4
$\begingroup$

Some friends and I are having trouble with a functional equation problem :

If $f : (0,1) \to \mathbb R$ is a positive continuous function satisfying $$\int_t^1 f(x)f\left(\frac{t}{x}\right)dx = \sqrt{t}$$ for all $t$, then $f(x) = \sqrt{\frac{2x}{\pi(1-x^2)}}$.

The function is not in $L^2$ or any nice Hilbert / normed space, so functional analysis techniques seem out of reach. I've tried various changes of variables, like $y = x /\sqrt t$ and $y = t/x$, but none of these yielded anything valuable. I've also tried integrating over u, but it didn't do much.

The problem might reduce to proving that two solutions $f,g$ have to be equal almost everywhere / on a dense subset of $(0,1)$, as we've been able to show that the given function is indeed a solution.

Any ideas would be welcome !

$\endgroup$
2
  • $\begingroup$ Welcome to Mathematics SE. Take a tour. You'll find that simple "Here's the statement of my question, solve it for me" posts will be poorly received. What is better is for you to add context (with an edit): What you understand about the problem, what you've tried so far, etc.; something both to show you are part of the learning experience and to help us guide you to the appropriate help. You can consult this link for further guidance. $\endgroup$ Commented Aug 2, 2022 at 18:15
  • 2
    $\begingroup$ Let $f_0(x) = \sqrt{2x/(\pi-\pi x^2)}$. Two quick ideas: 1. assume for contradiction $\overline{f}$ is another such function that differs from $f_0$ on a subset of $(0,1)$ of non-zero lebesgue measure. Since the functions are positive, you may be able to combine integrals and get some measure-theoretic contradiction 2. As $f$ is continuous, you may be able to use the first fundamental theorem of calculus. Although, you'd have to deal with $t$ being in the integrand. Maybe a way around this is to convert to double integrals with the function $g(x,t) = f(x)f(t/x)$ or something similar $\endgroup$ Commented Aug 2, 2022 at 19:20

1 Answer 1

1
$\begingroup$

Let's make the change of variables $x=e^{-\xi}$, $t=e^{-\tau}$ in the integral equation: $$ \int_t^{1}f(x)f\!\left(\frac{t}{x}\right)dx=\int_0^{\tau}f(e^{-\xi})f(e^{-(\tau-\xi)})e^{-\xi}\,d\xi=e^{-\tau/2}. \tag{1} $$ Defining $g(z):=f(e^{-z})$, we can rewrite $(1)$ as $$ \int_0^{\tau}g(\tau-\xi)g(\xi)e^{-\xi}\,d\xi=e^{-\tau/2}. \tag{2} $$ Taking the Laplace transform of both sides of $(2)$, we get $$ G(s)G(s+1)=\frac{1}{s+\frac{1}{2}}, \tag{3} $$ where $G(s):=\mathcal{L}\{g(\xi)\}(s)$.

The positive solution to the functional equation $(3)$ is (see the Appendix) $$ G(s)=\frac{1}{\sqrt{2}}\frac{\Gamma\!\left(\frac{s}{2}+\frac{1}{4}\right)}{\Gamma\!\left(\frac{s}{2}+\frac{3}{4}\right)}. \tag{4} $$

Finally, taking the inverse Laplace transform of $G(s)$, we obtain$^{(*)}$ $$ g(\xi)=\sqrt{\frac{2e^{-\xi}}{\pi\left(1-e^{-2\xi}\right)}}, \tag{5} $$ hence $$ f(x)=\sqrt{\frac{2x}{\pi\left(1-x^2\right)}}. \tag{6} $$


Appendix: Solution of Eq. $(3)$

To solve Eq. $(3)$, let's rewrite $G(s)$ as $$ G(s)=\frac{G(s)G(s+1)\cdots G(s+2n-1)}{G(s+1)G(s+2)\cdots G(s+2n)}\,G(s+2n)\qquad(n>1). \tag{A.1} $$ Using Eq. $(3)$, we can rewrite $(\text{A.}1)$ as \begin{align} G(s)&=\frac{\left(s+\frac{3}{2}\right)\left(s+\frac{7}{2}\right)\cdots\left(s+2n-\frac{1}{2}\right)} {\left(s+\frac{1}{2}\right)\left(s+\frac{5}{2}\right)\cdots\left(s+2n-\frac{3}{2}\right)}\,G(s+2n) \\ &=\frac{\left(\frac{s}{2}+\frac{3}{4}\right)\left(\frac{s}{2}+\frac{7}{4}\right)\cdots\left(\frac{s}{2}+n-\frac{1}{4}\right)} {\left(\frac{s}{2}+\frac{1}{4}\right)\left(\frac{s}{2}+\frac{5}{4}\right)\cdots\left(\frac{s}{2}+n-\frac{3}{4}\right)}\,G(s+2n). \tag{A.2} \end{align} Now, using the identity $x(x+1)\cdots(x+n-1)=\frac{\Gamma(x+n)}{\Gamma(x)}$, we can rewrite $(\text{A}.2)$ as $$ G(s)=\frac{\Gamma\!\left(\frac{s}{2}+\frac{1}{4}\right) \Gamma\!\left(\frac{s}{2}+n+\frac{3}{4}\right)} {\Gamma\!\left(\frac{s}{2}+\frac{3}{4}\right) \Gamma\!\left(\frac{s}{2}+n+\frac{1}{4}\right)}\,G(s+2n). \tag{A.3} $$ Since the RHS of $(\text{A}.3)$ does not depend on $n$, we can simplify it using the fact that Eq. $(3)$ implies $G(s+2n)\sim\frac{1}{\sqrt{s+2n}}$ as $n\to\infty$. Taking this limit with the help of WolframAlpha, we finally obtain $$ G(s)=\lim_{n\to\infty}\frac{\Gamma\!\left(\frac{s}{2}+\frac{1}{4}\right) \Gamma\!\left(\frac{s}{2}+n+\frac{3}{4}\right)} {\Gamma\!\left(\frac{s}{2}+\frac{3}{4}\right) \Gamma\!\left(\frac{s}{2}+n+\frac{1}{4}\right)}\,\frac{1}{\sqrt{n+2s}} =\frac{1}{\sqrt{2}}\frac{\Gamma\!\left(\frac{s}{2}+\frac{1}{4}\right)} {\Gamma\!\left(\frac{s}{2}+\frac{3}{4}\right)}. \tag{A.4} $$ Direct substitution shows that $(\text{A}.4)$ indeed satisfies Eq. $(3)$: $$ G(s)G(s+1)=\frac{1}{2}\frac{\Gamma\!\left(\frac{s}{2}+\frac{1}{4}\right)\Gamma\!\left(\frac{s+1}{2}+\frac{1}{4}\right)} {\Gamma\!\left(\frac{s}{2}+\frac{3}{4}\right)\Gamma\!\left(\frac{s+1}{2}+\frac{3}{4}\right)} =\frac{\Gamma\!\left(\frac{s}{2}+\frac{1}{4}\right)} {2\,\Gamma\!\left(\frac{s}{2}+\frac{5}{4}\right)} =\frac{1}{s+\frac{1}{2}}. \tag{A.5} $$


$^{(*)}$ A.P. Prudnikov, Yu.A. Brychkov, O.I. Marichev, Integrals and Series, Volume 5: Inverse Laplace Transforms, formula $3.1.1.1$.

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.