home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.num-analysis
- Path: sparky!uunet!cs.utexas.edu!hermes.chpc.utexas.edu!jonathan
- From: jonathan@chpc.utexas.edu (Jonathan Thornburg)
- Subject: Re: Solutions to quartic and cubic eqns.
- Message-ID: <1992Aug27.175551.23486@chpc.utexas.edu>
- Summary: numerical zero-finding is probably more accurate, try ZEROIN
- Keywords: quartic cubic quadratic equation zero root ill-condition ZEROIN
- Sender: jonathan@einstein.ph.utexas.edu
- Organization: U of Texas at Austin / Physics Dept / Center for Relativity
- References: <1992Aug25.161234.13275@magnus.acs.ohio-state.edu>
- Date: Thu, 27 Aug 92 17:55:51 GMT
- Lines: 44
-
- In article <1992Aug25.161234.13275@magnus.acs.ohio-state.edu>
- kiyer@magnus.acs.ohio-state.edu (Karthik A Iyer) writes:
- > I am looking for a routine(s) or a place where I can find
- >them (FTP sites, etc) to solve quartic and cubic eqations. I realize that
- >analytic solutions exist, but my problem is that the coefficients of the
- >powers vary by as much as 10-15 orders of magnitude. So, a small error is
- >solving the for the cubic's root leads to large errors when that root is
- >used to solve for the four roots of the quartic. Any help or leads would be
- >appreciated. I would prefer to get my hands on a FORTRAN double precision
- >routine. Thanks in advance,
-
- Unfortunately, the "classical" closed-form analytical solutions
- tend to be quite ill-conditioned, *especially* when the coefficients
- are widely varying in magnitude.
-
- I suspect you would get much more accurate results if you simply
- used a good numerical zero-finding code to numerically locate the
- (a) zero of your quartic/cubic. (Of course, you still have to be
- able to *evaluate* the quartic/cubic, which is itself rather tricky
- when the coefficients are so disparate in magnitude. But that's
- still a much easier problem than trying to find the roots directly.)
-
- An excellent zero-finding code is ZEROIN, available from netlib
- via
-
- echo 'send zeroin from fmm' | Mail netlib@ornl.gov
-
- This will get you a Fortran subroutine with all the documentation
- you need to use it. For more details, see either the (excellent,
- superb, ++good, highly recommended) book
-
- G. E. Forsythe, M. A. Malcolm, and C. B. Moler
- "Computer Methods for Mathematical Computations"
- Prentice-Hall, Englewood Cliffs, 1977
-
- or section 9.3 of Numerical Recipes. Be warned, though, that the
- code for ZEROIN given in Numerical Recipes (where it's called ZBRENT)
- has been "improved", and is of somewhat lower quality than the
- original one you'll get from netlib.
-
- - Jonathan Thornburg
- <jonathan@einstein.ph.utexas.edu> or <jonathan@hermes.chpc.utexas.edu>
- University of Texas at Austin / Physics Dept / Center for Relativity
- and (for a few more months) U of British Columbia / {Astronomy,Physics}
-