8
$\begingroup$

Because it comes up every so often, I was wondering if there was a nice general expansion for the $n$th derivative of $f(e^x)$.

$\frac{\mathrm d}{\mathrm dx}f(e^x)=e^xf'(e^x)$

$\frac{\mathrm d^2}{\mathrm dx^2}f(e^x)=e^xf'(e^x)+e^{2x}f''(e^x)$

$\frac{\mathrm d^3}{\mathrm dx^3}f(e^x)=e^xf'(e^x)+3e^{2x}f''(e^x)+e^{3x}f'''(e^x)$

$\vdots$

One can see that we end up with the form

$$\frac{\mathrm d^n}{\mathrm dx^n}f(e^x)=\sum_{k=1}^nC(n,k)e^{kx}f^{(k)}(e^x)$$

for some constants $C(n,k)$. I've found that it reduces to solving the following recurrence relation:

$$C(n,k)=\begin{cases}0,&n<k\\1,&n>0\land k=1\\kC(n-1,k)+C(n-1,k-1),&n>0\land k>1\end{cases}$$

From which you can deduce things such as $C(n,2)=2^{n-1}-1$.

Is there a nice general solution?

$\endgroup$
4
$\begingroup$

Short answer: No, there is no nice solution, at least in terms of elementary functions. But if you consider the Stirling numbers to be nice (which is a good practice), then the solution is very nice.

As mentioned in the comments, $C(n,k)$ are the Stirling numbers of the second kind, ${n \brace k}$. They have the following "closed form:" $$ {n\brace k}=\frac1{k!}\sum_j(-1)^j\binom{k}{j}(k-j)^n. $$ If you like, you can substitute that mess for $C(n,k)$, but I think it is better to leave it as ${n \brace k}$. This is as simple as it gets.

$\endgroup$

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.