home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!unipalm!uknet!pavo.csi.cam.ac.uk!camcus!gjm11
- From: gjm11@cus.cam.ac.uk (G.J. McCaughan)
- Newsgroups: sci.math
- Subject: Re: Help with Calculus
- Message-ID: <1992Sep4.214816.22940@infodev.cam.ac.uk>
- Date: 4 Sep 92 21:48:16 GMT
- References: <18531.2aa39479@levels.unisa.edu.au>
- Sender: news@infodev.cam.ac.uk (USENET news)
- Organization: U of Cambridge, England
- Lines: 41
- Nntp-Posting-Host: bootes.cus.cam.ac.uk
-
- In article <18531.2aa39479@levels.unisa.edu.au>, 8321207d@levels.unisa.edu.au writes:
-
- > 1) Let Phi(a)=Int from 0 to Inf of [exp(-ax)*Sinx/x]dx. Explain briefly
- > why it is plausible to write Phi(a)=Pi/2 - ArcTan a. Could this
- > result be used to evaluate Int from 0 to Inf of [Sinx/x]dx ?
-
- Well, how about differentiating under the integral sign? You get (d/da)Phi
- equal to integral of -x.exp(-ax).sin(x)/x dx
- = integral of -exp(-ax).sin(x) dx
- and you can do this, for instance, with complex numbers. (sin x is the imaginary
- part of exp(ix), so you want the imaginary part of the integral of -exp((a+i)x);
- this is dead easy.)
- Since you only ask for "Why it is plausible" I don't need to prove that it's
- *valid* to differentiate under the integral sign; but there are tests for when
- you can do it, and I'm sure one of them applies here...
-
- Um, you still have to work out what the constant is; try letting a tend to
- infinity; then the integral tends to 0 because it is dominated by the integral
- of exp(-ax), which tends to 0.
-
- If needed I can provide all the gory details.
-
- > 2) Let Psi(r)=Int from 0 to Inf of [Sin rx/x(1+x^2)]dx for r >= 0.
- > Explain briefly why it is plausible to suggest that
- > Psi''(r) - Psi(r) = Pi/2 with Psi(0)=0 and Psi'(0)=Pi/2. Hence
- > deduce that Psi(r) = Pi/2*(exp(r) - 1).
-
- Again, let's try differentiating under the integral sign.
- Psi'(r) = integral of x.cos(rx)/x(1+x^2) dx
- = integral of cos(rx)/(1+x^2) dx
- Psi''(r) = integral of x.-sin(rx)/(1+x^2) dx
- = integral of -sin(rx).[x/(1+x^2)] dx.
- So Psi''-Psi = integral of sin(rx)/(1+x^2).[x+1/x] dx
- = integral of sin(rx)/x dx
- and we know what this is by the previous part.
- The boundary conditions can't be hard. When r=0 the integrand for Psi is
- 0 everywhere, and that for Psi' is 1/(1+x^2); we know how to integrate that,
- or ought to.
-
- Hope all this helps... Good luck with whatever further studies you're
- intending.
-