Define the sequence given by the finite sum $$a_n:=\sum_{k=2}^{n+1}\binom{2k}k\binom{n+1}k\frac{k-1}{2^k\binom{4n}k}.$$


(1) Is $0<a_n<1$?

(2) Does the limit $\lim_{n\rightarrow\infty}a_n$ exist? If yes, what is its value?

  • 1
    $\begingroup$ Not having to deal with such sums in my work, it intrigues me how does one come about such a formula? $\endgroup$ – Asaf Karagila Jan 24 '17 at 7:47
  • 2
    $\begingroup$ Is this a research-level question...? Or did this site devolve? $\endgroup$ – user541686 Jan 24 '17 at 9:44
  • $\begingroup$ @AsafKaragila I don't know either, but I guess the reason is co.combinatorics. $\endgroup$ – Dirk Jan 24 '17 at 13:54
  • 2
    $\begingroup$ @Dirk: Is co-combinatorics just a long way of saying "mbinatorics"? :) $\endgroup$ – Asaf Karagila Jan 24 '17 at 14:03
  • $\begingroup$ @AsafKaragila Well, sometimes you encounter these kind of sums by counting the number of objects that you can construct somehow. We once met this one: $$ \sum_{k=0}^n \binom{n,k} \binom{k}{\lfloor \frac{k}{2} \rfloor} 2^{n-k}. $$ (Although it turned out to be a rather simple binomial coefficient.) $\endgroup$ – Daniel Soltész Jan 26 '17 at 16:40

For each specific $k$ we have $$\binom{2k}k\binom{n+1}k\frac{k-1}{2^k\binom{4n}k}\to \binom{2k}k\frac{k-1}{8^k},$$ that suggests that the sum tends to $$\sum_{k=2}^\infty \binom{2k}k\frac{k-1}{8^k}=1-\frac{1}{\sqrt 2},$$ as $$\sum_{k= 0}^\infty \binom{2k}k x^k =(1-4x)^{-1/2};\,\\\sum_{k= 0}^\infty k\binom{2k}k x^k =(x\frac{d}{dx})(1-4x)^{-1/2}=2x(1-4x)^{-3/2};\\ \sum_{k= 0}^\infty (k-1)\binom{2k}k x^k =(6x-1)(1-4x)^{-3/2}=|_{x=1/8}-\frac1{\sqrt{2}}.$$

In order to justify this limit changes (limit of a sum is a sum of limits) we majorate each term in our sum. Say, $$\frac{\binom{n+1}k}{\binom{4n}k}=\frac{(n+1)n(n-1)\dots(n-k+2)}{4n(4n-1)(4n-2)\dots (4n-k+1)}<4^{-k}(1+\frac{1}n)(1+\frac{1}{4n-1})<3\cdot 4^{-k}.$$ this allows to estimate the total sum of summands with $k\geqslant k_0$ by $3\sum_{k\geqslant k_0} \binom{2k}k\frac{k-1}{8^k}$, that tends to 0 for large $k_0$. Therefore the limit of your sum is indeed a sum of limits, moreover, the estimates are very specific and should prove that $0<a_n<1$ also after checking some small values of $n$.


Not an answer. First, Mathematica says that $$a_n = \frac{-4 n \, _2F_1\left(\frac{1}{2},-n-1;-4 n;2\right)+n \, _2F_1\left(\frac{3}{2},-n;1-4 n;2\right)+\, _2F_1\left(\frac{3}{2},-n;1-4 n;2\right)}{4 n}.$$

Secondly, for all reasonable values of $n$ it seems that $a_n < -0.7,$ and, indeed, I conjecture that $$\lim_{n\rightarrow \infty} a_n = -\frac{1}{\sqrt{2}}.$$

Now, this is a response to the original form of this question, with the sum going from $0.$ The first two terms contribute $-1$ to the sum, so for the current version, the conjectured limit is $$1-\frac{1}{\sqrt{2}}.$$

  • $\begingroup$ Do you mind altering your estimate to start at $k=2$? Sorry for the change made. If not, you can redefine and call it $b_n$ (in your solution) to start at $k=0$? $\endgroup$ – T. Amdeberhan Jan 24 '17 at 4:54
  • $\begingroup$ @T.Amdeberhan Well, $k=1$ contributes $0,$ while $k=0$ contributes $-1...$ $\endgroup$ – Igor Rivin Jan 24 '17 at 4:55
  • $\begingroup$ I know, this is just to avoid any confusion by potential readers. $\endgroup$ – T. Amdeberhan Jan 24 '17 at 4:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.