10.12.2012 Views

Prime Numbers

Prime Numbers

Prime Numbers

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

20 Chapter 1 PRIMES!<br />

It is not hard to see that Θ(n) for odd n is bounded below by a positive<br />

constant. This singular series can be given interesting alternative forms (see<br />

Exercise 1.68). Vinogradov’s effort is an example of analytic number theory<br />

par excellence (see Section 1.4.4 for a very brief overview of the core ideas).<br />

[Zinoviev 1997] shows that if one assumes the extended Riemann<br />

hypothesis (ERH) (Conjecture 1.4.2), then the ternary Goldbach conjecture<br />

holds for all odd n>10 20 . Further, [Saouter 1998] “bootstrapped” the then<br />

current bound of 4 · 10 11 for the binary Goldbach problem to show that the<br />

ternary Goldbach conjecture holds unconditionally for all odd numbers up<br />

to 10 20 . Thus, with the Zinoviev theorem, the ternary Goldbach problem is<br />

completely solved under the assumption of the ERH.<br />

It follows from the Vinogradov theorem that there is a number k such<br />

that every integer starting with 2 is a sum of k or fewer primes. This corollary<br />

was actually proved earlier by G. Shnirel’man in a completely different<br />

way. Shnirel’man used the Brun sieve method to show that the set of even<br />

numbers representable as a sum of two primes contains a subset with positive<br />

asymptotic density (this predated the results that almost all even numbers<br />

were so representable), and using just this fact was able to prove there is such<br />

anumberk. (See Exercise 1.44 for a tour of one proof method.) The least<br />

number k0 such that every number starting with 2 is a sum of k0 or fewer<br />

primes is now known as the Shnirel’man constant. If Goldbach’s conjecture is<br />

true, then k0 = 3. Since we now know that the ternary Goldbach conjecture<br />

is true, conditionally on the ERH, it follows that on this condition, k0 ≤ 4.<br />

The best unconditional estimate is due to O. Ramaré who showed that k0 ≤ 7<br />

[Ramaré 1995]. Ramaré’s proof used a great deal of computational analytic<br />

number theory, some of it joint with R. Rumely.<br />

1.2.4 The convexity question<br />

One spawning ground for curiosities about the primes is the theoretical issue<br />

of their density, either in special regions or under special constraints. Are there<br />

regions of integers in which primes are especially dense? Or especially sparse?<br />

Amusing dilemmas sometimes surface, such as the following one. There is an<br />

old conjecture of Hardy and Littlewood on the “convexity” of the distribution<br />

of primes:<br />

Conjecture 1.2.3. If x ≥ y ≥ 2, then π(x + y) ≤ π(x)+π(y).<br />

On the face of it, this conjecture seems reasonable: After all, since the primes<br />

tend to thin out, there ought to be fewer primes in the interval [x, x + y] than<br />

in [0,y]. But amazingly, Conjecture 1.2.3 is known to be incompatible with<br />

the prime k-tuples Conjecture 1.2.1 [Hensley and Richards 1973].<br />

So, which conjecture is true? Maybe neither is, but the current thinking is<br />

that the Hardy–Littlewood convexity Conjecture 1.2.3 is false, while the prime<br />

k-tuples conjecture is true. It would seem fairly easy to actually prove that the<br />

convexity conjecture is false; you just need to come up with numerical values of<br />

x and y where π(x+y),π(x),π(y) can be computed and π(x+y) >π(x)+π(y).

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

Saved successfully!

Ooh no, something went wrong!