home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!think.com!ames!agate!doc.ic.ac.uk!sot-ecs!dbc
- From: dbc@ecs.soton.ac.uk (Bryan Carpenter)
- Newsgroups: sci.math
- Subject: Re: DERIVATIVE OF X^X
- Message-ID: <13421@ecs.soton.ac.uk>
- Date: 6 Nov 92 18:55:25 GMT
- References: <1992Nov4.200746.11729@aio.jsc.nasa.gov> <1992Nov5.004445.21327@infodev.cam.ac.uk>
- Sender: news@ecs.soton.ac.uk
- Lines: 57
- Nntp-Posting-Host: louis
-
- In <1992Nov5.004445.21327@infodev.cam.ac.uk> gjm11@cus.cam.ac.uk (G.J. McCaughan) writes:
-
- >Do we really have to do this from first principles, as your post suggests?
- >If not:
-
- >f(x) = x^x = exp(x log x)
- >so by the chain rule
- >f'(x) = d/dx(x log x).exp(x log x)
- > = (log x + 1) exp(x log x)
- > = x^x + x^x.log x
-
- >But I suppose you want it done in gruesome detail, by hand, from first
- >principles. Well, here goes. In what follows, "O(h^2)" denotes stuff
- >which is bounded above by some constant times h^2, as h->0.
-
- >f(x+h) = (x+h)^(x+h)
- > = x^(x+h)(1+h/x)^(x+h)
- > = x^(x+h)(1+(x+h).(h/x)+O(h^2)) by the binomial theorem, if |h|<|x|
- > = x^(x+h)(1+h+O(h^2)) = x^(x+h)(1+h) + O(h^2)
- > = x^x.x^h.(1+h) + O(h^2).
-
- >So, f'(x) = limit as h->0 of [f(x+h)-f(x)]/h
- > = limit of [x^x.x^h.(1+h) - x^x]/h
- > = limit of x^x((x^h-1)/h + x^h)
- > = x^x.(limit of (x^h-1)/h + limit of x^h)
- > = x^x.(limit of (x^h-1)/h + 1)
-
- >so we are reduced to finding the limit, as h->0, of (x^h-1)/h.
- >From the chain-rule stuff above we know it's got to be log x;
- >and if it weren't for the requirement to do everything from first
- >principles this would also be an easy exercise in the chain rule.
- >I'm not at all sure, though, how we can do it from first principles.
-
- I think ``first principles'' ought to allow you to take
-
- Lim (1 + y/n)^n
- n -> inf
-
- as the definition of e^y. So, if y = log x and n is infinitely large
- (if you know what I mean),
-
- x = (1 + y/n)^n
-
- so
-
- y = n (x^(1/n) - 1)
-
- and you can put h = 1/n (so to speak). It gets the right answer, anyway.
-
- >...
-
- >--
- >Gareth McCaughan Dept. of Pure Mathematics & Mathematical Statistics,
- >gjm11@cus.cam.ac.uk Cambridge University, England. [Research student]
-
- Bryan
-
-