Singapore Mathematical Society Singapore Mathematical Olympiad (SMO) 2007 (Open Section, Round 1 Solutions)

1. Ans: 21

If A is the set of multiples of 5 in {1, 2, 3, ... , 100}, then IAI = 20 and la - bl ~ 5 for every two numbers in A. Thus, if IAI = 20, the existence such two numbers cannot be guaranteed. However, if IAI = 21, by the pigeonhole principle, there must be two numbers a, b in one of the following twenty sets:

{1,2,3,4,5},{6,7,8,9,10},···,{96,97,98,99,100},

and so la - bl ::; 4. Thus the answer is 21.

2. Ans: 24068

In such a binary sequence, O's either appear singly or in blocks of 2. If the sequence has exactly m blocks of double O's, then there are 10 - 2m single O's. The number of such binary sequences is

(11) x (11 - m ) .

m 10 - 2m

Thus the answer is

5 )

11 11-m

l=o (m) x (,0 _ 2m ~ 24068.

3. Ans: 5036

Consider such a figure as a plane graph G. Then the answer is equal to the number of interior faces of G. The number of vertices in G is 20 + C40). The sum

degrees is

(240)

20 x 21 + 4 x

and so the number of edges in G is

(240) .

10 x 21 + 2 x

Hence the number of interior faces, by Euler's formula, is

41

Second Solution: Let Pn be the number of such regions with ti points on the circumference. Then PI = 1, P2 = 2 and in general, for ti 2': 2, Pn+1 = Pn +n+ 2..:~12 i(ni). This can be obtained as follows. Suppose there are n + 1 points ao, ... ,an in that order on the circumference. The chords formed by aI, ... , an create Pn regions. The chord aOal adds one region. The chord aOal adds 1 + 1 x (n - 2) regions as this chord intersects the existing chords in 1 x (n - 2) points. Similarly, the chord aOa2 adds 1 + 2 x (n - 3) regions, etc. From this, it is easy to show that

Pn=l+ (;) + (n;2) +2(n;3) + +(n-3)(~)

=1+ (;) + (n;l) +1(n;3) + (n-4)(~)

= ... = 1 + (;) + (n; 1) + (n; 2) + ... + (~) = 1 + (;) + (:)

4. Ans: 2844

If only one digit appears, then there are 9 such numbers. If the two digits that appear are both nonzero, then the number of such numbers is

2 x (;) (~) = 2520.

If one of two digits that appear is 0, then the number of such numbers is

( (~) + (~) ) x (~) = 315.

Hence the answer is 9 + 2520 + 315 = 2844.

5. Ans: 20791 Let

A6 = {k E A: 61 k}; A2 = {k E A: 21 k, 6 t k}; A3 = {k E A: 31 k, 6 t k}.

Note that

IA61 = ll0600 J = 166; IA21 = ll02° J -l10600 J = 334;

IA31 = l10300 J -ll0600 J = 167.

42

For the product a x b to be divisible by 6, either (i) one or both of them are in A6 or (ii) one is in A2 and the other is in A3. Hence

(166)

m = 2 + 166 x (1000 - 166) + 334 x 167 = 207917.

6. Ans: 10626

Let v = 20 - (x + y + z + u). Then v 2: 0 if and only if x + y + z + u S; 20. Hence the answer is equal to the number of non-negative integer solutions of the following equation:

x + y + z u + v = 20 and the answer is (~4) = 10626.

7. Ans: 16800

Either one student receives three prizes; or two students are each awarded two prizes. Thus the answer is

(;) x 5! + (~) x (~)

1

x "2 x 5! = 16800.

8. Ans: 77

Assume that [CEF] = 1. Then, [AFC] = k and [BFC] = 11. Since AD = 7DB, k = [AFC] = 7[BFC] = 77.

A

D

B

E

c

9. Ans: 6544

For i = 0,1,2, let

Ai := {I S; k S; 50 : 3 I (k - i)}.

Then IAol = 16, IAII = 17 and IA21 = 17. It can be shown that 3 I (a + b + c) if and only if either {a, b, c} c::: Ai for some i or {a, b, c} n Ai i= 0 for all i = 1,2,3. Thus the answer is

43

10. Ans: 56841

We may assume Xl < X2 < ... < X1970· For 1 < Xi < Xj with i < i, we have xt + x] :s; xt + x] + 3(xj - Xi)(Xj + Xi - 1) = (Xi - 1)3 + (Xj + 1)3. Thus when Xl = X2 = ... = Xl969 = 1 and Xl970 = 38, the expression xr + x~ + ... + Xr970 attains its maximum value of 1969 + 383 = 56841.

11. Ans: 9

Substituting the first equation be = a2 - 8a + 7 into the second equation, we have (b + e)2 = (a - 1)2 so that b + e = ±(a - 1). That means band e are roots of the quadratic equation X2 ~ (a - l)x + (a2 - 8a + 7) = O. Thus its discriminant 6 = [~(a - 1)]2 - 4(a2 - 8a + 7) ~ 0, or equivalently, 1 :s; a :s; 9. For b = e = 4, a = 9 satisfies the two equations. Thus the largest value of a is 9.

12. Ans: 1506

Let a; = l i2 I 2007 J . Note that (~6~r - 2~~7 = 22~6i < 1 if and only if ti < 2006/2 = 1003. Since, al = 0, and alO03 = 501, we see that al, ... ,alO03 assume all the values from 0 to 501. We also conclude that alO04, ... ,a2007 are mutually distinct integers. Therefore, the answer is 502 + 1004 = 1506.

13. Ans: 112

Since the number (a + b) + (a - b) + ab + alb = (alb)(b + 1)2 is a perfect square with band b + 1 relatively prime, the number alb must be a perfect square. Let alb = n2. As a > b, the number ti ~ 2 so that a = bn2 :s; 200. From this, a can be determined once band ti are chosen. Hence, it suffices to count the number of pairs of (b, n) satisfying bn2 :s; 200 with b ~ 1 and n ~ 2. Hence the answer is l200/22J + ... + l200/142J = 112.

14. This question has been deleted.

15. Ans: 52

First we know that AE = AB cos A and AP = AC cos A. By cosine rule, EP2 = AE2 + AF2 - 2AE x AF cos A = cos2 A(AB2 + AC2 - 2AB x AC cos A) = BC2 cos2 A. Therefore EF = BC cos A. It is easy to see that A, E, H, F lie on a circle with diameter AH. Thus AH = sfn~ = BC cot A = 39 x 1 = 52.

A

44

'-'---,----'-_ ---~--------------------------

16. Ans: 12

Let R be the midpoint of ~C. Then PO = RO = RQ. Let LOPQ = x. Then LPOR = 2LB = 8x and LBOA = 2LC = 12x. Also, LROQ = LA = 1800 -lOx. It follows that LPQO = 1800 - x - 8x - (1800 - 1Ox) = x and therefore PO = OQ. Thus DOQR is equilateral, so LROQ = 1800 - lOx = 600 and x = 120.

c

17. Ans: 3

Let DE = x, AC = a > 5 and CD = u. Then BD = V42 + x2 and cos LCAD = cos LBAD = 3/5. By the cosine rule applied to DADC, we have u2 = (3 + x)2 +

a2 - 2a(3 + x)(3/5). Using the angle bisector theorem, we have 42rx2 = ~~. Thus 25[(3 + X)2 + a2 - 2a(3 + x)(3/5)] = a2(16 + x2).

A

This can be simplified to (a2 - 25)x2 + 30(a - 5)x - 9(a - 5)2 = O. Since a > 5, we can cancel a common factor (a - 5) to get (a + 5)x2 + 30x - 9(a - 5) = equivalently (x + 3)((a + 5)x - 3(a - 5)) = O. Thus x = 3(a - 5)/(a + 5). From this we obtain (AC+AB) ED = (a+5) x = 3.

, AC-AB a-5

18. Ans: 1

Note that tan(90° - e) = 1/ tan e for 00 < O < 900and that the product is positive. Setting j = 46 - k,

45 45 1

II tan(2k - 1)0 = II tan(90 - (2j - 1)t = 45 . 0 = 1.

k=l j=l TIj=l tan(2J - 1)

45

•

19. Ans: 20

The triangle is a right triangle with area A = 5 x 120/2 = 300. The semiperimeter is s = ~ x (50 + 120 + 130) = 150. Hence the inradius is AI s = 20.

20. Ans: 4 Set

u=(~,Vc=b,~)

v=( a , b , c ).

~Vc=b~

Then u . v = Ilullllvll and hence u = ov for some 0: 2: O. Thus

and

b-a c +b d-c

abc

Hence b] a = clb = dl c, i.e., a, b, c, d is a geometric progression a, ra, r2a, r3a = 2a. Now bed] a3 = r6 = (r3)2 = 4.

21. Ans: 2

Putting 1 I x in place of x in the given equation yields

f(~) - tf(x) = -logx.

Solving for f(x), we obtain f(x) = ~ logx. Hence f(1000) = 2.

22. Ans: 20 We have

OD [BOC] OE [AOC] OF [AOB]
AD [ABC]' BE [ABC]' CF [ABC]'
Thus
OD OE OF_1 =? OE = 20.
AD + BE + CF- 46

23. Ans: 53808

For ti 2': 3, among the an such integers, let bn denote the number of those that end with 1. By symmetry, the number of those that end with 2 is also equal to bn. Also the number of those that end with 0 or 3 are both an-I. Thus

Among the bn integers that end with 1, the number of those that end with 11 is bn-I while the number of those that end with 01 or 31 are both an-2. Thus

Solving, we get an = 3an-1 +2an-2. Since al = 3 and a2 = 10, we get ag = 73368.

24. Ans: 1005

Partition all the possible remainders when divided by 2006 as follows:

(0), (1,2006), (2,3005), ... , (1003,1004).

Suppose S has 1005 elements. If S has two elements with their difference divisible by 2007, we are done. Otherwise the elements of S have distinct remainders when divided by 2007. By the pigeonhole principle, S has two integers whose sum is divisible by 2007. The set {1, 2, ... , 1004} does not have 2 elements, x, y such that x + y or x - y is divisible by 2007.

25. Ans: 7040

For better understanding, we consider the general case when P has n vertices, where n 2': 9. We first count the number of such triangles S having a particular vertex A. The number is (n;-7). (This can be obtained as follows. Let the triangle be ABC ordered clockwise. Then B has a "left" neighbour and C has a "right" neighbour. The location of of Band C are uniquely determined by their neighbours. Besides A, B, C and the two vertices to the left and two vertices to the right of A, the two neighbours can be chosen from the remaining ti - 7 vertices. Thus the required answer is } (n;-7) = 7040.

47

Singapore Mathematical Society Singapore Mathematical Olympiad (SMO) 2007 (Open Section, Round 2 Solutions)

1. Without loss of generality, we may assume that all the o.; are positive, else we just change the sign of Xi. Since

we have L~=I o.; :::; Vii· There are k" integer sequences (tI, t2, ... , tn) satisfying o < ti < k -1 and for each such sequence we have 0 < L~=I aiti :::; (k -l)fo. Now divide the interval [0, (k -1) Vii] into k" -1 equal parts. By the pigeonhole principle, there must exist 2 nonnegative sequences (YI, Y2,·· ., Yn) and (ZI' Z2,·· ., zn) such

that I L~=I aiYi - L~=I aizi I < (k;;n12f· Set Xi = u. - Zi to satisfy the condition.

2. Suppose to the contrary that (x - aI)(x - a2) ... (x - an) - 1 = f(x )g(x) for some polynomials f(x) and g(x) with integer coefficients and deg (f(x)), deg (g(x)) ~ 1. Then f(ai)g(ai) = -1 for i = 1,2, ... , n implies that f(ai) = 1 and g(ai) = -1 or f(ai) = -1 and g(ai) = 1. Therefore, if we set h(x) = f(x) + g(x), then h(ai) = 0 for all i = 1,2, ... ,n. As deg(h(x)):::; max(deg(f(x)),deg(g(x))) < n, the polynomial equation h( x) = 0 cannot have ti distinct roots. It follows that h( x) must be the zero polynomial. Thus f(x) = -g(x), and therefore

for all real values of x. But this leads to a contradiction since we can choose a value for x large enough so that (x - ad (x - a2) ... (x - an) - 1 is positive.

2nd solution: We start off as in the first solution. Then instead of defining h( x), we proceed as follows. Let f(ai) = 1,1 :::; i:::; k and f(ai) = -1, k + 1 :::; i :::; n. Then, g(ai) = -1, 1 :::; i :::; k and g(ai) = 1, k + 1 :::; i :::; n, Therefore deg (f(x) - 1) = degf(x) .> max(k,n - k) .> k+(~-k) = ~. Similarly degg(x) .> ~. However deg f(x)+degg(x) = n, and thus n is even with deg f(x) = degg(x) = k = ~. Thus f(x)-l = h(x-ad(x-a2)'" (X-ak), and g(x)+l = b2(x-ad(x-a2)'" (x-ad for some bI, b2 E Z. Together we get f(x )g(x) + f(x) - g(x) -1 = b, b2[(x - ad(xa2)'" (x - ak)]2. By comparing coefficient of the x" term, hb2 = 1. This give us f(x) - 1 = g(x) + 1. Similarly, we have f(x) + 1 = g(x) - 1, a contradiction.

49

•

3. Let the point of contact of the two circles be P. First we show that AI, P and B2 are collinear. Let the common tangent at P meet C A at D and C B at E. Let LABC = b = LCAB = LAlCBl, LACB = c, LAlCP = x and LBlCP = y. Then x + y = band 2b + c = 1800• We have LP BIA = x, LBIP E = y. Therefore, by considering P BlBE, LBEP = 2x. Hence LEP B2 = x and consequently, LBI P B2 = X + y = b. This implies that AI, P and B2 are collinear. Similarly A2, P and Bl are collinear. Then ,6AlBC rv ,6AlCBl, and ,6CABI rv ,6AlCBl1 whence ,6AlBC r-;» ,6CABI. Thus AC/ABI = AIB/BC and BC2 = AlB· ABl, since AC = BC. Also ,6AA2Bl rv ,6BAlB2. Thus ABdAA2 = BB2/AIB and whence AA~ = AlB· ABI. Thus B is the midpoint of CB2. Since AB II A2B2, we have A2B2 = 2AB as required.

c

2nd solution: Let rl be the circumcircle of ,6AlBlC and its centre be 01) let the other circle T 2 has center O2, and the point of contact of the two circles be P. Now since CA2 and CB2 are tangent to r2, we have CA2 = CB2. Together with CA = CB, we have AA2 = BB2. This implies that ,6CAB rv ,6CA2B2 and AB II A2B2. Now LA202B2 = 1800 - LACB = 2LCAB = 2LAlCB1 = LAlOlBl· Thus, isosceles triangles AlOlBl and A202B2 are similar. Since AB II A2B2, AlOl II 02B2, also note that 01P02 is a straight line. Therefore, we have 2LAlBlP = LAIOlP = LP02B2 = 2LPA2B2. This implies that BIPA2 is a straight line. Similarly, AlPB2 is a straight line. Now we let rl intersects CA and CB at D and E respectively, and let DAI intersects EBI at G. By Pascal's Theorem on rl and the hexagons CEBIP AID, we have A2, G and B2 collinear. Using the fact that ,6ACBI rv ,6CAIBI rv BIAlC, we have LGAlBl = LDAIA = LDCBl = LCAIBI. Similarly, LGBIAI = LCBIAI. This implies that G is the image of C under reflection of line AB. Since G is on A2B2, A2B2 is twice as far as AB from C. Thus, A2B2 = 2AB.

50

3rd solution: Let r I be the circumcircle of L,Al H, C and its centre be 01, let the other circle r2 has center O2, Now since CA2 and CB2 are tangent to r2, we have CA2 = CB2. This implies that L,CAB rv L,CA2B2. Let us perform inversion with center C and radius CA. Let the image of AI, A2, Bl, B2 under this inversion be A~, A;, B~, B~ respectively. A, Band C remain invariant. The inversion keeps every line that passes through C invariant. Now the image of the line AAlBlB is the circumcircle of L,CAB, let it be r3, and the image of rl is the line AIBI. Thus the image of r2 is tangent to AlBl, AC and BC and is thus the incircle of L,ABC and touches the sides AC and BC at A; and B~, respectively. Thus

CA; _ CB~ - 1 hi h j Ii th t CA2 - CB2 - 2 HAd B th

CA - CB - 2' w IC Imp Ies a CA - CB - . ence an are e

midpoints of A2C and B2C, respectively. Thus A2B2 = 2AB.

4. We show that f is the identity function. First we observe that f is an injective function:

f(m) = f(n) =? f(m) + f(n) = f(n) + f(n)

=? f(J(m) + f(n)) = f(J(n) + f(n))

=? m+n=n+n

=? m= n

Let k > 1 be arbitrary. From the original equation, we have the equations

f(f(k+l)+ f(k-l)) = (k+l)+(k-l) = 2k, and f(f(k)+ f(k)) = k+]: = 2k.

Since f is injective, we have

f(k + 1) + f(k - 1) = f(k) + f(k) or f(k + 1) - f(k) = f(k) - f(k - 1).

This characterizes f as an arithmetic progression, so we may write f (n) = b (n - l)t where b = f(l) and t is the common difference. The original becomes b + [(b + (m - l)t) + (b + (n - l)t) - llt = m + n, which simplifies to (3b - 2t - 1) + (m + n)t = m + n, Comparing coefficients, we conclude that t = 1 and b = 1. Thus f (n) = n, as claimed. Clearly, this function satisfies the original functional equation.

51

•

5. The answer is x = 420.

Let Pl,P2,P3, ... be all the primes arranged in increasing order. By Bertrand's Postulate, we have Pi < Pi+l < 2Pi for all i E N, thus we have Pk+l < 2Pk < 4Pk-l < 8Pk-2 which implies that 64pkPk-1Pk-2 > Pk+13.

Let Pk :::; ijX < Pk+l for some kEN. Note that Pi I x for i = 1,2, ... , k. Suppose k ~ 5, then ijX ~ P5 = 11. Since 11 > 23 and 11 > 32, we have 2332 I x. Since k ~ 5, gcd(PkPk-1Pk-2, 2332) = 1 and thus 2332pkPk_1Pk_2 I x. This means we have x ~ 72pkPk-1Pk-2 > 64pkPk-1Pk-2 > Pk+13, implying Pk+l < ijX, which is a contradiction.Thus k < 5 and consequently, ijX < 11 or x < 1331.

Next, we notice that the integer 420 is divisible by all positive integers j, ij425, thus x ~ 420 ::::} ijX > 7. It then follows that x is divisible by 22 . 3 . 5 . 7 = 420.

Finally, suppose ijX ~ 9. We then have 23 .32 ·5·7 I x, i.e., x ~ 23 .32 ·5·7 = 2520, which is a contradiction since x < 1331. Thus ijX < 9,= or x < 729. Since 420 I x and x < 729, we have x = 420.

Alternatively, since x < 1331 and 420 I x, we only need to check the cases x = 420,840,1260.

52