🧮 Brain Teaser

A Polynomial Vanishing on All Lattice Points in a Strip

Let p(x,y)p(x, y) be a polynomial with real coefficients such that p(m,n)=0p(m, n) = 0 for every pair of integers (m,n)(m, n) with m0m \geq 0 and 0nm0 \leq n \leq m. Must pp be identically zero?

polynomialslattice pointsvanishingtwo-variable polynomialszeros

Answer: A Polynomial Vanishing on All Lattice Points in a Strip

Key Idea / Intuition

The lattice points {(m,n):m0, 0nm}\{(m,n) : m \geq 0,\ 0 \leq n \leq m\} form an infinite triangular array — there are infinitely many of them, but they don't "fill" the plane in the way needed to force a polynomial to vanish. The key insight is to think about what happens when you fix mm: for each fixed integer m0m \geq 0, the polynomial p(m,y)p(m, y) (a polynomial in yy alone) vanishes at m+1m+1 values n=0,1,,mn = 0, 1, \ldots, m. If the degree of pp in yy is dd, then once mdm \geq d, there are more zeros than the degree, forcing p(m,y)0p(m, y) \equiv 0 as a polynomial in yy for each sufficiently large integer mm. That means every coefficient (a polynomial in xx) vanishes at infinitely many xx-values, forcing them all to be zero.


Formal Proof / Solution

Setup. Write p(x,y)p(x,y) as a polynomial in yy with coefficients that are polynomials in xx:

p(x,y)=k=0dak(x)yk,p(x,y) = \sum_{k=0}^{d} a_k(x)\, y^k,

where each ak(x)a_k(x) is a polynomial in xx, and dd is the degree of pp in yy.

Step 1: Fix a large integer mm. For any integer mdm \geq d, the polynomial in yy:

qm(y):=p(m,y)=k=0dak(m)ykq_m(y) := p(m, y) = \sum_{k=0}^{d} a_k(m)\, y^k

has degree at most dd in yy. By hypothesis, qm(n)=0q_m(n) = 0 for n=0,1,2,,mn = 0, 1, 2, \ldots, m. That gives m+1d+1m+1 \geq d+1 zeros. Since a nonzero polynomial of degree d\leq d can have at most dd roots, we conclude:

qm(y)0as a polynomial in y.q_m(y) \equiv 0 \quad \text{as a polynomial in } y.

Step 2: Each coefficient vanishes at infinitely many integers. From Step 1, for every integer mdm \geq d we have ak(m)=0a_k(m) = 0 for all k=0,1,,dk = 0, 1, \ldots, d. This means the polynomial ak(x)a_k(x) vanishes at the infinite set {d,d+1,d+2,}\{d, d+1, d+2, \ldots\}.

Step 3: Conclude. A nonzero polynomial in one variable can only have finitely many roots. Since each ak(x)a_k(x) vanishes at infinitely many values of xx, we must have:

ak(x)0for all k=0,1,,d.a_k(x) \equiv 0 \quad \text{for all } k = 0, 1, \ldots, d.

Therefore p(x,y)0p(x,y) \equiv 0.

Answer: Yes, pp must be identically zero. \blacksquare


Why this is surprising: The lattice points only occupy the triangular region 0nm0 \leq n \leq m, which is a thin subset of Z2\mathbb{Z}^2 (in particular, a polynomial that vanishes on all of Z2\mathbb{Z}^2 would obviously be zero, but this is less obvious). The trick is that the triangular array is not thin in the critical direction: for each fixed mm, there are enough points to kill a polynomial in yy.

Source: Mathematical folklore / Putnam competition style

Type: PutnamSource: Mathematical folklore / Putnam competition styleEdit on GitHub ↗