Let P(x) and Q(x) be non-constant real polynomials such that for every positive integer m, there exists a positive integer n with P(m)=Q(n). (1) Prove that if deg(Q) divides deg(P), then there exists a real polynomial h(x) such that P(x)=Q(h(x)). (2) Prove that deg(Q) divides deg(P).
Solution
Proof 1: We denote by Ok(x) any polynomial of degree at most k.
(1) If v=degQ divides u=degP, let u=vL. Write P(x)=auxu+au−1xu−1+⋯+a0, Q(x)=bvxv+bv−1xv−1+⋯+b0.
For k=0, take c0=Lau/bv, so Q(c0xL)=auxu+⋯=P(x)+Ou−1(x). Assume gk−1(x)=c0xL+⋯+ck−1xL+1−k satisfies Q(gk−1(x))=P(x)+R(x), where R(x)=sxu−k+… has degree ≤u−k. Consider gk(x)=gk−1+ckxL−k: Q(gk(x))−Q(gk−1(x))=bv[(gk−1(x)+ckxL−k)v−(gk−1(x))v]+O(v−1)L−k(x)=bv[vgk−1(x)v−1ckxL−k+O(v−2)L+2(L−k)(x)]+O(v−1)L−k(x)=ckbvvc0v−1xvL−k+Ou−k−1(x). Choose ck such that ckbvvc0v−1=s (the coefficient of xvL−k in R(x)). Then gk(x)=gk−1+ckxL−k satisfies Q(gk(x))−P(x)=Ou−k−1(x).
The problem's condition guarantees that for every integer m≥M0, there exists n=nm such that Q(nm)=P(m), thus ∣Q(nm)−Q(h(m))∣=∣P(m)−Q(h(m))∣=∣R(m)∣≤C⋅mu−L−1. This implies that when M0 is sufficiently large, there exists a positive real number D such that for all integers m≥M0: ∣nm−h(m)∣<mD.(11)
Let nm=h(m)+rm. Since h(m) is a polynomial of degree L, we have k=0∑L+1(−1)k(kL+1)h(m+k)=0. However, since each nm is an integer, this implies Tm=k=0∑L+1(−1)k(kL+1)rm+k=k=0∑L+1(−1)k(kL+1)nm+k∈Z. From (11), there exists M1 such that when m≥M1, we have ∣rm∣<2L+11, and consequently Tm=0. This means rm is a recurrence sequence, and solving this recurrence shows that for m≥M1, rm is a polynomial in m with degree at most L. Moreover, from (11) we know {rm} is bounded, hence it must be identically zero. Therefore, P(m)−Q(h(m))=0 holds for all m≥M1, which implies P(x)≡Q(h(x)). Furthermore, since h(m)=nm always takes integer values, h(⋅) must be a polynomial with rational coefficients.
(2) Prove that deg(Q) divides deg(P).
Proof 2 for (2): Following similar analysis to Proof 1, we select f1(x)=xv and f2(x)=2xv, which yields rational-coefficient polynomials h1(x)=c1xu+… and h2(x)=c2xu+… satisfying P(xv)≡Q(h1(x)),P(2xv)≡Q(h2(x)). Comparing leading coefficients gives: au=bv⋅c1vandau⋅2u=bv⋅c2v, thus 2u=(c1c2)v, meaning 2u/v=c1c2 must be rational. Since fractional powers of 2 are irrational, vu must be an integer, i.e., deg(Q) divides deg(P).
Proof 3 for (2): Choosing f(x)=xv, there exists a rational-coefficient polynomial h(x)=cuxu+cu−1xu−1+⋯+c0 satisfying P(xv)≡Q(h(x)), where for large positive integers m=xv, nm=h(x) is also a positive integer. Grouping terms by powers modulo v, we express h(x) as: h(x)≡xv−1hv−1(xv)+xv−2hv−2(xv)+⋯+xh1(xv)+h0(xv), where each hk(⋅) is a rational-coefficient polynomial. For a large prime p, substituting x=p1/v yields h(p1/v)=D∈Z. By Eisenstein's criterion, the polynomial xv−p is irreducible and shares the root x=p1/v with h(x)−D, hence xv−p∣h(x)−D. For k=0,1,…,v−1, we have xv−p∣hk(xv)−hk(p), therefore: xv−p∣xv−1hv−1(p)+xv−2hv−2(p)+⋯+xh1(p)+h0(p)−D. The right-hand side has degree ≤v−1, so divisibility implies it's the zero polynomial. Thus for k=1,2,…,v−1, hk(p)=0 holds for all large primes p, making each hk(x) identically zero. Consequently, h(x)≡h0(xv), meaning the degree u must be a multiple of v, i.e., deg(Q) divides deg(P).
Proof 4 for (2): Again choosing f(x)=xv, there exists a rational-coefficient polynomial h(x)=cuxu+cu−1xu−1+⋯+c0 (cu=0) satisfying P(xv)≡Q(h(x)). We claim that all terms in h(x) must have exponents congruent to u modulo v or be constant terms. If not, let r be the largest exponent with cr=0 and r≡u(modv) (r>0). Since Q(x)=bvxv+bv−1xv−1+⋯, we have: Q(h(x))=bv(h(x))v+O(u(v−1)), where O(u(v−1)) denotes some polynomial of degree ≤u(v−1). In the expansion of bv(h(x))v, there exists a term: bv⋅(1v)(cuxu)v−1⋅crxr=bvvcuv−1cr⋅xuv−u+r, while other terms either have exponents divisible by v or exponents less than uv−u+r, leaving no cancellation possible. Thus P(xv)=Q(h(x)) would contain a non-zero xuv−u+r term whose exponent isn't divisible by v - a contradiction.
Let t be the smallest non-negative remainder of u modulo v. Then h(x) can be expressed as xtg(xv)+c0, where g(⋅) has rational coefficients. From P(xv)≡Q(h(x)), when y=xv is a large positive integer, h(x)=h(y1/v)=yt/vg(y)+c0 must also be integer-valued. This implies yt/v must be rational for all large positive integers y, forcing t=0. Therefore deg(Q)=v divides deg(P)=u. □
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: MathNet,
licensed CC-BY-4.0.
Statement and solution reproduced as published; topic and difficulty added by this site.