home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!cs.utexas.edu!rutgers!cmcl2!sbcs.sunysb.edu!hanche
- From: hanche@ams.sunysb.edu (Harald Hanche-Olsen)
- Newsgroups: sci.math
- Subject: Re: 1+1/2+1/3+1/4+...+1/n
- Message-ID: <HANCHE.92Nov23122110@ptolemy.ams.sunysb.edu>
- Date: 23 Nov 92 17:21:10 GMT
- References: <92324.132329K3032E2@ALIJKU11.BITNET>
- <HANCHE.92Nov19123151@ptolemy.ams.sunysb.edu>
- <BxzMuB.6Gr@cantua.canterbury.ac.nz>
- <1992Nov21.125910.24073@linus.mitre.org>
- Sender: usenet@sbcs.sunysb.edu (Usenet poster)
- Organization: University at Stony Brook, NY
- Lines: 44
- In-Reply-To: bs@gauss.mitre.org's message of Sat, 21 Nov 1992 12: 59:10 GMT
- Nntp-Posting-Host: ptolemy.ams.sunysb.edu
-
- >>>>> On Sat, 21 Nov 1992 12:59:10 GMT, bs@gauss.mitre.org (Robert D.
- >>>>> Silverman) said:
-
- bs> In article <BxzMuB.6Gr@cantua.canterbury.ac.nz>
- bs> jsv@math.canterbury.ac.nz (Julian Visch) writes:
-
- bs> :> evalf(sum(1./n,n=1..15092688622113788323693563264538101449859496));
- bs> :99.999999999999999999999999999999999999999999942747074071711130782[...]
-
- bs> Just HOW did Maple produce this sum?
-
- With the help of the Psi function, the heart of which does the
- following (this is computing Psi(x) -- I have simplified a bit and cut
- several branches that are not taken for the values of x and Digits
- considered here):
-
- z := evalf(x);
- t := Digits-trunc(z);
- s := ln(z);
- z := 1/z;
- t := -1/2*z;
- z := z^2;
- y := 1;
- for k from 2 by 2 while s+t <> s do
- s := s+t; y := y*z; t := -bernoulli(k)*y/k
- od;
- RETURN evalf(s)
-
- bs> How do you know it is right?
-
- You don't, of course, unless you repeat the computation independently.
-
- bs> I'm not saying it is wrong, but I am suspicious.
-
- As you should always be :-)
-
- bs> Using H_n = log(n) + gamma + O(1/n) [even getting many error terms
- bs> via Bernoulli polynomials won't give explicit constants on the
- bs> errors], is NOT sufficient.
-
- Ummm. Can't suitable constants be determined, you say? Then this
- calculation seems to be in trouble all right...
-
- - Harald
-