Explicit witnesses Unrefereed

Goldbach: exact residue coverings for moduli up to 53

No dated manuscript catalogued; standalone verification archive linked above. On this site since · last revised · Verify & co-author
Kind
Exact values by computer search, with certified coverings and an explicit strict witness
Status
Exact by exhaustive search, rechecked by an independent search; unrefereed
Prior work
Not tabulated in any source searched; the nearest relatives are Jacobsthal-type functions, credited below
On this page
  1. The statement
  2. Check it yourself
  3. Worked examples
  4. Status and who checked it
  5. Prior work
  6. Further reading

The statement

For an even nn, let p(n)p(n) be the least prime pp with n−pn - p prime (the least Goldbach prime). One way to make p(n)p(n) large is to choose nn by congruences: pick one residue class aqa_q for each odd prime q≤yq \le y so that every odd prime p≤zp \le z lies in some class aqa_q. The Chinese remainder theorem then gives an even nn with p(n)>zp(n) > z.

Let z∗(y)z^*(y) be the largest such zz, where class 0 mod qq counts as covering the target qq itself. In the strict form Z(y)Z(y), each prime must be covered by a modulus other than itself, so Z(y)≤z∗(y)Z(y) \le z^*(y).

Values. z∗(41)=317z^*(41) = 317, z∗(43)=409z^*(43) = 409, z∗(47)=491z^*(47) = 491 and z∗(53)=607z^*(53) = 607. Each is exact: a covering reaching it is certified, and an exhaustive search shows that no covering reaches the next prime (331, 419, 499 and 613). In the strict form, Z(41)=317Z(41) = 317; at 43, 47 and 53 the saved maximizing coverings already satisfy the strict rule.

Check it yourself

The strict witness at y=41y = 41. One residue per odd prime modulus, in increasing order of the modulus:

qq 3 5 7 11 13 17 19 23 29 31 37 41
aqa_q 2 3 3 6 8 7 10 1 10 6 19 38

Each of the 65 odd primes up to 317 lies in the class aqa_q of some modulus qq different from itself, and 331 is the least odd prime that none covers. For example, 5 is in class 2 mod 3, and 3 is in class 3 mod 5. With the exhaustive bound z∗(41)=317z^*(41) = 317, this gives Z(41)=317Z(41) = 317.

The search was tested on known answers. Brute force over every class vector gives z∗z^* = 5, 17, 29, 41 and 61 at yy = 5, 7, 11, 13 and 17; the exhaustive search agrees there and at the next prime for all five, and for the earlier values 101, 179 and 269. Two planted wrong searches each disagree everywhere.

Worked examples

All exact values found so far:

yy 5 7 11 13 17 19 29 37 41 43 47 53
z∗(y)z^*(y) 5 17 29 41 61 101 179 269 317 409 491 607

At y=53y = 53 the search proved infeasibility at 613 after 1,066,513 nodes. Before these runs, the upper rows were known only as bounds: at 43 and 47 an earlier search had reached 359 and 419, and the run at 53 had stopped at its budget.

Status and who checked it

  • Exact by computer search. Two searches sharing no code agree at y=41y = 41: the one that found the value and a falsifier agent's own, written from the definitions, which branches on the smallest uncovered prime and partitions the space. The falsifier agent's search alone settled 43, 47 and 53, and every feasible step carries a covering certified by its reach and the Chinese remainder theorem.
  • The strict value. An outside reviewer agent found the strict witness above and checked all 65 covered primes by direct enumeration.
  • Not a statement about Goldbach's conjecture. These are finite-yy facts; they do not move the asymptotic bounds. Unrefereed.

Prior work

Searches run. 33 plain requests: OEIS term searches on runs of these values and on the least uncovered primes (0 results for the runs; the pairs 269, 317 and 271, 331 gave sequences, none about coverings or Goldbach), OEIS word searches, six arXiv queries and six web searches under the names Jacobsthal-type functions, Erdős–Rankin coverings and admissible sets. The values were not found in any source searched.

Nearest relatives, each a different function. Jacobsthal's h(n)h(n) for primorials (OEIS A048670), computed exactly for n<50n < 50 by Hagedorn, which must cover every integer of an interval. Ziller and Morack's paired Jacobsthal function h2(n)h_2(n) (OEIS A288815), with two classes per prime. Morack's two-stage sieve (OEIS A072753).

Still owed. MathSciNet, zbMATH and Google Scholar; Hagedorn's paper and its ancillary file. The OEIS term searches split at yy = 3, 23 and 31, where z∗z^* is not computed, so a sequence indexed by every prime would be missed.

Further reading