betaFeedback

WorksPhysics and Astronomy

On a soliton with a non-normalisable zero mode

Abstract

We consider the holomorphic baby Skyrme model, which we propose as a simple tractable example of a soliton with a non-normalisable zero mode. The 11-soliton has the same structure as a lump of fixed size. The size modulus of the sigma model becomes massive, while the phase modulus remains massless. The latter is non-normalisable in the (0+1)(0+1)-dimensional sense because of the soliton tail. A suitable way to treat it is to consider the radial (1+1)(1+1)-dimensional effective action on the half-plane. The soliton core sets the boundary condition for the effective model, which we evaluate with the WKB approximation and also numerically. We discuss the analogue of the effective moduli space approximation with a 1+11+1 model and a boundary term at a certain cutoff scale. This may serve as a prototype for more complex systems featuring non-normalisable zero modes such as certain kinds of monopoles in 3+13+1 dimensions.

Introduction

Soliton zero modes play a fundamental role in determining soliton dynamics. Under suitable conditions, namely low velocities and negligible radiation into massless fields, the dynamics is governed by an effective action for the zero modes. The metric on the moduli space is induced by their L2L^2 inner product. This approach has applications in various dimensions and to different types of solitons.

A static soliton in d+1d+1 spacetime dimensions corresponds to a stable solution in dd Euclidean dimensions. If its zero modes are normalisable with respect to the dd-dimensional L2L^2 norm, its low-energy dynamics can be reduced to a (0+1)(0+1)-dimensional system [1]. What happens when a zero mode is non-normalisable? The corresponding (0+1)(0+1)-dimensional description is no longer useful: it assigns infinite inertia to the collective coordinate. Moving the entire soliton rigidly along the zero-mode direction would require infinite energy. A more physically meaningful question is what happens if we give the zero mode an initial push, with a velocity profile that decays sufficiently rapidly at large distances to have finite energy. Is there a low-energy effective description that is simpler than the full (d+1)(d+1)-dimensional dynamics?

A systematic treatment and a common framework for non-normalisable zero modes are still lacking. The best-known example is provided by lumps in (2+1)(2+1)-dimensional models with target space S2S^2 and a purely Dirichlet action. Another example is the magnetic monopole arising from the breaking SU(2)→U(1)SU(2) \to U(1) in hyperbolic space, with spacetime R×H3\mathbb{R} \times\mathbb{H}^3, see for example [2, 3]. A further important case is that of non-Abelian monopoles, for which the unbroken gauge group contains a non-Abelian subgroup, as in SU(3)→SU(2)×U(1)SU(3) \to SU(2) \times U(1) or others, see for example [4, 5, 6, 7, 8, 9]. Different approaches to these models have been developed, motivated by both mathematical and physical considerations.

We first study what may be the simplest example of a soliton with a non-normalisable zero mode. In a particular baby Skyrme model, the soliton solution coincides exactly with a lump of the S2S^2 sigma model at a fixed size [10, 11]. The size modulus is therefore stabilised, avoiding the difficulties it presents in lump dynamics [12, 13, 14, 15, 16]. The S1S^1 phase modulus remains, and its associated zero mode is non-normalisable.

We then consider s-wave fluctuations of the field around the soliton. We focus on the zero-angular-momentum modes associated with S1S^1 transformations. Away from the core, their radial equation has two linearly independent solutions that are non-normalisable as r→∞r \to\infty. When continued into the soliton core, one solution is normalisable at r=0r = 0, while the other one is not. This leads to an effective (1+1)(1 + 1)-dimensional theory for the radial s-wave S1S^1 mode, which can reach the core even at low energies. This effective theory, together with a boundary condition set by the soliton core, plays the role of the moduli-space approximation for a non-normalisable mode. A spatially constant, rigid motion still has infinite energy, whereas the scattering modes of physical interest have finite energy. Up to rescaling, the models form essentially a one-parameter family. We derive the boundary condition analytically in the low- and high-frequency regimes for arbitrary values of the parameter.

The paper is organised as follows. In § 2, we review certain aspects of the holomorphic baby Skyrme model. In § 3 we discuss the S1S^1 phase mode of this model. We conclude in § 4.

From lumps to the holomorphic baby Skyrmion

The baby Skyrme model is a modified non-linear sigma model with additional terms added in order, among other things, to prevent solutions from shrinking to zero size in finite time. We start by reviewing the pure non-linear sigma model mostly following [17]. The non-linear sigma model on the plane with target the round sphere S2S^2 has Lagrangian

L=14∂μϕ⋅∂μϕ(1)\mathcal{L} = \frac{1}{4}\partial_\mu\phi\cdot\partial^\mu\phi \tag*{(1)}

where ϕ:R2→S2⊂R3\phi: \mathbb{R}^2 \to S^2 \subset\mathbb{R}^3 must satisfy ∣ϕ∣2=ϕ12+ϕ22+ϕ32=1|\phi|^2 = \phi_1^2 + \phi_2^2 + \phi_3^2 = 1. The associated static energy is

U=14∫R2dϕ∧∗dϕ(2)U = \frac{1}{4}\int_{\mathbb{R}^2} \mathrm{d}\phi\wedge*\mathrm{d}\phi \tag*{(2)}

where ∗* is the Hodge operator on the domain R2\mathbb{R}^2. We also require ϕ\phi to converge to a constant unit vector at spatial infinity along any direction,

ϕ∞=lim⁡∣x∣→∞ϕ(x).(3)\phi_\infty= \lim_{|x|\to\infty}\phi(x). \tag*{(3)}

Because of (3) we can replace the domain with its one-point compactification and view ϕ\phi as a map ϕ:S2→S2\phi:S^2 \to S^2. As such it has a Brouwer degree which for a smooth field can be calculated via

deg⁡ϕ=14π∫R2ϕ∗(vol⁡S2),(4)\deg\phi=\frac{1}{4\pi}\int_{\mathbb{R}^2}\phi^*(\operatorname{vol}_{S^2}), \tag*{(4)}

where

vol⁡S2=1∣x∣2(x1dx2∧dx3+x2dx3∧dx1+x3dx1∧dx2).(5)\operatorname{vol}_{S^2}=\frac{1}{\lvert x\rvert^2}\left(x_1\mathrm{d}x_2\wedge\mathrm{d}x_3+x_2\mathrm{d}x_3\wedge\mathrm{d}x_1+x_3\mathrm{d}x_1\wedge\mathrm{d}x_2\right). \tag*{(5)}

Equivalently

N:=deg⁡ϕ=18π∫R2ϕ⋅(ϵij∂iϕ×∂jϕ)dx∧dy,(6)N:=\deg\phi=\frac{1}{8\pi}\int_{\mathbb{R}^2}\phi\cdot\left(\epsilon_{ij}\partial_i\phi\times\partial_j\phi\right)\mathrm{d}x\wedge\mathrm{d}y, \tag*{(6)}

where ϵij=−ϵji\epsilon_{ij}=-\epsilon_{ji} and ϵ12=1\epsilon_{12}=1. Since

∣∂iϕ±ϵijϕ×∂jϕ∣2=2∂iϕ⋅∂iϕ∓2ϕ⋅(ϵij∂iϕ×∂jϕ),\left\lvert\partial_i\phi\pm\epsilon_{ij}\phi\times\partial_j\phi\right\rvert^2=2\partial_i\phi\cdot\partial_i\phi\mp2\phi\cdot\left(\epsilon_{ij}\partial_i\phi\times\partial_j\phi\right),

one obtains

0≤18π∫R2∣∂iϕ±ϵijϕ×∂jϕ∣2dx∧dy=U∓2πN.(7)0\leq\frac{1}{8\pi}\int_{\mathbb{R}^2}\left\lvert\partial_i\phi\pm\epsilon_{ij}\phi\times\partial_j\phi\right\rvert^2\mathrm{d}x\wedge\mathrm{d}y=U\mp2\pi N. \tag*{(7)}

Therefore

U≥2π∣N∣,(8)U\geq2\pi\lvert N\rvert, \tag*{(8)}

a Bogomolny-type bound, saturated if and only if

∂iϕ±ϵijϕ×∂jϕ=0,(9)\partial_i\phi\pm\epsilon_{ij}\phi\times\partial_j\phi=0, \tag*{(9)}

with the upper (respectively lower) sign if N>0N>0 (N<0N<0). Solutions of (9) are absolute minima of the energy in each given topological sector. Of course solutions of the first order equation (9) also solve the second order variational equations associated to UU. For N=0N=0 both equations (9) need to be satisfied so ∂iϕ=0\partial_i\phi=0 and ϕ=ϕ∞\phi=\phi_\infty is constant. This is the vacuum state of the theory.

It is convenient to get rid of the constraint ∣ϕ∣2=1\lvert\phi\rvert^2=1 by using stereographic coordinates on the target space. Projecting from the south pole we define

w=ϕ1+iϕ21+ϕ3.(10)w=\frac{\phi_1+i\phi_2}{1+\phi_3}. \tag*{(10)}

We also write z=x+iyz=x+iy on the domain R2≃C\mathbb{R}^2\simeq\mathbb{C}. Then

U=∫Cdw∧∗dwˉ(1+∣w∣2)2=i∫C(∣∂zw∣2+∣∂zˉw∣2)dz∧dzˉ(1+∣w∣2)2,(11)U=\int_{\mathbb{C}}\frac{\mathrm{d}w\wedge*\mathrm{d}\bar{w}}{(1+\lvert w\rvert^2)^2} =i\int_{\mathbb{C}}\frac{\left(\lvert\partial_z w\rvert^2+\lvert\partial_{\bar{z}}w\rvert^2\right)\mathrm{d}z\wedge\mathrm{d}\bar{z}}{(1+\lvert w\rvert^2)^2}, \tag*{(11)}
N=i2π∫Cdw∧dwˉ(1+∣w∣2)2=i2π∫C(∣∂zw∣2−∣∂zˉw∣2)dz∧dzˉ(1+∣w∣2)2.(12)N=\frac{i}{2\pi}\int_{\mathbb{C}}\frac{\mathrm{d}w\wedge\mathrm{d}\bar{w}}{(1+\lvert w\rvert^2)^2} =\frac{i}{2\pi}\int_{\mathbb{C}}\frac{\left(\lvert\partial_z w\rvert^2-\lvert\partial_{\bar{z}}w\rvert^2\right)\mathrm{d}z\wedge\mathrm{d}\bar{z}}{(1+\lvert w\rvert^2)^2}. \tag*{(12)}

From (11), (12) we reobtain U≥2π∣N∣U\geq2\pi\lvert N\rvert, with equality if and only if ww is holomorphic (N>0)(N > 0) or anti-holomorphic (N<0)(N < 0). These conditions can be checked to be equivalent to (9). For N≥0N \ge0 solutions of the Bogomolny equations, often called lumps, are thus meromorphic functions on S2S^2, that is rational functions

w(z)=p(z)q(z)(13)w(z) = \frac{p(z)}{q(z)} \tag*{(13)}

with p≠0p \ne0, qq coprime polynomials with complex coefficients. Without loss of generality we can take ϕ∞=(0,0,1)\phi_\infty= (0,0,1), which translates to

w(∞)=0⇔deg⁡(p)<deg⁡(q),(14)w(\infty) = 0 \Leftrightarrow\deg(p) < \deg(q), \tag*{(14)}

and qq to be a monic polynomial. The general solution of degree NN then takes the form

w(z)=p0+p1z+⋯+pN−1zN−1q0+q1z+⋯+qN−1zN−1+zN.(15)w(z) = \frac{p_0 + p_1z + \cdots+ p_{N-1}z^{N-1}}{q_0 + q_1z + \cdots+ q_{N-1}z^{N-1} + z^N}. \tag*{(15)}

The 2N2N complex parameters {pi,qi}\{p_i,q_i\} are subject to the constraint imposing that the two polynomials have no common factor which is an open condition in C2N\mathbb{C}^{2N}. The moduli space of this sigma model is therefore a complex manifold of dimension 2N2N. Solutions with N<0N < 0 are obtained by taking the complex conjugate of (15).

For N=1N = 1 we have

w(z)=λeiχz−z0(16)w(z) = \frac{\lambda e^{i\chi}}{z-z_0} \tag*{(16)}

with z0∈Cz_0 \in\mathbb{C}, λ∈(0,∞)\lambda\in(0,\infty), χ∈[0,2π)\chi\in[0,2\pi). The energy of the single lump is

U=2λ2∫Cdx∧dy(∣z−z0∣2+λ2)2=4πλ2∫0∞ρ dρ(ρ2+λ2)2=2π,(17)U = 2\lambda^2 \int_{\mathbb{C}} \frac{\mathrm{d}x \wedge\mathrm{d}y}{(|z-z_0|^2+\lambda^2)^2} = 4\pi\lambda^2 \int_0^\infty\frac{\rho\,\mathrm{d}\rho}{(\rho^2+\lambda^2)^2} = 2\pi, \tag*{(17)}

having set x−x0=ρcos⁡θx-x_0 = \rho\cos\theta, y−y0=ρsin⁡θy-y_0 = \rho\sin\theta. The energy density (ρ2+λ2)−2(\rho^2+\lambda^2)^{-2} is strictly decreasing with ρ\rho with an absolute maximum at ρ=0⇔z=z0\rho= 0 \Leftrightarrow z = z_0 which can be interpreted as the position of the lump, giving two real moduli. The real parameter λ\lambda can be interpreted as the scale of the lump: exactly 12\frac{1}{2} of the energy is contained in the disk of centre z0z_0 and radius λ\lambda. The fourth real modulus is the internal phase χ\chi.

Introducing time, the kinetic energy of the model is

T=∫R2∣w˙∣2(1+∣w∣2)2 dx∧dy.(18)T = \int_{\mathbb{R}^2} \frac{|\dot{w}|^2}{(1+|w|^2)^2}\,\mathrm{d}x \wedge\mathrm{d}y. \tag*{(18)}

Consider the N=1N = 1 solution (16). Following the adiabatic approximation, let us assume that ww depends on time only through its moduli. If the only time dependent modulus is λ\lambda then the kinetic energy is, using polar coordinates centred at z0z_0,

T=2πλ˙2∫0∞ρ3 dρ(ρ2+λ2)2=2πλ˙212[log⁡(λ2+ρ2)−ρ2λ2+ρ2]0∞(19)T = 2\pi\dot{\lambda}^2 \int_0^\infty\frac{\rho^3\,\mathrm{d}\rho}{(\rho^2+\lambda^2)^2} = 2\pi\dot{\lambda}^2 \frac{1}{2}\left[\log(\lambda^2+\rho^2)-\frac{\rho^2}{\lambda^2+\rho^2}\right]_0^\infty \tag*{(19)}

which is logarithmically divergent for large ρ\rho. Similarly TT diverges if χ\chi varies in time. The integral for z0z_0 varying in time is finite instead. Hence in the adiabatic approximation the centre of the lump can move but its radius or phase cannot change.

For higher order solutions of the form w=(λ/z)Nw=(\lambda/z)^N the kinetic energy is some constant multiple of λ˙2\dot{\lambda}^{2} implying that, within the adiabatic approximation scheme, the parameter λ\lambda can vary linearly in time, leading to lumps collapsing to a point in finite time [18]. For N=1N=1 the adiabatic approximation is not available since λ\lambda has infinite inertia but collapse still occurs [19, 20]. This behaviour can be traced to the invariance of the theory under conformal transformations of the domain. In order to stabilise the model one needs to break this scale invariance. In the baby Skyrme model [21] the Lagrangian is modified by the addition of terms of order 4 and 0 in the derivatives,

L=14∂μϕ⋅∂μϕ−η4(∂μϕ×∂νϕ)⋅(∂μϕ×∂νϕ)−V(ϕ).(20)\mathcal{L}=\frac{1}{4}\partial^\mu\phi\cdot\partial_\mu\phi-\frac{\eta}{4}(\partial_\mu\phi\times\partial_\nu\phi)\cdot(\partial^\mu\phi\times\partial^\nu\phi)-V(\phi). \tag*{(20)}

Here η\eta is a positive constant and it is assumed that ∣ϕ∣2=1|\phi|^2=1. The topological structure of the baby Skyrme model is the same as that of the pure non-linear sigma model we just reviewed, but since all the additional energy contributions are, for a non-vacuum solution, positive, the Bogomolny bound (8) is never saturated. The term of order 4 in the derivatives is known as the Skyrme term. By Derrick’s argument, the Skyrme term alone does not allow for stable solitons in 2+12+1 dimensions, hence the addition of the potential term V(ϕ)V(\phi).

The potential term is somewhat arbitrary, with the qualitative behaviour of solutions depending strongly on it. Potentials which have been studied include

V(ϕ)=C(1−n⋅ϕ)k,(21)V(\phi)=C(1-n\cdot\phi)^k, \tag*{(21)}

with CC a positive constant and nn a unit vector. In this paper we will focus on the case k=4k=4, known as the holomorphic model [10, 22]. More precisely we will take

V(ϕ)=θ416(1−ϕ3)4.(22)V(\phi)=\frac{\theta_4}{16}(1-\phi_3)^4. \tag*{(22)}

Upon stereographic projection (10) the Lagrangian (20) becomes

L=1(1+∣w∣2)2(∣wt∣2−∣wx∣2−∣wy∣2)+2η(1+∣w∣2)4((wˉxwy−wxwˉy)2−(wˉtwx−wtwˉx)2−(wˉtwy−wtwˉy)2)−θ4(∣w∣21+∣w∣2)4,(23)\begin{aligned} \mathcal{L}={}&\frac{1}{(1+|w|^2)^2}\left(|w_t|^2-|w_x|^2-|w_y|^2\right)\\ &+\frac{2\eta}{(1+|w|^2)^4}\left((\bar{w}_xw_y-w_x\bar{w}_y)^2-(\bar{w}_tw_x-w_t\bar{w}_x)^2-(\bar{w}_tw_y-w_t\bar{w}_y)^2\right)\\ &-\theta_4\left(\frac{|w|^2}{1+|w|^2}\right)^4, \tag*{(23)} \end{aligned}

where wt=∂tww_t = \partial_t w, etc. The equations of motion obtained from (23) are

wtt−wxx−wyy−2wˉ1+∣w∣2(wt2−wx2−wy2)+4η(1+∣w∣2)2{2wˉtywtwy+2wˉtxwtwx−2wˉxywxwy+wˉxx(wy2−wt2)+wˉyy(wx2−wt2)−wˉtt(wx2+wy2)+wxx(∣wt∣2−∣wy∣2)+wyy(∣wt∣2−∣wx∣2)+wtt(∣wx∣2+∣wy∣2)+wxy(wˉxwy+wˉywx)−wtx(wˉtwx+wˉxwt)−wty(wˉtwy+wˉywt)}+8ηw(1+∣w∣2)3((wˉtwy−wˉywt)2+(wˉxwt−wˉtwx)2−(wˉxwy−wˉywx)2)+4θ4(∣w∣21+∣w∣2)3w=0.(24)\begin{aligned} w_{tt} - w_{xx} - w_{yy} - \frac{2\bar{w}}{1+|w|^2}\left(w_t^2-w_x^2-w_y^2\right) &+ \frac{4\eta}{(1+|w|^2)^2}\left\{2\bar{w}_{ty}w_tw_y+2\bar{w}_{tx}w_tw_x-2\bar{w}_{xy}w_xw_y \right.\\ &\quad+\bar{w}_{xx}\left(w_y^2-w_t^2\right)+\bar{w}_{yy}\left(w_x^2-w_t^2\right)-\bar{w}_{tt}\left(w_x^2+w_y^2\right)\\ &\quad+w_{xx}\left(|w_t|^2-|w_y|^2\right)+w_{yy}\left(|w_t|^2-|w_x|^2\right)+w_{tt}\left(|w_x|^2+|w_y|^2\right)\\ &\quad\left.+w_{xy}\left(\bar{w}_xw_y+\bar{w}_yw_x\right)-w_{tx}\left(\bar{w}_tw_x+\bar{w}_xw_t\right)-w_{ty}\left(\bar{w}_tw_y+\bar{w}_yw_t\right)\right\}\\ &+ \frac{8\eta w}{(1+|w|^2)^3}\left(\left(\bar{w}_tw_y-\bar{w}_yw_t\right)^2+\left(\bar{w}_xw_t-\bar{w}_tw_x\right)^2-\left(\bar{w}_xw_y-\bar{w}_yw_x\right)^2\right)\\ &+4\theta_4\left(\frac{|w|^2}{1+|w|^2}\right)^3w=0. \tag*{(24)} \end{aligned}

A particular feature of the holomorphic model is that it retains a 3-parameter family of the solutions of the Bogomolny equations for the pure sigma model: the function

w=λeiχz−z0(25)w=\frac{\lambda e^{i\chi}}{z-z_0} \tag*{(25)}

is a solution with N=1N=1 provided that

λ=(8ηθ4)1/4.(26)\lambda=\left(\frac{8\eta}{\theta_4}\right)^{1/4}. \tag*{(26)}

Hence compared with the N=1N=1 pure sigma model case, three of the four moduli survive while the lump’s radius is fixed by (26). Just as for the pure sigma model, χ\chi has infinite inertia. For ∣N∣>1|N|>1 rational functions do not solve (24). From now on we set z0=0z_0=0.

Effective Lagrangian for the non-normalisable mode

Let us consider the ansatz

w=w0eiα(t,r)(27)w=w_0e^{i\alpha(t,r)} \tag*{(27)}

where w0w_0 is the unperturbed solution (25) and α\alpha a small perturbation. We are taking the perturbation to be radial only, i.e. independent of the polar angle ϑ\vartheta on the domain R2\mathbb{R}^2. Substituting in (23) and dropping α\alpha-independent terms, which do not contribute to the equations of motion, gives

L=λ2r2(r2+λ2)2[(1+8ηλ2(r2+λ2)2)αt2−αr2].(28)\mathcal{L}=\frac{\lambda^2r^2}{(r^2+\lambda^2)^2}\left[\left(1+\frac{8\eta\lambda^2}{(r^2+\lambda^2)^2}\right)\alpha_t^2-\alpha_r^2\right]. \tag*{(28)}

Let us switch to the dimensionless variable x=r/λx = r/\lambda, τ=t/λ\tau= t/\lambda and write h′=λdh/drh' = \lambda dh/dr, h˙=dh/dτ\dot{h} = dh/d\tau. Then

Lvol=[x2(1+x2)2(α˙2−α′2)+μx2(1+x2)4α˙2]x dx∧dϑ=f(x)(h(x)α˙2−α′2)dx∧dϑ,(29)\begin{aligned} \mathcal{L}_{\mathrm{vol}} &= \left[\frac{x^2}{(1+x^2)^2}\left(\dot{\alpha}^2-\alpha'^2\right)+\frac{\mu x^2}{(1+x^2)^4}\dot{\alpha}^2\right]x\,dx\wedge d\vartheta\\ &=f(x)\left(h(x)\dot{\alpha}^2-\alpha'^2\right)dx\wedge d\vartheta, \tag*{(29)} \end{aligned}

where

μ=8ηλ2(30)\mu=\frac{8\eta}{\lambda^2} \tag*{(30)}

and

f(x)=x3(1+x2)2,h(x)=1+μ(1+x2)2.(31)f(x)=\frac{x^3}{(1+x^2)^2},\qquad h(x)=1+\frac{\mu}{(1+x^2)^2}. \tag*{(31)}

The corresponding equations of motion are

fhα¨−(fα′)′=f(hα¨−α′′)−f′α′=0.(32)fh\ddot{\alpha}-(f\alpha')'=f(h\ddot{\alpha}-\alpha'')-f'\alpha'=0. \tag*{(32)}

In particular if α¨=0\ddot{\alpha}=0 then

α(x)=C1∫dxf+C2.(33)\alpha(x)=C_1\int\frac{dx}{f}+C_2. \tag*{(33)}

For ff in (31) we get

α(x)=C1(x22+2log⁡x−12x2)+C2.(34)\alpha(x)=C_1\left(\frac{x^2}{2}+2\log x-\frac{1}{2x^2}\right)+C_2. \tag*{(34)}

Stationary solutions of the form α=a(x)exp⁡(iϵτ)\alpha=a(x)\exp(i\epsilon\tau), where ϵ=λω\epsilon=\lambda\omega is the dimensionless frequency, need to satisfy the ODE

ϵ2fha+(fa′)′=0.(35)\epsilon^2 f h a+(f a')'=0. \tag*{(35)}

Studying the ODE

We are now going to look at the approximate solutions which can be obtained in the two limits r≫λr\gg\lambda and r≪λr\ll\lambda, that is x≫1x\gg1 and x≪1x\ll1. We first rewrite (35) in the form

d2adx2+P(x)dadx+ϵ2Q(x)a=0,(36)\frac{d^2a}{dx^2}+P(x)\frac{da}{dx}+\epsilon^2Q(x)a=0, \tag*{(36)}

with

P=1fdfdx=3−x2x(1+x2),Q=1+μ(1+x2)2.(37)P=\frac{1}{f}\frac{df}{dx}=\frac{3-x^2}{x(1+x^2)},\qquad Q=1+\frac{\mu}{(1+x^2)^2}. \tag*{(37)}

For x≪1x \ll1 we have

Lvol=[x3((1+μ)α˙2−α′2)+O(x5)]dx∧dϑ.(38)\mathcal{L}_{\mathrm{vol}}=\left[x^{3}\left((1+\mu)\dot{\alpha}^{2}-\alpha'^{2}\right)+O\left(x^{5}\right)\right]\mathrm{d}x\wedge\mathrm{d}\vartheta. \tag*{(38)}

A stationary solution α¨=−ϵ2α\ddot{\alpha}=-\epsilon^{2}\alpha has to satisfy the approximate ODE

a′′+3a′x+ϵ2(1+μ)a=0.(39)a''+\frac{3a'}{x}+\epsilon^{2}(1+\mu)a=0. \tag*{(39)}

The solution can be written in terms of Bessel functions,

anear(x)=1x[C1J1(ϵ(1+μ)1/2x)+C2Y1(ϵ(1+μ)1/2x)].(40)a_{\mathrm{near}}(x)=\frac{1}{x}\left[C_{1}J_{1}\left(\epsilon(1+\mu)^{1/2}x\right)+C_{2}Y_{1}\left(\epsilon(1+\mu)^{1/2}x\right)\right]. \tag*{(40)}

Since, for small xx,

J1(Kx)=K2x+O(x3),Y1(Kx)=−2πKx+Kxπlog⁡(Kx2)+Kx2π(2γ−1)+O(x3log⁡x),(41)J_{1}(Kx)=\frac{K}{2}x+O\left(x^{3}\right),\qquad Y_{1}(Kx)=-\frac{2}{\pi Kx}+\frac{Kx}{\pi}\log\left(\frac{Kx}{2}\right)+\frac{Kx}{2\pi}(2\gamma-1)+O\left(x^{3}\log x\right), \tag*{(41)}

where γ\gamma is the Euler–Mascheroni constant, we need to take C2=0C_{2}=0 in order to have a solution which is well-defined at x=0x=0. Then aneara_{\mathrm{near}} is constant to leading order near x=0x=0,

anear=ϵC11+μ2+O(x2).(42)a_{\mathrm{near}}=\frac{\epsilon C_{1}\sqrt{1+\mu}}{2}+O\left(x^{2}\right). \tag*{(42)}

Since

lim⁡x→0(x−1J1(Kx))′=0,(43)\lim_{x\to0}\left(x^{-1}J_{1}(Kx)\right)'=0, \tag*{(43)}

while the derivative of x−1Y1(Kx)x^{-1}Y_{1}(Kx) diverges as x→0+x\to0^{+}, the condition C2=0C_{2}=0 is a boundary condition of Neumann type,

lim⁡x→0+anear′(x)=0.(44)\lim_{x\to0^{+}}a_{\mathrm{near}}'(x)=0. \tag*{(44)}

In the opposite regime x≫1x\gg1 we have

Lvol=[1x(α˙2−α′2)+O(x−3)]dx∧dϑ.(45)\mathcal{L}_{\mathrm{vol}}=\left[\frac{1}{x}\left(\dot{\alpha}^{2}-\alpha'^{2}\right)+O\left(x^{-3}\right)\right]\mathrm{d}x\wedge\mathrm{d}\vartheta. \tag*{(45)}

The corresponding stationary ODE is

a′′−a′x+ϵ2a=0,(46)a''-\frac{a'}{x}+\epsilon^{2}a=0, \tag*{(46)}

which has solution

afar(x)=x(A1J1(ϵx)+A2Y1(ϵx))(47)a_{\mathrm{far}}(x)=x\left(A_{1}J_{1}(\epsilon x)+A_{2}Y_{1}(\epsilon x)\right) \tag*{(47)}
=2x(A12+A22)πϵcos⁡(ϵx+Θ)+O(x−1/2),(48)=\sqrt{\frac{2x\left(A_{1}^{2}+A_{2}^{2}\right)}{\pi\epsilon}}\cos(\epsilon x+\Theta)+O\left(x^{-1/2}\right), \tag*{(48)}

where

tan⁡Θ=A1−A2A1+A2⟺A1A2=1+tan⁡Θ1−tan⁡Θ.(49)\tan\Theta= \frac{A_1-A_2}{A_1+A_2} \Longleftrightarrow\frac{A_1}{A_2} = \frac{1+\tan\Theta}{1-\tan\Theta}. \tag*{(49)}

Let us also look at the small xx behaviour of afara_{\mathrm{far}}. To second order the Taylor series near x=0x=0 is

afar=−2A2πϵ+ϵA2πx2log⁡(xϵ2)+ϵ2π[πA1+(2γ−1)A2]x2+O(x4).(50)a_{\mathrm{far}}=-\frac{2A_2}{\pi\epsilon}+\frac{\epsilon A_2}{\pi}x^2\log\left(\frac{x\epsilon}{2}\right)+\frac{\epsilon}{2\pi}\left[\pi A_1+(2\gamma-1)A_2\right]x^2+O(x^4). \tag*{(50)}

In particular

afar(0)=−2A2πϵ.(51)a_{\mathrm{far}}(0)=-\frac{2A_2}{\pi\epsilon}. \tag*{(51)}

Neumann conditions (44) for the “near” solution (40) imply a constraint for the “far” solution (47) and thus a relation between the two coefficients A1A_1, A2A_2. In Sections 3.2 and 3.3 we show how to derive an analytic approximation for this ratio in the large- and small-ϵ\epsilon regimes, see equations (88) and (98). To check our analytical approximation we also calculate the ratio numerically by integrating the exact ODE (36) using Mathematica. The initial conditions that we want to impose are a′(0)=0a'(0)=0 and, say, a(0)=1a(0)=1 (since the space of solutions is 1-dimensional, the value of a(0)a(0) has no effect on the asymptotic ratio A1/A2A_1/A_2). In practice, since the ODE coefficient (log⁡f)′(\log f)' diverges for x→0x\to0, we impose the initial conditions at x0=10−4x_0=10^{-4} using the expansion anear(x)=1−18ϵ2(1+μ)x2+O(x4)a_{\mathrm{near}}(x)=1-\frac{1}{8}\epsilon^2(1+\mu)x^2+O(x^4), normalised so that anear(0)=1a_{\mathrm{near}}(0)=1. We integrate (36) up to xmax⁡=xfit+12π/ϵx_{\max}=x_{\mathrm{fit}}+12\pi/\epsilon, where xfit=max⁡(15,10/ϵ)x_{\mathrm{fit}}=\max(15,10/\epsilon), and determine A1A_1, A2A_2 by a least-squares fit of the solution to xJ1(ϵx)xJ_1(\epsilon x) and xY1(ϵx)xY_1(\epsilon x) over the six periods in [xfit,xmax⁡][x_{\mathrm{fit}},x_{\max}]. Note that for ϵ≪1\epsilon\ll1 the lump core x≲1x\lesssim1 contributes a correction of order ϵ2\epsilon^2 relative to a(0)=1a(0)=1, so the integration tolerance must be well below this value.

The large-ϵ\epsilon regime

For ϵ≫1\epsilon\gg1 we can study the ODE (36) using the WKB method. First write

a=ψexp⁡(−12∫P(x) dx)=(1+x2x3/2)ψ,(52)a=\psi\exp\left(-\frac{1}{2}\int P(x)\,\mathrm{d}x\right)=\left(\frac{1+x^2}{x^{3/2}}\right)\psi, \tag*{(52)}

under which the full ODE becomes

ψ′′+Qψ=0,(53)\psi''+\mathcal{Q}\psi=0, \tag*{(53)}

where

Q=ϵ2Q−V,(54)\mathcal{Q}=\epsilon^2Q-V, \tag*{(54)}
V=12P′+14P2=34x2−8(1+x2)2.(55)V=\frac{1}{2}P'+\frac{1}{4}P^2=\frac{3}{4x^2}-\frac{8}{(1+x^2)^2}. \tag*{(55)}

and PP, QQ are given by (37). For perturbed spherical Schrödinger operators of this type see [23]. We also define

n=Q=1+μ(1+x2)2,n0=n(0)=1+μ.(56)n=\sqrt{Q}=\sqrt{1+\frac{\mu}{(1+x^2)^2}},\qquad n_0=n(0)=\sqrt{1+\mu}. \tag*{(56)}

Perturbative WKB expansion up to order O(ϵ−1)O(\epsilon^{-1}) for an ODE of the form (53) gives

ψ=ϵQ1/4(K1eiϕ+K2e−iϕ),(57)\psi=\frac{\sqrt{\epsilon}}{Q^{1/4}}\left(K_1e^{i\phi}+K_2e^{-i\phi}\right), \tag*{(57)}

where K1K_1, K2K_2 are arbitrary integration constants and

φ=ϵn+1ϵ(−V2n+3n′2−2nn′′8n3)+O(ϵ−3),(58)\varphi=\epsilon n+\frac{1}{\epsilon}\left(-\frac{V}{2n}+\frac{3n'^2-2nn''}{8n^3}\right)+O(\epsilon^{-3}), \tag*{(58)}
ϕ(x)=∫φ(x) dx.(59)\phi(x)=\int\varphi(x)\,\mathrm{d}x. \tag*{(59)}

By (52), for large xx, ψ=afar/x\psi=a_{\mathrm{far}}/\sqrt{x}. We can then use (48) to write

ψ=2(A12+A22)πϵcos⁡(ϵx+Θ)+O(x−1),(60)\psi=\sqrt{\frac{2(A_1^2+A_2^2)}{\pi\epsilon}}\cos(\epsilon x+\Theta)+O(x^{-1}), \tag*{(60)}

hence the constant Θ\Theta can be read from the asymptotic phase shift of ψ\psi,

Θ=lim⁡x→∞ϕ(x)−ϵx.(61)\Theta=\lim_{x\to\infty}\phi(x)-\epsilon x. \tag*{(61)}

We want to calculate Θ\Theta more explicitly. Up to order ϵ0\epsilon^0 the potential QQ is strictly positive (no turning points), so the term Q1/4Q^{1/4} is real and does not contribute to the phase of ψ\psi. We then simply have

Θ=ϵI(μ),(62)\Theta=\epsilon I(\mu), \tag*{(62)}

with

I(μ)=∫0∞(n−1) dx=∫0∞(1+μ(1+x2)2−1) dx.(63)I(\mu)=\int_0^\infty(n-1)\,\mathrm{d}x=\int_0^\infty\left(\sqrt{1+\frac{\mu}{(1+x^2)^2}}-1\right)\,\mathrm{d}x. \tag*{(63)}

From (49) we then see that A1/A2A_1/A_2 is quasi-periodic in ϵ\epsilon with (quasi) period

T=π∣dΘ/dϵ∣=πI(μ).(64)T=\frac{\pi}{|\mathrm{d}\Theta/\mathrm{d}\epsilon|}=\frac{\pi}{I(\mu)}. \tag*{(64)}

In particular A1/A2A_1/A_2 has poles whenever A2=0⇔Θ=π/4A_2=0\Leftrightarrow\Theta=\pi/4 (mod π\pi). Due to the quasi-periodicity of Θ\Theta the poles are also quasi-periodic, with spacing given by TT.

At order ϵ−1\epsilon^{-1} the potential QQ has a single zero x∗x_\ast given by

x∗=34(1+μ)1ϵ+O(ϵ−3),(65)x_\ast=\sqrt{\frac{3}{4(1+\mu)}\frac{1}{\epsilon}}+O(\epsilon^{-3}), \tag*{(65)}

so there is a phase contribution from the tunneling region 0≤x≤x∗0\leq x\leq x_\ast. For x≥x∗x\geq x_\ast let us define more generally the phase shift Θ(x)\Theta(x) as

Θ(x)=Θ(x∗)+∫x∗x(φ(t)−ϵ) dt=Θ(x∗)+∫x∗x[ϵ(n−1)+1ϵ(−V2n+3n′2−2nn′′8n3)]dt+O(ϵ−3),(66)\begin{aligned} \Theta(x)=\Theta(x_\ast)+\int_{x_\ast}^{x}\left(\varphi(t)-\epsilon\right)\,\mathrm{d}t \\ &=\Theta(x_\ast)+\int_{x_\ast}^{x}\left[\epsilon(n-1)+\frac{1}{\epsilon}\left(-\frac{V}{2n}+\frac{3n'^2-2nn''}{8n^3}\right)\right]\mathrm{d}t+O(\epsilon^{-3}), \tag*{(66)} \end{aligned}

where Θ(x∗)\Theta(x_\ast) represents the contribution to the asymptotic phase due to the tunneling region 0≤x<x∗0\leq x<x_\ast. Note that Θ≡Θ(∞)\Theta\equiv\Theta(\infty).

Near x=0x=0 we have

Q=ϵ2(1+μ)+8−34x2+O(x2),(67)Q=\epsilon^2(1+\mu)+8-\frac{3}{4x^2}+O(x^2), \tag*{(67)}

so the centrifugal term −34x2-\frac{3}{4x^2} dominates and cannot be treated perturbatively. Following a Bessel comparison-equation method [24] of a centrifugal singularity, we approximate ψ\psi near x=0x=0 by the exact solution of a reference ODE carrying an identical singularity. We therefore introduce

ψref′′+Qrefψref=0,(68)\psi_{\mathrm{ref}}''+Q_{\mathrm{ref}}\psi_{\mathrm{ref}}=0, \tag*{(68)}
Qref=κ2−34x2,(69)Q_{\mathrm{ref}}=\kappa^2-\frac{3}{4x^2}, \tag*{(69)}
κ=ϵ2(1+μ)+8=n0ϵ+4n0ϵ+O(ϵ−3),(70)\kappa=\sqrt{\epsilon^2(1+\mu)+8}=n_0\epsilon+\frac{4}{n_0\epsilon}+O(\epsilon^{-3}), \tag*{(70)}

whose potential reproduces the leading x→0x\to0 behaviour of QQ in (67). The solution of the ODE

ψ′′+(κ2−34x2)ψ=0(71)\psi''+\left(\kappa^2-\frac{3}{4x^2}\right)\psi=0 \tag*{(71)}

which is regular at x=0x=0 is

xJ1(κx)≃2κπcos⁡(κx−3π4+38κx).(72)\sqrt{x}J_1(\kappa x)\simeq\sqrt{\frac{2}{\kappa\pi}}\cos\left(\kappa x-\frac{3\pi}{4}+\frac{3}{8\kappa x}\right). \tag*{(72)}

The potential QrefQ_{\mathrm{ref}} has, to this order, the same turning point (65). Let xmx_m be a point such that xm≪ϵ−1/3x_m\ll\epsilon^{-1/3}, so that using the solution xJ1(κx)\sqrt{x}J_1(\kappa x) of the approximate ODE for small distances is justified, and ϵxm≫1\epsilon x_m\gg1, so that using the large distance asymptotics (72) is justified. Then we can write

ϕ(xm)=−3π4+κxm+38κxm+o(1)=−3π4+ϵn0xm+38ϵn0xm+4xmϵn0+O(ϵ−3)(73)\phi(x_m)=-\frac{3\pi}{4}+\kappa x_m+\frac{3}{8\kappa x_m}+o(1)=-\frac{3\pi}{4}+\epsilon n_0x_m+\frac{3}{8\epsilon n_0x_m}+\frac{4x_m}{\epsilon n_0}+O(\epsilon^{-3}) \tag*{(73)}

where the first equality comes from identifying ϕ(xm)\phi(x_m) with the argument of cos⁡\cos in (72), and the second one uses (70). In particular the WKB phase ϕref\phi_{\mathrm{ref}} accumulated by the reference solutions on top of its asymptotic κx\kappa x oscillatory behaviour is

lim⁡x→∞(ϕref(x)−κx)=−3π4.(74)\lim_{x\to\infty}\left(\phi_{\mathrm{ref}}(x)-\kappa x\right)=-\frac{3\pi}{4}. \tag*{(74)}

Supplying this constant, which is the contribution of the tunneling region to the WKB phase, is the reason for introducing ψref\psi_{\mathrm{ref}}. We then calculate

Θ=Θ(∞)−Θ(xm)+Θ(xm)=Θ(xm)+∫xm∞(φ−ϵ) dx.(75)\Theta=\Theta(\infty)-\Theta(x_m)+\Theta(x_m)=\Theta(x_m)+\int_{x_m}^{\infty}(\varphi-\epsilon)\,\mathrm{d}x. \tag*{(75)}

The term Θ(xm)\Theta(x_m) represents the phase accumulated up to xmx_m. Since ϵ−1≪xm≪ϵ−1/3\epsilon^{-1}\ll x_m\ll\epsilon^{-1/3}, its value can be read from the asymptotic behaviour (73) of the small xx solution:

Θ(xm)=ϕ(xm)−ϵxm=−3π4+ϵn0xm+38ϵn0xm+4xmϵn0−ϵxm+o(1).(76)\Theta(x_m)=\phi(x_m)-\epsilon x_m=-\frac{3\pi}{4}+\epsilon n_0x_m+\frac{3}{8\epsilon n_0x_m}+\frac{4x_m}{\epsilon n_0}-\epsilon x_m+o(1). \tag*{(76)}

We now calculate the integral in (75). By (58), it is given by

∫xm∞(φ−ϵ) dx=∫xm∞[ϵ(n−1)−12ϵnV+18ϵ(3n′2n3−2n′′n2)]dx+O(ϵ−3).(77)\int_{x_m}^{\infty}(\varphi-\epsilon)\,\mathrm{d}x=\int_{x_m}^{\infty}\left[\epsilon(n-1)-\frac{1}{2\epsilon n}V+\frac{1}{8\epsilon}\left(\frac{3n'^{2}}{n^{3}}-\frac{2n''}{n^{2}}\right)\right]\mathrm{d}x+O(\epsilon^{-3}). \tag*{(77)}

The first term is

ϵ∫xm∞(n−1) dx=ϵ∫0∞(n−1) dx−ϵ∫0xm(n−1) dx=ϵI(μ)−ϵ(n0−1)xm+O(ϵxm3)(78)\epsilon\int_{x_m}^{\infty}(n-1)\,\mathrm{d}x=\epsilon\int_{0}^{\infty}(n-1)\,\mathrm{d}x-\epsilon\int_{0}^{x_m}(n-1)\,\mathrm{d}x=\epsilon I(\mu)-\epsilon(n_0-1)x_m+O(\epsilon x_m^{3}) \tag*{(78)}

since in the second integral x≤xmx\leq x_m and

n=(1+μ(1+x2)2)1/2=(1+μ(1−2x2)+O(x4))1/2=n0+O(x2).(79)n=\left(1+\frac{\mu}{(1+x^{2})^{2}}\right)^{1/2}=\left(1+\mu(1-2x^{2})+O(x^{4})\right)^{1/2}=n_0+O(x^{2}). \tag*{(79)}

The VV term gives

−12ϵ∫xm∞Vn dx=−12ϵ∫xm∞(34nx2−8n(1+x2)2) dx=−12ϵ∫xm∞(34x2(1n−1n0)−8n(1+x2)2) dx−38n0ϵxm=−12ϵ∫0∞(34x2(1n−1n0)−8n(1+x2)2) dx+12ϵ∫0xm(34x2(1n−1n0)−8n(1+x2)2) dx−38n0ϵxm=−12ϵ∫0∞(34x2(1n−1n0)−8n(1+x2)2) dx+3μxm8ϵn03−4xmϵn0−38n0ϵxm+O(xm3/ϵ).(80)\begin{aligned} -\frac{1}{2\epsilon}\int_{x_m}^{\infty}\frac{V}{n}\,dx=-\frac{1}{2\epsilon}\int_{x_m}^{\infty}\left(\frac{3}{4nx^{2}}-\frac{8}{n(1+x^{2})^{2}}\right)\,dx \\ &=-\frac{1}{2\epsilon}\int_{x_m}^{\infty}\left(\frac{3}{4x^{2}}\left(\frac{1}{n}-\frac{1}{n_{0}}\right)-\frac{8}{n(1+x^{2})^{2}}\right)\,dx-\frac{3}{8n_{0}\epsilon x_{m}} \\ &=-\frac{1}{2\epsilon}\int_{0}^{\infty}\left(\frac{3}{4x^{2}}\left(\frac{1}{n}-\frac{1}{n_{0}}\right)-\frac{8}{n(1+x^{2})^{2}}\right)\,dx+\frac{1}{2\epsilon}\int_{0}^{x_{m}}\left(\frac{3}{4x^{2}}\left(\frac{1}{n}-\frac{1}{n_{0}}\right)-\frac{8}{n(1+x^{2})^{2}}\right)\,dx-\frac{3}{8n_{0}\epsilon x_{m}} \\ &=-\frac{1}{2\epsilon}\int_{0}^{\infty}\left(\frac{3}{4x^{2}}\left(\frac{1}{n}-\frac{1}{n_{0}}\right)-\frac{8}{n(1+x^{2})^{2}}\right)\,dx+\frac{3\mu x_{m}}{8\epsilon n_{0}^{3}}-\frac{4x_{m}}{\epsilon n_{0}}-\frac{3}{8n_{0}\epsilon x_{m}}+O(x_{m}^{3}/\epsilon). \tag*{(80)} \end{aligned}

Finally we have the O(ϵ−1)O(\epsilon^{-1}) WKB contribution,

18ϵ∫xm∞(3n′2n3−2n′′n2) dx=18ϵ∫0∞(3n′2n3−2n′′n2) dx−18ϵ∫0xm(3n′2n3−2n′′n2) dx=18ϵ∫0∞(−2n1/2 d(n′n3/2))−μ2ϵn03xm+O(xm3)=−18ϵ∫0∞n′2n3 dx−μ2ϵn03xm+O(xm3/ϵ).(81)\begin{aligned} \frac{1}{8\epsilon}\int_{x_m}^{\infty}\left(\frac{3n'^{2}}{n^{3}}-\frac{2n''}{n^{2}}\right)\,dx=\frac{1}{8\epsilon}\int_{0}^{\infty}\left(\frac{3n'^{2}}{n^{3}}-\frac{2n''}{n^{2}}\right)\,dx-\frac{1}{8\epsilon}\int_{0}^{x_m}\left(\frac{3n'^{2}}{n^{3}}-\frac{2n''}{n^{2}}\right)\,dx \\ &=\frac{1}{8\epsilon}\int_{0}^{\infty}\left(-\frac{2}{n^{1/2}}\,d\left(\frac{n'}{n^{3/2}}\right)\right)-\frac{\mu}{2\epsilon n_{0}^{3}}x_m+O(x_m^{3})=-\frac{1}{8\epsilon}\int_{0}^{\infty}\frac{n'^{2}}{n^{3}}\,dx-\frac{\mu}{2\epsilon n_{0}^{3}}x_m+O(x_m^{3}/\epsilon). \tag*{(81)} \end{aligned}

Summing the three terms gives

Θ(∞)=−3π4+ϵI(μ)−12ϵ∫0∞(34x2(1n−1n0)−8n(1+x2)2) dx−18ϵ∫0∞n′2n3 dx\Theta(\infty)=-\frac{3\pi}{4}+\epsilon I(\mu)-\frac{1}{2\epsilon}\int_{0}^{\infty}\left(\frac{3}{4x^{2}}\left(\frac{1}{n}-\frac{1}{n_{0}}\right)-\frac{8}{n(1+x^{2})^{2}}\right)\,dx-\frac{1}{8\epsilon}\int_{0}^{\infty}\frac{n'^{2}}{n^{3}}\,dx
−μxm8ϵn03.(82)-\frac{\mu x_m}{8\epsilon n_{0}^{3}}. \tag*{(82)}

The last term is actually o(ϵ−1)o(\epsilon^{-1}), so we can drop it. In fact in the intermediate region ϵ−1≪xm≪1\epsilon^{-1}\ll x_m\ll1 hence xm/ϵ≪1/ϵ⇒xm/ϵ=o(ϵ−1)x_m/\epsilon\ll1/\epsilon\Rightarrow x_m/\epsilon=o(\epsilon^{-1}) (the ϵ−1∼x∗≪xm\epsilon^{-1}\sim x_*\ll x_m inequality is not needed here but justifies using the oscillatory asymptotics of Bessel’s function).

Hence defining

I(x)=∫0x(n−1) dt,I=I(∞)\mathcal{I}(x)=\int_{0}^{x}(n-1)\,dt,\qquad I=\mathcal{I}(\infty)
K1(x)=−∫0x(38t2(1n−1n0)−4n(1+t2)2) dt,K1=K1(∞)\mathcal{K}_{1}(x)=-\int_{0}^{x}\left(\frac{3}{8t^{2}}\left(\frac{1}{n}-\frac{1}{n_{0}}\right)-\frac{4}{n(1+t^{2})^{2}}\right)\,dt,\qquad K_{1}=\mathcal{K}_{1}(\infty)
K2(x)=−18∫0xn′2n3 dx=−μ22∫0xt2 dt(1+t2)6n5,K2=K2(∞),(83)\mathcal{K}_{2}(x)=-\frac{1}{8}\int_{0}^{x}\frac{n'^{2}}{n^{3}}\,dx=-\frac{\mu^{2}}{2}\int_{0}^{x}\frac{t^{2}\,dt}{(1+t^{2})^{6}n^{5}},\qquad K_{2}=\mathcal{K}_{2}(\infty), \tag*{(83)}

we have

Θ=−3π4+ϵI+1ϵ(K1+K2)+o(ϵ−1).(84)\Theta=-\frac{3\pi}{4}+\epsilon I+\frac{1}{\epsilon}(K_{1}+K_{2})+o(\epsilon^{-1}). \tag*{(84)}

When convenient we will consider II, K1K_{1}, K2K_{2} as functions of the parameter μ\mu.

It is worth noting that this treatment shows that, to the order we are working at, we can define

Θ(x∗)=−3π4+ϵI(x∗)+1ϵ(K1(x∗)+K2(x∗))(85)\Theta(x_*) = -\frac{3\pi}{4} + \epsilon\mathcal{I}(x_*) + \frac{1}{\epsilon}\left(\mathcal{K}_1(x_*) + \mathcal{K}_2(x_*)\right) \tag*{(85)}

and consequently

Θ(x)=−3π4+ϵI(x)+1ϵ(K1(x)+K2(x)).(86)\Theta(x) = -\frac{3\pi}{4} + \epsilon\mathcal{I}(x) + \frac{1}{\epsilon}\left(\mathcal{K}_1(x) + \mathcal{K}_2(x)\right). \tag*{(86)}

We have thus managed to express the phase contributions from both the tunneling and oscillatory regions as the integral of the same smooth function. Interestingly we then see that

Θ(0)=−3π4,(87)\Theta(0) = -\frac{3\pi}{4}, \tag*{(87)}

so the regularity choice C2=0C_2 = 0 at x=0x = 0, see (72), can be interpreted as setting the initial value of Θ\Theta. The equation

A2A1=−tan⁡(Θ+3π4)=−tan⁡(Θ(∞)−Θ(0))(88)\frac{A_2}{A_1} = -\tan\left(\Theta+ \frac{3\pi}{4}\right) = -\tan\left(\Theta(\infty) - \Theta(0)\right) \tag*{(88)}

shows how A1/A2A_1/A_2 can be viewed as a phase shift.

The following Figure 1, Figure 2 show the ratio A1/A2A_1/A_2 and phase cot⁡Θ\cot\Theta obtained from the numerics and from the WKB estimate. The plots are shown for the value μ=1\mu= 1. The values of the three integrals are

Figure 1. The blue curve is the ratio A1/A2A_1/A_2 obtained from the numerics, the red curve is A1/A2A_1/A_2 obtained from (88), (84). The black vertical lines show the predicted asymptotic spacing of the poles, π/I(1)\pi/I(1).

Figure 2. The blue curve is the phase cot⁡Θ\cot\Theta obtained from the numeric solution of the ODE via (49), the red curve is cot⁡Θ\cot\Theta obtained from (84).

I(1)≃0.347,K1(1)≃2.300,K2(1)≃−0.007.(89)\mathcal{I}(1) \simeq0.347,\qquad\mathcal{K}_1(1) \simeq2.300,\qquad\mathcal{K}_2(1) \simeq-0.007. \tag*{(89)}

The small-ϵ\epsilon regime

Let us now consider the ϵ≪1\epsilon\ll1 regime. We use a matched asymptotic expansion with two regions: near region 0≤x≲10 \le x \lesssim1 with x=O(1)x = O(1) at the right region boundary, far region ϵx∼1\epsilon x \sim1 with x=O(1/ϵ)x = O(1/\epsilon) at the left boundary.

Note that the right boundary of the inner region has x∼1x \sim1 so we cannot simply take aneara_{\mathrm{near}}. Instead we expand perturbatively in ϵ2\epsilon^2. Write

a=a(0)+ϵ2a(2).(90)a = a^{(0)} + \epsilon^2 a^{(2)}. \tag*{(90)}

At lowest order in ϵ\epsilon (36) becomes

(a(0))′′+P(a(0))′=0(91)\left(a^{(0)}\right)'' + P\left(a^{(0)}\right)' = 0 \tag*{(91)}

which can be integrated explicitly,

a(0)(x)=C2(x2+4log⁡x−1x2)+C1.(92)a^{(0)}(x) = C_2\left(x^2 + 4\log x - \frac{1}{x^2}\right) + C_1. \tag*{(92)}

Regularity at x=0x=0 requires C2=0C_2=0, and we normalise a(0)(0)=1a^{(0)}(0)=1, so C1=1C_1=1. At order ϵ2\epsilon^2 we get the equation

(a(2))′′+P(a(2))′+Q=0.(93)\left(a^{(2)}\right)'' + P\left(a^{(2)}\right)' + Q = 0. \tag*{(93)}

This ODE can be integrated explicitly, but to do the matching we only need its behaviour for large xx, which is

a(2)=−x22(log⁡x+μ12−1)+O(log⁡2x).(94)a^{(2)} = -\frac{x^2}{2}\left(\log x + \frac{\mu}{12} - 1\right) + O\left(\log^2 x\right). \tag*{(94)}

Hence in the near region, for large xx, we have

a=1−ϵ2x22(log⁡x+μ12−1)+O(ϵ4).(95)a = 1 - \frac{\epsilon^2 x^2}{2}\left(\log x + \frac{\mu}{12} - 1\right) + O\left(\epsilon^4\right). \tag*{(95)}

In the far region we have x=O(1/ϵ)⇒x≫1x = O(1/\epsilon) \Rightarrow x \gg1. Since for large xx we have Q→1Q \to1 we can approximate ϵ2Q\epsilon^2 Q with ϵ2\epsilon^2, reobtaining (46). To do the matching we need the small xx behaviour of afara_{\mathrm{far}}, given by (50). The matching condition is then

1−ϵ2x22(log⁡x+μ12−1)=−2A2πϵ+ϵA2πx2log⁡x+ϵ2π[πA1+2A2log⁡(ϵ/2)+(2γ−1)A2]x2.(96)1 - \frac{\epsilon^2 x^2}{2}\left(\log x + \frac{\mu}{12} - 1\right) = -\frac{2A_2}{\pi\epsilon} + \frac{\epsilon A_2}{\pi}x^2\log x + \frac{\epsilon}{2\pi}\left[\pi A_1 + 2A_2\log(\epsilon/2) + (2\gamma- 1)A_2\right]x^2. \tag*{(96)}

Matching the constant term gives

A2=−πϵ2(97)A_2 = -\frac{\pi\epsilon}{2} \tag*{(97)}

which also solves the x2log⁡xx^2\log x matching. Matching the x2x^2 term then gives A1A_1, from which we get the ratio

A1A2=−2πlog⁡(ϵ/2)+μ6π−2γ+1π.(98)\frac{A_1}{A_2} = -\frac{2}{\pi}\log(\epsilon/2) + \frac{\mu}{6\pi} - \frac{2\gamma+ 1}{\pi}. \tag*{(98)}

In particular, A1A2\frac{A_1}{A_2} diverges logarithmically as ϵ→0+\epsilon\to0^+, with a slope which is independent of μ\mu. Plotting the result of the numerical integration of the full ODE shows full agreement with (98).

Figure 3. The red curve is the ratio A1/A2A_1/A_2 obtained from (98) for μ=1\mu=1, the blue dots show the ratio A1/A2A_1/A_2 obtained numerically. Note that the values are plotted against log⁡ϵ\log\epsilon for ϵ\epsilon varying between 10−410^{-4} and 10−210^{-2}.

The low-energy effective action

We can now write the low-energy effective action. This is the closest analogue to the zero-mode approximation dynamics [1] that we can find when the zero mode is non-normalisable. We consider a cutoff radius x~\tilde{x} that is bigger than the size of the soliton and yet smaller than 1/ϵ1/\epsilon

1≪x~≪1ϵ(99)1 \ll\tilde{x} \ll\frac{1}{\epsilon} \tag*{(99)}

We take only the leading terms at low-energy, or large radius, from (45) and we add a boundary term at x~\tilde{x} obtaining

Seff=∫dτ∫x~∞dxx(α˙(τ,x)2−α′(τ,x)2)+Meff(x~)2∫dτ α˙(τ,x~)2(100)S_{\mathrm{eff}}=\int\mathrm{d}\tau\int_{\tilde{x}}^{\infty}\frac{\mathrm{d}x}{x}\left(\dot{\alpha}(\tau,x)^2-\alpha'(\tau,x)^2\right)+\frac{M_{\mathrm{eff}}(\tilde{x})}{2}\int\mathrm{d}\tau\,\dot{\alpha}(\tau,\tilde{x})^2 \tag*{(100)}

The boundary term is a localised kinetic term for α(τ,x~)\alpha(\tau,\tilde{x}). It is the only term which would be present if the zero mode were normalisable. The boundary equation of motion from (100) is

α′(τ,x~)x~−Meff(x~)2α¨(τ,x~)=0.(101)\frac{\alpha'(\tau,\tilde{x})}{\tilde{x}}-\frac{M_{\mathrm{eff}}(\tilde{x})}{2}\ddot{\alpha}(\tau,\tilde{x})=0. \tag*{(101)}

From (50) we get, to leading order,

a′(x~)x~=2ϵA2πlog⁡(x~ϵ2)+ϵ(A1+2γA2π),(102)\frac{a'(\tilde{x})}{\tilde{x}} = \frac{2\epsilon A_{2}}{\pi}\log\left(\frac{\tilde{x}\epsilon}{2}\right) + \epsilon\left(A_{1}+\frac{2\gamma A_{2}}{\pi}\right), \tag*{(102)}

and

a(x~)=−2A2πϵ,(103)a(\tilde{x})=-\frac{2A_{2}}{\pi\epsilon}, \tag*{(103)}

so we have a clear correspondence between A1,2A_{1,2} and α(τ,x)=a(x)eiϵτ\alpha(\tau,x)=a(x)e^{i\epsilon\tau}. Substituting in (101) we obtain

2A2πlog⁡(x~ϵ2)+A1+2γA2π−Meff(x~)A2π=0.(104)\frac{2A_{2}}{\pi}\log\left(\frac{\tilde{x}\epsilon}{2}\right)+A_{1}+\frac{2\gamma A_{2}}{\pi}-M_{\mathrm{eff}}(\tilde{x})\frac{A_{2}}{\pi}=0. \tag*{(104)}

We can match (104) with (98) if we take the effective mass to be

Meff(x~)=2log⁡x~+μ6−1.(105)M_{\mathrm{eff}}(\tilde{x})=2\log\tilde{x}+\frac{\mu}{6}-1. \tag*{(105)}

This value for MeffM_{\mathrm{eff}} is perfectly consistent with the effective mass of the zero mode computed from (29) and (31) using x~\tilde{x} as an IR cutoff:

Meff(x~)=2∫0x~x3(1+x2)2(1+μ(1+x2)2) dx=log⁡(1+x~2)+x~2(x~4(μ−6)+3x~2(μ−4)−6)6(1+x~2)3=2log⁡x~+μ6−1+O(x~−2).(106)\begin{aligned} M_{\mathrm{eff}}(\tilde{x}) &= 2\int_{0}^{\tilde{x}}\frac{x^{3}}{(1+x^{2})^{2}}\left(1+\frac{\mu}{(1+x^{2})^{2}}\right)\,\mathrm{d}x\\ &=\log(1+\tilde{x}^{2})+\frac{\tilde{x}^{2}\left(\tilde{x}^{4}(\mu-6)+3\tilde{x}^{2}(\mu-4)-6\right)}{6(1+\tilde{x}^{2})^{3}}=2\log\tilde{x}+\frac{\mu}{6}-1+O(\tilde{x}^{-2}). \tag*{(106)} \end{aligned}

Conclusions and discussion

Non-normalisable zero modes fail to give a sensible (0+1)(0+1)-dimensional effective action, but in the model considered here they give a (1+1)(1+1)-dimensional effective action defined on a half-plane. At low energies these scattering zero modes can reach the soliton core, where the boundary conditions are determined. We have analyzed this scenario in the simplest possible model: the phase modulus of the holomorphic baby Skyrmion model. In future work we will consider more complex systems such as non-Abelian monopoles and monopoles in hyperbolic space. These systems are considerably more complex than the holomorphic baby Skyrmion. For the non-abelian monopole it is known that, in the Higgs phase, the non-normalizable modes become 1+11+1 dynamical fields on the non-abelian magnetic vortex string [8]. This is somehow similar to what we discussed in this paper, although for non-abelian monopoles a version of the 1+11+1 model in the Coulomb phase, like scattering of radial waves, has never been found so far.

There are similarities with the problem of fermionic fluctuations in a monopole background, known as the Callan–Rubakov effect [25, 26], which can be treated, like the Kondo effect, as a boundary problem [27]. In the presence of a magnetic monopole field, the angular momentum of fermions can be reduced to zero, producing an effective ss-wave particle that can penetrate the monopole core. The anomaly is not suppressed in the monopole sector of the theory, giving rise to the non-conservation of fermion quantum numbers. Results concerning these phenomena have been obtained by reducing the theory to a model for an ss-wave particle moving on a half-plane described by the radial and temporal coordinates. We have used a similar approach for non-normalisable zero modes in this work, focusing on scattering states of bosonic rather than fermionic fluctuations.

An interesting extension of this work would be to consider the limit in which the theory reduces to the pure nonlinear sigma model and the size modulus also becomes a non-normalisable zero mode. It would be interesting to see whether the (1+1)(1 + 1)-dimensional effective approach can also be used for the size modulus, in particular to understand the nature of the finite-time singularities observed in lump dynamics.

Acknowledgments

S. B. and G. F. are supported by the INFN special research project grant “GAST” (Gauge and String Theories). G. F. acknowledges support from GNFM (Gruppo Nazionale per la Fisica Matematica).

References

  1. [1]N. S. Manton. A remark on the scattering of BPS monopoles. Phys. Lett. B, 110(1):54–56, 1982.DOI
  2. [2]P. Sutcliffe. Boundary metrics on soliton moduli spaces. J. High Energy Phys., 01:118, 2022.DOI
  3. [3]G. Franchetti and D. Harland. L² geometry of hyperbolic monopoles, 2024.
  4. [4]E. J. Weinberg and P. Yi. Magnetic monopole dynamics, supersymmetry, and duality. Phys. Rep., 438:65–236, 2007.DOI
  5. [5]K. Lee, E. J. Weinberg, and P. Yi. Massive and massless monopoles with non-Abelian magnetic charges. Phys. Rev. D, 54:6351–6371, 1996.pubmed.ncbi.nlm.nih.gov/10020637
  6. [6]X. Chen, H. Guo, and E. J. Weinberg. Massless monopoles and the moduli space approximation. Phys. Rev. D, 64:125004, 2001.DOI
  7. [7]Roberto Auzzi, Stefano Bolognesi, Jarah Evslin, Kenichi Konishi, and Hitoshi Murayama. NonAbelian monopoles. Nucl. Phys. B, 701:207–246, 2004.DOI
  8. [8]Roberto Auzzi, Stefano Bolognesi, Jarah Evslin, and Kenichi Konishi. NonAbelian monopoles and the vortices that confine them. Nucl. Phys. B, 686:119–134, 2004.DOI
  9. [9]Muneto Nitta and Walter Vinci. Non-Abelian Monopoles in the Higgs Phase. Nucl. Phys. B, 848:121–154, 2011.DOI
  10. [10]R. A. Leese, M. Peyrard, and W. J. Zakrzewski. Soliton scattering in some relativistic models in (2 + 1) dimensions. Nonlinearity, 3(3):773–807, 1990.DOI
  11. [11]P. M. Sutcliffe. The interaction of Skyrme-like lumps in (2+1) dimensions. Nonlinearity, 4:1109–1121, 1991.DOI
  12. [12]R. S. Ward. Slowly-moving lumps in the CP^1 model in (2 + 1) dimensions. Phys. Lett. B, 158(5):424–428, 1985.DOI
  13. [13]R. A. Leese. Low-energy scattering of solitons in the CP^1 model. Nucl. Phys. B, 344:33–72, 1990.DOI
  14. [14]J. M. Speight. Low-energy dynamics of a CP^1 lump on the sphere. J. Math. Phys., 36(2):796–813, 1995.DOI
  15. [15]L. A. Sadun and J. M. Speight. Geodesic incompleteness in the CP^1 model on a compact Riemann surface. Lett. Math. Phys., 43(4):329–334, 1998.arxiv.org/abs/hep-th/9707169
  16. [16]J. M. Speight and I. A. B. Strachan. Gravity thaws the frozen moduli of the CP^1 lump. Phys. Lett. B, 457:17–22, 1999.DOI
  17. [17]N. S. Manton and P. M. Sutcliffe. Topological Solitons. Cambridge Monographs on Mathematical Physics. Cambridge University Press, Cambridge, 2004.DOI
  18. [18]R. A. Leese, M. Peyrard, and W. J. Zakrzewski. Soliton stability in the O(3) σ-model in (2 + 1) dimensions. Nonlinearity, 3(2):387–412, 1990.
  19. [19]P. Bizon, T. Chmaj, and Z. Tabor. Formation of singularities for equivariant (2 + 1)-dimensional wave maps into the two-sphere. Nonlinearity, 14(5):1041–1053, 2001.DOI
  20. [20]P. Raphaël and I. Rodnianski. Stable blow up dynamics for the critical co-rotational wave maps and equivariant Yang–Mills problems. Publ. Math. Inst. Hautes Études Sci., 115:1–122, 2012.arxiv.org/abs/0911.0692
  21. [21]B. M. A. G. Piette, B. J. Schroers, and W. J. Zakrzewski. Multisolitons in a two-dimensional Skyrme model. Z. Phys. C, 65:165–174, 1995.DOI
  22. [22]S. Bolognesi and W. Zakrzewski. Baby Skyrme model, near-BPS approximations and supersymmetric extensions. Phys. Rev. D, 91(4):045034, 2015.DOI
  23. [23]A. Kostenko and G. Teschl. Spectral asymptotics for perturbed spherical Schrödinger operators and applications to quantum scattering. Commun. Math. Phys., 322(1):255–275, 2013.
  24. [24]F. W. J. Olver. Asymptotics and Special Functions. AKP Classics. A K Peters, Wellesley, MA, 1997. Reprint of the 1974 Academic Press edition.
  25. [25]C. G. Callan, Jr. Disappearing dyons. Phys. Rev. D, 25(8):2141–2146, 1982.DOI
  26. [26]V. A. Rubakov. Monopole catalysis of proton decay. Rep. Prog. Phys., 51(2):189–241, 1988.DOI
  27. [27]I. Affleck and J. Sagi. Monopole-catalysed baryon decay: A boundary conformal field theory approach. Nucl. Phys. B, 417(3):374–402, 1994.

Paper details

vlnc:W_3qxst2ax
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents