Let k>0 satisfy f(k)=1. By step 1, f(k2)=f(f(k))=f(1), so by the lemma (step 3) with multiplier 1/k, f(k)=f(1/k); hence f(1/k)=1 too, so we may assume k≥1. Using the full form of (∗) with x=k2−1,y=k: f(k2−1+f(k))=f(k2−1)+f(k2)+2f(kk2−1), i.e. f(k2)=f(k2−1)+f(k2)+2f(kk2−1) (using f(k)=1), so f(k2−1)=−2f(kk2−1)≤0 by step 5's non-negativity. Since also f(k2−1)≥0, we get f(k2−1)=0, so by step 4, k2−1=0, i.e. k=1.