Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I would like to prove that this function (with the $a, b\in \mathbb{R}$ satisfying $a < b$),

$$f(x)=\begin{cases} \exp\left(-\frac{1}{x-a}+\frac{1}{x-b}\right), & \text{if },x\in (a,b)\,, \\ 0, & \text{if }x\notin (a,b)\text{} \end{cases}$$

is class $C^\infty (\mathbb{R}).$

But how?

share|cite|improve this question
Of course, the problem is at $a$ and $b$. You can show by induction that the $d$-th derivative of $f$ on $(a,b)$ is of the form $R_d(x)\exp\left(-\frac 1{x-a}+\frac 1{x-b}\right)$, where $R_d$ is a rational fraction. – Davide Giraudo Dec 31 '11 at 0:41
@DavideGiraudo The main obstacle I see here is that $R_d(x)$ has poles at $a$ and $b$. – Alex Becker Dec 31 '11 at 1:01
up vote 4 down vote accepted

Suppose $a=0$ and $b=1$, because the same reasoning applies in general. So we have $f(x)=\exp\left(\frac{1}{x(x-1)}\right)$ on $(0,1)$. Compositions of $C^\infty$ functions are $C^\infty$, so the only possible problems are at $0$ and $1$. Since $f(x)=f(1-x)$, it is enough to handle $0$.

Let's start by showing that $f'(0)=0$. The left-hand difference quotients at $0$ are $0$, so it is enough to show that $\lim\limits_{x\searrow 0}\frac{f(x)}{x}=0$. Note that $ \frac{1}{x(x-1)}<-\frac{1}{x}$ for $x\in(0,1)$, so it is enough to show that $\lim\limits_{x\searrow 0}\frac{e^{-1/x}}{x}=0$. This is probably easiest to see by a change of variables, $t=1/x$, to yield $\lim\limits_{x\searrow 0}\frac{e^{-1/x}}{x}=\lim\limits_{t\to\infty}\frac{e^{-t}}{1/t}=\lim\limits_{t\to\infty}\frac{t}{e^t}=0$.

So far we know that $f'$ exists everywhere. Now as Davide Giraudo indicates in a comment, you can show by induction that in $(0,1)$, $f^{(k)}(x)=R_k(x)f(x)$ for some rational function $R_k$ having poles only at $0$ and $1$. I'll omit proof of this. Suppose that for some $k$ we know that $f^{(k)}(0)=0$. To show that $f^{(k+1)}(0)=0$, we need to show that $\lim\limits_{x\searrow 0}\frac{R_k(x)f(x)}{x}=0$. Note that $R_k(x)=\frac{g(x)}{x^{n}}$ for some integer $n$ and a function $g$ that is continuous at $0$. Again, since $f(x)<e^{-1/x}$, it suffices to show that $\frac{e^{-1/x}}{x^{n+1}}\to 0$ as $x\searrow 0$, and this is straightforward from the same change of variables $t=1/x$. (The base case $k=1$ wasn't really necessary to show since we are already given the base case $k=0$, but I though it might be helpful to start with the simplest case and make it more explicit.)

Again, since $f(x)=f(1-x)$, this also gives $f^{(k)}(1)=0$ for all $k$, and this shows (in sketch form) why $f$ is infinitely differentiable.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.