ΘρϵηΠατπ

Hellinger-Toeplitz
If a linear operator T is defined on every element of a complex hilbert space H and T is self-adjoint then T is bounded

Suppose for the sake of contradiction that T is unbounded, therefore for each N∈ℕ1 we obtain an element of xN∈dom⁡(f) such that ‖TxN‖>N‖xN‖, therefore we can construct a sequence (yn) such that ‖yn‖=1 and ‖Tyn‖→∞

We now construct a sequence of functionals of the form fn(x)=⟨Tx,yn⟩=⟨x,Tyn⟩ where we note that each such fn is defined on all of H and is linear and moreover for each n by the Schwarz inequality we have that: |fn(x)|=|⟨x,Tyn⟩|≤‖Tyn‖‖x‖ which shows that the functional fn is bounded as a linear operator.

Additionally for any x∈H the real sequence (fn(x)) is also bounded because we have: |fn(x)|=|⟨Tx,yn⟩|≤‖Tx‖ therefore there exists some c∈ℝ+ such that ‖fn‖op⁡≤c for every n∈ℕ1 therefore for each x∈H we have that |fn(x)|≤‖fn‖op⁡‖x‖≤c‖x‖ thus taking x=Tyn we have that ‖Tyn‖2=⟨Tyn,Tyn⟩=|fn(Tyn)|≤c‖Tyn‖ therefore ‖Tyn‖≤c which is a contradiction because we assumed that ‖Tyn‖→∞

The above implies that given any unbounded linear operator it will never be defined on all elements of a hilbert space.

Extension of a Linear Operator
Suppose that S,T are linear operators, then we say that T is an extension of S if
  • dom⁡(S)⊆dom⁡(T)
  • S=T↾dom⁡(S)
Densely Defined Linear Operator
We say that a linear operator T is densely defined in H if dom⁡(T) is dense in H
Hilbert Adjoint Operator
Let T:X→H be a linear opeartor (possibly unbounded) denslely defined linear operator in a complex hilbert where H is a complex hilbert space, then the Hilbert adjoint operator T∗:Y→H of T where Y={y∈H:∃y∗∈H st ∀x∈H,⟨Tx,y⟩=⟨x,y∗⟩} then T∗(y) is defined to be equal to y∗
Adjoint Operators Respect the Subset Relation
Let S:X→H and T:Y→Y be densely defined linear operators in a complex hilbert space H then if S⊆T we have T∗⊆S∗

By the definition of T∗ we have that for all a∈Y,b∈dom⁡(T∗) that α:⟨Ta,b⟩=⟨a,T∗b⟩ and since S⊆T then also that for every s∈dom⁡(S) we have β:⟨Ss,b⟩=⟨s,T∗b⟩

Now we want to prove that dom⁡(T∗)⊆dom⁡(S∗). But recall by definition dom⁡(S∗)={y∈H:∃y∗∈H st ∀x∈H,⟨Sx,y⟩=⟨x,y∗⟩} Now the b's such that β holds must be a subset of dom⁡(S) because it is an even more specific representation, ie y∗ must exist and be of the form T∗y which is more restrictive, so we have that dom⁡(T∗)⊆dom⁡(S∗).

Finally we note that for any y∈dom⁡(T∗) we have that S∗y=T∗y by chaining α and β therefore we have that T∗⊆S∗

Double Adjoint Is a Superset
Suppose that T:X→H is a linear operator which is densely defined in a complex hilbert space H then if dom⁡(T∗) is dense in H then T⊆T∗∗
We know that for all x∈X and y∈dom⁡(T∗) that we have α:⟨Tx,y⟩=⟨x,T∗y⟩ therefore by taking complex conjugates that is: β:⟨T∗x,y⟩=⟨y,Tx⟩ since dom⁡(T∗) was dense then we know that T∗∗ exists and by definition we have that for all a∈dom⁡(T∗) and all b∈dom⁡(T∗∗) thus using α and β in the same way as the above prove we conclude that dom⁡(T)⊆dom⁡(T∗∗) and that T∗∗x=Tx for any x∈dom⁡(T), that is T⊆T∗∗
If a Linear Operator Is Injective and Has Dense Image Then the Adjoint Has an Inverse
Suppose that T:X→H is a densely defined linear operator in a dense hilbert space H, if T is injective and im⁡(T) is dense in H then T∗ is injective and (T∗)−1=(T−1)∗

We know that T∗ exists because T is densely defined in H, also that T−1 exists because T is injective. Also im⁡(T)=dom⁡(T−1) is dense in H therefore (T−1)∗ exists.

We must now show that (T∗)−1 exists and satisfies the equation as stated in the outset. So let y∈dom⁡(T∗), then for every x∈dom⁡(T−1) we have that T−1(x)∈dom⁡(T) satisfying α:⟨T−1x,T∗y⟩=⟨TT−1x,y⟩=⟨x,y⟩ then by the definition of the hilbert adjoint operator of T−1 we have for every x∈dom⁡(T−1) ⟨T−1x,T∗y⟩=⟨x,(T−1)∗T∗y⟩ so we conlcude that T∗y∈dom⁡((T−1)∗) combined with α we conclude that β:(T−1)∗T∗y=y when T∗y=0 we have that y=0 thus (T∗)−1:im⁡(T∗)→dom⁡(T∗) exists. Since (T∗)−1T∗ is the identity operator on dom⁡(T∗) then β implies that (T∗)−1⊆(T−1)∗

To complete the proof we must now show that (T∗)−1⊇(T−1)∗, so note that for any x∈dom⁡(T) and y∈dom⁡((T−1)∗), then Tx∈im⁡(T)=dom⁡(T−1) and ⟨Tx,(T−1)∗y⟩=⟨T−1Tx,y⟩=⟨x,y⟩ but also by the definition of the hilbert adjoint operator of T we note for each x∈dom⁡(T) ⟨Tx,(T−1)∗y⟩=⟨x,T∗(T−1)y⟩ so we have (T−1)∗y∈dom⁡(T∗) and that γ:T∗(T−1)∗y=y for every y∈dom⁡((T−1)∗).

Now we know that T∗(T∗)−1 is the identity operator on dom⁡((T∗)−1)=im⁡(T∗) and (T∗)−1:im⁡(T∗)→dom⁡(T∗) is surjective, so by γ we have that dom⁡((T∗)−1)⊇dom⁡((T−1)∗) to conclude that (T∗)−1⊇(T−1)∗ so we are done.