Maximizing Reciprocals

Algebra Level 4

Let x 1 , x 2 , , x n x_1, x_2, \dots, x_n be positive reals satisfying x 1 x 2 x n = 1 x_1x_2 \cdots x_n = 1 .

Find the maximum value of 1 n 1 + x 1 + 1 n 1 + x 2 + + 1 n 1 + x n . \dfrac{1}{n-1+x_1} + \dfrac{1}{n-1+x_2} + \dots + \dfrac{1}{n-1+x_n} .


The answer is 1.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

1 solution

Sharky Kesa
Jan 22, 2016

This question was in Romania 1999. The answer is 1, which we shall prove using Olympiad inequalities.

First, we prove a lemma:

The maximum of the sum occurs when n 1 n-1 of the x i x_i terms are equal. Consider f ( y ) = 1 k + e y f(y)=\frac{1}{k+e^y} for an arbitrary nonnegative constant k k . We have f ( y ) = e y ( k + e y ) 2 f'(y)=\frac {-e^y}{(k+e^y)^2} and f ( y ) = e y ( e y k ) ( k + e y ) 3 f''(y)=\frac{e^y (e^y-k)}{(k+e^y)^3} . Evidently, f ( y ) 0 e y k f''(y) \geq 0 \iff e^y \geq k . Hence, f ( y ) f(y) has a single inflexion point where y = ln ( k ) y=\ln(k) ., where f ( y ) f(y) is convex over the interval ( ln ( k ) , ) (\ln(k),\infty) . Now, we employ the substitution y i = ln ( x i ) y_i = \ln(x_i) so that y 1 + + y n = 0 y_1 + \ldots + y_n = 0 and

i = 1 n 1 n 1 + x i = i = 1 n f ( y i ) \displaystyle \sum_{i=1}^n \dfrac {1}{n-1+x_i} = \displaystyle \sum_{i=1}^n f(y_i)

We take k = n 1 k=n-1 and write k 0 = ln ( n 1 ) k_0 = \ln(n-1) . Suppose that y 1 y m k 0 y m + 1 x n y_1 \geq \ldots \geq y_m \geq k_0 \geq y_{m+1} \geq \ldots \geq x_n for some positive m m . Then, by Majorisation,

f ( y 1 ) + + f ( y m ) ( m 1 ) f ( k 0 ) + f ( y 1 + + y m ( m 1 ) k 0 ) f(y_1) + \ldots + f(y_m) \leq (m-1)f(k_0) + f(y_1 + \ldots + y_m-(m-1)k_0)

But then, also by Majorisation,

( m 1 ) f ( k 0 ) + f ( y m + 1 ) + + f ( y n ) ( n 1 ) f ( ( m 1 ) k 0 + y m + 1 + + y n n 1 (m-1)f(k_0)+f(y_{m+1})+\ldots+f(y_n) \leq (n-1)f(\dfrac{(m-1)k_0+y_{m+1}+\ldots+y_n}{n-1}

Otherwise, all of the y i y_i are less than k 0 k_0 . In that case, we may directly apply Majorisation to equate n 1 n-1 of the y i y_i whilst increasing the sum in question. Hence, the lemma is valid.

Applying this lemma, it would suffice to show

k k + x + 1 k + 1 x k 1 \dfrac {k}{k+x}+\dfrac {1}{k+\frac{1}{x^k}} \leq 1

Clearing the denominators,

( k 2 + k x k ) + ( k + x ) k 2 + k ( x + 1 x k ) + x 1 k \left (k^2 + \dfrac {k}{x^k} \right ) + (k + x) \leq k^2+ k \left (x+\dfrac{1}{x^k} \right ) + x^{1-k}

x k + x + k x 1 k -xk+x+k \leq x^{1-k}

But now this is evident. We have Bernoulli's inequality, since x 1 k = ( 1 + ( x 1 ) ) 1 k 1 + ( x 1 ) ( 1 k ) = x + k x k x^{1-k} = (1+(x-1))^{1-k} \geq 1+(x-1)(1-k) = x+k-xk . Equality holds only where x = 1 x=1 or n = 2 n=2 .

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...