Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Two questions came to mind when I was reading the proof for Bertrand's Postulate (there's always a prime between $n$ and $2n$):

(1) Can we change the proof somehow to show that: $\forall x > x_{0}$, there exists a prime $p$ $\in [x, ax]$, for some $a \in (1, 2)$?

(2) Suppose the (1) is true, what is the smallest value of $x_{0}$?

I'm not sure how to prove either of them, any input would be greatly appreciated! And correct me if any of the above statement is wrong. Thank you!

share|improve this question
    
I gave an answer here somewhere that talked about some of the improvements on Bertrand's postulate. Let's see if I can find it... –  mixedmath Sep 21 '12 at 19:15
1  
I must have remembered something longer. Here was that other answer. –  mixedmath Sep 21 '12 at 19:17
    
I'm not entirely certain how to parse statement (1). Are you asking whether $$(\exists x_0\in\mathbb{N})(\forall x\in\mathbb{N}, x>x_0)(\exists\mbox{prime }p)(\exists a\in(1,2))(p\in[x,ax])$$ (a rather trivial result) or whether $$(\forall a\in(1,2))(\exists x_0\in\mathbb{N})(\forall x\in\mathbb{N}, x>x_0)(\exists\mbox{prime }p)(p\in[x,ax])$$ or perhaps something else (perhaps changing the $\forall a$ in the last statement to $\exists a$)? –  jwodder Sep 21 '12 at 19:33
    
@jwodder: thanks for the correction. I believe I meant "for some a between 1 and 2". –  Lindsay Duran Sep 21 '12 at 19:40
    
@LindsayDuran: You still need to specify where in the sequence of quantifiers that $\exists a\in(1,2)$ appears. –  jwodder Sep 21 '12 at 19:41

2 Answers 2

up vote 5 down vote accepted

There are better results than Bertrand's Postulate. Pierre Dusart has proven better results, some of which can be found on wikipedia.

Less specifically, for any $k > 1$, $k \in \mathbb{R}$, one can show that $\lim_{n \to \infty} \dfrac{\pi(kn) - \pi(n)}{n/\ln n} = k - 1$ using the prime number theorem, which means that for any $k>1$, there is some $x_0$ such that for $x > x_0$, there is a prime between $x$ and $kx$.

share|improve this answer
    
Are you assuming that $k$ is an integer? –  M Turgeon Sep 21 '12 at 19:26
1  
@MTurgeon: No, $k$ can be any real number greater than $1$. –  mixedmath Sep 21 '12 at 19:29

I think you would enjoy the page PRIME GAPS.

My own version of the conjecture of Shanks, actually both a little stronger and a little weaker, is $$ p_{n+1} < p_n + 3 \,\left( \log p_n \right)^2, $$ for all primes $p_n \geq 2.$ This is true as high as has been checked.

Shanks conjectured that $$ \limsup \frac{p_{n+1} - p_n}{\left( \log p_n \right)^2} = 1, $$ while Granville later corrected the number $1$ on the right hand side to $2 e^{- \gamma} \approx 1.1229,$ see CRAMER GRANVILLE. There is no hope of proving this, but I enjoy knowing what seems likely as well as what little can be proved.

Here is a table from the third edition (2004) of Unsolved Problems in Number Theory by Richard K. Guy, in which $p = p_n$ is a prime but $n$ is not calculated, then $g = p_{n+1} - p_n,$ and $p = p(g),$ so $p_{n+1} = p + g.$

=-=-=-=-=-=-=-=-=-= enter image description here =-=-=-=-=-=-=-=-=-=

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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