Definition

Let T:D(T)XXT:\mathcal D(T)\subseteq X\to X be a densely defined on a , and suppose its ρ(T)\rho(T) is nonempty. The operator TT has compact resolvent if

(TλI)1:XX(T-\lambda I)^{-1}:X\longrightarrow X

is a for some λρ(T)\lambda\in\rho(T). The then implies compactness for every λρ(T)\lambda\in\rho(T), so the definition does not depend on the chosen resolvent point. Compact resolvent is stronger than closedness: it says that solving (TλI)x=y(T-\lambda I)x=y sends bounded sets of data to relatively compact sets in XX.

Compact domain embedding

Give D(T)\mathcal D(T) the xT=x+Tx\|x\|_T=\|x\|+\|Tx\|. When ρ(T)\rho(T)\neq\varnothing, TT has compact resolvent exactly when the inclusion

(D(T),T)X(\mathcal D(T),\|\cdot\|_T)\hookrightarrow X

is compact. This reformulation often proves compactness through an embedding theorem rather than by constructing the resolvent explicitly.

Spectral consequences

Every point of the of a compact-resolvent operator is an isolated eigenvalue of finite algebraic multiplicity, and the spectrum has no finite accumulation point. The spectrum may be empty for a general non-self-adjoint unbounded operator. If it is infinite, its eigenvalues can accumulate only at infinity.

Self-adjoint case and examples

For a DD on a , compact resolvent is equivalent to compactness of (1+D2)1/2(1+D^2)^{-1/2}. The spectral theorem then provides an of eigenvectors, with eigenvalues of finite multiplicity whose tend to infinity. on compact manifolds, subject to suitable elliptic boundary conditions when a boundary is present, are standard examples.

References