A1. Let En = (a1 - a2)(a1 - a3) ... (a1 - an) + (a2 - a1)(a2 - a3) ... (a2 - an) + ... + (an - a1)(an - a2) ... (an - an-1). Let Sn be the proposition that En ≥ 0 for all real ai. Prove that Sn is true for n = 3 and 5, but for no other n > 2.
A2. Let P1 be a convex polyhedron with vertices A1, A2, ... , A9. Let Pi be the polyhedron obtained from P1 by a translation that moves A1 to Ai. Prove that at least two of the polyhedra P1, P2, ... , P9 have an interior point in common.
A3. Prove that we can find an infinite set of positive integers of the form 2n - 3 (where n is a positive integer) every pair of which are relatively prime.
B1. All faces of the tetrahedron ABCD are acute-angled. Take a point X in the interior of the segment AB, and similarly Y in BC, Z in CD and T in AD. (a) If angle DAB + angle BCD ≠ angle CDA + angle ABC, then prove that none of the closed paths XYZTX has minimal length;
(b) If angle DAB + angle BCD =angle CDA + angle ABC, then there are infinitely many shortest paths XYZTX, each with length 2 AC sin k, where 2k = angle BAC + angle CAD + angle DAB.
B2. Prove that for every positive integer m we can find a finite set S of points in the plane, such that given any point A of S, there are exactly m points in S at unit distance from A.
B3. Let A = (aij), where i, j = 1, 2, ... , n, be a square matrix with all aij non-negative integers. For each i, j such that aij = 0, the sum of the elements in the ith row and the jth column is at least n. Prove that the sum of all the elements in the matrix is at least n2/2.
Solutions
Problem A1
Let En = (a1 - a2)(a1 - a3) ... (a1 - an) + (a2 - a1)(a2 - a3) ... (a2 - an) + ... + (an - a1)(an - a2) ... (an - an-1). Let Sn be the proposition that En ≥ 0 for all real ai.
Prove that Sn is true for n = 3 and 5, but for no other n > 2.
Solution
Take a1 < 0, and the remaining ai = 0. Then En = a1n-1 < 0 for n even, so the proposition is false for even n.
Suppose n ≥ 7 and odd. Take any c > a > b, and let a1 = a, a2 = a3 = a4= b, and a5 = a6 = ... = an = c. Then En = (a - b)3(a - c)n-4 < 0. So the proposition is false for odd n ≥ 7.
Assume a1 ≥ a2 ≥ a3. Then in E3 the sum of the first two terms is non-negative, because (a1 - a3) ≥ (a2 - a3). The last term is also non-negative. Hence E3 ≥ 0, and the proposition is true for n = 3.
It remains to prove S5. Suppose a1 ≥ a2 ≥ a3 ≥ a4 ≥ a5. Then the sum of the first two terms in E5 is (a1 - a2){(a1 - a3)(a1 - a4)(a1 - a5) - (a2 - a3)(a2 - a4)(a2 - a5)} ≥ 0. The third term is non-negative (the first two factors are non-positive and the last two non-negative). The sum of the last two terms is: (a4 - a5){(a1 - a5)(a2 - a5)(a3 - a5) - (a1 - a4)(a2 - a4)(a3 - a4)} ≥ 0. Hence E5 ≥ 0.
Problem A2
Let P1 be a convex polyhedron with vertices A1, A2, ... , A9. Let Pi be the polyhedron obtained from P1 by a translation that moves A1 to Ai. Prove that at least two of the polyhedra P1, P2, ... , P9 have an interior point in common.
Solution
The result is false for 8 vertices - for example, the cube. We get 8 cubes, with only faces in common, forming a cube 8 times as large.
This suggests a trick. Each Pi is contained in D, the polyhedron formed from P1 by doubling the scale. Take A1 as the origin and take the vertex Bi to have twice the coordinates of Ai. Given a point X inside P1, the midpoint of PiX must lie in P1 by convexity. Hence the point with doubled coordinates, which is obtained by adding the coordinates of Ai to the coordinates of X, lies in D. In other words every point of Pi lies in D. But the volume of D is 8 times the volume of P1, which is less than the sum of the volumes of P1, ... , P9.
Problem A3
Prove that we can find an infinite set of positive integers of the form 2n - 3 (where n is a positive integer) every pair of which are relatively prime.
Solution
We show how to enlarge a set of r such integers to a set of r+1. So suppose 2n1 - 3, ... , 2nr - 3 are all relatively prime. The idea is to find 2n - 1 divisible by m = (2n1 - 3) ... (2nr - 3), because then 2n - 3 must be relatively prime to all of the factors of m. At least two of 20, 21, ... , 2m must be congruent mod m. So suppose m1 > m2 and 2m1 = 2m2 (mod m), then we must have 2m1 - m2 - 1 = 0 (mod m), since m is odd. So we may take nr+1 to be m1 - m2.
Problem B1
All faces of the tetrahedron ABCD are acute-angled. Take a point X in the interior of the segment AB, and similarly Y in BC, Z in CD and T in AD.
(a) If ∠DAB + ∠BCD ≠ ∠CDA + ∠ABC, then prove that none of the closed paths XYZTX has minimal length;
(b) If ∠DAB + ∠BCD = ∠CDA + ∠ABC, then there are infinitely many shortest paths XYZTX, each with length 2 AC sin k, where 2k = ∠BAC + ∠CAD + ∠DAB.
Solution
The key is to pretend the tetrahedron is made of cardboard, cut it along three edges and unfold it. Suppose we do this to get the hexagon CAC'BDB'. Now the path is a line joining Y on B'C to Y' on the opposite side BC' of the hexagon. Clearly this line must be straight for a minimal path. If B'C and BC' are parallel, then we can take Y anywhere on the side and the minimal path length is the expression given.
But if they are not parallel, then the minimal path will come from an extreme position. Suppose CC' < BB'. If the interior angle CAC' is less than 180o, then the minimal path is obtained by taking Y at C. But this does not meet the requirement that Y be an interior point of the edge, so there is no minimal path in the permitted set. If the interior angle CAC' is greater than 180, then the minimal path is obtained by taking X and T at A. Again this is not permitted.
The problem therefore reduces to finding the condition for B'C and BC' to be parallel. This is evidently angles BCD + DCA + CAD + BAD + BAC + ACB = 360o. But DCA + CAD = 180o - ADC, and BAC + ACB = 180o - ABC, so we obtain the condition given.
Problem B2
Prove that for every positive integer m we can find a finite set S of points in the plane, such that given any point A of S, there are exactly m points in S at unit distance from A.
Solution
Take a1, a2, ... , am to be points a distance 1/2 from the origin O. Form the set of 2m points ±a1 ±a2 ± ... ±am. Given such a point, it is at unit distance from the m points with just one coefficient different. So we are home, provided that we can choose the ai to avoid any other pairs of points being at unit distance, and to avoid any degeneracy (where some of the 2m points coincide).
The distance between two points in the set is |c1a1 + c2a2 + ... + cmam|, where ci = 0, 2 or -2. So let us choose the ai inductively. Suppose we have already chosen up to m. The constraints on am+1 are that we do not have |c1a1 + c2a2 + ... + cmam + 2am+1| equal to 0 or 1 for any ci = 0, 2 or -2, apart from the trivial cases of all ci = 0. Each | | = 0 rules out a single point and each | | = 1 rules out a circle which intersects the circle radius 1/2 about the origin at 2 points and hence rules out two points. So the effect of the constraints is to rule out a finite number of points, whereas we have uncountably many to choose from.
Problem B3
Let A = (aij), where i, j = 1, 2, ... , n, be a square matrix with all aij non-negative integers. For each i, j such that aij = 0, the sum of the elements in the ith row and the jth column is at least n. Prove that the sum of all the elements in the matrix is at least n2/2.
Solution
Let x be the smallest row or column sum. If x >= n/2, then we are done, so assume x < n/2. Suppose it is a row. (If not, interchange rows and columns.) The number of non-zero elements in the row, y, must also satisfy y < n/2, since each non-zero element is at least 1. Now move across this row summing the columns. The y columns with a non-zero element have sum at least x (by the definition of x). The n - y columns with a zero have sum at least n - x. Hence the total sum is at least xy + (n - x)(n - y) = n2/2 + (n - 2x)(n - 2y)/2 > n2/2.
The result is evidently best possible, because we can fill the matrix alternately with zeros and ones (so that aij = 1 if i and j are both odd or both even, 0 otherwise). For n even, every row and column has n/2 1s, so the condition is certainly satisfied and the total sum is n2/2. For n odd, odd numbered rows have (n+1)/2 1s and even numbered one less. But the only zeros are in positions which have either the row or the column odd-numbered, so the sum in such cases is n as required. The total sum is n2/2 + 1/2. Alternatively, for n even, we could place n/2 down the main diagonal.
Labels:
IMO