1. What this unit proves 1
2. What it uses 7
3. Strategy 9
4. Steps 10
4.1 OS space, chambers, spectral calculus 10
4.2 Slab operators 12
4.3 The compressed algebra 13
4.4 Phase reflection and rotation invariance 14
4.5 The phase tables 16
4.6 Null spaces of the phases 17
4.7 Choosing a phase 18
5. Where the difficulty lies 20
6. Notation 21
7. What remains unexplained 22
The full account of unit V.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: V, phases and reconstruction. Lean: YM/OS, YM/Phase, YM/Ground, YM/Heat, YM/SNAG, YM/Paley, YM/Spectral, the work items B01–B18 in YM/Work, and the assembly files Bid/Phase.lean, Bid/Fetch.lean, Bid/Market.lean, Org/Phase.lean. Size: about 5,700 lines in declarations used by the main theorem, including about 2,700 in B01–B18, plus the assembly files.
Setting and notation.
Tables and words. \(S\) is the limit table of IV.1 (the mixed state \(S^{\text{mix}}\) of the main theorem). A label is a pair \((d, v)\): its degree \(d \in \mathbb{N}\) and a vector \(v\) in the finite-dimensional reference space \(V_d\). A word \(w = ((x_1, v_1), \ldots, (x_n, v_n))\) is a finite list of points of \(\mathbb{R}^4\) with labels; \(|w| = n\), and \(S (w) = S (x ; v)\) is its value, multilinear in the labels.
Time, slabs, buffers. Time is the last coordinate \(t (x)\), with unit vector \(e_t\); \(y \perp e_t\) denotes a spatial vector. For a different coordinate axis \(a\), replace \(t (x)\) by \(x^a\) in the following definitions.
A word is separated if its points are distinct, and positive if in addition all \(t (x_i) > 0\).
It lies in the slab \((0, d)\) if \(0 < t (x_i) < d\) for all \(i\).
It is \(r\)-buffered in that slab if its points are at mutual sup-distance \(\geqslant r\) and \(r \leqslant t (x_i) \leqslant d - r\).
Operations on words. \(w + a\) is the translate, \(\tau_d w = w + de_t\), \(w \cdot w'\) is concatenation, and \(\bar{w}\) conjugates only the labels.
Two reflections.
\(\Theta w\), the OS reflection: reverse the list, reflect the points in \(\{t = 0\}\), and apply to the labels the antilinear map (conjugation followed by the time-mirror action).
\(\vartheta w\), the linear time reflection: reflect the points and apply the time-mirror action to the labels, with neither conjugation nor reversal.
For real labels, \(\Theta w\) is a permutation of \(\vartheta w\).
Null spaces. \(N_d (S) \subset V_d\) consists of labels whose correlations with every separated family of spectators vanish. Write \(F_d (S) = V_d / N_d (S)\) and \(q_d : V_d \to F_d (S)\) for the quotient map. A table \(S'\) respects \(S\) if \(N_d (S) \subset N_d (S')\) for all \(d\).
Field algebra (VI.1). It consists of label derivatives \(\partial_j : F_d (S) \to F_{d + 1} (S)\), inclusions \(F_d (S) \to F_e (S)\) for \(d \leqslant e\), an action \(R\) of \(O (4)\) on these spaces, and a stress tensor \(J_{ab} \in F_4 (S)\). In a word, \(\partial_j v\) and \(R (K) v\) mean representatives of \(\partial_j q_d (v)\) and \(R (K) q_d (v)\). All tables and operators to which this notation is applied factor through the relevant quotients. Covariance means \(S (Kx_1, \ldots, Kx_n ; R (K) v_1, \ldots, R (K) v_n) = S (x ; v)\) for every \(K \in O (4)\) and separated configuration \(x \in \text{Conf}_n\).
Caps. For each fixed arity, degree list, label list and derivative order \(j\), there are \(A \in \mathbb{N}\) and \(C \geqslant 0\) such that \(\|D^j S (x ; v)\| \leqslant Cr^{- A}\) whenever \(0 < r \leqslant 1\) and the points have mutual sup-distance at least \(r\).
The OS space. The OS space \(\mathcal{H}\) of \(S\) for an axis \(a\) is the completion of the span of \(a\)-positive words modulo null vectors, with \(\langle b, b' \rangle = S (\Theta b \cdot b')\). \(|b \rangle\) denotes the vector of \(b\), and \(\Omega = | \emptyset \rangle\), with \(\| \Omega \| = 1\).
Nowhere does the unit assume clustering, a unique vacuum or a spectral gap.
Proposition V.1.1 (chambers and ground space). Let \(S\) satisfy the Euclidean hierarchy and let \(a\) be a coordinate axis. Then:
On the OS space \(\mathcal{H}\) there is a strongly continuous semigroup \((E_t)_{t \geqslant 0}\) of self-adjoint contractions with \(E_t |b \rangle = |b + te_a \rangle\).
There is a strongly continuous unitary group \((U_y)_{y \perp e_a}\) commuting with it, with \(U_y |b \rangle = |b + y \rangle\).
The triple is a chamber. Let \(\mathcal{H}_0 = \{\psi : E_t \psi = \psi \forall t\}\) be its ground space and \(P_0\) the orthogonal projection onto it. Then \(\Omega \in \mathcal{H}_0\), and \(E_t \psi \to P_0 \psi\) as \(t \to \infty\) for every \(\psi\).
Status: theorem. Lean: OS.realization, OS.implement, OS.Semigroup.emerges; definitions OS.Chamber, OS.Semigroup, Semigroup.ground.
Proposition V.1.2 (spectral calculus in a chamber). In a chamber, write formally \(E_t = e^{- tH}\).
(a) There is a strongly continuous unitary group \(s \mapsto e^{isH}\), determined by \(e^{isH} E_1 \psi = E_1^{\hspace{0.17em} 1 - is} \psi\). Here \(E_1^z\), for \(\text{Re} \hspace{0.17em} z > 0\), is the continuous functional calculus of the positive contraction \(E_1\). The map \(V (x) = e^{i \hspace{0.17em} x^a H} U_{x_{\perp}}\), with \(x_{\perp} = x - x^a e_a\), is a strongly continuous unitary representation of \(\mathbb{R}^4\).
(b) There is a projection-valued measure \(P\) on \(\mathbb{R}^4\) with \(\langle \psi, V (x) \psi \rangle = \int e^{ip \cdot x} \hspace{0.17em} d \mu_{\psi} (p)\), where \(\mu_{\psi} (B) = \langle \psi, P (B) \psi \rangle\) for Borel \(B\).
(c) Every \(\mu_{\psi}\) is supported in \(\{p^a \geqslant 0\}\).
(d) For \(\text{Im} \hspace{0.17em} z \geqslant 0\):
Status: theorem. It holds for every semigroup with the properties of V.1.1. No gap is assumed or proved. Lean: Semigroup.edge, Semigroup.unit, Spectral.Orbit.total, Orbit.P, Semigroup.nonneg, Semigroup.heat.
Proposition V.1.3 (slab operators). Assume the hierarchy and the bench (V.1.7). Then every chamber of \(S\) for the time axis carries a unique family of bounded operators \(T_d (w)\), one for each \(d > 0\) and each word \(w\) in the slab \((0, d)\), with
All-copy bound. For every \(D \in \mathbb{N}\) and \(\rho > 0\) there are \(C > 0\) and \(\kappa \in \mathbb{N}\) such that
for every \(d > 0\), every \(r \in (0, 1]\), and every word \(w\) that is \(r\)-buffered in \((0, d)\) with degrees \(\leqslant D\) and label norms \(\leqslant \rho\). The right side is \(1\) for the empty word.
Smoothness. If moreover \(S\) satisfies the derivative identities of the plan \(\partial\), then for fixed \(d\) and labels the map \(x \mapsto T_d (w (x))\) is \(C^{\infty}\) in operator norm on the open set where \(w (x)\) lies in the slab. Its derivatives are \(\partial_{x_p^j} T_d (w (x)) = T_d (w' (x))\), where \(w'\) replaces the label \(v_p\) by \(\partial_j v_p\).
Status: theorem, given the bench. More generally, smoothness follows for any slab family with the all-copy bound on a time-axis chamber of a table \(S'\) that satisfies the hierarchy, respects \(S\), and satisfies the derivative identities of the plan on \(S\). Lean: Work.B01.answer, Work.B02.answer; definitions Spread, Spread.AllCopy, Spread.Laws.
Proposition V.1.4 (the compressed vacuum algebra). Assume the hierarchy and \(O (4)\)-covariance of \(S\). Fix a time-axis chamber and any family \(T_d (w)\) with the defining identity of V.1.3. Then:
(a) \(U_y \psi = \psi\) for every \(\psi \in \mathcal{H}_0\) and every spatial \(y\).
(b) For a separated word \(w\), the operator \(M (w) := P_0 \hspace{0.17em} T_d (w + s) |_{\mathcal{H}_0}\) does not depend on the choice of \(s \in \mathbb{R}^4\) and \(d > 0\) with \(w + s\) in the slab \((0, d)\). For positive \(b\), \(M (b) \Omega = P_0 |b \rangle\). For every separated \(w\), \(\langle \Omega, M (w) \Omega \rangle = S (w)\).
(c) Cluster limit. Let \(q_1, \ldots, q_m\) be separated words and \(a_1 (n), \ldots, a_m (n) \in \mathbb{R}^4\) with \(t (a_j (n) - a_i (n)) \to + \infty\) for \(i < j\); the spatial components are arbitrary. Then for large \(n\) the word \((q_1 + a_1 (n)) \cdots (q_m + a_m (n))\) is separated, and its value under \(S\) tends to \(\langle \Omega, M (q_1) \cdots M (q_m) \Omega \rangle\).
(d) The norm-closed \(\ast\)-algebra \(\mathcal{A} \subset B (\mathcal{H}_0)\) generated by the \(M (w)\) is commutative, \(\mathcal{A} \Omega\) is dense in \(\mathcal{H}_0\), and \(p \in \mathcal{A}\) with \(p \Omega = 0\) implies \(p = 0\).
(e) Suppose in addition that the family satisfies the smoothness part of V.1.3. Then:
\(M\) is invariant under permutations of the word and under all translations; \(M (\emptyset) = 1\); a unit label may be dropped together with its point; and \(M (w)^{\ast} = M (\Theta w)\).
For \(x \in \text{Conf}_n\), the map \(v \mapsto M (x ; v)\) is multilinear and factors through \(F_{d_1} (S) \times \cdots \times F_{d_n} (S)\).
\(x \mapsto M (x ; v)\) is \(C^{\infty}\) in norm on \(\text{Conf}_n\).
\(\mathcal{A}\) is norm-separable.
Status: theorem. Lean: Ground.inert, Ground.overline, Ground.almost, Spread.insertValue, Ground.arcs, Ground.adjunction, Ground.full, Ground.attains, Work.B03.answer, Work.B12.answer; definitions Ground.M, Spread.algebra, Bid.Theme.
Proposition V.1.5 (decomposition into phases). Assume the setting of V.1.4(a)–(e). Then:
There are a compact metrizable space \(X\) (the Gelfand spectrum of \(\mathcal{A}\)), a Borel probability measure \(\nu\) on \(X\) with \(\nu (O) > 0\) for every nonempty open \(O\), and characters \(\chi_{\xi}\) (\(\xi \in X\)).
\(\xi \mapsto \chi_{\xi}\) is a bijection onto the characters of \(\mathcal{A}\); each \(\xi \mapsto \chi_{\xi} (p)\) is continuous; and \(\int_X \chi_{\xi} (p) \hspace{0.17em} d \nu (\xi) = \langle \Omega, p \Omega \rangle\) for every \(p \in \mathcal{A}\).
For each \(\xi\), the formula \(S^{\xi} (x ; v) = \chi_{\xi} (M (x ; v))\) on \(\text{Conf}_n\) (and \(0\) off it) defines a table.
\(S^{\xi} (w) = \chi_{\xi} (M (w))\) for every separated \(w\); \(\xi \mapsto S^{\xi} (w)\) is continuous; and the barycenter identity holds:
Status: theorem. Lean: Ground.resolution, Work.B07.state, Resolution.barycenter; definitions Spread.Resolution, Resolution.Out, Resolution.receives.
Proposition V.1.6 (every phase satisfies the hierarchy and the field laws). In addition to V.1.3–V.1.5 (with the smoothness laws for the plan of the field algebra), assume:
the Seed of \(S\) (IV.1);
the grid hypothesis (§2);
the laws of the field algebra for \(S\) (VI.1).
Then two operator identities hold before any character is applied:
(e) phase reflection: \(M (w) = M (\vartheta w)\) for every separated word with arbitrary complex labels;
(f) \(M (w) = M (Kw)\) for every separated \(w\) and every \(K \in O (4)\), where \(K\) acts on points and, through \(R (K)\), on labels.
As a consequence, for every \(\xi \in X\):
(a) \(S^{\xi}\) satisfies the Euclidean hierarchy:
permutation symmetry, \(S^{\xi} (\emptyset) = 1\), and unit labels;
invariance under signed permutations of the axes, and reality \(S^{\xi} (x ; \bar{v}) = \overline{S^{\xi} (x ; v)}\);
translation invariance;
reflection positivity for every coordinate axis, in point and smeared form;
smoothness on \(\text{Conf}_n\), and caps on all derivatives.
(b) \(S^{\xi}\) respects \(S\) and is \(O (4)\)-covariant for the same action \(R\). It satisfies the inclusion and derivative identities of the plan and the stress-tensor Ward identities of VI.1.
(c) Uniform cap: for every \(D \in \mathbb{N}\) and \(\rho > 0\) there are \(K > 0\) and \(\kappa \in \mathbb{N}\) with
for all \(\xi\), all \(r \in (0, 1]\), and all words with degrees \(\leqslant D\), label norms \(\leqslant \rho\) and points at mutual sup-distance \(\geqslant r\).
(d) \(N_d (S^{\xi}) = N_d (S)\) for every \(d\), hence \(F_d (S^{\xi}) = F_d (S)\).
Status: theorem. Lean: Work.B05.answer, B06.answer, B07.answer, B08.answer, B09.answer, B10.answer, B11.answer, Retain.lift_detect; definitions Phase.reflects, Bid.Garland, Phase.Scripts.
Proposition V.1.7 (bench). For every \(D \in \mathbb{N}\) and \(\rho > 0\) there are \(C_0 > 0\) and \(\kappa \in \mathbb{N}\) with the following property. Let \(d > 0\), \(r \in (0, 1]\), and let \(w \neq \emptyset\) be \(r\)-buffered in \((0, d)\), with degrees \(\leqslant D\) and label norms \(\leqslant \rho\). Then there are a lattice level \(b\) and a width \(s \leqslant d\) with \(s \in a_b \mathbb{N}\) and \(w\) in the slab \((0, s)\), such that every positive word \(v\) has a constant \(B_v\) with
Here \(v_n\) is \(v\) preceded by \(2 n\) consecutive slabs of width \(s\), holding alternately \(w\) and its reflection. Thus \(|v_n \rangle = (T_s (w)^{\ast} T_s (w))^n |v \rangle\) once \(T_s (w)\) exists, and \(S (\Theta v_n \cdot v_n)\) is a reflection-symmetric train of \(4 n\) copies of \(w\).
Status: a hypothesis of this unit, required of every limit table with a Seed. It is proved in VII.4 (Next.Phase.seats) from the cutoff family, the hierarchy, the point limits and the cell hypothesis of the Seed.
Chessboard estimate. The cell hypothesis bounds every signed or reflected field of degree \(\leqslant D\) by \(J \hspace{0.17em} a_k^{- (D + 1)}\) in chessboard norm. The scale can be chosen with \(a_k \geqslant c \hspace{0.17em} (r / |w|)^4\), where \(c > 0\) depends only on the fixed family and \((D, \rho)\). A finer scale handles the exterior word. The resulting train bound is \(B^{2 |v|} (Ja_k^{- (D + 1)})^{4 n |w|}\) (Next.Phase.travel_price). Thus one may take \(\kappa = 4 (D + 1)\); the exterior constant is independent of \(n\).
Lean: Bid.Bench, Bid.Seats, OS.QRead, Next.Phase.replicas, Next.Phase.rates.
Proposition V.1.8 (vacuum uniqueness in almost every phase). Let \((X, \nu, \chi)\) and \(S^{\xi}\) be as in V.1.5, with every \(S^{\xi}\) satisfying the hierarchy. For each \(\xi\) fix a time-axis chamber of \(S^{\xi}\), with vacuum \(\Omega_{\xi}\) and ground space \(\mathcal{H}_0^{\xi}\). Then:
\(\mathcal{H}_0^{\xi} =\mathbb{C} \hspace{0.17em} \Omega_{\xi}\) for \(\nu\)-almost every \(\xi\); in particular such phases exist.
Every time-axis chamber of every \(S^{\xi}\) carries a slab family \(T^{\xi}_d (w)\) with \(\|T^{\xi}_d (w)\| \leqslant \|T_d (w)\|\) for each valid slab word. If the original family has the all-copy bound, so does this family. Under V.1.6 it also has the smoothness laws.
Status: theorem. Uniqueness for every phase is proved later, in V.3 (EveryPhase.every), by an argument outside this unit. Lean: Cluster.exp_real, Cluster.lifts, Cluster.selects, Work.B13.answer, Work.B14.answer; definition Chamber.UniqueHeat.
Proposition V.1.9 (selected-phase certificate). Let \(C\) be a Wilson family with extraction order \(f\) and limit table \(S\). Assume:
the Seed of \(S\);
the grid hypothesis;
a field algebra on \(S\) with its laws.
Then there are:
a time-axis chamber of \(S\) whose ground space is fixed pointwise by spatial translations;
a slab family on it with the all-copy bound and the smoothness laws;
a decomposition \((X, \nu, \chi_{\xi}, S^{\xi})\) as in V.1.5, with every \(S^{\xi}\) satisfying V.1.6(a), (b), (d);
a phase \(\xi_{\ast}\), with a time-axis chamber of \(S^{\xi_{\ast}}\) whose ground space is \(\mathbb{C} \Omega_{\xi_{\ast}}\), and a slab family on that chamber with the smoothness laws.
Status: theorem.
Assembly. The bench theorem and B01–B18 give the slab and phase constructions (Bid.Looms); the selection argument produces the certificate (Phase.Loom.cleanse).
Use in the main theorem. The main theorem applies it in Pert.withFinal and MergeProof.Acme.glistens. The result is the phase field of Spec.Final.
Lean: definitions Phase.Loom, Phase.Stall, Phase.Vendor, OS.Selection, Selection.Certificate, Spec.Materializes.
IV.1 (limit table and Seed).
The limit table \(S\) with the Euclidean hierarchy: Lean Table.Hierarchy, with fields basic laws, real-spin, translations, point and smeared reflection positivity, smoothness, caps.
The Seed, Lean Spec.Seed. Its components:
(i) the cutoff family;
(ii) the ordered limit;
(iii) point limits at generic points (PointLimits);
(iv) the hierarchy;
(v) the cell hypothesis (CellHyp): reflection-positivity seams at every cutoff; and, for every degree \(d\) and \(\varepsilon, \rho > 0\), a threshold \(m_0\) and \(A \geqslant 0\) such that for every \(m \geqslant m_0\) and \(\eta > 0\), late in the ordered cutoff, every degree-\(d\) label of norm at most \(\rho\) is supported inside its centered cell and has chessboard root at most \(A \hspace{0.17em} a_m^{- d - \varepsilon} + \eta\), uniformly over cell centers;
(vi) the corrected-field data (RankWidths, Columns, Trials.Valid). For each degree \(d\): finitely many one- or two-point test rows and a basis of the free-field space of degree \(d\), for which the free test map is injective; and lattice "corrected" fields at levels \(k\) whose classes span \(F_d (S)\) for all sufficiently large \(k\), whose one-field differences from free values and centered two-field products minus free cumulants have chessboard bounds \(\text{gap}_k + \eta\) late in the cutoff for every \(\eta > 0\), with \(\text{gap}_k \to 0\). The free values use the rescaled test points.
The free table (Cells.standard).
Where these are used: (i)–(iv) in B04, B16 and the bench; (v) in the bench, B15 and the grid hypothesis; (vi) in B18 and lift_detect.
VII.4 (the bench). V.1.7, i.e. Next.Phase.seats.
III.11 (the grid hypothesis) Org.Grill. Proved from bounded ultraviolet approximants and the cell hypothesis by MergeProof.Accent.grilled.
Setup: take a degree bound \(d\) and a level \(m\). There is a level \(k \geqslant m\) such that the following holds for every word \(w\) with real labels of degree \(\leqslant d\) at generic points of the level-\(k\) grid.
Let \(F\) be the lattice product of the fields of \(w\), and \(\vartheta F\) the product of the mirrored fields at the reflected points.
The lattice variables \(f_j\) are the translates of \(F - \vartheta F\) by \(t \hspace{0.17em} j\) grid steps in every coordinate.
Conclusion: for each approximation index \(\varkappa\) the \(f_j\) admit real measurable approximants, bounded by \(R_{\varkappa}\) uniformly in the cutoff, whose pair integrals, for \(i < j\), differ from \(\mathbb{E} [\text{Re} (f_i f_j)]\) by at most \(\delta_{\varkappa} \to 0\). For a fixed grid word, \(R_{\varkappa}\) and \(\delta_{\varkappa}\) are independent of the translation, copy indices and cutoffs. The pair estimates hold for all sufficiently large \(t\), then late in the cutoff (Grid.Covering). It is used only in B05.
VI.1 (field algebra and its laws on \(S\)) (Spec.Algebra, Spec.Locals).
\(O (4)\)-covariance of \(S\) (Rotation.Oriented) is used in V.1.4(a), (d), in Ground.resolution, and in B06, B07.
The rotation laws (Rotation.Laws): group law, and \(R (s)\) equal to the mirror action of \(S\) for signed permutations \(s\). Used in B06, B07.
The derivative identities are used in B02 and B09; the inclusion identities and stress-tensor laws (StressLaws) in B09.
Order of dependence. VI.1 uses the chamber construction and the bounded heat functional calculus from the beginning of this unit, as well as elementary geometry and norm estimates. It does not use the compressed algebra, phase reflection or phase selection. Those constructions then use the \(O (4)\) action supplied by VI.1.
V.2. The Gram/completion construction of the OS space (Kernel.Gram, with by_bound, by_bud).
I.1, II.1. Lattice, generic points, label registers, the Wilson family.
Standard mathematics.
Osterwalder–Schrader reconstruction and the doubling bound for the semigroup: Osterwalder–Schrader, CMP 31 (1973), 42 (1975); Glimm–Jaffe, Quantum Physics, Ch. 19 (probable).
Chessboard estimates: Fröhlich–Israel–Lieb–Simon, CMP 62 (1978).
Pure phases. The classical analogue is the decomposition of a Nelson–Symanzik positive theory into pure phases, Fröhlich, Ann. Phys. 97 (1976) (probable), and the decomposition of states along abelian algebras, Bratteli–Robinson I, Ch. 4.
Operator algebras and measure theory:
Gelfand–Naimark for commutative C*-algebras and spectral permanence (Bratteli–Robinson I, §2.1–2.2);
Riesz–Markov (Reed–Simon I, Thm. IV.14);
continuous functional calculus;
Hahn–Banach;
dominated convergence.
Harmonic analysis: the Stone–Naimark–Ambrose–Godement theorem (Folland, A Course in Abstract Harmonic Analysis, §4.4) and a Paley–Wiener support theorem for bounded holomorphic functions in a half-plane (Reed–Simon II, §IX.3) (probable).
Geometry: Cartan–Dieudonné (reflections generate \(O (4)\)).
Imported constants and assumptions.
No numerical constant is fixed in the unit.
The all-copy bound uses the constants of the bench; one may choose \(\kappa = 4 (D + 1)\). These constants may depend on the fixed family and \((D, \rho)\), but not on the slab width, buffer, word length or number of copies.
The hypotheses are: the hierarchy and the Seed (IV.1), the grid hypothesis (III.11), the bench (VII.4), and the field-algebra laws (VI.1).
The limit table as a mixture. The table \(S\) is obtained by ordered limits of finite-volume Wilson correlations. Its OS semigroup may have a multidimensional ground space \(\mathcal{H}_0\). The unit decomposes the correlations through an algebra acting on that space.
The algebra. Inserting a separated word into a time slab gives a bounded operator; compressing it to \(\mathcal{H}_0\) gives \(M (w)\). These operators generate a commutative C*-algebra \(\mathcal{A}\), for which \(\Omega\) is cyclic and separating.
Phases. The Gelfand spectrum \(X\) of \(\mathcal{A}\), with the measure \(\nu\) representing \(p \mapsto \langle \Omega, p \Omega \rangle\), is the space of phases. Each character gives a table, \(S^{\xi} = \chi_{\xi} \circ M\), and \(S = \int S^{\xi} \hspace{0.17em} d \nu\).
Comparison with the classical case. The algebra describes observables at infinite time separation. It replaces the invariant observables used in the ergodic decomposition of a Euclidean measure. Here the construction uses only the OS space of the limit correlations. Operator identities give the hierarchy and field laws for every \(\xi\); vacuum uniqueness is first obtained almost everywhere.
Four inputs. Four inputs make this work, and each is used where the classical argument would use the measure.
(i) Boundedness. A chessboard estimate controls alternating trains of \(4 n\) copies with constants independent of \(n\). A power argument converts this estimate into the slab-operator bound \(C^{|w|} (|w| / r)^{\kappa |w|}\).
(ii) \(O (4)\)-covariance (VI.1). It shows that ground states are invariant under spatial translations, which makes \(M (w)\) well defined. Spatial displacements are rotated into time displacements, which are asymptotically invisible. It also makes \(\mathcal{A}\) commutative: exchanging the time axis with a spatial axis reverses the time order of two clusters inside a four-cluster limit.
(iii) Phase reflection \(M (w) = M (\vartheta w)\). Together with \(M (w)^{\ast} = M (\Theta w)\), this makes every phase value on real labels real. Put \(D = M (w) - M (\vartheta w)\) for such a word. Then \(D^{\ast} = - D\), so \(\langle \Omega, D^2 \Omega \rangle = - \|D \Omega \|^2\). Bounded real lattice approximants give the reverse inequality by positivity of the variance of many far-separated copies. Thus \(D = 0\). Conjugating time reflection gives invariance of \(M\) under \(O (4)\).
(iv) Chessboard detectors. For the corrected fields they show that the one- and two-point functions of every phase, on a spanning family of fields, are close to free-field values for which the test map is injective. Hence every phase has the null spaces of \(S\).
Selecting a phase. For a positive word \(b\), the quantity \(\|P_0^{\xi} |b \rangle \|^2 - |S^{\xi} (b) |^2 \geqslant 0\) has \(\nu\)-integral zero. This follows by dominated convergence along \(E_s\) and the barycenter identity. One countable family of words is total in every phase chamber, so a single null set suffices, and outside it every phase has a one-dimensional ground space. The spectral calculus of V.1.2 (SNAG, positivity of the energy from analyticity of the heat semigroup) is built without assuming a gap and is used downstream (V.2, V.3).
Lemma V.1.a (chamber). Under the hierarchy, V.1.1 holds except for the convergence \(E_t \to P_0\). Proof sketch.
Contraction (OS doubling trick). For \(p\) in the span of positive words, \(\|E_t p\|^2 = \langle p, E_{2 t} p \rangle \leqslant \|p\| \sup_s \|E_s p\|\). The supremum \(c\) is finite, because \(s \mapsto S (\Theta b \cdot \tau_s b')\) is bounded for \(s \geqslant 0\): the points stay at a fixed positive distance, and the zeroth cap applies. Hence \(c^2 \leqslant \|p\| c\), so \(\|E_t p\| \leqslant \|p\|\).
Self-adjointness and unitarity. Self-adjointness of \(E_t\) and unitarity of \(U_y\) follow from translation invariance.
Strong continuity. It follows from continuity of the Gram functions (smoothness) together with the uniform bounds.
Lean: OS.contract, OS.implement.
Lemma V.1.b (ground projection). For a semigroup of self-adjoint contractions, \(E_t \psi \to P_0 \psi\) for every \(\psi\). Proof sketch.
For \(u \leqslant v\), \(\langle E_u \psi, E_v \psi \rangle = \|E_{(u + v) / 2} \psi \|^2 \geqslant \|E_v \psi \|^2\). Hence \(\|E_u \psi - E_v \psi \|^2 \leqslant \|E_u \psi \|^2 - \|E_v \psi \|^2\), and the decreasing limit of \(\|E_t \psi \|^2\) makes \(E_t \psi\) Cauchy.
The limit is fixed by every \(E_s\), and \(\psi\) minus the limit is orthogonal to \(\mathcal{H}_0\), because \(\langle E_t \psi, u \rangle = \langle \psi, u \rangle\) for \(u \in \mathcal{H}_0\).
No spectral theorem is used.
Lean: Semigroup.emerges.
Lemma V.1.c (real-time group and SNAG). V.1.2(a), (b) hold. Proof sketch.
Real-time group. \(E_1 = E_{1 / 2}^2\) is positive, with spectrum in \([0, 1]\). The functional calculus gives \(E_1^z\), holomorphic for \(\text{Re} \hspace{0.17em} z > 0\), and \(E_1^t = E_t\) for \(t > 0\) by positive roots and continuity. If \(E_t \psi = 0\), analyticity makes \(E_s \psi = 0\) for every \(s > 0\), and continuity at zero gives \(\psi = 0\). Self-adjointness then makes the range of \(E_1\) dense. Since \(\|E_1^{1 - is} \psi \| = \|E_1 \psi \|\), the map \(E_1 \psi \mapsto E_1^{1 - is} \psi\) extends to an isometry. The power laws give the group law and inverse at \(- s\); continuity on the dense range gives strong continuity.
SNAG. Integrate \(V\) against Fourier transforms of Schwartz functions to obtain a \(\ast\)-representation of continuous functions on the one-point compactification of \(\mathbb{R}^4\). Riesz–Markov gives a projection-valued measure there. The point at infinity carries no mass: evaluate at \(x = 0\) along an approximate identity.
Lean: Semigroup.roots_heat, Semigroup.pillar, Semigroup.edge, Semigroup.unit, Spectral.Orbit.total, Spectral.PVM.
Lemma V.1.d (spectrum condition and Laplace formula). V.1.2(c), (d) hold. Proof sketch.
The function \(J\). Fix \(\psi\) and \(x\). The function \(J (z) = \langle \psi, V (x) e^{i (\text{Re} \hspace{0.17em} z) H} E_{\text{Im} \hspace{0.17em} z} \psi \rangle\) is continuous on \(\text{Im} \hspace{0.17em} z \geqslant 0\). In the interior it equals \(\langle \psi, V (x) E_1^{- iz} \psi \rangle\), so it is holomorphic there. It is bounded by \(\| \psi \|^2\). Its boundary values are the Fourier transform of \(e^{ip \cdot x} d \mu_{\psi} (p)\) in the variable \(p^a\).
Support theorem. Let \(\mu\) be a finite positive Borel measure on \(\mathbb{R}^4\), let \(\ell : \mathbb{R}^4 \to \mathbb{R}\) be continuous, and let \(F\) be bounded and holomorphic in the upper half-plane, continuous up to the boundary, with \(F (b) = \int e^{i \ell (y) b} d \mu (y)\). Then \(\text{supp} \hspace{0.17em} \mu \subset \{\ell \geqslant 0\}\). To see this, integrate \(F\) against the Gaussian \(e^{- ibz - \varepsilon z^2}\) on horizontal lines and shift the contour upward. For \(b < 0\), choosing height \(- b / (2 \varepsilon)\) bounds the mass where \(| \ell (y) - b| < |b| / 2\) by \(Ce^{- 3 b^2 / (16 \varepsilon)}\). Letting \(\varepsilon \downarrow 0\) excludes negative \(\ell\) from the support. Apply this with \(x = 0\) in \(J\) and \(\ell (p) = p^a\).
Laplace formula. Both sides of (d) are bounded and holomorphic with the same boundary values, so they agree.
Lean: Semigroup.altitude, Semigroup.nonneg, Semigroup.heat, Paley.supports, Paley.heat, Paley.Air.
Lemma V.1.e (power argument).
Setting. Let \(\Delta > 0\), let \(A, B\) be linear maps on a complex vector space, and let \(Y\) be a linear map into a Hilbert space, with \(\langle YAv, Yw \rangle = \langle Yv, YBw \rangle\) for all \(v, w\).
Statement. If every \(v\) has a constant \(K_v\) with \(\|Y (BA)^n v\| \leqslant K_v \Delta^{2 n}\) for all \(n\), then \(\|YAv\| \leqslant \Delta \|Yv\|\).
Proof sketch.
Put \(\delta_n = \|Y (BA)^n v\| / \Delta^{2 n}\). Then \(\delta_n^2 \leqslant \|Yv\| \hspace{0.17em} \delta_{2 n}\), so \(\sup_n \delta_n \leqslant \|Yv\|\).
Hence \(\|YAv\|^2 = \langle Yv, YBAv \rangle \leqslant \Delta^2 \|Yv\|^2\).
Apply this with \(A\) = insertion of \(w\), \(B\) = insertion of \(\tau_s \Theta w\), and \(Y\) = the OS map. The bench gives \(K_v = \sqrt{B_v}\) for single words; for finite sums use the triangle inequality. Extend by continuity.
Lean: Drive.against, OS.size, OS.producer.
Lemma V.1.f (existence and all-copy bound; B01). The first part of V.1.3 holds. Proof sketch.
Let \(w\) be \(r\)-buffered in \((0, d)\). The bench gives \(s \leqslant d\) with the train bound at width \(s\). V.1.e then gives a bounded operator \(T'\) with \(\|T' \| \leqslant A\) and \(T' |b \rangle = |w \cdot \tau_s b \rangle\).
Put \(T_d (w) = T' E_{d - s}\), so \(\|T_d (w)\| \leqslant A\); for the empty word, \(T_d (\emptyset) = E_d\).
Every word in a slab is buffered for some \((D, \rho, r)\). Operators agreeing on all positive-word vectors coincide, so the family is unique on valid slab words.
Lean: Work.B01.answer.
Lemma V.1.g (smoothness laws; B02).
Hypotheses. \(S'\) is a table respecting \(S\), both satisfying the hierarchy, and \(S'\) satisfies the derivative identities of the plan of \(S\). The slab family on a chamber of \(S'\) has the all-copy bound.
Conclusion. The family is \(C^{\infty}\) in operator norm, with derivatives given by differentiated labels.
Proof sketch.
Local bound on the candidate derivative. The candidate derivative is the operator with the label differentiated. It is locally bounded: an \(r\)-buffered configuration is \(r / 2\)-buffered on a neighbourhood, so the all-copy bound applies there.
Matrix elements. Between positive-word vectors, the derivative identity of \(S'\) gives the candidate's matrix elements as derivatives.
From matrix elements to norm. A mean-value estimate with constants uniform in the observing vectors gives first Lipschitz continuity in norm, then norm continuity of the candidate (apply the same to the next derivative), then Fréchet differentiability. Induction on the order gives \(C^{\infty}\).
Lean: Work.B02.answer, B02.norm_remainder, B02/Slab.
Lemma V.1.h (ground states are spatially invariant). Under the hierarchy and \(O (4)\)-covariance, \(U_y \psi = \psi\) for all \(\psi \in \mathcal{H}_0\). Proof sketch. Let \(p, q\) be positive words, \(y\) spatial, and \(u = y + e_t\).
Pure time shifts are invisible. For any \(d (s)\) with \(t (d (s)) \to \infty\) and any fixed \(\lambda\), \(S (\Theta p \cdot (q + d (s) + \lambda e_t)) - S (\Theta p \cdot (q + d (s))) \to 0\). Both terms are matrix elements \(\langle p', U_{z (s)} E_{\tau (s) + \lambda} q' \rangle\) and \(\langle p', U_{z (s)} E_{\tau (s)} q' \rangle\), with \(\tau (s) \to \infty\) and arbitrary spatial \(z (s)\), and \(E_{\tau} q'\) converges (V.1.b).
Rotate the spatial displacement into time. Choose a hyperplane reflection \(K\) with \(Ku = |u| e_t\) and \(t (Ke_t) > 0\). Covariance turns \(S (\Theta p \cdot (q + u + se_t)) - S (\Theta p \cdot (q + se_t))\) into a pure time shift by \(|u|\) at displacement \(sKe_t\), so it tends to \(0\).
Conclusion. Combining the two, \(\langle p, U_y E_s q \rangle - \langle p, E_s q \rangle \to 0\). Hence \(U_y P_0 |q \rangle = P_0 |q \rangle\), and this extends to all of \(\mathcal{H}_0\) by density.
Lean: Ground.inert, Ground.rays, Ground.across_axis, Ground.overtops.
Lemma V.1.i (the compression and its identities; B03). V.1.4(b) and (e), except separability, hold. Proof sketch.
Independence of the shift. For \(u \geqslant 0\) the defining identity gives \(T_{d + u} (w) = T_d (w) E_u\) and \(T_{u + d} (\tau_u w) = E_u T_d (w)\); spatial translations give \(U_y T_d (w) = T_d (w + y) U_y\). Together with \(P_0 E_u = P_0\) and V.1.h, these identities make \(M (w)\) independent of the shift and width.
Values. \(\langle \Omega, T_d (w) \Omega \rangle = S (w)\) because \(\Omega = | \emptyset \rangle\); translation invariance removes the shift.
Identities. Pairing against positive-word vectors gives \(\langle p, T_d (w) q \rangle = S (\Theta p \cdot w \cdot \tau_d q)\) with separated exterior words. Permutation symmetry, unit labels, multilinearity and dependence on \(F_d (S)\) therefore hold before compression.
Adjoint and smoothness. \(T_d (w)^{\ast} = T_d (\tau_d \Theta w)\) by translation invariance. Smoothness: near a configuration one shift and width serve (the slab condition is open), compression is bounded linear, and V.1.g applies.
Lean: Ground.at_any, Ground.overline, Ground.pins, Ground.slim, Work.B03.answer, Bid.depot.
Lemma V.1.j (cluster limits). V.1.4(c) holds. Proof sketch.
Reduction to positive clusters. Shift each cluster by a fixed vector to make it positive; this does not change \(M\).
Induction on \(m\). Shift the union so that \(q_1\) sits in its slab. Its vector is then \(T_d (q_1) U_{y_n} E_{u_n} | \text{rest}_n \rangle\) with \(u_n \to \infty\).
Key convergence. If \(\psi_n \to \psi\) and \(u_n \to \infty\), then \(U_{y_n} E_{u_n} \psi_n \to P_0 \psi\) for arbitrary spatial \(y_n\), by V.1.b and V.1.h. Inductively the vector converges to \(T_d (q_1) M (q_2) \cdots M (q_m) \Omega\), and pairing with \(\Omega\) inserts \(P_0\).
Lean: Ground.arcs, Ground.future_train, Ground.receives.
Lemma V.1.k (commutativity, cyclicity, separation). V.1.4(d) holds. Proof sketch. Write \(\langle q_1, \ldots, q_m \rangle = \langle \Omega, M (q_1) \cdots M (q_m) \Omega \rangle\).
Simultaneous rotation. If \(K \in O (4)\) maps some \(u\) with \(t (u) > 0\) to a vector with positive time, then \(\langle q_1, \ldots, q_m \rangle = \langle Kq_1, \ldots, Kq_m \rangle\). Place cluster \(j\) at \(n \hspace{0.17em} j \hspace{0.17em} u\), apply V.1.j to both orders, and use covariance of \(S\) at finite \(n\).
Exchange of two clusters. Let \(k\) exchange the time axis with the first spatial axis. Place four clusters at \(0\), \(n (e_t + 2 e_1)\), \(n (2 e_t + e_1)\) and \(n (3 e_t + 3 e_1)\), and apply \(k\): the two middle clusters exchange their time order. With the simultaneous-rotation step for \(u = e_t + e_1\), this gives \(\langle x, y, w, z \rangle = \langle x, w, y, z \rangle\).
Commutativity of generators. The vectors \(P_0 |f \rangle = M (f) \Omega\) for positive \(f\) have dense linear span in \(\mathcal{H}_0\), and \(M (f)^{\ast} = M (\Theta f)\). Thus \(\langle M (f) \Omega, M (u) M (v) M (p) \Omega \rangle = \langle \Theta f, u, v, p \rangle\) is symmetric in \(u, v\). Testing on both dense spans proves \(M (u) M (v) = M (v) M (u)\).
Extension to \(\mathcal{A}\). The generators are closed under adjoints, so \(\mathcal{A}\) is commutative.
Cyclicity and separation. \(\mathcal{A} \Omega \supset P_0 \{|b \rangle \}\) is dense. If \(p \Omega = 0\), then \(p \hspace{0.17em} a \Omega = a \hspace{0.17em} p \Omega = 0\) for all \(a \in \mathcal{A}\), so \(p = 0\).
Lean: Ground.same_up, Ground.exchanges, Ground.gens, Ground.adjunction, Ground.full, Ground.attains.
Lemma V.1.l (separability; Gelfand–Riesz; B12). \(\mathcal{A}\) is norm-separable, and the first part of V.1.5 holds. Proof sketch.
Separability. By multilinearity each generator is a finite combination of \(M (x ; e)\) with basis labels \(e\). For fixed arity, degrees and basis indices, \(x \mapsto M (x ; e)\) is continuous on the separable set \(\text{Conf}_n\). A countable union of separable images, finite products, adjoints and closure stay separable.
Gelfand–Riesz. Gelfand gives \(\mathcal{A} \cong C (X)\), and a countable dense sequence makes \(X\) metrizable. The vacuum state is represented by a probability measure \(\nu\).
Full support. If \(\nu (O) = 0\) for some open \(O \ni \xi_0\), a continuous bump supported in \(O\) corresponds to an element \(p \ne 0\) with \(\|p \Omega \|^2 = \int | \hat{p} |^2 d \nu = 0\), contradicting separation.
Lean: Work.B12.answer, Ground.resolution, Ground.tiny, Ground.supports.
Lemma V.1.m (grid passage; B04).
Setting. Assume the cutoff laws and generic point limits of the Seed. Let \(w\) have real labels at generic points of the level-\(k\) grid, and let \(w_j\) be its translate by \(t \hspace{0.17em} j\) grid steps in every coordinate.
Statement. For \(i < j\) and all large \(t\), the lattice pair expectation \(\mathbb{E}_{N, p} [\text{Re} (f_i f_j)]\) of §2 converges, along the ordered limit, to
Proof sketch. Expand the product into four point moments of the Wilson family. For large \(t\) the union of two translated grid words is generic, and the point limits of the Seed apply term by term. The reflection acts before the shift, so the shift stays outside it. Lean: Work.B04.answer, Grid.Expects, Grid.event.
Lemma V.1.n (phase reflection; B05).
Hypotheses. The Seed, the grid hypothesis, \(O (4)\)-covariance, V.1.h and V.1.i.
Statement. \(M (w) = M (\vartheta w)\) for every separated word with arbitrary labels.
Proof sketch.
(1) Real labels at generic grid points. Here \(\Theta w\) is a permutation of \(\vartheta w\), so \(M (w)^{\ast} = M (\vartheta w)\). Hence \(D := M (w) - M (\vartheta w)\) satisfies \(D^{\ast} = - D\) and \(a := \langle \Omega, D^2 \Omega \rangle = - \|D \Omega \|^2 \leqslant 0\).
(2) The variance bound. Fix \(q\) and \(\varkappa\). For all large \(t\), the approximants \(g_0, \ldots, g_q\) satisfy, late in the cutoff,
First take the ordered cutoff limit using V.1.m. Then let \(t \to \infty\); V.1.j gives \(e_{ij} (t) \to a\), so
Let \(q \to \infty\), then \(\varkappa \to \infty\). This gives \(a \geqslant 0\), hence \(D \Omega = 0\) and \(D = 0\) by V.1.k.
(3) Extension to all configurations. Approximate a generic configuration by grid configurations of finer levels; the grid hypothesis provides a cofinal set of levels. Both sides are continuous (V.1.i), so equality holds on the co-null set of generic configurations, and then on \(\text{Conf}_n\).
(4) Complex labels. Extend by multilinearity, since real basis vectors span.
Lean: Work.B05.answer, B05.real_depot, Grid.horizon, Grid.estimates, Grid.Covering.square in Phase/Square.lean, Grid.anti.
Lemma V.1.o (invariance under \(O (4)\); B06). Given phase reflection and the rotation laws, \(M (w) = M (Kw)\) for all \(K \in O (4)\). Proof sketch.
Products. The set of \(K\) with \(M (\cdot) = M (K \hspace{0.17em} \cdot)\) is closed under products. It contains the time reflection, because by the rotation law \(R (\vartheta)\) is the mirror action on \(F_d (S)\) and V.1.n applies.
Two-point functions. \(\langle p, x \rangle = \langle Kp, Kx \rangle\) for every \(K\). This is V.1.k's simultaneous-rotation step. If \(t (Ke_t) > 0\) use \(u = e_t\); if \(t (Ke_t) = 0\) use \(u = e_t + K^{- 1} e_t\); if \(t (Ke_t) < 0\) compose with the time reflection.
Conjugates of the time reflection. Rotating a pair back by \(J^{- 1}\), reflecting the second factor and rotating forward gives \(\langle p, x \rangle = \langle p, J \vartheta J^{- 1} x \rangle\) for all \(p\). Since \(M (\Theta p) = M (p)^{\ast}\), this gives \(M (x) \Omega = M (J \vartheta J^{- 1} x) \Omega\), hence equality by separation.
Generation. Every hyperplane reflection is such a conjugate, and reflections generate \(O (4)\) (Cartan–Dieudonné). No separate treatment of time–space rotations is needed.
Lean: Work.B06.answer, auxiliary lemmas in Work/B06.lean.
Lemma V.1.p (phase tables, symmetries, barycenter; B07).
Definition. \(S^{\xi} (x ; v) = \chi_{\xi} (M (x ; v))\) is a table.
Properties.
It is basic: permutation symmetry, \(S^{\xi} (\emptyset) = 1\), unit labels.
It is translation invariant, respects \(S\) (labels in \(N_d (S)\) give \(M = 0\)), and is \(O (4)\)-covariant (V.1.o).
It is real-spin. Signed permutations act through \(R\) by the rotation law. Reality: \(S^{\xi} (x ; \bar{v}) = \chi_{\xi} (M (\bar{w})) = \chi_{\xi} (M (\vartheta \bar{w})) = \chi_{\xi} (M (\Theta w)) = \overline{\chi_{\xi} (M (w))}\), using V.1.n and a permutation.
It is smooth: \(\chi_{\xi}\) is extended by Hahn–Banach to a bounded functional on \(B (\mathcal{H}_0)\) and composed with the smooth map \(x \mapsto M (x ; v)\).
Barycenter. \(\int S^{\xi} (w) \hspace{0.17em} d \nu = \langle \Omega, M (w) \Omega \rangle = S (w)\), by V.1.l and V.1.i.
Lean: Work.B07.answer, B07.state_spin, B07.state_smooth, Resolution.barycenter.
Lemma V.1.q (reflection positivity in every phase; B08). Every \(S^{\xi}\) is reflection positive for every axis, in point and smeared form. Proof sketch.
Time axis. Given finitely many positive words \(w_i\), choose \(d\) so they all lie in \((0, d)\). By V.1.i, \(M (\Theta w_i \cdot w_j) = Y_i^{\ast} Y_j\) with \(Y_i = T_d (w_i) |_{\mathcal{H}_0} : \mathcal{H}_0 \to \mathcal{H}\). Therefore \(\sum \bar{c}_i c_j M (\Theta w_i \cdot w_j) = Y^{\ast} Y \geqslant 0\), where \(Y = \sum c_i Y_i\). Spectral permanence makes its character value nonnegative.
Other axes. Real-spin exchanges the axis with time.
Smeared form. Realize the positive point kernel in a Hilbert space. A compactly supported test gives the integral of its word vectors against the test function. Smoothness makes these vector-valued functions continuous and integrable; the smeared quadratic form is the squared norm of a finite sum of these integrals.
Lean: Work.B08.answer, B08.char_nonneg, B08.smear.
Lemma V.1.r (derivative, inclusion and stress laws in every phase; B09). Every \(S^{\xi}\) satisfies the inclusion and derivative identities of the plan and the stress-tensor Ward identities. Proof sketch.
The evaluator. There is a bounded functional \(\varphi_{\xi}\) on \(B (\mathcal{H})\) with \(S^{\xi} (x ; v) = \varphi_{\xi} (T_D (w (x) + s))\) on a padded neighbourhood of each configuration; it is \(\chi_{\xi}\) extended by Hahn–Banach, composed with compression.
Derivatives and inclusions. The identities hold among slab operators. Derivatives come from V.1.g; pairing with positive-word vectors reduces inclusions to the identities of \(S\).
Stress laws. Choose one shift \(s\) and width \(D\) that put the configuration and the support of a smooth compactly supported window \(\chi\) in \((0, D)\), with \(\chi = 1\) near the distinguished point \(x_0\) and all spectators outside its support. The operators \(- \sum_a \int \partial_a \chi (y) \hspace{0.17em} T_D ((J_{ab} (y) B (x_0) \cdots) + s) \hspace{0.17em} dy\) are Bochner integrals. The shell weight vanishes near every possible collision, which gives integrability. Pairing with positive-word vectors adds only exterior points outside the window, so the Ward identities of \(S\) apply. The angular law uses the additional weight \((g (y - x_0))_b\) for a skew matrix \(g\). Apply \(\varphi_{\xi}\) to both operator identities.
Lean: Work.B09.answer, B09.has_evaluator, B09.Evaluates.flux.
Lemma V.1.s (caps in every phase; B10, B11).
(i) Uniform zeroth cap: V.1.6(c).
(ii) A smooth basic table that respects \(S\), satisfies the inclusion and derivative identities, and has the zeroth cap for every fixed list of labels, has caps on all derivatives.
Proof sketch.
(i) Translate \(w\) into a slab with buffer \(r\) and unit padding. Then \(| \chi_{\xi} (M (w)) | \leqslant \|M (w)\| \leqslant \|T_d (w)\|\), and the all-copy bound applies; the constants depend only on \((D, \rho)\).
(ii) A coordinate derivative replaces one label by its derivative label, raising its degree by one. Induct on the order, and bound each multilinear derivative norm by the sum over its finitely many coordinate directions. The constants may depend on the arity, labels and derivative order, but not on the positions or \(r\).
Lean: Work.B11.answer, B11.phase_cap, Work.B10.answer, B10.capped_tags, Bid.OneCap.
Lemma V.1.t (moment bounds give bounds in every phase). Let \(\tau \geqslant 0\) and \(p \in \mathcal{A}\). If \(\langle \Omega, (p^{\ast} p)^n \Omega \rangle \leqslant \tau^{2 n}\) for every \(n \in \mathbb{N}\), then \(| \chi_{\xi} (p) | \leqslant \tau\) for every \(\xi\). Proof sketch.
The hypothesis says \(\int | \chi_{\xi} (p) |^{2 n} d \nu \leqslant \tau^{2 n}\) for all \(n\).
If \(| \chi_{\xi_0} (p) | > \tau' > \tau\), then continuity gives an open \(O \ni \xi_0\) on which \(| \chi_{\xi} (p) | > \tau'\).
Full support gives \(\nu (O) > 0\), so \(\nu (O) {\tau'}^{2 n} \leqslant \tau^{2 n}\) for all \(n\), a contradiction.
Lean: Retain.bit, Retain.Tall.
Lemma V.1.u (detectors; B15–B18).
The detectors. Assume the Seed and phase reflection. Fix a degree \(d\), an index \(k\) of its corrected-field data, and a test row with a basis index. Let \(F\) be a one-point word at a prescribed grid position, with free value \(s\) at the rescaled position. For a two-point row, let \(G\) be its second one-point word, \(s'\) its free value and \(\gamma\) the free cumulant. There are detectors in \(\mathcal{A}\):
the one-point detector \(\Delta_1 = M (F) - s \cdot 1\);
the two-point detector \(\Delta_2 = M (F \cdot G) - s \hspace{0.17em} M (G) - s' M (F) + (ss' - \gamma) \cdot 1\).
Statement. For every slot there is a one-point detector; for a two-point row there is also the two-point detector. Each satisfies \(\langle \Omega, (\Delta^{\ast} \Delta)^n \Omega \rangle \leqslant \text{gap}_k^{2 n}\) for every \(n \in \mathbb{N}\), where \(\text{gap}_k \geqslant 0\) and \(\text{gap}_k \to 0\).
Proof sketch. Each detector is the cluster element \(c_E = \sum_i \lambda_i M (\text{word}_i)\) of a finite lattice episode \(E\), whose chessboard budget is \(\text{gap}_k\) (B18, from the corrected-field data). Then:
B15 (cutoff level). The absolute lattice expectation of the product of \(2 n\) diagonal translates of the episode function, alternately conjugated, is at most \((\text{gap}_k + \varepsilon)^{2 n}\) late in the cutoff. This is the chessboard estimate.
B16 (limit table). For fixed \(n\) and all sufficiently large translations, expand the product into finitely many monomials. Their points are generic: within each cluster this is part of the data, and between clusters each coordinate tie excludes at most one translation parameter. Point limits apply to each monomial, including the conjugated cutoff labels.
B17 (vacuum moments). Let the translation tend to infinity and apply V.1.j to each ordered monomial, using \(M (\bar{w}) = M (w)^{\ast}\) from V.1.i and V.1.n. The limit is \(\langle \Omega, (c_E^{\ast} c_E)^n \Omega \rangle\). Finally let \(\varepsilon \downarrow 0\).
Lean: Work.B15.answer, B16.answer, B17.answer, B18.answer, Bid.episode; definitions Bid.Episode, Episode.Sold, Retain.Δone, Retain.Δtwo, Retain.Funds.
Lemma V.1.v (retention of null spaces). Assume the corrected-field data and detector bounds of V.1.u. If a table \(S'\) is symmetric, respects \(S\), and satisfies \(S' (w) = \chi_{\xi} (M (w))\) for every separated word and some fixed \(\xi\), then \(N_d (S') = N_d (S)\) for every \(d\). Proof sketch.
Comparison maps. Let \(F_d^{\text{free}}\) be the free-field space in the corrected-field data. Define \(U_k : F_d^{\text{free}} \to V_d\) by sending its chosen basis to the corrected labels. Compose \(U_k\), the quotient map for \(S'\), and the finite family of \(S'\)-probes to obtain \(f_k : F_d^{\text{free}} \to \mathbb{C}^m\).
Injectivity. V.1.t and V.1.u give one-point errors at most \(\text{gap}_k\) and two-point errors at most \(\text{gap}_k (1 + |s| + |s' |)\). The free singleton values converge and are bounded, and the rescaled test points converge. Thus \(f_k\) converges in norm to the injective free test map. Injectivity is open in this finite-dimensional operator space, so \(f_k\) is injective for large \(k\).
Conclusion. For all large \(k\), \(q_d U_k\) is also surjective by the spanning hypothesis. If \(v \in N_d (S')\), choose \(z \in F_d^{\text{free}}\) with \(q_d U_k z = q_d v\). Since \(S'\) respects \(S\), \(U_k z \in N_d (S')\), hence \(f_k z = 0\). Injectivity gives \(z = 0\), so \(q_d v = 0\). The reverse inclusion is the respect hypothesis.
Lean: Retain.lift_detect, Retain.result, Retain.pair_near.
Lemma V.1.w (slab operators in a phase chamber; B13). For every \(\xi\), every time-axis chamber of \(S^{\xi}\) carries a slab family with \(\|T^{\xi}_d (w)\| \leqslant \|T_d (w)\|\). Proof sketch.
The train in the phase. In the phase chamber the train Gram is \(S^{\xi} (\Theta v_n \cdot v_n) = \chi_{\xi} (Y (v_n)^{\ast} Y (v_n))\), where \(Y (v_n) = (T_d (w)^{\ast} T_d (w))^n Y (v)\).
Bench in the phase. Hence \(|S^{\xi} (\Theta v_n \cdot v_n) | \leqslant \| \chi_{\xi} \| \hspace{0.17em} \|Y (v)\|^2 \hspace{0.17em} \Delta^{4 n}\) for every \(\Delta \geqslant \|T_d (w)\|\). This is the bench in the phase, with a constant depending on \(v\).
Conclusion. V.1.e gives norm \(\leqslant \Delta\) for every \(\Delta > \|T_d (w)\|\); the constant depending on \(v\) drops out. No operator inequality between the chambers is used.
Lean: Work.B13.answer, B13.buys.
Lemma V.1.x (one countable family; B14). There is a single sequence \((b_k)\) of positive words, depending only on the label register and the time axis, whose vectors have dense linear span in every OS Hilbert space realizing the positive-word Gram form of a symmetric smooth table. Proof sketch.
Parametrization. For fixed arity and degrees, positive words are parametrized by a separable space: positive separated configurations times label coefficients. Choose countable dense subsets, together with the empty word.
Continuity. Inner products of the vectors are values of the table at combined configurations, which are continuous in the points (smoothness) and in the coefficients (finite-dimensional multilinearity). So the vector map is continuous, and the closed span of the countable family contains all vectors.
Lean: Work.B14.universal, B14.answer.
Lemma V.1.y (almost every phase has a unique ground state).
Statement. In the setting of V.1.8, for each positive \(b\), \(P^{\xi}_0 |b \rangle_{\xi} = S^{\xi} (b) \hspace{0.17em} \Omega_{\xi}\) for \(\nu\)-almost every \(\xi\). This proves the uniqueness assertion of V.1.8.
Proof sketch.
The variance identity. Let \(\omega_{\xi} = \|P_0^{\xi} |b \rangle_{\xi} \|^2 = \lim_{n \to \infty} S^{\xi} (\Theta \tau_n b \cdot \tau_n b)\). For \(n \in \mathbb{N}\) these functions of \(\xi\) are continuous, nonnegative, decreasing in \(n\) and dominated by their value at \(n = 0\), which is integrable on compact \(X\). Their integrals are the corresponding quantities for \(S\). Dominated convergence gives \(\int \omega_{\xi} \hspace{0.17em} d \nu = \|P_0 |b \rangle \|^2 = \|M (b) \Omega \|^2 = \int | \chi_{\xi} (M (b)) |^2 d \nu\).
Almost-everywhere vanishing. Since \(\|P_0^{\xi} |b \rangle - S^{\xi} (b) \Omega_{\xi} \|^2 = \omega_{\xi} - |S^{\xi} (b) |^2 \geqslant 0\) has integral zero, it vanishes a.e.
Uniqueness. Apply this to the family \((b_k)\) of V.1.x; the union of the countably many null sets is null. Outside it, \(P_0^{\xi}\) maps a dense set into \(\mathbb{C} \Omega_{\xi}\), so \(\mathcal{H}_0^{\xi} =\mathbb{C} \Omega_{\xi}\). Since \(\nu (X) = 1\), such \(\xi\) exist; the choice is not canonical.
Lean: Cluster.exp_real, Cluster.champ, Cluster.fits, Cluster.lifts, Cluster.selects.
Lemma V.1.z (assembly). V.1.9 holds. Proof sketch.
Slab and phase constructions. B01 and B02 give the original slab family with its laws. V.1.h gives spatial invariance; B03 gives the compression identities; B04 and B05 give phase reflection; B15–B18 give the detector bounds; and B12 gives separability. For any decomposition, B06 and B07 construct the tables, and B08–B11 give their hierarchy and field laws. B13 and B02 give slab families in the phase chambers, and B14 supplies a common countable total family.
The certificate. Choose a chamber for every phase, select \(\xi_{\ast}\) by V.1.y, and apply V.1.v to obtain \(N_d (S^{\xi}) = N_d (S)\) for every \(\xi, d\).
Lean: Bid.Looms, Phase.Loom.cleanse, Pert.withFinal.
Distinctive arguments.
The decomposition uses the OS space of the limit correlations. Three properties require separate arguments:
spatial invariance of the ground states (V.1.h);
commutativity of \(\mathcal{A}\) (V.1.k);
reality of phase values (V.1.n).
Lattice input after the limit. It enters at three places only: the bench (V.1.7), phase reflection (V.1.n, through the grid hypothesis), and the detectors (V.1.u).
Phase reflection is the least standard argument. It is a variance-positivity statement for far-separated copies of a real lattice variable. It is equivalent to the claim that the abstract characters assign real values to real observables.
Detectors. They require quantitative UV input: one- and two-point chessboard deviations \(\text{gap}_k \to 0\) of corrected fields from free values. These estimates preserve the field null spaces in every phase.
Routine but long.
B02, B03, B09, B12, B14, B16, B18: smooth dependence on positions, label quotients, finite polynomial limits, separability.
SNAG, Paley, Heat (about 1,100 lines) reprove textbook theorems in an explicit form adapted to the library.
Unusual or worth checking.
Selection. The selected phase comes from a measure-theoretic existence argument and is not canonical. No finite bound on the number of phases is asserted. V.3 later proves unique ground states for every phase.
Sub-width in the bench. The bench gives the train bound only at a lattice-compatible width \(s \leqslant d\). The remaining width is absorbed by the contraction \(E_{d - s}\) (V.1.f).
Which symmetries are used where.
Commutativity uses covariance only under the exchange of the time axis with one spatial axis. This is a lattice symmetry, already contained in the real-spin part of the hierarchy.
Spatial invariance of \(\mathcal{H}_0\) needs arbitrary hyperplane reflections, i.e. the continuum \(O (4)\) action of VI.1. In the Lean, both steps draw on that action.
Both use improper elements of \(O (4)\) (reflections), not only rotations.
Scope of the certificate. Its unique-ground-state assertion concerns the time-axis chamber of the selected phase. It also records the hierarchy, field laws and null-space equality for every phase. The continuation linked to the selected Wightman realization uses that selected table.
Smoothness hypothesis in B02. B02 needs the derivative identities of the table on which the slab family lives. For phases these come from B09, which uses the slab laws of \(S\), so the two applications of B02 are ordered.
| object | code name | proposed notation |
| limit table, value on a word | Table, S.expect w | \(S = S^{\text{mix}}\), \(S (w)\) |
| OS reflection; linear time reflection of a word | reflect w timeAxis; Grid.anti | \(\Theta w\); \(\vartheta w\) |
| OS space, vector of a positive word | GNS.H, GNS.plant | \(\mathcal{H}\), \(\lvert b \rangle\) |
| chamber | Chamber | \((\mathcal{H}, E_t, U_y)\) |
| ground space, projection, vacuum | Semigroup.ground, orth, omega0 | \(\mathcal{H}_0\), \(P_0\), \(\Omega\) |
| real-time group; 4-parameter group; spectral measure | edge; unit; Orbit.P | \(e^{isH}\); \(V (x)\); \(P (dp)\) |
| slab, buffer | Slab, buffer | slab \((0, d)\), \(r\)-buffered |
| slab operators ("Spread") and all-copy bound | Spread, place, AllCopy, Laws | \(T_d (w)\); \(C^n (n / r)^{\kappa n}\) |
| bench, train bound | Bench, Seats, QRead | bench (V.1.7) |
| spatial invariance of \(\mathcal{H}_0\) | Steady | V.1.4(a) |
| compression; its multilinear form; algebra | M, atom; depot; Spread.algebra | \(M (w)\); \(M (x ; v)\); \(\mathcal{A}\) |
| identities of \(M\) ("Theme") | Theme | V.1.4(e) |
| phase reflection | Phase.reflects | \(M (w) = M (\vartheta w)\) |
| decomposition ("Resolution") | Spread.Resolution, X, μ, character | \(X\), \(\nu\), \(\chi_{\xi}\) |
| phase tables; their properties | ensemble, Garland, Scripts | \(S^{\xi}\) |
| \(S^{\xi}\) receives \(\chi_{\xi}\) | Resolution.receives | \(S^{\xi} (w) = \chi_{\xi} (M (w))\) |
| grid hypothesis, bounded approximants | Org.Grill, Grid.Covering | grid hypothesis |
| lattice episode, cluster element, budget | Episode, cluster, Sold | \(E\), \(c_E\) |
| detectors; moment bound | Δone, Δtwo; Tall, Funds | \(\Delta_1, \Delta_2\) |
| countable dense family | Cluster.horizon | \((b_k)\) |
| unique ground state | UniqueHeat | \(\mathcal{H}_0^{\xi} =\mathbb{C} \Omega_{\xi}\) |
| selected phase and certificate | Selection, atPoint, Certificate, Materializes | \(\xi_{\ast}\); V.1.9 |
| bundles of obligations | Loom, Stall, Vendor | assembly assumptions |
The measure \(\nu\) corresponds to \(\mu\) in the Lean.
Ultraviolet inputs. The chessboard geometry below Next.Phase.travel_price belongs to VII.4. Its train estimate and the subsequent scale choice give the bench used here; the supporting packing argument remains outside this unit. The bounded-approximation input to Accent.grilled belongs to III.11. Its output has the cutoff-independent bounds required in V.1.n, but the underlying ultraviolet estimates and the production of the corrected-field data remain inputs.
Order of limits. In V.1.n, fix the number of copies and approximation index, then fix a sufficiently large translation, take the ordered cutoff limit, and let the translation tend to infinity. Only then let the number of copies tend to infinity and the approximation error tend to zero. The detector argument similarly takes the cutoff limit before the large-translation limit and sends its budget slack to zero last.
Later vacuum uniqueness. V.1.8 gives uniqueness almost everywhere, which suffices to select \(\xi_{\ast}\). The stronger result for every phase, EveryPhase.every, uses reconstruction in V.3; its proof is outside this unit.
Placement of the spectral calculus. V.1.2 supplies V.2 and V.3 and can be placed at the start of reconstruction in the written proof. The \(O (4)\) action from VI.1 is needed only after the preliminary chamber and heat constructions.
Bibliography. The reference locations marked “probable” in §2 need bibliographic verification.