Maths Olympiad Prep

Library / /48 of 48

Algebra Difficulty 8.3 Shortlist Prove it Asia Pacific Mathematics Olympiad (APMO)

Let Z\mathbb{Z} denote the set of all integers. Find all polynomials P(x)P(x) with integer coefficients that satisfy the following property:
For any infinite sequence a1,a2,a_{1}, a_{2}, \ldots of integers in which each integer in Z\mathbb{Z} appears exactly once, there exist indices i<ji<j and an integer kk such that ai+ai+1++aj=P(k)a_{i}+a_{i+1}+\cdots+a_{j}=P(k).

Solution

Part 1: All polynomials with degP=1\operatorname{deg} P=1 satisfy the given property.
Suppose P(x)=cx+dP(x)=c x+d, and assume without loss of generality that c>d0c>d \geq 0. Denote si=a1+a2++ai(modc)s_{i}=a_{1}+a_{2}+ \cdots+a_{i} (\bmod c). It suffices to show that there exist indices ii and jj such that ji2j-i \geq 2 and sjsid(modc)s_{j}-s_{i} \equiv d \pmod{c}.

Consider c+1c+1 indices e1,e2,,ec+1>1e_{1}, e_{2}, \ldots, e_{c+1}>1 such that aeld(modc)a_{e_{l}} \equiv d \pmod{c}. By the pigeonhole principle, among the n+1n+1 pairs (se11,se1),(se21,se2),,(sn+11,sn+1)\left(s_{e_{1}-1}, s_{e_{1}}\right),\left(s_{e_{2}-1}, s_{e_{2}}\right), \ldots,\left(s_{n+1-1}, s_{n+1}\right), some two are equal, say (sm1,sm)\left(s_{m-1}, s_{m}\right) and (sn1,sn)\left(s_{n-1}, s_{n}\right). We can then take i=m1i=m-1 and j=nj=n.

Part 2: All polynomials with degP1\operatorname{deg} P \neq 1 do not satisfy the given property.
Lemma: If degP1\operatorname{deg} P \neq 1, then for any positive integers A,BA, B, and CC, there exists an integer yy with y>C|y|>C such that no value in the range of PP falls within the interval [yA,y+B][y-A, y+B].

Proof of Lemma: The claim is immediate when PP is constant or when degP\operatorname{deg} P is even since PP is bounded from below. Let P(x)=anxn++a1x+a0P(x)=a_{n} x^{n}+\cdots+a_{1} x+a_{0} be of odd degree greater than 1, and assume without loss of generality that an>0a_{n}>0. Since P(x+1)P(x)=annxn1+P(x+1)-P(x)=a_{n} n x^{n-1}+\ldots, and n1>0n-1>0, the gap between P(x)P(x) and P(x+1)P(x+1) grows arbitrarily for large xx. The claim follows. \square

Suppose degP1\operatorname{deg} P \neq 1. We will inductively construct a sequence {ai}\{a_{i}\} such that for any indices i<ji<j and any integer kk it holds that ai+ai+1++ajP(k)a_{i}+a_{i+1}+\cdots+a_{j} \neq P(k). Suppose that we have constructed the sequence up to aia_{i}, and mm is an integer with smallest magnitude yet to appear in the sequence. We will add two more terms to the sequence. Take ai+2=ma_{i+2}=m. Consider all the new sums of at least two consecutive terms; each of them contains ai+1a_{i+1}. Hence all such sums are in the interval [ai+1A,ai+1+B][a_{i+1}-A, a_{i+1}+B] for fixed constants A,BA, B. The lemma allows us to choose ai+1a_{i+1} so that all such sums avoid the range of PP.

Alternate Solution for Part 1: Again, suppose P(x)=cx+dP(x)=c x+d, and assume without loss of generality that c>d0c>d \geq 0. Let Si={aj+aj+1++ai(modc)j=1,2,,i}S_{i}=\{a_{j}+a_{j+1}+\cdots+a_{i} (\bmod c) \mid j=1,2, \ldots, i\}. Then Si+1={si+ai+1(modc)siSi}{ai+1(modc)}S_{i+1}=\{s_{i}+a_{i+1} (\bmod c) \mid s_{i} \in S_{i}\} \cup \{a_{i+1} (\bmod c)\}. Hence Si+1=Si|S_{i+1}|=|S_{i}| or Si+1=Si+1|S_{i+1}|=|S_{i}|+1, with the former occurring exactly when 0Si0 \in S_{i}. Since Sic|S_{i}| \leq c, the latter can only occur finitely many times, so there exists II such that 0Si0 \in S_{i} for all iIi \geq I. Let t>It>I be an index with atd(modc)a_{t} \equiv d \pmod{c}. Then we can find a sum of at least two consecutive terms ending at ata_{t} and congruent to d(modc)d \pmod{c}.

Alternate Construction when P(x)P(x) is constant or of even degree
If P(x)P(x) is of even degree, then PP is bounded from below or from above. In case PP is constant or bounded from above, then there exists a positive integer cc such that P(x)<cP(x)<c. Let {ai}\{a_{i}\} be the sequence
0,1,1,2,3,2,4,5,3, 0,1,-1,2,3,-2,4,5,-3, \cdots
which is given by a3n+1=2na_{3 n+1}=2 n, a3n+2=2n+1a_{3 n+2}=2 n+1, a3n+3=(n+1)a_{3 n+3}=-(n+1) for all n0n \geq 0. Notice that for any i<ji<j we have ai++aj0a_{i}+\cdots+a_{j} \geq 0. Then for the sequence {bn}\{b_{n}\} defined by bn=an+cb_{n}=a_{n}+c, clearly bi++bj(ai++aj)+2c>cb_{i}+\cdots+b_{j} \geq (a_{i}+\cdots+a_{j})+2c>c which is outside the range of P(x)P(x).

Now if PP is bounded from below, there exists a positive integer cc such that P(x)>cP(x)>-c. In this case, take bn=ancb_{n}=-a_{n}-c. Then for all i<ji<j we have bi++bj(ai++aj)2c<cb_{i}+\cdots+b_{j} \leq -(a_{i}+\cdots+a_{j})-2c<-c which is again outside the range of P(x)P(x).

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.