Chapter 12

Spectral Theory of Compact Operators

[43/93%]
Show
LaTeX
Problem 12.1

Find the spectrum and the eigenvalues of an operator AA in C([0,1])C(\left[0,1\right]), which is defined by the formula

?
(1)

(Ax)(t)=∫01e2t−sx(s)ds;(A x)(t) = \int_{0}^{1} e^{2 t-s} x(s) d s ;

(2)

(Ax)(t)=∫01(t2+ts)x(s)ds;(A x)(t) = \int_{0}^{1}\left(t^{2}+t s\right) x(s) d s ;

(3)

(Ax)(t)=3x(t)+∫0π(cos⁡t−2cos⁡s)x(s)ds(A x)(t) = 3 x(t)+\int_{0}^{\pi }(\cos t-2 \cos s) x(s) d s.

Problem 12.2

Find the spectrum, the eigenvalues, the eigenfunctions, and the norm of the operator AA in L2([0,1])L_{2}(\left[0,1\right]), defined by the formula (Ax)(t)=∫01min⁡(t,s)x(s)ds(A x)(t) = \int_{0}^{1} \min (t, s) x(s) d s.

?
Problem 12.3

Which of the following sets can be the spectrum of a compact operator in l2l_{2} :

?
(1)

[0,1]\left[0,1\right]

(2)

{0,1}\left\{ 0,1\right\}

(3)

{0}\left\{ 0\right\}

(4)

{1}\left\{ 1\right\}

(5)

{1n:n≥1}\left\{ \frac{1}{n}: n \geq 1\right\}

(6)

{0}∪{1n:n≥1}\left\{ 0\right\} \cup \left\{ \frac{1}{n}: n \geq 1\right\}

(7)

{0}∪{in:n≥1}\left\{ 0\right\} \cup \left\{ \frac{i}{n}: n \geq 1\right\}

?

Problem 12.4

Determine whether the operator (Ax)(t)=∫01ts3x(s)ds(A x)(t) = \int_{0}^{1} t s^{3} x(s) d s is compact, and find its spectrum:

?
(1)

in the space L2([0,1])L_{2}(\left[0,1\right]);

(2)

in the space C([0,1])C(\left[0,1\right]).

Problem 12.5

Let u,v∈C([a,b])u, v \in C(\left[a, b\right]). Determine whether the operator

(Ax)(t)=∫abu(t)v(s)x(s)ds (A x)(t) = \int _{a}^{b} u(t) v(s) x(s) d s

is compact, and find its spectrum:

?
(1)

in the space L2([a,b])L_{2}(\left[a, b\right]);

(2)

in the space C([a,b])C(\left[a, b\right]).

Problem 12.6

Find the spectrum of an operator AA in C([0,1])C(\left[0,1\right]).

?
(1)

(Ax)(t)=∫01et−sx(s)ds(A x)(t) = \int_{0}^{1} e^{t-s} x(s) d s;

(2)

(Ax)(t)=t3∫01s2x(s)ds(A x)(t) = t^{3} \int_{0}^{1} s^{2} x(s) d s;

(3)

(Ax)(t)=t∫01sin⁡sx(s)ds(A x)(t) = t \int_{0}^{1} \sin s x(s) d s;

(4)

(Ax)(t)=sin⁡t⋅x(π6)(A x)(t) = \sin t \cdot x\left(\frac{\pi }{6}\right);

(5)

(Ax)(t)=∫01sαx(s)ds,α≥0;(A x)(t) = \int_{0}^{1} s^{\alpha } x(s) d s, \alpha \geq 0 ;

(6)

(Ax)(t)=et∫01sx(s)ds(A x)(t) = e^{t} \int_{0}^{1} s x(s) d s;

(7)

(Ax)(t)=2x(0)+tx(1)(A x)(t) = 2 x(0)+t x(1);

(8)

(Ax)(t)=t∫01x(s)1+s2ds(A x)(t) = t \int_{0}^{1} \frac{x(s)}{1+s^{2}} d s.

Problem 12.7

Find the spectrum of an operator AA in L2([0,1])L_{2}(\left[0,1\right]).

?
(1)

(Ax)(t)=tα∫01x(s)ds,α>−12(A x)(t) = t^{\alpha } \int_{0}^{1} x(s) d s, \alpha > -\frac{1}{2};

(2)

(Ax)(t)=e2t∫01sx(s)ds(A x)(t) = e^{2 t} \int_{0}^{1} s x(s) d s;

(3)

(Ax)(t)=t5∫01sx(s)ds(A x)(t) = t^{5} \int_{0}^{1} s x(s) d s;

(4)

(Ax)(t)=∫01et−3sx(s)ds(A x)(t) = \int_{0}^{1} e^{t-3 s} x(s) d s;

(5)

(Ax)(t)=t2∫01s2x(s)ds(A x)(t) = t^{2} \int_{0}^{1} s^{2} x(s) d s;

(6)

(Ax)(t)=t∫01x(s)1+sds(A x)(t) = t \int_{0}^{1} \frac{x(s)}{1+s} d s;

(7)

(Ax)(t)=tα∫01sβx(s)ds,α,β≥0;(A x)(t) = t^{\alpha } \int_{0}^{1} s^{\beta } x(s) d s, \alpha , \beta \geq 0 ;

(8)

(Ax)(t)=χ[13,23](t)⋅∫01esx(s)ds(A x)(t) = \chi_{\left[\frac{1}{3}, \frac{2}{3}\right]}(t) \cdot \int_{0}^{1} e^{s} x(s) d s.

Problem 12.8

Find the spectrum of an operator A∈L(X)A \in \mathcal{L}(X).

?
(1)

(Ax)(t)=∫−11(3t2s−2ts3)x(s)ds,X=C([−1,1]);(A x)(t) = \int_{-1}^{1}\left(3 t^{2} s-2 t s^{3}\right) x(s) d s, X = C(\left[-1,1\right]) ;

(2)

(Ax)(t)=∫01(4t4s3+t)x(s)ds,X=Lp([0,1]),1≤p≤+∞;(A x)(t) = \int_{0}^{1}\left(4 t^{4} s^{3}+t\right) x(s) d s, X = L_{p}(\left[0,1\right]), 1 \leq p \leq +\infty ;

(3)

(Ax)(t)=∫01(8t3+5s4)x(s)ds,X=C([0,1]);(A x)(t) = \int_{0}^{1}\left(8 t^{3}+5 s^{4}\right) x(s) d s, X = C(\left[0,1\right]) ;

(4)

(Ax)(t)=∫01(s−3t2s3)x(s)ds,X=Lp([0,1]),1≤p≤+∞(A x)(t) = \int_{0}^{1}\left(s-3 t^{2} s^{3}\right) x(s) d s, X = L_{p}(\left[0,1\right]), 1 \leq p \leq +\infty.

Problem 12.9

Determine whether an operator AA is compact in L2([a,b])L_{2}(\left[a, b\right]), find its spectrum and norm.

?
(1)

(Ax)(t)=∫02π(cos⁡tcos⁡s+1)x(s)ds,[a,b]=[0,2π];(A x)(t) = \int_{0}^{2 \pi }(\cos t \cos s+1) x(s) d s,\left[a, b\right] = \left[0,2 \pi \right] ;

(2)

(Ax)(t)=∫02πsin⁡(t−s)x(s)ds,[a,b]=[0,2π];(A x)(t) = \int_{0}^{2 \pi } \sin (t-s) x(s) d s,\left[a, b\right] = \left[0,2 \pi \right] ;

(3)

(Ax)(t)=∫0πcos⁡(t+s)x(s)ds,[a,b]=[0,π];(A x)(t) = \int_{0}^{\pi } \cos (t+s) x(s) d s,\left[a, b\right] = \left[0, \pi \right] ;

(4)

(Ax)(t)=2x(t)+∫−11(ts−2t4s4)x(s)ds,[a,b]=[−1,1];(A x)(t) = 2 x(t)+\int_{-1}^{1}\left(t s-2 t^{4} s^{4}\right) x(s) d s,\left[a, b\right] = \left[-1,1\right] ;

(5)

(Ax)(t)=∫01(3ts+4t2s2)x(s)ds,[a,b]=[0,1];(A x)(t) = \int_{0}^{1}\left(3 t s+4 t^{2} s^{2}\right) x(s) d s,\left[a, b\right] = \left[0,1\right] ;

(6)

(Ax)(t)=∫−11(t2s−ts2)x(s)ds,[a,b]=[−1,1];(A x)(t) = \int_{-1}^{1}\left(t^{2} s-t s^{2}\right) x(s) d s,\left[a, b\right] = \left[-1,1\right] ;

(7)

(Ax)(t)=3ix(t)+∫0π2(cos⁡t−cos⁡s)x(s)ds,[a,b]=[0,π];(A x)(t) = 3 i x(t)+\int_{0}^{\pi } 2(\cos t-\cos s) x(s) d s,\left[a, b\right] = \left[0, \pi \right] ;

(8)

(Ax)(t)=∫01(t4+s4)x(s)ds,[a,b]=[0,1];(A x)(t) = \int_{0}^{1}\left(t^{4}+s^{4}\right) x(s) d s,\left[a, b\right] = \left[0,1\right] ;

(9)

(Ax)(t)=x(t)+∫−11(ts+3t2s2)x(s)ds,[a,b]=[−1,1];(A x)(t) = x(t)+\int_{-1}^{1}\left(t s+3 t^{2} s^{2}\right) x(s) d s,\left[a, b\right] = \left[-1,1\right] ;

(10)

(Ax)(t)=−x(t)+∫−ππcos⁡(t−s)x(s)ds,[a,b]=[−π,π](A x)(t) = -x(t)+\int_{-\pi }^{\pi } \cos (t-s) x(s) d s,\left[a, b\right] = \left[-\pi , \pi \right].

Problem 12.10

Determine whether an operator AA in lp,1≤p≤+∞l_{p}, 1 \leq p \leq +\infty, is compact, find its spectrum. Also find the norm of the operator in case p=2p = 2.

?
(1)

Ax=(x1+3x2,3x1+x2,2x3,5x4,5x5,…)A x = \left(x_{1}+3 x_{2}, 3 x_{1}+x_{2}, 2 x_{3}, 5 x_{4}, 5 x_{5}, \ldots \right);

(2)

Ax=(ix2,−ix1,2ix4,−2ix3,x5,0,0,…)A x = \left(i x_{2},-i x_{1}, 2 i x_{4},-2 i x_{3}, x_{5}, 0,0, \ldots \right);

(3)

Ax=(x1+x3,x2+2x4,x1+x3,2x2+x4,0,0,…)A x = \left(x_{1}+x_{3}, x_{2}+2 x_{4}, x_{1}+x_{3}, 2 x_{2}+x_{4}, 0,0, \ldots \right);

(4)

Ax=(x2+x3,x1+x3,x1+x2,2x4,2x5,…)A x = \left(x_{2}+x_{3}, x_{1}+x_{3}, x_{1}+x_{2}, 2 x_{4}, 2 x_{5}, \ldots \right).

Problem 12.11

Find the spectrum, the spectral radius, and the norm of the operator AA in l2l_{2}, defined by the formula Ax=(2x1+2x2,3x2,0,…)A x = \left(2 x_{1}+\sqrt{2} x_{2}, 3 x_{2}, 0, \ldots \right).

?
Problem 12.12

Let XX be a Banach space, A∈S∞(X)A \in S_{\infty }(X) and λ≠0\lambda \neq 0. Prove that R(A−λI)R(A-\lambda I) is a closed subspace in XX.

?
Problem 12.13

Provide an example of a compact operator in l2l_{2}, for which

?
(1)

0 is not an eigenvalue;

(2)

0 is an eigenvalue of finite multiplicity;

(3)

0 is an eigenvalue of infinite multiplicity.

Problem 12.14

Let AA be a compact self-adjoint operator in an infinite-dimensional Hilbert space. Prove that if AA has finitely many eigenvalues, then 0 is an eigenvalue of AA of infinite multiplicity.

?
Problem 12.15

Let {φn:n≥1}\left\{ \varphi_{n}: n \geq 1\right\} be an orthonormal basis in the Hilbert space H,A∈L(H)H, A \in \mathcal{L}(H) and Aφn=λnφn,n≥1A \varphi_{n} = \lambda_{n} \varphi_{n}, n \geq 1, where {λn:n≥1}\left\{ \lambda_{n}: n \geq 1\right\} is a bounded sequence of numbers. Prove that Ax=∑n=1∞λn(x,φn)φnA x = \sum_{n = 1}^{\infty } \lambda_{n}\left(x, \varphi_{n}\right) \varphi_{n} for all x∈Hx \in H.

?
Problem 12.16

Let A∈L(L2([0,π])),(Aφn)(t)=1n2φn(t),t∈[0,π]A \in \mathcal{L}\left(L_{2}(\left[0, \pi \right])\right),\left(A \varphi_{n}\right)(t) = \frac{1}{n^{2}} \varphi_{n}(t), t \in [0, \pi ], where φn(t)=sin⁡nt,n≥1\varphi_{n}(t) = \sin n t, n \geq 1. Prove that AA is a Hilbert-Schmidt integral operator. Find its kernel KK.

?
Problem 12.17

Let HH be a Hilbert space, {φn:n≥1}\left\{ \varphi_{n}: n \geq 1\right\} be an orthonormal sequence in H,{λn:n≥1}⊂C\{0}H,\left\{ \lambda_{n}: n \geq 1\right\} \subset \mathbb {C} \backslash \left\{ 0\right\} be a bounded sequence, Ax=∑n=1∞λn(x,φn)φn,x∈HA x = \sum_{n = 1}^{\infty } \lambda_{n}\left(x, \varphi_{n}\right) \varphi_{n}, x \in H.

Prove that

?
(1)

A∈L(H)A \in \mathcal{L}(H);

(2)

σp(A)\{0}={λn:n≥1}\sigma_{p}(A) \backslash \left\{ 0\right\} = \left\{ \lambda_{n}: n \geq 1\right\};

(3)

0∉σp(A)0 \notin \sigma_{p}(A) if and only if {φn:n≥1}\left\{ \varphi_{n}: n \geq 1\right\} is a basis in HH.

Under what conditions on the sequence {λn:n≥1}\left\{ \lambda_{n}: n \geq 1\right\} is the operator AA selfadjoint? nonnegative? normal? compact? a Hilbert-Schmidt operator?

Problem 12.18

Let AA be a self-adjoint compact operator in a Hilbert space H,A=∑n=1∞λn(⋅,φn)φnH, A = \sum_{n = 1}^{\infty } \lambda_{n}\left(\cdot , \varphi_{n}\right) \varphi_{n} be its Schmidt expansion. Prove that

?
(1)

∀k≥1:Akx=∑n=1∞λnk(x,φn)φn,x∈H\forall k \geq 1: A^{k} x = \sum_{n = 1}^{\infty } \lambda_{n}^{k}\left(x, \varphi_{n}\right) \varphi_{n}, x \in H;

(2)

∀λ∈C\{0,λn,n≥1}\forall \lambda \in \mathbb {C} \backslash \left\{ 0, \lambda_{n}, n \geq 1\right\} :

Rλ(A)y=∑n=1∞(y,φn)λn−λφn−1λPy=1λ(∑n=1∞λn(y,φn)λn−λφn−y),y∈H R_{\lambda }(A) y = \sum _{n = 1}^{\infty } \frac{\left(y, \varphi _{n}\right)}{\lambda _{n}-\lambda } \varphi _{n}-\frac{1}{\lambda } P y = \frac{1}{\lambda }\left(\sum _{n = 1}^{\infty } \frac{\lambda _{n}\left(y, \varphi _{n}\right)}{\lambda _{n}-\lambda } \varphi _{n}-y\right), y \in H

where PP is the orthogonal projector onto Ker⁡A\operatorname {Ker} A.

Problem 12.19

Let HH be a separable Hilbert space, AA be a compact self-adjoint operator in H,{φn:n≥1}H,\left\{ \varphi_{n}: n \geq 1\right\} be an orthonormal basis in HH, such that Aφn=λnφn,n≥1A \varphi_{n} = \lambda_{n} \varphi_{n}, n \geq 1. For fixed λ∈C\{0}\lambda \in \mathbb {C} \backslash \left\{ 0\right\} and y∈Hy \in H consider the equation (A−λI)x=y(A-\lambda I) x = y. Prove that

?
(1)

if λn≠λ\lambda_{n} \neq \lambda for all n≥1n \geq 1, then the solution is unique and has the form

x=1λ(∑n=1∞λnλn−λ(y,φn)φn−y) x = \frac{1}{\lambda }\left(\sum _{n = 1}^{\infty } \frac{\lambda _{n}}{\lambda _{n}-\lambda }\left(y, \varphi _{n}\right) \varphi _{n}-y\right)
(2)

if J={n≥1∣λn=λ}≠∅J = \left\{ n \geq 1 \mid \lambda_{n} = \lambda \right\} \neq \varnothing (the set JJ is finite, because the operator AA is compact), then the solution exists if and only if (y,φn)=0\left(y, \varphi_{n}\right) = 0 for all n∈Jn \in J. In this case, the general solution has the form

x=1λ(∑n∈N\Jλnλn−λ(y,φn)φn−y)+∑n∈JCnφn, x = \frac{1}{\lambda }\left(\sum _{n \in \mathbb {N} \backslash J} \frac{\lambda _{n}}{\lambda _{n}-\lambda }\left(y, \varphi _{n}\right) \varphi _{n}-y\right)+\sum _{n \in J} C_{n} \varphi _{n},

where Cn,n∈JC_{n}, n \in J, are arbitrary constants.

Problem 12.20

Let AA be a nonnegative compact operator in a Hilbert space HH. Prove that there exists a unique nonnegative compact operator BB such that B2=AB^{2} = A. In this case, if A=∑n=1∞λn(⋅,φn)φnA = \sum_{n = 1}^{\infty } \lambda_{n}\left(\cdot , \varphi_{n}\right) \varphi_{n} is the Schmidt expansion, then B=∑n=1∞λn(⋅,φn)φnB = \sum_{n = 1}^{\infty } \sqrt{\lambda_{n}}\left(\cdot , \varphi_{n}\right) \varphi_{n}.

?
Remark.
?

The operator BB is called square root of the operator AA. The square root of the operator AA is denoted as B=AB = \sqrt{A} or B=A12B = A^{\frac{1}{2}}. In problem 9.52* the existence and uniqueness of the square root of an arbitrary nonnegative operator AA is proven by another method without assumption that the operator AA is compact.

Problem 12.21

Let K(t,s)=∑n=1∞1n2sin⁡nπtsin⁡nπs,t,s∈[−1,1]K(t, s) = \sum_{n = 1}^{\infty } \frac{1}{n^{2}} \sin n \pi t \sin n \pi s, t, s \in [-1,1]. Consider the operator AA in L2([−1,1])L_{2}(\left[-1,1\right]), defined by the formula

(Ax)(t)=∫−11K(t,s)x(s)ds,t∈[−1,1] (A x)(t) = \int _{-1}^{1} K(t, s) x(s) d s, t \in [-1,1]
?
(1)

Prove that AA is a self-adjoint Hilbert-Schmidt operator (in particular, AA is a compact operator).

(2)

Find σ(A),r(A)\sigma (A), r(A) and ∥A∥\left\| A\right\|, as well as an orthonormal sequence of eigenfunctions of the operator AA.

(3)

Find the Schmidt expansion of AA.

Problem 12.22

Let KK be an even 2π2 \pi-periodic real-valued function that belongs to L2([−π,π]),K(t)=∑n∈Zcneint,t∈RL_{2}(\left[-\pi , \pi \right]), K(t) = \sum_{n \in \mathbb {Z}} c_{n} e^{i n t}, t \in \mathbb {R} be its Fourier series. Consider the operator AA in L2([−π,π])L_{2}(\left[-\pi , \pi \right]), defined by the formula

(Ax)(t)=∫−ππK(t−s)x(s)ds,t∈[−π,π] (A x)(t) = \int _{-\pi }^{\pi } K(t-s) x(s) d s, t \in [-\pi , \pi ]
?
(1)

Prove that AA is a self-adjoint Hilbert-Schmidt operator and the functions φn(t)=12πeint,t∈[−π,π],n∈Z\varphi_{n}(t) = \frac{1}{\sqrt{2 \pi }} e^{i n t}, t \in [-\pi , \pi ], n \in \mathbb {Z}, which form an orthonormal basis in the space L2([−π,π])L_{2}(\left[-\pi , \pi \right]), are the eigenfunctions of the operator AA.

(2)

Find σp(A),σ(A),r(A)\sigma_{p}(A), \sigma (A), r(A) and ∥A∥\left\| A\right\|.

(3)

Find the Schmidt expansion of AA.

Problem 12.23

Let KK be an even 2π2 \pi-periodic real-valued function that belongs to L2([−π,π]),K(t)=a02+∑n=1∞ancos⁡nt,t∈RL_{2}(\left[-\pi , \pi \right]), K(t) = \frac{a_{0}}{2}+\sum_{n = 1}^{\infty } a_{n} \cos n t, t \in \mathbb {R} be its Fourier series. Consider the operator AA in L2([−π,π])L_{2}(\left[-\pi , \pi \right]), defined by the formula

(Ax)(t)=∫−ππK(t+s)x(s)ds,t∈[−π,π] (A x)(t) = \int _{-\pi }^{\pi } K(t+s) x(s) d s, t \in [-\pi , \pi ]
?
(1)

Prove that AA is a self-adjoint Hilbert-Schmidt operator with the following eigenvalues and orthonormal eigenfunctions: λ0=πa0,φ0(t)=12π;λn(1)=πan,φn(1)(t)=cos⁡ntπ;λn(2)=−πan,φn(2)(t)=sin⁡ntπ,n≥1\lambda_{0} = \pi a_{0}, \varphi_{0}(t) = \frac{1}{\sqrt{2 \pi }} ; \lambda_{n}^{(1)} = \pi a_{n}, \varphi_{n}^{(1)}(t) = \frac{\cos n t}{\sqrt{\pi }} ; \lambda_{n}^{(2)} = -\pi a_{n}, \varphi_{n}^{(2)}(t) = \frac{\sin n t}{\sqrt{\pi }}, n \geq 1.

(2)

Find σ(A),r(A)\sigma (A), r(A) and ∥A∥\left\| A\right\|.

(3)

Find the Schmidt expansion of AA.

Problem 12.24

Let AA be the integral operator in L2([−π,π])L_{2}(\left[-\pi , \pi \right]) with a kernel K(t,s)K(t, s). Find the norm and the spectrum of AA and the eigenfunctions that correspond to nonzero eigenvalues.

?
(1)

K(t,s)=∑n=1Nλncos⁡n(t−s),N∈N,λ1,…,λN∈C\{0}K(t, s) = \sum_{n = 1}^{N} \lambda_{n} \cos n(t-s), N \in \mathbb {N}, \lambda_{1}, \ldots , \lambda_{N} \in \mathbb {C} \backslash \left\{ 0\right\};

(2)

K(t,s)=cos⁡4(t−s)K(t, s) = \cos^{4}(t-s);

(3)

K(t,s)=∑n=1∞1n2cos⁡ntcos⁡nsK(t, s) = \sum_{n = 1}^{\infty } \frac{1}{n^{2}} \cos n t \cos n s;

(4)

K(t,s)=∑n=1∞1n2sin⁡(2n+1)tsin⁡(2n+1)sK(t, s) = \sum_{n = 1}^{\infty } \frac{1}{n^{2}} \sin (2 n+1) t \sin (2 n+1) s;

(5)

K(t,s)=∑n=0∞1(2n+1)2cos⁡(2n+1)tcos⁡(2n+1)sK(t, s) = \sum_{n = 0}^{\infty } \frac{1}{(2 n+1)^{2}} \cos (2 n+1) t \cos (2 n+1) s;

(6)

K(t,s)=∑n=0∞1(2n+1)2cos⁡(2n+1)(t−s)K(t, s) = \sum_{n = 0}^{\infty } \frac{1}{(2 n+1)^{2}} \cos (2 n+1)(t-s).

Problem 12.25

Find the norm, the spectrum, the eigenvalues and the eigenfunctions of an operator AA in L2([a,b])L_{2}(\left[a, b\right]), defined by the formula (Ax)(t)=∫abK(t,s)x(s)ds,t∈[a,b](A x)(t) = \int_{a}^{b} K(t, s) x(s) d s, t \in \left[a, b\right], in the following cases:

?
(1)

K(t,s)={t(1−s),t≤s,s(1−t),s≤t,a=0,b=1K(t, s) = \begin{cases} t(1-s), & t \leq s, \\ s(1-t), & s \leq t, \end{cases} \quad a = 0, b = 1;

(2)

K(t,s)={sin⁡tcos⁡s,t≤s,cos⁡tsin⁡s,s≤t,a=0,b=πK(t, s) = \begin{cases} \sin t \cos s, & t \leq s, \\ \cos t \sin s, & s \leq t, \end{cases} \quad a = 0, b = \pi;

(3)

K(t,s)={cos⁡tsin⁡s,t≤s,sin⁡tcos⁡s,s≤t,a=0,b=πK(t, s) = \begin{cases} \cos t \sin s, & t \leq s, \\ \sin t \cos s, & s \leq t, \end{cases} \quad a = 0, b = \pi;

(4)

K(t,s)={(t+1)s,t≤s,t(s+1),s≤t,a=0,b=1K(t, s) = \begin{cases} (t+1) s, & t \leq s, \\ t(s+1), & s \leq t, \end{cases} \quad a = 0, b = 1;

(5)

K(t,s)={sin⁡tsin⁡(1−s),t≤s,sin⁡(1−t)sin⁡s,s≤t,a=0,b=1K(t, s) = \begin{cases} \sin t \sin (1-s), & t \leq s, \\ \sin (1-t) \sin s, & s \leq t, \end{cases} \quad a = 0, b = 1;

(6)

K(t,s)={(1+t)(1−s),t≤s,(1−t)(1+s),s≤t,a=−1,b=1K(t, s) = \begin{cases} (1+t)(1-s), & t \leq s, \\ (1-t)(1+s), & s \leq t, \end{cases} \quad a = -1, b = 1.

Problem 12.26

Find the norm of the operator AA in L2([0,1])L_{2}(\left[0,1\right]), defined by the formula (Ax)(t)=∫0tx(s)ds(A x)(t) = \int_{0}^{t} x(s) d s.

?
Problem 12.27

Let HH be a Hilbert space, A∈L(H)A \in \mathcal{L}(H) and A∗=AA^{*} = A. Prove that if σ(A)\sigma (A) is a countable set which has no nonzero limit points, and every nonzero point of its spectrum is an eigenvalue of finite multiplicity, then the operator AA is compact.

?
Problem 12.28

(Nikolsky) Prove that Theorem 12.1 remains valid for the operator A∈L(X)A \in \mathcal{L}(X), where XX is a Banach space, if the condition A∈S∞(X)A \in S_{\infty }(X) is replaced with the condition Ak∈S∞(X)A^{k} \in S_{\infty }(X) for some k∈Nk \in \mathbb {N}.

Theorem 12.1: Let AA be a compact operator in an infinite-dimensional Banach space. Then

?
(1)

0∈σ(A);0 \in \sigma (A) ;

(2)

σ(A)\sigma (A) is a countable set, the only possible limit point of which is 0 ;

(3)

any nonzero point of the spectrum is an eigenvalue of finite multiplicity, i.e., if λ∈σ(A)\{0}\lambda \in \sigma (A) \backslash \left\{ 0\right\}, then 0<dimKer⁡(A−λI)<+∞0 < \operatorname {dimKer}(A-\lambda I) < +\infty.

Problem 12.29

(Sudakov)

?
(1)

Let XX be a Banach space, M⊂X,A∈L(X),1∉σp(A),Ak∈S∞(X)M \subset X, A \in \mathcal{L}(X), 1 \notin \sigma_{p}(A), A^{k} \in S_{\infty }(X) for some k∈Nk \in \mathbb {N}, and there exists C>0C > 0 such that ∥Ax−x∥<C\left\| A x-x\right\| < C for all x∈Mx \in M. Prove that the set MM is bounded.

(2)

Prove that in the compactness criteria in Lp([a,b])L_{p}(\left[a, b\right]) by Kolmogorov and M. Riesz (see problems 10.81 and 10.82) condition (a) follows from condition (b).

Problem 12.30

Let HH be a separable Hilbert space, {φn:n≥1}\left\{ \varphi_{n}: n \geq 1\right\} and {ψn:n≥1}\left\{ \psi_{n}: n \geq 1\right\} be orthonormal sequences in H,{sn:n≥1}H,\left\{ s_{n}: n \geq 1\right\} be a sequence of numbers such that lim⁡n→∞sn=0\lim_{n \rightarrow \infty } s_{n} = 0. For any x∈Hx \in H put Anx=∑k=1nsk(x,φk)ψk,n≥1A_{n} x = \sum_{k = 1}^{n} s_{k}\left(x, \varphi_{k}\right) \psi_{k}, n \geq 1, and Ax=∑n=1∞sn(x,φn)ψnA x = \sum_{n = 1}^{\infty } s_{n}\left(x, \varphi_{n}\right) \psi_{n}. Prove that

?
(1)

An⇉A,n→∞A_{n} \rightrightarrows A, n \rightarrow \infty;

(2)

AA is a compact operator in HH;

(3)

AA is a Hilbert-Schmidt operator if and only if ∑n=1∞∣sn∣2<∞\sum_{n = 1}^{\infty }\left|s_{n}\right|^{2} < \infty.

Problem 12.31

Let AA be the operator from problem 12.30∘12.30^{\circ }. Find A∗,AA∗,A∗AA^{*}, A A^{*}, A^{*} A.

?
Problem 12.32

Prove that any compact operator AA in a Hilbert space HH can be represented as

A=∑n=1Nsn(⋅,φn)ψn, A = \sum _{n = 1}^{N} s_{n}\left(\cdot , \varphi _{n}\right) \psi _{n},

where N≤+∞,{φn:1≤n≤N}N \leq +\infty ,\left\{ \varphi_{n}: 1 \leq n \leq N\right\} and {ψn:1≤n≤N}\left\{ \psi_{n}: 1 \leq n \leq N\right\} are orthonormal sequences and {sn:1≤n≤N}\left\{ s_{n}: 1 \leq n \leq N\right\} are positive eigenvalues of the operator ∣A∣=A∗A\left|A\right| = \sqrt{A^{*} A} (square root is defined in problem 12.20). In this case, if N=+∞N = +\infty, then the series converges in operator norm. The numbers sn,1≤n≤Ns_{n}, 1 \leq n \leq N, are called the singular values of the operator AA.

?
Problem 12.33

Let HH be a Hilbert space, A∈S∞(H),∥A∗x∥≤∥Ax∥,x∈HA \in S_{\infty }(H),\left\| A^{*} x\right\| \leq \left\| A x\right\| , x \in H. Prove that A∗A=AA∗A^{*} A = A A^{*}.

?
Problem 12.34

Let HH be a Hilbert space, A∈S∞(H),A∗=A,{λn+:1≤n≤N+},{λn−:1≤n≤N−}A \in S_{\infty }(H), A^{*} = A,\left\{ \lambda_{n}^{+}: 1 \leq n \leq N^{+}\right\} ,\left\{ \lambda_{n}^{-}: 1 \leq n \leq N^{-}\right\}, where N+,N−≤+∞N^{+}, N^{-} \leq +\infty, are the sequences of positive and negative eigenvalues of the operator AA, numbered counting their multiplicities, and λ1−≤λ2−≤…<0<…≤λ2+≤λ1+\lambda_{1}^{-} \leq \lambda_{2}^{-} \leq \ldots < 0 < \ldots \leq \lambda_{2}^{+} \leq \lambda_{1}^{+}(one or both of the sequences {λn+},{λn−}\left\{ \lambda_{n}^{+}\right\} ,\left\{ \lambda_{n}^{-}\right\}may be empty). Also, let φn±\varphi_{n}^{ \pm }be the unit eigenvectors corresponding to the eigenvalues λn±\lambda_{n}^{ \pm }. Prove that

λ1+=max⁡{(Ax,x)∣∥x∥=1}=(Aφ1+,φ1+)λn+=max⁡{(Ax,x)∣∥x∥=1,(x,φk+)=0,1≤k≤n−1}==(Aφn+,φn+),2≤n≤N+. \begin{aligned} & \lambda _{1}^{+} = \max \left\{ (A x, x) \mid \left\| x\right\| = 1\right\} = \left(A \varphi _{1}^{+}, \varphi _{1}^{+}\right) \\ & \lambda _{n}^{+} = \max \left\{ (A x, x) \mid \left\| x\right\| = 1,\left(x, \varphi _{k}^{+}\right) = 0,1 \leq k \leq n-1\right\} = \\ & = \left(A \varphi _{n}^{+}, \varphi _{n}^{+}\right), 2 \leq n \leq N^{+}. \end{aligned}
?
Remark.
?

For the numbers λn−\lambda_{n}^{-}, similar formulas are true with max replaced with min⁡\min, because {−λn−:n≥1}\left\{ -\lambda_{n}^{-}: n \geq 1\right\} are the positive eigenvalues of the operator −A-A.

Problem 12.35

Let Ln\mathcal{L}_{n} be the family of all nn dimensional subspaces of H,n≥1H, n \geq 1. In notations from problem 12.34, prove that for all 1≤n≤N+1 \leq n \leq N^{+}

?
(1)

λn+=min⁡L∈Ln−1max⁡x∈L⊥,∥x∥=1(Ax,x)\lambda_{n}^{+} = \min_{L \in \mathcal{L}_{n-1}} \max_{x \in L^{\perp },\left\| x\right\| = 1}(A x, x);

(2)

λn+=max⁡L∈Lnmin⁡x∈L,∥x∥=1(Ax,x)\lambda_{n}^{+} = \max_{L \in \mathcal{L}_{n}} \min_{x \in L,\left\| x\right\| = 1}(A x, x).

Problem 12.36

(Cauchy interlacing theorem) Let HH be a Hilbert space, AA be a compact self-adjoint operator in H,G⊂HH, G \subset H be a subspace of codimension 1 (i.e., dim⁡G⊥=\operatorname {dim} G^{\perp } = 1), PP be the orthogonal projector onto GG. Prove that the eigenvalues λn±\lambda_{n}^{ \pm }of the operator AA and the eigenvalues μn±\mu_{n}^{ \pm }of the operator PAPP A P, numbered in the same way as in problem 12.34, satisfy the following inequalities: λ1+≥μ1+≥λ2+≥μ2+≥…>0\lambda_{1}^{+} \geq \mu_{1}^{+} \geq \lambda_{2}^{+} \geq \mu_{2}^{+} \geq \ldots > 0, if AA has positive eigenvalues, λ1−≤μ1−≤λ2−≤μ2−≤…<0\lambda_{1}^{-} \leq \mu_{1}^{-} \leq \lambda_{2}^{-} \leq \mu_{2}^{-} \leq \ldots < 0, if AA has negative eigenvalues.

?
Problem 12.37

Let HH be a Hilbert space, A∈S∞(H)A \in S_{\infty }(H). Denote by sn(A),1≤n≤NA,NA≤+∞s_{n}(A), 1 \leq n \leq N_{A}, N_{A} \leq +\infty, the singular values of the operator AA (see problem 12.32), written in the order of decreasing taking into account their multiplicities. In case NA<+∞N_{A} < +\infty put sn(A)=0,n>NAs_{n}(A) = 0, n > N_{A}. Prove that

?
(1)

sn(A∗)=sn(A),n≥1;s_{n}\left(A^{*}\right) = s_{n}(A), n \geq 1 ;

(2)

sn(A)=min⁡L:dim⁡L≤n−1max⁡x∈L⊥,∥x∥=1∥Ax∥,1≤n≤dim⁡Hs_{n}(A) = \min_{L: \operatorname {dim} L \leq n-1} \max_{x \in L^{\perp },\left\| x\right\| = 1}\left\| A x\right\| , 1 \leq n \leq \operatorname {dim} H, in particular, s1(A)=∥A∥s_{1}(A) = \left\| A\right\|;

(3)

if B∈L(H)B \in \mathcal{L}(H), then

sn(BA)≤∥B∥sn(A),sn(AB)≤∥B∥sn(A),n≥1; s_{n}(B A) \leq \left\| B\right\| s_{n}(A), s_{n}(A B) \leq \left\| B\right\| s_{n}(A), n \geq 1 ;
(4)

if B∈S∞(H),A≥B≥0B \in S_{\infty }(H), A \geq B \geq 0, then sn(A)≥sn(B),n≥1s_{n}(A) \geq s_{n}(B), n \geq 1;

(5)

sn(A)=min⁡∥A−T∥s_{n}(A) = \min \left\| A-T\right\|, where the minimum is taken over all operators T∈S0(H)T \in S_{0}(H), for which rank⁡T=dim⁡R(T)≤n−1,n≥1;\operatorname {rank} T = \operatorname {dim} R(T) \leq n-1, n \geq 1 ;

(6)

if B∈S∞(H)B \in S_{\infty }(H), then sm+n−1(A+B)≤sm(A)+sn(B),n,m≥1s_{m+n-1}(A+B) \leq s_{m}(A)+s_{n}(B), n, m \geq 1;

(7)

if B∈S∞(H)B \in S_{\infty }(H), then ∣sn(A)−sn(B)∣≤∥A−B∥,n≥1\left|s_{n}(A)-s_{n}(B)\right| \leq \left\| A-B\right\| , n \geq 1.

Problem 12.38

Let HH be a separable Hilbert space, A∈S∞(H),{sn(A):n≥1}A \in S_{\infty }(H),\left\{ s_{n}(A): n \geq 1\right\} be a sequence of singular values of the operator AA, numbered with respect to their multiplicities and, possibly, supplemented by zeros. Prove that AA is a Hilbert-Schmidt operator if and only if ∑n=1∞sn2(A)<+∞\sum_{n = 1}^{\infty } s_{n}^{2}(A) < +\infty. Moreover ∥A∥2=(∑n=1∞sn2(A))1/2\left\| A\right\|_{2} = \left(\sum_{n = 1}^{\infty } s_{n}^{2}(A)\right)^{1 / 2}.

?
Problem 12.39

Let HH be a separable Hilbert space, A∈S∞(H),{sn(A):n≥1}A \in S_{\infty }(H),\left\{ s_{n}(A): n \geq 1\right\} be a sequence of singular values of the operator AA. It is said that the operator AA belongs to the Schatten class Sp(H)S_{p}(H), where 1≤p<+∞1 \leq p < +\infty, if ∥A∥p:=(∑n=1∞snp(A))1/p<+∞\left\| A\right\|_{p}: = \left(\sum_{n = 1}^{\infty } s_{n}^{p}(A)\right)^{1 / p} < +\infty. Prove that

?
(1)

A∈Sp(H)A \in S_{p}(H) if and only if A∗∈Sp(H)A^{*} \in S_{p}(H);

(2)

∥A∥p≥∥A∥,A∈Sp(H)\left\| A\right\|_{p} \geq \left\| A\right\| , A \in S_{p}(H);

(3)

if A∈Sp(H)A \in S_{p}(H) and B∈L(H)B \in \mathcal{L}(H), then AB,BA∈Sp(H)A B, B A \in S_{p}(H), and

∥AB∥p≤∥A∥p⋅∥B∥,∥BA∥p≤∥A∥p⋅∥B∥; \left\| A B\right\| _{p} \leq \left\| A\right\| _{p} \cdot \left\| B\right\| , \quad \left\| B A\right\| _{p} \leq \left\| A\right\| _{p} \cdot \left\| B\right\| ;
(4)

∥A∥p=max⁡(∑n=1∞∣(Aen,gn)∣p)1/p,A∈S∞(H)\left\| A\right\|_{p} = \max \left(\sum_{n = 1}^{\infty }\left|\left(A e_{n}, g_{n}\right)\right|^{p}\right)^{1 / p}, A \in S_{\infty }(H), where the maximum is taken over all pairs of orthonormal bases {en:n≥1},{gn:n≥1}\left\{ e_{n}: n \geq 1\right\} ,\left\{ g_{n}: n \geq 1\right\};

(5)

if A,B∈Sp(H)A, B \in S_{p}(H), then A+B∈Sp(H)A+B \in S_{p}(H) and

∥A+B∥p≤∥A∥p+∥B∥p \left\| A+B\right\| _{p} \leq \left\| A\right\| _{p}+\left\| B\right\| _{p}
(6)

Sp(H)S_{p}(H) endowed with the norm ∥⋅∥p\left\| \cdot \right\|_{p} is a Banach space;

(7)

the set of finite-dimensional operators S0(H)S_{0}(H) is dense in Sp(H)S_{p}(H).

Remark.
?

In view of problem 12.38, the definition of the Schatten class Sp(H)S_{p}(H) for p=2p = 2 aligns with the definition of the set of Hilbert-Schmidt operators S2(H)S_{2}(H).

Problem 12.40

Let HH be a separable Hilbert space. An operator AA that belongs to the Schatten class S1(H)S_{1}(H) is called nuclear (or trace-class operator), and the number ∥A∥1\left\| A\right\|_{1} is called the nuclear norm. Prove the following statements:

?
(1)

If A∈S1(H)A \in S_{1}(H) and {en:n≥1}\left\{ e_{n}: n \geq 1\right\} is a complete orthonormal basis in HH, then the series ∑n=1∞(Aen,en)\sum_{n = 1}^{\infty }\left(A e_{n}, e_{n}\right) converges absolutely and its sum does not depend on the choice of basis {en:n≥1}\left\{ e_{n}: n \geq 1\right\}. This sum is denoted by tr⁡A\operatorname {tr} A and is called the trace of the operator AA.

(2)

If A∈S1(H)A \in S_{1}(H) and A≥0A \geq 0, then tr⁡A=∑n=1∞λn\operatorname {tr} A = \sum_{n = 1}^{\infty } \lambda_{n}, where λn,n≥1\lambda_{n}, n \geq 1, are the eigenvalues of the operator AA, numbered with respect to their multiplicities.

(3)

If A,B∈S2(H)A, B \in S_{2}(H), then AB,BA∈S1(H)A B, B A \in S_{1}(H),

∥AB∥1≤∥A∥2⋅∥B∥2,∥BA∥1≤∥A∥2⋅∥B∥2. \left\| A B\right\| _{1} \leq \left\| A\right\| _{2} \cdot \left\| B\right\| _{2}, \quad \left\| B A\right\| _{1} \leq \left\| A\right\| _{2} \cdot \left\| B\right\| _{2}.
(4)

If A,B∈S2(H)A, B \in S_{2}(H), then tr⁡AB=tr⁡BA\operatorname {tr} A B = \operatorname {tr} B A.

(5)

For every A∈S1(H)A \in S_{1}(H) there exist B,C∈S2(H)B, C \in S_{2}(H) such that A=BCA = B C.

(6)

The norm ∥A∥2\left\| A\right\|_{2} is defined by the scalar product (A,B):=tr⁡B∗A(A, B): = \operatorname {tr} B^{*} A.

(7)

S2(H)S_{2}(H) is a separable Hilbert space with an orthonormal basis, which consists of one-dimensional operators Am,n=(⋅,en)gm,m,n∈NA_{m, n} = \left(\cdot , e_{n}\right) g_{m}, m, n \in \mathbb {N}, where {en:n≥1}\left\{ e_{n}: n \geq 1\right\} and {gn:n≥1}\left\{ g_{n}: n \geq 1\right\} are orthonormal bases in HH.

Remark.
?
(2)

Lidskii's theorem states that the equality tr⁡A=∑n=1∞λn\operatorname {tr} A = \sum_{n = 1}^{\infty } \lambda_{n} is valid for all A∈S1(H)A \in S_{1}(H). In particular, if an operator AA has no eigenvalues, then tr⁡A=0\operatorname {tr} A = 0.

Problem 12.41

Let H=L2(T,μ)H = L_{2}(T, \mu ) be a separable Hilbert space. Prove that A∈S2(H)A \in S_{2}(H) if and only if AA is an integral operator of the form (Ax)(t)=∫TK(t,s)x(s)dμ(s),t∈T,x∈H(A x)(t) = \int_{T} K(t, s) x(s) d \mu (s), t \in T, x \in H, where K∈L2(T×T,μ×μ)K \in L_{2}(T \times T, \mu \times \mu ). Moreover ∥A∥22=∫T∫T∣K(t,s)∣2dμ(t)dμ(s)\left\| A\right\|_{2}^{2} = \int_{T} \int_{T}\left|K(t, s)\right|^{2} d \mu (t) d \mu (s).

?
Problem 12.42

Let H1,H2H_{1}, H_{2} be separable Hilbert spaces, {en:n≥1},{fn:n≥1}\left\{ e_{n}: n \geq 1\right\} ,\left\{ f_{n}: n \geq 1\right\} be orthonormal bases in H1H_{1} and H2H_{2}, respectively. An operator A∈L(H1,H2)A \in \mathcal{L}\left(H_{1}, H_{2}\right) is called a Hilbert-Schmidt operator if its Hilbert-Schmidt norm ∥A∥2=(∑j,k=1∞∣(Aek,fj)∣2)12\left\| A\right\|_{2} = \left(\sum_{j, k = 1}^{\infty }\left|\left(A e_{k}, f_{j}\right)\right|^{2}\right)^{\frac{1}{2}} is finite. Prove that

?
(1)

Hilbert-Schmidt norm does not depend on the choice of orthonormal bases in spaces H1H_{1} and H2H_{2};

(2)

functions sinh⁡t,eikt,t∈[−π,π],k∈Z\sinh t, e^{i k t}, t \in [-\pi , \pi ], k \in \mathbb {Z}, form a complete orthogonal system in the space W21([−π,π])W_{2}^{1}(\left[-\pi , \pi \right]) (see problems 2.32∘,2.33∗)2.32^{\circ }, 2.33^{*});

(3)

the embedding operator of the space W21([−π,π])W_{2}^{1}(\left[-\pi , \pi \right]) into L2([−π,π])L_{2}(\left[-\pi , \pi \right]) is a Hilbert-Schmidt operator.

Problem 12.43

Let α=(α1,α2,…),αk≥1\alpha = \left(\alpha_{1}, \alpha_{2}, \ldots \right), \alpha_{k} \geq 1 for all k≥1,l2,αk \geq 1, l_{2, \alpha } be the Hilbert space of all sequences x=(x1,x2,…)x = \left(x_{1}, x_{2}, \ldots \right), which satisfy the condition

∥x∥α=(∑n=1∞αnxn2)12<∞ \left\| x\right\| _{\alpha } = \left(\sum _{n = 1}^{\infty } \alpha _{n} x_{n}^{2}\right)^{\frac{1}{2}} < \infty

endowed with the corresponding scalar product. Prove the following statements:

?
(1)

l2,α⊂l2l_{2, \alpha } \subset l_{2}, and the norm of the embedding operator A:l2,α→l2,Ax=xA: l_{2, \alpha } \rightarrow l_{2}, A x = x, does not exceed one;

(2)

the operator AA is compact if and only if lim⁡n→∞αn=∞\lim_{n \rightarrow \infty } \alpha_{n} = \infty;

(3)

A:l2,α→l2A: l_{2, \alpha } \rightarrow l_{2} is a Hilbert-Schmidt operator if and only if ∑n=1∞1αn<∞\sum_{n = 1}^{\infty } \frac{1}{\alpha_{n}} < \infty.