CMS/SMC
Société mathématique du Canada
www.smc.math.ca
Société mathématique du Canada
  location: 
       
Problem Set 6

31.
Find all continuous functions f: R R such that
f(x + f(y)) = f(x) + y
for every x, y R.

View solution

32.
Let n 3 be a natural number. Prove that
1989 |nnnn - nnn   .

View solution

33.
Let n 3 be a positive integer. Are there n positive integers a1, a2, , an not all the same such that for each i with 3 i n we have
ai + Si = (ai, Si) + [ai, Si]  .
where Si = a1 + a2 + + an, and where (·, ·) and [ ·, ·] represent the greatest common divisor and least common multiple respectively?

View solution

34.
Find all integers p for which there are rational numbers a and b such that the polynomial x5 - px - 1 has at least one root in common with the polynomial x2 - ax - b.
View solution
35.
Let P be an arbitrary interior point of an equilateral triangle ABC. Prove that
|PAB - PAC | |PBC -PCB |  .
(Note: This is Problem 2255 in CM+MM, September, 1997.)

View solution

36.
Let n be a positive integer and x > 0. Prove that
(1 + x)n+1 (n+1)n+1
nn
x  .
(Note: This is Problem 2260 in CM+MM, September, 1997.)

View solution



Solutions to Problem Set 6

31.

31.
First solution. Setting (x, y) = (t, 0) yields f(t + f(0)) = f(t) for all real t. Setting (x, y) = (0, t) yields f(f(t)) = f(0) + t for all real t. Hence f(f(f(t))) = f(t) for all real t, i.e., f(f(z)) = z for each z in the image of f. Let (x, y) = (f(t), -f(0)). Then
f(f(t) + f(-f(0))) = f(f(t)) - f(0) = f(0) + t - f(0) = t
so that the image of f contains every real and so f(f(t)) t for all real t.

Taking (x, y) = (u, f(v)) yields
f(u + v) = f(u) + f(v)
since v = f(f(v)) for all real u and v. In particular, f(0) = 2f(0), so f(0) = 0 and 0 = f(-t + t) = f(-t) + f(t). By induction, it can be shown that for each integer n and each real t, f(nt) = nf(t). In particular, for each rational r/s, f(r/s) = rf(1/s) = (r/s)f(1). Since f is continuous, f(t) = f(t·1) = tf(1) for all real t. Let c = f(1). Then 1 = f(f(1)) = f(c) = cf(1) = c2 so that c = 1. Hence f(t) t or f(t) -t. Checking reveals that both these solutions work. (For f(t) -t, f(x + f(y)) = -x - f(y) = f(x) + y, as required.)

31.
Second solution. Taking (x, y) = (0, 0), we find that
f(0) = 0 + f(0) f(f(0)) = f(0 + f(0)) = f(0) + f(f(0)) f(0) = 0 .
Taking (x, y) = (0, t) yields f(f(t)) = t for all real t. Hence f(x + y) = f(x + f(f(y)) = f(x) + f(y) for all x, y, and we can complete the solution as in the First Solution.

31.
Third solution. Taking (x, y) = (0, 0) yields f(f(0)) = f(0), whence f(f(f(0))) = f(f(0)) = f(0). Taking (x, y) = (0, f(0)) yields f(f(f(0))) = 2f(0). Hence 2f(0) = f(0) so that f(0) = 0. Taking x = 0 yields f(f(y)) = y for each y. We can complete the solution as in the Second Solution.

32.

32.
First solution. Let N = nnnn - nnn. Since 1989 = 32 ·13 ·17,
N 0   (mod  1989) N 0   (mod    9, 13   &   17) .
We require the following facts:

(i) xu 0 (mod 9) whenever u 2 and x 0 (mod 3).
(ii) x6 1 (mod 9) whenever x \not 0 (mod 3).
(iii) xu 0 (mod 13) whenever x 0 (mod 13).
(iv) x12 1 (mod 13) whenever x \not 0 (mod 13), by Fermat's Little Theorem.
(v) xu 0 (mod 17) whenever x 0 (mod 17).
(vi) x16 1 (mod 17) whenever x \not 0 (mod 17), by FLT.
(vii) x4 1 (mod 16) whenever x = 2y + 1 is odd. (For, (2y+1)4 = 16y3(y+2) + 8y(3y+1) + 1 1 (mod 16).)
Note that
N = nnn

n(nnn - nn) - 1

= nnn

nnn (nnn - n - 1) - 1

 .

Modulo 17. If n 0 (mod 17), then nnn 0, and so N 0 (mod 17).
If n is even, n 4, then nn 0 (mod 16), so that
nnn(nnn - n - 1) 1(nnn - n - 1) 1
so N 0 (mod 17).

Suppose that n is odd. Then nn n (mod 4)
nn - n = 4r  for  some  r N
nnn - n = n4r 1  (mod 16)
nnn - n - 1 0  (mod 16)
nnn(nnn - n - 1) 1  (mod 17
N 0  (mod  17)  .
Hence N 0 (mod 17) for all n 3.

Modulo 13. If n 0 (mod 13), then nnn 0 and N 0 (mod 13).
Suppose that n is even. Then nn 0 (mod 4), so that nnn - nn 0 (mod 4). Suppose that n is odd. Then nnn - n - 1 0 (mod 16) and so nnn - nn 0 (mod 4).
If n 0 (mod 3), then nn 0 so nn(nnn - n - 1) 0 (mod 3). If n 1 (mod 3), then nnn - n 1 so nn(nnn-n - 1) 0 (mod 3). If n 2 (mod 3), then, as nn - n is always even, nnn-n 1 so nn(nnn-n - 1) 0 (mod 3). Hence, for all n, nnn - nn 0 (mod 3).
It follows that nnn - nn 0 (mod 12) for all values of n. Hence, when n is not a multiple of 13, n(nnn - n) 1 so N 0 (mod 13).
Modulo 9. If n 0 (mod 3), then nnn 0 (mod 9), so N 0 (mod 9). Let n \not 0 (mod 9). Since nnn - nn is divisible by 12, it is divisible by 6, and so n(nnn - nn) 1 and N 0 (mod 9). Hence N 0 (mod 9) for all n.
The required result follows.

33.

33.
First solution. Letting bi = (ai, Si), we find that
[ai , Si] = ai Si
(ai, Si)
= ai Si
bi
 .
The given condition is equivalent to ai + Si = bi + (ai Si/bi), which is equivalent to
0 = bi2 - (ai + Si)bi + ai Si = (bi - ai)(bi - Si) .
We can achieve the condition by making ai = (ai, Si) and Si = [ai, Si]. Let a1 = a2 = 1, ai = 2i-2 for i 3. Then
Si = 1 + i

j = 2 
2j-2 = 1 + (2i-1 - 1) = 2i-1
(ai, Si) = 2i-2 ,   [ai, Si] = 2i-1
for i 3.

33.
Second solution. Let ai = 1, a2 = 2 and ai = 3 ·2i-3 for i 3. Then
Si = 1 + 2 + 3 i-3

j = 0 
2j = 1 + 2 + 3(2i-2 - 1) = 3 ·2i-2
(ai, Si) = 3 ·2i-3 ,  [ai, Si] = 3 ·2i-2
for i 3.

33.
Third solution. [K. Purbhoo] Choose a1 at will, and let ai = Si-1 for i 2. Then Si = Si-1 + ai = 2ai, ai + Si = 3ai, (ai, Si) = ai and [ai, Si] = 2ai for i 2.
33.
Fourth solution. [K. Yeats] Let a1 = 1, a2 = 3 and an = 2n-1 for n 3. Then Sn = 2n for n 2 and, for each i with 3 i n, (ai, Si) = 2i-1 and [ai, Si] = 2i.

34.

34.
First solution. A common rational root must be 1 or -1, in which case the respective values of p are 0 and 2. These roots are shared with x2 - 1 (a = 0, b = 1). Suppose that there is a common nonrational root r. Then r2 = ar + b so that
r4 = a2 r2 + 2abr + b2 = (a3 + 2ab)r + (a2 b + b2)
whence
pr + 1 = r5 = (a3 + 2ab)r2 + (a2 b + b2)r = (a4 + 3a2 b + b2)r + b(a3 + 2ab) .
Since r is not rational,
p = a4 + 3a2 b + b2    and     1 = b(a3 + 2ab) .
Hence 2ap - 1 = 2a5 + 5a3 b, so that 5a3 b = 2ap - 2a5 - 1. Thus
1 =

-2a5 + 2ap - 1
5a3




a3 + -4a5 + 4pa - 2
5a2


from which we obtain that
25a5 = (-2a5 + 2pa - 1)(a5 + 4pa - 2) .
This simplifies to
a10 + 3pa6 + 11a5 - 4p2 a2 + 4pa - 1 = 0 .
This is a polynomial equation for a with integer coefficients. In order that a be rational, it must be an integer dividing 1. Hence a = 1. If a = 1, then 1 = b(1 + 2b), whence b = 1/2 or b = -1. The first value does not give an integer value of p, but the second yields p = -1. In this case, the two polynomials are x2 - x + 1 and x5 + x - 1 = (x2 - x + 1)(x3 + x2 - 1) with common roots -w and -w2, where w is an imaginary cube root of unity. If a = -1, we get the equation 2b2 + b + 1 = 0 which is satisfied by no rational value of b. Thus the possibilities are p = -1, 0, 2.

34.
Second solution. The rational root case can be handled as in the first solution. By long division, we find that
x5 - px - 1 = (x2 - ax - b)[x3 + ax2 + (a2 + b)x + (a3 + 2ab)]        
         + (a4 + 3a2b + b2 - p)x + (a3b + 2ab2 - 1) .
If r is a nonrational root of the quadratic and the quintic, then its quadratic conjugate is also. So the quadratic must divide the quintic and both coefficients of the foregoing division must vanish. Hence p = a4 + 3a2b + b2 and 1 = b(a3 + 2ab), and we can proceed as in the first solution.


35.

35.
First solution. The result is clear if P is on the bisector of the angle at A, since both sides of the inequality are 0.
Wolog, let P be closer to AB than AC, and let Q be the image of P under reflection in the bisector of the angle A. Then
PAQ = PAC - QAC = PAC - PAB
and
PCQ = QCB - PCB = PBC - PCB .
Thus, it is required to show that PAQ PCQ.

Produce PQ to meet AB in R and AC in S. Consider the reflection with axis RS. The circumcircle C of DARS is carried to a circle C with chord RS. Since RCS < 60 = RAS and the angle subtended at the major arc of C by RS is 60, the point C must lie outside of C. The circumcircle D of DAPQ is carried by to a circle D with chord PQ. Since D is contained in C, D must be contained in C, so C must lie outside of D. Hence PQC must be less than the angle subtended at the major arc of D by PQ, and this angle is equal to PAQ. The result follows.

36.

36.
First solution. By the Arithmetic-Geometric Means Inequality, we have that
1+x
n+1
= n(1/n) + x
n+1




1
n


n

 
x

[1/(n+1)]

 
so that
(1 + x)n+1
(n+1)n+1
x
nn
and the result follows.

36.
Second solution (by calculus). Let
f(x) = nn (1 + x)n+1 - (n+1)n+1x    for  x > 0 .
Then
f(x) = (n+1)[nn (1 + x)n - (n+1)n] = (n+1)nn[(1 + x)n - (1 + 1
n
)n ]
so that f(x) < 0 for 0 < x < 1/n and f(x) > 0 for 1/n < x. Thus f(x) attains its minimum value 0 when x = 1/n and so f(x) 0 when x > 0. The result follows.

36.
Third solution. Let u = nx - 1 so that x = (1 + u)/n. Then
(1 + x)n+1 - (n+1)n+1
nn
x
= (1 + 1
n
+ u
n
)n+1 - (1 + 1
n
)n+1(1 + u)
= (1 + 1
n
)n+1 + (n+1)(1 + 1
n
)n u
n
+

n+1
2


(1 + 1
n
)n-1( u
n
)2
       +

n+1
3


(1 + 1
n
)n-2( u
n
)3 + - (1 + 1
n
)n+1(1 + u)
=

n+1
2


(1 + 1
n
)n-1( u
n
)2+

n+1
3


(1 + 1
n
)n-2( u
n
)3 +  .
This is clearly nonnegative when u 0. Suppose that -1 < u < 0. For 1 k n/2, we have that


n+1
2k


(1 + 1
n
)n-2k+1( u
n
)2k+

n+1
2k+1


(1 + 1
n
)n-2k ( u
n
)2k+1                
= (n+1)!(1 + 1/n)n-2k
(2k+1)!(n+1-2k)!


u
n


2k

 
[(2k+1)(1 + 1
n
) +(n+1 - 2k)( u
n
) ]  .
This will be nonnegative if and only if the quantity in square brackets is nonnegative. Since u > -1, this quantity exceeds
(2k+1)(1 + 1
n
) - (n + 1 - 2k)( 1
n
) =

n+1
n


(2k + 1 - 1) - 2k
n
= 2k > 0 .
Thus, each consecutive pair of terms in the sequence


n+1
2


(1 + 1
n
)n-1( u
n
)2+

n+1
3


(1 + 1
n
)n-2( u
n
)3 +
has a positive sum and so the desired result follows.

© Société mathématique du Canada, 2014 : http://www.smc.math.ca/