Maths Olympiad Prep

Library / /366 of 397

Algebra Difficulty 7.1 National Olympiad, round 2 Prove it Taiwan

Let m,n2m, n \ge 2 be integers and let f(x1,,xn)f(x_1, \dots, x_n) be a polynomial with real coefficients such that for every x1,x2,,xn{0,1,,m1}x_1, x_2, \dots, x_n \in \{0, 1, \dots, m-1\},
f(x1,,xn)=[x1++xnm] f(x_1, \dots, x_n) = \left[ \frac{x_1 + \dots + x_n}{m} \right]
holds. Prove that the degree of ff is at least nn.

Solution

We transform the problem to a single variable question by the following.

Lemma. Let a1,,ana_1, \dots, a_n be nonnegative integers and let G(x)G(x) be a nonzero polynomial with degGa1++an\deg G \le a_1 + \dots + a_n. Suppose that some polynomial F(x1,,xn)F(x_1, \dots, x_n) satisfies
F(x1,,xn)=G(x1++xn) F(x_1, \dots, x_n) = G(x_1 + \dots + x_n)
for (x1,,xn){0,1,,a1}××{0,1,,an}(x_1, \dots, x_n) \in \{0, 1, \dots, a_1\} \times \dots \times \{0, 1, \dots, a_n\}.
Then FF cannot be zero polynomial, and degFdegG\deg F \ge \deg G.

For proving the lemma, we will use forward differences of polynomials. If p(x)p(x) is a polynomial with a single variable, then define
(Δp)(x)=p(x+1)p(x). (\Delta p)(x) = p(x + 1) - p(x).
It is well-known that if pp is a nonconstant polynomial then
degΔp=degp1. \deg \Delta p = \deg p - 1.
If p(x1,,xn)p(x_1, \dots, x_n) is a polynomial with nn variables and 1kn1 \le k \le n then let
Δk(p)(x1,,xn)=p(x1,,xk1,xk+1,xk+1,,xn)p(x1,,xn). \Delta_k(p)(x_1, \dots, x_n) = p(x_1, \dots, x_{k-1}, x_k+1, x_{k+1}, \dots, x_n) - p(x_1, \dots, x_n).

It is also well-known that either Δkp\Delta_k p is the zero polynomial or
deg(Δkp)degp1. \deg(\Delta_k p) \leq \deg p - 1.

Proof of the lemma. We apply induction on the degree of GG, If GG is a constant polynomial then we have F(0,,0)=G(0)0F(0, \dots, 0) = G(0) \neq 0, so FF cannot be the zero polynomial.
Suppose that degG1\deg G \geq 1 and the lemma holds true for lower degrees. Since
a1++andegG>0, a_1 + \dots + a_n \geq \deg G > 0,
at least one of a1,,ana_1, \dots, a_n is positive; without loss of generality suppose a11a_1 \geq 1.
Consider the polynomials F1=Δ1FF_1 = \Delta_1 F and G1=ΔGG_1 = \Delta G. On the grid {0,,a11}×{0,,a2}×{0,,an}\{0, \dots, a_1 - 1\} \times \{0, \dots, a_2\} \times \dots \{0, \dots, a_n\} we have
F1(x1,,xn)=F(x1+1,x2,,xn)F(x1,x2,,xn)=G(x1++xn+1)G(x1++xn)=G1(x1++xn). \begin{aligned} F_1(x_1, \dots, x_n) &= F(x_1 + 1, x_2, \dots, x_n) - F(x_1, x_2, \dots, x_n) \\ &= G(x_1 + \dots + x_n + 1) - G(x_1 + \dots + x_n) \\ &= G_1(x_1 + \dots + x_n). \end{aligned}
Since GG is nonconstant, we have
degG1=degG1(a11)+a2++an. \deg G_1 = \deg G - 1 \leq (a_1 - 1) + a_2 + \dots + a_n.
Therefore we can apply the induction hypothesis to F1F_1 and G1G_1 and conclude that F1F_1 is not the zero polynomial and degF1degG1\deg F_1 \geq \deg G_1. Hence,
degFdegF1+1degG1+1=degG. \deg F \geq \deg F_1 + 1 \geq \deg G_1 + 1 = \deg G.

To prove the problem statement, take the unique polynomial g(x)g(x) so that
g(x)=[xm] for x{0,1,,n(m1)} and deggn(m1). g(x) = \left[ \frac{x}{m} \right] \text{ for } x \in \{0, 1, \dots, n(m-1)\} \text{ and } \deg g \le n(m-1).
Notice that precisely n(m1)+1n(m-1)+1 values of gg are prescribed, so g(x)g(x) indeed exists and is unique. Notice further that the constraints g(0)=g(1)=0g(0) = g(1) = 0 and g(m)=1g(m) = 1 together enforce degg2\deg g \ge 2.
By applying the lemma to a1==an=m1a_1 = \dots = a_n = m-1 and the polynomials ff and gg, we achieve degfdegg\deg f \ge \deg g. Hence we just need a suitable lower bound on degg\deg g.

Consider the polynomial
h(x)=g(x+m)g(x)1. h(x) = g(x + m) - g(x) - 1.
The degree of g(x+m)g(x)g(x+m) - g(x) is degg11\deg g - 1 \ge 1, so
degh=degg11, \deg h = \deg g - 1 \geq 1,
and therefore hh cannot be the zero polynomial. On the other hand, hh vanishes at the points 0,1,,n(m1)m0, 1, \dots, n(m-1)-m, so hh has at least (n1)(m1)(n-1)(m-1) roots. Hence,
degfdegg=degh+1(n1)(m1)+1n. \deg f \geq \deg g = \deg h + 1 \geq (n-1)(m-1) + 1 \geq n.

At the point (b1,,bn)(b_1, \dots, b_n) we have
H(b1,,bn)=G(d)G0(d)0. H(b_1, \dots, b_n) = G(d) - G_0(d) \neq 0.
At all other points of the grid we have F=GF = G and therefore
H=GG0=0. H = G - G_0 = 0.
So, by the Alon-Füredi bound,
degHb1++bn=d. \deg H \geq b_1 + \dots + b_n = d.
Since degG0<d\deg G_0 < d, this implies
degF=deg(H+G0)=degHd=degG. \deg F = \deg (H + G_0) = \deg H \geq d = \deg G.

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 translated into English from zh; metadata (topic, difficulty) added by this project.