Fermat's
little theorem states that for any prime number N and integer a, not a multiple of N
(or, equivalently,[a]N belonging to the group of units UN of
ZN)aN-1 is congruent to 1 modulo N.
Now continuing our story the converse does not hold. There are in fact, infinitely many, composite numbers, the so-called Carmichael Numbers with the property that aN-1=1 mod N for any a relatively prime to N. The smallest example is 561=3*11*17, [the next 1105=5*13*17 and the third 1729=7*13*19, the Ramanujan-Hardy taxicab number]. Using the Chinese Remainder Theorem one easily shows that a composite N is a Carmichael number if and only if N is squarefree and p-1 divides N-1 for all prime factors p of N. So if we use the congruence relation aN-1=1 mod N as a primality test, it may occur that a composite number passes the test for all a.(!) Definite primality tests of a similar nature exist for numbers of a particular form, for instance: A number N will be called a Proth number, if N=k*2n + 1 with 0 < k < 2n and k odd.
To the right two stamps dedicated to the centenary of Sir William Rowan Hamilton's discovery of the skew field of Quaternions ,the first non-commutative division algebra ever.
Indeed in 1878 François Proth, a French farmer, published the following theorem in the "Comptes Rendus":
If there exists a number a such that- Proof. Let a and N satisfy the conditions above, x=ak and suppose, by way of contradiction, that N is composite. Let's say N=K*L with 1 < K < (or=) L and K and L relatively prime (so that the Chinese Remainder Theorem applies and UK is a homomorphic image of UN). Then, since k < 2n, we have K < 2n and the order of x in UK is a proper divisor of 2n. But now a(N-1)/2 = (x to the power 2n-1) = +1 mod K instead of the correct value -1.a(N-1)/2 = -1 mod N,
where N is a number now baring his name, then N is a prime.
- Here is another straightforward generalization of Proth's Theorem:
Let p be a prime number and
Fp(x) = (xp-1)/(x-1) =
1+x+x2+...+xp-1 be the corresponding cyclotomic polynomial.
Let N=k * pn + 1 with 0 < k < pn.
Then N is a prime, if there exists a number a (called a verificator) such that
Fp(a(N-1)/p) = 0 mod N.
Note that we may assume without loss of generality that k is not a multiple of p.
The proof is virtually the same as the one for the particular case p=2, which is Proth's result.
The important things to keep in mind is that under the given assumptions
the order of x=ak in UN is exactly
the maximal possible value pn and that, as before, K < pn,
if N were a product of K and L, with K and L relatively prime and 1 < K <
(or =) L.
- For k=2 and p=3 and n < 12 the number N=k * pn + 1 is prime for n=1,2,4,5 with verificator a equal to 2 and for n=7,9 with a=3. To eliminate the remaining cases note that 5 divides N iff n=3 mod 4, 7 divides N iff n=1 mod 6, 11 divides N iff n=3 mod 5, 17 divides N iff n=10 mod 16 etc. For n=12 and a=2 our N fails the prime test aN-1 = 1 mod N, mentioned above.
To the right you can see the first calculating machine ever, made by
Wilhelm Schickard around 1623.

- On the Prime Pages you can find and download a fast algorithm for testing Proth numbers for primality due to Yves Gallot. The internet community uses this algorithm to investigate interesting conjectures such as the famous Sierpinski Problem.
- Remember Euclid's proof of the infinity of the number of primes? In 2001 Chris Caldwell and Yves Gallot found a new prime of the form p#+1, where, for p prime, p# denotes the product of all primes up to and including p. It is 42209#+1, a 18241 digit number. [Math.Comp,71,pp.441-448]
- Albert Kluge has written some interesting related applets: Eratosthenes' sief and the Miller-Rabin probabilistic prime test and...