I am given the following. hint Consider $f(x)=e^{-x} \phi(x)$.

I am unsure how to approach this problem.


Note that $f'(x)=0$ for all values of $x$ by construction. Hence $f(x)$ is constant. Given that $\phi(0)=0$ it must be true that $f(x)=0$ for all values of $x$ and hence that $\phi(x)=0$ for all values of $x$.

  • 1
    $\begingroup$ Then take $\phi(x)=e^{a+x}-e^a e^x$ for the second part of the question. $\endgroup$ – JPi Feb 1 '14 at 21:22
  • $\begingroup$ I see, forgot to use the hypothesis part of the question $\endgroup$ – grayQuant Feb 1 '14 at 21:23

If $$ \phi'(t)=\phi(t), $$ then $$ \mathrm{e}^{-t}\big(\phi'(t)-\phi(t)\big)=0. $$ But $$ 0=\mathrm{e}^{-t}\big(\phi'(t)-\phi(t)\big)=\left(\mathrm{e}^{-t}\phi(t)\right)' $$ and thus $\mathrm{e}^{-t}\phi(t)$ is a constant function, i.e., $$ \mathrm{e}^{-t}\phi(t)=\mathrm{e}^{-0}\phi(0)=0. $$ Thus $\phi(t)=0$, for all $t$.

Then consider $\varphi(t)=\mathrm{e}^{a+t}-\mathrm{e}^a\mathrm{e}^t$. Clearly, $\varphi(0)=0$ and $\varphi'(t)=\varphi(t)$, and hence $\varphi(t)=0$.


If we carefully observe the answers given here, it will be found that the answers make use of three crucially important properties of $f(x) = e^{x}$:

1) $f'(x) = f(x)$

2) $f(0) = 1$

3) $f(x) \neq 0$ for all $x$

and then using these properties it is proven that $f(x + y) = f(x)f(y)$.

Justification of existence of a function $f(x)$ with the above properties is effectively equivalent to development of theory of exponential and logarithmic functions.

It is of interest to prove the result without using the properties of function $f(x) = e^{x}$. We proceed to provide such a solution as follows.

Case 1: Let us suppose that there is a constant $a$ such that $\phi(a) = k > 0$ and consider the function $\psi(x) = \phi(x + a)$. Clearly $\psi'(x) = \psi(x)$ and $\psi(0) = \phi(a) = k > 0$. We now claim that $\psi(x) > 0$ for all $x$. This will lead to a contradiction because $\psi(-a) = \phi(0) = 0$.

To establish our claim let's first consider positive values of $x$. If $\psi(x)$ vanishes for some positive $x$ then by continuity of $\psi(x)$ there is a least such positive value of $x$ for which $\psi(x) = 0$. Let this value be called $b$. Thus we have $\psi(b) = 0$ and $\psi(x) \neq 0$ for $x \in [0, b)$. Then by mean value theorem we have $\psi'(c) = (\psi(b) - \psi(0))/b = -k/b < 0$ for some $c \in (0, b)$. Clearly then $\psi(c) = \psi'(c) < 0$. Now $\psi(0) = k > 0, \psi(c) < 0$ and hence by intermediate value theorem $\psi(x)$ vanishes at least once in $(0, c) \subset (0, b)$ and this contradicts that $b$ is the least positive value of $x$ for which $\psi(x) = 0$. It follows that $\psi(x) \neq 0$ for all $x > 0$. Since $\psi(0) = k > 0$ it follows that $\psi(x) > 0$ for all $x > 0$ (if $\psi(x) < 0$ then by IVT $\psi(x)$ would vanish somewhere in between).

Now consider the function $h(x) = \psi(-x)\psi(x)$. Clearly $$h'(x) = \psi'(x)\psi(-x) - \psi(x)\psi'(-x) = \psi(x)\psi(-x) - \psi(x)\psi(-x) = 0$$ so that $h(x)$ is constant and hence $\psi(x)\psi(-x) = h(x) = h(0) = \psi(0)\psi(0) = k^{2} > 0$. It follows that $\psi(-x)$ has same sign as that $\psi(x)$. It now follows $\psi(x) > 0$ for all $x$.

We have thus completed case 1.

Case 2: Let us now suppose that there is some constant $a$ for which $\phi(a) = k < 0$. In this case proceeding as before we can show that the function $\psi(x) = \phi(x + a)$ is such that $\psi(x) < 0$ for all $x$ and this contradicts the fact that $\psi(-a) = \phi(0) = 0$.

Since both the cases are ruled out we are left with the only alternative that $\phi(x) = 0$ for all $x$.

Summing it up, we can see that if there is a function $\phi(x)$ which satisfies $\phi'(x) = \phi(x)$ for all $x$ then $\phi(x)$ is of constant sign for all $x$ i.e. either $\phi(x)$ is identically zero for all $x$ or $\phi(x)$ is positive for all $x$ or $\phi(x)$ is negative for all $x$.

The simple constraint $\phi'(x) = \phi(x)$ thus controls the behavior of the function $\phi(x)$ in a very strong way (and fortunately for us, in a very positive way leading to theory of exponential and logarithmic functions).

Note that the reasoning in case 1 above also shows that a function $f(x)$ satisfying $f'(x) = f(x)$ and $f(0) = 1$ is such that $f'(x) = f(x) > 0$ for all $x$. Thus $f(x)$ is strictly increasing and positive and hence admits an inverse $g(x)$ which is also strictly increasing and then $g(1) = 0$ and by rule of derivative of inverse functions $g'(x) = 1/x$ so that $g(x) = \int_{1}^{x}(1/t)\, dt$ and then it is easy to show that $g(xy) = g(x) + g(y)$ and this relation leads to $f(x + y) = f(x)f(y)$ and thereby it justifies the existence of $f(x) = e^{x}$ as well as the property $e^{x + y} = e^{x}e^{y}$ is also established.

Update: There is another approach which is simpler than the one used above. We are to prove that if $\phi'(x) = \phi(x)$ for all $x$ and $\phi(0) = 0$ then $\phi(x) = 0$ for all $x$. Let us suppose for example that $\phi(a) \neq 0$ and consider the function $\psi(x) = \phi(a + x)\phi(a - x)$. Clearly we have $$\psi'(x) = \phi'(a + x)\phi(a - x) - \phi(a + x)\phi'(a - x) = \phi(a + x)\phi(a - x) - \phi(a + x)\phi(a - x) = 0$$ so that $\psi(x)$ is a constant function and hence $\psi(x) = \psi(0) = \phi^{2}(a) > 0$. It follows that $\phi(a + x)\phi(a - x) > 0$. Putting $x = a$ in this equation we get $\phi(2a)\phi(0) > 0$ which is contrary to the fact that $\phi(0) = 0$.

What we have established so far is that if a function $\phi(x)$ satisfies $\phi'(x) = \phi(x)$ for all $x$ and is such that it takes a non-zero value at some point then it does not vanish for any value of $x$. Then by continuity $\phi(x)$ is of constant sign. Thus if we assume $\phi'(x) = \phi(x)$ and $\phi(x)$ not identically zero then we can define another function $f(x) = \phi(x)/\phi(0)$. Then $f'(x) = f(x), f(0) = 1$ and $f(x) > 0$ for all $x$. If $y$ is a constant and we define $g(x) = f(x + y)/f(x)$ then $$g'(x) = \frac{f(x)f'(x + y) - f(x + y)f'(x)}{\{f(x)\}^{2}} = \frac{f(x)f(x + y) - f(x + y)f(x)}{\{f(x)\}^{2}} = 0$$ so that $g(x)$ is a constant and hence $g(x) = g(0) = f(y)$. We thus have $f(x + y) = f(x)g(x) = f(x)f(y)$ so that the functional equation of $f(x)$ is proved.


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.