Home·The theorem·The proof in one reading·Architecture·Units·Manuscript·Glossary·FAQ·Work items·Data·Method·Provenance·Changes
← VII.2 The trace of the stress tensor and the trace anomaly (notes)Part VII·Unit outlineVII.4 Operator product expansion and contact extensions (notes) →

VII.3. The theory is not free: Gram positivity and exclusion of free-field profiles — detailed notes

Contents

1. What this unit proves 1

2. What it uses 7

3. Strategy 10

4. Steps 10

4.1 The Wick algebra 10

4.2 The tower Gram form 11

4.3 From towers to all Wick polynomials 13

4.4 Faithfulness and covariance of the free table 14

4.5 Degree four: rigidity and the position equations 15

4.6 Exclusion of Wick profiles (VII.3.7) 17

4.7 The formal lattice perturbation theory 19

5. Where the difficulty lies 20

6. Notation 21

7. What remains unexplained 21

The full account of unit VII.3: 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: YM/Extra/FREE_DETECTOR_ALGEBRA, YM/Pert/Position, YM/Pert/Rigid, YM/Pert/Lattice, and a few declarations of YM/Push/LOOP_COEFFICIENT. Proposition VII.3.7 is in YM/TraceExtension/Wick and TraceBridge.lean. Size: about 3,800 declaration lines on the main dependency path in the listed folders, excluding the trace-exclusion files.

The reflected Gram form of the free Maxwell Wick algebra is strictly positive. It identifies the free field spaces with on-shell polynomial labels and supplies injective two-point tests (VII.3.1–VII.3.5). At degree four, rigidity gives finite position equations with invertible linearization and a leading comparison with lattice perturbation theory (VII.3.6). Independently, the trace anomaly excludes every finite Wick profile (VII.3.7). The lattice series used in the position comparison are defined in §4.7.

1. What this unit proves

Setting.

Write \(\text{Conf}_n = \{(x_1, \ldots, x_n) \in (\mathbb{R}^4)^n : x_i \ne x_j \text{ for } i \ne j\}\).

Coefficients and jets.

Wick polynomials.

Free moments.

Reflected pairing.

\(\displaystyle \langle P, Q \rangle_{x, y} := \left\langle P (x) \hspace{0.17em} (\mathcal{R}_{x - y} \bar{Q}) (y) \right\rangle_0 .\)

Tower Gram form.

Proposition VII.3.1 (strict reflected positivity of the free Wick algebra). Let \(x \ne y\) in \(\mathbb{R}^4\). (a) For every \(P \in \mathcal{W}\): \(\langle P, P \rangle_{x, y} > 0\) if and only if \(P \ne 0\). (b) \(\overline{\langle P, Q \rangle_{x, y}} = \langle Q, P \rangle_{x, y}\) for all \(P, Q \in \mathcal{W}\). (c) For every finite \(\mathbb{C}\)-linearly independent family \((V_i)\) in \(\mathcal{W}\), the matrix \((\langle V_j, V_i \rangle_{x, y})_{ij}\) is positive definite. No simplicity or positive-dimension assumption is needed, and the polynomial degree is unrestricted. Self-pairings are real, so the inequality in (a) is a real inequality. Status: theorem. Only (a) lies on the path to the main theorem; (b) and (c) are proved but not used. Lean: (a) reflected_positive; (b) reflected_symm; (c) reflected_pd.

Proposition VII.3.2 (tower Gram form and Fock structure). Let \(x \ne y\), \(z = x - y\), \(n \in \mathbb{N}\). (a) \(B_z\) is symmetric and positive definite on \(\text{Tow}_n\). More precisely, there is a block-unitriangular automorphism \(\sigma_z\) of \(\text{Tow}_n\) (Lemma VII.3.f) such that

\(\displaystyle B_z (\sigma_z p, \sigma_z q) = \sum_{j < n} \pi_j (z) \hspace{0.17em} \langle p_j, q_j \rangle, \qquad \pi_j (z) = \frac{(j + 1) (j + 2) \hspace{0.17em} 2^j}{\pi^2 \hspace{0.17em} |z|^{2 j + 4}} > 0.\)

(b) Let \(b_1, \ldots, b_N \in \text{Tow}_n\) satisfy \(B_z (b_i, b_k) = d_i \delta_{ik}\), with any \(d_i \in \mathbb{R}\). Let \(\Phi_b : \mathbb{C} [Y_1, \ldots, Y_N] \to \mathcal{W}\) be the algebra homomorphism with \(Y_i \mapsto \ell_{b_i}\) (encode). Then

\(\displaystyle \langle \Phi_b p, \Phi_b q \rangle_{x, y} = \sum_{\alpha} p_{\alpha} \hspace{0.17em} \overline{q_{\alpha}} \hspace{0.27em} \alpha ! \hspace{0.17em} d^{\alpha}, \qquad \alpha ! \hspace{0.17em} d^{\alpha} := \prod_i \alpha_i ! \hspace{0.17em} d_i^{\alpha_i} .\)

Status: theorem. Lean: (a) towerB_sym, towerB_pos, shiftAll_gram, price, price_pos; (b) duplex_script, aligned_script.

Proposition VII.3.3 (Wick recursion; characterization of the free moments). Let \(n \geqslant 1\), \(x \in \text{Conf}_n\), \(W \in \mathcal{W}^n\), \(1 \leqslant i \leqslant n\) and \(s \in \textsf{S}\). Let \(\partial_s^{y, y'}\) be the \(\mathbb{C}\)-linear derivation of \(\mathcal{W}\) with \(\partial_s^{y, y'} X_t = C_{st} (y - y')\). Then

\(\displaystyle \langle W_1 (x_1) \cdots (W_i X_s) (x_i) \cdots W_n (x_n) \rangle_0 = \sum_{k \ne i} \langle W_1 (x_1) \cdots (\partial_s^{x_i, x_k} W_k) (x_k) \cdots W_n (x_n) \rangle_0,\)

where on the right the factor at \(x_i\) is \(W_i\). Conversely, for any \(n \geqslant 0\) and \(x \in \text{Conf}_n\), a \(\mathbb{C}\)-multilinear map \(T : \mathcal{W}^n \to \mathbb{C}\) with \(T (1, \ldots, 1) = 1\) satisfying this recursion for every \(i, s, W\) is the free moment at \(x\). Consequences:

Status: theorem. Lean: Pair.remove, Pair.extractVar; characterization byAnnihilators; consequences wick_rotate, wick_bar.

Proposition VII.3.4 (the free table is faithful on on-shell labels).

Setting:

Statements: (a) For every \(d \geqslant 0\), let \(P \in \text{sh}_d\). Suppose \(\langle P (x_0) Q_1 (x_1) \cdots Q_n (x_n) \rangle_0 = 0\) for every \(n \geqslant 0\), every choice of degrees \(d_1, \ldots, d_n \geqslant 0\), all \(Q_i \in \text{sh}_{d_i}\) and all \(x \in \text{Conf}_{n + 1}\). Then \(P = 0\). Hence:

(b) For every basis \((b_i)_{i \in I}\) of \(E_d\) and every \(x \ne y\), the map \(v \mapsto \left( \langle \varpi v (x) \hspace{0.17em} \varpi b_i (y) \rangle_0 \right)_{i \in I}\) is injective.

(c) Fix a Wilson family \(C\), any table \(S\) on its register, and a pair satisfying \(x_{\mu} \ne y_{\mu}\) for each coordinate \(\mu\), for example \((x, y) = (0, (1, 1, 1, 1))\). For every \(d \geqslant 0\) there is a column system (IV.1) for \(E_d\). It consists of:

Status: theorem. Part (a) enters the rank comparisons of III.11 and III.12; (c) supplies columns for III.11 and III.2. The embedding \(\varpi_d\) is used in VII.1. Lean: (a) no_residual (with the definitions residual, ProbeFamily), free_null, exactField, writes; (b) scanner_inj; (c) vane, vane_possible, RStage, recorded.

Proposition VII.3.5 (covariance of the free table). For every gauge group in the setting above, the free table has the following symmetries:

In addition, \(O (4)\) acts on each \(E_d\) by complex-linear isomorphisms satisfying \(\varpi_d (R \cdot v) = R \cdot (\varpi_d v)\). Status: theorem. The proof uses the \(O (4)\)-covariance of the free kernel, not any reconstruction theorem. Lean: table_spin, rideO, ride.

Proposition VII.3.6 (rigidity at degree four; position equations). Parts (a) and (c) assume that \(\mathfrak{g}\) is simple and nonabelian. Part (b) holds for any gauge group in the setting above.

Buffers. For a degree \(d\), a buffer (buffers) is a \(\mathbb{C}\)-linear map \(B : E_d \to E_d\) such that:

Buffers form a finite-dimensional real space \(\text{Buf}_d\).

Stationarity. \(B\) is stationary if

\(\displaystyle \sum_{j = 1}^r \langle v_1 (x_1) \cdots (Bv_j) (x_j) \cdots v_r (x_r) \rangle_0 = 0 \quad \text{for all } r, v \in E_d^r, x \in \text{Conf}_r .\)

Equivalently, the change of identification \(v \mapsto (1 + tB) v\) of degree-\(d\) fields leaves all free correlations unchanged to first order in \(t\).

(a) Rigidity. Every stationary buffer of degree \(4\) vanishes. In fact \(B \in \text{Buf}_4\) vanishes as soon as both of the following hold, with \(e\) the unit vector of coordinate axis \(0\):

For \(d < 4\), \(\text{Buf}_d = 0\).

(b) Square systems. Suppose every stationary buffer of degree \(d\) vanishes. Then there are finitely many sections \((r_i, p_i, f_i)_{1 \leqslant i \leqslant k}\), with \(r_i \geqslant 0\), \(p_i \in E_d^{r_i}\) and \(f_i \in C_c^{\infty} (\text{Conf}_{r_i} ; \mathbb{C})\), and a real-linear map \(g : \mathbb{R}^k \to \mathbb{R}^m\) with \(m = \dim_{\mathbb{R}} \text{Buf}_d\), such that the map

\(\displaystyle B \longmapsto g \left( \text{Re} \int f_i (x) \sum_j \langle p_{i 1} (x_1) \cdots (Bp_{ij}) (x_j) \cdots \rangle_0 \hspace{0.17em} dx \right)_{i \leqslant k}\)

is an isomorphism \(\text{Buf}_d \to \mathbb{R}^m\).

(c) Leading position comparison at degree four. Hypotheses:

Put \(B_k := \varphi (m_Q (k))\), where \(m_Q (k)\) are the tested rescaled correlations \(\int f (x) \langle U_k v_1 (r_k x_1) \cdots \rangle_S \hspace{0.17em} dx\) on basis fields. Then:

\(\displaystyle m_R^{B_k} (k) - \text{chart}_R (\pi_R^{(0)} (L, k), \varphi (\pi_Q^{(0)} (L, k))) \longrightarrow 0.\)

Here \(m_R^{B_k} (k)\) is the tested reading after replacing each field \(v\) by \(U_k (1 + B_k) v\), and \(\pi_R^{(0)} (L, k)\) is the array of order-zero lattice readings on basis fields at spacing \(L^{- k}\).

These conclusions, together with the comparisons for all clean systems described in VII.3.r, give clause (D15) of the detailed main theorem.

Status:

Lean: (a) Rigid.frozen_four, Rigid.first_frozen (unused), definition Frozen; (b) square_of_frozen, Conditions, chart, white, square, square_four; (c) residency, Resident, FourMeters, Crest, BoardAttains, four_on, normalized_four.

Proposition VII.3.7 (exclusion of finite Wick profiles; assigned to VII.2 and VIII.1).

Definitions:

(a) Suppose \(r^8 \log^2 (1 / r) \hspace{0.17em} G_{\Theta} (r) \to A\) as \(r \to 0^+\), with \(A > 0\). Then for every Wick profile \(W\) and every \(\delta > 0\) there is \(r \in (0, \delta)\) with \(G_{\Theta} (r) \ne W (r)\). Only this asymptotic law of \(G_{\Theta}\) is used. (b) Under the same asymptotic hypothesis, in the selected reconstructed theory, no finite sum of the corresponding mass-shell contraction integrals reproduces the Wightman two-point function of \(\Theta\) on every product \(\bar{f} \otimes g\) of Schwartz tests for which \(\psi_{f, g} = \overline{\hat{f}} \hspace{0.17em} \hat{g}\) is compactly supported, using the positive-sign Fourier convention. A product of mode kernels corresponds to testing \(\psi_{f, g}\) at the sum of their positive-energy on-shell momenta, with the product of their densities \(E^q h (p) / (2 E)\) (waveTest_product). Thus, for every proposed finite sum, some such pair \(f, g\) gives different readings.

With \(A = 3 \dim \mathfrak{g}/ \pi^4\) from VII.2, (a) is clause (T4) of the main theorem. Status: theorem. (a) is assigned to VII.2 and (b) to VIII.1. Lean: (a) Wick.euclidean_excludes, Wick.no_powerlog; public form publicTrace for Hierarchy.TraceTest. (b) Wick.polynomial_excludes. The asymptotic law and related polynomial-field exclusions occur in Headline.TraceExclusions.

The Euclidean conclusion concerns the single function \(G_{\Theta}\) along one axis. The excluded class allows arbitrary complex coefficients and continuous homogeneous numerators, but only finitely many modes and products. It does not include general continuous-mass generalized free fields or infinite Wick expansions.

2. What it uses

Inputs from other units.

  1. III.16 (analytic toolkit).

  2. IV.1. The free table, its on-shell image and spectator tests; correlation tables, null and quotient field spaces, real conjugation and signed-permutation symmetry, column systems and generic configurations.

  3. VI.2 (free Maxwell stress tensor). For simple nonabelian \(\mathfrak{g}\):

  4. VI.1. Scaling sequences of radii, stress normalizations and degreewise complex-linear frame changes.

  5. VII.1. Rig.zerothProfile: the order-zero coefficient of the lattice perturbative profile converges to the tested free reading. It supplies the lattice limit in VII.3.r.

  6. VIII.1.

  7. VII.2. For VII.3.7:

  8. V.3 and VII.2. The selected trace field has a positive-energy spectral measure \(\mu\), its two-point function is the Laplace transform of \(\mu\), and its Wightman two-point distribution is the corresponding momentum-space integral. Exponential damping is integrable for every positive time (TraceSpectrum, traceSpectrum_exists). This is used only in VII.3.7(b).

  9. I.1. The lattice, the Wilson action and the Haar measure (for §4.7).

Standard mathematics.

Imported constants and assumptions.

3. Strategy

The reflected covariance couples different jet orders. Hobson's formula identifies the field of an order-\(j\) multipole with \(s_j\) times its Kelvin transform. Translating the polynomial multipole by the negative inverted separation and rescaling its Taylor components produces a triangular change of probes whose response has no Taylor terms below order \(j\) at the second point. The reflected covariance becomes diagonal, with positive pivots \(\pi_j\). Wick recursion extends this to polynomial monomials, with factorial norms \(\alpha !d^{\alpha}\).

Strict positivity detects every nonzero on-shell label. It identifies the free quotient and gives injective two-point columns. At degree four, the explicit quadratic basis reduces buffer rigidity to a coefficient calculation: two-point stationarity gives weighted antisymmetry, while signed permutations give symmetry or zero. Compactly supported tests convert this rigidity into finitely many equations with invertible derivative. The implicit function theorem solves them along the scaling sequence.

Independently, finite Wick profiles have expansions in integer powers of distance and nonnegative powers of its logarithm. Their real parts cannot have the trace field's leading term \(Ar^{- 8} (\log (1 / r))^{- 2}\). Comparison along \(r = 2^{- n}\) proves the contradiction. The spectral representation transfers equality of tested Wightman two-point functions to equality of the Euclidean profiles.

4. Steps

4.1 The Wick algebra

Lemma VII.3.a (Wick recursion). Proposition VII.3.3. Proof sketch.

Lean: Pair.remove, Pair.remove_mon, bijections in Match.lean and Occurrences.lean; byAnnihilators, wick_rotate, wick_bar.

Lemma VII.3.b (two-point adjoint relation). For \(p \in \text{Tow}_n\) and \(x \ne y\), let \(D^{x, y}_p\) be the derivation of \(\mathcal{W}\) sending each variable at \(y\) to its free covariance with \(\ell_p\) at \(x\) (crash). For all \(P, Q \in \mathcal{W}\) and \(q \in \text{Tow}_n\):

Proof sketch. Apply VII.3.a with \(n = 2\): the sum over \(k \ne i\) has a single term. On linear polynomials, the resulting contraction is exactly \(B_{x - y} (p, q)\). Lean: join_crash, crash_relation, crash_test.

Lemma VII.3.c (exhaustion by towers). For every \(P \in \mathcal{W}\) there is \(n_0 \geqslant 0\) with the following property. For all \(n \geqslant n_0\) and every real basis \((b_i)_{1 \leqslant i \leqslant N}\) of \(\text{Tow}_n\), one has \(P = \Phi_b (p)\) for some \(p \in \mathbb{C} [Y_1, \ldots, Y_N]\). Proof sketch.

Lean: from_large, from_basis.

4.2 The tower Gram form

Lemma VII.3.d (Hobson's formula in \(\mathbb{R}^4\)). Let \(S \in \text{Sol}_j\) and \(x \ne 0\). Let \(S (\partial)\) be the differential operator with symbol \(S\), normalized with Taylor coefficients (a factor \(1 / j!\)). Then

\(\displaystyle S (\partial) \Gamma (x) = (- 2)^j \hspace{0.17em} |x|^{- 2 j} \hspace{0.17em} S (x) \hspace{0.17em} \Gamma (x) .\)

Proof sketch. Induct on \(j\). In the step from \(j\) to \(j + 1\), take \(S \in \text{Sol}_{j + 1}\) and let \(S_a \in \text{Sol}_j\) be its partial contraction with \(e_a\). Taylor normalization gives \(S (\partial) = \frac{1}{j + 1} \sum_a \partial_a S_a (\partial)\) (Sol.freeze). Apply the induction hypothesis and differentiate:

Lean: shell_green.

Lemma VII.3.e (the field of a multipole is a Kelvin transform). For \(P \in \text{Sol}_j\) and \(z \ne 0\), the curvature field \(R_P\) produced by \(P\) satisfies

\(\displaystyle R_P (z) = s_j \hspace{0.17em} |z|^{- 4} \hspace{0.17em} (\mathcal{R}_z \otimes \mathcal{R}_z) \hspace{0.17em} P (\check{z}) = s_j \hspace{0.17em} |z|^{- 2 j - 4} \hspace{0.17em} (\mathcal{R}_z \otimes \mathcal{R}_z) \hspace{0.17em} P (z), \qquad \check{z} = z / |z|^2 .\)

That is, \(R_P = s_j \hspace{0.17em} I^{\ast} P\), the pull-back of the 2-form \(P\) under the inversion \(I\), whose differential is \(|z|^{- 2} \mathcal{R}_z\). Proof sketch.

\(\displaystyle D_P (x) = - 2 (j + 1) (- 2)^j (4 \pi^2)^{- 1} |x|^{- 2 (j + 2)} \hspace{0.17em} \iota_x P (x)\)

(divP_value).

Lean: curl_kelvin, reply_kelvin, reply_radial, strength_sign.

Lemma VII.3.f (shifted towers). For \(x \ne 0\), \(j < n\) and \(P \in \text{Sol}_j\), define \(\sigma_x (P) := (Q_0, \ldots, Q_j, 0, \ldots) \in \text{Tow}_n\) with

\(\displaystyle Q_k := (s_j / s_k) \hspace{0.17em} T_k [P] (- \check{x}) .\)

Here \(T_k [P] (a) \in \text{Sol}_k\) is the order-\(k\) Taylor coefficient of \(P\) at \(a\). In particular \(Q_j = P\). For every \(y \ne 0\), the total field of the shifted tower is

\(\displaystyle \sum_k R_{Q_k} (y) = s_j |y|^{- 4} (\mathcal{R}_y \otimes \mathcal{R}_y) P (\check{y} - \check{x}) .\)

It vanishes to order \(j\) at \(y = x\):

\(\displaystyle T_k \left[ \left. \sum R_{Q_i} \right] (x) = 0 (k < j), \qquad T_j \left[ \sum R_{Q_i} \right] (x) = s_j |x|^{- 2 (j + 2)} \hspace{0.17em} \mathcal{R}_x P. \right.\)

Proof sketch.

Lean: shifts, towards, field_expands, fields_shift, signal_kelvin, invert_line, signal_lower, signal_top, newLeg.

Lemma VII.3.g (diagonal Gram form in shifted towers). For \(x \ne 0\), \(i, m < n\), \(P \in \text{Sol}_i\) and \(Q \in \text{Sol}_m\):

\(\displaystyle B_x (\sigma_x P, \sigma_x Q) = \left\{ \begin{array}{ll} \pi_i (x) \langle P, Q \rangle & i = m,\\ 0 & i \ne m. \end{array} \right.\)

Hence \(B_x (\sigma_x p, \sigma_x q) = \sum_i \pi_i (x) \langle p_i, q_i \rangle\), where \(\sigma_x p := \sum_i \sigma_x (p_i)\). Proof sketch.

Lean: slide_sum, slide_below, slide_top, newLeg_below, newLeg_diag, shiftAll_gram.

Lemma VII.3.h (VII.3.2(a)). For every \(z \ne 0\) and \(n \geqslant 0\), \(B_z\) is symmetric and positive definite on \(\text{Tow}_n\). Proof sketch.

Lean: scan_sym, towerB_sym, shiftAll_pos, shiftDual, towerB_pos.

4.3 From towers to all Wick polynomials

Lemma VII.3.i (Fock orthogonality). Let \(b\) be as in VII.3.2(b), and let \(\alpha, \beta \in \mathbb{N}^N\). Then, without conjugation,

\(\displaystyle \left\langle \Phi_b (Y^{\alpha}) (x) \hspace{0.17em} (\mathcal{R}_{x - y} \Phi_b (Y^{\beta})) (y) \right\rangle_0 = \delta_{\alpha \beta} \hspace{0.17em} \alpha ! \hspace{0.17em} d^{\alpha} .\)

Proof sketch.

This is the factorial orthogonality of Wick monomials. Lean: aligned_script.

Lemma VII.3.j (VII.3.2(b)). The monomials \(\Phi_b (Y^{\alpha})\) are real (bar_mono). Sesquilinear extension of VII.3.i then gives \(\langle \Phi_b p, \Phi_b q \rangle_{x, y} = \sum_{\alpha} p_{\alpha} \overline{q_{\alpha}} \alpha !d^{\alpha}\). Lean: duplex_script.

Lemma VII.3.k (VII.3.1). Proof sketch.

Lean: reflected_positive, script_pos, reflected_symm, reflected_pd.

4.4 Faithfulness and covariance of the free table

Lemma VII.3.l (VII.3.4(a)). Proof sketch.

Lean: no_residual, free_null, exactField.

Lemma VII.3.m (VII.3.4(b), (c)). Proof sketch of (b).

Proof sketch of (c).

Lean: scanner_inj, recorded, vane.

Lemma VII.3.n (VII.3.5). Proof sketch.

Lean: table_spin, ride.

4.5 Degree four: rigidity and the position equations

Lemma VII.3.o (coordinates on \(E_4\)). Assume that \(\mathfrak{g}\) is simple and nonabelian. Let \(e\) be the unit vector of coordinate axis \(0\). The basis and dual functionals below are from VI.2.

Basis. Quadratic products below are Wick-ordered. The space \(E_4\) has the basis \(\{\star_c \}\), \(c \in \{\varnothing\} \cup \{ \text{faces} \}\), where:

Dual functionals. The dual basis is \(\lambda_{\varnothing} (v) = \langle v \rangle_0\) and

\(\displaystyle \lambda_{\{P, Q\}} (v) = \frac{m (P, Q)}{\dim \mathfrak{g} \hspace{0.27em} \kappa (P) \kappa (Q)} \hspace{0.17em} \left\langle v (e) \hspace{0.17em} \langle F_P, F_Q \rangle (0) \right\rangle_0,\)

with \(m (P, Q) = \tfrac{1}{2}\) if \(P = Q\) and \(1\) otherwise, and \(\kappa (P) = \kappa_0 (P) = \mp \pi^{- 2}\).

Two-point formulas. For a quadratic index \(c = \{P, Q\}\), put \(\rho (c) = \dim \mathfrak{g} \hspace{0.17em} \kappa (P) \kappa (Q) / m (P, Q) \ne 0\). For every \(v \in E_4\):

Lean: price_tag, backwards, rate_ne, big_weight; inputs stand, separates.

Lemma VII.3.p (VII.3.6(a)). Let \(B \in \text{Buf}_4\) satisfy the one- and two-point conditions of VII.3.6(a). Put \(\text{coeff} (c, c') = \lambda_c (B \star_{c'})\). Then every coefficient vanishes, and so \(B = 0\). Proof sketch.

\(\displaystyle \rho (c') \hspace{0.17em} \text{coeff} (c', c) + \rho (c) \hspace{0.17em} \text{coeff} (c, c') = 0\)

(cancel).

Lean: coeff_barren, frozen_four, first_frozen, orient_transfer, wearer.

Lemma VII.3.q (VII.3.6(b)). Proof sketch.

Lean: square_of_frozen.

Lemma VII.3.r (VII.3.6(c)).

The data.

The hypothesis on \(\varphi\). It consists of four conditions:

Proof sketch.

The full degree-four position conclusion consists of:

Lean: residency, adjusted, testedMean, finite, readJet; packaged by four_on (VIII.1).

4.6 Exclusion of Wick profiles (VII.3.7)

Lemma VII.3.s (expansions; VII.2). Let \(\mathcal{B}\) be the real span of \(r^a (\log r)^k\) for \(a \in \mathbb{Z}\), \(k \in \mathbb{N}\), on \(r > 0\). Every integrated mode kernel, and each of the real and imaginary parts of every Wick profile, has an expansion to all orders in \(\mathcal{B}\): for every \(N \in \mathbb{Z}\) there is \(u \in \mathcal{B}\) with \(f - u = O (r^N)\) as \(r \to 0^+\). Proof sketch.

\(\displaystyle M_{j, k} (s) = \int_0^{\infty} t^{j + 2} E^{k - 1} e^{- sE} dt, \qquad E = \sqrt{\textsf{m}^2 + t^2} .\)

Lean: Expands, euler, derived, basic_expands, moment_expands, Track.moments_expand, real_profile, primitives.

Lemma VII.3.t (no power-log law; VII.2). Let \(f : (0, \infty) \to \mathbb{R}\) and \(A \in \mathbb{R}\). Suppose \(f = u + O (1)\) with \(u \in \mathcal{B}\) as \(s \to 0^+\), and \(s^8 \log^2 (1 / s) f (s) \to A\). Then \(A = 0\). Proof sketch.

Lean: no_powerlog.

Lemma VII.3.u (VII.3.7). Proof sketch.

Lean: euclidean_excludes, polynomial_excludes, publicTrace.

4.7 The formal lattice perturbation theory

Definitions VII.3.v. Fix a finite periodic four-dimensional lattice. The perturbation theory uses inversion-odd link potentials,

\(\displaystyle A (- x - e_{\mu}, \mu) = - A (x, \mu) .\)

Its definitions are:

Gauge-fixed linearization and Gaussian measure.

The links are \(U = \exp (A)\). The action \(S\) is the finite-regulator Wilson term plus the quadratic Feynman gauge-fixing term \(\tfrac{1}{4} \sum_x | \text{div} \hspace{0.17em} A (x) |^2\) (action).

The Haar Jacobian \(J\) and the normalized Faddeev–Popov determinant \(\text{FP}\) are smooth at \(0\), with \(J (0) = \text{FP} (0) = 1\). The latter is the determinant of the gauge-condition derivative on inversion-even site parameters fixed to zero at the root, divided by its flat value (Orb.FP). The matrix determinant, logarithmic determinant and inverse estimates come from the matrix calculus listed in the header.

The centered Gaussian probability has covariance equal to the inverse Hessian of this gauge-fixed action at \(0\) (Orb.data).

Formal series in the coupling.

Write \(S_j (A) = (j!)^{- 1} D^j S (0) [A, \ldots, A]\). After the rescaling \(A \mapsto gA\), the formal weight is

\(\displaystyle w (g, A) = \exp \hspace{-0.17em} \left( - \sum_{i \geqslant 1} g^i S_{i + 2} (A) \right) (J \hspace{0.17em} \text{FP}) (gA),\)

where the last factor denotes its formal Taylor series at \(g = 0\).

\(\mathbb{E}_0\) is the Gaussian expectation, taken coefficientwise.

The normalized expectation is \(\langle f \rangle_{\text{pert}} =\mathbb{E}_0 (wf) /\mathbb{E}_0 (w)\) (Orb.weight, Orb.expect, Orb.normalize).

Every coefficient is continuous with at most polynomial growth, hence Gaussian-integrable. Since \(w (0, A) = 1\), the denominator has constant coefficient \(1\) and is invertible as a formal series.

Lattice readers of labels. Substitute the lattice jet series plus \(iZ\) into each polynomial label, where \(Z\) is an independent centered real Gaussian jet with the prescribed covariance, and average in \(Z\) separately within each insertion. Multiply these averaged insertions and take the normalized formal expectation. The separate auxiliary averages implement Wick subtraction without auxiliary contractions between different insertions (Full.column, Full.amplitude).

Profiles. The profile is the lattice Riemann sum

\(\displaystyle \sum_o h^{4 r} f (\iota_h (o)) \hspace{0.17em} \text{amplitude} (o), \qquad h = L^{- m}, \text{side } 4 L^{m^2 + m + 1},\)

where \(o\) ranges over \(r\) lattice sites and \(\iota_h\) takes the chosen integer lifts relative to the quarter-side origin and multiplies them by \(h\) (Full.profile).

The order-zero profiles enter VII.3.r through the convergence theorem of VII.1. Their higher coefficients enter the formal position identities and the degree-four parity assertion assembled in VIII.1.

Orb.finite_agreement shows that retaining the interaction series, density series and insertions only through coupling order \(k\) preserves every normalized coefficient through order \(k\). The required action Taylor coefficients extend through order \(k + 2\).

5. Where the difficulty lies

The main calculation is the simultaneous diagonalization of all jet orders (VII.3.d–h). The Taylor normalization and the sum over ordered form indices determine the Kelvin constants and positive pivots; changing either convention changes the formulas. Recentring at the inverted separation eliminates all lower Taylor coefficients, so positivity follows from a triangular change of basis.

At degree four, the two-point form is indefinite before reflection. Rigidity uses its nonzero diagonal rates together with signed-permutation symmetry: equal reflection characters force symmetric coefficients and equal rates, while the two-point condition forces weighted antisymmetry. Full rotational equivariance of the buffer is unnecessary.

The position comparison is leading-order only. Its formal identities at every coupling order follow from the local implicit equation at a finite regulator; they give no all-order comparison with the interacting theory. The free quotient and its injective columns likewise give no interacting rank bound without the comparison estimates of III.11.

The non-freeness argument is independent of Gram positivity. Its analytic input is the power-log expansion of the massive Laplace moments, including the remainder estimates needed for integration. It excludes inverse-logarithmic leading behavior; logarithms with nonnegative powers already occur in free massive kernels.

6. Notation

object code name notation
on-shell \(j\)-jets (homogeneous Maxwell solutions) Sol \(\text{Sol}_j\)
towers of multipoles ProbeTower \(\text{Tow}_n\)
Wick polynomials WickPol, Slot \(\mathcal{W}=\mathbb{C} [X_s]\)
linear polynomial of a tower code, digit \(\ell_p\)
encoding in a basis of a tower encode \(\Phi_b\)
free moment Wick.moment, join \(\langle P_1 (x_1) \cdots \rangle_0\)
reflection in \(z^{\perp}\) face, pullSol, solO \(\mathcal{R}_z\)
reflected pairing duplex, aligned \(\langle P, Q \rangle_{x, y}\)
tower Gram form towerB, scan \(B_z\)
inversion; inverted point inversePoint, rr \(I (z) = \check{z} = z / \lvert z \rvert^2\)
field of a multipole reply \(R_P = s_j I^{\ast} P\)
Kelvin constant; pivot strength, price \(s_j\), \(\pi_j (z)\)
shifted tower shifts, newLeg, shiftAll \(\sigma_z\)
two-point annihilator crash, extractVar \(D^{x, y}_p\), \(\partial^{y, y'}_s\)
Fock weight tally \(\alpha ! \hspace{0.17em} d^{\alpha}\)
free-field map of labels; shell sector observable, shellSector \(\omega_d\), \(\text{sh}_d\)
field space of the free table; embedding (freeTable F).Field d, Position.E, writes \(E_d\), \(\varpi_d\)
buffers; corrected frame buffers, motor, adjusted \(\text{Buf}_d\), \(U_k (1 + B)\)
stationary buffers vanish Frozen, stationery rigidity
basis and dual basis of \(E_4\) stand, star, portLine, coeff \(\star_c\), \(\lambda_c\), \(\text{coeff} (c, c')\)
rate rate \(\rho (c)\)
test system; free readings; chart Conditions, white, chart, square \(Q\), \(w_Q\), \(\text{chart}_Q\)
measured tested readings measured, exactMean, testedMean \(m_Q (k)\)
leading position comparison FourMeters, Resident, Crest, BoardAttains (D15)
mode, track, Wick profile Mode, Track, wickProfile, atPoint \(K_{\textsf{m}, q, h}\), \(W\)
span of \(r^a (\log r)^k\); full expansion board, Expands, AtOrder \(\mathcal{B}\)
trace two-point function traceTwo, euclidean \(G_{\Theta} (r)\)
perturbative weight, expectation, normalization Orb.weight, expect, normalize \(w\), \(\mathbb{E}_0\), \(\langle \cdot \rangle_{\text{pert}}\)
lattice perturbative profile Full.profile, Position.finite \(\pi^{(n)}_Q (L, m)\)

7. What remains unexplained

  1. Image versus intrinsic invariant algebra. The statements use \(\text{sh}_d = \omega_d (\text{Label}_d)\) exactly. An intrinsic characterization as all gauge-invariant on-shell polynomials of weight at most \(d\) would require a separate surjectivity argument; no such characterization is needed here.

  2. Interacting rank estimates. The free columns enter the construction of determinant trials in III.11, and the free-null description enters the upper-rank comparison. The estimates that turn these into interacting rank bounds and compatible interacting field spaces are inputs from III.11 and III.12; their proofs are outside this unit's verification.

  3. Continuum Wick subtraction. The auxiliary imaginary Gaussian average in VII.3.v is the exact lattice definition. Its comparison with continuum Wick-ordered correlations uses VII.1's order-zero profile theorem. The underlying regulator-removal and Wick-subtraction estimates are accepted as VII.1 inputs; their proofs are not checked here.

  4. Further non-freeness statements. The larger trace-exclusion package also contains exclusions involving embeddings into Maxwell Fock spaces and invariant averages. Their proofs are not covered by VII.3.7, which treats finite mass-shell contraction profiles and their tested two-point identities.

← VII.2 The trace of the stress tensor and the trace anomaly (notes)Part VII·Unit outlineVII.4 Operator product expansion and contact extensions (notes) →