Home·The theorem·The proof in one reading·Architecture·Units·Manuscript·Glossary·FAQ·Work items·Data·Method·Provenance·Changes
← V.2 Osterwalder-Schrader reconstruction of a Wightman theory (notes)Part V·Unit outlineVI.1 Ward identities, the rotation action and the conserved stress tensor (notes) →

V.3. Reconstruction in every phase — detailed notes

Contents

1. What this unit proves 1

2. What it uses 5

3. Strategy 7

4. Steps 7

4.1 Slab bounds and the family of reconstructions 7

4.2 Continuity in the phase and barycenter 9

4.3 Surjectivity of the boundary isometry 11

4.4 Low-energy bound and heat decay 12

4.5 Every phase 14

4.6 Irreducibility 15

4.7 The continued stress tensor 16

4.8 Assembly 19

5. Where the difficulty lies 19

6. Notation 20

7. What remains unexplained 21

The full account of unit V.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: V, phases and reconstruction. Lean: YM/EveryPhase (34 files). Its hypotheses are recorded in YM/Spec/Closed.lean and YM/Phase/Mixture.lean. Size: 5,268 lines.

1. What this unit proves

Setting. Fix a Wilson family, an extraction and the limit table \(S = S^{\text{mix}}\), together with a record Final for them. The unit uses the following parts of it.

Notation.

Proposition V.3.1 (reconstruction in every phase). There is a family indexed by \(X\) with the following data for each \(\xi \in X\):

At \(\xi = \xi_{\ast}\) these are the data of V.1 and V.2. For every \(\xi \in X\), with no exceptional set, the following hold.

  1. Wightman axioms.

  2. Surjectivity. \(J_{\xi}\) is unitary.

  3. Euclidean vacuum. \(\mathcal{D}_{\xi} =\mathbb{C} \omega_{\xi}\).

  4. Zero energy. For every \(x \in \mathcal{H}_{\xi}\), \(P^0_{\xi} x = x\) if and only if \(x \in \mathbb{C} \Omega_{\xi}\).

  5. Low-energy bound. For every compactly supported finite sum of words \(A\) there is \(M \geqslant 0\) such that \(\mu^{\xi}_A (\{0 < p^4 \leqslant u\}) \leqslant Mu^2\) for all \(\xi \in X\) and \(u \in (0, 1]\).

  6. Heat decay. For every such \(A\) there is \(M \geqslant 0\) such that, for all \(\xi \in X\) and \(s \geqslant 1\),

\(\displaystyle 0 \leqslant \langle A \Omega_{\xi}, (e^{- sH_{\xi}} - P^0_{\xi}) A \Omega_{\xi} \rangle \leqslant Ms^{- 2} .\)
  1. Irreducibility. For a bounded complex-linear operator \(B\) on \(\mathcal{H}_{\xi}\), the following are equivalent: \(B \in \mathbb{C} \hspace{0.17em} 1\); and, for every \(a\), \(f\) and \(u, v \in \mathfrak{D}_{\xi}\), one has \(u \in \text{dom} \hspace{0.17em} \phi_{\xi} (a, f)^{\ast}\) and \(\langle \phi_{\xi} (a, f)^{\ast} u, Bv \rangle = \langle u, B \phi_{\xi} (a, f) v \rangle\).

  2. Stress tensor. The conclusions (T0)–(T6) of Proposition V.3.y hold for \(\mathfrak{W}_{\xi}\).

  3. Dependence on the phase.

In item 6, \(e^{- sH_{\xi}}\) is defined as \(J_{\xi}^{\ast} e^{- sH^{\text{OS}}_{\xi}} J_{\xi}\). Since \(J_{\xi}\) intertwines the spectral projections (V.3.b), this equals \(\int e^{- sp^4} \hspace{0.17em} dP_{\xi} (p)\). The projection subtracted in item 6 is the full zero-energy projection. Items 5 and 6 are therefore proved before item 4 and independently of it.

Status: theorem, for an arbitrary record Final. The record is constructed in the proof of the main theorem (VIII.1), so the statement is unconditional there.

Lean: all_characters, Completed (the statement; see §6 for the fields), member (weak commutant), Almanac (the data), almanac (existence).

Proposition V.3.2 (no exceptional phase). For every \(\xi \in X\) and \(x \in \mathcal{H}_{\xi}\),

\(\displaystyle P^0_{\xi} x = \langle \Omega_{\xi}, x \rangle \hspace{0.17em} \Omega_{\xi} .\)

Status: theorem (Lemma V.3.r). Items 3, 4 and 7 of V.3.1 are derived from it. Lean: all_plant, boundary_zero_energy.

Remark V.3.3 (recorded consequences, not used by the main theorem). Under V.3.1, for every \(\xi \in X\) and \(x \in \mathcal{H}_{\xi}\):

\(\displaystyle \sum_{i, j} \overline{c_i} c_j \hspace{0.17em} S^{\xi} (\theta_{O, b, a} w_i \hspace{0.17em} w_j) \geqslant 0 \qquad (c_i \in \mathbb{C}) .\)

Here \(\theta_{O, b, a}\) reverses the word, reflects its points across the hyperplane, and acts on labels by transporting the \(a\)-axis reflection through \(O\).

Lean: fixedHeat, exactly_fixed, recognizes, allPlanes.

2. What it uses

From other units.

Standard mathematics.

Imported constants and assumptions. The record Final is the only hypothesis. Its quantitative inputs are as follows.

\(\displaystyle \| \Pi (d, w)\| \leqslant C^{|w|} (|w| / r)^{Q |w|} .\)

This holds for every word \(w\) of \(|w|\) points with degrees \(\leqslant D\) and coefficients \(\leqslant R\), lying in the slab \([r, d - r]\), with points \(r\)-separated and \(0 < r \leqslant 1\).

The one-point bound is Boundary.Small.rated.

The unit's own constants are explicit:

The Fourier convention and constant are Boundary.Probe.hat_eq, Boundary.Probe.pref.

3. Strategy

V.2 reconstructs a relativistic theory at a selected phase \(\xi_{\ast}\). The same construction applies to every \(S^{\xi}\) before vacuum uniqueness is known. The norm of each phase’s slab operator is bounded by the corresponding norm for \(S^{\text{mix}}\) (V.3.a): repeated reflection gives a power bound, and the fixed norm of the character disappears on taking roots. This gives polynomial continuation bounds uniform in \(\xi\). Character continuity at imaginary times propagates to complex times by Vitali’s theorem, then to boundary distributions by a Taylor formula with an integrable remainder. Integration over phases commutes with both passages (V.3.d–V.3.h).

Vacuum uniqueness holds almost everywhere by the phase decomposition (V.3.c). To extend it to every phase, fix a compactly supported finite sum of words \(A\) and set

\(\displaystyle r_{\xi} (s) = \langle A \Omega_{\xi}, e^{- sH_{\xi}} A \Omega_{\xi} \rangle - | \langle \Omega_{\xi}, A \Omega_{\xi} \rangle |^2 .\)

Locality makes the vacuum commutator vanish at large spatial separation for \(|t| \leqslant 1\). Its time-averaged odd part is the spatial Fourier transform of a positive measure, weighted by \((1 - \cos E) / E\). A momentum bump of radius \(u\) bounds its shell mass by \(O (u^2)\): the bump contributes \(u^3\), while the weight contributes \(u\). Dyadic summation and a Laplace transform give decay \(Ms^{- 2}\), uniformly in \(\xi\), after subtracting the full zero-energy projection (V.3.k–V.3.o). Almost everywhere this is the vacuum projection. Since \(r_{\xi} (s)\) is continuous, the same estimate holds everywhere by full support of \(\nu\). Letting \(s \to \infty\) identifies the zero-energy projection with the vacuum projection at every phase (V.3.q–V.3.r). This needs no continuity of the zero-energy projection itself.

Separately, positive-height Euclidean vectors are images of Wightman vectors with exponentially weighted frequency tests, so \(J_{\xi}\) is onto (V.3.i–V.3.j). Vacuum uniqueness therefore holds on both sides. A Liouville argument gives irreducibility. Finally, continuation of the Euclidean Ward identities gives the local stress-charge commutator (V.3.u–V.3.y). The removal of the exceptional set uses uniform low-energy control from locality, rather than a mass gap.

4. Steps

4.1 Slab bounds and the family of reconstructions

Lemma V.3.a (slab operators of a phase are dominated). Fix \(\xi \in X\). Let \(\mathcal{K}\) be any OS space of \(S^{\xi}\) for the time axis, and \(\Pi_{\xi}\) any family of slab operators on it. Then for every thickness \(d > 0\) and every word \(w\) of pairwise distinct points in the slab \((0, d)\),

\(\displaystyle \| \Pi_{\xi} (d, w)\| \leqslant \| \Pi (d, w)\|,\)

where \(\Pi\) are the slab operators of \(S^{\text{mix}}\) on \(\mathcal{H}^{\text{OS}}\). Here \(\Pi (d, w)\) maps the OS vector of a positive-time word \(v\) to that of \(w\) followed by \(v\) translated by \(d\).

Sketch.

  1. Let \(u = \Pi (d, w)\) and \(D > \|u\|\). For a positive-time word \(v\), let \(X (v)\) be the slab operator of \(v\) restricted to the ground space \(\mathcal{D}\).

  2. The compression of the doubled word \(\theta v \hspace{0.17em} v\) is \(X (v)^{\ast} X (v) \in \mathfrak{A}\). Let \(r_n\) be \(v\) with \(n\) reflected pairs of copies of \(w\) inserted. Then \(X (r_n) = (u^{\ast} u)^n X (v)\).

  3. Hence the diagonal Gram entry of \(r_n\) in \(S^{\xi}\) satisfies

\(\displaystyle |S^{\xi} (\theta r_n \hspace{0.17em} r_n) | = | \chi_{\xi} (X (r_n)^{\ast} X (r_n)) | \leqslant \| \chi_{\xi} \| \hspace{0.17em} \|X (v)\|^2 D^{4 n} .\)
  1. The power argument (iterated Cauchy–Schwarz on the span of OS vectors) turns this into an operator on \(\mathcal{K}\) of norm at most \(D\) that inserts \(w\). The fixed factor \(\| \chi_{\xi} \| \hspace{0.17em} \|X (v)\|^2\) disappears in the limit.

  2. Slab operators are determined by their action on OS vectors, so this operator is \(\Pi_{\xi} (d, w)\). Finally let \(D \downarrow \|u\|\).

Lean: sameBound, with x_train, buys, actual, and OS.producer.

Lemma V.3.b (the family of reconstructions). For every \(\xi\) there are:

At \(\xi_{\ast}\) these are the data of V.1–V.2. Moreover:

Sketch.

Lean: storeys, almanac, Almanac.priced, Attachment.projections, Attachment.with_heat. (proof read for these; contour, compatible, establishing and Work.B02.answer also read; underlying analytic inputs from V.2)

Lemma V.3.c (unique Euclidean vacuum for almost every phase; from V.1). For \(\nu\)-almost every \(\xi\), \(\mathcal{D}_{\xi} =\mathbb{C} \omega_{\xi}\).

Sketch (a variance identity). For a positive-time word \(b\), let \(\Psi_{\xi} (b) \in \mathcal{H}^{\text{OS}}_{\xi}\) be its OS vector and \(Q_{\xi}\) the projection onto \(\mathcal{D}_{\xi}\).

  1. \(\|Q_{\xi} \Psi_{\xi} (b)\|^2 = \lim_{s \to \infty} S^{\xi} (\theta b_s \hspace{0.17em} b_s)\), where \(b_s\) is \(b\) translated by \(s\) in time. The terms are character values, continuous in \(\xi\), nonincreasing in \(s\) and dominated by the value at \(s = 0\).

  2. By dominated convergence and the barycenter identity, \(\int \|Q_{\xi} \Psi_{\xi} (b)\|^2 d \nu = \|Q \Psi (b)\|^2\).

  3. On the other hand \(S^{\xi} (b) = \chi_{\xi} (M_b)\) and \(M_b \omega = Q \Psi (b)\). Multiplicativity of characters gives \(\int |S^{\xi} (b) |^2 d \nu = \int \chi_{\xi} (M_b^{\ast} M_b) \hspace{0.17em} d \nu = \|Q \Psi (b)\|^2\).

  4. Hence the nonnegative function \(\|Q_{\xi} \Psi_{\xi} (b)\|^2 - |S^{\xi} (b) |^2 = \|Q_{\xi} \Psi_{\xi} (b) - S^{\xi} (b) \omega_{\xi} \|^2\) has integral zero, so \(Q_{\xi} \Psi_{\xi} (b) \in \mathbb{C} \omega_{\xi}\) almost everywhere.

  5. Let \(b\) run over a countable family of words whose OS vectors are total in every OS space (Work.B14.universal).

Lean: Almanac.almostEvery, Cluster.lifts, Cluster.exp_real.

4.2 Continuity in the phase and barycenter

Lemma V.3.d (polynomial bound, uniform in the phase). For every \(n\) and \(a \in \mathbb{F}^n\) there are \(m \in \mathbb{N}\) and \(C_0 \geqslant 0\) with the following property. For all \(\xi\), all real configurations \(x\), and all heights \(y\) with \(y_j - y_i \geqslant k\) for \(i < j\), where \(k \in (0, 1]\),

\(\displaystyle | \textsf{F}^{\xi}_n (a ; \hspace{0.17em} t (x) + iy, \hspace{0.17em} \vec{x}) | \leqslant C_0 \hspace{0.17em} k^{- m} .\)

Sketch.

Lean: power. (proof read; the matrix-element representation of the continuation comes from V.2)

Lemma V.3.e (strict heights). For all \(n\), \(a\), real \(x\) and \(y_1 < \ldots < y_n\), the map \(\xi \mapsto \textsf{F}^{\xi}_n (a ; t (x) + iy, \vec{x})\) is continuous and its \(\nu\)-integral is \(\textsf{F}^{\text{mix}}_n (a ; t (x) + iy, \vec{x})\).

Sketch.

  1. At purely imaginary times \(z_j = iy_j\), \(\textsf{F}^{\xi}_n\) is \(S^{\xi}\) of a word of distinct points, hence a character value. It is therefore continuous in \(\xi\), with barycenter \(S^{\text{mix}}\).

  2. Along the complex line \(c \mapsto iy + c \hspace{0.17em} t (x)\), for \(c\) in a strip \(| \text{Im} c| < \delta\) (where the heights stay ordered with gaps \(\geqslant k / 2\)), the functions are holomorphic and bounded uniformly in \(\xi\) by V.3.d.

  3. By Vitali's theorem, convergence along nets \(\xi' \to \xi\) on the imaginary segment through \(0\) propagates to the whole strip.

  4. The integral over \(X\) is holomorphic in \(c\). It agrees with the unselected function on that segment, hence on the strip by the identity theorem. Evaluate at \(c = 1\).

Lean: imaginary_evaluation, strict_evaluation, integral_holomorphic, propagate_limit.

Lemma V.3.f (tested strict heights). For \(y_1 < \ldots < y_n\) and \(g \in \mathcal{S} ((\mathbb{R}^4)^n)\), define \(\psi^{\xi} (a, y, g) = \int \textsf{F}^{\xi}_n (a ; t (x) + iy, \vec{x}) \hspace{0.17em} g (x) \hspace{0.17em} dx\). Then \(\xi \mapsto \psi^{\xi} (a, y, g)\) is continuous, and \(\int \psi^{\xi} \hspace{0.17em} d \nu = \psi^{\text{mix}}\).

Sketch. Dominated convergence with dominator \(C_0 k^{- m} |g|\) (V.3.d), and Fubini.

Lean: eager.

Lemma V.3.g (continuity and barycenter of the Wightman distributions). Item 9 of V.3.1, first part.

Sketch. Let \(\rho = (0, 1, \ldots, n - 1)\) and \(D = - i \sum_{j = 1}^n (j - 1) \partial_{t_j}\) on joint test functions. The Cauchy–Riemann equations and integration by parts give \(\frac{d}{dt} \psi (a, y + t \rho, g) = \psi (a, y + t \rho, Dg)\).

  1. With \(m\) as in V.3.d, the boundary value has the Taylor representation

\(\displaystyle W^{\xi}_n (a ; g) = \sum_{k = 0}^m \frac{(- 1)^k}{k!} \hspace{0.17em} \psi^{\xi} (a, \rho, D^k g) - \int_0^1 \frac{(- t)^m}{m!} \hspace{0.17em} \psi^{\xi} (a, t \rho, D^{m + 1} g) \hspace{0.17em} dt.\)
  1. By V.3.d the integrand is bounded by \(C_0 \|D^{m + 1} g\|_{L^1}\), uniformly in \(t\) and \(\xi\).

  2. Continuity then follows from V.3.f and dominated convergence in \(t\).

  3. For the barycenter: \(\psi^{\xi} (a, y, g) \to W^{\xi}_n (a ; g)\) as \(y \to 0\) among ordered heights, with a bound uniform in \(\xi\) and \(y\) given by the same representation. Integrate over \(X\) by dominated convergence, and use V.3.f at each \(y\).

Lean: word_continuity, barycenter, Waves.desired, Waves.meant. The derivative convention is Boundary.Waves.clock.

Lemma V.3.h (inner products). For finite sums of words \(A, A'\):

Sketch. \(\langle \phi_{\xi} (w) \Omega_{\xi}, \phi_{\xi} (w') \Omega_{\xi} \rangle = W^{\xi} (w^{\dagger} w')\), with the same star operation for all \(\xi\). Apply V.3.g, then use compactness of \(X\).

Lean: word_pair, sum_pair, word_bound.

4.3 Surjectivity of the boundary isometry

Lemma V.3.i (Euclidean vectors at positive heights are in the range). Fix \(n \geqslant 1\), \(\xi \in X\), labels \(a \in \mathbb{F}^n\), heights \(0 < y_1 < \ldots < y_n\) and \(p \in (0, 1)\) with \(4 p < y_1\) and \(4 p < y_{j + 1} - y_j\) for \(1 \leqslant j < n\).

\(\displaystyle R_{\xi, \ell} (z, \vec{x}) = e^{- (\text{Im} z - \ell) H^{\text{OS}}_{\xi}} e^{i \text{Re} zH^{\text{OS}}_{\xi}} U^{\text{OS}}_{\xi} (\vec{x}) .\)

For a real configuration \(q\), write \(z_j = q_j^4 + iy_j\) and define

\(\displaystyle \Psi^{\xi}_p (a ; q + iy) = R_{\xi, p} (z_1, \vec{q}_1) B_{\xi, p} (a_1) \left( \prod_{j = 2}^n R_{\xi, 2 p} (z_j - z_{j - 1}, \vec{q}_j - \vec{q}_{j - 1}) B_{\xi, p} (a_j) \right) \omega_{\xi},\)

where the factors are ordered by increasing \(j\) from left to right; the product is empty when \(n = 1\).

\(\displaystyle \int g (q) \hspace{0.17em} \Psi^{\xi}_p (a ; q + iy) \hspace{0.17em} dq = J_{\xi} \hspace{0.17em} \Phi_{\xi} (a ; g_y) .\)

Here \(\Phi_{\xi} (a ; \cdot)\) is the joint-test extension from V.3.b. Compact support of \(h\) ensures that the exponential multiplier preserves the Schwartz class.

Sketch.

  1. Call the two sides \(v\) and \(J_{\xi} \Phi\). Their squared norms and their inner product, \(\|v\|^2\), \(\langle v, J_{\xi} \Phi \rangle\) and \(\|J_{\xi} \Phi \|^2\), are each expressed through partial boundary values of \(W^{\xi}\) at heights \(y\).

  2. On tests of compact frequency support these partial boundary values equal \(W^{\xi}\) tested against the damped test functions, so the three numbers coincide and \(\|v - J_{\xi} \Phi \| = 0\).

  3. Density of compact-frequency tests and continuity in \(g\) extend the membership in \(\text{ran} \hspace{0.17em} J_{\xi}\) to all Schwartz \(g\).

  4. Testing against all \(g\), and using continuity in \(q\), gives the pointwise statement. The range is closed since \(J_{\xi}\) is an isometry.

Lean: vector_cooled, vector_in, shot_in, boundary_tilt, chilled. (proof read; V.2's partial-boundary identities statement only)

Lemma V.3.j (surjectivity). Item 2 of V.3.1.

Sketch.

  1. The empty word gives \(\omega_{\xi} = J_{\xi} \Omega_{\xi}\). Take a nonempty positive-time Euclidean word with strictly increasing times. Its OS vector is a vector \(\Psi^{\xi}_p\) of V.3.i (real times \(0\), heights equal to the Euclidean times), so it lies in the range.

  2. Take a general positive-time word. By symmetry its points may be reordered so that the times are nondecreasing, say \(r\). Then \(r + u \rho\) is strictly increasing for \(u > 0\).

  3. Its OS vector is the limit, as \(u \downarrow 0\), of OS vectors in the range, by continuity of OS vectors in the points. It therefore lies in the range, because the range is closed.

  4. These vectors span a dense subspace, so \(\text{ran} \hspace{0.17em} J_{\xi}\) is dense and closed.

Vacuum uniqueness is not used.

Lean: onto, ordered_covered, bright_shot, reaches.

4.4 Low-energy bound and heat decay

Lemma V.3.k (time-averaged commutator). Let \(\rho, \sigma\) be finite positive measures on \(\mathbb{R}^4\) supported in \(\bar{V}_+\), with \(\rho (\mathbb{R}^4) + \sigma (\mathbb{R}^4) \leqslant M\). Put

\(\displaystyle \hat{\rho} (\vec{x}, t) = \int e^{i (\vec{p} \cdot \vec{x} + p^4 t)} \hspace{0.17em} d \rho (p), \qquad a = \hat{\rho} (\cdot) - \hat{\sigma} (- \hspace{0.17em} \cdot), \qquad D (\vec{x}) = \int_0^1 \frac{a (\vec{x}, t) - a (\vec{x}, - t)}{2 i} \hspace{0.17em} dt.\)

Then \(|D| \leqslant M\). Write \(\mathcal{F}f (x) = \int e^{- i \vec{p} \cdot x} f (\vec{p}) \hspace{0.17em} d \vec{p}\), so \(c_3 = (2 \pi)^3\). If moreover \(D (\vec{x}) = 0\) for \(| \vec{x} | > R\), then for \(0 < u \leqslant 1\)

\(\displaystyle \rho (\{u / 2 < p^4 \leqslant u\}) \leqslant c (R, M) \hspace{0.17em} u^2, \qquad c (R, M) = 16 \pi \hspace{0.17em} |B_1 | \hspace{0.17em} |B_R | \hspace{0.17em} M / c_3 .\)

Sketch.

  1. By Fubini, \(D\) is the spatial Fourier transform of \(\rho^{\sharp} + \sigma^{\sharp}\). Here \(\rho^{\sharp}\) is the image under \(p \mapsto \vec{p}\) of \(\pi (p^4) \hspace{0.17em} d \rho (p)\), with \(\pi (E) = \int_0^1 \sin (Et) \hspace{0.17em} dt\), equal to \((1 - \cos E) / E\) for \(E > 0\) and to \(0\) at \(E = 0\), and \(\sigma^{\sharp}\) is the analogous image of \(\sigma\) under \(p \mapsto - \vec{p}\).

  2. On \(\bar{V}_+\) one has \(\pi (p^4) \geqslant 0\) at all energies, and \(\pi (E) \geqslant E / \pi\) for \(0 \leqslant E \leqslant 1\).

  3. Test \(D\) against \(\mathcal{F}f\), where \(f\) is a smooth spatial-momentum bump with \(0 \leqslant f \leqslant 1\), equal to \(1\) on the ball of radius \(u\) and supported in the ball of radius \(2 u\). Then \(|\mathcal{F}f| \leqslant |B_{2 u} | = 8 |B_1 | u^3\). Since \(D\) vanishes outside \(B_R\), the pairing is at most \(8 |B_1 | |B_R | M \hspace{0.17em} u^3\).

  4. The pairing equals \(c_3 \left( \int f \hspace{0.17em} d \rho^{\sharp} + \int f \hspace{0.17em} d \sigma^{\sharp} \right)\), which is at least \(c_3 \hspace{0.17em} \tfrac{u}{2 \pi} \hspace{0.17em} \rho (\{u / 2 < p^4 \leqslant u\})\). On that shell \(| \vec{p} | \leqslant p^4 \leqslant u\), so \(f = 1\) there, and \(\pi (p^4) \geqslant u / (2 \pi)\).

  5. Thus the exponent \(2\) is \(3\) (spatial dimension, from the volume of the momentum ball) minus \(1\) (first-order vanishing of the weight).

Lean: annulus, averaged_bound, fee, two_average, one_average, pulse_positive, pulse_lower.

Lemma V.3.l (dyadic summation). Let \(\rho\) be a finite measure and \(C \geqslant 0\) with \(\rho (\{u / 2 < p^4 \leqslant u\}) \leqslant Cu^2\) for all \(u \in (0, 1]\). Then \(\rho (\{0 < p^4 \leqslant u\}) \leqslant 2 Cu^2\) for \(u \in (0, 1]\).

Sketch. Cover \((0, u]\) by the shells with \(u_n = u 2^{- n}\). Then \(\sum_n Cu^2 \hspace{0.17em} 4^{- n} \leqslant \frac{4}{3} Cu^2\); the bound \(2\) follows already from the coarser majorant \(\sum_n 2^{- n} = 2\).

Lean: dyadic.

Lemma V.3.m (locality at large distance, uniformly in the phase). Let \(A\) be a compactly supported finite sum of words, and \(r \geqslant 0\) a bound for \(|t|\) and \(| \vec{x} |\) on the supports of all its test functions. Put \(R = 4 r + 1\). Then for all \(\xi\), \(|t| \leqslant 1\) and \(| \vec{x} | > R\):

\(\displaystyle \hat{\mu}^{\xi}_A (\vec{x}, t) - \hat{\mu}^{\xi}_{A^{\dagger}} (- \vec{x}, - t) = \langle \Omega_{\xi}, [A^{\dagger}, \alpha_{(\vec{x}, t)} A] \hspace{0.17em} \Omega_{\xi} \rangle = 0,\)

where \(\alpha\) denotes translation of the test functions. Hence the function \(D\) of V.3.k, built from \(\rho = \mu^{\xi}_A\) and \(\sigma = \mu^{\xi}_{A^{\dagger}}\), vanishes for \(| \vec{x} | > R\).

Sketch.

Lean: late, distant, comm_expect, op_comm.

Proposition V.3.n (low-energy bound). Item 5 of V.3.1, with \(M = 2 c (R, M_A + M_{A^{\dagger}})\).

Sketch. Apply V.3.k to \(\rho = \mu^{\xi}_A\) and \(\sigma = \mu^{\xi}_{A^{\dagger}}\). These are supported in \(\bar{V}_+\) by the spectrum condition, with mass bounds from V.3.h and \(R\) from V.3.m. Then apply V.3.l.

Lean: light_bound.

Proposition V.3.o (heat decay with the full zero-energy projection). Item 6 of V.3.1, with \(M = 2 \max (M_A, 2 c (R, M_A + M_{A^{\dagger}}))\).

Sketch.

  1. By the spectral theorem and \(\text{supp} \hspace{0.17em} \mu^{\xi}_A \subset \{p^4 \geqslant 0\}\):

\(\displaystyle \langle A \Omega_{\xi}, (e^{- sH_{\xi}} - P^0_{\xi}) A \Omega_{\xi} \rangle = \int_{\{p^4 > 0\}} e^{- sp^4} \hspace{0.17em} d \mu^{\xi}_A \geqslant 0.\)
  1. Let \(\rho\) be a finite measure with \(\rho (\{0 < p^4 \leqslant t\}) \leqslant Ct^2\) for all \(t > 0\); for \(t > 1\) this follows from the total-mass bound. Write \(e^{- sE} = \int_E^{\infty} se^{- st} \hspace{0.17em} dt\). Fubini gives

\(\displaystyle \int_{\{p^4 > 0\}} e^{- sp^4} d \rho = \int_0^{\infty} se^{- st} \rho (\{0 < p^4 \leqslant t\}) \hspace{0.17em} dt \leqslant Cs \int_0^{\infty} t^2 e^{- st} dt = \Gamma (3) \hspace{0.17em} C \hspace{0.17em} s^{- 2} .\)

Lean: residual_pair, decay, inverse_square, bHeat_pair.

Lemma V.3.p (continuity of the heat form). For every finite sum of words \(A\) and \(s > 0\), \(\xi \mapsto \langle A \Omega_{\xi}, e^{- sH_{\xi}} A \Omega_{\xi} \rangle = \int e^{- sp^4} \hspace{0.17em} d \mu^{\xi}_A\) is continuous.

Sketch.

  1. The Fourier transform \(\hat{\mu}^{\xi}_A (p) = \langle A \Omega_{\xi}, (\alpha_p A) \Omega_{\xi} \rangle\) is continuous in \(\xi\) by V.3.h, and bounded by \(M_A\).

  2. Hence \(\int h \hspace{0.17em} d \mu^{\xi}_A\) is continuous in \(\xi\) for every Schwartz \(h\), by Parseval and dominated convergence.

  3. On \(\bar{V}_+\) outside the ball of radius \(2 N\) one has \(p^4 \geqslant N\). Multiplying \(e^{- sp^4}\) by a bump equal to \(1\) on that ball changes the integral by at most \(M_A e^{- sN}\), uniformly in \(\xi\).

  4. A uniform limit of continuous functions is continuous.

Lean: uniform_heat, char_cont, test_cont.

4.5 Every phase

Lemma V.3.q (vacuum-subtracted decay at every phase). For every compactly supported finite sum of words \(A\) there is \(M \geqslant 0\) such that \(0 \leqslant r_{\xi} (s) \leqslant Ms^{- 2}\) for all \(\xi \in X\) and \(s \geqslant 1\), with \(r_{\xi} (s)\) as in §3.

Sketch.

  1. \(r_{\xi} (s)\) is continuous in \(\xi\) by V.3.p and V.3.h; \(\Omega_{\xi}\) is the vector of the empty word.

  2. Where \(\mathcal{D}_{\xi} =\mathbb{C} \omega_{\xi}\), which is \(\nu\)-almost everywhere by V.3.c, \(J_{\xi}\) intertwines the zero-energy projections. Then \(P^0_{\xi} x = \langle \Omega_{\xi}, x \rangle \Omega_{\xi}\), so \(r_{\xi} (s) = \langle A \Omega_{\xi}, (e^{- sH_{\xi}} - P^0_{\xi}) A \Omega_{\xi} \rangle \in [0, Ms^{- 2}]\) by V.3.o.

  3. The set \(\{\xi : r_{\xi} (s) \in [0, Ms^{- 2}]\}\) is closed and of full measure. Since \(\nu\) charges every nonempty open set, it is \(X\).

Lean: proj_eq, univ_of_closed, extends_decay.

Lemma V.3.r (zero-energy projection = vacuum projection). Proposition V.3.2.

Sketch.

  1. \(e^{- sH_{\xi}} \to P^0_{\xi}\) strongly as \(s \to \infty\), so V.3.q gives \(\langle x, P^0_{\xi} x \rangle = | \langle \Omega_{\xi}, x \rangle |^2\) for \(x = A \Omega_{\xi}\).

  2. Since \(P^0_{\xi}\) is an orthogonal projection with \(P^0_{\xi} \Omega_{\xi} = \Omega_{\xi}\), this gives \(\|P^0_{\xi} x - \langle \Omega_{\xi}, x \rangle \Omega_{\xi} \|^2 = \|P^0_{\xi} x\|^2 - | \langle \Omega_{\xi}, x \rangle |^2 = 0\).

  3. The identity extends from compactly supported to Schwartz test functions by cutoffs converging in \(\mathcal{S}\) and continuity of the word vectors in the joint test function.

  4. Bounded operators that agree on all word vectors are equal.

Lean: groundvac, bHeat_limit, cluster_zero, all_plant.

Lemma V.3.s (Euclidean vacuum and Wightman axioms at every phase). Items 1, 3 and 4 of V.3.1.

Sketch.

  1. Let \(v \in \mathcal{D}_{\xi}\). By V.3.j, \(v = J_{\xi} x\) for some \(x\).

  2. The projection onto \(\mathcal{D}_{\xi}\) equals the zero-energy spectral projection of the OS translation group: both are the strong limit of \(e^{- sH^{\text{OS}}_{\xi}}\).

  3. Intertwining with V.3.2 gives \(v = J_{\xi} P^0_{\xi} x = \langle \Omega_{\xi}, x \rangle \hspace{0.17em} \omega_{\xi}\). This is item 3.

  4. Item 1 then follows from V.3.b, and item 4 from V.3.2.

Lean: every, reflection_projection, boundary_realizes, boundary_zero_energy.

4.6 Irreducibility

Proposition V.3.t (trivial weak commutant). Item 7 of V.3.1.

Sketch. The weak commutant is first rewritten with \(\phi_{\xi} (a^{\ast}, \bar{f})\) in place of the Hilbert adjoint, since the former is the restriction of the latter to \(\mathfrak{D}_{\xi}\). The converse direction is the adjoint relation. Now let \(B\) be weakly central and write \(\Omega = \Omega_{\xi}\).

  1. For word vectors \(u = \phi (w) \Omega\) and \(v = \phi (w^{\dagger}) \Omega\), and \(\text{Im} z \geqslant 0\), set

\(\displaystyle p (z) = \langle J_{\xi} u, e^{izH^{\text{OS}}_{\xi}} J_{\xi} B \Omega \rangle, \qquad q (z) = \langle J_{\xi} B^{\ast} \Omega, e^{izH^{\text{OS}}_{\xi}} J_{\xi} v \rangle .\)

Both are bounded, continuous, and holomorphic for \(\text{Im} z > 0\). On the real axis \(p (t) = \langle u, U_{\xi} (t) B \Omega \rangle\), and at \(z = is\) one has \(p (is) = \langle u, e^{- sH_{\xi}} B \Omega \rangle\).

  1. Covariance and weak centrality give \(p (t) = q (- t)\) for real \(t\).

  2. Gluing \(p\) on the upper half-plane to \(q (- \hspace{0.17em} \cdot)\) on the lower one gives a bounded continuous function, holomorphic off \(\mathbb{R}\). It is entire by Morera's theorem and constant by Liouville's theorem.

  3. Hence \(\langle u, e^{- sH_{\xi}} B \Omega \rangle = \langle u, B \Omega \rangle\) for all \(s \geqslant 0\). Letting \(s \to \infty\) and using V.3.2 gives \(B \Omega = a \Omega\) with \(a = \langle \Omega, B \Omega \rangle\).

  4. Weak centrality then gives \(\langle \phi (c) \Omega, B \phi (b) \Omega \rangle = \langle \phi (b^{\dagger}) \phi (c) \Omega, B \Omega \rangle = a \langle \phi (c) \Omega, \phi (b) \Omega \rangle\) for all words, so \(B = a \hspace{0.17em} 1\).

The proof uses V.3.2 and the reconstruction data of V.3.b. It requires neither surjectivity of \(J_{\xi}\) nor invariance of \(\mathfrak{D}_{\xi}\) under \(B\).

Lean: commutant, saturated, paste, precise, irreducible.

4.7 The continued stress tensor

Conventions.

\(\displaystyle Q_K (f, c) = \sum_b T_{4 b} \left( K^b \hspace{0.17em} (c \otimes f) \right),\)

for a time profile \(f\) and a spatial cutoff \(c\).

Lemma V.3.u (Euclidean contact identity and current jump). Fix \(\xi \in X\), distinct Euclidean points \(x_1, \ldots, x_{n + 1}\), a distinguished label \(B \in F_d\) at \(x_k\), arbitrary spectator labels, and a Killing field \(K\). Let \(\chi \in C^{\infty}_c (\mathbb{R}^4 ; \mathbb{R})\) equal \(1\) near \(x_k\), with every other \(x_j\) outside \(\text{supp} \hspace{0.17em} \chi\). Then the following integral is absolutely convergent and

\(\displaystyle - \int \sum_{a, b} \partial_a \chi (y) \hspace{0.17em} K^b_{\text{E}} (y) \hspace{0.17em} S^{\xi} (\ldots, T^{\text{E}}_{ab} (y), \ldots) \hspace{0.17em} dy = S^{\xi} (\ldots, \Sigma_K B (x_k), \ldots) + \sum_b K^b_{\text{E}} (x_k) \hspace{0.17em} S^{\xi} (\ldots, \partial_b^{\text{E}} B (x_k), \ldots) .\)

If the exterior Euclidean times are strictly ordered and \(c \in C^{\infty}_c (\mathbb{R}^3 ; \mathbb{R})\) equals \(1\) near \(\vec{x}_k\), the difference of the currents \(\sum_b \int c (\vec{y}) K_{\text{E}}^b (\vec{y}, t) S^{\xi} (\ldots, T^{\text{E}}_{4 b} (\vec{y}, t), \ldots) \hspace{0.17em} d \vec{y}\), with the stress insertion immediately after \(B\) at \(t = x_k^4 + \varepsilon\) and immediately before \(B\) at \(t = x_k^4 - \varepsilon\), tends to the right side as \(\varepsilon \downarrow 0\).

Sketch. Decompose \(V\) into a spatial rotation and a boost. The translation and rotation Ward identities give the displayed formula, including the Wick factors. Choose \(\chi (y) = c (\vec{y}) h (y^4 - x_k^4)\) with a fixed smooth time cutoff supported between the neighboring exterior times. Conservation identifies the time derivative of the integrated current with the flux through \(\nabla c\). The fundamental theorem of calculus, applied on both sides of \(x_k^4\), gives the jump. For complex exterior times, the spacelike continuation germs in (T0) give equal limits for the two spatial fluxes; bounded derivatives give joint current limits. Vitali’s theorem then continues the jump identity from imaginary exterior times. Only spatially integrated currents have a jump limit here.

Lean: euclidean_shell, Stress.pure_clock, Stress.through_jump. The cutoff condition is Table.Window.

Lemma V.3.v (Minkowski jump identity). Fix \(\xi\), a Killing field \(K\), labels at \(n + 1\) points and a distinguished index \(k\). Let \(M \in C^{\infty}_c ((\mathbb{R}^4)^{n + 1})\), \(f \in C^{\infty}_c (\mathbb{R})\) and \(c \in C^{\infty}_c (\mathbb{R}^3 ; \mathbb{R})\). For every \(x \in \text{supp} \hspace{0.17em} M\), assume that \(c = 1\) near \(\vec{x}_k\) and

\(\displaystyle |s - x_k^4 | < | \vec{y} - \vec{x}_k | \qquad (s \in \text{supp} \hspace{0.17em} f, \vec{y} \in \text{supp} \hspace{0.17em} \partial_j c, j = 1, 2, 3) .\)

Then

\(\displaystyle \sum_b \zeta_b [W^{\xi} (\ldots, a_k, \mathcal{T}_b, \ldots) - W^{\xi} (\ldots, \mathcal{T}_b, a_k, \ldots)] = \left( \int f \right) \left[ W^{\xi} (\ldots, \Sigma_K a_k, \ldots ; M) + \sum_b \zeta_b W^{\xi} (\ldots, \partial_b^{\text{E}} a_k, \ldots ; K^b_k M) \right] .\)

The left side is tested jointly against \(M\) in the original variables and \(K^b (y) c (\vec{y}) f (y^4)\) in the inserted stress variable \(y\). Thus \(\mathcal{T}_b\) denotes the Euclidean label \(T^{\text{E}}_{4 b}\) inserted immediately after, respectively before, the \(k\)-th field. On the right, \(K^b_k M (x) = K^b (x_k) M (x)\).

Sketch.

Lean: Stress.verdict, Stress.partial_jump, Stress.Jury.

Lemma V.3.w (Ward identity with arbitrary spectators). In V.3.v take \(\int f = 1\) and a product test with a compactly supported factor \(m\) at the distinguished field. The resulting identity holds with arbitrary Schwartz spectator words before and after the inserted pair. The cutoff conditions need hold only for points of \(\text{supp} \hspace{0.17em} m\).

Sketch. Cut off the spectator test functions. Apply V.3.v. Pass to the limit by continuity of \(W^{\xi}\) on joint test functions.

Lean: Stress.Schwartz_shells, Stress.compact_shells.

Lemma V.3.x (the charge generates the transformation). Fix \(\xi \in X\), \(B \in F_d\) and a Killing field \(K\). Let \(\mathcal{O} \subset \mathbb{R}^4\) be compact, \(m \in C^{\infty}_c (\mathbb{R}^4)\) with \(\text{supp} \hspace{0.17em} m \subset \mathcal{O}\), and \(f \in C^{\infty}_c (\mathbb{R}; \mathbb{R})\) with \(\int f = 1\). Let \(c \in C^{\infty}_c (\mathbb{R}^3 ; \mathbb{C})\) equal \(1\) on an open ball such that, for every \(s \in \text{supp} \hspace{0.17em} f\), every point \((\vec{y}, s)\) over its complement is strictly spacelike to every point of \(\mathcal{O}\). Then, on \(\mathfrak{D}_{\xi}\),

\(\displaystyle i [Q_K (f, c), \phi_{\xi} (B, m)] = \phi_{\xi} (\Sigma_K B, m) - \sum_{a = 1}^4 \phi_{\xi} (B, \partial_a (K^a m)) .\)

The derivative includes the affine coefficient \(K^a\). Equivalently, the last term is \(\sum_a \zeta_a \phi_{\xi} (\partial_a^{\text{E}} B, K^a m)\).

Sketch.

Lean: Stress.local_action, Stress.measure_identity, Stress.Charge, Stress.Season, Stress.Reservation.

Proposition V.3.y (the stress tensor of \(\mathfrak{W}_{\xi}\)). For every \(\xi \in X\) the following hold. Statements (T1)–(T4) hold for all \(f \in \mathcal{S} (\mathbb{R}^4)\), \(a, b \in \{1, 2, 3, 4\}\) and \(u, v, x \in \mathfrak{D}_{\xi}\).

Sketch.

Lean: Stress.Force, Stress.force, Stress.Generated, Stress.tensor_lorentz, Stress.tensor_herm, Stress.tensor_conserved.

4.8 Assembly

Proof of V.3.1. Take the family of V.3.b. Then:

Lean: all_characters, Resolution.barycenter.

5. Where the difficulty lies

The delicate points are the common bounds near the analytic boundary, the recovery of positive-height Euclidean vectors from joint frequency tests, and the joint current limits in the stress-tensor argument. The last passage uses equal continuation germs only at spacelike separation; the current itself is spatially integrated. It gives a charge commutator for every admissible time profile, without defining a stress field at a fixed real time.

The low-energy estimate needs only locality, the spectrum condition and the two vector-norm bounds. Its positive weight avoids differentiating spectral integrals or imposing generator-domain bounds. The exponent \(2\) is compatible with the massless scalar measure, since \(\int_{| \vec{p} | \leqslant u} d^3 p / | \vec{p} | \sim u^2\). Thus this estimate does not imply a mass gap. For removal of the exceptional set, any decay rate tending to zero uniformly in \(\xi\) would suffice.

The remaining integrations, dyadic summation and assembly are routine once their uniform bounds have been established.

What looks unusual.

6. Notation

object code name proposed notation
phases, measure, selected phase F.α.space.X, F.α.space.μ, F.α.atPoint \(X\), \(\nu\), \(\xi_{\ast}\)
characters of the ground algebra F.α.space.character, Spread.algebra \(\chi_{\xi}\) on \(\mathfrak{A}\)
unselected table; table of phase \(\xi\) S; F.α.ensemble ξ \(S^{\text{mix}}\); \(S^{\xi}\)
OS space and slab operators of \(S^{\text{mix}}\) F.α.room, F.α.unselected \(\mathcal{H}^{\text{OS}}\), \(\Pi (d, w)\)
ground space, Euclidean vacuum Chamber.D, Chamber.omega0 \(\mathcal{D}_{\xi}\), \(\omega_{\xi}\)
one-dimensional ground space Chamber.UniqueHeat \(\mathcal{D}_{\xi} =\mathbb{C} \omega_{\xi}\)
family of reconstructions Almanac \((\mathcal{H}^{\text{OS}}_{\xi}, \Pi_{\xi}, \textsf{F}^{\xi}, \mathfrak{W}_{\xi}, J_{\xi})_{\xi \in X}\)
OS space, slab operators of \(S^{\xi}\) L.chambers ξ, L.spreads ξ \(\mathcal{H}^{\text{OS}}_{\xi}\), \(\Pi_{\xi} (d, w)\)
continuation and boundary values L.contours ξ (waves, world) \(\textsf{F}^{\xi}_n\), \(W^{\xi}_n\)
Wightman theory of phase \(\xi\) L.actors ξ \(\mathfrak{W}_{\xi} = (\mathcal{H}_{\xi}, \mathfrak{D}_{\xi}, \Omega_{\xi}, \phi_{\xi}, U_{\xi}, P_{\xi})\)
boundary isometry L.J ξ, L.links ξ (Attachment) \(J_{\xi}\)
Wightman data without vacuum uniqueness FieldCore —
heat semigroup on \(\mathcal{H}_{\xi}\) bHeat \(e^{- sH_{\xi}} = J_{\xi}^{\ast} e^{- sH^{\text{OS}}_{\xi}} J_{\xi}\)
zero-energy projection momentum.region zeroE \(P^0_{\xi}\)
finite sum of words; compact supports; adjoint; vector SumOfWords, finite, reverseP, stands \(A\), \(A^{\dagger}\), \(A \Omega_{\xi}\)
spectral measure of \(A \Omega_{\xi}\) νletter \(\mu^{\xi}_A\)
low-energy region; shell lowE, ringE \(\{0 < p^4 \leqslant u\}\); \(\{u / 2 < p^4 \leqslant u\}\)
vacuum commutator function delta \(\langle \Omega_{\xi}, [A^{\dagger}, \alpha_p A] \Omega_{\xi} \rangle\)
time-averaged odd part; weight averaged, pulse \(D (\vec{x})\); \(\pi (0) = 0\), \(\pi (E) = (1 - \cos E) / E\) for \(E > 0\)
strict-height values; tested values waves.restricted, waves.psi \(\textsf{F}^{\xi}_n (a ; t (x) + iy, \vec{x})\); \(\psi^{\xi} (a, y, g)\)
weak commutant member, weaklyCentral —
Minkowski stress labels and fields Stress.M, Stress.fieldM \(T^{\text{M}}_{ab}\), \(T_{ab} (f)\)
Killing field; charge; Lie derivative of a field Stress.Killing, Stress.Charge, Stress.action \(K\), \(Q_K (f, c)\), \(\phi (\Sigma_K B) + K^a \partial_a \phi (B)\)
stress record Stress.Force (fields appointment, orbital, symmetric, hermitian, conserve, commutator, rule) (T0)–(T6)
complete record Completed (fields isTheory, into_onto, euclid, mink, light, heat, commutant, continued_stress, depends, at_boundary, at_words) items 1, 2, 3, 4, 5, 6, 7, 8, 9, 9, 9 of V.3.1

7. What remains unexplained

  1. Analytic inputs from V.2. The constructions contour, compatible, establishing and their use here were checked in source. The underlying C01–C29 analytic estimates, including the partial-boundary identities used in V.3.i and the exchange continuations in (T0), remain inputs from V.2; their proofs are not audited here. The local statements and proof sketches have been checked against the declarations that consume them.

  2. Partial-boundary data. (T0) states the existence and compatibility needed for the current-jump argument. A detailed proof must specify the admissible tube faces and the spacelike neighborhoods of the tied pair; it must not replace integrated current limits by pointwise limits of the stress insertion.

← V.2 Osterwalder-Schrader reconstruction of a Wightman theory (notes)Part V·Unit outlineVI.1 Ward identities, the rotation action and the conserved stress tensor (notes) →