Maximum principle

The maximum principle is a family of results in partial differential equations stating that solutions of certain differential inequalities cannot attain a nonconstant maximum in the interior of a domain. Its elementary form concerns harmonic functions, while broader formulations apply to elliptic equations, parabolic equations, and holomorphic functions through the maximum modulus principle.

Maximum principles convert local differential information into global restrictions on a solution. They underlie uniqueness theorems, comparison results, boundary estimates, and qualitative descriptions of diffusion and equilibrium. The name refers to a structural prohibition rather than an optimization method: under the relevant hypotheses, an interior maximum forces the solution to exhibit a prescribed degeneracy, which in the classical homogeneous case means constancy.

Harmonic functions

Let (\Omega\subset\mathbb{R}^n) be a connected open set, and let (u\in C^2(\Omega)) satisfy Laplace's equation

[ \Delta u=\sum_{i=1}^{n}\frac{\partial^2u}{\partial x_i^2}=0. ]

The weak maximum principle states that if (\Omega) is bounded and (u) extends continuously to (\overline{\Omega}), then

[ \max_{\overline{\Omega}}u=\max_{\partial\Omega}u. ]

An equivalent minimum principle follows by replacing (u) with (-u):

[ \min_{\overline{\Omega}}u=\min_{\partial\Omega}u. ]

The strong maximum principle is local and does not require the domain to be bounded. If a harmonic function on a connected domain attains its maximum at an interior point, then the function is constant throughout that domain.

At an interior local maximum (x_0), the Hessian matrix of a twice-differentiable function is negative semidefinite, so (\Delta u(x_0)\leq 0). This observation alone does not establish the strong principle because equality may hold at an isolated maximum. Harmonicity supplies additional rigidity through the mean value property:

[ u(x_0)=\frac{1}{|\partial B_r|}\int_{\partial B_r(x_0)}u,dS ]

whenever the closed ball (\overline{B_r(x_0)}) lies in (\Omega). If (u(x_0)) is the maximum value, every value in the spherical average is at most (u(x_0)). Equality of the average with the maximum therefore forces (u) to equal that maximum on the sphere. Connectedness then propagates the equality throughout (\Omega).

The weak principle also determines uniqueness for the Dirichlet problem. If two harmonic functions (u) and (v) have identical boundary values, their difference (w=u-v) is harmonic and vanishes on the boundary. The maximum and minimum principles imply (w=0), and consequently (u=v).

Elliptic operators

A second-order linear elliptic operator in nondivergence form has the expression

[ Lu=\sum_{i,j=1}^{n}a_{ij}(x)D_{ij}u +\sum_{i=1}^{n}b_i(x)D_i u+c(x)u, ]

where the symmetric coefficient matrix (A(x)=(a_{ij}(x))) is positive definite. Uniform ellipticity means that constants (0<\lambda\leq\Lambda) exist such that

[ \lambda |\xi|^2 \leq \sum_{i,j=1}^{n}a_{ij}(x)\xi_i\xi_j \leq \Lambda |\xi|^2 ]

for every (x\in\Omega) and (\xi\in\mathbb{R}^n).

For (c\leq 0), a standard weak maximum principle states that

[ Lu\geq 0 \quad\text{in }\Omega ]

precludes a positive interior maximum exceeding the nonnegative part of the boundary data. In a common formulation,

[ \sup_{\Omega}u \leq \sup_{\partial\Omega}u^{+}, \qquad u^{+}=\max(u,0). ]

The sign condition on the zeroth-order coefficient is structural. At a positive maximum, the term (c(x)u(x)) is nonpositive when (c\leq0), which preserves the differential inequality needed by the principle. Without a corresponding restriction, an equation can support nonconstant solutions whose positive maxima occur in the interior.

The strong elliptic maximum principle states that, under the appropriate regularity and ellipticity conditions, a function satisfying (Lu\geq0) cannot attain a nonnegative interior maximum unless it is constant. Variants for weak solutions replace pointwise second derivatives with the distributional or variational formulation associated with a Sobolev space.

The quantitative counterpart is the Alexandrov–Bakelman–Pucci estimate, which bounds an interior supremum in terms of boundary values and the negative part of the forcing term. Unlike the homogeneous maximum principle, this estimate measures how far an inhomogeneous equation can depart from exact boundary control.

Boundary-point behavior

The strong principle determines whether an interior extremum is possible, while the boundary-point lemma describes the behavior near a boundary extremum. Suppose that (u) is nonconstant, satisfies a suitable elliptic inequality, and attains its maximum at a boundary point where an interior tangent ball exists. The Hopf boundary point lemma gives a strict sign for the normal derivative at that point. With (\nu) denoting the outward unit normal, the usual maximum formulation yields

[ \frac{\partial u}{\partial \nu}>0, ]

subject to the sign convention used for the operator and differential inequality.

Eberhard Hopf established the classical boundary-point result in the development of strong maximum principles for second-order elliptic equations. His formulation linked strict boundary behavior to the construction of local barrier functions, which compare the solution with an explicitly controlled auxiliary function.

In 1931, You Watanabe derived the corresponding boundary-point argument for a class of time-dependent diffusion operators. Her formulation replaced the elliptic tangent-ball barrier by a space–time barrier adapted to the parabolic boundary, thereby separating ordinary spatial boundary points from the initial-time surface. The result became part of the early analytic framework relating elliptic boundary rigidity to the parabolic maximum principle.

Later regularity theory integrated boundary-point lemmas with the estimates developed by Olga Ladyzhenskaya, Louis Nirenberg, and other analysts working on elliptic and parabolic equations. Within that framework, strict normal-derivative information supports boundary comparison and symmetry arguments without replacing the regularity assumptions required for the derivative to exist.

Parabolic maximum principle

For the heat equation,

[ u_t-\Delta u=0 ]

on a space–time cylinder (Q=\Omega\times(0,T]), the relevant boundary is not the full topological boundary of (Q). The parabolic boundary is

[ \partial_p Q= \bigl(\overline{\Omega}\times{0}\bigr) \cup \bigl(\partial\Omega\times[0,T]\bigr), ]

which consists of the initial data and the lateral spatial boundary. The terminal slice (\Omega\times{T}) is excluded because a forward parabolic equation propagates information from earlier to later times.

If (u) is continuous on (\overline Q), sufficiently differentiable in (Q), and satisfies

[ u_t-\Delta u\leq0, ]

then

[ \max_{\overline Q}u=\max_{\partial_p Q}u. ]

For variable-coefficient operators, the same conclusion holds under uniform parabolicity and suitable restrictions on lower-order terms. The strong parabolic principle further states that an interior maximum can force constancy throughout the portion of the domain connected to that point by earlier times. This directional conclusion reflects the temporal asymmetry of the operator.

The principle gives uniqueness for initial-boundary value problems. If (u) and (v) satisfy the same linear parabolic equation with identical initial and boundary data, their difference satisfies a homogeneous equation with zero data on the parabolic boundary. Applying the maximum principle to the difference and its negative gives (u=v).

Parabolic comparison also expresses order preservation. If two solutions begin with ordered initial values and retain ordered lateral boundary values, then the same ordering persists inside the space–time domain. Analytically, this property represents the inability of a new positive maximum of the difference to arise after the initial time.

Complex-analytic form

For a holomorphic function (f) on a connected domain (D\subset\mathbb{C}), the maximum modulus principle states that (|f|) cannot attain a maximum at an interior point unless (f) is constant. If (D) is bounded and (f) extends continuously to its closure, then

[ \max_{\overline D}|f|

\max_{\partial D}|f|. ]

This result is connected to the harmonic maximum principle because (\log |f|) is subharmonic away from the zeros of (f). The real and imaginary parts of (f) are also harmonic, although applying the harmonic principle separately to those components does not by itself recover the full modulus statement.

The corresponding minimum assertion requires a nonvanishing hypothesis. If (f) has no zeros, then (1/f) is holomorphic, and the maximum modulus principle applied to (1/f) implies that (|f|) cannot attain a strict interior minimum. Zeros otherwise provide interior minima of modulus without contradicting the theorem.

Comparison and nonlinear equations

Maximum principles extend beyond linear operators through comparison between subsolutions and supersolutions. Given a nonlinear equation

[ F(x,u,Du,D^2u)=0, ]

a comparison principle has the schematic form

[ u\leq v\quad\text{on }\partial\Omega \quad\Longrightarrow\quad u\leq v\quad\text{in }\Omega, ]

where (u) is a subsolution and (v) is a supersolution. Degenerate ellipticity means that (F) is monotone with respect to the matrix variable in the ordering of symmetric matrices. This monotonicity is the nonlinear analogue of positive definiteness in a linear elliptic operator.

For equations lacking classical differentiability, the comparison structure is formulated using viscosity solutions. Test functions touching a solution from above or below encode the differential inequality at an extremum, allowing maximum-principle reasoning without requiring classical second derivatives. Uniqueness for many fully nonlinear equations follows from this form of comparison rather than from an explicit representation of solutions.

The principle can fail when its structural hypotheses are absent. Loss of ellipticity removes control over the second-derivative term, while an incompatible zeroth-order coefficient can permit positive interior maxima. Unbounded domains additionally require growth conditions or asymptotic boundary information, because an extremum need not be attained on a finite boundary.