First suppose that r(A,B,n) is increasing for n>n0. For the sake of simplicity, let Aˉ=N∖A and Bˉ=N∖B. Then
n+1=r(N,N,n)=r(A,B,n)+r(Aˉ,B,n)+r(A,Bˉ,n)+r(Aˉ,Bˉ,n)⟺r(Aˉ,B,n)+r(A,Bˉ,n)+r(Aˉ,Bˉ,n)=n+1−r(A,B,n)
Since r(A,B,n) is increasing, there exists a constant c such that n+1−r(A,B,n)≤c for all n. In particular, for all n, r(Aˉ,N,n)=r(Aˉ,B,n)+r(Aˉ,Bˉ,n)<c, implying ∣Aˉ∣<c. In fact, if Aˉ={a1,a2,a3,…,al,…}, a1<a2<a3<…, for ak<n<ak+1, we have r(Aˉ,N,n)=k. Therefore if Aˉ were infinite, r(Aˉ,N,n) would be unbounded. Analogously Bˉ must be finite as well.
Conversely, suppose Aˉ and Bˉ are both finite. For all n>max(Aˉ)+max(Bˉ), if a∈Aˉ then n−a>maxBˉ, so n−a∈/Bˉ, hence n−a∈B. In a similar fashion, if b∈Bˉ then n−b∈/Aˉ, hence n−b∈A. Thus
r(A,B,n)=n+1−∣Aˉ∣−∣Bˉ∣<n+2−∣Aˉ∣−∣Bˉ∣=r(A,B,n+1)
So r(A,B,n) is increasing for n>max(Aˉ)+max(Bˉ).