vibe-qcDocumentation

Cell structure from a basis, the cubic Hessian and the independence of the splitting parameter

AICCM/CCM remains experimental. VibeQCAICCM/CellAnalytic.lean adds 58 claims. They close three gaps in the analytic chain behind Paper 1’s statements on the periodic Green function (its Eqs. nearfield and xiN), found in the review of 2026-11-13. No calculation is involved.

1. The cubic Hessian

near_cubic_tensor was 3x3 algebra under hypotheses, and no claim derived those hypotheses for a cubic lattice.

  • lat_regular_symmetry, lat_hessian_symmetry: a linear isometry that maps the lattice and the reciprocal set onto themselves leaves the regular part and its Hessian at the source invariant.

  • lat_cubic_hessian: for lattices closed under the coordinate reflections and exchanges (LatCubic), the Hessian is (Tr Q / 3) times the inner product. No smoothness, discreteness or cell hypothesis is used.

  • lat_cubic_hessian_value: with a cell structure, Q = 4 pi / (3 Omega).

  • lat_cubic_source_hessian: for the lattice a Z^3 with reciprocal set (2 pi / a) Z^3 and Omega = |a|^3, without any further hypothesis.

2. The cell structure is constructed

RegCell L K Omega (Regularity.lean) is a structure of hypotheses: a tiling cell, a quotient map with its fibres, a bijective labelling of the dual lattice, a character identity and the Haar scaling. No term of that type existed, so reg_theta_dual, reg_kernel_poisson and reg_source_trace were conditional.

  • latRegCell: every field from a real basis b and a basis bd with <bd_i, b_j> = 2 pi delta_ij. The cell is ZSpan.fundamentalDomain b.

  • latDualBasis, lat_dual_inner: that dual basis exists for every basis.

  • lat_cell_volume, lat_cell_volume_pos: the volume of the cell is |det b| and is positive.

  • lat_regcell_basis: a cell structure exists for every real basis, skew cells included, with L the integer span of b, K the integer span of the dual basis and Omega = |det b|.

  • lat_regcell_rect: the orthogonal cell in coordinate form.

3. The self-potential is tied to the cell and to the Poisson symbol

selfKernel and selfXi (Madelung.lean) take the lattice, the reciprocal set, the volume and the splitting parameter eta as unrelated inputs, and nothing related two values of eta.

  • lat_kernel_deriv: off the lattice, the derivative of the Ewald kernel with respect to eta is zero. The proof differentiates both sums term by term and cancels them with the theta identity reg_theta_dual.

  • lat_self_kernel_eta_indep, lat_self_xi_eta_indep: the kernel off the lattice and the self-potential do not depend on eta, given a cell structure. lat_self_xi_basis and lat_self_xi_rect state this without hypothesis for the lattice of a basis and for the orthogonal cell.

  • lat_elongated_positive: on the constructed 4 x 4 x 32 bohr cell the self-potential exceeds 1/8 bohr^-1 for every eta > 0. The earlier self_elongated_positive was an inequality at eta = 1/2 for any lattice containing one reciprocal vector.

  • lat_self_kernel_coeff, lat_self_kernel_coeff_basis: the cell Fourier coefficient of the Ewald kernel at a dual vector k is poissonWeight(|k|^2), with integrability on the cell. At k = 0 it is zero, the mean-zero gauge. The coefficient is the cell integral without the factor 1/Omega.

What is not covered

  • The Ewald kernel is not descended to an element of the torus L^2 space and identified with the Poisson solution of PeriodicPoisson.lean. The coefficient identity at every dual vector and the integrability are proved. The last bridge is missing.

  • The independence of eta needs the reciprocal set and the volume that belong to the lattice. For unrelated inputs it is false.

  • The kernel identity holds off the lattice, where the kernel is finite.

Check

python formal/check_proofs.py with the pinned Lake: 821 claims pass with the allowed axioms propext, Classical.choice and Quot.sound.