home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!spool.mu.edu!yale.edu!jvnc.net!princeton!mace.Princeton.EDU!tao
- From: tao@mace.Princeton.EDU (Terry Tao)
- Subject: Help with Tchebyshev polynomials, please...
- Message-ID: <1993Jan4.150052.3153@Princeton.EDU>
- Originator: news@nimaster
- Sender: news@Princeton.EDU (USENET News System)
- Nntp-Posting-Host: mace.princeton.edu
- Organization: Princeton University
- Date: Mon, 4 Jan 1993 15:00:52 GMT
- Lines: 27
-
-
- I've been toying with the problem of trying to maximize certain semi-norms
- on the space of polynomials of degree n or less which are bounded in
- modulus by 1 on the interval [-1, 1], i.e. the set of all f of degree \leq
- n such that |f(x)| \leq 1 for all x in [-1, 1]. The semi-norms I have in
- mind are things like the following: the L^p norm of f on the interval [1,
- 1], the derivative of f at a point x_0, the value of f at a point x_0
- (where x_0 need not be inside [-1, 1]), various coefficients of f, and so
- on. (actually some of these are not quite seminorms, you have to take the
- modulus of them first, but this is not important). In most of these cases
- I found that a maximum exists, because the space in question is convex,
- closed, and "bounded" in a sense, and the extremal value of f must attain
- the values -1 and 1 at least n times in the interval [-1, 1] and no more
- than n+1 times. If f attans the values -1 and 1 n+1 times, then I can show
- f is essentially a Tchebyshev (sic?) polynomial, f(x) = cos(n arccos(x)),
- which is nice. But how does one characterize the polynomials f which only
- attain the values -1 and 1 n times on the interval [-1, 1]? Some of these
- polynomials are merely dilations of the Tchebyshev, but others are slightly
- different. I have some characterizations of them, but they involve cosines
- of elliptic functions and this is not a friendly characterization to work
- with. I also have a recursion formula for the coefficients but they are
- also very messy to work with. Anyone know of a nice characterization of
- these polynomials, or at least be able to give me a reference?
-
- Email is preferred. Thank you.
-
- Terry
-