Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site yale.ARPA
Path: utzoo!utcs!lsuc!pesnta!amdcad!decwrl!decvax!yale!andrews
From: andrews@yale.ARPA (Thomas O. Andrews)
Newsgroups: net.math
Subject: Power Series Problem
Message-ID: <758@yale.ARPA>
Date: Wed, 20-Feb-85 12:36:12 EST
Article-I.D.: yale.758
Posted: Wed Feb 20 12:36:12 1985
Date-Received: Thu, 21-Feb-85 02:48:11 EST
Distribution: net
Organization: Yale University CS Dept., New Haven CT
Lines: 51

*** What's this line doing here? ***

Recently, the following problem was posted in net.math:

[Paraphrased - I can't find the original article.]

   Find sequence a1,a2,a3,... such that

	 oo
	----
	\       a     n
  f(x)=  \       n   x
	 /    -----                satisfies f(n)=a .
	/       n!                                 n
	----
	n=0


I believe I've found a nice, 'simple' function f that does the trick.

Let c be a complex root of    c=exp(c).  (Such roots exist.)


Let           n
	a  = c  .
	 n

Then f(x)= {summation above} = exp(cx) .  Hence f(n)=exp(cn).

			    n
But exp(c)=c, so exp(cn)=  c = a
				n.

We can bet a solution of this problem with the a 's real by taking
						n
	 __            ____
b = a  + a      (Where x+yi =x-yi.)
 n   n    n

Then we get
		 _
f(x)=exp(cx)+exp(cx).   In particular, if c=u+vi, then we can write

f(x)=exp(ux) (exp(vxi)+exp(-vxi)) = 2*exp(ux)*cos(vx).


-- 
					      Thomas Andrews

Have you killed a yellow pig today?