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.

How do we numerically integrate a rapidly decaying exponential function? A simple Gauss quadrature which is based on approximating the function by polynomial, I think will not work, since rapidly decaying exponential function will require more terms in the series. Is there a series approximation for the function?

share|improve this question
    
Did you try the variants of the basic Gauss quadrature like Gauss–Laguerre quadrature and Gauss–Hermite quadrature? –  user17762 Dec 24 '10 at 5:18
    
I haven't applied them yet but aren't they of different form? Will they work for a rapidly decaying function? I have the integral of the form $$\int f(x)e^{-kx}$$, where k lies in the range of [10,50] –  vijay Dec 24 '10 at 6:05
    
@vijay: is $f(x)$ smooth, or does it have singular behavior (e.g. at the left endpoint)? –  J. M. Dec 24 '10 at 8:08
    
@J.M. f(x) is smooth. Ok, I think I will be more clear here. The actual integration to be done goes like this,$$\int (\int f(v) e^{-kvx}dv \int g(v) e^{-kvx}dv)dx$$. Integrands with $f(v)$ & $g(v)$ can only be numerically evaluated. They have 10th order polynominal in $v$ in the denominator. –  vijay Dec 24 '10 at 8:58
    
...and how sure are you that the denominator polynomial has no zeroes within the interval of integration? –  J. M. Dec 24 '10 at 9:01
show 2 more comments

1 Answer

Variable substitution to make $k=1$, and use Gauss-Laguerre. For more advanced method, the double exponential transformation maybe relevant.

share|improve this answer
    
I would say that vanilla DE is overkill here if the integrand already has a decaying exponential factor; however, Mori and Takahasi do have a modified version suitable for integrands with decaying exponentials; see their paper for details. Otherwise, Gauss-Laguerre ought to work well here, unless the $f(x)$ in $f(x)\exp(-kx)$ has wonky behavior on its own. The problem with Gauss-Laguerre is that it doesn't admit a natural error estimate; one has to usually compare the results for $n$ nodes and $n+1$ nodes. –  J. M. Dec 24 '10 at 8:07
    
I'll give a try and get back to you, if you don't mind.. thanks! –  vijay Dec 24 '10 at 8:56
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.