1
$\begingroup$

Let $Tf(x):=\int_0^x f(t)dt$ be an integral Operator ($T:L_2[0,1]\rightarrow L_2[0,1]$). I am trying to find the eigenvalues and eigenfunctions of $S:=T^*T:L_2[0,1]\rightarrow L_2[0,1]$. So far I know that $T^*g(t)=\int_t^1 g(x)dx$ as well as $Sf(x)=\lambda f(x)\Leftrightarrow -f(x)=\lambda f''(x)$ and therefore $f(x)=Ce^{\frac{1}{\sqrt{\lambda}}x}$. Is this correct? How do I get $C$ and what should I do next?

$\endgroup$
  • 2
    $\begingroup$ You don't have to bother for C because an eigenvector is only known up to a mult.constant $\endgroup$ – Jean Marie May 25 '16 at 21:58
  • $\begingroup$ True, silly me! $\endgroup$ – Vorhang May 26 '16 at 5:57
4
$\begingroup$

Suppose you have a non-trivial solution of $$ T^*Tf = \int_{t}^{1}\int_{0}^{s}f(y)dy ds = \lambda f(t) $$ Then $\lambda \ne 0$ because the above would give $f=0$ after differentiating a couple of times. For $\lambda \ne 0$, any solution of the above must satisfy $$ \lambda f'' = -f \\ f(1)=0,\;\; f'(0)=0. $$ Any non-trivial solution of the above ODE cannot satisfy $f(0)=0$ because $f'(0)=0$ and $f(0)=0$ can hold for a solution of an ODE such as the above iff $f\equiv 0$. So a non-trivial solution may be normalized to satisfy $f(0)=1$ and $f'(0)=0$, which means any eigenfunction must be a non-zero scalar multiple of $$ \varphi_{\lambda}(t)=\cos(t/\sqrt{\lambda}). $$ The eigenvalue equation is then given by $\varphi_{\lambda}(1)=0$, or $\cos(1/\sqrt{\lambda})=0$, $\lambda \ne 0$. The eigenvalues are $$ \lambda_n = \frac{4}{(2n+1)^2\pi^2},\;\;\; n=0,1,2,3,\cdots, $$ and the corresponding eigenfunctions are $$ \cos\left(\frac{(2n+1)\pi}{2}t\right) $$ These are the only possible eigenfunctions, and you can check that they work.

$\endgroup$
  • $\begingroup$ Thanks! I still don't understand why $f$ must satisfy $f(1)=0$ and $f'(0)=0$. Where do these side conditions come from? $\endgroup$ – Vorhang May 26 '16 at 6:01
  • $\begingroup$ @Vorhang : Examine $f(t)=\frac{1}{\lambda}\int_{t}^{1}\int_{0}^{s}f(y)dyds$. Clearly $f(1)=0$. And $f'(t)=-\int_{0}^{t}f(y)dy$ must vanish at $t=0$. $\endgroup$ – DisintegratingByParts May 26 '16 at 6:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.