Professional Documents
Culture Documents
Imo 2002
Imo 2002
Olympiad Corner
Mathematical Games (I)
The 43rd International Mathematical Kin Y. Li
Olympiad 2002.
Problem 1. Let n be a positive integer. An invariant is a quantity that does not black pieces in the row or in the column
Let T be the set of points (x, y) in the change. A monovariant is a quantity that will become 8 – k, a change of (8 – k) –
plane where x and y are non-negative
keeps on increasing or keeps on k = 8 – 2 k black pieces on the board.
integers and x + y < n. Each point of T is
decreasing. In some mathematical Since 8 – 2 k is even, the parity of the
colored red or blue. If a point (x, y) is
red, then so are all points (x′, y′ ) of T games, winning often comes from number of black pieces stay the same
with both x′ ≤ x and y′ ≤ y. Define an understanding the invariants or the before and after the move. Since at the
X-set to be a set of n blue points having monovariants that are controlling the start, there are 32 black pieces, there
distinct x-coordinates, and a Y-set to be a games. cannot be 1 black piece left at any time.
set of n blue points having distinct
y-coordinates. Prove that the number of Example 1. (1974 Kiev Math Olympiad) Example 3. Four x’s and five o’s are
X-sets is equal to the number of Y-sets. written around the circle in an arbitrary
Numbers 1, 2, 3, l , 1974 are written on
a board. You are allowed to replace any order. If two consecutive symbols are the
Problem 2. Let BC be a diameter of the same, then insert a new x between them.
circle Γ with center O. Let A be a point two of these numbers by one number,
which is either the sum or the difference Otherwise insert a new o between them.
on Γ such that 0° < ∠AOB < 120°. Let D
of these numbers. Show that after 1973 Remove the old x’s and o’s. Keep on
be the midpoint of the arc AB not
containing C. The line through O parallel times performing this operation, the only repeating this operation. Is it possible to
to DA meets the line AC at J. The number left on the board cannot be 0. get nine o’s?
perpendicular bisector of OA meets Γ at
E and at F. Prove that J is the incentre of Solution. If we let x = 1 and o = – 1,
Solution. There are 987 odd numbers
the triangle CEF. then note that consecutive symbols are
on the board in the beginning. Every
replaced by their product. If we consider
Problem 3. Find all pairs of integers time the operation is performed, the
the product P of the nine values before
such that there exist infinitely many number of odd numbers left either stay
and after each operation, we will see that
positive integers a for which the same (when the numbers taken out
the new P is the square of the old P.
(continued on page 4) are not both odd) or decreases by two
Hence, P will always equal 1 after an
(when the numbers taken out are both
operation. So nine o’s yielding P = – 1
odd). So the number of odd numbers left
can never happen.
Editors: (CHEUNG Pak-Hong), Munsang College, HK on the board after each operation is
(KO Tsz-Mei)
always odd. Therefore, when one
(LEUNG Tat-Wing) Example 4. There are three piles of
(LI Kin-Yin), Dept. of Math., HKUST number is left, it must be odd and so it
stones numbering 19, 8 and 9,
(NG Keng-Po Roger), ITC, HKPU cannot be 0.
respectively. You are allowed to choose
Artist: (YEUNG Sau-Ying Camille), MFA, CU
two piles and transfer one stone from
Acknowledgment: Thanks to Elina Chiu, Math. Dept., HKUST Example 2. In an 8 × 8 board, there are each of these two piles to the third piles.
for general assistance.
32 white pieces and 32 black pieces, one After several of these operations, is it
On-line: http://www.math.ust.hk/mathematical_excalibur/
piece in each square. If a player can possible that each of the three piles has
The editors welcome contributions from all teachers and
students. With your submission, please include your name,
change all the white pieces to black and 12 stones?
address, school, email, telephone and fax numbers (if available). all the black pieces to white in any row
Electronic submissions, especially in MS Word, are encouraged.
or column in a single move, then is it Solution. No. Let the number of stones
The deadline for receiving material for the next issue is
November 2, 2002. possible that after finitely many moves, in the three piles be a, b and c,
For individual subscription for the next five issues for the 01-02 there will be exactly one black piece left respectively. Consider (mod 3) of these
academic year, send us five stamped self-addressed envelopes.
on the board? numbers. In the beginning, they are 1, 2,
Send all correspondence to:
stones are impossible. 1988. Both won a gold medal for the
IMO 2002 first time this year and both gold
Example 5. Two boys play the Kin Y. Li medallists scored 29 points.
following game with two piles of
candies. In the first pile, there are 12 The International Mathematical The IMO will be hosted by Japan
candies and in the second pile, there are Olympiad 2002 was held in Glasgow, next year at Keio University in Tokyo
13 candies. Each boy takes turn to United Kingdom from July 19 to 30. and the participants will stay in the
make a move consisting of eating two There were a total of 479 students from 84 Olympic village. Then Greece, Mexico,
candies from one of the piles or
countries and regions participated in the Slovenia will host in the following
transferring a candy from the first pile
Olympiad. years.
to the second. The boy who cannot
make a move loses. Show that the boy
who played second cannot lose. Can he The Hong Kong team members were Addendum. After the IMO, the
win? German leader Professor Gronau sent
Chao Khek Lun (St. Paul’s College) an email to inform all leaders about his
Solution. Consider S to be the number Chau Suk Ling (Queen Elizabeth School) updated webpage
of candies in the second pile minus the Cheng Kei Tsi (La Salle College) http://www.Mathematik-Olympiaden.de/
first. Initially, S = 13 – 12 = 1. After Ip Chi Ho (St. Joseph’s College) which contains IMO news and facts.
each move, S increases or decreases by Leung Wai Ying (Queen Elizabeth School) Clicking Internationale Olympiaden
2. So S (mod 4) has the pattern 1, 3, 1, Yu Hok Pun (SKH Bishop Baker on the left, then on that page, scrolling
3, l . Every time after the boy who Secondary School). down and clicking Top-Mathematikern,
played first made a move, S (mod 4) Die erfolgreichsten IMO-Teilnehmer in
would always be 3. Now a boy loses if blue on the right, we could find the
The team leader was K. Y. Li and the
and only if there are no candies left in
deputy leaders were Chiang Kin Nam and following past IMO participants who
the second pile, then S = 1 – 0 = 1. So
Luk Mee Lin. have also won the Fields medals, the
the boy who played second can always
make a move, hence he cannot lose. Nevanlinna prizes and the Wolf prizes:
The scores this year ranged from 0 to
Since either the total number of 42. The cutoffs for medals were 29 points Richard Borcherds (1977 IMO silver,
candies decreases or the number of for gold, 24 points for silver and 14 points 1978 IMO gold, 1998 Fields medal)
candies in the first pile decreases, so for bronze. The Hong Kong team received
eventually the game must stop, so the 1 gold medal (Yu Hok Pun), 2 silver Vladmir Drinfeld (1969 IMO gold,
boy who played second must win. medals (Leung Wai Ying and Cheng Kei 1990 Fields medal)
Tsi) and 2 bronze medals (Chao Khek Lun
Example 6. Each member of a club has and Ip Chi Ho). There were 3 perfect Tim Gowers (1981 IMO gold, 1998
at most three enemies in the club. (Here scores, two from China and one from Fields medal)
enemies are mutual.) Show that the Russia. After the 3 perfect scores, the
members can be divided into two Laurent Lafforgue (1984 IMO silver,
scores dropped to 36 with 9 students! This
groups so that each member in each
was due to the tough marking schemes, 1985 IMO silver, 2002 Fields medal)
group has at most one enemy in the
which intended to polarize the students’
group.
performance to specially distinguish those Gregori Margulis (1959 IMO member,
Solution. In the beginning, randomly who had close to complete solutions from 1962 IMO silver, 1978 Fields medal)
divide the members into two groups. those who should only deserve partial
Consider the number S of the pairs of points. Jean-Christoph Yoccoz (1974 IMO
enemies in the same group. If a gold, 1994 Fields medal)
member has at least two enemies in the The top five teams are China (212),
same group, then the member has at Russia (204), USA (171), Bulgaria (167) Alexander Razborov (1979 IMO gold,
most one enemy in the other group. and Vietnam (166). Hong Kong came in 1990 Nevanlinna prize)
Transferring the member to the other 24th (120), ahead of Australia, United
group, we will decrease S by at least Kingdom, Singapore, New Zealand, but Peter Shor (1977 IMO silver, 1998
one. Since S is a nonnegative integer, it behind Canada, France and Thailand this Nevanlinna prize)
cannot be decreased forever. So after
year.
finitely many transfers, each member
can have at most one enemy in the László Lovász (1963 IMO silver, 1964
same group. One piece of interesting coincidence
deserved to be pointed out. Both Hong IMO gold, 1965 IMO gold, 1966 IMO
Kong and New Zealand joined the IMO in gold, 1999 Wolf prize)
(Continued on page 4)
Mathematical Excalibur,
Excalibur Vol. 7, No. 3, Jul 02- Oct 02 Page 3
Problem 1. Let A be a subset of the set S = 已知一個複數 c0,並由此定義一 170 IF J >= 6 AND J < 9 THEN
{1, 2, …, 1000000} containing exactly 個複數數列 { cn },使 cn+1 = cn2 + c0, COLOR 1 : REM BLUE
101 elements. Prove that there exist 其中 n = 0, 1, 2, ……。如果這個數列 180 IF J >= 9 AND J < 12 THEN
numbers t1, t2, … , t100 such that the sets 有界,即可以找到一個正實數 M,使 COLOR 2 : REM GREEN
190 IF J >= 12 AND J < 15 THEN
對於一切的 n,| cn | < M,那麼 c0 便屬
aj = {x + tj | x∊A } for j = 1, 2, … , 100 COLOR 15 : REM WHITE
於曼德勃羅集(Mandelbrot Set)之內。 200 IF J >= 15 THEN
are pairwise disjoint. COLOR 12 : REM RED
210 PSET (X + LEFT, (TOP – Y)* M)
Problem 2. Determine all pairs of 220 NEXT X
positive integers (a, b) such that 230 NEXT Y
240 COLOR 15 : REM WHITE
a2 250 LINE (LEFT, (TOP – W / 2) * M)
2 ab − b 3 + 1
2
– (W + LEFT, (TOP – W / 2) * M)
260 LINE (W / 2 + LEFT, (TOP – W)
is a positive integer. * M) – (W / 2 + LEFT, TOP * M)
270 END
Problem 3. A convex hexagon is given
in which any two opposite sides have the 以下是這程序的解釋:
following property: the distance between
their midpoints is 3 / 2 times the sum 可以將以上定義寫成一個 W 紀錄在電腦畫面上將要畫出
of their lengths. Prove that all the angles GW-BASIC 程序(對不起!我本人始 圖形的大小。現將 W 設定為 360(見
of the hexagon are equal. (A convex 終都是喜歡最簡單的電腦語言,而且 第 10 行),表示打算在電腦畫面上一
hexagon ABCDEF has three pairs of 我認為將 GW-BASIC 程序翻譯成其他 個 360 × 360 的方格內畫出曼德勃羅集
opposite sides: AB and DE, BC and EF, 電腦語言亦不難),方法如下: (見第 50 及 60 行)。
CD and FA.)
10 LEFT = 150 : TOP = 380 : LEFT 是繪圖時左邊的起點,TOP
(continued on page 4) W = 360 : M = .833 是圖的最低的起點(見第 210、250 及
Editors: 張 百 康 (CHEUNG Pak-Hong), Munsang College, HK 20 R = 2.64 : S = 2 * R / W 260 行)。注意:在 GW-BASIC 中,畫
高 子 眉 (KO Tsz-Mei) 30 RECEN = 0 : IMCEN = 0 面坐標是由上至下排列的,並非像一
梁 達 榮 (LEUNG Tat-Wing) 40 SCREEN 9 : CLS
般的理解,將坐標由下至上排,因此
李 健 賢 (LI Kin-Yin), Dept. of Math., HKUST 50 FOR Y = 0 TO W
吳 鏡 波 (NG Keng-Po Roger), ITC, HKPU 60 FOR X = 0 TO W 要以 “TOP − Y” 的方法將常用的坐標
Artist: 楊 秀 英 (YEUNG Sau-Ying Camille), MFA, CU 70 REC = S * (X – W / 2) + RECEN : 轉換成電腦的坐標。
Acknowledgment: Thanks to Elina Chiu, Math. Dept., HKUST IMC = S * (Y – W / 2) + IMCEN
for general assistance. 80 RE = REC : IM = IMC 由於電腦畫面上的一點並非正
On-line: http://www.math.ust.hk/mathematical_excalibur/ 90 RE2 = RE * RE : IM2 = IM * IM : 方形,橫向和縱向的大小並不一樣,
The editors welcome contributions from all teachers and J=0 故引入 M(= 5
)來調節長闊比(見
100 WHILE RE2 + IM2 <= 256 AND 6
students. With your submission, please include your name,
address, school, email, telephone and fax numbers (if available). J < 15 第 10、210、250 及 260 行)。
Electronic submissions, especially in MS Word, are encouraged.
The deadline for receiving material for the next issue is
110 IM = 2 * RE * IM + IMC
120 RE = RE2 – IM2 + REC 留意 W 祇是「畫面上」的大小,
November 30, 2003.
For individual subscription for the next five issues for the 03-04 130 RE2 = RE * RE : 並非曼德勃羅集內每一個複數點的實
academic year, send us five stamped self-addressed envelopes. IM2 = IM * IM : J = J + 1 際坐標,故需要作出轉換。R 是實際
Send all correspondence to: 的數值(見第 20 行) ,即繪畫的範圍
140 WEND
Dr. Kin-Yin LI
150 IF J < 3 THEN GOTO 220 實軸由 −R 畫至 +R,同時虛軸亦由 −R
Department of Mathematics
The Hong Kong University of Science and Technology 160 IF J >= 3 AND J < 6 THEN 畫至 +R。S 計算 W 與 R 之間的比例,
Clear Water Bay, Kowloon, Hong Kong COLOR 14 : REM YELLOW 並應用於後面的計算之中(見第 20 及
Fax: (852) 2358 1643 70 行)。
Email: makyli@ust.hk
Mathematical Excalibur, Vol. 8, No. 4, Aug 03- Oct 03 Page 2
The 2003 Hong Kong IMO team from left to right: Wei Fei Fei (Guide), Leung Chit Wan (Deputy Leader), Chung Tat Chi, Siu
Tsz Hang, Kwok Tsz Chiu, Yu Hok Pun, Yeung Kai Sing, Lau Wai Shun, Leung Tat Wing (Leader).
Volume 9, Number 3 August 2004 – September 2004
Extending an IMO Problem notations r | s means s is divisible by r and for k = 1, 2, … By Lemma 2, we also
Hà Duy Hung u ≡ v (mod m) means u – v is divisible by have
m. ⎛y ⎞
Dept. of Math and Informatics gcd ⎜ k , p k ⎟ = 1
Hanoi Univ. of Education Lemma 2. Let a, b be distinct and ⎝ p ⎠
relatively prime positive integers, and p an for k = 1, 2, …. Moreover, we have
In this brief note we give a odd prime number which divides a + b .
generalization of a problem in the 41st xk ≥ p k . This leads us to
Then for any non-negative integer k,
International Mathematical Olympiad p −1
p k +1 | a m + b m , 2
+ ∑ [(a p ) 2i (b p ) p −1− 2i
p k ( p −1) k k
held in Taejon, South Korea in 2000. yk = b
where m = p k . i =1
IMO 2000/5. Determine whether or k k
− (a p ) 2i −1 (b p ) p − 2i ]
not there exists a positive integer n Proof. We prove the lemma by induction. k k
such that n is divisible by exactly It is clear that the lemma holds for k = 0 . > bp + ap
2000 different prime divisors, and Suppose the lemma holds for some = xk
2n + 1 is divisible by n. non-negative integer k, and we proceed to ≥ p k +1
the case k + 1 .
The answer to the question is k k
It follows that
Let x = a p and y = b p . Since
positive. This intriguing problem
yk
made me recall a well-known theorem p −1 ≥ pk > 1 .
due to O. Reutter in [1] as follows. x + y = ( x + y )∑ (−1) x
p p i p −1− i i
y , p
i =0
∑ (−1) x i p −1− i
yi
divisor qk of
yk
.
We frequently encounter the i =0
p
p −1
theorem in the case a = 2 . The
≡ ∑ (−1) 2i x p −1
theorem and the IMO problem i =0
We now have a sequence of odd
prompted me to think of more general ≡ px p −1 (mod p k +1 )
prime numbers {qk }k =1 satisfying the
+∞
problem. Can we replace the number 2
in the IMO problem by other positive completing the proof.
integers? The difficulty partly lies in following properties
the fact that the two original problems In the rest of this note we shall
are solved independently. After a long complete the proof of Theorem 2.
(i) gcd ( xk , qk ) = 1
time, I finally managed to prove a
Proof of Theorem 2. Without loss of
generalization as follows. (ii) gcd ( p, qk ) = 1
generality, let a > b . Since a + b is not a
power of 2, it has an odd prime factor p. (iii) qk | xk +1
Theorem 2. Let s, a, b be given
For natural number k, set
positive integers, such that a, b are (iv) xk | xk +1 .
relatively prime and a + b is not a pk pk x
xk = a + b , yk = k +1 .
power of 2. Then there exist infinitely xk We shall now show that the sequence
many positive integers n such that
Then yk is a positive integer and
{qk }k =1
+∞
• n has exactly s different prime consists of distinct prime
p −1
divisors; and
yk = ∑ (−1)i (a p ) p −1−i (b p )i
k k
Lemma 1. Let n be an odd positive yk by properties (iii) and (iv). But this
integer, and a, b be relative prime which implies that is a positive contradicts property (i).
p
positive integers. Then
Next, set n0 = p s q1 ...qs −1 and nk+1 =
integer. Also, we have
a +b
n n
pnk for k = 0, 1, 2, …. It is evident that
a+b yk ⎛ xk ⎞
≡b ( )
p k p −1
⎜ mod ⎟,
is an odd integer ≥ 1 , equality if and p ⎝ p ⎠ {nk }k =0 is a strictly increasing sequence
+∞
only if n = 1 or a = b = 1 .
so that
The proof of Lemma 1 is simple and is
left for the reader. ⎛x y ⎞ (continued on page 4)
gcd ⎜ k , k ⎟ = 1
Also, we remind readers the usual ⎝ p p⎠
Mathematical Excalibur, Vol. 9, No. 3, Aug. 04- Sept. 04 Page 3
Problem Corner tan A tan B tan C ∠ADF > ∠EDF. So we need ∠BAC
> [tan A] + [tan B] + [tan C], + ∠ABC > ∠BCA. Similarly, for BE to
We welcome readers to submit their coincide with CE and for CF to coincide
solutions to the problems posed below where [t] denotes the greatest integer less
with AF, we need ∠ ABC + ∠ BCA >
for publication consideration. The than or equal to t. Give a proof.
∠BAC and ∠BCA + ∠BAC > ∠ABC.
solutions should be preceded by the Solution. CHENG Hao (The Second So no angle of ∆ABC is 90 or more.
solver’s name, home (or email) address High School Attached to Beijing Normal Therefore, ∆ABC is acute.
and school affiliation. Please send University), CHEUNG Yun Kuen
submissions to Dr. Kin Y. Li, (HKUST, Math, Year 1) and YIM Wing
Yin (South Tuen Mun Government Conversely, if ∆ABC is acute, then
Department of Mathematics, The Hong Secondary School, Form 4). , Achilleas P. reversing the steps, we can see that the
Kong University of Science & PORFYRIADIS (American College of
Thessaloniki “Anatolia”, Thessaloniki, required tetrahedron can be obtained.
Technology, Clear Water Bay, Kowloon,
Hong Kong. The deadline for Greece),
Problem 203. (Due to José Luis
submitting solutions is October 20, From DÍAZ-BARRERO, Universitat Politec-
2004. nica de Catalunya, Barcelona, Spain)
tan C = tan (180○– A – B)
Problem 206. (Due to Zdravko F. Let a, b and c be real numbers such that
= – tan (A+B)
Starc, Vršac, Serbia and Montenegro) a + b + c ≠ 0. Prove that the equation
= – (tan A + tan B)/(1– tan A tan B),
Prove that if a, b are the legs and c is (a+b+c)x2 + 2(ab+bc+ca)x + 3abc = 0
the hypotenuse of a right triangle, then we get
has only real roots.
tan A + tan B + tan C = tan A tan B tan C.
(a + b) a + (a − b) b < 2 2c c.
Solution. CHAN Pak Woon (Wah Yan
Let x = tan A, y = tan B and z = tan C. If College, Kowloon, Form 6), CHENG Hao
Problem 207. Let A = { 0, 1, 2, …, 9} (The Second High School Attached to
xyz ≤ [x]+[y]+[z], then x+y+z ≤ [x] + [y]
and B1, B2, …, Bk be nonempty subsets Beijing Normal University), CHEUNG
of A such that Bi and Bj have at most 2 + [z]. As [t] ≤ t, x, y, z must be integers. Hoi Kit (SKH Lam Kau Mow Secondary
common elements whenever i ≠ j. Find School, Form 7), CHEUNG Yun Kuen
(HKUST, Math, Year 1), Murray
the maximum possible value of k. If triangle ABC is obtuse, say A > 90○, KLAMKIN (University of Alberta,
then x < 0 < 1 ≤ y ≤ z. This implies 1 ≤ yz Edmonton, Canada), Achilleas P.
Problem 208. In ∆ABC, AB > AC > BC. PORFYRIADIS (American College of
= (x + y + z)/x = 1 + (y + z)/x < 1, a Thessaloniki “Anatolia”, Thessaloniki,
Let D be a point on the minor arc BC of
contradiction. If triangle ABC is acute, Greece) and YIM Wing Yin (South Tuen
the circumcircle of ∆ABC. Let O be Mun Government Secondary School,
the circumcenter of ∆ABC. Let E, F be then we may assume 1 ≤ x ≤ y ≤ z. Now Form 4).
the intersection points of line AD with xy = (x+y+z)/z ≤ (3z)/z = 3. Checking the
The quadratic has real roots if and only
the perpendiculars from O to AB, AC, cases xy = 1, 2, 3, we see x+y+z = xyz can
if its discriminant
respectively. Let P be the intersection only happen when x=1, y=2 and z=3.
D = 4(ab+bc+ca)2 – 12(a+b+c)abc
of lines BE and CF. If PB = PC + PO, This corresponds to A = tan-1 1, B = tan-1
= 4[(ab)2+(bc)2+(ca)2–(a+b+c)abc]
then find ∠BAC with proof. 2 and C = tan-1 3. Reversing the steps, = 4[(ab–bc)2 +(bc–ca)2 +(ca–ab)2 ]
Problem 209. Prove that there are we see among nonright triangles, the is nonnegative, which is clear.
infinitely many positive integers n such inequality in the problem holds except
Other commended solvers: Jason CHENG
that 2n + 2 is divisible by n and 2n + 1 is only for triangles with angles equal 45○ Hoi Sing (SKH Lam Kau Mow Secondary
divisible by n – 1. = tan-1 1, tan-1 2 and tan-1 3. School, Form 7), POON Ho Yin (Munsang
College (Hong Kong Island), Form 4) and
Problem 210. Let a1 = 1 and Problem 202. (Due to LUK Mee Lin, La Anderson TORRES (Universidade de
Sao Paulo – Campus Sao Carlos).
a 1 Salle College) For triangle ABC, let D, E,
a n +1 = n + F be the midpoints of sides AB, BC, CA, Problem 204. Let n be an integer with
2 an
respectively. Determine which triangles n > 4. Prove that for every n distinct
for n = 1, 2, 3, … . Prove that for every ABC have the property that triangles ADF, integers taken from 1, 2, …, 2n, there
BED, CFE can be folded above the plane always exist two numbers whose least
integer n > 1, of triangle DEF to form a tetrahedron with common multiple is at most 3n + 6.
2 AD coincides with BD; BE coincides with Solution. CHENG Hao (The Second
a 2
− 2 CE; CF coincides with AF. High School Attached to Beijing
n
Normal University), CHEUNG Yun
is an integer. Solution. CHENG Hao (The Second Kuen (HKUST, Math, Year 1) and YIM
High School Attached to Beijing Normal Wing Yin (South Tuen Mun
University), CHEUNG Yun Kuen Government Secondary School, Form
(HKUST, Math, Year 1) and YIM Wing 4).
***************** Yin (South Tuen Mun Government
Solutions Secondary School, Form 4). Let S be the set of n integers taken and k
**************** Observe that ADEF, BEFD and CFDE are be the minimum of these integers. If k ≤
parallelograms. Hence ∠BDE =∠BAC, n, then either 2k is also in S or 2k is not
Problem 201. (Due to Abderrahim ∠ADF = ∠ABC and ∠EDF = ∠BCA. in S. In the former case, lcm(k,2k) = 2k
OUARDINI, Talence, France) Find In order for AD to coincide with BD in ≤ 2n < 3n+6. In the latter case, we
which nonright triangles ABC satisfy folding, we need to have ∠ BDE + replace k in S by 2k. Note this will not
Mathematical Excalibur, Vol. 9, No. 3, Aug. 04- Sept. 04 Page 4
decrease the least common multiple of p) implies p – 1 is divisible by d. Hence, ∠ PBC =∠ DBA and ∠ PDC =∠BDA
any pair of numbers. So if the new S gcd(n, p – 1) is divisible by d. Since p is Prove that ABCD is a cyclic quadrilateral
satisfies the problem, then the original S the smallest prime dividing n, we must if and only if AP = CP.
will also satisfy the problem. As we have gcd(n, p – 1) = 1. So d =1. By the
repeat this, the new minimum will definition of d, we get a – 1 is divisible by Problem 6. We call a positive integer
increase strictly so that we eventually
p. Therefore, gcd(a – 1, n) ≥ p > 1. alternating if every two consecutive digits
reach either k and 2k both in S, in which
in its decimal representation are of
case we are done, or the new S will consist
different parity.
of n+1, n+2, …, 2n. So we need to
consider the latter case only. Olympiad Corner Find all positive integers n such that n has
a multiple which is alternating.
If n > 4 is even, then 3(n+2)/2 is an integer (continued from page 1)
at most 2n and lcm(n+2, 3(n+2)/2) =
3n+6. If n > 4 is odd, then 3(n+1)/2 is an
integer at most 2n and lcm(n+1,3(n+1)/2) Extending an IMO Problem
= 3n+3. (continued from page 2)
Problem 205. (Due to HA Duy Hung, or any of the figures obtained by applying
of positive integers and each term of the
Hanoi University of Education, Vietnam) rotations and reflections to this figure.
sequence has exactly s distinct prime
Let a, n be integers, both greater than 1, Determine all m x n rectangles that can be divisors.
such that an – 1 is divisible by n. Prove covered with hooks so that It remains to show that
that the greatest common divisor (or ● the rectangle is covered without gaps nk | a nk + b nk
highest common factor) of a – 1 and n is and without overlaps; for k = 0, 1, 2, … Note that for odd
greater than 1. ● no part of a hook covers area outside the positive integers m, n with m | n , we have
Solution. CHENG Hao (The Second rectangle.
a m + b m | a n + b n . By property (iii), we
High School Attached to Beijing Normal
University), CHEUNG Yun Kuen Day 2 Time allowed: 4 hours 30 have, for 0 ≤ k < s ,
n n
(HKUST, Math, Year 1) and YIM Wing minutes. qk | xk +1 | xs | a n0 + b n0 | a j + b j
Yin (South Tuen Mun Government
Secondary School, Form 4). for j = 0, 1, 2, … . Now it suffices to show
Problem 4. Let n ≥ 3 be an integer. Let that
Let p be the smallest prime divisor of n. t1, t2, …, tn be positive real numbers such p k + s | a nk + b nk
Then an – 1 is divisibly by p so that an ≡ 1 that for k = 0, 1, 2, … . But this follows easily
(mod p). In particular, a is not divisible by n2 + 1 > (t1 + t2 + ⋯ + tn) from Lemma 2 since
p. Then, by Fermat’s little theorem, ap – 1 p s + k | xk + s | a nk + b nk .
⎛ 1 1 1 ⎞
≡1 (mod p). × ⎜ + + L + ⎟. This completes the proof of Theorem 2.
⎝ t1 2
t n
t ⎠
Let d be the smallest positive integer such Show that ti, tj, tk are side lengths of a References:
that ad ≡ 1 (mod p). Dividing n by d, we triangle for all i, j, k with 1≤ i <j <k ≤ n.
[1] O. Reutter, Elemente der Math., 18
get n = dq + r for some integers q, r with 0 (1963), 89.
≤ r < d. Then ar ≡ (ad)q ar = an ≡ 1 (mod Problem 5. In a convex quadrilateral
[2] W. Sierpinski, Elementary Theory of
p). By the definition of d, we get r = 0. ABCD the diagonal BD bisects neither the Numbers, English translation,
Then n is divisible by d. Similarly, angle ABC not the angle CDA. The point Warsaw, 1964.
dividing p – 1 by d, we see ap – 1 ≡ 1 (mod P lies inside ABCD and satisfies
2004 Hong Kong team to IMO: From left to right, Cheung Yun Kuen, Poon Ming Fung, Tang Chiu Fai, Cesar Jose Alaban
(Deputy Leader), Leung Tat Wing (Leader), Chung Tat Chi, Kwok Tsz Chiu & Wong Hon Yin.
Volume 14, Number 2 May - September, 2009
Likewise, be 60° or 90°. To prove the statement of induction step. Later the Russians
the problem, one may either use synthetic provided a solution by induction, by
AQ·QB = (R+OQ)(R−OQ) method or coordinate method. One separating the problem into sub-cases
= R2−OQ2. advantage of using the coordinate method min M < an or min M ≥ an, and then
These force OP2 = OQ2, or OP = OQ, is after showing the possible values of applying the principle-hole principle,
done! ∠ CAB, one can go back to show these etc judiciously to solve the problem.
values do work by suitable substitutions. Terry Tao said (jokingly) that the six
Some contestants lost marks either problems were easy. But in his blog, he
Problem 3. Suppose that s1, s2, s3, … because they missed some values of admitted that he had spent sometime
is a strictly increasing sequence of ∠ CAB or forgot to check the two reading the problem and he even wrote
positive integers such that the possible cases do work. an article about it (I have not seen the
subsequences article.)
ss1 , ss2 , ss3 ,... and ss1+1, ss2 +1, ss3 +1,... Problem 5. Determine all functions f
The two hard problems (3 and 6)
from the set of positive integers to the set
were more combinatorial and/or
are both arithmetic progressions. of positive integers such that, for all
algebraic in nature. I had a feeling that
Prove that the sequence s1, s2, s3, … is positive integers a and b, there exists a
this year the Jury has been trying to
itself an arithmetic progression. non-degenerate triangle with sides of
avoid hard number theory problems,
lengths a, f(b) and f(b+f(a)−1). (A triangle
which were essentially corollaries of
This was one of the two hard is non-degenerate if its vertices are not
deep theorems (for example, IMO
problems (3 and 6). Fortunately, it collinear.)
2003 problem 6 by the Chebotarev
turned out that it was still within reach. density theorem or IMO 2008 problem
The Jury worried if the word
3 by a theorem of H. Iwaniec) or hard
One trouble is of course the triangle may be allowed to be degenerate
geometry problem using sophisticated
notation. Of course, ss1 stands for the in some places. But I supposed all our
geometric techniques (like IMO 2008
s1th term of the si sequence and so on. secondary school students would consider
problem 6).
Starting from an arithmetic progression only non-degenerate triangles. This was a
(AP) with common difference d, then it nice problem in functional inequality
The Germans ran the program
is easy to check that both (triangle inequality). One proves the
vigorously (obstinately). They had an
problem by establishing several basic
organization (Bildung und Begabung)
ss1 , ss2 , ss3 ,... and ss1+1, ss2+1, ss3+1,... properties of f. Indeed the first step is to
that looked after the entire event. They
prove f(1)=1, which is not entirely easy.
had also prepared a very detailed
Then one proceeds to show that f is
are APs with common difference d2. shortlist problem set and afterwards
injective and/or f(f(x)) = x, etc, and finally
The question is essentially proving the prepared very detailed marking
shows that the only possible function is
“converse”. So the first step is to prove schemes for each problem. The
the identity function f(x) = x for all x.
that the common differences of the two coordinators were very professional
APs ssi and ssi +1 are in fact the same, and they studied the problems well.
Problem 6. Let a1, a2, …, an be distinct
say s. It is not too hard to prove and is Thus, there were not too many
positive integers and let M be a set of
intuitively clear, for two lines of arguments about how many points
n−1 positive integers not containing
different slopes will eventually meet should be awarded for each problem.
s=a1+a2+⋯+an. A grasshopper is to jump
and cross each other, violating the
along the real axis, starting from the point
condition of strictly increasing Three of the problems (namely 1,
O and making n jumps to the right with
sequence. The next step is the show 2 and 4) were relatively easy, problems
lengths a1, a2, …, an in some order. Prove
the difference between two 3 and 5 were not too hard, so although
that the order can be chosen in such a way
consecutive terms of si is indeed s , problem 6 was hard, contestants still
that the grasshopper never lands on any
(thus s is a square). One can achieve scored relatively high points. This
point in M.
this end by the method of descent, or explained why the cut-off scores were
max/min principle, etc. not low, 14 for bronze, 24 for silver and
It turned out that this problem was one 32 for gold.
Problem 4. Let ABC be a triangle with of the most difficult problems in IMO
AB = AC. The angle bisectors of history. Only three of the 564 contestants It might seem that we still didn’t
∠ CAB and ∠ ABC meet the sides BC received full scores. (Perhaps it was do the hard problems too well. But
and CA at D and E, respectively. Let K second to problem 3 posed in IMO 2007, after I discussed with my team
be the incenter of triangle ADC. for which only 2 contestants received full members, I found that they indeed had
Suppose that ∠ BEK = 45°. Find all scores.) the potential and aptitude to do the hard
possible values of ∠ CAB. problems. What may still be lacking
When I first read the solution provided are perhaps more sophisticated skills
This problem was also relatively by the Problem Committee, I felt I was and/or stronger will to tackle such
easy. It is interesting to observe that an reading a paper of analysis. Without problems.
isosceles triangle can be the starting reading the solution, of course I would say
point of an IMO problem. With we could try to prove the problem by
geometric software such as Sketchpad, induction, as the cases of small n were (continued on page 4)
one can easily see that ∠ CAB should easy. The trouble was how to establish the
Mathematical Excalibur, Vol. 14, No. 2, May-Sep. 09 Page 3
then prove that line O1O2 bisects line ∠ MQN = ∠ MQD + ∠ DQN counting region by region. On the other
segment AP. = ∠ NQE + ∠ DQN hand, it is also (4k2+6k3+⋯2nkn)+4n by
= ∠ DQE counting the number of edges around
Solution. Jim Robert STUDMAN
= ∠ DAE the k-points and around the vertices of
(Hanford, Washington, USA).
= ∠ MAN. M. The 4n term is due to the 2n edges
Let the circumcircle of ΔADE and the of M and each vertex of M (being not a
circumcircle of ΔABC intersect at A Comments: Some solvers used a bit of k-point) issues exactly one edge into
and Q. homothety to simplify the proof. the interior of M. So we have
Observe that line O1O2 bisects chord Other commended solvers: CHUNG 3P3+4P4+5P5+⋯=4k2+6k3+⋯2nkn+4n.
AQ and O1O2⊥AQ. Hence, line O1O2 Ping Ngai (La Salle College, Form 6), Subtracting the last two displayed
bisects line segment AP will follow if LAM Cho Ho (CUHK Math Year 1), NG equations, we can obtain
we can show that O1O2 || PQ, or Ngai Fung (STFA Leung Kau Kui
equivalently that PQ⊥AQ. College, Form 7). P3 + P4 + P5 +L= k2 + 2k3 + (n −1)kn + n +1.
A Problem 325. On a plane, n distinct lines Finally, the number of regions these n
are drawn. A point on the plane is called a lines divided the plane into is the limit
Q k-point if and only if there are exactly k of case r tends to infinity. Hence, it is
O1 exactly k2+2k3+⋯+(n−1)kn+n+1.
the n lines passing through the point. Let
E
k2, k3, …, kn be the numbers of 2-points,
3-points, …, n-points on the plane, Other commended solvers: CHUNG
D
N Ping Ngai (La Salle College, Form 6)
M O2 respectively.
P and YUNG Fai.
Determine the number of regions the n
B lines divided the plane into in terms of n,
k2, k3, …, kn. Remarks on IMO 2009
C
(Source: 1998 Jiangsu Province Math (continued from page 2)
Competition)
Let points M and N be the feet of As I found out from the stronger
perpendiculars from P to lines AB and Solution. LAM Cho Ho (CUHK Math teams (Chinese, Japanese, Korean, or
AC respectively. Since ∠ANP = 90° = Year 1). Thai, etc.), they were obviously more
∠AMP, points A, N, P, M lie on a circle Take a circle of radius r so that all heavily or vigorously trained. For
Г with AP as diameter. We claim that intersection points of the n lines are inside instance, a Thai boy/girl had to go
∠MQN =∠MAN. This would imply the circle and none of the n lines is tangent through more like 10 tests to be
Q is also on circle Г, and we would to the circle. Now each line intersects the selected as a team member.
have PQ⊥AQ as desired. circle at two points. These 2n points on Another thing I learned from the
the circle are the vertices of a convex meeting was several countries were
Since we are given ∠ ADP = ∠ AEP, 2n-gon (call it M) as we go around the
we get ∠ BDP = ∠ CEP. This interested to host the event (South-East
circle, say clockwise. Let the n lines Asia countries and Asia-Minor
combines with the given fact ∠ DPB = partition the interior of M into P3 triangles,
∠ EPC imply ∆DPB and ∆EPC are countries). In fact, one country is
P4 quadrilaterals, ⋯, Pj j-gons, ⋯. These going to host three international
similar, which yields DB/EC = polygonal regions are all convex since the
DP/EP=DM/EN. competitions of various subjects in a
angles of these regions, which were row for three years. Apparently they
formed by intersecting at least two lines, think hosting these events is good for
Since A,E,D,Q are concyclic, we have are all less than 180°. By convexity, no gifted education.
∠ BDQ =180°− ∠ ADQ two sides of any polygonal region are
=180°− ∠ AEQ = ∠ CEQ. parts of the same line. So we have Pj = 0 The first IMO was held in Romania
for j>3n. in 1959. Throughout these 51 years,
This and ∠ DBQ=∠ABQ =∠ACQ = only one year IMO was not held (1980).
∠ ECQ imply ∆DQB and ∆EQC are Consider the sum of all the angles of these To commemorate the fiftieth
similar. So we have QD/QE=DB/EC. regions partitioning M. On one hand, it is anniversary of IMO in 2009, six
Combining with the equation at the end 180°(P3+2P4+3P5+⋯) by counting region notable mathematicians related to IMO
of the last paragraph, we get by region. On the other hand, it also (B. Bollabas, T. Gowers, L. Lovasz, S.
equals 360°(k2+k3+⋯+kn)+(2n−2)180° by Smirnov, T. Tao and J. C. Yoccoz) were
QD/QE=DM/EN. counting all the angles around each invited to talk to the contestants. Of
vertices of the regions. Cancelling 180°,
Using ∆DQB and ∆EQC are similar, course, Yoccoz, Gowers and Tao were
we get
we get ∠ MDQ = ∠ BDQ = ∠ CEQ Fields medalists. The afternoon of
= ∠ NEQ. These imply ∆MDQ and P3+2P4+3P5+⋯=2(k2+k3+⋯+kn)+(2n−2). celebration then became a series of
∆NEQ are similar. Then ∠ MQD = (rather) heavy lectures (not bad). They
∠ NQE. Next, consider the total number of all the described the effects of IMOs on them
edges of these regions partitioned M (with and other things. The effect of IMO on
Finally, for the claim, we now have each of the edges inside M counted twice). the contestants is to be seen later, of
On one hand, it is 3P3+4P4+5P5+⋯ by course!
Volume 16, Number 3 November 2011 -January 2012
Problem 1. Given any set A={a1, a2, The 52nd IMO was held in Amsterdam, Now as we need a1 + a4 dividing a2 + a3
a3, a4} of four distinct positive integers, Netherlands, on 12-24, July, 2011. and also a2 + a3 dividing a1 + a4, we
we denote the sum a1+a2+a3+a4 by sA. Contestants took two 4½ hour exams must have a1 + a4 = a2 + a3. Putting a4=
Let nA denote the number of pairs (i,j) during the mornings of July 18 and 19. a2+a3−a1 into the equations a3 + a4 =
Each exam was consisted of 3 problems m(a1 + a2) and a2 + a4 = n(a1 + a3) with m
with 1≤i<j≤4 for which ai+aj divides
of varying degree of difficulty. The > n > 1, we eventually get (m,n) = (3,2)
sA. Find the sets A of four distinct
problems were first shortlisted by the or (4,2). Finally we get (a1, a2, a3, a4) =
positive integers which achieve the
host country, selected from problems (k, 5k, 7k, 11k) or (k, 11k, 19k, 29k),
largest possible value of nA.
submitted earlier by various countries. where k is a positive integer. As the
Problem 2. Let S be a finite set of at Leaders from 101 countries then picked derivation of the answers is rather
least two points in the plane. Assume the 2011 IMO problems (see Olympiad straight-forward, it does not pose any
that no three points of S are collinear. Corner). Traditionally an easy pair was serious difficulty.
A windmill is a process that starts with selected (Problems 1 and 4), then a hard
a line ℓ going through a single point P pair (Problems 3 and 6), with Problem 6 For problem 4, it is really quite easy if
∊S. The line rotates clockwise about usually selected as the “anchor one notes the proper recurrence relation.
the pivot P until the first time that the problem”, and finally the intermediate Indeed the weights 20, 21, 22, …, 2n−1
line meets some other point belonging pair (Problems 2 and 5). I would like to form a “super-increasing sequence”,
to S. This point, Q, takes over as the discuss first the problems selected, aim any weight is heavier than the sum of all
new pivot, and the line now rotates to provide something extra besides lighter weights. Denote by f(n) the
clockwise about Q, until it next meets a those which were provided by the number of ways of placing the weights.
point of S. This process continues solutions. However I would discuss the We consider first how to place the
indefinitely. problems by slightly different grouping. lightest weight (weight 1). Indeed if it is
placed in the first move, then it has to be
Show that we can choose a point P in S Problems 1 and 4 in the left pan. However if it is placed in
and a line ℓ going through P such that the second to the last move, then it
First the easy pair, problems 1 and 4.
the resulting windmill uses each point really doesn’t matter where it goes,
The problem selection committee
of S as a pivot infinitely many times. using the “super-increasing property”.
thought that both problems were quite
Hence altogether there are 2n−1
(continued on page 4) easy. It was nice to select one as a
possibilities of placing the weight of
problem of the contest. But if both
Editors: 張 百 康 (CHEUNG Pak-Hong), Munsang College, HK weight 1. Now placing the weights 21,
problems were selected, then the paper
高 子 眉 (KO Tsz-Mei) 22, …, 2n−1 clearly is the same as
would be too easy (or even disastrous).
梁 達 榮 (LEUNG Tat-Wing) placing the weights 20, 21, …, 2n−2.
李 健 賢 (LI Kin-Yin), Dept. of Math., HKUST Indeed eventually both problems were
There are f(n−1) ways of doing this.
吳 鏡 波 (NG Keng-Po Roger), ITC, HKPU selected. But it was not enough for
Thus we establish the recurrence
Artist: 楊 秀 英 (YEUNG Sau-Ying Camille), MFA, CU anyone to get a bronze medal even if he
relation f(n) = (2n−1)f(n−1). Using f(1)
could solve both problems (earning 14
Acknowledgment: Thanks to Elina Chiu, Math. Dept., = 1, by induction, we get
HKUST for general assistance. points) as the cut-off for bronze was 16.
On-line: f(n) = (2n −1)(2n −3)(2n −5)⋯1.
http://www.math.ust.hk/mathematical_excalibur/ In my opinion problem 1 is the easier of
The problem becomes a mere exercise
The editors welcome contributions from all teachers and the pair. Indeed we may without loss of
students. With your submission, please include your name, of recurrence relation if one notices how
generality assume a1 < a2 < a3 < a4. So if
address, school, email, telephone and fax numbers (if to place the lightest weight (minimum
available). Electronic submissions, especially in MS Word, the sum of one pair of the ai’s divides sA,
principle).
are encouraged. The deadline for receiving material for the then it will also divide the sum of the
next issue is February 28, 2012.
other pair. But clearly a bigger pair
For individual subscription for the next five issues for the It is slightly harder if we consider how
09-10 academic year, send us five stamped self-addressed cannot divide a smaller pair, so it is
to place the heaviest weight. Indeed if
envelopes. Send all correspondence to: impossible that a3 + a4 dividing a1 + a2,
the heaviest weight is to be placed in the
Dr. Kin-Yin LI, Math Dept., Hong Kong Univ. of Science nor is it possible that a2 + a4 dividing a1
ith move, then it has to be placed in the
and Technology, Clear Water Bay, Kowloon, Hong Kong + a3. Therefore the maximum possible
value of nA can only be 4. To achieve ⎛n−1⎞
Fax: (852) 2358 1643 left pan. There are ⎜ ⎟ ways of
⎝ i −1⎠
Email: makyli@ust.hk
this, it suffices to consider divisibility
© Department of Mathematics, The Hong Kong University
of Science and Technology
conditions among the other pairs.
Mathematical Excalibur, Vol. 16, No. 3, Nov.11 -Jan. 12 Page 2
choosing the previous i−1 weights and comes from one of our team members. following “continuity argument”. We
there are f(i−1) ways of placing them. Put y = z−x into the original equation consider only the case that there are an
After the heaviest weight is placed, it f(x+y) ≤ yf(x) + f(f(x)), one gets f(z) ≤ z odd number of points on the plane. Let
doesn’t matter how to place the other f(x) − xf (x) + f( f(x)). By letting z = f(k) in l be a line that goes through one of the
weights, and there are (n−i)!×2n−i ways the derived inequality one gets f(f(k)) ≤ points and that separates the other
of placing the remaining weights. Thus f(k) f(x) − xf (x) + f( f(x)). points into two equal halves. Note that
n
⎛ n − 1⎞ Interchanging k and x one then gets f(f(x)) such line clearly exists. Color one
f (n) = ∑ ⎜ n −i
⎟ f (i − 1)(n − i)!2 . ≤ f(k) f(x) − kf (k) + f( f(k)). Hence half-plane determined by the line
i =1 ⎝ i − 1 ⎠
orange (for Netherlands) and the other
f(x+y) ≤ y f(x) + f( f(x))
Replacing n by n −1 and by comparing half-plane blue. The color of the plane
≤ f(x)f(k) −kf(k) + f(f(k)).
the two expressions we again get f(n) = changes accordingly while the line is
(2n−1)f(n−1). We have no serious Letting y =f(k) − x in the inequality, we get turning. Note also that when the line
difficulty with this problem. moves to another pivot, the number of
f(f(k)) ≤ f(k) f(x) − xf (x) + points on the two sides remain the
Problems 3 and 5 f(k) f(x) − kf (k) + f( f(k)) same, except when two points are on
or 0 ≤ 2 f(k) f(x) − xf(x) −k f(k). Finally the line during the change of pivots. So
In my opinion both problems 3 and 5
letting k = 2 f(x) and simplifying, we consider what happen when the line
were of similar flavor. Both were
arrive at the important and essential turns 180°, (turning while changing
“functional equation” type of problems.
(hidden) inequality 0 ≤ −xf(x). This means pivots). The line will go through the
Problem 3 was slightly more involved
for x > 0, f(x) ≤ 0, and for x < 0, f(x) ≥ 0. same original starting point. Only the
and problem 5 more number theoretic.
But if there is an x0 < 0 such that f(x0) > 0, colors of the two sides of the line
One can of course put in many values
then putting x = x0 and y = 0 into the interchange! This means all the points
and obtain some equalities or
original equation, we gets 0<f(x0) ≤ f(f(x0)). have been visited at least once! A
inequalities. But the important thing is
However if f(x0) > 0, then f(f(x0)) ≤ 0, slightly modified argument works for
to substitute some suitable values so
hence a contradiction. This means for all the case there are an even number of
that one can derive important relevant
x < 0, f(x) = 0. Finally one has to prove f(0) points on the plane.
properties that can solve the problem.
= 0. We suppose first f(0) > 0. Put x = 0 Problem 6
In problem 5, indeed the condition and y < 0 sufficiently small into the
f(m−n) | (f(m) − f(n)) (*) poses very original equation, one gets f(y) < 0, a This was the most difficult problem of
serious restrictions on the image of f(x). contradiction. Suppose f(0) < 0. Take x, y the contest (the anchor problem), only
Putting n=0, one gets f(m) | (f(m) − f(0)), < 0. We get 6 out of more than 564 contestants
thus f(m) | f(0). Since f(0) can only 0 = f(x+y) ≤ y f(x) + f( f(x)) solved the problem. Curiously these
have finitely many factors, the image = y f(x) + f(0) = f(0) < 0, solvers were not necessarily from the
of f(x) must be finite. Putting m=0, one strongest teams. The problem is hard
gets f(−n) | f(n), and by interchanging n again contradiction! This implies f(0) = 0. and beautiful, and I feel that it may be a
and −n, one gets f(n) = f(−n). Now f(n) known problem because it is so nice.
| (f(2n) − f(n)), hence f(n) | f(2n), and Problem 2 However, I am not able to find any
by induction f(n) | f(mn). Put n = 1 into To me, problem 2 was one of a kind. The further detail. It is not convenient to
the relation. One gets f(1) | f(m). The problem was considered as “intermediate” reproduce the full solution here. But I
image of f(x) is therefore a finite and should not be too hard. However at still want to discuss the main idea used
sequence f(1) = a1< a2< ⋯ < ak = f(0). the end only 21 out of 564 contestants in the first official solution briefly.
One needs to show ai | ai+1. To scored full marks. It was essentially a
complete the proof, one needs to problem of computational geometry. We Lc Lb
analyze the sequence more carefully, know that if there is a line that goes H A
C' C"
say one may proceed by induction on k. through two or more of the points and B'
La
B" γ
But personally I like the following such that all other points are on the line or C
A'
argument. Let f(x) = ai and f(y) = ai+1. only on one side of the points, then by B L
We have f(x−y) | (f(y) − f(x)) < f(y) and repeatedly turning angles as indicated in T
f(y) − f(x) is positive, hence f (x − y) is the problem, the convex hull of the point
in the image of f(x) and therefore f(x−y) set will be constructed (so-called Jarvis’ A"
≤ ai= f(x). Now if f(x−y) < f(x), then f(x) march). Therefore some points may be
− f(x −y) > 0. Thus f(y) = f(x−(x−y)) | missed. So in order to solve the problem,
(f(x) − f(x −y)). we cannot start from the “boundary”. From ΔABC and the tangent line L at T,
In this case the right-hand side is Thus it is natural that we start from the we produce the reflecting lines La, Lb,
positive. We have f(y) ≤ f(x) − f(x −y)) < “center”, or a line going through a point and Lc. The reflecting lines meet at A”,
f(x) < f(y), a contradiction. So we have that separates the other points into equal B” and C” respectively. Now from A,
f(x−y) = f(x). Thus f(x) | f(y) as needed. halves (or differ by one). Indeed this idea we draw a circle of radius AT, meeting
is correct. The hard part is how to the circumcircle γ of ABC at A’.
It seems that Problem 3 is more substantiate the argument. Many Likewise we have BT=BB’ and
involved. However, by making useful contestants found it hard. Induction CT=CC’ (see the figure).
and clever substitutions, it is possible argument does not work because adding
to solve the problem in a relatively or deleting one point may change the
easy way. The following solution entire route. The proposer gives the (continued on page 4)
Mathematical Excalibur, Vol. 16, No. 3, Nov.11 -Jan. 12 Page 3
Problem Corner f(x) ≡ 0 (mod p) has solutions in integers, |z1+z2|≤|w1+w2| and |z1−z2|≤|w1−w2|.
but f(x) = 0 has no solution in integers.
We welcome readers to submit their Squaring both sides of these inequalities,
solutions to the problems posed below Solution. Alumni 2011 (Carmel Alison adding them and applying the
for publication consideration. The Lam Foundation Secondary School), parallelogram law, we get the desired
solutions should be preceded by the Maxim BOGDAN (“Mihai Eminescu” inequality. Next assume the case n=k
solver’s name, home (or email) address National College, Botosani, Romania), holds. Then for the n=k+1 case, we use
and school affiliation. Please send Koopa KOO and Andy LOO (St. Paul’s the 2k choices with ε1 = ε2 to get from the
submissions to Dr. Kin Y. Li, Co-educational College). n=k case that
Department of Mathematics, The Hong
Let f(x)=(x2−2)(x2−3)(x2−6). Then f(x) = 0 | z1 + z 2 |2 + | z3 |2 +L+ | z k +1 |2
Kong University of Science &
has no solution in integers. For p = 2 or 3, ≤ | w1 + w2 |2 + | w3 |2 + L+ | wk +1 |2 .
Technology, Clear Water Bay, Kowloon,
f(6) ≡ 0 (mod p). For a prime p > 3, if there
Hong Kong. The deadline for sending
exists x such that x2 ≡ 2 or 3 (mod p), then Similarly, using the other 2k choices
solutions is February 28, 2012.
f(x) ≡ 0 (mod p) has solutions in integers. with ε1 = −ε2, we get
Problem 381. Let k be a positive Otherwise, from Euler’s criterion, it
integer. There are 2k balls divided into follows that there will be x such that x2 ≡ 6 | z1 − z 2 | 2 + | z 3 | 2 + L+ | z k +1 | 2
a number of piles. For every two piles (mod p) and again f(x) ≡ 0 (mod p) has ≤ | w1 − w2 | 2 + | w3 | 2 + L+ | wk +1 | 2 .
A and B with p and q balls respectively, solutions in integers.
if p ≥ q, then we may transfer q balls Adding the last two inequalities and
from pile A to pile B. Prove that it is Comments: For readers not familiar with applying the parallelogram law, we get
always possible to make finitely many Euler’s criterion, we will give a bit more the n=k+1 case.
such transfers so as to have all the balls details. For c relatively prime to a prime p,
end up in one pile. by Fermat’s little theorem, we have Other commended solvers: Alumni
2011 (Carmel Alison Lam Foundation
(c(p−1)/2−1)(c(p−1)/2+1) = cp−1−1≡ 0 (mod p), Secondary School),Maxim BOGDAN
Problem 382. Let v0 = 0, v1 = 1 and
which implies c(p−1)/2 ≡ 1 or −1 (mod p). (“Mihai Eminescu” National College,
vn+1 = 8vn−vn−1 for n = 1,2,3,…. Botosani, Romania), O Kin Chit, Alex
If there exists x such that x2 ≡ c (mod p), (G.T.(Ellen Yeung) College) and
Prove that vn is divisible by 3 if and
then c(p−1)/2 ≡ xp−1 ≡ 1 (mod p). Conversely, Mohammad Reza SATOURI
only if vn is divisible by 7.
if c(p−1)/2 ≡ 1 (mod p), then there is x such (Bushehr, Iran).
Problem 383. Let O and I be the that x2 ≡ c (mod p). [This is because there
circumcenter and incenter of Δ ABC is a primitive root g (mod p) (see vol. 15, Problem 378. Prove that for all
respectively. If AB≠AC, points D, E are no. 1, p. 1 of Math Excalibur), so we get c positive integers m and n, there exists a
midpoints of AB, AC respectively and ≡ gi (mod p) for some positive integer i, positive integer k such that 2k −m has at
BC=(AB+AC)/2, then prove that the then gi(p−1)/2 ≡ 1 (mod p). Since g is a least n distinct positive prime divisors.
line OI and the bisector of ∠CAB are primitive root (mod p), so i(p−1)/2 is a
multiple of p−1, then i must be even, Solution. William PENG and Jeff
perpendicular.
hence c ≡ (gi/2)2 (mod p).] In above, if 2 PENG(Dallas,Texas, USA).
Problem 384. For all positive real and 3 are not squares (mod p), then For the case m is odd, we will prove the
numbers a,b,c satisfying a + b + c = 3, 6(p−1)/2=2(p−1)/23(p−1)/2 ≡ (−1)2 =1 (mod p), result by inducting on n. If n=1, then just
prove that hence 6 is a square (mod p). choose k large so that the odd number 2k
a2 + 3b2 b2 + 3c2 c2 + 3a2 −m is greater than 1. Next assume there
+ 2 + 2 ≥ 4. Problem 377. Let n be a positive integer. exists a positive integer k such that j =
ab (4 − ab) bc (4 − bc) ca (4 − ca)
2
For i=1,2,…,n, let zi and wi be complex 2k −m has at least n distinct positive
numbers such that for all 2n choices of ε1, prime divisors. Let s= k+φ(j2), where
Problem 385. To prepare for the IMO, ε2, …, εn equal to ±1, we have φ(j2) is the number of positive integers
in everyday of the next 11 weeks, Jack n n
at most j2 that are relatively prime to j2.
will solve at least one problem. If every ∑ε z
i =1
i i ≤ ∑ε w .
i =1
i i
Since j is odd, by Euler’s theorem,
week he can solve at most 12 problems,
then prove that for some positive n n 2 s − m ≡ 2 k × 1 − m = j (mod j 2 ).
integer n, there are n consecutive days
Prove that ∑ | zi |2 ≤ ∑ | wi |2 .
i =1 i =1 Then 2s − m is of the form j+tj2 for
in which he can solve a total of 21
some positive integer t. Hence it is
problems. Solution. William PENG and Jeff PENG divisible by j and (2s − m)/j is relatively
(Dallas,Texas, USA). prime to j. Therefore, 2s − m has at least
*****************
Solutions The case n = 1 is clear. Next, recall the n+1 distinct prime divisors.
**************** parallelogram law |a+b|2+|a−b|2=2|a|2+2|b|2,
which follows from adding the + and − For the case m is even, write m=2ir,
Problem 376. A polynomial is monic cases of the identity where i is a nonnegative integer and r is
if the coefficient of its greatest degree odd. Then as proved above there is k
(a ± b)(a ± b ) = aa ± ab ± ba + bb . such that 2k − r has at least n distinct
term is 1. Prove that there exists a
monic polynomial f(x) with integer prime divisors and so is 2i+k − m.
For n = 2, we have
coefficients such that for every prime p,
Mathematical Excalibur, Vol. 16, No. 3, Nov.11 -Jan. 12 Page 4
Other commended solvers: Maxim Problem 380. Let S = {1,2,…,2000}. If A Problem 6. Let ABC be an acute
BOGDAN (“Mihai Eminescu” and B are subsets of S, then let |A| and |B| triangle with circumcircle γ. Let L be a
National College, Botosani, Romania) denote the number of elements in A and in tangent line to γ, and let La, Lb and Lc be
B respectively. Suppose the product of |A| the line obtained by reflecting L in the
Problem 379. Let ℓ be a line on the and |B| is at least 3999. Then prove that lines BC, CA and AB, respectively.
plane of ∆ABC such that ℓ does not sets A−A and B−B contain at least one Show that the circumcircle of the
intersect the triangle and none of the common element, where X−X denotes triangle determined by the lines La, Lb
lines AB, BC, CA is perpendicular to ℓ. {s−t : s, t ∈ X and s ≠ t}. and Lc is tangent to the circle γ.
(Source: 2000 Hungarian-Israeli Math
Let A’, B’, C’ be the feet of the Competition)
perpendiculars from A, B, C to ℓ
respectively. Let A’’, B”, C” be the Solution. Maxim BOGDAN (“Mihai
feet of the perpendiculars from A’, B’, Eminescu” National College, Botosani, Remarks on IMO 2011
C’ to lines BC, CA, AB respectively. Romania) and William PENG and Jeff (continued from page 2)
PENG (Dallas,Texas, USA).
Prove that lines A’A”, B’B”, C’C” are The essential point is to observe that
concurrent. Note that the set T={(a,b): a∊A and b∊B} A”B”C” is in fact homothetic to
has |A|×|B| ≥ 3999 elements. Also, the set A’B’C’, with the homothetic center at
Solution. William PENG and Jeff W={a+b: a∊A and b∊B} is a subset of
PENG (Dallas, Texas, USA) and H, a point on γ , i.e. A”B”C” is an
{2,3,…4000}. If W = {2,3,…,4000}, then expansion of A’B’C’ at H by a constant
ZOLBAYAR Shagdar (9th Grade, 2 and 4000 in W imply sets A and B both
Orchlon Cambridge International centre. This implies the circumcircle
contain 1 and 2000. This leads to A−A of A”B”C” is tangent to γ at H.
School, Mongolia). and B−B both contain 1999.
C" B
A A lot of discussions were conducted
If W≠{2,3,…4000}, then W has less than concerning changing the format of the
3999 elements. By the pigeonhole Jury system during the IMO. At
C principle, there would exist (a,b) ≠ (a’,b’) present the leaders assemble to choose
B" in T such that a+b=a’+b’. This leads to six problems from the short-listed
l a−a’=b’−b in both A−A and B−B. problems. There are issues concerning
A' B'
C'
security and also financial matter (to
A" house the leaders in an obscure place
D
Problem 3. Comparing with problem 6, caused more trouble because of the Extend AX to meet the circumcircle of
I really found this problem harder to disputes about the marking scheme. First, ABC at A’, likewise extend BX to meet
approach! Nevertheless there were still by putting a=b=c=0, one gets f(0)=0. By the circle at B’. Now extend AB’ and
8 contestants who completely solved putting b=−a and c=0, one gets f(a) = BA’ to meet at H, which is exactly the
the problem. Among them three were f(−a). More importantly, by putting orthocentre of ABX and it lies on the
from the US team. That was an c=−(a+b) and solving f(a+b) = f(−(a+b)) extension of DC.
amazing achievement! as a quadratic equation of f(a) and f(b),
one gets Since BK2 = BC2 = BD·BA, we have
We can deal with this combinatorial ∆ABK~∆KBD, so ∠BKD =∠BAK =
f (a + b) = f (a ) + f (b) ± 2 f (a) f (b). ∠ BHD, which implies B, D, K, H
probabilistic problem as follows. Ask
Putting a=b and c=−2a into the original concyclic. So ∠BKH =∠BDH=90°.
repeatedly if x is 2k. If A answers no
k+1 times in a row, then the answer is
equation, one gets f(2a)=0 or f(2a)=4f(a). This implies HK2 = BH2 −BK2 = BH2−
honest and x≠2k. Otherwise B stops
Now the problem becomes getting all BD·BA = BH2 − BA’·BH = HA’·HB.
asking about 2k at the first time answer
possible solutions from these two Similarly HL2= HB’·HA. But HA’·HB
relations. Using the two conditions, one = HB’·HA. Hence HK=HL. Using
yes. He then asks, for each i=1,2,…,k,
checks that there are four types of similar arguments as above, we have
if the binary representation of x has a 0
solution: ∠ALH = 90° ( =∠BKH.) Along with
in the i-th digit. Whatever the answer is,
they are all inconsistent with a certain HK=HL, we see ∆MKH≅∆MLH.
(i) f1(x)≡0 , (ii) f2(x) = kx2 , Therefore, MK=ML.
number y in the set {0,1,2,⋯, 2k−1}.
The answer yes to 2k is also (iii) f ( x) = ⎧⎨0, x even and Problem 6. Clearing denominators of
inconsistent with y. Hence x≠y. 3
⎩ k , x odd
Otherwise the last k+1 answers are not 1 2 n
honest and that is impossible. So we ⎧ 0, x ≡ 0 (mod 4) a1
+ a2 + ... + an = 1,
find y and it can be eliminated. Or we (iv) ⎪ 3 3 3
f 4 ( x) = ⎨ k , x ≡ ±1(mod 4) .
can eliminate corresponding numbers ⎪ 4k , x ≡ 2 (mod 4) one gets x1+2x2+⋯+nxn=3a, where
⎩
with nonzero digits at higher end. x1,x2,…,xn are non-negative integer
Notice we may need to do some The “k” in the solutions is essentially f(1). powers of 3. Taking mod 2, one gets
re-indexing and asking more questions Indeed if f(1)=0, then f(2)=0, one then n(n+1)/2 ≡ 1(mod 2). This is the case
about the indices of the numbers show by induction f(x)=0 for all x. (Or by only when n ≡ 1, 2 (mod 4). The hard
subsequently. With these questions, we showing f(x) is periodic of period 1.) Now part is to prove the converse also holds.
can reduce the size of the set that x lies if f(1)=k, using the condition f(2a)=0, one The cases n=1 or 2 are easy. By trials,
until it lies in a set of size 2k. can show again by induction f(x) is k for x for n=5, (a1,…,a5)=(2,2,2,3,3) works.
odd and is 0 for x even. Now if f(1)=k and The official solution gave a systematic
Part 2 makes use of a function so that f(2)=4k, then f(4)=0 or 16k. In the first analysis of how to obtain solutions by
using the function, A can devise a case we get a function with period 4 and using identities 1/2a=1/2a+1+1/2a+1 and
strategy (to lie or not to lie, but lying arrive at the solution f4(x). In the second w/3a=u/3a+1+v/3a+1, where u+v=3w.
not more than k times consecutively) so case we get f2(x). (One needs to verify the For n=4k+1≥5, one can arrive at the
that no extra information will be details.) By checking the values of a, b solution a1=2=a3, a2=k+1, a4k = k+2 =
provided to B and hence B cannot and c mod 2 or 4, or other possible forms, a4k+1 and am = [m/4]+3 for 4≤m<4k.
eliminate anything for sure. Due to one can check the solutions are indeed Similarly, for n=4k+2≥6, one can
limit of space, I cannot provide all valid. Eventually if a contestant claimed arrive at the solution a1 = 2, a2 = k+1, a3
details here. that all the solutions are easy to check, but = a4 = 3, a4k+1 = k+2 = a4k+2 and am =
without checking, one point would be [(m−1)/4]+3 for 4<m≤4k. One can
It was decided that part 1 answered deducted. If a contestant says nothing check these are indeed solutions by
correctly alone was worth 3 points and about the solutions satisfy the functional math induction on k. In the inductive
part 2 alone worth 5 points. But equation and check nothing, then two steps of both cases, just notice a2, an−1,
altogether a problem is worth at most 7 points would be deducted! an are increased by 1 so to balance the
points. So 3 + 5 = 7! At the end it really new an+1, an+2, an+3, an+4 terms.
H
did not matter. After all, not too many
This reminds me of the 1978 USAMO
students did the problem right.
C A' problem: an integer n is called good if
B' we can write n=a1+a2+⋯+ak, where
The problem is noted to be related to X a1,a2,…,ak are positive integers (not
the Lovasz Local Lemma. See N. Alon
necessarily distinct) satisfying
et al, The Probabilistic Methods, Wiley, K L
M
1992. In the book it seems that there is 1 1 1
A B
+ +L+ = 1.
an example that deals with similar D a1 a2 ak
things. One may check how the lemma
Given 33 to 73 are good, prove that all
and the problem are related!
integer greater than 33 are good. The
idea there is to show if n is good, then
Problem 4. Despite being regarded as
2n+8 and 2n+9 are good by dividing
an easy problem, this problem is not at
both sides of the above equation by 2
all easy. It is much more involved than
Problem 5. The following solution was and adding the terms 1/4+1/4 and
expected. Also this problem eventually
obtained by one of our team members. 1/3+1/6 respectively.
Mathematical Excalibur, Vol. 17, No. 1, May-Aug. 12 Page 3
Other commended solvers: Simon LEE Therefore, the answer is n ≡ 2 (mod 4).
F
A (Carmel Alison Lam Foundation Secondary
School), and Alice WONG (Diocesan Other commended solvers: Alice
Girls’ School). WONG (Diocesan Girls’ School).
G
Problem 395. One frog is placed on every
O
vertex of a 2n-sided regular polygon,
E H where n is an integer at least 2. At a Olympiad Corner
M particular moment, each frog will jump to
B C (continued from page 1)
one of the two neighboring vertices (with
D X more than one frog at a vertex allowed). Problem 3. (Cont.) At the start of the
As AD bisects ∠BAC, D is the game A chooses integers x and N with
Find all n such that there exists a jumping 1≤x≤N. Player A keeps x secret, and
midpoint of arc BC. Hence, FD is the
of these frogs so that after the moment, all truthfully tells N to B. Player B now
perpendicular bisector of BC. Thus, (1)
lines connecting two frogs at different tries to obtain information about x by
FE || AH. Let line AH meet Γ again at X.
vertices do not pass through the center of asking player A questions as follows:
Since
the polygon. each question consists of B specifying
∠BCX=∠BAX=90°−∠ABC=∠BCH,
an arbitrary set S of positive integers
H is the mirror image of X with respect Solution. Kevin LAU (St. Paul’s
Co-educational College, S.3), Simon LEE (possibly one specified in some
to BC. Therefore, ∠HED=∠XDE= (Carmel Alison Lam Foundation Secondary previous question), and asking A
∠AFE. Thus, (2) AF || HE. By (1) and School), LI Jianhui (CNEC Christian whether x belongs to S. Player B may
(2), AFEH is a parallelogram. Hence, College, F.5) and Andy LOO (St. Paul’s ask as many such questions as he
G is the midpoint of AE. As M is also Co-educational College). wishes. After each question, player A
the midpoint of DE, we get GM || AD. If n ≡ 2 (mod 4), say n=4k+2, then label must immediately answer it with yes or
Since DF is the diameter of Γ, AD⊥AF, the 2n=8k+4 vertices from 1 to 8k+4 in no, but is allowed to lie as many times
hence GM⊥AF. clockwise direction. For j ≡ 1 or 2 (mod 4), as she wants; the only restriction is that,
let the frog at vertex j jump in the among any k + 1 consecutive answers,
Solution 2. Andy LOO (St. Paul’s
clockwise direction. For j ≡ 3 or 4 (mod 4), at least one answer must be truthful.
Co-educational College).
let the frog at vertex j jump in the After B has asked as many questions
Place the figure on the complex plane counter-clockwise direction. After the as he wants, he must specify a set X of
and let the circumcircle of ΔABC be jump, the frogs are at vertices 2, 6, …, at most n positive integers. If x belongs
the unit circle centered at the origin. 8k+2 and 3,7, …, 8k+3. No two of these to X, then B wins; otherwise, he loses.
Denote the complex number vertex numbers have a difference of the Prove that:
representing each point by the form 2 (mod 4). So no line through two
respective lower-case letter. Without different vertices with frogs will go 1. If n ≥2k, then B can guarantee a win.
loss of generality we may assume a = 1 through the center. 2. For all sufficiently large k, there
and that the points A, B and C lie on the exists an integer n ≥1.99k such that B
circle in anticlockwise order. Let b =u2 If n ≢2 (mod 4), then assume there is such cannot guarantee a win.
and c=v2, where |u|=|v|=1. Then d=uv a jump. We may exclude the cases all
and hence f =−uv. Next, E is the mirror frogs jump clockwise or all frogs jump Problem 4. Find all functions f: Z → Z
image of D with respect to BC means counter-clockwise, which clearly do not such that, for all integers a, b, c that
work. Hence, in this jump, there is a frog, satisfy a+b+c=0, the following
e−b ⎛ d −b ⎞ ,
=⎜ ⎟ equality holds:
c−b ⎝ c−b ⎠ say at vertex i, jumps in the
counter-clockwise direction, then the frog f (a)2 + f (b)2 + f (c)2
giving e = u2 − uv + v2. By the Euler at vertex i+m(n−2) (mod 2n) must jump
line theorem, h=a+b+c=1+u2+v2.Now = 2f (a) f (b) + 2f (b) f (c) + 2f (c) f (a).
in the same direction as the frog at vertex i
G on lines AE and FH means for m=1,2,…. (Here Z denotes the set of integers.)
g − a g − a and g − f g − f .
= = Problem 5. Let ABC be a triangle with
e−a e −a h− f h − f If n is odd, then gcd(n−2,2n) = 1. So there
∠BCA = 90°, and let D be the foot of
are integers a and b such that a(n−2) +
Solving these simultaneously for G, we the altitude from C. Let X be a point in
b(2n) = 1. For every integer q in [1,2n],
get g = (u2−uv+v2+1)/2. Also, m = letting m = (q−i)a, we have i+m(n−2) ≡ q
the interior of the segment CD. Let K
(b+c)/2 = (u2+v2)/2. be the point on the segment AX such
(mod 2n). This means all frogs jump in the
that BK=BC. Similarly, let L be the
counter-clockwise direction, which does
To show GM⊥AF, it suffices to prove point on the segment BX such that
not work.
AL=AC. Let M be the point of
that (m−g)/(f−a) is an imaginary
intersection of AL and BK. Show that
If n is divisible by 4, then gcd(n−2,2n) = 2.
number. Indeed, m − g = 1 ⋅ 1 − uv and MK=ML.
So there are integers c and d such that
f −a 2 1 + uv c(n−2)+d(2n)=2. Letting m=nc/2, we Problem 6. Find all positive integers n
1 1 have i+m(n−2)≡i+n (mod 2n). Then frogs for which there exist non-negative
⎛ m − g ⎞ 1 1 − u ⋅ v 1 uv − 1 m−g at vertices i and i+n jump in the
⎜ ⎟= ⋅ = ⋅ =− integers a1,a2,…,an such that
⎝ f − a ⎠ 2 1 + 1 ⋅ 1 2 uv + 1 f −a
u v counter-clockwise direction and the line
1 1 1 1 2 n
as desired. after the jump passes through the center, a1
+ a2 + ... + an = a1 + a2 + ... + an = 1.
contradiction. 2 2 2 3 3 3
Volume 17, Number 2 October 2012
July 10 On this first day of the contest, and marked the original as an alternative then aim to get Problem 5 by hook or
we had 3 problems to solve in 4.5 hours. solution. It turned out that Alice was also by crook.
Because questions could only be raised able to solve this problem with the
in the first 30 minutes, I had to AM-GM inequality. To my astonishment, Problem 4 was
understand all the problems quickly. much more involved than I had
Problem 3 The liar’s guessing game is a expected. On the other hand I felt I
Problem 1 Given triangle ABC the game played between two players A and B. could do Problem 5 with analytic tools:
point J is the center of the excircle The rules of the game depend on two
opposite the vertex A. This excircle is positive integers k and n which are known Problem 5 Let ABC be a triangle with
tangent to the side BC at M, and to the to both players. ∠BCA = 90°, and let D be the foot of
lines AB and AC at K and L, the altitude from C. Let X be a point in
respectively. The lines LM and BJ meet At the start of the game A chooses integers the interior of the segment CD. Let K
at F, and the lines KM and CJ meet at G. x and N with 1≤x≤N. Player A keeps x be the point on the segment AX such
Let S be the point of intersection of the secret, and truthfully tells N to B. Player B that BK=BC. Let L be the point on the
lines AF and BC, and let T be the point now tries to obtain information about x by segment BX such that AL=AC. Let M
asking player A questions as follows: each be the point of intersection of AL and
of intersection of the lines AG and BC. BK. Show that MK=ML.
Prove that M is the midpoint of ST. question consists of B specifying an
arbitrary set S of positive integers I proceeded to do coordinate geometry,
I decided to use my favorite method – (possibly one specified in some previous only to find out I was doomed after
complex numbers. Indeed, denote the question), and asking A whether x belongs
almost one hour. The reason was as
complex number representing each to S. Player B may ask as many such
questions as he wishes. After each follows. The expressions were
point by the corresponding small letter. quadratic in nature (as lengths took part
Setting j=0 and m=1, I found s=2k/(k+l) question, player A must immediately
answer it with yes or no, but is allowed to in the formulation of the problem),
and t=2l/(k+l) after a straightforward leading to the prevalence of square
lie as many times as she wants; the only
computation, and the result followed. roots. (As a side note, this also deterred
restriction is that, among any k+1
Problem 2 Let n ≥ 3 be an integer, and consecutive answers, at least one answer me from using complex numbers,
let a2,a3,…,an be positive real numbers must be truthful. where one may have difficulty in
such that a2a3⋯an =1. Prove that After B has asked as many questions as he
selecting the correct roots of the
quadratic equations.)
(1+a2)2(1+a3)3⋯(1+an)n > nn. wants, he must specify a set X of at most n
positive integers. If x belongs to X, then B As the old Chinese saying goes, one
Inequalities were once among the wins; otherwise, he loses. Prove that:
should “drop his cleaver and become a
hottest topics on the IMO but totally
1. If n ≥ 2k, then B can guarantee a win. Buddha (放下屠刀,立地成佛)”. I
disappeared in the last three years due
2. For all sufficiently large k, there exists decided to abandon Problem 5 for a
to the rising popularity of brute force
an integer n ≥1.99k such that B cannot moment and to reconsider Problem 4:
techniques, e.g. Muirhead’s inequality
guarantee a win.
and Schur’s inequality. But my firm Problem 4 Find all functions f: Z → Z
belief in the revival of inequalities has This problem was not only long, but also such that, for all integers a, b, c that
never been shaken, and instead was satisfy a + b + c = 0, the following
terribly difficult. In the end, only 8 equality holds:
only strengthened by Problem 5 of contestants managed to solve it. Despite
APMO 2012. Consequently I had done my effort, the only thing I was able to do f (a)2 + f (b)2 + f (c)2
appreciable preparation in this area was proving the k = 1 case in Part 1, with = 2f (a) f (b) + 2f (b) f (c) + 2f (c) f (a).
before the Olympiad. the hope of getting slim partial credits. (Here Z denotes the set of integers.)
In IMO history, this problem was quite Finally Day 1 of the contest was over. This was a problem with unusual
unique. For one, it is an n-variable Our team aced Problem 1. As for Problem answers. It took me quite a while to
inequality. For the other, it has no 2, Alice and I should be able to get 7’s write up a tidy solution and to ensure
equality case. Both features are while Albert’s partial analytic solution that no point could sneak away from
unparalleled according to my memory. would be subject to vigorous debate. Kit my hands. Thus it was 2.5 hours into
also finished the k = 1 case in Part 1 of Day 2. I still had Problems 5 and 6 left.
I spent about an hour attempting to Problem 3. Overall I was satisfied with
solve the problem using induction or my performance on Day 1. Problem 6 Find all positive integers n
analysis, with no avail. In despair, I for which there exist non-negative
took logarithm and applied Jensen’s July 11 The six IMO problems are integers a1,a2,…,an such that
inequality by appealing to concavity of usually partitioned into the four categories
1 1 1 1 2 n
the log function. Miraculously, it gave (algebra, combinatorics, geometry and + + ... + an = a1 + a2 + ... + an = 1.
precisely the inequality in the problem! 2a1 2a2 2 3 3 3
number theory) in the fashion of {1,5},
After checking that equality case {2,4}, {3} and {6} (up to permutation).
cannot satisfy the condition a2a3⋯an Judging from this pattern I would face an I quickly determined that Problem 6
=1, I was basically done. easy algebraic problem, an intermediate was hopeless. Turning to Problem 5
geometric problem and a hard number again, I spent all the remaining time
Then on a second thought, I realized theoretic problem on Day 2. I figured that expanding everything. I was finally
that I could actually convert my proof I would plausibly get a Gold medal for able to convince myself that my proof
into a logarithm-free one that involves solving two of them, a Silver medal for was complete.
the AM-GM inequality only. So I one and a Bronze medal for none. My
rewrote my solution in this new form (continued on page 4)
strategy was to guarantee Problem 4 and
Mathematical Excalibur, Vol. 17, No. 2, Oct. 12 Page 3
Problem 401. Suppose all faces of a Putting x=−y in (*), we get f(f(y)) = Problem 398. Let k be positive integer
convex polyhedron are parallelograms. (f(−y))2−yf(y)+y for all y. Call this (***). and m an odd integer. Show that there
Can it have exactly 2012 faces? Please exists a positive integer n for which the
Now f surjective implies there exists z
provide an explanation to your answer.
such that f(z)=0. Let x=y=z, then (*) number nn−m is divisible by 2k.
Problem 402. Let S be a 30 element yields f(2z2)=z. Putting (x,y)=(0,2z2) in (*), Solution. AN-anduud Problem
subset of {1,2,…,2012} such that every we get 0=2z2+f(0)2. Then z=0 and f(0)=0. Solving Group (Ulaanbaatar,
pair of elements in S are relatively So (**) reduces to f(f(y))=y for all y. Mongolia), Dusan DROBNJAK
Putting y = 0 in (*), since f(0) = 0, we get (Mathematical Grammar School,
prime. Prove that at least half of the Belgrade, Serbia), KWAN Chung
elements of S are prime numbers. f(x2) = (f(x))2. The last two sentences
Hang (Sir Ellis Kadoorie Secondary
reduce (***) to y = (f(y))2−yf(y)+y. This School (West Kowloon)), Kevin LAU
Problem 403. On the coordinate plane, simplifies to f(y) = 0 or f(y) = y for every y. (St. Paul’s Co-educational College,
1000 points are randomly chosen. Since f is injective and f(0) = 0, we get f(y) S.3), Simon LEE (Carmel Alison Lam
Prove that there exists a way of = y for all y. Conversely, a quick check Foundation Secondary School), Andy
coloring each of the points either red or shows f(y) = y for all y satisfies (*). LOO (Princeton University), Tobi
MOEKTIJONO (National University
blue (but not both) so that on every line of Singapore) and Maksim STOKIĆ
Other commended solvers: Tobi
parallel to the x-axis or y-axis, the MOEKTIJONO (National University of (Mathematical Grammar School,
number of red points minus the number Singapore). Belgrade, Serbia).
of blue points is equal to −1, 0 or 1. For k=1, let n=1. Suppose it is true for
Problem 397. Suppose in some set of 133
Problem 404. Let I be the incenter of case k (i.e. there exists n such that 2k |
distinct positive integers, there are at least
acute ∆ABC. Let Γ be a circle with nn−m). Now m odd implies n odd. For
799 pairs of relatively prime integers.
center I that lies inside ∆ABC. D, E, F case k+1, if 2k+1 | nn−m, then the same n
Prove that there exist a,b,c,d in the set
are the intersection points of circle Γ works for k+1. Otherwise, nn−m=2kl
such that gcd(a,b) = gcd(b,c) = gcd(c,d) =
with the perpendicular rays from I to for some odd integer l. Let v=2k. By
gcd(d,a) = 1.
sides BC, CA, AB respectively. Prove binomial theorem,
Solution. CHEUNG Ka Wai (Munsang
that lines AD, BE, CF are concurrent. College (Hong Kong Island)), Dusan (n+v)n+v = nn+v+(n+v)nn+v−1v+v2x
DROBNJAK (Mathematical Grammar = nn+v+vnn+v+v2y
Problem 405. Determine all functions School, Belgrade, Serbia), Kevin LAU
f,g: (0,+∞) → (0,+∞) such that for all (St. Paul’s Co-educational College, S.4), for some integers x,y. By Euler’s
positive number x, we have Simon LEE (Carmel Alison Lam theorem, since n is odd and φ(2k+1)=2k,
Foundation Secondary School), Andy k
x x LOO (Princeton University), Tobi n v = n 2 ≡ 1 (mod 2 k +1 ).
f ( g( x)) = and g( f ( x)) = .
xf ( x) − 2 xg( x) − 2 MOEKTIJONO (National University of Since l+nn is even, we have
Singapore) and Maksim STOKIĆ
(Mathematical Grammar School, (n+v)n+v = nn+v+vnn+v+v2y
***************** Belgrade, Serbia). ≡ nn+2knn = m+2k(l+nn)
Solutions ≡ m (mod 2k+1).
Let S={n1, n2,…,n133} be the set of these 133
****************
positive integers. From i=1 to 133, let Xi be So n+v works for k+1.
Problem 396. Determine (with proof) the set of all nk in S such that k≠i and
all functions f : ℝ→ℝ such that for all gcd(ni,nk)=1. Denote by |X| the number of Problem 399. Let ABC be a triangle
real numbers x and y, we have elements in set X. For k≠i, gcd(ni,nk)=1 for which ∠BAC=60°. Let P be the
implies ni∈Xk and nk∈Xi. Then N = |X1| + point of intersection of the bisector of
f (x2 + xy + f (y)) = (f(x))2 + xf(y) + y. |X2| + ⋯ + |X133| ≥ 2×799 = 1598. ∠ABC and the side AC. Let Q be the
Solution. AN-anduud Problem point of intersection of the bisector of
Define f(x) = x(x−1)/2. In a set X with j
Solving Group (Ulaanbaatar, ∠ACB and the side AB. Let r1 and r2 be
elements, there are exactly j(j−1)/2 = f(|X|)
Mongolia), CHEUNG Ka Wai the radii of the incircles of triangles
(Munsang College (Hong Kong pairs of distinct elements. Since f(x) is
ABC and APQ respectively. Find the
Island)), CHEUNG Wai Lam (Queen concave on ℝ, by Jensen’s inequality,
radius of the circumcircle of triangle
Elizabeth School), Dusan
APQ in terms of r1 and r2 with proof.
Mathematical Excalibur, Vol. 17, No. 2, Oct. 12 Page 4
Solution. Dusan DROBNJAK Letting M be the product of the July 13 The marking scheme
(Mathematical Grammar School, denominators, we see Q(x)=MP(x) has stipulated that any solution of Problem
Belgrade, Serbia), Kevin LAU (St.
Paul’s Co-educational College, S.4), integer coefficients. Let k be the leading 5 with coordinate geometry would
Andy LOO (Princeton University), coefficient of Q(x) and c be the constant score a 0 if not a 7. Despite our leaders’
MANOLOUDIS Apostolos (4° Lyk. term of P(x). Let p1, p2, p3, … be the relentless effort, the coordinators were
Korydallos, Piraeus, Greece), Tobi sequence of prime numbers. Let able to detect a fatal error of mine. So
MOEKTIJONO (National University P(x)=c+pi/M has solution ti∈ℚ. Then my Problem 5 was destined to be a 0.
of Singapore) and Maksim STOKIĆ
(Mathematical Grammar School, Q(x)−(cM+pi) has k as the leading
coefficient and −pi as constant term. Now On another note, Dr. Leung succeeded
Belgrade, Serbia).
Q(ti)=0, which implies ti=1/di or pi∕di for in getting 1 point for Alice on Problem
A
F some (not necessarily positive) divisor of 4, which in his words was “an
E k. Since P(ti)’s are distinct, so the ti’s are achievement”. Kevin’s Problem 4 was
S P finalized with a score of 4.
Q distinct. Hence, ti=1/di for at most as
many time as the number of divisors of k. July 14 We got up early in the morning
I So there must exist a divisor d of k such to enjoy the sunrise scene at the seaside.
B C that there are infinitely many times ti=pi∕d. Kevin had a pitiful blunder. His shoes
This imply that P(x)−(c+dx/M)=0 has and trousers were wetted by a sudden
Let I and S be the incenters of ΔABC infinitely many solutions. So the left side
and ΔAPQ respectively. (Note A,S,I strike of waves. That morning the last
is the zero polynomial. Then P(x)=ax+b coordination on Problem 2 was done.
are on the bisector of ∠BAC.) Now with a=d/M≠0 and b=c rational.
∠PIQ =∠CIB = 180°−(∠CBI +∠BCI) Albert was awarded 3 marks for his
= 180°− ½(∠CBA+∠BCA) = 120° Other commended solvers: Simon LEE analytic struggle. The uncertainties of
(Carmel Alison Lam Foundation Secondary our results then shifted from our actual
using ∠BAC=60°. So APIQ is cyclic. School). scores to the medal cutting scores.
Applying sine law to ΔAPI, we get
IP/(sin∠IAP) = 2R. So R = IP. By a We went shopping for souvenirs in the
IMO 2012 (Member Perspective) afternoon and as soon as we got back to
well-known property of incenter, we
have IP=IS (see vol.11, no.2, p.1 of (continued from page 2) the hotel, I learned from the Chinese
Math Excal.). Let the incircles of leaders that the cutting scores for Gold,
The arrival of Dr. Leung stirred up much Silver and Bronze Medals were 28, 21
ΔABC and ΔAPQ touch AC at E and F happiness after the contest. We reported
respectively. Then R=IP=IS = AI−AS and 14 respectively, all being multiples
on how we did. Albert and Jimmy shone of 7. I breathed a sigh of relief as my
= IE/sin30°−SF/sin30° = 2r1−2r2. on Day 2, solving Problems 4 and 5. Kit
Silver Medal was ultimately secure.
Other commended solvers: was also comfortable with Problem 4
AN-anduud Problem Solving Group while Kevin had some technical troubles July 15 In the afternoon we had the
(Ulaanbaatar, Mongolia), Ioan Viorel in one particular case. Nobody achieved Closing Ceremony followed by a chain
CODREANU (Secondary School anything substantial on Problem 6. of photo-taking. We won three Silver
Satulung, Maramure, Romania), Simon
LEE (Carmel Alison Lam Foundation We celebrated that evening at a Chinese Medals (Albert, Jimmy and me), one
Sec. School) and Mihai STOENESCU restaurant. It was especially memorable Bronze Medal (Alice) and two
(Bischwiller, France). that our deputy leader raised a couplet (對 Honorable Mentions (Kit and Kevin).
聯), which he regarded as an open puzzle July 16, 17 and 18 The six-hour bus
Problem 400. Determine (with proof)
for millenniums (千古絕對): journey from Mar del Plata to Buenos
all the polynomials P(x) with real
coefficients such that for every rational 望江樓,望江流 Aires passed rapidly in our dreams.
number r, the equation P(x) = r has a 望江樓上望江流 Then after a long flight, we were
rational solution. 江樓千古,江流千古 finally home in one piece and me with
several bonus pimples.
Solution. Tobi MOEKTIJONO It took me nearly an hour to come up with
(National University of Singapore), a so-so solution: In conclusion I shall stress one point –
Maksim STOKIĆ (Mathematical succinctly but with all the strength that
Grammar School, Belgrade, Serbia) 觀雨亭,觀雨停
and TAM Ka Yu (MIT). I command – one can never pay
觀雨亭下觀雨停 sufficient tribute to our IMO trainers,
We will show P(x) satisfies the desired 雨亭四方,雨停四方 who have so selflessly devoted
condition if and only if P(x)=ax+b, countless hours of their own time to
where a,b ∈ℚ and a ≠ 0. For the if-part, July 12 It was the contestants’ turn to
have fun and the leaders’ turn to work Mathematical Olympiad over the years.
P(x) = r ∈ℚ implies x = (r−b)/a ∈ℚ. I can find no words to thank them the
hard. At night, Dr. Leung briefed us on the
Conversely, let P(x) satisfy the desired progress of the first day of coordination. way they truly deserve.
condition and let n=deg P. For each r = In addition to our previous expectations, “Ask not what your country can do for
0,1,…, n, let P(xr)=r for some xr∈ℚ. Albert pocketed one point for proving the you; ask what you can do for your
By the Lagrange interpolation formula, necessary condition on Problem 6. country.” With this John F. Kennedy
n ⎛
x − xi ⎞ Regretfully, Kit lost one point on Problem
P( x) = ∑ ⎜⎜ r ∏ ⎟.
⎟ 4 for not having verified the feasibility of
exclamation I urge you all to support
r = 0 ⎝ 0 ≤ r ≤ n , r ≠ i xr − xi ⎠ the 2016 Hong Kong IMO by whatever
the functions obtained. Dr. Leung had means you can, so that together we can
Expanding the right side, we see P(x) refused to sign Alice’s and Kevin’s scores
has rational coefficients. make it an all-time success.
on Problem 4 in order to bargain later.