0
$\begingroup$

I'm a rather inexperienced researcher, I've been stuck on a question for a while. I would like to find the largest $N = f(n)$ that satisfies the following inequality:

$$\sum_{j = 0} ^ n p^{n - j} (-1)^j \binom{N}{j} + \sum_{j = n + 1} ^ N (-1)^j \binom{N}{j} > 0$$ for all sufficiently large integers $n$, where $p$ is a fixed prime, say $p = 3$. I suspect that $N$ will be some function close to $n$, such as $n + logn$ or $(1 + \epsilon)n$, but I don't know how to proceed with this problem and I have tried all naive methods.

Notably, if $N = n$ exactly then this is sum is $(p - 1)^n > 0$ exactly by the binomial theorem. Any solutions or tips are appreciated.

$\endgroup$
2
  • $\begingroup$ Can we just take $p=1/2$? $\endgroup$ Commented Jan 29, 2024 at 3:11
  • $\begingroup$ Apologies, I just edited. Assume p> 1 is a prime. $\endgroup$ Commented Jan 29, 2024 at 3:16

2 Answers 2

1
$\begingroup$

For some $p$ and $n$, such largest $N$ does not have to exist.

Indeed, the left-hand side of your inequality equals $$\begin{aligned}L_p(n,N) &:=\sum_{j=0}^n p^{n-j}(-1)^j \binom Nj +\sum_{j=n+1}^N (-1)^j \binom Nj \\ &=\sum_{j=0}^n p^{n-j}(-1)^j \binom Nj +\sum_{j=0}^N (-1)^j \binom Nj-\sum_{j=0}^n (-1)^j \binom Nj \\ &=\sum_{j=0}^n p^{n-j}(-1)^j \binom Nj +(1-1)^N-\sum_{j=0}^n (-1)^j \binom Nj \\ &=\sum_{j=0}^n p^{n-j}(-1)^j \binom Nj -\sum_{j=0}^n (-1)^j \binom Nj \\ &=\sum_{j=0}^n (p^{n-j}-1)(-1)^j \binom Nj . \end{aligned}\tag{1}\label{1}$$

So, if e.g. $p=2$ and $n=5$, then $$L_p(n,N)=L_2(5,N)=\frac{1}{24} \left(N^4-18 N^3+131 N^2-474 N+744\right)>0$$ for all $N$.

More generally, it follows from \eqref{1} that $L_p(n,N)$ is a polynomial in $N$ of degree $n-1$, with the leading coefficient $(-1)^{n-1}\frac{p-1}{(n-1)!}$. So, the set of the values of $N$ for which $L_p(n,N)>0$ is unbounded for each odd $n$.


Remark: Using the identity $\binom Nj=\binom{N-1}j+\binom{N-1}{j-1}$, one gets the recursion $$L_p(n,N)=L_p(n,N-1)-L_p(n-1,N-1),$$ with $L_p(0,N)=0$ and $L_p(n,0)=p^n-1$.

$\endgroup$
6
  • 1
    $\begingroup$ Thanks for the response, I am actually looking for a N such that the left hand side is greater than 0 for all sufficiently large integers n. In fact, I am interested in the asymptotics where N is a fixed function and n tends toward infinity. $\endgroup$ Commented Jan 29, 2024 at 4:22
  • $\begingroup$ @JasonZheng : As I understood, you now want to have a function $f$ such that $L_p(n,N)\le0$ for all large enough $n$ and all $N>f(n)$, right? It is shown in the answer that such a function $f$ does not exist. Indeed, if you take any, arbitrarily large odd $n$, then the set of the values of $N$ for which $L_p(n,N)>0$ will be unbounded. Let me know if this seems unclear. $\endgroup$ Commented Jan 29, 2024 at 4:41
  • $\begingroup$ @iosofPinelis Sorry, but how did you obtain the identity (1)? I am not understanding how the expression L_p(n,N) is equal to the left side of my inequality. $\endgroup$ Commented Jan 29, 2024 at 5:05
  • $\begingroup$ @JasonZheng It is because $\sum_{j=0}^N \binom Nj (-1)^j=0$. Binomial expansion for $(1-1)^N$. $\endgroup$ Commented Jan 29, 2024 at 5:29
  • $\begingroup$ @JasonZheng : Brendan McKay's comment explains how (1) was obtained. Is this detailed enough for you? Please let me know. $\endgroup$ Commented Jan 29, 2024 at 16:13
0
$\begingroup$

Since @Iosif Pinelis already gave the answer, this is too long for a comment.

Formula $(1)$ could be obtained simplifying for $N>1$ $$\left(\frac{p-1}{p}\right)^N\, p^n-\frac {(-1)^n }{N p}\binom{N}{n+1}\Bigg((n+1)p -N \, _2F_1\left(1,n-N+1;n+2;\frac{1}{p}\right)\Bigg)$$ where

$$ \, _2F_1\left(1,n-N+1;n+2;\frac{1}{p}\right)=\frac{(n+1)!}{(n-N)! }\,\sum_{k=0}^\infty \frac{(k+n-N)!}{(k+n+1)!}\, p^ {-n}$$

As far as I could see it, the problem is related to the parity of $n$. For example

$$L_3(6,N)=-\frac{1}{60} \left(N^5-30 N^4+415 N^3-3450 N^2+17584 N-43680\right)$$ is positive if $N \leq 9$ $$L_3(7,N)=\frac{1}{360}\left(N^6-39 N^5+715 N^4-8205 N^3+63724 N^2-318276 N+786960\right)$$ is always positive.

$\endgroup$
3
  • $\begingroup$ To get (1), I was indeed reasoning as in Brendan McKay's comment. $\endgroup$ Commented Jan 29, 2024 at 16:06
  • $\begingroup$ @Iosif Pinelis. I just wanted to see if I would be able to simplify my nasty expression and get your nice one. $\endgroup$ Commented Jan 30, 2024 at 2:24
  • $\begingroup$ I see; thank you for your response. $\endgroup$ Commented Jan 30, 2024 at 3:31

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.