Bounded Operators on a Hilbert Space
[40/0%]The completion of an inner product space is a Hilbert space. Make this statement more precise, and prove it. (See the proof of Theorem 12.40 for an application).
Suppose is a positive integer, , and . Prove that every Hilbert space inner product satisfies the identities
and
Generalize this: Which functions and measures on a set give rise to the identity
Assume and are in the closed unit ball of , and as . Prove that then .
Assume weakly, and . Prove that then .
Let be the dual space of ; define by
(See Theorem 12.5). Prove that is a Hilbert space, relative to the inner product
If satisfies for all and , prove that is an isomorphism of onto whose existence implies that is reflexive.
Suppose is a sequence of unit vectors in (that is ), and assume that
If is any sequence of scalars, prove that
and deduce that the following three properties of are equivalent to each other:
.
converges, in the norm of .
converges, for every .
This generalizes Theorem 12.6.
Suppose is a resolution of the identity, as in Section 12.17, and prove that
for all , and .
Suppose is unitary, and . Prove that scalars can be chosen so that
if is a proper subset of the unit circle, but that this norm is never less than 1 if covers the whole circle.
Prove Theorem 12.35 with in place of .
Suppose is the polar decomposition of an invertible . Prove that is normal if and only if .
Prove that every normal invertible is the exponential of some normal .
Suppose is normal, and is invertible. Prove that is normal if and only if commutes with .
Suppose and are normal, and . Prove that and are normal.
If, in addition, and (see Theorem 12.32), prove that and .
Show, however, that there exist and such that is not even normal (of course, then ). In fact, such examples exist if .
If is normal, show that , for some unitary . When is unique?
Assume and is a compact operator. Show that is then compact.
Find a noncompact such that . Can such an operator be normal?
Suppose is normal, and is a finite set. Deduce as much information about from this as you can.
Show, under the hypotheses of of Theorem 12.29, that the equation has a solution if and only if
(If for one , then must be 0, for this ).
The spectrum of can be divided into three disjoint pieces:
The point spectrum consists of all for which is not one-to-one.
The continuous spectrum consists of all such that is a one-to-one mapping of onto a dense proper subspace of .
The residual spectrum consists of all other .
Prove that every normal has empty residual spectrum.
Prove that the point spectrum of a normal is at most countable, if is separable.
Let and be the right and left shifts (as defined in Exercise 2 of Chapter 10), acting on the Hilbert space .
Prove that and that
Let and be as above. Prove that neither nor has polar decompositions , with unitary and .
Let be a positive measure on a measure space , let , with the usual inner product
For , define the multiplication operator by . Then .
Under what conditions on does have eigenvalues? Give an example in which . Show that every is normal. What is the relation between and the essential range of ? Show that is an isometric *-isomorphism of onto a closed subalgebra of . (Certain pathological measures have to be excluded in order to make this last statement correct). Is a maximal commutative sub-algebra of ? Hint: If and for all , and if , show that is a multiplication by and hence that .
Suppose is normal, is the closed subalgebra of generated by , , and , and can be approximated, in the norm topology of , by finite linear combinations of projections that belong to .
Under what (necessary and sufficient) conditions on does this happen?
Does every normal have a square root in ? What can you say about the cardinality of the set of all square roots of ? Can it happen that two square roots of the same do not commute? Can this happen when ?
Show that the Fourier transform is a unitary operator on . What is its spectrum? Suggestion: When , compute the Fourier transforms of
Show that any two infinite-dimensional separable Hilbert spaces are isometrically isomorphic (via countable orthonormal bases; see [23]). Show that the space in Theorem 12.38 is separable. Show that the answer to the question that precedes Theorem 12.38 is therefore negative for every infinite-dimensional , separable or not.
Suppose is normal, is a bounded Borel function on , and . If and are the spectral decompositions of and , respectively, prove that
for every Borel set .
If and , the notation means that , that is, that
for all . Prove the equivalence of the following four properties of a pair of self-adjoint projections and :
.
.
.
.
If is a resolution of the identity, it follows that if .
Suppose * is an involution in a complex algebra is an invertible element of such that and is defined by
for every . Show that is an involution in .
Let be the algebra of all complex 4 -by-4 matrices. If , let be the conjugate transpose of . Put
As in Exercise 27, define
Show that and are normal, with respect to the involution , that , but that .
Show that is not -normal.
Compare with .
Compute the spectral radius ; show that it is different from
Define so that otherwise. Compute ; it does not lie in .
Part (a) shows that Theorem 12.16 fails for some involutions. Part (b) does the same for part (a) of Exercise 12; (c), (d), and (e) show that various parts of Theorem 11.28 fail for the involution .
Let be the vector space of all trigonometric polynomials on the real line: these are functions of the form
where and , for . Show that
exists and defines an inner product on , that
and that the completion of is a nonseparable Hilbert space . Show that contains all uniform limits of trigonometric polynomials; these are the so-called " almost-periodic" functions on .
Let be an infinite-dimensional Hilbert space, with its weak topology. Prove that the inner product is a separately continuous function on which is not jointly continuous.
Assume for , and
for every . Does it follow that
for every ?
Let be a uniformly convex Banach space. This means, by definition, that the assumptions
imply that . For example, every Hilbert space is uniformly convex.
Prove that Theorem 12.3 holds in .
Assume , and . Prove that is a Cauchy sequence (in the norm-topology of ). Hint: Consider .
Show that the preceding four properties fail in certain Banach spaces (for instance, in , or in ). These are therefore not uniformly convex.
Assume that weakly and . Prove that . Hint: Reduce to the case . Consider , for a suitable .
Prove the assertion about the case made in the remark that follows Theorem 12.35.
Find an operator , with , which is neither unitary nor selfadjoint.
If is self-adjoint and , show that is unitary. Deduce from this, and from the fact that lies on the unit circle, that lies on the real axis.
Show that if is normal. Hint: Using Theorem 12.35, .
Define on by . Show that is self-adjoint and that is a dense proper subspace of .
Find a nonnormal such that
(This shows that Theorem 12.25 has no converse).
Show that and can have the same null space without being normal.
Let be a circular annulus in with center at 0. Define and as in Theorem 12.38. Prove that has no square root in by completing the following outline: Assume . Put . Since , induction shows that for all integers . It follows from the Laurent series expansion that for every . This leads to , i.e., for all , an impossibility.
Find the adjoint of . (Use Laurent series).