Identities inspired by Ramanujan Notebooks (part 2)
by Simon Plouffe
April 2006
These are new findings
related to the ones found in 1998. I had the idea that maybe the sums in the
first note I wrote were looking in the bad direction. They are good yes but
maybe there is better. Actually there is, instead of considering powers
of exp(2π) I should consider powers of exp(kπ).
After a few
experiments with Maple (computer algebra system) and PSLQ (integer relations
algorithm) here are the findings and remarks.
- I could go up to
Zeta(39) and powers of exp(30π) after that the numbers are too close together
for any computation, 1/exp(30π) is of the order of 10-40 and
even with 2000 decimal digits I get an error.
- The powers of
exp(kπ)±1 are related to the divisors of a number.
- The numbers that
appear to be expressible are so far : log(A) where A is an integer or simple
alebraic like arccosh(1/2) or arctanh(2/3), ζ(2n+1), π2n+1, ζ(2n+1)
and some formulas involving lnΓ(1/4) and ln(π) but not separately.
Formulas
for odd powers of π and ζ(2n+1)
That
one is the fastest converging series so far with 5.45 digits for each n.
Other
formulas
The last
formula for Zeta(5) is interesting since the lowest term gives a rate of
convergence of the order of 1/exp(3π) which is 4.09 decimal digits.
References
[1] Berndt, Bruce,
Ramanujan Notebooks (volumes I to V), Springer Verlag.
[2]
Plouffe, Simon, Identities
inspired from Ramanujan Notebooks II, 1998, unpublished notes.
[3]
Finch, Steven R., Mathematical Constants, Encyclopedia of Mathematics and its
Applications, 94 (2003).
[4]
Flajolet, Philippe. Recent articles from INRIA.
I take the occasion to
thank Bruce Berndt for the remarkable piece of work he did with these collected
works of S. Ramanujan.