Let $x_1,x_2,x_3 \ldots x_{2014}$ be positive real numbers such that $\sum_{j=1}^{2014} x_j=1$. Determine with proof the smallest constant $K$ such that \[K\sum_{j=1}^{2014}\frac{x_j^2}{1-x_j} \ge 1\]
Problem
Source:
Tags: inequalities
07.12.2014 14:52
Using titu's lemma , we get that $ \frac{1}{2014 -1} .K \ge 1 \Longrightarrow K \ge 2013 $
07.12.2014 19:28
it is too trivial assuming , $\sum x_i=1$ the sequences $x_i$'s and $(1-x_i)^{-1}$'s are sorted in a same way so applying Chebyshev's Ineq, $P=\sum \frac {x_i^2}{1-x_i} \geq \frac 1n (\sum x_i^2).(\sum \frac {1}{1-x_i})$ $ \implies P\geq \frac {1}{n^2} (\sum x_i)^2. \frac {n^2}{n-\sum x_i}$ by repeated application of CS implies , $P\geq \frac {1}{n-1}$ plugg $n=2014 \cdots $ Also jenesen works here too !
07.12.2014 20:02
Trivial application of Cauchy-Schwarz. Basically we want to find minimum value of $S=\sum\dfrac{x_j^2}{1-x_j}$ subject to $\sum x_j=1$. So $(\sum\dfrac{x_j^2}{1-x_j})(\sum {1-x_j})\geq (\sum{x_j})^2=1\iff S\geq\dfrac{1}{2013}\iff K=2013$.
08.12.2014 12:26
During the test,I did this in two ways.One using A.M-G.M and the other uses Cauchy.I shall present the A.M-G.M one. Note that $\frac{1}{1-x_j}-x_j=1+\frac{{x_j}^2}{1-x_j}$ Summing over all $j$ this yeilds $\sum_{j=1}^{2014}{\frac{1}{1-x_j}}-2015=\sum_{j=1}^{2014}{\frac{{x_j}^2}{1-x_j}}$. By A.M-G.M we get $(\prod_{i=1}^{2014}{(1-x_j)})^{\frac{1}{2014}} \le \frac{2013}{2014}$ Again using A.M-G.M and the last line we get $\sum_{j=1}^{2014}{\frac{1}{1-x_j}} \ge 2014{\frac{1}{\prod_{i=1}^{2014}{(1-x_j)}}}^{\frac{1}{2014}} \ge \frac{{2014}^2}{2013}$ Hence from the first equality we get $\sum_{j=1}^{2014}{\frac{{x_j}^2}{1-x_j} \ge \frac{{2014}^2}{2013}-2015=\frac{1}{2013}}$ Hence $K=2013$.Equality holds when every $x_i$ is equal.Hence $K$ cannot be made lesser.
08.12.2014 13:58
I think there is something wrong with the question. For example, consider any of the x's to be equal to 0.9. Then, x^2/1-x = .81/.1 = 8.1 the sum for all of the other x's should be positive, so when we take the final sum, it will be K(8.1 + something). In this case, K doesn't have to be greater than 1. Repeat this, and use 0.99, 0.999, etc. I believe we'll find that the value of K continuously decreases and tends to zero, and there is no minimum for k.
08.12.2014 14:06
That value for $K$ should work for all allowed values of those variables; you just exhibit some particular set of values, for which even a lesser $K$ will work ...
08.12.2014 14:10
I wrote the answer 2013, but they didn't specify that it has to work for every possible arrangement. It just said the smallest k such that the inequality CAN be satisfied. there is thus no bound on K.
08.12.2014 14:27
Just (typical) carelessness from the problem setters in wording the statement.
09.12.2014 04:48
By C-S , $\sum_{j=1}^{2014}\frac{x_j^2}{1-x_j} \ge\frac{(\sum_{j=1}^{2014}x_j)^2}{2014-\sum_{j=1}^{2014}x_j}=\frac{1}{2013}$, Equality holds when $x_1=x_2=\cdots =x_{2014}=\frac{1}{2014}$, hence $K_{min}=2013$ .
17.12.2014 20:06
$\Sigma_{j=1}^{2014}(1+x_j)$=2015 $\implies \Sigma_{j=1}^{2014}(\frac {1-x_j^2} {1-x_j})$=2015 $\implies \Sigma_{j=1}^{2014}(\frac {1} {1-x_j})$=2015 + $\Sigma_{j=1}^{2014}(\frac {x_j^2} {1-x_j})$ Now by AM-HM $\frac {\Sigma_{j=1}^{2014}(\frac {1} {1-x_j})} {2014} \ge {2014} {\Sigma_{j=1}^{2014}(1-x_j)}$ $\implies \Sigma_{j=1}^{2014} (\frac {1} {1-x_j}) \ge \frac {2014^2} {2013}$ $\implies$ 2015 + $\Sigma_{j=1}^{2014}(\frac {x_j^2} {1-x_j}) \ge \frac {2014^2} {2013}$ $\implies \Sigma_{j=1}^{2014}(\frac {x_j^2} {1-x_j}) \ge \frac {1} {2013}$ So K=2013
17.12.2014 20:27
From http://www.artofproblemsolving.com/Forum/viewtopic.php?f=51&t=585619 we can see that the answer is $\boxed{K=2013}$. Perhaps India copied my question!
26.01.2021 06:32
sorry for bumping the thread again but everybody is giving their different respnses but actually there is no need,simply apply titu and that should do it.
26.01.2021 08:29
Commander_Anta78 wrote: sorry for bumping the thread again but everybody is giving their different respnses but actually there is no need,simply apply titu and that should do it. Bro... Giving more than one solution which are not similar is not a sin, I guess... LOL
10.08.2021 12:40
Smathematician wrote: Commander_Anta78 wrote: sorry for bumping the thread again but everybody is giving their different respnses but actually there is no need,simply apply titu and that should do it. Bro... Giving more than one solution which are not similar is not a sin, I guess... LOL Oh I am sorry