[seqfan] A107008 = primes of the form 24k+1?

[Peter Munn]

Hi seqfans.

In A107008 (Primes of the form x^2+24*y^2), NJAS comments "Presumably this
is the same as primes congruent to 1 mod 24."

Can we come up with something to decide this?

It would help to establish when 24k+1 can be written as x^2+24*y^2. I
reckon this happens when the k in 24k+1 can be written as the sum of a
square, j, and a generalized pentagonal number, i, because setting y =
sqrt(j), x = sqrt(24i+1) can be shown to satisfy the equation. (See Zak
Seidov's 2008 comment in http://oeis.org/A001318, "Generalized pentagonal
numbers".) I believe the converse is true, also.

OEIS does not yet have the sequence "Numbers that can't be written as the
sum of a square and a generalized pentagonal number" [1], but the
sequence, S, starts 20, 29, 33, 34, 45, 46, 53, ... , and I reckon it has
positive asymptotic density.

So the question becomes: if k is a term of S, why should 24k+1 be
composite, at least up to the limit of Vladimir Orlovsky's check in
A107008? From the first 3 terms we get: 20*24 + 1 = 481 = 13*37; 29*24 + 1
= 697 = 17*41; 33*24 + 1 = 793 = 13*61.

Going through more terms, I saw a pattern emerge, prompting me to ask: is
this particular subset of "24k+1" numbers the same as "nonsquare numbers
of the form (24i + m) * (24j + m), 0 <= i < j, m in {13, 17, 19, 23}"?
This would be interesting anyway, and could be a clue.

However, I'm not sure I'm close to an answer, and there might be a much
easier route: does anyone have better ideas? Or know the answer already?

Best regards,

Peter

[1] I also tried looking for the number of ways positive integers _can_ be
so written, using "Search: seq:2,2,1,1,2,2,1,1,2,1,2,1,1,1,1,4" but it
draws a blank.