Introduction
Since its development in 1887 by the German physician Julius Petri [1 ] for the facilitation of cell culturing, extending the bacterial culture methods pioneered by his mentor Robert Koch [2 ] , the Petri dish has become an integral part of any biology laboratory. While still primarily used for culturing cells, providing storage space whilst reducing the risk of contamination, its simplicity and functionality allows it to be used in a wide range of other contexts: in chemistry to dry out precipitates and evaporate solvents (e.g. when studying Liesegang rings [3 , 4 ] ) or in entomology where they are convenient enclosures to study the behaviour of insects and small animals [5 , 6 ] . A Petri dish environment is also a simple and common setting in which to examine the locomotion of swimming organisms, particularly those whose body size is tens of microns to millimetres [7 , 8 , 9 , 10 , 11 ] . The boundary condition at the bottom surface of such a container can be approximated as no-slip, while the top of the fluid is stress-free. Hence, a general question is: how does confinement in a Petri dish alter the nature of the flow induced by motile organisms?
The framework to answer this question lies of course with Green’s functions. In low Reynolds number fluid mechanics governed by the Stokes equations [12 ] , the most important such function corresponds to the flow induced by a point force in an unbounded fluid and decays as 1 / r 1/r 1/ r . First written down by Lorentz [13 ] and later denoted a Stokeslet [14 ] , it has been used to solve a wide range of fluid dynamical problems (see Happel and Brenner [15 ] and Kim and Karrila [16 ] for general overviews). One powerful extension to the Stokeslet involves a multipole expansion similar to that in electrostatics. The fluid flow caused by the motion of an arbitrary rigid body through a viscous fluid can be represented as that from a collection of point forces at the surface of the body [16 ] . Expanding the Stokeslet produced at an arbitrary point on the body’s surface as a Taylor series about the center of the body and then summing these contributions in the far field, one obtains a perturbation expansion for the fluid flow induced by the body [17 ] . Regardless of the particular shape of the particle, the fluid velocity field will exhibit common features. The leading order 1 / r 1/r 1/ r term is still a Stokeslet, but at higher orders, one finds distinct singularities. In particular the 1 / r 2 1/r^{2} 1/ r 2 term, denoted a force dipole, can be separated into a symmetric part, denoted a stresslet [18 ] , that corresponds to a symmetric hydrodynamic stress applied locally to the fluid, and an anti-symmetric part, denoted a rotlet [19 ] (called a couplet by Batchelor [18 ] ), corresponding to a local hydrodynamic torque that produces rotational motion.
A well chosen distribution of such Stokes singularities that exploits the inherent symmetries of the system in question can be used to solve Stokes equations in a wide range of geometries and biological contexts [16 ] . Figure 1 illustrates the breadth of this approach, giving examples of biological flows associated with each of the low order Stokes singularities. Although classically in biological fluid dynamics the stresslet is the most common Stokes singularity considered [20 ] , one sees that all low order Stokes singularities arise in familiar contexts.
Figure 1. Stokes singularities in biological fluid mechanics. [a-g] Elementary singularities. Stokeslet flow is found in (a) far-field flow around Volvox carteri [21 ] . Source flows arise from injection of fluid from a micropipette into a Petri dish in studies of (b) dinoflagellates [22 ] and (c) plant-animal worms [23 ] . Rotlet flows arise from (d) the bacterium Escherichia coli under confinement, generating flow field in (e) [24 ] , (f) a magnetic nano stir bar [25 ] , and (g) a macroscopic stirrer [26 ] . [h-l] Composite singularities. Stresslets arise from (h) the pusher E. coli [9 ] , (i) the puller alga Chlamydomonas reinhardtii [21 ] , and (j) a phoretic Janus particle that changes from pusher to puller as a function of its aspect ratio [27 ] . A rotlet dipole flow is induced by (k) a circular mill of Symsagittifera roscoffensis [28 ] . A source is found in (l) the near-field flow induced by Volvox carteri after the Stokeslet contribution is subtracted [21 ] .
The key question addressed here is thus: what is the fluid flow resulting from any Stokes singularity placed in a fluid layer between a rigid lower no-slip boundary and an upper stress-free surface. Although a few cases have been investigated in the literature, there has not been a systematic breakdown of the possible cases that arise. This was first considered by Liron and Mochon [29 ] who derived an exact solution in integral form for a Stokeslet. Subsequent work on this problem includes a theoretical study of bacterial swarms on agar [30 ] , which contained a calculation of the leading order far field contribution to the flow from both a Stokeslet and a Rotlet when placed in a Petri dish configuration. This was further developed by Mathijssen, et. al. [31 ] , who derived a numerically tractable approximation for the flow field produced by a Stokeslet and hence the flow field produced by a force- and torque-free micro-swimmer in a Petri dish.
Table 1. Location of results for various singularities.
In this paper, paying particular attention to the far-field behavior, we systematically extend and generalize these works beyond Stokeslets by computing exact expressions for the flow components u j u_j u j generated in a Petri dish of height H H H by the biologically relevant low-order primary and composite singularities of Stokes flow:
The Stokeslet: u j k = λ F ( δ j k / r + x j x k / r 3 ) u_j^k = \lambda_F \left( \delta_{jk}/r + x_j x_k/r^3 \right) u j k = λ F ( δ j k / r + x j x k / r 3 ) , (1)
Rotlet: u j k = λ R ε j k p x p / r 3 u_j^k = \lambda_R \varepsilon_{jkp}x_p/r^3 u j k = λ R ε j k p x p / r 3 , (2)
Source: u j = λ S x j / r 3 u_j = \lambda_S x_j/r^3 u j = λ S x j / r 3 , (3)
General stresslet: u j k , l = λ C x j x k x l / r 5 u_j^{k,l} = \lambda_C x_j x_k x_l/r^5 u j k , l = λ C x j x k x l / r 5 , (4)
Rotlet dipole: u j k = λ R D ε j p k x k x p / r 5 u_j^k = \lambda_{RD}\varepsilon_{jpk}x_k x_p/r^5 u j k = λ R D ε j p k x k x p / r 5 , (5)
Source dipole: u j = λ S D ( δ j k / r 3 − 3 x j x k / r 5 ) u_j = \lambda_{SD}\left(\delta_{jk}/r^3 - 3x_jx_k/r^5\right) u j = λ S D ( δ j k / r 3 − 3 x j x k / r 5 ) . (6)
Note that here, j j j , k k k and l l l are free indices while the λ i \lambda_i λ i are dimensional constants denoting the strength of the singularities, with dimensions m 2 s − 1 \mathrm{m}^2\mathrm{s}^{-1} m 2 s − 1 for the Stokeslet, m 3 s − 1 \mathrm{m}^3\mathrm{s}^{-1} m 3 s − 1 for the rotlet, source and stresslet and m 4 s − 1 \mathrm{m}^4\mathrm{s}^{-1} m 4 s − 1 for the rotlet dipole and source dipole. For clarity, we only present in the main text analysis for a source and a Stokeslet, namely the simplest and the most common singularity respectively. The results for the other singularities are given in Appendices B-E. Table I lists the locations of all these results in the paper. We adopt the geometry of Figure 2 , with in-plane coordinates ( x 1 , x 2 ) (x_1,x_2) ( x 1 , x 2 ) , the no-slip surface at x 3 = 0 x_3 = 0 x 3 = 0 and the stress-free surface at x 3 = H x_3 = H x 3 = H .
Figure 2. Stokes singularity in a Petri dish. The positive singularity is located at z = s z=s z = s and labelled 1. Its reflection across the no-stress surface at z = H z=H z = H is labelled 2 and across the no-slip surface at z = 0 z=0 z = 0 is 2′, and so on. An alternate approach uses the full solution for a single no-slip surface and extends the domain to include a no-slip surface at z = 2 H z=2H z = 2 H .
In §III, we calculate for both a source and a Stokeslet a particular solution to the Stokes equations generated by summing the infinite image system of Stokes singularities that is formed by repeatedly reflecting the initial singularity in both of the vertical boundaries. Then in §IV, an auxiliary solution is calculated using a Fourier transform method so that the sum of the two solutions is an exact solution for the full boundary conditions. In §V, a contour integral approach is used to calculate the leading order term of the fluid velocity in the far-field of a source.
This methodology, applied to both the source and the Stokeslet in §IV-V, is applied to the rest of the most commonly used Stokes singularities, (namely a rotlet, a general stresslet, a rotlet dipole and a source dipole), in Appendices B–E. Finally, as an application of these results, §VIII reconsiders in the geometry of the Petri dish the problem of hydrodynamic bound states, first discovered using the green alga Volvox near a no-slip surface [32 ] and later rediscovered in multiple contexts. The concluding §VI summarises the main results of the paper.
In particular, we note that higher order in-plane Stokes singularities can be found by differentiating the solutions with respect to a horizontal coordinate x α x_\alpha x α . Since all other Stokes singularities can be expressed in terms of derivatives of these singularities, we conclude that the leading order contribution to the fluid velocity in the far field for an arbitrary Stokes singularity is separable in x 3 x_3 x 3 , either decaying exponentially radially or having x 3 x_3 x 3 dependence of the form x 3 ( 1 − x 3 / 2 H ) x_3(1-x_3/2H) x 3 ( 1 − x 3 /2 H ) . Hence, for many situations where the forcing can be modelled as a sum of Stokes singularities, the depth-averaged fluid flow can be captured by an associated Brinkman equation with a screening length proportional to H H H .
Singularity in a Petri dish
Consider, as in Figure 2 , a Stokes singularity f f f , located at the point ( x 1 , x 2 , x 3 ) = ( 0 , 0 , s ) (x_1,x_2,x_3)=(0,0,s) ( x 1 , x 2 , x 3 ) = ( 0 , 0 , s ) between a rigid lower surface at x 3 = 0 x_3=0 x 3 = 0 and an upper free surface at x 3 = H x_3=H x 3 = H , which generates a fluid flow u = ( u 1 , u 2 , u 3 ) \boldsymbol{u}=(u_1,u_2,u_3) u = ( u 1 , u 2 , u 3 ) . At x 3 = 0 x_3=0 x 3 = 0 , we impose the no-slip boundary conditions
u 1 = u 2 = u 3 = 0. (1) u_1=u_2=u_3=0. \tag*{(1)} u 1 = u 2 = u 3 = 0. (1)
The capillary length λ c a p \lambda_{\mathrm{cap}} λ cap for a water-air interface is λ c a p = γ w / ρ w g ≈ 2.73 m m \lambda_{\mathrm{cap}}=\sqrt{\gamma_w/\rho_w g}\approx2.73\ \mathrm{mm} λ cap = γ w / ρ w g ≈ 2.73 mm , where ρ w = 997 k g m − 3 \rho_w=997\ \mathrm{kg}\,\mathrm{m}^{-3} ρ w = 997 kg m − 3 is the density of water, γ w \gamma_w γ w as 72.8 m N m − 1 72.8\ \mathrm{mN}\,\mathrm{m}^{-1} 72.8 mN m − 1 is the air-water surface tension, and g = 9.81 m s − 2 g=9.81\ \mathrm{m}\,\mathrm{s}^{-2} g = 9.81 m s − 2 is the gravitational acceleration. Since in a Petri dish λ c a p \lambda_{\mathrm{cap}} λ cap and H H H are similar in size, at the free surface, surface tension and gravitational effects are of similar magnitudes. Together, they restrict the vertical deformation of the interface. Hence, we assume the limit of no deformation in the vertical direction, fixing H H H as a constant. The self-consistency of this assumption is explored later in §VI. The dynamic boundary condition u 3 = D H / D t u_3=DH/Dt u 3 = D H / D t thus simplifies to
u 3 = 0 at x 3 = H . (2) u_3=0 \quad\text{at } x_3=H. \tag*{(2)} u 3 = 0 at x 3 = H . (2)
A force balance at x 3 = H x_3 = H x 3 = H , σ x 1 x 3 = σ x 2 x 3 = 0 \sigma_{x_1x_3} = \sigma_{x_2x_3} = 0 σ x 1 x 3 = σ x 2 x 3 = 0 , implies
∂ u 1 ∂ x 3 = ∂ u 2 ∂ x 3 = 0 at x 3 = H . (3) \frac{\partial u_1}{\partial x_3} = \frac{\partial u_2}{\partial x_3} = 0 \quad\text{at } x_3 = H. \tag*{(3)} ∂ x 3 ∂ u 1 = ∂ x 3 ∂ u 2 = 0 at x 3 = H . (3)
We nondimensionalize this system, scaling lengths with H H H and velocities with U S U_S U S , where for a singularity of strength λ S \lambda_S λ S that decays in the far field like 1 / r n 1/r^n 1/ r n , U S = λ S H − n U_S = \lambda_S H^{-n} U S = λ S H − n . For notational simplicity, we define
( x , y , z ) = ( x 1 , x 2 , x 3 ) / H , (4) (x,y,z) = (x_1,x_2,x_3)/H, \tag*{(4)} ( x , y , z ) = ( x 1 , x 2 , x 3 ) / H , (4)
( u x , u y , u z ) = ( u 1 , u 2 , u 3 ) / U S , (5) (u_x,u_y,u_z) = (u_1,u_2,u_3)/U_S, \tag*{(5)} ( u x , u y , u z ) = ( u 1 , u 2 , u 3 ) / U S , (5)
h = s / H . (6) h = s/H. \tag*{(6)} h = s / H . (6)
The boundary conditions become
u x = u y = u z = 0 at z = 0 , (7) u_x = u_y = u_z = 0 \quad\text{at } z = 0, \tag*{(7)} u x = u y = u z = 0 at z = 0 , (7)
∂ u x ∂ z = ∂ u y ∂ z = u z = 0 at z = 1. (8) \frac{\partial u_x}{\partial z} = \frac{\partial u_y}{\partial z} = u_z = 0 \quad\text{at } z = 1. \tag*{(8)} ∂ z ∂ u x = ∂ z ∂ u y = u z = 0 at z = 1. (8)
Repeated Reflection Solution
We first examine the extent to which we can satisfy these boundary conditions through a distribution of image singularities. Following the canonical approach of Liron and Mochon [29 ] , for a singularity placed at x 3 = s x_3 = s x 3 = s [the green + labelled 1 in Figure 2 ], placing an image singularity of the same sign at x 3 = 2 H − s x_3 = 2H - s x 3 = 2 H − s (label 2) satisfies the free surface boundary condition at x 3 = H x_3 = H x 3 = H . Similarly, placing an image singularity of the opposite sign at x 3 = − s x_3 = -s x 3 = − s (2′) partially satisfies the no-slip boundary condition, but singularity 2 fails the no-slip boundary condition and thus must be reflected about x 3 = 0 x_3 = 0 x 3 = 0 , changing its sign at location 3, Similarly, singularity 2′ fails the free surface boundary condition and thus must be reflected in x 3 = H x_3 = H x 3 = H to give singularity 3′. Repeating this ad infinitum , namely inverting the sign when reflecting in the no-slip x 3 = 0 x_3 = 0 x 3 = 0 boundary and keeping the same sign when reflecting in the free surface x 3 = H x_3 = H x 3 = H boundary, gives an infinite series of singularities that constitutes the repeated reflection solution for that singularity.
In rescaled units, if we define the singularity locations r 1 n = ( x , y , z − h + 4 n ) \mathbf{r}_{1n} = (x,y,z-h+4n) r 1 n = ( x , y , z − h + 4 n ) , r 2 n = ( x , y , z − h + ( 4 n + 2 ) ) \mathbf{r}_{2n} = (x,y,z-h+(4n+2)) r 2 n = ( x , y , z − h + ( 4 n + 2 )) , R 1 n = ( x , y , z + h + 4 n ) \mathbf{R}_{1n} = (x,y,z+h+4n) R 1 n = ( x , y , z + h + 4 n ) , and R 2 n = ( x , y , z + h + ( 4 n + 2 ) ) \mathbf{R}_{2n} = (x,y,z+h+(4n+2)) R 2 n = ( x , y , z + h + ( 4 n + 2 )) , then the repeated reflection solution is but one case of the general function L ( f ) \mathcal{L}(f) L ( f ) for an arbitrary function f f f ,
L = ∑ n = − ∞ ∞ { f ( r 1 n ) − f ( r 2 n ) − f ( R 1 n ) + f ( R 2 n ) } . (9) \mathcal{L} = \sum_{n=-\infty}^{\infty} \{f(\mathbf{r}_{1n}) - f(\mathbf{r}_{2n}) - f(\mathbf{R}_{1n}) + f(\mathbf{R}_{2n})\}. \tag*{(9)} L = n = − ∞ ∑ ∞ { f ( r 1 n ) − f ( r 2 n ) − f ( R 1 n ) + f ( R 2 n )} . (9)
While intuitive, this series expansion is unwieldy. For the particular case f = 1 / r f = 1/r f = 1/ r , a Bessel function identity can be used to obtain the integral form
L ( 1 r ) = ∫ 0 ∞ d λ 2 J 0 ( λ ρ ) cosh ( λ ) { sinh h λ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , (10) \mathcal{L}\left(\frac{1}{r}\right) = \int_{0}^{\infty} d\lambda\frac{2J_0(\lambda\rho)}{\cosh(\lambda)}
\begin{cases}
\sinh h\lambda\cosh(1-z)\lambda,\\
\sinh z\lambda\cosh(1-h)\lambda,
\end{cases} \tag*{(10)} L ( r 1 ) = ∫ 0 ∞ d λ cosh ( λ ) 2 J 0 ( λ ρ ) { sinh hλ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , (10)
where ρ = x 2 + y 2 \rho= \sqrt{x^2+y^2} ρ = x 2 + y 2 and here and below the upper expression holds for z > h z > h z > h and the lower for z < h z < h z < h . Higher order solutions are obtained from this result through algebraic manipulation, as shown in Appendix A for the third and fifth order cases. From those results, we find the repeated reflection solution v j v_j v j for a source x j / r 3 x_j/r^3 x j / r 3 ,
v j = δ j α x α L ( 1 r 3 ) + δ j 3 L ( z r 3 ) = 2 x α ρ δ j α ∫ 0 ∞ λ d λ J 1 ( λ ρ ) cosh λ { sinh h λ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , + 2 δ j 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ { sinh h λ sinh ( 1 − z ) λ , − cosh z λ cosh ( 1 − h ) λ . (11) \begin{aligned}
v_j&=\delta_{j\alpha}x_\alpha\mathcal{L}\left(\frac{1}{r^3}\right)+\delta_{j3}\mathcal{L}\left(\frac{z}{r^3}\right)\\
&=\frac{2x_\alpha}{\rho}\delta_{j\alpha}\int_0^\infty\lambda\,\mathrm{d}\lambda\frac{J_1(\lambda\rho)}{\cosh\lambda}
\begin{cases}
\sinh h\lambda\cosh(1-z)\lambda,\\
\sinh z\lambda\cosh(1-h)\lambda,
\end{cases}\\
&\quad+2\delta_{j3}\int_0^\infty\lambda\,\mathrm{d}\lambda\frac{J_0(\lambda\rho)}{\cosh\lambda}
\begin{cases}
\sinh h\lambda\sinh(1-z)\lambda,\\
-\cosh z\lambda\cosh(1-h)\lambda.
\end{cases} \tag*{(11)}
\end{aligned} v j = δ j α x α L ( r 3 1 ) + δ j 3 L ( r 3 z ) = ρ 2 x α δ j α ∫ 0 ∞ λ d λ cosh λ J 1 ( λ ρ ) { sinh hλ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , + 2 δ j 3 ∫ 0 ∞ λ d λ cosh λ J 0 ( λ ρ ) { sinh hλ sinh ( 1 − z ) λ , − cosh z λ cosh ( 1 − h ) λ . (11)
Similarly, for a Stokeslet δ j k / r + x j x k / r 3 \delta_{jk}/r+x_jx_k/r^3 δ j k / r + x j x k / r 3 , we find
v j k = δ j k L ( 1 r ) + δ j α δ k β x α x β L ( 1 r 3 ) + ( δ j α δ k 3 + δ k α δ j 3 ) x α L ( z r 3 ) + δ j 3 δ k 3 L ( z 2 r 3 ) = 2 ( δ j k + δ j 3 δ k 3 ) ∫ 0 ∞ d λ J 0 ( λ ρ ) cosh λ { sinh h λ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , + 2 ( x α x β ρ δ j α δ k β − ρ δ j 3 δ k 3 ) ∫ 0 ∞ λ d λ J 1 ( λ ρ ) cosh λ × { sinh h λ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , + 2 x α ( δ j 3 δ k α + δ k 3 δ j α ) ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ × { sinh h λ sinh ( 1 − z ) λ , − cosh z λ cosh ( 1 − h ) λ . (12) \begin{aligned}
v_j^k&=\delta_{jk}\mathcal{L}\left(\frac{1}{r}\right)+\delta_{j\alpha}\delta_{k\beta}x_\alpha x_\beta\mathcal{L}\left(\frac{1}{r^3}\right)+(\delta_{j\alpha}\delta_{k3}+\delta_{k\alpha}\delta_{j3})x_\alpha\mathcal{L}\left(\frac{z}{r^3}\right)+\delta_{j3}\delta_{k3}\mathcal{L}\left(\frac{z^2}{r^3}\right)\\
&=2(\delta_{jk}+\delta_{j3}\delta_{k3})\int_0^\infty\mathrm{d}\lambda\frac{J_0(\lambda\rho)}{\cosh\lambda}
\begin{cases}
\sinh h\lambda\cosh(1-z)\lambda,\\
\sinh z\lambda\cosh(1-h)\lambda,
\end{cases}\\
&\quad+2\left(\frac{x_\alpha x_\beta}{\rho}\delta_{j\alpha}\delta_{k\beta}-\rho\delta_{j3}\delta_{k3}\right)\int_0^\infty\lambda\,\mathrm{d}\lambda\frac{J_1(\lambda\rho)}{\cosh\lambda}\times
\begin{cases}
\sinh h\lambda\cosh(1-z)\lambda,\\
\sinh z\lambda\cosh(1-h)\lambda,
\end{cases}\\
&\quad+2x_\alpha(\delta_{j3}\delta_{k\alpha}+\delta_{k3}\delta_{j\alpha})\int_0^\infty\lambda\,\mathrm{d}\lambda\frac{J_0(\lambda\rho)}{\cosh\lambda}\times
\begin{cases}
\sinh h\lambda\sinh(1-z)\lambda,\\
-\cosh z\lambda\cosh(1-h)\lambda.
\end{cases} \tag*{(12)}
\end{aligned} v j k = δ j k L ( r 1 ) + δ j α δ k β x α x β L ( r 3 1 ) + ( δ j α δ k 3 + δ k α δ j 3 ) x α L ( r 3 z ) + δ j 3 δ k 3 L ( r 3 z 2 ) = 2 ( δ j k + δ j 3 δ k 3 ) ∫ 0 ∞ d λ cosh λ J 0 ( λ ρ ) { sinh hλ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , + 2 ( ρ x α x β δ j α δ k β − ρ δ j 3 δ k 3 ) ∫ 0 ∞ λ d λ cosh λ J 1 ( λ ρ ) × { sinh hλ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , + 2 x α ( δ j 3 δ k α + δ k 3 δ j α ) ∫ 0 ∞ λ d λ cosh λ J 0 ( λ ρ ) × { sinh hλ sinh ( 1 − z ) λ , − cosh z λ cosh ( 1 − h ) λ . (12)
Similar expressions can be constructed for the other commonly used Stokes singularities (see Appendix § 1.11 for the rotlet, Appendix § 1.12 for the stresslet, Appendix § 1.13 for the rotlet dipole, and Appendix E for the source dipole).
These results obtained via the repeated reflection solution can also be found directly from Liron’s solution [29 ] for a point force between two no-slip walls by setting the separation in that calculation to be 2 H 2H 2 H , placing a second force at 2 H − s 2H-s 2 H − s and observing that the reflection symmetry of the problem about the midline at x 3 = H x_3=H x 3 = H guarantees a stress-free condition at the midline.
Due to the nature of the algebraic manipulations performed above, these integral expressions do not converge when in the horizontal plane of the singularity x 3 = s x_3=s x 3 = s . Instead, it transpires that the correct integral expression to use instead is ( v j k ∣ x 3 → s + + v j k ∣ x 3 → s − ) / 2 \left(v_j^k\big|_{x_3\to s^+}+v_j^k\big|_{x_3\to s^-}\right)/2 ( v j k x 3 → s + + v j k x 3 → s − ) /2 , the average of the integrals as x 3 x_3 x 3 tends to s s s from both directions.
Auxiliary Solution
In a scalar problem, such as a set of electric charges, the repeated reflection solution would solve the full system. However, our singularities are vectors and thus the repeated reflection solution does not satisfy all the boundary conditions. If we write the full fluid velocity field u j k u_j^k u j k as u j k = v j k + w j k u_j^k=v_j^k+w_j^k u j k = v j k + w j k , then the auxiliary solution w j k w_j^k w j k satisfies
μ ∇ 2 w j k = ∂ q ∂ x j , ∂ w j ∂ x j = 0 ⟶ ∇ 2 q = 0 , (13) \mu\nabla^2w_j^k=\frac{\partial q}{\partial x_j},\qquad\frac{\partial w_j}{\partial x_j}=0\quad\longrightarrow\quad\nabla^2q=0, \tag*{(13)} μ ∇ 2 w j k = ∂ x j ∂ q , ∂ x j ∂ w j = 0 ⟶ ∇ 2 q = 0 , (13)
for suitable effective pressure q q q , with boundary conditions
w j ∣ z = 0 = − v j ∣ z = 0 , w 3 ∣ z = 1 = − v 3 ∣ z = 1 , ∂ w α ∂ z ∣ z = 1 = − ∂ v α ∂ z ∣ z = 1 , (14) \left.w_j\right|_{z=0}=-\left.v_j\right|_{z=0},\qquad
\left.w_3\right|_{z=1}=-\left.v_3\right|_{z=1},\qquad
\left.\frac{\partial w_\alpha}{\partial z}\right|_{z=1}=-\left.\frac{\partial v_\alpha}{\partial z}\right|_{z=1}, \tag*{(14)} w j ∣ z = 0 = − v j ∣ z = 0 , w 3 ∣ z = 1 = − v 3 ∣ z = 1 , ∂ z ∂ w α z = 1 = − ∂ z ∂ v α z = 1 , (14)
where α ∈ [ 1 , 2 ] \alpha\in[1,2] α ∈ [ 1 , 2 ] and j ∈ [ 1 , 3 ] j \in[1,3] j ∈ [ 1 , 3 ] . For a source these are
w α ∣ z = 0 = w 3 ∣ z = 1 = ∂ w α ∂ z ∣ z = 1 = 0 , w 3 ∣ z = 0 = 2 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ cosh ( 1 − h ) λ (15) w_{\alpha}\big|_{z=0}=w_{3}\big|_{z=1}=\left.\frac{\partial w_{\alpha}}{\partial z}\right|_{z=1}=0,\qquad w_{3}\big|_{z=0}=2\int_{0}^{\infty}\lambda\,\mathrm{d}\lambda\frac{J_{0}(\lambda\rho)}{\cosh\lambda}\cosh(1-h)\lambda \tag*{(15)} w α z = 0 = w 3 z = 1 = ∂ z ∂ w α z = 1 = 0 , w 3 z = 0 = 2 ∫ 0 ∞ λ d λ cosh λ J 0 ( λ ρ ) cosh ( 1 − h ) λ (15)
Similarly for a Stokeslet, applying standard Bessel function identities, the auxiliary boundary conditions become
w j k ∣ z = 0 = 2 x α ( δ j 3 δ k α + δ k 3 δ j α ) ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh ( 1 − h ) λ cosh λ , (16) w_{j}^{k}\big|_{z=0}=2x_{\alpha}\left(\delta_{j3}\delta_{k\alpha}+\delta_{k3}\delta_{j\alpha}\right)\int_{0}^{\infty}\lambda\,\mathrm{d}\lambda J_{0}(\lambda\rho)\frac{\cosh(1-h)\lambda}{\cosh\lambda}, \tag*{(16)} w j k z = 0 = 2 x α ( δ j 3 δ k α + δ k 3 δ j α ) ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ cosh ( 1 − h ) λ , (16)
∂ w α k ∂ z ∣ z = H = 2 x α δ k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) λ sinh h λ cosh λ , (17) \left.\frac{\partial w_{\alpha}^{k}}{\partial z}\right|_{z=H}=2x_{\alpha}\delta_{k3}\int_{0}^{\infty}\lambda\,\mathrm{d}\lambda J_{0}(\lambda\rho)\frac{\lambda\sinh h\lambda}{\cosh\lambda}, \tag*{(17)} ∂ z ∂ w α k z = H = 2 x α δ k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ λ sinh hλ , (17)
w 3 k ∣ z = H = 2 δ k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) ( ∂ ∂ λ ( sinh h λ cosh λ ) − sinh h λ λ cosh λ ) . (18) w_{3}^{k}\big|_{z=H}=2\delta_{k3}\int_{0}^{\infty}\lambda\,\mathrm{d}\lambda J_{0}(\lambda\rho)\left(\frac{\partial}{\partial\lambda}\left(\frac{\sinh h\lambda}{\cosh\lambda}\right)-\frac{\sinh h\lambda}{\lambda\cosh\lambda}\right). \tag*{(18)} w 3 k z = H = 2 δ k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) ( ∂ λ ∂ ( cosh λ sinh hλ ) − λ cosh λ sinh hλ ) . (18)
We solve for w j w_j w j by taking the two dimensional Fourier transform of this system with respect to ( x , y ) (x,y) ( x , y ) , (namely w j ( x , y , z ) ⟹ w ^ j ( k 1 , k 2 , z ) w_j(x,y,z)\Longrightarrow\widehat{w}_j(k_1,k_2,z) w j ( x , y , z ) ⟹ w j ( k 1 , k 2 , z ) ), to arrive at
μ ( ∂ 2 w ^ j ∂ z 2 − k 2 w ^ j ) = δ j 3 ∂ q ^ ∂ z + i δ α j k α q ^ , (19) \mu\left(\frac{\partial^2\widehat{w}_j}{\partial z^2}-k^2\widehat{w}_j\right)=\delta_{j3}\frac{\partial\widehat{q}}{\partial z}+i\delta_{\alpha j}k_{\alpha}\widehat{q}, \tag*{(19)} μ ( ∂ z 2 ∂ 2 w j − k 2 w j ) = δ j 3 ∂ z ∂ q + i δ α j k α q , (19)
∂ w ^ 3 ∂ z + i k α w ^ α = 0 , (20) \frac{\partial\widehat{w}_3}{\partial z}+ik_{\alpha}\widehat{w}_{\alpha}=0, \tag*{(20)} ∂ z ∂ w 3 + i k α w α = 0 , (20)
∂ 2 q ^ ∂ z 2 − k 2 q ^ = 0 , (21) \frac{\partial^2\widehat{q}}{\partial z^2}-k^2\widehat{q}=0, \tag*{(21)} ∂ z 2 ∂ 2 q − k 2 q = 0 , (21)
where α ∈ [ 1 , 2 ] \alpha\in[1,2] α ∈ [ 1 , 2 ] and k 2 = k 1 2 + k 2 2 k^2=k_1^2+k_2^2 k 2 = k 1 2 + k 2 2 . From inspection, this has the general solution
q ^ = B ( k ) sinh k ( 1 − z ) + C ( k ) cosh k ( 1 − z ) , (22) \widehat{q}=B(k)\sinh k(1-z)+C(k)\cosh k(1-z), \tag*{(22)} q = B ( k ) sinh k ( 1 − z ) + C ( k ) cosh k ( 1 − z ) , (22)
2 μ w ^ j = B j ( k ) sinh k ( 1 − z ) + C j ( k ) cosh k ( 1 − z ) + ( z − 1 ) cosh k ( 1 − z ) ( δ j 3 C − δ α j i k α k B ) + z sinh k ( 1 − z ) ( δ j 3 B − δ α j i k α k C ) , (23) 2\mu\widehat{w}_j=B_j(k)\sinh k(1-z)+C_j(k)\cosh k(1-z)
+(z-1)\cosh k(1-z)\left(\delta_{j3}C-\delta_{\alpha j}\frac{ik_{\alpha}}{k}B\right)+z\sinh k(1-z)\left(\delta_{j3}B-\delta_{\alpha j}\frac{ik_{\alpha}}{k}C\right), \tag*{(23)} 2 μ w j = B j ( k ) sinh k ( 1 − z ) + C j ( k ) cosh k ( 1 − z ) + ( z − 1 ) cosh k ( 1 − z ) ( δ j 3 C − δ α j k i k α B ) + z sinh k ( 1 − z ) ( δ j 3 B − δ α j k i k α C ) , (23)
where { B , C , B j , C j } \{B,C,B_j,C_j\} { B , C , B j , C j } , with j ∈ [ 1 , 2 , 3 ] j \in[1,2,3] j ∈ [ 1 , 2 , 3 ] , are independent of z z z . From the continuity equation (20) they satisfy
C = k B 3 + k B − i k 1 C 1 − i k 2 C 2 , C=kB_3+kB-ik_1C_1-ik_2C_2, C = k B 3 + k B − i k 1 C 1 − i k 2 C 2 ,
B = k C 3 − k C − i k 1 B 1 − i k 2 B 2 . (24) B=kC_3-kC-ik_1B_1-ik_2B_2. \tag*{(24)} B = k C 3 − k C − i k 1 B 1 − i k 2 B 2 . (24)
These constants are found on a case by case basis by transforming the boundary conditions given in (14) and solving through matrix methods the resulting set of eight coupled simultaneous equations in terms of { k , h } \{k,h\} { k , h } . For a source, (15) transforms to give
w ^ α ∣ z = 0 = w ^ 3 ∣ z = 1 = ∂ w ^ α ∂ z ∣ z = 1 = 0 , w ^ 3 ∣ z = 0 = 4 π cosh ( 1 − h ) k cosh k , (25) \widehat{w}_{\alpha}\big|_{z=0}=\widehat{w}_3\big|_{z=1}=\left.\frac{\partial\widehat{w}_{\alpha}}{\partial z}\right|_{z=1}=0,\qquad\widehat{w}_3\big|_{z=0}=4\pi\frac{\cosh(1-h)k}{\cosh k}, \tag*{(25)} w α z = 0 = w 3 z = 1 = ∂ z ∂ w α z = 1 = 0 , w 3 z = 0 = 4 π cosh k cosh ( 1 − h ) k , (25)
with corresponding full solution for w ^ j \widehat{w}_j w j
w ^ 3 = 4 π cosh ( 1 − h ) k cosh k ( sinh 2 k − 2 k ) ( k ( z − 2 ) cosh k z + sinh k ( 2 − z ) − sinh k z + k z cosh k ( 2 − z ) ) , (26) \widehat{w}_3=\frac{4\pi\cosh(1-h)k}{\cosh k(\sinh2k-2k)}\left(k(z-2)\cosh kz+\sinh k(2-z)-\sinh kz+kz\cosh k(2-z)\right), \tag*{(26)} w 3 = cosh k ( sinh 2 k − 2 k ) 4 π cosh ( 1 − h ) k ( k ( z − 2 ) cosh k z + sinh k ( 2 − z ) − sinh k z + k z cosh k ( 2 − z ) ) , (26)
w ^ α = 4 π i k α cosh ( 1 − h ) k cosh k ( sinh 2 k − 2 k ) ( ( z − 2 ) sinh k z − z sinh k ( 2 − z ) ) . (27) \widehat{w}_{\alpha}=\frac{4\pi i k_{\alpha}\cosh(1-h)k}{\cosh k(\sinh2k-2k)}\left((z-2)\sinh kz-z\sinh k(2-z)\right). \tag*{(27)} w α = cosh k ( sinh 2 k − 2 k ) 4 π i k α cosh ( 1 − h ) k ( ( z − 2 ) sinh k z − z sinh k ( 2 − z ) ) . (27)
Similarly for a Stokeslet, (15) transforms to give
w ^ j k ∣ z = 0 = 4 π ( δ j 3 δ k α + δ k 3 δ j α ) k α k ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) , ∂ w ^ α k ∂ z ∣ z = 1 = 4 π δ k 3 k α k ∂ ∂ k ( k sinh h k cosh k ) , w ^ 3 k ∣ z = 1 = 4 π δ k 3 ( ∂ ∂ k ( sinh h k cosh h k ) − sinh h k k cosh k ) . (28) \begin{aligned}
\left.\widehat{w}^{k}_{j}\right|_{z=0} &= 4\pi\left(\delta_{j3}\delta_{k\alpha}+\delta_{k3}\delta_{j\alpha}\right)\frac{k_{\alpha}}{k}\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right),\\
\left.\frac{\partial\widehat{w}^{k}_{\alpha}}{\partial z}\right|_{z=1} &= 4\pi\delta_{k3}\frac{k_{\alpha}}{k}\frac{\partial}{\partial k}\left(\frac{k\sinh hk}{\cosh k}\right),\\
\left.\widehat{w}^{k}_{3}\right|_{z=1} &= 4\pi\delta_{k3}\left(\frac{\partial}{\partial k}\left(\frac{\sinh hk}{\cosh hk}\right)-\frac{\sinh hk}{k\cosh k}\right). \tag*{(28)}
\end{aligned} w j k z = 0 ∂ z ∂ w α k z = 1 w 3 k z = 1 = 4 π ( δ j 3 δ k α + δ k 3 δ j α ) k k α ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) , = 4 π δ k 3 k k α ∂ k ∂ ( cosh k k sinh hk ) , = 4 π δ k 3 ( ∂ k ∂ ( cosh hk sinh hk ) − k cosh k sinh hk ) . (28)
with corresponding full solution for w ^ j k \widehat{w}^{k}_{j} w j k
w ^ 3 3 = 8 π k cosh 2 k ( sinh 2 k − 2 k ) [ t ] ( k 2 sinh h k sinh k z + h k 2 cosh 2 k sinh h k sinh k z + h k 2 cosh k sinh k z sinh k ( 1 − h ) + k 2 z cosh k sinh h k sinh k ( 1 − z ) − h k 2 z cosh 3 k cosh k ( 1 − h − z ) + h k cosh 2 k cosh h k sinh k z + k z cosh 2 k sinh h k cosh k z − 2 k cosh k sinh k sinh h k sinh k z − cosh 2 k sinh h k sinh k z ) , (29) \widehat{w}^{3}_{3}=\frac{8\pi}{k\cosh^{2}k(\sinh2k-2k)}
\begin{aligned}[t]
\Big(&k^{2}\sinh hk\sinh kz+hk^{2}\cosh^{2}k\sinh hk\sinh kz\\
&+hk^{2}\cosh k\sinh kz\sinh k(1-h)+k^{2}z\cosh k\sinh hk\sinh k(1-z)\\
&-hk^{2}z\cosh^{3}k\cosh k(1-h-z)+hk\cosh^{2}k\cosh hk\sinh kz\\
&+kz\cosh^{2}k\sinh hk\cosh kz-2k\cosh k\sinh k\sinh hk\sinh kz\\
&-\cosh^{2}k\sinh hk\sinh kz\Big),
\end{aligned} \tag*{(29)} w 3 3 = k cosh 2 k ( sinh 2 k − 2 k ) 8 π [ t ] ( k 2 sinh hk sinh k z + h k 2 cosh 2 k sinh hk sinh k z + h k 2 cosh k sinh k z sinh k ( 1 − h ) + k 2 z cosh k sinh hk sinh k ( 1 − z ) − h k 2 z cosh 3 k cosh k ( 1 − h − z ) + hk cosh 2 k cosh hk sinh k z + k z cosh 2 k sinh hk cosh k z − 2 k cosh k sinh k sinh hk sinh k z − cosh 2 k sinh hk sinh k z ) , (29)
w ^ α 3 = 8 π i k α k cosh 2 k ( sinh 2 k − 2 k ) [ t ] ( z cosh 2 k sinh h k sinh k z + k sinh h k cosh k z + h k z cosh 2 k sinh h k cosh k z + h k z cosh 3 k sinh k ( 1 − h − z ) − cosh k sinh h k sinh k ( 1 + z ) − h cosh 2 k sinh k sinh k ( 1 − h − z ) + h k cosh k cosh k z sinh k ( 1 − h ) − k z cosh k sinh h k cosh k ( 1 − z ) ) , (30) \widehat{w}^{3}_{\alpha}=\frac{8\pi i k_{\alpha}}{k\cosh^{2}k(\sinh2k-2k)}
\begin{aligned}[t]
\Big(&z\cosh^{2}k\sinh hk\sinh kz+k\sinh hk\cosh kz\\
&+hkz\cosh^{2}k\sinh hk\cosh kz+hkz\cosh^{3}k\sinh k(1-h-z)\\
&-\cosh k\sinh hk\sinh k(1+z)-h\cosh^{2}k\sinh k\sinh k(1-h-z)\\
&+hk\cosh k\cosh kz\sinh k(1-h)-kz\cosh k\sinh hk\cosh k(1-z)\Big),
\end{aligned} \tag*{(30)} w α 3 = k cosh 2 k ( sinh 2 k − 2 k ) 8 π i k α [ t ] ( z cosh 2 k sinh hk sinh k z + k sinh hk cosh k z + hk z cosh 2 k sinh hk cosh k z + hk z cosh 3 k sinh k ( 1 − h − z ) − cosh k sinh hk sinh k ( 1 + z ) − h cosh 2 k sinh k sinh k ( 1 − h − z ) + hk cosh k cosh k z sinh k ( 1 − h ) − k z cosh k sinh hk cosh k ( 1 − z ) ) , (30)
w ^ 3 α = 4 π i k α k ( sinh 2 k − 2 k ) ( ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) , (31) \widehat{w}^{\alpha}_{3}=\frac{4\pi i k_{\alpha}}{k(\sinh2k-2k)}
\left(\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right)
\left(k(z-2)\cosh kz+kz\cosh k(2-z)+\sinh k(2-z)-\sinh kz\right), \tag*{(31)} w 3 α = k ( sinh 2 k − 2 k ) 4 π i k α ( ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) , (31)
w ^ β α = 4 π k α k β k ( sinh 2 k − 2 k ) ( ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) ( z sinh k ( 2 − z ) − ( z − 2 ) sinh k z ) . (32) \widehat{w}^{\alpha}_{\beta}=\frac{4\pi k_{\alpha}k_{\beta}}{k(\sinh2k-2k)}
\left(\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right)
\left(z\sinh k(2-z)-(z-2)\sinh kz\right). \tag*{(32)} w β α = k ( sinh 2 k − 2 k ) 4 π k α k β ( ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) ( z sinh k ( 2 − z ) − ( z − 2 ) sinh k z ) . (32)
Rewriting the inverse Fourier transform in terms of Hankel transforms, we obtain for the source
w 3 = 1 2 π H 0 ( w ^ 3 ) , w α = i x α 2 π ρ H 1 ( k w ^ α k α ) , (33) w_{3}=\frac{1}{2\pi}\mathcal{H}_{0}\left(\widehat{w}_{3}\right),\qquad
w_{\alpha}=\frac{ix_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(\frac{k\widehat{w}_{\alpha}}{k_{\alpha}}\right), \tag*{(33)} w 3 = 2 π 1 H 0 ( w 3 ) , w α = 2 π ρ i x α H 1 ( k α k w α ) , (33)
and for the Stokeslet
w 3 3 = 1 2 π H 0 ( w ^ 3 3 ) , w 3 α = i x α 2 π ρ H 1 ( k k α w ^ 3 α ) , w α α = i x α 2 π ρ H 1 ( k k α w ^ α α ) , (34) w^{3}_{3}=\frac{1}{2\pi}\mathcal{H}_{0}\left(\widehat{w}^{3}_{3}\right),\qquad
w^{\alpha}_{3}=\frac{ix_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(\frac{k}{k_{\alpha}}\widehat{w}^{\alpha}_{3}\right),\qquad
w^{\alpha}_{\alpha}=\frac{ix_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(\frac{k}{k_{\alpha}}\widehat{w}^{\alpha}_{\alpha}\right), \tag*{(34)} w 3 3 = 2 π 1 H 0 ( w 3 3 ) , w 3 α = 2 π ρ i x α H 1 ( k α k w 3 α ) , w α α = 2 π ρ i x α H 1 ( k α k w α α ) , (34)
w β α = 1 2 π ( δ α β ρ − 2 x α x β ρ 3 ) H 1 ( k k α k β w ^ β α ) + x α x β 2 π ρ H 0 ( k 2 k α k β w ^ β α ) , (35) w^{\alpha}_{\beta}=\frac{1}{2\pi}\left(\frac{\delta_{\alpha\beta}}{\rho}-\frac{2x_{\alpha}x_{\beta}}{\rho^{3}}\right)\mathcal{H}_{1}\left(\frac{k}{k_{\alpha}k_{\beta}}\widehat{w}^{\alpha}_{\beta}\right)+\frac{x_{\alpha}x_{\beta}}{2\pi\rho}\mathcal{H}_{0}\left(\frac{k^{2}}{k_{\alpha}k_{\beta}}\widehat{w}^{\alpha}_{\beta}\right), \tag*{(35)} w β α = 2 π 1 ( ρ δ α β − ρ 3 2 x α x β ) H 1 ( k α k β k w β α ) + 2 π ρ x α x β H 0 ( k α k β k 2 w β α ) , (35)
where α ∈ [ 1 , 2 ] \alpha\in[1,2] α ∈ [ 1 , 2 ] and H i \mathcal{H}_{i} H i is the Hankel transform of order i i i . Similar integral expressions in terms of Hankel transforms can be constructed for other Stokes singularities (see Appendix § 1.11 for the rotlet, Appendix § 1.12 for the stresslet, Appendix § 1.13 for the rotlet dipole, and Appendix E for the source dipole).
To illustrate the nature of these exact solutions, Figure 3 plots various components of the fluid velocity field induced by four of the main singularities, the rotlet, source, rotlet dipole and source dipole, as a function of vertical height z z z for a range of horizontal radial distances away from the singularities, in each case located at h = 0.4 h = 0.4 h = 0.4 .
Figure 3. The near field velocity u i u_i u i produced by a number of singularities at h = 0.4 h = 0.4 h = 0.4 as a function of z z z for a range of x ∈ { 0.19 , 0.25 , 0.3 , 0.4 } x \in\{0.19, 0.25, 0.3, 0.4\} x ∈ { 0.19 , 0.25 , 0.3 , 0.4 } , y = 0 y = 0 y = 0 , with darker colours denoting larger x x x . (a) Rotlet, i = θ i = \theta i = θ (green curves) (b) Source, i = r i = r i = r (green) or i = z i = z i = z (blue) (c) Rotlet dipole, i = θ i = \theta i = θ (green) (d) Source dipole, i = r i = r i = r (green) or i = z i = z i = z (blue). Note that here ( r , θ ) (r, \theta) ( r , θ ) are the polar coordinates for the horizontal plane i.e. x = r cos θ x = r \cos\theta x = r cos θ and y = r sin θ y = r \sin\theta y = r sin θ .
For the swirling component of the flow due to a rotlet, Figure 3 (a) illustrates clearly how the boundary conditions of no slip and no stress are satisfied, and the incipient divergence as the x x x location approaches that of the singularity. For the source in Figure 3 (b) the horizontal velocity u x u_x u x displays an increasing maximum as the observation point x x x approaches the singularity location, while the vertical velocity component u z u_z u z has a positive divergence for z → h + z \to h^{+} z → h + and a negative divergence as z → h − z \to h^{-} z → h − as expected for a source, while vanishing at the top and bottom boundaries, as required by (11). Both the rotlet dipole in Figure 3 (c) and the source dipole in Figure 3 (d) appear as derivatives of their corresponding monopoles.
Far-field Solutions
It is difficult to find the far-field (ρ ≫ 1 \rho\gg1 ρ ≫ 1 ) behaviour of these solutions when they are expressed as exact solutions in integral form as Hankel transforms. Following the approach of Liron and Mochon [29 ] , we may utilise a contour integration to express the exact solutions in series form. Given an even function f ( z ) f(z) f ( z ) decaying exponentially to zero on the real axis as z = x → ± ∞ z=x\to\pm\infty z = x → ± ∞ , consider the contour integral ∮ γ F \oint_{\gamma} F ∮ γ F where F = z i + 1 f ( z ) H i 1 ( ρ z ) F=z^{i+1}f(z)H_i^1(\rho z) F = z i + 1 f ( z ) H i 1 ( ρ z ) , H i 1 = J i + i Y i H_i^1=J_i+iY_i H i 1 = J i + i Y i with i ∈ [ 0 , 1 ] i\in[0,1] i ∈ [ 0 , 1 ] is a Hankel function of the 1st kind and γ = γ 0 + γ 1 + γ R + γ ϵ \gamma=\gamma_0+\gamma_1+\gamma_R+\gamma_\epsilon γ = γ 0 + γ 1 + γ R + γ ϵ is a notched semicircular contour centered at the origin (Figure 4 ). From Watson [38 ] , ∫ γ R F → 0 \int_{\gamma_R}F\to0 ∫ γ R F → 0 as R → ∞ R\to\infty R → ∞ . Hence, applying the residue theorem in the limit as R → ∞ R\to\infty R → ∞ and ϵ → 0 \epsilon\to0 ϵ → 0 yields
Figure 4. The notched semicircular contour γ \gamma γ .
∫ 0 ∞ λ i + 1 d λ J i ( λ ρ ) f = − 1 2 ∫ γ ϵ F + π i ∑ Residues of singularities of F in γ . (36) \int_0^\infty\lambda^{i+1}\mathrm{d}\lambda\,J_i(\lambda\rho)f=-\frac{1}{2}\int_{\gamma_\epsilon}F+\pi i\sum\text{Residues of singularities of }F\text{ in }\gamma. \tag*{(36)} ∫ 0 ∞ λ i + 1 d λ J i ( λ ρ ) f = − 2 1 ∫ γ ϵ F + π i ∑ Residues of singularities of F in γ . (36)
Using this method, the repeated reflection solutions v j v_j v j for all four primary Stokes singularities can be directly expressed in series form. For a source, v j v_j v j becomes
v 3 = − 2 π ∑ n = 1 , 3 , 5 , … ∞ n sin ( n π h 2 ) cos ( n π z 2 ) K 0 ( n π ρ 2 ) , (37) v_3=-2\pi\sum_{n=1,3,5,\ldots}^{\infty}n\sin\left(\frac{n\pi h}{2}\right)\cos\left(\frac{n\pi z}{2}\right)K_0\left(\frac{n\pi\rho}{2}\right), \tag*{(37)} v 3 = − 2 π n = 1 , 3 , 5 , … ∑ ∞ n sin ( 2 nπ h ) cos ( 2 nπ z ) K 0 ( 2 nπ ρ ) , (37)
v α = 2 π x α ρ ∑ n = 1 , 3 , 5 , … ∞ n sin ( n π h 2 ) sin ( n π z 2 ) K 1 ( n π ρ 2 ) . (38) v_\alpha=\frac{2\pi x_\alpha}{\rho}\sum_{n=1,3,5,\ldots}^{\infty}n\sin\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)K_1\left(\frac{n\pi\rho}{2}\right). \tag*{(38)} v α = ρ 2 π x α n = 1 , 3 , 5 , … ∑ ∞ n sin ( 2 nπ h ) sin ( 2 nπ z ) K 1 ( 2 nπ ρ ) . (38)
Note that for all four singularities, the dominant term in the far-field expansion (ρ ≫ 1 \rho\gg1 ρ ≫ 1 ) of the repeated reflection solution v j v_j v j comes from the n = 0 n=0 n = 0 terms and decays like exp ( − π ρ / 2 ) \exp(-\pi\rho/2) exp ( − π ρ /2 ) . Similarly, the integral expressions for the auxiliary solution w j w_j w j can be expressed in series form to obtain series expansions for the full flow field u j u_j u j . For a source, the corresponding complex function F F F has in γ \gamma γ poles of order 1 at z = π i ( n + 1 / 2 ) z=\pi i(n+1/2) z = π i ( n + 1/2 ) where n ∈ Z ≥ n\in\mathbb{Z}_{\geq} n ∈ Z ≥ and poles of order 1 at z = z 0 / 2 z=z_0/2 z = z 0 /2 where z 0 z_0 z 0 satisfies sinh z 0 = z 0 \sinh z_0=z_0 sinh z 0 = z 0 . Since ∫ γ ϵ \int_{\gamma_\epsilon} ∫ γ ϵ vanishes as ϵ → 0 \epsilon\to0 ϵ → 0 , when j = k = l = 3 j=k=l=3 j = k = l = 3 (36) simplifies to become
w 3 = 2 π ( 1 − z ) ∑ n = 1 , 3 , 5 , … ∞ n sin ( n π h 2 ) cos ( n π z 2 ) K 0 ( n π ρ 2 ) + O ( e − ρ y 1 / 2 ρ ) (39) w_3=2\pi(1-z)\sum_{n=1,3,5,\ldots}^{\infty}n\sin\left(\frac{n\pi h}{2}\right)\cos\left(\frac{n\pi z}{2}\right)K_0\left(\frac{n\pi\rho}{2}\right)+O\left(\frac{e^{-\rho y_1/2}}{\sqrt{\rho}}\right) \tag*{(39)} w 3 = 2 π ( 1 − z ) n = 1 , 3 , 5 , … ∑ ∞ n sin ( 2 nπ h ) cos ( 2 nπ z ) K 0 ( 2 nπ ρ ) + O ( ρ e − ρ y 1 /2 ) (39)
u 3 = v 3 + w 3 = − 2 π z ∑ n = 1 , 3 , 5 , … ∞ n sin ( n π h 2 ) cos ( n π z 2 ) K 0 ( n π ρ 2 ) . (40) \begin{aligned}
u_3&=v_3+w_3\\
&=-2\pi z\sum_{n=1,3,5,\ldots}^{\infty}n\sin\left(\frac{n\pi h}{2}\right)\cos\left(\frac{n\pi z}{2}\right)K_0\left(\frac{n\pi\rho}{2}\right). \tag*{(40)}
\end{aligned} u 3 = v 3 + w 3 = − 2 π z n = 1 , 3 , 5 , … ∑ ∞ n sin ( 2 nπ h ) cos ( 2 nπ z ) K 0 ( 2 nπ ρ ) . (40)
The first term dominates in the far-field, so
u 3 ≃ − 2 π z ρ cos ( π z 2 ) sin ( π h 2 ) e − ρ π / 2 + O ( e − ρ π / 2 ρ 3 / 2 ) , (41) u_3\simeq-\frac{2\pi z}{\sqrt{\rho}}\cos\left(\frac{\pi z}{2}\right)\sin\left(\frac{\pi h}{2}\right)e^{-\rho\pi/2}+O\left(\frac{e^{-\rho\pi/2}}{\rho^{3/2}}\right), \tag*{(41)} u 3 ≃ − ρ 2 π z cos ( 2 π z ) sin ( 2 π h ) e − ρ π /2 + O ( ρ 3/2 e − ρ π /2 ) , (41)
namely an exponential radial decay with z z z dependence z cos ( π z / 2 ) z\cos(\pi z/2) z cos ( π z /2 ) , vanishing at both surfaces. Furthermore, when j = α ∈ [ 1 , 2 ] j = \alpha\in[1,2] j = α ∈ [ 1 , 2 ] , the leading order contribution in the far field arises from γ ϵ \gamma_\epsilon γ ϵ , namely
u α = z ( 2 − z ) [ 3 x α ρ 2 ] , (42) u_\alpha= z(2-z)\left[\frac{3x_\alpha}{\rho^2}\right], \tag*{(42)} u α = z ( 2 − z ) [ ρ 2 3 x α ] , (42)
noting that the contribution from the poles at z = π i ( n + 1 / 2 ) z = \pi i(n+1/2) z = π i ( n + 1/2 ) in w α w_\alpha w α cancels out with v α v_\alpha v α . Similarly for a Stokeslet, F F F has poles of order 2 at z = π i ( n + 1 / 2 ) z = \pi i(n+1/2) z = π i ( n + 1/2 ) where n ∈ Z ≥ 0 n \in\mathbb{Z}_{\geq0} n ∈ Z ≥ 0 and poles of order 1 at z = z 0 / 2 z = z_0/2 z = z 0 /2 where z 0 z_0 z 0 satisfies sinh z 0 = z 0 \sinh z_0 = z_0 sinh z 0 = z 0 . When j = k = 3 j = k = 3 j = k = 3 , since γ ϵ \gamma_\epsilon γ ϵ vanishes as ϵ → 0 \epsilon\to0 ϵ → 0 , (36) simplifies to
w 3 3 = − ∑ n = 1 , 3 , 5 , … ∞ sin ( n π h 2 ) sin ( n π z 2 ) ( 8 K 0 ( n π ρ 2 ) − 2 n π ρ K 1 ( n π ρ 2 ) ) + ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 8 ( cosh z 0 − 1 ) ( w ^ 3 3 ( sinh 2 k − 2 k ) ) ∣ k = z 0 / 2 H 0 1 ( ρ z 0 2 ) w_3^3 = -\sum_{n=1,3,5,\ldots}^{\infty}\sin\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)\left(8K_0\left(\frac{n\pi\rho}{2}\right)-2n\pi\rho K_1\left(\frac{n\pi\rho}{2}\right)\right)
+\sum_{\substack{z_0\in\mathbb{H}:\\ z_0=\sinh z_0}}\frac{i z_0}{8(\cosh z_0-1)}\left(\widehat{w}_3^3(\sinh2k-2k)\right)\bigg|_{k=z_0/2}H_0^1\left(\frac{\rho z_0}{2}\right) w 3 3 = − n = 1 , 3 , 5 , … ∑ ∞ sin ( 2 nπ h ) sin ( 2 nπ z ) ( 8 K 0 ( 2 nπ ρ ) − 2 nπ ρ K 1 ( 2 nπ ρ ) ) + z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 ( cosh z 0 − 1 ) i z 0 ( w 3 3 ( sinh 2 k − 2 k ) ) k = z 0 /2 H 0 1 ( 2 ρ z 0 )
u 3 3 = v 3 3 + w 3 3 = ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 8 ( cosh z 0 − 1 ) × ( w ^ 3 3 ( sinh 2 k − 2 k ) ) ∣ k = z 0 / 2 H 0 1 ( ρ z 0 2 ) , (43) u_3^3=v_3^3+w_3^3=\sum_{\substack{z_0\in\mathbb{H}:\\ z_0=\sinh z_0}}\frac{i z_0}{8(\cosh z_0-1)}\times\left(\widehat{w}_3^3(\sinh2k-2k)\right)\bigg|_{k=z_0/2}H_0^1\left(\frac{\rho z_0}{2}\right), \tag*{(43)} u 3 3 = v 3 3 + w 3 3 = z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 ( cosh z 0 − 1 ) i z 0 × ( w 3 3 ( sinh 2 k − 2 k ) ) k = z 0 /2 H 0 1 ( 2 ρ z 0 ) , (43)
noting that the contribution from the poles of order 2 in w 3 3 w_3^3 w 3 3 cancels out with v 3 3 v_3^3 v 3 3 . The leading far-field behavior is
u 3 3 = O ( e − ρ y 1 / 2 ρ 1 / 2 ) , (44) u_{3}^{3} = \mathcal{O}\left(\frac{e^{-\rho y_{1}/2}}{\rho^{1/2}}\right), \tag*{(44)} u 3 3 = O ( ρ 1/2 e − ρ y 1 /2 ) , (44)
where y 1 = 7.498 … y_{1}=7.498\ldots y 1 = 7.498 … is the imaginary part of the first non-zero root to sinh z 0 = z 0 \sinh z_{0}=z_{0} sinh z 0 = z 0 in the first quadrant. Similarly for j = α j=\alpha j = α , k = 3 k=3 k = 3 and k = α k=\alpha k = α , j = 3 j=3 j = 3 where α ∈ [ 1 , 2 ] \alpha\in[1,2] α ∈ [ 1 , 2 ] , the leading order far-field contribution is
u α 3 , u 3 α = O ( x α e − ρ y 1 / 2 ρ 3 / 2 ) . (45) u_{\alpha}^{3},u_{3}^{\alpha}=\mathcal{O}\left(\frac{x_{\alpha}e^{-\rho y_{1}/2}}{\rho^{3/2}}\right). \tag*{(45)} u α 3 , u 3 α = O ( ρ 3/2 x α e − ρ y 1 /2 ) . (45)
When j = β j=\beta j = β and k = α k=\alpha k = α where α , β ∈ [ 1 , 2 ] \alpha,\beta\in[1,2] α , β ∈ [ 1 , 2 ] , the leading order contribution in the far-field arises from γ β ϵ \gamma_{\beta}^{\epsilon} γ β ϵ ,
u β α = z ( 2 − z ) [ − 3 h ( 2 − h ) ρ 2 ( δ α β − 2 x α x β ρ 2 ) ] . (46) u_{\beta}^{\alpha}=z(2-z)\left[-\frac{3h(2-h)}{\rho^{2}}\left(\delta_{\alpha\beta}-\frac{2x_{\alpha}x_{\beta}}{\rho^{2}}\right)\right]. \tag*{(46)} u β α = z ( 2 − z ) [ − ρ 2 3 h ( 2 − h ) ( δ α β − ρ 2 2 x α x β ) ] . (46)
Similar far-field approximations can be found for the other Stokes singularities (Appendix § 1.11 , rotlet; Appendix § 1.12 , stresslet; Appendix § 1.13 , rotlet dipole; Appendix E, source dipole). Figure Figure 5 plots streamlines of these far field flows in the horizontal plane z = 1 z=1 z = 1 . In Figure 5 (a), a Stokeslet orientated in the x x x direction generates a flow with a recirculating flow pattern of two loops decaying radially like 1 / ρ 2 1/\rho^{2} 1/ ρ 2 , namely a two dimensional source dipole (recalling that the source flow u s = x i / ρ 2 u_{s}=x_{i}/\rho^{2} u s = x i / ρ 2 leads to the source dipole flow u s d = δ i j / ρ 2 − 2 x i x j / ρ 4 u_{sd}=\delta_{ij}/\rho^{2}-2x_{i}x_{j}/\rho^{4} u s d = δ ij / ρ 2 − 2 x i x j / ρ 4 ). Confinement has fundamentally affected the unidirectionality of the flow by inducing recirculation in the y y y direction. This is a feature of the family of Stokes singularities that are derivatives of the Stokeslet, with higher order singularities having more recirculation loops. For example, a Stokes dipole has four loops while a Stokes quadrupole has six. In contrast, the spherical symmetry of a three dimensional source ensures that the new flow is still a source (Figure 5 (b)). Derivatives of the source, such as the source dipole, are also unchanged by confinement, and since the vertically orientated rotlet is independent of z z z , its streamlines are also unchanged, as seen in Figure 5 (c). Confinement breaks the symmetries of the horizontal rotlet and stresslet, leading to flows with the character a two dimensional source dipole for both a horizontally orientated rotlet (Figure 5 (d)) and a vertical stresslet (j = 1 j=1 j = 1 , k = 3 k=3 k = 3 , Figure 5 (e)) and a two dimensional source for a horizontal stresslet (j = k = 3 j=k=3 j = k = 3 , Figure 5 (f)), respectively.
Figure 5. Streamlines in the z = 1 z = 1 z = 1 plane for the flows generated by Stokes singularities in the far field thin-film limit (ρ ≫ H \rho\gg H ρ ≫ H ): (a) Stokeslet orientated in the positive x x x direction, (b) source, (c) and (d) Rotlet orientated in the z z z and x x x directions, respectively, (e) and (f) Stresslet u k , l u^{k,l} u k , l with k = 1 , l = 3 k = 1,l = 3 k = 1 , l = 3 and k = l = 1 k = l = 1 k = l = 1 , respectively. As streamlines in (f) depend on h h h , we have set h = 1 / 2 h = 1/2 h = 1/2 .
Leading Order Far Field Flow
Examining the cases given above in §V and in Appendices (B)-(E), we note that for the four primary Stokes singularities, the leading order far-field flow is separable in z z z (formally considering the limit where h h h , H H H , z z z are fixed while ρ \rho ρ is large). If the flow does not decay exponentially radially, the it has z z z dependence of the form z ( 1 − z / 2 ) z(1-z/2) z ( 1 − z /2 ) . Otherwise, the flow decays exponentially either as exp ( − ρ π / 2 ) \exp(-\rho\pi/2) exp ( − ρ π /2 ) , arising from a K 1 ( ρ π / 2 ) K_{1}(\rho\pi/2) K 1 ( ρ π /2 ) term with corresponding z z z dependence either sin π z / 2 \sin\pi z/2 sin π z /2 for horizontal flow or z cos π z / 2 z\cos\pi z/2 z cos π z /2 for vertical flow, or exp ( − ρ y 1 / 2 ) \exp(-\rho y_{1}/2) exp ( − ρ y 1 /2 ) where y 1 ≈ 7.498 y_{1}\approx7.498 y 1 ≈ 7.498 is the imaginary part of the first non-zero root to sinh z 0 = z 0 \sinh z_{0}=z_{0} sinh z 0 = z 0 in the upper half plane. All higher order Stokes singularities can be expressed as derivatives of these four primary Stokes singularities. These singularities must also either have leading order z z z dependence z ( 1 − z / 2 ) z(1-z/2) z ( 1 − z /2 ) or decay exponentially like exp ( − ρ π / 2 ) \exp(-\rho\pi/2) exp ( − ρ π /2 ) or exp ( − ρ y 1 / 2 ) \exp(-\rho y_{1}/2) exp ( − ρ y 1 /2 ) . This means that the leading order far field contribution to the flow from these singularities can be obtained directly by differentiating the far field flows for the primary Stokes singularities, namely the full exact solutions which quickly become very complicated do not need to be derived. For example, differentiating (46) once, (46) twice and (42) once recovers the far field flows for a Stokes dipole, a Stokes quadrupole and a source dipole respectively given in [31 ] , noting a sign error there in the expression given for a Stokes quadrupole (their equation (B8)), namely
[ u D d ] i = 6 ρ 4 ( x i + 2 x j δ i j − 4 x i x j 2 ρ 2 ) h z ( 2 − h ) ( 2 − z ) , (47) \left[u^{d}_{D}\right]_{i}=\frac{6}{\rho^{4}}\left(x_i+2x_j\delta_{ij}-\frac{4x_i x_j^2}{\rho^2}\right)hz(2-h)(2-z), \tag*{(47)} [ u D d ] i = ρ 4 6 ( x i + 2 x j δ ij − ρ 2 4 x i x j 2 ) h z ( 2 − h ) ( 2 − z ) , (47)
[ u Q j ] i = 18 ρ 4 ( δ i j − 4 x i x j ρ 2 − 4 x j 2 δ i j ρ 2 + 8 x i x j 3 ρ 4 ) × h z ( 2 − h ) ( 2 − z ) , (48) \left[u^{j}_{Q}\right]_{i}=\frac{18}{\rho^{4}}\left(\delta_{ij}-\frac{4x_i x_j}{\rho^2}-\frac{4x_j^2\delta_{ij}}{\rho^2}+\frac{8x_i x_j^3}{\rho^4}\right)\times hz(2-h)(2-z), \tag*{(48)} [ u Q j ] i = ρ 4 18 ( δ ij − ρ 2 4 x i x j − ρ 2 4 x j 2 δ ij + ρ 4 8 x i x j 3 ) × h z ( 2 − h ) ( 2 − z ) , (48)
[ u S j ] i = 6 z ρ 2 ( δ i j − 2 x i x j ρ 2 ) ( 1 − z 2 ) . (49) \left[u^{j}_{S}\right]_{i}=\frac{6z}{\rho^2}\left(\delta_{ij}-\frac{2x_i x_j}{\rho^2}\right)\left(1-\frac{z}{2}\right). \tag*{(49)} [ u S j ] i = ρ 2 6 z ( δ ij − ρ 2 2 x i x j ) ( 1 − 2 z ) . (49)
As a consistency check, (49) does indeed reproduce what was derived from first principles in Appendix E. Hence, for an arbitrary body whose free-space locomotion can be captured by a expansion in terms of Stokes singularities, the far field flow field is separable in z z z with either z z z dependence of the form z ( 1 − z / 2 ) z(1-z/2) z ( 1 − z /2 ) or the flow decays radially exponentially. The fluid velocity field u \boldsymbol{u} u can thus be factorised as u = f ( z ) U ( x h ) \boldsymbol{u}=f(z)\boldsymbol{U}(\boldsymbol{x}_h) u = f ( z ) U ( x h ) where x h = ( x 1 , x 2 ) \boldsymbol{x}_h=(x_1,x_2) x h = ( x 1 , x 2 ) and f ( z ) f(z) f ( z ) is normalised so that
∫ 0 H f d z = 1 \int_0^H f\,\mathrm{d}z=1 ∫ 0 H f d z = 1
(typically f f f is either 3 z ( 1 − z / 2 ) 3z(1-z/2) 3 z ( 1 − z /2 ) or π sin ( π z / 2 ) / 2 \pi\sin(\pi z/2)/2 π sin ( π z /2 ) /2 ). The 3D Stokes equations for u \boldsymbol{u} u reduces to a Brinkman-like equation for the vertically averaged fluid velocity U \boldsymbol{U} U
μ ( ∇ 2 − κ 2 ) U = ∇ p , (50) \mu\left(\nabla^2-\kappa^2\right)\boldsymbol{U}=\nabla p, \tag*{(50)} μ ( ∇ 2 − κ 2 ) U = ∇ p , (50)
with corresponding incompressibility condition ∇ ⋅ U = 0 \nabla\cdot\boldsymbol{U}=0 ∇ ⋅ U = 0 , where κ = ( ∂ f / ∂ z ∣ z = 0 ) 1 / 2 \kappa=(\left.\partial f/\partial z\right|_{z=0})^{1/2} κ = ( ∂ f / ∂ z ∣ z = 0 ) 1/2 plays the role of the inverse Debye screening length in screened electrostatics. We have thus reduced a 3D system to a 2D one that can be solved by transforming to an appropriate coordinate system that simplifies the boundary conditions. This method is equally applicable in the setup of Liron and Mochon [29 ] , namely a microfluidic environment between two horizontal rigid boundaries, where the corresponding far field z z z dependence for an non radially exponentially decaying flow is z ( 1 − z ) z(1-z) z ( 1 − z ) .
Self-consistency Check
A key assumption made above was that the combination of surface tension and gravitational effects restricts vertical deformation of the interface and hence H H H can be assumed constant. As a self-consistency check, using (41) , the leading order contribution to the stress σ x 3 x 3 \sigma_{x_3x_3} σ x 3 x 3 in the far field at the upper free surface boundary x 3 = H x_3=H x 3 = H that a source of strength λ S \lambda_S λ S (namely generating a flow u i = λ S x i / r 3 u_i=\lambda_S x_i/r^3 u i = λ S x i / r 3 ) at ( x 1 , x 2 , x 3 ) = ( 0 , 0 , s ) (x_1,x_2,x_3)=(0,0,s) ( x 1 , x 2 , x 3 ) = ( 0 , 0 , s ) produces is
σ x 3 x 3 = 2 μ ∂ u 3 ∂ x 3 = μ π 2 H 3 sin ( π s 2 H ) K 0 ( ρ π 2 H ) ≤ μ π 2 H 3 K 0 ( ρ π 2 H ) ≈ μ π 2 ρ 1 / 2 H 5 / 2 e − ρ π / 2 H , (51) \sigma_{x_3x_3}=2\mu\frac{\partial u_3}{\partial x_3}=\frac{\mu\pi^2}{H^3}\sin\left(\frac{\pi s}{2H}\right)K_0\left(\frac{\rho\pi}{2H}\right)
\leq\frac{\mu\pi^2}{H^3}K_0\left(\frac{\rho\pi}{2H}\right)\approx\frac{\mu\pi^2}{\rho^{1/2}H^{5/2}}e^{-\rho\pi/2H}, \tag*{(51)} σ x 3 x 3 = 2 μ ∂ x 3 ∂ u 3 = H 3 μ π 2 sin ( 2 H π s ) K 0 ( 2 H ρ π ) ≤ H 3 μ π 2 K 0 ( 2 H ρ π ) ≈ ρ 1/2 H 5/2 μ π 2 e − ρ π /2 H , (51)
when ρ ≫ 2 H / π \rho\gg2H/\pi ρ ≫ 2 H / π . Here, we have utilized the asymptotic large argument expansion for K α K_\alpha K α [39 ] together with the fact that ∣ sin ( π s / 2 H ) ∣ ≤ 1 ∀ s ∈ [ 0 , H ] \lvert\sin(\pi s/2H)\rvert\leq1\ \forall\ s\in[0,H] ∣ sin ( π s /2 H )∣ ≤ 1 ∀ s ∈ [ 0 , H ] . Hence, a measure M s M_s M s of the relative strength of the stresses at the free surface arising from the flow generated by the singularity that seek to deform this surface to the gravitational forces restricting vertical deformation is
M s = μ π 2 λ S ρ w ρ 1 / 2 g H 5 / 2 Δ H e − ρ π / 2 H . (52) M_s=\frac{\mu\pi^2\lambda_S}{\rho_w\rho^{1/2}gH^{5/2}\Delta H}e^{-\rho\pi/2H}. \tag*{(52)} M s = ρ w ρ 1/2 g H 5/2 Δ H μ π 2 λ S e − ρ π /2 H . (52)
Writing the strength of the source λ S \lambda_S λ S as λ S = U S H 2 \lambda_S=U_SH^2 λ S = U S H 2 , U S U_S U S scales with the typical velocities of flows in a Petri dish, namely U S ∼ 2 m m s − 1 U_S\sim2\ \mathrm{mm\,s^{-1}} U S ∼ 2 mm s − 1 . Hence, setting μ = 1 m P a s − 1 \mu=1\ \mathrm{mPa\,s^{-1}} μ = 1 mPa s − 1 , H = 5 m m H=5\ \mathrm{mm} H = 5 mm , Δ H = 0.1 m m \Delta H=0.1\ \mathrm{mm} Δ H = 0.1 mm , ρ = 1 c m \rho=1\ \mathrm{cm} ρ = 1 cm we find M s ≈ 1.2 × 10 − 4 ≪ 1 M_s\approx1.2\times10^{-4}\ll1 M s ≈ 1.2 × 1 0 − 4 ≪ 1 , so M s M_s M s is indeed small and thus the flat surface approximation is consistent for a source.
Case Study: Hydrodynamic Bound States
An instructive application of the results of this paper is exploring the notion of “hydrodynamic bound states”. First discovered by Drescher, et al. in 2009 using the green alga Volvox [32 ] , these are dynamical states exhibited by pairs of spherical chiral microswimmers near a surface. Volvox colonies have radius R ∼ 250 μ m R \sim250\,\mu\mathrm{m} R ∼ 250 μ m , with ∼ 10 3 \sim10^{3} ∼ 1 0 3 biflagellated somatic cells beating on their surface. This beating is primarily in the posterior-anterior direction, but has a modest orthogonal component that leads to spinning motion about the AP axis. While the organisms are slightly denser than the fluid surrounding them, the flagellar beating allows them to swim upwards against gravity. When a suspension of Volvox was placed in a glass-topped chamber, the colonies naturally swam upwards due to their bottom-heaviness, which aligned their AP axis with gravity. Pairs of colonies at the chamber top were found to move together while they continued to spin, eventually touching and orbiting about each other.
As shown schematically in Figure 6 , once the colonies have ascended as high as possible, their centers are a distance R = ϵ H R = \epsilon H R = ϵH (with ϵ ≪ 1 \epsilon\ll1 ϵ ≪ 1 ) below the upper no-slip surface. Due to their positive density offset relative to the surrounding ambient water, they are acted on by a downward gravitational force. Viewed from afar, each colony can be considered as a point force acting on a fluid: the resultant flow field is that of a downward-pointing Stokeslet of magnitude F = ( 4 π / 3 ) R 3 Δ ρ g F = (4\pi/3)R^3\Delta\rho g F = ( 4 π /3 ) R 3 Δ ρ g associated with the gravitational force. This geometry—two nearby Stokeslets directed away from a no-slip wall—is exactly that envisioned by Squires [33 ] in his analysis of surface-mediated interactions, who showed that the mutual advection of those Stokeslets toward each other is described by the dynamics of their separation r r r in the form
Figure 6. Geometry of hydrodynamic bound states. Two spherical, negatively buoyant microswimmers of radius R R R just below an upper surface, a horizontal distance r r r apart.
r ˙ = − 3 F π μ R r R 4 ( r 2 + 4 R 2 ) 5 / 2 , (53) \dot{r}=-\frac{3F}{\pi\mu R}\frac{rR^{4}}{(r^{2}+4R^{2})^{5/2}}, \tag*{(53)} r ˙ = − π μ R 3 F ( r 2 + 4 R 2 ) 5/2 r R 4 , (53)
expressed in a way that identifies the characteristic speed F / μ R F/\mu R F / μ R . Tracking of Volvox pairs showed precise quantitative agreement with this result [32 ] . While it was not clear a priori that the Stokeslet approximation was valid over the large range of inter-colony separations explored, direct measurements of the flow fields around freely swimming colonies [21 ] showed that the Stokeslet does indeed dominate all higher-order singularities beyond a few radii from the colony center.
This general phenomenon has been rediscovered several times: in suspensions of the fast-moving bacterium Thiovulum majus [35 ] , of the magnetotactic bacterium Magnetotacticum magnetotacticum [36 ] , and of starfish embryos [37 ] . In the latter case, the pairwise bound states occur at the air-water interface, which can be taken to be a stress-free boundary. In that case, and for an infinitely deep fluid, the image system for each Stokeslet is simply an opposite Stokeslet above the air-water interface - singularity 2 in Figure 2 . Thus, the lateral flow at ( x 1 , 0 , x 3 ) (x_1,0,x_3) ( x 1 , 0 , x 3 ) due to a downward Stokeslet at the origin is
u = F 8 π μ { x 1 ( x 3 − ( 1 − ϵ ) H ) [ x 1 2 + ( x 3 − ( 1 − ϵ ) H ) 2 ] 3 / 2 − x 1 ( x 3 − ( 1 + ϵ ) H ) [ x 1 2 + ( x 3 − ( 1 + ϵ ) H ) 2 ] 3 / 2 } . (54) u=\frac{F}{8\pi\mu}\left\{\frac{x_1\left(x_3-(1-\epsilon)H\right)}{\left[x_1^2+\left(x_3-(1-\epsilon)H\right)^2\right]^{3/2}}-\frac{x_1\left(x_3-(1+\epsilon)H\right)}{\left[x_1^2+\left(x_3-(1+\epsilon)H\right)^2\right]^{3/2}}\right\}. \tag*{(54)} u = 8 π μ F ⎩ ⎨ ⎧ [ x 1 2 + ( x 3 − ( 1 − ϵ ) H ) 2 ] 3/2 x 1 ( x 3 − ( 1 − ϵ ) H ) − [ x 1 2 + ( x 3 − ( 1 + ϵ ) H ) 2 ] 3/2 x 1 ( x 3 − ( 1 + ϵ ) H ) ⎭ ⎬ ⎫ . (54)
If we evaluate this flow at the Stokeslet location x 3 = ( 1 − ϵ ) H x_3=(1-\epsilon)H x 3 = ( 1 − ϵ ) H , and multiply by a factor of 2 we obtain the dynamics of the particle separation r r r in a form similar to the no-slip result (53) , but with a different power law exponent in the denominator,
r ˙ = − F 2 π μ R r R 2 ( r 2 + 4 R 2 ) 3 / 2 , (55) \dot{r}=-\frac{F}{2\pi\mu R}\frac{rR^2}{\left(r^2+4R^2\right)^{3/2}}, \tag*{(55)} r ˙ = − 2 π μ R F ( r 2 + 4 R 2 ) 3/2 r R 2 , (55)
where R = ϵ H R=\epsilon H R = ϵH . In each of (53) and (55) we can identify an effective potential energy V ( r ) V(r) V ( r ) such that r ˙ = − d V / d r \dot{r}=-\mathrm{d}V/\mathrm{d}r r ˙ = − d V / d r . A natural question is how the result (55) for a stress-free surface is modified in the geometry of a Petri dish. The three lengths R = ϵ H R=\epsilon H R = ϵH , H H H , and r r r must be compared to determine the appropriate asymptotic regime.
The dynamics (55) holds for r ≪ H r\ll H r ≪ H but without restriction on the relative sizes of r r r and R R R , except that the impenetrability of the colonies implies that this expression is only relevant for r > 2 R r>2R r > 2 R . Of course, the validity of the singularity approach itself will decrease for r ∼ R = ϵ H r\sim R=\epsilon H r ∼ R = ϵH , and thus it is fair to assert that (55) is physically valid for ϵ H ≪ r ≪ H \epsilon H\ll r\ll H ϵH ≪ r ≪ H , and in particular r ˙ ∼ ( F / 2 π μ R ) ( R / r ) 2 \dot{r}\sim(F/2\pi\mu R)(R/r)^2 r ˙ ∼ ( F /2 π μ R ) ( R / r ) 2 for r ≫ R r\gg R r ≫ R . Indeed, as a consistency check, the full integral expressions do indeed simplify to (55) in this limit as we now show. Working in the same horizontal plane as the singularity, after some contour integration the repeated reflection solution becomes
v α 3 = − x α ∫ 0 ∞ λ J 0 ( λ ρ ) cosh λ ( 1 − 2 h ) cosh λ d λ , = − 2 π x α ∑ n = 0 ∞ ( n + 1 2 ) sin 2 π ϵ ( n + 1 2 ) K 0 ( π ρ ( n + 1 2 ) ) . (56) \begin{aligned}
v_\alpha^3=-x_\alpha\int_0^\infty\lambda J_0(\lambda\rho)\frac{\cosh\lambda(1-2h)}{\cosh\lambda}\,\mathrm{d}\lambda, \\
&=-2\pi x_\alpha\sum_{n=0}^{\infty}\left(n+\frac{1}{2}\right)\sin2\pi\epsilon\left(n+\frac{1}{2}\right)K_0\left(\pi\rho\left(n+\frac{1}{2}\right)\right). \tag*{(56)}
\end{aligned} v α 3 = − x α ∫ 0 ∞ λ J 0 ( λ ρ ) cosh λ cosh λ ( 1 − 2 h ) d λ , = − 2 π x α n = 0 ∑ ∞ ( n + 2 1 ) sin 2 π ϵ ( n + 2 1 ) K 0 ( π ρ ( n + 2 1 ) ) . (56)
From (34) we find that for small ρ \rho ρ the auxiliary solution w α 3 w_\alpha^3 w α 3 is O ( x α ϵ / ρ 2 ) \mathcal{O}(x_\alpha\epsilon/\rho^2) O ( x α ϵ / ρ 2 ) and hence, for points with small ϵ \epsilon ϵ and ρ \rho ρ , the repeated reflection solution dominates the auxiliary solution. Expanding in powers of ϵ \epsilon ϵ we find
u α 3 = − 2 ϵ x α ρ 3 + 12 ϵ 3 x α ρ 5 + ⋯ , (57) u_\alpha^3=-\frac{2\epsilon x_\alpha}{\rho^3}+\frac{12\epsilon^3x_\alpha}{\rho^5}+\cdots, \tag*{(57)} u α 3 = − ρ 3 2 ϵ x α + ρ 5 12 ϵ 3 x α + ⋯ , (57)
a result that agrees precisely with an expansion in ϵ \epsilon ϵ and suitable nondimensionalization of (54) . The new regime of interest occurs when the separation r r r becomes comparable to or larger than the Petri dish depth H H H . Given for completeness in Appendix F, when r ≫ H r\gg H r ≫ H (ρ ≫ 1 \rho\gg1 ρ ≫ 1 ), the non-dimensional flow field u α 3 u_\alpha^3 u α 3 decays exponentially with an unusual sinusoidal form
u α 3 = A ϵ x α ρ 3 / 2 e − ρ y 1 / 2 sin ( x 1 ( ρ − ρ 0 ) / 2 ) , (58) u_\alpha^3=\frac{A\epsilon x_\alpha}{\rho^{3/2}}e^{-\rho y_1/2}\sin\left(x_1(\rho-\rho_0)/2\right), \tag*{(58)} u α 3 = ρ 3/2 A ϵ x α e − ρ y 1 /2 sin ( x 1 ( ρ − ρ 0 ) /2 ) , (58)
where z 1 = x 1 + i y 1 = 2.769 + 7.498 i z_1=x_1+\mathrm{i}y_1=2.769+7.498\mathrm{i} z 1 = x 1 + i y 1 = 2.769 + 7.498 i is the first root in the first quadrant to the equation sinh z 1 = z 1 \sinh z_1=z_1 sinh z 1 = z 1 , A = 38.340 A=38.340 A = 38.340 and ρ 0 = 0.298 \rho_0=0.298 ρ 0 = 0.298 . Figure 7 (a) explores this further, demonstrating how numerical solutions to the full flow field vary as a function of ρ \rho ρ for a range of values of h h h . Darker blue dots denote larger values of h h h i.e. the Volvox are closer to the free surface. For comparison, the asymptotic result (58) is superimposed on those numerical results. For clarity, all velocities are normalised by ( ρ / ϵ ) e ρ y 1 / 2 (\sqrt{\rho}/\epsilon)e^{\rho y_1/2} ( ρ / ϵ ) e ρ y 1 /2 to highlight the sinusoidal component of the flow field. As can be seen, the asymptotic result is a good fit for ρ ≳ 2 \rho\gtrsim2 ρ ≳ 2 , improving as ρ \rho ρ increases and as h → 1 h\to1 h → 1 .
Figure 7. The lateral flow leading to hydrodynamic bound states. (a) Numerically obtained horizontal fluid velocity field u 1 , n o r m 3 = ( ρ / ϵ ) e ρ y 1 / 2 u 1 3 u_{1,\mathrm{norm}}^{3}=(\sqrt{\rho}/\epsilon)e^{\rho y_{1}/2}u_{1}^{3} u 1 , norm 3 = ( ρ / ϵ ) e ρ y 1 /2 u 1 3 , normalized to highlight the asymptotical sinusoidal component, generated by a vertically orientated Stokeslet placed at ( 0 , 0 , H − R ) (0,0,H-R) ( 0 , 0 , H − R ) and evaluated as a function of ρ \rho ρ at the point ( ρ , 0 , h ) (\rho,0,h) ( ρ , 0 , h ) . Here, R / H ∈ [ 0.15 , 0.1 , 0.05 , 0.01 ] R/H \in[0.15,0.1,0.05,0.01] R / H ∈ [ 0.15 , 0.1 , 0.05 , 0.01 ] with darker shades of blue denoting smaller values of R / H R/H R / H . The similarly scaled asymptotic result (58) is shown as the solid line. (b) The velocity u 1 3 u_{1}^{3} u 1 3 as a function of r / R r/R r / R . Here, R / H ∈ [ 0.3 , 0.2 , 0.15 , 0.1 , 0.01 ] R/H \in[0.3,0.2,0.15,0.1,0.01] R / H ∈ [ 0.3 , 0.2 , 0.15 , 0.1 , 0.01 ] with darker shades of blue denoting smaller values of R / H R/H R / H . For comparison the asymptotic result (54) for an infinitely deep Petri dish is shown as the solid black line.
An interesting feature of the screened interaction is that the multiplicative power law ρ − 1 / 2 \rho^{-1/2} ρ − 1/2 differs from that underlying the unscreened form (55) , which falls off as ρ − 2 \rho^{-2} ρ − 2 . This is unlike the case in electrostatics, for example, where a screened Coulomb interaction in three dimensions decays as ∼ ( 1 / r ) e − r / λ \sim(1/r)e^{-r/\lambda} ∼ ( 1/ r ) e − r / λ , where λ \lambda λ is the screening length, and the unscreened interaction is ∼ 1 / r \sim1/r ∼ 1/ r . In the present case, the reason why we see a transition as r r r increases is that for small r r r the first reflection from the repeated reflection solution dominates, but as r r r increases the auxiliary solution generates terms that cancel out the repeated reflection solution, thus leaving lower order terms in the auxiliary solution to dominate, giving rise to an exponential decay.
Figure 7 (b) shows in a semilogarithmic plot the lateral fluid velocity u 1 3 u_{1}^{3} u 1 3 as a function of the dimensionless radial distance r / R r/R r / R for various values of R / H R/H R / H . The exponential cutoff of the power-law result (55) is evident. Even for the relatively large Petri dish depth H / R = 10 H/R=10 H / R = 10 the velocity is attenuated by many orders of magnitude relative to the unscreened case for r / R ∼ 8 r/R\sim8 r / R ∼ 8 , long before the sign oscillations are visible. Thus, while the corresponding evolution equation for the infalling of two colonies inherits the sign oscillations of the flow field (58) , they appear only in the limit of very strong vertical confinement. The screening would, however, lead to very marked slowing down of the infalling trajectories relative to the infinite-depth case, and additionally reduce the significance of further-neighbor flows on a given swimmer in dense surface aggregates.
Conclusion
In this paper we have comprehensively explored the flows induced when Stokes singularities are placed in a Petri dish configuration, namely in a fluid layer with a bottom no-slip boundary and a top free surface boundary. In particular, we have derived both exact integral expression and far-field approximations for the flow generated by the six primary Stokes singularities: the Stokeslet, the rotlet, the source, stresslet, rotlet dipole and source dipole. Since all Stokes singularities can be expressed as derivatives of these six singularities, we can thus can gain insight about more general flows generated in a Petri dish by particles whose free space swimming fluid velocity can be represented as a sum of Stokes singularities. In particular, since the leading order contribution to the fluid velocity for these flows is separable in z z z , the full three dimensional Stokes equations can be vertically averaged to yield a much simpler two dimensional Brinkman equation much more amenable to analytic progress. A good example of this technique in action is [23 ] , where the authors modeled a circular mill as a rotlet dipole, generating a radially exponentially decaying flow with z z z dependence sin ( π z / 2 ) \sin(\pi z/2) sin ( π z /2 ) , and then solve the resulting Brinkman equation in the limit that the circular mill is away from the centre of the Petri dish by transforming to bipolar coordinates. We expect similar simplifications to hold in the many contexts in which experiments are carried out in the geometry of a Petri dish.
Acknowledgments
This work was supported in part by the Engineering and Physical Sciences Research Council, through a Doctoral Training Fellowship (GTF), by EPSRC grant EP/W024012/1 (EL,GTF), the European Research Council under the European Union’s Horizon 2020 Research and Innovation Programme (Grant No. 682754, EL), EPSRC Established Career Fellowship EP/M017982/1, Grant No. 7523 from the Marine Microbiology Initiative of the Gordon and Betty Moore Foundation, and the John Templeton Foundation (REG)
Appendices Integral Notation and Higher Order Repeated Reflections Solutions
For clarity in Appendices B-E below, we define the functions F m , n F_{m,n} F m , n and G m , n G_{m,n} G m , n
F m , n = ∫ 0 ∞ d λ λ m J n ( λ ρ ) cosh λ { sinh h λ cosh ( 1 − z ) λ , z > h , sinh z λ cosh ( 1 − h ) λ , z < h . (59) F_{m,n}=\int_{0}^{\infty}d\lambda\,\frac{\lambda^{m}J_n(\lambda\rho)}{\cosh\lambda}
\begin{cases}
\sinh h\lambda\cosh(1-z)\lambda, & z>h,\\
\sinh z\lambda\cosh(1-h)\lambda, & z<h.
\end{cases} \tag*{(59)} F m , n = ∫ 0 ∞ d λ cosh λ λ m J n ( λ ρ ) { sinh hλ cosh ( 1 − z ) λ , sinh z λ cosh ( 1 − h ) λ , z > h , z < h . (59)
G m , n = ∫ 0 ∞ d λ λ m J n ( λ ρ ) cosh λ { sinh h λ sinh ( 1 − z ) λ , z > h , − cosh z λ cosh ( 1 − h ) λ , z < h . (60) G_{m,n}=\int_{0}^{\infty}d\lambda\,\frac{\lambda^{m}J_n(\lambda\rho)}{\cosh\lambda}
\begin{cases}
\sinh h\lambda\sinh(1-z)\lambda, & z>h,\\
-\cosh z\lambda\cosh(1-h)\lambda, & z<h.
\end{cases} \tag*{(60)} G m , n = ∫ 0 ∞ d λ cosh λ λ m J n ( λ ρ ) { sinh hλ sinh ( 1 − z ) λ , − cosh z λ cosh ( 1 − h ) λ , z > h , z < h . (60)
Shown in more detail elsewhere [28 ] , these functions allow (10) in the main text to be extended to obtain repeated reflection solutions at third and fifth order,
L ( 1 r 3 ) = − 1 ρ ∂ ∂ ρ L ( 1 r ) = 2 F 1 , 1 ρ , L ( z r 3 ) = 2 G 1 , 0 , L ( z 2 r 3 ) = 2 ( F 0 , 0 − ρ F 1 , 1 ) . (61) \mathcal{L}\left(\frac{1}{r^3}\right)=-\frac{1}{\rho}\frac{\partial}{\partial\rho}\mathcal{L}\left(\frac{1}{r}\right)=\frac{2F_{1,1}}{\rho},\qquad
\mathcal{L}\left(\frac{z}{r^3}\right)=2G_{1,0},\qquad
\mathcal{L}\left(\frac{z^2}{r^3}\right)=2(F_{0,0}-\rho F_{1,1}). \tag*{(61)} L ( r 3 1 ) = − ρ 1 ∂ ρ ∂ L ( r 1 ) = ρ 2 F 1 , 1 , L ( r 3 z ) = 2 G 1 , 0 , L ( r 3 z 2 ) = 2 ( F 0 , 0 − ρ F 1 , 1 ) . (61)
L ( 1 r 5 ) = 2 ( 2 F 1 , 1 − ρ F 2 , 0 ) 3 ρ 3 , L ( z r 5 ) = 2 G 2 , 1 3 ρ , \mathcal{L}\left(\frac{1}{r^5}\right)=\frac{2(2F_{1,1}-\rho F_{2,0})}{3\rho^3},\qquad
\mathcal{L}\left(\frac{z}{r^5}\right)=\frac{2G_{2,1}}{3\rho}, L ( r 5 1 ) = 3 ρ 3 2 ( 2 F 1 , 1 − ρ F 2 , 0 ) , L ( r 5 z ) = 3 ρ 2 G 2 , 1 ,
L ( z 2 r 5 ) = 2 ( ρ F 2 , 0 + F 1 , 1 ) 3 ρ , L ( z 3 r 5 ) = 2 ( 3 G 1 , 0 − ρ G 2 , 1 ) 3 . (62) \mathcal{L}\left(\frac{z^2}{r^5}\right)=\frac{2(\rho F_{2,0}+F_{1,1})}{3\rho},\qquad
\mathcal{L}\left(\frac{z^3}{r^5}\right)=\frac{2(3G_{1,0}-\rho G_{2,1})}{3}. \tag*{(62)} L ( r 5 z 2 ) = 3 ρ 2 ( ρ F 2 , 0 + F 1 , 1 ) , L ( r 5 z 3 ) = 3 2 ( 3 G 1 , 0 − ρ G 2 , 1 ) . (62)
Rotlet in a Petri Dish
The approach for a rotlet ( ϵ j k p x p / r 3 ) (\epsilon_{jkp}x_p/r^3) ( ϵ j k p x p / r 3 ) follows the procedure for the Stokeslet, with a repeated reflection solution
v j k = L ( ϵ j k p x p r 3 ) = x α ϵ j k α L ( 1 r 3 ) + ϵ j k 3 L ( z r 3 ) = 2 ϵ j k α x α F 1 , 1 ρ + 2 ϵ j k 3 G 1 , 0 , (63) v_j^k=\mathcal{L}\left(\epsilon_{jkp}\frac{x_p}{r^3}\right)=x_\alpha\epsilon_{jk\alpha}\mathcal{L}\left(\frac{1}{r^3}\right)+\epsilon_{jk3}\mathcal{L}\left(\frac{z}{r^3}\right)=\frac{2\epsilon_{jk\alpha}x_\alpha F_{1,1}}{\rho}+2\epsilon_{jk3}G_{1,0}, \tag*{(63)} v j k = L ( ϵ j k p r 3 x p ) = x α ϵ j k α L ( r 3 1 ) + ϵ j k 3 L ( r 3 z ) = ρ 2 ϵ j k α x α F 1 , 1 + 2 ϵ j k 3 G 1 , 0 , (63)
with the summation convention for α ∈ [ 1 , 2 ] \alpha\in[1,2] α ∈ [ 1 , 2 ] . The boundary conditions for the auxiliary solution w j k w_j^k w j k and transformed auxiliary solution w ^ j k \hat{w}_j^k w ^ j k become
w j k ∣ z = 0 = 2 ϵ j k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh ( 1 − h ) λ cosh λ ⟹ w ^ j k ∣ z = 0 = 4 π ϵ j k 3 cosh k ( 1 − h ) cosh k , (64) \left.w_j^k\right|_{z=0}=2\epsilon_{jk3}\int_0^\infty\lambda\,d\lambda\,J_0(\lambda\rho)\frac{\cosh(1-h)\lambda}{\cosh\lambda}\Longrightarrow\left.\hat{w}_j^k\right|_{z=0}=4\pi\epsilon_{jk3}\frac{\cosh k(1-h)}{\cosh k}, \tag*{(64)} w j k z = 0 = 2 ϵ j k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ cosh ( 1 − h ) λ ⟹ w ^ j k z = 0 = 4 π ϵ j k 3 cosh k cosh k ( 1 − h ) , (64)
∂ w α k ∂ z ∣ z = 1 = 2 ϵ α k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) λ sinh h λ cosh λ ⟹ ∂ w ^ α k ∂ z ∣ z = 1 = 4 π ϵ α k 3 k sinh h k cosh k , (65) \left.\frac{\partial w_\alpha^k}{\partial z}\right|_{z=1}=2\epsilon_{\alpha k3}\int_0^\infty\lambda\,d\lambda\,J_0(\lambda\rho)\frac{\lambda\sinh h\lambda}{\cosh\lambda}\Longrightarrow\left.\frac{\partial\hat{w}_\alpha^k}{\partial z}\right|_{z=1}=4\pi\epsilon_{\alpha k3}\frac{k\sinh hk}{\cosh k}, \tag*{(65)} ∂ z ∂ w α k z = 1 = 2 ϵ α k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ λ sinh hλ ⟹ ∂ z ∂ w ^ α k z = 1 = 4 π ϵ α k 3 cosh k k sinh hk , (65)
w 3 k ∣ z = 1 = − 2 ϵ 3 k α x α ρ ∫ 0 ∞ λ d λ J 1 ( λ ρ ) sinh h λ cosh λ ⟹ w ^ 3 k ∣ z = 1 = 4 π i k α ϵ 3 k α sinh h k k cosh k , (66) \left.w_3^k\right|_{z=1}=-\frac{2\epsilon_{3k\alpha}x_\alpha}{\rho}\int_0^\infty\lambda\,d\lambda\,J_1(\lambda\rho)\frac{\sinh h\lambda}{\cosh\lambda}\Longrightarrow\left.\hat{w}_3^k\right|_{z=1}=4\pi i k_\alpha\epsilon_{3k\alpha}\frac{\sinh hk}{k\cosh k}, \tag*{(66)} w 3 k z = 1 = − ρ 2 ϵ 3 k α x α ∫ 0 ∞ λ d λ J 1 ( λ ρ ) cosh λ sinh hλ ⟹ w ^ 3 k z = 1 = 4 π i k α ϵ 3 k α k cosh k sinh hk , (66)
When k = 3 k=3 k = 3 the boundary conditions are zero and w ^ j 3 = w j 3 = 0 \hat{w}_j^3=w_j^3=0 w ^ j 3 = w j 3 = 0 . When k = α ∈ [ 1 , 2 ] k=\alpha\in[1,2] k = α ∈ [ 1 , 2 ] , we find
w ^ 3 α = 4 π k γ i ϵ γ α 3 k cosh k ( sinh 2 k − 2 k ) ( 2 k cosh k ( 1 − h ) sinh k z − k z cosh k z sinh k ( 1 − h ) − k z sinh k ( 1 − h ) cosh k ( 2 − z ) − 2 cosh k sinh h k sinh k z ) , (67) \hat{w}_3^\alpha=\frac{4\pi k\gamma_i\epsilon_{\gamma\alpha3}}{k\cosh k(\sinh2k-2k)}\left(2k\cosh k(1-h)\sinh kz-kz\cosh kz\sinh k(1-h)-kz\sinh k(1-h)\cosh k(2-z)-2\cosh k\sinh hk\sinh kz\right), \tag*{(67)} w ^ 3 α = k cosh k ( sinh 2 k − 2 k ) 4 π k γ i ϵ γ α 3 ( 2 k cosh k ( 1 − h ) sinh k z − k z cosh k z sinh k ( 1 − h ) − k z sinh k ( 1 − h ) cosh k ( 2 − z ) − 2 cosh k sinh hk sinh k z ) , (67)
w ^ α α = 4 π k β k α ϵ β α 3 k cosh k ( sinh 2 k − 2 k ) ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) , (68) \hat{w}_\alpha^\alpha=\frac{4\pi k\beta_k\alpha\epsilon_{\beta\alpha3}}{k\cosh k(\sinh2k-2k)}\left(z\sinh k(h-1)\sinh k(2-z)+(2-z)\sinh(kz)\sinh k(h-1)\right), \tag*{(68)} w ^ α α = k cosh k ( sinh 2 k − 2 k ) 4 π k β k α ϵ β α 3 ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) , (68)
w ^ β α = 4 π k β 2 ϵ β α 3 k cosh k ( sinh 2 k − 2 k ) ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) + 4 π ϵ β α 3 cosh k ( sinh 2 k − 2 k ) ( sinh 2 k cosh k ( h + z − 1 ) − 2 k cosh k ( h + z − 1 ) ) , (69) \hat{w}_\beta^\alpha=\frac{4\pi k_\beta^2\epsilon_{\beta\alpha3}}{k\cosh k(\sinh2k-2k)}\left(z\sinh k(h-1)\sinh k(2-z)+(2-z)\sinh(kz)\sinh k(h-1)\right)+\frac{4\pi\epsilon_{\beta\alpha3}}{\cosh k(\sinh2k-2k)}\left(\sinh2k\cosh k(h+z-1)-2k\cosh k(h+z-1)\right), \tag*{(69)} w ^ β α = k cosh k ( sinh 2 k − 2 k ) 4 π k β 2 ϵ β α 3 ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) + cosh k ( sinh 2 k − 2 k ) 4 π ϵ β α 3 ( sinh 2 k cosh k ( h + z − 1 ) − 2 k cosh k ( h + z − 1 ) ) , (69)
where β ∈ [ 1 , 2 ] \beta\in[1,2] β ∈ [ 1 , 2 ] and β ≠ α \beta\ne\alpha β = α . Rewriting inverse Fourier transforms in terms of Hankel transforms, we find
v 3 α = i x α 2 π ρ H 1 ( k k α w ^ 3 α ) , w α α = − x α x β π ρ 3 H 1 ( k k α k β w ^ α α ) + x α x β 2 π ρ 2 H 0 ( k 2 k α k β w ^ α α ) , (70) v_3^\alpha=\frac{ix_\alpha}{2\pi\rho}\mathcal{H}_1\left(\frac{k}{k_\alpha}\hat{w}_3^\alpha\right),\qquad w_\alpha^\alpha=-\frac{x_\alpha x_\beta}{\pi\rho^3}\mathcal{H}_1\left(\frac{k}{k_\alpha k_\beta}\hat{w}_\alpha^\alpha\right)+\frac{x_\alpha x_\beta}{2\pi\rho^2}\mathcal{H}_0\left(\frac{k^2}{k_\alpha k_\beta}\hat{w}_\alpha^\alpha\right), \tag*{(70)} v 3 α = 2 π ρ i x α H 1 ( k α k w ^ 3 α ) , w α α = − π ρ 3 x α x β H 1 ( k α k β k w ^ α α ) + 2 π ρ 2 x α x β H 0 ( k α k β k 2 w ^ α α ) , (70)
w β α = 1 2 π ρ ( 1 − 2 x β 2 ρ 2 ) H 1 ( k w ^ 1 ) + x β 2 2 π ρ 2 H 0 ( k 2 w ^ 1 ) + 1 2 π H 0 ( w ^ 0 ) , (71) w_\beta^\alpha=\frac{1}{2\pi\rho}\left(1-\frac{2x_\beta^2}{\rho^2}\right)\mathcal{H}_1(k\hat{w}_1)+\frac{x_\beta^2}{2\pi\rho^2}\mathcal{H}_0(k^2\hat{w}_1)+\frac{1}{2\pi}\mathcal{H}_0(\hat{w}_0), \tag*{(71)} w β α = 2 π ρ 1 ( 1 − ρ 2 2 x β 2 ) H 1 ( k w ^ 1 ) + 2 π ρ 2 x β 2 H 0 ( k 2 w ^ 1 ) + 2 π 1 H 0 ( w ^ 0 ) , (71)
where β ∈ [ 1 , 2 ] \beta\in[1,2] β ∈ [ 1 , 2 ] , β ≠ α \beta\ne\alpha β = α and for notational simplicity we have decomposed w β α w_\beta^\alpha w β α as w ^ β α = w ^ 0 + k β 2 w ^ 1 \hat{w}_\beta^\alpha=\hat{w}_0+k_\beta^2\hat{w}_1 w ^ β α = w ^ 0 + k β 2 w ^ 1 . Using contour integration, as with the Stokeslet we find F F F has poles of order 1 at both z = π i ( n + 1 / 2 ) z=\pi i(n+1/2) z = π i ( n + 1/2 ) ( n ∈ Z ≥ ) (n\in\mathbb{Z}_{\geq}) ( n ∈ Z ≥ ) and z = z 0 / 2 z=z_0/2 z = z 0 /2 where z 0 z_0 z 0 satisfies sinh z 0 = z 0 \sinh z_0=z_0 sinh z 0 = z 0 . When k = 3 k=3 k = 3 , w ^ j 3 = w j 3 = 0 \hat{w}_j^3=w_j^3=0 w ^ j 3 = w j 3 = 0 , and thus the flow field u j 3 u_j^3 u j 3 satisfies
u j 3 = v j 3 = 2 π x α ϵ j 3 α ρ ∑ n = 1 , 3 , 5 , … n sin ( n π h 2 ) sin ( n π z 2 ) K 1 ( n π ρ 2 ) . (72) u_j^3=v_j^3=\frac{2\pi x_\alpha\epsilon_{j3\alpha}}{\rho}\sum_{n=1,3,5,\ldots}n\sin\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)K_1\left(\frac{n\pi\rho}{2}\right). \tag*{(72)} u j 3 = v j 3 = ρ 2 π x α ϵ j 3 α n = 1 , 3 , 5 , … ∑ n sin ( 2 nπ h ) sin ( 2 nπ z ) K 1 ( 2 nπ ρ ) . (72)
Hence in the far-field, the leading order contribution decays exponentially as
u j 3 = O ( ϵ j 3 α x α e − ρ π / 2 ρ 3 / 2 ) . (73) u_j^3=\mathcal{O}\left(\frac{\epsilon_{j3\alpha}x_\alpha e^{-\rho\pi/2}}{\rho^{3/2}}\right). \tag*{(73)} u j 3 = O ( ρ 3/2 ϵ j 3 α x α e − ρ π /2 ) . (73)
Since ∫ γ ϵ \int_{\gamma_\epsilon} ∫ γ ϵ vanishes as ϵ → 0 \epsilon\to0 ϵ → 0 , when j = 3 j=3 j = 3 and k = α k=\alpha k = α where α ∈ [ 1 , 2 ] \alpha\in[1,2] α ∈ [ 1 , 2 ] (36) simplifies to
w 3 α = 2 π x α γ ϵ γ 3 ρ ∑ n = 1 , 3 , 5 , … ∞ n sin ( n π h 2 ) sin ( n π z 2 ) K 1 ( n π ρ 2 ) − ∑ z 0 ∈ H : z 0 = sinh z 0 x α z 0 H 1 1 ( ρ z 0 2 ) 8 π ρ ( cosh z 0 − 1 ) ( ( sinh 2 k − 2 k ) w ^ 3 α k α ) ∣ k = z 0 / 2 (74) w_3^\alpha= \frac{2\pi x_\alpha\gamma_\epsilon\gamma_3}{\rho}\sum_{n=1,3,5,\ldots}^{\infty} n\sin\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)K_1\left(\frac{n\pi\rho}{2}\right)
-
\sum_{\substack{z_0\in\mathbb{H}:\\ z_0=\sinh z_0}}\frac{x_\alpha z_0 H_1^1\left(\frac{\rho z_0}{2}\right)}{8\pi\rho(\cosh z_0-1)}
\left(\left(\sinh2k-2k\right)\frac{\widehat{w}_3^\alpha}{k_\alpha}\right)\bigg|_{k=z_0/2} \tag*{(74)} w 3 α = ρ 2 π x α γ ϵ γ 3 n = 1 , 3 , 5 , … ∑ ∞ n sin ( 2 nπ h ) sin ( 2 nπ z ) K 1 ( 2 nπ ρ ) − z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 π ρ ( cosh z 0 − 1 ) x α z 0 H 1 1 ( 2 ρ z 0 ) ( ( sinh 2 k − 2 k ) k α w 3 α ) k = z 0 /2 (74)
u 3 α = v 3 α + w 3 α = − ∑ z 0 ∈ H : z 0 = sinh z 0 x α z 0 H 1 1 ( ρ z 0 2 ) 8 π ρ ( cosh z 0 − 1 ) ( ( sinh 2 k − 2 k ) w ^ 3 α k α ) ∣ k = z 0 / 2 , (75) u_3^\alpha=v_3^\alpha+w_3^\alpha
=
-
\sum_{\substack{z_0\in\mathbb{H}:\\ z_0=\sinh z_0}}\frac{x_\alpha z_0 H_1^1\left(\frac{\rho z_0}{2}\right)}{8\pi\rho(\cosh z_0-1)}
\left(\left(\sinh2k-2k\right)\frac{\widehat{w}_3^\alpha}{k_\alpha}\right)\bigg|_{k=z_0/2}, \tag*{(75)} u 3 α = v 3 α + w 3 α = − z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 π ρ ( cosh z 0 − 1 ) x α z 0 H 1 1 ( 2 ρ z 0 ) ( ( sinh 2 k − 2 k ) k α w 3 α ) k = z 0 /2 , (75)
The contribution from poles of order 1 at z = π i ( n + 1 / 2 ) z=\pi i(n+1/2) z = π i ( n + 1/2 ) , n ∈ Z ≥ n\in\mathbb{Z}_{\geq} n ∈ Z ≥ cancels out with v 3 α v_3^\alpha v 3 α , yielding
u 3 α = O ( ϵ γ α γ 3 x γ e − ρ y 1 / 2 ρ 3 / 2 ) . (76) u_3^\alpha=\mathcal{O}\left(\frac{\epsilon\gamma_\alpha\gamma_3x_\gamma e^{-\rho y_1/2}}{\rho^{3/2}}\right). \tag*{(76)} u 3 α = O ( ρ 3/2 ϵ γ α γ 3 x γ e − ρ y 1 /2 ) . (76)
Finally, when j , k ∈ [ 1 , 2 ] j,k\in[1,2] j , k ∈ [ 1 , 2 ] , the leading order contribution in the far-field arises from γ ϵ \gamma_\epsilon γ ϵ i.e.
w α α = z ( 2 − z ) [ ϵ β 3 6 x α x β ( 1 − h ) ρ 4 ] , w β α = z ( 2 − z ) [ − ϵ β 3 3 ( 1 − h ) ρ 2 ( 1 − 2 x β 2 ρ 2 ) ] , (77) w_\alpha^\alpha=z(2-z)\left[\epsilon\beta3\frac{6x_\alpha x_\beta(1-h)}{\rho^4}\right],\qquad
w_\beta^\alpha=z(2-z)\left[-\epsilon\beta3\frac{3(1-h)}{\rho^2}\left(1-\frac{2x_\beta^2}{\rho^2}\right)\right], \tag*{(77)} w α α = z ( 2 − z ) [ ϵ β 3 ρ 4 6 x α x β ( 1 − h ) ] , w β α = z ( 2 − z ) [ − ϵ β 3 ρ 2 3 ( 1 − h ) ( 1 − ρ 2 2 x β 2 ) ] , (77)
where β ∈ [ 1 , 2 ] \beta\in[1,2] β ∈ [ 1 , 2 ] and β ≠ α \beta\ne\alpha β = α .
Stresslet in a Petri Dish
While the most general stresslet form is { x j x k x l / r 5 } \{x_jx_kx_l/r^5\} { x j x k x l / r 5 } , for swimming microorganisms typically k = l k=l k = l . From fifth order repeated reflection solutions, v j k , l v_j^{k,l} v j k , l for a stresslet is
v j k , l = δ j α δ k β δ l δ x α x β x δ L ( 1 r 5 ) + δ j 3 δ k 3 δ l 3 L ( z 3 r 5 ) + ( δ j α δ k β δ l 3 + δ j α δ k 3 δ l β + δ j 3 δ k α δ l β ) x α x β L ( z r 5 ) + ( δ j α δ k 3 δ l 3 + δ j 3 δ k α δ l 3 + δ j 3 δ k 3 δ l α ) x α L ( z 2 r 5 ) = 2 x α F 1 , 1 3 ρ ( δ j α δ k 3 δ l 3 + δ j 3 δ k α δ l 3 + δ j 3 δ k 3 δ l α + 2 x β x δ ρ 2 δ j α δ k β δ l δ ) + 2 x α F 2 , 0 3 ( δ j α δ k 3 δ l 3 + δ j 3 δ k α δ l 3 + δ j 3 δ k 3 δ l α − x β x δ ρ 2 δ j α δ k β δ l δ ) + 2 ρ G 2 , 1 3 ( x α x β ρ 2 ( δ j α δ k β δ l 3 + δ j α δ k 3 δ l β + δ j 3 δ k α δ l β ) − δ j 3 δ k 3 δ l 3 ) + 2 δ j 3 δ k 3 δ l 3 G 1 , 0 , (78) \begin{aligned}
v_j^{k,l}
=
\delta_{j\alpha}\delta_{k\beta}\delta_{l\delta}x_\alpha x_\beta x_\delta\mathcal{L}\left(\frac{1}{r^5}\right)
+\delta_{j3}\delta_{k3}\delta_{l3}\mathcal{L}\left(\frac{z^3}{r^5}\right)
+\left(\delta_{j\alpha}\delta_{k\beta}\delta_{l3}+\delta_{j\alpha}\delta_{k3}\delta_{l\beta}+\delta_{j3}\delta_{k\alpha}\delta_{l\beta}\right)x_\alpha x_\beta\mathcal{L}\left(\frac{z}{r^5}\right) \\
&+\left(\delta_{j\alpha}\delta_{k3}\delta_{l3}+\delta_{j3}\delta_{k\alpha}\delta_{l3}+\delta_{j3}\delta_{k3}\delta_{l\alpha}\right)x_\alpha\mathcal{L}\left(\frac{z^2}{r^5}\right) \\
&=
\frac{2x_\alpha F_{1,1}}{3\rho}\left(\delta_{j\alpha}\delta_{k3}\delta_{l3}+\delta_{j3}\delta_{k\alpha}\delta_{l3}+\delta_{j3}\delta_{k3}\delta_{l\alpha}+\frac{2x_\beta x_\delta}{\rho^2}\delta_{j\alpha}\delta_{k\beta}\delta_{l\delta}\right) \\
&+\frac{2x_\alpha F_{2,0}}{3}\left(\delta_{j\alpha}\delta_{k3}\delta_{l3}+\delta_{j3}\delta_{k\alpha}\delta_{l3}+\delta_{j3}\delta_{k3}\delta_{l\alpha}-\frac{x_\beta x_\delta}{\rho^2}\delta_{j\alpha}\delta_{k\beta}\delta_{l\delta}\right) \\
&+\frac{2\rho G_{2,1}}{3}\left(\frac{x_\alpha x_\beta}{\rho^2}\left(\delta_{j\alpha}\delta_{k\beta}\delta_{l3}+\delta_{j\alpha}\delta_{k3}\delta_{l\beta}+\delta_{j3}\delta_{k\alpha}\delta_{l\beta}\right)-\delta_{j3}\delta_{k3}\delta_{l3}\right)+2\delta_{j3}\delta_{k3}\delta_{l3}G_{1,0}, \tag*{(78)}
\end{aligned} v j k , l = δ j α δ k β δ l δ x α x β x δ L ( r 5 1 ) + δ j 3 δ k 3 δ l 3 L ( r 5 z 3 ) + ( δ j α δ k β δ l 3 + δ j α δ k 3 δ l β + δ j 3 δ k α δ l β ) x α x β L ( r 5 z ) + ( δ j α δ k 3 δ l 3 + δ j 3 δ k α δ l 3 + δ j 3 δ k 3 δ l α ) x α L ( r 5 z 2 ) = 3 ρ 2 x α F 1 , 1 ( δ j α δ k 3 δ l 3 + δ j 3 δ k α δ l 3 + δ j 3 δ k 3 δ l α + ρ 2 2 x β x δ δ j α δ k β δ l δ ) + 3 2 x α F 2 , 0 ( δ j α δ k 3 δ l 3 + δ j 3 δ k α δ l 3 + δ j 3 δ k 3 δ l α − ρ 2 x β x δ δ j α δ k β δ l δ ) + 3 2 ρ G 2 , 1 ( ρ 2 x α x β ( δ j α δ k β δ l 3 + δ j α δ k 3 δ l β + δ j 3 δ k α δ l β ) − δ j 3 δ k 3 δ l 3 ) + 2 δ j 3 δ k 3 δ l 3 G 1 , 0 , (78)
where { α , β , δ } ∈ [ 1 , 2 ] \{\alpha,\beta,\delta\}\in[1,2] { α , β , δ } ∈ [ 1 , 2 ] . The boundary conditions for the transformed auxiliary solution w ^ j k , l \widehat{w}_j^{k,l} w j k , l simplify to
w ^ j k , l ∣ z = 0 = 4 π 3 ( δ j α δ k β δ l 3 + δ j α δ k 3 δ l β + δ j 3 δ k α δ l β ) ( δ α β cosh k ( 1 − h ) cosh k + k α k β k ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) + 4 π 3 δ j 3 δ k 3 δ l 3 ( cosh k ( 1 − h ) cosh k − k ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) , (79) \begin{aligned}
\left.\widehat{w}_j^{k,l}\right|_{z=0}
=
\frac{4\pi}{3}\left(\delta_{j\alpha}\delta_{k\beta}\delta_{l3}+\delta_{j\alpha}\delta_{k3}\delta_{l\beta}+\delta_{j3}\delta_{k\alpha}\delta_{l\beta}\right)
\left(\delta_{\alpha\beta}\frac{\cosh k(1-h)}{\cosh k}+\frac{k_\alpha k_\beta}{k}\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right) \\
&+\frac{4\pi}{3}\delta_{j3}\delta_{k3}\delta_{l3}\left(\frac{\cosh k(1-h)}{\cosh k}-k\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right), \tag*{(79)}
\end{aligned} w j k , l z = 0 = 3 4 π ( δ j α δ k β δ l 3 + δ j α δ k 3 δ l β + δ j 3 δ k α δ l β ) ( δ α β cosh k cosh k ( 1 − h ) + k k α k β ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) + 3 4 π δ j 3 δ k 3 δ l 3 ( cosh k cosh k ( 1 − h ) − k ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) , (79)
∂ w ^ α k , l ∂ z ∣ z = 1 = 4 π 3 ( δ k β δ l 3 + δ k 3 δ l β ) ( k δ α β sinh h k cosh k + k α k β k ∂ ∂ k ( k sinh h k cosh k ) ) , (80) \left.\frac{\partial\widehat{w}_\alpha^{k,l}}{\partial z}\right|_{z=1}
=
\frac{4\pi}{3}\left(\delta_{k\beta}\delta_{l3}+\delta_{k3}\delta_{l\beta}\right)
\left(k\delta_{\alpha\beta}\frac{\sinh hk}{\cosh k}+\frac{k_\alpha k_\beta}{k}\frac{\partial}{\partial k}\left(k\frac{\sinh hk}{\cosh k}\right)\right), \tag*{(80)} ∂ z ∂ w α k , l z = 1 = 3 4 π ( δ k β δ l 3 + δ k 3 δ l β ) ( k δ α β cosh k sinh hk + k k α k β ∂ k ∂ ( k cosh k sinh hk ) ) , (80)
w ^ 3 k , l ∣ z = 1 = − 4 π i k α 3 ( δ k α δ l 3 + δ k 3 δ l α ) ∂ ∂ k ( sinh h k cosh k ) . (81) \left.\widehat{w}_3^{k,l}\right|_{z=1}
=
-\frac{4\pi i k_\alpha}{3}\left(\delta_{k\alpha}\delta_{l3}+\delta_{k3}\delta_{l\alpha}\right)\frac{\partial}{\partial k}\left(\frac{\sinh hk}{\cosh k}\right). \tag*{(81)} w 3 k , l z = 1 = − 3 4 π i k α ( δ k α δ l 3 + δ k 3 δ l α ) ∂ k ∂ ( cosh k sinh hk ) . (81)
As in the main text for a source, we can thus solve for w ^ j k , l \hat{w}_{j}^{k,l} w ^ j k , l to give
w ^ 3 3 , 3 = 4 π 3 ( sinh 2 k − 2 k ) ( cosh k ( 1 − h ) cosh k − k ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) × ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) , \begin{aligned}
\hat{w}_{3}^{3,3}=\frac{4\pi}{3(\sinh2k-2k)}\left(\frac{\cosh k(1-h)}{\cosh k}-k\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right) \\
&\times\left(k(z-2)\cosh kz+kz\cosh k(2-z)+\sinh k(2-z)-\sinh kz\right),
\end{aligned} w ^ 3 3 , 3 = 3 ( sinh 2 k − 2 k ) 4 π ( cosh k cosh k ( 1 − h ) − k ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) × ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) ,
w ^ α 3 , 3 = 4 π k α i 3 ( sinh 2 k − 2 k ) ( cosh k ( 1 − h ) cosh k − k ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) × ( ( z − 2 ) sinh k z − z sinh k ( 2 − z ) ) , (82) \begin{aligned}
\hat{w}_{\alpha}^{3,3}=\frac{4\pi k_{\alpha}i}{3(\sinh2k-2k)}\left(\frac{\cosh k(1-h)}{\cosh k}-k\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right) \\
&\times\left((z-2)\sinh kz-z\sinh k(2-z)\right), \tag*{(82)}
\end{aligned} w ^ α 3 , 3 = 3 ( sinh 2 k − 2 k ) 4 π k α i ( cosh k cosh k ( 1 − h ) − k ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) × ( ( z − 2 ) sinh k z − z sinh k ( 2 − z ) ) , (82)
w ^ 3 β , δ = 4 π 3 ( sinh 2 k − 2 k ) ( δ β δ cosh k ( 1 − h ) cosh k + k β k δ k ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) × ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) , (83) \begin{aligned}
\hat{w}_{3}^{\beta,\delta}=\frac{4\pi}{3(\sinh2k-2k)}\left(\delta_{\beta\delta}\frac{\cosh k(1-h)}{\cosh k}+\frac{k_{\beta}k_{\delta}}{k}\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right) \\
&\times\left(k(z-2)\cosh kz+kz\cosh k(2-z)+\sinh k(2-z)-\sinh kz\right), \tag*{(83)}
\end{aligned} w ^ 3 β , δ = 3 ( sinh 2 k − 2 k ) 4 π ( δ β δ cosh k cosh k ( 1 − h ) + k k β k δ ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) × ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) , (83)
w ^ α β , δ = 4 π k α i 3 ( sinh 2 k − 2 k ) ( δ β δ cosh k ( 1 − h ) cosh k + k β k δ k ∂ ∂ k ( cosh k ( 1 − h ) cosh k ) ) × ( ( z − 2 ) sinh k z − z sinh k ( 2 − z ) ) , (84) \begin{aligned}
\hat{w}_{\alpha}^{\beta,\delta}=\frac{4\pi k_{\alpha}i}{3(\sinh2k-2k)}\left(\delta_{\beta\delta}\frac{\cosh k(1-h)}{\cosh k}+\frac{k_{\beta}k_{\delta}}{k}\frac{\partial}{\partial k}\left(\frac{\cosh k(1-h)}{\cosh k}\right)\right) \\
&\times\left((z-2)\sinh kz-z\sinh k(2-z)\right), \tag*{(84)}
\end{aligned} w ^ α β , δ = 3 ( sinh 2 k − 2 k ) 4 π k α i ( δ β δ cosh k cosh k ( 1 − h ) + k k β k δ ∂ k ∂ ( cosh k cosh k ( 1 − h ) ) ) × ( ( z − 2 ) sinh k z − z sinh k ( 2 − z ) ) , (84)
w ^ 3 α , 3 = 8 π i k α 3 cosh 2 k ( sinh 2 k − 2 k ) ( h k z cosh 3 k cosh k ( 1 − h − z ) − h k cosh k sinh k z sinh k ( 1 − h ) \hat{w}_{3}^{\alpha,3}=\frac{8\pi i k_{\alpha}}{3\cosh^{2}k(\sinh2k-2k)}\left(hkz\cosh^{3}k\cosh k(1-h-z)-hk\cosh k\sinh kz\sinh k(1-h)\right. w ^ 3 α , 3 = 3 cosh 2 k ( sinh 2 k − 2 k ) 8 π i k α ( hk z cosh 3 k cosh k ( 1 − h − z ) − hk cosh k sinh k z sinh k ( 1 − h )
− k sinh h k sinh k z − k z cosh k sinh h k sinh k ( 1 − z ) − h k z cosh 2 k sinh h k sinh k z \left.-k\sinh hk\sinh kz-kz\cosh k\sinh hk\sinh k(1-z)-hkz\cosh^{2}k\sinh hk\sinh kz
\right. − k sinh hk sinh k z − k z cosh k sinh hk sinh k ( 1 − z ) − hk z cosh 2 k sinh hk sinh k z
− z cosh 2 k sinh k cosh k ( 1 − h − z ) + cosh k sinh k z cosh k ( 1 + h ) -z\cosh^{2}k\sinh k\cosh k(1-h-z)+\cosh k\sinh kz\cosh k(1+h) − z cosh 2 k sinh k cosh k ( 1 − h − z ) + cosh k sinh k z cosh k ( 1 + h )
− h cosh 2 k cosh h k sinh k z ) , (85) \left.-h\cosh^{2}k\cosh hk\sinh kz\right), \tag*{(85)} − h cosh 2 k cosh hk sinh k z ) , (85)
w ^ β α , 3 = 8 π k α k β 3 k cosh 2 k ( sinh 2 k − 2 k ) ( k sinh h k cosh k z − k z cosh k sinh h k cosh k ( 1 − z ) + h k cosh k cosh k z sinh k ( 1 − h ) + h k z cosh 2 k sinh h k cosh k z + h k z cosh 3 k sinh k ( 1 − h − z ) \begin{aligned}
\hat{w}_{\beta}^{\alpha,3}=\frac{8\pi k_{\alpha}k_{\beta}}{3k\cosh^{2}k(\sinh2k-2k)}(k\sinh hk\cosh kz-kz\cosh k\sinh hk\cosh k(1-z) \\
&+hk\cosh k\cosh kz\sinh k(1-h)+hkz\cosh^{2}k\sinh hk\cosh kz+hkz\cosh^{3}k\sinh k(1-h-z)
\end{aligned} w ^ β α , 3 = 3 k cosh 2 k ( sinh 2 k − 2 k ) 8 π k α k β ( k sinh hk cosh k z − k z cosh k sinh hk cosh k ( 1 − z ) + hk cosh k cosh k z sinh k ( 1 − h ) + hk z cosh 2 k sinh hk cosh k z + hk z cosh 3 k sinh k ( 1 − h − z )
− cosh k sinh k sinh k ( h + z ) − ( h + z ) cosh 2 k sinh k sinh k ( 1 − h − z ) ) + 4 π δ α β 3 cosh k cosh k ( 1 − h − z ) , (86) \begin{aligned}
-\cosh k\sinh k\sinh k(h+z)-(h+z)\cosh^{2}k\sinh k\sinh k(1-h-z)) \\
&+\frac{4\pi\delta_{\alpha\beta}}{3}\cosh k\cosh k(1-h-z), \tag*{(86)}
\end{aligned} − cosh k sinh k sinh k ( h + z ) − ( h + z ) cosh 2 k sinh k sinh k ( 1 − h − z )) + 3 4 π δ α β cosh k cosh k ( 1 − h − z ) , (86)
where α \alpha α , β \beta β , δ ∈ [ 1 , 2 ] \delta\in[1,2] δ ∈ [ 1 , 2 ] . Hence, as above, we find the following integral expressions for w j k , l w_{j}^{k,l} w j k , l .
w 3 3 , 3 = 1 2 π H 0 ( w ^ 3 3 , 3 ) , w α 3 , 3 = i x α 2 π ρ H 1 ( k k α w ^ α 3 , 3 ) , w α 3 , 3 = i x α 2 π ρ H 1 ( k k α w ^ α 3 , 3 ) , (87) w_{3}^{3,3}=\frac{1}{2\pi}\mathcal{H}_{0}\left(\hat{w}_{3}^{3,3}\right),\qquad
w_{\alpha}^{3,3}=\frac{ix_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(\frac{k}{k_{\alpha}}\hat{w}_{\alpha}^{3,3}\right),\qquad
w_{\alpha}^{3,3}=\frac{ix_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(\frac{k}{k_{\alpha}}\hat{w}_{\alpha}^{3,3}\right), \tag*{(87)} w 3 3 , 3 = 2 π 1 H 0 ( w ^ 3 3 , 3 ) , w α 3 , 3 = 2 π ρ i x α H 1 ( k α k w ^ α 3 , 3 ) , w α 3 , 3 = 2 π ρ i x α H 1 ( k α k w ^ α 3 , 3 ) , (87)
w 3 β , δ = δ β δ 2 π H 0 ( w ^ 3 , 1 β , δ ) + x β x δ 2 π ρ 2 H 0 ( k 2 w ^ 3 , 2 β , δ ) + 1 2 π ρ ( δ β δ − 2 x β x δ ρ 2 ) H 1 ( k w ^ 3 , 2 β , δ ) , (88) w_{3}^{\beta,\delta}=\frac{\delta_{\beta\delta}}{2\pi}\mathcal{H}_{0}\left(\hat{w}_{3,1}^{\beta,\delta}\right)+\frac{x_{\beta}x_{\delta}}{2\pi\rho^{2}}\mathcal{H}_{0}\left(k^{2}\hat{w}_{3,2}^{\beta,\delta}\right)+\frac{1}{2\pi\rho}\left(\delta_{\beta\delta}-\frac{2x_{\beta}x_{\delta}}{\rho^{2}}\right)\mathcal{H}_{1}\left(k\hat{w}_{3,2}^{\beta,\delta}\right), \tag*{(88)} w 3 β , δ = 2 π δ β δ H 0 ( w ^ 3 , 1 β , δ ) + 2 π ρ 2 x β x δ H 0 ( k 2 w ^ 3 , 2 β , δ ) + 2 π ρ 1 ( δ β δ − ρ 2 2 x β x δ ) H 1 ( k w ^ 3 , 2 β , δ ) , (88)
w α β , δ = i δ β δ x α 2 π ρ H 1 ( k w ^ α , 1 β , δ ) + i x α x β x δ 2 π ρ 3 H 1 ( k 3 w ^ α , 2 β , δ ) + i 2 π ρ 3 ( x α δ β δ + x β δ α δ + x δ δ α β − 4 x α x β x δ ρ 2 ) ( 2 H 1 ( k w ^ α , 2 β , δ ) − ρ H 0 ( k 2 w ^ α , 2 β , δ ) ) , (89) \begin{aligned}
w_{\alpha}^{\beta,\delta}=\frac{i\delta_{\beta\delta}x_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(k\hat{w}_{\alpha,1}^{\beta,\delta}\right)+\frac{ix_{\alpha}x_{\beta}x_{\delta}}{2\pi\rho^{3}}\mathcal{H}_{1}\left(k^{3}\hat{w}_{\alpha,2}^{\beta,\delta}\right) \\
&+\frac{i}{2\pi\rho^{3}}\left(x_{\alpha}\delta_{\beta\delta}+x_{\beta}\delta_{\alpha\delta}+x_{\delta}\delta_{\alpha\beta}-\frac{4x_{\alpha}x_{\beta}x_{\delta}}{\rho^{2}}\right)
\left(2\mathcal{H}_{1}\left(k\hat{w}_{\alpha,2}^{\beta,\delta}\right)-\rho\mathcal{H}_{0}\left(k^{2}\hat{w}_{\alpha,2}^{\beta,\delta}\right)\right), \tag*{(89)}
\end{aligned} w α β , δ = 2 π ρ i δ β δ x α H 1 ( k w ^ α , 1 β , δ ) + 2 π ρ 3 i x α x β x δ H 1 ( k 3 w ^ α , 2 β , δ ) + 2 π ρ 3 i ( x α δ β δ + x β δ α δ + x δ δ α β − ρ 2 4 x α x β x δ ) ( 2 H 1 ( k w ^ α , 2 β , δ ) − ρ H 0 ( k 2 w ^ α , 2 β , δ ) ) , (89)
w β α , 3 = 1 2 π H 0 ( w ^ β , 1 α , 3 ) + x α x β 2 π ρ 2 H 0 ( k 2 w ^ β , 2 α , 3 ) + 1 2 π ρ ( δ α β − 2 x α x β ρ 2 ) H 1 ( k w ^ β , 2 α , 3 ) , (90) w_{\beta}^{\alpha,3}=\frac{1}{2\pi}\mathcal{H}_{0}\left(\hat{w}_{\beta,1}^{\alpha,3}\right)+\frac{x_{\alpha}x_{\beta}}{2\pi\rho^{2}}\mathcal{H}_{0}\left(k^{2}\hat{w}_{\beta,2}^{\alpha,3}\right)+\frac{1}{2\pi\rho}\left(\delta_{\alpha\beta}-\frac{2x_{\alpha}x_{\beta}}{\rho^{2}}\right)\mathcal{H}_{1}\left(k\hat{w}_{\beta,2}^{\alpha,3}\right), \tag*{(90)} w β α , 3 = 2 π 1 H 0 ( w ^ β , 1 α , 3 ) + 2 π ρ 2 x α x β H 0 ( k 2 w ^ β , 2 α , 3 ) + 2 π ρ 1 ( δ α β − ρ 2 2 x α x β ) H 1 ( k w ^ β , 2 α , 3 ) , (90)
where for notational simplicity, we have decomposed w ^ 3 β , δ \hat{w}_{3}^{\beta,\delta} w ^ 3 β , δ , w ^ α β , δ \hat{w}_{\alpha}^{\beta,\delta} w ^ α β , δ and w ^ β α , 3 \hat{w}_{\beta}^{\alpha,3} w ^ β α , 3 as
w ^ 3 β , δ = w ^ 3 , 1 β , δ + k β k δ w ^ 3 , 2 β , δ , w ^ α β , δ = k α w ^ α , 1 β , δ + k α k β k δ w ^ α , 2 β , δ , w ^ β α , 3 = w ^ β , 1 α , 3 + k α k β w ^ β , 2 α , 3 . (91) \hat{w}_{3}^{\beta,\delta}=\hat{w}_{3,1}^{\beta,\delta}+k_{\beta}k_{\delta}\hat{w}_{3,2}^{\beta,\delta},\qquad
\hat{w}_{\alpha}^{\beta,\delta}=k_{\alpha}\hat{w}_{\alpha,1}^{\beta,\delta}+k_{\alpha}k_{\beta}k_{\delta}\hat{w}_{\alpha,2}^{\beta,\delta},\qquad
\hat{w}_{\beta}^{\alpha,3}=\hat{w}_{\beta,1}^{\alpha,3}+k_{\alpha}k_{\beta}\hat{w}_{\beta,2}^{\alpha,3}. \tag*{(91)} w ^ 3 β , δ = w ^ 3 , 1 β , δ + k β k δ w ^ 3 , 2 β , δ , w ^ α β , δ = k α w ^ α , 1 β , δ + k α k β k δ w ^ α , 2 β , δ , w ^ β α , 3 = w ^ β , 1 α , 3 + k α k β w ^ β , 2 α , 3 . (91)
Similarly to the source above, F F F has in γ \gamma γ poles of order 2 2 2 at z = π i ( n + 1 / 2 ) z=\pi i(n+1/2) z = π i ( n + 1/2 ) where n ∈ Z ≥ 0 n\in\mathbb{Z}_{\geq0} n ∈ Z ≥ 0 and poles of order 1 1 1 at z = z 0 / 2 z=z_0/2 z = z 0 /2 where z 0 z_0 z 0 satisfies sinh z 0 = z 0 \sinh z_0=z_0 sinh z 0 = z 0 . Since ∫ γ ϵ \int_{\gamma_\epsilon} ∫ γ ϵ vanishes as ϵ → 0 \epsilon\to0 ϵ → 0 , when j = k = l = 3 j=k=l=3 j = k = l = 3 (36) simplifies to become
w 3 3 , 3 = 2 π 3 ∑ n = 1 , 3 , 5 , … ∞ n cos ( n π z 2 ) sin ( n π h 2 ) ( 3 K 0 ( n π ρ 2 ) − n π ρ 2 K 1 ( n π ρ 2 ) ) + ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 8 ( cosh z 0 − 1 ) ( w ^ 3 3 , 3 ( sinh 2 k − 2 k ) ) k = z 0 / 2 H 0 1 ( ρ z 0 2 ) . (92) w_3^{3,3}=\frac{2\pi}{3}\sum_{n=1,3,5,\ldots}^{\infty}n\cos\left(\frac{n\pi z}{2}\right)\sin\left(\frac{n\pi h}{2}\right)\left(3K_0\left(\frac{n\pi\rho}{2}\right)-\frac{n\pi\rho}{2}K_1\left(\frac{n\pi\rho}{2}\right)\right)
+\sum_{\substack{z_0\in\mathbb{H}:\\z_0=\sinh z_0}}\frac{i z_0}{8(\cosh z_0-1)}\left(\widehat{w}_3^{3,3}(\sinh2k-2k)\right)_{k=z_0/2}H_0^{1}\left(\frac{\rho z_0}{2}\right). \tag*{(92)} w 3 3 , 3 = 3 2 π n = 1 , 3 , 5 , … ∑ ∞ n cos ( 2 nπ z ) sin ( 2 nπ h ) ( 3 K 0 ( 2 nπ ρ ) − 2 nπ ρ K 1 ( 2 nπ ρ ) ) + z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 ( cosh z 0 − 1 ) i z 0 ( w 3 3 , 3 ( sinh 2 k − 2 k ) ) k = z 0 /2 H 0 1 ( 2 ρ z 0 ) . (92)
u 3 3 , 3 = v 3 3 , 3 + w 3 3 , 3 = O ( e − ρ y 1 / 2 ρ ) . (93) u_3^{3,3}=v_3^{3,3}+w_3^{3,3}=\mathcal{O}\left(\frac{e^{-\rho y_1/2}}{\sqrt{\rho}}\right). \tag*{(93)} u 3 3 , 3 = v 3 3 , 3 + w 3 3 , 3 = O ( ρ e − ρ y 1 /2 ) . (93)
noting that as for the Stokeslet, the contribution from the poles of order 2 in w 3 3 , 3 w_3^{3,3} w 3 3 , 3 cancels out with v 3 3 , 3 v_3^{3,3} v 3 3 , 3 . Similarly, the leading order contribution in the far-field when j = 3 j=3 j = 3 for the other cases for k k k and l l l are
u 3 3 , α = u 3 α , 3 = O ( x α e − ρ y 1 / 2 ρ 3 / 2 ) , u 3 β , δ = O ( δ β δ e − ρ y 1 / 2 ρ ) + O ( x β x δ e − ρ y 1 / 2 ρ 5 / 2 ) . (94) u_3^{3,\alpha}=u_3^{\alpha,3}=\mathcal{O}\left(\frac{x_\alpha e^{-\rho y_1/2}}{\rho^{3/2}}\right),\quad u_3^{\beta,\delta}=\mathcal{O}\left(\delta_{\beta\delta}\frac{e^{-\rho y_1/2}}{\sqrt{\rho}}\right)+\mathcal{O}\left(\frac{x_\beta x_\delta e^{-\rho y_1/2}}{\rho^{5/2}}\right). \tag*{(94)} u 3 3 , α = u 3 α , 3 = O ( ρ 3/2 x α e − ρ y 1 /2 ) , u 3 β , δ = O ( δ β δ ρ e − ρ y 1 /2 ) + O ( ρ 5/2 x β x δ e − ρ y 1 /2 ) . (94)
Finally, when j = α ∈ [ 1 , 2 ] j=\alpha\in[1,2] j = α ∈ [ 1 , 2 ] , the leading order contribution in the far-field arises from γ ϵ \gamma_\epsilon γ ϵ , namely
u α 3 , 3 = z x α ρ 2 ( 2 − z ) , u β α , 3 = z ( 2 − z ) [ − 1 − h ρ 2 ( δ α β − 2 x α x β ρ 2 ) ] . (95) u_\alpha^{3,3}=\frac{zx_\alpha}{\rho^2}(2-z),\quad u_\beta^{\alpha,3}=z(2-z)\left[-\frac{1-h}{\rho^2}\left(\delta_{\alpha\beta}-\frac{2x_\alpha x_\beta}{\rho^2}\right)\right]. \tag*{(95)} u α 3 , 3 = ρ 2 z x α ( 2 − z ) , u β α , 3 = z ( 2 − z ) [ − ρ 2 1 − h ( δ α β − ρ 2 2 x α x β ) ] . (95)
u α β , δ = z ( 2 − z ) ( x α ρ 2 δ β δ − 2 h ρ 4 ( 2 − h ) ( x α δ β δ + x β δ α δ + x δ δ α β − 4 x α x β x δ ρ 2 ) ) . (96) u_\alpha^{\beta,\delta}=z(2-z)\left(\frac{x_\alpha}{\rho^2}\delta_{\beta\delta}-\frac{2h}{\rho^4}(2-h)\left(x_\alpha\delta_{\beta\delta}+x_\beta\delta_{\alpha\delta}+x_\delta\delta_{\alpha\beta}-\frac{4x_\alpha x_\beta x_\delta}{\rho^2}\right)\right). \tag*{(96)} u α β , δ = z ( 2 − z ) ( ρ 2 x α δ β δ − ρ 4 2 h ( 2 − h ) ( x α δ β δ + x β δ α δ + x δ δ α β − ρ 2 4 x α x β x δ ) ) . (96)
Rotlet Dipole in a Petri Dish
From the fifth order repeated reflection solutions (Appendix A), v j k v_j^k v j k for a rotlet dipole is
v j k = L ( ϵ j p k x p x k r 5 ) = δ j 3 δ k α x α x β ϵ 3 β α L ( 1 r 5 ) + δ j α x β ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) L ( z r 3 ) = ϵ 3 β α 4 δ j 3 δ k α x α x β F 1 , 1 3 ρ 3 − ϵ 3 β α 2 δ j 3 δ k α x α x β F 2 , 0 3 ρ 2 + 2 δ j α x β G 2 , 1 3 ρ ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) , (97) \begin{aligned}
v_j^k=\mathcal{L}\left(\epsilon_{jpk}\frac{x_px_k}{r^5}\right)=\delta_{j3}\delta_{k\alpha}x_\alpha x_\beta\epsilon_{3\beta\alpha}\mathcal{L}\left(\frac{1}{r^5}\right)+\delta_{j\alpha}x_\beta\left(\delta_{k3}\epsilon_{\alpha\beta3}+\delta_{k\beta}\epsilon_{\alpha3\beta}\right)\mathcal{L}\left(\frac{z}{r^3}\right) \\
&=\epsilon_{3\beta\alpha}\frac{4\delta_{j3}\delta_{k\alpha}x_\alpha x_\beta F_{1,1}}{3\rho^3}-\epsilon_{3\beta\alpha}\frac{2\delta_{j3}\delta_{k\alpha}x_\alpha x_\beta F_{2,0}}{3\rho^2}+\frac{2\delta_{j\alpha}x_\beta G_{2,1}}{3\rho}\left(\delta_{k3}\epsilon_{\alpha\beta3}+\delta_{k\beta}\epsilon_{\alpha3\beta}\right), \tag*{(97)}
\end{aligned} v j k = L ( ϵ j p k r 5 x p x k ) = δ j 3 δ k α x α x β ϵ 3 β α L ( r 5 1 ) + δ j α x β ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) L ( r 3 z ) = ϵ 3 β α 3 ρ 3 4 δ j 3 δ k α x α x β F 1 , 1 − ϵ 3 β α 3 ρ 2 2 δ j 3 δ k α x α x β F 2 , 0 + 3 ρ 2 δ j α x β G 2 , 1 ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) , (97)
with boundary conditions for the corresponding auxiliary solution w j k w_j^k w j k and transformed auxiliary solution w ^ j k \widehat{w}_j^k w j k
w j k ∣ z = 0 = 2 δ j α x β 3 ρ ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) ∫ 0 ∞ d λ J 1 ( λ ρ ) λ 2 cosh ( 1 − h ) λ cosh λ ⟹ \left.w_j^k\right|_{z=0}=\frac{2\delta_{j\alpha}x_\beta}{3\rho}\left(\delta_{k3}\epsilon_{\alpha\beta3}+\delta_{k\beta}\epsilon_{\alpha3\beta}\right)\int_0^\infty d\lambda\,J_1(\lambda\rho)\frac{\lambda^2\cosh(1-h)\lambda}{\cosh\lambda}\Longrightarrow w j k z = 0 = 3 ρ 2 δ j α x β ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) ∫ 0 ∞ d λ J 1 ( λ ρ ) cosh λ λ 2 cosh ( 1 − h ) λ ⟹
w ^ j k ∣ z = 0 = − 4 π i δ j α k β 3 ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) cosh k ( 1 − h ) cosh k , (98) \left.\widehat{w}_j^k\right|_{z=0}=-\frac{4\pi i\delta_{j\alpha}k_\beta}{3}\left(\delta_{k3}\epsilon_{\alpha\beta3}+\delta_{k\beta}\epsilon_{\alpha3\beta}\right)\frac{\cosh k(1-h)}{\cosh k}, \tag*{(98)} w j k z = 0 = − 3 4 π i δ j α k β ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) cosh k cosh k ( 1 − h ) , (98)
∂ w α k ∂ z ∣ z = 1 = 2 x β 3 ρ ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) ∫ 0 ∞ d λ J 1 ( λ ρ ) λ 3 sinh h λ cosh λ ⟹ \left.\frac{\partial w_\alpha^k}{\partial z}\right|_{z=1}=\frac{2x_\beta}{3\rho}\left(\delta_{k3}\epsilon_{\alpha\beta3}+\delta_{k\beta}\epsilon_{\alpha3\beta}\right)\int_0^\infty d\lambda\,J_1(\lambda\rho)\frac{\lambda^3\sinh h\lambda}{\cosh\lambda}\Longrightarrow ∂ z ∂ w α k z = 1 = 3 ρ 2 x β ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) ∫ 0 ∞ d λ J 1 ( λ ρ ) cosh λ λ 3 sinh hλ ⟹
∂ w ^ α k ∂ z ∣ z = 1 = − 4 π k β 3 ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) k sinh h k cosh k , (99) \left.\frac{\partial\widehat{w}_\alpha^k}{\partial z}\right|_{z=1}=-\frac{4\pi k_\beta}{3}\left(\delta_{k3}\epsilon_{\alpha\beta3}+\delta_{k\beta}\epsilon_{\alpha3\beta}\right)\frac{k\sinh hk}{\cosh k}, \tag*{(99)} ∂ z ∂ w α k z = 1 = − 3 4 π k β ( δ k 3 ϵ α β 3 + δ k β ϵ α 3 β ) cosh k k sinh hk , (99)
w 3 k ∣ z = 1 = − 4 δ k α x α x β 3 ρ 3 ϵ 3 β α ∫ 0 ∞ d λ J 1 ( λ ρ ) λ sinh h λ cosh λ + 2 δ k α x α x β 3 ρ 2 ϵ 3 β α ∫ 0 ∞ d λ J 0 ( λ ρ ) λ 2 sinh h λ cosh λ . (100) \left.w_3^k\right|_{z=1}=-\frac{4\delta_{k\alpha}x_\alpha x_\beta}{3\rho^3}\epsilon_{3\beta\alpha}\int_0^\infty d\lambda\,J_1(\lambda\rho)\frac{\lambda\sinh h\lambda}{\cosh\lambda}+\frac{2\delta_{k\alpha}x_\alpha x_\beta}{3\rho^2}\epsilon_{3\beta\alpha}\int_0^\infty d\lambda\,J_0(\lambda\rho)\frac{\lambda^2\sinh h\lambda}{\cosh\lambda}. \tag*{(100)} w 3 k z = 1 = − 3 ρ 3 4 δ k α x α x β ϵ 3 β α ∫ 0 ∞ d λ J 1 ( λ ρ ) cosh λ λ sinh hλ + 3 ρ 2 2 δ k α x α x β ϵ 3 β α ∫ 0 ∞ d λ J 0 ( λ ρ ) cosh λ λ 2 sinh hλ . (100)
However, (D2c) is difficult to transform. Noting that α ≠ β \alpha\ne\beta α = β and utilising Bessel function identities, we find
w ^ 3 k ∣ z = 1 = − 2 δ k α ϵ 3 β α 3 k α k β k ∂ ∂ k ( 1 k ( 2 π ∫ 0 ∞ ∫ 0 ∞ d ρ d λ sinh h λ cosh λ J 0 ( k ρ ) ( 2 λ J 1 λ ρ ρ 2 − λ 2 J 0 ( λ ρ ) ρ ) ) ) = − 4 π δ k α ϵ 3 β α 3 k α k β k ∂ ∂ k ( 1 k ( g 1 − g 2 ) ) , (101) \begin{aligned}
\widehat{w}^{k}_{3}\big|_{z=1}
&=-\frac{2\delta_{k\alpha}\epsilon_{3\beta\alpha}}{3}\frac{k_{\alpha}k_{\beta}}{k}\frac{\partial}{\partial k}\left(\frac{1}{k}\left(2\pi\int_{0}^{\infty}\int_{0}^{\infty}d\rho d\lambda\frac{\sinh h\lambda}{\cosh\lambda}J_{0}(k\rho)\left(\frac{2\lambda J_{1}\lambda\rho}{\rho^{2}}-\frac{\lambda^{2}J_{0}(\lambda\rho)}{\rho}\right)\right)\right)\\
&=-\frac{4\pi\delta_{k\alpha}\epsilon_{3\beta\alpha}}{3}\frac{k_{\alpha}k_{\beta}}{k}\frac{\partial}{\partial k}\left(\frac{1}{k}(g_{1}-g_{2})\right), \tag*{(101)}
\end{aligned} w 3 k z = 1 = − 3 2 δ k α ϵ 3 β α k k α k β ∂ k ∂ ( k 1 ( 2 π ∫ 0 ∞ ∫ 0 ∞ d ρ d λ cosh λ sinh hλ J 0 ( k ρ ) ( ρ 2 2 λ J 1 λ ρ − ρ λ 2 J 0 ( λ ρ ) ) ) ) = − 3 4 π δ k α ϵ 3 β α k k α k β ∂ k ∂ ( k 1 ( g 1 − g 2 ) ) , (101)
where g 1 g_{1} g 1 and g 2 g_{2} g 2 are defined as satisfying respectively
g 1 = ∫ 0 ∞ ∫ 0 ∞ d ρ d λ λ 2 sinh h λ cosh λ J 0 ( λ ρ ) J 1 ( k ρ ) , g 2 = ∫ 0 ∞ ∫ 0 ∞ d ρ d λ 2 λ sinh ( h λ ) cosh λ J 1 ( λ ρ ) J 1 ( k ρ ) ρ . (102) g_{1}=\int_{0}^{\infty}\int_{0}^{\infty}d\rho d\lambda\frac{\lambda^{2}\sinh h\lambda}{\cosh\lambda}J_{0}(\lambda\rho)J_{1}(k\rho),\qquad
g_{2}=\int_{0}^{\infty}\int_{0}^{\infty}d\rho d\lambda\frac{2\lambda\sinh(h\lambda)}{\cosh\lambda}\frac{J_{1}(\lambda\rho)J_{1}(k\rho)}{\rho}. \tag*{(102)} g 1 = ∫ 0 ∞ ∫ 0 ∞ d ρ d λ cosh λ λ 2 sinh hλ J 0 ( λ ρ ) J 1 ( k ρ ) , g 2 = ∫ 0 ∞ ∫ 0 ∞ d ρ d λ cosh λ 2 λ sinh ( hλ ) ρ J 1 ( λ ρ ) J 1 ( k ρ ) . (102)
However, g 1 g_{1} g 1 simplifies to give
g 1 = ∫ 0 ∞ λ 2 sinh h λ cosh λ [ ∫ 0 ∞ ρ d ρ J 0 ( ρ λ ) ( J 1 ( ρ λ ) ρ ) ] = 1 k ∫ 0 k d λ λ 2 sinh h λ cosh λ . (103) g_{1}=\int_{0}^{\infty}\frac{\lambda^{2}\sinh h\lambda}{\cosh\lambda}\left[\int_{0}^{\infty}\rho d\rho J_{0}(\rho\lambda)\left(\frac{J_{1}(\rho\lambda)}{\rho}\right)\right]=\frac{1}{k}\int_{0}^{k}d\lambda\frac{\lambda^{2}\sinh h\lambda}{\cosh\lambda}. \tag*{(103)} g 1 = ∫ 0 ∞ cosh λ λ 2 sinh hλ [ ∫ 0 ∞ ρ d ρ J 0 ( ρ λ ) ( ρ J 1 ( ρ λ ) ) ] = k 1 ∫ 0 k d λ cosh λ λ 2 sinh hλ . (103)
Furthermore, g 2 g_{2} g 2 simplifies to give
g 2 = ∫ 0 ∞ d λ 2 λ sinh 2 h λ cosh λ [ ∫ 0 ∞ d ρ J 1 ( k ρ ) J 2 ( λ ρ ) ρ ] = ∫ 0 ∞ d λ 2 λ sinh h λ cosh λ [ λ k k 2 + λ 2 + ∣ k 2 − λ 2 ∣ ] = ∫ 0 k d λ λ 2 k sinh h λ cosh λ + ∫ k ∞ d λ k sinh h λ cosh λ . (104) \begin{aligned}
g_{2}
&=\int_{0}^{\infty}d\lambda\frac{2\lambda\sinh2h\lambda}{\cosh\lambda}\left[\int_{0}^{\infty}d\rho\frac{J_{1}(k\rho)J_{2}(\lambda\rho)}{\rho}\right]
=\int_{0}^{\infty}d\lambda\frac{2\lambda\sinh h\lambda}{\cosh\lambda}\left[\frac{\lambda k}{k^{2}+\lambda^{2}+|k^{2}-\lambda^{2}|}\right]\\
&=\int_{0}^{k}d\lambda\frac{\lambda^{2}}{k}\frac{\sinh h\lambda}{\cosh\lambda}+\int_{k}^{\infty}d\lambda\frac{k\sinh h\lambda}{\cosh\lambda}. \tag*{(104)}
\end{aligned} g 2 = ∫ 0 ∞ d λ cosh λ 2 λ sinh 2 hλ [ ∫ 0 ∞ d ρ ρ J 1 ( k ρ ) J 2 ( λ ρ ) ] = ∫ 0 ∞ d λ cosh λ 2 λ sinh hλ [ k 2 + λ 2 + ∣ k 2 − λ 2 ∣ λk ] = ∫ 0 k d λ k λ 2 cosh λ sinh hλ + ∫ k ∞ d λ cosh λ k sinh hλ . (104)
Putting this all together, (D3) becomes
w ^ 3 k ∣ z = 1 = 4 π δ k α ϵ 3 β α 3 k α k β k [ sinh h k cosh k ] . (105) \widehat{w}^{k}_{3}\big|_{z=1}=\frac{4\pi\delta_{k\alpha}\epsilon_{3\beta\alpha}}{3}\frac{k_{\alpha}k_{\beta}}{k}\left[\frac{\sinh hk}{\cosh k}\right]. \tag*{(105)} w 3 k z = 1 = 3 4 π δ k α ϵ 3 β α k k α k β [ cosh k sinh hk ] . (105)
Hence, as in the main text for a source, we can thus solve for w ^ j k \widehat{w}^{k}_{j} w j k to give
w ^ 3 3 = 0 , w ^ 3 α = − 4 π k β i ϵ α β 3 3 cosh k ( 1 − h − z ) cosh k , (106) \widehat{w}^{3}_{3}=0,\qquad\widehat{w}^{\alpha}_{3}=-\frac{4\pi k_{\beta}i\epsilon_{\alpha\beta3}}{3}\frac{\cosh k(1-h-z)}{\cosh k}, \tag*{(106)} w 3 3 = 0 , w 3 α = − 3 4 π k β i ϵ α β 3 cosh k cosh k ( 1 − h − z ) , (106)
w ^ 3 α = − 4 π k α k β i ϵ β α 3 3 k cosh k ( sinh 2 k − 2 k ) ( 2 k cosh k ( 1 − h ) sinh k z − k z cosh k z sinh k ( 1 − h ) − k z sinh k ( 1 − h ) cosh k ( 2 − z ) − 2 cosh k sinh h k sinh k z ) , (107) \widehat{w}^{\alpha}_{3}=-\frac{4\pi k_{\alpha}k_{\beta}i\epsilon_{\beta\alpha3}}{3k\cosh k(\sinh2k-2k)}\left(2k\cosh k(1-h)\sinh kz-kz\cosh kz\sinh k(1-h)-kz\sinh k(1-h)\cosh k(2-z)-2\cosh k\sinh hk\sinh kz\right), \tag*{(107)} w 3 α = − 3 k cosh k ( sinh 2 k − 2 k ) 4 π k α k β i ϵ β α 3 ( 2 k cosh k ( 1 − h ) sinh k z − k z cosh k z sinh k ( 1 − h ) − k z sinh k ( 1 − h ) cosh k ( 2 − z ) − 2 cosh k sinh hk sinh k z ) , (107)
w ^ α α = 4 π k α 2 k β i ϵ β α 3 3 k cosh k ( sinh 2 k − 2 k ) ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) , (108) \widehat{w}^{\alpha}_{\alpha}=\frac{4\pi k_{\alpha}^{2}k_{\beta}i\epsilon_{\beta\alpha3}}{3k\cosh k(\sinh2k-2k)}\left(z\sinh k(h-1)\sinh k(2-z)+(2-z)\sinh(kz)\sinh k(h-1)\right), \tag*{(108)} w α α = 3 k cosh k ( sinh 2 k − 2 k ) 4 π k α 2 k β i ϵ β α 3 ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) , (108)
w ^ β α = 4 π k α k β 2 i ϵ β α 3 3 k cosh k ( sinh 2 k − 2 k ) ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) + 4 π k α i ϵ β α 3 3 cosh k ( sinh 2 k − 2 k ) ( sinh 2 k cosh k ( h + z − 1 ) − 2 k cosh k ( h + z − 1 ) ) . (109) \begin{aligned}
\widehat{w}^{\alpha}_{\beta}
&=\frac{4\pi k_{\alpha}k_{\beta}^{2}i\epsilon_{\beta\alpha3}}{3k\cosh k(\sinh2k-2k)}\left(z\sinh k(h-1)\sinh k(2-z)+(2-z)\sinh(kz)\sinh k(h-1)\right)\\
&\quad+\frac{4\pi k_{\alpha}i\epsilon_{\beta\alpha3}}{3\cosh k(\sinh2k-2k)}\left(\sinh2k\cosh k(h+z-1)-2k\cosh k(h+z-1)\right). \tag*{(109)}
\end{aligned} w β α = 3 k cosh k ( sinh 2 k − 2 k ) 4 π k α k β 2 i ϵ β α 3 ( z sinh k ( h − 1 ) sinh k ( 2 − z ) + ( 2 − z ) sinh ( k z ) sinh k ( h − 1 ) ) + 3 cosh k ( sinh 2 k − 2 k ) 4 π k α i ϵ β α 3 ( sinh 2 k cosh k ( h + z − 1 ) − 2 k cosh k ( h + z − 1 ) ) . (109)
where β ∈ [ 1 , 2 ] \beta\in[1,2] β ∈ [ 1 , 2 ] and β ≠ α \beta\ne\alpha β = α . Rewriting the inverse Fourier transform in terms of Hankel transforms, we get the following integral expressions for w j α w_j^{\alpha} w j α
w 3 3 = 0 , w β 3 = i x β 2 π ρ H 1 ( k k β w ^ α 3 ) , w 3 α = − x α x β π ρ 3 H 1 ( k k α k β w ^ 3 α ) + x α x β 2 π ρ 2 H 0 ( k 2 k α k β w ^ 3 α ) (110) w_3^3=0,\quad w_\beta^3=\frac{ix_\beta}{2\pi\rho}\mathcal{H}_1\left(\frac{k}{k_\beta}\widehat{w}_\alpha^3\right),\quad w_3^\alpha=-\frac{x_\alpha x_\beta}{\pi\rho^3}\mathcal{H}_1\left(\frac{k}{k_\alpha k_\beta}\widehat{w}_3^\alpha\right)+\frac{x_\alpha x_\beta}{2\pi\rho^2}\mathcal{H}_0\left(\frac{k^2}{k_\alpha k_\beta}\widehat{w}_3^\alpha\right) \tag*{(110)} w 3 3 = 0 , w β 3 = 2 π ρ i x β H 1 ( k β k w α 3 ) , w 3 α = − π ρ 3 x α x β H 1 ( k α k β k w 3 α ) + 2 π ρ 2 x α x β H 0 ( k α k β k 2 w 3 α ) (110)
w β α = i x α 2 x β 2 π ρ 3 H 1 ( k 3 k α 2 k β w ^ α α ) + i 2 π ρ 3 ( x β − 4 x α 2 x β ρ 2 ) ( 2 H 1 ( k k α 2 k β w ^ α α ) − ρ H 0 ( k 2 k α 2 k β w ^ α α ) ) (111) w_\beta^\alpha=\frac{ix_\alpha^2x_\beta}{2\pi\rho^3}\mathcal{H}_1\left(\frac{k^3}{k_\alpha^2k_\beta}\widehat{w}_\alpha^\alpha\right)+\frac{i}{2\pi\rho^3}\left(x_\beta-\frac{4x_\alpha^2x_\beta}{\rho^2}\right)\left(2\mathcal{H}_1\left(\frac{k}{k_\alpha^2k_\beta}\widehat{w}_\alpha^\alpha\right)-\rho\mathcal{H}_0\left(\frac{k^2}{k_\alpha^2k_\beta}\widehat{w}_\alpha^\alpha\right)\right) \tag*{(111)} w β α = 2 π ρ 3 i x α 2 x β H 1 ( k α 2 k β k 3 w α α ) + 2 π ρ 3 i ( x β − ρ 2 4 x α 2 x β ) ( 2 H 1 ( k α 2 k β k w α α ) − ρ H 0 ( k α 2 k β k 2 w α α ) ) (111)
w β α = i x α 2 π ρ H 1 ( k w ^ 0 ) + i x α x β 2 2 π ρ 3 H 1 ( k 3 w ^ 1 ) + i 2 π ρ 3 ( x α − 4 x α x β 2 ρ 2 ) ( 2 H 1 ( k w ^ 1 ) − ρ H 0 ( k 2 w ^ 1 ) ) (112) w_\beta^\alpha=\frac{ix_\alpha}{2\pi\rho}\mathcal{H}_1\left(k\widehat{w}_0\right)+\frac{ix_\alpha x_\beta^2}{2\pi\rho^3}\mathcal{H}_1\left(k^3\widehat{w}_1\right)+\frac{i}{2\pi\rho^3}\left(x_\alpha-\frac{4x_\alpha x_\beta^2}{\rho^2}\right)\left(2\mathcal{H}_1\left(k\widehat{w}_1\right)-\rho\mathcal{H}_0\left(k^2\widehat{w}_1\right)\right) \tag*{(112)} w β α = 2 π ρ i x α H 1 ( k w 0 ) + 2 π ρ 3 i x α x β 2 H 1 ( k 3 w 1 ) + 2 π ρ 3 i ( x α − ρ 2 4 x α x β 2 ) ( 2 H 1 ( k w 1 ) − ρ H 0 ( k 2 w 1 ) ) (112)
where β ∈ [ 1 , 2 ] \beta\in[1,2] β ∈ [ 1 , 2 ] , β ≠ α \beta\ne\alpha β = α and for notational simplicity we have decomposed w ^ β α \widehat{w}_\beta^\alpha w β α as w ^ β α = k α ( w ^ 0 + k β 2 w ^ 1 ) \widehat{w}_\beta^\alpha=k_\alpha\left(\widehat{w}_0+k_\beta^2\widehat{w}_1\right) w β α = k α ( w 0 + k β 2 w 1 ) . When k = j = 3 k=j=3 k = j = 3 , u 3 3 = v 3 3 = w 3 3 = 0 u_3^3=v_3^3=w_3^3=0 u 3 3 = v 3 3 = w 3 3 = 0 . Furthermore, when k = 3 k=3 k = 3 and j = α j=\alpha j = α , we have
v α 3 = − x β ϵ α β 3 3 ρ ∑ n = 1 , 3 , 5 , … ∞ π 2 n 2 K 1 ( n π ρ 2 ) [ sin ( n π h 2 ) cos ( n π z 2 ) ] (113) v_\alpha^3=-\frac{x_\beta\epsilon_{\alpha\beta3}}{3\rho}\sum_{n=1,3,5,\ldots}^{\infty}\pi^2n^2K_1\left(\frac{n\pi\rho}{2}\right)\left[\sin\left(\frac{n\pi h}{2}\right)\cos\left(\frac{n\pi z}{2}\right)\right] \tag*{(113)} v α 3 = − 3 ρ x β ϵ α β 3 n = 1 , 3 , 5 , … ∑ ∞ π 2 n 2 K 1 ( 2 nπ ρ ) [ sin ( 2 nπ h ) cos ( 2 nπ z ) ] (113)
w α 3 = x β ϵ α β 3 3 ρ ∑ n = 1 , 3 , 5 , … ∞ π 2 n 2 K 1 ( n π ρ 2 ) [ sin ( n π h 2 ) cos ( n π z 2 ) + cos ( n π h 2 ) sin ( n π z 2 ) ] (114) w_\alpha^3=\frac{x_\beta\epsilon_{\alpha\beta3}}{3\rho}\sum_{n=1,3,5,\ldots}^{\infty}\pi^2n^2K_1\left(\frac{n\pi\rho}{2}\right)\left[\sin\left(\frac{n\pi h}{2}\right)\cos\left(\frac{n\pi z}{2}\right)+\cos\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)\right] \tag*{(114)} w α 3 = 3 ρ x β ϵ α β 3 n = 1 , 3 , 5 , … ∑ ∞ π 2 n 2 K 1 ( 2 nπ ρ ) [ sin ( 2 nπ h ) cos ( 2 nπ z ) + cos ( 2 nπ h ) sin ( 2 nπ z ) ] (114)
u α 3 = v α 3 + w α 3 = x β ϵ α β 3 3 ρ ∑ n = 1 , 3 , 5 , … ∞ π 2 n 2 K 1 ( n π ρ 2 ) [ cos ( n π h 2 ) sin ( n π z 2 ) ] . (115) u_\alpha^3=v_\alpha^3+w_\alpha^3=\frac{x_\beta\epsilon_{\alpha\beta3}}{3\rho}\sum_{n=1,3,5,\ldots}^{\infty}\pi^2n^2K_1\left(\frac{n\pi\rho}{2}\right)\left[\cos\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)\right]. \tag*{(115)} u α 3 = v α 3 + w α 3 = 3 ρ x β ϵ α β 3 n = 1 , 3 , 5 , … ∑ ∞ π 2 n 2 K 1 ( 2 nπ ρ ) [ cos ( 2 nπ h ) sin ( 2 nπ z ) ] . (115)
Hence in the far-field, the leading order contribution decays exponentially as
u α 3 = O ( ϵ α β 3 x β e − ρ π / 2 ρ 3 / 2 ) . (116) u_\alpha^3=\mathcal{O}\left(\frac{\epsilon_{\alpha\beta3}x_\beta e^{-\rho\pi/2}}{\rho^{3/2}}\right). \tag*{(116)} u α 3 = O ( ρ 3/2 ϵ α β 3 x β e − ρ π /2 ) . (116)
Since ∫ γ ϵ \int_{\gamma_\epsilon} ∫ γ ϵ vanishes as ϵ → 0 \epsilon\to0 ϵ → 0 , when j = 3 j=3 j = 3 and k = α k=\alpha k = α where α ∈ [ 1 , 2 ] \alpha\in[1,2] α ∈ [ 1 , 2 ] [31 ] becomes
w 3 α = − 4 x α x β ϵ β 3 α 3 ρ 2 ∑ n = 1 , 3 , 5 , … ∞ sin ( n π h 2 ) sin ( n π z 2 ) K 0 ( n π ρ 2 ) − 4 π x α x β ϵ β 3 α 3 ρ 3 ∑ n = 1 , 3 , 5 , … ∞ n sin ( n π h 2 ) sin ( n π z 2 ) K 1 ( n π ρ 2 ) − x α x β ρ 3 ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 2 H 1 1 ( ρ z 0 2 ) 8 ( cosh z 0 − 1 ) ( ( sinh 2 k − 2 k ) w ^ 3 α k α ) ∣ k = z 0 / 2 + x α x β ρ 2 ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 3 H 0 1 ( ρ z 0 2 ) 32 ( cosh z 0 − 1 ) ( ( sinh 2 k − 2 k ) w ^ 3 α k α ) ∣ k = z 0 / 2 , (117) \begin{aligned}
w_3^\alpha={}&-\frac{4x_\alpha x_\beta\epsilon_{\beta3\alpha}}{3\rho^2}\sum_{n=1,3,5,\ldots}^{\infty}\sin\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)K_0\left(\frac{n\pi\rho}{2}\right)\\
&-\frac{4\pi x_\alpha x_\beta\epsilon_{\beta3\alpha}}{3\rho^3}\sum_{n=1,3,5,\ldots}^{\infty}n\sin\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)K_1\left(\frac{n\pi\rho}{2}\right)\\
&-\frac{x_\alpha x_\beta}{\rho^3}\sum_{\substack{z_0\in\mathbb{H}:\\z_0=\sinh z_0}}\frac{iz_0^2H_1^1\left(\frac{\rho z_0}{2}\right)}{8(\cosh z_0-1)}\left(\left(\sinh2k-2k\right)\frac{\widehat{w}_3^\alpha}{k_\alpha}\right)\bigg|_{k=z_0/2}\\
&+\frac{x_\alpha x_\beta}{\rho^2}\sum_{\substack{z_0\in\mathbb{H}:\\z_0=\sinh z_0}}\frac{iz_0^3H_0^1\left(\frac{\rho z_0}{2}\right)}{32(\cosh z_0-1)}\left(\left(\sinh2k-2k\right)\frac{\widehat{w}_3^\alpha}{k_\alpha}\right)\bigg|_{k=z_0/2}, \tag*{(117)}
\end{aligned} w 3 α = − 3 ρ 2 4 x α x β ϵ β 3 α n = 1 , 3 , 5 , … ∑ ∞ sin ( 2 nπ h ) sin ( 2 nπ z ) K 0 ( 2 nπ ρ ) − 3 ρ 3 4 π x α x β ϵ β 3 α n = 1 , 3 , 5 , … ∑ ∞ n sin ( 2 nπ h ) sin ( 2 nπ z ) K 1 ( 2 nπ ρ ) − ρ 3 x α x β z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 ( cosh z 0 − 1 ) i z 0 2 H 1 1 ( 2 ρ z 0 ) ( ( sinh 2 k − 2 k ) k α w 3 α ) k = z 0 /2 + ρ 2 x α x β z 0 ∈ H : z 0 = s i n h z 0 ∑ 32 ( cosh z 0 − 1 ) i z 0 3 H 0 1 ( 2 ρ z 0 ) ( ( sinh 2 k − 2 k ) k α w 3 α ) k = z 0 /2 , (117)
u 3 α = v 3 α + w 3 α = − x α x β ρ 3 ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 2 H 1 1 ( ρ z 0 2 ) 8 ( cosh z 0 − 1 ) ( ( sinh 2 k − 2 k ) w ^ 3 α k α ) ∣ k = z 0 / 2 + x α x β ρ 2 ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 3 H 0 1 ( ρ z 0 2 ) 32 ( cosh z 0 − 1 ) ( ( sinh 2 k − 2 k ) w ^ 3 α k α ) ∣ k = z 0 / 2 . (118) \begin{aligned}
u_3^\alpha={}&v_3^\alpha+w_3^\alpha=-\frac{x_\alpha x_\beta}{\rho^3}\sum_{\substack{z_0\in\mathbb{H}:\\z_0=\sinh z_0}}\frac{iz_0^2H_1^1\left(\frac{\rho z_0}{2}\right)}{8(\cosh z_0-1)}\left(\left(\sinh2k-2k\right)\frac{\widehat{w}_3^\alpha}{k_\alpha}\right)\bigg|_{k=z_0/2}\\
&+\frac{x_\alpha x_\beta}{\rho^2}\sum_{\substack{z_0\in\mathbb{H}:\\z_0=\sinh z_0}}\frac{iz_0^3H_0^1\left(\frac{\rho z_0}{2}\right)}{32(\cosh z_0-1)}\left(\left(\sinh2k-2k\right)\frac{\widehat{w}_3^\alpha}{k_\alpha}\right)\bigg|_{k=z_0/2}. \tag*{(118)}
\end{aligned} u 3 α = v 3 α + w 3 α = − ρ 3 x α x β z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 ( cosh z 0 − 1 ) i z 0 2 H 1 1 ( 2 ρ z 0 ) ( ( sinh 2 k − 2 k ) k α w 3 α ) k = z 0 /2 + ρ 2 x α x β z 0 ∈ H : z 0 = s i n h z 0 ∑ 32 ( cosh z 0 − 1 ) i z 0 3 H 0 1 ( 2 ρ z 0 ) ( ( sinh 2 k − 2 k ) k α w 3 α ) k = z 0 /2 . (118)
noting that the contribution from the poles of order 1 at z = π i ( n + 1 / 2 ) z=\pi i(n+1/2) z = π i ( n + 1/2 ) where n ∈ Z n\in\mathbb{Z} n ∈ Z cancel out with v 3 α v_3^\alpha v 3 α . Hence,
u 3 α = O ( ϵ 3 β α x α x β e − ρ y 1 / 2 ρ 5 / 2 ) . (119) u_3^\alpha=\mathcal{O}\left(\frac{\epsilon_{3\beta\alpha}x_\alpha x_\beta e^{-\rho y_1/2}}{\rho^{5/2}}\right). \tag*{(119)} u 3 α = O ( ρ 5/2 ϵ 3 β α x α x β e − ρ y 1 /2 ) . (119)
Finally, when j , k ∈ [ 1 , 2 ] j,k\in[1,2] j , k ∈ [ 1 , 2 ] , the leading order contribution in the far-field arises from γ ϵ \gamma_\epsilon γ ϵ i.e.
w α α = z ( 2 − z ) [ − ϵ α β 3 2 x β ( 1 − h ) ρ 4 ( 1 − 4 x α 2 ρ 2 ) ] , w β α = z ( 2 − z ) [ − ϵ α β 3 2 x α ( 1 − h ) ρ 4 ( 1 − 4 x β 2 ρ 2 ) ] , (120) w_\alpha^\alpha=z(2-z)\left[-\epsilon_{\alpha\beta3}\frac{2x_\beta(1-h)}{\rho^4}\left(1-\frac{4x_\alpha^2}{\rho^2}\right)\right],\qquad
w_\beta^\alpha=z(2-z)\left[-\epsilon_{\alpha\beta3}\frac{2x_\alpha(1-h)}{\rho^4}\left(1-\frac{4x_\beta^2}{\rho^2}\right)\right], \tag*{(120)} w α α = z ( 2 − z ) [ − ϵ α β 3 ρ 4 2 x β ( 1 − h ) ( 1 − ρ 2 4 x α 2 ) ] , w β α = z ( 2 − z ) [ − ϵ α β 3 ρ 4 2 x α ( 1 − h ) ( 1 − ρ 2 4 x β 2 ) ] , (120)
where β ∈ [ 1 , 2 ] \beta\in[1,2] β ∈ [ 1 , 2 ] and β ≠ α \beta\ne\alpha β = α .
Source Dipole in a Petri Dish
Using fifth order repeated reflection solutions (Appendix A), v j k v_j^k v j k for a source dipole becomes
v j k = δ j k L ( 1 r 3 ) − 3 δ j α δ k β x α x β L ( 1 r 5 ) − 3 ( δ j α δ k 3 + δ k α δ j 3 ) x α L ( z r 5 ) − 3 δ j 3 δ k 3 L ( z 2 r 5 ) = 2 F 2 , 0 ( δ j α δ k β x α x β ρ 2 − δ j 3 δ k 3 ) + 2 F 1 , 1 ρ ( δ j k − 2 δ j α δ k β x α x β ρ 2 − δ j 3 δ k 3 ) − 2 x α G 2 , 1 ρ ( δ j 3 δ k α + δ k 3 δ j α ) . (121) \begin{aligned}
v_j^k={}&\delta_{jk}\mathcal{L}\left(\frac{1}{r^3}\right)-3\delta_{j\alpha}\delta_{k\beta}x_\alpha x_\beta\mathcal{L}\left(\frac{1}{r^5}\right)-3(\delta_{j\alpha}\delta_{k3}+\delta_{k\alpha}\delta_{j3})x_\alpha\mathcal{L}\left(\frac{z}{r^5}\right)-3\delta_{j3}\delta_{k3}\mathcal{L}\left(\frac{z^2}{r^5}\right)\\
={}&2F_{2,0}\left(\frac{\delta_{j\alpha}\delta_{k\beta}x_\alpha x_\beta}{\rho^2}-\delta_{j3}\delta_{k3}\right)+\frac{2F_{1,1}}{\rho}\left(\delta_{jk}-\frac{2\delta_{j\alpha}\delta_{k\beta}x_\alpha x_\beta}{\rho^2}-\delta_{j3}\delta_{k3}\right)-\frac{2x_\alpha G_{2,1}}{\rho}(\delta_{j3}\delta_{k\alpha}+\delta_{k3}\delta_{j\alpha}). \tag*{(121)}
\end{aligned} v j k = = δ j k L ( r 3 1 ) − 3 δ j α δ k β x α x β L ( r 5 1 ) − 3 ( δ j α δ k 3 + δ k α δ j 3 ) x α L ( r 5 z ) − 3 δ j 3 δ k 3 L ( r 5 z 2 ) 2 F 2 , 0 ( ρ 2 δ j α δ k β x α x β − δ j 3 δ k 3 ) + ρ 2 F 1 , 1 ( δ j k − ρ 2 2 δ j α δ k β x α x β − δ j 3 δ k 3 ) − ρ 2 x α G 2 , 1 ( δ j 3 δ k α + δ k 3 δ j α ) . (121)
The boundary conditions for the corresponding auxiliary solution w j k w_j^k w j k and transformed auxiliary solution w ^ j k \widehat{w}_j^k w j k become
w j k ∣ z = 0 = − 2 x α ρ ( δ j 3 δ k α + δ k 3 δ j α ) ∫ 0 ∞ λ 2 d λ J 1 ( λ ρ ) cosh ( 1 − h ) λ cosh λ ⟹ w ^ j k ∣ z = 0 = 4 π i k α ( δ j 3 δ k α + δ k 3 δ j α ) cosh k ( 1 − h ) cosh k . (122) \begin{aligned}
w_j^k\big|_{z=0}&=-\frac{2x_\alpha}{\rho}(\delta_{j3}\delta_{k\alpha}+\delta_{k3}\delta_{j\alpha})\int_0^\infty\lambda^2\,d\lambda\,J_1(\lambda\rho)\frac{\cosh(1-h)\lambda}{\cosh\lambda}\Longrightarrow\\
\widehat{w}_j^k\big|_{z=0}&=4\pi i k_\alpha(\delta_{j3}\delta_{k\alpha}+\delta_{k3}\delta_{j\alpha})\frac{\cosh k(1-h)}{\cosh k}. \tag*{(122)}
\end{aligned} w j k z = 0 w j k z = 0 = − ρ 2 x α ( δ j 3 δ k α + δ k 3 δ j α ) ∫ 0 ∞ λ 2 d λ J 1 ( λ ρ ) cosh λ cosh ( 1 − h ) λ ⟹ = 4 π i k α ( δ j 3 δ k α + δ k 3 δ j α ) cosh k cosh k ( 1 − h ) . (122)
∂ w α k ∂ z ∣ z = 1 = − 2 x α δ k 3 ρ ∫ 0 ∞ λ 2 d λ J 1 ( λ ρ ) λ sinh h λ cosh λ ⟹ ∂ w ^ α k ∂ z ∣ z = 1 = 4 π i k α δ k 3 k sinh h k cosh k , (123) \frac{\partial w_\alpha^k}{\partial z}\bigg|_{z=1}=-\frac{2x_\alpha\delta_{k3}}{\rho}\int_0^\infty\lambda^2\,d\lambda\,J_1(\lambda\rho)\frac{\lambda\sinh h\lambda}{\cosh\lambda}\Longrightarrow\frac{\partial\widehat{w}_\alpha^k}{\partial z}\bigg|_{z=1}=4\pi i k_\alpha\delta_{k3}\frac{k\sinh hk}{\cosh k}, \tag*{(123)} ∂ z ∂ w α k z = 1 = − ρ 2 x α δ k 3 ∫ 0 ∞ λ 2 d λ J 1 ( λ ρ ) cosh λ λ sinh hλ ⟹ ∂ z ∂ w α k z = 1 = 4 π i k α δ k 3 cosh k k sinh hk , (123)
w 3 k ∣ z = 1 = 2 δ k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) λ sinh h λ cosh λ ⟹ w ^ 3 k ∣ z = 1 = 4 π δ k 3 k sinh h k cosh k . (124) w_3^k\big|_{z=1}=2\delta_{k3}\int_0^\infty\lambda\,d\lambda\,J_0(\lambda\rho)\frac{\lambda\sinh h\lambda}{\cosh\lambda}\Longrightarrow\widehat{w}_3^k\big|_{z=1}=4\pi\delta_{k3}\frac{k\sinh hk}{\cosh k}. \tag*{(124)} w 3 k z = 1 = 2 δ k 3 ∫ 0 ∞ λ d λ J 0 ( λ ρ ) cosh λ λ sinh hλ ⟹ w 3 k z = 1 = 4 π δ k 3 cosh k k sinh hk . (124)
We thus obtain
w ^ 3 3 = 4 π k cosh k ( sinh 2 k − 2 k ) ( 2 cosh k sinh h k sinh k z − 2 k sinh k z cosh k ( 1 − h ) + 2 k z sinh k ( 1 − h ) cosh k cosh k ( 1 − z ) ) , (125) \widehat{w}_3^3=\frac{4\pi k}{\cosh k(\sinh2k-2k)}\left(2\cosh k\sinh hk\sinh kz-2k\sinh kz\cosh k(1-h)+2kz\sinh k(1-h)\cosh k\cosh k(1-z)\right), \tag*{(125)} w 3 3 = cosh k ( sinh 2 k − 2 k ) 4 π k ( 2 cosh k sinh hk sinh k z − 2 k sinh k z cosh k ( 1 − h ) + 2 k z sinh k ( 1 − h ) cosh k cosh k ( 1 − z ) ) , (125)
w ^ α 3 = 4 π i k α cosh k ( sinh 2 k − 2 k ) ( sinh 2 k cosh k ( 1 − h − z ) − 2 k cosh k z cosh k ( 1 − h ) − 2 k z cosh k sinh k ( 1 − z ) sinh k ( 1 − h ) ) , (126) \widehat{w}_\alpha^3=\frac{4\pi i k_\alpha}{\cosh k(\sinh2k-2k)}\left(\sinh2k\cosh k(1-h-z)-2k\cosh kz\cosh k(1-h)-2kz\cosh k\sinh k(1-z)\sinh k(1-h)\right), \tag*{(126)} w α 3 = cosh k ( sinh 2 k − 2 k ) 4 π i k α ( sinh 2 k cosh k ( 1 − h − z ) − 2 k cosh k z cosh k ( 1 − h ) − 2 k z cosh k sinh k ( 1 − z ) sinh k ( 1 − h ) ) , (126)
w ^ 3 α = 4 π i k α cosh k ( 1 − h ) cosh k ( sinh 2 k − 2 k ) ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) , (127) \widehat{w}_3^\alpha=\frac{4\pi i k_\alpha\cosh k(1-h)}{\cosh k(\sinh2k-2k)}\left(k(z-2)\cosh kz+kz\cosh k(2-z)+\sinh k(2-z)-\sinh kz\right), \tag*{(127)} w 3 α = cosh k ( sinh 2 k − 2 k ) 4 π i k α cosh k ( 1 − h ) ( k ( z − 2 ) cosh k z + k z cosh k ( 2 − z ) + sinh k ( 2 − z ) − sinh k z ) , (127)
w ^ β α = 4 π k α k β cosh k ( 1 − h ) cosh k ( sinh 2 k − 2 k ) ( z sinh k ( 2 − z ) − ( z − 2 ) sinh k z ) . (128) \widehat{w}_\beta^\alpha=\frac{4\pi k_\alpha k_\beta\cosh k(1-h)}{\cosh k(\sinh2k-2k)}\left(z\sinh k(2-z)-(z-2)\sinh kz\right). \tag*{(128)} w β α = cosh k ( sinh 2 k − 2 k ) 4 π k α k β cosh k ( 1 − h ) ( z sinh k ( 2 − z ) − ( z − 2 ) sinh k z ) . (128)
or, utilizing Hankel transforms,
w 3 3 = 1 2 π H 0 ( w ^ 3 3 ) , w α 3 = i x α 2 π ρ H 1 ( k k α α w ^ α 3 ) , w 3 α = i x α 2 π ρ H 1 ( k k α α w ^ 3 α ) , (129) w_{3}^{3}=\frac{1}{2\pi}\mathcal{H}_{0}\left(\widehat{w}_{3}^{3}\right),\quad w_{\alpha}^{3}=\frac{ix_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(\frac{k}{k_{\alpha\alpha}}\widehat{w}_{\alpha}^{3}\right),\quad w_{3}^{\alpha}=\frac{ix_{\alpha}}{2\pi\rho}\mathcal{H}_{1}\left(\frac{k}{k_{\alpha\alpha}}\widehat{w}_{3}^{\alpha}\right), \tag*{(129)} w 3 3 = 2 π 1 H 0 ( w 3 3 ) , w α 3 = 2 π ρ i x α H 1 ( k α α k w α 3 ) , w 3 α = 2 π ρ i x α H 1 ( k α α k w 3 α ) , (129)
w β α = 1 2 π ( δ α β ρ − 2 x α x β ρ 3 ) H 1 ( k k α β w ^ β α ) + x α x β 2 π ρ 2 H 0 ( k 2 k α β w ^ β α ) . (130) w_{\beta}^{\alpha}=\frac{1}{2\pi}\left(\frac{\delta_{\alpha\beta}}{\rho}-\frac{2x_{\alpha}x_{\beta}}{\rho^{3}}\right)\mathcal{H}_{1}\left(\frac{k}{k_{\alpha\beta}}\widehat{w}_{\beta}^{\alpha}\right)+\frac{x_{\alpha}x_{\beta}}{2\pi\rho^{2}}\mathcal{H}_{0}\left(\frac{k^{2}}{k_{\alpha\beta}}\widehat{w}_{\beta}^{\alpha}\right). \tag*{(130)} w β α = 2 π 1 ( ρ δ α β − ρ 3 2 x α x β ) H 1 ( k α β k w β α ) + 2 π ρ 2 x α x β H 0 ( k α β k 2 w β α ) . (130)
Similarly to the source, F F F has poles of order 1 at z = π i ( n + 1 / 2 ) z=\pi i(n+1/2) z = π i ( n + 1/2 ) where n ∈ Z ≥ n\in\mathbb{Z}_{\geq} n ∈ Z ≥ and poles of order 1 at z = z 0 / 2 z=z_{0}/2 z = z 0 /2 where sinh z 0 = z 0 \sinh z_{0}=z_{0} sinh z 0 = z 0 . When j = k = 3 j=k=3 j = k = 3 , since ∫ γ ϵ \int_{\gamma_{\epsilon}} ∫ γ ϵ vanishes as ϵ → 0 \epsilon\to0 ϵ → 0 , and (36) becomes
w 3 3 = − ∑ n = 1 , 3 , 5 , … ∞ π 2 n 2 sin ( n π h 2 ) sin ( n π z 2 ) K 0 ( n π ρ 2 ) + ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 8 ( cosh z 0 − 1 ) ( w ^ 3 3 ( sinh 2 k − 2 k ) ) ∣ k = z 0 / 2 H 0 1 ( ρ z 0 2 ) , w_{3}^{3}=-\sum_{n=1,3,5,\ldots}^{\infty}\pi^{2}n^{2}\sin\left(\frac{n\pi h}{2}\right)\sin\left(\frac{n\pi z}{2}\right)K_{0}\left(\frac{n\pi\rho}{2}\right)
+\sum_{\substack{z_{0}\in\mathbb{H}:\\ z_{0}=\sinh z_{0}}}\frac{iz_{0}}{8(\cosh z_{0}-1)}\left.\left(\widehat{w}_{3}^{3}(\sinh2k-2k)\right)\right|_{k=z_{0}/2}H_{0}^{1}\left(\frac{\rho z_{0}}{2}\right), w 3 3 = − n = 1 , 3 , 5 , … ∑ ∞ π 2 n 2 sin ( 2 nπ h ) sin ( 2 nπ z ) K 0 ( 2 nπ ρ ) + z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 ( cosh z 0 − 1 ) i z 0 ( w 3 3 ( sinh 2 k − 2 k ) ) k = z 0 /2 H 0 1 ( 2 ρ z 0 ) ,
⟹ u 3 3 = v 3 3 + w 3 3 = ∑ z 0 ∈ H : z 0 = sinh z 0 i z 0 8 ( cosh z 0 − 1 ) ( w ^ 3 3 ( sinh 2 k − 2 k ) ) ∣ k = z 0 / 2 H 0 1 ( ρ z 0 2 ) , (131) \Longrightarrow u_{3}^{3}=v_{3}^{3}+w_{3}^{3}=\sum_{\substack{z_{0}\in\mathbb{H}:\\ z_{0}=\sinh z_{0}}}\frac{iz_{0}}{8(\cosh z_{0}-1)}\left.\left(\widehat{w}_{3}^{3}(\sinh2k-2k)\right)\right|_{k=z_{0}/2}H_{0}^{1}\left(\frac{\rho z_{0}}{2}\right), \tag*{(131)} ⟹ u 3 3 = v 3 3 + w 3 3 = z 0 ∈ H : z 0 = s i n h z 0 ∑ 8 ( cosh z 0 − 1 ) i z 0 ( w 3 3 ( sinh 2 k − 2 k ) ) k = z 0 /2 H 0 1 ( 2 ρ z 0 ) , (131)
Hence, in the far-field the leading order contribution to u 3 3 u_{3}^{3} u 3 3 is
u 3 3 = O ( e − ρ y 1 / 2 ρ 1 / 2 ) . (132) u_{3}^{3}=\mathcal{O}\left(\frac{e^{-\rho y_{1}/2}}{\rho^{1/2}}\right). \tag*{(132)} u 3 3 = O ( ρ 1/2 e − ρ y 1 /2 ) . (132)
The leading order contribution in the far-field is
u α 3 , u 3 α = O ( x α e − ρ y 1 / 2 ρ 3 / 2 ) . (133) u_{\alpha}^{3},u_{3}^{\alpha}=\mathcal{O}\left(\frac{x_{\alpha}e^{-\rho y_{1}/2}}{\rho^{3/2}}\right). \tag*{(133)} u α 3 , u 3 α = O ( ρ 3/2 x α e − ρ y 1 /2 ) . (133)
When j = β j=\beta j = β and k = α k=\alpha k = α where α , β ∈ [ 1 , 2 ] \alpha,\beta\in[1,2] α , β ∈ [ 1 , 2 ] , the leading far-field contribution arises from γ ϵ \gamma_{\epsilon} γ ϵ ,
u β α = z ( 2 − z ) [ 3 ρ 2 ( δ α β − 2 x α x β ρ 2 ) ] . (134) u_{\beta}^{\alpha}=z(2-z)\left[\frac{3}{\rho^{2}}\left(\delta_{\alpha\beta}-\frac{2x_{\alpha}x_{\beta}}{\rho^{2}}\right)\right]. \tag*{(134)} u β α = z ( 2 − z ) [ ρ 2 3 ( δ α β − ρ 2 2 x α x β ) ] . (134)
Vertical Stokeslet Near the Free Surface Boundary
Here we find the leading term in (45) for the horizontal flow field at ( ρ , 0 , h ) (\rho,0,h) ( ρ , 0 , h ) produced by a vertical Stokeslet located at ( 0 , 0 , h ) (0,0,h) ( 0 , 0 , h ) in the limit that ϵ = 1 − h ≪ 1 \epsilon=1-h\ll1 ϵ = 1 − h ≪ 1 . From (32) we have
w 1 3 = − 4 ∫ γ k X ( cosh k ) 2 H 1 1 ( k ρ ) sinh 2 k − 2 k d k , (135) w_{1}^{3}=-4\int_{\gamma}\frac{kX}{(\cosh k)^{2}}\frac{H_{1}^{1}(k\rho)}{\sinh2k-2k}\,dk, \tag*{(135)} w 1 3 = − 4 ∫ γ ( cosh k ) 2 k X sinh 2 k − 2 k H 1 1 ( k ρ ) d k , (135)
where
X = h cosh 2 k sinh 2 h k + k sinh h k cosh h k + h 2 k cosh 2 k ( sinh h k cosh h k + cosh k sinh k ( 1 − 2 h ) ) − cosh k sinh h k sinh k ( 1 + h ) − h cosh 2 k sinh k sinh k ( 1 − 2 h ) + h k cosh k sinh k ( 1 − 2 h ) , ≃ ϵ ( k 2 ( 1 + cosh 2 k ) − 2 cosh 2 k sinh 2 k ) + O ( ϵ 2 ) . (136) \begin{aligned}
X=h\cosh^{2}k\sinh^{2}hk+k\sinh hk\cosh hk+h^{2}k\cosh^{2}k\left(\sinh hk\cosh hk+\cosh k\sinh k(1-2h)\right)
-\cosh k\sinh hk\sinh k(1+h)-h\cosh^{2}k\sinh k\sinh k(1-2h)+hk\cosh k\sinh k(1-2h), \\
&\simeq\epsilon\left(k^{2}(1+\cosh^{2}k)-2\cosh^{2}k\sinh^{2}k\right)+\mathcal{O}(\epsilon^{2}). \tag*{(136)}
\end{aligned} X = h cosh 2 k sinh 2 hk + k sinh hk cosh hk + h 2 k cosh 2 k ( sinh hk cosh hk + cosh k sinh k ( 1 − 2 h ) ) − cosh k sinh hk sinh k ( 1 + h ) − h cosh 2 k sinh k sinh k ( 1 − 2 h ) + hk cosh k sinh k ( 1 − 2 h ) , ≃ ϵ ( k 2 ( 1 + cosh 2 k ) − 2 cosh 2 k sinh 2 k ) + O ( ϵ 2 ) . (136)
The leading order term of u 1 3 u_{1}^{3} u 1 3 is that from w 1 3 w_{1}^{3} w 1 3 which is the sum of the residues at the first two roots in the upper half plane to the equation sinh 2 k = 2 k \sinh2k=2k sinh 2 k = 2 k i.e. k 0 + k_{0}^{+} k 0 + and k 0 − k_{0}^{-} k 0 − where
2 k 0 ± = ± x 1 + i y 1 = 2.769 + 7.498 i . (137) 2k_{0}^{\pm}=\pm x_{1}+iy_{1}=2.769+7.498i. \tag*{(137)} 2 k 0 ± = ± x 1 + i y 1 = 2.769 + 7.498 i . (137)
Hence, using the residue theorem we have
u 1 3 = − 4 π i ( ∑ k 0 ∈ [ k 0 + , k 0 − ] lim k → k 0 ( k − k 0 sinh 2 k − 2 k k H 1 1 ( k ρ ) cosh 2 k X ) ) = − 4 π i ( ∑ k 0 ∈ [ k 0 + , k 0 − ] lim k → k 0 ( 1 4 sinh 2 k k H 1 1 ( k ρ ) cosh 2 k ϵ k 2 sinh 2 k ) ) = − ϵ π i ( k 0 + sinh 2 k 0 + H 1 1 ( k 0 + ρ ) + k 0 − sinh 2 k 0 − H 1 1 ( k 0 − ρ ) ) . (138) u_{1}^{3}=-4\pi i\left(\sum_{k_{0}\in[k_{0}^{+},k_{0}^{-}]} \lim_{k\to k_{0}}\left(\frac{k-k_{0}}{\sinh2k-2k}\frac{kH_{1}^{1}(k\rho)}{\cosh^{2}k}X\right)\right)
= -4\pi i\left(\sum_{k_{0}\in[k_{0}^{+},k_{0}^{-}]} \lim_{k\to k_{0}}\left(\frac{1}{4\sinh^{2}k}\frac{kH_{1}^{1}(k\rho)}{\cosh^{2}k}\epsilon k^{2}\sinh^{2}k\right)\right)
= -\epsilon\pi i\left(k_{0}^{+}\sinh^{2}k_{0}^{+}H_{1}^{1}(k_{0}^{+}\rho)+k_{0}^{-}\sinh^{2}k_{0}^{-}H_{1}^{1}(k_{0}^{-}\rho)\right). \tag*{(138)} u 1 3 = − 4 π i k 0 ∈ [ k 0 + , k 0 − ] ∑ k → k 0 lim ( sinh 2 k − 2 k k − k 0 cosh 2 k k H 1 1 ( k ρ ) X ) = − 4 π i k 0 ∈ [ k 0 + , k 0 − ] ∑ k → k 0 lim ( 4 sinh 2 k 1 cosh 2 k k H 1 1 ( k ρ ) ϵ k 2 sinh 2 k ) = − ϵ π i ( k 0 + sinh 2 k 0 + H 1 1 ( k 0 + ρ ) + k 0 − sinh 2 k 0 − H 1 1 ( k 0 − ρ ) ) . (138)
However, recall the standard result (e.g. see equation 9.2.3 of [39 ] ) that
H 1 1 ( z ) ∼ 2 π z e i ( z − 3 π / 4 ) when ∣ z ∣ → ∞ and − π < arg z < 2 π . (139) H_{1}^{1}(z)\sim\frac{2}{\pi z}e^{i(z-3\pi/4)}\quad\text{when}\quad|z|\to\infty\quad\text{and}\quad-\pi<\arg z<2\pi. \tag*{(139)} H 1 1 ( z ) ∼ π z 2 e i ( z − 3 π /4 ) when ∣ z ∣ → ∞ and − π < arg z < 2 π . (139)
Hence, (F4) simplifies to become
u 1 3 = − π ϵ ρ ( 1 − i ) e − ρ y 1 / 2 ( k 0 + sinh 2 k 0 + e i ρ x 1 / 2 + k 0 − sinh 2 k 0 − e − i ρ x 1 / 2 ) = − π ϵ ρ e − ρ y 1 / 2 ( g + i g ⋆ ) ( 1 − i ) = − 2 π ϵ ρ e − ρ y 1 / 2 ( Re ( g ) + Im ( g ) ) , (140) u_{1}^{3}=-\frac{\sqrt{\pi\epsilon}}{\sqrt{\rho}}(1-i)e^{-\rho y_{1}/2}\left(k_{0}^{+}\sinh^{2}k_{0}^{+}e^{i\rho x_{1}/2}+k_{0}^{-}\sinh^{2}k_{0}^{-}e^{-i\rho x_{1}/2}\right)
= -\frac{\sqrt{\pi\epsilon}}{\sqrt{\rho}}e^{-\rho y_{1}/2}(g+ig^{\star})(1-i)
= -\frac{2\sqrt{\pi\epsilon}}{\sqrt{\rho}}e^{-\rho y_{1}/2}(\operatorname{Re}(g)+\operatorname{Im}(g)), \tag*{(140)} u 1 3 = − ρ π ϵ ( 1 − i ) e − ρ y 1 /2 ( k 0 + sinh 2 k 0 + e i ρ x 1 /2 + k 0 − sinh 2 k 0 − e − i ρ x 1 /2 ) = − ρ π ϵ e − ρ y 1 /2 ( g + i g ⋆ ) ( 1 − i ) = − ρ 2 π ϵ e − ρ y 1 /2 ( Re ( g ) + Im ( g )) , (140)
where g g g satisfies
g = k 0 + sinh 2 k 0 + e i ρ x 1 / 2 = e i ρ x 1 / 2 ( − 2.782 + 7.1238 i ) . (141) g=\sqrt{k_{0}^{+}\sinh^{2}k_{0}^{+}e^{i\rho x_{1}/2}}=e^{i\rho x_{1}/2}(-2.782+7.1238i). \tag*{(141)} g = k 0 + sinh 2 k 0 + e i ρ x 1 /2 = e i ρ x 1 /2 ( − 2.782 + 7.1238 i ) . (141)
Rearranging this expression gives the relation in (52), namely
u α 3 = A ϵ x α ρ 3 / 2 e − ρ y 1 / 2 sin ( x 1 ( ρ − ρ 0 ) / 2 ) , (142) u_{\alpha}^{3}=\frac{A\epsilon x_{\alpha}}{\rho^{3/2}}e^{-\rho y_{1}/2}\sin\left(x_{1}(\rho-\rho_{0})/2\right), \tag*{(142)} u α 3 = ρ 3/2 A ϵ x α e − ρ y 1 /2 sin ( x 1 ( ρ − ρ 0 ) /2 ) , (142)
where A = 38.340 A=38.340 A = 38.340 and ρ 0 = 0.298 \rho_{0}=0.298 ρ 0 = 0.298 .
References