07.01.2013 Views

Dietzfelbinger M. Primality testing in polynomial time ... - tiera.ru

Dietzfelbinger M. Primality testing in polynomial time ... - tiera.ru

Dietzfelbinger M. Primality testing in polynomial time ... - tiera.ru

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.

46 3. Fundamentals from Number Theory<br />

Theorem 3.6.2 (The Prime Number Theorem).<br />

π(x)<br />

lim<br />

x→∞ x/ ln x =1<br />

The prime number theorem should be read as follows: asymptotically,<br />

that means for large enough x, about a fraction of 1 <strong>in</strong> ln x of the numbers<br />

≤ x will be primes, or, the density of prime numbers among the <strong>in</strong>tegers <strong>in</strong><br />

the neighborhood of x is around 1 <strong>in</strong> ln x. Actually, the figure x/(ln x − 1)<br />

is an even better approximation. We can thus estimate that the percentage<br />

of primes <strong>in</strong> numbers that can be written with up to 50 decimal digits is<br />

about 1/ ln(10 50 − 1) = 1/(50 ln 10 − 1) ≈ 1/114 or 0.88 percent; for 100<br />

decimal digits the percentage is about 1/ ln(10 100 − 1) = 1/(100 ln 10 − 1) ≈<br />

1/229. In general, doubl<strong>in</strong>g the number of digits will approximately halve the<br />

percentage of prime numbers. Readers who wish to see a full proof of the<br />

prime number theorem are referred to [6]; for details on the quality of the<br />

approximation see [16].<br />

We cannot prove the prime number theorem here, and really we do not<br />

need it. Rather, we are content with show<strong>in</strong>g that π(x) =Θ(x/ log x), which<br />

is sufficient for our purposes. The proofs for these weaker upper and lower<br />

bounds are both classical gems and quite clear and should give the reader<br />

a good <strong>in</strong>tuitive understand<strong>in</strong>g of why the density of prime numbers <strong>in</strong><br />

{1,...,N} is Θ(1/log N). We will have the opportunity to use a variant<br />

of these bounds (Proposition 3.6.9) <strong>in</strong> the analysis of the determ<strong>in</strong>istic primality<br />

test. Moreover, lower bounds on the density of the prime numbers<br />

are important for analyz<strong>in</strong>g the <strong>ru</strong>nn<strong>in</strong>g <strong>time</strong> of randomized procedures for<br />

generat<strong>in</strong>g large prime numbers.<br />

Theorem 3.6.3. For all <strong>in</strong>tegers N ≥ 2 we have<br />

N<br />

3N<br />

− 2 ≤ π(N) ≤<br />

log N log N .<br />

We prove the lower bound first, and then turn to the upper bound.<br />

Proof of Theorem 3.6.3 — The Lower Bound. First, we focus on even<br />

numbers N, oftheformN =2n. In the center of the lower bound proof<br />

stands the b<strong>in</strong>omial coefficient<br />

� �<br />

2n<br />

=<br />

n<br />

(2n)! 2n(2n − 1) ···(n +1)<br />

= .<br />

n! · n! n(n − 1) ···1<br />

(For a discussion of factorials n! and b<strong>in</strong>omial coefficients � � n<br />

k see Appendix<br />

A.1.) Recall that � � 2n<br />

n is the number of n-element subsets of a 2nelement<br />

set and as such is a natural number. In comparison to 2n the number<br />

� � 2n<br />

2n<br />

n is very large, namely very close to 2 . Now consider the prime<br />

decomposition

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

Saved successfully!

Ooh no, something went wrong!