Alessandro Palliccia

Real and Functional Analysis

Exercises on the continuity, norm, kernel and range of linear operators

1,032 words, , published

The exercises study linear operators between normed spaces, with λ\lambda the Lebesgue measure on R.{\mathbb{R}.}

Exercises

The operator Tf=∫−11fg dλTf = \int_{-1}^{1} f g \, d\lambda

Example 14.1. Prove that the map T ⁣:(C([−1,1]),∥⋅∥∞)→(R,∣⋅∣){T \colon (C([-1, 1]), \|\cdot\|_\infty) \to (\mathbb{R}, |\cdot|)} given by

Tf:=∫−11fg dλ,whereg(x)={−1−1≤x≤0,10<x≤1,Tf := \int_{-1}^{1} f g \, d\lambda, \qquad \text{where} \quad g(x) = \begin{cases} -1 & -1 \le x \le 0, \\ 1 & 0 < x \le 1, \end{cases}

is linear and continuous.

For the operator TT of Example 14.1, let α,β∈R{\alpha, \beta \in \mathbb{R}} and f,h∈C([−1,1]).{f, h \in C([-1, 1]).} Then

T(αf+βh)=∫−11(αf+βh)g dλ=∫−11(αfg+βhg) dλ=∫−11αfg dλ+∫−11βhg dλ=α∫−11fg dλ+β∫−11hg dλ=αTf+βTh,\begin{aligned} T(\alpha f + \beta h) &= \int_{-1}^{1} (\alpha f + \beta h) g \, d\lambda \\ &= \int_{-1}^{1} (\alpha f g + \beta h g) \, d\lambda \\ &= \int_{-1}^{1} \alpha f g \, d\lambda + \int_{-1}^{1} \beta h g \, d\lambda \\ &= \alpha \int_{-1}^{1} f g \, d\lambda + \beta \int_{-1}^{1} h g \, d\lambda = \alpha Tf + \beta Th, \end{aligned}

where the third and the fourth equalities are the linearity of the integral, Remark 5.13. Hence TT is linear.

To prove that TT is continuous, we prove that it is bounded. For a linear operator TT from a normed space (X,∥⋅∥X){(X, \|\cdot\|_X)} to a normed space (Y,∥⋅∥Y){(Y, \|\cdot\|_Y)} this means, as in Definition 13.5 (ii), that there exists C>0{C > 0} such that ∥Tf∥Y≤C∥f∥X{\|Tf\|_Y \le C \|f\|_X} for every f∈X.{f \in X.} Here we need C>0{C > 0} such that ∣Tf∣≤C∥f∥∞{|Tf| \le C \|f\|_\infty} for every f∈C([−1,1]).{f \in C([-1, 1]).} Let f∈C([−1,1]).{f \in C([-1, 1]).} Then, since ∣f∣≤∥f∥∞{|f| \le \|f\|_\infty} and ∣g∣=1,{|g| = 1,}

∣Tf∣=∣∫−11fg dλ∣≤∫−11∣f∣ ∣g∣ dλ≤∥f∥∞∫−11∣g∣ dλ=2∥f∥∞.\begin{aligned} |Tf| = \biggl| \int_{-1}^{1} f g \, d\lambda \biggr| &\le \int_{-1}^{1} |f| \, |g| \, d\lambda \\ &\le \|f\|_\infty \int_{-1}^{1} |g| \, d\lambda \\ &= 2 \|f\|_\infty. \end{aligned}

Hence TT is bounded, and TT is continuous by Theorem 13.8. We now compute ∥T∥L,{\|T\|_{\mathcal{L}},} the norm of TT in L(C([−1,1]),R).{\mathcal{L}(C([-1, 1]), \mathbb{R}).}

Remark 14.2. Let (X,∥⋅∥X){(X, \|\cdot\|_X)} and (Y,∥⋅∥Y){(Y, \|\cdot\|_Y)} be normed spaces and let T∈L(X,Y){T \in \mathcal{L}(X, Y)} with T≠0.{T \neq 0.} By Proposition 13.10, ∥T∥L\|T\|_{\mathcal{L}} is the least constant C>0{C > 0} such that ∥Tx∥Y≤C∥x∥X{\|Tx\|_Y \le C \|x\|_X} for every x∈X.{x \in X.} In general, to compute ∥T∥L:{\|T\|_{\mathcal{L}}{:}}
(i) try to estimate ∥Tx∥Y≤C∥x∥X{\|Tx\|_Y \le C \|x\|_X} for every x∈X,{x \in X,} keeping CC as small as possible;
(ii) (a) find x0∈X{x_0 \in X} with x0≠0{x_0 \neq 0} such that ∥Tx0∥Y=C∥x0∥X,{\|Tx_0\|_Y = C \|x_0\|_X,} or
(b) find (xn)⊂X∖{0}{(x_n) \subset X \setminus \{0\}} such that ∥Txn∥Y/∥xn∥X→C.{\|Tx_n\|_Y/\|x_n\|_X \to C.}

For Example 14.1 we choose (b) of Remark 14.2. For every integer n≥1{n \ge 1} let fn∈C([−1,1]){f_n \in C([-1, 1])} be equal to −1-1 on [−1,−1/n],{[-1, -1/n],} linear on [−1/n,1/n][-1/n, 1/n] and equal to 11 on [1/n,1],{[1/n, 1],} as in Figure 14.1, so that ∥fn∥∞=1.{\|f_n\|_\infty = 1.}

−1−1/n1/n11−1fn

Figure 14.1. The drawing of Example 14.1: the function fnf_n for the nn of the slider, equal to −1-1 on [−1,−1/n],{[-1, -1/n],} linear on [−1/n,1/n][-1/n, 1/n] and equal to 11 on [1/n,1].{[1/n, 1].}

Since fn g=1{f_n \, g = 1} on [−1,−1/n]{[-1, -1/n]} and on [1/n,1],{[1/n, 1],}

∣Tfn∣=∣∫−11fn g dλ∣=∣∫−1−1/nfn g dλ+∫−1/n1/nfn g dλ+∫1/n1fn g dλ∣=∣−1n+1+1−1n+∫−1/n1/nfn g dλ∣≥2−∣−2n+∫−1/n1/nfn g dλ∣.\begin{aligned} |Tf_n| &= \biggl| \int_{-1}^{1} f_n \, g \, d\lambda \biggr| \\ &= \biggl| \int_{-1}^{-1/n} f_n \, g \, d\lambda + \int_{-1/n}^{1/n} f_n \, g \, d\lambda + \int_{1/n}^{1} f_n \, g \, d\lambda \biggr| \\ &= \biggl| -\frac{1}{n} + 1 + 1 - \frac{1}{n} + \int_{-1/n}^{1/n} f_n \, g \, d\lambda \biggr| \\ &\ge 2 - \biggl| -\frac{2}{n} + \int_{-1/n}^{1/n} f_n \, g \, d\lambda \biggr|. \end{aligned}

Here −2/n→0{-2/n \to 0} and

∣∫−1/n1/nfn g dλ∣≤∥fn∥∞ 2n=2n⟶0.\biggl| \int_{-1/n}^{1/n} f_n \, g \, d\lambda \biggr| \le \|f_n\|_\infty \, \frac{2}{n} = \frac{2}{n} \longrightarrow 0.

Hence

∣Tfn∣∥fn∥∞=∣Tfn∣≥2−∣−2n+∫−1/n1/nfn g dλ∣→n→+∞2.\frac{|Tf_n|}{\|f_n\|_\infty} = |Tf_n| \ge 2 - \biggl| -\frac{2}{n} + \int_{-1/n}^{1/n} f_n \, g \, d\lambda \biggr| \xrightarrow[n \to +\infty]{} 2.

In detail, ∣Tfn∣≤2∥fn∥∞=2{|Tf_n| \le 2 \|f_n\|_\infty = 2} by the bound proved for the continuity of T,{T,} so the ratio tends to 2,{2,} and ∥T∥L=2{\|T\|_{\mathcal{L}} = 2} by (b) of Remark 14.2.

The operators Tif=∫−11f x/(1+x2) dλT_i f = \int_{-1}^{1} f \, x/(1 + x^2) \, d\lambda

Example 14.3. Let X1=L∞((−1,1)){X_1 = L^\infty((-1, 1))} and X2=C([−1,1]),{X_2 = C([-1, 1]),} and for i=1,2{i = 1, 2} let Ti ⁣:Xi→Xi{T_i \colon X_i \to X_i} be given by

Tif:=∫−11f x1+x2 dλ.T_i f := \int_{-1}^{1} f \, \frac{x}{1 + x^2} \, d\lambda.

Note 14.4. For every f∈Xi,{f \in X_i,} the image TifT_i f is a constant function, and therefore ∥Tif∥Xi=∥Tif∥∞=∣Tif∣.{\|T_i f\|_{X_i} = \|T_i f\|_\infty = |T_i f|.}

(1) Is TiT_i continuous? By Note 14.4, for every f∈Xi,{f \in X_i,}

∥Tif∥∞=∣Tif∣=∣∫−11x1+x2 f dλ∣≤∥f∥∞∫−11∣x∣1+x2 dλ=∥f∥∞∫012x1+x2 dλ=(log⁡2) ∥f∥∞.\begin{aligned} \|T_i f\|_\infty = |T_i f| &= \biggl| \int_{-1}^{1} \frac{x}{1 + x^2} \, f \, d\lambda \biggr| \\ &\le \|f\|_\infty \int_{-1}^{1} \frac{|x|}{1 + x^2} \, d\lambda \\ &= \|f\|_\infty \int_0^1 \frac{2x}{1 + x^2} \, d\lambda = (\log 2) \, \|f\|_\infty. \end{aligned}

Hence TiT_i is bounded, hence continuous, for i=1,2.{i = 1, 2.}

(2) What is ∥Ti∥L(Xi,Xi)?{\|T_i\|_{\mathcal{L}(X_i, X_i)}?} For i=1,{i = 1,} that is, X1=L∞((−1,1)),{X_1 = L^\infty((-1, 1)),} we choose

f0(x)={−1−1≤x≤0,10<x≤1,f_0(x) = \begin{cases} -1 & -1 \le x \le 0, \\ 1 & 0 < x \le 1, \end{cases}

which belongs to L∞((−1,1)){L^\infty((-1, 1))} and satisfies ∥f0∥∞=1.{\|f_0\|_\infty = 1.} Then

∥T1f0∥∞∥f0∥∞=∥T1f0∥∞=∣T1f0∣=∣∫−11x1+x2 f0 dλ∣=∣∫−11∣x∣1+x2 dλ∣=∫−11∣x∣1+x2 dλ=log⁡2.\begin{aligned} \frac{\|T_1 f_0\|_\infty}{\|f_0\|_\infty} &= \|T_1 f_0\|_\infty = |T_1 f_0| \\ &= \biggl| \int_{-1}^{1} \frac{x}{1 + x^2} \, f_0 \, d\lambda \biggr| \\ &= \biggl| \int_{-1}^{1} \frac{|x|}{1 + x^2} \, d\lambda \biggr| \\ &= \int_{-1}^{1} \frac{|x|}{1 + x^2} \, d\lambda = \log 2. \end{aligned}

Hence, by (1) and (a) of Remark 14.2, ∥T1∥L(L∞,L∞)=log⁡2.{\|T_1\|_{\mathcal{L}(L^\infty, L^\infty)} = \log 2.} For i=2,{i = 2,} that is, X2=C([−1,1]),{X_2 = C([-1, 1]),} we take the functions fnf_n of Figure 14.1, with ∥fn∥∞=1.{\|f_n\|_\infty = 1.} Then

∥T2fn∥∞∥fn∥∞=∣T2fn∣=∣∫−11x1+x2 fn dλ∣.\frac{\|T_2 f_n\|_\infty}{\|f_n\|_\infty} = |T_2 f_n| = \biggl| \int_{-1}^{1} \frac{x}{1 + x^2} \, f_n \, d\lambda \biggr|.

Moreover fn→H{f_n \to H} in L1((−1,1)),{L^1((-1, 1)),} where

H(x)={−1−1≤x<0,10<x<1.H(x) = \begin{cases} -1 & -1 \le x < 0, \\ 1 & 0 < x < 1. \end{cases}

Hence

lim⁡n→∞∣T2fn∣=∣∫−11x1+x2 H dλ∣=∫−11∣x∣1+x2 dλ=log⁡2.\begin{aligned} \lim_{n \to \infty} |T_2 f_n| &= \biggl| \int_{-1}^{1} \frac{x}{1 + x^2} \, H \, d\lambda \biggr| \\ &= \int_{-1}^{1} \frac{|x|}{1 + x^2} \, d\lambda = \log 2. \end{aligned}

Indeed, ∣x∣/(1+x2)≤1{|x|/(1 + x^2) \le 1} on (−1,1),{(-1, 1),} so the integrals of xfn/(1+x2){x f_n/(1 + x^2)} and of xH/(1+x2){x H/(1 + x^2)} differ at most by ∥fn−H∥1,{\|f_n - H\|_1,} which tends to 0.{0.} Hence, by (1) and (b) of Remark 14.2, ∥T2∥L(C,C)=log⁡2.{\|T_2\|_{\mathcal{L}(C, C)} = \log 2.}

(3) Let g(x)=x2{g(x) = x^2} for x∈(−1,1).{x \in (-1, 1).} Is g∈Ker⁡(T1)?{g \in \operatorname{Ker}(T_1)?} Is T1T_1 injective? We compute

T1g=∫−11x1+x2 x2 dλ=∫−11x31+x2 dλ=0,T_1 g = \int_{-1}^{1} \frac{x}{1 + x^2} \, x^2 \, d\lambda = \int_{-1}^{1} \frac{x^3}{1 + x^2} \, d\lambda = 0,

because the function x3/(1+x2){x^3/(1 + x^2)} is odd. Hence g∈Ker⁡(T1).{g \in \operatorname{Ker}(T_1).}

Remark 14.5. Let TT be a linear operator between normed spaces. Then TT is injective if and only if Ker⁡(T)={0}.{\operatorname{Ker}(T) = \{0\}.}

By Remark 14.5, since g≠0,{g \neq 0,} the operator T1T_1 is not injective.

The operator (Tf)(x)=e−x∫0xey f(y) dy(Tf)(x) = e^{-x} \int_0^x e^y \, f(y) \, dy

Example 14.6. Let T ⁣:L∞((0,1))→L∞((0,1)){T \colon L^\infty((0, 1)) \to L^\infty((0, 1))} be given by

(Tf)(x):=e−x∫0xey f(y) dyfor every f∈L∞((0,1)) and every x∈(0,1).(Tf)(x) := e^{-x} \int_0^x e^y \, f(y) \, dy \qquad \text{for every } f \in L^\infty((0, 1)) \text{ and every } x \in (0, 1).

(1) Is TT continuous?

For the operator TT of Example 14.6, let f∈L∞((0,1)){f \in L^\infty((0, 1))} and x∈(0,1).{x \in (0, 1).} Then

∣(Tf)(x)∣=e−x∣∫0xey f(y) dy∣≤e−x∫0xey ∣f(y)∣ dy≤∥f∥∞ e−x∫0xey dy=∥f∥∞(1−e−x)≤(1−e−1)∥f∥∞,\begin{aligned} |(Tf)(x)| &= e^{-x} \biggl| \int_0^x e^y \, f(y) \, dy \biggr| \\ &\le e^{-x} \int_0^x e^y \, |f(y)| \, dy \\ &\le \|f\|_\infty \, e^{-x} \int_0^x e^y \, dy \\ &= \|f\|_\infty (1 - e^{-x}) \le (1 - e^{-1}) \|f\|_\infty, \end{aligned}

where the integral of eye^y over (0,x)(0, x) is ex−1,{e^x - 1,} and e−x≥e−1.{e^{-x} \ge e^{-1}.} Hence ∥Tf∥∞≤(1−e−1)∥f∥∞{\|Tf\|_\infty \le (1 - e^{-1}) \|f\|_\infty} for every f∈L∞((0,1)),{f \in L^\infty((0, 1)),} and TT is continuous.

(2) What is ∥T∥L?{\|T\|_{\mathcal{L}}?} We choose f0=1,{f_0 = 1,} so that ∥f0∥∞=1.{\|f_0\|_\infty = 1.} Then

∥Tf0∥∞∥f0∥∞=∥Tf0∥∞=∥e−x∫0xey dy∥∞=∥1−e−x∥∞=1−e−1.\begin{aligned} \frac{\|Tf_0\|_\infty}{\|f_0\|_\infty} = \|Tf_0\|_\infty &= \biggl\| e^{-x} \int_0^x e^y \, dy \biggr\|_\infty \\ &= \|1 - e^{-x}\|_\infty = 1 - e^{-1}. \end{aligned}

For the last equality, Figure 14.2 draws the graphs of e−x,{e^{-x},} which decreases from 1,{1,} and of 1−e−x,{1 - e^{-x},} which increases from 0.{0.} Hence, by (1) and (a) of Remark 14.2, ∥T∥L=1−e−1.{\|T\|_{\mathcal{L}} = 1 - e^{-1}.}

1e−x1 − e−x

Figure 14.2. The drawings of (2) in Example 14.6: on the left the graph of e−x,{e^{-x},} on the right the graph of 1−e−x,{1 - e^{-x},} for x∈[0,1].{x \in [0, 1].}

(3) Is TT injective? The idea is to impose f∈Ker⁡(T),{f \in \operatorname{Ker}(T),} that is, Tf=0,{Tf = 0,} and to look for necessary conditions on f.{f.} For f∈L∞((0,1)){f \in L^\infty((0, 1))} let

F(x):=∫0xey f(y) dyfor x∈[0,1],F(x) := \int_0^x e^y \, f(y) \, dy \qquad \text{for } x \in [0, 1],

so that (Tf)(x)=e−x F(x){(Tf)(x) = e^{-x} \, F(x)} for x∈(0,1).{x \in (0, 1).} Let f∈Ker⁡(T).{f \in \operatorname{Ker}(T).} Then e−x F(x)=0{e^{-x} \, F(x) = 0} for almost every x∈(0,1).{x \in (0, 1).}

In fact, e−x F(x)=0{e^{-x} \, F(x) = 0} for every x∈(0,1).{x \in (0, 1).} The function y↦ey f(y){y \mapsto e^y \, f(y)} belongs to L∞((0,1))⊂L1((0,1)),{L^\infty((0, 1)) \subset L^1((0, 1)),} so FF is continuous by the first fundamental theorem of calculus, Theorem 9.9. Hence TfTf is continuous, and it vanishes at every point because f∈Ker⁡(T).{f \in \operatorname{Ker}(T).} Indeed, a continuous function that vanishes almost everywhere on (0,1)(0, 1) vanishes at every point of (0,1),{(0, 1),} since every nonempty open interval has positive measure.

Since e−x>0,{e^{-x} > 0,} it follows that F(x)=0{F(x) = 0} for almost every x∈(0,1).{x \in (0, 1).} Moreover, since y↦ey f(y){y \mapsto e^y \, f(y)} belongs to L1((0,1)),{L^1((0, 1)),} the function FF belongs to AC([0,1]),{\mathrm{AC}([0, 1]),} and F′(x){F'(x)} exists and equals ex f(x){e^x \, f(x)} for almost every x∈(0,1).{x \in (0, 1).} In detail, FF vanishes at every point of (0,1),{(0, 1),} because TfTf does, so F′=0{F' = 0} on (0,1).{(0, 1).} Hence 0=ex f(x){0 = e^x \, f(x)} for almost every x,{x,} where ex>0.{e^x > 0.} So f(x)=0{f(x) = 0} for almost every x,{x,} that is, f=0{f = 0} in L∞((0,1)).{L^\infty((0, 1)).} Therefore Ker⁡(T)={0},{\operatorname{Ker}(T) = \{0\},} and TT is injective by Remark 14.5.

(4) Is TT surjective? The idea is to look for necessary conditions on TfTf as ff varies in L∞((0,1)).{L^\infty((0, 1)).}

(i) Let f∈L∞((0,1)).{f \in L^\infty((0, 1)).} In (Tf)(x)=e−x F(x){(Tf)(x) = e^{-x} \, F(x)} the factor e−xe^{-x} is continuous, and F∈C([0,1]){F \in C([0, 1])} by the first fundamental theorem of calculus, because ey f(y)e^y \, f(y) belongs to L1((0,1)).{L^1((0, 1)).} Hence Tf∈C([0,1]){Tf \in C([0, 1])} for every f∈L∞((0,1)).{f \in L^\infty((0, 1)).} Therefore R(T)⊂C([0,1])⊊L∞((0,1)),{R(T) \subset C([0, 1]) \subsetneq L^\infty((0, 1)),} and TT is not surjective.

(ii) We can observe that R(T)⊂AC([0,1]).{R(T) \subset \mathrm{AC}([0, 1]).} Both factors of (Tf)(x)=e−x F(x){(Tf)(x) = e^{-x} \, F(x)} belong to AC([0,1]):{\mathrm{AC}([0, 1]){:}} the function e−x,{e^{-x},} and the function FF of (3).

Remark 14.7. Let a<b{a < b} be real numbers, let v1,v2∈AC([a,b]){v_1, v_2 \in \mathrm{AC}([a, b])} and let v:=v1 v2.{v := v_1 \, v_2.} Then v∈AC([a,b]).{v \in \mathrm{AC}([a, b]).}

Proof (complete). Let ε>0.{\varepsilon > 0.} By Definition 9.11, applied to v1v_1 and to v2v_2 with ε/(1+∥v1∥∞+∥v2∥∞){\varepsilon/(1 + \|v_1\|_\infty + \|v_2\|_\infty)} in place of ε,{\varepsilon,} for i=1,2{i = 1, 2} there exists δi>0{\delta_i > 0} such that, for every finite family {(aj,bj):j=1,…,N}{\{ (a_j, b_j) : j = 1, \dots, N \}} of pairwise disjoint subintervals of [a,b],{[a, b],}

∑j=1N(bj−aj)<δi  ⟹  ∑j=1N∣vi(bj)−vi(aj)∣<ε1+∥v1∥∞+∥v2∥∞.\sum_{j=1}^{N} (b_j - a_j) < \delta_i \implies \sum_{j=1}^{N} |v_i(b_j) - v_i(a_j)| < \frac{\varepsilon}{1 + \|v_1\|_\infty + \|v_2\|_\infty}.

We take δ=min⁡{δ1,δ2}{\delta = \min\{\delta_1, \delta_2\}} and estimate. If (b1−a1)+⋯+(bN−aN)<δ,{(b_1 - a_1) + \dots + (b_N - a_N) < \delta,} then

∑j=1N∣v(bj)−v(aj)∣=∑j=1N∣v1(bj)v2(bj)−v1(aj)v2(aj)∣≤∑j=1N(∣v1(bj)∣ ∣v2(bj)−v2(aj)∣+∣v2(aj)∣ ∣v1(bj)−v1(aj)∣)≤∥v1∥∞∑j=1N∣v2(bj)−v2(aj)∣+∥v2∥∞∑j=1N∣v1(bj)−v1(aj)∣≤ε1+∥v1∥∞+∥v2∥∞(∥v1∥∞+∥v2∥∞)<ε.\begin{aligned} &\sum_{j=1}^{N} |v(b_j) - v(a_j)| = \sum_{j=1}^{N} |v_1(b_j) v_2(b_j) - v_1(a_j) v_2(a_j)| \\ &\quad \le \sum_{j=1}^{N} \bigl( |v_1(b_j)| \, |v_2(b_j) - v_2(a_j)| + |v_2(a_j)| \, |v_1(b_j) - v_1(a_j)| \bigr) \\ &\quad \le \|v_1\|_\infty \sum_{j=1}^{N} |v_2(b_j) - v_2(a_j)| + \|v_2\|_\infty \sum_{j=1}^{N} |v_1(b_j) - v_1(a_j)| \\ &\quad \le \frac{\varepsilon}{1 + \|v_1\|_\infty + \|v_2\|_\infty} \bigl( \|v_1\|_\infty + \|v_2\|_\infty \bigr) < \varepsilon. \end{aligned}

In the first inequality, v1(bj)v2(aj){v_1(b_j) v_2(a_j)} is added and subtracted inside each absolute value. The second uses ∣v1(bj)∣≤∥v1∥∞{|v_1(b_j)| \le \|v_1\|_\infty} and ∣v2(aj)∣≤∥v2∥∞,{|v_2(a_j)| \le \|v_2\|_\infty,} and the third the choice of δ.{\delta.} The last inequality holds because ∥v1∥∞+∥v2∥∞<1+∥v1∥∞+∥v2∥∞.{\|v_1\|_\infty + \|v_2\|_\infty < 1 + \|v_1\|_\infty + \|v_2\|_\infty.} Hence v∈AC([a,b]).{v \in \mathrm{AC}([a, b]).} ■\blacksquare

By Remark 14.7, applied on [0,1][0, 1] to the functions e−xe^{-x} and F,{F,} the function TfTf belongs to AC([0,1]){\mathrm{AC}([0, 1])} for every f∈L∞((0,1)).{f \in L^\infty((0, 1)).} That is, R(T)⊂AC([0,1]).{R(T) \subset \mathrm{AC}([0, 1]).}

The operator (Tf)(x)=(2+χ[1,2](x))f(x+1)(Tf)(x) = (2 + \chi_{[1, 2]}(x)) f(x + 1)

Example 14.8. Let T ⁣:L1((0,∞))→L1((0,∞)){T \colon L^1((0, \infty)) \to L^1((0, \infty))} be given by

(Tf)(x):=(2+χ[1,2](x))f(x+1)for almost every x>0.(Tf)(x) := \big( 2 + \chi_{[1, 2]}(x) \big) f(x + 1) \qquad \text{for almost every } x > 0.

(1) Is TT continuous?

For the operator TT of Example 14.8 and every f∈L1((0,∞)),{f \in L^1((0, \infty)),}

∥Tf∥1=∫0∞∣Tf∣ dλ=∫0∞(2+χ[1,2](x)) ∣f(x+1)∣ dλ=∫0∞2 ∣f(x+1)∣ dλ+∫12∣f(x+1)∣ dλ=2∫1∞∣f(x)∣ dλ+∫23∣f(x)∣ dλ≤3∫0∞∣f∣ dλ=3∥f∥1.\begin{aligned} \|Tf\|_1 = \int_0^\infty |Tf| \, d\lambda &= \int_0^\infty \big( 2 + \chi_{[1, 2]}(x) \big) \, |f(x + 1)| \, d\lambda \\ &= \int_0^\infty 2 \, |f(x + 1)| \, d\lambda + \int_1^2 |f(x + 1)| \, d\lambda \\ &= 2 \int_1^\infty |f(x)| \, d\lambda + \int_2^3 |f(x)| \, d\lambda \\ &\le 3 \int_0^\infty |f| \, d\lambda = 3 \|f\|_1. \end{aligned}

Hence TT is bounded, hence continuous.

(2) What is ∥T∥L?{\|T\|_{\mathcal{L}}?} Let f0∈L1((0,∞)){f_0 \in L^1((0, \infty))} be a function with f0≠0{f_0 \neq 0} that vanishes outside [2,3],{[2, 3],} as in Figure 14.3.

23f0

Figure 14.3. The drawing of (2) in Example 14.8, an instance: a function f0f_0 of L1((0,∞)){L^1((0, \infty))} with f0≠0,{f_0 \neq 0,} equal to 0 outside [2,3],{[2, 3],} with the parts where it is 0 thick on the axis.

By the computation of (1),

∥Tf0∥1=2∫1∞∣f0∣ dλ+∫23∣f0∣ dλ=2∫23∣f0∣ dλ+∫23∣f0∣ dλ=3∫23∣f0∣ dλ=3∫0∞∣f0∣ dλ=3∥f0∥1.\begin{aligned} \|Tf_0\|_1 &= 2 \int_1^\infty |f_0| \, d\lambda + \int_2^3 |f_0| \, d\lambda \\ &= 2 \int_2^3 |f_0| \, d\lambda + \int_2^3 |f_0| \, d\lambda \\ &= 3 \int_2^3 |f_0| \, d\lambda = 3 \int_0^\infty |f_0| \, d\lambda = 3 \|f_0\|_1. \end{aligned}

Hence, by (1) and (a) of Remark 14.2, ∥T∥L=3.{\|T\|_{\mathcal{L}} = 3.}

(3) Is TT injective? Let f∈Ker⁡(T),{f \in \operatorname{Ker}(T),} that is, Tf=0.{Tf = 0.} Then (2+χ[1,2](x))f(x+1)=0{(2 + \chi_{[1, 2]}(x)) f(x + 1) = 0} for almost every x>0,{x > 0,} where 2+χ[1,2](x)>0.{2 + \chi_{[1, 2]}(x) > 0.} Hence f(x+1)=0{f(x + 1) = 0} for almost every x>0.{x > 0.} That is, f(y)=0{f(y) = 0} for almost every y>1.{y > 1.} Then we take a function f0∈L1((0,∞)){f_0 \in L^1((0, \infty))} with f0≠0{f_0 \neq 0} that vanishes on (1,∞),{(1, \infty),} as in Figure 14.4.

1f0

Figure 14.4. The drawing of (3) in Example 14.8, an instance: a function f0f_0 of L1((0,∞)){L^1((0, \infty))} with f0≠0,{f_0 \neq 0,} equal to 0 on (1,∞),{(1, \infty),} where it is thick on the axis.

Then (Tf0)(x)=(2+χ[1,2](x))f0(x+1){(Tf_0)(x) = (2 + \chi_{[1, 2]}(x)) f_0(x + 1)} for almost every x>0.{x > 0.} Notice that f0(x+1)=0{f_0(x + 1) = 0} for every x>0.{x > 0.} Hence (Tf0)(x)=0{(Tf_0)(x) = 0} for almost every x,{x,} that is, Tf0=0{Tf_0 = 0} in L1((0,∞)).{L^1((0, \infty)).} So f0∈Ker⁡(T){f_0 \in \operatorname{Ker}(T)} and f0≠0,{f_0 \neq 0,} and TT is not injective.

(4) Is TT surjective? Let u∈R(T),{u \in R(T),} so that there exists f∈L1((0,∞)){f \in L^1((0, \infty))} with Tf=u.{Tf = u.} Then u(x)=(2+χ[1,2](x))f(x+1){u(x) = (2 + \chi_{[1, 2]}(x)) f(x + 1)} for almost every x>0,{x > 0,} and, since 2+χ[1,2](x)>0,{2 + \chi_{[1, 2]}(x) > 0,}

f(x+1)=u(x)2+χ[1,2](x)for almost every x>0,f(y)=u(y−1)2+χ[1,2](y−1)for almost every y>1,\begin{aligned} f(x + 1) &= \frac{u(x)}{2 + \chi_{[1, 2]}(x)} && \text{for almost every } x > 0, \\ f(y) &= \frac{u(y - 1)}{2 + \chi_{[1, 2]}(y - 1)} && \text{for almost every } y > 1, \end{aligned}

where the second line is the first with y=x+1.{y = x + 1.} Then, if u∈L1((0,∞)),{u \in L^1((0, \infty)),} consider

f(y)={0for almost every y∈(0,1),u(y−1)2+χ[1,2](y−1)for almost every y>1.f(y) = \begin{cases} 0 & \text{for almost every } y \in (0, 1), \\[4pt] \dfrac{u(y - 1)}{2 + \chi_{[1, 2]}(y - 1)} & \text{for almost every } y > 1. \end{cases}

(i) f∈L1((0,∞)).{f \in L^1((0, \infty)).} Indeed, since 2+χ[1,2](y−1)≥2,{2 + \chi_{[1, 2]}(y - 1) \ge 2,}

∫0∞∣f∣ dλ=∫1∞∣u(y−1)∣2+χ[1,2](y−1) dλ≤12∫1∞∣u(y−1)∣ dλ=12∫0∞∣u∣ dλ<+∞.\begin{aligned} \int_0^\infty |f| \, d\lambda &= \int_1^\infty \frac{|u(y - 1)|}{2 + \chi_{[1, 2]}(y - 1)} \, d\lambda \\ &\le \frac{1}{2} \int_1^\infty |u(y - 1)| \, d\lambda \\ &= \frac{1}{2} \int_0^\infty |u| \, d\lambda < +\infty. \end{aligned}

(ii) For almost every x>0,{x > 0,}

(Tf)(x)=(2+χ[1,2](x))f(x+1)=(2+χ[1,2](x))u(x+1−1)2+χ[1,2](x+1−1)=u(x).\begin{aligned} (Tf)(x) &= \big( 2 + \chi_{[1, 2]}(x) \big) f(x + 1) \\ &= \big( 2 + \chi_{[1, 2]}(x) \big) \frac{u(x + 1 - 1)}{2 + \chi_{[1, 2]}(x + 1 - 1)} \\ &= u(x). \end{aligned}

Hence for every u∈L1((0,∞)){u \in L^1((0, \infty))} there exists f∈L1((0,∞)){f \in L^1((0, \infty))} such that Tf=u,{Tf = u,} and TT is surjective.

The operator (Tx)(k)=kx(k)/(1+k2)(Tx)^{(k)} = k x^{(k)}/(1 + k^2)

Example 14.9. Let T ⁣:ℓ2→ℓ2{T \colon \ell^2 \to \ell^2} be given by

(Tx)(k)=k1+k2 x(k)for every x∈ℓ2 and every k∈N, k≥0.(Tx)^{(k)} = \frac{k}{1 + k^2} \, x^{(k)} \qquad \text{for every } x \in \ell^2 \text{ and every } k \in \mathbb{N}, \ k \ge 0.

(1) Is TT continuous?

For the operator TT of Example 14.9 and every x∈ℓ2,{x \in \ell^2,}

∥Tx∥22=∑k=0∞(k1+k2 x(k))2=∑k=1∞(k1+k2)2(x(k))2.\begin{aligned} \|Tx\|_2^2 &= \sum_{k=0}^{\infty} \Big( \frac{k}{1 + k^2} \, x^{(k)} \Big)^2 \\ &= \sum_{k=1}^{\infty} \Big( \frac{k}{1 + k^2} \Big)^2 \bigl( x^{(k)} \bigr)^2. \end{aligned}

Note 14.10. For k=0{k = 0} the factor k/(1+k2){k/(1 + k^2)} equals 0.{0.}

By Note 14.10, only the indices k≥1{k \ge 1} count, and we check that k/(1+k2)≤1/2{k/(1 + k^2) \le 1/2} for every k≥1.{k \ge 1.} This holds if and only if 2k≤1+k2,{2k \le 1 + k^2,} that is, 0≤k2−2k+1=(k−1)2,{0 \le k^2 - 2k + 1 = (k - 1)^2,} which is true. Hence

∥Tx∥22=∑k=1∞(k1+k2)2(x(k))2≤14∑k=1∞(x(k))2≤14∥x∥22,\begin{aligned} \|Tx\|_2^2 &= \sum_{k=1}^{\infty} \Big( \frac{k}{1 + k^2} \Big)^2 \bigl( x^{(k)} \bigr)^2 \\ &\le \frac{1}{4} \sum_{k=1}^{\infty} \bigl( x^{(k)} \bigr)^2 \\ &\le \frac{1}{4} \|x\|_2^2, \end{aligned}

that is, ∥Tx∥2≤12∥x∥2{\|Tx\|_2 \le \frac{1}{2} \|x\|_2} for every x∈ℓ2,{x \in \ell^2,} and TT is continuous.

(2) What is ∥T∥L?{\|T\|_{\mathcal{L}}?} Let x0=(0,1,0,0,… )∈ℓ2,{x_0 = (0, 1, 0, 0, \dots) \in \ell^2,} so that ∥x0∥2=1.{\|x_0\|_2 = 1.} Since x0(k)≠0{x_0^{(k)} \neq 0} only for k=1,{k = 1,} where x0(1)=1,{x_0^{(1)} = 1,}

∥Tx0∥2∥x0∥2=∥Tx0∥2=(∑k=0∞(k1+k2)2(x0(k))2)12=((11+1)2(x0(1))2)12=12.\begin{aligned} \frac{\|Tx_0\|_2}{\|x_0\|_2} = \|Tx_0\|_2 &= \biggl( \sum_{k=0}^{\infty} \Big( \frac{k}{1 + k^2} \Big)^2 \bigl( x_0^{(k)} \bigr)^2 \biggr)^{\frac{1}{2}} \\ &= \biggl( \Big( \frac{1}{1 + 1} \Big)^2 \bigl( x_0^{(1)} \bigr)^2 \biggr)^{\frac{1}{2}} = \frac{1}{2}. \end{aligned}

Hence, by (1) and (a) of Remark 14.2, ∥T∥L=12.{\|T\|_{\mathcal{L}} = \frac{1}{2}.}