It can be shown, and is a problem in Rudin's Principles of Mathematical Analysis (Chapter 8), that when $f$ is continuous, and $f(x)f(y) = f(x+y)$, $f$ is a function of the form $e^{cx}$.

Must this necessarily be true when the right hand side is a different function $g$, and then $g = f$? If so, is there a good reason why, in this particular functional equation, it doesn't matter whether it's $f$ or $g$ on each side--but it matters in others?

EDIT: $f$ should also be assumed to be never zero.

  • $\begingroup$ Well I would say $e^{x} \cdot e^{y}=e^{x+y}$ still satisfies $f(x) \cdot f(y)=g(x+y)$ (because g could equal f) but there could be other functions to satisfy it also. For some reason I was thinking about some trig identities but I haven't found any that work yet. $\endgroup$ – randomgirl Apr 1 '15 at 19:11
  • $\begingroup$ If any of the two functions are $0$ at any point, they're both the constant function $0$. $\endgroup$ – Alice Ryhl Apr 1 '15 at 19:12
  • $\begingroup$ If $f$ is $1$ at any point, then $g(x)=f(x+\delta)$ for some $\delta\in\mathbb R$. $\endgroup$ – Alice Ryhl Apr 1 '15 at 19:13
  • $\begingroup$ Then let's set them nonzero. In fact Rudin does, in that problem, but I neglected to write it. The question's edited now. Also, randomgirl, what I'm wondering is not what the solution is, but if the solution can be DIFFERENT from $e^{cx}$. That's edited too. $\endgroup$ – Drew N Apr 1 '15 at 19:16

For every $x$,

$$ f(x)f(0) = g(x + 0) = g(x) $$

so $g(x) = Cf(x)$, where $C = f(0)$.

Then, you have

$$ f(x)f(y) = Cf(x + y) $$

and further, we can obviously write $f(x) = \dfrac{1}{C}f(x)f(0)$.

Now, if we use the limit definition of the derivative:

$$ f'(x) = \lim_{h \to 0} \frac{f(x+h) - f(x)}{h} = \frac{\dfrac{1}{C}f(x)f(h) - \dfrac{1}{C}f(x)f(0)}{h} = \dfrac{1}{C}f(x)\dfrac{f(h) - f(0)}{h} = \dfrac{1}{C}f'(0)f(x) $$

So solutions are just $f(x) = Ce^{\alpha x}$; i.e. you get basically the same result, except that you are not forced to take $f(0) = 1$.

  • 1
    $\begingroup$ Ah, of course. Yes. A concrete example is $(2e^{x})(2e^{y}) = 4e^{x+y} = C*2e^{x+y}$, where C = 2. $\endgroup$ – Drew N Apr 1 '15 at 19:32
  • 1
    $\begingroup$ ...And what if $f$ is not differentiable? Just replace the equation $f(x) f(y) = Cf(x+y)$ by the equivalent equation $\frac{f(x)}{C} \frac{f(y)}{C} = \frac{f(x+y)}{C}$ and apply the already-known result to $f/C$. $\endgroup$ – Najib Idrissi Apr 1 '15 at 19:47
  • $\begingroup$ Nice addition @NajibIdrissi thanks! $\endgroup$ – BaronVT Apr 1 '15 at 20:08

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.