4
$\begingroup$

Let $V$ be the vector space of all functions from $\Bbb{R}$ into $\Bbb{R}$ which are continuous, i.e. $V=C^0(\Bbb{R},\Bbb{R})$. Let $T$ be the linear operator on $V$ defined by $$\forall f \in V, \forall x\in \Bbb{R}, \quad [T(f)](x)=\int_0^xf(t)dt.$$ Prove that $T$ has no characteristic values.

My attempt: Assume towards contradiction, $\lambda \in F$ such that $\lambda$ is eigenvalue of $T$. Then $\exists g\in V\setminus \{0_V\}$ such that $T(g)=\lambda\cdot g$. So $[T(g)](x)=\int_0^xg(t)dt=(\lambda \cdot g)(x)=\lambda\cdot g(x)$, $\forall x\in \Bbb{R}$. By elementary property of integration, $[T(g)](0)=0=\lambda \cdot g(0)$. Which implies $\lambda =0$ or $g(0)=0$. Since $g\neq 0_V$, $\exists y\in \Bbb{R}$ such that $g(y)\neq 0$. So $[T(g)](y)=\int_0^yg(t)dt=\lambda \cdot g(y)$. How to reach contradiction?

$\endgroup$
2
  • $\begingroup$ If $g$ is an eigefunction corresponding to $\lambda$ then $g=\lambda g'$ and $g(0)=0.$ Thus $g(x)=e^{x/\lambda}+c.$ $\endgroup$ Commented Nov 7, 2022 at 19:14
  • $\begingroup$ If $g=\lambda g'$ and $g(0)=0$, then $g=0$ which is a contradiction because $\lambda$ is only an eigenvalue of $g$ if $g$ is not $0_V$. That's your contradiction $\endgroup$
    – Sam
    Commented Nov 7, 2022 at 19:26

3 Answers 3

2
$\begingroup$

If $\lambda=0$, you have $$\tag1 \int_0^xg(t)\,dt=0 $$ for all $x$. Differentiating, since $g$ is continuous, gives $g(x)=0$. This can be done for any $x$, so $g=0$.

If $\lambda\ne0$, from $$\tag2 \int_0^xg(t)\,dt=\lambda\,g(x), $$ the continuity of $g$ makes the left-hand-side differentiable. We also get $g(0)=0$. Differentiating $(2)$, $$ g (x)=\lambda \,g'(x). $$ This gives $g(x)=k\,e^{x/\lambda }$. The condition $g(0)=0$ then gives $g=0$.

$\endgroup$
4
  • $\begingroup$ You were faster than me, but $g(x)=ke^{x/\lambda}.$ On the other hand I have forgotten the multiplicative constant in my comment. $\endgroup$ Commented Nov 7, 2022 at 19:23
  • $\begingroup$ Hi Ryszard, thanks for noticing! I've made that mistake many times :) $\endgroup$ Commented Nov 7, 2022 at 19:31
  • $\begingroup$ Hi Martin, now I feel like a hawk eye on hard surface tennis court. :) $\endgroup$ Commented Nov 7, 2022 at 19:44
  • $\begingroup$ :D $ \ \ \ \ \ $ $\endgroup$ Commented Nov 7, 2022 at 21:49
1
$\begingroup$

Suppose $\lambda$ is an eigenvalue for $T$ and let $g$ be a corrisponding eigenvector i.e.

$$\label{eq}\tag{1}T(g)=\int_0^x g(t)dt=\lambda g(x)$$

Since $g$ is continuous $T(g)$ is differentiable because is a primitive of $g$, since $T(g)=\lambda g(x)$ you get that $g$ is $C^1$ (actually $C^\infty$). Differentiating both sides of \eqref{eq} you get the equation $$\label{eq2}\tag{2}g=\lambda g'$$ with starting value $g(0)=0$ as you noted. (Notice that $\lambda=0$ here implies $g\equiv 0$, so $0$ is not an eigenvalue).

If $\lambda \neq 0$ we may now \eqref{eq2}:

$$(\log g)'=1/\lambda \implies \log g(x)= x/\lambda + C\implies g(x)=Ae^{x/\lambda}$$

and you get contradiction by evaluating $g$ at $0$.

$\endgroup$
2
  • $\begingroup$ Thank you so much for the answer. It’s been long time I have done integration. I can’t fill in details of following: $(1)$ $T(g)$ is differentiable because is a primitive of $g$. $(2)$ since $T(g)=\lambda g(x)$ you get that $g$ is $C^1$ (actually $C^\infty$). $(3)$ $[T(g)]’=g$. $(4)$ $(\log g)'=1/\lambda \implies \log g(x)= x/\lambda + C$. I’m sorry for asking so elementary question. $\endgroup$
    – user264745
    Commented Nov 8, 2022 at 8:21
  • 1
    $\begingroup$ (1) and (3) This is just the fundamental theorem of calculus. (2) You get $g\in C^\infty$ because the left hand side of $T(g)=\lambda g$ has "one derivative more" then the right hand-side and the right hand-side has the same number of derivatives as the left, if it makes sense. (4) Is basic integration w.r.t. the $x$ variable (keeping in mind that $\int f'=f$). $\endgroup$
    – theGrey
    Commented Nov 8, 2022 at 12:10
1
$\begingroup$

Remark that $T(g) '(x) =g(x) $, So $g=\lambda g' $, Now $\lambda =0\implies g=0$,So $\lambda \neq 0$,$$g(x)=c\exp(\frac{1}{\lambda} x) $$But $0=Tg(0)=\lambda g(0)=\lambda c$, hence $c=0$,Contradiction!

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .