MathLabs

Problem 3

For each integer k≥2k\ge2, determine all infinite sequences of positive integers a1,a2,…a_1,a_2,\ldots for which there exists a polynomial PP of the form P(x)=xk+ck−1xk−1+⋯+c1x+c0P(x)=x^k+c_{k-1}x^{k-1}+\cdots+c_1x+c_0, where c0,c1,…,ck−1c_0,c_1,\ldots,c_{k-1} are non-negative integers, such that P(an)=an+1an+2⋯an+kP(a_n)=a_{n+1}a_{n+2}\cdots a_{n+k} for every integer n≥1n\ge1.
Step 6 of 6: The whole sequence is an arithmetic progression
In plain words

Once P is identified with this specific product, the output tuple of increments it forces is unique, so eventually the sequence must repeat the same increment forever, and induction backward shows it was always arithmetic.

an=a1+(n−1)d for all na_n=a_1+(n-1)d\ \text{for all } n
Detailed analysis

There is a finite set of possible tuples, and for each tuple ee let Qe(X)=∏j=1k(X+e1+⋯+ej)Q_e(X)=\prod_{j=1}^k(X+e_1+\cdots+e_j). The tuple from Step 5 satisfies Qe=PQ_e=P. Every other tuple can occur only finitely often: otherwise the distinct values ana_n at those indices would give infinitely many roots of the nonzero polynomial Qe−PQ_e-P. Thus, for all sufficiently large nn, δ(n)=(d1,…,dk)\delta(n)=(d_1,\ldots,d_k). Applying this also to δ(n+1)\delta(n+1) shows (d2,…,dk,an+k+1−an+k)=(d1,…,dk)(d_2,\ldots,d_k,a_{n+k+1}-a_{n+k})=(d_1,\ldots,d_k), so d1=⋯=dk=:dd_1=\cdots=d_k=:d and the tail is an arithmetic progression. Consequently P(X)=F(X):=∏j=1k(X+jd)P(X)=F(X):=\prod_{j=1}^k(X+jd). To extend the progression backwards, suppose an+1,…,an+ka_{n+1},\ldots,a_{n+k} already have common difference dd. Then F(an)=P(an)=∏j=1k(an+1+(j−1)d)=F(an+1−d)F(a_n)=P(a_n)=\prod_{j=1}^k(a_{n+1}+(j-1)d)=F(a_{n+1}-d). The function F(t)F(t) is strictly increasing for t>−dt>-d, and both arguments lie in that interval, so an=an+1−da_n=a_{n+1}-d. Downward induction reaches n=1n=1 and proves an=a1+(n−1)da_n=a_1+(n-1)d for every nn, with common difference dd.