3

I came across the following finite sum involving (generalized) binomial coefficients:

$$ 2^q \sum_{k=0}^r \binom{r}{k} \binom{k/2}{q} (-1)^k .$$

Putting this into Mathematica gives me:

$$ (-1)^q 2^{r-q} \left( \binom{2q-r-1}{q-1} - \binom{2q-r-1}{q} \right) $$

and I'm interested in how could this solution be derived. There seems to be some binomial coefficient magic going on which I don't understand.

So far I have made very little progress, I noticed that the $2^q \binom{k/2}{q} = \frac{1}{q!} \prod_{i=0}^{q-1} (k-2i)$ -term looks a bit like a double factorial but this didn't get me very far. There also seems to be a lot of identities for sums involving $ \binom{r}{k} (-1)^k $ but I haven't found anything useful for this case.

Grigory M
  • 17,478
655321
  • 527

2 Answers2

4

Since I am a Bear of Very Little Brain, and long proofs Bother me, let me post slightly shorter version of essentially the same proof.

$2^q\sum_{k=0}^r(-1)^k\binom rk\binom{k/2}q$ is the coefficient of $z^q$ in the expansion of $(1-\sqrt{1+2z})^r=\left(\frac{-2z}{1+\sqrt{1+2z}}\right)^r$.

Now we want to substitue $\sqrt{1+2z}$ by $1+w$. There is a purely algebraic lemma for this, but one way to establish it is to write this coefficient as a (complex) integral and apply the change of variables formula for integrals: \begin{multline} \DeclareMathOperator{\res}{res} %\res\,(1-\sqrt{1+2z})^r\frac{dz}{z^{q+1}}= \res\left(\frac{-2z}{1+\sqrt{1+2z}}\right)^r\frac{dz}{z^{q+1}}= (-2)^r\res\frac1{(1+\sqrt{1+2z})^r}\frac{dz}{z^{q-r+1}}=\\ (-2)^r\res\frac1{2+w}\frac{dw+w\,dw}{(w+w^2/2)^{q-r+1}}= (-1)^r2^{2r-q-1} \res\frac1{(2+w)^{q-r+1}}\frac{dw+w\,dw}{w^{q-r+1}}. \end{multline} (here $\res_z=\frac1{2\pi i}\oint_{|z|=\epsilon}$, if you will; since $z=w+w^2/2$, $dz=dw+w\,dw$).

So we get a sum of two binomial coefficients (each multiplied by $(-1)^\cdots2^\cdots$) — that's the answer Mathematica gave you.

Grigory M
  • 17,478
  • Is there like a book or an online resource that would thoroughly explain these representations of coefficients as complex integrals (residues)? I'm having bit of a hard time with those (not this one in particular, but in general). – 655321 Jan 15 '15 at 02:53
  • @655321 I'm afraid I don't know a good reference... – Grigory M Jan 15 '15 at 19:22
  • @655321 though there is one example that is explained in detail in many books (e.g. in Enumerative Combinatorics): Lagrange inversion formula – Grigory M Jan 15 '15 at 19:24
0

Let $\mathbb{N}_0=\{0,1,2,\dotsc\}$.

  • For $m,n\in\mathbb{N}_0$, we have \begin{equation}\label{Wilf-Lemma2.2}\tag{1} \sum_{k=0}^{n}(-1)^{k} \binom{n}{k}\binom{k/2}{m} = \begin{cases} 0, & n>m\in\mathbb{N}_0;\\ \displaystyle (-1)^{m}n!\frac{[2(m-n)-1]!!}{(2m)!!}\binom{2m-n-1}{2(m-n)}, & m\ge n\in\mathbb{N}_0. \end{cases} \end{equation}
  • For $m,n\in\mathbb{N}_0$, we have \begin{equation}\tag{2} \sum_{k=0}^{n}(-1)^{k}\binom{n}{k}\binom{2k}{m} = \begin{cases} 0, & n>m\in\mathbb{N}_0;\\\displaystyle (-1)^n\binom{n}{m-n}2^{2n-m}, & m\ge n\in\mathbb{N}_0. \end{cases} \end{equation}
  • For $m\ge n\in\mathbb{N}_0$, we have \begin{equation}\tag{3} \sum_{n=0}^m\sum_{k=0}^{n}(-1)^{k} \binom{n}{k}\binom{k/2}{m} =(-1)^m\frac{(2m-1)!!}{(2m)!!}. \end{equation}

(A1) The proof of the identity \eqref{Wilf-Lemma2.2} can be found in Lemma 2.2 and its proof in the paper [1] below.

References

  1. Feng Qi and Mark Daniel Ward, Closed-form formulas and properties of coefficients in Maclaurin's series expansion of Wilf's function, arXiv (2021), available online at https://arxiv.org/abs/2110.08576v1.
  2. Siqintuya Jin, Bai-Ni Guo, and Feng Qi, Partial Bell polynomials, falling and rising factorials, Stirling numbers, and combinatorial identities, Computer Modeling in Engineering & Sciences (2022), in press; accepted on 24 January 2022; available online at https://dx.doi.org/10.32604/cmes.2022.019941 or https://www.researchgate.net/publication/358050501.
qifeng618
  • 1,691
  • Could you provide a proof? – user Oct 05 '21 at 11:38
  • @user I will provide a suitable proof, please keep patiently. Not only this answer, but also several other conclusions have been proved in a draft of a manuscript of mine. – qifeng618 Oct 06 '21 at 13:35
  • A detailed proof of this answer is the proof of Lemma 2.2 in "Feng Qi and Mark Daniel Ward, Closed-form formulas and properties of coefficients in Maclaurin's series expansion of Wilf's function, arXiv (2021), available online at https://arxiv.org/abs/2110.08576v1." – qifeng618 Oct 19 '21 at 03:23
  • I would suggest you to give the reference as the part of the answer rather than a comment. It would be also useful to refer the particular page or the formula number in the manuscript. – user Oct 19 '21 at 13:52
  • @user I will publish all proofs in officially published papers, but not here. One of reasons is that these proofs are not short and independent. – qifeng618 Oct 20 '21 at 12:09
  • I don't understand your reasons giving the reference in a comment rather than in the answer. – user Oct 21 '21 at 07:58
  • @user I told you the reasons above, but you didn't understand, I don't know why you didn't understand. I believe that, after reading the reference at https://arxiv.org/abs/2110.08576v1, you would and should understand all. I mainly aim to pubulish papers in mathematical journals, but don't only aim to post answers here. – qifeng618 Oct 22 '21 at 08:34
  • Then you can equally delete this answer. Without a proof it is of no use. – user Oct 24 '21 at 12:26
  • I don't think so. Currently their proofs are not suitable to directly post here. As you said, you didn't understand why. In my opinion, we shouldn't talk about what we don't understand. – qifeng618 Oct 24 '21 at 12:45
  • If you do not understand such simple things the further explanations are useless. Have a nice day. – user Oct 24 '21 at 12:49
  • Have nice days, everyone. – qifeng618 Oct 24 '21 at 13:06