🧮 Brain Teaser

The Polynomial That Stays Small on Integers

Let p(x)p(x) be a polynomial of degree nn with leading coefficient 11 (i.e., monic). Prove that

max⁡0≤x≤n∣p(x)∣≥n!2n.\max_{0 \le x \le n} |p(x)| \ge \frac{n!}{2^n}.

In other words, no monic degree-nn polynomial can be uniformly small on the integers 0,1,…,n0, 1, \ldots, n.

finite differencesmonic polynomialcombinatoricsextremal polynomial

Answer: The Polynomial That Stays Small on Integers

Key Idea / Intuition

The integers 0,1,…,n0, 1, \ldots, n give us n+1n+1 points, and a monic degree-nn polynomial is completely pinned by its values there — in fact, the unique monic degree-nn polynomial that passes through prescribed values at 0,1,…,n0, 1, \ldots, n is determined by the Newton forward difference formula. The key insight is that the leading coefficient of a monic polynomial forces a specific alternating sum of its values at these integer points to equal exactly n!n!, so those values can't all be small.


Formal Proof / Solution

Step 1: The finite difference operator.

Define the nn-th finite difference of any function ff at 00 by

Δnf(0)=∑k=0n(−1)n−k(nk)f(k).\Delta^n f(0) = \sum_{k=0}^{n} (-1)^{n-k} \binom{n}{k} f(k).

A key fact from finite difference calculus: if p(x)p(x) is a polynomial of degree nn with leading coefficient ana_n, then

Δnp(0)=n!⋅an.\Delta^n p(0) = n! \cdot a_n.

Quick proof: It suffices to check on the basis {1,x,x2,…,xn}\{1, x, x^2, \ldots, x^n\}. One can show Δnxn=n!\Delta^n x^n = n! and Δnxj=0\Delta^n x^j = 0 for j<nj < n.

Step 2: Apply to our monic polynomial.

Since p(x)p(x) is monic of degree nn, we have an=1a_n = 1, so

Δnp(0)=∑k=0n(−1)n−k(nk)p(k)=n!.\Delta^n p(0) = \sum_{k=0}^{n} (-1)^{n-k} \binom{n}{k} p(k) = n!.

Step 3: Bound via the maximum.

Let M=max⁡0≤k≤n∣p(k)∣M = \max_{0 \le k \le n} |p(k)|. Then by the triangle inequality:

n!=∣∑k=0n(−1)n−k(nk)p(k)∣≤∑k=0n(nk)∣p(k)∣≤M∑k=0n(nk)=M⋅2n.n! = \left|\sum_{k=0}^{n} (-1)^{n-k} \binom{n}{k} p(k)\right| \le \sum_{k=0}^{n} \binom{n}{k} |p(k)| \le M \sum_{k=0}^{n} \binom{n}{k} = M \cdot 2^n.

Step 4: Conclude.

Rearranging gives

M=max⁡0≤k≤n∣p(k)∣≥n!2n.M = \max_{0 \le k \le n} |p(k)| \ge \frac{n!}{2^n}.

■\blacksquare


Sharpness: The bound is essentially achieved by the polynomial p(x)=∏k=0n−1(x−k/1)⋅(shifted)p(x) = \prod_{k=0}^{n-1}(x - k/1) \cdot \text{(shifted)}... more precisely, the Chebyshev-like polynomial on {0,…,n}\{0,\ldots,n\} comes close. A clean equality case is p(x)=x(x−1)(x−2)⋯(x−n+1)⋅p(x) = x(x-1)(x-2)\cdots(x-n+1) \cdot (adjusted), but note q(x)=x(x−1)⋯(x−n+1)q(x) = x(x-1)\cdots(x-n+1) is already monic of degree nn and max⁡0≤k≤n∣q(k)∣=0\max_{0\le k \le n} |q(k)| = 0 since qq vanishes at all of 0,1,…,n−10,1,\ldots,n-1... wait, q(n)=n!q(n) = n!. So qq itself achieves the bound with equality: max⁡∣q(k)∣=n!≥n!/2n\max |q(k)| = n! \ge n!/2^n. The bound is non-trivial precisely because we're claiming n!/2nn!/2^n is a lower bound, and qq exceeds it.

Source: Classical Putnam/olympiad folklore; related to Putnam techniques on polynomial lower bounds

Type: PutnamSource: Classical Putnam/olympiad folklore; related to Putnam techniques on polynomial lower boundsEdit on GitHub ↗