Math Problem Book I

compiled by

Kin Y. Li

Hong Kong Mathematical Society International Mathematical Olympiad Hong Kong Committee

(Supported by the Quality Education Fund)

Math Problem Book I
compiled by

Kin Y. Li
Department of Mathematics Hong Kong University of Science and Technology

Copyright c 2001 Hong Kong Mathematical Society IMOHK Committee. Printed in Hong Kong

We hope the problems collected in this book will stimulate them and seduce them to deeper understanding of what mathematics is all about. In fact. We are especially pleased with the e orts of these students. Through problem solving. but there are still some that escape us at the moment. especially those who worked so hard to become reserve or team members. This book is for students who have creative minds and are interested in mathematics. The problems in this book came from many sources. in future editions of the book we can ll in these missing sources with the help of the knowledgeable readers. For those involved in international math competitions. We hope the international math communities support our e orts for using these brilliant problems and solutions to attract our young students to mathematics. We tried to identify the sources whenever possible. the original motivation for writing the book was to reward them in some ways. Hopefully. Most of the problems have been used in practice sessions for students participated in the Hong Kong IMO training program. dating back to the late 1800's in Hungary's case. they no doubt will recognize many of these problems. It is only tting to list their names along with their solutions. Russia and Romania. mathematical competitions have a long history.Preface There are over fty countries in the world nowadays that hold mathematical olympiads at the secondary school level annually. In Hungary. Many professional or amateur mathematicians developed their interest in math by working on these olympiad problems in their youths and some in their adulthoods as well. Again there are unsung heros iii . they will learn a great deal more than school curricula can o er and will sharpen their analytical skills.

who contributed solutions. will involve the dedication of more than one person. Li Hong Kong April. which provided the support that made this book possible. Kin Y. this is a problem book. 2001 iv . we would like to express deep gratitude to the Hong Kong Quality Education Fund. for certain. but whose names we can only hope to identify in future editions. This. From the results of the Hong Kong IMO preliminary contests. As the title of the book suggest. So very little introduction materials can be found. Finally. Maybe the next problem book in the series will be written by our students. Also. We do promise to write another book presenting the materials covered in the Hong Kong IMO training program. this is the rst of a series of problem books we hope. we can see waves of new creative minds appear in the training program continuously and they are younger and younger.

While thinking about the problems. This helps you to develop strategies for dealing with other problems.Advices to the Readers The only way to learn mathematics is to do mathematics. you will nd many math problems. Always try to write out your solution in a clear and concise manner. 4. Have protractor. see how the given facts are necessary for solving the problem. Sometimes may be you can even guess the answers from some cases. you can create something you can be proud of. You may not succeed in solving all the problems. In this book. This may help you to solve related problems later on. Along the way. ruler and compass ready to measure angles and lengths. to unleash your potential of creativity. Very few people can solve them all. To start thinking about a problem. Other things you can try in tackling a problem include changing the given conditions a little or experimenting with some special cases rst. The purposes of the book are to expose you to many interesting and useful mathematical ideas. ranging from simple to challenging problems. Finally. you may discover things you never know before and putting in your ideas. very often it is helpful to look at the initial cases. when you gure out the solutions. which can solve the whole problem. the ideas in these cases allow you to see a pattern. For geometry problems. 5: These cases are simple enough to let you get a feeling of the situations. 3. you will polish the argument and see the steps of the solutions more clearly. such as when n = 2. to develop your skills in analyzing problems and most important of all. always draw a picture as accurate as possible rst. then you can study the form of the answers and trace backward. Sometimes. v . You should try to generalize the problem. don't just stop there.

We hope this book will increase your interest in math. Read the statements of the problems and start with the ones that interest you the most. Finally. Also if you have something you cannot understand. We recommend inspecting the list of miscellaneous problems rst. we will o er one last advice. Have a fun time. please feel free to contact us by email. please send it to us by email at makyli@ust. If you have a nice elegant solution to a problem and would like to share with others in future editions of this book.hk .The solutions presented in the book are by no means the only ways to do the problems. vi . Don't start with problem 1.

Table of Contents Preface :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : iii Advices to the Readers : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : v Contributors : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : ix Algebra Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 1 Geometry Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 10 Number Theory Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 18 Combinatorics Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 24 Miscellaneous Problems : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 28 Solutions to Algebra Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 35 Solutions to Geometry Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 69 Solutions to Number Theory Problems : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 98 Solutions to Combinatorics Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 121 Solutions to Miscellaneous Problems :: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 135 .

.

2000 Hong Kong team member Poon Wai Hoi. 1999 Hong Kong team reserve member Lam Pei Fung. 1997. 2000 Hong Kong team member Fung Ho Yin. 1997 Hong Kong team reserve member Ng Ka Wing. 2000 Hong Kong team member Law Siu Lung. 1999. 1996 Hong Kong team member Poon Wing Chi. 2000 Hong Kong team member Wong Him Ting. 1997 Hong Kong team reserve member Ho Wing Yip. 1999. 1997 Hong Kong team member Yung Fai. 1997. 1998 Hong Kong team member Law Ka Ho. 1997 Hong Kong team member Ng Ka Man. 1996 Hong Kong team member Lee Tak Wing. 1998. 1991. 2001 Hong Kong team member Chan Ming Chiu. 2001 Hong Kong team member Leung Wing Chung. 1995. 1998 Hong Kong team member Fan Wai Tong. 1997. 1992 Hong Kong team member Lau Lap Ming. 1999 Hong Kong team reserve member To Kar Keung. 1994. 1994. 1998. 1993 Hong Kong team reserve member Leung Wai Ying.Contributors Chan Kin Hang. 1994. 1997 Hong Kong team reserve member Chao Khek Lun. 1993 Hong Kong team member ix . 1999. 1996 Hong Kong team member Kee Wing Tao. 1995. 1995 Hong Kong team member Yu Ka Chun. 2000. 2001 Hong Kong team member Cheung Pok Man. 1997 Hong Kong team reserve member Lam Po Leung. 1998 Hong Kong team member Mok Tze Tao. 1992 Hong Kong team member Wong Chun Wai. 2001 Hong Kong team member Cheng Kei Tsi. 1999. 1996. 1995. 1997 Hong Kong team reserve member Tam Siu Lung.

.

Problems .

.

gx. 6. if P x = x1 . 4. 3. : : : of nonzero real numbers such that for n = 1. 1995 Russian Math Olympiad Is it possible to nd three quadratic polynomials f x. B x. fnx with real coe cients. 1. 1990 Putnam Exam Is there an in nite sequence a0 . 1 is divisible by x + 13 : 2. 3. 1991 Austrian-Polish Math Competition Let P x be a polynomial with real coe cients such that P x  0 for 0  x  1: Show that there are polynomials Ax. n: Find the value of P n + 1: 3. x. a1. 4. 6. 2. f2 x. 7. : : :. the polynomial Pnx = a0 + a1 x + a2 x2 +    + an xn has exactly n distinct real roots? 7. x2 +1 . C x with real coe cients such that a Ax  0. 2. : : : . then P x = 0+ x1 . xx2 : 1 . 13 and px . Problem 7 Find without calculus a fth degree polynomial px such that px + 1 is divisible by x . 1999 Putnam Exam Let P x be a polynomial with real coe cients such that P x  0 for every real x: Prove that P x = f1 x2 + f2 x2 +    + fnx2 for some polynomials f1 x. 5. 2. B x  0. Crux Mathematicorum. 8? 5. : : :. C x  0 for all real x and b P x = Ax + xB x + 1 . 1968 Putnam Exam Determine all polynomials whose coe cients are all 1 that have only real roots.Algebra Problems Polynomials 1. A polynomial P x of the n-th degree satis es P k = 2k for k = 0. hx such that the equation f ghx = 0 has the eight roots 1. xC x for all real x: For example. a2.

8. 1993 IMO Let f x = xn + 5xn,1 + 3; where n 1 is an integer. Prove that f x cannot be expressed as a product of two polynomials, each has integer coe cients and degree at least 1. 9. Prove that if the integer a is not divisible by 5, then f x = x5 , x + a cannot be factored as the product of two nonconstant polynomials with integer coe cients. 10. 1991 Soviet Math Olympiad Given 2n distinct numbers a1; a2; : : : ; an; b1; b2; : : :; bn; an n  n table is lled as follows: into the cell in the i-th row and j -th column is written the number ai + bj : Prove that if the product of each column is the same, then also the product of each row is the same. 11. Let a1; a2; : : : ; an and b1; b2; : : : ; bn be two distinct collections of n positive integers, where each collection may contain repetitions. If the two collections of integers ai + aj 1  i j  n and bi + bj 1  i j  n are the same, then show that n is a power of 2.
Recurrence Relations

12. The sequence xn is de ned by + xn ; n = 1; 2; 3; : : :: x1 = 2; xn+1 = 12, 2x
n

or 0 for all n and the terms of the sequence are all Prove that xn 6= 1 2 distinct. 13. 1988 Nanchang City Math Competition De ne a1 = 1; a2 = 7 and a2 , 1 for positive integer n: Prove that 9a a + 1 is a an+2 = n+1 n n+1 an perfect square for every positive integer n: 14. Proposed by Bulgaria for 1988 IMO De ne a0 = 0; a1 = 1 and an = 2an,1 + an,2 for n 1: Show that for positive integer k; an is divisible by 2k if and only if n is divisible by 2k : 2

15. American Mathematical Monthly, Problem E2998 Let x and y be n , yn x distinct complex numbers such that x , y is an integer for some n , yn x four consecutive positive integers n: Show that x , y is an integer for all positive integers n:
Inequalities

16. For real numbers a1; a2; a3; : : :; if an,1 + an+1  2an for n = 2; 3; : : :; then prove that

An,1 + An+1  2An for n = 2; 3; : : :;
where An is the average of a1 ; a2; : : :; an: 17. Let a; b; c 0 and abc  1: Prove that

a + b + c  a + b + c: c a b
18. 1982 Moscow Math Olympiad Use the identity 13 + 23 +    + n3 = n2n + 12 to prove that for distinct positive integers a ; a ; : : :; a ; 1 2 n 2
7 7 5 5 5 3 3 3 2 a7 1 + a2 +    + an  + a1 + a2 +    + an   2a1 + a2 +    + an  :

Can equality occur? 19. 1997 IMO shortlisted problem Let a1      an  an+1 = 0 be a sequence of real numbers. Prove that

v u n n u tX ak  X pkpak , pak :
k=1 k=1
+1

3

20. 1994 Chinese Team Selection Test For 0  a  b  c  d  e and a + b + c + d + e = 1; show that ad + dc + cb + be + ea  1 5: 21. 1985 Wuhu City Math Competition Let x; y; z be real numbers such that x + y + z = 0: Show that 6x3 + y3 + z 3 2  x2 + y2 + z 23 : 22. 1999 IMO Let n be a xed integer, with n  2: a Determine the least constant C such that the inequality

X

1 

i jn

xi xj xi + xj   C
2 2

X
1 

in

xi

4

holds for all nonnegative real numbers x1 ; x2; : : : ; xn: b For this constant C; determine when equality holds. 23. 1995 Bulgarian Math Competition Let n  2 and 0  xi  1 for i = 1; 2; : : :; n: Prove that hni x1 + x2 +    + xn , x1 x2 + x2 x3 +    + xn,1 xn + xn x1   2 ; where x is the greatest integer less than or equal to x: 24. For every triplet of functions f; g; h : 0; 1 ! R; prove that there are numbers x; y; z in 0; 1 such that jf x + gy + hz  , xyz j  1 3: 25. Proposed by Great Britain for 1987 IMO If x; y; z are real numbers such that x2 + y2 + z 2 = 2; then show that x + y + z  xyz + 2: 4

a2. b. : : :. an and b1. c 0 and m be a positive integer. 1  k  n: Show that b1 + b2 +    + bn  a1 + a2 +    + an: 28. d. Proposed by USA for 1993 IMO Prove that for positive real numbers a. Proposed by Greece for 1987 IMO Let a. b.26. prove that . bn be 2n positive real numbers such that a a1  a2      an and b b1 b2    bk  a1 a2    ak for all k. : : :. Let a1 . c. b2. b c d 2: a + + +  b + 2c + 3d c + 2d + 3a d + 2a + 3b a + 2b + 3c 3 27.

an a1 + a2 +    + an. d 0 and 1 + 1 + 1 + 1 = 1: 1 + a4 1 + b4 1 + c4 1 + d4 5 . : : :. Prove that if a. c. 1982 West German Math Olympiad If a1. a2. an be distinct positive integers. then show that n X ai  n : i 2a . b. then b a + +c c+a a+b 2 32. a2. show that a1 + a2 +    + an  1 . : : : . c 0. Let a1. ai 2n . 1 : 2 8 n2n 2n 30. b. 1 =1 0 and a = 3 b3 + c3  a2 + b2 + c2 : 31. Let a. am + bm + cm  3 a + b + c m.1 : b+c c+a a+b 2 3 29.

namely 1 + 1 + 1 + 1. there exists an 1? integer m such that . P  x  n : P0 x = 0 and for n  0. Pn+1x = Pn x + 2 Prove that p 2 0  x . 2 + 2. where a. Show that 1 ++ 1 1 + n : a1 an 1951 34. d arep complex numbers with jaj = jbj = jcj = jdj = 1: Show that jP z j  6 for at least one complex number z satisfying jz j = 1: 37. Thus p4 = 5. American Mathematical Monthly. and 4: 6 . b. 1 : 35.Prove that abcd  3: 33. A sequence Pn  of polynomials is de ned recursively as follows: 2 x . Pn x  n + 1 for every nonnegative integer n and all x in 0. 1 + 3. 1979 British Math Olympiad If n is a positive integer. P x = ax3 + bx2 + cx + d: Prove that if jP xj  1 for all x such that jxj  1. 1 + 1 + 2. 1997 Hungarian-Israeli Math Competition Find all real numbers with the following property: for any positive integer n. then jaj + jbj + jcj + jdj  7: 36. Problem 4426 Let P z  = az 3 + bz 2 + cz + d. m n 3n 38. c. : : :. 1996 IMO shortlisted problem Let P x be the real polynomial function. an is less than 1951. as there are ve di erent ways of expressing 4 in terms of positive integers. The least common multiple of any two of these is greater than 1951. Due to Paul Erdos Each of the positive integers a1. denote by pn the number of ways of expressing n as the sum of one or more positive integers.

y: Prove that f x = x for every real number x: 42. 1  0 for each n 1: Functional Equations 39. f a + b = f a + f b for all a. Find all polynomials f satisfying f x2  + f xf x + 1 = 0: 40. 1996 IMO shortlisted problem Let R denote the real numbers and f : R ! .1. b 2 R and f xf  1 x  = 1 for x 6= 0: Show that f x = x for all x: 43. 1 ! R be a function such that a f is strictly increasing. 1997 Greek Math Olympiad Let f : 0. 1979 Eotvos-Kursch ak Math Competition The function f is de ned for all real numbers and satis es f x  x and f x + y  f x + f y for all real x. 1992 Polish Math Olympiad Let Q+ be the positive rational numbers. 1 b f x . x for all x 0 and 1  = 1 for all x 0: c f xf f x + x Find f 1: 41. Proposed by Ireland for 1989 IMO Suppose f : R ! R satis es f 1 = 1.Prove that pn + 1 . Determine all functions f : Q+ ! Q+ such that f x + 1 = f x + 1 and f x3  = f x3 for every x 2 Q+ : 44. 2pn + pn . 1 satisfy 1 +f x+ 1 f x + 13 + f  x  = f x + 42 6 7 .

.

.

e. for every x 2 R: Show that f is a periodic function. i. Suppose s : N ! N is an increasing function such that ssn = 3n for all n 2 N: Find all possible values of s1997: 7 . Let N denote the positive integers. there is a nonzero real number T such that f x + T  = f x for every x 2 R: 45.

Is there a function f : N ! N such that f 1996n = 2n for all n 2 N. y in R: 50. 8 . Find all functions f : R ! R such that f f x + y = f x + y + f xf y . 1993 Czechoslovak Math Olympiad Let Z be the integers.46. Let R be the real numbers. 1995 Byelorussian Math Olympiad Let R be the real numbers. f y = f f y + xf y + f x . xy for all x. 1999 IMO Determine all functions f : R ! R such that . where f 1x = f x and f k+1 x = f f k x? 47. 48. y 2 R: 51.  f x . American Mathematical Monthly. 2f m2 + n2  = f m2 + f n2 . 2 or show no such function can exist. 1995 South Korean Math Olympiad Let A be the set of non-negative integers. Find all functions f : Z ! Z such that f . 1 for all x. y: 52. Find all functions f : A ! A satisfying the following two conditions: a For any m. 2xy for all integers x. f xf y + x = xy + f x: 49. Find all functions f : R ! R such that f f x = x2 .1 = f 1 and f x + f y = f x + 2xy + f y . Problem E984 Let R denote the real numbers. n 2 A. Let N be the positive integers. Find all functions f : R ! R such that for all real numbers x and y.

f m2   f n2 : 53. Problem E2176 Let Q denote the rational numbers. n 2 A with m  n.b For any m. Find all functions f : Q ! Q such that f 2 = 2 and . American Mathematical Monthly.

: : :. 9 . 23 + n . : : :. x2. f y for x 6= y: 54. y = f x . de ne f n = 1n + 2n. Problem 1552 Find all functions f : R ! R such that f x + yf x = f x + xf y for all x. y in R: Maximum Minimum 55. 1985 Austrian Math Olympiad For positive integers n. 1996 Putnam Exam Given that fx1 .1 + 3n. x + y f x + f y f x .2 +    + n . 12 + n: What is the minimum of f n + 1=f n? 56. Mathematics Magazine. nd the largest possible value of x1 x2 + x2 x3 +    + xn. 2. xn g = f1. ng.1xn + xnx1 in terms of n with n  2.

Let the tangents at Y and Z intersect at X and lines Y O1 and ZO2 intersect at P: Let the circumcircle of 4O1O2B have center at O and intersect line XB at B and Q: Prove that PQ is a diameter of the circumcircle of 4O1O2 B: 61. Prove that two of the sides have the same length. Prove that the triangle is a right triangle. which is not a diameter. 63. The length of each side divides the sum of the other three lengths. there are four points such that the distance between every pair of them is greater than the radius of the disk. 10 . Prove that there is a pair of perpendicular diameters such that exactly one of the four points lies inside each of the four quarter disks formed by the diameters. The lengths of the sides of a quadrilateral are positive integers. Chords A1 B1 and A2B2 intersect at the midpoint P of AB: Let the tangents to the circle at A1 and B1 intersect at C1: Similarly. 1988 Leningrad Math Olympiad Squares ABDE and BCFG are drawn outside of triangle ABC: Prove that triangle ABC is isosceles if DG is parallel to AC: 59. respectively. 1988 Sichuan Province Math Competition Suppose the lengths of the three sides of 4ABC are integers and the inradius of the triangle is 1. let the tangents to the circle at A2 and B2 intersect at C2: Prove that C1 C2 is parallel to AB: 60. 62. 1995 British Math Olympiad Triangle ABC has a right angle at C: The internal bisectors of angles BAC and ABC meet BC and CA at P and Q respectively. 1981 Beijing City Math Competition In a disk with center O. The points M and N are the feet of the perpendiculars from P and Q to AB: Find angle MCN: 58. Z. 1991 Hunan Province Math Competition Two circles with centers O1 and O2 intersect at points A and B: A line through A intersects the circles with centers O1 and O2 at points Y. AB is a chord of a circle.Geometry Problems 57.

Quadrilaterals ABCP and A0 B 0 C 0 P 0 are inscribed in two concentric circles. 1985 IMO A circle has center on the side AB of the cyclic quadrilateral ABCD: The other three sides are tangent to the circle. Prove that AD + BC = AB: 65. side CA at E and side AB at F: Let G be the foot of perpendicular from D to FG = BF : EF: Show that EG CE 69. prove that P 0 A2 + P 0 B 2 + P 0 C 2 = PA02 + PB 02 + PC 02 : 68. 1995 Russian Math Olympiad Circles S1 and S2 with centers O1. O2 respectively intersect each other at points A and B: Ray O1 B intersects S2 at point F and ray O2B intersects S1 at point E: The line parallel to EF and passing through B intersects S1 and S2 at points M and N. Point C lies on the minor arc AB of the circle centered at O: Suppose the tangent line at C cuts the perpendiculars to chord AB through A at E and through B at F: Let D be the intersection of chord AB and radius OC: Prove that CE  CF = AD  BD and CD2 = AE  BF: 67. respectively. If triangles ABC and A0B 0 C 0 are equilateral. Let the inscribed circle of triangle ABC touchs side BC at D. 1998 IMO shortlisted problem Let ABCDEF be a convex hexagon such that 6 B + 6 D + 6 F = 360 and AB  CD  EF = 1: BC DE FA Prove that BC  AE  FD = 1: CA EF DB 11 . Prove that B is the incenter of 4EAF and MN = AE + AF: 66.Geometric Equations 64.

1999 IMO Two circles . .1 at C and D. R be the points of intersection of AC and BD. respectively. BRP. CE and DF.2 meets . Hexagon ABCDEF is inscribed in a circle so that AB = CD = EF: Let P. 1998 IMO shortlisted problem Let ABCD be a cyclic quadrilateral. Proposed by India for 1992 IMO Circles G1 and G2 touch each other externally at a point W and are inscribed in a circle G: A.. respectively. at the distinct points M and N. EA and FB respectively. and R are arbitrary points on the sides BC. Two circles intersect at points A and B: An arbitrary line through B intersects the rst circle again at C and the second circle again at D: The tangents to the rst circle at C and to the second circle at D intersect at M: The parallel to CM which passes through the point of intersection of AM and CD intersects AC at K: Prove that BK is tangent to the second circle.2 : 75. 1984 British Math Olympiad P.1 passes through the center of . and CPQ form a triangle similar to triangle ABC: 71. Q. and are tangent to .2 : The line passing through the two points of intersection of .1 and . C are 12 . and AB respectively of triangle ABC: Prove that the three circumcentres of triangles AQR. respectively. respectively. Let E and F be variable points on the sides AB and CD.1 and .Similar Triangles 70. Q. B. CA. 72. such that AE : EB = CF : FD: Let P be the point on the segment EF such that PE : PF = AB : CD: Prove that the ratio between the areas of triangles APD and BPC does not depend on the choice of E and F: Tangent Lines 73. The lines MA and MB meets . Prove that triangles PQR and BDF are similar. at A and B.2 are contained inside the circle . 74. Prove that CD is tangent to .

1982 IMO Diagonals AC and CE of the regular hexagon ABCDEF are divided by the inner points M and N. 1996 Putnam Exam Let C1 and C2 be circles whose centers are 10 units apart. Suppose A is a point inside a given circle and is di erent from the center. respectively.points on G such that A. B. and whose radii are 1 and 3. respectively. Find the locus of the circumcenter of 4PDE as P moves around the circle. Find the locus of all points M for which there exists points X on C1 and Y on C2 such that M is the midpoint of the line segment XY: Collinear or Concyclic Points 80. 81. C. Given 4ABC: Let line EF bisects 6 BAC and AE  AF = AB  AC: Find the locus of the intersection P of lines BE and CF: 79. E. so that AM = CN = r: AC CE Determine r if B. 13 . 77. prove that the four points are either collinear or concyclic. D are four distinct points such that every circle through A and B intersects or coincides with every circle through C and D. Perpendiculars from a point P on the circumcircle of 4ABC are drawn to lines AB. which is also tangent to G1 and G2 : Suppose AW is also tangent to G1 and G2: Prove that W is the incenter of triangle ABC: Locus 76. G1 and G2 are on the same side of chord BC. Consider all chords excluding the diameter passing through A: What is the locus of the intersection of the tangent lines at the endpoints of these chords? 78. BC with feet at D. M and N are collinear. 1965 Putnam Exam If A.

respectively. the diagonals AC and BD are perpendicular and the opposite sides AB and DC are not parallel. O2. Prove that EFZY is a cyclic quadrilateral. X is a point inside triangle ABC such that the incircle of triangle XBC touches BC at D also. BCP. then it is a cyclic quadrilateral. 83. respectively. in that order. PC and AD are concurrent. is inside ABCD: Prove that ABCD is a cyclic quadrilateral if and only if the triangles ABP and CDP have equal areas. O1. b. suppose AB AC: Let P and Q be the feet of the perpendiculars from B and C to the angle bisector of 6 BAC. 88. 84. ABP. 87. Prove that OP. O2O4 are concurrent. and touches CX and XB at Y and Z respectively. Let D be on line BC such that DA ? AP: Prove that lines BQ. where the perpendicular bisectors of AB and DC meet. CDP and DAP be O.82. 1995 IMO Let A. 85. then the points are collinear. E and F respectively. O3 and O4 . In 4ABC. 1970 Putnam Exam Show p that if a convex quadrilateral with sidelengths a. prove that if all the distances between every pair are integers. 1957 Putnam Exam Given an in nite number of points in a plane. 1995 IMO shortlisted problem The incircle of triangle ABC touches BC. Concurrent Lines 86. The circles with diameters AC and BD intersect at the points X and Y: The line XY meets BC at the point Z: Let P be a point on the line XY di erent from Z: The line CP intersects the circle with 14 . c. 1998 IMO In the convex quadrilateral ABCD. O1O3. CA and AB at D. C and D be four distinct points on a line. 1990 Chinese National Math Competition Diagonals AC and BD of a cyclic quadrilateral ABCD meets at P: Let the circumcenters of ABCD. d and area abcd has an inscribed circle. B. Suppose the point P.

Show that the diagonals AD. B1. and the line BP intersects the circle with diameter BD at the points B and N: Prove that the lines AM. respectively. at A meet the side BC: Let Q and P be the points in which the perpendicular at N to NA meets MA and BA. FD. are concurrent. B2 C2 meets at X . A: Suppose B1 C1. 89. C2. 92. CC 0 and HH 0 are concurrent. Perpendicular Lines 93.diameter AC at the points C and M. Q. Prove that QO is perpendicular to BC: 15 . respectively. BE. C1. where H and H 0 are the orthocenters of triangles ABC and AB 0 C 0 . respectively. Q. C2. CZ are concurrent. C. BC. A2. A2B2 meets at Z: Show that AX. 1988 Chinese Math Olympiad Training Test ABCDEF is a hexagon inscribed in a circle. BE. B2 . CF are concurrent if and only if AB  CD  EF = BC  DE  FA: 91. 1998 APMO Let ABC be a triangle and D the foot of the altitude from A: Let E and F be on a line passing through D such that AE is perpendicular to BE. and E and F are di erent from D: Let M and N be the midpoints of the line segments BC and EF. then show that the perpendicular from P. 1995 IMO shortlisted problem A circle passing through vertices B and C of triangle ABC intersects sides AB and AC at C 0 and B 0 . B1 . 2000 APMO Let ABC be a triangle. AD. EF. respectively. B2 . B. R are the midpoints of DE. 90. AF is perpendicular to CF. where the order of appearance along the triangle is A. and O the point in which the perpendicular at P to BA meets AN produced. C1A1 . A circle intersects a triangle ABC at six points A1 . respectively. A1. R to AB. DN and XY are concurrent. C2A2 meets at Y and A1 B1 . C1. CF are the altitudes of 4ABC: If P. Prove that BB 0 . respectively. BY. A2. respectively. Let M and N be the points in which the median and the angle bisector. Prove that AN is perpendicular to NM: 94. CA.

E to side BC respectively. respectively. AD. DE meet lines BC. 1985 IMO A circle with center O passes through the vertices A and C of triangle ABC and intersects the segments AB and AC again at distinct points K and N. P2. ! .. 1996 Chinese Team Selection Test The semicircle with side BC of 4ABC as diameter intersects sides AB. .95. 6 C of triangle ABC intersect its circumcircle at P. FD. N are collinear and the line through them is perpendicular to the line joining the orthocenter H and circumcenter O of 4ABC: Geometric Inequalities. CA. respectively. Show that L. AB in points L. 99. M. Let BB 0 and CC 0 be altitudes of triangle ABC: Assume that AB 6= AC: Let M be the midpoint of BC. E. respectively. respectively. N. N. Maximum Minimum 100. 1973 IMO Let P1 . P2n+1 be distinct points on some half of the unit circle centered at the origin O: Show that . 1997 Chinese Senoir High Math Competition A circle with center O is internally tangent to two circles inside it at points S and T: Suppose the two circles inside intersect at M and N with N closer to ST: Show that OM ? MN if and only if S. Q. T are collinear.. G be the feet of the perpendiculars from D. 6 B. Let M be the intersection of DG and EF: Prove that AM ? BC: 97. CF are the altitudes of 4ABC: Lines EF. Prove that AP + BQ + CR BC + CA + AB: 16 . .. ! . The circumcircles of triangles ABC and KBN intersect at exactly two distinct points B and M: Prove that OM ? MB: 98. . ! jOP 1 + OP2 +    + OP2n+1 j  1: 101. BE. R. AC at points D. Let F. : : : . Let the angle bisectors of 6 A. M. H the orthocenter of ABC and D the intersection of B 0 C 0 and BC: Prove that DH ? AM: 96.

mb . 17 . Q be points on EG such that BP and CQ are perpendicular to BC: Prove that BP + CQ  BC + EG: When does equality hold? 105. respectively. Solid or Space Geometry 106. mc are the lengths of the angle bisectors internal to the triangle and Ma . Proposed by Italy for 1967 IMO Which regular polygons can be obtained and how by cutting a cube with a plane? 107. Prove that 6 AIO  90 if and only if 2BC  AB + CA: 104. in general position i. they are not coplanar and any three are not collinear. 1995 Israeli Math Olympiad Four points are given in space. where ma . lc = mc =Mc .. sin2 A sin2 B sin2 C and that equality holds i ABC is equilateral. Problem 1506 Let I and O be the incenter and circumcenter of 4ABC. Prove that la + lb + lc  3. A plane  is called an equalizing plane if all four points have the same distance from : Find the number of equalizing planes. Mc are the lengths of the angle bisectors extended until they meet the circle. Assume 4ABC is not equilateral so I 6= O. 1997 APMO Let ABC be a triangle inscribed in a circle and let la = ma =Ma. Mb . 103. Mathematics Magazine. Point P is inside 4ABC: Determine points D on side AB and E on side AC such that BD = CE and PD + PE is minimum.e.102. lb = mb =Mb. Squares ABDE and ACFG are drawn outside 4ABC: Let P.

n2 an+1 + 2 + n2 an for all n  1: Do there exist indices p. 112. a3b3 . do the integers in An form an arithmetic progression? 18 . q and r such that apaq = ar ? Prime Factorization 114. : : :. Does the sequence a1. : : : become periodic after a nite number of terms? Modulo Arithmetic 111. : : :. 1995 Czech-Slovak Match Let a1 . b3. 113. 109.Number Theory Problems Digits 108. b2. a2. American Mathematical Monthly. Does there exist a positive integer a such that the sum of the digits in base 10 of a is 1999 and the sum of the digits in base 10 of a2 is 19992? 110. : : :. 11: Show that if each of the numbers a1b1. Problem E2684 Let An be the set of positive integers which are less than n and are relatively prime to n: For which n 1. Let a1. a11 and b1. 1956 Putnam Exam Prove that the number of odd binomial coe cients in any row of the Pascal triangle is a power of 2. a3. 2. : : :. 3. : : : be a sequence satisfying a1 = 2. a2b2. : : :. a2. a11b11 is divided by 11. then at least two of them will have the same remainder. 1956 Putnam Exam Prove that every positive integer has a multiple whose decimal representation involves all ten digits. an. a2 = 5 and an+2 = 2 . Proposed by USSR for 1991 IMO Let an be the last nonzero digit in the decimal representation of the number n!. a2. b11 be two permutations of the natural numbers 1.

at least one of the subsets contains three numbers a. 3. show that there exists a positive integer m such that the rightmost k digits of mr . Find all positive integers which cannot be written as the sum of two or more consecutive positive integers. 1983 IMO Can you choose 1983 pairwise distinct nonnegative integers less than 105 such that no three are in arithmetic progression? 118.115. : : : where the brackets denote the greatest integer function contains an in nite number of integral powers of 2. Find all even natural numbers n which can be written as a sum of two odd composite numbers. : : :. 4. 3 k = 2. 1971 IMO Prove that the set of integers of the form 2k . 2. b and c not necessarily distinct such that ab = c: Base n Representations 117. Representations 120. are all 1's. ng is partitioned into the union of two subsets. 3. Problem 2486 Let p be an odd prime number and r be a positive integer not divisible by p: For any positive integer k. 119. 122. Proposed by Romania p for 1985 IMO Show that the sequence fang de ned by an = n 2 for n = 1. when expressed in the base p. Proposed by Australia for 1990 IMO Observe that 9 = 4+5 = 2+3+4: Is there an integer N which can be written as a sum of 1990 consecutive positive integers and which can be written as a sum of more than one consecutive integers in exactly 1990 ways? 19 . 116. 1988 Chinese Math Olympiad Training Test Determine the smallest value of the natural number n 3 with the property that whenever the set Sn = f3. : : : contains an in nite subset in which every two members are relatively prime. 121. American Mathematical Monthly.

and show that 17 does not have this property. Due to Paul Erdos and M. Chinese Remainder Theorem 127. 125. b such that b 2 and 2a + 1 is divisible by 2b .. an is called quadratic if for each i 2 f1. 128. Find all positive integers a. : : :. Prove that every integer greater than 17 can be represented as a sum of three integers 1 which are pairwise relatively prime. Sur anyi Prove that every integer k can be represented in in nitely many ways in the form k = 12  22    m2 for some positive integer m and some choice of signs + or . where m. k are positive integers and k  2? Divisibility 129. : : :. 2. Show that if p 3 is prime. there exists a natural number n and a quadratic sequence with a0 = b and an = c: b Find the least natural number n for which there exists a quadratic sequence with a0 = 0 and an = 1996: 126.123. Proposed by North Korea for 1992 IMO Does there exist a set M with the following properties: a The set M consists of 1992 natural numbers. 1996 IMO shortlisted problem A nite sequence of integers a0. then pn cannot be the sum of two positive cubes for any n  1: What about p = 2 or 3? 124. 1: 20 .1 j = i2: a Prove that for any two integers b and c. jai .1 + 2 for all positive integers n: Prove that an integer value can be chosen for x0 so that x100 is divisible by 1998. 1988 Chinese Team Selection Test De ne xn = 3xn. a1. ai. b Every element in M and the sum of any number of elements in M have the form mk . ng.

1995 Russian Math Olympiad Is there a sequence of natural numbers in which every natural number occurs just once and moreover.2 from left to right.1 and An. Prove that there are in nitely many positive integers n such that 2n +1 is divisible by n: Find all such n's that are prime numbers. Determine all n such that An is divisible by 11. and so forth. let f n be the largest integer k such that 2k divides n and gn be the sum of the digits in the binary representation of n: Prove that for any positive integer n. y is an integer and divides 1995. Show that for any positive integer n. 1998 Putnam Exam Let A1 = 0 and A2 = 1: For n 2. For example. 1998 Romanian Math Olympiad Find all positive integers x. 3. 1995 Bulgarian Math Competition Find all pairs of positive integers 2 2 x + y x.1 + 1: Use this to nd all prime numbers p and q such that 2p + 2q is divisible by pq: 137.1 . 134. the number An is de ned by concatenating the decimal expansions of An. Show that there are in nitely many composite n such that 3n. 136. each of which is 1 or 2. A3 = A2 A1 = 10. A4 = A3 A2 = 101. A5 = A4 A3 = 10110. n such that xn + 2n + 1 is a divisor of xn+1 + 2n+1 + 1: 133. 2. : : : the sum of the rst k terms is divisible by k? 135. y for which x .1 is divisible by n: 131. 132. 1995 Bulgarian Math Competition If k 1. 2n.130. there is a number whose decimal representation contains n digits. and which is divisible by 2n: 138. For a positive integer n. for any k = 1. a f n! = n. show that k does not divide 2k.

g n. b 4 divides 2n n!n! 21 . n = 2n! if and only if n is not a power of 2. .

Proposed by Romania for 1985 IMO For k  2. n3 2n2 . 22 . 1969 Eotvos-Kursch Competition Let n be a positive integer. 1972 Putnam Exam Show that if n is an integer greater than 1. y such that a x and y are relatively prime. n1 2nk . b y divides x2 + m. b. there exist an in nite number of pairs of integers x. c is 1. c be positive integers such that 1 + a b c common divisor of a. 1. Proposed by Australia for 1992 IMO Prove that for any positive integer m. b. Find all integers n by n: 1 such that 1n + 2n +    + n . nk 2nk. 146. 1. c x divides y2 + m: 140. 1. Perfect Cubes 1 = 1 : If the greatest 145. 1n is divisible 141. Let a. then n does not divide 2n . 1: Prove that n1 = n2 =    = nk = 1: 143. Perfect Squares. p ak Math 2 Show that if 2 + 2 28n + 1 is an integer. : : :. 1997 Ukrainian Math Olympiad Find the smallest integer n such that among any n integers with possible repetitions.1 . : : : . then it is a square. nk be positive integers such that n2 2n1 . then prove that a + b must be a perfect square. n2. there exist 18 integers whose sum is divisible by 18. 1: 142. let n1 . 1998 APMO Determine the largest of all integer n with thepproperty that n is divisible by all positive integers that are less than 3 n: 144.139.

Show that if the equation x2 + y2 + 1 = xyz has positive integral solutions x. 7. y and z such that x  y  z and xy + yz = z x : 151. 157. also 1985 Moscow Math Olympiad If n  3. y odd positive integers. 153. c be integers such that a + b c a cube of an integer. Find all positive integral solutions to the equation xy + yz + zx = xyz + 2: 155. z. 1995 IMO shortlisted problem Let k be a positive integer. c. for any integers a. 1995 Czech-Slovak Match Find all pairs of nonnegative integers x and y which solve the equation px . 152. then prove that 2n can be represented in the form 2n = 7x2 + y2 with x. 1995 IMO shortlisted problem Find all positive integers x and y such that x + y2 + z 3 = xyz. where z is the greatest common divisor of x and y: 154. Find all integer solutions of the system of equations x + y + z = 3 and x3 + y3 + z 3 = 3: 23 . b. b. Let a. Find all sets of positive integers x. Diophantine Equations 150. where p is a given odd prime. b + c = 3: Prove that abc is the 149.147. then z = 3: 156. Sierpinski in 1955 Find all positive integral solutions of 3x + 4y = 5z . there exists a positive integer n such that n3 + an2 + bn + c is not an integer. 1998 Putnam Exam Prove p that. yp = 1. Due to Euler. where n is a positive integer. Due to W. Prove that there are in nitely many perfect squares of the form n2k . 148. y.

2. : : :. Show that one side of the triangle has length less than 1. nd the number of words of n letters which contain an even number of a's. c. The cells of a 7  7 square are colored with two colors. If 2n is replaced by 2n . 2g. For n 1. The set f1. is the statement still true in general? 165. 2. 1. 49g is partitioned into three subsets. 164. Find the number of n-words from the alphabet A = f0. Pigeonhole Principle 161. b. p  2. Find the number of possible colorings of these points with p colors. A2. if any two neighbors can di er by at most 1. Show that there are four pieces among them that formed the vertices of a parallelogram. 159. 160. : : :. let 2n chess pieces be placed at the centers of 2n squares of an n  n chessboard. 1. Show that at least one of the subsets contains three di erent numbers a. 1995 Romanian Math Olympiad Let A1 . 1996 Italian Mathematical Olympiad Given an alphabet with three letters a. 1989 Polish Math Olympiad Suppose a triangle can be placed inside a square of unit area in such a way that the center of the square is not inside the triangle. such that any two neighboring points have distinct colors. c such that a + b = c: 24 .Combinatorics Problems Counting Methods 158. An be points on a circle. Prove that there exist at least 21 rectangles with vertices of the same color and with sides parallel to the sides of the square. : : :. 163. 3000g contain a subset A consisting of 2000 numbers such that x 2 A implies 2x 62 A? 162. 1987 Austrian-Polish Math Competition Does the set f1. b.

: : : . lk are on a plane such that no two are parallel and no three are concurrent. : : :. l2. 2. b 1978 Austrian-Polish Math Competition There are 1978 clubs. 170. mg to Bn = f1. 2. 1996 Tournaments of the Towns In a lottery game. : : :. ng: 167. k . then prove that all 1978 clubs have a common member. 171. If no beetle makes either two horizontal moves or two vertical moves in succession. 1 so that in each of the lines l1 . It is known that for any subject exactly three students get the best score in the subject. 2. : : :. : : :. every beetle creeps to a horizontal or vertical adjacent square. there will be at least two beetles in the same square. Let m  n 0: Find the number of surjective functions from Bm = f1. 1 appear exactly once if and only if k is even. If every two clubs have exactly one common member. 3. 65 beetles are placed at di erent squares of a 9  9 square board. The 25 . 2. Determine the smallest n so that the above conditions imply that exactly one student gets the best score in every one of the n subjects. 2. k . a 1999 Hong Kong China Math Olympiad Students have taken a test paper in each of n n  3 subjects. 1995 Byelorussian Math Olympiad In the begining. Find the maximal number of 3-element subsets of A. Show that we k intersection points of these lines by the numbers can label the C2 1. Each has 40 members. In each move. a person must select six distinct numbers from 1. and for any two subjects excatly one student gets the best score in every one of these two subjects. Let A be a set with 8 elements.Inclusion-Exclusion Principle 166. such that the intersection of any two of them is not a 2element set. 168. lk the numbers 1. 36 to put on a ticket. l2. : : :. : : :. 1995 Greek Math Olympiad Lines l1. Combinatorial Designs 169. show that after some moves.

1991 Australian Math Olympiad There are n points given on a plane such that the area of the triangle formed by every 3 of them is at most 1. the triangle can be divided into 36 smaller equilateral triangles. 1998 Putnam Exam Let F be a nite collection of open discs in the plane whose union covers a set E: Show that there is a pairwise disjoint subcollection D1 . Is the statement true if 6 is replaced by 5? Covering. 1995 Byelorussian Math Olympiad By dividing each side of an equilateral triangle into 6 equal parts. A2 . Show that the n points lie on or inside some triangle of area at most 4. rn satisfying the following two conditions: a no three of the points A1. 36: Any ticket with numbers not containing any of these six numbers is a winning ticket. where 3D is the disc with the same center as D but having three times the radius. A2. A beetle is placed on each vertex of these triangles at the same time. Then the beetles move along di erent edges with the same speed. and real numbers r1 . When they get to a vertex. : : :. An lie on a line. : : : . An in the plane. j. Show that there is a scheme of buying 9 tickets guaranteeing at least a winning ticket. 3Dn covers E. : : : . 2. but 8 tickets is not enough to guarantee a winning ticket in general. they must make a 60 or 120 turn. 174. 172. 1995 IMO Determine all integers n 3 for which there exist n points A1. r2. : : :. 176. : : : . 3. k 1  i j k  n the triangle Ai Aj Ak has area equal to ri + rj + rk : 26 .lottery commitee will then draw six distinct numbers randomly from 1. Dn in F such that the union of 3D1 . : : :. b for each triple i. Convex Hull 173. 1969 Putnam Exam Show that any continuous curve of unit length can be covered by a closed rectangles of area 1=4: 175. Prove that at some moment two beetles must meet at some vertex.

177. the perpendicular bisector of the line segment AB is an axis of symmetry of S: 27 . 1999 IMO Determine all nite sets S of at least three points in the plane which satisfy the following condition: for any two distinct points A and B in S.

1 buttons. Show that there is 28 . they have just enough gas for one car to complete a lap. A boy takes n rides. 1995 USAMO A calculator is broken so that the only keys that still work are the sin.1 . 1995 Israeli Math Olympiad Two players play a game on an in nite board that consists of 1  1 squares. 1995 Russian Math Olympiad There are n seats at a merry-go-around. one from each school. cos. 179. tan. he moves clockwise a certain number less than n of places to a new horse. There are n identical cars on a circular track. the game is a tie. 181. 1977 Eotvos-Kursch ak Math Competition Each of three schools is attended by exactly n students. 182. cos. Is it possible to write a positive integer into each square of the rst quadrant such that each column and each row contains every positive integer exactly once? 184.Miscellaneous Problems 178. Each time he moves a di erent number of places. It is assumed that acquaintance is mutual. Show that player II can prevent player I from winning. who know one another. They play until one of the players marks a row or a column of 5 consecutive squares. and tan.1 . Is there a way to pack 250 1  1  4 bricks into a 10  10  10 box? 183. Find all n for which the boy ends up riding each horse. Prove that one can pick three students. 180. and this player wins the game. If no player can achieve this. player II chooses another square and marks it with X. show that pressing some nite sequence of buttons will yield q: Assume that the calculator does real number calculations with in nite precision. Each student has exactly n + 1 acquaintances in the other two schools. Among all of them. sin. Between each ride. All functions are in terms of radians. Then. The display initially shows 0: Given any positive rational number q. Player I chooses a square and marks it with an O.

Each lattice point of the plane is labeled by a positive integer. left. Each of these numbers is the arithmetic mean of its four neighbors above. n such that pk+1 divides p1 + p2 +    + pk for k = 1. there exists a permutation p1 . It is observed that in each pair. pn of 1. one can perform a sequence of these operations that will transform X to a one-digit number. 1984 Tournament of the Towns In a party. p2 . Can the sum of the numbers written at the vertices be the same as the sum of the numbers written at the edges? 186. 1991 Leningrad Math Olympiad One may perform the following two operations on a positive integer: a multiply it by any positive integer and b delete zeros in its decimal representation.a car which can complete a lap by collecting gas from the other cars on its way around the track in the clockwise direction. : : :. 1991 German Mathematical Olympiad Show that for every positive integer n  2. 190. below. n: 191. 1: 189. : : :. the di erence in height is less than 10 cm. 1995 Russian Math Olympiad Is it possible to ll in the cells of a 9  9 table with positive integers ranging from 1 to 81 in such a way that the sum of the elements of every 3  3 square is the same? 188. : : :. and on each of its edges is written the greatest common divisor of the numbers at the endpoints of the edge. Prove that for every positive integer X. 1996 Russian Math Olympiad At the vertices of a cube are written eight pairwise distinct natural numbers. 2. n . n boys and n girls are paired. 29 . 2. 185. right. : : :. Can the positive integers be partitioned into in nitely many subsets such that each subset is obtained from any other subset by adding the same integer to each element of the other subset? 187. Show that the di erence in height of the k-th tallest boy and the k-th tallest girl is also less than 10 cm for k = 1. 2. Show that all the numbers are equal.

17 red chips on a table. c . : : :. Prove that 1 no matter which sides of the coins are up initially. The following operations are permitted with the quadratic polynomial ax2 + bx + c: a switch a and c. c. adj. 2 in the above procedure. can you transform x2 . Is it possible after 1996 such steps to have numbers a. On each step we simultaneously replace each number by the di erence between this number and next number on the circle in a given direction that is. d. jab . 2. In one step. whether the head or the tail sides turned up at the end will depend on the initial placement of the coins. Can all chips become the same color after some steps? 196. c. b replace x by x + t. b. 195. d . d are replaced by a . 3. the k-th person is allowed turn over any k coins. the second person is allowed turn over any two coins. Proposed by India for 1992 IMO Show that there exists a convex polygon of 1992 sides satisfying the following conditions: a its sides are 1. b the polygon is circumscribable about a circle. 1? 30 . b. 194. x . : : : . the 1989 persons can come up with a procedure turning all coins the same sides up at the end of the operations. d such that the numbers jbc . b . where t is a real number. you may choose 2 chips of di erent colors and replace each one by a chip of the third color. 1992 in some order. 2 into x2 . bdj. the numbers a. There are 13 white. cdj are primes? 193.192. 1996 IMO shortlisted problem Four integers are marked on a circle. a. c. the 1989th person is allowed to turn over every coin. Some are placed with the head sides up and some the tail sides up. By repeating these operations. 15 black. A group of 1989 persons will perform the following operations: the rst person is allowed turn over any one coin. b. : : : . 1989 Nanchang City Math Competition There are 1989 coins on a table. jac . x .

we may turn over three coins which are mutually adjacent.197. d have the same color. : : :. then the intervals are colored the same way. can the numbers 21. 64. b . 1. 5 are written on a blackboard. 180. c. 31 . For intervals a. the goal is to make all of the coins turned tails up. the cells with at least 2 infected neighbors having a common side become infected. If this operation is performed repeatedly. 27. Nine 1  1 cells of a 10  10 square are infected. if a. which means a + x and c + x have the same color for x = 0. 540 ever appear on the board? 198. Initially. all of the coins are turned heads up. 1990 Chinese Team Selection Test Every integer is colored with one of 100 colors and all 100 colors are used.1990 and 1990 have di erent colors. On each turn. a: Prove that . 3. d having integers endpoints and same lengths. In one unit time. A student may erase any two of the numbers a and b on the board and write the numbers a + b and ab replacing them. Five numbers 1. 1997 Colombian Math Olympiad We play the following game with an equilateral triangle of nn + 1=2 dollar coins with n coins on each side. Can the infection spread to the whole square? What if nine is replaced by ten? 199. b . For which values of n can this be done? 200. c have the same color and b. 2. 4.

.

Solutions .

.

1: Comparing the degree 2 and 4 coe cients. 13 and px . So px+ p.x = 0 for all x: Then the even degree term coe cients of px are zero. Due to Law Ka Ho.1 x+1 divides px+p. 1 divides their sum px + p. A = 0. 1.x: Also x + 1 divides px .9=8: This yields px = . Ng Ka Wing. 15x=8: 3 3 3 3 3 3 3 2 5 3 2.3x =8 + 5x =4 . Tam Siu Lung Note x . Now px+1 = x. 3A = 0 and 3 + 3B . which is of degree at most 5. 1 divides px + 1 and p. 1 is divisible by x + 13 : Solution.x.x + 1.Solutions to Algebra Problems Polynomials 1. n: Find the value of P n + 1: r r! is of degree r: Consider the degree n polynomial . 1. Problem 7 Find without calculus a fth degree polynomial px such that px + 1 is divisible by x . Crux Mathematicorum.1 Ax +Bx. A polynomial P x of the n-th degree satis es P k = 2k for k = 0.x . we get B . 1 and p. 2. : : :. which implies A = . so x + 1 divides px+p. so x .x: Then x.3=8 and B = .

x .

x .

1.x Qx = 0 + 1 +    + n : By the binomial theorem. Qk = 1 + 1k = 2k for k = 0. 2. : : :. n: So P x = Qx for all x: Then .

n + 1 .

n + 1 .

1: 3. 1999 Putnam Exam Let P x be a polynomial with real coe cients such that P x  0 for every real x: Prove that P x = f1 x2 + f2 x2 +    + fnx2 35 . n + 1 P n +1 = Qn +1 = 0 + 1 +    + n = 2n+1 .

1    x . the polynomial = . r + 1 Solution.x xx . For 0  r  n.

2. b are symmetric with respect to the axis of the parabola. h: Then h1. g.for some polynomials f1 x.zk : Then a  0: Since P x  0 for every real x and a factor x . h8 will be the roots of the 4-th degree polynomial f gx: Since ha = hb. fnx with real coe cients. this would force A = 0. V x are polynomials with real coe cients. So Rx = f x2 for some polynomial f x with real coe cients. : : : . 5. 4. h2 = h7. Next pick one factor from each conjugate pair of nonreal factors and let the product of these factors x . 7. where U x. Solution. r repeated according to multiplicities and C x is the product of all conjugate pairs of nonreal root factors x. a = 6 b if and only if a. h4 = h5 and the parabola y = hx is symmetric with respect to x = 9=2: Also. 5. zk  be equal to U x + iV x. 8? Solution.zk x. it follows that h1 = h8. gh3. Due to Cheung Pok Man Write P x = aRxC x. We have P x = af x2 U x + iV xU x . 1995 Russian Math Olympiad Is it possible to nd three quadratic polynomials f x. where a is the coe cient of the highest degree term. 36 . Rx is the product of all real root factors x . 3. r n 2 +1 would change sign near a real root r of odd multiplicity. a contradiction. 6. gh2. iV x p p =  af xU x2 +  af xV x2 : 4. f2 x. gh4 are the roots of the quadratic polynomial f x. which implies h1+ h4 = h2+ h3: For hx = Ax2 + Bx + C. h3 = h6. we have either h1 h2 h3 h4 or h1 h2 h3 h4: Now gh1. hx such that the equation f ghx = 0 has the eight roots 1. : : : . each real root of P must have even multiplicity. 1968 Putnam Exam Determine all polynomials whose coe cients are all 1 that have only real roots. so gh1 = gh4 and gh2 = gh3. gx. h2. Suppose there are such f.

Take a = 1. 1990 Putnam Exam Is there an in nite sequence a0 . +1 7. we get a . x2 . C x with real coe cients such that a Ax  0. a1. we must have a = . x2 + x . C x  0 for all real x and b P x = Ax + xB x + 1 . 1991 Austrian-Polish Math Competition Let P x be a polynomial with real coe cients such that P x  0 for 0  x  1: Show that there are polynomials Ax. then by the AM-GM inequality. there must be another 0 1 0 +1 +1 +1 +1 +1 +1 +1 +1 sign change beyond T . Pn x has a sign change very close to every root of Pn x and has the same sign as Pn x at T: Since Pn x and Pn x take on di erent sign when x ! 1. a2.1 and n  3: By simple 0 1 1 2 1 2 2 1 2 2 1 0 2 2 1 2 1 2 2 checking. T .Solution. a = . : : :.1n =M . where M is chosen to be very large so that T n =M is very small. : : : . x2 . a = 1. then so is its negative. Hence we may assume a = 1: Let r .T. xC x for all real x: For example.T. Below all polynomials have real coe cients. Yes. x + 1: 6.xn =M is very close to Pn x on .T. Then Pn x = Pn x + . Then r +    + rn = a . 1. x3 . 2. we get the list 1 2 2 x . If a polynomial a xn + a xn. xx2 : Solution. then c  0 37 . 1. 3. x + 1. x3 + x2 . rn be the roots. 2a =n  an=n: Since a . x2 +1 . : : :. 1.1 and proceed by induction. x .1 as x ! 1 depending upon whether n is even or odd. Let an = . x . So Pn+1 x must have n + 1 real roots. x. Pn xj  T n =M for every x on . : : : of nonzero real numbers such that for n = 1. +   + an is such a polynomial. Suppose a . Suppose the roots of Pn x is in the interval . 2a and r    rn = an: If the roots are all real. T : So. then P x = 0+ x1 . an have been chosen so that Pn x has n distinct real roots and Pn x ! 1 or . We induct on the degree of P x: If P x is a constant polynomial c. if P x = x1 . B x  0. B x. T because jPn x . the polynomial Pnx = a0 + a1 x + a2 x2 +    + an xn has exactly n distinct real roots? Solution. 1.

5 . B x. Solution. x +1 . x +x C | 11 z. +1: Case x0 in .and we can take Ax = c. x0 Qx and Qx is of degree n with Qx  0 for all x n 0. zr j = 38 r Y i=1 jzin. then simply let Ax = P x. C0x  0 for all x in 0.xC0 x. we may assume b0 = 1 and bx = xr +     1: Since f 1 6= 0. For the case P x is of degree n + 1: If P x  0 for all real x. 1 : 8.1 + 3. 1993 IMO Let f x = xn + 5xn. 1 : Then P x = Q1 . : : :. Prove that f x cannot be expressed as a product of two polynomials. z1     . Ax Bx C x where the polynomials Ax. +1: Consider Qx = P 1 . each has integer coe cients and degree at least 1.1. xC1 x. where the polynomials A1 x. B x. x = x1 . x = A | 11z.5 . x  . 1 : Using x1 .1zi + 5j = 3r  9: . 1 : So Qx = A0x+xB0 x+1. 0 or 1.5j = j. B1 x. P x has a root x0 in . where n 1 is an integer.1. xC0 x 2 2 = . x0 A0 x + xB0 x + 1 . C x  0 for all x in 0. where A0x. | z Ax Bx C x where the polynomials Ax. x0 B0 x +1 . x 0 C0 x + x B0 x. x: This reduces to the previous case. x B | 11 z. Suppose f x = bxcx for nonconstant polynomials bx and cx with integer coe cients. B x = C x = 0: Next suppose the degree n case is true. 0 : Then P x = x . C x  0 for all x in 0. x 0 A0 x+ x B0 x +x A0 x . r 1: Let z . B x = C x = 0: Otherwise. Since f 0 = 3. C1 x  0 for all x in 0. x C0 x z | z | 2 + 1 . We have Qx = A1x + xB1 x + 1 . B0 x. x2 + 1 . 1 : Case x0 in 1. we have P x = x . x. xx2 . zr be the roots of bx: Then jz    zr j = jb0j = 1 and 1 1 jb.

P x = c for all x and so c = P . where each collection may contain repetitions. If the two 39 . : : : . x + a by x . then f x = x5 . b2. then also the product of each row is the same. bn be two distinct collections of n positive integers. 4 2 2 3 2 2 0 = bb4 + 3b2 c + c2 . a2. : : :. 2. for j = 1. 3a implies 3a = b5 . Let a1. Prove that if the integer a is not divisible by 5. 10. b2     x . n: Then the product of each row is . Solution. bgx or f x = x .1n+1 c: 11. Suppose f can be factored. : : :. x + a cannot be factored as the product of two nonconstant polynomials with integer coe cients. bx + c is 2 5 5 5 5 2 a factor of f x. b1. In the latter case. an. b2. a contradiction. b = a. 1x + b c + 2bc + a: Since x . an n  n table is lled as follows: into the cell in the i-th row and j -th column is written the number ai + bj : Prove that if the product of each column is the same. then deg P n: Now P bj  = bj + a1bj + a2    bj + an  = c. divding f x = x . b1x . both coe cients equal 0. 2. we get the remainder b + 3b c + c . a contradiction. 4 mod 5: Then 5 divides b . bx + c. 5bc2 is divisible by 5. 2. a2. some constant. b . : : : . bn: Therefore. : : : . c has distinct roots b1.5 = 3. n: So P x . : : :. Then a would be divisible by 5.1n+1 ai + b1ai + b2     ai + bn  for i = 1. b . 3b3c + 2bc2 + a = b5 . Solution. 1. an and b1. b . 1 . Let P x = x + a1x + a2     x + an .    . b + a = f b = 0: Now b b mod 5 by Fermat's little theorem or simply checking the cases b 0. then f x = x . b2. 9. 3. x .5 also divides f . 5bc2 . b.However. bn. bn. bx + cgx: In the fomer case.ai  = . Finally. 1991 Soviet Math Olympiad Given 2n distinct numbers a1. a contradiction.

3. Suppose xn = 0 and n is the least such index. then show that n is a power of 2. 1k Qx = x + 1 Qx : Setting x = 1. xn+1 = 12. then xn = tan n 1 +1 +1 +1 1 2 by induction. gx2 : Since f 1 . so f x . 2pn. which implies every qn is odd by induction. Now f x = 2 X n X i=1 x2ai + 2 X 1 i jn xai +aj = f x2  + 2 1 X i jn xai +aj : Since the ai + aj 's and the bi + bj 's are the same. Solution. 2pn: So qn divides qn . Write xn = pn =qn. n = 1. n = 0. g  x   x .1 : Recurrence Relations 12. f and g i+1 i=1 are distinct polynomials. Due to Wong Chun Wai The terms xn's are clearly rational by induction. 2xn Prove that xn 6= 1 or 0 for all n and the terms of the sequence are all 2 distinct. where pn. 1k Qx for some k  1 and polynomial Q such that Q1 6= 0: Then 2 2 2 k Qx2  2 f  x  . every xn = 6 : Next.collections of integers ai + aj 1  i j  n and bi + bj 1  i j  n are the same. 2. f x2  = gx2 . : : :: x1 = 2. Hence. If n = 40 . let = arctan 2. to show every xn = 6 0. g1 = n . we have n = 2k. 1 Q  x k f x + gx = f x . The sequence xn is de ned by + xn . Due to Law Siu Lung Consider the functions f x = n n X xai and gx = xbi : Since the ai's and bi 's are distinct. Solution . gx = x . so f x2 . gx = x . qn are relatively prime integers and qn 0: Then q = 1 and pn =qn = 2qn + pn=qn .

1 n +1 . b n n . then 0 = x2m = tan 2m = 2xm =1 . Due to Chan Kin Hang Since an depends on an and +2 +1 . Therefore the terms are nonzero and distinct.2m is even. 13. writing out the rst few terms of 9anan+1 + 1 will suggest that 9anan+1 + 1 = an + an+12 : The case n = 1 is true as 9  7 + 1 = 1 + 72 : Suppose this is true for n = k: Using the recurrence relations 2 and * 2a2 k+1 . xm  would imply xm = 1  5=2 is irrational. ak 2 = ak+1 + ak+22 : 41 Solution.1 + bn. ak + ak+1  + 2 2 = 62a2 k+1 . if xm = xn for some m n. ak + 2 2 = 64a2 k+1 . 16ak+1ak + ak by * = 8ak+1 . bn+2 = 7bn+1 . 2ak ak+1 .2ak . 1988 Nanchang City Math Competition De ne a1 = 1. c0 = . then 2 2 b 7 b .2 = 2ak ak+2 = 14ak ak+1 . Finally. b2 = a2. a contradiction. 1 for positive integer n: Prove that 9a a + 1 is a an+2 = an+1 n n+1 an perfect square for every positive integer n: an. a contradiction. ak  + 1 2 = 63a2 k+1 .2: Then . bn for n  1: Then b3 = 48 = a3 : Suppose ak = bk for k  n + 1. If n = 2m + 1 is odd. bn = bn+2: So ak = bk for all k: Next. 1 an+2 = b = bn n = 49bn . x2 m  would imply xm = 0.2 = 2xm =1 . a2 = 7 and 2 . it is plausible that the sequence satis es a linear recurrence relation an+2 = can+1 + c0an : If this is so. then 0 = x2m+1 = tan + 2m  = 2 + x2m =1 . 1 . then using the rst 4 terms. then xm. we nd c = 7. 2x2m  would p imply 2 x2m = . we get the case n = k +1 as follow: 9ak+1ak+2 + 1 = 9ak+17ak+1 .n = tanm . a contradiction to n being least. xn=1 + xm xn = 0.1: De ne b1 = a1. 14bn. n  = xm .2 = 7bn+1 .

American Mathematical Monthly. m + 1. b must be an integer.x + y.1 is 1.1 of degree n . Since m c +1 is integer.e. pBn is n+1 p 2 2mod n odd. 2Bn = . Since b is a root of the integer coe cient polynomial fm z  . c = xy: Suppose tn is an integer for m.1 b for some polynomial fn. let tn = xn . Nextp suppose case k is true. Bn 2: Multiplying these 2 equations. For nonnegative integer n. m + 1. where b = . m + 3: Since cn = xyn = t2 n +2 . Next tm+1 tm+2 : b = tm tm+3 . Proposed by Bulgaria for 1988 IMO De ne a0 = 0. we nd that an = Bn : Now write n = 2k m. i. in fact.1n : This implies An is always odd. 1 with integer coe cients. Problem E2998 Let x and y be n n . 42 . so B2n = 2AnBn : Then it follows case k implies case k + 1: 15. it follows by induction that tn = fn. y: So t0 = 0. Since 1+ 2 = An + Bn 22 = A2n + B2n 2. By the binomial theorem.1 + an. m + 2.2 for n 1: Show that for positive integer k.e. y is an integer for all positive integers n: Solution.1 in fn. cm So b is rational. yn x four consecutive positive integers n: Show that x . y is an integer for some distinct complex numbers such that xx . a1 = 1 and an = 2an. Using characteristic 2 2 equation method to solve the given recurrence relations on an. if 1 + 2 = A + B 2. tm+1 = 0. fn. So the recurrence relation implies all tn 's are integers.14. 2 = An . an is divisible by 2k if and only if n is divisible by 2k : p n p Solution .1 is monic. so c is rational. t1 = 1 and we have a recurrence relation tn+2 + btn+1 + ctn = 0. n is odd if and only if 2Bn = A2 4. We have k = 0 2 i. where m is odd. yn =x . Note the coe cient of xn. then n n p p n 1 .e. we get An . i. y n . c must. be an integer. tn tn+2 is an integer for n = m. From the recurrence relation.

a2. : : :. n + n . : : :. Due to Leung Wai Ying Since abc  1.1 . +2n .Inequalities 16. 2b=a + a=c  3b and 2c=b + b=a  3c: Adding these and dividing by 3. Let a. an: Solution. 1=ac  b and 1=ab  c: By the AM-GM inequality. then prove that An. 1 X = 2 ak. we easily see the pattern. b.1 + An+1  2An for n = 2. Expressing in ak . we get the desired inequality. 2 a + nn . we get 1=bc  a. 2ak + ak+2  0: k=2 17. the required inequality is equivalent to a +    + a . 3.1 + an+1  2an for n = 2.1 . 3. : : :. : : :. 2 an + nn2. where An is the average of a1 . a2. if an. 3. 1 an+1 n kk . For real numbers a1. 43 r . c 0 and abc  1: Prove that a + b + c  a + b + c: c a b Solution.1 2 2 n 2 n+1 From the cases n = 2. a3. We have 2 n a1 +    + an. 2a + c = a + a + c  3 3 a2  3a: c b c c b bc Similarly. 1 a  0: 1 n.

without loss of generality. a + a . 1997 IMO shortlisted problem Let a1      an  an+1 = 0 be a sequence of real numbers. Prove that v u n n u tX ak  X pkpak . 2a  = a a . For n = 1. 1982 Moscow Math Olympiad Use the identity 13 + 23 +    + n3 = n2n + 12 to prove that for distinct positive integers a . b = z 2 x. we get a + b + c = x2 + z 2 + y2 c a b yz xy zx . we may assume a1 a2 : : : ak+1: Now 3 2 3 3 2 2a3 1 +    + ak+1  . 1  2 2  a k +1 . For the case n = k + 1. a . Equality occurs if and only if a1.2 and z = 9 b4 c=a2 : Alternatively. : : :. a . an are 1. y = 9 c4a=b p We have a = x2 y. a2. Suppose the case n = k is true. : : : . c = y2 z and xyz = 3 abc  1: Using this and the rearrangement inequality. let x = 9 a4b=c2. pak : k=1 k=1 +1 44 . x2 + z2 + y2  = x3 + y3 + z3 xyz yz xy zx p p p x2 y + y2 z + z 2x = a + b + c: 18. 2. 1  0 and so case 7 1 5 1 3 2 1 5 1 1 2 n = 1 is true. n: 19. 2a1 +    + ak  3 3 3 =2a6 k+1 + 4ak+1a1 +    + ak  3 3 3 3 2a6 k+1 + 4ak+11 + 2 +    + ak+1 . : : :. 1 2 n 2 7 7 5 5 5 3 3 3 2 a7 1 + a2 +    + an  + a1 + a2 +    + an   2a1 + a2 +    + an  : Can equality occur? Solution. 1 ak 6 3 7 5 =2ak+1 + 4ak+1 = a k +1 + ak+1 : 2 7 5 5 3 3 2 So a7 1 +    + ak+1  + a1 +    + ak+1   2a1 +    + ak+1  follows.

we get the desired inequality. Due to Lee Tak Wing Let xk = pak . ad + dc + cb + be + ea d + da + c + ea + b = ad + e + bc + e + cb + 2 e + b + d + a + c + a + b  a + b + c + d + ed + e + c + 10 =2 5: 21. Due to Lau Lap Ming Since a  b  c  d  e. 1985 Wuhu City Math Competition Let x. n n n X X X X ak = xk + xk +    + xn  = kxk + 2 ixixj +1 2 +1 k=1  k=1 n X k=1 +1 2 2 kxk + 2 2 X p 1 Taking square root of both sides. y.x + y and so x2 + y2 + z 23 = x2 + y2 + x + y2 3 . z be real numbers such that x + y + z = 0: Show that 6x3 + y3 + z 3 2  x2 + y2 + z 23 : Solution. x + y2 3 = 27 x + y4 x + y2  3 2 8 .  . 20. Due to Ng Ka Wing We have z = . pak : Then ak = xk + xk +    + xn  : So.Solution. so d + e  c + e  b + d  a + c  a + b: By Chebysev's inequality. x + y3 2 = 6x3 + y3 + z 32 : 45 . 1994 Chinese Team Selection Test For 0  a  b  c  d  e and a + b + c + d + e = 1. ijxi xj = kxk : 2 k=1 1 i jn k=1 Solution. show that ad + dc + cb + be + ea  1 5: i jn n p  X .  27 p  8 2 xy 4 x + y2 = 6 3xyx + y 2 = 6 x3 + y3 .

it is well-known that the discriminant 4 = x . 27c2  0: For the cubic polynomial f w = w3 + bw + c. xw . with n  2: a Determine the least constant C such that the inequality X 1 i jn xi xj xi + xj   C 2 2 X 1 in xi 4 holds for all nonnegative real numbers x1 .2b and 0 = f x+ f y+ f z  = x3 + y3 + z 3 + bx + y + z  + 3c implies x3 + y3 + z 3 = . 1999 IMO Let n be a xed integer. Overall. determine when equality holds. yw . 2xy + yz + zx = . X 1 in xi = 4 X 1  2 2 =8 1 s X X X 1 in xi + 2 2 X i jn 1 1 i jn xi xj X x2 i xi xj 2 2 8 i jn i jn 2 xi xj x2 1 +    + xn  2 xi xj x2 i + xj : in 1 Equality holds in the second inequality if and only if at least n . z 2 z .Comments. x2. Let f w = w . 22. 46 . Then equality holds in the rst inequality if and only if the remaining pair of xi 's are equal.4b3 . : : : . z  = w3 + bw + c: Then x2 + y2 + z 2 = x + y + z 2 . xn: b For this constant C.3c: So the inequality is the same as 2.4b3 . equality holds if and only if two of the xi 's are equal and the others are zeros. Solution. y2 y . 27c2 : The inequality follows easily from this. x2 equals . Due to Law Ka Ho and Ng Ka Wing We will show the least C is 1=8: By the AM-GM inequality. 2 of the xi 's are zeros.

x1 x2 + x2 x3 +    + xn.1 xn + xn x1   2 . jf 0 + g0 + h0j 3 3 1: jf x + g0 + hz j 1 . y. x2  . the expression above is at most n: So 1 1 2 2 3 1 2 . : : : . z in 0. j f  x  + g  y  + h 0 j 3 3 47 . : : :. x2 x3 . where x is the greatest integer less than or equal to x: Solution. x1 x2 . xn1 . xyz j  1 3: Solution. x1 1 . 1 . 1 . every xi 's is 0 or 1 and the value is an integer.1 xn + xn x1  =n . x1  . For every triplet of functions f. 1995 Bulgarian Math Competition Let n  2 and 0  xi  1 for i = 1. x x + x x +    + xn. prove that there are numbers x. jf x + gy + hz  . so the maximum value is attained when x1 = 0 or 1. So when the left side is maximum.  hni . xn are xed. Now 2 x1 +    + xn  . x1 x2 + x2 x3 +    + xn. x3  .23. Suppose for all x. the left side is a degree one polynomial in x1 . x2 1 . : : : . z in 0. 1 such that jf x + gy + hz  . h : 0. xn + xn x   2 : 1 24. xnx1 : Since 0  xi  1. When x . 1 . xyz j 1=3: Then 1 .  max x +    + xn  . : : :. y. 1 . 2. 1 ! R. jf 0 + gy + hz j 1 . The situation is similar for the other xi 's. n: Prove that hni x1 + x2 +    + xn . g.

the sum of j . f 1+ g1+ h1 . then 2 + xyz . jf 1 + g0 + h1j. 1j. jf x + gy + hz j 2=3: In particular. g0 . then show that x + y + z  xyz + 2: say z. say 0 x  y  z: If z  1. z = 1 . z is nonpositive. j. z are real numbers such that x2 + y2 + z 2 = 2. xy  0 because x + y  2x2 + y2   2 and xy  x2 + y2 =2  1: So we may assume x.f 0 .Since + hz  f x + gy + hz  = f 0 + g2y + hz  + f x + g0 2 y + h0 + . Proposed by Great Britain for 1987 IMO If x. y . By the triangle inequality. jf 0 + g1 + h1j. b. c. x1 . jf 1 + g1 + h1j 2=3: However. a contradiction. y . x . There is a simpler proof. + f x + g 2 2 by the triangle inequality. Due to Chan Ming Chiu If one of x. j. one of them is at least 2=6: 25. then x + y + z  2x + y2 + z 2 = 2 xy + 1  xy + 2  xyz + 2: 26. we get 1 1. y + 1 . j1 . 1j is at least 2. then Solution. g1 . z = 2 . y . Proposed by USA for 1993 IMO Prove that for positive real numbers a. f 0 + g0 + h0j. jf 1+ g1+ h0j. d. So. f 1+ g1+ h1 . h0 . x . z 1 . 2 + xyz . y. f 1 . x . a b c d 2: + + +  b + 2c + 3d c + 2d + 3a d + 2a + 3b a + 2b + 3c 3 48 p p p . z 1 . y. xy  0: If z 1. h1j 1=3: Adding these two inequality and applying the triangle inequality to the left side. y. z are positive.

x2 1 +    + x4 y1 +    + y4   a + b + c + d : To 2 2 2 2 nish. c + c +    + ck =k  p k c c    ck  1. 1a + c . : : :. bn be 2n positive real numbers such that a a1  a2      an and b b1 b2    bk  a1 a2    ak for all k.Solution. 1an =d a + d . a  + d a . an and b1. an =d a . d2 + c . y3 = cd + 2a + 3b. a  +    + dnan  0: 1 2 1 2 1 2 1 1 2 1 2 1 2 2 1 1 1 2 1 2 1 1 2 2 2 3 49 . 1a +    + cn . it su ces to show a + b + c + d2 =y1 + y2 + y3 + y4   2=3: This follows from 2 2 2 2 3a + b + c + d2 . y2 = bc + 2d + 3a. c2 + b . 1+c . y4 = da + 2b + 3c: 2 2 2 The inequality to be proved is x2 1 + x2 + x3 + x4  2=3: By the Cauchy2 2 2 2 Schwarz inequality. r c p x3 = d + 2a + 3b . Let ck = bk =ak and dk = c . 1+    +ck . : : :. b2. r d p x4 = a + 2b + 3c . Let p b x2 = c + 2d + 3a . d2 + b . p Solution. d2  0: 27. dn. Let a1 . 2y1 + y2 + y3 + y4  =3a + b + c + d2 . r r y1 = ab + 2c + 3d. which implies dk  0: Finally. 1 for 1  k  n: By the AM-GM inequality and b. c2 + a . 1  k  n: Show that b1 + b2 +    + bn  a1 + a2 +    + an: a x1 = b + 2c + 3d . d a +    + dn . b + b +    + bn  . a2. a + a +    + an =c . b2 + a . 8ab + ac + ad + bc + bd + cd =a .

28. c 0 and m be a positive integer. Proposed by Greece for 1987 IMO Let a. b. prove that .

show that a1 + a2 +    + an  1 . 1 . 9 b+c m  a +3 2a + b + c . a . an into increasing order as b . : : : . 1 : 2 8 n2n 2n 50 1 2 . which implies b + c c+a a+b Solution. by the rearrangement inequalpositive integers. : : :. Let a1.1: =3 2 3 29. we have am + bm + cm  1 am + bm + cm . a + b + c m. bn: Then bn  n because they are distinct 1 2 1 . 1 : 2 8 n2n 2n Solution.1 : b+c c+a a+b 2 3 inequality. : : :. Due to Chan Kin Hang Arrange a . 1 . assume a  b  c: So a + b  1  1  1 : By the AM-HM c + a  b + c. a2. b . b + c + c + a + a + b  3 1 + 1 + 1 : 3 b+c c+a a+b 9 1 + 1 + 1  This yields b + c c + a a + b 2a + b + c : By the Chebysev inequality and the power mean inequality respectively. Since 2 8 n2n ity. am + bm + cm  3 a + b + c m. Without loss of generality. 1 + 1 + 1  b+c c+a a+b 3 b+c c+a a+b  . an be distinct positive integers. : : :. a1 + a2 +    + an  b1 + b2 +    + bn 2 8 n2n 2 8 n 2n 1 + 2 + + n = 1 .

a : For m = 0. ai i=1 n X n na X i : Since  2a . b. we get i=1 a = 1 + 1 ai . a2. ai 2n . 1982 West German Math Olympiad If a1. let a = a if i j mod n: i j 2a . ai 2a . a+b b+c c+a b+c c+a a+b a3 + b3 + c3  a3 + b3 + c3 : c+a a+b b+c b+c c+a a+b 51 Adding these n inequalities. by the rearrangement inequality.30. we get 2a . Due to Ho Wing Yip By symmetry. 1 =1 1 0 and a = Solution. then b + c + c + a + a + b  : 2 Solution. we get 2 1 n a X m i=1 n X ai :  2a . ai 2a . we may assume a  a  : : :  an: Then 1  : : :  1 For convenience. : : : . an 2a . ai 2 2 2a . Prove that if a. we may assume a  1  1  1 : By the b  c: Then a + b  c + a  b + c: So b + c c+a a+b rearrangement inequality. n . c 0. ai a i=1 . then show that n X ai  n : i 2a . we have a3 + b3 + c3  a3 + b3 + c3 . ai + i n n n+1X ai  n X ai : 2 2 i=1 2a . an a1 + a2 +    + an. By symmetry. 3 3 3 2 2 2 a b c a + b + c 31. ai Solving for the sum. 1. we get the desired inequality. : : :. ai i=1 2a . 1.

since x3 + y3 =x + y = x2 . The least common multiple of any two of these is greater than 1951. we have a2 + b2 + c2 = 1 . xy + y2  x2 + y2 =2. a2 + b2 + b2 + c2 + c2 + a2  2 2 2 2 2 3 3 3 3 3 3 . we get 1 . d = tan : Then cos2 + cos2 + cos2 + cos2 = 1: By the AM-GM inequality.Adding these. 1 a + b b + c c + a  2 a+b + b+c + c+a 3 b3 + c3 : +  ba +c c+a a+b 32. c = tan . b. : : :. Let a. sin2 = cos2 + cos2 + cos2  3cos cos cos 2=3 : Multiplying this and three other similar inequalities. an is less than 1951. Let a = tan . d 0 and 1 + 1 + 1 + 1 = 1: 1 + a4 1 + b4 1 + c4 1 + d4 Prove that abcd  3: Solution. b = tan . we have sin2 sin2 sin2 sin2  81 cos2 cos2 cos2 cos2 : Then abcd = tan tan tan tan  3: 33. then dividing by 2. c. Due to Paul Erdos Each of the positive integers a1. a3 + b3 + b3 + c3 + c3 + a3   a3 + b3 + c3 : 2 a+b b+c c+a b+c c+a a+b Finally. Show that 1 ++ 1 1 + n : a1 an 1951 52 2 2 2 2 p .

1951 is 1951=ai : So we have 1951=a +  + 1951=an  1951: Since x . 1951 is a common multiple of more than one ai 's.1. P x = 1. : 2 2 k k=0 k=0 p By induction. we can show that 0  Pn x  x  1 for all x in 0. 34. : : :. px  n+1 . P x n . Pn+1x = Pn x + x . 2. The number of multiples of ai among 1. 1 : .1 px . 2. Pn x  n + 1 for every nonnegative integer n and all x in 0. : : :. 1 1951: a an 1 Then px . 1 . 1 x . px  +  2 + 1 . P x  px. px  Y Y n n k +1 k px = px .1 . so 1 Dividing by 1951 and moving the negative terms to the right. 1951 . px n 0 n px + 1 . Due to Wong Chun Wai For x in 0. P2n x : Prove that p 2 0  x . Observe that none of the numbers 1. we have px . px + P x  . we get the desired inequality. P x =  x . 1 +    + . p p n : x . 1951 .Solution.  1 . n+1 n 2 Multiplying both sides by x and applying the AM-GM inequality. px + P x  . P x n . P x 1 . 1 : Solution. A sequence Pn  of polynomials is de ned recursively as follows: 2 P0 x = 0 and for n  0.

n 2 2 2 2 n n+1 2 . n n+1  2 : =n n+1 n+1 53 p .

then jaj + jbj + jcj + jdj = a + b . 1  P 1 . we see that the maximum 7 is obtained by P x = 4x3 . 2P 0  3: Case c in .35. c. c .1. Problem 4426 Let P z  = az 3 + bz 2 + cz + d. 3x only.1 . where a. 8 P  1  + 8 P . 36. +1: If d  0. then jaj + jbj + jcj + jdj = a + b . Tracing the equality cases. b  0: Case c in 0. So we may assume a. d 1  + 4 P . b. 0: If d  0. 4 P  3 2 3 2 5 + 4 + 4 = 7: 3 3 Comments.d = P 1 . 1996 IMO shortlisted problem Let P x be the real polynomial function. American Mathematical Monthly. b  0: The problem stays the same if P x is replaced by this polynomial. then jaj + jbj + jcj + jdj = a + b + c + d + 2. Note the four polynomials P x satisfy the same conditions as P x: One of these have a. d arep complex numbers with jaj = jbj = jcj = jdj = 1: Show that jP z j  6 for at least one complex number z satisfying jz j = 1: 54 . 1  =5 P 1 . then jaj + jbj + jcj + jdj = a + b + c + d = P 1  1: If d 0. c + d 1 P . =4 3 3 3 2 3 2 1 + 8 + 8 = 7: 4 + 3 3 3 3 If d 0. then jaj + jbj + jcj + jdj  7: Solution. P x = ax3 + bx2 + cx + d: Prove that if jP xj  1 for all x such that jxj  1.

then for any n. mk+1 j=2k+1 . that is mk =2k = mk+1 =2k+1 for every k: Then jx . we get Solution. the inequalities imply it is 0. then jP z j2 = 4 + 2Re Qz : Now we use the roots of unity trick! Let ! be a cube root of unity not equal to 1. jP !2 z j is at least 6: 37. 2. 38. take m = nx: Conversely. d: Using w + w = 2Re w. If x . suppose the condition holds for x: Let mk be the integer corresponding to n = 2k . Due to Yung Fai We have aa = jaj = 1 and similarly for 2 jP z j2 = az 3 + bz 2 + cz + daz 3 + bz 2 + cz + d = 4 + 2Re adz 3 + ac + bdz 2 + ab + bc + cdz : . denote by pn the number of ways of expressing n as the sum of one or more positive 55 Solution. 1997 Hungarian-Israeli Math Competition Find all real numbers with the following property: for any positive integer n.b. then jz j = 1 and Re Qz  + Re Q!z  + Re Q!2 z  = 3: So jP z j2 + jP !zp j2 + jP !2 z j2 = 18: Then one of jP z j. mk =2k j  1=3  2k  for every k: Therefore. mk+1  mk . jP !z j. Since 1 + ! + !2 = 0 and 1 + !2 + !4 = 0. The condition holds if and only if x is an integer. there exists an 1? integer m such that . c. 1979 British Math Olympiad If n is a positive integer. mk+1 2k 2k+1 2k 2k+1 3  2k 3  2k+1 2k+1 Since the leftmost expression is j2mk . : : :: By the triangle inequality. x = m0 is an integer. x + x . m0 j = jx . k = 0.  Let Qz  = adz 3 + ac + bdz 2 + ab + bc + cdz. so Qz  + Q!z  + Q!2 z  =3adz 3 + ac + bd1 + ! + !2  + ad + bc + cd1 + !2 + !4 z =3adz 3: If we now choose z to be a cube root of ad. 1. m n 3n is an integer. 1 + 1 = 1 : mk .

f x = 0 or . n: If c = 6 0. pn . removing that 1 will results in a sum of n . Thus p4 = 5. 1 will yield pn . namely 1 + 1 + 1 + 1. we see that z and z . 1. 1n for some real c and nonnegative integers m. 1j also equals 0 or 1. 1 b f x . It follows that z = 0 or 1. then f is 0 or . then after simplifying the functional equation. The required inequality can be written as pn + 1 . pn . 1: Note that adding a 1 to each pn . Since z is a root implies z . 1: So pn . for each sum of n whose least summand is 1. If f is constant. 1 is the number of sums of n whose least summands are at least 2. 1 ! R be a function such that a f is strictly increasing.xn 1 . Since f has nitely many roots and z are all roots. pn . 1997 Greek Math Olympiad Let f : 0.integers. xn for nonnegative integer n: 2 2 2 2 40. 1 sums of n: Conversely. 1 is a root. as there are ve di erent ways of expressing 4 in terms of positive integers. Then f x = cxm x . then let z be a root of f: Setting x = z and x = z . 1: Functional Equations 39. 1 + 3. so we must have jz j = 0 or 1. increasing the largest summand by 1 will give a sum of n + 1 with least summand at least 2. we will see that n = m and c = 1: Therefore. and 4: Prove that pn + 1 . pn . 1n are also roots. pn  pn . 2 + 2. Find all polynomials f satisfying f x2  + f xf x + 1 = 0: Solution. 2pn + pn . x for all x 0 and 1  = 1 for all x 0: c f xf f x + x Find f 1: 56 . So pn + 1 . For every one of these pn . pn  pn . 1 sums of n. 1 + 1 + 2.1: If f is not constant. 1 sums of n . respectively. respectively. 1  0 for each n 1: Solution. jz .

we get f t +1f f t + 1+ t  = 1: Then f  t + t  = t = f 1 increasing.f x: By induction. we get f . Proposed by Ireland for 1989 IMO Suppose f : R ! R satis es f 1 = 1. Due to Yung Fai From f 0 + 0 = f 0 + f 0. so x  . Let t = f 1: Setting x = 1 in c. p: Since f is strictlyp t + t = 1: Solving. we have 1 1 1 1 j 2: So thereqis w such that w + w = nx : We have jf  nx j = j nx 1 1 jf w + f  w j  2 f wf  w  = 2: So jf xj = 1 1  21 : Then njf  nx j n lim f x = 0 = f 0: x!0 57 .x = .x.p f 1 = t = 1 .x = 0. we get f 0 = 0: From 0 = f x + . 5=2: Note f x = 1 .x  .x  x + . The so-called Cauchy's equation f a + b = f a + f b implies f x = f 1x for rational x: Next we will show f is continuous at 0: For 0 jxj 21n . f nx = nf x for positive integer n: For x = n . a contradiction.f . 0 = f x + . Therefore. f x = x for all real x: 42. so 0  f 0: Since f 0  0 also. m 1 1 1 1 1 1  = nf  n : Then f  n = n and f  n  = f m n = 1 = f 1 = f n n 1 m mf  n  = n : So f x = x for rational x: The argument up to this point is well-known. then 1 t = fp 1 f 1 + t = t 1. hence . y: Prove that f x = x for every real number x: Solution. b 2 R and f xf  1 x  = 1 for x 6= 0: Show that f x = x for all x: Solution. we get f 0 = 0: For all real x. we get t = 1  5=2: If t = 1 + 5=2 0.x = 0: So f x + f .f .x  f x + f .x = f x for all real x: Since f .x = f x + f . Due to Ng Ka Wing Since f 0 + 0  f 0 + f 0.Solution. 1979 Eotvos-Kursch ak Math Competition The function f is de ned for all real numbers and satis es f x  x and f x + y  f x + f y for all real x.x. f a + b = f a + f b for all a. 1 1 +1 1 +1 1 1 +1 1 41.x = f x  x: Therefore. 5=2x is such a function. we get tf t + 1 = 1: So t= 6 0 and f t +1 = 1=t: Setting x = t +1 in c.

1. + p3 + 3p2 + 3pq3 + q6 = t3 + 3p2 + 3pq3 + q6 2 3 + q   = f  f  p q q3 and on the other hand. we get f x + n = f x + n for p all positive integer n: For q 2 Q . From f x + 1 = f x + 1. Then nlim !1x . let t = f  p q : On one hand. f x = nlim !1f x . p  + q2 3 = t3 + 3t2q2 + 3tq4 + q6 : 2 3 f  p + q   =  f  q q Equating the right sides and simplifying the equation to a quadratic in + t. 1992 Polish Math Olympiad Let Q+ be the positive rational numbers. f x = x for all x: This rst and third paragraphs show the Cauchy equation with continuity at a point has the unique solution f x = f 1x: 43. rn  + f rn  = nlim !1 rn = x: Therefore. 1996 IMO shortlisted problem Let R denote the real numbers and f : R ! .Now for every real x. we get the only positive root t = p q : So f x = x for all x 2 Q : 44. rn = 0: By continuity at 0. Determine all functions f : Q+ ! Q+ such that f x + 1 = f x + 1 and f x3  = f x3 for every x 2 Q+ : Solution. let rn be a rational number agreeing with x to n places after the decimal point. 1 satisfy 1 +f x+ 1 f x + 13 + f  x  = f x + 42 6 7 .

.

.

we will get f w + 7  + f w  =  for all w: Setting w = z + k for k = 0. we get 6 equations. 5. 6 f w + 1 + f w + 1 7 7 6 58 . 1. : : :. : : :. there is a nonzero real number T such that f x + T  = f x for every x 2 R: Solution.e. i. for every x 2 R: Show that f is a periodic function. 1. 8 Adding these and cancelling terms. Setting x = w + k for k = 0.

say c: If c 6= 0. then f n = 2ef q: To de ne such a function. 1: Then s2  3k + j  = ss3k + j  = 33k +j : Since 1997 = 236 +539 37. a contradiction. 1 + f z  Solution. then 3m = ssm = ssn = 3n implies m = n: From this.1. 3990 if q 3991 mod 3992 1997 59 . 1 is a constant. we get 7 equations. we see that s is strictly increasing. where e. f z . we will get f z +2+ f z  = 2f z +1 for all z: Rewriting this as f z +2 . 23k  and 2  3k . : : :. 1 s1 3 implies s1 = 2 and s2 = ss1 = 3: With the help of s3n = sssn = 3sn. 3k . then f z + k  = k X = kc + f z  62 . 1 for large k. we need to de ne it at odd integer q: Now de ne q+2 if q 1. q are nonnegative integers and q odd. 3.1 integers in each of the open intervals 3k . so s1997 = 336 +539 = 3804: 46. where f 1x = f x and f k+1 x = f f k x? Solution. we see that f z + n . f z +1 = f z +1 . Due to Chan Kin Hang Note that if sm = sn. Is there a function f : N ! N such that f 1996n = 2n for all n 2 N. 3k+1: Since f is strictly increasing. Next we have n sn for all n otherwise sn  n for some n. Adding these and cancelling terms. we get s3k  = 2  3k and s2  3k  = ss3k  = 3k+1: Now there are 3k . we must have s3k + j  = 2  3k + j for j = 1. So c = 0 and f z + 1 = f z  for all z: 45. For such a function f 2n = f n = f 1996f n = 2f n: So if n = 2e q. which yields the contradiction that 3n = ssn  sn  n: Then sn ssn = 3n: In particular. f z +n . 2.in this new equation. f z + n . Let N denote the positive integers. Let N be the positive integers. Suppose s : N ! N is an increasing function such that ssn = 3n for all n 2 N: Find all possible values of s1997: n=1 f z + n . : : :. 3989 mod 3992 f q = 2 q .

1 . then f 1996n = 2e f 1996q = 2e 2q = 2n: 47. 2 and suppose f f x = gx: Put hx = ggx = x .1 or 2.jq = q + 21996 . x. f 2g = f.1 + 5=2. since f induces a bijection S ! S andp ga = a .1 and 2: The set of pxed points of h contains the xed points of g and is S = f.  f a = f f y + 1 = y + f 1 = .e.1: 60 Solution 1.1 . Find all functions f : R ! R such that f f x = x2 . 2. : : :. Let R be the real numbers. 1.and f 2n = 2f n for positive integer n: If q = 3992m + 2j .e. 2 or show no such function can exist. i.1 3992m + 1 = 23992m + 1 + 2j . 4x + 2: The xed points of g i. f 1 . 1 = 2q: If q = 3992m + 3991.  f xf y + x = xy + f x: and letting a = f y + 1. j  = 3992m + 3991.jq = 23992m + 1 and f 1996q = 2f j. j = 1. we get .1. 2. Let gx = x . . y 2 S and f x = f y imply x = hx = hy = y: So f is a bijection S ! S: If c = . Find all functions f : R ! R such that for all real numbers x and y. 1995. Due to Leung Wai Ying Putting x = 1. y = .1  5=2g: Now observe that x 2 S implies hf x = f hx = f x. Problem E984 Let R denote the real numbers. 5=2: It follows that f b = a and we 2 4 2 2 have a contradiction a = f b = f f a = ga: 48. the solutions of the equation gx = x are . then f 1996. f 1997. 2 = 6 a implies f a = 6 a.1. Solution. American Mathematical Monthly. 2g: For a = . we must have f a = b = . .1. f x 2 S: Also. then f q = 23992m + 1 and f 1996 q = 2f 19953992m + 1 = 23992m + 1 + 2  1995 = 2q: So f 1996q = 2q for odd q: If n = 2eq. then gf c = f f f c = f gc  = f c and p consequently ff .

 f x .1 = f b and so .Putting y = a and letting b = f 0.  f xf y + x = 0 = f 0: By injectivity. Taking x = 1.  = f f d + 1 = d + f 1: . Setting x = 1.x. let y = a . f y = f f y + xf y + f x .  f . f x = . there is b such that f b = .x: We can easily check both are solutions. we get f f y + 1 = y + f 1: . then f f y + 1 = a and f is surjective.  For every real number a. we nd f x = x or f x = .1: Also.  Solution 2. y = 0. then . ax + b = xy . we have a2xy . 1999 IMO Determine all functions f : R ! R such that . In particular.f x=x. f 1. so f x = .f x=x = b for every x 6= 0: That is.  So c = d and f is injective.  c + f 1 = f f c + 1 .bx: Putting this into the given equation. we get f 0 = 0: For x 6= 0. 1 61 . which are easily checked to be solutions. so f x = x or f x = . let y = . abx . then . we get a = 1 and b = 0. 49.ax + b: Putting this into the equation. . ax + b: Equating coe cients. if f c = f d. we get xf y + x = 0: Then .f x=x = f y = . we get f f 0 + 1 = f 1: Since f is injective. we get b = f xf a + x = ax + f x.

we get f f x = 1 + f 0f x: Replacing x by x + y. 1995 Byelorussian Math Olympiad Let R be the real numbers. f x = 1. y2 c + 1 . A = fy1 . xy: Putting y = 1 in *. we get 1 + f 0f x + y = f f x + y = f x + y + f xf y . f x : x 2 Rg = fcx + f c . 1 2 2 1 . which simpli es to * f 0f x + y = f xf y . y in R: Solution. Find all functions f : R ! R such that f f x + y = f x + y + f xf y . from the equation. 2 However. 1 : x 2 Rg = R because c 6= 0: This means A . f x = c+1 . y2 : Then f x = f y1 . y2 : y1 . y 2 R: constant. 2 : = c. x: Putting y = . 1: So c 6= 0: For x = f y 2 A. for x 2 A. Putting y = 0. if we set y = 0. let y1 .1 2 2 x  y 1 . xy. c .for all x. y2 2 A be such that x = y1 .1 and replacing 62 Solution. y1 = c+ + y y + 2 2 12 2 2 . y2  = c. x22 2 for all x 2 R: 50. Let A be the range of f and c = f 0: Setting x = y = 0. y2 2 Ag = R: Now for an arbitrary x 2 R. we get f . Due to Yung Fai Clearly. xy for all x. y2  = f y2  + y1 y2 + f y1  . x22 : So c = 1: Therefore. we get +1 2 2 ff x . we get f 0f x + 1 = f xf 1 . f x = c .c = f c + c . x2 : Next. f x is not .

then the functional equation is clearly satis ed. we can factor n = 2em. 2x: 63 . f 1f . m are nonnegative integers and m odd. we get f 0f x = f x +1f .1. Finally. f 1f . this gives f 1+2n = f . then similarly. we nd a = 1 and b = 0. f 2. y = 2d n. We have * f 1 + f n = f 1 + 2n + f . f xf y = xy: Then f xf 1 = x for all x 2 R: So f 2 0 .e. then f x + f y = f 2e  + f 2d  = f x1 + 2y + f y1 .n and f n + f . we get f n = f x = f x +2xy = f n. If we let x = .1 = 0.1 = .n+f . The functional equation gives f n = f y = f y . y = n. If x = 2e m.1x + f 0: If f 2 0 . f 1. we check such functions satisfy the equations. Therefore.x by x +1 in *. where m. n are odd. f 2 0 . 2xy for all integers x. 2xy = f n4k + 3: If we let x = n. then x and x + 2xy are odd.1f x = f 0 . Finally. we get f 2 0 . : : : which may be arbitrary. 1993 Czechoslovak Math Olympiad Let Z be the integers. f 8. f 4.1+ x +1: Eliminating f x + 1 in the last two equations. f . we get f 0 = 0. f .4k . f 1f .1 = f . then putting x = 0 in the last equation. f x = x for all x 2 R: 51.n for every integer n: So we need to nd f n for nonnegative integers n only.1 = f 1: If x or y = 0. 1 = f n4k +1: So f n = f nm for every odd m: For a positive integer n. By *. Then f n = f 2e : So any such function f is dtermined by the values f 0. substituting f x = ax + b into the original equation. i. Clearly.2k + 1.1+2n: Since f . y: Solution.1 6= 0 and f x is a degree 1 polynomial. All other values are given by f n = f 2e  as above. y = . f 16.1 + 2n for every integer n: So f k has the same value for every odd k: Then equation * implies f n = f . where e.2k + 1. resulting in a contradiction.1 = f 1.1 = f 1 and f x + f y = f x + 2xy + f y . f 1f . Find all functions f : Z ! Z such that f .

which implies f 2 = kf 12 : Using 2f n2k+1= f n2 repeatedly or by induction. we kget f 2 = 1 : Assume f 2  k= 1: Then for +1 k +1 n = 22 k. b For any m. which implies f 0 = 0 or 1. 1995 South Korean Math Olympiad Let A be the set of non-negative integers. we get f 1 = 1: 2 For m k= 1 = n. we get 2f n2  = f 02 + f n2 : Let m n. they are the only solutions. so f n = 1 for all n: k!1 Case f 0 = 0: Then 2f n2  = f n2 : So f n is even for all n: For m = 1 = n. f n = 0 and f n = 2n are solutions. For m = 0. n 2 A. Therefore. then f 22k  = 22k+1 : Now f m + 12 = 2f m + 12  = 2f m2 + 2m + 1 2f m2 + 1 = f m2 + f 12 f m2 : As f n is always even. Setting m = 0 = n. Problem E2176 Let Q denote the rational numbers. If f 1 = 0. f m2   f n2 : Solution. we get 2f 2 = f 12 + f 12 . we k k k k 2 get f n  2n: Since f 22  = 22 +1 = 2  22 for all k. then f m2 . we get 22 . we get f m + 1  f m + 2: By induction. 53. Case f 0 = 1: Then 2f n2  = 1 + f n2 : For n = 1k. Find all functions f : A ! A satisfying the following two conditions: a For any m. f n2   0: So f m  f n: This means f is nondecreasing. we get 2f 0 = f 02 + f 02 . 2f m2 + n2  = f m2 + f n2 . Find all functions f : Q ! Q such that . f n2 = 2f m2  .52. klim 2 =1 !1 and f is nondecreasing. we get 2f 22  = 1 + f 22 2 = 2: So f 22  = 1: Since 2 lim 2 = 1 and f is nondecreasing. k then klim 22 = 1 and f nondecreasing imply f n = 0 for all n: If !1 f 1 = 2. American Mathematical Monthly. so f n = 2n for all n: It is easy to check that f n = 1. n 2 A with m  n.1 f 22k  = f 12 : Now 2f 1 = f 12 implies f 1 = 0 or 2.

x + y f x + f y f 2 = 2 and f x . y = f x . f y for x 6= y: 64 .

x = . So. which yields f 1 . x 6= 0. f 0. 1 Setting y = 0.Solution. we get 0 = f x + f . We will show f x = x is the only solution by a series of observations. so f . c 6= 1. f 1 = 1 and then f 0 = 0: 2 Setting y = . 1f x = f 01 + f 1: Now f is not constant because the denominator in the equation cannot equal 0. we get f 1 = f x + f 0=f x .f x: 3 Setting y = cx. we get f x + f cx = f .x.x.

f c which implies f cx = f cf x: Taking c = q. otherwise f x + 2=x . then f x . 1 . The condition f 2 = 2 can also be deduced from the functional equation as shown below in 5. 2 = n . 65 . f 3: Substituting the equations for f 3 and f 5 in terms of f 2 and simplifying. 2. 1 = n . y 0: This implies f is strictly increasing for positive real numbers. 2=f x . So. Therefore. x = p=q. 2: If f n . f 2 = 2: 6 Note f x 6= 0 for x 0. 1. 1 + c = 1 + f c . f cx 1. f 5 = f 3 . 1 = f x + f x . f x . we get f p=q = f p=f q: 4 Setting y = x . 0 = 1 will force f to be constant. f 22 = f 4 = f 5+ f 3=f 5 . 5 We have 2 f 3 + f 2 f 2 +1 : f 2 + 1 f 3 = f 2 .c 1 . then this equation implies f n = n: Since f 1 = 1 and f 2 = 2. if x 0. then f n = n for all positive integers by induction and 2. then f x = f  x2 0: If x y  0. otherwise f cx =p 0 for c 6= 1 will force f to be constant. f x . then again the only solution is still f x = x as shown below in 6 and 7. f 22 Also. 2 6= 0 and f n . If rational numbers are replaced by real numbers. we get f 22 = 2f 2: Now f 2 6= 0. f y = f x + f y=f x + y=x . 3 will imply f x = x for all x 2 Q: Comments. 2 = f x + 0=f x . f 2 = 1 + 2f 2 . we get f x .

7 For x 0; if x f x; then picking r 2 Q such that x r f x will give the contradiction that f x f r = r f x: Similarly, f x x will also lead to a contradiction. So, f x = x for all x: 54. Mathematics Magazine, Problem 1552 Find all functions f : R ! R such that

f x + yf x = f x + xf y for all x; y in R:

Solution. It is easy to check that f x = 0 and f x = x are solutions.

Suppose f is a solution that is not the zero function. We will show f x = x for all x: Step 1. Setting y = 0; x = 1; we get f 0 = 0: If f x = 0; then 0 = xf y for all y; which implies x = 0 as f is not the zero function. So f x = 0 if and only if x = 0: Step 2. Setting x = 1; we get the equation * f 1+yf 1 = f 1+f y  for all y: If f 1 6= 1; then setting y = 1=1 , f 1 in *, we get f y = f 1 + f y; resulting in f 1 = 0; contradicting step 1. So f 1 = 1 and * becomes f 1 + y = f 1 + f y; which implies f n = n for every integer n: Step 3. For integer n; real z; setting x = n; y = z , 1 in the functional equation, we get

f nz  = f n + z , 1f n = n + nf z , 1 = nf z :
Step 4. If a = ,b; then f a = f ,b = ,f b implies f a + f b = 0 = f a + b: If a 6= ,b; then a + b 6= 0 and f a + b 6= 0 by step 1. b ; we get Setting x = a + b=2; y = a , b=2f  a+ 2

, b + a , b f  a + b  = f  a + b  + a + b f , a , b ; f a = f a + b b 2 2 2 2 2f  a+ 2f  a+ 2 2 , b + b , a f  a + b  = f  a + b  + a + b f , b , a : f b = f a + b b 2 2 2 2 2f  a+ 2f  a+ 2 2
66

b  = f a + b by step 3. Adding these, we get f a + f b = 2f  a+ 2 Step 5. Applying step 4 to the functional equation, we get the equation ** f yf x = xf y: Setting y = 1; we get f f x = x: Then f is bijective. Setting z = f x in **, we get f yz  = f yf z  for all y; z: Step 6. Setting z = y in the last equation, we get f y2  = f y2  0: Setting z = ,y; we get f ,y2  = ,f y2  = ,f y2  0: So f a 0 if and only if a 0: Step 7. Setting y = ,1 in the functional equation, we get f x , f x = f x , x: Since x , f x and f x , x are of opposite signs, by step 6, we must have x , f x = 0 for all x; i.e. f x = x for all x:

Maximum Minimum

55. 1985 Austrian Math Olympiad For positive integers n; de ne

f n = 1n + 2n,1 + 3n,2 +    + n , 23 + n , 12 + n:
What is the minimum of f n + 1=f n?

Solution. For n = 1; 2; 3; 4; 5; 6; f n + 1=f n = 3; 8=3; 22=8; 65=22;
209=65; 732=209; respectively. The minimum of these is 8=3: For n 6; we will show f n + 1=f n 3 8=3: This follows from

f n + 1 1n+1 + 2n + 3n,1 + 4n,2 + 5n,3 + 6n,4 +    + n , 13 + n2 1n+1 + 2n + 3n,1 + 4n,2 + 5n,3 + 36n,5 +    + n , 12 + n =1n+1 +    + 5n,3 + 3f n , 1n , 2n,1 + 3n,2 + 4n,3 + 5n,4  =3f n + 25n,4 , 1 + 2n,12n,5 , 1 3f n:
Therefore, 8=3 is the answer. 56. 1996 Putnam Exam Given that fx1 ; x2; : : :; xn g = f1; 2; : : :; ng; nd the largest possible value of x1 x2 + x2 x3 +    + xn,1xn + xnx1 in terms of n with n  2. 67

Solution. Let Mn be the largest such cyclic sum for x ; x ; : : :; xn:
1 2

In case n = 2; we have M2 = 4: Next suppose Mn is attained by some permutation of 1; 2; : : :; n: Let x; y be the neighbors of n: Then removing n from the permutation, we get a permutation of 1; 2; : : :; n , 1: The di erence of the cyclic sums before and after n is removed is nx + ny , xy = n2 , n , xn , y  n2 , 2: Equality holds if and only if x; y are n , 1; n , 2: So Mn , n2 , 2  Mn,1 : Then
3 2 , 11n + 18 : Mn  M2 +32 , 2+42 , 2+    +n2 , 2 = 2n + 3n 6

Following the equality case above, we should consider the permutation constructed as follows: starting with 1; 2, we put 3 between 1 and 2 to get 1; 3; 2; then put 4 between 3 and 2 to get 1; 3; 4; 2, then put 5 between 3 and 4 to get 1; 3; 5; 4; 2 and so on. If n is odd, the permutation is 1; 3; : : :; n , 2; n; n , 1; : : :; 4; 2: If n is even, the permutation is 1; 3; : : :; n , 1; n; n , 2; : : :; 4; 2: The cyclic sum for each of these two permutations is 2n3 + 3n2 , 11n + 18=6 because of the equality case at each stage. Therefore, Mn = 2n3 + 3n2 , 11n + 18=6:

68

ON ? A B : Then O. 6 KPM = 6 OPM. 1 1 1 2 2 2 1 1 1 1 1 1 2 2 2 2 1 1 2 2 1 2 1 2 69 . The points M and N are the feet of the perpendiculars from P and Q to AB: Find angle MCN: be 45 : To prove this. we have PC = PM: Let L be the foot of the perpendicular from C to AB: Then PM k CL: So 6 PCM = 6 PMC = 6 MCL: Similarly. we get OM  OC = OA = OB = ON  OC : By the converse of the intersecting chord theorem. Ng Ka Wing. observe that since P is on the bisector of 6 BAC. which is not a diameter. Chords A1 B1 and A2B2 intersect at the midpoint P of AB: Let the tangents to the circle at A1 and B1 intersect at C1: Similarly. OC intersects A B at N: and OC intersect AB at K: Since OC is a perpendicular bisector of A B . Due to Ng Ka Man. which implies AB = CB: 59. 1988 Leningrad Math Olympiad Squares ABDE and BCFG are drawn outside of triangle ABC: Prove that triangle ABC is isosceles if DG is parallel to AC: Solution. 16 6 QCN = 6 NCL: So 6 MCN = 2 PCQ = 45 : 58. the right triangles ABX and BDY are congruent and AX = BY: Similarly. Due to Poon Wai Hoi Using protractor.6 DBY = 6 BDY and AB = BD. Due to Poon Wai Hoi Let OC intersects A B at M. the angle should Solution. N. the right triangles CBX and BGY are congruent and BY = CX: So AX = CX. we have 6 OKP = 6 ONM: From the right triangles OA C and OB C . So 6 ONM = 6 OPM: Since 6 OKP = 90 . let the tangents to the circle at A2 and B2 intersect at C2: Prove that C1 C2 is parallel to AB: Solution. AB is a chord of a circle. M are concyclic. P. 1995 British Math Olympiad Triangle ABC has a right angle at C: The internal bisectors of angles BAC and ABC meet BC and CA at P and Q respectively. draw a perpendicular to AC and hence also perpendicular to DG: Let it intersect AC at X and DG at Y: Since 6 ABX = 90 .Solutions to Geometry Problems 57. so OM ? A B : Similarly. Yung Fai From B .

1991 Hunan Province Math Competition Two circles with centers O1 and O2 intersect at points A and B: A line through A intersects the circles with centers O1 and O2 at points Y. So 6 OC1C2 = 6 ONM = 6 OKP: Then C1C2 k KP. O2 are concyclic. respectively. 6 O1 AY + 6 O2AZ  = 6 O1AO2 = 6 O1 BO2 : So B. Y. C2 are concyclic. Then we will show 90 = 6 QBP = 6 XBP: Since 6 XY P. P. Then 6 QBP = 6 XBP = 180 . Z. Since 6 XY P = 6 XZP = 90 . First we need to show P. O .we get M. Let the tangents at Y and Z intersect at X and lines Y O1 and ZO2 intersect at P: Let the circumcircle of 4O1O2B have center at O and intersect line XB at B and Q: Prove that PQ is a diameter of the circumcircle of 4O1O2 B: Solution. 6 O1PZ = 180 .  1 2 2 = 180 . P. there are four points such that the distance between every pair of them is greater than the radius of the disk. 6 O1 Y Z + 6 O2ZA = 180 . 6 AY B + 6 AZB  6 AO1 B + 1 6 AO2 B  = 180 . O . N. 1 2 it su ces to show X. 6 XZP = 90 : Therefore. that is C1 C2 is parallel to AB: 60. C1 . O1. Connect O1 A and O2 A: In 4Y PZ. 6 PZX are both 90 . P. Connect BY and BZ: Then 6 Y BZ = 180 . B are concyclic. X. 1981 Beijing City Math Competition In a disk with center O. B are concyclic. P. so the points Y. 6 BO1 O2 + 6 BO2 O1  = 6 O1BO2 = 6 O1PZ = 6 Y PZ: So Y. B. Z. Z. Prove that there is a pair of perpendicular diameters such that exactly one of the four points lies inside each of the four quarter disks formed by the diameters. PQ is a diameter of the circumcircle of 4O1O2B: 61. 70 . B are concyclic. Z are concyclic.

four points. sides are a. Now EF and its perpendicular through O divide the disk into four quarter disks. Prove that the triangle is a right triangle. 6 BOC. The length of each side divides the sum of the other three lengths. D in the order a rotating radius encountered them. OB. 3  60 = 180 : Let EF be the diameter bisecting 6 AOB and with A. 2  60 = 240 and 6 EOB + 6 BOC 240 . 1988 Sichuan Province Math Competition Suppose the lengths of the three sides of 4ABC are integers and the inradius of the triangle is 1. b = CA. y = 3d=b and z = 3d=c: Then a b c d implies x y z 3: So z  4. 6 AOE  240 . Therefore. C each is on a di erent quarter disk. c = AB be the side lengths. Similarly. D will lie on the remaining quarter disk. d with a b c d: Since d a + b + c 3d and d divides a + b + c. we get 6 AOB + 6 BOC 360 . 6 BAO: Hence 6 AOB 60 : Similarly. b. r be the inradius and s = a + b + c=2: Since the area 71 . Due to Chao Khek Lun and Leung Wai Ying Suppose the Solution. c dvides a + b + c + d = 3d: Let x = 3d=a. 63. 90 6 EOD = 6 EOA + 6 AOD 180 : Therefore. so 6 AOB 6 OBA. Let us name the points A. c. we have a + b + c = 2d: Now each of a. a contradiction. y  5. B. The lengths of the sides of a quadrilateral are positive integers. B. Due to Lee Tak Wing By the distance condition on the Solution. We have 90 = 30 + 60 6 EOB + 6 BOC: 6 AOB In the case 60 120 . C. Since AB OA. Due to Chan Kin Hang Let a = BC. B. E on the same half disk. 6 COD. b. Solution. 62. Prove that two of the sides have the same length. none of them equals O and no pair of them are on the same radius. 6 DOA 60 : Let 6 AOB be the largest among them. two of the sides are equal. x  6: Then 2d = a + b + c  36d + 35d + 34d 2d. 120 =2 = 180 : So A. then 60 6 AOB 360 . we get 6 EOB + 6 BOC 60 + 120 = 180 : In the case 120  6 AOB 180 .

c: In mod 2. a.of the triangle is rs. z = s . 1985 IMO A circle has center on the side AB of the cyclic quadrilateral ABCD: The other three sides are tangent to the circle. M. Geometric Equations p 64. c are the same. 2b2s . as . 2a2s . Prove that B is the incenter of 4EAF and MN = AE + AF: 72 Solution. ABC = CDA = 6 CDO: . a + s . y = 2. a. b = 4. bs . we get ss . a = 3. D. a + b . Then a + b + c is even and s is an integer. bs . s = 6. each of b + c . the triangle is a right triangle. DAM = AMD + ADM : 2 2 2 So 6 AMD = 6 ADM: Therefore. AM = AD and AB = AM + MB = AD + BC: 65. y = s . c satisfy xyz = x + y + z: Suppose x  y  z: Then yz  3 for otherwise xyz 3x  x + y + z: This implies x = 3. b + s . z = 1. they are all even. O2 respectively intersect each other at points A and B: Ray O1 B intersects S2 at point F and ray O2B intersects S1 at point E: The line parallel to EF and passing through B intersects S1 and S2 at points M and N. c = 5: Therefore. 1995 Russian Math Olympiad Circles S1 and S2 with centers O1. Then 6 6 6 6 6 AMD = 6 OCD = DCB = 180 . respectively. Prove that AD + BC = AB: 2 2 This implies C. So either they are all odd or all even. ba + b . b. b. c = s = s . Since their product is even. The positive integers x = s . Let M be on AB such that MB = BC: Then 6 6 6 CMB = 180 . c + a . O are concyclic. c: Now 4a + b + c = 8s = 2s . as . 2c = b + c . c = 1  s = s: Then s . ac + a .

If triangles ABC and A0B 0 C 0 are equilateral. which implies B is the incenter 2 of 4EAF because the incenter is the unique point P on the bisector 6 EAF . Due to Wong Chun Wai Note that 6 EAD. by cosine law. 6 FCD. Then 6 AEB = 6 BEF = of 6 EAF such that EPF = 90 .Solution. 16 EAF: Then 6 EBF = 90 + 6 EAF. E are concyclic and B. So A. 6 BDF = 6 BCF = 6 BAC = 6 DEC.6 O1BA = 180 +6 O1BE = 270 . D. C. 2 16 AB bisects 6 EAF and 6 O1 BE = 90 . F are concyclic. we get PA0 + PB 0 = 2PE + EB 0 : From 4PC 0 E with cevian PO note C 0O = 2 2 2 2 73 . Then 6 ADE = 6 ACE = 6 ABC = 6 DFC. say the measure of these angles is : Also. 6 EAB = 90 . MN = MB + NB = AE + AF: 66. say CE  CF = DE cos DF cos  = DE cos DF cos  = AD  BD. 6 FBO2 = 6 BAF. 1 2 6 EBM since EF k MN: So EBAM is an isosceles trapezoid. Hence EA = MB: Similarly FA = NB: Therefore. CD2 = DE sin DF sin  = DE sin DF sin  = AE  BF: 67. Since . Let O be the center of both circles and E be the midpoint of A0 B 0 : From 4PA0 B 0 with median PE. Quadrilaterals ABCP and A0 B 0 C 0 P 0 are inscribed in two concentric circles. Point C lies on the minor arc AB of the circle centered at O: Suppose the tangent line at C cuts the perpendiculars to chord AB through A at E and through B at F: Let D be the intersection of chord AB and radius OC: Prove that CE  CF = AD  BD and CD2 = AE  BF: 6 EAB = 1 6 EO1 B = 90 Solution. 6 O1BE = 90 . D. 6 ECD. 2 EAF: Now 6 EBA+6 FBA = 6 EBA+180 . the measure of these angle is : Then 6 FBD are right angles. prove that P 0 A2 + P 0 B 2 + P 0 C 2 = PA02 + PB 02 + PC 02 : Solution. C.

Without loss of generality. A0 be at rei. we get PA02 + PB 02 + PC 02 = 2EB 02 + OE 2  + 3PO2 + 4OE 2 = 2B 0 O2 + 3PO2 + C 0O2 = 3PO2 + P 0 O2: Similarly. by cosine law again. we get PC 02 + 2PE 2 = 3PO2 + 2OE 2: Putting these together. P 0 A2 + P 0 B 2 + PC 02 = 3PO2 + P 0 O2: Alternatively. let the center be at the origin. the problem can be solved using complex numbers.2OE .

= rcos .

+ i sin .

rei. and P be at Rei : Let ! = e2i=3: We have PA02 + PB 02 + PC 02 =jRei .

j2 + jRei . rei.

!j2 + jRei . rei.

2Re Rrei . !2 j2 .  =3R2 .

6 DEG = 6 DIF = 6 DIB: So 4BDI. Due to Wong Chun Wai Let I be the incenter of 4ABC: Then 6 BDI = 90 = 6 EGD: Also. 4CDI. 4EGD are similar. P 0 A2 + P 0 B 2 + P 0 C 2 = 3R2 + 3r2: 68. Then BD=ID = DG=EG: Likewise. side CA at E and side AB at F: Let G be the foot of perpendicular from D to FG = BF : EF: Show that EG CE 1 2 Solution. 1998 IMO shortlisted problem Let ABCDEF be a convex hexagon such that 6 B + 6 D + 6 F = 360 and AB  CD  EF = 1: BC DE FA 74 . 4FGD are similar and CD=ID = DG=FG: Therefore. Let the inscribed circle of triangle ABC touchs side BC at D. 1 + ! + !2  + 3r2 =3R2 + 3r2: Similarly. FG = DG=EG = BD=ID = BD = BF : EG DG=FG CD=ID CD CE 69.

C. 4PDC are similar. B. f  = b . 6 D and 6 F are the arguments of the complex numbers a . respectively. Consequently.Prove that BC  AE  FD = 1: CA EF DB Solution. Let P be such that 6 FEA = 6 DEP and 6 EFA = 6 EDP. a . AB = DE  FA = DP : BC CD  EF CD Then 4ABC. since 6 BCD = 6 ACP. c . 4AEP are similar. de . cf . d  e . we can get an algebraic solution. E. eb . 6 B. Using complex numbers. CD=DE and EF=FA: Since AB=BC CD=DE EF=F A = 1. c. d: 75 . 4ACP are similar. f . b  c . d . So FA = DP and  EF = EA : EF PE ED EP Since 6 B + 6 D + 6 F = 360 . b=c . respectively. Also. c . 4FED. Since ABCDEF is convex. bc . we have a . d=e . f  . ef . It is equal to the product of their absolute values AB=BC. The origin may be taken anywhere on the plane. by **. we get the desired equation. be . Then the condition 6 B + 6 D + 6 F = 360 implies that the product of these three complex numbers is a positive real number. b. So AE=EF = PA=FD and BC=CA = DB=PA: Multiplying these and moving all factors to the left side. where P is on the opposite side of lines DE and CD as A: Then 4FEA. 4BCD. Let a.f So 0 = a . f = 1: c. f =a . F. D.b e. by *. d and e . we get 6 ABC = 6 PDC: Also. b.d a. d. we get 6 BCA = 6 DCP and ** CB=CD = CA=CP: Since 6 FED = 6 AEP. 4DEP are similar. f denote the complex numbers corresponding to A. ca . da . e.

N be the projections of O on BD. FB. d = 1: CA EF DB a . Hexagon ABCDEF is inscribed in a circle so that AB = CD = EF: Let P. EA and FB respectively. N are midpoints of BD. BRP and CPQ be Solution. Due to Ng Ka Wing Let O be the center of the circle and let L. respectively. 6 A0 B 0 C 0 = 6 ABC and 6 B 0 C 0 A0 = 6 BCA: So. c  a . 6 CAB = 6 ABC + 6 BCA and 6 RXP = 180 . we get FB = AE and hence 76 . respectively. which implies X is on the circumcircle of triangle CPQ: Now 2 2 1 = 2 6 QA0 R = 6 CAB: Similarly. triangles A0B 0 C 0 and ABC are similar. d Similar Triangles 70. 71. Let S be the projection of O on AE: Since AB = EF. and AB respectively of triangle ABC: Prove that the three circumcentres of triangles AQR. A good drawing suggests the circles pass through a common point! To prove this. R be the points of intersection of AC and BD.Then BC  AE  FD = b . CE and DF. DF. respectively. BRP. e  f . e b . Let the circumcenters of triangles AQR. 6 BCA. 6 ABC = 6 CAB + 6 BCA: So 6 PXQ = 360 . CA. FB. Q. 6 C 0A0 B 0 = 6 C 0 A0X + 6 XA0B 0 = 1 6 QA0 X + 1 6 RA0 X Solution. Q. Then L. let circumcircles of triangles AQR and BRP intersect at R and X: Then 6 QXR = 180 . 6 QXR . 1984 British Math Olympiad P. and R are arbitrary points on the sides BC. 6 RXP = 180 . M. DF. c f . Prove that triangles PQR and BDF are similar. M. and CPQ form a triangle similar to triangle ABC: A0. B 0 and C 0 .

dP. the image R0 of R lies on the line ON and the image P 0 of P lies on line OL: Then ON=OR0 = OL=OP 0 = OM=OQ0 = cos : So 4P 0Q0 R0 . we have Solution. Let E and F be variable points on the sides AB and CD. such that AE : EB = CF : FD: Let P be the point on the segment EF such that PE : PF = AB : CD: Prove that the ratio between the areas of triangles APD and BPC does not depend on the choice of E and F: denote the distance from X to line Y Z: We have AE : EB = CF : FD = a : b. DF. M. Therefore. Since L. respectively. BC . AD = ABCD d  E. 6 POL = 6 QOM = and OL=OP = OM=OQ = cos : Next rotate 4PQR around O at angle so that the image Q0 of Q lies on the line OM. AD. we have 4LMN. respectively. FB. dP. 4PQR. N are midpoints of BD. 4BDF are similar. where a + b = 1: Since PE : PF = AB : CD. 4LMN are similar.ON = OS: Let 6 AOB = 6 COD = 6 EOF = 2 : Then 6 RON = 1 6 16 6 ARB = 4 SON = 1  AOB + 6 EOF  = : Hence. BC  + + CD AB + CD AB BFC BEC + BPC = ABCD + CD AB + CD b  CD BAC + a  AB BDC : = AB + CD AB + CD 77 . Let UV W denote the area of 4UV W and let dX. 72. 1998 IMO shortlisted problem Let ABCD be a cyclic quadrilateral. ON=OR = 2 2 cos : Similarly. Y Z  AB dF. 4BDF are similar. AD  + + CD AB + CD AB AFD AED + APD = ABCD + CD AB + CD a  CD ABD + b  AB ACD . = AB + CD AB + CD AB dF. BC  = ABCD d  E.

B. respectively. 6 CAB = 6 CBA = 6 LBA: So A. so A. 78 . 6 KCL . B. 1999 IMO Two circles . C. Then 6 KBA = 6 KLA = 6 CMA = 6 CDA = 6 BDA: Therefore.Since A. 6 ACB . D are concyclic. D: 74. APD = a  CD  ABD + b  AB  ACD BPC b  CD  BAC + a  AB  BDC  CD  AB  AD  sin 6 BAD + b  AB  CD  AD  sin 6 ADC =a b  CD  AB  BC  sin 6 ABC + a  AB  CD  BC  sin 6 BCD AD  a  sin 6 BAD + b  sin 6 ADC = AD : = BC b  sin 6 ABC + a  sin 6 BCD BC Tangent Lines 73. Since LK k MC. Two circles intersect at points A and B: An arbitrary line through B intersects the rst circle again at C and the second circle again at D: The tangents to the rst circle at C and to the second circle at D intersect at M: The parallel to CM which passes through the point of intersection of AM and CD intersects AC at K: Prove that BK is tangent to the second circle. C. at the distinct points M and N. sin 6 BAD = sin 6 BCD and sin 6 ABC= sin 6 ADC.1 and . Solution. L. D are concyclic. 6 LKC = 180 . BK is tangent to the circle passing through A. 6 LCM = 180 . So.2 are contained inside the circle . M. K are concyclic. and are tangent to . Let L be the intersection of AM and CD: Since 6 CMD + 6 CAD = 6 CMD + 6 CAB + 6 DAB = 6 CMD + 6 BCM + 6 BDM = 180 .. B. 6 KCM = 180 .

6 CBD subtend equal arcs. 4PBD are similar. respectively. 6 ECY = 90 . G1 and G2 are on the same side of chord BC. E are concyclic. B.2. D are collinear because the tangent at P namely BC  and the tangent at D are parallel. The lines MA and MB meets . .1 : Similarly. at A and B. 1 6 CXY 2 = 90 . 1 2 Solution. M.1 and . DB 2 = DP  DQ = DL2: So DL = DB = DC: Then D has the same power DB 2 = DC 2 with respect to G1 and G2 : Hence D is on the radical axis AW of G1 and G2 : So L = W and DW is tangent to G1 and G2: 79 Solution.1 at M: Then 6 UCM = 6 ANM = 6 UMC: So CE is tangent to . C are points on G such that A.2 : respectively. 6 CY Q = 6 CY Q: These imply 4CY E. CE is tangent to . Due to Wong .2 : The line passing through the two points of intersection of .2 at E: Since AB is the radical axis of . Proposed by India for 1992 IMO Circles G1 and G2 touch each other externally at a point W and are inscribed in a circle G: A. Let D be the midpoint of the complementary arc BC of G not containing A and L be a point on G1 so that DL is tangent to G1 and intersects segment PC: Considering the homothety with center Q that maps G1 onto G.2 meets . Prove that CD is tangent to .1 passes through the center of . . Extend Y X to meet . 90 .1 at C and D. N.2 : Now Y E = Y Q and 6 CY E = 90 . we see that Q. Y be the centers of . P. Let U be the intersection of line CE with the tangent to .2 at Q: Join AN to meet . ! Chun Wai Let X. so AC  AM = AE  AN: This implies C. CD is tangent to . Let P and Q be the points of tangency of G with BC and 1 .1 . . .2 : 75. Hence 6 CQY = 6 CEY = 90 : Similarly 6 DQY = 90 : Therefore. Since 6 BQD. respectively.. 4CY Q are congruent. . which is also tangent to G1 and G2 : Suppose AW is also tangent to G1 and G2: Prove that W is the incenter of triangle ABC: arc BAC. 4BQD. Hence DB=DP = DQ=DB: By the intersecting chord theorem. respectively.

let the chord through A perpendicular to OD0 intersect the circle at B and C . B. Perpendiculars from a point P on the circumcircle of 4ABC are drawn to lines AB. Solution. 6 BAD = 6 WBD . Consider all chords excluding the diameter passing through A: What is the locus of the intersection of the tangent lines at the endpoints of these chords? Solution. where D is the midpoint of BC: Since 6 OBD0 = 90 . D. OD  OD0 = OB 2= OA  OA0: So 4OAD is similar to 4OD0A0: Since 6 ODA = 90 . respectively.Since D is the midpoint of arc BC. By symmetry. Find the locus of the circumcenter of 4PDE as P moves around the circle. 6 ABW = 6 BWD . then OM ? PB: It follows that the locus of M is the circle with OB as a diameter. P. Hence. D0 is on the line OD. The rst part of the proof can also be done by inversion with respect to the circle centered at D and of radius DB = DC: It maps arc BC onto the chord BC: Both G1 and G2 are invariant because the power of D with respect to them is DB 2 = DC 2 : Hence W is xed and so DW is tangent to both G1 and G2 : Locus 76. Then PB is a diameter and the circumcenter of 4PDE is at the midpoint M of PB: Let O be the circumcenter of 4ABC. so AW bisects 6 BAC: Also. Due to Wong Him Ting Let O be the center and and r be the radius. for D0 on L. D0 is on the line L perpendicular to OA at A0 : Conversely. Let D be the intersection of 80 . E. E are concyclic. 6 CBD = 6 CBW and BW bisects 6 ABC: Therefore W is the incenter of 4ABC: Comments. 77. Suppose A is a point inside a given circle and is di erent from the center. Since 6 PDB + 6 PEB = 180 . Let A0 be the point on OA extended beyond A such that OA  OA0 = r : Suppose BC is one such chord passing through 2 A and the tangents at B and C intersect at D0 . BC with feet at D. the circumcircle of 4PDE passes through B always.

Find the locus of all points M for which there exists points X on C1 and Y on C2 such that M is the midpoint of the line segment XY: respectively. so 4BAE. the locus of P is the circle C : 79. M will 1 trace a circle . Then AB  AC = AE  AF: Therefore. respectively. Due to Poon Wai Hoi Let O . since AB=AE = AF=AC and 6 BAE = 6 FAC. Then 6 AEP = 6 PCA: So A. so 6 EPF = 6 EAC: Hence A. so that AM = CN = r: AC CE 81 . Hence 6 BPC = 6 CAE = 6 BAC=2: Therefore. O be the centers of C . 1996 Putnam Exam Let C1 and C2 be circles whose centers are 10 units apart. 1982 IMO Diagonals AC and CE of the regular hexagon ABCDEF are divided by the inner points M and N. P are concyclic. P is on the circle C whose points X satisfy 6 BXC = 6 BAC=2 and whose center is on the same side of line BC as A: Conversely. 4FAC are similar. and whose radii are 1 and 3. Let BP. F. then as X moves around C1 . So 6 BEA = 6 FCA: Also 6 BAE = 6 FAC: So 4BAE. P are concyclic. Given 4ABC: Let line EF bisects 6 BAC and AE  AF = AB  AC: Find the locus of the intersection P of lines BE and CF: Solution. respectively. which implies 6 OBD0 = 6 OCD0 = 90 : Therefore. 1 2 1 2 80. mY will trace a circle of radius 3 centered at 2 the midpont P of O1 O2: So the locus is the solid annulus centered at 3 3 P with inner radius 2 . CP intersect the angle bisector of 6 BAC at E. E.Y with radius 2 centered at the midpoint mY of O1Y: As Y moves around C2. If we x Y on C2.the chord with OD0 : Now 4OAD. Since 6 BPC = 6 BCA=2. 4OD0A0 are similar right triangles. E. C.1 = 1 and outer radius 2 +1 = 2: 2 2 Collinear or Concyclic Points Solution. For such a point P. This shows the locus is the line L: 78. D0 is on the locus. So OD  OD0 = OA  OA0 = OB 2 = OC 2. 4FAC are similar. C. C . for P on C .

1965 Putnam Exam If A. the circles through A. D are neither concyclic nor collinear. midway between the parallel lines AB and CD: Clearly. 1. r: Let XY Z denote the area of 4XY Z: Since 6 NCM = 60 . D. a contradiction.1 or on the perpendicular bisector Hr of AB: Similarly. prove that if all the distances between every pair are integers. XC j = j g for j = 0. C such that AB = r and AC = s: Observe that if P is one of the other points. 2. C. so 2 x2 1 . again a contradiction. then the points are collinear. Suppose A. : : :. CN = xr. PB j = 0. then lines AB and CD are also parallel. s: So P is on one of the sets Kj = fX : jXA . respectively.XB j = ig for i = 1. 2. p Solution. Solution.Determine r if B. 1. 1957 Putnam Exam Given an in nite number of points in a plane. r: Hence P would be on the line H joining A. r. Consider points P and Q on p and q. C. B. If lines p and q are parallel. Q have no common point. the other through C and D. r sin 60 = xp r: 2 2 3 2 3 1: Cancelling x2 and solving for r. Suppose there are three noncollinear points A. B. their common point is the center of two concentric circles. 6 BCM = 30 and BCM + CMN = BCN . one through A and B. : : :. p r sin 30 + x2 r1 . B. jPA . then BC = x= 3. 1. D are four distinct points such that every circle through A and B intersects or coincides with every circle through C and D. then bt the triangle inequality. : : :. M and N are collinear. : : :. 2. Then the perpendicular bisector p of AB cannot coincide with the perpendicular bisector q of CD: If lines p and q intersect. PC j = 0. prove that the four points are either collinear or concyclic. CM = x1 . s: Since 0 82 . jPA . P and C. we get r = p 3 81. B or on one of the hyperbolas Hi = fX : jXA. B. Due to Lee Tak Wing Let AC = x. 82. Solution.

Therefore. q2 . . Suppose the point P. C. CP = DP = s: Then the coordinates of A. CY = CD = CE and AF=EA = 1 = XZ=Y X. p2 p p q . q . r2 . D are . 2 2 2 2 2 2 Solution. C. PE  PF = PD = PY  PZ: Hence EFZY is a cyclic quadrilateral by the converse of the intersecting chord theorem. the diagonals AC and BD are perpendicular and the opposite sides AB and DC are not parallel. p. s2 . B. 1998 IMO In the convex quadrilateral ABCD. . we get XZ  BP  CY =ZB  PC  Y X  = 1: By the converse of the Menelaus' theorem. a contradiction. 1995 IMO shortlisted problem The incircle of triangle ABC touches BC. If lines EF and BC intersect at P. Due to Leung Wing Chung Set the origin at P: Suppose p p p p respectively. 4CDP have equal areas if and only if 1 . is inside ABCD: Prove that ABCD is a cyclic quadrilateral if and only if the triangles ABP and CDP have equal areas. 83. Y. q. and touches CX and XB at Y and Z respectively. r . then by Menelaus' theorem.lines AB and AC are distinct. P are collinear. 2 84. p . X is a point inside triangle ABC such that the incircle of triangle XBC touches BC at D also. then AB = AC and AD is an axis of symmetry of EFZY: Hence EFZY is a cyclic quadrilateral. E and F respectively. p . CA and AB at D. So there can only be nitely many points that are integral distances from A. B. If EF k BC. the given points must be collinear. every intersection Hi Kj is only a nite set.q 83 p s2 . Prove that EFZY is a cyclic quadrilateral. r . Now 4ABP. Z. Solution. p. where the perpendicular bisectors of AB and DC meet. A and C are on the line y = p and B and D are on the line x = q: Let AP = BP = r.  s . s2 . AF  BP  CE =FB  PC  EA = 1: Since BZ = BD = BF. By the intersecting chord theorem. q. q2 . p2 2 q p p 2p 2 = 1 2 r .

Let M be the intersection of PC and AD: Let B 0 be the mirror image of B with respect to line AP: Since BB 0 ? AP and 84 . cos2  + c2 d2 1 . q2 . q . b. cos2  + 2abcd sin sin : Using *. respectively. c.e.pq=2 = . x2 . which implies + = 180 : Therefore. suppose AB AC: Let P and Q be the feet of the perpendiculars from B and C to the angle bisector of 6 BAC. c = b + d: Let k be the length of a diagonal and angles and selected so that k2 = a2 + b2 . Let D be on line BC such that DA ? AP: Prove that lines BQ. the quadrilateral is cyclic. we can cancel abcd to obtain the equation 4 = 1 + cos 1 . we get the equation * p ab1 .pq=2: Since f x = . 2 cos + . Solution. d2 and divide by 2. cos  = cd1 . Since the quadrilateral has an inscribed circle. which is equivalent to ABCD cyclic. we have a + 86. In 4ABC.i. the determinants are equal if and only if r = s. PC and AD are concurrent. s . 1970 Putnam Exam Show p that if a convex quadrilateral with sidelengths a. cos  + 2 sin sin = 2 . then it is a cyclic quadrilateral. p s2 . cos : From the area ab sin + cd sin =2 = abcd. d and area abcd has an inscribed circle. 2cd cos : If we subtract a . cos  + 1 + cos 1 . Concurrent Lines Solution.p r p . . b2 = c . r p . q2 . p2 x2 . pq=2 is strictly decreasing when x  jpj and q. 2 2 2 2 2 2 p p p p 85. 2ab cos = c2 + d2 . we get 4abcd = a2b21 .

The circles with diameters AC and BD intersect at the points X and Y: The line XY meets BC at the point Z: Let P be a point on the line XY di erent from Z: The line CP intersects the circle with diameter AC at the points C and M. and the line BP intersects the circle with diameter BD at the points B and N: Prove that the lines AM. 2 2 PO4 and OO2 are parallel. we get BZ===AZ = CZ==DZ: Since 4CZP is similar to 4DZR0 and 4BZP is similar to 4AZR. BCP. Solution. This implies 6 BQP = 6 MQA: So line BQ passes through M. 1990 Chinese National Math Competition Diagonals AC and BD of a cyclic quadrilateral ABCD meets at P: Let the circumcenters of ABCD. ABP. Let line PO intersect the circumcircle of 4BCP and segment AD at points Q and R. O2O4 are concurrent. so PC intersects AD at its midpoint M: Now AQ = MC = AM = AM : PQ PC B 0 P BP 4BPQ. too. Due to Yu Chun Ling Let AR be parallel to BP and DR0 be parallel to CP. O2. Similarly. C and D be four distinct points on a line. so OO ? AD: Hence PO and OO are parallel.AD ? AP. Prove that OP. DN and XY are concurrent. O2O4 concur at G: 4 2 4 88. O1O3. PO1 OO3 is a parallelogram and diagonal O1O3 passes through G: Therefore. so BB 0 k AD: Then 4BCB 0. CDP and DAP be O. O3 and O4 . Since P is the midpoint of BB 0 . Now 6 PDR = 6 BCA = 6 PQB and 6 DPR = 6 QPB: So 6 DRP = 6 QBP = 90 and PO ? AD: Next circumcircles of ABCD and DAP share the common chord AD. 4DAC are similar. where R and R0 are points on line XY: Since BZ  ZD = XZ  ZY = CZ  ZA. OP. 87. So PO2 OO4 is a parallelogram and diagonal O2O4 passes through the midpoint G of OP: Similarly. 1995 IMO Let A. respectively. Solution 1. 4MAQ are similar. so ZP = BZ = CZ = ZP : ZR AZ DZ ZR0 85 . respectively. O1O3. B. O1. in that order.

x2 2 the intersection of DN with XY. Q. respectively. Let Y be the foot of the perpendicular from Q to BC and H be the orthocenter of 4ABC: Note that 6 ACF = 90 . R to AB. y0 y = x2 1 . so 6 EDH = 6 RQY: Hence. AMC. y0 : Since 0 . AD. Let the coordinates of P be 0. Let AM intersect XY at Q and DN intersect XY at Q0 : Observe that the right triangles AZQ. R are the midpoints of DE. so AZ=QZ = PZ=CZ: Then QZ = AZ  CZ=PZ = XZ  Y Z=PZ: Similarly. Solution. Q. Then the coordinates of A and C are x1 . r1 : Let Q be the intersection of 2 2 r 1 . x1 AM with XY. C. CF are the altitudes of 4ABC: If P. H. EF. 4BPC is similar to 4ARD: Since XY ? AD. Q. CA. CM k DR. 0 and x1 + r1 . since 6 ACF = 6 ABE. Therefore. let Q0 be 0 2 . 0 and x2 . x y 1 + r1 2 works out to be x1 + r1x . 6 CAB = 6 ABE: Since 6 CEH = 90 = 6 CDH. x2 1 = ZX = r2 . then Q0 has coordinates 0. 6 ABE = 6 FDH = 6 PQY: Next. BC. then Q has coordinates 0. Solution 2. BC. PZC are similar. x2 . DN. CA concur at the incenter of 4PQR: 86 90. the equation of AM AM ? CP and the slope of CP is . R to AB.Hence R and R0 must coincide. respectively. DN ? BN and BN k AR. BE. 0. y : Similarly. r2 y : 0 2 2 2 2 0 Since r1 . it follows the perpendiculars from P. ED k QR. XY are the extensions of the altitudes of 4ARD. QY bisects 6 PQR: From these. respectively. 0. r1. then show that the perpendicular from P. FD. 1988 Chinese Math Olympiad Training Test ABCDEF is a hexagon . E are concyclic. AM ? CM. the lines AM. respectively. So 6 ACF = 6 EDH: Now AH k QY. Due to Mok Tze Tao Set the origin at Z and the xaxis on line AD: Let the coordinates of the circumcenters of triangles AMC and BND be x1 . so Q = Q : Solution 3. are concurrent. D. 6 RQY = 6 EDH = 6 ACF: Similarly. hence they must be concurrent. and the circumradii be r1 and r2. Q0 Z = XZ  Y Z=PZ: Therefore Q = Q0 : 89.

6 CBG = 6 EFI. A2. we get AB  CD  EF 0 = BC  DE  F 0 A: Dividing this by *. CF are concurrent if and only if  CH  EI = AB  CD  EF : 1 = AG CG  EH  AI BC  DE  FA 91. we have EF 0 =EF = F 0 A=FA: If F 0 is on open arc AF. C1. we get AB=DE = BX=DX: Similarly. we can use Ceva's theorem and its converse. BE. B. so AB  CD  EF = BC  DE  FA: For the converse. EA and CF meet at I: Let h. A1. a contradiction. then FA F 0 A and EF 0 EF yielding EF 0 =EF 1 F 0 A=FA. we get AB  CD  EF =DE  FA  BC  = 1. A circle intersects a triangle ABC at six points A1 . Show that the diagonals AD. 6 CDH = 6 AFI: By Ceva's theorem and its converse. 87 . Suppose * AB  CD  EF = BC  DE  FA and AD intersect BE at X: Now let CX meet the circle again at F 0 : By the rst part. C to BE. Then Solution. B1 .inscribed in a circle. CF are concurrent if and only if AB  CD  EF = BC  DE  FA: From similar triangles ABX and EDX. C2. CE and AD meet at H. BE. CD=FA = DX=FX and EF=BC = FX=BX: Multiplying thse. C2. A2. AD. BE. If F 0 is on the open arc EF. we use the so-called method of false position. a contradiction. k be the distances from A. where the order of appearance along the triangle is A. CF concurs at X: AG = h = AB sin 6 ABG : CG k BC sin 6 CBG Similarly. respectively. then F 0 A FA and EF EF 0 yielding F 0 A=FA 1 EF 0 =EF. Let AC and BE meet at G. Due to Yu Ka Chun Suppose AD. C1. So F 0 = F: Alternatively. CH = CD sin 6 CDH and EI = EF sin 6 EFI : EH DE sin 6 EDH AI FA sin 6 AFI Now 6 ABG = 6 EDH. B2 .

sin Y BA sin C2C1 A1  sin C1A2 C2 sin ACZ = sin B1 A1 B2  sin A2B2 A1 : sin ZCB sin A2 A1 B1  sin A1 B2 A2 Using 6 C1 B1 C2 = 6 C1 A2C2 and similar angle equality. Due to Lam Po Leung Let dX. Then sin sin XAC sin DAB1 sin B2 B1 C1  sin B1 C2 B2 sin CBY = sin A1 C1 A2  sin C2A2 A1 . where H and H 0 are the orthocenters of triangles ABC and AB 0 C 0 . A: Suppose B1 C1. L  H 0Y H 0Y 2 Solution. By the converse of the trigonometric form of Ceva's theorem. 92. L1=dH 0 . 1995 IMO shortlisted problem A circle passing through vertices B and C of triangle ABC intersects sides AB and AC at C 0 and B 0 . L1 = dH. Y. L2 intersect at P forming four angles with vertex P . B2 . L2 . CZ are concurrent. BY. P. then H. Y. C2A2 meets at Y and A1 B1 . we see that the product of the three equations involving X. H 0 lie on an opposite pair of these angles. H 0 are collinear. BY. CZ are concurrent. L2 + 1 = HY + 1 = HH 0 : =d H 0 . B2 C2 are concurrent. by the trigonometric form of Ceva's theorem. Let D be the intersection of AX and B C : Since AX. Z above is equal to 1. Let HH 0 intersect L1 .C. L denote the distance 88 . CC 0 and HH 0 are concurrent. Then HH 0 = HX + 1 = dH. L1 dH. respectively. from a point X to a line L: For the problem. B C . L2 at X. Let lines L1. B1. sin C2AD  sin B2 B1 C1  sin B1 C2 B2 : 2  B1 B2  AC1 1 = DC = DB1  AB2  C1 C2 sin DAB1  sin C1B1 C2  sin B2 C2C1 BAX = sin C2 AD = sin C1 B1 C2  sin B2 C2 C1 : Similarly. respectively. C1A1 . B2 C2 meets at X . Proof. Prove that BB 0 . 1 2 1 1 Solution. we see that AX. A2B2 meets at Z: Show that AX. L1 + 1 H 0X H 0X dH 0 . Suppose H. we will use the following lemma. If dH. L2=dH 0. respectively. Lemma.

6 A = 6 AC 0 H 0. 4B 0H 0C 0 are similar because 6 CBH = 90 . HH 0 also passes through P: Perpendicular Lines 93. Set A at the origin and let y = b = 6 0 be the equation of the line . So KH=B 0H 0 = LH=C 0H 0: Since BH k C 0 H 0 and CH k B 0 H 0. let BB 0 . K. AF is perpendicular to CF. C. b. E. P. respectively. 6 B 0 CP = 6 ACH . hence it is P: Therefore. so dH. F be d.So X = Y is on both L1 and L2 . 6 A . CC 0: By the lemma. BB 0  = dH. L are concyclic. 4BHC. CC 0 intersect at L and CH. Since BE ? AE and slope of AE is b=e. E. b. BB 0=dH 0 . Also. 6 AC 0 B 0 = 6 C 0B 0 H 0 and similarly 6 BCH = 6 B 0 C 0 H 0: Therefore 4LHK. Then 4LHK. b2 + f 2  = 0 and the equation of line BC is dx + by . so the equation of line AE is ex + by . 6 C 0 BP = 90 . e. 6 B 0 CP = 6 PCH: So. H. Due to Cheung Pok Man There are many di erent pictures. the equation of line CF is fx + by . Prove that AN is perpendicular to NM: through D. CC 0=dH 0 . CC 0 intersect at P: Since 6 ABH = 90 . b2 + d2  = 0: 89 Solution. respectively. b. F: Note the case b = 0 implies D = E = F. 6 ACB = 90 . BB 0 intersect at K: Now 6 PBH = 6 ABH . which is not allowed. 4BHC are similar. CH k B 0 H 0: Let BH. B. and E and F are di erent from D: Let M and N be the midpoints of the line segments BC and EF. H 0 are collinear. so it is better to use coordinate geometry to cover all cases. 4B 0H 0C 0 are similar. f. b2 + e2 = 0: Similarly. 1998 APMO Let ABC be a triangle and D the foot of the altitude from A: Let E and F be on a line passing through D such that AE is perpendicular to BE. For the problem. Let the coordinates of D. so BH k C 0H 0 : Similarly.

N are d + 2 . de+df . b. respectively. B 0. b .1: Therefore. respectively. we may assume AB AC: Let AN intersect the circumcircle of 4ABC at D: Then 6 DBC = 6 DAC = 1 6 BAC = 6 DAB = 6 DCB: 2 So DB = DC and MD ? BC: With A as the center of homothety. Prove that QO is perpendicular to BC: x-axis on line NO: Let the equation of line AB be y = ax + b. Let the equation of BC be y = cx: Then B has cob . b and y = . at A meet the side BC: Let Q and P be the points in which the perpendicular at N to NA meets MA and BA.From these. a c. Then the coordinates of M. Due to Wong Chun Wai Set the origin at N and the . a ordinates ab. then 1 x + b. O. K are concyclic. A has coordinates . b . 0 and Q has coordinates 0. QO ? BC: Solution 2.a c+a c+a abc b . b. respectively. 2000 APMO Let ABC be a triangle. M has ordinates  c . b . O has coab coordinates  c2 . B to B 0 and C 16 to C 0: Then 6 OB 0 C 0 = 2 BAC = 6 OC 0B 0 and BC k B 0 C 0: Let B 0 C 0 cut PN at K: Then 6 OB 0K = 6 DAB = 6 OPK: So P. a respectively. So the slope of AN is 2b=e + f  and the 2b 2 +df e+f slope of MN is . C are d+e. AN ? MN: 94. 2b : The product of these slopes is . bc .  e+f . and O the point in which the perpendicular at P to BA meets AN produced. Without loss of generality. de b+f. a2 . C has coordinates . 1 c : Therefore.ax . the equation of lines AC and PO are y = . we found the coordinates of B. c2 . Hence 6 B 0 KO = 6 B 0 PO = 90 and so B 0 K = C 0 K: Since 90 Solution 1. b . bc . Due to Poon Wai Hoi The case AB = AC is clear. de2 b =d = . Let M and N be the points in which the median and the angle bisector. d+ e df f. . ab c : Then BC has slope c and QO has slope . respectively. 0. a2 . shrink D to O.

M has the midpoint Y of HM as center. points A. Let M be the intersection of DG and EF: Prove that AM ? BC: similarly EG = BC cos C sin C: Now GM = EG = cos C sin C = cos C AB : MD FD cos B sin B cos B AC Since BH = AB cos B. Let H be the foot of the perpendicular from A to BC: Now . 1996 Chinese Team Selection Test The semicircle with side BC of 4ABC as diameter intersects sides AB. B 0 . So D and H  are on the radical axis of . so by the intersecting chord theorem. K = Q: Since 6 B 0 KO = 90 and BC k B 0 C 0 . C 0. A0. Therefore. AC at points D. E to side BC respectively. respectively. G be the feet of the perpendiculars from D. DB 0  DC 0 = DA0  DM: Since 6 AC 0 H = 90 = 6 AB 0 H.BC k B 0 C 0 .2 are the same. this implies A. Since DB 0  DC 0 = DA0  DM. . M are collinear. Since 6 HA0M = 90 . H are collinear. M.1 and . AM ? BC: 97. HG = AE cos C.2 through H.1 with the midpoint X of AH as center. 1985 IMO A circle with center O passes through the vertices A and C of triangle ABC and intersects the segments AB and AC again at 91 6 BDC = 90 = 6 BEC: So DF = BD sin B = BC cos B sin B and Solution.2 : Then DH ? XY: By the midpoint theorem. H. K. we get BH = AB cos B = AC cos B and BH GM DA = 1: HG AE cos C AD cos C HG MD AB By the converse of Menelaus' theorem on 4BDG. the powers of D with respect to . so the circle . XY k AM: Therefore. C 0M lie on the nine-point circle of 4ABC. Let F.1 . DH ? AM: 96. we get QO ? BC: 95. Therefore. H the orthocenter of ABC and D the intersection of B 0 C 0 and BC: Prove that DH ? AM: Solution. Let A0 be the foot of the altitude from A to BC: Since A0. Let BB 0 and CC 0 be altitudes of triangle ABC: Assume that AB 6= AC: Let M be the midpoint of BC. B 0 lie on a circle . points A. E.

1997 Chinese Senoir High Math Competition A circle with center O is internally tangent to two circles inside it at points S and T: Suppose the two circles inside intersect at M and N with N closer to ST: Show that OM ? MN if and only if S. then 6 OMK = 90 = 6 OSK implies M. O. Then 6 SKT = 180 . K are collinear. T. N. 6 BAC = 180 . K. so OM ? KL: Hence OM ? BM: 98. S. K. 6 KTN: 92 . contradicting N is closer to ST: Let the tangent lines meet at K. then 6 OSK = 90 = 6 OTK implies O. S. Due to Leung Wai Ying Consider the tangent lines at S . So M. if OM ? MN. 6 NSK . T. The circumcircles of triangles ABC and KBN intersect at exactly two distinct points B and M: Prove that OM ? MB: Solution. so BM is parallel to KL: Now 6 LKM = 6 LKN + 6 NKM = 6 LCN + 6 NBM = 180 . respectively. O are concyclic. O are concyclic. N. KS 2 = KT 2 implies K is on the radical axis MN of the two inside circles.distinct points K and N. T are collinear. K. Then 6 OMN = 6 OMK = 6 OSK = 90 : Conversely. T are collinear. 6 SMT = 180 . 6 KAC = 6 KLC. Let CM intersect the circle with center O at a point L: Since 6 BMC = 180 . 6 TMN = 180 . 6 BMC = 6 BAC = 6 KLM: Then KM = LM: Since KO = LO. 6 SMN . N. then S. and at T: Suppose they are parallel. T lie on a circle with diameter OK: Also. S. then 6 SMT = 6 SMN + 6 TMN = 6 NSK + 6 KTN = 180 Solution. T will be collinear so that M and N will be equidistant from ST. If S. 6 SKT: So M.

! Then jOP k j = jOR + OS j  jOR j  1: 2 +3 1 2 2 +2 1 2 +3 2 +3 1 2 +3 1 2 +3 1 2 +3 101. DE meet lines BC. P2n+1 be distinct points on some half of the unit circle centered at the origin O: Show that . ! lies inside 6 P OP k : Since By the case n = k. . 6 SKT . .! :. So L. . Prove that AP + BQ + CR BC + CA + AB: 93 . 4DEF are the same. .. 6 B. 1973 IMO Let P1 . Show that L. By the intersecting chord theorem. 6 TNS = 360 . . CA. ! +    + OP . AB in points L.! . N are collinear. j OR . M. B.. P2. 99. respectively. so the line through L.. M. . When n = 0. are arranged . AD. Since 6 ADB = 90 = 6 AEB. Let the angle bisectors of 6 A. !j = 1 = jOP . ! . ... Since the circumcircle of 4DEF is the nine-point circle of 4ABC. OS bisects 6 P OP k : Hence 6 ROS  90 : jOP k . Maximum Minimum 100. Solution. Q. . ! . M..6 KTN = 180 : Therefore.. . T are collinear. ! .. FD. ! jOP 1 + OP2 +    + OP2n+1 j  1: 1 1 2 . 6 C of triangle ABC intersect its circumcircle at P. . !j = 1: Suppose the case n = k is true.. N.. . . . : : : . OR . !j  1: Also. M are also on this radical axis. 6 NSK . BE. ! . ! . we may assume P . L... M.. D. !: clockwise. Hence N is on the radical axis of these circles.. .. . : :. E are concyclic. P . ! . N. respectively. the center N of the nine-point circle is the midpoint of H and O: Since the radical axis is perpendicular to the line of centers O and N. Let OR = OP +    + OP and OS = OP + OP k k . !j. NA  NB = ND  NE: So the power of N with respect to the circumcircles of 4ABC. ! ! .Thus. .P k.. S. A. CF are the altitudes of 4ABC: Lines EF. N is perpendicular to the line HO: Geometric Inequalities. N are collinear and the line through them is perpendicular to the line joining the orthocenter H and circumcenter O of 4ABC: Solution. R. then jOP For the case n = k + 1. Similarly.

sin2 A sin2 B sin2 C and that equality holds i ABC is equilateral. 1997 APMO Let ABC be a triangle inscribed in a circle and let la = ma =Ma. then BQ is a diameter and we again get AB + CB = 2AB 2BQ: Similarly. where ma . mb . we get the desired inequality. which is equivalent to 6 A = 6 B = 6 C: 94 . 2AB  BQ cos 6 ABQ CQ2 = CB 2 + BQ2 . then subtracting these and simplifying. Due to Fung Ho Yin Let A0 be the point the angle bisector of 6 A extended to meet the circle. BC + AC 2CR and CA + BA 2AP: Adding these inequalities and dividing by 2. 102. we get A  = ma = sin B: So 2 2 ma = la = M a sin B sin C A   sin B sin C:  sin C + sinB + A 2 2 By the AM-GM inequality. Mc are the lengths of the angle bisectors extended until they meet the circle. la + lb + lc  sin B sin C + sin C sin A + sin A sin B  3 sin2 A sin2 B sin2 C sin2 A sin2 B sin2 C with equality if and only if sin A = sin B = sin C and C + A = B+ A = 2 2    = 90 . AB= sinC + Solution. lc = mc =Mc . Prove that la + lb + lc  3. we get AB + CB = 2BQ cos 6 ABQ 2BQ: If AB = CB. Applying sine law to 4ABA0. we have AQ = CQ: By cosine law.Solution. mc are the lengths of the angle bisectors internal to the triangle and Ma . 2CB  BQ cos 6 CBQ: If AB 6= CB. Mb . lb = mb =Mb. Due to Lau Lap Ming Since 6 ABQ = 6 CBQ. AQ2 = AB 2 + BQ2 . we get AB= sin C = Ma = sinB + A : Applying sine law to 4ABD.

EG ? MA: Let L be on MN such that AL k EG: Since NL k PB. respectively. Q be points on EG such that BP and CQ are perpendicular to BC: Prove that BP + CQ  BC + EG: When does equality hold? Solution. Prove that 6 AIO  90 if and only if 2BC  AB + CA: DC = DB because 6 CAD = 6 BAD and DB = DI because 6 BID = 6 BAD + 6 ABI = 6 CAD + 6 CBI = 6 DBC + 6 CBI = 6 DBI: Since ABDC is a cyclic quadrilateral. by Ptolemy's theorem. 104. so LNOA is a parallelogram. so MA ? AL. 6 AIO = 90 if and only if DI  AD=2. i. EG. 4GHA will coincide with 4ABC and O will move to M: So HA = BC. respectively. the converse of the well-known fact is true. Translating by GA.! Let H be the point so that HEAG is a parallelogram. Assume 4ABC is not equilateral so I 6= O. PB ? BC.e. O be midpoints of BC. which implies ML  MA: Therefore BP + CQ = 2MN = 2LN + ML 2AO + MA = 2BM + OE  = BC + EG: Equality holds if and only if L coincides with A. Due to Wong Chun Wai Let D be the intersection of ray AI and the circumcircle of 4ABC: It is well-known that DC = DB = DI: Solution. HA ? BC. PQ.e. which is equivalent to 2BC  AB + AC: Comments. Let M. i. then rotating by 90 about A. Then AO = LN: Since MA ? EG. AB = AC: 95 . N. BC ? HA. OE = OG = MA. we see 6 AIO = 90 if and only if 2BC = AB + AC: Also. Squares ABDE and ACFG are drawn outside 4ABC: Let P. Mathematics Magazine. In the solution above. Problem 1506 Let I and O be the incenter and circumcenter of 4ABC. since 6 DBC = 6 BAD always. AD  BC = AB  DC + AC  DB = AB + AC   DI: Then DI = AD  BC=AB + AC : Since 4AOD is isosceles..103. the point I on AD such that DC = DB = DI is the incenter of 4ABC: This is because 6 BID = 6 DBI if and only if 6 CBI = 6 ABI.

but no two sides of a regular pentagon are parallel. Then PD0 = GE 0 and PD0 + PE 0 = GE 0 + PE 0 GP: Solid or Space Geometry 106. Then the plane through A. So BD = CE and PD + PE = GE + PE = GP: For other D0 . by the de nition of G. Since a cube has 6 faces.105. Proposed by Italy for 1967 IMO Which regular polygons can be obtained and how by cutting a cube with a plane? Observe that if two sides of a polygon is on a face of the cube. 4CEG are congruent. 1995 Israeli Math Olympiad Four points are given in space. 96 Solution. To see this. Solution. otherwise the ve sides will be on ve faces of the cube implying two of the sides are on parallel planes. a regular pentagon is impossible.e. B. we have 4BPD0 .. Finally. C 0D0 . D0 with A0 on top of A. Due to Fan Wai Tong. we have 4BDP. CC 0. BC. D0 A0. Kee Wing Tao and Tam Siu Lung . 4. The perpendicular bisecting plane to edge AA0 cuts a square. 5 or 6 sides are possible. B 0 on top of B and so on. only regular polygon with 3. AC. 107. consider the point G such that G lies on the opposite side of line AC as B. 6 ABP = 6 ACG and CG = BP: Let E be the intersection of lines AC and PG: Let D be the intersection of AB with the circumcircle of APE: Since ADPE is a cyclic quadrilateral. Let the vertices of the bottom face of the cube be A. The plane through the midpoints of edges AB. 6 BDP = 6 AEP = 6 CEG: Using the de nition of G. 4CGE 0 are congruent. Point P is inside 4ABC: Determine points D on side AB and E on side AC such that BD = CE and PD + PE is minimum. B 0 . E 0 on sides AB. in general position i. B 0 . The minimum is attained when ADPE is a cyclic quadri- lateral. A0 A cuts a regular hexagon. they are not coplanar and any three are not collinear. then the whole polygon lies on the face. C 0. D0 cuts an equilateral triangle. C. respectively such that BD0 = CE 0 . D and the vertices on the top face be A0 .

plane. each containing one of the line and is parallel to the other line. again there is exactly one equalizing plane. otherwise they would lie in a plane parallel to the equalizing plane. The equalizing plane is the plane midway between these two plane.A plane  is called an equalizing plane if all four points have the same distance from : Find the number of equalizing planes. which is parallel to the plane P containing the three points and passing through the midpoint of the segment joining the fourth point x and the foot of the perpendicular from x to P: In the latter case. there is exactly one equalizing plane. Since there are 4 + 3 = 7 ways of dividing the four points into these two cases. Consider the two planes. The four points cannot all lie on one side of an equalizing 97 . Hence either three lie on one side and one on the other or two lie on each side. The two pair of points determine two skew lines in space. there are exactly 7 equalizing planes. Solution. In the former case.

Suppose after N terms. Does there exist a positive integer a such that the sum of the digits in base 10 of a is 1999 and the sum of the digits in base 10 of a2 is 19992? Solution. Yes. 10N . the sequence becomes periodic with period T: Then ai jT = ai for i  N. 10N . there is such a number whose digits consist of 1 2 k 0's and 1's. where k is so large that 10k n: Then the n consecutive integers p + 1. Let n be a positive integer and p = 1234567890  10k . 3 : : :: By the pigeonhole principle. Let k = 1999: Consider a = 10 + 10 +    + 10 : Then the sum of the digits of a is k: Now 2 2 2 2 2 a = 10 + 10 +    + 10 2 2 3 2 k+1 +2 1 X i jk 102i+2j : Observe that the exponent are all di erent by the uniqueness of base 2 representation. p + n have decimal representations beginning with 1234567890    and one of them is a multiple of n: 109. : : : become periodic after a nite number of terms? Solution. In fact. : : : that have the same remainder when divided by T. there are two numbers among 10N . the sum of the digits of a2 in base 10 is k = k2: k + 2C2 110. Does the sequence a1. 10m = jT for some integer j: + +1 +2 +3 98 . j = 1. say 10m 10k mod T  with N m k: Then 10k .Solutions to Number Theory Problems Digits 108. 1956 Putnam Exam Prove that every positive integer has a multiple whose decimal representation involves all ten digits. Therefore. Solution.    . p + 2. Proposed by USSR for 1991 IMO Let an be the last nonzero digit in the decimal representation of the number n!. 2. an. a2. : : :.

a2b2. 1+ jT: Since 10k .1 an mod 10: This implies an = 5: However. a 1 1 2 2 11 11 1 1 2 2 11 11 2 contradiction. Suppose a b . nS = 2 : Since there are 2 i2S k subsets of f1. : : :. : : :. 112. 11. a2. where for each subset S of f1. for i = 2.a b . kg. By symmetry. 1! implies a10k = a10k. x 10! 10 mod 11: On the other hand. kg. 2. : : :. where the ai 's are distinct 2 2 nonnegative integers. 2. so 9an = 9an. Xweaiget the sum of xnSk. b3. : : :. b11 be two permutations of the natural numbers 1. the exponent of 2 is greater than the exponent of 5. 11: Show that if each of the numbers a1b1. Let a1. 1 + x m 1 + x m mod 2: If we write n in base 2. By induction. we may assume a b 0 mod 11: Let x = a b     a b : On one hand. a11b11 is divided by 11. since ai bi 6 0 mod 11. an+1 = a10k+jT = a10k .1 : Let n = 10k . say n = 2a1 + 2a2 +    + 2ak . : : :. we get a = 11 = b : So x = a    a b    b  = 10! 10 1 mod 11. Solution. in the prime factorization of n!. 1956 Putnam Exam Prove that the number of odd binomial coe cients in any row of the Pascal triangle is a power of 2. 1 11 2 11 2 2 1 99 . a contradiction. 1  N. each with coe cient 1. 2. : : :. then at least two of them will have the same remainder. then 1 + xn = 1 + x2a1    1 + x2ak 1 + x2a1     1 + x2ak  mod 2: In expanding the expression in front of mod 2. : : :. a3. b2. Modulo Arithmetic 111. This implies there are exactly 2k odd binomial coe cients in the n-th row of the Pascal triangle. : : :. there are exactly 2 terms. a11 and b1.Observe that 10k ! = 10k 10k . 3.1+jT = an: Since n + 1! = 2  10k . a3b3 . Solution. 10m n! = 199    90    0n!.a b have distinct remainders when divided by 11. which implies an is even.

so 1 + kp . Suppose An is an arithmetic progression. : : :. 3 2 An . An = f1. since 1. then 1. . then 1. a 2 mod 3: If an. 2 2 An implies An = f1. If n is odd and n  3. 2. r as 4 6 2 mod 3: 1 2 +1 +2 2 2 Prime Factorization 114. : : :: Since the di erences of consecutive terms are multiples of 3. a power of 2 or n = 6: . 5. 1 is divisible by 3 and so 2p . Problem E2684 Let An be the set of positive integers which are less than n and are relatively prime to n: For which n 1. do the integers in An form an arithmetic progression? Solution. then an 2 .766. 1 = n . all an 2 mod 3: Then apaq = 6 ar for all p. if n is even and divisible by 3. 1 62 An: Consequently. 3. q. 11. 3 k = 2. 3. 1 for some k: This implies n 2 mod 6. 1g. An is an arithmetic progression if and only if n is a prime. a contradiction. 1971 IMO Prove that the set of integers of the form 2k . 1995 Czech-Slovak Match Let a1 . q and r such that apaq = ar ? Solution. an 2 mod 3. n2 an+1 + 2 + n2 an for all n  1: Do there exist indices p. 1 2 An . Due to Lau Lap Ming The rst few terms are 2.113. : : :. If n is even and not divisible by 3. n . : : : contains an in nite subset in which every two members are relatively prime. pg. 2 62 An imply An = f1. we suspect an 2 mod 3 for all n: Clearly. which implies n = 6 by 115. 5. : : : be a sequence satisfying a1 = 2. a . a2. 65. n 2 + 2 + n 2 = 8 2 mod 3: So by induction. 100 considering the prime factorization of n: Therefore. So p 5 mod 6: Then 2p . Finally. then let p be the smallest prime not dividing n: Either p 1 mod 6 or p 5 mod 6: In the former case. American Mathematical Monthly. a2 = 5 and an+2 = 2 . which implies n is a power of 2. 8. which implies n is prime. n . 1g. p are the rst two elements of An and n .

2 . 3 = 6 2ki . : : :. hence 2 . each of which does not contain products of its elements. either 3  81 = 243 is in X1 or 9  27 = 243 is in X2: In either case we are done. then we are done. . 2s: At least two of these will be congruent mod s. 3 is relatively prime to s: This implies 2 . the Euler -function of s. .Solution. 2 .3: This inductive construction can be repeated to form an in nite 1 2 1 2 0 1 2 1 2 +1 sequence. X2: Without loss of generality. . ng is partitioned into the union of two subsets. : : :. . which equals the number of positive integers less than s that are relatively prime to s. 9 is in X2 : If 92 = 81 is in X2. let 3 be in X1: If 32 = 9 is in X1 . an = 2kn . Due to Lam Pei Fung We rst show that 3 = 243 has the 5 . 2.3. De ne C to be prime" in S242 if C is not the product of elements of S242: The primes" in S242 consist of 4. 2p where p 242 is a usual prime number. Let a = 2 . p. . . 1988 Chinese Math Olympiad Training Test Determine the smallest value of the natural number n 3 with the property that whenever the set Sn = f3. i = 1. Otherwise. we will show that S242 can be partitioned into two subsets. so it must divide 2 . 1 is divisible by s and s is odd. : : :. 27 is in X2: Finally. 1. : : : .3 = 1: Suppose we have n pairwise relatively prime numbers a = 2k1 . Suppose S243 is partitioned into two subsets X1. each relatively prime to all the others. 2 . : : :. Now consider the s + 1 numbers 2 . Otherwise. then we will show it is the least solution. By Euler's theorem. n: So we may de ne an = 2 . We shall give a recipe for actually constructing an in nite set of integers of the form ai = 2ki . 3. then we are done. Comments. To show 243 is the smallest. at least one of the subsets contains three numbers a. Otherwise. say 2 2 mod s. 4. 3 for i = 1. 2. 1 = ms for some integer m: The odd number s does not divide 2 . then 2 s 1 mod s: 116. which is odd. or equivalently 2 2 . 1 = ls for some integer l: Since 2 . . we may take the exponent . a = 2k2 .3: We form the product s = a a    an. b and c not necessarily distinct such that ab = c: property. to be s. 8. 81 is in X1: If 81=3 = 27 is in X1. Since the smallest prime in S242 is 101 Solution.3. then we are done.

suppose mr . 1001. a contradiction. 118.1 + apk + higher terms: . 1 and at each step add the smallest integer which is not in arithmetic progression with any two preceding terms. are all 1's. then that digit for z is a 2. take m = 1: Next. Solution. 102 mr = 1 + p +    + pk. Since 1982 in base 2 is 11110111110. 4. no number in S242 is the product of more than four primes". so switching this from base 3 to base 10. 3. 1. We get 0. 100.e. z with x y z are 3 terms of the sequence in arithmetic progression. 10. so their products can be written with at least four prime" factors. 12. 1000. : : :: sequence: start with 0. 111. Consider the rightmost digit in base 3 where x di ers from y. and let X2 = S242 n X1: No products in X2 are in X2 because numbers in X2 have at least two prime" factors. 11. American Mathematical Monthly. Next looking at the product of 4. 9. 1983 IMO Can you choose 1983 pairwise distinct nonnegative integers less than 105 such that no three are in arithmetic progression? Solution. we get the 1983th term of the sequence is 87843 105 : To see this sequence work. ends in k 1's. suppose x.3. We prove by induction on k: For k = 1. 13. we see that a product of two primes" cannot be factored into a product of four primes". 1. 10. show that there exists a positive integer m such that the rightmost k digits of mr . y. when expressed in the base p. 8. 101. Put all the primes" and numbers that can be written as products of four primes" in one subset X1. 28. So no products in X1 are in X1: Base n Representations 117. 27. 2p p odd prime 242. Problem 2486 Let p be an odd prime number and r be a positive integer not divisible by p: For any positive integer k. this sequence is Note this sequence is the nonnegative integers in base 2. i. p. 110. in base p. We consider the greedy algorithm for constructing such a 0. : : :: In base 3.

27 will do. 24 = 9 + 15. then p k . then n = 9+n .1 cpk + higher terms. then n = 27 + n . the congruence a + rmr. Proposed by Romania p for 1985 IMO Show that the sequence fang de ned by an = n 2 for n = 1. Find all positive integers which cannot be written as the sum of two or more consecutive positive integers.1 + a + rmr. If d = 2. 2 = m 2k. 0 1 2 3 1 1 p 2 in base 2 as b :b b b : : : . gcdm. Write p Since 2 is irrational. then m + c0pk r will end in k + 1 1's as required. 33 will do. 34 = 9 + 25. p = 1: Then m + cpk r = mr + rmr. 2 : p p p p k k Multiplying byp 2 and adding 2. :bk    Let m = 2 2 . 1 2k.1 c 1 mod p is solvable for c: If c0 is a solution. 2 .Clearly. 36 = 9 + 27 can be so expressed. 2 . 103 . then n = 15 + n . 119. If d = 0. Solution. 21 will do. Find all even natural numbers n which can be written as a sum of two odd composite numbers. Let n  40 and d be its units digit. If d = 4. 121. For n 40. k . 15 will do.1 cpk +    + cr pkr i = 1 + p +    + pk. 30 = 9 + 21. Since gcdmr. If d = 6. 2. If d = 8. direct checking shows only 18 = 9 + 9. where each bi = 0 or 1. then n = 33+n . then in base 2. p = 1. 3. : : : where the brackets denote the greatest integer function contains an in nite number of integral powers of 2. Solution . 9 will do. there are in nitely many bk = p1: If bk = 1. then n = 21+n . 2 = b    bk. we get 2 m + 1 2 2 + : k 0 1 1 1 1 Then m + 1 2 = 2 : 2 2 Representations 120. 2 p p 1 p 2k.

Solution. Due to Cheung Pok Man For odd integer n = 2k + 1  3; n = k + k + 1: For even integer n  2; suppose n = m + m + 1 +    +m + r = 2m + rr +1=2 with m; r  1: Then 2m + r; r +1  2

and one of 2m + r; r + 1 is odd. So n must have an odd divisor greater than 1. In particular, n = 2j ; j = 0; 1; 2; : : :; cannot be written as the sum of consecutive positive integers. For the other even integers, n = 2j 2k +1 with j; k  1: If 2j k; then n = 2j , k+2j , k +1+   +2j + k: If 2j  k; then n = k , 2j +1+k , 2j +2+    +k +2j :

122. Proposed by Australia for 1990 IMO Observe that 9 = 4+5 = 2+3+4: Is there an integer N which can be written as a sum of 1990 consecutive positive integers and which can be written as a sum of more than one consecutive integers in exactly 1990 ways?

n + 2k : positive integer pairs n; k such that N = n + i= k + 1 2 i=0 Hence 2N can be factored as k +12n + k in exactly 1990 ways. Note if 2N = ab with 2  a b; then n = 1 + b , a=2; k = a , 1: This means 2N has exactly 2  1991 = 2  11  181 positive divisors. Now write 2N in prime factorization as 2  5e1  199e2     : Then we get 2  11  181 = 2e1 + 1e2 + 1    : So fe1; e2g = f10; 180g: Therefore, N = 510  199180 or 5180  19910 : As all the steps can be reversed, these are the only answers.
123. Show that if p 3 is prime, then pn cannot be the sum of two positive cubes for any n  1: What about p = 2 or 3?

Solution. For such N, we have N = m + i = 9952m + 1989: So i N is odd and is divisible by 995 = 5  199: Also, there are exactly 1990
=0

1989

X

k X

Solution. Suppose n is the smallest positive integer such that pn is the sum of two positive cubes, say pn = a + b = a + ba , ab + b : Then a + b = pk and a , ab + b = pn,k : Since a + b  2; so k 0: Since a , ab + b  ab 1; so n k: Now 3ab = a + b , a , ab + b  = p k , pn,k and 0 k n; so p j 3ab: Since p 3; so p j a or p j b: Since a + b = pk ; so p j a and p j b; say a = pA and b = pB: Then
3 3 2 2 2 2 2 2 2 2 2 2

A3 + B 3 = pn,3 ; contradicting the smallest property of n:
104

For p = 2; suppose a3 + b3 = 2n : If a + b 2; then 2 j a; 2 j b and b 3 = 2n,3 : So a = b = 2k and n = 3k + 1: a 3 +  2 2 For p = 3; suppose a3 + b3 = 3n : If a + b = 3k and a2 , ab + b2 = b 3 = 3n,3: 3 +  3 3n,k  32 ; then 9 j 3ab implies 3 j a; 3 j b and  a 3 Otherwise we have 3 = a2 , ab + b2  ab and then a + b = 3: So in this case, a; b are 2  3k ; 3k and n = 3k + 2: 124. Due to Paul Erdos and M. Sur anyi Prove that every integer k can be represented in in nitely many ways in the form k = 12  22    m2 for some positive integer m and some choice of signs + or ,.

Solution. We rst show every integer can be so represented in at least one way. If k can be represented, then changing all the signs, we see ,k also can be represented. So it su ces to do the nonnegative cases.
The key observation is the identity m + 12 , m + 22 , m + 32 + m + 42 = 4: Now 0 = 4 , 4 = ,12 , 22 + 32  , 42 + 52 + 62 , 72 ; 1 = 12 ; 2 = ,12 , 22 , 32 +42 ; 3 = ,12 +22 : By the identity, if k can be represented, then k + 4 can be represented. So by induction, every nonnegative integer and hence every integer can be represented. To see there are in nitely many such representations, we use the identity again. Observe 0 = 4 , 4 = m + 12 , m + 22 , m + 32 + m + 42 , m + 52 + m + 62 + m + 72 , m + 82 : So for every representation, we can add 8 more terms to get another representation. 125. 1996 IMO shortlisted problem A nite sequence of integers a0; a1; : : :; an is called quadratic if for each i 2 f1; 2; : : :; ng; jai , ai,1 j = i2: a Prove that for any two integers b and c; there exists a natural number n and a quadratic sequence with a0 = b and an = c: b Find the least natural number n for which there exists a quadratic sequence with a0 = 0 and an = 1996:

Solution. Part a follows from the last problem by letting k = c ,
105

b: For part b, consider ak in such a quadratic sequence. We have

ak  12 + 22 +    + k2 = kk + 12k + 1=6: So a17  1785: Also ak 12 + 22 +    + k2 mod 2: Since 12 + 22 +    + 182 is odd, n  19: To construct such a quadratic sequence with n = 19; rst note 12 + 22 +    + 192 = 2470: Now we write 2470 , 1996=2 = 237 = 142 + 52 + 42 : Then
1996 = 12 + 22 + 32 , 42 , 52 + 62 +    + 132 , 142 + 152 +    + 192: 126. Prove that every integer greater than 17 can be represented as a sum of three integers 1 which are pairwise relatively prime, and show that 17 does not have this property.

For 6k + 1; we split into cases 12k0 + 1 and 12k0 + 7: We have 12k0 +1 = 9+6k0 , 1+6k0 , 7 and 12k0 +7 = 3+6k0 , 1+6k0 +5: For 6k + 3; we split into cases 12k0 + 3 and 12k0 + 9: We have 12k0 +3 = 3+6k0 , 1+6k0 +1 and 12k0 +9 = 9+6k0 , 1+6k0 +1: For 6k + 5; we split into cases 12k0 + 5 and 12k0 + 11: We have 12k0 +5 = 9+6k0 , 5+6k0 +1 and 12k0 +11 = 3+6k0 +1+6k0 +7: Finally, 17 does not have the property. Otherwise, 17 = a + b + c; where a; b; c are relatively prime and a b c: Then a; b; c are odd. If a = 3; then 3 + 5 + 7 a + b + c 3 + 5 + 11 shows this is impossible. If a  5; then b  7; c  9 and a + b + c  21 17; again impossible.
Chinese Remainder Theorem

Solution. Due to Chan Kin Hang and Ng Ka Wing Let k  3: From 18 = 2+3+13; we see 2+3+6k , 5 works for 6k: From 20 = 3+4+13; we see 3 + 4 + 6k , 4 works for 6k + 2: From 22 = 2 + 3 + 17; we see 2 + 3 + 6k , 1 works for 6k + 4:

127. 1988 Chinese Team Selection Test De ne xn = 3xn,1 + 2 for all positive integers n: Prove that an integer value can be chosen for x0 so that x100 is divisible by 1998. 106

1 to be divisible by 1998 = 4  7  71. Find all positive integers a. : : :. which means x 0 mod 4. k are positive integers and k  2? Solution. 2d. where m. 1 2b . x 1 mod 7. hence is also a perfect power. Due to Cheung Pok Man Let n = 1+2+    +1992: Choose n distinct prime numbers p . : : :. 1992d are all perfect powers and any sum of them is a multiple of d. d is a pi -th power. Since e +1. pn : Let d = 2e2 3e3 4e4    nen . we get 2a + 1 = 2a. x 45 mod 71: 1 0 2 0 100 0 100 0 0 0 Since 4. b +    + 2a. e . Since b 2. 71 are pairwise relatively prime. then y = y + Solution.b + 2a.2 . b such that b 2 and 2a + 1 is divisible by 2b . b Every element in M and the sum of any number of elements in M have the form mk . by the Chinese remainder theorem. 1: Solution. 0  r b. It follows d. such x0 exists. 1 2 107 . less than or equal to nd. then by division. p .1 mod pi  and x 0 mod pj  for every 1  j  n.qb + 2r + 1 : 2b . : : : . 128. which implies n . : : :. 7. Divisibility 2 3 1 2 3 129. so 2b . en 0 mod p . e . Let yn = x n n . j = 6 i: Since the pi 's are pairwise relatively prime. en 1 2 0 mod p2 . 1 2a + 1. hence b a: Let a = qb + r. 1: We want x = x + 13 . such a solution exists by the Chinese remainder theorem. 2d is a p2 -th power and so on. where ei is a solution of the n equations x . Since e . Proposed by North Korea for 1992 IMO Does there exist a set M with the following properties: a The set M consists of 1992 natural numbers. 3n 3n 2 + + 2 : yn = y + 2 + 3 3 3n This gives xn = x + 13n .

Now 2 + 1 = 2 + 12 . Due to Cheng Kei Tsi and Leung Wai Ying For x = 1. 2n .1 . 1 1. Then n j 2n + 1 . completing the induction. then we get xxn + 2n + 1 xn+1 + 2n+1 + 1: Now xn+1 + 2n+1 + 1 =x .Since 0 2r + 1 2b . so 2 . . We use the fact x . y j x n = 3 t .1 . 2n.t we can tshow 2t j 3 t . Prove that there are in nitely many positive integers n such that 2n +1 is divisible by n: Find all such n's that are prime numbers. Solution. 2 = 3. 1998 Romanian Math Olympiad Find all positive integers x. yk for positive integer k: Consider Solution . 3: For x  4. 1 = 2t k: So n = 3 . by Euler's theorem. 33n + 2n + 1 3n+1 + 2n+1 + 1 23n + 2n + 1: So there are no solutions with x = 1. 2. 130. 3. . 2 + 1: By case k. if n  2. then by Fermat's little theorem. there are no solutions. we should have n j 2 k+1 k k k case k is true. 3 = 3 t  1 mod 2t : t t t Then n . 2n. 2 k = 3n. Show that there are in nitely many composite n such that 3n.1 is divisible by n: k . k = n + 1: The case k = 0 is clear. : : :: By induction. 2 t for t = 2. . n j 2n . If a prime n divides 2n +1. so n = 3: 23 132. 2 k + 1 . 2. Looking at the cases n = 1 to 10 suggest for n = 3k . n such that xn + 2n + 1 is a divisor of xn+1 + 2n+1 + 1: 21n +2n +1 1n+1 +2n+1 +1 1n +2n +1: For x = 2. : : :. 2. too. 22n +2n +1 2n+1 + 2n+1 + 1 2n + 2n + 1: For x = 3. 2 j 3 k .1 + 1 0 mod 3: So 3 3 3 2 3 3 3 2 3 2 + 1 is divisible by 3k+1 . : 2 2 2 2 2  2 2 2 2 2 1 1 131. 1 = n . 1: Alternatively. 1xn + 2n + 1 108 Solution. 1xn + 2n + 1 + xn . k k 2 k+1 1 mod 3. 1. Suppose 0. 2n + 1x + 3  2n + 2 x .

y : Completing squares. We may assume x y. n mod k +2: We can increase m by a large multiple of k + 1k + 2 to ensure it is positive and not equal to anyone of a1 . A4 = A3 A2 = 109 Solution.  . A3 = A2 A1 = 10. y. 1995 Russian Math Olympiad Is there a sequence of natural numbers in which every natural number occurs just once and moreover. ak : Let ak+1 = m and ak+2 = n: The sequence constructed this way have the property. then by the fact that prime p 3 mod 4 dividing x + y implies p divides x or y. 7. 2c. y = 3c. Now xn + 2n + 1 = x + 3 is a divisor of xn+1 + 2n+1 + 1 = x2 + 5 = x . we must have x + y = 5x .because for n = 2. we may cancel p to get an equation x + y = k x . y  = 3. : : :. c. where k j 1995 = 3  5  7  19: If p = 3 or 7 or 19 divides k.a1 . otherwise consider y. : : : the sum of the rst k terms is divisible by k? the property. 2n + 1x + 2n+1 = x2 .y. xn . there is an integer m such that m . ak . x = 4 or 11. for any k = 1. 1998 Putnam Exam Let A1 = 0 and A2 = 1: For n 2. Solution. 2c. 3x + 3 + 14 if and only if x + 3 is a divisor of 14. we get 2x . c. the number An is de ned by concatenating the decimal expansions of An. Let a = 1: Suppose a . y is an integer and divides 1995. y is such a pair. So the solutions are x. 1 0: Hence only n = 1 and x  4 are possible. 1 and 11.1 . ak has been chosen to have 1 1 . For example. 1: It follows x. c. Since x + y x x x . y  with k not divisible by 3. c.  . ak mod k +1 and m . 3. 19. y = 4.2 from left to right. 2n + 1x  x4n. xn . 1995 Bulgarian Math Competition Find all pairs of positive integers 2 2 x + y x. 2. 1 or 2. x: Then x + y = kx .1 and An. 3c. where c is a positive divisor of 3  7  19: 2 2 2 2 2 0 2 2 0 0 2 0 2 2 0 2 0 2 0 2 0 2 0 0 0 0 0 0 0 0 0 0 0 0 134. 1: 133. Suppose x. Since x + 3  7. 5x + 8 0 and for n  3. 5 + 2y + 5 = 50.a1 . which gives x . 135. Let n be the smallest natural number not yet appeared. 2n . By the Chinese remainder theorem. y for which x . : : :.

2pq . show that k does not divide 2k. q are prime and 2p + 2q is divisible by pq: Then 2p . mod 11: By induction. If p = 2. each of which is 1 or 2. Determine all n such that An is divisible by 11.2q mod p: If p.1 mod pj : j Then 2 pj . 2. then 2 .1 + 1: Use this to nd all prime numbers p and q such that 2p + 2q is divisible by pq: implies q = 3: Therefore. then 2p 2 mod p by Fermat's little theorem and 2q .101. 2. +An.1 mod pq.2 An. However. q 1 mod pj . Suppose p. The Fibonacci numbers Fn is de ned by F = 1. contradicting the rst part of the problem. . Let pi . . and so forth. Write m k = pe1    pe rr .2 An. then q = 2 or q 2: If q 2.2p .1Fn. q = 2 q qj . 3. there is a number whose decimal representation contains n digits. We will prove that the 2n numbers with n digits of 1's or 110 2's have di erent remainders when divided by 2n: Hence one of them is . q: Since pj j k andmk j 2 + 1. 0. so 2 q = 2k. In fact. . 2pj . Let mj = minfm . + Fn. Show that for any positive integer n. +An. + 1 for some k 1: Then k is odd. where pi 's are distinct primes. 1. A5 = A4 A3 = 10110. which 1 1 1 1 2 1  1 2 1  1 1 2 137. : : : . a contradiction. So we have the recursion An = 10Fn.2q . : : :: The rst eight terms of An mod 11 are 0. .1 mod pj  because qj is odd.2p . F = 1 and Fn = Fn. 1. for n 2: Note An has Fn digits. 1. we i get pe so k = 2mj q m +j 1 for some positive integer i 1 mod mj  and k .1. 3. 0. 1.2 mod q by Fermat's little theorem. So An is divisible by 11 if and only 1 2 1 2 1 2 1 2 if n = 6k + 1 for some positive integer k: +6 136. 1. the sequence Fn mod 2 is 1. 1995 Bulgarian Math Competition If k 1. the solutions are p. 2: Solution . Suppose k j 2k. the recursion implies An An mod 11 by induction. and which is divisible by 2n: Solution. 1. pj  = 1: So 2 pj . q are odd. Solution.2 mod p: So 2pq . q = 2. 2. mr g: Since mi 1 mod mj . 0. 1: Note the numbers start to repeat after the sixth term. 1 = 2 i qi withe qi odd. 1 mod pj  by Fermat's little theorem since gcd2.2 mod p: Similarly.2 mod q: Then 2pq.

b are k + 1-digit numbers. 138. then the units digits of a. where each digit equals 1 or 2. a f n! = n. completing the induction. and a b mod 2k+1 . b0: Since a0 . then 2k+1 divides a . b = 10b0 + i. b = 10a0 .divisible by 2n: For n = 1. this is clear. b0 are k-digit numbers with digits equal 1 or 2. If a = 10a0 + i. b0 is equivalent to 2k divides a0 . b are the same. Suppose this is true for n = k: Now if a. we have a0 = b0: So a = b. For a positive integer n. let f n be the largest integer k such that 2k divides n and gn be the sum of the digits in the binary representation of n: Prove that for any positive integer n. where i is the units digit.

, g n; n = 2n! if and only if n is not a power of 2. b 4 divides 2n n!n!

Solution. Due to Ng Ka Man and Poon Wing Chi a Write n in base 2 as ar ar,    a  : Then 

n  , 2 n : ai = ar    ai+1ai 2 , ar    ai+102 = 2 i 2i+1
r X i=0
+1

1

0 2

So

gn =

r 
n   r n X X  n ai = = n, 2i , 2 2i 2i = n , f n!: i=0 i=0 

b Let Mn = 2n!=n!2: Since g2n = gn; using a, we get

f Mn  = f 2n! , 2f n! = 2gn , g2n = gn:
So the largest k such that 2k divides Mn is k = gn: Now 4 divides Mn if and only if gn  2; which is equivalent to n not being a power of 2. 139. Proposed by Australia for 1992 IMO Prove that for any positive integer m; there exist an in nite number of pairs of integers x; y such that 111

a x and y are relatively prime; b y divides x2 + m; c x divides y2 + m:

Solution. Note x; y = 1; 1 is such a pair. Now, if x; y is such a pair with x  y; then consider y; z ; where y + m = xz: Then
2

every common divisor of z and y is a divisor of m; and hence of x: So gcdz; y = 1: Now

x2 z 2 + m = y2 + m + x2 m = y4 + 2my2 + mx2 + m
is divisible by y: Since gcdx; y = 1; y j z 2 + m; so y; z  is another such pair with y  y2 =x z: This can be repeated in nitely many times. 140. Find all integers n 1 such that 1n + 2n +    + n , 1n is divisible by n:

Solution. For odd n = 2j + 1 1; since n , kn kn mod n for 1  k  j; so 1n +2n +    +n , 1n is divisible by n: For even n; write n = 2st; where t is odd. Then 2s j 1n + 2n +    + n , 1n: Now if k is even and less than n;s then 2s j kn : If k is odd and less than n; then by Euler's theorem, k , 1 mod 2s ; so kn 1 mod 2s: Then 0 1n + 2n +    + n , 1n n mod 2s; which implies 2s j n; a
2 1

contradiction. So only odd n 1 has the property.

2

+1

141. 1972 Putnam Exam Show that if n is an integer greater than 1, then n does not divide 2n , 1:

Solution. Suppose n j 2n , 1 for some n 1: Since 2n , 1 is odd, so n is odd. Let p be the smallest prime divisor of n: Then p j 2n , 1; so 2n 1 mod p: By Fermat's little theorem, 2p, 1 mod p: Let k be the smallest positive integer such that 2k 1 mod p: Then k j n because otherwise n = kq + r with 0 r k and 1 2n = 2k q 2r 2r mod p; contradicting k being smallest. Similarly k j p , 1: So k j gcdn; p , 1: Now d = gcdn; p , 1 must be 1 since d j n; d  p , 1 and p is the smallest prime divisor of n: So k = 1 and 2 = 2k
1

1 mod p; a contradiction.

112

142. Proposed by Romania for 1985 IMO For k  2; let n1 ; n2; : : :; nk be positive integers such that

n2 2n1 , 1; n3 2n2 , 1; : : : ; nk 2nk,1 , 1; n1 2nk , 1: Prove that n1 = n2 =    = nk = 1:

Solution. Observe that if ni = 1 for some i; then ni will equal 1 and the chain e ect causes all of them to be 1. So assume no ni is 1. Let pk be the smallest prime number dividing nk : Then pk j 2nk,1 , 1: So 2nk+1 1 mod pk : Let mk be the smallest positive integer m such that 2m 1 mod pk : Then mk j nk, and mk j pk , 1 by Fermat's little theorem. In particular 1 mk  pk , 1 pk and so the smallest prime divisor pk, of nk, is less than pk : Then we get the contradiction that pk pk,    p pk :
+1 1 1 1 1 1

143. 1998 APMO Determine the largest of all integer n with thepproperty that n is divisible by all positive integers that are less than 3 n:

Solution. Due to Lau Lap Ming The largest n is 420. Since 420 = p 3  4  5  7 and 7 3 n 8; 420 has the property. Next, if n has the property and n 420; then 3; 4; 5; 7 divide n: Hence n  840 729 = 9 : Then 5; 7; 8; 9 divide n; so n  5  7  8  9 = 2420 2197 = 13 : Then 5; 7; 8; 9; 11; 13 divide n; so n  5p  7  8  9  11  13 k 3 8000 19 : Let k be the integer such that 19 n 19k : Then
3 3

5k ; 7k ; 9k ; 11k ; 13k ; 16k ; 17k ; 19k divide n; and we get the following contradiction n  5k 7k 9k 11k 13k 16k 17k 19k = 19k 4  5k 3  7k 2  11k 2  13k 3  17k 193k+3  n:

3

+1

144. 1997 Ukrainian Math Olympiad Find the smallest integer n such that among any n integers with possible repetitions, there exist 18 integers whose sum is divisible by 18.

Solution. Taking seventeen 0's and seventeen 1's, we see that the
smallest such integer n cannot be 34 or less. We will show 35 is the 113

in expansion. the terms ae 1    ap have exponent sum e1 +    + ep  p . since 18 = 2  32. The statement is true for every positive integer k: All we have to consider is the case k = p is prime. we can take out 2k1 . 1 integers. 1 sums s1 . ap: By Fermat's little theorem.answer. then the sum of those three will be 0 mod 3. where the sum is over all m pickings a1. Consider the statement among any 2k . Since the case k = k2 is true. then consider X S = a1 +    + app. Suppose 2p . 1 integers. from the 2k2 .1. 1 integers to the remaining integers and repeat the taking and picking. for 2k1 k2 . 1 groups. There are p p .1 . Totally we we will get 2k2 . consider mod 3 of the integers. Among 2  3 . If 0. considering the numbers di = si =gcdk1 . there are two odd or two even numbers. Comments. 1: Hence the numbers of nonzero exponents ei in the terms will be positive integers j  p . otherwise there are two choices for 5 integers and three of them will be congruent mod 3. Among 2  2 .1 of the groups. 1: Since p . 1 = 5 integers. S 1 +    + 1 = m 6 0 mod p: ep 1 On the other hand. there exist k of them whose sum is divisible by k:" We will rst show that if the statement is true for k = k1 and k2 . we can get k2 of them whose sum is divisible by k2 : The union of the k2 groups with sum si 's consists of k1k2 numbers whose sum is then divisible by k1 k2 : To nish the problem. 1 = 3 numbers. we have to show the statement is true for k = 2 and 3. 1 of them and pick k1 of them with sum divisible by k1 to form a group. s2k2 . 1 integers are given. : : :. Then return the other k1 . j of the ei is 114 . then it is true for k = k1 k2 : Suppose it is true for k = k1 and k2 : Since the case k = k1 is true.2 each appears. k2. : : :. their sum is even. If no p of them have a sum divisible by p. 1! ways in picking p of them. their sum is 0 mod 3.

12p .2p . 1 2p . 2    p + 1 m= = .

0. j ai's whose exponents ei = 0 is . the coe cient of the term equals to the number of ways of choosing the p .

j + 1 = . where p. b.1 y mod 7. a contradiction. b. p. j 2p . 2 : This implies m is even. By algebra. qp+q. 1 = 7y : In the former case. c = q .c c p a . q are positive integers and gcdp. j    p    p . 1969 Eotvos-Kursch Competition Let n be a positive integer. say k = 2j: Then 7n = j j . we have a = pp + q. Perfect Squares. b = qp + q and c = pq: Therefore a + b = p + q 2 : 146. then it is a square. 2 115 . say m = 2k: So 28n = k . q= 1 implies gcdpp+q. a b c c b. 1 . which is impossible. j . 1 + 1 = 1 is equivalent to a .j p . j . c be positive integers such that 1 a b c common divisor of a.2p . Perfect Cubes 1 = 1 : If the greatest + 145. j . then 428n +1 = m . So 1: Then p + q q p+q p a = b = c: pp + q qp + q pq Now gcdp.2k: This implies k is even. Let a. 1 . then prove that a + b must be a perfect square. c is 1. If 2+2 28n + 1 = m. 1 = y or j = x . we get . 1: Since gcdj. p ak Math 2 Show that if 2 + 2 28n + 1 is an integer. q = a = c and b = c by simple algebra. 2 2 2 2 m = 2k = 4j = 4x2 is a square. In the latter case. pq = 1: Since gcda. c = 1. c pose c = b . hence S 0 mod p. b. j ! which is divisible by p: So all coe cients are divisible by p. either j = 7x . an integer. c = c : SupSolution. 1 = 1. 2 2 2 2 2 2 p Solution.

1998 Putnam Exam Prove p that. c = 1. let ak = ak : In the latter case. It su ces to show there is a sequence of positive integers ak such that ak . let p be a prime divisor of abc: Since gcda.7 mod 2k  and the ak 's have no maximum. b. b. for any integers a. Otherwise. 7 for all k. Let P x = x + ax + bx + c and n = jbj + 1: Observe that P n P n + 2 mod 2: Suppose both P n and P n + 2 are perfect squares. Let a. the sequence has no maximum. 7 mod 2k : In the former case. c = 1: . Otherwise.7 mod 2k : Since ak  2k . ak + 2k ak ak + 2k . where n is a positive integer. we may assume p divides 116 Solution. Without loss of generality. 1995 IMO shortlisted problem Let k be a positive integer. b. Since perfect squares are congruent to 0 or 1 mod 4. P n = 2n + 2b 3 2 2 Solution. c0 = c=d. Solution. Prove that there are in nitely many perfect squares of the form n2k . b. P n + 2 . either P n or P n + 2 is not an integer. we may assume gcda. Multiplying by abc. b0. c: By symmetry.7 mod 2k : Then either ak . 148. c. c be integers such that a + b c a cube of an integer. so P n P n + 2 mod 4: However. c. is not divisible by 4. then we are done. let ak = ak +2k. the equation still holds for a0 . Therefore. : Then since k  3 and ak is odd. Let a = a = a + 3 = 1: For k  3. there exists a positive integer n such that n3 + an2 + bn + c is not an integer. b. the new equation implies that p divides exactly two of a. if d = gcda. b. b0 = b=d.147.7 mod 2k  or ak 2k . p So either P pn or P n + 2 is not a perfect square. 2 1 2 2 2 +1 2 +1 +1 +1 1 2 +1 2 2 2 2 2 +1 2 b + c = 3: Prove that abc is the 149. then for a0 = a=d. 7. ak = ak + 2k ak + 2 k. we get a new equation a2c+b2 a+c2 b = 3abc: If abc = 1. c0 and a0b0 c0 is a cube if and only if abc is a cube. a contradiction. suppose ak .

We will show there is exactly one set of solution. Due to Cheung Pok Man Since 3 = 4 = 5 =    . a contradiction.1y 0 mod 3 and . If n 2m. since their sum is not divisible by 3. 2. then 5 5w +2y = 3x 1 or 3 mod 8. 2y = 1. Due to W. 3 1 3 1 4 1 5 2 is not a solution to xy + yz = z x : The second equation has no solution either since 2z z: The rst equation leads to the unique solution 1. . Solution. 2y = 1 implies w = 1 and z = 2. but not c: Suppose the largest powers of p dividing a. respectively. y are 1. b2 c. b2a. To simplify the equation. From these. the only possible values for x. We have 1 = 0 + 1y 3x + 4y = 5z . namely x = y = z = 2. Finally. 1. If y 2. say z = 2w. 117 . 3abc: Hence pn+1 j c2b. 1+2z = z and 4+2z = z : The third equation has no solution since 2z  z for z  4 and 2. . Therefore. 1. 2 and 2. b are m. So y = 2. Then. 2: 2 151. we consider modulo 3. Sierpinski in 1955 Find all positive integral solutions of 3x + 4y = 5z . we have yz  z y if y  3: Hence the equation has no solution if y  3: Since 1  x  y. a contradiction. 2y : Now 5w + 2y and 5w . pjabc Diophantine Equations 150. y and z such that x  y  z and xy + yz = z x : Solution. 3abc: Hence pY j a2c.a. 5w + 2y = 3x and 5w . It follows that z must be even.1y 1 mod 3. n. 4y = 5w + 2y 5w .1z mod 3. n = 2m and abc = p3m is a cube. If n 2m. b. a contradiction. forcing p j c. we get w is odd and y is even.1w . 2y are not both divisible by 3. Then 5w . forcing p j c. So. 1. we get x = 2.1w + . then n  2m+1 and 2m+1 p2m+1 j c2 b. then n + 1  2m and pn+1 j a2c. 2: These lead to the equations 1+1 = z. Then 3x = 5z . Find all sets of positive integers x.

yj=2 should be a solution for the case n + 1: If x + y=2 is even. yj=2. then the other is also odd. j7x . y = 4. a pattern begins to emerge. 3: If z = 2.1 2 + 32 = 4b + 1 + 33 : It follows or 5. we have a solution x. 7x . Due to Euler. y = 5. yj=2 is odd.1 2 + 1 = b + 1 + 2 : It follows that b = 2 or 3. If x. 153. x. 3 for case n = 4: Suppose in case n. Suppose x. exactly one of x + y=2. jx . Also. then 7 x + 2 2 . which gives c = bz 2.  . 1995 IMO shortlisted problem Find all positive integers x and y such that x + y2 + z 3 = xyz. 6: 118 Solution. 2 is even. then prove that 2n can be represented in the form 2n = 7x2 + y2 with x. then jx . Let x = az. y is a pair of solution. y 2 = 14x2 + 2y2 = 2n+1: Therefore. After working out solutions for the rst few cases. y = bz. 1 b. y is a solution to case n. if x. y is odd. 2 c = bb . we observe that since x + y=2 + jx . b are positive integers and gcda. 1 4b . so x. then the pattern suggests the following: If x + y=2 is odd. so x. yj 2 = 14x2 + 2y2 = 2n+1: If x + y=2 odd. y odd positive integers. yj=2 = maxx.152. For the case n = 3. y 2 + . x. j7x . y = 1. y=2 should be a solution for the case n +1: Before we con rm this. Now we con rm the pattern by induction. b = 1. where z is the greatest common divisor of x and y: where a. j x . . then we have c + b2 + z = cbz 2. Similarly. y j . y=2. exactly one of 7x . yj=2 is odd. y: If x + y=2 is . y is a solution and one of Solution. then 16c = 164bb . The equation implies a + b2z + z 2 = abz 2: Hence a = cz for some integer c and b2 + z : If z = 1. 2 or 4. also 1985 Moscow Math Olympiad If n  3. then x + y=2. 1 that b = 1 or 3. j7x . 1 with 1 + 1=2 = 1 leads to a solution 1. then 7 2 + 7x 2 the pattern is true for all cases by induction. y is odd.

z  is also a positive integral solution with y  x and z = 6 3: However y x 2 2 2 0 2 0 0 0 2 0 0 0 1 2 0 0 2 0 1 0 0 0 2 0 0 0 0 0 0 1 0 contradicts x0 being least possible. 4. 6. z = 4: Therefore. y. 4z . 2.1 bz 2 . 119 0 1 0 0 0 . let x . b = In general. 3. Find all positive integral solutions to the equation xy + yz + zx = xyz + 2: 1 + 1 + 1 = 1 + 2 : So sides by xyz. y.2 2 z + z 3 = b + b + z 3 : Now integer c2 z . it is between z 2 . As the equation is symmetric in x. The other root is x = y z . 3. z  be a solution with x least possible. 3: 154. y = 4. z  with x  y and z = 6 3. z = 3. 3. 2. 1 2 2 integer solution of w . then the equation implies y = z = 1: If x = 2. 4. then z 2 . z + 1 : If z  3. 1 z. 3. 2 = 1 and so x = 1. 4. If x = 1. x . Now x . 2. z with z = 6 3: Then x = 6 y for otherwise 2x +1 = x z would give x z . However. y z x + y + 1 = 0 has x as a root. 1 z. 1. 2 and 5. 1. 4. 4. 2: 155.1 z2 .1 2 2 b + z  z + z 2. z. y. 2. so c = 1: Now b is an so b  z: It follows that c = bz 2. z w + z +1 = 0: So the discriminant z 4 . 5. Show that if the equation x2 + y2 + 1 = xyz has positive integral solutions x. a contradiction. z + 1 z + 1. Therefore. 2. y . 2. 3. then z = 3: Solution. c2z = bbz 2. y. By symmetry. 4 is a square. 1 = 1 + 1 : So 1 1 + 1 = 1 + 1  2 : Then y 4: Simthen 1 + z y 2 yz 2 2 yz z y y ple checkings yield y = 3. Due to Chan Kin Hang Suppose the equation has positive integral solutions x. we get z y x xyz 2 = 1 + 1 + 1  3: 1 1 + xyz z y x x Then x = 1 or 2. we may assume x y: Among the positive integral solutions x. y. 4. 3. bz 2 . z  = 1. the required solutions are x. 1 2 3 b + z  1 implies b  z . we may assume x  y  z: Dividing both Solution. x = y + 1=x : We have 0 x = y + 1=x  y + 1=y + 1  y : Now y . 4. 12 and z 2 2 . the only solutions are x.

156.1 . then x. y = 0. 0 and p may be arbitrary. y + 1 and so y + 1 = pn for some n: If n = 0. where p is a given odd prime. px = yp + 1 = y + 1yp. 1p + 1 . yp = 1. 1995 Czech-Slovak Match Find all pairs of nonnegative integers x and y which solve the equation px . px = pn . Otherwise.    + y2 .

p .

p = pnp .1 + 2 pnp. all of the binomial coe cients are divisible by p: Hence all terms are divisible by pn+1. p . 2 p2n + p  pn : Since p is prime. p  pnp. and all but the last by pn+2 : Therefore the highest power of p dividing the right side is pn+1 and so x = n + 1: We also have .2 +    .

p .

we get 8 = 3 . z  + 3 . the only solutions are x. 4. y = 2. 3  32n . y = 0. y: Since 6 = 3 . 0 for all odd prime p and x. which implies n = 1 and p is not divisible by p2 . 1. 1. then Solution. y. p. . 4. 2 but the lat on the right is divisible by p2n+2. z  is a solution. From the identity x + y + z  . we see that the solutions are 1. . 4. 2 p2n: For p = 3. 4. p 0 = pnp . y = 2. y: Checking the factorization of 8. this is a contradiction. x3 . this gives 0 =. Since 0 is divisible by p2n+2. Find all integer solutions of the system of equations x + y + z = 3 and x3 + y3 + z 3 = 3: Solution. If x. 2 for p = 3: 157. x + y + z  = 3x + yy + z z + x. 4. Therefore.5.1 + 2 pnp.5: 3 3 3 3 120 .5. while the last term is not. p . p  pnp. so every term x. . x + 3 .2 +    . z 3 . Suppose x. 2: For p  5. 4. y is a solution. 33n .

the number of n letter words with 2k a's n 2n. 1 or 2 in front of an n-word. such that any two neighboring points have distinct colors. yn is also the number of n-words satisfying the condition and beginning with 2. we get xn+1 = 3xn . Solution 1. we 2n. 1 = 0 hasp roots p n = 1. 1. Let xn be the number of n-words satisfying the condition. 2xn+1 . Tn. 1996 Italian Mathematical Olympiad Given an alphabet with three letters a. Solution 2. Due to Tam Siu Lung Let Sn be the number of n letter words with even number of a's and Tn be the number of n letter words with odd number of a's. Tn = S1 . : : :. 2yn and yn+1 = xn . n+1 where 1+ 2=2. = 1 . we get xn+2 . Sn = 3n + 1=2: 159. So x1 = 3. Due to Chao Khek Lun and Ng Ka Wing For a nonnegative even integer 2k  n. c. Find the number of n-words from the alphabet A = f0. A2.1: Similarly Tn = Sn. which can be is C2 k simpli ed to 2 + 1n + 2 .1 : So Sn . Considering a 0. xn = 0: For convenience. then substituting into the second equation. Then Sn + Tn = 3n: Among the Sn words.p 2r . set x0 = x2 . 1n =2 using binomial expansion. p  2. 2x1 = 1: Since r2 . By interchanging 0 and 2. T1 = 2 . 2 =2: Solution.2k : The answer is the sum of these numbers. x2 = 7: Let yn be the number of n-words satisfying the condition and beginning with 0. x1 = 3. An be points on a circle. Find the number of possible colorings of these points with p colors. 1995 Romanian Math Olympiad Let A1 . yn : Solving for yn in the rst equation. xn = 1+ 2 + p = n +1 1 . b.1 + 2Tn. there are Tn.1 + 2Sn. 121 . 160.1 : Subtracting these. 2=2: Therefore. we get Sn . if any two neighbors can di er by at most 1.1 .1 words ended in a and 2Sn.Solutions to Combinatorics Problems Counting Methods 158. 1 = 1: Therefore. 2g.1 words ended in b or c: So we get Sn = Tn. 1  2 and x0 p pget xn = 1 + 2 + 1 p. Tn = Sn. nd the number of words of n letters which contain an even number of a's.

: : :. Let Cn be the answer for n points. 1Cn + Cn. By the pigeonhole principle. 1 ways. the triangle can lie in at most two adjacent quadrilaterals. 162. 2Cn + p . 2. C + C  = pp . Let A be the subset of S = f1. : : :. for n  3. Show that one side of the triangle has length less than 1. form a set Sx = fx. We have C = p. A simple count shows A0 has 1999 elements.1n p . Due to To Kar Keung Through the center c of the square. then A . An can be colored in Cn ways. 1n + . 2. C = pp . 1n. So Cn = p . while An can be colored in p . 1 for n 3 by induction. 1 and C = pp . hence of ratio 2. 1n : Then Cn = p . : : :. by the pigeonhole principle. 2xg S: Note the union of all Sx 's contains S: So. 3000g contain a subset A consisting of 2000 numbers such that x 2 A implies 2x 62 A? numbers of the form 4nk. then A . 2 ways. 1 2 3 1 1 +1 1 1 1 +1 +1 1 +1 +1 2 1 3 2 Pigeonhole Principle 161. 1p . Now the furthest distance between two points 122 Solution. if A and An have di erent colors.Solution. 3000g containing all 0 Solution. Now for each x 2 A0 . : : :. So no subset of 2000 numbers in S has the property. ways and An can be colored in p . two of the vertices of the triangle must belong to the same quadrilateral. draw a line L1 parallel to the closest side of the triangle and a second line L2 perpendicular to L1 at c. which can be written as Cn + Cn = p . 2: For n + 1 points. where n is a nonnegative integer and k is an odd positive integer. any subset of S having more than 1999 elements must contain a pair in some Sx . : This implies Cn + Cn = p . The lines L1 and L2 divide the square into four congruent quadrilaterals. An can be colored in Cn. Since c is not inside the triangle. If A and An have the same color. . Then no two elements of A0 have ratio 2. 1987 Austrian-Polish Math Competition Does the set f1. 1989 Polish Math Olympiad Suppose a triangle can be placed inside a square of unit area in such a way that the center of the square is not inside the triangle. 1Cn.

is the statement still true in general? at most 1 piece. locate the leftmost square that contains a piece. Then there are C k + C . These 4 pieces yield a parallelogram. Due to Ho Wing Yip Let m be the number of rows that have at least 2 pieces.e. 1  63 . then there are at least ji . there are at least n + m . This implies there are at least two rows each containing 2 pieces that are of the same distance apart. suppose there are k black cells and 7 . 1. at least two of these distances are the same. 2. 1 rectangles for that combination. For n 1. For a row. 42 = 21 2 7 2 2 7 2 42 such rectangles. Prove that there exist at least 21 rectangles with vertices of the same color and with sides parallel to the sides of the square. The cells of a 7  7 square are colored with two colors. k white cells. number of squares between this piece and the other pieces on the same row. let 2n chess pieces be placed at the centers of 2n squares of an n  n chessboard. The distances can only be 1. Then each of the remaining n . Record the distances i. Since the number of pieces in these m rows altogether is at least 2n . m = n + m. which is at most 1. if there are ji pairs in the same combination. Next there are C = 21 pairs of columns.k = k . n . 163. For each of these m rows. n .in the quadrilateral is the distance between two opposite vertices. For combination i = 1 to 42. 1 because there are n columns. Since the sum of X the ji's is at least 63. So there are 21  2 = 42 combinations of color and pair of columns. Show that there are four pieces among them that formed the vertices of a parallelogram. so there are at least ji . By the pigeonhole principle. Due to Wong Chun Wai Let the colors be black and white. 123 Solution. : : :. So there are at least 7  9 = 63 pairs of cells on the same row with the same color. So the side of the triangle with two vertices lying in the same quadrilateral must have length less than 1. If 2n is replaced by 2n . m rows contains . i=1 164. 7k + 21  9 pairs of cells with the same color. Solution. m = n distances recorded altogether for these m rows.

Thus. Let m  n 0: Find the number of surjective functions from Bm = f1. z1 . If one of the remaining di erences belongs to X. one of the subsets. placing 2n . : : :. z5 . c in one of X. then we are done. yj1 = xijk . say Y 6= X . xij . belong to X. xi1 are di erent elements of X and xij . y3 . b. The remaining di erence zk . Y or Z such that a + b = c: Inclusion-Exclusion Principle Solution. must contain at least 49=3 elements. For i = 1. y1 . 2. z4 . xi1  + xi1 = xij and we are done. : : :. n. c 1 2 17 2 1 3 1 17 1 1 166. x . z1 . yij1 must belong to one of X. then let xi1 . If one of them. belong to the remainging subset Z and say z1 z2 : : : z5 : Form the di erence z2 . z1 = yjk . one of the subsets. say x x : : : x : Form the di erences x . yj1 . ng: Solution. by the pigeonhole principle again. 2. By the pigeonhole principle. say y1 y2 : : : y8 : Consider the di erences y2 . z1 . y1 and remove y1 and xi1 if they appear on the list. : : :. let Ai be the set of functions f : Bm ! 124 . z1 and remove z1. y1 . we get three distinct elements a. we may assume 5 of these di erences zk = yjk . : : :. b. 165. must contain at least 15=2 elements from these di erences yj = xij . xi1 6= xi1 . z3 . If one of these di erences belong to Y. Show that at least one of the subsets contains three di erent numbers a. say X. c such that a + b = c: are to be di erent if it appears on the list. x .For the second question. xij . 49g is partitioned into three subsets. xij1 if they appear on the list. : : :. say yj . : : :. b. The set f1. x . Otherwise. mg to Bn = f1. 2. y1 . then we are done. x and remove x because a. y1 = xij . x1 . y8 . xij . 2. 1 pieces on the squares of the rst row and rst column shows there are no parallelograms. x . Y or Z: As above.

: : : . f m: By the inclusion-exclusion principle. jA1    An j X X = jAi j . jAi Aj j + in .Bn such that i 6= f 1.

 n 1 1 X .

2 n .    : The number of surjections from Bm to Bn is . 3m . 1m . 2m + 3 n .ni j 1 i j n = 1 n .

1i .    nm .n kn jAi A2 A3j . jA1    Anj = n X i=0 .

Let jS j denote the number of elements in a set S: Let B . Let A be a set with 8 elements. such that the intersection of any two of them is not a 2element set.    . Bn . i  i 167.n m:  n . Solution. Find the maximal number of 3-element subsets of A.

6g. a 1999 Hong Kong China Math Olympiad Students have taken a test paper in each of n n  3 subjects. 3. 7g: 1 1 1 1 5 2 6 3 7 4 8 168. just consider B = f1. : : :. we get k  3: From this. B = f1. 4. so n  8: To show 8 is possible. j = 1. 6. 125 . 5g. 2. then jBi Bj j = 1 for i. Then 3n  8  3. 3g. 5. B = f3. B = f8. 4. jBi Bj j 6= 2 for i. we see that every element of A is in at most 3 Bi 's. B = f1. Bk . B = f8. 7g. and for any two subjects excatly one student gets the best score in every one of these two subjects. k: Since 8 = jAj  jB    Bk j = 1 + 2k. B = f8. n: If a 2 A belongs to B . A be such that jBi j = 3. : : :. 4g. j = 1. : : :. It is known that for any subject exactly three students get the best score in the subject. Determine the smallest n so that the above conditions imply that exactly one student gets the best score in every one of the n subjects. 7g. 6g. 5. 2. B = f2.

Let x be one student who is best in most number of subjects. which implies n  7: Therefore. suppose x is in S1. 2. let Si be the set of students who get the best score in the i-th subject. a Due to Fan Wai Tong For i = 1. say m m n subjects. Suppose nobody gets the best score in every one of the n subjects. S2. jS1 S2    Snj X X = jSij . If every two clubs have exactly one common member. m. then prove that all 1978 clubs have a common member.b 1978 Austrian-Polish Math Competition There are 1978 clubs. 2. Solution. : : :. so m  3: This means each student is best in at most three subjects. let Si0 = Si n fxg: Then the m sets Si0 are pairwise disjoint and so each shares a distinct common member with Sm+1: Since Sm+1 has three members. Each has 40 members. n. : : : . then there is at least one . : : :. Sm: For i = 1. By the inclusion-exclusion principle. 2 + jS1 S2    Sn j. if n  8. we give an example of the case n = 7 with nobody best in all subjects:  3n . There is exactly one such students because only one student gets the best score in a pair of subjects. jSi Sj j + 1 X 1 in student who get the best score in every one of the n subjects. Finally. Without loss of generality.

the k lists of the k members of C1 are 126 . x4 . form a list of the indices of the other clubs that this member also belongs to. x4 . x7g: b Let n = 1978 and k = 40: Let C1. x3 g. S3 = fx2 . x4 . : : : . x6 . x2. For each member of C1. S5 = fx1 . x5 . x7g. x5. C2. S2 = fx1 . x6 g. x7g. x5g. Cn be the n clubs. S4 = fx3 . S6 = fx2 . Since C1 and any other club Ci have exactly one common member.n 1 i jn i j kn jSi Sj Sk j S1 = fx1 . S7 = fx3 . x6g.

In each move. will contain at least m = d n . show that after some moves. Comments. k + 1 = 1561 n = 1978. So x must be a member of all n clubs. each having k members and each pair having exactly one common member. respectively. 1995 Byelorussian Math Olympiad In the begining. the inclusion-exclusion principle in a will be less e ective. k + 1  n: Since k . 65 beetles are placed at di erent squares of a 9  9 square board. k Next we will show this x is a member of all n clubs. k numbers. Type A consists of squares with both a and b odd. The argument in part b is more convenient and shows that for n sets. a contradiction. k 2 2 k . If no beetle makes either two horizontal moves or two vertical moves in succession. there will be at least two beetles in the same square. As the number of members in the sets gets large. then all n sets have a common member. B and C are 25. Combinatorial Designs 169. b = 1. say x's list. beetles in type A squares will be in type B squares. k + 1. : : :. b to each square with a.disjoint and together contain all integers from 2 to n: By the pigeonhole 1e principle. type B consists of squares with both a and b even and type C consists of the remaining squares. every beetle creeps to a horizontal or vertical adjacent square. which implies Since Ci has k members. so k  m + 1  n. 16 and 40. 9: Divide the 81 squares into 3 types. It is clear that the two problems are essentially the same. The numbers of squares of the types A. 1 + 1. The notation means m is the least integer greater than or 1 : equal to n . Due to Cheung Pok Man and Yung Fai Assign an ordered pair a. 2. if n k2 . After two successive moves. this is a contradiction. So the number of beetles in 127 . Solution. Assume no collision occurs. one of the lists. Suppose x is not a member of some club Ci: Then each of the m + 1 clubs that x belong to will share a di erent member with Ci otherwise two of the m + 1 clubs will share a member y in Ci and also x.

1: For 1  i j  k . Show that we k intersection points of these lines by the numbers can label the C2 1. 1 when i + j  k. 36 to put on a ticket. k when 2i k: Alternatively. then the following labeling works: for 1  i j  k . l2. For the intersection of lines lk and li i = 1. lk the numbers 1. k . l2. The lottery commitee will then draw six distinct numbers randomly from 1. : : :.type A squares are at most 16 at any time. : : :. Due to Ng Ka Wing If such labeling exists for an integer . 2. 2. but 8 tickets is not enough to guarantee a winning ticket in general. beetles in type C squares will go to type A or type B squares. 3. 36: Any ticket with numbers not containing any of these six numbers is a winning ticket. Hence there are at most 64 beetles on the board.e. k is even. give the intersection of lines li and lj the label of that vertex. k when i + j k: For the intersection of lines lk and li i = 1. the label 2i . give the intersection of lines li and lj the label i + j . k . : : : . 1. 3. 1995 Greek Math Olympiad Lines l1. Hence there are k=2 1's. : : :. Conversely. : : :. 128 Solution. k . k . 1. after one move. 1 so that in each of the lines l1 . 2. k. 1996 Tournaments of the Towns In a lottery game. : : :. 170. 1 appear exactly once if and only if k is even. 2. : : :. the label i + j . give the label 2i . 2. we can make use of the symmetry of an odd number sided regular polygon to construct the labeling as follows: for k even. a person must select six distinct numbers from 1. give the label i: 171. k . if k is even. 1. 1 sided regular polygon with the vertices labeled 1. i. Then there are at most 32 beetles in type A or type B squares at any time. a contradiction. 2. : : :. Also. So there are at most 32 beetles in type C squares at any time. 2. consider the k . the perpendicular bisector of the segment joining vertices i and j passes through a unique vertex. 1 when 2i  k. lk are on a plane such that no two are parallel and no three are concurrent. Show that there is a scheme of buying 9 tickets guaranteeing at least a winning ticket. then the label 1 must occur once on each line and each point labeled 1 lies on exactly 2 lines. : : :. 1.

resulting in no winning tickets. 2. 12 and one number from each of the four remaining tickets may be the six numbers drawn by the committee. 13. they must make a 60 or 120 turn. 4. then two of the six sumbers must be 10. Then 1. 4. 11. 2. if they are not winning. 27. 7. say they are 1.  a a  6. 172. 25. 7. 9. 9. 9: For the next three tickets. The remaining 10 numbers on them will miss at least one of the numbers 2. We put coordinates at the vertices so that a. at least one of the nine tickets is a winning ticket. 20. 16. 36: Since only six numbers are drawn. When they get to a vertex. 2. 23. if they are not winning. 6. Consider the nine tickets with numbers 1. 0. for 0  b  . 14. 13. 31. 3. 19. if one number appears in three tickets. : : :. 8. 24. 2. b. 6: 129 . 10.Solution. 15. then three of the six numbers must be among 19. : : :. corresponds to the position of b in the Pascal triangle. resulting in no winning tickets. Prove that at some moment two beetles must meet at some vertex. 2. 2. 28. 36: For the rst three tickets. if they are not winning. 14. 18. 4. 2. First mark the vertices 0. : : :. 17. Now 12 appears in two other tickets. : : :. then this number and one number from each of the ve remaining tickets may be the six numbers drawn. 12: Consider the two tickets with 1 on them. 10. 6. 18. say 12. 2. Then the beetles move along di erent edges with the same speed. 3. the triangle can be divided into 36 smaller equilateral triangles. 11. So of the 48 numbers on the eight tickets. 16. : : :. Is the statement true if 6 is replaced by 5? Solution. 0. 34. 5. 18: For the last three tickets. 11. 5. 6. we may assume at least 12 appeared exactly 2 times. 2. 22. 1. 30. 33. 17. 15. 32. 6. For any eight tickets. 12. then two of the six numbers drawn must be among 1. 6. 4. 0. 29. 4. 4. 4. 8. 35. 20. 26. 6. 1995 Byelorussian Math Olympiad By dividing each side of an equilateral triangle into 6 equal parts. 12. 21. 12. 0. A beetle is placed on each vertex of these triangles at the same time. 3.

Covering. f4. 2. otherwise 4Pj Pk Pl has larger area than 4PiPj Pk : Similarly. 0. 1. If 6 is replaced by 5. Due to Lee Tak Wing Let the n points be P . Pn: Suppose 4Pi Pj Pk have the maximum area among all triangles with vertices from these n points. 3. 3. then the 10 beetles at the marked vertices will move to 10 unmarked vertices and 10 other beetles will move to the marked vertices. After another move. 3. 174. Solution. 5. 1g. No Pl can lie on the opposite side of the line through Pi parallel to Pj Pk as Pj Pk . f4. 3. then divide the vertices into groups as follows: f0. f2. 0. P . 1. 0. 4. 5. 0. 1991 Australian Math Olympiad There are n points given on a plane such that the area of the triangle formed by every 3 of them is at most 1. 5. 5. 1. Since the area of 4PiPj Pk is at most 1. Show that the n points lie on or inside some triangle of area at most 4. no Pl can lie on the opposite side 1 2 of the line through Pj parallel to Pi Pk as Pi Pk or on the opposite side of the line through Pk parallel to Pi Pj as Pi Pj : Therefore. 2. Since there are only 18 unmarked vertices. 3. two of them will meet. : : :. 0. each of the n points lie in the interior or on the boundary of the triangle having Pi . 2. f4. 3. 1969 Putnam Exam Show that any continuous curve of unit length can be covered by a closed rectangles of area 1=4: 130 . if no beetles meet. 5. 4.After one move. 2g. 5. these 20 beetles will be at unmarked vertices. 4. so the area of this triangle is at most 4. Then the beetles will not meet at any moment. 1g. Pk as midpoints of its sides. 5. 2g. Convex Hull 173. Pj . 3. 4g: Let the beetles in each group move in the counterclockwise direction along the vertices in the group. 1g. 1. 0. f2.

: : : . Let its horizontal and vertical dimensions be a and b. suppose x is not in the union of D . the centers is at most r + rj units apart. A2. where 3D is the disc with the same center as D but having three times the radius. : : : . : : : . 3Dn covers E. and real numbers r1 . 1995 IMO Determine all integers n 3 for which there exist n points A1. : : :. The polygonal line P0 P1 P2 P3 P4 P5 has length at most one. The horizontal projections of the segments of this polygonal line add up to at least a. Dj has been picked. The vertical projections of the segments of this polygonal line add up to at least 2b. Solution. : : : . the algorithm will stop at a nal disc Dn : For x in E. Since F is a nite collection. Let P1 . 3Dn : 176. Dn: Then x is in some disc D of radius r in F : Now D is not one of the Dj 's implies it intersects some disc Dj of radius rj  r: 1 1 +1 1 1 By the triangle inequality. Dn in F such that the union of 3D1 . Let P0 and P5 be its endpoints. : : :.Then take the smallest rectangle with sides parallel to the axes which covers the curve. Solution. P4 be the points on the curve. Place the curve so that its endpoints lies on the x-axis. Then we pick a disc Dj disjoint from each of D . in the order named. respectively. P3 . rn satisfying the following two conditions: 131 . An in the plane. r2. p So the polygonal line has length at least a2 + 4b2  1: By the AM-GM inequality. E is contained in the union of 3D1 . 1998 Putnam Exam Let F be a nite collection of open discs in the plane whose union covers a set E: Show that there is a pairwise disjoint subcollection D1 . 4ab  a2 + 4b2  1 and so the area is at most 1=4: 175. : : :. P2. Let D be a disc of largest radius in F : Suppose D . which lie one on each of the four sides of the rectangle. We construct such Di 's by the greedy algorithm. Then D is contained in 3Dj : In particular. x is in 3Dj : Therefore. since the endpoints are on the x-axis and the line also has points on the top and bottom side of the rectangle. : : :. since the line has points on the left and right sides of the rectangle. Dj and has the largest possible radius.

Quadrilateral Case. Triangular Case. Due to Ho Wing Yip For n = 4. So 2S = 3r1 + r2 + r3 + r4 : Also S = A1A2 A5 + A2A3 A5 + A3A4 A5 + A4 A1A5 = 2r1 + r2 r3 + r4  + r5 : From the last equation. : : :. A3 not collinear implies one side of 6 A1A5 A3 is less than 180 : Then one of the quadrilaterals A1 A5 A3A4 132 . A = 0. An lie on a line. r3. 0. with A4 . A = 1. 0. A5 inside such that A5 is outside 4A1A2A4 and A4 is outside 4A1A3 A5: Denote the area of 4XY Z by XY Z : We get a contradiction as follows: Solution.a no three of the points A1. r4 . A2. Next we will show there are no solutions for n  5: Suppose 1 2 3 4 1 2 3 4 A1 A4A5 + A1A2A3 = r1 + r4 + r5  + r1 + r2 + r3  = r1 + r2 + r4  + r1 + r3 + r5  = A1 A2A4 + A1A3 A5 A1 A2A3 : Pentagonal Case. We may assume A5 is inside the convex hull. A2 is on line L or on the half plane of L opposite A3 . First obseve that r1 + r3 = R2 + r4 : This is because r1 + r2 + r3  + r3 + r4 + r1  = r1 + r2 + r4 + r2 + r4 + r3 is the area S of the convex hull. There are three cases. we get r5 = .r1 + r2 + r3 + r4 =8 = . A3 are the vertices of the convex hull. note A = 0. A2. 1. we get 6 A2 A1 A5 180 contradicting convexity. consider the convex hull of A1 .S=12 0: Next observe that A1. r = r = r = r = 1=6 satisfy the conditions. 1. r2 . A5 . b for each triple i. A5: This is the smallest convex set containing the ve points. A2 . We may assume r1 = minfr1. k 1  i j k  n the triangle Ai Aj Ak has area equal to ri + rj + rk : the contrary. r5g: Draw line L through A1 parallel to A3A4 : Since A1A3 A4 = r1 + r3 + r4  r2 + r3 + r4 = A2A3 A4 . A = 1. A4. A3. We may assume the points are named so A1 . A4 and similarly for A5 : Since A1. A5 cannot all be on L. A2 . j.

Place it at the origin and the furthest point Q of S from P on the positive real axis. ! 1 +    + OXn OG = : n 133 .! . we also get r5 = r1 or r3 : Therefore. considering A2. : : :. Since the perpendicular bisector of PQ is an axis of symmetry. the mirror image of Pj will be a point in S further than Q from P. 1999 IMO Determine all nite sets S of at least three points in the plane which satisfy the following condition: for any two distinct points A and B in S. we see that 6 Pi. Since the origin P is in the interior of the convex polygon. So there exists a vertex Pj to the left of the y-axis. which is the smallest convex set containing S: Since S is nite. Xng is such a set. Let H be the convex hull of such a set.. A4 not collinear. . A5.1 Pi+1 : So Pi. n. Comments. such a set clearly has the required property. P2 . 2. r3. The o cial solution is shorter and goes as follows: Suppose S = fX1 . 177. Pn belonging to S: Let Pi = Pj if i j mod n: For i = 1. To see this. we get r5 = r2 or r4: Similarly. the perpendicular bisector of the line segment AB is an axis of symmetry of S: considering the perpendicular bisector of the two furthest points of S on that line. ! OX . there cannot be any point P of S inside the regular polygon. we will get a contradiction. assume such a P exists. r2 . the boundary of H is a polygon with the vertices P1 . a contradiction. not all the vertices can lie on or to the right of the y-axis. which is the point G such that Solution.1 Pi+2 . Clearly. three of the numbers r1. . r4 .1 Pi Pi+1 = 6 Pi Pi+1 Pi+2 : So the boundary of H is a regular polygon. : : :. no three points of such a set is collinear otherwise . : : : . Consider the barycenter of S. where ri = r4 or r2 : Since r1 + r3 = r2 + r4 . r5 are negative. r1 + r3 = r5 + ri .or A1 A5A3 A2 is convex.1 Pi = Pi Pi+1 : Considering the perpendicular bisector of Pi.. a contradiction. the condition on the set implies Pi is on the perpendicular bisector of Pi. By the rst observation of this case. but the area of the corresponding triangle is positive. Next. Conversely. So S is the set of vertices of some regular polygon.

O i! . . . To see this. after re ection with respect to the perpendicular bisector of every segment XiXj .. . which implies G is on every such perpendicular bisector. n: 0G 0 is the average of O 0X using another origin O0. ! 0 0 0 : Adding O 0G 0 Subtracting the two averages. . : : :. Hence G is equidistant from all Xi's. . ! = OG0 . ! for i = 1.e. O G = OO . the Xi's are concyclic. ! . i. . we get OG . So G is unchanged. suppose we get a point G . considering the perpendicular bisector of segment XiXk .. Xj .. the points of S are permuted only. 134 . ! . . say Xi . . . Xk . For three consecutive points of S.. ! . so G = G0 : to both sides. ! . we get OG By the condition on S. we have XiXj = Xj Xk : It follows that the points of S are the vertices of a regular polygon and the converse is clear.The case n = 2 yields the midpoint of segment X1 X2 and the case n = 3 yields the centroid of triangle X1 X2X3: Note the barycenter 0 does not depend on the origin. Therefore. on the circle.

player II could prevent player I from winning by choosing the other square in the tile.1 .1 . after player I chose a square. Between each ride. player II chooses another square and marks it with X. his last ride will repeat the rst horse. and tan. n + 1: 2 179. Show that player II can prevent player I from winning. A boy takes n rides. 180. 3. show that pressing some nite sequence of buttons will yield q: Assume that the calculator does real number calculations with in nite precision. n . the bisecting segment in one will be horizontal and the other vertical. Player I chooses a square and marks it with an O. this is possible by moving forward 1. 2. 4. n . 1 places corresponding to horses 1. The display initially shows 0: Given any positive rational number q.1 buttons. 135 . Due to Chao Khek Lun Divide the board into 2  2 blocks. 3. Then. All functions are in terms of radians. : : : . n. cos. sin. n . Since nn . he moves clockwise a certain number less than n of places to a new horse. cos. the game is a tie. If n is even. 1 = nn . Find all n for which the boy ends up riding each horse. If n is odd. 1=2 is divisible by n. Then bisect each 2  2 block into two 1  2 tiles so that for every pair of blocks sharing a common edge. Since every ve consecutive squares on the board contains a tile. and this player wins the game. the boy's travel 1+2+    +n . Solution. 1995 Israeli Math Olympiad Two players play a game on an in nite board that consists of 1  1 squares. 1995 USAMO A calculator is broken so that the only keys that still work are the sin. : : :. 1995 Russian Math Olympiad There are n seats at a merry-go-around. If no player can achieve this. They play until one of the players marks a row or a column of 5 consecutive squares. Each time he moves a di erent number of places. 1=2 places between the rst and the last rides. Solution.Solutions to Miscellaneous Problems 178. 1. 2. tan. n .

n are positive integers. these include all positive rationals. where p m. can be displayed by induction on k = m + n: p 2 Since r=s = r =s2 . We will show that all numbers of the form m=n. For k = 2. Suppose the statement is true for integer less than k: Observe that if x is displayed. pressing cos will display 1. then using the facts .Solution.

1 x implies cos.1 sin . = tan.

. = =2 .

and tan =2 . .

However. m=mp can be displayed. k acquaintances in school Y: Adding the number of acquaintances of x and z in school Y. we can display m=n: This completes the induction. in the third school Z: Now z has at most k acquaintances in school X and hence z has at least n +1 . x must have at least one acquaintance. we may assume m n: By the inducp tion step. Prove that one can pick three students. m=m and cos = m=n. y. k = n +1 n and so x and z must have a common acquaintance y in school Y: 182. we get k +n +1 . Assign coordinate x. So such a packing is impossible. y. z = 0. 1977 Eotvos-Kursch ak Math Competition Each of three schools is attended by exactly n students. Call this student x. n k implies n . Due to Chan Kin Hang Consider a student who has the is possible. his school X and the k acquaintances in school Y: Since n +1 n  k. a contradiction. a direct counting shows there are 251 cells of color 0. Is there a way to pack 250 1  1  4 bricks into a 10  10  10 box? Solution. If the packing . we can display 1=x: Therefore. 1 = tan n . to display pm=n with k = m + n. one from each school. 1. = 1=x: So. 136 Solution. then there are exactly 250 cells of each color. Then using p p . 181. z  to each of the cells. : : :. highest number. z  be given color x + y + z mod 4: Note each 1  1  4 brick contains all 4 colors exactly once. where x. 9: Let the cell x. say z. who know one another. Each student has exactly n + 1 acquaintances in the other two schools. It is assumed that acquaintance is mutual. say k. y. of acquaintances in another school.

1: So 1 2 3 +1 . Is it possible to write a positive integer into each square of the rst quadrant such that each column and each row contains every positive integer exactly once? Solution. De ne A = 1 and An 1 where the entries of Bn are those of An plus 2n. it is possible.183. Yes.

n n 04 3 2 11 .B A = An Bn .

There are n identical cars on a circular track. So. Among all of them. then the gas for all cars would not be enough for completing a lap. then every column and every row of Bn will contain 2n. So there is a car that can complete a lap. the case n = k is true. 2. A = 1 2 . For the case n = k + 1.1 +1. Let us empty the gas of B into A and remove B: Then the k cars left satisfy the condition. This same car will also be able to complete the lap collecting gas from other cars when B is included because when this car gets to car A.1 exactly once. Due to Chan Kin Hang The case n = 1 is clear. 184. : : :. 2n. 1996 Russian Math Olympiad At the vertices of a cube are written eight pairwise distinct natural numbers. Now ll the rst quadrant using the An 's. 2. rst observe that there is a car A which can reach the next car B: If no car can reach the next car. :::: 1 2 3 4 Note that if every column and every row of An contain 1. 2n exactly once. A = B @3 2 1 4 3A. : : :. and on each of its edges is written the greatest common divisor of the numbers at the endpoints 137 Solution. every column and every row of An+1 will contain 1. 2n exactly once. Suppose .2 1 4 1 2C A 1. they have just enough gas for one car to complete a lap. the gas collected from car A will be enough to get it to car B: 185. : : : . Show that there is a car which can complete a lap by collecting gas from the other cars on its way around the track in the clockwise direction.

Consider this table and its 90 rotation. pn of 1. 1991 German Mathematical Olympiad Show that for every positive integer n  2. : : :. 12345 = 2040 + 10305 2 2040 + B: 187. Can the sum of the numbers written at the vertices be the same as the sum of the numbers written at the edges? of the edge numbers. Then every 3  3 square has sum 36. For each cell. Let A be the set of positive integers whose odd digit positions from the right are zeros. n . 0. a contradiction. Let B be the set of positive integers whose even digit positions from the right are zeros. Observe that if a b. : : :. 7. 8. there exists a permutation p1 . 5. 4. At every vertex. 1. 8 on the rst. b  b and gcda. 1: 138 Solution. 0. Can the positive integers be partitioned into in nitely many subsets such that each subset is obtained from any other subset by adding the same integer to each element of the other subset? Solution. then gcda. 1. fth and eigth rows. 2. b = a + b=3 for every pair of adjacent vertex numbers. then we will have gcda. 2. 6. 6. 3. n such that pk+1 divides p1 + p2 +    + pk for k = 1. b  a + b: If the sum of the vertex numbers equals the sum Solution. 5 on the third. 4. : : :. fourth and seventh . Place 0. there are 3 adjacent vertices. 1995 Russian Math Olympiad Is it possible to ll in the cells of a 9  9 table with positive integers ranging from 1 to 81 in such a way that the sum of the elements of every 3  3 square is the same? rows. 2 on the second. 2. 2. 186. 7. 7. where a is the number in the cell originally and b is the number in the cell after the table is rotated by 90 : By inspection. which implies a = 2b or b = 2a at the two ends of every edge. ll it with the number 9a + b + 1. 8. b  a=2: So 3gcda. Place 3. Place 6. sixth and ninth rows. p2 . 4. 5. The a = 2b or b = 2a condition implies two of these adjacent vertex numbers must be the same.of the edge. 1 to 81 appears exactly once each and every 3  3 square has sum 9  36 + 36 + 9 = 369: 188. Then A and B are in nite set and the set of positive integers is the union of a + B = fa + b : b 2 B g as a ranges over the elements of A: For example. 1. Yes. 3.

we have gi . m.Solution. m + 2. bj  gk . However. then a + b + c + d=4 = m and a. 1984 Tournament of the Towns In a party. m: For odd n = 2m + 1. if one end is labelled m. gk j 10 for all k: 1 2 1 2 139 . bk  10. b. gi . If the four neighboring numbers are a. 5 suggest the following permutations. below. 2. consider the permutation m + 1. Show that all the numbers are equal. c. 3. d. : : : . c. consider the permutation m + 1. where k  j  n: By the pigeonhole principle. : : :. gk  10 is handled similarly. Each lattice point of the plane is labeled by a positive integer. right. 1. Show that the di erence in height of the k-th tallest boy and the k-th tallest girl is also less than 10 cm for k = 1. m + m. 2m + 1: If k = 2j . the di erence in height is less than 10 cm. n: Solution. where 1  i  k. Consider the smallest number m labelled at a lattice point. every number equals m: 190. d  m imply a = b = c = d = m: Since any two lattice points can be connected by horizontal and vertical segments. then along this path all numbers will equal to m: Therefore. The case bk . Due to Andy Liu. 1  j  m then m +1+1+    +m + j  = j m + j : If k = 2j. For even n = 2m. bk  10 for 1  i  k and k  j  n: Consider the girls of height gi . : : :. for that pair. n boys and n girls are paired. 1. 1  j  m then m +1+1+    +m + j + j = j m + j +1: 189. m + 2. The cases n = 2. So jbk . jbk . b. Each of these numbers is the arithmetic mean of its four neighbors above. Solution. left. Suppose for some k. 1. 2. 4. It is observed that in each pair. m + m. 2. gk j  10: In the case gk . Canada Let b  b      bn be the heights of the boys and g  g      gn be those of the girls. University of Alberta. and the boys of height bj . two of these n + 1 people must be paired originally. bj  10 contradicts the hypothesis.

By the pigeonhole principle. d. 27    4 = 108. so jbc . 2delete the zero and multiply by 9 to get a number 8200    08. d . we do the following steps: 1 multiply by 82 to get a number 911    102. 3 delete the zeros to get 828. 4 now 828  25 = 20700. 2z : From that point on. then after 4 steps. 11. jac . d = z. a. 140 . d such that the numbers jbc . : : : . b. b. 2x + 3y . 2w + 3y . a. b. 1996 IMO shortlisted problem Four integers are marked on a circle. bdj. we get a number of the form 11    100    0. On each step we simultaneously replace each number by the di erence between this number and next number on the circle in a given direction that is. jab . c . d will always be even. cdj will always be divisible by 4. jac . c = 2y . 192. 2w. bdj. Solution. jab . c = y. adj. b. Then perform operation b to delete the zeros if any. d are replaced by a . 2y + 3z . which is a multiple of X: Perform operation a on X to get such a multiple. Due to Ng Ka Man and Ng Ka Wing If the initial numbers a = 2w . c. b .191. 111 | z  1 have the same remainder when divided by X: So taking the di erence of two of these numbers. the numbers a. we get the single digit 9. the numbers will be X +1 digits Solution. c. 18  5 = 90 and delete zero. 1991 Leningrad Math Olympiad One may perform the following two operations on a positive integer: a multiply it by any positive integer and b delete zeros in its decimal representation. adj. c. 2x. at least two of the X + 1 numbers 1. 111. one can perform a sequence of these operations that will transform X to a one-digit number. 2z + 3w . 2z . b = 2x . b = x. If the new number has more than one digits. d = 2z . c. cdj are primes? are a = w. Prove that for every positive integer X. Is it possible after 1996 such steps to have numbers a.

we have n = 2k + 1 coins. Some are placed with the head sides up and some the tail sides up. For i = 1. 1989 Nanchang City Math Competition There are 1989 coins on a table. Mark these two coins. First. Suppose the case k is true. Then each coin is ipped k + 1 times and at the end all coins will be the same side up. Due to Chan Kin Hang 1 The number 1989 may not . in particular. be special. 2 in the above procedure. A group of 1989 persons will perform the following operations: the rst person is allowed turn over any one coin. : : :. then in some initial placements of the coins. 2 are true. suppose all coins are the same side up initially. 2. the second person is allowed turn over any two coins. For the case k + 1. whether the head or the tail sides turned up at the end will depend on the initial placement of the coins. The 2k-th person ips the 2k coins the same side up and the 2k + 1-st person ips all coins and this subcase is solved. i coins. let the i-th person ip any i coins and let the 2k +1 . the 1989 persons can come up with a procedure turning all coins the same sides up at the end of the operations. Let the rst 2k . Then there is one coin head up and another tail up.193. So we suspect the statement is true for odd n and do induction on k. the coins may be ipped with 141 Solution. k. the 1989th person is allowed to turn over every coin. 1 coins the same side up by the case k: Then there are exactly 2k coins the same side up and one coin opposite side up. the k-th person is allowed turn over any k coins. case n = 1989 is true. 2 If the procedure does not depend on the initial placement. but the case n = 2 is false when both coins are heads up initially. 1: The cases k = 1. suppose not all coins are the same sides up initially. So the k + 1 case is true in either way and the induction step is complete. Prove that 1 no matter which sides of the coins are up initially. where n = 2k . : : : . Next. So let us replace it by a variable n: The cases n = 1 and 3 are true. 1 persons ip the other 2k . i-th person ips the remaining 2k + 1 . : : : .

: : :. 1989. 3. for each coin to go from head up to tail up. However. Reversing the ippings on the heads up case. 194. 8.1 x1992 + 2 tan f r = 2 tan. a contradiction. 6. 1992 in some order. 7.all heads up and may also be ipped with all tails up. b the polygon is circumscribable about a circle. there exists r such that f r = 2: For such r. Solution. each must be ipped an odd number of times and the 1989 coins must total to an odd number of ippings. 2. 1 if n 1. de ne 8 n . : : : . we can then go from all coins heads up to all tails up in 21 + 2 +    + 1989 ippings. : : :. 2. 3. 1991: Consider a circle centered at O with large radius r and wind a polygonal line A1A2    A1992A1993 with length Ai Ai+1 = ai around the circle so that the segments Ai Ai+1 are tangent to the circle at some point Pi p 2 with Ai Pi = xi and Pi Ai+1 = xi+1: Then OA1 = x2 1 + r = OA1993: De ne 1 . For n = 1. 4. 142 . 3 mod 4 xn = : 3=2 if n 2 mod 4 1=2 if n 0 mod 4 and an = xn + xn+1 with x1993 = x1 : The sequence an is 1. Proposed by India for 1992 IMO Show that there exists a convex polygon of 1992 sides satisfying the following conditions: a its sides are 1. 2.1 x r r r = 6 A1 OP1 + 6 P1992 OA1993 + 6 P1 OP2 + 6 P2 OP3 +    + 6 P1991OP1992: Now f is continuous. 1992. 1992. 1990. A1993 will coincide with A1 . 5.1 x2 +    + 2 tan. resulting in the desired polygon. r! lim f r = 1992 and r! lim f r = 0: By the +1 0+ intermediate value theorem.

4. we have 45. 196. c + 2: Observe that in all 3 cases. b + 2. By repeating these operations. 1. 1? Solution. x . c. 2 2 2 2 2 2 2 2  = 5: So the answer is no. 4ca = b . The key fact that makes the polygon exists is that there is a permutation a1. 1 . So the answer is no. a1992 of 1. So from a. 1 a . c red chips. 1 or a . we can get a +2. then at the end.  = b . b mod 3 is an invariant. 2. can you transform x2 . 1 . b. b = 13 . b . 45: So a . 4ac: After operation b. 0 or 0. a . Consider the discriminant  = b . : : :. x2 + x3 = a2. In one step. A student may erase any two of the numbers a and b on the board and write the 143 . 15 black. in one step. b . 4ac: After operation a. However. 0. you may choose 2 chips of di erent colors and replace each one by a chip of the third color. 1. There are 13 white. 1. 2 2 2 197. b . x . 3. the di erence a + 2 . For x . Can all chips become the same color after some steps? Solution. a2.Comments. 4aat + bt + c = b . b . c . b black. The following operations are permitted with the quadratic polynomial ax2 + bx + c: a switch a and c. : : :. a . 0.  = 9: For x . 45. a . 1. 4ac: So  is an invariant. 2. ax + t + bx + t + c = ax + 2at + bx + at + bt + c and  = 2at + b . 1. : : : . 1992 such that the system of equations x1 + x2 = a1. Five numbers 1. 0 or 0. If all chips become the same color. b replace x by x + t. Write a. 2 into x2 . c . 17 red chips on a table. x . b. 1 or a . 5 are written on a blackboard. 195. b 0 mod 3 at the end. x1992 + x1 = a1992 have positive real solutions. c for a white. 1. b +2. 15 6 0 mod 3 in the beginning. b mod 3: So a . x . 2. where t is a real number.

the fth number must be 2 mod 3: Since 64 1 mod 3. Since in the beginning. If nine is replaced by ten. Observe that the perimeter of the black region cannot increase. 4. If a cell has four. then it is possible as the ten diagonal cells when infected can spread to the whole square. On each turn. 64. 3. b are multiples of 3. all of the coins are turned heads up. 64. when that cell is colored black. the cells with at least 2 infected neighbors having a common side become infected. Can the infection spread to the whole square? What if nine is replaced by ten? record the perimeter of the black region at every unit time. ab will also be multiples of 3. then it will not be infected. If a cell has one or no infected neighbors. then a + b 0 mod 3 and ab 2 mod 3: Now note there is one multiple of 3 in f1. 5g and four multiples of 3 in f21. three. namely when one of a or b is 1 mod 3 and the other is 2 mod 3. the goal is to 144 Solution. 0. If one of them is a multiple of 3. The number of multiples of 3 can increase in only one way. 180. the infection cannot spread to the whole square. 198.numbers a + b and ab replacing them. 180. 2. 540g: So when the number of multiples of 3 increases to four. Observe that the number of multiples of 3 among the ve Solution. Initially. 199. two infected neighbors. the perimeter of the black region is at most 9  4 = 36. Due to Cheung Pok Man Color the infected cells black and . we may turn over three coins which are mutually adjacent. If a. respectively. then ab will also be a multiple of 3. In one unit time. 2. If this operation is performed repeatedly. 27. can the numbers 21. then it will become infected and the perimeter will decrease by 4. 27. Nine 1  1 cells of a 10  10 square are infected. and a 10  10 black region has perimeter 40. 27. 540 can never appear on the board. so 21. then a + b. 64. 180. 1997 Colombian Math Olympiad We play the following game with an equilateral triangle of nn + 1=2 dollar coins with n coins on each side. 540 ever appear on the board? numbers on the blackboard cannot decrease after each operation.

As we ip the triangles. d having integers endpoints and same lengths. Since the interior coins have side length n . 200. white and blue so that adjacent coins have di erent colors and any three coins in a row have di erent colors. 2. For n = 2. all of them can be ipped so to have tails up. say red. We will show that x. So the parities of the red and white coins stay di erent. 100 and let f x be the color number of x: Since all 100 colors were used. we will mean three coins which are mutually adjacent. which implies . y have the same color if and only if x y mod 100.1990 and 1990 have di erent colors. a: Prove that . by the induction step. Next suppose n 1 mod 3: Color the heads of each coin red. So all these coins will have tails up. 1002: 145 Solution. : : :. which means a + x and c + x have the same color for x = 0.make all of the coins turned tails up. For intervals a. ip every triangle. The coins on the sides not corners are ipped three times each. : : :. : : :. 100: Let M = minm1 . 2. This can be done only for all n 0. Then the coins at the corners are ipped once. if a. So this cannot happen. Let the colors be 1.1990 and 1990 have di erent colors. d have the same color. b . c. b . So the odd or even parities of the red and white coins are di erent in the beginning. 3. there is an integer mi such that f mi  = i for i = 1. m2 . . at each turn. Then the coins in the corner have the same color. c have the same color and b. In the case all coins are tails up. the number of red and white coins would be zero and the parities would be the same. m100  . then the intervals are colored the same way. 1. For n 0. The interior coins are ipped six times each and have heads up. either a both red and white coins increase by 1 or b both decrease by 1 or c one increases by 1 and the other decreases by 1. : : :. For which values of n can this be done? a triangle. clearly it can be done and for n = 3. A simple count shows that there are one more red coins than white or blue coins. 2 mod 3 and n 3. 1990 Chinese Team Selection Test Every integer is colored with one of 100 colors and all 100 colors are used. ip each of the four triangles. 2 mod 3: Below by Solution.

there is an interval a + i1. f x = f x + c . a + i + n i = 0. f must have period 100 on the set of integers. Finally. d and n is arbitrary. say f b = f c with a +1002  b c  a + 1002 + 100: For every x  a + 1002 + 100. Since all 100 colors are used. the union of the intervals a + i1. f a +1002 +100 are the same. In particular.Consider a xed integer a  M and an arbitrary positive integer n: Since there are 1002 ways of coloring a pair of integers. say n1 . a + i1 + nk . at least two of the pairs a + i. which means f a + i1  = f a + i2  and f a + i1 + n = f a + i2 + n for some integers i1. In particular. 1002 are colored the same way. f x = f x + d: So f x has period d when x  a + 1002 : Since a  M. a + i1 + d + nk are colored the same way. b  100 when x  a + 1002 + 100: So the least period of f x for x  a + 1002 + 100 must be 100. f a +1002 + 1. intervals b. c + md are colored the same way. so intervals a + i1 . b + md : Since f b + md = f b = f c = f c + md. b: So f x has period c . a + i1 + d. a + i1 + nk containing x: Since f a + i1 = f a + i1 + d and f a + i1 + nk  = f a + i1 + d + nk . b + md . 146 . since a can be as close to . 2. : : : . having the same i1 's and d's. choose a large integer m so that x is in b. a + i1 + nk will contain every integer x  a + 1002: So for every such x. n2. : : :. : : :. i2 such that 0  i1 i2  1002: Let d = i2 . there are in nitely many n's. Since these nk 's may be arbitrarily large. by the pigeonhole principle.1 as we like. two of f a +1002 . 100 colors are used for the integers x  a + 1002 and so d  100: Consider the least possible such period d: Next. no two of 100 consecutive intgers can have the same color. 1. c. i1: Since there are nitely many combinations of ordered pairs i1.

Master your semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master your semester with Scribd & The New York Times

Cancel anytime.