mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Math

Reply
 
Thread Tools
Old 2011-07-24, 16:06   #1
wblipp
 
wblipp's Avatar
 
"William"
May 2003
New Haven

23×103 Posts
Default Prime Density Correction Terms?

Is there a known correction term to the density primes being 1/ln(x)? There are well known improvements to the nth prime being n*ln(n), but I think these are mostly improved approximations for the Log Integral (Li). What I'm asking about would be an improved integrand for the Log Integral.

I know this isn't stated in a mathematically rigorous manner, but I think the question is clear enough.
wblipp is offline   Reply With Quote
Old 2011-07-24, 20:23   #2
CRGreathouse
 
CRGreathouse's Avatar
 
Aug 2006

135338 Posts
Default

Quote:
Originally Posted by wblipp View Post
Is there a known correction term to the density primes being 1/ln(x)?
I'm pretty sure the answer is no. Better estimates would probably be stronger than RH.
CRGreathouse is offline   Reply With Quote
Old 2011-07-25, 11:43   #3
R.D. Silverman
 
R.D. Silverman's Avatar
 
Nov 2003

22×5×373 Posts
Default

Quote:
Originally Posted by CRGreathouse View Post
I'm pretty sure the answer is no. Better estimates would probably be stronger than RH.
Indeed. One needs to get a count of the number of primes in a
short interval. Cramer's conjecture asserts the existence of a prime
in the (short) interval x, x + O(log^2 x), but says nothing
about how many there are.
R.D. Silverman is offline   Reply With Quote
Old 2011-07-25, 20:42   #4
mart_r
 
mart_r's Avatar
 
Dec 2008
you know...around...

3·229 Posts
Default

I was on that too and came up with 1/ln(x+1/2*ln(x)*sqrt(x)) (ln being the natural logarithm here, as being used in the OP) which was met with about the same criticism as in this thread.
Sure, I understand why this is a moot point, but there are some numerical and heuristical indications...

Last fiddled with by mart_r on 2011-07-25 at 20:47
mart_r is offline   Reply With Quote
Old 2011-07-30, 01:34   #5
LiquidNitrogen
 
LiquidNitrogen's Avatar
 
Jun 2011
Henlopen Acres, Delaware

7·19 Posts
Default

Quote:
Originally Posted by wblipp View Post
Is there a known correction term to the density primes being 1/ln(x)?
pi(x) = x/[ln(x) - 1.08] was offered as one "improvement", very long time ago.
LiquidNitrogen is offline   Reply With Quote
Old 2011-07-30, 10:54   #6
Arkadiusz
 
Dec 2009

33 Posts
Default

Quote:
Originally Posted by LiquidNitrogen View Post
pi(x) = x/[ln(x) - 1.08] was offered as one "improvement", very long time ago.
http://oeis.org/A193257
Arkadiusz is offline   Reply With Quote
Old 2011-07-30, 16:37   #7
CRGreathouse
 
CRGreathouse's Avatar
 
Aug 2006

175B16 Posts
Default

Arkadiusz, I didn't realize you posted here! (Or lurked, as the case may be.)
CRGreathouse is offline   Reply With Quote
Old 2011-09-01, 03:59   #8
CRGreathouse
 
CRGreathouse's Avatar
 
Aug 2006

3×1,993 Posts
Default

Quote:
Originally Posted by mart_r View Post
I was on that too and came up with 1/ln(x+1/2*ln(x)*sqrt(x)) (ln being the natural logarithm here, as being used in the OP) which was met with about the same criticism as in this thread.
Sure, I understand why this is a moot point, but there are some numerical and heuristical indications...
This seems like arguing about whether sin x for large x is more like 0 or 1/x: it seems to miss the point that the osculating portion is large compared to the correction.

As an example, between 10^10 and 10^10 + 10^7 the standard 1/log x predicts 434,285 primes, while this predicts 2 fewer. But there are 434,650, so the actual errors are 365 and 367.

IIRC |pi(x) - li(x)| is known to be >> sqrt(x)/log x infinitely often, while under RH it's << sqrt(x) log x. So it's a priori possible that the correction could be meaningful, but I'm not holding my breath.
CRGreathouse is offline   Reply With Quote
Old 2011-09-02, 22:19   #9
mart_r
 
mart_r's Avatar
 
Dec 2008
you know...around...

3×229 Posts
Default

Just this question:
Does the Riemann prime counting formula / RH if true imply that \int (Li(x)-\pi(x)-\frac{\sqrt x}{\log x}) is - most of the time - closer to zero than \int (Li(x)-\pi(x))?
mart_r is offline   Reply With Quote
Old 2011-09-03, 00:09   #10
Zeta-Flux
 
Zeta-Flux's Avatar
 
May 2003

110000010112 Posts
Default

William,

I just noticed this question. I'll take a look at my files on this when I get back to work on Tuesday. (If you don't hear from me, remind me.)
Zeta-Flux is offline   Reply With Quote
Old 2011-09-03, 02:26   #11
CRGreathouse
 
CRGreathouse's Avatar
 
Aug 2006

10111010110112 Posts
Default

Quote:
Originally Posted by mart_r View Post
Just this question:
Does the Riemann prime counting formula / RH if true imply that \int (Li(x)-\pi(x)-\frac{\sqrt x}{\log x}) is - most of the time - closer to zero than \int (Li(x)-\pi(x))?
I don't think that's known to be true under the RH.
CRGreathouse is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
@ Supermods: Error Correction please Raman Forum Feedback 72 2013-06-22 07:24
Asymptotic density of k-almost primes CRGreathouse Math 1 2010-08-22 23:47
Prime k-value density rating and factorizations gd_barnes No Prime Left Behind 25 2009-07-30 12:06
Density of Mersenne divisors Random Poster Math 1 2008-12-15 01:14
prime density vs. sieving vs. prp time jasong Math 18 2006-03-31 03:14

All times are UTC. The time now is 07:36.


Wed Dec 8 07:36:54 UTC 2021 up 138 days, 2:05, 1 user, load averages: 1.15, 1.70, 1.67

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.