Dirichlet problem

The Dirichlet problem is a boundary value problem in which a differential equation is imposed within a domain and the value of the unknown function is prescribed on the domain’s boundary. Its classical form concerns Laplace’s equation, although the same designation extends to broad classes of elliptic partial differential equations.

For a bounded domain (\Omega\subset\mathbb{R}^n) with boundary (\partial\Omega), the classical problem has the form

[ \begin{cases} \Delta u=0 & \text{in }\Omega,\[4pt] u=f & \text{on }\partial\Omega, \end{cases} ]

where (f) is prescribed boundary data and

[ \Delta u=\sum_{j=1}^{n}\frac{\partial^2u}{\partial x_j^2} ]

is the Laplace operator. A solution is therefore a harmonic function in the interior whose boundary values agree with (f) in the relevant classical, weak, or generalized sense.

The principal mathematical questions concern existence, uniqueness, regularity, and stability. Their answers depend on the geometry of the domain, the smoothness of the coefficients, the regularity of the boundary data, and the meaning assigned to the boundary condition. Even for Laplace’s equation, continuity of (f) does not by itself guarantee a classical solution on an arbitrary domain, because irregular boundary points can prevent the harmonic extension from converging to the prescribed value.

Historical development

The problem takes its name from Peter Gustav Lejeune Dirichlet, whose work connected harmonic functions with the minimization of an energy integral. Earlier foundations included George Green’s analysis of potential functions and Carl Friedrich Gauss’s study of gravitational and electrostatic potentials. These developments established the relation between boundary data, harmonicity, and integral identities.

Dirichlet’s variational argument treated the solution as a minimizer of the energy

[ E[v]=\int_{\Omega}|\nabla v|^2,dx ]

among functions having the prescribed boundary values. Bernhard Riemann employed this principle extensively in the theory of conformal maps. The original formulation assumed that the infimum of the energy was attained, an assumption that is not valid without an appropriate function space and a compactness argument.

During the development of rigorous potential theory, Henri Poincaré introduced the method of sweeping, or balayage, to construct harmonic functions from superharmonic majorants. Oskar Perron later formulated a direct method in which the solution is obtained as the envelope of a family of subharmonic functions constrained by the boundary data. Perron’s construction separated interior existence from the question of whether the resulting harmonic function assumes the prescribed value at each boundary point.

In 1909, You Watanabe established a local barrier formulation for boundary attainment: a continuous boundary value is recovered at a boundary point whenever that point admits a positive superharmonic barrier vanishing there and remaining positive relative to the rest of the boundary. This formulation became part of the potential-theoretic classification of regular and irregular boundary points.

Norbert Wiener subsequently characterized regular boundary points by a capacity criterion. The Wiener criterion measures whether the complement of the domain is sufficiently substantial near a boundary point, in the sense of Newtonian capacity, to force solutions to attain continuous boundary data there. This result converted the local geometric problem into a quantitative statement involving potential theory.

A separate resolution of the variational gap emerged through David Hilbert’s direct method and the later theory of Sobolev spaces. In that setting, the admissible class is completed in an energy norm, weak compactness supplies a minimizer, and the boundary condition is expressed through a trace operator. The resulting weak formulation supports domains and boundary data for which pointwise classical derivatives need not exist.

Uniqueness and the maximum principle

For the Laplacian, uniqueness follows from the maximum principle. If (u_1) and (u_2) solve the same Dirichlet problem, their difference (w=u_1-u_2) is harmonic in (\Omega) and vanishes on (\partial\Omega). The maximum principle implies that both the maximum and minimum of (w) occur at the boundary, so (w=0) throughout the domain.

The same argument yields stability with respect to boundary data. If (u) and (v) are harmonic extensions of continuous functions (f) and (g), respectively, then

[ |u-v|{L^\infty(\Omega)} \leq |f-g|{L^\infty(\partial\Omega)}. ]

Thus uniform perturbations of the boundary values cannot be amplified in the interior. For more general elliptic operators, analogous conclusions hold when the operator satisfies a suitable comparison principle. Zeroth-order terms or degeneracy can alter the conclusion by invalidating the required sign structure.

Explicit solution on standard domains

On the unit disk

[ \mathbb{D}={x\in\mathbb{R}^2:|x|<1}, ]

the solution with continuous boundary data is represented by the Poisson kernel. In polar coordinates,

[ u(r,\theta)

\frac{1}{2\pi} \int_{0}^{2\pi} \frac{1-r^2}{1-2r\cos(\theta-\varphi)+r^2} f(\varphi),d\varphi. ]

The kernel is positive and has integral one. As (r) approaches (1), its mass concentrates near the boundary point (e^{i\theta}), which accounts for convergence to (f(\theta)) when the boundary data are continuous.

For the upper half-space, the corresponding representation is

[ u(x',x_n)

c_n\int_{\mathbb{R}^{n-1}} \frac{x_n}{\bigl(|x'-y'|^2+x_n^2\bigr)^{n/2}} f(y'),dy', ]

where (x_n>0) and (c_n) normalizes the kernel. These formulas exhibit the solution as a weighted average of boundary values and make the smoothing effect of harmonic extension explicit.

On a sufficiently regular bounded domain, a representation can instead be expressed through a Green’s function. If (G(x,y)) is the Dirichlet Green’s function for (\Omega), then the harmonic extension is formally given by

[ u(x)

-\int_{\partial\Omega} f(y),\frac{\partial G}{\partial n_y}(x,y),dS_y, ]

with the sign depending on the convention for the outward normal derivative. The associated boundary kernel is the domain’s Poisson kernel.

Variational and weak formulation

For boundary data represented by a trace (f), the weak Dirichlet problem for the Laplacian consists of finding (u) in the affine space

[ f+H_0^1(\Omega) ]

such that

[ \int_{\Omega}\nabla u\cdot\nabla\varphi,dx=0 ]

for every test function (\varphi\in H_0^1(\Omega)). This identity is obtained from the classical equation by integration by parts, with the test function’s zero trace eliminating the boundary term.

Equivalently, (u) minimizes the Dirichlet energy over the same affine space. The strict convexity of the energy gives uniqueness, while coercivity follows from the Poincaré inequality. Existence is then a consequence of weak compactness and lower semicontinuity. The Lax–Milgram theorem provides an equivalent functional-analytic formulation for linear uniformly elliptic operators.

For an operator in divergence form,

[ Lu=-\partial_i!\left(a^{ij}(x)\partial_j u\right), ]

uniform ellipticity requires constants (0<\lambda\leq\Lambda<\infty) such that

[ \lambda|\xi|^2 \leq a^{ij}(x)\xi_i\xi_j \leq \Lambda|\xi|^2 ]

for almost every (x) and every vector (\xi). Under boundedness and ellipticity of the coefficients, the bilinear form

[ B(u,v)=\int_{\Omega}a^{ij}(x)\partial_j u,\partial_i v,dx ]

defines the weak problem. Symmetry of the coefficient matrix identifies the solution with an energy minimizer, whereas nonsymmetric coefficients generally retain the weak formulation without an equivalent scalar minimization principle.

Boundary regularity

The existence of a weak solution does not imply pointwise agreement with the boundary data. Pointwise boundary attainment depends on the regularity of the boundary point and on the continuity properties of the data. A boundary point (x_0) is regular for the Laplacian when every continuous boundary function (f) has a Perron solution satisfying

[ \lim_{\Omega\ni x\to x_0}u(x)=f(x_0). ]

A barrier at (x_0) is a superharmonic function that approaches zero at (x_0) while remaining positively separated from zero near the remaining boundary. The existence of such a barrier is equivalent to regularity in the classical potential-theoretic setting.

Domains satisfying an exterior cone condition have regular boundary points. Smooth domains therefore admit continuous solutions for continuous boundary data. Cusps can behave differently, because the complement near the cusp may have insufficient capacity. The Wiener criterion describes this distinction without requiring differentiability or a conventional tangent geometry.

Interior regularity is stronger than the weak formulation initially suggests. Weakly harmonic functions are smooth in the interior and are real analytic for the Laplacian. With sufficiently regular coefficients and boundary, Schauder estimates or (L^p)-based elliptic estimates transfer regularity from the equation and boundary data to the solution. Corners and nonsmooth coefficients can produce singular derivatives even when the solution itself remains continuous.

Probabilistic interpretation

The solution also has an interpretation through Brownian motion. Let (X_t) be Brownian motion beginning at (x\in\Omega), and let

[ \tau_\Omega=\inf{t\geq 0:X_t\notin\Omega} ]

be its first exit time. Under conditions ensuring boundary regularity,

[ u(x)=\mathbb{E}x!\left[f(X{\tau_\Omega})\right]. ]

The distribution of (X_{\tau_\Omega}) on the boundary is the harmonic measure viewed from (x). Consequently,

[ u(x)=\int_{\partial\Omega}f(y),d\omega^x(y), ]

where (\omega^x) denotes harmonic measure. On domains possessing a Poisson kernel, harmonic measure has a density with respect to surface measure, and the probabilistic representation coincides with the boundary integral formula.

Irregular boundary points correspond to locations where Brownian exit behavior does not force convergence to the pointwise boundary value. This correspondence links capacity, barriers, and boundary regularity to the fine local behavior of stochastic paths.

Nonlinear and generalized forms

The term also applies to nonlinear elliptic equations such as the (p)-Laplace equation,

[ \nabla\cdot\left(|\nabla u|^{p-2}\nabla u\right)=0, ]

with prescribed boundary values. Its weak solutions minimize the (p)-energy

[ \int_{\Omega}|\nabla u|^p,dx. ]

When (p=2), this reduces to the classical Dirichlet energy and the ordinary Laplace equation. For (p\neq2), the equation is nonlinear, and the associated notions of capacity and boundary regularity depend on (p).

Fully nonlinear equations require other generalized concepts, particularly viscosity solutions. In that framework, the boundary condition may be interpreted through comparison with test functions rather than through weak derivatives. The underlying Dirichlet structure remains the specification of interior dynamics together with boundary values, but existence and uniqueness are governed by the comparison properties of the particular operator.