Definition
Core of a closed operator
A subdomain whose graph closure recovers the entire closed operator.
Definition
Let be a closed linear operator on a Banach space. A linear subspace is a core for if the closure of the restricted operator is . Equivalently, is dense in for the graph norm
Thus every admits with both and in . Density of in alone is not enough.
Graph interpretation
The graph of is
The core condition says that its closure is exactly the graph of . Consequently, values of on the smaller test domain determine the closed operator uniquely. This is why differential operators are often first computed on smooth compactly supported functions and then recovered by graph closure.
How cores are used
Suppose is a closable operator with and on . Then is a core for exactly when . In particular, proving that a proposed test domain is a core justifies checking identities and approximation arguments there before passing to the full domain Schmüdgen, Chapter 1.
Warning: ambient density versus graph density
For an unbounded operator, convergence does not control . A dense invariant subspace can therefore fail to be a core. Likewise, “essential domain” is sometimes used as a synonym for a core, but in discussions of symmetric operators the nearby phrase “essentially self-adjoint” adds a separate assertion about the closure being self-adjoint.
References
- Tosio Kato, Perturbation Theory for Linear Operators, 2nd ed., Springer, 1995. Publisher record. Relevant: Chapter III on closability, graph norms, and cores.
- Konrad Schmüdgen, Unbounded Self-adjoint Operators on Hilbert Space, Springer, 2012. Publisher record. Relevant: Chapter 1 on closed operators and core domains.