Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Static updates: one-edge energy is a quadratic form in a unit-modulus phase

Open
QAlgorithms.Papers.Independent.Intakethird100.P57.single_edge_energy

by Gohan · Nov 1, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Fix admissible RAGE data with N,D≥1N,D\ge1N,D≥1, nonzero raw MPS, a distinct pair j,kj,kj,k, and a Hermitian Hamiltonian HHH. Undo the final local rotations in HHH and freeze all parameters except this edge. Remove its old phase from the pre-rotation vector and split the resulting vector into its P11P_{11}P11​ and complementary components v,uv,uv,u. With HV=V†HVH_V=V^\dagger HVHV​=V†HV, put c0=⟨u,HVu⟩+⟨v,HVv⟩c_0=\langle u,H_Vu\rangle+\langle v,H_Vv\ranglec0​=⟨u,HV​u⟩+⟨v,HV​v⟩, c1=⟨u,HVv⟩c_1=\langle u,H_Vv\ranglec1​=⟨u,HV​v⟩, and Z=∥M(A)∥2Z=\|M(A)\|^2Z=∥M(A)∥2. For every ∣z∣=1|z|=1∣z∣=1, the trial norm squared is ZZZ and

E(z)=c0+2Re⁡(c1z)Z.E(z)=\frac{c_0+2\operatorname{Re}(c_1z)}{Z}.E(z)=Zc0​+2Re(c1​z)​.

If c1≠0c_1\ne0c1​=0, z∗=−c1‾/∣c1∣z_*=-\overline{c_1}/|c_1|z∗​=−c1​​/∣c1​∣ minimizes this expression; if c1=0c_1=0c1​=0, every unit-modulus phase is optimal. For a supplied sum of mmm terms of fixed support at most sss, scalar programs chosen before the numerical data compute c0,c1,Zc_0,c_1,Zc0​,c1​,Z and the total energy in at most Cs(m+1)ND5C_s(m+1)ND^5Cs​(m+1)ND5 operations. This is a single-edge optimum with a unit-circle constraint.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakethird100_P57
Formal statement
noncomputable section

open scoped ComplexConjugate Matrix

namespace QAlgorithms.Papers.Independent.Intakethird100.P57

attribute [local instance] Classical.propDecidable

/-- PDF p.3, Static updates: a constrained quadratic single-edge minimum, together
with exact coefficient and finite local-Hamiltonian evaluation. The coefficient calculation
costs O((m+1)ND⁵) for m terms of fixed support, using supports enlarged by the two edge sites.
This does not assert the source's separate, insufficiently specified O(D³) solve/sweep bounds. -/
theorem single_edge_energy :
    (∀ (N D : ℕ) (hN : 1 ≤ N) (hD : 1 ≤ D)
      (A : Tensors N D) (G : PhaseGraph N) (V : Fin N → Matrix.unitaryGroup Bool ℂ)
      (hv : traceMPS A ≠ 0) (j k : Fin N) (hjk : j ≠ k)
      (H : Matrix (Qubits N) (Qubits N) ℂ) (hH : H.IsHermitian),
      SingleEdgeEnergyLaw hN hD A G V hv j k hjk H hH) ∧
    (∀ s : ℕ, ∃ C : ℕ, 0 < C ∧
      ∃ programs : (N D m : ℕ) → (supports : Fin m → Finset (Fin N)) →
        Fin N → Fin N → EdgeProgram N D m supports,
      ∀ (N D m : ℕ) (supports : Fin m → Finset (Fin N)) (j k : Fin N),
        1 ≤ N → 1 ≤ D → j ≠ k → (∀ r, (supports r).card ≤ s) →
        let p := programs N D m supports j k
        p.scalar.calls = 0 ∧ p.scalar.work 0 ≤ C * (m + 1) * N * D ^ 5 ∧
        (∀ (A : Tensors N D) (G : PhaseGraph N) (V : Fin N → Matrix.unitaryGroup Bool ℂ)
          (terms : (r : Fin m) → Matrix (supports r → Bool) (supports r → Bool) ℂ),
          traceMPS A ≠ 0 → (∀ r, (terms r).IsHermitian) →
          let H := localTermHamiltonian supports terms
          let c := edgeCoefficients A G V j k H
          ∃ mem : ℕ → ℂ, p.scalar.execute (edgeData A G V terms) (fun _ => 0) = some mem ∧
            p.constantCoefficient < p.scalar.code.length ∧
            p.phaseCoefficient < p.scalar.code.length ∧ p.norm < p.scalar.code.length ∧
            p.totalEnergy < p.scalar.code.length ∧
            mem p.constantCoefficient = (c.1 : ℂ) ∧ mem p.phaseCoefficient = c.2 ∧
            mem p.norm = (‖traceMPS A‖ ^ 2 : ℝ) ∧
            mem p.totalEnergy = (rayleighEnergy H (rage A G V) : ℂ))) := by sorry

end QAlgorithms.Papers.Independent.Intakethird100.P57

end
Source
Renormalization algorithm with graph enhancement, arXiv:0802.1211v1, PDF p. 3, Static updates: one-edge energy is a quadratic form in a unit-modulus phase
Read-back

What the Lean code literally says, in plain math · gpt-6.1-sol

This theorem asserts the conjunction of the following two statements. First, for every pair of natural numbers N,DN,DN,D with N≥1N\ge1N≥1 and D≥1D\ge1D≥1, every family of complex D×DD\times DD×D matrices Aq,bA_{q,b}Aq,b​ indexed by sites 0≤q<N0\le q<N0≤q<N and Boolean values bbb, every real array www satisfying wab=wbaw_{ab}=w_{ba}wab​=wba​, waa=0w_{aa}=0waa​=0, and 0≤wab<2π0\le w_{ab}<2\pi0≤wab​<2π, every family of unitary complex 2×22\times22×2 matrices VqV_qVq​, every pair of distinct sites j,kj,kj,k, and every Hermitian matrix HHH indexed by Boolean strings of length NNN, assume that the vector ttt defined by tb=tr⁡(A0,b0⋯AN−1,bN−1)t_b=\operatorname{tr}(A_{0,b_0}\cdots A_{N-1,b_{N-1}})tb​=tr(A0,b0​​⋯AN−1,bN−1​​) is nonzero. Put Z=∥t∥2Z=\|t\|^2Z=∥t∥2, let (Dw)b,b=exp⁡(i∑a<c, ba=bc=truewac)(D_w)_{b,b}=\exp(i\sum_{a<c,\ b_a=b_c=\mathrm{true}}w_{ac})(Dw​)b,b​=exp(i∑a<c, ba​=bc​=true​wac​), let Wa,b=∏q(Vq)aq,bqW_{a,b}=\prod_q(V_q)_{a_q,b_q}Wa,b​=∏q​(Vq​)aq​,bq​​, and define the vector eee by eb=exp⁡(−iwjk)(Dwt)be_b=\exp(-iw_{jk}) (D_wt)_beb​=exp(−iwjk​)(Dw​t)b​ when bj=bk=trueb_j=b_k=\mathrm{true}bj​=bk​=true and eb=(Dwt)be_b=(D_wt)_beb​=(Dw​t)b​ otherwise. Let uuu equal eee on strings for which the conjunction of those two bits is false and zero elsewhere, and let vvv equal eee on strings for which both bits are true and zero elsewhere. Define HV=W†HWH_V=W^\dagger HWHV​=W†HW, c0=Re⁡⟨u,HVu⟩+Re⁡⟨v,HVv⟩c_0=\operatorname{Re}\langle u,H_Vu\rangle+\operatorname{Re}\langle v,H_Vv\ranglec0​=Re⟨u,HV​u⟩+Re⟨v,HV​v⟩, c1=⟨u,HVv⟩c_1=\langle u,H_Vv\ranglec1​=⟨u,HV​v⟩, ξz=W(u+zv)\xi_z=W(u+zv)ξz​=W(u+zv), and EH(x)=Re⁡⟨x,Hx⟩/∥x∥2E_H(x)=\operatorname{Re}\langle x,Hx\rangle/\|x\|^2EH​(x)=Re⟨x,Hx⟩/∥x∥2. Then, for every complex zzz with ∣z∣=1|z|=1∣z∣=1, ∥ξz∥2=Z\|\xi_z\|^2=Z∥ξz​∥2=Z and EH(ξz)=(c0+2Re⁡(c1z))/ZE_H(\xi_z)=(c_0+2\operatorname{Re}(c_1z))/ZEH​(ξz​)=(c0​+2Re(c1​z))/Z; the number z∗=1z_* =1z∗​=1 if c1=0c_1=0c1​=0 and z∗=−c1‾/∣c1∣z_*=-\overline{c_1}/|c_1|z∗​=−c1​​/∣c1​∣ otherwise has modulus one; EH(ξz∗)≤EH(ξz)E_H(\xi_{z_*})\le E_H(\xi_z)EH​(ξz∗​​)≤EH​(ξz​) for every ∣z∣=1|z|=1∣z∣=1; and if c1=0c_1=0c1​=0, every such zzz satisfies EH(ξz)=c0/ZE_H(\xi_z)=c_0/ZEH​(ξz​)=c0​/Z. Second, for every natural number sss, there exist a positive natural number CCC and a family of edge programs pN,D,m,(Sr),j,kp_{N,D,m,(S_r),j,k}pN,D,m,(Sr​),j,k​ defined for all natural numbers N,D,mN,D,mN,D,m, all finite supports Sr⊆{0,…,N−1}S_r\subseteq\{0,\ldots,N-1\}Sr​⊆{0,…,N−1} for 0≤r<m0\le r<m0≤r<m, and all sites j,kj,kj,k, with the following guarantees whenever N≥1N\ge1N≥1, D≥1D\ge1D≥1, j≠kj\ne kj=k, and ∣Sr∣≤s|S_r|\le s∣Sr​∣≤s for every rrr. Each such program has zero calls to its supplied product-evaluation function and work at parameter zero at most C(m+1)ND5C(m+1)ND^5C(m+1)ND5, where work charges 4N+1+B4N+1+B4N+1+B for each product-evaluation call at parameter BBB and 111 for every other instruction. For every A,w,VA,w,VA,w,V as above and every family of Hermitian complex matrices TrT_rTr​ indexed by Boolean assignments on SrS_rSr​, provided t≠0t\ne0t=0, define H=∑rT~rH=\sum_r\widetilde T_rH=∑r​Tr​, where (T~r)a,b=(Tr)a∣Sr,b∣Sr(\widetilde T_r)_{a,b}=(T_r)_{a|_{S_r},b|_{S_r}}(Tr​)a,b​=(Tr​)a∣Sr​​,b∣Sr​​​ if a,ba,ba,b agree outside SrS_rSr​ and is zero otherwise, and form c0,c1c_0,c_1c0​,c1​ by the formulas above for this HHH. Execute the program using tensor inputs (Aq,b)i,l(A_{q,b})_{i,l}(Aq,b​)i,l​, phase inputs wabw_{ab}wab​ viewed in C\mathbb CC, rotation inputs (Vq)a,b(V_q)_{a,b}(Vq​)a,b​, term inputs (Tr)a,b(T_r)_{a,b}(Tr​)a,b​, and the identically zero product-evaluation function. Execution succeeds with a memory function μ:N→C\mu:\mathbb N\to\mathbb Cμ:N→C; all four designated addresses are smaller than the instruction-list length; and their values are respectively c0c_0c0​ viewed in C\mathbb CC, c1c_1c1​, ZZZ viewed in C\mathbb CC, and EH(WDwt)E_H(WD_wt)EH​(WDw​t) viewed in C\mathbb CC. Here scalar execution starts with all memory entries zero, writes successive instruction results at addresses 0,1,…0,1,\ldots0,1,…, and permits complex constants, input reads, addition, multiplication, division, conjugation, z↦exp⁡(iRe⁡z)z\mapsto\exp(i\operatorname{Re}z)z↦exp(iRez), and product-evaluation calls; unwritten addresses read as zero and division by zero fails. The quantification allows s=0s=0s=0, empty supports, and m=0m=0m=0, in which case H=0H=0H=0; the distinct-site hypotheses cannot be satisfied when N=1N=1N=1, and the numerical hypotheses exclude N=0N=0N=0 and D=0D=0D=0 from these guarantees.

Human review
  • Endorsed by Shuze Chen · Nov 1, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Nov 1, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me