Image Charge for a Grounded Plane

This is the method of images' signature performance — the problem every physicist solves first and remembers forever. A point charge $q$ hovers a distance $d$ above an infinite grounded conducting plane. The plane's surface teems with induced charge whose distribution nobody prescribed; the previous articles built the entire boundary-value apparatus for exactly this impasse. And then the solution turns out to be one line: delete the plane, place $-q$ at the mirror point, done. The two-charge potential vanishes on the mid-plane by symmetry, uniqueness stamps the certificate, and a problem with an unknown continuous charge distribution collapses into freshman superposition. But the one-liner is only the opening move; the riches are in the harvest, and this article collects all of it with full arithmetic. The induced surface density comes out in closed form, $\sigma(s) = -qd/2\pi(s^2+d^2)^{3/2}$ — a perfect bell of negative charge crowding under the point charge, with half the total inside radius $\sqrt{3}\,d$ . Integrated, it delivers exactly $-q$ , confirming the totals theorem of the method article. The charge is pulled toward the plane by $q^2/16\pi\varepsilon_0 d^2$ — computable two independent ways, as the image's Coulomb tug and as the integrated electrostatic pressure of the surface-charge article, which agree to the digit. And the energy, $-q^2/16\pi\varepsilon_0 d$ , earned here by an honest work integral, is precisely half the naive charge-image value — the half-price theorem performing live. One fictitious charge, four exact physical results. When the same trick meets a sphere in the next article, the mirror will bend — but every harvesting tool sharpened here will transfer intact.

Key takeaways

Setup: charge $q$ at height $d$ above an infinite grounded plane; image: $-q$ at depth $d$ on the far side. $V = \tfrac{1}{4\pi\varepsilon_0}(q/r_+ - q/r_-)$ vanishes on the plane and is exact for $z\ge0$ .
Induced surface density: $\sigma(s) = -\dfrac{qd}{2\pi(s^2+d^2)^{3/2}}$ — peak $-q/2\pi d^2$ directly beneath the charge, falling like $1/s^3$ far out.
The map integrates to exactly $-q$ (all of $q$ 's flux lands on the plane), with half the induced charge inside radius $s_{1/2} = \sqrt{3}\,d$ .
Force on the charge: $F = \dfrac{q^2}{16\pi\varepsilon_0 d^2}$ toward the plane — equal to the image's Coulomb pull and to the integrated $\sigma^2/2\varepsilon_0$ surface pressure (two-route check).
Energy: $W = -\dfrac{q^2}{16\pi\varepsilon_0 d}$ by the work integral — half the naive two-charge energy, the linear-response half-price rule made explicit.
KEY IDEA: one mirror charge converts an unknown induced-charge problem into closed forms for potential, charge map, force, and energy — each independently checkable, and all four checked here.

The core insight

A charge $q$ at height $d$ above a grounded plane is exactly solved (for $z\ge0$ ) by its mirror image $-q$ : the potential vanishes on the plane, the induced density $-qd/2\pi(s^2+d^2)^{3/2}$ integrates to $-q$ , the attraction is $q^2/16\pi\varepsilon_0d^2$ , and the stored energy $-q^2/16\pi\varepsilon_0 d$ is half the naive image value.

Where this lives today

The plane image runs quietly inside a surprising amount of technology. Every electron gun and field-emission microscope lives with the Schottky effect: an electron escaping a metal feels its own image attraction, potential $-e^2/16\pi\varepsilon_0 z$ — this article's energy formula read as a function of height — which tilts against any applied extraction field $F$ and lowers the escape barrier by $\Delta W = \sqrt{e^3F/4\pi\varepsilon_0}$ : about $1.2\,\mathrm{eV}$ at $F = 10^9\,\mathrm{V/m}$ , a huge bite out of a $4.5\,\mathrm{eV}$ work function and a term sitting explicitly inside the Fowler–Nordheim equation that calibrates every field-emission source. Electrostatic chucks — the devices that hold silicon wafers flat during lithography without touching their front side — are engineered image attraction: electrode, dielectric, wafer, and a force budget computed from exactly the $\sigma^2/2\varepsilon_0$ integral of Derivation B. And whenever an antenna engineer mounts a quarter-wave monopole on a conducting ground plane, the plane's image supplies the missing half of the dipole — the radiation pattern above the plane is the dipole's, at half the input impedance. From photocopier toner clinging to a drum, to dust adhesion on solar panels, to the tip–sample force floor in atomic-force microscopy: wherever charge meets conductor, this article's four numbers — $\sigma(s)$ , $-q$ , $q^2/16\pi\varepsilon_0d^2$ , $-q^2/16\pi\varepsilon_0 d$ — are the first estimate, and often the last.

Origins · 1 minJames Clerk Maxwell and how this was discovered

James Clerk Maxwell

1831–1879 · Cambridge and the Cavendish

Maxwell is the reason generations could see this solution. His 1873 Treatise devoted a chapter to Thomson's images and closed with engraved plates of breathtaking field-line diagrams — among them the charge-over-plane pattern, every curve computed by hand from the two-term potential of this article — that became the canonical mental furniture of electromagnetism; Feynman learned fields from redrawn Maxwell plates, and so, at whatever remove, did you. The choice was philosophical as much as pedagogical: Maxwell had translated Faraday's lines of force into mathematics, and images were his favorite demonstration that a hard boundary problem becomes visible geometry — "the image is the optical appearance of the electrical state of the surface," he wrote, taking Thomson's borrowed optical metaphor fully seriously. First Cavendish Professor, unifier of electricity, magnetism, and light, dead at forty-eight: he spent his last decade building the institutional and visual language in which results like $\sigma(s) = -qd/2\pi(s^2+d^2)^{3/2}$ stopped being formulas and became pictures.

History

1828

— Green's Essay solves the half-space with the kernel that contains the mirror charge

1845

— Thomson's Paris letters to Liouville debut electrical images, plane and sphere alike

1873

— Maxwell's Treatise engraves the charge-over-plane field pattern into the visual canon of physics

1914

— Schottky applies the image energy to electron emission: barrier lowering by

\sqrt{e^3F/4\pi\varepsilon_0}

1928

— Fowler and Nordheim build the image correction into the quantum theory of field emission

The physical picture

Hold a positive charge above a grounded metal floor and watch the field rearrange. Far above, its field lines fan out as if free; but every line heading downward must end perpendicular on the metal (the conductor articles' boundary condition), and the plane, wired to Earth's inexhaustible reservoir, pulls up whatever negative charge is needed to terminate them. The result is a field that looks exactly like the upper half of a dipole pattern — because mathematically it is one: the unique field with $q$ 's singularity above and $V=0$ on the plane coincides, in the upper half-space, with the field of the pair $\{+q$ at $d$ , $-q$ at $-d\}$ .

Figure

The mirror construction — a dipole's upper half, a conductor's exact field

The picture already answers the qualitative questions. Where does the induced charge sit? Where the field lines land — overwhelmingly in the patch of radius $\sim d$ beneath the charge; the density must peak at the foot of the perpendicular and die off in the far field like the dipole it mimics. Which way is the force? Toward the plane, always: the charge is attracted to the oppositely-signed screen it itself induced (flip $q$ 's sign and the screen flips with it — attraction survives, as the $q^2$ in the force formula insists). How big is the pull at close range? Growing like $1/d^2$ with the doubled distance $2d$ in the denominator — the mirror keeps pace as you approach, which is why surfaces grab charged dust, toner, and scanning-probe tips with such enthusiasm. One geometric subtlety deserves its own paragraph: the force and energy involve $2d$ , but the harvested surface charge is a function of the real geometry only. The image is scaffolding; every physical output — $\sigma(s)$ , total $-q$ , field at the plane — refers to the actual half-space. The derivations keep the scaffolding and the building rigorously separate.

The construction and the harvest

Coordinates: plane at $z=0$ , charge $q$ at $(0,0,d)$ , cylindrical radius $s$ from the foot of the perpendicular. The claimed solution for the physical region $z\ge0$ :

V(s,z) = \frac{q}{4\pi\varepsilon_0}\left[\frac{1}{\sqrt{s^2+(z-d)^2}} - \frac{1}{\sqrt{s^2+(z+d)^2}}\right] \qquad (1)

where: the first term is the real charge's Coulomb potential ( $r_+$ the distance from it), the second the image's ( $r_-$ from the mirror point $(0,0,-d)$ ). On the plane $z=0$ the two distances are equal and $V$ vanishes identically — the boundary condition holds on the entire surface, not just under the charge. Off the plane the two terms never cancel, and near the real charge the first term's singularity is exactly the prescribed $q$ — Poisson's equation with the right source. With $V\to0$ at infinity, the uniqueness checklist closes: equation (1) is the potential in the upper half-space (and pure fiction below it). The harvest headline is the induced-charge map, obtained from the boundary condition of the surface-charge article, $\sigma = -\varepsilon_0\,\partial V/\partial n$ :

\sigma(s) = -\frac{q\,d}{2\pi\,(s^2+d^2)^{3/2}} \qquad (2)

where: $s$ — distance along the plane from the foot of the perpendicular. Read it physically: at $s=0$ the density is $-q/2\pi d^2$ — halve the height and the peak quadruples; by $s=d$ it has already fallen to $2^{-3/2}\approx0.35$ of the peak; far out it decays like $-qd/2\pi s^3$ , the signature of the dipole's far field. Negative everywhere, exactly as the field-line picture demanded: nowhere on the plane do lines leave.

Figure

The induced-charge bell — sigma along the plane, and where the screen lives

Derivation AConstruction, certification, and the induced-charge map

We build equation (1) from symmetry, certify it against the uniqueness checklist, and differentiate our way to the surface-charge bell (2). The construction step is short by design — the whole point of images is that finding the candidate is easy here — so the care goes into the certification (each clause of the method article's theorem, checked concretely) and into keeping signs honest in the harvest, where the normal derivative is a classic sign trap.

Step 1. Construct by mirror symmetry: the pair $\{+q$ at $(0,0,d)$ , $-q$ at $(0,0,-d)\}$ has a nodal plane. Any point of $z=0$ is equidistant from the two charges, $r_+ = r_- = \sqrt{s^2+d^2}$ , so their potentials cancel point by point: $V=0$ on the whole plane, for free, by geometry.

Step 2. Certify the three clauses. Sources in $\Omega$ ( $z>0$ ): only the real $q$ — the image sits at $z=-d$ , outside. Boundary: $V=0$ on $z=0$ , Step 1. Infinity: a finite dipole's potential dies like $1/r^2$ , faster than required. All hypotheses of the first uniqueness theorem hold; (1) is exact for $z\ge0$ . Why the mirror had to carry charge exactly $-q$ at depth exactly $d$ : any other magnitude or depth breaks the equidistance cancellation somewhere on the plane. The construction has zero adjustable slack — a feature, since there is also exactly one solution to hit.

Step 3. Compute the field at the plane. Differentiate (1) with respect to $z$ and evaluate at $z=0$ :

\frac{\partial V}{\partial z}\bigg|_{z=0} = \frac{q}{4\pi\varepsilon_0}\left[\frac{-(z-d)}{r_+^3} + \frac{(z+d)}{r_-^3}\right]_{z=0} = \frac{q}{4\pi\varepsilon_0}\cdot\frac{2d}{(s^2+d^2)^{3/2}}.

The tangential components vanish on the plane (equal and opposite by the mirror symmetry) — consistent with a conductor surface, and a good silent check.

Step 4. Harvest $\sigma$ with the normal anchored out of the conductor, i.e. $\hat{\mathbf{n}} = +\hat{z}$ into the region:

\sigma(s) = -\varepsilon_0\,\frac{\partial V}{\partial z}\bigg|_{z=0} = -\frac{q\,d}{2\pi\,(s^2+d^2)^{3/2}}.

Why the sign is trustworthy: negative for positive $q$ — field lines land, they don't launch — and the field just above the plane, $E_z = \sigma/\varepsilon_0 < 0$ , points into the metal, matching the picture. Both sanity checks would flip visibly if the normal had been misanchored.

Step 5. Profile anatomy for the record: peak $\sigma(0) = -q/2\pi d^2$ ; fall-off to $35\%$ at $s=d$ ; asymptote $-qd/2\pi s^3$ . The map's moments are the next derivation's business. Physical cross-checks: the tangential field vanishes on the plane (as required for a conductor surface), $\sigma<0$ everywhere for positive $q$ (field lines land, never launch), and the dimensions work out: $[qd/s^3] = \mathrm{C/m^2}$ ✓.

Go deeper · Derivation B · 2 minThe three payoffs — total charge, force by two routes, and the half-price energy

Derivation BThe three payoffs — total charge, force by two routes, and the half-price energy

Now the map pays out. We integrate $\sigma$ to confirm the totals theorem's $-q$ ; we compute the force on the charge twice — as the image's Coulomb pull and, independently, as the integrated electrostatic pressure on the plane — and watch the routes agree; and we earn the energy by the work integral, catching the famous factor of two in the act. Three payoffs, three independent checks on one construction.

Step 1. Total induced charge. Integrate (2) over the plane in rings, $\mathrm{d}a = 2\pi s\,\mathrm{d}s$ , substituting $u = s^2+d^2$ , $\mathrm{d}u = 2s\,\mathrm{d}s$ :

Q_{\text{ind}} = -\int_{d^2}^{\infty} \frac{qd}{2\pi u^{3/2}}\,\pi\,\mathrm{d}u = -\frac{qd}{2}\left[\frac{-2}{\sqrt{u}}\right]_{d^2}^{\infty} = -\frac{qd}{2}\cdot\frac{2}{d} = -q.

Exactly the image charge, as the totals theorem promised — every line of flux from $q$ ends on the plane. A bonus moment: the charge inside radius $s$ is $-q\left[1 - d/\sqrt{s^2+d^2}\right]$ , which reaches $-q/2$ at $s_{1/2} = \sqrt{3}\,d$ — half the screen lives within $1.73$ heights of the foot.

Step 2. Force, route one — the image pull. In the physical region the induced charge's field is the image's field, so the force on $q$ is Coulomb attraction across the doubled distance $2d$ :

F = \frac{1}{4\pi\varepsilon_0}\,\frac{q^2}{(2d)^2} = \frac{q^2}{16\pi\varepsilon_0\,d^2}, \quad\text{toward the plane.}

Step 3. Force, route two — integrate the skin tension. The plane's surface feels the outward electrostatic pressure $\sigma^2/2\varepsilon_0$ (derived in Surface Charge and Field at a Conductor); by Newton's third law its total equals the force on the charge. With the same $u$ -substitution:

F = \int_0^\infty \frac{\sigma^2}{2\varepsilon_0}\,2\pi s\,\mathrm{d}s = \frac{q^2d^2}{4\pi\varepsilon_0}\int_0^\infty \frac{s\,\mathrm{d}s}{(s^2+d^2)^3} = \frac{q^2d^2}{4\pi\varepsilon_0}\cdot\frac{1}{4d^4} = \frac{q^2}{16\pi\varepsilon_0\,d^2}.

Why this agreement is worth the integral: the two routes share no intermediate algebra — one never mentions $\sigma$ , the other never mentions the image — yet land on the same number, exactly the cross-method consistency the uniqueness article told us to exploit.

Step 4. Energy by the work integral. Drag $q$ in from infinity along the axis, paying against the (position-dependent!) image force $F(z') = q^2/16\pi\varepsilon_0 z'^2$ :

W = -\int_\infty^d F(z')\,\mathrm{d}z' = -\frac{q^2}{16\pi\varepsilon_0}\int_d^\infty \frac{\mathrm{d}z'}{z'^2} = -\frac{q^2}{16\pi\varepsilon_0\,d}.

Step 5. Confront the naive answer. A fixed pair $\pm q$ at separation $2d$ stores $-q^2/8\pi\varepsilon_0 d$ — twice ours. The discrepancy is not an error but the method article's half-price theorem caught red-handed: as $q$ approached, its "partner" was being conjured in proportion, so the average bill ran at half rate. Equivalently: the naive formula double-counts by including field energy in the fictional region below the plane, where the true field is zero. Check the general rule $W = \tfrac12 qV_{\text{im}}$ : $V_{\text{im}} = -q/4\pi\varepsilon_0(2d)$ , so $\tfrac12 qV_{\text{im}} = -q^2/16\pi\varepsilon_0 d$ ✓ — three expressions, one number.

Limiting cases

Case	Result	Meaning
Directly beneath the charge	$\sigma(0) = -q/2\pi d^2$	peak density; quadruples when height halves
Far along the plane, $s\gg d$	$\sigma \to -qd/2\pi s^3$	dipole far field; total converges
Half-charge radius	$s_{1/2} = \sqrt{3}\,d$	the screen is compact: half within $1.73d$
Whole plane	$Q_{\text{ind}} = -q$ exactly	totals theorem confirmed by integration
Approach, $d\to0$	$F = q^2/16\pi\varepsilon_0d^2 \to \infty$	surfaces grab charges; adhesion of toner and dust
Retreat, $d\to\infty$	$F\to0$ as $1/d^2$ , $W\to0$ as $1/d$	plane forgotten; free Coulomb charge recovered

Common trapThe energy is just the Coulomb energy of the charge and its image

The single most-committed error in this problem: $W \ne -q^2/8\pi\varepsilon_0 d$ . That naive two-charge formula describes a world with a real $-q$ assembled at fixed strength — our world has an induced screen that grew from zero as $q$ arrived, and the honest work integral (or the region-restricted field-energy integral, which excludes the fictional lower half-space) gives exactly half: $W = -q^2/16\pi\varepsilon_0 d$ . Meanwhile the force is the full, unhalved image force — energy halves, forces never do. If you differentiate: $-\mathrm{d}W/\mathrm{d}d = -q^2/16\pi\varepsilon_0 d^2$ ... magnitude $q^2/16\pi\varepsilon_0d^2$ ✓ — consistent, because the image recedes as you retreat. The two ledgers agree only if you keep each one honest.

Worked examples

1 · The full harvest with numbers (confidence builder)

A charge $q = 5.0\,\mathrm{nC}$ sits $d = 2.0\,\mathrm{cm}$ above a grounded plane. Find the peak induced density, the potential at the point halfway up the axis, the force on the charge, and the energy of the configuration.

Step 1. Peak density, equation (2) at $s=0$ :

\sigma(0) = -\frac{q}{2\pi d^2} = -\frac{5.0\times10^{-9}}{2\pi(0.020)^2} = -2.0\times10^{-6}\,\mathrm{C/m^2} = -2.0\,\mu\mathrm{C/m^2}.

Step 2. Potential on the axis at $z = 1.0\,\mathrm{cm}$ (with $kq = 44.95\,\mathrm{V\cdot m}$ ): $r_+ = 1.0\,\mathrm{cm}$ , $r_- = 3.0\,\mathrm{cm}$ :

V = 44.95\left(\frac{1}{0.010} - \frac{1}{0.030}\right) = 44.95\,(100 - 33.3) = 3.0\times10^{3}\,\mathrm{V}.

(The real charge's term alone would say $4.5\,\mathrm{kV}$ — the induced screen shaves off a third even at half-height.)

Step 3. Force, with $kq^2 = (8.99\times10^{9})(5.0\times10^{-9})^2 = 2.25\times10^{-7}\,\mathrm{N\cdot m^2}$ :

F = \frac{kq^2}{(2d)^2} = \frac{2.25\times10^{-7}}{(0.040)^2} = 1.4\times10^{-4}\,\mathrm{N},

toward the plane — about the weight of a housefly, applied to a nanocoulomb.

Step 4. Energy: $W = -kq^2/4d = -(2.25\times10^{-7})/(0.080) = -2.8\times10^{-6}\,\mathrm{J}$ . It costs $2.8\,\mu\mathrm{J}$ to drag this charge off to infinity — and the naive image-pair formula would have billed $5.6\,\mu\mathrm{J}$ .

2 · Reading the induced-charge map (exam level)

For the same configuration, characterize the screen quantitatively: the density and surface field at $s = d$ , the radius containing half the induced charge, the radius containing $90\%$ , and the fraction of the screen a grounded disk of radius $10\,\mathrm{cm}$ would capture.

Step 1. At $s = d = 2.0\,\mathrm{cm}$ : $(s^2+d^2)^{3/2} = (2d^2)^{3/2} = 2\sqrt{2}\,d^3$ , so

\sigma(d) = \frac{\sigma(0)}{2\sqrt{2}} = \frac{-2.0}{2.83} = -0.70\,\mu\mathrm{C/m^2}, \qquad E = \frac{|\sigma|}{\varepsilon_0} = \frac{7.0\times10^{-7}}{8.854\times10^{-12}} = 7.9\times10^{4}\,\mathrm{V/m}

pointing into the plane — the surface-charge article's boundary condition, read in reverse as a field map.

Step 2. Enclosed-charge function from Derivation B: $Q(s) = -q\left[1 - d/\sqrt{s^2+d^2}\right]$ . Half charge: $d/\sqrt{s^2+d^2} = \tfrac12 \Rightarrow s_{1/2} = \sqrt{3}\,d = 3.5\,\mathrm{cm}$ .

Step 3. Ninety percent: $d/\sqrt{s^2+d^2} = 0.1 \Rightarrow s_{90} = d\sqrt{100-1} = 9.95\,d \approx 20\,\mathrm{cm}$ . The screen is compact at its core but long-tailed: the last ten percent is smeared over everything beyond ten heights.

Step 4. A grounded disk of radius $10\,\mathrm{cm} = 5d$ : $Q(5d)/(-q) = 1 - 1/\sqrt{26} = 1 - 0.196 = 0.804$ . An $80\%$ screen — which is why "infinite plane" results work so well over lab-sized ground planes: a plate a few heights across already captures the great bulk of the flux, and force errors fall off faster still: they scale as the fourth power of the missing-charge fraction, so the $19.6\%$ of charge lying beyond $5d$ costs only about $0.15\%$ of the force. (It is also why the construction quietly fails for a charge near the edge of a real plate — no mirror symmetry, no image, back to the grids of the earlier articles.)

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §3.2.1–3.2.3 · Jackson, Classical Electrodynamics, 3rd ed., §2.1 · Purcell & Morin, Electricity and Magnetism, 3rd ed., §3.4 · Zangwill, Modern Electrodynamics, §8.3. The key results can be confirmed by four self-contained checks: (1) $V=0$ on $z=0$ follows immediately from $r_+=r_-$ at the plane; (2) $\sigma=-\varepsilon_0\partial_z V|_{z=0}$ reproduces equation (2); (3) $\int\sigma\,\mathrm{d}a=-q$ via the substitution $u=s^2+d^2$ ; (4) $F=q^2/16\pi\varepsilon_0 d^2$ and $W=-q^2/16\pi\varepsilon_0 d$ , with the factor-of-two relative to the naive two-charge energy accounted for by the half-price theorem.

Builds on: Method of Images · Uniqueness Theorems for the Potential · Surface Charge and Field at a Conductor · Boundary Conditions for the Potential

Leads to: Image Charges for a Conducting Sphere · Laplace's Equation (E2) · Separation of Variables in Cartesian Coordinates

Why is the charge attracted to the plane even though the plane is neutral overall... wait, is it neutral?

It isn't — that's the resolution. The plane is grounded: it traded charge with the Earth, importing exactly $-q$ (Derivation B's integral) to terminate the incoming flux. The attraction is ordinary Coulomb attraction to that imported screen. Even an ungrounded neutral conductor would attract the charge (induced charge separates, negative crowding closer than positive — the polarization force), but the grounded plane's pull is stronger and cleaner: a net $-q$ arranged at effective distance $2d$ .

Does the sign of q matter anywhere in the results?

Only where it should. Flip $q\to-q$ : the image flips too, the induced screen turns positive, $\sigma$ flips sign point by point, and the potential map reverses. But the force ( $\propto q^2$ ) stays attractive and the energy ( $\propto q^2$ ) stays negative — a conductor pulls on either sign of charge, which is why image adhesion is universal (dust, toner, ions of both polarities). Any result linear in $q$ flips; anything quadratic survives.

Real ground planes are finite. When is the infinite-plane answer good enough?

When the plate extends several heights $d$ in every direction from the foot of the perpendicular. Worked example 2 quantifies it: a disk of radius $5d$ already holds $80\%$ of the screen, $10d$ holds $90\%$ ; the force error scales roughly with the missing flux. The approximation dies near edges — a charge hovering by a plate's rim has no mirror symmetry to exploit, edge fields kick in, and you are back to numerical methods. Rule of thumb: center of a plate wider than $\sim10d$ , trust the image; within $\sim2d$ of an edge, don't.

Where does the induced charge physically come from, and how fast does it arrive?

Through the ground wire, from the Earth (or whatever reservoir "ground" is). The conductor's own mobile electrons rearrange first, and the wire makes up the net deficit. Timescale: the charge-relaxation time of the shielding article, $\tau = \varepsilon_0/\sigma_c \sim 10^{-19}\,\mathrm{s}$ for metals — so for any mechanical motion of the charge (even GHz vibrations), the screen follows adiabatically and the electrostatic formulas hold instant by instant. That adiabatic tracking is what lets AFM interpret its measured force–distance curves with static image formulas.

Is there any regime where the image force on a single electron is physically wrong?

Two, both instructive. Very close (sub-nanometer): the classical $1/z$ image potential must round off — the electron's wavefunction overlaps the surface, exchange and correlation take over, and the true surface barrier saturates instead of diverging; surface physics splices the classical tail onto a quantum core. Very fast: if the charge moves so quickly that the screen can't follow (approaching the metal's plasma frequency response), the image weakens and lags — "image potential dynamics." In between — from AFM tips to Schottky emission — the nineteenth-century formula is quantitatively excellent, which is itself remarkable: it even binds electrons into the image-potential surface states mentioned in the method article.

Can I get the force by differentiating the energy, and do the factors of two work out?

Yes, and it is the best consistency drill in the problem. $W(d) = -q^2/16\pi\varepsilon_0 d$ , so $F = -\mathrm{d}W/\mathrm{d}d = -q^2/16\pi\varepsilon_0 d^2$ — i.e. magnitude $q^2/16\pi\varepsilon_0d^2$ pulling toward decreasing $d$ : exactly the full image force, no residual factor. It works because $W(d)$ is the true assembly energy at every height (the half is already inside it), and as the charge moves, the image's motion is automatically encoded in the $d$ -dependence. Differentiate the naive $-q^2/8\pi\varepsilon_0d$ instead and you get twice the real force — the two errors advertise each other.

What changes for a charge between two grounded parallel planes?

The hall of mirrors from the method article: each plane images the charge, then images the other plane's images, generating an infinite ladder of alternating charges at spacings set by the gap. The series for force and potential converge (alternating, $\sim1/n^2$ for the force), and near either plane the single-image result dominates. This is the standard model for an ion in a parallel-plate trap or a charged particle in a capacitor gap — and a clean example of a problem where "one image" fails but the image method still beats separation of variables on speed.