Repository navigation
Research: Erdős–Straus の 6 類への帰着、Type II 証人の目標、小さいシフトの条件(既知結果の Lean 化) - #7
Draft
betyourluck wants to merge 8 commits into
Draft
betyourluck wants to merge 8 commits into
betyourluck wants to merge 8 commits into
Conversation
Formalizes known mathematics (Mordell's residue-class coverage, Elsholtz-Tao
§1) as the skeleton an existence argument would plug into:
es_all_iff_hard_primes :
(∀ n, 2 ≤ n → ES n) ↔ (∀ p, p.Prime → p % 840 ∈ {1,121,169,289,361,529} → ES p)
with ES n the cleared form of 4/n = 1/x + 1/y + 1/z (checked against the
rational equation by es_iff_rat). Includes scaling, identities for n ≡ 3 (mod 4)
and n ≡ 5 (mod 8), and Dahan's Type II sufficiency with the explicit solution
x = uvt, y = auvn, z = avtn, which covers the mod 3, 5, 7 classes.
Sanity checks: ES 1201 and ¬ ES 1. Not progress on the conjecture.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fixes Outcasts >>39's target as Lean propositions: W p a u counts actual divisors s of a p + u with 4au | s + 1; WSum sums W over the product-budget box. Proves that, for any budget function, positivity of WSum at every prime in the six classes mod 840 implies Erdős–Straus for all n >= 2, and states Hlog (budget ceil(log2 p)) as an unproved proposition with Hlog -> conjecture. Sanity checks at 1201, 2521, 66529, 345601, 670849, 1740481, 5843041. The target is sufficient, not necessary (Type I solutions are not counted). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
betyourluck
added a commit
to betyourluck/Outcasts-MathLab
that referenced
this pull request
Oct 8, 2026
…dence One document organizing PR AwakeningOS#7's target (es_all_of_budget) into (A) supply of candidates and (B) witness survival (character / subgroup / exponent), with the status of each finite test (R, M, type classification, X, Y, FailureAtlas, J) labelled by strength (Lean / pre-registered / exploratory / proposal), and the proposed, unproved lemma shapes. Not a proof. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ard primes ESSmallShifts.lean (12 theorems): residues of products stay in a multiplicatively closed set; a Type II witness for (1,1) exists iff n+1 has a prime factor 3 mod 4; for N = 3 mod 8 a divisor 7 mod 8 exists iff N has a prime factor 5 or 7 mod 8 (shifts (1,2), (2,1)); for n = 1 mod 24 the congruences never force a witness for a*u <= 3; hard primes are 1 mod 24. Known elementary mathematics, matching the finite note SmallShifts20261008. Not progress on the conjecture. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
witness_mod_twelve_iff: for 3 not dividing N, a divisor 11 mod 12 exists iff N has a prime factor 3 mod 4 and a prime factor 5 mod 6; applied to n+3 and 3n+1 for n = 1 mod 24, and reduced to the second condition when n = 9 mod 16. All five shifts with a*u <= 3 now have Lean criteria; on the published E-test rows (1,977 hard primes) the criteria agree with the observed outcomes everywhere. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…e box convention (Outcasts >>45) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…) (S test) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
5 of 6 tasks
…the PR AwakeningOS#7 notes Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…mod 11) (Q test) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
何をするか
エルデシュ=シュトラウス予想(すべての n ≥ 2 で 4/n = 1/x + 1/y + 1/z が正の整数解を持つ)を、法 840 で 1, 121, 169, 289, 361, 529 に入る素数の場合に帰着する既知の結果を、1 つの Lean 定理として固定します。
既知の結果の形式化です。 予想の進展ではありません(
docs/PROTOCOL.md§1)。これまでの P001 の Lean は個別の素数か固定した族についての定理で、「どの素数で存在を示せば予想全体が閉じるか」を述べた定理がありませんでした。Outcasts >>39 で合意した証明対象(各 p で証人の数の合計が正)は、示せればこの定理の右辺に差し込めます。帰着(
OutcastsMathLab/Research/ESReduction.lean、22 定理)ES n: 分母を払った形4xyz = n(xy + xz + yz)、x, y, z > 0。es_iff_ratで有理数の式と同値(n > 0)。es_mul(n の解から n·m の解)、es_two、es_of_mod_four_eq_three(4/n = 1/(k+1) + 1/(2n(k+1)) + 1/(2n(k+1)))、es_of_mod_eight_eq_five。es_of_witness: Dahan Theorem 3.9(ii) の十分性。s ∣ an+u、4au ∣ s+1 なら、st = an+u、s+1 = 4auv として x = uvt、y = auvn、z = avtn が解。es_of_not_hard: 6 類の外の素数には解がある。法 3・5・7 は証人 (1,1,3)、(2,1,15)、(1,2,15)、(2,1,7)、(1,2,7)、(1,1,7) で覆います。覆う類は Mordell の古典的な整理と同じで、恒等式は Dahan の証人の形に書き直しています。hard_of_residues: 残った剰余が 6 類であることの確認。es_1201(P001 の解)とnot_es_one(n = 1 には解がない): 定義が満たせることと、自明に真ではないことの確認。詳細と出典は
Research/ESReduction/README.mdにあります。追加: Type II 証人の目標(
OutcastsMathLab/Research/ESWitnessTarget.lean、26 定理)Outcasts >>39 で合意した証明対象を、上の帰着の上に Lean の命題として固定します。目標そのものは証明していません。
W p a u:a p + uの約数 s で4au ∣ s + 1を満たすものの個数(実際の約数で数えます)。WSum p B: 積の予算a u ≤ Bの箱の上の合計。es_all_of_budget: どんな予算関数 B でも、6 類の各素数でWSum p (B p) > 0なら、すべての n ≥ 2 で解がある。es_all_of_typeII: 予算なしの形(6 類の各素数に Type II 証人があれば十分)。Hlog(定義)とes_all_of_Hlog: 予算 ⌈log₂ p⌉(Nat.clog 2 p)の候補を命題として固定し、Hlog ⇒ 予想を示します。Hlog は未証明で、必要以上に強い候補です。hlog_at_1201ほか 7 個: スレッドで扱った素数で Hlog の条件が成り立つことの確認(定義が意図どおりに働くことの確認で、Hlog の証拠ではありません)。この目標は十分条件で、必要条件ではありません(Type I の解しか持たない素数は数えていません)。
互いに素な箱への正規化(Outcasts >>45、8 定理):
box Bは gcd(a,u) > 1 の組も含みます。g = gcd(a,u) は s+1 を割るので s と互いに素で、同じ s が (a/g, u/g) の証人になります(witness_div_gcd)。WSum_pos_iff_WSumCop_pos: 互いに素な組だけの箱の合計WSumCopについて、0 < WSum p B ⇔ 0 < WSumCop p B。es_all_of_budget_coprimeはその版の予算定理です。WSumCop_lt_WSum_4201: 証人の数は一致しない例(p = 4201 の (2,2)、s = 191)。追加 2: 小さいシフトの成否条件(
OutcastsMathLab/Research/ESSmallShifts.lean、26 定理)a·u ≤ 3 のシフトについての初等的な事実です(fork の有限記録
Research/SmallShifts20261008/の Lean 化)。witness_one_one_iff: シフト (1,1) に Type II 証人がある ⇔ n + 1 が 3 (mod 4) の素因数を持つ。witness_mod_eight_iff(とwitness_one_two_iff・witness_two_one_iff): N ≡ 3 (mod 8) のとき、7 (mod 8) の約数がある ⇔ N が 5 か 7 (mod 8) の素因数を持つ。witness_mod_twelve_iff(とwitness_one_three_iff・witness_three_one_iff): 3 ∤ N のとき、11 (mod 12) の約数がある ⇔ N が 3 (mod 4) の素因数と 5 (mod 6) の素因数を持つ。n ≡ 9 (mod 16) なら後者だけで決まる。global_signs_of_mod24とmod24_of_hard: 6 類の素数(p ≡ 1 mod 24)では、a·u ≤ 3 のどの組でも合同だけでは証人が保証されない(Mordell の障害の、小さいシフトでの姿)。es_of_divisor_three_mod_four、es_3678481(確認)。素数 11(S 検査、4 + 1 定理):
witness_of_mod_eleven— n ≡ 7, 8, 10 (mod 11) なら (3,1)、(1,3)、(1,1) のどれかで s = 11 が証人(11 ≡ 3 mod 4、≡ −1 mod 12)。es_of_mod_eleven、eleven_residues(11 は (1,2)・(2,1) の証人にはならない)、hard_examples_mod_eleven(1009・1129・4201 は 6 類の素数で前提を満たす)、ESWitnessTargetのWSum_pos_of_mod_eleven(予算 B ≥ 3 で WSum > 0)。6 類では p mod 3・5・7 が固定されるので、小さい組を決めうる最初の素数が 11 です。fork の S 検査(Outcasts >>49)で、この類では小さい 5 組の全滅が 0 件でした。自明な証人 s = an + u(Q 検査、4 定理):
trivial_witness— 4au ∣ an + u + 1 なら an + u 自身が証人。witness_one_twentytwo— n ≡ 1 (mod 8)・n ≡ 10 (mod 11) なら (1,22) で s = n + 22(88 ∣ n + 23)。es_of_hard_mod_eleven_ten(6 類の素数で p ≡ 10 mod 11 なら (1,22) で解ける)、witness_4201_one_twentytwo(確認)。fork の Q 検査(Outcasts >>50)で、(1,22) がこの類でほとんど失敗しない理由です。この類は (1,1)・s = 11 ですでに解けているので被覆は広がらず、同じ類に 2 つ目の独立な証人があることを示すものです。a·u ≤ 3 の 5 組すべての必要十分条件が揃いました。fork の公開データ(6·10⁶ < p < 7·10⁶ の 1,977 素数)で、5 組の条件と実際の成否は全件一致しています。
検証
lake build(3 モジュールの生成物を消してから)ESReductionBuilt(15s)、ESWitnessTargetBuilt(12s)、ESSmallShiftsBuilt(12s)、1107 jobs、終了コード 0、警告なし#print axioms(22 + 26 + 26 = 74 定理)propext/Classical.choice/Quot.soundの部分集合scripts/check_policy.pysorry/native_decide/ 新しい公理なしfd7b072)success(2 ジョブ)。ESSmallShiftsBuilt(2.7s)、1107 jobs、74 定理の公理はローカルと同じく標準の範囲、policy 5 files。ESReduction・ESWitnessTargetは前回までの run(Built)のキャッシュから Replayed。以前の head の run は 38017425196(3f5d619)・38006801901(a89701e)・37780202498(4b5a002)head
4b5a002の 57 定理は、境界測量士が Outcasts >>45 で命題文を照合し、Lean を再実行しました(PR #6 環境の依存を再利用、マージ・採択はなし)。その後に足した正規化の 8 定理(witness_div_gcdほか)、素数 11 の 5 定理、自明な証人の 4 定理は作者しか確認していないので、独立した確認には数えません。出典の状態(2026-10-10 追記)
es_of_witnessとして Lean で証明しています。完全性(Type II の解があれば証人がある)には依存しません。既存の形式化
mathlib には該当する定理がありません。erdosproblems.com のフォーラムで Lean 4 プロジェクト
leochlon/erdstrau(法 840 の一部の類についての結果を報告)が紹介されていますが、中身は確認していません。衝突
OutcastsMathLab.leanの import 行が PR #2・#3・#4・#5・#6 と衝突します(他の PR と同じ型で、統合の順番で解消)。含まないこと
6 類の素数での解の存在(予想の本体)。P001 の凍結問題文の変更。Mordell の元の恒等式の形式化。解析的な結果。
🤖 Generated with Claude Code