Home·The theorem·The proof in one reading·Architecture·Units·Manuscript·Glossary·FAQ·Work items·Data·Method·Provenance·Changes
← VI.2 The free Maxwell stress tensor and the local Ward charge (notes)Part VII·Unit outlineVII.2 The trace of the stress tensor and the trace anomaly (notes) →

VII.1. Short distances: convergence to free Maxwell fields — detailed notes

Contents

1. What this unit proves 2

2. What it uses 7

3. Strategy 9

4. Steps 9

4.1 The free lattice covariance 9

4.2 Order zero as a Gaussian recursion 11

4.3 Tested sums 12

4.4 Detection on the free side 13

4.5 Corrected charts and bookkeeping 15

5. Where the difficulty lies 16

6. Notation 17

7. What remains unexplained 17

The full account of unit VII.1: every proposition with its proof sketch, checked against the Lean in two rounds and edited. The condensed outline and the map of the Lean are on the unit page.

Unit page · All units · Overview

Part: VII, identification as Yang–Mills. Lean: the chain of YM/Pert (folders Fourier, Align, Warp, Rig, the jet stencils of Lattice, the statement Position/Crest.finite_leading, and Dense), YM/Push/FINITE_DETECTION, YM/SharpGram/Weak, YM/SharpGram/Dense, YM/RG08 (about thirty small sub-folders), YM/PertExtension, YM/OPE/Levy. Size: 8,117 lines. They divide as follows:

The main result identifies the order-zero lattice correlations with the free Maxwell hierarchy. The short-distance limit of the interacting correlations is stated in FirstExtension (VI.2), established by firstTarget in VIII.1. The results are:

1. What this unit proves

Setting.

Gauge group and continuum objects.

The free Maxwell hierarchy (III.16).

The lattice prescription at order zero (definitions from VII.3).

\(\displaystyle R_h (p, o) (A) =\mathbb{E}_{\xi} \hspace{0.17em} p (\ell^h_o (A) + i \xi),\)

where \(\xi\) is an independent centred Gaussian on the keys of rank \(\leqslant d\) with covariance

\(\displaystyle S_h (a, b) = (2 \pi)^{- 4} \int_{[- \pi, \pi]^4} \hat{\Delta} (k)^{- 1} \sum_c \lambda_a (c, k) \lambda_b (c, k) \hspace{0.17em} dk.\)

Here \(\hat{\Delta} (k) = \sum_i 4 \sin^2 (k_i / 2)\), \(c\) runs over cosine/sine, axis and colour channels, and \(\lambda_a (c, \cdot)\) is \(\ell^h_{0, a}\) applied to the corresponding plane wave on \(\mathbb{Z}^4\). So \(S_h\) is the covariance of the jet stencils at one site for the Feynman-gauge Gaussian on the infinite lattice \(\mathbb{Z}^4\), and \(R_h (p, o) = \hspace{0.17em} : \hspace{-0.17em} p (\ell^h_o) \hspace{-0.17em} :_{S_h}\) is Wick-ordered with respect to the infinite-lattice self-contraction (column, link).

\(\displaystyle \Pi_m (f ; p) = h_m^{4 r} \sum_{o \in \Lambda_{N_m}^r} f (h_m (o - c_m)) \hspace{0.17em} \left\langle \prod_j R_{h_m} (\text{sec}_d \hspace{0.17em} p_j, o_j) \right\rangle^{(0)}_{N_m}\)

(profile, row).

Proposition VII.1.1 (regulator removal at order zero). Fix an integer \(L \geqslant 2\), \(d, r \in \mathbb{N}\), label frames, and \(p_1, \ldots, p_r \in F_d\). Let \(f\) be continuous with compact support and \(\text{supp} \hspace{0.17em} f \subset \text{Conf}_r\), where support is closed. Then

\(\displaystyle \lim_{m \to \infty} \Pi_m (f ; p) = \int f (x) \hspace{0.17em} \mathcal{W}_x (\omega_d p_1, \ldots, \omega_d p_r) \hspace{0.17em} dx.\)

In board form: let \(Q\) be a position board whose profiles are smooth test functions compactly supported in \(\text{Conf}\) ("clean profiles"). Then the order-zero board readings converge to the free readings, slot by slot and for every tuple of basis labels. Status: theorem. There is no hypothesis on the construction; it concerns lattice Gaussian integrals and the continuum kernel only. Lean: Rig.zerothProfile. Board form: Position.finite_leading, using the definitions finite, white and CleanProfiles (all VII.3).

Proposition VII.1.2 (pointwise convergence of order-zero correlations). Let \((N_i, h_i)\) be an itinerary with \(4 \mid N_i\). Let \(z \in \text{Conf}_r\), and let \(o_{i, j} \in \Lambda_{N_i}\) be homing to \(z_j\) for each \(j\). Then for all \(p_1, \ldots, p_r \in F_d\),

\(\displaystyle \left\langle \prod_j R_{h_i} (\text{sec}_d p_j, o_{i, j}) \right\rangle^{(0)}_{N_i} \longrightarrow \mathcal{W}_z (\omega_d p_1, \ldots, \omega_d p_r) .\)

Status: theorem. Lean: Warp.row_limit, through poly_limit and earth.

Proposition VII.1.3 (corrected two-leg covariance). Let \((N_i, h_i)\) be an itinerary with \(4 \mid N_i\). Let \(\iota\) be any index set, let \(z : \iota \to \mathbb{R}^4\) be injective, and for every \(j \in \iota\) let \(o_{i, j}\) be sites homing to \(z_j\). For indices \(j, l \in \iota\) and arbitrary keys \(a, b\), define

\(\displaystyle \kappa_h (j, a ; l, b) = \text{Cov}_{\mu_N} (\ell^h_{o_j, a}, \ell^h_{o_l, b}) - \delta_{jl} \hspace{0.17em} S_h (a, b) .\)

Then

\(\displaystyle \kappa_{h_i} (j, a ; l, b) \to \left\{ \begin{array}{ll} K_{ab} (z_j - z_l), & j \ne l,\\ 0, & j = l. \end{array} \right.\)

Status: theorem. Lean: Align.corrected_limit, from orbit_limit, orbit_contact and shell_exact.

Proposition VII.1.4 (two-point detection). Let \(e_{\text{t}}\) be the unit vector of the time axis, \(\sigma\) the time reflection \(x_{\text{t}} \mapsto - x_{\text{t}}\) acting on labels, and \(\bar{v}\) the complex conjugate label. If \(v \in V_d \setminus \mathcal{I}_d\), then

\(\displaystyle \text{Re} \hspace{0.27em} \mathcal{W}_{(e_{\text{t}}, 0)} (\rho v, \rho (\sigma \bar{v})) > 0.\)

More generally, let \(\Phi\) and \(\Psi\) be the algebra maps sending shell coordinates to the diagonalizing and reflected probe variables of Lemma VII.1.m. For every nonzero \(q \in \text{WickPol}\), \(\text{Re} \hspace{0.17em} \mathcal{W}_{(e_{\text{t}}, 0)} (\Phi q, \Psi \bar{q}) > 0\). Here \(\Phi\) is surjective, and \(\rho (\sigma \bar{v}) = \Psi (\bar{q})\) whenever \(\rho v = \Phi q\). Status: theorem, about the free kernel only. No Wilson family and no reflection positivity of a measure is involved. Lean: Free.two_detect, Free.positive, detectors.

Proposition VII.1.5 (Kelvin formulas for the free kernel).

(a) Let \(A\) be a real symmetric \(j\)-tensor on \(\mathbb{R}^4\) whose diagonal \(x \mapsto A (x, \ldots, x)\) is harmonic. Then for \(x \ne 0\)

\(\displaystyle \frac{1}{j!} A (\partial, \ldots, \partial) \hspace{0.17em} \Gamma (x) = \frac{(- 2)^j}{4 \pi^2} \hspace{0.17em} |x|^{- 2 (j + 1)} A (x, \ldots, x) .\)

(b) Let \(Y \in \text{Sh}_j\) be \(\mathfrak{g}\)-valued, with homogeneous Maxwell solution \(Y^{\natural} (x) = Y (x, \ldots, x)\). Contract the first leg of the curvature kernel at \(x\) with \(Y\), using its rank-\(j\) Taylor coefficient and summing over colours, ordered form indices and jet indices, without a factor \(1 / 2\). For \(x \ne 0\) the result equals

\(\displaystyle \gamma_j \hspace{0.17em} (\iota^{\ast} Y^{\natural}) (x), \qquad \gamma_j = \frac{2^j (j + 1) (j + 2)}{\pi^2} .\)

Here \(\iota (x) = - x / |x|^2\) and \(\iota^{\ast}\) is the pullback of 2-forms:

\(\displaystyle (\iota^{\ast} F)_{km} (x) = |x|^{- 4} \left[ B_{km} - 2| x|^{- 2} \left( x_k \sum_a x_a B_{am} - x_m \sum_a x_a B_{ak} \right) \right], \qquad B_{ab} = F_{ab} (\iota x) .\)

Status: theorem. Lean: Kelvin.harmonic_Green (with R \(= |x|^{- 2}\)), and Kelvin.response_eq (with inverted and gain).

Proposition VII.1.6 (corrected measuring charts). Fix a Wilson family \(C\) with a realization of its defining conditions, a construction record \(U\) with limit table \(S\) and extraction order \(f\), and OPE table data as in VII.4. Let \(D \in \mathbb{N}\), and let \(u\) be a nontrivial filter on \(\mathbb{N} \times \mathbb{N}\) finer than the extraction filter of \(f\).

Assume a weak measuring portfolio \(X\) through degree \(D\). Its data include the observable bounds, Haar bounds, a Haar chart ratio, three price panels, and a constant dominating the label costs through \(D\). It supplies an index \(g (k) \to \infty\), finite-regulator linear label maps, and limiting maps \(\Phi_{k, d}\) on the reference label-coordinate spaces. Eventually in \(k\):

Let \(\lambda_d\) be the fixed equivariant representative of the free quotient in reference label coordinates, and let \(\pi^S_d\) be projection to \(S_d\). Set \(D_{k, d} = \pi^S_d \Phi_{k, d} \lambda_d\). Define \(\chi_k (d) = D_{k, d}\) when this map is bijective, and choose a fixed isomorphism \(F_d \to S_d\) otherwise. Then there are complex linear isomorphisms \(p_k (d) : F_d \to S_d\), for every \(k, d \in \mathbb{N}\), such that:

  1. For all sufficiently large \(k\), simultaneously for \(d \leqslant D\), \(p_k (d)\) commutes with signed permutations and conjugation and preserves the unit. For all \(d \leqslant e \leqslant D\),

    \(\displaystyle p_k (e) \circ \iota^F_{d, e} = \iota^S_{d, e} \circ p_k (d) .\)
  2. For every \(d \leqslant D\) and \(v \in F_d\),

    \(\displaystyle \chi_k (d)^{- 1} p_k (d) v \longrightarrow v.\)

The simultaneous finite-regulator conditions are Poised, and the representative is ladder.

Status: theorem under the stated construction and portfolio hypotheses. Lemma VII.1.q supplies the required two-point convergence. Lean: Weak.Portfolio.tune_grades, with Portfolio and chart.

Proposition VII.1.7 (backward windows and local expansions).

(a) Fix the RG constants of III.2, Haar bounds, a Haar chart ratio satisfying the local ratio bounds, and a relay of admissible price panels with activation bounds \(u_j \leqslant L_j\). Each degree is eventually active; its three panels satisfy the comparison inequalities and the analytic bounds required for the rank estimates. Let \(U\) be a wait with window

\(\displaystyle w_j (t) = \left\lfloor \frac{\log (\bar{\beta} + c_G t)}{\log L_j} \right\rfloor,\)

where \(c_G > 0\) is the coupling slope. There is a wait \(W\) with no smaller minimum stage length or entry thresholds and with the same window, together with panels and cost bounds for every degree, having the following property. For every realized Wilson family satisfying \(W\), the prescribed observable bounds and the polynomial-representative conditions, and every \(D\), all sufficiently late steps \(k\) whose scale exceeds the stage entry threshold and whose backward window has length at most \(k\) and lies in one stage admit simultaneous finite-regulator rank-comparison data through \(D\), as in VII.1.6. The returned panels satisfy the three comparison inequalities at every degree, and each cost bound dominates the label costs through that degree.

(b) Let \(U\) have the window in (a), and let \(V\) have no smaller stage-length bounds, entry thresholds or windows. Assume, for every stage \(j\), that the entry threshold of \(V\) at \(j\) is at least that of \(U\) at \(j + 1\), and that \(\bar{\beta} + c_G t \geqslant c_G\) above the \(V\) threshold at \(j\). Every RG trajectory satisfying \(V\) admits \(g (k) \to \infty\) such that eventually \(g (k) \leqslant k\), the step \(g (k)\) is above its \(U\) entry threshold, its backward \(U\) window has length at most \(g (k)\) and lies in one stage, and

\(\displaystyle \text{scale} (k) \leqslant \text{scale} (g (k)) + \log \hat{\beta}_k .\)

These threshold conditions are LateEnough.

(c) Fix a finite index set \(I\), a marker subset \(T \subset I\), an anchor \(p \in T\), \(D, k, E, q \in \mathbb{N}\), a support parameter \(R \in \mathbb{R}\), and \(v > 0\), with \(q \leqslant k + E\). There are \(0 < w \leqslant v\) and \(C_0 \geqslant 0\), independent of the layout and its bound \(B\), with the following property. Suppose a layout has marker weights \(a_i \in \mathbb{N}\) with \(\sum_{i \in T} a_i = k\), depends only on the occupied markers, vanishes on the empty marker set, and satisfies the prescribed tensor symmetry, rank and analytic realization conditions (Sense). Suppose also that every coefficient of expansion order \(i \leqslant E\), at every leg rank, has the anchored \(v\)-weighted absolute-sum bound \(B\). Translate the root to \(x_p\) and shift the order by \(q\), setting coefficients with \(q + n < k\) to zero. Through leg rank \(k + E - q\), the resulting coefficients are exactly the Taylor coefficients of a finite sum of curvature expressions of ranks

\(\displaystyle k \dot{-} q, \ldots, k + E - q,\)

where \(k \dot{-} q = \max (k - q, 0)\). Each expression depends only on \(T\), retains the required symmetries and analytic realization, and has the anchored absolute-sum bound \(BC_0\) with weight \(\exp (w \hspace{0.17em} \text{girth})\), including the exponential weight on relative marker positions.

Status: theorems. Lean: Dense.slim_wait_for, Dense.nearby, Dart.coin_pre.

2. What it uses

From other units.

Standard mathematics.

Imported constants and assumptions.

3. Strategy

At order zero, the normalized perturbative expectation is Gaussian. Wick ordering introduces an independent imaginary Gaussian for each insertion, so the pair contraction is the torus stencil covariance minus the infinite-lattice self-covariance at a common insertion. Fourier analysis identifies the contractions at separated points with the Maxwell kernel. At one insertion, the remaining finite-volume error tends to zero because the period grows faster than every inverse power of the spacing. The lattice moments and the continuum Wick moments satisfy the same deletion recursion; induction on the number of distinguishable legs gives VII.1.2. Compactness and dominated convergence then give the tested limit VII.1.1.

For detection, inversion and translation turn a homogeneous Maxwell solution of degree \(j\) into a field whose Taylor expansion at the detector begins at order \(j\). Adding the corresponding lower-rank probes gives a triangular change of variables. In an orthogonal shell basis, the reflected one-leg pairing becomes diagonal with positive entries. Wick's rule then makes distinct monomials orthogonal and gives every monomial a positive squared norm, proving VII.1.4.

The interacting comparison uses the rank estimates of III.11. They imply convergence of all mixed two-point readings in one portfolio. Free detection makes the raw charts eventually invertible and forces their filtration defects to tend to zero. Fixed equivariant retractions of the free inclusions remove these defects recursively, giving the exactly compatible charts of VII.1.6. Proposition VII.1.7 supplies the backward-window and local-expansion estimates used in this construction.

4. Steps

4.1 The free lattice covariance

Lemma VII.1.a (covariance of link forms). For links \(a = (x, i)\), \(b = (y, j)\) and colours \(u, t\),

\(\displaystyle \text{Cov}_{\mu_N} (\langle A (a), u \rangle, \langle A (b), t \rangle) = \langle u, t \rangle \hspace{0.17em} \boldsymbol{1}_{i = j} \hspace{0.17em} [g_N (y - x) - g_N (y + x + e_i)] .\)

Here \(g_N\) is the mean-zero torus Green function, \(- \Delta g_N = \delta_0 - N^{- 4}\) and \(\sum_x g_N = 0\), written as a Fourier sum over \((\mathbb{Z}/ N)^4\) with the zero mode omitted (toroid, lap_eq). Proof sketch. Summation by parts gives \(\textsf{Q}_N (A, B) = - \tfrac{1}{2} \sum \langle A (x, i), \Delta B_i (x) \rangle\), the Feynman-gauge identity curl² + div² = \(- A \cdot \Delta A\) (total). The response to a unit forcing at \(a\) is \(g_N (\cdot - x) e_i\) minus the same function at the image link. The constants \(N^{- 4}\) cancel between the two images, and oddness of the test field doubles the pairing (inverse_eq). Lean: Orb.covar_link.

Lemma VII.1.b (separated legs). Let \((N_i, h_i)\) be an itinerary. For a torus site \(x\), write \([x]_N\) for its centred lift in \((- N / 2, N / 2]^4\). Let \(x_i, y_i\) satisfy \(h_i [x_i - y_i + p]_{N_i} \to z \ne 0\) and \(\|h_i [x_i + y_i + p]_{N_i} \| \to \infty\) for every fixed \(p \in \mathbb{Z}^4\). Then for real linear forms \(P\) on \(T_j\) and \(Q\) on \(T_k\),

\(\displaystyle \text{Cov}_{\mu_{N_i}} (P \circ \ell^{h_i}_{x_i}, Q \circ \ell^{h_i}_{y_i}) \to K (P \circ \kappa, Q \circ \kappa) (z) .\)

The image condition holds automatically when both families are homing and \(4 \mid N_i\), since then each coordinate of \(x_i + y_i\) is \(N_i / 2 + O (h_i^{- 1})\) modulo \(N_i\) (finale_home). Proof sketch.

  1. Average over \(W_4\) and expand the forms in component functionals. By Lemma VII.1.a each component pairing becomes \(h^{- (|D| + 2)}\) times a forward difference \(\nabla^+_D\) of \(g_N\), evaluated at \(x - y\) (direct term) and at \(x + y + e\) (image term). The Fourier symbol of \(\nabla^+_D\) is \(\prod_{k \in D} (e^{ip_k} - 1)\).

  2. Direct term (joint_casual). At physical momentum scale \(s\), set \(t = hs\) and write lattice momentum as \(tk\). The unscaled Fourier multiplier becomes \(N^{- 4} t^{|D| - 2}\) times a rescaled symbol; after multiplication by \(h^{- (|D| + 2)}\), the prefactor is \((Nh)^{- 4} s^{|D| - 2}\). The rescaled symbol is jointly smooth near \((0, k)\) for \(k \ne 0\) and equals \(\prod_{l \in D} (ik_l) / |k|^2\) at \(t = 0\). Split momentum space with a smooth bump into a low-momentum piece, dyadic annuli and a tail.

  3. Image term. The same decomposition gives \(0\) when the norm of the centred image displacement tends to infinity in physical units (joint_far).

  4. The limit is identified with \(K\) on rotated on-shell jets, and \(W_4\)-averaging returns \(K (P \circ \kappa, Q \circ \kappa)\).

Lean: Align.chart_limit, Align.orbit_limit. (proof read through the Fourier bounds; remaining analytic inputs in §7)

Lemma VII.1.c (coincident legs). Let \((N_i, h_i)\) be an itinerary and let \(x_i \in \Lambda_{N_i}\) satisfy \(\|h_i [2 x_i + p]_{N_i} \| \to \infty\) for every fixed \(p \in \mathbb{Z}^4\). For real linear forms \(P\) on \(T_j\) and \(Q\) on \(T_k\),

\(\displaystyle \text{Cov}_{\mu_{N_i}} (P \circ \ell^{h_i}_{x_i}, Q \circ \ell^{h_i}_{x_i}) -\mathcal{S}_{h_i} (P, Q) \to 0,\)

where \(\mathcal{S}_h\) is the \(\mathbb{Z}^4\) covariance of the same stencils. In particular \(S_h (a, b) =\mathcal{S}_h (x_a, x_b)\) (shell_exact). Proof sketch. The image term vanishes as in Lemma VII.1.b. The direct term is a finite combination of values \(g_N (a)\) at fixed lattice offsets \(a\), compared with the \(\mathbb{Z}^4\) Green function \(G_{\mathbb{Z}^4} (a)\) and multiplied by \(h^{- (j + k + 4)}\). The estimate for each fixed offset \(a\) (toroid_cost) gives constants \(C, A \geqslant 0\), independent of \(N\) and \(r\), such that for \(N\) even and \(0 < r \leqslant 1\), \(8 / N \leqslant r\),

\(\displaystyle |g_N (a) - G_{\mathbb{Z}^4} (a) | \leqslant C \hspace{0.17em} r^2 + A / (Nr^4) .\)

It is proved by viewing \(g_N (a)\) as a Riemann sum of \(\cos (2 \pi a \cdot q) / \hat{\Delta} (2 \pi q)\):

With \(r = h^{n + 1}\), where \(n = j + k + 4\) is the power of \(h^{- 1}\) to be beaten, the error times \(h^{- n}\) is at most \(Ch^{n + 2} + A / (Nh^{5 n + 4}) \to 0\). This comparison uses the condition \(N_i h_i^k \to \infty\) for every \(k\) (strike_two). Lean: Align.chart_return, Align.orbit_contact.

Lemma VII.1.d (VII.1.3). This is Lemma VII.1.b for \(j \ne l\) (with scene_home and finale_home) and Lemma VII.1.c for \(j = l\). The covariance matrix \(\textsf{Q}_N^{- 1}\) evaluated on the stencil vectors equals the orbital pairing (key_orb). Lean: corrected_limit.

4.2 Order zero as a Gaussian recursion

Lemma VII.1.e (order zero is a joint Gaussian expectation). Fix \(h > 0\), an even period \(N \geqslant 4\), insertion sites \(o_1, \ldots, o_r\), and \(d \in \mathbb{N}\). Let \(s\) be a finite set of distinguishable occurrences, with insertion map \(\tau : s \to \{1, \ldots, r\}\) and key map \(a : s \to \{ \text{keys of rank at most } d\}\). Repeated keys are allowed. Set \(\text{root} (s)_j = \prod_{\nu : \tau (\nu) = j} x_{a (\nu)}\). Then

\(\displaystyle \left\langle \prod_j R_h (\text{root} (s)_j, o_j) \right\rangle^{(0)} = E_h (s) := \mathbb{E} \left[ \prod_{\nu \in s} \left( \ell^h_{o_{\tau (\nu)}, a (\nu)} + i \hspace{0.17em} \xi^{(\tau (\nu))}_{a (\nu)} \right) \right],\)

under \(\mu_N \otimes \bigotimes_{j = 1}^r \mathcal{N} (0, S_h)\). Proof sketch. The weight and the vacuum both have constant coefficient \(1\), so the order-zero coefficient of the normalized expectation is a \(\mu_N\)-integral. The constant coefficient of each lattice jet is its linear part \(\ell^h\). Fubini over the independent auxiliary copies finishes the proof. Lean: Warp.amplitude_zero. The pair covariance of two legs is \(\kappa_h\) (node_wage): the auxiliary copies contribute \(i^2 S_h = - S_h\) for legs assigned to the same insertion and \(0\) otherwise, even if two insertion sites happen to coincide.

Lemma VII.1.f (symmetric deletion identity). Let \(\mu\) be a centred Gaussian on a finite-dimensional real space (possibly degenerate) and \(P_a\), \(a \in s\), real linear functionals with values in \(\mathbb{C}\). Then

\(\displaystyle |s| \hspace{0.17em} \mathbb{E} \prod_{a \in s} P_a = \sum_{a \in s} \sum_{b \in s \setminus \{a\}} \mathbb{E} [P_a P_b] \hspace{0.27em} \mathbb{E} \prod_{c \in s \setminus \{a, b\}} P_c .\)

The products are bilinear, without complex conjugation. For the legs of Lemma VII.1.e, write \(\kappa_h (\nu, \mu) = \kappa_h (\tau (\nu), a (\nu) ; \tau (\mu), a (\mu))\); then \(|s| E_h (s) = \sum_{\nu \ne \mu} \kappa_h (\nu, \mu) E_h (s \setminus \{\nu, \mu\})\). Proof sketch. For powers of a single real functional \(L\) of variance \(v\), \(\mathbb{E}L^{n + 1} = nv \hspace{0.17em} \mathbb{E}L^{n - 1}\) for \(n \geqslant 1\), with \(\mathbb{E}L = 0\), read off from the moment generating function (ev_next). The multilinear case follows by polarization (packet_cancel). The complex case follows by splitting into real and imaginary parts. Lean: exchange_s, Warp.eq_cancel.

Lemma VII.1.g (the continuum recursion). For \(z \in \text{Conf}_r\) set \(c_z (s) =\mathcal{W}_z (\rho \circ \text{root} (s))\). Then for every \(\nu \in s\),

\(\displaystyle c_z (s) = \sum_{\mu \in s \setminus \{\nu\}} \boldsymbol{1}_{\tau (\mu) \ne \tau (\nu)} \hspace{0.17em} K_{a (\nu), a (\mu)} (z_{\tau (\nu)} - z_{\tau (\mu)}) \hspace{0.17em} c_z (s \setminus \{\nu, \mu\}),\)

and summing over \(\nu \in s\) gives the analogue of Lemma VII.1.f. Proof sketch. Removing one on-shell variable \(\rho (x_{a (\nu)})\) at the point \(z_{\tau (\nu)}\) from a Wick moment amounts to applying, at every other point, the derivation that replaces one variable by its pairing with \(a (\nu)\) (take_line). On a monomial the derivation produces the sum over its variables (D_kill). There are no same-point pairings. Lean: cut_one, cut_cancel.

Lemma VII.1.h (convergence of leg products). In the setting of VII.1.2, \(E_{h_i} (s) \to c_z (s)\) for every finite set of legs. Proof sketch. The proof is by strong induction on \(|s|\).

Lean: Warp.earth.

Lemma VII.1.i (monomials, polynomials, labels; VII.1.2).

Lean: mono_single, mono_limit, poly_limit, row_limit.

4.3 Tested sums

Lemma VII.1.j (Riemann sums with continuous convergence). Hypotheses:

Writing \(\jmath_m (o) = t_m [o - c_{N_m}]_{N_m}\) for the embedding, one has \(t_m^{4 r} \sum_o f (\jmath_m (o)) \mathcal{K}_m (o) \to \int f\mathcal{G}\). Proof sketch.

Lean: Rig.grid_limit, with Arrives.

Lemma VII.1.k (VII.1.1). Apply Lemma VII.1.j with \(\omega = \text{Conf}_r\), \(\mathcal{K}_m\) the order-zero correlations and \(\mathcal{G}=\mathcal{W} (\omega_d p)\), along the regulator sequence \(h_m = L^{- m}\), \(N_m = 4 L^{m^2 + m + 1}\) (routine).

Lean: Rig.zerothProfile, Position.finite_leading.

Downstream use (not part of this unit). The assembly in VIII.1 uses the tested order-zero limit in two ways:

The position-comparison data are described in VII.3.

4.4 Detection on the free side

Lemma VII.1.l (Wick rule under linear substitution). Fix \(x \in \text{Conf}_r\) and a finite list of variable occurrences assigned to its insertions. Replace each occurrence \(a\) by a finite complex linear combination \(u_a\) of shell coordinates, allowing different Taylor ranks. Extend \(K\) to these combinations by complex bilinearity. The Wick moment of the resulting products equals

\(\displaystyle \sum_{\operatorname{matchings}} \prod_{\operatorname{pairs}} K (u_a, u_{a'}) (x_{i (a)} - x_{i (a')}) .\)

Only matchings between distinct insertions are included. Proof sketch. Expand the linear forms in the coordinate basis. Multilinearity reduces the formula to the defining Wick sum on coordinate monomials. Lean: Free.linear_substitution.

Lemma VII.1.m (diagonal one-leg kernel). With the bilinear extension of \(K\) from Lemma VII.1.l, choose in each \(\text{Sh}_j\) a basis \((Y_a)\) orthogonal for full component contraction, and set \(\lambda_a = \gamma_j \langle Y_a, Y_a \rangle > 0\). For \(Y \in \text{Sh}_j\), let \(Y_i\) be the order-\(i\) Taylor coefficient of \(Y^{\natural}\) at \(e_{\text{t}}\) and let \(\langle Y_i, \cdot \rangle\) denote the full-contraction probe. Define

\(\displaystyle \epsilon_a = \sum_{i = 0}^j \frac{\gamma_j}{\gamma_i} \langle (Y_a)_i, \cdot \rangle, \qquad \epsilon^{\sigma}_a = \epsilon_a \circ \sigma .\)

Then

\(\displaystyle K (\epsilon_a, \epsilon^{\sigma}_b) (e_{\text{t}}) = \delta_{ab} \lambda_a .\)

Proof sketch. Expand the translated field \(x \mapsto Y^{\natural} (e_{\text{t}} + x)\) into its homogeneous Taylor pieces and apply the Kelvin formula to each piece. Since \(e_{\text{t}} + \iota (e_{\text{t}}) = 0\), its inverted field has no Taylor coefficient at \(e_{\text{t}}\) below order \(j\); its order-\(j\) coefficient is \((- 1)^j\) times the time-reflected tensor \(Y\). The weighted sum defining \(\epsilon_a\) produces exactly \(\gamma_j\) times this inverted translated field. The second-endpoint sign cancels \((- 1)^j\) on the diagonal. Thus pairings with lower ranks vanish and equal-rank pairings are \(\gamma_j\) times full contraction. Symmetry gives vanishing for the opposite rank ordering, and the chosen orthogonal bases give the displayed matrix.

Lean: Free.pairing_var, with center_lower and center_diag.

Lemma VII.1.n (orthogonality and positivity; VII.1.4). For finitely supported exponent vectors \(\alpha, \beta\) in the probe variables,

\(\displaystyle \mathcal{W}_{(e_{\text{t}}, 0)} (\Phi x^{\alpha}, \Psi x^{\beta}) = \delta_{\alpha \beta} s_{\alpha}, \qquad s_{\alpha} = \prod_a \alpha_a ! \lambda_a^{\alpha_a} > 0.\)

The empty monomial has \(s_0 = 1\). Proof sketch.

Lean: pair_test, wick_orth, positive, two_detect.

Lemma VII.1.o (Kelvin formulas; VII.1.5). Proof sketch for (a). The case \(j = 0\) is the definition of \(\Gamma\). For \(j \geqslant 1\), use

\(\displaystyle \tfrac{1}{j!} A (\partial) = \tfrac{1}{j} \sum_i \partial_i \hspace{0.17em} \tfrac{1}{(j - 1) !} A (e_i, \cdot) (\partial) .\)

Each contraction \(A (e_i, \cdot)\) is again symmetric with harmonic diagonal. Then use \(\partial_i |x|^{- 2 j} = - 2 j \hspace{0.17em} x_i |x|^{- 2 (j + 1)}\) and the vanishing of \(\sum_i \partial_i A (e_i, x, \ldots, x)\) (harmonicity). The contraction identity \(\sum_i x_i A (e_i, x, \ldots) = A (x, \ldots, x)\) closes the step. Proof sketch for (b). Work colour by colour. Each component \(Y_{ab}^{\natural}\) is harmonic (closed and coclosed), so (a) applies to every term of the curvature kernel. The terms are recombined using the divergence equation and \(\partial_k (x \lrcorner Y^{\natural})_m - \partial_m (x \lrcorner Y^{\natural})_k = (j + 2) Y^{\natural}_{km}\) (Bianchi plus Euler). Lean: harmonic_Green, response_eq.

Lemma VII.1.p (finite detection). Let \(\beta\) be a basis of \(F_d\). The map \(v \mapsto (\mathcal{W}_{(e_{\text{t}}, 0)} (\omega_d v, \omega_d \beta_i))_i\) is injective on \(F_d\). A label whose two-point functions with all labels vanish at the detector pair is null in the free table. Proof sketch. If all basis probes vanish, linearity gives vanishing against every second label. For a representative of \(v\), take its time-reflected conjugate as the second label and apply VII.1.4. Conversely, \(\rho v = 0\) makes every free correlation containing that label vanish. Thus the free null space in \(V_d\) is \(\mathcal{I}_d\). Lean: free_criterion, free_null, free_injective. In the same spirit, OPE.Levy.free_positive shows that the reflected free two-point form at the detector pair is positive on \(F_d \setminus 0\).

For bounded complex linear maps \(P : E \to E\), \(X : V \to E\), \(A : E \to Z\) and \(H : V \to Z\) between complex normed spaces, the perturbation lemma test_projected is not used by the main theorem. It states that a lower bound \(a \|x\| \leqslant \|Hx\|\) for every \(x \in V\) and \(\|H - AX\| + \|(A - AP) X\| \leqslant \delta\) imply \((a - \delta) \|x\| \leqslant \|APXx\|\).

4.5 Corrected charts and bookkeeping

Lemma VII.1.q (two-point convergence at moving detectors). Under the hypotheses of VII.1.6, let \(\delta_k = r_k y_k\) be the portfolio detector pair, where \(r_k > 0\), \(r_k \to 0\), and \(y_k \to (e_{\text{t}}, 0)\). For fixed reference labels \(v_0, v_1\) of respective weights at most \(d_0, d_1 \leqslant D\), put \(\hat{v}_{k, j} = \pi^S_{d_j} \Phi_{k, d_j} v_j\). Then

\(\displaystyle S_{d_0, d_1} (\delta_k ; \hat{v}_{k, 0}, \hat{v}_{k, 1}) \longrightarrow \mathcal{W}_{(e_{\text{t}}, 0)} (\rho v_0, \rho v_1),\)

where \(\rho v_j\) means the on-shell image of the label polynomial represented by \(v_j\). Proof sketch. The finite-regulator estimate Read.full_cost converts bounds on all nonempty marker components into a moment bound. The portfolio supplies these component bounds, and passage to the common extraction subfilter gives actualRead. The resulting comparison compareC bounds the difference between the interacting moment at \(\delta_k\) and the free moment at \(y_k\) by \(C \varepsilon_{g (k)}\), with \(\varepsilon_{g (k)} \to 0\) by III.11. Continuity of the free kernel at the detector pair gives the limit. Lean: Portfolio.twoC. (proof read through the rank-comparison input)

Lemma VII.1.r (raw charts are eventually bijective and nearly compatible). Under the hypotheses of VII.1.6:

  1. The read-out maps \(D_{k, d}\) are eventually bijective for every \(d \leqslant D\), so \(\chi_k (d) = D_{k, d}\). The real part of the reflected measured form converges in operator norm to a free form bounded below by \(\delta \|v\|^2\) for some \(\delta > 0\). It is therefore bounded below by \((\delta / 2) \|v\|^2\) eventually. This proves injectivity; the comparator isomorphism in the construction record gives equality of dimensions and hence surjectivity (freely, certified, chartDraw).

  2. For every \(d < D\) and \(v \in F_d\), \(\chi_k (d + 1)^{- 1} \iota^S \chi_k (d) v \to \iota^F v\) (stepFits). Both sides are probed by two-point functions against \(\chi_k (d + 1) \beta_i\), where \((\beta_i)\) is a basis of \(F_{d + 1}\); Lemma VII.1.q gives convergence of the probes, and injectivity of the free probe (Lemma VII.1.p), through the finite-dimensional fitting lemma of §2, gives convergence of the arguments.

  3. Eventually each \(\chi_k (d)\) commutes with conjugation and signed permutations (symmetrical), and it preserves the unit.

Lemma VII.1.s (VII.1.6). Under the hypotheses of VII.1.6:

  1. Take a retraction \(P_d : F_{d + 1} \to F_d\) of the free inclusion that is equivariant and commutes with conjugation (averaging over \(W_4\) and symmetrizing with the conjugation).

  2. For fixed \(k\), define recursively \(A_0 = \chi_k (0)\) and

\(\displaystyle A_{d + 1} = \iota^S A_d P_d + \chi_k (d + 1) (1 - \iota^F P_d) .\)
  1. For each \(d \leqslant D\), the operators \(\chi_k (d)^{- 1} A_d\) tend to the identity by induction using Lemma VII.1.r(2). They are eventually invertible. Set \(p_k (d) = A_d\) when it is bijective and \(p_k (d) = \chi_k (d)\) otherwise.

  2. Exact compatibility follows from the construction together with VII.1.r(3), and \(\chi_k^{- 1} p_k \to 1\) from VII.1.r(2).

Lean: tune_grades.

Lemma VII.1.t (VII.1.7).

Parts (a) and (b) supply the common portfolio used in VII.1.6; the assembly is in III.13. Part (c) supplies local curvature expansions for the RG estimates.

(proof read through the stated station and curvature-lifting inputs)

5. Where the difficulty lies

The main analytic work is the control of finite differences of the torus Green function, uniformly as the spacing shrinks and the volume grows. The Fourier argument must retain oscillatory cancellation; an absolute bound on the full momentum integral would not give the required derivative limits.

Wick ordering uses the infinite-lattice covariance. Consequently the same-insertion contraction is not exactly zero at finite volume. For each fixed stencil offset, the available estimate is \(Cr^2 + A / (Nr^4)\), multiplied by \(h^{- n}\). Taking \(r = h^{n + 1}\) gives \(Ch^{n + 2} + A / (Nh^{5 n + 4})\). The chosen period handles every fixed \(n\) at once. This is a sufficient growth condition; no optimal volume rate is asserted.

The reflection-odd field space removes constant zero modes but introduces an image term. Reporting around the quarter period sends that image to infinite physical distance. Detection requires a separate algebraic argument: the translated inversion formula and the triangular probe substitution establish strict positivity on every nonzero on-shell polynomial, including constants. The chart correction then uses finite-dimensional perturbation theory and the interacting estimates of III.11.

6. Notation

object code name proposed notation
reflection-odd link fields, Feynman-gauge form, Gaussian LinkP, form, data \(\mathcal{A}_N\), \(\textsf{Q}_N\), \(\mu_N\)
torus Green function (mean zero) toroid \(g_N\)
linear lattice curvature jet slope \(\ell^h_x\)
\(\mathbb{Z}^4\) self-covariance of the jets (Wick subtraction) link, shellCov \(S_h\)
Wick-ordered representative column \(R_h (p, o)\)
corrected and limit two-leg covariances corrected, indirect \(\kappa_h\), \(K_{ab}\)
itinerary; homing sites Itinerary, Home \((N_i, h_i)\); \(h_i (o_i - c_{N_i}) \to z\)
regulator sequence, centre, profile topSpacing, period, center, profile \(h_m\), \(N_m\), \(c_m\), \(\Pi_m (f ; p)\)
shell, on-shell reduction, equation ideal Shell, reduceEq, equationIdeal \(\text{Sh}_j\), \(\rho\), \(\mathcal{I}_d\)
Green function, curvature jet kernel Green, legKernel \(\Gamma\), \(K (P, Q)\)
Wick moment Wick.moment \(\mathcal{W}_x\)
detector pair, time reflection detectors, Kelvin.σ \((e_{\text{t}}, 0)\), \(\sigma\)
inversion pullback, gain inverted, gain \(\iota^{\ast}\), \(\gamma_j\)
weak measuring portfolio, raw chart, corrected chart Portfolio, chart, tune_grades \(X\), \(\chi_k (d)\), \(p_k (d)\)

7. What remains unexplained

  1. Analytic inputs below the Fourier limits. The symbol smoothness, higher-difference estimate, continuum annular decay, tail estimate and fixed-offset quadrature estimate support Lemmas VII.1.b–c. The remaining lower-level estimates for periodic annular sums, the low-momentum bound and the annular Riemann-sum comparison have not been checked line by line. Their uniformity in the period and spacing remains an input to the sketches. (probable)

  2. Imported interacting estimates. The two-point limit follows from the simultaneous portfolio conditions and the rank-comparison error bound. The full analytic proofs supplying those conditions, including the station estimates and curvature-lifting estimates used in VII.1.7, are outside this verification. (probable)

  3. Other assigned packages. The remaining RG local-expansion packages and the perturbative-extension packages are not described here; no proof outline for their internal estimates is supplied. Their placement relative to III.13, III.15 and VII.6 remains to be settled. (tentative)

  4. Scope of the comparison. The Gaussian \(\mu_N\) is defined independently of the RG Gaussian measures. Proposition VII.1.1 requires no identification with them. The interacting short-distance theorem and its combination with the tested order-zero limit are the linked results in VIII.1; no all-orders perturbative comparison follows from this unit.

  5. References and size estimates. The literature citations and attributed source-line counts have not been independently verified. (tentative)

← VI.2 The free Maxwell stress tensor and the local Ward charge (notes)Part VII·Unit outlineVII.2 The trace of the stress tensor and the trace anomaly (notes) →