$$f(x_{2}-x_{1})= f(x_{2}-y)f(y-x_{1})$$ I need to find a unique solution to this algebraic equation. Any hints as to how to proceed. The Exponential function works in this case, but I explicitly need to see the details. Thanks in advance.

| cite | improve this question | | | | |
  • $\begingroup$ There is no unique solution, as $f(x)=b^x$ works for any $b\in(0,1)\cup(1,\infty)$. $\endgroup$ – Ben W Jan 22 '19 at 11:33
  • $\begingroup$ @BenW Why do you exclude the case $b=1$? $\endgroup$ – Jose Brox Jan 22 '19 at 11:39
  • $\begingroup$ Are $x_1,x_2$ supposed to be fixed real numbers, or is the equation supposed to hold for all $x_1,x_2\in\mathbb{R}$? $\endgroup$ – Jose Brox Jan 22 '19 at 11:44
  • $\begingroup$ @JoseBrox just habit $\endgroup$ – Ben W Jan 22 '19 at 16:39

I'm going to suppose that $x_1,x_2\in\mathbb{R}$ are variables and not fixed real numbers, and that $f$ is differentiable.

First, if $y=x_2$ then $f(x_2-x_1)=f(0)f(x_2-x_1)$, hence $f(0)=1$ unless $f(x_2-x_1)=0$ for every $x_1,x_2$, i.e., unless $f(x)=0$ in all of $\mathbb{R}$.

Now suppose $f\neq0$ and let us differentiate the equation with respect to $y$:


so $$f'(x_2-y)f(y-x_1)=f(x_2-y)f'(x_1-y)$$ and substituting $y=x_1$ we get


and therefore

$$f'(x)=\alpha_f f(x) \,\,\,\, (1)$$

for some constant $\alpha_f:=f'(0)\in\mathbb{R}$.

Since the derivative of $f$ is proportional to $f$, the unique family of solutions for (1) when $f\neq0$ is $f(x)=a^x$ for $a\in(0,\infty)$.

Finally we check that $f(x)=0$ and $f(x)=a^x$, $a>1$ are solutions to the original equation, so they are the only ones.

| cite | improve this answer | | | | |
  • $\begingroup$ OP is not assuming smoothness. $\endgroup$ – Kavi Rama Murthy Jan 22 '19 at 12:05
  • $\begingroup$ @KaviRamaMurthy Oh, yes, I forgot to add that explicitly. I'll edit, thanks! $\endgroup$ – Jose Brox Jan 22 '19 at 12:14

The equation is equivalent to $f(a+b)=f(a)f(b)$ which is closely related to $g(a+b)=g(a)+g(b)$. It is well know that there are pathological examples of functions satisfying the equation which are not even measurable. However the only continuous solutions are the exponential functions mentioned in the comments. [Note that $f(a)=0$ for some $a$ implies $f \equiv 0$. If $f$ is continuous and never vanishes then it is always positive and we can take logarithms to reduce the given equation to $g(a+b)=g(a)+g(b)$ where $g=\log \, f$. In this case $g(x)=cx$ for some $c$ and $f(x)=e^{cx}$].

| cite | improve this answer | | | | |

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.