Documentation

CsdLean4.CV.SupportSpreading

CV-8 (ii)–(iv): support spreading and the coupling-graph light cone #

Category: CV (continuous variables — the multi-mode field).

CV-6 proved the free field never spreads support. This module bounds what an interaction can do:

This retires CV-6's open boundary: interacting-drive spreading is now a theorem with a bound, not an unclaimed scope. The norm pricing landed as CV-9 (CV/InteractionPrice.lean, covering non-diagonal perturbations too), and the non-diagonal KICKED cone landed as CV-11 (CV/LocalAlgebraClosed.lean, the exp-closure route delivered); what remains open is the full-exponential cone — Lieb–Robinson velocity bounds, the promoted Stage-5 headline (specs/eft-stage4-plan.md, horizon note).

References #

CV/LocalAlgebra.lean (CV-8 (i)); CV/DynamicalLocality.lean (CV-6, the free case); CV/Interaction.lean (CV-7, interactingU); specs/cv-stage3-plan.md §3b; specs/future-work.md (row CV-8). Haag, Local Quantum Physics (1992); Lieb–Robinson (1972) for the continuum-flavoured analogue of the light cone.

Conjugation by a supported unitary #

theorem CSD.CV.phaseDiagU_supportedOn {K N : } {T : Finset (Fin K)} {f : FieldConfig K N} (hf : ∀ (c c' : FieldConfig K N), (∀ kT, c k = c' k)f c = f c') :

A diagonal phase depending only on the modes in T is a T-supported unitary.

theorem CSD.CV.heisenberg_supportedOn_union {K N : } {S T : Finset (Fin K)} {U : (Matrix.unitaryGroup (FieldConfig K N) )} {A : Matrix (FieldConfig K N) (FieldConfig K N) } (hU : SupportedOn T U) (hA : SupportedOn S A) :

Conjugation by a T-supported unitary maps the S-algebra into the S ∪ T-algebra — from the LocalAlgebra closure alone.

theorem CSD.CV.heisenberg_eq_of_disjoint {K N : } {S T : Finset (Fin K)} (hST : Disjoint S T) {U : (Matrix.unitaryGroup (FieldConfig K N) )} {A : Matrix (FieldConfig K N) (FieldConfig K N) } (hU : SupportedOn T U) (hA : SupportedOn S A) :

A coupling disjoint from the support acts trivially: support only grows into modes the interaction actually couples.

The coupling graph #

def CSD.CV.graphNeighborhood {K : } (E : Finset (Fin K × Fin K)) (R : Finset (Fin K)) :

The one-step neighbourhood: R together with both endpoints of every edge that touches R.

Equations
Instances For
    theorem CSD.CV.subset_graphNeighborhood {K : } (E : Finset (Fin K × Fin K)) (R : Finset (Fin K)) :
    theorem CSD.CV.mem_graphNeighborhood_fst {K : } {E : Finset (Fin K × Fin K)} {R : Finset (Fin K)} {e : Fin K × Fin K} (he : e E) (ht : e.1 R e.2 R) :
    theorem CSD.CV.mem_graphNeighborhood_snd {K : } {E : Finset (Fin K × Fin K)} {R : Finset (Fin K)} {e : Fin K × Fin K} (he : e E) (ht : e.1 R e.2 R) :
    def CSD.CV.graphBall {K : } (E : Finset (Fin K × Fin K)) (R : Finset (Fin K)) :
    Finset (Fin K)

    The n-ball: n applications of the one-step neighbourhood — the light cone of R after n periods.

    Equations
    Instances For
      @[simp]
      theorem CSD.CV.graphBall_zero {K : } (E : Finset (Fin K × Fin K)) (R : Finset (Fin K)) :
      graphBall E R 0 = R
      @[simp]
      theorem CSD.CV.graphBall_succ {K : } (E : Finset (Fin K × Fin K)) (R : Finset (Fin K)) (n : ) :
      theorem CSD.CV.graphBall_mono {K : } (E : Finset (Fin K × Fin K)) (R : Finset (Fin K)) {i j : } (hij : i j) :
      graphBall E R igraphBall E R j

      The balls grow with the period count.

      The phase-difference collapse for edge sums #

      theorem CSD.CV.graph_sum_sub_of_agree {K N : } {R : Finset (Fin K)} (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) {x y : FieldConfig K N} (hxy : kR, x k = y k) :
      eE, g e (y e.1) (y e.2) - eE, g e (x e.1) (x e.2) = eE with e.1 R e.2 R, (g e (y e.1) (y e.2) - g e (x e.1) (x e.2))

      Off-R agreement collapses an edge-sum difference to the edges that touch R.

      The conjugating phase pair of any real phase: e^{+ia}·e^{-ib} = e^{-i(b−a)}.

      theorem CSD.CV.addGraph_phase_diff_eq {K N : } {R : Finset (Fin K)} (h : Fin KFin N) (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) {x y x' y' : FieldConfig K N} (hxy : kR, x k = y k) (hxy' : kR, x' k = y' k) (hxx' : kgraphNeighborhood E R, x k = x' k) (hyy' : kgraphNeighborhood E R, y k = y' k) :
      k : Fin K, h k (y k) + eE, g e (y e.1) (y e.2) - (k : Fin K, h k (x k) + eE, g e (x e.1) (x e.2)) = k : Fin K, h k (y' k) + eE, g e (y' e.1) (y' e.2) - (k : Fin K, h k (x' k) + eE, g e (x' e.1) (x' e.2))

      The phase difference of a mode-additive-plus-edge-sum phase depends only on the modes of graphNeighborhood E R, for configuration pairs agreeing off R.

      The one-period spreading bound #

      theorem CSD.CV.heisenberg_addGraphPhase_supportedOn {K N : } {R : Finset (Fin K)} {A : Matrix (FieldConfig K N) (FieldConfig K N) } (h : Fin KFin N) (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) (hA : SupportedOn R A) :
      SupportedOn (graphNeighborhood E R) (heisenberg (phaseDiagU fun (c : FieldConfig K N) => k : Fin K, h k (c k) + eE, g e (c e.1) (c e.2)) A)

      One period of a diagonal interacting drive spreads support at most onto the one-step neighbourhood: for a phase = mode-additive part + edge sum over E, conjugation maps the R-algebra into the graphNeighborhood E R-algebra.

      The interacting drive and its light cone #

      def CSD.CV.graphPotential {K N : } (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) :

      The graph potential: a sum of pair couplings over the edges of E.

      Equations
      Instances For
        noncomputable def CSD.CV.graphInteractingU (K N : ) (τ lam : ) (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) :

        The graph-interacting drive: the CV-7 interacting step at the graph potential — the free field plus pair couplings along the edges of E.

        Equations
        Instances For
          theorem CSD.CV.heisenberg_graphInteractingU_supportedOn {K N : } {R : Finset (Fin K)} {A : Matrix (FieldConfig K N) (FieldConfig K N) } (τ lam : ) (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) (hA : SupportedOn R A) :

          One interacting period spreads support at most one graph neighbourhood.

          theorem CSD.CV.heisenberg_graphInteractingU_pow_supportedOn {K N : } {R : Finset (Fin K)} {A : Matrix (FieldConfig K N) (FieldConfig K N) } (τ lam : ) (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) (n : ) (hA : SupportedOn R A) :
          SupportedOn (graphBall E R n) (heisenberg (graphInteractingU K N τ lam E g ^ n) A)

          ★★ The light cone: after n periods of the interacting drive, support lies inside the coupling graph's n-ball — information propagates at most one graph edge per period.

          theorem CSD.CV.commute_heisenberg_graphInteractingU_pow {K N : } {R T : Finset (Fin K)} {A B : Matrix (FieldConfig K N) (FieldConfig K N) } (τ lam : ) (E : Finset (Fin K × Fin K)) (g : Fin K × Fin KFin NFin N) (n : ) (hRT : Disjoint (graphBall E R n) (graphBall E T n)) (hA : SupportedOn R A) (hB : SupportedOn T B) :
          heisenberg (graphInteractingU K N τ lam E g ^ n) A * heisenberg (graphInteractingU K N τ lam E g ^ n) B = heisenberg (graphInteractingU K N τ lam E g ^ n) B * heisenberg (graphInteractingU K N τ lam E g ^ n) A

          ★★ Locality outside the light cones: observables whose n-balls are still disjoint — not yet causally connected through the coupling graph — still commute after n interacting periods.

          Non-vacuity: interaction genuinely spreads #

          noncomputable def CSD.CV.spreadPotential :

          The witness pair potential at K = N = 2: phase π exactly when both modes are excited.

          Equations
          Instances For

            The witness kick: the diagonal unitary of spreadPotential.

            Equations
            Instances For
              noncomputable def CSD.CV.raiseMat :
              Matrix (Fin 2) (Fin 2)

              The witness single-mode operator: the raising matrix |1⟩⟨0|-style !![0, 1; 0, 0].

              Equations
              Instances For

                Non-vacuity — the converse of CV-6: the pair kick moves the single-mode operator OFF the single-mode algebra. The evolved entries at spectator occupation 0 and 1 are 1 and −1, so indep fails: interaction genuinely spreads support, and the light cone bounds a real phenomenon.