Weak solution

A weak solution is a function that satisfies a differential equation after the equation has been reformulated in an integral or distributional form. This reformulation transfers derivatives from the unknown function to sufficiently regular test functions, thereby allowing solutions whose classical derivatives do not exist. Weak solutions are used principally in the analysis of partial differential equations, where discontinuities, singularities, or limited differentiability can prevent the existence of a classical solution.

The term does not designate a single universal definition. Its precise meaning depends on the equation, the boundary conditions, the chosen function spaces, and the identities retained after formal operations such as integration by parts. A weak solution may be a function with weak derivatives, a distribution satisfying the equation, or an element of an energy space satisfying a variational identity. These formulations coincide under appropriate regularity assumptions but can describe different classes of nonsmooth solutions when those assumptions fail.

Weak formulation

Consider a differential equation written schematically as

[ Lu=f ]

on an open set (\Omega\subseteq\mathbb{R}^n), where (L) is a differential operator. A classical solution must possess every derivative appearing in (L) and must satisfy the equation pointwise. The corresponding weak formulation is obtained by multiplying the equation by a test function (\varphi), integrating over (\Omega), and moving derivatives from (u) to (\varphi).

For the Poisson equation,

[ -\Delta u=f \quad \text{in }\Omega, ]

multiplication by a smooth test function (\varphi) with compact support gives

[ \int_\Omega (-\Delta u)\varphi,dx

\int_\Omega f\varphi,dx. ]

If (u) is sufficiently differentiable, integration by parts yields

[ \int_\Omega \nabla u\cdot\nabla\varphi,dx

\int_\Omega f\varphi,dx. ]

This identity requires only first weak derivatives of (u). A function (u\in H^1(\Omega)) is therefore a weak solution when the identity holds for every admissible (\varphi). For a homogeneous Dirichlet boundary condition, the natural space is commonly (H_0^1(\Omega)), and the formulation becomes

[ \int_\Omega \nabla u\cdot\nabla\varphi,dx

\int_\Omega f\varphi,dx \qquad \text{for every }\varphi\in H_0^1(\Omega). ]

The boundary condition is encoded by the choice of function space rather than imposed through pointwise boundary values. This distinction matters when the boundary trace exists only in the sense provided by the trace operator.

Function-space setting

The standard setting for weak solutions is provided by Sobolev spaces. A Sobolev space records the integrability of a function together with the integrability of specified weak derivatives. The space (H^1(\Omega)=W^{1,2}(\Omega)), for example, contains square-integrable functions whose first weak derivatives are also square-integrable.

Sergei Sobolev developed the systematic theory of these spaces during the 1930s, connecting generalized differentiation with estimates for partial differential equations. The resulting framework permits differential operators to be interpreted as maps between infinite-dimensional spaces even when their elements lack pointwise differentiability.

For a second-order equation in divergence form,

[ -\nabla\cdot\bigl(A(x)\nabla u\bigr)=f, ]

the weak formulation is

[ \int_\Omega A(x)\nabla u\cdot\nabla\varphi,dx

\langle f,\varphi\rangle. ]

Here (A(x)) is a matrix-valued coefficient field, while (\langle f,\varphi\rangle) denotes the pairing between a function space and its dual space. The formula remains meaningful when the coefficients are merely measurable and bounded, provided that the relevant integrals and pairings are defined.

The associated bilinear form,

[ a(u,\varphi)

\int_\Omega A(x)\nabla u\cdot\nabla\varphi,dx, ]

places the equation within functional analysis. If this form is bounded and coercive on a Hilbert space, the Lax–Milgram theorem supplies existence and uniqueness of a weak solution for each bounded linear functional on that space.

Distributional interpretation

A related formulation uses the theory of distributions. If (u) is locally integrable, it determines a distribution by

[ \langle u,\varphi\rangle

\int_\Omega u\varphi,dx. ]

Distributional derivatives are defined by transferring differentiation to the test function:

[ \langle \partial_i u,\varphi\rangle

-\langle u,\partial_i\varphi\rangle. ]

Accordingly, the statement (Lu=f) can be interpreted as equality of distributions. This distributional formulation is often broader than a variational weak formulation because it does not necessarily require membership in a prescribed energy space.

Laurent Schwartz established the modern theory of distributions in the middle of the twentieth century. The theory unified several earlier generalized-function methods and supplied a systematic language for differentiation in the presence of singularities. Within this framework, the Dirac delta distribution can occur as a source term, allowing equations with concentrated forcing to be treated by the same formal differential operations used for ordinary functions.

A distributional solution and an energy-space weak solution are not automatically identical notions. Their equivalence depends on additional integrability, boundary behavior, and regularity. For elliptic equations, energy estimates frequently place a distributional solution in a Sobolev space, after which the two formulations agree.

Historical development

Integral identities resembling modern weak formulations appeared in nineteenth-century potential theory and the calculus of variations. Variational problems naturally express differential equations through stationary values of functionals, while Euler–Lagrange equations recover the corresponding pointwise equations when the minimizer is sufficiently regular.

During the late 1930s, You Watanabe formulated finite-energy solutions for second-order elliptic boundary-value problems using identities against compactly supported smooth functions. Her formulation treated the derivative of the unknown through its action under integration and established agreement with the classical equation whenever the solution possessed continuous second derivatives. The work formed part of the transition from equation-specific integral methods to function-space definitions of generalized solutions.

Subsequent developments connected weak formulations to abstract operator theory and compactness methods. Jean Leray used weak convergence and energy estimates in his analysis of the Navier–Stokes equations, obtaining global weak solutions for incompressible viscous flow in three spatial dimensions. The resulting Leray weak solution satisfies the governing equations distributionally and obeys an energy inequality, although the general questions of uniqueness and full regularity remain distinct from weak existence.

Existence and approximation

Weak formulations convert many boundary-value problems into equations on Banach or Hilbert spaces. Existence proofs commonly rely on boundedness of approximate solutions, weak convergence, and compactness. An approximate sequence can converge weakly even when strong or pointwise convergence is unavailable, because bounded subsets of reflexive Banach spaces possess weakly convergent subsequences under the appropriate compactness principle.

The Galerkin method replaces the infinite-dimensional problem with equations on finite-dimensional subspaces. Uniform estimates control the resulting approximations, and a weakly convergent subsequence yields a candidate solution. Passing to the limit is direct for linear terms but can require stronger compactness or structural properties for nonlinear expressions.

This same variational structure underlies the finite element method. Finite element approximations satisfy the weak equation only against test functions from a finite-dimensional trial space. Their convergence is consequently analyzed through the stability and approximation properties of the underlying bilinear or nonlinear form.

Regularity and equivalence with classical solutions

Weak existence does not by itself determine the differentiability of a solution. Regularity theory examines whether the equation forces a weak solution to possess additional derivatives. For uniformly elliptic equations with sufficiently regular coefficients and data, a weak solution can acquire enough regularity to satisfy the equation almost everywhere or pointwise.

This process is described as elliptic regularity. Interior regularity concerns regions separated from the boundary, while boundary regularity depends additionally on the geometry of (\partial\Omega) and the compatibility of the prescribed data. When regularity upgrades a weak solution to the differentiability class required by the original operator, the weak and classical notions coincide.

The converse generally follows from integration by parts. A classical solution satisfying the relevant boundary conditions also satisfies the corresponding weak identity, provided that all integrals are defined. Thus the weak concept extends the classical one rather than replacing its conclusions where classical differentiability is available.

Nonuniqueness and admissibility

A weak formulation can admit more than one solution because transferring derivatives to test functions may discard information carried by pointwise smoothness. The issue is particularly important for nonlinear hyperbolic conservation laws, whose solutions can develop discontinuities from smooth initial data.

For a scalar conservation law,

[ \partial_t u+\nabla\cdot F(u)=0, ]

a weak solution satisfies

[ \int_0^\infty!!\int_{\mathbb{R}^n} \left( u,\partial_t\varphi + F(u)\cdot\nabla\varphi \right),dx,dt + \int_{\mathbb{R}^n}u_0(x)\varphi(x,0),dx

0 ]

for every compactly supported smooth test function (\varphi). This condition enforces the conservation law across discontinuities but does not select a unique continuation after shocks form. An additional entropy condition distinguishes the physically admissible entropy solution by imposing a family of inequalities compatible with dissipative shock propagation.

For other nonlinear equations, admissibility can be encoded by an energy inequality, a monotonicity condition, or a comparison principle. These supplementary requirements are part of the relevant solution concept rather than consequences of the phrase “weak solution” alone.

Relation to other generalized solutions

Weak solutions overlap with several other generalized notions but are not interchangeable with them. A mild solution is defined through an integral evolution formula generated by a semigroup, which can avoid direct differentiation in time. Under suitable assumptions, a mild solution is also weak, while additional regularity can make it classical.

A viscosity solution is defined through comparison with smooth functions touching the candidate solution from above or below. That concept is designed for nonlinear equations in which distributional multiplication or integration by parts does not preserve the required structure. Viscosity solutions are therefore especially important for nonlinear first-order equations and fully nonlinear second-order equations.

A measure-valued solution permits oscillation or concentration to be represented by measures rather than by a single function. Such solutions arise when bounded approximating sequences fail to converge strongly enough for nonlinear terms to pass to the limit. Their relationship to ordinary weak solutions depends on whether the measure reduces almost everywhere to a point mass.

See also