6

I want to prove the product of even/odd power with a combinatorial number: \begin{aligned} \sum_{i=1}^n i \binom{2n}{n-i}&=\frac12(n+1) \binom{2n}{n-1}, \\ \sum_{i=1}^n i^2 \binom{2n}{n-i}&=2^{2n-2} n. \end{aligned} I am sure these results are correct since WolframAlpha verifies them. I wonder if the first formula can be proved, and also the general version can be proved: Formula related to combinatorial number.


Below is proof for the second formula:

First, note that $r\binom{n}{r}=n\binom{n-1}{r-1}$, and $\sum_{i=0}^n\binom{n}{i}=2^n$. We thus have \begin{aligned} \sum_{i=0}^n i\binom{n}{i}&=2^{n-1}n, \\ \sum_{i=0}^n i^2\binom{n}{i} &= \sum_{i=0}^n i(i-1)\binom{n}{i}+\sum_{i=0}^n i\binom{n}{i} \\ &= 2^{n-2}(n^2-n) + 2^{n-1} n \\ &= n(n+1)2^{n-2}. \end{aligned} Therefore, \begin{aligned} \sum_{i=0}^{2n}(n-i)^2\binom{2n}{i} &= n^2\sum_{i=0}^{2n}\binom{2n}{i}-2n\sum_{i=0}^{2n}i\binom{2n}{i}+\sum_{i=0}^{2n}i^2 \binom{2n}{i} \\ &= 2^{2n}n^2-2^{2n+1}n^2+n(2n+1)2^{2n-1} \\ &= 2^{2n-1} n, \end{aligned} and thus \begin{aligned} \sum_{i=0}^n i^2 \binom{2n}{n-i} = \sum_{i=0}^n (n-i)^2 \binom{2n}{i} = \frac12 \sum_{i=0}^{2n} (n-i)^2 \binom{2n}{i} = 2^{2n-2} n. \end{aligned}

RobPratt
  • 45,619
lsstat
  • 145

4 Answers4

3

\begin{align} \sum_{i=1}^n i\binom{2n}{n-i} &= \sum_{i=0}^n i\binom{2n}{n-i} \\ &= \sum_{i=0}^n (n-i)\binom{2n}{i} \\ &= n\sum_{i=0}^n \binom{2n}{i} - \sum_{i=1}^n i\binom{2n}{i} \\ &= \frac{n}{2}\left(\binom{2n}{n}+\sum_{i=0}^{2n} \binom{2n}{i}\right) - 2n\sum_{i=1}^n \binom{2n-1}{i-1} \\ &= \frac{n}{2}\left(\binom{2n}{n}+2^{2n}\right) - 2n\sum_{i=0}^{n-1}\binom{2n-1}{i} \\ &= \frac{n}{2}\left(\binom{2n}{n}+2^{2n}\right) - \frac{2n}{2}\sum_{i=0}^{2n-1}\binom{2n-1}{i} \\ &= \frac{n}{2}\left(\binom{2n}{n}+2^{2n}\right) - n2^{2n-1}\\ &= \frac{n}{2}\binom{2n}{n}\\ &= \frac{n+1}{2}\binom{2n}{n-1} \end{align}

RobPratt
  • 45,619
1

Let $$S_n=\sum_{k=0}^{n} {2n \choose k}$$ and let $$S'_n=\sum_{k=0}^{n} k {2n \choose n-k}=\sum_{k=0}^{n} (n-k) {2n \choose k}=nS_n-T_n$$ $$2^{2n}=\sum_{k=0}^{2n} {2n \choose k}=\sum_{k=0}^{n} {2n \choose k}+\sum_{k=n+1}^{2n} {2n \choose k}$$ $$\implies S_n+\sum_{k=n+1}^{2n} {2n \choose k}=2^{2n}$$ Change $k$ to $2n-j$, then $$S_n+\sum_{j=n-1}^{0} {2n \choose 2n-j}=2^{2n}$$ $$\implies S_n+ \sum_{j=0}^{n} {2n \choose j}-{2n \choose n}=2^{2n}$$ $$\implies S_n=2^{2n-1}+\frac{1}{2}{2n \choose n}.$$ Next let $$T_n=\sum_{k=0}^{n} k {2n \choose k}=\sum_{k=0}^{n} k\frac{(2n)!}{k!(2n-k)!}=2n\sum_{k=0}^{n}{2n-1 \choose k-1}=2n\sum_{j=0}^{n-1}{2n-1 \choose j} $$ We have $$2^{2n-1}= \sum_{k=0}^{2n-1} {2n-1\choose k}=\sum_{k=0}^{n-1} {2n-1 \choose k}+\sum_{k=n}^{2n-1} {2n-1 \choose k}=\sum_{k=0}^{n-1} {2n-1 \choose k}+\sum_{j=0}^{n-1} {2n-1 \choose n+j}$$ $$2^{2n-1}\implies \sum_{k=0}^{n-1} {2n-1 \choose k}+\sum_{j=0}^{n-1} {2n-1 \choose n+j}=\sum_{k=0}^{n-1} {2n-1 \choose k}+\sum_{j=0}^{n-1} {2n-1 \choose n-1-j}$$ $$\implies 2^{2n-1}= \sum_{k=0}^{n-1} {2n-1 \choose k}+\sum_{m=0}^{n-1} {2n-1 \choose m} \implies \sum_{k=0}^{n-1} {2n-1 \choose k}=2^{2n-2}$$ Finally, we have $$S'_n=n2^{2n-1}+\frac{n}{2} {2n \choose n}-n2^{2n-1}=\frac{n}{2}{2n \choose n}$$

Z Ahmed
  • 43,235
1

Related Equation

Just for a fun alternative, we have the following equation that we will prove below.

$$ \sum_{i=2}^{n} (n+i)\binom{2n}{n+i} - \sum_{i=1}^{n} (n-i)\binom{2n}{n-i} = 0 $$

Combinatorial Proof

One wish to colour $2n$ distinguishable balls such that exactly one is pink, some (can be none) are black, and some (can be none) are white. The two terms on LHS are, respectively:

  1. Number of possible colouring such that there are at least $n+1$ black balls
  2. Number of possible colouring such that there are at most $n-2$ white balls

Easy to see that (1) equals (2) so their difference is zero.

Relation to Original Question

$$ \begin{align} \sum_{i=1}^{n} i\binom{2n}{n-i} &= \frac{1}{2} \left[ \sum_{i=1}^{n} (n+i)\binom{2n}{n+i} - \sum_{i=1}^{n} (n-i)\binom{2n}{n-i} \right] \\\\ &= \frac{n+1}{2}\binom{2n}{n+1} \end{align} $$

Potential Extension

Instead of one pink, colour $m$ balls pink.

$$ \sum_{i=m+1}^{n}\binom{n+i}{m}\binom{2n}{n+i} - \sum_{i=1}^{n}\binom{n-i}{m}\binom{2n}{n-i} = 0 $$

The two terms are, respectively:

  1. Number of possible colouring such that there are at least $n+1$ black balls.
  2. Number of possible colouring such that there are at most $n-m-1$ white balls.

These two are the same, thus the difference is zero and we can write:

$$ \sum_{i=1}^{n}\left[\binom{n+i}{m}-\binom{n-i}{m}\right]\binom{2n}{n-i}= \sum_{i=1}^{m}\binom{n+i}{m}\binom{2n}{n-i} $$

For odd $m$, $\binom{n+i}{m}-\binom{n-i}{m}$ has the term $i^{m}$

acat3
  • 11,897
  • Thanks! This is a very clever idea. I wonder whether this idea can be used to prove the general version. I use $\sum_{i=1}^n i\binom{2n}{n-i}$ to prove $\sum_{i=1}^n i^3\binom{2n}{n-i}$ and it can be done. But using induction (i.e. using case $i^r$ to prove case $i^{r+2}$) seems complicated. Do you have any idea? – lsstat Jan 08 '23 at 10:59
  • 1
    @lsstat I added some possible extension in the answer. The first part of the answer is a special case of $m=1$, for this value we can get quite elegant solution. For higher $m$, I think it's not elegant anymore, so I didn't write the answer for higher $m$ before – acat3 Jan 08 '23 at 15:06
  • 1
    @lsstat btw, I posted a question for the identity in my answer. In case you are also interested in other proof, here is the link: https://math.stackexchange.com/q/4614246/751970 – acat3 Jan 08 '23 at 15:33
  • Thanks! You are quite correct. For higher $m$, it may be much more difficult than we thought. I was thinking of the generating functions for $i^m\binom{2n}{n-i}$. By WolframAlpha, it is $i^m\binom{0}{-i} _3F_2(\frac12,1,1;1-i,i+1;4z)$. The partial sum of hypergeometric functions may give some useful results, but I haven't found them yet. See page 223 section 5.7 in the book: https://www.csie.ntu.edu.tw/~r97002/temp/Concrete%20Mathematics%202e.pdf – lsstat Jan 09 '23 at 03:28
0

I don't konw how to prove it, but for general case I found something:

Integral Representation and the Computation of Combinatorial Sums by G.P.EGORYCHEV Page77   says that:

enter image description here


verify the formula$\sum\limits_{i=1}^n i\left(\begin{array}{c} 2 n \\ n-i \end{array}\right)=\frac{1}{2}(n+1)\left(\begin{array}{c} 2 n \\ n-1 \end{array}\right)$ with Mathematica

Sum[Binomial[i, 1]*Binomial[2*n, n - i], {i, 1, n}]

$$ \frac{1}{2} (n+1) \binom{2 n}{n-1} $$

138 Aspen
  • 880
  • Thanks! Mathematica indeed gives the answer, but I would like to know the proof. For the book, I guess what I need is $k^x$ instead of $x^k$ on the right-hand side of (2.42). – lsstat Jan 05 '23 at 04:21
  • no, just let $x=1$ would be ok. – 138 Aspen Jan 05 '23 at 04:29
  • OK. It seems we can choose $a_1=2,a_4=1,a_5=-1,a_2=a_3=a_6=0$. But how can we choose $b$ to make the second combinatorial number become $k^r$? – lsstat Jan 05 '23 at 04:55