From bd7add28181243f76bddd4b56f166572d5fb7942 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 29 Sep 2026 16:21:02 +0000 Subject: [PATCH 1/4] =?UTF-8?q?math(candidate):=20C6=20successor=20?= =?UTF-8?q?=E2=80=94=20super-exponential=20Rouche-degree=20tail;=20O(r^3?= =?UTF-8?q?=20[L/log=20L]^2)=20in=20the=20plane,=20O(r^3=20[L/log=20L]^d)?= =?UTF-8?q?=20in=20dimension=20d=20given=20G=5Fd=20(scientific=20effect=20?= =?UTF-8?q?NONE)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Successor to CL-C6-FACTORIAL-20260929-v1 (#140), whose PROOF.md is not modified. Writes out the route proposed by OpenAI in #140 comment 5893989971 (arithmetic checked in 5894002968): - Lemma G: E_{Q_r} sup over a complex ball of radius R of |grad f|^2 and |D grad f|^2 is <= C e^{C R^2}, uniformly in the pin family (Fourier + completing the square + bounded E_r regression rows). - Lemma M+: with Cauchy radius sqrt(m) (fixed protected annuli and the reviewed Lemma D unchanged), P(m* > m) <= C exp(-m log m/(8d)). - Theorem S_2: E N(N-1) <= C r^3 [L/log L]^2, L = log(1/r), and the factorial-moment analogue (unconditional, via I5). - Theorem S_d: the same with exponent d in every fixed d >= 2, conditional on Theorem G_d of #141 (under review); uses the d-dim forms of Lemmas R (Bezout m^d), D and E written in section 4 and the accepted parent's W/Z bounds in fixed d. Theta(r^3) is not claimed. Companion: 6 checks, 4 mutants, -O identical. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_017Mi3hxjaxV45x6zo6o1ee3 --- .github/workflows/c6-sharpened.yml | 59 ++++++ frontiers/c6_sharpened_20260929/PROOF.md | 200 ++++++++++++++++++ frontiers/c6_sharpened_20260929/RESULTS.json | 13 ++ .../c6_sharpened_20260929/SOURCE_FILES.json | 52 +++++ .../c6_sharpened_20260929/sharp_check.py | 134 ++++++++++++ 5 files changed, 458 insertions(+) create mode 100644 .github/workflows/c6-sharpened.yml create mode 100644 frontiers/c6_sharpened_20260929/PROOF.md create mode 100644 frontiers/c6_sharpened_20260929/RESULTS.json create mode 100644 frontiers/c6_sharpened_20260929/SOURCE_FILES.json create mode 100644 frontiers/c6_sharpened_20260929/sharp_check.py diff --git a/.github/workflows/c6-sharpened.yml b/.github/workflows/c6-sharpened.yml new file mode 100644 index 000000000..3a8a205e2 --- /dev/null +++ b/.github/workflows/c6-sharpened.yml @@ -0,0 +1,59 @@ +name: C6 sharpened successor checks + +'on': + pull_request: + paths: + - 'frontiers/c6_sharpened_20260929/**' + - '.github/workflows/c6-sharpened.yml' + workflow_dispatch: + +permissions: + contents: read + +jobs: + exact-and-probe: + runs-on: ubuntu-latest + timeout-minutes: 10 + steps: + - uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 + with: + persist-credentials: false + - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 + with: + python-version: '3.11.16' + - name: Verify sources, replay both modes, reject mutants + env: + PYTHONDONTWRITEBYTECODE: '1' + run: | + set -euo pipefail + python -B -S - <<'PY' + import hashlib, json, pathlib, subprocess, sys + root = pathlib.Path('frontiers/c6_sharpened_20260929').resolve() + manifest = json.loads((root / 'SOURCE_FILES.json').read_text())['files'] + present = sorted(p.name for p in root.iterdir()) + expected_names = sorted([e['path'] for e in manifest] + ['SOURCE_FILES.json']) + if present != expected_names: + raise ValueError('packet tree differs from manifest: ' + repr(present)) + for entry in manifest: + path = root / entry['path'] + if pathlib.PurePosixPath(entry['path']).name != entry['path'] or path.is_symlink(): + raise ValueError('flat regular source required: ' + entry['path']) + data = path.read_bytes() + if len(data) != entry['bytes'] or hashlib.sha256(data).hexdigest() != entry['sha256']: + raise ValueError('source identity mismatch: ' + entry['path']) + script = str(root / 'sharp_check.py') + expected = (root / 'RESULTS.json').read_bytes() + mutants = ['fixed-radius', 'wrong-cutoff', 'planar-cap', 'drop-loglog'] + for flags in (['-B', '-S'], ['-B', '-O', '-S']): + out = subprocess.run([sys.executable, *flags, script], + check=True, capture_output=True, timeout=300).stdout + if out != expected: + raise RuntimeError('output differs from RESULTS.json under ' + ' '.join(flags)) + for m in mutants: + rc = subprocess.run([sys.executable, *flags, script, '--mutant', m], + capture_output=True, timeout=300).returncode + if rc != 1: + raise RuntimeError('mutant ' + m + ' not rejected') + print(json.dumps({'scientific_effect': 'NONE', 'mathematical_acceptance': False})) + PY + git diff --exit-code diff --git a/frontiers/c6_sharpened_20260929/PROOF.md b/frontiers/c6_sharpened_20260929/PROOF.md new file mode 100644 index 000000000..5f96d0f67 --- /dev/null +++ b/frontiers/c6_sharpened_20260929/PROOF.md @@ -0,0 +1,200 @@ +# C6 successor: a super-exponential Rouché-degree tail, and the bound in every dimension + +Object: CL-C6-SHARPENED-20260929-v1. +Author: Anthropic Claude (Claude Code session `session_017Mi3hxjaxV45x6zo6o1ee3`). +Disposition: AUTHOR-SIDE CANDIDATE; nonauthor review required. +Scientific effect: NONE. No register, graph, catalog, lemma flag or prize changes. + +**Relation to #140.** This is a successor to CL-C6-FACTORIAL-20260929-v1 (Math-#140, +`frontiers/c6_factorial_moment_20260929/PROOF.md`). That file is **not modified**; this note cites it as [C6]. +- The route in §3 was proposed by OpenAI in + [#140 comment 5893989971](https://github.com/d6g8k5htny-coder/Math-/pull/140#issuecomment-5893989971). +- Its arithmetic was checked in [5894002968](https://github.com/d6g8k5htny-coder/Math-/pull/140#issuecomment-5894002968). +- This note writes the route out in full and adds the dimension-`d` form. + +## 1. Statements + +The setting, law `Q_r^W`, window `I_r` and count `N` (window critical points off the pins) are as in [C6] §1, on +`T^d = ℝ^d/Tℤ^d`. The pins, the weight `W_r = F_d(H_M)F_{d−1}(H_S)` and the normalizer `Z_r` are those of +`imports/lifetime_parent_20260925` [LP] in fixed dimension `d`. Put `L = log(1/r)`. + +**Theorem S₂ (`d = 2`).** There are `C` and `0 < r_* < e^{−e}` such that for `0 < r ≤ r_*` + + E_{Q_r^W} N(N−1) ≤ C r³ [L / log L]², + E_{Q_r^W} (N)_q ≤ C_q r³ [L / log L]^{2(q−1)} (each fixed integer q ≥ 2). (S2) + +This sharpens [C6]'s `r³ L²`. + +**Theorem S_d (every fixed `d ≥ 2`, conditional).** Assume Theorem G_d of Math-#141: +`E_{Q_r^W} N ≤ C r³`, which is under review. Then + + E_{Q_r^W} N(N−1) ≤ C r³ [L / log L]^d, E_{Q_r^W} (N)_q ≤ C_q r³ [L / log L]^{d(q−1)}. (Sd) + +For `d = 2`, G_2 is the reviewed (I5), so S₂ is unconditional. + +**Consequence.** With the recorded lower obstruction `E N(N−1) ≥ 2c r³`, the planar optimal order lies between `r³` +and `r³[L/log L]²`. `Θ(r³)` is **not** claimed. The Palm route of [C6] §9 remains the separate route to it. + +**Inputs.** + +From [C6], used verbatim: +- the pathwise cap (§2); +- Lemma R (§3); +- the protected cover (§5), including the measurability paragraph; +- Lemma D and Lemma E (§6, fixed annuli, unchanged), which OpenAI review 5355120953 accepts in `d = 2`. + +Their `d`-dimensional forms are in §4 below. + +From [LP], accepted in the D1 reconciliation, all in fixed `d`: +- `W_r/r²` has bounded moments of every order ((4.1), (5.1)); +- `Z_r ≥ z_* r²` ((5.5)). + +The first moment: (I5) in `d = 2`; Theorem G_d of #141 in `d ≥ 3`. + +## 2. Growth lemma for the complexified gradient + +**Lemma G.** Uniformly in `r ≤ r_*`, marks, frames and centres `c`, for every `R ≥ 1`, + + E_{Q_r} [ sup_{z ∈ Ω_R(c)} |F(z)|² + sup_{z ∈ Ω_R(c)} ‖DF(z)‖² ] ≤ C e^{C R²}. (G) + +Here `F` is the complexified gradient and `Ω_R(c)` the Hermitian ball in `ℂ^d`. + +*Proof.* +- **Series.** Write `f = Σ_n σ_n(ξ_n cos(n'·x) + ζ_n sin(n'·x))`, with `n' = 2πn/T`, + `σ_n = κ e^{−|n'|²/4}` and i.i.d. standard `ξ, ζ`. +- **Mode bounds.** For `|Im z| ≤ R`, each derivative of order `k ≤ 2` of `cos(n'·z)` or `sin(n'·z)` has modulus + `≤ |n'|^k e^{R|n'|}`. +- **Minkowski.** This gives `‖sup_{Ω_R}|∂^α f|‖_{L²} ≤ C Σ_n |n'|^k e^{−|n'|²/4 + R|n'|}`. +- **Completing the square.** `−s²/4 + Rs = R² − (s/2 − R)²`. The lattice sum of `|n'|^k e^{−(|n'|/2 − R)²}` is + `O((1+R)^{k+d})`: shells of width `O(1)` around radius `2R` carry `O((1+R)^{d−1})` points. Hence + `E sup |∂^α f|² ≤ C (1+R)^{2k+2d} e^{2R²} ≤ C e^{3R²}`. +- **Under `Q_r`.** The regression correction is `c(z)(e − E_r(f))`, with `c(z) = Cov(f(z), E_r)Cov(E_r)^{−1}`. + - `Cov(∂^α f(z), ℓ) = ℓ(∂^α K_T(z − ·))`, and for complex `h = a + ib`, + `|e^{−(h+Tm)·(h+Tm)/2}| = e^{−|a+Tm|²/2 + |b|²/2}`. So these entries grow at most like `C(1+R)^{C} e^{R²/2}`. + - The functionals `E_r` have `C⁴`-dual norms bounded uniformly in `r`, and `Cov(E_r)^{−1}` is bounded ([C6] §4, + [LP] §3). The targets `e` are bounded, and `E|E_r(f)|²` is bounded. +- (G) follows. ∎ + +## 3. Lemma M⁺: a super-exponential degree tail + +Keep [C6]'s cover, its protected annuli `A'_j = {η_j/2 ≤ |z − c_j| ≤ 3η_j}`, its test spheres `ρ ∈ [η_j, 2η_j]` and its +Lemma D, **unchanged**. Change only the Cauchy radius used for the Taylor remainder. + +**Lemma M⁺.** There are `C, m_0` (depending on `d`, `T`, `η`) such that for `m ≥ m_0`, all `j`, `r ≤ r_*`, marks and +frames, + + Q_r(m_j* > m) ≤ C exp(−(m log m)/(8d)). (M+) + +*Proof.* Take the [C6] §5 cover with `η ≤ 1/8`. This is allowed, because `η ∈ (0, T/20)` is free there, and it gives +every `η' := η_j ≤ 1/8`. Fix `j` and write `c = c_j`. + +**(a) Remainder with a growing radius.** +- Put `R_m = √m`, with `m ≥ 16`, so that `2η'/R_m ≤ 1/2`, and `S_m = sup_{Ω_{R_m}(c)} |F|`. +- Cauchy's estimate on complex lines ([C6] §6 (a)) gives, on `Ω_{2η'}(c)`, + + |F − T_m| ≤ S_m Σ_{k>m} (2η'/R_m)^k ≤ S_m a_m, a_m := 2 (2η'/√m)^{m+1} ≤ m^{−m/2}. (3.1) + + The last inequality uses `2η' ≤ 1/4`: `a_m ≤ 2·4^{−m−1} m^{−(m+1)/2} ≤ m^{−m/2}`. + +- By Lemma G, `E S_m² ≤ C e^{Cm}`. The Lipschitz constant `L = sup_{Ω_{3η'}(c)} ‖DF‖` is on the **fixed** annulus + and has bounded moments. + +**(b) Failure means near-zeros on every sphere.** This is [C6] §6 (b), with new cutoffs: + + λ_m = m^{m/(8d)}, ε_m = 2 λ_m a_m, δ_m = ε_m / λ_m = 2 a_m. + +- If `m_j* > m`, every sphere `∂Ω_ρ`, `ρ ∈ [η', 2η']`, has a point `z_ρ` with `|T_m(z_ρ)| ≤ S_m a_m`. +- On `{S_m ≤ λ_m, L ≤ λ_m}` this gives `|F(z_ρ)| ≤ ε_m`, and then `|F| ≤ 2ε_m` on `B(z_ρ, δ_m)`. +- Radii `2δ_m` apart give disjoint balls inside `A'_j` once `δ_m ≤ η'/2`, which holds for `m ≥ m_0`. So + + V := Leb_{2d}{z ∈ A'_j : |F(z)| ≤ 2ε_m} ≥ c_d η' δ_m^{2d−1}. (3.2) + +**(c) Markov.** The `d`-dimensional Lemma D (§4) gives `E V ≤ C ε_m^{2d}(1 + log(1/ε_m))`. Hence + + Q_r(m_j* > m, S_m ≤ λ_m, L ≤ λ_m) ≤ C λ_m^{2d−1} ε_m (1 + log(1/ε_m)) / η' + ≤ C λ_m^{2d} a_m · m log m ≤ C m^{−m/4} m log m. (3.3) + +Here `λ_m^{2d} = m^{m/4}` and `log(1/ε_m) ≤ m log m`. + +**(d) Truncation.** + + Q_r(S_m > λ_m) ≤ C e^{Cm} m^{−m/(4d)}, Q_r(L > λ_m) ≤ C m^{−m/(4d)}. (3.4) + +**(e) Conclusion.** Adding (3.3) and (3.4) gives `C e^{Cm} m^{−m/(4d)} ≤ C exp(−(m log m)/(8d))` once +`log m ≥ 8dC`. So `m_0` is at least `e^{8dC}`: large, but a constant. ∎ + +**Why the sharpening appears.** With [C6]'s fixed radius `4η'`, the remainder factor is only `2^{−m}`, so the tail is +merely exponential. With radius `√m`, the remainder factor `m^{−m/2}` is super-exponential. The price is the +`e^{Cm}` growth of `S_m` (Lemma G), and that price is still super-exponentially small against `m^{−m/(4d)}`. + +## 4. The `d`-dimensional forms of [C6]'s Lemmas R, D and E, and of the cover + +- **Lemma R in `ℂ^d`.** + - Rouché for holomorphic maps `ℂ^d → ℂ^d` on `Ω_ρ` is the degree argument of [C6] §3, word for word. + - A compact analytic subset of `Ω_ρ` is finite. + - The refined Bézout inequality bounds the isolated zeros of `d` polynomials of degree `≤ m` in `d` variables by + `m^d`, counted with multiplicity. + - So a ball whose Rouché test passes at degree `m` carries at most `m^d` real critical points. +- **Lemma E on `ℤ^d`.** Shifts `τ_{e_1}, …, τ_{e_d}` commute, and distinct characters of `ℤ^d` have distinct + eigenvalue `d`-tuples. The proof of [C6] §6 is unchanged. +- **Lemma D in `d` dimensions.** + - `G = (Re F(z), Im F(z)/t) ∈ ℝ^{2d}`, with `t = |Im z|`. Its `t = 0` limit is `(∇f(x), D²f(x)u)`. + - At `t = 0` the polynomial symbols `i a·n'` and `−(b·n')(u·n')` vanish only for `a = b = 0`. + - At `t > 0` the points `z, z̄` are non-real and distinct from the pins mod `Tℤ^d`. + - Compactness then gives a uniform floor for the conditional covariance. Hence + `Q_r(|F(z)| ≤ 2ε) ≤ C ε^d (ε/t)^d = C ε^{2d}/t^d`, capped by 1. + - Over the annulus, `∫_{|t|≤R} min(1, ε^{2d}/|t|^d) d^d t = |S^{d−1}| ε^{2d}(1/d + log(R/ε²))`. + So `E V ≤ C ε^{2d}(1 + log(1/ε))`. +- **Cover.** [C6] §5's construction (big ball at the midpoint, a small-ball net of `{|x| ≥ 7η/8}`) is dimension-free. + The real slices of the annuli stay `≥ η/4` from the pins. + +## 5. Proof of Theorems S₂ and S_d + +Put `Ψ = Σ_j (m_j*)^d`, which is `≥` the total critical count by §4 (Lemma R). By (M+), + + Q_r(Ψ > x) ≤ Σ_j Q_r(m_j* > (x/J')^{1/d}) ≤ C exp(−c x^{1/d} log x), J' = number of balls. (5.1) + +In particular `sup_r E_{Q_r} Ψ^p < ∞` for all `p`. + +**The cap.** As in [C6] (2.1), `N(N−1) ≤ λN + Ψ² 1{Ψ > λ}` pathwise. Then: +- `E_{Q_r^W} N ≤ C r³`, by (I5) for `d = 2` and by G_d for `d ≥ 3`. +- `E_{Q_r^W}[Ψ² 1{Ψ>λ}] ≤ Z_r^{−1}‖W_r‖_{L²}‖Ψ² 1{Ψ>λ}‖_{L²} ≤ C (E Ψ⁸)^{1/4} Q_r(Ψ > λ)^{1/4}`, by [LP] (5.1) and + (5.5). + +**Choice of `λ`.** Take `λ = A^d [L/log L]^d`, with `A` large. For `r ≤ r_*`, +`log λ = d(log A + log L − log log L) ≥ log L`, so `λ^{1/d} log λ ≥ A L`. By (5.1), +`Q_r(Ψ > λ)^{1/4} ≤ C exp(−c A L/4) ≤ r³` once `A ≥ 12/c`. Hence + + E_{Q_r^W} N(N−1) ≤ C r³ λ + C r³ ≤ C r³ [L/log L]^d. + +**Factorial moments.** The `(N)_q` bound is identical, using `(N)_q ≤ λ^{q−1}N + Ψ^q 1{Ψ > λ}`. ∎ + +## 6. Scope + +**Claimed.** Lemmas G and M⁺, the `d`-dimensional forms of §4, Theorem S₂, and Theorem S_d **conditional on G_d** +(#141). All constants are existential, for fixed `T` and compact marks. + +**Not claimed.** +- `Θ(r³)`; +- the witness-conditioned (M′) or the marked upgrade (H) of [C6] §9; +- numerical constants; +- uniformity in `T`, `d`, or as `k ↓ 0`; +- any register, graph or catalog change; +- any change to [C6]'s reviewed bytes. + +A defect in Lemma G's uniformity under `Q_r` blocks both theorems. + +**Prior art.** Gass–Stecconi (arXiv:2305.17586, PTRF 2024) prove finiteness of moments of critical-point counts. No +novelty claim is made. + +## 7. Finite companion + +`sharp_check.py` (stdlib) checks: +- the completing-square identity; +- the lattice-shell sum growth against `(1+R)^{k+d}e^{R²}` for `d = 2, 3`; +- the remainder bound (3.1) exactly at perfect-square `m`; +- the exponent bookkeeping `λ_m^{2d} = m^{m/4}` and `λ_m² = m^{m/(4d)}`; +- the Lemma D radial integral in `d = 2, 3`; +- the choice `λ = A^d[L/log L]^d`, with `λ^{1/d} log λ ≥ A L`, at `r = 2^{−k}`. + +Mutants `fixed-radius`, `wrong-cutoff`, `planar-cap` and `drop-loglog` must fail. diff --git a/frontiers/c6_sharpened_20260929/RESULTS.json b/frontiers/c6_sharpened_20260929/RESULTS.json new file mode 100644 index 000000000..f5adc3ed1 --- /dev/null +++ b/frontiers/c6_sharpened_20260929/RESULTS.json @@ -0,0 +1,13 @@ +{ + "checks": { + "EXPONENTS": true, + "LAMBDA": true, + "RADIAL": true, + "REMAINDER": true, + "SHELL": true, + "SQUARE": true + }, + "object": "CL-C6-SHARPENED-20260929-v1", + "passed": true, + "scope": "finite identities, exponent bookkeeping and quadratures only; the proof is PROOF.md" +} diff --git a/frontiers/c6_sharpened_20260929/SOURCE_FILES.json b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json new file mode 100644 index 000000000..a93c118c3 --- /dev/null +++ b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json @@ -0,0 +1,52 @@ +{ + "schema": 1, + "scientific_effect": "NONE", + "disposition": "AUTHOR_SIDE_CANDIDATE", + "successor_of": "CL-C6-FACTORIAL-20260929-v1", + "consumes": [ + { + "id": "C6", + "path": "frontiers/c6_factorial_moment_20260929/PROOF.md", + "sha256": "ae00625a460b2d5597e3e4669bb468ffb60e85b5a259854f856577397d9ebd2d", + "git_blob": "6cb705ef44100d5a32d4c852cf7e460d520f90f5", + "pr": "Math-#140" + }, + { + "id": "LP", + "path": "imports/lifetime_parent_20260925/UNIFORM_MATRIX_CAP_AND_LIFETIME.md", + "sha256": "9350ad6eaba6626b93c3dedeef9e2ff816e5cdf1c8318e85fb27499141c84bc7" + }, + { + "id": "I5", + "records": [ + "reviews/d5_i5_planar_f_20260928/REVIEW.md", + "reviews/d5_intermediate_window_claude_20260928/REVIEW.md" + ] + }, + { + "id": "G_d", + "pr": "Math-#141", + "status": "under review; Theorem S_d is conditional on it" + } + ], + "files": [ + { + "path": "PROOF.md", + "bytes": 10903, + "sha256": "8ffbf45d29625619d48fb6857289be1b2d846a4b8f2dd88a10878170b612a1cf", + "git_blob": "5f96d0f67faa471b9f1e7f700a31dde36dfa6a24" + }, + { + "path": "RESULTS.json", + "bytes": 305, + "sha256": "72690d42292c2d11dff2f77c608eaeeb5b246a3bb7da31dcbb16f511a47e00a7", + "git_blob": "f5adc3ed12bddd5858bbb4b3164f813766f6c9be" + }, + { + "path": "sharp_check.py", + "bytes": 5477, + "sha256": "749a36db92223defbda33ef9bbbd3676578f8a48d56d5e0b657b05008f8744e8", + "git_blob": "407967d30843620622870d7f569d845825139e4e" + } + ] +} diff --git a/frontiers/c6_sharpened_20260929/sharp_check.py b/frontiers/c6_sharpened_20260929/sharp_check.py new file mode 100644 index 000000000..407967d30 --- /dev/null +++ b/frontiers/c6_sharpened_20260929/sharp_check.py @@ -0,0 +1,134 @@ +"""Finite companion for CL-C6-SHARPENED-20260929-v1 (PROOF.md). Standard library only. + + SQUARE -s^2/4 + R s = R^2 - (s/2 - R)^2 exactly (Lemma G's completing the square). + SHELL the lattice sum sum_{n in Z^d} |n'|^k exp(-|n'|^2/4 + R|n'|) (n' = 2 pi n, T = 1) stays below + (1+R)^(k+d) e^(R^2) times a fixed constant for R = 1..6, d = 2, 3 (Lemma G's growth rate). + REMAINDER a_m = 2 (2 eta'/sqrt m)^(m+1) <= m^(-m/2) exactly at perfect squares m = k^2, eta' = 1/8 (3.1). + EXPONENTS lambda_m = m^(m/(8d)): (lambda_m^(2d)) exponent m/4 and lambda_m^2 exponent m/(4d), exactly; the Markov + term m^(-m/4) m log m and the truncation term e^(Cm) m^(-m/(4d)) both fall below exp(-m log m/(8d)) once + log m >= 8dC (checked in logs for C = 1 at m = e^(8d+1), e^(8d+4), e^(8d+8)). + RADIAL int_{|t|<=R} min(1, eps^(2d)/|t|^d) d^d t = |S^(d-1)| eps^(2d) (1/d + log(R/eps^2)), d = 2, 3, quadrature. + LAMBDA lambda = A^d [L/log L]^d satisfies lambda^(1/d) log lambda >= A L at r = 2^-k, L = log(1/r), d = 2, 3. +Mutants (each must fail): fixed-radius, wrong-cutoff, planar-cap, drop-loglog. +""" +import argparse +import itertools +import json +import math +import sys +from fractions import Fraction as F + +MUTANTS = ("fixed-radius", "wrong-cutoff", "planar-cap", "drop-loglog") +MUT = None + + +def check_square(): + ok = True + for R in (F(1), F(5, 2), F(7)): + for s in (F(0), F(1, 3), F(4), F(19, 2)): + ok &= -s * s / 4 + R * s == R * R - (s / 2 - R) ** 2 + return ok + + +def shell_sum(R, d, k, cutoff): + tot = 0.0 + rng = range(-cutoff, cutoff + 1) + for n in itertools.product(rng, repeat=d): + s = 2 * math.pi * math.sqrt(sum(x * x for x in n)) + tot += s ** k * math.exp(-s * s / 4 + R * s) + return tot + + +def check_shell(): + ok = True + for d, cutoff in ((2, 12), (3, 7)): + ratios = [] + for R in (1, 2, 3, 4, 5, 6): + val = shell_sum(R, d, 2, cutoff) + ratios.append(val / ((1 + R) ** (2 + d) * math.exp(R * R))) + ok &= max(ratios) < 50 and all(x > 0 for x in ratios) + return ok + + +def check_remainder(): + ok = True + eta = F(1, 8) + for k in (4, 5, 6, 8, 10): + m = k * k + a_m = 2 * (2 * eta / k) ** (m + 1) # sqrt(m) = k exactly + if MUT == "fixed-radius": + a_m = 2 * (2 * eta / (4 * eta)) ** (m + 1) # radius 4 eta': factor 2^-m only + ok &= a_m <= F(1, k ** m) # m^(-m/2) = k^(-m) + return ok + + +def check_exponents(): + ok = True + for d in (2, 3, 4): + lam_exp = F(1, 8 * d) if MUT != "wrong-cutoff" else F(1, 4 * d) + markov = 2 * d * lam_exp - F(1, 2) # exponent of m^(m*.) in lambda^(2d) a_m + trunc = -2 * lam_exp # exponent in the lambda^-2 truncation + ok &= markov == -F(1, 4) and trunc == -F(1, 4 * d) + for shift in (1, 4, 8): # m >= e^(8d C) with C = 1: the regime m >= m_0 + m = math.exp(8 * d + shift) + lm = math.log(m) + target = -m * lm / (8 * d) + ok &= float(markov) * m * lm + math.log(m * lm) <= target + ok &= m + float(trunc) * m * lm <= target # e^(Cm) with C = 1 + return ok + + +def radial(eps, R, d, n=6000): + a, b = math.log(eps * eps), math.log(R) + h = (b - a) / n + area = 2 * math.pi if d == 2 else 4 * math.pi # |S^(d-1)| + g = lambda u: min(1.0, eps ** (2 * d) / math.exp(u * d)) * area * math.exp(u * d) + s = g(a) + g(b) + sum((4 if i % 2 else 2) * g(a + i * h) for i in range(1, n)) + inner = area * eps ** (2 * d) / d # |t| <= eps^2, where min(...) = 1 + return inner + s * h / 3 + + +def check_radial(): + ok = True + for d in (2, 3): + area = 2 * math.pi if d == 2 else 4 * math.pi + for e in (1e-1, 1e-2, 1e-3): + closed = area * e ** (2 * d) * (1 / d + math.log(1.0 / e ** 2)) + ok &= abs(radial(e, 1.0, d) - closed) <= 1e-6 * closed + return ok + + +def check_lambda(): + ok = True + A = 12.0 + for d in (2, 3): + cap_d = 2 if MUT == "planar-cap" else d + for k in (2 ** 10, 2 ** 20, 2 ** 40): + L = k * math.log(2) + base = L if MUT == "drop-loglog" else L / math.log(L) + lam = (A * base) ** cap_d + lhs = lam ** (1.0 / d) * math.log(lam) + ok &= lhs >= A * L + if MUT != "drop-loglog": + ok &= lam <= (A * L / math.log(L)) ** d * 1.000001 # the claimed size, no larger + else: + ok &= lam <= (A * L / math.log(L)) ** d + return ok + + +def main(): + global MUT + ap = argparse.ArgumentParser() + ap.add_argument("--mutant", choices=MUTANTS) + MUT = ap.parse_args().mutant + checks = {"SQUARE": check_square(), "SHELL": check_shell(), "REMAINDER": check_remainder(), + "EXPONENTS": check_exponents(), "RADIAL": check_radial(), "LAMBDA": check_lambda()} + passed = all(checks.values()) and len(checks) == 6 + print(json.dumps({"object": "CL-C6-SHARPENED-20260929-v1", "checks": checks, "passed": passed, + "scope": "finite identities, exponent bookkeeping and quadratures only; the proof is PROOF.md"}, + indent=2, sort_keys=True)) + return 0 if passed else 1 + + +if __name__ == "__main__": + sys.exit(main()) From aee04cd1eca597cfd45ca88df4b2fdc1728717f9 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 29 Sep 2026 16:30:16 +0000 Subject: [PATCH 2/4] C6 successor v1.1: cite [C6] v1.2; (I5) and G_d inputs at exact rows (scientific effect NONE) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - Cites CL-C6-FACTORIAL v1.2 (09763ca) and refreshes its consumed hash. - S2 does not depend on #141; it consumes (I5) exactly as [C6] v1.2 §1 does (window first moment, D5 reading rule, inherited (P2) continuum HOLD). - G_d input records the xAI Slice A split (algebra ACCEPT, continuum HOLD). - Lemma R in C^d points to the explicit component step of [C6] v1.2. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_017Mi3hxjaxV45x6zo6o1ee3 --- frontiers/c6_sharpened_20260929/PROOF.md | 17 ++++++++++++---- .../c6_sharpened_20260929/SOURCE_FILES.json | 20 +++++++++++-------- 2 files changed, 25 insertions(+), 12 deletions(-) diff --git a/frontiers/c6_sharpened_20260929/PROOF.md b/frontiers/c6_sharpened_20260929/PROOF.md index 5f96d0f67..7c46f977e 100644 --- a/frontiers/c6_sharpened_20260929/PROOF.md +++ b/frontiers/c6_sharpened_20260929/PROOF.md @@ -3,6 +3,8 @@ Object: CL-C6-SHARPENED-20260929-v1. Author: Anthropic Claude (Claude Code session `session_017Mi3hxjaxV45x6zo6o1ee3`). Disposition: AUTHOR-SIDE CANDIDATE; nonauthor review required. +Version: v1.1. It cites [C6] at v1.2 (Math-#140 `09763ca`) and aligns the inputs of §1 with [C6] v1.2 §1. There is no +theorem change. Scientific effect: NONE. No register, graph, catalog, lemma flag or prize changes. **Relation to #140.** This is a successor to CL-C6-FACTORIAL-20260929-v1 (Math-#140, @@ -30,7 +32,9 @@ This sharpens [C6]'s `r³ L²`. E_{Q_r^W} N(N−1) ≤ C r³ [L / log L]^d, E_{Q_r^W} (N)_q ≤ C_q r³ [L / log L]^{d(q−1)}. (Sd) -For `d = 2`, G_2 is the reviewed (I5), so S₂ is unconditional. +For `d = 2`, G_2 is (I5), so S₂ does not depend on #141. It consumes (I5) exactly as [C6] v1.2 §1 does: as a +window first moment, at the scope of the D5 reading rule `reviews/d5_reconciliation_20260929/`, including the +(P2) continuum HOLD (#111) that (I5) inherits through (C2). **Consequence.** With the recorded lower obstruction `E N(N−1) ≥ 2c r³`, the planar optimal order lies between `r³` and `r³[L/log L]²`. `Θ(r³)` is **not** claimed. The Palm route of [C6] §9 remains the separate route to it. @@ -39,7 +43,7 @@ and `r³[L/log L]²`. `Θ(r³)` is **not** claimed. The Palm route of [C6] §9 r From [C6], used verbatim: - the pathwise cap (§2); -- Lemma R (§3); +- Lemma R (§3), with the v1.2 explicit form of its step 1; - the protected cover (§5), including the measurability paragraph; - Lemma D and Lemma E (§6, fixed annuli, unchanged), which OpenAI review 5355120953 accepts in `d = 2`. @@ -49,7 +53,11 @@ From [LP], accepted in the D1 reconciliation, all in fixed `d`: - `W_r/r²` has bounded moments of every order ((4.1), (5.1)); - `Z_r ≥ z_* r²` ((5.5)). -The first moment: (I5) in `d = 2`; Theorem G_d of #141 in `d ≥ 3`. +The first moment: +- `d = 2`: (I5), cited at the xAI row "I5 … ACCEPT as corollary of C2 + I4 + D4 A" of + `reviews/d5_i5_planar_f_20260928/REVIEW.md`, under the D5 reading rule; +- `d ≥ 3`: Theorem G_d of #141. It is under review, and its continuum count carries an xAI HOLD at the Slice A + verdict ([#141 comment 5894274124](https://github.com/d6g8k5htny-coder/Math-/pull/141#issuecomment-5894274124)). ## 2. Growth lemma for the complexified gradient @@ -131,7 +139,8 @@ merely exponential. With radius `√m`, the remainder factor `m^{−m/2}` is sup - **Lemma R in `ℂ^d`.** - Rouché for holomorphic maps `ℂ^d → ℂ^d` on `Ω_ρ` is the degree argument of [C6] §3, word for word. - - A compact analytic subset of `Ω_ρ` is finite. + - A compact analytic subset of `Ω_ρ` is finite. As in [C6] v1.2 step 1, no positive-dimensional component of + `Z(P)` meets `Ω_ρ` once `Z(P) ∩ ∂Ω_ρ = ∅`. - The refined Bézout inequality bounds the isolated zeros of `d` polynomials of degree `≤ m` in `d` variables by `m^d`, counted with multiplicity. - So a ball whose Rouché test passes at degree `m` carries at most `m^d` real critical points. diff --git a/frontiers/c6_sharpened_20260929/SOURCE_FILES.json b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json index a93c118c3..e0893f18a 100644 --- a/frontiers/c6_sharpened_20260929/SOURCE_FILES.json +++ b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json @@ -7,9 +7,10 @@ { "id": "C6", "path": "frontiers/c6_factorial_moment_20260929/PROOF.md", - "sha256": "ae00625a460b2d5597e3e4669bb468ffb60e85b5a259854f856577397d9ebd2d", - "git_blob": "6cb705ef44100d5a32d4c852cf7e460d520f90f5", - "pr": "Math-#140" + "sha256": "c5dcc611810c24b115f44ef9cce468e83618f3f63de385e7157a46076a4762d1", + "git_blob": "af6c336301d1ec6cf3b2021e8b7478543200d400", + "pr": "Math-#140", + "version": "v1.2 (09763ca)" }, { "id": "LP", @@ -18,23 +19,26 @@ }, { "id": "I5", + "path": "frontiers/intermediate_window_20260928/PROOF.md", + "sha256": "b3eb9456d7058b5b78149cfca4e072679beda85cd1a0293d209664a542db66cf", "records": [ "reviews/d5_i5_planar_f_20260928/REVIEW.md", "reviews/d5_intermediate_window_claude_20260928/REVIEW.md" - ] + ], + "reading_rule": "reviews/d5_reconciliation_20260929/RECONCILIATION.md" }, { "id": "G_d", "pr": "Math-#141", - "status": "under review; Theorem S_d is conditional on it" + "status": "under review (xAI Slice A: algebra ACCEPT, continuum count HOLD); Theorem S_d is conditional on it" } ], "files": [ { "path": "PROOF.md", - "bytes": 10903, - "sha256": "8ffbf45d29625619d48fb6857289be1b2d846a4b8f2dd88a10878170b612a1cf", - "git_blob": "5f96d0f67faa471b9f1e7f700a31dde36dfa6a24" + "bytes": 11756, + "sha256": "10dd2bd411479ce68dcf4a4f5053881f6ac530d2be2e192fa8b09f3e404cc56c", + "git_blob": "7c46f977eb8015659839d8dda1bdc1c2e45ca98a" }, { "path": "RESULTS.json", From 0a4890c0f981321c86b495fb2fa56fd96366bd89 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 29 Sep 2026 16:34:27 +0000 Subject: [PATCH 3/4] C6 successor: cite [C6] v1.3 and refresh its consumed hash (scientific effect NONE) Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_017Mi3hxjaxV45x6zo6o1ee3 --- frontiers/c6_sharpened_20260929/PROOF.md | 2 +- frontiers/c6_sharpened_20260929/SOURCE_FILES.json | 12 ++++++------ 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/frontiers/c6_sharpened_20260929/PROOF.md b/frontiers/c6_sharpened_20260929/PROOF.md index 7c46f977e..3cccb18e7 100644 --- a/frontiers/c6_sharpened_20260929/PROOF.md +++ b/frontiers/c6_sharpened_20260929/PROOF.md @@ -3,7 +3,7 @@ Object: CL-C6-SHARPENED-20260929-v1. Author: Anthropic Claude (Claude Code session `session_017Mi3hxjaxV45x6zo6o1ee3`). Disposition: AUTHOR-SIDE CANDIDATE; nonauthor review required. -Version: v1.1. It cites [C6] at v1.2 (Math-#140 `09763ca`) and aligns the inputs of §1 with [C6] v1.2 §1. There is no +Version: v1.1. It cites [C6] at v1.3 (Math-#140) and aligns the inputs of §1 with [C6] §1 (v1.2 onward). There is no theorem change. Scientific effect: NONE. No register, graph, catalog, lemma flag or prize changes. diff --git a/frontiers/c6_sharpened_20260929/SOURCE_FILES.json b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json index e0893f18a..bebdeee4c 100644 --- a/frontiers/c6_sharpened_20260929/SOURCE_FILES.json +++ b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json @@ -7,10 +7,10 @@ { "id": "C6", "path": "frontiers/c6_factorial_moment_20260929/PROOF.md", - "sha256": "c5dcc611810c24b115f44ef9cce468e83618f3f63de385e7157a46076a4762d1", - "git_blob": "af6c336301d1ec6cf3b2021e8b7478543200d400", + "sha256": "b19927563e2e9ff8f38648b61cd3f2c09ee1a081494b98cc285283554746e6d3", + "git_blob": "f5bd013b7134be9efd246fc9bbb8d18ad5a728bc", "pr": "Math-#140", - "version": "v1.2 (09763ca)" + "version": "v1.3 (b3e8823)" }, { "id": "LP", @@ -36,9 +36,9 @@ "files": [ { "path": "PROOF.md", - "bytes": 11756, - "sha256": "10dd2bd411479ce68dcf4a4f5053881f6ac530d2be2e192fa8b09f3e404cc56c", - "git_blob": "7c46f977eb8015659839d8dda1bdc1c2e45ca98a" + "bytes": 11755, + "sha256": "72d50c977190a45e5796325ae03fe1bdd37eec58807f8813dbc5da8f7827fbe6", + "git_blob": "3cccb18e7bc726c1782ae6c17c255b508e30605b" }, { "path": "RESULTS.json", From 2a48c13154df29c190c5e8e86fefba99f12c43ab Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 29 Sep 2026 16:37:17 +0000 Subject: [PATCH 4/4] C6 successor v1.2: Lemma G for real centres; exact-formula log(1/eps_m) step with enlarged m_0 (scientific effect NONE) From the OpenAI pickup 5894431345 on #142: - Lemma G is stated for real centres c in R^d (every cover centre is real); for Im c != 0 no R-only bound holds. - (3.3): log(1/eps_m) <= m log m is derived from the exact eps_m = 4 m^(m/(8d)) (2 eta'/sqrt m)^(m+1), valid once log m >= 4 log(1/(2 eta')); m_0 >= max(e^(8dC), (4/eta)^4, 16). - sharp_check.py: new LOGEPS group and small-m0 mutant (5 mutants rejected, output identical under -O); workflow mutant list updated; manifest refreshed. No theorem change. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_017Mi3hxjaxV45x6zo6o1ee3 --- .github/workflows/c6-sharpened.yml | 2 +- frontiers/c6_sharpened_20260929/PROOF.md | 32 +++++++++++++++---- frontiers/c6_sharpened_20260929/RESULTS.json | 1 + .../c6_sharpened_20260929/SOURCE_FILES.json | 18 +++++------ .../c6_sharpened_20260929/sharp_check.py | 28 +++++++++++++--- 5 files changed, 61 insertions(+), 20 deletions(-) diff --git a/.github/workflows/c6-sharpened.yml b/.github/workflows/c6-sharpened.yml index 3a8a205e2..363d116be 100644 --- a/.github/workflows/c6-sharpened.yml +++ b/.github/workflows/c6-sharpened.yml @@ -43,7 +43,7 @@ jobs: raise ValueError('source identity mismatch: ' + entry['path']) script = str(root / 'sharp_check.py') expected = (root / 'RESULTS.json').read_bytes() - mutants = ['fixed-radius', 'wrong-cutoff', 'planar-cap', 'drop-loglog'] + mutants = ['fixed-radius', 'wrong-cutoff', 'planar-cap', 'drop-loglog', 'small-m0'] for flags in (['-B', '-S'], ['-B', '-O', '-S']): out = subprocess.run([sys.executable, *flags, script], check=True, capture_output=True, timeout=300).stdout diff --git a/frontiers/c6_sharpened_20260929/PROOF.md b/frontiers/c6_sharpened_20260929/PROOF.md index 3cccb18e7..70ba19726 100644 --- a/frontiers/c6_sharpened_20260929/PROOF.md +++ b/frontiers/c6_sharpened_20260929/PROOF.md @@ -3,8 +3,13 @@ Object: CL-C6-SHARPENED-20260929-v1. Author: Anthropic Claude (Claude Code session `session_017Mi3hxjaxV45x6zo6o1ee3`). Disposition: AUTHOR-SIDE CANDIDATE; nonauthor review required. -Version: v1.1. It cites [C6] at v1.3 (Math-#140) and aligns the inputs of §1 with [C6] §1 (v1.2 onward). There is no -theorem change. +Version: v1.2. +- v1.1 cites [C6] at v1.3 (Math-#140) and aligns the inputs of §1 with [C6] §1 (v1.2 onward). +- v1.2 applies the two wording checks from the OpenAI pickup ([#142 comment 5894431345](https://github.com/d6g8k5htny-coder/Math-/pull/142#issuecomment-5894431345)): + - Lemma G is stated for real centres; + - the bound `log(1/ε_m) ≤ m log m` in (3.3) is derived from the exact `a_m`, with `m_0` enlarged to absorb + `log(1/(2η'))`. +- There is no theorem change. Scientific effect: NONE. No register, graph, catalog, lemma flag or prize changes. **Relation to #140.** This is a successor to CL-C6-FACTORIAL-20260929-v1 (Math-#140, @@ -61,11 +66,14 @@ The first moment: ## 2. Growth lemma for the complexified gradient -**Lemma G.** Uniformly in `r ≤ r_*`, marks, frames and centres `c`, for every `R ≥ 1`, +**Lemma G.** Uniformly in `r ≤ r_*`, marks, frames and **real** centres `c ∈ ℝ^d`, for every `R ≥ 1`, E_{Q_r} [ sup_{z ∈ Ω_R(c)} |F(z)|² + sup_{z ∈ Ω_R(c)} ‖DF(z)‖² ] ≤ C e^{C R²}. (G) Here `F` is the complexified gradient and `Ω_R(c)` the Hermitian ball in `ℂ^d`. +- **Real centres.** The centres must be real, as every centre of the [C6] cover is. For real `c`, the ball `Ω_R(c)` + lies in `{|Im z| ≤ R}`, which is what the mode bounds below use. +- **Complex centres fail.** For `Im c ≠ 0`, the ball reaches `|Im z| = |Im c| + R`, and no bound in `R` alone holds. *Proof.* - **Series.** Write `f = Σ_n σ_n(ξ_n cos(n'·x) + ζ_n sin(n'·x))`, with `n' = 2πn/T`, @@ -122,14 +130,24 @@ every `η' := η_j ≤ 1/8`. Fix `j` and write `c = c_j`. Q_r(m_j* > m, S_m ≤ λ_m, L ≤ λ_m) ≤ C λ_m^{2d−1} ε_m (1 + log(1/ε_m)) / η' ≤ C λ_m^{2d} a_m · m log m ≤ C m^{−m/4} m log m. (3.3) -Here `λ_m^{2d} = m^{m/4}` and `log(1/ε_m) ≤ m log m`. +Here `λ_m^{2d} = m^{m/4}`. + +**The logarithm.** The bound `log(1/ε_m) ≤ m log m` uses the exact formula for `a_m`, not merely `a_m ≤ m^{−m/2}`. +An upper bound on `a_m` does not bound `log(1/ε_m)` from above. From `ε_m = 2λ_m a_m = 4 m^{m/(8d)}(2η'/√m)^{m+1}`, + + log(1/ε_m) = (m+1)(½ log m + ℓ') − (m log m)/(8d) − log 4, ℓ' := log(1/(2η')) ≥ log 4. + +For `m ≥ 3` we have `m + 1 ≤ (4/3)m`, so `log(1/ε_m) ≤ m((2/3) log m + (4/3)ℓ') ≤ m log m` once `log m ≥ 4ℓ'`. +The cover has finitely many radii (`η_0 = η`, `η_j = η/8`), so `m_0` is enlarged once, to absorb +`ℓ'_max = log(4/η)`. **(d) Truncation.** Q_r(S_m > λ_m) ≤ C e^{Cm} m^{−m/(4d)}, Q_r(L > λ_m) ≤ C m^{−m/(4d)}. (3.4) **(e) Conclusion.** Adding (3.3) and (3.4) gives `C e^{Cm} m^{−m/(4d)} ≤ C exp(−(m log m)/(8d))` once -`log m ≥ 8dC`. So `m_0` is at least `e^{8dC}`: large, but a constant. ∎ +`log m ≥ 8dC`. So `m_0 ≥ max(e^{8dC}, (4/η)^4, 16)`, where the middle term comes from the logarithm step in (c): +large, but a constant. ∎ **Why the sharpening appears.** With [C6]'s fixed radius `4η'`, the remainder factor is only `2^{−m}`, so the tail is merely exponential. With radius `√m`, the remainder factor `m^{−m/2}` is super-exponential. The price is the @@ -203,7 +221,9 @@ novelty claim is made. - the lattice-shell sum growth against `(1+R)^{k+d}e^{R²}` for `d = 2, 3`; - the remainder bound (3.1) exactly at perfect-square `m`; - the exponent bookkeeping `λ_m^{2d} = m^{m/4}` and `λ_m² = m^{m/(4d)}`; +- the logarithm step of (3.3), from the exact `ε_m`, for the cover radii `η' = 1/8, 1/64`, at `m ≥ max(3, e^{4ℓ'})`; - the Lemma D radial integral in `d = 2, 3`; - the choice `λ = A^d[L/log L]^d`, with `λ^{1/d} log λ ≥ A L`, at `r = 2^{−k}`. -Mutants `fixed-radius`, `wrong-cutoff`, `planar-cap` and `drop-loglog` must fail. +Mutants `fixed-radius`, `wrong-cutoff`, `planar-cap`, `drop-loglog` and `small-m0` must fail. `small-m0` is the +logarithm step at `m = 16` with no enlargement of `m_0`. diff --git a/frontiers/c6_sharpened_20260929/RESULTS.json b/frontiers/c6_sharpened_20260929/RESULTS.json index f5adc3ed1..b5b7ae466 100644 --- a/frontiers/c6_sharpened_20260929/RESULTS.json +++ b/frontiers/c6_sharpened_20260929/RESULTS.json @@ -2,6 +2,7 @@ "checks": { "EXPONENTS": true, "LAMBDA": true, + "LOGEPS": true, "RADIAL": true, "REMAINDER": true, "SHELL": true, diff --git a/frontiers/c6_sharpened_20260929/SOURCE_FILES.json b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json index bebdeee4c..05e0e88d8 100644 --- a/frontiers/c6_sharpened_20260929/SOURCE_FILES.json +++ b/frontiers/c6_sharpened_20260929/SOURCE_FILES.json @@ -36,21 +36,21 @@ "files": [ { "path": "PROOF.md", - "bytes": 11755, - "sha256": "72d50c977190a45e5796325ae03fe1bdd37eec58807f8813dbc5da8f7827fbe6", - "git_blob": "3cccb18e7bc726c1782ae6c17c255b508e30605b" + "bytes": 13278, + "sha256": "9f45274ab59d4f4417d1a51e9c97f5916240fdb0a4640ca6f0d3e9e0aea37dd7", + "git_blob": "70ba19726a9114bfceae04d021095e6c8eed5026" }, { "path": "RESULTS.json", - "bytes": 305, - "sha256": "72690d42292c2d11dff2f77c608eaeeb5b246a3bb7da31dcbb16f511a47e00a7", - "git_blob": "f5adc3ed12bddd5858bbb4b3164f813766f6c9be" + "bytes": 325, + "sha256": "c8ca0fee386ed4ed260ccb00ba40b2b9d4866f8488c7cc1d3cbfb6eece098bf0", + "git_blob": "b5b7ae466c10d5f079411bf4993db5c8d994ba0c" }, { "path": "sharp_check.py", - "bytes": 5477, - "sha256": "749a36db92223defbda33ef9bbbd3676578f8a48d56d5e0b657b05008f8744e8", - "git_blob": "407967d30843620622870d7f569d845825139e4e" + "bytes": 6392, + "sha256": "36b169407ceb4a9b32fbc9882a77cbe02ffa3b889e4d69c6f55227f3d0983504", + "git_blob": "f6bed29b6a882c785f268df67fd681db08a01073" } ] } diff --git a/frontiers/c6_sharpened_20260929/sharp_check.py b/frontiers/c6_sharpened_20260929/sharp_check.py index 407967d30..f6bed29b6 100644 --- a/frontiers/c6_sharpened_20260929/sharp_check.py +++ b/frontiers/c6_sharpened_20260929/sharp_check.py @@ -7,9 +7,11 @@ EXPONENTS lambda_m = m^(m/(8d)): (lambda_m^(2d)) exponent m/4 and lambda_m^2 exponent m/(4d), exactly; the Markov term m^(-m/4) m log m and the truncation term e^(Cm) m^(-m/(4d)) both fall below exp(-m log m/(8d)) once log m >= 8dC (checked in logs for C = 1 at m = e^(8d+1), e^(8d+4), e^(8d+8)). + LOGEPS log(1/eps_m) <= m log m from the exact eps_m = 4 m^(m/(8d)) (2 eta'/sqrt m)^(m+1), for eta' = 1/8, 1/64 + and m >= max(3, e^(4 log(1/(2 eta')))) (the enlarged m_0 of (3.3)); d = 2, 3, 4. RADIAL int_{|t|<=R} min(1, eps^(2d)/|t|^d) d^d t = |S^(d-1)| eps^(2d) (1/d + log(R/eps^2)), d = 2, 3, quadrature. LAMBDA lambda = A^d [L/log L]^d satisfies lambda^(1/d) log lambda >= A L at r = 2^-k, L = log(1/r), d = 2, 3. -Mutants (each must fail): fixed-radius, wrong-cutoff, planar-cap, drop-loglog. +Mutants (each must fail): fixed-radius, wrong-cutoff, planar-cap, drop-loglog, small-m0. """ import argparse import itertools @@ -18,7 +20,7 @@ import sys from fractions import Fraction as F -MUTANTS = ("fixed-radius", "wrong-cutoff", "planar-cap", "drop-loglog") +MUTANTS = ("fixed-radius", "wrong-cutoff", "planar-cap", "drop-loglog", "small-m0") MUT = None @@ -78,6 +80,23 @@ def check_exponents(): return ok +def log_inv_eps(m, eta, d): + return (m + 1) * (0.5 * math.log(m) + math.log(1 / (2 * eta))) - m * math.log(m) / (8 * d) - math.log(4) + + +def check_logeps(): + ok = True + for eta in (1 / 8, 1 / 64): # eta_0 = eta = 1/8, eta_j = eta/8 + m0 = max(3.0, math.exp(4 * math.log(1 / (2 * eta)))) + if MUT == "small-m0": + m0 = 16.0 # no enlargement of m_0 + for d in (2, 3, 4): + for mult in (1, 2, 10, 1000): + m = m0 * mult + ok &= log_inv_eps(m, eta, d) <= m * math.log(m) + return ok + + def radial(eps, R, d, n=6000): a, b = math.log(eps * eps), math.log(R) h = (b - a) / n @@ -122,8 +141,9 @@ def main(): ap.add_argument("--mutant", choices=MUTANTS) MUT = ap.parse_args().mutant checks = {"SQUARE": check_square(), "SHELL": check_shell(), "REMAINDER": check_remainder(), - "EXPONENTS": check_exponents(), "RADIAL": check_radial(), "LAMBDA": check_lambda()} - passed = all(checks.values()) and len(checks) == 6 + "EXPONENTS": check_exponents(), "LOGEPS": check_logeps(), "RADIAL": check_radial(), + "LAMBDA": check_lambda()} + passed = all(checks.values()) and len(checks) == 7 print(json.dumps({"object": "CL-C6-SHARPENED-20260929-v1", "checks": checks, "passed": passed, "scope": "finite identities, exponent bookkeeping and quadratures only; the proof is PROOF.md"}, indent=2, sort_keys=True))