Section III.5

Hilbert Spaces

[41/49%]
Show
LaTeX
Problem 462
?
(a)

Prove that the correspondence L→L~L\to \widetilde L constructed in 4.1 of Chapter III of the Theory part defines a covariant functor from the category of pre-Hilbert spaces to the category of Hilbert spaces.

(b)

Show that L~\widetilde L is a universal object in a suitable category.

Here a pre-Hilbert space may have a positive semidefinite scalar product. Set L0={x∈L:⟨x,x⟩=0}L_0=\left\{ x\in L: \left\langle x, x \right\rangle =0\right\}, give L/L0L/L_0 the induced scalar product, and let L~\widetilde L be its completion in the norm ∥x∥=⟨x,x⟩1/2\left\| x\right\| = \left\langle x, x \right\rangle^{1/2}. The scalar product extends continuously to L~\widetilde L.

Problem 463

Prove that the system of functions en(x)=e2πinxe_n(x)=e^{2\pi inx}, n∈Zn\in \mathbb {Z}, is a Hilbert basis in L2((0,1))L^{2}\left({(0,1)}\right).

?
Problem 464

Apply the orthogonalization process to the sequence of monomials 1,x,x2,…1,x,x^2,\ldots in the following Hilbert spaces:

?
(a)

L2([−1,1],dx)L^{2}\left({[-1,1],dx}\right);

(b)

L2([−1,1],dx1−x2)L^{2}\left({[-1,1],\frac{dx}{\sqrt{1-x^2}}}\right);

(c)

L2[[0,∞),e−x dx]L^{2}[{[0,\infty ),e^{-x}\, dx}];

(d)

L2((−∞,∞),e−x2 dx)L^{2}\left({(-\infty ,\infty ),e^{-x^2}\, dx}\right).

Problem 465

Apply the orthogonalization process to the sequence of monomials 1,z,z2,…1,z,z^2,\ldots in the pre-Hilbert spaces of polynomials with the following scalar products:

?
(a)

⟨P,Q⟩=∬∣z∣≤RP(z)Q(z)‾ dx dy\displaystyle \left\langle P, Q \right\rangle =\iint_{\left|z\right|\le R}P(z)\overline{Q(z)}\, dx\, dy;

(b)

⟨P,Q⟩=∬CP(z)Q(z)‾e−∣z∣2 dx dy\displaystyle \left\langle P, Q \right\rangle =\iint_{\mathbb {C}}P(z)\overline{Q(z)}e^{-\left|z\right|^2}\, dx\, dy.

Problem 466

Express the linear functional Fx(f)=f(x)F_x(f)=f(x) in the spaces of Problem 465 in the form of a scalar product.

?
Problem 467
[NO TASK]
Problem 468

Prove that the orthogonal complement of the system of functions en(x)=e2πinxe_n(x)=e^{2\pi inx}, n∈Zn\in \mathbb {Z}, in the space L2([a,b],dx)L^{2}\left({[a,b],dx}\right)

?
(a)

is equal to zero for ∣b−a∣≤1\left|b-a\right|\le 1;

(b)

is different from zero for ∣b−a∣>1\left|b-a\right|>1.

Problem 469

Let B0B_0 be the space of all trigonometric polynomials (i.e., finite linear combinations of the functions eiλxe^{i\lambda x}, λ∈R\lambda \in \mathbb {R}). Define in B0B_0 the scalar product

⟨f,g⟩=lim⁡A→+∞12A∫−AAf(x)g(x)‾ dx. \left\langle f, g \right\rangle =\lim _{A\to +\infty }\frac1{2A}\int _{-A}^A f(x)\overline{g(x)}\, dx.
?
(a)

Prove that the corresponding Hilbert space BB is nonseparable and has the continuum Hilbert basis {eiλx}λ∈R\left\{ e^{i\lambda x}\right\}_{\lambda \in \mathbb {R}}.

(b)

Prove that BB contains the space of almost periodic continuous functions, i.e., the closure of B0B_0 in the uniform norm ∥f∥=sup⁡x∈R∣f(x)∣\left\| f\right\| =\sup_{x\in \mathbb {R}}\left|f(x)\right|, but does not coincide with it.

Problem 470

Let μ\mu be the measure on the line defined by the formula μ(A)=card⁡A\mu (A)=\operatorname {card}A (the number of points in AA). Prove that L2(R,μ)L^{2}\left({\mathbb {R},\mu }\right) is nonseparable and isomorphic to the space of Problem 469.

?
Problem 471

Prove that L2([a,b])L^{2}\left({[a,b]}\right) contains complete orthonormal systems consisting of:

?
(a)

polynomials;

(b)

step functions;

(c)

trigonometric polynomials;

(d)

functions lying in any given dense subspace.

Problem 472

Find the orthogonal complements in L2((1,0))L^{2}\left({(1,0)}\right) of the following sets:

?
(a)

the polynomials in xx;

(b)

the polynomials in x2x^2;

(c)

the polynomials with zero free term;

(d)

the polynomials with sum of coefficients equal to zero.

Problem 473

Find the orthogonal complements of the following spaces in the pre-Hilbert space C[−1,1]C[-1,1] of all continuous functions on [−1,1][-1,1] with the scalar product ⟨f,g⟩=∫−11f(x)g(x)‾ dx\left\langle f, g \right\rangle =\int_{-1}^1 f(x)\overline{g(x)}\, dx:

?
(a)

the functions equal to zero for x≤0x\le 0;

(b)

the functions equal to zero at the point x=0x=0.

Is the theorem on orthogonal complements true in these cases? The theorem on orthogonal complements states that if H1H_1 is a closed subspace of a Hilbert space HH and H2=H1⊥H_2=H_1^\perp, then H=H1⊕H2H=H_1\oplus H_2.

Problem 474

Let ete_t be the characteristic function of the closed interval [0,t][0,t], t≥0t\ge 0. Find the angles between two chords of the curve ete_t in L2((0,∞))L^{2}\left({(0,\infty )}\right):

?
(a)

if the chords have a common endpoint and opposite directions with respect to tt;

(b)

if the chords have a common endpoint and the same direction with respect to tt.

Problem 475
?
(a)

Prove that in every pre-Hilbert space the parallelogram law is satisfied:

∥x−y∥2+∥x+y∥2=2∥x∥2+2∥y∥2. \left\| x-y\right\| ^2+\left\| x+y\right\| ^2=2\left\| x\right\| ^2+2\left\| y\right\| ^2.
(b)

Prove that in every normed linear space over R\mathbb {R} or C\mathbb {C} in which this law is satisfied it is possible to introduce a scalar product such that ∥x∥2=⟨x,x⟩\left\| x\right\|^2= \left\langle x, x \right\rangle.

Problem 476

Prove that in a complex Hilbert space the following equalities hold:

?
(a)

⟨x,y⟩=1N∑k=1N∥x+e2πik/Ny∥2e2πik/N\displaystyle \left\langle x, y \right\rangle =\frac1N\sum_{k=1}^N\left\| x+e^{2\pi ik/N}y\right\|^2e^{2\pi ik/N} for N≥3N\ge 3;

(b)

⟨x,y⟩=12π∫02π∥x+eiθy∥2eiθ dθ\displaystyle \left\langle x, y \right\rangle =\frac1{2\pi }\int_0^{2\pi }\left\| x+e^{i\theta }y\right\|^2e^{i\theta }\, d\theta.

Problem 477

Suppose that the family {xn}\left\{ x_n\right\} of vectors in a Hilbert space HH has the properties that ∥xn∥=1\left\| x_n\right\| =1, ⟨xn,xm⟩=c\left\langle x_n, x_m \right\rangle =c for m≠nm\ne n. Prove that the sequence {xn}\left\{ x_n\right\} has a weak limit.

?
Problem 478

Prove that the following conditions are equivalent for an orthogonal system {xn}\left\{ x_n\right\} of vectors:

  1. ∑nxn\sum_nx_n is strongly convergent;

  2. ∑nxn\sum_nx_n is weakly convergent;

  3. ∑n∥xn∥2\sum_n\left\| x_n\right\|^2 converges.

?
Problem 479

Let SS be any subset of a Hilbert space HH. Prove that (S⊥)⊥(S^\perp )^\perp coincides with the closure of the linear span of SS.

?
Problem 480

Let LL be a subspace of a Hilbert space HH, and f0f_0 a continuous linear functional on LL. Prove that f0f_0 has a unique extension to HH with the same norm.

?
Problem 481
?
(a)

Prove that every operator AA on a complex Hilbert space HH has a unique representation in the form A=B+iCA=B+iC, where BB and CC are Hermitian operators. (This is sometimes written B=Re⁡AB=\operatorname {Re}A, C=Im⁡AC=\operatorname {Im}A.)

(b)

Verify that AA is normal if and only if Re⁡A\operatorname {Re}A and Im⁡A\operatorname {Im}A commute.

(c)

Prove that AA is unitary if and only if it is normal and (Re⁡A)2+(Im⁡A)2=1(\operatorname {Re}A)^2+(\operatorname {Im}A)^2=1.

Problem 482

Prove that:

?
(a)

every operator PP on a Hilbert space having the property that P2=P∗=PP^2=P^*=P is an orthogonal projection, i.e., a projection onto a closed subspace H1H_1 parallel to its orthogonal complement H2H_2;

(b)

every operator SS having the property S−1=S∗=SS^{-1}=S^*=S is an orthogonal reflection in some subspace.

Problem 483

Prove the equality ∥A∗A∥=∥AA∗∥=∥A∥2=∥A∗∥2\left\| A^*A\right\| =\left\| AA^*\right\| =\left\| A\right\|^2=\left\| A^*\right\|^2 for any bounded operator AA on a Hilbert space HH.

?
Problem 484

Let AA be a positive operator on a pre-Hilbert space HH, and xx any vector in HH.

?
(a)

Prove that the numerical sequence a(k)=ln⁡⟨Akx,x⟩a(k)=\ln \left\langle A^kx, x \right\rangle is convex, i.e., a[(k+l)/2]≤[a(k)+a(l)]/2a[(k+l)/2]\le [a(k)+a(l)]/2.

(b)

Prove the inequality ∥Ax∥2≤QA(x)⋅∥A∥\left\| Ax\right\|^2\le Q_A(x)\cdot \left\| A\right\|, where QA(x)=⟨Ax,x⟩Q_A(x)= \left\langle Ax, x \right\rangle.

Problem 485

Let {An}\left\{ A_n\right\} be a monotone bounded sequence of operators on a Hilbert space HH. Prove that s-lim⁡n→∞An\operatorname {s-lim}_{n\to \infty }A_n exists.

?
Problem 486

Suppose that H1H_1 is a closed subspace of a Hilbert space HH, H2=H1⊥H_2=H_1^\perp, PP is the orthogonal projection onto H1H_1, and AA is an operator on HH. Express the following statements in the form of algebraic relations between AA and PP:

?
(a)

H1H_1 is invariant under AA;

(b)

H1H_1 and H2H_2 are invariant under AA.

Problem 487

A pair (L1,L2)(L_1,L_2) of finite-dimensional subspaces of a Hilbert space HH (over R\mathbb {R} or C\mathbb {C}) is said to be congruent to a pair (M1,M2)(M_1,M_2) if there exists a unitary operator UU on HH that carries L1L_1 to M1M_1 and L2L_2 to M2M_2. If such a congruence exists, then dim⁡L1=dim⁡M1\dim L_1=\dim M_1, dim⁡L2=dim⁡M2\dim L_2=\dim M_2. The latter conditions are assumed to hold in what follows. Let PiP_i be the orthogonal projection onto LiL_i, and QiQ_i the orthogonal projection onto MiM_i, i=1,2i=1,2.

?
(a)

Suppose that dim⁡Li=dim⁡Mi=1\dim L_i=\dim M_i=1. Prove that the pairs (L1,L2)(L_1,L_2) and (M1,M2)(M_1,M_2) are congruent in a real space if and only if the angles between the vectors generating these spaces are equal.

(b)

Express the angle between the vectors generating the spaces L1L_1 and L2L_2 in terms of the projections P1P_1 and P2P_2.

(c)

Find a criterion for the congruence of two pairs of one-dimensional subspaces of a complex space.

Problem 488

In the notation of Problem 487 let λ1,…,λn,…\lambda_1,\ldots ,\lambda_n,\ldots be the eigenvalues of the operator P1P2P1P_1P_2P_1.

?
(a)

Prove that the λi\lambda_i are real numbers between 00 and 11.

(b)

Prove that there are not more than k=min⁡(dim⁡L1,dim⁡L2)k=\min (\dim L_1,\dim L_2) nonzero numbers λi\lambda_i. Arrange them in decreasing order and set φi=arccos⁡λi\varphi_i=\arccos \sqrt{\lambda_i}, i=1,2,…,ki=1,2,\ldots ,k. The numbers φi\varphi_i are called the angles between the subspaces L1L_1 and L2L_2.

(c)

If dim⁡L1=1\dim L_1=1, dim⁡M1=1\dim M_1=1, then there is a unique angle φ1\varphi_1 between L1L_1 and L2L_2 and a unique angle ψ1\psi_1 between M1M_1 and M2M_2. Prove that the pairs (L1,L2)(L_1,L_2) and (M1,M2)(M_1,M_2) are congruent if and only if φ1=ψ1\varphi_1=\psi_1.

(d)

Prove the following general criterion for congruence: The pair (L1,L2)(L_1,L_2) is congruent to the pair (M1,M2)(M_1,M_2) if and only if the angles between L1L_1 and L2L_2 are equal to the corresponding angles between M1M_1 and M2M_2.

(e)

The opening between the subspaces L1L_1 and L2L_2 is defined to be the number ∥P1−P2∥\left\| P_1-P_2\right\|. Express this number in terms of the angles between L1L_1 and L2L_2.

Problem 489

Prove that for an operator U∈L(H1,H2)U\in \mathcal L(H_1,H_2) to be unitary it is necessary that UU carry any Hilbert basis in H1H_1 into a basis in H2H_2, and sufficient that UU carry some basis in H1H_1 into a basis in H2H_2.

?
Problem 490

For any operator AA on a Hilbert space prove the relations:

?
(a)

(im⁡A)⊥=ker⁡A∗(\operatorname {im}A)^\perp =\ker A^*;

(b)

(ker⁡A)⊥=im⁡A∗‾(\ker A)^\perp =\overline{\operatorname {im}A^*}

(the bar denotes closure).

Problem 491

Suppose that the sequence {An}\left\{ A_n\right\} of operators on a Hilbert space HH converges weakly to AA and, moreover, that ∥Anx∥→∥Ax∥\left\| A_nx\right\| \to \left\| Ax\right\| for any x∈Hx\in H. Prove that {An}\left\{ A_n\right\} converges strongly to AA. In particular, this means that strong and weak convergence to a unitary limit coincide for sequences of unitary operators.

?
Problem 492

Let L(H)\mathcal L(H) be an algebra of bounded operators on a Hilbert space HH. Prove that every automorphism of L(H)\mathcal L(H) that commutes with the operation of taking adjoints has the form A↦UAU−1A\mapsto UAU^{-1}, where UU is a unitary operator in HH.

?
Problem 493

Find all the norm-closed ideals in the algebra L(H)\mathcal L(H) (see Problem 492) for a separable Hilbert space HH.

?
Problem 494

Derive the following interlacing principle from the Courant theorem. Let AA be a compact Hermitian operator on a Hilbert space HH, H1H_1 a closed subspace of codimension 11 in HH (i.e., dim⁡H1⊥=1\dim H_1^\perp =1), and PP the orthogonal projection onto H1H_1. Then the eigenvalues {λi}\left\{ \lambda_i\right\} of AA and the eigenvalues {μi}\left\{ \mu_i\right\} of PAPPAP, numbered as in Courant's theorem, satisfy

λ−1≤μ−1≤λ2≤⋯≤0≤⋯≤μ2≤λ2≤μ1≤λ1. \lambda _{-1}\le \mu _{-1}\le \lambda _2\le \cdots \le 0\le \cdots \le \mu _2\le \lambda _2\le \mu _1\le \lambda _1.

Courant's theorem numbers the nonzero eigenvalues, with multiplicities, as

λ−1≤λ−2≤⋯<0<⋯≤λ2≤λ1. \lambda _{-1}\le \lambda _{-2}\le \cdots <0<\cdots \le \lambda _2\le \lambda _1.

Writing QA(x)=⟨Ax,x⟩Q_A(x)= \left\langle Ax, x \right\rangle, it gives

λn=inf⁡Hn−1sup⁡x⊥Hn−1QA(x)∥x∥2,λ−n=sup⁡Hn−1inf⁡x⊥Hn−1QA(x)∥x∥2, \lambda _n=\inf _{H_{n-1}}\sup _{x\perp H_{n-1}}\frac{Q_A(x)}{\left\| x\right\| ^2},\qquad \lambda _{-n}=\sup _{H_{n-1}}\inf _{x\perp H_{n-1}}\frac{Q_A(x)}{\left\| x\right\| ^2},

where Hn−1H_{n-1} runs through all (n−1)(n-1)-dimensional subspaces of HH and the quotients are taken for nonzero xx.

?
Problem 495
?
(a)

Prove that every positive operator AA on a Hilbert space has a positive square root BB, which can be obtained as the strong limit of the sequence BnB_n determined by the initial condition B0=0B_0=0 and the recursion formula Bn+1=Bn+[(A−Bn2)/(2∥A∥)]B_{n+1}=B_n+[(A-B_n^2)/(2\sqrt{\left\| A\right\| })].

(b)

Prove that the positive square root BB of a positive operator AA is unique.

Problem 496
?
(a)

Prove that every operator AA on a finite-dimensional Hilbert space HH admits representations in the form A=RUA=RU and A=VSA=VS, where RR and SS are positive operators, and UU and VV are unitary operators. These expressions are called polar decompositions of AA. (In the case dim⁡H=1\dim H=1 this reduces to the expression of a complex number aa in the form reiφre^{i\varphi }.)

(b)

Prove that RR and SS are uniquely determined by AA. Is this true for UU and VV?

(c)

Verify that the one-sided shift operator TT on l2l_2 does not admit a polar decomposition in the sense of part (a).

Problem 497

Let UU be an operator on a Hilbert space HH, H1H_1 the orthogonal complement of ker⁡U\ker U, and H2H_2 the closure of im⁡U\operatorname {im}U. The operator UU is called a partial isometry if it maps H1H_1 isometrically onto H2H_2. Express this property in terms of the orthogonal projections P1P_1, P2P_2 onto H1H_1, H2H_2.

?
Problem 498

Let AA be an operator on a Hilbert space HH. Prove that there is a unique representation of AA in the form A=RUA=RU, where RR is a positive operator and UU is a partial isometry (see Problem 497) such that ker⁡U=ker⁡A\ker U=\ker A. This representation is also called the polar decomposition of AA (cf. Problem 496).

?
Problem 499

An operator AA is called a Hilbert–Schmidt operator if the series ∑β∈B∥Axβ∥2\sum_{\beta \in B}\left\| Ax_\beta \right\|^2 converges for some Hilbert basis {xβ}β∈B\left\{ x_\beta \right\}_{\beta \in B}.

?
(a)

Prove that if AA is a Hilbert--Schmidt operator, then the quantity ∥A∥2=(∑β∈B∥Axβ∥)1/2\left\| A\right\|_{2}=(\sum_{\beta \in B}\left\| Ax_\beta \right\| )^{1/2} does not depend on the choice of basis and defines in the space L2(H)\mathcal L_2(H) of Hilbert--Schmidt operators a norm which majorizes the usual operator norm.

(b)

Prove that the norm ∥⋅∥2\left\| \cdot \right\|_{2} is generated by the scalar product ⟨A,B⟩L2(H)=∑γ∈Γ(Ayγ,Byγ)H\langle A,B\rangle_{\mathcal L_2(H)}=\sum_{\gamma \in \Gamma }(Ay_\gamma ,By_\gamma )_H, where {yγ}γ∈Γ\left\{ y_\gamma \right\}_{\gamma \in \Gamma } is any Hilbert basis in HH.

(c)

Prove that every Hilbert--Schmidt operator is compact.

(d)

Construct an isomorphism between L2(H)\mathcal L_2(H) and the Hilbert tensor product of HH and H′H'.

(e)

Prove that every Hilbert--Schmidt operator on L2(X,μ)L^{2}\left({X,\mu }\right) is an integral operator with kernel K\inL2(X×X,μ×μ)K\inL^{2}\left({X\times X,\mu \times \mu }\right).

Problem 500

An operator AA in a Hilbert space HH is said to be nuclear if it can be represented in the form A=BCA=BC, where BB and CC are Hilbert--Schmidt operators. Prove that:

?
(a)

if AA is a nuclear operator, then for any basis {yγ}γ∈Γ\left\{ y_\gamma \right\}_{\gamma \in \Gamma } in HH the series ∑γ⟨Ayγ,yγ⟩\sum_\gamma \left\langle Ay_\gamma , y_\gamma \right\rangle converges, and its sum does not depend on the choice of basis (it is denoted by tr⁡A\operatorname {tr}A and called the trace of the operator AA);

(b)

if AA is a nuclear operator and BB is a bounded operator, then ABAB and BABA are nuclear, and tr⁡AB=tr⁡BA\operatorname {tr}AB=\operatorname {tr}BA;

(c)

the collection of all nuclear operators forms a Banach space L1(H)\mathcal L_1(H) with respect to the norm ∥⋅∥1\left\| \cdot \right\|_{1} defined by the formula ∥A∥=tr⁡R\left\| A\right\| =\operatorname {tr}R, where A=RUA=RU is the polar decomposition of AA;

(d)

the isomorphisms K(H)′≃L1(H)\mathcal K(H)'\simeq \mathcal L_1(H), L1(H)′≃L(H)\mathcal L_1(H)'\simeq \mathcal L(H) are valid, where K(H)\mathcal K(H) is the space of compact operators on HH with the usual norm.

Problem 501

Let HH be a Hilbert space, and XX a set with measure μ\mu. A collection of unit vectors {ξx}x∈X\left\{ \xi_x\right\}_{x\in X} in HH is called a continuous basis (or a coherent or overfilled system) if for any ξ∈H\xi \in H the function x↦⟨ξ,ξx⟩x\mapsto \left\langle \xi , \xi_x \right\rangle is μ\mu-measurable and

∥ξ∥2=∫X∣⟨ξ,ξx⟩∣2 dμ(x). \left\| \xi \right\| ^2=\int _X\left| \left\langle \xi , \xi _x \right\rangle \right|^2\, d\mu (x).
?
(a)

Construct continuous bases in the spaces of Problem 465.

(b)

Prove that the mapping ξ↦⟨ξ,ξx⟩\xi \mapsto \left\langle \xi , \xi_x \right\rangle is an isometry of HH into L2(X,μ)L^{2}\left({X,\mu }\right).

(c)

Prove that dim⁡H=μ(X)\dim H=\mu (X).

(d)

Prove that tr⁡A=∫X⟨Aξx,ξx⟩ dμ(x)\operatorname {tr}A=\int_X \left\langle A\xi_x, \xi_x \right\rangle \, d\mu (x) for a nuclear operator AA.

Problem 502

Let AA be a nuclear integral operator on L2([0,1])L^{2}\left({[0,1]}\right) with continuous kernel K(x,y)K(x,y). Prove the identity

tr⁡A=∫01K(x,x) dx. \operatorname {tr}A=\int _0^1K(x,x)\, dx.
?