1

I am now trying to find proof for the following, which are significant to establishing proof for the Prime number relation that was originally stated in the question I posted here:

$$\Bigl \lfloor \frac{n+1}{\lfloor \sqrt{n+1} \rfloor}\Bigr\rfloor-\Bigl \lfloor \frac{n}{\lfloor \sqrt{n} \rfloor}\Bigr\rfloor \in {\{-1,0,1}\} \quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad(A)$$

$$\frac{1}{2}\Biggl(\Biggl\lfloor \frac{(n+1)^2-1}{\Bigl\lfloor \sqrt{(n+1)^2-1} \Bigr\rfloor} \Biggr\rfloor-n\Biggr)-1=0\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad(B)$$

$$\Biggl\lfloor\frac{n+\bigl\lfloor(\frac{1}{2}n-\frac{1}{4n})^2\bigr\rfloor+1}{\bigl\lfloor \sqrt{n+\lfloor(\frac{1}{2}n-\frac{1}{4n})^2\rfloor+1} \bigr\rfloor}\Biggr\rfloor-\Biggl\lfloor\frac{n+\lfloor(\frac{1}{2}n-\frac{1}{4n})^2\rfloor}{\bigl\lfloor \sqrt{n+\bigl\lfloor(\frac{1}{2}n-\frac{1}{4n})^2\bigr\rfloor} \bigr\rfloor}\Biggr\rfloor-1=0\quad\quad\quad\quad(C)$$

$(B)$ and $(C)$ were conjectured from numerical observations of $(A)$

Adam Ledger
  • 1,240

1 Answers1

1

(A) We divide in two cases:

Case 1: $n=m^2-1$ for some $m\in\mathbb N$.

In this case, we have that $$ \left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor=\left\lfloor \dfrac{m^2}{\lfloor \sqrt{m^2} \rfloor} \right\rfloor-\left\lfloor \dfrac{m^2-1}{\lfloor \sqrt{m^2-1} \rfloor} \right\rfloor $$ and since $\lfloor \sqrt{m^2} \rfloor=m$ and $\lfloor \sqrt{m^2-1} \rfloor=m-1$, $$ \left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor=\left\lfloor \dfrac{m^2}{m} \right\rfloor-\left\lfloor \dfrac{m^2-1}{m-1} \right\rfloor = m-(m+1)=-1 $$

Case 2: $n$ is not of the form $m^2-1$.

Pick $m\in\mathbb N$ such that $m^2\leq n \leq (m+1)^2-1$. Since $n\neq (m+1)^2-1$, we have $m^2\leq n \leq (m+1)^2-2$, and consequently, $$ m^2\leq n,\ n+1<(m+1)^2\ \Rightarrow\ m\leq \sqrt n,\ \sqrt{n+1}<m+1\ \Rightarrow\ \lfloor\sqrt n\rfloor=\lfloor\sqrt{n+1}\rfloor=m. $$ Then $$ \left|\frac{n+1}{\lfloor \sqrt{n+1}\rfloor}-\frac{n}{\lfloor \sqrt n\rfloor}\right|=\left|\frac{n+1}{m}-\frac{n}{m}\right|=\frac{1}{m}\leq 1, $$ and this implies that $$ \left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor \in \{-1,0,1\}. $$

(B) We have that $\lfloor \sqrt{(n+1)^2-1}\rfloor=n$, because $$ n^2\leq(n+1)^2-1<(n+1)^2, $$ so $$ \frac{(n+1)^2-1}{\lfloor \sqrt{(n+1)^2-1}\rfloor} = \frac{n^2+2n}{n}=n+2, $$ from which follows the identity (B).

(C) First, we shall characterize the numbers such that $$ \left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor=1. $$ By the Case 1 at the proof of (A), we already know that $n$ has not the form $m^2-1$. Pick $m\in\mathbb N$ such that $m^2\leq n\leq (m+1)^2-1$. By the Case 2 at the proof of (A), we know that $\lfloor\sqrt n\rfloor=\lfloor\sqrt{n+1}\rfloor=m$.

Put $n= m^2+k$, with $0\leq k \leq 2m$. Then $$ \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} = m +\frac{k+1}{m}, $$ and $$ \dfrac{n}{\lfloor \sqrt{n} \rfloor} = m +\frac{k}{m}. $$ So, $$ \left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor = \left\lfloor m+\dfrac{k+1}{m} \right\rfloor-\left\lfloor m+\dfrac{k}{m} \right\rfloor = \left\lfloor \dfrac{k+1}{m} \right\rfloor-\left\lfloor \dfrac{k}{m} \right\rfloor, $$ and the cases that this equals $1$ are precisely when $k=m-1$ and $k=2m-1$.

This way we proved that the only numbers $n$ that satisfiy $$ \left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor=1, $$ Are precisely the $n$'s of the form $m^2+m-1$ or $m^2+2m-1$, for some natural $m$.

Now it only lasts to verify that the numbers $n+\left\lfloor\left(\dfrac{n}{2}-\dfrac{1}{4n}\right)^2\right\rfloor$ have one of these forms.

Let us do it. First, note that $$ \left(n-\frac{1}{2n}\right)^2=n^2-1+\frac{1}{4n^2}, $$ and since $0<\dfrac {1}{4n^2}<1$, $$ n^2-1<\left(n-\frac{1}{2n}\right)^2< n^2. $$ We divide in two cases:

Case 1. $n=2k$.

Then $$ \begin{array}{rcl} 4k^2-1<\left(n-\frac{1}{2n}\right)^2<4k^2 & \Rightarrow & k^2-1<k^2-\frac{1}{4}<\frac{1}{4}\left(n-\frac{1}{2n}\right)^2 < k^2 \\ & \Rightarrow & k^2-1<\left(\frac{n}{2}-\frac{1}{4n}\right)^2< k^2 \\ & \Rightarrow & \left\lfloor\left(\frac{n}{2}-\frac{1}{4n}\right)^2\right\rfloor = k^2-1 \\ & \Rightarrow & n+ \left\lfloor\left(\frac{n}{2}-\frac{1}{4n}\right)^2\right\rfloor = n + k^2 - 1 = k^2 + 2k -1. \end{array} $$ Case 2. $n=2k-1$.

Then $$ \begin{array}{rcl} 4k^2-4k<\left(n-\frac{1}{2n}\right)^2<4k^2 -4k+1 & \Rightarrow & k^2-k<\frac{1}{4}\left(n-\frac{1}{2n}\right)^2 < k^2 -k +\frac{1}{4} < k^2 -k +1\\ & \Rightarrow & k^2-k<\left(\frac{n}{2}-\frac{1}{4n}\right)^2< k^2 -k+1\\ & \Rightarrow & \left\lfloor\left(\frac{n}{2}-\frac{1}{4n}\right)^2\right\rfloor = k^2-k\\ & \Rightarrow & n+ \left\lfloor\left(\frac{n}{2}-\frac{1}{4n}\right)^2\right\rfloor = n + k^2 - k = k^2 +k-1. \end{array} $$

André Porto
  • 1,855
  • haha well sure looks good so far wouldn't mind the proof for $\lfloor \sqrt{n+1}\rfloor = \lfloor \sqrt n\rfloor$ for $n$ not of the form $m^2-1$ while you are there anyway but sure I don't think I've got one for $\lfloor \sqrt{m^2-1} \rfloor=m-1$ either – Adam Ledger Oct 27 '18 at 23:13
  • Well I probably have one somewhere but still the one you've got after dinner would be a winner – Adam Ledger Oct 27 '18 at 23:18
  • Well, pick $m\in\mathbb N$ such that $m^2\leq n <(m+1)^2$. Then $m^2\leq n \leq(m+1)^2-1$. But since $n\neq (m+1)^2-1$, we have that $m^2\leq n \leq (m+1)^2-2$. Thus, both $n, n+1$ are numbers between $m^2$ and $(m+1)^2-1$... – André Porto Oct 27 '18 at 23:34
  • sure looks good well with $(C)$ I got as far as to find that it's because $n(n-1)+1$ and $n$ are always coprime for $n \gt 1$, but it could probably be done in the same manner you've done $(B)$, only that it's the case that $\left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor=1$ that implies $(C)$ , as how $\left\lfloor \dfrac{n+1}{\lfloor \sqrt{n+1} \rfloor} \right\rfloor-\left\lfloor \dfrac{n}{\lfloor \sqrt{n} \rfloor} \right\rfloor=-1$ implies $(B)$ – Adam Ledger Oct 27 '18 at 23:48
  • 1
    I incorporated the proof of the fact that $\lfloor\sqrt{n}\rfloor=\lfloor\sqrt{n+1}\rfloor$ – André Porto Oct 28 '18 at 00:04
  • Now there's a characterization of (C) that hopefully will help you to prove it. – André Porto Oct 28 '18 at 00:42
  • oh right I was thinking about that you mean the one I've put on this page? https://math.stackexchange.com/questions/2645263/on-the-conjecture-that-for-every-n-lfloor-e-fracp-n2-p-n2-1/2973172#2973172 I basically started and decided 5 minutes later I need to do something else but still it's relevant perhaps – Adam Ledger Oct 28 '18 at 00:47
  • oh $(C)$ as in this page I thought you meant $\mathbb C$ my bad – Adam Ledger Oct 28 '18 at 00:48
  • Ah right I see what you mean now with you last few lines that you've added.

    I'm thinking rearrangement to $\frac{(2n^2-1)^2}{16n^2}$ and a determination of their greatest common divisor ought to be the way to go I think I mean for $n \gt 2$ you are assured the floor nor the fractional part is ever zero, so surely this would allow us to make assertions that can allow the removal of the floor function in place of some integer multiple of one of the two forms you mentioned

    – Adam Ledger Oct 28 '18 at 01:54
  • For example a characteristic of the arithmetic function we are dealing with here is such that:

    $f(n)=\left( \frac{1}{2}n-\frac{1}{4n}\right) ^{2}$

    we have:

    $\frac{1}{f(2n+1)}-\Bigl\lfloor \frac{1}{f(2n+1)} \Bigr\rfloor=f(n+2)$

    – Adam Ledger Oct 28 '18 at 02:06
  • It's disappointing when it turns out a problem involves a whole lot of mucking about exploiting particular things about the problem in specific that don't make use of any of the more powerful of ideas, but yes sadly we have to grind through them I guess and hope that next time it's a lot cooler than a bunch of migraines – Adam Ledger Oct 28 '18 at 02:12
  • Essentially the direction I was going with this was to attempt to apply Legendre's postulate regarding the existence of a prime between $n^2$ and $(n+1)^2$ as a means of proof for the stuff I've written in the question I've linked to this one, I really hate the amount of times I've changed direction with this one but you will immediately get what I mean if you take a look – Adam Ledger Oct 28 '18 at 02:20
  • Yes, the arguments I've used are too specific. It would be nice if we could derive those from just one property. – André Porto Oct 28 '18 at 13:14
  • 1
    So, just for completeness, I finished the proof of (C) at the answer. I hope it was useful. – André Porto Oct 28 '18 at 13:47
  • sorry I just need to correct what I stated before to $ \left( f \left( n+2 \right) \right) ^{2}-f \left( n+2 \right) \lfloor f \left( n+2 \right) \rfloor -1=0 $ and then I will have to come back and read you final take.

    Thanks very much for this by the way I do appreciate it immensely

    – Adam Ledger Oct 28 '18 at 13:50
  • Absolutely very useful indeed – Adam Ledger Oct 28 '18 at 13:51
  • Porto I am just studying your solutions now and wanted to let you know I've come across three identities Ramanujuan published, the proofs for which are in this book which is paywalled atm, but your approach should still be the most valid for these tyoes – Adam Ledger Nov 12 '18 at 22:14
  • They are listed under solved problems on the Wikipedia page too actually – Adam Ledger Nov 12 '18 at 22:15
  • https://en.wikipedia.org/wiki/Floor_and_ceiling_functions – Adam Ledger Nov 12 '18 at 22:15