Let T:D(T)⊆H→KT:\mathcal D(T)\subseteq H\to K be a between complex , with inner products linear in the first variable. The adjoint T∗T^* has domain

D(T∗)={y∈K:∃z∈H such that ⟨Tx,y⟩K=⟨x,z⟩H for all x∈D(T)},\mathcal D(T^*)= \left\{y\in K:\exists z\in H\ \text{such that}\ \langle Tx,y\rangle_K=\langle x,z\rangle_H \ \text{for all }x\in\mathcal D(T)\right\},

and T∗y=zT^*y=z. Density of D(T)\mathcal D(T) makes zz unique. Equivalently, y∈D(T∗)y\in\mathcal D(T^*) exactly when the functional x↦⟨Tx,y⟩Kx\mapsto\langle Tx,y\rangle_K is bounded in the HH-norm; the then supplies T∗yT^*y.

Closedness and closability

The adjoint T∗T^* is always closed, although its domain need not be dense. The original operator TT is closable exactly when D(T∗)\mathcal D(T^*) is dense in KK. In that case its satisfies

T‾=T∗∗.\overline T=T^{**}.

These identities include equality of domains as well as equality of the operator actions.

Domain-sensitive algebra

Adjoints reverse products only with domain qualifications. For densely defined SS and TT such that STST is densely defined, one generally has

(ST)∗⊇T∗S∗,(ST)^*\supseteq T^*S^*,

and equality requires additional hypotheses. Likewise, the formal integration-by-parts expression for a differential operator does not by itself determine the adjoint: boundary terms and the chosen domain determine D(T∗)\mathcal D(T^*).

Symmetry and self-adjointness

For T:D(T)⊆H→HT:\mathcal D(T)\subseteq H\to H, symmetry means T⊆T∗T\subseteq T^*, including D(T)⊆D(T∗)\mathcal D(T)\subseteq\mathcal D(T^*). Self-adjointness requires the stronger equality T=T∗T=T^*, especially D(T)=D(T∗)\mathcal D(T)=\mathcal D(T^*). A symmetric differential operator on a small test domain can therefore fail to be self-adjoint even when its formal expression is real.

References