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.

134 Chapter 3 RECOGNIZING PRIMES AND COMPOSITES<br />

Korselt felt sure that no examples could possibly exist, and developed the<br />

criterion as a first step toward proving this.<br />

Proof. First, suppose n is a Carmichael number. Then n is composite. Let p<br />

be a prime factor of n. Fromp n ≡ p (mod n), we see that p 2 does not divide<br />

n. Thus,n is squarefree. Let a be a primitive root modulo p. Sincea n ≡ a<br />

(mod n), we have a n ≡ a (mod p), from which we see that a n−1 ≡ 1(modp).<br />

But a (mod p) has order p − 1, so that p − 1 divides n − 1.<br />

Now, conversely, assume that n is composite, squarefree, and for each<br />

prime p dividing n, wehavep − 1 dividing n − 1. We are to show that a n ≡ a<br />

(mod n) for every integer a. Sincen is squarefree, it suffices to show that<br />

a n ≡ a (mod p) for every integer a and for each prime p dividing n. So suppose<br />

that p|n and a is an integer. If a is not divisible by p,wehavea p−1 ≡ 1(modp)<br />

(by (3.3)), and since p − 1 divides n − 1, we have a n−1 ≡ 1(modp). Thus,<br />

a n ≡ a (mod p). But this congruence clearly holds when a is divisible by p,<br />

so it holds for all a. This completes the proof of the theorem. ✷<br />

Are there infinitely many Carmichael numbers? Again, unfortunately for<br />

primality testing, the answer is yes. This was shown in [Alford et al. 1994a].<br />

P. Erdős had given a heuristic argument in 1956 that not only are there<br />

infinitely many Carmichael numbers, but they are not as rare as one might<br />

expect. That is, if C(x) denotes the number of Carmichael numbers up to the<br />

bound x, thenErdős conjectured that for each ε>0, there is a number x0(ε)<br />

such that C(x) >x 1−ε for all x ≥ x0(ε). The proof of Alford, Granville, and<br />

Pomerance starts from the Erdős heuristic and adds some new ingredients.<br />

Theorem 3.4.7. (Alford, Granville, Pomerance). There are infinitely many<br />

Carmichael numbers. In particular, for x sufficiently large, the number C(x)<br />

of Carmichael numbers not exceeding x satisfies C(x) >x 2/7 .<br />

The proof is beyond the scope of this book; it may be found in [Alford et al.<br />

1994a].<br />

The “sufficiently large” in Theorem 3.4.7 has not been calculated, but<br />

probably it is the 96th Carmichael number, 8719309. From calculations in<br />

[Pinch 1993] it seems likely that C(x) >x 1/3 for all x ≥ 10 15 .Alreadyat<br />

10 15 , there are 105212 Carmichael numbers. Though Erdős has conjectured<br />

that C(x) > x 1−ε for x ≥ x0(ε), we know no numerical value of x with<br />

C(x) >x 1/2 .<br />

Is there a “Carmichael number theorem,” which like the prime number<br />

theorem would give an asymptotic formula for C(x)? So far there is not even<br />

a conjecture for what this formula may be. However, there is a somewhat<br />

weaker conjecture.<br />

Conjecture 3.4.1 (Erdős, Pomerance). The number C(x) of Carmichael<br />

numbers not exceeding x satisfies<br />

as x →∞.<br />

1−(1+o(1)) ln ln ln x/ ln ln x<br />

C(x) =x

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

Saved successfully!

Ooh no, something went wrong!