Approximating sum by Gaussian integral - how big is the error?

1.4k Views Asked by At

I have the following infinite sum:

$$S=\sum_{n=1}^{\infty}e^{-an^2}$$

Where $a$ is a positive constant. Is there a simple way to estimate the error when approximating $S$ by:

$$S \approx \int_0^ \infty e^{-ax^2}dx .$$

Does this depend at all on the value of $a$?

1

There are 1 best solutions below

0
On BEST ANSWER

A very simple estimate (which is what you were asking for) would be the following: $f(x) = e^{-a x^2}$ is strictly decreasing, therefore $$ f(n+1) < \int_n^{n+1} f(x)\,dx < f(n) $$ for all $n \in \mathbb N_0$. Summation gives $$ \sum_{n=1}^\infty f(n) < \int_0^\infty f(x)\,dx < f(0) + \sum_{n=1}^\infty f(n) $$ So the difference between sum and integral is at most $f(0) = 1$.