Shortlisted Problems and Solutions - International Mathematical ...
Shortlisted Problems and Solutions - International Mathematical ...
Shortlisted Problems and Solutions - International Mathematical ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
51 st <strong>International</strong> <strong>Mathematical</strong> Olympiad<br />
Astana, Kazakhstan 2010<br />
<strong>Shortlisted</strong> <strong>Problems</strong> with <strong>Solutions</strong>
Contents<br />
Note of Confidentiality 5<br />
Contributing Countries & Problem Selection Committee 5<br />
Algebra 7<br />
Problem A1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7<br />
Problem A2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8<br />
Problem A3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10<br />
Problem A4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12<br />
Problem A5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13<br />
Problem A6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15<br />
Problem A7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17<br />
Problem A8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19<br />
Combinatorics 23<br />
Problem C1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23<br />
Problem C2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />
Problem C3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28<br />
Problem C4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />
Problem C4½. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />
Problem C5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32<br />
Problem C6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />
Problem C7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38<br />
Geometry 44<br />
Problem G1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44<br />
Problem G2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46<br />
Problem G3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50<br />
Problem G4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52<br />
Problem G5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54<br />
Problem G6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56<br />
Problem G6½. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56<br />
Problem G7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60<br />
Number Theory 64<br />
Problem N1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64<br />
Problem N1½. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64<br />
Problem N2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66<br />
Problem N3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68<br />
Problem N4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70<br />
Problem N5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71<br />
Problem N6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Note of Confidentiality<br />
The <strong>Shortlisted</strong> <strong>Problems</strong> should be kept<br />
strictly confidential until IMO 2011.<br />
Contributing Countries<br />
The Organizing Committee <strong>and</strong> the Problem Selection Committee of IMO 2010 thank the<br />
following 42 countries for contributing 158 problem proposals.<br />
Armenia, Australia, Austria, Bulgaria, Canada, Columbia, Croatia,<br />
Cyprus, Estonia, Finl<strong>and</strong>, France, Georgia, Germany, Greece,<br />
Hong Kong, Hungary, India, Indonesia, Iran, Irel<strong>and</strong>, Japan,<br />
Korea (North), Korea (South), Luxembourg, Mongolia, Netherl<strong>and</strong>s,<br />
Pakistan, Panama, Pol<strong>and</strong>, Romania, Russia, Saudi Arabia, Serbia,<br />
Slovakia, Slovenia, Switzerl<strong>and</strong>, Thail<strong>and</strong>, Turkey, Ukraine,<br />
United Kingdom, United States of America, Uzbekistan<br />
Problem Selection Committee<br />
Yerzhan Baissalov<br />
Ilya Bogdanov<br />
Géza Kós<br />
Nairi Sedrakyan<br />
Damir Yeliussizov<br />
Kuat Yessenov
Algebra<br />
A1. Determine all functions f : RR such that the equality<br />
fÔÖx×yÕfÔxÕÖfÔyÕ×. (1)<br />
holds for all x, yÈR. Here, byÖx×we denote the greatest integer not exceeding x.<br />
Answer. fÔxÕconstC, where C0 or 1C2.<br />
fÔ0ÕfÔ0ÕÖfÔyÕ× (2)<br />
(France)<br />
Solution 1. First, setting x0 in (1) we get<br />
for all yÈR. Now, two cases are possible.<br />
Case 1. Assume that fÔ0Õ0. Then from (2) we conclude thatÖfÔyÕ×1 for all<br />
yÈR. Therefore, equation (1) becomes fÔÖx×yÕfÔxÕ, <strong>and</strong> substituting y0we have<br />
fÔxÕfÔ0ÕC0. Finally, fromÖfÔyÕ×1ÖC×we obtain that 1C2.<br />
Case 2. Now we have fÔ0Õ0. Here we consider two subcases.<br />
NÖz×<br />
Subcase 2a. Suppose that there exists 0α1 such that fÔαÕ0. Then setting xα<br />
in (1) we obtain 0fÔ0ÕfÔαÕÖfÔyÕ×for all yÈR. Hence,ÖfÔyÕ×0 for all yÈR. Finally,<br />
substituting x1in (1) provides fÔyÕ0for all yÈR, thus contradicting the condition<br />
fÔαÕ0.<br />
Subcase 2b. Conversely, we have fÔαÕ0for all 0α1. Consider any real z; there<br />
exists an integer N such that αz<br />
1Õ(one may set 1 if z0<strong>and</strong><br />
NÈÖ0, NÖzס1<br />
otherwise). Now, from (1) we get fÔzÕfÔÖN×αÕfÔNÕÖfÔαÕ×0 for all zÈR.<br />
Finally, a straightforward check shows that all the<br />
fÔxÕfÔÖx×aÕ<br />
obtained functions satisfy (1).<br />
Solution 2. Assume thatÖfÔyÕ×0 for some y; then the substitution x1 provides<br />
fÔyÕfÔ1ÕÖfÔyÕ×0. Hence, ifÖfÔyÕ×0for all y, then fÔyÕ0for all y. This function<br />
obviously satisfies the problem conditions.<br />
So we are left to consider the case whenÖfÔaÕ×0 for some a. Then we have<br />
fÔÖx×aÕfÔxÕÖfÔaÕ×, or<br />
This means that fÔx 1ÕfÔx 2ÕwheneverÖx fÔxÕfÔ0Õ 1×Öx 2×, hence fÔxÕfÔÖx×Õ, <strong>and</strong> we may assume<br />
that a is an integer.<br />
Now we have<br />
fÔaÕf 2a¤1<br />
2¨fÔ2aÕf<br />
1<br />
2¨fÔ2aÕÖfÔ0Õ×;<br />
this impliesÖfÔ0Õ×0, so we may even assume that a0. Therefore equation (3) provides<br />
ÖfÔ0Õ×C0<br />
for each x. Now, condition (1) becomes equivalent to the equation CCÖC×which holds<br />
exactly whenÖC×1.<br />
ÖfÔaÕ×. (3)
8<br />
A2. Let the real numbers a, b, c, d satisfy the relations a b c d6 <strong>and</strong> a 2 b 2 c 2 d 212.<br />
Prove that<br />
364Ôa 3 b 3 c 3 d 3Õ¡Ôa 4 b 4 c 4 d 4Õ48.<br />
¡ 4Õ¡ Ôd¡1Õ4¨ dÕ<br />
(Ukraine)<br />
Solution 1. Observe that<br />
4Ôa 3 b 3 c 3 d 3Õ¡Ôa 4 b 4 c 4 d Ôa¡1Õ4 6Ôa 2 b 2 c 2 d 2Õ¡4Ôa b c 4<br />
Ôa¡1Õ4 Ôb¡1Õ4 Ôc¡1Õ4 52.<br />
dÕ<br />
Now, introducing xa¡1, yb¡1, zc¡1, td¡1, we need to prove the inequalities<br />
16x 4 y 4 z 4 t 44,<br />
under the constraint<br />
x 2 y 2 z 2 t 2Ôa 2 b 2 c 2 d 2Õ¡2Ôa b c 44 (1)<br />
(we will not use the value of x y z t though it can be found).<br />
Now the rightmost inequality in (1) follows from the power 4Õ mean inequality:<br />
x 4 y 4 z 4 t 4Ôx2 y 2 z 2 t 2Õ2<br />
4.<br />
4<br />
For the other one, exp<strong>and</strong>ing the brackets we note that<br />
Ôx 2 y 2 z 2 t 2Õ2Ôx 4 y 4 z 4 t q,<br />
where q is a nonnegative number, so<br />
<strong>and</strong> we are done.<br />
x 4 y 4 z 4 t 4Ôx 2<br />
Ôb¡1Õ4 Ôc¡1Õ4 Ôd¡1Õ4¨<br />
y 2 z 2 t 2Õ216,<br />
Comment 1. The estimates are sharp; the lower <strong>and</strong> upper bounds are attained atÔ3,1,1,1Õ<strong>and</strong><br />
Ô0,2,2,2Õ, respectively.<br />
Comment 2. After the change of variables, one can finish the solution in several different ways.<br />
The latter estimate, for instance, it can be performed by moving the variables – since we need only<br />
the second of the two shifted conditions.<br />
Solution 2. First, we claim that 0a, b, c, d3. Actually, we have<br />
a b c6¡d, a 2 b 2 c 212¡d 2 ,<br />
hence the power mean inequality<br />
a 2 b 2 c 2Ôa b cÕ2<br />
3<br />
rewrites as<br />
2dÔd¡3Õ0,<br />
12¡d 2Ô6¡dÕ2<br />
3
which implies the desired inequalities for d; since the conditions are symmetric, we also have<br />
the same estimate for the other variables.<br />
Now, to prove the rightmost inequality, we use the obvious inequality x 2Ôx¡2Õ20for<br />
each real 4Õ 4Õ x; this inequality 4Õ rewrites 4Õ as 4x 3¡x44x2 . It follows 2Õ that<br />
Ô4a 3¡a 2 b 2 c 2 d 2Õ48,<br />
as desired.<br />
Now we prove the leftmost inequality in an analogous way. For each xÈÖ0, 3×, we have<br />
Ôx 1ÕÔx¡1Õ2Ôx¡3Õ0 which is equivalent to 4x 3¡x42x2 4x¡3. This implies that<br />
Ô4a 3¡a Ô4b 3¡b Ô4c 3¡c Ô4d 3¡d 4Õ2Ôa 2 b 2 c 2 d 4Ôa b c dÕ¡1236,<br />
as desired.<br />
4Õ Ô4b 3¡b<br />
4Õ Ô4c 3¡c<br />
Ô4d 3¡d 4Õ4Ôa<br />
Comment. It is easy to guess the extremal pointsÔ0,2,2,2Õ<strong>and</strong>Ô3,1,1,1Õfor this inequality. This<br />
provides a method of finding the polynomials used in Solution 2. Namely, these polynomials should<br />
have the form x 4¡4x3 ax 2 bx c; moreover, the former polynomial should have roots at 2 (with<br />
an even multiplicity) <strong>and</strong> 0, while the latter should have roots at 1 (with an even multiplicity) <strong>and</strong> 3.<br />
These conditions determine the polynomials uniquely.<br />
Solution 3. First, exp<strong>and</strong>ing 484Ôa 3 2Õ 2 b 3 2 c 3 2Õ 2Õ 2© 2Õ<br />
2 d 2Õ<strong>and</strong> applying the AM–GM inequality,<br />
we have<br />
a 4 b 4 c 4 d 4 48Ôa 4 4a Ôb 4 4b Ôc 4 4c Ôd 4 4d<br />
2¡a 4¤4a b 2 4¤4b c 2 4¤4c d 2 4¤4d<br />
4Ôa b c 3 b 3 c 3 d 3Õ,<br />
c<br />
which establishes the rightmost inequality.<br />
To prove the leftmost inequality, we first show that a, b, c, dÈÖ0, 3×as in the previous<br />
solution. Moreover, we can assume that 0abcd. Then we have a bb 2<br />
3Ôb c dÕ2<br />
3¤64.<br />
Next, we show 2Õ that 4b¡b24c¡c2 . Actually, this inequality rewrites asÔc¡bÕÔb c¡4Õ0,<br />
which follows from the previous estimate. The inequality 4a¡a24b¡b2 can be proved<br />
analogously.<br />
Further, the inequalities abctogether with 4a¡a24b¡b24c¡c2 allow us to<br />
apply the Chebyshev inequality obtaining<br />
a 2Ô4a¡a 4Õ 2Õ b 2Ô4b¡b 4Õ c<br />
2Ô4c¡c 4Õ 2Õ 2ÕÕ<br />
2Õ¨<br />
2Õ1<br />
3Ôa 2 b 2 c 4Ôa b cÕ¡Ôa 2 b 2 c<br />
Ô12¡d 2ÕÔ4Ô6¡dÕ¡Ô12¡d .<br />
3<br />
This implies that<br />
12Õ<br />
Ô4a 3¡a Ô4b 3¡b Ô4c 3¡c Ô4d 3¡d 4ÕÔ12¡d 2ÕÔd2¡4d 4d 3¡d 4<br />
3<br />
144¡48d 16d 3¡4d4<br />
4<br />
36<br />
3Ô3¡dÕÔd¡1ÕÔd 2¡3Õ. (2)<br />
3<br />
Finally, we have d 21<br />
4Ôa2 b 2 c 2 d 2Õ3(which implies d1); so, the expression<br />
3Ô3¡dÕÔd¡1ÕÔd 4<br />
2¡3Õin the right-h<strong>and</strong> part of (2) is nonnegative, <strong>and</strong> the desired inequality<br />
is proved.<br />
Comment. The rightmost inequality is easier than the leftmost one. In particular, <strong>Solutions</strong> 2 <strong>and</strong> 3<br />
show that only the condition a 2 b 2 c 2 d 212 is needed for the former one.<br />
d 3Õ4Ôa<br />
9
10<br />
A3. Let x 1 , . . .,x 100 be nonnegative real numbers such that x i x i 1 x i 21 for all<br />
i1, . . .,100 (we put x 101x 1 , x 102x 2 ). Find the maximal possible value of the sum<br />
S100<br />
x i x i 2.<br />
ôi1<br />
Answer. 25<br />
2 .<br />
1Õ 2i x 2i<br />
(Russia)<br />
Solution 1. Let x 2i0, x 2i¡11<br />
for all i1, . . ., 50. Then we have S50¤<br />
1Õ<br />
1<br />
2 2¨225<br />
. So, 2<br />
we are left to show that S25<br />
for all values of x 2 i’s satisfying the problem conditions.<br />
Consider any 1i50. By the problem condition, we get x 2i¡11¡x 2i¡x 2i 1 <strong>and</strong><br />
x 2i 21¡x 2i¡x 1Õ<br />
2i 1. Hence by the AM–GM inequality we get<br />
x 2i¡1x 2i 1 x 2i x 2i 2Ô1¡x 2i¡x 2i 1Õx 2i 1 x 2iÔ1¡x 2i¡x 2i<br />
1Õ¢Ôx Ô1¡x 2i¡x Ôx 2i x 2i 1ÕÔ1¡x 2i<br />
Å21<br />
2i¡x 2i<br />
2 4 .<br />
Summing up these inequalities for i1, 2, . . ., 50, we get the desired inequality<br />
50ôi1Ôx 2i¡1x 2i 1 x 2i x 2i 2Õ50¤1<br />
425<br />
2 .<br />
Comment. This solution shows that a bit more general fact holds. Namely, consider 2n nonnegative<br />
numbers x 1 ,... ,x 2n in a row (with no cyclic notation) <strong>and</strong> suppose that x i x i 1 x i 21 for all<br />
i1,2,... ,2n¡2. Then<br />
2n¡2 ôi1<br />
x i x i 2n¡1<br />
.<br />
4<br />
The proof is the same as above, though if might be easier to find it (for instance, applying<br />
induction). The original estimate can be obtained from this version by considering the sequence<br />
x 1 ,x 2 ,... ,x 100 ,x 1 ,x 2 .<br />
Solution 2. We present another proof of the estimate. From the problem condition, we get<br />
S100<br />
x i x i 2100 ôi1x iÔ1¡x i¡x i 1Õ100 ôi1x i¡100<br />
xi¡100<br />
ôi1 ôi1 2 x i x i 1<br />
ôi1<br />
100 100 ôi1x i¡1 ôi1Ôx i x 1Õ2 i .<br />
2<br />
By the AM–QM inequality, we haveÔx i x i 1Õ21<br />
100<br />
Ôx<br />
S100 ôi1x i¡1 ôi1Ôx i x i 1Õ«2100<br />
200£100<br />
And finally, by the AM–GM inequality<br />
i x 1Õ¨2 i<br />
x i«2<br />
ôi1<br />
ôi1 x i¡2<br />
100£100<br />
2 ôi1x i«£100<br />
100£100<br />
ôi1x i««22<br />
, so<br />
2¡100 ôi1<br />
x i«.<br />
100<br />
S2<br />
x i<br />
100¤£1<br />
2£100<br />
2¡100 ôi1<br />
100¤¢100<br />
4Å225<br />
2 .
11<br />
Comment. These solutions are not as easy as they may seem at the first sight. There are two<br />
different optimal configurations in which the variables have different values, <strong>and</strong> not all of sums of<br />
three consecutive numbers equal 1. Although it is easy to find the value 25 2<br />
, the estimates must be<br />
done with care to preserve equality in the optimal configurations.
12<br />
x<br />
¤¤¤<br />
A4. A sequence x 1 , x 2 , . . . is defined by x 11<strong>and</strong> x 2k¡x k , x 2k¡1Ô¡1Õk k1. Prove that x 1 x 2<br />
n0 for all n1.<br />
1<br />
x k for all<br />
(Austria)<br />
Solution 1. We start with some observations. First, from the definition of x i it follows that<br />
for each positive integer k we have<br />
x 4k¡3x 2k¡1¡x 4k¡2 2kx k . (1)<br />
Hence, denoting S nn<br />
i1 x i, we have<br />
S Ôx 4k¡3 x 4kkôi1 4k¡2Õ Ôx 4k¡1 x 4kÕ¨kôi1Ô0 2x kÕ2S k , (2)<br />
S 4k 2S 4k Ôx 4k 1 x 4k 2ÕS 4k . (3)<br />
Observe also that S nn<br />
i1 x in<br />
i1 1nÔmod 2Õ.<br />
Now we prove by induction on k that S i0for all i4k. The base case is valid since<br />
x 1x 3x 41, x 2¡1. For the induction step, assume that S i0for all i4k. Using<br />
the relations (1)–(3), we obtain<br />
4k 2 S 4k<br />
S 4k 42S k 10, S 4k 2S 4k0, S 4k 3S 4k 2 x 4k<br />
1<br />
4<br />
3S 0.<br />
2<br />
So, we are left to prove that S 4k 10. If k is odd, then S 4k2S k0; since k is odd, S k<br />
is odd as well, so we have S 4k2 <strong>and</strong> hence S 4k 1S 4k x 4k 11.<br />
Conversely, if k is even, then we have x 4k 1x 2k 1x k 1, hence S 4k 1S 4k x 4k<br />
2S k x k 1S k S k 10. The step is proved.<br />
<strong>and</strong> x 4k¡1x 4k¡x<br />
Solution 2. We will use the notation of S n <strong>and</strong> the relations (1)–(3) from the previous<br />
solution.<br />
Assume the contrary <strong>and</strong> consider the minimal n such that S n 10; surely n1, <strong>and</strong><br />
from S n0 we get S n0, x n 1¡1. Hence, we are especially interested in the set<br />
MØn : S n0Ù; our aim is to prove that x n 11whenever nÈM thus coming to a<br />
contradiction.<br />
For this purpose, we first describe the set M inductively. We claim that (i) M consists only<br />
of even numbers, (ii) 2ÈM, <strong>and</strong> (iii) for every even n4we have nÈMÖnß4×ÈM.<br />
Actually, (i) holds since S nnÔmod 2Õ, (ii) is straightforward, while (iii) follows from the<br />
relations S 4k 2S 4k2S k .<br />
Now, we are left to prove that x n 11if nÈM. We use the induction on n. The base<br />
case is n2, that is, the minimal element of M; here we have x 31, as desired.<br />
For the induction step, consider some 4nÈM <strong>and</strong> let mÖnß4×ÈM; then m is even,<br />
<strong>and</strong> x m 11by the induction hypothesis. We prove that x n 1x m 11. If n4m then we<br />
have x n 1x 2m 1x m 1 since m is even; otherwise, n4m 2, <strong>and</strong> x n 1¡x 2x m 1,<br />
as desired. The proof is complete.<br />
2m<br />
Comment. Using the inductive definition of set M, one can describe it explicitly. Namely, M consists<br />
exactly of all positive integers not containing digits 1 <strong>and</strong> 3 in their 4-base representation.
13<br />
A5. Denote by Q the set of all positive rational numbers. Determine all functions f : Q Q<br />
which satisfy the following equation for all x, yÈQ :<br />
f fÔxÕ2 y¨x 3<br />
fÔxyÕ. (1)<br />
(Switzerl<strong>and</strong>)<br />
fÔxÕ1<br />
Answer. The only such function is<br />
x .<br />
Solution. By substituting y1, we get<br />
fÔxÕ2¨<br />
fÔyÕ2¨<br />
f fÔxÕ2¨x 3 fÔxÕ. (2)<br />
Then, whenever fÔxÕfÔyÕ, we have<br />
x 3f f<br />
y 3<br />
which implies xy, so the function f is injective.<br />
Now replace x by xy in (2), <strong>and</strong> apply (1) twice, second time to y, fÔxÕ2¨instead ofÔx, yÕ:<br />
Since f is injective, we get<br />
f fÔxyÕ2¨ÔxyÕ3<br />
fÔxyÕy 3 f fÔxÕ2 y¨f<br />
fÔxyÕfÔxÕfÔyÕ.<br />
Therefore, f is multiplicative. This also implies fÔ1Õ1<br />
fÔxyÕ2fÔxÕ2 fÔyÕ2 ,<br />
fÔxÕ2<br />
fÔyÕ2¨.<br />
<strong>and</strong> fÔxnÕfÔxÕn for all integers n.<br />
Then the function equation (1) can be re-written as<br />
f fÔxÕ¨2 fÔxÕ<br />
fÔyÕx xfÔxÕ¨5ß2 fÔxÕ¨<br />
3 fÔxÕfÔyÕ,<br />
f fÔxÕ¨x 3 fÔxÕ. (3)<br />
Let gÔxÕxfÔxÕ. Then, by (3), we have<br />
ÐÓÓÓÓÑÓÓÓÓÒÔxÕ...¨© g gÔxÕ¨g xfÔxÕ¨xfÔxÕ¤f xfÔxÕ¨xfÔxÕ2 f<br />
xfÔxÕ2x 3 gÔxÕ¨5ß2 ,<br />
<strong>and</strong>, by induction,<br />
gÔxÕ¨Ô5ß2Õn<br />
g¡g ...g<br />
(4)<br />
n 1<br />
for every positive integer n.<br />
Consider (4) for<br />
ÐÓÓÓÓÑÓÓÓÓÒ<br />
gÔxÕ¨Ô5ß2Õn a fixed x. The left-h<strong>and</strong> side is always rational, so must be<br />
rational for every n. We show that this is possible only if gÔxÕ1. Suppose that gÔxÕ1,<br />
<strong>and</strong> let the prime factorization of gÔxÕbe gÔxÕp α 1<br />
1 . . .p α k<br />
k<br />
where p 1 , . . .,p k are distinct primes<br />
<strong>and</strong> α 1 , . . .,α k are nonzero integers. Then the unique prime factorization of (4) is<br />
g¡g ...g gÔxÕ¨Ô5ß2ÕnpÔ5ß2Õn α ¤¤¤pÔ5ß2Õn 1 α k<br />
1 k<br />
n 1<br />
ÔxÕ...¨©
14<br />
where the exponents should be integers. But this is not true for large values of n, for example<br />
. Therefore, gÔxÕ1 is impossible.<br />
yÕ y<br />
Hence, gÔxÕ1 <strong>and</strong> for all x.<br />
fÔxÕ1<br />
The function satisfies the equation (1):<br />
x fÔfÔxÕ2 1 1<br />
fÔxÕ2 x¨2 1<br />
yx3<br />
xyx 3 fÔxyÕ.<br />
fÔxÕ1<br />
Comment. Among R R functions, is not the only solution. Another solution is<br />
x<br />
f 1ÔxÕx . Using transfinite tools, infinitely many other solutions can be constructed.<br />
Ô5<br />
2Õn α 1 cannot be a integer number when 2 nЧα<br />
3ß2<br />
fÔxÕ1<br />
thus<br />
x<br />
1
gÔnÕ fÔgÔnÕÕfÔnÕ gÔfÔnÕÕ<br />
15<br />
A6. Suppose that f <strong>and</strong> g are two functions defined on the set of positive integers <strong>and</strong> taking<br />
positive integer values. Suppose also that the equations 1 <strong>and</strong><br />
1 hold for all positive integers. Prove that fÔnÕgÔnÕfor all positive integer n.<br />
(Germany)<br />
kÔxÕ¨gÔxÕ k¡1ÔxÕ¨ 1¤¤¤fÔxÕ ÐÓÓÓÓÑÓÓÓÓÒ<br />
Solution 1. Throughout the solution, by N we denote the set of all positive integers. For<br />
any function h: NN<strong>and</strong> for any positive integer k, define h kÔxÕh h . . .hÔxÕ...¨¨(in<br />
k<br />
particular, h 0ÔxÕx).<br />
Observe that f g kÔxÕ¨f g k for any positive integer k, <strong>and</strong><br />
similarly g f k. Now let a <strong>and</strong> b are the minimal values attained by f <strong>and</strong> g,<br />
respectively; say fÔn fÕa, gÔn gÕb. Then we have f g kÔn fÕ¨a k, g f kÔn gÕ¨b k, so<br />
the function f attains all values from the set N fØa, a 1, . . .Ù, while g attains all the values<br />
from the set N gØb, b 1, . . .Ù.<br />
Next, note that fÔxÕfÔyÕimplies gÔxÕg fÔxÕ¨¡1g fÔyÕ¨¡1gÔyÕ; surely, the<br />
converse implication also holds. Now, we say that x <strong>and</strong> y are similar (<strong>and</strong> write xy) if<br />
fÔxÕfÔyÕ(equivalently, gÔxÕgÔyÕ). For every xÈN, we defineÖx×ØyÈN:xyÙ;<br />
surely, y 1y 2 for all y 1 , y 2ÈÖx×, soÖx×Öy×whenever yÈÖx×.<br />
Ð<br />
Now we investigate the structure of the setsÖx×.<br />
Claim 1. Suppose that fÔxÕfÔyÕ; then xy, that is, fÔxÕfÔyÕ. Consequently, each<br />
classÖx×contains at most one element from N f , as well as at most one element from N g .<br />
Proof. If fÔxÕfÔyÕ, then we have gÔxÕg fÔxÕ¨¡1g fÔyÕ¨¡1gÔyÕ, so xy. The<br />
second statement follows now from the sets of values of f <strong>and</strong> g.<br />
Next, we clarify which classes do not contain large elements.<br />
fÔyÕ<br />
Claim 2. For any xÈN, we haveÖx×Ø1,<br />
gÔyÕ¨fÔyÕ gÔyÕ¨<br />
2, . . .,b¡1Ùif <strong>and</strong> only if fÔxÕa. Analogously,<br />
Ð<br />
Öx×Ø1, 2, . . ., a¡1Ùif <strong>and</strong> only if gÔxÕb.<br />
Proof. We will prove thatÖx×Ø1, 2, . . .,b¡1ÙfÔxÕa; the proof of the second<br />
statement is similar.<br />
Note that fÔxÕaimplies that there exists some y satisfying fÔyÕfÔxÕ¡1, so f<br />
1fÔxÕ, <strong>and</strong> hence xgÔyÕb. Conversely, if bcxthen cgÔyÕfor some yÈN,<br />
which in turn follows fÔxÕf 1a 1, <strong>and</strong> hence fÔxÕa.<br />
Claim 2 implies that there exists exactly one class contained inØ1, . . .,a¡1Ù(that is, the<br />
classÖn g×), as well as exactly one class contained inØ1, . . .,b¡1Ù(the classÖn f×). Assume for a<br />
moment that ab; thenÖn g×is contained inØ1, . . .,b¡1Ùas well, hence it coincides withÖn g×.<br />
So, we get that<br />
Ð<br />
Claim 3. ab.<br />
Proof. By Claim 2, we haveÖa×Ön f×, soÖa×should contain some element a½b by Claim 2<br />
again. If aa½, thenÖa×contains two elementsawhich is impossible by Claim 1. Therefore,<br />
aa½b. Similarly, ba.<br />
Now we are ready to prove the problem statement. First, we establish the following<br />
fÕ Ð<br />
d.<br />
Proof. Induction on d. For d0, the statement follows from (1) <strong>and</strong> Claim 3. Next, for d1<br />
f g dÔn fÕ¨fÔn da d.<br />
d is analogous.<br />
fÔxÕagÔxÕbxn fn g . (1)<br />
Claim 4. For every integer d0, f d 1Ôn fÕgd 1Ôn fÕa from the induction hypothesis we have f d 1Ôn fÕf dÔn fÕ¨f The equality g d 1Ôn fÕa
16<br />
Finally, for each xÈN, we have fÔxÕa d for some d0, so fÔxÕf g dÔn fÕ¨<strong>and</strong><br />
hence xgdÔn fÕ. It follows that gÔxÕg g dÔn fÕ¨gd 1Ôn fÕa fÔaÕ¨gÔaÕ<br />
dfÔxÕby Claim 4.<br />
Solution 2. We start with the same observations, introducing the relation<strong>and</strong> proving<br />
Claim<br />
fÔaÕfÔxÕ<br />
1 from the previous solution.<br />
Note that fÔaÕa since otherwise we have fÔaÕa<strong>and</strong> hence gÔaÕg Ð1,<br />
which is false.<br />
Claim 2½. ab.<br />
Proof. We can assume that ab. Since fÔaÕa 1, there exists some xÈNsuch that<br />
1, which is equivalent to fÔaÕf gÔxÕ¨<strong>and</strong> agÔxÕ. fÔaÕfÔxÕ<br />
Since gÔxÕba, by<br />
Claim 1 we have agÔxÕb, which together with ab proves the Claim.<br />
Now, almost the same method allows to find the values fÔaÕ<strong>and</strong> gÔaÕ.<br />
Claim 3½. fÔaÕgÔaÕa 1.<br />
Proof. Assume the contrary; then fÔaÕa 2, hence there exist some x, yÈNsuch that<br />
fÔxÕfÔaÕ¡2 <strong>and</strong> fÔyÕgÔxÕ(as gÔxÕab). Now we get 2f g 2ÔxÕ¨,<br />
so ag2ÔxÕa, <strong>and</strong> by Claim<br />
gÔxÕ¨fÔxÕ<br />
1 we get ag2ÔxÕg 1Õ<br />
fÔyÕ¨1 gÔyÕ1 a; this is<br />
impossible. The equality gÔaÕa 1 is similar.<br />
Now, we are prepared for the proof of the problem statement. First, we prove it for na.<br />
Claim 4½. For each integer xa, we have fÔxÕgÔxÕx 1.<br />
Proof. Induction on x. The base case xais provided by Claim 3½, while the induction<br />
step<br />
fÔnÕ gÔnÕ¨gÔnÕ fÔnÕ fÔfÔnÕÕ 1<br />
follows from fÔx 1Õf 1Ôx 1 <strong>and</strong> the similar computation<br />
for gÔx 1Õ.<br />
Finally, for an arbitrary nÈNwe have gÔnÕa, so by Claim 4½we have<br />
f 1, hence fÔnÕgÔnÕ.<br />
Comment. It is not hard now to describe all the functions f : NN satisfying the property<br />
1. For each such function, there exists n 0ÈN such that fÔnÕn 1 for all nn 0 , while for<br />
each nn 0 , fÔnÕis an arbitrary number greater than of equal to n 0 (these numbers may be different<br />
for different nn 0 ).
17<br />
A7. Let a 1 , . . .,a r be positive real numbers. For nr, we inductively define<br />
a nmax<br />
1kn¡1Ôa k a n¡kÕ. (1)<br />
Prove that there exist positive integers lr <strong>and</strong> N such that a na n¡l a l for all nN.<br />
¤¤¤<br />
(Iran)<br />
Solution 1. First, from the problem conditions we have that each a n (nr) can be expressed<br />
as a na j1 a j2 with j 1 , j 2n, j 1 j 2n. If, say, j 1r then we can proceed in the same<br />
way with a j1 , <strong>and</strong> so on. Finally, we represent a n in a form<br />
a na i1 a ik , (2)<br />
i kn. (3)<br />
Moreover, if a i1<br />
¤¤¤<br />
<strong>and</strong> a i2 are the numbers in (2) obtained on the last step, then i 1 i 2r.<br />
Hence we can adjust (3) as<br />
i kn, i 1 i 2r.<br />
¤¤¤<br />
(4)<br />
On the other h<strong>and</strong>, suppose that the indices i 1 , . . .,i k satisfy the conditions (4). Then,<br />
denoting s ji 1 i j , from (1) we have<br />
a na ska sk¡1 ¤¤¤<br />
a ika sk¡2 a ik¡1 a ik¤¤¤a i1 a ik<br />
Ð<br />
.<br />
Summarizing these observations we get the following<br />
Claim. For every nr, we have<br />
a nmaxØa i1 a ik : the collectionÔi 1 , . . .,i kÕsatisfies (4)Ù.<br />
¤¤¤<br />
Now we denote<br />
a i<br />
smax<br />
1ir i<br />
l<br />
<strong>and</strong> fix some index lr such that sa l .<br />
Consider some nr2 l 2r <strong>and</strong> choose an expansion of a n in the form (2), (4). Then we have<br />
ni 1 i krk,<br />
¤¤¤ ¤¤¤<br />
so knßrrl 2. Suppose that none of the numbers i 3 , . . ., i k equals l.<br />
Then by the pigeonhole principle there is an index 1jr which appears among i 3 , . . .,i k<br />
at least l times, <strong>and</strong> surely jl. Let us delete these l occurrences of j fromÔi 1 , . . ., i kÕ, <strong>and</strong><br />
add j occurrences of l instead, obtaining a sequenceÔi 1 , i 2 , i½3 , . . .,i½k½Õalso satisfying (4). By<br />
Claim, we have<br />
a i1 a ika na i1 a i2 a i½3<br />
a i½k½,<br />
or, after removing the coinciding terms, la jja¤¤¤ l , so a l j<br />
. By the definition of l, this<br />
la j<br />
means that la jja l , hence<br />
a na i1 a i2<br />
¤¤¤<br />
a i½3<br />
a i½k½.<br />
Thus, for every nr2 l 2r we have found a representation of the form (2), (4) with i jl<br />
for some j3. Rearranging the indices we may assume<br />
ik¡1Õ<br />
that i kl.<br />
Finally, observe that in this representation, the indicesÔi 1 , . . ., i k¡1Õsatisfy the conditions<br />
(4) with n replaced by n¡l. Thus, from the Claim we get<br />
a n¡l a lÔa i1 a a la n ,<br />
which by (1) implies<br />
a na n¡l a l for each nr 2 l 2r,<br />
as desired.<br />
1i jr, i 1 ¤¤¤<br />
1i jr, i 1 ¤¤¤
18<br />
Solution 2. As in the previous solution, we involve the expansion (2), (3), <strong>and</strong> we fix some<br />
index 1lr such that<br />
a l a i<br />
lsmax<br />
1ir i .<br />
Now, we introduce the sequenceÔb nÕas b na n¡sn; then b l0.<br />
We prove by induction on n that b n0, <strong>and</strong>Ôb nÕsatisfies the same recurrence relation<br />
asÔa nÕ. The base cases nrfollow from the definition of s. Now, for nrfrom the<br />
induction hypothesis we have<br />
b nmax<br />
1kn¡1Ôa k a n¡kÕ¡nsmax<br />
1kn¡1Ôb k b n¡k nsÕ¡nsmax<br />
1kn¡1Ôb k b n¡kÕ0,<br />
as required.<br />
Now, if b k0 for all 1kr, then b n0 for all n, hence a nsn, <strong>and</strong> the statement is<br />
trivial. Otherwise, define<br />
Mmax<br />
1irb i,<br />
εminØb i:1ir, b i0Ù.<br />
Then for nr we obtain<br />
b nmax<br />
1kn¡1Ôb k b n¡kÕb<br />
l<br />
b n¡lb n¡l,<br />
so<br />
¤¤¤<br />
contained in a set<br />
b<br />
We claim that this set is finite. Actually, for any xÈT, let xb i1 b ik (i 1 , . . .,i kr).<br />
Then among b ij ’s there are at most M nonzero terms (otherwise xM<br />
Thus<br />
ε ε¤Ô¡εÕ¡M).<br />
x can be expressed in the same way with kM<br />
, <strong>and</strong> there is only a finite number of such<br />
ε<br />
sums.<br />
0b nb ¤¤¤<br />
n¡lb n¡2l¤¤¤¡M.<br />
Thus, in view of the expansion (2), (3) applied to the sequenceÔb nÕ, 0×<br />
we get that each b n is<br />
TØb i1 b i2 ik : i 1 , . . .,i krÙÖ¡M,<br />
Finally, for every t1, 2, . . ., l we get that the sequence<br />
b r t, b r t l, b r t 2l, . . .<br />
is non-decreasing <strong>and</strong> attains the finite number of values; therefore it is constant from some<br />
index. Thus, the sequenceÔb nÕis periodic with period l from some index N, which means that<br />
b nb Ôn¡lÕsÕ<br />
n¡lb n¡l b l for all nN l,<br />
<strong>and</strong> hence<br />
a nb n nsÔb n¡l Ôb l lsÕa n¡l a l for all nN l,<br />
as desired.
TÕ<br />
19<br />
A8. Given six positive numbers a, b, c, d, e, f such that abcdef. Let a c eS<br />
<strong>and</strong> b d fT. Prove that<br />
2ST3ÔS SÔbd bf dfÕ TÔac ae ceÕ¨. (1)<br />
fÕÔx¡aÕÔx¡cÕÔx¡eÕ eÕÔx¡bÕÔx¡dÕÔx¡fÕ<br />
(South Korea)<br />
Solution 1. We define also σac ce ae, τbd bf df. The idea of the solution is to<br />
interpret (1) as a natural inequality on the roots of an appropriate polynomial.<br />
Actually, consider the polynomial<br />
PÔxÕÔb d Ôa c<br />
σx¡aceÕ TÔx 3¡Sx 2 SÔx 3¡Tx 2 τx¡bdfÕ. (2)<br />
Surely, P is cubic with leading coefficient S T0. Moreover, we have<br />
PÔaÕSÔa¡bÕÔa¡dÕÔa¡fÕ0,<br />
PÔeÕSÔe¡bÕÔe¡dÕÔe¡fÕ0,<br />
PÔcÕSÔc¡bÕÔc¡dÕÔc¡fÕ0,<br />
PÔfÕTÔf¡aÕÔf¡cÕÔf¡eÕ0.<br />
Hence, each of the intervalsÔa, cÕ,Ôc, eÕ,Ôe, fÕcontains at least one root of PÔxÕ. Since there<br />
are at most three roots at all, we obtain that there is exactly one root in each interval (denote<br />
them by αÈÔa, cÕ, βÈÔc, eÕ, γÈÔe, fÕ). Moreover, the polynomial P can be factorized as<br />
PÔxÕÔT SÕÔx¡αÕÔx¡βÕÔx¡γÕ. (3)<br />
Equating the coefficients in the two representations (2) <strong>and</strong> (3) of PÔxÕprovides<br />
α β γ2TS<br />
T S ,<br />
αβ αγ βγSτ Tσ<br />
T S .<br />
Now, since the numbers α, β, γ are distinct, we have<br />
0Ôα¡βÕ2 which implies<br />
β<br />
or<br />
4S 2 T 23ÔT SÕÔTσ SτÕ,<br />
4S 2 T 2<br />
ÔT SÕ2Ôα<br />
Ôα¡γÕ2 Ôβ¡γÕ22Ôα β γÕ2¡6Ôαβ αγ βγÕ,<br />
TσÕ<br />
γÕ23Ôαβ αγ βγÕ3ÔSτ<br />
,<br />
T S<br />
which is exactly what we need.<br />
Comment 1. In fact, one can locate the roots of PÔxÕmore narrowly: they should lie in the intervals<br />
Ôa,bÕ,Ôc,dÕ,Ôe,fÕ.<br />
Surely, if we change all inequality signs<br />
fÕ<br />
in the problem statement to non-strict ones, the (non-strict)<br />
inequality will also hold by continuity. One can also find when the equality is achieved. This happens<br />
in that case when PÔxÕis a perfect cube, which immediately implies that bcdeÔαβγÕ,<br />
together with the additional condition that P¾ÔbÕ0. Algebraically,<br />
6ÔT SÕb¡4TS0<br />
3bÔa fÕ2Ôa 2bÕÔ2b<br />
fbÔ4b¡aÕ 3Ôb¡aÕ<br />
2a bb¢1<br />
2a bÅb.<br />
This means that for every pair of numbers a, b such that 0ab, there exists fb such that the<br />
pointÔa,b,b,b,b,fÕis a point of equality.
20<br />
Solution 2. Let<br />
U1<br />
Ôe¡aÕ2 Ôc¡aÕ2 Ôe¡cÕ2¨S 2¡3Ôac ae ceÕ<br />
2<br />
TÕ<br />
<strong>and</strong><br />
2¡VÕ TÕ<br />
V1<br />
Ôf¡bÕ2 Ôf¡dÕ2 Ôd¡bÕ2¨T 2¡3Ôbd bf dfÕ.<br />
2<br />
ThenÔL.H.S.Õ2¡ÔR.H.S.Õ2Ô2STÕ2¡ÔS S¤3Ôbd bf dfÕ ceÕ¨<br />
T¤3Ôac ae<br />
4S 2 T 2¡ÔS SÔT SVÕ<br />
TÔS 2¡UÕ¨ÔS TÕÔSV TUÕ¡STÔT¡SÕ2 ,<br />
<strong>and</strong> the statement is equivalent with<br />
ÔS TÕÔSV TUÕSTÔT¡SÕ2 . (4)<br />
By the Cauchy-Schwarz inequality,<br />
ÔS TÕÔTU<br />
S¤TU<br />
T¤SV¨2ST<br />
U V¨2<br />
Ôe¡aÕ Ôc¡aÕ Ôf¡bÕ Ôf¡dÕ 2Å<br />
. (5)<br />
Estimate the quantitiesU <strong>and</strong>V by the QM–AM inequality with the positive termsÔe¡cÕ2<br />
<strong>and</strong>Ôd¡bÕ2 being omitted:<br />
ÔT¡SÕ 2Ôe¡dÕ 2¡a© Ôf¡bÕ2 U<br />
VÔe¡aÕ2 Ôc¡aÕ2<br />
Ôf¡dÕ2<br />
2 2<br />
¡e c<br />
2 2¡b<br />
2<br />
Ôc¡bÕÔc¡dÕ<br />
3<br />
Ôe¡fÕÔe¡dÕ<br />
3<br />
(6)<br />
2Ôc¡bÕT¡S.<br />
The estimates (5) <strong>and</strong> (6) prove (4) <strong>and</strong> hence the statement.<br />
Solution 3. We keep using the notations σ <strong>and</strong> τ from Solution 1. Moreover, let sc e.<br />
Note that<br />
Ôe¡fÕÔc¡bÕ0,<br />
since each summ<strong>and</strong> is negative. This rewrites as<br />
Ôbd bf dfÕ¡Ôac ce aeÕÔc eÕÔb d f¡a¡c¡eÕ, or<br />
τ¡σsÔT¡SÕ. (7)<br />
Then we have<br />
Sτ TσSÔτ¡σÕ<br />
SsÔT¡SÕ<br />
TÕσSsÔT¡SÕ ÔS ÔS TÕÔce asÕ<br />
TÕsÅ<br />
TÕ¢s 2<br />
ÔS¡sÕsÅs¢2ST¡3<br />
ÔS TÕsÅ.<br />
4 4ÔS Using this inequality together with the AM–GM inequality we get<br />
3<br />
TσÕ3<br />
TÕÔSτ TÕs¢2ST¡3<br />
4ÔS TÕ<br />
4ÔS 4ÔS 3<br />
4ÔS TÕs 2ST¡3<br />
4ÔS TÕs<br />
ST.<br />
2<br />
Hence,<br />
2ST3ÔS SÔbd bf dfÕ TÔac ae ceÕ¨.<br />
2<br />
¢f¡d
Comment 2. The expression (7) can be found by considering the sum of the roots of the quadratic<br />
polynomial qÔxÕÔx¡bÕÔx¡dÕÔx¡fÕ¡Ôx¡aÕÔx¡cÕÔx¡eÕ.<br />
Solution 4. We introduce the expressions σ <strong>and</strong> τ as in the previous solutions. The idea of<br />
the solution is to change the values of variables a, . . .,f keeping the left-h<strong>and</strong> side unchanged<br />
<strong>and</strong> increasing the right-h<strong>and</strong> side; it will lead to a simpler inequality which can be proved in<br />
a direct way.<br />
abcd<br />
Namely, we change the variables (i) keeping the (non-strict) inequalities<br />
x½<br />
ef; (ii) keeping the values of sums S <strong>and</strong> T unchanged; <strong>and</strong> finally (iii) increasing<br />
zÕ<br />
the<br />
values of σ <strong>and</strong> τ. Then the left-h<strong>and</strong> side of (1) remains unchanged, while the right-h<strong>and</strong><br />
side increases. Hence, the inequality (1) (<strong>and</strong> even a non-strict version of (1)) for the changed<br />
values would imply the same (strict) inequality for the original values.<br />
First, we find<br />
zÕzÔx½<br />
the sufficient conditions for (ii) <strong>and</strong> (iii) to be satisfied.<br />
y½Õ x½y½zÔx½<br />
Lemma. Let x, y, z0; denote UÔx, y, zÕx y z, υÔx, y, zÕxy xz yz. Suppose that<br />
y½x y butx¡yx½¡y½; then we<br />
yÕ y½Õ Ôx½<br />
have UÔx½, y½, zÕUÔx, y, zÕ<strong>and</strong> υÔx½, y½,<br />
υÔx, y, zÕwith equality achieved only whenx¡yx½¡y½.<br />
Proof. The first equality is obvious. For the second, we have<br />
y½Õ2¡Ôx½¡y½Õ2<br />
υÔx½, y½,<br />
4<br />
Ôx yÕ2¡Ôx¡yÕ2<br />
zÔx υÔx, y, zÕ,<br />
4<br />
with the equality achieved only forÔx½¡y½Õ2Ôx¡yÕ2x½¡y½x¡y, as desired.<br />
ÐNow, we apply Lemma several times making the following changes. For each change, we<br />
denote the new values by the same letters to avoid cumbersome notations.<br />
1. Let kd¡c<br />
2 . ReplaceÔb, c, d, eÕbyÔb k, c k, d¡k, e¡kÕ. After the change we have<br />
abcdef, the values of S, T remain unchanged, but σ, τ strictly increase by<br />
Lemma.<br />
2. Let le¡d<br />
2 . ReplaceÔc, d, e, fÕbyÔc l, d l, e¡l, f¡lÕ. After the change we<br />
dÕ<br />
have<br />
abcdef, the values of S, T remain unchanged, but σ, τ strictly increase by the<br />
Lemma.<br />
3. Finally, let mc¡b<br />
3 . ReplaceÔa, b, c, d, e, fÕbyÔa 2m, b 2m, c¡m, d¡m, e¡m, f¡mÕ.<br />
After the change, we have abcdef <strong>and</strong> S, T are unchanged. To check (iii),<br />
we observe that our change can be considered as a composition of two changes:Ôa, b, c,<br />
Ôa m, b m, c¡m, d¡mÕ<strong>and</strong>Ôa, b, e, fÕÔa fÕ<br />
m, b m, e¡m, f¡mÕ. It is easy to see that<br />
each of these two consecutive changes satisfy the conditions of the Lemma, hence the values<br />
of σ <strong>and</strong> τ increase.<br />
Finally, we come to the situation when abcdef, <strong>and</strong> we need to prove the<br />
inequality<br />
2Ôa 2bÕÔ2b 2bfÕ 2Õ¨<br />
fÕ3Ôa Ô2b fÕÔ2ab b<br />
Now, observe that<br />
fÕ¤<br />
4b<br />
3bÔa 4b<br />
2¤2Ôa 2bÕÔ2b fÕ3bÔa 4b fÕ<br />
Ôa 2bÕÔb<br />
2 2bÕÔb<br />
Ôa<br />
Ôa 2bÕÔb<br />
2fÕ<br />
2fÕ<br />
21<br />
Ô2b fÕÔ2a bÕ¨. (8)<br />
Ô2a bÕÔ2b fÕ¨.
22<br />
HenceÔ4Õrewrites as<br />
3bÔa 4b fÕ<br />
2fÕ<br />
fÕ¤ 23bÔa 4b<br />
Ôa 2bÕÔb<br />
which is simply the AM–GM inequality.<br />
fÕ¨<br />
Ô2a bÕÔ2b<br />
2fÕ<br />
Ôa 2bÕÔb<br />
Ô2b fÕÔ2a bÕ¨,<br />
Comment 3. Here, we also can find all the cases of equality. Actually, it is easy to see that if<br />
some two numbers among b,c,d,e are distinct then one can use Lemma to increase the right-h<strong>and</strong> side<br />
of (1). Further, if bcde, then we need equality inÔ4Õ; this means that we apply AM–GM to<br />
equal numbers, that is,<br />
3bÔa 4b 2fÕ fÕÔa 2bÕÔb Ô2a bÕÔ2b fÕ,<br />
which leads to the same equality as in Comment 1.
Combinatorics<br />
C1. In a concert, 20 singers will perform. For each singer, there is a (possibly empty) set of<br />
other singers such that he wishes to perform later than all the singers from that set. Can it<br />
happen that there are exactly 2010 orders of the singers such that all their wishes are satisfied<br />
Answer. Yes, such an example exists.<br />
(Austria)<br />
Solution. We say that an order of singers is good if it satisfied all their wishes. Next, we<br />
say that a number N is realizable by k singers (or k-realizable) if for some set of wishes of<br />
these singers there are exactly N good orders. Thus, we have to prove that a number 2010 is<br />
20-realizable.<br />
We start with the following simple<br />
Lemma. Suppose that numbers n 1 , n 2 are realizable by respectively k 1 <strong>and</strong> k 2 singers. Then<br />
the number n 1 n 2 isÔk 1 k 2Õ-realizable.<br />
Proof. Let the singers A 1 , . . ., A k1 (with some wishes among them) realize n 1 , <strong>and</strong> the singers B 1 ,<br />
..., B k2 (with some wishes among them) realize n 2 . Add to each singer B i the wish to perform<br />
later than all the singers A j . Then, each good order of the obtained set of singers has the form<br />
ÔA i1 , . . .,A ik1 , B j1 , . . .,B jk2Õ, whereÔA i1 , . . .,A ik1Õis a good order for A i ’s <strong>and</strong>ÔB j1 , . . .,B jk2Õ<br />
Ð<br />
is a good order for B j ’s. Conversely, each order of this form is obviously good. Hence, the<br />
number of good orders is n 1 n 2 .<br />
In view of Lemma, we show how to construct sets of singers containing 4, 3 <strong>and</strong> 13 singers<br />
<strong>and</strong> realizing the numbers 5, 6 <strong>and</strong> 67, respectively. Thus the number 20106¤5¤67 will be<br />
realizable by 4 3 1320 singers. These companies of singers are shown in Figs. 1–3; the<br />
wishes are denoted by arrows, <strong>and</strong> the number of good orders for each Figure st<strong>and</strong>s below in<br />
the brackets.<br />
a 3<br />
a 2<br />
a 1<br />
a<br />
b<br />
a 4<br />
c<br />
(5)<br />
Fig. 1<br />
d<br />
(3)<br />
Fig. 2<br />
y<br />
a 5<br />
a 6<br />
a 7<br />
a 8<br />
a 9<br />
a 10<br />
x<br />
a 11<br />
(67)<br />
Fig. 3<br />
For Fig. 1, there are exactly 5 good ordersÔa, b, c, dÕ,Ôa, b, d, cÕ,Ôb, a, c, dÕ,Ôb, a, d, cÕ,<br />
Ôb, d, a, cÕ. For Fig. 2, each of 6 orders is good since there are no wishes.
24<br />
Finally, for Fig. 3, the order of a 1 , . . .,a 11 is fixed; in this line, singer x can st<strong>and</strong> before<br />
each of a i (i9), <strong>and</strong> singer y can st<strong>and</strong> after each of a j (j5), thus resulting in 9¤763<br />
cases. Further, the positions of x <strong>and</strong> y in this line determine the whole order uniquely unless<br />
both of them come between the same pairÔa i , a i 1Õ(thus 5i8); in the latter cases, there<br />
are two orders instead of 1 due to the order of x <strong>and</strong> y. Hence, the total number of good orders<br />
is 63 467, as desired.<br />
Comment. The number 20 in the problem statement is not sharp <strong>and</strong> is put there to respect the<br />
original formulation. So, if necessary, the difficulty level of this problem may be adjusted by replacing<br />
20 by a smaller number. Here we present some improvements of the example leading to a smaller<br />
number of singers.<br />
Surely, each example with20 singers can be filled with some “super-stars” who should perform<br />
at the very end in a fixed order. Hence each of these improvements provides a different solution for<br />
the problem. Moreover, the large variety of ideas st<strong>and</strong>ing behind these examples allows to suggest<br />
that there are many other examples.<br />
1. Instead of building the examples realizing 5 <strong>and</strong> 6, it is more economic to make an example<br />
realizing 30; it may seem even simpler. Two possible examples consisting of 5 <strong>and</strong> 6 singers are shown<br />
in Fig. 4; hence the number 20 can be decreased to 19 or 18.<br />
For Fig. 4a, the order of a 1 ,... ,a 4 is fixed, there are 5 ways to add x into this order, <strong>and</strong> there<br />
are 6 ways to add y into the resulting order of a 1 ,...,a 4 ,x. Hence there are 5¤630 good orders.<br />
On Fig. 4b, for 5 singers a,b 1 ,b 2 ,c 1 ,c 2 there are 5!120 orders at all. Obviously, exactly one half<br />
of them satisfies the wish b 1b 2 , <strong>and</strong> exactly one half of these orders satisfies another wish c 1c 2 ;<br />
hence, there are exactly 5!ß430 good orders.<br />
a 1<br />
a 2<br />
a 3<br />
x<br />
y<br />
b 1<br />
a<br />
c 2<br />
b 2<br />
c 1<br />
y<br />
b 2 b<br />
b 3 1<br />
b 4 b 5<br />
a 6<br />
a7<br />
a8<br />
a 9<br />
x<br />
b 1 b 2<br />
b 3 b 4<br />
a 5<br />
y a 6<br />
a7 x<br />
a8<br />
c 9 c 10<br />
a 4<br />
(30)<br />
a)<br />
(30)<br />
b)<br />
a 11<br />
(2010)<br />
a 10<br />
(2010)<br />
Fig. 4 Fig. 5 Fig. 6<br />
c 11<br />
2. One can merge the examples for 30 <strong>and</strong> 67 shown in Figs. 4b <strong>and</strong> 3 in a smarter way, obtaining<br />
a set of 13 singers representing 2010. This example is shown in Fig. 5; an arrow from/to group<br />
Øb 1 ,...,b 5Ùmeans that there exists such arrow from each member of this group.<br />
Here, as in Fig. 4b, one can see that there are exactly 30 orders of b 1 ,... ,b 5 ,a 6 ,...,a 11 satisfying<br />
all their wishes among themselves. Moreover, one can prove in the same way as for Fig. 3 that each<br />
of these orders can be complemented by x <strong>and</strong> y in exactly 67 ways, hence obtaining 30¤672010<br />
good orders at all.<br />
Analogously, one can merge the examples in Figs. 1–3 to represent 2010 by 13 singers as is shown<br />
in Fig. 6.
25<br />
a 5<br />
a 1<br />
a 2<br />
a 3<br />
b 5<br />
b 6<br />
(67)<br />
b 4<br />
a 4<br />
b 1<br />
b 2<br />
b 3<br />
a 4<br />
b 3<br />
b 2<br />
b 4<br />
a 3<br />
a<br />
a 6<br />
2<br />
a 1<br />
b 1<br />
(2010)<br />
Fig. 7 Fig. 8<br />
3. Finally, we will present two other improvements; the proofs are left to the reader. The graph in<br />
Fig. 7 shows how 10 singers can represent 67. Moreover, even a company of 10 singers representing 2010<br />
can be found; this company is shown in Fig. 8.
26<br />
C2. On some planet, there are 2 N countries (N4). Each country has a flag N units wide<br />
<strong>and</strong> one unit high composed of N fields of size 1¢1, each field being either yellow or blue. No<br />
two countries have the same flag.<br />
We say that a set of N flags is diverse if these flags can be arranged into an N¢N square so<br />
that all N fields on its main diagonal will have the same color. Determine the smallest positive<br />
integer M such that among any M distinct flags, there exist N flags forming a diverse set.<br />
Answer. M2 N¡2 1.<br />
(Croatia)<br />
Solution. When speaking about the diagonal of a square, we will always mean the main<br />
diagonal.<br />
Let M N be the smallest positive integer satisfying the problem condition. First, we show<br />
that M N2N¡2 . Consider the collection of all 2 N¡2 flags having yellow first squares <strong>and</strong> blue<br />
second ones. Obviously, both colors appear on the diagonal of each N¢N square formed by<br />
these flags.<br />
We are left to show that M N2N¡2 1, thus obtaining the desired answer. We start with<br />
establishing this statement for N4.<br />
Suppose that we have 5 flags of length 4. We decompose each flag into two parts of 2 squares<br />
each; thereby, we denote each flag as LR, where the 2¢1 flags L, RÈSØBB, BY, YB, YYÙ<br />
are its left <strong>and</strong> right parts, respectively. First, we make two easy observations on the flags 2¢1<br />
which can be checked manually.<br />
(i) For each AÈS, there exists only one 2¢1 flag CÈS (possibly CA) such that A<br />
<strong>and</strong> C cannot form a 2¢2 square with monochrome diagonal (for part BB, that is YY, <strong>and</strong><br />
for BY that is YB).<br />
(ii) Let A 1 , A 2 , A 3ÈS be three distinct elements; then two of them can form a 2¢2 square<br />
with yellow diagonal, <strong>and</strong> two of them can form a 2¢2 square with blue diagonal (for all parts<br />
but BB, a pair (BY, YB) fits for both statements, while for all parts but BY, these pairs are<br />
(YB, YY) <strong>and</strong> (BB, YB)).<br />
Now, let l <strong>and</strong> r be the numbers of distinct left <strong>and</strong> right parts of our 5 flags, respectively.<br />
The total number of flags is 5rl, hence one of the factors (say, r) should be at least 3. On<br />
the other h<strong>and</strong>, l, r4, so there are two flags with coinciding right part; let them be L 1 R 1<br />
<strong>and</strong> L 2 R 1 (L 1L 2 ).<br />
Next, since r3, there exist some flags L 3 R 3 <strong>and</strong> L 4 R 4 such that R 1 , R 3 , R 4 are distinct.<br />
Let L½R½be the remaining flag. By (i), one of the pairsÔL½, L 1Õ<strong>and</strong>ÔL½, L 2Õcan form a<br />
2¢2 square with monochrome diagonal; we can assume that L½, L 2 form a square with a blue<br />
diagonal. Finally, the right parts of two of the flags L 1 R 1 , L 3 R 3 , L 4 R 4 can also form a 2¢2<br />
square with a blue diagonal by (ii). Putting these 2¢2 squares on the diagonal of a 4¢4<br />
square, we find a desired arrangement of four flags.<br />
We are ready to prove the problem statement by induction on N; actually, above we have<br />
proved the base case N4. For the induction step, assume that N4, consider any 2 N¡2 1<br />
flags of length N, <strong>and</strong> arrange them into a large flag of sizeÔ2N¡2 1Õ¢N. This flag contains<br />
a non-monochrome column since the flags are distinct; we may assume that thisÎ2 column is the<br />
first one. By the pigeonhole principle, this column contains at leastÊ2 N¡2 1 N¡3 1<br />
2<br />
squares of one color (say, blue). We call the flags with a blue first square good.<br />
Consider all the good flags <strong>and</strong> remove the first square from each of them. We obtain at<br />
least 2 N¡3 1M N¡1 flags of length N¡1; by the induction hypothesis, N¡1 of them
27<br />
can form a square Q with the monochrome diagonal. Now, returning the removed squares, we<br />
obtain a rectangleÔN¡1Õ¢N, <strong>and</strong> our aim is to supplement it on the top by one more flag.<br />
If Q has a yellow diagonal, then we can take each flag with a yellow first square (it exists<br />
by a choice of the first column; moreover, it is not used in Q). Conversely, if the diagonal of Q<br />
is blue then we can take any of the2 N¡3 1¡ÔN¡1Õ0remaining good flags. So, in both<br />
cases we get a desired N¢N square.<br />
Solution 2. We present a different proof of the estimate M N2N¡2 1. We do not use the<br />
induction, involving Hall’s lemma on matchings instead.<br />
Consider arbitrary 2 N¡2 1 distinct flags <strong>and</strong> arrange them into a largeÔ2N¡2 1Õ¢N<br />
flag. Construct two bipartite graphs G yÔVV½, E yÕ<strong>and</strong> G bÔVV½, E bÕwith the<br />
common set of vertices as follows. Let V <strong>and</strong> V½be the set of columns <strong>and</strong> the set of flags<br />
under consideration, respectively. Next, let the edgeÔc, fÕappear in E y if the intersection of<br />
column c <strong>and</strong> flag f is yellow, <strong>and</strong>Ôc, fÕÈE otherwise. Then we have to prove exactly that<br />
one of the graphs G y <strong>and</strong> G b contains a matching with all the vertices of V involved.<br />
b<br />
two sets of columns S y , S<br />
yÕ<br />
bV such thatE y<br />
yÔS yÕS flag is connected to c either in G y or in G b , hence E yÔS yÕE 2 N¡2 1V½E yÔS E bÔS bÕS S<br />
So, we have S yS b∅. Let yS bS yÔS yÕE<br />
Assume that these matchings do not exist. By Hall’s lemma, it means that there exist<br />
y¡1 <strong>and</strong>E bÔS bÕS b¡1 (in the<br />
left-h<strong>and</strong> sides, E yÔS yÕ<strong>and</strong> E bÔS bÕdenote respectively the sets of all vertices connected to S y<br />
<strong>and</strong> S b in the corresponding graphs). Our aim is to prove that this is impossible. Note that<br />
S y , S bV since N2N¡2 1.<br />
First, suppose that S yS b∅,<br />
V¾V½Þ<br />
so there exists some cÈS yS b . Note that each<br />
bÔS bÕV½. Hence we have<br />
b¡22N¡4; this is impossible for N4.<br />
y, b. From the construction of our graph,<br />
bÕ<br />
we have that all the flags in the set E bÔS bÕ¨have blue squares in the<br />
columns of S y <strong>and</strong> yellow squares in the columns of S b . Hence the only undetermined positions<br />
in these flags are the remaining N¡y¡b ones, so 2 N¡y¡bV¾V½¡E yÔS yÕ¡E bÔS 2 N¡2 1¡Ôy¡1Õ¡Ôb¡1Õ, or, denoting cy b, 2 N¡c c2N¡2 2. This is impossible<br />
since Nc2.
28<br />
C3. 2500 chess kings have to be placed on a 100¢100 chessboard so that<br />
(i) no king can capture any other one (i.e. no two kings are placed in two squares sharing<br />
a common vertex);<br />
(ii) each row <strong>and</strong> each column contains exactly 25 kings.<br />
Find the number of such arrangements. (Two arrangements differing by rotation or symmetry<br />
are supposed to be different.)<br />
Answer. There are two such arrangements.<br />
(Russia)<br />
Solution. Suppose that we have an arrangement satisfying the problem conditions. Divide the<br />
board into 2¢2 pieces; we call these pieces blocks. Each block can contain not more than one<br />
king (otherwise these two kings would attack each other); hence, by the pigeonhole principle<br />
each block must contain exactly one king.<br />
Now assign to each block a letter T or B if a king is placed in its top or bottom half,<br />
respectively. Similarly, assign to each block a letter L or R if a king st<strong>and</strong>s in its left or right<br />
half. So we define T-blocks, B-blocks, L-blocks, <strong>and</strong> R-blocks. We also combine the letters; we call<br />
a block a TL-block if it is simultaneously T-block <strong>and</strong> L-block. Similarly we define TR-blocks,<br />
BL-blocks, <strong>and</strong> BR-blocks. The arrangement of blocks determines uniquely the arrangement of<br />
kings; so in the rest of the solution we consider the 50¢50 system of blocks (see Fig. 1). We<br />
identify the blocks by their coordinate pairs; the pairÔi, jÕ, where 1i, j50, refers to the<br />
jth block in the ith row (or the ith block in the jth column). The upper-left block isÔ1, 1Õ.<br />
The system of blocks has the following properties..<br />
(i½) IfÔi, jÕis a B-block thenÔi 1, jÕis a B-block: otherwise the kings in these two blocks<br />
can take each other. Similarly: ifÔi, jÕis a T-block thenÔi¡1, jÕis a T-block; ifÔi, jÕis an<br />
L-block thenÔi, j¡1Õis an L-block; ifÔi, jÕis an R-block thenÔi, j 1Õis an R-block.<br />
(ii½) Each column contains exactly 25 L-blocks <strong>and</strong> 25 R-blocks, <strong>and</strong> each row contains<br />
exactly 25 T-blocks <strong>and</strong> 25 B-blocks. In particular, the total number of L-blocks (or R-blocks,<br />
or T-blocks, or B-blocks) is equal to 25¤501250.<br />
Consider any B-block of the formÔ1, jÕ. By (i½), all blocks in the jth column are B-blocks;<br />
so we call such a column B-column. By (ii½), we have 25 B-blocks in the first row, so we obtain<br />
25 B-columns. These 25 B-columns contain 1250 B-blocks, hence all blocks in the remaining<br />
columns are T-blocks,<br />
<br />
<strong>and</strong> we obtain 25 T-columns. Similarly, there are exactly 25 L-rows <strong>and</strong><br />
exactly 25 R-rows.<br />
Now consider an arbitrary pair of a T-column <strong>and</strong> a neighboring B-column (columns with<br />
numbers j <strong>and</strong> j 1).<br />
<br />
<br />
1 2 3<br />
j j+1<br />
1 TL BL TL<br />
i<br />
L<br />
2 TL BR TR<br />
i+1<br />
3 BL BL TR<br />
T B<br />
Fig. 1 Fig. 2<br />
Case 1. Suppose that the jth column is a T-column, <strong>and</strong> theÔj 1Õth column is a B-<br />
column. Consider some index i such that the ith row is an L-row; thenÔi, j 1Õis a BL-block.<br />
Therefore,Ôi 1, jÕcannot be a TR-block (see Fig. 2), henceÔi 1, jÕis a TL-block, thus the
29<br />
Ôi 1Õth row is an L-row. Now, choosing the ith row to be the topmost L-row, we successively<br />
obtain that all rows from the ith to the 50th are L-rows. Since we have exactly 25 L-rows, it<br />
follows that the rows from the 1st to the 25th are R-rows, <strong>and</strong> the rows from the 26th to the<br />
50th are L-rows.<br />
Now consider the neighboring R-row <strong>and</strong> L-row (that are the rows with numbers 25 <strong>and</strong><br />
26). Replacing in the previous reasoning rows by columns <strong>and</strong> vice versa, the columns from the<br />
1st to the 25th are T-columns, <strong>and</strong> the columns from the 26th to the 50th are B-columns. So<br />
we have a unique arrangement of blocks that leads to the arrangement of kings satisfying the<br />
condition of the problem (see Fig. 3).<br />
<br />
<br />
<br />
<br />
TR TR BR BR<br />
TR TR BR BR<br />
<br />
<br />
TL TL BL BL<br />
TL TL BL BL<br />
Fig. 3 Fig. 4<br />
Case 2. Suppose that the jth column is a B-column, <strong>and</strong> theÔj 1Õth column is a T-column.<br />
Repeating the arguments from Case 1, we obtain that the rows from the 1st to the 25th are<br />
L-rows (<strong>and</strong> all other rows are R-rows), the columns from the 1st to the 25th are B-columns<br />
(<strong>and</strong> all other columns are T-columns), so we find exactly one more arrangement of kings (see<br />
Fig. 4).
30<br />
C4. Six stacks S 1 , . . ., S 6 of coins are st<strong>and</strong>ing in a row. In the beginning every stack contains<br />
a single coin. There are two types of allowed moves:<br />
Move 1: If stack S k with 1k5contains at least one coin, you may remove one coin<br />
from S k <strong>and</strong> add two coins to S k 1.<br />
Move 2: If stack S k with 1k4contains at least one coin, then you may remove<br />
one coin from S k <strong>and</strong> exchange stacks S k 1 <strong>and</strong> S k 2.<br />
Decide whether it is possible to achieve by a sequence of such moves that the first five stacks<br />
are empty, whereas the sixth stack S 6 contains exactly 2010 20102010 coins.<br />
C4½. Same as Problem C4, but the constant 2010 20102010 is replaced by 2010 2010 .<br />
2, 0ÕÔa¡1,<br />
(Netherl<strong>and</strong>s)<br />
Answer. Yes (in both variants of the problem). There exists such a sequence of moves.<br />
Solution. Denote byÔa 1 , a 2 , . . .,a nÕÔa½1, a½2, . . .,a½nÕthe following: if some consecutive stacks<br />
contain a 1 , . . .,a n coins, then it is possible to perform several allowed moves such that the stacks<br />
contain a½1 , . . .,a½n coins respectively, whereas the contents of the other stacks remain unchanged.<br />
Let A2010 2010 or A2010 20102010 , respectively. Our goal is to show that<br />
Ô1, 1, 1, 1, 1, 1ÕÔ0, 0, 0, 0, 0, AÕ.<br />
First we prove two auxiliary observations.<br />
Lemma 1.Ôa, 0, 0ÕÔ0, 2 a , 0Õfor every a1.<br />
Proof. We prove by induction thatÔa, 0, 0ÕÔa¡k, 2 k , 0Õfor every 1ka. For k1,<br />
apply Move 1 to the first stack:<br />
Ôa, 0, 0ÕÔa¡1, 2 1 , 0Õ.<br />
Now assume that ka<strong>and</strong> the statement holds for some ka. Starting fromÔa¡k, 2 k , 0Õ,<br />
apply Move 1 to the middle stack 2 k times, until it becomes empty. Then apply Move 2 to the<br />
first stack:<br />
ÐÓÑÓÒ Ð<br />
Ôa¡k, 2 k , 0ÕÔa¡k, 2 k 1 , 0Õ.<br />
Hence,<br />
Ôa, 0, 0ÕÔa¡k, 2 k 1 , 0Õ. Lemma 2. For every positive integer n, let P n2 2...2<br />
(e.g. P 322216). Then<br />
n<br />
2 k¡1, 2Õ¤¤¤Ôa¡k,<br />
2 k , 0ÕÔa¡k¡1,<br />
0, 2 k 1ÕÔa¡k¡1,<br />
Ôa, 0, 0, 0ÕÔ0, P a , 0, 0Õfor every a1.<br />
Proof. Similarly to Lemma 1, we prove thatÔa, 0, 0, 0ÕÔa¡k,<br />
For k1, apply Move 1 to the first stack:<br />
Ôa, 0, 0, 0ÕÔa¡1,<br />
2, 0, 0ÕÔa¡1,<br />
P k , 0, 0Õfor every 1ka.<br />
P 1 , 0, 0Õ.<br />
Ð<br />
P k 1, 0, 0Õ.<br />
k 1, 0, 0Õ.<br />
Now assume that the lemma holds for some ka. Starting fromÔa¡k, P k , 0, 0Õ, apply<br />
Lemma 1, then apply Move 1 to the first stack:<br />
Therefore,<br />
Ôa¡k, P k , 0, 0ÕÔa¡k,<br />
0, 2 P k<br />
, 0ÕÔa¡k,<br />
Ôa, 0, 0, 0ÕÔa¡k,<br />
P k , 0, 0ÕÔa¡k¡1,<br />
0, P k 1, 0ÕÔa¡k¡1,<br />
P
31<br />
0Õ<br />
Now we prove the statement of the problem.<br />
First apply Move 1 to stack 5, then apply Move 2 to stacks S 4 , S 3 , S 2 <strong>and</strong> S 1 in this order.<br />
Then apply Lemma 2 twice:<br />
Ô1, 1, 1, 1, 1, 1ÕÔ1, 1, 1, 1, 0, 3ÕÔ1, 1, 1, 0, 3, 0ÕÔ1, 1, 0, 3, 0, 0ÕÔ1, 0, 3, 0, 0,<br />
Ô0, 3, 0, 0, 0, 0ÕÔ0, 0, P 3 , 0, 0, 0ÕÔ0, 0, 16, 0, 0, 0ÕÔ0, 0, 0, P 16 , 0, 0Õ.<br />
We already have more than A coins in stack S 4 , since<br />
A2010 20102010Ô2 11Õ201020102 11¤201020102 201020112Ô211Õ20112 211¤20112 2215P 16 .<br />
To decrease the number of coins in stack S 4<br />
0Õ<br />
, apply Move 2 to this stack repeatedly until its<br />
size decreases to Aß4. (In every step, we remove a coin from S 4 <strong>and</strong> exchange the empty stacks<br />
0, 0, P 16¡1, 0, 0ÕÔ0, 0, 0, P 16¡2, 0,<br />
¤¤¤Ô0, 0, 0, Aß4, 0, 0Õ.<br />
Finally, apply Move 1 repeatedly to empty stacks S 4 <strong>and</strong> S 5 :<br />
Ô0, 0, 0, Aß4, 0, 0Õ¤¤¤Ô0, 0, 0, 0, Aß2, 0Õ¤¤¤Ô0, 0, 0, 0, 0, AÕ.<br />
Comment 1. Starting with only 4 stack, it is not hard to check manually that we can achieve at<br />
most 28 coins in the last position. However, around 5 <strong>and</strong> 6 stacks the maximal number of coins<br />
Ô1,1,1,1,1,1ÕÔ0,3,1,1,1,1ÕÔ0,1,5,1,1,1ÕÔ0,1,1,9,1,1Õ<br />
explodes. With 5 stacks it is possible to achieve more than 2 214 coins. With 6 stacks the maximum is<br />
greater than P . P2 14<br />
It is not hard to show that the numbers 2010 2010 <strong>and</strong> 2010 20102010 in the problem can be replaced<br />
by any nonnegative integer up to P . P2 14<br />
Comment 2. The simpler variant C4½of the problem can be solved without Lemma 2:<br />
Ô0,1,1,1,17,1ÕÔ0,1,1,1,0,35ÕÔ0,1,1,0,35,0ÕÔ0,1,0,35,0,0Õ<br />
Ô0,0,35,0,0,0ÕÔ0,0,1,2 34 ,0,0ÕÔ0,0,1,0,2 234 ,0ÕÔ0,0,0,2 ,0,0Õ<br />
234<br />
S 5 <strong>and</strong> S 6 .)Ô0, 0, 0, P 16 , 0, 0ÕÔ0,<br />
Ô0,0,0,2 234¡1,0,0Õ...Ô0,0,0,Aß4,0,0ÕÔ0,0,0,0,Aß2,0ÕÔ0,0,0,0,0,AÕ.<br />
For this reason, the PSC suggests to consider the problem C4 as well. Problem C4 requires more<br />
invention <strong>and</strong> technical care. On the other h<strong>and</strong>, the problem statement in C4½hides the fact that the<br />
resulting amount of coins can be such incredibly huge <strong>and</strong> leaves this discovery to the students.
32<br />
C5. n4 players participated in a tennis tournament. Any two players have played exactly<br />
one game, <strong>and</strong> there was no tie game. We call a company of four players bad if one player<br />
was defeated by the other three players, <strong>and</strong> each of these three players won a game <strong>and</strong> lost<br />
another game among themselves. Suppose that there is no bad company in this tournament.<br />
Let w i <strong>and</strong> l i be respectively the number of wins <strong>and</strong> losses of the ith player. Prove that<br />
nôi1Ôw i¡l iÕ30. (1)<br />
(South Korea)<br />
Solution. For any tournament T satisfying the problem condition, denote by SÔTÕsum under<br />
consideration, namely<br />
SÔTÕnôi1Ôw i¡l iÕ3 .<br />
4<br />
4<br />
First, we show that the statement holds if a tournament T has only 4 players. Actually, let<br />
AÔa 1 , a 2 , a 3 , a 4Õbe the number of wins of the players; we may assume that a 1a 2a 3a 4 .<br />
We have a 1 a 2 a 3 a 2¨6, hence a 41. If a 40, then we cannot have<br />
a 1a 2a 32, otherwise the company of all players is bad. Hence we should have<br />
AÔ3, 2, 1, 0Õ, <strong>and</strong> SÔTÕ33 1 3 Ô¡1Õ3 Ô¡3Õ30. On the other h<strong>and</strong>, if a 41, then<br />
only two possibilities, AÔ3, 1, 1, 1Õ<strong>and</strong> AÔ2, 2, 1, 1Õcan take place. In the former case we<br />
have SÔTÕ33 3¤Ô¡2Õ30, while in the latter one SÔTÕ13 1 3 Ô¡1Õ3 SÔTÕô<br />
Ô¡1Õ30, as<br />
desired.<br />
Now we turn to the general problem. Consider a tournament T with no bad companies <strong>and</strong><br />
enumerate the players by the numbers from 1 to n. For every 4 players i 1 , i 2 , i 3 , i 4 consider a<br />
“sub-tournament” T i1 i 2 i 3 i 4<br />
consisting of only these players <strong>and</strong> the games which they performed<br />
with each other. By the abovementioned, we have SÔT i1 i 2 i 3 i 4Õ0. Our aim is to prove that<br />
SÔT i1 i 2 i 3 i 4Õ, (2)<br />
i 1 ,i 2 ,i 3 ,i 4<br />
ε ij«3<br />
ôε ij1 ε ij2 ε ij3 .<br />
j 1 ,j 2 ,j 3i<br />
where the sum is taken over all 4-tuples of distinct numbers from the setØ1, . . .,nÙ. This way<br />
the problem statement will be established.<br />
We interpret the numberÔw i¡l iÕ3 as following. For ij, let ε ij1 if the ith player wins<br />
against the jth one, <strong>and</strong> ε ij¡1 otherwise. Then<br />
Ôw i¡l iÕ3£ôji SÔTÕ ô<br />
Hence,<br />
ij1 ε ij2 ε ij3 .<br />
,j 2 ,j 3Ùε<br />
SÔTÕ ô<br />
iÊØj 1<br />
To simplify this expression, consider all the terms in this sum where two indices are equal.<br />
Øi,j 1 ,j 2 ,j 3Ù4<br />
If, for instance, j 1j 2 , then the term contains ε 2 ij 11, so we can replace this term by ε ij3 .<br />
Make such replacements for each such term; obviously, after this change each term of the form<br />
ε ij3 will appear PÔTÕtimes, hence<br />
PÔTÕS 2ÔTÕ.<br />
ε<br />
1ÔTÕ ij1 ε ij2 ε ij3 ε PÔTÕôij<br />
ijS
33<br />
We show that S 2ÔTÕ0<strong>and</strong> hence SÔTÕS 1ÔTÕfor each tournament. Actually, note that<br />
ε ij¡ε ji , <strong>and</strong> the whole sum can<br />
1ÔTÕô<br />
be split into such pairs. Since the sum in each pair is 0, so<br />
is S 2ÔTÕ.<br />
Thus the desired equality (2) rewrites as<br />
S S 1ÔT i1 i 2 i 3 i 4Õ. (3)<br />
i 1 ,i 2 ,i 3 ,i 4<br />
Now, if all the numbers j 1 , j 2 , j 3 are distinct, then the setØi, j 1 , j 2 , j 3Ùis contained in exactly<br />
one 4-tuple, hence the term ε ij1 ε ij2 ε ij3 appears in the right-h<strong>and</strong> part of (3) exactly once, as<br />
well as in the left-h<strong>and</strong> part. Clearly, there are no other terms in both parts, so the equality is<br />
established.<br />
Solution 2. Similarly to the first solution, we call the subsets of players as companies, <strong>and</strong><br />
the k-element subsets will be called as k-companies.<br />
In any company of the players, call a player the local champion of the company if he defeated<br />
all other members of the company. Similarly, if a player lost all his games against the others<br />
in the company then call him the local loser of the company. Obviously every company has<br />
at most one local champion <strong>and</strong> at most one local loser. By the condition of the problem,<br />
whenever a 4-company has a local loser, then this company has a local champion as well.<br />
Suppose that k is some positive integer, <strong>and</strong> let us count all cases when a player is the local<br />
champion of some k-company. The ith player won against w i other player. To be the local<br />
champion of a k-company, he must be a member of the company, <strong>and</strong> the other k¡1 members<br />
must be chosen from those whom he defeated. Therefore, the ith player is the local champion<br />
of¢w i<br />
k¡1Åk-companies. Hence, the total number of local champions of all k-companies is<br />
nôi1¢w i<br />
k¡1Å.<br />
Similarly, the total number of local losers of the k-companies is<br />
Now apply this for k2, 3 <strong>and</strong> 4.<br />
Since every game has a winner <strong>and</strong> a loser, we have<br />
w<br />
nôi1<br />
inôi1<br />
nôi1¢l i<br />
k¡1Å.<br />
i¢n<br />
l <strong>and</strong> hence<br />
2Å,<br />
nôi1 w i¡l i¨0. (4)<br />
In every 3-company, either the players defeated one another in a cycle or the company has<br />
both a local champion <strong>and</strong> a local loser. Therefore, the total number of local champions <strong>and</strong><br />
nôi1¢w i<br />
local losers in the 3-companies is the same,<br />
2Ånôi1¢l i<br />
2Å. So we have<br />
nôi1£¢w 2Å¡¢l i<br />
2Å«0. i<br />
(5)<br />
In every 4-company, by the problem’s condition, the number of local losers is less than or<br />
equal to the number of local champions. Then the same holds for the total numbers of local
34<br />
champions <strong>and</strong> local losers in all 4-companies, so<br />
nôi1¢w i<br />
3Ånôi1¢l i<br />
3Å. Hence,<br />
3Å«<br />
nôi1£¢w 3Å¡¢l i<br />
3Å«0. i<br />
(6)<br />
2Å«¡<br />
3Å«<br />
Ôx¡yÕ324£¢x 3Å¡¢y 24£¢x 2Å¡¢y<br />
2Å«¡<br />
3Ôx yÕ2¡4¨Ôx¡yÕ.<br />
iÕ324£¢w 3Å¡¢l<br />
Ôw i<br />
24£¢w 2Å¡¢l i i<br />
i¡l ÐÓÓÓÓÓÓÓÓÓÑÓÓÓÓÓÓÓÓÓÒ 3Å« ÐÓÓÓÓÓÓÓÓÓÑÓÓÓÓÓÓÓÓÓÒ 2Å«<br />
3Ôn¡1Õ2¡4¨¡w<br />
ÐÓÓÓÓÓÑÓÓÓÓÓÒ<br />
i¡l i©.<br />
nôi1£¢w nôi1£¢w 3Å¡¢l i i<br />
2Å¡¢l i i<br />
Now we establish the problem statement (1) as a linear combination of (4), (5) <strong>and</strong> (6). It<br />
is easy check that<br />
Apply this identity to xw 1 <strong>and</strong> yl i . Since every player played n¡1 games, we have<br />
w i l in¡1, <strong>and</strong> thus<br />
i<br />
i©<br />
Then<br />
nôi1Ôw i¡l iÕ324<br />
3Ôn¡1Õ2¡4¨nôi1¡w i¡l 0.<br />
0<br />
24<br />
0<br />
¡<br />
0
35<br />
C6. Given a positive integer k <strong>and</strong> other two integers bw1. There are two strings of<br />
pearls, a string of b black pearls <strong>and</strong> a string of w white pearls. The length of a string is the<br />
number of pearls on it.<br />
One cuts these strings in some steps by the following rules. In each step:<br />
(i) The strings are ordered by their lengths in a non-increasing order. If there are some<br />
strings of equal lengths, then the white ones precede the black ones. Then k first ones (if they<br />
consist of more than one pearl) are chosen; if there are less than k strings longer than 1, then<br />
one chooses all of them.<br />
(ii) Next, one cuts each chosen string into two parts differing in length by at most one.<br />
(For instance, if there are strings of 5, 4, 4, 2 black pearls, strings of 8, 4, 3 white pearls <strong>and</strong><br />
k4, then the strings of 8 white, 5 black, 4 white <strong>and</strong> 4 black pearls are cut into the parts<br />
Ô4, 4Õ,Ô3, 2Õ,Ô2, 2Õ<strong>and</strong>Ô2, 2Õ, respectively.)<br />
The process stops immediately after the step when a first isolated white pearl appears.<br />
Prove that at this stage, there will still exist a string of at least two black pearls.<br />
(Canada)<br />
Solution 1. Denote the situation after the ith step by A i ; hence A 0 is the initial situation, <strong>and</strong><br />
A i¡1A i is the ith step. We call a string containing m pearls an m-string; it is an m-w-string<br />
or a m-b-string if it is white or black, respectively.<br />
We continue the process until every string consists of a single pearl. We will focus on three<br />
moments of the process: (a) the first stage A s when the first 1-string (no matter black or<br />
white) appears; (b) the first stage A t where the total number of strings is greater than k (if<br />
such moment does not appear then we put t); <strong>and</strong> (c) the first stage A f when all black<br />
pearls are isolated. It is sufficient to prove that in A f¡1 (or earlier), a 1-w-string appears.<br />
We start with some easy properties of the situations under consideration. Obviously, we<br />
have sf. Moreover, all b-strings from A f¡1 become single pearls in the fth step, hence all<br />
of them are 1- or 2-b-strings.<br />
Next, observe that in each step A iA i 1 with it¡1, allÔ1Õ-strings were cut since<br />
there are not more than k strings at all; if, in addition, is, then there were no 1-string, so<br />
all the strings were cut in this step.<br />
Now, let B i <strong>and</strong> b i be the lengths of the longest <strong>and</strong> the shortest b-strings in A i , <strong>and</strong><br />
let W i <strong>and</strong> w i be the same for w-strings. We show by induction on iminØs, tÙthat (i) the<br />
situation A i contains exactly 2 i black <strong>and</strong> 2 i white strings, (ii) B iW i , <strong>and</strong> (iii) b iw i .<br />
The base case i0is obvious. For the induction step, if iminØs, tÙthen in the ith step,<br />
each string is cut, thus the claim (i) follows from the induction hypothesis; next, we have<br />
B iÖB i¡1ß2×ÖW i¡1ß2×W i <strong>and</strong> b iØb i¡1ß2ÙØw i¡1ß2Ùw i , thus establishing (ii)<br />
<strong>and</strong> (iii).<br />
For the numbers s, t, f, two cases are possible.<br />
Case 1. Suppose that stor ft 1 (<strong>and</strong> hence st 1); in particular, this is true<br />
when t. Then in A s¡1 we have B s¡1W s¡1, b s¡1w s¡11as s¡1minØs, tÙ.<br />
Now, if sf, then in A s¡1, there is no 1-w-string as well as noÔ2Õ-b-string. That is,<br />
2B s¡1W s¡1b s¡1w s¡11, hence all these numbers equal 2. This means that<br />
in A s¡1, all strings contain 2 pearls, <strong>and</strong> there are 2 s¡1 black <strong>and</strong> 2 s¡1 white strings, which<br />
means b2¤2s¡1w. This contradicts the problem conditions.<br />
Hence we have sf¡1 <strong>and</strong> thus st. Therefore, in the sth step each string is cut<br />
into two parts. Now, if a 1-b-string appears in this step, then from w s¡1b s¡1 we see that a
36<br />
1-w-string appears as well; so, in each case in the sth step a 1-w-string appears, while not all<br />
black pearls become single, as desired.<br />
Case 2. Now assume that t 1s<strong>and</strong> t 2f. Then in A t we have exactly 2 t white<br />
<strong>and</strong> 2 t black strings, all being larger than 1, <strong>and</strong> 2 t 1k2t (the latter holds since 2 t is the<br />
total number of strings in A t¡1). Now, in theÔt 1Õst step, exactly k strings are cut, not more<br />
than 2 t of them being black; so the number of w-strings in A t 1 is at least 2 t Ôk¡2tÕk.<br />
Since the number of w-strings does not decrease in our process, in A f¡1 we have at least k<br />
white strings as well.<br />
Finally, in A f¡1, all b-strings are not larger than 2, <strong>and</strong> at least one 2-b-string is cut in<br />
the fth step. Therefore, at most k¡1 white strings are cut in this step, hence there exists a<br />
w-string W which is not cut in the fth step. On the other h<strong>and</strong>, since a 2-b-string is cut, all<br />
Ô2Õ-w-strings should also be cut in the fth step; hence W should be a single pearl. This is<br />
exactly what we needed.<br />
Comment. In this solution, we used the condition bw only to avoid the case bw2 t . Hence,<br />
if a number bw is not a power of 2, then the problem statement is also valid.<br />
Solution 2. We use the same notations as introduced in the first paragraph of the previous<br />
solution. We claim that at every stage, there exist a u-b-string <strong>and</strong> a v-w-string such that<br />
either<br />
(i) uv1, or<br />
(ii) 2uv2u, <strong>and</strong> there also exist k¡1 ofÔvß2Õ-strings other than considered<br />
above.<br />
First, we notice that this statement implies the problem statement. Actually, in both<br />
cases (i) <strong>and</strong> (ii) we have u1, so at each stage there exists aÔ2Õ-b-string, <strong>and</strong> for the last<br />
stage it is exactly what we need.<br />
Now, we prove the claim by induction on the number of the stage. Obviously, for A 0 the<br />
condition (i) holds since bw. Further, we suppose that the statement holds for A i , <strong>and</strong> prove<br />
it for A i 1. Two cases are possible.<br />
Case 1. Assume that in A i , there are a u-b-string <strong>and</strong> a v-w-string with uv. We can<br />
assume that v is the length of the shortest w-string in A i ; since we are not at the final stage,<br />
we have v2. Now, in theÔi 1Õst step, two subcases may occur.<br />
Subcase 1a. Suppose that either no u-b-string is cut, or both some u-b-string <strong>and</strong> some<br />
v-w-string are cut. Then in A i 1, we have either a u-b-string <strong>and</strong> aÔvÕ-w-string (<strong>and</strong> (i) is<br />
valid), or we have aÖuß2×-b-string <strong>and</strong> aØvß2Ù-w-string. In the latter case, from uv we get<br />
Öuß2×Øvß2Ù, <strong>and</strong> (i) is valid again.<br />
Subcase 1b. Now, some u-b-string is cut, <strong>and</strong> no v-w-string is cut (<strong>and</strong> hence all the strings<br />
which are cut are longer than v). If u½Öuß2×v, then the condition (i) is satisfied since we<br />
have a u½-b-string <strong>and</strong> a v-w-string in A i 1. Otherwise, notice that the inequality uv2<br />
implies u½2. Furthermore, besides a fixed u-b-string, other k¡1 ofÔv 1Õ-strings should<br />
be cut in theÔi 1Õst step, hence providing at least k¡1 ofÔÖÔv 1Õß2×Õ-strings, <strong>and</strong><br />
ÖÔv 1Õß2×vß2. So, we can put v½v, <strong>and</strong> we have u½vu2u½, so the condition (ii)<br />
holds for A i 1.<br />
Case 2. Conversely, assume that in A i there exist a u-b-string, a v-w-string (2uv2u)<br />
<strong>and</strong> a set S of k¡1 other strings larger than vß2 (<strong>and</strong> hence larger than 1). In theÔi 1Õst<br />
step, three subcases may occur.<br />
Subcase 2a. Suppose that some u-b-string is not cut, <strong>and</strong> some v-w-string is cut. The latter<br />
one results in aØvß2Ù-w-string, we have v½Øvß2Ùu, <strong>and</strong> the condition (i) is valid.
37<br />
Subcase 2b. Next, suppose that no v-w-string is cut (<strong>and</strong> therefore no u-b-string is cut as<br />
uv). Then all k strings which are cut have the lengthv, so each one results in aÔvß2Õstring.<br />
Hence in A i 1, there exist kk¡1 ofÔvß2Õ-strings other than the considered u- <strong>and</strong><br />
v-strings, <strong>and</strong> the condition (ii) is satisfied.<br />
Subcase 2c. In the remaining case, all u-b-strings are cut. This means that allÔuÕ-strings<br />
are cut as well, hence our v-w-string is cut. Therefore in A i 1 there exists aÖuß2×-b-string<br />
together with aØvß2Ù-w-string. Now, if u½Öuß2×Øvß2Ùv½then the condition (i) is<br />
fulfilled. Otherwise, we have u½v½u2u½. In this case, we show that u½2. If, to the<br />
contrary, u½1 (<strong>and</strong> hence u2), then all black <strong>and</strong> whiteÔ2Õ-strings should be cut in the<br />
Ôi 1Õst step, <strong>and</strong> among these strings there are at least a u-b-string, a v-w-string, <strong>and</strong> k¡1<br />
strings in S (k 1 strings altogether). This is impossible.<br />
Hence, we get 2u½v½2u½. To reach (ii), it remains to check that in A i 1, there exists<br />
a set S½of k¡1 other strings larger than v½ß2. These will be exactly the strings obtained from<br />
the elements of S. Namely, each sÈS was either cut in theÔi 1Õst step, or not. In the former<br />
case, let us include into S½the largest of the strings obtained from s; otherwise we include s<br />
itself into S½. All k¡1 strings in S½are greater than vß2v½, as desired.
38<br />
C7. Let P 1 , . . .,P s be arithmetic progressions of integers, the following conditions being<br />
satisfied:<br />
(i) each integer belongs to at least one of them;<br />
(ii) each progression contains a number which does not belong to other progressions.<br />
Denote by n the least common multiple of steps of these progressions; let np α 1<br />
1 . . .p α k<br />
k<br />
be<br />
its prime factorization. Prove that<br />
s1 kôi1α iÔp i¡1Õ.<br />
N<br />
(Germany)<br />
Solution 1. First, we prove the key lemma, <strong>and</strong> then we show how to apply it to finish the<br />
k grid we mean the set<br />
ØÔa 1 , . . ., a kÕ:a iÈZ, 0a in i¡1Ù; the elements of N will be referred to as points. In this<br />
grid, we define a subgrid as a subset of the form<br />
: b i1x i1 , . . .,b itx itÙ, (1)<br />
where IØi 1 , . . .,i tÙis an arbitrary nonempty set of indices, <strong>and</strong> x ijÈÖ0, n ij¡1×(1jt)<br />
are fixed integer numbers. Further, we say that a subgrid (1) is orthogonal to the ith coordinate<br />
axis if iÈI, <strong>and</strong> that it is parallel to the ith coordinate axis otherwise.<br />
Lemma. Assume that the grid N is covered by subgrids L 1 , L 2 , . . .,L s (this means Näs<br />
i1 L i)<br />
so that<br />
(ii½) each subgrid contains a point which is not covered by other subgrids;<br />
(iii) for each coordinate axis, there exists a subgrid L i orthogonal to this axis.<br />
Then<br />
solution.<br />
Let n 1 , . . .,n k be positive integers. By an n 1¢n 2¢¤¤¤¢n<br />
LØÔb 1 , . . .,b kÕÈN<br />
s1 kôi1Ôn i¡1Õ.<br />
Proof. Assume to the contrary that siÔn i¡1Õs½. Our aim is to find a point that is not<br />
covered by L 1 , . . ., L s .<br />
The idea of the proof is the following. Imagine that we exp<strong>and</strong> each subgrid to some maximal<br />
subgrid so that for the ith axis, there will be at most n i¡1 maximal subgrids orthogonal to<br />
this axis. Then the desired point can be found easily: its ith coordinate should be that not<br />
covered by the maximal subgrids orthogonal to the ith axis. Surely, the conditions for existence<br />
of such expansion are provided by Hall’s lemma on matchings. So, we will follow this direction,<br />
although we will apply Hall’s lemma to some subgraph instead of the whole graph.<br />
Construct a bipartite graph GÔVV½, EÕas follows. Let VØL 1 , . . .,L sÙ, <strong>and</strong> let<br />
V½Øv ij : 1is, 1jn i¡1Ùbe some set of s½elements. Further, let the edgeÔL m , v<br />
appear iff L m is orthogonal to the ith axis.<br />
For each subset WV , denote<br />
ijÕ<br />
fÔWÕØvÈV½:ÔL, vÕÈE for some LÈWÙ.<br />
Notice that fÔVÕV½by (iii).<br />
Now, consider the set WV containing the maximal number of elements such thatW<br />
fÔWÕ; if there is no such set then we set W∅. Denote W½fÔWÕ, UVÞW, U½V½ÞW½.
39<br />
By our<br />
WXW<br />
assumption <strong>and</strong><br />
XfÔWÕ<br />
the Lemma condition,fÔVÕV½V, hence WV <strong>and</strong> U∅.<br />
Permuting the coordinates, we can assume that U½Øv ij : 1ilÙ, W½Øv ij : l 1ikÙ.<br />
Consider the induced subgraph G½of G on the vertices UU½. We claim that for every<br />
XU, we getfÔXÕU½X(so G½satisfies the conditions of Hall’s lemma). Actually, we<br />
haveWfÔWÕ, so ifXfÔXÕU½for some XU, then we have<br />
fÔXÕU½fÔWÕÔfÔXÕU½ÕfÔWXÕ.<br />
This contradicts the maximality ofW.<br />
Thus, applying Hall’s lemma, we can assign to each LÈU some vertex v ijÈU½so that to<br />
distinct elements of U, distinct vertices of U½are assigned. In this situation, we say that LÈU<br />
corresponds to the ith axis, <strong>and</strong> write gÔLÕi. Since there are n i¡1 vertices of the form v ij ,<br />
we get that for each 1il, not more than n i¡1 subgrids correspond to the ith axis.<br />
Finally, we are ready to present the desired point. Since WV , there exists a point<br />
bÔb 1 , b 2 , . . .,b kÕÈNÞÔLÈWLÕ. On the other h<strong>and</strong>, for every 1il, consider any subgrid<br />
LÈU with gÔLÕi. This means exactly that L is orthogonal to the ith axis, <strong>and</strong> hence all<br />
its elements have the same ith coordinate c L . Since there are at most n i¡1 such subgrids,<br />
there exists a number 0a in i¡1 which is not contained in a setØc L : gÔLÕiÙ. Choose<br />
such number for every 1il. Now we claim that point aÔa 1 , . . .,a l , b l 1, . . .,b kÕis not<br />
covered, hence contradicting the Lemma condition.<br />
Surely, point a cannot lie in some LÈU, since all the points in L have gÔLÕth coordinate<br />
c La gÔLÕ. Ð<br />
On the other h<strong>and</strong>, suppose that aÈL for some LÈW; recall that bÊL. But the<br />
points a <strong>and</strong> b differ only at first l coordinates, so L should be orthogonal to at least one of<br />
the first l axes, <strong>and</strong> hence our graph contains some edgeÔL, v ijÕfor il. It contradicts the<br />
definition of W½. The Lemma is proved.<br />
Now we turn to the problem. Let d j be the step of the progression P j . Note that since<br />
nl.c.m.Ôd ÐÓÓÓÓÓÓÑÓÓÓÓÓÓÒ<br />
1 , . . .,d sÕ, for each<br />
ÐÓÓÓÓÓÓÑÓÓÓÓÓÓÒ<br />
1ik there exists an index jÔiÕsuch that p α i<br />
i§d jÔiÕ. rÔmÕ<br />
We<br />
assume that n1; otherwise the problem statement is trivial.<br />
For each 0mn¡1 <strong>and</strong> 1ik, let m i be the residue of m modulo p α i<br />
i , <strong>and</strong> let<br />
m ir iαi . . .r i1 be the base p i representation of m i (possibly, with some leading zeroes). Now,<br />
we put into correspondence to m the sequence rÔmÕÔr 11 , . . .,r 1α1 , r 21 , . . .,r kαkÕ. Hence<br />
lies in a p 1¢¤¤¤¢p 1 ¢¤¤¤¢p k¢¤¤¤¢p k grid N.<br />
α 1 times<br />
α k times<br />
Surely, if rÔmÕrÔm½Õthen p α i<br />
i§m i¡m½i , which follows pα i<br />
i§m¡m½for all 1ik;<br />
consequently, n§m¡m½. So, when m runs over the setØ0, . . .,n¡1Ù, the sequences rÔmÕdo<br />
not repeat; sinceNn, this means that r is a bijection betweenØ0, . . ., n¡1Ù<strong>and</strong> N. Now<br />
we will show that for each 1is, the set L iØrÔmÕ:mÈP iÙis a subgrid, <strong>and</strong> that for<br />
each axis there exists<br />
<br />
a subgrid orthogonal to this axis. Obviously, these subgrids cover N, <strong>and</strong><br />
the condition (ii½) follows directly from (ii). Hence the Lemma provides exactly the estimate<br />
we need.<br />
Consider some 1js<strong>and</strong> let d jp rÔqÕ<br />
γ 1<br />
1 . . .pγ k<br />
k . Consider some qÈP j <strong>and</strong> let<br />
Ôr 11 , . . ., r kαkÕ. Then for an arbitrary q½, setting rÔq½ÕÔr½11, . . .,r½kα kÕwe have<br />
p γ i<br />
i§q¡q½for each 1ik r i,tr½i,t for all tγ i .<br />
Hence L jØÔr½11 , . . .,r½kα kÕÈN : r i,tr½i,t for all tγ iÙwhich means that L j is a subgrid<br />
containing rÔqÕ. Moreover, in L jÔiÕ, all the coordinates corresponding to p i are fixed, so it is<br />
orthogonal to all of their axes, as desired.<br />
q½ÈP j
40<br />
Comment 1. The estimate in the problem is sharp for every n. One of the possible examples is the<br />
following one. For each 1ik, 0jα i¡1, 1kp¡1, let<br />
P i,j,kkp j i<br />
p j 1<br />
i<br />
Z,<br />
<strong>and</strong> add the progression P 0nZ. One can easily check that this set satisfies all the problem conditions.<br />
There also exist other examples.<br />
On the other h<strong>and</strong>, the estimate can be adjusted in the following sense. For every 1ik, let<br />
0α i0 ,α i1 ,... ,α ihi be all the numbers of the form ord piÔd jÕin an increasing order (we delete the<br />
repeating occurences of a number, <strong>and</strong> add a number 0α i0 if it does not occur). Then, repeating<br />
the arguments from the solution one can obtain that<br />
s1 kôi1<br />
h i ôj1Ôp α j¡α j¡1¡1Õ.<br />
Note that p α¡1αÔp¡1Õ, <strong>and</strong> the equality is achieved only for α1. Hence, for reaching the<br />
minimal number of the progressions, one should have α i,jj for all i,j. In other words, for each<br />
1jα i , there should be an index t such that ord piÔd tÕj.<br />
Solution 2. We start with introducing some notation. For positive integer r, we denote<br />
Ör×Ø1, 2, . . .,rÙ. Next, we say that a set of progressions PØP 1 , . . .,P sÙcover Z if each<br />
integer belongs to some of them; we say that this covering is minimal if no proper subset of P<br />
covers Z. Obviously, each covering contains a minimal subcovering.<br />
Next, for a minimal coveringØP 1 , . . .,P sÙ<strong>and</strong> for every 1is, let d i be the step of<br />
progression P i , <strong>and</strong> h i be some number which is contained in P i but in none of the other<br />
progressions. We assume that n1, otherwise the problem is trivial. This implies d i1,<br />
otherwise the progression P i covers all the numbers, <strong>and</strong> n1.<br />
tÙÖk×<br />
We will prove a more general statement, namely the following<br />
Claim. Assume that the progressions P 1 , . . .,P s <strong>and</strong> number np α 1<br />
1 . . .pα k<br />
k1 are chosen as<br />
in the problem statement. Moreover, choose some nonempty set of indices IØi 1 , . . .,i<br />
<strong>and</strong> some positive integer β iα i for every iÈI. Consider the set of indices<br />
Tj : 1js, <strong>and</strong> p α i¡β i 1<br />
i<br />
§d j for some iÈIµ.<br />
Then<br />
ôiÈI<br />
T1 β iÔp i¡1Õ. (2)<br />
Observe that the Claim for IÖk×<strong>and</strong> β iα i implies the problem statement, since the<br />
left-h<strong>and</strong> side in (2) is not greater than s. Hence, it suffices to prove the Claim.<br />
1. First, we prove the Claim assuming that all d j ’s are prime numbers. If for some 1ik<br />
we have at least p i progressions with the step p i , then they do not intersect <strong>and</strong> hence cover all<br />
the integers; it means that there are no other progressions, <strong>and</strong> np i ; the Claim is trivial in<br />
this case.<br />
Now assume that for every 1ik, there are not more than p i¡1 progressions with<br />
step p i ; each such progression covers the numbers with a fixed residue modulo p i , therefore<br />
there exists a residue q i mod p i which is not touched by these progressions. By the Chinese<br />
Remainder Theorem, there exists a number q such that qq iÔmod p iÕfor all 1ik; this<br />
number cannot be covered by any progression with step p i , hence it is not covered at all. A<br />
contradiction.
41<br />
2. Now, we assume that the general Claim is not valid, <strong>and</strong> hence we consider a counterexampleØP<br />
1 , . . .,P sÙfor the Claim; we can choose it to be minimal in the following sense:<br />
the number n is minimal possible among all the counterexamples;<br />
the sumi d i is minimal possible among all the counterexamples having the chosen value<br />
of n.<br />
As was mentioned above, not all numbers d i are primes; hence we can assume that d 1 is<br />
composite, say p 1§d 1 <strong>and</strong> d½1d 1<br />
p 11. Consider a progression P½1 having the step d½1, <strong>and</strong><br />
containing P 1 . We will focus on two coverings constructed as follows.<br />
(i) Surely, the progressions P½1, P 2 , . . .,P s cover Z, though this covering in not necessarily<br />
minimal. So, choose some minimal subcovering P½in it; surely P½1ÈP½since h 1 is not covered<br />
by P 2 , . . .,P s , so we may assume that P½ØP½1 P 2, . . ., P s½Ùfor some s½s. Furthermore, the<br />
,<br />
period of the covering P½can appear to be less than n; so we denote this period by<br />
l.c.m. d½1, d 2 , . . ., d s½¨.<br />
n½p α 1¡σ 1<br />
1 . . .p α k¡σ k<br />
k<br />
Observe that for each P jÊP½, we have h jÈP½1, otherwise h j would not be covered by P.<br />
(ii) On the other h<strong>and</strong>, each nonempty set of the form R iP iP½1 (1is) is also a<br />
progression with a step r il.c.m.Ôd i , d½1Õ, <strong>and</strong> such sets cover P½1. Scaling these progressions<br />
with the ratio 1ßd½1 , we obtain the progressions Q i with steps q ir ißd½1 which cover Z. Now we<br />
choose a minimal subcovering Q of this covering; again we should have Q 1ÈQ by the reasons<br />
of h 1 . Now, denote the period of Q by<br />
n¾l.c.m.Øq i : Q iÈQÙl.c.m.Ør i : Q iÈQÙ<br />
d½1<br />
p γ 1<br />
1 . . .p γ k<br />
k<br />
.<br />
d½1<br />
Note that if h jÈP½1 , then the image of h j under the scaling can be covered by Q j only; so, in<br />
this case we have Q jÈQ.<br />
Our aim is to find the desired number of progressions in coverings P <strong>and</strong> Q. First, we have<br />
nn½, <strong>and</strong> the sum of the steps in P½is less than that in P; hence the Claim is valid for P½.<br />
We apply it to the set of indices I½ØiÈI : β iσ iÙ<strong>and</strong> the exponents β½iβ i¡σ i ; hence<br />
the set under consideration is<br />
T½j : 1js½, <strong>and</strong> i¡σ pÔα iÕ¡β½i<br />
1p α i¡β i 1<br />
i i<br />
§d j for some iÈI½µTÖs½×,<br />
<strong>and</strong> we obtain that<br />
ôiÈIÔβ whereÔxÕ<br />
iÕ<br />
xÔx¡yÕ i¡1Õ1 i¡σ Ôp i¡1Õ,<br />
maxØx, 0Ù; the latter equality holds as for<br />
i¡1Õ<br />
iÊI½we have β iσ i .<br />
TTÖs½× TØs½<br />
Observe that<br />
TØs½<br />
minØx, yÙfor all x, y. So, if we find at least<br />
minØβ GôiÈI i , σ iÙÔp iÕ<br />
indices in 1, . . ., sÙ, then we would have<br />
1, . . .,<br />
ôiÈI sÙ1 Ôβ i¡σ minØβ i , σ<br />
ôiÈI<br />
iÙ¨Ôp i¡1Õ1 β iÔp i¡1Õ,<br />
ôiÈI½Ôβ TÖs½×T½1 i¡σ iÕÔp<br />
thus leading to a contradiction with the choice of P. We will find those indices among the<br />
indices of progressions in Q.
42<br />
3. Now denote I¾ØiÈI : σ i0Ù<strong>and</strong> consider some iÈI¾; then p α i<br />
the<br />
other h<strong>and</strong>, there exists an index jÔiÕsuch that p α i<br />
i§d jÔiÕ; this means that d jÔiÕЧn½<strong>and</strong> hence<br />
P jÔiÕcannot appear in P½, so jÔiÕs½. Moreover, we have observed before that in this case<br />
h jÔiÕÈP½1 , hence Q jÔiÕÈQ. This means that q jÔiÕ§n¾, therefore γ iα i for each iÈI¾(recall<br />
here that q ir ißd½1 <strong>and</strong> hence d jÔiÕ§r jÔiÕ§d½1n¾).<br />
Let d½1p τ 1<br />
1 . . .p τ k<br />
k<br />
. Then n¾p γ 1¡τ 1<br />
1 . . .p γ i¡τ i<br />
k<br />
. Now, if iÈI¾, then for every β the condition<br />
i¡τ pÔγ iÕ¡β 1<br />
i<br />
§q j is equivalent to p α i¡β 1<br />
i<br />
§r j .<br />
Note that n¾nßd½1n, hence we can apply the Claim to the covering Q. We perform<br />
this with the set of indices I¾<strong>and</strong> the exponents β¾iminØβ i , σ iÙ0. So, the set under<br />
consideration is<br />
iÙ iÙ iÈI¾µ<br />
T¾j : Q jÈQ, <strong>and</strong> i¡τ iÕ¡minØβ i<br />
T¾T<br />
,σ 1<br />
pÔγ i<br />
§q j<br />
Ø1ÙØs½<br />
for some<br />
j : Q jÈQ, <strong>and</strong> p α i¡minØβ i ,σ 1<br />
i<br />
§r j for some iÈI¾µ,<br />
<strong>and</strong> we obtainT¾1 G. Finally, we claim that 1, . . .,sÙ¨; then we<br />
will obtainTØs½<br />
iÙ iÙ<br />
1, . . .,sÙG, which is exactly what we need.<br />
To prove this, consider any jÈT¾. Observe first that α i¡minØβ i , σ 1α i¡σ iτ i ,<br />
hence from p α i¡minØβ i ,σ 1<br />
1<br />
i<br />
§d j , which means that<br />
jÈT. Next, the exponent of p i in d j is greater than that in n½, which means that P jÊP½. This<br />
may appear only if j1 or js½, as desired. This completes the proof.<br />
i<br />
§r jl.c.m.Ôd½1, d jÕwe have p α i¡minØβ i ,σ iÙ<br />
Comment 2. A grid analogue of the Claim is also valid. It reads as following.<br />
Claim. Assume that the grid N is covered by subgrids L 1 ,L 2 ,... ,L s so that<br />
(ii½) each subgrid contains a point which is not covered by other subgrids;<br />
(iii) for each coordinate axis, there exists a subgrid L i orthogonal to this axis.<br />
Choose some set of indices IØi 1 ,... ,i tÙÖk×, <strong>and</strong> consider the set of indices<br />
TØj : 1js,<strong>and</strong> L j is orthogonal to the ith axis for some iÈIÙ.<br />
Then ôiÈIÔn T1 i¡1Õ.<br />
This Claim may be proved almost in the same way as in Solution 1.<br />
iЧn½. On
Geometry<br />
G1. Let ABC be an acute triangle with D, E, F the feet of the altitudes lying on BC, CA, AB<br />
respectively. One of the intersection points of the line EF <strong>and</strong> the circumcircle is P. The<br />
lines BP <strong>and</strong> DF meet at point Q. Prove that APAQ.<br />
(United Kingdom)<br />
Solution 1. The line EF intersects the circumcircle at two points. Depending on the choice<br />
of P, there are two different cases to consider.<br />
Case 1: The point P lies on the ray EF (see Fig. 1).<br />
LetCABα,ABCβ <strong>and</strong>BCAγ. The quadrilaterals BCEF <strong>and</strong> CAFD are<br />
cyclic due to the right angles at D, E <strong>and</strong> F. So,<br />
BDF180¥¡FDCCAFα,<br />
AFE180¥¡EFBBCEγ,<br />
DFB180¥¡AFDDCAγ.<br />
Since P lies on the arc AB of the circumcircle,PBABCAγ. Hence, we have<br />
PBD BDFPBA ABD BDFγ β α180¥,<br />
<strong>and</strong> the point Q must lie on the extensions of BP <strong>and</strong> DF beyond the points P <strong>and</strong> F,<br />
respectively.<br />
From the cyclic quadrilateral APBC we get<br />
QPA180¥¡APBBCAγDFBQFA.<br />
Hence, the quadrilateral AQPF is cyclic. ThenAQP180¥¡PFAAFEγ.<br />
We obtained thatAQPQPAγ, so the triangle AQP is isosceles, APAQ.<br />
Q<br />
γ<br />
A<br />
A<br />
α<br />
P<br />
γ<br />
γ<br />
F γ<br />
γ<br />
E<br />
F<br />
γ<br />
γ<br />
γ<br />
E<br />
γ<br />
P<br />
B<br />
γ<br />
β<br />
α<br />
D<br />
γ<br />
C<br />
B<br />
Q<br />
D<br />
γ<br />
C<br />
Fig. 1 Fig. 2
45<br />
Case 2: The point P lies on the ray FE (see Fig. 2). In this case the point Q lies inside<br />
the segment FD.<br />
Similarly to the first case, we have<br />
QPABCAγDFB180¥¡AFQ.<br />
Hence, the quadrilateral AFQP is cyclic.<br />
ThenAQPAFPAFEγQPA. The triangle AQP is isosceles again,<br />
AQPQPA <strong>and</strong> thus APAQ.<br />
Comment. Using signed angles, the two possible configurations can be h<strong>and</strong>led simultaneously, without<br />
investigating the possible locations of P <strong>and</strong> Q.<br />
Solution 2. For arbitrary points X, Y on the circumcircle, denote byXY the central angle<br />
of the arc XY .<br />
Let P <strong>and</strong> P½be the two points where the line EF meets the circumcircle; let P lie on<br />
the arc AB <strong>and</strong> let P½lie on the arc CA. Let BP <strong>and</strong> BP½meet the line DF <strong>and</strong> Q <strong>and</strong> Q½,<br />
respectively (see Fig. 3). We will prove that APAP½AQAQ½.<br />
Q<br />
A<br />
P<br />
F<br />
γ<br />
γ<br />
γ<br />
E<br />
P ′<br />
Q ′<br />
γ<br />
B D<br />
C<br />
Fig. 3<br />
Like in the first solution, we haveAFEBFPDFBBCAγ from the<br />
cyclic quadrilaterals BCEF <strong>and</strong> CAFD.<br />
ByPB P½A2AFP½2γ2BCAAP PB, we have<br />
APP½A,PBAABP½<br />
Ô1Õ<br />
<strong>and</strong> APAP½.<br />
Ô2Õ<br />
Due toAPP½A, the lines BP <strong>and</strong> BQ½are symmetrical about line AB.<br />
Similarly, byBFPQ½FB, the lines FP <strong>and</strong> FQ½are symmetrical about AB. It<br />
follows that also the points P <strong>and</strong> P½are symmetrical to Q½<strong>and</strong> Q, respectively. Therefore,<br />
APAQ½<strong>and</strong> AP½AQ.<br />
The relations (1) <strong>and</strong> (2) together prove APAP½AQAQ½.
46<br />
G2. Point P lies inside triangle ABC. Lines AP, BP, CP meet the circumcircle of ABC<br />
again at points K, L, M, respectively. The tangent to the circumcircle at C meets line AB<br />
at S. Prove that SCSP if <strong>and</strong> only if MKML.<br />
(Pol<strong>and</strong>)<br />
Solution 1. We assume that CACB, so point S lies on the ray AB.<br />
From the similar triangles △PKM△PCA <strong>and</strong> △PLM△PCB we get<br />
KMPA PM<br />
CA<br />
<strong>and</strong><br />
PMCB LM . Multiplying these two equalities, we get<br />
PB<br />
LM<br />
KMCB<br />
CA¤PA<br />
PB .<br />
Hence, the relation MKML is equivalent to CB<br />
CAPB<br />
PA .<br />
Denote by E the foot of the bisector of angle B in triangle ABC. Recall that the locus of<br />
points X for which<br />
XBCA XA is the Apollonius circle Ω with the center Q on the line AB,<br />
CB<br />
<strong>and</strong> this circle passes through C <strong>and</strong> E. Hence, we have MKML if <strong>and</strong> only if P lies on Ω,<br />
that is QPQC.<br />
L<br />
C<br />
Ω<br />
K<br />
P<br />
A<br />
E<br />
M<br />
Fig. 1<br />
Now we prove that SQ, thus establishing the problem statement. We haveCES<br />
CAE ACEBCS ECBECS, so SCSE. Hence, the point S lies on AB<br />
as well as on the perpendicular bisector of CE <strong>and</strong> therefore coincides with Q.<br />
Solution 2. As in the previous solution, we assume that S lies on the ray AB.<br />
1. Let P be an arbitrary point inside both the circumcircle ω of the triangle ABC <strong>and</strong> the<br />
angle ASC, the points K, L, M defined as in the problem. We claim that SPSC implies<br />
MKML.<br />
Let E <strong>and</strong> F be the points of intersection of the line SP with ω, point E lying on the<br />
segment SP (see Fig. 2).<br />
B<br />
S
47<br />
F<br />
L<br />
ω<br />
P<br />
K<br />
C<br />
M<br />
E<br />
A<br />
B<br />
Fig. 2<br />
S<br />
We have SP 2SC2SA¤SB, so<br />
SBSA SP<br />
SP , <strong>and</strong> hence △PSA△BSP. Then<br />
BPSSAP. Since 2BPSBE LF <strong>and</strong> 2SAPBE EK we have<br />
LFEK. (1)<br />
On the other h<strong>and</strong>, fromSPCSCP we haveEC MFEC EM, or<br />
MFEM. (2)<br />
From (1) <strong>and</strong> (2) we getǑMFLMF FLME EKǑMEK <strong>and</strong> hence MKML.<br />
The claim is proved.<br />
2. We are left to prove the converse. So, assume that MKML, <strong>and</strong> introduce the<br />
points E <strong>and</strong> F as above. We have SC 2SE¤SF; hence, there exists a point P½lying on the<br />
segment EF such that SP½SC (see Fig. 3).<br />
L ′<br />
L<br />
C<br />
F<br />
ω<br />
P ′<br />
P<br />
K ′<br />
K<br />
E<br />
A<br />
B<br />
S<br />
M ′<br />
M<br />
Fig. 3
M½<br />
48<br />
Assume that PP½. Let the lines AP½, BP½, CP½meet ω again at points K½, L½,<br />
respectively. Now, if P½lies on the segment PF then by the first part of the solution we have<br />
ǑM½FL½ǑM½EK½. On the other h<strong>and</strong>, we haveǑMFLǑM½FL½ǑM½EK½ǑMEK, therefore<br />
ǑMFLǑMEK which contradicts MKML.<br />
Similarly, if point P½lies on the segment EP then we getǑMFLǑMEK which is impossible.<br />
Therefore, the points P <strong>and</strong> P½coincide <strong>and</strong> hence SPSP½SC.<br />
Solution 3. We present a different proof of the converse direction, that is, MKML<br />
SPSC. As in the previous solutions we assume that CACB, <strong>and</strong> line meets ω<br />
at E <strong>and</strong> F.<br />
From MLMK we getǑMEKǑMFL. Now we claim thatMEMF <strong>and</strong>EKFL.<br />
To the contrary, suppose first thatMEMF; thenEKǑMEK¡MEǑMFL¡MF<br />
FL. Now, the inequalityMEMF implies 2SCMEC MEEC MF2SPC<br />
<strong>and</strong> hence SPSC. On the other h<strong>and</strong>, the inequalityEKFL implies 2SPK<br />
EK AFFL AF2ABL, hence<br />
SPA180¥¡SPK180¥¡ABLSBP.<br />
L<br />
C<br />
F<br />
K<br />
P<br />
E<br />
A<br />
A ′<br />
B<br />
S<br />
M<br />
ω<br />
Fig. 4<br />
Consider the point A½on the ray SA for whichSPA½SBP; in our case, this point lies<br />
on the segment SA (see Fig. 4). Then △SBP△SPA½<strong>and</strong> SP 2SB¤SA½SB¤SASC 2 .<br />
Therefore, SPSC which contradicts SPSC.<br />
Similarly, one can prove that the inequalityMEMF is also impossible. So, we get<br />
MEMF <strong>and</strong> therefore 2SCMEC MEEC MF2SPC, which implies<br />
SCSP.
50<br />
G3. Let A 1 A 2 . . .A n be a convex polygon. Point P inside this polygon is chosen so that its<br />
projections P 1 , ..., P n onto lines A 1 A 2 , ..., A n A 1 respectively lie on the sides of the polygon.<br />
Prove that for arbitrary points X 1 , ..., X n on sides A 1 A 2 , ..., A n A 1 respectively,<br />
maxX 1 X 2<br />
P 1 P 2<br />
, . . ., X nX 1<br />
P n P 11.<br />
(Armenia)<br />
Solution 1. Denote P n 1P 1 , X n 1X 1 , A n 1A 1 .<br />
Lemma. Let point Q lies inside A 1 A 2 . . .A n . Then it is contained in at least one of the circumcircles<br />
of triangles X 1 A 2 X 2 , ..., X n A 1 X 1 .<br />
Proof. If Q lies in one<br />
2Õ ¤¤¤<br />
of the triangles X 1 A 2 X 2 , ..., X n A 1 X 1 , the claim is obvious. Otherwise<br />
Q lies inside the polygon X 1 X 2 . . .X n<br />
1Õ 1Õ ¤¤¤<br />
(see Fig. 1). Then we have<br />
X 1 QX ÔX n A 1 X 1 X n QX<br />
X n A 1 X ÔX 1 QX 2<br />
Ð<br />
2πnπ,<br />
hence there exists an index i such thatX i A i 1X i X 1 i QX i 1πn<br />
nπ. Since the<br />
quadrilateral QX i A i 1X i 1 is convex, this means exactly that Q is contained the circumcircle<br />
of △X i A i 1X i 1, as desired.<br />
ÔX 1 A 2 X 2<br />
¤¤¤ ÔX 1 A 1 X 2<br />
X n QX 1ÕÔn¡2Õπ<br />
Now we turn to the solution. Applying lemma, we get that P lies inside the circumcircle of<br />
triangle X i A i 1X i 1 for some i. Consider the circumcircles ω <strong>and</strong> Ω of triangles P i A i 1P i 1 <strong>and</strong><br />
X i A i 1X i 1 respectively (see Fig. 2); let r <strong>and</strong> R be their radii. Then we get 2rA i 1P2R<br />
(since P lies inside Ω), hence<br />
QED.<br />
P i P i 12r sinP i A i 1P i 12R sinX i A i 1X i 1X i X i 1,<br />
X 4<br />
A 4<br />
X 3<br />
Ω<br />
A 3<br />
P<br />
A 5 X i+1<br />
Q<br />
X 2 ω<br />
P i+1<br />
X 5<br />
A 1 X 1 A P i 2<br />
X i A i+1<br />
Fig. 1 Fig. 2
51<br />
Solution 2. As in Solution 1, we assume that all indices of points are considered modulo n.<br />
We will prove a bit stronger inequality, namely<br />
1 X 2<br />
maxX<br />
cosα 1 , . . ., X nX 1<br />
cosα n1,<br />
P 1 P 2 P n P 1<br />
where α i (1in) is the angle between lines X i X i 1 <strong>and</strong> P i P i 1. We denote β iA i P i P i¡1<br />
<strong>and</strong> γ iA i 1P i P i 1 for all 1in.<br />
i<br />
Suppose that for some 1in, point X i lies on the segment A i P i , while point X i 1 lies on<br />
the segment P i 1A i 2. Then the projection of the segment X i X i 1 onto the line P i P i 1 contains<br />
segment P i P i 1, since γ i <strong>and</strong> β i 1 are acute angles (see Fig. 3). Therefore, X i X i 1 cosα<br />
P i P i 1, <strong>and</strong> in this case the statement is proved.<br />
So, the only case left is when point X i lies on segment P i A i 1 for all 1in(the case<br />
when each X i lies on segment A i P i<br />
Y½ Y½<br />
1Y½<br />
is completely analogous).<br />
Now, assume to the contrary that the inequality<br />
X i X i 1 cosα iP i P i 1 (1)<br />
holds for every 1in. Let Y i <strong>and</strong> i 1 be the projections of X i <strong>and</strong> X i 1 onto P i P i 1. Then<br />
inequality (1) means exactly that Y i i 1P i P i 1, or P i Y iP i i 1 (again since γ i <strong>and</strong> β i 1<br />
are acute; see Fig. 4). Hence, we have<br />
X i P i cosγ iX i 1P i<br />
1¨<br />
1 cosβ i 1, 1in.<br />
Multiplying these inequalities, we get<br />
cos γ 1 cos γ 2¤¤¤cosγ ncosβ 1 cosβ 2¤¤¤cosβ n . (2)<br />
On the other h<strong>and</strong>, the sines theorem applied to triangle PP i P i 1 provides<br />
π<br />
PP i 2¡β i i 1<br />
.<br />
PP<br />
π<br />
i 1sin<br />
sin 2¡γ i¨cosβ γ i<br />
Multiplying these equalities we get<br />
which contradicts (2).<br />
1cos β 2<br />
cos γ 1¤cos β 3<br />
cos γ 2¤¤¤cosβ 1<br />
cosγ n<br />
α i<br />
X i+1<br />
P<br />
P i+1<br />
β i+1<br />
γ i<br />
X i P i A i+1<br />
P i+1<br />
γ i<br />
β i+1<br />
α i<br />
A i+1<br />
X i<br />
P<br />
X i−1<br />
X i+1<br />
Fig. 3 Fig. 4<br />
Y i<br />
Y ′<br />
i+1
52<br />
G4. Let I be the incenter of a triangle ABC <strong>and</strong> Γ be its circumcircle. Let the line AI<br />
intersect Γ at a point DA. Let F <strong>and</strong> E be points on side BC <strong>and</strong> arc BDC respectively<br />
such thatBAFCAE1<br />
2BAC. Finally, let G be the midpoint of the segment IF.<br />
Prove that the lines DG <strong>and</strong> EI intersect on Γ.<br />
(Hong Kong)<br />
Solution 1. Let X be the second point of intersection of line EI with Γ, <strong>and</strong> L be the foot<br />
of the bisector of angle BAC. Let G½<strong>and</strong> T be the points of intersection of segment DX with<br />
lines IF <strong>and</strong> AF, respectively. We are to prove that GG½, or IG½G½F. By the Menelaus<br />
theorem applied to triangle AIF <strong>and</strong> line DX, it means that we need the relation<br />
1G½F<br />
IG½TF<br />
AT¤AD<br />
ID , or TF<br />
ATID<br />
AD .<br />
Let the line AF intersect Γ at point KA(see Fig. 1); Furthermore,DCLDCB<br />
sinceBAKCAE we have<br />
BKCE, hence KE ‖ BC. Notice thatIATDAKEADEXDIXT, so<br />
the points I, A, X, T are concyclic. Hence we haveITAIXAEXAEKA, so<br />
IT ‖ KE ‖ BC. Therefore we obtain TF<br />
ATIL<br />
AI .<br />
Since CI is the bisector ofACL, we get<br />
AICL IL<br />
AC .<br />
DABCAD1<br />
2BAC, hence the triangles DCL <strong>and</strong> DAC are similar; therefore we get<br />
CL<br />
. Finally, it is known that the midpoint D of arc BC is equidistant from points I,<br />
ACDC<br />
AD<br />
B, C, hence DC<br />
ADID<br />
AD .<br />
Summarizing all these equalities, we get<br />
as desired.<br />
TF<br />
ATIL<br />
AICL<br />
ACDC<br />
ADID<br />
AD ,<br />
X<br />
A<br />
A<br />
I<br />
T<br />
I<br />
B<br />
C<br />
B<br />
F<br />
G ′<br />
L<br />
C<br />
D<br />
K<br />
E<br />
D<br />
L<br />
Fig. 1 Fig. 2
53<br />
Comment. The equality AI is known <strong>and</strong> can be obtained in many different ways. For<br />
ILAD<br />
DI<br />
instance, one can consider the inversion with center D <strong>and</strong> radius DCDI. This inversion takes<br />
<br />
ǑBAC to the segment BC, so point A goes to L. Hence<br />
DIAI IL<br />
<br />
, which is the desired equality.<br />
AD<br />
Solution 2. As in the previous solution, we introduce the points X, T <strong>and</strong> K <strong>and</strong> note that<br />
it suffice to prove the equality<br />
TF<br />
TF AT<br />
ATDI DI AD AT AF<br />
AD AT AD ADDI AD .<br />
SinceFADEAI <strong>and</strong>TDAXDAXEAIEA, we get that the triangles<br />
ATD <strong>and</strong> AIE are similar, therefore<br />
ADAI AT<br />
AE .<br />
Next, we also use the relation DBDCDI. Let J be the point on the extension<br />
of segment AD over point D such that DJDIDC (see Fig. 2). ThenDJC<br />
JCD1<br />
2Ôπ¡JDCÕ1<br />
2ADC1<br />
2ABCABI. Moreover,BAIJAC, hence<br />
triangles ABI <strong>and</strong> AJC are similar, so AB<br />
AJAI<br />
AC , or AB¤ACAJ¤AIÔDI ADÕ¤AI.<br />
<br />
On the other h<strong>and</strong>, we getABFABCAEC <strong>and</strong>BAFCAE, so triangles<br />
ABF <strong>and</strong> AEC are also similar, which implies AF , or AB¤ACAF¤AE.<br />
ACAB<br />
AE<br />
Summarizing we get<br />
AI AF<br />
AE <br />
AT AF<br />
AD<br />
ÔDI ADÕ¤AIAB¤ACAF¤AE<br />
as desired.<br />
Comment. In fact, point J is an excenter of triangle ABC.<br />
AD DI<br />
AD DI ,
54<br />
G5. Let ABCDE be a convex pentagon such that BC ‖ AE, ABBC AE, <strong>and</strong>ABC<br />
CDE. Let M be the midpoint of CE, <strong>and</strong> let O be the circumcenter of triangle BCD. Given<br />
thatDMO90¥, prove that 2BDACDE.<br />
(Ukraine)<br />
Solution 1. Choose point T on ray AE such that ATAB; then from AE ‖ BC we have<br />
CBTATBABT, so BT is the bisector ofABC. On the other h<strong>and</strong>, we have<br />
ETAT¡AEAB¡AEBC, hence quadrilateral<br />
BDCCDEABC180¥¡<br />
BCTE is a parallelogram, <strong>and</strong> the<br />
midpoint M of its diagonal CE is also the midpoint of the other diagonal BT.<br />
Next, let point K be symmetrical to D with respect to M. Then OM is the perpendicular<br />
bisector of segment DK, <strong>and</strong> hence ODOK, which means that point K lies on the circumcircle<br />
of triangle BCD. Hence we haveBDCBKC. On the other h<strong>and</strong>, the angles<br />
BKC <strong>and</strong> TDE are symmetrical with respect to M, soTDEBKCBDC.<br />
Therefore,BDTBDE EDTBDE<br />
BAT. This means that the points A, B, D, T are concyclic, <strong>and</strong> henceADBATB<br />
1<br />
2ABC1<br />
2CDE, as desired.<br />
B<br />
C<br />
B<br />
C<br />
O 2ϕ α<br />
D<br />
β<br />
α + β<br />
K<br />
M<br />
2ϕ − β − γ<br />
D<br />
γ<br />
A E<br />
T<br />
A<br />
E<br />
Solution 2. LetCBDα,BDCβ,ADEγ, <strong>and</strong>ABCCDE2ϕ. Then<br />
we haveADB2ϕ¡β¡γ,BCD180¥¡α¡β,AED360¥¡BCD¡CDE<br />
180¥¡2ϕ α β, <strong>and</strong> finallyDAE180¥¡ADE¡AED2ϕ¡α¡β¡γ.<br />
2ϕ − α − β − γ<br />
2ϕ − α − β<br />
B<br />
C<br />
B<br />
C<br />
O<br />
M<br />
N<br />
D<br />
M<br />
N<br />
O<br />
D<br />
E<br />
E<br />
Let N be the midpoint of CD; thenDNO90¥DMO, hence points M, N lie on<br />
the circle with diameter OD. Now, if points O <strong>and</strong> M lie on the same side of CD, we have<br />
DMNDON1<br />
2DOCα; in the other case, we haveDMN180¥¡DONα;
55<br />
so, in both casesDMNα(see Figures). Next, since MN is a midline triangle CDE,<br />
we haveMDEDMNα <strong>and</strong>NDM2ϕ¡α.<br />
Now we apply the sine rule to the triangles ABD, ADE (twice), BCD <strong>and</strong> MND obtaining<br />
AB<br />
AE<br />
γ DE<br />
ADsinÔ2ϕ¡β¡γÕ<br />
sinÔ2ϕ¡αÕ,<br />
ADsinÔ2ϕ¡α¡βÕ,<br />
ADsinÔ2ϕ¡α¡β¡γÕ<br />
sinÔ2ϕ¡α¡βÕ,<br />
BC β<br />
CDsin α ,<br />
α<br />
DECDß2<br />
DEß2ND<br />
NMsinÔ2ϕ¡αÕ,<br />
which implies<br />
β¤sinÔ2ϕ¡α¡β¡γÕ<br />
ADBC<br />
CD¤CD<br />
DE¤DE<br />
ADsin<br />
sinÔ2ϕ¡αÕ¤sinÔ2ϕ¡α¡βÕ.<br />
Hence, the condition ABAE BC, or equivalently<br />
ADAE AB BC<br />
, after multiplying<br />
AD<br />
by the common denominator rewrites as<br />
sinÔ2ϕ¡α¡βÕ¤sinÔ2ϕ¡β¡γÕsin γ¤sinÔ2ϕ¡αÕ sin β¤sinÔ2ϕ¡α¡β¡γÕ<br />
γ¡αÕ cosÔγ¡αÕ¡cosÔ4ϕ¡2β¡α¡γÕcosÔ2ϕ¡α¡2β¡γÕ¡cosÔ2ϕ<br />
cosÔγ¡αÕ cosÔ2ϕ γ¡αÕcosÔ2ϕ¡α¡2β¡γÕ cosÔ4ϕ¡2β¡α¡γÕ<br />
cosϕ¤cosÔϕ γ¡αÕcosϕ¤cosÔ3ϕ¡2β¡α¡γÕ<br />
cosÔϕ γ¡αÕ¡cosÔ3ϕ¡2β¡α¡γÕ¨0<br />
cosϕ¤sinÔ2ϕ¡β¡αÕ¤sinÔϕ¡β¡γÕ0.<br />
Since 2ϕ¡β¡α180¥¡AED180¥<strong>and</strong> ϕ1<br />
2ABC90¥, it follows that ϕβ γ,<br />
henceBDA2ϕ¡β¡γϕ1<br />
2CDE, as desired.
56<br />
G6. The vertices X, Y , Z of an equilateral triangle XY Z lie respectively on the sides BC,<br />
CA, AB of an acute-angled triangle ABC. Prove that the incenter of triangle ABC lies inside<br />
triangle XY Z.<br />
G6½.<br />
The vertices X, Y , Z of an equilateral triangle XY Z lie respectively on the sides<br />
BC, CA, AB of a triangle ABC. Prove that if the incenter of triangle ABC lies outside<br />
triangle XY Z, then one of the angles of triangle ABC is greater than 120¥.<br />
(Bulgaria)<br />
Solution 1 for G6. We will prove a stronger fact; namely, we will show that the incenter I of<br />
triangle ABC lies inside the incircle of triangle XY Z (<strong>and</strong> hence surely inside triangle XY Z<br />
itself). We denote by dÔU, V WÕthe distance between point U <strong>and</strong> line V W.<br />
Denote by O the incenter of △XY Z <strong>and</strong> by r, r½<strong>and</strong> R½the inradii of triangles ABC, XY Z<br />
<strong>and</strong> the circumradius of XY Z, respectively. Then we have R½2r½, <strong>and</strong> the desired inequality<br />
is OIr½. We assume that OI; otherwise the claim is trivial.<br />
Let the incircle of △ABC touch its sides BC, AC, AB at points A 1 , B 1 , C 1 respectively.<br />
The lines IA 1 , IB 1 , IC 1 cut the plane into 6 acute angles, each one containing one of the<br />
points A 1 , B 1 , C 1<br />
BCÕ<br />
on its border. We may assume that O lies in an angle defined by lines IA 1 ,<br />
IC 1 <strong>and</strong> containing point C 1 (see Fig. 1). Let A½<strong>and</strong> C½be the projections of O onto lines IA 1<br />
<strong>and</strong> IC 1 , respectively.<br />
Since OXR½, we have dÔO, BCÕR½. Since OA½‖ BC, it follows that dÔA½,<br />
A½I rR½, or A½IR½¡r. On the other h<strong>and</strong>, the incircle of △XY Z lies inside △ABC,<br />
hence dÔO, ABÕr½, <strong>and</strong> analogously we get dÔO, ABÕC½C 1r¡IC½r½, or IC½r¡r½.<br />
B<br />
X<br />
C 1<br />
C ′<br />
Z<br />
O C′<br />
I<br />
A ′<br />
O<br />
I<br />
A<br />
B 1 Y C<br />
A ′<br />
ÔR½¡rÕ Fig. 1 Fig. 2<br />
Finally, the quadrilateral IA½OC½is circumscribed due to the right angles at A½<strong>and</strong><br />
(see Fig. 2). On its circumcircle, we haveǑA½OC½2A½IC½180¥OC½I, hence 180¥<br />
IC½A½O. This means that IC½A½O. Finally, we have OIIA½ A½OIA½ IC½ Ôr¡r½ÕR½¡r½r½, as desired.<br />
Solution 2 for G6. Assume the contrary. Then the incenter I should lie in one of triangles<br />
AY Z, BXZ, CXY — assume that it lies in △AY Z. Let the incircle ω of △ABC touch<br />
sides BC, AC at point A 1 , B 1 respectively. Without loss of generality, assume that point A 1<br />
lies on segment CX. In this case we will show thatC90¥thus leading to a contradiction.<br />
Note that ω intersects each of the segments XY <strong>and</strong> Y Z at two points; let U, U½<strong>and</strong> V ,<br />
V½be the points of intersection of ω with XY <strong>and</strong> Y Z, respectively (UYU½Y , V YV½Y ;<br />
see Figs. 3 <strong>and</strong> 4). Note that 60¥XY Z1<br />
2ÔUV¡U½V½Õ1<br />
2UV , henceUV120¥.<br />
A 1<br />
C ′
On the other h<strong>and</strong>, since I lies in △AY Z, we getǑV UV½180¥, henceǑUA 1 U½ǑUA 180¥¡UV60¥.<br />
Now, two cases are possible due to the order of points Y , B 1 on segment AC.<br />
1<br />
V½<br />
57<br />
A<br />
ω C 1<br />
B 1<br />
C 1<br />
Y<br />
V ′<br />
Y<br />
I<br />
V ′<br />
I<br />
U ′<br />
V<br />
U ′<br />
B<br />
Z<br />
1 V<br />
Z<br />
U<br />
ω<br />
U<br />
C A 1 X<br />
B C A 1 X B<br />
Fig. 3 Fig. 4<br />
Case 1. Let point Y lie on the segment AB 1 (see Fig. 3). Then we haveY XC<br />
1A 2 1 U½¡A U¨1<br />
1 2ǑUA 1 U½30¥; analogously, we getXY C1<br />
2ǑUA 1<br />
X<br />
U½30¥. Therefore,<br />
Y CX180¥¡Y XC¡XY C120¥, as desired.<br />
Case 2. Now let point Y lie on the segment CB 1 (see Fig. 4). Analogously, we obtain<br />
Y XC30¥. Next,IY XZY X60¥, butIY XIY B 1 , since Y B 1 is a tangent<br />
<strong>and</strong> Y X is a secant line to circle ω from point Y . Hence, we get 120¥IY B 1 IY<br />
B 1 Y XY XC CX30¥ Y Y CX, henceY CX120¥¡30¥90¥, as desired.<br />
Comment. In the same way, one can prove a more general<br />
Claim. Let the vertices X, Y , Z of a triangle XY Z lie respectively on the sides BC, CA, AB of a<br />
triangle ABC. Suppose that the incenter of triangle ABC lies outside triangle XY Z, <strong>and</strong> α is the<br />
least angle of △XY Z. Then one of the angles of triangle ABC is greater than 3α¡90¥.<br />
A<br />
Solution for G6½. Assume the contrary. As in Solution 2, we assume that the incenter I of<br />
△ABC lies in △AY Z, <strong>and</strong> the tangency point A 1 of ω <strong>and</strong> BC lies on segment CX. Surely,<br />
Y ZA180¥¡Y ZX120¥, hence points I <strong>and</strong> Y lie on one side of the perpendicular<br />
bisector to XY ; therefore IXIY . Moreover, ω intersects segment XY at two points, <strong>and</strong><br />
therefore the projection M of I onto XY lies on the segment XY . In this case, we will prove<br />
thatC120¥.<br />
Let Y K, Y L be two tangents from point Y to ω (points K <strong>and</strong> A 1 lie on one side of XY ;<br />
if Y lies on ω, we say KLY ); one of the points K <strong>and</strong> L is in fact a tangency point B 1<br />
of ω <strong>and</strong> AC. From symmetry, we haveY IKY IL. On the other h<strong>and</strong>, since IXIY ,<br />
we get XMXY which impliesA 1<br />
MIKÕ<br />
XYKY X.<br />
Next, we haveMIY90¥¡IY X90¥¡ZY X30¥. Since IA 1ÃA 1 X, IMÃXY ,<br />
IKÃY K we getMIA 1A 1 XYKY XMIK. Finally, we get<br />
A 1 IKA 1 IM ÔKIY ILÕ Y<br />
2MIK 2KIY2MIY60¥.<br />
1 ILÔA<br />
Hence,A 1 IB 160¥, <strong>and</strong> thereforeACB180¥¡A 1 IB 1120¥, as desired.
58<br />
X A 1 K(= B 1 )<br />
Y<br />
M<br />
C<br />
A 1 M Y<br />
L(= B 1 )<br />
X<br />
I<br />
B<br />
Z<br />
A<br />
I<br />
Fig. 5 Fig. 6<br />
Comment 1. The estimate claimed in G6½is sharp. Actually, ifBAC120¥, one can consider an<br />
equilateral triangle XY Z with ZA, YÈAC, XÈBC (such triangle exists sinceACB60¥). It<br />
intersects with the angle bisector ofBAC only at point A, hence it does not contain I.<br />
Comment 2. As in the previous solution, there is a generalization for an arbitrary triangle XY Z,<br />
but here we need some additional condition. The statement reads as follows.<br />
Claim. Let the vertices X, Y , Z of a triangle XY Z lie respectively on the sides BC, CA, AB of a<br />
triangle ABC. Suppose that the incenter of triangle ABC lies outside triangle XY Z, α is the least<br />
angle of △XY Z, <strong>and</strong> all sides of triangle XY Z are greater than 2r cot α, where r is the inradius<br />
of △ABC. Then one of the angles of triangle ABC is greater than 2α.<br />
The additional condition is needed to verify that XMY M since it cannot be shown in the<br />
original way. Actually, we haveMY Iα, IMr, hence Y Mr cot α. Now, if we have<br />
XYXM YM2r cot α, then surely XMY M.<br />
On the other h<strong>and</strong>, this additional condition follows easily from the conditions of the original<br />
problem. Actually, if IÈ△AY Z, then the diameter of ω parallel to Y Z is contained in △AY Z <strong>and</strong><br />
is thus shorter than Y Z. Hence Y Z2r2r cot 60¥.
60<br />
G7. Three circular arcs γ 1 , γ 2 , <strong>and</strong> γ 3 connect the points A <strong>and</strong> C. These arcs lie in the same<br />
half-plane defined by line AC in such a way that arc γ 2 lies between the arcs γ 1 <strong>and</strong> γ 3 . Point<br />
B lies on the segment AC. Let h 1 , h 2 , <strong>and</strong> h 3 be three rays starting at B, lying in the same<br />
half-plane, h 2 being between h 1 <strong>and</strong> h 3 . For i, j1, 2, 3, denote by V ij the point of intersection<br />
of h i <strong>and</strong> γ j (see the Figure below).<br />
Denote byǑV ij V kjǑV kl V il the curved quadrilateral, whose sides are the segments V ij V il , V kj V kl<br />
<strong>and</strong> arcs V ij V kj <strong>and</strong> V il V kl . We say that this quadrilateral is circumscribed if there exists a circle<br />
touching these two segments <strong>and</strong> two arcs.<br />
Prove that if the curved quadrilateralsǑV 11 V 21ǑV 22 V 12 ,ǑV 12 V 22ǑV 23 V 13 ,ǑV 21 V 31ǑV 32 V 22 are circumscribed,<br />
then the curved quadrilateralǑV 22 V 32ǑV 33 V 23 is circumscribed, too.<br />
h<br />
V 2 23<br />
h 1<br />
V 22<br />
γ 3<br />
V 13<br />
V 33<br />
V 12<br />
V 11<br />
V 32<br />
h 3<br />
A<br />
γ 2<br />
V 21 V 31<br />
γ 1<br />
B<br />
Fig. 1<br />
C<br />
(Hungary)<br />
Solution. Denote by O i <strong>and</strong> R i the center <strong>and</strong> the radius of γ i , respectively. Denote also by H<br />
the half-plane defined by AC which contains the whole configuration. For every point P in<br />
the half-plane H, denote by dÔPÕthe distance between P <strong>and</strong> line AC. Furthermore, for any<br />
r0, denote by ΩÔP, rÕthe circle with center P <strong>and</strong> radius r.<br />
Lemma 1. For every 1ij3, consider those circles ΩÔP, rÕin the half-plane H which<br />
are tangent to h i <strong>and</strong> h j .<br />
(a) The locus of the centers of these circles is the angle bisector β ij between h i <strong>and</strong> h j .<br />
(b) There is a constant u ij such that ru ij¤dÔPÕfor<br />
Ð<br />
all such circles.<br />
Proof. Part (a) is obvious. To prove part (b), notice that the circles which are tangent to h i<br />
<strong>and</strong> h j are homothetic with the common homothety center B (see Fig. 2). Then part (b) also<br />
becomes trivial.<br />
Lemma 2. For every 1ij3, consider those circles ΩÔP, rÕin the half-plane H which<br />
are externally tangent to γ i <strong>and</strong> internally tangent to γ j .<br />
(a) The locus of the centers of these circles is an ellipse arc ε ij with end-points A <strong>and</strong> C.<br />
(b) There is a constant v ij such that rv ij¤dÔPÕfor all such circles.<br />
Proof. (a) Notice that the circle ΩÔP, rÕis externally tangent to γ i <strong>and</strong> internally tangent to γ j<br />
if <strong>and</strong> only if O i PR i r <strong>and</strong> O jR j¡r. Therefore, for each such circle we have<br />
O i P O j PO i A O j AO i C O j CR i R j .<br />
Such points lie on an ellipse with foci O i <strong>and</strong> O j ; the diameter of this ellipse is R i R j , <strong>and</strong> it<br />
passes through the points A <strong>and</strong> C. Let ε ij be that arc AC of the ellipse which runs inside the<br />
half plane H (see Fig. 3.)<br />
This ellipse arc lies between the arcs γ i <strong>and</strong> γ j . Therefore, if some point P lies on ε ij ,<br />
then O i PR i <strong>and</strong> O j PR j . Now, we choose rO i P¡R iR j¡O j P0; then the
61<br />
Ω(P, r)<br />
γ j<br />
h i<br />
β ij<br />
r<br />
P<br />
R j<br />
ε ij<br />
P<br />
r<br />
P ′<br />
r ′<br />
d(P ′ )<br />
h j<br />
A<br />
d(P)<br />
B<br />
⃗ρ<br />
Fig. 2 Fig. 3<br />
r, i j<br />
circle ΩÔP, rÕtouches γ i externally <strong>and</strong> touches γ j<br />
,P<br />
internally, so P belongs to the locus under<br />
investigation.<br />
(b) Let AP, ⃗ρ AO i , <strong>and</strong> ⃗ρ AO j ; let d ijO i O j , <strong>and</strong> let ⃗v be a unit vector<br />
orthogonal to AC <strong>and</strong> directed toward H. Then we have⃗ρ iR i ,⃗ρ jR j O i<br />
⃗ρ¡⃗ρ iR i O j<br />
jÕ P⃗ρ¡⃗ρ jR j¡r,<br />
jÕ<br />
hence<br />
Ô⃗ρ¡⃗ρ iÕ2¡Ô⃗ρ¡⃗ρ jÕ2ÔR i rÕ2¡ÔR j¡rÕ2 ,<br />
Ô⃗ρ i¡⃗ρ 2 2 2⃗ρ¤Ô⃗ρ j¡⃗ρ iÕÔR i¡R 2 2 2rÔR i R jÕ,<br />
d ij¤dÔPÕd ij<br />
r<br />
⃗v¤⃗ρÔ⃗ρ j¡⃗ρ iÕ¤⃗ρrÔR i R jÕ.<br />
Therefore,<br />
d R i R j¤dÔPÕ,<br />
<strong>and</strong> the value v ij<br />
d<br />
ijǑ Ǒ Ð<br />
ij<br />
does not depend on P.<br />
R i R j<br />
Lemma 3. The curved quadrilateral Q V i,j V i 1,jV i 1,j 1V i,j 1 is circumscribed if <strong>and</strong> only<br />
if u i,i 1v j,j 1.<br />
Proof. First suppose that the curved quadrilateral Q ij is circumscribed <strong>and</strong> ΩÔP, rÕis its inscribed<br />
circle. By Lemma 1 <strong>and</strong> Lemma 2 we have ru i,i 1¤dÔPÕ<strong>and</strong> rv j,j 1¤dÔPÕas<br />
well. Hence, u i,i 1v j,j 1.<br />
To prove the opposite direction, suppose u i,i 1v j,j<br />
Ð<br />
1. Let P be the intersection of the<br />
angle bisector β i,i 1 <strong>and</strong> the ellipse arc ε j,j 1. Choose ru i,i 1¤dÔPÕv j,j 1¤dÔPÕ. Then<br />
the circle ΩÔP, rÕis tangent to the half lines h i <strong>and</strong> h i 1 by Lemma 1, <strong>and</strong> it is tangent to the<br />
arcs γ j <strong>and</strong> γ j 1 by Lemma 2. Hence, the curved quadrilateral Q ij is circumscribed.<br />
By Lemma 3, the statement of the problem can be reformulated to an obvious fact: If the<br />
equalities u 12v 12 , u 12v 23 , <strong>and</strong> u 23v 12 hold, then u 23v 23 holds as well.<br />
⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ⃗ρ<br />
r<br />
⃗ρ j<br />
⃗ρ i<br />
O j<br />
O i<br />
R i<br />
γ i<br />
⃗v<br />
C
62<br />
Comment 1. Lemma 2(b) (together with the easy Lemma 1(b)) is the key tool in this solution.<br />
If one finds this fact, then the solution can be finished in many ways. That is, one can find a circle<br />
touching three of h 2 , h 3 , γ 2 , <strong>and</strong> γ 3 , <strong>and</strong> then prove that it is tangent to the fourth one in either<br />
synthetic or analytical way. Both approaches can be successful.<br />
Here we present some discussion about this key Lemma.<br />
1. In the solution above we chose an analytic proof for Lemma 2(b) because we expect that most<br />
students will use coordinates or vectors to examine the locus of the centers, <strong>and</strong> these approaches are<br />
less case-sensitive.<br />
Here we outline a synthetic proof. We consider only the case when P does not lie in the line O i O j .<br />
The other case can be obtained as a limit case, or computed in a direct way.<br />
Let S be the internal homothety center between the circles of γ i <strong>and</strong> γ j , lying on O i O j ; this point<br />
does not depend on P. Let U <strong>and</strong> V be the points of tangency of circle σΩÔP,rÕwith γ i <strong>and</strong> γ j ,<br />
respectively (then rPUPV ); in other words, points U <strong>and</strong> V are the intersection points of<br />
rays O i P, O j P with arcs γ i , γ j respectively (see Fig. 4).<br />
Due to the theorem on three homothety centers (or just to the Menelaus theorem applied to<br />
triangle O i O j<br />
P<br />
P), the points U, V <strong>and</strong> S are collinear. Let T be the intersection point of line AC <strong>and</strong><br />
the common tangent to σ <strong>and</strong> γ i at U; then T is the radical center of σ, γ i <strong>and</strong> γ j , hence TV is the<br />
common tangent to σ <strong>and</strong> γ j<br />
Ð<br />
.<br />
Let Q be the projection of P onto the line AC. By the right angles, the points U, V <strong>and</strong> Q lie on<br />
the circle with diameter PT. From this fact <strong>and</strong> the equality PUPV we getUQPUV<br />
V UPSUO i . Since O i S ‖ PQ, we haveSO i UQPU. Hence, the triangles SO i U <strong>and</strong> UPQ<br />
r<br />
i S i S<br />
are similar <strong>and</strong> thus<br />
; the last expression is constant since S is a constant<br />
dÔPÕPU<br />
PQO O i UO R i<br />
point.<br />
l<br />
V<br />
P<br />
σ<br />
γ j<br />
d l (A)<br />
P<br />
d l (P)<br />
O j<br />
ε ij<br />
U<br />
γ i<br />
d(P)<br />
O j<br />
S<br />
A<br />
C<br />
T<br />
A<br />
Q<br />
C<br />
O i<br />
O i<br />
Fig. 4 Fig. 5<br />
2. Using some known facts about conics, the same statement can be proved in a very short way.<br />
Denote by l the directrix of ellipse of ε ij related to the focus O j ; since ε ij is symmetrical about O i O j ,<br />
we have l ‖ AC. Recall that for each point PÈε ij , we have PO jǫ¤d lÔPÕ, where d lÔPÕis the<br />
distance from P to l, <strong>and</strong> ǫ is the eccentricity of ε ij<br />
Ð<br />
(see Fig. 5).<br />
Now we have<br />
rR j¡ÔR j¡rÕAO j¡PO jǫ d lÔAÕ¡d lÔPÕ¨ǫ dÔPÕ¡dÔAÕ¨ǫ¤dÔPÕ,<br />
<strong>and</strong> ǫ does not depend on P.
63<br />
Comment 2. One can find a spatial interpretations of the problem <strong>and</strong> the solution.<br />
For every pointÔx,yÕ<strong>and</strong> radius r0, represent the circle Ω Ôx,yÕ,r¨by the pointÔx,y,rÕ<br />
in space. This point is the apex of the cone with base circle Ω Ôx,yÕ,r¨<strong>and</strong> height r. According to<br />
Lemma 1, the circles which are tangent to h i <strong>and</strong> h j correspond to the points of a half line β½ij , starting<br />
at B.<br />
Now we translate Lemma 2. Take some 1ij3, <strong>and</strong> consider those circles which are<br />
internally tangent to γ j . It is easy to see that the locus of the points which represent these circles is<br />
a subset of a cone, containing γ j . Similarly, the circles which are externally tangent to γ i correspond<br />
to the points on the extension of another cone, which has its apex on the opposite side of the base<br />
plane Π. (See Fig. 6; for this illustration, the z-coordinates were multiplied by 2.)<br />
The two cones are symmetric to each other (they have the same aperture, <strong>and</strong> their axes are<br />
parallel). As is well-known, it follows that the common points of the two cones are co-planar. So the<br />
intersection of the two cones is a a conic section — which is an ellipse, according to Lemma 2(a). The<br />
points which represent the circles touching γ i <strong>and</strong> γ j is an ellipse arc ε½ij with end-points A <strong>and</strong> C.<br />
β 12<br />
′<br />
ε ′ ij<br />
γ i<br />
γ j<br />
Π<br />
Σ<br />
ε ′ 23<br />
ε ′ 12<br />
β ′ 23<br />
Fig. 6 Fig. 7<br />
Thus, the curved quadrilateral Q ij is circumscribed if <strong>and</strong> only if β½i,i 1 <strong>and</strong> ε½j,j 1 intersect, i.e. if<br />
they are coplanar. If three of the four curved quadrilaterals are circumscribed, it means that ε½12 , ε½23 ,<br />
<strong>and</strong> β½23 lie in the same plane Σ, <strong>and</strong> the fourth intersection comes to existence, too (see Fig. 7).<br />
β½12<br />
A connection between mathematics <strong>and</strong> real life:<br />
the Palace of Creativity “Shabyt” (“Inspiration”) in Astana
Number Theory<br />
N1. Find the least positive integer n for which there exists a setØs 1 , s 2 , . . .,s nÙconsisting of<br />
n distinct positive integers such that<br />
¢1¡1<br />
1Å¢1¡1<br />
s s 2Å...¢1¡1<br />
s nÅ51<br />
2010 .<br />
N1½. Same as Problem N1, but the constant<br />
51<br />
2010<br />
is replaced by<br />
42<br />
2010 . (Canada)<br />
Answer for Problem N1. n39.<br />
Solution for Problem N1. Suppose that for some n there exist the desired numbers; we<br />
may assume that s 1s 2¤¤¤s n . Surely s 11since<br />
nÅ<br />
otherwise 1¡1<br />
So we have<br />
s 10.<br />
2s 1s 2¡1¤¤¤s n¡Ôn¡1Õ, hence s ii 1 for each i1, . . .,n. Therefore<br />
51<br />
2010¢1¡1<br />
1Å¢1¡1<br />
s s 2Å...¢1¡1<br />
1<br />
s<br />
¢1¡1 2Å¢1¡1<br />
3Å...¢1¡1<br />
1Å1<br />
n 2¤2<br />
3¤¤¤n 1<br />
n n 1 ,<br />
which implies<br />
n 12010<br />
51670<br />
1739,<br />
so n39.<br />
Now we are left to show that n39 fits. Consider the setØ2, 3, . . ., 33, 35, 36, . . ., 40, 67Ù<br />
Ô1Õ<br />
which contains exactly 39 numbers. We have<br />
1<br />
2¤2<br />
3¤¤¤32<br />
33¤34<br />
35¤¤¤39<br />
40¤66<br />
671<br />
33¤34<br />
40¤66<br />
6717<br />
67051<br />
2010 ,<br />
hence for n39 there exists a desired example.<br />
Comment. One can show that the exampleÔ1Õis unique.<br />
Answer for Problem N1½. n48.<br />
Solution for Problem N1½. Suppose that for some n there exist the desired numbers. In<br />
the same way we obtain that s ii 1. Moreover, since the denominator of the fraction<br />
42<br />
20107<br />
335 is divisible by 67, some of s i’s should be divisible by 67, so s ns i67. This<br />
3¤¤¤n¡1<br />
means that<br />
42<br />
20101<br />
2¤2<br />
n¤¢1¡1<br />
67Å66<br />
67n ,
65<br />
which implies<br />
n2010¤66<br />
42¤67330<br />
747,<br />
so n48.<br />
Now we are left to show that n48 fits. Consider the setØ2, 3, . . ., 33, 36, 37, . . ., 50, 67Ù<br />
which contains exactly 48 numbers. We have<br />
1<br />
2¤2<br />
3¤¤¤32<br />
33¤35<br />
36¤¤¤49<br />
50¤66<br />
671<br />
33¤35<br />
50¤66<br />
677<br />
33542<br />
2010 ,<br />
hence for n48 there exists a desired example.<br />
Comment 1. In this version of the problem, the estimate needs one more step, hence it is a bit<br />
harder. On the other h<strong>and</strong>, the example in this version is not unique. Another example is<br />
1<br />
2¤2<br />
3¤¤¤46<br />
47¤66<br />
67¤329<br />
3301<br />
67¤66<br />
330¤329<br />
477<br />
67¤542<br />
2010 .<br />
Comment 2. N1½was the Proposer’s formulation of the problem. We propose N1 according to the<br />
number of current IMO.
66<br />
N2. Find all pairsÔm, nÕof nonnegative integers for which<br />
m 2 2¤3 nm 2 n 1¡1¨. (1)<br />
Answer.Ô6, 3Õ,Ô9, 3Õ,Ô9, 5Õ,Ô54, 5Õ.<br />
(Australia)<br />
Solution. For fixed values of n, the equation (1) is a simple quadratic equation in m. For<br />
n5 the solutions are listed in the following table.<br />
case equation discriminant integer roots<br />
n0 m 2¡m 20 ¡7 none<br />
n1 m 2¡3m 60 ¡15 none<br />
n2 m 2¡7m 180 ¡23 none<br />
n3 m 2¡15m 540 9 m6 <strong>and</strong> m9<br />
n4 m 2¡31m 1620 313 none<br />
n5 m 2¡63m 4860 202545 2 m9 <strong>and</strong> m54<br />
We prove that there is no solution for n6.<br />
Suppose thatÔm, nÕsatisfies (1) <strong>and</strong> n6. Since m§2¤3nm 2 n 1¡1¨¡m<br />
m3p with some 0pn or m2¤3q with some 0qn.<br />
In the first case, let qn¡p; then<br />
In the second case let pn¡q. Then<br />
2 n 1¡1m 2¤3 n<br />
m3 p 2¤3 q .<br />
2 , we have<br />
2 n 2¤3 1¡1m n<br />
m2¤3 q 3 p .<br />
1Õ<br />
Hence, in both cases we need to find the nonnegative integer solutions of<br />
3 p 2¤3 q2 n 1¡1, p qn. (2)<br />
Next, we prove<br />
1Õ<br />
1Õ<br />
bounds for p, q. From (2) we get<br />
3 p2 n 18 n 1<br />
39 n 1<br />
33 1Õ<br />
2Ôn 3<br />
<strong>and</strong><br />
2¤3 q2 n 12¤8 n 32¤9 n 32¤3 2n 32¤3 2Ôn 3 ,<br />
so p, q2Ôn . Combining these inequalities with p qn, we obtain<br />
3<br />
n¡2<br />
q2Ôn p, . (3)<br />
3 3<br />
Now let hminÔp, qÕ. By (3) we have hn¡2<br />
3 ; in particular, we have h1. On the<br />
left-h<strong>and</strong> side of (2), both terms are divisible by 3 h , therefore 9§3 h§2 n 1¡1. It is easy check<br />
that ord 9Ô2Õ6, so 9§2 n 1¡1 if <strong>and</strong> only if 6§n 1. Therefore, n 16r for some positive<br />
integer r, <strong>and</strong> we can write<br />
2 n 1¡14 3r¡1Ô4 2r 4 r 1ÕÔ2<br />
r¡1ÕÔ2 r 1Õ. (4)
Notice that the factor 4 2r 4 r 1Ô4 divisible by 9. The other two factors in (4), 2 r¡1 <strong>and</strong> 2 r 1 are coprime: both are odd <strong>and</strong><br />
67<br />
r¡1Õ2 3¤4r is divisible by 3, but it is never<br />
their difference is 2. Since the whole product is divisible by 3 h , we have either 3 h¡1§2 r¡1 or<br />
3 h¡1§2 r 1. In any case, we have 3 h¡12 r 1. Then<br />
3 h¡12 r 13 r3 n 1<br />
6 ,<br />
n¡2<br />
1<br />
,<br />
3¡1h¡1n 6<br />
n11.<br />
But this is impossible since we assumed n6, <strong>and</strong> we proved 6§n 1.
¤¤¤<br />
68<br />
N3. Find the smallest number n such that there exist polynomials f 1 , f 2 , ..., f n with rational<br />
coefficients satisfying<br />
x 2 7f 1ÔxÕ2 f 2ÔxÕ2 f nÔxÕ2 .<br />
(Pol<strong>and</strong>)<br />
Answer. The smallest n is 5.<br />
Solution 1. The equality x 2 7x2 2 2 1 2 1 2 1 2 shows that n5. It remains to<br />
show that x 2 7 is not a sum of four (or less) squares of polynomials with rational coefficients.<br />
Suppose by way of contradiction that x 2 7f 1ÔxÕ2 f 2ÔxÕ2 f 3ÔxÕ2 f 4ÔxÕ2 , where the<br />
coefficients of polynomials f 1 , f 2 , f 3 <strong>and</strong> f 4 are rational (some of these polynomials may be<br />
zero).<br />
Clearly, the degrees of f 1 , f 2 , f 3 <strong>and</strong> f 4 are at most 1. Thus f iÔxÕa i x b i for i1, 2, 3, 4<br />
<strong>and</strong> some rationals a 1 , b 1 , a 2 , b 2 , a 3 , b 3 , a 4 , b 4 . It follows that x 2 74<br />
i1Ôa i x b iÕ2 <strong>and</strong><br />
hence<br />
4ôi1<br />
a 2 i1,<br />
4ôi1<br />
a i b i0,<br />
Let p ia i b i <strong>and</strong> q ia i¡b i for i1, 2, 3, 4. Then<br />
4ôi1<br />
b 2 i7. (1)<br />
4ôi1<br />
p 2 i4ôi1 a 2 i 2<br />
4ôi1<br />
a i b i<br />
4ôi1b 2 i8,<br />
4ôi1<br />
q 2 i4ôi1 a 2 i¡2<br />
4ôi1<br />
a i b i<br />
4ôi1b 2 i8<br />
<strong>and</strong><br />
4ôi1<br />
p i q i4ôi1<br />
a 2 i¡4ôi1 b 2 i¡6,<br />
which means that there exist a solution in integers x 1 , y 1 , x 2 , y 2 , x 3 , y 3 , x 4 , y 4 <strong>and</strong> m0of<br />
the system of equations<br />
4ôi1<br />
4ôi1<br />
4ôi1<br />
(i)<br />
x 2 i8m 2 ,<br />
(ii)<br />
y 2 i8m 2 ,<br />
(iii)<br />
x i y i¡6m 2 .<br />
We will show that such a solution does not exist.<br />
Assume the contrary <strong>and</strong> consider a solution with minimal m. Note that if an integer x is<br />
odd then x 21Ômod 8Õ. Otherwise (i.e., if x is even) we have x 20Ômod 8Õor x 24<br />
Ômod 8Õ. Hence, by (i), we get that x 1 , x 2 , x 3 <strong>and</strong> x 4 are even. Similarly, by (ii), we get that<br />
y 1 , y 2 , y 3 <strong>and</strong> y 4 are even. Thus the LHS of (iii) is divisible by 4 <strong>and</strong> m is also even. It follows<br />
thatÔx 1<br />
2<br />
, y 1<br />
, x 2<br />
2 2<br />
, y 2<br />
, x 3<br />
2 2<br />
, y 3<br />
, x 4<br />
2 2<br />
, y 4<br />
, m 2 2Õis a solution of the system of equations (i), (ii) <strong>and</strong> (iii),<br />
which contradicts the minimality of m.<br />
Solution 2. We prove that n4 is impossible. Define the numbers a i , b i for i1, 2, 3, 4 as<br />
in the previous solution.<br />
By Euler’s identity we have<br />
Ôa 2 1 a 2 2 a 2 3 a 2 4ÕÔb 2 1 b 2 2 b 2 3 b 2 4ÕÔa<br />
1<br />
b 1 a 2 b 2 a 3 b 3 a 4 b 4Õ2 Ôa 1 b 3¡a 3 b 1 a 4 b 2¡a 2 b 4Õ2<br />
Ôa 1 b 2¡a 2 b 1 a 3 b 4¡a 4 b 3Õ2<br />
Ôa 1 b 4¡a 4 b 1 a 2 b 3¡a 3 b 2Õ2 .
69<br />
So, using the relations (1) from the Solution 1 we get that<br />
where<br />
7¡m 1<br />
m©2 ¡m 2<br />
m©2 ¡m 3<br />
m©2<br />
, (2)<br />
m 1<br />
1 b 2¡a 2 b 1 a 3 b 4¡a 4 b 3 ,<br />
ma m 2<br />
1 b 3¡a 3 b 1 a 4 b 2¡a 2 b 4 ,<br />
ma m 3<br />
1 b 4¡a 4 b 1 a 2 b 3¡a 3 b 2<br />
ma <strong>and</strong> m 1 , m 2 , m 3ÈZ,<br />
2¨<br />
mÈN.<br />
Let m be a minimum positive integer number for which (2) holds. Then<br />
8m 2m 2 1 m 2 2 m 2 3 m 2 .<br />
m<br />
As in the previous solution, we get that m 1 , m 2 , m 3 , m are all even numbers. Then 1<br />
, m 2<br />
, m 3<br />
, m<br />
2 2 2<br />
is also a solution of (2) which contradicts the minimality of m. So, we have n5. The example<br />
with n5 is already shown in Solution 1.
70<br />
N4. Let a, b be integers, <strong>and</strong> let PÔxÕax 3 bx. For any positive integer n we say that the<br />
pairÔa, bÕis n-good if n§PÔmÕ¡PÔkÕimplies n§m¡k for all integers m, k. We say that<br />
Ôa, bÕis very good ifÔa, bÕis n-good for infinitely many positive integers n.<br />
(a) Find a pairÔa, bÕwhich is 51-good, but not very good.<br />
(b) Show that all 2010-good pairs are very good.<br />
(Turkey)<br />
Solution. (a) We show that the pairÔ1,¡51 2Õis good but not very good. Let PÔxÕx3¡512 x.<br />
Since PÔ51ÕPÔ0Õ, the pairÔ1,¡51 2Õis not n-good for any positive integer that does not<br />
divide 51. Therefore,Ô1,¡51 2Õis not very good.<br />
On the other h<strong>and</strong>, if PÔmÕPÔkÕÔmod 51Õ, then m 3k3Ômod 51Õ. By Fermat’s<br />
theorem, from this we obtain<br />
mm 3k 3kÔmod 3Õ<strong>and</strong> mm 33k 33kÔmod 17Õ.<br />
Hence we have mkÔmod 51Õ. ThereforeÔ1,¡51 2Õis 51-good.<br />
PÔm½ÕPÔmÕPÔkÕPÔk½Õ<br />
(b) We will show that if a pairÔa, bÕis 2010-good thenÔa, bÕis 67 i -good for all positive<br />
integer i.<br />
Ð<br />
Claim 1. IfÔa, bÕis 2010-good thenÔa, bÕis 67-good.<br />
Proof. Assume that PÔmÕPÔkÕÔmod 67Õ. Since 67 <strong>and</strong> 30 are coprime, there exist integers<br />
m½<strong>and</strong> k½such that k½kÔmod 67Õ, k½0Ômod 30Õ, <strong>and</strong> m½mÔmod 67Õ, m½0<br />
Ômod 30Õ. Then we have PÔm½ÕPÔ0ÕPÔk½ÕÔmod 30Õ<strong>and</strong><br />
Ômod 67Õ, hence PÔm½ÕPÔk½ÕÔmod 2010Õ. This implies m½k½Ômod 2010ÕasÔa, bÕis<br />
2010-good. It follows that mm½k½kÔmod 67Õ. Therefore,Ôa, bÕis 67-good.<br />
Claim 2. IfÔa, bÕis 67-good then 3Õ 67§a.<br />
bÔm¡kÕÔm¡kÕ 2Õ b¨<br />
Proof. Suppose that 67Чa. Consider the setsØat 2Ômod 67Õ:0t33Ù<strong>and</strong>Ø¡3as 2¡b<br />
mod 67 : 0s33Ù. Since a0Ômod 67Õ, each of these sets has 34 elements. Hence they<br />
have at least one element in common. If at 2¡3as2¡bÔmod 67Õthen for mt¨s, k©2s<br />
we have<br />
PÔmÕ¡PÔkÕaÔm 3¡k aÔm 2 mk k<br />
Ôt¨3sÕÔat 2 3as 2 bÕ0Ômod 67Õ.<br />
SinceÔa, bÕis 67-good, we must have mkÔmod 67Õin both cases, that is, t3sÔmod 67Õ<br />
<strong>and</strong> t¡3sÔmod i§PÔmÕ¡PÔkÕÔm¡kÕ 67Õ. This means ts0Ômod 67Õ<strong>and</strong> b¡3as 2¡at20Ômod 67Õ.<br />
But then 67§PÔ7Õ¡PÔ2Õ67¤5a 2Õ 2Õ 5b <strong>and</strong> 67Ч7¡2, contradicting thatÔa, bÕis 67-good.Ð<br />
bÕ Ð<br />
Claim 3. IfÔa, bÕis 2010-good thenÔa, bÕis 67 i -good all i1.<br />
Proof. By Claim 2, we have 67§a. If 67§b, then PÔxÕPÔ0ÕÔmod 67Õfor all x, contradicting<br />
thatÔa, bÕis 67-good. Hence, 67Чb.<br />
Suppose that 67 aÔm2 mk k b¨. Since 67§a <strong>and</strong> 67Чb,<br />
the second factor aÔm2 mk k b is coprime to 67 <strong>and</strong> hence 67 i§m¡k. Therefore,Ôa,<br />
is 67 i -good.<br />
Comment 1. In the proof of Claim 2, the following reasoning can also be used. Since 3 is not<br />
a quadratic residue modulo 67, either au 2¡bÔmod 67Õor 3av 2¡bÔmod 67Õhas a solution.<br />
The settingsÔm,kÕÔu,0Õin the first case <strong>and</strong>Ôm,kÕÔv,¡2vÕin the second case lead to b0<br />
Ômod 67Õ.<br />
Comment 2. The pairÔ67,30Õis n-good if <strong>and</strong> only if nd¤67 i , where d§30 <strong>and</strong> i0. It shows<br />
that in part (b), one should deal with the large powers of 67 to reach the solution. The key property<br />
of number 67 is that it has the form 3k 1, so there exists a nontrivial cubic root of unity modulo 67.
fÔmÕ n¨<br />
71<br />
N5. Let N be the set of all positive integers. Find all functions f : NNsuch that the<br />
number m fÔnÕ¨is a square for all m,<br />
fÔmÕ n¨<br />
nÈN.<br />
fÔnÕ<br />
(U.S.A.)<br />
Answer. All functions of the form fÔnÕn c, where cÈNØ0Ù.<br />
Solution. First, it is clear that all functions of<br />
fÔlÕfÔkÕ<br />
the form fÔnÕn c with a constant nonnegative<br />
integer c satisfy the problem conditions since m¨Ôn m cÕ2 is a<br />
square.<br />
We are left to prove that there are no other functions. We start with the following<br />
Lemma. Suppose that p§fÔkÕ¡fÔlÕfor some prime p <strong>and</strong> positive integers k, l. Then p§k¡l.<br />
p§<br />
Proof.<br />
fÔnÕ fÔkÕ<br />
Suppose<br />
k¨¡ n¨<br />
first<br />
fÔnÕ fÔnÕ fÔlÕ<br />
that p 2§fÔkÕ¡fÔlÕ, fÔnÕ n¨ fÔnÕ fÔnÕ<br />
so p 2 a for some integer a. Take some<br />
positive integer DmaxØfÔkÕ, fÔlÕÙwhich is not divisible by p <strong>and</strong> set npD¡fÔkÕ. Then<br />
the positive numbers n fÔkÕpD <strong>and</strong> n fÔlÕpD fÔlÕ¡fÔkÕ¨pÔD paÕare<br />
both divisible by p but not by p 2 . Now, applying the problem conditions, we get that both the<br />
numbers k¨<strong>and</strong> l¨are squares divisible by p (<strong>and</strong> thus<br />
by p 2 ); this<br />
p§<br />
means<br />
fÔnÕ<br />
that<br />
k¨¡<br />
the<br />
fÔnÕ fÔnÕ fÔkÕ fÔnÕ fÔlÕ n<br />
multipliers k <strong>and</strong> l are also divisible by p, therefore<br />
Ð<br />
l¨k¡l as well.<br />
On the other h<strong>and</strong>, if fÔkÕ¡fÔlÕis divisible by p but not by p 2 , then choose the same<br />
number D <strong>and</strong> set np3 D¡fÔkÕ. Then the positive numbers np3 D <strong>and</strong><br />
p 3 D fÔlÕ¡fÔkÕ¨are respectively divisible by p 3 (but not by p 4 ) <strong>and</strong> by p (but not by p 2 ).<br />
Hence in analogous way we obtain that the numbers k <strong>and</strong> l are divisible by p,<br />
therefore l¨k¡l.<br />
We turn to the problem. First, suppose that fÔkÕfÔlÕfor some k, lÈN. Then by Lemma<br />
we<br />
fÔnÕ¨qfÔ1Õ fÔnÕfÔ1Õ<br />
have that k¡l is divisible by every prime number, so k¡l0, or kl. Therefore, the<br />
function f is injective.<br />
Next, consider the<br />
fÔnÕfÔ1Õ fÔnÕfÔn¡2ÕfÔ1Õ 1Õ<br />
numbers fÔkÕ<strong>and</strong> fÔk 1Õ. Since the numberÔk 1Õ¡k1has no<br />
prime divisors, by Lemma the same holds for fÔk 1Õ¡fÔkÕ; fÔnÕÔfÔ1Õ¡1Õ fÔnÕfÔ1Õ nfÔ1Õ<br />
thusfÔk 1Õ¡fÔkÕ1.<br />
Now, let fÔ2Õ¡fÔ1Õq,q1. Then we prove by induction that qÔn¡1Õ.<br />
The base for n1, 2 holds by the definition of q. For the step, if n1we have qÔn¡1Õ¨q. Since qÔn¡2Õ, we get qn,<br />
as desired.<br />
Finally, we have qÔn¡1Õ. Then q cannot be¡1 since otherwise for 1<br />
we have fÔnÕ0 which is impossible. Hence q1 <strong>and</strong> n for each nÈN,<br />
<strong>and</strong> fÔ1Õ¡10, as desired.
72<br />
N6. The rows <strong>and</strong> columns of a 2 n¢2 n table are numbered from 0 to 2 n¡1. The cells of the<br />
table have been colored with the following property being satisfied: for each 0i, j2 n¡1,<br />
the jth cell in the ith row <strong>and</strong> theÔi jÕth cell in the jth row have the same color. (The<br />
indices of the cells in a row are considered modulo 2 n .)<br />
Prove that the maximal possible number of colors is 2 n .<br />
jÕ<br />
(Iran)<br />
Solution. Throughout the solution we denote the cells of the table by coordinate pairs;Ôi,<br />
refers to the jth cell in the ith row.<br />
Consider the directed graph, whose vertices are the cells of the board, <strong>and</strong> the edges are<br />
the arrowsÔi, jÕÔj, i jÕfor all 0i, j2n¡1. From each vertexÔi, jÕ, exactly one edge<br />
passes (toÔj, i j mod 2 nÕ); conversely, to each cellÔj, kÕexactly one edge is directed (from<br />
the cellÔk¡j mod 2 n , jÕÕ. Hence, the graph splits into cycles.<br />
Now, in any coloring considered, the vertices of each cycle should have the same color by<br />
the problem condition. On the other h<strong>and</strong>, if each cycle has its own color, the obtained coloring<br />
obviously satisfies the problem conditions. Thus, the maximal possible number of colors is the<br />
same as the number of cycles, <strong>and</strong> we have to prove that this number is 2 n .<br />
Next, consider any cycleÔi 1 , j 1Õ,Ôi 2 , j 2Õ, . . .; we will describe it in other terms. Define a<br />
sequenceÔa 0 , a 1 , . . .Õby the relations a 0i 1 , a 1j 1 , a n 1a n a n¡1 2<br />
for all n1(we<br />
say that such a sequence is a Fibonacci-type sequence). Then an obvious induction shows<br />
that i ka k¡1Ômod 2 nÕ, j ka kÔmod 2 nÕ. Hence we need to investigate the behavior of<br />
Fibonacci-type sequences modulo 2 n .<br />
Denote by F 0 , F 1 , . . . the Fibonacci numbers defined by F 00, F 11, <strong>and</strong> F n<br />
F n 1 F n for n0. We also set F¡11 according to the recurrence relation.<br />
νÔmÕ k<br />
For every positive integer m, denote by νÔmÕthe exponent of 2 in the prime factorization<br />
of m, i.e. for which 2 νÔmÕ§m but 2 1Чm.<br />
Lemma 1. For every Fibonacci-type sequence a 0 , a 1 , a 2 , . . ., <strong>and</strong> every k0, we have a<br />
F k¡1a 0 F k a 1 .<br />
Proof. Apply induction on k. The base cases<br />
1Õ Ð<br />
k0, 1 are trivial. For the step, from the<br />
induction hypothesis we get<br />
a k 1a k a k¡1ÔF k¡1a 0 F k a ÔF k¡2a 0 F k¡1a 1ÕF k a 0 F k 1a 1 .<br />
Lemma 2. For every m3,<br />
(a) we have νÔF m¡2Õm;<br />
3¤2 (b) d3¤2m¡2 is the least positive index for which 2 m§F d ;<br />
(c) F 3¤2m¡2 11 2 m¡1Ômod 2 mÕ.<br />
Proof. Apply induction on m. In the base case m3we have νÔF m¡2ÕF 3¤2 68,<br />
so<br />
νÔF m¡2ÕνÔ8Õ3, 3¤2 the preceding Fibonacci-numbers are not divisible by 8, <strong>and</strong> indeed<br />
F 3¤2m¡2 1F 7131 4Ômod 8Õ.<br />
Now suppose that m3 <strong>and</strong> let k3¤2m¡3 . By applying Lemma 1 to the Fibonacci-type<br />
2Ôm¡1ÕÔm¡1Õ<br />
sequence F k , F k 1, . . . we get<br />
F 2kF k¡1F k F k F k 1ÔF k 1¡F kÕF k F k 1F k2F k 1F k¡F kÕ<br />
k 2 ,<br />
F 2k 1F k¤F k F k 1¤F k 1F k 2 Fk 2 1.<br />
By the induction hypothesis, νÔF kÕm¡1, <strong>and</strong> F k 1 is odd. Therefore we get νÔF 2<br />
1νÔ2F k F k 1Õ, which implies νÔF 2kÕm, establishing statement (a).
73<br />
Moreover, since F k 11 2 m¡2 a2 m¡1 for some integer a, we get<br />
F 2k 1F k 2 Fk 2 10 Ô1 2 m¡2 a2 m¡1Õ21 2 m¡1Ômod 2 mÕ,<br />
as desired in statement (c).<br />
We are left to prove that 2 mЧF l for l2k. Assume the contrary. Since 2 m¡1§F l<br />
Ð<br />
, from<br />
the induction hypothesis it follows that lk. But then we have F lF k¡1F l¡k F k F l¡k 1,<br />
where the second summ<strong>and</strong> is divisible by 2 m¡1 but the first one is not (since F k¡1 is odd <strong>and</strong><br />
l¡kk). Hence the sum is not divisible even by 2 m¡1 . A contradiction.<br />
Now, for every pair of integersÔa, bÕÔ0, 0Õ, let µÔa, bÕminØνÔaÕ, νÔbÕÙ. By an obvious induction,<br />
for every Fibonacci-type sequence AÔa 0 , a 1 , . . .Õwe have µÔa 0 , a 1ÕµÔa 1 , a 2Õ. . .;<br />
denote this common value by µÔAÕ. Also denote by p nÔAÕthe period of this sequence modulo<br />
2 n , that is, the least p0 such that a k pa kÔmod 2 nÕfor all k0.<br />
Lemma 3. Let AÔa 0 , a 1 , . . .Õbe a Fibonacci-type sequence such that µÔAÕkn. Then<br />
p nÔAÕ3¤2n¡1¡k .<br />
Proof. First, we note that the sequenceÔa 0 , a 1 , . . .Õhas period p modulo 2 n if <strong>and</strong> only if the<br />
sequenceÔa 0ß2k , a 1ß2k , . . .Õhas period p modulo 2 n¡k . Hence, passing to this sequence we can<br />
.Õ<br />
assume that k0.<br />
We prove the statement by induction on n. It is easy to see that for n1, 2 the claim<br />
is true; actually, each Fibonacci-type sequence A with µÔAÕ0behaves as 0, 1, 1, 0, 1, 1, . . .<br />
modulo 2, <strong>and</strong> as 0, 1, 1, 2, 3, 1, 0, 1, 1, 2, 3, 1, ... modulo 4 (all pairs of residues from which at<br />
least one is odd appear as a pair of consecutive terms in this sequence).<br />
Now suppose that n3<strong>and</strong> consider an arbitrary Fibonacci-type sequence AÔa 0 , a 1 , . .<br />
with µÔAÕ0. Obviously we should have p n¡1ÔAÕ§p nÔAÕ, or, using the induction hypothesis,<br />
0<br />
s3¤2n¡2§p nÔAÕ. Next, we may suppose that a 0 is even; hence a 1 is odd, <strong>and</strong> a 02b 0 ,<br />
a 12b 1 1 for some integers b 0 , b 1 .<br />
Consider the Fibonacci-type sequence BÔb 0 , b 1 , . . .Õstarting withÔb 0 , b 1Õ. Since a<br />
2b 0 F 0 , a 12b 1 F 1 , by an easy induction we get a k2b k F k for all k0. By<br />
the induction hypothesis, we have p n¡1ÔBÕ§s, hence the sequenceÔ2b 0 , 2b 1 , . . .Õis s-periodic<br />
modulo 2 n . On the other h<strong>and</strong>, by Lemma 2 we have F s 11 2 n¡1Ômod 2 nÕ, F 2s0<br />
Ômod 2 nÕ, F 2s 11Ômod 2 nÕ, hence<br />
a s 12b s 1 F s 12b 1 1 2 n¡12b 1 1a 1Ômod 2 nÕ,<br />
a 2s2b 2s F 2s2b 0 0a 0Ômod 2 nÕ,<br />
a 2s 12b 2s 1 F 2s 12b 1 1a 1Ômod<br />
Ð<br />
2 nÕ.<br />
The first line means that A is not s-periodic, while the other two provide that a 2sa 0 ,<br />
a 2s 1a 1 <strong>and</strong> hence a 2s ta t for all t0. Hence s§p nÔAÕ§2s <strong>and</strong> p nÔAÕs, which means<br />
that p nÔAÕ2s, as desired.<br />
0Õ<br />
Finally, Lemma 3 provides a straightforward method of counting the number of cycles.<br />
Actually, take any number 0kn¡1 <strong>and</strong> consider all the cellsÔi, jÕwith µÔi, jÕk. The<br />
total number of such cells is 22Ôn¡kÕ¡22Ôn¡k¡1Õ3¤22n¡2k¡2 . On the other h<strong>and</strong>, they are split<br />
into cycles, <strong>and</strong> by Lemma 3 the length of each cycle is<br />
¤¤¤<br />
3¤2n¡1¡k . Hence the number of cycles<br />
2n¡2k¡2<br />
3¤2 consisting of these cells is exactly<br />
3¤2n¡1¡k2 n¡k¡1 . Finally, there is only one cellÔ0,<br />
which is not mentioned in the previous computation, <strong>and</strong> it forms a separate cycle. So the total<br />
number of cycles is<br />
n¡1<br />
1 2<br />
ôk0<br />
n¡1¡k1 Ô1 2 4 2 n¡1Õ2 n .
74<br />
Comment. We outline a different proof for the essential part of Lemma 3. That is, we assume that<br />
k0<strong>and</strong> show that in this case the period ofÔa iÕmodulo 2 n coincides with the period of the Fibonacci<br />
numbers modulo 2 n ; then the proof can be finished by the arguments from Lemma 2..<br />
Note that p is a (not necessarily minimal) period of the<br />
0Õ<br />
sequenceÔa iÕmodulo 2 n if <strong>and</strong> only if we<br />
have a 0a pÔmod 2 nÕ, a 1a p 1Ômod 2 nÕ, that is,<br />
a 0a pF p¡1a 0 F p a 1F pÔa 1¡a F p 1a 0Ômod 2 nÕ,<br />
a 1a p 1F p a 0 F p 1a 1Ômod 2 nÕ.<br />
(1)<br />
Now, If p is a period ofÔF iÕthen we have F pF 00Ômod 2 nÕ<strong>and</strong> F p 1F 11Ômod 2 nÕ, which<br />
by (1) implies that p is a period ofÔa iÕas well.<br />
0Õ<br />
Conversely, suppose that p is a period ofÔa iÕ. Combining the relations of (1) we<br />
0Õ 1Õ<br />
get<br />
0¨<br />
0a 1¤a 0¡a 0¤a 1a 1 F pÔa 1¡a F p 1a 0¨¡a 0ÔF p a 0 F p 1a<br />
F pÔa 2 1¡a 1 a 0¡a 2 0ÕÔmod 2 nÕ,<br />
a 2 1¡a 1 a 0¡a 2 0Ôa 1¡a 0Õa 1¡a 0¤a 0Ôa 1¡a 0ÕÔF p a 0 F p 1a 1Õ¡a 0 F pÔa 1¡a F p 1a<br />
F p 1Ôa 2 1¡a 1 a 0¡a 2 0ÕÔmod 2 nÕ.<br />
Since at least one of the numbers a 0 , a 1 is odd, the number a 2 1¡a 1 a 0¡a2 0 is odd as well. Therefore the<br />
previous relations are equivalent with F p0Ômod 2 nÕ<strong>and</strong> F p 11Ômod 2 nÕ, which means exactly<br />
that p is a period ofÔF 0 ,F 1 ,...Õmodulo 2 n .<br />
So, the sets of periods ofÔa iÕ<strong>and</strong>ÔF iÕcoincide, <strong>and</strong> hence the minimal periods coincide as well.