10.12.2012 Views

Prime Numbers

Prime Numbers

Prime Numbers

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

1.5 Exercises 69<br />

1.64. Perform computations that connect the distribution of primes with the<br />

Riemann critical zeros by way of the ψ function defined in (1.22). Starting<br />

with the classical exact relation (1.23), obtain a numerical table of the first 2K<br />

critical zeros (K of them having positive imaginary part), and evaluate the<br />

resulting numerical approximation to ψ(x) for, say, noninteger x ∈ (2, 1000).<br />

As a check on your computations, you should find, for K = 200 zeros and<br />

denoting by ψ (K) the approximation obtained via said 2K zeros, the amazing<br />

fact that <br />

ψ(x) (200) <br />

− ψ (x) < 5<br />

throughout the possible x values. This means—heuristically speaking—<br />

that the first 200 critical zeros and their conjugates determine the prime<br />

occurrences in (2, 1000) “up to a handful,” if you will. Furthermore, a plot of<br />

the error vs. x is nicely noisy around zero, so the approximation is quite good<br />

in some sense of average. Try to answer this question: For a given range on x,<br />

about how many critical zeros are required to effect an approximation as good<br />

as |ψ − ψ (K) | < 1 across the entire range? And here is another computational<br />

question: How numerically good is the approximation (based on the Riemann<br />

hypothesis)<br />

ψ(x) =x +2 √ x <br />

t<br />

sin(t ln x)<br />

t<br />

+ O √ x ,<br />

with t running over the imaginary parts of the critical zeros [Ellison and<br />

Ellison 1985]? For an analogous analytic approach to actual prime-counting,<br />

see Section 3.7 and especially Exercise 3.50.<br />

1.65. This, like Exercise 1.64, also requires a database of critical zeros of the<br />

Riemann zeta function. There exist some useful tests of any computational<br />

scheme attendant on the critical line, and here is one such test. It is a<br />

consequence of the Riemann hypothesis that we would have an exact relation<br />

(see [Bach and Shallit 1996, p. 214])<br />

<br />

ρ<br />

1<br />

=2+γ − ln(4π),<br />

|ρ| 2<br />

where ρ runs over all the zeros on the critical line. Verify this relation<br />

numerically, to as much accuracy as possible, by:<br />

(1) Performing the sum for all zeros ρ =1/2+it for |t| ≤T , some T of choice.<br />

(2) Performing such a sum for |t| ≤ T but appending an estimate of<br />

the remaining, infinite, tail of the sum, using known formulae for the<br />

approximate distribution of zeros [Edwards 1974], [Titchmarsh 1986], [Ivić<br />

1985].<br />

Note in this connection Exercises 1.61 (for actual calculation of ζ values)<br />

and 8.34 (for more computations relating to the Riemann hypothesis).<br />

1.66. There are attractive analyses possible for some of the simpler<br />

exponential sums. Often enough, estimates—particularly upper bounds—on

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!