home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!usc!sol.ctr.columbia.edu!samsung!balrog!web.ctron.com!wilson
- From: wilson@web.ctron.com (David Wilson)
- Newsgroups: sci.math
- Subject: Re: Strange Constant
- Message-ID: <4723@balrog.ctron.com>
- Date: 13 Aug 92 14:09:47 GMT
- Sender: usenet@balrog.ctron.com
- Reply-To: wilson@web.ctron.com (David Wilson)
- Organization: Cabletron Systems INc.
- Lines: 129
- Nntp-Posting-Host: web
- Originator: wilson@web
-
-
-
- > I have some questions about a constant, I don't know it's name,
- > but here's how it is obtained.
- >
- > 1 1 1 1 1
- > <constant> = lim --- + --- + --- + ... + --- + --- - ln n
- > n->oo 1 2 3 n-1 n
- >
- > <constant> is approximately equal to the following
- >
- > 0.577216164900715256180774304084479808807373046875...
- >
- > My questions are :
- >
- > 1. What is this constant?
- > 2. Who first obtained it?
- > 3. Why is the way it is?
- > 4. How is it used (If in fact it is used anywhere)?
- >
- > Thanks for the information.
- > Could you please post a respsonse, mail isn't working at this end.
-
-
- 1. As has been posted, this constant is known as Euler's constant,
- is generally denoted by a lower case gamma, and has the value
-
- 0.57721 56649 01532 86060 6512
-
- Euler's constant is hard to approximate by naive methods. I
- don't know if sophisticated methods are known for its accurate
- computation.
-
- 2. Euler.
-
- 3. I assume you are asking why the limit approaches a constant
- value. If f(x) is the greatest integer less than or equal to x,
- we have
-
- f(x) <= x < f(x)+1
-
- If x >= 1, then f(x) >= 1 and
-
- 1/(f(x)+1) < 1/x <= 1/(f(x))
-
- ==> 0 < 1/x - 1/(f(x)+1) <= 1/(f(x)) - 1/(f(x)+1)
-
- ==> 0 < 1/x - 1/(f(x)+1) <= 1/(f(x)(f(x)+1))
-
- Letting n be a positive integer, we integrate the above between
- 1 and n, getting
-
- n-1 n-1
- 0 <= ln n - sum 1/(k+1) <= sum 1/(k(k+1))
- k = 1 k = 1
-
- From here, a few algebraic manipultions bound
-
- n
- sum 1/k - ln n
- k = 1
-
- by a constant on both sides, for sufficient n. Finally, noting
- that the above expression is monotonically decreasing in n, we
- deduce that a limit exists.
-
- The precise value of that limit is harder to explain. It
- involves analysis of the digamma function, generally denoted
- by the lowercase greek letter psi, which is the first
- derivative of log of the gamma function. The important property
- of the digamma function is the recurrence:
-
- digamma(n+1) = digamma(n) + 1/n
-
- From which we deduce that
-
- n
- sum 1/k = digamma(n+1) - digamma(1).
- k = 1
-
- The Euler-Maclaurin sum formula gives us an asymptotic expansion
- for digamma(n) for large n, which is
-
- digamma(n) = ln n + 0(1/n)
-
- That is, digamma(n) approachs ln(n) asymptotically for large n.
- Thus, for large integer n, we have
-
-
- n
- sum 1/k = digamma(n+1) - digamma(1)
- k = 1
- = ln(n+1) + 0(1/n) - digamma(1)
- = (ln(n) + 0(1/n)) + 0(1/n) - digamma(1)
- = ln(n) + 0(1/n) - digamma(1)
-
- And so
-
- n
- sum 1/k - ln(n) = -digamma(1) + 0(1/n)
- k = 1
-
- So that the limit of the left side approaches -digamma(1) as
- n grows without bound. Playing with the definition of digamma,
- reveals that digamma(1) is the derivative of the gamma function
- at x = 1. Therefore, Euler's constant is the negative of the
- derivative of the gamma function at x = 1.
-
- 4. Euler's constant shows up in various definite integrals,
- which pop up from time to time, e.g.:
-
- / inf -x
- gamma = - | e log x dx
- / 0
-
- gamma pops up from time to time in unexpected places. I would
- go so far as to say that gamma is the third most pervasive
- "transcendental" constant after e (and its clan, such as
- log 2), and pi (and its clan, such as sqrt(2/pi)), although
- it is still unknown whether gamma is transcendental (or
- irrational, for that matter, though you know where my money
- falls.)
-
-
- --
- David W. Wilson (wilson@ctron.com)
-
- Disclaimer: "Truth is just truth...You can't have opinions about truth."
- - Peter Schikele, introduction to P.D.Q. Bach's oratorio "The Seasonings."
-