[Contents]

The Electrophorus,
and Other Variable-Geometry Capacitors

John Denker

— HALF-FINISHED DRAFT —

Let’s see if we can figure out how an electrophorus works. This is quite an open-ended question. I find the analysis to be interesting without being overly complicated.

There may be some good explanations in the literature, but they seem to be hopelessly outnumbered by incomplete and confusing explanations. This version in reference 1 is typical, insofar as the diagram shows an unbalanced negative charge on the dielectric ... without showing what the field lines are doing. What’s worse, the distribution of charges on the moveable conducting plate is incorrect, presumably based on a seriously incorrect guess as to what the field lines are doing. Furthermore, some critical steps in the procedure are not diagrammed.

We start by doing some warm-up exercises, to establish some terminology and to build up a baseline understanding of exotic capacitors, including situations where there is both a mutual capacitance and a self-capacitance.

## *   Contents

1  Warm-Up Exercise: Gorging and Disgorging a Capacitor
2  Warm-Up Exercise: Capacitor Near a Wall, Or Not
2.1  The Situation
2.2  Two-Terminal Mutual Capacitance
2.3  Single-Plate Self-Capacitance
2.4  Combining the Results
3  An Actual Electrophorus
4  Self-Capacitance of a Sphere and a Disk
4.1  Sphere
4.2  Disk
5  Asymmetric Capacitor : Disk + Plate
6  References

## 1  Warm-Up Exercise: Gorging and Disgorging a Capacitor

Consider the situation shown in part A of figure 1. We have a sphere that carries a charge of +4 units on the upper hemisphere and another +4 units on the lower hemisphere. Presumably there is – somewhere – a counter-charge of −8 units, but we assume it is so far away that the details don’t matter. The sphere has been cut along the equatorial plane, so that that there is a tiny gap of size g between the upper hemisphere and the lower hemisphere. However, for present purposes the gap doesn’t matter very much. Basically what we have is a sphere with a total charge of +8 units.

Figure 1: A Capacitor, with Charge and Gorge

We now turn to part B of figure 1. As before, there is a charge of +4 units on the upper hemisphere ... but now there is zero charge on the lower hemisphere. We assume the gap is so small that the mutual capacitance between the hemispheres is enormous compared to the self-capacitance. We can figure out (by symmetry or otherwise) that the 4 units of charge will distribute themselves evenly over the sphere. In the far field, the field exhibits (to leading order) a monopole field pattern, which is what we would expect from +4 units of charge. There will also be a dipole term, due to the voltage difference between the two hemispheres, but in the far field, compared to the monopole term, the dipole term will be negligible. We can understand this as follows: There will be a significant field in the gap, but because the gap is small the mutual capacitance is large, so the voltage difference between the hemispheres will be small.

Finally, we turn to part C of figure 1. As before, there is a charge of +4 units on the upper hemisphere ... but now the charge on the lower hemisphere is −4 units. In the far field, this looks like a dipole. It won’t be a very strong dipole, because the separation between the positive and negative charges is not large, but in the far field this small dipole term is all there is.

Now we get to the tricky part. Consider the following contrast:

• Suppose we say object C is a sphere, and we ask how much charge is on the sphere. The answer, obviously, is zero charge. To be more specific, you could talk about the total charge or the net charge or the honest-to-goodness charge ... but it seems simpler and better to just call it charge.
• Suppose we say object C is a capacitor, and we ask how much «charge» is on the capacitor. The answer, according to long-established convention, is that the capacitor has +4 units of «charge». I write «charge» in scare quotes here, because we have a problem.

The problem is, the word “charge” is being used in two inconsistent ways. To solve this problem, let’s define the gorge, G, so that

G :=
 q1 − q2 2
(gorge)
Q := q1 + q2     (charge)
(1)

Now we can unambiguously say that in part C of the figure, the object has zero units of charge and four units of gorge.

Meanwhile, in part B we have four units of charge and two units of gorge. In part A we have eight units of charge and zero units of gorge.

The mapping in equation 1 can be reversed as follows, as you can easily verify:

 q1 = Q/2 + G q2 = Q/2 − G
(2)

Note that according to the laws of physics, the charge Q is conserved. Specifically, Q obeys a strict local conservation law. In contrast, gorge is not conserved. Not at all. If you put many units of gorge on a capacitor – or a battery – and come back later, you might find that all of the gorge has disappeared, due to some internal leakage process.

Note that what we have here is just a special case of the three-terminal capacitor: The upper hemisphere, the lower hemisphere, and the far-away counter-electrode. The terminology of charge and gorge is not particularly good for handling three-terminal capacitors in general ... but the case we have considered here is so exceedingly common that it is worth having special tools for handling it.

## 2  Warm-Up Exercise: Capacitor Near a Wall, Or Not

### 2.1  The Situation

Consider the situation in figure 2. There is a room with conducting walls (shown in blue). The floor of the room is X by X. The height of the room is Y. Centered over the middle of the floor is a thin metallic disk of radius r (shown in red). The disk is horizontal at a height h above the floor. There is a charge Q on the disk ... and correspondingly a charge −Q on the walls of the room. The disk is not very large, and the room is much wider than it is tall, so that XYr.

Figure 2: Capacitance of a Disk in a Room

Your mission, should you decide to accept it, is to find an algebraic expression for the voltage on the disk, as a function of h. Voltage is measured relative to the walls, which we designate to serve as “chassis ground”.

The disk has an insulating handle, so you can change h arbitrarily without changing Q.

Make whatever reasonable simplifying assumptions you like. (Replacing the conducting walls with non-conducting walls would not be considered reasonable.)

As a specific numerical instance, set

 X = 10 m Y = 2 m r = 0.2 m h = 1×10−5 m initially
(3)

Charge up the disk to 1.5 volts using a battery, and then disconnect the battery. Raise the disk (maintaining constant Q during phase of the process). What happens to the voltage?

The usual jsd puzzle rules apply: Everything I have said is (to the best of my knowledge) true and non-misleading, although some details may be irrelevant. On the other hand, I have not told you everything I know. I have not told you the answer or even the method of solution ... and more importantly, the problem may be underspecified.

Again, the point is that after doing this warm-up exercise you should have a fighting chance of figuring out how an electrophorus works.

### 2.2  Two-Terminal Mutual Capacitance

We boldly assume that the bottom of the disk combines with the floor to form a capacitor ... and that the top of the disk combines with the ceiling to form another capacitor. We have a circuit with these two capacitors in parallel.

Let Q1 be the charge on the bottom capacitor, and Q2 be the charge on the top capacitor. Let the corresponding capacitances be C1 and C2.

 Q = Q1 + Q2
(4)

The two capacitors see the same voltage, so

 Q = (C1  + C2) V
(5)

The formula for a parallel-plate capacitor with an area A and a small gap g is

C =
є0
 A g
(6a)
=
є0
 π r2 g
for a disk    (6b)

If we temporarily assume that we can ignore fringing fields, we can rewrite equation 5 as

Q =
є0 π r2 (
 1 h
 1 Y−h
V
(7)

hence

V =
 1 є0 π r2

 h(Y−h) Y
Q
(8)

This result is shown by the dashed red curve in figure 3.

Figure 3: Inverse Capacitance as a Function of Height

### 2.3  Single-Plate Self-Capacitance

The result in section 2.2 is contingent on a number of assumptions.

Before we rely too much on the formula for a parallel-plate capacitor, we should consider what happens when there is a small disk in the middle of a large room. The capacitance is then dominated by the self-capacitance of disk, namely

 C = 8 є0 r
(9)

A derivation of this formula can be found in reference 2.

The inverse self-capacitance is shown by the dashed magenta line in figure 3.

It should be obvious that under no circumstances can the capacitance of a disk be less than the self-capacitance. Comparing equation 6b to equation 9 implies we cannot possibly apply the parallel-plate capacitance formula unless the gap is quite small:

g <
 π 8
r        (for mutual-capacitance model)
g >
 π 8
r        (for self-capacitance model)
(10)

### 2.4  Combining the Results

The solid green curve in figure 3 is an approximate interpolation between the two simple models. This interpolation is done by simply adding the two capacitances. This is surely an overestimate of the capacitance, and therefore an underestimate of the V/Q, i.e. a lower bound on the voltage we can achieve by manipulating the disk. The dashed curves in figure 3 serve as an upper bound. A general discussion of circular disk parallel-plate capacitor, including fringing fields, can be found in reference 3. There is no simple closed-form solution for the general case.

A better estimate is obtained by adding the two capacitances in quadrature. This is shown by the blue curve in the figure. This corresponds to using the Riesz (or Lebesgue) L2 norm. It sits in between simple addition (the L1 norm) and taking the lesser of the two contributions (the L norm).

TODO: Compare this approximate algebraic approach to doing the actual integrals, as in reference 3.

This tells us that under a rather wide range of conditions, the height of the room is almost irrelevant. However, we would not have known this is we had not done the calculation in section 2.2. The relevant length-scales are either {r and h} or {r and Yh} ... unless the disk is very large and/or the ceiling is very low. To say the same thing another way, the integral over the fringing fields converges quite quickly.

## 3  An Actual Electrophorus

It is important that the dielectric sit on an insulating support, a goodly distance from any grounded object. In other words, in the situation shown in figure 4, the capacitance is dominated by the single-plate self-capacitance. Also note that the charge on the dielectric is not mobile. The electric field lines (shown in green) extend above and below the dielectric, heading off “to infinity” as they should in any self-capacitance situation.

Figure 4: Electrophorus – Step 1

In the situation shown in figure 5, there is a tricky combination of two things going on. Half of the story involves treating the two plates as a parallel-plate capacitor, with the usual sort of mutual capacitance. Each plate of this capacitor bears a charge of ±4 units. The other half of the story involves treating the two plates together as if they were a single slightly-thicker plate, which exhibits a self-capacitance. This combined object bears a charge of −8 units.

Figure 5: Electrophorus – Step 2

Figure 6 shows the situation after the moveable plate has been grounded, and after the grounding wire has been removed. The situation is entirely dominated by the two-plate mutual capacitance. The dielectric is at a slight voltage relative to ground, but since the capacitor gap is small and the capacitance is enormous, this voltage is rather small.

Figure 6: Electrophorus – Step 3

Figure 7: Electrophorus – Step 4

Once the dielectric has been charged up, it can under some conditions be used over and over again to induce charge on the moveable disk.

## 4  Self-Capacitance of a Sphere and a Disk

### 4.1  Sphere

We start by considering the self-capacitance of a sphere:

Csphere = 4πє0 r    (11a)
=
 4π r2 r
є0
(11b)
=
 area gap
є0
(11c)

To make sense of equation 11c, we introduce the concept of effective gap, as follows: The field lines are most concentrated near the sphere, and then peter out as we get farther away. The concentrated field lines contribute the lion’s share to the total capacitance. We approximate the total energy in the field by overestimating the energy within a distance r of the surface while discarding the more distant contributions. This is shown in figure 8. The energy of the approximation (the red-shaded region) turns out to be exactly equal to the actual energy (the blue-shaded region).

Figure 8: Spherical Capacitor Energy

That means we can consider the sphere to be analogous to a parallel plate capacitor, with a suitable area and gap:

 parallel-plate capacitor spherical capacitor Area: The area of one plate The total area of the sphere (not counting the other plate). (not counting the far-distant counterelectrode). Gap: The actual gap between the plates. An “effective gap” equal to the radius of the sphere.

This can be considered a mnemonic for remembering the formula for the capacitance of a sphere. It produces the exactly-correct formula.

### 4.2  Disk

The usual way to calculate the capacitance of a thin disk is to start with a sphere and then transform it into an ellipsoid using suitable orthogonal coordinates. This can be considered a conformal transformation. See reference 2.

However, in the spirit of qualitative reasoning, we can get a rather good estimate of the capacitance of the disk using a simpler transformation. This is in some sense a scaling argument. We shrink the sphere in the Z direction but not in the X or Y direction.

We start from the parallel-plate capacitor formula (equation 6a) and the corresponding spherical-capacitor formula (equation 11c). Let’s see whether the disk upholds the same pattern, namely a capacitance in proportion to the area divided by some effective gap.

Cdisk = 0 r    (12a)
=
 4 π
 2π r2 r
є0
(12b)
=
 4 π
 area gap
є0
(12c)

We find that the disk does uphold the pattern, except for a fudge factor of 4/π, if you set the effective gap equal to the radius. That’s off by less than 30%, which is not bad for a qualitative argument.

Another way of interpreting the same result is to say that the charge on the disk is more concentrated (not as evenly spread out) compared to the sphere, so the effective gap is smaller than you might have guessed, smaller by a factor of π/4.

Note that the appropriate area includes both sides of the disk.

## 5  Asymmetric Capacitor : Disk + Plate

This can be rather well solved using "qualitative reasoning". Some people are really good at qualitative reasoning, but others not so much. It is a skill that can be taught (and learned). This example can serve as nice teachable moment.

Specifically, consider the situation shown in figure 9. We have an asymmetric capacitor consisting of a moderate-sized flat disk somewhere near an enormous flat slab.

Figure 9: Disk + Plate Capacitor

Step 1: My physics intuition says that when the disk is very near the slab, the capacitance is the same as it would be for a simple symmetric two-disk parallel-plate capacitor.

In this case "near" means g is very small compared to (π/8) r, where g is the capacitor gap and r is the radius of the disk.

Rationale: Draw the picture. Energy principles say that virtually all the relevant field lines will be in the gap. This accounts for virtually all the charge and virtually all the field lines. Experience playing with field-calculating software helps with this. See reference 4.

Step 2: When the disk is far away from the slab, the capacitance of the setup as a whole must be the same as the self-capacitance of the disk.

Rationale: For capacitances in series, we should be able to add the elastances (i.e. inverse capacitances). This is analogous to the rule for adding resistances in series. The slab has zero elastance (infinite capacitance) so the elastance of the overall situation is dominated by the elastance of the disk.

Step 3: Even if you don’t know how to find the exact result for an arbitrary-sized gap, you can get a good approximation by interpolating between the small-gap asymptote and the large-gap asymptote, as worked out in the previous steps. A preliminary version of this is discussed and diagrammed at www.av8n.com/physics/electrophorus.htm#fig-pre-electrophorus-plot

Step 4: The formula for a symmetrical capacitor is simple and well known for the case of a small gap. It is neither simple nor super-widely known – but still known (reference 3) – for an arbitrary-sized gap. We should be able to (a) invoke symmetry and (b) invoke the method of images to reduce the asymmetrical case to the corresponding symmetrical case ... /provided/ we keep track of various factors of two. The original situation should have half the elastance of the situation shown in figure 10.

Figure 10: Disk + Plate + Image

The factor of 2 arises for a prosaic reason, namely that we are now measuring the voltage from the bottom disk to the top disk, rather from slab to disk. Twice as much V per unit Q means twice as much elastance.

By symmetry and/or by the method of images, that must be the same as

Figure 11: Disk + Disk (Double Gap)

In the large-gap limit, the elastance here is just the series combination of the self-elastance of two disks. From this we obtain the final result, namely that the capacitance of the original asymmetrical situation (disk plus slab) should be double the capacitance of the corresponding symmetrical situation /with a double gap/.

In the small-gap case the two factors of two cancel out, so that disk + slab has the same capacitance as disk + disk (with the same gap) as expected. As the gap gets larger, the asymmetrical case starts to approach the large-gap asymptote twice as soon as you might have guessed by naively looking at the symmetrical formula.

Step 5: There is a piece of intuition that is implicit in all the previous steps. Consider the extreme large-gap limit. Zoom out your point of view so that the disk becomes essentially a single point. You know how the electric field of a point charge falls off with distance. You know that the Poyinting field energy goes like the square of the field. If you integrate this energy over all space, starting some not-too-small distance R away from the disk, then the integral converges. So the intuition is that whatever is going on in the far field never matters very much for electrostatics problems.

Now let’s consider what happens if we extend the plate to include the floor, ceiling, and four walls of a room. This raises questions about the “hall of mirrors” effect. I’m not worried too much about that, given that the overall charge is neutral. The images we see in the “hall of mirror” can be paired to form dipoles, and the dipole field falls off even faster than the monopole field. The lattice sum converges quite quickly, even with an infinitude of sources. Sure, it contributes a correction term in principle, but under a wide range of practical situations (electrophorus disk small compared to the size of the room) this won’t change the story in any qualitative way.

For more about the physics of capacitance, including odd-shaped capacitors, see reference 5

## 6  References

1.
Wikipedia, article, “Electrophorus”
http://en.wikipedia.org/wiki/Electrophorus

2.
Konstantin K. Likharev
“CLASSICAL ELECTRODYNAMICS”
http://mysbfiles.stonybrook.edu/~klikharev/505-506/F10-S11/
The self-capacitance of a thin disk is covered in chapter 2:
http://mysbfiles.stonybrook.edu/~klikharev/505-506/F10-S11/Ch02.pdf

3.
G. T. Carlson and B. L. Illman
“The circular disk parallel plate capacitor”
American Journal of Physics 62 12, pp. 1099 (December 1994)
http://ajp.aapt.org/resource/1/ajpias/v62/i12/p1099_s1

4.
John Denker,
“Electric Charge and Potential: Spreadsheets for solving Laplace’s equation”
www.av8n.com/physics/laplace.htm

5.
John Denker, “Capacitance” www.av8n.com/physics/capacitance.htm
[Contents]