Let $k:[0,1]^2\rightarrow (0,+\infty)$ be a continuous function and let $f,g:[0,1]\rightarrow (0,+\infty)$ be measurable functions. We assume that $$\forall x\in [0,1],\quad f(x)=\int_0^1 k(x,y) g(y) dy,\hskip15pt g(x)=\int_0^1 k(x,y) f(y) dy. $$ We want to prove that $f=g$.

Note that defining the operator $K$ by $(Ku)(x)=\int_0^1 k(x,y) u(y) dy$, this amounts to proving that the assumption $f=K^2f$ implies $f=Kf$ for a positive $f$. In other words, if $(Id-K^2)f=0$, it should imply $(Id-K)f=0$ whenever $f$ is positive. If the operator $Id+K$ were one-to-one, it would imply the sought result, but I do not see why this invertibility property should hold.

  • $\begingroup$ Something wrong with my answer? $\endgroup$
    – username
    Apr 19 at 19:49

This is related to Leray's Eigenvalue Problem . Under your hypothesis, there is at least one positive eigenvalue, with a corresponding non negative eigenvector $u$ such that $$ Ku = \lambda u, $$ by Arzela-Ascoli's Theorem and Brouwer's Fixed point applied to $$ F(u)(x)= \frac{\int_0^1 k(x,y) u(y) dy}{\int_0^1\int_0^1 k(t,y) u(y) dy dt}. $$ Now, you have $K^2$ a positive, compact, linear operator acting on $L^{\ldots}(0,1)$ such that both $$ K^2(f)=f \mbox{ and } K^2(u) = \lambda^2 u, $$ that is, two stricly positive eigenvalues with positive eigenvectors. By the uniqueness property of such eigenpairs (Perron-Frobenius / Krein-Rutman Theorem) $\lambda=1$, and $f=u$, so indeed, $Kf=f$.


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.