[seqfan] Where are the primes?
David W. Wilson
wilson at cabletron.com
Tue May 12 21:53:09 CEST 1998
Richard Guy wrote:
> You can get an earlier AP of 12 primes if you don't insist on the minimum
> common difference. E.g., a = 23143, d = 30030 [V.A. Golubev, probably
> one of the references in A5 in UPINT.] R.
My computations were in regard to finding an AP with the theoretically
smallest possible difference. If the Prime Patterns conjecture is true, then
an upper bound on this minimal difference is n# (n primorial) where n is
the number of elements in the AP. Prior to my search, a 12-element AP
of primes theoretically had minimal difference 2310, whereas now this is
established fact.
In brief, if the difference of an AP of n primes is less than n#, then some
prime p <= n does not divide this difference, and therefore divides one of
the elements. This element, being prime, must be p itself. Therefore, an
AP of primes either has difference >= n# or contains a prime element
<= n.
The latter case includes a finite number of AP's which can at worst be
exhaustively checked (but we can streamline the search). This search
can produce a AP of primes of difference < n#, as with the AP of 7 primes
starting at 7 of difference 150.
If there is no AP with difference < n#, then we consider an AP of n
elements with difference n#. Such an AP does not cover all the residues
of any prime (for p <= n, all elements will all have the same residue,
since p divides n#, for p > n, the n elements cannot possibly cover all p
residues). These are precisely the conditions under which the Prime
Patterns Conjecture asserts an infinitude of prime patterns in the
form of a n-element AP of difference n#.
In his article "Prime Arithmetic Sequence", Weisstein originally had
an incomplete sequence of theoretical minimum differences for a
prime AP of n elements. As an early element was missing, Sloane did
not have the sequence (? is not a valid element in a Sloane sequence).
Based upon the above argument assuming PPC, I filled in the gaps,
and extended the sequence to Sloane's A033188.
Finding a starter element that confirms a given minimum difference
is, of course, rather another matter. I easily computed a starters for
n <= 10, and so began A033189. Weisstein then blindsided me by
finding a(11), but I fought back with a(12) and a(13). The idea of
A033189 is to find the smallest starter element that confirms the
corresponding difference in A033188.
More information about the SeqFan
mailing list