Suppose a prime p satisfies pa∣m. Then from the given conditions, there exists an infinite subset A1 of A such that p is coprime to every element in A1.
By the pigeonhole principle, there is an infinite subset A2 of A1, such that x≡a(modmn), for each element x∈A2, where a is a positive integer and p∤a.
Since (m,n)=1, we have (pa,pamn)=1. By the Chinese Remainder Theorem, we know that
{x≡a−1(modpa),x≡0(modpamn)①
have infinitely many solutions. Among them we take one as x. Next define Bp as the set of the first x elements in A2, and Sp as the sum of all elements in Bp. Then we have Sp≡ax(modmn). By ① we have
Sp≡ax≡1(modpa),Sp≡0(modpamn).
Suppose that m=p1a1⋯pkak, and for every pi (1≤i≤k) select a finite subset Bi of A, where Bi⊂A∖(B1∪⋯∪Bi−1), such that Spi, the sum of all the elements in Bi, satisfies
Spi≡1(modpiai),Spi≡0(modpiaimn).2◯
Let B=⋃i=1kBi, whose sum of elements then satisfies S=∑i=1kSi. According to ②, we have S≡1(modpiai) (1≤i≤k), and S≡0(modn). So B is the required subset, and the proof is complete.
Solution 2:
Divide every element in A by mn, and let the remainders which occur infinitely be (in order) α1,α2,…,αk. We claim that
(α1,α2,…,αk,m)=1.3◯
Otherwise, assume that p∣(α1,α2,…,αk,m). Then p∤n, since (m,n)=1. By the given conditions we know that there exist infinitely many elements of A not divisible by p. But by the definition of α1,α2,…,αk, the number of such elements are finite. Then by contradiction, ③ holds. Consequently, there are x1,x2,…,xk,y, satisfying α1x1+α2x2+⋯+αkxk−ym=1. Choose a suitable positive integer r such that rn≡1(modm). Then
α1(rnx1)+α2(rnx2)+⋯+αk(rnxk)=rn+rmny.
We select in order rxi elements from A such that the remainders by mn are αi (i=1,2,…,k). The set of all these elements is acquired.