Determine all functions f:Z→Z such that f(f(a)−b)+bf(2a) is a perfect square for all integers a and b.
A number or a short expression. Spacing and $ signs are ignored.
Solution
There are two families of functions which satisfy the condition: (1) f(n)={0 any perfect square if n is even, and if n is odd (2) f(n)=n2, for every integer n. It is straightforward to verify that the two families of functions are indeed solutions. Now, suppose that f is any function which satisfies the condition that f(f(a)−b)+bf(2a) is a perfect square for every pair (a,b) of integers. We denote this condition by (∗). We will show that f must belong to either Family (1) or Family (2). Claim 1. f(0)=0 and f(n) is a perfect square for every integer n. Proof. Plugging (a,b)→(0,f(0)) in (∗) shows that f(0)(f(0)+1)=z2 for some integer z. Thus, (2f(0)+1−2z)(2f(0)+1+2z)=1. Therefore, f(0) is either -1 or 0 . Suppose, for sake of contradiction, that f(0)=−1. For any integer a, plugging (a,b)→(a,f(a)) implies that f(a)f(2a)−1 is a square. Thus, for each a∈Z, there exists x∈Z such that f(a)f(2a)=x2+1 This implies that any prime divisor of f(a) is either 2 or is congruent to 1(mod4), and that 4∤f(a), for every a∈Z. Plugging (a,b)→(0,3) in (∗) shows that f(−4)−3 is a square. Thus, there is y∈Z such that f(−4)=y2+3. Since 4∤f(−4), we note that f(−4) is a positive integer congruent to 3(mod4), but any prime dividing f(−4) is either 2 or is congruent to 1(mod4). This gives a contradiction. Therefore, f(0) must be 0 . For every integer n, plugging (a,b)→(0,−n) in (∗) shows that f(n) is a square. Replacing b with f(a)−b, we find that for all integers a and b, f(b)+(f(a)−b)f(2a) is a square. (**) Now, let S be the set of all integers n such that f(n)=0. We have two cases: - Case 1: S is unbounded from above. We claim that f(2n)=0 for any integer n. Fix some integer n, and let k∈S with k>f(n). Then, plugging (a,b)↦(n,k) in (∗∗) gives us that f(k)+(f(n)−k)f(2n)=(f(n)−k)f(2n) is a square. But f(n)−k<0 and f(2n) is a square by Claim 1. This is possible only if f(2n)=0. In summary, f(n)=0 whenever n is even and Claim 1 shows that f(n) is a square whenever n is odd. - Case 2: S is bounded from above. Let T be the set of all integers n such that f(n)=n2. We show that T is unbounded from above. In fact, we show that 2p+1∈T for all primes p big enough. Fix a prime number p big enough, and let n=2p+1. Plugging (a,b)↦(n,2n) in (∗∗) shows us that f(2n)(f(n)−2n+1) is a square for any integer n. For p big enough, we have 2n∈/S, so f(2n) is a non-zero square. As a result, when p is big enough, f(n) and f(n)−2n+1=f(n)−p are both squares. Writing f(n)=k2 and f(n)−p=m2 for some k,m≥0, we have (k+m)(k−m)=k2−m2=p⟹k+m=p,k−m=1⟹k=n,m=n−1 Thus, f(n)=k2=n2, giving us n=2p+1∈T. Next, for all k∈T and n∈Z, plugging (a,b)↦(n,k) in (∗∗) shows us that k2+(f(n)−k)f(2n) is a square. But that means (2k−f(2n))2−(f(2n)2−4f(n)f(2n))=4(k2+(f(n)−k)f(2n)) is also a square. When k is large enough, we have f(2n)2−4f(n)f(2n)+1<∣2k−f(2n)∣. As a result, we must have f(2n)2=4f(n)f(2n) and thus f(2n)∈{0,4f(n)} for all integers n. Finally, we prove that f(n)=n2 for all integers n. Fix n, and take k∈T big enough such that 2k∈/S. Then, we have f(k)=k2 and f(2k)=4f(k)=4k2. Plugging (a,b)↦(k,n) to (∗∗) shows us that f(n)+(k2−n)4k2=(2k2−n)2+(f(n)−n2) is a square. Since T is unbounded from above, we can take k∈T such that 2k∈/S and also 2k2−n>f(n)−n2. This forces f(n)=n2, giving us the second family of solution.
Want a route through all this instead of an archive? The track
puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.
Source: Omni-MATH,
licensed Apache-2.0.
Statement and solution reproduced as published; topic and difficulty added by this site.