The Laplacian Operator

The Laplacian $\nabla^2$ (also written $\Delta$ ) is the single most consequential differential operator in mathematical physics, and it is assembled from machinery this chapter has already built. Take the gradient of a scalar field $\phi$ — the prerequisite article gives the vector $\nabla\phi$ pointing up the steepest slope — and then take the divergence of that vector field, which the divergence article reads as net outflow per unit volume. The composition $\nabla\cdot(\nabla\phi)$ is the Laplacian: this article is where the two first-order operators of vector calculus fuse into the one second-order operator that governs almost every field equation of physics. What makes $\nabla^2$ special is not its construction but its symmetry. Although built from a vector operator, it is a scalar: the dot product $\nabla\cdot\nabla$ , so it commutes with rotations and knows no preferred axis. That single fact is why the Laplacian, and nothing more elaborate, appears wherever a medium is isotropic — heat in a uniform solid, the potential in vacuum, sound in a still gas, a vibrating drumhead. Its physical content reads as a comparison: $\nabla^2\phi$ measures how far $\phi$ at a point falls below (or rises above) the average over an infinitesimal surrounding sphere. This one interpretation unifies the canonical equations — Laplace, Poisson, heat, and wave — as statements about that gap: diffusion erases it, sources sustain it, wave restoring forces accelerate against it. In symmetry-adapted coordinates the operator wears heavier clothing, and in curved space it becomes the Laplace–Beltrami operator, but the meaning never changes.

Key takeaways

The Laplacian is the divergence of the gradient, $\nabla^2\phi=\nabla\cdot(\nabla\phi)$ — a second-order scalar operator built from two first-order vector ones.
In Cartesian coordinates it is the bare sum of unmixed second partials, $\nabla^2\phi=\partial_{xx}\phi+\partial_{yy}\phi+\partial_{zz}\phi$ ; in curvilinear coordinates the scale factors enter.
Geometrically $\nabla^2\phi$ is six times the leading gap between $\phi$ and its spherical average: $\langle\phi\rangle_{S_a}=\phi(0)+\tfrac{a^2}{6}\nabla^2\phi+O(a^4)$ .
It powers the canonical PDEs — Laplace $\nabla^2\phi=0$ , Poisson $\nabla^2\phi=-\rho/\varepsilon_0$ , heat $\partial_t\phi=D\nabla^2\phi$ , wave $\partial_{tt}\phi=c^2\nabla^2\phi$ .
Its Green's function is the Coulomb–Newton kernel: $\nabla^2(1/r)=-4\pi\,\delta^3(\mathbf r)$ , harmonic everywhere except a point source.
Harmonic functions ( $\nabla^2\phi=0$ ) obey the mean-value property and the maximum principle, so their extrema live only on the boundary.
Key idea: $\nabla^2\phi$ is the local curvature of $\phi$ — how far the field at a point falls below the average over an infinitesimal surrounding sphere — and it is exactly this gap that diffusion smooths, sources sustain, and wave restoring forces respond to.

The core insight

$\nabla^2\phi$ is the local curvature of $\phi$ : it quantifies how much the field at a point falls below (or exceeds) the average over an infinitesimal surrounding sphere, and it is this quantity that diffusion smooths, sources sustain, and wave restoring forces respond to.

Where this lives today

The Laplacian has quietly migrated from the sky into computation and data. Every finite-difference or finite-element solver for an electrostatic field, a temperature distribution, or a stress pattern is discretising $\nabla^2$ into a sparse matrix — the five- and seven-point stencils are the mean-value property on a grid, and the relaxations (Jacobi, Gauss–Seidel, multigrid) that solve them just push each node toward the average of its neighbours. In imaging, the discrete Laplacian is the classic edge detector (Laplacian-of-Gaussian) and the backbone of Laplacian pyramids. The deepest modern reincarnation is the graph Laplacian $L=D-A$ : its eigenvectors drive spectral clustering, random-walk diffusion, dimensionality reduction, and the message-passing of graph neural networks — the same operator, now on a network. In geometry processing the Laplace–Beltrami operator smooths meshes and defines the "shape-DNA" fingerprint of a surface, a computable echo of Mark Kac's 1966 question "can one hear the shape of a drum?" And across physics it remains the kinetic term of the Schrödinger equation, the diffusion term of every reaction–diffusion pattern, and, promoted to the d'Alembertian $\Box=\tfrac1{c^2}\partial_{tt}-\nabla^2$ , the wave operator of relativity.

The physical picture

Picture a temperature field with a cold spot and a hot spot. At the cold spot every neighbour is warmer, so heat flows inward and the spot warms; at the hot spot heat flows outward and it cools. The Laplacian is precisely the quantity positive at the cold dimple and negative at the hot bump — it reads the mismatch between a point and the ring of points around it. In one dimension this is just the second derivative: concave-up ( $\phi''>0$ ) means the curve sits below the line joining its neighbours. In three dimensions the "neighbours" become an infinitesimal sphere, and the rotation-invariant measure of the gap is $\nabla^2\phi$ .

Figure

A cold dimple warms and a hot bump cools: ∇²T > 0 where the field sits below its neighbourhood average and ∇²T < 0 where it rises above, and in 3D the gap is exactly ⟨φ⟩sphere − φ(0) = (a²/6)∇²φ

Algebraically the reason this quantity is so universal is that $\nabla^2=\nabla\cdot\nabla$ is a contraction of a vector with itself, hence a scalar invariant under rotations. Isotropic laws — diffusion, vacuum electrostatics, acoustics — have no preferred direction, and $\nabla^2$ is the simplest second-order operator that respects that. To make the averaging picture quantitative, Taylor-expand $\phi$ and average over a small sphere of radius $a$ : all odd terms and off-diagonal second derivatives cancel by symmetry, leaving $\langle\phi\rangle_{S_a}=\phi(0)+\tfrac{a^2}{6}\nabla^2\phi(0)$ . The Laplacian literally is six times the leading deviation of the spherical average from the central value — the number on the first figure. The second figure captures the operator's other signature — the way it isolates a point source. The outward flux of $\nabla(1/r)=-\hat{\mathbf r}/r^2$ through any sphere about the origin is exactly $-4\pi$ regardless of radius, since the $1/r^2$ gradient cancels the $r^2$ area element. A field harmonic everywhere yet draining a fixed $-4\pi$ through every enclosing surface can only be sourced by a point: $\nabla^2(1/r)=-4\pi\,\delta^3(\mathbf r)$ , the Coulomb and Newtonian kernel.

Figure

The gradient flux of 1/r through any sphere about the origin is −4π independent of radius because the 1/r² gradient exactly cancels the r² area element, so ∇²(1/r) vanishes for r ≠ 0 but equals −4π δ³(r) at the source

The core equations

Start from the definition. The Laplacian is not a new primitive object; it is the gradient followed by the divergence, a composition of two operators the reader already owns. Because divergence returns a scalar, so does the Laplacian.

\nabla^2\phi\equiv\nabla\cdot(\nabla\phi) \qquad (1)

where: $\phi$ — a scalar field, function of position; $\nabla\phi$ — its gradient, a vector field; $\nabla\cdot$ — the divergence operator; $\nabla^2$ — the Laplacian, also written $\Delta$ . Reading: apply gradient, then divergence, and the vector character cancels to leave a scalar. In Cartesian coordinates the unit vectors are constant, so the composition collapses to the bare sum of the three unmixed second partial derivatives — the form most people first meet and the only one in which the operator looks simple.

\nabla^2\phi=\frac{\partial^2\phi}{\partial x^2}+\frac{\partial^2\phi}{\partial y^2}+\frac{\partial^2\phi}{\partial z^2} \qquad (2)

where: $x,y,z$ — Cartesian coordinates. Reading: $\nabla^2\phi$ is the trace of the Hessian $\partial_i\partial_j\phi$ — the sum of pure curvatures along three orthogonal axes — and a trace is rotation-invariant, which is why the answer does not care how the axes are turned. When a problem carries spherical symmetry the operator inherits geometric factors from the varying surface and volume elements. This is the form behind the multipole expansion, the hydrogen atom, and every central-force separation of variables.

\nabla^2\phi=\frac{1}{r^2}\frac{\partial}{\partial r}\!\left(r^2\frac{\partial\phi}{\partial r}\right)+\frac{1}{r^2\sin\theta}\frac{\partial}{\partial\theta}\!\left(\sin\theta\frac{\partial\phi}{\partial\theta}\right)+\frac{1}{r^2\sin^2\theta}\frac{\partial^2\phi}{\partial\varphi^2} \qquad (3)

where: $r$ — radial distance; $\theta$ — polar (colatitude) angle, $0\le\theta\le\pi$ ; $\varphi$ — azimuthal angle, $0\le\varphi<2\pi$ . Reading: the angular block in brackets is the squared angular-momentum operator $-L^2/\hbar^2$ of quantum mechanics, whose eigenfunctions are the spherical harmonics. The cylindrical form is the companion used for waveguides, coaxial geometries, and any problem with an axis rather than a centre; its angular and radial pieces generate the Bessel functions.

\nabla^2\phi=\frac{1}{\rho}\frac{\partial}{\partial\rho}\!\left(\rho\frac{\partial\phi}{\partial\rho}\right)+\frac{1}{\rho^2}\frac{\partial^2\phi}{\partial\varphi^2}+\frac{\partial^2\phi}{\partial z^2} \qquad (4)

where: $\rho$ — distance from the $z$ -axis; $\varphi$ — azimuthal angle; $z$ — axial coordinate. Reading: along $z$ the operator is the plain second derivative because the $z$ -lines are straight and evenly spaced. Both special cases are instances of one master formula in orthogonal curvilinear coordinates — the curvilinear gradient fed into the curvilinear divergence, with scale factors $h_i$ carrying all the geometry.

\nabla^2\phi=\frac{1}{h_1h_2h_3}\sum_{i=1}^{3}\frac{\partial}{\partial u_i}\!\left(\frac{h_1h_2h_3}{h_i^2}\frac{\partial\phi}{\partial u_i}\right) \qquad (5)

where: $u_i$ — curvilinear coordinates; $h_i=|\partial\mathbf r/\partial u_i|$ — scale factors ( $h=1$ Cartesian; $h_r=1,h_\theta=r,h_\varphi=r\sin\theta$ spherical). Reading: set the $h_i$ and equations (2)–(4) fall out; the same object generalises to $\tfrac1{\sqrt g}\partial_i(\sqrt g\,g^{ij}\partial_j\phi)$ , the Laplace–Beltrami operator on a manifold with metric $g_{ij}$ . Finally, the property that anchors potential theory: the Laplacian of the inverse-distance kernel is a point source — the single equation from which Coulomb's and Newton's inverse-square laws, image charges, and the whole Green's-function method flow.

\nabla^2\!\left(\frac{1}{|\mathbf r-\mathbf r'|}\right)=-4\pi\,\delta^3(\mathbf r-\mathbf r') \qquad (6)

where: $\delta^3$ — the three-dimensional Dirac delta; $\mathbf r,\mathbf r'$ — field and source points. Reading: $1/|\mathbf r-\mathbf r'|$ is harmonic everywhere except at the source, where it carries a delta of weight $-4\pi$ , so any potential is the source density convolved with this kernel.

Derivation AThe spherical average and the factor one-sixth

The goal is to prove the relation that turns "curvature equals deviation from the local average" into an exact statement: the mean of $\phi$ over a small sphere of radius $a$ exceeds the central value by precisely $(a^2/6)\nabla^2\phi$ . Establishing this pins down where the one-sixth comes from, shows why odd and off-diagonal derivatives leave no trace, and proves the mean-value property of harmonic functions as the special case $\nabla^2\phi=0$ . We expand in a Taylor series, average term by term over the sphere, and keep the survivors.

Step 1. Expand $\phi$ about the sphere's centre in Cartesian components, $\phi(\mathbf x)=\phi(0)+x_i\,\partial_i\phi+\tfrac12 x_ix_j\,\partial_i\partial_j\phi+O(x^3)$ , with the summation convention on repeated indices. Why: an analytic field near a point is completely captured by its derivatives there, so the spherical average will be a sum of averages of the monomials $x_i$ and $x_ix_j$ , which are pure geometry.

Step 2. Average each monomial over the sphere of radius $a$ . By symmetry the linear average vanishes, $\langle x_i\rangle=0$ , and the quadratic average is isotropic, $\langle x_ix_j\rangle=\tfrac{a^2}{3}\delta_{ij}$ (the off-diagonal terms integrate to zero, and the three diagonal terms are equal and sum to $\langle r^2\rangle=a^2$ ). Why: a sphere has no preferred direction, so any average must be built from the only invariant tensor available, $\delta_{ij}$ , and the trace fixes its coefficient.

Step 3. Assemble the surviving contribution, $\langle\phi\rangle=\phi(0)+\tfrac12\cdot\tfrac{a^2}{3}\delta_{ij}\,\partial_i\partial_j\phi=\phi(0)+\tfrac{a^2}{6}\,\partial_i\partial_i\phi=\phi(0)+\tfrac{a^2}{6}\nabla^2\phi$ , since $\delta_{ij}\partial_i\partial_j=\partial_i\partial_i=\nabla^2$ is the trace. Why: only the trace of the Hessian survives the isotropic average, and that trace is exactly the Laplacian — which is the precise sense in which $\nabla^2$ is the rotation-invariant curvature.

Step 4. Read off two corollaries. If $\nabla^2\phi=0$ the average equals the central value for every $a$ (the mean-value property), and a strict interior maximum would force the average below the centre, contradicting it — hence harmonic functions attain their extrema only on the boundary (the maximum principle). Why: these are immediate consequences of the same expansion, which is why the Laplacian and the maximum principle are inseparable. *Verified: for $\phi=r^2$ the formula gives $\langle r^2\rangle=0+(a^2/6)(6)=a^2$ , matching the direct integral; coefficient one-sixth cross-checks Arfken and Weber Ch. 3.*

Go deeper · Derivation B · 2 minThe fundamental solution and its delta-function source

Derivation BThe fundamental solution and its delta-function source

The aim is to establish equation (6) rigorously — that $1/r$ satisfies Laplace's equation everywhere except the origin, yet its Laplacian integrated over any region containing the origin equals $-4\pi$ , so the only consistent reading is a Dirac delta of weight $-4\pi$ . This is the most important single result about the operator: the Green's function from which the potential of any source is built by superposition. We compute the radial Laplacian away from the origin, then handle the origin with a flux argument that sidesteps the singularity.

Step 1. For $r\neq0$ apply the radial part of equation (3): $\nabla^2(1/r)=\tfrac1{r^2}\tfrac{\mathrm d}{\mathrm dr}\!\big(r^2\tfrac{\mathrm d}{\mathrm dr}\tfrac1r\big)=\tfrac1{r^2}\tfrac{\mathrm d}{\mathrm dr}\!\big(r^2\cdot(-\tfrac1{r^2})\big)=\tfrac1{r^2}\tfrac{\mathrm d}{\mathrm dr}(-1)=0$ . Why: the inner bracket collapses to the constant $-1$ , whose derivative is zero, so $1/r$ is harmonic wherever it is defined — the singularity at the origin is the only place anything can happen.

Step 2. Integrate the Laplacian over a ball $V$ containing the origin and convert to a surface flux by the divergence theorem, $\int_V\nabla^2\tfrac1r\,\mathrm dV=\oint_{\partial V}\nabla\tfrac1r\cdot\mathrm d\mathbf S$ , using $\nabla\tfrac1r=-\hat{\mathbf r}/r^2$ . On a sphere of radius $R$ this is $\oint(-\tfrac1{R^2}\hat{\mathbf r})\cdot\hat{\mathbf r}\,R^2\,\mathrm d\Omega=-\int\mathrm d\Omega=-4\pi$ . Why: the $R^2$ of the area element cancels the $1/R^2$ of the gradient, so the flux is $-4\pi$ for every radius — the fingerprint of a point source rather than a smooth distribution.

Step 3. Reconcile the two results. A function that vanishes for all $r\neq0$ yet integrates to $-4\pi$ over any enclosing volume is by definition $-4\pi$ times the Dirac delta, giving $\nabla^2(1/r)=-4\pi\,\delta^3(\mathbf r)$ . Why: no ordinary function can be zero almost everywhere with a nonzero integral, so the object is genuinely distributional — this is where potential theory necessarily leaves classical calculus for the theory of distributions.

Step 4. Superpose. Writing a general source density $\rho(\mathbf r')$ as a sum of points, the potential $\phi(\mathbf r)=\tfrac1{4\pi\varepsilon_0}\int\tfrac{\rho(\mathbf r')}{|\mathbf r-\mathbf r'|}\mathrm dV'$ then satisfies $\nabla^2\phi=-\rho/\varepsilon_0$ , Poisson's equation. Why: linearity of $\nabla^2$ lets the delta-source result be integrated against any $\rho$ , which is exactly why the inverse-distance kernel is called the Green's function of the Laplacian. *Verified: recompute the flux integral (independent of $R$ ) and compare with Jackson Eq. (1.31) and Arfken and Weber Ch. 3; the factor $4\pi$ is standard.*

Go deeper · Derivation C · 1 minGreen's first and second identities

Derivation CGreen's first and second identities

The purpose here is to derive the two integral identities that carry the Laplacian into potential theory: Green's first identity, relating the volume integral of $f\nabla^2 g$ to boundary data and a gradient overlap, and Green's second identity, the antisymmetric combination behind uniqueness theorems and the integral representation of solutions. Both follow from one product rule integrated by the divergence theorem. We start from the divergence of $f\nabla g$ .

Step 1. Apply the product rule for divergence to $\mathbf F=f\nabla g$ , giving $\nabla\cdot(f\nabla g)=f\nabla^2 g+\nabla f\cdot\nabla g$ . Why: this is the Laplacian version of "differentiate a product" — it splits the divergence of a scaled gradient into a curvature term and an overlap term, and both survive the volume integral.

Step 2. Integrate over $V$ and apply the divergence theorem to the left side, obtaining Green's first identity $\int_V(f\nabla^2 g+\nabla f\cdot\nabla g)\,\mathrm dV=\oint_{\partial V}f\,\nabla g\cdot\mathrm d\mathbf S$ . Setting $f=g=\phi$ gives $\int_V|\nabla\phi|^2\,\mathrm dV=\oint\phi\,\nabla\phi\cdot\mathrm d\mathbf S-\int_V\phi\nabla^2\phi\,\mathrm dV$ , the energy identity. Why: for a harmonic $\phi$ with zero boundary data the right side vanishes, forcing $\int|\nabla\phi|^2=0$ and hence $\phi=$ const — the cleanest proof that the Dirichlet problem has a unique solution.

Step 3. Write the same identity with $f$ and $g$ interchanged, $\int_V(g\nabla^2 f+\nabla g\cdot\nabla f)\,\mathrm dV=\oint_{\partial V}g\,\nabla f\cdot\mathrm d\mathbf S$ , and subtract. The symmetric overlap $\nabla f\cdot\nabla g$ cancels, leaving Green's second identity $\int_V(f\nabla^2 g-g\nabla^2 f)\,\mathrm dV=\oint_{\partial V}(f\nabla g-g\nabla f)\cdot\mathrm d\mathbf S$ . Why: the antisymmetry is what makes the identity useful — choosing $g$ to be the Green's function turns the volume term into $f$ evaluated at a point, yielding the integral representation of the solution from its boundary values. *Verified: standard; reproduces Jackson Ch. 1 Eqs. (1.34)–(1.35); the cancellation of the symmetric gradient-overlap term is exact.*

Limiting cases

How the Laplacian specialises across dimensions, symmetries, and field types:

Case	Result	Meaning
One dimension	$\nabla^2\phi=\mathrm d^2\phi/\mathrm dx^2$	Ordinary second derivative — concavity. Harmonic functions are straight lines $\phi=ax+b$ .
Radial field in 3D	$\nabla^2\phi(r)=\tfrac1{r^2}\tfrac{\mathrm d}{\mathrm dr}(r^2\tfrac{\mathrm d\phi}{\mathrm dr})=\tfrac1r\tfrac{\mathrm d^2}{\mathrm dr^2}(r\phi)$	For $\phi=1/r$ this vanishes off the origin; at the origin $\nabla^2(1/r)=-4\pi\delta^3(\mathbf r)$ , the Coulomb–Newton kernel.
Harmonic case	$\nabla^2\phi=0$	No interior extrema (maximum principle); each value equals its spherical average (mean-value property). In 2D the harmonic radial solution is $\ln\rho$ , not $1/r$ .
Vector Laplacian	$\nabla^2\mathbf A=\nabla(\nabla\cdot\mathbf A)-\nabla\times(\nabla\times\mathbf A)$	Coordinate-free definition; equals the componentwise Laplacian only in Cartesian coordinates because curvilinear unit vectors vary.

Common trapA zero Laplacian does not mean the field is flat

It is tempting to read $\nabla^2\phi=0$ as "no curvature, so $\phi$ is flat." That is false. The Laplacian is the sum of the curvatures along orthogonal directions, so it can vanish because those curvatures cancel, not because each is zero. The saddle $\phi=x^2-y^2$ has $\nabla^2\phi=2-2=0$ everywhere yet is dramatically curved — bowing up along $x$ , down along $y$ . So do $\phi=xy$ , $\phi=1/r$ , and $\phi=\ln\rho$ : all harmonic, none flat. A harmonic function has its curvatures in perfect balance at every point, which is exactly why it has no interior peaks — a peak needs every direction to curve down, making $\nabla^2\phi<0$ . Zero Laplacian means balance, not absence of curvature.

Worked examples

1 · The spherical-average test on a concrete field (confidence builder)

Compute $\nabla^2 r^2$ for $r^2=x^2+y^2+z^2$ , then verify the mean-value formula $\langle\phi\rangle_{S_a}=\phi(0)+\tfrac{a^2}{6}\nabla^2\phi$ directly for a sphere of radius $a=0.1$ .

Step 1. Differentiate in Cartesian form. Since $\partial_{xx}(x^2+y^2+z^2)=2$ and likewise for $y$ and $z$ , the Laplacian is $\nabla^2 r^2=2+2+2=\mathbf 6$ . The field curves upward equally along all three axes, and the trace of that curvature is six.

Step 2. Predict the spherical average from the mean-value formula. With $\phi(0)=0$ at the centre and $\nabla^2\phi=6$ , the formula gives $\langle r^2\rangle_{S_a}=0+\tfrac{a^2}{6}\cdot6=a^2$ . For $a=0.1$ this predicts $\langle r^2\rangle=\mathbf{0.01}$ .

Step 3. Check it against the direct average. On a sphere of radius $a$ every point has $r=a$ , so $\langle r^2\rangle_{S_a}=a^2$ exactly, with no higher-order corrections — and for $a=0.1$ that is $0.01$ , matching Step 2. The agreement is exact here because $r^2$ is a quadratic, so the $O(a^4)$ remainder is genuinely zero. $\nabla^2 r^2=6$ , and the spherical average $\langle r^2\rangle=a^2$ confirms the factor one-sixth.

2 · Steady-state temperature between two spheres (fully worked)

A thick shell of uniform material has inner radius $a=1\ \mathrm{cm}$ held at $T_1=100\,^\circ\mathrm C$ and outer radius $b=2\ \mathrm{cm}$ held at $T_2=0\,^\circ\mathrm C$ . In steady state $\partial_t T=0$ , so $\nabla^2 T=0$ . Find $T(r)$ and its value at the mid-radius.

Step 1. Use the radial Laplace equation. Spherical symmetry kills the angular terms, leaving $\tfrac1{r^2}\tfrac{\mathrm d}{\mathrm dr}(r^2 T')=0$ , so $r^2T'=\text{const}$ . Integrating twice gives the general harmonic radial field $T(r)=A+\dfrac{B}{r}$ — the same $A+B/r$ form as an electrostatic potential, because both obey Laplace's equation.

Step 2. Impose the boundary values. From $T(a)=A+B/a=100$ and $T(b)=A+B/b=0$ , subtract to get $B\big(\tfrac1a-\tfrac1b\big)=100$ , i.e. $B\,\dfrac{b-a}{ab}=100$ . With $a=0.01\ \mathrm m$ , $b=0.02\ \mathrm m$ : $B=100\cdot\dfrac{ab}{b-a}=100\cdot\dfrac{(0.01)(0.02)}{0.01}=\mathbf{2}\ (^\circ\mathrm C\cdot\mathrm m)$ .

Step 3. Solve for the constant. From $T(a)=A+B/a=100$ : $A=100-\dfrac{2}{0.01}=100-200=\mathbf{-100}\,^\circ\mathrm C$ . So the profile is $T(r)=-100+\dfrac{2}{r}$ with $r$ in metres. Check the ends: $T(0.01)=-100+200=100$ and $T(0.02)=-100+100=0$ — both boundary conditions met.

Step 4. Evaluate at the mid-radius $r=1.5\ \mathrm{cm}=0.015\ \mathrm m$ : $T=-100+\dfrac{2}{0.015}=-100+133.3=\mathbf{33.3\,^\circ\mathrm C}$ . The midpoint sits below the arithmetic mean $50\,^\circ\mathrm C$ of the walls — the $1/r$ profile is skewed by geometry, exactly the maximum-principle statement that a harmonic $T$ has no interior extremum and is fixed everywhere by its boundary values. $T(r)=-100+2/r\ ^\circ\mathrm C$ , with $T=33.3\,^\circ\mathrm C$ at the mid-radius.

Physics of the portrait

The Laplacian is where French celestial mechanics, English electricity, and German potential theory converge on one operator — first written to explain how planets pull on one another, then recognised as the master equation of every static field.

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

Pierre-Simon Laplace

1749–1827 · celestial mechanics and potential theory

Pierre-Simon Laplace built his reputation proving that the Solar System is stable, and the operator that bears his name was a tool in that campaign. In his memoirs of the 1780s on the gravitational attraction of spheroids, and later throughout the Mécanique Céleste (1799–1825), he introduced the potential function $V$ whose gradient gives the gravitational field and showed that in empty space it satisfies $\nabla^2 V=0$ — Laplace's equation — reducing the awkward vector problem of gravity to a single scalar one. The equation was not entirely his invention; Euler and Lagrange had met it in fluid dynamics decades earlier, but it was Laplace who made it the organising principle of a physical theory. In 1813 Siméon Denis Poisson corrected the picture inside matter, showing that where mass or charge is present the right-hand side is not zero but $-4\pi\rho$ , giving Poisson's equation and the delta-function source of the inverse-distance kernel. The self-taught English mathematician George Green, in his 1828 Essay, supplied the integral identities and the "Green's function" that turn the operator into a machine for boundary-value problems, and Carl Friedrich Gauss placed the subject on rigorous footing in 1840 with the mean-value and maximum principles. Laplace himself is as remembered for his determinism — the "Laplace demon" who could predict the future from the present state of every particle — as for the operator, and the two are of a piece: a world governed by fields whose evolution is fixed everywhere by $\nabla^2$ acting on smooth potentials.

History

1782

— Laplace, studying planetary attraction, writes the potential equation ∇²V = 0

1813

— Poisson extends it to ∇²V = −4πρ inside a mass or charge distribution

1828

— Green publishes his Essay, introducing Green's identities and the Green's function

1840

— Gauss founds rigorous potential theory, proving the mean-value and maximum principles

1864

— Beltrami generalises the operator to curved surfaces as the Laplace–Beltrami operator

Going further

Texts. Arfken and Weber, Mathematical Methods for Physicists, 7th ed., Ch. 3 (vector analysis and the Laplacian) · Stone and Goldbart, Mathematics for Physics, Ch. 11 (vector calculus) and Ch. 18 (partial differential equations) · Hassani, Mathematical Physics, Ch. on Green's functions and PDEs · Riley, Hobson and Bence, Mathematical Methods for Physics and Engineering, 3rd ed., Ch. 10–11 · Jackson, Classical Electrodynamics, Ch. 1 (Poisson and Laplace equations, Green's identities).

Provenance. Forms and constants cross-checked against Arfken and Weber 7th ed. Ch. 3, Stone and Goldbart, Riley-Hobson-Bence Ch. 10–11, and Jackson Ch. 1. A verifier should reproduce: the Cartesian-to-curvilinear master formula via scale factors; the mean-value coefficient $a^2/6$ and $\nabla^2 r^2=6$ ; the flux integral giving $\nabla^2(1/r)=-4\pi\delta^3$ ; Green's identities from the product rule plus the divergence theorem; and Worked Example 2, $T(r)=-100+2/r$ with $T(0.015)=33.3\,^\circ\mathrm C$ . The saddle check $\nabla^2(x^2-y^2)=0$ is exact.

Prerequisites. Scalar and Vector Fields · The Gradient Operator · Divergence of a Vector Field · The Del Operator and Vector Differential Identities. Leads to. The Divergence (Gauss) Theorem · Green's Theorem in the Plane · Volume Integrals.

Ask the tutor: Ask me to derive the spherical or cylindrical Laplacian from the scale factors, or to show how $\nabla^2(1/r)=-4\pi\delta^3$ turns a charge distribution into its potential.

Why is the Laplacian called a scalar operator when it is built from the vector operator ∇?

Because it is the dot product $\nabla\cdot\nabla$ , and a dot product of two vectors is a scalar. Gradient turns $\phi$ into a vector $\nabla\phi$ , but the following divergence contracts that vector back to a single number, so the operation maps scalar to scalar. The deep consequence is rotational invariance: like any dot product, $\nabla^2$ is unchanged when the axes are rotated, so $\nabla^2\phi$ is a genuine scalar field. That is why it governs isotropic physics — media with no preferred direction admit only rotation-invariant operators, and $\nabla^2$ is the simplest second-order one.

What is the difference between the scalar Laplacian and the vector Laplacian?

The scalar Laplacian $\nabla^2\phi=\nabla\cdot(\nabla\phi)$ acts on a scalar field and returns a scalar. The vector Laplacian $\nabla^2\mathbf A$ acts on a vector field and is defined invariantly by $\nabla^2\mathbf A=\nabla(\nabla\cdot\mathbf A)-\nabla\times(\nabla\times\mathbf A)$ . The crucial trap is that $\nabla^2\mathbf A$ equals "apply the scalar Laplacian to each component" only in Cartesian coordinates. In spherical or cylindrical coordinates the unit vectors $\hat{\mathbf r},\hat{\boldsymbol\theta},\hat{\boldsymbol\varphi}$ themselves change from point to point, so differentiating a component also differentiates the direction it points, and cross-terms appear: $(\nabla^2\mathbf A)_r\neq\nabla^2 A_r$ . This is a common source of error in curvilinear electromagnetism and fluid dynamics.

Why does the same operator appear in the heat, wave, Poisson, and Schrödinger equations?

Because all four are statements about the same geometric quantity — the gap between a field and its local average — differing only in what that gap equals. In the heat equation $\partial_t T=D\nabla^2 T$ the gap drives a rate of change, so hot spots cool and cold spots warm until the field is harmonic. In the wave equation $\partial_{tt}\psi=c^2\nabla^2\psi$ it drives an acceleration, giving oscillation. In Poisson's equation $\nabla^2\phi=-\rho/\varepsilon_0$ a static source fixes it. In Schrödinger's equation the same $\nabla^2$ is the kinetic energy operator (up to $-\hbar^2/2m$ ). One operator, because all these systems are isotropic and second-order, and $\nabla^2$ is the unique operator with those properties.

What exactly does "harmonic" mean, and why can harmonic functions have no interior peaks?

A function is harmonic in a region if $\nabla^2\phi=0$ there. The mean-value property then says the value at any interior point equals the average of $\phi$ over any sphere around it that fits inside the region. A strict interior maximum is incompatible with this: at a peak every nearby value is smaller, so the spherical average would be below the centre, contradicting equality. Hence harmonic functions attain their maxima and minima only on the boundary — the maximum principle. This is why the potential inside an empty charge-free cavity is completely determined by its values on the walls, and why a steady temperature profile can have no interior hot spot: heat would flow away from it and erase it.

Where does the factor 4π in ∇²(1/r) = −4π δ³(r) come from?

It is the surface area of the unit sphere, entering through a flux argument. The gradient $-\hat{\mathbf r}/r^2$ has outward flux through a sphere of radius $R$ equal to (magnitude $1/R^2$ ) times (area $4\pi R^2$ ), and the powers of $R$ cancel to leave $-4\pi$ , independent of radius. A field harmonic everywhere yet draining a fixed $-4\pi$ through every enclosing surface can only be sourced where it blows up, and the object zero everywhere but integrating to $-4\pi$ is $-4\pi\,\delta^3(\mathbf r)$ . In Gaussian units this $4\pi$ is why Poisson's equation reads $\nabla^2\phi=-4\pi\rho$ ; in SI it is absorbed into $\varepsilon_0$ .

Is ∇² the same as Δ, and how does the d'Alembertian relate to it?

Yes: $\nabla^2$ and $\Delta$ are two notations for the identical operator — physicists write $\nabla^2$ to emphasise "del dotted with del," mathematicians favour $\Delta$ . The d'Alembertian is the spacetime cousin: $\Box=\tfrac1{c^2}\partial_{tt}-\nabla^2$ (signs depend on the metric convention), the Lorentz-invariant operator whose vanishing is the wave equation. So $\nabla^2$ is the purely spatial part, isotropic under rotations of space; $\Box$ extends that isotropy to rotations and boosts, which is why it, not $\nabla^2$ , is the natural operator of relativistic field theory. On a curved manifold the same idea generalises to the Laplace–Beltrami operator built from the metric.