home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!stanford.edu!snorkelwacker.mit.edu!bloom-picayune.mit.edu!athena.mit.edu!frisch1
- From: frisch1@athena.mit.edu (Jonathan Katz)
- Subject: ODE problem...
- Message-ID: <1992Nov20.184102.14068@athena.mit.edu>
- Sender: news@athena.mit.edu (News system)
- Nntp-Posting-Host: m4-035-2.mit.edu
- Organization: Massachusetts Institute of Technology
- Date: Fri, 20 Nov 1992 18:41:02 GMT
- Lines: 30
-
- >The following ODE problem came up recently.
- >I know how to solve it by the power series method, but was wondering
- >if anyone could figure out an easier way of solving it (maybe a nice
- >substitution?).
- >(x and y are functions of t, a is a constant)
- >x'=(a)(x)cost+(a)(y)sint
- >y'=(a)(x)sint-(a)(y)cost.
-
- Hi, I was the one to originally pose this problem, and I'm glad to see the
- interest it has sparked.
- Thanks to all those who sent email.
-
- However, while playing with this system, I came upon a neat 'trick' which no one
- has yet pointed out.
-
- multiply the first equation by y and the second by x. Then add the two equations
- together, getting:
- yx'+xy'= xy(acos(t)+asin(t))
- letting f=xy, note that this simplifies to:
- f'=f(acos(t)+asin(t)) which is easily solvable.
-
- Now subtract the two equations, and divide the resulting equation by y^2. Then
- let g=x/y, and the equation becomes:
- g'=some function of g (sorry I don't have my work in front of me, but it's
- simple enough to carry out)
-
- Now we have xy=some function of t and
- x/y=some other function of t.
-
- x and y can then be solved for easily.
-