- View SourceMany thanks to everyone replied my message.

I did not know before the equivalent criteria

Paul Joblin cited.

My claim that the twin primes' rule holds

was based on the following arguments:

(A) n is prime iff [(n-1)/2]!^2 = +/- 1 mod n

(B) (n,n+2) is a prime pair iff

[(n-1)/2]!^2 = +/- 1 mod n = -/+4 mod (n+2)

(note that the sign on the left side of / is for n=4k-1

and the sign on the rigth side of / is for n=4k+1)

If you agree with (A) and (B), whose proofs are skipped, then

(C) when n=4k+1

considering the identity [(n-3)/2]n - 1 = [(n-5)/2](n+2) + 4

it follows that [(n-3)/2]n - 1 = - 1 mod n = +4 mod (n+2)

hence, by (B), [(n-1)/2]!^2 =[(n-3)/2]n - 1 mod n and mod (n+2)

this means that [(n-1)/2]!^2 =[(n-3)/2]n - 1 mod n(n+2)

which leads to 2[(n-1)/2]!^2 = -(5n+2) mod n(n+2)

(D) when n=4k-1

considering the identity [(n+7)/2]n + 1 = [(n+5)/2](n+2) - 4

and proceeding as in (C)

2[(n-1)/2]!^2 = +(5n+2) mod n(n+2) is obtained at last.

Could anyone verify or not the above reasoning?

I would be very grateful if someone send me a copy of Joseph B. Dence

& Thomas P. Dence's paper where twin prime criteria were proved.

Best regards

Flavio Torasso

-----Messaggio originale-----

Da: Alan Powell [mailto:powella@...]

Inviato: giovedÃ¬ 29 marzo 2001 21.00

A: Flavio Torasso

Cc: primenumbers@yahoogroups.com

Oggetto: [PrimeNumbers] Twin primes Criteria (Correction)

Hi Flavio

As Paul Jobling pointed out, your conjecture is equivalent to

the criteria (2) and (3) below. A more thorough investigation

shows that both these equivalent versions are indeed correct.

Mike Oakes appears to be mistaken or I am missing something?

I base this on the following Mathematica snippet:

Do[p=Prime[i];

If[Mod[p,4]==1, r=-1, r=+1];

If[PrimeQ[p]&&PrimeQ[p+2],

m=Mod[2((p-1)/2)!^2 - r(5*p+2),p(p+2)];

If[m!=0,Print["False"];

];

];

,{i,2,PrimePi[10^4]}];

In 1949 using Wilson's Theorem* P. A. Clement published

a proof that p and p+2 are both prime ("twin primes")

if and only if the following congruence holds:

(1) 4((p-1)! + 1) + p = 0 [mod p(p+2)]

Furthermore in 1995 Joseph B. Dence & Thomas P. Dence

reduced the Clement criterion to the following two criteria:

(2) 2((((p-1)/2)!)^2 + 1) + 5p = 0 [mod p(p+2)]

if and only if p and p+2 are primes and p=4k+1

(3) 2((((p-1)/2)!)^2 - 1) - 5p = 0 [mod p(p+2)]

if and only if p and p+2 are primes and p=4k-1

A similar criterion exists for the primality of p and p+d:

If p>1 and d>1 are both integers, then p and p+d are both

prime if and only if:

( 1 ((-1)^d)d! ) 1 1

(4) (p-1)! ( - + ---------- ) + - + --- is an integer

( p p+d ) p p+d

In addition P. A. Clement's paper proves similar necessary and

sufficient criteria for prime triples p, p+2 and p+6 as well

as prime quadruplets p, p+2, p+6, p+8. If these are of general

interest I can post them to the list.

Regards

Alan Powell

* Wilson's Theorem:

An integer p>1 is prime if and only if (p-1)! + 1 = 0 [mod p]

At 09:06 AM 3/27/01, torasso.flavio@... wrote:

> I found the following rule concerning twin primes:

> n and n+2 are both prime iff

> 2 [(n-1)/2]!^2 = \pm (5n+2) mod n(n+2)

> the sign being "+" when n=4k-1, "-" when n=4k+1.

>

> Is it an interesting or trivial result?

> Are there similar congruences for other prime pairs?

>

> Thanks for any comments