Maths Olympiad Prep

Library / United States — Team Selection Test /37 of 40

United States — Team Selection Test

Number theory Difficulty 9.0 IMO level Prove it United States

Let pp be a prime. We say that a sequence of integers {zn}n=0∞\{z_n\}_{n=0}^{\infty} is a pp-pod if for each e≥0e \ge 0, there is an N≥0N \ge 0 such that whenever m≥Nm \ge N, pep^e divides the sum
∑k=0m(−1)k(mk)zk. \sum_{k=0}^{m} (-1)^k \binom{m}{k} z_k.

Prove that if both sequences {xn}n=0∞\{x_n\}_{n=0}^{\infty} and {yn}n=0∞\{y_n\}_{n=0}^{\infty} are pp-pods, then the sequence {xnyn}n=0∞\{x_n y_n\}_{n=0}^{\infty} is a pp-pod.

Solution

Let
Xn=∑i=0n(−1)i(ni)xiandYn=∑i=0n(−1)i(ni)yi. X_n = \sum_{i=0}^{n} (-1)^i \binom{n}{i} x_i \quad \text{and} \quad Y_n = \sum_{i=0}^{n} (-1)^i \binom{n}{i} y_i.
For nonnegative integers i≤ji \le j, consider the expression
∑k=ij(−1)k(ki)(jk). \sum_{k=i}^{j} (-1)^k \binom{k}{i} \binom{j}{k}.
Viewing (jk)\binom{j}{k} as the number of ways to choose a kk-element subset of a jj-element set and (ki)\binom{k}{i} as the number of ways to choose an ii-element subset which is contained in this chosen kk-element set, the principle of inclusion-exclusion shows that (−1)j(-1)^j times this sum is the number of ways to choose an ii-element subset from a jj-element set that coincides with the set itself. Thus, the sum is (−1)i(-1)^i if i=ji = j and 00 otherwise. Therefore, we may write
∑r=0n(−1)r(nr)xryr=∑i=0n∑j=0n(−1)i+jxiyj∑r=0n(−1)r(nr)[∑k=ir(−1)k(ki)(rk)][∑ℓ=jr(−1)ℓ(ℓj)(rℓ)]=∑k=0n∑ℓ=0n[∑i=0k(−1)k−i(ki)xi][∑j=0ℓ(−1)ℓ−j(ℓj)yj][∑r=0n(−1)r(nr)(rk)(rℓ)]=∑k=0n∑ℓ=0n(−1)n+k+ℓXkYℓ[∑r=0n(−1)n−r(nr)(rk)(rℓ)]. \begin{aligned} \sum_{r=0}^{n} (-1)^r \binom{n}{r} x_r y_r &= \sum_{i=0}^{n} \sum_{j=0}^{n} (-1)^{i+j} x_i y_j \sum_{r=0}^{n} (-1)^r \binom{n}{r} \left[ \sum_{k=i}^{r} (-1)^k \binom{k}{i} \binom{r}{k} \right] \left[ \sum_{\ell=j}^{r} (-1)^\ell \binom{\ell}{j} \binom{r}{\ell} \right] \\ &= \sum_{k=0}^{n} \sum_{\ell=0}^{n} \left[ \sum_{i=0}^{k} (-1)^{k-i} \binom{k}{i} x_i \right] \left[ \sum_{j=0}^{\ell} (-1)^{\ell-j} \binom{\ell}{j} y_j \right] \left[ \sum_{r=0}^{n} (-1)^r \binom{n}{r} \binom{r}{k} \binom{r}{\ell} \right] \\ &= \sum_{k=0}^{n} \sum_{\ell=0}^{n} (-1)^{n+k+\ell} X_k Y_\ell \left[ \sum_{r=0}^{n} (-1)^{n-r} \binom{n}{r} \binom{r}{k} \binom{r}{\ell} \right]. \end{aligned}
Again by the principle of inclusion-exclusion, the final expression in brackets counts the number of ways to choose a kk-element subset and an ℓ\ell-element subset of an nn-element set whose union is the entire set (again by the principle of inclusion-exclusion), so it is 00 if k+ℓ<nk + \ell < n.

Now, let ee be arbitrary, let NN be so large that pep^e divides both XmX_m and YmY_m whenever m≥N/2m \ge N/2 (such an NN exists by the definition of pp-pod), and take n≥Nn \ge N. Then, whenever k+ℓ≥n≥Nk + \ell \ge n \ge N, at least one of kk or ℓ\ell is at least N/2N/2, so pep^e will divide either XkX_k or YℓY_\ell. We conclude that pep^e divides every term in the sum, and that therefore for each n≥Nn \ge N, pep^e divides
∑r=0n(−1)r(nr)xryr, \sum_{r=0}^{n} (-1)^r \binom{n}{r} x_r y_r,
hence {xnyn}n=0∞\{x_n y_n\}_{n=0}^\infty is a pp-pod.

Solution 2:

For a sequence {zn}n=0∞\{z_n\}_{n=0}^\infty and an integer m≥0m \ge 0, define
Δmzn=∑k=0m(−1)m−k(mk)zn+k, \Delta^m z_n = \sum_{k=0}^{m} (-1)^{m-k} \binom{m}{k} z_{n+k},
so that {Δmzn}n=0∞\{\Delta^m z_n\}_{n=0}^\infty is the sequence of mm-fold finite differences in {zn}\{z_n\}. With this notation, {zn}\{z_n\} is a pp-pod if and only if for every e>0e > 0 there is an N≥0N \ge 0 such that whenever m≥Nm \ge N, pe∣Δmz0p^e \mid \Delta^m z_0.

Lemma 1. The sequence {zn}\{z_n\} is a pp-pod if and only if for every e>0e > 0 there is an N≥0N \ge 0 such that whenever m≥Nm \ge N, pep^e divides Δmzn\Delta^m z_n for every n≥0n \ge 0.

Proof. Indeed, recall (or verify) that for any m,n≥0m, n \ge 0, we have
Δm+1zn=Δmzn+1−Δmzn,orΔmzn+1=Δmzn+Δm+1zn.(33) \Delta^{m+1} z_n = \Delta^m z_{n+1} - \Delta^m z_n, \quad \text{or} \quad \Delta^m z_{n+1} = \Delta^m z_n + \Delta^{m+1} z_n. \qquad (33)
A trivial induction using (33) shows that the same NN as in the definition of pp-pod works for all n≥0n \ge 0. □\square

Now, let {xn}\{x_n\} and {yn}\{y_n\} be pp-pods. By Lemma 1, there exists an NN such that for all nn, Δmxn≡0(modpe)\Delta^m x_n \equiv 0 \pmod{p^e} for m≥Nm \ge N. Let f(t)f(t) be the monic degree NN polynomial with rational coefficients such that f(n)=xnf(n) = x_n for n=0,1,…,Nn = 0, 1, \dots, N. By definition, we see that Δmx0=Δmf(0)\Delta^m x_0 = \Delta^m f(0) for m≤Nm \le N; further, because Δmf(n)=0\Delta^m f(n) = 0 for m≥Nm \ge N, we see that Δmx0≡Δmf(0)(modpe)\Delta^m x_0 \equiv \Delta^m f(0) \pmod{p^e} for all m≥0m \ge 0. An easy induction using (33) then shows that Δmxn≡Δmf(n)(modpe)\Delta^m x_n \equiv \Delta^m f(n) \pmod{p^e} for all m,n≥0m, n \ge 0. Taking m=0m = 0 in particular gives that f(n)≡xn(modpe)f(n) \equiv x_n \pmod{p^e} for all n≥0n \ge 0. Similarly, we find a polynomial g(t)g(t) such that g(n)≡yn(modpe)g(n) \equiv y_n \pmod{p^e} for all n≥0n \ge 0.

But then h(t)=f(t)g(t)h(t) = f(t)g(t) has the property that h(n)≡xnyn(modpe)h(n) \equiv x_n y_n \pmod{p^e} for all n≥0n \ge 0. Therefore, if h(t)h(t) has degree MM, then for any m≥Mm \ge M and n≥0n \ge 0 we have
Δm{xnyn}≡Δmh(n)≡0(modpe), \Delta^m \{x_n y_n\} \equiv \Delta^m h(n) \equiv 0 \pmod{p^e},
and we are done by Lemma 1.

Want a route through all this instead of an archive? The track puts 2,604 problems in a working order, from Junior Challenge level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.