{
  "id": "laplace-operator",
  "title": "Laplace Operator",
  "lang": "en",
  "source_lang": "en",
  "summary": "A second-order differential operator that measures local spatial variation and underlies equations of potential theory, diffusion, waves, and quantum mechanics.",
  "category": "mathematics",
  "subcategory": "analysis",
  "type": "concept",
  "aliases": [
    "Laplacian",
    "Laplacian operator"
  ],
  "generated_at": "2026-10-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/laplace-operator/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/laplace-operator/",
      "json": "https://aiwiki.page/zh/laplace-operator.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The Laplace operator, or Laplacian, is a second-order differential operator that maps a scalar function to the sum of its second spatial derivatives in orthogonal Cartesian coordinates. Usually denoted by \\(\\Delta\\) or \\(\\nabla^2\\), it is a central operator in partial differential equations, describing equilibrium potentials, diffusion, and other spatial processes. Its geometric generalization acts on functions on curved spaces, while discrete analogues act on grids and graphs. (web.stanford.edu)"
    },
    {
      "heading": "Definition and basic properties",
      "text": "For a twice continuously differentiable function \\(f\\) on an open subset of Euclidean space \\(\\mathbb{R}^n\\), the Laplacian is\n\n\\[\\Delta f=\\sum_{i=1}^{n}\\frac{\\partial^2 f}{\\partial x_i^2}.\\]\n\nThus, in one dimension it is the ordinary second derivative; in three dimensions,\n\n\\[\\Delta f=f_{xx}+f_{yy}+f_{zz}.\\]\n\nAn equivalent definition is\n\n\\[\\Delta f=\\operatorname{div}(\\nabla f),\\]\n\nthe divergence of the gradient. It is also the trace of the Hessian matrix. These expressions describe the same operator in Cartesian coordinates. (web.stanford.edu)\n\nThe operator is linear:\n\n\\[\\Delta(af+bg)=a\\Delta f+b\\Delta g\\]\n\nfor constant scalars \\(a,b\\). It is invariant under translations and orthogonal changes of Cartesian coordinates: rotating the coordinate axes does not change the scalar quantity it computes. Unlike the Hessian, the Laplacian retains only the sum of directional second derivatives, not their full directional distribution. (math.mit.edu)\n\nSome mathematical texts define the Laplacian with the opposite sign, as \\(-\\operatorname{div}\\nabla\\). This entry uses \\(\\Delta=\\operatorname{div}\\nabla\\), so that \\(-\\Delta\\), with appropriate boundary conditions, is nonnegative. Sign conventions must be checked when comparing formulas, particularly in spectral theory and geometry. (arxiv.org)"
    },
    {
      "heading": "Local interpretation",
      "text": "The Laplacian measures how a function differs from its nearby average. If \\(f\\) is sufficiently smooth and \\(S^{n-1}\\) is the unit sphere, Taylor expansion gives\n\n\\[\\frac{1}{|S^{n-1}|}\\int_{S^{n-1}}f(x+r\\omega)\\,dS(\\omega) = f(x)+\\frac{r^2}{2n}\\Delta f(x)+o(r^2).\\]\n\nConsequently, a positive Laplacian indicates that the average over a sufficiently small surrounding sphere exceeds the value at its center to leading order; a negative Laplacian indicates the reverse. (web.stanford.edu)\n\nAt an interior local minimum of a twice differentiable real-valued function, \\(\\Delta f\\geq0\\); at an interior local maximum, \\(\\Delta f\\leq0\\). The converses do not hold: the sum of second derivatives does not determine whether the point is an extremum. For example,\n\n\\[f(x,y)=x^2-y^2\\]\n\nhas \\(\\Delta f=0\\), although the origin is a saddle point. (web.stanford.edu)\n\nA function satisfying\n\n\\[\\Delta f=0\\]\n\nis called a harmonic function. Harmonic functions have an exact mean-value property on spheres and balls contained in their domain. On a connected domain, a nonconstant harmonic function cannot attain an interior maximum or minimum. These facts help explain the uniqueness of many equilibrium boundary-value problems. (web.stanford.edu)"
    },
    {
      "heading": "Coordinate expressions",
      "text": "The Cartesian formula cannot be transferred unchanged to curvilinear coordinates: scale factors and volume elements must be included. In plane polar coordinates \\((r,\\theta)\\),\n\n\\[\\Delta f = \\frac{\\partial^2f}{\\partial r^2} +\\frac1r\\frac{\\partial f}{\\partial r} +\\frac1{r^2}\\frac{\\partial^2f}{\\partial\\theta^2}.\\]\n\nIn spherical coordinates \\((r,\\theta,\\phi)\\), with \\(\\theta\\) the polar angle,\n\n\\[\\Delta f = \\frac1{r^2}\\frac{\\partial}{\\partial r} \\left(r^2\\frac{\\partial f}{\\partial r}\\right) + \\frac1{r^2\\sin\\theta}\\frac{\\partial}{\\partial\\theta} \\left(\\sin\\theta\\frac{\\partial f}{\\partial\\theta}\\right) + \\frac1{r^2\\sin^2\\theta} \\frac{\\partial^2f}{\\partial\\phi^2}.\\]\n\nThe apparent singularities at the origin and poles arise from the coordinates rather than from the Euclidean operator itself. (ocw.mit.edu)\n\nFor a radial function \\(f(x)=F(r)\\) in \\(\\mathbb{R}^n\\), where \\(r=|x|>0\\),\n\n\\[\\Delta f=F''(r)+\\frac{n-1}{r}F'(r).\\]\n\nThis reduces rotationally symmetric Laplace and Poisson equations to ordinary differential equations. (math.stanford.edu)"
    },
    {
      "heading": "Differential equations and physical applications",
      "text": "The Laplacian appears in several fundamental equations:\n\n- Laplace’s equation: \\(\\Delta u=0\\), describing source-free equilibrium fields.\n- Poisson’s equation: \\(\\Delta u=f\\), describing an equilibrium field with a prescribed source.\n- Heat equation: \\(\\partial_tu=\\kappa\\Delta u\\), where \\(\\kappa>0\\), describing homogeneous isotropic diffusion.\n- Wave equation: \\(\\partial_t^2u=c^2\\Delta u\\), describing waves in a homogeneous isotropic medium. (math.mit.edu)\n\nIn electrostatics, the electric potential \\(V\\) in a medium with constant permittivity satisfies\n\n\\[\\Delta V=-\\frac{\\rho}{\\varepsilon},\\]\n\nwhere \\(\\rho\\) is electric charge density. In a charge-free region this becomes Laplace’s equation. Analogous equations govern Newtonian gravitational potentials. (web.stanford.edu)\n\nIn nonrelativistic quantum mechanics, the Laplacian represents the spatial part of the kinetic-energy operator for a particle of mass \\(m\\):\n\n\\[\\widehat T=-\\frac{\\hbar^2}{2m}\\Delta.\\]\n\nThe Schrödinger equation therefore takes the form\n\n\\[i\\hbar\\partial_t\\psi = -\\frac{\\hbar^2}{2m}\\Delta\\psi+V\\psi.\\]\n\nHere \\(\\psi\\) is the wave function and \\(V\\) is potential energy. (damtp.cam.ac.uk)\n\nThe ordinary Laplacian assumes an isotropic spatial response. When transport coefficients vary in space, or transport is direction-dependent, the relevant operator is generally\n\n\\[\\operatorname{div}(A(x)\\nabla u),\\]\n\nwhere \\(A(x)\\) is a coefficient matrix, rather than a constant multiple of \\(\\Delta\\). (math.stanford.edu)"
    },
    {
      "heading": "Boundary conditions and the energy identity",
      "text": "On a bounded domain \\(\\Omega\\), a differential equation involving the Laplacian generally requires boundary data. Common choices in a boundary-value problem include:\n\n- Dirichlet conditions, prescribing \\(u\\) on the boundary;\n- Neumann conditions, prescribing its outward normal derivative \\(\\partial_nu\\);\n- Robin conditions, prescribing a linear combination of the two. (arxiv.org)\n\nThe divergence theorem yields Green’s first identity. For sufficiently smooth real-valued \\(u,v\\),\n\n\\[\\int_\\Omega v\\,\\Delta u\\,dx = -\\int_\\Omega\\nabla v\\cdot\\nabla u\\,dx +\\int_{\\partial\\Omega}v\\,\\partial_nu\\,dS.\\]\n\nTaking \\(v=u\\), with boundary conditions that eliminate the boundary term, gives\n\n\\[\\int_\\Omega u(-\\Delta u)\\,dx = \\int_\\Omega|\\nabla u|^2\\,dx\\geq0.\\]\n\nThis connects the Laplacian to an energy functional measuring spatial variation. (math.stanford.edu)\n\nAs an operator on an Hilbert space such as \\(L^2(\\Omega)\\), the Laplacian is not specified by its differential expression alone: its domain and boundary conditions are essential. Suitable Dirichlet and Neumann realizations of \\(-\\Delta\\) are self-adjoint operators, but they have different spectra and kernels. On a bounded connected domain, constants form the Neumann kernel; homogeneous Dirichlet conditions exclude nonzero constants. (arxiv.org)"
    },
    {
      "heading": "Fourier and spectral descriptions",
      "text": "With the Fourier transform convention using \\(e^{-ix\\cdot\\xi}\\),\n\n\\[\\widehat{\\Delta f}(\\xi)=-|\\xi|^2\\widehat f(\\xi).\\]\n\nThe Laplacian thus multiplies each spatial frequency by a factor proportional to the square of its magnitude. For the heat equation, the corresponding time-evolution factor is \\(e^{-\\kappa t|\\xi|^2}\\), explaining why diffusion suppresses short-wavelength variations especially rapidly. (arxiv.org)\n\nOn a bounded smooth domain, standard Dirichlet or Neumann conditions lead to the eigenvalue problem\n\n\\[-\\Delta\\phi_k=\\lambda_k\\phi_k.\\]\n\nIts eigenfunctions can be chosen as an orthonormal basis of \\(L^2(\\Omega)\\). Under homogeneous boundary conditions, an initial heat distribution can be expanded into these modes:\n\n\\[u(x,t)=\\sum_k a_k e^{-\\kappa\\lambda_kt}\\phi_k(x).\\]\n\nThe same spatial modes determine vibration frequencies through \\(\\omega_k=c\\sqrt{\\lambda_k}\\). (web.stanford.edu)\n\nThe spectral behavior depends on the underlying space. On all of \\(\\mathbb{R}^n\\), the standard operator \\(-\\Delta\\) has continuous spectrum \\([0,\\infty)\\), rather than the discrete eigenvalue sequence characteristic of bounded domains. (arxiv.org)"
    },
    {
      "heading": "Geometric generalization",
      "text": "On a manifold equipped with a Riemannian metric, the corresponding scalar operator is the Laplace–Beltrami operator:\n\n\\[\\Delta_gf=\\operatorname{div}_g(\\nabla_gf).\\]\n\nIn local coordinates,\n\n\\[\\Delta_gf = \\frac1{\\sqrt{|g|}} \\sum_{i,j} \\frac{\\partial}{\\partial x^i} \\left( \\sqrt{|g|}\\,g^{ij} \\frac{\\partial f}{\\partial x^j} \\right),\\]\n\nwhere \\(|g|=\\det(g_{ij})\\) and \\((g^{ij})\\) is the inverse metric matrix. This is an intrinsic definition: its coordinate expressions describe the same geometrically defined operator. When the metric is Euclidean, it recovers the ordinary Laplacian. (math.mit.edu)\n\nLaplacians also act on differential forms and other geometric objects. Such extensions require specifying the structure used to define the operator; a scalar Laplacian applied componentwise is not automatically an intrinsic operator on arbitrary tensor fields. (math.mit.edu)"
    },
    {
      "heading": "Discrete and numerical forms",
      "text": "The finite difference method approximates the two-dimensional Laplacian on a square grid of spacing \\(h\\) by the five-point stencil\n\n\\[(\\Delta_hu)_{i,j} = \\frac{ u_{i+1,j}+u_{i-1,j} +u_{i,j+1}+u_{i,j-1} -4u_{i,j} }{h^2}.\\]\n\nThis compares a grid value with its four nearest neighbors and converts a differential equation into a sparse system of linear equations. For a sufficiently smooth function, Taylor expansion shows that the interior approximation has error \\(O(h^2)\\); the full numerical solution also depends on boundary treatment and stability. (math.mit.edu)\n\nIn graph theory, the graph Laplacian of an undirected weighted graph is\n\n\\[L=D-A,\\]\n\nwhere \\(A\\) is its symmetric adjacency matrix and \\(D\\) contains the weighted vertex degrees. It acts on vertex values by\n\n\\[(Lf)_i=\\sum_jw_{ij}(f_i-f_j).\\]\n\nIts energy identity is\n\n\\[f^\\mathsf{T}Lf = \\sum_{\\{i,j\\}\\in E}w_{ij}(f_i-f_j)^2.\\]\n\nFor positive edge weights, \\(L\\) is positive semidefinite, and its zero eigenspace consists of functions constant on each connected component. This convention corresponds to the sign of \\(-\\Delta\\). Graph-Laplacian eigenvectors are used in spectral graph drawing and partitioning, extending the relationship between spatial variation, energy, and eigenmodes to discrete networks. (cs.yale.edu)"
    }
  ],
  "links": [
    "partial-differential-equation",
    "open-set",
    "euclidean-space",
    "gradient",
    "matrix-trace",
    "hessian-matrix",
    "harmonic-function",
    "laplace-equation",
    "poisson-equation",
    "heat-equation",
    "wave-equation",
    "electrostatics",
    "electric-potential",
    "quantum-mechanics",
    "schrodinger-equation",
    "boundary-value-problem",
    "divergence-theorem",
    "hilbert-space",
    "self-adjoint-operator",
    "fourier-transform",
    "eigenvalues-and-eigenvectors",
    "manifold",
    "metric-tensor",
    "finite-difference-method",
    "graph-theory",
    "graph-laplacian"
  ],
  "sources": [
    {
      "title": "Lectures on PDE",
      "url": "https://math.stanford.edu/~lms/lecs-on-pde.pdf"
    },
    {
      "title": "Math 220B Lecture Notes",
      "url": "https://web.stanford.edu/class/math220b/lecturenotes.html"
    },
    {
      "title": "155, Differential Analysis I, Fall 2021",
      "url": "https://math.mit.edu/~dyatlov/18.155-F21/"
    },
    {
      "title": "Lecture Notes for 18.155",
      "url": "https://math.mit.edu/~dyatlov/18.155/155-notes.pdf"
    },
    {
      "title": "303 Fall 2009",
      "url": "https://math.mit.edu/classes/18.303/index.html"
    },
    {
      "title": "Spectral Theory of Partial Differential Equations — Lecture Notes",
      "url": "https://arxiv.org/abs/1203.2344"
    },
    {
      "title": "Lecture 11: The Laplacian in Polar Coordinates",
      "url": "https://ocw.mit.edu/courses/18-152-introduction-to-partial-differential-equations-fall-2005/resources/lecture11/"
    },
    {
      "title": "MIT Mathematics: 18.085 Fall 2010",
      "url": "https://math.mit.edu/classes/18.085/2010fall.html"
    },
    {
      "title": "Spectral Graph Theory, Lecture 2: The Laplacian",
      "url": "https://www.cs.yale.edu/homes/spielman/561/lect02-15.pdf"
    },
    {
      "title": "Spectral Graph Theory, Fall 2019: Syllabus",
      "url": "https://cs.yale.edu/homes/spielman/462/"
    }
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}