Solutions to the Shortlisted Problems of IMO 1992

4.33 Solutions to the Shortlisted Problems of IMO 1992

1. Assume that a pair (x, y) with x < y satisfies the required conditions. We claim that the pair (y, x 1 ) also satisfies the conditions, where x 1 = y 2 +m x (note that x 1 >y is a positive integer). This will imply the desired result, since starting from the

pair (1, 1) we can obtain arbitrarily many solutions. First, we show that gcd (x 1 , y) = 1. Suppose to the contrary that gcd(x 1 , y) = d >

1. Then d

2 |x 2 1 |y + m ⇒ d | m, which implies d | y | x + m ⇒ d | x. But this last is impossible, since gcd (x, y) = 1. Thus it remains to show that x 1 |y 2 +m

and y 2 |x 1 + m. The former relation is obvious. Since gcd(x, y) = 1, the latter is equivalent to y | (xx 1 ) 2 + mx 2 =y 4 + 2my 2 +m 2 + mx 2 , which is true because y | m(m + x 2 ) by the assumption. Hence (y, x 1 ) indeed satisfies all the required conditions.

Remark. The original problem asked to prove the existence of a pair (x, y) of positive integers satisfying the given conditions such that x + y ≤ m + 1. The problem in this formulation is trivial, since the pair x = y = 1 satisfies the con- ditions. Moreover, this is sometimes the only solution with x + y ≤ m + 1. For

example, for m = 3 the least nontrivial solution is (x 0 ,y 0 ) = (1, 4).

2. Let us define x n inductively as x n = f (x n −1 ), where x 0 ≥ 0 is a fixed real number. It follows from the given equation in f that x n +2 = −ax n +1 + b(a + b)x n . The general solution to this equation is of the form

n = λ 1 b n + λ 2 (−a − b) ,

where λ 1 , λ 2 ∈ R satisfy x 0 = λ 1 + λ 2 and x 1 = λ 1 b − λ 2 (a + b). In order to have x n ≥ 0 for all n we must have λ 2 = 0. Hence x 0 = λ 1 and f (x 0 )=x 1 = λ 1 b = bx 0 . Since x 0 was arbitrary, we conclude that f (x) = bx is the only possible solution of the functional equation. It is easily verified that this is indeed a solution.

3. Consider two squares AB ′ CD ′ and A ′ BC ′ D . Since AC ⊥ BD, these two squares are homothetic, which implies that the lines AA ′ , BB ′ ,CC ′ , DD ′ are concurrent at

a certain point O. Since the rotation about A by 90 ◦

takes ∆ ABK into ∆ AFD , it follows that

K BK ⊥ DF. Denote by T the intersec-

tion of BK and DF. The rotation about A some point X by 90 ◦ maps BK into DF

TL if and only if T X bisects an angle be- tween BK and DF. Therefore ∠FTA =

D ∠AT K = 45 ◦ . Moreover, the quadri-

B ′ C lateral BA ′

A DT is cyclic. Therefore ∠BTA ′ = BDA ′ = 45 ◦ and consequently that the points A , T, A ′ are collinear. It

follows that the point O lies on a bisector of ∠BT D and therefore the rotation R about O by 90 ◦ takes BK into DF. Analogously, R maps the lines CE , DG, AI

into AH , BJ,CL. Hence the quadrilateral P 1 Q 1 R 1 S 1 is the image of the quadrilat- eral P 2 Q 2 R 2 S 2 , and the result follows.

564 4 Solutions

4. There are 36 possible edges in total. If not more than 3 edges are left undrawn, then we can choose 6 of the given 9 points no two of which are connected by an undrawn edge. These 6 points together with the edges between them form a two-colored complete graph, and thus by a well-known result there exists at least one monochromatic triangle. It follows that n ≤ 33. In order to show that n = 33, we shall give an example of a graph with 32 edges that does not contain a monochromatic triangle. Let us start with a complete

graph C 5 with 5 vertices. Its edges can be colored in two colors so that there is no monochromatic triangle ( Fig. 1 ). Furthermore, given a graph H with k vertices without monochromatic triangles, we can add to it a new vertex, join it to all vertices of H except A, and color each edge BX in the same way as AX . The obtained graph obviously contains no monochromatic triangles. Applying

this construction four times to the graph C 5 we arrive to an example like that shown on Fig. 2 .

Fig. 1 Fig. 2 Second solution. For simplicity, we call the colors red and blue.

Let r (k, l) be the least positive integer r such that each complete r-graph whose edges are colored in red and blue contains either a complete red k-graph or a complete blue l-graph. Also, let t (n, k) be the greatest possible number of edges in a graph with n vertices that does not contain a complete k-graph. These num- bers exist by the theorems of Ramsey and Turán. Let us assume that r (k, l) < n. Every graph with n vertices and t(n, r(k, l)) +1 edges contains a complete subgraph with r (k, l) vertices, and this subgraph con- tains either a red complete k-graph or a blue complete l-graph. We claim that t (n, r(k, l)) + 1 is the smallest number of edges with the above property. By the definition of r (k, l) there exists a coloring of the complete graph

H with r (k, l) − 1 vertices in two colors such that no red complete k-graph or blue complete l-graph exists. Let c ij

be the color in which the edge (i, j) of H is colored, 1 ≤ i < j ≤ r(k,l) − 1. Consider a complete r(k,l) − 1-partite graph

G with n vertices and exactly t (n, r(k, l)) edges and denote its partitions by P i ,

i = 1, . . . , r(k, l) − 1. If we color each edge of H between P i and P j (j < i) in the color c ij , we obviously obtain a graph with n vertices and t (n, r(k, l)) edges in two colors that contains neither a red complete k-graph nor a blue complete l -graph. Therefore the answer to our problem is t (9, r(3, 3)) + 1 = t(9, 6) + 1 = 33.

4.33 Shortlisted Problems 1992 565

5. Denote by K , L, M, and N the midpoints of the sides AB, BC,CD, and DA, re- spectively. The quadrilateral KLMN is a rhombus. We shall prove that O 1 O 3 k KM . Similarly, O 2 O 4 k LN, and the desired result follows immediately.

We have −−−→ O 1 O 3 = −−→ KM + O 1 K + MO −−→ 3 . Assume that ABCD is positively ori- ented. A rotational homothety R with angle −90 ◦

√ and coefficient 1 /

3 takes the vectors −→ BK and CM −→ into −−→ O 1 K and −−→ MO 3 respectively. Therefore

−−−→ O O KM −−→

Since LN ⊥ KM, it follows that R(LN) is parallel to KM and so is O 1 O 3 .

6. It is easy to see that f is injective and surjective. From f (x 2 + f (y)) = f ((−x) 2 +

f (y)) it follows that f (x) 2 = ( f (−x)) 2 , which implies f (−x) = − f (x) be- cause f is injective. Furthermore, there exists z ∈ R such that f (z) = 0. From

f (−z) = − f (z) = 0 we deduce that z = 0. Now we have f (x 2 ) = f (x 2 + f (0)) =

, and consequently f (x) = f ( x ) 2 > 0 for all x > 0. It also follows that f (x) < 0 for x < 0. In other words, f preserves sign. Now setting x > 0 and y = − f (x) in the given functional equation we obtain

0 + ( f (x)) 2 = f (x) 2

f (x − f (x)) = f ( x + f (−x)) = −x + f ( x ) 2 = −(x − f (x)). But since f preserves sign, this implies that f (x) = x for x > 0. Moreover, since

f (−x) = − f (x), it follows that f (x) = x for all x. It is easily verified that this is indeed a solution.

7. Let G 1 ,G 2 touch the chord BC at P , Q and touch the circle G at R, S respect- ively. Let D be the midpoint of the complementary arc BC of G. The homo- thety centered at R mapping G 1 onto G also maps the line BC onto a tangent of G parallel to BC. It follows that this line touches G at point D, which is therefore the image of P under the homothety. Hence R , P, and D are collinear. Since ∠DBP = ∠DCB = ∠DRB, it follows that △DBP ∼ △DRB and conse-

quently that DP · DR = DB 2 . Similarly, points S , Q, D are collinear and satisfy DQ ·DS = DB 2 = DP · DR. Hence D lies on the radical axis of the circles G 1 and

G 2 , i.e., on their common tangent AW , which also implies that AW bisects the √ angle BAD. Furthermore, since DB = DC = DW = DP · DR, it follows from the lemma of (SL99-14) that W is the incenter of △ABC.

Remark. According to the third solution of (SL93-3), both PW and QW contain the incenter of △ABC, and the result is immediate. The problem can also be solved by inversion centered at W .

8. For simplicity, we shall write n instead of 1992. Lemma. There exists a tangent n-gon A 1 A 2 ...A n with sides A 1 A 2 =a 1 ,A 2 A 3 =

a 2 ,...,A n A 1 =a n if and only if the system x 1 +x 2 =a 1 ,x 2 +x 3 =a 2 ,,...,x n +x 1 =a n

566 4 Solutions

has a solution (x 1 ,...,x n ) in positive reals.

Proof. Suppose that such an n-gon A 1 A 2 ...A n exists. Let the side A i A i +1 touch the inscribed circle at point P i (where A n +1 =A 1 ). Then x 1 =A 1 P n =A 1 P 1 , x 2 =A 2 P 1 =A 2 P 2 , ...,x n =A n P n −1 =A n P n is clearly a positive solution of the system (1). Now suppose that the system (1) has a positive real solution (x 1 ,...,x n ). Let us draw a polygonal line A 1 A 2 ...A n +1 touching a circle of radius r at points P 1 ,P 2 ,...,P n respectively such that A 1 P 1 =A n +1 P n =x 1 and

A i P i =A i P i −1 =x i for i = 2, . . . , n. Observe that

OA 1 = OA n +1 = x 2 1 +r 2

A A 1 n +1 ∠A 2 OA 3 + ··· + ∠A n OA n +1 =2·

and the function f (r) = ∠A 1 OA 2 +

r + ··· + arctan x n

(arctan x 1 ) is con-

A 2 O closed simple polygonal line if and

tinuous. Thus A 1 A 2 ...A n +1 is a

P 2 P n −1 only if f (r) = 360 ◦ . But such an r

A n −1 exists, since f (r) → 0 when r → ∞,

and f (r) → ∞ when r → 0. This proves the second direction of the lemma.

For n = 4k, the system (1) is solvable in positive reals if a i = i for i ≡ 1,2 (mod 4), a i = i + 1 for i ≡ 3 and a i = i − 1 for i ≡ 0 (mod 4). Indeed, one solution is given by x i = 1/2 for i ≡ 1, x i = 3/2 for i ≡ 3 and x i = i − 3/2 for i ≡ 0,2 (mod 4).

Remark. For n = 4k + 2 there is no such n-gon. In fact, solvability of the system (1) implies a 1 +a 3 + ··· = a 2 +a 4 + ···, while in the case n = 4k + 2 the sum

a 1 +a 2 + ··· + a n is odd. √

9. Since the equation x 3 − x − c = 0 has only one real root for every c > 2/(3 3 ), α is the unique real root of x 3 − x − 33 1992 = 0. Hence f n ( α )=f( α )= α .

Remark. Consider any irreducible polynomial g (x) in the place of x 3 −x−

33 1992 . The problem amounts to proving that if α and f ( α ) are roots of g, then any f (n) ( α ) is also a root of g. In fact, since g( f (x)) vanishes at x = α , it must be divisible by the minimal polynomial of α , that is, g (x). It follows by induction that g (f (n)

(x)) is divisible by g(x) for all n ∈ N, and hence g( f (n) ( α )) = 0.

10. Let us set S (x) = {(y,z) | (x,y,z) ∈ V }, S y (x) = {z | (x,z) ∈ S y } and S z (x) = {y | (x, y) ∈ S z }. Clearly S(x) ⊂ S x and S (x) ⊂ S y (x) × S z (x). It follows that

|V | = ∑ |S(x)| ≤ ∑ |S x ||S y (x)||S z (x)|

= |S x | ∑ |S y (x)||S z (x)|.

4.33 Shortlisted Problems 1992 567 Using the Cauchy–Schwarz inequality we also get

∑ |S y (x)||S z (x)| ≤ ∑ |S y (x)| ∑ |S z (x)| = |S y ||S z |.

Now (1) and (2) together yield |V | ≤ |S x ||S y ||S z |.

11. Let I be the incenter of △ABC. Since 90 ◦ + α /2 = ∠BIC = ∠DIE = 138 ◦ , we obtain that ∠A = 96 ◦ .

AE

B E ′ D ′ C Let D ′ and E ′

be the points symmetric to D and E with respect to CE and BD re- spectively, and let S be the intersection point of ED ′ and BD. Then ∠BDE ′ = 24 ◦ and ∠D ′ DE ′ = ∠D ′ DE − ∠E ′ DE = 24 ◦ ,which means that DE ′ bisects the angle SDD ′ . Moreover, ∠E ′ SB = ∠ESB = ∠EDS + ∠DES = 60 ◦ and hence SE ′ bisects the angle D ′ SB . It follows that E ′ is the excenter of △D ′ DS and

consequently ∠D ′ DC = ∠DD ′ C = ∠SD ′ E ′ = (180 ◦ − 72 ◦ )/2 = 54 ◦ . Finally, ∠C = 180 ◦ − 2 · 54 ◦ = 72 ◦ and ∠B = 12 ◦ .

12. Let us set deg f = n and deg g = m. We shall prove the result by induction on n. If n < m, then deg x [ f (x) − f (y)] < deg x [g(x) − g(y)], which implies that f (x) −

f (y) = 0, i.e., that f is constant. The statement trivially holds. Assume now that n ≥ m. Transition to f 1 (x) = f (x) − f (0) and g 1 (x) = g(x) −

g (0) allows us to suppose that f (0) = g(0) = 0. Then the given condition for y = 0 gives us f (x) = f 1 (x)g(x), where f 1 (x) = a(x, 0) and deg f 1 = n − m. We now have

a (x, y)(g(x) − g(y)) = f (x) − f (y) = f 1 (x)g(x) − f 1 (y)g(y) =[f 1 (x) − f 1 (y)]g(x) + f 1 (y)[g(x) − g(y)].

Since g (x) is relatively prime to g(x) − g(y), it follows that f 1 (x) − f 1 (y) =

b (x, y)(g(x) − g(y)) for some polynomial b(x,y). By the induction hypothe- sis there exists a polynomial h 1 such that f 1 (x) = h 1 (g(x)) and consequently

f (x) = g(x) · h 1 (g(x)) = h(g(x)) for h(t) = th 1 (t). Thus the induction is com- plete.

13. Let us define

(pqr − 1)

F (p, q, r) = (p − 1)(q − 1)(r − 1)

p −1 q −1 r −1

+ . (p − 1)(q − 1) (q − 1)(r − 1) (r − 1)(p − 1)

568 4 Solutions Obviously F is a decreasing function of p , q, r. Suppose that 1 < p < q < r are

integers for which F (p, q, r) is an integer. Observe that p, q, r are either all even or all odd. Indeed, if for example p is odd and q is even, then pqr − 1 is odd while (p − 1)(q − 1)(r − 1) is even, which is impossible. Also, if p,q,r are even then F (p, q, r) is odd. If p ≥ 4, then 1 < F(p,q,r) ≤ F(4,6,8) = 191/105 < 2, which is impossible. Hence p ≤ 3. Let p = 2. Then q, r are even and 1 < F(2, q, r) ≤ F(2,4,6) = 47/15 < 4. There- fore F (2, q, r) = 3. This equality reduces to (q − 3)(r − 3) = 5, with the unique solution q = 4, r = 8. Let p = 3. Then q, r are odd and 1 < F(3, q, r) ≤ F(3,5,7) = 104/48 < 3. There- fore F (3, q, r) = 2. This equality reduces to (q − 4)(r − 4) = 11, which leads to q = 5, r = 15. Hence the only solutions (p, q, r) of the problem are (2, 4, 8) and (3, 5, 15).

14. We see that x 1 =2 0 . Suppose that for some m , r ∈ N we have x m =2 r . Then inductively x m +i =2 r −i (2i + 1) for i = 1, 2, . . . , r and x m +r+1 =2 r +1 . Since every natural number can be uniquely represented as the product of an odd number and

a power of two, we conclude that every natural number occurs in our sequence exactly once. Moreover, it follows that 2k −1=x k (k+1)/2 . Thus x n = 1992 = 2 3 · 249 implies that x n +3 = 255 = 2 · 128 − 1 = x 128 ·129/2 =x 8256 . Hence n = 8253.

15. The result follows from the following lemma by taking n = 1992 ·1993 2 and M = {d,2d,...,1992d}. Lemma. For every n ∈ N there exists a natural number d such that all the

numbers d , 2d, . . . , nd are of the form m k (m , k ∈ N, k ≥ 2). Proof. Let p 1 ,p 2 ,...,p n

be distinct prime numbers. We shall find d in the form

d =2 α 2 3 α 3 ···n α n , where α i ≥ 0 are integers such that kd is a perfect p k th power. It is sufficient to find α i ,i = 2, 3, . . . , n, such that α i ≡ 0 (mod p j ) if i 6= j and α i ≡ −1 (mod p j ) if i = j. But the existence of such α i ’s is an immediate consequence of the Chinese remainder theorem.

16. Observe that x 4 +x 3 +x 2 + x + 1 = (x 2 + 3x + 1) 2 − 5x(x + 1) 2 . Thus for x =5 25 we have

N =x 4 +x 3 +x 2 +x+1

= (x 2 + 3x + 1 − 5 13 (x + 1))(x 2 + 3x + 1 + 5 13 (x + 1)) = A · B. Clearly, both A and B are positive integers greater than 1.

17. (a) Let n =∑ k i =1 2 a i , so that α (n) = k. Then

n 2 = ∑ 2 2a i + ∑ 2 a i +a j +1

i <j

has at most k (k+1) + k

2 = k 2 binary ones.

(b) The above inequality is an equality for all numbers n k =2 k .

4.33 Shortlisted Problems 1992 569 (c) Put n m

m =2 2 m −1 −∑ m =1 2 2 −2 j j , where m > 1. It is easy to see that α (n m )=

2 m − m. On the other hand, squaring and simplifying yields n 2

m =1+ ∑ <j 2 2 +1−2 −2 . Hence α (n i 2 m )=1+ m (m+1) 2 and thus

m +1 ij

α (n 2 m ) 2 + m(m + 1)

→ 0 as m → ∞.

α (n m )

2 (2 m

− m)

Solution to the alternative parts.

(d) Let n i +2 j +1 =∑ 2 2 +1

i =1 2 . Then n =∑ i =1 2 +∑ i <j 2 2 has exactly

k (k+1)

binary ones, and therefore α(n 2 α(n) ) = 2k k (k+1) → ∞.

(e) Consider the sequence n i constructed in part (c). Let θ > 1 be a con- stant to be chosen later, and let N i =2 m i n i − 1 where m i > α (n i ) is such that m i / α (n i )→ θ as i → ∞. Then α (N i )= α (n i )+m i − 1, whereas

N 2 =2 2m i i n 2 i −2 m i +1 n i + 1 and α (N 2 i )= α (n 2 i )− α (n i )+m i . It follows that

α (N 2 α

(n 2 i )+( − 1) α (n i ) θ −1

lim

= lim

i →∞ α (N i ) i →∞

(1 + θ ) α (n i )

θ +1 which is equal to γ ∈ [0,1] for θ 1 = +γ 1 −γ (for γ = 1 we set m i / α (n i ) → ∞). (f) Let be given a sequence (n i ) ∞ i =1 with α (n 2 i )/ α (n i )→ γ . Taking m i > α (n i )

and N i =2 m i n i + 1 we easily find that α (N i )= α (n i ) + 1 and α (N 2 i )= α (n 2 i )+ α (n i )+ 1. Hence α (N 2 i )/ α (N i )= γ + 1. Continuing this procedure we can construct a sequence t i such that α (t 2 i )/ α (t i )= γ +k for an arbitrary k ∈ N.

18. Let us define inductively f 1 (x) = f (x) = 1 x n +1 and f (x) = f ( f n −1 (x)), and let

g (x) = x + f (x) + f n 2 (x) + ··· + f n (x). We shall prove first the following state- ment. Lemma. The function g n (x) is strictly increasing on [0, 1], and g n −1 (1) =

F 1 /F 2 +F 2 /F 3 + ··· + F n /F n +1 . Proof. Since f

y (x) − f (y) = −x (1+x)(1+y) is smaller in absolute value than x − y, it follows that x > y implies f 2k (x) > f 2k (y) and f 2k +1 (x) < f 2k +1 (y), and moreover that for every integer k ≥ 0,

[f 2k (x) − f 2k (y)] + [ f 2k +1 2k (x) − f +1 (y)] > 0. Hence if x > y, we have g n (x) − g n (y) = (x − y) + [ f (x) − f (y)] + ··· +

[f n (x) − f n (y)] > 0, which yields the first part of the lemma. The second part follows by simple induction, since f k (1) = F k +1 /F k +2 .

If some x i = 0 and consequently x j = 0 for all j ≥ i, then the problem reduces to the problem with i − 1 instead of n. Thus we may assume that all x 1 ,...,x n are different from 0. If we write a i = [1/x i ], then x i = 1 a i +x i +1 . Thus we can regard x i

as functions of x n depending on a 1 ,...,a n −1 .

Suppose that x n ,a n −1 ,...,a 3 ,a 2 are fixed. Then x 2 ,x 3 ,...,x n are all fixed, and x 1 = 1 a 1 +x 2 is maximal when a 1 = 1. Hence the sum S = x 1 +x 2 + ··· + x n is maximized for a 1 = 1.

570 4 Solutions We shall show by induction on i that S is maximized for a 1 =a 2 = ··· = a i = 1.

In fact, assuming that the statement holds for i − 1 and thus a 1 = ··· = a i −1 = 1, having x n ,a n −1 ,...,a i +1 fixed we have that x n ,...,x i +1 are also fixed, and that x i −1 = f (x i ), . . . , x 1 =f i −1 (x i ). Hence by the lemma, S = g i −1 (x i )+x i +1 + ··· + x n is maximal when x i = 1 a i +x i +1 is maximal, that is, for a i = 1. Thus the induction is complete. It follows that x 1 + ··· + x n is maximal when a 1 = ··· = a n −1 = 1, so that x 1 + ··· + x n =g n −1 (x 1 ). By the lemma, the latter does not exceed g n −1 (1). This completes the proof.

Remark. The upper bound is the best possible, because it is approached by taking x

1 n close to 1 and inductively (in reverse) defining x i −1 = 1 1 +x i = a i +x i .

√ + 2(1 − 2 )x √ 2 √ √ +3+

19. Observe that f (x) = (x 4 + 2x 2 + 3) 2 − 8(x 2 − 1) 2 = [x 4

2 2 ][x 4 + 2(1 + 2 )x 2 +3−2 2 ]. Now it is easy to find that the roots of f are √ 4 √ 4

x 1 ,2,3,4 = ±i i 2 ±1

and x 5 ,6,7,8 = ±i

2 ±1 . In other words, x k = α i + β j , where α 2 i = −1 and β 4 j = 2.

We claim that any root of f can be obtained from any other using rational func- tions. In fact, we have

x 3 =− α i −3 β j +3 α i β 2 j + β 3 j ,

x 5 = 11 α i +7 β j − 10 α i β 2 j − 10 β 3 j x 7 = −71 α i − 49 β j + 35 α i β 2 j 3 + 37 β j ,

from which we easily obtain that α = 24 −1 (127x + 5x 3 i + 19x 5 + 5x 7 ), β j = 24 −1 (151x + 5x 3 + 19x 5 + 5x 7 ). Since all other values of α and β can be obtained as rational functions of α i and

β j , it follows that all the roots x l are rational functions of a particular root x k .

We now note that if x 1 is an integer such that f (x 1 ) is divisible by p, then p > 3 and x 1 ∈Z p is a root of the polynomial f . By the previous consideration, all remaining roots x 2 ,...,x 8 of f over the field Z p are rational functions of x 1 , since 24 is invertible in Z p . Then f (x) factors as

f (x) = (x − x 1 )(x − x 2 ) ··· (x − x 8 ), and the result follows.

20. Denote by U the point of tangency of the circle C and the line l. Let X and U ′ be the points symmetric to U with respect to S and M respectively; these points do not depend on the choice of P. Also, let C ′

be the excircle

4.33 Shortlisted Problems 1992 571 of △PQR corresponding to P, S ′ the

center of C ′ , and W ,W ′ the points of X tangency of C and C ′ with the line PQ respectively. Obviously,

S △W ′ S ′ P . Since SX kS ′ U ′ and SX :

△W SP ∼ W

′ W ′ = SP : S ′ = SW : S R P , we de- duce that ∆ SX P ∼ ∆ S ′ U ′ P , and conse-

quently that P lies on the line XU ′ . On the other hand, it is easy to show that

each point P of the ray U ′ X over X sat- W ′

isfies the required condition. Thus the desired locus is the extension of U ′

S X ′ over X .

21. (a) Representing n 2 as a sum of n 2 − 13 squares is equivalent to representing

13 as a sum of numbers of the form x 2 − 1, x ∈ N, such as 0,3,8,15,.... But it is easy to check that this is impossible, and hence s (n) ≤ n 2 − 14. (b) Let us prove that s (13) = 13 2 − 14 = 155. Observe that

Given any representation of n 2 as a sum of m squares one of which is even, we can construct a representation as a sum of m + 3 squares by dividing the even square into four equal squares. Thus the first equality enables us to construct representations with 5 , 8, 11, . . ., 155 squares, the second to con- struct ones with 7 , 10, 13, . . ., 154 squares, and the third with 9, 12, . . . , 153

squares. It remains only to represent 13 2 as a sum of k = 2, 3, 4, 6 squares. This can be done as follows:

= 11 2 +4 2 +4 2 +4 2 = 12 2 +3 2 +2 2 +2 2 +2 2 +2 2 . (c) We shall prove that whenever s (n) = n 2 − 14 for some n ≥ 13, it also holds

that s (2n) = (2n) 2 − 14. This will imply that s(n) = n 2 − 14 for any n =

2 t · 13. If n 2 =x 2 1 + ··· + x 2 r , then we have

(2n) 2 = (2x 1 ) 2 + ··· + (2x r ) 2 . Replacing (2x i ) 2 with x 2 i +x 2 i +x 2 i +x 2 i as long as it is possible we can

obtain representations of (2n) 2 consisting of r , r + 3, . . . , 4r squares. This gives representations of (2n) 2 into k squares for any k

≤ 4n 2 − 62. Further, we observe that each number m ≥ 14 can be written as a sum of k ≥ m

numbers of the form x 2 − 1, x ∈ N, which is easy to verify. Therefore if

572 4 Solutions k ≤ 4n 2 − 14, it follows that 4n 2 − k is a sum of k numbers of the form

x 2 − 1 (since k ≥ 4n 2 − k ≥ 14), and consequently 4n 2 is a sum of k squares. Remark. One can find exactly the value of s (n) for each n: 

 1 , if n has no prime divisor congruent to 1 mod 4; s (n) = 2 ,

if n is of the form 5 ·2 k , k a positive integer;  n 2 − 14, otherwise.

4.34 Shortlisted Problems 1993 573

Dokumen yang terkait

Analisis Komparasi Internet Financial Local Government Reporting Pada Website Resmi Kabupaten dan Kota di Jawa Timur The Comparison Analysis of Internet Financial Local Government Reporting on Official Website of Regency and City in East Java

19 819 7

ANTARA IDEALISME DAN KENYATAAN: KEBIJAKAN PENDIDIKAN TIONGHOA PERANAKAN DI SURABAYA PADA MASA PENDUDUKAN JEPANG TAHUN 1942-1945 Between Idealism and Reality: Education Policy of Chinese in Surabaya in the Japanese Era at 1942-1945)

1 29 9

Implementasi Prinsip-Prinsip Good Corporate Governance pada PT. Mitra Tani Dua Tujuh (The Implementation of the Principles of Good Coporate Governance in Mitra Tani Dua Tujuh_

0 45 8

Improving the Eighth Year Students' Tense Achievement and Active Participation by Giving Positive Reinforcement at SMPN 1 Silo in the 2013/2014 Academic Year

7 202 3

An Analysis of illocutionary acts in Sherlock Holmes movie

27 148 96

The Effectiveness of Computer-Assisted Language Learning in Teaching Past Tense to the Tenth Grade Students of SMAN 5 Tangerang Selatan

4 116 138

The correlation between listening skill and pronunciation accuracy : a case study in the firt year of smk vocation higt school pupita bangsa ciputat school year 2005-2006

9 128 37

Existentialism of Jack in David Fincher’s Fight Club Film

5 71 55

Phase response analysis during in vivo l 001

2 30 2

1 BAB I PENDAHULUAN A. Latar Belakang - Penerapan model Problem Based Instruction (PBI) terhadap pemahaman konsep dan hasil belajar siswa pokok bahasan tekanan Kelas VIII Semester II di SMPN Palangka Raya Tahun Ajaran 2015/2016 - Digital Library IAIN Pala

0 3 80