Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Riemann Hypothesis and Primes

  1. Jun 25, 2005 #1
    Hi,

    When I hear about the Riemann hypothesis, it seems like the first thing I hear about it is its importance to the distribution of prime numbers. However, looking online this seems to be a very difficult thing to explain. I understand that the Riemann Hypothesis asserts that the zeroes of the zeta function all have real part equal to 1/2, but how is this related to prime numbers? I've looked online and the explanations seem very confusing.

    Thank you.
     
  2. jcsd
  3. Jun 25, 2005 #2

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    google the internet or search these very forums for the answer to this oft asked question
     
  4. Jul 1, 2005 #3
    Basically Riemann gave a formula for the prime counting function that includes a sum over all zeros of the zeta function (well, not exactly, it's actually a sum of x to the power of all zeros of the zeta function), and if all of the zeros lie on the critical line than we can get a good estimate of the error between Li(x) and the prime counting function.
     
  5. Jul 24, 2005 #4
    wouldn,t be easier to solve the interal equation for Pi(x) (i propesed a formula for solving this integral equation) to get the value of Pi(x)?
     
  6. Jul 24, 2005 #5

    Hurkyl

    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    Wolfram seems to think that the Lagarias-Odlyzko algorithm is the best to use: that's what's implemented in Mathematica.
     
  7. Jul 24, 2005 #6

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    ui think the answer is almost certainly "no" still, cos you've never proven otherwise. but riemann stated something abuot the estimate of the difference between the *estiamte* via Li(x) compared to Pi(x). You understand Li(x) is an asymptotic estimate, right?
     
  8. Jul 25, 2005 #7
    but as far as i know riemann hypothesis being true you stil have the error term that would go as O(x^{0.5}log(x)) so the error diverges, another question is supposed that the formula:

    Pi(x)=Li(x)+O(x^{0.5}log(x)) is an asymptotic stimation isn,t it?...

    so i don,t see why is better Riemann,s formula for Pi(x) rather than an exact method to calculate it,the approximation of Riemann would only be valid to compute Pi(x) for big x and don,t even that because we don,t know what the function is exactly...

    Hurkyl..wher could i find the lagarias Odlyzko algorithm for calculating Pi(x)?
     
    Last edited: Jul 25, 2005
  9. Jul 25, 2005 #8

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    no one has ever claimed that Li(x) tends to Pi(x), waht is the point of your question?
     
  10. Jul 25, 2005 #9
    didn't Hardy and Littlewood prove that Li(x) occilates above and below Pi(x) an infinate number of times?
     
  11. Jul 26, 2005 #10
    i got by solving the integral equation for Pi(x) a solution in the form:

    [tex]\frac{1}{2\pi}\int_{-\infty}^{\infty}t^{-iw}F(w)/G(w)dw [/tex] with

    [tex]F(w)=\int_{-\infty}^ {\infty}i(2+is)^{-iw-1}Ln\zeta(2+2s) [/tex]

    [tex]G(w)=\int_{-\infty}^ {2}e^{r}e^{ir}/exp(exp(ir))-1 [/tex]

    so here you are a true function for PI(x),i have never understood mathematician and their "mania" to express always the formulae as f(x)=g(x)+O(h(x)) this does not prove anything, as there is no real way to calculate f(x) from the former expression,don,t even an approximation you only get that f(x)-g(x)<ch(x) this is useless..(at least in my opinion) unless for big x h(x) tends to a constant a and with that you get only that for big x f(x)-g(x)<ca...

    Note:there is a mistake the upper limit in G(w) is 2 not oo
     
    Last edited: Jul 26, 2005
  12. Jul 26, 2005 #11
    sorry it was:

    [tex]F(w)=\int_{-\infty}^ {\infty}i(2+is)^{-iw-1}Ln\zeta(2+is) [/tex]

    [tex]G(w)=\int_{-\infty}^ {2}e^{r}e^{ir}/exp(exp(ir))-1 [/tex]


    [tex]\frac{1}{2\pi}\int_{-\infty}^ {\infty}t^{-iw}F(w)/G(w)=\pi(e^t) [/tex]

    with that you have an exact integral formula for solving it.......
     
  13. Jul 26, 2005 #12

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    We do not have a mania (so far mathematicians are snobs and manics; i think we have shown you considerable patience and indulgence given your less than complimentary attitude). We do care about knowing the error in approximation via some method (something you could do with learning about). Again you appear to think that mathematicians are of the opinion that Li(x) is used to get Pi(x). It is a nice result, one of significant importance, and one that took many years (100) to prove from the original conjecture of Gauss (he made the conjecture when he was 15) based upon observations of tables of data. It is one of the high points of Analytic Number Theory. And as LeVeque says, the proof is too hard to include here.

    Any time you actually want to "solve" those integrals go right ahead. I for one have a much simple method for "solving" Pi(x): let s(x)=1 if x is prime and 0 otherwise, then

    [tex]\text{pi}(x)= \sum_{2}^{\floor{x}}s(x)[/tex]

    very simple and "solves" pi(x)....
     
    Last edited: Jul 26, 2005
  14. Jul 26, 2005 #13

    Zurtex

    User Avatar
    Science Advisor
    Homework Helper

    Here is my favorite representation:

    [tex]\pi(n) = -1 + \sum_{j=3}^n \left((j-2)! - j \Biggl{\lfloor} \frac{(j-2)!}{j} \Biggr{\rfloor}\right) \quad \text{for} \, \, n > 3[/tex]

    Not exactly efficient, but quite pretty :smile:
     
  15. Jul 26, 2005 #14
    Matt and Zurtex: the problem you have with your approach is that either you don,t know what s(x) is (the analytic form) or it,s a sum hard to calculate (i don,t know why they published the formula zurtex gave but not mine :( ).

    Another question is not RH proved (riemann hypothesis) as far as you know from the functional equation: [tex]\zeta(1-s)=g(s)\zeta(s) [/tex] putting s=1/2+it you get:

    [tex]\zeta(1/2-it)=g(s)\zeta(1/2+it) [/tex] from that would,nt you get RH?...as you know that Zeroes are symmetrically placed
     
  16. Jul 26, 2005 #15

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    and your integrals are easy to calculate are they? i haven't seen you do one yet. you do understand that we were being sarcastic? the formula zurtex gave, if it were ever published, would have been published because it was a new idea at the time. It may also have been one small result in one paper. yuor idea is not new, the method is not new, the result is not new, and it is very small. it may also have been published because its proof was of sigificant importance for a variety of reasons.

    let me give you another example:
    let f be any function, then we can work out f(x) via some transform....

    that is all you have done, ok? that is not interesting. we know that many functions can be evaluated through inverse functions. you havwe not defined a new function or a new transform or a new way of evaluating anything! and you have royally pissed off anyone who might have wnated to help you make sense of your work.

    and now you have a one line proof of the riemann hypothesis (yes i am tired of trying to be nice). and no, the RH does not follow from you observation. (g(s) being what? and why may that not be zero? by your logic ALL zeroes lie on the line Re(s) = 1/2 and we know there are ones that do not lie on the line)
     
    Last edited: Jul 26, 2005
  17. Jul 27, 2005 #16
    you needn,t be sarcastic or unpolite,i have a proof for getting Pi(x) choose the method you want to solve the integral (yes you wll have a series but a finite series simpler than the series Zurtex gave) i give the formula for Pi(x) by means of a fourier transform and you can use the algorithm of FFT (fast fourier transform) to solve it so you can improve the time to solve the integrals, i gave an integral representation for Pi(x) and this is new.

    In the formula R(1-s)=g(s)R(s) g(s) is a formula involving Gama function and pi^x and 2^{1-x} i,ve never said i had a proof of Riemann Hypothesis,i only gave a reasoning to prove it...
     
  18. Jul 27, 2005 #17

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    you have given *a* different formula for pi(x). there are many. and an integral representation for it is not new as has been pointed out to you with a refereve to a paper by Ozlydsko. no one disputes what you have we are attempting to explain to you how it fits into the lagrer scheme of mathematics.
     
  19. Jul 28, 2005 #18

    shmoe

    User Avatar
    Science Advisor
    Homework Helper

    No, all this shows (after examining the g term) is that if you have a zero at 1/2+it then you have a zero at 1/2-it (and vice versa), it does not guarantee that any non-trivial zero must be of this form. This is a trivial and well known fact. Actually more symmetry between the zeros is known-reflection principle plus the functional equation tells you non-trivial ones come in sets of 4 symmetric about s=1/2 (collapses to 2 zeros where the zero is on the critical line). This is why they only look for zeros in the upper half of the critical strip, the part below the real axis comes for "free".
     
  20. Jul 29, 2005 #19
    and what about studying the integral:

    [tex]\int_0^{\infty}\int_{c-i\infty}^{c+i\infty}\frac{x^{s-1-z}}{\zeta(s)s}dsdx[/tex]

    the singularities (the values when is infinite) would be the place where zeros are
     
  21. Jul 29, 2005 #20

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    that doesn't make sense, does it?
     
  22. Jul 29, 2005 #21
    why not you can check this integral is equal to 1/R(s) so when the integral is infinite so will be the value of 1/R(s)

    another question: i have checked several webpages about Lagarias-odlyzko method for pi(x) and does not appear an integral by any side is only a small modification of Lehmer,s method to calculate Pi(x) and this implies calculatig a sum not an integral....
     
    Last edited: Jul 29, 2005
  23. Jul 29, 2005 #22

    shmoe

    User Avatar
    Science Advisor
    Homework Helper

    I really hope you aren't trying to imply again that an algorithm for computing pi by analytic methods doesn't currently exist. I've told you to look at "Computing pi(x):An Analytic Method", by Lagarias and Odlyzko. If you can't be bothered to actually do this and see how it differs from variants of the Meissel-Lehmer method (Odlyzko and Lagarias have written about the Meissel-Lehmer method as well) there's really no point at all in me wasting my time replying to you.


    I just noticed this:

    It was Littlewood who did this. There's an upper bound for where the first sign change is guaranteed to happen. I can't recall what the current best upper bound is, but it's still well out of computational reach (the first was by Skewes and was something like 10^34 digits, this has been greatly improved though).
     
  24. Jul 30, 2005 #23

    matt grime

    User Avatar
    Science Advisor
    Homework Helper

    i merely meant are you sure you're using valid expansions and relations for the zeta function with correct convergence issues.
     
  25. Aug 1, 2005 #24
  26. Aug 1, 2005 #25

    shmoe

    User Avatar
    Science Advisor
    Homework Helper

    De Branges method is known to be flawed. Conrey and Li published a counter example years ago.

    In your first link, he's attempting to claim that the real and imaginary parts of zeta(s) and zeta(1-s) cannot be equal (and hence both zero) if real part of s is not 1/2 by looking at their respective Dirichlet series (which don't quite give zeta anyways, this is the alternating guy from Hardy, theres a factor missing). He concludes that u and u' must not be equal because...? They 'look' different? This is actually false anyways, for s in the critical strip off the critical line it is false to claim the real parts of zeta(s) and zeta(1-s) cannot be equal (likewise for the imaginary parts).

    One of my favorites is in your second link, by Pradas, which can prove RH on the very false assumption that Zeta is one to one. I've looked at a couple others on that list and have not found anything worthwhile. I'm going to hazard a guess that time will tell all on that page are false. It's likely that all have already been checked by mathematicians and found false. Don't take silence (in some of the cases) to mean that mathematicians are stunned into shame-it's more likely it's not worth the time to comment.
     
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook