Definition

Let T:D(T)XYT:D(T)\subseteq X\to Y be a linear operator between . Its graph is the

G(T)={(x,Tx):xD(T)}X×Y.\mathcal G(T)=\{(x,Tx):x\in D(T)\}\subseteq X\times Y.

The graph records both the action of TT and its domain: the first-coordinate projection maps G(T)\mathcal G(T) bijectively onto D(T)D(T), and TT is recovered by following its inverse with the second-coordinate projection. For a , density concerns D(T)D(T), while closedness and closability concern G(T)\mathcal G(T) inside the .

Closedness and closability

If XX and YY are , TT is closed exactly when G(T)\mathcal G(T) is closed in X×YX\times Y. Equivalently, whenever xnxx_n\to x and TxnyTx_n\to y, one has xD(T)x\in D(T) and Tx=yTx=y. The operator is closable exactly when the closure of G(T)\mathcal G(T) is itself the graph of an operator; that operator is the closure T\overline T.

Graph norm

When XX and YY are normed spaces, the on D(T)D(T) may be defined by

xT=xX+TxY.\|x\|_T=\|x\|_X+\|Tx\|_Y.

The map x(x,Tx)x\mapsto(x,Tx) identifies this normed space with the graph equipped with the corresponding product norm. If XX and YY are Banach spaces, TT is closed precisely when D(T)D(T) is complete in the graph norm.

Why the domain matters

The same formula can define different operators when assigned different domains, and their graphs then differ. For example, differentiation on L2L^2 has distinct closed realizations determined by boundary conditions. Graph language makes this domain dependence explicit and is therefore essential for unbounded operators.

References
  1. Tosio Kato, Perturbation Theory for Linear Operators, 2nd ed., Springer, 1976. DOI record. Relevant: Chapter III on closed operators and their graphs.
  2. Konrad Schmüdgen, Unbounded Self-adjoint Operators on Hilbert Space, Springer, 2012. DOI record. Relevant: Chapter 1 on domains, graphs, and closures.