Let n=p1a1p2a2⋯prar,ai≥0 be the prime factorization of the number n. Then the number of divisors of n is (a1+1)(a2+1)⋯(ar+1).
Let k be a positive integer and let f(2k)=p1a1p2a2⋯prar. Then ai=2bi−1 for some integers bi≥0, i=1,2,…,r and ∑i=1rbi=k.
Let pi,pj be any two distinct prime factors of f(2k). If we replace pi2bi−1 with pi2bi+1−1 and pj2bj−1 with pj2bj−1−1 in the prime factorization of f(2k) we get a number that also has 2k divisors (as f(2k)) because 2bi+1⋅2bj−1=2bi⋅2bj. Minimality of f(2k) implies
pi2bi−1⋅pj2bj−1<pi2bi+1−1⋅pj2bj−1−1,
or equivalently
pj2bj−1<pi2bi.(∙)
If we allow some of the integers ci or bi to be zero, we can write f(2k+1)=p12c1−1p22c2−1⋯pr2cr−1.
Since ∑i=1rci=k+1>k=∑i=1rbi, there is a positive integer s such that cs>bs.
Let t be a positive integer different than s. Using (∙) for the pair (ps,pt), and for the pair (pt,ps) gives
pt2ct>ps2cs−1≥ps2bs>pt2bt−1.
We conclude ct>bt−1, i.e. ct≥bt for all t=1,…,r, which implies that f(2k+1) divides f(2k).