Documentation

LeanPool.ArchonFirstProofResults.FirstProof6.Problem6

Problem 6: Large epsilon-light vertex subsets -- Main Proof #

This file contains the main theorem and its supporting lemmas. All auxiliary infrastructure is in Problem6Aux.lean.

Main theorems #

BSS coloring infrastructure #

The BSS framework follows the informal proof's dynamic approach:

  1. Spectral setup: Lhalf with Lhalf² = L (via spectral theorem)
  2. Normalized edge Laplacians A_e = Lhalf_pinv * L_e * Lhalf_pinv
  3. Coloring with one-sided barrier tracking monochromatic matrix M_t
  4. Pullback: M_k ≤ u_k·I → each color class ε-light (via eps_light_of_loewner_bound)
theorem Problem6.unitary_diag_mul {V : Type u_1} [Fintype V] [DecidableEq V] (U : Matrix V V ) (hstarU : star U * U = 1) (f g : V) :

Product of unitary-diagonal-conjugate matrices: (U * D(f) * U★) * (U * D(g) * U★) = U * D(f·g) * U★. Used for spectral calculus (computing products via eigenvalue multiplication).

theorem Problem6.spectral_sqrt_exists {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] :
∃ (Lhalf : Matrix V V ), Lhalf.IsHermitian Lhalf * Lhalf = graphLaplacian G

Sub-lemma 1 (Spectral setup): For any graph G, there exists a Hermitian matrix Lhalf with Lhalf * Lhalf = graphLaplacian G (the PSD square root), constructed via the spectral theorem. This is Step 1 of the informal proof. The spectral decomposition L = U * diag(λ) * U* gives Lhalf = U * diag(√λ) * U*, which satisfies Lhalf² = L and IsHermitian.

theorem Problem6.hermitian_pseudo_inverse_exists {V : Type u_1} [Fintype V] (M : Matrix V V ) (hM : M.IsHermitian) :
∃ (M_pinv : Matrix V V ), M_pinv.IsHermitian M * M_pinv * M = M M_pinv * M * M_pinv = M_pinv M * M_pinv * (M * M_pinv) = M * M_pinv M * M_pinv = M_pinv * M

For a Hermitian matrix M, there exists a Hermitian pseudo-inverse M_pinv satisfying the Moore-Penrose conditions: (1) M * M_pinv * M = M (2) M_pinv * M * M_pinv = M_pinv (3) M * M_pinv is idempotent (a projection) Constructed via spectral decomposition: if M = U * diag(λ) * U★, then M_pinv = U * diag(λ⁻¹) * U★ where λ⁻¹(i) = 0 if λ(i) = 0.

Dynamic BSS coloring (Steps 1–5 combined) #

The informal proof (Steps 3–5) uses a dynamic coloring process:

  1. Per-edge matrices: A_e := L^{-1/2} L_e L^{-1/2} with ∑_e A_e = I on range(L)
  2. Track M_t := ∑_{monochromatic edges at time t} A_e (DYNAMIC, depends on current coloring)
  3. At time t: color vertex v with γ, increment M_{t+1} = M_t + B_v^γ where B_v^γ = ∑{u ∈ T, col(u)=γ, uv∈E} A{uv}
  4. BSS barrier (Lemma 6.1) maintains M_t ≺ u_t·I at each step
  5. After k = n/4 steps: M_k = ∑a L^{-1/2} L{S_a} L^{-1/2} (by construction)
  6. Since M_k ≺ u_k·I ≺ ε·I, each color class S_a is ε-light The proof is decomposed into: (A) dynamic_barrier_coloring: The core BSS dynamic coloring induction that produces a PartialColoring with per-color barrier bounds on normalized monochromatic sums. (B) bss_dynamic_coloring: Combines (A) with pinv_pullback_eq and eps_light_of_loewner_bound to convert barrier bounds to ε-lightness.
theorem Problem6.cross_edge_off_diagonal {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (r : ) (pc : PartialColoring V r) (S : Fin rFinset V) (hS : S = fun (γ : Fin r) => {wpc.colored | pc.color w = γ}) (G_same : SimpleGraph V) [DecidableRel G_same.Adj] (hG_same : ∀ (a b : V), G_same.Adj a b G.Adj a b (a pc.colored b pc.colored)) (DS : Matrix V V ) (hDS : DS = v : V with vpc.colored, γ : Fin r, (inducedLaplacian G (S γ {v}) - inducedLaplacian G (S γ))) (i j : V) (hij : i j) :
(graphLaplacian G - DS) i j = graphLaplacian G_same i j

Off-diagonal entry equality for the cross-edge decomposition: for i ≠ j, (graphLaplacian G - DS) i j = (graphLaplacian G_same) i j.

theorem Problem6.cross_edge_diagonal {V : Type u_1} [Fintype V] (L DS R : Matrix V V ) (h_off_diag : ∀ (a b : V), a b(L - DS) a b = R a b) (h_row_LHS : ∀ (a : V), b : V, (L - DS) a b = 0) (h_row_RHS : ∀ (a : V), b : V, R a b = 0) (i : V) :
(L - DS) i i = R i i

Diagonal entry derivation for the cross-edge decomposition: derives diagonal from row sums and off-diagonal matching.

theorem Problem6.cross_edge_sum_le_graphLaplacian {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (r : ) (pc : PartialColoring V r) :
(graphLaplacian G - v : V with vpc.colored, γ : Fin r, (inducedLaplacian G ({wpc.colored | pc.color w = γ} {v}) - inducedLaplacian G ({wpc.colored | pc.color w = γ}))).PosSemidef

The cross-edge sum decomposition is PSD: the difference between the graph Laplacian and the double sum ∑{v uncolored} ∑γ (L{S_γ ∪ {v}} - L{S_γ}) is PSD, equaling the Laplacian of the "same-status" subgraph (edges where both endpoints have the same colored/uncolored status).

theorem Problem6.total_barrier_bound_base {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (r : ) (hr_pos : 0 < r) (Lhalf_pinv : Matrix V V ) (h_lap_le1 : ∀ (T : Finset V), T.card 1inducedLaplacian G T = 0) :
∃ (pc : PartialColoring V r), pc.colored.card = 0 ((ε / 2 + 0 * (ε / (Fintype.card V))) 1 - γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc.colored | pc.color v = γ}) * Lhalf_pinv).PosDef barrierPotential (ε / 2 + 0 * (ε / (Fintype.card V))) (∑ γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc.colored | pc.color v = γ}) * Lhalf_pinv) 2 * (Fintype.card V) / ε

Base case of BSS induction: empty coloring, M_0 = 0, Φ_{ε/2}(0) = 2n/ε.

theorem Problem6.good_pair_exists {V : Type u_1} [Fintype V] [DecidableEq V] [Nonempty V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (r : ) (hr_def : r = 16 / ε⌉₊) (hn : 4 Fintype.card V) (Lhalf Lhalf_pinv : Matrix V V ) (hLhalf_herm : Lhalf.IsHermitian) (hpinv_herm : Lhalf_pinv.IsHermitian) (hLhalf_sq : Lhalf * Lhalf = graphLaplacian G) (hMP3 : Lhalf * Lhalf_pinv * (Lhalf * Lhalf_pinv) = Lhalf * Lhalf_pinv) (hMP4 : Lhalf * Lhalf_pinv = Lhalf_pinv * Lhalf) (pc_t : PartialColoring V r) (t : ) (ht : t + 1 Fintype.card V / 4) (hcard_t : pc_t.colored.card = t) (hpd_t : ((ε / 2 + t * (ε / (Fintype.card V))) 1 - γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc_t.colored | pc_t.color v = γ}) * Lhalf_pinv).PosDef) (hphi_t : barrierPotential (ε / 2 + t * (ε / (Fintype.card V))) (∑ γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc_t.colored | pc_t.color v = γ}) * Lhalf_pinv) 2 * (Fintype.card V) / ε) :
have M_t := γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc_t.colored | pc_t.color v = γ}) * Lhalf_pinv; have u_t := ε / 2 + t * (ε / (Fintype.card V)); have u' := ε / 2 + (t + 1) * (ε / (Fintype.card V)); have U := (u' 1 - M_t)⁻¹; ∃ (v₀ : V) (_ : v₀pc_t.colored) (γ₀ : Fin r), have S_γ₀ := {vpc_t.colored | pc_t.color v = γ₀}; have B := Lhalf_pinv * (inducedLaplacian G (S_γ₀ {v₀}) - inducedLaplacian G S_γ₀) * Lhalf_pinv; (B * U).trace < 1 (B * U).trace + (B * U * U).trace / (barrierPotential u_t M_t - barrierPotential u' M_t) 1

BSS averaging: ∃ good (v₀, γ₀) pair with barrier conditions. Core by_contra argument of the BSS dynamic coloring.

theorem Problem6.total_barrier_bound_step {V : Type u_1} [Fintype V] [DecidableEq V] [Nonempty V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (r : ) (hr_def : r = 16 / ε⌉₊) (hn : 4 Fintype.card V) (Lhalf Lhalf_pinv : Matrix V V ) (hLhalf_herm : Lhalf.IsHermitian) (hpinv_herm : Lhalf_pinv.IsHermitian) (hLhalf_sq : Lhalf * Lhalf = graphLaplacian G) (hMP3 : Lhalf * Lhalf_pinv * (Lhalf * Lhalf_pinv) = Lhalf * Lhalf_pinv) (hMP4 : Lhalf * Lhalf_pinv = Lhalf_pinv * Lhalf) (t : ) (ht : t + 1 Fintype.card V / 4) (pc_t : PartialColoring V r) (hcard_t : pc_t.colored.card = t) (hpd_t : ((ε / 2 + t * (ε / (Fintype.card V))) 1 - γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc_t.colored | pc_t.color v = γ}) * Lhalf_pinv).PosDef) (hphi_t : barrierPotential (ε / 2 + t * (ε / (Fintype.card V))) (∑ γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc_t.colored | pc_t.color v = γ}) * Lhalf_pinv) 2 * (Fintype.card V) / ε) :
∃ (pc : PartialColoring V r), pc.colored.card = t + 1 ((ε / 2 + (t + 1) * (ε / (Fintype.card V))) 1 - γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc.colored | pc.color v = γ}) * Lhalf_pinv).PosDef barrierPotential (ε / 2 + (t + 1) * (ε / (Fintype.card V))) (∑ γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc.colored | pc.color v = γ}) * Lhalf_pinv) 2 * (Fintype.card V) / ε

BSS induction step: extend coloring by one vertex using good_pair_exists.

theorem Problem6.total_barrier_bound {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (_hε1 : ε 1) (hn : 4 Fintype.card V) (Lhalf Lhalf_pinv : Matrix V V ) (hLhalf_herm : Lhalf.IsHermitian) (hpinv_herm : Lhalf_pinv.IsHermitian) (hLhalf_sq : Lhalf * Lhalf = graphLaplacian G) (_hMP1 : Lhalf * Lhalf_pinv * Lhalf = Lhalf) (hMP3 : Lhalf * Lhalf_pinv * (Lhalf * Lhalf_pinv) = Lhalf * Lhalf_pinv) (hMP4 : Lhalf * Lhalf_pinv = Lhalf_pinv * Lhalf) :
have r := 16 / ε⌉₊; have k := Fintype.card V / 4; have n := Fintype.card V; have u_k := ε / 2 + k * (ε / n); ∃ (pc : PartialColoring V r), pc.colored.card = k (u_k 1 - γ : Fin r, Lhalf_pinv * inducedLaplacian G ({vpc.colored | pc.color v = γ}) * Lhalf_pinv).PosSemidef

Total barrier bound (Steps 3-4 of informal proof, core BSS): Given the spectral setup, there exists a partial coloring of k = n/4 vertices with r = ⌈16/ε⌉ colors such that the TOTAL monochromatic normalized sum ∑_γ M_γ ≤ u_k · I. The per-color bounds then follow from psd_sub_sum_le (extraction lemma). The proof constructs the coloring step by step using one_sided_barrier:

  1. Track M_t = ∑_γ M_t^γ (total monochromatic matrix, eq. 38)
  2. At step t, for each (v,γ) ∈ R × [r], the increment B_v^γ increases M_t^γ
  3. Averaging: ∑_{v,γ} B_v^γ ≤ P ≤ I (sum of cross-edge contributions)
  4. By trace_mul_le_of_loewner + barrier_potential_decrease + averaging: average barrier cost < 1, so some (v,γ) satisfies the barrier condition (eq. 35)
  5. Apply one_sided_barrier to propagate M_t ≺ u_t·I → M_{t+1} ≺ u_{t+1}·I
  6. After k steps: (u_k·I - M_k).PosDef implies (u_k·I - ∑_γ M_k^γ).PosSemidef
theorem Problem6.dynamic_barrier_coloring {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (hε1 : ε 1) (hn : 4 Fintype.card V) (Lhalf Lhalf_pinv : Matrix V V ) (hLhalf_herm : Lhalf.IsHermitian) (hpinv_herm : Lhalf_pinv.IsHermitian) (hLhalf_sq : Lhalf * Lhalf = graphLaplacian G) (hMP1 : Lhalf * Lhalf_pinv * Lhalf = Lhalf) (hMP3 : Lhalf * Lhalf_pinv * (Lhalf * Lhalf_pinv) = Lhalf * Lhalf_pinv) (hMP4 : Lhalf * Lhalf_pinv = Lhalf_pinv * Lhalf) :
have r := 16 / ε⌉₊; have k := Fintype.card V / 4; have n := Fintype.card V; have u_k := ε / 2 + k * (ε / n); ∃ (pc : PartialColoring V r), pc.colored.card = k ∀ (γ : Fin r), (u_k 1 - Lhalf_pinv * inducedLaplacian G ({vpc.colored | pc.color v = γ}) * Lhalf_pinv).PosSemidef

Dynamic barrier coloring (Steps 3-4 of informal proof): Given the spectral setup (Lhalf, Lhalf_pinv satisfying Moore-Penrose conditions), there exists a partial coloring of k = n/4 vertices with r = ⌈16/ε⌉ colors such that each color class γ has its normalized monochromatic matrix bounded: Lhalf_pinv * L_{S_γ} * Lhalf_pinv ≼ u_k · I. Proof: Apply total_barrier_bound for the total sum bound ∑_γ M_γ ≤ u_k·I, then psd_sub_sum_le to extract per-color bounds (since each M_γ is PSD).

theorem Problem6.bss_dynamic_coloring {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (hε1 : ε 1) (hn : 4 Fintype.card V) (Lhalf : Matrix V V ) (hLhalf_herm : Lhalf.IsHermitian) (hLhalf_sq : Lhalf * Lhalf = graphLaplacian G) :
have r := 16 / ε⌉₊; have k := Fintype.card V / 4; ∃ (pc : PartialColoring V r), pc.colored.card = k ∀ (γ : Fin r), IsEpsLight G ε ({vpc.colored | pc.color v = γ})

Dynamic BSS coloring result (Steps 1–5 of informal proof): Given the spectral setup (Lhalf with Lhalf² = L), there exists a partial coloring of k = n/4 vertices with r = ⌈16/ε⌉ colors such that each color class is ε-light. Proof: Obtain Lhalf_pinv from hermitian_pseudo_inverse_exists, apply dynamic_barrier_coloring for the barrier bound, then use pinv_pullback_eq and eps_light_of_loewner_bound to convert the barrier bound to ε-lightness.

theorem Problem6.bss_coloring_eps_light {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (hε1 : ε 1) (hn : 4 Fintype.card V) :
have r := 16 / ε⌉₊; have k := Fintype.card V / 4; ∃ (pc : PartialColoring V r), pc.colored.card = k ∀ (γ : Fin r), IsEpsLight G ε ({vpc.colored | pc.color v = γ})

The BSS coloring result (correctly stated, dynamic approach). There exists a partial coloring of n/4 vertices with r = ⌈16/ε⌉ colors such that each color class is ε-light. This captures Steps 1–5 of the informal proof:

  1. spectral_sqrt_exists: Construct Lhalf with Lhalf² = L (spectral setup)
  2. bss_dynamic_coloring: Dynamic coloring with per-edge tracking

The coloring induction result #

theorem Problem6.coloring_induction_exists {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (hε1 : ε 1) (hn : 4 Fintype.card V) :
have n := Fintype.card V; have r := 16 / ε⌉₊; have k := n / 4; ∃ (pc : PartialColoring V r), pc.colored.card = k ∀ (γ : Fin r), IsEpsLight G ε ({vpc.colored | pc.color v = γ})

The result of the BSS coloring induction: there exists a partial coloring of k vertices with r colors such that each color class is ε-light. This packages Steps 1-5 of the informal proof, delegating to bss_coloring_eps_light which uses:

  1. spectral_sqrt_exists: Lhalf with Lhalf² = L
  2. bss_dynamic_coloring: Dynamic per-edge coloring with BSS barrier

Proof of the main theorem #

theorem Problem6.eps_light_large_n {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (hε1 : ε 1) (hn : 4 Fintype.card V) :
∃ (S : Finset V), IsEpsLight G ε S ε / 256 * (Fintype.card V) S.card

For n ≥ 4, the coloring process produces an ε-light set of size ≥ εn/256.

theorem Problem6.eps_light_small_n {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) ( : 0 < ε) (hε1 : ε 1) (hn : Fintype.card V < 4) (hne : 0 < Fintype.card V) :
∃ (S : Finset V), IsEpsLight G ε S ε / 256 * (Fintype.card V) S.card

For 1 ≤ n ≤ 3, a single vertex is ε-light and satisfies the size bound.

Main theorem #

theorem Problem6.exists_eps_light_subset {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] (ε : ) (hε_pos : 0 < ε) (hε_le : ε 1) :
∃ (S : Finset V), IsEpsLight G ε S ε / 256 * (Fintype.card V) S.card

Main Theorem (Problem 6): For every simple graph G on a finite vertex set V and every ε ∈ (0, 1], there exists an ε-light subset S with |S| ≥ ε/256 · |V|.