From a6b67882344cc39c76c29f5a6a6787a0e243c630 Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 1 Oct 2026 19:45:32 +0000 Subject: [PATCH 1/2] Candidate density with remainder l^(3/5): reflection parity removes the l^(1/2) term (Theorem P) Author-side proof candidate CL-CANDIDATE-PARITY-RATE-20261001-v1. Scientific effect: NONE. nu_cand = c l^(-1/3) + B_{d,L} + I^cand l^(1/4) + c2 l^(1/3) + O(l^(3/5)) for every d >= 2, improving O(l^(4/11)) (#218) and O(l^(3/7)) (#229); Corollary P': no l^(1/2) term. New: Lemma R (birth integration; independent even/odd parts; point reflection k -> -k; evenness in r), Lemma Pi (Pi = 36k^2 Delta^2 + r^2 (12k Delta V - U^2) + O(r^4)), Lemma K, Lemma D'' (two small eigenvalues), Lemma Lambda (the gap inside the cusp layer costs O(k^2/(1 + kappa))), Lemma U (one two-scale kernel formula for k >= r^2 with error O(r^2 + r min(k, 1/k)); at the cusp scale the O(r) correction vanishes). Consumes merged #207 (f6df5a73 at 566b1a1) and #218 (70ca57ef at fb6ee97), [R], [P], [Z], #191, #198; cites #229, #232, #216, #223, #214, #220. Exact stdlib controls Q1-Q6 (14 pinned fields in d = 1-4), mutants M1-M8, byte-identical on CPython 3.10-3.14 and -O. Two clean-context same-family referees with delta checks (A: ACCEPT; B: ACCEPT WITH MINOR FIXES, applied); zero organizational independence; nonauthor review required. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01NMeKEismAyeqgdB4sy2NJU --- .github/workflows/candidate-parity-rate.yml | 162 +++++ .../candidate_parity_rate_20261001/PROOF.md | 587 +++++++++++++++ .../candidate_parity_rate_20261001/README.md | 119 +++ .../RESULTS.json | 110 +++ .../SOURCES.json | 240 +++++++ .../parity_check.py | 676 ++++++++++++++++++ 6 files changed, 1894 insertions(+) create mode 100644 .github/workflows/candidate-parity-rate.yml create mode 100644 frontiers/candidate_parity_rate_20261001/PROOF.md create mode 100644 frontiers/candidate_parity_rate_20261001/README.md create mode 100644 frontiers/candidate_parity_rate_20261001/RESULTS.json create mode 100644 frontiers/candidate_parity_rate_20261001/SOURCES.json create mode 100644 frontiers/candidate_parity_rate_20261001/parity_check.py diff --git a/.github/workflows/candidate-parity-rate.yml b/.github/workflows/candidate-parity-rate.yml new file mode 100644 index 000000000..894bf68e0 --- /dev/null +++ b/.github/workflows/candidate-parity-rate.yml @@ -0,0 +1,162 @@ +name: Candidate parity rate 3/5 exact controls + +'on': + pull_request: + paths: + - 'frontiers/candidate_parity_rate_20261001/**' + - '.github/workflows/candidate-parity-rate.yml' + # bound merged sources (consumed and cited in SOURCES.json) and the pinned unmerged sources + - 'frontiers/three_fronts_20260924/LIFETIME_REMAINDER.md' + - 'imports/lifetime_parent_20260925/UNIFORM_MATRIX_CAP_AND_LIFETIME.md' + - 'imports/lifetime_parent_20260925/ERRATUM_CONGRUENCE.md' + - 'reviews/d1_section9_borel_repair_20260925/REPAIR.md' + - 'reviews/d1_chain_reconciliation_20260928/RECONCILIATION.md' + - 'frontiers/c7_zero_gap_limit_20260929/PROOF.md' + - 'frontiers/remainder_vanishing_20260930/PROOF.md' + - 'frontiers/remainder_rate_20260930/PROOF.md' + - 'frontiers/cusp_second_order_20261001/PROOF.md' + - 'frontiers/c2_finite_jet_transfer_20261001/PROOF.md' + - 'frontiers/c2_exact_20261001/NOTE.md' + - 'frontiers/candidate_third_order_20261001/PROOF.md' + - 'frontiers/third_order_rate_20261001/PROOF.md' + - 'frontiers/third_order_coefficient_20261001/NOTE.md' + - 'frontiers/d1_third_order_law_20261001/PROOF.md' + - 'frontiers/elder_third_order_20261001/PROOF.md' + workflow_dispatch: + +permissions: + contents: read + pull-requests: read + +jobs: + exact-controls: + 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' + GH_TOKEN: ${{ github.token }} + GH_REPOSITORY: ${{ github.repository }} + run: | + set -euo pipefail + python -B -S - <<'PY' + import base64, hashlib, json, os, pathlib, subprocess, sys, urllib.parse, urllib.request + root = pathlib.Path('frontiers/candidate_parity_rate_20261001').resolve() + manifest = json.loads((root / 'SOURCES.json').read_text()) + present = sorted(p.name for p in root.iterdir()) + expected_names = sorted([e['path'] for e in manifest['files']] + ['SOURCES.json']) + if present != expected_names: + raise ValueError('packet tree differs from manifest: ' + repr(present)) + for entry in manifest['files']: + path = root / entry['path'] + if path.is_symlink(): + raise ValueError('symlink: ' + entry['path']) + data = path.read_bytes() + if len(data) != entry['bytes'] or hashlib.sha256(data).hexdigest() != entry['sha256']: + raise ValueError('packet file identity mismatch: ' + entry['path']) + def same(data, entry): + framed = b'blob ' + str(len(data)).encode() + b'\0' + data + return len(data) == entry['bytes'] and hashlib.sha256(data).hexdigest() == entry['sha256'] and hashlib.sha1(framed).hexdigest() == entry['git_blob'] + def tree_entry(path): + # the checked-out commit's tree entry for path: (mode, type, blob) or None; symlinks have mode 120000 + out = subprocess.run(['git', 'ls-tree', '-z', 'HEAD', '--', path], check=True, capture_output=True).stdout + if not out: + return None + meta, name = out.rstrip(b'\0').split(b'\t', 1) + if name.decode() != path: + raise ValueError('unexpected tree entry for ' + path) + mode, kind, sha = meta.decode().split() + return mode, kind, sha + def tree_bytes(entry): + # a source read from the checkout must be a regular file whose tree blob is the recorded one + te = tree_entry(entry['path']) + if te is None: + return None + mode, kind, sha = te + if mode != '100644' or kind != 'blob' or sha != entry['git_blob']: + raise ValueError('tree entry is not the recorded regular-file blob (mode/type/blob ' + ' '.join(te) + '): ' + entry['path']) + path = pathlib.Path(entry['path']) + if path.is_symlink() or not path.is_file(): + raise ValueError('checkout path is not a regular file: ' + entry['path']) + return path.read_bytes() + for entry in manifest['consumed'] + manifest['cited_only']: + data = tree_bytes(entry) + if data is None or not same(data, entry): + raise ValueError('merged source drifted or absent: ' + entry['path']) + # unmerged sources are bound to their recorded commit, path and blob through the repository API: the head must + # resolve, the path in that head's tree must be a file with the recorded blob id and size, and the bytes must have + # the recorded sha256; a consumed dependency that has landed is instead read from the tree and must be identical + # to the pinned version (else this packet needs rebinding) + api = 'https://api.github.com/repos/' + os.environ['GH_REPOSITORY'] + def get_json(url): + req = urllib.request.Request(url, headers={'Authorization': 'Bearer ' + os.environ['GH_TOKEN'], + 'Accept': 'application/vnd.github+json'}) + with urllib.request.urlopen(req, timeout=60) as resp: + return json.load(resp) + def bound(entry): + commit = get_json(api + '/commits/' + entry['head']) + if commit.get('sha') != entry['head']: + raise ValueError('recorded head does not resolve: ' + entry['head']) + node = get_json(api + '/contents/' + urllib.parse.quote(entry['path']) + '?ref=' + entry['head']) + if (node.get('type') != 'file' or node.get('path') != entry['path'] or node.get('sha') != entry['git_blob'] + or node.get('size') != entry['bytes']): + raise ValueError('recorded head/path does not map to the recorded blob: ' + entry['path']) + blob = get_json(api + '/git/blobs/' + entry['git_blob']) + if blob.get('encoding') != 'base64': + raise ValueError('unexpected blob encoding for ' + entry['path']) + return base64.b64decode(blob['content']) + def current(entry): + # drift gate (as on #220, after Copilot's review there): the recorded head keeps resolving after a dependency PR advances, so + # also resolve the dependency PR now. Open: the file at its current head must be the pinned blob. Merged: + # the file at its merge commit must be the pinned blob. Closed unmerged: rebinding is required. + pr = get_json(api + '/pulls/' + entry['pr'].split('#', 1)[1]) + if pr.get('merged'): + ref = pr.get('merge_commit_sha') + elif pr.get('state') == 'open': + ref = pr['head']['sha'] + else: + raise ValueError('dependency ' + entry['pr'] + ' closed without merge: rebind required') + node = get_json(api + '/contents/' + urllib.parse.quote(entry['path']) + '?ref=' + ref) + if node.get('type') != 'file' or node.get('sha') != entry['git_blob'] or node.get('size') != entry['bytes']: + raise ValueError('dependency ' + entry['pr'] + ' advanced: ' + entry['path'] + ' at ' + ref + + ' is not the pinned blob (rebind required)') + return ref + for kind in ('consumed_unmerged', 'cited_unmerged'): + for entry in manifest.get(kind, []): + # a landed consumed dependency must be the recorded regular-file blob in this tree (symlinks and any + # other mode are rejected by tree_bytes); otherwise the source is bound through the API + landed = tree_bytes(entry) if kind == 'consumed_unmerged' else None + if landed is not None: + data, where = landed, 'tree' + else: + if os.path.lexists(entry['path']) and kind == 'consumed_unmerged': + raise ValueError('path present in the checkout but not as a tracked blob: ' + entry['path']) + data, where = bound(entry), 'head/path/blob api' + if not same(data, entry): + raise ValueError(kind + ' source differs from its pinned blob (rebind required): ' + entry['path'] + ' via ' + where) + now = current(entry) if kind == 'consumed_unmerged' else None + print(json.dumps({kind + '_bound': entry['path'], 'git_blob': entry['git_blob'], 'head': entry['head'], + 'via': where, 'current_ref_checked': now})) + script = str(root / 'parity_check.py') + expected = (root / 'RESULTS.json').read_bytes() + for flags in (['-B', '-S'], ['-B', '-O', '-S']): + out = subprocess.run([sys.executable, *flags, script], check=True, capture_output=True, timeout=600).stdout + if out != expected: + raise RuntimeError('output differs from RESULTS.json under ' + ' '.join(flags)) + for m in manifest['mutants']: + rc = subprocess.run([sys.executable, *flags, script, '--mutant', m], capture_output=True, timeout=600).returncode + if rc != 1: + raise RuntimeError('mutant ' + m + ' not rejected') + rc = subprocess.run([sys.executable, *flags, script, '--mutant', 'M99'], capture_output=True, timeout=600).returncode + if rc != 2: + raise RuntimeError('unknown mutant label not rejected with exit 2') + print(json.dumps({'scientific_effect': 'NONE', 'mathematical_acceptance': False})) + PY + git diff --exit-code diff --git a/frontiers/candidate_parity_rate_20261001/PROOF.md b/frontiers/candidate_parity_rate_20261001/PROOF.md new file mode 100644 index 000000000..97298f32f --- /dev/null +++ b/frontiers/candidate_parity_rate_20261001/PROOF.md @@ -0,0 +1,587 @@ +# The candidate density with remainder `ℓ^{3/5}`: a reflection parity removes the `ℓ^{1/2}` term + +Object: CL-CANDIDATE-PARITY-RATE-20261001-v1. +Author: Anthropic Claude (claude.ai session `session_01NMeKEismAyeqgdB4sy2NJU`), 1 October 2026. +Disposition: AUTHOR-SIDE PROOF CANDIDATE; NONAUTHOR REVIEW REQUIRED. Scientific effect: NONE — no register, graph, +STATUS, PROOF_INDEX, prize or Boolean change; no numerical constant is certified. Same GitHub account as every lane; +zero organizational independence. + +**What is new.** Math- #218 proves the candidate lifetime density to third order with remainder `O(ℓ^{4/11})`; Math- #229 +improves this to `O(ℓ^{3/7})`. #229 Remark 1 explains why a two-region (fold/cusp) argument stops at `3/7`: its error +terms come in two matching pairs whose leading coefficients are not identified, namely the boundary layer of the fold +expansion against the `κ^{−1}` tail of the cusp kernel, and the truncation `A₂(k) − A₂(0) = O(k)` against the linear +cusp error `rκ = k`. This note uses a single region and proves `O(ℓ^{3/5})`. In particular **the candidate density has +no term of order `ℓ^{1/2}`**, although #218 §0 and #229 §0 record `ℓ^{1/2}` as the formal next order. Three ingredients: +- **Parity and reflection** (§§1–2). After the birth height is integrated out, the field splits into independent even + and odd parts, the pin rows are even in the separation `r`, and point reflection `z ↦ −z` maps the problem at gap `k` + to the one at `−k`. Parts of this are on file: #191 Lemma E Step 3 records that `det H_M det H_S/r²` is even in `r`, + and #232 §§1, 3 uses the same birth-integrated basis and proves that the fold coefficient is even in `k` (Theorem J). + New here is the use of the reflection for the conditional laws inside the cusp layer (Lemma Λ; Lemma U, Step U2). +- **A uniform two-scale kernel** (§4, Lemma U). For `k ≥ r²` the typed kernel equals the fixed-cone kernel plus a + *layer* term, and the layer is, up to `O(r² + r min(k, 1/k))`, the cusp kernel's own tail `𝓐(κ) − 𝐀₂(0)`. So one + formula, `r^{−2}𝐀_0 + [𝐀₂(k) − 𝐀₂(0)] + 𝓐(κ)`, approximates the kernel from the fold scale through the cusp scale, + and no fold/cusp split point is needed. At the cusp scale it shows that the `O(r)` correction of the candidate cusp + kernel vanishes. +- **A layer comparison lemma** (§3, Lemma Λ). Moving the gap parameter from `k` to `0` inside the layer costs + `O(k²/(1 + κ))`, not `O(k²)`: the first order vanishes by reflection, and the kink of the weight is controlled through + the density of its location (with Lemma D″, a bound for two small eigenvalues). + +**Dependencies (consumed; both merged on 1 October 2026 with the reviewed blobs).** +- Math- #207 (`frontiers/cusp_second_order_20261001/PROOF.md`, blob `f6df5a73`; merged at `566b1a1`): §0 + (`Y`, `Δ`, `E₀`); §4 (4.1) (`𝒜^{cand}`, `𝒜^{con}`); §7: Lemma L (7.1)–(7.2), the decomposition of `B_{d,L}`, `I^{cand}` + ((CU′.1), (7.3)). +- Math- #218 (`frontiers/candidate_third_order_20261001/PROOF.md`, blob `70ca57ef`; merged at `fb6ee97`): §0 (`A₂`, + the coefficient `c₂` (0.1), the cusp kernels restated); §1 (Lemma D (1.1) and its proof; `T` and (1.2)); Step F1 + ((2.1)–(2.2) and the free-jet nondegeneracy); Step F3; (F.2) and the surrogate remark after Lemma F; Lemma C's Step C1 + (its three cases); the proof of Lemma O; §4 (the decomposition and the bounds on `J₂`–`J₅`). + +**Merged inputs.** [R] (§2 rows and target; (R2)–(R5)); [P] (§2 finite-jet rank) with [E1], [E2], [REC]; [Z] +((Z2)–(Z3): `B_{d,L}`); Math- #191 (`441152df`: `r_0^*`, (2.1) and Lemma E Step 4 directly, Step 3 for attribution; +Lemma E Steps 1–3 through #218); Math- #198 (`abfb98ae`: (W.4) directly; (3.1), the sign window of §3 and Lemma W through +#218 and #207). +**Cited only:** #229 (Theorem T⁺, superseded here for `ν_cand`; Remarks 1–2), #232 (merged at `7fe06b0`: an independent +derivation of the same second coefficient (2.3), and the birth-integrated basis), #216 (values of `c₂`; the formal law (0.2); its +Monte Carlo), #223 (`c` and `c₂` in closed form, with certified enclosures), #214 (`d = 1`), #220 (the elder density; not +used). + +## 0. Statement + +Setting and notation are those of #218 §0 and #207 §0: fixed `d ≥ 2`, `m = d − 1`, `L > 0`, the [P] field `f` on the torus +`R^d/(LZ^d)`; a unit vector `u` with an orthonormal frame `Θ` of `u^⊥` (coordinate `x` along `u`, `y` along `Θ`); pins +`M = −ru/2` (height `b`, a maximum) and `S = ru/2` (height `b − kr³`, an index-`(d−1)` saddle); `κ := k/r`; the scaled +endpoint Hessians `K_M`, `K_S` with `det K_i = det H_i/r`; the candidate kernel `A_r(b, k, u) = 12π_r(v_r)E_Q[F_d(K_M) +F_{d−1}(K_S)]`, so that `ν_cand(ℓ) = ∫_0^{r_0}∫_R∫_{S^{d−1}} r^{−2}A_r(b, ℓ/r³, u) dσ db dr + ν_cand^{far,r_0}(ℓ)`; the +contact kernel `A_0`; the fold coefficient `A₂` and `c₂` of #218 (0.1); the cusp kernels `𝒜^{cand}`, `𝒜^{con}` of #207 (4.1) +(restated in #218 §0); `I^{cand}` of #207 (CU′.1); `c = c_{d,L}` and `B_{d,L}`. + +*Jets at `0`.* `A := ∇_Θ²f(0)`, `B := ∂_uA`, `C := ∂_u²A`, `γ := ∇_Θ∂_u²f(0)`, `η := ∇_Θ∂_u³f(0)`, `f₄ := ∂_u⁴f(0)`, +`f₅ := ∂_u⁵f(0)`; `Δ := det A` and `A^♯ := adj A`; `Y′ := (f₄/12)Δ − γᵀA^♯γ/4`, the cusp variable `Y` of #207 §0 built from +the jets of `f` at `0`. Jets of even order (`A`, `C`, `η`, `f₄`) are called *even*, jets of odd order (`B`, `γ`, `f₅`) +*odd*; the parity map `𝒫` negates the odd ones. + +*Birth-integrated kernels.* `𝐓_r(k, u) := ∫_R r^{−2}A_r(b, k, u) db`; `𝐀_0(k, u)` and `𝐀₂(k, u)`, defined for `k ∈ R` in +Lemma K and equal to `∫A_0 db` and `∫A₂ db` for `k ≥ 0`; `𝓐(κ, u) := ∫_R(𝒜^{cand} − 𝒜^{con})(b, κ, u) db`. + +**Theorem P (the candidate density with remainder `ℓ^{3/5}`).** For every `d ≥ 2` and `L > 0`, as `ℓ ↓ 0`, + + ν_cand(ℓ) = c ℓ^{−1/3} + B_{d,L} + I^{cand} ℓ^{1/4} + c₂ ℓ^{1/3} + O(ℓ^{3/5}). (P.1) + +**Corollary P′ (no `ℓ^{1/2}` term).** If `ν_cand(ℓ) = cℓ^{−1/3} + B_{d,L} + I^{cand}ℓ^{1/4} + c₂ℓ^{1/3} + eℓ^{1/2} + o(ℓ^{1/2})`, +then `e = 0`. (Indeed `ℓ^{3/5} = o(ℓ^{1/2})`.) + +**What is not claimed.** +- Nothing about the elder or rejected densities beyond #229. Lemma U is not proved for the elder kernel (Remark 3), and + the Monte Carlo of #216 suggests that the elder and rejected densities carry further terms that the adjacent-pair + density does not (Remark 4). +- No sharpness of `3/5`. The ledger's remaining terms are `ρ³ + ℓ²ρ^{−7} + ℓρ^{−2}` at the split `ρ = ℓ^{1/5}`, and + `ℓ^{2/3}` from the fold-scale part of Lemma U (Remark 2). +- No certified numerical value of `I^{cand}` or `B_{d,L}` (those of `c`, `c₂` are #223's); no uniformity in `d` or `L`; + no finite-radius band. +- Nothing beyond the existential scope of [R], [P], [Z], #191, #198, #207 and #218. + +## 1. Birth integration, parity and reflection + +**Rows.** [R] §2's rows at the pins `x_M := −r/2` and `x_S := r/2` on the axis `u` are `(f(x_M) + f(x_S))/2`, +`(f(x_S) − f(x_M))/r`, `(f_x(x_S) − f_x(x_M))/r`, `T_r := (6/r²)[f_x(x_M) + f_x(x_S) − 2(f(x_S) − f(x_M))/r]` and the +transverse rows `(f_{y_j}(x_M) + f_{y_j}(x_S))/2`, `(f_{y_j}(x_S) − f_{y_j}(x_M))/r`, with target +`(b − kr³/2, −kr², 0, 12k, 0, …, 0)` and Jacobian factor `12`. Replace the second row by +`(f_x(x_M) + f_x(x_S))/2 = (f(x_S) − f(x_M))/r + r²T_r/12`. This adds a multiple of `T_r` to it (determinant `1`), and +its target is `−kr² + r²·12k/12 = 0` (control Q2). The first row is the midpoint height `b′ = b − kr³/2`. What remains is + + V_r := ((∇f(x_M) + ∇f(x_S))/2, (∇f(x_S) − ∇f(x_M))/r, T_r) ∈ R^{2d+1}, target v(k) := (0_d, 0_d, 12k). (1.1) + +For `k ∈ R` write `p_{V_r}` for the density of `V_r` and `Q̄_{r,k}` for the law of the field given `V_r = v(k)` (Gaussian +regression). Put `f_e(z) := (f(z) + f(−z))/2` and `f_o(z) := (f(z) − f(−z))/2`. + +**Lemma R (birth integration, parity, reflection).** For `0 < r ≤ r_0^*`, `k ∈ R` and `u ∈ S^{d−1}`: +- (R.1) *Birth integration.* For `k ≥ 0`, `𝐓_r(k, u) = 12p_{V_r}(v(k))E_{Q̄_{r,k}}[r^{−2}Π 1_{typed}]` with + `Π := −det K_M det K_S`. The fixed-cone surrogate `𝐒_r(k, u) := 12p_{V_r}(v(k))E_{Q̄_{r,k}}[r^{−2}Π1{A<0}]`, defined for + `k ∈ R`, equals `∫_R 12π_r(v_r)E_Q[r^{−2}Π1{A<0}] db` for `k ≥ 0`. +- (R.2) *Parity.* Under `Q̄_{r,k}` the fields `f_e` and `f_o` are independent; `f_e` is centred and its law does not depend + on `k`; `f_o` has a covariance independent of `k` and a mean linear in `k`. So the even jets have a centred law that does + not depend on `k`, the odd jets have means `k·w` (`w` fixed) and a covariance independent of `k`, and the two families + are independent. Moreover `p_{V_r}(v(−k)) = p_{V_r}(v(k))`. +- (R.3) *Reflection.* For `f̃(z) := f(−z)`, `V_r(f̃) = 𝔭V_r(f)`, where `𝔭` negates the first and last blocks. So + `f ~ Q̄_{r,k}` implies `f̃ ~ Q̄_{r,−k}`. The jets of `f̃` at `0` are those of `f` with `𝒫` applied; `Π(f̃) = Π(f)` and + `A(f̃) = A(f)`; typed and anti-typed configurations are exchanged. +- (R.4) *Evenness in `r`.* The rows of `V_r`, extended to `0 < |r| ≤ r_0^*` by the same formulas, are even in `r`, and + `V_0 := lim_{r→0}V_r = (∇f(0), ∂_u∇f(0), ∂_u³f(0))`. For every finite vector `J₀` of free jets at `0`, the covariance of + `(V_r, J₀)` is a smooth even function of `r ∈ [−r_0^*, r_0^*]`, uniformly nondegenerate (also in `u`). Hence + `p_{V_r}(v(k))` and the `Q̄_{r,k}`-law of `J₀` (mean and covariance) are smooth even functions of `r`; at `r = 0` the law is + the conditional law given `V_0 = v(k)`. Pathwise, `Π` is even in `r`. + +*Proof.* (R.1) In [R]'s Kac–Rice representation, `π_r(v_r)E_Q[·]` is the density of [R]'s pin vector at its target times +the conditional expectation. The row change is unimodular, with the target computed above, so this equals +`p_{(b′,V_r)}(b′, v(k))E[· | b′, V_r = v(k)]`. Integrating over `b′ ∈ R` gives `p_{V_r}(v(k))E[· | V_r = v(k)]`: by Tonelli +for `𝐓_r`, whose integrand is nonnegative, and by Fubini for `𝐒_r`, whose integrand is absolutely integrable. +(R.2) `Cov(f(z) + f(−z), f(w) − f(−w)) = 0` because the covariance is even, so the Gaussian fields `f_e`, `f_o` are +independent. Since `x_M = −x_S`, the average gradient is `∇f_o(x_S)`, and `T_r` is a functional of `f_o` (it is built from +`f_x(x_M) + f_x(x_S) = 2∂_xf_o(x_S)` and `f(x_S) − f(x_M) = 2f_o(x_S)`), while the gradient difference is `2∇f_e(x_S)/r`. +So `V_r = v(k)` conditions `f_e` on a zero target and `f_o` on a target linear in `k`, and Gaussian regression gives means +linear in the targets and covariances that do not depend on them. Jets of even order at `0` are functionals of `f_e`, +those of odd order of `f_o`. Finally the law of `V_r` is invariant under `𝔭` by (R.3), and `𝔭v(k) = v(−k)`. +(R.3) `∇f̃(z) = −∇f(−z)` and `D²f̃(z) = D²f(−z)`. So the average gradient and `T_r` change sign, the gradient difference +does not, and the two endpoint Hessians are exchanged; `Π` is symmetric in them, and `A(f̃) = ∇_Θ²f(0)`. The +unconditioned field is invariant in law under `z ↦ −z`, and Gaussian regression is equivariant under the linear +isomorphism `𝔭`. Exchanging the endpoint Hessians exchanges the two index conditions, that is, typed and anti-typed pairs. +(R.4) Each row is unchanged under `r ↦ −r`: the endpoints are exchanged and every rule is centred. Taylor's formula gives +the limits ([R] §2: `T_r = f_xxx(0) + (r²/40)f₅ + O(r⁴)`). The covariances are centred difference quotients of +derivatives of the smooth covariance function, hence smooth and even in `r`. For each `r ∈ [0, r_0^*]` and frame they are +nondegenerate by [P] §2 (distinct derivative functionals at distinct sites, or distinct jets at one point, are linearly +independent), as in #218 Step F1, and uniformly so by compactness. `Π = −det H(x_M) det H(x_S)/r²` is symmetric under +the exchange `r ↦ −r`; #191 Lemma E Step 3 records this as `det H_M(−r) = det H_S(r)`. ∎ + +The birth-integrated basis (1.1), with its evenness in `r` and the independence of the odd and even families, is also the +one Math- #232 §1 uses; nothing here depends on #232. The typed indicator is not even in `r` or `k`: both `r ↦ −r` and +the reflection exchange it with the anti-typed one. + +## 2. The product and its second coefficient + +With the jets of §0 put `Δ_B := D det(A)[B] = tr(A^♯B)`, `Δ_C := tr(A^♯C)`, `Δ_BB := D²det(A)[B, B]` and +`A^♯_B := D adj(A)[B]` (polynomials, defined for singular `A`), and + + U := Y′ + 3kΔ_B, + V := (3k/4)(Δ_C + Δ_BB) + (f₄/24)Δ_B + (f₅/120)Δ − (1/12)γᵀA^♯η − (1/8)γᵀA^♯_Bγ. (2.1) + +Write `V = V_o + kV_e` with `V_e := (3/4)(Δ_C + Δ_BB)`; `V_e` is even and `V_o` is odd under `𝒫`. Let `T` be #218 §1's +dominating variable. Under `Q̄_{r,k}`, `E[T^p] ≤ C_p(1 + k)^p`: mix #218's bound `C_p(1 + |b′| + k)^p` over the conditional +law of the midpoint height `b′ = f_e(x_S)`, which is centred Gaussian and does not depend on `k` (Lemma R (R.2)). Likewise, +for a finite jet vector `J₀` with uniformly invertible covariance, (1.2) gives `E[T^p | J₀] ≤ C_p(1 + k + |J₀|)^p`. + +**Lemma Π.** Pathwise, + + det K_M = −6kΔ + rU − r²V + O(r³(1 + k)T^N), det K_S = 6kΔ + rU + r²V + O(r³(1 + k)T^N), + Π = 36k²Δ² + r²(12kΔV − U²) + O(r⁴(1 + k)^N T^N). (2.2) + +Under `(k, jets) ↦ (−k, 𝒫jets)`, `Δ`, `Y′` and `U` are invariant and `V` changes sign; `det K_M` and `det K_S` are affine in +`k`. + +*Proof.* By #218 Step F1, `Π` equals a polynomial `Φ(J′; r, k)` in the free jets `J′` up to `O(r⁴T^{n₀})`, and the +coefficients of `Φ` are identities of polynomials, which may be read off on exactly pinned polynomial fields. On those +fields, with the height row pinned at the midpoint `b′`, every pinned jet is an even polynomial in `r` (Lemma R (R.4); +with [R]'s row at fixed `b`, `f(0)` would carry a term `−kr³/2`, which does not enter the Hessians). So `Π` is even in `r`, +and `Φ` has no odd powers of `r`. The pins give `f_xx(0) = −r²f₄/24 + O(r⁴)`, `f_xxx(0) = 12k − r²f₅/40 + O(r⁴)` and +`∇_yf_x(0) = −r²η/24 + O(r⁴)`. Expanding `α_i = f_xx(x_i)/r`, `β_i = ∇_yf_x(x_i)/r` and `A_i = ∇_Θ²f(x_i)` and using +`det K_i = α_i det A_i − rβ_iᵀadj(A_i)β_i` gives the first two lines; their product gives (2.2), the `r`-odd terms +cancelling. Only `α_i` depends on `k`, affinely, so `det K_i` is affine in `k`. The parity statement is read off (2.1): +`B`, `γ`, `f₅` are odd and `A`, `C`, `η`, `f₄` even, so `Δ_B` and `A^♯_B` are odd and `Δ_BB` (quadratic in `B`) is even. +Control Q3 checks (2.2) exactly on fourteen pinned polynomial fields in `d = 1, 2, 3, 4`, with general free jets; control +Q2 checks the evenness, the reflection and the affinity in `k`. ∎ + +(2.2) is #232's (2.3) (`W₂ = 12kΔV − U²`, with remainder `O(r³)` there), obtained independently here; the absence of the +`r³` term is the pathwise evenness of Lemma R (R.4). + +**Lemma K (the surrogate; `𝐀₂` is even in `k`).** There are `C, c, N` such that for `0 < r ≤ r_0^*`, `k ∈ R` and `u`, + + r²𝐒_r(k, u) = 𝐀_0(k, u) + r²𝐀₂(k, u) + ϱ_S, |ϱ_S| ≤ C r⁴ (1 + |k|)^N e^{−ck²}, (K.1) + +where `𝐀_0` and `𝐀₂` are smooth even functions of `k ∈ R`, equal for `k ≥ 0` to `∫A_0(b, k, u) db` and `∫A₂(b, k, u) db`. +In particular `|𝐀₂(k, u) − 𝐀₂(0, u)| ≤ Ck²` for all `k ≥ 0`, uniformly in `u`. + +*Proof.* Condition on `A`. Given `(V_r, A) = (v(k), A)`, the entries of the Hessians at `x_M` and `x_S` are jointly Gaussian, +with means affine in `(k, A)` and a covariance independent of `(k, A)` and smooth in `r`: `(V_r, A)` is uniformly +nondegenerate (Lemma R (R.4) with `J₀ ⊃ A`), and the covariances of the endpoint Hessian entries with `(V_r, A)` and +with each other are smooth in `r` by the argument of (R.4). So +`Ψ(r; k, A) := E[−det H_M det H_S | A, V_r = v(k)]` is a polynomial in `(k, A)` whose coefficients are smooth functions of +`r`, even because `r ↦ −r` exchanges the endpoints and preserves the product. Pathwise `−det H_M det H_S = r²Π`, so `Ψ` +vanishes at `r = 0`, and `Ψ = r²Ψ̃` with `Ψ̃(r; k, A) = E[Π | A, V_r = v(k)]` smooth and even in `r`. Hence + + r²𝐒_r(k, u) = 12 ∫_{A<0} p_{(V_r, A)}(v(k), A) Ψ̃(r; k, A) dA + +is smooth and even in `r`, with fourth `r`-derivative bounded by `C(1 + |k|)^Ne^{−ck²}` uniformly in `u` (the frames form a +compact set); this gives (K.1). At `r = 0` it is `12p_{V_0}(v(k))·36k²E_{Q̄_{0,k}}[Δ²1{A<0}]`, which is `∫A_0 db` for `k ≥ 0` +(disintegrate in `b′` as in (R.1)). By #218's surrogate remark after Lemma F (through Step F3), at fixed `b` the surrogate +is `A_0 + r²A₂ + O(r³(1 + k)²P^Ne^{−c(b²+k²)})`. Integrating in `b` and comparing with (K.1) identifies the `r²`-coefficient +as `∫A₂ db`; the `r³` terms at fixed `b` (the birth shift `b ↦ b′` and #218's `r³Φ₃`) disappear after the `b`-integral, which +is even in `r`. Evenness in `k`: by Lemma R (R.2)–(R.3), `r²𝐒_r(−k, u) = r²𝐒_r(k, u)`. Finally `∂_k𝐀₂(0, u) = 0`, +`sup_{|k|≤1, u}|∂_k²𝐀₂| < ∞`, and `|𝐀₂(k)| + |𝐀₂(0)| ≤ C` for `k ≥ 1`. ∎ + +(For the Gaussian kernel, control Q4 computes exactly, in `d = 1, 2, 3`, the `A`-integrand `H(k, A)` of the birth-integrated +`r²` coefficient (#232 (3.2)): `H = h₀ + h₂k² + h₄k⁴`, with no odd powers, and the density exponent is `a = 12`. This is +#232's Theorem J; the `d = 1, 2` polynomials agree with its §4.) + +## 3. Layer tools + +Throughout, `X := Y² − 36κ²Δ²` for a cusp variable `Y ∈ {Y′, U}`, and the *layer* is `{A < 0, X > 0} = {A < 0, |Y| > 6κ|Δ|}`. + +**Lemma D′ (bounded density of the determinant).** Under the hypothesis of #218 Lemma D, for every interval `I ⊂ R`, +`E[(1 + ‖G‖)^n 1{det G ∈ I}] ≤ C(1 + |μ|)^{n′}|I|`. + +*Proof.* #218's proof of Lemma D's second bound, with `(−ε, ε)` replaced by `I`: on `{det G ∈ I}`, an eigenvalue `λ_{i*}` +of least modulus lies in an interval of length `|I|/D`, `D := ∏_{j≠i*}|λ_j|`, and the Vandermonde factor satisfies +`|𝒱(λ)| ≤ 2^{m−1}D|𝒱′|`; `D` cancels. ∎ + +**Lemma D″ (two small eigenvalues).** Let `m ≥ 2`, let `G` be as in #218 Lemma D, and let `D_G` be the product of the +`m − 1` largest eigenvalue moduli of `G`. Then for `0 < ε ≤ 1`, `E[(1 + ‖G‖)^n 1{D_G < ε}] ≤ C(1 + |μ|)^{n′}ε²`. + +*Proof.* Use Weyl coordinates, as in #218's proof of Lemma D, with the eigenvalues ordered by modulus, +`x₁ = |λ_(1)| ≤ x₂ = |λ_(2)| ≤ … ≤ x_m`, so that `D_G = x₂⋯x_m`. Since `|λ_(1) − λ_(2)| ≤ 2x₂` and `|λ_(i) − λ_(j)| ≤ 2x_j` +for `i ≤ 2 < j`, the Vandermonde factor is `|𝒱(λ)| ≤ 2^{2m−3}x₂∏_{j≥3}x_j²|𝒱″|`, where `𝒱″` is the Vandermonde factor of +`λ_(3), …, λ_(m)`. Integrating `λ_(1)` over `[−x₂, x₂]` gives at most `2^{2m−2}D_G²|𝒱″| ≤ 2^{2m−2}ε²|𝒱″|` on `{D_G < ε}`. +Then `λ_(2)` ranges over an interval of length `≤ 2x₃` (of length `≤ 2ε` if `m = 2`), and the remaining integral, with +the Gaussian weight, is `≤ C(1 + |μ|)^{n′}`. ∎ + +**Lemma Λ₀ (layer probability).** Consider a Gaussian law of the jets under which `A` is centred, with covariance in a +fixed compact set of positive definite matrices, and the other jets have, given `A`, means affine in `(A, k)` with bounded +coefficients and bounded covariances. Examples are `Q̄_{r,k}` for `0 ≤ r ≤ r_0^*` and `k ≥ 0` (Lemma R (R.2), (R.4)) and the +linear interpolation of the means and covariances of two such laws. For nonnegative functions `p` and `ω` of the jets +that are bounded by polynomials, and `κ > 0`, + + E[p 1{A < 0, |Y| ≥ 6κ|Δ| − ω}] ≤ C(1 + k)^N / (1 + κ), + +with `C` and `N` depending on `p` and `ω` but not on `κ`, `k`, `r` or `u`. + +*Proof.* For `κ ≤ 1` bound the indicator by `1`. For `κ ≥ 1` the event lies in `{|Δ| ≤ (|Y| + ω)/(6κ)}`. Given `A`, the +other jets are Gaussian with means affine in `(A, k)`, so `P(|Y| + ω ≥ t | A) ≤ C_pt^{−p}(1 + k + ‖A‖)^{n_p}`. Dyadic +layers of `|Y| + ω` and Lemma D give the bound, as in #218's proof of Lemma O. ∎ + +**Lemma Λ (the gap inside the layer).** Let `κ > 0` be a parameter and, for `s ∈ R`, +`Λ(s) := 12p_{V_0}(v(s))E_{Q̄_{0,s}}[(U_s² − 36κ²Δ²)₊1{A<0}]`, where `U_s := Y′ + 3sΔ_B`. Then `Λ` is even, and for +`0 ≤ k ≤ 1`, with `C` independent of `κ`, `k` and `u`, + + |Λ(k) − Λ(0)| ≤ C k² / (1 + κ). (Λ.1) + +*Proof.* *Evenness:* by Lemma R (R.2)–(R.3), `U_{−s}(𝒫·) = U_s(·)`, `Δ(𝒫·) = Δ(·)` and `p_{V_0}(v(−s)) = p_{V_0}(v(s))`. + +*Reduction.* `p_{V_0}(v(s)) = p_{V_0}(v(0))e^{−as²}`, with `a > 0` bounded above and below. Hence the density factor +contributes `|e^{−ak²} − 1|·12p_{V_0}(v(0))E_{Q̄_{0,k}}[(U_k² − 36κ²Δ²)₊1{A<0}]`, which is `≤ Ck²/(1 + κ)` by Lemma Λ₀. +By (R.2), under `Q̄_{0,s}` the free jets can be realized as `ζ₀ + s·w`, with `ζ₀ ~ Q̄_{0,0}` and `w` supported on the odd jets. +Then `U_s(ζ₀ + sw) = Y₀ + sY₁ + s²Y₂`, with `Y₀ := Y′(ζ₀)`, `Y₁ := 3tr(A^♯B₀) − w_γᵀA^♯γ₀/2` and +`Y₂ := 3tr(A^♯w_B) − w_γᵀA^♯w_γ/4`. Here `γ₀`, `B₀` are the odd jets of `ζ₀`, and `w_γ`, `w_B` are the components of `w`. +`Y₁` is odd, `Y₂` is even, and neither involves `f₄`. Since `Λ(s)e^{as²}` is even in `s`, it remains to bound + + E[½(φ(Y₀ + kY₁ + k²Y₂) + φ(Y₀ − kY₁ + k²Y₂)) − φ(Y₀)], φ(y) := (y² − θ²)₊, θ := 6κ|Δ|, on {A < 0}. + +*Integrate `f₄` first.* Given the other jets, `f₄` is Gaussian with mean `m₄`, linear in the other even jets, and a +variance `σ₄²` bounded above and below (a Schur complement; #218 Step F1). Write `Y₀ = (Δ/12)f₄ + β` with +`β := −γ₀ᵀA^♯γ₀/4`. Then `Φ(t) := E_{f₄}[φ(Y₀ + t)]` is `E[φ(Z + t)]` for `Z ~ N(μ_Z, σ_Z²)`, where `μ_Z := (Δ/12)m₄ + β` +and `σ_Z := |Δ|σ₄/12`. `φ` is convex, `φ″ = 2` off `{|y| = θ}`, and `φ′` jumps by `2θ` at `±θ`. Hence `Φ` is convex and +smooth, with + + Φ″(t) = 2P(|Z + t| > θ) + 2θ[p_Z(θ − t) + p_Z(−θ − t)], θ p_Z(x) = (72κ/(√(2π)σ₄)) e^{−(x − μ_Z)²/(2σ_Z²)}. (3.1) + +(`Δ` cancels in the second identity.) In particular `Φ″ ≤ 2 + 288κ/(√(2π)σ₄) ≤ C(1 + κ)` (two bumps). Write the quantity to be bounded +as `E[Φ(t₀) − Φ(0)] + E[½(Φ(t₀ + h) + Φ(t₀ − h)) − Φ(t₀)]`, with `t₀ := k²Y₂` and `h := kY₁`. + +*The case `κ ≤ 1`.* `|Φ′(t)| ≤ 2E_{f₄}|Z + t|`, and the second difference is at most `h²·sup Φ″ ≤ Ch²`. So the quantity is +`O(k²)`. + +*The case `κ ≥ 1`.* +- *First part.* `|Φ(t₀) − Φ(0)| ≤ |t₀|·sup_{|t|≤|t₀|}|Φ′(t)|` and `|Φ′(t)| ≤ 2E_{f₄}[(|Z| + |t₀|)1{|Z| ≥ θ − |t₀|}]`. By + Lemma Λ₀ (with `ω = |t₀|`) this part is `≤ Ck²κ^{−1}`. +- *Second part, bulk.* The symmetric second difference of a convex function is `½∫_{−|h|}^{|h|}(|h| − |τ|)Φ″(t₀ + τ)dτ`. + The first term of (3.1) contributes `≤ h²P_{f₄}(|Z| ≥ θ − |t₀| − |h|)`, which is `≤ Ck²κ^{−1}` by Lemma Λ₀ (with + `ω = |t₀| + |h|`). +- *Second part, kink.* By (3.1) the second term contributes + `K := (72κ/(√(2π)σ₄))∫(|h| − |τ|)₊ Σ_± e^{−g_±²/(2σ_Z²)} dτ`, where `g_± := ±θ − μ_Z − t₀ − τ`. Note that `K ≤ Cκh²`. + - *Coordinates.* Use Weyl coordinates for `A`, as in Lemma D′: `λ` is an eigenvalue of least modulus and `D` the product + of the other moduli (`D := 1` if `m = 1`), so `|Δ| = |λ|D`, and `λ < 0` on `{A < 0}`. Condition on the eigenvectors, + the other eigenvalues and the free jets `𝐠` other than `A` and `f₄`. Then `λ` ranges over `[−x₂, 0)`, where `x₂` is the + second smallest modulus (over `(−∞, 0)` if `m = 1`), with density `ρ(λ) ≤ C2^{m−1}D|𝒱′|p̄`. Here `𝒱′` is the Vandermonde + factor of the other eigenvalues and `p̄` the supremum over `λ` of the Gaussian density of `A` given `𝐠`. + - *Dependence on `λ`.* `θ = −6κDλ` is linear in `λ`; `A^♯`, `β`, `m₄`, `Y₁` and `Y₂` are affine; `μ_Z` is quadratic. Let + `P := C₀(1 + ‖A‖ + |𝐠|)^{N₀}`, with `C₀`, `N₀` chosen so that `|μ_Z′| + |t₀′| ≤ P/2` and `|m₄| ≤ P` on the range of `λ`. + For `m ≥ 2`, `‖A‖ = x_m` does not depend on `λ`; for `m = 1` the conditions below cut out an interval of `λ`. + - *Main set* `𝔐 := {P ≤ κD and P ≤ κ}`. Here `|g_±′| ≥ 6κD − P/2 ≥ 3κD`, so each `g_±` is monotone in `λ`, and + `|Δm₄|/12 ≤ θ/72`. Where `|g| ≥ θ/2`, `e^{−g²/(2σ_Z²)} ≤ e^{−648κ²/σ₄²}`, and the contribution is + `≤ Cκh²e^{−648κ²/σ₄²} ≤ Ch²κ^{−1}`. Where `|g| < θ/2`, `|β + t₀ + τ| > θ/2 − θ/72 > θ/4`. On the support + `|τ| < |h(λ)|` of the weight, `|β + t₀ + τ| ≤ M₀ := sup_λ(|β| + |t₀| + |h|)`, a polynomial in `(‖A‖, 𝐠)` that does + not depend on `λ`; so `|λ| < M₀/(1.5κD)`. Hence + `σ_Z < σ̄ := M₀σ₄/(18κ)` there. Bound `(|h| − |τ|)₊ ≤ (H − |τ|)₊` with `H := sup_λ|h|`, and change variables `λ ↦ g` + (`|dλ/dg| ≤ 1/(3κD)`): `∫e^{−g²/(2σ̄²)}ρ dλ ≤ C D|𝒱′|p̄·√(2π)σ̄/(3κD) = C|𝒱′|p̄M₀σ₄/κ²`. With the prefactor and + `∫(H − |τ|)₊dτ = H²` this gives `≤ C|𝒱′|p̄M₀H²/κ`, and integrating over the conditioning variables gives + `E[K1_𝔐] ≤ Ck²(1 + k)^Nκ^{−1}`. No largeness of `κ` is used beyond `κ ≥ 1`. + - *Exceptional set* `𝔐^c ⊂ {D < P/κ} ∪ {P > κ}`. Use `K ≤ Cκh²`. For `m ≥ 2`, Lemma D″, applied given `𝐠` in dyadic + layers of `‖A‖`, gives `E[h²1{D < P/κ}] ≤ Ck²κ^{−2}`. For `m = 1`, `D = 1` and this set is `{P > κ}`. By Markov, + `E[h²1{P > κ}] ≤ C_pk²κ^{−p}`. So `E[K1_{𝔐^c}] ≤ Ck²κ^{−1}`. + +Summing gives (Λ.1). ∎ + +## 4. A uniform two-scale kernel + +**Lemma U.** There are `C, r₃ > 0` such that for `0 < r ≤ r₃`, `k ≥ r²` (that is, `κ ≥ r`) and `u ∈ S^{d−1}`, + + |𝐓_r(k, u) − r^{−2}𝐀_0(k, u) − [𝐀₂(k, u) − 𝐀₂(0, u)] − 𝓐(k/r, u)| ≤ C(r² + r min(k, 1/k)). (U.1) + +*Proof.* Terms evaluated at the target `v(k)` carry the factor `p_{V_r}(v(k)) ≤ Ce^{−ck²}`, which absorbs polynomial growth +in `k`. The zero-gap terms `Λ(0)`, `𝐀₂(0)` and `𝓐` do not carry it; they are compared directly. + +*Step U1 (the surrogate).* By Lemma K, `𝐒_r = r^{−2}𝐀_0 + 𝐀₂ + O(r²(1 + k)^Ne^{−ck²})`. + +*Step U2 (the typed kernel is the surrogate plus the layer).* Work under `Q̄ := Q̄_{r,k}`, with `T` as in §2, +`m_i := det K_i/r` and `s := (−1)^m`. Put `X := U² − 36κ²Δ²` and `ε := −r^{−2}Π − X`. By Lemma Π, `ε = −12kΔV + ε₂` with +`|ε₂| ≤ Cr²(1 + k)^NT^N`. By (R.1), + + 𝐓_r − 𝐒_r = 12p_{V_r}(v(k)) (E_Q̄[X₊1{A<0}] + E_Q̄[G]), G := r^{−2}Π(1_{typed} − 1{A<0}) − X₊1{A<0}. + +Split according to #218 Step C1's cases, with `A = ∇_Θ²f(0)` in place of #218's `A_0`. Step C1 uses `‖A_i − A‖ ≤ rT` for +the endpoint matrices `A_i := ∇_Θ²f(x_i)`, which holds since `‖A_i − A‖ ≤ (r/2)‖f‖_{C³} ≤ rT`, and the expansions of +`det K_i`, which Lemma Π supplies with `A` in place of `A_0`. +- *Case 2* (`λ_max(A) > rT`): the pair is not typed and `A ≮ 0`, so `G = 0`. +- *Case 3* (`|λ_max(A)| ≤ rT`). Here `|Δ| ≤ rT^m`, and #218's Case 3 gives `|r^{−2}Π|1_{typed} = |m_Mm_S|1_{typed} ≤ + C(k + r(1 + k))²T^N`. The surrogate weight itself is not small here (`−r^{−2}Π ≈ U²`, and `U` is not small when `Δ` is), + so compare with the layer integrand. On `{A < 0}`, `r^{−2}Π(1_{typed} − 1) − X₊ = r^{−2}Π1_{typed} − (−X)₊ + ε` + (control Q5), with `(−X)₊ ≤ 36κ²Δ² ≤ 36k²T^{2m}` and `|ε| ≤ C(k|Δ| + r²)(1 + k)^NT^N ≤ C(kr + r²)(1 + k)^NT^N`. On + `{0 < λ_max(A) ≤ rT}` only `r^{−2}Π1_{typed}` is present. So `|G| ≤ C(k + r)²(1 + k)^NT^N` on Case 3. Its probability, + with the weight `T^N`, is `≤ Cr(1 + k)^N` (Lemma D with dyadic layers in `T`, and (1.2) with `J₀ = A`). So Case 3 + contributes `O(r(k + r)²(1 + k)^N)`. +- *Case 1* (`λ_max(A) < −rT`). Here `A, A_M, A_S < 0`, and by Haynsworth's inertia additivity (#218 Step C1; #191 Step 4) + the pair is typed iff `s m_M < 0 < s m_S`. Since `r^{−2}Π = −m_Mm_S`, `r^{−2}Π(1_{typed} − 1) = (m_Mm_S)₊ − + |m_Mm_S|1_{anti}`, where `anti := {s m_M > 0 > s m_S}` (control Q5). With `m_Mm_S = X + ε`, + + G = [(X + ε)₊ − X₊ − ε1{X>0}] + ε1{X>0} − |m_Mm_S|1_{anti}. + + - (a) The bracket is at most `|ε|1{|X| ≤ |ε|}` in modulus (control Q5). On the dyadic layer `{T ∈ [2^j, 2^{j+1})}`, + `|ε| ≤ ε̄_j := C(k|Δ| + r²)(1 + k)^N2^{(j+1)N}`, which does not involve `f₄`; and by (1.2), with `J₀` the finite jet + vector, `P(T ≥ 2^j | J₀) ≤ C_p2^{−jp}(1 + k + |J₀|)^p`. Integrate `f₄` given the other jets. `U` is affine in `f₄` with + slope `Δ/12`, and `f₄` has a bounded conditional density (#218 Step F1), also after multiplication by the polynomial + weight `(1 + k + |J₀|)^p`, which contains `|f₄|^p`. So `{|X| ≤ ε̄_j}` has conditional probability + `≤ C min(1, ε̄_j/(κΔ²))` when `36κ²Δ² > 2ε̄_j`, and `≤ C min(1, ε̄_j^{1/2}/|Δ|)` otherwise. + - On the first set `κ|Δ| ≥ cr`, and there `E[ε̄_j min(1, ε̄_j/(κΔ²))] ≤ C2^{2jN}(1 + k)^N(kr + r³log(2/r))` (use + `k/κ = r` and Lemma D in dyadic layers). + - The second set needs `|Δ| ≤ C2^{jN}r/κ`, which has probability `≤ C2^{jN}r/κ`; there the cost is + `≤ ε̄_j ≤ C2^{jN}(1 + k)^Nr²`. Since `κ ≥ r`, `r³/κ ≤ r²`. + + Summing over `j`, (a) is `O((1 + k)^N(kr + r²))`. + - (b) `ε1{X>0}`. First replace `1_{Case 1}` by `1{A<0}`, at cost `≤ E[|ε|1_{Case 3}] ≤ Cr(kr + r²)(1 + k)^N`. The `ε₂` + part is `≤ Cr²(1 + k)^N/(1 + κ)` by (1.2) and Lemma Λ₀. What remains, `−12kE[ΔV1{X>0}1{A<0}]`, is a Gaussian expectation + of finitely many free jets. By (R.2), their law has a covariance independent of `k` and odd means `k·w`. + - `12k²|E[ΔV_e1{X>0}1{A<0}]|`: on `{X > 0}`, `|Δ| < |U|/(6κ)`, so by Lemma Λ₀ this is + `≤ Ck²(1 + k)^N/(1 + κ)² ≤ C(1 + k)^Nr²`. + - `12kE[ΔV_o1{X>0}1{A<0}]`, in three steps: + - (i) With the law at `k = 0` (which is `𝒫`-invariant by (R.2)) and `Y′` in place of `U` in `X`, the expectation + vanishes: `ΔV_o` is odd and `1{Y′² > 36κ²Δ²}1{A<0}` is even. + - (ii) Moving the odd mean from `0` to `k·w` costs, by the Gaussian score (no derivative of an indicator is + taken), `≤ Ck·sup E[|ΔV_o|(1 + |jets|)1{|Y′| > 6κ|Δ|}1{A<0}] ≤ Ck(1 + k)^N/(1 + κ)²`. On the layer + `|Δ| < |Y′|/(6κ)`, so Lemma Λ₀ applies along the interpolation. + - (iii) Replacing `Y′` by `U` in the indicator (`U − Y′ = 3kΔ_B`). The two indicators differ only where `Y′` is + within `3k|Δ_B|` of `±6κ|Δ|`, that is, for `f₄` in two windows `W` of total length `≤ 72k|Δ_B|/|Δ|`, centred at + `f₄ = 12(±6κ|Δ| − β)/Δ` with `β := −γᵀA^♯γ/4`. Write `ΔV_o = Δ(a₁f₄ + a₀)` with `a₁ = Δ_B/24` and `a₀` free of + `f₄`. Since `sup_x(1 + |x|)p_{f₄}(x) ≤ C(1 + |m₄|)`, the factor `Δ` cancels the `1/|Δ|` of the window length: + `E_{f₄}[|ΔV_o|1_W] ≤ Ck(1 + |m₄|)(|a₁| + |a₀|)|Δ_B|`, uniformly in `Δ`. For `k ≤ 1` and + `|Δ| > (|β| + 3|Δ_B|)/(3κ)`, on `W` we have `|Δf₄|/12 = |Y′ − β| ≥ 6κ|Δ| − 3|Δ_B| − |β| > 3κ|Δ|`, so + `W ⊂ {|f₄| > 36κ}`. There `(1 + |x|)p_{f₄}(x) ≤ C(1 + |m₄|)^Nκ^{−N}`, and this part costs `O(kκ^{−N})`. The + complementary set `{|Δ| ≤ (|β| + 3|Δ_B|)/(3κ)}` has probability `≤ C/κ`, with the polynomial weights (Lemma D in + dyadic layers of `|β| + |Δ_B|`). So this step costs `≤ Ck(1 + k)^N/(1 + κ)`. For `k > 1`, Lemma Λ₀ applied to + each indicator gives `≤ C(1 + k)^N/(1 + κ)`, and the Gaussian factor absorbs the missing `k`. + + Hence `|12kE[ΔV_o1{X>0}1{A<0}]| ≤ Ck²(1 + k)^N/(1 + κ)`, and `k²/(1 + κ) = k²r/(k + r) ≤ kr`. + + In all, (b) is `O((1 + k)^N(kr + r²))`. + - (c) *The anti-typed sliver.* By Lemma Π, `s m_S − s m_M = 12κ|Δ| + O(r(1 + k)T^N)`. On `anti` the left side is + negative, so `|m_M| + |m_S| = −(s m_S − s m_M) ≤ Cr(1 + k)T^N`. Hence `|U| = |m_M + m_S|/2 + O(r²(1 + k)T^N) ≤ + Cr(1 + k)T^N`. Integrating `f₄`, then using Lemma D in layers, this event has probability `≤ Cr log(2/r)(1 + k)^N`. + Its weight is `|m_Mm_S| ≤ Cr²(1 + k)^NT^N`, so (c) is `O(r³log(2/r)(1 + k)^N)`. + + Collecting, with the Gaussian factor (control Q6): + - for `k ≤ 1`, `kr = r min(k, 1/k)` and `r(k + r)² ≤ 2kr + 2r²`; + - for `k ≥ 1`, `(1 + k)^Ne^{−ck²}(kr + r²) ≤ C(r² + r/k)`. + + Hence `𝐓_r − 𝐒_r = 12p_{V_r}(v(k))E_{Q̄_{r,k}}[X₊1{A<0}] + O(r² + r min(k, 1/k))`, the main term being over all of + `{A < 0}`. + +*Step U3 (the layer across laws).* Let `L_r(k) := 12p_{V_r}(v(k))E_{Q̄_{r,k}}[X₊1{A<0}]`. Throughout, `κ = k/r` is a fixed +parameter of the weight. +- *`k ≥ 1`.* `L_r(k) ≤ Ce^{−ck²}(1 + k)^N/(1 + κ)` and `Λ(0) ≤ C/(1 + κ)` (Lemma Λ₀), so `|L_r(k) − Λ(0)| ≤ C/(1 + κ) ≤ Cr/k`. +- *`k ≤ 1`, from `r` to `0` at fixed `(k, κ)`.* `Λ(k)` is the same functional at `r = 0`. By Lemma R (R.4), the laws of the + finitely many jets involved under `Q̄_{r,k}` and `Q̄_{0,k}` have means and covariances that differ by `O(r²(1 + k))`, and + `p_{V_r}(v(k)) − p_{V_0}(v(k)) = O(r²)`. The Gaussian score along the linear interpolation, with Lemma Λ₀ applied to each + interpolating law, gives `|L_r(k) − Λ(k)| ≤ Cr²/(1 + κ)`. +- *`k ≤ 1`, from `k` to `0` at `r = 0`.* By Lemma Λ, `|Λ(k) − Λ(0)| ≤ Ck²/(1 + κ) ≤ Ckr`. + +*Step U4 (the identity).* By #218 (F.2) and #218 §0, `A₂(b, 0, u) = −12π_0E_0[Y²1{A<0} | b]` and +`(𝒜^{cand} − 𝒜^{con})(b, κ, u) = −12π_0E_0[min(Y², 36κ²Δ²)1{A<0} | b]`, with `π_0 = π_0(v_0(b, 0))`. Since +`y² − min(y², c²) = (y² − c²)₊` (control Q5), the difference is `12π_0E_0[(Y² − 36κ²Δ²)₊1{A<0} | b]`. The contact law at +`v_0(b, 0)` conditions `(f(0), V_0)` on `(b, v(0))`, and `Y = Y′` there. So integrating in `b` (as in (R.1)) gives +`Λ(0) = 𝓐(κ, u) − 𝐀₂(0, u)`. + +Steps U1–U4 together give (U.1). ∎ + +At the cusp scale (`κ` fixed, `k = κr ≤ 1`, `r ≤ min(κ, r₃)`), (U.1) and Lemma K give, after the `b`-integral, +`𝐓_r − r^{−2}𝐀_0 = 𝓐(κ) + O((1 + κ)²r²)`. So the `O(r)` correction allowed by #218 Lemma C and #229 Lemma C⁺ vanishes. In +the fold regime (`k` fixed, `r → 0`), the layer `𝓐(κ) − 𝐀₂(0) = O(κ^{−1}) = O(r/k)` is the boundary layer of #218 (F.1), +now identified rather than bounded. That is how Lemma U resolves the matching pairs of #229 Remark 1. + +## 5. Proof of Theorem P + +Put `ρ := ℓ^{1/5}` and take `ℓ` so small that `ρ ≤ r₃` and `ρ < r_0 ≤ r_0^*`. By #191 (2.1) and #207 §7's decomposition of +`B_{d,L}`, as in #218 §4 with its two inner regions merged into one (`F + K`), + + ν_cand − cℓ^{−1/3} − B_{d,L} = ∫_0^ρ∫_S[𝐓_r(ℓ/r³, u) − r^{−2}𝐀_0(ℓ/r³, u)] dσ dr + J₂ + J₃ + J₄ + J₅, + +with `J₂ = −∫_0^ρ∫∫r^{−2}A_r(b, 0, u)`, `J₃ = ∫_ρ^{r_0}∫∫r^{−2}[A_r(b, ℓ/r³, u) − A_r(b, 0, u)]`, +`J₄ = −∫_ρ^∞∫∫r^{−2}A_0` and `J₅ = ∫_{dist(0,y)≥r_0}∫(Ψ_ℓ^{cand} − Ψ_0)`. + +*The main term.* For `r ≤ ρ`, `k = ℓ/r³ ≥ r²`, so Lemma U applies. +- `∫_0^ρ[𝐀₂(ℓ/r³) − 𝐀₂(0)]dr = (ℓ^{1/3}/3)∫_{ℓρ^{−3}}^∞[𝐀₂(k) − 𝐀₂(0)]k^{−4/3}dk`. By (0.1) of #218 this is `c₂ℓ^{1/3}` + (after `∫dσ`) minus `(ℓ^{1/3}/3)∫_0^{ℓρ^{−3}}`, which is `O(ℓ^{1/3}(ℓρ^{−3})^{5/3}) = O(ℓ²ρ^{−5})` by Lemma K. +- `∫_0^ρ𝓐(ℓ/r⁴)dr = ℓ^{1/4}∫_0^{ρℓ^{−1/4}}𝓐(s^{−4})ds = I^{cand}ℓ^{1/4} − O(ℓ^{1/4}(ρℓ^{−1/4})^{−7}) = + I^{cand}ℓ^{1/4} − O(ℓ²ρ^{−7})`. This uses `|𝒜^{cand} − 𝒜^{con}| ≤ Cκ²` (#218 §4) and the convergence at `s → 0`, where + `𝓐 → 𝐀₂(0)`; `I^{cand} = ∫_S∫_0^∞𝓐(s^{−4}) ds dσ` (#207 §7). +- The error of (U.1) integrates to `C∫_0^ρ(r² + r min(ℓ/r³, r³/ℓ))dr ≤ C(ρ³ + (6/5)ℓ^{2/3})` (control Q1). + +*The rest* (#218 §4, #207 §7, #198). +- `|J₂| ≤ Cρ³` (#198 (W.4)). +- `|J₃| ≤ C(ℓ²ρ^{−7} + ℓρ^{−2})` (#207 (7.1); here `k ≤ ℓρ^{−3} ≤ 1`). +- `0 ≤ −J₄ ≤ Cℓ²ρ^{−7}`. +- `|J₅| ≤ Cℓ` (#207 (7.2)). + +*The ledger* (control Q1; exponents of `ℓ` at `ρ = ℓ^{1/5}`): + +| Term | Source | Exponent | +|---|---|---| +| `ρ³` | the `r²` part of Lemma U; `J₂` | **3/5** | +| `ℓ²ρ^{−7}` | the cusp tail; `J₃`; `J₄` | **3/5** | +| `ℓρ^{−2}` | `J₃` | **3/5** | +| `ℓ^{2/3}` | the `r min(k, 1/k)` part of Lemma U (fold scale) | 2/3 | +| `ℓ²ρ^{−5}` | the finite part of `𝐀₂` | 1 | +| `ℓ` | `J₅` | 1 | + +The least exponent is `3/5`. The split is optimal for this ledger, and it is exactly the edge `κ = r` of Lemma U's range. +This proves (P.1). ∎ + +(Lemma K's `O(k²)` is not needed for (P.1). With #218's `C¹` bound `𝐀₂(k) − 𝐀₂(0) = O(k)`, the finite-part row becomes +`ℓρ^{−2}`, again of exponent `3/5` (control Q1). Lemma K is used for the cusp-scale statement after Lemma U.) + +## 6. Remarks + +1. **The `ℓ^{1/2}` order.** #218 §0 ("Formally the next term is of order `ℓ^{1/2}`") and #229 anticipated an `ℓ^{1/2}` term. + #229 §0 says the same; its Remark 1 says "formally `𝒜₃` produces the next term, of order `ℓ^{1/2}`", where `𝒜₃` is the + `O(r)` correction of the cusp kernel. For the candidate density the term is absent. After the `b`-integral, the + cusp-scale expansion of the candidate kernel has no `O(r)` term (Lemma U with Lemma K). The would-be term is + `12κrΔV_o` in `r^{−2}Π` (Lemma Π), which is odd under `𝒫`, while the dependence of the law and of `U` on `k` is even + (Lemmas R and Λ). + - Reflection alone does not make the expansion even to all orders. It exchanges typed and anti-typed pairs, whose + `O(r³log(2/r))` contribution (Step U2 (c)) is not expanded here. + - #216's formal law (0.2) concerns the elder density, with remainder `O(ℓ^{1/2})`. Theorem P neither confirms nor + contradicts it (Remark 3). +2. **What the method leaves.** + - *The fold scale.* There `𝓐(κ) − 𝐀₂(0) − [𝐓_r − 𝐒_r]` is `O(r min(k, 1/k))`. This comes from the `k`-dependence of + the layer coefficient (Lemma Λ) and from Step U2 (b), and it integrates to `O(ℓ^{2/3})`. It is plausibly the next + term, but neither its non-vanishing nor the sharpness of Lemma Λ is proved. The referee's numerics support a nonzero + `k²/κ` coefficient for the Gaussian kernel in `d = 2` (README, review record). + - *What fixes `3/5`.* Three terms balance at `ρ = ℓ^{1/5}`: + - `ρ³`, from the `r²` part of Lemma U and from `J₂`; + - `ℓ²ρ^{−7}`, from the cusp tail, `J₃` and `J₄`; + - `ℓρ^{−2}`, from `J₃` via #207 (7.1). + + They are fixed together with the cap `ρ ≤ ℓ^{1/5}`, which Lemma U's range `κ ≥ r` imposes. Sharpening the `r²` part of + Lemma U alone would not change `3/5`. + - *A route to `2/3` (not attempted).* An exponent `2/3` would follow from a version of Lemma U for `𝐓_r(k) − 𝐓_r(0)` + with three properties: + - it absorbs `J₂`; + - it holds for all `k > 0`, so that `ρ` can be fixed; + - its error is `O(r²min(1, κ) + r min(k, 1/k))`. + + With `ρ` fixed, the remaining terms are `O(ℓ^{3/4})`, `O(ℓ^{2/3})` and `O(ℓ)`. +3. **The elder density.** + - *Formally.* The leading elder window `{|Y′| < 2κ|Δ|}` (#229 Remark 2) is `𝒫`-even, so formally the elder cusp kernel + has no `O(r)` term either. + - *Why the proof does not transfer.* Reflection does not make the elder kernel even in `k`, since it exchanges typed and + anti-typed configurations. Lemma U is not available for the elder kernel: the elder decision's bad events (#220 + Lemma Q; #229 Lemma CE⁺: `κ²r^{3/2}`, `κ³r²`) are not uniform in the fold regime. The intermediate elder mass is also + only known to be `O(ℓ^{4/9}log(1/ℓ))` (#229). + - Theorem P says nothing new about `ν_eld` or `ρ_rej`. +4. **Numerical evidence** (exploration; project archive `V2_2/frontiers_candidate_parity_rate_20261001/exploration/`; not + part of the proof). + - *The cusp kernel's `O(r)` coefficient.* `d = 2`, Gaussian kernel. + - *Method.* `r^{−2}A_r(b, κr)` was computed by exact Gaussian conditioning of the two endpoint Hessians on the pin + rows (mpmath, 60 digits), then by Monte Carlo with common random numbers over ten separations + `r ≤ c := 0.1/(1 + κ)²` (`cusp_parity_mc.py`, `cusp_parity_fine.py`). + - *Fits.* Polynomial fits `g₀ + g₁r + g₂r² + …` of degree 2–4 were made for `κ ∈ {1/2, 1, 2, 4}` and `b ∈ {0, 1}`. + Across the grid the linear term changes the kernel by `|g₁c/g₀| ≤ 5·10⁻⁵`, against `|g₂c²/g₀| ≈ 3–8·10⁻³` for the + quadratic term. + - *Check.* `g₀` agrees with a Monte Carlo value of `𝒜^{cand}(κ)` within `0.6%` (about two standard errors of that + value; `g₀` is lower in all eight cases). + - *The layer* (`layer_match_mc.py`). At fixed `k ∈ {0.05, 0.1, 0.2, 0.4, 0.8}` and `r ∈ [0.005, 0.04]`, the + typed-minus-surrogate difference agrees with the cusp tail `𝒜(κ) − A₂(0)` within `O(rk)`, as (U.1) predicts. The + relative discrepancy is `O(k²)`, and the factor `p_{V_r}(v(k)) = O(e^{−12k²})` makes it of order one at `k ≈ 1`, the + fold scale. + - *The full-field Monte Carlo of #216.* + - *Data.* Binned counts in the project archive, `V2_2/numerics_lifetime_mc_20261001/results/mc{2,3}_final.json`; + fits by `refit_mc.py` and `scan_p.py`. + - *What it measures.* The *adjacent*-pair density, not `ν_cand`. Formally the two have the same `ℓ^{−1/3}`, + `ℓ^{1/4}` and `ℓ^{1/3}` coefficients, but adjacent pairs have no `B_{d,L}` term, and nothing on file proves the + adjacent-pair expansion (#218 Remark 3). So the comparison concerns the common formal coefficients only. + - *Adjacent pairs.* The residual about the three-term law gives `ℓ^{1/2}` coefficients, fitting `ℓ^{1/2}` and `ℓ^{3/4}` + jointly, of: + - `−0.003 ± 0.005` (`d = 2`) and `+0.003 ± 0.007` (`d = 3`) on `[10⁻⁴, 0.3]`; + - `+0.000 ± 0.020` and `−0.023 ± 0.030` on `[10⁻⁴, 0.1]`. + + All are consistent with zero. A single extra power (scanned over `p ∈ [0.10, 1.20]`) is best fitted at + `p ≈ 0.7–0.8` (`d = 2`) and `0.68` (`d = 3`, on `[10⁻⁴, 0.3]`). + - *Elder and rejected pairs.* These residuals behave differently: they are not consistent with zero. A single extra + power fits best at `p ≈ 0.14–0.33` in six of eight fits; the others are `0.54` (elder, `d = 3`, `[10⁻⁴, 0.1]`) and + a poor fit (rejected, `d = 2`, `[10⁻⁴, 0.3]`). So the data do not single out `ℓ^{1/2}`. In the joint + `ℓ^{1/2} + ℓ^{3/4}` fits the `ℓ^{1/2}` coefficients are `+0.033 ± 0.004` and `+0.054 ± 0.005` (elder), and + `−0.038 ± 0.002` and `−0.049 ± 0.004` (rejected). + - *Conclusion.* This is consistent with Theorem P and with Remark 3. Explaining the elder residual (non-analytic + decision effects, or Monte Carlo systematics) is open. +5. **`d = 1`.** In `d = 1` the layer is exponentially small in `κ` (the window is on `f₄` alone), and there is no + transverse layer. The same reflection removes the `O(r)` cusp term. This is consistent with #214 §0 ("the numerics of + §6.2 suggest the next terms are `O(h^{3/4})`"). +6. **Consistency.** + - (P.1) implies #229 (T⁺.1) and #218 (T.1). + - Lemma U at fixed `κ` implies, after the `b`-integral, #229 (C⁺.1) for `r ≤ min(κ, r₃)`, in the stronger form + `O((1 + κ)²r²)`. + - Lemma K and (2.2) agree with #232 (J1) and (2.3). + - #229 Remark 2 bounded the kink of the candidate weight by `O(κk²)` and expected `O(k²)`. Lemma Λ controls the kink + through the density of its location instead, and gets `O(k²/(1 + κ))` for the layer part. + - With `c` and `c₂` from #223 (certified enclosures there) and `I^{cand}` from #207 (exploration), the SIDE24 candidate + density is `ν_cand ≈ 0.0417759ℓ^{−1/3} + B_{3,24} − 0.1394041ℓ^{1/4} + 0.1612340ℓ^{1/3} + O(ℓ^{3/5})`. + +## 7. Sources (exact identities in `SOURCES.json`) + +| Tag | Path | Role | +|---|---|---| +| [R] | `frontiers/three_fronts_20260924/LIFETIME_REMAINDER.md` (`247b3ecf`) | §2 rows, target, Jacobian 12; (R2)–(R5) — consumed | +| [P] | `imports/lifetime_parent_20260925/UNIFORM_MATRIX_CAP_AND_LIFETIME.md` (`dfed3b8d`) | §2 finite-jet rank — consumed | +| [E1] | `imports/lifetime_parent_20260925/ERRATUM_CONGRUENCE.md` (`213594d6`) | reading rule | +| [E2] | `reviews/d1_section9_borel_repair_20260925/REPAIR.md` (`fe9b9ce4`) | §9 replacement | +| [REC] | `reviews/d1_chain_reconciliation_20260928/RECONCILIATION.md` (`75da2597`) | reading rule | +| [Z] | `frontiers/c7_zero_gap_limit_20260929/PROOF.md` (`5b6328ea`) | (Z2)–(Z3): `Ψ_0`, `B_{d,L}` — consumed | +| #191 | `frontiers/remainder_vanishing_20260930/PROOF.md` (`441152df`, merged) | `r_0^*`, (2.1), Lemma E Step 4; Lemma E Steps 1–3 through #218 — consumed | +| #198 | `frontiers/remainder_rate_20260930/PROOF.md` (`abfb98ae`, merged) | (W.4); (3.1), §3 sign window and Lemma W through #218 and #207 — consumed | +| #207 | `frontiers/cusp_second_order_20261001/PROOF.md` (`f6df5a73`, merged at `566b1a1`) | §0 (`Y`, `Δ`, `E₀`); §4 (4.1); §7 Lemma L (7.1)–(7.2), the `B_{d,L}` decomposition, `I^{cand}` ((CU′.1), (7.3)) — consumed | +| #218 | `frontiers/candidate_third_order_20261001/PROOF.md` (`70ca57ef`, merged at `fb6ee97`) | §0 (`A₂`, `c₂`); §1 (Lemma D, `T`, (1.2)); Steps F1 and F3, (F.2), the surrogate remark; Lemma C Step C1; Lemma O's proof; §4 — consumed | +| #229 | `frontiers/third_order_rate_20261001/PROOF.md` (unmerged) | Theorem T⁺, (C⁺.1), Remarks 1–2 — cited | +| #232 | `frontiers/c2_finite_jet_transfer_20261001/PROOF.md` (`54cc4a1a`, merged at `7fe06b0`) | (2.3), §1, (3.2), §4, Theorem J — cited (independent agreement) | +| #216, #223, #214, #220 | as in `SOURCES.json` | `c₂` and #216's formal law (0.2); `c`, `c₂` enclosures; `d = 1`; the elder density — cited | + +## 8. Exact controls (`parity_check.py`; stdlib; exact rationals; byte-identical under `-O` and on CPython 3.10–3.14) + +| Control | Checks | +|---|---| +| Q1 | the ledger of §5: least `3/5` at `ρ = ℓ^{1/5}`, attained by `ρ³`, `ℓ²ρ^{−7}`, `ℓρ^{−2}`; the split is optimal on a grid and is the edge `k = r²`; the ledger without Lemma K also gives `3/5`; the closed form `∫_0^∞ r min(k, 1/k) dr = (6/5)ℓ^{2/3}` (antiderivatives evaluated exactly at three values); `3/5 > 1/2 > 3/7 > 4/11` | +| Q2 | Lemma R and the evenness of Lemma Π on fourteen exactly pinned polynomial fields (`d = 1, 2, 3, 4`): pinned jets even in `r`; `Π` even; the new row of §1 has target `0` (a consistency identity); point reflection maps the pins at `k` to those at `−k` and preserves `Π`; `det K_M` and `det K_S` affine in `k` (at `d + 1` values of `k`, which determine a polynomial of degree `≤ d`) | +| Q3 | (2.2) exactly on the same fields: `det K_{M,S} = ∓6kΔ + rU ∓ r²V + …`, `Π = 36k²Δ² + r²(12kΔV − U²) + …`, with `Δ_BB = ∂_t²det(A + tB)` and `A^♯_B = ∂_t adj(A + tB)` at `t = 0` in general form (so `d = 4` tests the terms that vanish or simplify for `m ≤ 2`); `U` invariant and `V` odd under `(k, odd jets) ↦ (−k, −odd jets)` | +| Q4 | Lemma K for the Gaussian kernel: the `A`-integrand `H(k, A)` of the birth-integrated `r²` coefficient (exact Gaussian conditioning and Isserlis) has only `k⁰, k², k⁴` terms in `d = 1, 2, 3`, and `a = 12`; the `d = 1, 2` polynomials equal #232 §4's | +| Q5 | Step U2's pointwise algebra on rational grids: `y² − min(y², c²) = (y² − c²)₊`; `|(x + e)₊ − x₊ − e1{x>0}| ≤ |e|1{|x| ≤ |e|}`; the Case 1 sign decomposition with the anti-typed term; the Case 3 identity `P(1_{typed} − 1) − X₊ = P1_{typed} − (−X)₊ + ε` | +| Q6 | the inequality `(r² + min(1, k²))/(1 + κ) ≤ 2(r² + r min(k, 1/k))` for `k ≥ r²` on a rational grid; and, as consistency identities that cannot fail, `k/(1 + κ) ≤ r`, `r³/κ ≤ r²` (equivalent to `k ≥ r²`) and `kr = r min(k, 1/k)` for `k ≤ 1` | + +Mutants `M1`–`M8` each fail only their own control: +- M1: the split `ℓ^{1/6}`; +- M2: the height of `M` in place of the midpoint; +- M3: `U` without `3kΔ_B`; +- M4: `V` without `f₅`; +- M5: a constant offset in `T_r`'s target; +- M6: the decomposition without the anti-typed term; +- M7: a false bookkeeping inequality; +- M8: the saddle's Hessian at `r/2 + r⁴`, which breaks the endpoint symmetry of `Π` (its evenness in `r` and its + reflection invariance, both checked in Q2). + +An unknown label exits 2. + + python3 -B -S parity_check.py # exit 0, output = RESULTS.json + python3 -B -S parity_check.py --mutant M3 # exit 1 + +The controls check exact algebra, symmetry and arithmetic. The Gaussian layer estimates of §§3–4 (Lemmas D″, Λ₀, Λ, U) +are proved in prose only. + +## 9. Review slices + +- **A** §§1–2: Lemma R (birth integration, parity, reflection, evenness in `r`), Lemma Π, Lemma K. +- **B** §3: Lemmas D′, D″, Λ₀ and Λ (the coupling, the `f₄`-smoothing, the kink: main set, change of variables, + exceptional set). +- **C** §§4–5: Lemma U (Steps U1–U4: the case analysis, (a)–(c), the comparison of laws) and the proof of Theorem P with + its ledger. diff --git a/frontiers/candidate_parity_rate_20261001/README.md b/frontiers/candidate_parity_rate_20261001/README.md new file mode 100644 index 000000000..717f30a88 --- /dev/null +++ b/frontiers/candidate_parity_rate_20261001/README.md @@ -0,0 +1,119 @@ +# The candidate density with remainder `ℓ^{3/5}`: no `ℓ^{1/2}` term + +**Author-side proof candidate (Anthropic Claude). Scientific effect: NONE. Nonauthor review required.** + +Object `CL-CANDIDATE-PARITY-RATE-20261001-v1`. Full text: [`PROOF.md`](PROOF.md). + +## Result + +For every `d ≥ 2` and `L > 0`, as `ℓ ↓ 0`: + + ν_cand(ℓ) = c ℓ^{−1/3} + B_{d,L} + I^{cand} ℓ^{1/4} + c₂ ℓ^{1/3} + O(ℓ^{3/5}) (P.1) + +Math- #218 proved this with `O(ℓ^{4/11})` and Math- #229 with `O(ℓ^{3/7})`. Since `3/5 > 1/2`, **the candidate density +has no term of order `ℓ^{1/2}`** (Corollary P′), although #218 §0 and #229 §0 name `ℓ^{1/2}` as the formal next order. + +## The three ingredients + +| | Statement | Role | +|---|---|---| +| **Parity and reflection** (Lemmas R, Π, K) | After the birth height is integrated out, the field splits into independent even and odd parts, and the pin rows are even in the separation `r`. Hence `Π = −det K_M det K_S = 36k²Δ² + r²(12kΔV − U²) + O(r⁴)`. Point reflection `z ↦ −z` maps the problem at gap `k` to the one at `−k`, so the fold coefficient `𝐀₂` is even in `k`. | The pathwise evenness is in #191 Lemma E Step 3. The basis and the `k`-parity of `𝐀₂` are in #232 (§§1, 3, Theorem J). New: the reflection is applied to the conditional laws inside the cusp layer. | +| **Lemma U** (uniform two-scale kernel) | For `k ≥ r²`, the birth-integrated kernel is `r^{−2}𝐀_0 + [𝐀₂(k) − 𝐀₂(0)] + 𝓐(k/r)` up to `O(r² + r min(k, 1/k))`. Here `𝓐` is the birth-integrated cusp kernel `𝒜^{cand} − 𝒜^{con}`. | One region replaces #218's fold/cusp split. The boundary layer of the fold expansion is identified as the cusp kernel's tail, and the `O(r)` cusp correction vanishes. | +| **Lemma Λ** (the gap inside the layer) | Moving the gap from `k` to `0` inside the layer `{|U| > 6κ|det A|}` costs `O(k²/(1 + κ))`. The first order vanishes by reflection. The kink is controlled through the density of its location, using Lemma D″ (two small eigenvalues) for the exceptional set. | Replaces #229 Remark 2's kink bound `O(κk²)`. | + +#229 Remark 1 explains why a two-region argument stops at `3/7`: two matching pairs of error terms have unidentified +leading coefficients. Lemma U removes both pairs. There is no split point and so no truncation, and the linear cusp +error vanishes inside Lemma U (Step U2 (b), Lemma Λ). + +The remaining terms `ρ³`, `ℓ²ρ^{−7}` and `ℓρ^{−2}` balance at `ρ = ℓ^{1/5}`, the edge `k = r²` of Lemma U's range. Lemma K's +`O(k²)` is not needed for (P.1): #218's `C¹` bound gives the same exponent. + +## Dependencies + +| | Sources | +|---|---| +| Consumed (all merged) | #218 (`70ca57ef`, merged 1 Oct at `fb6ee97`): §0; §1 (Lemma D, `T`, (1.2)); Steps F1 and F3; (F.2) and the surrogate remark; Step C1; Lemma O's proof; §4. #207 (`f6df5a73`, merged 1 Oct at `566b1a1`): §0, §4 (4.1), §7 (Lemma L, `I^{cand}`). [R], [P] with [E1]/[E2]/[REC], [Z], #191 (`r_0^*`, (2.1), Lemma E Step 4), #198 ((W.4)) | +| Cited | #229 (Theorem T⁺, superseded for `ν_cand`; Remarks 1–2); #232 (merged at `7fe06b0`: an independent derivation of (2.2) and of the `k`-parity of `𝐀₂`); #216 and #223 (`c₂`; #223 also certifies `c`); #214 (`d = 1`); #220 (the elder density; not used) | + +Every consumed source is merged, with the reviewed blob. The workflow checks every pin in the tree, and binds the +cited unmerged sources to their recorded heads. + +## Controls + +`parity_check.py` uses the standard library and exact rationals. Its output is `RESULTS.json`, byte-identical under `-O` +and on CPython 3.10–3.14 (a few seconds). + +| Control | Checks | +|---|---| +| Q1 | The ledger: least `3/5` at `ρ = ℓ^{1/5}`, attained by `ρ³`, `ℓ²ρ^{−7}`, `ℓρ^{−2}`. The split is optimal. The exponent is still `3/5` without Lemma K. The closed form of `∫ r min(k, 1/k) dr` is evaluated (a consistency identity). | +| Q2 | On fourteen exactly pinned polynomial fields (`d = 1, 2, 3, 4`): pinned jets are even in `r`; `Π` is even; the reflection maps `k` to `−k`; `det K_M` and `det K_S` are affine in `k`, checked at `d + 1` values. | +| Q3 | (2.2) exactly on the same fields, with the general `Δ_BB` and `A^♯_B`. The `d = 4` fields fail the old `m ≤ 2` shortcut, so they test the `m ≥ 3` terms. The parities of `U` and `V` are checked. | +| Q4 | Lemma K for the Gaussian kernel. The `A`-integrand `H(k, A)` of the birth-integrated `r²` coefficient has only `k⁰, k², k⁴` terms (`d = 1, 2, 3`). The `d = 1, 2` polynomials are #232's. | +| Q5, Q6 | The pointwise algebra of Step U2, including the Case 3 identity, and the inequalities that collect Lemma U's errors. Three of Q6's four relations are identities, recorded as consistency checks. | + +Mutants `M1`–`M8` each fail only their own control, and an unknown label exits 2. M8 moves the saddle's Hessian to +`r/2 + r⁴`, which breaks the endpoint symmetry of `Π`: its evenness in `r` and its reflection invariance, both in Q2. + + python3 -B -S parity_check.py # exit 0, output = RESULTS.json + python3 -B -S parity_check.py --mutant M3 # exit 1 + +The controls check exact algebra, symmetry and arithmetic. The Gaussian layer estimates (Lemmas D″, Λ₀, Λ, U) are proved +in prose only. + +## Review record + +Two clean-context same-family referees (Anthropic Claude subagents) reviewed the note before submission. Referee A took +§§0–3 and referee B took §§4–8, the controls and this README. Each read the whole note against all sources at their +declared blobs. + +**First pass: MAJOR REVISION, narrow.** Each referee found local defects with explicit repairs. None falsified Theorem P, +but two left steps of its written proof incomplete or incorrect; both are repaired. +- Lemma Λ's exceptional set did not cover `d ≥ 4`, so the proof was incomplete there. It is repaired with Lemma D″, a + full Vandermonde bound `P(D < ε) ≤ Cε²`. +- Step U2's Case 3 bullet was false as written, because `U` is not small there. It now compares with the layer + integrand. +- Remark 2 misidentified what fixes `3/5`, and its route to `ℓ^{2/3}` was wrong. It is rewritten. + +**Other fixes applied.** Seventeen minor findings and thirteen nits, including: +- the kink computation of Lemma Λ, made rigorous with a change of variables; +- the smoothness argument for Lemma K, by conditioning on `A`; +- Lemma R's quantifiers and the even/odd decomposition (R.2); +- Lemma Λ₀ stated for random shifts and interpolated laws; +- Step U2 (a)'s first-piece bound near `κ = r`, and Step U2 (b)'s restriction to Case 1; +- Step U3, which no longer relies on #229 Remark 2; +- attribution to #191 Step 3 and #232; +- citations: #207 §4 (4.1), #214 §0, and #223's certified values; +- notation clashes; +- the controls: Q3 with general `m` and `d = 4`, `K_S` affinity, the new M8, and honest labels for Q1, Q2, Q4 and Q6; +- `SOURCES.json` and the workflow, added before submission. + +**Second pass (delta check of the revision).** Referee A: **ACCEPT** for §§0–3, with six nits. Referee B: **ACCEPT WITH +MINOR FIXES** for §§4–8, with three minor items and six nits. All are applied, including: +- the one substantive item, Step U2 (b)(iii): the factor `Δ` of `ΔV_o` cancels the `1/|Δ|` of the `f₄`-window length, + uniformly in `Δ`; +- the wording on the elder and rejected residuals: the single-power scan does not single out `ℓ^{1/2}`; +- the archive of the exploration scripts. + +**The referees' independent checks.** +- (2.1)–(2.2) were re-derived symbolically for `d = 1, 2, 3` and checked exactly for `d = 2`–`5`. +- The typed criterion and the Case 1 decomposition were checked against true inertias on 80,000 random matrices + (`d = 2`–`5`), with no mismatch. +- Exact finite-`r` conditioning confirmed Lemma K and Q4 for `d = 1, 2`. +- Lemma Λ toys (`m = 1, 2, 3`) show `κ(Λ(k) − Λ(0))/k²` flat in `κ`, and `P(D < ε) ∝ ε^{2.5–3}`. +- A `d = 2` Gaussian-kernel Monte Carlo with exact finite-`r` conditioning matched `𝐓_r − 𝐒_r` to the layer term at 18 + points, to 1–2%. At fixed `κ` it found no `O(r)` term: the relative linear coefficient was `−0.007 ± 0.011` at `κ = 1` + and `−0.065 ± 0.041` at `κ = 2`. The `k²/κ` coefficient of the layer was nonzero, about `+0.005`. +- In the same model, the cost of Step U2 (b)(iii) satisfies `κG(k)/k ∈ [0.0002, 0.0066]` for `κ ∈ {5, 20, 80}`. +- Control replay was byte-identical on 3.10–3.14 with and without `-O`, for the reviewed and the revised packet, and the + mutant matrix was clean. + +**Independence.** The referees are the same provider and the same GitHub account as the author, so they carry zero +organizational independence. Nonauthor review is required; the review slices are in PROOF §9. + +## Not claimed + +- sharpness of `3/5`. The next ledger term is `ℓ^{2/3}`, and reaching it would need Lemma U for `𝐓_r(k) − 𝐓_r(0)` on all + `k > 0` (PROOF Remark 2). +- anything new about the elder or rejected densities. The Monte Carlo of #216 suggests that they carry further terms + that the adjacent-pair density does not (PROOF Remarks 3–4). +- certified values of `I^{cand}` or `B_{d,L}`; uniformity in `d` or `L`. diff --git a/frontiers/candidate_parity_rate_20261001/RESULTS.json b/frontiers/candidate_parity_rate_20261001/RESULTS.json new file mode 100644 index 000000000..7e7d467fe --- /dev/null +++ b/frontiers/candidate_parity_rate_20261001/RESULTS.json @@ -0,0 +1,110 @@ +{ + "controls": { + "Q1_ledger": true, + "Q2_reflection_evenness": true, + "Q3_second_coefficient": true, + "Q4_k_parity_reference": true, + "Q5_layer_algebra": true, + "Q6_bookkeeping": true + }, + "ledger": { + "attained_by": [ + "l rho^-2 (J3)", + "l^2 rho^-7 (cusp tail; J3; J4)", + "rho^3 (Lemma U, r^2 part; J2)" + ], + "least": "3/5", + "least_without_lemma_K": "3/5", + "optimal_split": "1/5", + "split": "1/5", + "terms": { + "l (J5)": "1", + "l rho^-2 (J3)": "3/5", + "l^(2/3) (Lemma U, r min(k,1/k) part)": "2/3", + "l^2 rho^-5 (A_2 finite part)": "1", + "l^2 rho^-7 (cusp tail; J3; J4)": "3/5", + "rho^3 (Lemma U, r^2 part; J2)": "3/5" + } + }, + "object": "CL-CANDIDATE-PARITY-RATE-20261001-v1", + "pinned_fields_by_d": { + "1": 4, + "2": 4, + "3": 4, + "4": 2 + }, + "pinned_fields_checked": 14, + "reference_H_polynomials": { + "d1": { + "0": [ + [ + [], + "-5/24" + ] + ], + "2": [ + [ + [], + "9/2" + ] + ], + "4": [ + [ + [], + "108" + ] + ] + }, + "d2": { + "0": [ + [ + [ + 0 + ], + "-3/4" + ], + [ + [ + 2 + ], + "-13/96" + ], + [ + [ + 4 + ], + "-1/256" + ] + ], + "2": [ + [ + [ + 0 + ], + "-18" + ], + [ + [ + 2 + ], + "57/8" + ], + [ + [ + 4 + ], + "-9/64" + ] + ], + "4": [ + [ + [ + 2 + ], + "108" + ] + ] + } + }, + "scientific_effect": "NONE" +} diff --git a/frontiers/candidate_parity_rate_20261001/SOURCES.json b/frontiers/candidate_parity_rate_20261001/SOURCES.json new file mode 100644 index 000000000..95ecb2845 --- /dev/null +++ b/frontiers/candidate_parity_rate_20261001/SOURCES.json @@ -0,0 +1,240 @@ +{ + "schema": 1, + "scientific_effect": "NONE", + "object": "CL-CANDIDATE-PARITY-RATE-20261001-v1", + "kind": "author-side proof candidate: the candidate lifetime density with remainder O(l^(3/5)) in every d >= 2 (Theorem P): nu_cand = c l^(-1/3) + B_{d,L} + I^cand l^(1/4) + c2 l^(1/3) + O(l^(3/5)), improving O(l^(4/11)) (Math- #218) and O(l^(3/7)) (Math- #229); Corollary P-prime: no l^(1/2) term. New: Lemma R (birth integration; independent even/odd parts; point reflection k -> -k; evenness in r), Lemma Pi (Pi = 36 k^2 Delta^2 + r^2 (12 k Delta V - U^2) + O(r^4)), Lemma K (the surrogate is smooth and even in r; A_2 even in k), Lemma D-double-prime (two small eigenvalues, eps^2), Lemma Lambda (the gap inside the cusp layer costs O(k^2/(1 + kappa))), Lemma U (one uniform two-scale kernel formula for k >= r^2 with error O(r^2 + r min(k, 1/k)); at the cusp scale the O(r) correction vanishes)", + "author": "Anthropic Claude", + "session": "https://claude.ai/code/session_01NMeKEismAyeqgdB4sy2NJU", + "read_at": { + "branch_base": "fb6ee97bc7916edcebba1f236f65762f2eb50570" + }, + "consumed": [ + { + "path": "frontiers/three_fronts_20260924/LIFETIME_REMAINDER.md", + "bytes": 17734, + "sha256": "380b7d0abdb0fe2de5a6564565af9560d3f1b1ce22fd5538927db0c52f4f3a4a", + "git_blob": "247b3ecf80bfbe896948d5d489b2d5842a81c481", + "tag": "R", + "use": "section 2 rows, target (b - k r^3/2, -k r^2, 0, 12k, 0, ...) and Jacobian 12; (R2)-(R5): coupling, moments, pin densities" + }, + { + "path": "imports/lifetime_parent_20260925/UNIFORM_MATRIX_CAP_AND_LIFETIME.md", + "bytes": 40261, + "sha256": "9350ad6eaba6626b93c3dedeef9e2ff816e5cdf1c8318e85fb27499141c84bc7", + "git_blob": "dfed3b8d318a3ab1950957f393307733a4bef3f2", + "tag": "P", + "use": "section 2 finite-jet rank (nondegeneracy in Lemma R (R.4), as in #218 Step F1)" + }, + { + "path": "imports/lifetime_parent_20260925/ERRATUM_CONGRUENCE.md", + "bytes": 1782, + "sha256": "bad7ef609c4ad8c41ad6af562c1b6807921e19a9d556ed793ad1a0db6e202028", + "git_blob": "213594d6ca6a86fb938110f4d166d9ce275a02d0", + "tag": "E1", + "use": "reading rule" + }, + { + "path": "reviews/d1_section9_borel_repair_20260925/REPAIR.md", + "bytes": 9062, + "sha256": "845abf9f9c99d672c2a10a887b5a2e7206a3d2de3d876f35f75ff6e2dc13e62f", + "git_blob": "fe9b9ce4999908bb3814b500ee2d0ceb0c6f704a", + "tag": "E2", + "use": "section 9 replacement" + }, + { + "path": "reviews/d1_chain_reconciliation_20260928/RECONCILIATION.md", + "bytes": 23312, + "sha256": "451b9d7ffee072a73fc904cab891b6693e3f233808b89df10cce1b57036b65da", + "git_blob": "75da2597971510f843f8d90c743950cb8c177342", + "tag": "REC", + "use": "reading rule" + }, + { + "path": "frontiers/c7_zero_gap_limit_20260929/PROOF.md", + "bytes": 17592, + "sha256": "f0e563c45c41333f2bb6e5bf5cb98f0d3b592e2abf2af1c1b098c21ad77cb777", + "git_blob": "5b6328ea3d6afcbeb731698b3834e536f72d1bcb", + "tag": "Z", + "use": "(Z2)-(Z3): the equal-height kernel Psi_0 and B_{d,L}" + }, + { + "path": "frontiers/remainder_vanishing_20260930/PROOF.md", + "bytes": 25353, + "sha256": "0cb5b949261476f10e539677e2744e3288752474393af25ba787e67b1350f32a", + "git_blob": "441152dfc1d8d00cb268cebdb0394212a5bfacdb", + "tag": "191", + "use": "Math- #191 (merged 2026-10-01): r_0^*, section 2 (2.1) and Lemma E Step 4 (Haynsworth) directly; Lemma E Step 3 for attribution (pathwise evenness of det H_M det H_S / r^2); Lemma E Steps 1-3 through #218" + }, + { + "path": "frontiers/remainder_rate_20260930/PROOF.md", + "bytes": 33131, + "sha256": "0b690df6ac40707ca06783f5641d8b480277c716df75b1031d84ee53a0575f72", + "git_blob": "abfb98ae41d66a86d3cb094162db6547d0f29e7e", + "tag": "198", + "use": "Math- #198 (merged 2026-10-01): (W.4) directly (|J2| <= C rho^3); (3.1), the sign window of section 3 and Lemma W through #218 and #207" + }, + { + "path": "frontiers/cusp_second_order_20261001/PROOF.md", + "bytes": 52605, + "sha256": "cffc1b83ac364da58bd650c3c301ea1babd38e19af54f74010bdc120edce38a8", + "git_blob": "f6df5a7356a54771b0052e42a864a140709903ad", + "tag": "207", + "use": "Math- #207 (merged 2026-10-01 19:29 UTC at 566b1a1 from head 74a3943; blob unchanged from the reviewed 96f8aad): section 0 (Y, Delta, E_0); section 4 (4.1) (A^cand, A^con); section 7: Lemma L (7.1)-(7.2), the decomposition of B_{d,L}, I^cand ((CU-prime.1), (7.3))" + }, + { + "path": "frontiers/candidate_third_order_20261001/PROOF.md", + "bytes": 37220, + "sha256": "6c1279164499d4fe2fbeee244b40dafa2159758d20f04826cd67d7602ea97fe4", + "git_blob": "70ca57ef05199e75417d7d57928c1e1c93bf944e", + "tag": "218", + "use": "Math- #218 (merged 2026-10-01 19:39 UTC at fb6ee97 from head 94758c0, the integrator's guarded main-only refresh of the reviewed 0cf048d; blob unchanged): section 0 (A_2, the coefficient c2 (0.1), the cusp kernels restated); section 1 (Lemma D (1.1) and its proof; the dominating variable T and (1.2)); Step F1 ((2.1)-(2.2) and the free-jet nondegeneracy); Step F3; (F.2) and the surrogate remark after Lemma F; Lemma C Step C1 (its three cases); the proof of Lemma O; section 4 (the decomposition and the bounds on J2-J5)" + } + ], + "consumed_unmerged": [], + "cited_only": [ + { + "path": "frontiers/c2_finite_jet_transfer_20261001/PROOF.md", + "bytes": 25581, + "sha256": "971101de946fd711207454c2cd4bbc36729ca9b7c07262e122ff1d54bf5dfa16", + "git_blob": "54cc4a1ae274842bc2c01f9c207549e7b07de7bc", + "tag": "232", + "use": "Math- #232 (merged 2026-10-01 at 7fe06b0): (2.3), section 1 (the birth-integrated basis), (3.2), section 4 (the d = 1, 2 polynomials), Theorem J; an independent derivation of (2.2) and of the k-parity of A_2 (cited)" + }, + { + "path": "frontiers/c2_exact_20261001/NOTE.md", + "bytes": 14480, + "sha256": "598fc86a7c6b21bad438ef087e5aa7521f0fae08e25fa4f40f9b2ba635846dbb", + "git_blob": "a2798b87674ed5fc0d2144697e77ccb9b185b4ab", + "tag": "223", + "use": "Math- #223 (merged 2026-10-01 at f9ebba1): c and c2 in closed form with certified enclosures (Remark 6)" + } + ], + "cited_unmerged": [ + { + "path": "frontiers/third_order_rate_20261001/PROOF.md", + "bytes": 37626, + "sha256": "5de5a6c7b942a71675c812000ade30c16766f02ee388e8dc3e1c51dde5297f7d", + "git_blob": "110ed33a57902bef11873807e722fde3888b59fb", + "tag": "229", + "use": "Theorem T+ (superseded here for nu_cand), (C+.1), Remarks 1-2 (the matching pairs; the kink bound O(kappa k^2)), Lemma CE+ and the elder window (Remark 3)", + "head": "5dbe15cdc5f6d67b37dcefb176e9ad7433d6f77e", + "pr": "Math- #229" + }, + { + "path": "frontiers/third_order_coefficient_20261001/NOTE.md", + "bytes": 24186, + "sha256": "ae15c0ee173b9d01d258536a11780cc4b128c57811d1fb5db9db391ffc759ee5", + "git_blob": "aa078a9cbfc3aa859c089596892e34a8293831eb", + "tag": "216", + "use": "numerical values of c2; the formal elder law (0.2); the full-field Monte Carlo of section 3 (Remark 4)", + "head": "04c08e1534122f210be888742fcd7a568e45b097", + "pr": "Math- #216" + }, + { + "path": "frontiers/d1_third_order_law_20261001/PROOF.md", + "bytes": 44378, + "sha256": "279ed88ff2debd570e6426aec53cc93aabc7a298731cd9b25c45667a2c11077b", + "git_blob": "873532b955dbf7684edc1a53ba169c7216f778ab", + "tag": "214", + "use": "the d = 1 analogue: section 0 quote on the next terms O(h^(3/4)) (Remark 5)", + "head": "f4a58df9247dfa9ada2b756ecd7629c0b2faf5b5", + "pr": "Math- #214" + }, + { + "path": "frontiers/elder_third_order_20261001/PROOF.md", + "bytes": 70544, + "sha256": "74b5f031ba5be6f50b5e543636136df1d9b5fad0fcd2c8210320cfef02f80e37", + "git_blob": "c8767ddecf1041d234f685e95af582073a7be393", + "tag": "220", + "use": "the elder density: Lemma Q (Remark 3); not used", + "head": "70dcf310fc68f7d2b6949f035a09acf54f0cff7d", + "pr": "Math- #220" + } + ], + "cited_external": [ + { + "tag": "Haynsworth", + "reference": "E. V. Haynsworth, inertia additivity of the Schur complement (1968)", + "use": "through #218 Lemma C Step C1 and #191 Lemma E Step 4 (Lemma U, Step U2)" + }, + { + "tag": "Weyl", + "reference": "Weyl integration formula on symmetric matrices", + "use": "Lemmas D-prime, D-double-prime and Lambda, as in #218 Lemma D" + }, + { + "tag": "Anderson", + "reference": "T. W. Anderson, Proc. Amer. Math. Soc. 6 (1955) 170-176", + "use": "through #218 (1.2)" + }, + { + "tag": "Isserlis", + "reference": "Isserlis' (Wick's) theorem for Gaussian moments", + "use": "control Q4" + } + ], + "mutants": [ + "M1", + "M2", + "M3", + "M4", + "M5", + "M6", + "M7", + "M8" + ], + "replayed": [ + "parity_check.py" + ], + "referee": { + "kind": "two clean-context same-family referees (Anthropic Claude subagents) before submission, each followed by a delta check of the revision: A on sections 0-3, B on sections 4-8, the controls and the README", + "verdict": "first pass: A MAJOR REVISION (narrow; one major: the exceptional set of Lemma Lambda did not cover d >= 4, repaired by Lemma D-double-prime), B MAJOR REVISION (narrow; two majors: Step U2 Case 3 false as written, repaired by comparing with the layer integrand; Remark 2 rewritten); 17 minor findings and 13 nits applied. Delta: A ACCEPT (six nits, applied); B ACCEPT WITH MINOR FIXES (three minor items, including a one-line repair of Step U2 (b)(iii), and six nits; applied)", + "reports": [ + { + "report": "project archive V2_2/frontiers_candidate_parity_rate_20261001/REFEREE_A.md", + "sha256": "5d81d6282a6c436743e07dee6311984502e500db050acbe37be925cbcbe2a55e", + "bytes": 26270 + }, + { + "report": "project archive V2_2/frontiers_candidate_parity_rate_20261001/REFEREE_A_DELTA.md", + "sha256": "77493c38bd3217e540bc22c97a7c5d3fd679696f8fae0947d00d02e73b2d8624", + "bytes": 15896 + }, + { + "report": "project archive V2_2/frontiers_candidate_parity_rate_20261001/REFEREE_B.md", + "sha256": "f6c8687293d898b1d55fcce82c14c10d93f2548ac54e815126fef67f2aea53d5", + "bytes": 34949 + }, + { + "report": "project archive V2_2/frontiers_candidate_parity_rate_20261001/REFEREE_B_DELTA.md", + "sha256": "91043600090bc304da0d5b66bbe678544274a9d8d6826ed2d3f4de8102e0e80b", + "bytes": 25762 + } + ], + "independent_checks": "(2.1)-(2.2) symbolic for d = 1, 2, 3 and exact for d = 2-5; the typed criterion against true inertias on 80,000 random matrices (no mismatch); exact finite-r conditioning for Lemma K and Q4 (d = 1, 2); Lemma Lambda toys m = 1, 2, 3 (kappa (Lambda(k) - Lambda(0))/k^2 flat in kappa); P(D < eps) ~ eps^(2.5-3); a d = 2 Gaussian-kernel Monte Carlo with exact finite-r conditioning (T_r - S_r matches the layer term to 1-2% at 18 points; no O(r) term at fixed kappa: -0.007 +- 0.011 at kappa = 1); the cost of Step U2 (b)(iii) (kappa G/k in [0.0002, 0.0066]); control replay byte-identical on CPython 3.10-3.14 with and without -O; the mutant matrix", + "independence": "same provider and GitHub account as the author; zero organizational independence" + }, + "delivered_under": "explicit owner instructions of Dylan Roy in claude.ai session session_01NMeKEismAyeqgdB4sy2NJU, given after OWNER_STOP (2026-09-27): 2026-09-30 \"Continue to Work on closing as many mathematical items as possible while working with and supporting the other models in the GitHub. Also advance the mathematics in every place that's justified based on what we currently know\", followed by \"Continue\"; standing: \"Any request or permissions are preapproved in advance by me Dylan Roy\", pull requests only, never self-merge, no register/GRAPH/STATUS/lemma_closed/prize/premise edits, OWNER_STOP.md untouched. OWNER_STOP.md addresses Cursor agents and is untouched; nothing is merged by its author.", + "files": [ + { + "path": "PROOF.md", + "bytes": 50928, + "sha256": "aa4200100fcdf23135f82b7ac6f69621dedc65983d6707970f93d0343c6fca77" + }, + { + "path": "README.md", + "bytes": 9263, + "sha256": "e8e6c4bc50c858c5d29be390a52178f22945e456fa2b6a006ad6874298e7e1b1" + }, + { + "path": "parity_check.py", + "bytes": 26836, + "sha256": "fbea93a4810d21130a6dc95a593f36d09756368798f6b7ccf608520c68331470" + }, + { + "path": "RESULTS.json", + "bytes": 1433, + "sha256": "2af10a58aae284635de8973840594a71f18f1243415e1e8a3eb9d8a64ee669b1" + } + ] +} diff --git a/frontiers/candidate_parity_rate_20261001/parity_check.py b/frontiers/candidate_parity_rate_20261001/parity_check.py new file mode 100644 index 000000000..3e0e1a382 --- /dev/null +++ b/frontiers/candidate_parity_rate_20261001/parity_check.py @@ -0,0 +1,676 @@ +#!/usr/bin/env python3 +"""Exact controls for CL-CANDIDATE-PARITY-RATE-20261001-v1 (Theorem P: the candidate density with remainder l^(3/5)). + +Standard library only; exact rationals throughout; deterministic (a fixed linear congruential generator). +Output: RESULTS.json (sorted keys), byte-identical under -O. Usage: + python3 -B -S parity_check.py # exit 0, prints RESULTS.json + python3 -B -S parity_check.py --mutant M3 # exit 1 (only its control fails) + unknown mutant label -> exit 2 + +Controls + Q1 the exponent ledger of section 5 at rho = l^(1/5): least exponent 3/5, attained by rho^3, l^2 rho^-7, l rho^-2; + the split is optimal on a grid and is exactly the edge k = r^2 of Lemma U's range; the ledger without Lemma K + (finite part l rho^-2) also gives 3/5; the closed form of the integral of r min(k, 1/k) (antiderivatives + evaluated exactly) is (6/5) l^(2/3); 3/5 > 1/2 > 3/7 > 4/11 + Q2 Lemma R and the evenness of Lemma Pi on exactly pinned polynomial fields (d = 1, 2, 3, 4): in the + birth-integrated (midpoint) coordinates every pinned jet is an even polynomial in r; Pi(r) = + -det H(-r/2) det H(r/2) / r^2 is even in r; the new row of section 1 has target 0 (a consistency identity); + point reflection maps the pins at k to the pins at -k and preserves Pi; det K_M and det K_S are affine in k + (checked at d + 1 values of k, which determine a polynomial of degree <= d) + Q3 Lemma Pi, (2.2), on the same fields: det K_(M,S) = -/+ 6 k Delta + r U -/+ r^2 V + O(r^3) and + Pi = 36 k^2 Delta^2 + r^2 (12 k Delta V - U^2) + O(r^4), with U, V of (2.1) built from the general + Delta_BB = d^2/dt^2 det(A + t B) and A#_B = d/dt adj(A + t B) at t = 0 (so d = 4 tests the m >= 3 terms); + U is invariant and V changes sign under (k, odd jets) -> (-k, -odd jets) + Q4 Lemma K (the k-parity of A_2) on an instance: for the reference kernel exp(-|z|^2/2) in d = 1, 2, 3 the + A-integrand H(k, A) of the birth-integrated r^2 coefficient is h0 + h2 k^2 + h4 k^4 exactly (no odd powers), + a = 12, and the d = 1, 2 polynomials are those of Math- #232 section 4 + Q5 the pointwise algebra of Lemma U, Step U2, on rational grids (including the Case 3 identity) + Q6 the inequalities that collect Lemma U's error terms into C(r^2 + r min(k, 1/k)) for k >= r^2: (U.c) on a grid, + and the identities (U.a), (U.b), (U.d), recorded as consistency checks (they cannot fail) +""" +import json +import sys +from fractions import Fraction as Fr +from itertools import product + +MUTANTS = ('M1', 'M2', 'M3', 'M4', 'M5', 'M6', 'M7', 'M8') +MUT = None + + +# ----------------------------------------------------------------------------------------------- deterministic rationals +class LCG: + def __init__(self, seed): + self.s = seed % (2 ** 61 - 1) + + def nxt(self): + self.s = (self.s * 437799614237992725 + 1) % (2 ** 61 - 1) + return self.s + + def rat(self, num, den): + return Fr(self.nxt() % (2 * num + 1) - num, 1 + self.nxt() % den) + + +# ----------------------------------------------------------------------------------------------- polynomials in r +class RP: + __slots__ = ('c',) + + def __init__(self, c=None): + c = [Fr(x) for x in (c or [])] + while c and c[-1] == 0: + c.pop() + self.c = c + + def __add__(self, o): + o = o if isinstance(o, RP) else RP([o]) + n = max(len(self.c), len(o.c)) + return RP([(self.c[i] if i < len(self.c) else 0) + (o.c[i] if i < len(o.c) else 0) for i in range(n)]) + + __radd__ = __add__ + + def __neg__(self): + return RP([-x for x in self.c]) + + def __sub__(self, o): + return self + (-(o if isinstance(o, RP) else RP([o]))) + + def __mul__(self, o): + if not isinstance(o, RP): + return RP([x * o for x in self.c]) + if not self.c or not o.c: + return RP() + out = [Fr(0)] * (len(self.c) + len(o.c) - 1) + for i, a in enumerate(self.c): + if a: + for j, b in enumerate(o.c): + out[i + j] += a * b + return RP(out) + + __rmul__ = __mul__ + + def co(self, n): + return self.c[n] if n < len(self.c) else Fr(0) + + def div_r(self, n=1): + if any(self.co(i) != 0 for i in range(n)): + raise ArithmeticError('not divisible by r^%d' % n) + return RP(self.c[n:]) + + def even(self): + return all(self.co(i) == 0 for i in range(1, len(self.c), 2)) + + +def fact(n): + out = 1 + for i in range(2, n + 1): + out *= i + return out + + +def monomials(m, D, tmax=None): + """exponent tuples (a, b_1..b_m) of total degree <= D; with tmax, transverse degree b_1 + ... + b_m <= tmax + (second derivatives on the axis y = 0 see only transverse degree <= 2).""" + out = [] + for tot in range(D + 1): + for a in range(tot + 1): + for bs in product(range(tot - a + 1), repeat=m): + if sum(bs) == tot - a and (tmax is None or sum(bs) <= tmax): + out.append((a,) + bs) + return out + + +_POW = {} + + +def xpow(x, n): + """x^n for an RP x, memoized (x is one of the fixed evaluation points HALF, MHALF, SADDLE_M8).""" + key = (tuple(x.c), n) + v = _POW.get(key) + if v is None: + v = RP([1]) if n == 0 else xpow(x, n - 1) * x + _POW[key] = v + return v + + +def dat(field, alpha, x): + """partial^alpha f at (x, 0, ..., 0); x an RP in r.""" + m = len(alpha) - 1 + tot = RP() + for mono, coef in field.items(): + if mono[0] < alpha[0] or any(mono[j] != alpha[j] for j in range(1, m + 1)): + continue + k = fact(mono[0]) // fact(mono[0] - alpha[0]) + for j in range(1, m + 1): + k *= fact(mono[j]) + tot = tot + coef * k * xpow(x, mono[0] - alpha[0]) + return tot + + +HALF, MHALF = RP([0, Fr(1, 2)]), RP([0, Fr(-1, 2)]) +SADDLE_M8 = RP([0, Fr(1, 2), 0, 0, 1]) # r/2 + r^4: an asymmetric displacement of the saddle (mutant M8) + + +def pinned(m): + e = lambda j: tuple(1 if i == j else 0 for i in range(m)) + return [(0,) + (0,) * m, (1,) + (0,) * m, (2,) + (0,) * m, (3,) + (0,) * m] + [(0,) + e(j) for j in range(m)] + [(1,) + e(j) for j in range(m)] + + +def row(field, m, i): + """the i-th centred pin row of [R] section 2 (0: midpoint height, 1: (f(+) - f(-))/r, 2: (f_x(+) - f_x(-))/r, + 3: T_r, then the m transverse averages and the m transverse differences), as an RP in r.""" + z = (0,) * m + f = lambda al, x: dat(field, al, x) + if i == 0: + return (f((0,) + z, MHALF) + f((0,) + z, HALF)) * Fr(1, 2) + if i in (1, 3): + R2 = (f((0,) + z, HALF) - f((0,) + z, MHALF)).div_r(1) + return R2 if i == 1 else ((f((1,) + z, MHALF) + f((1,) + z, HALF) - R2 * 2) * 6).div_r(2) + if i == 2: + return (f((1,) + z, HALF) - f((1,) + z, MHALF)).div_r(1) + j = i - 4 + ej = tuple(1 if t == j % m else 0 for t in range(m)) + if j < m: + return (f((0,) + ej, MHALF) + f((0,) + ej, HALF)) * Fr(1, 2) + return (f((0,) + ej, HALF) - f((0,) + ej, MHALF)).div_r(1) + + +def rows(field, m): + return [row(field, m, i) for i in range(4 + 2 * m)] + + +def pin_field(free, m, bp, k): + """Solve the pinned monomials (as polynomials in r) for target (bp, -k r^2, 0, 12 k, 0...) of [R] section 2, + written in the midpoint height bp; mutant M2 writes the height of M (bp + k r^3/2) in place of the midpoint.""" + pins = pinned(m) + field = {mo: RP([v]) for mo, v in free.items()} + for mo in pins: + field[mo] = RP() + h0 = RP([bp]) if MUT != 'M2' else RP([bp, 0, 0, k / 2]) + tgt = [h0, RP([0, 0, -k]), RP(), RP([12 * k])] + [RP()] * (2 * m) + order = [2, 0, 3, 1] + [4 + j for j in range(m)] + [4 + m + j for j in range(m)] + for _ in range(3): + for ri in order: + mo = pins[ri] + field[mo] = RP() + rest = row(field, m, ri) + field[mo] = RP([1]) + unit = row(field, m, ri) - rest + if len(unit.c) != 1 or unit.c[0] == 0: + raise ArithmeticError('row %d not unit-triangular' % ri) + field[mo] = (tgt[ri] - rest) * (1 / unit.c[0]) + chk = rows(field, m) + if any((chk[i] - tgt[i]).c for i in range(len(chk))): + raise ArithmeticError('pins not satisfied') + return field + + +def rdet(M): + n = len(M) + if n == 1: + return M[0][0] + s = RP() + for j in range(n): + minor = [[M[i][jj] for jj in range(n) if jj != j] for i in range(1, n)] + t = M[0][j] * rdet(minor) + s = s + (t if j % 2 == 0 else -t) + return s + + +def hess(field, m, x): + d = m + 1 + H = [[None] * d for _ in range(d)] + for i in range(d): + for j in range(d): + al = [0] * d + al[i] += 1 + al[j] += 1 + H[i][j] = dat(field, tuple(al), x) + return H + + +def Pi_of(field, m): + KM = rdet(hess(field, m, MHALF)).div_r(1) + KS = rdet(hess(field, m, HALF if MUT != 'M8' else SADDLE_M8)).div_r(1) + return -(KM * KS), KM, KS + + +# ----------------------------------------------------------------------------------------------- matrices over Q +def mdet(M): + n = len(M) + if n == 0: + return Fr(1) + if n == 1: + return M[0][0] + s = Fr(0) + for j in range(n): + minor = [[M[i][jj] for jj in range(n) if jj != j] for i in range(1, n)] + s += (-1) ** j * M[0][j] * mdet(minor) + return s + + +def madj(M): + n = len(M) + if n == 1: + return [[Fr(1)]] + return [[(-1) ** (i + j) * mdet([[M[a][b] for b in range(n) if b != i] for a in range(n) if a != j]) for j in range(n)] for i in range(n)] + + +def quad(u, M, v): + return sum((u[i] * M[i][j] * v[j] for i in range(len(u)) for j in range(len(v))), Fr(0)) + + +def jets_of(field, m): + g = lambda mono: field[mono].co(0) if mono in field else Fr(0) + e = lambda j: tuple(1 if i == j else 0 for i in range(m)) + + def sym(xp, md, mo): + return [[(md * g((xp,) + tuple(2 if t == i else 0 for t in range(m))) if i == j else + mo * g((xp,) + tuple(1 if t in (i, j) else 0 for t in range(m)))) for j in range(m)] for i in range(m)] + return (sym(0, 2, 1), sym(1, 2, 1), sym(2, 4, 2), [2 * g((2,) + e(j)) for j in range(m)], + [6 * g((3,) + e(j)) for j in range(m)], 24 * g((4,) + (0,) * m), 120 * g((5,) + (0,) * m)) + + +def det_adj_t(A, B): + """det(A + tB) and adj(A + tB) as polynomials in t (the RP class, with t in place of r).""" + m = len(A) + M = [[RP([A[i][j], B[i][j]]) for j in range(m)] for i in range(m)] + if m == 1: + return M[0][0], [[RP([1])]] + adj = [[rdet([[M[a][b] for b in range(m) if b != i] for a in range(m) if a != j]) * (-1) ** (i + j) + for j in range(m)] for i in range(m)] + return rdet(M), adj + + +def UV(A, B, C, gam, eta, f4, f5, k): + """U, V of (2.1); Delta_B = D det(A)[B], Delta_BB = D^2 det(A)[B, B], A#_B = D adj(A)[B] in general form.""" + m = len(A) + if m == 0: + D, J, DB, DC, DBB, JB = Fr(1), [], Fr(0), Fr(0), Fr(0), [] + else: + dt, at = det_adj_t(A, B) + D, DB, DBB = dt.co(0), dt.co(1), 2 * dt.co(2) + J = [[x.co(0) for x in row] for row in at] + JB = [[x.co(1) for x in row] for row in at] + DC = sum((J[i][j] * C[j][i] for i in range(m) for j in range(m)), Fr(0)) + U = f4 / 12 * D - quad(gam, J, gam) / 4 + (3 * k * DB if MUT != 'M3' else 0) + V = (Fr(3, 4) * k * (DC + DBB) + f4 / 24 * DB + (f5 / 120 * D if MUT != 'M4' else 0) - quad(gam, J, eta) / 12 + - (quad(gam, JB, gam) / 8 if m >= 2 else 0)) + return D, U, V + + +def control_Q2_Q3(): + q2 = True + q3 = True + n = 0 + by_d = {} + rng = LCG(20261001) + for m, deg, trials, tmax in ((0, 9, 4, None), (1, 9, 4, None), (2, 8, 2, None), (2, 8, 2, None), (3, 9, 2, 2)): + pins = set(pinned(m)) + for _ in range(trials): + free = {mo: rng.rat(9, 5) for mo in monomials(m, deg, tmax) if mo not in pins} + k = rng.rat(6, 4) + if k == 0: + k = Fr(1, 3) + bp = rng.rat(5, 3) + fld = pin_field(free, m, bp, k) + q2 &= all(fld[mo].even() for mo in pinned(m)) + Pi, KM, KS = Pi_of(fld, m) + q2 &= Pi.even() + # the new row of section 1: (f_x(-) + f_x(+))/2 = R2 + r^2 R4 / 12 has target -k r^2 + k r^2 = 0 + z = (0,) * m + avg = (dat(fld, (1,) + z, MHALF) + dat(fld, (1,) + z, HALF)) * Fr(1, 2) + R = rows(fld, m) + q2 &= not (avg - (R[1] + RP([0, 0, Fr(1, 12)]) * R[3])).c and not avg.c + # point reflection f(-z): coefficients times (-1)^{|alpha|}; it satisfies the pins with -k, same Pi + refl = {mo: c * (-1) ** sum(mo) for mo, c in fld.items()} + Rr = rows(refl, m) + tgt = [RP([bp]), RP([0, 0, k]), RP(), RP([-12 * k])] + [RP()] * (2 * m) + q2 &= all(not (Rr[i] - tgt[i]).c for i in range(len(Rr))) + Pr, _, _ = Pi_of(refl, m) + q2 &= not (Pr - Pi).c + # det K_M and det K_S affine in k: both are polynomials of degree <= d in k (the pinned coefficients are + # affine in k), so d + 1 values t k, t = 1..d+1, decide; all second differences must vanish + KK = [Pi_of(pin_field(free, m, bp, t * k), m) for t in range(1, m + 3)] + for idx in (1, 2): + vals = [x[idx] for x in KK] + q2 &= all(not (vals[i] - vals[i + 1] * 2 + vals[i + 2]).c for i in range(len(vals) - 2)) + # Q3: the coefficients of Pi and the parity of U, V + A, B, C, gam, eta, f4, f5 = jets_of(fld, m) + D, U, V = UV(A, B, C, gam, eta, f4, f5, k) + q3 &= Pi.co(0) == 36 * k * k * D * D and KM.co(0) == -6 * k * D and KS.co(0) == 6 * k * D + q3 &= Pi.co(2) == 12 * k * D * V - U * U + q3 &= KM.co(1) == U and KS.co(1) == U and KM.co(2) == -V and KS.co(2) == V + neg = lambda M: [[-x for x in row] for row in M] + D2, U2, V2 = UV(A, neg(B), C, [-x for x in gam], eta, f4, -f5, -k) + q3 &= U2 == U and V2 == -V and D2 == D + n += 1 + by_d[str(m + 1)] = by_d.get(str(m + 1), 0) + 1 + return q2, q3, n, by_d + + +# ----------------------------------------------------------------------------------------------- Q4: reference kernel +class MP: + def __init__(self, n, t=None): + self.n = n + self.t = {e: Fr(c) for e, c in (t or {}).items() if c != 0} + + @staticmethod + def var(n, i): + e = [0] * n + e[i] = 1 + return MP(n, {tuple(e): 1}) + + @staticmethod + def cst(n, c): + return MP(n, {(0,) * n: c}) + + def __add__(self, o): + o = o if isinstance(o, MP) else MP.cst(self.n, o) + out = dict(self.t) + for e, c in o.t.items(): + out[e] = out.get(e, 0) + c + return MP(self.n, out) + + __radd__ = __add__ + + def __neg__(self): + return MP(self.n, {e: -c for e, c in self.t.items()}) + + def __sub__(self, o): + return self + (-(o if isinstance(o, MP) else MP.cst(self.n, o))) + + def __mul__(self, o): + if not isinstance(o, MP): + return MP(self.n, {e: c * o for e, c in self.t.items()}) + out = {} + for e1, c1 in self.t.items(): + for e2, c2 in o.t.items(): + e = tuple(a + b for a, b in zip(e1, e2)) + out[e] = out.get(e, 0) + c1 * c2 + return MP(self.n, out) + + __rmul__ = __mul__ + + +def He0(n): + if n % 2: + return 0 + v = 1 + for j in range(1, n, 2): + v *= j + return (-1) ** (n // 2) * v + + +def kcov(a, b): + s = (-1) ** sum(a) + for x, y in zip(a, b): + s *= He0(x + y) + return Fr(s) + + +def inv(M): + n = len(M) + A = [[Fr(x) for x in row] + [Fr(int(i == j)) for j in range(n)] for i, row in enumerate(M)] + for c in range(n): + p = next(i for i in range(c, n) if A[i][c] != 0) + A[c], A[p] = A[p], A[c] + pv = A[c][c] + A[c] = [x / pv for x in A[c]] + for i in range(n): + if i != c and A[i][c] != 0: + f = A[i][c] + A[i] = [x - f * y for x, y in zip(A[i], A[c])] + return [row[n:] for row in A] + + +def isserlis(poly, ns, S): + memo = {} + + def mom(ex): + if ex in memo: + return memo[ex] + idx = [i for i, p in enumerate(ex) for _ in range(p)] + if len(idx) % 2: + memo[ex] = Fr(0) + return Fr(0) + + def pr(lst): + if not lst: + return Fr(1) + tot = Fr(0) + for j in range(1, len(lst)): + c = S[lst[0]][lst[j]] + if c: + tot += c * pr(lst[1:j] + lst[j + 1:]) + return tot + memo[ex] = pr(idx) + return memo[ex] + out = {} + for e, c in poly.t.items(): + v = mom(e[ns:]) + if v: + out[e[:ns]] = out.get(e[:ns], 0) + c * v + return MP(ns, out) + + +def ref_H(d): + """H(k, A) of the birth-integrated r^2 coefficient for exp(-|z|^2/2), as a polynomial in (k, A_upper).""" + m = d - 1 + + def mi(xo, ys): + al = [0] * d + al[0] = xo + for j in ys: + al[1 + j] += 1 + return tuple(al) + G = [mi(1, [])] + [mi(0, [j]) for j in range(m)] + T3 = mi(3, []) + Hu = [mi(2, [])] + [mi(1, [j]) for j in range(m)] + Aidx = [(i, j) for i in range(m) for j in range(i, m)] + Aj = [mi(0, [i, j]) for (i, j) in Aidx] + O = G + [T3] + E0 = Hu + Aj + C = O + E0 + F = [mi(4, []), mi(5, [])] + [mi(1, [i, j]) for (i, j) in Aidx] + [mi(2, [i, j]) for (i, j) in Aidx] \ + + [mi(2, [j]) for j in range(m)] + [mi(3, [j]) for j in range(m)] + SCi = inv([[kcov(a, b) for b in C] for a in C]) + SFC = [[kcov(a, b) for b in C] for a in F] + K = [[sum((SFC[i][t] * SCi[t][j] for t in range(len(C))), Fr(0)) for j in range(len(C))] for i in range(len(F))] + Sc = [[kcov(F[i], F[j]) - sum((K[i][t] * SFC[j][t] for t in range(len(C))), Fr(0)) for j in range(len(F))] for i in range(len(F))] + ns = 1 + len(Aidx) + nv = ns + len(F) + kv = MP.var(nv, 0) + Av = [MP.var(nv, 1 + i) for i in range(len(Aidx))] + t3 = 12 * kv + (1 if MUT == 'M5' else 0) + cv = [MP.cst(nv, 0)] * len(G) + [t3] + [MP.cst(nv, 0)] * len(Hu) + Av + jet = [] + for i in range(len(F)): + mu = MP.cst(nv, 0) + for t in range(len(C)): + if K[i][t]: + mu = mu + cv[t] * K[i][t] + jet.append(mu + MP.var(nv, ns + i)) + f4, f5 = jet[0], jet[1] + nA = len(Aidx) + Bv, Cv = jet[2:2 + nA], jet[2 + nA:2 + 2 * nA] + gm, et = jet[2 + 2 * nA:2 + 2 * nA + m], jet[2 + 2 * nA + m:] + + def sy(vals): + M = [[None] * m for _ in range(m)] + for t, (i, j) in enumerate(Aidx): + M[i][j] = vals[t] + M[j][i] = vals[t] + return M + Am, Bm, Cm = sy(Av), sy(Bv), sy(Cv) + one, zero = MP.cst(nv, 1), MP.cst(nv, 0) + if m == 0: + D, DB, DC, DBB, J, JB = one, zero, zero, zero, [], [] + elif m == 1: + D, DB, DC, DBB, J, JB = Am[0][0], Bm[0][0], Cm[0][0], zero, [[one]], [[zero]] + else: + D = Am[0][0] * Am[1][1] - Am[0][1] * Am[0][1] + J = [[Am[1][1], -Am[0][1]], [-Am[0][1], Am[0][0]]] + DB = Am[1][1] * Bm[0][0] + Am[0][0] * Bm[1][1] - 2 * Am[0][1] * Bm[0][1] + DC = Am[1][1] * Cm[0][0] + Am[0][0] * Cm[1][1] - 2 * Am[0][1] * Cm[0][1] + DBB = 2 * (Bm[0][0] * Bm[1][1] - Bm[0][1] * Bm[0][1]) + JB = [[Bm[1][1], -Bm[0][1]], [-Bm[0][1], Bm[0][0]]] + + def qd(u, M, v): + s = zero + for i in range(len(u)): + for j in range(len(v)): + s = s + u[i] * M[i][j] * v[j] + return s + U = f4 * Fr(1, 12) * D - qd(gm, J, gm) * Fr(1, 4) + 3 * kv * DB + V = Fr(3, 4) * kv * (DC + DBB) + f4 * Fr(1, 24) * DB + f5 * Fr(1, 120) * D - qd(gm, J, et) * Fr(1, 12) - qd(gm, JB, gm) * Fr(1, 8) + EW2 = isserlis(12 * kv * D * V - U * U, ns, Sc) + + def shift(al): + return (al[0] + 2,) + al[1:] + dO = [(shift(a), Fr(1, 8)) for a in G] + [(shift(T3), Fr(1, 40))] + dE = [(shift(a), Fr(1, 24)) for a in Hu] + [(None, 0) for _ in Aj] + + def S2(base, dl): + n = len(base) + return [[(dl[i][1] * kcov(dl[i][0], base[j]) if dl[i][0] else 0) + (dl[j][1] * kcov(base[i], dl[j][0]) if dl[j][0] else 0) + for j in range(n)] for i in range(n)] + Si = inv([[kcov(a, b) for b in O] for a in O]) + Ti = inv([[kcov(a, b) for b in E0] for a in E0]) + S2m, T2m = S2(O, dO), S2(E0, dE) + + def mm(X, Y): + return [[sum((X[i][t] * Y[t][j] for t in range(len(Y))), Fr(0)) for j in range(len(Y[0]))] for i in range(len(X))] + SS, TT = mm(Si, S2m), mm(Ti, T2m) + SSS, TTT = mm(SS, Si), mm(TT, Ti) + nO = len(O) + score0 = -sum((SS[i][i] for i in range(nO)), Fr(0)) / 2 - sum((TT[i][i] for i in range(len(E0))), Fr(0)) / 2 + k2 = 72 * SSS[nO - 1][nO - 1] + kvs = MP.var(ns, 0) + Avs = [MP.var(ns, 1 + i) for i in range(len(Aidx))] + xA = [MP.cst(ns, 0)] * len(Hu) + Avs + sq = MP.cst(ns, 0) + for i in range(len(E0)): + for j in range(len(E0)): + if TTT[i][j]: + sq = sq + xA[i] * xA[j] * TTT[i][j] + score = MP.cst(ns, score0) + kvs * kvs * k2 + sq * Fr(1, 2) + Ds = MP.cst(ns, 1) if m == 0 else (Avs[0] if m == 1 else Avs[0] * Avs[2] - Avs[1] * Avs[1]) + return EW2 + 36 * kvs * kvs * Ds * Ds * score, 72 * Si[nO - 1][nO - 1] + + +def control_Q4(): + ok = True + polys = {} + for d in (1, 2, 3): + H, a = ref_H(d) + powers = sorted({e[0] for e in H.t}) + ok &= set(powers) <= {0, 2, 4} and a == 12 + polys[d] = {str(p): sorted([[list(e[1:]), str(c)] for e, c in H.t.items() if e[0] == p]) for p in (0, 2, 4)} + # the d = 1 and d = 2 polynomials (Math- #232 (4.1) and the planar recovery; derived here independently) + ok &= polys[1] == {'0': [[[], '-5/24']], '2': [[[], '9/2']], '4': [[[], '108']]} + ok &= polys[2]['0'] == sorted([[[0], '-3/4'], [[2], '-13/96'], [[4], '-1/256']]) + ok &= polys[2]['2'] == sorted([[[0], '-18'], [[2], '57/8'], [[4], '-9/64']]) + ok &= polys[2]['4'] == [[[2], '108']] + return ok, {'d1': polys[1], 'd2': polys[2]} + + +# ----------------------------------------------------------------------------------------------- Q1, Q5, Q6 +def control_Q1(): + s = Fr(1, 5) if MUT != 'M1' else Fr(1, 6) + terms = {'rho^3 (Lemma U, r^2 part; J2)': 3 * s, 'l^(2/3) (Lemma U, r min(k,1/k) part)': Fr(2, 3), + 'l^2 rho^-5 (A_2 finite part)': 2 - 5 * s, 'l^2 rho^-7 (cusp tail; J3; J4)': 2 - 7 * s, + 'l rho^-2 (J3)': 1 - 2 * s, 'l (J5)': Fr(1)} + least = min(terms.values()) + attained = sorted(t for t, e in terms.items() if e == least) + + def worst(sig): + return min(3 * sig, Fr(2, 3), 2 - 5 * sig, 2 - 7 * sig, 1 - 2 * sig, Fr(1)) + grid = [Fr(i, 1200) for i in range(150, 301)] + best = max(grid, key=worst) + ok = least == Fr(3, 5) and best == Fr(1, 5) and worst(Fr(1, 5)) == Fr(3, 5) + # Lemma U is applied for r <= rho with k = l / r^3 >= r^2, i.e. r <= l^(1/5): the split is the boundary + ok &= s >= Fr(1, 5) + # without Lemma K (only the C^1 bound A_2(k) - A_2(0) = O(k)) the finite part is l rho^-2: still 3/5 + no_k = dict(terms) + no_k['l^2 rho^-5 (A_2 finite part)'] = 1 - 2 * s + least_no_k = min(no_k.values()) + ok &= least_no_k == Fr(3, 5) + # closed form of int_0^inf r min(k, 1/k) dr, k = l/r^3, l = q^3 (crossover r = q): + # antiderivatives r^5/(5 q^3) on [0, q] and -q^3/r on [q, inf), evaluated exactly + for q in (Fr(1, 10), Fr(7, 3), Fr(5)): + ok &= (q ** 5 / (5 * q ** 3) - 0) + (0 - (-(q ** 3) / q)) == Fr(6, 5) * q ** 2 + ok &= Fr(3, 5) > Fr(1, 2) > Fr(3, 7) > Fr(4, 11) + return ok, {'split': str(s), 'terms': {t: str(e) for t, e in terms.items()}, 'least': str(least), 'attained_by': attained, + 'optimal_split': str(best), 'least_without_lemma_K': str(least_no_k)} + + +def control_Q5(): + ok = True + grid = [Fr(i, 3) for i in range(-7, 8)] + for y, c in product(grid, grid): + if c >= 0: + ok &= y * y - min(y * y, c * c) == max(y * y - c * c, 0) + for x, e in product(grid, grid): + lhs = abs(max(x + e, 0) - max(x, 0) - (e if x > 0 else 0)) + ok &= lhs <= (abs(e) if abs(x) <= abs(e) else 0) + nz = [g for g in grid if g != 0] + for mM, mS, s in product(nz, nz, (1, -1)): + typed = s * mM < 0 < s * mS + anti = s * mM > 0 > s * mS + lhs = -mM * mS * ((1 if typed else 0) - 1) + rhs = max(mM * mS, 0) - (abs(mM * mS) if (anti and MUT != 'M6') else 0) + ok &= lhs == rhs + # Case 3 on {A < 0}: with P := r^-2 Pi and eps := -P - X, P (1_typed - 1) - X_+ = P 1_typed - (-X)_+ + eps + for P, X, t in product(grid, grid, (0, 1)): + eps = -P - X + ok &= P * (t - 1) - max(X, 0) == P * t - max(-X, 0) + eps + return ok + + +def control_Q6(): + ok = True + rs = [Fr(1, 2 ** i) for i in range(1, 9)] + [Fr(3, 5), Fr(1)] + ks = [Fr(1, 10 ** 4), Fr(1, 300), Fr(1, 17), Fr(1, 3), Fr(1), Fr(5, 2), Fr(9), Fr(40)] + for r, k in product(rs, ks): + if k < r * r: + continue + kap = k / r + mk = min(k, 1 / k) + ok &= k * k / (1 + kap) ** 2 <= r * r # (U.a): identity, k/(1 + kap) = kr/(k + r) <= r + ok &= r ** 3 / kap <= r * r # (U.b): identity, equivalent to k >= r^2 + lhs = (r * r + min(Fr(1), k * k)) / (1 + kap) + ok &= lhs <= 2 * (r * r + r * mk) # (U.c) + if MUT == 'M7': + ok &= k * k / (1 + kap) <= r * r + if k <= 1: + ok &= k * r <= r * mk # (U.d): identity, kr = r min(k, 1/k) + return ok + + +def main(argv): + global MUT + if len(argv) == 2 and argv[0] == '--mutant': + if argv[1] not in MUTANTS: + print('unknown mutant label', file=sys.stderr) + return 2 + MUT = argv[1] + elif argv: + print('usage: parity_check.py [--mutant M1..M8]', file=sys.stderr) + return 2 + q1, ledger = control_Q1() + q2, q3, nfields, by_d = control_Q2_Q3() + q4, polys = control_Q4() + q5 = control_Q5() + q6 = control_Q6() + res = {'object': 'CL-CANDIDATE-PARITY-RATE-20261001-v1', 'scientific_effect': 'NONE', + 'controls': {'Q1_ledger': q1, 'Q2_reflection_evenness': q2, 'Q3_second_coefficient': q3, + 'Q4_k_parity_reference': q4, 'Q5_layer_algebra': q5, 'Q6_bookkeeping': q6}, + 'ledger': ledger, 'pinned_fields_checked': nfields, 'pinned_fields_by_d': by_d, + 'reference_H_polynomials': polys} + ok = all(res['controls'].values()) + if MUT is None: + print(json.dumps(res, indent=1, sort_keys=True)) + return 0 if ok else 1 + + +if __name__ == '__main__': + sys.exit(main(sys.argv[1:])) From 0add03bb806b6237e6aa0b29b06bedec54863e79 Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 1 Oct 2026 22:31:05 +0000 Subject: [PATCH 2/2] Theorem P v1.1: apply the six C55 nonblocking wording corrections CL-CANDIDATE-PARITY-RATE-20261001-v1.1 (wording only; no estimate or conclusion changes), after the C55 nonauthor reviews on a6b6788 (OpenAI Codex: Slice A 5385194600, Slice B 5385169112, Slice C 5385136758 + addendum 5940107723, all PASS_TECHNICAL): 1. Lemma Pi: det K_i is affine in k at fixed free jets, not along the laws Q-bar_{r,k} (statement, proof, Q2 rows, checker docstring). 2. Moment bounds before Lemma Pi and Lemma Pi: signed k with 1 + |k|. 3. Lemma Lambda: phi'' = 2*1{|y|>theta} + 2 theta (delta_-theta + delta_theta). 4. Lemma Lambda kink: rho is the unnormalized Weyl-coordinate slice, integrated over d lambda' dO dP_g. 5. The case m = 1 of that step (adj A = 1). 6. Step U2 (b)(iii): two one-sided threshold-crossing strips of length |a|, a = 3k Delta_B, centred at +-6 kappa |Delta| - a/2 (total 72k|Delta_B|/|Delta|). RESULTS.json byte-identical (CPython 3.10-3.14, with and without -O); mutants M1-M8 unchanged. README and SOURCES.json record the reviews. Scientific effect: NONE. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01NMeKEismAyeqgdB4sy2NJU --- .../candidate_parity_rate_20261001/PROOF.md | 70 ++++++++++++------- .../candidate_parity_rate_20261001/README.md | 27 ++++++- .../SOURCES.json | 40 ++++++++--- .../parity_check.py | 9 +-- 4 files changed, 108 insertions(+), 38 deletions(-) diff --git a/frontiers/candidate_parity_rate_20261001/PROOF.md b/frontiers/candidate_parity_rate_20261001/PROOF.md index 97298f32f..a97bf528b 100644 --- a/frontiers/candidate_parity_rate_20261001/PROOF.md +++ b/frontiers/candidate_parity_rate_20261001/PROOF.md @@ -1,7 +1,15 @@ # The candidate density with remainder `ℓ^{3/5}`: a reflection parity removes the `ℓ^{1/2}` term -Object: CL-CANDIDATE-PARITY-RATE-20261001-v1. +Object: CL-CANDIDATE-PARITY-RATE-20261001-v1.1. Author: Anthropic Claude (claude.ai session `session_01NMeKEismAyeqgdB4sy2NJU`), 1 October 2026. +**v1.1 (wording only; no estimate or conclusion changes):** the six nonblocking corrections of the C55 nonauthor reviews +(OpenAI Codex, on head `a6b6788`: Slice A 5385194600, Slice B 5385169112, Slice C 5385136758 with addendum +5940107723) are applied. (1) In Lemma Π, `det K_i` is affine in `k` *at fixed free jets*, not along the laws `Q̄_{r,k}`, +under which the odd free jets have `k`-dependent means. (2) The moment bounds before Lemma Π, and Lemma Π, are stated +for signed `k` with `1 + |k|`. (3) In Lemma Λ, `φ″` is given as a distribution. (4) In Lemma Λ's kink step, `ρ` is the +unnormalized Weyl-coordinate slice, and the measure is stated. (5) The case `m = 1` of that step is spelled out. +(6) In Step U2 (b)(iii), the `f₄`-windows are the two one-sided threshold-crossing strips, centred at the shifted +points. Disposition: AUTHOR-SIDE PROOF CANDIDATE; NONAUTHOR REVIEW REQUIRED. Scientific effect: NONE — no register, graph, STATUS, PROOF_INDEX, prize or Boolean change; no numerical constant is certified. Same GitHub account as every lane; zero organizational independence. @@ -144,17 +152,21 @@ With the jets of §0 put `Δ_B := D det(A)[B] = tr(A^♯B)`, `Δ_C := tr(A^♯C) V := (3k/4)(Δ_C + Δ_BB) + (f₄/24)Δ_B + (f₅/120)Δ − (1/12)γᵀA^♯η − (1/8)γᵀA^♯_Bγ. (2.1) Write `V = V_o + kV_e` with `V_e := (3/4)(Δ_C + Δ_BB)`; `V_e` is even and `V_o` is odd under `𝒫`. Let `T` be #218 §1's -dominating variable. Under `Q̄_{r,k}`, `E[T^p] ≤ C_p(1 + k)^p`: mix #218's bound `C_p(1 + |b′| + k)^p` over the conditional -law of the midpoint height `b′ = f_e(x_S)`, which is centred Gaussian and does not depend on `k` (Lemma R (R.2)). Likewise, -for a finite jet vector `J₀` with uniformly invertible covariance, (1.2) gives `E[T^p | J₀] ≤ C_p(1 + k + |J₀|)^p`. +dominating variable. Under `Q̄_{r,k}`, for `k ∈ R`, `E[T^p] ≤ C_p(1 + |k|)^p`: mix #218's bound `C_p(1 + |b′| + |k|)^p` +over the conditional law of the midpoint height `b′ = f_e(x_S)`, which is centred Gaussian and does not depend on `k` +(Lemma R (R.2)). #218 states its bound for `k ≥ 0`; the Gaussian regression behind it is linear in the targets, so for +signed `k` it holds with `|k|`. Likewise, for a finite jet vector `J₀` with uniformly invertible covariance, (1.2) gives +`E[T^p | J₀] ≤ C_p(1 + |k| + |J₀|)^p`. For `k ≥ 0` these read `1 + k`, as written where only `k ≥ 0` occurs. **Lemma Π.** Pathwise, - det K_M = −6kΔ + rU − r²V + O(r³(1 + k)T^N), det K_S = 6kΔ + rU + r²V + O(r³(1 + k)T^N), - Π = 36k²Δ² + r²(12kΔV − U²) + O(r⁴(1 + k)^N T^N). (2.2) + det K_M = −6kΔ + rU − r²V + O(r³(1 + |k|)T^N), det K_S = 6kΔ + rU + r²V + O(r³(1 + |k|)T^N), + Π = 36k²Δ² + r²(12kΔV − U²) + O(r⁴(1 + |k|)^N T^N). (2.2) -Under `(k, jets) ↦ (−k, 𝒫jets)`, `Δ`, `Y′` and `U` are invariant and `V` changes sign; `det K_M` and `det K_S` are affine in -`k`. +Under `(k, jets) ↦ (−k, 𝒫jets)`, `Δ`, `Y′` and `U` are invariant and `V` changes sign. At fixed free jets, `det K_M` and +`det K_S` are affine in `k`. This is a statement about the pinned polynomial fields of the proof, not about the laws +`Q̄_{r,k}` at different `k`: under those, the odd free jets have means `k·w` (Lemma R (R.2)), and these mean shifts are +kept wherever they matter (Step U2 (b)(ii)). *Proof.* By #218 Step F1, `Π` equals a polynomial `Φ(J′; r, k)` in the free jets `J′` up to `O(r⁴T^{n₀})`, and the coefficients of `Φ` are identities of polynomials, which may be read off on exactly pinned polynomial fields. On those @@ -163,10 +175,11 @@ with [R]'s row at fixed `b`, `f(0)` would carry a term `−kr³/2`, which does n and `Φ` has no odd powers of `r`. The pins give `f_xx(0) = −r²f₄/24 + O(r⁴)`, `f_xxx(0) = 12k − r²f₅/40 + O(r⁴)` and `∇_yf_x(0) = −r²η/24 + O(r⁴)`. Expanding `α_i = f_xx(x_i)/r`, `β_i = ∇_yf_x(x_i)/r` and `A_i = ∇_Θ²f(x_i)` and using `det K_i = α_i det A_i − rβ_iᵀadj(A_i)β_i` gives the first two lines; their product gives (2.2), the `r`-odd terms -cancelling. Only `α_i` depends on `k`, affinely, so `det K_i` is affine in `k`. The parity statement is read off (2.1): +cancelling. At fixed free jets only `α_i` depends on `k`, affinely, so there `det K_i` is affine in `k`. The parity +statement is read off (2.1): `B`, `γ`, `f₅` are odd and `A`, `C`, `η`, `f₄` even, so `Δ_B` and `A^♯_B` are odd and `Δ_BB` (quadratic in `B`) is even. Control Q3 checks (2.2) exactly on fourteen pinned polynomial fields in `d = 1, 2, 3, 4`, with general free jets; control -Q2 checks the evenness, the reflection and the affinity in `k`. ∎ +Q2 checks the evenness, the reflection and the affinity in `k` at fixed free jets. ∎ (2.2) is #232's (2.3) (`W₂ = 12kΔV − U²`, with remainder `O(r³)` there), obtained independently here; the absence of the `r³` term is the pathwise evenness of Lemma R (R.4). @@ -255,8 +268,8 @@ Then `U_s(ζ₀ + sw) = Y₀ + sY₁ + s²Y₂`, with `Y₀ := Y′(ζ₀)`, `Y *Integrate `f₄` first.* Given the other jets, `f₄` is Gaussian with mean `m₄`, linear in the other even jets, and a variance `σ₄²` bounded above and below (a Schur complement; #218 Step F1). Write `Y₀ = (Δ/12)f₄ + β` with `β := −γ₀ᵀA^♯γ₀/4`. Then `Φ(t) := E_{f₄}[φ(Y₀ + t)]` is `E[φ(Z + t)]` for `Z ~ N(μ_Z, σ_Z²)`, where `μ_Z := (Δ/12)m₄ + β` -and `σ_Z := |Δ|σ₄/12`. `φ` is convex, `φ″ = 2` off `{|y| = θ}`, and `φ′` jumps by `2θ` at `±θ`. Hence `Φ` is convex and -smooth, with +and `σ_Z := |Δ|σ₄/12`. `φ` is convex, and as a distribution `φ″ = 2·1{|y| > θ} + 2θ(δ_{−θ} + δ_θ)`: `φ″ = 0` on +`|y| < θ`, `φ″ = 2` on `|y| > θ`, and `φ′` jumps by `2θ` at `±θ`. Hence `Φ` is convex and smooth, with Φ″(t) = 2P(|Z + t| > θ) + 2θ[p_Z(θ − t) + p_Z(−θ − t)], θ p_Z(x) = (72κ/(√(2π)σ₄)) e^{−(x − μ_Z)²/(2σ_Z²)}. (3.1) @@ -275,10 +288,14 @@ as `E[Φ(t₀) − Φ(0)] + E[½(Φ(t₀ + h) + Φ(t₀ − h)) − Φ(t₀)]`, - *Second part, kink.* By (3.1) the second term contributes `K := (72κ/(√(2π)σ₄))∫(|h| − |τ|)₊ Σ_± e^{−g_±²/(2σ_Z²)} dτ`, where `g_± := ±θ − μ_Z − t₀ − τ`. Note that `K ≤ Cκh²`. - *Coordinates.* Use Weyl coordinates for `A`, as in Lemma D′: `λ` is an eigenvalue of least modulus and `D` the product - of the other moduli (`D := 1` if `m = 1`), so `|Δ| = |λ|D`, and `λ < 0` on `{A < 0}`. Condition on the eigenvectors, - the other eigenvalues and the free jets `𝐠` other than `A` and `f₄`. Then `λ` ranges over `[−x₂, 0)`, where `x₂` is the - second smallest modulus (over `(−∞, 0)` if `m = 1`), with density `ρ(λ) ≤ C2^{m−1}D|𝒱′|p̄`. Here `𝒱′` is the Vandermonde - factor of the other eigenvalues and `p̄` the supremum over `λ` of the Gaussian density of `A` given `𝐠`. + of the other moduli (`D := 1` if `m = 1`), so `|Δ| = |λ|D`, and `λ < 0` on `{A < 0}`. In the Weyl-coordinate + integral, fix the eigenvectors `O`, the other eigenvalues `λ′` and the free jets `𝐠` other than `A` and `f₄`, and + use the *unnormalized* slice `ρ(λ)dλ := c_m q_{A|𝐠}(O diag(λ, λ′)Oᵀ)|𝒱(λ, λ′)|dλ`, to be integrated afterwards over + `dλ′ dO dP_𝐠`. Here `q_{A|𝐠}` is the Gaussian density of `A` given `𝐠` on the symmetric matrices and `𝒱` the + Vandermonde factor. (A normalized conditional density of `λ` would carry the inverse of the slice integral, which is + not bounded uniformly; it is not used.) Then `λ` ranges over `[−x₂, 0)`, where `x₂` is the second smallest modulus + (over `(−∞, 0)` if `m = 1`), and `ρ(λ) ≤ C2^{m−1}D|𝒱′|p̄`, where `𝒱′` is the Vandermonde factor of the other + eigenvalues and `p̄ := sup_λ q_{A|𝐠}(O diag(λ, λ′)Oᵀ)`. - *Dependence on `λ`.* `θ = −6κDλ` is linear in `λ`; `A^♯`, `β`, `m₄`, `Y₁` and `Y₂` are affine; `μ_Z` is quadratic. Let `P := C₀(1 + ‖A‖ + |𝐠|)^{N₀}`, with `C₀`, `N₀` chosen so that `|μ_Z′| + |t₀′| ≤ P/2` and `|m₄| ≤ P` on the range of `λ`. For `m ≥ 2`, `‖A‖ = x_m` does not depend on `λ`; for `m = 1` the conditions below cut out an interval of `λ`. @@ -286,10 +303,11 @@ as `E[Φ(t₀) − Φ(0)] + E[½(Φ(t₀ + h) + Φ(t₀ − h)) − Φ(t₀)]`, `|Δm₄|/12 ≤ θ/72`. Where `|g| ≥ θ/2`, `e^{−g²/(2σ_Z²)} ≤ e^{−648κ²/σ₄²}`, and the contribution is `≤ Cκh²e^{−648κ²/σ₄²} ≤ Ch²κ^{−1}`. Where `|g| < θ/2`, `|β + t₀ + τ| > θ/2 − θ/72 > θ/4`. On the support `|τ| < |h(λ)|` of the weight, `|β + t₀ + τ| ≤ M₀ := sup_λ(|β| + |t₀| + |h|)`, a polynomial in `(‖A‖, 𝐠)` that does - not depend on `λ`; so `|λ| < M₀/(1.5κD)`. Hence + not depend on `λ`. (For `m = 1`, `adj A = 1`, so `β`, `Y₁` and `Y₂`, hence `t₀` and `h`, do not depend on `λ`, and + `M₀` and `H` below are finite although `λ` ranges over `(−∞, 0)`.) So `|λ| < M₀/(1.5κD)`. Hence `σ_Z < σ̄ := M₀σ₄/(18κ)` there. Bound `(|h| − |τ|)₊ ≤ (H − |τ|)₊` with `H := sup_λ|h|`, and change variables `λ ↦ g` (`|dλ/dg| ≤ 1/(3κD)`): `∫e^{−g²/(2σ̄²)}ρ dλ ≤ C D|𝒱′|p̄·√(2π)σ̄/(3κD) = C|𝒱′|p̄M₀σ₄/κ²`. With the prefactor and - `∫(H − |τ|)₊dτ = H²` this gives `≤ C|𝒱′|p̄M₀H²/κ`, and integrating over the conditioning variables gives + `∫(H − |τ|)₊dτ = H²` this gives `≤ C|𝒱′|p̄M₀H²/κ`, and integrating over `dλ′ dO dP_𝐠` gives `E[K1_𝔐] ≤ Ck²(1 + k)^Nκ^{−1}`. No largeness of `κ` is used beyond `κ ≥ 1`. - *Exceptional set* `𝔐^c ⊂ {D < P/κ} ∪ {P > κ}`. Use `K ≤ Cκh²`. For `m ≥ 2`, Lemma D″, applied given `𝐠` in dyadic layers of `‖A‖`, gives `E[h²1{D < P/κ}] ≤ Ck²κ^{−2}`. For `m = 1`, `D = 1` and this set is `{P > κ}`. By Markov, @@ -354,11 +372,15 @@ the endpoint matrices `A_i := ∇_Θ²f(x_i)`, which holds since `‖A_i − A - (ii) Moving the odd mean from `0` to `k·w` costs, by the Gaussian score (no derivative of an indicator is taken), `≤ Ck·sup E[|ΔV_o|(1 + |jets|)1{|Y′| > 6κ|Δ|}1{A<0}] ≤ Ck(1 + k)^N/(1 + κ)²`. On the layer `|Δ| < |Y′|/(6κ)`, so Lemma Λ₀ applies along the interpolation. - - (iii) Replacing `Y′` by `U` in the indicator (`U − Y′ = 3kΔ_B`). The two indicators differ only where `Y′` is - within `3k|Δ_B|` of `±6κ|Δ|`, that is, for `f₄` in two windows `W` of total length `≤ 72k|Δ_B|/|Δ|`, centred at - `f₄ = 12(±6κ|Δ| − β)/Δ` with `β := −γᵀA^♯γ/4`. Write `ΔV_o = Δ(a₁f₄ + a₀)` with `a₁ = Δ_B/24` and `a₀` free of - `f₄`. Since `sup_x(1 + |x|)p_{f₄}(x) ≤ C(1 + |m₄|)`, the factor `Δ` cancels the `1/|Δ|` of the window length: - `E_{f₄}[|ΔV_o|1_W] ≤ Ck(1 + |m₄|)(|a₁| + |a₀|)|Δ_B|`, uniformly in `Δ`. For `k ≤ 1` and + - (iii) Replacing `Y′` by `U` in the indicator (`U − Y′ = a := 3kΔ_B`). The two indicators differ only where `Y′` + lies in one of the two one-sided threshold-crossing strips between `±6κ|Δ|` and `±6κ|Δ| − a`, each of length + `|a|` and centred at `±6κ|Δ| − a/2`. (If `|a| > 12κ|Δ|`, the strips overlap and the difference set is smaller; + only this containment is used.) Since `Y′ = (Δ/12)f₄ + β` with `β := −γᵀA^♯γ/4`, this is `f₄` in two windows `W` + of total length `≤ 72k|Δ_B|/|Δ|`, centred at `f₄ = 12(±6κ|Δ| − a/2 − β)/Δ`. (Windows centred at the unshifted + points `12(±6κ|Δ| − β)/Δ` would need total length `144k|Δ_B|/|Δ|`; the bounds below hold with either.) Write + `ΔV_o = Δ(a₁f₄ + a₀)` with `a₁ = Δ_B/24` and `a₀` free of `f₄`. Since `sup_x(1 + |x|)p_{f₄}(x) ≤ C(1 + |m₄|)`, + the factor `Δ` cancels the `1/|Δ|` of the window length: `E_{f₄}[|ΔV_o|1_W] ≤ Ck(1 + |m₄|)(|a₁| + |a₀|)|Δ_B|`, + uniformly in `Δ`. For `k ≤ 1` and `|Δ| > (|β| + 3|Δ_B|)/(3κ)`, on `W` we have `|Δf₄|/12 = |Y′ − β| ≥ 6κ|Δ| − 3|Δ_B| − |β| > 3κ|Δ|`, so `W ⊂ {|f₄| > 36κ}`. There `(1 + |x|)p_{f₄}(x) ≤ C(1 + |m₄|)^Nκ^{−N}`, and this part costs `O(kκ^{−N})`. The complementary set `{|Δ| ≤ (|β| + 3|Δ_B|)/(3κ)}` has probability `≤ C/κ`, with the polynomial weights (Lemma D in @@ -553,7 +575,7 @@ This proves (P.1). ∎ | Control | Checks | |---|---| | Q1 | the ledger of §5: least `3/5` at `ρ = ℓ^{1/5}`, attained by `ρ³`, `ℓ²ρ^{−7}`, `ℓρ^{−2}`; the split is optimal on a grid and is the edge `k = r²`; the ledger without Lemma K also gives `3/5`; the closed form `∫_0^∞ r min(k, 1/k) dr = (6/5)ℓ^{2/3}` (antiderivatives evaluated exactly at three values); `3/5 > 1/2 > 3/7 > 4/11` | -| Q2 | Lemma R and the evenness of Lemma Π on fourteen exactly pinned polynomial fields (`d = 1, 2, 3, 4`): pinned jets even in `r`; `Π` even; the new row of §1 has target `0` (a consistency identity); point reflection maps the pins at `k` to those at `−k` and preserves `Π`; `det K_M` and `det K_S` affine in `k` (at `d + 1` values of `k`, which determine a polynomial of degree `≤ d`) | +| Q2 | Lemma R and the evenness of Lemma Π on fourteen exactly pinned polynomial fields (`d = 1, 2, 3, 4`): pinned jets even in `r`; `Π` even; the new row of §1 has target `0` (a consistency identity); point reflection maps the pins at `k` to those at `−k` and preserves `Π`; `det K_M` and `det K_S` affine in `k` at fixed free jets (at `d + 1` values of `k`, which determine a polynomial of degree `≤ d`) | | Q3 | (2.2) exactly on the same fields: `det K_{M,S} = ∓6kΔ + rU ∓ r²V + …`, `Π = 36k²Δ² + r²(12kΔV − U²) + …`, with `Δ_BB = ∂_t²det(A + tB)` and `A^♯_B = ∂_t adj(A + tB)` at `t = 0` in general form (so `d = 4` tests the terms that vanish or simplify for `m ≤ 2`); `U` invariant and `V` odd under `(k, odd jets) ↦ (−k, −odd jets)` | | Q4 | Lemma K for the Gaussian kernel: the `A`-integrand `H(k, A)` of the birth-integrated `r²` coefficient (exact Gaussian conditioning and Isserlis) has only `k⁰, k², k⁴` terms in `d = 1, 2, 3`, and `a = 12`; the `d = 1, 2` polynomials equal #232 §4's | | Q5 | Step U2's pointwise algebra on rational grids: `y² − min(y², c²) = (y² − c²)₊`; `|(x + e)₊ − x₊ − e1{x>0}| ≤ |e|1{|x| ≤ |e|}`; the Case 1 sign decomposition with the anti-typed term; the Case 3 identity `P(1_{typed} − 1) − X₊ = P1_{typed} − (−X)₊ + ε` | diff --git a/frontiers/candidate_parity_rate_20261001/README.md b/frontiers/candidate_parity_rate_20261001/README.md index 717f30a88..178d63dcc 100644 --- a/frontiers/candidate_parity_rate_20261001/README.md +++ b/frontiers/candidate_parity_rate_20261001/README.md @@ -2,7 +2,11 @@ **Author-side proof candidate (Anthropic Claude). Scientific effect: NONE. Nonauthor review required.** -Object `CL-CANDIDATE-PARITY-RATE-20261001-v1`. Full text: [`PROOF.md`](PROOF.md). +Object `CL-CANDIDATE-PARITY-RATE-20261001-v1.1`. Full text: [`PROOF.md`](PROOF.md). + +**v1.1 (wording only).** The six nonblocking corrections of the C55 nonauthor reviews are applied; no estimate or +conclusion changes (see *Review record*). `parity_check.py` changed in one docstring and one comment; `RESULTS.json` is +byte-identical to v1. ## Result @@ -46,7 +50,7 @@ and on CPython 3.10–3.14 (a few seconds). | Control | Checks | |---|---| | Q1 | The ledger: least `3/5` at `ρ = ℓ^{1/5}`, attained by `ρ³`, `ℓ²ρ^{−7}`, `ℓρ^{−2}`. The split is optimal. The exponent is still `3/5` without Lemma K. The closed form of `∫ r min(k, 1/k) dr` is evaluated (a consistency identity). | -| Q2 | On fourteen exactly pinned polynomial fields (`d = 1, 2, 3, 4`): pinned jets are even in `r`; `Π` is even; the reflection maps `k` to `−k`; `det K_M` and `det K_S` are affine in `k`, checked at `d + 1` values. | +| Q2 | On fourteen exactly pinned polynomial fields (`d = 1, 2, 3, 4`): pinned jets are even in `r`; `Π` is even; the reflection maps `k` to `−k`; at fixed free jets, `det K_M` and `det K_S` are affine in `k`, checked at `d + 1` values. | | Q3 | (2.2) exactly on the same fields, with the general `Δ_BB` and `A^♯_B`. The `d = 4` fields fail the old `m ≤ 2` shortcut, so they test the `m ≥ 3` terms. The parities of `U` and `V` are checked. | | Q4 | Lemma K for the Gaussian kernel. The `A`-integrand `H(k, A)` of the birth-integrated `r²` coefficient has only `k⁰, k², k⁴` terms (`d = 1, 2, 3`). The `d = 1, 2` polynomials are #232's. | | Q5, Q6 | The pointwise algebra of Step U2, including the Case 3 identity, and the inequalities that collect Lemma U's errors. Three of Q6's four relations are identities, recorded as consistency checks. | @@ -110,6 +114,25 @@ MINOR FIXES** for §§4–8, with three minor items and six nits. All are applie **Independence.** The referees are the same provider and the same GitHub account as the author, so they carry zero organizational independence. Nonauthor review is required; the review slices are in PROOF §9. +**Nonauthor review (C55: OpenAI Codex, delegated AI review, on head `a6b6788`, PROOF blob `97298f32`).** All three +slices are PASS_TECHNICAL at their stated scope. Organizational independence is 0, and Dylan Roy's personal reading is +pending. +- *Slice A* (§§1–2: Lemmas R, Π, K), review 5385194600. Two nonblocking precision notes. +- *Slice B* (§3: Lemmas D′, D″, Λ₀, Λ), review 5385169112. Two nonblocking prose corrections and one clarification. +- *Slice C* (§§4–5: Lemma U and the assembly), review 5385136758, with the addendum 5940107723. One nonblocking + correction. + +Each correction is applied in v1.1, as follows. + +| Review item | v1.1 | +|---|---| +| A: `det K_i` affine in `k` | at fixed free jets only, not along the laws `Q̄_{r,k}` (Lemma Π, its proof, Q2) | +| A: signed-`k` moments | `1 + \|k\|` in the moment bounds before Lemma Π and in Lemma Π | +| B-01: `φ″ = 2` off `{\|y\| = θ}` | `φ″ = 2·1{\|y\| > θ} + 2θ(δ_{−θ} + δ_θ)` as a distribution | +| B-02: the density `ρ(λ)` | the unnormalized Weyl slice `c_m q_{A\|𝐠}(O diag(λ, λ′)Oᵀ)\|𝒱\|dλ`, integrated over `dλ′ dO dP_𝐠` | +| B-02: the case `m = 1` | `adj A = 1`, so `β`, `Y₁`, `Y₂` do not depend on `λ`, and the suprema are finite | +| C-N1: the `f₄`-windows | two one-sided crossing strips of length `\|a\|`, `a = 3kΔ_B`, centred at `±6κ\|Δ\| − a/2` (total `72k\|Δ_B\|/\|Δ\|`; `144` if centred at the old thresholds) | + ## Not claimed - sharpness of `3/5`. The next ledger term is `ℓ^{2/3}`, and reaching it would need Lemma U for `𝐓_r(k) − 𝐓_r(0)` on all diff --git a/frontiers/candidate_parity_rate_20261001/SOURCES.json b/frontiers/candidate_parity_rate_20261001/SOURCES.json index 95ecb2845..ac3210288 100644 --- a/frontiers/candidate_parity_rate_20261001/SOURCES.json +++ b/frontiers/candidate_parity_rate_20261001/SOURCES.json @@ -1,7 +1,7 @@ { "schema": 1, "scientific_effect": "NONE", - "object": "CL-CANDIDATE-PARITY-RATE-20261001-v1", + "object": "CL-CANDIDATE-PARITY-RATE-20261001-v1.1", "kind": "author-side proof candidate: the candidate lifetime density with remainder O(l^(3/5)) in every d >= 2 (Theorem P): nu_cand = c l^(-1/3) + B_{d,L} + I^cand l^(1/4) + c2 l^(1/3) + O(l^(3/5)), improving O(l^(4/11)) (Math- #218) and O(l^(3/7)) (Math- #229); Corollary P-prime: no l^(1/2) term. New: Lemma R (birth integration; independent even/odd parts; point reflection k -> -k; evenness in r), Lemma Pi (Pi = 36 k^2 Delta^2 + r^2 (12 k Delta V - U^2) + O(r^4)), Lemma K (the surrogate is smooth and even in r; A_2 even in k), Lemma D-double-prime (two small eigenvalues, eps^2), Lemma Lambda (the gap inside the cusp layer costs O(k^2/(1 + kappa))), Lemma U (one uniform two-scale kernel formula for k >= r^2 with error O(r^2 + r min(k, 1/k)); at the cusp scale the O(r) correction vanishes)", "author": "Anthropic Claude", "session": "https://claude.ai/code/session_01NMeKEismAyeqgdB4sy2NJU", @@ -218,23 +218,47 @@ "files": [ { "path": "PROOF.md", - "bytes": 50928, - "sha256": "aa4200100fcdf23135f82b7ac6f69621dedc65983d6707970f93d0343c6fca77" + "bytes": 53499, + "sha256": "9d58a29bf725d0756c8d11b5a9aa538782e83bee62242bcb1a869d2b95ccce94" }, { "path": "README.md", - "bytes": 9263, - "sha256": "e8e6c4bc50c858c5d29be390a52178f22945e456fa2b6a006ad6874298e7e1b1" + "bytes": 11002, + "sha256": "515714528b4257378239093bcaa05f83a7d7714f0666856471f6b2c6292f495c" }, { "path": "parity_check.py", - "bytes": 26836, - "sha256": "fbea93a4810d21130a6dc95a593f36d09756368798f6b7ccf608520c68331470" + "bytes": 26894, + "sha256": "1d103f0f00c1752d96aed59b581965d01a0dbbf26f3832a7a398bc0f5f8f5c5c" }, { "path": "RESULTS.json", "bytes": 1433, "sha256": "2af10a58aae284635de8973840594a71f18f1243415e1e8a3eb9d8a64ee669b1" } - ] + ], + "revision_history": "v1.1 (wording only; no estimate or conclusion changes) applies the six nonblocking corrections of the C55 nonauthor reviews on head a6b67882344cc39c76c29f5a6a6787a0e243c630: (1) Lemma Pi: det K_i is affine in k at fixed free jets, not along the laws Q-bar_{r,k} (statement, proof, Q2 row, README, checker docstring); (2) the moment bounds before Lemma Pi and Lemma Pi are stated for signed k with 1 + |k|; (3) Lemma Lambda: phi'' = 2*1{|y| > theta} + 2 theta (delta_{-theta} + delta_theta) as a distribution; (4) Lemma Lambda kink step: rho is the unnormalized Weyl-coordinate slice c_m q_{A|g}(O diag(lambda, lambda') O^T) |V| d lambda, integrated over d lambda' dO dP_g; (5) the case m = 1 of that step (adj A = 1); (6) Step U2 (b)(iii): the f4-windows are the two one-sided threshold-crossing strips of length |a|, a = 3k Delta_B, centred at +-6 kappa |Delta| - a/2 (total length 72k|Delta_B|/|Delta|; 144 if centred at the unshifted thresholds). RESULTS.json is byte-identical to v1 and keeps its object label; parity_check.py changed in one docstring and one comment.", + "nonauthor_reviews": [ + { + "slice": "A (sections 1-2: Lemmas R, Pi, K)", + "record": "Math- #237 review 5385194600", + "verdict": "PASS_TECHNICAL at the stated scope; two nonblocking precision notes (fixed-free-jet affinity; 1 + |k|), applied in v1.1" + }, + { + "slice": "B (section 3: Lemmas D', D'', Lambda_0, Lambda)", + "record": "Math- #237 review 5385169112", + "verdict": "PASS_TECHNICAL at the stated scope; nonblocking prose corrections B-01 (phi'') and B-02 (unnormalized Weyl slice; m = 1 sentence), applied in v1.1" + }, + { + "slice": "C (sections 4-5: Lemma U and the assembly)", + "record": "Math- #237 review 5385136758 and addendum comment 5940107723", + "verdict": "PASS_TECHNICAL at the stated scope; nonblocking correction C55-C-N1 (one-sided crossing strips), applied in v1.1" + } + ], + "nonauthor_reviews_common": { + "reviewer": "OpenAI Codex under \"Dylan Roy - delegated AI review\" (C55 session; pickup comment 5939770763, delivery comment 5940196613)", + "head": "a6b67882344cc39c76c29f5a6a6787a0e243c630", + "proof_blob": "97298f32f64bbed591e3067865629ccb838d4e9d", + "limits": "organizational independence 0; Dylan Roy personal reading PENDING; reviews the candidate density only, no elder or rejected-density claim; scientific effect NONE" + } } diff --git a/frontiers/candidate_parity_rate_20261001/parity_check.py b/frontiers/candidate_parity_rate_20261001/parity_check.py index 3e0e1a382..9138b55a7 100644 --- a/frontiers/candidate_parity_rate_20261001/parity_check.py +++ b/frontiers/candidate_parity_rate_20261001/parity_check.py @@ -15,8 +15,8 @@ Q2 Lemma R and the evenness of Lemma Pi on exactly pinned polynomial fields (d = 1, 2, 3, 4): in the birth-integrated (midpoint) coordinates every pinned jet is an even polynomial in r; Pi(r) = -det H(-r/2) det H(r/2) / r^2 is even in r; the new row of section 1 has target 0 (a consistency identity); - point reflection maps the pins at k to the pins at -k and preserves Pi; det K_M and det K_S are affine in k - (checked at d + 1 values of k, which determine a polynomial of degree <= d) + point reflection maps the pins at k to the pins at -k and preserves Pi; at fixed free jets, det K_M and + det K_S are affine in k (checked at d + 1 values of k, which determine a polynomial of degree <= d) Q3 Lemma Pi, (2.2), on the same fields: det K_(M,S) = -/+ 6 k Delta + r U -/+ r^2 V + O(r^3) and Pi = 36 k^2 Delta^2 + r^2 (12 k Delta V - U^2) + O(r^4), with U, V of (2.1) built from the general Delta_BB = d^2/dt^2 det(A + t B) and A#_B = d/dt adj(A + t B) at t = 0 (so d = 4 tests the m >= 3 terms); @@ -327,8 +327,9 @@ def control_Q2_Q3(): q2 &= all(not (Rr[i] - tgt[i]).c for i in range(len(Rr))) Pr, _, _ = Pi_of(refl, m) q2 &= not (Pr - Pi).c - # det K_M and det K_S affine in k: both are polynomials of degree <= d in k (the pinned coefficients are - # affine in k), so d + 1 values t k, t = 1..d+1, decide; all second differences must vanish + # at fixed free jets, det K_M and det K_S are affine in k: both are polynomials of degree <= d in k (the + # pinned coefficients are affine in k), so d + 1 values t k, t = 1..d+1, decide; all second differences + # must vanish KK = [Pi_of(pin_field(free, m, bp, t * k), m) for t in range(1, m + 3)] for idx in (1, 2): vals = [x[idx] for x in KK]