Laplace's Equation

Set the source term of Poisson's equation to zero and what remains looks almost too bare to study: $\nabla^2 V = 0$ . No charge, no scale, no parameters. Yet this — Laplace's equation — is arguably the most consequential PDE in classical physics, because most of space is empty of charge. Between the plates, outside the wire, in the vacuum gap of every instrument, the potential obeys Laplace's equation, shaped entirely by sources it never touches. What makes the equation worth an article is the astonishing rigidity of its solutions. A function with $\nabla^2V=0$ — a harmonic function — cannot do as it pleases. Its value at any point is exactly the average of its values on any surrounding sphere. It can have no peaks and no pits in the interior: every extreme lives on the boundary. It is infinitely smooth even if the boundary data is rough. And — the property that powers everything in the next chapter — once its boundary values are pinned, the whole interior is unique. Guess a solution by any trick whatsoever, verify it fits the boundary, and it is the solution. Two famous consequences fall out almost for free. Earnshaw's theorem: no arrangement of static charges can trap a charged particle stably — the potential simply has no interior minimum to sit in, which is why ion traps must cheat with oscillating fields. And the relaxation method: since every interior point is the average of its neighbours, a computer can solve electrostatics by literally averaging a grid until it stops changing. One equation, no parameters — and it dictates the texture of every field map you will ever draw.

Key takeaways

Wherever the local charge density vanishes, Poisson's equation reduces to Laplace's equation $\nabla^2V=0$ ; its solutions are called harmonic functions. Charge-free does not mean field-free — the sources sit on the boundary.
Mean-value property: $V$ at any point equals its average over any sphere centred there (in 1D: straight lines, the average of the endpoints; in 2D: circles).
Consequently $V$ has no local maxima or minima in the interior — extremes live on the boundary — and Earnshaw's theorem follows: no stable electrostatic trap for a charged particle exists in free space.
Uniqueness: two solutions of Laplace's equation with the same boundary values are identical. Boundary data determines the interior — the license behind images, guessing, and every E2 method.
Harmonic functions are the smoothest interpolants of their boundary data: they minimize the stored field energy and are analytic in the interior.
On a grid, "each point equals the average of its neighbours" becomes the relaxation algorithm — electrostatics solved by repeated averaging.
Key idea: Laplace's equation $\nabla^2V=0$ governs charge-free regions; its harmonic solutions obey the mean-value property, hence have no interior extrema, and are uniquely fixed by their boundary data.

The core insight

Laplace's equation $\nabla^2 V = 0$ describes charge-free regions; its harmonic solutions obey the mean-value property, hence have no interior extrema, and are uniquely fixed by their boundary data.

Where this lives today

Every time a chip designer, an accelerator physicist, or an antenna engineer presses "solve" on a field simulation with no space charge, a Laplace solver runs — and its inner loop is this article's mean-value property, discretized: set each grid node to the average of its neighbours, sweep until converged. Modern multigrid refinements solve billion-node electrode geometries in seconds, but the update rule a 2026 GPU executes is recognizably the theorem Gauss would have written on a blackboard. The uniqueness theorem is what makes the whole enterprise trustworthy: whatever iteration path the solver takes, there is exactly one answer for it to find. Earnshaw's theorem, meanwhile, quietly organizes a surprising amount of modern hardware. Trapped-ion quantum computers hold their qubits in Paul traps precisely because a static hold is illegal — the radio-frequency saddle rotation is Earnshaw-evasion, engineered. Penning traps at antimatter facilities add magnetic fields for the same reason, storing antiprotons for months in a vacuum the theorem forbids them from sitting still in electrostatically. And the frog that famously levitated in a Nijmegen magnet floats through the diamagnetic loophole — induced moments, not fixed charges. One bare equation with no parameters, and it dictates both how we compute fields and how we are permitted to hold matter.

The physical picture

Picture the steady temperature in a metal block with no heater inside. Heat flows from hot to cold until nothing changes anymore, and in that final state no interior point can be hotter than all its surroundings — any hot spot would still be shedding heat, contradicting steadiness. Every point settles at the average of its neighbours, and the temperature field becomes the smoothest possible interpolation of whatever temperatures are held on the boundary. That is Laplace's equation in the flesh; swap temperature for potential and "no heater" for "no charge" and the picture transfers exactly. The E0 chapter's Second Derivatives and the Laplacian built the local version of this intuition — the Laplacian as an averaging meter, reading how a function's value compares to the average of its immediate neighbours. This article is what happens when the meter reads zero everywhere: the local statement globalizes into exact theorems about spheres of any size, forbidden extrema, and uniqueness.

Figure

Harmonic terrain — averages over spheres, saddles allowed, bowls forbidden

The saddle in the picture carries the deepest message. Because the Laplacian is a sum of the three direction-curvatures, $\nabla^2V=0$ says the curvatures must cancel: if $V$ curves up along $x$ , it must curve down along $y$ or $z$ . A bowl — curving up in all three directions, the shape a stable trap requires — would need all three curvatures positive, which no harmonic function can supply. Equilibrium points exist (the field can vanish), but they are always saddles or flats, never bowls. That single geometric fact is Earnshaw's theorem, and it is why the ion-trap builders of the modern era had to be clever.

The equation and its rigid solutions

The starting point is inherited directly from the previous article: in any region where $\rho=0$ ,

\nabla^2 V = 0 \qquad (1)

where: $\nabla^2$ — the Laplacian, $\partial_x^2+\partial_y^2+\partial_z^2$ in Cartesian coordinates (radial-angular forms in the E0 coordinate articles); $V$ — the potential, in volts, sourced by charges that live outside the region or on its boundary. Solutions of (1) are called harmonic functions, and the name should carry a warning label: harmonic functions form a far more exclusive club than "functions with some second derivatives that cancel" suggests. Membership imposes global structure. The club's charter is the mean-value property — the first derivation below proves it, and everything else in the article descends from it:

V(\mathbf{r}) = \frac{1}{4\pi R^2}\oint_{\text{sphere }R} V\,\mathrm{d}a \qquad (2)

where: the sphere of radius $R$ is centred at $\mathbf{r}$ and must lie entirely within the charge-free region; $\mathrm{d}a$ — its area element. Read the claim slowly, because it is outrageous: the potential at a point equals its exact average over a surrounding sphere of any radius — not approximately, not for small spheres, but exactly and for every admissible $R$ . The one-dimensional shadow of the theorem is familiar: $\mathrm{d}^2V/\mathrm{d}x^2=0$ forces $V=ax+b$ , a straight line, whose value at any point is the average of its values at equal distances either side. In two dimensions the average runs over circles. Dimension changes the geometry, never the principle. Two corollaries fall out immediately. First, no interior extrema: a strict maximum would exceed the average over a small surrounding sphere, contradicting (2), so maxima and minima are exiled to the boundary. Second — proved as the third derivation — uniqueness:

\nabla^2V_1=\nabla^2V_2=0 \ \text{in}\ \Omega, \quad V_1=V_2 \ \text{on the boundary} \ \Longrightarrow\ V_1=V_2 \ \text{everywhere in}\ \Omega \qquad (3)

where: $\Omega$ — the charge-free region. This is the first uniqueness theorem, and it converts electrostatics from an integration problem into a guessing license: any function, found by any method — mirror-image charges, series, inspired doodling — that is harmonic inside and matches the boundary data is the answer, with no further justification owed. The entire E2 chapter (the boundary-value twin of this article, Method of Images, Multipole Expansion of the Potential) is an organized collection of such guessing schemes; the machinery of Neumann conditions and Green identities is stated and wielded there. One more structural fact deserves its equation. Among all functions taking the prescribed boundary values, the harmonic one minimizes the stored field energy,

W = \frac{\varepsilon_0}{2}\int_\Omega |\nabla V|^2\,\mathrm{d}\tau \qquad (4)

where: $W$ — the energy the field holds in the region (the energy-density story is told properly in Energy Density of the Electric Field, two articles ahead). Nature's static configuration is the laziest one compatible with the boundary — harmonic functions are what "relaxed" means, and the numerical method named relaxation is this variational fact executed on a grid.

Which fact answers which question

you want	use	notes
is this function a legal vacuum potential?	check $\nabla^2V=0$	curvatures must cancel direction-by-direction
value at a point from surroundings	mean value (2)	any admissible sphere; also the grid update rule
can a trap/extremum exist here?	no-extremum corollary	interior of charge-free region: never
is my guessed solution THE solution?	uniqueness (3)	harmonic + correct boundary values = done
solve with given electrode potentials	E2 machinery	separation of variables, images, Green functions
quick sanity on any field map	equipotentials never close in empty space	a closed equipotential in vacuum would trap an extremum

The mean-value property even comes with a built-in algorithm. Chop the region into a grid, hold the boundary nodes at their prescribed potentials, and repeatedly replace every interior node by the average of its nearest neighbours; the grid converges to the discrete solution of Laplace's equation. This relaxation method — previewed here, industrialized in the today-section — is worth seeing before the formal proofs, because it is the mean-value theorem doing manual labor:

Figure

Relaxation — the mean-value property as an algorithm on a grid

Derivation AThe mean-value property from the divergence theorem

The goal is equation (2): a harmonic function's value at a point equals its average over any sphere around that point. The strategy is elegant and short — write the average over a sphere of radius $R$ as a function $\bar V(R)$ , differentiate with respect to the radius, and show the derivative vanishes; the average then cannot depend on $R$ , and shrinking $R\to0$ identifies its constant value as $V$ at the centre. The move to watch is the exchange of the $R$ -derivative for a flux integral, which is where $\nabla^2V=0$ gets spent.

Step 1. Centre coordinates on the point and write the spherical average over solid angle:

\bar V(R) = \frac{1}{4\pi R^2}\oint V\,\mathrm{d}a = \frac{1}{4\pi}\int V(R,\theta,\phi)\,\mathrm{d}\Omega

Why solid angle: dividing out the $R^2$ of the area element makes the integration domain the fixed unit sphere, so $R$ survives only inside $V$ 's argument — differentiating under the integral becomes clean.

Step 2. Differentiate with respect to $R$ . The only $R$ -dependence is in $V$ 's first slot, and $\partial V/\partial R$ on the sphere is the radial component of the gradient, $\hat{\mathbf{r}}\cdot\nabla V$ :

\frac{\mathrm{d}\bar V}{\mathrm{d}R} = \frac{1}{4\pi}\int \hat{\mathbf{r}}\cdot\nabla V\,\mathrm{d}\Omega = \frac{1}{4\pi R^2}\oint \nabla V\cdot\mathrm{d}\mathbf{a}

Step 3. Convert the flux to a volume integral by the divergence theorem, and spend the hypothesis:

\oint \nabla V\cdot\mathrm{d}\mathbf{a} = \int_{\text{ball}} \nabla\cdot(\nabla V)\,\mathrm{d}\tau = \int_{\text{ball}} \nabla^2 V\,\mathrm{d}\tau = 0

Why this is Gauss's law wearing a math gown: the flux of $\nabla V$ (which is $-\mathbf{E}$ ) through the sphere counts the charge inside, and there is none. The mean-value theorem is the charge-free statement of Gauss's law, integrated over radii.

Step 4. So $\mathrm{d}\bar V/\mathrm{d}R=0$ for every admissible $R$ : the average is one number for all spheres. As $R\to0$ the sphere closes on the centre and, $V$ being continuous, $\bar V(R)\to V(\text{centre})$ . Hence $\bar V(R)=V(\text{centre})$ for all $R$ — equation (2). Verified: reproduces Griffiths §3.1.4; direct check on the harmonic $V=\kappa(x^2+y^2-2z^2)$ of worked example 1 — over a sphere of radius $R$ , the averages of $x^2$ , $y^2$ , $z^2$ each equal $R^2/3$ , so $\bar V = \kappa(R^2/3+R^2/3-2R^2/3)=0=V(0)$ , exactly.

Go deeper · Derivation B · 2 minNo interior extrema, and Earnshaw's impossibility theorem

Derivation BNo interior extrema, and Earnshaw's impossibility theorem

The goal is two statements riding one argument: a harmonic function has no strict interior maximum or minimum, and therefore no static arrangement of charges can hold a charged particle in stable equilibrium in vacuum — Earnshaw's theorem, the reason electrostatic levitation is a fantasy and ion traps oscillate. The strategy: derive the extremum ban from the mean-value property by contradiction, then translate "stable trap" into "interior minimum of the potential energy" and find none available. Watch how little is needed — the whole theorem is the previous derivation plus one sentence of mechanics.

Step 1. Suppose $V$ had a strict local maximum at an interior point $P$ : some small sphere around $P$ lies in the region with $V<V(P)$ everywhere on it. Averaging over that sphere gives $\bar V < V(P)$ — but the mean-value property demands $\bar V = V(P)$ . Contradiction; no such maximum exists. The identical argument (flip the inequality) bans minima. Why "strict" matters: a constant $V$ trivially attains its extreme everywhere; the theorem forbids isolated peaks and pits, which is exactly what stability questions ask about.

Step 2. Translate to mechanics. A charge $q$ at rest at $P$ is in stable equilibrium only if its potential energy $U(\mathbf{r})=qV(\mathbf{r})$ has a local minimum at $P$ : displaced in any direction, it must be pushed back. In the charge-free neighbourhood a trap requires, $V$ is harmonic — and $U$ , being $V$ scaled by the constant $q$ , is harmonic too (for either sign of $q$ ).

Step 3. Apply Step 1 to $U$ : no interior minimum exists. The best any equilibrium can be is a saddle — restoring along some directions, expelling along at least one. A marble on a mountain pass, never a marble in a bowl.

Step 4. Read the curvature version for intuition: stability in all directions at an equilibrium demands $\partial_x^2U$ , $\partial_y^2U$ , $\partial_z^2U$ all positive, while harmonicity forces their sum to vanish. The three curvatures cannot all be positive; whatever confines in two directions must expel in the third. Why the loopholes are where the modern traps live: oscillate the saddle fast enough and the time-average confines (the Paul trap); add a magnetic field (the Penning trap); use polarizable matter in field gradients — diamagnets levitate legally because induced dipoles evade the theorem's premise of fixed charge in a source-free harmonic potential. *Verified: matches Griffiths §3.1.5 and Purcell §2.12; the quadrupole saddle of worked example 1 exhibits the theorem concretely — equilibrium at the origin, confining in $x$ and $y$ , expelling along $z$ , curvatures $+2\kappa,+2\kappa,-4\kappa$ summing to zero.*

Go deeper · Derivation C · 1 minBoundary data determine the interior — the first uniqueness theorem

Derivation CBoundary data determine the interior — the first uniqueness theorem

The goal is equation (3): two harmonic functions agreeing on the boundary of a region agree everywhere inside — stated here with the proof sketch the no-extremum property makes almost free; the full boundary-value arsenal (Neumann data, Green identities, conductors with floating potentials) is built in the E2 chapter, where it earns its keep. The strategy is the difference trick, the single most reused move in linear PDE theory: subtract the two candidates and show the difference has nowhere to be anything but zero.

Step 1. Let $V_1$ and $V_2$ both be harmonic in $\Omega$ with $V_1=V_2$ on the boundary, and set $W=V_1-V_2$ . Linearity of the Laplacian makes $W$ harmonic ( $\nabla^2W=\nabla^2V_1-\nabla^2V_2=0$ ), and by construction $W=0$ everywhere on the boundary. Why the difference trick works only because the equation is linear: for a nonlinear equation the difference of two solutions solves nothing in particular, and uniqueness becomes genuinely hard.

Step 2. Apply the extremum ban to $W$ . Its maximum over the closed region is attained on the boundary — where $W=0$ . So $W\le0$ inside. Its minimum is also attained on the boundary — also $0$ . So $W\ge0$ inside.

Step 3. Squeezed: $W\equiv0$ , that is, $V_1=V_2$ throughout $\Omega$ . Boundary data pin the interior completely.

Step 4. Register what has been purchased. Any harmonic function matching the boundary values is the potential — however disreputable the method that produced it. This is the legal foundation for the method of images (replace a conductor by a fictitious mirror charge; if the boundary values come out right, the fiction is exact in the physical region) and for series expansions fitted to electrodes. Uniqueness converts cleverness into rigor. Verified: reproduces Griffiths §3.1.5 (first uniqueness theorem); the Green-identity route to the same result, which extends to Neumann conditions and mixed boundaries, is developed in the E2 boundary-value chapter — one theorem, two proofs, deliberately split across the two articles.

Limiting cases

case	result	meaning
charge switched on	$\nabla^2V=-\rho/\varepsilon_0$	back to Poisson's equation; Laplace is its $\rho=0$ specialization
one dimension	$V=ax+b$	straight lines: value = average of neighbours, the mean-value property naked
spherical symmetry	$V=A/r+B$	the unique radial harmonic pair; with $V(\infty)=0$ , the Coulomb form
two dimensions	$V=\mathrm{Re}\,f(z)$ , $f$ analytic	harmonic = real parts of complex-analytic functions; conformal mapping solves electrostatics
cylindrical symmetry (2D radial)	$V=A\ln s+B$	the line-charge exterior potential; $\ln$ replaces $1/r$ in two dimensions
constant boundary values	$V=\text{const}$ , $\mathbf{E}=0$ inside	uniqueness with the trivial guess: the shielded cavity of Conductors in Electrostatic Equilibrium

The last row is the punchline of the whole theory delivered early: a closed conductor holds its surface at one potential, the constant function matches that boundary data and is harmonic, so — by uniqueness — the cavity field is exactly zero, no matter what storm rages outside. Faraday's cage is a two-line corollary of this article.

Common trapNo charge here means no field here — the most tempting false theorem in electrostatics

Laplace's equation says the local charge is zero; it says nothing about the field, which is imposed by sources on and beyond the boundary. The vacuum gap of a 10 kV capacitor is charge-free and carries a ferocious field; the region obeys Laplace's equation everywhere. The equation constrains the field's texture, not its size: no bumps, extremes on the boundary, averages exact. A related seduction: "the field is zero at an equilibrium point, so a charge placed there is trapped." Zero force is not stability — Earnshaw guarantees the equilibrium is a saddle, and the particle escapes along the expelling direction at the first nudge. When you catch yourself drawing a closed equipotential bubble in empty space, stop: it would enclose an extremum, and there are none to enclose.

Worked examples

1 · Certifying a quadrupole saddle (confidence builder)

A region near the origin is charge-free, and the potential is modelled as $V(x,y,z)=\kappa\,(x^2+y^2-2z^2)$ with $\kappa=100\ \mathrm{V/m^2}$ . (a) Verify this is a legal vacuum potential. (b) Find the field at $P=(0.10,\,0.10,\,0.10)$ m. (c) Is the origin a trap for a positive ion?

Step 1. Legality is the Laplacian: $\partial_x^2V=2\kappa$ , $\partial_y^2V=2\kappa$ , $\partial_z^2V=-4\kappa$ ; sum $=2\kappa+2\kappa-4\kappa=0$ ✓. Harmonic — the curvatures cancel exactly, the up-bending in $x$ and $y$ paid for by twice the down-bending in $z$ . (This is the electrostatic quadrupole potential, the workhorse geometry of real ion optics.)

Step 2. The field is minus the gradient: $\mathbf{E}=-\kappa\,(2x,\ 2y,\ -4z)$ . At $P$ : $\mathbf{E}=-(100)(0.20,\ 0.20,\ -0.40)=(-20,\ -20,\ +40)$ V/m, magnitude

|\mathbf{E}| = \sqrt{20^2+20^2+40^2} = \sqrt{2400} = \boxed{49\ \mathrm{V/m}}

Step 3. Trap audit at the origin. $\mathbf{E}(0)=0$ — an equilibrium. For a positive ion, $U=qV$ curves up along $x$ and $y$ (restoring) but down along $z$ (expelling): a saddle, exactly as Earnshaw demands. Released off-axis in $z$ , the ion accelerates away. Flipping the sign of the charge only relabels the trapdoor: a negative ion is confined along $z$ and expelled radially.

Step 4. Bonus check of the mean-value property: over a sphere of radius $R$ about the origin, $\langle x^2\rangle=\langle y^2\rangle=\langle z^2\rangle=R^2/3$ , so $\langle V\rangle=\kappa(R^2/3+R^2/3-2R^2/3)=0=V(0)$ ✓ — the average is exact, for every radius. (The Paul trap rescues this saddle by spinning it at radio frequency; the time-averaged pseudo-potential then does confine — dodging Earnshaw rather than defeating him.)

2 · The spherical capacitor by pure guessing (exam level)

A conducting sphere of radius $a=5.0$ cm held at $100$ V sits concentric inside a grounded shell of radius $b=20$ cm. Find $V(r)$ in the gap, the field at the inner surface, and the charge on the inner sphere — without a single superposition integral.

Step 1. The gap is charge-free and spherically symmetric, so try the radial harmonic catalogue: $V(r)=A/r+B$ (the unique spherically symmetric solutions of Laplace's equation — integrate $\left(r^2V'\right)'=0$ twice). Uniqueness says: fit the boundary values and we are done, no further justification owed.

Step 2. Fit. $V(a)=A/a+B=100$ and $V(b)=A/b+B=0$ . Subtracting: $A\,(1/a-1/b)=100$ , with $1/a=20\ \mathrm{m^{-1}}$ and $1/b=5\ \mathrm{m^{-1}}$ , so $A=100/15=6.67\ \mathrm{V\,m}$ and $B=-A/b=-33.3$ V:

V(r) = \frac{6.67}{r} - 33.3 \ \text{volts} \quad (a\le r\le b)

Spot-check: $V(0.05)=133.3-33.3=100$ V ✓, $V(0.20)=33.3-33.3=0$ ✓.

Step 3. Field and surface charge. $E_r=-\mathrm{d}V/\mathrm{d}r=A/r^2$ ; at the inner surface $E(a)=6.67/(0.05)^2=\boxed{2.7\ \mathrm{kV/m}}$ , pointing outward. The conductor-surface rule (Conductors in Electrostatic Equilibrium) gives $\sigma=\varepsilon_0E(a)=2.36\times10^{-8}\ \mathrm{C/m^2}$ , so

q = \sigma\cdot4\pi a^2 = 4\pi\varepsilon_0 A = (1.113\times10^{-10})(6.67) = \boxed{0.74\ \mathrm{nC}}

Step 4. Close the loop with a known formula: the capacitance follows as $C=q/\Delta V=7.4\times10^{-10}/100$ ... careful — $q=7.42\times10^{-10}$ C at $\Delta V=100$ V gives $C=7.4$ pF, and the textbook spherical-capacitor formula $C=4\pi\varepsilon_0\,ab/(b-a)=1.113\times10^{-10}\times(0.01/0.15)=7.4$ pF ✓. A boundary-value guess, two constants, and the whole electrostatics of the device follows. That economy — impossible by direct integration over unknown surface charges — is what uniqueness sells.

Origins · 1 minPierre-Simon Laplace and how this was discovered

Pierre-Simon Laplace

1749–1827 · Paris, the Académie and the École

Laplace arrived in Paris a Norman farmer's son with a letter of introduction and left French science its most imposing monument: the five-volume Mécanique Céleste, which set out to deduce the solar system's entire behaviour from gravitation alone. The equation bearing his name appeared in that program — studying the attraction of spheroids, he showed the gravitational potential in empty space satisfies $\nabla^2V=0$ , turning force calculations into potential theory. His confidence in determinism was total ("an intelligence knowing all forces and positions... nothing would be uncertain"), and his political instincts matched: he served the Revolution, Napoleon (briefly, badly, as interior minister — Napoleon complained he brought "the spirit of the infinitesimally small" into administration), and the restored Bourbons, collecting a marquisate on the way. When asked why God did not appear in the Mécanique Céleste, he is said to have replied, "I had no need of that hypothesis." Poisson, his student, would shortly show the hypothesis his equation needed was only charge: inside matter, $\nabla^2V$ stops vanishing.

History

1785

— Laplace, computing the attraction of spheroids, introduces the potential equation

\nabla^2V=0

for empty space.

1813

— Poisson supplies the source term, marking off Laplace's equation as the charge-free case.

1828

— Green's Essay proves boundary data determine the interior, founding the uniqueness industry.

1842

— Earnshaw shows no static field can stably trap a charged particle — the theorem that later forces ion traps to oscillate.

1946

— Southwell's relaxation methods turn the mean-value property into an industrial algorithm, ancestor of today's multigrid solvers.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §3.1–3.4 · Jackson, Classical Electrodynamics, 3rd ed., Ch. 1–3 · Purcell & Morin, Electricity and Magnetism, 3rd ed., §2.12 · Zangwill, Modern Electrodynamics, Ch. 7

Provenance. Equations and theorems follow Griffiths (Introduction to Electrodynamics, 4th ed., Ch. 3) and Jackson (Classical Electrodynamics, 3rd ed., Ch. 1–3). Earnshaw's theorem and the extremum principle are textbook-standard (Griffiths §3.1.4–3.1.5, Purcell §2.12). A verifier should: (1) rerun the mean-value derivation and its example-1 check $\langle x^2\rangle=R^2/3$ giving $\bar V=0=V(0)$ ; (2) confirm the quadrupole numbers: curvatures $(+200,+200,-400)\ \mathrm{V/m^2}$ summing to zero, $\mathbf{E}(P)=(-20,-20,+40)$ V/m, $|\mathbf{E}|=\sqrt{2400}=49.0$ V/m; (3) recompute the spherical capacitor: $A=100/(20-5)=6.67$ V·m, $B=-33.3$ V, $E(a)=2.67$ kV/m, $q=4\pi\varepsilon_0A=0.742$ nC, against $C=4\pi\varepsilon_0ab/(b-a)=7.4$ pF; (4) verify the uniqueness sketch (difference harmonic, zero on boundary, squeezed by the extremum ban). The Green-identity proof and Neumann-condition extensions are deliberately deferred to the E2 boundary-value chapter.

Builds on: Gauss's Law in Differential Form · The Electrostatic Potential · The Curl of the Electrostatic Field · Potential of a Localized Charge Distribution Leads to: Method of Images · Multipole Expansion of the Potential · Boundary-Value Problems in Electrostatics · Applications of Gauss's Law and Symmetry

If there is no charge in the region, why is there any field at all?

Because the sources live outside the region or on its boundary, and their influence reaches in. Laplace's equation is a statement about the local charge — zero — not about the field, which can be enormous: the vacuum between capacitor plates at 10 kV is charge-free and obeys $\nabla^2V=0$ while carrying megavolt-per-metre fields. What the equation does constrain is the field's texture: the potential in the region is the smoothest interpolation of its boundary values, with no interior bumps. Think of the region as a drumhead whose rim is held at prescribed heights — the membrane inside can tilt and warp dramatically, but every feature it has is dictated by the rim.

How can the value at a point equal the average over a big sphere and a small one at the same time?

Because the averages over all admissible spheres are forced to agree with each other — that is exactly what the derivation shows: the radius-derivative of the spherical average vanishes, so $\bar V(R)$ is one constant number, and continuity identifies that constant as the central value. It feels miraculous because generic functions certainly do not behave this way; harmonicity is a powerful global constraint dressed as a local one. The 1D case demystifies it: for a straight line, the average of the endpoint values is the midpoint value whether the endpoints are 1 mm or 1 km apart. Spheres in 3D play the role of endpoint pairs in 1D.

Earnshaw forbids electrostatic traps — so how do ion traps, magnetic levitation, and optical tweezers work?

Each one breaks a premise of the theorem rather than the theorem itself. The Paul trap breaks static: it spins a quadrupole saddle at radio frequency, and the time-averaged dynamics of a particle in an oscillating saddle is confining (a pseudo-potential minimum) even though every instantaneous potential is saddle-shaped. The Penning trap breaks purely electrostatic by adding a magnetic field, which bends the escape trajectories into bounded orbits. Diamagnetic levitation (the famous frog) and optical tweezers break fixed charge: they act on induced moments, whose energy is not a harmonic function of position — for diamagnets, field-magnitude minima are allowed in free space and stable levitation is legal. Earnshaw is not a dead letter; it is the design constraint that explains why every one of these machines looks the way it does.

Is a conductor's interior field being zero the same theorem as the shielded cavity?

They are two applications of the same uniqueness logic at different places. Inside the conducting material, mobile charges rearrange until the field vanishes — that is dynamics, told in Conductors in Electrostatic Equilibrium. Inside an empty cavity enclosed by the conductor, there are no mobile charges to do any rearranging; what protects the cavity is this article's mathematics. The cavity walls are one connected conductor at a single potential $V_0$ ; the constant function $V=V_0$ is harmonic and matches the boundary; uniqueness says that is the cavity potential, so $\mathbf{E}=0$ inside, whatever fields rage outside. Faraday cages work by theorem, not by absorption.

Do I ever need the energy-minimization picture, or is it a curiosity?

It earns its keep three ways. Conceptually it explains why fields look relaxed — no wiggles that boundary data do not force, since any extra wiggle costs $\int|\nabla V|^2$ energy. Numerically it underwrites a whole solver family: minimizing the Dirichlet energy over trial functions is the finite-element method's starting point, and "relaxation" converges because each averaging sweep strictly lowers the energy toward the harmonic minimum. And theoretically it gives an independent route to uniqueness (two minimizers with the same boundary data would tie, forcing their difference's gradient energy to vanish). When E2 introduces variational estimates for capacitance, this is the principle doing the lifting.

What is special about two dimensions and complex numbers?

In 2D, a function is harmonic exactly when it is locally the real part of an analytic function of $z=x+\mathrm{i}y$ — differentiate the Cauchy–Riemann equations once and the Laplacian falls out. This marriage is spectacularly productive: any conformal (angle-preserving) map of the plane drags harmonic functions to harmonic functions, so a hard electrode geometry can be mapped to an easy one (a wedge to a half-plane, two cylinders to two planes), solved there, and mapped back. Fringing fields at capacitor edges and transmission-line impedances were computed this way for a century before numerics took over, and the method still delivers exact closed forms no grid can match. Three dimensions has no comparable structure — one of several ways 2D electrostatics is secretly complex analysis.

Where does this article stop and the E2 boundary-value chapter begin?

This article establishes what harmonic functions are — mean value, no extrema, uniqueness — plus the two-constant catalogues ( $A/r+B$ , $A\ln s+B$ ) that symmetry alone can fit. What it deliberately does not do is construct solutions for real electrode geometries: fitting infinite series of separable solutions (Legendre polynomials in spherical problems, sinusoids in Cartesian slots), placing image charges, and building Green functions are E2's business, opened by its own Laplace's-equation article and Method of Images. The division of labor is honest: everything in E2 is a scheme for producing a harmonic function with the right boundary values, and everything here — especially uniqueness — is why producing one is the same as solving the problem.