OFFSET
4,1
COMMENTS
Mod 2, odd primes p are 1 and mod 4 or mod 6, p=-+1, so that p^2==p^4==1 (mod 2*4*6). Moreover, mod 5, p==-+1, -+2 for p>5, implying p^2==-+1 or p^4==1, so that finally p^4==1 (mod 2*4*6*5), i.e., 240 divides (p^4 - 1) for p>5.
From Jean-Claude Babois, Jan 13 2012: (Start)
From Simon Plouffe's web site we know that sum_{n >= 1} n^3/(exp(2*n*Pi / 7) - 1) = 10.0000000000000001901617..., very close to a(1). Extensive calculations suggest that more generally, for any prime p >= 7, Sum_{n >=1} n^3/(exp(2*n*Pi / p) - 1) is similarly very close to (p^4-1)/240.
Victor Miller replied on Jan 29 2012 via email, with an explanation of this observation. The following is an abridged version of his reply:
Let q = exp(2*Pi*i*z). Define the Eisenstein series E_4(z) = 1 + 240*sum_{n >= 1} n^3*q^n/(1-q^n). For your observation we take z = i/p, so that q = exp(-2*Pi / p). So what you've evaluated numerically is (E_4(i/p) - 1)/240.
The Eisenstein series obeys the transformation law E_4(-1/z) = z^4*E_4(z), or E_4(i/p) = p^4*E_4(i*p). Your observation reduces to showing that E_4(i*p) is very close to 1. In this case q = exp(-2*Pi*p), so E_4(i*p) - 1 is bounded by a geometric series in q. In your first case, when p = 7, q is around exp(-44), which is already quite small. (End)
LINKS
Charles R Greathouse IV, Table of n, a(n) for n = 4..10000
MATHEMATICA
Select[(Prime[Range[50]]^4-1)/240, IntegerQ] (* Harvey P. Dale, Nov 28 2015 *)
PROG
(PARI) a(n)=(prime(n)^4 - 1)/240 \\ Charles R Greathouse IV, May 31 2013
CROSSREFS
KEYWORD
nonn
AUTHOR
Lekraj Beedassy, Nov 12 2003
EXTENSIONS
More terms from Ray Chandler, Nov 12 2003
STATUS
approved