On Sunday 01 October 2006 18:28, Soo Reams wrote:
> Brian Beesley wrote:
> >
> > Doesn't this depend on the (AFAIK unproved) Riemann hypothesis?
>
> Not certain of this, but I think Pi(x) is defined as the number of
> primes less than or equal to x. The Riemann hypothesis is concerned with
> a particular approximation to Pi(x), a more accurate one than Pi(x) = x
> / lg x.

Isn't it _perfectly_ accurate _unless_ RH is _false_?

The problem being that the computation of pi(x) involves evaluating an 
infinite series of terms which converge rather slowly (especially for large 
x), so that any computed value is indeed an approximation. If we had a 
complete table of the zeros of the zeta function we could do this computation 
quickly and accurately; but then the existence of such a table would render 
the truth or otherwise of the Riemann hypothesis moot.
> >
> > This is where the possible interest is. If Allen's algorithm turns out to
> > be genuinely different, proveably correct and returns the same results as
> > Pi(x) then there may be a handle on proving Riemann. Or even the extended
> > Riemann hypothesis!
>
> Hmm, unlikely. Allen's algorithm sounds like it's at least as
> computationally intensive as finding the primes individually.

So what? There are two points here:

(1) Advance in mathematics often comes from linking seemingly unrelated 
fields; and, whilst an infinite amount of computation may still not be 
sufficient to generate a proof, many types of proof depend on a strictly 
finite (and sometimes surprisingly small) amount of computation. The trick is 
in finding the right combination of already proved results, manipulating 
these to derive something "new". A new algorithm for pi(x) might (but doesn't 
necessarily) result in a new basis for proving RH and/or a new method of 
manipulating known results to derive a proof.

(2) A new method of computing pi(x) _may_ lead to a new method of computing 
the zeros of zeta(s) which _may_ result in the finding of at least one zero 
which isn't on the critical line. If this helps you to think about this, 
AFAIK no-one has yet managed to prove even that there at most a finite number 
of zeros off the critical line. Suppose there are two (or more) classes of 
zero, one (or more) on the critical line and one (or more) off it. So happens 
that the methods we have at the moment only find zeros on the critical 
line ... we can calculate the "first" trillion or so zeros and find them all 
on the critical line, as we have already done, generating evidence of the 
truth of RH without constituting a proof of any sort.

I personally believe that RH is true - but this is only blind faith, which a 
demonstration by counterexample would shatter. 

Regards
Brian Beesley
_______________________________________________
Prime mailing list
[email protected]
http://hogranch.com/mailman/listinfo/prime

Reply via email to