On Wed, 15 Feb 2012 17:52:08 -0800 (PST), RichD
<
r_dela...@yahoo.com> wrote:
>I was reading a piece on the Rieman hypothesis,
>and that 300 million zeta function zeros have been
>computed.
Is that what it said, or did it just say that the RH had been
verified for the first 300 million zeroes? It makes a big
difference!
I don't know how people actually do these computations.
But I can explain how at least theoretically it's possible
to know that a zero of the zeta function lies _exactly_
on the line s = 1/2, in spite of roundoff errors in
the calculations:
1. Say f is an analytic function and c is a closed curve.
You can verify numerically that f has no zero _on_ c
be first getting a bound of the derivative of f, then
calculating f at a finite number of points to suffiicient
precision.
A simpler example of how that might go: Suppose
you have f defined on [0,1]. Suppose you know
that |f'(t)| <= C everywhere, where C is _approximately_ 1.
Suppose you know that f(0) is _approximately_ 5.
Then (assuming those approximations are good enough)
you know that f has no zero on [0,1], because you
know that |f(t) - f(0)| >= |f(0)| - Ct >= 4, approximately.
2. Magic: Given an analytic function f and a closed
curve c, such that f has no zero on c (which you can
know in spite of roundoff error - see (1) above),
there is a certain integral that gives exatly the
number of zeroes of f inside c. So you can calculate
the number of zeroes inside c exactly in spite of
roundoff error, because the number is an integer;
if there are 2 zeroes inside c, with an error less than 1/2,
then there are exactly 2 zeroes inside c.
3. More magic: The zeta function satisfies certain
identites, showing that if there was a zero in
the critical strip but not on the critical line, then
there would be another zero nearby. SO: You
calculate that there is exactly one zero inside
a certain tiny circle (see (2)) and it follows that that
zero is _exactly_ on the line s = 1/2.