betaFeedback

WorksPhysics and Astronomy

On the motion of a point charge in a plate capacitor considering influence effects

Abstract

A point charge between the plates of a capacitor generates an influence charge distribution on the plates that modify the electric field acting upon the point charge. This effect is described by the well-known Dirichlet Green's function for the two parallel conducting plate problem for which we derive an infinite mirror charge series representation. At the line perpendicular to the plates and passing through the point charge this Green's function and hence the total force can be explicitly evaluated in terms of the psi function. For the motion of the point charge we develop an analytical approximation and compare it with the numerical integration of the exact equations of motion. The correction due to influence effects is shown to be of order O(−λlog⁡λ)O(-λ\log λ) where λλ denotes the relative strength of the Green's function compared with the pure capacitor potential.

Introduction

Many supposedly simple problems in physics turn out to be not so simple on closer inspection. As an example of this we look at the motion of a point charge in the field of a plate capacitor. At first glance, this is an example of motion in a constant force field that could be covered in physics lessons at school. In this simplified view, however, the sources of the field, i. e. the surface charges on the capacitor plates, are assumed to be fixed. In reality, however, the point charge acts back on the sources and generates “influence charges” on both plates, which have an opposite sign relative to the point charge. These influence charges change the field, and hence also the acceleration of the point charge, and in turn depend on the location of the point charge. The overall result is a complex interplay between the motion of the point charge and influence effects, in which the distinction between cause and effect becomes blurred.

An intermediate step in the theoretical treatment of the problem outlined is the calculation of the (Dirichlet) Green’s function G(r,r0)G(\mathbf{r},\mathbf{r}_0) for the plate capacitor. This is essentially the electric potential of a point charge at the position r0\mathbf{r}_0 between the plates with the Dirichlet boundary condition that G(r,r0)G(\mathbf{r},\mathbf{r}_0) vanishes if r\mathbf{r} lies on the surface of the capacitor plates. From this, the potential energy for the motion of the point charge can be obtained by subtracting its Coulomb potential from G(r,r0)G(\mathbf{r},\mathbf{r}_0), adding the pure capacitor potential and setting r=r0\mathbf{r}=\mathbf{r}_0 in the result. Of course, we assume that the influenced surface charges react instantaneously to the motion of the point charge without any delay. The Green’s function for two parallel conducting plates is known; it is given in the literature as a double series expansion according to cylindrically symmetric eigenfunctions of the Laplace operator, see [1], Problem 3.17a, or as a mixed series/integral representation, see [1], Problem 3.17b. An early source is [2], §19.

However, the double series representation is less suitable for our problem, see Appendix B, and also the series/integral representation appears opaque at first sight. We therefore propose a different way of calculating G(r,r0)G(\mathbf{r},\mathbf{r}_0), which is based on the idea of mirror charges. The approach of calculating the potential of a point charge in a plate capacitor by means of infinitely many mirror charges is occasionally discussed on the internet and in lecture notes, but to our knowledge has rarely been carried out explicitly. Exceptions are [3], where the special case is considered that the point charge is located exactly in the middle between the plates, and [4]. It is well known that, in the presence of only one conducting plate, the Green’s function can be constructed by assuming, in addition to the point charge qq at a distance xx from the plate, a fictitious “mirror charge” −q-q at the same distance xx behind the plate and defining G(r,r0)G(\mathbf{r},\mathbf{r}_0) as the sum of the potentials of both charges. If two plates are present, one can work with two “primary” mirror charges accordingly. Each of these mirror charges ensures a constant potential on the plate assigned to it, but distorts the potential on the other plate. To compensate for this, additional “secondary” mirror charges can be introduced, but then the analogous problem will arise, albeit to a lesser extent. Overall, the Green’s function results as an (initially formal) series of Coulomb potentials of an infinite number of mirror charges at ever greater distances from the plates, in addition to the Coulomb potential of the actual point charge, see section II. The convergence of this series representation is proven in the appendix A. Similar approaches involving infinite many mirror charges have been applied to the problem of a point charge on the axis of a conducting cylinder [3] or to find the self- and mutual capacitances for two spheres [5], section 5.08 and 5.081.

We have thus obtained an alternative series representation of Green’s function for the plate capacitor, which is somewhat simpler than the representations known from the literature and whose terms have a clear meaning as potentials of mirror charges. The motion of the point charge can be restricted to a straight line LL perpendicular to the plates, because the motion perpendicular to it is rectilinear and uniform. We will refer to this straight line LL, on which the mirror charges also lie, as the “central line” in this paper and choose it as the ξ\xi axis in a rectangular coordinate system (ξ,η,ζ)(\xi,\eta,\zeta). As explained above, the special values G(ξ,ξ0)G(\xi,\xi_0) of the restricted Green’s function are required for the motion of the point charge on the central line. It now turns out that the series representation for G(ξ,ξ0)G(\xi,\xi_0) can be explicitly evaluated in terms of the so-called psi function ψ(z)=Γ′(z)/Γ(z)\psi(z)=\Gamma'(z)/\Gamma(z), see section III. An alternative derivation using the mixed series/integral representation of the Green’s function is given in Appendix B. This also allows the force on the point charge to be calculated and the corresponding equation of motion for x(t)x(t) to be numerically integrated. In principle, we have thus solved the problem that we set ourselves in this paper.

In the following, we investigate the possibilities of analytical approximations for x(t)x(t), which are offered by our approach. In the vicinity of the left-hand plate, the influence of all mirror charges can be approximated by that of the single primary mirror charge, which is located at a distance x(t)x(t) behind the plate, see subsection IV A. This leads to the motion in a potential of the form V(x)=−qUxd−q24πϵ01xV(x)=-qU\frac{x}{d}-\frac{q^2}{4\pi\epsilon_0}\frac{1}{x}, or, in dimensionless quantities, V(x)=−x−λxV(x)=-x-\frac{\lambda}{x}. This simple potential, superposition of a linear and a Coulomb potential, is known in the literature as “Cornell potential” and is used in connection with confinement in quantum chromodynamics, see [6]. Another application of this potential is the Stark effect, see [7]. It turns out that the one-dimensional motion in the Cornell potential can be described by Legendre elliptic integrals of the 1st and 2nd kind for the inverse function t(x)t(x), see subsection IV B 1 and Appendix C. This also applies to the motion near the right-hand plate in the approximated potential V(x)=−x−λ1−xV(x)=-x-\frac{\lambda}{1-x}, see subsection IV B 2. We therefore divide the range 0≤x≤10\leq x\leq1 between the plates into three equal parts and approximate the exact potential in the middle range 1/3≤x≤2/31/3\leq x\leq2/3 by its second-order Taylor expansion at the point x=1/2x=1/2. The approximated equation of motion can also be solved exactly in this range, see subsection IV C. The quality of the analytical approximation just described for the motion x(t)x(t) of the point charge is excellent for the parameters investigated here, see subsection IV D.

The relative strength of the influence effects is described by the dimensionless parameter λ\lambda. For λ=0\lambda=0, the (dimensionless) transit time T(λ)T(\lambda) of the point charge between the plates assumes the simple value T(0)=2(E+1−E)T(0)=\sqrt{2}\left(\sqrt{E+1}-\sqrt{E}\right), where EE denotes the total energy. At first glance one would expect that the influence correction δT=T(0)−T(λ)\delta T=T(0)-T(\lambda) approaches zero with the order O(λ)O(\lambda). In fact, however, the portions of the potential of the form −λx-\frac{\lambda}{x} and −λ1−x-\frac{\lambda}{1-x}, which describe the influence of the primary mirror charges, are arbitrarily large for x→0x\to0, resp. x→1x\to1, as long as λ>0\lambda>0. It is therefore plausible that δT\delta T approaches zero more slowly than with the order O(λ)O(\lambda). Numerical and analytical investigations rather show a λ\lambda-dependence of δT\delta T of the order O(−λlog⁡λ)O(-\lambda\log\lambda), see section V.

After the successful theoretical treatment of the motion of a point charge in a plate capacitor, the question of possible experimental applications arises. In the last decades, experiments with a so-called “electrostatic pendulum” have been reported, as well as approaches to their theoretical description, see [8, 9, 10, 11, 12, 13]. This is a charged hollow sphere that oscillates between the plates of a capacitor and is recharged after contact with one plate at a time. The complete description of this experiment is difficult because it would have to include influence effects on the capacitor plates as well as on the sphere. The present work can be seen as a theoretical description of a limiting case of these pendulum experiments in which, firstly, the pendulum length is large compared to the distance of the plates (and therefore the motion is almost rectilinear) and, secondly, the radius of the charged sphere is so small that it can be described as a point charge. We will return to this aspect of our work in the Summary and Outlook section VI.

Infinite Sequence of Mirror Charges

The potential of a point charge qq located at a distance xx in front of an infinitely extended conductive plate can be calculated using the well-known mirror charge method. This consists of assuming a fictitious mirror charge with opposite sign −q-q at the same distance xx behind the plate and adding the fields of the point charge and the mirror charge, see Figure 1. The field lines of the resulting field are then perpendicular to the surface of the plate, which therefore forms an equipotential surface of the resulting potential, see Figure 2, left panel. If a second conductive parallel plate is added so that the point charge is located between the plates of a capacitor with the distance dd, it is obvious to also introduce a second mirror charge −q-q at a distance d−xd-x behind the second plate, see Figure 1. These two mirror charges are referred to as “primary” mirror charges.

Figure 1. Sketch of the problem considered in this paper: We show two capacitor plates at a distance dd. A particle with charge qq moving between the plates has at momentary distance xx from the left plate, and at distance d−xd-x from the right plate. We consider a “primary” left mirror charge −q-q at distance xx from the left plate and another primary right mirror charge −q-q at a distance d−xd-x from the right-hand plate. Both primary mirror charges generate series of further “secondary” mirror charges with alternating signs, which are located at the central line (dashed black line) to the right or left of the capacitor plates.

Figure 2. Left panel: Sketch of the field generated by a point charge (magenta dot) and a left mirror charge (magenta circle). The field lines are perpendicular on the left plate (blue line). Right panel: Sketch of the field generated by a point charge and two “primary” mirror charges (magenta circles). The field lines are approximately perpendicular on the plates (blue and red line) only in a central region close to the central line joining the charges.

However, the field lines of the system, which consists of a point charge and two mirror charges, are not exactly perpendicular to the surfaces of the two plates, see Figure 2, right panel, because the field of the left mirror charge is not perpendicular to the right plate and, conversely, the field of the right mirror charge is not perpendicular to the left plate. To further approximate a solution to the electrostatic problem, additional “secondary” mirror charges of the primary mirror charges can be introduced. But then the analogous problem arises again, albeit to a lesser extent. It is therefore expected that the exact potential of a point charge between two parallel conducting plates can only be obtained by adding the potentials of an infinite number of mirror charges at ever greater distances from the plates, see Figure 3.

Figure 3. Sketch of the field generated by a point charge, two “primary” mirror charges and 8 further “secondary” mirror charges of alternate signs, of which only 4 are visible. The total field is vertical on both plates to a very good approximation. The actual field of the point charge will be obtained by the limit of the infinitely many mirror charges, restricted to the space between the plates. The fictitious field outside this space is shown here for illustrative purposes only.

To calculate the positions of the various mirror charges it is advisable to divide the set of mirror charges into two “generations” depending on whether we start with the left or the right primary mirror charge. Let R1(n)R_1(n) denote the distance of the nnth mirror charge qnq_n, n=1,2,…n=1,2,\ldots of the first generation to the right plate, analogously L1(n)L_1(n) for the distance to the left plate, and R2(n)R_2(n) and L2(n)L_2(n) for the mirror charges of the second generation, see Figure 4 and Figure 5. Then the distances of the primary mirror charges, i. e., for n=1n=1, satisfy

Figure 4. Sketch of the first five mirror charges qnq_n, n=1,…,5n = 1, \ldots, 5 of the first generation. The first row contains their numbers n=1,…,5n = 1, \ldots, 5. The odd charges q1,q3,q5,…q_1, q_3, q_5, \ldots are reflected at the right plate of the capacitor which leads to the even charges q2,q4,…q_2, q_4, \ldots. The corresponding distances to the right plate are therefore the same, for example R1(4)=R1(3)R_1(4) = R_1(3). The distance of the even charges to the left-hand plate is increased by dd, for example L1(4)=R1(4)+d=R1(3)+dL_1(4) = R_1(4) + d = R_1(3) + d. The even charges q2,q4,…q_2, q_4, \ldots are reflected at the left plate, resulting in the odd charges q3,q5,…q_3, q_5, \ldots. Therefore, the corresponding distances to the left plate are equal, for example L1(5)=L1(4)L_1(5) = L_1(4). In addition, the distance of the odd charges to the right plate is increased by dd, for example R1(5)=L1(5)+d=L1(4)+dR_1(5) = L_1(5) + d = L_1(4) + d. This illustrates the recurrence relations (3) and (4) for the first generation of mirror charges.

Figure 5. Sketch of the first five mirror charges qnq_n, n=1,…,5n = 1,\ldots,5 of the second generation. The first row contains their numbers n=1,…,5n = 1,\ldots,5. The odd charges q1,q3,q5,…q_1,q_3,q_5,\ldots are reflected at the left plate of the capacitor which leads to the even charges q2,q4,…q_2,q_4,\ldots. The corresponding distances to the left plate are therefore the same, for example L2(4)=L2(3)L_2(4) = L_2(3). The distance of the even charges to the right plate is increased by dd, for example R2(4)=L2(4)+d=L2(3)+dR_2(4) = L_2(4) + d = L_2(3) + d. The even charges q2,q4,…q_2,q_4,\ldots are reflected at the right plate, resulting in the odd charges q3,q5,…q_3,q_5,\ldots. Therefore, the corresponding distances to the right plate are equal, for example R2(5)=R2(4)R_2(5) = R_2(4). In addition, the distance of the odd charges to the left plate is increased by dd, for example L2(5)=R2(5)+d=R2(4)+dL_2(5) = R_2(5) + d = R_2(4) + d. This illustrates the recurrence relations (5) and (6) for the second generation of mirror charges.

R1(1)=d+x,L1(1)=x,(1)R_1(1) = d+x,\quad L_1(1) = x, \tag*{(1)}
L2(1)=2d−x,R2(1)=d−x,(2)L_2(1) = 2d-x,\quad R_2(1) = d-x, \tag*{(2)}

see Figure 1. The mirror charges have alternating signs (−1)nq(-1)^nq for both generations so that, assuming q>0q>0, all negative charges are at the left hand side of the left plate for the first generation, and at the right hand side of the right plate for the second generation, see the Figure 4 and Figure 5. Moreover, the four sequences n↦Li(n),Ri(n),i=1,2n \mapsto L_i(n), R_i(n), i = 1, 2, satisfy the recursion relations

L1(n+1)={L1(n): n evenR1(n)+d: n odd(3)\begin{aligned} L_1(n+1) &= \begin{cases} L_1(n) & :\ n\ \text{even}\\ R_1(n)+d & :\ n\ \text{odd} \end{cases} \tag*{(3)} \end{aligned}
R1(n+1)={L1(n)+d: n evenR1(n): n odd(4)\begin{aligned} R_1(n+1) &= \begin{cases} L_1(n)+d & :\ n\ \text{even}\\ R_1(n) & :\ n\ \text{odd} \end{cases} \tag*{(4)} \end{aligned}
L2(n+1)={R2(n)+d: n evenL2(n): n odd(5)\begin{aligned} L_2(n+1) &= \begin{cases} R_2(n)+d & :\ n\ \text{even}\\ L_2(n) & :\ n\ \text{odd} \end{cases} \tag*{(5)} \end{aligned}
R2(n+1)={R2(n): n evenL2(n)+d: n odd(6)\begin{aligned} R_2(n+1) &= \begin{cases} R_2(n) & :\ n\ \text{even}\\ L_2(n)+d & :\ n\ \text{odd} \end{cases} \tag*{(6)} \end{aligned}

These relations can be obtained by generalizing the relations considered in the Figure 4 and Figure 5. The initial conditions (1),(2) and the recursion relations (3)-(6) lead to the following sequences with typical pairwise constant sequence members:

L1=(x,2d+x,2d+x,4d+x,4d+x,…),(7)L_1 = (x,2d+x,2d+x,4d+x,4d+x,\ldots), \tag*{(7)}
R1=(d+x,d+x,3d+x,3d+x,5d+x,5d+x,…),(8)R_1 = (d+x,d+x,3d+x,3d+x,5d+x,5d+x,\ldots), \tag*{(8)}
L2=(2d−x,2d−x,4d−x,4d−x,6d−x,6d−x,…),(9)L_2 = (2d-x,2d-x,4d-x,4d-x,6d-x,6d-x,\ldots), \tag*{(9)}
R2=(d−x,3d−x,3d−x,5d−x,5d−x,…).(10)R_2 = (d-x,3d-x,3d-x,5d-x,5d-x,\ldots). \tag*{(10)}

Their explicit representation can be given in the following form:

L1(n)=2⌊n2⌋d+x,(11)L_1(n) = 2\left\lfloor\frac{n}{2}\right\rfloor d+x, \tag*{(11)}
R1(n)=(2⌊n+12⌋−1)d+x,(12)R_1(n) = \left(2\left\lfloor\frac{n+1}{2}\right\rfloor-1\right)d+x, \tag*{(12)}
L2(n)=2⌊n+12⌋d−x,(13)L_2(n) = 2\left\lfloor\frac{n+1}{2}\right\rfloor d-x, \tag*{(13)}
R2(n)=(2⌊n2⌋+1)d−x,(14)R_2(n) = \left(2\left\lfloor\frac{n}{2}\right\rfloor+1\right)d-x, \tag*{(14)}

for n=1,2,…n = 1,2,\ldots. The proof is straightforward. In order to give an example we will only prove that (3) is satisfied by (11) and (12): First, let nn be even. Then

L1(n+1)=(11)2⌊n+12⌋d+x=(n even)2⌊n2⌋d+x=(11)L1(n).(15)L_1(n+1) \mathrel{\overset{(11)}{=}} 2\left\lfloor\frac{n+1}{2}\right\rfloor d+x \mathrel{\overset{(n\ \mathrm{even})}{=}} 2\left\lfloor\frac{n}{2}\right\rfloor d+x \mathrel{\overset{(11)}{=}} L_1(n). \tag*{(15)}

Next, let nn be odd. Then

L1(n+1)=(11)2⌊n+12⌋d+x=(2⌊n+12⌋−1)d+x+d=(12)R1(n)+d.(16)L_1(n+1) \mathrel{\overset{(11)}{=}} 2\left\lfloor\frac{n+1}{2}\right\rfloor d+x = \left(2\left\lfloor\frac{n+1}{2}\right\rfloor-1\right)d+x+d \mathrel{\overset{(12)}{=}} R_1(n)+d. \tag*{(16)}

This completes the proof of (3).

To determine the coordinates of the mirror charges we define the central line connecting the mirror charges as a ξ\xi-axis with the origin in the center exactly between the two plates, so that the original point charge has the coordinate

ξ0=x−d2.(17)\xi_0 = x-\frac{d}{2}. \tag*{(17)}

Then a short calculation gives the ξ\xi-coordinates of the mirror charges of the two generations in the following form:

X1(n)=(−1)n((n−12)d+x),(18)X_1(n)=(-1)^n\left(\left(n-\frac{1}{2}\right)d+x\right), \tag*{(18)}
X2(n)=(−1)n(−(n+12)d+x),(19)X_2(n)=(-1)^n\left(-\left(n+\frac{1}{2}\right)d+x\right), \tag*{(19)}

for n=1,2,…n=1,2,\ldots.

We extend the ξ\xi-axis to a 3-dimensional Cartesian coordinate system with coordinates r=(ξ,η,ζ)\mathbf{r}=(\xi,\eta,\zeta) such that the mirror charges obtain the coordinates

R1(n)=(X1(n),0,0)(20)\mathbf{R}_1(n)=(X_1(n),0,0) \tag*{(20)}
R2(n)=(X2(n),0,0)(21)\mathbf{R}_2(n)=(X_2(n),0,0) \tag*{(21)}

for n=1,2,…n=1,2,\ldots and the point charge has the coordinates

r0=(ξ0,0,0).(22)\mathbf{r}_0=(\xi_0,0,0). \tag*{(22)}

The electrical potential of the point charge in the plate capacitor is formally given by the following infinite sum of Coulomb potentials

Φ(r,r0)=q4πϵ0(∑i=1,2∑n=1∞(−1)n∣r−Ri(n)∣+1∣r−r0∣)(23)\Phi(\mathbf{r},\mathbf{r}_0)=\frac{q}{4\pi\epsilon_0}\left(\sum_{i=1,2}\sum_{n=1}^{\infty}\frac{(-1)^n}{\lvert\mathbf{r}-\mathbf{R}_i(n)\rvert}+\frac{1}{\lvert\mathbf{r}-\mathbf{r}_0\rvert}\right) \tag*{(23)}

We have labeled this solution as “formal” since questions of convergence of the series have not yet been considered. For this see Appendix § 4.1.

Next we address the question whether the series representation (23) can be considered as the Dirichlet Green’s function of the plate capacitor problem. For this it is required as a boundary condition that the potential assumes the value zero at both plates. This would then automatically entail that the field lines of the corresponding electrical field are perpendicular to the surfaces of the plates. Anticipating the considerations in Appendix § 4.1 we arrange the contributions to (23) into pairs. For the second generation, it is obvious that the pairs of mirror charges numbered (1, 2), (3, 4), . . . are symmetric with respect to the left plate, see Figure 5, and, since they have opposite charges, that the sum of the potentials of each pair vanishes at the left plate. For the first generation the situation is analogous for the pairs with numbers (2, 3), (4, 5), . . ., see Figure 4, such that the contribution of these pairs also vanishes at the left plate. The remaining terms correspond to the primary mirror charge with number n=1n=1 and to the point charge. But also their potentials add to zero at the left plate. The argumentation for the vanishing of (23) at the right-hand plate is completely analogous. This shows that, upon arranging the series (23) pairwise in the manner indicated above, it represents a potential that satisfies the requirement of a Dirichlet Green’s function, i. e., G(r,r0)=Φ(r,r0)G(\mathbf{r},\mathbf{r}_0)=\Phi(\mathbf{r},\mathbf{r}_0) if setting q4πϵ0=1\frac{q}{4\pi\epsilon_0}=1.

Evaluation of the Green’s Function at the Central Line

We will explicitly evaluate the series representation (23) of Φ(r,r0)\Phi(\mathbf{r},\mathbf{r}_0) for the special values of r=(ξ,0,0)\mathbf{r}=(\xi,0,0), that is, at the central line joining the mirror charges. Based on the considerations in the Appendix § 4.1 on the conditional convergence of the series (23), we will arrange the terms for both generations of mirror charges pairwise.

Starting with the first generation and recalling the definitions (17), (20) and (22) we obtain for even n=2mn=2m

∣R1(2m)−r∣=2md+ξ0−ξ,(24)\lvert\mathbf{R}_1(2m)-\mathbf{r}\rvert=2md+\xi_0-\xi, \tag*{(24)}

while for odd n=2m−1n=2m-1

∣R1(2m−1)−r∣=(2m−1)d+ξ0+ξ.(25)\lvert\mathbf{R}_1(2m-1)-\mathbf{r}\rvert=(2m-1)d+\xi_0+\xi. \tag*{(25)}

This entails

1∣R1(2m)−r∣−1∣R1(2m−1)−r∣=2ξ−d(2md−ξ+ξ0)((2m−1)d+ξ+ξ0).(26)\frac{1}{\lvert\mathbf{R}_1(2m)-\mathbf{r}\rvert}-\frac{1}{\lvert\mathbf{R}_1(2m-1)-\mathbf{r}\rvert} =\frac{2\xi-d}{(2md-\xi+\xi_0)((2m-1)d+\xi+\xi_0)}. \tag*{(26)}

The infinite sum of these terms can be explicitly calculated and gives the result

φ1:=∑m=1∞1∣R1(2m)−r∣−1∣R1(2m−1)−r∣=12d(−ψ(2d−ξ+ξ02d)+ψ(d+ξ+ξ02d))(27)\varphi_1 := \sum_{m=1}^{\infty} \frac{1}{\lvert\mathbf{R}_1(2m)-\mathbf{r}\rvert} - \frac{1}{\lvert\mathbf{R}_1(2m-1)-\mathbf{r}\rvert} = \frac{1}{2d}\left(-\psi\left(\frac{2d-\xi+\xi_0}{2d}\right)+\psi\left(\frac{d+\xi+\xi_0}{2d}\right)\right) \tag*{(27)}

where ψ(z)=Γ′(z)/Γ(z)\psi(z)=\Gamma'(z)/\Gamma(z) denotes the psi function, see [14], eq. 5.2.2., which is a special case of the polygamma functions ψ(n)(z)\psi^{(n)}(z), see [14], § 5.15.

The result (27) has been derived with the aid of computer algebra software; for similar series evaluations see [14], chapter 5.7.

Analogously, we obtain for the second generation

1∣R2(2m)−r∣−1∣R2(2m−1)−r∣=12dm+ξ−ξ0−1d(2m−1)−ξ−ξ0=d+2ξ(2dm+ξ−ξ0)(−2dm+d+ξ+ξ0)(28)\frac{1}{\lvert\mathbf{R}_2(2m)-\mathbf{r}\rvert} - \frac{1}{\lvert\mathbf{R}_2(2m-1)-\mathbf{r}\rvert} = \frac{1}{2dm+\xi-\xi_0} - \frac{1}{d(2m-1)-\xi-\xi_0} = \frac{d+2\xi}{(2dm+\xi-\xi_0)(-2dm+d+\xi+\xi_0)} \tag*{(28)}

and the infinite sum of these terms is given by

φ2:=∑m=1∞1∣R2(2m)−r∣−1∣R2(2m−1)−r∣=12d(ψ(d−ξ−ξ02d)−ψ(2d+ξ−ξ02d))(29)\varphi_2 := \sum_{m=1}^{\infty} \frac{1}{\lvert\mathbf{R}_2(2m)-\mathbf{r}\rvert} - \frac{1}{\lvert\mathbf{R}_2(2m-1)-\mathbf{r}\rvert} = \frac{1}{2d}\left(\psi\left(\frac{d-\xi-\xi_0}{2d}\right)-\psi\left(\frac{2d+\xi-\xi_0}{2d}\right)\right) \tag*{(29)}

Figure 6. The potential of a point charge and an infinite number of mirror charges φ(ξ,ξ0)\varphi(\xi,\xi_0), restricted to the ξ\xi-axis according to (30) (arbitrary units). The coordinate of the point charge is chosen as ξ0=−d/3\xi_0=-d/3 where φ(ξ,ξ0)\varphi(\xi,\xi_0) diverges according to the contribution of the Coulomb potential. Remarkably, φ(ξ,ξ0)\varphi(\xi,\xi_0) vanishes for ξ→±d/2\xi\to\pm d/2 as it is required for a Dirichlet Green’s function.

Finally, we add the contribution of both generations and the Coulomb potential of the point charge to obtain

φ(ξ,ξ0):=φ1+φ2+1∣ξ−ξ0∣,(30)\varphi(\xi,\xi_0):=\varphi_1+\varphi_2+\frac{1}{\lvert\xi-\xi_0\rvert}, \tag*{(30)}

which, up to the pre-factor q4πε0\frac{q}{4\pi\varepsilon_0}, represents an explicit evaluation of the series representation (23) restricted to the ξ\xi-axis.

The limit value of the potential (23) for ξ→±d/2\xi\to\pm d/2 has already be shown to vanish in Section II. As a test one can check that this also applies for its evaluation (30) at the ξ\xi-axis. Interestingly, this test has to use the recurrence equation ψ(z+1)=ψ(z)+1/z\psi(z+1)=\psi(z)+1/z for the psi function, see [14], eq. 5.5.2.

Motion of the point charge in the plate capacitor

Approximation of the potential energy

At this point, it is advisable to introduce dimensionless variables. We therefore consider the following quantities with their problem-oriented units:

LengthMassChargeEnergyTime
ddm>0m>0q>0q>0qUqUmqUd\sqrt{\frac{m}{qU}}d

Table 1.

Here mm denotes the mass of the point charge and U>0U>0 the voltage between the plates of the capacitor, assuming that the left plate is positively charged and the right one negatively. We will use the same letters to designate the dimensionless variables as before, without the risk of confusion. The potential energy of a pair of mirror charges with distance 2d2d divided by the energy unit qUqU will define a dimensionless parameter λ\lambda typical for the relative strength of influence effects:

λ:=q24πε02d1qU=q8πε0Ud.(31)\lambda:= \frac{q^2}{4\pi\varepsilon_0 2d}\frac{1}{qU}=\frac{q}{8\pi\varepsilon_0Ud}. \tag*{(31)}

For the potential energy of the point charge corresponding to the mirror charges we obtain

V1=2λ(φ1+φ2)∣ξ=ξ0=(27,29)λ(ψ(12−ξ0)+ψ(ξ0+12)+2γ)=(17)λ(ψ(1−x)+ψ(x)+2γ),(32)V_1=2\lambda\left(\varphi_1+\varphi_2\right)\big|_{\xi=\xi_0}\overset{\text{(27,29)}}{=}\lambda\left(\psi\left(\frac{1}{2}-\xi_0\right)+\psi\left(\xi_0+\frac{1}{2}\right)+2\gamma\right)\overset{\text{(17)}}{=}\lambda\left(\psi(1-x)+\psi(x)+2\gamma\right), \tag*{(32)}

where γ=0.57721…\gamma=0.57721\ldots denotes the Euler-Mascheroni constant.

For the motion of the point charge the potential energy of the plate capacitor without influence effects has to be added which gives the total potential energy

V(x)=V1−x=λ(ψ(1−x)+ψ(x)+2γ)−x.(33)V(x)=V_1-x=\lambda\left(\psi(1-x)+\psi(x)+2\gamma\right)-x. \tag*{(33)}

From this the total force F(x)F(x) upon the point charge can be obtained by

F(x)=−ddxV(x)=1+λ(ψ(1)(1−x)−ψ(1)(x)).(34)F(x)=-\frac{d}{dx}V(x)=1+\lambda\left(\psi^{(1)}(1-x)-\psi^{(1)}(x)\right). \tag*{(34)}

It is obvious that the corresponding equation of motion cannot be solved in closed form but only numerically. Therefore it is sensible to devise an analytical approximation of V(x)V(x) which allows for an exact solution of the equation of motion. For small xx one can derive the asymptotic expansion

V1(x)=−λx−2λζ(3)x2+O(x4),(35)V_1(x)=-\frac{\lambda}{x}-2\lambda\zeta(3)x^2+O\left(x^4\right), \tag*{(35)}

using the zeta function ζ(z)\zeta(z), see [14], chapter 25. The leading term −λx-\frac{\lambda}{x} represents the potential energy due to the left primary mirror charge. Analogously, in the neighbourhood of the right plate we have

V1(x)=−λ1−x−2λζ(3)(x−1)2+O((x−1)4),(36)V_1(x)=-\frac{\lambda}{1-x}-2\lambda\zeta(3)(x-1)^2+O\left((x-1)^4\right), \tag*{(36)}

where λx−1\frac{\lambda}{x-1} represents the potential energy due to the right primary mirror charge. The potential energy, which is only due to the secondary mirror charges will be defined by

V2(x)=V1(x)+λ(1x+11−x).(37)V_2(x)=V_1(x)+\lambda\left(\frac{1}{x}+\frac{1}{1-x}\right). \tag*{(37)}

It is almost constant, see Figure 7.

Figure 7. The potential energy V1(x)V_1(x) of a point charge due to an infinite number of mirror charges (dashed curve) and its analytical approximation Vapp(x)V_{\mathrm{app}}(x) consisting of three parts according to (39) (green, red and cyan curves). Moreover, we show the potential energy V2(x)V_2(x) only due to the secondary mirror charges according to (37), see the black curve. The parameter λ\lambda has been chosen as λ=0.01\lambda= 0.01.

In the middle between the plates the potential V1V_1 can be better approximated by the Taylor expansion

V1(x)=−2λlog⁡(4)−14λζ(3)(x−12)2+O((x−12)4),(38)V_1(x)=-2\lambda\log(4)-14\lambda\zeta(3)\left(x-\frac{1}{2}\right)^2+O\left(\left(x-\frac{1}{2}\right)^4\right), \tag*{(38)}

which represents the influence of all mirror charges in the vicinity of the point x=1/2x = 1/2 at which the forces of all mirror charges cancel each other out. Hence we will choose the following analytical approximation VappV_{\mathrm{app}} of the partial potential V1(x)V_1(x)

Vapp(x)={−λx:0<x≤1/3,−2λlog⁡(4)−14λζ(3)(x−12)2:1/3<x<2/3,−λ1−x:2/3≤x<1.(39)V_{\mathrm{app}}(x) = \begin{cases} -\frac{\lambda}{x} & : 0 < x \le1/3, \\ -2\lambda\log(4)-14\lambda\zeta(3)\left(x-\frac{1}{2}\right)^2 & : 1/3 < x < 2/3, \\ -\frac{\lambda}{1-x} & : 2/3 \le x < 1 . \end{cases} \tag*{(39)}

For an example of the quality of this approximation see Figure 7.

Motion under the influence of a single mirror charge

The calculations considered in this subsection can be viewed on the one hand as an approximation of the motion of a point charge in the domain 0<x<1/30 < x < 1/3 or 2/3<x<12/3 < x < 1 of the plate capacitor according to the analytical approximation of V(x)V(x) in this domain, see (39). On the other hand, these calculations can also be interpreted independently as the solution of the exact equation of motion of a point charge in the presence of only one conductive plate. We will consider the two cases of motion near the left or right plate separately.

Motion under the influence of the left plate, large energy case

Again using dimensionless quantities we can write the potential energy as

V(x)=−x−λx.(40)V(x)=-x-\frac{\lambda}{x}. \tag*{(40)}

As with all one-dimensional mechanical problems with constant total energy EE, the inverse function t(x)t(x) can be expressed by the indefinite integral

t(x)=∫dx2(E−V(x))=12∫dxE+x+λx=12∫2yE+y2+λy2 dy=2∫y2Ey2+y4+λ dy,(41)t(x)=\int\frac{dx}{\sqrt{2(E-V(x))}}=\frac{1}{\sqrt{2}}\int\frac{dx}{\sqrt{E+x+\frac{\lambda}{x}}}=\frac{1}{\sqrt{2}}\int\frac{2y}{\sqrt{E+y^2+\frac{\lambda}{y^2}}}\,dy=\sqrt{2}\int\frac{y^2}{\sqrt{Ey^2+y^4+\lambda}}\,dy, \tag*{(41)}

where we have substituted x=y2x=y^2. It is clear that this kind of expression can be written in terms of elliptical integrals. Here we will only discuss the large energy case, that is,

E>2λ.(42)E > 2\sqrt{\lambda}. \tag*{(42)}

For the other cases see Appendix § 5. In order to transform the integral (41) into some kind of standard form we factor the polynomial in the square root according to

Ey2+y4+λ=(A2+y2)(B2+y2),(43)E y^2+y^4+\lambda=\left(A^2+y^2\right)\left(B^2+y^2\right), \tag*{(43)}

where

A2=E2−E24−λandB2=E2+E24−λ.(44)A^2=\frac{E}{2}-\sqrt{\frac{E^2}{4}-\lambda} \qquad\text{and}\qquad B^2=\frac{E}{2}+\sqrt{\frac{E^2}{4}-\lambda}. \tag*{(44)}

Note that E24−λ>0\frac{E^2}{4}-\lambda>0 by virtue of (42) and hence 0<A<B0<A<B. Upon the substitution y=Auy=Au the integral (41) can be further transformed into

t=(43)2∫y2(A2+y2)(B2+y2) dy=A32∫u2(A2+A2u2)(B2+A2u2) du(45)t \overset{(43)}{=} \sqrt{2}\int\frac{y^2}{\sqrt{\left(A^2+y^2\right)\left(B^2+y^2\right)}}\,\mathrm{d}y =A^3\sqrt{2}\int\frac{u^2}{\sqrt{\left(A^2+A^2u^2\right)\left(B^2+A^2u^2\right)}}\,\mathrm{d}u \tag*{(45)}
=A2B2∫u2(1+k2u2)(1+u2) duwithk2:=A2B2<1.(46)=\frac{A^2}{B}\sqrt{2}\int\frac{u^2}{\sqrt{\left(1+k^2u^2\right)\left(1+u^2\right)}}\,\mathrm{d}u \qquad\text{with}\qquad k^2:=\frac{A^2}{B^2}<1. \tag*{(46)}

This integral can be evaluated with the aid of computer algebra software which gives

t=-\mathrm{i}B\sqrt{2}\left(E\left(\mathrm{i}\sinh^{-1}(u)\mathrel{\left|}k^2\right.\right)-F\left(\mathrm{i}\sinh^{-1}(u)\mathrel{\left|}k^2\right.\right)\right) \tag*{(47)}
=-\mathrm{i}B\sqrt{2}\left(E\left(\mathrm{i}\sinh^{-1}\left(\frac{\sqrt{x}}{A}\right)\mathrel{\left|}k^2\right.\right)-F\left(\mathrm{i}\sinh^{-1}\left(\frac{\sqrt{x}}{A}\right)\mathrel{\left|}k^2\right.\right)\right), \tag*{(48)}

where F(z∣k2)F(z|k^2) and E(z∣k2)E(z|k^2) denote the Legendre’s elliptic integrals of the first and second kind, see [14], §19, and the above substitutions have been undone. It is possible to transform (47) into a manifest real form, which looks a little more complicated, but this is not necessary because (48) can be evaluated numerically without any problems.

Motion under the influence of the right plate, large energy case

The mathematical treatment is largely analogous to that of the preceding subsection. The potential energy now assumes the form

V(x)=−x−λ1−x,(49)V(x)=-x-\frac{\lambda}{1-x}, \tag*{(49)}

and t(x)t(x) is given by the integral

t(x)=12∫dxE+x+λ1−x=−12∫2yE+1−y2+λy2 dy=−2∫y2(E+1)y2−y4+λ dy,(50)t(x)=\frac{1}{\sqrt{2}}\int\frac{\mathrm{d}x}{\sqrt{E+x+\frac{\lambda}{1-x}}} =-\frac{1}{\sqrt{2}}\int\frac{2y}{\sqrt{E+1-y^2+\frac{\lambda}{y^2}}}\,\mathrm{d}y =-\sqrt{2}\int\frac{y^2}{\sqrt{(E+1)y^2-y^4+\lambda}}\,\mathrm{d}y, \tag*{(50)}

where we have substituted 1−x=y21-x=y^2. We factor the polynomial in the square root according to

(E+1)y2−y4+λ=(A~2−y2)(B~2+y2),(51)(E+1)y^2-y^4+\lambda=\left(\widetilde{A}^2-y^2\right)\left(\widetilde{B}^2+y^2\right), \tag*{(51)}

where

A~2=E+12+(E+1)24+λandB~2=−E+12+(E+1)24+λ,(52)\widetilde{A}^2=\frac{E+1}{2}+\sqrt{\frac{(E+1)^2}{4}+\lambda} \qquad\text{and}\qquad \widetilde{B}^2=-\frac{E+1}{2}+\sqrt{\frac{(E+1)^2}{4}+\lambda}, \tag*{(52)}

such that 0<B~<A~0<\widetilde{B}<\widetilde{A} if the large energy condition holds in the slightly weakened form

E>0.(53)E>0. \tag*{(53)}

Upon the substitution y=A~uy=\widetilde{A}u the integral (50) can be further transformed into

t=(51)−2∫y2(A~2−y2)(B~2+y2) dy=−A~32∫u2(A~2−A~2u2)(B~2+A~2u2) du(54)t \overset{(51)}{=}-\sqrt{2}\int\frac{y^2}{\sqrt{\left(\widetilde{A}^2-y^2\right)\left(\widetilde{B}^2+y^2\right)}}\,\mathrm{d}y =-\widetilde{A}^3\sqrt{2}\int\frac{u^2}{\sqrt{\left(\widetilde{A}^2-\widetilde{A}^2u^2\right)\left(\widetilde{B}^2+\widetilde{A}^2u^2\right)}}\,\mathrm{d}u \tag*{(54)}
=−A~2B~2∫u2(1+k~2u2)(1−u2) duwithk~2:=A~2B~2>1.(55)=-\frac{\widetilde{A}^2}{\widetilde{B}}\sqrt{2}\int\frac{u^2}{\sqrt{\left(1+\widetilde{k}^2u^2\right)\left(1-u^2\right)}}\,\mathrm{d}u \qquad\text{with}\qquad \widetilde{k}^2:=\frac{\widetilde{A}^2}{\widetilde{B}^2}>1. \tag*{(55)}

This integral is evaluated as

t=−B~2(E(arcsin⁡(u)∣−k2)−F(arcsin⁡(u)∣−k2))(56)t=-\widetilde{B}\sqrt{2}\left(E\left(\arcsin(u)\mid-k^2\right)-F\left(\arcsin(u)\mid-k^2\right)\right) \tag*{(56)}
=−B~2(E(arcsin⁡(1−xB)∣−k2)−F(arcsin⁡(1−xB)∣−k2)),(57)=-\widetilde{B}\sqrt{2}\left(E\left(\arcsin\left(\frac{\sqrt{1-x}}{B}\right)\mid-k^2\right)-F\left(\arcsin\left(\frac{\sqrt{1-x}}{B}\right)\mid-k^2\right)\right), \tag*{(57)}

where the above substitutions have been undone.

Motion under the influence of both plates

In the domain 1/3<x<2/31/3<x<2/3 the motion of the point charge will be approximated by a second order Taylor expansion of the exact potential according to (39). The equation of motion can be solved exactly for this approximation, and the inverse function t(x)t(x), which is considered in analogy to the other cases, reads

t(x)=12λ7ζ(3)log⁡(2λ7ζ(3)2e+λlog⁡(256)+7λ(1−2x)2ζ(3)+2x+14λ(2x−1)ζ(3)+1).(58)t(x)=\frac{1}{2\sqrt{\lambda}\sqrt{7\zeta(3)}}\log\left(2\sqrt{\lambda}\sqrt{7\zeta(3)}\sqrt{2e+\lambda\log(256)+7\lambda(1-2x)^2\zeta(3)+2x+14\lambda(2x-1)\zeta(3)+1}\right). \tag*{(58)}

Comparison of numerical integration and analytical approximation

Finally, we compare the numerical integration of the equation of motion for the point charge with the exact potential energy (33) with the three analytical approximations obtained so far and find excellent agreement for the choice of parameters λ=0.01\lambda=0.01 and E=0.4E=0.4, see Figure 8. The time TT required for the complete motion from x=0x=0 to x=1x=1 is calculated by numerical integration as Tnum=0.731182T_{\mathrm{num}}=0.731182 and by analytical approximations as Taa=0.731476T_{\mathrm{aa}}=0.731476, which corresponds to a relative deviation of ≈4×10−4\approx4\times10^{-4}. For the λ\lambda-dependence of TT see § 7.

Figure 8. We show the numerical integration x(t)x(t) of the equation of motion for the point charge with the exact potential energy (33) (dashed black curve), with the choice of parameters λ=0.01\lambda=0.01 and E=0.4E=0.4. This result is compared with the analytical approximations of the motion corresponding to the intervals 0≤x<1/30\leq x<1/3 (green curve), 1/3≤x≤2/31/3\leq x\leq2/3 (red curve), and 2/3<x≤12/3<x\leq1 (cyan curve), as explained in the main text. The difference between the two types of curves is not visible.

On the λ\lambda-dependence of the influence correction

Without influence effects, that is, for λ=0\lambda=0, the time required for the complete motion from x=0x=0 to x=dx=d can be elementarily calculated as

∫0ddx2m(E+Uqdx)=2md(E+qU−E)qU,(59)\int_{0}^{d}\frac{\mathrm{d}x}{\sqrt{\frac{2}{m}\left(E+\frac{Uq}{d}x\right)}}=\frac{\sqrt{2m}d\left(\sqrt{E+qU}-\sqrt{E}\right)}{qU}, \tag*{(59)}

or, using dimensionless quantities, as

T(0)=2(E+1−E).(60)T(0)=\sqrt{2}\left(\sqrt{E+1}-\sqrt{E}\right). \tag*{(60)}

Taking into account influence effects, this time can be written as

T(λ)=∫0112(E−V(x)) dx,(61)T(\lambda)=\int_{0}^{1}\frac{1}{\sqrt{2\left(E-V(x)\right)}}\,\mathrm{d}x, \tag*{(61)}

with the exact potential energy V(x)V(x) given by (33). We define the “influence correction” by

δT:=T(0)−T(λ).(62)\delta T:=T(0)-T(\lambda). \tag*{(62)}

and will investigate its λ\lambda-dependence.

Naively, one would think that δT\delta T goes linearly with λ\lambda to 00, so that δT=O(λ)\delta T=O(\lambda). However, the numerical results cast doubt on this expectation, see Figure 9, and rather suggest that δT=O(−λlog⁡λ)\delta T=O(-\lambda\log\lambda).

Figure 9. Plot of the influence correction δT\delta T according to (62) divided by λ=e−na\lambda=e^{-na} as a function of nn. The total energy EE has been chosen as E=0.4E=0.4. We observe an asymptotically linear dependence for n→∞n\to\infty, indicated by a red line, which suggests that δT=O(−λlog⁡λ)\delta T=O(-\lambda\log\lambda).

It remains to verify this conjecture analytically. The obvious attempt to develop the integrand of (61) into a Taylor series with respect to λ\lambda fails because already the linear term cannot be integrated over xx due to the divergences of the form ∼1/x\sim1/x and ∼1/(1−x)\sim1/(1-x).

On the other hand, one can derive from this failure the hope that the O(−λlog⁡λ)O(-\lambda\log\lambda) behavior of the influence correction can already be derived for the analytical approximations of V(x)V(x), which was considered in § 4 A for the domains close to x=0x=0 and x=1x=1, see (39). For this purpose, we introduce the influence correction due to the analytical approximations of V(x)V(x) as ΔT=ΔLT+ΔRT\Delta T = \Delta_L T + \Delta_R T and write these terms as integrals using the substitution x=y2x = y^2 introduced above:

ΔT=ΔLT+ΔRT=∫01/2(y2E+y2−y2E+y2+λ/y2) dy+∫1/21(y2E+y2−y2E+y2+λ/(1−y2)) dy.(63)\Delta T = \Delta_L T + \Delta_R T = \int_0^{1/2} \left( \frac{y^2}{\sqrt{E+y^2}} - \frac{y^2}{\sqrt{E+y^2+\lambda/y^2}} \right) \, \mathrm{d}y + \int_{1/2}^{1} \left( \frac{y^2}{\sqrt{E+y^2}} - \frac{y^2}{\sqrt{E+y^2+\lambda/(1-y^2)}} \right) \, \mathrm{d}y . \tag*{(63)}

First, let us consider

ΔLT=∫01/2(y2E+y2−y2E+y2+λ/y2) dy=∫01/2H(y,λ) dy.(64)\Delta_L T = \int_0^{1/2} \left( \frac{y^2}{\sqrt{E+y^2}} - \frac{y^2}{\sqrt{E+y^2+\lambda/y^2}} \right) \, \mathrm{d}y = \int_0^{1/2} H(y,\lambda) \, \mathrm{d}y . \tag*{(64)}

The family of functions H(y,λ)H(y,\lambda) seems to asymptotically scale with λ\sqrt{\lambda} w. r. t. argument and function value, see Figure Figure 10. Hence we introduce the scaled family of functions

Family of functions H and K

Figure 10. Left panel: The family of functions H(y,λ)H(y,\lambda) according to (64) for λ=10−n\lambda= 10^{-n}, n=2,…,5n = 2,\ldots,5, starting with the blue curve for n=2n = 2. The total energy has been chosen as E=0.4E = 0.4. Right panel: The corresponding family of scaled functions K(z,λ)K(z,\lambda) according to (65), together with the limit function K(z,0)K(z,0) for n→∞n \to\infty (dashed black curve). K(z,0)K(z,0) has its maximum α=1.06166\alpha= 1.06166 at z=β=1.24301z = \beta= 1.24301, in accordance with (68).

K(z,λ)=1λH(λz,λ),(65)K(z,\lambda) = \frac{1}{\sqrt{\lambda}} H(\sqrt{\lambda}z,\lambda) , \tag*{(65)}

such that

ΔLT=∫01/2H(y,λ) dy=λ∫01/(2λ)K(z,λ) dz.(66)\Delta_L T = \int_0^{1/2} H(y,\lambda) \, \mathrm{d}y = \lambda\int_0^{1/(2\sqrt{\lambda})} K(z,\lambda) \, \mathrm{d}z . \tag*{(66)}

For λ→0\lambda\to0 the function K(z,λ)K(z,\lambda) assumes the limit value

K(z,0)=2z(1E−zEz2+1),(67)K(z,0) = \sqrt{2}z \left( \frac{1}{\sqrt{E}} - \frac{z}{\sqrt{Ez^2+1}} \right) , \tag*{(67)}

with a maximum of

α=K(β,0)=3−51+5Eatz=β=5−E2E,(68)\alpha= K(\beta,0) = \frac{3-\sqrt{5}}{\sqrt{1+\sqrt{5}E}} \quad\text{at} \quad z = \beta= \sqrt{\frac{\sqrt{5}-E}{2E}} , \tag*{(68)}

see Figure Figure 10, right panel. It is possible to explicitly calculate the integral for ΔLT\Delta_L T using the limit value of K(z,0)K(z,0):

∫01/(2λ)K(z,0) dz=E−E(E+4λ)42E3/2λ+log⁡(E+E+4λ2λ)2E3/2.(69)\int_0^{1/(2\sqrt{\lambda})} K(z,0) \, \mathrm{d}z = \frac{E-\sqrt{E(E+4\lambda)}}{4\sqrt{2}E^{3/2}\lambda} + \frac{\log\left(\frac{\sqrt{E}+\sqrt{E+4\lambda}}{2\sqrt{\lambda}}\right)}{\sqrt{2}E^{3/2}} . \tag*{(69)}

The first term of (69) has the series expansion

E−E(E+4λ)42E3/2/λ=−122E3/2+λ22E5/2+O(λ2),(70)\frac{E-\sqrt{E(E+4\lambda)}}{4\sqrt{2}E^{3/2}/\lambda}=-\frac{1}{2\sqrt{2}E^{3/2}}+\frac{\lambda}{2\sqrt{2}E^{5/2}}+O(\lambda^2), \tag*{(70)}

and hence does not contribute to the order O(−λlog⁡λ)O(-\lambda\log\lambda) we are looking for. However, the second term has the asymptotic expansion

log⁡(E+E+4λ2λ)2E3/2=log⁡(E)−log⁡(λ)22E3/2+λ2E5/2+O(λ2),(71)\frac{\log\left(\frac{\sqrt{E}+\sqrt{E+4\lambda}}{2\sqrt{\lambda}}\right)}{\sqrt{2}E^{3/2}}=\frac{\log(E)-\log(\lambda)}{2\sqrt{2}E^{3/2}}+\frac{\lambda}{\sqrt{2}E^{5/2}}+O(\lambda^2), \tag*{(71)}

and thus yields, together with (B6), the desired result ΔLT=O(−λlog⁡λ)\Delta_L T=O(-\lambda\log\lambda). The calculations to prove ΔRT=O(−λlog⁡λ)\Delta_R T=O(-\lambda\log\lambda) are largely analogous and can be omitted here.

Summary and Outlook

In theoretical physics, it is common to describe the motion of so-called “test particles” in external fields in such a way that the influence of the particle on the sources of the fields is neglected. An example of this is the motion of a point charge in the field of a plate capacitor, which is assumed to be constant. In this case, the neglected influence consists of the formation of induced charges on the capacitor surfaces, which change the motion of the point charge. In this example, however, it is also possible to set up an exact equation of motion taking into account the influence charges. Although its integration can still only be done numerically, analytical approximations are possible, which allow an approximate description of the motion using elliptical integrals. These calculations would probably be practically impossible without the use of computer algebra software.

There are various mathematical and physical aspects that are interesting beyond the specific problem dealt with. On the one hand, there is the aforementioned fact that there is a real interaction between the source and the particle, which can nevertheless be treated exactly.

The Green’s function of the plate capacitor, known from the literature by a double series or mixed series/integral representation using Bessel and trigonometric functions, could also be obtained in this work by a series representation of the Coulomb potentials of infinitely many mirror charges. Both representations have been shown to agree at the central line passing to the point charge and the mirror charges. The sum of infinitely many Coulomb potentials of equidistant identical charges diverges; therefore, for a converging series representation, the Coulomb potentials of pairs of mirror charges of different signs must be combined. We have thus found a rare case of only conditional convergence in a physical problem.

For the motion of the point charge, the potential of the mirror charges at the location of the point charge is required. Remarkably, the series representation of the potential can be summed up for this special case, resulting in another application of the psi function in physics. Other such applications are known from the theory of the Casimir effect, see [15], eq. (5.28), and from the evaluation of Feynman diagrams, see [16]. To a first approximation, the attractive Coulomb potential of the directly neighboring mirror charge acts in the vicinity of the plates. Although the divergence of this potential stems from the un-physical limiting case of the fusion of charge and mirror charge, the strong attraction by the influence charges can certainly be seen as a rudimentary consideration of the work of emission that prevents a charge from leaving a capacitor plate without further ado. Of course, this is only a rough model of the exact interaction between the point charge and the many-body system of the capacitor plate, limited by the continuum description of the macroscopic charge distribution in electrostatics.

In the approximation of the Coulomb potential of the directly neighboring mirror charge, the motion of the point charge is described by a so-called Cornell potential, which also appears in the classical Stark effect. The one-dimensional motion in this potential leads to elliptic integrals. This is an important special case of motion in the three-dimensional Cornell potential, which is an integrable problem of classical mechanics (with the conserved quantities energy, angular momentum around the symmetry axis and another constant of motion related to the generalized Runge–Lenz vector, see [7], [17]). We have calculated the motion of the point charge in the Cornell-potential approximation using elliptic integrals of Legendre’s form as well as of Carlson’s form. The latter are a modern variant of elliptic integrals that allow uniform representations without having to deal with case distinctions.

The ratio of the Coulomb energy of the neighboring mirror charge to the typical capacitor energy is defined as a dimensionless parameter λ=q28πε0Ud\lambda=\frac{q^2}{8\pi\varepsilon_0Ud}. This makes the transit time TT from, say, the left to the right plate, a function of λ\lambda. Contrary to the naive expectation, λ↦T(λ)\lambda\mapsto T(\lambda) is not analytical at λ=0\lambda=0, but T(λ)=T(0)+O(−λlog⁡λ)T(\lambda)=T(0)+O(-\lambda\log\lambda). Here too, the singular character of the Coulomb potentials of the primary mirror charges means that the influence correction behaves differently than expected.

If one inserts the elementary charge for qq into the above formula for λ\lambda and chooses the problem-oriented parameters UU, dd of order 1 in SI units, one gets λ∼10−9\lambda\sim10^{-9}, a very small value. However, if we choose a macroscopic charge value of, say, q=4×10−9 Asq = 4 \times10^{-9}\,\mathrm{As}, together with d=0.04 md = 0.04\,\mathrm{m}, m=0.003 kgm = 0.003\,\mathrm{kg} and U=103 VU = 10^{3}\,\mathrm{V}, we obtain λ∼0.45\lambda\sim0.45. This means that in the experiments mentioned in the Introduction with an electrostatic pendulum, influence effects cannot be neglected. For example, if the motion of the point charge has the total energy of E=2×10−5 JE = 2 \times10^{-5}\,\mathrm{J}, which is E=5E = 5 in dimensionless quantities, we would obtain a transit time of T(0)≈0.66 sT(0) \approx0.66\,\mathrm{s} compared with T(λ)≈0.58 sT(\lambda) \approx0.58\,\mathrm{s}, which corresponds to a relative influence correction of δT/T(0)≈12%\delta T/T(0) \approx12\%.

However, our theory is only suitable for describing the effects on the plates of the capacitor, not on a conductive sphere that moves between the plates. Such influence effects could be described in the lowest order by dipole moments induced by the external electric field. The electric dipole potential leads to additional influence effects on the plates, which could be calculated with an analog method considering an infinite number of mirror dipoles. This would be a possible extension and application of our work in the future.

Convergence of the series representation of the Green’s function

In order to investigate the convergence of the series representation of the putative (Dirichlet) Green’s function (23) it is sensible to split off the three terms Φ0(r,r0)\Phi_0(\mathbf{r},\mathbf{r}_0) corresponding to the Coulomb potential of the point charge and of the two primary mirror charges. This part can be considered as a distribution satisfying the Poisson equation

ΔΦ0(r,r0)=ΔΦ(r,r0)=−qε0δ(r−r0)(72)\Delta\Phi_0(\mathbf{r},\mathbf{r}_0)=\Delta\Phi(\mathbf{r},\mathbf{r}_0)=-\frac{q}{\varepsilon_0}\delta(\mathbf{r}-\mathbf{r}_0) \tag*{(72)}

and can be regularized by cancelling two Coulomb potentials in the case of x=0x=0 (when the point charge hits the left plate and coincides with the left primary mirror charge) or x=dx=d (analogously for the right plate). Then it will be appropriate to postulate (A1) only for the open domain 0<x<d0<x<d between the plates.

After having subtracted Φ0(r,r0)\Phi_0(\mathbf{r},\mathbf{r}_0) from the series representation (23) we consider only the remaining contributions from the first generation of mirror charges, i. e.,

Φ1(r,r0)=q4πε0∑n=2∞(−1)n∣r−R1(n)∣(73)\Phi_1(\mathbf{r},\mathbf{r}_0)=\frac{q}{4\pi\varepsilon_0}\sum_{n=2}^{\infty}\frac{(-1)^n}{\lvert\mathbf{r}-\mathbf{R}_1(n)\rvert} \tag*{(73)}

such that −d/2≤ξ,ξ0≤d/2-d/2\leq\xi,\xi_0\leq d/2 for r=(ξ,η,ζ)\mathbf{r}=(\xi,\eta,\zeta) and r0=(ξ0,0,0)=(x−d/2,0,0)\mathbf{r}_0=(\xi_0,0,0)=(x-d/2,0,0). The contribution from the second generation can be treated completely analogously.

It is obvious that the series (A2) does not converge unconditionally, i. e., independent of all rearrangements of the members of the series, since the partial series for the odd and even nn diverge in the same way as the harmonic series ∑n=1∞1n\sum_{n=1}^{\infty}\frac{1}{n}. However, convergence can be achieved if we group together the nn-th and the (n+1)(n+1)-th term of (A2) for even nn, having different signs. This corresponds to a mirror charge qq with coordinate X1(n)=nd+x−d/2>d/2X_1(n)=nd+x-d/2>d/2 and its mirror −q-q with coordinate X1(n+1)=−nd−x−d/2<−d/2X_1(n+1)=-nd-x-d/2<-d/2. We consider the contribution to (A2) corresponding to this pair of mirror charges (ignoring the pre-factor q4πε0\frac{q}{4\pi\varepsilon_0}):

ϕn:=1∣r−R1(n)∣−1∣r−R1(n+1)∣,n even.(74)\phi_n:=\frac{1}{\lvert\mathbf{r}-\mathbf{R}_1(n)\rvert}-\frac{1}{\lvert\mathbf{r}-\mathbf{R}_1(n+1)\rvert},\qquad n\ \text{even}. \tag*{(74)}

We use the notation explained in Figure 11 and write

Sketch of the geometry underlying the contribution of a pair of mirror charges to the Green’s function

Figure 11. Sketch of the geometry underlying the contribution (A3) of a pair of mirror charges to the Green’s function Φ(r,r0)\Phi(\mathbf{r},\mathbf{r}_0). The plates of the capacitor are indicated by blue and red vertical lines. We display the triangle with the edges ana_n, bnb_n and cnc_n spanned by two mirror charges with numbers nn, n+1n+1, nn even, of the first generation (magenta dot and circle) and the point with coordinates r=(−d/2+x1,η,ζ)\mathbf{r}=(-d/2+x_1,\eta,\zeta), where the potential is evaluated. The height of the triangle is the cylinder coordinate ρ=η2+ζ2\rho=\sqrt{\eta^2+\zeta^2}. Note that bn≤anb_n\leq a_n always applies because the left capacitor plate (blue line) lies exactly in the middle between the mirror charges.

ϕn=1bn−1an=an−bnanbn=(an−bn)(an+bn)anbn(an+bn)=an2−bn2anbn(an+bn)≥0,(75)\phi_n=\frac{1}{b_n}-\frac{1}{a_n}=\frac{a_n-b_n}{a_nb_n}=\frac{(a_n-b_n)(a_n+b_n)}{a_nb_n(a_n+b_n)}=\frac{a_n^2-b_n^2}{a_nb_n(a_n+b_n)}\geq0, \tag*{(75)}

since an≥bna_n\geq b_n always applies and an=bna_n=b_n only for x=0x=0, see Figure 11. This term can be further evaluated with the help of Pythagoras theorem which yields

an2=ρ2+(cn2+x1)2,bn2=ρ2+(cn2−x1)2,(76)a_n^2=\rho^2+\left(\frac{c_n}{2}+x_1\right)^2,\qquad b_n^2=\rho^2+\left(\frac{c_n}{2}-x_1\right)^2, \tag*{(76)}

see Figure 11, and hence

an2−bn2=2cnx1.(77)a_n^2-b_n^2=2c_nx_1. \tag*{(77)}

We conclude

ϕn=1bn−1an=(A4,A6)2cnx1anbn(an+bn)≤2x1anbn≤2danbn≤2d(cn2+x1)(cn2−x1)(78)\phi_n=\frac{1}{b_n}-\frac{1}{a_n}\overset{\mathrm{(A4,A6)}}{=}\frac{2c_nx_1}{a_nb_n(a_n+b_n)}\leq\frac{2x_1}{a_nb_n}\leq\frac{2d}{a_nb_n}\leq\frac{2d}{\left(\frac{c_n}{2}+x_1\right)\left(\frac{c_n}{2}-x_1\right)} \tag*{(78)}
=2d14cn2−x12≤8dcn2−4d2=8d(2dn+x)2−4d2≤8d4d2n2−4d2=2d(n2−1)<2d(n−1)2.(79)=\frac{2d}{\frac{1}{4}c_n^2-x_1^2}\leq\frac{8d}{c_n^2-4d^2}=\frac{8d}{(2dn+x)^2-4d^2}\leq\frac{8d}{4d^2n^2-4d^2}=\frac{2}{d(n^2-1)}<\frac{2}{d(n-1)^2}. \tag*{(79)}

Here we have used in line (A7) that, due to the triangle inequality, cn≤an+bnc_n\leq a_n+b_n and that 0≤x1≤d0\leq x_1\leq d due to the restriction of r\mathbf{r} to the space between the plates. In the line (A8) we have further used the equation

cn=X1(n)−X1(n+1)=(nd+x−d/2)−(−nd−x−d/2)=2dn+x≥2dn,(80)c_n=X_1(n)-X_1(n+1)=(nd+x-d/2)-(-nd-x-d/2)=2dn+x\geq2dn, \tag*{(80)}

and that n2−1>(n−1)2n^2-1>(n-1)^2 for n≥2n\geq2.

Summarizing, we have found that the series ∑n=2,4,6,…∞ϕn\sum_{n=2,4,6,\ldots}^{\infty}\phi_n is majorized by the convergent series 2d∑n=2,4,6,…∞1(n−1)2=π24d\frac{2}{d}\sum_{n=2,4,6,\ldots}^{\infty}\frac{1}{(n-1)^2}=\frac{\pi^2}{4d}. Hence also Φ1(r,r0)\Phi_1(\mathbf{r},\mathbf{r}_0) converges w. r. t. the arrangement of the series according to the pairs ϕn\phi_n. Since the upper bound in (A8) does not depend on r\mathbf{r} or r0\mathbf{r}_0 the convergence is uniform in the region −d/2≤ξ,ξ0≤d/2-d/2\leq\xi,\xi_0\leq d/2.

In order to verify the Poisson equation (A1) we have to show

ΔΦ1=Δ(Φ−Φ0)=0,for−d2<ξ<d2.(81)\Delta\Phi_1=\Delta\left(\Phi-\Phi_0\right)=0,\quad\text{for}\quad-\frac{d}{2}<\xi<\frac{d}{2}. \tag*{(81)}

This follows because that the sources of the potential Φ1\Phi_1 consist of an infinite number of secondary mirror charges outside the space between the plates.

Comparison with other representations of the Green’s function

As mentioned in the Introduction there exist other series representations of the Green’s function G(r,r′)G(\mathbf{r},\mathbf{r}') of the plate capacitor, especially in [1], problem 3.17a and 3.17b. The latter is a mixed series/integral representation. It will be instructive to compare these with our explicit form (30) obtained for the special case where both, r\mathbf{r} and r′\mathbf{r}', are chosen to lie on the ξ\xi-axis between the plates.

We first consider the double series representation of G(r,r′)G(\mathbf{r},\mathbf{r}') according to [1], problem 3.17a:

G(r,r′)=4L∑n=1∞∑m=−∞∞eim(ϕ−ϕ′)sin⁡(nπzL)sin⁡(nπz′L)Im(nπLρ<)Km(nπLρ>).(82)G(\mathbf{r},\mathbf{r}')=\frac{4}{L}\sum_{n=1}^{\infty}\sum_{m=-\infty}^{\infty}e^{\mathrm{i}m(\phi-\phi')}\sin\left(\frac{n\pi z}{L}\right)\sin\left(\frac{n\pi z'}{L}\right)I_m\left(\frac{n\pi}{L}\rho_<\right)K_m\left(\frac{n\pi}{L}\rho_>\right). \tag*{(82)}

Here z,ρ,ϕz,\rho,\phi are cylindrical coordinates of r\mathbf{r}, analogously for z′,ρ′,ϕ′z',\rho',\phi' and the two plates of the capacitor lie at z=0z=0 and z=Lz=L. (Hence the zz-axis corresponds to our ξ\xi-axis and the LL to our dd.) The Im(z)I_m(z) and Km(z)K_m(z) denote the modified Bessel functions of the 1st resp. 2nd kind. ρ<\rho_< denotes the minimum of ρ\rho and ρ′\rho' and ρ>\rho_> the corresponding maximum. If both position vectors r\mathbf{r} and r′\mathbf{r}' are located at the zz-axis we have ρ=ρ′=0\rho=\rho'=0 and hence ρ<=ρ>=0\rho_<=\rho_>=0. Due to Im(0)=0I_m(0)=0 for m≠0m\ne0 and I0(0)=1I_0(0)=1 only the m=0m=0-term of the sum survives and the term eim(ϕ−ϕ′)e^{\mathrm{i}m(\phi-\phi')} can be replaced by 1. This is highly plausible due to the azimuthal symmetry of the problem. However, the term K0(nπLρ>)K_0\left(\frac{n\pi}{L}\rho_>\right) diverges for ρ>→0\rho_>\to0 and hence the series representation (B1) is not suitable for comparison with (30), at least if we adopt the obvious choice and identify the central line with the zz-axis. It may be, nevertheless, useful for evaluating the Green’s function outside the zz-axis, see below.

Next we consider the mixed series/integral representation of G(r,r′)G(\mathbf{r},\mathbf{r}') according to [1], problem 3.17b:

G(r,r′)=2∑m=−∞∞∫0∞dk eim(ϕ−ϕ′)Jm(kρ)Jm(kρ′)sinh⁡(kz<)sinh⁡(k(L−z>))sinh⁡(kL),(83)G(\mathbf{r},\mathbf{r}')=2\sum_{m=-\infty}^{\infty}\int_{0}^{\infty}dk\,e^{\mathrm{i}m(\phi-\phi')}J_m(k\rho)J_m(k\rho')\frac{\sinh(kz_<)\sinh(k(L-z_>))}{\sinh(kL)}, \tag*{(83)}

where the Jm(z)J_m(z) denote the Bessel functions and z<z_<, resp. z>z_>, the minimum, resp. maximum, of zz and z′z'. Again, we restrict ourselves to the case ρ=ρ′=0\rho=\rho'=0 and hence the sum shrinks to the term with m=0m=0. The remaining integral can be expressed in terms of the hypergeometric functions 2F1(a,b,c;z){}_2F_1(a,b,c;z), as far as the indefinite form is considered, and the definite form can be simplified to

G(r,r′)=12L(ψ(1−z<+z>2L)−ψ(1+z<−z>2L)−ψ(z>−z<2L)+ψ(z<+z>2L)).(84)G(\mathbf{r},\mathbf{r}')=\frac{1}{2L}\left(\psi\left(1-\frac{z_<+z_>}{2L}\right)-\psi\left(1+\frac{z_<-z_>}{2L}\right)-\psi\left(\frac{z_>-z_<}{2L}\right)+\psi\left(\frac{z_<+z_>}{2L}\right)\right). \tag*{(84)}

This expression can be shown to agree with (30) for both cases, z≤z′z\leq z' and z≥z′z\geq z', by using the equations z=ξ+L/2z=\xi+L/2, z′=ξ0+L/2z'=\xi_0+L/2, L=dL=d and ψ(z+1)=ψ(z)+1/z\psi(z+1)=\psi(z)+1/z. For the latter see [14], eq. 5.5.2.

Finally, we want to use the series representation (B1) to calculate the total influenced charge on the left plate. To do this, we choose ρ=0\rho=0 again, differentiate the resulting expression w. r. t. z′z' and set z′=0z'=0. This gives essentially the normal component of the electric field at the left hand plate and is hence proportional to the surface density σ1\sigma_1:

σ1=−q4π∂∂z′G(r,r′)∣z′=0=−qπL∑n=1∞sin⁡nπzL(nπLcos⁡nπz′L⏟1)K0(nπρL)=−qL2∑n=1∞nsin⁡nπzLK0(nπρL).(85)\sigma_1=-\frac{q}{4\pi}\left.\frac{\partial}{\partial z'}G(\mathbf{r},\mathbf{r}')\right|_{z'=0}=-\frac{q}{\pi L}\sum_{n=1}^{\infty}\sin\frac{n\pi z}{L}\left(\frac{n\pi}{L}\underbrace{\cos\frac{n\pi z'}{L}}_{1}\right)K_0\left(\frac{n\pi\rho}{L}\right)=-\frac{q}{L^2}\sum_{n=1}^{\infty}n\sin\frac{n\pi z}{L}K_0\left(\frac{n\pi\rho}{L}\right). \tag*{(85)}

The total influenced charge Q1Q_1 on the left hand plate is obtained by the integral

Q1=∫0∞σ1 2πρ dρ.(86)Q_1=\int_0^\infty\sigma_1\,2\pi\rho\,d\rho. \tag*{(86)}

In a sub-calculation we obtain

∫0∞K0(nπρL)2πρ dρ=2πL2n2π2∫0∞K0(r)r dr⏟1=2L2n2π.(87)\int_0^\infty K_0\left(\frac{n\pi\rho}{L}\right)2\pi\rho\,d\rho=\frac{2\pi L^2}{n^2\pi^2}\int_0^\infty\underbrace{K_0(r)r\,dr}_{1}=\frac{2L^2}{n^2\pi}. \tag*{(87)}

using the substitution r=nπρLr=\frac{n\pi\rho}{L} and [14], 10.43.19. Inserting this result into (B5) we finally obtain

Q1=−qL22L2π∑n=1∞1nsin⁡nπzL=−q(1−zL),(88)Q_1=-\frac{q}{L^2}\frac{2L^2}{\pi}\sum_{n=1}^{\infty}\frac{1}{n}\sin\frac{n\pi z}{L}=-q\left(1-\frac{z}{L}\right), \tag*{(88)}

using the Fourier series of the sawtooth wave [18]. By an analogous calculation we obtain for the influenced charge Q2Q_2 on the right hand plate

Q2=−qzL,(89)Q_2=-q\frac{z}{L}, \tag*{(89)}

such that Q1+Q2=−qQ_1+Q_2=-q. These results agree with [1], problem 1.13, where a solution using Green’s reciprocity theorem is suggested. A calculation using the set of mirror charges for the plate capacitor is much more difficult, see [4].

Motion under the influence of the left plate, arbitrary energy

If we consider the motion of a point charge in a potential V(x)=−x−λ/xV(x) = -x - \lambda/x for arbitrary energies EE, it is more convenient to use the Carlson symmetric forms of elliptic integrals, see [14], section 19, and [19], instead of the Legendre form. We will quote the corresponding equations (19.29.20), (19.29.22), (19.16.5) in [14], adapted to our notation:

∫y0y1y2 dyQ1(y)Q2(y)=13a1a2RD(U2+a1b2,U2+a2b1,U2)+y0y1U(90)\int_{y_0}^{y_1} \frac{y^2 \,\mathrm{d}y}{\sqrt{Q_1(y)Q_2(y)}} = \frac{1}{3}a_1a_2R_D\left(U^2+a_1b_2,U^2+a_2b_1,U^2\right) + \frac{y_0y_1}{U} \tag*{(90)}

where

Qj(y)=aj+bjy2,j=1,2,(91)Q_j(y) = a_j + b_jy^2,\quad j = 1,2, \tag*{(91)}
(y12−y02)U=y1Q1(y0)Q2(y0)+y0Q1(y1)Q2(y1)(92)(y_1^2-y_0^2)U = y_1\sqrt{Q_1(y_0)Q_2(y_0)} + y_0\sqrt{Q_1(y_1)Q_2(y_1)} \tag*{(92)}

and

RD(x,y,z)=32∫0∞dt(t+x)(t+y)(t+z)(t+z)(93)R_D(x,y,z) = \frac{3}{2}\int_0^\infty\frac{\mathrm{d}t}{\sqrt{(t+x)(t+y)(t+z)}(t+z)} \tag*{(93)}

These equations are especially suited to express the integral (41) in terms of the standard Carlson elliptic integral RD(x,y,z)R_D(x,y,z) (a degenerate case of the symmetric Carlson elliptic integral RJR_J of the second kind) for all choices of the energy EE (except for the special value E=−2λE = -2\sqrt{\lambda}, see below). In our case we have, see (44),

a1=E2−E24−λ,b1=1(94)a_1 = \frac{E}{2} - \sqrt{\frac{E^2}{4} - \lambda}, \qquad b_1 = 1 \tag*{(94)}
a2=E2+E24−λ,b2=1(95)a_2 = \frac{E}{2} + \sqrt{\frac{E^2}{4} - \lambda}, \qquad b_2 = 1 \tag*{(95)}

and

U=y0Ey12+λ+y14+y1Ey02+λ+y04y12−y02(96)U = \frac{y_0\sqrt{Ey_1^2+\lambda+y_1^4}+y_1\sqrt{Ey_0^2+\lambda+y_0^4}}{y_1^2-y_0^2} \tag*{(96)}

In the range −2λ<E<2λ-2\sqrt{\lambda} < E < 2\sqrt{\lambda}, see Figure 12, the parameters a1a_1 and a2a_2 will be complex, but the equation (C1) can nevertheless be used to calculate the integral (41), see Figure 14, left panel.

Sketch of various cases for the evaluation of the integral

Figure 12. Sketch of various cases for the evaluation of the integral (41). The parameter λ\lambda is chosen as λ=0.01\lambda= 0.01. The potential energy V(x)=−x−λ/xV(x) = -x - \lambda/x (black curve) has its maximum Emax⁡=−2λE_{\max} = -2\sqrt{\lambda} at x=λx = \sqrt{\lambda}. For E<−2λE < -2\sqrt{\lambda} there are two turning points x1=12(−E−E2−4λ)<x2=12(−E+E2−4λ)x_1 = \frac{1}{2}(-E - \sqrt{E^2 - 4\lambda}) < x_2 = \frac{1}{2}(-E + \sqrt{E^2 - 4\lambda}) where E=V(xi)E = V(x_i), i=1,2i = 1, 2, such that the point charge cannot enter the range x1<x<x2x_1 < x < x_2. We have indicated the example of E=−7λE = -7\sqrt{\lambda} by dotted black lines.

Comparison of the motion $x(t)$ of the point charge

Figure 14. Comparison of the motion x(t)x(t) of the point charge calculated analytically according to (C1), resp., (C8) (blue dashed curves) with the numerical results (red curves). We have chosen λ=0.01\lambda= 0.01. Left panel: Case E=−0.5E = -0.5 with complex aja_j, j=1,2j = 1, 2. Right panel: Case E=−2λ=−0.2E = -2\sqrt{\lambda} = -0.2. The two branches of the plot correspond to the motion approaching or leaving the unstable rest position of x=λ=0.1x = \sqrt{\lambda} = 0.1 where V(x)V(x) has a global maximum for t→±∞t \to\pm\infty.

We will compare the analytical form of the motion of the point charge with the results from numerical integration for two different values of EE, see the Figure 13 and Figure 14, left panel. The agreement is perfect.

Figure 13. Comparison of the motion x(t)x(t) of the point charge calculated analytically according to (C1) (blue dashed curves) with the numerical results (red curves). We have chosen λ=0.01\lambda= 0.01 and E=−0.7E = -0.7. There are two turning points x1=0.0145898x_1 = 0.0145898 and x2=0.68541x_2 = 0.68541 where the velocity of the point charge vanishes, corresponding to the example given in Figure 12. Left panel: Range 0≤x≤x10 \leq x \leq x_1. Right panel: Range x2≤x≤1x_2 \leq x \leq1.

In the special case of E=−2λE = -2\sqrt{\lambda} the two polynomials Q1(y)Q_1(y) and Q2(y)Q_2(y) coincide and the integral (41) can be evaluated in a simpler way without using elliptic integrals. This case is comparable to the motion of a mathematical pendulum which reaches its unstable resting position for t→±∞t \to\pm\infty. For E=−2λE = -2\sqrt{\lambda} we obtain

t(x)=2(x−λ)(x−14λ4log⁡((x4+λ)2(x−λ4)2))x−2λ+λx+x(97)t(x) = \frac{\sqrt{2}(x-\sqrt{\lambda})\left(\sqrt{x}-\frac{1}{4}\sqrt[4]{\lambda}\log\left(\frac{(\sqrt[4]{x}+\sqrt{\lambda})^2}{(\sqrt{x}-\sqrt[4]{\lambda})^2}\right)\right)}{\sqrt{x}\sqrt{-2\sqrt{\lambda}+\frac{\lambda}{x}+x}} \tag*{(97)}

For a comparison of this solution with numerical results see Figure 14, right panel.

The case of the motion under the influence of the right plate and arbitrary energy can be analogously treated using the Carlson symmetric forms of elliptic integrals and will not be considered here.

References

  1. [1]J. D. Jackson, Classical electrodynamics, 3rd edition, Hoboken, John Wiley & Sons, 2021.
  2. [2]J. Dougall, The determination of Green’s Function by means of cylindrical or spherical harmonics, Proceedings of the Edinburgh Mathematical Society 18, 33 – 83, (1900)
  3. [3]M. Majic, The image of a point charge in an infinite conducting cylinder, Preprint, arXiv:2001.10651 (2020).arxiv.org/abs/2001.10651
  4. [4]W. A. Newcomb, Trouble with the method of images, Am. J. Phys. 50 (7), 601 – 607, (1982)DOI
  5. [5]W. R. Smythe, Static and dynamic electricity, 3rd edition, revised printing, London, Taylor & Francis, 1998.
  6. [6]M. Seetharaman, S. Raghavan, and S. S. Vasan, Analytic WKB energy expression for the linear plus Coulomb potential, J. Phys. A: Math. Gen. 51, 455 – 462, (1983)
  7. [7]M. Courtney, N. Spellmeyer, H. Jiao, and D. Kleppner, Classical, semiclassical, and quantum dynamics in the lithium Stark system, Phys. Rev. A 16 (5), 3604 – 3620, (1995)
  8. [8]D. H. Martin and R. L. Longbrake, An electrostatic oscillator experiment, Am. J. Phys. 40 (3), 395 – 398, (1972)DOI
  9. [9]K. Asano, On the theory of an electrostatic pendulum oscillator, Am. J. Phys. 43 (5), 423 – 427, (1975)DOI
  10. [10]A. T. Pérez, Charge and force on a conducting sphere between two parallel electrodes, Journal of Electrostatics 56, 199 – 217, (2002)DOI
  11. [11]V. A. Saranin, Electrostatic oscillators, Physics-Uspekhi 55 (7) , 700 – 708, (2012)DOI
  12. [12]M. Forjan, M. Marhl, and V. Grubelnіk, Mathematical modelling of the electrostatic pendulum in school and undergraduate education, Eur. J. Phys. 35, 015022, (2014)
  13. [13]D. S. Goodman, R. D. Fischetti, A. Hodges, and H. A. Tekalp, Image-charge effects on an electrostatic pendulum in mechanical equilibrium, Am. J. Phys. 88 (3), 222 – 228, (2020)
  14. [14]NIST Digital Library of Mathematical Functions. http://dlmf.nist.gov/ , Release 1.1.1 of 2021-03-15. F. W. J. Olver, A. B. Olde Daalhuis, D. W. Lozier, B. I. Schneider, R. F. Boisvert, C. W. Clark, B. R. Miller, B. V. Saunders, H. S. Cohl, and M. A. McClain, eds.
  15. [15]E. Elizalde, Ten physical applications of spectral zeta functions, Lecture Notes in Physics 855, Berlin - Heidelberg - New York, Springer, 1995.
  16. [16]M. W. Coffey, On one-dimensional digamma and polygamma series related to the evaluation of Feynman diagrams, J. Comp. Appl. Math. 183 (1), 84 – 100, (2005)
  17. [17]M. A. Liberman, Hydrogen Atom in Electric and Magnetic Fields: Dynamical Symmetries, Superintegrable and Integrable Systems, Exact Solutions, Preprint, arXiv:2203.02730 (2022).DOI
  18. [18]E. W. Weisstein, Fourier Series–Sawtooth Wave, From MathWorld–A Wolfram Web Resource, https://mathworld.wolfram.com/FourierSeriesSawtoothWave.html, (2022)
  19. [19]B. C. Carlson, Normal elliptic integrals of the first and second kinds, Duke Math. J. 31 (3), 405 – 419, (1964)DOI

Paper details

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

views · 0

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