st

32 International Mathematical Olympiad

IMO 1991

1
 32st International Mathematical Olympiad
IMO 1991 Problems and Solutions

Problem A1
Given a triangle ABC, let I be the incenter. The internal bisectors of
angles A, B, C meet the opposite sides in A', B', C' respectively. Prove
that:
1/4 < AI·BI·CI/(AA'·BB'·CC') ≤ 8/27.
Solution
Consider the areas of the three triangles ABI, BCI, CAI. Taking base BC
we conclude that (area ABI + area CAI)/area ABC = AI/AA'. On the other
hand, if r is the radius of the in-circle, then area ABI = AB.r/2 and
similarly for the other two triangles. Hence AI/AA' = (CA + AB)/p, where
p is the perimeter. Similarly BI/BB' = (AB + BC)/p and CI/CC' = (BC +
CA)/p. But the arithmetic mean of (CA + AB)/p, (AB + BC)/p and (BC +
CA)/p is 2/3. Hence their product is at most (2/3)3 = 8/27.
Let AB + BC - CA = 2z, BC + CA - AB = 2x, CA + AB - BC = 2y. Then x, y,
z are all positive and we have AB = y + z, BC = z + x, CA = x + y.
Hence (AI/AA')(BI/BB')(CI/CC') = (1/2 + y/p)(1/2 + z/p)(1/2 + x/p) > 1/8
+ (x+y+z)/(4p) = 1/8 + 1/8 = 1/4.
Solutions are also available in István Reiman, International
Mathematical Olympiad 1959-1999, ISBN 189-8855-48-X.

32nd IMO 1991

© John Scholes
jscholes@kalva.demon.co.uk
2 Nov 1998

Problem A2
Let n > 6 be an integer and let a1, a2, ... , ak be all the positive integers
less than n and relatively prime to n. If a2 - a1 = a3 - a2 = ... = ak - ak-1 >
0, prove that n must be either a prime number or a power of 2.
Solution
by anon
If n is odd, then 1 and 2 are prime to n, so all integers < n are prime to
n, and hence is prime.
If n = 4k, then 2k-1 and 2k+1 are prime to n, so all odd integers < n are
prime to n, and hence n must be a power of 2.
If n = 4k+2, then 2k+1 divides n, but 2k+3 and 2k+5 are prime to n.
But if n > 6, then 2k+5 < n, so this cannot be a solution.
Solutions are also available in István Reiman, International
Mathematical Olympiad 1959-1999, ISBN 189-8855-48-X.

32nd IMO 1991

2
© John Scholes
jscholes@kalva.demon.co.uk
7 Sep 1999

Problem A3
Let S = {1, 2, 3, ... 280}. Find the smallest integer n such that each n-
element subset of S contains five numbers which are pairwise relatively
prime.
Solution
Answer: 217.
Let A be the subset of all multiples of 2, 3, 5 or 7. Then A has 216
members and every 5-subset has 2 members with a common factor. [To
show that |A| = 216, let an be the number of multiples of n in S. Then a2
= 140, a3 = 93, a5 = 56, a6 = 46, a10 = 28, a15 = 18, a30 = 9. Hence the
number of multiples of 2, 3 or 5 = a2 + a3 + a5 - a6 - a10 - a15 + a30 = 206.
There are ten additional multiples of 7: 7, 49, 77, 91, 119, 133, 161,
203, 217, 259.]
Let P be the set consisting of 1 and all the primes < 280. Define:
A1 = {2·41, 3·37, 5·31, 7·29, 11·23, 13·19}
A2 = {2·37, 3·31, 5·29, 7·23, 11·19, 13·17}
A3 = {2··31, 3·29, 5·23, 7·19, 11·17, 13·13}
B1 = {2·29, 3·23, 5·19, 7·17, 11·13}
B2 = {2·23, 3·19, 5·17, 7·13, 11·11}
Note that these 6 sets are disjoint subsets of S and the members of any
one set are relatively prime in pairs. But P has 60 members, the three
As have 6 each, and the two Bs have 5 each, a total of 88. So any
subset T of S with 217 elements must have at least 25 elements in
common with their union. But 6·4 = 24 < 25, so T must have at least 5
elements in common with one of them. Those 5 elements are the
required subset of elements relatively prime in pairs.
Solutions are also available in István Reiman, International
Mathematical Olympiad 1959-1999, ISBN 189-8855-48-X.

32nd IMO 1991

© John Scholes
jscholes@kalva.demon.co.uk
14 Sep 1999

Problem B1
Suppose G is a connected graph with k edges. Prove that it is possible to
label the edges 1, 2, ... , k in such a way that at each vertex which
belongs to two or more edges, the greatest common divisor of the
integers labeling those edges is 1.

3
[A graph is a set of points, called vertices, together with a set of edges
joining certain pairs of distinct vertices. Each pair of edges belongs to at
most one edge. The graph is connected if for each pair of distinct
vertices x, y there is some sequence of vertices x = v0, v1, ... , vm = y,
such that each pair vi, vi+1 (0 ≤ i < m) is joined by an edge.]
Solution
The basic idea is that consecutive numbers are relatively prime.
We construct a labeling as follows. Pick any vertex A and take a path
from A along unlabeled edges. Label the edges consecutively 1, 2, 3, ...
as the path is constructed. Continue the path until it reaches a vertex
with no unlabeled edges. Let B be the endpoint of the path. A is now
guaranteed to have the gcd (= greatest common divisor) of its edges 1,
because one of its edges is labeled 1. All the vertices between A and B
are guaranteed to have gcd 1 because they have at least one pair of
edges with consecutive numbers. Finally, either B has only one edge, in
which case its gcd does not matter, or it is also one of the vertices
between A and B, in which case its gcd is 1.
Now take any vertex C with an unlabeled edge and repeat the process.
The same argument shows that all the new vertices on the new path
have gcd 1. The endpoint is fine, because either it has only one edge (in
which case its gcd does not matter) or it has already got gcd 1.
Repeat until all the edges are labeled.
Solutions are also available in István Reiman, International
Mathematical Olympiad 1959-1999, ISBN 189-8855-48-X.

32nd IMO 1991

© John Scholes
jscholes@kalva.demon.co.uk
19 Nov 1998

Problem B2
Let ABC be a triangle and X an interior point of ABC. Show that at least
one of the angles XAB, XBC, XCA is less than or equal to 30o.
Solution
By Marcin Mazur, University of Illinois at Urbana-Champaign
Let P, Q, R be the feet of the perpendiculars from X to BC, CA, AB
respectively. Use A, B, C to denote the interior angles of the triangle
(BAC, CBA, ACB). We have PX = BX sin XBC = CX sin(C - XCA), QX = CX
sin XCA = AX sin(A - XAB), RX = AX sin XAB = BX sin(B - XBC).
Multiplying: sin(A - XAB) sin(B - XBC) sin(C - XCA) = sin A sin B sin C.
Now observe that sin(A - x)/sin x = sin A cot x - cos A is a strictly
decreasing function of x (over the range 0 to π), so if XAB, XBC and XCA
are all greater than 30, then sin(A - 30) sin(B - 30o) sin(C - 30o) > sin330o
= 1/8.
But sin(A - 30o) sin(B - 30o) = (cos(A - B) - cos(A + B - 60o))/2 ≤ (1 -
cos(A + B - 60o))/2 = (1 - sin(C - 30o))/2, since (A - 30o) + (B - 30o) + (C -
30o) = 90o. Hence sin(A - 30o) sin(B - 30o) sin(C - 30o) ≤ 1/2 (1 - sin(C -
4
30o)) sin(C - 30o) = 1/2 (1/4 - (sin(C - 30o) - 1/2)2) ≤ 1/8. So XAB, XBC,
XCA cannot all be greater than 30o.
By Jean-Pierre Ehrmann
P, Q, R as above. Area ABX + area BCX + area CAX = area ABC, so
AB·XR + BC·XP + CA·XQ = 2 area ABC ≤ BC·AP ≤ BC(AX + XP). Hence
AB·XR/AX + CA·XQ/AX ≤ BC.
Squaring and using (λ + μ)2 ≥ 4 λμ, we have: BC2 ≥ 4 AB·CA. XR·XQ/AX2.
Similarly: CA2 ≥ 4 BC·AB·XP·XR/BX2, and AB2 ≥ 4 AB·BC·XQ·XP/CX2.
Multiplying these three inequalities together gives: 1 ≥ 64
(XR/AX)2(XP/BX)2(XQ/CX)2, and hence: (XR/AX) (XP/BX) (XQ/CX) ≤ 1/8, or
sin XAB sin XBC sin XCA ≤ 1/8. So not all XAB, XBC, XCA are greater
than 30o.
Gerard Gjonej noted that the result follows almost immediately from the
Erdos-Mordell inequality: XA + XB + XC ≥ 2(XP + XQ + XR). [For if all
the angles are greater than 30, then XR/XA, XP/XB, XQ/XC are all
greater than sin 30o = 1/2.]. This result was notoriously hard to prove -
Erdos hawked it around a large number of mathematicians before
Mordell found a proof - but the proof now appears fairly innocuous, at
least if you do not have to rediscover it:
Let R1, Q1 be the feet of the perpendiculars from P to AB, CA
respectively. Similarly, let P2, R2 be the feet of the perpendiculars from Q
to BC, AB, and Q3, P3 the feet of the perpendiculars from R to CA, BC.
Then P2P3 is the projection of QR onto BC, so P2P3/QR ≤ 1. Similarly,
Q3Q1/RP ≤ 1, and R1R2/PQ ≤ 1. Hence XA + XB + XC ≥ XA.P2P3/QR +
XB·Q3Q1/RP + XC·R1R2/PQ (*)
Now BPXR is cyclic, because BPX and XRB are both right angles. Hence
angle BXR = angle BPR = angle RPP3, so triangles XBR and PRP3 are
similar. Hence PP3 = PR.XR/XB.
Similarly, QQ1 = QP·XP/XC, RR2 = RQ·XQ/XA, and PP2 = PQ·XQ/XC, QQ3 =
QR·XR/XA, RR1 = RP·XP/XB. Substituting into (*), we obtain:
XA + XB + XC ≥ XA( PQ/QR XQ/XC + PR/QR XR/XB ) + XB( QR/RP XR/XA
+ QP/RP XP/XC ) + XC( RP/PQ XP/XB + RQ/PQ XQ/XA ).
On the right hand side, the terms involving XP are: XP( QP/RP XB/XC +
RP/PQ XC/XB ), which has the form XP (x + 1/x) and hence is at least 2
XP. Similarly for the terms involving XQ and XR.
Solutions are also available in István Reiman, International
Mathematical Olympiad 1959-1999, ISBN 189-8855-48-X.
32nd IMO 1991
© John Scholes
jscholes@kalva.demon.co.uk
7 Sep 1999
Problem B3
Given any real number a > 1 construct a bounded infinite sequence x0,
x1, x2, ... such that |xn - xm| |n - m|a ≥ 1 for every pair of distinct n, m.
[An infinite sequence x0, x1, x2, ... of real numbers is bounded if there is
a constant C such that |xn| < C for all n.]
Solution
5
By Marcin Mazur, University of Illinois at Urbana-Champaign
Let t = 1/2a. Define c = 1 - t/(1 - t). Since a > 1, c > 0. Now given any
integer n > 0, take the binary expansion n = ∑i bi 2i, and define xn = 1/c
∑bi>0 ti. For example, taking n = 21 = 24 + 22 + 20, we have x21 = (t4 + t2
+ t0)/c. We show that for any unequal n, m, |xn - xm| |n - m|a ≥ 1. This
solves the problem, since the xn are all positive and bounded by (∑ tn )/c
= 1/(1 - 2t).
Take k to be the highest power of 2 dividing both n and m. Then |n - m|
≥ 2k. Also, in the binary expansions for n and m, the coefficients of 20,
21, ... , 2k-1 agree, but the coefficients for 2k are different. Hence c |xn -
xm| = tk + ∑i>k yi, where yi = 0, ti or - ti. Certainly ∑i>k yi > - ∑i>k ti = tk+1/(1
- t), so c |xn - xm| > tk(1 - t/(1 - t)) = c tk. Hence |xn - xm| |n - m|a > tk 2ak =
1.
Solutions are also available in István Reiman, International
Mathematical Olympiad 1959-1999, ISBN 189-8855-48-X.

32nd IMO 1991

© John Scholes
jscholes@kalva.demon.co.uk
14 Sep 1999