# Elliptic operator

A solution to Laplace's equation defined on an annulus. The Laplace operator is the most famous example of an elliptic operator.

In the theory of partial differential equations, elliptic operators are differential operators that generalize the Laplace operator. They are defined by the condition that the coefficients of the highest-order derivatives be positive, which implies the key property that the principal symbol is invertible, or equivalently that there are no real characteristic directions.

Elliptic operators are typical of potential theory, and they appear frequently in electrostatics and continuum mechanics. Elliptic regularity implies that their solutions tend to be smooth functions (if the coefficients in the operator are smooth). Steady-state solutions to hyperbolic and parabolic equations generally solve elliptic equations.

## Definitions

A linear differential operator L of order m on a domain $\Omega$ in Rd given by

$Lu = \sum_{|\alpha| \le m} a_\alpha(x)\partial^\alpha u\,$

(where $\alpha = (\alpha_1, ..., \alpha_n)$ is a multi-index, and $\partial^\alpha u = \partial^{\alpha_1} \cdots \partial^{\alpha_n} u$) is called elliptic if for every x in $\Omega$ and every non-zero $\xi$ in Rd,

$\sum_{|\alpha| = m} a_\alpha(x)\xi^\alpha \neq 0,\,$

where $\xi^\alpha = \xi_1^{\alpha_1} \cdots \xi_n^{\alpha_n}$.

In many applications, this condition is not strong enough, and instead a uniform ellipticity condition may be imposed for operators of degree m = 2k:

$(-1)^k\sum_{|\alpha| = 2k} a_\alpha(x) \xi^\alpha > C |\xi|^{2k},\,$

where C is a positive constant. Note that ellipticity only depends on the highest-order terms.[1]

A nonlinear operator

$L(u) = F(x, u, (\partial^\alpha u)_{|\alpha| \le 2k})\,$

is elliptic if its first-order Taylor expansion with respect to u and its derivatives about any point is a linear elliptic operator.

Example 1
The negative of the Laplacian in Rd given by
$- \Delta u = -\sum_{i=1}^d \partial_i^2u\,$
is a uniformly elliptic operator. The Laplace operator occurs frequently in electrostatics. If ρ is the charge density within some region Ω, the potential Φ must satisfy the equation
$- \Delta \Phi = 4\pi\rho.\,$
Example 2
Given a matrix-valued function A(x) which is symmetric and positive definite for every x, having components aij, the operator
$Lu = -\partial_i(a^{ij}(x)\partial_ju) + b^j(x)\partial_ju + cu\,$
is elliptic. This is the most general form of a second-order divergence form linear elliptic differential operator. The Laplace operator is obtained by taking A = I. These operators also occur in electrostatics in polarized media.
Example 3
For p a non-negative number, the p-Laplacian is a nonlinear elliptic operator defined by
$L(u) = -\sum_{i = 1}^d\partial_i (|\nabla u|^{p - 2}\partial_i u).\,$
A similar nonlinear operator occurs in glacier mechanics. The Cauchy stress tensor of ice, according to Glen's flow law, is given by
$\tau_{ij} = B\left(\sum_{k,l = 1}^3(\partial_lu_k)^2\right)^{-\frac{1}{3}}\cdot\frac{1}{2}(\partial_ju_i + \partial_iu_j)\,$
for some constant B. The velocity of an ice sheet in steady state will then solve the nonlinear elliptic system
$\sum_{j = 1}^3\partial_j\tau_{ij} + \rho g_i - \partial_ip = Q,\,$
where ρ is the ice density, g is the gravitational acceleration vector, p is the pressure and Q is a forcing term.

## Elliptic regularity theorem

Let L be an elliptic operator of order 2k with coefficients having 2k continuous derivatives. The Dirichlet problem for L is to find a function u, given a function f and some appropriate boundary values, such that Lu = f and such that u has the appropriate boundary values and normal derivatives. The existence theory for elliptic operators, using Gårding's inequality and the Lax–Milgram lemma, only guarantees that a weak solution u exists in the Sobolev space Hk.

This situation is ultimately unsatisfactory, as the weak solution u might not have enough derivatives for the expression Lu to even make sense.

The elliptic regularity theorem guarantees that, provided f is square-integrable, u will in fact have 2k square-integrable weak derivatives. In particular, if f is infinitely-often differentiable, then so is u.

Any differential operator exhibiting this property is called a hypoelliptic operator; thus, every elliptic operator is hypoelliptic. The property also means that every fundamental solution of an elliptic operator is infinitely differentiable in any neighborhood not containing 0.

As an application, suppose a function $f$ satisfies the Cauchy-Riemann equations. Since the Cauchy-Riemann equations form an elliptic operator, it follows that $f$ is smooth.

## General definition

Let $D$ be a (possibly nonlinear) differential operator between vector bundles of any rank. Take its principal symbol $\sigma_\xi(D)$ with respect to a one-form $\xi$. (Basically, what we are doing is replacing the highest order covariant derivatives $\nabla$ by vector fields $\xi$.)

We say $D$ is weakly elliptic if $\sigma_\xi(D)$ is a linear isomorphism for every non-zero $\xi$.

We say $D$ is (uniformly) strongly elliptic if for some constant $c>0$,

$([\sigma_\xi(D)](v),v) \geq c\|v\|^2$

for all $\|\xi\|=1$ and all $v$. It is important to note that the definition of ellipticity in the previous part of the article is strong ellipticity. Here $(\cdot,\cdot)$ is an inner product. Notice that the $\xi$ are covector fields or one-forms, but the $v$ are elements of the vector bundle upon which $D$ acts.

The quintessential example of a (strongly) elliptic operator is the Laplacian (or its negative, depending upon convention). It is not hard to see that $D$ needs to be of even order for strong ellipticity to even be an option. Otherwise, just consider plugging in both $\xi$ and its negative. On the other hand, a weakly elliptic first-order operator, such as the Dirac operator can square to become a strongly elliptic operator, such as the Laplacian. The composition of weakly elliptic operators is weakly elliptic.

Weak ellipticity is nevertheless strong enough for the Fredholm alternative, Schauder estimates, and the Atiyah–Singer index theorem. On the other hand, we need strong ellipticity for the maximum principle, and to guarantee that the eigenvalues are discrete, and their only limit point is infinity.