Integral of e-x2


By James Thimont (P369) on Tuesday, April 20, 1999 - 10:13 pm :
Does anyone know the technique for integrating òe-x2 dx? Supposedly it uses techniques "not taught at school" so I'm not sure how to do it.
James
By Dave Sheridan (Dms22) on Tuesday, April 20, 1999 - 10:51 pm :
Hm. This depends on what limits you want to integrate between (there is no algebraic answer so you can only do a definite integral). I was taught a method for this at school, as part of Probability/Statistics A-Level. The ideas are quite straightforward.

Let I=ò-¥¥ e-x2 dx. Now square it. To avoid getting confused, we'll label one integral dy.

I2=ò-¥¥ò-infty¥ e-x2-y2dx dy

Now transform to polar co-ordinates (please tell me if you would like these explained). r2=x2+y2 and tanq = y/x.

I2=ò0¥ò02pe-r2r dr dq = p

So
I=
Ö
 

p
 

. By symmetry, replacing one of the limits by 0 will halve the result.

However, for any other limit there is no explicit way of calculating it - just approximating the integral, although this can be done to any degree of accuracy you want.

Hope this answers your question in part - if not, give some more information and we'll see what we can do.

Also, how did you come across this function? It normally occurs as either an example of a difficult integral, or involved in the Normal distribution; the latter has some very interesting properties. This is a discussion after all, so where would you like it to head?

-Dave


By James Thimont (P369) on Monday, April 26, 1999 - 08:01 pm :

Thanks Dave. I wanted to know if there was an algebraic method; to evaluate the indefinite integral.
James


By Dave Sheridan (Dms22) on Monday, April 26, 1999 - 09:25 pm :

...in which case I hope you didn't mind my long talk about the (rather restricted) definite one.

Just to re-emphasise the point, there is no nice way to express the indefinite integral; it can just be calculated to good (as good as you need!) precision. There are a variety of methods to do this but I don't imagine that's much use to you.

One reason it's not given in school is the notion of converging approximations isn't particularly developed below degree level; without it, numerical analysis isn't particularly helpful (try to answer the question "Why?" and you'll see what I mean...)

Thanks for an interesting question and sorry for there not being a completely satisfactory answer!

-Dave