10

How to evaluate $1 - \frac{\binom{n^2}{1}}{\binom{n+1}{1}} + \frac{\binom{n^2}{2}}{\binom{n+2}{2} } - \frac{\binom{n^2}{3}}{\binom{n+3}{3}} + \frac{\binom{n^2}{4}}{\binom{n+4}{4}} - ......$

I really have no idea how to proceed in this question. Expanding it by the formula isn't helping as much as I can see. And the options are extremely sofisticated as well like 1/n, 1/(n+1), 1. How to proceed?

  • 1
    Why do you think it should have a closed form simplification? – Hw Chu Jan 01 '18 at 17:40
  • 1
    Mainly because of the options. The options are 1/n , 1/(n+1), 1 and none of these. I know it can be none of these, but the options make me think the it will be transformed into closed form. I was thinking of trying to expand it using the formula and try to make a difference of series or something but wasn't able to. – Avinash Bhawnani Jan 01 '18 at 17:46
  • Is the sum really infinite? Because if it is as the question seems to show, then in the numerator, when the number of terms gets large enough, eventually the $k$ term in ${n^2 \choose k}$ will be larger than $n^2$, which would create a problem. – Cade Reinberger Jan 01 '18 at 17:49
  • 2
    No problem: those terms are all zero. – kimchi lover Jan 01 '18 at 17:51
  • Is it $\lim_{n\to\infty}$ ? – jonsno Jan 01 '18 at 18:03
  • Which book/problem set is this from? – Rick Jan 01 '18 at 18:04
  • 1
    Actually I am an amateur teacher. A senior teacher from my academy gave a problem sheet in the top batch, and one of the students from that batch asked me this. So, don't know the exact source. – Avinash Bhawnani Jan 01 '18 at 18:08
  • Giving a law of small numbers answer... the first two terms are $\frac 12,\frac 13$... – abiessu Jan 01 '18 at 18:25

2 Answers2

11

A variation. We obtain \begin{align*} \color{blue}{\sum_{j=0}^{n^2}}&\color{blue}{(-1)^j\binom{n^2}{j}\binom{n+j}{j}^{-1}}\\ &=\sum_{j=0}^{n^2}\binom{n^2}{j}\binom{-n-1}{j}^{-1}\tag{1}\\ &=\sum_{j=0}^{n^2}\binom{n^2}{j}(-n)\int_{0}^1t^j(1-t)^{-n-1-j}\,dt\tag{2}\\ &=(-n)\int_{0}^1(1-t)^{-n-1}\sum_{j=0}^{n^2}\binom{n^2}{j}\left(\frac{t}{1-t}\right)^j\,dt\\ &=(-n)\int_{0}^1(1-t)^{-n-1}\left(1+\frac{t}{1-t}\right)^{n^2}\,dt\\ &=(-n)\int_{0}^1(1-t)^{-n^2-n-1}\,dt\\ &=\frac{-n}{-n^2-n}\\ &\color{blue}{=\frac{1}{n+1}} \end{align*}

Comment:

  • In (1) we use the binomial identity $\binom{-p}{q}=\binom{p+q-1}{q}(-1)^q$.

  • In (2) we apply the identity $\binom{n}{r}^{-1}=(n+1)\int_{0}^1t^r(1-t)^{n-r}\,dt$.

Markus Scheuer
  • 108,315
  • 1
    @MathLover: Thanks, MathLover. :-) – Markus Scheuer Jan 01 '18 at 20:54
  • Hey, just can you give me links with proofs of those comments? I will really appreciate that. The 2nd one seems provable using a trigonometric substitution, but the first one seems awkward and it's the first time I have seen anything like that. – Avinash Bhawnani Jan 03 '18 at 06:13
  • @Avinash Bhawnani : if you replace the binomial coefficients with the explicit formula $\binom{\alpha}{k} = \frac{\alpha(\alpha-1)\cdot\cdot\cdot (\alpha-k+1)}{k!}$, (1) can easily be obtained. On the other hand, I don't know if you'll manage to prove the second one with a mere trigo substitution (see https://en.wikipedia.org/wiki/Beta_function#Relationship_between_gamma_function_and_beta_function for a proof involving double integrals, and see https://math.stackexchange.com/questions/3528/beta-function-derivation/3540#3540 for a proof involving intuition) – charmd Jan 03 '18 at 16:33
  • Thank you so much. I actually didn't try, but just had a feeling that with trigo substis I will be able to get it. Obviously, I was far from it. – Avinash Bhawnani Jan 03 '18 at 17:04
  • @charMD: Thanks for the nice references. – Markus Scheuer Jan 03 '18 at 21:58
6

Below is a quite brutal approach : whenever I see inverses of binomial coefficients, I try to use the following relation between the Gamma and Beta functions (I can provide a link if needed) :

$$\mbox{With } 0 < m \le n,\quad \quad \ \frac{1}{\binom{m+n}{m}} = \frac{mn}{m+n} \cdot \displaystyle{\int_0^1} t^{m-1}(1-t)^{n-1} dt$$

Your sum becomes :

$$S_n = 1 + \sum \limits_{k=1}^{n^2} (-1)^k \binom{n^2}{k} \frac{kn}{k+n} \displaystyle{\int_0^1} t^{k-1}(1-t)^{n-1} dt$$

$$S_n = 1 + n \displaystyle{\int_0^1} (1-t)^{n-1} \cdot \Big( \sum \limits_{k=1}^{n^2} \frac{(-1)^k k}{k+n} \binom{n^2}{k} t^{k-1} \Big) dt$$

Short interlude : we can simplify this sum. Denote $f : t \mapsto t^{n+1} \sum \limits_{k=1}^{n^2} \frac{(-1)^k k}{k+n} \binom{n^2}{k} t^{k-1}$. Then $f'(t) = \sum \limits_{k=1}^{n^2} (-1)^k k \binom{n^2}{k} t^{k+n-1} = t^n \frac{d}{dt} \big( \sum \limits_{k=0}^{n^2} \binom{n^2}{k} (-t)^k \Big) = t^n \frac{d}{dt} \Big( (1-t)^{n^2} \Big) = -n^2 t^n (1-t)^{n^2-1}$.

Hence $$S_n = 1 + n \displaystyle{\int_0^1} (1-t)^{n-1} \cdot \frac{1}{t^{n+1}} \displaystyle{\int_0^{t}}-n^2 s^n (1-s)^{n^2-1} ds dt$$

and by exchanging the integrals : $$S_n = 1 - n^3 \displaystyle{\int_0^1} s^n (1-s)^{n^2-1} \cdot \displaystyle{\int_s^1} \frac{(1-t)^{n-1}}{t^{n+1}} dtds$$

$$S_n = 1 - n^3 \displaystyle{\int_0^1} s^n (1-s)^{n^2-1} \cdot \frac{(1-s)^n}{n s^n}ds$$

$$S_n = 1 -n^2 \displaystyle{\int_0^1} (1-s)^{n^2+n-1} = 1 - \frac{n^2}{n^2+n}$$

There might (must) be a more elegant way, but with this we can conclude that $S_n = \frac{1}{n+1}$.

charmd
  • 6,050
  • 1
    I've been trying hard to get to the answer without using the gamma or beta function. Is there anyway it can be simplified from here: https://i.stack.imgur.com/zUNrV.jpg ? – Rick Jan 01 '18 at 20:49
  • 1
    Thanks ! And btw @Rick, to conclude with your method, it boils down to $\sum \limits_{k=0}^{n-1} (-1)^k \binom{n^2+n}{k} = \frac{(-1)^{n+1}}{n+1} \binom{n^2+n}{n}$. Wolfram says this is true, but I don't know how it does. In any case, there are more chances to prove this with some counting argument – charmd Jan 01 '18 at 21:40
  • alright, let me know if you get a proof for that last step :) – Rick Jan 01 '18 at 21:56
  • @Rick: A proof can be found at this MSE post. In fact a clever multiplication with $\pm 1$ does the job. – Markus Scheuer Jan 03 '18 at 22:04
  • @Markus Scheuer Great, thank you ! – charmd Jan 03 '18 at 23:19
  • @charMD: You're welcome, thanks. – Markus Scheuer Jan 04 '18 at 00:37