Banach Algebras and Spectral Theory
[113/0%]Use the identity to prove that and always have the same spectral radius.
If and are invertible in , prove that is invertible.
If and are invertible in , prove that and are invertible. (The commutative case of this was tacitly used in the proofs of Theorems 10.13 and 10.28).
Show that it is possible to have but . For example, consider the right and left shifts and , defined on some Banach space of functions on the nonnegative integers by
If and , show that is a nontrivial idempotent. (This means that ).
Prove that every finite-dimensional is isomorphic to an algebra of matrices. Hint: The proof of Theorem 10.2 shows that every is isomorphic to a subalgebra of . Conclude that implies if .
Prove that is invertible in if is invertible. Suggestion: Put , write as a geometric series (assume ), and use the identity stated in Exercise 1 to obtain a finite formula for in terms of . Then show that this formula works without any restrictions on or .
If , and , prove that . Thus . Show, however, that is not always equal to .
Let and be the algebras of all complex 2-by-2 matrices of the form
Prove that every two-dimensional complex algebra with unit is isomorphic to one of these and that is not isomorphic to . Hint: Show that has a basis in which for some . Distinguish between the cases . Show that there exists a three-dimensional noncommutative Banach algebra.
Let be the algebra of all complex functions on which are 0 except at finitely many points, with pointwise addition and multiplication and norm
Show that multiplication is left-continuous (hence also right-continuous, since is commutative) but not jointly continuous. (Adjunction of a unit element, as suggested in Section 10.1, would make no difference). Show, in fact, that there is a sequence in so that but , as .
Let be the space of all complex functions on [0, 1] whose second derivative is continuous. Choose , and define
Show that this makes into a Banach space, for every choice of , but that the Banach algebra axioms hold if and only if . (For necessity, consider and ).
What happens if the process of adjoining a unit (described in Section 10.1) is applied to an algebra which already has a unit? Clearly, the result cannot be an algebra with two units. Explain.
Suppose that is open in and that and are holomorphic. Prove that is holomorphic.
This was used in the proof of Theorem 10.13, with .
Another proof that can be based on Liouville's theorem 3.32 and the fact that as . Complete the details.
Call a topological divisor of zero if there is a sequence in , with , such that
Prove that every boundary point of the set of all invertible elements of is a topological divisor of zero. Hint: Take , where .
In which Banach algebras is 0 the only topological divisor of 0?
Find the spectrum of the operator given by
Suppose . Put . Let be the smallest closed subalgebra of that contains 1 and . Let be the smallest closed subalgebra of that contains and . Describe the spectra and .
Fubini's theorem was applied to vector-valued integrals in the proof of Theorem 10.29. Justify this.
Construct a proof of Theorem 10.29 that uses no contour integrals, as follows: Prove the theorem first for polynomials , then for rational functions with no poles in , and obtain the general case from Runge's theorem.
Suppose is a Banach space, is compact, and for all . Prove that the point spectrum of is not empty.
Let and define by
Show that . Hence for every , but if , then 0 ; hence
The extra hypotheses in (c) and (d) of Theorem 10.33 are thus needed.
Suppose that the spectrum of some is not connected. Prove that contains a nontrivial idempotent 2. (This is defined in Exercise 2).
Prove also that , where
Suppose is open in is an isolated boundary point of is a holomorphic -valued function (where is some complex Banach space), and there is a smallest nonnegative integer such that
is bounded as . If , then is said to have a pole of order at .
Suppose and has a pole at every point of . Prove that there is a nontrivial polynomial such that .
As a special case of , assume and has a pole of order at 0. Prove that .
In part (a): this can happen only when is a finite set.
Let be the right shift, acting on , as in Exercise 2. Let be a sequence of complex numbers such that but as . Define by
and define by .
Compute , for .
Show that .
Show that has no eigenvalue. (Its point spectrum is therefore empty, although its spectrum consists of a single point!)
Show that does not have a pole at 0.
Show that is a compact operator.
Suppose , and . Suppose is an open set in that contains a component of . Prove that intersects for all sufficiently large . (This strengthens Theorem 10.20). Hint: If , where is an open set disjoint from , consider the function that is 1 in in .
Let be the algebra of all real continuous functions on , with the supremum norm. This satisfies all requirements of a Banach algebra, except that the scalars are now real.
If , then , and if is invertible in , but is not multiplicative.
If and are defined in as in Theorem 10.34, show that is a group of order 2.
The analogues of Theorem 10.9 and (d) of Theorem 10.34 are thus false for real scalars. Exactly where would the proof of (d) of Theorem 10.34 break down?
Suppose , the algebra of all continuous complex functions on the unit circle , with the supremum norm. Show that two invertible members of are in the same coset of if and only if they are homotopic mappings of into the set of all nonzero complex numbers. Deduce from this that is isomorphic to the additive group of the integers. (The notation is as in Theorem 10.34).
Suppose , the convolution algebra of all complex Borel measures on the real line; see (e) of Example 10.3. Supply the details in the following proof that is uncountable: If , let be the unit mass concentrated at . Assume . Then for some ; hence, for ,
where is an integer. Since is a bounded function, . Thus is the only in . No coset of in contains therefore more than one .
Prove that is commutative if there is a constant such that for all and in A. Hint: if is invertible in . Replace by , where and . Continue as in Theorem 12.16.
Prove that is commutative if for every . Hint: Show that . Use Exercise 1 to deduce that . Continue as in (a).
As regards the invariant subspace problem, described in the introduction to Theorem 10.35, explain why the problem is
trivial in ,
different in ,
uninteresting if is not separable.
How should Lomonosov's theorem be reformulated when ?
Let be the right shift on , as in Exercise 2. Prove that 0 is the only compact that commutes with . Hint: If then
does not tend to 0 when .
Prove Proposition 11.2.
State and prove an analogue of Wiener's lemma 11.6 for power series that converge absolutely on the closed unit disc.
If is a compact Hausdorff space, show that there is a natural one-to-one correspondence between closed subsets of and closed ideals of .
Prove that the polynomials are dense in the polydisc algebra . (See Theorem 11.7). Suggestion: If , and is defined by , then is the sum of an absolutely (hence uniformly) convergent multiple power series on .
Suppose is a commutative Banach algebra, , and is holomorphic in some open set that contains the range of . Prove that there exists such that , that is, such that for every complex homomorphism of . Prove that is uniquely determined by and if is semisimple.
Suppose and are commutative Banach algebras, is semisimple, is a homomorphism whose range is dense in , and is defined by
Prove that is a homeomorphism of onto a compact subset of . [The fact that is dense in implies that is one-to-one and that the topology of is the weak topology induced by the Gelfand transforms of the elements , for .
Let be the disc algebra, let , where is an arc in the unit disc, and let be the restriction mapping of into . This example shows that may be a proper subset of , even if is one-to-one.
Find an example in which but .
In Example it was asserted that . Find several proofs of this.
Which properties of Lebesgue measure are used in Example 11.13(f)? Can Lebesgue measure be replaced by any positive measure, without changing any of the results?
Supply the details for the last paragraph in Example 11.13(f).
Let be the algebra of all continuously differentiable complex functions on the unit interval , with pointwise multiplication, normed by
Show that is a semisimple commutative Banach algebra. Find its maximal ideal space.
Fix ; let be the set of all for which . Show that is a closed ideal in and that is a two-dimensional algebra which has a one-dimensional radical. (This gives an example of a semisimple algebra with a quotient algebra that is not semisimple). To which of the two algebras described in Exercise 5 of Chapter 10 is isomorphic?
Let be the disc algebra. Associate to each a function by the formula
Then is an involution on .
Does this involution turn into a -algebra?
Does always lie in the real axis?
Which complex homomorphisms of are positive functionals, with respect to this involution?
If is a positive finite Borel measure on , then
is a positive functional on . Are there any others?
Show that commuting idempotents have distance . Explicitly, if , for some and in a Banach algebra, then either or . Show that this may fail if .
If for some and in a Banach algebra, prove that
Let be a large positive number, and define a norm on by
Let be the algebra of all complex 2-by-2 matrices, with the corresponding operator norm:
For , let be the conjugate transpose of . Consider a fixed , namely,
Prove the following statements.
; hence .
.
.
.
Therefore commutativity is required in Theorem 11.23 and in Exercise 12.
If is the sum of the four entries in , for , then is a positive functional on .
The equality [see (e) of Theorem 11.31] does not hold, because and , so that .
If is the set of all positive functionals on that satisfy (as in Theorem 11.33), then has many extreme points, although 0 is the only multiplicative linear functional on . Commutativity is therefore required in the implication of Theorem 11.33.
A complex function , defined on , is said to be positive-definite if
for every choice of , in and for every choice of complex numbers .
Show that for every .
Show that the Fourier transform of every finite positive Borel measure on is positive-definite.
Complete the following outline of the converse of (b) (Bochner's theorem): If is continuous and positive-definite, then is the Fourier transform of a finite positive Borel measure.
Let be the convolution algebra , with a unit attached, as described in of Section 10.3 and of Section 11.13. Define . Show that
is an involution on and that
is a positive functional on . By Theorem 11.32 and (e) of Section 11.13, there is a positive measure on the one-point compactification of , such that
If is the restriction of to , it follows that
for every . Hence . (Actually, is already concentrated on , so that ).
Let be the set of all continuous positive-definite functions on that satisfy . Find all extreme points of this convex set.
Let be the maximal ideal space of a commutative Banach algebra . Call a closed set an -boundary if the maximum of on equals its maximum on , for every . (Trivially, is an -boundary).
Prove that the intersection of all -boundaries is an -boundary. is called the Shilov boundary of . The terminology is suggested by the maximum modulus property of holomorphic functions. For instance, when is the disc algebra, then is the unit circle, which is the topological boundary of , the closed unit disc.
Outline of proof: Show first that there is an -boundary which is minimal in the sense that no proper subset of is an -boundary. (Partially order the collection of -boundaries by set inclusion, etc). Then pick , pick with , and put
Since is minimal, there exists with and on . If and is sufficiently large, then on , for all . Hence . Conclude from this first that only in , hence that intersects every -boundary , and finally that . Thus , and .
Suppose is a Banach algebra, is an integer, , and
for every . Show that there exist constants , for , such that
(This extends Theorem 11.12).
Suppose are positive numbers such that and
for all integers and . Let be the set of all complex functions on the integers for which the norm
is finite. Define multiplication in by
Show that each is a commutative Banach algebra.
Show that exists and is finite, by showing that .
Show similarly that exists and that .
Put . Show that can be identified with the maximal ideal space of and that the Gelfand transforms are absolutely convergent Laurent series on .
Consider the following choices for :
-
.
-
.
-
if if .
-
.
-
if if . For which of these is a circle? For which choices is self-adjoint, in the sense that is closed under complex conjugation?
Is always semisimple?
Is there an , with the unit circle, such that consists entirely of infinitely differentiable functions?
Let be the maximal ideal space of , as in Section 11.13. Show that
has no isolated point, and
contains no convergent sequence of distinct points. Hint: If , are distinct points in , none of which is a limit point of the others, and if is any bounded sequence of numbers, then there exist pairwise disjoint open sets in such that , and there exists a function such that on .
Let be as above and prove: If and in the weak topology of , then for every . Show, by constructing an example, that the converse is false.
Prove the following partial converse of Theorem 11.31: If is a bounded linear functional on a -algebra , and , then is positive.
Suggestion: Choose , put ,
for . Use Theorem 11.28 to show that and that therefore
Proceed as in Lemma 5.26.
In , let consist of all points , and let consist of the points . Of these circles, show that is polynomially convex but is not. How about ?
Show that a 3-by-3 matrix commutes with
Deduce from this that centralizers (see Section 11.21) need not be commutative.
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).
The associative law has been used freely throughout this chapter. Prove it. Prove also that implies and .
Let be a densely defined operator in . Prove that has a closed extension if and only if is dense in . In that case, prove that is an extension of .
By Theorem for a densely defined operator in if and only if is dense in . Show that this can actually happen.
Suggestion: Let be an orthonormal basis of ; let be a dense subset of ; define ; and extend linearly to , the set of all finite linear combinations of the basis vectors . Show that the graph of this is dense in .
Suppose is a densely defined, closed operator in , and . Does it follow that is normal?
Suppose is a densely defined operator in , and for every . Does it follow that for every ?
If is an operator in , define
If is dense, prove that
If is also closed, prove that
This generalizes Theorem 12.10.
Consider the following three boundary value problems. The differential equation is
where is given. The choices of boundary conditions are
(i) .
(ii) .
(iii) and .
Show that each of these problems has a unique solution such that is absolutely continuous and . Hint: Combine Example 13.4 with Theorem 13.13.
Do this also by solving the problems explicitly.
Prove the self-adjointness of the operator in , defined by , with consisting of all absolutely continuous such that .
Hint: You may need to know that as for every . Prove this. Or prove more, namely, that every is the Fourier transform of an -function.
Fix . Use Theorem 13.13 to prove that the equation
has a unique absolutely continuous solution , which has , , and absolutely continuous.
Prove also, by direct calculation, that
This solution can also be found by means of Fourier transforms.
Let be the space of all holomorphic functions in the open unit disc that satisfy
Show that is a Hilbert space which is isomorphic to via the one-to-one correspondence .
Define by . Show that is the Cayley transform of the symmetric operator in , given by
Find the ranges of and of ; show that one is and one has codimension 1. (Compare with Example 13.21).
With as in Exercise 9, define now by
Show that is an isometry which is the Cayley transform of a closed symmetric operator in , whose deficiency indices are 0 and .
Prove part (c) of Lemma 13.18.
In the context of Theorem 13.24, how are the operators and related?
If and are measurable and is bounded, prove that maps into .
Prove that if and only if a.e. [E], that is, if and only if
Is the operator that occurs in the proof of Theorem 13.33 normal?
Prove that every normal operator in , bounded or not, has a polar decomposition
where is unitary, is self-adjoint, . Moreover, .
Prove the following extension of Theorem 12.16: If , if and are normal operators in , and if , then also .
Suppose is a closed operator in , and for every . Prove that is normal. Hint: Begin by proving that
Prove that the spectrum of any operator in is a closed subset of . (See Definition 13.26). Hint: If , and , then is a bounded inverse of , for small .
Put . Define , where , by
so that , etc. (Note that is presented in its polar decomposition ).
Find . Compute that
Conclude that is one-to-one, that is dense in , and that . Define , with domain , by
Prove that is empty.
Let be as in Example 13.4, put
and define for all .
Prove the following assertions:
Every is in the point spectrum of .
consists of the numbers , where runs through the integers; each of these is in the point spectrum of .
has codimension 1 for every . Hence . The point spectrum of is empty.
is empty.
Hint: Study the differential equation . This illustrates how sensitive the spectrum of a differential operator is to its domain (in this case, to the boundary conditions that are imposed).
Show that every nonempty closed subset of is the spectrum of some normal operator in (if ).
Define , where , by
Prove that each is unitary.
Prove that satisfies the conditions stated in Definition 13.34.
If is the infinitesimal generator of , prove that if and only if (where is the Fourier transform of ) and that for all .
Prove that is the imaginary axis. More precisely, show that is one-to-one for every , that lies in the resolvent set of if and only if is not pure imaginary, and that the range of is a proper dense subspace of if is pure imaginary.
Hint: if and only if and also is in .
If (see Exercise 9) and , define
Show that each is self-adjoint (and positive). Find the infinitesimal generator of the semigroup . Is self-adjoint? Show that has pure point spectrum, at the points .
For , define
and put . Show that satisfies the conditions imposed in Definition 13.34 and that for all .
Find the domain of the infinitesimal generator of , and prove that
where is the Hilbert transform (Chapter 7, Exercise 24). Prove that is positive and self-adjoint.
The integral represents a harmonic function in the upper half-plane, with boundary values . The semigroup property of can be deduced from this, as well as from a look at the Fourier transforms of the functions .
Show that every isometric operator in has a closed isometric extension.
On the other hand, show that some symmetric operators in have no closed symmetric extension by completing the following outline.
Let be an orthonormal basis of . Let be the set of all finite sums , subject to the condition . Prove that is a dense subspace of . Define by
and let be the restriction of to . Then is an isometry, with , and is one-to-one on . Hence is the Cayley transform of a symmetric operator . Any closed symmetric extension of would correspond to a closed isometric extension of , with one-to-one. But is dense in ; hence has only one closed isometric extension, namely , and is not one-to-one.