For each positive integer n, we write n in binary representation as n=2a1+2a2+⋯+2ak, where 0≤a1<a2<⋯<ak. Define a set T(n)={2a1,…,2ak}, T(0) is considered empty set.
By Lucas' theorem, Cni is odd if and only if T(i)≤T(n), hence
f(n,q)=A⊆T(n)∑qσ(A)=a∈T(n)∏(1+qa),
where σ(A) denotes the sum of all elements of A.
For m,n and q as given by assumption, we show that if
f(m,q)=a∈T(m)∏(1+qa)∣a∈T(n)∏(1+qa)=f(n,q),
then T(m)⊆T(n), and consequently, f(m,r)∣f(n,r) for every r.
For any integers i,j, 0≤i<j, we have the following factorization:
q2j−1=(q2j−1+1)⋯(q2+1)(q2−1),
therefore
(q2j+1,q2i+1)=(q2i+1,2)∣2.
Let s(k) be the largest odd divisor of a positive integer k, then it follows that s(q2i+1) and s(q2j+1) are coprime. Clearly q>1. If i>0, q2i+1≡1(mod2), and q2j+1>2, thus s(q2i+1)>1. If i=0, since q+1 is not a power of 2, we have s(q+1)>1. For any a∈T(m), s(qa+1)∣∏b∈T(n)s(qb+1). Since s(1+qa)>1, we have a∈T(n), hence T(m)⊆T(n), which completes the proof!