac{1}{\varepsilon}\partial_{\theta}K_D\right)=\frac{3}{4\pi}\left(W_{\theta,1}+W_{\theta,2}+W_{\theta,3}+W_{\theta,4}\right), W θ ( ψ ; ε 1 ∂ θ K D ) = 4 π 3 ( W θ , 1 + W θ , 2 + W θ , 3 + W θ , 4 ) ,
W θ , 1 = ( R ⋅ ψ ) ( R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 , W_{\theta,1}=\frac{(R\cdot\psi)(R\cdot n')}{\lvert R\rvert^{5}}-\frac{(\overline{R}\cdot(\Phi^{-1}\psi))(\overline{R}\cdot\overline{n}')}{\lvert\overline{R}\rvert^{5}}, W θ , 1 = ∣ R ∣ 5 ( R ⋅ ψ ) ( R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ ( Φ − 1 ψ )) ( R ⋅ n ′ ) ,
W θ , 2 = ( R ⋅ e θ ) ( 1 ε ∂ θ R ⋅ ψ ) ( R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e ‾ θ ) ( 1 ε ∂ θ R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 , W_{\theta,2}=\frac{(R\cdot e_{\theta})\left(\frac{1}{\varepsilon}\partial_{\theta}R\cdot\psi\right)(R\cdot n')}{\lvert R\rvert^{5}}-\frac{(\overline{R}\cdot\overline{e}_{\theta})\left(\frac{1}{\varepsilon}\partial_{\theta}\overline{R}\cdot(\Phi^{-1}\psi)\right)(\overline{R}\cdot\overline{n}')}{\lvert\overline{R}\rvert^{5}}, W θ , 2 = ∣ R ∣ 5 ( R ⋅ e θ ) ( ε 1 ∂ θ R ⋅ ψ ) ( R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e θ ) ( ε 1 ∂ θ R ⋅ ( Φ − 1 ψ ) ) ( R ⋅ n ′ ) ,
W θ , 3 = ( R ⋅ e θ ) ( R ⋅ ψ ) ( 1 ε ∂ θ R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e ‾ θ ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( 1 ε ∂ θ R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 , W_{\theta,3}=\frac{(R\cdot e_{\theta})(R\cdot\psi)\left(\frac{1}{\varepsilon}\partial_{\theta}R\cdot n'\right)}{\lvert R\rvert^{5}}-\frac{(\overline{R}\cdot\overline{e}_{\theta})(\overline{R}\cdot(\Phi^{-1}\psi))\left(\frac{1}{\varepsilon}\partial_{\theta}\overline{R}\cdot\overline{n}'\right)}{\lvert\overline{R}\rvert^{5}}, W θ , 3 = ∣ R ∣ 5 ( R ⋅ e θ ) ( R ⋅ ψ ) ( ε 1 ∂ θ R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e θ ) ( R ⋅ ( Φ − 1 ψ )) ( ε 1 ∂ θ R ⋅ n ′ ) ,
W θ , 4 = − 5 ( R ⋅ e θ ) ( R ⋅ ψ ) ( R ⋅ n ′ ) ( 1 ε ∂ θ R ⋅ R ) ∣ R ∣ 7 + 5 ( R ‾ ⋅ e ‾ θ ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ‾ ′ ) ( 1 ε ∂ θ R ‾ ⋅ R ‾ ) ∣ R ‾ ∣ 7 . W_{\theta,4}=-\frac{5(R\cdot e_{\theta})(R\cdot\psi)(R\cdot n')\left(\frac{1}{\varepsilon}\partial_{\theta}R\cdot R\right)}{\lvert R\rvert^{7}}+\frac{5(\overline{R}\cdot\overline{e}_{\theta})(\overline{R}\cdot(\Phi^{-1}\psi))(\overline{R}\cdot\overline{n}')\left(\frac{1}{\varepsilon}\partial_{\theta}\overline{R}\cdot\overline{R}\right)}{\lvert\overline{R}\rvert^{7}}. W θ , 4 = − ∣ R ∣ 7 5 ( R ⋅ e θ ) ( R ⋅ ψ ) ( R ⋅ n ′ ) ( ε 1 ∂ θ R ⋅ R ) + ∣ R ∣ 7 5 ( R ⋅ e θ ) ( R ⋅ ( Φ − 1 ψ )) ( R ⋅ n ′ ) ( ε 1 ∂ θ R ⋅ R ) .
We may again use (212) , (213) , and (250) to expand each W θ , j W_{\theta,j} W θ , j as
W θ , 1 = ( 2 ε sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ 2 Q t ⋅ ψ − ε s ^ Q r ⋅ ψ ) + s ^ 2 Q R n ′ ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ∣ 5 W_{\theta,1}=\frac{\left(2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q_{Rn'}\right)\left(\widehat{s}^{2}Q_t\cdot\psi-\varepsilon\widehat{s}Q_r\cdot\psi\right)+\widehat{s}^{2}Q_{Rn'}\left(\widehat{s}e_t\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)}{\lvert R\rvert^{5}} W θ , 1 = ∣ R ∣ 5 ( 2 ε sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( s 2 Q t ⋅ ψ − ε s Q r ⋅ ψ ) + s 2 Q R n ′ ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ )
− ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) 2 ε sin 2 ( θ ^ 2 ) ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) , -\left(\frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\overline{R}\rvert^{5}}\right)2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\left(\widehat{s}e_t\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right), − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) 2 ε sin 2 ( 2 θ ) ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ,
W θ , 2 = ( ε s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ) ( − 2 ε sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ′ ) + ε sin ( θ ^ ) s ^ 2 Q R n ′ ∣ R ∣ 5 ( ψ ⋅ e θ ) + ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) ε sin ( θ ^ ) ( − 2 ε sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ′ ) ( ψ ⋅ e θ ) , \begin{aligned}
W_{\theta,2}=\frac{\left(\varepsilon\widehat{s}\kappa_3\cos(\theta-\widehat{\theta})+\widehat{s}^{2}Q_{R\theta}\right)\left(-2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)+\widehat{s}^{2}Q_{Rn'}\right)+\varepsilon\sin(\widehat{\theta})\widehat{s}^{2}Q_{Rn'}}{\lvert R\rvert^{5}}(\psi\cdot e_{\theta}) \\
&+\left(\frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\overline{R}\rvert^{5}}\right)\varepsilon\sin(\widehat{\theta})\left(-2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)+\widehat{s}^{2}Q_{Rn'}\right)(\psi\cdot e_{\theta}),
\end{aligned} W θ , 2 = ∣ R ∣ 5 ( ε s κ 3 cos ( θ − θ ) + s 2 Q R θ ) ( − 2 ε sin 2 ( 2 θ ) + s 2 Q R n ′ ) + ε sin ( θ ) s 2 Q R n ′ ( ψ ⋅ e θ ) + ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) ε sin ( θ ) ( − 2 ε sin 2 ( 2 θ ) + s 2 Q R n ′ ) ( ψ ⋅ e θ ) ,
W θ , 3 = ( ε s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ) ( s ^ Q ~ t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ( − sin ( θ ^ ) + s ^ Q R θ 3 ) ∣ R ∣ 5 + ε sin ( θ ^ ) s ^ [ ( sin ( θ ^ ) − s ^ Q R θ 3 ) ( s ^ Q t ⋅ ψ − ε Q r ⋅ ψ ) + Q R θ 3 ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ] ∣ R ∣ 5 \begin{aligned}
W_{\theta,3}=\frac{\left(\varepsilon\widehat{s}\kappa_3\cos(\theta-\widehat{\theta})+\widehat{s}^{2}Q_{R\theta}\right)\left(\widehat{s}\widetilde{Q}_t\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\left(-\sin(\widehat{\theta})+\widehat{s}Q_{R\theta3}\right)}{\lvert R\rvert^{5}} \\
&+\frac{\varepsilon\sin(\widehat{\theta})\widehat{s}\left[\left(\sin(\widehat{\theta})-\widehat{s}Q_{R\theta3}\right)\left(\widehat{s}Q_t\cdot\psi-\varepsilon Q_r\cdot\psi\right)+Q_{R\theta3}\left(\widehat{s}e_t\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\right]}{\lvert R\rvert^{5}}
\end{aligned} W θ , 3 = ∣ R ∣ 5 ( ε s κ 3 cos ( θ − θ ) + s 2 Q R θ ) ( s Q t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ( − sin ( θ ) + s Q R θ 3 ) + ∣ R ∣ 5 ε sin ( θ ) s [ ( sin ( θ ) − s Q R θ 3 ) ( s Q t ⋅ ψ − ε Q r ⋅ ψ ) + Q R θ 3 ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ]
− ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) ε sin 2 ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) -\left(\frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\overline{R}\rvert^{5}}\right)\varepsilon\sin^{2}(\widehat{\theta})\left(\widehat{s}e_t\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right) − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) ε sin 2 ( θ ) ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ )
W θ , 4 = 5 ( s ^ Q ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( ϵ s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 + 5 ϵ sin ( θ ^ ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( − 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ′ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 − 5 ϵ sin ( θ ^ ) s ^ 2 Q R n ′ ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( ϵ sin ( θ ^ ) + s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ∣ R ∣ 7 + 10 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ 2 Q R θ 1 + ϵ s ^ Q R θ 2 ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ∣ 7 + ( 1 ∣ R ∣ 7 − 1 ∣ R ∣ 7 ) 10 ϵ 3 sin 2 ( θ ^ 2 ) sin 2 ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) . W_{\theta,4} =
\frac{5(\widehat{s} \widehat{\mathbf{Q}}_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(\epsilon\widehat{s}\kappa_3\cos(\theta-\widehat{\theta})+\widehat{s}^{2}Q R_\theta)(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q R_{n'}) (\epsilon\sin(\widehat{\theta})+\widehat{s}^{2}Q R_{\theta1}+\epsilon\widehat{s}Q R_{\theta2})}{\lvert R\rvert^{7}}
+\frac{5\epsilon\sin(\widehat{\theta})(\widehat{s}^{2}\mathbf{Q}_t\cdot\psi-\epsilon\widehat{s}\mathbf{Q}_r\cdot\psi)(-2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)+\widehat{s}^{2}Q R_{n'})(\epsilon\sin(\widehat{\theta})+\widehat{s}^{2}Q R_{\theta1}+\epsilon\widehat{s}Q R_{\theta2})}{\lvert R\rvert^{7}}
-\frac{5\epsilon\sin(\widehat{\theta})\widehat{s}^{2}Q R_{n'}(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(\epsilon\sin(\widehat{\theta})+\widehat{s}^{2}Q R_{\theta1}+\epsilon\widehat{s}Q R_{\theta2})}{\lvert R\rvert^{7}}
+\frac{10\epsilon^{2}\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})(\widehat{s}^{2}Q R_{\theta1}+\epsilon\widehat{s}Q R_{\theta2})(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)}{\lvert R\rvert^{7}}
+\left(\frac{1}{\lvert R\rvert^{7}}-\frac{1}{\lvert R\rvert^{7}}\right)10\epsilon^{3}\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin^{2}(\widehat{\theta})(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi). W θ , 4 = ∣ R ∣ 7 5 ( s Q t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( ϵ s κ 3 cos ( θ − θ ) + s 2 Q R θ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( ϵ sin ( θ ) + s 2 Q R θ 1 + ϵ s Q R θ 2 ) + ∣ R ∣ 7 5 ϵ sin ( θ ) ( s 2 Q t ⋅ ψ − ϵ s Q r ⋅ ψ ) ( − 2 ϵ sin 2 ( 2 θ ) + s 2 Q R n ′ ) ( ϵ sin ( θ ) + s 2 Q R θ 1 + ϵ s Q R θ 2 ) − ∣ R ∣ 7 5 ϵ sin ( θ ) s 2 Q R n ′ ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( ϵ sin ( θ ) + s 2 Q R θ 1 + ϵ s Q R θ 2 ) + ∣ R ∣ 7 10 ϵ 2 sin 2 ( 2 θ ) sin ( θ ) ( s 2 Q R θ 1 + ϵ s Q R θ 2 ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) + ( ∣ R ∣ 7 1 − ∣ R ∣ 7 1 ) 10 ϵ 3 sin 2 ( 2 θ ) sin 2 ( θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) .
As before, using the expansions (209), (210) for 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 \frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\overline{R}\rvert^{5}} ∣ R ∣ 5 1 − ∣ R ∣ 5 1 and 1 ∣ R ∣ 7 − 1 ∣ R ‾ ∣ 7 \frac{1}{\lvert R\rvert^{7}}-\frac{1}{\lvert\overline{R}\rvert^{7}} ∣ R ∣ 7 1 − ∣ R ∣ 7 1 and applying Lemma 4.4 yields the bounds
∣ W ( ψ ; 1 ϵ ∂ θ K D ) ⋅ e ‾ θ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) max j ( ∥ Q j ∥ C 0 0 , γ + + ∥ Q j ∥ C 0 γ + ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ , \left\lvert W(\psi;\tfrac{1}{\epsilon}\partial_\theta K_D)\cdot\overline{e}_\theta\right\rvert_{\dot{C}^{0,\gamma}}
\leq c(\kappa_*,c_\Gamma)\max_j\left(\lVert Q_j\rVert_{C_0^{0,\gamma_+}}+\lVert Q_j\rVert_{C_0^{\gamma_+}}\right)\lVert\psi\rVert_{C^{0,\gamma}}
\leq c(\kappa_*,\gamma_+,c_\Gamma)\epsilon^{1-\gamma_+}\lVert\psi\rVert_{C^{0,\gamma}}, W ( ψ ; ϵ 1 ∂ θ K D ) ⋅ e θ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) j max ( ∥ Q j ∥ C 0 0 , γ + + ∥ Q j ∥ C 0 γ + ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ ,
∣ W ( ψ ; 1 ϵ ∂ θ ( K D ( a ) − K D ( b ) ) ) ⋅ e ‾ θ ∣ C ˙ 0 , γ ≤ c ( κ ∗ ( a ) , γ + , κ ∗ ( b ) , γ + , c Γ ) ϵ 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (254) \left\lvert W(\psi;\tfrac{1}{\epsilon}\partial_\theta(K_D^{(a)}-K_D^{(b)}))\cdot\overline{e}_\theta\right\rvert_{\dot{C}^{0,\gamma}}
\leq c(\kappa_*^{(a)},\gamma_+,\kappa_*^{(b)},\gamma_+,c_\Gamma)\epsilon^{1-\gamma_+}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\gamma_+}}\lVert\psi\rVert_{C^{0,\gamma}}. \tag*{(254)} W ( ψ ; ϵ 1 ∂ θ ( K D ( a ) − K D ( b ) )) ⋅ e θ C ˙ 0 , γ ≤ c ( κ ∗ ( a ) , γ + , κ ∗ ( b ) , γ + , c Γ ) ϵ 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (254)
Combining (250), (251), (252), and (253), we obtain
∣ 1 ϵ ∂ θ R D , 1 [ ψ ] ∣ C 0 , γ = ∣ W ( ψ ; 1 ϵ ∂ θ K D ) ∣ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ , \left\lvert\tfrac{1}{\epsilon}\partial_\theta\mathcal{R}_{D,1}[\psi]\right\rvert_{C^{0,\gamma}}
=\left\lvert W(\psi;\tfrac{1}{\epsilon}\partial_\theta K_D)\right\rvert_{C^{0,\gamma}}
\leq c(\kappa_*,\gamma_+,c_\Gamma)\epsilon^{1-\gamma_+}\lVert\psi\rVert_{C^{0,\gamma}}, ϵ 1 ∂ θ R D , 1 [ ψ ] C 0 , γ = W ( ψ ; ϵ 1 ∂ θ K D ) C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ ,
∣ 1 ϵ ∂ θ R D , 1 ( a ) [ ψ ] − 1 ϵ ∂ θ R D , 1 ( b ) [ ψ ] ∣ C 0 , γ ≤ c ( κ ∗ ( a ) , γ + , κ ∗ ( b ) , γ + , c Γ ) ϵ 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (255) \left\lvert\tfrac{1}{\epsilon}\partial_\theta\mathcal{R}_{D,1}^{(a)}[\psi]-\tfrac{1}{\epsilon}\partial_\theta\mathcal{R}_{D,1}^{(b)}[\psi]\right\rvert_{C^{0,\gamma}}
\leq c(\kappa_*^{(a)},\gamma_+,\kappa_*^{(b)},\gamma_+,c_\Gamma)\epsilon^{1-\gamma_+}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\gamma_+}}\lVert\psi\rVert_{C^{0,\gamma}}. \tag*{(255)} ϵ 1 ∂ θ R D , 1 ( a ) [ ψ ] − ϵ 1 ∂ θ R D , 1 ( b ) [ ψ ] C 0 , γ ≤ c ( κ ∗ ( a ) , γ + , κ ∗ ( b ) , γ + , c Γ ) ϵ 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (255)
We next turn to 1 ϵ ∂ θ R D , 2 [ ψ ] \tfrac{1}{\epsilon}\partial_\theta\mathcal{R}_{D,2}[\psi] ϵ 1 ∂ θ R D , 2 [ ψ ] , which we note may be written
1 ϵ ∂ θ R D , 2 [ ψ ] = − W [ ψ ; κ ^ ∂ θ ^ K D ] − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π κ ^ ∂ θ ^ K ‾ D ( Φ − 1 ψ ) ( s − s ^ , θ − θ ^ ) ϵ d θ ^ d s ^ . \tfrac{1}{\epsilon}\partial_\theta\mathcal{R}_{D,2}[\psi]
=-W[\psi;\widehat{\kappa}\partial_{\widehat{\theta}}K_D]
-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\widehat{\kappa}\partial_{\widehat{\theta}}\overline{K}_D(\Phi^{-1}\psi)(s-\widehat{s},\theta-\widehat{\theta})\epsilon\,d\widehat{\theta}\,d\widehat{s}. ϵ 1 ∂ θ R D , 2 [ ψ ] = − W [ ψ ; κ ∂ θ K D ] − p.v. ∫ − 1/2 1/2 ∫ − π π κ ∂ θ K D ( Φ − 1 ψ ) ( s − s , θ − θ ) ϵ d θ d s .
Note the additional factor of ϵ κ ^ \epsilon\widehat{\kappa} ϵ κ is due to the Jacobian factor (8). Using that (254) implies
∣ W [ ψ ; κ ^ ∂ θ ^ K D ] ∣ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ , \left\lvert W[\psi;\widehat{\kappa}\partial_{\widehat{\theta}}K_D]\right\rvert_{C^{0,\gamma}}
\leq c(\kappa_*,\gamma_+,c_\Gamma)\epsilon^{1-\gamma_+}\lVert\psi\rVert_{C^{0,\gamma}}, W [ ψ ; κ ∂ θ K D ] C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ ,
∣ W [ ψ ; κ ^ ( a ) ∂ θ ^ K D ( a ) − κ ^ ( b ) ∂ θ ^ K D ( b ) ] ∣ C 0 , γ ≤ c ( κ ∗ ( a ) , γ + , κ ∗ ( b ) , γ + , c Γ ) ϵ 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ , \left\lvert W[\psi;\widehat{\kappa}^{(a)}\partial_{\widehat{\theta}}K_D^{(a)}-\widehat{\kappa}^{(b)}\partial_{\widehat{\theta}}K_D^{(b)}]\right\rvert_{C^{0,\gamma}}
\leq c(\kappa_*^{(a)},\gamma_+,\kappa_*^{(b)},\gamma_+,c_\Gamma)\epsilon^{1-\gamma_+}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\gamma_+}}\lVert\psi\rVert_{C^{0,\gamma}}, W [ ψ ; κ ( a ) ∂ θ K D ( a ) − κ ( b ) ∂ θ K D ( b ) ] C 0 , γ ≤ c ( κ ∗ ( a ) , γ + , κ ∗ ( b ) , γ + , c Γ ) ϵ 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ ,
it remains to bound the second term of the 1 ϵ ∂ θ R D , 2 [ ψ ] \tfrac{1}{\epsilon}\partial_\theta\mathcal{R}_{D,2}[\psi] ϵ 1 ∂ θ R D , 2 [ ψ ] expression. Using the form (196) of R ‾ \overline{R} R , we may write
e z ⋅ ( κ ^ ∂ θ ^ K ‾ D ( Φ − 1 ψ ) ) = 3 ϵ κ ^ 4 π ( − 2 ϵ sin 2 ( θ ^ 2 ) s ^ ( e θ ⋅ ψ ) ∣ R ‾ ∣ 5 − s ^ sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 5 + 10 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) s ^ ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 7 ) e_z\cdot\left(\widehat{\kappa}\partial_{\widehat{\theta}}\overline{K}_D(\Phi^{-1}\psi)\right)
=\frac{3\epsilon\widehat{\kappa}}{4\pi}\left(
-\frac{2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\widehat{s}(e_\theta\cdot\psi)}{\lvert\overline{R}\rvert^{5}}
-\frac{\widehat{s}\sin(\widehat{\theta})(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)}{\lvert\overline{R}\rvert^{5}}
+\frac{10\epsilon^{2}\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})\widehat{s}(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)}{\lvert\overline{R}\rvert^{7}}
\right) e z ⋅ ( κ ∂ θ K D ( Φ − 1 ψ ) ) = 4 π 3 ϵ κ − ∣ R ∣ 5 2 ϵ sin 2 ( 2 θ ) s ( e θ ⋅ ψ ) − ∣ R ∣ 5 s sin ( θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) + ∣ R ∣ 7 10 ϵ 2 sin 2 ( 2 θ ) sin ( θ ) s ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ )
e ‾ r ⋅ ( κ ^ ∂ θ ^ K ‾ D ( Φ − 1 ψ ) ) = 3 ϵ κ ^ 4 π ( − 2 ϵ sin 2 ( θ ^ 2 ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ + ϵ sin ( θ ^ ) e θ ⋅ ψ ) ∣ R ‾ ∣ 5 − ϵ sin 2 ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 5 + 10 ϵ 3 sin 2 ( θ ^ 2 ) sin 2 ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 7 ) \overline{e}_r\cdot\left(\widehat{\kappa}\partial_{\widehat{\theta}}\overline{K}_D(\Phi^{-1}\psi)\right)
=\frac{3\epsilon\widehat{\kappa}}{4\pi}\left(
-\frac{2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi+\epsilon\sin(\widehat{\theta})e_\theta\cdot\psi)}{\lvert\overline{R}\rvert^{5}}
-\frac{\epsilon\sin^{2}(\widehat{\theta})(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)}{\lvert\overline{R}\rvert^{5}}
+\frac{10\epsilon^{3}\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin^{2}(\widehat{\theta})(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)}{\lvert\overline{R}\rvert^{7}}
\right) e r ⋅ ( κ ∂ θ K D ( Φ − 1 ψ ) ) = 4 π 3 ϵ κ − ∣ R ∣ 5 2 ϵ sin 2 ( 2 θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ + ϵ sin ( θ ) e θ ⋅ ψ ) − ∣ R ∣ 5 ϵ sin 2 ( θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) + ∣ R ∣ 7 10 ϵ 3 sin 2 ( 2 θ ) sin 2 ( θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ )
e ‾ θ ⋅ ( κ ^ ∂ θ K ‾ D ( Φ − 1 ψ ) ) = 3 ε κ ^ 4 π ( − 4 ε 2 sin 4 ( θ ^ 2 ) ( e θ ⋅ ψ ) ∣ R ‾ ∣ 5 − 2 ε sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 5 + 20 ε 3 sin 4 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 7 ) . \overline{e}_{\theta}\cdot\left(\widehat{\kappa}\partial_{\theta}\overline{K}_{D}\left(\Phi^{-1}\psi\right)\right)=\frac{3\varepsilon\widehat{\kappa}}{4\pi}\left(-\frac{4\varepsilon^{2}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)\left(e_{\theta}\cdot\psi\right)}{\lvert\overline{R}\rvert^{5}}-\frac{2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})\left(\widehat{s}e_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)}{\lvert\overline{R}\rvert^{5}}+\frac{20\varepsilon^{3}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})\left(\widehat{s}e_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)}{\lvert\overline{R}\rvert^{7}}\right). e θ ⋅ ( κ ∂ θ K D ( Φ − 1 ψ ) ) = 4 π 3 ε κ − ∣ R ∣ 5 4 ε 2 sin 4 ( 2 θ ) ( e θ ⋅ ψ ) − ∣ R ∣ 5 2 ε sin 2 ( 2 θ ) sin ( θ ) ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) + ∣ R ∣ 7 20 ε 3 sin 4 ( 2 θ ) sin ( θ ) ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) .
Using Lemma 4.4, we thus obtain
∣ 1 ε ∂ θ R D , 2 [ ψ ] ∣ C ˙ 0 , γ ≤ ∣ W [ ψ ; κ ^ ∂ θ K D ] ∣ C ˙ 0 , γ + ∣ ∫ − 1 / 2 1 / 2 ∫ − π π κ ^ ∂ θ K ‾ D ( Φ − 1 ψ ) ε d θ ^ d s ^ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ε − γ + ∥ ψ ∥ C 0 , γ , \begin{aligned}
\left\lvert\frac{1}{\varepsilon}\partial_{\theta}\mathcal{R}_{D,2}[\psi]\right\rvert_{\dot{C}^{0,\gamma}}\leq\left\lvert W[\psi;\widehat{\kappa}\partial_{\theta}K_{D}]\right\rvert_{\dot{C}^{0,\gamma}}+\left\lvert\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\widehat{\kappa}\partial_{\theta}\overline{K}_{D}\left(\Phi^{-1}\psi\right)\varepsilon\,d\widehat{\theta}\,d\widehat{s}\right\rvert_{\dot{C}^{0,\gamma}} \\
&\leq c\left(\kappa_{*,\gamma+},c_{\Gamma}\right)\varepsilon^{-\gamma+}\lVert\psi\rVert_{C^{0,\gamma}},
\end{aligned} ε 1 ∂ θ R D , 2 [ ψ ] C ˙ 0 , γ ≤ ∣ W [ ψ ; κ ∂ θ K D ] ∣ C ˙ 0 , γ + ∫ − 1/2 1/2 ∫ − π π κ ∂ θ K D ( Φ − 1 ψ ) ε d θ d s C ˙ 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ε − γ + ∥ ψ ∥ C 0 , γ ,
∣ 1 ε ∂ θ R D , 2 ( a ) [ ψ ] − 1 ε ∂ θ R D , 2 ( b ) [ ψ ] ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ε − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (256) \left\lvert\frac{1}{\varepsilon}\partial_{\theta}\mathcal{R}_{D,2}^{(a)}[\psi]-\frac{1}{\varepsilon}\partial_{\theta}\mathcal{R}_{D,2}^{(b)}[\psi]\right\rvert_{\dot{C}^{0,\gamma}}\leq c\left(\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma}\right)\varepsilon^{-\gamma+}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^{2,\gamma+}}\lVert\psi\rVert_{C^{0,\gamma}}. \tag*{(256)} ε 1 ∂ θ R D , 2 ( a ) [ ψ ] − ε 1 ∂ θ R D , 2 ( b ) [ ψ ] C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ε − γ + X ( a ) − X ( b ) C 2 , γ + ∥ ψ ∥ C 0 , γ . (256)
We now turn to C ˙ 0 , γ \dot{C}^{0,\gamma} C ˙ 0 , γ estimates for the s s s -derivatives ∂ s R D , 1 \partial_{s}\mathcal{R}_{D,1} ∂ s R D , 1 and ∂ s R D , 2 \partial_{s}\mathcal{R}_{D,2} ∂ s R D , 2 along Γ ε \Gamma_{\varepsilon} Γ ε . These may be written as
∂ s R D , 1 [ ψ ] = W [ ψ ; ∂ s K D ] + [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ε d θ ^ d s ^ \partial_{s}\mathcal{R}_{D,1}[\psi]=W[\psi;\partial_{s}K_{D}]+[\partial_{s},\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{D}\psi\,\varepsilon\,d\widehat{\theta}\,d\widehat{s} ∂ s R D , 1 [ ψ ] = W [ ψ ; ∂ s K D ] + [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ψ ε d θ d s
∂ s R D , 2 [ ψ ] = − Φ − 1 p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ∂ s K D ψ ε 2 κ ^ d θ ^ d s ^ − [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ε 2 κ ^ d θ ^ d s ^ . \partial_{s}\mathcal{R}_{D,2}[\psi]=-\Phi^{-1}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\partial_{s}K_{D}\psi\,\varepsilon^{2}\widehat{\kappa}\,d\widehat{\theta}\,d\widehat{s}-[\partial_{s},\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{D}\psi\,\varepsilon^{2}\widehat{\kappa}\,d\widehat{\theta}\,d\widehat{s}. ∂ s R D , 2 [ ψ ] = − Φ − 1 p.v. ∫ − 1/2 1/2 ∫ − π π ∂ s K D ψ ε 2 κ d θ d s − [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ψ ε 2 κ d θ d s .
We may estimate W [ ψ ; ∂ s K D ] W[\psi;\partial_{s}K_{D}] W [ ψ ; ∂ s K D ] by first using the definition (243) of the integrands W j W_{j} W j to write
W t ( ψ ; ∂ s K D ) = 3 4 π ( W t , 1 + W t , 2 + W t , 3 ) , W_{t}(\psi;\partial_{s}K_{D})=\frac{3}{4\pi}\left(W_{t,1}+W_{t,2}+W_{t,3}\right), W t ( ψ ; ∂ s K D ) = 4 π 3 ( W t , 1 + W t , 2 + W t , 3 ) ,
W t , 1 = ( ( 1 − ε κ ^ ) ( R ⋅ ψ ) + ( R ⋅ e t ) ( ∂ s R ⋅ ψ ) ) R ⋅ n ′ ∣ R ∣ 5 − ( R ‾ ⋅ ( Φ − 1 ψ ) + ( R ‾ ⋅ e z ) ( e t ⋅ ψ ) ) R ‾ ⋅ n ‾ ′ ∣ R ‾ ∣ 5 , W_{t,1}=\frac{\left((1-\varepsilon\widehat{\kappa})(R\cdot\psi)+(R\cdot e_{t})(\partial_{s}R\cdot\psi)\right)R\cdot n'}{\lvert R\rvert^{5}}-\frac{\left(\overline{R}\cdot(\Phi^{-1}\psi)+(\overline{R}\cdot e_{z})(e_{t}\cdot\psi)\right)\overline{R}\cdot\overline{n}'}{\lvert\overline{R}\rvert^{5}}, W t , 1 = ∣ R ∣ 5 ( ( 1 − ε κ ) ( R ⋅ ψ ) + ( R ⋅ e t ) ( ∂ s R ⋅ ψ ) ) R ⋅ n ′ − ∣ R ∣ 5 ( R ⋅ ( Φ − 1 ψ ) + ( R ⋅ e z ) ( e t ⋅ ψ ) ) R ⋅ n ′ ,
W t , 2 = ( R ⋅ e t ) ( R ⋅ ψ ) ( ∂ s R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e z ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( ∂ s R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 , W_{t,2}=\frac{(R\cdot e_{t})(R\cdot\psi)(\partial_{s}R\cdot n')}{\lvert R\rvert^{5}}-\frac{(\overline{R}\cdot e_{z})(\overline{R}\cdot(\Phi^{-1}\psi))(\partial_{s}\overline{R}\cdot\overline{n}')}{\lvert\overline{R}\rvert^{5}}, W t , 2 = ∣ R ∣ 5 ( R ⋅ e t ) ( R ⋅ ψ ) ( ∂ s R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e z ) ( R ⋅ ( Φ − 1 ψ )) ( ∂ s R ⋅ n ′ ) ,
W t , 3 = − 5 ( R ⋅ e t ) ( R ⋅ ψ ) ( R ⋅ n ′ ) R ⋅ ∂ s R ∣ R ∣ 7 + 5 ( R ‾ ⋅ e z ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ‾ ′ ) R ‾ ⋅ ∂ s R ‾ ∣ R ‾ ∣ 7 . W_{t,3}=-\frac{5(R\cdot e_{t})(R\cdot\psi)(R\cdot n')R\cdot\partial_{s}R}{\lvert R\rvert^{7}}+\frac{5(\overline{R}\cdot e_{z})(\overline{R}\cdot(\Phi^{-1}\psi))(\overline{R}\cdot\overline{n}')\overline{R}\cdot\partial_{s}\overline{R}}{\lvert\overline{R}\rvert^{7}}. W t , 3 = − ∣ R ∣ 7 5 ( R ⋅ e t ) ( R ⋅ ψ ) ( R ⋅ n ′ ) R ⋅ ∂ s R + ∣ R ∣ 7 5 ( R ⋅ e z ) ( R ⋅ ( Φ − 1 ψ )) ( R ⋅ n ′ ) R ⋅ ∂ s R .
Using the expansions (211), (212), and (231), we may expand each W t , j W_{t,j} W t , j as follows:
W t , 1 = ( s ^ 2 Q t ⋅ ψ − ε s ^ Q r ⋅ ψ + ε κ ^ ( s ^ Q ^ t ⋅ ψ − 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ) ( 2 ε sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ∣ R ∣ 5 − ( ( ε s ^ κ ^ + s ^ 2 Q R t ) ( ( 1 − ε κ ^ ) e t ⋅ ψ + ε κ 3 e θ ⋅ ψ ) − ε s ^ ( κ ^ e t ⋅ ψ − κ 3 e θ ⋅ ψ ) ) ( 2 ε sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ∣ R ∣ 5 + 2 s ^ 2 Q R n ′ ( s ^ e t ⋅ ψ + ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ∣ 5 − ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) 4 ε sin 2 ( θ ^ 2 ) ( s ^ e t ⋅ ψ + ε sin ( θ ^ 2 ) e θ ⋅ ψ ) , \begin{aligned}
W_{t,1}={}&\frac{\left(\widehat{s}^{2}Q_{t}\cdot\psi-\varepsilon\widehat{s}Q_{r}\cdot\psi+\varepsilon\widehat{\kappa}\left(\widehat{s}\widehat{Q}_{t}\cdot\psi-2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\right)\left(2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q_{Rn'}\right)}{\lvert R\rvert^{5}}\\
&-\frac{\left((\varepsilon\widehat{s}\widehat{\kappa}+\widehat{s}^{2}Q_{Rt})\left((1-\varepsilon\widehat{\kappa})e_{t}\cdot\psi+\varepsilon\kappa_{3}e_{\theta}\cdot\psi\right)-\varepsilon\widehat{s}\left(\widehat{\kappa}e_{t}\cdot\psi-\kappa_{3}e_{\theta}\cdot\psi\right)\right)\left(2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q_{Rn'}\right)}{\lvert R\rvert^{5}}\\
&+\frac{2\widehat{s}^{2}Q_{Rn'}\left(\widehat{s}e_{t}\cdot\psi+\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)}{\lvert R\rvert^{5}}-\left(\frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\overline{R}\rvert^{5}}\right)4\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\left(\widehat{s}e_{t}\cdot\psi+\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right),
\end{aligned} W t , 1 = ∣ R ∣ 5 ( s 2 Q t ⋅ ψ − ε s Q r ⋅ ψ + ε κ ( s Q t ⋅ ψ − 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ) ( 2 ε sin 2 ( 2 θ ) − s 2 Q R n ′ ) − ∣ R ∣ 5 ( ( ε s κ + s 2 Q R t ) ( ( 1 − ε κ ) e t ⋅ ψ + ε κ 3 e θ ⋅ ψ ) − ε s ( κ e t ⋅ ψ − κ 3 e θ ⋅ ψ ) ) ( 2 ε sin 2 ( 2 θ ) − s 2 Q R n ′ ) + ∣ R ∣ 5 2 s 2 Q R n ′ ( s e t ⋅ ψ + ε sin ( 2 θ ) e θ ⋅ ψ ) − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) 4 ε sin 2 ( 2 θ ) ( s e t ⋅ ψ + ε sin ( 2 θ ) e θ ⋅ ψ ) ,
W t , 2 = ( ε s ^ κ ^ + s ^ 2 Q R t ) ( s ^ Q ^ t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ κ ^ − κ 3 ε sin ( θ ^ ) + ε s ^ Q R s 3 ) ∣ R ∣ 5 − s ^ ( s ^ 2 Q t ⋅ ψ − ε s ^ Q r ⋅ ψ ) ( s ^ κ ^ − κ 3 ε sin ( θ ^ ) + ε s ^ Q R s 3 ) ∣ R ∣ 5 + ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) s ^ ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ κ ^ − κ 3 ε sin ( θ ^ ) + ε s ^ Q R s 3 ) \begin{aligned}
W_{t,2}={}&\frac{(\varepsilon\widehat{s}\widehat{\kappa}+\widehat{s}^{2}Q_{Rt})(\widehat{s}\widehat{Q}_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi)(\widehat{s}\widehat{\kappa}-\kappa_{3}\varepsilon\sin(\widehat{\theta})+\varepsilon\widehat{s}Q_{Rs3})}{\lvert R\rvert^{5}}\\
&-\frac{\widehat{s}(\widehat{s}^{2}Q_{t}\cdot\psi-\varepsilon\widehat{s}Q_{r}\cdot\psi)(\widehat{s}\widehat{\kappa}-\kappa_{3}\varepsilon\sin(\widehat{\theta})+\varepsilon\widehat{s}Q_{Rs3})}{\lvert R\rvert^{5}}\\
&+\left(\frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\overline{R}\rvert^{5}}\right)\widehat{s}\left(\widehat{s}e_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi\right)\left(\widehat{s}\widehat{\kappa}-\kappa_{3}\varepsilon\sin(\widehat{\theta})+\varepsilon\widehat{s}Q_{Rs3}\right)
\end{aligned} W t , 2 = ∣ R ∣ 5 ( ε s κ + s 2 Q R t ) ( s Q t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ( s κ − κ 3 ε sin ( θ ) + ε s Q R s 3 ) − ∣ R ∣ 5 s ( s 2 Q t ⋅ ψ − ε s Q r ⋅ ψ ) ( s κ − κ 3 ε sin ( θ ) + ε s Q R s 3 ) + ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) s ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ( s κ − κ 3 ε sin ( θ ) + ε s Q R s 3 )
W t , 3 = 5 ( ε s ^ κ ^ + s ^ 2 Q R t ) ( s ^ Q ~ t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ( 2 ε sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ Q ~ R + 2 ε 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 − 5 s ^ ( s ^ 2 Q t ⋅ ψ − ε s ^ Q r ⋅ ψ ) ( 2 ε sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ Q ~ R + 2 ε 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 − 5 s ^ 3 Q R n ′ ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ Q ~ R + 2 ε 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 + 10 s ^ ε sin 2 ( θ ^ 2 ) ( ε s Q R s 1 + s ^ 2 Q R s 2 ) ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ∣ 7 + ( 1 ∣ R ∣ 7 − 1 ∣ R ‾ ∣ 7 ) 10 s ^ ε sin 2 ( θ ^ 2 ) ( s ^ e t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ + 2 ε 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) . \begin{aligned}
W_{t,3}={}&\frac{5(\varepsilon\widehat{s}\widehat{\kappa}+\widehat{s}^{2}Q_{Rt})(\widehat{s}\widetilde{Q}_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi)(2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}QR_{n'}) (\widehat{s}\widetilde{Q}_{R}+2\varepsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta}))}{|R|^{7}}\\
&-\frac{5\widehat{s}(\widehat{s}^{2}Q_{t}\cdot\psi-\varepsilon\widehat{s}Q_{r}\cdot\psi)(2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}QR_{n'})(\widehat{s}\widetilde{Q}_{R}+2\varepsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta}))}{|R|^{7}}\\
&-\frac{5\widehat{s}^{3}QR_{n'}(\widehat{s}e_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi)(\widehat{s}\widetilde{Q}_{R}+2\varepsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta}))}{|R|^{7}}\\
&+\frac{10\widehat{s}\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\varepsilon sQ_{Rs1}+\widehat{s}^{2}QR_{s2})(\widehat{s}e_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi)}{|R|^{7}}\\
&+\left(\frac{1}{|R|^{7}}-\frac{1}{|\overline{R}|^{7}}\right)10\widehat{s}\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\widehat{s}e_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi)(\widehat{s}+2\varepsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})).
\end{aligned} W t , 3 = ∣ R ∣ 7 5 ( ε s κ + s 2 Q R t ) ( s Q t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ( 2 ε sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( s Q R + 2 ε 2 sin ( 2 θ ) sin ( θ )) − ∣ R ∣ 7 5 s ( s 2 Q t ⋅ ψ − ε s Q r ⋅ ψ ) ( 2 ε sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( s Q R + 2 ε 2 sin ( 2 θ ) sin ( θ )) − ∣ R ∣ 7 5 s 3 Q R n ′ ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ( s Q R + 2 ε 2 sin ( 2 θ ) sin ( θ )) + ∣ R ∣ 7 10 s ε sin 2 ( 2 θ ) ( ε s Q R s 1 + s 2 Q R s 2 ) ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) + ( ∣ R ∣ 7 1 − ∣ R ∣ 7 1 ) 10 s ε sin 2 ( 2 θ ) ( s e t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ( s + 2 ε 2 sin ( 2 θ ) sin ( θ )) .
As above, Q ~ t = e t − s ^ Q t + ε Q r \widetilde{Q}_{t}=e_{t}-\widehat{s}Q_{t}+\varepsilon Q_{r} Q t = e t − s Q t + ε Q r , and here we define Q ~ R = 1 + ε Q R s 1 + s ^ Q R s 2 \widetilde{Q}_{R}=1+\varepsilon QR_{s1}+\widehat{s}QR_{s2} Q R = 1 + εQ R s 1 + s Q R s 2 . Expanding 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 \frac{1}{|R|^{5}}-\frac{1}{|\overline{R}|^{5}} ∣ R ∣ 5 1 − ∣ R ∣ 5 1 and 1 ∣ R ∣ 7 − 1 ∣ R ‾ ∣ 7 \frac{1}{|R|^{7}}-\frac{1}{|\overline{R}|^{7}} ∣ R ∣ 7 1 − ∣ R ∣ 7 1 using (209) and (210), we may use Lemma 4.4 to estimate
∣ W ( ψ ; ∂ s K D ) ⋅ e z ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) max j ( ∥ Q j ∥ C ˙ 0 , γ + + ∥ Q j ∥ C 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ε γ + ∥ ψ ∥ C 0 , γ . (257) \left|W(\psi;\partial_{s}K_{D})\cdot e_{z}\right|_{\dot{C}^{0,\gamma}}
\leq c(\kappa_{*},c_{\Gamma})\max_{j}\left(\|Q_{j}\|_{\dot{C}^{0,\gamma_{+}}}+\|Q_{j}\|_{C^{0,\gamma}}\right)\|\psi\|_{C^{0,\gamma}}
\leq c(\kappa_{*},\gamma_{+},c_{\Gamma})\varepsilon^{\gamma_{+}}\|\psi\|_{C^{0,\gamma}}. \tag*{(257)} ∣ W ( ψ ; ∂ s K D ) ⋅ e z ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) j max ( ∥ Q j ∥ C ˙ 0 , γ + + ∥ Q j ∥ C 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ε γ + ∥ ψ ∥ C 0 , γ . (257)
Furthermore, using that each remainder term Q i Q_{i} Q i satisfies a Lipschitz estimate of the form (213), by the above expansions and Lemma 4.4 we have
∣ W ( ψ ; ∂ s K D ( a ) − ∂ s K D ( b ) ) ⋅ e z ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ε γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (258) \left|W(\psi;\partial_{s}K_{D}^{(a)}-\partial_{s}K_{D}^{(b)})\cdot e_{z}\right|_{\dot{C}^{0,\gamma}}
\leq c(\kappa_{*,\gamma_{+}}^{(a)},\kappa_{*,\gamma_{+}}^{(b)},c_{\Gamma})\varepsilon^{\gamma_{+}}\|X^{(a)}-X^{(b)}\|_{C^{2,\gamma_{+}}}\|\psi\|_{C^{0,\gamma}}. \tag*{(258)} W ( ψ ; ∂ s K D ( a ) − ∂ s K D ( b ) ) ⋅ e z C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ε γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . (258)
We next consider W r ( ψ ; ∂ s K D ) W_{r}(\psi;\partial_{s}K_{D}) W r ( ψ ; ∂ s K D ) as given by (243). We write
W r ( ψ ; ∂ s K D ) = 3 4 π ( W r , 1 + W r , 2 + W r , 3 ) , W_{r}(\psi;\partial_{s}K_{D})=\frac{3}{4\pi}(W_{r,1}+W_{r,2}+W_{r,3}), W r ( ψ ; ∂ s K D ) = 4 π 3 ( W r , 1 + W r , 2 + W r , 3 ) ,
W r , 1 = ( R ⋅ e r ) ( ∂ s R ⋅ ψ ) R ⋅ n ′ ∣ R ∣ 5 − ( R ‾ ⋅ e ‾ r ) ( e t ⋅ ψ ) R ‾ ⋅ n ‾ ′ ∣ R ‾ ∣ 5 , W_{r,1}=\frac{(R\cdot e_{r})(\partial_{s}R\cdot\psi)R\cdot n'}{|R|^{5}}-\frac{(\overline{R}\cdot\overline{e}_{r})(e_{t}\cdot\psi)\overline{R}\cdot\overline{n}'}{|\overline{R}|^{5}}, W r , 1 = ∣ R ∣ 5 ( R ⋅ e r ) ( ∂ s R ⋅ ψ ) R ⋅ n ′ − ∣ R ∣ 5 ( R ⋅ e r ) ( e t ⋅ ψ ) R ⋅ n ′ ,
W r , 2 = ( R ⋅ e r ) ( R ⋅ ψ ) ( ∂ s R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e ‾ r ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( ∂ s R ‾ ⋅ n ‾ ′ ) ∣ R ‾ ∣ 5 , W_{r,2}=\frac{(R\cdot e_{r})(R\cdot\psi)(\partial_{s}R\cdot n')}{|R|^{5}}-\frac{(\overline{R}\cdot\overline{e}_{r})(\overline{R}\cdot(\Phi^{-1}\psi))(\partial_{s}\overline{R}\cdot\overline{n}')}{|\overline{R}|^{5}}, W r , 2 = ∣ R ∣ 5 ( R ⋅ e r ) ( R ⋅ ψ ) ( ∂ s R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e r ) ( R ⋅ ( Φ − 1 ψ )) ( ∂ s R ⋅ n ′ ) ,
W r , 3 = − 5 ( R ⋅ e r ) ( R ⋅ ψ ) ( R ⋅ n ′ ) R ⋅ ∂ s R ∣ R ∣ 7 + 5 ( R ‾ ⋅ e ‾ r ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ‾ ′ ) R ‾ ⋅ ∂ s R ‾ ∣ R ‾ ∣ 7 . W_{r,3}=-\frac{5(R\cdot e_{r})(R\cdot\psi)(R\cdot n')R\cdot\partial_{s}R}{|R|^{7}}+\frac{5(\overline{R}\cdot\overline{e}_{r})(\overline{R}\cdot(\Phi^{-1}\psi))(\overline{R}\cdot\overline{n}')\overline{R}\cdot\partial_{s}\overline{R}}{|\overline{R}|^{7}}. W r , 3 = − ∣ R ∣ 7 5 ( R ⋅ e r ) ( R ⋅ ψ ) ( R ⋅ n ′ ) R ⋅ ∂ s R + ∣ R ∣ 7 5 ( R ⋅ e r ) ( R ⋅ ( Φ − 1 ψ )) ( R ⋅ n ′ ) R ⋅ ∂ s R .
Using (211), (212), and (231), we may expand each W r , j W_{r,j} W r , j as
W r , 1 = − s ^ 2 Q R n ( ( 1 − ε κ ^ ) e t ⋅ ψ + ε κ 3 e θ ⋅ ψ ) ( 2 ε sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ∣ R ∣ 5 + 2 ε sin 2 ( θ ^ 2 ) ( ε κ ^ e t ⋅ ψ − ε κ 3 e θ ⋅ ψ ) ( 2 ε sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ∣ R ∣ 5 + 2 ε sin 2 ( θ ^ 2 ) s ^ 2 Q R n ′ ( e t ⋅ ψ ) ∣ R ∣ 5 − ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) 4 ε 2 sin 4 ( θ ^ 2 ) ( e t ⋅ ψ ) , \begin{aligned}
W_{r,1}={}&-\frac{\widehat{s}^{2}QR_{n}((1-\varepsilon\widehat{\kappa})e_{t}\cdot\psi+\varepsilon\kappa_{3}e_{\theta}\cdot\psi)(2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}QR_{n'})}{|R|^{5}}\\
&+\frac{2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\varepsilon\widehat{\kappa}e_{t}\cdot\psi-\varepsilon\kappa_{3}e_{\theta}\cdot\psi)(2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}QR_{n'})}{|R|^{5}}\\
&+\frac{2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\widehat{s}^{2}QR_{n'}(e_{t}\cdot\psi)}{|R|^{5}}-\left(\frac{1}{|R|^{5}}-\frac{1}{|\overline{R}|^{5}}\right)4\varepsilon^{2}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)(e_{t}\cdot\psi),
\end{aligned} W r , 1 = − ∣ R ∣ 5 s 2 Q R n (( 1 − ε κ ) e t ⋅ ψ + ε κ 3 e θ ⋅ ψ ) ( 2 ε sin 2 ( 2 θ ) − s 2 Q R n ′ ) + ∣ R ∣ 5 2 ε sin 2 ( 2 θ ) ( ε κ e t ⋅ ψ − ε κ 3 e θ ⋅ ψ ) ( 2 ε sin 2 ( 2 θ ) − s 2 Q R n ′ ) + ∣ R ∣ 5 2 ε sin 2 ( 2 θ ) s 2 Q R n ′ ( e t ⋅ ψ ) − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) 4 ε 2 sin 4 ( 2 θ ) ( e t ⋅ ψ ) ,
W r , 2 = s ^ 2 Q R n ( s ^ Q ~ t ⋅ ψ + 2 ε sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ κ ^ − κ 3 ε sin ( θ ^ ) + ε s ^ Q R s 3 ) ∣ R ∣ 5 − 2 ε sin 2 ( θ ^ 2 ) ( s ^ 2 Q t ⋅ ψ − ε s ^ Q r ⋅ ψ ) ( s ^ κ ^ − κ 3 ε sin ( θ ^ ) + ε s ^ Q R s 3 ) ∣ R ∣ 5 + ( 1 ∣ R ∣ 5 − 1 ∣ R ~ ∣ 5 ) 2 ϵ sin 2 ( θ ^ 2 ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ κ ^ − κ 3 ϵ sin ( θ ^ ) + ϵ s ^ Q R s 3 ) \begin{aligned}
W_{r,2}={}&\frac{\widehat{s}^{2}QR_{n}(\widehat{s}\widetilde{Q}_{t}\cdot\psi+2\varepsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\psi)(\widehat{s}\widehat{\kappa}-\kappa_{3}\varepsilon\sin(\widehat{\theta})+\varepsilon\widehat{s}QR_{s3})}{|R|^{5}}\\
&-\frac{2\varepsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\widehat{s}^{2}Q_{t}\cdot\psi-\varepsilon\widehat{s}Q_{r}\cdot\psi)(\widehat{s}\widehat{\kappa}-\kappa_{3}\varepsilon\sin(\widehat{\theta})+\varepsilon\widehat{s}QR_{s3})}{|R|^{5}} \\
&+\left(\frac{1}{|R|^{5}}-\frac{1}{|\widetilde{R}|^{5}}\right)2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\widehat{s}e_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(\widehat{s}\widehat{\kappa}-\kappa_3\epsilon\sin(\widehat{\theta})+\epsilon\widehat{s}Q R_{s3})
\end{aligned} W r , 2 = ∣ R ∣ 5 s 2 Q R n ( s Q t ⋅ ψ + 2 ε sin ( 2 θ ) e θ ⋅ ψ ) ( s κ − κ 3 ε sin ( θ ) + ε s Q R s 3 ) − ∣ R ∣ 5 2 ε sin 2 ( 2 θ ) ( s 2 Q t ⋅ ψ − ε s Q r ⋅ ψ ) ( s κ − κ 3 ε sin ( θ ) + ε s Q R s 3 ) + ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) 2 ϵ sin 2 ( 2 θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( s κ − κ 3 ϵ sin ( θ ) + ϵ s Q R s 3 )
W r , 3 = 5 s ^ 2 Q R n ( s ^ Q ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ Q ~ R + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 − 10 ϵ sin 2 ( θ ^ 2 ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ Q ~ R + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 − 10 ϵ sin 2 ( θ ^ 2 ) s ^ 2 Q R n ′ ( s ^ e ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ Q ~ R + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 + 20 ϵ 2 sin 4 ( θ ^ 2 ) ( ϵ s ^ Q R s 1 + s ^ 2 Q R s 2 ) ( s ^ e ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ∣ 7 + ( 1 ∣ R ∣ 7 − 1 ∣ R ~ ∣ 7 ) 20 ϵ 2 sin 4 ( θ ^ 2 ) ( s ^ e ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) . W_{r,3}=\frac{5\widehat{s}^{2}Q R_n(\widehat{s}\widehat{Q}_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q R_{n'})(\widehat{s}\widetilde{Q}_R+2\epsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta}))}{|R|^{7}}
-\frac{10\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\widehat{s}^{2}Q_t\cdot\psi-\epsilon\widehat{s}Q_r\cdot\psi)(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q R_{n'})(\widehat{s}\widetilde{Q}_R+2\epsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta}))}{|R|^{7}}
-\frac{10\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\widehat{s}^{2}Q R_{n'}(\widehat{s}\widehat{e}_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(\widehat{s}\widetilde{Q}_R+2\epsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta}))}{|R|^{7}}
+\frac{20\epsilon^{2}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)(\epsilon\widehat{s}Q R_{s1}+\widehat{s}^{2}Q R_{s2})(\widehat{s}\widehat{e}_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)}{|R|^{7}}
+\left(\frac{1}{|R|^{7}}-\frac{1}{|\widetilde{R}|^{7}}\right)20\epsilon^{2}\sin^{4}\left(\frac{\widehat{\theta}}{2}\right)(\widehat{s}\widehat{e}_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(\widehat{s}+2\epsilon^{2}\sin\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})). W r , 3 = ∣ R ∣ 7 5 s 2 Q R n ( s Q t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( s Q R + 2 ϵ 2 sin ( 2 θ ) sin ( θ )) − ∣ R ∣ 7 10 ϵ sin 2 ( 2 θ ) ( s 2 Q t ⋅ ψ − ϵ s Q r ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) ( s Q R + 2 ϵ 2 sin ( 2 θ ) sin ( θ )) − ∣ R ∣ 7 10 ϵ sin 2 ( 2 θ ) s 2 Q R n ′ ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( s Q R + 2 ϵ 2 sin ( 2 θ ) sin ( θ )) + ∣ R ∣ 7 20 ϵ 2 sin 4 ( 2 θ ) ( ϵ s Q R s 1 + s 2 Q R s 2 ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) + ( ∣ R ∣ 7 1 − ∣ R ∣ 7 1 ) 20 ϵ 2 sin 4 ( 2 θ ) ( s e t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( s + 2 ϵ 2 sin ( 2 θ ) sin ( θ )) .
Again we have Q ~ t = e t − s ^ Q ^ t + ϵ Q r \widetilde{Q}_t=e_t-\widehat{s}\widehat{Q}_t+\epsilon Q_r Q t = e t − s Q t + ϵ Q r and Q ~ R = 1 + ϵ Q R s 1 + s ^ Q R s 2 \widetilde{Q}_R=1+\epsilon Q R_{s1}+\widehat{s}Q R_{s2} Q R = 1 + ϵ Q R s 1 + s Q R s 2 . Expanding 1 ∣ R ∣ 5 − 1 ∣ R ~ ∣ 5 \frac{1}{|R|^{5}}-\frac{1}{|\widetilde{R}|^{5}} ∣ R ∣ 5 1 − ∣ R ∣ 5 1 and 1 ∣ R ∣ 7 − 1 ∣ R ~ ∣ 7 \frac{1}{|R|^{7}}-\frac{1}{|\widetilde{R}|^{7}} ∣ R ∣ 7 1 − ∣ R ∣ 7 1 via (209) and (210), by Lemma 4.4 we have
∣ W ( ψ ; ∂ s K D ) ⋅ e r ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) max j ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) e − γ + t ∥ ψ ∥ C 0 , γ , (259) \left|W(\psi;\partial_s K_D)\cdot e_r\right|_{\dot{C}^{0,\gamma}}
\le c(\kappa_*,c_\Gamma)\max_j\left(\|Q_j\|_{C_1^{0,\gamma_+}}+\|Q_j\|_{C_2^{0,\gamma}}\right)\|\psi\|_{C^{0,\gamma}}
\le c(\kappa_{*,\gamma_+},c_\Gamma)e^{-\gamma_+t}\|\psi\|_{C^{0,\gamma}}, \tag*{(259)} ∣ W ( ψ ; ∂ s K D ) ⋅ e r ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) j max ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) e − γ + t ∥ ψ ∥ C 0 , γ , (259)
∣ W ( ψ ; ∂ s K D ( a ) − ∂ s K D ( b ) ) ⋅ e r ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) e − γ + t ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . \left|W(\psi;\partial_s K_D^{(a)}-\partial_s K_D^{(b)})\cdot e_r\right|_{\dot{C}^{0,\gamma}}
\le c(\kappa_{*,\gamma_+}^{(a)},\kappa_{*,\gamma_+}^{(b)},c_\Gamma)e^{-\gamma_+t}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma_+}}\|\psi\|_{C^{0,\gamma}}. W ( ψ ; ∂ s K D ( a ) − ∂ s K D ( b ) ) ⋅ e r C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) e − γ + t X ( a ) − X ( b ) C 2 , γ + ∥ ψ ∥ C 0 , γ .
We finally turn to W θ ( ψ ; ∂ s K D ) W_\theta(\psi;\partial_s K_D) W θ ( ψ ; ∂ s K D ) as in (243), which we may write as
W θ ( ψ ; ∂ s K D ) = 3 4 π ( W θ , 1 + W θ , 2 + W θ , 3 ) , W_\theta(\psi;\partial_s K_D)=\frac{3}{4\pi}(W_{\theta,1}+W_{\theta,2}+W_{\theta,3}), W θ ( ψ ; ∂ s K D ) = 4 π 3 ( W θ , 1 + W θ , 2 + W θ , 3 ) ,
W θ , 1 = ( ϵ κ 3 ( R ⋅ ψ ) + ( R ⋅ e θ ) ( ∂ s R ⋅ ψ ) ) R ⋅ n ′ ∣ R ∣ 5 − ( R ‾ ⋅ e θ ‾ ) ( e t ⋅ ψ ) R ‾ ⋅ n ′ ‾ ∣ R ‾ ∣ 5 , W_{\theta,1}=\frac{(\epsilon\kappa_3(R\cdot\psi)+(R\cdot e_\theta)(\partial_sR\cdot\psi))R\cdot n'}{|R|^{5}}-\frac{(\overline{R}\cdot\overline{e_\theta})(e_t\cdot\psi)\overline{R}\cdot\overline{n'}}{|\overline{R}|^{5}}, W θ , 1 = ∣ R ∣ 5 ( ϵ κ 3 ( R ⋅ ψ ) + ( R ⋅ e θ ) ( ∂ s R ⋅ ψ )) R ⋅ n ′ − ∣ R ∣ 5 ( R ⋅ e θ ) ( e t ⋅ ψ ) R ⋅ n ′ ,
W θ , 2 = ( R ⋅ e θ ) ( R ⋅ ψ ) ( ∂ s R ⋅ n ′ ) ∣ R ∣ 5 − ( R ‾ ⋅ e θ ‾ ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( ∂ s R ‾ ⋅ n ′ ‾ ) ∣ R ‾ ∣ 5 , W_{\theta,2}=\frac{(R\cdot e_\theta)(R\cdot\psi)(\partial_sR\cdot n')}{|R|^{5}}-\frac{(\overline{R}\cdot\overline{e_\theta})(\overline{R}\cdot(\Phi^{-1}\psi))(\partial_s\overline{R}\cdot\overline{n'})}{|\overline{R}|^{5}}, W θ , 2 = ∣ R ∣ 5 ( R ⋅ e θ ) ( R ⋅ ψ ) ( ∂ s R ⋅ n ′ ) − ∣ R ∣ 5 ( R ⋅ e θ ) ( R ⋅ ( Φ − 1 ψ )) ( ∂ s R ⋅ n ′ ) ,
W θ , 3 = − 5 ( R ⋅ e θ ) ( R ⋅ ψ ) ( R ⋅ n ′ ) R ⋅ ∂ s R ∣ R ∣ 7 + 5 ( R ‾ ⋅ e θ ‾ ) ( R ‾ ⋅ ( Φ − 1 ψ ) ) ( R ‾ ⋅ n ′ ‾ ) R ‾ ⋅ ∂ s R ‾ ∣ R ‾ ∣ 7 . W_{\theta,3}=-\frac{5(R\cdot e_\theta)(R\cdot\psi)(R\cdot n')R\cdot\partial_sR}{|R|^{7}}+\frac{5(\overline{R}\cdot\overline{e_\theta})(\overline{R}\cdot(\Phi^{-1}\psi))(\overline{R}\cdot\overline{n'})\overline{R}\cdot\partial_s\overline{R}}{|\overline{R}|^{7}}. W θ , 3 = − ∣ R ∣ 7 5 ( R ⋅ e θ ) ( R ⋅ ψ ) ( R ⋅ n ′ ) R ⋅ ∂ s R + ∣ R ∣ 7 5 ( R ⋅ e θ ) ( R ⋅ ( Φ − 1 ψ )) ( R ⋅ n ′ ) R ⋅ ∂ s R .
By (211), (212), and (231), we may expand W θ , j W_{\theta,j} W θ , j as
W θ , 1 = − ( ϵ s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ) ( ( 1 − ϵ κ ^ ) e t ⋅ ψ + ϵ κ 3 e θ ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ∣ R ∣ 5 − ϵ κ 3 ( s ^ Q ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ∣ R ∣ 5 + ϵ sin ( θ ^ ) ( ϵ κ ^ e t ⋅ ψ + ϵ κ 3 e θ ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ∣ R ∣ 5 + ϵ sin ( θ ^ ) s ^ 2 Q R n ′ ( e t ⋅ ψ ) ∣ R ∣ 5 − ( 1 ∣ R ∣ 5 − 1 ∣ R ‾ ∣ 5 ) 2 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( e t ⋅ ψ ) , W_{\theta,1}=-\frac{(\epsilon\widehat{s}\kappa_3\cos(\theta-\widehat{\theta})+\widehat{s}^{2}Q R_\theta)((1-\epsilon\widehat{\kappa})e_t\cdot\psi+\epsilon\kappa_3e_\theta\cdot\psi)(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q R_{n'})}{|R|^{5}}
-\frac{\epsilon\kappa_3(\widehat{s}\widehat{Q}_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q R_{n'})}{|R|^{5}}
+\frac{\epsilon\sin(\widehat{\theta})(\epsilon\widehat{\kappa}e_t\cdot\psi+\epsilon\kappa_3e_\theta\cdot\psi)(2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)-\widehat{s}^{2}Q R_{n'})}{|R|^{5}}
+\frac{\epsilon\sin(\widehat{\theta})\widehat{s}^{2}Q R_{n'}(e_t\cdot\psi)}{|R|^{5}}
-\left(\frac{1}{|R|^{5}}-\frac{1}{|\overline{R}|^{5}}\right)2\epsilon^{2}\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})(e_t\cdot\psi), W θ , 1 = − ∣ R ∣ 5 ( ϵ s κ 3 cos ( θ − θ ) + s 2 Q R θ ) (( 1 − ϵ κ ) e t ⋅ ψ + ϵ κ 3 e θ ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) − ∣ R ∣ 5 ϵ κ 3 ( s Q t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) + ∣ R ∣ 5 ϵ sin ( θ ) ( ϵ κ e t ⋅ ψ + ϵ κ 3 e θ ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ) − s 2 Q R n ′ ) + ∣ R ∣ 5 ϵ sin ( θ ) s 2 Q R n ′ ( e t ⋅ ψ ) − ( ∣ R ∣ 5 1 − ∣ R ∣ 5 1 ) 2 ϵ 2 sin 2 ( 2 θ ) sin ( θ ) ( e t ⋅ ψ ) ,
W θ , 2 = ( ϵ s ^ κ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ) ( s ^ Q ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ κ ^ − κ 3 ϵ sin ( θ ^ ) + ϵ s ^ Q R s 3 ) ∣ R ∣ 5 (260) W_{\theta,2}=\frac{(\epsilon\widehat{s}\kappa_3\cos(\theta-\widehat{\theta})+\widehat{s}^{2}Q R_\theta)(\widehat{s}\widehat{Q}_t\cdot\psi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_\theta\cdot\psi)(\widehat{s}\widehat{\kappa}-\kappa_3\epsilon\sin(\widehat{\theta})+\epsilon\widehat{s}Q R_{s3})}{|R|^{5}} \tag*{(260)} W θ , 2 = ∣ R ∣ 5 ( ϵ s κ 3 cos ( θ − θ ) + s 2 Q R θ ) ( s Q t ⋅ ψ + 2 ϵ sin ( 2 θ ) e θ ⋅ ψ ) ( s κ − κ 3 ϵ sin ( θ ) + ϵ s Q R s 3 ) (260)
− ϵ sin ( θ ^ ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( s ^ κ ^ − κ 3 ϵ sin ( θ ^ ) + ϵ s ^ Q R s 3 ) ∣ R ∣ 5 + ( 1 ∣ R ∣ 5 − 1 ∣ R ˉ ∣ 5 ) ϵ sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ κ ^ − κ 3 ϵ sin ( θ ^ ) + ϵ s ^ Q R s 3 ) -\frac{\epsilon\sin(\hat{\theta})(\hat{s}^{2}Q_{\mathsf{t}}\cdot\psi-\epsilon\hat{s}Q_{\mathsf{r}}\cdot\psi)(\hat{s}\hat{\kappa}-\kappa_{3}\epsilon\sin(\hat{\theta})+\epsilon\hat{s}Q R_{\mathsf{s3}})}{\lvert R\rvert^{5}}
+\left(\frac{1}{\lvert R\rvert^{5}}-\frac{1}{\lvert\bar{R}\rvert^{5}}\right)\epsilon\sin(\hat{\theta})(\hat{s}e_{\mathsf{t}}\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_{\theta}\cdot\psi)(\hat{s}\hat{\kappa}-\kappa_{3}\epsilon\sin(\hat{\theta})+\epsilon\hat{s}Q R_{\mathsf{s3}}) − ∣ R ∣ 5 ϵ sin ( θ ^ ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( s ^ κ ^ − κ 3 ϵ sin ( θ ^ ) + ϵ s ^ Q R s3 ) + ( ∣ R ∣ 5 1 − ∣ R ˉ ∣ 5 1 ) ϵ sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) ( s ^ κ ^ − κ 3 ϵ sin ( θ ^ ) + ϵ s ^ Q R s3 )
W θ , 3 = 5 ( ϵ s ^ κ ^ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ) ( s ^ Q ~ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ Q ~ R + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 − 5 ϵ sin ( θ ^ ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ Q ~ R + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 − 5 ϵ sin ( θ ^ ) s ^ 2 Q R n ′ ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ Q ~ R + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) ∣ R ∣ 7 + 10 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( ϵ s ^ Q R s 1 + s ^ 2 Q R s 2 ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ∣ 7 + ( 1 ∣ R ∣ 7 − 1 ∣ R ˉ ∣ 7 ) 10 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ( s ^ + 2 ϵ 2 sin ( θ ^ 2 ) sin ( θ ^ ) ) . W_{\theta,3}=
\frac{5(\epsilon\hat{s}\hat{\kappa}_{3}\cos(\theta-\hat{\theta})+\hat{s}^{2}Q R_{\theta})(\hat{s}\widetilde{Q}_{\mathsf{t}}\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_{\theta}\cdot\psi)(2\epsilon\sin^{2}\left(\frac{\hat{\theta}}{2}\right)-\hat{s}^{2}Q R_{\mathsf{n'}})(\hat{s}\widetilde{Q}R+2\epsilon^{2}\sin\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta}))}{\lvert R\rvert^{7}}
-\frac{5\epsilon\sin(\hat{\theta})(\hat{s}^{2}Q_{\mathsf{t}}\cdot\psi-\epsilon\hat{s}Q_{\mathsf{r}}\cdot\psi)(2\epsilon\sin^{2}\left(\frac{\hat{\theta}}{2}\right)-\hat{s}^{2}Q R_{\mathsf{n'}})(\hat{s}\widetilde{Q}R+2\epsilon^{2}\sin\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta}))}{\lvert R\rvert^{7}}
-\frac{5\epsilon\sin(\hat{\theta})\hat{s}^{2}Q R_{\mathsf{n'}}(\hat{s}e_{\mathsf{t}}\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_{\theta}\cdot\psi)(\hat{s}\widetilde{Q}R+2\epsilon^{2}\sin\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta}))}{\lvert R\rvert^{7}}
+\frac{10\epsilon^{2}\sin^{2}\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta})(\epsilon\hat{s}Q R_{\mathsf{s1}}+\hat{s}^{2}Q R_{\mathsf{s2}})(\hat{s}e_{\mathsf{t}}\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_{\theta}\cdot\psi)}{\lvert R\rvert^{7}}
+\left(\frac{1}{\lvert R\rvert^{7}}-\frac{1}{\lvert\bar{R}\rvert^{7}}\right)10\epsilon^{2}\sin^{2}\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta})(\hat{s}e_{\mathsf{t}}\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_{\theta}\cdot\psi)(\hat{s}+2\epsilon^{2}\sin\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta})). W θ , 3 = ∣ R ∣ 7 5 ( ϵ s ^ κ ^ 3 cos ( θ − θ ^ ) + s ^ 2 Q R θ ) ( s ^ Q t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ^ ) − s ^ 2 Q R n ′ ) ( s ^ Q R + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) − ∣ R ∣ 7 5 ϵ sin ( θ ^ ) ( s ^ 2 Q t ⋅ ψ − ϵ s ^ Q r ⋅ ψ ) ( 2 ϵ sin 2 ( 2 θ ^ ) − s ^ 2 Q R n ′ ) ( s ^ Q R + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) − ∣ R ∣ 7 5 ϵ sin ( θ ^ ) s ^ 2 Q R n ′ ( s ^ e t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) ( s ^ Q R + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) + ∣ R ∣ 7 10 ϵ 2 sin 2 ( 2 θ ^ ) sin ( θ ^ ) ( ϵ s ^ Q R s1 + s ^ 2 Q R s2 ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) + ( ∣ R ∣ 7 1 − ∣ R ˉ ∣ 7 1 ) 10 ϵ 2 sin 2 ( 2 θ ^ ) sin ( θ ^ ) ( s ^ e t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) ( s ^ + 2 ϵ 2 sin ( 2 θ ^ ) sin ( θ ^ )) .
We again expand 1 ∣ R ∣ 7 − 1 ∣ R ˉ ∣ 7 \frac{1}{\lvert R\rvert^{7}}-\frac{1}{\lvert\bar{R}\rvert^{7}} ∣ R ∣ 7 1 − ∣ R ˉ ∣ 7 1 and 1 ∣ R ∣ 7 − 1 ∣ R ˉ ∣ 7 \frac{1}{\lvert R\rvert^{7}}-\frac{1}{\lvert\bar{R}\rvert^{7}} ∣ R ∣ 7 1 − ∣ R ˉ ∣ 7 1 using (209) and (210), and apply Lemma 4.4 to obtain
∣ W ( ψ ; ∂ s K D ) ⋅ e θ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) max j ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ , (261) \left\lvert W(\psi;\partial_{s}K_{D})\cdot e_{\theta}\right\rvert_{\dot{C}^{0,\gamma}}
\le c(\kappa_{*},c_{\Gamma})\max_{j}\left(\lVert Q_{j}\rVert_{C_{1}^{0,\gamma+}}+\lVert Q_{j}\rVert_{C_{2}^{0,\gamma}}\right)\lVert\psi\rVert_{C^{0,\gamma}}
\le c(\kappa_{*,\gamma+},c_{\Gamma})\epsilon^{-\gamma+}\lVert\psi\rVert_{C^{0,\gamma}}, \tag*{(261)} ∣ W ( ψ ; ∂ s K D ) ⋅ e θ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , c Γ ) j max ( ∥ Q j ∥ C 1 0 , γ + + ∥ Q j ∥ C 2 0 , γ ) ∥ ψ ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ , (261)
∣ W ( ψ ; ∂ s K D ( a ) − ∂ s K D ( b ) ) ⋅ e θ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . \left\lvert W(\psi;\partial_{s}K_{D}^{(a)}-\partial_{s}K_{D}^{(b)})\cdot e_{\theta}\right\rvert_{\dot{C}^{0,\gamma}}
\le c(\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\epsilon^{-\gamma+}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^{2,\gamma+}}\lVert\psi\rVert_{C^{0,\gamma}}. W ( ψ ; ∂ s K D ( a ) − ∂ s K D ( b ) ) ⋅ e θ C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + X ( a ) − X ( b ) C 2 , γ + ∥ ψ ∥ C 0 , γ .
Next, noting that, by the form (211) of R ⋅ n ′ R\cdot n' R ⋅ n ′ , we have
K D = 3 4 π R ⊗ R ( R ⋅ n ′ ) ∣ R ∣ 5 = 3 4 π R ⊗ R ( − 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ′ ) ∣ R ∣ 5 , (262) K_{D}=\frac{3}{4\pi}\frac{R\otimes R(R\cdot n')}{\lvert R\rvert^{5}}=\frac{3}{4\pi}\frac{R\otimes R(-2\epsilon\sin^{2}\left(\frac{\hat{\theta}}{2}\right)+\hat{s}^{2}Q R_{\mathsf{n'}})}{\lvert R\rvert^{5}}, \tag*{(262)} K D = 4 π 3 ∣ R ∣ 5 R ⊗ R ( R ⋅ n ′ ) = 4 π 3 ∣ R ∣ 5 R ⊗ R ( − 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ′ ) , (262)
we may use the commutator bounds (78) and Lemmas 4.2 and 4.4 to estimate the second term of R D , 1 \mathcal{R}_{D,1} R D , 1 involving [ ∂ s , Φ − 1 ] [\partial_{s},\Phi^{-1}] [ ∂ s , Φ − 1 ] :
∣ [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ϵ d θ ^ d s ^ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ ) ∥ ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ϵ d θ ^ d s ^ ∥ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − γ ∥ ψ ∥ L ∞ , \left\lvert[\partial_{s},\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{D}\psi\,\epsilon\,d\hat{\theta}\,d\hat{s}\right\rvert_{\dot{C}^{0,\gamma}}
\le c(\kappa_{*,\gamma})\left\lVert\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{D}\psi\,\epsilon\,d\hat{\theta}\,d\hat{s}\right\rVert_{C^{0,\gamma}}
\le c(\kappa_{*,\gamma},c_{\Gamma})\epsilon^{-\gamma}\lVert\psi\rVert_{L^{\infty}}, [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ψ ϵ d θ ^ d s ^ C ˙ 0 , γ ≤ c ( κ ∗ , γ ) ∫ − 1/2 1/2 ∫ − π π K D ψ ϵ d θ ^ d s ^ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − γ ∥ ψ ∥ L ∞ ,
∣ [ ∂ s , ( Φ ( a ) ) − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ( a ) ψ ϵ d θ ^ d s ^ − [ ∂ s , ( Φ ( b ) ) − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ( b ) ψ ϵ d θ ^ d s ^ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ ψ ∥ L ∞ . (263) \left\lvert[\partial_{s},(\Phi^{(a)})^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{D}^{(a)}\psi\,\epsilon\,d\hat{\theta}\,d\hat{s}-[\partial_{s},(\Phi^{(b)})^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{D}^{(b)}\psi\,\epsilon\,d\hat{\theta}\,d\hat{s}\right\rvert_{\dot{C}^{0,\gamma}}
\le c(\kappa_{*,\gamma}^{(a)},\kappa_{*,\gamma}^{(b)},c_{\Gamma})\epsilon^{-\gamma}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^{2,\gamma}}\lVert\psi\rVert_{L^{\infty}}. \tag*{(263)} [ ∂ s , ( Φ ( a ) ) − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ( a ) ψ ϵ d θ ^ d s ^ − [ ∂ s , ( Φ ( b ) ) − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ( b ) ψ ϵ d θ ^ d s ^ C ˙ 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − γ X ( a ) − X ( b ) C 2 , γ ∥ ψ ∥ L ∞ . (263)
We may then combine (256), (257), (258), (259), and (261) to obtain
∣ ∂ s R D , 1 [ ψ ] ∣ C ˙ 0 , γ ≤ ∣ W ( ψ ; ∂ s K D ) ∣ C ˙ 0 , γ + ∣ [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ϵ d θ ^ d s ^ ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ , (264) \left\lvert\partial_{s}\mathcal{R}_{D,1}[\psi]\right\rvert_{\dot{C}^{0,\gamma}}
\le\left\lvert W(\psi;\partial_{s}K_{D})\right\rvert_{\dot{C}^{0,\gamma}}+
\left\lvert[\partial_{s},\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{D}\psi\,\epsilon\,d\hat{\theta}\,d\hat{s}\right\rvert_{\dot{C}^{0,\gamma}}
\le c(\kappa_{*,\gamma+},c_{\Gamma})\epsilon^{-\gamma+}\lVert\psi\rVert_{C^{0,\gamma}}, \tag*{(264)} ∣ ∂ s R D , 1 [ ψ ] ∣ C ˙ 0 , γ ≤ ∣ W ( ψ ; ∂ s K D ) ∣ C ˙ 0 , γ + [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ψ ϵ d θ ^ d s ^ C ˙ 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ , (264)
∣ ∂ s R D , 1 ( a ) [ ψ ] − ∂ s R D , 1 ( b ) [ ψ ] ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ ψ ∥ C 0 , γ . \left\lvert\partial_{s}\mathcal{R}_{D,1}^{(a)}[\psi]-\partial_{s}\mathcal{R}_{D,1}^{(b)}[\psi]\right\rvert_{\dot{C}^{0,\gamma}}
\le c(\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\epsilon^{-\gamma+}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^{2,\gamma+}}\lVert\psi\rVert_{C^{0,\gamma}}. ∂ s R D , 1 ( a ) [ ψ ] − ∂ s R D , 1 ( b ) [ ψ ] C ˙ 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + X ( a ) − X ( b ) C 2 , γ + ∥ ψ ∥ C 0 , γ .
We finally turn to ∂ s R D , 2 [ ψ ] \partial_s\mathcal{R}_{D,2}[\psi] ∂ s R D , 2 [ ψ ] , which we write as
∂ s R D , 2 [ ψ ] = − W [ ψ ; ϵ κ ^ ∂ s K ‾ D ] − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ϵ κ ^ ∂ s K ‾ D ( Φ − 1 ψ ) ( s − s ^ , θ − θ ^ ) ϵ θ ^ d θ ^ d s ^ \partial_s\mathcal{R}_{D,2}[\psi]=-\mathcal{W}[\psi;\epsilon\hat{\kappa}\partial_s\overline{K}_D]-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\epsilon\hat{\kappa}\partial_s\overline{K}_D(\Phi^{-1}\psi)(s-\hat{s},\theta-\hat{\theta})\epsilon\hat{\theta}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s} ∂ s R D , 2 [ ψ ] = − W [ ψ ; ϵ κ ^ ∂ s K D ] − p.v. ∫ − 1/2 1/2 ∫ − π π ϵ κ ^ ∂ s K D ( Φ − 1 ψ ) ( s − s ^ , θ − θ ^ ) ϵ θ ^ d θ ^ d s ^
− [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ . -[\partial_s,\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_D\psi\,\epsilon^2\hat{\kappa}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s}. − [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ .
We may use that (256), (258), and (259) imply that
∣ W [ ψ ; ϵ κ ^ ∂ s K D ] ∣ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ , \left|\mathcal{W}[\psi;\epsilon\hat{\kappa}\partial_sK_D]\right|_{C^0,\gamma}\leq c(\kappa_{*,\gamma},c_\Gamma)\epsilon^{1-\gamma^+}\|\psi\|_{C^0,\gamma}, ∣ W [ ψ ; ϵ κ ^ ∂ s K D ] ∣ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ 1 − γ + ∥ ψ ∥ C 0 , γ ,
∣ W [ ψ ; ϵ κ ^ ( a ) ∂ s K D ( a ) − ϵ κ ^ ( b ) ∂ s K D ( b ) ] ∣ C 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ ψ ∥ C 0 , γ . \left|\mathcal{W}\left[\psi;\epsilon\hat{\kappa}^{(a)}\partial_sK_D^{(a)}-\epsilon\hat{\kappa}^{(b)}\partial_sK_D^{(b)}\right]\right|_{C^0,\gamma}\leq c(\kappa_{*,\gamma}^{(a)},\kappa_{*,\gamma}^{(b)},c_\Gamma)\epsilon^{1-\gamma^+}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\gamma}}\|\psi\|_{C^0,\gamma}. W [ ψ ; ϵ κ ^ ( a ) ∂ s K D ( a ) − ϵ κ ^ ( b ) ∂ s K D ( b ) ] C 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ 1 − γ + X ( a ) − X ( b ) C 2 , γ ∥ ψ ∥ C 0 , γ .
Furthermore, using (78), (260), and Lemma 4.2, the third term involving the commutator [ ∂ s , Φ − 1 ] [\partial_s,\Phi^{-1}] [ ∂ s , Φ − 1 ] satisfies
∣ [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ ∣ C 0 , γ ≤ c ( κ ∗ , γ ) ∣ ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ ∣ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ 1 − γ ∥ ψ ∥ L ∞ , \begin{aligned}
\left|[\partial_s,\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_D\psi\,\epsilon^2\hat{\kappa}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s}\right|_{C^0,\gamma}
\leq c(\kappa_{*,\gamma})\left|\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_D\psi\,\epsilon^2\hat{\kappa}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s}\right|_{C^0,\gamma} \\
&\leq c(\kappa_{*,\gamma},c_\Gamma)\epsilon^{1-\gamma}\|\psi\|_{L^\infty},
\end{aligned} [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ C 0 , γ ≤ c ( κ ∗ , γ ) ∫ − 1/2 1/2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ 1 − γ ∥ ψ ∥ L ∞ ,
∣ [ ∂ s , ( Φ ( a ) ) − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ( a ) ψ ϵ 2 κ ^ ( a ) d θ ^ d s ^ − [ ∂ s , ( Φ ( b ) ) − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ( b ) ψ ϵ 2 κ ^ ( b ) d θ ^ d s ^ ∣ C 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ ψ ∥ L ∞ . \begin{aligned}
\left|[\partial_s,(\Phi^{(a)})^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_D^{(a)}\psi\,\epsilon^2\hat{\kappa}^{(a)}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s}
-[\partial_s,(\Phi^{(b)})^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_D^{(b)}\psi\,\epsilon^2\hat{\kappa}^{(b)}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s}\right|_{C^0,\gamma} \\
&\leq c(\kappa_{*,\gamma}^{(a)},\kappa_{*,\gamma}^{(b)},c_\Gamma)\epsilon^{1-\gamma}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\gamma}}\|\psi\|_{L^\infty}.
\end{aligned} [ ∂ s , ( Φ ( a ) ) − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ( a ) ψ ϵ 2 κ ^ ( a ) d θ ^ d s ^ − [ ∂ s , ( Φ ( b ) ) − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ( b ) ψ ϵ 2 κ ^ ( b ) d θ ^ d s ^ C 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ 1 − γ X ( a ) − X ( b ) C 2 , γ ∥ ψ ∥ L ∞ .
It thus remains to bound the middle term of the above expression for ∂ s R D , 2 [ ψ ] \partial_s\mathcal{R}_{D,2}[\psi] ∂ s R D , 2 [ ψ ] . Using the form (196) of R ‾ \overline{R} R , we may write
e z ⋅ ( ϵ κ ^ ∂ s K ‾ D ( Φ − 1 ψ ) ) = 3 ϵ κ ^ 4 π ( 10 ϵ sin 2 ( θ ^ 2 ) s ^ 2 ( e ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 7 − 4 ϵ sin 2 ( θ ^ 2 ) ( e ^ t ⋅ ψ + ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 5 ) \mathbf{e}_z\cdot\left(\epsilon\hat{\kappa}\partial_s\overline{K}_D(\Phi^{-1}\psi)\right)
=\frac{3\epsilon\hat{\kappa}}{4\pi}\left(\frac{10\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)\hat{s}^2(\hat{\mathbf{e}}_t\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\mathbf{e}_\theta\cdot\psi)}{|\overline{R}|^7}
-\frac{4\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)(\hat{\mathbf{e}}_t\cdot\psi+\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\mathbf{e}_\theta\cdot\psi)}{|\overline{R}|^5}\right) e z ⋅ ( ϵ κ ^ ∂ s K D ( Φ − 1 ψ ) ) = 4 π 3 ϵ κ ^ ∣ R ∣ 7 10 ϵ sin 2 ( 2 θ ^ ) s ^ 2 ( e ^ t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) − ∣ R ∣ 5 4 ϵ sin 2 ( 2 θ ^ ) ( e ^ t ⋅ ψ + ϵ sin ( 2 θ ^ ) e θ ⋅ ψ )
e ‾ r ⋅ ( ϵ κ ^ ∂ s K ‾ D ( Φ − 1 ψ ) ) = 3 ϵ κ ^ 4 π ( 20 ϵ 2 sin 4 ( θ ^ 2 ) s ^ ( e ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 7 − 4 ϵ 2 sin 4 ( θ ^ 2 ) ( e t ⋅ ψ ) ∣ R ‾ ∣ 5 ) \overline{\mathbf{e}}_r\cdot\left(\epsilon\hat{\kappa}\partial_s\overline{K}_D(\Phi^{-1}\psi)\right)
=\frac{3\epsilon\hat{\kappa}}{4\pi}\left(\frac{20\epsilon^2\sin^4\left(\frac{\hat{\theta}}{2}\right)\hat{s}(\hat{\mathbf{e}}_t\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\mathbf{e}_\theta\cdot\psi)}{|\overline{R}|^7}
-\frac{4\epsilon^2\sin^4\left(\frac{\hat{\theta}}{2}\right)(\mathbf{e}_t\cdot\psi)}{|\overline{R}|^5}\right) e r ⋅ ( ϵ κ ^ ∂ s K D ( Φ − 1 ψ ) ) = 4 π 3 ϵ κ ^ ∣ R ∣ 7 20 ϵ 2 sin 4 ( 2 θ ^ ) s ^ ( e ^ t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) − ∣ R ∣ 5 4 ϵ 2 sin 4 ( 2 θ ^ ) ( e t ⋅ ψ )
e ‾ θ ⋅ ( ϵ κ ^ ∂ s K ‾ D ( Φ − 1 ψ ) ) = 3 ϵ κ ^ 4 π ( 10 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) s ^ ( e ^ t ⋅ ψ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ ψ ) ∣ R ‾ ∣ 7 − 2 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( e t ⋅ ψ ) ∣ R ‾ ∣ 5 ) . \overline{\mathbf{e}}_\theta\cdot\left(\epsilon\hat{\kappa}\partial_s\overline{K}_D(\Phi^{-1}\psi)\right)
=\frac{3\epsilon\hat{\kappa}}{4\pi}\left(\frac{10\epsilon^2\sin^2\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta})\hat{s}(\hat{\mathbf{e}}_t\cdot\psi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\mathbf{e}_\theta\cdot\psi)}{|\overline{R}|^7}
-\frac{2\epsilon^2\sin^2\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta})(\mathbf{e}_t\cdot\psi)}{|\overline{R}|^5}\right). e θ ⋅ ( ϵ κ ^ ∂ s K D ( Φ − 1 ψ ) ) = 4 π 3 ϵ κ ^ ∣ R ∣ 7 10 ϵ 2 sin 2 ( 2 θ ^ ) sin ( θ ^ ) s ^ ( e ^ t ⋅ ψ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ ψ ) − ∣ R ∣ 5 2 ϵ 2 sin 2 ( 2 θ ^ ) sin ( θ ^ ) ( e t ⋅ ψ ) .
By Lemma 4.4, we thus have
∣ ∂ s R D , 2 [ ψ ] ∣ C 0 , γ ≤ ∣ W [ ψ ; ϵ κ ^ ∂ s K D ] ∣ C 0 , γ + ∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ϵ κ ^ ∂ s K ‾ D ( Φ − 1 ψ ) ϵ θ ^ d θ ^ d s ^ ∣ C 0 , γ + ∣ [ ∂ s , Φ − 1 ] ∫ − 1 / 2 1 / 2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ ∣ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ , ( 263 ) \begin{aligned}
\left|\partial_s\mathcal{R}_{D,2}[\psi]\right|_{C^0,\gamma}\leq\left|\mathcal{W}[\psi;\epsilon\hat{\kappa}\partial_sK_D]\right|_{C^0,\gamma}
+\left|\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\epsilon\hat{\kappa}\partial_s\overline{K}_D(\Phi^{-1}\psi)\epsilon\hat{\theta}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s}\right|_{C^0,\gamma} \\
&+\left|[\partial_s,\Phi^{-1}]\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_D\psi\,\epsilon^2\hat{\kappa}\,\mathrm{d}\hat{\theta}\mathrm{d}\hat{s}\right|_{C^0,\gamma}
\leq c(\kappa_{*,\gamma},c_\Gamma)\epsilon^{-\gamma^+}\|\psi\|_{C^0,\gamma},
\qquad(263)
\end{aligned} ∣ ∂ s R D , 2 [ ψ ] ∣ C 0 , γ ≤ ∣ W [ ψ ; ϵ κ ^ ∂ s K D ] ∣ C 0 , γ + p.v. ∫ − 1/2 1/2 ∫ − π π ϵ κ ^ ∂ s K D ( Φ − 1 ψ ) ϵ θ ^ d θ ^ d s ^ C 0 , γ + [ ∂ s , Φ − 1 ] ∫ − 1/2 1/2 ∫ − π π K D ψ ϵ 2 κ ^ d θ ^ d s ^ C 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − γ + ∥ ψ ∥ C 0 , γ , ( 263 )
∣ ∂ s R D , 2 ( a ) [ ψ ] − ∂ s R D , 2 ( b ) [ ψ ] ∣ C 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ ψ ∥ C 0 , γ . \left|\partial_s\mathcal{R}_{D,2}^{(a)}[\psi]-\partial_s\mathcal{R}_{D,2}^{(b)}[\psi]\right|_{C^0,\gamma}
\leq c(\kappa_{*,\gamma}^{(a)},\kappa_{*,\gamma}^{(b)},c_\Gamma)\epsilon^{-\gamma^+}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\gamma}}\|\psi\|_{C^0,\gamma}. ∂ s R D , 2 ( a ) [ ψ ] − ∂ s R D , 2 ( b ) [ ψ ] C 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − γ + X ( a ) − X ( b ) C 2 , γ ∥ ψ ∥ C 0 , γ .
Combining the estimates (244), (245), (246), (247), (248), (254), (255), (262), and (263), we obtain Lemma 1.11. □
Single layer applied to constant-in-arclength
This section is devoted to the proof of Lemma 1.12 regarding the angle-averaged single layer operator applied to functions that are constant in s s s . We restate the lemma here for convenience.
Lemma 4.7 (Single layer applied to constant-in-s s s ). Let 0 < α < β < 1 0 < \alpha< \beta< 1 0 < α < β < 1 and consider a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) \mathbf{X}(s) \in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Given h ˉ ( θ ) ∈ C 0 , α ( 2 π T ) \bar{\mathbf{h}}(\theta) \in C^{0,\alpha}(2\pi\mathbb{T}) h ˉ ( θ ) ∈ C 0 , α ( 2 π T ) of the form h ˉ = h 1 ( θ ) e z + h 2 ( θ ) e x + h 3 ( θ ) e y \bar{\mathbf{h}} = h_1(\theta)\mathbf{e}_z + h_2(\theta)\mathbf{e}_x + h_3(\theta)\mathbf{e}_y h ˉ = h 1 ( θ ) e z + h 2 ( θ ) e x + h 3 ( θ ) e y , and recalling the map Φ \Phi Φ (27), we may decompose
S [ Φ h ˉ ( θ ) ] = H ϵ [ Φ h ˉ ( θ ) ] + H + [ Φ h ˉ ( θ ) ] \mathcal{S}[\Phi\bar{\mathbf{h}}(\theta)] = \mathcal{H}_\epsilon[\Phi\bar{\mathbf{h}}(\theta)] + \mathcal{H}_+[\Phi\bar{\mathbf{h}}(\theta)] S [ Φ h ˉ ( θ )] = H ϵ [ Φ h ˉ ( θ )] + H + [ Φ h ˉ ( θ )] where
∥ H ϵ [ Φ h ˉ ] ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ˉ ∥ L ∞ , ∥ H ϵ [ Φ h ˉ ] ∥ C s 1 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ˉ ∥ C 0 , α , \left\|\mathcal{H}_\epsilon[\Phi\bar{\mathbf{h}}]\right\|_{C^{0,\alpha}} \leq c(\kappa_{*,\alpha},c_\Gamma)\epsilon^{2-\alpha}\|\bar{\mathbf{h}}\|_{L^\infty}, \qquad\left\|\mathcal{H}_\epsilon[\Phi\bar{\mathbf{h}}]\right\|_{C_s^{1,\alpha}} \leq c(\kappa_{*,\alpha^+},c_\Gamma)\epsilon^{1-\alpha^+}\|\bar{\mathbf{h}}\|_{C^{0,\alpha}}, H ϵ [ Φ h ˉ ] C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ˉ ∥ L ∞ , H ϵ [ Φ h ˉ ] C s 1 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ˉ ∥ C 0 , α , ∥ H + [ Φ h ˉ ] ∥ C 0 , γ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ h ˉ ∥ L ∞ , ∥ H + [ Φ h ˉ ] ∥ C s 1 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ˉ ∥ C 0 , α . (265) \left\|\mathcal{H}_+[\Phi\bar{\mathbf{h}}]\right\|_{C^{0,\gamma}} \leq c(\kappa_{*,\alpha},c_\Gamma)\epsilon^{1-\alpha}\|\bar{\mathbf{h}}\|_{L^\infty}, \qquad\left\|\mathcal{H}_+[\Phi\bar{\mathbf{h}}]\right\|_{C_s^{1,\beta}} \leq c(\kappa_{*,\beta},c_\Gamma)\epsilon^{1-2\beta}\|\bar{\mathbf{h}}\|_{C^{0,\alpha}}. \tag*{(265)} H + [ Φ h ˉ ] C 0 , γ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ h ˉ ∥ L ∞ , H + [ Φ h ˉ ] C s 1 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ˉ ∥ C 0 , α . (265) for any α + ∈ ( α , β ] \alpha^+ \in(\alpha,\beta] α + ∈ ( α , β ] . In addition, given two nearby curves X ( a ) , X ( b ) \mathbf{X}^{(a)},\mathbf{X}^{(b)} X ( a ) , X ( b ) satisfying Lemma 1.1, the corresponding differences H ϵ ( a ) − H ϵ ( b ) \mathcal{H}_\epsilon^{(a)}-\mathcal{H}_\epsilon^{(b)} H ϵ ( a ) − H ϵ ( b ) and H + ( a ) − H + ( b ) \mathcal{H}_+^{(a)}-\mathcal{H}_+^{(b)} H + ( a ) − H + ( b ) satisfy
∥ H ϵ ( a ) [ Φ ( a ) h ˉ ] − H ϵ ( b ) [ Φ ( b ) h ˉ ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ˉ ∥ L ∞ , \left\|\mathcal{H}_\epsilon^{(a)}[\Phi^{(a)}\bar{\mathbf{h}}]-\mathcal{H}_\epsilon^{(b)}[\Phi^{(b)}\bar{\mathbf{h}}]\right\|_{C^{0,\alpha}} \leq c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_\Gamma)\epsilon^{2-\alpha}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha}}\|\bar{\mathbf{h}}\|_{L^\infty}, H ϵ ( a ) [ Φ ( a ) h ˉ ] − H ϵ ( b ) [ Φ ( b ) h ˉ ] C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 2 − α X ( a ) − X ( b ) C 2 , α ∥ h ˉ ∥ L ∞ , ∥ H ϵ ( a ) [ Φ ( a ) h ˉ ] − H ϵ ( b ) [ Φ ( b ) h ˉ ] ∥ C s 1 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ h ˉ ∥ C 0 , α , \left\|\mathcal{H}_\epsilon^{(a)}[\Phi^{(a)}\bar{\mathbf{h}}]-\mathcal{H}_\epsilon^{(b)}[\Phi^{(b)}\bar{\mathbf{h}}]\right\|_{C_s^{1,\alpha}} \leq c(\kappa_{*,\alpha^+}^{(a)},\kappa_{*,\alpha^+}^{(b)},c_\Gamma)\epsilon^{1-\alpha^+}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha^+}}\|\bar{\mathbf{h}}\|_{C^{0,\alpha}}, H ϵ ( a ) [ Φ ( a ) h ˉ ] − H ϵ ( b ) [ Φ ( b ) h ˉ ] C s 1 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + X ( a ) − X ( b ) C 2 , α + ∥ h ˉ ∥ C 0 , α , ∥ H + ( a ) [ Φ ( a ) h ˉ ] − H + ( b ) [ Φ ( b ) h ˉ ] ∥ C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ˉ ∥ L ∞ , \left\|\mathcal{H}_+^{(a)}[\Phi^{(a)}\bar{\mathbf{h}}]-\mathcal{H}_+^{(b)}[\Phi^{(b)}\bar{\mathbf{h}}]\right\|_{C^{0,\alpha}} \leq c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_\Gamma)\epsilon^{1-\alpha}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\alpha}}\|\bar{\mathbf{h}}\|_{L^\infty}, H + ( a ) [ Φ ( a ) h ˉ ] − H + ( b ) [ Φ ( b ) h ˉ ] C 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ 1 − α X ( a ) − X ( b ) C 2 , α ∥ h ˉ ∥ L ∞ , ∥ H + ( a ) [ Φ ( a ) h ˉ ] − H + ( b ) [ Φ ( b ) h ˉ ] ∥ C s 1 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ h ˉ ∥ C 0 , α . (266) \left\|\mathcal{H}_+^{(a)}[\Phi^{(a)}\bar{\mathbf{h}}]-\mathcal{H}_+^{(b)}[\Phi^{(b)}\bar{\mathbf{h}}]\right\|_{C_s^{1,\beta}} \leq c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)},c_\Gamma)\epsilon^{1-2\beta}\left\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\|_{C^{2,\beta}}\|\bar{\mathbf{h}}\|_{C^{0,\alpha}}. \tag*{(266)} H + ( a ) [ Φ ( a ) h ˉ ] − H + ( b ) [ Φ ( b ) h ˉ ] C s 1 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β X ( a ) − X ( b ) C 2 , β ∥ h ˉ ∥ C 0 , α . (266)
Proof of Lemma 1.12. For the proof of this lemma, it will be convenient to keep our integrands in terms of s ′ s' s ′ and θ ′ \theta' θ ′ instead of s ^ = s − s ′ \hat{s}=s-s' s ^ = s − s ′ and θ ^ = θ − θ ′ \hat{\theta}=\theta-\theta' θ ^ = θ − θ ′ . The expansions and bounds of section 4.1 will thus be applied using s − s ′ s-s' s − s ′ and θ − θ ′ \theta-\theta' θ − θ ′ in place of s ^ , θ ^ \hat{s},\hat{\theta} s ^ , θ ^ .
We will first define an auxiliary function akin to the function R t \mathbf{R}_t R t appearing in the proof of Lemma 1.10. Let
R r = X ( s ) − X ( s ′ ) + ϵ ( e r ( s , θ ) − e r ( s , θ ′ ) ) = X ( s ) − X ( s ′ ) + 2 ϵ sin ( θ − θ ′ 2 ) e θ ( s , θ + θ ′ 2 ) , (267) \begin{aligned}
\mathbf{R}_r=\mathbf{X}(s)-\mathbf{X}(s')+\epsilon\left(\mathbf{e}_r(s,\theta)-\mathbf{e}_r(s,\theta')\right) \\
&=\mathbf{X}(s)-\mathbf{X}(s')+2\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)\mathbf{e}_\theta\left(s,\frac{\theta+\theta'}{2}\right), \tag*{(267)}
\end{aligned} R r = X ( s ) − X ( s ′ ) + ϵ ( e r ( s , θ ) − e r ( s , θ ′ ) ) = X ( s ) − X ( s ′ ) + 2 ϵ sin ( 2 θ − θ ′ ) e θ ( s , 2 θ + θ ′ ) , (267)
and note that we have
R − R r = ϵ ( e r ( s , θ ′ ) − e r ( s ′ , θ ′ ) ) = ϵ ( s − s ′ ) Q r \mathbf{R}-\mathbf{R}_r=\epsilon\left(\mathbf{e}_r(s,\theta')-\mathbf{e}_r(s',\theta')\right)=\epsilon(s-s')\mathbf{Q}_r R − R r = ϵ ( e r ( s , θ ′ ) − e r ( s ′ , θ ′ ) ) = ϵ ( s − s ′ ) Q r
∣ R ∣ 2 − ∣ R r ∣ 2 = 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2 |\mathbf{R}|^2-|\mathbf{R}_r|^2=2\epsilon(s-s')\mathbf{Q}_r\cdot\mathbf{R}_r+\epsilon^2(s-s')^2|\mathbf{Q}_r|^2 ∣ R ∣ 2 − ∣ R r ∣ 2 = 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2
1 ∣ R ∣ k − 1 ∣ R r ∣ k = − 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2 ∣ R ∣ ∣ R r ∣ ( ∣ R ∣ + ∣ R r ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R ∣ ℓ ∣ R r ∣ k − 1 − ℓ \frac{1}{|\mathbf{R}|^k}-\frac{1}{|\mathbf{R}_r|^k}
=-\frac{2\epsilon(s-s')\mathbf{Q}_r\cdot\mathbf{R}_r+\epsilon^2(s-s')^2|\mathbf{Q}_r|^2}{|\mathbf{R}||\mathbf{R}_r|(|\mathbf{R}|+|\mathbf{R}_r|)}
\sum_{\ell=0}^{k-1}\frac{1}{|\mathbf{R}|^\ell|\mathbf{R}_r|^{k-1-\ell}} ∣ R ∣ k 1 − ∣ R r ∣ k 1 = − ∣ R ∣∣ R r ∣ ( ∣ R ∣ + ∣ R r ∣ ) 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2 ℓ = 0 ∑ k − 1 ∣ R ∣ ℓ ∣ R r ∣ k − 1 − ℓ 1
1 ∣ R r ( a ) ∣ k − 1 ∣ R r ( b ) ∣ k = 2 s ^ 2 ϵ sin ( θ ^ 2 ) A a b + s ^ 4 B a b ∣ R r ( a ) ∣ ∣ R r ( b ) ∣ ( ∣ R r ( a ) ∣ + ∣ R r ( b ) ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R r ( a ) ∣ ℓ ∣ R r ( b ) ∣ k − 1 − ℓ . (268) \frac{1}{|\mathbf{R}_r^{(a)}|^k}-\frac{1}{|\mathbf{R}_r^{(b)}|^k}
=\frac{2\hat{s}^2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)A_{ab}+\hat{s}^4B_{ab}}{|\mathbf{R}_r^{(a)}||\mathbf{R}_r^{(b)}|(|\mathbf{R}_r^{(a)}|+|\mathbf{R}_r^{(b)}|)}
\sum_{\ell=0}^{k-1}\frac{1}{|\mathbf{R}_r^{(a)}|^\ell|\mathbf{R}_r^{(b)}|^{k-1-\ell}}. \tag*{(268)} ∣ R r ( a ) ∣ k 1 − ∣ R r ( b ) ∣ k 1 = ∣ R r ( a ) ∣∣ R r ( b ) ∣ ( ∣ R r ( a ) ∣ + ∣ R r ( b ) ∣ ) 2 s ^ 2 ϵ sin ( 2 θ ^ ) A ab + s ^ 4 B ab ℓ = 0 ∑ k − 1 ∣ R r ( a ) ∣ ℓ ∣ R r ( b ) ∣ k − 1 − ℓ 1 . (268)
where A a b = e θ ( a ) ⋅ Q t ( a ) − e θ ( b ) ⋅ Q t ( b ) A_{ab}=\mathbf{e}_\theta^{(a)}\cdot\mathbf{Q}_t^{(a)}-\mathbf{e}_\theta^{(b)}\cdot\mathbf{Q}_t^{(b)} A ab = e θ ( a ) ⋅ Q t ( a ) − e θ ( b ) ⋅ Q t ( b ) and B a b = 2 Q t ( a ) − 2 Q t ( b ) + ∣ Q t ( b ) ∣ 2 − ∣ Q t ( a ) ∣ 2 B_{ab}=2\mathbf{Q}_t^{(a)}-2\mathbf{Q}_t^{(b)}+|\mathbf{Q}_t^{(b)}|^2-|\mathbf{Q}_t^{(a)}|^2 B ab = 2 Q t ( a ) − 2 Q t ( b ) + ∣ Q t ( b ) ∣ 2 − ∣ Q t ( a ) ∣ 2 . In addition, R r \mathbf{R}_r R r may be shown to satisfy each of Lemmas 4.1, 4.2, 4.3, and 4.4.
Using R r \mathbf{R}_r R r , we will then write S [ Φ h ˉ ( θ ) ] \mathcal{S}[\Phi\bar{\mathbf{h}}(\theta)] S [ Φ h ˉ ( θ )] as
S [ Φ h ˉ ( θ ) ] = 1 8 π ∫ T ∫ 0 2 π ( H 1 + H 2 + H 3 ) ( Φ h ˉ ) ( s ′ , θ ′ ) ϵ d θ ′ d s ′ , \mathcal{S}[\Phi\bar{\mathbf{h}}(\theta)]=\frac{1}{8\pi}\int_{\mathbb{T}}\int_0^{2\pi}(H_1+H_2+H_3)(\Phi\bar{\mathbf{h}})(s',\theta')\,\epsilon\,d\theta'\,ds', S [ Φ h ˉ ( θ )] = 8 π 1 ∫ T ∫ 0 2 π ( H 1 + H 2 + H 3 ) ( Φ h ˉ ) ( s ′ , θ ′ ) ϵ d θ ′ d s ′ ,
H 1 = I ∣ R r ∣ + R r ⊗ R r ∣ R r ∣ 3 H_1=\frac{I}{\lvert R_r\rvert}+\frac{R_r\otimes R_r}{\lvert R_r\rvert^3} H 1 = ∣ R r ∣ I + ∣ R r ∣ 3 R r ⊗ R r
H 2 = R ⊗ R ∣ R ∣ 3 − R r ⊗ R r ∣ R r ∣ 3 = ( R − R r ) ⊗ R + R r ⊗ ( R − R r ) ∣ R ∣ 3 + ( 1 ∣ R ∣ 3 − 1 ∣ R r ∣ 3 ) R r ⊗ R r H_2=\frac{R\otimes R}{\lvert R\rvert^3}-\frac{R_r\otimes R_r}{\lvert R_r\rvert^3}
=\frac{(R-R_r)\otimes R+R_r\otimes(R-R_r)}{\lvert R\rvert^3}
+\left(\frac{1}{\lvert R\rvert^3}-\frac{1}{\lvert R_r\rvert^3}\right)R_r\otimes R_r H 2 = ∣ R ∣ 3 R ⊗ R − ∣ R r ∣ 3 R r ⊗ R r = ∣ R ∣ 3 ( R − R r ) ⊗ R + R r ⊗ ( R − R r ) + ( ∣ R ∣ 3 1 − ∣ R r ∣ 3 1 ) R r ⊗ R r
H 3 = − ( I ∣ R ∣ + R ⊗ R ∣ R ∣ 3 ) ϵ κ ^ ( s ′ , θ ′ ) H_3=-\left(\frac{I}{\lvert R\rvert}+\frac{R\otimes R}{\lvert R\rvert^3}\right)\widehat{\epsilon\kappa}(s',\theta') H 3 = − ( ∣ R ∣ I + ∣ R ∣ 3 R ⊗ R ) ϵ κ ( s ′ , θ ′ )
Using (268) along with Lemmas 4.2 and 4.4 , we have
∥ ∫ T ∫ 0 2 π H 1 ( Φ ‾ h ‾ ) ϵ d θ ′ d s ′ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ Φ ‾ h ‾ ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ h ‾ ∥ L ∞ \left\lVert\int_{\mathbb{T}}\int_{0}^{2\pi}H_1(\overline{\Phi}\overline{h})\,\epsilon\,d\theta'\,ds'\right\rVert_{C^{0,\alpha}}
\le c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{1-\alpha}\lVert\overline{\Phi}\overline{h}\rVert_{L^\infty}
\le c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{1-\alpha}\lVert\overline{h}\rVert_{L^\infty} ∫ T ∫ 0 2 π H 1 ( Φ h ) ϵ d θ ′ d s ′ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ Φ h ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 1 − α ∥ h ∥ L ∞
∥ ∫ T ∫ 0 2 π H 2 ( Φ ‾ h ‾ ) ϵ d θ ′ d s ′ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ Φ ‾ h ‾ ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ‾ ∥ L ∞ \left\lVert\int_{\mathbb{T}}\int_{0}^{2\pi}H_2(\overline{\Phi}\overline{h})\,\epsilon\,d\theta'\,ds'\right\rVert_{C^{0,\alpha}}
\le c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{2-\alpha}\lVert\overline{\Phi}\overline{h}\rVert_{L^\infty}
\le c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{2-\alpha}\lVert\overline{h}\rVert_{L^\infty} ∫ T ∫ 0 2 π H 2 ( Φ h ) ϵ d θ ′ d s ′ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ Φ h ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ∥ L ∞
∥ ∫ T ∫ 0 2 π H 3 ( Φ ‾ h ‾ ) ϵ d θ ′ d s ′ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ Φ ‾ h ‾ ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ‾ ∥ L ∞ . \left\lVert\int_{\mathbb{T}}\int_{0}^{2\pi}H_3(\overline{\Phi}\overline{h})\,\epsilon\,d\theta'\,ds'\right\rVert_{C^{0,\alpha}}
\le c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{2-\alpha}\lVert\overline{\Phi}\overline{h}\rVert_{L^\infty}
\le c(\kappa_{*},\alpha,c_{\Gamma})\epsilon^{2-\alpha}\lVert\overline{h}\rVert_{L^\infty}. ∫ T ∫ 0 2 π H 3 ( Φ h ) ϵ d θ ′ d s ′ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ Φ h ∥ L ∞ ≤ c ( κ ∗ , α , c Γ ) ϵ 2 − α ∥ h ∥ L ∞ .
In addition, using (268) along with the fact that each remainder Q Q Q satisfies (214) , we may show that for two curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1 , we have
∥ ∫ T ∫ 0 2 π ( H 1 ( a ) ( Φ ( a ) h ‾ ) − H 1 ( b ) ( Φ ( b ) h ‾ ) ) ϵ d θ ′ d s ′ ∥ C 0 , α ≤ c ( κ ∗ ( a ) , α , κ ∗ ( b ) , α , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ‾ ∥ L ∞ \left\lVert\int_{\mathbb{T}}\int_{0}^{2\pi}\left(H_1^{(a)}(\Phi^{(a)}\overline{h})-H_1^{(b)}(\Phi^{(b)}\overline{h})\right)\epsilon\,d\theta'\,ds'\right\rVert_{C^{0,\alpha}}
\le c(\kappa_{*}^{(a)},\alpha,\kappa_{*}^{(b)},\alpha,c_{\Gamma})\epsilon^{1-\alpha}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\alpha}}\lVert\overline{h}\rVert_{L^\infty} ∫ T ∫ 0 2 π ( H 1 ( a ) ( Φ ( a ) h ) − H 1 ( b ) ( Φ ( b ) h ) ) ϵ d θ ′ d s ′ C 0 , α ≤ c ( κ ∗ ( a ) , α , κ ∗ ( b ) , α , c Γ ) ϵ 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ∥ L ∞
∥ ∫ T ∫ 0 2 π ( H 2 ( a ) ( Φ ( a ) h ‾ ) − H 2 ( b ) ( Φ ( b ) h ‾ ) ) ϵ d θ ′ d s ′ ∥ C 0 , α ≤ c ( κ ∗ ( a ) , α , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ‾ ∥ L ∞ \left\lVert\int_{\mathbb{T}}\int_{0}^{2\pi}\left(H_2^{(a)}(\Phi^{(a)}\overline{h})-H_2^{(b)}(\Phi^{(b)}\overline{h})\right)\epsilon\,d\theta'\,ds'\right\rVert_{C^{0,\alpha}}
\le c(\kappa_{*}^{(a)},\alpha,\kappa_{*}^{(b)},\alpha,c_{\Gamma})\epsilon^{2-\alpha}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\alpha}}\lVert\overline{h}\rVert_{L^\infty} ∫ T ∫ 0 2 π ( H 2 ( a ) ( Φ ( a ) h ) − H 2 ( b ) ( Φ ( b ) h ) ) ϵ d θ ′ d s ′ C 0 , α ≤ c ( κ ∗ ( a ) , α , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ∥ L ∞
∥ ∫ T ∫ 0 2 π ( H 3 ( a ) ( Φ ( a ) h ‾ ) − H 3 ( b ) ( Φ ( b ) h ‾ ) ) ϵ d θ ′ d s ′ ∥ C 0 , α ≤ c ( κ ∗ ( a ) , α , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ‾ ∥ L ∞ . \left\lVert\int_{\mathbb{T}}\int_{0}^{2\pi}\left(H_3^{(a)}(\Phi^{(a)}\overline{h})-H_3^{(b)}(\Phi^{(b)}\overline{h})\right)\epsilon\,d\theta'\,ds'\right\rVert_{C^{0,\alpha}}
\le c(\kappa_{*}^{(a)},\alpha,\kappa_{*}^{(b)},\alpha,c_{\Gamma})\epsilon^{2-\alpha}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\alpha}}\lVert\overline{h}\rVert_{L^\infty}. ∫ T ∫ 0 2 π ( H 3 ( a ) ( Φ ( a ) h ) − H 3 ( b ) ( Φ ( b ) h ) ) ϵ d θ ′ d s ′ C 0 , α ≤ c ( κ ∗ ( a ) , α , κ ∗ ( b ) , α , c Γ ) ϵ 2 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ h ∥ L ∞ .
We next turn to estimates for ∂ s S [ Φ ‾ h ‾ ( θ ) ] \partial_sS[\overline{\Phi}\overline{h}(\theta)] ∂ s S [ Φ h ( θ )] . We first note that, since h ‾ ( θ ) \overline{h}(\theta) h ( θ ) is independent of s s s , the only s s s -dependence in ( Φ ‾ h ‾ ) ( s , θ ) (\overline{\Phi}\overline{h})(s,\theta) ( Φ h ) ( s , θ ) is due to the map Φ \Phi Φ , given by (27) . We may calculate
∂ s ( Φ ‾ h ‾ ) ( s , θ ) = ( ∂ s e t ) ( h ‾ ⋅ e z ) + ( ∂ s e n 1 ) ( h ‾ ⋅ e x ) + ( ∂ s e n 2 ) ( h ‾ ⋅ e y ) . \partial_s(\overline{\Phi}\overline{h})(s,\theta)=(\partial_se_t)(\overline{h}\cdot e_z)+(\partial_se_{n_1})(\overline{h}\cdot e_x)+(\partial_se_{n_2})(\overline{h}\cdot e_y). ∂ s ( Φ h ) ( s , θ ) = ( ∂ s e t ) ( h ⋅ e z ) + ( ∂ s e n 1 ) ( h ⋅ e x ) + ( ∂ s e n 2 ) ( h ⋅ e y ) .
In particular, we may bound
∥ ∂ s ( Φ ‾ h ‾ ) ∥ L ∞ ( Γ ϵ ) ≤ c ( κ ∗ ) ∥ h ‾ ∥ L ∞ ( 2 π T ) , ∥ ∂ s ( Φ ‾ h ‾ ) ∥ C 0 , α ( Γ ϵ ) ≤ c ( κ ∗ , α ) ∥ h ‾ ∥ L ∞ ( 2 π T ) . (269) \lVert\partial_s(\overline{\Phi}\overline{h})\rVert_{L^\infty(\Gamma_\epsilon)}
\le c(\kappa_*)\lVert\overline{h}\rVert_{L^\infty(2\pi\mathbb{T})},
\qquad
\lVert\partial_s(\overline{\Phi}\overline{h})\rVert_{C^{0,\alpha}(\Gamma_\epsilon)}
\le c(\kappa_*,\alpha)\lVert\overline{h}\rVert_{L^\infty(2\pi\mathbb{T})}. \tag*{(269)} ∥ ∂ s ( Φ h ) ∥ L ∞ ( Γ ϵ ) ≤ c ( κ ∗ ) ∥ h ∥ L ∞ ( 2 π T ) , ∥ ∂ s ( Φ h ) ∥ C 0 , α ( Γ ϵ ) ≤ c ( κ ∗ , α ) ∥ h ∥ L ∞ ( 2 π T ) . (269)
In addition, we note the following expansion for ∂ s R r \partial_sR_r ∂ s R r :
− ∂ s ′ R r = e t ( s ′ ) -\partial_{s'}R_r=e_t(s') − ∂ s ′ R r = e t ( s ′ )
∂ s R r = e t ( s ) + 2 ϵ sin ( θ − θ ′ 2 ) ( ∂ θ κ ^ ( s , θ ) e t ( s ) − κ 3 e r ( s , θ ) ) \partial_sR_r=e_t(s)+2\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)\left(\partial_\theta\widehat{\kappa}(s,\theta)e_t(s)-\kappa_3e_r(s,\theta)\right) ∂ s R r = e t ( s ) + 2 ϵ sin ( 2 θ − θ ′ ) ( ∂ θ κ ( s , θ ) e t ( s ) − κ 3 e r ( s , θ ) )
= : − ∂ s ′ R r + ( s − s ′ ) Q t ( s , s ′ ) + ϵ sin ( θ − θ ′ 2 ) Q X 1 . (270) \mathrel{=:}-\partial_{s'}R_r+(s-s')Q_t(s,s')+\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)Q_{X_1}. \tag*{(270)} =: − ∂ s ′ R r + ( s − s ′ ) Q t ( s , s ′ ) + ϵ sin ( 2 θ − θ ′ ) Q X 1 . (270)
The closeness of ∂ s R r \partial_sR_r ∂ s R r to − ∂ s ′ R r -\partial_{s'}R_r − ∂ s ′ R r will be of particular importance in estimating the kernel H 1 H_1 H 1 . Finally, we record the following expansions of ∂ s R \partial_sR ∂ s R and ∂ s ( R − R r ) \partial_s(R-R_r) ∂ s ( R − R r ) :
∂ s R = ( 1 − ϵ κ ^ ( s , θ ) ) e t ( s ) + ϵ κ 3 e θ ( s , θ ) = : e t ( s ) + ϵ Q X 2 ( s , θ ) \partial_sR=(1-\epsilon\widehat{\kappa}(s,\theta))e_t(s)+\epsilon\kappa_3e_\theta(s,\theta)
\mathrel{=:}e_t(s)+\epsilon Q_{X_2}(s,\theta) ∂ s R = ( 1 − ϵ κ ( s , θ )) e t ( s ) + ϵ κ 3 e θ ( s , θ ) =: e t ( s ) + ϵ Q X 2 ( s , θ )
∂ s ( R − R r ) = ϵ κ ^ ( s , θ ′ ) e t ( s ) + ϵ κ 3 e θ ( s , θ ′ ) = : ϵ Q X 2 ( s , θ ′ ) . (271) \partial_s(R-R_r)=\epsilon\widehat{\kappa}(s,\theta')e_t(s)+\epsilon\kappa_3e_\theta(s,\theta')
\mathrel{=:}\epsilon Q_{X_2}(s,\theta'). \tag*{(271)} ∂ s ( R − R r ) = ϵ κ ( s , θ ′ ) e t ( s ) + ϵ κ 3 e θ ( s , θ ′ ) =: ϵ Q X 2 ( s , θ ′ ) . (271)
Here, as usual, the remainders Q X 1 Q_{X_1} Q X 1 and Q X 2 Q_{X_2} Q X 2 satisfy (214) .
We may use the expansion (270) of ∂ s R r \partial_sR_r ∂ s R r to write H 1 H_1 H 1 as
∂ s H 1 = − I ( ∂ s R r ⋅ R r ) ∣ R r ∣ 3 + ∂ s R r ⊗ R r + R r ⊗ ∂ s R r ∣ R r ∣ 3 − 3 R r ⊗ R r ( ∂ s R r ⋅ R r ) ∣ R r ∣ 5 = − ∂ s ′ H 1 + ( ( s − s ′ ) Q t + ϵ sin ( θ − θ ′ 2 ) Q X 1 ) ⊗ R r + R r ⊗ ( ( s − s ′ ) Q t + ϵ sin ( θ − θ ′ 2 ) Q X 1 ) ∣ R r ∣ 3 − I ( ( s − s ′ ) Q t + ϵ sin ( θ − θ ′ 2 ) Q X 1 ) ⋅ R r ∣ R r ∣ 3 − 3 R r ⊗ R r ( ( s − s ′ ) Q t + ϵ sin ( θ − θ ′ 2 ) Q X 1 ) ⋅ R r ∣ R r ∣ 5 . \begin{aligned}
\partial_sH_1=-\frac{I(\partial_sR_r\cdot R_r)}{\lvert R_r\rvert^3}
+\frac{\partial_sR_r\otimes R_r+R_r\otimes\partial_sR_r}{\lvert R_r\rvert^3}
-\frac{3R_r\otimes R_r(\partial_sR_r\cdot R_r)}{\lvert R_r\rvert^5} \\
&=-\partial_{s'}H_1+\frac{\left((s-s')Q_t+\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)Q_{X1}\right)\otimes R_r+R_r\otimes\left((s-s')Q_t+\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)Q_{X1}\right)}{|R_r|^3}
-\frac{\mathbb{I}\left((s-s')Q_t+\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)Q_{X1}\right)\cdot R_r}{|R_r|^3}
-\frac{3R_r\otimes R_r\left((s-s')Q_t+\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)Q_{X1}\right)\cdot R_r}{|R_r|^5}.
\end{aligned} ∂ s H 1 = − ∣ R r ∣ 3 I ( ∂ s R r ⋅ R r ) + ∣ R r ∣ 3 ∂ s R r ⊗ R r + R r ⊗ ∂ s R r − ∣ R r ∣ 5 3 R r ⊗ R r ( ∂ s R r ⋅ R r ) = − ∂ s ′ H 1 + ∣ R r ∣ 3 ( ( s − s ′ ) Q t + ϵ sin ( 2 θ − θ ′ ) Q X 1 ) ⊗ R r + R r ⊗ ( ( s − s ′ ) Q t + ϵ sin ( 2 θ − θ ′ ) Q X 1 ) − ∣ R r ∣ 3 I ( ( s − s ′ ) Q t + ϵ sin ( 2 θ − θ ′ ) Q X 1 ) ⋅ R r − ∣ R r ∣ 5 3 R r ⊗ R r ( ( s − s ′ ) Q t + ϵ sin ( 2 θ − θ ′ ) Q X 1 ) ⋅ R r .
For the kernel − ∂ s ′ H 1 -\partial_{s'}H_1 − ∂ s ′ H 1 , we may integrate by parts in s ′ s' s ′ and use Lemma 4.4, case (1), along with (268) to obtain
∣ p.v. ∫ T ∫ 0 2 π ∂ s ′ H 1 ( Φ h ‾ ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , β = ∣ ∫ T ∫ 0 2 π H 1 ∂ s ′ ( Φ h ‾ ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ‾ ∥ L ∞ . \left|\operatorname{p.v.}\int_{\mathbb{T}}\int_0^{2\pi}\partial_{s'}H_1(\overline{\Phi h})\,\epsilon\,d\theta'\,ds'\right|_{\dot{C}^{0,\beta}}
=
\left|\int_{\mathbb{T}}\int_0^{2\pi}H_1\,\partial_{s'}(\overline{\Phi h})\,\epsilon\,d\theta'\,ds'\right|_{\dot{C}^{0,\beta}}
\leq c(\kappa_{*,\beta},c_{\Gamma})\epsilon^{1-2\beta}\|\overline{h}\|_{L^\infty}. p.v. ∫ T ∫ 0 2 π ∂ s ′ H 1 ( Φ h ) ϵ d θ ′ d s ′ C ˙ 0 , β = ∫ T ∫ 0 2 π H 1 ∂ s ′ ( Φ h ) ϵ d θ ′ d s ′ C ˙ 0 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ∥ L ∞ .
For the remaining terms in the above expansion of ∂ s H 1 \partial_sH_1 ∂ s H 1 , we may use Lemma 4.4, case (1), to estimate
∣ p.v. ∫ T ∫ 0 2 π ∂ s H 1 ( Φ h ‾ ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ Φ h ‾ ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ‾ ∥ L ∞ . \left|\operatorname{p.v.}\int_{\mathbb{T}}\int_0^{2\pi}\partial_sH_1(\overline{\Phi h})\,\epsilon\,d\theta'\,ds'\right|_{\dot{C}^{0,\beta}}
\leq c(\kappa_{*,\beta},c_{\Gamma})\epsilon^{1-2\beta}\|\overline{\Phi h}\|_{L^\infty}
\leq c(\kappa_{*,\beta},c_{\Gamma})\epsilon^{1-2\beta}\|\overline{h}\|_{L^\infty}. p.v. ∫ T ∫ 0 2 π ∂ s H 1 ( Φ h ) ϵ d θ ′ d s ′ C ˙ 0 , β ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ Φ h ∥ L ∞ ≤ c ( κ ∗ , β , c Γ ) ϵ 1 − 2 β ∥ h ∥ L ∞ .
Using the expansion for ∂ s H 1 \partial_sH_1 ∂ s H 1 and that each remainder Q Q Q satisfies (213), we may use a similar series of arguments to show that, for two curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1, we have
∣ p.v. ∫ T ∫ 0 2 π ( ∂ s H 1 ( a ) ( Φ ( a ) h ‾ ) − ∂ s H 1 ( b ) ( Φ ( b ) h ‾ ) ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ h ‾ ∥ L ∞ . \left|\operatorname{p.v.}\int_{\mathbb{T}}\int_0^{2\pi}\left(\partial_sH_1^{(a)}(\overline{\Phi^{(a)}h})-\partial_sH_1^{(b)}(\overline{\Phi^{(b)}h})\right)\epsilon\,d\theta'\,ds'\right|_{\dot{C}^{0,\beta}}
\leq c(\kappa_{*,\beta}^{(a)},\kappa_{*,\beta}^{(b)},c_{\Gamma})\epsilon^{1-2\beta}\|X^{(a)}-X^{(b)}\|_{C^{2,\beta}}\|\overline{h}\|_{L^\infty}. p.v. ∫ T ∫ 0 2 π ( ∂ s H 1 ( a ) ( Φ ( a ) h ) − ∂ s H 1 ( b ) ( Φ ( b ) h ) ) ϵ d θ ′ d s ′ C ˙ 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) , c Γ ) ϵ 1 − 2 β ∥ X ( a ) − X ( b ) ∥ C 2 , β ∥ h ∥ L ∞ .
We next consider the kernel H 2 H_2 H 2 . Using (267) and (270), we may write
∂ s H 2 = ∂ s R ⊗ ( R − R r ) ∣ R ∣ 3 + R ⊗ ∂ s ( R − R r ) ∣ R ∣ 3 − 3 R ⊗ ( R − R r ) ( ∂ s R ⋅ R ) ∣ R ∣ 5 + ∂ s ( R − R r ) ⊗ R r ∣ R ∣ 3 + ( R − R r ) ⊗ ∂ s R r ∣ R ∣ 3 − 3 ( R − R r ) ⊗ R r ( ∂ s R ⋅ R ) ∣ R ∣ 5 − 3 ∂ s ( R − R r ) ⋅ R + ∂ s R r ⋅ ( R − R r ) ∣ R ∣ 5 R r ⊗ R r + ( 1 ∣ R ∣ 3 − 1 ∣ R r ∣ 3 ) ( ∂ s R r ⊗ R r + R r ⊗ ∂ s R r ) − 3 ( 1 ∣ R ∣ 5 − 1 ∣ R r ∣ 5 ) ( ∂ s R r ⋅ R r ) R r ⊗ R r = ϵ ( s − s ′ ) ∂ s R ⊗ Q r ∣ R ∣ 3 + ϵ R ⊗ Q X 2 ∣ R ∣ 3 − 3 ϵ ( s − s ′ ) R ⊗ Q r ( ∂ s R ⋅ R ) ∣ R ∣ 5 + ϵ Q X 2 ⊗ R r ∣ R ∣ 3 + ϵ ( s − s ′ ) Q r ⊗ ∂ s R r ∣ R ∣ 3 − 3 ϵ ( s − s ′ ) Q r ⊗ R r ( ∂ s R ⋅ R ) ∣ R ∣ 5 − 3 ϵ Q X 2 ⋅ R + ϵ ( s − s ′ ) ∂ s R r ⋅ Q r ∣ R ∣ 5 R r ⊗ R r − ( 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2 ∣ R ∣ ∣ R r ∣ ( ∣ R ∣ + ∣ R r ∣ ) ∑ ℓ = 0 2 1 ∣ R ∣ ℓ ∣ R r ∣ 2 − ℓ ) ( ∂ s R r ⊗ R r + R r ⊗ ∂ s R r ) + 3 ( 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2 ∣ R ∣ ∣ R r ∣ ( ∣ R ∣ + ∣ R r ∣ ) ∑ ℓ = 0 4 1 ∣ R ∣ ℓ ∣ R r ∣ 4 − ℓ ) ( ∂ s R r ⋅ R r ) R r ⊗ R r . \begin{aligned}
\partial_sH_2={}&\frac{\partial_sR\otimes(R-R_r)}{|R|^3}+\frac{R\otimes\partial_s(R-R_r)}{|R|^3}-\frac{3R\otimes(R-R_r)(\partial_sR\cdot R)}{|R|^5}\\
&+\frac{\partial_s(R-R_r)\otimes R_r}{|R|^3}+\frac{(R-R_r)\otimes\partial_sR_r}{|R|^3}-\frac{3(R-R_r)\otimes R_r(\partial_sR\cdot R)}{|R|^5}\\
&-3\frac{\partial_s(R-R_r)\cdot R+\partial_sR_r\cdot(R-R_r)}{|R|^5}R_r\otimes R_r\\
&+\left(\frac{1}{|R|^3}-\frac{1}{|R_r|^3}\right)(\partial_sR_r\otimes R_r+R_r\otimes\partial_sR_r)\\
&-3\left(\frac{1}{|R|^5}-\frac{1}{|R_r|^5}\right)(\partial_sR_r\cdot R_r)R_r\otimes R_r\\
={}&\frac{\epsilon(s-s')\partial_sR\otimes Q_r}{|R|^3}+\frac{\epsilon R\otimes Q_{X2}}{|R|^3}-\frac{3\epsilon(s-s')R\otimes Q_r(\partial_sR\cdot R)}{|R|^5}\\
&+\frac{\epsilon Q_{X2}\otimes R_r}{|R|^3}+\frac{\epsilon(s-s')Q_r\otimes\partial_sR_r}{|R|^3}-\frac{3\epsilon(s-s')Q_r\otimes R_r(\partial_sR\cdot R)}{|R|^5}\\
&-3\frac{\epsilon Q_{X2}\cdot R+\epsilon(s-s')\partial_sR_r\cdot Q_r}{|R|^5}R_r\otimes R_r\\
&-\left(\frac{2\epsilon(s-s')Q_r\cdot R_r+\epsilon^2(s-s')^2|Q_r|^2}{|R||R_r|(|R|+|R_r|)}\sum_{\ell=0}^{2}\frac{1}{|R|^\ell|R_r|^{2-\ell}}\right)(\partial_sR_r\otimes R_r+R_r\otimes\partial_sR_r)\\
&+3\left(\frac{2\epsilon(s-s')Q_r\cdot R_r+\epsilon^2(s-s')^2|Q_r|^2}{|R||R_r|(|R|+|R_r|)}\sum_{\ell=0}^{4}\frac{1}{|R|^\ell|R_r|^{4-\ell}}\right)(\partial_sR_r\cdot R_r)R_r\otimes R_r.
\end{aligned} ∂ s H 2 = = ∣ R ∣ 3 ∂ s R ⊗ ( R − R r ) + ∣ R ∣ 3 R ⊗ ∂ s ( R − R r ) − ∣ R ∣ 5 3 R ⊗ ( R − R r ) ( ∂ s R ⋅ R ) + ∣ R ∣ 3 ∂ s ( R − R r ) ⊗ R r + ∣ R ∣ 3 ( R − R r ) ⊗ ∂ s R r − ∣ R ∣ 5 3 ( R − R r ) ⊗ R r ( ∂ s R ⋅ R ) − 3 ∣ R ∣ 5 ∂ s ( R − R r ) ⋅ R + ∂ s R r ⋅ ( R − R r ) R r ⊗ R r + ( ∣ R ∣ 3 1 − ∣ R r ∣ 3 1 ) ( ∂ s R r ⊗ R r + R r ⊗ ∂ s R r ) − 3 ( ∣ R ∣ 5 1 − ∣ R r ∣ 5 1 ) ( ∂ s R r ⋅ R r ) R r ⊗ R r ∣ R ∣ 3 ϵ ( s − s ′ ) ∂ s R ⊗ Q r + ∣ R ∣ 3 ϵ R ⊗ Q X 2 − ∣ R ∣ 5 3 ϵ ( s − s ′ ) R ⊗ Q r ( ∂ s R ⋅ R ) + ∣ R ∣ 3 ϵ Q X 2 ⊗ R r + ∣ R ∣ 3 ϵ ( s − s ′ ) Q r ⊗ ∂ s R r − ∣ R ∣ 5 3 ϵ ( s − s ′ ) Q r ⊗ R r ( ∂ s R ⋅ R ) − 3 ∣ R ∣ 5 ϵ Q X 2 ⋅ R + ϵ ( s − s ′ ) ∂ s R r ⋅ Q r R r ⊗ R r − ( ∣ R ∣∣ R r ∣ ( ∣ R ∣ + ∣ R r ∣ ) 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2 ℓ = 0 ∑ 2 ∣ R ∣ ℓ ∣ R r ∣ 2 − ℓ 1 ) ( ∂ s R r ⊗ R r + R r ⊗ ∂ s R r ) + 3 ( ∣ R ∣∣ R r ∣ ( ∣ R ∣ + ∣ R r ∣ ) 2 ϵ ( s − s ′ ) Q r ⋅ R r + ϵ 2 ( s − s ′ ) 2 ∣ Q r ∣ 2 ℓ = 0 ∑ 4 ∣ R ∣ ℓ ∣ R r ∣ 4 − ℓ 1 ) ( ∂ s R r ⋅ R r ) R r ⊗ R r .
Note that each term contains an additional factor of ϵ \epsilon ϵ . Using Lemma 4.4, case (2), we thus have
∣ p.v. ∫ T ∫ 0 2 π ∂ s H 2 ( Φ h ‾ ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ Φ h ‾ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ‾ ∥ C 0 , α . \left\lvert\operatorname{p.v.}\int_{\mathbb{T}}\int_{0}^{2\pi}\partial_s H_2(\overline{\Phi h})\,\epsilon\,d\theta'\,ds'\right\rvert_{\dot{C}^{0,\alpha}} \le c(\kappa_{*,\alpha+},c_{\Gamma})\epsilon^{1-\alpha+}\lVert\overline{\Phi h}\rVert_{C^{0,\alpha}} \le c(\kappa_{*,\alpha+},c_{\Gamma})\epsilon^{1-\alpha+}\lVert\overline{h}\rVert_{C^{0,\alpha}}. p.v. ∫ T ∫ 0 2 π ∂ s H 2 ( Φ h ) ϵ d θ ′ d s ′ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ Φ h ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ∥ C 0 , α .
In addition, using that each of the remainder terms Q Q Q satisfy (213), we may use the above expansion of H 2 H_2 H 2 to show that, for two curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1, we have
∣ p.v. ∫ T ∫ 0 2 π ( ∂ s H 2 ( a ) ( Φ ( a ) h ‾ ) − ∂ s H 2 ( b ) ( Φ ( b ) h ‾ ) ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ h ‾ ∥ C 0 , α . \left\lvert\operatorname{p.v.}\int_{\mathbb{T}}\int_{0}^{2\pi}\left(\partial_s H_2^{(a)}(\Phi^{(a)}\overline{h})-\partial_s H_2^{(b)}(\Phi^{(b)}\overline{h})\right)\epsilon\,d\theta'\,ds'\right\rvert_{\dot{C}^{0,\alpha}}
\le c(\kappa_{*,\alpha+}^{(a)},\kappa_{*,\alpha+}^{(b)},c_{\Gamma})\epsilon^{1-\alpha+}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\alpha+}}\lVert\overline{h}\rVert_{C^{0,\alpha}}. p.v. ∫ T ∫ 0 2 π ( ∂ s H 2 ( a ) ( Φ ( a ) h ) − ∂ s H 2 ( b ) ( Φ ( b ) h ) ) ϵ d θ ′ d s ′ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ h ∥ C 0 , α .
Finally, we turn to the kernel H 3 H_3 H 3 . We have
∂ s H 3 = ( I ∂ s R ⋅ R ∣ R ∣ 3 + ∂ s R ⊗ R + R ⊗ ∂ s R ∣ R ∣ 3 + 3 ( R ⊗ R ) ∂ s R ⋅ R ∣ R ∣ 5 ) ϵ κ ^ ( s ′ , θ ′ ) . \partial_s H_3=\left(\frac{\mathbf{I}\,\partial_s\mathbf{R}\cdot\mathbf{R}}{|\mathbf{R}|^3}+\frac{\partial_s\mathbf{R}\otimes\mathbf{R}+\mathbf{R}\otimes\partial_s\mathbf{R}}{|\mathbf{R}|^3}+3\frac{(\mathbf{R}\otimes\mathbf{R})\,\partial_s\mathbf{R}\cdot\mathbf{R}}{|\mathbf{R}|^5}\right)\epsilon\widehat{\kappa}(s',\theta'). ∂ s H 3 = ( ∣ R ∣ 3 I ∂ s R ⋅ R + ∣ R ∣ 3 ∂ s R ⊗ R + R ⊗ ∂ s R + 3 ∣ R ∣ 5 ( R ⊗ R ) ∂ s R ⋅ R ) ϵ κ ( s ′ , θ ′ ) .
Noting the additional factor of ϵ \epsilon ϵ and using the form (270) of ∂ s R \partial_s\mathbf{R} ∂ s R , we may use Lemma 4.4 to obtain
∣ p.v. ∫ T ∫ 0 2 π ∂ s H 3 ( Φ h ‾ ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ Φ h ‾ ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ‾ ∥ C 0 , α . \left\lvert\operatorname{p.v.}\int_{\mathbb{T}}\int_{0}^{2\pi}\partial_s H_3(\overline{\Phi h})\,\epsilon\,d\theta'\,ds'\right\rvert_{\dot{C}^{0,\alpha}} \le c(\kappa_{*,\alpha+},c_{\Gamma})\epsilon^{1-\alpha+}\lVert\overline{\Phi h}\rVert_{C^{0,\alpha}} \le c(\kappa_{*,\alpha+},c_{\Gamma})\epsilon^{1-\alpha+}\lVert\overline{h}\rVert_{C^{0,\alpha}}. p.v. ∫ T ∫ 0 2 π ∂ s H 3 ( Φ h ) ϵ d θ ′ d s ′ C ˙ 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ Φ h ∥ C 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ 1 − α + ∥ h ∥ C 0 , α .
Furthermore, for two nearby curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) , we have
∣ p.v. ∫ T ∫ 0 2 π ( ∂ s H 3 ( a ) ( Φ ( a ) h ‾ ) − ∂ s H 3 ( b ) ( Φ ( b ) h ‾ ) ) ϵ d θ ′ d s ′ ∣ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ h ‾ ∥ C 0 , α . \left\lvert\operatorname{p.v.}\int_{\mathbb{T}}\int_{0}^{2\pi}\left(\partial_s H_3^{(a)}(\Phi^{(a)}\overline{h})-\partial_s H_3^{(b)}(\Phi^{(b)}\overline{h})\right)\epsilon\,d\theta'\,ds'\right\rvert_{\dot{C}^{0,\alpha}}
\le c(\kappa_{*,\alpha+}^{(a)},\kappa_{*,\alpha+}^{(b)},c_{\Gamma})\epsilon^{1-\alpha+}\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\alpha+}}\lVert\overline{h}\rVert_{C^{0,\alpha}}. p.v. ∫ T ∫ 0 2 π ( ∂ s H 3 ( a ) ( Φ ( a ) h ) − ∂ s H 3 ( b ) ( Φ ( b ) h ) ) ϵ d θ ′ d s ′ C ˙ 0 , α ≤ c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ h ∥ C 0 , α .
Defining
H + [ Φ h ‾ ] = 1 8 π ∫ T ∫ 0 2 π H 1 ( Φ h ‾ ) ( s ′ , θ ′ ) ϵ d θ ′ d s ′ , H ϵ [ Φ h ‾ ] = 1 8 π ∫ T ∫ 0 2 π ( H 2 + H 3 ) ( Φ h ‾ ) ( s ′ , θ ′ ) ϵ d θ ′ d s ′ , \mathcal{H}_{+}[\overline{\Phi h}]=\frac{1}{8\pi}\int_{\mathbb{T}}\int_{0}^{2\pi}H_1(\overline{\Phi h})(s',\theta')\,\epsilon\,d\theta'\,ds',\qquad\mathcal{H}_{\epsilon}[\overline{\Phi h}]=\frac{1}{8\pi}\int_{\mathbb{T}}\int_{0}^{2\pi}(H_2+H_3)(\overline{\Phi h})(s',\theta')\,\epsilon\,d\theta'\,ds', H + [ Φ h ] = 8 π 1 ∫ T ∫ 0 2 π H 1 ( Φ h ) ( s ′ , θ ′ ) ϵ d θ ′ d s ′ , H ϵ [ Φ h ] = 8 π 1 ∫ T ∫ 0 2 π ( H 2 + H 3 ) ( Φ h ) ( s ′ , θ ′ ) ϵ d θ ′ d s ′ ,
we obtain Lemma 1.12. □
Bounds for full Neumann data
This section is concerned with the bounds for the full Neumann data w ( s , θ ) = − σ [ u ] n ∣ Γ ϵ \left.w(s,\theta)=-\sigma[u]n\right|_{\Gamma_\epsilon} w ( s , θ ) = − σ [ u ] n ∣ Γ ϵ stated in Lemma 1.13. We restate the lemma below for convenience.
Lemma 5.1 (Bounds for w ( s , θ ) w(s,\theta) w ( s , θ ) , restated). Let 0 < α < γ < β < 1 0<\alpha<\gamma<\beta<1 0 < α < γ < β < 1 and consider a slender body Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) X(s)\in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) . Given θ \theta θ -independent Dirichlet data v ( s ) ∈ C 1 , α ( T ) v(s)\in C^{1,\alpha}(\mathbb{T}) v ( s ) ∈ C 1 , α ( T ) along the filament surface Γ ϵ \Gamma_\epsilon Γ ϵ , let u ( x ) u(x) u ( x ) , x ∈ Ω ϵ x\in\Omega_\epsilon x ∈ Ω ϵ , denote the corresponding solution to the Stokes equations in Ω ϵ \Omega_\epsilon Ω ϵ , and let w ( s , θ ) w(s,\theta) w ( s , θ ) denote the corresponding Neumann boundary value w = − σ [ u ] n ∣ Γ ϵ \left.w=-\sigma[u]n\right|_{\Gamma_\epsilon} w = − σ [ u ] n ∣ Γ ϵ . Then w w w may be decomposed as
w ( s , θ ) = w ϵ ( s , θ ) + w + ( s , θ ) w(s,\theta)=w_\epsilon(s,\theta)+w_+(s,\theta) w ( s , θ ) = w ϵ ( s , θ ) + w + ( s , θ ) where
∥ w ϵ ∥ C 0 , α ≤ c ( κ ∗ , c Γ ) ∥ v ∥ C 1 , α \lVert w_\epsilon\rVert_{C^{0,\alpha}}\le c(\kappa_*,c_{\Gamma})\lVert v\rVert_{C^{1,\alpha}} ∥ w ϵ ∥ C 0 , α ≤ c ( κ ∗ , c Γ ) ∥ v ∥ C 1 , α ∥ w + ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ (272) \lVert w_+\rVert_{C^{0,\gamma}}\le c(\epsilon,\kappa_{*,\gamma+},c_{\Gamma})\lVert v\rVert_{C^{0,\gamma}} \tag*{(272)} ∥ w + ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ (272)
for any γ + ∈ ( γ , β ] \gamma^{+} \in(\gamma,\beta] γ + ∈ ( γ , β ] . In addition, given two nearby filaments Σ ϵ ( a ) , Σ ϵ ( b ) \Sigma_{\epsilon}^{(a)},\Sigma_{\epsilon}^{(b)} Σ ϵ ( a ) , Σ ϵ ( b ) with centerlines X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1, we have
∥ w ϵ ( a ) − w ϵ ( b ) ∥ C 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ∥ v ∥ C 1 , α ∥ w + ( a ) − w + ( b ) ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ , (273) \begin{aligned}
\left\|w_{\epsilon}^{(a)}-w_{\epsilon}^{(b)}\right\|_{C^{0,\alpha}} &\leq c\left(\kappa_{*}^{(a)},\kappa_{*}^{(b)},c_{\Gamma}\right)\left\|X^{(a)}-X^{(b)}\right\|_{C^{2}}\|v\|_{C^{1,\alpha}} \\
\left\|w_{+}^{(a)}-w_{+}^{(b)}\right\|_{C^{0,\gamma}} &\leq c\left(\epsilon,\kappa_{*,\gamma^{+}}^{(a)},\kappa_{*,\gamma^{+}}^{(b)},c_{\Gamma}\right)\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma^{+}}}\|v\|_{C^{0,\gamma}}, \tag*{(273)}
\end{aligned} w ϵ ( a ) − w ϵ ( b ) C 0 , α w + ( a ) − w + ( b ) C 0 , γ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) X ( a ) − X ( b ) C 2 ∥ v ∥ C 1 , α ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) X ( a ) − X ( b ) C 2 , γ + ∥ v ∥ C 0 , γ , (273)
where w ( a ) , w ( b ) w^{(a)},w^{(b)} w ( a ) , w ( b ) denote the Neumann data along filaments Σ ϵ ( a ) \Sigma_{\epsilon}^{(a)} Σ ϵ ( a ) and Σ ϵ ( b ) \Sigma_{\epsilon}^{(b)} Σ ϵ ( b ) , respectively.
In both (73) and (74) , the ϵ \epsilon ϵ -dependence is explicit in the bounds for w ϵ w_{\epsilon} w ϵ but not for w + w_{+} w + .
The proof of Lemma 1.13 will again rely on a boundary integral formulation of the full Stokes Dirichlet-to-Neumann map along the slender body surface Γ ϵ \Gamma_{\epsilon} Γ ϵ . However, it will be more convenient to use a different formulation from the one presented in § 1.2.3 in order to extract more precise information.
Following [62 , 61 ] , given Dirichlet data v ( x ) v(x) v ( x ) on Γ ϵ \Gamma_{\epsilon} Γ ϵ , the Stokes flow outside of the slender body Σ ϵ \Sigma_{\epsilon} Σ ϵ may be represented by
u ( x ) = D [ φ ] ( x ) + V [ φ ] ( x ) (274) u(x)=\mathcal{D}[\varphi](x)+\mathcal{V}[\varphi](x) \tag*{(274)} u ( x ) = D [ φ ] ( x ) + V [ φ ] ( x ) (274)
where D \mathcal{D} D is the double layer operator given by (39) and V \mathcal{V} V is a completion flow serving the following purpose. Along the slender body surface, the density φ \varphi φ must satisfy the integral equation
v ( x ) = ( 1 2 I + D ) [ φ ] ( x ) + V [ φ ] ( x ) , x ∈ Γ ϵ , (275) v(x)=\left(\frac{1}{2}I+\mathcal{D}\right)[\varphi](x)+\mathcal{V}[\varphi](x),\qquad x\in\Gamma_{\epsilon}, \tag*{(275)} v ( x ) = ( 2 1 I + D ) [ φ ] ( x ) + V [ φ ] ( x ) , x ∈ Γ ϵ , (275)
where the 1 2 I \frac{1}{2}I 2 1 I arises from the jump condition for the double layer operator on Γ ϵ \Gamma_{\epsilon} Γ ϵ . Since the operator 1 2 I + D \frac{1}{2}I+\mathcal{D} 2 1 I + D has a well-known 6-dimensional nullspace along Γ ϵ \Gamma_{\epsilon} Γ ϵ consisting of rigid motions [62 ] , an appropriate choice of completion flow V \mathcal{V} V is necessary for invertibility.
There are many possible choices for V \mathcal{V} V (see [16 , 61 ] ). We will use a form of V \mathcal{V} V proposed by [35 ] that is particularly well-suited for slender filaments. Consider x ′ ∈ Γ ϵ x^{\prime}\in\Gamma_{\epsilon} x ′ ∈ Γ ϵ , denoted by x ′ = X ( s ′ ) + ϵ e r ( s ′ , θ ′ ) x^{\prime}=X(s^{\prime})+\epsilon e_{r}(s^{\prime},\theta^{\prime}) x ′ = X ( s ′ ) + ϵ e r ( s ′ , θ ′ ) , and consider the velocity field V \mathcal{V} V given by
V [ φ ] ( x ) = ∫ Γ ϵ G ( x , X ( s ′ ) ) φ ( x ′ ) d S x ′ + ∫ Γ ϵ l R ( x , X ( s ′ ) ) × ( ( x ′ − X ( s ′ ) ) × φ ( x ′ ) ) d S x ′ . (276) \mathcal{V}[\varphi](x)=\int_{\Gamma_{\epsilon}}G(x,X(s^{\prime}))\varphi(x^{\prime})\,dS_{x^{\prime}}+\int_{\Gamma_{\epsilon}}l_{R}(x,X(s^{\prime}))\times\left((x^{\prime}-X(s^{\prime}))\times\varphi(x^{\prime})\right)\,dS_{x^{\prime}}. \tag*{(276)} V [ φ ] ( x ) = ∫ Γ ϵ G ( x , X ( s ′ )) φ ( x ′ ) d S x ′ + ∫ Γ ϵ l R ( x , X ( s ′ )) × ( ( x ′ − X ( s ′ )) × φ ( x ′ ) ) d S x ′ . (276)
We emphasize that X ( s ′ ) X(s^{\prime}) X ( s ′ ) corresponds to x ′ x^{\prime} x ′ while x x x is independent. Here l R l_{R} l R is a rotlet , given by
l R ( x , y ) = − x − y ∣ x − y ∣ 3 , (277) l_{R}(x,y)=-\frac{x-y}{\lvert x-y\rvert^{3}}, \tag*{(277)} l R ( x , y ) = − ∣ x − y ∣ 3 x − y , (277)
and describes the Stokes flow due to a point torque at y y y [15 ] . The completion flow (276) may be understood as the Stokes flow due to a line density of point forces and point torques distributed along the filament centerline. Note that V \mathcal{V} V is a smooth function of x x x along Γ ϵ \Gamma_{\epsilon} Γ ϵ since these singularities lie along the centerline of the filament rather than the surface.
Given the representation (274) , (275) of the Stokes flow u ( x ) u(x) u ( x ) in Ω ϵ \Omega_{\epsilon} Ω ϵ with Dirichlet data v ( x ) v(x) v ( x ) , we may calculate the corresponding Neumann boundary value for u u u along Γ ϵ \Gamma_{\epsilon} Γ ϵ . We begin by noting the following expression for the stress due to the double layer D [ φ ] ( x ) \mathcal{D}[\varphi](x) D [ φ ] ( x ) , x ∈ Ω ϵ x\in\Omega_{\epsilon} x ∈ Ω ϵ :
σ [ D [ φ ] ] ( x ) = ∇ x D [ φ ] + ( ∇ x D [ φ ] ) T − p D [ φ ] ( x ) I , \sigma[\mathcal{D}[\varphi]](x)=\nabla_{x}\mathcal{D}[\varphi]+\left(\nabla_{x}\mathcal{D}[\varphi]\right)^{\mathsf{T}}-p^{\mathcal{D}}[\varphi](x)I, σ [ D [ φ ]] ( x ) = ∇ x D [ φ ] + ( ∇ x D [ φ ] ) T − p D [ φ ] ( x ) I ,
p D [ φ ] ( x ) = ∫ Γ ϵ p D ( x , x ′ ) ⋅ φ ( x ′ ) d S x ′ , p^{\mathcal{D}}[\varphi](x)=\int_{\Gamma_{\epsilon}}p^{\mathcal{D}}(x,x^{\prime})\cdot\varphi(x^{\prime})\,dS_{x^{\prime}}, p D [ φ ] ( x ) = ∫ Γ ϵ p D ( x , x ′ ) ⋅ φ ( x ′ ) d S x ′ ,
p D ( x , x ′ ; n ( x ′ ) ) = 1 2 π ( − n ( x ′ ) ∣ x − x ′ ∣ 3 + 3 ( x − x ′ ) ( x − x ′ ) ⋅ n ( x ′ ) ∣ x − x ′ ∣ 5 ) . (278) p^{\mathcal{D}}(x,x^{\prime};n(x^{\prime}))=\frac{1}{2\pi}\left(-\frac{n(x^{\prime})}{\lvert x-x^{\prime}\rvert^{3}}+3\frac{(x-x^{\prime})(x-x^{\prime})\cdot n(x^{\prime})}{\lvert x-x^{\prime}\rvert^{5}}\right). \tag*{(278)} p D ( x , x ′ ; n ( x ′ )) = 2 π 1 ( − ∣ x − x ′ ∣ 3 n ( x ′ ) + 3 ∣ x − x ′ ∣ 5 ( x − x ′ ) ( x − x ′ ) ⋅ n ( x ′ ) ) . (278)
Here p D p^{D} p D is the pressure associated with the double layer representation of u \boldsymbol{u} u and p D \boldsymbol{p}^{D} p D is the vector-valued pressure kernel for this representation [15 , 47 , 62 ] .
Letting R = x − x ′ \boldsymbol{R}=\boldsymbol{x}-\boldsymbol{x}' R = x − x ′ , n = n ( x ) \boldsymbol{n}=\boldsymbol{n}(\boldsymbol{x}) n = n ( x ) , and n ′ = n ( x ′ ) \boldsymbol{n}'=\boldsymbol{n}(\boldsymbol{x}') n ′ = n ( x ′ ) , we define the following kernel corresponding to the double layer stress tensor (278) :
K T ( x , x ′ ; n , n ′ ) = 1 4 π ( 2 n ⊗ n ′ ∣ R ∣ 3 + 3 ( R ⋅ n ′ ) R ⊗ n + ( n ⋅ n ′ ) R ⊗ R ∣ R ∣ 5 + 3 ( R ⋅ n ) ( R ⋅ n ′ ) I + ( R ⋅ n ) n ′ ⊗ R ∣ R ∣ 5 − 30 ( R ⋅ n ) ( R ⋅ n ′ ) R ⊗ R ∣ R ∣ 7 ) . (279) \boldsymbol{K}_{T}(\boldsymbol{x},\boldsymbol{x}';\boldsymbol{n},\boldsymbol{n}')=\frac{1}{4\pi}\left(2\frac{\boldsymbol{n}\otimes\boldsymbol{n}'}{\lvert\boldsymbol{R}\rvert^{3}}+3\frac{(\boldsymbol{R}\cdot\boldsymbol{n}')\boldsymbol{R}\otimes\boldsymbol{n}+(\boldsymbol{n}\cdot\boldsymbol{n}')\boldsymbol{R}\otimes\boldsymbol{R}}{\lvert\boldsymbol{R}\rvert^{5}}+3\frac{(\boldsymbol{R}\cdot\boldsymbol{n})(\boldsymbol{R}\cdot\boldsymbol{n}')\boldsymbol{I}+(\boldsymbol{R}\cdot\boldsymbol{n})\boldsymbol{n}'\otimes\boldsymbol{R}}{\lvert\boldsymbol{R}\rvert^{5}}-30\frac{(\boldsymbol{R}\cdot\boldsymbol{n})(\boldsymbol{R}\cdot\boldsymbol{n}')\boldsymbol{R}\otimes\boldsymbol{R}}{\lvert\boldsymbol{R}\rvert^{7}}\right). \tag*{(279)} K T ( x , x ′ ; n , n ′ ) = 4 π 1 ( 2 ∣ R ∣ 3 n ⊗ n ′ + 3 ∣ R ∣ 5 ( R ⋅ n ′ ) R ⊗ n + ( n ⋅ n ′ ) R ⊗ R + 3 ∣ R ∣ 5 ( R ⋅ n ) ( R ⋅ n ′ ) I + ( R ⋅ n ) n ′ ⊗ R − 30 ∣ R ∣ 7 ( R ⋅ n ) ( R ⋅ n ′ ) R ⊗ R ) . (279)
Then for x ∈ Γ ϵ \boldsymbol{x}\in\Gamma_{\epsilon} x ∈ Γ ϵ , on the surface of the filament, we may consider the stress associated to the double layer operator as
σ [ D [ φ ] ] n ( x ) = p.v. ∫ Γ ϵ K T ( x , x ′ ; n , n ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ . (280) \boldsymbol{\sigma}[\mathcal{D}[\boldsymbol{\varphi}]]\boldsymbol{n}(\boldsymbol{x})=\operatorname{p.v.}\int_{\Gamma_{\epsilon}}\boldsymbol{K}_{T}(\boldsymbol{x},\boldsymbol{x}';\boldsymbol{n},\boldsymbol{n}')\left(\boldsymbol{\varphi}(\boldsymbol{x}')-\boldsymbol{\varphi}(\boldsymbol{x})\right)\,dS_{\boldsymbol{x}'}. \tag*{(280)} σ [ D [ φ ]] n ( x ) = p.v. ∫ Γ ϵ K T ( x , x ′ ; n , n ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ . (280)
Here the normal vectors n \boldsymbol{n} n and n ′ \boldsymbol{n}' n ′ both point out of the slender body Σ ϵ \Sigma_{\epsilon} Σ ϵ , into the fluid domain Ω ϵ \Omega_{\epsilon} Ω ϵ .
To obtain the representation (280) , we use that both the double layer kernel K D \boldsymbol{K}_{D} K D (38) and the corresponding pressure kernel p D \boldsymbol{p}^{D} p D (278) integrate to zero within the exterior domain Ω ϵ \Omega_{\epsilon} Ω ϵ , i.e.
∫ Γ ϵ K D ( y , x ′ ) d S x ′ = 0 , ∫ Γ ϵ p D ( y , x ′ ; n ( x ′ ) ) d S x ′ = 0 , y ∈ Ω ϵ . \int_{\Gamma_{\epsilon}}\boldsymbol{K}_{D}(\boldsymbol{y},\boldsymbol{x}')\,dS_{\boldsymbol{x}'}=0,\qquad\int_{\Gamma_{\epsilon}}\boldsymbol{p}^{D}(\boldsymbol{y},\boldsymbol{x}';\boldsymbol{n}(\boldsymbol{x}'))\,dS_{\boldsymbol{x}'}=0,\qquad\boldsymbol{y}\in\Omega_{\epsilon}. ∫ Γ ϵ K D ( y , x ′ ) d S x ′ = 0 , ∫ Γ ϵ p D ( y , x ′ ; n ( x ′ )) d S x ′ = 0 , y ∈ Ω ϵ .
For any constant-in-y \boldsymbol{y} y vector-valued surface density φ ( x ) \boldsymbol{\varphi}(\boldsymbol{x}) φ ( x ) , x ∈ Γ ϵ \boldsymbol{x}\in\Gamma_{\epsilon} x ∈ Γ ϵ , we then have
∇ y ∫ Γ ϵ K D ( y , x ′ ) φ ( x ) d S x ′ = 0 , ∫ Γ ϵ p D ( y , x ′ ; n ( x ′ ) ) ⋅ φ ( x ) d S x ′ = 0 ; y ∈ Ω ϵ , x ∈ Γ ϵ . \nabla_{\boldsymbol{y}}\int_{\Gamma_{\epsilon}}\boldsymbol{K}_{D}(\boldsymbol{y},\boldsymbol{x}')\boldsymbol{\varphi}(\boldsymbol{x})\,dS_{\boldsymbol{x}'}=0,\qquad\int_{\Gamma_{\epsilon}}\boldsymbol{p}^{D}(\boldsymbol{y},\boldsymbol{x}';\boldsymbol{n}(\boldsymbol{x}'))\cdot\boldsymbol{\varphi}(\boldsymbol{x})\,dS_{\boldsymbol{x}'}=0;\qquad\boldsymbol{y}\in\Omega_{\epsilon},\ \boldsymbol{x}\in\Gamma_{\epsilon}. ∇ y ∫ Γ ϵ K D ( y , x ′ ) φ ( x ) d S x ′ = 0 , ∫ Γ ϵ p D ( y , x ′ ; n ( x ′ )) ⋅ φ ( x ) d S x ′ = 0 ; y ∈ Ω ϵ , x ∈ Γ ϵ .
Putting these together, for x ∈ Γ ϵ \boldsymbol{x}\in\Gamma_{\epsilon} x ∈ Γ ϵ , we may understand σ [ D [ φ ] ] n ( x ) \boldsymbol{\sigma}[\mathcal{D}[\boldsymbol{\varphi}]]\boldsymbol{n}(\boldsymbol{x}) σ [ D [ φ ]] n ( x ) as
σ [ D [ φ ] ] n ( x ) = lim y → x ( ∇ y ∫ Γ ϵ K D ( y , x ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ + ( ∇ y ∫ Γ ϵ K D ( y , x ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ ) T − I ∫ Γ ϵ p D ( y , x ′ ; n ( x ′ ) ) ⋅ ( φ ( x ′ ) − φ ( x ) ) d S x ′ ) n ( x ) = lim y → x ∫ Γ ϵ K T ( y , x ′ ; n , n ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ , y ∈ Ω ϵ . \begin{aligned}
\boldsymbol{\sigma}[\mathcal{D}[\boldsymbol{\varphi}]]\boldsymbol{n}(\boldsymbol{x})={}&\lim_{\boldsymbol{y}\to\boldsymbol{x}}\left(\nabla_{\boldsymbol{y}}\int_{\Gamma_{\epsilon}}\boldsymbol{K}_{D}(\boldsymbol{y},\boldsymbol{x}')\left(\boldsymbol{\varphi}(\boldsymbol{x}')-\boldsymbol{\varphi}(\boldsymbol{x})\right)\,dS_{\boldsymbol{x}'}\right.\\
&+\left(\nabla_{\boldsymbol{y}}\int_{\Gamma_{\epsilon}}\boldsymbol{K}_{D}(\boldsymbol{y},\boldsymbol{x}')\left(\boldsymbol{\varphi}(\boldsymbol{x}')-\boldsymbol{\varphi}(\boldsymbol{x})\right)\,dS_{\boldsymbol{x}'}\right)^{\mathrm{T}}\\
&\left.-\boldsymbol{I}\int_{\Gamma_{\epsilon}}\boldsymbol{p}^{D}(\boldsymbol{y},\boldsymbol{x}';\boldsymbol{n}(\boldsymbol{x}'))\cdot\left(\boldsymbol{\varphi}(\boldsymbol{x}')-\boldsymbol{\varphi}(\boldsymbol{x})\right)\,dS_{\boldsymbol{x}'}\right)\boldsymbol{n}(\boldsymbol{x})\\
={}&\lim_{\boldsymbol{y}\to\boldsymbol{x}}\int_{\Gamma_{\epsilon}}\boldsymbol{K}_{T}(\boldsymbol{y},\boldsymbol{x}';\boldsymbol{n},\boldsymbol{n}')\left(\boldsymbol{\varphi}(\boldsymbol{x}')-\boldsymbol{\varphi}(\boldsymbol{x})\right)\,dS_{\boldsymbol{x}'},\qquad\boldsymbol{y}\in\Omega_{\epsilon}.
\end{aligned} σ [ D [ φ ]] n ( x ) = = y → x lim ( ∇ y ∫ Γ ϵ K D ( y , x ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ + ( ∇ y ∫ Γ ϵ K D ( y , x ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ ) T − I ∫ Γ ϵ p D ( y , x ′ ; n ( x ′ )) ⋅ ( φ ( x ′ ) − φ ( x ) ) d S x ′ ) n ( x ) y → x lim ∫ Γ ϵ K T ( y , x ′ ; n , n ′ ) ( φ ( x ′ ) − φ ( x ) ) d S x ′ , y ∈ Ω ϵ .
We may also calculate the stress corresponding to the completion flow V \boldsymbol{V} V , given by
σ [ V [ φ ] ] ( x ) = ∇ x V [ φ ] + ( ∇ x V [ φ ] ) T + p V ( x ) I , \boldsymbol{\sigma}[\boldsymbol{V}[\boldsymbol{\varphi}]](\boldsymbol{x})=\nabla_{\boldsymbol{x}}\boldsymbol{V}[\boldsymbol{\varphi}]+(\nabla_{\boldsymbol{x}}\boldsymbol{V}[\boldsymbol{\varphi}])^{\mathrm{T}}+p^{V}(\boldsymbol{x})\boldsymbol{I}, σ [ V [ φ ]] ( x ) = ∇ x V [ φ ] + ( ∇ x V [ φ ] ) T + p V ( x ) I ,
p V ( x ) = 1 4 π ∫ Γ ϵ ( x − X ( s ′ ) ) ⋅ φ ( x ′ ) ∣ x − X ( s ′ ) ∣ 3 d S x ′ . p^{V}(\boldsymbol{x})=\frac{1}{4\pi}\int_{\Gamma_{\epsilon}}\frac{(\boldsymbol{x}-\boldsymbol{X}(s'))\cdot\boldsymbol{\varphi}(\boldsymbol{x}')}{\lvert\boldsymbol{x}-\boldsymbol{X}(s')\rvert^{3}}\,dS_{\boldsymbol{x}'}. p V ( x ) = 4 π 1 ∫ Γ ϵ ∣ x − X ( s ′ ) ∣ 3 ( x − X ( s ′ )) ⋅ φ ( x ′ ) d S x ′ .
Here p V p^{V} p V is the pressure corresponding to the flow V \boldsymbol{V} V [47 ] . For x ∈ Γ ϵ \boldsymbol{x}\in\Gamma_{\epsilon} x ∈ Γ ϵ , we define R X = x − X ( s ′ ) \boldsymbol{R}_{X}=\boldsymbol{x}-\boldsymbol{X}(s') R X = x − X ( s ′ ) and let n = n ( x ) \boldsymbol{n}=\boldsymbol{n}(\boldsymbol{x}) n = n ( x ) . Along Γ ϵ \Gamma_{\epsilon} Γ ϵ , we may then write
σ [ V [ φ ] ] n ( x ) = 3 8 π ∫ Γ ϵ ( − 2 ( R X ⋅ n ) ( R X ⋅ φ ( x ′ ) ) R X ∣ R X ∣ 5 + R X ⋅ n ∣ R X ∣ 5 R X × ( ( x ′ − X ( s ′ ) ) × φ ( x ′ ) ) + R X ∣ R X ∣ 5 n ⋅ ( R X × ( ( x ′ − X ( s ′ ) ) × φ ( x ′ ) ) ) ) d S x ′ . (281) \boldsymbol{\sigma}[\boldsymbol{V}[\boldsymbol{\varphi}]]\boldsymbol{n}(\boldsymbol{x})=\frac{3}{8\pi}\int_{\Gamma_{\epsilon}}\left(-2\frac{(\boldsymbol{R}_{X}\cdot\boldsymbol{n})(\boldsymbol{R}_{X}\cdot\boldsymbol{\varphi}(\boldsymbol{x}'))\boldsymbol{R}_{X}}{\lvert\boldsymbol{R}_{X}\rvert^{5}}+\frac{\boldsymbol{R}_{X}\cdot\boldsymbol{n}}{\lvert\boldsymbol{R}_{X}\rvert^{5}}\boldsymbol{R}_{X}\times\left((\boldsymbol{x}'-\boldsymbol{X}(s'))\times\boldsymbol{\varphi}(\boldsymbol{x}')\right)+\frac{\boldsymbol{R}_{X}}{\lvert\boldsymbol{R}_{X}\rvert^{5}}\boldsymbol{n}\cdot\left(\boldsymbol{R}_{X}\times\left((\boldsymbol{x}'-\boldsymbol{X}(s'))\times\boldsymbol{\varphi}(\boldsymbol{x}')\right)\right)\right)\,dS_{\boldsymbol{x}'}. \tag*{(281)} σ [ V [ φ ]] n ( x ) = 8 π 3 ∫ Γ ϵ ( − 2 ∣ R X ∣ 5 ( R X ⋅ n ) ( R X ⋅ φ ( x ′ )) R X + ∣ R X ∣ 5 R X ⋅ n R X × ( ( x ′ − X ( s ′ )) × φ ( x ′ ) ) + ∣ R X ∣ 5 R X n ⋅ ( R X × ( ( x ′ − X ( s ′ )) × φ ( x ′ ) ) ) ) d S x ′ . (281)
Given the representations (279) and (280), we define the hypersingular operator T \mathcal{T} T to be the Neumann boundary value associated with the flow (273), given by
T [ φ ] ( x ) = − σ [ D [ φ ] ] n ( x ) − σ [ V [ φ ] ] n ( x ) , x ∈ Γ ϵ . (282) \mathcal{T}[\varphi](x)=-\sigma[\mathcal{D}[\varphi]]\boldsymbol{n}(x)-\sigma[\mathcal{V}[\varphi]]\boldsymbol{n}(x),\qquad x\in\Gamma_{\epsilon}. \tag*{(282)} T [ φ ] ( x ) = − σ [ D [ φ ]] n ( x ) − σ [ V [ φ ]] n ( x ) , x ∈ Γ ϵ . (282)
Here we note that the negative sign arises since n \boldsymbol{n} n points into the fluid domain, out of the slender body Σ ϵ \Sigma_{\epsilon} Σ ϵ .
Given Dirichlet data v ( x ) \boldsymbol{v}(x) v ( x ) along Γ ϵ \Gamma_{\epsilon} Γ ϵ , we may use the completed double layer representation (273), (274) of u \boldsymbol{u} u to calculate the corresponding Neumann boundary value w \boldsymbol{w} w via
w ( x ) = T [ ( 1 2 I + D + V ) − 1 [ v ] ] ( x ) , x ∈ Γ ϵ . (283) \boldsymbol{w}(x)=\mathcal{T}\left[\left(\frac{1}{2}\mathcal{I}+\mathcal{D}+\mathcal{V}\right)^{-1}[\boldsymbol{v}]\right](x),\qquad x\in\Gamma_{\epsilon}. \tag*{(283)} w ( x ) = T [ ( 2 1 I + D + V ) − 1 [ v ] ] ( x ) , x ∈ Γ ϵ . (283)
We will use the representation (282) of the full Neumann boundary value w \boldsymbol{w} w to prove Lemma 1.13. In order to do so, we will rely on the following series of lemmas bounding each of the components of (282). We state each lemma below and then immediately use the lemmas to prove Lemma 1.13. The proof of each component lemma is then given in subsequent sections 5.1, 5.2, and 5.3.
We begin with bounds for the completed double layer operator (274) along Γ ϵ \Gamma_{\epsilon} Γ ϵ .
Lemma 5.2 (Completed double layer smoothing). Let 0 < γ < β < 1 0<\gamma<\beta<1 0 < γ < β < 1 and consider a filament Σ ϵ \Sigma_{\epsilon} Σ ϵ with centerline X ∈ C 2 , β \boldsymbol{X}\in C^{2,\beta} X ∈ C 2 , β . Given φ ∈ C 1 , α ( Γ ϵ ) \varphi\in C^{1,\alpha}(\Gamma_{\epsilon}) φ ∈ C 1 , α ( Γ ϵ ) , the double layer operator D \mathcal{D} D (39) and completion flow V \mathcal{V} V (275) along Γ ϵ \Gamma_{\epsilon} Γ ϵ satisfy
∥ ( D + V ) [ φ ] ∥ C 1 , γ ( Γ ϵ ) ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ ( Γ ϵ ) (284) \left\|(\mathcal{D}+\mathcal{V})[\varphi]\right\|_{C^{1,\gamma}(\Gamma_{\epsilon})}\leq c(\kappa_{*,\gamma^{+}},c_{\Gamma})\epsilon^{-1-\gamma}\left\|\varphi\right\|_{C^{0,\gamma}(\Gamma_{\epsilon})} \tag*{(284)} ∥ ( D + V ) [ φ ] ∥ C 1 , γ ( Γ ϵ ) ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ ( Γ ϵ ) (284) for any γ + ∈ ( γ , β ] \gamma^{+}\in(\gamma,\beta] γ + ∈ ( γ , β ] . Furthermore, given two nearby filaments with centerlines X ( a ) \boldsymbol{X}^{(a)} X ( a ) , X ( b ) \boldsymbol{X}^{(b)} X ( b ) satisfying Lemma 1.1, we have
∥ ( D ( a ) + V ( a ) − D ( b ) − V ( b ) ) [ φ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ φ ∥ C 0 , γ . (285) \left\|(\mathcal{D}^{(a)}+\mathcal{V}^{(a)}-\mathcal{D}^{(b)}-\mathcal{V}^{(b)})[\varphi]\right\|_{C^{1,\gamma}}\leq c(\kappa_{*,\gamma^{+}}^{(a)},\kappa_{*,\gamma^{+}}^{(b)},c_{\Gamma})\epsilon^{-1-\gamma}\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{2,\gamma^{+}}}\left\|\varphi\right\|_{C^{0,\gamma}}. \tag*{(285)} ( D ( a ) + V ( a ) − D ( b ) − V ( b ) ) [ φ ] C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ X ( a ) − X ( b ) C 2 , γ + ∥ φ ∥ C 0 , γ . (285)
The proof of Lemma 5.2 appears in section 5.1. We will additionally rely on the following uniform bound for ( 1 2 I + D + V ) − 1 (\frac{1}{2}\mathcal{I}+\mathcal{D}+\mathcal{V})^{-1} ( 2 1 I + D + V ) − 1 with respect to the filament centerline shape.
Lemma 5.3 (Inverse completed double layer). Let 0 < γ < β < 1 0<\gamma<\beta<1 0 < γ < β < 1 , and let ϵ > 0 \epsilon>0 ϵ > 0 be fixed. Given any filament Σ ϵ \Sigma_{\epsilon} Σ ϵ with centerline X ( s ) ∈ C 2 , β ( T ) \boldsymbol{X}(s)\in C^{2,\beta}(\mathbb{T}) X ( s ) ∈ C 2 , β ( T ) satisfying the non-self-intersection condition (2), the operator ( 1 2 I + D + V ) − 1 (\frac{1}{2}\mathcal{I}+\mathcal{D}+\mathcal{V})^{-1} ( 2 1 I + D + V ) − 1 is uniformly bounded with respect to the filament centerline shape. In particular, for φ ∈ C 0 , γ ( Γ ϵ ) \varphi\in C^{0,\gamma}(\Gamma_{\epsilon}) φ ∈ C 0 , γ ( Γ ϵ ) , we have
∥ ( 1 2 I + D + V ) − 1 [ φ ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ φ ∥ C 0 , γ (286) \left\|(\tfrac{1}{2}\mathcal{I}+\mathcal{D}+\mathcal{V})^{-1}[\varphi]\right\|_{C^{0,\gamma}}\leq c(\epsilon,\kappa_{*,\gamma^{+}},c_{\Gamma})\left\|\varphi\right\|_{C^{0,\gamma}} \tag*{(286)} ( 2 1 I + D + V ) − 1 [ φ ] C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ φ ∥ C 0 , γ (286) for any γ + ∈ ( γ , β ] \gamma^{+}\in(\gamma,\beta] γ + ∈ ( γ , β ] . Moreover, given two nearby filaments with centerlines X ( a ) \boldsymbol{X}^{(a)} X ( a ) , X ( b ) \boldsymbol{X}^{(b)} X ( b ) satisfying Lemma 1.1, we have
∥ ( 1 2 I + D ( a ) + V ( a ) ) − 1 [ φ ] − ( 1 2 I + D ( b ) + V ( b ) ) − 1 [ φ ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ φ ∥ C 0 , γ . (287) \left\|(\tfrac{1}{2}\mathcal{I}+\mathcal{D}^{(a)}+\mathcal{V}^{(a)})^{-1}[\varphi]-(\tfrac{1}{2}\mathcal{I}+\mathcal{D}^{(b)}+\mathcal{V}^{(b)})^{-1}[\varphi]\right\|_{C^{0,\gamma}}
\leq c(\epsilon,\kappa_{*,\gamma^{+}}^{(a)},\kappa_{*,\gamma^{+}}^{(b)},c_{\Gamma})\left\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\right\|_{C^{2,\gamma^{+}}}\left\|\varphi\right\|_{C^{0,\gamma}}. \tag*{(287)} ( 2 1 I + D ( a ) + V ( a ) ) − 1 [ φ ] − ( 2 1 I + D ( b ) + V ( b ) ) − 1 [ φ ] C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) X ( a ) − X ( b ) C 2 , γ + ∥ φ ∥ C 0 , γ . (287)
The proof of Lemma 5.3 is given in section 5.2. Note that the ϵ \epsilon ϵ -dependence here is not explicit.
The final ingredient used in the proof of Lemma 1.13 involves the mapping properties of the hypersingular operator T \mathcal{T} T given by (281). To determine these mapping properties, we will first need the following auxiliary lemma. The bound below is a more general version of the result appearing in [59 Lemma 4.3] .
Lemma 5.4 (Hypersingular operator: estimating ∣ ⋅ ∣ C ˙ 0 , α |\cdot|_{\dot{C}^{0,\alpha}} ∣ ⋅ ∣ C ˙ 0 , α seminorms). Let ℓ , k , n \ell,k,n ℓ , k , n be nonnegative integers satisfying ℓ + k + 3 = n \ell+k+3=n ℓ + k + 3 = n and ℓ + k \ell+k ℓ + k is even. Suppose φ ( s , θ ) ∈ C 1 , α ( Γ ϵ ) \varphi(s,\theta)\in C^{1,\alpha}(\Gamma_\epsilon) φ ( s , θ ) ∈ C 1 , α ( Γ ϵ ) and g ( s , θ , s − s ^ , θ − θ ^ ) ∈ C 0 , α + ( Γ ϵ × Γ ϵ ) g(s,\theta,s-\widehat{s},\theta-\widehat{\theta})\in C^{0,\alpha^+}(\Gamma_\epsilon\times\Gamma_\epsilon) g ( s , θ , s − s , θ − θ ) ∈ C 0 , α + ( Γ ϵ × Γ ϵ ) for any 0 < α < α + < 1 0<\alpha<\alpha^+<1 0 < α < α + < 1 . For R \boldsymbol{R} R as in (206) , let I ℓ k n ( s , θ , s ^ , θ ^ ) I_{\ell kn}(s,\theta,\widehat{s},\widehat{\theta}) I ℓ k n ( s , θ , s , θ ) denote
I φ ; ℓ k n ( s , θ , s ^ , θ ^ ) : = s ℓ ( ϵ sin ( θ ^ 2 ) ) k g ( s , θ , s − s ^ , θ − θ ^ ) ∣ R ( s , θ , s ^ , θ ^ ) ∣ n ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) ) . (288) I_{\varphi;\ell kn}(s,\theta,\widehat{s},\widehat{\theta}):=\frac{s^\ell\left(\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\right)^k g(s,\theta,s-\widehat{s},\theta-\widehat{\theta})}{\lvert\boldsymbol{R}(s,\theta,\widehat{s},\widehat{\theta})\rvert^n}\left(\varphi(s-\widehat{s},\theta-\widehat{\theta})-\varphi(s,\theta)\right). \tag*{(288)} I φ ; ℓ k n ( s , θ , s , θ ) := ∣ R ( s , θ , s , θ ) ∣ n s ℓ ( ϵ sin ( 2 θ ) ) k g ( s , θ , s − s , θ − θ ) ( φ ( s − s , θ − θ ) − φ ( s , θ ) ) . (288) Then for any ( s 0 , θ 0 ) ∈ [ − 1 / 2 , 1 / 2 ] × [ − π , π ] (s_0,\theta_0)\in[-1/2,1/2]\times[-\pi,\pi] ( s 0 , θ 0 ) ∈ [ − 1/2 , 1/2 ] × [ − π , π ] with s 0 2 + θ 0 2 ≠ 0 s_0^2+\theta_0^2\ne0 s 0 2 + θ 0 2 = 0 we have
∣ p . v . ∫ − s 0 + 1 / 2 − s 0 − 1 / 2 ∫ − θ 0 + π − θ 0 − π I φ ; ℓ k n ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) ϵ d θ ^ d s ^ − p . v . ∫ − 1 / 2 1 / 2 ∫ − π π I φ ; ℓ k n ( s , θ , s ^ , θ ^ ) ϵ d θ ^ d s ^ ∣ \left|\mathop{\mathrm{p.v.}}\int_{-s_0+1/2}^{-s_0-1/2}\int_{-\theta_0+\pi}^{-\theta_0-\pi} I_{\varphi;\ell kn}(s_0+s,\theta_0+\theta,s_0+\widehat{s},\theta_0+\widehat{\theta})\,\epsilon\,\mathrm{d}\widehat{\theta}\mathrm{d}\widehat{s}
-\mathop{\mathrm{p.v.}}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}I_{\varphi;\ell kn}(s,\theta,\widehat{s},\widehat{\theta})\,\epsilon\,\mathrm{d}\widehat{\theta}\mathrm{d}\widehat{s}\right| p.v. ∫ − s 0 + 1/2 − s 0 − 1/2 ∫ − θ 0 + π − θ 0 − π I φ ; ℓ k n ( s 0 + s , θ 0 + θ , s 0 + s , θ 0 + θ ) ϵ d θ d s − p.v. ∫ − 1/2 1/2 ∫ − π π I φ ; ℓ k n ( s , θ , s , θ ) ϵ d θ d s ≤ c ( κ ∗ , c Γ ) ( ( ∥ ∂ s φ ∥ L ∞ + ∥ 1 ϵ ∂ θ φ ∥ L ∞ ) ( ∥ g ∥ C 1 0 , α + + ∥ g ∥ C 2 0 , α ) + ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α , (289) \le c(\kappa_*,c_\Gamma)\left(\left(\|\partial_s\varphi\|_{L^\infty}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{L^\infty}\right)\left(\|g\|_{C_1^{0,\alpha^+}}+\|g\|_{C_2^{0,\alpha}}\right)+\left(\|\partial_s\varphi\|_{C_0,\alpha}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C_0,\alpha}\right)\|g\|_{L^\infty}\right)\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\,\alpha}, \tag*{(289)} ≤ c ( κ ∗ , c Γ ) ( ( ∥ ∂ s φ ∥ L ∞ + ϵ 1 ∂ θ φ L ∞ ) ( ∥ g ∥ C 1 0 , α + + ∥ g ∥ C 2 0 , α ) + ( ∥ ∂ s φ ∥ C 0 , α + ϵ 1 ∂ θ φ C 0 , α ) ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α , (289) where ∥ ⋅ ∥ C 1 0 , α + \|\cdot\|_{C_1^{0,\alpha^+}} ∥ ⋅ ∥ C 1 0 , α + and ∥ ⋅ ∥ C 2 0 , α \|\cdot\|_{C_2^{0,\alpha}} ∥ ⋅ ∥ C 2 0 , α are as in (200) .
The proof of Lemma 5.4 appears in Appendix D. Using Lemma 5.4, we may then show the final ingredient needed in the proof of Lemma 1.13. In particular, we show the following bounds for the hypersingular operator T \mathcal{T} T given by (282) .
Lemma 5.5 (Hypersingular operator bounds). Let 0 < α < β < 1 0<\alpha<\beta<1 0 < α < β < 1 . Given a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ∈ C 2 , β \boldsymbol{X}\in C^{2,\beta} X ∈ C 2 , β , we consider the hypersingular operator T \mathcal{T} T given by (282) . Given φ ∈ C 1 , α ( Γ ϵ ) \varphi\in C^{1,\alpha}(\Gamma_\epsilon) φ ∈ C 1 , α ( Γ ϵ ) , we may decompose T [ φ ] \mathcal{T}[\varphi] T [ φ ] as
T [ φ ] = T ϵ [ φ ] + T + [ φ ] \mathcal{T}[\varphi]=\mathcal{T}_\epsilon[\varphi]+\mathcal{T}_+[\varphi] T [ φ ] = T ϵ [ φ ] + T + [ φ ] where
∥ T ϵ [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ C 0 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 0 , α ) \|\mathcal{T}_\epsilon[\varphi]\|_{C_0^{0,\alpha}}\le c(\kappa_*,c_\Gamma)\left(\|\partial_s\varphi\|_{C_0^{0,\alpha}}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C_0^{0,\alpha}}\right) ∥ T ϵ [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ C 0 0 , α + ϵ 1 ∂ θ φ C 0 0 , α ) ∥ T + [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ − 1 − α + ∥ φ ∥ C 1 (290) \|\mathcal{T}_+[\varphi]\|_{C_0^{0,\alpha}}\le c(\kappa_*,\alpha^+,c_\Gamma)\epsilon^{-1-\alpha^+}\|\varphi\|_{C^1} \tag*{(290)} ∥ T + [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ , α + , c Γ ) ϵ − 1 − α + ∥ φ ∥ C 1 (290) for any α + ∈ ( α , β ] \alpha^+\in(\alpha,\beta] α + ∈ ( α , β ] . In addition, given two nearby curves X ( a ) \boldsymbol{X}^{(a)} X ( a ) and X ( b ) \boldsymbol{X}^{(b)} X ( b ) satisfying Lemma 1.1, we have the Lipschitz estimate
∥ T ϵ ( a ) [ φ ] − T ϵ ( b ) [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ( ∥ ∂ s φ ∥ C 0 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 0 , α ) \|\mathcal{T}^{(a)}_\epsilon[\varphi]-\mathcal{T}^{(b)}_\epsilon[\varphi]\|_{C_0^{0,\alpha}}\le c(\kappa_*^{(a)},\kappa_*^{(b)},c_\Gamma)\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2}}\left(\|\partial_s\varphi\|_{C_0^{0,\alpha}}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C_0^{0,\alpha}}\right) ∥ T ϵ ( a ) [ φ ] − T ϵ ( b ) [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ( ∥ ∂ s φ ∥ C 0 0 , α + ϵ 1 ∂ θ φ C 0 0 , α ) ∥ T + ( a ) [ φ ] − T + ( b ) [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ ( a ) , α + , κ ∗ ( b ) , α + , c Γ ) ϵ − 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 1 . (291) \|\mathcal{T}^{(a)}_+[\varphi]-\mathcal{T}^{(b)}_+[\varphi]\|_{C_0^{0,\alpha}}\le c(\kappa_*^{(a)},\alpha^+,\kappa_*^{(b)},\alpha^+,c_\Gamma)\epsilon^{-1-\alpha^+}\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha^+}}\|\varphi\|_{C^1}. \tag*{(291)} ∥ T + ( a ) [ φ ] − T + ( b ) [ φ ] ∥ C 0 0 , α ≤ c ( κ ∗ ( a ) , α + , κ ∗ ( b ) , α + , c Γ ) ϵ − 1 − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ φ ∥ C 1 . (291)
The proof of Lemma 5.5 appears in section 5.3.
Given Lemmas 5.2, 5.3, and 5.5, we may proceed to the proof of the main result of this section, Lemma 1.13.
Proof of Lemma 1.13: bounds on full Neumann boundary value. Given θ \theta θ -independent Dirichlet data v ( s ) v(s) v ( s ) , we may solve for the intermediate density φ ( x ) = φ ( s , θ ) \varphi(x)=\varphi(s,\theta) φ ( x ) = φ ( s , θ ) in the completed double layer representation (275) as
φ ( s , θ ) = 2 v ( s ) − M [ v ( s ) ] , M = ( D + V ) ( 1 2 I + D + V ) − 1 . \varphi(s,\theta)=2v(s)-\mathcal{M}[v(s)],\qquad\mathcal{M}=(\mathcal{D}+\mathcal{V})\left(\frac{1}{2}\mathcal{I}+\mathcal{D}+\mathcal{V}\right)^{-1}. φ ( s , θ ) = 2 v ( s ) − M [ v ( s )] , M = ( D + V ) ( 2 1 I + D + V ) − 1 .
Using Lemmas 5.2 and 5.3, we may estimate
∥ M [ v ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ∥ ( 1 2 I + D + V ) − 1 [ v ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ , 0 < γ < γ + < 1. \|\mathcal{M}[v]\|_{C^{1,\gamma}} \le c(\kappa_{*,\gamma^+},c_{\Gamma})\epsilon^{-1-\gamma}\left\|\left(\frac{1}{2}I+D+V\right)^{-1}[v]\right\|_{C^{0,\gamma}}
\le c(\epsilon,\kappa_{*,\gamma^+},c_{\Gamma})\|v\|_{C^{0,\gamma}}, \qquad0<\gamma<\gamma^+<1. ∥ M [ v ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ( 2 1 I + D + V ) − 1 [ v ] C 0 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ , 0 < γ < γ + < 1.
In addition, for two curves X ( a ) , X ( b ) \mathbf{X}^{(a)},\mathbf{X}^{(b)} X ( a ) , X ( b ) satisfying Lemma 1.1, we obtain the estimate
∥ M ( a ) [ v ] − M ( b ) [ v ] ∥ C 1 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ . \|\mathcal{M}^{(a)}[v]-\mathcal{M}^{(b)}[v]\|_{C^{1,\gamma}}
\le c(\epsilon,\kappa_{*,\gamma^+}^{(a)},\kappa_{*,\gamma^+}^{(b)},c_{\Gamma})
\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma^+}}\|v\|_{C^{0,\gamma}}. ∥ M ( a ) [ v ] − M ( b ) [ v ] ∥ C 1 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ .
Using Lemma 5.5, we may write w ( s , θ ) w(s,\theta) w ( s , θ ) from (282) as
w ( s , θ ) = T [ 2 v ( s ) ] − T [ M [ v ( s ) ] ] = 2 T ϵ [ v ( s ) ] + 2 T + [ v ( s ) ] − T ϵ [ M [ v ( s ) ] ] − T + [ M [ v ( s ) ] ] . \begin{aligned}
w(s,\theta)=\mathcal{T}[2v(s)]-\mathcal{T}[\mathcal{M}[v(s)]] \\
&=2\mathcal{T}_{\epsilon}[v(s)]+2\mathcal{T}_{+}[v(s)]-\mathcal{T}_{\epsilon}[\mathcal{M}[v(s)]]-\mathcal{T}_{+}[\mathcal{M}[v(s)]].
\end{aligned} w ( s , θ ) = T [ 2 v ( s )] − T [ M [ v ( s )]] = 2 T ϵ [ v ( s )] + 2 T + [ v ( s )] − T ϵ [ M [ v ( s )]] − T + [ M [ v ( s )]] .
For v ∈ C 1 , α ( T ) v\in C^{1,\alpha}(\mathbb{T}) v ∈ C 1 , α ( T ) , we have
2 ∥ T ϵ [ v ] ∥ C 0 , α ≤ c ( κ ∗ , c Γ ) ∥ v ∥ C 1 , α , 2\|\mathcal{T}_{\epsilon}[v]\|_{C^{0,\alpha}}\le c(\kappa_*,c_{\Gamma})\|v\|_{C^{1,\alpha}}, 2∥ T ϵ [ v ] ∥ C 0 , α ≤ c ( κ ∗ , c Γ ) ∥ v ∥ C 1 , α ,
2 ∥ T ϵ ( a ) [ v ] − T ϵ ( b ) [ v ] ∥ C 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ∥ v ∥ C 1 , α , 2\|\mathcal{T}_{\epsilon}^{(a)}[v]-\mathcal{T}_{\epsilon}^{(b)}[v]\|_{C^{0,\alpha}}
\le c(\kappa_*^{(a)},\kappa_*^{(b)},c_{\Gamma})\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|v\|_{C^{1,\alpha}}, 2∥ T ϵ ( a ) [ v ] − T ϵ ( b ) [ v ] ∥ C 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ∥ v ∥ C 1 , α ,
while each of the remaining terms are smoother. In particular, for 0 < α < γ < γ + < 1 0<\alpha<\gamma<\gamma^+<1 0 < α < γ < γ + < 1 , we may estimate
∥ T + [ v ] ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ + ∥ v ∥ C 1 \|\mathcal{T}_{+}[v]\|_{C^{0,\gamma}}\le c(\kappa_{*,\gamma^+},c_{\Gamma})\epsilon^{-1-\gamma^+}\|v\|_{C^1} ∥ T + [ v ] ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ + ∥ v ∥ C 1
∥ T ϵ [ M [ v ] ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , c Γ ) ∥ M [ v ] ∥ C 1 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ \|\mathcal{T}_{\epsilon}[\mathcal{M}[v]]\|_{C^{0,\gamma}}
\le c(\epsilon,\kappa_*,c_{\Gamma})\|\mathcal{M}[v]\|_{C^{1,\gamma}}
\le c(\epsilon,\kappa_{*,\gamma^+},c_{\Gamma})\|v\|_{C^{0,\gamma}} ∥ T ϵ [ M [ v ]] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , c Γ ) ∥ M [ v ] ∥ C 1 , γ ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ
∥ T + [ M [ v ] ] ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ + ∥ M [ v ] ∥ C 1 ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ . \|\mathcal{T}_{+}[\mathcal{M}[v]]\|_{C^{0,\gamma}}
\le c(\kappa_{*,\gamma^+},c_{\Gamma})\epsilon^{-1-\gamma^+}\|\mathcal{M}[v]\|_{C^1}
\le c(\epsilon,\kappa_{*,\gamma^+},c_{\Gamma})\|v\|_{C^{0,\gamma}}. ∥ T + [ M [ v ]] ∥ C 0 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ + ∥ M [ v ] ∥ C 1 ≤ c ( ϵ , κ ∗ , γ + , c Γ ) ∥ v ∥ C 0 , γ .
In addition, we obtain the Lipschitz estimates
∥ T + ( a ) [ v ] − T + ( b ) [ v ] ∥ C 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 1 \|\mathcal{T}_{+}^{(a)}[v]-\mathcal{T}_{+}^{(b)}[v]\|_{C^{0,\gamma}}
\le c(\kappa_{*,\gamma^+}^{(a)},\kappa_{*,\gamma^+}^{(b)},c_{\Gamma})\epsilon^{-1-\gamma^+}
\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma^+}}\|v\|_{C^1} ∥ T + ( a ) [ v ] − T + ( b ) [ v ] ∥ C 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 1
∥ T ϵ ( a ) [ M ( a ) [ v ] ] − T ϵ ( b ) [ M ( b ) [ v ] ] ∥ C 0 , γ ≤ ∥ T ϵ ( a ) [ M ( a ) [ v ] − M ( b ) [ v ] ] ∥ C 0 , γ + ∥ T ϵ ( a ) [ M ( b ) [ v ] ] − T ϵ ( b ) [ M ( b ) [ v ] ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ( ∥ M ( a ) [ v ] − M ( b ) [ v ] ∥ C 1 , γ + ∥ X ( a ) − X ( b ) ∥ C 2 ∥ M ( b ) [ v ] ∥ C 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ . \begin{aligned}
\|\mathcal{T}_{\epsilon}^{(a)}[\mathcal{M}^{(a)}[v]]-\mathcal{T}_{\epsilon}^{(b)}[\mathcal{M}^{(b)}[v]]\|_{C^{0,\gamma}}
&\le\|\mathcal{T}_{\epsilon}^{(a)}[\mathcal{M}^{(a)}[v]-\mathcal{M}^{(b)}[v]]\|_{C^{0,\gamma}}\\
&\quad+\|\mathcal{T}_{\epsilon}^{(a)}[\mathcal{M}^{(b)}[v]]-\mathcal{T}_{\epsilon}^{(b)}[\mathcal{M}^{(b)}[v]]\|_{C^{0,\gamma}}\\
&\le c(\epsilon,\kappa_*^{(a)},\kappa_*^{(b)},c_{\Gamma})
\left(\|\mathcal{M}^{(a)}[v]-\mathcal{M}^{(b)}[v]\|_{C^{1,\gamma}}
+\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^2}\|\mathcal{M}^{(b)}[v]\|_{C^{1,\gamma}}\right)\\
&\le c(\epsilon,\kappa_{*,\gamma^+}^{(a)},\kappa_{*,\gamma^+}^{(b)},c_{\Gamma})
\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma^+}}\|v\|_{C^{0,\gamma}}.
\end{aligned} ∥ T ϵ ( a ) [ M ( a ) [ v ]] − T ϵ ( b ) [ M ( b ) [ v ]] ∥ C 0 , γ ≤ ∥ T ϵ ( a ) [ M ( a ) [ v ] − M ( b ) [ v ]] ∥ C 0 , γ + ∥ T ϵ ( a ) [ M ( b ) [ v ]] − T ϵ ( b ) [ M ( b ) [ v ]] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ( ∥ M ( a ) [ v ] − M ( b ) [ v ] ∥ C 1 , γ + ∥ X ( a ) − X ( b ) ∥ C 2 ∥ M ( b ) [ v ] ∥ C 1 , γ ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ .
∥ T + ( a ) [ M ( a ) [ v ] ] − T + ( b ) [ M ( b ) [ v ] ] ∥ C 0 , γ ≤ ∥ T + ( a ) [ M ( a ) [ v ] − M ( b ) [ v ] ] ∥ C 0 , γ + ∥ T + ( a ) [ M ( b ) [ v ] ] − T + ( b ) [ M ( b ) [ v ] ] ∥ C 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ + ( ∥ M ( a ) [ v ] − M ( b ) [ v ] ∥ C 1 + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ M ( b ) [ v ] ∥ C 1 ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ . \begin{aligned}
\|\mathcal{T}_{+}^{(a)}[\mathcal{M}^{(a)}[v]]-\mathcal{T}_{+}^{(b)}[\mathcal{M}^{(b)}[v]]\|_{C^{0,\gamma}}
&\le\|\mathcal{T}_{+}^{(a)}[\mathcal{M}^{(a)}[v]-\mathcal{M}^{(b)}[v]]\|_{C^{0,\gamma}}\\
&\quad+\|\mathcal{T}_{+}^{(a)}[\mathcal{M}^{(b)}[v]]-\mathcal{T}_{+}^{(b)}[\mathcal{M}^{(b)}[v]]\|_{C^{0,\gamma}}\\
&\le c(\kappa_{*,\gamma^+}^{(a)},\kappa_{*,\gamma^+}^{(b)},c_{\Gamma})\epsilon^{-1-\gamma^+}
\left(\|\mathcal{M}^{(a)}[v]-\mathcal{M}^{(b)}[v]\|_{C^1}
+\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma^+}}\|\mathcal{M}^{(b)}[v]\|_{C^1}\right)\\
&\le c(\epsilon,\kappa_{*,\gamma^+}^{(a)},\kappa_{*,\gamma^+}^{(b)},c_{\Gamma})
\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma^+}}\|v\|_{C^{0,\gamma}}.
\end{aligned} ∥ T + ( a ) [ M ( a ) [ v ]] − T + ( b ) [ M ( b ) [ v ]] ∥ C 0 , γ ≤ ∥ T + ( a ) [ M ( a ) [ v ] − M ( b ) [ v ]] ∥ C 0 , γ + ∥ T + ( a ) [ M ( b ) [ v ]] − T + ( b ) [ M ( b ) [ v ]] ∥ C 0 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ + ( ∥ M ( a ) [ v ] − M ( b ) [ v ] ∥ C 1 + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ M ( b ) [ v ] ∥ C 1 ) ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ∥ C 0 , γ .
Combining the above bounds, we obtain Lemma 1.13. □
In the subsequent sections we prove each of the building block lemmas used in the proof of Lemma 1.13. Section 5.1 contains the proof of Lemma 5.2, section 5.2 contains the proof of Lemma 5.3, and section 5.3 contains the proof of Lemma 5.5.
Proof of Lemma 5.2: completed double layer
We begin by considering the double layer operator D D D as
D [ φ ] = D ‾ [ Φ − 1 φ ] + R D [ φ ] \mathcal{D}[\varphi]=\overline{\mathcal{D}}[\Phi^{-1}\varphi]+\mathcal{R}_{D}[\varphi] D [ φ ] = D [ Φ − 1 φ ] + R D [ φ ]
where D ‾ \overline{\mathcal{D}} D and R D \mathcal{R}_{D} R D are as in (64) . By Lemma 1.11, we have that R D \mathcal{R}_{D} R D satisfies
∥ R D [ φ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ φ ∥ C 0 , γ , ∥ R D ( a ) [ φ ] − R D ( b ) [ φ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ φ ∥ C 0 , γ . (292) \begin{aligned}
\|\mathcal{R}_{D}[\varphi]\|_{C^{1,\gamma}}&\leq c(\kappa_{*,\gamma+},c_{\Gamma})\epsilon^{-\gamma+}\|\varphi\|_{C^{0,\gamma}},\\
\|\mathcal{R}_{D}^{(a)}[\varphi]-\mathcal{R}_{D}^{(b)}[\varphi]\|_{C^{1,\gamma}}&\leq c(\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\epsilon^{-\gamma+}\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma+}}\|\varphi\|_{C^{0,\gamma}}. \tag*{(292)}
\end{aligned} ∥ R D [ φ ] ∥ C 1 , γ ∥ R D ( a ) [ φ ] − R D ( b ) [ φ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − γ + ∥ φ ∥ C 0 , γ , ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − γ + ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ φ ∥ C 0 , γ . (292)
To estimate D ‾ \overline{\mathcal{D}} D , we use the form (38) of the kernel K D K_{D} K D and the form (212) of R ‾ ⋅ n ′ \overline{\mathbf{R}}\cdot\mathbf{n}' R ⋅ n ′ to write
D ‾ [ Φ − 1 φ ] = 3 4 π ∫ − 1 / 2 1 / 2 ∫ − π π R ‾ ( R ‾ ⋅ n ′ ) R ‾ ⋅ ( Φ − 1 φ ) ∣ R ‾ ∣ 5 ϵ d θ ^ d s ^ = − 3 2 π ∫ − 1 / 2 1 / 2 ∫ − π π ϵ sin 2 ( θ ^ 2 ) R ‾ R ‾ ⋅ ( Φ − 1 φ ) ∣ R ‾ ∣ 5 ϵ d θ ^ d s ^ . \begin{aligned}
\overline{\mathcal{D}}[\Phi^{-1}\varphi]
&=\frac{3}{4\pi}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\overline{\mathbf{R}}(\overline{\mathbf{R}}\cdot\mathbf{n}')\overline{\mathbf{R}}\cdot(\Phi^{-1}\varphi)}{|\overline{\mathbf{R}}|^{5}}\epsilon\,d\widehat{\theta}\,d\widehat{s}\\
&=-\frac{3}{2\pi}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\overline{\mathbf{R}}\overline{\mathbf{R}}\cdot(\Phi^{-1}\varphi)}{|\overline{\mathbf{R}}|^{5}}\epsilon\,d\widehat{\theta}\,d\widehat{s}.
\end{aligned} D [ Φ − 1 φ ] = 4 π 3 ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 5 R ( R ⋅ n ′ ) R ⋅ ( Φ − 1 φ ) ϵ d θ d s = − 2 π 3 ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 5 ϵ sin 2 ( 2 θ ) R R ⋅ ( Φ − 1 φ ) ϵ d θ d s .
Then, by Lemma 4.2, we have
∥ D ‾ [ Φ − 1 φ ] ∥ ≤ c ∥ φ ∥ L ∞ , \|\overline{\mathcal{D}}[\Phi^{-1}\varphi]\|\leq c\|\varphi\|_{L^{\infty}}, ∥ D [ Φ − 1 φ ] ∥ ≤ c ∥ φ ∥ L ∞ ,
∥ D ‾ [ ( Φ ( a ) ) − 1 φ − ( Φ ( b ) ) − 1 φ ] ∥ ≤ c ∥ X ( a ) − X ( b ) ∥ C 1 ∥ φ ∥ L ∞ . (293) \|\overline{\mathcal{D}}[(\Phi^{(a)})^{-1}\varphi-(\Phi^{(b)})^{-1}\varphi]\|\leq c\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{1}}\|\varphi\|_{L^{\infty}}. \tag*{(293)} ∥ D [( Φ ( a ) ) − 1 φ − ( Φ ( b ) ) − 1 φ ] ∥ ≤ c ∥ X ( a ) − X ( b ) ∥ C 1 ∥ φ ∥ L ∞ . (293)
Next, taking θ \theta θ -derivatives along Γ ϵ \Gamma_{\epsilon} Γ ϵ and using the expansions (196), (212) , (212), (249), we may write
1 ϵ ∂ θ D ‾ [ Φ − 1 φ ] = 3 4 π p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( R ‾ ( R ‾ ⋅ n ′ ) ( e θ ⋅ φ ) ∣ R ‾ ∣ 5 + R ‾ ( R ‾ ⋅ ( Φ − 1 φ ) ) ( 1 ϵ ∂ θ R ‾ ⋅ n ′ ) ∣ R ‾ ∣ 5 + 5 R ‾ ( R ‾ ⋅ ( Φ − 1 φ ) ) ( R ‾ ⋅ n ′ ) ( 1 ϵ ∂ θ R ‾ ⋅ R ‾ ) ∣ R ‾ ∣ 7 ) ϵ d θ ^ d s ^ = 3 4 π p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( − 2 ϵ sin 2 ( θ ^ 2 ) ( s ^ e z + 2 ϵ sin ( θ ^ 2 ) e θ ) ( e θ ⋅ φ ) ∣ R ‾ ∣ 5 − sin ( θ ^ ) ( s ^ e z + 2 ϵ sin ( θ ^ 2 ) e θ ) ( s ^ e t ⋅ φ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ φ ) ∣ R ‾ ∣ 5 − 10 ϵ 2 sin 2 ( θ ^ 2 ) sin ( θ ^ ) ( s ^ e z + 2 ϵ sin ( θ ^ 2 ) e θ ) ( s ^ e t ⋅ φ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ φ ) ∣ R ‾ ∣ 7 ) ϵ d θ ^ d s ^ . \begin{aligned}
\frac{1}{\epsilon}\partial_{\theta}\overline{\mathcal{D}}[\Phi^{-1}\varphi]
={}&\frac{3}{4\pi}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{\overline{\mathbf{R}}(\overline{\mathbf{R}}\cdot\mathbf{n}')(e_{\theta}\cdot\varphi)}{|\overline{\mathbf{R}}|^{5}}+\frac{\overline{\mathbf{R}}(\overline{\mathbf{R}}\cdot(\Phi^{-1}\varphi))(\frac{1}{\epsilon}\partial_{\theta}\overline{\mathbf{R}}\cdot\mathbf{n}')}{|\overline{\mathbf{R}}|^{5}}\right.\\
&\left.\qquad+\frac{5\overline{\mathbf{R}}(\overline{\mathbf{R}}\cdot(\Phi^{-1}\varphi))(\overline{\mathbf{R}}\cdot\mathbf{n}')(\frac{1}{\epsilon}\partial_{\theta}\overline{\mathbf{R}}\cdot\overline{\mathbf{R}})}{|\overline{\mathbf{R}}|^{7}}\right)\epsilon\,d\widehat{\theta}\,d\widehat{s}\\
={}&\frac{3}{4\pi}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{-2\epsilon\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)(\widehat{s}\mathbf{e}_{z}+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\mathbf{e}_{\theta})(e_{\theta}\cdot\varphi)}{|\overline{\mathbf{R}}|^{5}}\right.\\
&\qquad-\frac{\sin(\widehat{\theta})(\widehat{s}\mathbf{e}_{z}+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\mathbf{e}_{\theta})(\widehat{s}\mathbf{e}_{t}\cdot\varphi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\varphi)}{|\overline{\mathbf{R}}|^{5}}\\
&\left.\qquad-\frac{10\epsilon^{2}\sin^{2}\left(\frac{\widehat{\theta}}{2}\right)\sin(\widehat{\theta})(\widehat{s}\mathbf{e}_{z}+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\mathbf{e}_{\theta})(\widehat{s}\mathbf{e}_{t}\cdot\varphi+2\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)e_{\theta}\cdot\varphi)}{|\overline{\mathbf{R}}|^{7}}\right)\epsilon\,d\widehat{\theta}\,d\widehat{s}.
\end{aligned} ϵ 1 ∂ θ D [ Φ − 1 φ ] = = 4 π 3 p.v. ∫ − 1/2 1/2 ∫ − π π ( ∣ R ∣ 5 R ( R ⋅ n ′ ) ( e θ ⋅ φ ) + ∣ R ∣ 5 R ( R ⋅ ( Φ − 1 φ )) ( ϵ 1 ∂ θ R ⋅ n ′ ) + ∣ R ∣ 7 5 R ( R ⋅ ( Φ − 1 φ )) ( R ⋅ n ′ ) ( ϵ 1 ∂ θ R ⋅ R ) ) ϵ d θ d s 4 π 3 p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 5 − 2 ϵ sin 2 ( 2 θ ) ( s e z + 2 ϵ sin ( 2 θ ) e θ ) ( e θ ⋅ φ ) − ∣ R ∣ 5 sin ( θ ) ( s e z + 2 ϵ sin ( 2 θ ) e θ ) ( s e t ⋅ φ + 2 ϵ sin ( 2 θ ) e θ ⋅ φ ) − ∣ R ∣ 7 10 ϵ 2 sin 2 ( 2 θ ) sin ( θ ) ( s e z + 2 ϵ sin ( 2 θ ) e θ ) ( s e t ⋅ φ + 2 ϵ sin ( 2 θ ) e θ ⋅ φ ) ϵ d θ d s .
Using Lemma 4.4, case (2), we then obtain
∥ 1 ϵ ∂ θ D ‾ [ Φ − 1 φ ] ∥ C ˙ 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ , \left\|\frac{1}{\epsilon}\partial_{\theta}\overline{\mathcal{D}}[\Phi^{-1}\varphi]\right\|_{\dot{C}^{0,\gamma}}\leq c(\kappa_{*,\gamma},c_{\Gamma})\epsilon^{-1-\gamma}\|\varphi\|_{C^{0,\gamma}}, ϵ 1 ∂ θ D [ Φ − 1 φ ] C ˙ 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ ,
∥ 1 ϵ ∂ θ D ‾ [ ( Φ ( a ) ) − 1 φ − ( Φ ( b ) ) − 1 φ ] ∥ C ˙ 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ φ ∥ C 0 , γ . (294) \left\|\frac{1}{\epsilon}\partial_{\theta}\overline{\mathcal{D}}[(\Phi^{(a)})^{-1}\varphi-(\Phi^{(b)})^{-1}\varphi]\right\|_{\dot{C}^{0,\gamma}}\leq c(\kappa_{*,\gamma}^{(a)},\kappa_{*,\gamma}^{(b)},c_{\Gamma})\epsilon^{-1-\gamma}\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma}}\|\varphi\|_{C^{0,\gamma}}. \tag*{(294)} ϵ 1 ∂ θ D [( Φ ( a ) ) − 1 φ − ( Φ ( b ) ) − 1 φ ] C ˙ 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ φ ∥ C 0 , γ . (294)
Here the additional ϵ − γ \epsilon^{-\gamma} ϵ − γ arises due to the dependence on the curved frame vector e θ ( s , θ − θ ^ 2 ) e_{\theta}(s,\theta-\frac{\widehat{\theta}}{2}) e θ ( s , θ − 2 θ ) in the above expansion. Taking s s s -derivatives along Γ ϵ \Gamma_{\epsilon} Γ ϵ , we next have
∂ s D ‾ [ Φ − 1 φ ] = 3 4 π p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( R ‾ ( R ‾ ⋅ n ′ ) ( e t ⋅ φ ) ∣ R ‾ ∣ 5 + R ‾ ( R ‾ ⋅ ( Φ − 1 φ ) ) ( ∂ s R ‾ ⋅ n ′ ) ∣ R ‾ ∣ 5 + 5 R ‾ ( R ‾ ⋅ ( Φ − 1 φ ) ) ( R ‾ ⋅ n ′ ) ( ∂ s R ‾ ⋅ R ‾ ) ∣ R ‾ ∣ 7 ) ϵ d θ ^ d s ^ = 3 4 π p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( − 2 ϵ sin 2 ( θ ^ 2 ) ( s ^ e z + 2 ϵ sin ( θ ^ 2 ) e θ ) ( e t ⋅ φ ) ∣ R ‾ ∣ 5 − 10 ϵ sin 2 ( θ ^ 2 ) ( s ^ + 2 ϵ sin ( θ ^ 2 ) sin ( θ ^ ) ) ( s ^ e z + 2 ϵ sin ( θ ^ 2 ) e θ ) ( s ^ e t ⋅ φ + 2 ϵ sin ( θ ^ 2 ) e θ ⋅ φ ) ∣ R ‾ ∣ 7 ) ϵ d θ ^ d s ^ . \begin{aligned}
\partial_{s}\overline{\mathcal{D}}[\Phi^{-1}\varphi]=\frac{3}{4\pi}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{\overline{\mathbf{R}}(\overline{\mathbf{R}}\cdot\mathbf{n}')(e_{t}\cdot\varphi)}{|\overline{\mathbf{R}}|^{5}}+\frac{\overline{\mathbf{R}}(\overline{\mathbf{R}}\cdot(\Phi^{-1}\varphi))(\partial_{s}\overline{\mathbf{R}}\cdot\mathbf{n}')}{|\overline{\mathbf{R}}|^{5}}+\frac{5\overline{\mathbf{R}}(\overline{\mathbf{R}}\cdot(\Phi^{-1}\varphi))(\overline{\mathbf{R}}\cdot\mathbf{n}')(\partial_{s}\overline{\mathbf{R}}\cdot\overline{\mathbf{R}})}{|\overline{\mathbf{R}}|^{7}}\right)\epsilon\,d\widehat{\theta}\,d\widehat{s} \\
&= \frac{3}{4\pi}\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{-2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)(\hat{s}e_z+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta)(e_t\cdot\varphi)}{|\overline{R}|^5}
-\frac{10\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)(\hat{s}+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\sin(\hat{\theta}))(\hat{s}e_z+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta)(\hat{s}e_t\cdot\varphi+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta\cdot\varphi)}{|\overline{R}|^7}\right)\epsilon\,d\hat{\theta}d\hat{s}.
\end{aligned} ∂ s D [ Φ − 1 φ ] = 4 π 3 p.v. ∫ − 1/2 1/2 ∫ − π π ( ∣ R ∣ 5 R ( R ⋅ n ′ ) ( e t ⋅ φ ) + ∣ R ∣ 5 R ( R ⋅ ( Φ − 1 φ )) ( ∂ s R ⋅ n ′ ) + ∣ R ∣ 7 5 R ( R ⋅ ( Φ − 1 φ )) ( R ⋅ n ′ ) ( ∂ s R ⋅ R ) ) ϵ d θ d s = 4 π 3 p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 5 − 2 ϵ sin 2 ( 2 θ ^ ) ( s ^ e z + 2 ϵ sin ( 2 θ ^ ) e θ ) ( e t ⋅ φ ) − ∣ R ∣ 7 10 ϵ sin 2 ( 2 θ ^ ) ( s ^ + 2 ϵ sin ( 2 θ ^ ) sin ( θ ^ )) ( s ^ e z + 2 ϵ sin ( 2 θ ^ ) e θ ) ( s ^ e t ⋅ φ + 2 ϵ sin ( 2 θ ^ ) e θ ⋅ φ ) ϵ d θ ^ d s ^ .
Again using Lemma 4.4 , case (2), we have
∥ ∂ s ^ D ‾ [ Φ − 1 φ ] ∥ C ˙ 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ \left\|\partial_{\hat{s}}\overline{D}[\Phi^{-1}\varphi]\right\|_{\dot{C}^{0,\gamma}}\leq c(\kappa_{*},\gamma,c_{\Gamma})\epsilon^{-1-\gamma}\|\varphi\|_{C^{0,\gamma}} ∂ s ^ D [ Φ − 1 φ ] C ˙ 0 , γ ≤ c ( κ ∗ , γ , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ
∥ ∂ s ^ D ‾ [ ( Φ ( a ) ) − 1 φ − ( Φ ( b ) ) − 1 φ ] ∥ C ˙ 0 , γ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , γ , c Γ ) ϵ − 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ φ ∥ C 0 , γ . (295) \left\|\partial_{\hat{s}}\overline{D}\left[(\Phi^{(a)})^{-1}\varphi-(\Phi^{(b)})^{-1}\varphi\right]\right\|_{\dot{C}^{0,\gamma}}\leq c(\kappa_{*}^{(a)},\kappa_{*}^{(b)},\gamma,c_{\Gamma})\epsilon^{-1-\gamma}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma}}\|\varphi\|_{C^{0,\gamma}} . \tag*{(295)} ∂ s ^ D [ ( Φ ( a ) ) − 1 φ − ( Φ ( b ) ) − 1 φ ] C ˙ 0 , γ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , γ , c Γ ) ϵ − 1 − γ X ( a ) − X ( b ) C 2 , γ ∥ φ ∥ C 0 , γ . (295)
Here again we lose a factor of ϵ − γ \epsilon^{-\gamma} ϵ − γ from the single e θ ( s , θ − θ ^ 2 ) e_\theta(s,\theta-\frac{\hat{\theta}}{2}) e θ ( s , θ − 2 θ ^ ) term above. Combining (292) , (293) , (294) , and (295) , we obtain
∥ D ‾ [ φ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ \left\|\overline{D}[\varphi]\right\|_{C^{1,\gamma}}\leq c(\kappa_{*,\gamma+},c_{\Gamma})\epsilon^{-1-\gamma}\|\varphi\|_{C^{0,\gamma}} D [ φ ] C 1 , γ ≤ c ( κ ∗ , γ + , c Γ ) ϵ − 1 − γ ∥ φ ∥ C 0 , γ
∥ D ‾ ( a ) [ φ ] − D ‾ ( b ) [ φ ] ∥ C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ φ ∥ C 0 , γ . (296) \left\|\overline{D}^{(a)}[\varphi]-\overline{D}^{(b)}[\varphi]\right\|_{C^{1,\gamma}}\leq c(\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\epsilon^{-1-\gamma}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\gamma+}}\|\varphi\|_{C^{0,\gamma}} . \tag*{(296)} D ( a ) [ φ ] − D ( b ) [ φ ] C 1 , γ ≤ c ( κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ϵ − 1 − γ X ( a ) − X ( b ) C 2 , γ + ∥ φ ∥ C 0 , γ . (296)
It remains to estimate the completion flow V [ φ ] V[\varphi] V [ φ ] , given by (276) . For x ∈ Γ ϵ x\in\Gamma_\epsilon x ∈ Γ ϵ , it will be convenient to rewrite R X = x ( s , θ ) − X ( s ′ ) R_X=x(s,\theta)-X(s') R X = x ( s , θ ) − X ( s ′ ) in terms of s ^ = s − s ′ \hat{s}=s-s' s ^ = s − s ′ and s , θ s,\theta s , θ . Using the expansions (198) , we may write R X R_X R X as
R X ( s , s ^ , θ ) = s ^ e t ( s ) − s ^ 2 Q t ( s , s ^ ) + ϵ e r ( s , θ ) . R_X(s,\hat{s},\theta)=\hat{s}e_t(s)-\hat{s}^{2}Q_t(s,\hat{s})+\epsilon e_r(s,\theta). R X ( s , s ^ , θ ) = s ^ e t ( s ) − s ^ 2 Q t ( s , s ^ ) + ϵ e r ( s , θ ) .
For two filaments X ( a ) X^{(a)} X ( a ) , X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1 , we note the expansions
R X ( a ) − R X ( b ) = s ^ ( e t ( a ) − e t ( b ) − s ^ ( Q t ( a ) − Q t ( b ) ) ) + ϵ ( e r ( a ) − e r ( b ) ) R_X^{(a)}-R_X^{(b)}=\hat{s}(e_t^{(a)}-e_t^{(b)}-\hat{s}(Q_t^{(a)}-Q_t^{(b)}))+\epsilon(e_r^{(a)}-e_r^{(b)}) R X ( a ) − R X ( b ) = s ^ ( e t ( a ) − e t ( b ) − s ^ ( Q t ( a ) − Q t ( b ) )) + ϵ ( e r ( a ) − e r ( b ) )
1 ∣ R X ( a ) ∣ k − 1 ∣ R X ( b ) ∣ k = ( − 2 s ^ ( Q t ( b ) ⋅ e t ( b ) − Q t ( a ) ⋅ e t ( a ) ) − 2 ϵ ( Q t ( b ) ⋅ e r ( b ) − Q t ( a ) ⋅ e r ( a ) ) + s ^ 2 ( ∣ Q t ( b ) ∣ 2 − ∣ Q t ( a ) ∣ 2 ) ) s ^ 2 ∣ R X ( a ) ∣ ∣ R X ( b ) ∣ ( ∣ R X ( a ) ∣ + ∣ R X ( b ) ∣ ) ∑ ℓ = 0 k − 1 1 ∣ R X ( a ) ∣ ℓ ∣ R X ( b ) ∣ k − 1 − ℓ . \frac{1}{|R_X^{(a)}|^k}-\frac{1}{|R_X^{(b)}|^k}=\left(-2\hat{s}(Q_t^{(b)}\cdot e_t^{(b)}-Q_t^{(a)}\cdot e_t^{(a)})-2\epsilon(Q_t^{(b)}\cdot e_r^{(b)}-Q_t^{(a)}\cdot e_r^{(a)})+\hat{s}^{2}(|Q_t^{(b)}|^2-|Q_t^{(a)}|^2)\right)\frac{\hat{s}^{2}}{|R_X^{(a)}||R_X^{(b)}|(|R_X^{(a)}|+|R_X^{(b)}|)}\sum_{\ell=0}^{k-1}\frac{1}{|R_X^{(a)}|^\ell|R_X^{(b)}|^{k-1-\ell}} . ∣ R X ( a ) ∣ k 1 − ∣ R X ( b ) ∣ k 1 = ( − 2 s ^ ( Q t ( b ) ⋅ e t ( b ) − Q t ( a ) ⋅ e t ( a ) ) − 2 ϵ ( Q t ( b ) ⋅ e r ( b ) − Q t ( a ) ⋅ e r ( a ) ) + s ^ 2 ( ∣ Q t ( b ) ∣ 2 − ∣ Q t ( a ) ∣ 2 ) ) ∣ R X ( a ) ∣∣ R X ( b ) ∣ ( ∣ R X ( a ) ∣ + ∣ R X ( b ) ∣ ) s ^ 2 ℓ = 0 ∑ k − 1 ∣ R X ( a ) ∣ ℓ ∣ R X ( b ) ∣ k − 1 − ℓ 1 .
From [52 Lemma 3.1] , we have
∣ R X ∣ ≥ c ( κ ∗ , c Γ ) s ^ 2 + ϵ 2 ; (297) |R_X|\geq c(\kappa_{*},c_{\Gamma})\sqrt{\hat{s}^{2}+\epsilon^{2}}; \tag*{(297)} ∣ R X ∣ ≥ c ( κ ∗ , c Γ ) s ^ 2 + ϵ 2 ; (297)
in particular, ∣ s ^ ∣ R X ∣ ∣ ≤ c ( κ ∗ , c Γ ) \left|\frac{\hat{s}}{|R_X|}\right|\leq c(\kappa_{*},c_{\Gamma}) ∣ R X ∣ s ^ ≤ c ( κ ∗ , c Γ ) . Then, using the forms (37) , (277) of G \mathcal{G} G and l R l_R l R as well as the fact that ∣ R X ∣ ≥ ϵ |R_X|\geq\epsilon ∣ R X ∣ ≥ ϵ while ∣ x ′ − X ( s ′ ) ∣ = ϵ |x'-X(s')|=\epsilon ∣ x ′ − X ( s ′ ) ∣ = ϵ , we have
∣ V [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ Γ ϵ 1 ϵ d S x ′ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ , |V[\varphi]|\leq c(\kappa_{*})\|\varphi\|_{L^\infty}\int_{\Gamma_\epsilon}\frac{1}{\epsilon}\,dS_{x'}\leq c(\kappa_{*})\|\varphi\|_{L^\infty}, ∣ V [ φ ] ∣ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ Γ ϵ ϵ 1 d S x ′ ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ,
∣ V ( a ) [ φ ] − V ( b ) [ φ ] ∣ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ . |V^{(a)}[\varphi]-V^{(b)}[\varphi]|\leq c(\kappa_{*}^{(a)},\kappa_{*}^{(b)},c_{\Gamma})\left\|X^{(a)}-X^{(b)}\right\|_{C^2}\|\varphi\|_{L^\infty}. ∣ V ( a ) [ φ ] − V ( b ) [ φ ] ∣ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) X ( a ) − X ( b ) C 2 ∥ φ ∥ L ∞ .
Here we recall that d S x ′ = J ϵ ( s ′ , θ ′ ) d θ ′ d s ′ dS_{x'}=\mathcal{J}_\epsilon(s',\theta')\,d\theta'ds' d S x ′ = J ϵ ( s ′ , θ ′ ) d θ ′ d s ′ .
We next take derivatives of V [ φ ] V[\varphi] V [ φ ] along Γ ϵ \Gamma_\epsilon Γ ϵ . Here we use ∂ j \partial_j ∂ j to denote ∂ s \partial_s ∂ s or 1 ϵ ∂ θ \frac{1}{\epsilon}\partial_\theta ϵ 1 ∂ θ . We have
∂ j V [ φ ] = ∫ Γ ϵ ∂ j G ( x , X ( s ′ ) ) φ ( x ′ ) d S x ′ + ∫ Γ ϵ ∂ j l R ( x , X ( s ′ ) ) × ( ( x ′ − X ( s ′ ) ) × φ ( x ′ ) ) d S x ′ . \partial_jV[\varphi]=\int_{\Gamma_\epsilon}\partial_j\mathcal{G}(x,X(s'))\varphi(x')\,dS_{x'}+\int_{\Gamma_\epsilon}\partial_jl_R(x,X(s'))\times((x'-X(s'))\times\varphi(x'))\,dS_{x'}. ∂ j V [ φ ] = ∫ Γ ϵ ∂ j G ( x , X ( s ′ )) φ ( x ′ ) d S x ′ + ∫ Γ ϵ ∂ j l R ( x , X ( s ′ )) × (( x ′ − X ( s ′ )) × φ ( x ′ )) d S x ′ .
From the forms (37) , (277) of G \mathcal{G} G and l R l_R l R , we may calculate
∂ j G ( x , X ( s ′ ) ) = − ( ∂ j R X ⋅ R X ) I ∣ R X ∣ 3 + ∂ j R X ⊗ R X + R X ⊗ ∂ j R X ∣ R X ∣ 3 − 3 ( ∂ j R X ⋅ R X ) R X ⊗ R X ∣ R X ∣ 5 ; \partial_j\mathcal{G}(x,X(s'))=-\frac{(\partial_jR_X\cdot R_X)I}{|R_X|^3}+\frac{\partial_jR_X\otimes R_X+R_X\otimes\partial_jR_X}{|R_X|^3}-3\frac{(\partial_jR_X\cdot R_X)R_X\otimes R_X}{|R_X|^5}; ∂ j G ( x , X ( s ′ )) = − ∣ R X ∣ 3 ( ∂ j R X ⋅ R X ) I + ∣ R X ∣ 3 ∂ j R X ⊗ R X + R X ⊗ ∂ j R X − 3 ∣ R X ∣ 5 ( ∂ j R X ⋅ R X ) R X ⊗ R X ;
e n d s m i d = 1 {ends_mid=1} e n d s m i d = 1
∂ j l R ( x , X ( s ′ ) ) = − ∂ j R X ∣ R X ∣ 3 + 3 ( ∂ j R X ⋅ R X ) R X ∣ R X ∣ 5 \partial_j l_R(x,X(s'))=-\frac{\partial_j R_X}{|R_X|^3}+3\frac{(\partial_j R_X\cdot R_X)R_X}{|R_X|^5} ∂ j l R ( x , X ( s ′ )) = − ∣ R X ∣ 3 ∂ j R X + 3 ∣ R X ∣ 5 ( ∂ j R X ⋅ R X ) R X
∂ s R X = ( 1 − ϵ κ b ( s , θ ) ) e t ( s ) + ϵ κ 3 e θ ( s , θ ) , 1 ϵ ∂ θ R X = e θ ( s , θ ) . \partial_s R_X=(1-\epsilon\kappa b(s,\theta))e_t(s)+\epsilon\kappa_3e_\theta(s,\theta),\qquad\frac{1}{\epsilon}\partial_\theta R_X=e_\theta(s,\theta). ∂ s R X = ( 1 − ϵ κb ( s , θ )) e t ( s ) + ϵ κ 3 e θ ( s , θ ) , ϵ 1 ∂ θ R X = e θ ( s , θ ) .
Using the above expressions as well as the form (45) of the C ˙ 0 , γ \dot{C}^{0,\gamma} C ˙ 0 , γ seminorm, we obtain the bounds
∣ ∂ j V [ φ ] ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ ) ∥ φ ∥ L ∞ ( ∫ Γ ϵ 1 ϵ 2 + γ d S x ′ + ∫ Γ ϵ 1 ϵ 3 + γ ϵ d S x ′ ) ≤ c ( κ ∗ , γ ) ϵ − 1 − γ ∥ φ ∥ L ∞ , |\partial_j V[\varphi]|_{\dot{C}^{0,\gamma}}\le c(\kappa_{*,\gamma})\|\varphi\|_{L^\infty}\left(\int_{\Gamma_\epsilon}\frac{1}{\epsilon^{2+\gamma}}\,dS_{x'}+\int_{\Gamma_\epsilon}\frac{1}{\epsilon^{3+\gamma}}\epsilon\,dS_{x'}\right)\le c(\kappa_{*,\gamma})\epsilon^{-1-\gamma}\|\varphi\|_{L^\infty}, ∣ ∂ j V [ φ ] ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ ) ∥ φ ∥ L ∞ ( ∫ Γ ϵ ϵ 2 + γ 1 d S x ′ + ∫ Γ ϵ ϵ 3 + γ 1 ϵ d S x ′ ) ≤ c ( κ ∗ , γ ) ϵ − 1 − γ ∥ φ ∥ L ∞ ,
∣ ∂ j V ( a ) [ φ ] − ∂ j V ( b ) [ φ ] ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ φ ∥ L ∞ . |\partial_j V^{(a)}[\varphi]-\partial_j V^{(b)}[\varphi]|_{\dot{C}^{0,\gamma}}\le c(\kappa_{*,\gamma}^{(a)},\kappa_{*,\gamma}^{(b)},c_\Gamma)\epsilon^{-1-\gamma}\|\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\|_{C^{2,\gamma}}\|\varphi\|_{L^\infty}. ∣ ∂ j V ( a ) [ φ ] − ∂ j V ( b ) [ φ ] ∣ C ˙ 0 , γ ≤ c ( κ ∗ , γ ( a ) , κ ∗ , γ ( b ) , c Γ ) ϵ − 1 − γ ∥ X ( a ) − X ( b ) ∥ C 2 , γ ∥ φ ∥ L ∞ .
Combining the estimates for V V V with the bounds for D \mathcal{D} D , we obtain Lemma 5.2. □
Proof of Lemma 5.3: inverse double layer
The proof of the bound (285) relies on a compactness argument and is nearly identical to the Laplace setting [59 Section 4.2] .
Given fixed constants κ ∗ , γ + \kappa_{*,\gamma^+} κ ∗ , γ + and c Γ c_\Gamma c Γ , we begin by fixing a nonzero function φ = φ ( s , θ ) \varphi=\varphi(s,\theta) φ = φ ( s , θ ) in C 0 , γ ( Γ ϵ ) C^{0,\gamma}(\Gamma_\epsilon) C 0 , γ ( Γ ϵ ) , where the space C 0 , γ ( Γ ϵ ) C^{0,\gamma}(\Gamma_\epsilon) C 0 , γ ( Γ ϵ ) is understood with respect to the norm (45); in particular, without reference to the filament centerline shape. We will show that
∥ ( 1 2 I + D + V ) − 1 [ φ ] ∥ C 0 , γ ≤ c φ ( κ ∗ , γ + , c Γ ) ∥ φ ∥ C 0 , γ (298) \left\|\left(\frac{1}{2}\mathcal{I}+\mathcal{D}+\mathcal{V}\right)^{-1}[\varphi]\right\|_{C^{0,\gamma}}\le c_\varphi(\kappa_{*,\gamma^+},c_\Gamma)\|\varphi\|_{C^{0,\gamma}} \tag*{(298)} ( 2 1 I + D + V ) − 1 [ φ ] C 0 , γ ≤ c φ ( κ ∗ , γ + , c Γ ) ∥ φ ∥ C 0 , γ (298)
for any filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline curve X ( s ) \mathbf{X}(s) X ( s ) satisfying ∥ X s s ∥ C 0 , γ + ≤ κ ∗ , γ + \|\mathbf{X}_{ss}\|_{C^{0,\gamma^+}}\le\kappa_{*,\gamma^+} ∥ X ss ∥ C 0 , γ + ≤ κ ∗ , γ + and the non-self-intersection condition (2) with the fixed choice of c Γ c_\Gamma c Γ . Since φ \varphi φ is arbitrary, the uniform boundedness principle then implies that the bound (297) holds independent of φ \varphi φ .
To show (297), we fix ∥ φ ∥ C 0 , γ = 1 \|\varphi\|_{C^{0,\gamma}}=1 ∥ φ ∥ C 0 , γ = 1 . Proceeding by contradiction, if the bound (297) does not hold, then there exists a sequence of curves X j \mathbf{X}_j X j , j ∈ N j\in\mathbb{N} j ∈ N , satisfying (2) and ∥ ( X j ) s s ∥ C 0 , γ + ≤ κ ∗ , γ + \|(\mathbf{X}_j)_{ss}\|_{C^{0,\gamma^+}}\le\kappa_{*,\gamma^+} ∥ ( X j ) ss ∥ C 0 , γ + ≤ κ ∗ , γ + such that
∥ ( 1 2 I + D j + V j ) − 1 [ φ ] ∥ C 0 , γ > j . \left\|\left(\frac{1}{2}\mathcal{I}+\mathcal{D}_j+\mathcal{V}_j\right)^{-1}[\varphi]\right\|_{C^{0,\gamma}}>j. ( 2 1 I + D j + V j ) − 1 [ φ ] C 0 , γ > j .
Here D j + V j \mathcal{D}_j+\mathcal{V}_j D j + V j denotes the double layer operator and completion flow corresponding to curve j j j . We define
h j = φ ∥ ( 1 2 I + D j + V j ) − 1 [ φ ] ∥ C 0 , γ , g j = ( 1 2 I + D j + V j ) − 1 [ φ ] ∥ ( 1 2 I + D j + V j ) − 1 [ φ ] ∥ C 0 , γ . h_j=\frac{\varphi}{\left\|\left(\frac{1}{2}\mathcal{I}+\mathcal{D}_j+\mathcal{V}_j\right)^{-1}[\varphi]\right\|_{C^{0,\gamma}}},\qquad
g_j=\frac{\left(\frac{1}{2}\mathcal{I}+\mathcal{D}_j+\mathcal{V}_j\right)^{-1}[\varphi]}{\left\|\left(\frac{1}{2}\mathcal{I}+\mathcal{D}_j+\mathcal{V}_j\right)^{-1}[\varphi]\right\|_{C^{0,\gamma}}}. h j = ( 2 1 I + D j + V j ) − 1 [ φ ] C 0 , γ φ , g j = ( 2 1 I + D j + V j ) − 1 [ φ ] C 0 , γ ( 2 1 I + D j + V j ) − 1 [ φ ] .
Since ∥ g j ∥ C 0 , γ = 1 \|g_j\|_{C^{0,\gamma}}=1 ∥ g j ∥ C 0 , γ = 1 , we then have
( 1 2 I + D j + V j ) [ g j ] = h j → 0 in C 0 , γ . (299) \left(\frac{1}{2}\mathcal{I}+\mathcal{D}_j+\mathcal{V}_j\right)[g_j]=h_j\to0\quad\text{in }C^{0,\gamma}. \tag*{(299)} ( 2 1 I + D j + V j ) [ g j ] = h j → 0 in C 0 , γ . (299)
Using Lemma 5.2, we have that ( D j + V j ) [ g j ] (\mathcal{D}_j+\mathcal{V}_j)[g_j] ( D j + V j ) [ g j ] is smoother, in fact, bounded in C 1 , γ C^{1,\gamma} C 1 , γ uniformly in j j j . Thus (298) implies that along a subsequence j ℓ j_\ell j ℓ ,
g j ℓ − 2 h j ℓ → g ∞ strongly in C 0 , γ (300) g_{j_\ell}-2h_{j_\ell}\to g_\infty\qquad\text{strongly in }C^{0,\gamma} \tag*{(300)} g j ℓ − 2 h j ℓ → g ∞ strongly in C 0 , γ (300)
for some limit g ∞ g_\infty g ∞ with ∥ g ∞ ∥ C 0 , γ = 1 \|g_\infty\|_{C^{0,\gamma}}=1 ∥ g ∞ ∥ C 0 , γ = 1 .
Since ∥ ( X j ) s s ∥ C 0 , γ + ≤ κ ∗ , γ + \|(\mathbf{X}_j)_{ss}\|_{C^{0,\gamma^+}}\le\kappa_{*,\gamma^+} ∥ ( X j ) ss ∥ C 0 , γ + ≤ κ ∗ , γ + , there exists a subsequence of curves X j k \mathbf{X}_{j_k} X j k converging (after possible translation) strongly in C 2 , γ C^{2,\gamma} C 2 , γ , 0 < γ < γ + 0<\gamma<\gamma^+ 0 < γ < γ + , to a limit curve X ∞ ( s ) \mathbf{X}_\infty(s) X ∞ ( s ) satisfying (2). Using the form (39) of D \mathcal{D} D , for any ψ ∈ C 0 , γ ( Γ ϵ ) \psi\in C^{0,\gamma}(\Gamma_\epsilon) ψ ∈ C 0 , γ ( Γ ϵ ) , we have
( D j k − D ∞ ) [ ψ ] = 3 4 π ∫ − 1 / 2 1 / 2 ∫ − π π ( X ^ j k + ϵ e ^ r , j k ) ( X ^ j k + ϵ e ^ r , j k ) ⋅ e r , j k ( X ^ j k + ϵ e ^ r , j k ) ⋅ ψ ( s ′ , θ ′ ) ∣ X ^ j k + ϵ e ^ r , j k ∣ 5 J ϵ , j k d θ ′ d s ′ (\mathcal{D}_{j_k}-\mathcal{D}_\infty)[\psi]
=\frac{3}{4\pi}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
\frac{(\widehat{\mathbf{X}}_{j_k}+\epsilon\widehat{\mathbf{e}}_{r,j_k})(\widehat{\mathbf{X}}_{j_k}+\epsilon\widehat{\mathbf{e}}_{r,j_k})\cdot\mathbf{e}_{r,j_k}(\widehat{\mathbf{X}}_{j_k}+\epsilon\widehat{\mathbf{e}}_{r,j_k})\cdot\psi(s',\theta')}
{|\widehat{\mathbf{X}}_{j_k}+\epsilon\widehat{\mathbf{e}}_{r,j_k}|^5}\mathcal{J}_{\epsilon,j_k}\,d\theta'\,ds' ( D j k − D ∞ ) [ ψ ] = 4 π 3 ∫ − 1/2 1/2 ∫ − π π ∣ X j k + ϵ e r , j k ∣ 5 ( X j k + ϵ e r , j k ) ( X j k + ϵ e r , j k ) ⋅ e r , j k ( X j k + ϵ e r , j k ) ⋅ ψ ( s ′ , θ ′ ) J ϵ , j k d θ ′ d s ′
− 3 4 π ∫ − 1 / 2 1 / 2 ∫ − π π ( X ^ ∞ + ϵ e ^ r , ∞ ) ( X ^ ∞ + ϵ e ^ r , ∞ ) ⋅ e r , ∞ ( X ^ ∞ + ϵ e ^ r , ∞ ) ⋅ ψ ( s ′ , θ ′ ) ∣ X ^ ∞ + ϵ e ^ r , ∞ ∣ 5 J ϵ , ∞ d θ ′ d s ′ , -\frac{3}{4\pi}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{(\widehat{X}_{\infty}+\epsilon\widehat{e}_{r,\infty})(\widehat{X}_{\infty}+\epsilon\widehat{e}_{r,\infty})\cdot e_{r,\infty}(\widehat{X}_{\infty}+\epsilon\widehat{e}_{r,\infty})\cdot\psi(s',\theta')}{\lvert\widehat{X}_{\infty}+\epsilon\widehat{e}_{r,\infty}\rvert^{5}}\mathcal{J}_{\epsilon,\infty}\,d\theta'\,ds', − 4 π 3 ∫ − 1/2 1/2 ∫ − π π ∣ X ∞ + ϵ e r , ∞ ∣ 5 ( X ∞ + ϵ e r , ∞ ) ( X ∞ + ϵ e r , ∞ ) ⋅ e r , ∞ ( X ∞ + ϵ e r , ∞ ) ⋅ ψ ( s ′ , θ ′ ) J ϵ , ∞ d θ ′ d s ′ ,
where we denote X ^ j k ( s , s ′ ) = X j k ( s ) − X j k ( s ′ ) \widehat{X}_{j_k}(s,s')=X_{j_k}(s)-X_{j_k}(s') X j k ( s , s ′ ) = X j k ( s ) − X j k ( s ′ ) , e ^ r , j k ( s , θ , s ′ , θ ′ ) = e r , j k ( s , θ ) − e r , j k ( s ′ , θ ′ ) \widehat{e}_{r,j_k}(s,\theta,s',\theta')=e_{r,j_k}(s,\theta)-e_{r,j_k}(s',\theta') e r , j k ( s , θ , s ′ , θ ′ ) = e r , j k ( s , θ ) − e r , j k ( s ′ , θ ′ ) , and each of X ^ j \widehat{X}_j X j , e r , j e_{r,j} e r , j , and J ϵ , j \mathcal{J}_{\epsilon,j} J ϵ , j are defined along the curve X j X_j X j . From the form of these expressions, we can see that ∣ ( D j k − D ∞ ) [ ψ ] ∣ → 0 \lvert(D_{j_k}-D_{\infty})[\psi]\rvert\to0 ∣( D j k − D ∞ ) [ ψ ]∣ → 0 as j k → ∞ j_k\to\infty j k → ∞ , since X j k → X ∞ X_{j_k}\to X_{\infty} X j k → X ∞ . Similarly, using the form (275) of the completion flow V V V , we have
( V j k − V ∞ ) [ ψ ] = ∫ − 1 / 2 1 / 2 ∫ − π π ( J ϵ , j k ( s ′ , θ ′ ) ∣ X ^ j k + ϵ e r , j k ( s , θ ) ∣ − J ϵ , ∞ ( s ′ , θ ′ ) ∣ X ^ ∞ + ϵ e r , ∞ ( s , θ ) ∣ ) ψ ( s ′ , θ ′ ) d θ ′ d s ′ + ∫ − 1 / 2 1 / 2 ∫ − π π ( ( X ^ j k + ϵ e r , j k ) ⊗ ( X ^ j k + ϵ e r , j k ) J ϵ , j k ( s ′ , θ ′ ) ∣ X ^ j k + ϵ e r , j k ( s , θ ) ∣ 3 − ( X ^ ∞ + ϵ e r , ∞ ) ⊗ ( X ^ ∞ + ϵ e r , ∞ ) J ϵ , ∞ ( s ′ , θ ′ ) ∣ X ^ ∞ + ϵ e r , ∞ ( s , θ ) ∣ 3 ) ψ ( s ′ , θ ′ ) d θ ′ d s ′ − ∫ − 1 / 2 1 / 2 ∫ − π π X ^ j k + ϵ e r , j k ∣ X ^ j k + ϵ e r , j k ( s , θ ) ∣ 3 × ( e r , j k ( s ′ , θ ′ ) × ψ ( s ′ , θ ′ ) ) J ϵ , j k ( s ′ , θ ′ ) d θ ′ d s ′ + ∫ − 1 / 2 1 / 2 ∫ − π π X ^ ∞ + ϵ e r , ∞ ∣ X ^ ∞ + ϵ e r , ∞ ( s , θ ) ∣ 3 × ( e r , ∞ ( s ′ , θ ′ ) × ψ ( s ′ , θ ′ ) ) J ϵ , ∞ ( s ′ , θ ′ ) d θ ′ d s ′ , \begin{aligned}
(V_{j_k}-V_{\infty})[\psi]={}&\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{\mathcal{J}_{\epsilon,j_k}(s',\theta')}{\lvert\widehat{X}_{j_k}+\epsilon e_{r,j_k}(s,\theta)\rvert}-\frac{\mathcal{J}_{\epsilon,\infty}(s',\theta')}{\lvert\widehat{X}_{\infty}+\epsilon e_{r,\infty}(s,\theta)\rvert}\right)\psi(s',\theta')\,d\theta'\,ds'\\
&+\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(\frac{(\widehat{X}_{j_k}+\epsilon e_{r,j_k})\otimes(\widehat{X}_{j_k}+\epsilon e_{r,j_k})\mathcal{J}_{\epsilon,j_k}(s',\theta')}{\lvert\widehat{X}_{j_k}+\epsilon e_{r,j_k}(s,\theta)\rvert^{3}}\right.\\
&\qquad\left.-\frac{(\widehat{X}_{\infty}+\epsilon e_{r,\infty})\otimes(\widehat{X}_{\infty}+\epsilon e_{r,\infty})\mathcal{J}_{\epsilon,\infty}(s',\theta')}{\lvert\widehat{X}_{\infty}+\epsilon e_{r,\infty}(s,\theta)\rvert^{3}}\right)\psi(s',\theta')\,d\theta'\,ds'\\
&-\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\widehat{X}_{j_k}+\epsilon e_{r,j_k}}{\lvert\widehat{X}_{j_k}+\epsilon e_{r,j_k}(s,\theta)\rvert^{3}}\times(e_{r,j_k}(s',\theta')\times\psi(s',\theta'))\mathcal{J}_{\epsilon,j_k}(s',\theta')\,d\theta'\,ds'\\
&+\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{\widehat{X}_{\infty}+\epsilon e_{r,\infty}}{\lvert\widehat{X}_{\infty}+\epsilon e_{r,\infty}(s,\theta)\rvert^{3}}\times(e_{r,\infty}(s',\theta')\times\psi(s',\theta'))\mathcal{J}_{\epsilon,\infty}(s',\theta')\,d\theta'\,ds',
\end{aligned} ( V j k − V ∞ ) [ ψ ] = ∫ − 1/2 1/2 ∫ − π π ( ∣ X j k + ϵ e r , j k ( s , θ )∣ J ϵ , j k ( s ′ , θ ′ ) − ∣ X ∞ + ϵ e r , ∞ ( s , θ )∣ J ϵ , ∞ ( s ′ , θ ′ ) ) ψ ( s ′ , θ ′ ) d θ ′ d s ′ + ∫ − 1/2 1/2 ∫ − π π ( ∣ X j k + ϵ e r , j k ( s , θ ) ∣ 3 ( X j k + ϵ e r , j k ) ⊗ ( X j k + ϵ e r , j k ) J ϵ , j k ( s ′ , θ ′ ) − ∣ X ∞ + ϵ e r , ∞ ( s , θ ) ∣ 3 ( X ∞ + ϵ e r , ∞ ) ⊗ ( X ∞ + ϵ e r , ∞ ) J ϵ , ∞ ( s ′ , θ ′ ) ) ψ ( s ′ , θ ′ ) d θ ′ d s ′ − ∫ − 1/2 1/2 ∫ − π π ∣ X j k + ϵ e r , j k ( s , θ ) ∣ 3 X j k + ϵ e r , j k × ( e r , j k ( s ′ , θ ′ ) × ψ ( s ′ , θ ′ )) J ϵ , j k ( s ′ , θ ′ ) d θ ′ d s ′ + ∫ − 1/2 1/2 ∫ − π π ∣ X ∞ + ϵ e r , ∞ ( s , θ ) ∣ 3 X ∞ + ϵ e r , ∞ × ( e r , ∞ ( s ′ , θ ′ ) × ψ ( s ′ , θ ′ )) J ϵ , ∞ ( s ′ , θ ′ ) d θ ′ d s ′ ,
and again we see that ( V j k − V ∞ ) [ ψ ] → 0 (V_{j_k}-V_{\infty})[\psi]\to0 ( V j k − V ∞ ) [ ψ ] → 0 as j k → ∞ j_k\to\infty j k → ∞ . Together, we have ∣ ( D j k + V j k ) [ ψ ] − ( D ∞ + V ∞ ) [ ψ ] ∣ → 0 \lvert(D_{j_k}+V_{j_k})[\psi]-(D_{\infty}+V_{\infty})[\psi]\rvert\to0 ∣( D j k + V j k ) [ ψ ] − ( D ∞ + V ∞ ) [ ψ ]∣ → 0 . Therefore, by a diagonalization argument, we have
( D j + V j ) [ g j ] → ( D ∞ + V ∞ ) [ g ∞ ] (D_j+V_j)[g_j]\to(D_{\infty}+V_{\infty})[g_{\infty}] ( D j + V j ) [ g j ] → ( D ∞ + V ∞ ) [ g ∞ ]
along a subsequence, and, using (299), ( 1 2 I + D ∞ + V ∞ ) [ g ∞ ] = 0 (\frac{1}{2}I+D_{\infty}+V_{\infty})[g_{\infty}]=0 ( 2 1 I + D ∞ + V ∞ ) [ g ∞ ] = 0 . By injectivity of the completed double layer operator (see discussion below (274)), we then have g ∞ = 0 g_{\infty}=0 g ∞ = 0 , contradicting ∥ g ∞ ∥ C 0 , γ = 1 \lVert g_{\infty}\rVert_{C^{0,\gamma}}=1 ∥ g ∞ ∥ C 0 , γ = 1 . We thus obtain (285).
We now turn to the Lipschitz bound (286). Consider two nearby filaments X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1. We aim to bound the difference v ( a ) − v ( b ) v^{(a)}-v^{(b)} v ( a ) − v ( b ) where v ( a ) , v ( b ) v^{(a)},v^{(b)} v ( a ) , v ( b ) satisfy
( 1 2 I + D ( a ) + V ( a ) ) [ v ( a ) ] = φ = ( 1 2 I + D ( b ) + V ( b ) ) [ v ( b ) ] . \left(\frac{1}{2}I+D^{(a)}+V^{(a)}\right)[v^{(a)}]=\varphi=\left(\frac{1}{2}I+D^{(b)}+V^{(b)}\right)[v^{(b)}]. ( 2 1 I + D ( a ) + V ( a ) ) [ v ( a ) ] = φ = ( 2 1 I + D ( b ) + V ( b ) ) [ v ( b ) ] .
Rewriting, we have
v ( a ) − v ( b ) = 2 ( D ( b ) + V ( b ) ) [ v ( b ) ] − 2 ( D ( a ) + V ( a ) ) [ v ( a ) ] = 2 ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] + 2 ( D ( a ) + V ( a ) ) [ v ( b ) − v ( a ) ] , \begin{aligned}
v^{(a)}-v^{(b)}={}&2(D^{(b)}+V^{(b)})[v^{(b)}]-2(D^{(a)}+V^{(a)})[v^{(a)}]\\
={}&2(D^{(b)}+V^{(b)}-D^{(a)}-V^{(a)})[v^{(b)}]+2(D^{(a)}+V^{(a)})[v^{(b)}-v^{(a)}],
\end{aligned} v ( a ) − v ( b ) = = 2 ( D ( b ) + V ( b ) ) [ v ( b ) ] − 2 ( D ( a ) + V ( a ) ) [ v ( a ) ] 2 ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] + 2 ( D ( a ) + V ( a ) ) [ v ( b ) − v ( a ) ] ,
or, moving all v ( a ) − v ( b ) v^{(a)}-v^{(b)} v ( a ) − v ( b ) terms to the left hand side,
v ( a ) − v ( b ) = ( 1 2 I + D ( a ) + V ( a ) ) − 1 ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] . v^{(a)}-v^{(b)}=\left(\frac{1}{2}I+D^{(a)}+V^{(a)}\right)^{-1}(D^{(b)}+V^{(b)}-D^{(a)}-V^{(a)})[v^{(b)}]. v ( a ) − v ( b ) = ( 2 1 I + D ( a ) + V ( a ) ) − 1 ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] .
Then, using the single filament bound (285) and Lemma 5.2, we have
∥ v ( a ) − v ( b ) ∥ C 0 , γ = ∥ ( 1 2 I + D ( a ) + V ( a ) ) − 1 ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , c Γ ) ∥ ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ( b ) ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ φ ∥ C 0 , γ . \begin{aligned}
\lVert v^{(a)}-v^{(b)}\rVert_{C^{0,\gamma}}
&=\left\lVert\left(\frac{1}{2}I+D^{(a)}+V^{(a)}\right)^{-1}(D^{(b)}+V^{(b)}-D^{(a)}-V^{(a)})[v^{(b)}]\right\rVert_{C^{0,\gamma}}\\
&\le c(\epsilon,\kappa_{*,\gamma+}^{(a)},c_{\Gamma})\left\lVert(D^{(b)}+V^{(b)}-D^{(a)}-V^{(a)})[v^{(b)}]\right\rVert_{C^{0,\gamma}}\\
&\le c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\gamma+}}\lVert v^{(b)}\rVert_{C^{0,\gamma}}\\
&\le c(\epsilon,\kappa_{*,\gamma+}^{(a)},\kappa_{*,\gamma+}^{(b)},c_{\Gamma})\lVert X^{(a)}-X^{(b)}\rVert_{C^{2,\gamma+}}\lVert\varphi\rVert_{C^{0,\gamma}}.
\end{aligned} ∥ v ( a ) − v ( b ) ∥ C 0 , γ = ( 2 1 I + D ( a ) + V ( a ) ) − 1 ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , c Γ ) ( D ( b ) + V ( b ) − D ( a ) − V ( a ) ) [ v ( b ) ] C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ v ( b ) ∥ C 0 , γ ≤ c ( ϵ , κ ∗ , γ + ( a ) , κ ∗ , γ + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , γ + ∥ φ ∥ C 0 , γ .
Proof of Lemma 5.5: hypersingular operator bounds
We use the expansions (206) and (212) to write the kernel K T K_{\mathcal{T}} K T (278) in terms of s , s ^ , θ , θ ^ s,\hat{s},\theta,\hat{\theta} s , s ^ , θ , θ ^ as
K T = 1 4 π ( K T , 0 + K T , 1 + K T , 2 ) . (301) K_{\mathcal{T}}=\frac{1}{4\pi}\left(K_{\mathcal{T},0}+K_{\mathcal{T},1}+K_{\mathcal{T},2}\right). \tag*{(301)} K T = 4 π 1 ( K T , 0 + K T , 1 + K T , 2 ) . (301)
K T , 0 = 2 e r ⊗ e r ∣ R ∣ 3 + 3 ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⊗ ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ∣ R ∣ 5 K_{\mathcal{T},0}=2\frac{e_r\otimes e_r}{\lvert R\rvert^3}+3\frac{\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)\otimes\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)}{\lvert R\rvert^5} K T , 0 = 2 ∣ R ∣ 3 e r ⊗ e r + 3 ∣ R ∣ 5 ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⊗ ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r )
K T , 1 = − 3 s ^ 2 [ Q t ⊗ ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) + ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⊗ Q t ] ∣ R ∣ 5 − 2 s ^ e r ⊗ Q r + 2 sin ( θ ^ 2 ) e r ⊗ e θ ∣ R ∣ 3 − 3 ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ⊗ e r ∣ R ∣ 5 + 3 ( 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ) e r ⊗ ( s ^ e t + 2 ϵ sin ( θ ^ 2 ) e θ + ϵ s ^ Q r ) ∣ R ∣ 5 \begin{aligned}
K_{\mathcal{T},1}={}&-3\frac{\hat{s}^2\left[Q_t\otimes\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)+\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)\otimes Q_t\right]}{\lvert R\rvert^5}\\
&-2\frac{\hat{s}e_r\otimes Q_r+2\sin\left(\frac{\hat{\theta}}{2}\right)e_r\otimes e_\theta}{\lvert R\rvert^3}
-3\frac{\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)-\hat{s}^2Q_{Rn'}\right)\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)\otimes e_r}{\lvert R\rvert^5}\\
&+3\frac{\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)+\hat{s}^2Q_{Rn}\right)e_r\otimes\left(\hat{s}e_t+2\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta+\epsilon\hat{s}Q_r\right)}{\lvert R\rvert^5}
\end{aligned} K T , 1 = − 3 ∣ R ∣ 5 s ^ 2 [ Q t ⊗ ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) + ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⊗ Q t ] − 2 ∣ R ∣ 3 s ^ e r ⊗ Q r + 2 sin ( 2 θ ^ ) e r ⊗ e θ − 3 ∣ R ∣ 5 ( 2 ϵ sin 2 ( 2 θ ^ ) − s ^ 2 Q R n ′ ) ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r ) ⊗ e r + 3 ∣ R ∣ 5 ( 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ) e r ⊗ ( s ^ e t + 2 ϵ sin ( 2 θ ^ ) e θ + ϵ s ^ Q r )
K T , 2 = 6 s ^ 4 Q t ⊗ Q t ∣ R ∣ 5 + 3 s ^ 2 [ ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) Q t ⊗ e r − ( 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ) e r ⊗ Q t ] ∣ R ∣ 5 − 3 ( 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ) [ ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) I + ( s ^ Q r + 2 sin ( θ ^ 2 ) e θ ) ⊗ R ] ∣ R ∣ 5 − 3 [ ( 2 sin 2 ( θ ^ 2 ) − s ^ 2 Q 0 , 5 ) ∣ R ∣ 2 − 10 ( 2 ϵ sin 2 ( θ ^ 2 ) + s ^ 2 Q R n ) ( 2 ϵ sin 2 ( θ ^ 2 ) − s ^ 2 Q R n ′ ) ] R ⊗ R ∣ R ∣ 7 . \begin{aligned}
K_{\mathcal{T},2}={}&6\frac{\hat{s}^4Q_t\otimes Q_t}{\lvert R\rvert^5}
+3\frac{\hat{s}^2\left[\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)-\hat{s}^2Q_{Rn'}\right)Q_t\otimes e_r-\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)+\hat{s}^2Q_{Rn}\right)e_r\otimes Q_t\right]}{\lvert R\rvert^5}\\
&-3\frac{\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)+\hat{s}^2Q_{Rn}\right)\left[\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)-\hat{s}^2Q_{Rn'}\right)I+\left(\hat{s}Q_r+2\sin\left(\frac{\hat{\theta}}{2}\right)e_\theta\right)\otimes R\right]}{\lvert R\rvert^5}\\
&-3\frac{\left[\left(2\sin^2\left(\frac{\hat{\theta}}{2}\right)-\hat{s}^2Q_{0,5}\right)\lvert R\rvert^2-10\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)+\hat{s}^2Q_{Rn}\right)\left(2\epsilon\sin^2\left(\frac{\hat{\theta}}{2}\right)-\hat{s}^2Q_{Rn'}\right)\right]R\otimes R}{\lvert R\rvert^7}.
\end{aligned} K T , 2 = 6 ∣ R ∣ 5 s ^ 4 Q t ⊗ Q t + 3 ∣ R ∣ 5 s ^ 2 [ ( 2 ϵ sin 2 ( 2 θ ^ ) − s ^ 2 Q R n ′ ) Q t ⊗ e r − ( 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ) e r ⊗ Q t ] − 3 ∣ R ∣ 5 ( 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ) [ ( 2 ϵ sin 2 ( 2 θ ^ ) − s ^ 2 Q R n ′ ) I + ( s ^ Q r + 2 sin ( 2 θ ^ ) e θ ) ⊗ R ] − 3 ∣ R ∣ 7 [ ( 2 sin 2 ( 2 θ ^ ) − s ^ 2 Q 0 , 5 ) ∣ R ∣ 2 − 10 ( 2 ϵ sin 2 ( 2 θ ^ ) + s ^ 2 Q R n ) ( 2 ϵ sin 2 ( 2 θ ^ ) − s ^ 2 Q R n ′ ) ] R ⊗ R .
Accordingly, for j = 0 , 1 , 2 j=0,1,2 j = 0 , 1 , 2 , we define
T j [ φ ] = − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π K T , j ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) ) J ϵ ( s − s ^ , θ − θ ^ ) d θ ^ d s . \mathcal{T}_j[\varphi]=-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{\mathcal{T},j}\left(\varphi(s-\hat{s},\theta-\hat{\theta})-\varphi(s,\theta)\right)\mathcal{J}_{\epsilon}(s-\hat{s},\theta-\hat{\theta})\,d\hat{\theta}ds. T j [ φ ] = − p.v. ∫ − 1/2 1/2 ∫ − π π K T , j ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) ) J ϵ ( s − s ^ , θ − θ ^ ) d θ ^ d s .
We begin with bounds for T 2 \mathcal{T}_2 T 2 . Using Lemma 4.2 , we have
∣ T 2 [ φ ] ∣ ≤ 2 ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π ∣ K T , 2 ∣ ∣ J ϵ ∣ d θ ^ d s ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π 1 + ϵ − 1 + ϵ − 2 ∣ R ∣ ϵ d θ ^ d s ≤ c ( κ ∗ , c Γ ) ϵ − 1 ∥ φ ∥ L ∞ . \begin{aligned}
\left\lvert\mathcal{T}_2[\varphi]\right\rvert\leq2\lVert\varphi\rVert_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\lvert K_{\mathcal{T},2}\rvert\lvert\mathcal{J}_{\epsilon}\rvert\,d\hat{\theta}ds \\
&\leq c(\kappa_*)\lVert\varphi\rVert_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\frac{1+\epsilon^{-1}+\epsilon^{-2}}{\lvert R\rvert}\epsilon\,d\hat{\theta}ds\leq c(\kappa_*,c_\Gamma)\epsilon^{-1}\lVert\varphi\rVert_{L^\infty}.
\end{aligned} ∣ T 2 [ φ ] ∣ ≤ 2 ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ∣ K T , 2 ∣ ∣ J ϵ ∣ d θ ^ d s ≤ c ( κ ∗ ) ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ 1 + ϵ − 1 + ϵ − 2 ϵ d θ ^ d s ≤ c ( κ ∗ , c Γ ) ϵ − 1 ∥ φ ∥ L ∞ .
Furthermore, using that the remainder terms Q i Q_i Q i above each satisfy an estimate of the form (214) , and using the expansion (211) , for two curves X ( a ) X^{(a)} X ( a ) and X ( b ) X^{(b)} X ( b ) satisfying Lemma 1.1 , we have
∣ T 2 ( a ) [ φ ] − T 2 ( b ) [ φ ] ∣ ≤ 2 ∥ φ ∥ L ∞ ∫ − 1 / 2 1 / 2 ∫ − π π ∣ K T , 2 ( a ) J ϵ ( a ) − K T , 2 ( b ) J ϵ ( b ) ∣ d θ ^ d s ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ − 1 ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ . \begin{aligned}
\left\lvert\mathcal{T}_2^{(a)}[\varphi]-\mathcal{T}_2^{(b)}[\varphi]\right\rvert\leq2\lVert\varphi\rVert_{L^\infty}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left\lvert K_{\mathcal{T},2}^{(a)}\mathcal{J}_{\epsilon}^{(a)}-K_{\mathcal{T},2}^{(b)}\mathcal{J}_{\epsilon}^{(b)}\right\rvert\,d\hat{\theta}ds \\
&\leq c(\kappa_*^{(a)},\kappa_*^{(b)},c_\Gamma)\epsilon^{-1}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^2}\lVert\varphi\rVert_{L^\infty}.
\end{aligned} T 2 ( a ) [ φ ] − T 2 ( b ) [ φ ] ≤ 2 ∥ φ ∥ L ∞ ∫ − 1/2 1/2 ∫ − π π K T , 2 ( a ) J ϵ ( a ) − K T , 2 ( b ) J ϵ ( b ) d θ ^ d s ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ − 1 X ( a ) − X ( b ) C 2 ∥ φ ∥ L ∞ .
In addition, by Lemma 4.4 , we have
∣ T 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ − 1 − α ∥ φ ∥ C 0 , α \left\lvert\mathcal{T}_2[\varphi]\right\rvert_{\dot{C}^{0,\alpha}}\leq c(\kappa_*,\alpha,c_\Gamma)\epsilon^{-1-\alpha}\lVert\varphi\rVert_{C^{0,\alpha}} ∣ T 2 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ − 1 − α ∥ φ ∥ C 0 , α
∣ T 2 ( a ) [ φ ] − T 2 ( b ) [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ − 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ C 0 , α . \left\lvert\mathcal{T}_2^{(a)}[\varphi]-\mathcal{T}_2^{(b)}[\varphi]\right\rvert_{\dot{C}^{0,\alpha}}\leq c(\kappa_*^{(a)},\kappa_*^{(b)},c_\Gamma)\epsilon^{-1-\alpha}\left\lVert X^{(a)}-X^{(b)}\right\rVert_{C^{2,\alpha}}\lVert\varphi\rVert_{C^{0,\alpha}}. T 2 ( a ) [ φ ] − T 2 ( b ) [ φ ] C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ϵ − 1 − α X ( a ) − X ( b ) C 2 , α ∥ φ ∥ C 0 , α .
Here we use case (2) of Lemma 4.4 to estimate terms of the form sin 2 ( θ ^ / 2 ) ∣ R ∣ 3 \frac{\sin^{2}(\hat{\theta}/2)}{\lvert R\rvert^{3}} ∣ R ∣ 3 s i n 2 ( θ ^ /2 ) , treating one factor of sin ( θ ^ / 2 ) \sin(\hat{\theta}/2) sin ( θ ^ /2 ) as a coefficient, which allows us to save a factor of ϵ − α \epsilon^{-\alpha} ϵ − α . We may use case (1) to estimate the remaining terms.
We next estimate T 1 [ φ ] \mathcal{T}_{1}[\varphi] T 1 [ φ ] . We may first use Lemma 4.3 to obtain
∣ T 1 [ φ ] ∣ ≤ ∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π K T , 1 ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) ) J ϵ d θ ^ d s ∣ ≤ c ( κ ∗ , c Γ ) ϵ α − 1 ( max i ∥ Q i ∥ C 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ − 1 ∥ φ ∥ C 0 , α . \lvert\mathcal{T}_{1}[\varphi]\rvert\le\left\lvert\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi} K_{\mathcal{T},1}(\varphi(s-\hat{s},\theta-\hat{\theta})-\varphi(s,\theta))\mathcal{J}_{\epsilon}\,\mathrm{d}\hat{\theta}\mathrm{d}s\right\rvert
\le c(\kappa_{*},c_{\Gamma})\epsilon^{\alpha-1}\left(\max_{i}\lVert Q_{i}\rVert_{C^{0,\alpha}}\right)\lVert\varphi\rVert_{C^{0,\alpha}}
\le c(\kappa_{*,\alpha},c_{\Gamma})\epsilon^{-1}\lVert\varphi\rVert_{C^{0,\alpha}}. ∣ T 1 [ φ ]∣ ≤ p.v. ∫ − 1/2 1/2 ∫ − π π K T , 1 ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ )) J ϵ d θ ^ d s ≤ c ( κ ∗ , c Γ ) ϵ α − 1 ( i max ∥ Q i ∥ C 0 , α ) ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ − 1 ∥ φ ∥ C 0 , α .
As usual, here we use Q i Q_i Q i as catch-all notation for the remainder coefficients in the above expansion. Similarly, using the expressions (213) and (210), for two curves X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) satisfying Lemma 1.1, we have
∣ T 1 ( a ) [ φ ] − T 1 ( b ) [ φ ] ∣ ≤ ∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( K T , 1 ( a ) J ϵ ( a ) − K T , 1 ( b ) J ϵ ( b ) ) ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) ) d θ ^ d s ∣ ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ − 1 ∥ X ( a ) − X ( b ) ∥ C 2 , α ∥ φ ∥ C 0 , α . \left\lvert\mathcal{T}_{1}^{(a)}[\varphi]-\mathcal{T}_{1}^{(b)}[\varphi]\right\rvert\le\left\lvert\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(K_{\mathcal{T},1}^{(a)}\mathcal{J}_{\epsilon}^{(a)}-K_{\mathcal{T},1}^{(b)}\mathcal{J}_{\epsilon}^{(b)}\right)(\varphi(s-\hat{s},\theta-\hat{\theta})-\varphi(s,\theta))\,\mathrm{d}\hat{\theta}\mathrm{d}s\right\rvert
\le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_{\Gamma})\epsilon^{-1}\left\lVert\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\rVert_{C^{2,\alpha}}\lVert\varphi\rVert_{C^{0,\alpha}}. T 1 ( a ) [ φ ] − T 1 ( b ) [ φ ] ≤ p.v. ∫ − 1/2 1/2 ∫ − π π ( K T , 1 ( a ) J ϵ ( a ) − K T , 1 ( b ) J ϵ ( b ) ) ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ )) d θ ^ d s ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ − 1 X ( a ) − X ( b ) C 2 , α ∥ φ ∥ C 0 , α .
We next turn to C ˙ 0 , α \dot{C}^{0,\alpha} C ˙ 0 , α estimates. Using Lemma 4.4, case (2), we obtain
∣ T 1 [ φ ] ∣ C ˙ 0 , α ≤ ∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π K T , 1 ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) ) J ϵ d θ ^ d s ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) max i ( ∥ Q i ∥ C 1 0 , α + ∥ Q i ∥ C 2 0 , α ) ϵ − 1 ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ − 1 + α ∥ φ ∥ C 0 , α , \left\lvert\mathcal{T}_{1}[\varphi]\right\rvert_{\dot{C}^{0,\alpha}} \le\left\lvert\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi} K_{\mathcal{T},1}(\varphi(s-\hat{s},\theta-\hat{\theta})-\varphi(s,\theta))\mathcal{J}_{\epsilon}\,\mathrm{d}\hat{\theta}\mathrm{d}s\right\rvert_{\dot{C}^{0,\alpha}}
\le c(\kappa_{*},c_{\Gamma})\max_{i}\left(\lVert Q_{i}\rVert_{C_{1}^{0,\alpha}}+\lVert Q_{i}\rVert_{C_{2}^{0,\alpha}}\right)\epsilon^{-1}\lVert\varphi\rVert_{C^{0,\alpha}}
\le c(\kappa_{*,\alpha},c_{\Gamma})\epsilon^{-1+\alpha}\lVert\varphi\rVert_{C^{0,\alpha}}, ∣ T 1 [ φ ] ∣ C ˙ 0 , α ≤ p.v. ∫ − 1/2 1/2 ∫ − π π K T , 1 ( φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ )) J ϵ d θ ^ d s C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) i max ( ∥ Q i ∥ C 1 0 , α + ∥ Q i ∥ C 2 0 , α ) ϵ − 1 ∥ φ ∥ C 0 , α ≤ c ( κ ∗ , α , c Γ ) ϵ − 1 + α ∥ φ ∥ C 0 , α ,
∣ T 1 ( a ) [ φ ] − T 1 ( b ) [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ − 1 + α ∥ X ( a ) − X ( b ) ∥ C 2 , α 2 ∥ φ ∥ C 0 , α . \left\lvert\mathcal{T}_{1}^{(a)}[\varphi]-\mathcal{T}_{1}^{(b)}[\varphi]\right\rvert_{\dot{C}^{0,\alpha}}
\le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_{\Gamma})\epsilon^{-1+\alpha}\left\lVert\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\rVert_{C_{2,\alpha}^{2}}\lVert\varphi\rVert_{C^{0,\alpha}}. T 1 ( a ) [ φ ] − T 1 ( b ) [ φ ] C ˙ 0 , α ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ − 1 + α X ( a ) − X ( b ) C 2 , α 2 ∥ φ ∥ C 0 , α .
Finally, we estimate T 0 [ φ ] \mathcal{T}_{0}[\varphi] T 0 [ φ ] . Here we will use that φ ∈ C 1 , α \varphi\in C^{1,\alpha} φ ∈ C 1 , α and we may thus expand
φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) = s ^ Q φ , 1 + ϵ θ ^ Q φ , 2 , \varphi(s-\hat{s},\theta-\hat{\theta})-\varphi(s,\theta)=\hat{s}Q_{\varphi,1}+\epsilon\hat{\theta}Q_{\varphi,2}, φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) = s ^ Q φ , 1 + ϵ θ ^ Q φ , 2 ,
∥ Q φ , 1 ∥ Y ≤ c ∥ ∂ s φ ∥ Y , ∥ Q φ , 2 ∥ Y ≤ c ∥ 1 ϵ ∂ θ φ ∥ Y , (302) \lVert Q_{\varphi,1}\rVert_{Y}\le c\lVert\partial_{s}\varphi\rVert_{Y},\qquad\lVert Q_{\varphi,2}\rVert_{Y}\le c\left\lVert\frac{1}{\epsilon}\partial_{\theta}\varphi\right\rVert_{Y}, \tag*{(302)} ∥ Q φ , 1 ∥ Y ≤ c ∥ ∂ s φ ∥ Y , ∥ Q φ , 2 ∥ Y ≤ c ϵ 1 ∂ θ φ Y , (302)
where the norms Y Y Y are as in (199) with α \alpha α in place of β \beta β . Note that since φ \varphi φ is a function of just ( s , θ ) (s,\theta) ( s , θ ) , ∥ ∂ s φ ∥ C μ 0 , α ≤ ∥ ∂ s φ ∥ C 0 , α \lVert\partial_{s}\varphi\rVert_{C_{\mu}^{0,\alpha}}\le\lVert\partial_{s}\varphi\rVert_{C^{0,\alpha}} ∥ ∂ s φ ∥ C μ 0 , α ≤ ∥ ∂ s φ ∥ C 0 , α for μ = 1 , 2 \mu=1,2 μ = 1 , 2 . Using (300) along with Lemma 4.3, we obtain the L ∞ L^{\infty} L ∞ estimates
∣ T 0 [ φ ] ∣ ≤ ∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π K T , 0 ( s ^ Q φ , 1 + ϵ θ ^ Q φ , 2 ) J ϵ d θ ^ d s ∣ ≤ c ( κ ∗ , α , c Γ ) ϵ α ( ∥ Q φ , 1 ∥ C 2 0 , α + ∥ Q φ , 2 ∥ C 2 0 , α ) ≤ c ( κ ∗ , α , c Γ ) ϵ α ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) . \lvert\mathcal{T}_{0}[\varphi]\rvert\le\left\lvert\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}K_{\mathcal{T},0}(\hat{s}Q_{\varphi,1}+\epsilon\hat{\theta}Q_{\varphi,2})\mathcal{J}_{\epsilon}\,\mathrm{d}\hat{\theta}\mathrm{d}s\right\rvert
\le c(\kappa_{*,\alpha},c_{\Gamma})\epsilon^{\alpha}\left(\lVert Q_{\varphi,1}\rVert_{C_{2}^{0,\alpha}}+\lVert Q_{\varphi,2}\rVert_{C_{2}^{0,\alpha}}\right)
\le c(\kappa_{*,\alpha},c_{\Gamma})\epsilon^{\alpha}\left(\lVert\partial_{s}\varphi\rVert_{C^{0,\alpha}}+\left\lVert\frac{1}{\epsilon}\partial_{\theta}\varphi\right\rVert_{C^{0,\alpha}}\right). ∣ T 0 [ φ ]∣ ≤ p.v. ∫ − 1/2 1/2 ∫ − π π K T , 0 ( s ^ Q φ , 1 + ϵ θ ^ Q φ , 2 ) J ϵ d θ ^ d s ≤ c ( κ ∗ , α , c Γ ) ϵ α ( ∥ Q φ , 1 ∥ C 2 0 , α + ∥ Q φ , 2 ∥ C 2 0 , α ) ≤ c ( κ ∗ , α , c Γ ) ϵ α ( ∥ ∂ s φ ∥ C 0 , α + ϵ 1 ∂ θ φ C 0 , α ) .
In addition, for two curves X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) satisfying Lemma 1.1, we have
∣ T 0 ( a ) [ φ ] − T 0 ( b ) [ φ ] ∣ ≤ ∣ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π ( K T , 0 ( a ) J ϵ ( a ) − K T , 0 ( b ) J ϵ ( b ) ) ( s ^ Q φ , 1 + ϵ θ ^ Q φ , 2 ) d θ ^ d s ∣ ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ α ∥ X ( a ) − X ( b ) ∥ C 2 , α ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) . \left\lvert\mathcal{T}_{0}^{(a)}[\varphi]-\mathcal{T}_{0}^{(b)}[\varphi]\right\rvert\le\left\lvert\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}\left(K_{\mathcal{T},0}^{(a)}\mathcal{J}_{\epsilon}^{(a)}-K_{\mathcal{T},0}^{(b)}\mathcal{J}_{\epsilon}^{(b)}\right)(\hat{s}Q_{\varphi,1}+\epsilon\hat{\theta}Q_{\varphi,2})\,\mathrm{d}\hat{\theta}\mathrm{d}s\right\rvert
\le c(\kappa_{*,\alpha}^{(a)},\kappa_{*,\alpha}^{(b)},c_{\Gamma})\epsilon^{\alpha}\left\lVert\mathbf{X}^{(a)}-\mathbf{X}^{(b)}\right\rVert_{C^{2,\alpha}}\left(\lVert\partial_{s}\varphi\rVert_{C^{0,\alpha}}+\left\lVert\frac{1}{\epsilon}\partial_{\theta}\varphi\right\rVert_{C^{0,\alpha}}\right). T 0 ( a ) [ φ ] − T 0 ( b ) [ φ ] ≤ p.v. ∫ − 1/2 1/2 ∫ − π π ( K T , 0 ( a ) J ϵ ( a ) − K T , 0 ( b ) J ϵ ( b ) ) ( s ^ Q φ , 1 + ϵ θ ^ Q φ , 2 ) d θ ^ d s ≤ c ( κ ∗ , α ( a ) , κ ∗ , α ( b ) , c Γ ) ϵ α X ( a ) − X ( b ) C 2 , α ( ∥ ∂ s φ ∥ C 0 , α + ϵ 1 ∂ θ φ C 0 , α ) .
Next, using Lemma 5.4, we have
∣ T 0 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ( ( ∥ ∂ s φ ∥ L ∞ + ∥ 1 ϵ ∂ θ φ ∥ L ∞ ) max i ( ∥ Q i ∥ C 1 0 , α + + ∥ Q i ∥ C 2 0 , α ) + ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) max i ∥ Q i ∥ L ∞ ) ≤ c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) + c ( κ ∗ , α + , c Γ ) ϵ − α + ( ∥ ∂ s φ ∥ L ∞ + ∥ 1 ϵ ∂ θ φ ∥ L ∞ ) . \begin{aligned}
|\mathcal{T}_0[\varphi]|_{\dot{C}^{0,\alpha}} \leq c(\kappa_*,c_\Gamma)\left(\left(\|\partial_s\varphi\|_{L^\infty}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{L^\infty}\right)\max_i\left(\|Q_i\|_{C_1^{0,\alpha_+}}+\|Q_i\|_{C_2^{0,\alpha}}\right)+\left(\|\partial_s\varphi\|_{C^{0,\alpha}}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C^{0,\alpha}}\right)\max_i\|Q_i\|_{L^\infty}\right) \\
&\leq c(\kappa_*,c_\Gamma)\left(\|\partial_s\varphi\|_{C^{0,\alpha}}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C^{0,\alpha}}\right)+c(\kappa_*,\alpha_+,c_\Gamma)\epsilon^{-\alpha_+}\left(\|\partial_s\varphi\|_{L^\infty}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{L^\infty}\right).
\end{aligned} ∣ T 0 [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ , c Γ ) ( ( ∥ ∂ s φ ∥ L ∞ + ϵ 1 ∂ θ φ L ∞ ) i max ( ∥ Q i ∥ C 1 0 , α + + ∥ Q i ∥ C 2 0 , α ) + ( ∥ ∂ s φ ∥ C 0 , α + ϵ 1 ∂ θ φ C 0 , α ) i max ∥ Q i ∥ L ∞ ) ≤ c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ C 0 , α + ϵ 1 ∂ θ φ C 0 , α ) + c ( κ ∗ , α + , c Γ ) ϵ − α + ( ∥ ∂ s φ ∥ L ∞ + ϵ 1 ∂ θ φ L ∞ ) .
Here the coefficients Q i Q_i Q i include ϵ − 1 J ϵ \epsilon^{-1}\mathcal{J}_\epsilon ϵ − 1 J ϵ , ϵ Q r \epsilon Q_r ϵ Q r , and e r , e θ e_r,e_\theta e r , e θ , which incur an additional ϵ − α + \epsilon^{-\alpha_+} ϵ − α + due to θ \theta θ -dependence. Similarly, we have
∣ T 0 ( a ) [ φ ] − T 0 ( b ) [ φ ] ∣ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) + c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ − α + ∥ X ( a ) − X ( b ) ∥ C 2 , α + ( ∥ ∂ s φ ∥ L ∞ + ∥ 1 ϵ ∂ θ φ ∥ L ∞ ) . \left|\mathcal{T}_0^{(a)}[\varphi]-\mathcal{T}_0^{(b)}[\varphi]\right|_{\dot{C}^{0,\alpha}}\leq c(\kappa_*^{(a)},\kappa_*^{(b)},c_\Gamma)\left\|X^{(a)}-X^{(b)}\right\|_{C^2}\left(\|\partial_s\varphi\|_{C^{0,\alpha}}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C^{0,\alpha}}\right)+c(\kappa_{*,\alpha_+}^{(a)},\kappa_{*,\alpha_+}^{(b)},c_\Gamma)\epsilon^{-\alpha_+}\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\alpha_+}}\left(\|\partial_s\varphi\|_{L^\infty}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{L^\infty}\right). T 0 ( a ) [ φ ] − T 0 ( b ) [ φ ] C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , c Γ ) X ( a ) − X ( b ) C 2 ( ∥ ∂ s φ ∥ C 0 , α + ϵ 1 ∂ θ φ C 0 , α ) + c ( κ ∗ , α + ( a ) , κ ∗ , α + ( b ) , c Γ ) ϵ − α + X ( a ) − X ( b ) C 2 , α + ( ∥ ∂ s φ ∥ L ∞ + ϵ 1 ∂ θ φ L ∞ ) .
In addition to the stress σ [ D [ φ ] ] n ( x ) = T 0 + T 1 + T 2 \boldsymbol{\sigma}[\mathcal{D}[\varphi]]\boldsymbol{n}(x)=\mathcal{T}_0+\mathcal{T}_1+\mathcal{T}_2 σ [ D [ φ ]] n ( x ) = T 0 + T 1 + T 2 , we need to estimate the stress σ [ V [ φ ] ] n ( x ) \boldsymbol{\sigma}[\mathcal{V}[\varphi]]\boldsymbol{n}(x) σ [ V [ φ ]] n ( x ) due to the completion flow, given by (281) .
Using that d S x = J ϵ d θ d s dS_x=\mathcal{J}_\epsilon\,d\theta ds d S x = J ϵ d θ d s and ∣ R x ∣ ≥ ϵ |R_x|\geq\epsilon ∣ R x ∣ ≥ ϵ while ∣ x ′ − X ( s ′ ) ∣ ≤ ϵ |x'-X(s')|\leq\epsilon ∣ x ′ − X ( s ′ ) ∣ ≤ ϵ , we have
∣ σ [ V [ φ ] ] n ∣ ≤ c ∫ Γ ϵ ∣ φ ∣ ϵ 2 d S x ′ ≤ c ( κ ∗ ) ϵ − 1 ∥ φ ∥ L ∞ , |\boldsymbol{\sigma}[\mathcal{V}[\varphi]]\boldsymbol{n}|\leq c\int_{\Gamma_\epsilon}\frac{|\varphi|}{\epsilon^2}\,dS_{x'}\leq c(\kappa_*)\epsilon^{-1}\|\varphi\|_{L^\infty}, ∣ σ [ V [ φ ]] n ∣ ≤ c ∫ Γ ϵ ϵ 2 ∣ φ ∣ d S x ′ ≤ c ( κ ∗ ) ϵ − 1 ∥ φ ∥ L ∞ ,
∣ σ [ V ( a ) [ φ ] ] n ( a ) − σ [ V ( b ) [ φ ] ] n ( b ) ∣ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ − 1 ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ . \left|\boldsymbol{\sigma}[\mathcal{V}^{(a)}[\varphi]]\boldsymbol{n}^{(a)}-\boldsymbol{\sigma}[\mathcal{V}^{(b)}[\varphi]]\boldsymbol{n}^{(b)}\right|\leq c(\kappa_*^{(a)},\kappa_*^{(b)})\epsilon^{-1}\left\|X^{(a)}-X^{(b)}\right\|_{C^2}\|\varphi\|_{L^\infty}. σ [ V ( a ) [ φ ]] n ( a ) − σ [ V ( b ) [ φ ]] n ( b ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ − 1 X ( a ) − X ( b ) C 2 ∥ φ ∥ L ∞ .
In addition, we have
∣ σ [ V [ φ ] ] n ∣ C ˙ 0 , α ≤ c ( κ ∗ ) ϵ − 1 − α ∥ φ ∥ L ∞ , |\boldsymbol{\sigma}[\mathcal{V}[\varphi]]\boldsymbol{n}|_{\dot{C}^{0,\alpha}}\leq c(\kappa_*)\epsilon^{-1-\alpha}\|\varphi\|_{L^\infty}, ∣ σ [ V [ φ ]] n ∣ C ˙ 0 , α ≤ c ( κ ∗ ) ϵ − 1 − α ∥ φ ∥ L ∞ ,
∣ σ [ V ( a ) [ φ ] ] n ( a ) − σ [ V ( b ) [ φ ] ] n ( b ) ∣ C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ − 1 − α ∥ X ( a ) − X ( b ) ∥ C 2 ∥ φ ∥ L ∞ . \left|\boldsymbol{\sigma}[\mathcal{V}^{(a)}[\varphi]]\boldsymbol{n}^{(a)}-\boldsymbol{\sigma}[\mathcal{V}^{(b)}[\varphi]]\boldsymbol{n}^{(b)}\right|_{\dot{C}^{0,\alpha}}\leq c(\kappa_*^{(a)},\kappa_*^{(b)})\epsilon^{-1-\alpha}\left\|X^{(a)}-X^{(b)}\right\|_{C^2}\|\varphi\|_{L^\infty}. σ [ V ( a ) [ φ ]] n ( a ) − σ [ V ( b ) [ φ ]] n ( b ) C ˙ 0 , α ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ϵ − 1 − α X ( a ) − X ( b ) C 2 ∥ φ ∥ L ∞ .
Combining the estimates for σ [ V ] \boldsymbol{\sigma}[\mathcal{V}] σ [ V ] with the estimates for T 0 \mathcal{T}_0 T 0 , T 1 \mathcal{T}_1 T 1 and T 2 \mathcal{T}_2 T 2 , we obtain Lemma 5.5. □
Slender body NtD in Hölder spaces
Here we prove Lemma 1.14 regarding Hölder space estimates for the slender body Neumann-to-Dirichlet map L ϵ \mathcal{L}_\epsilon L ϵ about a curved filament. As in the Laplace setting [59 section 5] , we will rely on a Campanato-type argument.
We begin by recalling the L 2 L^2 L 2 -based solution theory for the slender body boundary value problem (11) , developed in [52 Theorem 1.2] :
Lemma 6.1 (L 2 L^2 L 2 -based SB BVP solution theory). Consider a filament Σ ϵ \Sigma_\epsilon Σ ϵ with centerline X ( s ) ∈ C 2 ( T ) X(s)\in C^2(\mathbb{T}) X ( s ) ∈ C 2 ( T ) . Given slender body Neumann data f ( s ) ∈ L 2 ( T ) f(s)\in L^2(\mathbb{T}) f ( s ) ∈ L 2 ( T ) , the solution ( u , p ) (u,p) ( u , p ) to the slender body boundary value problem (11) belongs to D 1 , 2 ( Ω ϵ ) × L 2 ( Ω ϵ ) D^{1,2}(\Omega_\epsilon)\times L^2(\Omega_\epsilon) D 1 , 2 ( Ω ϵ ) × L 2 ( Ω ϵ ) , where
D 1 , 2 ( Ω ϵ ) : = { u ∈ L 6 ( Ω ϵ ) : ∇ u ∈ L 2 ( Ω ϵ ) } , D^{1,2}(\Omega_\epsilon):=\{u\in L^6(\Omega_\epsilon):\nabla u\in L^2(\Omega_\epsilon)\}, D 1 , 2 ( Ω ϵ ) := { u ∈ L 6 ( Ω ϵ ) : ∇ u ∈ L 2 ( Ω ϵ )} , and satisfies the bound
∥ u ∥ D 1 , 2 ( Ω ϵ ) + ∥ p ∥ L 2 ( Ω ϵ ) ≡ ∥ ∇ u ∥ L 2 ( Ω ϵ ) + ∥ p ∥ L 2 ( Ω ϵ ) ≤ c ( κ ∗ , c Γ ) ∣ log ϵ ∣ 1 / 2 ∥ f ∥ L 2 ( T ) . (303) \|u\|_{D^{1,2}(\Omega_\epsilon)}+\|p\|_{L^2(\Omega_\epsilon)}\equiv\|\nabla u\|_{L^2(\Omega_\epsilon)}+\|p\|_{L^2(\Omega_\epsilon)}\leq c(\kappa_*,c_\Gamma)|\log\epsilon|^{1/2}\|f\|_{L^2(\mathbb{T})}. \tag*{(303)} ∥ u ∥ D 1 , 2 ( Ω ϵ ) + ∥ p ∥ L 2 ( Ω ϵ ) ≡ ∥∇ u ∥ L 2 ( Ω ϵ ) + ∥ p ∥ L 2 ( Ω ϵ ) ≤ c ( κ ∗ , c Γ ) ∣ log ϵ ∣ 1/2 ∥ f ∥ L 2 ( T ) . (303)
Recalling the definition (6) of r ∗ = r ∗ ( c Γ , κ ∗ ) r_*=r_*(c_\Gamma,\kappa_*) r ∗ = r ∗ ( c Γ , κ ∗ ) , we may define the curved annular region
O r ∗ = { x ∈ Ω ϵ : ϵ < dist ( x , Γ 0 ) < r ∗ } (304) \mathcal{O}_{r_*}=\{x\in\Omega_\epsilon:\epsilon<\operatorname{dist}(x,\Gamma_0)<r_*\} \tag*{(304)} O r ∗ = { x ∈ Ω ϵ : ϵ < dist ( x , Γ 0 ) < r ∗ } (304)
about Σ ϵ \Sigma_{\epsilon} Σ ϵ . Within O r ∗ \mathcal{O}_{r_*} O r ∗ , the solution ( u , p ) (u,p) ( u , p ) to (11) then satisfies
∥ u ∥ H 1 ( O r ∗ ) + ∥ p ∥ L 2 ( O r ∗ ) ≤ c ( κ ∗ , c Γ , ∣ O r ∗ ∣ ) ∣ log ϵ ∣ 1 / 2 ∥ f ∥ L 2 ( T ) . (305) \lVert u\rVert_{H^1(\mathcal{O}_{r_*})}+\lVert p\rVert_{L^2(\mathcal{O}_{r_*})}\leq c(\kappa_*,c_\Gamma,\lvert\mathcal{O}_{r_*}\rvert)\lvert\log\epsilon\rvert^{1/2}\lVert f\rVert_{L^2(\mathbb{T})}. \tag*{(305)} ∥ u ∥ H 1 ( O r ∗ ) + ∥ p ∥ L 2 ( O r ∗ ) ≤ c ( κ ∗ , c Γ , ∣ O r ∗ ∣) ∣ log ϵ ∣ 1/2 ∥ f ∥ L 2 ( T ) . (305)
We consider the weak form of the slender body BVP (11) in O r ∗ \mathcal{O}_{r_*} O r ∗ , given by
∫ O r ∗ ( 2 E ( u ) : ∇ ξ − p div ( ξ ) ) d x = ∫ T f ( s ) ⋅ ξ ( s ) d s , E ( u ) : = 1 2 ( ∇ u + ∇ u T ) , (306) \int_{\mathcal{O}_{r_*}}\left(2\mathcal{E}(u):\nabla\xi-p\operatorname{div}(\xi)\right)\,dx=\int_{\mathbb{T}}f(s)\cdot\xi(s)\,ds,\qquad\mathcal{E}(u):=\frac{1}{2}\left(\nabla u+\nabla u^{\mathsf{T}}\right), \tag*{(306)} ∫ O r ∗ ( 2 E ( u ) : ∇ ξ − p div ( ξ ) ) d x = ∫ T f ( s ) ⋅ ξ ( s ) d s , E ( u ) := 2 1 ( ∇ u + ∇ u T ) , (306)
for any ξ ∈ H 1 ( O r ∗ ) \xi\in H^1(\mathcal{O}_{r_*}) ξ ∈ H 1 ( O r ∗ ) satisfying ξ ∣ Γ ϵ = ξ ( s ) \left.\xi\right|_{\Gamma_\epsilon}=\xi(s) ξ ∣ Γ ϵ = ξ ( s ) and ξ ∣ ∂ O r ∗ ∖ Γ ϵ = 0 \left.\xi\right|_{\partial\mathcal{O}_{r_*}\setminus\Gamma_\epsilon}=0 ξ ∣ ∂ O r ∗ ∖ Γ ϵ = 0 .
Writing x ∈ O r ∗ x\in\mathcal{O}_{r_*} x ∈ O r ∗ as x = X ( s ) + r e r ( s , θ ) x=X(s)+r e_r(s,\theta) x = X ( s ) + r e r ( s , θ ) , we define a C 1 , β C^{1,\beta} C 1 , β change of variables
Ψ ( x ) = s e z + r e ˉ r ( θ ) = s e z + r cos θ e x + r sin θ e y \Psi(x)=s e_z+r\bar e_r(\theta)=s e_z+r\cos\theta\,e_x+r\sin\theta\,e_y Ψ ( x ) = s e z + r e ˉ r ( θ ) = s e z + r cos θ e x + r sin θ e y
mapping the curved annulus O r ∗ \mathcal{O}_{r_*} O r ∗ to the straight annulus Ψ ( O r ∗ ) \Psi(\mathcal{O}_{r_*}) Ψ ( O r ∗ ) about the straight cylinder C ϵ \mathcal{C}_\epsilon C ϵ . Here ( e z , e x , e y ) (e_z,e_x,e_y) ( e z , e x , e y ) are the standard Cartesian basis vectors. We may calculate
∇ Ψ − 1 = ( 1 − r κ ^ ) e t ⊗ e z + r κ 3 e θ ( s , θ ) ⊗ e z + e ˉ θ ( θ ) ⊗ e θ ( s , θ ) + e ˉ r ( θ ) ⊗ e r ( s , θ ) , (307) \nabla\Psi^{-1}=(1-r\widehat{\kappa})e_t\otimes e_z+r\kappa_3e_\theta(s,\theta)\otimes e_z+\bar e_\theta(\theta)\otimes e_\theta(s,\theta)+\bar e_r(\theta)\otimes e_r(s,\theta), \tag*{(307)} ∇ Ψ − 1 = ( 1 − r κ ) e t ⊗ e z + r κ 3 e θ ( s , θ ) ⊗ e z + e ˉ θ ( θ ) ⊗ e θ ( s , θ ) + e ˉ r ( θ ) ⊗ e r ( s , θ ) , (307)
as well as
∇ Ψ ( ∇ Ψ ) T ∘ Ψ − 1 = 1 + r 2 κ 3 2 ( 1 − r κ ^ ) 2 e z ⊗ e z − r κ 3 1 − r κ ^ ( e ˉ θ ( θ ) ⊗ e z + e z ⊗ e ˉ θ ( θ ) ) + e ˉ θ ( θ ) ⊗ e ˉ θ ( θ ) + e ˉ r ( θ ) ⊗ e ˉ r ( θ ) (308) \nabla\Psi(\nabla\Psi)^{\mathsf{T}}\circ\Psi^{-1}
=\frac{1+r^2\kappa_3^2}{(1-r\widehat{\kappa})^2}e_z\otimes e_z-\frac{r\kappa_3}{1-r\widehat{\kappa}}\left(\bar e_\theta(\theta)\otimes e_z+e_z\otimes\bar e_\theta(\theta)\right)+\bar e_\theta(\theta)\otimes\bar e_\theta(\theta)+\bar e_r(\theta)\otimes\bar e_r(\theta) \tag*{(308)} ∇Ψ ( ∇Ψ ) T ∘ Ψ − 1 = ( 1 − r κ ) 2 1 + r 2 κ 3 2 e z ⊗ e z − 1 − r κ r κ 3 ( e ˉ θ ( θ ) ⊗ e z + e z ⊗ e ˉ θ ( θ ) ) + e ˉ θ ( θ ) ⊗ e ˉ θ ( θ ) + e ˉ r ( θ ) ⊗ e ˉ r ( θ ) (308)
and ∣ det ∇ Ψ ∣ − 1 = 1 − r κ ^ \lvert\det\nabla\Psi\rvert^{-1}=1-r\widehat{\kappa} ∣ det ∇Ψ ∣ − 1 = 1 − r κ . We denote
A ( s , r , θ ) = ( 1 ∣ det ∇ Ψ ∣ ∇ Ψ ( ∇ Ψ ) T ) ∘ Ψ − 1 , B ( s , r , θ ) = ( 1 ∣ det ∇ Ψ ∣ ∇ Ψ ) ∘ Ψ − 1 , (309) A(s,r,\theta)=\left(\frac{1}{\lvert\det\nabla\Psi\rvert}\nabla\Psi(\nabla\Psi)^{\mathsf{T}}\right)\circ\Psi^{-1},\qquad B(s,r,\theta)=\left(\frac{1}{\lvert\det\nabla\Psi\rvert}\nabla\Psi\right)\circ\Psi^{-1}, \tag*{(309)} A ( s , r , θ ) = ( ∣ det ∇Ψ ∣ 1 ∇Ψ ( ∇Ψ ) T ) ∘ Ψ − 1 , B ( s , r , θ ) = ( ∣ det ∇Ψ ∣ 1 ∇Ψ ) ∘ Ψ − 1 , (309)
and note that, using the additional C 3 , α C^{3,\alpha} C 3 , α regularity of X X X , we have
1 c ( κ ∗ , β ) ≤ ∥ A ∥ C 0 , β ( Ψ ( O r ∗ ) ) ≤ c ( κ ∗ , β ) , \frac{1}{c(\kappa_*,\beta)}\leq\lVert A\rVert_{C^{0,\beta}(\Psi(\mathcal{O}_{r_*}))}\leq c(\kappa_*,\beta), c ( κ ∗ , β ) 1 ≤ ∥ A ∥ C 0 , β ( Ψ ( O r ∗ )) ≤ c ( κ ∗ , β ) ,
1 c ( ∥ X s s ∥ C 1 , α ) ≤ ∥ B ∥ C 1 , α ( Ψ ( O r ∗ ) ) ≤ c ( ∥ X s s ∥ C 1 , α ) , (310) \frac{1}{c(\lVert X_{ss}\rVert_{C^{1,\alpha}})}\leq\lVert B\rVert_{C^{1,\alpha}(\Psi(\mathcal{O}_{r_*}))}\leq c(\lVert X_{ss}\rVert_{C^{1,\alpha}}), \tag*{(310)} c (∥ X ss ∥ C 1 , α ) 1 ≤ ∥ B ∥ C 1 , α ( Ψ ( O r ∗ )) ≤ c (∥ X ss ∥ C 1 , α ) , (310)
with smooth dependence on θ \theta θ . Here we have used that, using (3) , we may write
( κ 1 ) s = X s s s ⋅ e n 1 + κ 2 2 , ( κ 2 ) s = X s s s ⋅ e n 2 + κ 1 2 , (\kappa_1)_s=X_{sss}\cdot e_{n_1}+\kappa_2^2,\qquad(\kappa_2)_s=X_{sss}\cdot e_{n_2}+\kappa_1^2, ( κ 1 ) s = X sss ⋅ e n 1 + κ 2 2 , ( κ 2 ) s = X sss ⋅ e n 2 + κ 1 2 ,
from which we may obtain ∥ κ 1 ∥ C 1 , α , ∥ κ 2 ∥ C 1 , α ≤ c ∥ X s s ∥ C 1 , α \lVert\kappa_1\rVert_{C^{1,\alpha}},\ \lVert\kappa_2\rVert_{C^{1,\alpha}}\leq c\lVert X_{ss}\rVert_{C^{1,\alpha}} ∥ κ 1 ∥ C 1 , α , ∥ κ 2 ∥ C 1 , α ≤ c ∥ X ss ∥ C 1 , α .
Let ψ = ξ ∘ Ψ − 1 \psi=\xi\circ\Psi^{-1} ψ = ξ ∘ Ψ − 1 and note that along the fiber surface, ψ ∣ y ∈ Ψ ( Γ ϵ ) = ξ ∣ x ∈ Γ ϵ \left.\psi\right|_{y\in\Psi(\Gamma_\epsilon)}=\left.\xi\right|_{x\in\Gamma_\epsilon} ψ ∣ y ∈ Ψ ( Γ ϵ ) = ξ ∣ x ∈ Γ ϵ is still θ \theta θ -independent. Letting u ˉ = u ∘ Ψ − 1 \bar u=u\circ\Psi^{-1} u ˉ = u ∘ Ψ − 1 and p ˉ = p ∘ Ψ − 1 \bar p=p\circ\Psi^{-1} p ˉ = p ∘ Ψ − 1 , we may rewrite (304) in straightened variables y = Ψ ( x ) y=\Psi(x) y = Ψ ( x ) as
∫ Ψ ( O r ∗ ) ( ( ∇ u ˉ A + B T ∇ u ˉ T ∇ Ψ T ) : ∇ ψ − p ˉ div ( B ψ ) ) d y = ∫ T f ( s ) ⋅ ψ ( s ) d s . (311) \int_{\Psi(\mathcal{O}_{r_*})}\left((\nabla\bar u A+B^{\mathsf{T}}\nabla\bar u^{\mathsf{T}}\nabla\Psi^{\mathsf{T}}):\nabla\psi-\bar p\operatorname{div}(B\psi)\right)\,dy=\int_{\mathbb{T}}f(s)\cdot\psi(s)\,ds. \tag*{(311)} ∫ Ψ ( O r ∗ ) ( ( ∇ u ˉ A + B T ∇ u ˉ T ∇ Ψ T ) : ∇ ψ − p ˉ div ( B ψ ) ) d y = ∫ T f ( s ) ⋅ ψ ( s ) d s . (311)
We will use the formulation (311) along with the analogous Campanato-type argument used in [59 ] to show that u = u ˉ ∘ Ψ u=\bar u\circ\Psi u = u ˉ ∘ Ψ satisfies Lemma 1.14 along Γ ϵ \Gamma_\epsilon Γ ϵ .
We will make use of a Campanato-type function space tailored to the slender body BVP, which was defined in [59 ] following the more classical constructions in [28 , 48 ] . Given ρ ≤ r ∗ 2 \rho\leq\frac{r_*}{2} ρ ≤ 2 r ∗ and s 0 ∈ T s_0\in\mathbb{T} s 0 ∈ T , we define the following annular region about the straight cylinder C ϵ \mathcal{C}_\epsilon C ϵ :
A ρ ( s 0 ) = { y = s e z + r e ˉ r ( θ ) ∈ Ψ ( O r ∗ ) : s 0 − ρ < s < s 0 + ρ , ϵ ≤ r < ρ + ϵ , 0 ≤ θ < 2 π } . A_\rho(s_0)=\left\{y=s e_z+r\bar e_r(\theta)\in\Psi(\mathcal{O}_{r_*}):s_0-\rho<s<s_0+\rho,\ \epsilon\leq r<\rho+\epsilon,\ 0\leq\theta<2\pi\right\}. A ρ ( s 0 ) = { y = s e z + r e ˉ r ( θ ) ∈ Ψ ( O r ∗ ) : s 0 − ρ < s < s 0 + ρ , ϵ ≤ r < ρ + ϵ , 0 ≤ θ < 2 π } .
We emphasize that the volume of the region A ρ ( s 0 ) A_\rho(s_0) A ρ ( s 0 ) scales like ρ 2 \rho^2 ρ 2 as ρ → 0 \rho\to0 ρ → 0 . Given g ( y ) \mathbf{g}(\mathbf{y}) g ( y ) , y ∈ A ρ ( s 0 ) \mathbf{y} \in A_\rho(s_0) y ∈ A ρ ( s 0 ) , we use the notation g s 0 , ρ \mathbf{g}_{s_0,\rho} g s 0 , ρ to denote the mean of g \mathbf{g} g over A ρ ( s 0 ) A_\rho(s_0) A ρ ( s 0 ) :
\mathbf{g}_{s_0,\rho} = \fint_{A_\rho(s_0)} \mathbf{g}(\mathbf{y})\,\mathrm{d}y . \tag*{(312)}
For 0 < α < 1 0 < \alpha< 1 0 < α < 1 , we define the seminorm [ ⋅ ] A 2 , α [\cdot]_{A^{2,\alpha}} [ ⋅ ] A 2 , α and full norm ∥ ⋅ ∥ A 2 , α \|\cdot\|_{A^{2,\alpha}} ∥ ⋅ ∥ A 2 , α by
[\mathbf{g}]_{A^{2,\alpha}} = \sup_{\substack{s_0\in\mathbb{T},\,0<\rho\le r_*/2}} \rho^{-\alpha}\left(\fint_{A_\rho(s_0)}\lvert\mathbf{g}-\mathbf{g}_{s_0,\rho}\rvert^2\,\mathrm{d}y\right)^{1/2}, \tag*{(313)}
∥ g ∥ A 2 , α = ∥ g ∥ L 2 ( Ψ ( O r ∗ ) ) + [ g ] A 2 , α . \|\mathbf{g}\|_{A^{2,\alpha}} = \|\mathbf{g}\|_{L^2(\Psi(\mathcal{O}_{r_*}))} + [\mathbf{g}]_{A^{2,\alpha}} . ∥ g ∥ A 2 , α = ∥ g ∥ L 2 ( Ψ ( O r ∗ )) + [ g ] A 2 , α .
We may relate the function space defined by the norm ∥ ⋅ ∥ A 2 , α \|\cdot\|_{A^{2,\alpha}} ∥ ⋅ ∥ A 2 , α to the Hölder space C 0 , α C^{0,\alpha} C 0 , α about the straight filament C ϵ \mathcal{C}_\epsilon C ϵ in the following way.
Lemma 6.2 (Campanato-type norm bounds). Consider g \mathbf{g} g in Ψ ( O r ∗ ) \Psi(\mathcal{O}_{r_*}) Ψ ( O r ∗ ) satisfying g ∣ C ϵ = g ( s ) \mathbf{g}|_{\mathcal{C}_\epsilon}=\mathbf{g}(s) g ∣ C ϵ = g ( s ) , a function of arclength only. Then
∥ g ∥ C 0 , α ( T ) ≤ c ∥ g ∥ A 2 , α ≤ c ∥ g ∥ C 0 , α ( Ψ ( O r ∗ ) ) . (314) \|\mathbf{g}\|_{C^{0,\alpha}(\mathbb{T})} \le c\|\mathbf{g}\|_{A^{2,\alpha}} \le c\|\mathbf{g}\|_{C^{0,\alpha}(\Psi(\mathcal{O}_{r_*}))} . \tag*{(314)} ∥ g ∥ C 0 , α ( T ) ≤ c ∥ g ∥ A 2 , α ≤ c ∥ g ∥ C 0 , α ( Ψ ( O r ∗ )) . (314) Here the left-hand-side C 0 , α C^{0,\alpha} C 0 , α norm is taken over the filament centerline (s ∈ T s \in\mathbb{T} s ∈ T ), while the right-hand-side C 0 , α C^{0,\alpha} C 0 , α norm is over the entire annular region Ψ ( O r ∗ ) \Psi(\mathcal{O}_{r_*}) Ψ ( O r ∗ ) .
The proof of Lemma 6.2 is given in [59 Lemma 5.2] .
We next consider a frozen-coefficients version of the slender body BVP (311) . Taking 0 < R ≤ r ∗ 2 0<R\le\frac{r_*}{2} 0 < R ≤ 2 r ∗ and s 0 ∈ T s_0\in\mathbb{T} s 0 ∈ T , we consider the annular region A R ( s 0 ) A_R(s_0) A R ( s 0 ) about C ϵ \mathcal{C}_\epsilon C ϵ . Let A 0 ( θ ) = A ( s 0 , ϵ , θ ) \mathbf{A}_0(\theta)=\mathbf{A}(s_0,\epsilon,\theta) A 0 ( θ ) = A ( s 0 , ϵ , θ ) , B 0 ( θ ) = B ( s 0 , ϵ , θ ) \mathbf{B}_0(\theta)=\mathbf{B}(s_0,\epsilon,\theta) B 0 ( θ ) = B ( s 0 , ϵ , θ ) , and ∇ Ψ 0 ( θ ) = ∇ Ψ ( s 0 , ϵ , θ ) \nabla\Psi_0(\theta)=\nabla\Psi(s_0,\epsilon,\theta) ∇ Ψ 0 ( θ ) = ∇Ψ ( s 0 , ϵ , θ ) , and note that A 0 ( θ ) \mathbf{A}_0(\theta) A 0 ( θ ) , B 0 ( θ ) \mathbf{B}_0(\theta) B 0 ( θ ) , and ∇ Ψ 0 ( θ ) \nabla\Psi_0(\theta) ∇ Ψ 0 ( θ ) are smooth functions of θ \theta θ satisfying 1 c ( κ ∗ ) ≤ ∣ ∂ θ ℓ A 0 ∣ , ∣ ∂ θ ℓ B 0 ∣ , ∣ ∂ θ ℓ ∇ Ψ 0 ∣ ≤ c ( κ ∗ ) \frac{1}{c(\kappa_*)}\le\lvert\partial_\theta^\ell\mathbf{A}_0\rvert,\lvert\partial_\theta^\ell\mathbf{B}_0\rvert,\lvert\partial_\theta^\ell\nabla\Psi_0\rvert\le c(\kappa_*) c ( κ ∗ ) 1 ≤ ∣ ∂ θ ℓ A 0 ∣ , ∣ ∂ θ ℓ B 0 ∣ , ∣ ∂ θ ℓ ∇ Ψ 0 ∣ ≤ c ( κ ∗ ) for c ( κ ∗ ) > 0 c(\kappa_*)>0 c ( κ ∗ ) > 0 and 0 ≤ ℓ ∈ Z 0\le\ell\in\mathbb{Z} 0 ≤ ℓ ∈ Z . We consider ( h , q ) (\mathbf{h},q) ( h , q ) belonging to the space
{ ( h , q ) ∈ H 1 × L 2 ( A R ( s 0 ) ) : div ( B 0 ( θ ) h ) = 0 , h ∣ C ϵ = h ( s ) , h ∣ ∂ A R ∖ C ϵ = u , ∫ A R q d y = 0 } \left\{(\mathbf{h},q)\in H^1\times L^2(A_R(s_0)):\ \operatorname{div}(\mathbf{B}_0(\theta)\mathbf{h})=0,\ \mathbf{h}|_{\mathcal{C}_\epsilon}=\mathbf{h}(s),\ \mathbf{h}|_{\partial A_R\setminus\mathcal{C}_\epsilon}=\mathbf{u},\ \int_{A_R}q\,\mathrm{d}y=0\right\} { ( h , q ) ∈ H 1 × L 2 ( A R ( s 0 )) : div ( B 0 ( θ ) h ) = 0 , h ∣ C ϵ = h ( s ) , h ∣ ∂ A R ∖ C ϵ = u , ∫ A R q d y = 0 }
and satisfying
∫ A R ( s 0 ) ( ( ∇ h A 0 ( θ ) + B 0 ( θ ) T ∇ h T ∇ Ψ 0 ( θ ) T ) : ∇ ψ − q div ( B 0 ( θ ) ψ ) ) d y = f ( s 0 ) ⋅ ∫ ∣ s − s 0 ∣ ≤ R ψ ( s ) d s (315) \int_{A_R(s_0)}\left((\nabla\mathbf{h}\mathbf{A}_0(\theta)+\mathbf{B}_0(\theta)^{T}\nabla\mathbf{h}^{T}\nabla\Psi_0(\theta)^{T}):\nabla\boldsymbol{\psi}-q\operatorname{div}(\mathbf{B}_0(\theta)\boldsymbol{\psi})\right)\,\mathrm{d}y
= \mathbf{f}(s_0)\cdot\int_{\lvert s-s_0\rvert\le R}\boldsymbol{\psi}(s)\,\mathrm{d}s \tag*{(315)} ∫ A R ( s 0 ) ( ( ∇ h A 0 ( θ ) + B 0 ( θ ) T ∇ h T ∇ Ψ 0 ( θ ) T ) : ∇ ψ − q div ( B 0 ( θ ) ψ ) ) d y = f ( s 0 ) ⋅ ∫ ∣ s − s 0 ∣ ≤ R ψ ( s ) d s (315)
for any ψ ∈ H 1 ( A R ( s 0 ) ) \boldsymbol{\psi}\in H^1(A_R(s_0)) ψ ∈ H 1 ( A R ( s 0 )) with ψ ∣ C ϵ = ψ ( s ) \boldsymbol{\psi}|_{\mathcal{C}_\epsilon}=\boldsymbol{\psi}(s) ψ ∣ C ϵ = ψ ( s ) and ψ ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 \boldsymbol{\psi}|_{\partial A_R(s_0)\setminus\mathcal{C}_\epsilon}=0 ψ ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 . We may show that ( h , q ) (\mathbf{h},q) ( h , q ) satisfies the following proposition.
Proposition 6.3 (Higher regularity for ( h , q ) (\mathbf{h},q) ( h , q ) ). For m ∈ N m\in\mathbb{N} m ∈ N and any 0 < ρ < R 0<\rho<R 0 < ρ < R , the solution ( h , q ) (\mathbf{h},q) ( h , q ) to (315) satisfies
∥ h ∥ H ˙ m + 1 ( A ρ ( s 0 ) ) + ∥ q ∥ H ˙ m ( A ρ ( s 0 ) ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − ρ ) m ( ∥ h ∥ H 1 ( A R ( s 0 ) ) + ∥ q ∥ L 2 ( A R ( s 0 ) ) ) . (316) \|\mathbf{h}\|_{\dot{H}^{m+1}(A_\rho(s_0))}+\|q\|_{\dot{H}^{m}(A_\rho(s_0))}\le\frac{c(\epsilon,\kappa_*,c_\Gamma)}{(R-\rho)^m}\left(\|\mathbf{h}\|_{H^1(A_R(s_0))}+\|q\|_{L^2(A_R(s_0))}\right) . \tag*{(316)} ∥ h ∥ H ˙ m + 1 ( A ρ ( s 0 )) + ∥ q ∥ H ˙ m ( A ρ ( s 0 )) ≤ ( R − ρ ) m c ( ϵ , κ ∗ , c Γ ) ( ∥ h ∥ H 1 ( A R ( s 0 )) + ∥ q ∥ L 2 ( A R ( s 0 )) ) . (316)
Proof. Fixing R ′ < R R'<R R ′ < R , we begin by defining a cutoff function
η ( x ) = { 1 , x ∈ A R ′ ( s 0 ) , 0 , x ∉ A R ( s 0 ) , R ′ < R , (317) \eta(\mathbf{x})=
\begin{cases}
1, & \mathbf{x}\in A_{R'}(s_0),\\
0, & \mathbf{x}\notin A_R(s_0),
\end{cases}
\qquad R'<R , \tag*{(317)} η ( x ) = { 1 , 0 , x ∈ A R ′ ( s 0 ) , x ∈ / A R ( s 0 ) , R ′ < R , (317)
with ∣ ∇ η ∣ ≤ 1 R − R ′ |\nabla\eta|\leq\frac{1}{R-R'} ∣∇ η ∣ ≤ R − R ′ 1 . We would like to use ∂ k ( η 2 ∂ k h ) \partial_k(\eta^2\partial_k h) ∂ k ( η 2 ∂ k h ) , k = s , θ k=s,\theta k = s , θ as test functions in (315) (this may be justified rigorously using finite differences), but first it will be convenient to correct for the divergence. For k = s , θ k=s,\theta k = s , θ , we consider d k ∈ H 0 1 ( A R ( s 0 ) ) d_k\in H^1_0(A_R(s_0)) d k ∈ H 0 1 ( A R ( s 0 )) satisfying
div ( B 0 d k ) = div ( B 0 η 2 ∂ k h ) in A R ( s 0 ) \operatorname{div}(B_0d_k)=\operatorname{div}(B_0\eta^2\partial_kh)\quad\text{in }A_R(s_0) div ( B 0 d k ) = div ( B 0 η 2 ∂ k h ) in A R ( s 0 )
∥ ∇ d k ∥ L 2 ( A R ( s 0 ) ) ≤ c ( κ ∗ , c Γ ) ∥ div ( B 0 η 2 ∂ k h ) ∥ L 2 ( A R ( s 0 ) ) . (318) \|\nabla d_k\|_{L^2(A_R(s_0))}\leq c(\kappa_*,c_\Gamma)\|\operatorname{div}(B_0\eta^2\partial_kh)\|_{L^2(A_R(s_0))}. \tag*{(318)} ∥∇ d k ∥ L 2 ( A R ( s 0 )) ≤ c ( κ ∗ , c Γ ) ∥ div ( B 0 η 2 ∂ k h ) ∥ L 2 ( A R ( s 0 )) . (318)
Such a d k d_k d k exists by [24 Section III.3] , since B 0 B_0 B 0 is invertible. In addition, the constant appearing in the bound (318) is independent of ϵ \epsilon ϵ due to [52 Appendix A.2.5] , although we will not keep track of ϵ \epsilon ϵ -dependence throughout. We further note that, since div ( B 0 h ) = 0 \operatorname{div}(B_0h)=0 div ( B 0 h ) = 0 in A R ( s 0 ) A_R(s_0) A R ( s 0 ) , and div commutes with ∂ s \partial_s ∂ s and ∂ θ \partial_\theta ∂ θ in cylindrical coordinates, we have
∥ div ( B 0 η 2 ∂ k h ) ∥ L 2 ( A R ( s 0 ) ) ≤ ∥ div ( B 0 η 2 ) ∂ k h ∥ L 2 ( A R ( s 0 ) ) + ∥ η 2 div ( ( ∂ k B 0 ) h ) ∥ L 2 ( A R ( s 0 ) ) + ∥ η 2 ∂ k div ( B 0 h ) ∥ L 2 ( A R ( s 0 ) ) ≤ c ( κ ∗ ) ∥ η h ∥ H 1 ( A R ( s 0 ) ) , \begin{aligned}
\|\operatorname{div}(B_0\eta^2\partial_kh)\|_{L^2(A_R(s_0))}
&\leq\|\operatorname{div}(B_0\eta^2)\partial_kh\|_{L^2(A_R(s_0))}+\|\eta^2\operatorname{div}((\partial_kB_0)h)\|_{L^2(A_R(s_0))}\\
&\quad+\|\eta^2\partial_k\operatorname{div}(B_0h)\|_{L^2(A_R(s_0))}\\
&\leq c(\kappa_*)\|\eta h\|_{H^1(A_R(s_0))},
\end{aligned} ∥ div ( B 0 η 2 ∂ k h ) ∥ L 2 ( A R ( s 0 )) ≤ ∥ div ( B 0 η 2 ) ∂ k h ∥ L 2 ( A R ( s 0 )) + ∥ η 2 div (( ∂ k B 0 ) h ) ∥ L 2 ( A R ( s 0 )) + ∥ η 2 ∂ k div ( B 0 h ) ∥ L 2 ( A R ( s 0 )) ≤ c ( κ ∗ ) ∥ η h ∥ H 1 ( A R ( s 0 )) ,
so, in particular, for k = s , θ k=s,\theta k = s , θ , we have
∥ ∇ d k ∥ L 2 ( A R ( s 0 ) ) ≤ c ( κ ∗ , c Γ ) ∥ η h ∥ H 1 ( A R ( s 0 ) ) . \|\nabla d_k\|_{L^2(A_R(s_0))}\leq c(\kappa_*,c_\Gamma)\|\eta h\|_{H^1(A_R(s_0))}. ∥∇ d k ∥ L 2 ( A R ( s 0 )) ≤ c ( κ ∗ , c Γ ) ∥ η h ∥ H 1 ( A R ( s 0 )) .
We then take ∂ k ( η 2 ∂ k h − d k ) \partial_k(\eta^2\partial_kh-d_k) ∂ k ( η 2 ∂ k h − d k ) , k = s , θ k=s,\theta k = s , θ , as a test function in (315) . We have
∫ A R ( s 0 ) ( ( ( ∇ h A 0 + B 0 T ∇ h T ∇ Ψ 0 T ) : ∇ ( ∂ k ( η 2 ∂ k h − d k ) ) − q div ( B 0 ∂ k ( η 2 ∂ k h − d k ) ) ) d y = f ( s 0 ) ⋅ ∫ ∣ s − s 0 ∣ ≤ R ∂ k η 2 ∂ k h ( s ) d s = 0. (319) \begin{aligned}
\int_{A_R(s_0)}\left(\left((\nabla hA_0+B_0^{\mathsf T}\nabla h^{\mathsf T}\nabla\Psi_0^{\mathsf T}):\nabla(\partial_k(\eta^2\partial_kh-d_k)\right)-q\operatorname{div}\left(B_0\partial_k(\eta^2\partial_kh-d_k)\right)\right)\,dy \\
&=f(s_0)\cdot\int_{|s-s_0|\leq R}\partial_k\eta^2\partial_kh(s)\,ds=0. \tag*{(319)}
\end{aligned} ∫ A R ( s 0 ) ( ( ( ∇ h A 0 + B 0 T ∇ h T ∇ Ψ 0 T ) : ∇ ( ∂ k ( η 2 ∂ k h − d k ) ) − q div ( B 0 ∂ k ( η 2 ∂ k h − d k ) ) ) d y = f ( s 0 ) ⋅ ∫ ∣ s − s 0 ∣ ≤ R ∂ k η 2 ∂ k h ( s ) d s = 0. (319)
Before proceeding, we note that, using the expressions (307) –(309) for A A A , B B B , and ∇ Ψ \nabla\Psi ∇Ψ , for any v ∈ L 2 ( A R ( s 0 ) ) v\in L^2(A_R(s_0)) v ∈ L 2 ( A R ( s 0 )) , we may write
( ∇ v A 0 + B 0 T ∇ v T ∇ Ψ 0 T ) = ( ∇ v + ∇ v T ) − E 0 ( ∇ v ) , (\nabla vA_0+B_0^{\mathsf T}\nabla v^{\mathsf T}\nabla\Psi_0^{\mathsf T})=(\nabla v+\nabla v^{\mathsf T})-E_0(\nabla v), ( ∇ v A 0 + B 0 T ∇ v T ∇ Ψ 0 T ) = ( ∇ v + ∇ v T ) − E 0 ( ∇ v ) ,
∣ E 0 ∣ ≤ 2 max { ∣ ϵ κ ^ b ( s 0 , θ ) ∣ , ∣ ϵ κ 3 ∣ , ∣ ϵ κ ^ b ( s 0 , θ ) ∣ + ϵ 2 κ 3 2 ∣ 1 − ϵ κ ^ b ( s 0 , θ ) ∣ } ∣ ∇ v ∣ ≤ ϵ c ( κ ∗ ) ∣ ∇ v ∣ . (320) |E_0|\leq2\max\left\{|\epsilon\widehat{\kappa}_b(s_0,\theta)|,\ |\epsilon\kappa_3|,\ \frac{|\epsilon\widehat{\kappa}_b(s_0,\theta)|+\epsilon^2\kappa_3^2}{|1-\epsilon\widehat{\kappa}_b(s_0,\theta)|}\right\}|\nabla v|\leq\epsilon c(\kappa_*)|\nabla v|. \tag*{(320)} ∣ E 0 ∣ ≤ 2 max { ∣ ϵ κ b ( s 0 , θ ) ∣ , ∣ ϵ κ 3 ∣ , ∣1 − ϵ κ b ( s 0 , θ ) ∣ ∣ ϵ κ b ( s 0 , θ ) ∣ + ϵ 2 κ 3 2 } ∣∇ v ∣ ≤ ϵc ( κ ∗ ) ∣∇ v ∣. (320)
Furthermore, for any v v v satisfying v = 0 v=0 v = 0 on ∂ A R ( s 0 ) ∖ C ϵ \partial A_R(s_0)\setminus C_\epsilon ∂ A R ( s 0 ) ∖ C ϵ , we may use the homogeneous Korn inequality in A R ( s 0 ) A_R(s_0) A R ( s 0 ) to bound
∫ A R ( s 0 ) ∣ ∇ v ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ E ( v ) ∣ 2 d y , E ( v ) = 1 2 ( ∇ v + ∇ v T ) , (321) \int_{A_R(s_0)}|\nabla v|^2\,dy\leq c(c_\Gamma,\kappa_*)\int_{A_R(s_0)}|\mathcal{E}(v)|^2\,dy,\qquad\mathcal{E}(v)=\frac{1}{2}(\nabla v+\nabla v^{\mathsf T}), \tag*{(321)} ∫ A R ( s 0 ) ∣∇ v ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ E ( v ) ∣ 2 d y , E ( v ) = 2 1 ( ∇ v + ∇ v T ) , (321)
where, by [52 Lemma 2.6] , c ( c Γ , κ ∗ ) c(c_\Gamma,\kappa_*) c ( c Γ , κ ∗ ) is independent of ϵ \epsilon ϵ .
Applying the above bounds, we may estimate η ∇ ∂ k h \eta\nabla\partial_kh η ∇ ∂ k h , k = s , θ k=s,\theta k = s , θ , in A R ( s 0 ) A_R(s_0) A R ( s 0 ) as
∫ A R ( s 0 ) η 2 ∣ ∇ ∂ k h ∣ 2 d y ≤ ∫ A R ( s 0 ) ∣ ∇ ( η ∂ k h ) ∣ 2 d y + ∫ A R ( s 0 ) ∣ ∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ E ( η ∂ k h ) ∣ 2 d y + ∫ A R ( s 0 ) ∣ ∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 ∣ E ( ∂ k h ) ∣ 2 d y + c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ ∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y = c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 E ( ∂ k h ) : ∇ ( ∂ k h ) d y + c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ ∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y = c ( Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 ( ∇ ( ∂ k h ) A 0 + B 0 T ∇ ( ∂ k h ) ∇ Ψ 0 T ) : ∇ ( ∂ k h ) d y + c ( Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 E 0 ( ∂ k h ) : ∇ ( ∂ k h ) d y + c ( Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ ∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y . \begin{aligned}
\int_{A_R(s_0)}\eta^2|\nabla\partial_kh|^2\,dy
&\leq\int_{A_R(s_0)}|\nabla(\eta\partial_kh)|^2\,dy+\int_{A_R(s_0)}|\nabla\eta|^2|\partial_kh|^2\,dy\\
&\leq c(c_\Gamma,\kappa_*)\int_{A_R(s_0)}|\mathcal{E}(\eta\partial_kh)|^2\,dy+\int_{A_R(s_0)}|\nabla\eta|^2|\partial_kh|^2\,dy\\
&\leq c(c_\Gamma,\kappa_*)\int_{A_R(s_0)}\eta^2|\mathcal{E}(\partial_kh)|^2\,dy+c(c_\Gamma,\kappa_*)\int_{A_R(s_0)}|\nabla\eta|^2|\partial_kh|^2\,dy\\
&=c(c_\Gamma,\kappa_*)\int_{A_R(s_0)}\eta^2\mathcal{E}(\partial_kh):\nabla(\partial_kh)\,dy+c(c_\Gamma,\kappa_*)\int_{A_R(s_0)}|\nabla\eta|^2|\partial_kh|^2\,dy \\
&= c(\Gamma, \kappa_*) \int_{A_R(s_0)} \eta^2 \left(\nabla(\partial_k h) A_0 + B_0^{\mathsf T}\nabla(\partial_k h)\nabla\Psi_0^{\mathsf T}\right) : \nabla(\partial_k h)\,dy
+ c(\Gamma, \kappa_*) \int_{A_R(s_0)} \eta^2 E_0(\partial_k h) : \nabla(\partial_k h)\,dy
+ c(\Gamma, \kappa_*) \int_{A_R(s_0)} |\nabla\eta|^2 |\partial_k h|^2\,dy .
\end{aligned} ∫ A R ( s 0 ) η 2 ∣∇ ∂ k h ∣ 2 d y ≤ ∫ A R ( s 0 ) ∣∇ ( η ∂ k h ) ∣ 2 d y + ∫ A R ( s 0 ) ∣∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ E ( η ∂ k h ) ∣ 2 d y + ∫ A R ( s 0 ) ∣∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 ∣ E ( ∂ k h ) ∣ 2 d y + c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y = c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 E ( ∂ k h ) : ∇ ( ∂ k h ) d y + c ( c Γ , κ ∗ ) ∫ A R ( s 0 ) ∣∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y = c ( Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 ( ∇ ( ∂ k h ) A 0 + B 0 T ∇ ( ∂ k h ) ∇ Ψ 0 T ) : ∇ ( ∂ k h ) d y + c ( Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 E 0 ( ∂ k h ) : ∇ ( ∂ k h ) d y + c ( Γ , κ ∗ ) ∫ A R ( s 0 ) ∣∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y .
For ϵ \epsilon ϵ sufficiently small (depending on c Γ c_\Gamma c Γ and κ ∗ \kappa_* κ ∗ ), the middle term may be absorbed into the left hand side to yield
∫ A R ( s 0 ) η 2 ∣ ∇ ∂ k h ∣ 2 d y ≤ c ( Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 ( ∇ ( ∂ k h ) A 0 + B 0 T ∇ ( ∂ k h ) ∇ Ψ 0 T ) : ∇ ( ∂ k h ) d y (322) \int_{A_R(s_0)} \eta^2 |\nabla\partial_k h|^2\,dy
\le c(\Gamma, \kappa_*) \int_{A_R(s_0)} \eta^2 \left(\nabla(\partial_k h) A_0 + B_0^{\mathsf T}\nabla(\partial_k h)\nabla\Psi_0^{\mathsf T}\right) : \nabla(\partial_k h)\,dy \tag*{(322)} ∫ A R ( s 0 ) η 2 ∣∇ ∂ k h ∣ 2 d y ≤ c ( Γ , κ ∗ ) ∫ A R ( s 0 ) η 2 ( ∇ ( ∂ k h ) A 0 + B 0 T ∇ ( ∂ k h ) ∇ Ψ 0 T ) : ∇ ( ∂ k h ) d y (322)
+ c ( Γ , κ ∗ ) ∫ A R ( s 0 ) ∣ ∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y . \qquad+ c(\Gamma, \kappa_*) \int_{A_R(s_0)} |\nabla\eta|^2 |\partial_k h|^2\,dy . + c ( Γ , κ ∗ ) ∫ A R ( s 0 ) ∣∇ η ∣ 2 ∣ ∂ k h ∣ 2 d y .
To estimate η ∇ ∂ k h \eta\nabla\partial_k h η ∇ ∂ k h in A R ( s 0 ) A_R(s_0) A R ( s 0 ) , we will thus require a bound for the integral J 0 J_0 J 0 , which we may obtain from (317) after some rearrangement.
Again, since we are using cylindrical coordinates, ∂ s \partial_s ∂ s and ∂ θ \partial_\theta ∂ θ commute with ∇ \nabla ∇ and div \operatorname{div} div , but we pick up additional terms in moving ∂ θ \partial_\theta ∂ θ throughout the coefficients A 0 ( θ ) A_0(\theta) A 0 ( θ ) , B 0 ( θ ) B_0(\theta) B 0 ( θ ) , and ∇ Ψ 0 ( θ ) \nabla\Psi_0(\theta) ∇ Ψ 0 ( θ ) . After commuting, integrating by parts, and using that div ( B 0 ( η 2 ∂ k h − d k ) ) = 0 \operatorname{div}\left(B_0(\eta^2\partial_k h-d_k)\right)=0 div ( B 0 ( η 2 ∂ k h − d k ) ) = 0 for k = s , θ k=s,\theta k = s , θ , we obtain
∣ J 0 ∣ ≤ ∫ A R ( s 0 ) ( ( ∣ A 0 ∣ + ∣ B 0 ∣ ∣ ∇ Ψ 0 ∣ ) ∣ ∇ ∂ k h ∣ ( ∣ ∇ ( η 2 ) ∣ ∣ ∂ k h ∣ + ∣ ∇ d k ∣ ) + ( ∣ ∂ k A 0 ∣ + ∣ ∂ k B 0 ∣ ∣ ∇ Ψ 0 ∣ + ∣ B 0 ∣ ∣ ∂ k ∇ Ψ 0 ∣ ) ∣ ∇ h ∣ ( η 2 ∣ ∇ ∂ k h ∣ + ∣ ∇ ( η 2 ) ∣ ∣ ∂ k h ∣ + ∣ ∇ d k ∣ ) ) d y + ∫ A R ( s 0 ) ∣ q ∣ ( ∣ ∇ ∂ k B 0 ∣ ( η 2 ∣ ∂ k h ∣ + ∣ d k ∣ ) + ∣ ∂ k B 0 ∣ ( η 2 ∣ ∇ ∂ k h ∣ + ∣ ∇ ( η 2 ) ∣ ∣ ∂ k h ∣ + ∣ ∇ d k ∣ ) ) d y ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) ( ∥ η ∇ ∂ k h ∥ L 2 ( A R ) + ∥ η h ∥ H 1 ( A R ) ) , \begin{aligned}
|J_0| \le{}& \int_{A_R(s_0)}
\Bigl((|A_0|+|B_0|\,|\nabla\Psi_0|)|\nabla\partial_k h|
\bigl(|\nabla(\eta^2)|\,|\partial_k h|+|\nabla d_k|\bigr) \\
&\qquad+ (|\partial_k A_0|+|\partial_k B_0|\,|\nabla\Psi_0|+|B_0|\,|\partial_k\nabla\Psi_0|)|\nabla h|
\bigl(\eta^2|\nabla\partial_k h|+|\nabla(\eta^2)|\,|\partial_k h|+|\nabla d_k|\bigr)\Bigr)\,dy \\
&\quad+ \int_{A_R(s_0)} |q|\Bigl(|\nabla\partial_k B_0|
\bigl(\eta^2|\partial_k h|+|d_k|\bigr)
+|\partial_k B_0|\bigl(\eta^2|\nabla\partial_k h|+|\nabla(\eta^2)|\,|\partial_k h|+|\nabla d_k|\bigr)\Bigr)\,dy \\
\le{}& c(\epsilon,\kappa_*,c_\Gamma)(R-R')^{-1}
\left(\|h\|_{H^1(A_R)}+\|q\|_{L^2(A_R)}\right)
\left(\|\eta\nabla\partial_k h\|_{L^2(A_R)}+\|\eta h\|_{H^1(A_R)}\right),
\end{aligned} ∣ J 0 ∣ ≤ ≤ ∫ A R ( s 0 ) ( ( ∣ A 0 ∣ + ∣ B 0 ∣ ∣∇ Ψ 0 ∣ ) ∣∇ ∂ k h ∣ ( ∣∇ ( η 2 ) ∣ ∣ ∂ k h ∣ + ∣∇ d k ∣ ) + ( ∣ ∂ k A 0 ∣ + ∣ ∂ k B 0 ∣ ∣∇ Ψ 0 ∣ + ∣ B 0 ∣ ∣ ∂ k ∇ Ψ 0 ∣ ) ∣∇ h ∣ ( η 2 ∣∇ ∂ k h ∣ + ∣∇ ( η 2 ) ∣ ∣ ∂ k h ∣ + ∣∇ d k ∣ ) ) d y + ∫ A R ( s 0 ) ∣ q ∣ ( ∣∇ ∂ k B 0 ∣ ( η 2 ∣ ∂ k h ∣ + ∣ d k ∣ ) + ∣ ∂ k B 0 ∣ ( η 2 ∣∇ ∂ k h ∣ + ∣∇ ( η 2 ) ∣ ∣ ∂ k h ∣ + ∣∇ d k ∣ ) ) d y c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) ( ∥ η ∇ ∂ k h ∥ L 2 ( A R ) + ∥ η h ∥ H 1 ( A R ) ) ,
since ∣ ∇ η ∣ ≲ ( R − R ′ ) − 1 |\nabla\eta|\lesssim(R-R')^{-1} ∣∇ η ∣ ≲ ( R − R ′ ) − 1 . Combining with (320) and using Young’s inequality, over the smaller annulus A R ′ ( s 0 ) A_{R'}(s_0) A R ′ ( s 0 ) , we obtain the bounds
∥ ∇ ∂ k h ∥ L 2 ( A R ′ ( s 0 ) ) ≤ c ( ϵ , c Γ , κ ∗ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ( s 0 ) ) + ∥ q ∥ L 2 ( A R ( s 0 ) ) ) , k = s , θ . (323) \|\nabla\partial_k h\|_{L^2(A_{R'}(s_0))}
\le c(\epsilon,c_\Gamma,\kappa_*)(R-R')^{-1}
\left(\|h\|_{H^1(A_R(s_0))}+\|q\|_{L^2(A_R(s_0))}\right),
\qquad k=s,\theta. \tag*{(323)} ∥∇ ∂ k h ∥ L 2 ( A R ′ ( s 0 )) ≤ c ( ϵ , c Γ , κ ∗ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ( s 0 )) + ∥ q ∥ L 2 ( A R ( s 0 )) ) , k = s , θ . (323)
We next consider tangential derivatives ∂ k q \partial_k q ∂ k q , k = s , θ k=s,\theta k = s , θ , of the frozen-coefficient pressure q q q . We consider g k ∈ H 0 1 ( A R ( s 0 ) ) g_k\in H^1_0(A_R(s_0)) g k ∈ H 0 1 ( A R ( s 0 )) satisfying
div ( B 0 g k ) = ∂ k ( η 2 q ) in A R ( s 0 ) \operatorname{div}(B_0g_k)=\partial_k(\eta^2q)
\qquad\text{in }A_R(s_0) div ( B 0 g k ) = ∂ k ( η 2 q ) in A R ( s 0 )
∥ ∇ g k ∥ L 2 ( A R ( s 0 ) ) ≤ c ( κ ∗ , c Γ ) ∥ ∂ k ( η 2 q ) ∥ L 2 ( A R ( s 0 ) ) . (324) \|\nabla g_k\|_{L^2(A_R(s_0))}
\le c(\kappa_*,c_\Gamma)\|\partial_k(\eta^2q)\|_{L^2(A_R(s_0))}. \tag*{(324)} ∥∇ g k ∥ L 2 ( A R ( s 0 )) ≤ c ( κ ∗ , c Γ ) ∥ ∂ k ( η 2 q ) ∥ L 2 ( A R ( s 0 )) . (324)
Again, this may be justified by finite differences, and, since ∂ k ( η 2 q ) \partial_k(\eta^2q) ∂ k ( η 2 q ) has mean zero on A R ( s 0 ) A_R(s_0) A R ( s 0 ) and B 0 B_0 B 0 is invertible, such a g k g_k g k exists by [24 Section III.3] .
We now take η 2 ∂ k g k \eta^2\partial_kg_k η 2 ∂ k g k , k = s , θ k=s,\theta k = s , θ , as test functions in (313):
∫ A R ( s 0 ) ( ( ∇ h A 0 + B 0 T ∇ h T ∇ Ψ 0 T ) : ∇ ( η 2 ∂ k g k ) − q div ( B 0 ( η 2 ∂ k g k ) ) ) d y = 0. \int_{A_R(s_0)}
\left(\left(\nabla h A_0+B_0^{\mathsf T}\nabla h^{\mathsf T}\nabla\Psi_0^{\mathsf T}\right):\nabla(\eta^2\partial_kg_k)
-q\operatorname{div}\left(B_0(\eta^2\partial_kg_k)\right)\right)\,dy=0 . ∫ A R ( s 0 ) ( ( ∇ h A 0 + B 0 T ∇ h T ∇ Ψ 0 T ) : ∇ ( η 2 ∂ k g k ) − q div ( B 0 ( η 2 ∂ k g k ) ) ) d y = 0.
Commuting ∂ k \partial_k ∂ k through the equations (and noting that this picks up a factor of ∂ θ A 0 \partial_\theta A_0 ∂ θ A 0 , ∂ θ B 0 \partial_\theta B_0 ∂ θ B 0 , and ∂ θ ∇ Ψ 0 \partial_\theta\nabla\Psi_0 ∂ θ ∇ Ψ 0 ), after integrating by parts and using the definition (322) of g k g_k g k , we obtain
∫ A R ( s 0 ) ∣ ∂ k ( η 2 q ) ∣ 2 d y ≤ ∫ A R ( s 0 ) ( ∣ A 0 ∣ + ∣ B 0 ∣ ∣ ∇ Ψ 0 ∣ ) ( 2 ∣ ∇ h ∣ ∣ ∇ g k ∣ ∣ ∇ ( η 2 ) ∣ + η 2 ∣ ∇ g k ∣ ∣ ∇ ∂ k h ∣ ) d y + ∫ A R ( s 0 ) η 2 ∣ ∇ g k ∣ ( ∣ ∂ k A 0 ∣ + ∣ ∂ k B 0 ∣ ∣ ∇ Ψ 0 ∣ + ∣ B 0 ∣ ∣ ∂ k ∇ Ψ 0 ∣ ) ∣ ∇ h ∣ d y + ∫ A R ( s 0 ) ∣ q ∣ ( η 2 ( ∣ ∇ ∂ k B 0 ∣ ∣ g k ∣ + ∣ ∂ k B 0 ∣ ∣ ∇ g k ∣ ) + ∣ B 0 ∣ ∣ ∂ k g k ∣ ∣ ∇ ( η 2 ) ∣ ) d y . \begin{aligned}
\int_{A_R(s_0)} \left|\partial_k(\eta^2q)\right|^2\,dy
&\le\int_{A_R(s_0)} \left(|A_0|+|B_0||\nabla\Psi_0|\right)
\left(2|\nabla h||\nabla g_k||\nabla(\eta^2)|+\eta^2|\nabla g_k||\nabla\partial_k h|\right)\,dy \\
&\quad+\int_{A_R(s_0)}\eta^2|\nabla g_k|
\left(|\partial_k A_0|+|\partial_k B_0||\nabla\Psi_0|+|B_0||\partial_k\nabla\Psi_0|\right)|\nabla h|\,dy \\
&\quad+\int_{A_R(s_0)}|q|
\left(\eta^2\left(|\nabla\partial_k B_0||g_k|+|\partial_k B_0||\nabla g_k|\right)
+|B_0||\partial_k g_k||\nabla(\eta^2)|\right)\,dy .
\end{aligned} ∫ A R ( s 0 ) ∂ k ( η 2 q ) 2 d y ≤ ∫ A R ( s 0 ) ( ∣ A 0 ∣ + ∣ B 0 ∣∣∇ Ψ 0 ∣ ) ( 2∣∇ h ∣∣∇ g k ∣∣∇ ( η 2 ) ∣ + η 2 ∣∇ g k ∣∣∇ ∂ k h ∣ ) d y + ∫ A R ( s 0 ) η 2 ∣∇ g k ∣ ( ∣ ∂ k A 0 ∣ + ∣ ∂ k B 0 ∣∣∇ Ψ 0 ∣ + ∣ B 0 ∣∣ ∂ k ∇ Ψ 0 ∣ ) ∣∇ h ∣ d y + ∫ A R ( s 0 ) ∣ q ∣ ( η 2 ( ∣∇ ∂ k B 0 ∣∣ g k ∣ + ∣ ∂ k B 0 ∣∣∇ g k ∣ ) + ∣ B 0 ∣∣ ∂ k g k ∣∣∇ ( η 2 ) ∣ ) d y .
Using Young’s inequality and the bound (322), and recalling that ∣ ∇ η ∣ ≲ 1 R − R ′ |\nabla\eta|\lesssim\frac{1}{R-R'} ∣∇ η ∣ ≲ R − R ′ 1 , over the smaller annulus A R ′ ( s 0 ) A_{R'}(s_0) A R ′ ( s 0 ) , we then have
∥ ∂ k q ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ( ∥ η ∇ ∂ k h ∥ L 2 ( A R ) + ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) . (325) \begin{aligned}
\|\partial_kq\|_{L^2(A_{R'})}
&\le c(\epsilon,\kappa_*,c_\Gamma)\left(\|\eta\nabla\partial_kh\|_{L^2(A_R)}
+(R-R')^{-1}\left(\|h\|_{H^1(A_R)}+\|q\|_{L^2(A_R)}\right)\right) \\
&\le c(\epsilon,\kappa_*,c_\Gamma)(R-R')^{-1}
\left(\|h\|_{H^1(A_R)}+\|q\|_{L^2(A_R)}\right). \tag*{(325)}
\end{aligned} ∥ ∂ k q ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ( ∥ η ∇ ∂ k h ∥ L 2 ( A R ) + ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) . (325)
for k = s , θ k=s,\theta k = s , θ . Note that for ∂ s q \partial_sq ∂ s q , instead of (322), we may instead consider g s ∈ H 0 1 ( A R ′ ( s 0 ) ) g_s\in H^1_0(A_{R'}(s_0)) g s ∈ H 0 1 ( A R ′ ( s 0 )) satisfying
div ( B 0 g s ) = ∂ s q − ( ∂ s q ) s 0 , R ′ in A R ′ ( s 0 ) \operatorname{div}(B_0g_s)=\partial_sq-(\partial_sq)_{s_0,R'}\quad\text{in }A_{R'}(s_0) div ( B 0 g s ) = ∂ s q − ( ∂ s q ) s 0 , R ′ in A R ′ ( s 0 )
∥ ∇ g s ∥ L 2 ( A R ′ ( s 0 ) ) ≤ c ( κ ∗ , c Γ ) ∥ ∂ s q − ( ∂ s q ) s 0 , R ′ ∥ L 2 ( A R ′ ( s 0 ) ) . (326) \|\nabla g_s\|_{L^2(A_{R'}(s_0))}
\le c(\kappa_*,c_\Gamma)\|\partial_sq-(\partial_sq)_{s_0,R'}\|_{L^2(A_{R'}(s_0))}. \tag*{(326)} ∥∇ g s ∥ L 2 ( A R ′ ( s 0 )) ≤ c ( κ ∗ , c Γ ) ∥ ∂ s q − ( ∂ s q ) s 0 , R ′ ∥ L 2 ( A R ′ ( s 0 )) . (326)
where ( ∂ s q ) s 0 , R ′ (\partial_sq)_{s_0,R'} ( ∂ s q ) s 0 , R ′ is given by (310). Using ∂ s g s \partial_sg_s ∂ s g s as our test function in (313), we may then estimate
∫ A R ′ ( s 0 ) ∣ ∂ s q − ( ∂ s q ) s 0 , R ′ ∣ 2 d y ≤ ∫ A R ′ ( s 0 ) ( ∣ A 0 ∣ + ∣ B 0 ∣ ∣ ∇ Ψ 0 ∣ ) ∣ ∇ ∂ s h ∣ ∣ ∇ g s ∣ d y ≤ c ( ϵ , κ ∗ , c Γ ) ∥ ∇ ∂ s h ∥ L 2 ( A R ′ ) ∥ ∂ s q − ( ∂ s q ) s 0 , R ′ ∥ L 2 ( A R ′ ) . \begin{aligned}
\int_{A_{R'}(s_0)}\left|\partial_sq-(\partial_sq)_{s_0,R'}\right|^2\,dy
&\le\int_{A_{R'}(s_0)}\left(|A_0|+|B_0||\nabla\Psi_0|\right)|\nabla\partial_sh||\nabla g_s|\,dy \\
&\le c(\epsilon,\kappa_*,c_\Gamma)\|\nabla\partial_sh\|_{L^2(A_{R'})}
\|\partial_sq-(\partial_sq)_{s_0,R'}\|_{L^2(A_{R'})}.
\end{aligned} ∫ A R ′ ( s 0 ) ∣ ∂ s q − ( ∂ s q ) s 0 , R ′ ∣ 2 d y ≤ ∫ A R ′ ( s 0 ) ( ∣ A 0 ∣ + ∣ B 0 ∣∣∇ Ψ 0 ∣ ) ∣∇ ∂ s h ∣∣∇ g s ∣ d y ≤ c ( ϵ , κ ∗ , c Γ ) ∥∇ ∂ s h ∥ L 2 ( A R ′ ) ∥ ∂ s q − ( ∂ s q ) s 0 , R ′ ∥ L 2 ( A R ′ ) .
In particular, over A R ′ ( s 0 ) A_{R'}(s_0) A R ′ ( s 0 ) , we may estimate
∥ ∂ s q − ( ∂ s q ) s 0 , R ′ ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ∥ ∇ ∂ s h ∥ L 2 ( A R ′ ) . (327) \|\partial_sq-(\partial_sq)_{s_0,R'}\|_{L^2(A_{R'})}
\le c(\epsilon,\kappa_*,c_\Gamma)\|\nabla\partial_sh\|_{L^2(A_{R'})}. \tag*{(327)} ∥ ∂ s q − ( ∂ s q ) s 0 , R ′ ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ∥∇ ∂ s h ∥ L 2 ( A R ′ ) . (327)
We now turn to the normal direction derivatives ( ∂ r r h , ∂ r q ) (\partial_{rr}h,\partial_rq) ( ∂ r r h , ∂ r q ) , for which we must use the strong form of the equations (313) within A R ( s 0 ) A_R(s_0) A R ( s 0 ) . First, using the form (307) of B 0 = B ( s 0 , ϵ , θ ) B_0=B(s_0,\epsilon,\theta) B 0 = B ( s 0 , ϵ , θ ) and ∇ Ψ \nabla\Psi ∇Ψ (305), we may write the condition ∂ r div ( B 0 h ) = 0 \partial_r\operatorname{div}(B_0h)=0 ∂ r div ( B 0 h ) = 0 in cylindrical coordinates as
∂ r r h r = − ∂ r ( 1 ∣ det ∇ Ψ ∣ ( 1 r h r + 1 r ∂ θ ( h θ + a 1 ( θ ) h z ) + a 2 ( θ ) ∂ s h z ) ) . \partial_{rr}h_r=-\partial_r\left(\frac{1}{|\det\nabla\Psi|}\left(\frac{1}{r}h_r+\frac{1}{r}\partial_\theta(h_\theta+a_1(\theta)h_z)+a_2(\theta)\partial_sh_z\right)\right). ∂ r r h r = − ∂ r ( ∣ det ∇Ψ∣ 1 ( r 1 h r + r 1 ∂ θ ( h θ + a 1 ( θ ) h z ) + a 2 ( θ ) ∂ s h z ) ) .
where a 1 ( θ ) = − ϵ κ 3 1 − ϵ κ ( s 0 , θ ) a_1(\theta)=-\frac{\epsilon\kappa_3}{1-\epsilon\kappa(s_0,\theta)} a 1 ( θ ) = − 1 − ϵ κ ( s 0 , θ ) ϵ κ 3 and a 2 ( θ ) = 1 1 − ϵ κ ( s 0 , θ ) a_2(\theta)=\frac{1}{1-\epsilon\kappa(s_0,\theta)} a 2 ( θ ) = 1 − ϵ κ ( s 0 , θ ) 1 . In particular, by (321) and the boundedness of a 1 a_1 a 1 and a 2 a_2 a 2 , we have
∥ ∂ r r h r ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) . \|\partial_{rr}h_r\|_{L^2(A_{R'})}
\le c(\epsilon,\kappa_*,c_\Gamma)(R-R')^{-1}
\left(\|h\|_{H^1(A_R)}+\|q\|_{L^2(A_R)}\right). ∥ ∂ r r h r ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) .
Bounds for the remaining directions ∂ r r ( h ⋅ e θ ) \partial_{rr}(h\cdot e_\theta) ∂ r r ( h ⋅ e θ ) , ∂ r r ( h ⋅ e z ) \partial_{rr}(h\cdot e_z) ∂ r r ( h ⋅ e z ) , and ∂ r q \partial_rq ∂ r q come from the form of the frozen-coefficient straightened Stokes equations in A R ( s 0 ) A_R(s_0) A R ( s 0 ) . We have that ( h , q ) (h,q) ( h , q ) satisfies
div ( ∇ h A 0 ) + F ( ∇ h ) + G ( h ) = B 0 T ∇ q in A R ( s 0 ) . (328) \operatorname{div}(\nabla hA_0)+F(\nabla h)+G(h)=B_0^{\mathsf T}\nabla q\quad\text{in }A_R(s_0). \tag*{(328)} div ( ∇ h A 0 ) + F ( ∇ h ) + G ( h ) = B 0 T ∇ q in A R ( s 0 ) . (328)
Here the lower order terms F F F and G G G arise from the frozen-coefficient symmetric gradient term div ( B 0 T ∇ h T ∇ Ψ 0 T ) \operatorname{div}(B_0^{\mathsf T}\nabla h^{\mathsf T}\nabla\Psi_0^{\mathsf T}) div ( B 0 T ∇ h T ∇ Ψ 0 T ) and the divergence-free condition div ( B 0 h ) = 0 \operatorname{div}(B_0h)=0 div ( B 0 h ) = 0 . Using the form (307) of A 0 = A ( s 0 , ϵ , θ ) A_0=A(s_0,\epsilon,\theta) A 0 = A ( s 0 , ϵ , θ ) and B 0 B_0 B 0 , we may write the normal direction components of the equation (326) as
∣ det ∇ Ψ ∣ div ( ∇ h A 0 ) ⋅ e r = 1 r ∂ r ( r ∂ r h r ) + 1 r 2 ∂ θ ( ∂ θ h r − h θ ) + ∂ s s h r , |\det\nabla\Psi|\operatorname{div}(\nabla hA_0)\cdot e_r
=\frac{1}{r}\partial_r(r\partial_rh_r)+\frac{1}{r^2}\partial_\theta(\partial_\theta h_r-h_\theta)+\partial_{ss}h_r, ∣ det ∇Ψ∣ div ( ∇ h A 0 ) ⋅ e r = r 1 ∂ r ( r ∂ r h r ) + r 2 1 ∂ θ ( ∂ θ h r − h θ ) + ∂ ss h r ,
∣ det ∇ Ψ ∣ ( B 0 T ∇ q ) ⋅ e r = ∂ r q , |\det\nabla\Psi|(B_0^{\mathsf T}\nabla q)\cdot e_r=\partial_rq, ∣ det ∇Ψ∣ ( B 0 T ∇ q ) ⋅ e r = ∂ r q ,
from which we may bound ∂ r q \partial_{r}q ∂ r q by ∂ r r h r \partial_{rr}h_{r} ∂ r r h r , ∂ θ θ h r \partial_{\theta\theta}h_{r} ∂ θ θ h r , ∂ s s h r \partial_{ss}h_{r} ∂ ss h r , ∇ h \nabla h ∇ h , and h h h to obtain a full gradient bound for q q q :
∥ ∇ q ∥ L 2 ( A R ′ ( s 0 ) ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) . (329) \lVert\nabla q\rVert_{L^{2}(A_{R'}(s_{0}))} \le c(\epsilon,\kappa_{*},c_{\Gamma})(R-R')^{-1}\left(\lVert h\rVert_{H^{1}(A_{R})}+\lVert q\rVert_{L^{2}(A_{R})}\right). \tag*{(329)} ∥ ∇ q ∥ L 2 ( A R ′ ( s 0 )) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) . (329)
In the e θ e_{\theta} e θ , e z e_{z} e z directions, we obtain
∣ det ∇ Ψ ∣ div ( ∇ h A 0 ) ⋅ e θ = 1 r ∂ r ( r ( ∂ r h θ + a 1 ∂ r h z ) ) + 1 r 2 ∂ θ ( ∂ θ h θ + h r + a 1 ∂ s h z ) + ∂ s s ( h θ + a 1 h z ) , \lvert\det\nabla\Psi\rvert\operatorname{div}(\nabla h A_{0})\cdot e_{\theta}=\frac{1}{r}\partial_{r}\left(r(\partial_{r}h_{\theta}+a_{1}\partial_{r}h_{z})\right)+\frac{1}{r^{2}}\partial_{\theta}(\partial_{\theta}h_{\theta}+h_{r}+a_{1}\partial_{s}h_{z})+\partial_{ss}(h_{\theta}+a_{1}h_{z}), ∣ det ∇Ψ ∣ div ( ∇ h A 0 ) ⋅ e θ = r 1 ∂ r ( r ( ∂ r h θ + a 1 ∂ r h z ) ) + r 2 1 ∂ θ ( ∂ θ h θ + h r + a 1 ∂ s h z ) + ∂ ss ( h θ + a 1 h z ) ,
∣ det ∇ Ψ ∣ ( B 0 T ∇ q ) ⋅ e θ = 1 r ∂ θ q , \lvert\det\nabla\Psi\rvert(B_{0}^{\mathsf{T}}\nabla q)\cdot e_{\theta}=\frac{1}{r}\partial_{\theta}q, ∣ det ∇Ψ ∣ ( B 0 T ∇ q ) ⋅ e θ = r 1 ∂ θ q ,
∣ det ∇ Ψ ∣ div ( ∇ h A 0 ) ⋅ e r = 1 r ∂ r ( r ( a 1 ∂ r h θ + a 3 ∂ r h z ) ) + 1 r 2 ∂ θ ( a 1 ∂ θ h θ + a 1 h r + a 3 ∂ s h z ) + ∂ s s ( a 1 h θ + a 3 h z ) , \lvert\det\nabla\Psi\rvert\operatorname{div}(\nabla h A_{0})\cdot e_{r}=\frac{1}{r}\partial_{r}\left(r(a_{1}\partial_{r}h_{\theta}+a_{3}\partial_{r}h_{z})\right)+\frac{1}{r^{2}}\partial_{\theta}(a_{1}\partial_{\theta}h_{\theta}+a_{1}h_{r}+a_{3}\partial_{s}h_{z})+\partial_{ss}(a_{1}h_{\theta}+a_{3}h_{z}), ∣ det ∇Ψ ∣ div ( ∇ h A 0 ) ⋅ e r = r 1 ∂ r ( r ( a 1 ∂ r h θ + a 3 ∂ r h z ) ) + r 2 1 ∂ θ ( a 1 ∂ θ h θ + a 1 h r + a 3 ∂ s h z ) + ∂ ss ( a 1 h θ + a 3 h z ) ,
∣ det ∇ Ψ ∣ ( B 0 T ∇ q ) ⋅ e z = a 2 ∂ s q + a 1 ∂ θ q . \lvert\det\nabla\Psi\rvert(B_{0}^{\mathsf{T}}\nabla q)\cdot e_{z}=a_{2}\partial_{s}q+a_{1}\partial_{\theta}q. ∣ det ∇Ψ ∣ ( B 0 T ∇ q ) ⋅ e z = a 2 ∂ s q + a 1 ∂ θ q .
where a 1 ( θ ) a_{1}(\theta) a 1 ( θ ) and a 2 ( θ ) a_{2}(\theta) a 2 ( θ ) are as above and a 3 ( θ ) = 1 + ϵ 2 κ 3 2 ( 1 − ϵ κ b ( s 0 , θ ) ) 2 a_{3}(\theta)=\frac{1+\epsilon^{2}\kappa_{3}^{2}}{(1-\epsilon\kappa_{b}(s_{0},\theta))^{2}} a 3 ( θ ) = ( 1 − ϵ κ b ( s 0 , θ ) ) 2 1 + ϵ 2 κ 3 2 . Using (328) , we obtain a system of two equations for the unknown directions ∂ r r h θ \partial_{rr}h_{\theta} ∂ r r h θ and ∂ r r h z \partial_{rr}h_{z} ∂ r r h z in terms of the remaining directions. Together, we may thus bound
∥ ∂ r r h θ ∥ L 2 ( A R ′ ) , ∥ ∂ r r h z ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) . \lVert\partial_{rr}h_{\theta}\rVert_{L^{2}(A_{R'})},\lVert\partial_{rr}h_{z}\rVert_{L^{2}(A_{R'})}\le c(\epsilon,\kappa_{*},c_{\Gamma})(R-R')^{-1}\left(\lVert h\rVert_{H^{1}(A_{R})}+\lVert q\rVert_{L^{2}(A_{R})}\right). ∥ ∂ r r h θ ∥ L 2 ( A R ′ ) , ∥ ∂ r r h z ∥ L 2 ( A R ′ ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ) + ∥ q ∥ L 2 ( A R ) ) .
Combining the tangential regularity bound (321) with the pressure gradient bound (327) and the bounds for ∂ r r h \partial_{rr}h ∂ r r h , we obtain the full H 2 ( A R ′ ( s 0 ) ) × H 1 ( A R ′ ( s 0 ) ) H^{2}(A_{R'}(s_{0}))\times H^{1}(A_{R'}(s_{0})) H 2 ( A R ′ ( s 0 )) × H 1 ( A R ′ ( s 0 )) bound
∥ h ∥ H 2 ( A R ′ ( s 0 ) ) + ∥ q ∥ H 1 ( A R ′ ( s 0 ) ) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ( s 0 ) ) + ∥ q ∥ L 2 ( A R ( s 0 ) ) ) . (330) \lVert h\rVert_{H^{2}(A_{R'}(s_{0}))}+\lVert q\rVert_{H^{1}(A_{R'}(s_{0}))}\le c(\epsilon,\kappa_{*},c_{\Gamma})(R-R')^{-1}\left(\lVert h\rVert_{H^{1}(A_{R}(s_{0}))}+\lVert q\rVert_{L^{2}(A_{R}(s_{0}))}\right). \tag*{(330)} ∥ h ∥ H 2 ( A R ′ ( s 0 )) + ∥ q ∥ H 1 ( A R ′ ( s 0 )) ≤ c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) − 1 ( ∥ h ∥ H 1 ( A R ( s 0 )) + ∥ q ∥ L 2 ( A R ( s 0 )) ) . (330)
This procedure may be iterated over smaller and smaller annuli, using test functions of the form ∂ k ∂ ℓ ( η 1 2 ( ∂ k ∂ ℓ h ) ) \partial_{k}\partial_{\ell}(\eta_{1}^{2}(\partial_{k}\partial_{\ell}h)) ∂ k ∂ ℓ ( η 1 2 ( ∂ k ∂ ℓ h )) , ∂ k ∂ ℓ ∂ m ( η 2 2 ( ∂ k ∂ ℓ ∂ m h ) ) \partial_{k}\partial_{\ell}\partial_{m}(\eta_{2}^{2}(\partial_{k}\partial_{\ell}\partial_{m}h)) ∂ k ∂ ℓ ∂ m ( η 2 2 ( ∂ k ∂ ℓ ∂ m h )) , etc., where η j \eta_{j} η j are appropriate cutoffs and similar corrections for div ( B 0 ⋅ ) \operatorname{div}(B_{0}\cdot) div ( B 0 ⋅ ) -freeness are made. We thus obtain Proposition 6.3. □
Recalling the notation g s 0 , R g_{s_{0},R} g s 0 , R for the mean of a function g g g over the annular region A R ( s 0 ) A_{R}(s_{0}) A R ( s 0 ) , we may now use Proposition 6.3 to show that the solution ( h , q ) (h,q) ( h , q ) to (313) satisfies the following.
Corollary 6.4 (Caccioppoli inequality). For 0 < R ′ < R 0<R'<R 0 < R ′ < R , the solution ( h , q ) (h,q) ( h , q ) to (313) satisfies
∫ A R ′ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ≤ c ( ϵ , κ ∗ ) 1 ( R − R ′ ) 2 ∫ A R ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , R ∣ 2 d y . (331) \int_{A_{R'}(s_{0})}\lvert\nabla_{\partial_{s}h}\rvert^{2}\,dy\le c(\epsilon,\kappa_{*})\frac{1}{(R-R')^{2}}\int_{A_{R}(s_{0})}\lvert\partial_{s}h-(\partial_{s}h)_{s_{0},R}\rvert^{2}\,dy. \tag*{(331)} ∫ A R ′ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ≤ c ( ϵ , κ ∗ ) ( R − R ′ ) 2 1 ∫ A R ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , R ∣ 2 d y . (331)
Proof. Using that the coefficients A 0 ( θ ) A_{0}(\theta) A 0 ( θ ) , B 0 ( θ ) B_{0}(\theta) B 0 ( θ ) , ∇ Ψ 0 ( θ ) \nabla\Psi_{0}(\theta) ∇ Ψ 0 ( θ ) are independent of s s s , for ( h , q ) (h,q) ( h , q ) satisfying (313), we have that ( ∂ s h , ∂ s q − ( ∂ s q ) s 0 , R ) (\partial_{s}h,\partial_{s}q-(\partial_{s}q)_{s_{0},R}) ( ∂ s h , ∂ s q − ( ∂ s q ) s 0 , R ) satisfies
∫ A R ( s 0 ) ( ( ∇ ∂ s h A 0 ( θ ) + B 0 T ( θ ) ( ∇ ∂ s h ) T ∇ Ψ 0 T ) : ∇ ψ − ( ∂ s q − ( ∂ s q ) s 0 , R ) div ( B 0 ( θ ) ψ ) ) d y = 0 (332) \int_{A_{R}(s_{0})}\left((\nabla\partial_{s}h A_{0}(\theta)+B_{0}^{\mathsf{T}}(\theta)(\nabla\partial_{s}h)^{\mathsf{T}}\nabla\Psi_{0}^{\mathsf{T}}):\nabla\psi-(\partial_{s}q-(\partial_{s}q)_{s_{0},R})\operatorname{div}(B_{0}(\theta)\psi)\right)\,dy=0 \tag*{(332)} ∫ A R ( s 0 ) ( ( ∇ ∂ s h A 0 ( θ ) + B 0 T ( θ ) ( ∇ ∂ s h ) T ∇ Ψ 0 T ) : ∇ ψ − ( ∂ s q − ( ∂ s q ) s 0 , R ) div ( B 0 ( θ ) ψ ) ) d y = 0 (332)
for any ψ ∈ H 1 ( A R ( s 0 ) ) \psi\in H^{1}(A_{R}(s_{0})) ψ ∈ H 1 ( A R ( s 0 )) with ψ ∣ C ϵ = ψ ( s ) \psi|_{C_{\epsilon}}=\psi(s) ψ ∣ C ϵ = ψ ( s ) and ψ ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 \psi|_{\partial A_{R}(s_{0})\setminus C_{\epsilon}}=0 ψ ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 . Furthermore, since adding a constant does not affect the θ \theta θ -independence of ∂ s h ∣ C ϵ \partial_{s}h|_{C_{\epsilon}} ∂ s h ∣ C ϵ , we have that ( ∂ s h − λ , ∂ s q − ( ∂ s q ) s 0 , R ) (\partial_{s}h-\lambda,\partial_{s}q-(\partial_{s}q)_{s_{0},R}) ( ∂ s h − λ , ∂ s q − ( ∂ s q ) s 0 , R ) also satisfies (330) for any constant λ \lambda λ .
For 0 < R ′ < R 0<R'<R 0 < R ′ < R , taking η \eta η to be a cutoff function as in (315), we will use ψ = η 2 ( ∂ s h − λ ) \psi=\eta^{2}(\partial_{s}h-\lambda) ψ = η 2 ( ∂ s h − λ ) as a test function in (330) with ( ∂ s h − λ , ∂ s q − ( ∂ s q ) s 0 , R ) (\partial_{s}h-\lambda,\partial_{s}q-(\partial_{s}q)_{s_{0},R}) ( ∂ s h − λ , ∂ s q − ( ∂ s q ) s 0 , R ) . We then have
0 = ∫ A R ( s 0 ) ( ( ∇ ( ∂ s h − λ ) A 0 ( θ ) + B 0 T ( θ ) ∇ ( ∂ s h − λ ) T ∇ Ψ 0 T ) : ∇ ( η 2 ( ∂ s h − λ ) ) 0=\int_{A_{R}(s_{0})}((\nabla(\partial_{s}h-\lambda)A_{0}(\theta)+B_{0}^{\mathsf{T}}(\theta)\nabla(\partial_{s}h-\lambda)^{\mathsf{T}}\nabla\Psi_{0}^{\mathsf{T}}):\nabla(\eta^{2}(\partial_{s}h-\lambda)) 0 = ∫ A R ( s 0 ) (( ∇ ( ∂ s h − λ ) A 0 ( θ ) + B 0 T ( θ ) ∇ ( ∂ s h − λ ) T ∇ Ψ 0 T ) : ∇ ( η 2 ( ∂ s h − λ ))
− ( ∂ s q − ( ∂ s q ) s 0 , R ) div ( B 0 ( θ ) η 2 ( ∂ s h − λ ) ) d y = ∫ A R ( s 0 ) ( η 2 ( ∇ ( ∂ s h − λ ) A 0 ( θ ) + B 0 T ( θ ) ∇ ( ∂ s h − λ ) ∇ Ψ 0 T ) : ∇ ( ∂ s h − λ ) + ( ∇ ( ∂ s h − λ ) A 0 ( θ ) + B 0 T ( θ ) ∇ ( ∂ s h − λ ) ∇ Ψ 0 T ) : ( ∇ ( η 2 ) ⊗ ( ∂ s h − λ ) ) \begin{aligned}
-\bigl(\partial_s q-(\partial_s q)_{s_0,R}\bigr)\operatorname{div}\bigl(B_0(\theta)\eta^2(\partial_s h-\lambda)\bigr)\,dy
=
\int_{A_R(s_0)}\Bigl(\eta^2\bigl(\nabla(\partial_s h-\lambda)A_0(\theta)+B_0^{T}(\theta)\nabla(\partial_s h-\lambda)\nabla\Psi_0^{T}\bigr):\nabla(\partial_s h-\lambda) \\
&+\bigl(\nabla(\partial_s h-\lambda)A_0(\theta)+B_0^{T}(\theta)\nabla(\partial_s h-\lambda)\nabla\Psi_0^{T}\bigr):\bigl(\nabla(\eta^2)\otimes(\partial_s h-\lambda)\bigr)
\end{aligned} − ( ∂ s q − ( ∂ s q ) s 0 , R ) div ( B 0 ( θ ) η 2 ( ∂ s h − λ ) ) d y = ∫ A R ( s 0 ) ( η 2 ( ∇ ( ∂ s h − λ ) A 0 ( θ ) + B 0 T ( θ ) ∇ ( ∂ s h − λ ) ∇ Ψ 0 T ) : ∇ ( ∂ s h − λ ) + ( ∇ ( ∂ s h − λ ) A 0 ( θ ) + B 0 T ( θ ) ∇ ( ∂ s h − λ ) ∇ Ψ 0 T ) : ( ∇ ( η 2 ) ⊗ ( ∂ s h − λ ) )
− η 2 ( ∂ s q − ( ∂ s q ) s 0 , R ) div ( B 0 ( θ ) ( ∂ s h − λ ) ) − ( ∂ s q − ( ∂ s q ) s 0 , R ) ∇ ( η 2 ) ⋅ ( B 0 ( θ ) ( ∂ s h − λ ) ) ) d y . -\eta^2\bigl(\partial_s q-(\partial_s q)_{s_0,R}\bigr)\operatorname{div}\bigl(B_0(\theta)(\partial_s h-\lambda)\bigr)-\bigl(\partial_s q-(\partial_s q)_{s_0,R}\bigr)\nabla(\eta^2)\cdot\bigl(B_0(\theta)(\partial_s h-\lambda)\bigr)\Bigr)\,dy. − η 2 ( ∂ s q − ( ∂ s q ) s 0 , R ) div ( B 0 ( θ ) ( ∂ s h − λ ) ) − ( ∂ s q − ( ∂ s q ) s 0 , R ) ∇ ( η 2 ) ⋅ ( B 0 ( θ ) ( ∂ s h − λ ) ) ) d y .
Since div ( B 0 ( θ ) ( ∂ s h − λ ) ) = 0 \operatorname{div}\bigl(B_0(\theta)(\partial_s h-\lambda)\bigr)=0 div ( B 0 ( θ ) ( ∂ s h − λ ) ) = 0 , after using an analogous argument to (320) to bound η ∇ ( ∂ s h − λ ) \eta\nabla(\partial_s h-\lambda) η ∇ ( ∂ s h − λ ) by the first term in the final expression above, we may estimate
∫ A R ( s 0 ) η 2 ∣ ∇ ( ∂ s h − λ ) ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∇ η ∣ ∣ ∂ s h − λ ∣ ( ∣ η ∇ ( ∂ s h − λ ) ∣ + ∣ ∂ s q − ( ∂ s q ) s 0 , R ∣ ) d y . \int_{A_R(s_0)}\eta^2\lvert\nabla(\partial_s h-\lambda)\rvert^2\,dy
\le c(\epsilon,\kappa_*,c_\Gamma)\int_{A_R(s_0)}\lvert\nabla\eta\rvert\lvert\partial_s h-\lambda\rvert\bigl(\lvert\eta\nabla(\partial_s h-\lambda)\rvert+\lvert\partial_s q-(\partial_s q)_{s_0,R}\rvert\bigr)\,dy. ∫ A R ( s 0 ) η 2 ∣ ∇ ( ∂ s h − λ ) ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∇ η ∣ ∣ ∂ s h − λ ∣ ( ∣ η ∇ ( ∂ s h − λ )∣ + ∣ ∂ s q − ( ∂ s q ) s 0 , R ∣ ) d y .
Using Young’s inequality, the support of η \eta η , and ∇ η ≲ 1 ∣ R − R ′ ∣ \nabla\eta\lesssim\frac{1}{\lvert R-R'\rvert} ∇ η ≲ ∣ R − R ′ ∣ 1 from (315), we obtain
∫ A R ′ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ≤ c ( δ ) c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) 2 ∫ A R ( s 0 ) ∣ ∂ s h − λ ∣ 2 d y + δ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∂ s q − ( ∂ s q ) s 0 , R ∣ 2 d y \int_{A_{R'}(s_0)}\lvert\nabla\partial_s h\rvert^2\,dy
\le\frac{c(\delta)c(\epsilon,\kappa_*,c_\Gamma)}{(R-R')^2}\int_{A_R(s_0)}\lvert\partial_s h-\lambda\rvert^2\,dy+\delta c(\epsilon,\kappa_*,c_\Gamma)\int_{A_R(s_0)}\lvert\partial_s q-(\partial_s q)_{s_0,R}\rvert^2\,dy ∫ A R ′ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ≤ ( R − R ′ ) 2 c ( δ ) c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∂ s h − λ ∣ 2 d y + δ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∂ s q − ( ∂ s q ) s 0 , R ∣ 2 d y
for any small δ > 0 \delta>0 δ > 0 . Using the estimate (325), we have
∫ A R ( s 0 ) ∣ ∂ s q − ( ∂ s q ) s 0 , R ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ; \int_{A_R(s_0)}\lvert\partial_s q-(\partial_s q)_{s_0,R}\rvert^2\,dy\le c(\epsilon,\kappa_*,c_\Gamma)\int_{A_R(s_0)}\lvert\nabla\partial_s h\rvert^2\,dy; ∫ A R ( s 0 ) ∣ ∂ s q − ( ∂ s q ) s 0 , R ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ;
so, in particular,
∫ A R ′ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ≤ c ( δ ) c ( ϵ , κ ∗ , c Γ ) ( R − R ′ ) 2 ∫ A R ( s 0 ) ∣ ∂ s h − λ ∣ 2 d y + δ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y . (333) \int_{A_{R'}(s_0)}\lvert\nabla\partial_s h\rvert^2\,dy\le\frac{c(\delta)c(\epsilon,\kappa_*,c_\Gamma)}{(R-R')^2}\int_{A_R(s_0)}\lvert\partial_s h-\lambda\rvert^2\,dy+\delta c(\epsilon,\kappa_*,c_\Gamma)\int_{A_R(s_0)}\lvert\nabla\partial_s h\rvert^2\,dy. \tag*{(333)} ∫ A R ′ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y ≤ ( R − R ′ ) 2 c ( δ ) c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∂ s h − λ ∣ 2 d y + δ c ( ϵ , κ ∗ , c Γ ) ∫ A R ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y . (333)
We now recall the following useful proposition.
Proposition 6.5 (Lemma 6.1 from [29 ] ). Let Z ( ρ ) Z(\rho) Z ( ρ ) be a bounded, nonnegative function in the interval [ R 1 , R 2 ] [R_1,R_2] [ R 1 , R 2 ] . If for R 1 ≤ R ′ ≤ R ≤ R 2 R_1\le R'\le R\le R_2 R 1 ≤ R ′ ≤ R ≤ R 2 we have
Z ( R ′ ) ≤ M ( R − R ′ ) − m + δ Z ( R ) Z(R')\le M(R-R')^{-m}+\delta Z(R) Z ( R ′ ) ≤ M ( R − R ′ ) − m + δ Z ( R ) for M , m > 0 M,m>0 M , m > 0 and 0 ≤ δ < 1 0\le\delta<1 0 ≤ δ < 1 , then
Z ( R 1 ) ≤ c ( m , δ ) M ( R 2 − R 1 ) − m . (334) Z(R_1)\le c(m,\delta)M(R_2-R_1)^{-m}. \tag*{(334)} Z ( R 1 ) ≤ c ( m , δ ) M ( R 2 − R 1 ) − m . (334)
Taking δ \delta δ sufficiently small, we may apply Proposition 6.5 to (331) with Z ( ρ ) = ∫ A ρ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y Z(\rho)=\int_{A_\rho(s_0)}\lvert\nabla\partial_s h\rvert^2\,dy Z ( ρ ) = ∫ A ρ ( s 0 ) ∣ ∇ ∂ s h ∣ 2 d y . Taking λ = ( ∂ s h ) s 0 , R \lambda=(\partial_s h)_{s_0,R} λ = ( ∂ s h ) s 0 , R , we thus obtain (329). □
Finally, using Proposition 6.3 and Corollary 6.4, we show that oscillation of ∂ s h \partial_s h ∂ s h satisfies the following scaling.
Lemma 6.6 (Oscillation of ∂ s h \partial_s h ∂ s h ). Consider ( h , q ) (h,q) ( h , q ) satisfying (313). For 0 < ρ < R / 2 0<\rho<R/2 0 < ρ < R /2 , we have that ∂ s h \partial_s h ∂ s h satisfies
\mathop{\fint}_{A_\rho(s_0)}\lvert\partial_s h-(\partial_s h)_{s_0,\rho}\rvert^2\,dy\le c(\epsilon,\kappa_*,c_\Gamma)\left(\frac{\rho}{R}\right)^2\mathop{\fint}_{A_R(s_0)}\lvert\partial_s h-(\partial_s h)_{s_0,R}\rvert^2\,dy. \tag*{(335)}
Proof. Using a Poincaré inequality over A ρ ( s 0 ) A_{\rho}(s_{0}) A ρ ( s 0 ) , Sobolev embedding, Proposition 6.3 , estimate (327) , and Corollary 6.4 , we may estimate
∫ A ρ ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , ρ ∣ 2 d y ≤ c ( ϵ ) ρ 2 ∫ A ρ ( s 0 ) ∣ ∇ σ h ∣ 2 d y ≤ c ( ϵ , κ ∗ ) ρ 2 ∫ 0 2 π ( sup ∣ s − s 0 ∣ ≤ ρ , 0 ≤ r ≤ ρ ∣ ∇ σ h ∣ 2 ) d θ ≤ c ( ϵ , κ ∗ ) ρ 2 ∫ 0 2 π ( sup ∣ s − s 0 ∣ ≤ R / 2 , 0 ≤ r ≤ R / 2 ∣ ∇ σ h ∣ 2 ) d θ ≤ c ( ϵ , κ ∗ ) ρ 2 ( R − 2 ∥ ∇ σ h ∥ L 2 ( A R / 2 ( s 0 ) ) 2 + R 2 ∥ ∇ σ h ∥ H ˙ 2 ( A R / 2 ( s 0 ) ) 2 ) ≤ c ( ϵ , κ ∗ , c Γ ) ρ 2 ∫ A 3 R / 4 ( s 0 ) ( ∣ ∇ σ h ∣ 2 + ∣ ∂ s q − ( ∂ s q ) s 0 , 3 R / 4 ∣ 2 ) d y ≤ c ( ϵ , κ ∗ , c Γ ) ρ 2 ∫ A 3 R / 4 ( s 0 ) ∣ ∇ σ h ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ( ρ R ) 2 ∫ A R ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , R ∣ 2 d y . \begin{aligned}
\int_{A_{\rho}(s_{0})}\left|\partial_{s}h-(\partial_{s}h)_{s_{0},\rho}\right|^{2}\,dy
&\leq c(\epsilon)\rho^{2}\int_{A_{\rho}(s_{0})}|\nabla_{\sigma}h|^{2}\,dy \\
&\leq c(\epsilon,\kappa_{*})\rho^{2}\int_{0}^{2\pi}\left(\sup_{|s-s_{0}|\leq\rho,\,0\leq r\leq\rho}|\nabla_{\sigma}h|^{2}\right)\,d\theta\\
&\leq c(\epsilon,\kappa_{*})\rho^{2}\int_{0}^{2\pi}\left(\sup_{|s-s_{0}|\leq R/2,\,0\leq r\leq R/2}|\nabla_{\sigma}h|^{2}\right)\,d\theta\\
&\leq c(\epsilon,\kappa_{*})\rho^{2}\left(R^{-2}\|\nabla_{\sigma}h\|_{L^{2}(A_{R/2}(s_{0}))}^{2}+R^{2}\|\nabla_{\sigma}h\|_{\dot{H}^{2}(A_{R/2}(s_{0}))}^{2}\right) \\
&\leq c(\epsilon,\kappa_{*},c_{\Gamma})\rho^{2}\int_{A_{3R/4}(s_{0})}\left(|\nabla_{\sigma}h|^{2}+\left|\partial_{s}q-(\partial_{s}q)_{s_{0},3R/4}\right|^{2}\right)\,dy \\
&\leq c(\epsilon,\kappa_{*},c_{\Gamma})\rho^{2}\int_{A_{3R/4}(s_{0})}|\nabla_{\sigma}h|^{2}\,dy
\leq c(\epsilon,\kappa_{*},c_{\Gamma})\left(\frac{\rho}{R}\right)^{2}\int_{A_{R}(s_{0})}\left|\partial_{s}h-(\partial_{s}h)_{s_{0},R}\right|^{2}\,dy.
\end{aligned} ∫ A ρ ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , ρ ∣ 2 d y ≤ c ( ϵ ) ρ 2 ∫ A ρ ( s 0 ) ∣ ∇ σ h ∣ 2 d y ≤ c ( ϵ , κ ∗ ) ρ 2 ∫ 0 2 π ( ∣ s − s 0 ∣ ≤ ρ , 0 ≤ r ≤ ρ sup ∣ ∇ σ h ∣ 2 ) d θ ≤ c ( ϵ , κ ∗ ) ρ 2 ∫ 0 2 π ( ∣ s − s 0 ∣ ≤ R /2 , 0 ≤ r ≤ R /2 sup ∣ ∇ σ h ∣ 2 ) d θ ≤ c ( ϵ , κ ∗ ) ρ 2 ( R − 2 ∥ ∇ σ h ∥ L 2 ( A R /2 ( s 0 )) 2 + R 2 ∥ ∇ σ h ∥ H ˙ 2 ( A R /2 ( s 0 )) 2 ) ≤ c ( ϵ , κ ∗ , c Γ ) ρ 2 ∫ A 3 R /4 ( s 0 ) ( ∣ ∇ σ h ∣ 2 + ∂ s q − ( ∂ s q ) s 0 , 3 R /4 2 ) d y ≤ c ( ϵ , κ ∗ , c Γ ) ρ 2 ∫ A 3 R /4 ( s 0 ) ∣ ∇ σ h ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ( R ρ ) 2 ∫ A R ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , R ∣ 2 d y .
In the fourth inequality we have used scaling along each ( s , r ) (s,r) ( s , r ) slice in the annulus A R / 2 ( s 0 ) A_{R/2}(s_{0}) A R /2 ( s 0 ) to obtain the R R R -dependence of the Sobolev inequality. In the fifth inequality we have used that ( ∂ s h , ∂ s q − ( ∂ s q ) s 0 , 3 R / 4 ) (\partial_{s}h,\partial_{s}q-(\partial_{s}q)_{s_{0},3R/4}) ( ∂ s h , ∂ s q − ( ∂ s q ) s 0 , 3 R /4 ) satisfies (315) with zero boundary data, and we may thus apply the bound from Proposition 6.3 at the level of ( ∂ s h , ∂ s q − ( ∂ s q ) s 0 , 3 R / 4 ) (\partial_{s}h,\partial_{s}q-(\partial_{s}q)_{s_{0},3R/4}) ( ∂ s h , ∂ s q − ( ∂ s q ) s 0 , 3 R /4 ) . □
Equipped with Lemmas 6.2 and 6.6 , we may proceed to the proof of Lemma 1.14 .
Proof of Lemma 1.14. Consider the weak solution ( u ‾ , p ‾ ) (\overline{u},\overline{p}) ( u , p ) to the slender body BVP satisfying (306) within the straightened region Ψ ( O r ∗ ) \Psi(\mathcal{O}_{r_{*}}) Ψ ( O r ∗ ) . Let 0 < R ≤ r ∗ 2 0<R\leq\frac{r_{*}}{2} 0 < R ≤ 2 r ∗ and s 0 ∈ T s_{0}\in\mathbb{T} s 0 ∈ T , and consider the annular region A R ( s 0 ) ⊂ Ψ ( O r ∗ ) A_{R}(s_{0})\subset\Psi(\mathcal{O}_{r_{*}}) A R ( s 0 ) ⊂ Ψ ( O r ∗ ) .
Within A R ( s 0 ) A_{R}(s_{0}) A R ( s 0 ) , we decompose ( u ‾ , p ‾ ) (\overline{u},\overline{p}) ( u , p ) as u ‾ = h + b \overline{u}=h+b u = h + b and p ‾ = q + ζ \overline{p}=q+\zeta p = q + ζ , where ( h , q ) (h,q) ( h , q ) satisfies the frozen-coefficient slender body PDE (315) . Rewriting (306) as
∫ A R ( ∇ u ‾ ( A − A 0 ) + ( B − B 0 ) T ∇ u ‾ T ∇ Ψ T + B 0 T ∇ u ‾ T ( ∇ Ψ − ∇ Ψ 0 ) T ) : ∇ ψ d y + ∫ A R ( ( ∇ h + ∇ b ) A 0 + B 0 T ( ∇ h + ∇ b ) T ∇ Ψ 0 T ) : ∇ ψ d y − ∫ A R ( p ‾ − p ‾ s 0 , R ) div ( ( B − B 0 ) ψ ) d y − ∫ A R ( q + ζ − p ‾ s 0 , R ) div ( B 0 ψ ) d y = ∫ T f ( s ) ψ ( s ) d s \begin{aligned}
&\int_{A_{R}}\left(\nabla\overline{u}(A-A_{0})+(B-B_{0})^{\mathsf{T}}\nabla\overline{u}^{\mathsf{T}}\nabla\Psi^{\mathsf{T}}+B_{0}^{\mathsf{T}}\nabla\overline{u}^{\mathsf{T}}(\nabla\Psi-\nabla\Psi_{0})^{\mathsf{T}}\right):\nabla\psi\,dy \\
&\quad+\int_{A_{R}}\left((\nabla h+\nabla b)A_{0}+B_{0}^{\mathsf{T}}(\nabla h+\nabla b)^{\mathsf{T}}\nabla\Psi_{0}^{\mathsf{T}}\right):\nabla\psi\,dy \\
&\quad-\int_{A_{R}}(\overline{p}-\overline{p}_{s_{0},R})\operatorname{div}((B-B_{0})\psi)\,dy-\int_{A_{R}}(q+\zeta-\overline{p}_{s_{0},R})\operatorname{div}(B_{0}\psi)\,dy
=\int_{\mathbb{T}}f(s)\psi(s)\,ds
\end{aligned} ∫ A R ( ∇ u ( A − A 0 ) + ( B − B 0 ) T ∇ u T ∇ Ψ T + B 0 T ∇ u T ( ∇Ψ − ∇ Ψ 0 ) T ) : ∇ ψ d y + ∫ A R ( ( ∇ h + ∇ b ) A 0 + B 0 T ( ∇ h + ∇ b ) T ∇ Ψ 0 T ) : ∇ ψ d y − ∫ A R ( p − p s 0 , R ) div (( B − B 0 ) ψ ) d y − ∫ A R ( q + ζ − p s 0 , R ) div ( B 0 ψ ) d y = ∫ T f ( s ) ψ ( s ) d s
for any ψ \psi ψ with ψ ∣ C ϵ = ψ ( s ) \left.\psi\right|_{\mathcal{C}_{\epsilon}}=\psi(s) ψ ∣ C ϵ = ψ ( s ) and ψ ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 \left.\psi\right|_{\partial A_{R}(s_{0})\setminus\mathcal{C}_{\epsilon}}=0 ψ ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 , we may use that ( h , q ) (h,q) ( h , q ) satisfies (315) to obtain an equation for b b b :
∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ ψ d y = ∫ T ( f ( s ) − f ( s 0 ) ) ψ ( s ) d s − ∫ A R ( ∇ u ‾ ( A − A 0 ) + ( B − B 0 ) T ∇ u ‾ T ∇ Ψ T + B 0 T ∇ u ‾ T ( ∇ Ψ − ∇ Ψ 0 ) T ) : ∇ ψ d y + ∫ A R ( p ‾ − p ‾ s 0 , R ) div ( ( B − B 0 ) ψ ) d y + ∫ A R ( ζ − p ‾ s 0 , R ) div ( B 0 ψ ) d y . (336) \begin{aligned}
\int_{A_{R}}\left(\nabla bA_{0}+B_{0}^{\mathsf{T}}\nabla b^{\mathsf{T}}\nabla\Psi_{0}^{\mathsf{T}}\right):\nabla\psi\,dy
&=\int_{\mathbb{T}}(f(s)-f(s_{0}))\psi(s)\,ds \\
&\quad-\int_{A_{R}}\left(\nabla\overline{u}(A-A_{0})+(B-B_{0})^{\mathsf{T}}\nabla\overline{u}^{\mathsf{T}}\nabla\Psi^{\mathsf{T}}+B_{0}^{\mathsf{T}}\nabla\overline{u}^{\mathsf{T}}(\nabla\Psi-\nabla\Psi_{0})^{\mathsf{T}}\right):\nabla\psi\,dy \\
&\quad+\int_{A_{R}}(\overline{p}-\overline{p}_{s_{0},R})\operatorname{div}((B-B_{0})\psi)\,dy+\int_{A_{R}}(\zeta-\overline{p}_{s_{0},R})\operatorname{div}(B_{0}\psi)\,dy. \tag*{(336)}
\end{aligned} ∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ ψ d y = ∫ T ( f ( s ) − f ( s 0 )) ψ ( s ) d s − ∫ A R ( ∇ u ( A − A 0 ) + ( B − B 0 ) T ∇ u T ∇ Ψ T + B 0 T ∇ u T ( ∇Ψ − ∇ Ψ 0 ) T ) : ∇ ψ d y + ∫ A R ( p − p s 0 , R ) div (( B − B 0 ) ψ ) d y + ∫ A R ( ζ − p s 0 , R ) div ( B 0 ψ ) d y . (336)
If, in addition, we require that the test function ψ \psi ψ satisfies div ( B 0 ψ ) = 0 \operatorname{div}(B_0\psi)=0 div ( B 0 ψ ) = 0 , we obtain the equation
∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ ψ d y = ∫ T ( f ( s ) − f ( s 0 ) ) ψ ( s ) d s − ∫ A R ( ∇ u ‾ ( A − A 0 ) + ( B − B 0 ) T ∇ u ‾ T ∇ Ψ T + B 0 T ∇ u ‾ T ( ∇ Ψ − ∇ Ψ 0 ) T ) : ∇ ψ d y + ∫ A R ( p ‾ − p ‾ s 0 , R ) div ( ( B − B 0 ) ψ ) d y . (337) \int_{A_R} \left(\nabla b A_0+B_0^{\mathsf T}\nabla b^{\mathsf T}\nabla\Psi_0^{\mathsf T}\right):\nabla\psi\,dy=\int_{\mathbb{T}}\left(f(s)-f(s_0)\right)\psi(s)\,ds
-\int_{A_R}\left(\nabla\overline{u}(A-A_0)+(B-B_0)^{\mathsf T}\nabla\overline{u}^{\mathsf T}\nabla\Psi^{\mathsf T}+B_0^{\mathsf T}\nabla\overline{u}^{\mathsf T}\left(\nabla\Psi-\nabla\Psi_0\right)^{\mathsf T}\right):\nabla\psi\,dy
+\int_{A_R}\left(\overline{p}-\overline{p}_{s_0,R}\right)\operatorname{div}\left((B-B_0)\psi\right)\,dy. \tag*{(337)} ∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ ψ d y = ∫ T ( f ( s ) − f ( s 0 ) ) ψ ( s ) d s − ∫ A R ( ∇ u ( A − A 0 ) + ( B − B 0 ) T ∇ u T ∇ Ψ T + B 0 T ∇ u T ( ∇Ψ − ∇ Ψ 0 ) T ) : ∇ ψ d y + ∫ A R ( p − p s 0 , R ) div ( ( B − B 0 ) ψ ) d y . (337)
Noting that b = u ‾ − h b=\overline{u}-h b = u − h already satisfies the boundary conditions b ∣ C ϵ = b ( s ) b|_{C_\epsilon}=b(s) b ∣ C ϵ = b ( s ) and b ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 b|_{\partial A_R(s_0)\setminus C_\epsilon}=0 b ∣ ∂ A R ( s 0 ) ∖ C ϵ = 0 , we would like to use b b b as a test function above. It will be convenient to use (337) , so we must first correct for the divergence. We consider g ∈ H 0 1 ( A R ( s 0 ) ) g\in H^1_0(A_R(s_0)) g ∈ H 0 1 ( A R ( s 0 )) satisfying
div ( B 0 g ) = div ( B 0 b ) in A R ( s 0 ) \operatorname{div}(B_0g)=\operatorname{div}(B_0b)\qquad\text{in }A_R(s_0) div ( B 0 g ) = div ( B 0 b ) in A R ( s 0 )
∥ ∇ g ∥ L 2 ( A R ( s 0 ) ) ≤ c ( κ ∗ , c Γ ) ∥ div ( B 0 b ) ∥ L 2 ( A R ( s 0 ) ) . (338) \|\nabla g\|_{L^2(A_R(s_0))}\leq c(\kappa_*,c_\Gamma)\|\operatorname{div}(B_0b)\|_{L^2(A_R(s_0))}. \tag*{(338)} ∥∇ g ∥ L 2 ( A R ( s 0 )) ≤ c ( κ ∗ , c Γ ) ∥ div ( B 0 b ) ∥ L 2 ( A R ( s 0 )) . (338)
As before, such a g g g exists due to [24 Section III.3] and [52 Appendix A.2.5] . Using that div ( B u ‾ ) = 0 \operatorname{div}(B\overline{u})=0 div ( B u ) = 0 and div ( B 0 h ) = 0 \operatorname{div}(B_0h)=0 div ( B 0 h ) = 0 in A R ( s 0 ) A_R(s_0) A R ( s 0 ) , we may write
div ( B 0 b ) = − div ( ( B − B 0 ) u ‾ ) , \operatorname{div}(B_0b)=-\operatorname{div}\left((B-B_0)\overline{u}\right), div ( B 0 b ) = − div ( ( B − B 0 ) u ) ,
so, making use of the extra regularity assumption (308) on the coefficient B B B , we have
∥ ∇ g ∥ L 2 ( A R ( s 0 ) ) ≤ c ( κ ∗ , c Γ ) ∥ div ( B 0 b ) ∥ L 2 ( A R ( s 0 ) ) ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) R α ∥ u ‾ ∥ H 1 ( A R ( s 0 ) ) . (339) \|\nabla g\|_{L^2(A_R(s_0))}\leq c(\kappa_*,c_\Gamma)\|\operatorname{div}(B_0b)\|_{L^2(A_R(s_0))}\leq c(\epsilon,\|X_{ss}\|_{C^{1,\alpha}},c_\Gamma)R^\alpha\|\overline{u}\|_{H^1(A_R(s_0))}. \tag*{(339)} ∥∇ g ∥ L 2 ( A R ( s 0 )) ≤ c ( κ ∗ , c Γ ) ∥ div ( B 0 b ) ∥ L 2 ( A R ( s 0 )) ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) R α ∥ u ∥ H 1 ( A R ( s 0 )) . (339)
We may now use b − g b-g b − g as a test function in (337) . We first note that
∣ ∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ b d y ∣ ≤ ∫ A R ( ∣ A 0 ∣ + ∣ B 0 ∣ ∣ ∇ Ψ 0 ∣ ) ∣ ∇ b ∣ ∣ ∇ g ∣ d y + ∫ A R ( ∣ A − A 0 ∣ + ∣ B − B 0 ∣ ∣ ∇ Ψ ∣ + ∣ B 0 ∣ ∣ ∇ Ψ − ∇ Ψ 0 ∣ ) ∣ ∇ u ‾ ∣ ( ∣ ∇ b ∣ + ∣ ∇ g ∣ ) d y + ∫ A R ∣ p ‾ − p ‾ s 0 , R ∣ ( ∣ B − B 0 ∣ + ∣ ∇ ( B − B 0 ) ∣ ) ( ∣ ∇ b ∣ + ∣ ∇ g ∣ + ∣ b ∣ + ∣ g ∣ ) d y + ∫ T ∣ f ( s ) − f ( s 0 ) ∣ ∣ b ∣ C ϵ d s . \begin{aligned}
\left|\int_{A_R}\left(\nabla bA_0+B_0^{\mathsf T}\nabla b^{\mathsf T}\nabla\Psi_0^{\mathsf T}\right):\nabla b\,dy\right|
\leq\int_{A_R}\left(|A_0|+|B_0||\nabla\Psi_0|\right)|\nabla b||\nabla g|\,dy \\
&+\int_{A_R}\left(|A-A_0|+|B-B_0||\nabla\Psi|+|B_0||\nabla\Psi-\nabla\Psi_0|\right)|\nabla\overline{u}|\left(|\nabla b|+|\nabla g|\right)\,dy \\
&+\int_{A_R}|\overline{p}-\overline{p}_{s_0,R}|\left(|B-B_0|+|\nabla(B-B_0)|\right)\left(|\nabla b|+|\nabla g|+|b|+|g|\right)\,dy \\
&+\int_{\mathbb{T}}|f(s)-f(s_0)||b|_{C_\epsilon}\,ds.
\end{aligned} ∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ b d y ≤ ∫ A R ( ∣ A 0 ∣ + ∣ B 0 ∣∣∇ Ψ 0 ∣ ) ∣∇ b ∣∣∇ g ∣ d y + ∫ A R ( ∣ A − A 0 ∣ + ∣ B − B 0 ∣∣∇Ψ∣ + ∣ B 0 ∣∣∇Ψ − ∇ Ψ 0 ∣ ) ∣∇ u ∣ ( ∣∇ b ∣ + ∣∇ g ∣ ) d y + ∫ A R ∣ p − p s 0 , R ∣ ( ∣ B − B 0 ∣ + ∣∇ ( B − B 0 ) ∣ ) ( ∣∇ b ∣ + ∣∇ g ∣ + ∣ b ∣ + ∣ g ∣ ) d y + ∫ T ∣ f ( s ) − f ( s 0 ) ∣∣ b ∣ C ϵ d s .
Using the decomposition (318) and the homogeneous Korn inequality (319), we may bound
∫ A R ∣ ∇ b ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ b d y . \int_{A_R}|\nabla b|^2\,dy\leq c(c_\Gamma,\kappa_*)\int_{A_R}\left(\nabla bA_0+B_0^{\mathsf T}\nabla b^{\mathsf T}\nabla\Psi_0^{\mathsf T}\right):\nabla b\,dy. ∫ A R ∣∇ b ∣ 2 d y ≤ c ( c Γ , κ ∗ ) ∫ A R ( ∇ b A 0 + B 0 T ∇ b T ∇ Ψ 0 T ) : ∇ b d y .
Then, using the regularity of the coefficients A A A , B B B , and ∇ Ψ \nabla\Psi ∇Ψ as well as the bound (337) for g g g , we obtain
∫ A R ∣ ∇ b ∣ 2 d y ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) ( R α ∥ ∇ b ∥ L 2 ( A R ) ∥ u ‾ ∥ H 1 ( A R ) + R 2 α ∥ u ‾ ∥ H 1 ( A R ) 2 + R α ∥ p ‾ − p ‾ s 0 , R ∥ L 2 ( A R ) ( ∥ ∇ b ∥ L 2 ( A R ) + R α ∥ u ‾ ∥ H 1 ( A R ) ) + R 1 / 2 + α ∥ f ∥ C 0 , α ∥ b ∣ C ϵ ∥ L 2 ( ∣ s − s 0 ∣ < R ) ) . \int_{A_R}|\nabla b|^2\,dy\leq c(\epsilon,\|X_{ss}\|_{C^{1,\alpha}},c_\Gamma)\left(R^\alpha\|\nabla b\|_{L^2(A_R)}\|\overline{u}\|_{H^1(A_R)}+R^{2\alpha}\|\overline{u}\|_{H^1(A_R)}^2
+R^\alpha\|\overline{p}-\overline{p}_{s_0,R}\|_{L^2(A_R)}\left(\|\nabla b\|_{L^2(A_R)}+R^\alpha\|\overline{u}\|_{H^1(A_R)}\right)+R^{1/2+\alpha}\|f\|_{C^{0,\alpha}}\|b|_{C_\epsilon}\|_{L^2(|s-s_0|<R)}\right). ∫ A R ∣∇ b ∣ 2 d y ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) ( R α ∥∇ b ∥ L 2 ( A R ) ∥ u ∥ H 1 ( A R ) + R 2 α ∥ u ∥ H 1 ( A R ) 2 + R α ∥ p − p s 0 , R ∥ L 2 ( A R ) ( ∥∇ b ∥ L 2 ( A R ) + R α ∥ u ∥ H 1 ( A R ) ) + R 1/2 + α ∥ f ∥ C 0 , α ∥ b ∣ C ϵ ∥ L 2 ( ∣ s − s 0 ∣ < R ) ) .
By a scaling argument (see [59 proof of Lemma 1.8] ), the following trace inequality holds in A R ( s 0 ) A_R(s_0) A R ( s 0 ) :
∥ b ∣ C ϵ ∥ L 2 ( ∣ s − s 0 ∣ < R ) 2 ≤ c ( ϵ ) R ∥ ∇ b ∥ L 2 ( A R ( s 0 ) ) 2 . \|b|_{C_\epsilon}\|_{L^2(|s-s_0|<R)}^2\leq c(\epsilon)R\|\nabla b\|_{L^2(A_R(s_0))}^2. ∥ b ∣ C ϵ ∥ L 2 ( ∣ s − s 0 ∣ < R ) 2 ≤ c ( ϵ ) R ∥∇ b ∥ L 2 ( A R ( s 0 )) 2 .
Using Young’s inequality, we thus obtain the bound
∫ A R ∣ ∇ b ∣ 2 d y ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) R 2 α ( ∥ u ‾ ∥ H 1 ( A R ) + ∥ p ‾ − p ‾ s 0 , R ∥ L 2 ( A R ) 2 + R 2 ∥ f ∥ C 0 , α ( T ) 2 ) . (340) \int_{A_R} |\nabla b|^2\,dy \le c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_\Gamma})R^{2\alpha}\left(\|\overline{u}\|_{H^1(A_R)}+\|\overline{p}-\overline{p}_{s_0,R}\|_{L^2(A_R)}^2+R^2\|f\|_{C^{0,\alpha}(\mathbb{T})}^2\right). \tag*{(340)} ∫ A R ∣∇ b ∣ 2 d y ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) R 2 α ( ∥ u ∥ H 1 ( A R ) + ∥ p − p s 0 , R ∥ L 2 ( A R ) 2 + R 2 ∥ f ∥ C 0 , α ( T ) 2 ) . (340)
Returning to the equation (334) for b b b , we may estimate the pressure term ζ − p ‾ s 0 , R \zeta-\overline{p}_{s_0,R} ζ − p s 0 , R by choosing a test function ψ ∈ H 0 1 ( A R ( s 0 ) ) \psi\in H^1_0(A_R(s_0)) ψ ∈ H 0 1 ( A R ( s 0 )) satisfying
div ( B 0 ψ ) = ζ − p ‾ s 0 , R in A R ( s 0 ) \operatorname{div}(B_0\psi)=\zeta-\overline{p}_{s_0,R}\qquad\text{in }A_R(s_0) div ( B 0 ψ ) = ζ − p s 0 , R in A R ( s 0 )
∥ ∇ ψ ∥ L 2 ( A R ( s 0 ) ) ≤ c ( κ ∗ , C Γ ) ∥ ζ − p ‾ s 0 , R ∥ L 2 ( A R ( s 0 ) ) . \|\nabla\psi\|_{L^2(A_R(s_0))}\le c(\kappa_*,C_\Gamma)\|\zeta-\overline{p}_{s_0,R}\|_{L^2(A_R(s_0))}. ∥∇ ψ ∥ L 2 ( A R ( s 0 )) ≤ c ( κ ∗ , C Γ ) ∥ ζ − p s 0 , R ∥ L 2 ( A R ( s 0 )) .
Using this ψ \psi ψ in (334), we obtain
∫ A R ( s 0 ) ∣ ζ − p ‾ s 0 , R ∣ 2 d y ≤ ∫ A R ( s 0 ) ( ( ∣ A 0 ∣ + ∣ B 0 ∣ ∣ ∇ Ψ 0 ∣ ) ∣ ∇ b ∣ ∣ ∇ ψ ∣ + ∣ p ‾ − p ‾ s 0 , R ∣ ∣ div ( ( B − B 0 ) ψ ) ∣ \int_{A_R(s_0)}|\zeta-\overline{p}_{s_0,R}|^2\,dy\le\int_{A_R(s_0)}\left((|\mathbf{A}_0|+|B_0|\lvert\nabla\Psi_0\rvert)|\nabla b||\nabla\psi|+|\overline{p}-\overline{p}_{s_0,R}||\operatorname{div}((B-B_0)\psi)|
\right. ∫ A R ( s 0 ) ∣ ζ − p s 0 , R ∣ 2 d y ≤ ∫ A R ( s 0 ) ( ( ∣ A 0 ∣ + ∣ B 0 ∣ ∣ ∇ Ψ 0 ∣) ∣∇ b ∣∣∇ ψ ∣ + ∣ p − p s 0 , R ∣∣ div (( B − B 0 ) ψ ) ∣
+ ( ∣ A − A 0 ∣ + ∣ B − B 0 ∣ ∣ ∇ Ψ ∣ + ∣ B 0 ∣ ∣ ∇ Ψ − ∇ Ψ 0 ∣ ) ∣ ∇ u ‾ ∣ ∣ ∇ ψ ∣ ) d y ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) ( ∥ ∇ b ∥ L 2 ( A R ) 2 + R 2 α ∥ ∇ u ‾ ∥ L 2 ( A R ) 2 + R 2 α ∥ p ‾ − p ‾ s 0 , R ∥ L 2 ( A R ) 2 ) . \begin{aligned}
\left.
+(|\mathbf{A}-\mathbf{A}_0|+|B-B_0|\lvert\nabla\Psi\rvert+|B_0|\lvert\nabla\Psi-\nabla\Psi_0\rvert)|\nabla\overline{u}||\nabla\psi|\right)\,dy \\
&\le c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_\Gamma})\left(\|\nabla b\|_{L^2(A_R)}^2+R^{2\alpha}\|\nabla\overline{u}\|_{L^2(A_R)}^2+R^{2\alpha}\|\overline{p}-\overline{p}_{s_0,R}\|_{L^2(A_R)}^2\right).
\end{aligned} + ( ∣ A − A 0 ∣ + ∣ B − B 0 ∣ ∣ ∇Ψ ∣ + ∣ B 0 ∣ ∣ ∇Ψ − ∇ Ψ 0 ∣) ∣∇ u ∣∣∇ ψ ∣ ) d y ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) ( ∥∇ b ∥ L 2 ( A R ) 2 + R 2 α ∥∇ u ∥ L 2 ( A R ) 2 + R 2 α ∥ p − p s 0 , R ∥ L 2 ( A R ) 2 ) .
Combining with (338), we have
∫ A R ( s 0 ) ∣ ∇ b ∣ 2 d y + ∫ A R ( s 0 ) ∣ ζ − p ‾ s 0 , R ∣ 2 d y ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) ( R 2 α ∥ u ‾ ∥ H 1 ( A R ) 2 + R 2 α ∥ p ‾ − p ‾ s 0 , R ∥ L 2 ( A R ) 2 + R 2 + 2 α ∥ f ∥ C 0 , α ( T ) 2 ) . (341) \begin{aligned}
\int_{A_R(s_0)}|\nabla b|^2\,dy+\int_{A_R(s_0)}|\zeta-\overline{p}_{s_0,R}|^2\,dy \\
&\le c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_\Gamma})\left(R^{2\alpha}\|\overline{u}\|_{H^1(A_R)}^2+R^{2\alpha}\|\overline{p}-\overline{p}_{s_0,R}\|_{L^2(A_R)}^2+R^{2+2\alpha}\|f\|_{C^{0,\alpha}(\mathbb{T})}^2\right). \tag*{(341)}
\end{aligned} ∫ A R ( s 0 ) ∣∇ b ∣ 2 d y + ∫ A R ( s 0 ) ∣ ζ − p s 0 , R ∣ 2 d y ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) ( R 2 α ∥ u ∥ H 1 ( A R ) 2 + R 2 α ∥ p − p s 0 , R ∥ L 2 ( A R ) 2 + R 2 + 2 α ∥ f ∥ C 0 , α ( T ) 2 ) . (341)
Next, using Proposition 6.3 and recalling that A ρ A_\rho A ρ scales like ρ 2 \rho^2 ρ 2 for small ρ \rho ρ , for ρ < R / 2 \rho<R/2 ρ < R /2 , we have that h = u ‾ − b h=\overline{u}-b h = u − b and q = p ‾ − ζ q=\overline{p}-\zeta q = p − ζ satisfy
∫ A ρ ( s 0 ) ( ∣ ∇ h ∣ 2 + ∣ h ∣ 2 + ∣ q ∣ 2 ) d y ≤ c ρ 2 ∫ 0 2 π sup ∣ s − s 0 ∣ ≤ ρ , 0 ≤ r ≤ ρ ( ∣ ∇ h ∣ 2 + ∣ h ∣ 2 + ∣ q ∣ 2 ) d θ ≤ c ρ 2 ∫ 0 2 π sup ∣ s − s 0 ∣ ≤ R / 2 , 0 ≤ r ≤ R / 2 ( ∣ ∇ h ∣ 2 + ∣ h ∣ 2 + ∣ q ∣ 2 ) d θ ≤ c ( ϵ , κ ∗ , R ) ρ 2 ( ∥ h ∥ H 3 ( A R / 2 ( s 0 ) ) 2 + ∥ q ∥ H 2 ( A R / 2 ( s 0 ) ) 2 ) ≤ c ( ϵ , κ ∗ ) ( ρ R ) 2 ( ∥ h ∥ H 1 ( A R ( s 0 ) ) 2 + ∥ q ∥ L 2 ( A R ( s 0 ) ) 2 ) , \begin{aligned}
\int_{A_\rho(s_0)}\left(|\nabla h|^2+|h|^2+|q|^2\right)\,dy\le c\rho^2\int_0^{2\pi}\sup_{|s-s_0|\le\rho,\,0\le r\le\rho}\left(|\nabla h|^2+|h|^2+|q|^2\right)\,d\theta\\
&\le c\rho^2\int_0^{2\pi}\sup_{|s-s_0|\le R/2,\,0\le r\le R/2}\left(|\nabla h|^2+|h|^2+|q|^2\right)\,d\theta\\
&\le c(\epsilon,\kappa_*,R)\rho^2\left(\|h\|_{H^3(A_{R/2}(s_0))}^2+\|q\|_{H^2(A_{R/2}(s_0))}^2\right) \\
&\le c(\epsilon,\kappa_*)\left(\frac{\rho}{R}\right)^2\left(\|h\|_{H^1(A_R(s_0))}^2+\|q\|_{L^2(A_R(s_0))}^2\right),
\end{aligned} ∫ A ρ ( s 0 ) ( ∣∇ h ∣ 2 + ∣ h ∣ 2 + ∣ q ∣ 2 ) d y ≤ c ρ 2 ∫ 0 2 π ∣ s − s 0 ∣ ≤ ρ , 0 ≤ r ≤ ρ sup ( ∣∇ h ∣ 2 + ∣ h ∣ 2 + ∣ q ∣ 2 ) d θ ≤ c ρ 2 ∫ 0 2 π ∣ s − s 0 ∣ ≤ R /2 , 0 ≤ r ≤ R /2 sup ( ∣∇ h ∣ 2 + ∣ h ∣ 2 + ∣ q ∣ 2 ) d θ ≤ c ( ϵ , κ ∗ , R ) ρ 2 ( ∥ h ∥ H 3 ( A R /2 ( s 0 )) 2 + ∥ q ∥ H 2 ( A R /2 ( s 0 )) 2 ) ≤ c ( ϵ , κ ∗ ) ( R ρ ) 2 ( ∥ h ∥ H 1 ( A R ( s 0 )) 2 + ∥ q ∥ L 2 ( A R ( s 0 )) 2 ) ,
by a similar scaling argument to the proof of Lemma 6.6. We thus have
∥ u ‾ ∥ H 1 ( A ρ ) 2 + ∥ p ‾ − p ‾ s 0 , ρ ∥ L 2 ( A ρ ) 2 ≤ 2 ( ∥ h ∥ H 1 ( A ρ ) 2 + ∥ b ∥ H 1 ( A ρ ) 2 + ∥ q ∥ L 2 ( A ρ ) 2 + ∥ ζ − p ‾ s 0 , ρ ∥ L 2 ( A ρ ) 2 ) ≤ c ( ϵ , κ ∗ ) ( ρ R ) 2 ( ∥ h ∥ H 1 ( A R ) 2 + ∥ q ∥ L 2 ( A R ) 2 ) + 2 ∥ b ∥ H 1 ( A ρ ) 2 + 2 ∥ ζ − p ‾ s 0 , ρ ∥ L 2 ( A ρ ) 2 ≤ c ( ϵ , κ ∗ ) ( ρ R ) 2 ( ∥ u ‾ ∥ H 1 ( A R ) 2 + ∥ p ‾ − p ‾ s 0 , R ∥ L 2 ( A R ) 2 ) + c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) ( R 2 α ( ∥ u ‾ ∥ H 1 ( A R ) 2 + ∥ p ‾ − p ‾ s 0 , R ∥ L 2 ( A R ) 2 ) + R 2 + 2 α ∥ f ∥ C 0 , α ( T ) 2 ) , (342) \begin{aligned}
\|\overline{u}\|_{H^1(A_\rho)}^2+\|\overline{p}-\overline{p}_{s_0,\rho}\|_{L^2(A_\rho)}^2\le2\left(\|h\|_{H^1(A_\rho)}^2+\|b\|_{H^1(A_\rho)}^2+\|q\|_{L^2(A_\rho)}^2+\|\zeta-\overline{p}_{s_0,\rho}\|_{L^2(A_\rho)}^2\right) \\
&\le c(\epsilon,\kappa_*)\left(\frac{\rho}{R}\right)^2\left(\|h\|_{H^1(A_R)}^2+\|q\|_{L^2(A_R)}^2\right)+2\|b\|_{H^1(A_\rho)}^2+2\|\zeta-\overline{p}_{s_0,\rho}\|_{L^2(A_\rho)}^2 \\
&\le c(\epsilon,\kappa_*)\left(\frac{\rho}{R}\right)^2\left(\|\overline{u}\|_{H^1(A_R)}^2+\|\overline{p}-\overline{p}_{s_0,R}\|_{L^2(A_R)}^2\right) \\
&+c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_\Gamma})\left(R^{2\alpha}\left(\|\overline{u}\|_{H^1(A_R)}^2+\|\overline{p}-\overline{p}_{s_0,R}\|_{L^2(A_R)}^2\right)+R^{2+2\alpha}\|f\|_{C^{0,\alpha}(\mathbb{T})}^2\right), \tag*{(342)}
\end{aligned} ∥ u ∥ H 1 ( A ρ ) 2 + ∥ p − p s 0 , ρ ∥ L 2 ( A ρ ) 2 ≤ 2 ( ∥ h ∥ H 1 ( A ρ ) 2 + ∥ b ∥ H 1 ( A ρ ) 2 + ∥ q ∥ L 2 ( A ρ ) 2 + ∥ ζ − p s 0 , ρ ∥ L 2 ( A ρ ) 2 ) ≤ c ( ϵ , κ ∗ ) ( R ρ ) 2 ( ∥ h ∥ H 1 ( A R ) 2 + ∥ q ∥ L 2 ( A R ) 2 ) + 2∥ b ∥ H 1 ( A ρ ) 2 + 2∥ ζ − p s 0 , ρ ∥ L 2 ( A ρ ) 2 ≤ c ( ϵ , κ ∗ ) ( R ρ ) 2 ( ∥ u ∥ H 1 ( A R ) 2 + ∥ p − p s 0 , R ∥ L 2 ( A R ) 2 ) + c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) ( R 2 α ( ∥ u ∥ H 1 ( A R ) 2 + ∥ p − p s 0 , R ∥ L 2 ( A R ) 2 ) + R 2 + 2 α ∥ f ∥ C 0 , α ( T ) 2 ) , (342)
where we have used the estimate (339).
We may now apply a useful proposition from [28 ] :
Proposition 6.7 (Lemma 5.13 from [28 ] ). Given m > ν > 0 m>\nu>0 m > ν > 0 and a nondecreasing function Ξ : R + → R + \Xi:\mathbb{R}^{+}\to\mathbb{R}^{+} Ξ : R + → R + satisfying
Ξ ( ρ ) ≤ c 1 ( ρ R ) m + ω ) Ξ ( R ) + c 2 R ν \Xi(\rho)\leq c_{1}(\frac{\rho}{R})^{m}+\omega)\Xi(R)+c_{2}R^{\nu} Ξ ( ρ ) ≤ c 1 ( R ρ ) m + ω ) Ξ ( R ) + c 2 R ν for all 0 < ρ ≤ R ≤ r ∗ 2 0<\rho\leq R\leq\frac{r_{*}}{2} 0 < ρ ≤ R ≤ 2 r ∗ , for ω \omega ω sufficiently small (depending on c 1 , m , ν c_{1},m,\nu c 1 , m , ν ), we in fact have
Ξ ( ρ ) ≤ c 3 ( Ξ ( R ) R ν + c 2 ) ρ ν (343) \Xi(\rho)\leq c_{3}\left(\frac{\Xi(R)}{R^{\nu}}+c_{2}\right)\rho^{\nu} \tag*{(343)} Ξ ( ρ ) ≤ c 3 ( R ν Ξ ( R ) + c 2 ) ρ ν (343) for all 0 ≤ ρ ≤ R ≤ r ∗ 2 0\leq\rho\leq R\leq\frac{r_{*}}{2} 0 ≤ ρ ≤ R ≤ 2 r ∗ .
Applying Proposition 6.7 to (340) with Ξ ( ρ ) = ∥ u ˉ ∥ H 1 ( A ρ ) 2 + ∥ p ˉ − p ˉ s 0 , ρ ∥ L 2 ( A ρ ) 2 \Xi(\rho)=\|\bar{u}\|_{H^{1}(A_{\rho})}^{2}+\|\bar{p}-\bar{p}_{s_{0},\rho}\|_{L^{2}(A_{\rho})}^{2} Ξ ( ρ ) = ∥ u ˉ ∥ H 1 ( A ρ ) 2 + ∥ p ˉ − p ˉ s 0 , ρ ∥ L 2 ( A ρ ) 2 , m = 2 m=2 m = 2 , and ν = 2 − δ \nu=2-\delta ν = 2 − δ for δ > 0 \delta>0 δ > 0 , for R R R sufficiently small we obtain
∥ u ˉ ∥ H 1 ( A ρ ( s 0 ) ) 2 + ∥ p ˉ ∥ L 2 ( A ρ ( s 0 ) ) 2 ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) ( R − ( 2 − δ ) ( ∥ u ˉ ∥ H 1 ( A R ) + ∥ p ˉ − p ˉ s 0 , R ∥ L 2 ( A R ) 2 ) + R 2 α + δ ∥ f ∥ C 0 , α ( T ) 2 ) ρ 2 − δ (344) \begin{aligned}
\|\bar{u}\|_{H^{1}(A_{\rho}(s_{0}))}^{2}+\|\bar{p}\|_{L^{2}(A_{\rho}(s_{0}))}^{2} \\
&\leq c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_{\Gamma}})\left(R^{-(2-\delta)}\left(\|\bar{u}\|_{H^{1}(A_{R})}+\|\bar{p}-\bar{p}_{s_{0},R}\|_{L^{2}(A_{R})}^{2}\right)+R^{2\alpha+\delta}\|f\|_{C^{0,\alpha}(\mathcal{T})}^{2}\right)\rho^{2-\delta} \tag*{(344)}
\end{aligned} ∥ u ˉ ∥ H 1 ( A ρ ( s 0 )) 2 + ∥ p ˉ ∥ L 2 ( A ρ ( s 0 )) 2 ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) ( R − ( 2 − δ ) ( ∥ u ˉ ∥ H 1 ( A R ) + ∥ p ˉ − p ˉ s 0 , R ∥ L 2 ( A R ) 2 ) + R 2 α + δ ∥ f ∥ C 0 , α ( T ) 2 ) ρ 2 − δ (344)
for 0 ≤ ρ ≤ R ≤ r ∗ 2 0\leq\rho\leq R\leq\frac{r_{*}}{2} 0 ≤ ρ ≤ R ≤ 2 r ∗ .
We may now estimate the oscillation of ∂ s u ˉ \partial_{s}\bar{u} ∂ s u ˉ over A ρ ( s 0 ) A_{\rho}(s_{0}) A ρ ( s 0 ) . For 0 < ρ ≤ R ′ ≤ R ≤ r ∗ 2 0<\rho\leq R'\leq R\leq\frac{r_{*}}{2} 0 < ρ ≤ R ′ ≤ R ≤ 2 r ∗ , we have
∫ A ρ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , ρ ∣ 2 d y ≤ ∫ A ρ ( s 0 ) 2 ( ∣ ∂ s h − ( ∂ s h ) s 0 , ρ ∣ 2 + ∣ ∂ s b − ( ∂ s b ) s 0 , ρ ∣ 2 ) d y ≤ c ( ϵ , κ ∗ , C Γ ) ( ρ R ′ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , R ′ ∣ 2 d y + 4 ∫ A ρ ( s 0 ) ∣ ∇ b ∣ 2 d y ≤ c ( ϵ , κ ∗ , C Γ ) ( ρ R ′ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ′ ∣ 2 d y + c ( ϵ , κ ∗ ) ∫ A R ′ ( s 0 ) ∣ ∇ b ∣ 2 d y ≤ c ( ϵ , κ ∗ , C Γ ) ( ρ R ′ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ′ ∣ 2 d y + c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) ( ( R ′ ) 2 α ( ∥ u ˉ ∥ H 1 ( A R ′ ) + ∥ p ˉ − p ˉ s 0 , R ′ ∥ L 2 ( A R ′ ) 2 ) + ( R ′ ) 2 + 2 α ∥ f ∥ C 0 , α ( T ) 2 ) ≤ c ( ϵ , κ ∗ , C Γ ) ( ρ R ′ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ′ ∣ 2 d y + c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) ( R ′ ) 2 + 2 α − δ M , (345) \begin{aligned}
\int_{A_{\rho}(s_{0})}|\partial_{s}\bar{u}-(\partial_{s}\bar{u})_{s_{0},\rho}|^{2}\,dy\leq\int_{A_{\rho}(s_{0})}2\left(|\partial_{s}h-(\partial_{s}h)_{s_{0},\rho}|^{2}+|\partial_{s}b-(\partial_{s}b)_{s_{0},\rho}|^{2}\right)\,dy \\
&\leq c(\epsilon,\kappa_{*},C_{\Gamma})\left(\frac{\rho}{R'}\right)^{4}\int_{A_{R'}(s_{0})}|\partial_{s}h-(\partial_{s}h)_{s_{0},R'}|^{2}\,dy+4\int_{A_{\rho}(s_{0})}|\nabla b|^{2}\,dy \\
&\leq c(\epsilon,\kappa_{*},C_{\Gamma})\left(\frac{\rho}{R'}\right)^{4}\int_{A_{R'}(s_{0})}|\partial_{s}\bar{u}-(\partial_{s}\bar{u})_{s_{0},R'}|^{2}\,dy+c(\epsilon,\kappa_{*})\int_{A_{R'}(s_{0})}|\nabla b|^{2}\,dy \\
&\leq c(\epsilon,\kappa_{*},C_{\Gamma})\left(\frac{\rho}{R'}\right)^{4}\int_{A_{R'}(s_{0})}|\partial_{s}\bar{u}-(\partial_{s}\bar{u})_{s_{0},R'}|^{2}\,dy \\
&+c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_{\Gamma}})\left((R')^{2\alpha}\left(\|\bar{u}\|_{H^{1}(A_{R'})}+\|\bar{p}-\bar{p}_{s_{0},R'}\|_{L^{2}(A_{R'})}^{2}\right)+(R')^{2+2\alpha}\|f\|_{C^{0,\alpha}(\mathcal{T})}^{2}\right) \\
&\leq c(\epsilon,\kappa_{*},C_{\Gamma})\left(\frac{\rho}{R'}\right)^{4}\int_{A_{R'}(s_{0})}|\partial_{s}\bar{u}-(\partial_{s}\bar{u})_{s_{0},R'}|^{2}\,dy+c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_{\Gamma}})(R')^{2+2\alpha-\delta}M, \tag*{(345)}
\end{aligned} ∫ A ρ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , ρ ∣ 2 d y ≤ ∫ A ρ ( s 0 ) 2 ( ∣ ∂ s h − ( ∂ s h ) s 0 , ρ ∣ 2 + ∣ ∂ s b − ( ∂ s b ) s 0 , ρ ∣ 2 ) d y ≤ c ( ϵ , κ ∗ , C Γ ) ( R ′ ρ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s h − ( ∂ s h ) s 0 , R ′ ∣ 2 d y + 4 ∫ A ρ ( s 0 ) ∣∇ b ∣ 2 d y ≤ c ( ϵ , κ ∗ , C Γ ) ( R ′ ρ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ′ ∣ 2 d y + c ( ϵ , κ ∗ ) ∫ A R ′ ( s 0 ) ∣∇ b ∣ 2 d y ≤ c ( ϵ , κ ∗ , C Γ ) ( R ′ ρ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ′ ∣ 2 d y + c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) ( ( R ′ ) 2 α ( ∥ u ˉ ∥ H 1 ( A R ′ ) + ∥ p ˉ − p ˉ s 0 , R ′ ∥ L 2 ( A R ′ ) 2 ) + ( R ′ ) 2 + 2 α ∥ f ∥ C 0 , α ( T ) 2 ) ≤ c ( ϵ , κ ∗ , C Γ ) ( R ′ ρ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ′ ∣ 2 d y + c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) ( R ′ ) 2 + 2 α − δ M , (345)
where M = R − ( 2 − δ ) ( ∥ u ˉ ∥ H 1 ( A R ) + ∥ p ˉ − p ˉ s 0 , R ∥ L 2 ( A R ) 2 ) + R δ ∥ f ∥ C 0 , α ( T ) 2 M=R^{-(2-\delta)}\left(\|\bar{u}\|_{H^{1}(A_{R})}+\|\bar{p}-\bar{p}_{s_{0},R}\|_{L^{2}(A_{R})}^{2}\right)+R^{\delta}\|f\|_{C^{0,\alpha}(\mathcal{T})}^{2} M = R − ( 2 − δ ) ( ∥ u ˉ ∥ H 1 ( A R ) + ∥ p ˉ − p ˉ s 0 , R ∥ L 2 ( A R ) 2 ) + R δ ∥ f ∥ C 0 , α ( T ) 2 .
Using Proposition 6.7 with Ξ ( ρ ) = ∫ A ρ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , ρ ∣ 2 d y \Xi(\rho)=\int_{A_{\rho}(s_{0})}|\partial_{s}\bar{u}-(\partial_{s}\bar{u})_{s_{0},\rho}|^{2}\,dy Ξ ( ρ ) = ∫ A ρ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , ρ ∣ 2 d y , m = 4 m=4 m = 4 , and ν = 2 + 2 α − δ \nu=2+2\alpha-\delta ν = 2 + 2 α − δ , we obtain
∫ A ρ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , ρ ∣ 2 d y ≤ ρ 2 + 2 α − δ c ( ϵ , ∥ X s s ∥ C 1 , α , C Γ ) ( R − ( 2 + 2 α − δ ) ∫ A R ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ∣ 2 d y + M ) \begin{aligned}
\int_{A_{\rho}(s_{0})}|\partial_{s}\bar{u}-(\partial_{s}\bar{u})_{s_{0},\rho}|^{2}\,dy \\
&\leq\rho^{2+2\alpha-\delta}c(\epsilon,\|\mathbf{X}_{ss}\|_{C^{1,\alpha},C_{\Gamma}})\left(R^{-(2+2\alpha-\delta)}\int_{A_{R}(s_{0})}|\partial_{s}\bar{u}-(\partial_{s}\bar{u})_{s_{0},R}|^{2}\,dy+M\right)
\end{aligned} ∫ A ρ ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , ρ ∣ 2 d y ≤ ρ 2 + 2 α − δ c ( ϵ , ∥ X ss ∥ C 1 , α , C Γ ) ( R − ( 2 + 2 α − δ ) ∫ A R ( s 0 ) ∣ ∂ s u ˉ − ( ∂ s u ˉ ) s 0 , R ∣ 2 d y + M )
for any 0 ≤ ρ ≤ R ≤ r ∗ 2 0\leq\rho\leq R\leq\frac{r_{*}}{2} 0 ≤ ρ ≤ R ≤ 2 r ∗ . Using Lemma 6.2, we have ∂ s u ˉ ∈ C 0 , α − δ 2 \partial_{s}\bar{u}\in C^{0,\alpha-\frac{\delta}{2}} ∂ s u ˉ ∈ C 0 , α − 2 δ for δ > 0 \delta>0 δ > 0 small. Furthermore, using that u ˉ ∣ Ψ ( Γ ϵ ) \bar{u}|_{\Psi(\Gamma_{\epsilon})} u ˉ ∣ Ψ ( Γ ϵ ) is independent of θ \theta θ , we also have ∂ θ u ˉ ∣ Ψ ( Γ ϵ ) ∩ ∂ A R ( s 0 ) ∈ C 0 , α − δ 2 \partial_{\theta}\bar{u}|_{\Psi(\Gamma_{\epsilon})\cap\partial A_{R}(s_{0})}\in C^{0,\alpha-\frac{\delta}{2}} ∂ θ u ˉ ∣ Ψ ( Γ ϵ ) ∩ ∂ A R ( s 0 ) ∈ C 0 , α − 2 δ . Using classical elliptic regularity theory, we then have ( u ‾ , p ‾ − p ‾ s 0 , R ) ∈ C 1 , α − δ 2 ( A R ( s 0 ) ) × C 0 , α − δ 2 ( A R ( s 0 ) ) (\overline{u},\overline{p}-\overline{p}_{s_0,R})\in C^{1,\alpha-\frac{\delta}{2}}(A_R(s_0))\times C^{0,\alpha-\frac{\delta}{2}}(A_R(s_0)) ( u , p − p s 0 , R ) ∈ C 1 , α − 2 δ ( A R ( s 0 )) × C 0 , α − 2 δ ( A R ( s 0 )) with
∥ u ‾ ∥ C 1 , α − δ 2 ( A R ( s 0 ) ) + ∥ p ‾ − p ‾ s 0 , R ∥ C 0 , α − δ 2 ( A R ( s 0 ) ) ≤ c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) ( ∥ u ‾ ∥ H 1 ( Ψ ( O r ∗ ) ) + ∥ p ‾ ∥ L 2 ( Ψ ( O r ∗ ) ) + ∥ f ∥ C 0 , α ( T ) ) . (346) \|\overline{u}\|_{C^{1,\alpha-\frac{\delta}{2}}(A_R(s_0))}+\|\overline{p}-\overline{p}_{s_0,R}\|_{C^{0,\alpha-\frac{\delta}{2}}(A_R(s_0))}
\leq c(\epsilon,\|\boldsymbol{X}_{ss}\|_{C^{1,\alpha}},c_{\Gamma})\left(\|\overline{u}\|_{H^1(\Psi(O_{r_*}))}+\|\overline{p}\|_{L^2(\Psi(O_{r_*}))}+\|\boldsymbol{f}\|_{C^{0,\alpha}(\mathbb{T})}\right). \tag*{(346)} ∥ u ∥ C 1 , α − 2 δ ( A R ( s 0 )) + ∥ p − p s 0 , R ∥ C 0 , α − 2 δ ( A R ( s 0 )) ≤ c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) ( ∥ u ∥ H 1 ( Ψ ( O r ∗ )) + ∥ p ∥ L 2 ( Ψ ( O r ∗ )) + ∥ f ∥ C 0 , α ( T ) ) . (346)
In particular, we thus have ( ∇ u ‾ , p ‾ − p ‾ s 0 , R ) ∈ L ∞ ( A R ( s 0 ) ) × L ∞ ( A R ( s 0 ) ) (\nabla\overline{u},\overline{p}-\overline{p}_{s_0,R})\in L^\infty(A_R(s_0))\times L^\infty(A_R(s_0)) ( ∇ u , p − p s 0 , R ) ∈ L ∞ ( A R ( s 0 )) × L ∞ ( A R ( s 0 )) and we may estimate
∫ A R ( s 0 ) ( ∣ ∇ u ‾ ∣ 2 + ∣ u ‾ ∣ 2 + ∣ p ‾ − p ‾ s 0 , R ∣ 2 ) d y ≤ c R 2 sup y ∈ A R ( s 0 ) ( ∣ ∇ u ‾ ∣ 2 + ∣ u ‾ ∣ 2 + ∣ p ‾ − p ‾ s 0 , R ∣ 2 ) . \int_{A_R(s_0)}\left(|\nabla\overline{u}|^2+|\overline{u}|^2+|\overline{p}-\overline{p}_{s_0,R}|^2\right)\,dy\leq cR^2\sup_{y\in A_R(s_0)}\left(|\nabla\overline{u}|^2+|\overline{u}|^2+|\overline{p}-\overline{p}_{s_0,R}|^2\right). ∫ A R ( s 0 ) ( ∣∇ u ∣ 2 + ∣ u ∣ 2 + ∣ p − p s 0 , R ∣ 2 ) d y ≤ c R 2 y ∈ A R ( s 0 ) sup ( ∣∇ u ∣ 2 + ∣ u ∣ 2 + ∣ p − p s 0 , R ∣ 2 ) .
The estimate (345) may thus be improved to
∫ A ρ ( s 0 ) ∣ ∂ s u ‾ − ( ∂ s u ‾ ) s 0 , ρ ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ( ρ R ′ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u ‾ − ( ∂ s u ‾ ) s 0 , R ′ ∣ 2 d y + c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) ( R ′ ) 2 + 2 α M 2 , (347) \int_{A_\rho(s_0)}|\partial_s\overline{u}-(\partial_s\overline{u})_{s_0,\rho}|^2\,dy
\leq c(\epsilon,\kappa_*,c_{\Gamma})\left(\frac{\rho}{R'}\right)^4\int_{A_{R'}(s_0)}|\partial_s\overline{u}-(\partial_s\overline{u})_{s_0,R'}|^2\,dy+c(\epsilon,\|\boldsymbol{X}_{ss}\|_{C^{1,\alpha}},c_{\Gamma})(R')^{2+2\alpha}M_2, \tag*{(347)} ∫ A ρ ( s 0 ) ∣ ∂ s u − ( ∂ s u ) s 0 , ρ ∣ 2 d y ≤ c ( ϵ , κ ∗ , c Γ ) ( R ′ ρ ) 4 ∫ A R ′ ( s 0 ) ∣ ∂ s u − ( ∂ s u ) s 0 , R ′ ∣ 2 d y + c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) ( R ′ ) 2 + 2 α M 2 , (347)
where M 2 = sup y ∈ A R ( s 0 ) ( ∣ ∇ u ‾ ∣ 2 + ∣ u ‾ ∣ 2 + ∣ p ‾ − p ‾ s 0 , R ∣ 2 ) + ∥ f ∥ C 0 , α ( T ) 2 M_2=\sup_{y\in A_R(s_0)}\left(|\nabla\overline{u}|^2+|\overline{u}|^2+|\overline{p}-\overline{p}_{s_0,R}|^2\right)+\|\boldsymbol{f}\|_{C^{0,\alpha}(\mathbb{T})}^2 M 2 = sup y ∈ A R ( s 0 ) ( ∣∇ u ∣ 2 + ∣ u ∣ 2 + ∣ p − p s 0 , R ∣ 2 ) + ∥ f ∥ C 0 , α ( T ) 2 . By Proposition 6.7, we then have
∫ A ρ ( s 0 ) ∣ ∂ s u ‾ − ( ∂ s u ‾ ) s 0 , ρ ∣ 2 d y ≤ ρ 2 α c ( ϵ , ∥ X s s ∥ C 1 , α , c Γ ) ( R − 2 α ∫ A R ( s 0 ) ∣ ∂ s u ‾ − ( ∂ s u ‾ ) s 0 , ρ ∣ 2 d y + M 2 ) . (348) \int_{A_\rho(s_0)}|\partial_s\overline{u}-(\partial_s\overline{u})_{s_0,\rho}|^2\,dy
\leq\rho^{2\alpha}c(\epsilon,\|\boldsymbol{X}_{ss}\|_{C^{1,\alpha}},c_{\Gamma})\left(R^{-2\alpha}\int_{A_R(s_0)}|\partial_s\overline{u}-(\partial_s\overline{u})_{s_0,\rho}|^2\,dy+M_2\right). \tag*{(348)} ∫ A ρ ( s 0 ) ∣ ∂ s u − ( ∂ s u ) s 0 , ρ ∣ 2 d y ≤ ρ 2 α c ( ϵ , ∥ X ss ∥ C 1 , α , c Γ ) ( R − 2 α ∫ A R ( s 0 ) ∣ ∂ s u − ( ∂ s u ) s 0 , ρ ∣ 2 d y + M 2 ) . (348)
By covering the region Ψ ( O r ∗ ) \Psi(O_{r_*}) Ψ ( O r ∗ ) with annuli A R ( s 0 ) A_R(s_0) A R ( s 0 ) , s 0 ∈ T s_0\in\mathbb{T} s 0 ∈ T , we thus have that ∂ s u ‾ \partial_s\overline{u} ∂ s u belongs to the Campanato-type space A 2 , α \mathcal{A}^{2,\alpha} A 2 , α (313) . Using Lemma 6.2 along with u = u ‾ ∘ Ψ u=\overline{u}\circ\Psi u = u ∘ Ψ , we obtain that u ∣ Γ ϵ = u ( s ) ∈ C 1 , α ( T ) u|_{\Gamma_\epsilon}=u(s)\in C^{1,\alpha}(\mathbb{T}) u ∣ Γ ϵ = u ( s ) ∈ C 1 , α ( T ) . The estimate (75) follows from the bounds (346) and (305) .
Finally, we show the Lipschitz estimate (76) . Given two nearby filaments with centerlines X ( a ) \boldsymbol{X}^{(a)} X ( a ) , X ( b ) \boldsymbol{X}^{(b)} X ( b ) satisfying Lemma 1.1, we aim to bound the difference v ( a ) ( s ) − v ( b ) ( s ) \boldsymbol{v}^{(a)}(s)-\boldsymbol{v}^{(b)}(s) v ( a ) ( s ) − v ( b ) ( s ) for v ( a ) ( s ) , v ( b ) ( s ) \boldsymbol{v}^{(a)}(s),\boldsymbol{v}^{(b)}(s) v ( a ) ( s ) , v ( b ) ( s ) satisfying
( L ϵ ( a ) ) − 1 [ v ( a ) ] ( s ) = f ( s ) = ( L ϵ ( b ) ) − 1 [ v ( b ) ] ( s ) . (\mathcal{L}_{\epsilon}^{(a)})^{-1}[\boldsymbol{v}^{(a)}](s)=\boldsymbol{f}(s)=(\mathcal{L}_{\epsilon}^{(b)})^{-1}[\boldsymbol{v}^{(b)}](s). ( L ϵ ( a ) ) − 1 [ v ( a ) ] ( s ) = f ( s ) = ( L ϵ ( b ) ) − 1 [ v ( b ) ] ( s ) .
Rearranging, we may write
v ( a ) − v ( b ) = L ϵ ( a ) [ ( L ϵ ( b ) ) − 1 [ v ( b ) ] − ( L ϵ ( a ) ) − 1 [ v ( b ) ] ] . \boldsymbol{v}^{(a)}-\boldsymbol{v}^{(b)}=\mathcal{L}_{\epsilon}^{(a)}\left[(\mathcal{L}_{\epsilon}^{(b)})^{-1}[\boldsymbol{v}^{(b)}]-(\mathcal{L}_{\epsilon}^{(a)})^{-1}[\boldsymbol{v}^{(b)}]\right]. v ( a ) − v ( b ) = L ϵ ( a ) [ ( L ϵ ( b ) ) − 1 [ v ( b ) ] − ( L ϵ ( a ) ) − 1 [ v ( b ) ] ] .
We may then apply the estimate (75) for L ϵ ( a ) \mathcal{L}_{\epsilon}^{(a)} L ϵ ( a ) and L ϵ ( b ) \mathcal{L}_{\epsilon}^{(b)} L ϵ ( b ) , and the Lipschitz estimate (31) for L ϵ − 1 \mathcal{L}_{\epsilon}^{-1} L ϵ − 1 to obtain
∥ v ( a ) − v ( b ) ∥ C 1 , α ( T ) = ∥ L ϵ ( a ) [ ( L ϵ ( b ) ) − 1 [ v ( b ) ] − ( L ϵ ( a ) ) − 1 [ v ( b ) ] ] ∥ C 1 , α ( T ) ≤ c ( ϵ , ∥ X s s ( a ) ∥ C 1 , α , c Γ ) ∥ ( L ϵ ( b ) ) − 1 [ v ( b ) ] − ( L ϵ ( a ) ) − 1 [ v ( b ) ] ∥ C 0 , α ( T ) ≤ c ( ϵ , ∥ X s s ( a ) ∥ C 1 , α , κ ∗ , α + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ v ( b ) ∥ C 1 , α ( T ) ≤ c ( ϵ , ∥ X s s ( a ) ∥ C 1 , α , ∥ X s s ( b ) ∥ C 1 , α , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ f ∥ C 0 , α ( T ) . \begin{aligned}
\|\boldsymbol{v}^{(a)}-\boldsymbol{v}^{(b)}\|_{C^{1,\alpha}(\mathbb{T})}
=\left\|\mathcal{L}_{\epsilon}^{(a)}\left[(\mathcal{L}_{\epsilon}^{(b)})^{-1}[\boldsymbol{v}^{(b)}]-(\mathcal{L}_{\epsilon}^{(a)})^{-1}[\boldsymbol{v}^{(b)}]\right]\right\|_{C^{1,\alpha}(\mathbb{T})} \\
&\leq c(\epsilon,\|\boldsymbol{X}_{ss}^{(a)}\|_{C^{1,\alpha}},c_{\Gamma})\left\|(\mathcal{L}_{\epsilon}^{(b)})^{-1}[\boldsymbol{v}^{(b)}]-(\mathcal{L}_{\epsilon}^{(a)})^{-1}[\boldsymbol{v}^{(b)}]\right\|_{C^{0,\alpha}(\mathbb{T})} \\
&\leq c(\epsilon,\|\boldsymbol{X}_{ss}^{(a)}\|_{C^{1,\alpha}},\kappa_{*,\alpha+}^{(b)},c_{\Gamma})\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha+}}\|\boldsymbol{v}^{(b)}\|_{C^{1,\alpha}(\mathbb{T})} \\
&\leq c(\epsilon,\|\boldsymbol{X}_{ss}^{(a)}\|_{C^{1,\alpha}},\|\boldsymbol{X}_{ss}^{(b)}\|_{C^{1,\alpha}},c_{\Gamma})\|\boldsymbol{X}^{(a)}-\boldsymbol{X}^{(b)}\|_{C^{2,\alpha+}}\|\boldsymbol{f}\|_{C^{0,\alpha}(\mathbb{T})}.
\end{aligned} ∥ v ( a ) − v ( b ) ∥ C 1 , α ( T ) = L ϵ ( a ) [ ( L ϵ ( b ) ) − 1 [ v ( b ) ] − ( L ϵ ( a ) ) − 1 [ v ( b ) ] ] C 1 , α ( T ) ≤ c ( ϵ , ∥ X ss ( a ) ∥ C 1 , α , c Γ ) ( L ϵ ( b ) ) − 1 [ v ( b ) ] − ( L ϵ ( a ) ) − 1 [ v ( b ) ] C 0 , α ( T ) ≤ c ( ϵ , ∥ X ss ( a ) ∥ C 1 , α , κ ∗ , α + ( b ) , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ v ( b ) ∥ C 1 , α ( T ) ≤ c ( ϵ , ∥ X ss ( a ) ∥ C 1 , α , ∥ X ss ( b ) ∥ C 1 , α , c Γ ) ∥ X ( a ) − X ( b ) ∥ C 2 , α + ∥ f ∥ C 0 , α ( T ) .
Proof of Lemma 1.1: nearby curves
Given the two curves X ( a ) ( s ) , X ( b ) ( s ) \mathbf{X}^{(a)}(s),\mathbf{X}^{(b)}(s) X ( a ) ( s ) , X ( b ) ( s ) satisfying (9), we begin by framing both curves with a corresponding Bishop frame ( e t , e ~ 1 , e ~ 2 ) (\mathbf{e}_t,\widetilde{\mathbf{e}}_1,\widetilde{\mathbf{e}}_2) ( e t , e 1 , e 2 ) [9 ] , which, given an initial choice of normal vector e ~ 1 ( 0 ) \widetilde{\mathbf{e}}_1(0) e 1 ( 0 ) , satisfies the ODE
d d s ( e t e ~ 1 e ~ 2 ) = ( 0 κ ~ 1 κ ~ 2 − κ ~ 1 0 0 − κ ~ 2 0 0 ) ( e t e ~ 1 e ~ 2 ) . (349) \frac{d}{ds}
\begin{pmatrix}
\mathbf{e}_t\\
\widetilde{\mathbf{e}}_1\\
\widetilde{\mathbf{e}}_2
\end{pmatrix}
=
\begin{pmatrix}
0 & \widetilde{\kappa}_1 & \widetilde{\kappa}_2\\
-\widetilde{\kappa}_1 & 0 & 0\\
-\widetilde{\kappa}_2 & 0 & 0
\end{pmatrix}
\begin{pmatrix}
\mathbf{e}_t\\
\widetilde{\mathbf{e}}_1\\
\widetilde{\mathbf{e}}_2
\end{pmatrix}. \tag*{(349)} d s d e t e 1 e 2 = 0 − κ 1 − κ 2 κ 1 0 0 κ 2 0 0 e t e 1 e 2 . (349)
Here for curve (a) we choose the initial normal vector e ~ 1 ( a ) ( 0 ) \widetilde{\mathbf{e}}_1^{(a)}(0) e 1 ( a ) ( 0 ) arbitrarily, while e ~ 1 ( b ) ( 0 ) \widetilde{\mathbf{e}}_1^{(b)}(0) e 1 ( b ) ( 0 ) is chosen to satisfy
e ~ 1 ( b ) ( 0 ) = inf v ⊥ e t ( b ) ( 0 ) , ∣ v ∣ = 1 ∣ v − e ~ 1 ( a ) ( 0 ) ∣ ; \widetilde{\mathbf{e}}_1^{(b)}(0)=\inf_{\substack{\mathbf{v}\perp\mathbf{e}_t^{(b)}(0),\lvert\mathbf{v}\rvert=1}}\left\lvert\mathbf{v}-\widetilde{\mathbf{e}}_1^{(a)}(0)\right\rvert; e 1 ( b ) ( 0 ) = v ⊥ e t ( b ) ( 0 ) , ∣ v ∣ = 1 inf v − e 1 ( a ) ( 0 ) ;
i.e. at the closest pair of points X ( a ) ( 0 ) \mathbf{X}^{(a)}(0) X ( a ) ( 0 ) and X ( b ) ( 0 ) \mathbf{X}^{(b)}(0) X ( b ) ( 0 ) , we choose the normal vectors to be as close as possible. Given that X ( a ) \mathbf{X}^{(a)} X ( a ) and X ( b ) \mathbf{X}^{(b)} X ( b ) satisfy (9), the difference in initial normal vectors must then satisfy
∣ e ~ 1 ( a ) ( 0 ) − e ~ 1 ( b ) ( 0 ) ∣ ≤ δ . (350) \left\lvert\widetilde{\mathbf{e}}_1^{(a)}(0)-\widetilde{\mathbf{e}}_1^{(b)}(0)\right\rvert\leq\delta. \tag*{(350)} e 1 ( a ) ( 0 ) − e 1 ( b ) ( 0 ) ≤ δ . (350)
Although the pure Bishop frame is not necessarily periodic in s s s , its coefficients κ ~ 1 \widetilde{\kappa}_1 κ 1 and κ ~ 2 \widetilde{\kappa}_2 κ 2 control the coefficients from the periodicized frame (3). In particular, using the Bishop frame (347), we define the angle − π ≤ ϕ < π -\pi\leq\phi<\pi − π ≤ ϕ < π to be
ϕ = arccos ( e ~ 1 ( 0 ) ⋅ e ~ 1 ( 1 ) ) , \phi=\arccos\left(\widetilde{\mathbf{e}}_1(0)\cdot\widetilde{\mathbf{e}}_1(1)\right), ϕ = arccos ( e 1 ( 0 ) ⋅ e 1 ( 1 ) ) ,
i.e. the mismatch from periodicity. Take κ 3 = − ϕ \kappa_3=-\phi κ 3 = − ϕ and note that for the curves (a) and (b) we have
∣ κ 3 ( a ) − κ 3 ( b ) ∣ ≤ c ∥ e ~ 1 ( a ) − e ~ 1 ( b ) ∥ L ∞ . (351) \left\lvert\kappa_3^{(a)}-\kappa_3^{(b)}\right\rvert\leq c\left\lVert\widetilde{\mathbf{e}}_1^{(a)}-\widetilde{\mathbf{e}}_1^{(b)}\right\rVert_{L^\infty}. \tag*{(351)} κ 3 ( a ) − κ 3 ( b ) ≤ c e 1 ( a ) − e 1 ( b ) L ∞ . (351)
Furthermore, given κ 3 \kappa_3 κ 3 as above, we may define
( e n 1 ( s ) e n 2 ( s ) ) = ( cos ( s κ 3 ) − sin ( s κ 3 ) sin ( s κ 3 ) cos ( s κ 3 ) ) ( e ~ 1 ( s ) e ~ 2 ( s ) ) . \begin{pmatrix}
\mathbf{e}_{n_1}(s)\\
\mathbf{e}_{n_2}(s)
\end{pmatrix}
=
\begin{pmatrix}
\cos(s\kappa_3)&-\sin(s\kappa_3)\\
\sin(s\kappa_3)&\cos(s\kappa_3)
\end{pmatrix}
\begin{pmatrix}
\widetilde{\mathbf{e}}_1(s)\\
\widetilde{\mathbf{e}}_2(s)
\end{pmatrix}. ( e n 1 ( s ) e n 2 ( s ) ) = ( cos ( s κ 3 ) sin ( s κ 3 ) − sin ( s κ 3 ) cos ( s κ 3 ) ) ( e 1 ( s ) e 2 ( s ) ) .
It may be seen that ( e t ( s ) , e n 1 ( s ) , e n 2 ( s ) ) (\mathbf{e}_t(s),\mathbf{e}_{n_1}(s),\mathbf{e}_{n_2}(s)) ( e t ( s ) , e n 1 ( s ) , e n 2 ( s )) satisfy the ODE (3) with
( κ 1 ( s ) κ 2 ( s ) ) = ( cos ( s κ 3 ) − sin ( s κ 3 ) sin ( s κ 3 ) cos ( s κ 3 ) ) ( κ ~ 1 ( s ) κ ~ 2 ( s ) ) . \begin{pmatrix}
\kappa_1(s)\\
\kappa_2(s)
\end{pmatrix}
=
\begin{pmatrix}
\cos(s\kappa_3)&-\sin(s\kappa_3)\\
\sin(s\kappa_3)&\cos(s\kappa_3)
\end{pmatrix}
\begin{pmatrix}
\widetilde{\kappa}_1(s)\\
\widetilde{\kappa}_2(s)
\end{pmatrix}. ( κ 1 ( s ) κ 2 ( s ) ) = ( cos ( s κ 3 ) sin ( s κ 3 ) − sin ( s κ 3 ) cos ( s κ 3 ) ) ( κ 1 ( s ) κ 2 ( s ) ) .
For both ℓ = 1 , 2 \ell=1,2 ℓ = 1 , 2 , we then have
∥ κ ℓ ( a ) − κ ℓ ( b ) ∥ C 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ∣ κ 3 ( a ) − κ 3 ( b ) ∣ + c ∥ κ ~ 1 ( a ) − κ ~ 1 ( b ) ∥ C 0 , β + c ∥ κ ~ 2 ( a ) − κ ~ 2 ( b ) ∥ C 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ( ∥ κ ~ 1 ( a ) − κ ~ 1 ( b ) ∥ C 0 , β + ∥ κ ~ 2 ( a ) − κ ~ 2 ( b ) ∥ C 0 , β ) . (352) \begin{aligned}
\left\lVert\kappa_\ell^{(a)}-\kappa_\ell^{(b)}\right\rVert_{C^{0,\beta}}
\leq c(\kappa_{\ast,\beta}^{(a)},\kappa_{\ast,\beta}^{(b)})\left\lvert\kappa_3^{(a)}-\kappa_3^{(b)}\right\rvert
+c\left\lVert\widetilde{\kappa}_1^{(a)}-\widetilde{\kappa}_1^{(b)}\right\rVert_{C^{0,\beta}}
+c\left\lVert\widetilde{\kappa}_2^{(a)}-\widetilde{\kappa}_2^{(b)}\right\rVert_{C^{0,\beta}} \\
&\leq c(\kappa_{\ast,\beta}^{(a)},\kappa_{\ast,\beta}^{(b)})\left(\left\lVert\widetilde{\kappa}_1^{(a)}-\widetilde{\kappa}_1^{(b)}\right\rVert_{C^{0,\beta}}+\left\lVert\widetilde{\kappa}_2^{(a)}-\widetilde{\kappa}_2^{(b)}\right\rVert_{C^{0,\beta}}\right). \tag*{(352)}
\end{aligned} κ ℓ ( a ) − κ ℓ ( b ) C 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) κ 3 ( a ) − κ 3 ( b ) + c κ 1 ( a ) − κ 1 ( b ) C 0 , β + c κ 2 ( a ) − κ 2 ( b ) C 0 , β ≤ c ( κ ∗ , β ( a ) , κ ∗ , β ( b ) ) ( κ 1 ( a ) − κ 1 ( b ) C 0 , β + κ 2 ( a ) − κ 2 ( b ) C 0 , β ) . (352)
It thus suffices to show the desired bound (10) for the coefficients κ ~ 1 , κ ~ 2 \widetilde{\kappa}_1,\widetilde{\kappa}_2 κ 1 , κ 2 of the pure Bishop frame. For both curves (a) and (b), we may rewrite the ODEs (347) for e ~ 1 \widetilde{\mathbf{e}}_1 e 1 and e ~ 2 \widetilde{\mathbf{e}}_2 e 2 as
e ~ ˙ 1 = − ( e ~ 1 ⋅ e ˙ t ) e t \dot{\widetilde{\mathbf{e}}}_1=-(\widetilde{\mathbf{e}}_1\cdot\dot{\mathbf{e}}_t)\mathbf{e}_t e ˙ 1 = − ( e 1 ⋅ e ˙ t ) e t
e ~ ˙ 2 = − ( e ~ 2 ⋅ e ˙ t ) e t . (353) \dot{\widetilde{\mathbf{e}}}_2=-(\widetilde{\mathbf{e}}_2\cdot\dot{\mathbf{e}}_t)\mathbf{e}_t. \tag*{(353)} e ˙ 2 = − ( e 2 ⋅ e ˙ t ) e t . (353)
where we use e ˙ \dot{\mathbf{e}} e ˙ to denote d d s e \frac{d}{ds}\mathbf{e} d s d e . Letting w = e ~ 1 ( a ) − e ~ 1 ( b ) \mathbf{w}=\widetilde{\mathbf{e}}_1^{(a)}-\widetilde{\mathbf{e}}_1^{(b)} w = e 1 ( a ) − e 1 ( b ) , we then have
w ˙ = − ( w ⋅ e ˙ t ( a ) ) e t ( a ) − ( e ~ 1 ( b ) ⋅ ( e ˙ t ( a ) − e ˙ t ( b ) ) ) e t ( a ) − ( e ~ 1 ( b ) ⋅ e ˙ t ( b ) ) ( e t ( a ) − e t ( b ) ) . (354) \dot{\mathbf{w}}=-(\mathbf{w}\cdot\dot{\mathbf{e}}_t^{(a)})\mathbf{e}_t^{(a)}
-(\widetilde{\mathbf{e}}_1^{(b)}\cdot(\dot{\mathbf{e}}_t^{(a)}-\dot{\mathbf{e}}_t^{(b)}))\mathbf{e}_t^{(a)}
-(\widetilde{\mathbf{e}}_1^{(b)}\cdot\dot{\mathbf{e}}_t^{(b)})(\mathbf{e}_t^{(a)}-\mathbf{e}_t^{(b)}). \tag*{(354)} w ˙ = − ( w ⋅ e ˙ t ( a ) ) e t ( a ) − ( e 1 ( b ) ⋅ ( e ˙ t ( a ) − e ˙ t ( b ) )) e t ( a ) − ( e 1 ( b ) ⋅ e ˙ t ( b ) ) ( e t ( a ) − e t ( b ) ) . (354)
In particular, using that ∥ e ~ t ( a ) − e ~ t ( b ) ∥ C 1 , β ≤ ∥ X ( a ) − X ( b ) ∥ C 2 , β = δ \left\|\widetilde{e}_t^{(a)}-\widetilde{e}_t^{(b)}\right\|_{C^{1,\beta}}\leq\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}}=\delta e t ( a ) − e t ( b ) C 1 , β ≤ X ( a ) − X ( b ) C 2 , β = δ , we have
1 2 d d s ∣ w ∣ 2 = c ( κ ∗ ( a ) ) ∣ w ∣ 2 + δ ∣ w ∣ + c ( κ ∗ ( b ) ) δ ∣ w ∣ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∣ w ∣ 2 + δ 2 ) . \frac{1}{2}\frac{\mathrm{d}}{\mathrm{d}s}|w|^2=c(\kappa_*^{(a)})|w|^2+\delta|w|+c(\kappa_*^{(b)})\delta|w|
\leq c(\kappa_*^{(a)},\kappa_*^{(b)})(|w|^2+\delta^2). 2 1 d s d ∣ w ∣ 2 = c ( κ ∗ ( a ) ) ∣ w ∣ 2 + δ ∣ w ∣ + c ( κ ∗ ( b ) ) δ ∣ w ∣ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∣ w ∣ 2 + δ 2 ) .
Using Grönwall’s inequality and (350) , we then have
∣ w ∣ 2 ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∣ w ( 0 ) ∣ 2 + δ 2 ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) δ 2 . |w|^2\leq c(\kappa_*^{(a)},\kappa_*^{(b)})(|w(0)|^2+\delta^2)\leq c(\kappa_*^{(a)},\kappa_*^{(b)})\delta^2. ∣ w ∣ 2 ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ( ∣ w ( 0 ) ∣ 2 + δ 2 ) ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) δ 2 .
Returning to the ODE (354) for w w w , we also obtain
∣ w ˙ ∣ ≤ c ( κ ∗ ( a ) ) ∣ w ∣ + δ + c ( κ ∗ ( b ) ) δ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) δ . |\dot{w}|\leq c(\kappa_*^{(a)})|w|+\delta+c(\kappa_*^{(b)})\delta\leq c(\kappa_*^{(a)},\kappa_*^{(b)})\delta. ∣ w ˙ ∣ ≤ c ( κ ∗ ( a ) ) ∣ w ∣ + δ + c ( κ ∗ ( b ) ) δ ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) δ .
In total, recalling that δ = ∥ X ( a ) − X ( b ) ∥ C 2 , β \delta=\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}} δ = X ( a ) − X ( b ) C 2 , β and w = e ~ 1 ( a ) − e ~ 1 ( b ) w=\widetilde{e}_1^{(a)}-\widetilde{e}_1^{(b)} w = e 1 ( a ) − e 1 ( b ) , we have
∥ e ~ 1 ( a ) − e ~ 1 ( b ) ∥ C 1 ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) ∥ X ( a ) − X ( b ) ∥ C 2 , β . \left\|\widetilde{e}_1^{(a)}-\widetilde{e}_1^{(b)}\right\|_{C^1}\leq c(\kappa_*^{(a)},\kappa_*^{(b)})\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}}. e 1 ( a ) − e 1 ( b ) C 1 ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) ) X ( a ) − X ( b ) C 2 , β .
We may then use the ODEs (349) , (353) to obtain bounds on the coefficients κ ~ 1 \widetilde{\kappa}_1 κ 1 . We have
κ ~ 1 ( a ) − κ ~ 1 ( b ) = ( e ~ 1 ( a ) − e ~ 1 ( b ) ) ⋅ e ˙ t ( a ) + e ~ 1 ( b ) ⋅ ( e ˙ t ( a ) − e ˙ t ( b ) ) , \widetilde{\kappa}_1^{(a)}-\widetilde{\kappa}_1^{(b)}=(\widetilde{e}_1^{(a)}-\widetilde{e}_1^{(b)})\cdot\dot{e}_t^{(a)}+\widetilde{e}_1^{(b)}\cdot(\dot{e}_t^{(a)}-\dot{e}_t^{(b)}), κ 1 ( a ) − κ 1 ( b ) = ( e 1 ( a ) − e 1 ( b ) ) ⋅ e ˙ t ( a ) + e 1 ( b ) ⋅ ( e ˙ t ( a ) − e ˙ t ( b ) ) ,
and thus
∥ κ ~ 1 ( a ) − κ ~ 1 ( b ) ∥ C 0 , β ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , κ ∗ ( a ) , κ ∗ ( b ) ) ∥ X ( a ) − X ( b ) ∥ C 2 , β . \left\|\widetilde{\kappa}_1^{(a)}-\widetilde{\kappa}_1^{(b)}\right\|_{C^{0,\beta}}\leq c(\kappa_*^{(a)},\kappa_*^{(b)},\kappa_*^{(a)},\kappa_*^{(b)})\left\|X^{(a)}-X^{(b)}\right\|_{C^{2,\beta}}. κ 1 ( a ) − κ 1 ( b ) C 0 , β ≤ c ( κ ∗ ( a ) , κ ∗ ( b ) , κ ∗ ( a ) , κ ∗ ( b ) ) X ( a ) − X ( b ) C 2 , β .
Using the ODE (353) for e ~ 2 \widetilde{e}_2 e 2 , the same bound may be shown for κ ~ 2 ( a ) − κ ~ 2 ( b ) \widetilde{\kappa}_2^{(a)}-\widetilde{\kappa}_2^{(b)} κ 2 ( a ) − κ 2 ( b ) . Then by (351) and (352) , we obtain Lemma 1.1. □
Appendix B. Calculation of straight single and double layer symbols
Here we derive the expressions stated in Lemma 1.7 for the single and double layer operators S ‾ \overline{S} S and D ‾ \overline{D} D about the straight filament C ϵ C_\epsilon C ϵ .
Proof of Lemma 1.7. For the straight filament, periodicity of the operators S ‾ \overline{S} S and D ‾ \overline{D} D is enforced by considering only densities φ ( s , θ ) \varphi(s,\theta) φ ( s , θ ) which are 1-periodic in s s s . For such a density φ \varphi φ , we write S ‾ \overline{S} S and D ‾ \overline{D} D as
S ‾ [ φ ] ( s , θ ) = ∫ − ∞ ∞ ∫ 0 2 π G ‾ ( s , θ , s ′ , θ ′ ) φ ( s ′ , θ ′ ) ϵ d θ ′ d s ′ \overline{S}[\varphi](s,\theta)=\int_{-\infty}^{\infty}\int_0^{2\pi}\overline{G}(s,\theta,s',\theta')\varphi(s',\theta')\,\epsilon\,\mathrm{d}\theta'\mathrm{d}s' S [ φ ] ( s , θ ) = ∫ − ∞ ∞ ∫ 0 2 π G ( s , θ , s ′ , θ ′ ) φ ( s ′ , θ ′ ) ϵ d θ ′ d s ′
D ‾ [ φ ] ( s , θ ) = ∫ − ∞ ∞ ∫ 0 2 π K ‾ D ( s , θ , s ′ , θ ′ ) φ ( s ′ , θ ′ ) ϵ d θ ′ d s ′ , \overline{D}[\varphi](s,\theta)=\int_{-\infty}^{\infty}\int_0^{2\pi}\overline{K}_D(s,\theta,s',\theta')\varphi(s',\theta')\,\epsilon\,\mathrm{d}\theta'\mathrm{d}s', D [ φ ] ( s , θ ) = ∫ − ∞ ∞ ∫ 0 2 π K D ( s , θ , s ′ , θ ′ ) φ ( s ′ , θ ′ ) ϵ d θ ′ d s ′ ,
where the kernels G ‾ \overline{G} G and K ‾ D \overline{K}_D K D will be written in more detail below. Recalling the general form of the Stokeslet G G G (37) and double layer kernel K D K_D K D (38) , we note that about C ϵ C_\epsilon C ϵ we may write
x ‾ − x ‾ ′ = ( s − s ′ ) e z + 2 ϵ sin ( θ − θ ′ 2 ) e θ ( θ + θ ′ 2 ) , ( x ‾ − x ‾ ′ ) ⋅ e r ( θ ′ ) = − 2 ϵ sin 2 ( θ − θ ′ 2 ) , e θ ( θ + θ ′ 2 ) = − sin ( θ + θ ′ 2 ) e x + cos ( θ + θ ′ 2 ) e y . (355) \begin{aligned}
\overline{x}-\overline{x}'&=(s-s')e_z+2\epsilon\sin\left(\frac{\theta-\theta'}{2}\right)e_\theta\left(\frac{\theta+\theta'}{2}\right),\\
(\overline{x}-\overline{x}')\cdot e_r(\theta')&=-2\epsilon\sin^2\left(\frac{\theta-\theta'}{2}\right),\\
e_\theta\left(\frac{\theta+\theta'}{2}\right)&=-\sin\left(\frac{\theta+\theta'}{2}\right)e_x+\cos\left(\frac{\theta+\theta'}{2}\right)e_y. \tag*{(355)}
\end{aligned} x − x ′ ( x − x ′ ) ⋅ e r ( θ ′ ) e θ ( 2 θ + θ ′ ) = ( s − s ′ ) e z + 2 ϵ sin ( 2 θ − θ ′ ) e θ ( 2 θ + θ ′ ) , = − 2 ϵ sin 2 ( 2 θ − θ ′ ) , = − sin ( 2 θ + θ ′ ) e x + cos ( 2 θ + θ ′ ) e y . (355)
Here again we use overline notation to emphasize that x ‾ \overline{x} x lies along the straight filament C ϵ C_\epsilon C ϵ . Given a continuous density φ \varphi φ along C ϵ C_\epsilon C ϵ , we may use (355) to write out the integrand (37) of the single layer operator about C ϵ C_{\epsilon} C ϵ as
8 π ( G ‾ φ ) ⋅ e z = φ ⋅ e z ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) + ( s − s ′ ) 2 ( φ ⋅ e z ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 3 + 2 ϵ ( s − s ′ ) sin ( θ − θ ′ 2 ) ( − sin ( θ + θ ′ 2 ) ( φ ⋅ e x ) + cos ( θ + θ ′ 2 ) ( φ ⋅ e y ) ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 3 , 8 π ( G ‾ φ ) ⋅ e x = φ ⋅ e x ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) − 2 ϵ ( s − s ′ ) sin ( θ − θ ′ 2 ) sin ( θ + θ ′ 2 ) ( φ ⋅ e z ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 3 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) ( sin 2 ( θ + θ ′ 2 ) ( φ ⋅ e x ) − sin ( θ + θ ′ 2 ) cos ( θ + θ ′ 2 ) ( φ ⋅ e y ) ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 3 , 8 π ( G ‾ φ ) ⋅ e y = φ ⋅ e y ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) + 2 ϵ ( s − s ′ ) sin ( θ − θ ′ 2 ) cos ( θ + θ ′ 2 ) ( φ ⋅ e z ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 3 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) ( cos 2 ( θ + θ ′ 2 ) ( φ ⋅ e y ) − sin ( θ + θ ′ 2 ) cos ( θ + θ ′ 2 ) ( φ ⋅ e x ) ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 3 . (356) \begin{aligned}
8\pi(\overline{G}\varphi)\cdot e_z
&= \frac{\varphi\cdot e_z}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}}+
\frac{(s-s')^2(\varphi\cdot e_z)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^3}\\
&\quad+\frac{2\epsilon(s-s')\sin\left(\frac{\theta-\theta'}{2}\right)\left(-\sin\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_x)+\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_y)\right)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^3},\\
8\pi(\overline{G}\varphi)\cdot e_x
&= \frac{\varphi\cdot e_x}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}}-
\frac{2\epsilon(s-s')\sin\left(\frac{\theta-\theta'}{2}\right)\sin\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_z)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^3}\\
&\quad+\frac{4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)\left(\sin^2\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_x)-\sin\left(\frac{\theta+\theta'}{2}\right)\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_y)\right)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^3},\\
8\pi(\overline{G}\varphi)\cdot e_y
&= \frac{\varphi\cdot e_y}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}}+
\frac{2\epsilon(s-s')\sin\left(\frac{\theta-\theta'}{2}\right)\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_z)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^3}\\
&\quad+\frac{4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)\left(\cos^2\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_y)-\sin\left(\frac{\theta+\theta'}{2}\right)\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_x)\right)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^3}. \tag*{(356)}
\end{aligned} 8 π ( G φ ) ⋅ e z 8 π ( G φ ) ⋅ e x 8 π ( G φ ) ⋅ e y = ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) φ ⋅ e z + ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 3 ( s − s ′ ) 2 ( φ ⋅ e z ) + ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 3 2 ϵ ( s − s ′ ) sin ( 2 θ − θ ′ ) ( − sin ( 2 θ + θ ′ ) ( φ ⋅ e x ) + cos ( 2 θ + θ ′ ) ( φ ⋅ e y ) ) , = ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) φ ⋅ e x − ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 3 2 ϵ ( s − s ′ ) sin ( 2 θ − θ ′ ) sin ( 2 θ + θ ′ ) ( φ ⋅ e z ) + ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 3 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) ( sin 2 ( 2 θ + θ ′ ) ( φ ⋅ e x ) − sin ( 2 θ + θ ′ ) cos ( 2 θ + θ ′ ) ( φ ⋅ e y ) ) , = ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) φ ⋅ e y + ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 3 2 ϵ ( s − s ′ ) sin ( 2 θ − θ ′ ) cos ( 2 θ + θ ′ ) ( φ ⋅ e z ) + ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 3 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) ( cos 2 ( 2 θ + θ ′ ) ( φ ⋅ e y ) − sin ( 2 θ + θ ′ ) cos ( 2 θ + θ ′ ) ( φ ⋅ e x ) ) . (356)
Similarly, using (355) in (38) the integrand of the straight double layer operator may be written
( K ‾ D φ ) ⋅ e z = − 3 π ϵ 2 ( s − s ′ ) sin 3 ( θ − θ ′ 2 ) ( − sin ( θ + θ ′ 2 ) ( φ ⋅ e x ) + cos ( θ + θ ′ 2 ) ( φ ⋅ e y ) ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 5 − 3 4 π 2 ϵ ( s − s ′ ) sin 2 ( θ − θ ′ 2 ) ( φ ⋅ e z ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 5 , ( K ‾ D φ ) ⋅ e x = − 6 π ϵ 3 sin 4 ( θ − θ ′ 2 ) ( sin 2 ( θ + θ ′ 2 ) ( φ ⋅ e x ) − sin ( θ + θ ′ 2 ) cos ( θ + θ ′ 2 ) ( φ ⋅ e y ) ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 5 + 3 π ϵ 2 ( s − s ′ ) sin 3 ( θ − θ ′ 2 ) sin ( θ + θ ′ 2 ) ( φ ⋅ e z ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 5 , ( K ‾ D φ ) ⋅ e y = − 6 π ϵ 3 sin 4 ( θ − θ ′ 2 ) ( cos 2 ( θ + θ ′ 2 ) ( φ ⋅ e y ) − sin ( θ + θ ′ 2 ) cos ( θ + θ ′ 2 ) ( φ ⋅ e x ) ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 5 − 3 π ϵ 2 ( s − s ′ ) sin 3 ( θ − θ ′ 2 ) cos ( θ + θ ′ 2 ) ( φ ⋅ e z ) ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( θ − θ ′ 2 ) 5 . (357) \begin{aligned}
(\overline{K}_D\varphi)\cdot e_z
&= -\frac{3}{\pi}\frac{\epsilon^2(s-s')\sin^3\left(\frac{\theta-\theta'}{2}\right)\left(-\sin\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_x)+\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_y)\right)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^5}\\
&\quad-\frac{3}{4\pi}\frac{2\epsilon(s-s')\sin^2\left(\frac{\theta-\theta'}{2}\right)(\varphi\cdot e_z)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^5},\\
(\overline{K}_D\varphi)\cdot e_x
&= -\frac{6}{\pi}\frac{\epsilon^3\sin^4\left(\frac{\theta-\theta'}{2}\right)\left(\sin^2\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_x)-\sin\left(\frac{\theta+\theta'}{2}\right)\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_y)\right)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^5}\\
&\quad+\frac{3}{\pi}\frac{\epsilon^2(s-s')\sin^3\left(\frac{\theta-\theta'}{2}\right)\sin\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_z)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^5},\\
(\overline{K}_D\varphi)\cdot e_y
&= -\frac{6}{\pi}\frac{\epsilon^3\sin^4\left(\frac{\theta-\theta'}{2}\right)\left(\cos^2\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_y)-\sin\left(\frac{\theta+\theta'}{2}\right)\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_x)\right)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^5}\\
&\quad-\frac{3}{\pi}\frac{\epsilon^2(s-s')\sin^3\left(\frac{\theta-\theta'}{2}\right)\cos\left(\frac{\theta+\theta'}{2}\right)(\varphi\cdot e_z)}{\sqrt{(s-s')^2+4\epsilon^2\sin^2\left(\frac{\theta-\theta'}{2}\right)}^5}. \tag*{(357)}
\end{aligned} ( K D φ ) ⋅ e z ( K D φ ) ⋅ e x ( K D φ ) ⋅ e y = − π 3 ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 5 ϵ 2 ( s − s ′ ) sin 3 ( 2 θ − θ ′ ) ( − sin ( 2 θ + θ ′ ) ( φ ⋅ e x ) + cos ( 2 θ + θ ′ ) ( φ ⋅ e y ) ) − 4 π 3 ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 5 2 ϵ ( s − s ′ ) sin 2 ( 2 θ − θ ′ ) ( φ ⋅ e z ) , = − π 6 ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 5 ϵ 3 sin 4 ( 2 θ − θ ′ ) ( sin 2 ( 2 θ + θ ′ ) ( φ ⋅ e x ) − sin ( 2 θ + θ ′ ) cos ( 2 θ + θ ′ ) ( φ ⋅ e y ) ) + π 3 ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 5 ϵ 2 ( s − s ′ ) sin 3 ( 2 θ − θ ′ ) sin ( 2 θ + θ ′ ) ( φ ⋅ e z ) , = − π 6 ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 5 ϵ 3 sin 4 ( 2 θ − θ ′ ) ( cos 2 ( 2 θ + θ ′ ) ( φ ⋅ e y ) − sin ( 2 θ + θ ′ ) cos ( 2 θ + θ ′ ) ( φ ⋅ e x ) ) − π 3 ( s − s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ − θ ′ ) 5 ϵ 2 ( s − s ′ ) sin 3 ( 2 θ − θ ′ ) cos ( 2 θ + θ ′ ) ( φ ⋅ e z ) . (357)
Before computing anything involving (356) and (357) , we make note of a series of identities involving integrals of the second-kind modified Bessel functions K 0 K_0 K 0 and K 1 K_1 K 1 . In particular, the following identities hold:
∫ 0 2 π K 0 ( z sin ( θ 2 ) ) d θ = 2 π I 0 ( z 2 ) K 0 ( z 2 ) \int_{0}^{2\pi} K_0\left(z\sin\left(\frac{\theta}{2}\right)\right)\,d\theta
=2\pi I_0\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right) ∫ 0 2 π K 0 ( z sin ( 2 θ ) ) d θ = 2 π I 0 ( 2 z ) K 0 ( 2 z )
∫ 0 2 π sin ( θ 2 ) K 1 ( z sin ( θ 2 ) ) d θ = π ( I 0 ( z 2 ) K 1 ( z 2 ) − I 1 ( z 2 ) K 0 ( z 2 ) ) \int_{0}^{2\pi} \sin\left(\frac{\theta}{2}\right)K_1\left(z\sin\left(\frac{\theta}{2}\right)\right)\,d\theta
=\pi\left(I_0\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)-I_1\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)\right) ∫ 0 2 π sin ( 2 θ ) K 1 ( z sin ( 2 θ ) ) d θ = π ( I 0 ( 2 z ) K 1 ( 2 z ) − I 1 ( 2 z ) K 0 ( 2 z ) )
∫ 0 2 π sin 2 ( θ 2 ) K 0 ( z sin ( θ 2 ) ) d θ = π ( I 0 ( z 2 ) K 0 ( z 2 ) − I 1 ( z 2 ) K 1 ( z 2 ) ) \int_{0}^{2\pi} \sin^2\left(\frac{\theta}{2}\right)K_0\left(z\sin\left(\frac{\theta}{2}\right)\right)\,d\theta
=\pi\left(I_0\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)-I_1\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)\right) ∫ 0 2 π sin 2 ( 2 θ ) K 0 ( z sin ( 2 θ ) ) d θ = π ( I 0 ( 2 z ) K 0 ( 2 z ) − I 1 ( 2 z ) K 1 ( 2 z ) )
∫ 0 2 π sin 3 ( θ 2 ) K 1 ( z sin ( θ 2 ) ) d θ = π ( I 0 ( z 2 ) K 1 ( z 2 ) − I 1 ( z 2 ) K 0 ( z 2 ) ) − 2 π z I 1 ( z 2 ) K 1 ( z 2 ) \int_{0}^{2\pi} \sin^3\left(\frac{\theta}{2}\right)K_1\left(z\sin\left(\frac{\theta}{2}\right)\right)\,d\theta
=\pi\left(I_0\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)-I_1\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)\right)-\frac{2\pi}{z}I_1\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right) ∫ 0 2 π sin 3 ( 2 θ ) K 1 ( z sin ( 2 θ ) ) d θ = π ( I 0 ( 2 z ) K 1 ( 2 z ) − I 1 ( 2 z ) K 0 ( 2 z ) ) − z 2 π I 1 ( 2 z ) K 1 ( 2 z )
∫ 0 2 π sin 4 ( θ 2 ) K 0 ( z sin ( θ 2 ) ) d θ = π ( I 0 ( z 2 ) K 0 ( z 2 ) − I 1 ( z 2 ) K 1 ( z 2 ) ) + π z ( I 0 ( z 2 ) K 1 ( z 2 ) − I 1 ( z 2 ) K 0 ( z 2 ) ) − 4 π z 2 I 1 ( z 2 ) K 1 ( z 2 ) \begin{aligned}
\int_{0}^{2\pi} \sin^4\left(\frac{\theta}{2}\right)K_0\left(z\sin\left(\frac{\theta}{2}\right)\right)\,d\theta
={}&\pi\left(I_0\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)-I_1\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)\right)\\
&+\frac{\pi}{z}\left(I_0\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)-I_1\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)\right)\\
&-\frac{4\pi}{z^2}I_1\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)
\end{aligned} ∫ 0 2 π sin 4 ( 2 θ ) K 0 ( z sin ( 2 θ ) ) d θ = π ( I 0 ( 2 z ) K 0 ( 2 z ) − I 1 ( 2 z ) K 1 ( 2 z ) ) + z π ( I 0 ( 2 z ) K 1 ( 2 z ) − I 1 ( 2 z ) K 0 ( 2 z ) ) − z 2 4 π I 1 ( 2 z ) K 1 ( 2 z )
∫ 0 2 π sin 5 ( θ 2 ) K 1 ( z sin ( θ 2 ) ) d θ = π ( I 0 ( z 2 ) K 1 ( z 2 ) − I 1 ( z 2 ) K 0 ( z 2 ) ) + π z ( I 0 ( z 2 ) K 0 ( z 2 ) − 3 I 1 ( z 2 ) K 1 ( z 2 ) ) + 4 π z 2 ( I 0 ( z 2 ) K 1 ( z 2 ) − I 1 ( z 2 ) K 0 ( z 2 ) ) − 16 π z 3 I 1 ( z 2 ) K 1 ( z 2 ) . (358) \begin{aligned}
\int_{0}^{2\pi} \sin^5\left(\frac{\theta}{2}\right)K_1\left(z\sin\left(\frac{\theta}{2}\right)\right)\,d\theta
={}&\pi\left(I_0\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)-I_1\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)\right)\\
&+\frac{\pi}{z}\left(I_0\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)-3I_1\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)\right)\\
&+\frac{4\pi}{z^2}\left(I_0\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right)-I_1\left(\frac{z}{2}\right)K_0\left(\frac{z}{2}\right)\right)
-\frac{16\pi}{z^3}I_1\left(\frac{z}{2}\right)K_1\left(\frac{z}{2}\right). \tag*{(358)}
\end{aligned} ∫ 0 2 π sin 5 ( 2 θ ) K 1 ( z sin ( 2 θ ) ) d θ = π ( I 0 ( 2 z ) K 1 ( 2 z ) − I 1 ( 2 z ) K 0 ( 2 z ) ) + z π ( I 0 ( 2 z ) K 0 ( 2 z ) − 3 I 1 ( 2 z ) K 1 ( 2 z ) ) + z 2 4 π ( I 0 ( 2 z ) K 1 ( 2 z ) − I 1 ( 2 z ) K 0 ( 2 z ) ) − z 3 16 π I 1 ( 2 z ) K 1 ( 2 z ) . (358)
Here I 0 I_0 I 0 and I 1 I_1 I 1 are first-kind modified Bessel functions. The first integral identity is shown in [59 ] ; the rest may be seen by differentiating the first identity with respect to z z z , using that K 0 ′ ( z ) = − K 1 ( z ) K'_0(z)=-K_1(z) K 0 ′ ( z ) = − K 1 ( z ) and K 1 ′ ( z ) = − K 0 ( z ) − K 1 ( z ) z K'_1(z)=-K_0(z)-\frac{K_1(z)}{z} K 1 ′ ( z ) = − K 0 ( z ) − z K 1 ( z ) .
Equipped with the expressions (356) and (357) and the identities (358) , we may show Lemma 1.7. We begin with the expression (47) , i.e. how the double layer acts on θ \theta θ -independent functions v ( s ) = e 2 π i k s e z v(s)=e^{2\pi iks}e_z v ( s ) = e 2 π ik s e z tangent to the filament centerline. Using (357) and (358) , we may calculate
D ‾ [ e 2 π i k s e z ] ⋅ e z = − 3 2 π ∫ − ∞ ∞ ∫ 0 2 π ϵ sin 2 ( θ ′ 2 ) ( s ′ ) 2 ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) 5 / 2 e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = − ∫ 0 2 π ( ϵ ∣ k ∣ sin ( θ ′ 2 ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) − 4 π ϵ 2 ∣ k ∣ 2 sin 2 ( θ ′ 2 ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) ) d θ ′ e 2 π i k s = − ( − 4 π 2 ϵ 2 ∣ k ∣ 2 ( I 0 K 0 − I 1 K 1 ) + π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) ) e 2 π i k s , \begin{aligned}
\overline{D}[e^{2\pi iks}e_z]\cdot e_z
={}&-\frac{3}{2\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}
\frac{\epsilon\sin^2\left(\frac{\theta'}{2}\right)(s')^2}
{\left((s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)\right)^{5/2}}
e^{2\pi ik(s-s')}\epsilon\,d\theta'\,ds'\\
={}&-\int_0^{2\pi}\left(\epsilon|k|\sin\left(\frac{\theta'}{2}\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)\right.\\
&\left.\qquad-4\pi\epsilon^2|k|^2\sin^2\left(\frac{\theta'}{2}\right)K_0\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)\right)d\theta'\,e^{2\pi iks}\\
={}&-\left(-4\pi^2\epsilon^2|k|^2(I_0K_0-I_1K_1)+\pi\epsilon|k|(I_0K_1-I_1K_0)\right)e^{2\pi iks},
\end{aligned} D [ e 2 π ik s e z ] ⋅ e z = = = − 2 π 3 ∫ − ∞ ∞ ∫ 0 2 π ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) ) 5/2 ϵ sin 2 ( 2 θ ′ ) ( s ′ ) 2 e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ − ∫ 0 2 π ( ϵ ∣ k ∣ sin ( 2 θ ′ ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) − 4 π ϵ 2 ∣ k ∣ 2 sin 2 ( 2 θ ′ ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) ) d θ ′ e 2 π ik s − ( − 4 π 2 ϵ 2 ∣ k ∣ 2 ( I 0 K 0 − I 1 K 1 ) + π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) ) e 2 π ik s ,
D ‾ [ e 2 π i k s e z ] ⋅ e r = − 3 π ∫ − ∞ ∞ ∫ 0 2 π ϵ 2 s ′ sin 4 ( θ ′ 2 ) ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) 5 / 2 e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = ∫ 0 2 π 4 π i ϵ 2 k 2 sin 3 ( θ ′ 2 ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s = − i ( 4 π 2 ϵ 2 k 2 ( I 1 K 0 − I 0 K 1 ) + 2 π ϵ ∣ k ∣ I 1 K 1 ) e 2 π i k s . \begin{aligned}
\overline{D}[e^{2\pi iks}e_z]\cdot e_r
={}&-\frac{3}{\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}
\frac{\epsilon^2s'\sin^4\left(\frac{\theta'}{2}\right)}
{\left((s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)\right)^{5/2}}
e^{2\pi ik(s-s')}\epsilon\,d\theta'\,ds'\\
={}&\int_0^{2\pi}4\pi i\epsilon^2k^2\sin^3\left(\frac{\theta'}{2}\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'\,e^{2\pi iks}\\
={}&-i\left(4\pi^2\epsilon^2k^2(I_1K_0-I_0K_1)+2\pi\epsilon|k|I_1K_1\right)e^{2\pi iks}.
\end{aligned} D [ e 2 π ik s e z ] ⋅ e r = = = − π 3 ∫ − ∞ ∞ ∫ 0 2 π ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) ) 5/2 ϵ 2 s ′ sin 4 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ ∫ 0 2 π 4 π i ϵ 2 k 2 sin 3 ( 2 θ ′ ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s − i ( 4 π 2 ϵ 2 k 2 ( I 1 K 0 − I 0 K 1 ) + 2 π ϵ ∣ k ∣ I 1 K 1 ) e 2 π ik s .
Here each I j I_j I j , K j K_j K j in the final expressions is evaluated at 2 π ϵ ∣ k ∣ 2\pi\epsilon|k| 2 π ϵ ∣ k ∣ . We also note that D ‾ [ e 2 π i k s e z ] ⋅ e θ = 0 \overline{D}[e^{2\pi iks}e_z]\cdot e_\theta=0 D [ e 2 π ik s e z ] ⋅ e θ = 0 .
To calculate the expression (49) for how S ‾ − 1 \overline{S}^{-1} S − 1 acts on zero modes in θ \theta θ , we first determine how S ‾ \overline{S} S acts on functions of the form e 2 π i k s e z e^{2\pi iks}e_z e 2 π ik s e z and e 2 π i k s e r e^{2\pi iks}e_r e 2 π ik s e r . Using (356) and (358) , for the e z e_z e z direction we may calculate
S ‾ [ e 2 π i k s e z ] ⋅ e r = 1 8 π ∫ − ∞ ∞ ∫ 0 2 π 2 ϵ s ′ sin 2 ( θ ′ 2 ) ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) 3 / 2 e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = − ∫ 0 2 π i ϵ 2 ∣ k ∣ sin 2 ( θ ′ 2 ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s = − i π ϵ 2 ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) e 2 π i k s , \begin{aligned}
\overline{S}[e^{2\pi iks}e_z]\cdot e_r
=\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}
\frac{2\epsilon s'\sin^2\left(\frac{\theta'}{2}\right)}
{\left((s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)\right)^{3/2}}
e^{2\pi ik(s-s')}\epsilon\,d\theta'\,ds' \\
&=-\int_0^{2\pi}i\epsilon^2|k|\sin^2\left(\frac{\theta'}{2}\right)K_0\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'\,e^{2\pi iks} \\
&= -i\pi\epsilon^2 |k| (I_0K_0-I_1K_1)e^{2\pi i k s},
\end{aligned} S [ e 2 π ik s e z ] ⋅ e r = 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) ) 3/2 2 ϵ s ′ sin 2 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ = − ∫ 0 2 π i ϵ 2 ∣ k ∣ sin 2 ( 2 θ ′ ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s = − iπ ϵ 2 ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) e 2 π ik s ,
S ‾ [ e 2 π i k s e z ] ⋅ e z = 1 8 π ∫ − ∞ ∞ ∫ 0 2 π ( 2 ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) − 4 ϵ 2 sin 2 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) 3 ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = ( 1 2 π ∫ 0 2 π K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) ϵ d θ ′ − ∫ 0 2 π ϵ 2 ∣ k ∣ sin ( θ ′ 2 ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ ) e 2 π i k s = ϵ ( I 0 K 0 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) ) e 2 π i k s . \begin{aligned}
\overline{S}[e^{2\pi i k s}e_z]\cdot e_z
= \frac{1}{8\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\left(\frac{2}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}}-\frac{4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)^3}}\right)e^{2\pi i k(s-s')}\epsilon\,d\theta'\,ds' \\
&=\left(\frac{1}{2\pi}\int_0^{2\pi}K_0\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)\epsilon\,d\theta'-\int_0^{2\pi}\epsilon^2|k|\sin\left(\frac{\theta'}{2}\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'\right)e^{2\pi i k s} \\
&=\epsilon\left(I_0K_0+\pi\epsilon|k|(I_1K_0-I_0K_1)\right)e^{2\pi i k s}.
\end{aligned} S [ e 2 π ik s e z ] ⋅ e z = 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 2 − ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 3 4 ϵ 2 sin 2 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ = ( 2 π 1 ∫ 0 2 π K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) ϵ d θ ′ − ∫ 0 2 π ϵ 2 ∣ k ∣ sin ( 2 θ ′ ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ ) e 2 π ik s = ϵ ( I 0 K 0 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) ) e 2 π ik s .
For the e r e_r e r direction, we may calculate
S ‾ [ e 2 π i k s e r ] ⋅ e r = 1 8 π ∫ − ∞ ∞ ∫ 0 2 π ( 1 − 2 sin 2 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) − 4 ϵ 2 sin 4 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) 3 ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = 1 4 π ∫ 0 2 π K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) ( 1 − 2 sin 2 ( θ ′ 2 ) ) ϵ d θ ′ e 2 π i k s \begin{aligned}
\overline{S}[e^{2\pi i k s}e_r]\cdot e_r
=\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\left(\frac{1-2\sin^2\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}}-\frac{4\epsilon^2\sin^4\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)^3}}\right)e^{2\pi i k(s-s')}\epsilon\,d\theta'\,ds' \\
&=\frac{1}{4\pi}\int_0^{2\pi}K_0\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)\left(1-2\sin^2\left(\frac{\theta'}{2}\right)\right)\epsilon\,d\theta'\,e^{2\pi i k s}
\end{aligned} S [ e 2 π ik s e r ] ⋅ e r = 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 1 − 2 sin 2 ( 2 θ ′ ) − ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 3 4 ϵ 2 sin 4 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ = 4 π 1 ∫ 0 2 π K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) ( 1 − 2 sin 2 ( 2 θ ′ ) ) ϵ d θ ′ e 2 π ik s
− ∫ 0 2 π ϵ 2 ∣ k ∣ sin 3 ( θ ′ 2 ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s = ϵ ( I 1 K 1 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) ) e 2 π i k s , \begin{aligned}
-\int_0^{2\pi}\epsilon^2|k|\sin^3\left(\frac{\theta'}{2}\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'\,e^{2\pi i k s} \\
&=\epsilon\left(I_1K_1+\pi\epsilon|k|(I_1K_0-I_0K_1)\right)e^{2\pi i k s},
\end{aligned} − ∫ 0 2 π ϵ 2 ∣ k ∣ sin 3 ( 2 θ ′ ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s = ϵ ( I 1 K 1 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) ) e 2 π ik s ,
S ‾ [ e 2 π i k s e r ] ⋅ e z = − 1 8 π ∫ − ∞ ∞ ∫ 0 2 π 2 ϵ s ′ sin 2 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) 3 e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = ∫ 0 2 π i ϵ 2 ∣ k ∣ sin 2 ( θ ′ 2 ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s = i π ϵ 2 ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) e 2 π i k s . \begin{aligned}
\overline{S}[e^{2\pi i k s}e_r]\cdot e_z
=-\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\frac{2\epsilon s'\sin^2\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)^3}}e^{2\pi i k(s-s')}\epsilon\,d\theta'\,ds' \\
&=\int_0^{2\pi}i\epsilon^2|k|\sin^2\left(\frac{\theta'}{2}\right)K_0\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'\,e^{2\pi i k s} \\
&=i\pi\epsilon^2|k|(I_0K_0-I_1K_1)e^{2\pi i k s}.
\end{aligned} S [ e 2 π ik s e r ] ⋅ e z = − 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 3 2 ϵ s ′ sin 2 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ = ∫ 0 2 π i ϵ 2 ∣ k ∣ sin 2 ( 2 θ ′ ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s = iπ ϵ 2 ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) e 2 π ik s .
Each I j I_j I j and K j K_j K j in the final expressions is again evaluated at 2 π ϵ ∣ k ∣ 2\pi\epsilon|k| 2 π ϵ ∣ k ∣ . We further note that S ‾ [ e 2 π i k s e z ] ⋅ e θ = S ‾ [ e 2 π i k s e r ] ⋅ e θ = 0 \overline{S}[e^{2\pi i k s}e_z]\cdot e_\theta=\overline{S}[e^{2\pi i k s}e_r]\cdot e_\theta=0 S [ e 2 π ik s e z ] ⋅ e θ = S [ e 2 π ik s e r ] ⋅ e θ = 0 .
In total, we obtain the following expression for the action of S ‾ \overline{S} S on the relevant zero modes in θ \theta θ :
S ‾ ( ( e 2 π i k s e z e 2 π i k s e r ) ) = ϵ ( I 0 K 0 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) i π ϵ ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) − i π ϵ ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) I 1 K 1 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) ) ( ( e 2 π i k s e z e 2 π i k s e r ) ) . (359) \overline{S}\left(\begin{pmatrix}e^{2\pi i k s}e_z\\e^{2\pi i k s}e_r\end{pmatrix}\right)
=\epsilon\begin{pmatrix}
I_0K_0+\pi\epsilon|k|(I_1K_0-I_0K_1)&i\pi\epsilon|k|(I_0K_0-I_1K_1)\\
-i\pi\epsilon|k|(I_0K_0-I_1K_1)&I_1K_1+\pi\epsilon|k|(I_1K_0-I_0K_1)
\end{pmatrix}
\left(\begin{pmatrix}e^{2\pi i k s}e_z\\e^{2\pi i k s}e_r\end{pmatrix}\right). \tag*{(359)} S ( ( e 2 π ik s e z e 2 π ik s e r ) ) = ϵ ( I 0 K 0 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) − iπ ϵ ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) iπ ϵ ∣ k ∣ ( I 0 K 0 − I 1 K 1 ) I 1 K 1 + π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) ) ( ( e 2 π ik s e z e 2 π ik s e r ) ) . (359)
Here I j = I j ( 2 π ϵ ∣ k ∣ ) I_j=I_j(2\pi\epsilon|k|) I j = I j ( 2 π ϵ ∣ k ∣ ) and K j = K j ( 2 π ϵ ∣ k ∣ ) K_j=K_j(2\pi\epsilon|k|) K j = K j ( 2 π ϵ ∣ k ∣ ) . Inverting the matrix (359) , we obtain the expression (49) and (50) for the multiplier M S , t − 1 ( k ) M^{-1}_{S,t}(k) M S , t − 1 ( k ) in the tangential direction along the filament.
We next consider θ \theta θ -independent data in directions normal to the filament centerline. We consider data of the form e 2 π i k s e x e^{2\pi i k s}e_x e 2 π ik s e x and note that, by symmetry, data of the form e 2 π i k s e y e^{2\pi i k s}e_y e 2 π ik s e y behaves analogously. Note that, using the identities e x = cos θ e r − sin θ e θ e_x=\cos\theta e_r-\sin\theta e_\theta e x = cos θ e r − sin θ e θ and e y = sin θ e r + cos θ e θ e_y=\sin\theta e_r+\cos\theta e_\theta e y = sin θ e r + cos θ e θ , functions of this form actually correspond to one-modes in θ \theta θ .
We begin by calculating the expressions (51) for the double layer multipliers. Using the form (355) of the integrand of D ‾ \overline{D} D as well as the identities (356), we may calculate
D ‾ [ e 2 π i k s e x ] ⋅ e z = − 3 4 π ∫ − ∞ ∞ ∫ 0 2 π 4 ϵ 2 s ′ sin 4 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) 5 e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = ∫ 0 2 π i 4 π ϵ 2 k 2 sin 3 ( θ ′ 2 ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s cos θ = i ( 4 π 2 ϵ 2 k 2 ( I 0 K 1 − I 1 K 0 ) − 2 π ϵ ∣ k ∣ I 1 K 1 ) e 2 π i k s cos θ , \begin{aligned}
\overline{D}[e^{2\pi i k s}e_x]\cdot e_z
=-\frac{3}{4\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\frac{4\epsilon^2s'\sin^4\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)^5}}e^{2\pi i k(s-s')}\epsilon\,d\theta'\,ds'\cos\theta\\
&=\int_0^{2\pi}i4\pi\epsilon^2k^2\sin^3\left(\frac{\theta'}{2}\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'\,e^{2\pi i k s}\cos\theta\\
&=i\left(4\pi^2\epsilon^2k^2(I_0K_1-I_1K_0)-2\pi\epsilon|k|I_1K_1\right)e^{2\pi i k s}\cos\theta,
\end{aligned} D [ e 2 π ik s e x ] ⋅ e z = − 4 π 3 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 5 4 ϵ 2 s ′ sin 4 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = ∫ 0 2 π i 4 π ϵ 2 k 2 sin 3 ( 2 θ ′ ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s cos θ = i ( 4 π 2 ϵ 2 k 2 ( I 0 K 1 − I 1 K 0 ) − 2 π ϵ ∣ k ∣ I 1 K 1 ) e 2 π ik s cos θ ,
D ‾ [ e 2 π i k s e x ] ⋅ e x = − 3 4 π ∫ − ∞ ∞ ∫ 0 2 π 8 ϵ 3 sin 4 ( θ ′ 2 ) ( sin 2 θ − sin 2 ( θ ′ 2 ) ) ( sin 2 θ − cos 2 θ ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = − ∫ 0 2 π 4 π ϵ 2 k 2 sin 2 ( θ ′ 2 ) K 2 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s sin 2 θ + ∫ 0 2 π 4 π ϵ 2 k 2 sin 4 ( θ ′ 2 ) K 2 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s ( sin 2 θ − cos 2 θ ) = − ( π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1 ) e 2 π i k s sin 2 θ + ( 4 π 2 ϵ 2 ∣ k ∣ 2 ( I 1 K 1 − I 0 K 0 ) + 3 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1 ) e 2 π i k s cos 2 θ . \begin{aligned}
\overline{D}[e^{2\pi i k s}e_x]\cdot e_x=-\frac{3}{4\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\frac{8\epsilon^3\sin^4\left(\frac{\theta'}{2}\right)\left(\sin^2\theta-\sin^2\left(\frac{\theta'}{2}\right)\right)\left(\sin^2\theta-\cos^2\theta\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}}e^{2\pi i k(s-s')}\epsilon\,d\theta' ds' \\
&=-\int_0^{2\pi}4\pi\epsilon^2 k^2\sin^2\left(\frac{\theta'}{2}\right)K_2\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta' e^{2\pi i k s}\sin^2\theta
+\int_0^{2\pi}4\pi\epsilon^2 k^2\sin^4\left(\frac{\theta'}{2}\right)K_2\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta' e^{2\pi i k s}\left(\sin^2\theta-\cos^2\theta\right) \\
&=-\left(\pi\epsilon|k|\left(I_1K_0-I_0K_1\right)+2I_1K_1\right)e^{2\pi i k s}\sin^2\theta
+\left(4\pi^2\epsilon^2|k|^2\left(I_1K_1-I_0K_0\right)+3\pi\epsilon|k|\left(I_1K_0-I_0K_1\right)+2I_1K_1\right)e^{2\pi i k s}\cos^2\theta.
\end{aligned} D [ e 2 π ik s e x ] ⋅ e x = − 4 π 3 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 8 ϵ 3 sin 4 ( 2 θ ′ ) ( sin 2 θ − sin 2 ( 2 θ ′ ) ) ( sin 2 θ − cos 2 θ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ = − ∫ 0 2 π 4 π ϵ 2 k 2 sin 2 ( 2 θ ′ ) K 2 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s sin 2 θ + ∫ 0 2 π 4 π ϵ 2 k 2 sin 4 ( 2 θ ′ ) K 2 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s ( sin 2 θ − cos 2 θ ) = − ( π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1 ) e 2 π ik s sin 2 θ + ( 4 π 2 ϵ 2 ∣ k ∣ 2 ( I 1 K 1 − I 0 K 0 ) + 3 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1 ) e 2 π ik s cos 2 θ .
Here we have also used the Bessel function identity K 2 ( z ) = K 0 ( z ) + 2 K 1 ( z ) z K_2(z)=K_0(z)+2\frac{K_1(z)}{z} K 2 ( z ) = K 0 ( z ) + 2 z K 1 ( z ) . As before, in the final expressions, each I j I_j I j and K j K_j K j is evaluated at 2 π ϵ ∣ k ∣ 2\pi\epsilon|k| 2 π ϵ ∣ k ∣ . Altogether, the double layer operator D ‾ \overline{D} D can be seen to act on θ \theta θ -independent functions in directions normal to the filament centerline via
D ‾ [ e 2 π i k s e x ] = D ‾ [ e 2 π i k s ( cos θ e r − sin θ e θ ) ] = ( Q N cos θ e r − Q O sin θ e θ + i Q P cos θ e z ) e 2 π i k s , \overline{D}[e^{2\pi i k s}e_x]=\overline{D}[e^{2\pi i k s}\left(\cos\theta e_r-\sin\theta e_\theta\right)]=\left(Q_N\cos\theta e_r-Q_O\sin\theta e_\theta+iQ_P\cos\theta e_z\right)e^{2\pi i k s}, D [ e 2 π ik s e x ] = D [ e 2 π ik s ( cos θ e r − sin θ e θ ) ] = ( Q N cos θ e r − Q O sin θ e θ + i Q P cos θ e z ) e 2 π ik s ,
Q N = 4 π 2 ϵ 2 ∣ k ∣ 2 ( I 1 K 1 − I 0 K 0 ) + 3 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1 Q_N=4\pi^2\epsilon^2|k|^2\left(I_1K_1-I_0K_0\right)+3\pi\epsilon|k|\left(I_1K_0-I_0K_1\right)+2I_1K_1 Q N = 4 π 2 ϵ 2 ∣ k ∣ 2 ( I 1 K 1 − I 0 K 0 ) + 3 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 2 I 1 K 1
Q O = − π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) − 2 I 1 K 1 Q_O=-\pi\epsilon|k|\left(I_1K_0-I_0K_1\right)-2I_1K_1 Q O = − π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) − 2 I 1 K 1
Q P = 4 π 2 ϵ 2 k 2 ( I 0 K 1 − I 1 K 0 ) − 2 π ϵ ∣ k ∣ I 1 K 1 . Q_P=4\pi^2\epsilon^2k^2\left(I_0K_1-I_1K_0\right)-2\pi\epsilon|k|I_1K_1. Q P = 4 π 2 ϵ 2 k 2 ( I 0 K 1 − I 1 K 0 ) − 2 π ϵ ∣ k ∣ I 1 K 1 .
We next turn to the single layer operator. Given the form of the expressions arising in the double layer operator, we will need to calculate a full matrix-valued multiplier for S ‾ \overline{S} S in order to determine how S ‾ − 1 \overline{S}^{-1} S − 1 acts on functions of the form (358). Recalling the series of identities (356) as well as the form (354) of the integrand of S ‾ \overline{S} S , we begin by calculating
S ‾ [ e 2 π i k s e x ] ⋅ e x = S ‾ [ e 2 π i k s cos θ e r ] ⋅ e x − S ‾ [ e 2 π i k s sin θ e θ ] ⋅ e x = 1 8 π ∫ − ∞ ∞ ∫ 0 2 π ( 1 ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) + 4 ϵ 2 sin 2 ( θ ′ 2 ) ( sin 2 θ − sin 2 ( θ ′ 2 ) ) ( sin 2 θ − cos 2 θ ) ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) 3 ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ = 1 4 π ∫ 0 2 π K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) ϵ d θ ′ d s ′ e 2 π i k s + ∫ 0 2 π ϵ 2 ∣ k ∣ sin ( θ ′ 2 ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s sin 2 θ + ∫ 0 2 π ϵ 2 ∣ k ∣ sin 3 ( θ ′ 2 ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s ( cos 2 θ − sin 2 θ ) = ϵ 2 ( I 0 K 0 + I 1 K 1 ) e 2 π i k s sin 2 θ + ϵ ( π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) + 1 2 ( I 0 K 0 − I 1 K 1 ) ) e 2 π i k s cos 2 θ . \begin{aligned}
\overline{S}[e^{2\pi i k s}e_x]\cdot e_x=\overline{S}[e^{2\pi i k s}\cos\theta e_r]\cdot e_x-\overline{S}[e^{2\pi i k s}\sin\theta e_\theta]\cdot e_x \\
&=\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\left(\frac{1}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}}+\frac{4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)\left(\sin^2\theta-\sin^2\left(\frac{\theta'}{2}\right)\right)\left(\sin^2\theta-\cos^2\theta\right)}{\left(\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}\right)^3}\right)e^{2\pi i k(s-s')}\epsilon\,d\theta' ds' \\
&=\frac{1}{4\pi}\int_0^{2\pi}K_0\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)\epsilon\,d\theta' ds'e^{2\pi i k s}
+\int_0^{2\pi}\epsilon^2|k|\sin\left(\frac{\theta'}{2}\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'e^{2\pi i k s}\sin^2\theta\\
&+\int_0^{2\pi}\epsilon^2|k|\sin^3\left(\frac{\theta'}{2}\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)d\theta'e^{2\pi i k s}\left(\cos^2\theta-\sin^2\theta\right) \\
&=\frac{\epsilon}{2}\left(I_0K_0+I_1K_1\right)e^{2\pi i k s}\sin^2\theta+\epsilon\left(\pi\epsilon|k|\left(I_0K_1-I_1K_0\right)+\frac{1}{2}\left(I_0K_0-I_1K_1\right)\right)e^{2\pi i k s}\cos^2\theta.
\end{aligned} S [ e 2 π ik s e x ] ⋅ e x = S [ e 2 π ik s cos θ e r ] ⋅ e x − S [ e 2 π ik s sin θ e θ ] ⋅ e x = 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 1 + ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) ) 3 4 ϵ 2 sin 2 ( 2 θ ′ ) ( sin 2 θ − sin 2 ( 2 θ ′ ) ) ( sin 2 θ − cos 2 θ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ = 4 π 1 ∫ 0 2 π K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) ϵ d θ ′ d s ′ e 2 π ik s + ∫ 0 2 π ϵ 2 ∣ k ∣ sin ( 2 θ ′ ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s sin 2 θ + ∫ 0 2 π ϵ 2 ∣ k ∣ sin 3 ( 2 θ ′ ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s ( cos 2 θ − sin 2 θ ) = 2 ϵ ( I 0 K 0 + I 1 K 1 ) e 2 π ik s sin 2 θ + ϵ ( π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) + 2 1 ( I 0 K 0 − I 1 K 1 ) ) e 2 π ik s cos 2 θ .
As usual, I j = I j ( 2 π ϵ ∣ k ∣ ) I_j=I_j(2\pi\epsilon|k|) I j = I j ( 2 π ϵ ∣ k ∣ ) and K j = K j ( 2 π ϵ ∣ k ∣ ) K_j=K_j(2\pi\epsilon|k|) K j = K j ( 2 π ϵ ∣ k ∣ ) in the final expressions. We may also calculate
S ‾ [ e 2 π i k s cos θ e r ] ⋅ e x = 1 8 π ∫ − ∞ ∞ ∫ 0 2 π ( 8 ϵ 2 ( sin 4 ( θ ′ 2 ) − sin 6 ( θ ′ 2 ) ) ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) 3 + 4 sin 2 ( θ ′ 2 ) − 4 sin 4 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ sin 2 θ \begin{aligned}
\overline{S}[e^{2\pi i k s}\cos\theta e_r]\cdot e_x \\
&=\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\left(\frac{8\epsilon^2\left(\sin^4\left(\frac{\theta'}{2}\right)-\sin^6\left(\frac{\theta'}{2}\right)\right)}{\left(\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}\right)^3}+\frac{4\sin^2\left(\frac{\theta'}{2}\right)-4\sin^4\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}}\right)e^{2\pi i k(s-s')}\epsilon\,d\theta' ds'\sin^2\theta
\end{aligned} S [ e 2 π ik s cos θ e r ] ⋅ e x = 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) ) 3 8 ϵ 2 ( sin 4 ( 2 θ ′ ) − sin 6 ( 2 θ ′ ) ) + ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 4 sin 2 ( 2 θ ′ ) − 4 sin 4 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ sin 2 θ
− 1 8 π ∫ − ∞ ∞ ∫ 0 2 π ( 4 ϵ 2 sin 4 ( θ ′ 2 ) − 8 ϵ 2 sin 6 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) − 1 − 4 sin 2 ( θ ′ 2 ) + 4 sin 4 ( θ ′ 2 ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ cos 2 θ = ∫ 0 2 π ( ϵ π ( sin 2 ( θ ′ 2 ) − sin 4 ( θ ′ 2 ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) + 2 ϵ 2 ∣ k ∣ ( sin 3 ( θ ′ 2 ) − sin 5 ( θ ′ 2 ) ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) ) d θ ′ e 2 π i k s sin 2 θ + ∫ 0 2 π ( ϵ 4 π ( 1 − 4 sin 2 ( θ ′ 2 ) + 4 sin 4 ( θ ′ 2 ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) − ϵ 2 ∣ k ∣ ( sin 3 ( θ ′ 2 ) − 2 sin 5 ( θ ′ 2 ) ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) ) d θ ′ e 2 π i k s cos 2 θ = ϵ ( 3 4 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 3 4 π 2 ϵ 2 k 2 I 1 K 1 − 1 2 ( I 0 K 0 − I 1 K 1 ) ) e 2 π i k s sin 2 θ − ϵ ( 3 4 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 3 4 π 2 ϵ 2 k 2 I 1 K 1 − π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) − ( I 0 K 0 − I 1 K 1 ) ) e 2 π i k s cos 2 θ , (360) \begin{aligned}
-\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_0^{2\pi}\left(\frac{4\epsilon^2\sin^4\left(\frac{\theta'}{2}\right)-8\epsilon^2\sin^6\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}}-\frac{1-4\sin^2\left(\frac{\theta'}{2}\right)+4\sin^4\left(\frac{\theta'}{2}\right)}{\sqrt{(s')^2+4\epsilon^2\sin^2\left(\frac{\theta'}{2}\right)}}\right)e^{2\pi i k(s-s')}\epsilon\,d\theta' ds'\cos^2\theta\\
&=\int_0^{2\pi}\left(\frac{\epsilon}{\pi}\left(\sin^2\left(\frac{\theta'}{2}\right)-\sin^4\left(\frac{\theta'}{2}\right)\right)K_0\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)
+2\epsilon^2|k|\left(\sin^3\left(\frac{\theta'}{2}\right)-\sin^5\left(\frac{\theta'}{2}\right)\right)K_1\left(4\pi\epsilon|k|\sin\left(\frac{\theta'}{2}\right)\right)\right)d\theta'e^{2\pi i k s}\sin^2\theta\\
&+\int_{0}^{2\pi}\left(\frac{\epsilon}{4\pi}\left(1-4\sin^{2}\left(\frac{\theta'}{2}\right)+4\sin^{4}\left(\frac{\theta'}{2}\right)\right)K_{0}\left(4\pi\epsilon\lvert k\rvert\sin\left(\frac{\theta'}{2}\right)\right)
-\epsilon^{2}\lvert k\rvert\left(\sin^{3}\left(\frac{\theta'}{2}\right)-2\sin^{5}\left(\frac{\theta'}{2}\right)\right)K_{1}\left(4\pi\epsilon\lvert k\rvert\sin\left(\frac{\theta'}{2}\right)\right)\right)d\theta' e^{2\pi iks}\cos^{2}\theta\\
&=\epsilon\left(\frac{3}{4\pi\epsilon\lvert k\rvert}(I_{1}K_{0}-I_{0}K_{1})+\frac{3}{4\pi^{2}\epsilon^{2}k^{2}}I_{1}K_{1}-\frac{1}{2}(I_{0}K_{0}-I_{1}K_{1})\right)e^{2\pi iks}\sin^{2}\theta
-\epsilon\left(\frac{3}{4\pi\epsilon\lvert k\rvert}(I_{1}K_{0}-I_{0}K_{1})+\frac{3}{4\pi^{2}\epsilon^{2}k^{2}}I_{1}K_{1}-\pi\epsilon\lvert k\rvert(I_{0}K_{1}-I_{1}K_{0})-(I_{0}K_{0}-I_{1}K_{1})\right)e^{2\pi iks}\cos^{2}\theta, \tag*{(360)}
\end{aligned} − 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 4 ϵ 2 sin 4 ( 2 θ ′ ) − 8 ϵ 2 sin 6 ( 2 θ ′ ) − ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 1 − 4 sin 2 ( 2 θ ′ ) + 4 sin 4 ( 2 θ ′ ) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ cos 2 θ = ∫ 0 2 π ( π ϵ ( sin 2 ( 2 θ ′ ) − sin 4 ( 2 θ ′ ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) + 2 ϵ 2 ∣ k ∣ ( sin 3 ( 2 θ ′ ) − sin 5 ( 2 θ ′ ) ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) ) d θ ′ e 2 π ik s sin 2 θ + ∫ 0 2 π ( 4 π ϵ ( 1 − 4 sin 2 ( 2 θ ′ ) + 4 sin 4 ( 2 θ ′ ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) − ϵ 2 ∣ k ∣ ( sin 3 ( 2 θ ′ ) − 2 sin 5 ( 2 θ ′ ) ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) ) d θ ′ e 2 π ik s cos 2 θ = ϵ ( 4 π ϵ ∣ k ∣ 3 ( I 1 K 0 − I 0 K 1 ) + 4 π 2 ϵ 2 k 2 3 I 1 K 1 − 2 1 ( I 0 K 0 − I 1 K 1 ) ) e 2 π ik s sin 2 θ − ϵ ( 4 π ϵ ∣ k ∣ 3 ( I 1 K 0 − I 0 K 1 ) + 4 π 2 ϵ 2 k 2 3 I 1 K 1 − π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) − ( I 0 K 0 − I 1 K 1 ) ) e 2 π ik s cos 2 θ , (360)
from which we may then obtain
S ‾ [ e 2 π i k s sin θ e θ ] ⋅ e x = ϵ ( 3 4 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 3 4 π 2 ϵ 2 k 2 I 1 K 1 − I 0 K 0 ) e 2 π i k s sin 2 θ + ϵ ( − 3 4 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) − 3 4 π 2 ϵ 2 k 2 I 1 K 1 + 1 2 ( I 0 K 0 − I 1 K 1 ) ) e 2 π i k s cos 2 θ . \overline{S}[e^{2\pi iks}\sin\theta e_{\theta}]\cdot e_{x}
=\epsilon\left(\frac{3}{4\pi\epsilon\lvert k\rvert}(I_{1}K_{0}-I_{0}K_{1})+\frac{3}{4\pi^{2}\epsilon^{2}k^{2}}I_{1}K_{1}-I_{0}K_{0}\right)e^{2\pi iks}\sin^{2}\theta
+\epsilon\left(-\frac{3}{4\pi\epsilon\lvert k\rvert}(I_{1}K_{0}-I_{0}K_{1})-\frac{3}{4\pi^{2}\epsilon^{2}k^{2}}I_{1}K_{1}+\frac{1}{2}(I_{0}K_{0}-I_{1}K_{1})\right)e^{2\pi iks}\cos^{2}\theta. S [ e 2 π ik s sin θ e θ ] ⋅ e x = ϵ ( 4 π ϵ ∣ k ∣ 3 ( I 1 K 0 − I 0 K 1 ) + 4 π 2 ϵ 2 k 2 3 I 1 K 1 − I 0 K 0 ) e 2 π ik s sin 2 θ + ϵ ( − 4 π ϵ ∣ k ∣ 3 ( I 1 K 0 − I 0 K 1 ) − 4 π 2 ϵ 2 k 2 3 I 1 K 1 + 2 1 ( I 0 K 0 − I 1 K 1 ) ) e 2 π ik s cos 2 θ .
In the e z e_{z} e z direction, we additionally have
S ‾ [ e 2 π i k s cos θ e r ] ⋅ e z = − 1 8 π ∫ − ∞ ∞ ∫ 0 2 π 2 ϵ s ′ ( sin 2 ( θ ′ 2 ) − 2 sin 4 ( θ ′ 2 ) ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = − ∫ 0 2 π i ϵ 2 k ( − sin 2 ( θ ′ 2 ) + 2 sin 4 ( θ ′ 2 ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s cos θ = − ϵ i ( π ϵ ∣ k ∣ ( I 1 K 1 − I 0 K 0 ) + 1 2 ( I 1 K 0 − I 0 K 1 ) + 1 2 π ϵ ∣ k ∣ I 1 K 1 ) e 2 π i k s cos θ , \begin{aligned}
\overline{S}[e^{2\pi iks}\cos\theta e_{r}]\cdot e_{z}
=-\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_{0}^{2\pi}
\frac{2\epsilon s'(\sin^{2}(\frac{\theta'}{2})-2\sin^{4}(\frac{\theta'}{2}))}{\sqrt{(s')^{2}+4\epsilon^{2}\sin^{2}(\frac{\theta'}{2})}}
e^{2\pi ik(s-s')}\epsilon\,d\theta'\,ds'\cos\theta\\
&=-\int_{0}^{2\pi}i\epsilon^{2}k\left(-\sin^{2}\left(\frac{\theta'}{2}\right)+2\sin^{4}\left(\frac{\theta'}{2}\right)\right)K_{0}\left(4\pi\epsilon\lvert k\rvert\sin\left(\frac{\theta'}{2}\right)\right)d\theta' e^{2\pi iks}\cos\theta\\
&=-\epsilon i\left(\pi\epsilon\lvert k\rvert(I_{1}K_{1}-I_{0}K_{0})+\frac{1}{2}(I_{1}K_{0}-I_{0}K_{1})+\frac{1}{2\pi\epsilon\lvert k\rvert}I_{1}K_{1}\right)e^{2\pi iks}\cos\theta,
\end{aligned} S [ e 2 π ik s cos θ e r ] ⋅ e z = − 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 2 ϵ s ′ ( sin 2 ( 2 θ ′ ) − 2 sin 4 ( 2 θ ′ )) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = − ∫ 0 2 π i ϵ 2 k ( − sin 2 ( 2 θ ′ ) + 2 sin 4 ( 2 θ ′ ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s cos θ = − ϵ i ( π ϵ ∣ k ∣ ( I 1 K 1 − I 0 K 0 ) + 2 1 ( I 1 K 0 − I 0 K 1 ) + 2 π ϵ ∣ k ∣ 1 I 1 K 1 ) e 2 π ik s cos θ ,
S ‾ [ e 2 π i k s sin θ e θ ] ⋅ e z = − 1 8 π ∫ − ∞ ∞ ∫ 0 2 π 4 ϵ s ′ ( sin 2 ( θ ′ 2 ) − sin 4 ( θ ′ 2 ) ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = − ∫ 0 2 π i 2 ϵ 2 k ( − sin 2 ( θ ′ 2 ) + sin 4 ( θ ′ 2 ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s cos θ = − ϵ i ( 1 2 ( I 1 K 0 − I 0 K 1 ) + 1 2 π ϵ ∣ k ∣ I 1 K 1 ) e 2 π i k s cos θ , \begin{aligned}
\overline{S}[e^{2\pi iks}\sin\theta e_{\theta}]\cdot e_{z}
=-\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_{0}^{2\pi}
\frac{4\epsilon s'(\sin^{2}(\frac{\theta'}{2})-\sin^{4}(\frac{\theta'}{2}))}{\sqrt{(s')^{2}+4\epsilon^{2}\sin^{2}(\frac{\theta'}{2})}}
e^{2\pi ik(s-s')}\epsilon\,d\theta'\,ds'\cos\theta\\
&=-\int_{0}^{2\pi}i2\epsilon^{2}k\left(-\sin^{2}\left(\frac{\theta'}{2}\right)+\sin^{4}\left(\frac{\theta'}{2}\right)\right)K_{0}\left(4\pi\epsilon\lvert k\rvert\sin\left(\frac{\theta'}{2}\right)\right)d\theta' e^{2\pi iks}\cos\theta\\
&=-\epsilon i\left(\frac{1}{2}(I_{1}K_{0}-I_{0}K_{1})+\frac{1}{2\pi\epsilon\lvert k\rvert}I_{1}K_{1}\right)e^{2\pi iks}\cos\theta,
\end{aligned} S [ e 2 π ik s sin θ e θ ] ⋅ e z = − 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 4 ϵ s ′ ( sin 2 ( 2 θ ′ ) − sin 4 ( 2 θ ′ )) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = − ∫ 0 2 π i 2 ϵ 2 k ( − sin 2 ( 2 θ ′ ) + sin 4 ( 2 θ ′ ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s cos θ = − ϵ i ( 2 1 ( I 1 K 0 − I 0 K 1 ) + 2 π ϵ ∣ k ∣ 1 I 1 K 1 ) e 2 π ik s cos θ ,
S ‾ [ e 2 π i k s cos θ e z ] ⋅ e z = 1 8 π ∫ − ∞ ∞ ∫ 0 2 π ( 2 ( 1 − 2 sin 2 ( θ ′ 2 ) ) ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) − 4 ϵ 2 ( sin 2 ( θ ′ 2 ) − 2 sin 4 ( θ ′ 2 ) ) ( ( s ′ ) 2 + 4 ϵ 2 sin 2 ( θ ′ 2 ) ) 3 / 2 ) e 2 π i k ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = 1 2 π ∫ 0 2 π ( 1 − 2 sin 2 ( θ ′ 2 ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) ϵ d θ ′ e 2 π i k s cos θ − ∫ 0 2 π ϵ 2 ∣ k ∣ ( sin ( θ ′ 2 ) − 2 sin 3 ( θ ′ 2 ) ) K 1 ( 4 π ϵ ∣ k ∣ sin ( θ ′ 2 ) ) d θ ′ e 2 π i k s cos θ = ϵ ( π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) ) e 2 π i k s cos θ . \begin{aligned}
\overline{S}[e^{2\pi iks}\cos\theta e_{z}]\cdot e_{z}
=\frac{1}{8\pi}\int_{-\infty}^{\infty}\int_{0}^{2\pi}
\left(\frac{2(1-2\sin^{2}(\frac{\theta'}{2}))}{\sqrt{(s')^{2}+4\epsilon^{2}\sin^{2}(\frac{\theta'}{2})}}
-\frac{4\epsilon^{2}(\sin^{2}(\frac{\theta'}{2})-2\sin^{4}(\frac{\theta'}{2}))}{((s')^{2}+4\epsilon^{2}\sin^{2}(\frac{\theta'}{2}))^{3/2}}\right)e^{2\pi ik(s-s')}\epsilon\,d\theta'\,ds'\cos\theta\\
&=\frac{1}{2\pi}\int_{0}^{2\pi}\left(1-2\sin^{2}\left(\frac{\theta'}{2}\right)\right)K_{0}\left(4\pi\epsilon\lvert k\rvert\sin\left(\frac{\theta'}{2}\right)\right)\epsilon\,d\theta' e^{2\pi iks}\cos\theta
-\int_{0}^{2\pi}\epsilon^{2}\lvert k\rvert\left(\sin\left(\frac{\theta'}{2}\right)-2\sin^{3}\left(\frac{\theta'}{2}\right)\right)K_{1}\left(4\pi\epsilon\lvert k\rvert\sin\left(\frac{\theta'}{2}\right)\right)d\theta' e^{2\pi iks}\cos\theta\\
&=\epsilon\left(\pi\epsilon\lvert k\rvert(I_{0}K_{1}-I_{1}K_{0})\right)e^{2\pi iks}\cos\theta.
\end{aligned} S [ e 2 π ik s cos θ e z ] ⋅ e z = 8 π 1 ∫ − ∞ ∞ ∫ 0 2 π ( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) 2 ( 1 − 2 sin 2 ( 2 θ ′ )) − (( s ′ ) 2 + 4 ϵ 2 sin 2 ( 2 θ ′ ) ) 3/2 4 ϵ 2 ( sin 2 ( 2 θ ′ ) − 2 sin 4 ( 2 θ ′ )) e 2 π ik ( s − s ′ ) ϵ d θ ′ d s ′ cos θ = 2 π 1 ∫ 0 2 π ( 1 − 2 sin 2 ( 2 θ ′ ) ) K 0 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) ϵ d θ ′ e 2 π ik s cos θ − ∫ 0 2 π ϵ 2 ∣ k ∣ ( sin ( 2 θ ′ ) − 2 sin 3 ( 2 θ ′ ) ) K 1 ( 4 π ϵ ∣ k ∣ sin ( 2 θ ′ ) ) d θ ′ e 2 π ik s cos θ = ϵ ( π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) ) e 2 π ik s cos θ .
To summarize, we may write the action of S ‾ \overline{S} S on the vectors ( cos θ e r , sin θ e θ , cos θ e z ) (\cos\theta e_r,\sin\theta e_\theta,\cos\theta e_z) ( cos θ e r , sin θ e θ , cos θ e z ) associated with the e x e_x e x direction as
S ‾ ( e 2 π i k s cos θ e r e 2 π i k s sin θ e θ e 2 π i k s cos θ e z ) = ϵ ( Q H Q I − i Q J Q I Q K − i Q L i Q J i Q L Q M ) ( e 2 π i k s cos θ e r e 2 π i k s sin θ e θ e 2 π i k s cos θ e z ) , \overline{S}\begin{pmatrix} e^{2\pi iks}\cos\theta e_r \\ e^{2\pi iks}\sin\theta e_\theta\\ e^{2\pi iks}\cos\theta e_z \end{pmatrix}
=
\epsilon
\begin{pmatrix}
Q_H & Q_I & -iQ_J \\
Q_I & Q_K & -iQ_L \\
iQ_J & iQ_L & Q_M
\end{pmatrix}
\begin{pmatrix} e^{2\pi iks}\cos\theta e_r \\ e^{2\pi iks}\sin\theta e_\theta\\ e^{2\pi iks}\cos\theta e_z \end{pmatrix}, S e 2 π ik s cos θ e r e 2 π ik s sin θ e θ e 2 π ik s cos θ e z = ϵ Q H Q I i Q J Q I Q K i Q L − i Q J − i Q L Q M e 2 π ik s cos θ e r e 2 π ik s sin θ e θ e 2 π ik s cos θ e z ,
Q H = − 3 4 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) − 3 4 π 2 ϵ 2 k 2 I 1 K 1 + π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) + ( I 0 K 0 − I 1 K 1 ) Q_H=-\frac{3}{4\pi\epsilon|k|}(I_1K_0-I_0K_1)-\frac{3}{4\pi^2\epsilon^2k^2}I_1K_1+\pi\epsilon|k|(I_0K_1-I_1K_0)+(I_0K_0-I_1K_1) Q H = − 4 π ϵ ∣ k ∣ 3 ( I 1 K 0 − I 0 K 1 ) − 4 π 2 ϵ 2 k 2 3 I 1 K 1 + π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) + ( I 0 K 0 − I 1 K 1 )
Q I = − 3 4 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) − 3 4 π 2 ϵ 2 k 2 I 1 K 1 + 1 2 ( I 0 K 0 − I 1 K 1 ) Q_I=-\frac{3}{4\pi\epsilon|k|}(I_1K_0-I_0K_1)-\frac{3}{4\pi^2\epsilon^2k^2}I_1K_1+\frac{1}{2}(I_0K_0-I_1K_1) Q I = − 4 π ϵ ∣ k ∣ 3 ( I 1 K 0 − I 0 K 1 ) − 4 π 2 ϵ 2 k 2 3 I 1 K 1 + 2 1 ( I 0 K 0 − I 1 K 1 )
Q J = π ϵ ∣ k ∣ ( I 1 K 1 − I 0 K 0 ) + 1 2 ( I 1 K 0 − I 0 K 1 ) + 1 2 π ϵ ∣ k ∣ I 1 K 1 Q_J=\pi\epsilon|k|(I_1K_1-I_0K_0)+\frac{1}{2}(I_1K_0-I_0K_1)+\frac{1}{2\pi\epsilon|k|}I_1K_1 Q J = π ϵ ∣ k ∣ ( I 1 K 1 − I 0 K 0 ) + 2 1 ( I 1 K 0 − I 0 K 1 ) + 2 π ϵ ∣ k ∣ 1 I 1 K 1
Q K = − ( 3 4 π ϵ ∣ k ∣ ( I 1 K 0 − I 0 K 1 ) + 3 4 π 2 ϵ 2 k 2 I 1 K 1 − I 0 K 0 ) Q_K=-\left(\frac{3}{4\pi\epsilon|k|}(I_1K_0-I_0K_1)+\frac{3}{4\pi^2\epsilon^2k^2}I_1K_1-I_0K_0\right) Q K = − ( 4 π ϵ ∣ k ∣ 3 ( I 1 K 0 − I 0 K 1 ) + 4 π 2 ϵ 2 k 2 3 I 1 K 1 − I 0 K 0 )
Q L = 1 2 ( I 1 K 0 − I 0 K 1 ) + 1 2 π ϵ ∣ k ∣ I 1 K 1 Q_L=\frac{1}{2}(I_1K_0-I_0K_1)+\frac{1}{2\pi\epsilon|k|}I_1K_1 Q L = 2 1 ( I 1 K 0 − I 0 K 1 ) + 2 π ϵ ∣ k ∣ 1 I 1 K 1
Q M = π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) , (361) Q_M=\pi\epsilon|k|(I_0K_1-I_1K_0), \tag*{(361)} Q M = π ϵ ∣ k ∣ ( I 0 K 1 − I 1 K 0 ) , (361)
with each I j I_j I j , K j K_j K j evaluated at 2 π ϵ ∣ k ∣ 2\pi\epsilon|k| 2 π ϵ ∣ k ∣ . Inverting the matrix (361) , we arrive at the expression (54) , (55) for M S , n − 1 ( k ) M_{\mathrm{S,n}}^{-1}(k) M S , n − 1 ( k ) stated in Lemma 1.7. The same expression for the y y y -direction follows by symmetry. □
Appendix C. Proof of Proposition 2.1: commutator estimate
Here we show the commutator estimate (88) for the multiplier L ϵ ∂ s 2 \mathcal{L}_{\epsilon}\partial_s^2 L ϵ ∂ s 2 . We will rely on the tools of section 3.1. Before proving Proposition 2.1, we show the following auxiliary lemma.
Lemma C.1 (Order one commutator estimate). Let M ^ ( ξ ) \widehat{M}(\xi) M ( ξ ) , ξ ∈ R \xi\in\mathbb{R} ξ ∈ R , be a matrix-valued W l o c 2 , ∞ W_{\mathrm{loc}}^{2,\infty} W loc 2 , ∞ function satisfying
∣ ϕ j ( ξ ) M ^ ( ξ ) ∣ ≤ b 0 2 j , ∣ ∂ ξ 2 ( ϕ j ( ξ ) M ^ ( ξ ) ) ∣ ≤ b 1 2 − j , (362) \left|\phi_j(\xi)\widehat{M}(\xi)\right|\le b_0 2^j,\qquad\left|\partial_\xi^2\left(\phi_j(\xi)\widehat{M}(\xi)\right)\right|\le b_1 2^{-j}, \tag*{(362)} ϕ j ( ξ ) M ( ξ ) ≤ b 0 2 j , ∂ ξ 2 ( ϕ j ( ξ ) M ( ξ ) ) ≤ b 1 2 − j , (362) for ϕ j ( ξ ) \phi_j(\xi) ϕ j ( ξ ) as in (106) . Given matrix-valued A ∈ C n + 1 , α ( R ) \mathbf{A}\in C^{n+1,\alpha}(\mathbb{R}) A ∈ C n + 1 , α ( R ) and vector-valued g ∈ C n , α ( R ) \mathbf{g}\in C^{n,\alpha}(\mathbb{R}) g ∈ C n , α ( R ) , 0 ≤ n ∈ Z 0\le n\in\mathbb{Z} 0 ≤ n ∈ Z , 0 < α < 1 0<\alpha<1 0 < α < 1 , the commutator T M ^ ( A g ) − A ( T M ^ g ) T_{\widehat{M}}(\mathbf{A}\mathbf{g})-\mathbf{A}(T_{\widehat{M}}\mathbf{g}) T M ( Ag ) − A ( T M g ) satisfies
∥ T M ^ ( A g ) − A ( T M ^ g ) ∥ C n , α ≤ c ( b 0 b 1 + b 1 ) ∥ A ∥ C n + 1 , α ∥ g ∥ C n , α . (363) \left\|T_{\widehat{M}}(\mathbf{A}\mathbf{g})-\mathbf{A}(T_{\widehat{M}}\mathbf{g})\right\|_{C^{n,\alpha}}\le c\left(\sqrt{b_0b_1}+b_1\right)\|\mathbf{A}\|_{C^{n+1,\alpha}}\|\mathbf{g}\|_{C^{n,\alpha}}. \tag*{(363)} ∥ T M ( Ag ) − A ( T M g ) ∥ C n , α ≤ c ( b 0 b 1 + b 1 ) ∥ A ∥ C n + 1 , α ∥ g ∥ C n , α . (363)
Proof. Following [5 Chapter 2.8] , we begin by defining a decomposition of the product A g \mathbf{A}\mathbf{g} Ag into three pieces capturing interactions among different modes in the Littlewood-Paley decomposition (107) . We write
A g = T L H ( A , g ) + T H L ( A , g ) + T H H ( A , g ) , (364) \mathbf{A}\mathbf{g}=T_{LH}(\mathbf{A},\mathbf{g})+T_{HL}(\mathbf{A},\mathbf{g})+T_{HH}(\mathbf{A},\mathbf{g}), \tag*{(364)} Ag = T L H ( A , g ) + T H L ( A , g ) + T H H ( A , g ) , (364)
where the low-high, high-low, and high-high interactions, respectively, are given by
T L H ( A , g ) = ∑ j ≥ 1 P j A P ≤ j − 2 g T_{LH}(\mathbf{A},\mathbf{g})=\sum_{j\ge1}P_j\mathbf{A}P_{\le j-2}\mathbf{g} T L H ( A , g ) = j ≥ 1 ∑ P j A P ≤ j − 2 g
T H L ( A , g ) = ∑ j ≥ 1 P ≤ j − 2 A P j g T_{HL}(\mathbf{A},\mathbf{g})=\sum_{j\ge1}P_{\le j-2}\mathbf{A}P_j\mathbf{g} T H L ( A , g ) = j ≥ 1 ∑ P ≤ j − 2 A P j g
T H H ( A , g ) = P ≤ 0 A P ≤ 0 g + ∑ ∣ i − j ∣ ≤ 1 i + j ≥ 1 P j A P i g . (365) T_{HH}(\mathbf{A},\mathbf{g})=P_{\le0}\mathbf{A}P_{\le0}\mathbf{g}+\sum_{\substack{|i-j|\le1\\i+j\ge1}}P_j\mathbf{A}P_i\mathbf{g}. \tag*{(365)} T H H ( A , g ) = P ≤ 0 A P ≤ 0 g + ∣ i − j ∣ ≤ 1 i + j ≥ 1 ∑ P j A P i g . (365)
Using the definitions (365) , we decompose the commutator expression (363) into three parts:
T ~ M ~ ( A g ) − A ( T ~ M ~ g ) = J L H + J H L + J H H , J L H = T ~ M ~ T ~ L H ( A , g ) − T ~ L H ( A , T ~ M ~ g ) J H L = T ~ M ~ T ~ H L ( A , g ) − T ~ H L ( A , T ~ M ~ g ) J H H = T ~ M ~ T ~ H H ( A , g ) − T ~ H H ( A , T ~ M ~ g ) . (366) \begin{aligned}
\widetilde{T}_{\widetilde{M}}(Ag)-A(\widetilde{T}_{\widetilde{M}}g)&=J_{LH}+J_{HL}+J_{HH},\\
J_{LH}&=\widetilde{T}_{\widetilde{M}}\widetilde{T}_{LH}(A,g)-\widetilde{T}_{LH}(A,\widetilde{T}_{\widetilde{M}}g)\\
J_{HL}&=\widetilde{T}_{\widetilde{M}}\widetilde{T}_{HL}(A,g)-\widetilde{T}_{HL}(A,\widetilde{T}_{\widetilde{M}}g)\\
J_{HH}&=\widetilde{T}_{\widetilde{M}}\widetilde{T}_{HH}(A,g)-\widetilde{T}_{HH}(A,\widetilde{T}_{\widetilde{M}}g). \tag*{(366)}
\end{aligned} T M ( A g ) − A ( T M g ) J L H J H L J H H = J L H + J H L + J H H , = T M T L H ( A , g ) − T L H ( A , T M g ) = T M T H L ( A , g ) − T H L ( A , T M g ) = T M T H H ( A , g ) − T H H ( A , T M g ) . (366)
We first compute the low-high interactions J L H J_{LH} J L H . For k ≥ 1 k \ge1 k ≥ 1 , we have
P k J L H = P k ( T ~ M ~ ∑ j = k − 2 k + 2 P ≤ j − 2 A P j g − ∑ j = k − 2 k + 2 P ≤ j − 2 A P j ( T ~ M ~ g ) ) = P k ( P k − 5 ≤ ℓ ≤ k + 5 T ~ M ~ ∑ j = k − 2 k + 2 P ≤ j − 2 A P j g − ∑ j = k − 2 k + 2 P ≤ j − 2 A P k − 5 ≤ ℓ ≤ k + 5 T ~ M ~ P j g ) . \begin{aligned}
P_kJ_{LH}=P_k\left(\widetilde{T}_{\widetilde{M}}\sum_{j=k-2}^{k+2}P_{\leq j-2}AP_jg-\sum_{j=k-2}^{k+2}P_{\leq j-2}AP_j(\widetilde{T}_{\widetilde{M}}g)\right) \\
&=P_k\left(P_{k-5\leq\ell\leq k+5}\widetilde{T}_{\widetilde{M}}\sum_{j=k-2}^{k+2}P_{\leq j-2}AP_jg-\sum_{j=k-2}^{k+2}P_{\leq j-2}AP_{k-5\leq\ell\leq k+5}\widetilde{T}_{\widetilde{M}}P_jg\right).
\end{aligned} P k J L H = P k T M j = k − 2 ∑ k + 2 P ≤ j − 2 A P j g − j = k − 2 ∑ k + 2 P ≤ j − 2 A P j ( T M g ) = P k P k − 5 ≤ ℓ ≤ k + 5 T M j = k − 2 ∑ k + 2 P ≤ j − 2 A P j g − j = k − 2 ∑ k + 2 P ≤ j − 2 A P k − 5 ≤ ℓ ≤ k + 5 T M P j g .
Let M k M_k M k denote the physical space kernel corresponding to the operator P k − 5 ≤ ℓ ≤ k + 5 T ~ M ~ P_{k-5\leq\ell\leq k+5}\widetilde{T}_{\widetilde{M}} P k − 5 ≤ ℓ ≤ k + 5 T M . For each j j j , we may write
P k − 5 ≤ ℓ ≤ k + 5 T ~ M ~ P ≤ j − 2 A P j g − P ≤ j − 2 A P k − 5 ≤ ℓ ≤ k + 5 T ~ M ~ P j g = ∫ R ( M k ( s − s ′ ) P ≤ j − 2 A ( s ′ ) P j g ( s ′ ) − P ≤ j − 2 A ( s ) M k ( s − s ′ ) P j g ( s ′ ) ) d s ′ . P_{k-5\leq\ell\leq k+5}\widetilde{T}_{\widetilde{M}}P_{\leq j-2}AP_jg-P_{\leq j-2}AP_{k-5\leq\ell\leq k+5}\widetilde{T}_{\widetilde{M}}P_jg
=\int_{\mathbb{R}}\left(M_k(s-s')P_{\leq j-2}A(s')P_jg(s')-P_{\leq j-2}A(s)M_k(s-s')P_jg(s')\right)\,ds'. P k − 5 ≤ ℓ ≤ k + 5 T M P ≤ j − 2 A P j g − P ≤ j − 2 A P k − 5 ≤ ℓ ≤ k + 5 T M P j g = ∫ R ( M k ( s − s ′ ) P ≤ j − 2 A ( s ′ ) P j g ( s ′ ) − P ≤ j − 2 A ( s ) M k ( s − s ′ ) P j g ( s ′ ) ) d s ′ .
We may then estimate
∣ ∫ R M k ( s − s ′ ) ( P ≤ j − 2 A ( s ′ ) − P ≤ j − 2 A ( s ) ) P j g ( s ′ ) d s ′ ∣ ≤ c ∫ R ∣ M k ( s − s ′ ) ∣ ∣ s − s ′ ∣ ∥ ∂ s ( P ≤ j − 2 A ) ∥ L ∞ ∥ P j g ∥ L ∞ d s ′ ≤ c ∥ ∣ M k ∣ s ∥ L 1 2 j ∥ P ≤ j − 2 A ∥ L ∞ ∥ P j g ∥ L ∞ . (367) \begin{aligned}
\left|\int_{\mathbb{R}}M_k(s-s')\left(P_{\leq j-2}A(s')-P_{\leq j-2}A(s)\right)P_jg(s')\,ds'\right| \\
&\leq c\int_{\mathbb{R}}|M_k(s-s')||s-s'|\|\partial_s(P_{\leq j-2}A)\|_{L^\infty}\|P_jg\|_{L^\infty}\,ds' \\
&\leq c\|\,|M_k|s\|_{L^1}2^j\|P_{\leq j-2}A\|_{L^\infty}\|P_jg\|_{L^\infty}. \tag*{(367)}
\end{aligned} ∫ R M k ( s − s ′ ) ( P ≤ j − 2 A ( s ′ ) − P ≤ j − 2 A ( s ) ) P j g ( s ′ ) d s ′ ≤ c ∫ R ∣ M k ( s − s ′ ) ∣∣ s − s ′ ∣∥ ∂ s ( P ≤ j − 2 A ) ∥ L ∞ ∥ P j g ∥ L ∞ d s ′ ≤ c ∥ ∣ M k ∣ s ∥ L 1 2 j ∥ P ≤ j − 2 A ∥ L ∞ ∥ P j g ∥ L ∞ . (367)
We note that (362) implies that
∫ R ∣ M k ( s ) ∣ ∣ s ∣ d s ≤ R k ∫ ∣ s ∣ ≤ R k ∣ M k ∣ d s + ∫ ∣ s ∣ > R k ∣ M k ∣ ∣ s ∣ d s ≤ R k ∫ ∣ s ∣ ≤ R k ∣ M k ∣ d s + ( ∫ ∣ s ∣ > R k ∣ M k ∣ 2 ∣ s ∣ 4 d s ) 1 / 2 ( ∫ ∣ s ∣ > R k ∣ s ∣ − 2 d s ) 1 / 2 ≤ R k ∥ M k ∥ L 1 ( R ) + c R k − 1 / 2 ∥ ∂ ξ 2 M ^ k ∥ L 2 ( R ) ≤ c R k 2 k b 0 b 1 + c R k − 1 / 2 2 − k / 2 b 1 , \begin{aligned}
\int_{\mathbb{R}}|M_k(s)|\,|s|\,ds\leq R_k\int_{|s|\leq R_k}|M_k|\,ds+\int_{|s|>R_k}|M_k|\,|s|\,ds \\
&\leq R_k\int_{|s|\leq R_k}|M_k|\,ds+\left(\int_{|s|>R_k}|M_k|^2|s|^4\,ds\right)^{1/2}\left(\int_{|s|>R_k}|s|^{-2}\,ds\right)^{1/2} \\
&\leq R_k\|M_k\|_{L^1(\mathbb{R})}+cR_k^{-1/2}\|\partial_\xi^2\widehat{M}_k\|_{L^2(\mathbb{R})}\leq cR_k2^k\sqrt{b_0b_1}+cR_k^{-1/2}2^{-k/2}b_1,
\end{aligned} ∫ R ∣ M k ( s ) ∣ ∣ s ∣ d s ≤ R k ∫ ∣ s ∣ ≤ R k ∣ M k ∣ d s + ∫ ∣ s ∣ > R k ∣ M k ∣ ∣ s ∣ d s ≤ R k ∫ ∣ s ∣ ≤ R k ∣ M k ∣ d s + ( ∫ ∣ s ∣ > R k ∣ M k ∣ 2 ∣ s ∣ 4 d s ) 1/2 ( ∫ ∣ s ∣ > R k ∣ s ∣ − 2 d s ) 1/2 ≤ R k ∥ M k ∥ L 1 ( R ) + c R k − 1/2 ∥ ∂ ξ 2 M k ∥ L 2 ( R ) ≤ c R k 2 k b 0 b 1 + c R k − 1/2 2 − k /2 b 1 ,
where we have used Lemma 3.1 . Choosing R k ∼ 2 − k R_k\sim2^{-k} R k ∼ 2 − k , we have
∫ R ∣ M k ( s ) ∣ ∣ s ∣ d s ≤ c ( b 0 b 1 + b 1 ) , \int_{\mathbb{R}}|M_k(s)|\,|s|\,ds\leq c(\sqrt{b_0b_1}+b_1), ∫ R ∣ M k ( s ) ∣ ∣ s ∣ d s ≤ c ( b 0 b 1 + b 1 ) ,
so, in particular,
∣ ∫ R M k ( s − s ′ ) ( P ≤ j − 2 A ( s ′ ) − P ≤ j − 2 A ( s ) ) P j g ( s ′ ) d s ′ ∣ ≤ c ( b 0 b 1 + b 1 ) 2 j ∥ P ≤ j − 2 A ∥ L ∞ ∥ P j g ∥ L ∞ . \left|\int_{\mathbb{R}}M_k(s-s')\left(P_{\leq j-2}A(s')-P_{\leq j-2}A(s)\right)P_jg(s')\,ds'\right|\leq c(\sqrt{b_0b_1}+b_1)2^j\|P_{\leq j-2}A\|_{L^\infty}\|P_jg\|_{L^\infty}. ∫ R M k ( s − s ′ ) ( P ≤ j − 2 A ( s ′ ) − P ≤ j − 2 A ( s ) ) P j g ( s ′ ) d s ′ ≤ c ( b 0 b 1 + b 1 ) 2 j ∥ P ≤ j − 2 A ∥ L ∞ ∥ P j g ∥ L ∞ .
Summing in j j j and applying P k P_k P k , k ≥ 1 k\geq1 k ≥ 1 , we thus obtain
2 k ( n + α ) ∥ P k J L H ∥ L ∞ ≤ c ( b 0 b 1 + b 1 ) 2 k ( n + α ) ∑ j = k − 2 k + 2 2 j ∥ P ≤ j − 2 A ∥ L ∞ ∥ P j g ∥ L ∞ ≤ c ( b 0 b 1 + b 1 ) ∥ A ∥ C 1 ∥ g ∥ C n , α . \begin{aligned}
2^{k(n+\alpha)}\|P_kJ_{LH}\|_{L^\infty}\leq c(\sqrt{b_0b_1}+b_1)2^{k(n+\alpha)}\sum_{j=k-2}^{k+2}2^j\|P_{\leq j-2}A\|_{L^\infty}\|P_jg\|_{L^\infty} \\
&\leq c(\sqrt{b_0b_1}+b_1)\|A\|_{C^1}\|g\|_{C^{n,\alpha}}.
\end{aligned} 2 k ( n + α ) ∥ P k J L H ∥ L ∞ ≤ c ( b 0 b 1 + b 1 ) 2 k ( n + α ) j = k − 2 ∑ k + 2 2 j ∥ P ≤ j − 2 A ∥ L ∞ ∥ P j g ∥ L ∞ ≤ c ( b 0 b 1 + b 1 ) ∥ A ∥ C 1 ∥ g ∥ C n , α .
For P ≤ 0 J L H P_{\leq0}J_{LH} P ≤ 0 J L H , we have ∥ P ≤ 0 J L H ∥ L ∞ ≤ c ( b 0 b 1 + b 1 ) ∥ A ∥ L ∞ ∥ g ∥ L ∞ \|P_{\leq0}J_{LH}\|_{L^\infty} \leq c(\sqrt{b_0b_1}+b_1)\|A\|_{L^\infty}\|g\|_{L^\infty} ∥ P ≤ 0 J L H ∥ L ∞ ≤ c ( b 0 b 1 + b 1 ) ∥ A ∥ L ∞ ∥ g ∥ L ∞ .
We next consider the high-low interactions J H L J_{HL} J H L in (366) . Similar to before, for k ≥ 1 k \geq1 k ≥ 1 , we have
P k J H L = P k ( P k − 5 ≤ ℓ ≤ k + 5 T M ^ ∑ j = k − 2 k + 2 P j A P ≤ j − 2 g − ∑ j = k − 2 k + 2 P j A P k − 5 ≤ ℓ ≤ k + 5 T M ^ P ≤ j − 2 g ) . P_kJ_{HL}=P_k\left(P_{k-5\leq\ell\leq k+5}T_{\widehat{M}}\sum_{j=k-2}^{k+2}P_jAP_{\leq j-2}g-\sum_{j=k-2}^{k+2}P_jAP_{k-5\leq\ell\leq k+5}T_{\widehat{M}}P_{\leq j-2}g\right). P k J H L = P k P k − 5 ≤ ℓ ≤ k + 5 T M j = k − 2 ∑ k + 2 P j A P ≤ j − 2 g − j = k − 2 ∑ k + 2 P j A P k − 5 ≤ ℓ ≤ k + 5 T M P ≤ j − 2 g .
For each j j j , the analogous integral to (367) is then
∣ ∫ R ( M k ( s − s ′ ) P j A ( s ′ ) P ≤ j − 2 g ( s ′ ) − P j A ( s ) M k ( s − s ′ ) P ≤ j − 2 g ( s ′ ) ) d s ′ ∣ \left|\int_{\mathbb{R}}\left(M_k(s-s')P_jA(s')P_{\leq j-2}g(s')-P_jA(s)M_k(s-s')P_{\leq j-2}g(s')\right)\,ds'\right| ∫ R ( M k ( s − s ′ ) P j A ( s ′ ) P ≤ j − 2 g ( s ′ ) − P j A ( s ) M k ( s − s ′ ) P ≤ j − 2 g ( s ′ ) ) d s ′
≤ 2 ∥ M k ∥ L 1 ∥ P j A ∥ L ∞ ∥ P ≤ j − 2 g ∥ L ∞ ≤ c b 0 b 1 2 k ∥ P j A ∥ L ∞ ∥ P ≤ j − 2 g ∥ L ∞ , \leq2\|M_k\|_{L^1}\|P_jA\|_{L^\infty}\|P_{\leq j-2}g\|_{L^\infty}\leq c\sqrt{b_0b_1}2^k\|P_jA\|_{L^\infty}\|P_{\leq j-2}g\|_{L^\infty}, ≤ 2∥ M k ∥ L 1 ∥ P j A ∥ L ∞ ∥ P ≤ j − 2 g ∥ L ∞ ≤ c b 0 b 1 2 k ∥ P j A ∥ L ∞ ∥ P ≤ j − 2 g ∥ L ∞ ,
where we have used Lemma 3.1 to bound ∥ M k ∥ L 1 \|M_k\|_{L^1} ∥ M k ∥ L 1 . Summing in j j j and applying P k P_k P k , we have
2 k ( n + α ) ∥ P k J H L ∥ L ∞ ≤ c 2 k ( n + α ) ∑ j = k − 2 k + 2 c b 0 b 1 2 k ∥ P j A ∥ L ∞ ∥ P ≤ j − 2 g ∥ L ∞ ≤ c b 0 b 1 ∥ A ∥ C n + 1 , α ∥ g ∥ L ∞ . \begin{aligned}
2^{k(n+\alpha)}\|P_kJ_{HL}\|_{L^\infty}\leq c2^{k(n+\alpha)}\sum_{j=k-2}^{k+2}c\sqrt{b_0b_1}2^k\|P_jA\|_{L^\infty}\|P_{\leq j-2}g\|_{L^\infty} \\
&\leq c\sqrt{b_0b_1}\|A\|_{C^{n+1,\alpha}}\|g\|_{L^\infty}.
\end{aligned} 2 k ( n + α ) ∥ P k J H L ∥ L ∞ ≤ c 2 k ( n + α ) j = k − 2 ∑ k + 2 c b 0 b 1 2 k ∥ P j A ∥ L ∞ ∥ P ≤ j − 2 g ∥ L ∞ ≤ c b 0 b 1 ∥ A ∥ C n + 1 , α ∥ g ∥ L ∞ .
We additionally have ∥ P ≤ 0 J H L ∥ L ∞ ≤ c b 0 b 1 ∥ A ∥ L ∞ ∥ g ∥ L ∞ \|P_{\leq0}J_{HL}\|_{L^\infty}\leq c\sqrt{b_0b_1}\|A\|_{L^\infty}\|g\|_{L^\infty} ∥ P ≤ 0 J H L ∥ L ∞ ≤ c b 0 b 1 ∥ A ∥ L ∞ ∥ g ∥ L ∞ .
Similarly, for J H H J_{HH} J H H , for k ≥ 3 k\geq3 k ≥ 3 we have
P k J H H = P k ( P k − 5 ≤ ℓ ≤ k + 5 T M ^ ∑ j ≥ k − 2 , ∣ i − j ∣ ≤ 1 , i + j ≥ 1 P j A P i g − ∑ j ≥ k − 2 , ∣ i − j ∣ ≤ 1 , i + j ≥ 1 P j A P k − 5 ≤ ℓ ≤ k + 5 T M ^ P i g ) . P_kJ_{HH}=P_k\left(P_{k-5\leq\ell\leq k+5}T_{\widehat{M}}\sum_{\substack{j\geq k-2,\,|i-j|\leq1,\,i+j\geq1}}P_jAP_ig-\sum_{\substack{j\geq k-2,\,|i-j|\leq1,\,i+j\geq1}}P_jAP_{k-5\leq\ell\leq k+5}T_{\widehat{M}}P_ig\right). P k J H H = P k P k − 5 ≤ ℓ ≤ k + 5 T M j ≥ k − 2 , ∣ i − j ∣ ≤ 1 , i + j ≥ 1 ∑ P j A P i g − j ≥ k − 2 , ∣ i − j ∣ ≤ 1 , i + j ≥ 1 ∑ P j A P k − 5 ≤ ℓ ≤ k + 5 T M P i g .
For each j , i j,i j , i satisfying j ≥ k − 2 j\geq k-2 j ≥ k − 2 , ∣ i − j ∣ ≤ 1 |i-j|\leq1 ∣ i − j ∣ ≤ 1 , i + j ≥ 1 i+j\geq1 i + j ≥ 1 , as in (367) , we must estimate the integral
∣ ∫ R ( M k ( s − s ′ ) P j A ( s ′ ) P i g ( s ′ ) − P j A ( s ) M k ( s − s ′ ) P i g ( s ′ ) ) d s ′ ∣ \left|\int_{\mathbb{R}}\left(M_k(s-s')P_jA(s')P_ig(s')-P_jA(s)M_k(s-s')P_ig(s')\right)\,ds'\right| ∫ R ( M k ( s − s ′ ) P j A ( s ′ ) P i g ( s ′ ) − P j A ( s ) M k ( s − s ′ ) P i g ( s ′ ) ) d s ′
≤ 2 ∥ M k ∥ L 1 ∥ P j A ∥ L ∞ ∥ P i g ∥ L ∞ ≤ c b 0 b 1 2 k ∥ P j A ∥ L ∞ ∥ P i g ∥ L ∞ . \leq2\|M_k\|_{L^1}\|P_jA\|_{L^\infty}\|P_ig\|_{L^\infty}\leq c\sqrt{b_0b_1}2^k\|P_jA\|_{L^\infty}\|P_ig\|_{L^\infty}. ≤ 2∥ M k ∥ L 1 ∥ P j A ∥ L ∞ ∥ P i g ∥ L ∞ ≤ c b 0 b 1 2 k ∥ P j A ∥ L ∞ ∥ P i g ∥ L ∞ .
Again summing in j j j and i i i and applying P k P_k P k , we have
2 k ( n + α ) ∥ P k J H H ∥ L ∞ ≤ c b 0 b 1 2 k ( n + α ) ∑ j ≥ k − 2 , ∣ i − j ∣ ≤ 1 , i + j ≥ 1 2 k ∥ P j A ∥ L ∞ ∥ P i g ∥ L ∞ ≤ c b 0 b 1 ∑ j ≥ k − 2 2 k ( n + 1 + α ) 2 − j ( n + 1 + α ) ∥ A ∥ C n + 1 , α 2 − j α ∥ g ∥ C 0 , α ≤ c b 0 b 1 ∥ A ∥ C n + 1 , α ∥ g ∥ C 0 , α . \begin{aligned}
2^{k(n+\alpha)}\|P_kJ_{HH}\|_{L^\infty}\leq c\sqrt{b_0b_1}2^{k(n+\alpha)}\sum_{\substack{j\geq k-2,\,|i-j|\leq1,\,i+j\geq1}}2^k\|P_jA\|_{L^\infty}\|P_ig\|_{L^\infty} \\
&\leq c\sqrt{b_0b_1}\sum_{j\geq k-2}2^{k(n+1+\alpha)}2^{-j(n+1+\alpha)}\|A\|_{C^{n+1,\alpha}}2^{-j\alpha}\|g\|_{C^{0,\alpha}} \\
&\leq c\sqrt{b_0b_1}\|A\|_{C^{n+1,\alpha}}\|g\|_{C^{0,\alpha}}.
\end{aligned} 2 k ( n + α ) ∥ P k J H H ∥ L ∞ ≤ c b 0 b 1 2 k ( n + α ) j ≥ k − 2 , ∣ i − j ∣ ≤ 1 , i + j ≥ 1 ∑ 2 k ∥ P j A ∥ L ∞ ∥ P i g ∥ L ∞ ≤ c b 0 b 1 j ≥ k − 2 ∑ 2 k ( n + 1 + α ) 2 − j ( n + 1 + α ) ∥ A ∥ C n + 1 , α 2 − j α ∥ g ∥ C 0 , α ≤ c b 0 b 1 ∥ A ∥ C n + 1 , α ∥ g ∥ C 0 , α .
Furthermore, for P ≤ 2 J H H P_{\leq2}J_{HH} P ≤ 2 J H H we have ∥ P ≤ 2 J H H ∥ L ∞ ≤ c b 0 b 1 ∥ A ∥ L ∞ ∥ g ∥ L ∞ \|P_{\leq2}J_{HH}\|_{L^\infty}\leq c\sqrt{b_0b_1}\|A\|_{L^\infty}\|g\|_{L^\infty} ∥ P ≤ 2 J H H ∥ L ∞ ≤ c b 0 b 1 ∥ A ∥ L ∞ ∥ g ∥ L ∞ .
In total, we obtain Lemma C.1. □
We may now prove Proposition 2.1 by showing that the operator L ϵ ∂ s 2 \mathcal{L}_\epsilon\partial_s^2 L ϵ ∂ s 2 satisfies the assumptions (362) of Lemma C.1.
Proof of Proposition 2.1. Recalling the definition (186) of the multiplier M ^ ϵ ( 1 ) ( ξ ) \widehat{M}_\epsilon^{(1)}(\xi) M ϵ ( 1 ) ( ξ ) , we may write L ϵ ∂ s 2 \mathcal{L}_\epsilon\partial_s^2 L ϵ ∂ s 2 as T M ^ T_{\widehat{M}} T M where
M ^ ( ξ ) = M ^ ϵ ( 1 ) ( ξ ) ξ 2 . (368) \widehat{M}(\xi)=\widehat{M}_\epsilon^{(1)}(\xi)\xi^2. \tag*{(368)} M ( ξ ) = M ϵ ( 1 ) ( ξ ) ξ 2 . (368)
Using Lemmas 3.4 and 3.5, we immediately have that
∣ ϕ j ( ξ ) M ^ ( ξ ) ∣ ≤ c ϵ − 1 ∣ log ϵ ∣ 2 j , ∣ ∂ ξ 2 ( ϕ j ( ξ ) M ^ ( ξ ) ) ∣ ≤ c ϵ − 1 ∣ log ϵ ∣ 2 − j . (369) \left|\phi_j(\xi)\widehat{M}(\xi)\right|\leq c\epsilon^{-1}|\log\epsilon|2^j,\qquad\left|\partial_\xi^2\left(\phi_j(\xi)\widehat{M}(\xi)\right)\right|\leq c\epsilon^{-1}|\log\epsilon|2^{-j}. \tag*{(369)} ϕ j ( ξ ) M ( ξ ) ≤ c ϵ − 1 ∣ log ϵ ∣ 2 j , ∂ ξ 2 ( ϕ j ( ξ ) M ( ξ ) ) ≤ c ϵ − 1 ∣ log ϵ ∣ 2 − j . (369)
We may thus apply Lemma C.1 with b 0 = b 1 = c ϵ − 1 ∣ log ϵ ∣ b_0=b_1=c\epsilon^{-1}|\log\epsilon| b 0 = b 1 = c ϵ − 1 ∣ log ϵ ∣ to obtain Proposition 2.1. □ \square □
Appendix D. Proof of Lemma 5.4: hypersingular bounds
Using that φ ∈ C 1 , α ( Γ ϵ ) \varphi\in C^{1,\alpha}(\Gamma_\epsilon) φ ∈ C 1 , α ( Γ ϵ ) , we may expand φ \varphi φ about ( s , θ ) (s,\theta) ( s , θ ) as
φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) = − s ^ ∂ s φ ( s , θ ) − ϵ θ ^ 1 ϵ ∂ θ φ ( s , θ ) + ∣ s ^ ∣ 1 + α Q φ , s ( s , θ , s ^ , θ ^ ) + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ( s , θ , s ^ , θ ^ ) , (370) \varphi(s-\hat{s},\theta-\hat{\theta})-\varphi(s,\theta)=-\hat{s}\partial_s\varphi(s,\theta)-\epsilon\hat{\theta}\frac{1}{\epsilon}\partial_\theta\varphi(s,\theta)+|\hat{s}|^{1+\alpha}Q_{\varphi,s}(s,\theta,\hat{s},\hat{\theta})+|\epsilon\hat{\theta}|^{1+\alpha}Q_{\varphi,\theta}(s,\theta,\hat{s},\hat{\theta}), \tag*{(370)} φ ( s − s ^ , θ − θ ^ ) − φ ( s , θ ) = − s ^ ∂ s φ ( s , θ ) − ϵ θ ^ ϵ 1 ∂ θ φ ( s , θ ) + ∣ s ^ ∣ 1 + α Q φ , s ( s , θ , s ^ , θ ^ ) + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ( s , θ , s ^ , θ ^ ) , (370)
where
∥ Q φ , s ∥ L ∞ ≤ c ∥ ∂ s φ ∥ C 0 , α , ∥ Q φ , θ ∥ L ∞ ≤ c ∥ 1 ϵ ∂ θ φ ∥ C 0 , α . (371) \|Q_{\varphi,s}\|_{L^\infty}\le c\|\partial_s\varphi\|_{C^{0,\alpha}},\qquad\|Q_{\varphi,\theta}\|_{L^\infty}\le c\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C^{0,\alpha}}. \tag*{(371)} ∥ Q φ , s ∥ L ∞ ≤ c ∥ ∂ s φ ∥ C 0 , α , ∥ Q φ , θ ∥ L ∞ ≤ c ϵ 1 ∂ θ φ C 0 , α . (371)
Similarly, we may expand φ \varphi φ about ( s 0 + s , θ 0 + θ ) (s_0+s,\theta_0+\theta) ( s 0 + s , θ 0 + θ ) as
φ ( s − s ^ , θ − θ ^ ) − φ ( s 0 + s , θ 0 + θ ) = ( s 0 + s ^ ) H φ , s + ϵ ( θ 0 + θ ^ ) H φ , θ , (372) \varphi(s-\hat{s},\theta-\hat{\theta})-\varphi(s_0+s,\theta_0+\theta)=(s_0+\hat{s})H_{\varphi,s}+\epsilon(\theta_0+\hat{\theta})H_{\varphi,\theta}, \tag*{(372)} φ ( s − s ^ , θ − θ ^ ) − φ ( s 0 + s , θ 0 + θ ) = ( s 0 + s ^ ) H φ , s + ϵ ( θ 0 + θ ^ ) H φ , θ , (372)
but make note of two different expressions for the remainder terms H φ H_\varphi H φ . The first is given by
( s 0 + s ^ ) H φ , s = − ( s 0 + s ^ ) ∂ s φ ( s 0 + s , θ 0 + θ ) + ∣ s 0 + s ^ ∣ 1 + α Q φ , s ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) , ϵ ( θ 0 + θ ^ ) H φ , θ = − ϵ ( θ 0 + θ ^ ) 1 ϵ ∂ θ φ ( s 0 + s , θ 0 + θ ) + ∣ ϵ ( θ 0 + θ ^ ) ∣ 1 + α Q φ , θ ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) , (373) \begin{aligned}
(s_0+\hat{s})H_{\varphi,s}&=-(s_0+\hat{s})\partial_s\varphi(s_0+s,\theta_0+\theta)\\
&\quad+|s_0+\hat{s}|^{1+\alpha}Q_{\varphi,s}(s_0+s,\theta_0+\theta,s_0+\hat{s},\theta_0+\hat{\theta}),\\
\epsilon(\theta_0+\hat{\theta})H_{\varphi,\theta}&=-\epsilon(\theta_0+\hat{\theta})\frac{1}{\epsilon}\partial_\theta\varphi(s_0+s,\theta_0+\theta)\\
&\quad+|\epsilon(\theta_0+\hat{\theta})|^{1+\alpha}Q_{\varphi,\theta}(s_0+s,\theta_0+\theta,s_0+\hat{s},\theta_0+\hat{\theta}), \tag*{(373)}
\end{aligned} ( s 0 + s ^ ) H φ , s ϵ ( θ 0 + θ ^ ) H φ , θ = − ( s 0 + s ^ ) ∂ s φ ( s 0 + s , θ 0 + θ ) + ∣ s 0 + s ^ ∣ 1 + α Q φ , s ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) , = − ϵ ( θ 0 + θ ^ ) ϵ 1 ∂ θ φ ( s 0 + s , θ 0 + θ ) + ∣ ϵ ( θ 0 + θ ^ ) ∣ 1 + α Q φ , θ ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) , (373)
i.e. the expansion (368) evaluated at ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) (s_0+s,\theta_0+\theta,s_0+\hat{s},\theta_0+\hat{\theta}) ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) . The second is given by
( s 0 + s ^ ) H φ , s = − ( s 0 + s ^ ) ∂ s φ ( s , θ ) + ∣ s 0 ∣ 1 + α Q φ , s ( s , θ , s 0 + s , θ 0 + θ ) + ∣ s ^ ∣ 1 + α Q φ , s ( s , θ , s , θ ^ ) , ϵ ( θ 0 + θ ^ ) H φ , θ = − ϵ ( θ 0 + θ ^ ) 1 ϵ ∂ θ φ ( s , θ ) + ∣ ϵ θ 0 ∣ 1 + α Q φ , θ ( s , θ , s 0 + s , θ 0 + θ ) + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ( s , θ , s , θ ^ ) . (374) \begin{aligned}
(s_0+\hat{s})H_{\varphi,s}&=-(s_0+\hat{s})\partial_s\varphi(s,\theta)+|s_0|^{1+\alpha}Q_{\varphi,s}(s,\theta,s_0+s,\theta_0+\theta)\\
&\quad+|\hat{s}|^{1+\alpha}Q_{\varphi,s}(s,\theta,s,\hat{\theta}),\\
\epsilon(\theta_0+\hat{\theta})H_{\varphi,\theta}&=-\epsilon(\theta_0+\hat{\theta})\frac{1}{\epsilon}\partial_\theta\varphi(s,\theta)+|\epsilon\theta_0|^{1+\alpha}Q_{\varphi,\theta}(s,\theta,s_0+s,\theta_0+\theta)\\
&\quad+|\epsilon\hat{\theta}|^{1+\alpha}Q_{\varphi,\theta}(s,\theta,s,\hat{\theta}). \tag*{(374)}
\end{aligned} ( s 0 + s ^ ) H φ , s ϵ ( θ 0 + θ ^ ) H φ , θ = − ( s 0 + s ^ ) ∂ s φ ( s , θ ) + ∣ s 0 ∣ 1 + α Q φ , s ( s , θ , s 0 + s , θ 0 + θ ) + ∣ s ^ ∣ 1 + α Q φ , s ( s , θ , s , θ ^ ) , = − ϵ ( θ 0 + θ ^ ) ϵ 1 ∂ θ φ ( s , θ ) + ∣ ϵ θ 0 ∣ 1 + α Q φ , θ ( s , θ , s 0 + s , θ 0 + θ ) + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ( s , θ , s , θ ^ ) . (374)
In both (371) and (372), the terms Q φ , μ Q_{\varphi,\mu} Q φ , μ are as in (369).
Let I ^ φ \hat{I}_\varphi I ^ φ denote the quantity of interest, i.e.
I ^ φ = ∣ p.v. ∫ − 1 / 2 − s 0 1 / 2 − s 0 ∫ − π − θ 0 π − θ 0 I φ , ℓ k n ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) ϵ d θ ^ d s ^ − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π I φ , ℓ k n ( s , θ , s , θ ^ ) ϵ d θ ^ d s ^ ∣ . \hat{I}_\varphi=\left|\operatorname{p.v.}\int_{-1/2-s_0}^{1/2-s_0}\int_{-\pi-\theta_0}^{\pi-\theta_0}I_{\varphi,\ell kn}(s_0+s,\theta_0+\theta,s_0+\hat{s},\theta_0+\hat{\theta})\,\epsilon\,d\hat{\theta}\,d\hat{s}
-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}I_{\varphi,\ell kn}(s,\theta,s,\hat{\theta})\,\epsilon\,d\hat{\theta}\,d\hat{s}\right|. I ^ φ = p.v. ∫ − 1/2 − s 0 1/2 − s 0 ∫ − π − θ 0 π − θ 0 I φ , ℓ k n ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) ϵ d θ ^ d s ^ − p.v. ∫ − 1/2 1/2 ∫ − π π I φ , ℓ k n ( s , θ , s , θ ^ ) ϵ d θ ^ d s ^ .
We begin by considering the case s 0 2 + ϵ 2 θ 0 2 ≥ ϵ \sqrt{s_0^2+\epsilon^2\theta_0^2}\ge\epsilon s 0 2 + ϵ 2 θ 0 2 ≥ ϵ . Using the first expansion (371) for φ \varphi φ , we have
∣ I ^ φ ∣ ≤ 2 ∥ p.v. ∫ − 1 / 2 1 / 2 ∫ − π π s ^ ℓ ( ϵ sin ( θ ^ / 2 ) ) k g ∣ R ∣ n ( s ^ ∂ s φ ( s , θ ) + ϵ θ ^ 1 ϵ ∂ θ φ ( s , θ ) ) ϵ d θ ^ d s ^ ∥ L ∞ + 2 ∥ ∫ − 1 / 2 1 / 2 ∫ − π π s ^ ℓ ( ϵ sin ( θ ^ / 2 ) ) k g ∣ R ∣ n ( ∣ s ^ ∣ 1 + α Q φ , s + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ) ϵ d θ ^ d s ^ ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ϵ α ( ∥ g ∥ C 2 0 , α ( ∥ ∂ s φ ∥ L ∞ + ∥ 1 ϵ ∂ θ φ ∥ L ∞ ) + ∥ g ∥ L ∞ ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) ) ≤ c ( κ ∗ , c Γ ) s 0 2 + ϵ 2 θ 0 2 α ( ∥ g ∥ C 2 0 , α ( ∥ ∂ s φ ∥ L ∞ + ∥ 1 ϵ ∂ θ φ ∥ L ∞ ) + ∥ g ∥ L ∞ ( ∥ ∂ s ϕ ∥ C 0 , α + ∥ 1 ϵ ∂ θ ϕ ∥ C 0 , α ) . \begin{aligned}
|\hat{I}_\varphi|
&\le2\|\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
\frac{\hat{s}^{\ell}(\epsilon\sin(\hat{\theta}/2))^k g}{|R|^n}
(\hat{s}\partial_s\varphi(s,\theta)+\epsilon\hat{\theta}\frac{1}{\epsilon}\partial_\theta\varphi(s,\theta))
\,\epsilon\,d\hat{\theta}\,d\hat{s}\|_{L^\infty}\\
&\quad+2\|\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
\frac{\hat{s}^{\ell}(\epsilon\sin(\hat{\theta}/2))^k g}{|R|^n}
(|\hat{s}|^{1+\alpha}Q_{\varphi,s}+|\epsilon\hat{\theta}|^{1+\alpha}Q_{\varphi,\theta})
\,\epsilon\,d\hat{\theta}\,d\hat{s}\|_{L^\infty}\\
&\le c(\kappa_*,c_\Gamma)\epsilon^\alpha(\|g\|_{C_2^{0,\alpha}}(\|\partial_s\varphi\|_{L^\infty}+\|\frac{1}{\epsilon}\partial_\theta\varphi\|_{L^\infty})+\|g\|_{L^\infty}(\|\partial_s\varphi\|_{C^{0,\alpha}}+\|\frac{1}{\epsilon}\partial_\theta\varphi\|_{C^{0,\alpha}}))\\
&\le c(\kappa_*,c_\Gamma)\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\,\alpha}(\|g\|_{C_2^{0,\alpha}}(\|\partial_s\varphi\|_{L^\infty}+\|\frac{1}{\epsilon}\partial_\theta\varphi\|_{L^\infty}) \\
&+ \lVert g\rVert_{L^\infty}\left(\lVert\partial_s\phi\rVert_{C^{0,\alpha}}+\left\lVert\frac{1}{\epsilon}\partial_\theta\phi\right\rVert_{C^{0,\alpha}}\right).
\end{aligned} ∣ I ^ φ ∣ ≤ 2∥ p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ n s ^ ℓ ( ϵ sin ( θ ^ /2 ) ) k g ( s ^ ∂ s φ ( s , θ ) + ϵ θ ^ ϵ 1 ∂ θ φ ( s , θ )) ϵ d θ ^ d s ^ ∥ L ∞ + 2∥ ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ n s ^ ℓ ( ϵ sin ( θ ^ /2 ) ) k g ( ∣ s ^ ∣ 1 + α Q φ , s + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ) ϵ d θ ^ d s ^ ∥ L ∞ ≤ c ( κ ∗ , c Γ ) ϵ α ( ∥ g ∥ C 2 0 , α ( ∥ ∂ s φ ∥ L ∞ + ∥ ϵ 1 ∂ θ φ ∥ L ∞ ) + ∥ g ∥ L ∞ ( ∥ ∂ s φ ∥ C 0 , α + ∥ ϵ 1 ∂ θ φ ∥ C 0 , α )) ≤ c ( κ ∗ , c Γ ) s 0 2 + ϵ 2 θ 0 2 α ( ∥ g ∥ C 2 0 , α ( ∥ ∂ s φ ∥ L ∞ + ∥ ϵ 1 ∂ θ φ ∥ L ∞ ) + ∥ g ∥ L ∞ ( ∥ ∂ s ϕ ∥ C 0 , α + ϵ 1 ∂ θ ϕ C 0 , α ) .
Here we have used Lemmas 4.3 and 4.2 and the fact that ϵ ≤ s 0 2 + ϵ 2 θ 0 2 \epsilon\le\sqrt{s_0^2+\epsilon^2\theta_0^2} ϵ ≤ s 0 2 + ϵ 2 θ 0 2 .
We next consider the case s 0 2 + ϵ 2 θ 0 2 < ϵ \sqrt{s_0^2+\epsilon^2\theta_0^2}<\epsilon s 0 2 + ϵ 2 θ 0 2 < ϵ . Throughout, we will use the notation
R = R ( s , θ , s ^ , θ ^ ) , R 0 = R ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) , R=R(s,\theta,\hat{s},\hat{\theta}),\qquad R_0=R(s_0+s,\theta_0+\theta,s_0+\hat{s},\theta_0+\hat{\theta}), R = R ( s , θ , s ^ , θ ^ ) , R 0 = R ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) ,
g 0 = g ( s 0 + s , θ 0 + θ , s − s ^ , θ − θ ^ ) , g 0 0 = g 0 ∣ ( s 0 + s ^ , θ 0 + θ ^ ) = ( 0 , 0 ) , g 0 = g ∣ ( s ^ , θ ^ ) = ( 0 , 0 ) g_0=g(s_0+s,\theta_0+\theta,s-\hat{s},\theta-\hat{\theta}),\qquad
g_0^0=\left.g_0\right|_{(s_0+\hat{s},\theta_0+\hat{\theta})=(0,0)},\qquad
g^0=\left.g\right|_{(\hat{s},\hat{\theta})=(0,0)} g 0 = g ( s 0 + s , θ 0 + θ , s − s ^ , θ − θ ^ ) , g 0 0 = g 0 ∣ ( s 0 + s ^ , θ 0 + θ ^ ) = ( 0 , 0 ) , g 0 = g ∣ ( s ^ , θ ^ ) = ( 0 , 0 )
to condense the exposition. Using the above notation as well as the expansions (370) and (372) , we may write
I ^ φ = J φ , s + J φ , θ , \widehat{I}_\varphi=J_{\varphi,s}+J_{\varphi,\theta}, I φ = J φ , s + J φ , θ ,
J φ , s = p.v. ∫ − 1 / 2 − s 0 1 / 2 − s 0 ∫ − π − θ 0 π − θ 0 ( s 0 + s ^ ) ℓ ( ϵ sin ( θ 0 + θ ^ 2 ) ) k g 0 ∣ R 0 ∣ n ( s 0 + s ^ ) H φ , s ϵ d θ ^ d s ^ − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π s ^ ℓ ( ϵ sin ( θ ^ 2 ) ) k g ∣ R ∣ n ( − s ^ ∂ s φ ( s , θ ) + ∣ s ^ ∣ 1 + α Q φ , s ( s , θ , s ^ , θ ^ ) ) ϵ d θ ^ d s ^ , J_{\varphi,s}=\operatorname{p.v.}\int_{-1/2-s_0}^{1/2-s_0}\int_{-\pi-\theta_0}^{\pi-\theta_0}
\frac{(s_0+\hat{s})^\ell\left(\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right)^k g_0}{\lvert R_0\rvert^n}(s_0+\hat{s})H_{\varphi,s}\,\epsilon\,d\hat{\theta}\,d\hat{s}
-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
\frac{\hat{s}^\ell\left(\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\right)^k g}{\lvert R\rvert^n}
\left(-\hat{s}\partial_s\varphi(s,\theta)+\lvert\hat{s}\rvert^{1+\alpha}Q_{\varphi,s}(s,\theta,\hat{s},\hat{\theta})\right)\epsilon\,d\hat{\theta}\,d\hat{s}, J φ , s = p.v. ∫ − 1/2 − s 0 1/2 − s 0 ∫ − π − θ 0 π − θ 0 ∣ R 0 ∣ n ( s 0 + s ^ ) ℓ ( ϵ sin ( 2 θ 0 + θ ^ ) ) k g 0 ( s 0 + s ^ ) H φ , s ϵ d θ ^ d s ^ − p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ n s ^ ℓ ( ϵ sin ( 2 θ ^ ) ) k g ( − s ^ ∂ s φ ( s , θ ) + ∣ s ^ ∣ 1 + α Q φ , s ( s , θ , s ^ , θ ^ ) ) ϵ d θ ^ d s ^ ,
J φ , θ = p.v. ∫ − 1 / 2 − s 0 1 / 2 − s 0 ∫ − π − θ 0 π − θ 0 ( s 0 + s ^ ) ℓ ( ϵ sin ( θ 0 + θ ^ 2 ) ) k g 0 ∣ R 0 ∣ n ϵ ( θ 0 + θ ^ ) H φ , θ ϵ d θ ^ d s ^ − p.v. ∫ − 1 / 2 1 / 2 ∫ − π π s ^ ℓ ( ϵ sin ( θ ^ 2 ) ) k g ∣ R ∣ n ( − ϵ θ ^ 1 ϵ ∂ θ φ ( s , θ ) + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ( s , θ , s ^ , θ ^ ) ) ϵ d θ ^ d s ^ . J_{\varphi,\theta}=\operatorname{p.v.}\int_{-1/2-s_0}^{1/2-s_0}\int_{-\pi-\theta_0}^{\pi-\theta_0}
\frac{(s_0+\hat{s})^\ell\left(\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right)^k g_0}{\lvert R_0\rvert^n}
\epsilon(\theta_0+\hat{\theta})H_{\varphi,\theta}\,\epsilon\,d\hat{\theta}\,d\hat{s}
-\operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
\frac{\hat{s}^\ell\left(\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\right)^k g}{\lvert R\rvert^n}
\left(-\epsilon\hat{\theta}\frac{1}{\epsilon}\partial_\theta\varphi(s,\theta)+\lvert\epsilon\hat{\theta}\rvert^{1+\alpha}Q_{\varphi,\theta}(s,\theta,\hat{s},\hat{\theta})\right)\epsilon\,d\hat{\theta}\,d\hat{s}. J φ , θ = p.v. ∫ − 1/2 − s 0 1/2 − s 0 ∫ − π − θ 0 π − θ 0 ∣ R 0 ∣ n ( s 0 + s ^ ) ℓ ( ϵ sin ( 2 θ 0 + θ ^ ) ) k g 0 ϵ ( θ 0 + θ ^ ) H φ , θ ϵ d θ ^ d s ^ − p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R ∣ n s ^ ℓ ( ϵ sin ( 2 θ ^ ) ) k g ( − ϵ θ ^ ϵ 1 ∂ θ φ ( s , θ ) + ∣ ϵ θ ^ ∣ 1 + α Q φ , θ ( s , θ , s ^ , θ ^ ) ) ϵ d θ ^ d s ^ .
In order to bound these terms, we will need to introduce a function R e v e n R_{\mathrm{even}} R even satisfying
∣ R e v e n ∣ 2 : = ∣ R ‾ ∣ 2 + s ^ 2 Q R , 0 ( s , θ ) + κ 3 ϵ 2 s ^ sin ( θ ^ ) , (375) \lvert R_{\mathrm{even}}\rvert^2:=\lvert\overline{R}\rvert^2+\hat{s}^2Q_{R,0}(s,\theta)+\kappa_3\epsilon^2\hat{s}\sin(\hat{\theta}), \tag*{(375)} ∣ R even ∣ 2 := ∣ R ∣ 2 + s ^ 2 Q R , 0 ( s , θ ) + κ 3 ϵ 2 s ^ sin ( θ ^ ) , (375)
where Q R , 0 Q_{R,0} Q R , 0 is as in (208) , and crucially does not depend on s ^ \hat{s} s ^ or θ ^ \hat{\theta} θ ^ . In particular, since ∣ R ‾ ∣ \lvert\overline{R}\rvert ∣ R ∣ is even about both s ^ = 0 \hat{s}=0 s ^ = 0 and θ ^ = 0 \hat{\theta}=0 θ ^ = 0 , this means that ∣ R e v e n ∣ \lvert R_{\mathrm{even}}\rvert ∣ R even ∣ is also even about s ^ = 0 \hat{s}=0 s ^ = 0 and θ ^ = 0 \hat{\theta}=0 θ ^ = 0 . For nonnegative integers n n n and m m m , we thus have that the expression
s ^ n ( ϵ sin ( θ ^ 2 ) ) m ∣ R e v e n ∣ n + m + 2 , n + m odd , (376) \frac{\hat{s}^n\left(\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\right)^m}{\lvert R_{\mathrm{even}}\rvert^{n+m+2}},\qquad n+m\ \text{odd}, \tag*{(376)} ∣ R even ∣ n + m + 2 s ^ n ( ϵ sin ( 2 θ ^ ) ) m , n + m odd , (376)
is globally odd about s ^ = θ ^ = 0 \hat{s}=\hat{\theta}=0 s ^ = θ ^ = 0 . Therefore we have that
p.v. ∫ − 1 / 2 1 / 2 ∫ − π π s ^ n ( ϵ sin ( θ ^ 2 ) ) m ∣ R e v e n ∣ n + m + 2 ϵ d θ ^ d s ^ = 0 , (377) \operatorname{p.v.}\int_{-1/2}^{1/2}\int_{-\pi}^{\pi}
\frac{\hat{s}^n\left(\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\right)^m}{\lvert R_{\mathrm{even}}\rvert^{n+m+2}}\epsilon\,d\hat{\theta}\,d\hat{s}=0, \tag*{(377)} p.v. ∫ − 1/2 1/2 ∫ − π π ∣ R even ∣ n + m + 2 s ^ n ( ϵ sin ( 2 θ ^ ) ) m ϵ d θ ^ d s ^ = 0 , (377)
and thus kernels of the form (376) with constant-in-s ^ \hat{s} s ^ -and-θ ^ \hat{\theta} θ ^ coefficients may be inserted into the integral expressions for J φ , s J_{\varphi,s} J φ , s and J φ , θ J_{\varphi,\theta} J φ , θ . We additionally note that
∣ R ∣ 2 = ∣ R e v e n ∣ 2 + s ^ 4 Q R , 1 + ϵ s ^ 3 Q R , 2 + ϵ 2 s ^ 2 sin ( θ ^ 2 ) Q R , 3 , \lvert R\rvert^2=\lvert R_{\mathrm{even}}\rvert^2+\hat{s}^4Q_{R,1}+\epsilon\hat{s}^3Q_{R,2}+\epsilon^2\hat{s}^2\sin\left(\frac{\hat{\theta}}{2}\right)Q_{R,3}, ∣ R ∣ 2 = ∣ R even ∣ 2 + s ^ 4 Q R , 1 + ϵ s ^ 3 Q R , 2 + ϵ 2 s ^ 2 sin ( 2 θ ^ ) Q R , 3 ,
1 ∣ R ∣ − 1 ∣ R e v e n ∣ = s ^ 4 Q R , 1 + ϵ s ^ 3 Q R , 2 + ϵ 2 s ^ 2 sin ( θ ^ 2 ) Q R , 3 ∣ R e v e n ∣ ∣ R ∣ ( ∣ R ∣ + ∣ R e v e n ∣ ) , (378) \frac{1}{\lvert R\rvert}-\frac{1}{\lvert R_{\mathrm{even}}\rvert}
=
\frac{\hat{s}^4Q_{R,1}+\epsilon\hat{s}^3Q_{R,2}+\epsilon^2\hat{s}^2\sin\left(\frac{\hat{\theta}}{2}\right)Q_{R,3}}
{\lvert R_{\mathrm{even}}\rvert\lvert R\rvert(\lvert R\rvert+\lvert R_{\mathrm{even}}\rvert)}, \tag*{(378)} ∣ R ∣ 1 − ∣ R even ∣ 1 = ∣ R even ∣ ∣ R ∣ (∣ R ∣ + ∣ R even ∣) s ^ 4 Q R , 1 + ϵ s ^ 3 Q R , 2 + ϵ 2 s ^ 2 sin ( 2 θ ^ ) Q R , 3 , (378)
where each Q R , j Q_{R,j} Q R , j is as in (208) and, as in Lemma 4.1 ,
∣ R e v e n ∣ ≥ c ( κ ∗ , c Γ ) ∣ R ‾ ∣ . (379) \lvert R_{\mathrm{even}}\rvert\ge c(\kappa_*,c_\Gamma)\lvert\overline{R}\rvert. \tag*{(379)} ∣ R even ∣ ≥ c ( κ ∗ , c Γ ) ∣ R ∣ . (379)
We will use the notation R e v e n , 0 R_{\mathrm{even},0} R even , 0 to denote R e v e n ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) R_{\mathrm{even}}(s_0+s,\theta_0+\theta,s_0+\hat{s},\theta_0+\hat{\theta}) R even ( s 0 + s , θ 0 + θ , s 0 + s ^ , θ 0 + θ ^ ) .
Now, in order to bound J φ , s J_{\varphi,s} J φ , s and J φ , θ J_{\varphi,\theta} J φ , θ , we split the integrals into two regions:
R 1 = { ( s ^ , θ ^ ) : ( s 0 + s ^ ) 2 + ϵ 2 ( θ 0 + θ ^ ) 2 ≤ 4 s 0 2 + ϵ 2 θ 0 2 } , R_1=\left\{(\widehat{s},\widehat{\theta}):\sqrt{(s_0+\widehat{s})^2+\epsilon^2(\theta_0+\widehat{\theta})^2}\leq4\sqrt{s_0^2+\epsilon^2\theta_0^2}\right\}, R 1 = { ( s , θ ) : ( s 0 + s ) 2 + ϵ 2 ( θ 0 + θ ) 2 ≤ 4 s 0 2 + ϵ 2 θ 0 2 } ,
R 2 = { ( s ^ , θ ^ ) : ( s 0 + s ^ ) 2 + ϵ 2 ( θ 0 + θ ^ ) 2 > 4 s 0 2 + ϵ 2 θ 0 2 } . R_2=\left\{(\widehat{s},\widehat{\theta}):\sqrt{(s_0+\widehat{s})^2+\epsilon^2(\theta_0+\widehat{\theta})^2}>4\sqrt{s_0^2+\epsilon^2\theta_0^2}\right\}. R 2 = { ( s , θ ) : ( s 0 + s ) 2 + ϵ 2 ( θ 0 + θ ) 2 > 4 s 0 2 + ϵ 2 θ 0 2 } .
For j = s , θ j=s,\theta j = s , θ , let J φ , j , 1 J_{\varphi,j,1} J φ , j , 1 denote the integral of the integrand of J φ , j J_{\varphi,j} J φ , j over the region R 1 R_1 R 1 . Note that s ^ 2 + ϵ 2 θ ^ 2 \sqrt{\widehat{s}^2+\epsilon^2\widehat{\theta}^2} s 2 + ϵ 2 θ 2 then belongs to the region R 1 ′ R'_1 R 1 ′ where s ^ 2 + ϵ 2 θ ^ 2 ≤ 5 s 0 2 + ϵ 2 θ 0 2 \sqrt{\widehat{s}^2+\epsilon^2\widehat{\theta}^2}\leq5\sqrt{s_0^2+\epsilon^2\theta_0^2} s 2 + ϵ 2 θ 2 ≤ 5 s 0 2 + ϵ 2 θ 0 2 .
After inserting ∣ R e v e n ∣ |R_{\mathrm{even}}| ∣ R even ∣ and ∣ R e v e n , 0 ∣ |R_{\mathrm{even},0}| ∣ R even , 0 ∣ and using the first expansion (371) for φ \varphi φ about ( s 0 + s , θ 0 + θ ) (s_0+s,\theta_0+\theta) ( s 0 + s , θ 0 + θ ) , we may bound
∣ J φ , s , 1 ∣ ≤ ∣ ∂ s φ ( s 0 + s , θ 0 + θ ) p.v. ∬ R 1 ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( θ 0 + θ ^ 2 ) ) k ( g 0 ∣ R 0 ∣ n − g 0 0 ∣ R e v e n , 0 ∣ n ) ϵ d θ ^ d s ^ ∣ + ∣ ∂ s φ ( s , θ ) p.v. ∬ R 1 ′ s ^ ℓ + 1 ( ϵ sin ( θ ^ 2 ) ) k ( g ∣ R ∣ n − g 0 ∣ R e v e n ∣ n ) ϵ d θ ^ d s ^ ∣ + ∬ R 1 ∣ s 0 + s ^ ∣ ℓ + 1 + α ∣ ϵ sin ( θ 0 + θ ^ 2 ) ∣ k ∣ g 0 ∣ ∣ R 0 ∣ n ∣ Q φ , s ∣ ϵ d θ ^ d s ^ + ∬ R 1 ′ ∣ s ^ ∣ ℓ + 1 + α ∣ ϵ sin ( θ ^ 2 ) ∣ k ∣ g ∣ ∣ R ∣ n ∣ Q φ , s ∣ ϵ d θ ^ d s ^ ≤ ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α ∬ R 1 1 ∣ R 0 ∣ 2 − α ϵ d θ ^ d s ^ + ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α ∬ R 1 ′ 1 ∣ R ∣ 2 − α ϵ d θ ^ d s ^ + ∥ ∂ s φ ∥ L ∞ ∣ p.v. ∬ R 1 ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( θ 0 + θ ^ 2 ) ) k ( 1 ∣ R 0 ∣ n − 1 ∣ R e v e n , 0 ∣ n ) ϵ d θ ^ d s ^ ∣ g 0 0 + ∥ ∂ s φ ∥ L ∞ ∣ p.v. ∬ R 1 ′ s ^ ℓ + 1 ( ϵ sin ( θ ^ 2 ) ) k ( 1 ∣ R ∣ n − 1 ∣ R e v e n ∣ n ) ϵ d θ ^ d s ^ ∣ g 0 + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ( ∬ R 1 1 ∣ R 0 ∣ 2 − α ϵ d θ ^ d s ^ + ∬ R 1 ′ 1 ∣ R ∣ 2 − α ϵ d θ ^ d s ^ ) ≤ c ( ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ) ( ∬ R 1 1 ∣ R 0 ∣ 2 − α ϵ d θ ^ d s ^ + ∬ R 1 ′ 1 ∣ R ∣ 2 − α ϵ d θ ^ d s ^ ) ≤ c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ) ( ∬ ρ ≤ 4 s 0 2 + ϵ 2 θ 0 2 1 ρ 2 − α ρ d ρ d ϕ + ∬ ρ ≤ 5 s 0 2 + ϵ 2 θ 0 2 1 ρ 2 − α ρ d ρ d ϕ ) ≤ c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α . \begin{aligned}
|J_{\varphi,s,1}| \leq{}&
\left|\partial_s\varphi(s_0+s,\theta_0+\theta)\operatorname{p.v.}\iint_{R_1}(s_0+\widehat{s})^{\ell+1}\left(\epsilon\sin\left(\frac{\theta_0+\widehat{\theta}}{2}\right)\right)^k\left(\frac{g_0}{|R_0|^n}-\frac{g_0^0}{|R_{\mathrm{even},0}|^n}\right)\epsilon\,d\widehat{\theta}d\widehat{s}\right|\\
&+\left|\partial_s\varphi(s,\theta)\operatorname{p.v.}\iint_{R'_1}\widehat{s}^{\ell+1}\left(\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\right)^k\left(\frac{g}{|R|^n}-\frac{g^0}{|R_{\mathrm{even}}|^n}\right)\epsilon\,d\widehat{\theta}d\widehat{s}\right|\\
&+\iint_{R_1}\frac{|s_0+\widehat{s}|^{\ell+1+\alpha}\left|\epsilon\sin\left(\frac{\theta_0+\widehat{\theta}}{2}\right)\right|^k|g_0|}{|R_0|^n}|Q_{\varphi,s}|\epsilon\,d\widehat{\theta}d\widehat{s}
+\iint_{R'_1}\frac{|\widehat{s}|^{\ell+1+\alpha}\left|\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\right|^k|g|}{|R|^n}|Q_{\varphi,s}|\epsilon\,d\widehat{\theta}d\widehat{s}\\
\leq{}&
\|\partial_s\varphi\|_{L^\infty}\|g\|_{C_2^{0,\alpha}}\iint_{R_1}\frac{1}{|R_0|^{2-\alpha}}\epsilon\,d\widehat{\theta}d\widehat{s}
+\|\partial_s\varphi\|_{L^\infty}\|g\|_{C_2^{0,\alpha}}\iint_{R'_1}\frac{1}{|R|^{2-\alpha}}\epsilon\,d\widehat{\theta}d\widehat{s}\\
&+\|\partial_s\varphi\|_{L^\infty}\left|\operatorname{p.v.}\iint_{R_1}(s_0+\widehat{s})^{\ell+1}\left(\epsilon\sin\left(\frac{\theta_0+\widehat{\theta}}{2}\right)\right)^k\left(\frac{1}{|R_0|^n}-\frac{1}{|R_{\mathrm{even},0}|^n}\right)\epsilon\,d\widehat{\theta}d\widehat{s}\right|g_0^0\\
&+\|\partial_s\varphi\|_{L^\infty}\left|\operatorname{p.v.}\iint_{R'_1}\widehat{s}^{\ell+1}\left(\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\right)^k\left(\frac{1}{|R|^n}-\frac{1}{|R_{\mathrm{even}}|^n}\right)\epsilon\,d\widehat{\theta}d\widehat{s}\right|g^0\\
&+\|\partial_s\varphi\|_{C_0^{0,\alpha}}\|g\|_{L^\infty}\left(\iint_{R_1}\frac{1}{|R_0|^{2-\alpha}}\epsilon\,d\widehat{\theta}d\widehat{s}+\iint_{R'_1}\frac{1}{|R|^{2-\alpha}}\epsilon\,d\widehat{\theta}d\widehat{s}\right)\\
\leq{}&
c\left(\|\partial_s\varphi\|_{L^\infty}\|g\|_{C_2^{0,\alpha}}+\|\partial_s\varphi\|_{C_0^{0,\alpha}}\|g\|_{L^\infty}\right)\left(\iint_{R_1}\frac{1}{|R_0|^{2-\alpha}}\epsilon\,d\widehat{\theta}d\widehat{s}+\iint_{R'_1}\frac{1}{|R|^{2-\alpha}}\epsilon\,d\widehat{\theta}d\widehat{s}\right)\\
\leq{}&
c(\kappa_*,c_\Gamma)\left(\|\partial_s\varphi\|_{L^\infty}\|g\|_{C_2^{0,\alpha}}+\|\partial_s\varphi\|_{C_0^{0,\alpha}}\|g\|_{L^\infty}\right)
\left(\iint_{\rho\leq4\sqrt{s_0^2+\epsilon^2\theta_0^2}}\frac{1}{\rho^{2-\alpha}}\rho\,d\rho d\phi
+\iint_{\rho\leq5\sqrt{s_0^2+\epsilon^2\theta_0^2}}\frac{1}{\rho^{2-\alpha}}\rho\,d\rho d\phi\right)\\
\leq{}&
c(\kappa_*,c_\Gamma)\left(\|\partial_s\varphi\|_{L^\infty}\|g\|_{C_2^{0,\alpha}}+\|\partial_s\varphi\|_{C_0^{0,\alpha}}\|g\|_{L^\infty}\right)\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\,\alpha}.
\end{aligned} ∣ J φ , s , 1 ∣ ≤ ≤ ≤ ≤ ≤ ∂ s φ ( s 0 + s , θ 0 + θ ) p.v. ∬ R 1 ( s 0 + s ) ℓ + 1 ( ϵ sin ( 2 θ 0 + θ ) ) k ( ∣ R 0 ∣ n g 0 − ∣ R even , 0 ∣ n g 0 0 ) ϵ d θ d s + ∂ s φ ( s , θ ) p.v. ∬ R 1 ′ s ℓ + 1 ( ϵ sin ( 2 θ ) ) k ( ∣ R ∣ n g − ∣ R even ∣ n g 0 ) ϵ d θ d s + ∬ R 1 ∣ R 0 ∣ n ∣ s 0 + s ∣ ℓ + 1 + α ϵ sin ( 2 θ 0 + θ ) k ∣ g 0 ∣ ∣ Q φ , s ∣ ϵ d θ d s + ∬ R 1 ′ ∣ R ∣ n ∣ s ∣ ℓ + 1 + α ϵ sin ( 2 θ ) k ∣ g ∣ ∣ Q φ , s ∣ ϵ d θ d s ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α ∬ R 1 ∣ R 0 ∣ 2 − α 1 ϵ d θ d s + ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α ∬ R 1 ′ ∣ R ∣ 2 − α 1 ϵ d θ d s + ∥ ∂ s φ ∥ L ∞ p.v. ∬ R 1 ( s 0 + s ) ℓ + 1 ( ϵ sin ( 2 θ 0 + θ ) ) k ( ∣ R 0 ∣ n 1 − ∣ R even , 0 ∣ n 1 ) ϵ d θ d s g 0 0 + ∥ ∂ s φ ∥ L ∞ p.v. ∬ R 1 ′ s ℓ + 1 ( ϵ sin ( 2 θ ) ) k ( ∣ R ∣ n 1 − ∣ R even ∣ n 1 ) ϵ d θ d s g 0 + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ( ∬ R 1 ∣ R 0 ∣ 2 − α 1 ϵ d θ d s + ∬ R 1 ′ ∣ R ∣ 2 − α 1 ϵ d θ d s ) c ( ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ) ( ∬ R 1 ∣ R 0 ∣ 2 − α 1 ϵ d θ d s + ∬ R 1 ′ ∣ R ∣ 2 − α 1 ϵ d θ d s ) c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ) ( ∬ ρ ≤ 4 s 0 2 + ϵ 2 θ 0 2 ρ 2 − α 1 ρ d ρ d ϕ + ∬ ρ ≤ 5 s 0 2 + ϵ 2 θ 0 2 ρ 2 − α 1 ρ d ρ d ϕ ) c ( κ ∗ , c Γ ) ( ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α + ∥ ∂ s φ ∥ C 0 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α .
Here in the second-to-last line we have used that ∣ R ∣ ≥ c ( κ ∗ , c Γ ) s ^ 2 + ϵ 2 θ ^ 2 |R|\geq c(\kappa_*,c_\Gamma)\sqrt{\widehat{s}^2+\epsilon^2\widehat{\theta}^2} ∣ R ∣ ≥ c ( κ ∗ , c Γ ) s 2 + ϵ 2 θ 2 (see [59 Lemma 3.1] ) to convert to polar coordinates ( ρ , ϕ ) (\rho,\phi) ( ρ , ϕ ) over the patch ρ ≤ c s 0 2 + ϵ 2 θ 0 2 < c ϵ \rho\leq c\sqrt{s_0^2+\epsilon^2\theta_0^2}<c\epsilon ρ ≤ c s 0 2 + ϵ 2 θ 0 2 < cϵ . Similarly, we may show that J φ , θ , 1 J_{\varphi,\theta,1} J φ , θ , 1 satisfies
∣ J φ , θ , 1 ∣ ≤ c ( κ ∗ , c Γ ) ( ∥ 1 ϵ ∂ θ φ ∥ L ∞ ∥ g ∥ C 2 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α . |J_{\varphi,\theta,1}|\leq c(\kappa_*,c_\Gamma)\left(\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{L^\infty}\|g\|_{C_2^{0,\alpha}}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C_0^{0,\alpha}}\|g\|_{L^\infty}\right)\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\,\alpha}. ∣ J φ , θ , 1 ∣ ≤ c ( κ ∗ , c Γ ) ( ϵ 1 ∂ θ φ L ∞ ∥ g ∥ C 2 0 , α + ϵ 1 ∂ θ φ C 0 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α .
Finally, we bound J φ , s , 2 J_{\varphi,s,2} J φ , s , 2 and J φ , θ , 2 J_{\varphi,\theta,2} J φ , θ , 2 over the region R 2 R_2 R 2 . We now use the second expansion (372) of φ \varphi φ about ( s 0 + s , θ 0 + θ ) (s_0+s,\theta_0+\theta) ( s 0 + s , θ 0 + θ ) to write J φ , s , 2 J_{\varphi,s,2} J φ , s , 2 as
J φ , s , 2 = − ∂ s φ ( s , θ ) ∬ R 2 ( ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( θ 0 + θ ^ 2 ) ) k ( g 0 − g 0 0 ) ∣ R 0 ∣ n − s ^ ℓ + 1 ( ϵ sin ( θ ^ 2 ) ) k ( g − g 0 ) ∣ R ∣ n ) ϵ d θ ^ d s ^ J_{\varphi,s,2}=-\partial_s\varphi(s,\theta)\iint_{R_2}\left(\frac{(s_0+\widehat{s})^{\ell+1}\left(\epsilon\sin\left(\frac{\theta_0+\widehat{\theta}}{2}\right)\right)^k(g_0-g_0^0)}{|R_0|^n}-\frac{\widehat{s}^{\ell+1}\left(\epsilon\sin\left(\frac{\widehat{\theta}}{2}\right)\right)^k(g-g^0)}{|R|^n}\right)\epsilon\,d\widehat{\theta}d\widehat{s} J φ , s , 2 = − ∂ s φ ( s , θ ) ∬ R 2 ∣ R 0 ∣ n ( s 0 + s ) ℓ + 1 ( ϵ sin ( 2 θ 0 + θ ) ) k ( g 0 − g 0 0 ) − ∣ R ∣ n s ℓ + 1 ( ϵ sin ( 2 θ ) ) k ( g − g 0 ) ϵ d θ d s
− ∂ s φ ( s , θ ) ∬ R 2 ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( θ 0 + θ ^ 2 ) ) k ( g 0 0 − g 0 ) ∣ R 0 ∣ n ϵ d θ ^ d s ^ − ∂ s φ ( s , θ ) g 0 ∬ R 2 ( ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( θ 0 + θ ^ 2 ) ) k ( 1 ∣ R 0 ∣ n − 1 ∣ R e v e n , 0 ∣ n ) + s ^ ℓ + 1 ( ϵ sin ( θ ^ 2 ) ) k ( 1 ∣ R e v e n ∣ n − 1 ∣ R ∣ n ) ) ϵ d θ ^ d s ^ + ∣ s 0 ∣ 1 + α Q φ , s ( s , θ , s 0 + s , θ 0 + θ ) ∬ R 2 ( s 0 + s ^ ) ℓ ( ϵ sin ( θ 0 + θ ^ 2 ) ) k g 0 ∣ R 0 ∣ n ϵ d θ ^ d s ^ . \begin{aligned}
-\partial_s\varphi(s,\theta)\iint_{R_2}\frac{(s_0+\hat{s})^{\ell+1}\left(\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right)^k(g_0^0-g^0)}{|R_0|^n}\,\epsilon\,d\hat{\theta}d\hat{s}\\
-\partial_s\varphi(s,\theta)g^0\iint_{R_2}\left((s_0+\hat{s})^{\ell+1}\left(\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right)^k\left(\frac{1}{|R_0|^n}-\frac{1}{|R_{\mathrm{even},0}|^n}\right)\right.\\
\left.+\hat{s}^{\ell+1}\left(\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\right)^k\left(\frac{1}{|R_{\mathrm{even}}|^n}-\frac{1}{|R|^n}\right)\right)\epsilon\,d\hat{\theta}d\hat{s}\\
+|s_0|^{1+\alpha}Q_{\varphi,s}(s,\theta,s_0+s,\theta_0+\theta)\iint_{R_2}\frac{(s_0+\hat{s})^\ell\left(\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right)^kg_0}{|R_0|^n}\,\epsilon\,d\hat{\theta}d\hat{s}.
\end{aligned} − ∂ s φ ( s , θ ) ∬ R 2 ∣ R 0 ∣ n ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( 2 θ 0 + θ ^ ) ) k ( g 0 0 − g 0 ) ϵ d θ ^ d s ^ − ∂ s φ ( s , θ ) g 0 ∬ R 2 ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( 2 θ 0 + θ ^ ) ) k ( ∣ R 0 ∣ n 1 − ∣ R even , 0 ∣ n 1 ) + s ^ ℓ + 1 ( ϵ sin ( 2 θ ^ ) ) k ( ∣ R even ∣ n 1 − ∣ R ∣ n 1 ) ϵ d θ ^ d s ^ + ∣ s 0 ∣ 1 + α Q φ , s ( s , θ , s 0 + s , θ 0 + θ ) ∬ R 2 ∣ R 0 ∣ n ( s 0 + s ^ ) ℓ ( ϵ sin ( 2 θ 0 + θ ^ ) ) k g 0 ϵ d θ ^ d s ^ .
Using (376) to expand the differences 1 ∣ R 0 ∣ n − 1 ∣ R e v e n , 0 ∣ n \frac{1}{|R_0|^n}-\frac{1}{|R_{\mathrm{even},0}|^n} ∣ R 0 ∣ n 1 − ∣ R even , 0 ∣ n 1 and 1 ∣ R e v e n ∣ n − 1 ∣ R ∣ n \frac{1}{|R_{\mathrm{even}}|^n}-\frac{1}{|R|^n} ∣ R even ∣ n 1 − ∣ R ∣ n 1 , we have
∣ J φ , s , 2 ∣ ≤ ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α ∣ ∬ R 2 ( ( s 0 + s ^ ) ℓ + 1 + α ( ϵ sin ( θ 0 + θ ^ 2 ) ) k ∣ R 0 ∣ n − s ^ ℓ + 1 + α ( ϵ sin ( θ ^ 2 ) ) k ∣ R ∣ n ) ϵ d θ ^ d s ^ ∣ + c ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 1 0 , α + ∬ R 2 s 0 2 + ϵ 2 θ 0 2 α + ∣ s 0 + s ^ ∣ ℓ + 1 ∣ ϵ sin ( θ 0 + θ ^ 2 ) ∣ k ∣ R 0 ∣ n ϵ d θ ^ d s ^ + ∥ ∂ s φ ∥ L ∞ ∥ g ∥ L ∞ ∣ ∬ R 2 ( s ^ 5 Q R , 1 + ϵ s ^ 4 Q R , 2 + ϵ 2 s ^ 3 sin ( θ ^ 2 ) Q R , 3 ∣ R e v e n ∣ ∣ R ∣ ( ∣ R ∣ + ∣ R e v e n ∣ ) ∑ j = 0 n − 1 s ^ ℓ + 1 ( ϵ sin ( θ ^ 2 ) ) k ∣ R ∣ j ∣ R e v e n ∣ n − 1 − j − ( s 0 + s ^ ) 5 Q R , 1 + ϵ ( s 0 + s ^ ) 4 Q R , 2 + ϵ 2 ( s 0 + s ^ ) 3 sin ( θ 0 + θ ^ 2 ) Q R , 3 ∣ R e v e n , 0 ∣ ∣ R 0 ∣ ( ∣ R 0 ∣ + ∣ R e v e n , 0 ∣ ) ∑ j = 0 n − 1 ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( θ 0 + θ ^ 2 ) ) k ∣ R 0 ∣ j ∣ R e v e n , 0 ∣ n − 1 − j ) ϵ d θ ^ d s ^ ∣ + c ∣ s 0 ∣ 1 + α ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ∬ R 2 1 ∣ R 0 ∣ 3 ϵ d θ ^ d s ^ ≤ c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α ∬ R 2 s 0 2 + ϵ 2 θ 0 2 α + ∣ s 0 + s ^ ∣ ℓ + 1 ∣ R 0 ∣ 3 − α ϵ d θ ^ d s ^ + ∥ g ∥ C 1 0 , α + ∬ R 2 s 0 2 + ϵ 2 θ 0 2 α + ∣ R 0 ∣ 2 ϵ d θ ^ d s ^ ) + c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ∥ g ∥ L ∞ ( ∬ R 2 s 0 2 + ϵ 2 θ 0 2 ∣ R 0 ∣ 2 ϵ d θ ^ d s ^ + ∬ R 2 s 0 2 + ϵ 2 θ 0 2 α ∣ R 0 ∣ ϵ d θ ^ d s ^ ) + c ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ∬ R 2 ∣ s 0 ∣ 1 + α ∣ R 0 ∣ 3 ϵ d θ ^ d s ^ ≤ c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) ∬ ρ > 4 s 0 2 + ϵ 2 θ 0 2 ( s 0 2 + ϵ 2 θ 0 2 ρ 3 − α + s 0 2 + ϵ 2 θ 0 2 α + ρ 2 ) ρ d ρ d ϕ + c ( κ ∗ ) ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ∬ ρ > 4 s 0 2 + ϵ 2 θ 0 2 ( s 0 2 + ϵ 2 θ 0 2 ρ 2 + s 0 2 + ϵ 2 θ 0 2 α ρ + ∣ s 0 ∣ 1 + α ρ 3 ) ρ d ρ d ϕ ≤ c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) ( s 0 2 + ϵ 2 θ 0 2 ( 1 + s 0 2 + ϵ 2 θ 0 2 − 1 + α ) + s 0 2 + ϵ 2 θ 0 2 α + ( 1 + ∣ log ( s 0 2 + ϵ 2 θ 0 2 ) ∣ ) ) + c ( κ ∗ ) ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ( s 0 2 + ϵ 2 θ 0 2 ( 1 + ∣ log ( s 0 2 + ϵ 2 θ 0 2 ) ∣ ) + s 0 2 + ϵ 2 θ 0 2 α ( 1 + s 0 2 + ϵ 2 θ 0 2 ) + ∣ s 0 ∣ 1 + α ( 1 + s 0 2 + ϵ 2 θ 0 2 − 1 ) ) ≤ c ( κ ∗ ) ( ( ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) + ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α ) . \begin{aligned}
|J_{\varphi,s,2}| \leq{}&\|\partial_s\varphi\|_{L^\infty}\|g\|_{C_2^{0,\alpha}}\left|\iint_{R_2}\left(\frac{(s_0+\hat{s})^{\ell+1+\alpha}\left(\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right)^k}{|R_0|^n}\right.\right.\\
&\left.\left.\qquad-\frac{\hat{s}^{\ell+1+\alpha}\left(\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\right)^k}{|R|^n}\right)\epsilon\,d\hat{\theta}d\hat{s}\right|\\
&+c\|\partial_s\varphi\|_{L^\infty}\|g\|_{C_1^{0,\alpha+}}\iint_{R_2}\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha+}|s_0+\hat{s}|^{\ell+1}\left|\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right|^k}{|R_0|^n}\,\epsilon\,d\hat{\theta}d\hat{s}\\
&+\|\partial_s\varphi\|_{L^\infty}\|g\|_{L^\infty}\left|\iint_{R_2}\left(\frac{\hat{s}^5Q_{R,1}+\epsilon\hat{s}^4Q_{R,2}+\epsilon^2\hat{s}^3\sin\left(\frac{\hat{\theta}}{2}\right)Q_{R,3}}{|R_{\mathrm{even}}||R|(|R|+|R_{\mathrm{even}}|)}\sum_{j=0}^{n-1}\frac{\hat{s}^{\ell+1}\left(\epsilon\sin\left(\frac{\hat{\theta}}{2}\right)\right)^k}{|R|^j|R_{\mathrm{even}}|^{n-1-j}}\right.\right.\\
&\left.\left.\qquad-\frac{(s_0+\hat{s})^5Q_{R,1}+\epsilon(s_0+\hat{s})^4Q_{R,2}+\epsilon^2(s_0+\hat{s})^3\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)Q_{R,3}}{|R_{\mathrm{even},0}||R_0|(|R_0|+|R_{\mathrm{even},0}|)}\sum_{j=0}^{n-1}\frac{(s_0+\hat{s})^{\ell+1}\left(\epsilon\sin\left(\frac{\theta_0+\hat{\theta}}{2}\right)\right)^k}{|R_0|^j|R_{\mathrm{even},0}|^{n-1-j}}\right)\epsilon\,d\hat{\theta}d\hat{s}\right|\\
&+c|s_0|^{1+\alpha}\|\partial_s\varphi\|_{C_0,\alpha}\|g\|_{L^\infty}\iint_{R_2}\frac{1}{|R_0|^3}\,\epsilon\,d\hat{\theta}d\hat{s}\\
&\leq c(\kappa_*)\|\partial_s\varphi\|_{L^\infty}\left(\|g\|_{C_2^{0,\alpha}}\iint_{R_2}\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha+}|s_0+\hat{s}|^{\ell+1}}{|R_0|^{3-\alpha}}\,\epsilon\,d\hat{\theta}d\hat{s}+\|g\|_{C_1^{0,\alpha+}}\iint_{R_2}\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha+}}{|R_0|^2}\,\epsilon\,d\hat{\theta}d\hat{s}\right)\\
&+c(\kappa_*)\|\partial_s\varphi\|_{L^\infty}\|g\|_{L^\infty}\left(\iint_{R_2}\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}}{|R_0|^2}\,\epsilon\,d\hat{\theta}d\hat{s}+\iint_{R_2}\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha}}{|R_0|}\,\epsilon\,d\hat{\theta}d\hat{s}\right)\\
&+c\|\partial_s\varphi\|_{C_0,\alpha}\|g\|_{L^\infty}\iint_{R_2}\frac{|s_0|^{1+\alpha}}{|R_0|^3}\,\epsilon\,d\hat{\theta}d\hat{s}\\
&\leq c(\kappa_*)\|\partial_s\varphi\|_{L^\infty}\left(\|g\|_{C_2^{0,\alpha}}+\|g\|_{C_1^{0,\alpha+}}\right)\iint_{\rho>4\sqrt{s_0^2+\epsilon^2\theta_0^2}}\left(\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}}{\rho^{3-\alpha}}+\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha+}}{\rho^2}\right)\rho\,d\rho d\phi\\
&+c(\kappa_*)\|\partial_s\varphi\|_{C_0,\alpha}\|g\|_{L^\infty}\iint_{\rho>4\sqrt{s_0^2+\epsilon^2\theta_0^2}}\left(\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}}{\rho^2}+\frac{\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha}}{\rho}+\frac{|s_0|^{1+\alpha}}{\rho^3}\right)\rho\,d\rho d\phi\\
&\leq c(\kappa_*)\|\partial_s\varphi\|_{L^\infty}\left(\|g\|_{C_2^{0,\alpha}}+\|g\|_{C_1^{0,\alpha+}}\right)\left(\sqrt{s_0^2+\epsilon^2\theta_0^2}\left(1+\sqrt{s_0^2+\epsilon^2\theta_0^2}^{-1+\alpha}\right)\right.\\
&\left.\qquad+\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha+}\left(1+\left|\log(s_0^2+\epsilon^2\theta_0^2)\right|\right)\right)\\
&+c(\kappa_*)\|\partial_s\varphi\|_{C_0,\alpha}\|g\|_{L^\infty}\left(\sqrt{s_0^2+\epsilon^2\theta_0^2}\left(1+\left|\log(s_0^2+\epsilon^2\theta_0^2)\right|\right)\right.\\
&\left.\qquad+\sqrt{s_0^2+\epsilon^2\theta_0^2}^{\alpha}\left(1+\sqrt{s_0^2+\epsilon^2\theta_0^2}\right)+|s_0|^{1+\alpha}\left(1+\sqrt{s_0^2+\epsilon^2\theta_0^2}^{-1}\right)\right) \\
&\le c(\kappa_*)\left(\left(\|\partial_s\varphi\|_{L^\infty}\left(\|g\|_{C_2^{0,\alpha}}+\|g\|_{C_1^{0,\alpha^+}}\right)+\|\partial_s\varphi\|_{C^{0,\alpha}}\|g\|_{L^\infty}\right)\sqrt{s_0^2+\epsilon^2\theta_0^{2\alpha}}\right).
\end{aligned} ∣ J φ , s , 2 ∣ ≤ ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 2 0 , α ∬ R 2 ∣ R 0 ∣ n ( s 0 + s ^ ) ℓ + 1 + α ( ϵ sin ( 2 θ 0 + θ ^ ) ) k − ∣ R ∣ n s ^ ℓ + 1 + α ( ϵ sin ( 2 θ ^ ) ) k ϵ d θ ^ d s ^ + c ∥ ∂ s φ ∥ L ∞ ∥ g ∥ C 1 0 , α + ∬ R 2 ∣ R 0 ∣ n s 0 2 + ϵ 2 θ 0 2 α + ∣ s 0 + s ^ ∣ ℓ + 1 ϵ sin ( 2 θ 0 + θ ^ ) k ϵ d θ ^ d s ^ + ∥ ∂ s φ ∥ L ∞ ∥ g ∥ L ∞ ∬ R 2 ∣ R even ∣∣ R ∣ ( ∣ R ∣ + ∣ R even ∣ ) s ^ 5 Q R , 1 + ϵ s ^ 4 Q R , 2 + ϵ 2 s ^ 3 sin ( 2 θ ^ ) Q R , 3 j = 0 ∑ n − 1 ∣ R ∣ j ∣ R even ∣ n − 1 − j s ^ ℓ + 1 ( ϵ sin ( 2 θ ^ ) ) k − ∣ R even , 0 ∣∣ R 0 ∣ ( ∣ R 0 ∣ + ∣ R even , 0 ∣ ) ( s 0 + s ^ ) 5 Q R , 1 + ϵ ( s 0 + s ^ ) 4 Q R , 2 + ϵ 2 ( s 0 + s ^ ) 3 sin ( 2 θ 0 + θ ^ ) Q R , 3 j = 0 ∑ n − 1 ∣ R 0 ∣ j ∣ R even , 0 ∣ n − 1 − j ( s 0 + s ^ ) ℓ + 1 ( ϵ sin ( 2 θ 0 + θ ^ ) ) k ϵ d θ ^ d s ^ + c ∣ s 0 ∣ 1 + α ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ∬ R 2 ∣ R 0 ∣ 3 1 ϵ d θ ^ d s ^ ≤ c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α ∬ R 2 ∣ R 0 ∣ 3 − α s 0 2 + ϵ 2 θ 0 2 α + ∣ s 0 + s ^ ∣ ℓ + 1 ϵ d θ ^ d s ^ + ∥ g ∥ C 1 0 , α + ∬ R 2 ∣ R 0 ∣ 2 s 0 2 + ϵ 2 θ 0 2 α + ϵ d θ ^ d s ^ ) + c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ∥ g ∥ L ∞ ( ∬ R 2 ∣ R 0 ∣ 2 s 0 2 + ϵ 2 θ 0 2 ϵ d θ ^ d s ^ + ∬ R 2 ∣ R 0 ∣ s 0 2 + ϵ 2 θ 0 2 α ϵ d θ ^ d s ^ ) + c ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ∬ R 2 ∣ R 0 ∣ 3 ∣ s 0 ∣ 1 + α ϵ d θ ^ d s ^ ≤ c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) ∬ ρ > 4 s 0 2 + ϵ 2 θ 0 2 ( ρ 3 − α s 0 2 + ϵ 2 θ 0 2 + ρ 2 s 0 2 + ϵ 2 θ 0 2 α + ) ρ d ρ d ϕ + c ( κ ∗ ) ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ∬ ρ > 4 s 0 2 + ϵ 2 θ 0 2 ( ρ 2 s 0 2 + ϵ 2 θ 0 2 + ρ s 0 2 + ϵ 2 θ 0 2 α + ρ 3 ∣ s 0 ∣ 1 + α ) ρ d ρ d ϕ ≤ c ( κ ∗ ) ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) ( s 0 2 + ϵ 2 θ 0 2 ( 1 + s 0 2 + ϵ 2 θ 0 2 − 1 + α ) + s 0 2 + ϵ 2 θ 0 2 α + ( 1 + log ( s 0 2 + ϵ 2 θ 0 2 ) ) ) + c ( κ ∗ ) ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ( s 0 2 + ϵ 2 θ 0 2 ( 1 + log ( s 0 2 + ϵ 2 θ 0 2 ) ) + s 0 2 + ϵ 2 θ 0 2 α ( 1 + s 0 2 + ϵ 2 θ 0 2 ) + ∣ s 0 ∣ 1 + α ( 1 + s 0 2 + ϵ 2 θ 0 2 − 1 ) ) ≤ c ( κ ∗ ) ( ( ∥ ∂ s φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) + ∥ ∂ s φ ∥ C 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α ) .
Here in the second inequality we have used the closeness of the kernels with respect to s 0 s_0 s 0 and ϵ θ 0 \epsilon\theta_0 ϵ θ 0 , and in the third inequality we have again switched to polar coordinates. In the fourth inequality we use that α + > α \alpha^+>\alpha α + > α to absorb the logarithmic term in s 0 2 + ϵ 2 θ 0 2 α + ∣ log ( s 0 2 + ϵ 2 θ 0 2 ) ∣ \sqrt{s_0^2+\epsilon^2\theta_0^{2\alpha^+}}\left|\log(s_0^2+\epsilon^2\theta_0^2)\right| s 0 2 + ϵ 2 θ 0 2 α + log ( s 0 2 + ϵ 2 θ 0 2 ) .
By an analogous series of estimates, it may be shown that J φ , θ , 2 J_{\varphi,\theta,2} J φ , θ , 2 satisfies
c ( κ ∗ ) ( ∥ 1 ϵ ∂ θ φ ∥ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α . c(\kappa_*)\left(\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{L^\infty}\left(\|g\|_{C_2^{0,\alpha}}+\|g\|_{C_1^{0,\alpha^+}}\right)+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C^{0,\alpha}}\|g\|_{L^\infty}\right)\sqrt{s_0^2+\epsilon^2\theta_0^{2\alpha}}. c ( κ ∗ ) ( ϵ 1 ∂ θ φ L ∞ ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) + ϵ 1 ∂ θ φ C 0 , α ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α .
Combining the estimates from cases 1, 2a, and 2b, we have
∣ I φ ^ ∣ ≤ c ( κ ∗ , c Γ ) ( ( ∥ ∂ s φ ∥ L ∞ + ∥ 1 ϵ ∂ θ φ ∥ L ∞ ) ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) + ( ∥ ∂ s φ ∥ C 0 , α + ∥ 1 ϵ ∂ θ φ ∥ C 0 , α ) ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α . \left|\widehat{I_\varphi}\right|\le c(\kappa_*,c_\Gamma)\left(\left(\|\partial_s\varphi\|_{L^\infty}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{L^\infty}\right)\left(\|g\|_{C_2^{0,\alpha}}+\|g\|_{C_1^{0,\alpha^+}}\right)+\left(\|\partial_s\varphi\|_{C^{0,\alpha}}+\left\|\frac{1}{\epsilon}\partial_\theta\varphi\right\|_{C^{0,\alpha}}\right)\|g\|_{L^\infty}\right)\sqrt{s_0^2+\epsilon^2\theta_0^{2\alpha}}. I φ ≤ c ( κ ∗ , c Γ ) ( ( ∥ ∂ s φ ∥ L ∞ + ϵ 1 ∂ θ φ L ∞ ) ( ∥ g ∥ C 2 0 , α + ∥ g ∥ C 1 0 , α + ) + ( ∥ ∂ s φ ∥ C 0 , α + ϵ 1 ∂ θ φ C 0 , α ) ∥ g ∥ L ∞ ) s 0 2 + ϵ 2 θ 0 2 α .
Acknowledgments. LO acknowledges support from NSF grant DMS-2406003 and from the Wisconsin Alumni Research Foundation.
Department of Mathematics, University of Wisconsin - Madison, Madison, WI 53706
Email address: lohm2@wisc.edu