Olympiad Maths Prep

Track / Stage 9 / 42 of 80 #1922 of 2000

Problem 1922

IMO P2/P5; hard shortlist
Geometry Difficulty 9.1 Prove it IMO · United States

Let nn be a positive integer. Consider
S={(x,y,z):x,y,z{0,1,,n},x+y+z>0}. S = \{(x, y, z) : x, y, z \in \{0, 1, \dots, n\}, x + y + z > 0\}.
as a set of (n+1)31(n+1)^3 - 1 points in three-dimensional space. Determine the smallest possible number of planes, the union of which contains SS but does not include (0,0,0)(0, 0, 0).
(This problem was suggested by the Netherlands.)

This one wants a proof. Work it on paper, read the official solution, then mark yourself honestly — the ladder only means something if the record is true.

Official solutions — 2

Solution 1

We establish the following key lemma.

Lemma 1. Consider a nonzero polynomial P(x1,,xk)P(x_1, \dots, x_k) in kk variables. Suppose that PP vanishes at all points (x1,,xk)(x_1, \dots, x_k) such that x1,,xk{0,1,,n}x_1, \dots, x_k \in \{0, 1, \dots, n\} and x1++xk>0x_1 + \dots + x_k > 0, while P(0,0,,0)0P(0, 0, \dots, 0) \ne 0. Then degPkn\deg P \ge kn.

*Proof.* We induct on kk. The base case k=0k=0 is clear since P0P \ne 0. We assume that the statement is true for k=1k = \ell - 1 for some positive integer \ell. Now we consider the case k=k = \ell. Denote for clarity y=xk=xy = x_k = x_\ell.

Let R(x1,,xk1,y)R(x_1, \dots, x_{k-1}, y) be the residue of PP modulo Q(y)=y(y1)(yn)Q(y) = y(y-1)\dots(y-n). Polynomial Q(y)Q(y) vanishes at each y=0,1,,ny=0, 1, \dots, n, hence P(x1,,xk1,y)=R(x1,,xk1,y)P(x_1, \dots, x_{k-1}, y) = R(x_1, \dots, x_{k-1}, y) for all x1,,xk1,y{0,1,,n}x_1, \dots, x_{k-1}, y \in \{0, 1, \dots, n\}. Therefore, RR also satisfies the condition of the Lemma; moreover, degyRn\deg_y R \le n. Clearly, degRdegP\deg R \le \deg P, so it suffices to prove that degRnk\deg R \ge nk.

Now, expand polynomial RR in the powers of yy:
R(x1,,xk1,y)=Rn(x1,,xk1)yn+Rn1(x1,,xk1)yn1++R0(x1,,xk1). R(x_1, \dots, x_{k-1}, y) = R_n(x_1, \dots, x_{k-1})y^n + R_{n-1}(x_1, \dots, x_{k-1})y^{n-1} + \dots + R_0(x_1, \dots, x_{k-1}).
We show that the polynomial Rn(x1,,xk1)R_n(x_1, \dots, x_{k-1}) satisfies the condition of the induction hypothesis.

Consider the polynomial T(y)=R(0,,0,y)T(y) = R(0, \dots, 0, y) of degree n\le n. This polynomial has nn roots y=1,,ny = 1, \dots, n; on the other hand, T(y)0T(y) \ne 0 since T(0)0T(0) \ne 0. Hence degT=n\deg T = n, and its leading coefficient is Rn(0,0,,0)0R_n(0, 0, \dots, 0) \ne 0. (For example, in the case k=1k=1 we obtain that coefficient RnR_n is nonzero.)

Similarly, take any numbers a1,,ak1{0,1,,n}a_1, \dots, a_{k-1} \in \{0, 1, \dots, n\} with a1++ak1>0a_1 + \dots + a_{k-1} > 0. Substituting xi=aix_i = a_i into R(x1,,xk1,y)R(x_1, \dots, x_{k-1}, y), we get a polynomial in yy which vanishes at all points y=0,,ny = 0, \dots, n and has degree n\le n. Therefore, this polynomial is null, hence Ri(a1,,ak1)=0R_i(a_1, \dots, a_{k-1}) = 0 for all i=0,1,,ni = 0, 1, \dots, n. In particular, Rn(a1,,ak1)=0R_n(a_1, \dots, a_{k-1}) = 0.

Thus, the polynomial Rn(x1,,xk1)R_n(x_1, \dots, x_{k-1}) satisfies the condition of the induction hypothesis. So, we have degRn(k1)n\deg R_n \ge (k-1)n and degPdegRdegRn+nkn\deg P \ge \deg R \ge \deg R_n + n \ge kn. \square

Now we can finish the solution. Suppose that there are NN planes covering all the points of SS but not containing the origin. Let their equations be aix+biy+ciz+di=0a_i x + b_i y + c_i z + d_i = 0. Consider the polynomial
P(x,y,z)=i=1N(aix+biy+ciz+di). P(x, y, z) = \prod_{i=1}^{N} (a_i x + b_i y + c_i z + d_i).
It has total degree NN. This polynomial has the property that P(x0,y0,z0)=0P(x_0, y_0, z_0) = 0 for any (x0,y0,z0)S(x_0, y_0, z_0) \in S, while P(0,0,0)0P(0, 0, 0) \ne 0. Hence by Lemma 1 we get N=degP3nN = \deg P \ge 3n, as desired.

Solution 2

Suppose rr planes are given. As in the first solution, let A1,,ArA_1, \dots, A_r be (nonzero) linear functions over R3\mathbb{R}^3 such that the equations of the planes are Ai(x,y,z)=0A_i(x, y, z) = 0. Define the polynomial P0=A1A2ArP_0 = A_1A_2\cdots A_r, whose degree is rr. If a lattice point (x,y,z)(x, y, z) satisfies 0xa0 \le x \le a, 0yb0 \le y \le b, 0zc0 \le z \le c, and x+y+z>0x + y + z > 0, then P0(x,y,z)=0P_0(x, y, z) = 0; however, P0(0,0,0)0P_0(0, 0, 0) \ne 0.

Polynomials P1,,PaP_1, \dots, P_a are defined recursively by Pi+1(x,y,z)=Pi(x+1,y,z)Pi(x,y,z)P_{i+1}(x, y, z) = P_i(x+1, y, z) - P_i(x, y, z). By induction, we see that Pi(x,y,z)=0P_i(x, y, z) = 0 if 0xai0 \le x \le a-i, 0yb0 \le y \le b, 0zc0 \le z \le c, and x+y+z>0x+y+z > 0, while Pi(0,0,0)0P_i(0, 0, 0) \ne 0. Furthermore, if PiP_i is a nonzero polynomial, then its degree is rir-i.

Let Q0=PaQ_0 = P_a, and construct polynomials Q1,,QbQ_1, \dots, Q_b as above. Specifically, define Qi+1(x,y,z)=Qi(x,y+1,z)Qi(x,y,z)Q_{i+1}(x, y, z) = Q_i(x, y+1, z) - Q_i(x, y, z). Again, we have Qi(x,y,z)=0Q_i(x, y, z) = 0 provided x=0,0ybi,0zcx = 0, 0 \le y \le b-i, 0 \le z \le c, and x+y+z>0x+y+z > 0, while Qi(0,0,0)0Q_i(0, 0, 0) \ne 0. If QiQ_i is nonzero, then its degree is rair-a-i.

Finally, let R0=QbR_0 = Q_b, and define R1,,RcR_1, \dots, R_c by Ri+1(x,y,z)=Ri(x,y,z+1)Ri(x,y,z)R_{i+1}(x, y, z) = R_i(x, y, z+1) - R_i(x, y, z). As above, Ri(x,y,z)=0R_i(x, y, z) = 0 if x=y=0,0zcix = y = 0, 0 \le z \le c-i, and x+y+z>0x+y+z > 0. If RiR_i is nonzero then its degree is rabir-a-b-i.

Consider the polynomial Rc(x,y,z)R_c(x, y, z). Its value at (0,0,0)(0, 0, 0) is nonzero, so it is a nonzero polynomial. Its degree, which must be nonnegative, is rabcr-a-b-c. Therefore, we have ra+b+cr \ge a+b+c, as desired.

Therefore, a+b+ca+b+c planes are necessary, and it is possible to cover all the points with a+b+ca+b+c planes. In the original problem, a=b=c=na=b=c=n, so the answer is 3n3n.

Real numbers x0,,xa,y0,,yb,z0,,zcx_0, \dots, x_a, y_0, \dots, y_b, z_0, \dots, z_c are given such that the xix_i are distinct, the yiy_i are distinct, and the ziz_i are distinct. If a collection of planes covers all points (xi,yj,zk)(x_i, y_j, z_k) except (x0,y0,z0)(x_0, y_0, z_0), the collection contains at least a+b+ca+b+c points.

To prove this we need a more general form of the method of finite differences used above.

Lemma 2. Suppose distinct reals t0,,tnt_0, \dots, t_n are given. Then there exist weights w0,,wnw_0, \dots, w_n, with w00w_0 \neq 0, such that for each nonnegative integer i<ni < n, we have w0t0i++wntni=0w_0 t_0^i + \dots + w_n t_n^i = 0, while w0t0n++wntnn0w_0 t_0^n + \dots + w_n t_n^n \neq 0.

*Proof.* The n+1n+1 vectors (1,tj,tj2,,tjn1)(1, t_j, t_j^2, \dots, t_j^{n-1}) are linearly dependent over Rn\mathbb{R}^n, so there exist weights wjw_j such that w0t0i++wntni=0w_0 t_0^i + \dots + w_n t_n^i = 0 for each i<ni < n. However, since the tjt_j are distinct, the vectors (1,tj,tj2,,tjn)(1, t_j, t_j^2, \dots, t_j^n) are linearly independent over Rn+1\mathbb{R}^{n+1}. (This follows from the fact that their determinant, the Vandermonde determinant, is nonzero.) Therefore, we must have w0t0n++wntnn0w_0 t_0^n + \dots + w_n t_n^n \neq 0. Also, the vectors (1,tj,,tjn1)(1, t_j, \dots, t_j^{n-1}), for j=1,,nj = 1, \dots, n, are independent. Thus we cannot have w0=0w_0 = 0. \square

Next we prove an essential result about the weights found above.

Lemma 3. Let reals t0,,tnt_0, \dots, t_n be given, and weights w0,,wnw_0, \dots, w_n be defined as above. Let P(t)P(t) be any polynomial, and define Q(t)=w0P(t+t0)++wnP(t+tn)Q(t) = w_0 P(t+t_0) + \dots + w_n P(t+t_n). If the degree of P(t)P(t) is less than nn, then Q(t)=0Q(t) = 0; otherwise, the degree of QQ is nn less than the degree of PP.

*Proof.* First we prove the result for P(t)=tsP(t) = t^s. The trt^r coefficient of QQ is s!r!(sr)!(w0t0sr++wntnsr)\frac{s!}{r!(s-r)!} (w_0 t_0^{s-r} + \dots + w_n t_n^{s-r}). This is zero if r>snr > s-n and nonzero if rsnr \le s-n, provided sns \ge n. If s<ns < n, then all coefficients of QQ are zero.

Now suppose PP is a general polynomial of degree ss. Then the tst^s term of PP is nonzero. Also, the tsnt^{s-n} term of QQ depends only on the tst^s term of PP, so it is also nonzero. Thus the degree of QQ is sns-n. \square

Finally, we solve the general form of IMO Problem 6. Suppose there are rr planes in the collection, so there exists a polynomial PP of degree rr such that P(xi,yj,zk)=0P(x_i, y_j, z_k) = 0 unless i=j=k=0i = j = k = 0. Apply Lemma 3 to PP (regarded as a polynomial in its first variable) with t0=x0,,ta=xat_0 = x_0, \dots, t_a = x_a, to obtain a polynomial P1P_1 of degree rar-a such that P1(0,yj,zk)=0P_1(0, y_j, z_k) = 0 unless j=k=0j = k = 0. Apply Lemma 3 again (this time on the second variable) to produce a polynomial P2P_2 of degree rabr-a-b such that P2(0,0,zk)=0P_2(0, 0, z_k) = 0 unless k=0k = 0. Apply Lemma 3 one more time (on the third variable) to produce a polynomial P3P_3 of degree rabcr-a-b-c such that P3(0,0,0)0P_3(0, 0, 0) \neq 0. Since P3P_3 is a nonzero polynomial, we have ra+b+cr \ge a+b+c. This completes the proof.

Source: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.