Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $a_1,a_2,\ldots ,a_n$, be positive real numbers and S = $a_1+a_2+a_3+\cdots+a_n$. Use the Cauchy-Schwarz inequality to prove that

$$\sum_{i=1}^{i=n}\frac{a_i}{S-a_i}\geqslant \frac{n}{n-1}$$ and $$\sum_{i=1}^{i=n}\frac{S-a_i}{a_i}\geqslant n(n-1)$$

Hint: Apply the inequality to $\sum \frac{S}{a_i}$ and $\sum \frac{a_i}{S}$ for one case and $\sum a_i(S-a_i)$ and $\sum \frac{a_i}{S-a_i}$ for the other case.

share|improve this question
2  
Have you written down the version of Cauchy-Schwarz you want to use? If not, do it! –  martini Mar 27 '12 at 10:08
3  
I think the hint says clearly what you have to do. You only need to write the corresponding Cauchy Schwarz inequality and you are done. –  Beni Bogosel Mar 27 '12 at 10:17
    
Faisal, there is a "homework" tag, which might be a better fit than "education" if this is indeed homework. –  The Chaz 2.0 Mar 27 '12 at 13:15
    
@Community has a bug? \lodts should be \ldots. –  Jonas Teuwen Mar 27 '12 at 13:29
add comment

2 Answers

up vote 4 down vote accepted
+50

Cauchy Schwarz Inequality:

$$ \begin{align*} \left( \frac{1}{a_1}+\frac{1}{a_2}+\dots +\frac{1}{a_n}\right)\left( a_1+a_2+\dots+a_n \right) &\geq \left(\frac{1}{\sqrt{a_1}}\sqrt{a_1}+\frac{1}{\sqrt{a_2}}\sqrt{a_2}+\dots +\frac{1}{\sqrt{a_n}}\sqrt{a_n} \right)^2 \\\ &\geq \left( 1+1+\dots+1\right)^2 = n^2 \end{align*} $$

which when applied here $$ \begin{align*} \displaystyle\sum_{i=1}^n \frac{S}{a_i} &= \left( \frac{1}{a_1}+\frac{1}{a_2}+\dots +\frac{1}{a_n}\right)S \geq n^2\\ \end{align*} $$

Therefore $$ \begin{align*} \displaystyle\sum_{i=1}^n \frac{S-a_i}{a_i} &= \displaystyle\sum_{i=1}^n \frac{S}{a_i} -n\\ &\geq n^2 - n = n(n-1) \tag{1} \end{align*} $$

Similarly, re-writing the other summation

$$ \begin{align*} \displaystyle\sum_{i=1}^n \frac{a_i}{S-a_i} &= \frac{S-\displaystyle\sum_{j \neq 1}a_j}{\displaystyle\sum_{j \neq 1}a_j} + \frac{S-\displaystyle\sum_{j \neq 2}a_j}{\displaystyle\sum_{j \neq 2}a_j}+\dots +\frac{S-\displaystyle\sum_{j \neq n}a_j}{\displaystyle\sum_{j \neq n}a_j}\\ &= \displaystyle\sum_{i=1}^n \left(\frac{S}{\displaystyle\sum_{j \neq i}a_j}\right)-n \tag{2} \end{align*} $$

Also since

$$ \displaystyle\sum_{j \neq 1}a_j+\displaystyle\sum_{j \neq 2}a_j+\dots+\displaystyle\sum_{j \neq n}a_j = (n-1)S $$

By Cauchy-Schwarz Inequality $$ \begin{align*} \displaystyle\sum_{i=1}^n \left(\frac{1}{\displaystyle\sum_{j \neq i}a_j}\right)(n-1)S \geq n^2\\ \Longrightarrow \displaystyle\sum_{i=1}^n \left(\frac{1}{\displaystyle\sum_{j \neq i}a_j}\right) \geq \frac{n^2}{(n-1)S} \end{align*} $$

From $(2)$

$$ \begin{align*} \displaystyle\sum_{i=1}^n \frac{a_i}{S-a_i} &\geq \left(\frac{n^2}{n-1}\right) - n\\ &\geq \frac{n}{n-1} \end{align*} $$

share|improve this answer
add comment

The left sides are the values of $\sum f(p_i)$ and $\sum g(p_i)$ with $\sum p_i = 1$ being constant, where

$f(x) = \dfrac{x} {1-x}$ $g(x) = \dfrac 1 x - 1$, and $p_i = \dfrac {a_i} S $.

Both functions are convex for $x>0$, so that each sum of values is minimized when all arguments $p_i$ (and therefore all $a_i$) are equal.

Another line of argument that makes this obvious is the Rearrangement Inequality: the average of the sequence $x_i y_i$ is larger or equal to the product of the averages of $x_i$ and $y_i$ if the two sequences are in the same order. In the first inequality the sequences $a_i$ and $(S-a_i)^{-1}$ are in the same order, and in the second inequality, $(S-a_i)$ and $1/a_i$ are in the same order. The averages of $a_i$ and $(S-a_i)$ are $S/n$ and $(n-1)S/n$, which depend only on $S$. The other averages lead to the standard problem of minimizing a sum of reciprocals given the sum of the variables, which happens when the summands are equal.

share|improve this answer
    
@Peter , I asked a meta question about \Sigma vs \sum . –  zyx Apr 9 '12 at 21:17
    
Link? What is the issue? –  Pedro Tamaroff Apr 9 '12 at 21:18
    
@Peter, meta.math.stackexchange.com/questions/3932/… . No issue, a question. –  zyx Apr 9 '12 at 21:21
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.