9

If $\displaystyle p=\sum^{r}_{k=0}\binom{n}{2k}\binom{n-2k}{r-k}$ and $\displaystyle q=\sum^{n}_{k=r}\binom{n}{k}\binom{2k}{2r}\bigg(\frac{3}{4}\bigg)^{n-k}\bigg(\frac{1}{2}\bigg)^{2k-2r}$ , where $n\geq 2r$.

Then value of $\displaystyle \frac{p}{q}$

First I am try to simplify $p$ and $q$

$\displaystyle p=\sum^{r}_{k=0}\frac{n!}{2k!\cdot (n-2k)!}\times \frac{(n-2k)!}{(r-k)!\cdot (n-k-r)!}=\sum^{r}_{k=0}\frac{n!}{(2k)!\cdot (r-k)!\cdot (n-r-k)!}$

But I have seems that there is no closed form for $p$

And $\displaystyle q=\sum^{n}_{k=r}\frac{n!}{k!\cdot (n-k)!}\cdot \frac{(2k)!}{(2r)!\cdot (2k-2r)!}\bigg(\frac{3}{4}\bigg)^{n-k}\bigg(\frac{1}{2}\bigg)^{2k-2r}$

I did not understand how can I simplify $p$ and $q$ seems that something tricky in simplifying $p$ and $q$

Please have a look on that problem

  • I think these 2 questions should be split in 2 posts, to be more easily markable as duplicates. – Anne Bauval Mar 27 '23 at 03:21
  • 3
    @AnneBauval Unsure if I am interpreting your comment correctly. If I am, splitting into two posts may not be appropriate. That is, unless you already know the answer to at least one of the questions, it is theoretically possible that there is no (? elegant ?) closed form expression for either $~p~$ or $~q,~$ while there might be one for $~(p+q).$ Personally, I am totally ignorant in this area, so I have no opinion. – user2661923 Mar 27 '23 at 03:27
  • 1
    I am very sorry actually it is ratio of p and q not sum of p and q – Priti Bisht Mar 27 '23 at 03:30
  • 1
    No it is $\displaystyle \binom{n}{2k}$ – Priti Bisht Mar 27 '23 at 03:41

3 Answers3

11

It's just the same old Snake Oil method. Multiply through by $x^r$, sum over $r$, exchange the order of summation, find the closed form, extract the coefficient at $x^r$ or compare the generating functions, and then do whatever you want with the resulting $p$ and $q$. In fact, I will rename them $p_r$ and $q_r$ to mark their dependence on $r$.

In other words, $$ \begin{split} \sum_{r=0}^{\infty}{p_rx^r} &=\sum_{r=0}^{\infty}\sum_{k=0}^{r}\binom{n}{2k}\binom{n-2k}{r-k}x^r\\ &=\sum_{0\le k\le r}\binom{n}{2k}\binom{n-2k}{r-k}x^r\\ &=\sum_{k=0}^{\infty}\sum_{r=k}^{\infty}\binom{n}{2k}\binom{n-2k}{r-k}x^r\\ &=\sum_{k=0}^{\infty}\sum_{r=0}^{\infty}\binom{n}{2k}\binom{n-2k}{r}x^{r+k}\\ &=\sum_{k=0}^{\infty}\binom{n}{2k}x^k\left(\sum_{r=0}^{\infty}\binom{n-2k}{r}x^r\right)\\ &=\sum_{k=0}^{\infty}\binom{n}{2k}x^k(1+x)^{n-2k}\\ &=(1+x)^n\sum_{k=0}^{\infty}\binom{n}{2k}\left(\frac{\sqrt{x}}{1+x}\right)^{2k} \end{split} $$ Now $$ \sum_{k=0}^{\infty}\binom{n}{2k}t^{2k}=\frac{1}{2}\left(\sum_{k=0}^{\infty}\binom{n}{k}t^k+\sum_{k=0}^{\infty}\binom{n}{k}(-t)^k\right)=\frac{1}{2}\left((1+t)^n+(1-t)^n\right), $$ so the sum above is $$ \begin{split} \sum_{r=0}^{\infty}{p_rx^r}&=\frac{1}{2}(1+x)^n\left(\left(1+\frac{\sqrt{x}}{1+x}\right)^n+\left(1-\frac{\sqrt{x}}{1+x}\right)^n\right)\\ &=\frac{\left(1+\sqrt{x}+x\right)^n+\left(1-\sqrt{x}+x\right)^n}{2}. \end{split} $$ Similarly, $$ \begin{split} \sum_{r=0}^{\infty}q_rx^r &=\sum_{r=0}^{\infty}\sum_{k=r}^{n}\binom{n}{k}\binom{2k}{2r}\left(\frac{3}{4}\right)^{n-k}\left(\frac{1}{2}\right)^{2k-2r}x^r\\ &=\sum_{r=0}^{\infty}\sum_{k=0}^{n-r}\binom{n}{k}\binom{2n-2k}{2r}\left(\frac{3}{4}\right)^{k}\left(\frac{1}{2}\right)^{2n-2r-2k}x^r\\ &=\sum_{r=0}^{\infty}\sum_{k=0}^{n-r}\binom{n}{k}\binom{2n-2k}{2r}\left(\frac{3}{4}\right)^{k}\left(\frac{1}{4}\right)^{n-r-k}x^r\\ &=\sum_{k=0}^{n}\sum_{r=0}^{n-k}\binom{n}{k}\left(\frac{3}{4}\right)^{k}\left(\frac{1}{4}\right)^{n-k}\binom{2n-2k}{2r}4^rx^r\\ &=\frac{1}{4^n}\sum_{k=0}^{n}\binom{n}{k}3^k\left(\sum_{r=0}^{n-k}\binom{2n-2k}{2r}\left(2\sqrt{x}\right)^{2r}\right)\\ &=\frac{1}{2\cdot4^n}\sum_{k=0}^{n}\binom{n}{k}3^k\left((1+2\sqrt{x})^{2n-2k}+(1-2\sqrt{x})^{2n-2k}\right)\\ &=\frac{1}{2\cdot4^n}\left(\left(3+(1+2\sqrt{x})^2\right)^n+\left(3+(1-2\sqrt{x})^2\right)^n\right)\\ &=\frac{1}{2\cdot4^n}\left(\left(4+4\sqrt{x}+4x\right)^n+\left(4-4\sqrt{x}+4x\right)^n\right)\\ &=\frac{\left(1+\sqrt{x}+x\right)^n+\left(1-\sqrt{x}+x\right)^n}{2}. \end{split} $$

Thus, the generating functions of the sequences $(p_r)$ and $(q_r)$ are equal for any $n$, so $p_r=q_r$ for any $n$ and $r$, and thus their ratio is $1$ whenever they are nonzero.

The triangular array of values of $p(n,r)=q(n,r)$ for $0\le r\le n$ is the OEIS sequence A056241.

3

We will show that using Egorychev method

$$\sum_{k=0}^r {n\choose 2k} {n-2k\choose r-k} = \sum_{k=r}^n {n\choose k} {2k\choose 2r} \left(\frac{3}{4}\right)^{n-k} \left(\frac{1}{2}\right)^{2k-2r}.$$

Computation for LHS

We get for LHS

$$\sum_{k=0}^r {n\choose n-2k} {n-2k\choose r-k} \\ = [z^n] (1+z)^n [w^r] (1+w)^n \sum_{k\ge 0} z^{2k} \frac{w^k}{(1+w)^{2k}}.$$

Here we have extended to infinity due to the coefficient extractor in $w$. We obtain

$$[z^n] (1+z)^n [w^r] (1+w)^n \frac{1}{1-z^2 w/(1+w)^2} \\ = [z^n] (1+z)^n [w^r] (1+w)^{n+2} \frac{1}{(1+w)^2-wz^2} \\ = [z^n] (1+z)^n [w^{2r}] (1+w^2)^{n+2} \frac{1}{(1+w^2)^2-w^2z^2} \\ = [z^n] (1+z)^n [w^{2r}] (1+w^2)^{n+2} \frac{1}{1+wz+w^2}\frac{1}{1-wz+w^2} \\ = - [z^n] (1+z)^n [w^{2r+2}] (1+w^2)^{n+2} \frac{1}{z+(1+w^2)/w}\frac{1}{z-(1+w^2)/w}.$$

The contribution from $z$ is

$$\;\underset{z}{\mathrm{res}}\; \frac{1}{z^{n+1}} (1+z)^n \frac{1}{z+(1+w^2)/w}\frac{1}{z-(1+w^2)/w}.$$

Here the residue at infinity is zero so we may use minus the residues at $z=\pm (1+w^2)/w.$ We get first

$$[w^{2r+2}] (1+w^2)^{n+2} (-1)^{n+1} \frac{w^{n+1}}{(1+w^2)^{n+1}} \frac{(-1+w-w^2)^n}{w^n} \frac{1}{-2(1+w^2)/w} \\ = \frac{1}{2} [w^{2r}] (w^2-w+1)^n = \frac{1}{2} (-1)^{2r} [w^{2r}] (w^2-w+1)^n \\ = \frac{1}{2} [w^{2r}] (w^2+w+1)^n.$$

and second

$$[w^{2r+2}] (1+w^2)^{n+2} \frac{w^{n+1}}{(1+w^2)^{n+1}} \frac{(w^2+w+1)^n}{w^n} \frac{1}{2(1+w^2)/w} \\ = \frac{1}{2} [w^{2r}] (w^2+w+1)^n.$$

Collecting everything,

$$[w^{2r}] (w^2+w+1)^n.$$

Computation for RHS

We have for RHS

$$\left(\frac{3}{4}\right)^n \left(\frac{1}{2}\right)^{-2r} \sum_{k=r}^n {n\choose n-k} {2k\choose 2r} \left(\frac{4}{3}\right)^k \left(\frac{1}{2}\right)^{2k} \\ = \left(\frac{3}{4}\right)^n \left(\frac{1}{2}\right)^{-2r} [z^n] (1+z)^n [w^{2r}] \sum_{k=r}^n z^k (1+w)^{2k} \left(\frac{1}{3}\right)^k$$

We may lower $k$ to zero owing to the coefficient extractor in $w$ and raise to infinity due to the extractor in $z$ to get

$$\left(\frac{3}{4}\right)^n \left(\frac{1}{2}\right)^{-2r} [z^n] (1+z)^n [w^{2r}] \frac{1}{1-z(1+w)^2/3} \\ = 3^n 2^{2r-2n} [z^n] (1+z)^n [w^{2r}] \frac{1}{1-z(1+w)^2/3} \\ = 2^{2r-2n} [z^n] (1+3z)^n [w^{2r}] \frac{1}{1-z(1+w)^2} \\ = - 2^{2r-2n} [z^n] (1+3z)^n [w^{2r}] \frac{1}{(1+w)^2} \frac{1}{z-1/(1+w)^2}.$$

The contribution from $z$ is

$$\;\underset{z}{\mathrm{res}}\; \frac{1}{z^{n+1}} (1+3z)^n \frac{1}{z-1/(1+w)^2}.$$

Here the residue at infinity is zero so we may use minus the residue at $z=1/(1+w)^2$ to get

$$ 2^{2r-2n} [w^{2r}] \frac{1}{(1+w)^2} (1+w)^{2n+2} (1+3/(1+w)^2)^n \\ = 2^{2r-2n} [w^{2r}] (4+2w+w^2)^n = 2^{-2n} [w^{2r}] (4+4w+4w^2)^n \\ = [w^{2r}] (w^2+w+1)^n.$$

We have equality and thus the claim. Here we had some help from OEIS A005714.

Marko Riedel
  • 61,317
3

We transform the binomial expressions into hypergeometric series and show this way equality.

Left-hand side:

Starting with the left-hand side we obtain using rising factorials $(a)_{k}:=a(a+1)\cdots(a+k-1)$ \begin{align*} \color{blue}{\sum_{k=0}^r}&\color{blue}{\underbrace{\binom{n}{2k}\binom{n-2k}{r-k}}_{=:t_k}}\\ &=\sum_{k=0}^r t_k\\ &= t_0\sum_{k=0}^r\prod_{j=0}^{k-1}\frac{t_{j+1}}{t_j}\\ &=\binom{n}{r}\sum_{k=0}^r\prod_{j=0}^{k-1}\left(\binom{n}{2j+2}\binom{n-2j-2}{r-j-1}\binom{n}{2j}^{-1}\binom{n-2j}{r-j}^{-1}\right)\\ &=\binom{n}{r}\sum_{k=0}^r\prod_{j=0}^{k-1}\left(\frac{1}{4}\frac{(j-r)(j-n+r)}{(j+1)\left(j+\frac{1}{2}\right)}\right)\\ &=\binom{n}{r}\sum_{k=0}^{r}\frac{(-r)_k(r-n)_{k}}{\left(\frac{1}{2}\right)_k}\left(\frac{1}{4}\right)^k\frac{1}{k!}\\ &\,\,\color{blue}{= \binom{n}{r}\ _2F_1\left(-r,r-n;\frac{1}{2};\frac{1}{4}\right)}\tag{1} \end{align*}

Right-hand side:

We obtain \begin{align*} \color{blue}{\sum_{k=r}^n}&\color{blue}{ \binom{n}{k}\binom{2k}{2r}\left(\frac{3}{4}\right)^{n-k}\left(\frac{1}{2}\right)^{2k-2r}}\\ &=\sum_{k=0}^{n-r}\binom{n}{k+r}\binom{2k+2r}{2r}\left(\frac{3}{4}\right)^{n-k-r}\left(\frac{1}{4}\right)^k\\ &=\left(\frac{3}{4}\right)^{n-r}\sum_{k=0}^{n-r}\underbrace{\binom{n}{k+r}\binom{2k+2r}{2r}\left(\frac{1}{3}\right)^k}_{=:t_k}\\ &=\left(\frac{3}{4}\right)^{n-r}\sum_{k=0}^{n-r} t_k\\ &= \left(\frac{3}{4}\right)^{n-r}t_0\sum_{k=0}^{n-r}\prod_{j=0}^{k-1}\frac{t_{j+1}}{t_j}\\ &=\left(\frac{3}{4}\right)^{n-r}\binom{n}{r}\sum_{k=0}^{n-r}\prod_{j=0}^{k-1} \left(\binom{n}{j+1+r}\binom{2j+2+2r}{2r}\left(\frac{1}{3}\right)^{j+1}\right.\\ &\qquad\qquad\qquad\qquad\qquad\qquad\left.\cdot\binom{n}{j+r}^{-1}\binom{2j+2r}{2r}^{-1}\left(\frac{1}{3}\right)^{-j}\right)\\ &=\left(\frac{3}{4}\right)^{n-r}\binom{n}{r}\sum_{k=0}^{n-r}\prod_{j=0}^{k-1} \left(\frac{\left(j+r+\frac{1}{2}\right)(j-n+r)\left(-\frac{1}{3}\right)}{(j+1)\left(j+\frac{1}{2}\right)}\right)\\ &=\left(\frac{3}{4}\right)^{n-r}\binom{n}{r}\sum_{k=0}^{n-r}\frac{\left(r+\frac{1}{2}\right)_k(r-n)_k} {\left(\frac{1}{2}\right)_k}\left(-\frac{1}{3}\right)^k\frac{1}{k!}\\ &\,\,\color{blue}{=\left(\frac{3}{4}\right)^{n-r}\binom{n}{r}\ _2F_1\left(r+\frac{1}{2},r-n;\frac{1}{2};-\frac{1}{3}\right)}\tag{2} \end{align*}

Pfaff Transformation:

Applying to (1) the Pfaff transformation \begin{align*} \,\,\color{blue}{_2F_1\left(a,b;c;z\right)=(1-z)^{-b}\ _2F_1\left(c-a,b;c;\frac{z}{z-1}\right)} \end{align*} and equality of (1) and (2) follows.

Markus Scheuer
  • 108,315