51st International Mathematical Olympiad Astana, Kazakhstan 2010

Shortlisted Problems with Solutions

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem C2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem N4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem C7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5 7 7 8 10 12 13 15 17 19 23 23 26 28 30 30 32 35 38 44 44 46 50 52 54 56 56 60 64 64 64 66 68 70 71 72 Combinatorics Problem C1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Geometry Problem Problem Problem Problem Problem Problem Problem Problem G1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem C3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem N5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem C4½ Problem C5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem C6 . . . . . . G6½ G7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Number Theory Problem N1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem N6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Contents Note of Confidentiality Contributing Countries & Problem Selection Committee Algebra Problem Problem Problem Problem Problem Problem Problem Problem A1 A2 A3 A4 A5 A6 A7 A8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem C4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem N3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G5 . . . . . . . . . . . Problem N1½ Problem N2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G3 . . . . . . . . .

.

Note of Confidentiality The Shortlisted Problems should be kept strictly confidential until IMO 2011. Panama. Croatia. Saudi Arabia. Slovenia. Luxembourg. Bulgaria. Hungary. Slovakia. Russia. Hong Kong. Canada. Ireland. Turkey. Georgia. France. India. Switzerland. Korea (North). Poland. Korea (South). Ukraine. Cyprus. Columbia. Thailand. Uzbekistan Problem Selection Committee Yerzhan Baissalov Ilya Bogdanov G´ eza K´ os Nairi Sedrakyan Damir Yeliussizov Kuat Yessenov . Austria. Estonia. Iran. Germany. Contributing Countries The Organizing Committee and the Problem Selection Committee of IMO 2010 thank the following 42 countries for contributing 158 problem proposals. Armenia. Romania. Serbia. Australia. Indonesia. Finland. United Kingdom. Pakistan. Netherlands. Greece. United States of America. Mongolia. Japan.

.

Subcase 2a. (3) f Ôx2 Õ whenever Öx1 × f ÔaÕ f 2a ¤   1 2 Öx2 ×. Finally. or f ÔxÕ f ÔÖx×aÕ Öf ÔaÕ× . Assume that f Ô0Õ 0. Solution 1. Assume that Öf Ôy Õ× 0 for some y . Finally. then the substitution x 1 provides f Ôy Õ f Ô1ÕÖf Ôy Õ× 0. So we are left to consider the case when Öf ÔaÕ× 0 for some a. thus contradicting the condition f ÔαÕ 0. if Öf Ôy Õ× 0 for all y . f ÔxÕ R such that the equality f ÔÖx×y Õ f ÔxÕÖf Ôy Õ×. y Answer. substituting x 1 in (1) provides f Ôy Õ 0 for all y È R. Then setting x α in (1) we obtain 0 f Ô0Õ f ÔαÕÖf Ôy Õ× for all y È R. Finally. hence f ÔxÕ  1¨ 2 f ÔÖx×Õ. from (1) we get f Ôz Õ f ÔÖN ×αÕ f ÔN ÕÖf ÔαÕ× 0 for all z È R. then f Ôy Õ 0 for all y . Case 2. 0 for all 0 α 1. Now we have f Ô0Õ 0. where C 0 or 1 f Ô0 Õ C 2. by Öx× we denote the greatest integer not exceeding x. Here. 0. we have f ÔαÕ z exists an integer N such that α È Ö0. Then from (2) we conclude that Öf Ôy Õ× 1 for all y È R. Now. Solution 2. . two cases are possible. 1Õ (one may set N Öz×  1 if z 0 and N Özס 1 N otherwise). so we may even assume that a f ÔxÕ C for each x. Öf Ôy Õ× 0 for all y È R. Case 1. First. equation (1) becomes f ÔÖx×y Õ f ÔxÕ. condition (1) becomes equivalent to the equation C exactly when ÖC × 1. (1) (France) const C . Now we have this implies Öf Ô0Õ× f ÔxÕÖf ÔaÕ×. Hence. from Öf Ôy Õ× 1 ÖC × we obtain that 1 C 2. Determine all functions f : R holds for all x. This function obviously satisfies the problem conditions. and substituting y 0 we have f ÔxÕ f Ô0Õ C 0. Here we consider two subcases. Now.Algebra A1. È R. Hence. there Subcase 2b. setting x 0 in (1) we get f Ô0ÕÖf Ôy Õ× (2) for all y È R. Then we have f ÔÖx×aÕ This means that f Ôx1 Õ that a is an integer. Therefore equation (3) provides 0 C ÖC × which holds 0. Suppose that there exists 0 α 1 such that f ÔαÕ 0. a straightforward check shows that all the obtained functions satisfy (1). Now. Conversely. and we may assume ¨ f Ô2aÕ f f Ô0Õ Öf Ô0Õ× f Ô2aÕÖf Ô0Õ×. Therefore. Consider any real z .

(Ukraine) Solution 1. where q is a nonnegative number. 1Õ and Ô0. 1. 4 (1) under the constraint Ôa2   b2   c2   d2Õ ¡ 2Ôa   b   c   dÕ   4 (we will not use the value of x   y   z   t though it can be found). 3. respectively. c. a2   b2   c2 a2   b2   c2 rewrites as 12 ¡ d2 Ôa   b   cÕ2 3 2dÔd ¡ 3Õ 0. The latter estimate. 12. Comment 2. it can be performed by moving the variables – since we need only the second of the two shifted conditions. for instance. b. introducing x ¡ Ôa ¡ 1Õ4   Ôb ¡ 1Õ4   Ôc ¡ 1Õ4   Ôd ¡ 1Õ4   52. Ôx2   y2   z2   t2Õ2 16. 2. t d ¡ 1. expanding the brackets we note that Ôx2   y2   z2   t2Õ2 Ôx4   y4   z4   t4 Õ   q. one can finish the solution in several different ways. we have 12 ¡ d2 . Ô6 ¡ dÕ2 3 . Let the real numbers a. Solution 2. we need to prove the inequalities 16 x4   y 4   z 4   t4 4. d satisfy the relations a   b   c   d 36 4Ôa3   b3   c3   d3 Õ ¡ Ôa4   b4   c4   d4 Õ 6 and a2   b2   c2   d2 48. For the other one. Comment 1. y b ¡ 1. a ¡ 1. the lower and upper bounds are attained at Ô3. b. d 6 ¡ d. z c ¡ 1. x 4 x2   y 2   z 2   t2 Now the rightmost inequality in (1) follows from the power mean inequality:  y  z  t 4 4 4 Ôx2   y2   z2   t2 Õ2 4 4. we claim that 0 a b c hence the power mean inequality a. so x4   y 4   z 4   t4 and we are done. Actually. 2. After the change of variables. First. The estimates are sharp.8 Prove that A2. 1. c. 2Õ. Observe that 4Ôa3   b3   c3   d3 Õ ¡ Ôa4   b4   c4   d4 Õ ¡ Ôa ¡ 1Õ4   Ôb ¡ 1Õ4   Ôc ¡ 1Õ4   Ôd ¡ 1Õ4   6Ôa2   b2   c2   d2Õ ¡ 4Ôa   b   c   dÕ   4   ¨   ¨ Now.

This implies that Ô4a3 ¡ a4 Õ Ô4b3 ¡ b4 Õ Ô4c3 ¡ c4 Õ Ô4d3 ¡ d4Õ as desired. 3× as in the previous solution.9 which implies the desired inequalities for d. we have d2 Ôa   b2   c2   d2Õ 3 (which implies d 1). we show that 4b ¡ b2 4c ¡ c2 . In particular. First. 3×. 2. this inequality rewrites as 4x3 ¡ x4 4x2 . these polynomials should have the form x4 ¡ 4x3   ax2   bx   c. to prove the rightmost inequality. For each x È Ö0. Now we prove the leftmost inequality in an analogous way. we can assume that 0 a b c d. we have Ôx   1ÕÔx ¡ 1Õ2Ôx ¡ 3Õ 0 which is equivalent to 4x3 ¡ x4 2x2   4x ¡ 3. Solutions 2 and 3 show that only the condition a2   b2   c2   d2 12 is needed for the former one. To prove the leftmost inequality. as desired. 2Õ and Ô3. we first show that a. while the latter should have roots at 1 (with an even multiplicity) and 3. 1Õ for this inequality. 2. 4 Ôa¡   4a2 Õ   Ôb4   4b2 Õ   Ôc4   4c2Õ   Ôd4   4d2© Õ 2 a4 ¤ 4a2   b4 ¤ 4b2   c4 ¤ 4c2   d4 ¤ 4d2 4Ô a3   b3   c3   d3 Õ 4Ôa3   b3   c3   d3 Õ. we use the obvious inequality x2 Ôx ¡ 2Õ2 0 for each real x. this inequality rewrites as Ôc ¡ bÕÔb   c ¡ 4Õ 0. Actually. moreover. the inequalities a b c together with 4a ¡ a2 apply the Chebyshev inequality obtaining   ¨ 1 2 a2 Ô4a ¡ a2 Õ   b2 Ô4b ¡ b2 Õ   c2 Ô4c ¡ c2 Õ Ô a   b2   c2 Õ 4Ôa   b   cÕ ¡ Ôa2   b2   c2 Õ 3 Ô12 ¡ d2 ÕÔ4Ô6 ¡ dÕ ¡ Ô12 ¡ d2ÕÕ . since the conditions are symmetric. (2) 3 3 1 2 Finally. we also have the same estimate for the other variables. These conditions determine the polynomials uniquely. the expression 4 4 Ô3 ¡ dÕÔd ¡ 1ÕÔd2 ¡ 3Õ in the right-hand part of (2) is nonnegative. Comment. Then we have a   b b   c 2 Ôb   c   d Õ 2 ¤ 6 4. It is easy to guess the extremal points Ô0. This provides a method of finding the polynomials used in Solution 2. Solution 3. 4b ¡ b2 4c ¡ c2 allow us to Further. 3 3 Next. Now. c. Moreover. and the desired inequality 3 is proved. the former polynomial should have roots at 2 (with an even multiplicity) and 0. so. which establishes the rightmost inequality. d È Ö0. Namely. 3 This implies that 2 2 Ô4a3 ¡ a4 Õ   Ô4b3 ¡ b4 Õ   Ô4c3 ¡ c4Õ   Ô4d3 ¡ d4Õ Ô12 ¡ d ÕÔd3 ¡ 4d   12Õ   4d3 ¡ d4 144 ¡ 48d   16d3 ¡ 4d4 4 36   Ô3 ¡ dÕÔd ¡ 1ÕÔd2 ¡ 3Õ. It follows that Ô4a3 ¡ a4 Õ   Ô4b3 ¡ b4 Õ   Ô4c3 ¡ c4Õ   Ô4d3 ¡ d4Õ 4Ôa2   b2   c2   d2 Õ 48. expanding 48 we have a4   b4   c4   d4   48 4Ôa2   b2   c2   d2 Õ and applying the AM–GM inequality. 1. b. The inequality 4a ¡ a2 4b ¡ b2 can be proved analogously. which follows from the previous estimate. 2Ôa2   b2   c2   d2 Õ  4Ôa   b   c   dÕ¡ 12 36. . Comment. The rightmost inequality is easier than the leftmost one. 1.

. Namely. applying induction). . i 1. . x100 be nonnegative real numbers such that xi i 1. Hence by the AM–GM inequality we get x2i¡1 x2i 1   x2i x2i 2 1 ¡ x2i   ¨2 25 . . 4 Ôx2i   x2i 1 Ô1 ¡ x2i ¡ x2i 1 Õx2i 1   x2i Ô1 ¡ x2i ¡ x2i 1 Õ ¢ ª2 Ô x2i   x2i 1 Õ   Ô1 ¡ x2i ¡ x2i 1 Õ ÕÔ1 ¡ x ¡ x Õ 2i 2i 1   2 Summing up these inequalities for i 50 ô 1. . ¡ x2i 1 and 1 . Let x2i 0. Let x1 . Solution 2. . . . . . The original estimate can be obtained from this version by considering the sequence x1 . x2n in a row (with no cyclic notation) and suppose that xi   xi 1   xi 2 1 for all 2ô n¡2 n¡1 xi xi 2 .10 A3. ¨2 «2 1 Ôxi   xi 1Õ 2 100 xi £ 100 ô i 1 . we have S 100 ô i 1 xi ¡ 1 200 £ Ôxi   xi 1 Õ2 «2 100 ô i 1 1 100   1 2i Ôxi   xi 1 Õ2. x102 S 100 ô i 1   xi 1   xi 2 1 for all x2 ). though if might be easier to find it (for instance. Then 4 i 1 The proof is the same as above. 2 Ôx2i¡1 x2i 1   x2ix2i 2 Õ i 1 Comment. . We present another proof of the estimate. consider 2n nonnegative numbers x1 . . Then we have S 50 ¤ 1 2 2 we are left to show that S 25 for all values of x ’s satisfying the problem conditions. x1 . 2n ¡ 2. . x100 . 50. by the AM–GM inequality S 2 100 £ £ 100 ô i 1 100 ô i 1 100 ô «£ i 1 100 2 ¡ 100 ô i 1 « xi . 2 . so 100 ô Ôxi   xi 1Õ xi ¡ £ xi i 1 2 100 And finally. 2. . By the problem condition. i 2 i 50. we get S 100 ô i 1 xi xi 2 100 ô i 1 xi Ô1 ¡ xi ¡ xi 1 Õ 100 ô i 1 100 ô i 1 xi ¡ xi ¡ 100 ô x2 i ¡ 100 ô i 1 xi xi 1 i 1 100 ô By the AM–QM inequality. x2 . x2i¡1 1 for all i 1. . 50. 100 (we put x101 x1 . 2 So. (Russia) Answer. 25 . . Find the maximal possible value of the sum xi xi 2 . . we get the desired inequality 50 ¤ 1 4 25 . From the problem condition. . ¤ 1 2 xi   100 2 ¡ ««2 xi 2 100 ¤ ¢ 100 4 ª2 25 . 2 Solution 1. . . we get x2i¡1 Consider any 1 x2i 2 1 ¡ x2i ¡ x2i 1 . . . . . . x2 . This solution shows that a bit more general fact holds. . . . 2.

the estimates must be done with care to preserve equality in the optimal configurations. Although it is easy to find the value 25 2 . . There are two different optimal configurations in which the variables have different values. and not all of sums of three consecutive numbers equal 1. These solutions are not as easy as they may seem at the first sight.11 Comment.

Using the inductive definition of set M . The proof is complete. then S4k 2Sk 0. If n 4m then we have xn 1 x2m 1 xm 1 since m is even. The step is proved. 1 if n È M . We prove that xn 1 xm 1 1. Hence. denoting Sn S4k x2k¡1 n i 1 k ô   i 1 ¡x4k¡2 and x4k¡1 x4k ¡x2k Ô0   2xk Õ xk . hence S4k 1 S4k   x4k 1 2Sk   xk 1 Sk   Sk 1 0. S4k 2 S4k 0. we obtain S4k 4 2Sk 1 0. The base Now. n 4m   2. First. Prove that x1   x2   ¤ ¤ ¤   xn 1 and x2k 0 for all n 1. (ii) 2 È M . if k is even. x2 ¡1. (i) holds since Sn n Ômod 2Õ. Assume the contrary and consider the minimal n such that Sn 1 0.12 A4. from the definition of xi it follows that for each positive integer k we have x4k¡3 Hence. M consists exactly of all positive integers not containing digits 1 and 3 in their 4-base representation. our aim is to prove that xn 1 1 whenever n È M thus coming to a contradiction. otherwise. one can describe it explicitly. x2k¡1 Ô¡1Õk 1xk for all (Austria) Solution 1. We use the induction on n. assume that Si 0 for all i 4k . since k is odd. i 1 xi i 11 Now we prove by induction on k that Si 0 for all i 4k . xn 1 ¡1. is defined by x1 k 1. we first describe the set M inductively. (ii) is straightforward. Using the relations (1)–(3). so we have S4k 2 and hence S4k 1 S4k   x4k 1 1. . Conversely. We start with some observations. we are left to prove that S4k 1 0. and (iii) for every even n 4 we have n È M Önß4× È M . Comment. and from Sn 0 we get Sn 0. surely n 1. ¡xk . we have ¨ k ô Ôx4k¡3   x4k¡2 Õ   Ôx4k¡1   x4k Õ S4k 2 2 Sk . Sk is odd as well. consider some 4 n È M and let m Önß4× È M . If k is odd. . as desired. . For the induction step. and xn 1 ¡x2m 2 xm 1 . Actually. here we have x3 1. A sequence x1 . (2) (3) S4k   Ôx4k 1   x4k 2 Õ i 1 S4k . while (iii) follows from the relations S4k 2 S4k 2Sk . (1) xi . then m is even. For the induction step. then we have x4k 1 x2k 1 xk 1 . . For this purpose. S4k 3 S4k 2   x4k 3 S4k 2   S4k 4 2 0. we are especially interested in the set M Øn : Sn 0Ù. x2 . So. as desired. Solution 2. We claim that (i) M consists only of even numbers. We will use the notation of Sn and the relations (1)–(3) from the previous solution. and xm 1 1 by the induction hypothesis. n n Observe also that Sn n Ômod 2Õ. the minimal element of M . Namely. The base case is valid since x1 x3 x4 1. we are left to prove that xn 1 case is n 2. that is.

. αk 1 and let the prime factorization of g ÔxÕ be g ÔxÕ pα 1 . by induction. we have g g ÔxÕ   ¨ g xf ÔxÕ xf ÔxÕ2   ¨ x3 f ÔxÕ xf ÔxÕ ¤ f xf ÔxÕ     xf ÔxÕ   ¨5ß2 g ÔxÕ ¨5ß2 . . f ÔxÕf Ôy Õ. ÔxyÕ3f ÔxyÕ f Ôxy Õ2 f Ôxy Õ y 3f f ÔxÕ2 y   ¨ f f ÔxÕ2 f Ôy Õ2 . Determine all functions f : Q  which satisfy the following equation for all x. g ÔxÕ . we get f f ÔxÕ2 f Ôy Õ. . . . . y Õ: f f Ôxy Õ2   ¨ y . x3 f ÔxÕ. . . . Q  (1) (Switzerland) Answer. ÐÓÓÓÓÑÓÓÓÓÒ n 1 g g . . αk are nonzero integers. The left-hand side is always rational. x x3 f ÔxÕ. we have x3 which implies x f f ÔxÕ2 f ÔxÕ     ¨ 1 . 1 and f Ôxn Õ f ÔxÕn for all integers n. whenever f ÔxÕ 1. f f Ôy Õ2 f Ôy Õ   ¨ (2) ¨ y3   ¨ Now replace x by xy in (2). . . By substituting y Then. second time to y. This also implies f Ô1Õ Then the function equation (1) can be re-written as f f ÔxÕ f Ôy Õ f f ÔxÕ   ¨   ¨2 x3 f ÔxÕf Ôy Õ. . The only such function is f ÔxÕ Solution. pk where p1 .   ¨ Since f is injective. . f is multiplicative. We show that this is possible only if g ÔxÕ 1. ÐÓÓÓÓÑÓÓÓÓÒ n 1 ¡   ¨©   g ÔxÕ ¨Ô5ß2Õn p1 Ô5ß2Õn α1 ¤ ¤ ¤ pÔ5ß2Õn αk k   . (4) and. Therefore. ¨ (3) xf ÔxÕ2 f f ÔxÕ     ¨ Let g ÔxÕ xf ÔxÕ. . y f f ÔxÕ2 y   ¨ È Q : x3 f Ôxy Õ. f ÔxÕ2 instead of Ôx. so the function f is injective. Denote by Q  the set of all positive rational numbers. and apply (1) twice. . so g ÔxÕ rational for every n. Suppose that g ÔxÕ 1. g ÔxÕ . pk are distinct primes and α1 . . Then. Then the unique prime factorization of (4) is g g . by (3).   ¨Ô5ß2Õn must be Consider (4) for a fixed x.13 A5. we get f ÔxÕ2 f Ôy Õ2 . . ¡   ¨© g ÔxÕ ¨Ô5ß2Õn   for every positive integer n.

Using transfinite tools. Among R  1 f ÔxÕ2 y   1 ¨2 x 1 y x3 xy x3 f Ôxy Õ.14 where the exponents should be integers. Therefore. gÔxÕ 1 is impossible. 2 1 for all x. Another solution is x x3ß2 . f ÔxÕ . g ÔxÕ 1 and thus f ÔxÕ x 1 The function f ÔxÕ satisfies the equation (1): x f Ôf ÔxÕ2 y Õ Comment. infinitely many other solutions can be constructed. Hence. for example Ô5 Õnα1 cannot be a integer number when 2n § α1. f1 ÔxÕ 1 is not the only solution. But§ this is not true for large values of n. R  functions.

a   1. we say that x and y are similar (and write x y ) if f ÔxÕ f Ôy Õ (equivalently. . The second statement follows now from the sets of values of f and g . the statement follows and Claim ¨ (1)   ¨ 3. . Prove that f ÔnÕ g ÔnÕ for all positive integer n. Now. . . as well as exactly one class contained in Ø1. . . . while g attains all the values from the set Ng Øb. surely. so the function f attains all values from the set Nf Øa. then Öng × is contained in Ø1. Suppose that f ÔxÕ f Ôy Õ. If a a½ . Proof. Öx× Ø1. Conversely. b ¡ 1Ù as well. (1) f ÔxÕ a k Claim 3. . so Öa× should contain some element a½ b by Claim 2 again. f ÔxÕ f Ôy Õ. a a½ b. as well as at most one element from Ng . Ð Now we are ready to prove the problem statement. . First. and ¨   similarly g f k ÔxÕ g ÔxÕ   k . we get that g ÔxÕ b x nf ng . Claim 1. Proof. the proof of the second statement is similar. If f ÔxÕ f Ôy Õ. Consequently. y2 È Öx×.   ¨ Note that f ÔxÕ a implies that there exists some y satisfying f Ôy Õ f ÔxÕ¡ 1. . so x y . (Germany) Solution 1. . . 2. For every x È N. g ÔxÕ g Ôy Õ).   minimal   k ¨ values attained ¨ k respectively. Now let a and b are the by f and g . . we establish the following Claim 4. the converse implication also holds. g Ông Õ b. b ¡ 1Ù (the class Önf ×). we have Öx× Ø1. which in turn follows f ÔxÕ f g Ôy Õ f Ôy Õ   1 a   1. Assume for a moment that a b. . Suppose that f and g are two functions defined on the set of positive integers and taking positive integer values. . . hence it coincides with Öng ×. each class Öx× contains at most one element from Nf . Ù. Then we have f g Ônf Õ a   k . surely.15 A6. . For any x È N. then we have g ÔxÕ g f ÔxÕ ¡ 1 g f Ôy Õ ¡ 1 g Ôy Õ. Therefore. Ð Next. . Induction on d. Similarly.     ¨ ¨ Next. Analogously. For every integer d 0. 2. . that is. so Öx× Öy × whenever y È Öx×. Now we investigate the structure of the sets Öx×. Next. a ¡ 1Ù if and only if gÔxÕ b. . by N we denote the set of all positive integers. for d 1   from d 1 d d from the induction hypothesis we have f Ônf Õ f f Ônf Õ f g Ônf Õ f Ônf Õ  d a   d. Ð Claim 2 implies that there exists exactly one class contained in Ø1. we have Öa× Önf ×. Throughout the solution. Proof. d 1 The equality g Ônf Õ a   d is analogous. . . define h ÔxÕ h h . Claim 2. . . For d 0. 2 . b a. b ¡ 1 Ù f ÔxÕ a. . . b   1. .     ¨¨ For k N and for any positive integer k . a b. we define Öx× Øy È N : x yÙ. and hence f ÔxÕ a. . . we clarify which classes do not contain large elements. then Öa× contains two elements a which is impossible by Claim 1. f d 1 Ônf Õ g d 1 Ônf Õ a   d. say f Ônf Õ a. ¨   k¡1 ¨ k Observe that f g Ô x Õ f g ÔxÕ   1 ¤ ¤ ¤ f ÔxÕ   k for any positive integer k . b ¡ 1Ù if and only if f ÔxÕ a. . y1 y2 for all y1 . and hence x  g Ôy Õ ¨ b. So. a ¡ 1Ù (that is. . Suppose also that the equations f Ôg ÔnÕÕ f ÔnÕ   1 and g Ôf ÔnÕÕ g ÔnÕ   1 hold for all positive integers. h0 ÔxÕ  x). . note that f ÔxÕ f Ôy Õ implies g ÔxÕ g f ÔxÕ ¡ 1 g f Ôy Õ ¡ 1 g Ôy Õ. hÔxÕ . so f g Ôy Õ f Ôy Õ  1 f ÔxÕ. the class Öng ×). then x y . By Claim 2. Ð . . Ù. .     ¨ ¨ Proof. (in any function h : N ÐÓÓÓÓÑÓÓÓÓÒ particular. if b c x then c g Ôy Õ for some y È N. g f Ông Õ b   k . We will prove that Öx× Ø1 .

there exists n0 È N such that f ÔnÕ n   1 for all n n0 . Claim 4½ . while for each n n0 . Now we get f Ô a Õ f Ô x Õ   2 f g ÔxÕ . we are prepared for the proof of the problem statement. almost the same method allows to find the values f ÔaÕ and g ÔaÕ. For each integer x a. there exists some x È N such that f ÔaÕ f ÔxÕ   1. which is false. It follows that g ÔxÕ g g d Ônf Õ g d 1 Ônf Õ 0. which together with a b proves the Claim. Assume the contrary. Claim 3½ . Induction on x. for each x È N. Solution 2.   ¨ N satisfying the property f Ôf ÔnÕÕ Comment. f ÔnÕ is an arbitrary number greater than of equal to n0 (these numbers may be different for different n n0 ). The equality g ÔaÕ a   1 is similar. hence there exist some x. Since f ÔaÕ   a   ¨ 1. then f ÔaÕ a   2. Ð Now. We start with the same observations. so f ÔxÕ f g dÔnf Õ and a   d f ÔxÕ by Claim 4. y È N such ¨   2 that f ÔxÕ f ÔaÕ ¡ 2 and f Ôy Õ g ÔxÕ (as g ÔxÕ a b). Proof. We can assume that a b. First. we have f ÔxÕ   a   ¨ d for some d hence x g d Ônf Õ. introducing the relation and proving Claim 1 from the previous solution. which is equivalent to f ÔaÕ f g ÔxÕ and a g ÔxÕ. while the induction   case ¨ x step follows from f Ôx   1Õ f g ÔxÕ f ÔxÕ   1 Ôx   1Õ   1 and the similar computation for g Ôx   1Õ. so by Claim 4½ we have f ÔnÕ   1   Finally. Since g ÔxÕ b a. Proof. this is impossible.16 Finally. Proof. Claim 2½ . by Claim 1 we have a g ÔxÕ b. . hence f ÔnÕ g ÔnÕ. The base a is provided by Claim 3½ . for an arbitrary n È N we have g ÔnÕ a. we prove it for n a.   ¨ Note that f ÔaÕ a since otherwise we have f ÔaÕ a and hence g ÔaÕ g f ÔaÕ g ÔaÕ   1. Now. we have f ÔxÕ g ÔxÕ x   1. For each such function. and by Claim 1 we get a g 2 ÔxÕ g f Ôy Õ 1   g Ôy Õ 1   a. ¨ f g ÔnÕ g ÔnÕ   1. f ÔaÕ g ÔaÕ a   1. ¨   so a g 2ÔxÕ a. a b. It is not hard now to describe all the functions f : N f ÔnÕ   1.

. (Iran) Prove that there exist positive integers ℓ r and N such that an Solution 1. . j1   j2 n. Moreover. Suppose that none of the numbers i3 . . r . max ai r i Ð aℓ and fix some index ℓ r such that s . i1   ¤ ¤ ¤   ik n. By the definition of ℓ. ai1   ¤ ¤ ¤   ai . If. i2 . we inductively define (1) an¡ℓ   aℓ for all n N. . aℓ aj jaℓ . i1   ¤ ¤ ¤   ik k n. . Let us delete these ℓ occurrences of j from Ôi1 . the indices Ôi1 . so or. Then. (4) On the other hand. .17 A7. . . . .   i2 (2) (3) r. i½3 . Finally. and so on. from the problem conditions we have that each an (n r ) can be expressed as an aj1   aj2 with j1 . . . . Then by the pigeonhole principle there is an index 1 j r which appears among i3 . j1 r then we can proceed in the same way with aj1 . k Summarizing these observations we get the following Claim. Rearranging the indices we may assume that ik ℓ. ik at least ℓ times. (4). . j2 n. ik¡1 Õ satisfy the conditions (4) with n replaced by n ¡ ℓ. an¡ℓ   aℓ for each n r 2 ℓ   2r. . if ai1 and ai2 are the numbers in (2) obtained on the last step. . (4) with ij ℓ for some j 3. For n an 1 k n 1 max ¡ Ôak   an¡k Õ. . we represent an in a form an 1 ij r. First. so k nßr rℓ   2. Thus. . and surely j ℓ. ℓ 2 Consider some n r ℓ   2r and choose an expansion of an in the form (2). . . i½k½ Õ also satisfying (4). from (1) we have an ask ask¡1   ai k ask¡2   ai ¡   ai k 1 k ¤¤¤ ai1   ¤ ¤ ¤   ai . and add j occurrences of ℓ instead. Then we have n i1  ¤ ¤ ¤  ik rk . ik Õ. suppose that the indices i1 . then i1 Hence we can adjust (3) as 1 ij r. after removing the coinciding terms. By Claim. we have an Now we denote s 1 i maxØai1   ¤ ¤ ¤   ai k : the collection Ôi1 . Let a1 . For every n r . ik satisfy the conditions (4). from the Claim we get an¡ℓ   aℓ   ai   ai½   ¤ ¤ ¤   ai½ ½ . . . . ik equals ℓ. this ℓ j means that ℓaj jaℓ . 2 3 k Ôai   ¤ ¤ ¤   ai ¡ Õ   aℓ 1 k 1 an . ik Õ satisfies (4)Ù. hence an 2 ai1 Thus. for every n r ℓ   2r we have found a representation of the form (2). obtaining a sequence Ôi1 . . . say. Finally. ℓaj . i1   i2 r. denoting sj i1   ¤ ¤ ¤   ij . . . ar be positive real numbers. . . . . we have ai1   ¤ ¤ ¤   aik an ai1   ai2   ai½3   ¤ ¤ ¤   ai½k½ . observe that in this representation. . which by (1) implies an as desired. .

the sequence Ôbn Õ is periodic with period ℓ from some index N .18 Solution 2. . Finally. we involve the expansion (2). ℓ we get that the sequence br t . . for all n N Ôbn¡ℓ   Ôn ¡ ℓÕsÕ   Ôbℓ   ℓsÕ an¡ℓ   aℓ   ℓ. and there is only a finite number of such ε sums. Otherwise. then bn ε sn. and we fix some index 1 ℓ r such that aℓ ai s max . bn   ns bn¡ℓ bn¡ℓ   bℓ for all n N   ℓ. . ik 1 k j M x can be expressed in the same way with k . . br t ℓ . . ik r ). . 1 k n 1 max ¡ Ôbk   bn¡k Õ bn¡ℓ bn¡2ℓ bℓ   bn¡ℓ bn¡ℓ . define M Then for n r we obtain bn so r . M M nonzero terms (otherwise x Then among bi ’s there are at most ¤ Ô¡εÕ ¡M ). which means that bn and hence an as desired. . . then bℓ 0. in view of the expansion (2). therefore it is constant from some index. . Now. for every t 1. 0 bn ¤ ¤ ¤ ¡M. for n r from the induction hypothesis we have bn 1 k n 1 max ¡ Ôak   an¡k Õ ¡ ns k 1 k n 1 max ¡ Ôbk   bn¡k   nsÕ ¡ ns 0 for all n. and Ôbn Õ satisfies the same recurrence relation We prove by induction on n that bn as Ôan Õ. Thus. let x bi   ¤ ¤ ¤   bi (i1 . 1 i r i ℓ Now. 1 i r max bi . rÙ Thus. . as required. is non-decreasing and attains the finite number of values. br t 2ℓ . As in the previous solution. and the statement is 0Ù. . . . The base cases n r follow from the definition of s. Now. hence an minØ bi : 1 i r. . . 0. 2. Thus ε ε T 1 2 k Øbi   bi   ¤ ¤ ¤   bi : i1 . we get that each bn is contained in a set Ö¡M. 0× We claim that this set is finite. we introduce the sequence Ôbn Õ as bn an ¡ sn. . for any x È T . bi 1 k n 1 max ¡ Ôbk   bn¡k Õ 0. Actually. (3) applied to the sequence Ôbn Õ. (3). if bk 0 for all 1 trivial.

6ÔT   S Õb ¡ 4T S 0 f 3bÔa   4b   f Õ bÔ4b ¡ aÕ 2a   b a b 1  ¢ 2Ôa   2bÕÔ2b   f Õ 3Ôb ¡ aÕ 2a   b ª b. f Õ is a point of equality. cÕ. T  S 3Ôαβ   αγ   βγ Õ 3ÔT   S ÕÔT σ   Sτ Õ. (2) Surely. we have P ÔaÕ S Ôa ¡ bÕÔa ¡ dÕÔa ¡ f Õ 0. which immediately implies that b c d eÔ α β γ Õ. T  S P ÔxÕ P ÔxÕ ÔT   S ÕÔx ¡ αÕÔx ¡ β ÕÔx ¡ γ Õ. Comment 1. cÕ. In fact. one can locate the roots of P ÔxÕ more narrowly: they should lie in the intervals Ôa. bÕ. γ È Ôe. 3ÔSτ   T σ Õ . if we change all inequality signs in the problem statement to non-strict ones. Since there are at most three roots at all. b such that the This means that for every pair of numbers a. there exists f . Moreover. f such that a and b   d   f T . P Ôf Õ T Ôf ¡ aÕÔf ¡ cÕÔf ¡ eÕ 0. we obtain that there is exactly one root in each interval (denote them by α È Ôa. Hence. β . f Õ). each of the intervals Ôa. Let a  c   e S (1)   df Õ   T Ôac   ae   ceÕ . we have 0 which implies 4S 2 T 2 ÔT   S Õ2 or 4S 2 T 2 which is exactly what we need. P ÔeÕ S Ôe ¡ bÕÔe ¡ dÕÔe ¡ f Õ 0. the (non-strict) inequality will also hold by continuity. The idea of the solution is to interpret (1) as a natural inequality on the roots of an appropriate polynomial. P ÔcÕ S Ôc ¡ bÕÔc ¡ dÕÔc ¡ f Õ 0. the polynomial P can be factorized as Equating the coefficients in the two representations (2) and (3) of P ÔxÕ provides α β γ 2T S . eÕ. Actually. dÕ. β È Ôc. γ are distinct. Surely. τ bd   bf   df . b such that 0 point Ôa. We define also σ ac   ce   ae. Prove that 2ST 3ÔS   T Õ S Ôbd   bf   b c d e ¨ f . (3) Now. Ôc. since the numbers α. Ôe. Ôc. (South Korea) Solution 1. f Õ contains at least one root of P ÔxÕ. Moreover. Ôe. b. d. together with the additional condition that P ¾ ÔbÕ 0. f Õ. consider the polynomial Ôb   d   f ÕÔx ¡ aÕÔx ¡ cÕÔx ¡ eÕ   Ôa   c   eÕÔx ¡ bÕÔx ¡ dÕÔx ¡ f Õ T Ôx3 ¡ Sx2   σx ¡ aceÕ   S Ôx3 ¡ T x2   τ x ¡ bdf Õ. Ôα ¡ β Õ2   Ôα ¡ γ Õ2   Ôβ ¡ γ Õ2 Ôα   β   γ Õ2 2Ôα   β   γ Õ2 ¡ 6Ôαβ   αγ   βγ Õ.19 A8. c. One can also find when the equality is achieved. b. e. eÕ. Algebraically. This happens in that case when P ÔxÕ is a perfect cube. b. b. Given six positive numbers a. T  S αβ   αγ   βγ Sτ   T σ . b. P is cubic with leading coefficient S   T 0. b.

¨ ÔL. τ ¡ σ sÔT ¡ S Õ.Õ2  Ô2ST Õ2 ¡ ÔS   T Õ S ¤ ¨3Ôbd   bf   df Õ   T ¤ 3Ôac   ae   ceÕ 4S 2 T 2 ¡ ÔS   T Õ S ÔT 2 ¡ V Õ   T ÔS 2 ¡ U Õ ÔS   T ÕÔSV   T U Õ ¡ ST ÔT ¡ S Õ2 . and the statement is equivalent with ÔS   T ÕÔSV   T U Õ By the Cauchy-Schwarz inequality. Then we have Sτ or (7)   Tσ S Ôτ ¡ σÕ   ÔS   T Õσ ¢ SsÔT ¡ S Õ   ª ÔS   T¢ÕÔce   asÕ ª 2 s 3 SsÔT ¡ S Õ   ÔS   T Õ   ÔS ¡ sÕs s 2ST ¡ 4 ÔS   T Õs . ¨ Hence.20 Solution 2.Õ2 ¡ ÔR.S.H. since each summand is negative.S. let s Note that Ôc ¡ bÕÔc ¡ dÕ   Ôe ¡ f ÕÔe ¡ dÕ   Ôe ¡ f ÕÔc ¡ bÕ 0. 2ST 3ÔS   T Õ S Ôbd   bf     df Õ   T Ôac   ae   ceÕ . ¨2 (4)   ÔS   T ÕÔT U   SV Õ Estimate the quantities U and and Ôd ¡ bÕ2 being omitted: U   V by the QM–AM inequality with the positive terms Ôe ¡ cÕ2 S ¤ TU   ST U   V ¨2 .H. (5)   V Ôe ¡ aÕ   Ôc ¡ aÕ   Ôf ¡ bÕ   Ôf ¡ dÕ 2 2 3 Ô e ¡ dÕ   Ôc ¡ bÕ ÔT ¡ S Õ   3 2 2 T Ôe ¡ aÕ2   Ôc ¡ aÕ2   Ôf ¡ bÕ2   Ôf ¡ dÕ2 2 2 ¢ f ¡ ¡ d 2 b 2 ª   ¡e c   ¡a 2 2 © ¡ S. ST ÔT T ¤ SV ¡ S Õ2. . The estimates (5) and (6) prove (4) and hence the statement. Let U and V Then ¨ 1  e ¡ aÕ2   Ôc ¡ aÕ2   Ôe ¡ cÕ2 Ô 2 S 2 ¡ 3Ôac   ae   ceÕ T 2 ¡ 3Ôbd   bf 1  Ôf 2 ¡ bÕ2   Ôf ¡ dÕ2   Ôd ¡ bÕ2   ¨   df Õ. We keep using the notations σ and τ from Solution 1. Moreover. Solution 3. 4 3 ÔS   T ÕÔSτ 4 Using this inequality together with the AM–GM inequality we get   T σÕ 3 4 ÔS   T Õs   2ST ¡ 3 ÔS   T Õs 4 2 3 ÔS   T Õs 2ST 4 ¢ ¡ 3 ÔS   T Õ s 4 ª ST. This rewrites as Ôbd   bf   df Õ ¡ Ôac   ce   aeÕ Ôc   eÕÔb   d   f ¡ a ¡ c ¡ eÕ. (6) c   e.

but σ . Let k . c¡b 3. c¡m. z Õ x   y   z . z Õ. denote U Ôx. Hence. f Õ by Ôa 2m. After the change we have 2 a b c d e f . c. y ½ . d¡c 1. e. f Õ Ôa   m. For each change. we have υ Ôx½ . d. e ¡ k Õ. . 3 After the change. z Õ U Ôx. c. The expression (7) can be found by considering the sum of the roots of the quadratic polynomial q ÔxÕ Ôx ¡ bÕÔx ¡ dÕÔx ¡ f Õ ¡ Ôx ¡ aÕÔx ¡ cÕÔx ¡ eÕ. f Õ by Ôc   ℓ. y. (ii) keeping the values of sums S and T unchanged. Let ℓ . then we have U Ôx½ . Replace Ôa. c. and we need to prove the inequality 2Ôa   2bÕÔ2b   f Õ 3Ôa   4b   f Õ Ôa   2bÕÔb2   2bf Õ   Ô2b   f ÕÔ2ab   b2 Õ 3bÔa   4b   f Õ ¤ Now. It is easy to see that each of these two consecutive changes satisfy the conditions of the Lemma. c ¡ m. For the second. b. d ¡ k. dÕ Ôa   m. hence the values of σ and τ increase. d. υ Ôx. . The first equality is obvious. we change the variables (i) keeping the (non-strict) inequalities a b c d e f . e ¡ m. x ¡ y . y ½ . e. and finally (iii) increasing the values of σ and τ . τ strictly increase by the Lemma. f ¡mÕ. . b   m. b 2m. y. Suppose that x½   y ½ x   y but x ¡ y x½ ¡ y ½ . e ¡ ℓ. y. c   k. z Õ xy   xz   yz . d   ℓ. z Õ and υ Ôx½ . To check (iii). z Õ υ Ôx. but σ . . d. Finally. d¡m. while the right-hand side increases. e¡m. observe that 2 ¤ 2Ôa   2bÕÔ2b   f Õ 3bÔa   4b   f Õ       ¨ ¨ Ôa   2bÕÔb   2f Õ   Ô2b   f ÕÔ2a   bÕ . We introduce the expressions σ and τ as in the previous solutions. Namely. we come to the situation when a b c d e f . e¡d 2. eÕ by Ôb   k. . we find the sufficient conditions for (ii) and (iii) to be satisfied. y ½ . e. we observe that our change can be considered as a composition of two changes: Ôa. After the change we have 2 a b c d e f . Solution 4. b   m. f ¡ ℓÕ. b. T are unchanged. Proof.21 Comment 2. Replace Ôc. Lemma. The idea of the solution is to change the values of variables a. T remain unchanged. b. Replace Ôb. T remain unchanged. we have a b c d e f and S . the values of S . z Õ z Ôx½   y ½ Õ   x½ y ½ z Ôx½   y ½ Õ   Ôx½   y½Õ2 ¡ Ôx½ ¡ y½Õ2 4 4 x½ ¡ y ½ Ôx   yÕ2 ¡ Ôx ¡ yÕ2 z Ôx   y Õ   Ôx ¡ y Õ2 υ Ôx. let m . ¨ (8)   Ôa   2bÕÔb   2f Õ   Ô2a   bÕÔ2b   f Õ . y. we apply Lemma several times making the following changes. First. Then the left-hand side of (1) remains unchanged. τ strictly increase by Lemma. as desired. y. we denote the new values by the same letters to avoid cumbersome notations. f ¡ mÕ. it will lead to a simpler inequality which can be proved in a direct way. the values of S . d ¡ mÕ and Ôa. y. the inequality (1) (and even a non-strict version of (1)) for the changed values would imply the same (strict) inequality for the original values. f keeping the left-hand side unchanged and increasing the right-hand side. Ð with the equality achieved only for Ôx½ ¡ y ½ Õ2 Now. Finally. Let x. z Õ with equality achieved only when x ¡ y x½ ¡ y ½ . z 0.

we also can find all the cases of equality.22 Hence Ô4Õ rewrites as 3bÔa   4b   f Õ     Ôa   2bÕÔb   2f Õ   Ô2a   bÕÔ2b   f Õ   ¨ 2 3bÔa   4b   f Õ ¤ Ôa   2bÕÔb   2f Õ   Ô2b   f ÕÔ2a   bÕ . 3bÔa   4b   f Õ Ôa   2bÕÔb   2f Õ   Ô2a   bÕÔ2b   f Õ. Here. e are distinct then one can use Lemma to increase the right-hand side of (1). . which leads to the same equality as in Comment 1. Comment 3. d. this means that we apply AM–GM to equal numbers. if b c d e. c. ¨ which is simply the AM–GM inequality. then we need equality in Ô4Õ. Actually. that is. Further. it is easy to see that if some two numbers among b.

. Aik1 Õ is a good order for Ai’s and ÔBj1 . Conversely. Ð In view of Lemma. . dÕ. dÕ. d. . 3 and 13 singers and realizing the numbers 5. . 1 Fig. we show how to construct sets of singers containing 4. Then. . In a concert. c. We start with the following simple Lemma. Solution. . we say that a number N is realizable by k singers (or k -realizable) if for some set of wishes of these singers there are exactly N good orders. Thus. Ôb. each good order of the obtained set of singers has the form ÔAi1 . where ÔAi1 . . cÕ. Proof. d. Can it happen that there are exactly 2010 orders of the singers such that all their wishes are satisfied? (Austria) Answer. Bj1 . Let the singers A1 . Bjk2 Õ is a good order for Bj ’s. . . the wishes are denoted by arrows. cÕ. a1 a2 a3 a4 a b y c (5) d (3) a5 a6 a7 a8 a9 a10 a11 (67) x Fig. For each singer. . 1. Then the number n1 n2 is Ôk1   k2 Õ-realizable. b. For Fig. . Add to each singer Bi the wish to perform later than all the singers Aj . Next. . . . . 1–3. Bk2 (with some wishes among them) realize n2 .Combinatorics C1. . there are exactly 5 good orders Ôa. . each of 6 orders is good since there are no wishes. and the number of good orders for each Figure stands below in the brackets. . we have to prove that a number 2010 is 20-realizable. a. Ôb. n2 are realizable by respectively k1 and k2 singers. a. Thus the number 2010 6 ¤ 5 ¤ 67 will be realizable by 4   3   13 20 singers. such an example exists. there is a (possibly empty) set of other singers such that he wishes to perform later than all the singers from that set. b. respectively. We say that an order of singers is good if it satisfied all their wishes. Bjk2 Õ. Suppose that numbers n1 . These companies of singers are shown in Figs. . . c. each order of this form is obviously good. 2 Fig. . Aik1 . . 20 singers will perform. Ôa. and the singers B1 . Yes. . 3 For Fig. . cÕ. the number of good orders is n1 n2 . Hence. 6 and 67. d. Ak1 (with some wishes among them) realize n1 . . a. Ôb. 2.

3 that each of these orders can be complemented by x and y in exactly 67 ways. it may seem even simpler. . exactly one half of them satisfies the wish b1  b2 . Surely. Hence. b5 .24 Finally. 5 Fig. . . Comment. the positions of x and y in this line determine the whole order uniquely unless both of them come between the same pair Ôai . for Fig. the total number of good orders is 63   4 67. . . hence obtaining 30 ¤ 67 2010 good orders at all. and exactly one half of these orders satisfies another wish c1  c2 . . . . . 6 2. as in Fig. . the large variety of ideas standing behind these examples allows to suggest that there are many other examples. a4 . 1. c1 . 4b and 3 in a smarter way. b1 . . . . Here we present some improvements of the example leading to a smaller number of singers. Here. a1 x a2 a a3 y a4 (30) a) b2 (30) b) c2 y b1 c1 b2 b4 a6 b1 b3 b5 a5 y x a8 a9 a10 b1 b3 a6 a7 b2 b4 a7 x a8 c9 c11 c10 a11 (2010) (2010) Fig. 4b. 1–3 to represent 2010 by 13 singers as is shown in Fig. . for 5 singers a. . and there are 6 ways to add y into the resulting order of a1 . Moreover. 5. there are exactly 5!ß4 30 good orders. it is more economic to make an example realizing 30. . The number 20 in the problem statement is not sharp and is put there to respect the original formulation. So. . a4 is fixed. the order of a1 . c2 there are 5! 120 orders at all. in the latter cases. Instead of building the examples realizing 5 and 6. This example is shown in Fig. if necessary. x. . each example with 20 singers can be filled with some “super-stars” who should perform at the very end in a fixed order. a11 satisfying all their wishes among themselves. one can merge the examples in Figs. Obviously. 4 Fig. a11 is fixed. For Fig. . b5 Ù means that there exists such arrow from each member of this group. singer x can stand before each of ai (i 9). hence. there are two orders instead of 1 due to the order of x and y . b2 . Hence each of these improvements provides a different solution for the problem. Analogously. Two possible examples consisting of 5 and 6 singers are shown in Fig. 4b. an arrow from/to group Øb1 . and singer y can stand after each of aj (j 5). 3. obtaining a set of 13 singers representing 2010. the difficulty level of this problem may be adjusted by replacing 20 by a smaller number. hence the number 20 can be decreased to 19 or 18. On Fig. one can prove in the same way as for Fig. the order of a1 . . One can merge the examples for 30 and 67 shown in Figs. 4. in this line. thus resulting in 9 ¤ 7 63 cases. as desired. Moreover. Further. . . . ai 1 Õ (thus 5 i 8). . Hence there are 5 ¤ 6 30 good orders. 6. there are 5 ways to add x into this order. a6 . . 4a. one can see that there are exactly 30 orders of b1 .

this company is shown in Fig. 8. Finally. 8 3. even a company of 10 singers representing 2010 can be found. we will present two other improvements. Moreover. The graph in Fig.25 a1 a2 a3 b1 b4 b5 b6 (67) a4 b2 b3 a5 a4 a3 a2 a1 (2010) a6 b2 b1 b3 b4 Fig. 7 shows how 10 singers can represent 67. . 7 Fig. the proofs are left to the reader.

YB. respectively. and for BY that is YB). L2 Õ can form a 2 ¢ 2 square with monochrome diagonal. we find a desired arrangement of four flags. R È S ØBB. assume that N 4. C2. We are left to show that MN 2N ¡2   1. r ) should be at least 3. R4 are distinct. one of the pairs ÔL½ . thereby. we will always mean the main diagonal. We start with establishing this statement for N 4. where the 2 ¢ 1 flags L. there exists only one 2 ¢ 1 flag C È S (possibly C and C cannot form a 2 ¢ 2 square with monochrome diagonal (for part BB. R3 . Obviously. YB)). so there are two flags with coinciding right part. L4 R4 can also form a 2 ¢ 2 square with a blue diagonal by (ii). Consider all the good flags and remove the first square from each of them. A2 . YY) and (BB. these pairs are (YB. First. we show that MN 2N ¡2 . there are 2N countries (N Solution. and two of them can form a 2 ¢ 2 square with blue diagonal (for all parts but BB. there exist N flags forming a diverse set. thus obtaining the desired answer. Determine the smallest positive integer M such that among any M distinct flags. Finally. there exist some flags L3 R3 and L4 R4 such that R1 . We say that a set of N flags is diverse if these flags can be arranged into an N ¢ N square so that all N fields on its main diagonal will have the same color. Let L½ R½ be the remaining flag. Suppose that we have 5 flags of length 4. YYÙ are its left and right parts. this column contains at least 2 squares of one color (say. We obtain at least 2N ¡3   1 MN ¡1 flags of length N ¡ 1. above we have proved the base case N 4. hence one of the factors (say. Consider the collection of all 2N ¡2 flags having yellow first squares and blue second ones. we make two easy observations on the flags 2 ¢ 1 which can be checked manually. We are ready to prove the problem statement by induction on N . consider any 2N ¡2   1 flags of length N . Let MN be the smallest positive integer satisfying the problem condition. L2 form a square with a blue diagonal. The total number of flags is 5 rℓ. M 2N ¡2   1. A) such that A (i) For each A È S .26 4). r 4. since r 3. we may assume this Î column is the Ê that 2N ¡2   1 2N ¡3   1 first one. blue). L1 Õ and ÔL½ . BY. Next. ℓ. each field being either yellow or blue. A3 È S be three distinct elements. actually. then two of them can form a 2 ¢ 2 square with yellow diagonal. On some planet. let them be L1 R1 and L2 R1 (L1 L2 ). By the pigeonhole principle. On the other hand. When speaking about the diagonal of a square. First. We call the flags with a blue first square good. while for all parts but BY. we denote each flag as LR. the right parts of two of the flags L1 R1 . let ℓ and r be the numbers of distinct left and right parts of our 5 flags. Each country has a flag N units wide and one unit high composed of N fields of size 1 ¢ 1. We decompose each flag into two parts of 2 squares each. a pair (BY. respectively. we can assume that L½ . YB) fits for both statements. (Croatia) Answer. For the induction step. N ¡ 1 of them . (ii) Let A1 . Putting these 2 ¢ 2 squares on the diagonal of a 4 ¢ 4 square. that is YY. This flag contains a non-monochrome column since the flags are distinct. and arrange them into a large flag of size Ô2N ¡2   1Õ ¢ N . both colors appear on the diagonal of each N ¢ N square formed by these flags. Now. L3 R3 . No two countries have the same flag. By (i). by the induction hypothesis.

returning the removed squares. and Ôc. in both cases we get a desired N ¢ N square. Note that Sy . Assume that these matchings do not exist. so there exists some c È Sy Sb . Ey Õ and Gb ÔV V ½ . We present a different proof of the estimate MN 2N ¡2   1. respectively. this is impossible for N 4. it is not used in Q). By Hall’s lemma. Consider arbitrary 2N ¡2   1 distinct flags and arrange them into a large Ô2N ¡2   1Õ ¢ N flag. So. If Q has a yellow diagonal. Let V and V ½ be the set of columns and the set of flags under consideration. Hence the only undetermined positions in these flags are the remaining N ¡ y ¡ b ones. From the ¨construction of our graph. hence Ey ÔSy Õ Eb ÔSb Õ V ½ . if the diagonal of Q is blue then we can take any of the 2N ¡3   1 ¡ÔN ¡ 1Õ 0 remaining good flags. We do not use the induction. b   Sb . Note that each flag is connected to c either in Gy or in Gb . or.27 can form a square Q with the monochrome diagonal. let the edge Ôc. Now. Conversely. Ey ÔSy Õ and Eb ÔSb Õ denote respectively the sets of all vertices connected to Sy and Sb in the corresponding graphs). Then we have to prove exactly that one of the graphs Gy and Gb contains a matching with all the vertices of V involved. and our aim is to supplement it on the top by one more flag. then we can take each flag with a yellow first square (it exists by a choice of the first column. so 2N ¡y¡b V¾ V ½ ¡ Ey ÔSy Õ ¡ EbÔSb Õ 2N ¡2   1 ¡ Ôy ¡ 1Õ ¡ Ôb ¡ 1Õ. Next. Solution 2. Construct two bipartite graphs Gy ÔV V ½. it means that there exist two sets of columns Sy . EbÕ with the common set of vertices as follows. involving Hall’s lemma on matchings instead. This is impossible since N c 2. . First. ¾ V ½ Þ Ey ÔSy Õ Eb ÔSb Õ have blue squares in the we have that all the flags in the set V columns of Sy and yellow squares in the columns of Sb . 2N ¡c   c 2N ¡2   2. moreover. f Õ È Eb otherwise. Hence we have 2N ¡2   1 V½ Ey ÔSy Õ   Eb ÔSb Õ Sy   Sb ¡ 2 2N ¡ 4. we obtain a rectangle ÔN ¡ 1Õ ¢ N . Sb V since N 2N ¡2   1. Let y Sy . suppose that Sy Sb ∅. f Õ appear in Ey if the intersection of column c and flag f is yellow. denoting c y   b. Our aim is to prove that this is impossible. Sb V such that Ey ÔSy Õ Sy ¡ 1 and Eb ÔSb Õ Sb ¡ 1 (in the left-hand sides. So. we have Sy Sb ∅.

L-blocks. 1 Fig. 1). Consider any B-block of the form Ô1. and we obtain 25 T-columns. These 25 B-columns contain 1250 B-blocks. refers to the j th block in the ith row (or the ith block in the j th column). by the pigeonhole principle each block must contain exactly one king. Consider some index i such that the ith row is an L-row. no two kings are placed in two squares sharing a common vertex). thus the . j 50. or T-blocks. then Ôi. respectively. j Õ cannot be a TR-block (see Fig. the pair Ôi. By (i½ ). j Õ is a T-block. Find the number of such arrangements. Suppose that we have an arrangement satisfying the problem conditions.) (Russia) Answer. 2500 chess kings have to be placed on a 100 ¢ 100 chessboard so that TL i BR TR i+1 L BL TR T B Fig. if Ôi. By (ii½ ). and the Ôj   1Õth column is a Bcolumn.e. j Õ is a T-block then Ôi ¡ 1. 2). Similarly we define TR-blocks. Ôi   1. there are exactly 25 L-rows and exactly 25 R-rows. so we obtain 25 B-columns. 2 Case 1. j ¡ 1Õ is an L-block. (ii) each row and each column contains exactly 25 kings. assign to each block a letter L or R if a king stands in its left or right half. j Õ. if Ôi. j Õ is an R-block then Ôi. (ii½ ) Each column contains exactly 25 L-blocks and 25 R-blocks. In particular. and BR-blocks. j Õ is a B-block: otherwise the kings in these two blocks can take each other. Similarly: if Ôi. We identify the blocks by their coordinate pairs. j Õ is a B-block then Ôi   1. the total number of L-blocks (or R-blocks. hence.. and R-blocks. and each row contains exactly 25 T-blocks and 25 B-blocks. Similarly. Each block can contain not more than one king (otherwise these two kings would attack each other). The arrangement of blocks determines uniquely the arrangement of kings. There are two such arrangements. we call a block a TL-block if it is simultaneously T-block and L-block. j   1Õ is an R-block. or B-blocks) is equal to 25 ¤ 50 1250. (Two arrangements differing by rotation or symmetry are supposed to be different. We also combine the letters. So we define T-blocks. B-blocks. The upper-left block is Ô1. 1 1 2 3 TL TL BL 2 BL 3 j j +1 C3. Now consider an arbitrary pair of a T-column and a neighboring B-column (columns with numbers j and j   1). where 1 i. Now assign to each block a letter T or B if a king is placed in its top or bottom half. so in the rest of the solution we consider the 50 ¢ 50 system of blocks (see Fig. j Õ. so we call such a column B-column. The system of blocks has the following properties. Divide the board into 2 ¢ 2 pieces. we have 25 B-blocks in the first row. hence all blocks in the remaining columns are T-blocks. j Õ is an L-block then Ôi. BL-blocks. (i½ ) If Ôi. hence Ôi   1. Similarly. j   1Õ is a BL-block. all blocks in the j th column are B-blocks. we call these pieces blocks. Solution. 1Õ.28 (i) no king can capture any other one (i. Therefore. Suppose that the j th column is a T-column. j Õ is a TL-block.

4). Replacing in the previous reasoning rows by columns and vice versa. we obtain that the rows from the 1st to the 25th are L-rows (and all other rows are R-rows). so we find exactly one more arrangement of kings (see Fig. Repeating the arguments from Case 1. TR TR BR BR TR TR BR BR TL TL TL TL BL BL BL BL Fig. Now. the columns from the 1st to the 25th are B-columns (and all other columns are T-columns). the columns from the 1st to the 25th are T-columns. 3). 4 Case 2. Now consider the neighboring R-row and L-row (that are the rows with numbers 25 and 26). and the columns from the 26th to the 50th are B-columns. and the Ôj   1Õth column is a T-column. 3 Fig. choosing the ith row to be the topmost L-row. it follows that the rows from the 1st to the 25th are R-rows. we successively obtain that all rows from the ith to the 50th are L-rows. Since we have exactly 25 L-rows. and the rows from the 26th to the 50th are L-rows.29 Ôi   1Õth row is an L-row. So we have a unique arrangement of blocks that leads to the arrangement of kings satisfying the condition of the problem (see Fig. . Suppose that the j th column is a B-column.

2k . Lemma 1. Pk 1. . . 0. 1Õ Ô0. 2a. For every positive integer n. 0. . S6 of coins are standing in a row. 0Õ. . Pa. but the constant 20102010 ½ 2010 is replaced by 20102010 . Starting from Ôa ¡ k. Proof.g. . Yes (in both variants of the problem). 0. Ôa. 4 contains at least one coin. 2k . 0Õ Ôa ¡ k. 0Õ for every a 1. Six stacks S1 . . until it becomes empty. apply Move 1 to the middle stack 2k times. Ôa. 0. 2 ÐÓ ÓÑÓ ÓÒ (e. 0Õ Ôa ¡ k ¡ 1. 0Õ for every a Ôa ¡ k. 0. 0Õ Ôa ¡ k. C4 . There exists such a sequence of moves. . 0. 0. a½n coins respectively. We prove by induction that Ôa. . 0. Then 1. Ð . (Netherlands) Answer. 0. Pk . an coins. 2k . 0Õ. 2010 Let A 20102010 or A 20102010 . 0Õ Ôa ¡ 1. Pk . . 0Õ Ôa ¡ k. 1. 0. Then apply Move 2 to the first stack: Ôa. 0Õ for every 1 k a. we prove that Ôa. Now assume that the lemma holds for some k Lemma 1. apply Ôa ¡ k. 0Õ Ôa ¡ k. Same as Problem C4. Similarly to Lemma 1. 0. 0. Ôa½1 . 0Õ Ôa ¡ 1. 0. 0. For k 1. 0. 0Õ. 0. 2k 1. . 2. apply Move 1 to the first stack: Ôa. 0Õ for every 1 apply Move 1 to the first stack: k a. a2 . 0. you may remove one coin from Sk and add two coins to Sk 1 . 0Õ Therefore. AÕ. 2k 1Õ . 0Õ. 0Õ. 0. There are two types of allowed moves: Move 1 : If stack Sk with 1 k 5 contains at least one coin. a½nÕ the following: if some consecutive stacks Solution. Pk . Proof. 0Õ. 2 Ôa ¡ k ¡ 1. . Ôa. 2. Pk 1. respectively. . First we prove two auxiliary observations. 0. then apply Move 1 to the first stack: k a. . Starting from Ôa ¡ k. whereas the sixth stack S6 contains exactly 20102010 coins. 1. P1. 0. 2P . . . 21. whereas the contents of the other stacks remain unchanged. . .30 C4. 0. an Õ contain a1 . In the beginning every stack contains a single coin. 0. . 0. 0Õ For k 1. 0Õ Ôa ¡ k. 0Õ Ôa ¡ 1. 0Õ. 2Õ ¤ ¤ ¤ Ôa. Pk 1. Our goal is to show that Ô1. 2k 1. 2k . Now assume that k a and the statement holds for some k a. 0Õ Ô0. then you may remove Move 2 : If stack Sk with 1 k one coin from Sk and exchange stacks Sk 1 and Sk 2 . 0. Pk . let Pn 16). . 0Õ Ôa ¡ 1. . Ôa ¡ k. 1. 0Õ Hence. then it is possible to perform several allowed moves such that the stacks contain a½1 . 0Õ Ôa ¡ k. Denote by Ôa1 . 0. P3 n 2. Decide whether it is possible to achieve by a sequence of such moves that the first five stacks 2010 are empty. 0. 1. Ôa ¡ k. 0Õ. 0.. 0. Ôa ¡ k ¡ 1. 0Õ Ô0. a½2. Ôa ¡ k ¡ 1. Ð 22 2 Lemma 2. 2k ¡ 1.

1Õ Ô0. 1. P16 ¡ 1. 1. 0. 0Õ ¤ ¤ ¤ Ô0. 0Õ Ô0. 0.) Ô0. 17. 0. since A 20102010 2010 Ô211 Õ2010 2010 211¤2010 2010 22010 2011 11 2011 2Ô2 Õ 22 11 2011 ¤ 22 215 P16 . P16 ¡ 2. 0. 0. 1. 0. 0. 0. 1. 0. 1. 0Õ Ô0. 0Õ. Finally. The simpler variant C4½ of the problem can be solved without Lemma 2: Ô1. 0. 0Õ Ô1. 0. 1. 1Õ Ô1. 0. 1. 3. 16. 0. 1. 0. 1. 0. . 0. 0. 0. 3Õ Ô1. 0Õ Ô0. 0Õ Ô0. apply Move 1 repeatedly to empty stacks S4 and S5 : Ô0. 0. 34 34 34 For this reason. 0. 0. 1. 0. 35. We already have more than A coins in stack S4 . 0. it is not hard to check manually that we can achieve at most 28 coins in the last position. 9. 1Õ Ô0. 1. 0. P3. 1. 0Õ Ô0. Ô0. 0Õ Ô0. 1. 234 . 0. S3 . Aß4. 3. 3. P16. 0. 0. AÕ. 1. 0Õ . 0Õ Ô0. 0Õ Ô0. However. 1. 0. 1. 3. 0. 0Õ ¤ ¤ ¤ Ô0. S2 and S1 in this order. the PSC suggests to consider the problem C4 as well. 0. 1. To decrease the number of coins in stack S4 . 1. 1. 35Õ Ô0. 1. 1. 1. 0. 0. 22 . 0. 0. Comment 1. Then apply Lemma 2 twice: Ô1. 2 It is not hard to show that the numbers 20102010 and 20102010 by any nonnegative integer up to PP214 . 0.31 Now we prove the statement of the problem. 2010 in the problem can be replaced Comment 2. 0. then apply Move 2 to stacks S4 . 0Õ Ô1. 0. 35. 0. 0Õ Ô0. 0. 0Õ. AÕ. 1. (In every step. Aß2. 0Õ Ô0. 0. 0Õ Ô0. we remove a coin from S4 and exchange the empty stacks S5 and S6 . 0. 0. On the other hand. 0. 1. apply Move 2 to this stack repeatedly until its size decreases to Aß4. 1. 0. Aß4. 0Õ ¤ ¤ ¤ Ô0. . the problem statement in C4½ hides the fact that the resulting amount of coins can be such incredibly huge and leaves this discovery to the students. 22 . 3. 0. 0. 0. Starting with only 4 stack. 0. 5. 0. 0. 0. 0. 0. 0. 1Õ Ô0. 0. 0. 0Õ Ô0. 0. 1. Aß2. 0. 1. 0. 0. Aß4. First apply Move 1 to stack 5. . 1. 0. 1. 22 ¡ 1. 0. 0Õ Ô0. around 5 and 6 stacks the maximal number of coins 14 explodes. 0. 0. 1. 0. 1. 1Õ Ô0. 1. 0. 1. 1Õ Ô0. With 6 stacks the maximum is greater than PP 14 . 0Õ Ô0. With 5 stacks it is possible to achieve more than 22 coins. 0. 35. 0. 1. Problem C4 requires more invention and technical care. P16.

denote by S ÔT Õ sum under consideration. £ ô j i « 3 εij ô ô j1 . let εij 1 if the ith player wins against the j th one. and εij ¡1 otherwise. hence S ÔT Õ ô Øi. then only two possibilities. otherwise the company of all players is bad. S ÔT Õ i ÊØj1 . If. 2. while in the latter one S ÔT Õ 13   13   Ô¡1Õ3   Ô¡1Õ3 0. . Hence we should have A Ô3.j3 Ù εij1 εij2 εij3   P ÔT Õ 4 ô i j εij S1 ÔT Õ   P ÔT ÕS2ÔT Õ. we may assume that a1 a2 a3 a4 . 1. as desired. for instance. 0Õ. and S ÔT Õ 33   13   Ô¡1Õ3   Ô¡3Õ3 0.j3 Ù εij1 εij2 εij3 . i2 . Consider a tournament T with no bad companies and enumerate the players by the numbers from 1 to n. we have S ÔTi1 i2 i3 i4 Õ 0. Let wi and ℓi be respectively the number of wins and losses of the ith player. hence a4 1. if a4 1.j1 . By the abovementioned. . Actually. For any tournament T satisfying the problem condition. i3 . To simplify this expression. . Then Ôwi ¡ ℓiÕ3 Hence. we show that the statement holds if a tournament T has only 4 players. n 4 players participated in a tennis tournament. then we cannot have 2 a1 a2 a3 2. a3 . On the other hand. ij1 Make such replacements for each such term. This way the problem statement will be established. Now we turn to the general problem. We interpret the number Ôwi ¡ ℓi Õ3 as following. after this change each term of the form εij3 will appear P ÔT Õ times.32 C5. 1. then the term contains ε2 1. . Suppose that there is no bad company in this tournament. let A Ôa1 . 1Õ and A Ô2. i4 consider a “sub-tournament” Ti1 i2 i3 i4 consisting of only these players and the games which they performed with each other. a4 Õ be the number of ¨ wins of the players. consider all the terms in this sum where two indices are equal. i 1 First.i2 .j2. (2) i1 . . For every 4 players i1 . 1.j2 . 1. obviously.i4 where the sum is taken over all 4-tuples of distinct numbers from the set Ø1. nÙ. 4 We have a1   a2   a3   a4 6. A Ô3. In the former case we have S ÔT Õ 33   3 ¤ Ô¡2Õ3 0. Prove that n ô Ôwi ¡ ℓiÕ3 0. 1Õ can take place. so we can replace this term by εij3 .i3 . and there was no tie game. a2 . namely S ÔT Õ n ô Ôwi ¡ ℓiÕ3. 2. and each of these three players won a game and lost another game among themselves. Any two players have played exactly one game. Our aim is to prove that S ÔT Õ ô S ÔTi1 i2 i3 i4 Õ. We call a company of four players bad if one player was defeated by the other three players. If a4 0.j2. For i j . (1) (South Korea) i 1 Solution. j1 j2 .j3 i εij1 εij2 εij3 .

and let us count all cases when a player is the local champion of some k -company. n ô i 1 n ô   i 1 n ô i 1 n ô ¢ i 1 k¡1 ¢ ª ℓi ª . j3 are distinct.i3 . by the problem’s condition. we have w i ¡ ℓi ¨ wi ℓi n . 3 and 4. if a player lost all his games against the others in the company then call him the local loser of the company . the total number of local champions of all k -companies is k¡1 ª n ¢ ô wi . Actually. the number of local losers is less than or equal to the number of local champions. as well as in the left-hand part. the total number of local losers of the k -companies is Now apply this for k 2. Since the sum in each pair is 0. To be the local champion of a k -company. k¡1 i 1 Similarly. and hence 2 (4) 0. and the other k ¡ 1 members must from those whom he defeated. call a player the local champion of the company if he defeated all other members of the company. and the whole sum can be split into such pairs. By the condition of the problem. and the k -element subsets will be called as k -companies . . Since every game has a winner and a loser. So we have 2 2 i 1 i 1 n ô £¢ i 1 wi 2 ª ¡ ¢ ª« ℓi 2 0. (3) Now. so is S2 ÔT Õ. Solution 2. Thus the desired equality (2) rewrites as S1 ÔT Õ ô i1 . we call the subsets of players as companies. Similarly to the first solution. (5) In every 4-company. note that εij ¡εji. In every 3-company. then this company has a local champion as well. Clearly. Hence. j2 . if all the numbers j1 . In any company of the players. the ith player is the local champion ¢ be chosen ª wi of k -companies. j1 . j3 Ù is contained in exactly one 4-tuple. Then the same holds for the total numbers of local .33 We show that S2 ÔT Õ 0 and hence S ÔT Õ S1 ÔT Õ for each tournament. so the equality is established. whenever a 4-company has a local loser. there are no other terms in both parts.i4 S1 ÔTi1 i2 i3 i4 Õ. The ith player won against wi other player.i2 . the total number of local champions and n ¢ ª n ¢ ª ô ô wi ℓi local losers in the 3-companies is the same. Suppose that k is some positive integer. Similarly. he must be a member of the company. Obviously every company has at most one local champion and at most one local loser. then the set Øi. Therefore. hence the term εij1 εij2 εij3 appears in the right-hand part of (3) exactly once. Therefore. j2 . either the players defeated one another in a cycle or the company has both a local champion and a local loser.

and thus w1 and y « ª ¢ ª ℓi . . (5) and (6). © Ôw i ¡ ℓ i Õ 3 24 « £ n ¢ ª ¢ ª ô wi ℓi i 1 ÐÓÓÓÓÓÓÓÓÓÑÓÓÓÓÓÓÓÓÓÒ 0 i 1 3 ¡ 3  24 « £ n ¢ ª ¢ ª ô wi ℓi i 1 ÐÓÓÓÓÓÓÓÓÓÑÓÓÓÓÓÓÓÓÓÒ 0 2 ¡ 2 ¡ 3Ôn ¡ 1Õ ¡ 4 2 n ¡ ¨ô i 1 ÐÓÓÓÓÓÑÓÓÓÓÓÒ 0 w i ¡ ℓi © 0. Since every player played n ¡ 1 games. ℓi .34 n ¢ ª ô wi i 1 n ô n ô i 1 ¢ ª champions and local losers in all 4-companies. so £¢ 3 0. Hence. It is easy check that Ôx ¡ yÕ £ « ¢ ª ¢ ª 3 24 x 3 ¡ y 3   24 £ « ¢ ª ¢ ª x 2 ¡ y 2 ¡ 3Ôx   yÕ2 ¡ 4 Ôx ¡ yÕ. we have Ôwi ¡ ℓiÕ3 Then n ô £ ¢ 24 wi 3 ¡ ℓi 3   24 £ ¢ wi 2 « ª ¢ ª ¡ ℓi 2 ¡ 3Ôn ¡ 1Õ2 ¡ 4     ¨¡ w i ¡ ℓi .   ¨ Apply this identity to x wi   ℓi n ¡ 1. 3 (6) i 1 wi 3 ª ¡ ¢ ª« ℓi 3 Now we establish the problem statement (1) as a linear combination of (4).

thus establishing (ii) Bi and (iii). in particular. Then in As¡1 we have Bs¡1 Ws¡1 . tÙ. it is an m-w-string Ai¡1 or a m-b-string if it is white or black.) The process stops immediately after the step when a first isolated white pearl appears. Obviously. If there are some strings of equal lengths.or 2-b-strings. Therefore. f . 4. We will focus on three moments of the process: (a) the first stage As when the first 1-string (no matter black or white) appears. Case 1. a string of b black pearls and a string of w white pearls. then the strings of 8 white. one cuts each chosen string into two parts differing in length by at most one. respectively. Prove that at this stage. 2Õ. this is true when t . and there are 2s¡1 black and 2s¡1 white strings.35 C6. We show by induction on i minØs. 4 white and 4 black pearls are cut into the parts Ô4. if there are less than k strings longer than 1. we step. There are two strings of pearls. Then k first ones (if they consist of more than one pearl) are chosen. strings of 8. there will still exist a string of at least two black pearls. so Now. (For instance. if i minØs. (Canada) Solution 1. then from ws¡1 bs¡1 we see that a For the numbers s. a 1-w-string appears. in the sth step each string is cut into two parts. hence all were cut since no 1-string. Denote the situation after the ith step by Ai . then one chooses all of them. if a 1-b-string appears in this step. bs¡1 ws¡1 1 as s ¡ 1 minØs. and Ai is the ith step. let Bi and bi be the lengths of the longest and the shortest b-strings in Ai . if there are strings of 5. Ô2. 2 black pearls. if. which means b 2 ¤ 2s¡1 w . One cuts these strings in some steps by the following rules. all Ô 1Õ-strings Next. respectively. Suppose that s t or f t   1 (and hence s t   1). Moreover. all strings contain 2 pearls. and let Wi and wi be the same for w-strings. hence all these numbers equal 2. This means that in As¡1 . and (iii) bi wi . there is no 1-w-string as well as no Ô 2Õ-b-string. That is. Hence we have s f ¡ 1 and thus s t. The base case i 0 is obvious. . For the induction step. Ai 1 with i t ¡ 1. (ii) Next. two cases are possible. in addition. 2 Bs¡1 Ws¡1 bs¡1 ws¡1 1. next. observe that in each step Ai there are not more than k strings at all. i s. Given a positive integer k and other two integers b w 1. 5 black. (b) the first stage At where the total number of strings is greater than k (if such moment does not appear then we put t ). Ô3. Now. This contradicts the problem conditions. 4. 4. have s f . and (c) the first stage Af when all black pearls are isolated. Now. tÙ that (i) the situation Ai contains exactly 2i black and 2i white strings. all b-strings from Af ¡1 become single pearls in the f th of them are 1. hence A0 is the initial situation. In each step: (i) The strings are ordered by their lengths in a non-increasing order. 4Õ. We continue the process until every string consists of a single pearl. then the white ones precede the black ones. we have ÖBi¡1 ß2× ÖWi¡1 ß2× Wi and bi Øbi¡1 ß2Ù Øwi¡1 ß2Ù wi. each string is cut. It is sufficient to prove that in Af ¡1 (or earlier). t. then there were all the strings were cut in this step. (ii) Bi Wi . thus the claim (i) follows from the induction hypothesis. We call a string containing m pearls an m-string. tÙ then in the ith step. 2Õ. The length of a string is the number of pearls on it. if s f . then in As¡1 . We start with some easy properties of the situations under consideration. 3 white pearls and k 4. 2Õ and Ô2.

Hence. This is exactly what we needed. so at each stage there exists a Ô 2Õ-b-string. besides a fixed u-b-string. assume that in Ai there exist a u-b-string. three subcases may occur. then the problem statement is also valid. Subcase 1a. we have v ½ Øv ß2Ù u. We can assume that v is the length of the shortest w-string in Ai . In this solution. two subcases may occur. Further. a v -w-string (2 u v 2u) and a set S of k ¡ 1 other strings larger than v ß2 (and hence larger than 1). so the condition (ii) holds for Ai 1 . while not all black pearls become single. and (i) is valid again. Case 2. We claim that at every stage. First. in Af ¡1 . there are a u-b-string and a v -w-string with u v . Therefore. not more than 2t of them being black. On the other hand. and at least one 2-b-string is cut in the f th step. so the number of w-strings in At 1 is at least 2t   Ôk ¡ 2t Õ k . Suppose that some u-b-string is not cut. Actually. Since the number of w-strings does not decrease in our process. or both some u-b-string and some v -w-string are cut. We use the same notations as introduced in the first paragraph of the previous solution. Case 2. Solution 2. as desired. for A0 the condition (i) holds since b w . Two cases are possible. so. we prove the claim by induction on the number of the stage. and the condition (i) is valid. Now assume that t   1 s and t   2 f . Furthermore. and ÖÔv   1Õß2× vß2. w 2t . at most k ¡ 1 white strings are cut in this step. in the Ôi   1Õst step. Now. from u v we get Öuß2× Øvß2Ù. Then in Ai 1 .36 1-w-string appears as well. all being larger than 1. and some v -w-string is cut. we have either a u-b-string and a Ô v Õ-w-string (and (i) is valid). The latter one results in a Øv ß2Ù-w-string. Suppose that either no u-b-string is cut. Now. Then in At we have exactly 2t white and 2t black strings. In the Ôi   1Õst step. hence W should be a single pearl. then the condition (i) is satisfied since we have a u½ -b-string and a v -w-string in Ai 1 . exactly k strings are cut. in both cases (i) and (ii) we have u 1. Subcase 2a. Subcase 1b. Conversely. and we have u½ v u 2u½. Otherwise. and for the last stage it is exactly what we need. So. hence there exists a w-string W which is not cut in the f th step. since a 2-b-string is cut. Obviously. since we are not at the final stage. and 2t 1 k 2t (the latter holds since 2t is the total number of strings in At¡1 ). . all Ô 2Õ-w-strings should also be cut in the f th step. Assume that in Ai . Now. in each case in the sth step a 1-w-string appears. In the latter case. we have v 2. and prove it for Ai 1 . and no v -w-string is cut (and hence all the strings which are cut are longer than v ). we notice that this statement implies the problem statement. some u-b-string is cut. Finally. If u½ Öuß2× v . Now. hence providing at least k ¡ 1 of Ô ÖÔv   1Õß2×Õ-strings. we can put v½ v. or we have a Öuß2×-b-string and a Øv ß2Ù-w-string. Comment. or (ii) 2 u v 2u. we used the condition b w only to avoid the case b if a number b w is not a power of 2. in the Ôt   1Õst step. and there also exist k ¡ 1 of Ô v ß2Õ-strings other than considered above. all b-strings are not larger than 2. in Af ¡1 we have at least k white strings as well. Case 1. there exist a u-b-string and a v -w-string such that either (i) u v 1. notice that the inequality u v 2 implies u½ 2. we suppose that the statement holds for Ai . other k ¡ 1 of Ô v   1Õ-strings should be cut in the Ôi   1Õst step.

then all black and white Ô 2Õ-strings should be cut in the Ôi   1Õst step. let us include into S ½ the largest of the strings obtained from s. If. Subcase 2c. This is impossible. Now. or not. To reach (ii). so each one results in a Ô v ß2Õstring. In this case. as desired. it remains to check that in Ai 1 . if u½ ½ ½ ½ v u 2u . we have u contrary. otherwise we include s itself into S ½ . All k ¡ 1 strings in S ½ are greater than v ß2 v ½ . to the fulfilled. u½ 1 (and hence u 2). Therefore in Ai 1 there exists a Öuß2×-b-string Öuß2× Øvß2Ù v½ then the condition (i) is together with a Øv ß2Ù-w-string. Then all k strings which are cut have the length v . we get 2 u½ v ½ 2u½ . In the remaining case. Next.and v -strings. Namely. and k ¡ 1 strings in S (k   1 strings altogether). all u-b-strings are cut. In the former case. there exists a set S ½ of k ¡ 1 other strings larger than v ½ ß2. Otherwise. . a v-w-string. each s È S was either cut in the Ôi   1Õst step. Hence. there exist k k ¡ 1 of Ô v ß2Õ-strings other than the considered u.37 Subcase 2b. and the condition (ii) is satisfied. hence our v -w-string is cut. and among these strings there are at least a u-b-string. This means that all Ô uÕ-strings are cut as well. we show that u½ 2. Hence in Ai 1 . These will be exactly the strings obtained from the elements of S . suppose that no v -w-string is cut (and therefore no u-b-string is cut as u v ).

. 0 ai ni ¡ 1Ù. U V ÞW . . . . . Let V ØL1 . Then s 1  k ô Ôni ¡ 1Õ. Ls . . and that it is parallel to the ith coordinate axis otherwise. (Germany) Solution 1. Assume that the grid N is covered by subgrids L1 . 1 j ni ¡ 1Ù be some set of s½ elements. nk be positive integers. let n pα 1 . vÕ È E for some L È W Ù. pk be its prime factorization. Let n1 . . . Further. there will be at most ni ¡ 1 maximal subgrids orthogonal to this axis. Ps be arithmetic progressions of integers. First. and then we show how to apply it to finish the solution. . Denote W ½ f ÔW Õ. we prove the key lemma. . . (1) where I Øi1 . . U ½ V ½ ÞW ½ . . E Õ as follows. . . . By an n1 ¢ n2 ¢ ¤ ¤ ¤ ¢ nk grid we mean the set N ØÔa1 . . the elements of N will be referred to as points. For each subset W V . and xij È Ö0. αk 1 Denote by n the least common multiple of steps of these progressions. . . Construct a bipartite graph G ÔV V ½. Ls (this means N i 1 Li ) so that (ii½ ) each subgrid contains a point which is not covered by other subgrids. the conditions for existence of such expansion are provided by Hall’s lemma on matchings. ak Õ : ai È Z. Assume to the contrary that s s½ . we say that a subgrid (1) is orthogonal to the ith coordinate axis if i È I . nij ¡ 1× (1 j t) are fixed integer numbers. there exists a subgrid Li orthogonal to this axis. äs Lemma. The idea of the proof is the following.38 C7. . . bit xit Ù. . . (iii) for each coordinate axis. we define a subgrid as a subset of the form L ØÔb1 . consider the set W V containing the maximal number of elements such that W f ÔW Õ . Notice that f ÔV Õ V ½ by (iii). . In this grid. we will follow this direction. . Surely. LsÙ. Now. the following conditions being satisfied: (i) each integer belongs to at least one of them. Let P1 . L2 . . . denote f ÔW Õ Øv È V ½ : ÔL. . . bk Õ È N : bi 1 xi1 . . . . Prove that s 1  k ô i 1 αi Ôpi ¡ 1Õ. . Further. let the edge ÔLm . Imagine that we expand each subgrid to some maximal subgrid so that for the ith axis. if there is no such set then we set W ∅. Our aim is to find a point that is not i Ôni ¡ 1Õ covered by L1 . although we will apply Hall’s lemma to some subgraph instead of the whole graph. . So. i 1 Proof. and let V ½ Øvij : 1 i s. it Ù is an arbitrary nonempty set of indices. (ii) each progression contains a number which does not belong to other progressions. vij Õ appear iff Lm is orthogonal to the ith axis. Then the desired point can be found easily: its ith coordinate should be that not covered by the maximal subgrids orthogonal to the ith axis. . . . . .

aℓ . . . Now. . . Now we claim that point a Ôa1 .39 By our assumption and the Lemma condition. for each 1 i k there exists an index j ÔiÕ such that pα dj ÔiÕ . . f ÔV Õ V½ V . This contradicts the maximality of W . and that for each axis there exists a subgrid orthogonal to this axis. . . . . . On the other hand. and the condition (ii½ ) follows directly from (ii). point a cannot lie in some L È U . Obviously. rkαk Õ. . Then for an arbitrary q½ . The Lemma is proved. ds Õ. . Note §that since i § n l. we can assume that U ½ Øvij : 1 i ℓÙ. Since there are ni ¡ 1 vertices of the form vij . we get f ÔX Õ U ½ X (so G½ satisfies the conditions of Hall’s lemma). and hence our graph contains some edge ÔL. let mi be the residue of m modulo pα For each 0 i . n m ¡ m . In this situation. Hence the Lemma provides exactly the estimate we need. the sequences r ÔmÕ do not repeat. . . . we say that L È U corresponds to the ith axis. there exists a point b Ôb1 . r½ Õ we have Ôr11 . and write g ÔLÕ i. otherwise the problem statement is trivial. We claim that for every X U . . we get that for each 1 i ℓ. we are ready to present the desired point. . Choose such number for every 1 i ℓ. hence W V and U ∅.Ôd1 . Since W V . Moreover. Finally. in Lj ÔiÕ . and let mi riαi . . Permuting the coordinates. . . . consider any subgrid L È U with g ÔLÕ i. ri1 be the base pi representation of mi (possibly. Thus. .t r½ for all t γi Ù which means that Lj is a subgrid Hence Lj ØÔr11 i. bℓ 1 . not more than ni ¡ 1 subgrids correspond to the ith axis. as desired. b2 .c. all the coordinates corresponding to pi are fixed. . pk . But the points a and b differ only at first ℓ coordinates. Consider some q È Pj and let r Ôq Õ ½ . . W ½ Øvij : ℓ 1 i k Ù. . which follows pα m ¡ m½ for all 1 i k. . . . this means that r is a bijection between Ø0. On the other hand. . we have W f ÔW Õ . bk Õ È N ÞÔ LÈW LÕ.t ri.t kα containing r Ôq Õ. . Hence r ÔmÕ lies in a p pk ¢ ¤ ¤ ¤ ¢ pk grid N . suppose that a È L for some L È W . § q ¡ q ½ for each 1 i k pγ ri. . for every 1 i ℓ. we can assign to each L È U some vertex vij È U ½ so that to distinct elements of U . applying Hall’s lemma. then we have W X W   X f ÔW Õ   f ÔX Õ U½ f ÔW Õ Ô f ÔX Õ U ½Õ f ÔW XÕ . . . since N n. distinct vertices of U ½ are assigned. Now we will show that for each 1 i s. so L should be orthogonal to at least one of the first ℓ axes. . r½ Õ È N : ri. there exists a number 0 ai ni ¡ 1 which is not contained in a set ØcL : g ÔLÕ iÙ. so it is i k q½ § orthogonal to all of their axes. setting rÔq½Õ Ôr11 kαk α1 times αk times § § ½ for all t γ . . hence contradicting the Lemma condition. we put into correspondence to m the sequence r ÔmÕ Ôr11 . the set Li Ør ÔmÕ : m È Pi Ù is a subgrid. Let dj be the step of the progression Pj . so if X f ÔX Õ U ½ for some X U . Surely. . . i m n ¡ 1 and 1 i k . This means exactly that L is orthogonal to the ith axis. if r Ôm i i §Õ ½ § consequently. γk 1 Consider some 1 j s and let dj pγ 1 . Consider the induced subgraph G½ of G on the vertices U U ½ . these subgrids cover N . and hence all its elements have the same ith coordinate cL . since all the points in L have g ÔLÕth coordinate cL agÔLÕ . n ¡ 1Ù. rkαk Õ. . n ¡ 1Ù and N . .m. r21 . 1 ¢ ¤ ¤ ¤ ¢ p1 ¢ ¤ ¤ ¤ ¢ ÐÓÓÓÓÓÓÑÓÓÓÓÓÓÒ ÐÓÓÓÓÓÓÑÓÓÓÓÓÓÒ i § i § r Ôm½ Õ then pα mi ¡ m½i .t È Pj i i ½ . vij Õ for i ℓ. Surely. So. . . . . r1α1 . . recall that b Ê L. Since there are at most ni ¡ 1 such subgrids. . bk Õ is not covered. We i assume that n 1. . . Ð Now we turn to the problem. . with some leading zeroes). It contradicts the definition of W ½ . Actually. . when m runs over the set Ø0.

40
Comment 1. The estimate in the problem is sharp for every n. One of the possible examples is the following one. For each 1 i k, 0 j αi ¡ 1, 1 k p ¡ 1, let Pi,j,k kpj i

  pij  1Z,

and add the progression P0 nZ. One can easily check that this set satisfies all the problem conditions. There also exist other examples. On the other hand, the estimate can be adjusted in the following sense. For every 1 i k, let 0 αi0 , αi1 , . . . , αihi be all the numbers of the form ordpi Ôdj Õ in an increasing order (we delete the repeating occurences of a number, and add a number 0 αi0 if it does not occur). Then, repeating the arguments from the solution one can obtain that s 1 
hi k ô ô

Ôpα ¡α ¡ ¡ 1Õ.
j j 1

i 1j 1

Note that pα ¡ 1 αÔp ¡ 1Õ, and the equality is achieved only for α 1. Hence, for reaching the j for all i, j . In other words, for each minimal number of the progressions, one should have αi,j 1 j αi , there should be an index t such that ordpi Ôdt Õ j .

Solution 2. We start with introducing some notation. For positive integer r , we denote Ör× Ø1, 2, . . . , rÙ. Next, we say that a set of progressions P ØP1 , . . . , PsÙ cover Z if each integer belongs to some of them; we say that this covering is minimal if no proper subset of P covers Z. Obviously, each covering contains a minimal subcovering. i s, let di be the step of Next, for a minimal covering ØP1 , . . . , Ps Ù and for every 1 progression Pi , and hi be some number which is contained in Pi but in none of the other progressions. We assume that n 1, otherwise the problem is trivial. This implies di 1, otherwise the progression Pi covers all the numbers, and n 1. We will prove a more general statement, namely the following αk 1 Claim. Assume that the progressions P1 , . . . , Ps and number n pα 1 are chosen as 1 . . . pk in the problem statement. Moreover, choose some nonempty set of indices I Øi1 , . . . , it Ù Ök × and some positive integer βi αi for every i È I . Consider the set of indices T Then T j:1 j
i ¡βi  1 § s, and pα dj for some i È I . i

§

µ

ô
i I

Observe that the Claim for I Ök× and βi αi implies the problem statement, since the left-hand side in (2) is not greater than s. Hence, it suffices to prove the Claim. 1. First, we prove the Claim assuming that all dj ’s are prime numbers. If for some 1 i k we have at least pi progressions with the step pi , then they do not intersect and hence cover all the integers; it means that there are no other progressions, and n pi ; the Claim is trivial in this case. Now assume that for every 1 i k , there are not more than pi ¡ 1 progressions with step pi ; each such progression covers the numbers with a fixed residue modulo pi , therefore there exists a residue qi mod pi which is not touched by these progressions. By the Chinese Remainder Theorem, there exists a number q such that q qi Ômod pi Õ for all 1 i k ; this number cannot be covered by any progression with step pi , hence it is not covered at all. A contradiction.

È

βi Ôpi ¡ 1Õ.

(2)

41 2. Now, we assume that the general Claim is not valid, and hence we consider a counterexample ØP1 , . . . , Ps Ù for the Claim; we can choose it to be minimal in the following sense: ­ the number n is minimal possible among all the counterexamples; ­ the sum i di is minimal possible among all the counterexamples having the chosen value of n. As was mentioned above, not all numbers di are primes; hence we can assume that d1 is § d1 ½ having the step d½ , and § 1. Consider a progression P1 composite, say p1 d1 and d½1 1 p1 containing P1 . We will focus on two coverings constructed as follows. ½ , P2 , . . . , Ps cover Z, though this covering in not necessarily (i) Surely, the progressions P1 ½ È P ½ since h1 is not covered minimal. So, choose some minimal subcovering P ½ in it; surely P1 ½ , P2 , . . . , Ps½ Ù for some s½ s. Furthermore, the by P2 , . . . , Ps , so we may assume that P ½ ØP1 period of the covering P ½ can appear to be less than n; so we denote this period by n½
αk ¡σk 1 ¡σ1 pα . . . pk 1

¨   l.c.m. d½1 , d2 , . . . , ds½ .

½ , otherwise hj would not be covered by P . Observe that for each Pj Ê P ½ , we have hj È P1 ½ (1 i s) is also a (ii) On the other hand, each nonempty set of the form Ri Pi P1 ½ ½ progression with a step ri l.c.m.Ôdi , d1Õ, and such sets cover P1 . Scaling these progressions ½ with the ratio 1ßd1 , we obtain the progressions Qi with steps qi ri ßd½1 which cover Z. Now we choose a minimal subcovering Q of this covering; again we should have Q1 È Q by the reasons of h1 . Now, denote the period of Q by
n¾ l.c.m.Øqi : Qi

È QÙ

l.c.m.Øri : Qi d½1

È QÙ

γk 1 pγ 1 . . . pk . d½1

½ , then the image of hj under the scaling can be covered by Qj only; so, in Note that if hj È P1 this case we have Qj È Q. Our aim is to find the desired number of progressions in coverings P and Q. First, we have n n½ , and the sum of the steps in P ½ is less than that in P ; hence the Claim is valid for P ½ . We apply it to the set of indices I ½ Øi È I : βi σi Ù and the exponents βi½ βi ¡ σi ; hence the set under consideration is
T½ j:1 j

Ôα ¡σ Õ¡β ½  1 s½ , and p i i i
i

i βi 1 pα i

¡   § § dj for some i È I ½

ô
i I

µ

T

Ös½×,

and we obtain that T

Ös½×

ô
i I½

where ÔxÕ  maxØx, 0Ù; the latter equality holds as for i Ê I ½ we have βi σi . Observe that x Ôx ¡ y Õ    minØx, y Ù for all x, y . So, if we find at least G indices in T T T
ô
i I

È

Ôβi ¡ σiÕÔpi ¡ 1Õ

È

Ôβi ¡ σi Õ Ôpi ¡ 1Õ,

Øs½   1, . . . , sÙ, then we would have ô  ¨ Ös½×   T Øs½   1, . . . , sÙ 1   Ôβi ¡ σiÕ    minØβi , σiÙ Ôpi ¡ 1Õ
i I

È

minØβi , σi ÙÔpi ¡ 1Õ

È

ô
i I

È

βi Ôpi ¡ 1Õ,

thus leading to a contradiction with the choice of P . We will find those indices among the indices of progressions in Q.

42

½ i § Øi È I : σi 0Ù and consider 3. Now denote I ¾ some i È I ¾ ; then pα i § n . On the § i § other hand, there exists an index j ÔiÕ such that pα dj ÔiÕ ; this means that dj ÔiÕ § n½ and hence i Pj ÔiÕ cannot appear in P ½ , so j ÔiÕ s½ . Moreover, § we have observed before that in this case ½ § ¾ hj ÔiÕ È P1 , hence Qj ÔiÕ È Q. This means that q αi for each i È I ¾ (recall j ÔiÕ n , therefore γi § § here that qi ri ßd½1 and hence dj ÔiÕ § rj ÔiÕ § d½1 n¾ ). τk ¾ pγ1 ¡τ1 . . . pγi ¡τi . Now, if i È I ¾, then for every β the condition 1 Let d½1 pτ 1 1 . . . pk . Then n k § Ôγ ¡τ Õ¡β  1 § q is equivalent to pαi ¡β  1 § § rj . pi i i j i Note that n¾ nßd½1 n, hence we can apply the Claim to the covering Q. We perform this with the set of indices I ¾ and the exponents βi¾ minØβi , σi Ù 0. So, the set under consideration is

§

È Q, and piÔγ ¡τ Õ¡minØβ ,σ Ù 1 § qj for some i È I ¾ µ § α ¡minØβ ,σ Ù 1 § ¾ j : Qj È Q, and pi rj for some i È I ,
j : Qj
i i i i i i i

§

µ

1   G. Finally, we claim that T ¾ T Ø1Ù Øs½   1, . . . , sÙ ; then we and we obtain T ¾ will obtain T Øs½   1, . . . , sÙ G, which is exactly what we need. To prove this, consider §any j È T ¾ . Observe first that αi ¡ minØβi , §σi Ù   1 αi ¡ σi τi , α ¡minØβi ,σi Ù 1 § α ¡minØβi ,σi Ù 1 § rj l.c.m.Ôd½1 , dj Õ we have pi i dj , which means that hence from pi i j È T . Next, the exponent of pi in dj is greater than that in n½ , which means that Pj Ê P ½ . This may appear only if j 1 or j s½ , as desired. This completes the proof.
Comment 2. A grid analogue of the Claim is also valid. It reads as following. Claim. Assume that the grid N is covered by subgrids L1 , L2 , . . . , Ls so that (ii½ ) each subgrid contains a point which is not covered by other subgrids; (iii) for each coordinate axis, there exists a subgrid Li orthogonal to this axis. Choose some set of indices I Øi1 , . . . , it Ù Ök×, and consider the set of indices T Then T

 

¨

Øj : 1

j

s, and Lj is orthogonal to the ith axis for some i È I Ù . 1 
ô
i I

È

Ôni ¡ 1Õ.

This Claim may be proved almost in the same way as in Solution 1.

43 .

the quadrilateral AQP F is cyclic. From the cyclic quadrilateral AP BC we get QP A 180¥ ¡ AP B BCA γ DF B QF A. E and F . γ. there are two different cases to consider. 1 Fig. Prove that AP AQ. (United Kingdom) Solution 1. E. CA. AP AQ. Hence. Case 1 : The point P lies on the ray EF (see Fig. F the feet of the altitudes lying on BC. AB respectively. we have 180¥ . One of the intersection points of the line EF and the circumcircle is P . The lines BP and DF meet at point Q.Geometry G1. Depending on the choice of P . AF E γ . BDF AF E DF B 180¥ ¡ F DC 180¥ ¡ EF B 180¥ ¡ AF D CAF BCE DCA P BA α. respectively. Let CAB α. Since P lies on the arc AB of the circumcircle. BCA γ . so the triangle AQP is isosceles. 2 . So. 1). Q γ A α A γ P γ B F γ γ γ E F γ γ γ E γ P β α D γ C B Q D γ C Fig. The quadrilaterals BCEF and CAF D are cyclic due to the right angles at D . Then AQP 180¥ ¡ P F A We obtained that AQP QP A γ . The line EF intersects the circumcircle at two points. Hence. γ. P BD   BDF P BA   ABD   BDF γ β α and the point Q must lie on the extensions of BP and DF beyond the points P and F . ABC β and BCA γ . Let ABC be an acute triangle with D.

by BF P Q½ F B . By P B   P ½ A 2 AF P ½ 2γ 2 BCA AP P ½ A. the quadrilateral AF QP is cyclic. 3 Like in the first solution.45 Case 2 : The point P lies on the ray F E (see Fig. let P lie on the arc AB and let P ½ lie on the arc CA. In this case the point Q lies inside the segment F D. Let BP and BP ½ meet the line DF and Q and Q½ . the lines F P and F Q½ are symmetrical about AB . the lines BP and BQ½ are symmetrical about line AB . Similarly to the first case. we have AF E cyclic quadrilaterals BCEF and CAF D. 2). For arbitrary points X. Using signed angles. we have and AP AP ½ . AQ AQ½ . Then AQP AF P AF E γ AQP QP A and thus AP AQ. respectively. Ô2Õ The relations (1) and (2) together prove AP . P BA BF P AP DF B BCA γ from the   P B. 3). The triangle AQP is isosceles again. We will prove that AP AP ½ AQ AQ½ . we have QP A BCA γ DF B 180¥ ¡ AF Q. ABP ½ Ô1Õ Due to AP P ½ A. Let P and P ½ be the two points where the line EF meets the circumcircle. denote by XY the central angle of the arc XY . Y on the circumcircle. Similarly. the two possible configurations can be handled simultaneously. AP AQ½ and AP ½ AP ½ AQ. QP A. without investigating the possible locations of P and Q. Solution 2. Hence. It follows that also the points P and P ½ are symmetrical to Q½ and Q. Comment. Q A P F γ γ γ Q′ E P′ γ D C B Fig. Therefore. respectively (see Fig.

that is QP QC . Let P be an arbitrary point inside both the circumcircle ω of the triangle ABC and the angle ASC . We claim that SP SC implies MK ML. thus establishing the problem statement. CA PA Denote by E the foot of the bisector of angle B in triangle ABC . △P CB we get PM KM PA CA △P CA and △P LM From the similar triangles △P KM LM CB and . ML is equivalent to L Ω C C C C K P P P S A E B M Fig. As in the previous solution. CA P B CB PB . The tangent to the circumcircle at C meets line AB at S . Hence. Prove that SC SP if and only if MK ML. L. Recall that the locus of CA XA points X for which is the Apollonius circle Ω with the center Q on the line AB . the points K . we get PM PB LM KM Hence. . the point S lies on AB as well as on the perpendicular bisector of CE and therefore coincides with Q. respectively. L. we assume that S lies on the ray AB . 2). point E lying on the segment SP (see Fig. 1 Now we prove that S Q. We have CES CAE   ACE BCS   ECB ECS . Point P lies inside triangle ABC . XB CB and this circle passes through C and E . 1. BP . (Poland) Solution 1. Let E and F be the points of intersection of the line SP with ω . so SC SE . Multiplying these two equalities. M defined as in the problem. Solution 2. M . the relation MK CB P A ¤ . CP meet the circumcircle of ABC again at points K . We assume that CA CB . we have MK ML if and only if P lies on Ω.46 G2. Hence. so point S lies on the ray AB . Lines AP .

and introduce the points E and F as above. assume that MK ML. and hence △P SA △BSP . L L′ C F ′ ′ ′ K K K′′ ′ ′ ′ P P P′′ ω P K E A B S M′ M Fig. there exists a point P ½ lying on the segment EF such that SP ½ SC (see Fig. or (2) Ǒ and hence MK MEK (1) On the other hand. from SP C SCP we have EC   MF MF EM . EC   EM . MF  FL ML. We are left to prove the converse. So. ME   EK Ǒ L From (1) and (2) we get MF The claim is proved. 2. 3). hence.47 F L K ω P P P M E A B S C Fig. Since 2 BP S SA SP . 2 We have SP 2 BP S SC 2 SA ¤ SB . Then SB SP BE   LF and 2 SAP BE   EK we have LF EK. 3 . We have SC 2 SE ¤ SF . so SAP .

BP ½ . On the other hand.48 Assume that P P ½ . Ǒ ¡ ME MF Ǒ L ¡ MF To the contrary. in our case. suppose first that ME MF . Similarly. Let the lines AP ½ . the inequality EK F L implies 2 SP K EK   AF F L   AF 2 ABL. that is. the points P and P ½ coincide and hence SP SP ½ SC . which implies SC SP . Now we claim that ME MF and EK F L. we have MF ½ F L½ M ½ EK ½ MEK Ǒ Ǒ Ǒ Ǒ Ǒ Ǒ . Then △SBP △SP A and SP 2 SB ¤ SA½ SB ¤ SA SC 2 . CP ½ meet ω again at points K ½ . L½ . one can prove that the inequality ME ME MF and therefore 2 SCM EC   ME EC   MF 2 SP C . we get Similarly. and the line SP meets ω at E and F . if point P lies on the segment EP then we get MF L MEK Therefore. MK ML SP SC . this point lies ½ on the segment SA (see Fig. So. the inequality ME MF implies 2 SCM EC   ME EC   MF 2 SP C and hence SP SC . ½ Ǒ Ǒ which is impossible. Therefore. Now. 4). M ½ respectively. Solution 3. As in the previous solutions we assume that CA CB . We present a different proof of the converse direction. hence SP A 180¥ ¡ SP K 180¥ ¡ ABL SBP. if P ½ lies on the segment P F then by the first part of the solution we have ½ F L½ M ½ EK ½ . Ǒ Ǒ From ML MK we get MEK MF L. SP SC which contradicts SP SC . Now. . On the other hand. MF is also impossible. 4 Consider the point A½ on the ray SA for which SP A½ SBP . then EK MEK F L. therefore M L M Ǒ Ǒ which contradicts MK MF L MEK ML. L F P C K K K E A A′ B S M ω Fig.

49 .

. . we get that P lies inside the circumcircle of triangle Xi Ai 1 Xi 1 for some i. Denote Pn 1 P1 . Point P inside this polygon is chosen so that its projections P1 .. . Consider the circumcircles ω and Ω of triangles Pi Ai 1 Pi 1 and Xi Ai 1 Xi 1 respectively (see Fig. . Otherwise Q lies inside the polygon X1 X2 .50 G3. . . . 2 . Xn on sides A1 A2 . An 1 A1 . . Let point Q lies inside A1 A2 . An be a convex polygon. Since the n quadrilateral QXi Ai 1 Xi 1 is convex. . this means exactly that Q is contained the circumcircle Ð of △Xi Ai 1 Xi 1 . An . . If Q lies in one of the triangles X1 A2 X2 . hence Pi Pi 1 QED. X n X1 X 1 X2 . . Lemma. . Then we get 2r Ai 1 P 2R (since P lies inside Ω). let r and R be their radii. Xn (see Fig. Xn 1 X1 . the claim is obvious. Pn onto lines A1 A2 . An A1 respectively lie on the sides of the polygon. . .. . . 2). 1 Fig. Xn A1 X1 . . . 2r sin Pi Ai 1 Pi 1 2R sin Xi Ai 1 Xi 1 Xi Xi 1 . Then we have Ô X1 A2 X2   X1 QX2 Õ   ¤ ¤ ¤   Ô Xn A1 X1   Xn QX1 Õ Ô X1 A1X2   ¤ ¤ ¤   XnA1 X1Õ   Ô X1 QX2   ¤ ¤ ¤   Xn QX1Õ Ôn ¡ 2Õπ   2π nπ. πn hence there exists an index i such that Xi Ai 1 Xi 1   Xi QXi 1 π . . . . . Applying lemma.. 1). . (Armenia) Solution 1. A4 X4 X3 A3 P Q X2 X5 Pi Xi ω Pi+1 Ω A5 Xi+1 A1 X1 A2 Ai+1 Fig. Now we turn to the solution. Xn A1 X1 . . . Proof. max P1 P2 Pn P1 ¶ 1. as desired. An A1 respectively. . . Let A1 A2 . .. . . Prove that for arbitrary points X1 . Then it is contained in at least one of the circumcircles of triangles X1 A2 X2 .

since γi and βi 1 are acute angles (see Fig. and in this case the statement is proved. We denote βi Ai Pi Pi¡1 and γi Ai 1 Pi Pi 1 for all 1 i n. the sines theorem applied to triangle P Pi Pi 1 provides sin π ¡ βi ¨ 1 2   π sin 2 ¡ γi cos βi 1 . 4 . . Therefore. we assume that all indices of points are considered modulo n. (2) Xi Xi 1 cos αi Pi Pi 1 (1) cos γ1 cos γ2 ¤ ¤ ¤ cos γn P Pi P Pi 1   cos β1 cos β2 ¤ ¤ ¤ cos βn . We will prove a bit stronger inequality. 3). see Fig. Suppose that for some 1 i n. . . Xi Xi 1 cos αi Pi Pi 1 . the only case left is when point Xi lies on segment Pi Ai 1 for all 1 i n (the case when each Xi lies on segment Ai Pi is completely analogous). Then the projection of the segment Xi Xi 1 onto the line Pi Pi 1 contains segment Pi Pi 1 . we get Xi 1 Pi 1 cos βi 1 . As in Solution 1. cos αn max 1. 3 Fig. Now. Let Yi and Yi½ 1 be the projections of Xi and Xi 1 onto Pi Pi 1 . assume to the contrary that the inequality holds for every 1 i n. . Xi+1 Yi′+1 Pi+1 βi+1 αi Xi+1 P Pi+1 βi+1 γi Xi Pi Ai+1 Pi−1 Xi−1 P γi Pi Yi Xi αi Ai+1 cos β2 cos β3 ¤ cos γ1 cos γ2 cos β1 ¤ ¤ ¤ cos γ n βi Ai Fig. cos γi Multiplying these equalities we get 1 which contradicts (2). 4). namely ¶ X1 X2 Xn X1 cos α1 . Then inequality (1) means exactly that YiYi½ 1 Pi Pi 1 . while point Xi 1 lies on the segment Pi 1 Ai 2 . Hence. So. we have Xi Pi cos γi Multiplying these inequalities. or Pi Yi Pi 1 Yi½ 1 (again since γi and βi 1 are acute. ¨ On the other hand. P1 P2 Pn P1 where αi (1 i n) is the angle between lines Xi Xi 1 and Pi Pi 1 .51 Solution 2. 1 i n. point Xi lies on the segment Ai Pi .

and L be the foot of the bisector of angle BAC . Let I be the incenter of a triangle ABC and Γ be its circumcircle. we get TF AT as desired. B . it means that we need the relation 1 G½ F IG½ TF AT ¤ AD . AT AI CL IL . Finally. it is known that the midpoint D of arc BC is equidistant from points I . Let G½ and T be the points of intersection of segment DX with lines IF and AF . Hence we have IT A IXA EXA EKA. AD CAE we have Let the line AF intersect Γ at point K A (see Fig. so TF IL IT KE BC . we get DCB AI AC 1 DAB CAD 2 BAC .52 G4. We are to prove that G G½ . Notice that IAT DAK EAD EXD IXT . Furthermore. therefore we get CL DC . Let the line AI intersect Γ at a point D A. Finally. let G be the midpoint of the segment IF . (Hong Kong) Solution 1. hence KE BC . AC AD DC ID . A. hence the triangles DCL and DAC are similar. since BAK BK CE . X . T are concyclic. or IG½ G½ F . Let F and E be points on side BC and arc BDC respectively 1 such that BAF CAE BAC . so the points I . By the Menelaus theorem applied to triangle AIF and line DX . ID or TF AT ID . 1). X A A IL AI CL AC DC AD ID . 2 . Therefore we obtain . hence AD AD Summarizing all these equalities. Let X be the second point of intersection of line EI with Γ. C . 2 Prove that the lines DG and EI intersect on Γ. DCL Since CI is the bisector of ACL. 1 Fig. respectively. AD I T T G′ F B K D L I I I I B C D C E L Fig.

AC AE Summarizing we get ÔDI   ADÕ ¤ AI as desired. or AB ¤ AC AF ¤ AE . so point A goes to L. Then DJC 1 1 1 JCD Ô π ¡ JDC Õ ADC ABC ABI . which is the desired equality. As in the previous solution. DI AD Solution 2. point J is an excenter of triangle ABC . AD   DI Comment. For IL DI DI . gles AT D and AIE are similar. hence 2 2 2 AI AB . Let J be the point on the extension of segment AD over point D such that DJ DI DC (see Fig. T and K and note that it suffice to prove the equality TF AT Since F AD DI AD EAI and T F   AT AT T DA DI   AD AD AT AD AF . This inversion takes instance. we introduce the points X . AB ¤ AC AF ¤ AE AI AE AF AD   DI AT AD AF . so trianAB AF gles ABF and AEC are also similar. BAI JAC . therefore AD AE Next. The equality is known and can be obtained in many different ways. Moreover. or AB ¤ AC AJ ¤ AI ÔDI   AD Õ ¤ AI . we get ABF ABC AEC and BAF CAE . we get that the trianAI AT . which implies .53 AD AI Comment. we also use the relation DB DC DI . . DI   AD XDA XEA IEA. triangles ABI and AJC are similar. Hence BAC . so AJ AC On the other hand. one can consider the inversion with center D and radius DC IL AI Ǒ to the segment BC . In fact. 2).

hence quadrilateral BCT E is a parallelogram. BDT BDE   EDT BDE   BDC CDE ABC 180¥ ¡ BAT . BDC we have ADB 2ϕ ¡ β ¡ γ . Let M be the midpoint of CE . so BT is the bisector of ABC . AED 360 ¡ BCD ¡ CDE 180¥ ¡ ADE ¡ AED 2ϕ ¡ α ¡ β ¡ γ . 2 . we have 1 DMN DON DOC α. and hence ADB AT B 1 1 ABC CDE . This means that the points A. 2 2 B C B O D K M M M β 2ϕ − β − γ D γ γ A E T A α−β 2ϕ − −γ −α 2ϕ −β 2ϕ α C α + β E Solution 2. D . prove that 2 BDA CDE . then DNO 90¥ DMO .54 G5. Choose point T on ray AE such that AT CBT AT B ABT . Let CBD α. ADE γ . and hence OD OK . On the other hand. which means that point K lies on the circumcircle of triangle BCD. and the midpoint M of its diagonal CE is also the midpoint of the other diagonal BT . we have ET AT ¡ AE AB ¡ AE BC . T are concyclic. and ABC CDE . Then ¥ ¥ 180 ¡ α ¡ β . Therefore. Let ABCDE be a convex pentagon such that BC AE . then from AE BC we have Solution 1. Now. in the other case. (Ukraine) AB . B . and let O be the circumcenter of triangle BCD. we have DMN 180¥ ¡ DON α. let point K be symmetrical to D with respect to M . as desired. AB BC   AE . if points O and M lie on the same side of CD. and finally DAE B C N N N O M M D β . Next. the angles BKC and T DE are symmetrical with respect to M . On the other hand. Then OM is the perpendicular bisector of segment DK . Given that DMO 90¥ . BCD 180¥ ¡ 2ϕ   α   β . N lie on the circle with diameter OD . hence points M . and ABC CDE 2ϕ. so T DE BKC BDC . Hence we have BDC BKC . B C N N N M M M D D D O O O E E Let N be the midpoint of CD.

as desired. . since MN is a midline in triangle CDE . sinÔ2ϕ ¡ αÕ AD sinÔ2ϕ ¡ α ¡ β Õ AD sinÔ2ϕ ¡ α ¡ β Õ CD BC sin β CDß2 ND sin α .   ¨ 0 β   γ. 2 90¥ . . after multiplying AD sinÔ2ϕ ¡ α ¡ β Õ ¤ sinÔ2ϕ ¡ β ¡ γ Õ sin γ ¤ sinÔ2ϕ ¡ αÕ   sin β ¤ sinÔ2ϕ ¡ α ¡ β ¡ γ Õ cosÔγ ¡ αÕ ¡ cosÔ4ϕ ¡ 2β ¡ α ¡ γ Õ cosÔ2ϕ ¡ α ¡ 2β ¡ γ Õ ¡ cosÔ2ϕ   γ ¡ αÕ cosÔγ ¡ αÕ   cosÔ2ϕ   γ ¡ αÕ cosÔ2ϕ ¡ α ¡ 2β ¡ γ Õ   cosÔ4ϕ ¡ 2β ¡ α ¡ γ Õ cos ϕ ¤ cosÔϕ   γ ¡ αÕ cos ϕ ¤ cosÔ3ϕ ¡ 2β ¡ α ¡ γ Õ cos ϕ ¤ cosÔϕ   γ ¡ αÕ ¡ cosÔ3ϕ ¡ 2β ¡ α ¡ γ Õ cos ϕ ¤ sinÔ2ϕ ¡ β ¡ αÕ ¤ sinÔϕ ¡ β ¡ γ Õ 0. it follows that ϕ . or equivalently AD by the common denominator rewrites as AE   BC . we have MDE DMN α and NDM 2ϕ ¡ α.55 so. Now we apply the sine rule to the triangles ABD . . in both cases DMN α (see Figures). the condition AB AE   BC . BCD and MND obtaining AB AD sinÔ2ϕ ¡ β ¡ γ Õ sin γ sinÔ2ϕ ¡ α ¡ β ¡ γ Õ AE DE . CD sin α DE DE ß2 NM sinÔ2ϕ ¡ αÕ BC AD BC CD DE ¤ ¤ CD DE AD sin β ¤ sinÔ2ϕ ¡ α ¡ β ¡ γ Õ . ADE (twice). sinÔ2ϕ ¡ αÕ ¤ sinÔ2ϕ ¡ α ¡ β Õ which implies AB Hence. Next. Since 2ϕ ¡ β ¡ α 180¥ ¡ AED 180¥ and ϕ 1 ABC 2 hence BDA 2ϕ ¡ β ¡ γ ϕ 1 CDE .

Y . B C1 Z ′ ′ ′ C C C′′ C I I G6 . Finally. Prove that the incenter of triangle ABC lies inside triangle XY Z . U ½ and V . AB Õ r ½ . V ½ be the points of intersection of ω with XY and Y Z . BC Õ Since OX A½ I   r R½ . and the desired inequality is OI r ½ . hence dÔO. V Y V ½Y . Y . The vertices X . Assume the contrary. On the other hand. let U . assume that point A1 lies on segment CX . 2 . the incircle of △XY Z lies inside △ABC . AC . We may assume that O lies in an angle defined by lines IA1 . 3 and 4). Since OA½ BC . AB of an acute-angled triangle ABC . r ½ and R½ the inradii of triangles ABC . Z of an equilateral triangle XY Z lie respectively on the sides BC . AB Õ C ½ C1 r ¡ IC ½ r ½ . we will show that the incenter I of triangle ABC lies inside the incircle of triangle XY Z (and hence surely inside triangle XY Z itself). the quadrilateral IA½ OC ½ is circumscribed due to the right angles at A½ and C ½ ½ OC ½ 2 A½ IC ½ 180¥ OC ½I . we have A IC ½ A½ O . The lines IA1 . we have dÔO. In this case we will show that C 90¥ thus leading to a contradiction. R½ . Then we have R½ 2r ½ . as desired. We will prove a stronger fact. Let the incircle ω of △ABC touch sides BC . ½ X A1 O O C′ I A′ A B1 Y C A′ Fig. then one of the angles of triangle ABC is greater than 120¥ . 2 Finally. or IC ½ r ¡ r ½ . it follows that dÔA½ . Z of an equilateral triangle XY Z lie respectively on the sides BC . B1 . IC1 cut the plane into 6 acute angles. or A½ I R½ ¡ r . otherwise the claim is trivial. 1 1 ¥ see Figs. Solution 2 for G6. and analogously we get dÔO. 1 Fig. Then the incenter I should lie in one of triangles AY Z . On its circumcircle. B1 respectively. This means that IC ½ A½ O . Prove that if the incenter of triangle ABC lies outside triangle XY Z . CA. (Bulgaria) Solution 1 for G6. AC at point A1 .56 G6. Let the incircle of △ABC touch its sides BC . We denote by dÔU. CA. V W Õ the distance between point U and line V W . XY Z and the circumradius of XY Z . BXZ . 1). C1 on its border. IB1 . hence UV 120¥. Denote by O the incenter of △XY Z and by r . Note that ω intersects each of the segments XY and Y Z at two points. BC Õ R½ . Without loss of generality. B1 . Note that 60 XY Z ÔUV ¡ U ½ V ½ Õ 2 UV . hence 180¥ Ǒ (see Fig. C1 respectively. respectively (UY U ½Y . respectively. we have OI IA½   A½ O IA½   IC ½ ÔR½ ¡ rÕ   Ôr ¡ r½Õ R½ ¡ r½ r½. namely. IC1 and containing point C1 (see Fig. We assume that O I . CXY — assume that it lies in △AY Z . respectively. 2). Let A½ and C ½ be the projections of O onto lines IA1 and IC1 . AB at points A1 . The vertices X . each one containing one of the points A1 . AB of a triangle ABC .

we assume that the incenter I of △ABC lies in △AY Z . ACB 180¥ ¡ A1 IB1 120¥ . we get XM XY which implies A1 XY KY X . 4). 2 2 Y CX ¡ 180¥ ¡ Y XC ¡ XY C 120¥ . as desired. AB of a triangle ABC . we get XY C UA1 U 30¥ . From symmetry. In the same way. two cases are possible due to the order of points Y . Hence. Let ¨ point 1 Ǒ½ 1 Ǒ½ ½ A1 U A1 U UA1 U 30¥ . we have Y IK Y IL. 3). as desired. and therefore the projection M of I onto XY lies on the segment XY . Then we have Y XC   Case 1. In this case. one can prove a more general Claim. IK Ã Y K we get MIA1 A1 XY KY X MIK . one of the points K and L is in fact a tangency point B1 of ω and AC . Finally. Surely. Y ZA 180¥ ¡ Y ZX 120¥ . Moreover. we will prove that C 120¥ . hence UA 1U ¥ ¥ 180 ¡ UV 60 . since Y B1 is a tangent and Y X is a secant line to circle ω from point Y . we get A1 IK A1 IL Ô A1 IM   MIK Õ   Ô KIY   Y ILÕ 2 MIK   2 KIY 2 MIY 60¥ . hence points I and Y lie on one side of the perpendicular bisector to XY . as desired.57 ½ Ǒ Ǒ On the other hand. therefore IX IY . Let the vertices X . we have MIY 90 ¡ IY X 90 ¡ ZY X 30¥ . and α is the least angle of △XY Z . Hence. Solution for G6½ . analogously. since IX IY . On the other hand. Y . we obtain Y XC 30¥ . Assume the contrary. Y L be two tangents from point Y to ω (points K and A1 lie on one side of XY . Since IA1 Ã A1 X . Case 2. we say K L Y ). CA. hence Y CX 120¥ ¡ 30¥ 90¥ . we get V UV ½ 180¥ . As in Solution 2. since I lies in △AY Z . Therefore. Comment. 4 1 2 Y lie on the segment AB1 (see Fig. we get 120¥ IY B1   IY X B1 Y X Y XC   Y CX 30¥   Y CX . Now let point Y lie on the segment CB1 (see Fig. Z of a triangle XY Z lie respectively on the sides BC . ω intersects segment XY at two points. if Y lies on ω . Let Y K . and therefore . B1 on segment AC . Analogously. IM Ã XY . Now. and the tangency point A1 of ω and BC lies on segment CX . Then one of the angles of triangle ABC is greater than 3α ¡ 90¥ . IY X ZY X 60¥ . ¥ ¥ Next. Next. A1 IB1 60¥ . Suppose that the incenter of triangle ABC lies outside triangle XY Z . but IY X IY B1 . A A ω ′ ′ ′ V V′′ V Y V ′′ U U U′′′ U B1 ½ Ǒ UA 1V C1 I V U U U Z U′ B1 ′ ′ ′ V Y V V′′ V C1 I V Z U A1 X B ω B C C A1 X Fig. 3 Fig.

Claim. CA. and all sides of triangle XY Z are greater than 2r cot α. IM r . Comment 2. hence it does not contain I . Now. Actually. Then one of the angles of triangle ABC is greater than 2α. Z of a triangle XY Z lie respectively on the sides BC . . Suppose that the incenter of triangle ABC lies outside triangle XY Z . but here we need some additional condition. if we have XM   Y M 2r cot α. AB of a triangle ABC . this additional condition follows easily from the conditions of the original problem. Hence Y Z 2r 2r cot 60¥ . Actually. Let the vertices X . X È BC (such triangle exists since ACB 60¥ ). α is the least angle of △XY Z . then the diameter of ω parallel to Y Z is contained in △AY Z and is thus shorter than Y Z . where r is the inradius of △ABC . Y . if I È △AY Z . then surely XM Y M . 5 Fig. if BAC 120¥ . hence Y M r cot α. we have M Y I α. As in the previous solution. The estimate claimed in G6½ is sharp. Y È AC . there is a generalization for an arbitrary triangle XY Z . It intersects with the angle bisector of BAC only at point A. one can consider an equilateral triangle XY Z with Z A. XY On the other hand. Actually. The statement reads as follows. Y M since it cannot be shown in the The additional condition is needed to verify that XM original way. 6 Comment 1.58 A1 M A1 X I B Z A I C M M M Y M L(= B1 ) K (= B1 ) X Y Fig.

59 .

Such points lie on an ellipse with foci Oi and Oj . Ð Lemma 2. then the curved quadrilateral V22 V32 V33 V23 is circumscribed. Ǒ Ǒ Denote by V ij Vkj Vkℓ Viℓ the curved quadrilateral. Let εij be that arc AC of the ellipse which runs inside the half plane H (see Fig. Point B lies on the segment AC . V21 V31 V32 V22 are circumǑ Ǒ scribed. We say that this quadrilateral is circumscribed if there exists a circle touching these two segments and two arcs. (b) There is a constant uij such that r uij ¤ dÔP Õ for all such circles. (a) The locus of the centers of these circles is the angle bisector βij between hi and hj . Vkj Vkℓ and arcs Vij Vkj and Viℓ Vkℓ . 3. j 1. Ǒ Ǒ ǑǑ ǑǑ Prove that if the curved quadrilaterals V 11 V21 V22 V12 . h2 being between h1 and h3 . 2). Then part (b) also becomes trivial. V12 V22 V23 V13 .) This ellipse arc lies between the arcs γi and γj . for each such circle we have Oi P   Oj P Oi A   Oj A Oi C   Oj C Ri   Rj . Part (a) is obvious. then Oi P Ri and Oj P Rj . and γ3 connect the points A and C . For i. Proof. lying in the same half-plane. denote by Vij the point of intersection of hi and γj (see the Figure below). whose sides are the segments Vij Viℓ . for any r 0. we choose r Oi P ¡ Ri Rj ¡ Oj P 0. For every 1 i j 3. h1 V13 V22 γ3 γ2 A γ1 V12 V11 B V21 V31 C V32 V23 h2 h3 V33 Fig. (a) The locus of the centers of these circles is an ellipse arc εij with end-points A and C . Therefore. Three circular arcs γ1 . h2 . consider those circles ΩÔP. r Õ in the half-plane H which are tangent to hi and hj . Proof. (b) There is a constant vij such that r vij ¤ dÔP Õ for all such circles. Denote by Oi and Ri the center and the radius of γi . r Õ in the half-plane H which are externally tangent to γi and internally tangent to γj . too. r Õ is externally tangent to γi and internally tangent to γj if and only if Oi P Ri   r and Oj Rj ¡ r . Denote also by H the half-plane defined by AC which contains the whole configuration. 1 (Hungary) Solution. To prove part (b). γ2 . Let h1 . 2. (a) Notice that the circle ΩÔP. Now. notice that the circles which are tangent to hi and hj are homothetic with the common homothety center B (see Fig. Lemma 1. For every point P in the half-plane H . Furthermore. r Õ the circle with center P and radius r . For every 1 i j 3. Therefore. then the . the diameter of this ellipse is Ri   Rj . These arcs lie in the same half-plane defined by line AC in such a way that arc γ2 lies between the arcs γ1 and γ3 . denote by ΩÔP. denote by dÔP Õ the distance between P and line AC . and it passes through the points A and C . consider those circles ΩÔP. 3.60 G7. if some point P lies on εij . and h3 be three rays starting at B . respectively.

i 1 ¤ dÔP Õ vj.i 1 angle bisector βi.j  1 ¤ dÔP Õ as well. hence Ôρ ¡ ρiÕ2 ¡ Ôρ ¡ ρj Õ2 ÔRi   rÕ2 ¡ ÔRj ¡ rÕ2. u12 v23 . does not depend on P . Then the circle ΩÔP. let dij OiOj .61 Ω(P. and u23 v12 hold. r Õ touches γi externally and touches γj internally. The curved quadrilateral Qij if ui. (b) Let ρ AP . Ǒ Vi.j  1. vj. OiP orthogonal to AC and directed toward H . dij ¤ dÔP Õ dij v ¤ ρ Ôρj ¡ ρi Õ ¤ ρ r ÔRi   Rj Õ. so P belongs to the locus under investigation. Let P be the intersection of the To prove the opposite direction.j  1. Ð Proof.j Vi 1Ǒ . Hence. Therefore. and it is tangent to the Ð arcs γj and γj  1 by Lemma 2. ui.i 1 ¤ dÔP Õ and r vj.j  1 Vi. 2 Fig. the statement of the problem can be reformulated to an obvious fact: If the equalities u12 v12 .j  1 is circumscribed if and only . Lemma 3.i 1 and the ellipse arc εj. Hence. ρj Rj . Then we have ρi ρ ¡ ρi Ri   r . 3 circle ΩÔP. suppose ui. and let v be a unit vector Ri .j  1 ¤ dÔP Õ.j Vi 1. then u23 v23 holds as well. By Lemma 1 and Lemma 2 we have r ui. r and the value vij Ri   Rj dij Ri   Rj dij ¤ dÔP Õ.j  1.i 1 vj. Choose r ui.j  1. the curved quadrilateral Qij is circumscribed.i 1 vj. By Lemma 3. ρi AOi. Oj P ρ ¡ ρj Rj ¡ r . First suppose that the curved quadrilateral Qij is circumscribed and ΩÔP. Ôρi2 ¡ ρj2Õ   2ρ ¤ Ôρj ¡ ρiÕ ÔRi2 ¡ Rj2 Õ   2rÔRi   Rj Õ. r Õ is tangent to the half lines hi and hi 1 by Lemma 1. and ρj AOj . r Õ is its inscribed circle. r) hi P′ r′ hj r d(P ′ ) d(P ) B ρ ρ ρ ρj P A ρi Oi Ri βij r Rj r Oj γi γj P εij v C Fig.

respectively (then r PU P V ). Denote by ℓ the directrix of ellipse of εij related to the focus Oj . Ð . If one finds this fact. By the right angles. Oj P with arcs γi . Let Q be the projection of P onto the line AC . 5). then T is the radical center of σ . then the solution can be finished in many ways. this point does not depend on P . and then prove that it is tangent to the fourth one in either synthetic or analytical way. Using some known facts about conics. h3 . we have SOi U QP U . lying on Oi Oj . since εij is symmetrical about Oi Oj . and ǫ does not depend on P . we have P Oj distance from P to ℓ. the points U . the triangles SOi U and U P Q PU Oi S Oi S r . Due to the theorem on three homothety centers (or just to the Menelaus theorem applied to triangle Oi Oj P ). where dℓ ÔP Õ is the we have ℓ AC . the last expression is constant since S is a constant are similar and thus dÔP Õ PQ Oi U Ri point. or computed in a direct way. Ð ℓ γj σ V P U U U γi Oj S T A Q Oi C A Oi C P dℓ (A) d( (P P) ) d Oj dℓ (P ) εij Fig. the points U . one can find a circle touching three of h2 . 1. Let T be the intersection point of line AC and the common tangent to σ and γi at U . and these approaches are less case-sensitive. 4). The other case can be obtained as a limit case. γj respectively (see Fig. points U and V are the intersection points of rays Oi P .62 Comment 1. Lemma 2(b) (together with the easy Lemma 1(b)) is the key tool in this solution. 5 2. Hence. From this fact and the equality P U V UP SU Oi . Here we present some discussion about this key Lemma. V and Q lie on P V we get U QP UV P the circle with diameter P T . We consider only the case when P does not lie in the line Oi Oj . and γ3 . in other words. Here we outline a synthetic proof. Both approaches can be successful. r Õ with γi and γj . hence T V is the common tangent to σ and γj . Let S be the internal homothety center between the circles of γi and γj . Since Oi S P Q. In the solution above we chose an analytic proof for Lemma 2(b) because we expect that most students will use coordinates or vectors to examine the locus of the centers. the same statement can be proved in a very short way. Recall that for each point P È εij . and ǫ is the eccentricity of εij (see Fig. Let U and V be the points of tangency of circle σ ΩÔP. V and S are collinear. γi and γj . Now we have r Rj ¡ Ô Rj ¡ r Õ AOj ¡ P Oj ǫ dℓ ÔAÕ ¡ dℓ ÔP Õ   ¨ ǫ dÔP Õ ¡ dÔAÕ   ¨ ǫ ¤ dÔP Õ. That is. 4 Fig. ǫ ¤ dℓ ÔP Õ. γ2 .

β12 23 ½ ½ i. for this illustration. y Õ. r by the point Ôx. Similarly. As is well-known. it means that ε12 . y Õ .63 Comment 2. which has its apex on the opposite side of the base plane Π. y Õ and radius r   ¨ in space. r Õ Ô For every point Ôx. This point is the apex of the cone with base circle Ω Ôx. according to Lemma 2(a). the circles which are tangent to hi and hj correspond to the points of a half line βij at B . 6. Σ ε′ij ′ β12 ′ β23 ε′23 ε′12 γi γj Π they are coplanar. y. One can find a spatial interpretations of the problem   and the ¨ solution. According to ½ .i  1 and εj. Take some 1 i j 3. 6 Fig. If three of the four curved quadrilaterals are circumscribed. if Thus. starting Lemma 1. the z -coordinates were multiplied by 2. ε23 . and their axes are parallel).j  1 intersect. So the intersection of the two cones is a a conic section — which is an ellipse. represent the circle Ω x. The points which represent the circles touching γi and γj is an ellipse arc ε½ ij with end-points A and C . too (see Fig. the circles which are externally tangent to γi correspond to the points on the extension of another cone.e. 0. Now we translate Lemma 2. the curved quadrilateral Qij is circumscribed if and only if βi. containing γj . (See Fig. ½ ½ Fig. 7). It is easy to see that the locus of the points which represent these circles is a subset of a cone. 7 A connection between mathematics and real life: the Palace of Creativity “Shabyt” (“Inspiration”) in Astana . it follows that the common points of the two cones are co-planar. and consider those circles which are internally tangent to γj . and the fourth intersection comes to existence.) The two cones are symmetric to each other (they have the same aperture. ½ and β ½ lie in the same plane Σ. r and height r .

which implies so n 39. This 2010 335 means that ¢ ª 1 2 n¡1 1 66 42 ¤ ¤ ¤ ¤ ¤ 1¡ . Therefore 51 2010 ¢ ¢ 1¡ 1¡ 1 1 1 1¡ . 40. 3. we s2 ¤ ¤ ¤ sn. 1 34 66 ¤ ¤ 33 40 67 17 670 51 . ½ 51 42 is replaced by . 36. Same as Problem N1.. Surely s1 1 since otherwise 1 ¡ s1 0. n Solution for Problem N1½ . but the constant Answer for Problem N1. We have 1 2 32 34 39 66 ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ 2 3 33 35 40 67 hence for n 39 there exists a desired example. so sn si 67. . since the denominator of the fraction 7 42 is divisible by 67. 2010 Ô1Õ Comment. So we have may assume that s1 1 2 s1 s2 ¡ 1 ¤ ¤ ¤ sn ¡ Ôn ¡ 1Õ.. 67Ù which contains exactly 39 numbers.. 1 ¡ ¢ 1 sn ª 51 . . . . . 2010 2010 (Canada) Solution for Problem N1. 33. n. 48. 1 ¡ s1 s2 sn ª¢ ª ¢ ª 1 1 1 1¡ . . n 39. some of si ’s should be divisible by 67. . . . 2010 N1 . Consider the set Ø2. snÙ consisting of 1¡ 1 s1 ª¢ 1¡ 1 s2 ª . 2010 2 3 n 67 67n . . s2 . . . n 1 1 .. . hence si i   1 for each i 1. Now we are left to show that n 39 fits.. . Suppose that for some n there exist the desired numbers. 1 ¡ 2 3 n 1 n 1 2010 51 670 17 ª¢ ª ¢ ª 1 2 n ¤ ¤ ¤ ¤ 2 3 n 1 39. . .Number Theory n distinct positive integers such that ¢ N1. One can show that the example Ô1Õ is unique. Suppose that for some n there exist the desired numbers. Answer for Problem N1½ . Find the least positive integer n for which there exists a set Øs1 . 35.. Moreover. In the same way we obtain that si i   1.

1 35 66 ¤ ¤ 33 50 67 7 335 42 . 50. . 36. We propose N1 according to the number of current IMO. . 2010 Comment 2. Now we are left to show that n 48 fits. 37. . hence it is a bit harder. . . Another example is 1 2 46 66 329 ¤ ¤ ¤ ¤ ¤ ¤ 2 3 47 67 330 1 66 329 ¤ ¤ 67 330 47 7 67 ¤ 5 42 . . so n 48. 33. Consider the set Ø2. 67Ù which contains exactly 48 numbers. . N1½ was the Proposer’s formulation of the problem. We have 1 2 32 35 49 66 ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ 2 3 33 36 50 67 hence for n 48 there exists a desired example. 3. . On the other hand. the estimate needs one more step. the example in this version is not unique. . 2010 Comment 1. In this version of the problem.65 which implies n 2010 ¤ 66 42 ¤ 67 330 7 47.

q 2 ¤ 3q 2n 1 2 ¤ 93 2¤33 2n 2¤3 2 n 1 3 Ô  Õ . then 2 n  1 ¡ 1 In the second case let p n ¡ q . case n 0 n 1 n 2 n 3 n 4 n 5 equation m2 ¡ m   2 0 m2 ¡ 3m   6 0 m2 ¡ 7m   18 0 m2 ¡ 15m   54 0 m2 ¡ 31m   162 0 m2 ¡ 63m   486 0 discriminant ¡7 ¡15 ¡23 9 313 2025 452 6. n. p q Hence. For fixed values of n. Then 2 n  1 ¡ 1 3p   2 ¤ 3q m  m  2 ¤ 3n m 2 ¤ 3n m 3p   2 ¤ 3q . n. let q n ¡ p. so 9 § 2n 1 ¡ 1 if and only if 6 § n   1. 3Õ.66 N2. both terms are divisible by 3 . For n 5 the solutions are listed in the following table. the equation (1) is a simple quadratic equation in m. 3 n¡2 Now let h . we obtain (3) 2Ôn   1Õ . Since m § 2 ¤ 3n m 2n 1 ¡ 1 m 3p with some 0 p n or m 2 ¤ 3q with some 0 q n. In the first case. (2) Next.   ¨ (1) (Australia) Solution. Ô9. q Õ. q 2 n 1 3 2n 1 2 ¤ 83 n 8 n 1 3   9 n n 1 3   3 2 n 1 3 Ô  Õ Ô   Õ . we have 2 ¤ 3q   3p . Find all pairs Ôm. q . m 2n 1 ¡ 1 . therefore 9 § § that ord9 Ô2Õ 6. On the 3 § § we have h h § 3h § 2n 1 ¡ 1. and we can write 2n 1 ¡ 1 43r ¡ 1 Ô42r   4r   1ÕÔ2r ¡ 1ÕÔ2r   1Õ. we prove bounds for p. Ô6. We prove that there is no solution for n ¡ m2 . in particular. Combining these inequalities with p   q n¡2 3 p. § integer roots none none none m 6 and m none m 9 and m   9 54 ¨ Suppose that Ôm. By (3) we have h 1. Ô9. nÕ of nonnegative integers for which m2   2 ¤ 3n Answer. 5Õ. minÔp. Ô54. in both cases we need to find the nonnegative integer solutions of 2n 1 ¡ 1. n   1 6r for some positive integer r . 3Õ. nÕ satisfies (1) and n 6. (4) . 5Õ. From (2) we get 3p and so p. Therefore. It is easy check left-hand side of (2).

But this is impossible since we assumed n 3 6 . 2r ¡ 1 and 2r   1 are coprime: both are § odd and h h¡1 § r their §difference is 2. . Then 3h¡1 2r   1 3r n¡2 ¡1 h¡1 3 n 11. we have either 3 2 ¡ 1 or 3h¡1 § 2r   1. but it is never divisible by 9. Since the whole product is divisible by 3 . 6 § n 1   6. and we proved 6 § n   1. In any case. The other two factors in (4).67 Notice that the factor 42r   4r   1 Ô4r ¡ 1Õ2   3 ¤ 4r is divisible by 3. n 1 . we have 3h¡1 2r   1.

2 2 which contradicts the minimality of m. we get that x1 . x2 . It remains to show that x2   7 is not a sum of four (or less) squares of polynomials with rational coefficients. 4 as 2 2 2 2 2 2 2 Ôa2 Ôa1 b1   a2b2   a3 b3   a4 b4 Õ2   Ôa1 b2 ¡ a2b1   a3 b4 ¡ a4 b3 Õ2 1   a2   a3   a4 ÕÔb1   b2   b3   b4 Õ  Ôa1 b3 ¡ a3b1   a4 b2 ¡ a2 b4 Õ2   Ôa1 b4 ¡ a4b1   a2 b3 ¡ a3 b2 Õ2. (iii) 4 ô i 1 xi yi ¡6m2. By Euler’s identity we have 4 is impossible. by (ii). where the coefficients of polynomials f1 . It follows 3 that Ô x21 . Similarly. x3 . x4 . f3 and f4 are at most 1. The equality x2   7 x2   22   12   12   12 shows that n 5. y2 . y4 and m the system of equations (i) 4 ô i 1 x2 i 8m2 . 3. Define the numbers ai . 4 ô b2 i 7. y2 . x24 .e. . y3 . Then 4 ô i 1 4 ô i 1 4 ô i 1 p2 i 2 qi a2 i  2 4 ô i 1 4 ô ai bi   ai bi   4 ô i 1 4 ô i 1 b2 i b2 i 8. The smallest n is 5. b2 . x . . 3. 4 4 2 and some rationals a1 . f2 . Find the smallest number n such that there exist polynomials f1 .. Hence. the degrees of f1 . . (1) Let pi ai   bi and qi ai ¡ bi for i 4 ô i 1 4 ô i 1 4 ô i 1 i 1 i 1 i 1 1. if x is even) we have x2 0 Ômod 8Õ or x2 4 odd then x2 Ômod 8Õ. by (i). bi for i 1. Assume the contrary and consider a solution with minimal m. Otherwise (i. x2 . (ii) 4 ô i 1 2 yi 8m2 . y1 . Note that if an integer x is 1 Ômod 8Õ. m Õ is a solution of the system of equations (i). It follows that x2   7 i 1 Ôai x   bi Õ and hence 4 ô a2 i 1. 8 a2 i ¡2 a2 i ¡ and pi qi i 1 4 ô b2 i i 1 ¡6. 4 ô ai bi 0. Thus the LHS of (iii) is divisible by 4 and m is also even. fn with rational coefficients satisfying x2   7 f1 ÔxÕ2   f2 ÔxÕ2   ¤ ¤ ¤   fn ÔxÕ2 .68 N3. We prove that n in the previous solution. 3. b4 . Thus fi ÔxÕ ai x   bi for i 1. . y24 . . b1 . We will show that such a solution does not exist. x22 . Suppose by way of contradiction that x2   7 f1 ÔxÕ2   f2 ÔxÕ2   f3 ÔxÕ2   f4 ÔxÕ2 . we get that y1 . f2 . 2. (Poland) Answer. y21 . a4 . 2. Solution 2. f2 . 2. 4. a3 . a2 . b3 . y23 . f3 and f4 are rational (some of these polynomials may be zero). y22 . x3 and x4 are even. (ii) and (iii). 0 of which means that there exist a solution in integers x1 . Solution 1. Clearly. y3 and y4 are even.

m3 . The example with n 5 is already shown in Solution 1. Then m . m2 .69 So. 2 2 is also a solution of (2) which contradicts the minimality of m. (2) and m1 .   ¨ . Let m be a minimum positive integer number for which (2) holds. m2 . we get that m1 . a1 b4 ¡ a4 b1   a2 b3 ¡ a3 b2 ¡ m ©2 1 m   ¡ m ©2 2 m   ¡ m ©2 3 m . we have n 5. So. Then 8m2 2 2 2 m2 1   m2   m3   m . using the relations (1) from the Solution 1 we get that 7 where m1 m m2 m m3 m a1 b2 ¡ a2 b1   a3 b4 ¡ a4 b3 . m È N. m3 È Z. 2 . m are all even numbers. a1 b3 ¡ a3 b1   a4 b2 ¡ a2 b4 . 2 . 1 m2 m3 m As in the previous solution.

Proof. The pair Ô67. Ôa. § Comment 2. Therefore. Therefore. bÕ is very good if Ôa. . ¡512 Õ is good but not very good. Ôa. ¡512 Õ is not very good. Ð Claim 2. Let a. Ð Claim 3. k . If at2 ¡3as2 ¡ b Ômod 67Õ then for m t ¨ s. Consider the sets Øat2 Ômod 67Õ : 0 mod 67 : 0 s 33Ù. k ½ 0 Ômod 30Õ. (a) We show that the pair Ô1. Let P ÔxÕ x3 ¡ 512x. Therefore Ô1. (b) We will show that if a pair Ôa. one should deal with the large powers of 67 to reach the solution. By Fermat’s theorem. 67 b. b be integers. Then we have P Ôm Õ P Ô0Õ P Ôk Õ Ômod 30Õ and P Ôm Õ P ÔmÕ P ÔkÕ P Ôk Õ Ômod 67Õ. Since 67 § a and 67 § b. where d § 30 and i 0. Therefore. that is. § Hence. Assume that P ÔmÕ P Ôk Õ Ômod 67Õ. contradicting that Ôa. bÕ is 67-good. bÕ is 2010-good then Ôa. kÕ Ôu. This means t s 0 § Ômod 67Õ and b ¡3as2 ¡ at2 0 Ômod 67Õ. bÕ is 67-good.70 ax3   bx. we must have m k Ômod 67Õ in both cases. 30Õ is n-good if and only if n d ¤ 67i . bÕ is 67-good then 67 § a. Hence they have at least one element in common. § t 33Ù and Ø¡3as2 ¡ b Proof. so there exists a nontrivial cubic root of unity modulo 67. Since Ôa. The settings Ôm. k ©2s we have P ÔmÕ ¡ P Ôk Õ aÔm3 ¡ k 3 Õ   bÔm ¡ k Õ   ¨ § Ôm ¡ kÕ aÔm2   mk   k2 Õ   b Ôt ¨ 3sÕÔat2   3as2   bÕ 0 Ômod 67Õ. It shows that in part (b). On the other hand. bÕ is n-good for infinitely many positive integers n. bÕ is 67-good. we have 67 § a§ . bÕ is 67-good. § § Proof. The key property of number 67 is that it has the form 3k   1. (Turkey) Solution. from this we obtain m Hence we have m m3 k3 k k Ômod 51Õ. the second factor aÔm2   mk   k 2 Õ   b is coprime to 67 and hence 67i § m ¡ k . Since P Ô51Õ P Ô0Õ. N4. bÕ is 67i-good all i 1. Since 67 and 30 are coprime. bÕ is 67-good. bÕ is 67i -good. If 67 § b. If Ôa. We say that Ôa. bÕ is n-good if n § P ÔmÕ ¡ P Ôk Õ implies n § m ¡ k for all integers m. contradicting § that Ôa. Suppose that 67 § a. It follows that m m½ k ½ k Ômod 67Õ. bÕ is 2010-good then Ôa. each of these sets has 34 elements. and m½ m Ômod 67Õ. if P ÔmÕ P Ôk Õ Ômod 51Õ. 0Õ in the first case and Ôm. Claim 1. ¡2v Õ in the second case lead to b 0 Ômod 67Õ. bÕ is 67i -good for all positive integer i. But then 67 § P Ô7Õ¡ P Ô2Õ 67 ¤ 5a   5b and 67 § 7 ¡ 2. If Ôa. t 3s Ômod 67Õ and t ¡3s§ Ômod 67Õ. then m3 k 3 Ômod 51Õ. either au2 ¡b Ômod 67Õ or 3av2 ¡b Ômod 67Õ has a solution. but not very good. § §   ¨ i § Suppose that 67 P ÔmÕ ¡ P Ôk Õ Ôm ¡ k Õ aÔm2   mk   k 2 Õ   b §. (a) Find a pair Ôa. By Claim 2. § For any positive integer n we say that the pair Ôa. ¡512Õ is 51-good. This implies m½ k½ Ômod 2010Õ as Ôa. Ô1. kÕ Ôv. ¡512Õ is not n-good for any positive integer that does not divide 51. bÕ which is 51-good. Ð Comment 1. (b) Show that all 2010-good pairs are very good. m½ 0 ½ ½ ½ ½ Ômod 30Õ. Since a 0 Ômod 67Õ. bÕ is 2010-good. there exist integers m½ and k ½ such that k ½ k Ômod 67Õ. In the proof of Claim 2. If Ôa. bÕ is 2010-good then Ôa. the following reasoning can also be used. Since 3 is not a quadratic residue modulo 67. then P ÔxÕ P Ô0Õ Ômod 67Õ for all x.§ and let P ÔxÕ Ômod 3Õ and m m33 k 33 k Ômod 17Õ. hence P Ôm½Õ P Ôk½Õ Ômod 2010Õ. the pair Ô1.

so f ÔℓÕ f Ôk Õ   p2 a for some integer a. Solution. we get that both the     applying ¨  the problem ¨ numbers f Ôk Õ   n f ÔnÕ   k and f ÔℓÕ   n f ÔnÕ   ℓ are squares divisible by p (and thus by p2 ). Then q cannot be ¡1 since otherwise for n f Ô1Õ  1 we have f ÔnÕ 0 which is impossible. ℓ. then choose the same number D and set n p3 D ¡ f Ôk Õ. We start with the following § § Lemma. First. thus f Ôk   1Õ ¡ f Ôk Õ 1. Then the positive numbers f Ôk Õ   n p3 D and f ÔℓÕ   n   ¨ p3 D   f ÔℓÕ ¡ f Ôk Õ are respectively divisible by p3 (but not by p4 ) and by p (but not by p2 ). Now. we get f ÔnÕ f Ô1Õ  qn. All functions of the form f ÔnÕ n   c. q 1. Then we prove by induction that f ÔnÕ f Ô1Õ   q Ôn ¡ 1Õ. Answer. the function f is injective. Let   N be the ¨  set of all ¨positive integers. For the step. . Then by Lemma we have that k ¡ ℓ is divisible by every prime number. Since the number Ôk   1Õ ¡ k 1 has no prime divisors. as desired. where c È N N such that the (U. § Proof. We are left to prove that there are no other functions.A. this means §  ¨   that the ¨ multipliers f ÔnÕ   k and f ÔnÕ   ℓ are also divisible by p.S. consider the numbers f Ôk Õ and f Ôk   1Õ. therefore § p f ÔnÕ   k ¡ f ÔnÕ   ℓ k ¡ ℓ as well. it is clear that all functions of the   form f ÔnÕ ¨  n   c with ¨ a constant nonnegative integer c satisfy the problem conditions since f ÔmÕ   n f ÔnÕ   m Ôn   m   cÕ2 is a square. Therefore. as desired. Hence in analogous way the numbers f ÔnÕ   k and f ÔnÕ   ℓ are divisible by p. suppose that f Ôk Õ f ÔℓÕ for some k. First. 2 holds by the definition of q . Then p § k ¡ ℓ. if n 1 we have f Ôn   1Õ f ÔnÕ¨ q f Ô1Õ  q Ôn ¡ 1Õ¨ q . The base for n 1. Finally. Then the positive numbers n   f Ôk Õ pD and n   f ÔℓÕ pD   f ÔℓÕ ¡ f Ôk Õ pÔD   paÕ are 2 both divisible by p ¨  but not by¨p . let f Ô2Õ ¡ f Ô1Õ q . by Lemma the same holds for f Ôk   1Õ ¡ f Ôk Õ. Since f ÔnÕ f Ôn ¡ 2Õ f Ô1Õ  q Ôn ¡ 2Õ. ℓ È N. Suppose first that p2 § f Ôk Õ ¡ f ÔℓÕ. Find all functions f : N number f ÔmÕ   n m   f ÔnÕ is a square for all m. or k ℓ. if f Ôk Õ ¡ f ÔℓÕ is divisible by p but not by p2 . f ÔℓÕÙ which is not divisible by   p and set n ¨ pD ¡ f Ôk Õ. §  ¨ we   obtain that ¨ § therefore p f ÔnÕ   k ¡ f ÔnÕ   ℓ k ¡ ℓ.71 N5. Next. On the other hand. conditions. Now. Take some positive integer D maxØf Ôk Õ. and f Ô1Õ ¡ 1 0. so k ¡ ℓ 0. n È N. Hence q 1 and f ÔnÕ Ôf Ô1Õ ¡ 1Õ   n for each n È N. Ð We turn to the problem. Suppose that p § f Ôk Õ¡ f ÔℓÕ for some prime p and positive integers k .) Ø0Ù. we have f ÔnÕ f Ô1Õ  q Ôn ¡ 1Õ.

a1 . . and the edges are the arrows Ôi. a2 . we have ak Lemma 1. j Õ refers to the j th cell in the ith row. By applying Lemma 1 to the Fibonacci-type sequence Fk . Denote by F0 . the j th cell in the ith row and the Ôi   j Õth cell in the j th row have the same color. Now. . in any coloring considered. 0. Fk 1 . For every positive integer m. Thus. and we have to prove that this number is 2n . (c) F3¤2m¡2  1 1   2m¡1 Ômod 2m Õ. For the step. In the base case m ν ÔF3¤2m¡2 Õ ν Ô8Õ 3. 2 2 F2k 1 Fk ¤ Fk   Fk 1 ¤ Fk 1 Fk   Fk 1 . . . the graph splits into cycles. Õ by the relations a0 i1 . i   j Õ for all 0 i. (The indices of the cells in a row are considered modulo 2n . Next. ν ÔFk Õ 2Ôm ¡ 1Õ Ôm ¡ 1Õ   1 ν Ô2Fk Fk 1 Õ. the maximal possible number of colors is the same as the number of cycles. j 2n ¡ 1. For every Fibonacci-type sequence a0 . . Ôi. From each vertex Ôi. The cells of the ÔFk¡1a0   Fk a1 Õ   ÔFk¡2a0   Fk¡1a1Õ Fk a0   Fk 1 a1 . (Iran) Solution. for which 2 m but 2 m. the Fibonacci numbers defined by F0 0. establishing statement (a). k Õ exactly one edge is directed (from the cell Ôk ¡ j mod 2n . if each cycle has its own color. § (b) d 3 ¤ 2m¡2 is the least positive index for which 2m § Fd . whose vertices are the cells of the board. Define a sequence Ôa0 . j 2n ¡ 1. . 0. . consider any cycle Ôi1 . Now suppose that m 3 and let k 3 ¤ 2m¡3 .e. Apply induction on m. Hence we need to investigate the behavior of that ik Fibonacci-type sequences modulo 2n . . a1 j1 . the preceding Fibonacci-numbers are not divisible by 8. i. The base cases k induction hypothesis we get ak 1 ak   ak¡1 N6. j2 Õ. j1 Õ. a1 . and Fn 2 Fn 1   Fn for n 0. j Õ Ôj. (a) we have ν ÔF3¤2m¡2 Õ m. Hence. Consider the directed graph. . 2 m ¡ 1. . j Õ. Ôi2 .72 table have been colored with the following property being satisfied: for each 0 i. i   j mod 2n Õ). . . 3 we have ν ÔF3¤2m¡2 Õ F6 8. which implies ν ÔF2k Õ m. we get F2k 2 Fk¡1 Fk   Fk Fk 1 ÔFk 1 ¡ Fk ÕFk   Fk 1 Fk 2Fk 1Fk ¡ Fk . . F1 . The rows and columns of a 2n ¢ 2n table are numbered from 0 to 2n ¡ 1. . F1 1. from the Proof. exactly one edge passes (to Ôj. On the other hand. . For every m 3. Then an obvious induction shows ak¡1 Ômod 2n Õ. 1 are trivial. Throughout the solution we denote the cells of the table by coordinate pairs. . the obtained coloring obviously satisfies the problem conditions. an 1 an   an¡1 for all n 1 (we say that such a sequence is a Fibonacci-type sequence). we will describe it in other terms. conversely. the vertices of each cycle should have the same color by the problem condition. Therefore we get ν ÔFk Õ By the induction hypothesis. . j ÕÕ. denote by § § ν ÔmÕ the exponent of 2 in the prime factorization ν ÔmÕ § ν ÔmÕ 1 § of m. Apply induction on k . to each cell Ôj. . and every k Fk¡1 a0   Fk a1 . Ð Lemma 2.) Prove that the maximal possible number of colors is 2n . and indeed F3¤2m¡2  1 F7 13 1   4 Ômod 8Õ. so Proof. and Fk 1 is odd. We also set F¡1 1 according to the recurrence relation. jk ak Ômod 2n Õ.

. . since Fk 1 F2k 1 1   2m¡2   a2m¡1 for some integer a. 1. denote this common value by µÔAÕ. 3. . Obviously we should have pn¡1 ÔAÕ § pn ÔAÕ. modulo 4 (all pairs of residues from which at least one is odd appear as a pair of consecutive terms in this sequence). 2. . they are split into cycles.. The total number of such cells is 22Ôn¡kÕ ¡ 22Ôn¡k¡1Õ 3 ¤ 22n¡2k¡2 . a2s 2b2s   F2s 2b0   0 a0 Ômod 2n Õ. b1 . . . . a1 Õ µÔa1 . . a1 ß2k . Õ has period p modulo 2n if and only if the sequence Ôa0 ß2k . Proof. bÕ minØν ÔaÕ. Let A Ôa0 . . and it forms a separate cycle. 1. Lemma 3 provides a straightforward method of counting the number of cycles. But then we have Fℓ Fk¡1 Fℓ¡k   Fk Fℓ¡k 1 . we note that the sequence Ôa0 . hence a1 is odd. Actually. 3. F2s 1 1 Ômod 2 Õ. Assume the contrary. Õ starting with Ôb0 . First. . . using the induction hypothesis. which means that pn ÔAÕ 2s. Õ has period p modulo 2n¡k . and as 0. s 3 ¤ 2n¡2 § pn ÔAÕ. So the total number of cycles is 1  n 1 ô k 0 ¡ 2n¡1¡k 1   Ô1   2   4   ¤ ¤ ¤   2n¡1Õ 2n . . By an obvious induction. 1. by Lemma 2 we have Fs 1 n n Ômod 2 Õ. Õ with µÔAÕ § 0. The first line means that A is not s-periodic. each Fibonacci-type sequence A with µÔAÕ 0 behaves as 0. take any number 0 k n ¡ 1 and consider all the cells Ôi. By § the induction hypothesis. . Hence the number of cycles 3 ¤ 22n¡2k¡2 consisting of these cells is exactly 2n¡k¡1. ν ÔbÕÙ. . Finally. from the induction hypothesis it follows that ℓ k . j Õ k . that is. the least p 0 such that ak p ak Ômod 2n Õ for all k 0. and a0 2b0 . Hence the sum is not divisible even by 2m¡1 . by an easy induction we get ak 2bk   Fk for all k 0. actually. a1 . 1. as desired in statement (c). Hence. a2s 1 2b2s 1   F2s 1 2b1   1 a1 Ômod 2n Õ. a1 . Õ be a Fibonacci-type sequence such that µÔAÕ k n. Now suppose that n 3 and consider an arbitrary § Fibonacci-type sequence A Ôa0 . Since 2m¡1 § Fℓ . 2b1 . 1. 1. . Then pn ÔAÕ 3 ¤ 2n¡1¡k . where the second summand is divisible by 2m¡1 but the first one is not (since Fk¡1 is odd and ℓ ¡ k k ). Ð Finally. 1. j Õ with µÔi. b1Õ. Since a0 2b0   F0 . . we have pn¡1 ÔB Õ § s. 0Õ 3 ¤ 2n¡1¡k which is not mentioned in the previous computation. let µÔa. bÕ Ô0. 0. we get 0   Ô1   2m¡2   a2m¡1 Õ2 2 Fk   Fk2 1 1   2m¡1 Ômod 2mÕ. . . Also denote by pn ÔAÕ the period of this sequence modulo 2n . . a2 Õ . for every pair of integers Ôa. hence as 1 2bs 1   Fs 1 2b1   1   2n¡1 2b1   1 a1 Ômod 2n Õ. Next. . passing to this sequence we can assume that k 0. Õ is s-periodic 1   2n¡1 Ômod 2n Õ. § § We are left to prove that 2m § Fℓ for ℓ 2k . or. Hence s § pn ÔAÕ § 2s and pn ÔAÕ s. . as desired. We prove the statement by induction on n. a1 2b1   F1 . . there is only one cell Ô0. we may suppose that a0 is even. 2. and by Lemma 3 the length of each cycle is 3 ¤ 2n¡1¡k . A contradiction. On the other hand. . 2 the claim is true. hence the sequence Ô2b0 . 1. a1 2b1   1 for some integers b0 . Consider the Fibonacci-type sequence B Ôb0 . 1. . a1 . while§ the other two provide that a2s § a2s 1 a1 and hence a2s t at for all t 0. . modulo 2. a1 . It is easy to see that for n 1. . 0. . 0Õ. . 1. a0 . On the other hand. . Õ we have µÔa0 . b1 . . Lemma 3. for every Fibonacci-type sequence A Ôa0 .73 Moreover. F2s 0 modulo 2n . Ð Now.

0 a1 ¤ a0 ¡ a0 ¤ a1 a1 Fp Ôa1 ¡ a0 Õ   Fp 1 a0   . Note that p is a (not necessarily minimal) period of the sequence Ôai Õ modulo 2n if and only if we have a0 ap Ômod 2n Õ. Conversely. we assume that k 0 and show that in this case the period of Ôai Õ modulo 2n coincides with the period of the Fibonacci numbers modulo 2n . then the proof can be finished by the arguments from Lemma 2.74 Comment. Combining the relations of (1) we get Ômod 2nÕ. That is. and hence the minimal periods coincide as well. So. the number a2 1 ¡ a1 a0 ¡ a0 is odd as well. . We outline a different proof for the essential part of Lemma 3. a1 is odd. Õ modulo 2n . F0 0 Ômod 2n Õ and Fp 1 ¨ Fp Ôa1 ¡ a0 Õ   Fp 1 a0 Ômod 2nÕ.. which by (1) implies that p is a period of Ôai Õ as well.   ¨ 2 a2 Ôa1 ¡ a0Õa1 ¡ a0 ¤ a0 Ôa1 ¡ a0ÕÔFp a0   Fp 1a1 Õ ¡ a0 Fp Ôa1 ¡ a0 Õ   Fp 1 a0 1 ¡ a1 a0 ¡ a0 2 n Fp 1 Ôa2 1 ¡ a1 a0 ¡ a0 Õ Ômod 2 Õ. If p is a period of ÔFi Õ then we have Fp F1 1 Ômod 2n Õ. . suppose that p is a period of Ôai Õ. Therefore the n previous relations are equivalent with Fp 0 Ômod 2 Õ and Fp 1 1 Ômod 2n Õ. (1) ¡ a0ÔFp a0   Fp 1 a1Õ 2 n Fp Ôa2 1 ¡ a1 a0 ¡ a0 Õ Ômod 2 Õ. that is. the sets of periods of Ôai Õ and ÔFi Õ coincide. a0 a1 ap ap 1 Fp¡1 a0   Fp a1 Fp a0   Fp 1 a1 Now. 2 Since at least one of the numbers a0 . F1 . which means exactly that p is a period of ÔF0 . . a1 ap 1 Ômod 2n Õ.

75 .

76 .

Sign up to vote on this title
UsefulNot useful