9.3.1 Definition and Existence
Given a linear operator on an inner product space, its adjoint is the operator obtained by moving it from one argument of the inner product to the other. Throughout this section is a finite-dimensional inner product space.
Theorem 9.3.1 (Existence and uniqueness of the adjoint). Let be a linear operator on a finite-dimensional inner product space. There is a unique map such that and is linear.
Proof. Existence. Fix . The map is a linear functional on , being the composite of with the linear functional . By the Riesz representation theorem, there exists a unique with for all . Define .
Linearity. For , , and all : By the uniqueness in the Riesz representation theorem, .
Uniqueness. If also satisfies for all , then for all , so . ∎
Definition 9.3.2 (Adjoint operator). Let be a linear operator on a finite-dimensional inner product space. The adjoint of is the operator of the theorem, characterized by Taking complex conjugates of both sides (and renaming the variables), this is equivalent to for all .
Note. The adjoint is the dual map transported by the inner product. Let , , be the bijection of the Riesz representation theorem, and let , , be the dual map (Definition 6.2.9). The defining property says so Over , is linear and is simply the transpose read through the isomorphism . Over , is conjugate-linear, and is the complex conjugate of the transpose: in an orthonormal basis with dual basis we have , so if has matrix , then has matrix and has matrix (see Theorem 9.3 below).
9.3.2 Properties of the Adjoint
Proposition 9.3.3 (Algebraic properties of the adjoint). Let be linear operators on a finite-dimensional inner product space, and . Then:
- .
- .
- .
- .
- .
- If is invertible, then is invertible and .
Proof. Each follows from the defining equation and the uniqueness of the adjoint.
(i) . So satisfies the defining property of .
(ii) .
(iii) .
(iv) .
(v) .
(vi) Applying (iv) and (v) to gives . ∎
Note. Property (iv) shows that taking adjoints reverses the order of products, just like taking transposes or inverses. So is a conjugate-linear map that reverses products and is its own inverse.
Matrix of the Adjoint
Theorem 9.3.4 (Matrix of the adjoint). Let be an orthonormal basis of , and let be the matrix of in this basis. Then the matrix of is where denotes the conjugate transpose.
Proof. The -entry of is the coefficient of in , which is (coordinates in an orthonormal basis). Likewise the -entry of is So ; over the bar has no effect. ∎
Note. The orthonormality of the basis matters: in a general basis, the matrix of is not .
Proposition 9.3.5 (Kernel and image of the adjoint). For :
- .
- .
Proof. (i) Since ,
(ii) Apply (i) to : , i.e., . Taking orthogonal complements and using Corollary 9.1.32: . ∎
Corollary 9.3.6 (Rank equality). .
Proof. , by Lemma 9.1.30 and rank–nullity. ∎
The following lemma is the key to all the structure theorems below: it lets us split off an invariant subspace and continue on its orthogonal complement.
Lemma 9.3.7 (Invariant subspaces and the adjoint). Let and let be a subspace.
- If is -invariant, then is -invariant.
- If is invariant under both and , then, for the restricted inner product on , .
Proof. (i) Let and . Then , so . Thus .
(ii) The restriction of to is an inner product, and maps to . For , , so satisfies the defining property of the adjoint of ; by uniqueness, . ∎