3
$\begingroup$

During studying for an exam, I came across the following exercise:

For which $z \in \mathbb{C}$ does the series $$\sum\limits_{n=1}^{\infty} \frac{n^n z^n}{n!}$$ converge?

By using the ratio test (or alternatively: calculating the radius of convergence), I can see that the series converges for $|z|<\frac{1}{e}$ and diverges for $|z|>\frac{1}{e}$. However, I am not able to find out what happens if $|z|=\frac{1}{e}$. Does anyone have an idea?

$\endgroup$
  • $\begingroup$ See Stirling's approximation. $\endgroup$ – Lucian Feb 3 '15 at 23:14
  • $\begingroup$ @Lucian: Ok, thanks for the hint. Stirling approximation is something which I do not know yet. If this is the only way to solve the problem, then the problem is currently beyond my skills. $\endgroup$ – russoo Feb 3 '15 at 23:42
2
$\begingroup$

Claim: if $|z|=\frac1e$ but $z\ne\frac1e$, then the series converges.

Write $z=e^{i\theta-1}$. We can show the claim using Dirichlet's test, with $a_n = n^n/(e^nn!)$ and $b_n = e^{in\theta}$, if we can prove that $\{n^n/(e^nn!)\}$ is a decreasing sequence whose limit equals $0$. (The geometric series formula for $\sum_{n=1}^N e^{in\theta}$ shows that the boundedness condition is satisfied.) Equivalently, one needs to show that $$ \bigg\{\log \frac{e^nn!}{n^n} \bigg\} = \bigg\{ \sum_{k=1}^n \log k - (n \log n - n) \bigg\} $$ is increasing and tends to $\infty$. This can be done by looking at the difference sequence \begin{multline*} \bigg( \sum_{k=1}^n \log k - (n \log n - n) \bigg) - \bigg( \sum_{k=1}^{n-1} \log k - ((n-1) \log (n-1) - (n-1)) \bigg) \\ = \log n - \int_{n-1}^n \log x\,dx. \end{multline*} That difference is the area of the almost triangle-shaped region above $y=\log x$ and $y=\log n$ between $n-1\le x\le n$; it is certainly greater than the area of the rectangle with upper-left corner $(n-1,\log n)$ and lower-right corner $(n-\frac12,\log(n-\frac12))$, which one can show (using the mean value theorem) is greater than a constant times $\frac1n$. This lower bound for the difference sequence forces the original sequence to increase to infinity by comparison with the harmonic series.

The original series diverges when $z=\frac1e$, but that's difficult to motivate without knowing Stirling's approximation. Nevertheless you can show, using a similar method, that $$ \frac{n^n}{e^nn!}\cdot n $$ increases to infinity, which is enough to show that $\sum \frac{n^n}{e^nn!}$ diverges.

$\endgroup$
  • $\begingroup$ Unfortunately, neither did we study Dirichlet's test in our lecture (althought can understand how it works), nor did we study integration yet. Nevertheless, +1 since I trust you that your solution is correct. $\endgroup$ – russoo Feb 4 '15 at 0:28
1
$\begingroup$

You need to use something like the Generalized Dirichlet Test (proven in this answer).

This answer shows that $\left(1+\frac1n\right)^n$ is an increasing sequence. Therefore, $$ \frac{\cfrac{(n+1)^{n+1}}{e^{n+1}(n+1)!}}{\cfrac{n^n}{e^nn!}}=\frac{\left(1+\frac1n\right)^n}e\lt1 $$ and so $\dfrac{n^n}{e^nn!}$ is a decreasing sequence.

When $\left|z\right|=\frac1e$, we can write $z=\frac{e^{i\theta}}e$. Thus, $$ \begin{align} \sum_{n=0}^\infty\frac{n^nz^n}{n!} &=\sum_{n=0}^\infty\frac{n^n}{e^nn!}e^{in\theta} \end{align} $$ converges when $$ \begin{align} \left|\,\sum_{n=0}^{N-1}e^{in\theta}\,\right| &=\left|\frac{e^{iN\theta}-1}{e^{i\theta}-1}\right|\\ &\le\frac2{\left|e^{i\theta}-1\right|}\\ &=\frac2{\left|ez-1\right|} \end{align} $$ is finite; i.e. $z\ne\frac1e$.

Therefore, $$ \sum_{n=0}^\infty\frac{n^nz^n}{n!} $$ converges when $\left|z\right|=\frac1e$, except when $z=\frac1e$.

$\endgroup$
-1
$\begingroup$

Apply ratio test: $$\left| {\frac{z \cdot n}{{n+1}}}\right| $$ Converges when $|z|<1$, diverges when $|z|\le1$

$\endgroup$
  • $\begingroup$ The ratio test is inconclusive in the case that the question is about. $\endgroup$ – robjohn Nov 28 '15 at 10:16

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.