cp's OEIS Frontend

The sequence comes from a problem suggested by Fred Lunnon on the math-fun mailing list on Sep 24 2008. a(n) is also the number of polynomials of degree at most n-1 with all coefficients equal to 1 or -1 which are not invertible modulo x^n - 1. I have a proof that a(n) = 2 if n is an odd prime. - W. Edwin Clark

_Max Alekseyev_'s proof that a(2*p) = 2*binomial(2*p,p) if p is an odd prime:

First notice that A144926(2p) equals the number of such (1,-1)-polynomials f(x) of degree 2p-1 that are divisible by at least one of the cyclotomic polynomials of orders 1, 2, p, or 2p (that are divisors of 2p). These cyclotomic polynomials are: c(1,x) = x - 1, c(2,x) = x + 1, c(p,x) = x^(p-1) + x^(p-2) + ... + x + 1, and c(2p,x) = x^(p-1) - x^(p-2) + ... - x + 1.

Our goal is to prove that (i) f(x) may be divisible only by (x-1) or (x+1) but not both; (ii) if c(p,x) divides f(x) then so does either (x-1) or (x+1) ; (iii) if c(2p,x) divides f(x) then so does either (x-1) or (x+1). Conditions (i), (ii), (iii) all follow from the following Lemma (that in turn directly follows from the definition of f(x)):

Lemma. The values of f(1) and f(-1) are even but different modulo 4.

To prove (i), it is enough to notice that if f(x) were divisible by (x - 1) and (x + 1), then f(1) = f(-1) = 0, a contradiction to the Lemma. To prove (ii), suppose that f(x) = c(p,x) * q(x) for some integer polynomial q. Then f(1) = p * q(1), together with the Lemma and primality of p implying that f(1) = -2p, 0, or 2p. But it is easy to see that if f(1)=0, then (x-1) divides f(x); while if f(1) = -2p or 2p, then (x+1) divides f(x). Condition (iii) is proved similarly to (ii).

From (i), (ii), (iii), it follows that to compute A144926(2p) it is enough to compute the number of such f(x) that are divisible by (x-1) and the number of such f(x) that are divisible by (x+1) and add up these numbers. Clearly, each of these numbers equals binomial(2p,p) that completes the proof. - Vladeta Jovovic, Oct 02 2008