X International Zhautykov Olympiad in Sciences Kazakhstan, January 12-18, 2014 Presentation of the Mathematics Section, by Dan Schwarz First

Day (January 14, 2014) Problems

1

Problem 1. Points M, N, K lie on the sides BC, CA, AB respectively, of a triangle ABC , and are dierent from its vertices. The triangle M N K is called beautiful if the triangles M N K and ABC are similar (with the vertices respectively in this order). Show that if in the triangle ABC there are two beautiful triangles with a common vertex, then ABC is right-angled. Problem 2. Do there exist? functions f : R R such that a) f is a surjective function; and b) f (f (x)) = (x 1)f (x) + 2 for all x real.

Problem 3. There are given 100 distinct positive integers. We call a pair of integers among them good if the ratio of its elements is either 2 or 3. What is the maximum number g of good pairs that these 100 numbers can form? (A same number can be used in several pairs.)

Second Day (January 15, 2014) Problems

Problem 4. Does it exist? a polynomial P (x) with integer coecients, such that P (1 + 3) = 2 + 3 and P (3 + 5) = 3 + 5. Problem 5. Let U = {1, 2, 3, . . . , 2014}. For all a, b, c N let f (a, b, c) be the number of ordered sextuplets (X1 , X2 , X3 , Y1 , Y2 , Y3 ) of subsets of U , satisfying the following conditions (i) Y1 X1 U and |X1 | = a; (ii) Y2 X2 U \ Y1 and |X2 | = b; (iii) Y3 X3 U \ (Y1 Y2 ) and |X3 | = c. Prove f ( (a), (b), (c)) does not change, for permutations of a, b, c. Problem 6. Let ABCD be a convex quadrilateral, partitioned by four lines as in the picture below, such that the meeting points of these lines all lie on the diagonals of ABCD. Prove that if the corner quadrilaterals 1, 2, 3 and the center quadrilateral 4 are all tangential, then the corner quadrilateral 5 is also tangential.
Prezentarea problemelor din cadrul sect iunii matematice a concursului Zhautykov din acest an este f acut a si popularizat a cu scopul de a vedea nc a un alt concurs internat ional, pentru a putea c ant ari dicultatea probelor, n comparat ie cu propriile noastre concursuri, de select ie sau de palmares. Deoarece concursul este internat ional, am preferat a face prezentarea n limba englez a, dar comentariile vor f acute n limba rom an a, ca si anunt area rezultatelor, la ncheierea acestui material. Enunt urile de pe prima pagin a sunt urmate de solut ii detaliate si comentarii. Rezultate complete la http://izho.kz/.
1
1

Zhautykov 2014

D. Schwarz

empty

Figure for Problem 6.

< spoiler > Solut ii detaliate si comentarii vin pe paginile urm atoare. Incercat i s a rezolvat i problemele nainte de a merge mai departe.

D. Schwarz

Zhautykov 2014

First Day Solutions

Problem 1. Points M, N, K lie on the sides BC, CA, AB respectively, of a triangle ABC , and are dierent from its vertices. The triangle M N K is called beautiful if the triangles M N K and ABC are similar (with the vertices respectively in this order). Show that if in the triangle ABC there are two beautiful triangles with a common vertex, then ABC is right-angled.

Solution. (L. Ploscaru) Say M N1 K1 and M N2 K2 are such two beautiful triangles. Let T = N1 K1 N2 K2 ; T exists, and belongs to the interior of ABC , since angles at M are equal to A. Then N1 M N2 = K1 M K2 . We then have M N1 T = M N2 T = B and M K1 T = M K2 T = C . It follows that M N2 N1 T and M K1 K2 T are cyclic quadrilaterals. So K2 K1 T = K2 M T and AN1 T = N2 N1 T = N2 M T , whence A = AK1 N1 + AN1 K1 = K2 M N2 = A, yielding A = A, so A = /2 . Moreover, notice it forces M to be the midpoint M0 of BC ; conversely then, all triangles M0 N K with KM0 N = /2 are beautiful. ste, n English mathematical parlance, Comentarii. Ceea ce se nume simple angle chasing. Reciproca merita si ea a remarcat a. Problem 2. Do there exist? functions f : R R such that a) f is a surjective function; and b) f (f (x)) = (x 1)f (x) + 2 for all x real.

Solution. We have f (f (f (x))) = f f (f (x)) = f ((x 1)f (x) + 2) and also f (f (f (x))) = (f (x) 1)f (f (x)) + 2 = (f (x) 1)((x 1)f (x) + 2) + 2 = (f (x) 1)(x 1)f (x) + 2(f (x) 1) + 2 = f (x)((f (x) 1)(x 1) + 2), so Let f (a) = 0; then for x = a we get f (0) = 2, and then for x = 0 we get f (2) = 0. Now for x = 1 we get f (1) = 0. Taking f (b) = 1 leads us to b = 1, and then to f (1) = 1. Finally, taking f (c) = 1 leads to c = 2, but that means 0 = f (2) = 1, absurd. Thus the answer is No . f ((x 1)f (x) + 2) = f (x)((x 1)f (x) + 2 (x 1)).

an a la obt inerea Comentarii. Jonglerii cu valori particulare (mici), p unei contradict ii ... Important este c a undeva pe parcurs trebuie exprimat f (f (f (z ))) n dou a feluri diferite, ceea ce este tipic pentru ecuat ii funct ionale cont in and iterata funct iei. Ar interesant de v azut n ce m asur a putem relaxa condit iile, sau ce fenomen se ascunde aici!?! F ar a condit ia a) de surjectivitate, o solut ie banal a este f (0) = 2, f (x) = 0 pentru x = 0. Putem oare g asi toate solut iile? Problem 3. There are given 100 distinct positive integers. We call a pair of integers among them good if the ratio of its elements is either 2 or 3. What is the maximum number g of good pairs that these 100 numbers can form? (A same number can be used in several pairs.) 3

Zhautykov 2014

D. Schwarz

Solution. Like so often in Russian problems, numbers are used instead of generic symbols. Let us therefore denote 10 = n > 1, 2 = k > 1, 3 = > 1, with the extra condition both k and arent powers of a same number. Consider the digraph G whose set of vertices V (G) is made of v = n2 distinct positive integers, and whose set of edges E (G) is made by the pairs (a, b) V (G) V (G) with a | b. For each positive integer m consider now the (not induced) spanning subdigraph Gm of G (with V (Gm ) = V (G) and so vm = v = n2 vertices), and whose edges are the pairs (a, b) G G with b = ma. Moreover, it is clear that E (Gm ) E (Gm ) = for m = m (since if (a, b) E (G) then (a, b) E (Gb/a ) only), and also E ( Gm ) = E ( G)
m1

(but that is irrelevant). Since the good pairs are precisely the edges of Gk and G together, we need to maximize their number g . A digraph Gm is clearly a union of some nm disjoint (directed) paths Pm,i ,
nm

with lengths (Pm,i ) = m,i , 0 m,i n2 1, such that

nm

(m,i +1) = n2 ,
i=1

and containing em =
i=1

m,i edges (zero-length paths, i.e. isolated vertices,

are possible, allowed, and duly considered). The defect of the graph Gm is taken to be v em = nm . We therefore need to maximize g = ek + e , or equivalently, to minimize the defect = nk + n . Using the model V (G) = Vx = {k i1 j 1 x | 1 i, j n}, we have nk = n = n, therefore = 2n, so g = 2n(n 1) . To prove value 2n is a minimum for is almost obvious. We have k,i n 1 for all 1 i nk (by the condition on k and , we have |Pk,i P,j | 1 for all 1 i nk and
nk nk

n2 nk n , and so = nk + n 2 nk n = 2n. Moreover, we see equality occurs if and only if nk = n = n and k,i = ,i = n 1 for all 1 i n, thus only for the sets Vx described above. R aspunsul este deci g = 180 . a ce ideea vine, problema este aproape trivial a, cu Comentarii. Odat detaliile tehnice ind aproape fort ate. Valorile particulare folosite arunc a doar un v al de umbr a asupra situat iei de fapt (mai ales ocultul 100 = 102 )! Laticea de divizibilitate a celor n2 numere este considerat a n mod natural, si conduce la facila num ar atoare de mai sus.
2Say |P P | 2; then take a | b P P , so that b = k x a and b = y a. Then k,i ,j k,i ,j
s s
i pk i and = i p i . Then for each 1 i s

1 j n ), so n nk = ek =

i=1

k,i

i=1

(n 1) = nk n nk , therefore

kx = y , wlog with gcd(x, y ) = 1. Say k =

i=1 s s

i i we have pxk = py i i , so xki = yi , therefore x | i and y | ki , with ki /y = i /x N . Let

i=1

r=
i=1

pi i

k /y

=
i=1

pi i

/x

; then k = ry and = rx , in contradiction with the condition that

both k and arent simultaneously powers of a same number.

D. Schwarz

Zhautykov 2014

Second Day Solutions

Problem 4. Does it exist? a polynomial P (x) with integer coecients, such that P (1 + 3) = 2 + 3 and P (3 + 5) = 3 + 5. Solution. The answer is No . The polynomial P (x) x has root 3 + 5, and since it has integer coecients, it also has root 3 5. The quadratic having these two roots is x2 6x +4. Therefore P (x) x = (x2 6x +4)Q( x) for some polynomial Q ( x ) with integer coecients. Plugging in x = 1 + 3 we also we nd 1 = (2 4 3)Q(1+ 3). From known conjugates properties do have 1 = (2+4 3)Q(1 3), so (multiply) 1 = 44Q(1+ 3)Q(1 3), impossible, since again by the properties of conjugates Q(1 + 3)Q(1 3) = Q(1 + 3)Q(1 + 3) = Q(1 + 3)Q(1 + 3) must be some integer k , and we cannot have 1 = 44k . ii legate de r ad acinile conjugate ale unor Comentarii. Simple considerat irat ionale p atratice pentru un polinom cu coecient i ntregi. Multiple solut ii, bazate toate pe aceste propriet a ti, sunt posibile, iar problema este de o tehnicalitate elementar a si articial a. Problem 5. Let U = {1, 2, 3, . . . , 2014}. For all a, b, c N let f (a, b, c) be the number of ordered sextuplets (X1 , X2 , X3 , Y1 , Y2 , Y3 ) of subsets of U , satisfying the following conditions (i) Y1 X1 U and |X1 | = a; (ii) Y2 X2 U \ Y1 and |X2 | = b; (iii) Y3 X3 U \ (Y1 Y2 ) and |X3 | = c. Prove f ( (a), (b), (c)) does not change, for permutations of a, b, c. Solution. In order to avoid any confusion between the letters a, b, c and their numerical values (as cardinalities of sets), the most convenient way will be to denote by || the cardinality symbolized by any such letter . We can now consider the true 3-element set {a, b, c}, and the canonical bijection : {1, 2, 3} {a, b, c} given by (1) = a, (2) = b, (3) = c. Let us now consider any permutation of {a, b, c}. We will denote by F the family of ordered sextuplets (X1 , X2 , X3 , Y1 , Y2 , Y3 ) of subsets of U satisfying the conditions of the statement but, under our notations, having |X1 | = | (a)|, |X2 | = | (b)|, |X3 | = | (c)|. We will also denote by F the family of doubletons {(Xa , Xb , Xc ), Y }, with Xa , Xb , Xc subsets of U having |Xa | = |a|, |Xb | = |b|, |Xc | = |c|, and Y X = Xa Xb Xc . The set Y uniquely partitions into 7 classes (some of them maybe empty), by indexed the non-empty subsets S of {a, b, c} via YS = Y 5
S

( U \ X ) .

Zhautykov 2014

D. Schwarz

We will now establish a bijection between F and F . This will show that f ( (a), (b), (c)) = |F| is constant over all permutations . We send an element {(Xa , Xb , Xc ), Y } F into the sextuplet (X1 , X2 , X3 , Y1 , Y2 , Y3 ) given by, and easily veried it actually belongs to F , X1 = X(a) , X2 = X(b) , X3 = X(c) (i.e. Xi = X((i)) for i {1, 2, 3}), Y1 = Y{(a)} , Y2 = Y{(b)} Y{(a),(b)} , Y3 = Y{(c)} Y{(a),(c)} Y{(b),(c)} Y{(a),(b),(c)} . We also send an element (X1 , X2 , X3 , Y1 , Y2 , Y3 ) F into the doubleton {(Xa , Xb , Xc ), Y } given by, and easily veried it actually belongs to F , with the permutation of {1, 2, 3} induced by via (i) = 1 ( ((i))) for all i {1, 2, 3}, Xa = X (1) , Xb = X (2) , Xc = X (3) (i.e. X(i) = X (i) for i {1, 2, 3}), Y = Y1 Y2 Y3 . It is immediate to see this mapping is a bijection, due to the unicity of the partitioning described in the above. Visualizing the Venn diagrams should tremendously help in understanding our considerations. The only diculty resides in providing a luminous write-up of the argumentation, the underlying phenomenon being in fact almost trivial. The key element of this solution is to consider the unique partitioning of the set Y = Y1 Y2 Y3 induced by the three sets of cardinalities a, b, c. a tenie de enunt ! Iar cerint a este aproape Comentarii. Pouah ... ce ur evident a, doar c a solut ia este cam lung de scris, formalizat, si explicat. Evident, valoarea 2014 nu joac a niciun rol. Problem 6. Let ABCD be a convex quadrilateral, partitioned by four lines as in the picture below, such that the meeting points of these lines all lie on the diagonals of ABCD. Prove that if the corner quadrilaterals 1, 2, 3 and the center quadrilateral 4 are all tangential, then the corner quadrilateral 5 is also tangential.

Solution. It seems the corner quadrilateral 5 is tangential if and only if quadrilateral ABCD is itself tangential; but no progress is thus yet made. Not being a geometer, I will wait for a decent solution to present itself (and it did! see the presentation on the last page). 6

D. Schwarz

Zhautykov 2014

Comentarii Finale. Au participat 57 de echipe, cu 368 de participant i din 16 ta ri (pentru trei discipline), printre care Bielorusia, Bulgaria, India, Indonezia, Kazahstan, Mongolia, Rusia, Rom ania (C. N. Tudor Vianu), Serbia, Ucraina. Rezultatele delegat iei din Rom ania (C. N. Tudor Vianu) Disciplina Nume Rezultat

Matematic a Irina Cristali Lucian Nit a Horia Banciu Fizic a Cristian Dut escu Bogdan Cionca

2/42 1/42 1/42 7,6/40 7,5/40

Informatic a

Petru-Eric Stavarache 495/600 Aur Alexandru V aleanu 334/600 Bronz

Echipa s-a clasat pe locul 22/57 (mult umit a de departe rezultatelor de la Informatic a rezultatele de la Matematic a au fost dezam agitor de slabe; cu at at mai caraghios este faptul c a ambii profesori nsot itori au fost ale si de la aceast a disciplin a)! Marele Premiu a fost luat de echipa din Moscova. La Matematic a au existat dou a punctaje perfecte 42/42, ambele din Rusia. Nici problemele, nici cu at at mai put in solut iile ociale, nu au fost a sate, dar lista complet a a participant ilor si rezultatele nale pot g asite la http://izho.kz/2014/. Urm atorul concurs internat ional este 7th Master of Mathematics 2014 g azduit chiar de C. N. Tudor Vianu! Site-ul ocial http://rmms.lbi.ro/ nc a nu face anunt ul concursului, care ar trebui s a se desf a soare n perioada 26 februarie 3 martie 2014, dar probleme logistice ar putea s a-i pun a n pericol fragila sa existent a ... S a ne bucur am deci c a 3rd EGMO 2014 este n bun a stare de s an atate si se va desf a sura n perioada 10 16 aprilie 2014 (imediat dup a Etapa Nat ional a a Olimpiadei de Matematic a de la Cluj), n Antalia, Turcia site ocial https://www.egmo.org/egmos/egmo3/ (vezi anunt urile deja postate pe http://www.viitoriolimpici.ro/). In ultimul moment, am intrat n posesia solut iilor ociale (curtoazie Ionut Oni sor). Paginile urm atoare cont in comentarii suplimentare (generate de aceste solut ii), solut ia ocial a la Problema 6, si c ateva statistici referitoare la rezultatele sect iunii Matematic a a concursului. Dan Schwarz, 21 ianuarie 2014 7

Zhautykov 2014

D. Schwarz

Statistici asupra punctajelor, pe probleme. Problema/Puncte P1 P2 P3 P4 P5 P6 7 115 78 14 60 14 7 6 0 0 3 7 2 1 5 1 0 5 4 3 0 4 4 1 0 2 1 0 3 1 2 0 0 0 2 2 1 0 Medie 1 8 41 4, 9 3 25 103 3, 4 4 22 123 1, 0 0 5 93 2, 9 0 24 127 0, 9 0 6 155 0, 4

Dup a cum am ment ionat mai sus, au fost 2 scoruri perfecte de 42/42, din Rusia. Cele 7 scoruri maxime la problema 6 au fost si ele toate din Rusia. Ziua a doua s-a dovedit mult mai grea dec at prima. S-au acordat 18 medalii de aur (42-29 puncte), 23 medalii de argint (28-22 puncte) si 32 medalii de bronz (21-15 puncte); n total 73 de medalii pentru 171 de participant i, mult mai put in dec at proport ia de 50% aplicat a de obicei (din cele ce v ad, s-au aplicat mai degrab a medal cut-os cele mai frecvente de la OIM, si nu procentajele; probabil datorit a nivelului sc azut al celor mai mult i dintre concurent i comparativ cu nivelul problemelor date spre rezolvare).

Comentarii Suplimentare. Problema 1. Autorii nu atrag atent ia c a v arful comun a dou a triunghiuri frumoase va cu trebuint a mijloc de latur a, si c a atunci orice triunghi dreptunghic de acest v arf este frumos. Problema 2. S i autorii consider a posibilitatea ca f s a nu e surjectiv a, x2 si ofer a at at exemplul dat si de mine, c at si exemplul f (x) = pentru x1 a chestiunea g asirii tuturor solut iilor x = 1, f (1) = 0. Dar nici ei nu consider pentru acest caz (tsk, tsk, Igor Voronovich ...). Problema 3. Solut ia ocial a a Problemei 3 o reformuleaz a astfel Fiind dat un num ar de 100 = n2 puncte laticiale, care este num arul maxim de perechi dintre aceste puncte, aate la distant a 1 unul de cel alalt? Echivalent a celor dou a enunt uri se face prin corespondent a (i, j ) 2i 3j pentru punctele laticiale date; invers, pentru numerele ntregi date, ecare astfel de num ar b se poate scrie b = 2i 3j a, cu i, j 0 si cmmdc(a, 6) = 1. Acum, pentru ecare valoare a, facem corespondent a 2i 3j a (i + ua , j + va ), unde ua , va sunt alese astfel nc at pentru a = a s a nu avem puncte de tipul (i + ua , j + va ) la distant a 0 sau 1 de puncte de tipul (i + ua , j + va ). 8

D. Schwarz

Zhautykov 2014

Se demonstreaz a apoi c a num arul maxim de perechi dintre cele n2 puncte laticiale, cu puncte aate la distant a 1 unul de cel alalt, se obt ine c and punctele formeaz a un p atrat de latur a n; metoda este practic identic a celei folosite de mine, n contextul u sor diferit al rat ionamentului meu. De remarcat c a aceast a solut ie folose ste din plin faptul c a 2 si 3 sunt numere prime iar a si, solut ia nu mai merge pentru valori k si ca cele folosite de mine, c aci atunci factorizarea unui num ar b ca b = k i j a nu mai este neap arat unic a, si corespondent a cu puncte laticiale este primejduit a. Mai mult, nici m acar faptul c a lucr am cu n2 numere nu este chiar esent ial; acelea si rat ionamente a pentru un num ar N , cu valoarea optim a funct ioneaz 2 ind n acest caz 2 N ( N 1), valoarea particular a N = n permit and doar exprimarea rezultatului ntr-o form a simpl a. Problema 4. Solut ia ocial a este inutil de complicat a, dar ideea merit a ment ionat a. Pentru orice polinom P Z [x], numerele ntregi a, b, c, d pentru si P (3 + 5) = c + d 5 sunt evident unic care P (1 + 3) = a + b 3 determinate. Numim polinomul regulat dac a a c b d 0 (mod 2). Se arat a u sor apoi c a dac aP si Q sunt polinoame regulate, atunci at at orice combinat ie liniar a cu coecient i ntregi kP + Q, c at si produsul P Q sunt polinoame regulate. Dar deoarece polinomul constant 1(x) = 1 si polinomul X(x) = x sunt regulate, si deoarece 1 si X genereaz a inelul Z[x], rezult a c a orice polinom P Z[x] este regulat; polinomul cerut nu este ns a regulat, c aci a c = 1 0 (mod 2). Il b anuiesc pe Alexander Golovanov ca f apta s la acest efort ... de altfel meritoriu, dar vezi comentariul care urmeaz a relativ la adev arata sa valoare. Articiul este ns a de ret inut, folositor poate pentru alte instant e asem an atoare. De si metoda este elegant a, dac a am avut ns a P (1 + 3) = 3 + 3, polinomul ar fost regulat, si solut ia de mai sus nu mai tine, n timp ce banala solut ie cu conjugate conduce la relat ia imposibil a 4 = 44k . Dup a mine, acest fenomen este echivalent cu o not a de u sor ridicul ... o solut ie sosticat a, dar care funct ioneaz a n (mult) mai put ine cazuri dec at solut ia mitoc aneasc a. Problema 5. Solut ia ocial a asociaz a ec arui sextuplet (care satisface condit iile) (X1 , X2 , X3 , Y1 , Y2 , Y3 ) sirurile A = (ai )1i2014 , B = (bi )1i2014 , C = (ci )1i2014 , denite prin ai = 2 dac a i Y1 , ai = 1 dac a i X1 \ Y 1 , si ai = 0 altfel; la fel sirurile B (pentru indicele 2) si C (pentru indicele 3). Sunt apoi interpretate cele trei condit ii din enunt n termeni relativi la aceste siruri, si se stabile ste c a f (a, b, c) este num arul de triplete (A, B, C ) cu aceste propriet a ti. Finalmente, sunt descrise biject ii ntre tripletele corespunz and ordinii (a, b, c) si cele corespunz and ordinii (b, a, c), duc and la f (a, b, c) = f (b, a, c), si ntre tripletele corespunz and ordinii (a, b, c) si cele corespunz and ordinii (a, c, b), duc and la f (a, b, c) = f (a, c, b), si aceasta este deajuns. 9

Zhautykov 2014

D. Schwarz

De fapt, sirurile A, B, C sunt un fel de extensie a funct iilor indicatoare; de exemplu A este secvent a valorilor unei funct ii indicatoare a sistemului de mult imi Y1 X1 U . Cu toate acestea, metoda este extrem de articial a si ne-intuitiv a comparat a cu ideea demonstrat iei date de mine. Vericarea transform arilor date de si este delicat a, ar at and c a alegerea sirurilor a fost f acut a cum se zice cu mintea omului cea de pe urm a. Problema 6. Let ABCD be a convex quadrilateral, partitioned by four lines as in the picture below, such that the meeting points of these lines all lie on the diagonals of ABCD. Prove that if the corner quadrilaterals 1, 2, 3 and the center quadrilateral 4 are all tangential, then the corner quadrilateral 5 is also tangential.

Ocial Solution. We use the following Lemma.3 A convex quadrilateral XY ZT is tangential if and only if 1 1 1 1 tan Y XZ tan T ZX = tan Y ZX tan T XZ. 2 2 2 2 Proof. From what I see, a simple derivation from Pitots theorem, which equivalates the tangentiality of XY ZT with XY + ZT = Y Z + T X . Applying the Lemma to quadrilaterals 1, 4 and 3 shows that quadrilateral ABCD is tangential. Applying again the Lemma to quadrilaterals 2, 4 and ABCD shows that quadrilateral 5 is tangential. In other words, if the inner/outer quadrilateral, and any three of the corner quadrilaterals are tangential, so will be the outer/inner one, and the remaining fourth corner quadrilateral; the conguration is symmetric. S a remarc am c a celelalte patru patrulatere prezente n gur a nu sunt neap arat circumscriptibile (dar pot ). Am o oarecare b anuial a c a aceast a ar putut arhicunoscut Lema a n anumite cercuri (Rusia este o fort a n geometria sintetic a de olimpiade). Enough said ... D. S. 22 ianuarie 2014
3O cercetare pe Internet este de fapt cunoscut mi descoper a c a aceast a Lema a ca

rezultat al lui M. Iosifescu, G.M.-B (1953-1954)! vezi http://en.wikipedia.org/wiki/Tangential quadrilateral/ (wiki) si/sau http://forumgeom.fau.edu/FG2011volume11/FG201108.pdf (pages 75-77).

10