The Laplace Equation and Harmonic Functions

One-line statement: Laplace's equation

\nabla^{2} u = 0

characterizes harmonic functions — the smoothest, most "balanced" scalar fields, equal at every point to the average of their surroundings. Key result in words: A twice-differentiable function whose Laplacian vanishes has no local maxima or minima in the interior, is infinitely differentiable, and is uniquely determined by its boundary values. Where it applies: Steady-state heat flow, electrostatics, gravitation, incompressible fluid flow, complex analysis, and probability (Brownian motion). Who/when: Pierre-Simon Laplace (1780s, celestial mechanics); foundations laid by Euler, Gauss, Green, Riemann, and Poincaré through the 19th century.

The Laplace equation is the second-order partial differential equation $\nabla^{2} u = 0$ , where $\nabla^{2}$ (the Laplacian) is the sum of unmixed second partial derivatives. A solution $u$ is called a harmonic function. These functions matter because they describe equilibrium: the steady state a diffusing quantity reaches when nothing changes in time, when sources and sinks have balanced out. They sit at the crossroads of physics, geometry, and analysis — the real and imaginary parts of any holomorphic function are harmonic, linking Laplace's equation directly to complex analysis. The surprising fact: harmonic functions are rigid. Knowing a harmonic function on the boundary of a region fixes it everywhere inside, and a harmonic function that is bounded on all of $R^{n}$ must be constant.

The Core Insight

A harmonic function is one with zero net curvature in the averaged sense: at every interior point its value equals the average of its values on any surrounding sphere. Curvature in one direction is exactly cancelled by opposite curvature in the perpendicular directions, so $u$ "bows up" in some directions only by "bowing down" equally in others. This single averaging property — the mean value property — generates almost everything else: smoothness, the maximum principle, uniqueness, and rigidity.

Formal Statement

The Laplacian and Laplace's Equation

For a scalar function $u : Ω \subseteq R^{n} \to R$ that is twice continuously differentiable ( $u \in C^{2}$ ), the Laplacian is

\nabla^{2} u = Δ u = i = 1 \sum n \frac{\partial ^{2} u}{\partial x _{i}^{2}} = \frac{\partial ^{2} u}{\partial x _{1}^{2}} + \dots + \frac{\partial ^{2} u}{\partial x _{n}^{2}} .

⟳The Laplacian is the divergence of the gradient,

Δ u = \nabla \cdot (\nabla u)

; expand

\nabla \cdot (\partial_{x_{1}} u, \dots, \partial_{x_{n}} u)

component-by-component.

Laplace's equation is the homogeneous PDE

Δ u = 0 on Ω.

⟳Set the Laplacian to zero; this is the source-free special case of Poisson's equation

Δ u = f

with

f \equiv 0

A function satisfying [2] on an open set $Ω$ is harmonic on $Ω$ . Here $Ω$ is the open domain, $\partial Ω$ its boundary, and the symbols $x_{i}$ are Cartesian coordinates.

Laplacian in Other Coordinate Systems

In polar coordinates $(r, θ)$ on $R^{2}$ :

Δ u = \frac{\partial ^{2} u}{\partial r ^{2}} + \frac{1}{r} \frac{\partial u}{\partial r} + \frac{1}{r ^{2}} \frac{\partial ^{2} u}{\partial θ ^{2}} .

⟳Apply the chain rule to

x = r cos θ

y = r sin θ

, compute

\partial_{x}, \partial_{y}

in terms of

\partial_{r}, \partial_{θ}

, square and add; cross terms cancel.

In spherical coordinates $(r, θ, φ)$ on $R^{3}$ (with $θ$ the polar angle):

Δ u = \frac{1}{r ^{2}} \frac{\partial}{\partial r} (r^{2} \frac{\partial u}{\partial r}) + \frac{1}{r ^{2} sin θ} \frac{\partial}{\partial θ} (sin θ \frac{\partial u}{\partial θ}) + \frac{1}{r ^{2} sin ^{2} θ} \frac{\partial ^{2} u}{\partial φ ^{2}} .

⟳Use

Δ u = \frac{1}{g} \partial_{i} (g g^{ij} \partial_{j} u)

with the spherical metric

g = diag (1, r^{2}, r^{2} sin^{2} θ)

Mean Value Property

Let $u$ be harmonic on an open set $Ω \subseteq R^{n}$ . For any ball $\overline{B (x_{0}, r)} \subset Ω$ , the value at the center equals the average over the sphere and over the ball:

u (x_{0}) = \frac{1}{∣ \partial B ( x _{0} , r ) ∣} \int_{\partial B (x_{0}, r)} u d S = \frac{1}{∣ B ( x _{0} , r ) ∣} \int_{B (x_{0}, r)} u d V .

⟳Define

ϕ (r)

as the spherical average; differentiate under the integral, rescale to the unit sphere, and apply the divergence theorem to

\int_{B} Δ u d V = 0

to show

ϕ^{'} (r) = 0

Conversely, any continuous function satisfying the mean value property on every ball in $Ω$ is harmonic.

Maximum Principle

If $u$ is harmonic on a bounded open set $Ω$ and continuous on $\overline{Ω}$ , then $u$ attains both its maximum and minimum on the boundary:

\overline{Ω} max u = \partial Ω max u, \overline{Ω} min u = \partial Ω min u .

⟳If

u

attained an interior maximum

M

x_{0}

, the mean value property [5] forces

u \equiv M

on a surrounding sphere; a connectedness argument propagates

u \equiv M

throughout

Ω

The strong maximum principle states that if a non-constant harmonic function on a connected domain attains an interior maximum, that is impossible — interior extrema occur only for constants.

Uniqueness for the Dirichlet Problem

The boundary value problem $Δ u = 0$ in $Ω$ , $u = g$ on $\partial Ω$ (with $g$ continuous, $Ω$ bounded) has at most one solution.

⟳If

u_{1}, u_{2}

both solve it, then

w = u_{1} - u_{2}

is harmonic with

w = 0

\partial Ω

; the maximum principle [6] forces

max w = min w = 0

, so

w \equiv 0

Why It Holds / Derivation

Mean Value Property from Laplace's Equation

Step 1. Fix

x_{0} \in Ω

and define the spherical average

ϕ (r) = \frac{1}{∣ \partial B ( x _{0} , r ) ∣} \int_{\partial B (x_{0}, r)} u (y) d S (y) .

Step 2. Change variables to the unit sphere via

y = x_{0} + r z

∣ z ∣ = 1

d S (y) = r^{n - 1} d S (z)

ϕ (r) = \frac{1}{∣ \partial B ( 0 , 1 ) ∣} \int_{∣ z ∣ = 1} u (x_{0} + r z) d S (z) .

Step 3. Differentiate under the integral sign:

ϕ^{'} (r) = \frac{1}{∣ \partial B ( 0 , 1 ) ∣} \int_{∣ z ∣ = 1} \nabla u (x_{0} + r z) \cdot z d S (z) .

Here

z

is the outward unit normal of the unit sphere, so

\nabla u \cdot z = \partial u / \partial ν

is the normal derivative. Step 4. Convert back to the sphere of radius

r

and apply the divergence theorem:

ϕ^{'} (r) = \frac{1}{∣ \partial B ( x _{0} , r ) ∣} \int_{\partial B (x_{0}, r)} \frac{\partial u}{\partial ν} d S = \frac{1}{∣ \partial B ( x _{0} , r ) ∣} \int_{B (x_{0}, r)} Δ u d V .

Step 5. Since

u

is harmonic,

Δ u = 0

, hence

ϕ^{'} (r) = 0

. Thus

ϕ

is constant in

r

. Step 6. As

r \to 0^{+}

, continuity gives

ϕ (r) \to u (x_{0})

. Therefore

ϕ (r) = u (x_{0})

for all admissible

r

, proving the spherical mean value property [5]. Integrating the spherical version over

r

yields the solid (ball) version.

■

Smoothness of Harmonic Functions

Step 1. Let

u

satisfy the mean value property. Choose a smooth radial mollifier

η_{ε}

supported in

B (0, ε)

with

\int η_{ε} = 1

and depending only on

∣ x ∣

. Step 2. Form the convolution

u_{ε} = u * η_{ε}

, which is automatically

C^{\infty}

wherever defined. Step 3. Because

η_{ε}

is radial and

u

satisfies the mean value property over every sphere, the weighted average

u_{ε} (x)

collapses to

u (x)

itself:

u_{ε} = u

on the interior at distance

> ε

from

\partial Ω

. Step 4. Hence

u

agrees locally with a

C^{\infty}

function, so

u \in C^{\infty} (Ω)

. In fact

u

is real-analytic.

■

Pierre-Simon Laplace | 1749–1827

Laplace introduced the operator now bearing his name in his monumental Traité de mécanique céleste, where he studied the gravitational potential of celestial bodies. He showed that outside any mass distribution the potential $V$ satisfies $Δ V = 0$ , reducing the calculation of gravitational fields to solving a single elegant equation. Famously, when Napoleon asked why his treatise made no mention of God, Laplace is said to have replied, "I had no need of that hypothesis." His work on potential theory, probability, and celestial mechanics shaped the analytic foundations of mathematical physics for the next century.

Building Intuition

Imagine a thin metal plate whose edge is held at fixed temperatures. After a long time the interior settles into a steady state: heat no longer flows on net, and the temperature distribution is harmonic. At any interior point the temperature equals the average of its neighbors — if it were higher, heat would flow out and cool it; if lower, heat would flow in and warm it. Equilibrium is the mean value property.

This also explains why harmonic functions have no interior bumps. A local hot spot would be hotter than its surrounding average, violating [5]. So the hottest and coldest points always sit on the boundary. A taut soap film stretched across a wire loop is the same idea: with no pressure difference, the film's height is harmonic and the surface has no interior peaks or pits — it is a saddle everywhere or flat.

A harmonic saddle: curvature in $x$ cancels curvature in $y$

Worked Examples

Steady Temperature on a Disk

Given: A circular plate of radius

R = 2

m. The boundary temperature is held at

g (θ) = 100 cos θ

°C. Find the steady-state temperature

u (r, θ)

inside, and its value at the center and at

(r, θ) = (1, 0)

. Step 1 — General harmonic solution (Poisson/Fourier on a disk). Bounded harmonic functions on a disk have the form

u (r, θ) = a_{0} + n = 1 \sum \infty r^{n} (a_{n} cos n θ + b_{n} sin n θ),

because each

r^{n} cos n θ

and

r^{n} sin n θ

satisfies [3]. Step 2 — Match boundary data. At

r = R = 2

we need

u (2, θ) = 100 cos θ

. Only the

n = 1

cosine term survives:

a_{1} R^{1} cos θ = 100 cos θ \Rightarrow a_{1} \cdot 2 = 100 \Rightarrow a_{1} = 50.

All other coefficients vanish. Step 3 — Write the solution.

u (r, θ) = 50 r cos θ = 50 x .

Step 4 — Evaluate. Center

(r = 0)

u = 0

°C — consistent with the mean value property, since

\frac{1}{2 π} \int_{0}^{2 π} 100 cos θ d θ = 0

. At

(r, θ) = (1, 0)

u = 50 (1) cos 0 = 50

u (r, θ) = 50 r cos θ °C, u (center) = 0 °C, u (1, 0) = 50 °C

Verifying and Building Harmonic Functions

Given: Test whether

u (x, y) = e^{x} cos y

is harmonic, and find its harmonic conjugate

v

so that

f = u + i v

is holomorphic. Step 1 — Compute second derivatives.

u_{xx} = e^{x} cos y, u_{y y} = - e^{x} cos y .

Step 2 — Sum them.

Δ u = u_{xx} + u_{y y} = e^{x} cos y - e^{x} cos y = 0. ✓ Harmonic.

Step 3 — Find conjugate via Cauchy–Riemann. Require

v_{y} = u_{x} = e^{x} cos y

and

v_{x} = - u_{y} = e^{x} sin y

. Integrate the first:

v = e^{x} sin y + h (x)

. Differentiate:

v_{x} = e^{x} sin y + h^{'} (x)

. Compare with

v_{x} = e^{x} sin y

, so

h^{'} (x) = 0

h = C

. Step 4 — Identify the holomorphic function.

f (z) = e^{x} cos y + i e^{x} sin y = e^{x} (cos y + i sin y) = e^{x + i y} = e^{z} .

u = e^{x} cos y is harmonic; v = e^{x} sin y + C; f (z) = e^{z}

Common Misconceptions

" $Δ u = 0$ just means $u$ is a constant or a plane." False. Harmonic functions form an infinite-dimensional space:

x^{2} - y^{2}

x y

e^{x} cos y

ln r

arctan (y / x)

, and infinitely many spherical harmonics are all non-trivial harmonic functions. What is true is that a harmonic function cannot have interior strict extrema. "Harmonic implies bounded or implies decaying." Not on unbounded domains. On

R^{2}

u = x

is harmonic and unbounded. Liouville's theorem only forbids functions that are both harmonic and bounded on all of

R^{n}

— those must be constant. "The Laplacian is just the second derivative." In one dimension yes, but in higher dimensions it is a sum of second derivatives. The cancellation between directions (positive in one, negative in another) is precisely what makes the harmonic constraint non-trivial.

Harmonic in 2D but Not in 3D

The function $u (x, y) = ln x^{2} + y^{2} = ln r$ is harmonic on $R^{2} ∖ {0}$ :

Δ (ln r) = \frac{d ^{2}}{d r ^{2}} ln r + \frac{1}{r} \frac{d}{d r} ln r = - \frac{1}{r ^{2}} + \frac{1}{r} \cdot \frac{1}{r} = 0.

But the same logarithm of the radial coordinate is not harmonic in 3D — there the fundamental solution is

1/ r

instead:

Δ (\frac{1}{r}) = 0 on R^{3} ∖ {0}, whereas Δ (ln r) = \frac{1}{r ^{2}} \neq = 0 in 3D .

Lesson: The form of the fundamental solution depends critically on dimension —

ln r

in 2D,

r^{2 - n}

in dimension

n \geq 3

. Naively reusing a 2D harmonic function in 3D fails.

Connections

Laplace's equation is the hub of an enormous web of mathematics:

Complex Analysis: The real and imaginary parts of every holomorphic function are harmonic, linked by the Cauchy-Riemann Equations. Conformal maps preserve harmonicity, giving a powerful method to solve boundary value problems.
Poisson's Equation: The inhomogeneous version $Δ u = f$ models fields with sources; Laplace's equation is the source-free case $f = 0$ .
The Heat Equation: $u_{t} = Δ u$ describes diffusion; its time-independent (steady-state) solutions are exactly the harmonic functions.
The Wave Equation: $u_{tt} = c^{2} Δ u$ shares the Laplacian but is hyperbolic, behaving very differently (no maximum principle, finite propagation speed).
Fourier Series and Separation of Variables: The principal tools for constructing explicit harmonic solutions in rectangles, disks, and spheres.
Brownian Motion: A harmonic function's value at a point equals the expected boundary value reached by a random walker started there — a probabilistic incarnation of the mean value property.

Real-World Applications

Electrostatics — In a charge-free region the electric potential $V$ satisfies $Δ V = 0$ ; solving it gives the field around conductors, capacitors, and microchip interconnects.
Steady-state heat conduction — Temperature distribution in a furnace wall, engine block, or building insulation after thermal equilibrium is reached.
Incompressible, irrotational fluid flow — The velocity potential $ϕ$ of ideal flow obeys $Δ ϕ = 0$ ; used in aerodynamics (airfoil lift via conformal mapping) and groundwater hydrology.
Gravitation — Outside a mass distribution, the Newtonian gravitational potential is harmonic; foundational to celestial mechanics and geodesy (the shape of Earth's geoid).
Image processing — Harmonic inpainting fills missing image regions by solving Laplace's equation with the surrounding pixels as boundary data, producing smooth interpolations.
Membrane and soap-film modeling — The small-deflection shape of a stretched elastic membrane (or minimal soap film) over a wire frame is harmonic, used in tensile architecture.
Electrical Impedance / Tomography (EIT) — Reconstructing internal conductivity from boundary voltage-current data requires solving Laplace-type equations; used in medical and geophysical imaging.

Historical Development

1752 — Euler writes down what is essentially the Laplacian condition for the velocity potential of incompressible fluid flow.
1782–1785 — Laplace systematically uses $Δ V = 0$ for gravitational potentials in his celestial mechanics; the equation acquires his name.
1813 — Poisson generalizes to $Δ V = - 4 π ρ$ (Poisson's equation), correctly handling regions containing mass or charge.
1828 — George Green introduces the Green's function and potential-theoretic identities (Green's theorems), enabling boundary integral solutions.
1839–1850s — Gauss develops the theory of the potential and the mean value / maximum principle; Dirichlet poses the boundary value problem now bearing his name.
1851 — Riemann uses the Dirichlet principle (minimizing the energy integral) to assert existence of harmonic functions, fueling his theory of complex functions.
1870 — Weierstrass exposes a gap in the Dirichlet principle (the minimizer need not exist), triggering decades of rigor in the calculus of variations.
1887–1900 — Schwarz, Neumann, and Poincaré develop rigorous existence methods (alternating method, sweeping-out / balayage).
1900–1923 — Hilbert rehabilitates the Dirichlet principle; Perron (1923) gives the elegant subharmonic-function method that resolves existence for general domains.

Subtleties & Limits

While the theory of harmonic functions is classically complete, several deep and active threads remain:

Regularity of boundaries. The Dirichlet problem is solvable for nice domains, but boundary behavior can fail at irregular points (e.g., the tip of a sharp inward cusp — the classical Lebesgue spine). Characterizing solvability via the Wiener criterion connects to potential theory and capacity, and analogues in non-smooth and fractal domains are still studied.
Higher-order and nonlinear analogues. The biharmonic equation $Δ^{2} u = 0$ (plate bending) lacks a maximum principle, and the p-Laplacian $\nabla \cdot (∣\nabla u ∣^{p - 2} \nabla u) = 0$ generalizes harmonicity nonlinearly with subtle regularity theory.
Geometric harmonic maps. Harmonic maps between Riemannian manifolds generalize harmonic functions; existence, uniqueness, and singularity structure (bubbling) remain research frontiers.
Discrete and probabilistic harmonicity. On graphs and lattices, discrete harmonic functions connect to random walks, electrical networks, and the convergence of numerical schemes; sharp results on infinite graphs are ongoing.
Unique continuation. How strongly a harmonic function's local behavior controls it globally (and quantitative estimates thereof) ties into open questions in spectral geometry and nodal-set bounds (Yau's conjecture territory).