home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!news.claremont.edu!ucivax!orion.oac.uci.edu!beckman.com!dn66!a_rubin
- Newsgroups: sci.math
- Subject: Re: int(x*log(x)*((1-x)^n), x=0..1)
- Message-ID: <a_rubin.714329152@dn66>
- From: a_rubin@dsg4.dse.beckman.com (Arthur Rubin)
- Date: 20 Aug 92 16:45:52 GMT
- References: <dld.714296391@bruce.cs.monash.edu.au>
- Keywords: Hairy integral, Beta function
- Nntp-Posting-Host: dn66.dse.beckman.com
- Lines: 67
-
- In <dld.714296391@bruce.cs.monash.edu.au> dld@cs.monash.edu.au (David L Dowe) writes:
-
- >Let I(n) = int(x*log(x)*((1-x)^n), x=0..1);
- >to use Maple's notation. n is a positive integer. log is ln .
-
- >I seek as simple an expression as possible or, that failing, as rapidly
- >converging an expression as possible for I(n).
-
- >Before you respond (to me rather than the newsgroup, please), let me tell
- >you what I do know:
- >By parts, int((x^n)*log(x), x=0..1) = -1/((n+1)^2) (1)
-
- >Expanding (1-x)^n binomially and using (1) enables us to get a
- >fairly simple expression for I(n). But, I would like (if possible) a simpler
- >or more rapidly converging expression.
-
- >o I also know that, re-writing I(n) as int((1-x)*log(1-x)*(x^n)),x=0..1);
- > Taylor expanding log(x) = Sum(i=1,infinity) -(x^i)/i gives
- > I(n) = int(-(x^(n+1)) + Sum(i=2,infinity) (x^(n+i))/((i-1)*i) , x=0..1) ;
- >and so I(n) = -1/(n+2) + Sum(i=2,infinity) 1/((i-1)*i*(n+i+1)) .
- > This also converges more slowly than I would like it to.
-
- >o I also "know" that I(n) is rational. I(0) = -1/4, I(1) = -5/36,
- > I(2) = -13/144, I(3) = -77/1200, I(4) = -29/600, I(5) = -223/5880,
- > I(6) = -481/15680, and Maple Vn4.3 gave me a mammoth expression for
- > I(50) in 15 seconds.
- > When I gave Maple 4.3 the integral I(n) (n unspecified), it merely
- > returned some seconds later with the same initial expression I had given it.
-
- >o I also "know" that n*(n-1)*I(n) + log(n + 1/2) converges to
- > (I forget whether it's + or - ) +/- 0.42... ,
- > the 0.42... possibly (a guess) being something to do with Euler's constant.
-
- > I would appreciate (a reference to)
- > a proof that I(n) is rational,
- > a proof of the convergence result, and the value of "0.42..." , and
- > (most importantly) as simple an expression as possible for I(n) .
- >__
-
-
- Mathematica 2.0 for HP/Apollo Domain OS reports:
- Integrate[(1-x) Log[1-x] x^n,{x,0,1}] = (n^3 -4 n -2)/(n (n+1)^2 (n+2)^2) -
- (EulerGamma + PolyGamma[n])/((n+1)(n+2))
-
- (PolyGamma is the digamma function "psi", and EulerGamma is Euler's constant.)
-
- (1) You "know" I(n) is rational by using your expansion (1).
-
- (2) EulerGamma + PolyGammma[n] is Sum[1/m,{m,n-1}], which produces a fairly
- simple (but not rapidly convergent) rational expression.
-
- (3) PolyGamma[n] = log(n) + 1/2/n - 1/12/n^2 + 1/120/n^4 - 1/252/n^6 +
- 1/240/n^8 -1/132/n^10 + ..., at least as an asymtotic series, (which can be
- used for rapid calculation given the value of gamma) so your constant limit
- of n(n-1)I(n) + Log(n+1/2) or (n+1)(n+2)I(n) + log(n) is 1 - EulerGamma.
-
-
-
-
-
-
-
- --
- Arthur L. Rubin: a_rubin@dsg4.dse.beckman.com (work) Beckman Instruments/Brea
- 216-5888@mcimail.com 70707.453@compuserve.com arthur@pnet01.cts.com (personal)
- My opinions are my own, and do not represent those of my employer.
- My interaction with our news system is unstable; if you want to be sure I see a post, mail it.
-