Take the 2-minute tour ×
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.

I am trying to analyze the convergence of multidimensional infinite sums such as those in the following form:

$$\sum_{n=0}^{\infty}\sum_{m=0}^{\infty}\frac{1}{1+\alpha\exp (\beta n) \exp (\gamma m)}$$

where $\alpha,\beta,\gamma\in(0,\infty)$.

I'm quickly realizing though, that I have no experience with analyzing sums like this. I realize that because $\mathbb{N}^2$ is countable, I could in principle re-write this sum as a related one:

$\sum_{n=0}^{\infty} f(n)$

where $f(n)$ could even be guarenteed monotone decreasing. However, I've been having trouble getting anywhere with this.

Is there some well-established method of attacking convergence of multidimensional sums like this one? If so, I would very much appreciate being pointed in the right direction.

share|improve this question
add comment

2 Answers

up vote 7 down vote accepted

The question of absolute convergence can be answered by studying how quickly the function decays as $(n, m) \in \mathbb{N}_{\ge 0}^2$ gets further away from the origin. For example, let $A_k = \{ (n, m) : n + m = k \}$. Then we can write the sum as

$$\sum_{k \ge 0} \sum_{(n, m) \in A_k} \frac{1}{1 + \alpha \exp (\beta n) \exp (\gamma m)}.$$

The inner sum is a sum over $k + 1$ terms, each of which is bounded above by $\frac{1}{1 + \alpha \exp (\text{min}(\beta, \gamma) k)}$, so the whole sum is bounded by

$$\sum_{k \ge 0} \frac{C_1 k}{\exp (C_2 k)}$$

for some constants $C_1, C_2$, and this obviously converges.

The basic idea is to reduce the question to a one-dimensional question in a natural way. It is closely related to, for example, determining whether a multidimensional integral absolutely converges by bounding its integral over larger and larger spheres.

The question of conditional convergence is much more subtle, since it strongly depends on what order you sum the terms in (for example, in general you can't switch the inner and outer sums), and unlike the one-dimensional case there isn't a privileged order.

share|improve this answer
    
Does it matter that both of these are true: (1) it converges absolutely, and (2) all terms are non-negative? –  rwong May 9 '11 at 4:58
    
@rwong: I don't understand your question. –  Qiaochu Yuan May 9 '11 at 5:00
    
sorry, i'm referring to "the question of conditional convergence". please let me know if i misunderstood something. –  rwong May 9 '11 at 5:02
    
@rwong: that second comment refers to the general question of dealing with multidimensional sums, not to this particular sum. –  Qiaochu Yuan May 9 '11 at 5:16
add comment

If you just want to know if this converges or not, the answer is that it converges since $\alpha>0$ implies $$\frac1{1 + \alpha \exp(\beta n) \exp(\gamma m)} < \frac1{\alpha \exp(\beta n) \exp(\gamma m)}$$

$$\sum_{n=0}^{\infty} \sum_{m=0}^{\infty} \frac1{1 + \alpha \exp(\beta n) \exp(\gamma m)} < \sum_{n=0}^{\infty} \sum_{m=0}^{\infty} \frac1{\alpha} \exp(-\beta n) \exp(-\gamma m)$$

It is a geometric series and since $\beta,\gamma > 0$ the series converges and we have $$\sum_{n=0}^{\infty} \sum_{m=0}^{\infty} \frac1{1 + \alpha \exp(\beta n) \exp(\gamma m)} < \frac1{\alpha} \frac1{1-\exp(-\beta)} \frac1{1 - \exp(-\gamma)}$$

Hence the series converges.

share|improve this answer
add comment

Your Answer

 
discard

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.