home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!elroy.jpl.nasa.gov!ames!purdue!mentor.cc.purdue.edu!pop.stat.purdue.edu!hrubin
- From: hrubin@pop.stat.purdue.edu (Herman Rubin)
- Newsgroups: sci.math
- Subject: Re: Power series solution to ordinary differential equation
- Message-ID: <57125@mentor.cc.purdue.edu>
- Date: 18 Aug 92 13:34:03 GMT
- References: <1992Aug16.231044.4401@pellns.alleg.edu> <57030@mentor.cc.purdue.edu> <25544@dog.ee.lbl.gov>
- Sender: news@mentor.cc.purdue.edu
- Distribution: na
- Organization: Purdue University Statistics Department
- Lines: 36
-
- In article <25544@dog.ee.lbl.gov> sichase@csa2.lbl.gov writes:
- >In article <57030@mentor.cc.purdue.edu>, hrubin@pop.stat.purdue.edu (Herman Rubin) writes...
-
- >>The method of power series is actually the oldest method of numerical
- >>solution of differential equations, but is now usually considered "not
- >>in style." There ARE many cases in which it is far better than the
- >>usual ones, and of course many where it is not. HOW is not much of
- >>a problem; as to WHEN, this is a matter of art, not science.
-
- >Interesting. Why is is "not in style"? As a physicist, I use only
- >two methods of solving DE's. I guess the solution that I already know or
- >I use the Frobenius method. With rare exception, I have not needed more.
- >(Once or twice I have had to solve a DE with a contour integral, but that's
- >unusual.) In fact, the only time I have ever needed to know any of the
- >host of standard techniques was when taking a Diff Eq. class as an
- >undergrad.
-
- What do you mean by "solve"? Numerical methods of solving a differential
- equation can be quite a problem. But if one looks at the current numerical
- analysis books, the emphasis is almost entirely on methods which do not
- involve computing derivatives anywhere. These methods propagate rounding
- errors, etc., and getting lots of accuracy is VERY difficult.
-
- On the other hand, power series methods run into problems with radius of
- convergence. But a good nasty example is the Thomas-Fermi equation, where
- the boundary conditions make real problems. Now one can carry out numerical
- approximations, and eventually get something reasonable, but a method, for
- which I believe there is a paper by Hille, points out that by using two
- power series, one in x^.5 about 0, and one which, apart from a simple factor,
- is in x^a, where a is a quadratic irrational, about infinity, and matching
- them, these major problems can be avoided.
- --
- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
- Phone: (317)494-6054
- hrubin@pop.stat.purdue.edu (Internet, bitnet)
- {purdue,pur-ee}!pop.stat!hrubin(UUCP)
-