Let $n,k$ be given positive integers with $n>k$. Prove that: \[ \frac{1}{n+1} \cdot \frac{n^n}{k^k (n-k)^{n-k}} < \frac{n!}{k! (n-k)!} < \frac{n^n}{k^k(n-k)^{n-k}} \]
Problem
Source: APMO 2000
Tags: inequalities, induction, function, probability
02.04.2006 15:46
For $k=n-1$ or $k=1$, the first inequality becomes: $n^{n}<n(n+1)((n-1)^{n-1})$, or: $(1+\frac{1}{n-1})^{n-1}<n+1$. But LHS$<e<3\leq n$. The second inequality becomes: $n<\frac{n^{n}}{(n-1)^{n-1}}$, or: $1<(1+\frac{1}{n-1})^{n-1}$, evidenty true. Now return to the initial problem. For $n=3$, the inequality is true, since we proved it for $k=1, 2=3-1$. Consider the inequalities true for all $n,k$ such that $k<n\leq t-1$. We'll prove the inequality for $t$. Consider $k>1$. The second inequality becomes: $\frac{t!}{k!(t-k)!}<\frac{t^{t}}{k^{k}(t-k)^{t-k}} \Leftrightarrow$, $\frac{(t-1)!}{(k-1)!(t-k)!)}<\frac{t^{t-1}}{k^{k-1}(t-k)^{t-k}}$. By the induction hypothesis: $\frac{(t-1)!}{(k-1)!(t-k)!}<\frac{(t-1)^{t-1}}{(k-1)^{k-1}(t-k)^{t-k}}$. So it suffices to prove that: $\frac{(t-1)^{t-1}}{(k-1)^{k-1}(t-k)^{t-k}}<\frac{t^{t-1}}{k^{k-1}(t-k)^{t-k}}$, or: $(1+\frac{1}{k-1})^{k-1}<(1+\frac{1}{t-1})^{t-1}$. But the function $(1+\frac{1}{x})^{x}$ is increasing. Unfortunately, the first inequality doesn't result in a similar way
18.12.2010 04:38
shobber wrote: Let $n,k$ be given positive integers with $n>k$. Prove that: \[ \frac{1}{n+1} \cdot \frac{n^n}{k^k (n-k)^{n-k}} < \frac{n!}{k! (n-k)!} < \frac{n^n}{k^k(n-k)^{n-k}} \] Lemma: Let $a$ and $b$ be any positive integers. For any integer $i$ between 0 and $a+b$, $\binom{a+b}{i} a^i b^{a+b-i} \leq \binom{a+b}{a} a^a b^b$, with equality not holding when $i=0$ or when $i = a+b.$ Proof: Let $f(i) = \binom{a+b}{i} a^i b^{a+b-i}$. We observe that $f(a) = \binom{a+b}{a} a^a b^b$. We will show that $\frac{f(i)}{f(i+1)} \geq 1$ whenever $i \geq a$ and that $\frac{f(i)}{f(i+1)} \leq 1$ whenever $i \leq a$. We have \[ \frac{f(i)}{f(i+1)} = \frac{\binom{a+b}{i} a^i b^{a+b-i}}{\binom{a+b}{i+1} a^{i+1} b^{a+b-i-1}} = \frac{\frac{(a+b)!}{i! (a+b-i)!}}{\frac{(a+b)!}{(i+1)! (a+b-i-1)!}} \cdot \frac{b}{a} = \frac{i}{a+b-i} \cdot \frac{b}{a}.\] If $i \geq a$, then \[ \frac{i}{a+b-i} \cdot \frac{b}{a} = \frac{1}{\frac{a+b}{i} - 1} \cdot \frac{b}{a} \geq \frac{1}{\frac{a+b}{a} - 1} \cdot \frac{b}{a} = 1, \] and if $i \leq a$, then \[ \frac{i}{a+b-i} \cdot \frac{b}{a} = \frac{1}{\frac{a+b}{i} - 1} \cdot \frac{b}{a} \leq \frac{1}{\frac{a+b}{a} - 1} \cdot \frac{b}{a} = 1,\] as desired. To show that $f(0) < f(a)$ or $f(a+b) < f(a)$, we suppose without loss of generality $a \geq b$. Then \[ \frac{f(a)}{f(0)} = \frac{a^a b^b \binom{a+b}{a}}{b^{a+b}} = \left(\frac{a}{b}\right)^a \binom{a+b}{a} > 1.\] Let $j = n-k$. We wish to show that \[ \frac{1}{j+k+1}{\cdot \frac{(j+k)^{j+k}}{j^j k^k} < \frac{(j+k)!}{j! k!} < \frac{(j+k)^{j+k}}{j^j k^k} }\] We have \[ (j+k)^{j+k} = \sum_{i=0}^{j+k} \binom{j+k}{i} j^i k^{j+k-i} \&> \binom{j+k}{j} j^j k^k \] which rearranges to \[ \frac{(j+k)!}{j! k!} < \frac{(j+k)^{j+k}}{j^j k^k}. \] In addition, by the lemma, we have \[ (j+k)^{j+k} = \sum_{i=0}^{j+k} \binom{j+k}{i} j^i k^{j+k-i} > \sum_{i=0}^{j+k} \binom{j+k}{j} j^j k^k = (j+k+1) \binom{j+k}{j} j^j k^k, \] which rearranges to \[ \frac{1}{j+k+1}{\cdot \frac{(j+k)^{j+k}}{j^j k^k} < \frac{(j+k)!}{j! k!}, }\] so we are done.
23.02.2011 07:59
well, let k/n = p. Then, it is pretty easy to see that the claim is equivalent to showing that 1/(n+1) < (nCk)(p^k*(1-p)^(n-k) < 1 however, simply considering a binomial distribution with n trials, a probability of success p, we are nearly instantly done (a binomial distribution is maximized at the mean, np = k.)
06.10.2021 20:57
Lets consider $n^n=(k+n-k)^n.$ The expansion of $n^n=\sum _{i=0}^n\binom{n}{i}k^k(n-k)^{n-k}.$ One of the term of the expansion is $n\binom{n}{i}k^k{(n-k)}^{n-k}$ which is strictly less than $n^n$. Dividing by $k^k\times (n-k)^{n-k}$ we get the second inequality. For the first inequality, we have to show that the highest term of the expansion of $n^n$ is $\binom{n}{k}k^k{(n-k)}^{n-k}$. For this let $T_r=\binom{n}{r}k^r{(n-k)}^{n-r}$ and $T_{r+1}=\binom{n}{r+1}k^{r+1}{(n-k)}^{n-r-1}$. Notice that $T_r>T_{r+1}$ for $r\geq k$ and smaller when $r<k$. Therefore we get $(n+1)\binom{n}{k}k^k{(n-k)}^{n-k}>n^n$. Dividing both side by $k^k\times (n-k)^{n-k}$ and rearranging we get our desired first inequality.
14.02.2022 00:58
Abdullahil_Kafi wrote: Lets consider $n^n=(k+n-k)^n.$ The expansion of $n^n=\sum _{i=0}^n\binom{n}{i}k^k(n-k)^{n-k}.$ One of the term of the expansion is $n\binom{n}{i}k^k{(n-k)}^{n-k}$ which is strictly less than $n^n$. Dividing by $k^k\times (n-k)^{n-k}$ we get the second inequality. For the first inequality, we have to show that the highest term of the expansion of $n^n$ is $\binom{n}{k}k^k{(n-k)}^{n-k}$. For this let $T_r=\binom{n}{r}k^r{(n-k)}^{n-r}$ and $T_{r+1}=\binom{n}{r+1}k^k{(n-k)}^{n-r-1}$. Notice that $T_r>T_{r+1}$ for $r>m$ and smaller when $r<m$. Therefore we get $(n+1)\binom{n}{k}k^k{(n-k)}^{n-k}>n^n$. Dividing both side by $k^k\times (n-k)^{n-k}$ and rearranging we get our desired first inequality. can u explain how did you prove the first ineq it is unclear to me
14.02.2022 04:01
TFIRSTMGMEDALIST wrote: can u explain how did you prove the first ineq it is unclear to me In any $n$ power binomial, there are exactly $n+1$ terms, and all the terms are not equal. Therefore if we sum the biggest term $n+1$ times then it is obviously strictly bigger than the original and hence done.
14.02.2022 12:16
yes but before this how did you prove that the highest term of the expansion is ..... you didn't define m and there's a mistake in the expansion of n up to n
14.02.2022 20:43
TFIRSTMGMEDALIST wrote: yes but before this how did you prove that the highest term of the expansion is ..... you didn't define m and there's a mistake in the expansion of n up to n Ok, $m$ is actually $k$ (I edited).
05.03.2023 18:37
being an exact bound rather than an asymptotic formula) we have: \begin{align*}\frac{n!}{k! (n-k)!}&>\frac{\sqrt {2\pi n}\left(\frac{n}{e}\right)^{n}e^{\frac{1}{12n+1}}}{\left(\sqrt {2\pi k}\left(\frac{k}{e}\right)^{k}e^{\frac{1}{12k}}\right)\cdot \left(\sqrt {2\pi (n-k)}\left(\frac{n-k}{e}\right)^{n-k}e^{\frac{1}{12(n-k)}}\right)} \\ &=\frac{n^{n}}{k^{k}(n-k)^{n-k}}\cdot \frac{\sqrt{n}}{\sqrt{k(n-k)}}\cdot\frac{1}{\sqrt{2 \pi}} \cdot e^{\frac{1}{12n+1}-\frac{1}{12k}-\frac{1}{12n-12k}}\\ &>\frac{n^{n}}{k^{k}(n-k)^{n-k}}\cdot \frac{\sqrt{n}}{\sqrt{k(n-k)}}\cdot\frac{1}{\sqrt{2 \pi}} \cdot e^{-1/6}\\ &>\frac{n^{n}}{k^{k}(n-k)^{n-k}}\cdot \frac{4}{n+1}\cdot \frac{1}{3}\\ &>\frac{n^{n}}{k^{k}(n-k)^{n-k}}\cdot \frac{1}{n+1}\\ \end{align*}where we've used the lower bound approximation for $n!$ and the upper bound for $k!$ and $(n-k)!$ as they're in the denominator. Also, to carry out the inequality we've used the following (far from strict) bounds: \[\frac{1}{12n+1}-\frac{1}{12k}-\frac{1}{12n-12k}>-\frac{1}{12k}-\frac{1}{12(n-k)}\geq-\frac{1}{12}-\frac{1}{12}=-\frac{1}{6}\]\[\frac{e^{-1/6}}{\sqrt{2 \pi}}=\frac{1}{\sqrt[6]{8\pi^3e}}>\frac{1}{3}\]\[\frac{n}{2}=\frac{k+(n-k)}{2}\geq \sqrt{k(n-k)}\Longrightarrow \frac{\sqrt{n}}{\sqrt{k(n-k)}}\geq\frac{\sqrt{n}}{n/2}=\frac{2}{\sqrt{n}}\geq \frac{2}{\frac{n+1}{2}}=\frac{4}{n+1}\]We prove the right-hand inequality in the same manner: \begin{align*}\frac{n!}{k! (n-k)!}&<\frac{\sqrt {2\pi n}\left(\frac{n}{e}\right)^{n}e^{\frac{1}{12n}}}{\left(\sqrt {2\pi k}\left(\frac{k}{e}\right)^{k}e^{\frac{1}{12k-1}}\right)\cdot \left(\sqrt {2\pi (n-k)}\left(\frac{n-k}{e}\right)^{n-k}e^{\frac{1}{12(n-k)+1}}\right)} \\ &=\frac{n^{n}}{k^{k}(n-k)^{n-k}}\cdot \frac{\sqrt{n}}{\sqrt{k(n-k)}}\cdot\frac{1}{\sqrt{2 \pi}} \cdot e^{\frac{1}{12n}-\frac{1}{12k+1}-\frac{1}{12n-12k+1}}\\ &<\frac{n^{n}}{k^{k}(n-k)^{n-k}}\cdot \frac{\sqrt{n}}{\sqrt{k(n-k)}}\cdot\frac{1}{\sqrt{2 \pi}} \cdot 1\\ &<\frac{n^{n}}{k^{k}(n-k)^{n-k}}\cdot \frac{\sqrt{n}}{\sqrt{n-1}}\cdot \frac{1}{2}\\ &<\frac{n^{n}}{k^{k}(n-k)^{n-k}}\\ \end{align*}where we've used the upper bound approximation for $n!$ and the lower bound for $k!$ and $(n-k)!$ as they're in the denominator. Also, to carry out the inequality we've used the following (far from strict) bounds:
06.10.2024 17:27
Any combinatorial solution?