Solution:
a. If d is a divisor of n then n/d is also a divisor of n. It follows that for every divisor d of n=k2 with 0<d<k there exists a corresponding divisor n/d with n/d=k2/d>k, and conversely for every divisor d of k2 with d>k there exists a corresponding divisor 0<n/d<k.
The positive divisors of k2 less than k are therefore just as many as those greater than k, as required.
Alternative solution. If n is a perfect square, its prime factorization has the form n=p12α1⋅p22α2⋯pi2αi. Its divisors are numbers of the form p1β1⋯piβi, with 0≤βj≤2αj for every j=1,…,i. In particular k=p1α1⋅p2α2⋯piαi. If we write n=d⋅dn as a product of two of its divisors, we must redistribute the prime factors of n between the two divisors considered. We will thus have that the two factors are of the form
d=p1β1⋯piβi,dn=p12α1−β1⋯pi2αi−βi
which we can write, in terms of k, as
d=k⋅p1β1−α1⋯piβi−αi,dn=k⋅p1α1−β1⋯piαi−βi:
we are multiplying k by a number and by its reciprocal, so one of d and n/d will be greater than k and the other will be less than k. In particular, for every divisor less than k there is one greater than k.
b. By what was shown in part (a), the divisors in total are twice those less than k plus 1 (that is, the divisor k). The positive integers less than k are k−1, so n can have at most 2(k−1)+1=2k−1 divisors.
c. Continuing the reasoning made in part (b), saying that n=k2 has 2k−1 divisors means that all positive integers less than or equal to k divide n. Observe that for k=1 we have n=1, which has a unique positive divisor (itself), so indeed the number of divisors is 2k−1=2⋅1−1. Similarly, for k=2 we have n=4, whose divisors are 1,2 and 4, as required: 2⋅2−1=3. We then observe that an integer n has no divisors >1 in common with n−1. If k≥3 this leads to a contradiction: we observed that k−1>1 must divide k2, and this is absurd because k−1 divides n−1=k2−1=(k+1)(k−1). There are therefore no solutions for k≥3, so the desired k are only k=1 and k=2.