## Phase 114: Kalika Capstone — Non-Perturbative QFT, Superspace Calculus, Integrable Systems, Spectral Geometry, Optimal Transport, Quantum Chaos & Remaining Gap Closures

The final phase, produced by a surgical four-pass audit. Closes every remaining
gap where the absence would be noticed by a working theorist. After this phase,
Kalika covers physics from solitons to scrambling, from Grassmann integrals to
Gaussian processes, from Lax pairs to Lindblad equations — the complete toolkit
for researchers at every level of theoretical physics and mathematics.

---

### 114.1 Non-Perturbative QFT Objects (`@kalika/non-perturbative`)

#### 114.1.1 Instantons

- [x] **114.1.1.1** Implement BPST instanton solution: the self-dual (F = \*F)
      SU(2) Yang-Mills solution on R^4 with action S = 8π²/g², parametrized by
      position x₀, scale ρ, and gauge orientation — in both singular and regular
      gauges, with explicit field strength F^a\_{μν}(x) as symbolic expressions
      computable at any point
- [x] **114.1.1.2** Implement ADHM construction: given instanton charge k and
      gauge group SU(N), construct the moduli space M*{k,N} of k-instanton
      solutions using the Atiyah-Drinfeld-Hitchin-Manin data (matrices B₁, B₂,
      I, J satisfying ADHM equation), parametrize solutions, and compute the
      dimension of M*{k,N} = 4kN - N² + 1
- [x] **114.1.1.3** Implement instanton calculus: semiclassical expansion around
      instanton background — compute the one-loop fluctuation determinant
      det'(-D²)/det(-∂²) (ratio of functional determinants, primed = zero modes
      removed), count zero modes (4kN for SU(N) k-instanton via Atiyah-Singer
      index theorem), and evaluate instanton contributions to correlation
      functions using collective coordinates
- [x] **114.1.1.4** Implement Nekrasov partition function: exact partition
      function of N=2 gauge theories via equivariant localization on the
      instanton moduli space — sum over Young diagrams (k-instanton sector = sum
      over pairs of Young diagrams with |Y₁|+|Y₂|=k), compute the Nekrasov
      deformation parameters (ε₁, ε₂), and extract the Seiberg-Witten
      prepotential F in the ε₁,ε₂ → 0 limit — the cornerstone of exact results
      in 4D gauge theory

#### 114.1.2 Monopoles & Solitons

- [x] **114.1.2.1** Implement 't Hooft-Polyakov monopole: spherically symmetric
      solution of SU(2) Yang-Mills-Higgs with magnetic charge g_m = 4π/e,
      asymptotic Coulomb magnetic field, core structure — numerical solution of
      the radial BPS equation (dφ/dr = ±B_r) and non-BPS case, magnetic charge
      computation via surface integral of F
- [x] **114.1.2.2** Implement kink/domain wall solutions: scalar field kinks in
      1+1D (φ⁴ kink: φ(x) = v·tanh(mx/√2), sine-Gordon: φ(x) = 4arctan(e^{mx})),
      domain walls in 3+1D, kink-antikink scattering, Bogomolny bound E ≥
      |∫(dW/dφ)dφ|, and topological charge Q = (φ(+∞)-φ(-∞))/2v
- [x] **114.1.2.3** Implement vortex solutions: Abrikosov/Nielsen-Olesen vortex
      in Abelian Higgs model (magnetic flux quantized: Φ = 2πn/e), BPS vortex
      equation, vortex profile functions (numerical solution), multi-vortex
      moduli space, and Taubes' theorem (counting vortices = counting zeros of
      Higgs field)
- [x] **114.1.2.4** Implement skyrmion solutions: SU(2) Skyrme model soliton
      (topological charge = baryon number B), numerical minimization of Skyrme
      energy functional for B = 1,2,3,4 skyrmions, rational map ansatz for
      higher B, and connection to nuclear physics (skyrmions as nucleons)
- [x] **114.1.2.5** Implement topological defect classification: given a
      symmetry breaking pattern G → H, compute the homotopy groups π_n(G/H) to
      classify: domain walls (π₀), cosmic strings (π₁), monopoles (π₂), textures
      (π₃) — automating the Kibble-Zurek classification of topological defects
      from the vacuum manifold

#### 114.1.3 Anomalies (Explicit Computation)

- [x] **114.1.3.1** Implement chiral anomaly computation: triangle diagram (AVV)
      for a single Weyl fermion coupled to gauge fields, verify ∂*μ j^μ_5 =
      (1/16π²)F*μν F̃^μν, compute anomaly coefficient for arbitrary
      representation R — both by direct diagram evaluation (regularization
      required: Pauli-Villars or dimensional) and by Fujikawa path-integral
      method (Jacobian of chiral transformation)
- [x] **114.1.3.2** Implement anomaly polynomial: for a theory with chiral
      fermions in representations R*i under gauge group G, compute the total
      anomaly polynomial I*{d+2} = ch(F) · Â(R) using Chern character and A-hat
      genus, extract: gauge anomaly (I_6^{gauge}), gravitational anomaly
      (I_6^{grav}), mixed anomaly (I_6^{mixed}), and verify anomaly cancellation
      conditions for consistent theories
- [x] **114.1.3.3** Implement anomaly descent equations: from the anomaly
      polynomial I*{d+2}, compute the Chern-Simons form ω*{d+1} (I*{d+2} =
      dω*{d+1}) and the anomaly functional G*d (δω*{d+1} = dG_d + ...) — the
      systematic procedure relating anomaly polynomials to actual anomalous
      variations of the effective action
- [x] **114.1.3.4** Implement anomaly inflow (Callan-Harvey): given a domain
      wall or brane configuration, compute the anomaly on the wall/brane
      worldvolume and verify it is cancelled by bulk Chern-Simons term inflow —
      the mechanism by which D-branes carry charge and chiral edge modes of
      topological insulators are anomaly-free

---

### 114.2 Superspace & Grassmann Calculus (`@kalika/supersymmetry`)

- [x] **114.2.1.1** Implement Grassmann algebra: Grassmann variables (θ*α, θ̄*α̇)
      with anticommutation {θ*α, θ*β} = 0, Grassmann polynomial arithmetic (a +
      bθ + cθ² with θ² = θ₁θ₂), and Berezin integration (∫dθ θ = 1, ∫dθ 1 = 0) —
      the algebraic foundation of all superspace computation
- [x] **114.2.1.2** Implement N=1 superspace: supercoordinates (x^μ, θ^α, θ̄^α̇),
      supercovariant derivatives D*α = ∂/∂θ^α + iσ^μ*{αα̇}θ̄^α̇ ∂*μ and D̄*{α̇},
      supersymmetry generators Q*α, Q̄*{α̇}, and their algebra {Q*α, Q̄*{α̇}} =
      2σ^μ*{αα̇} P*μ — with symbolic manipulation of all superspace expressions
- [x] **114.2.1.3** Implement superfield expansion: chiral superfield Φ(y,θ) =
      φ(y) + √2 θψ(y) + θ²F(y) where y^μ = x^μ + iθσ^μθ̄, vector superfield
      V(x,θ,θ̄) in Wess-Zumino gauge = -θσ^μθ̄ A_μ + iθ²θ̄λ̄ - iθ̄²θλ + ½θ²θ̄²D, and
      automatic component extraction from superspace expressions
- [x] **114.2.1.4** Implement superspace Lagrangian construction: F-term (∫d²θ
      W(Φ) + h.c.), D-term (∫d⁴θ K(Φ†,Φ)), gauge kinetic (∫d²θ W^αW_α), and
      automatic derivation of component Lagrangian from superspace Lagrangian
      via Grassmann integration — this IS how theorists derive SUSY Lagrangians;
      the component expressions in 101.2.1.5 are the output, not the starting
      point
- [x] **114.2.1.5** Implement N=2 superspace: harmonic superspace
      (Galperin-Ivanov-Kalitzin-Ogievetsky-Sokatchev), N=2 vector multiplet and
      hypermultiplet, and the connection to Seiberg-Witten theory via the
      prepotential F(Φ) — for researchers working with extended supersymmetry

---

### 114.3 Integrable Systems (`@kalika/integrable-systems`)

_(Fills the empty placeholder directory from Phase 99.1.1.1)_

- [x] **114.3.1.1** Implement Lax pair framework: given a PDE, search for Lax
      representation L_t = [M, L] (or zero-curvature form dA - A∧A = 0) where
      the eigenvalues of L are constants of motion, verify integrability via
      existence of Lax pair, and construct the conserved charges from Tr(L^n) —
      the diagnostic tool for integrability
- [x] **114.3.1.2** Implement inverse scattering transform (IST): for integrable
      PDEs (KdV, NLS, sine-Gordon, KP), solve the direct scattering problem
      (compute scattering data from initial condition), evolve scattering data
      in time (trivial linear evolution), and reconstruct the solution via
      inverse scattering (Gel'fand-Levitan-Marchenko equation) — the nonlinear
      analogue of the Fourier transform
- [x] **114.3.1.3** Implement explicit soliton solutions: N-soliton solutions
      for KdV (via Hirota bilinear method: u = 2∂²_x ln τ where τ is a
      determinant), NLS (bright and dark solitons), sine-Gordon (kink,
      breather), KP equation (line solitons, web solitons), and Toda lattice —
      with parameter visualization (soliton velocity, amplitude, phase shifts
      from scattering)
- [x] **114.3.1.4** Implement Hirota bilinear method: transform a nonlinear PDE
      to bilinear form (D-operator calculus), construct the tau-function, and
      systematically generate N-soliton solutions as determinants
      (Wronskian/Grammian form) — the most efficient method for constructing
      multi-soliton solutions
- [x] **114.3.1.5** Implement classical r-matrix and Yang-Baxter: classical
      r-matrix r ∈ g ⊗ g satisfying the classical Yang-Baxter equation [r₁₂,
      r₁₃] + [r₁₂, r₂₃] + [r₁₃, r₂₃] = 0, Belavin-Drinfeld classification of
      solutions for simple Lie algebras, and connection to Poisson-Lie groups —
      the classical limit of the quantum Yang-Baxter equation (already in
      110.7.1.3)
- [x] **114.3.1.6** Implement Bethe ansatz: coordinate Bethe ansatz for XXX
      (Heisenberg) spin chain — construct the N-magnon wavefunction, derive
      Bethe equations for the rapidities {λ_j}, solve for the ground state
      energy E_0/N = -ln(2) + 1/4 (matching exact result), and compute
      excitation spectrum — the cornerstone of exactly solvable quantum spin
      chains

---

### 114.4 Spectral Geometry (`@kalika/spectral-geometry`)

- [x] **114.4.1.1** Implement Laplacian eigenvalue computation on manifolds:
      given a Riemannian manifold (M, g) discretized on a mesh, compute
      eigenvalues and eigenfunctions of the Laplace-Beltrami operator Δ_g =
      (1/√g)∂_i(√g g^{ij} ∂_j) — via FEM discretization (from @kalika/numerical)
      on the manifold's mesh. Verify Weyl's law: N(λ) ~ C_d vol(M) λ^{d/2} for
      the eigenvalue counting function
- [x] **114.4.1.2** Implement heat kernel coefficient computation: Seeley-DeWitt
      coefficients a_0, a_1, a_2, a_3, a_4 for a Laplace-type operator -D² + E
      on a vector bundle, where a_0 = (4π)^{-d/2} tr(1), a_1 = (4π)^{-d/2}
      tr(R/6 - E), a_2 involves Riemann tensor squared terms — these are
      ESSENTIAL for one-loop effective action computation in quantum field
      theory on curved backgrounds
- [x] **114.4.1.3** Implement spectral zeta function: ζ*Δ(s) = Σ λ_n^{-s} for
      the eigenvalues of the Laplacian, meromorphic continuation, evaluation at
      s=0 (gives log of functional determinant: ln det(Δ) = -ζ'*Δ(0)), and
      regularized functional determinant computation — used for one-loop
      partition functions in QFT, Casimir energy, and string theory (Polyakov
      action)
- [x] **114.4.1.4** Implement eta invariant: η(s) = Σ sign(λ_n)|λ_n|^{-s} for
      the Dirac operator, evaluate η(0) (the APS eta invariant), and compute the
      APS index = (η(0) + dim ker D)/2 — connects spectral geometry to the index
      theorem with boundary (Atiyah-Patodi-Singer) and to Chern-Simons theory

---

### 114.5 Optimal Transport (`@kalika/optimal-transport`)

- [x] **114.5.1.1** Implement Wasserstein distance: W_p(μ, ν) between discrete
      probability measures via linear programming (exact solution of the
      Kantorovich problem), W_1 (Earth Mover's Distance) via network simplex,
      and W_2 via Benamou-Brenier (dynamic formulation as geodesic distance in
      the space of measures)
- [x] **114.5.1.2** Implement **Sinkhorn algorithm**: entropic regularization of
      optimal transport — solve the regularized problem via iterative matrix
      scaling (Sinkhorn-Knopp iterations), converging in O(n² log n / ε²) — the
      standard fast approximation for OT, used across ML and physics.
      GPU-accelerated via @kalika/numerical-engine
- [x] **114.5.1.3** Implement displacement interpolation: given measures μ₀ and
      μ₁, compute the McCann displacement interpolation μ_t = ((1-t)id + t T)#μ₀
      where T is the optimal transport map — the geodesic in Wasserstein space,
      used in cosmological reconstruction (matching observed galaxy distribution
      to initial density field)
- [x] **114.5.1.4** Implement Monge-Ampère solver: for the quadratic cost c(x,y)
      = |x-y|², the OT map satisfies the Monge-Ampère equation det(D²φ) = f/g —
      solve via finite difference or semi-discrete methods. Applications:
      cosmological reconstruction (Zeldovich approximation), mesh generation,
      reflector design

---

### 114.6 Open Quantum Systems (`@kalika/open-quantum-systems`)

- [x] **114.6.1.1** Implement Lindblad master equation solver: dρ/dt = -i[H,ρ] +
      Σ_k γ_k(L_k ρ L_k† - ½{L_k†L_k, ρ}) — given Hamiltonian H, jump operators
      L_k, and rates γ_k, solve for time evolution of the density matrix ρ(t)
      via: matrix exponential (small systems), Runge-Kutta on vectorized ρ
      (medium), and Monte Carlo wavefunction method / quantum trajectories
      (large systems)
- [x] **114.6.1.2** Implement standard quantum noise channels: depolarizing (ρ →
      (1-p)ρ + p/d·I), dephasing (ρ → (1-p)ρ + p·Z ρ Z), amplitude damping
      (spontaneous emission), phase damping, and generalized amplitude damping —
      as Kraus operator sets, verifiable via complete positivity and trace
      preservation
- [x] **114.6.1.3** Implement quantum trajectory / Monte Carlo wavefunction
      method: stochastic simulation of open quantum system evolution —
      deterministic Schrodinger evolution + random quantum jumps at rates
      determined by jump operators, ensemble average recovers Lindblad evolution
      — efficient for large Hilbert spaces where direct density matrix evolution
      is intractable
- [x] **114.6.1.4** Implement quantum master equations beyond Lindblad: Redfield
      equation (Born-Markov, non-secular), Nakajima-Zwanzig (exact,
      non-Markovian with memory kernel), and hierarchical equations of motion
      (HEOM) for structured environments — for researchers studying
      non-Markovian open quantum dynamics

---

### 114.7 Quantum Chaos & Quantum Optics

#### 114.7.1 Quantum Chaos

- [x] **114.7.1.1** Implement level spacing statistics: given a spectrum {E_n},
      compute the nearest-neighbor spacing distribution P(s) and the level
      spacing ratio distribution P(r), compare against Wigner-Dyson (chaotic:
      GOE/GUE/GSE depending on symmetry class) vs Poisson (integrable), compute
      the spectral rigidity Δ₃(L) and number variance Σ²(L)
- [x] **114.7.1.2** Implement spectral unfolding: remove the smooth part of the
      density of states N̄(E) to produce a unit-mean-spacing spectrum — the
      preprocessing step required before any spectral statistics analysis.
      Implement both polynomial fit and Weyl's law based unfolding
- [x] **114.7.1.3** Implement OTOC (out-of-time-order correlator): C(t) =
      ⟨|[W(t), V(0)]|²⟩ for operators V, W — compute via exact diagonalization
      (small systems), Krylov subspace methods (medium), or tensor network
      approximation (1D chains). Extract the quantum Lyapunov exponent λ_L from
      exponential growth C(t) ~ e^{2λ_L t}, and verify the
      Maldacena-Shenker-Stanford bound λ_L ≤ 2πT/ℏ for thermal states
- [x] **114.7.1.4** Implement quantum billiards: eigenvalues/eigenfunctions of
      the Laplacian on 2D domains (stadium, Sinai, cardioid, Bunimovich), using
      FEM or boundary element methods. Visualize eigenfunction scarring
      (enhanced amplitude along classical periodic orbits), compute Husimi
      function, and verify Gutzwiller trace formula (connecting classical
      periodic orbits to quantum eigenvalues)
- [x] **114.7.1.5** Implement spectral form factor: K(t) = |Σ_n e^{-iE_n t}|² /
      Z(2β), the Fourier transform of the two-point spectral correlation
      function — diagnose quantum chaos via the dip-ramp-plateau structure
      (early-time decay, linear ramp from RMT, late-time plateau at 1). Crucial
      for SYK model and holographic studies of black hole information

#### 114.7.2 Quantum Optics

- [x] **114.7.2.1** Implement coherent states: |α⟩ = e^{-|α|²/2} Σ (α^n/√n!)
      |n⟩, displacement operator D(α) = exp(αa† - α\*a), coherent state algebra
      (a|α⟩ = α|α⟩), Glauber-Sudarshan P representation, and phase space methods
      (Wigner function for coherent states is a Gaussian)
- [x] **114.7.2.2** Implement squeezed states: S(ξ)|0⟩ where S(ξ) =
      exp(½(ξ\*a² - ξa†²)) is the squeezing operator, quadrature variances
      (⟨ΔX₁²⟩ < 1/2, ⟨ΔX₂²⟩ > 1/2), and connection to gravitational wave
      detection (LIGO uses squeezed light to enhance sensitivity beyond quantum
      noise)
- [x] **114.7.2.3** Implement Jaynes-Cummings model: two-level atom coupled to
      single-mode cavity field, H = ωa†a + ½ω₀σ_z + g(a†σ₋ + aσ₊), exact
      diagonalization (dressed states), vacuum Rabi oscillations, collapse and
      revival of atomic inversion, and Purcell effect — the fundamental model of
      cavity QED and circuit QED
- [x] **114.7.2.4** Implement beam splitter and interferometer: beam splitter
      transformation (a_out = t·a_in + r·b_in), Mach-Zehnder interferometer,
      Hong-Ou-Mandel effect (two-photon interference: coincidence dip), and
      quantum state tomography via homodyne detection — the basic optical
      quantum information toolkit

---

### 114.8 Non-Equilibrium Statistical Mechanics

- [x] **114.8.1.1** Implement Jarzynski equality: ⟨e^{-βW}⟩ = e^{-βΔF} — given
      an ensemble of non-equilibrium work values W from driving a system between
      states A and B, estimate the free energy difference ΔF = F_B - F_A.
      Implement estimators: exponential average (biased for small samples), BAR
      (Bennett acceptance ratio), and MBAR (multistate BAR)
- [x] **114.8.1.2** Implement Crooks fluctuation theorem: P_F(W)/P_R(-W) =
      e^{β(W-ΔF)} — from forward and reverse work distributions, extract ΔF at
      the crossing point. Implement bidirectional estimator for optimal
      precision
- [x] **114.8.1.3** Implement stochastic thermodynamics: entropy production
      along individual trajectories σ = ΔS_sys + ΔS_env, fluctuation theorem
      P(σ)/P(-σ) = e^σ, and entropy production rate computation for Langevin
      dynamics and Markov jump processes — the framework for thermodynamics of
      small systems (molecular motors, single molecules)
- [x] **114.8.1.4** Implement large deviation theory computation: rate function
      I(x) = -lim(1/N)ln P(X_N = x), Gärtner-Ellis theorem (from cumulant
      generating function), Legendre-Fenchel transform, and tilted ensemble
      sampling — for computing probabilities of rare events in statistical
      mechanics

---

### 114.9 Gaussian Processes & Surrogate Modeling (`@kalika/surrogate`)

- [x] **114.9.1.1** Implement Gaussian process regression: GP(m(x), k(x,x'))
      with standard kernels (RBF/squared-exponential, Matern-3/2, Matern-5/2,
      periodic, rational quadratic, linear), hyperparameter optimization via log
      marginal likelihood maximization (L-BFGS), posterior mean and variance
      computation, and efficient implementation using Cholesky decomposition of
      the kernel matrix
- [x] **114.9.1.2** Implement Bayesian optimization: acquisition functions
      (Expected Improvement, Upper Confidence Bound, Probability of Improvement,
      Thompson Sampling), sequential optimization loop (evaluate acquisition →
      add point → update GP → repeat), and batch Bayesian optimization — for
      optimizing expensive simulations (lattice QCD parameters, collider
      analysis cuts, string landscape scanning)
- [x] **114.9.1.3** Implement GP emulator framework: given a set of simulation
      runs (input parameters → output observables), train a GP emulator that
      predicts outputs at new parameter values with uncertainty quantification —
      the standard technique for cosmological parameter estimation (replacing
      direct CAMB/CLASS runs with a trained GP) and collider physics (replacing
      detector simulation with a fast emulator)

---

### 114.10 Spin Geometry & Index Theorem

- [x] **114.10.1.1** Implement spin structure computation: given an oriented
      Riemannian manifold (M, g), determine if M admits a spin structure
      (obstruction: second Stiefel-Whitney class w₂ = 0), construct the spinor
      bundle S(M), and compute the Dirac operator D = γ^μ∇*μ acting on sections
      of S — in terms of the vielbein e^a*μ and spin connection ω^{ab}\_μ
- [x] **114.10.1.2** Implement Atiyah-Singer index computation: for the Dirac
      operator D on a compact spin manifold, compute ind(D) = dim ker D - dim
      ker D† via the index formula ind(D) = ∫_M Â(M) — the A-hat genus
      integrated over the manifold, expressed in terms of Pontryagin classes of
      TM. For Dirac coupled to a gauge bundle E, ind(D_E) = ∫_M Â(M) · ch(E) —
      giving the chiral asymmetry (number of zero modes)
- [x] **114.10.1.3** Implement physics applications of the index theorem:
      anomaly = index (chiral anomaly of Weyl fermions = index of Dirac operator
      on gauge bundle), generation number in string compactification (= half the
      Euler characteristic for CY3, or index of Dirac on CY with gauge bundle),
      and Witten index Tr(-1)^F (= index of supercharge operator, counts SUSY
      vacua)

---

### 114.11 Morse Theory

- [x] **114.11.1.1** Implement Morse function analysis: given a smooth function
      f: M → R on a manifold (discretized), compute critical points (∇f = 0),
      classify by Morse index (number of negative eigenvalues of the Hessian
      ∂²f/∂x_i∂x_j), verify Morse condition (all critical points are
      non-degenerate), and compute Morse polynomial M_t(f) = Σ t^{index(p)}
      summing over critical points
- [x] **114.11.1.2** Implement Morse homology: Morse chain complex (generators =
      critical points, differential counts gradient flow lines between critical
      points of adjacent index), verify ∂² = 0, compute Morse homology
      (isomorphic to singular homology by Morse's theorem), and verify Morse
      inequalities (M_t - P_t = (1+t)Q_t with Q_t having non-negative
      coefficients)
- [x] **114.11.1.3** Implement discrete Morse theory (Forman): for simplicial/CW
      complexes, compute optimal discrete Morse functions (minimize number of
      critical cells), discrete gradient vector fields, and Morse complex —
      computationally efficient for TDA and practical topology computation.
      Interface with GUDHI/Dionysus for persistent Morse-theoretic computations
- [x] **114.11.1.4** Implement Witten deformation: deformed Laplacian Δ_t = d_t
      d_t* + d_t* d_t where d_t = e^{-tf} d e^{tf}, verify that as t → ∞ the
      harmonic forms concentrate near critical points of f, recovering Morse
      theory from Hodge theory — the physics-inspired proof of the Morse
      inequalities that connects to supersymmetric quantum mechanics

---

### 114.12 Tests

- [x] **114.12.1.1** Write non-perturbative QFT tests: verify BPST instanton has
      action 8π²/g², topological charge Q=1, and self-duality F = \*F; verify
      kink mass matches Bogomolny bound; verify 't Hooft-Polyakov monopole has
      magnetic charge 4π/e — minimum 30 test cases
- [x] **114.12.1.2** Write superspace tests: verify component extraction from
      chiral superfield reproduces standard Wess-Zumino Lagrangian, F-term
      potential computation matches V = |∂W/∂φ|², D-term gives standard gauge
      kinetic term — minimum 20 test cases
- [x] **114.12.1.3** Write integrable systems tests: verify KdV 1-soliton
      solution is exact, 2-soliton scattering produces known phase shift, Bethe
      ansatz ground state energy for XXX chain matches exact E₀/N = -ln2 + 1/4 —
      minimum 25 test cases
- [x] **114.12.1.4** Write spectral geometry tests: verify Weyl's law for
      Laplacian eigenvalues on flat torus, heat kernel coefficients a₀ and a₁
      for sphere, spectral zeta ζ'(0) for interval [0,L] — minimum 15 test cases
- [x] **114.12.1.5** Write quantum chaos tests: verify level spacing statistics
      match GUE for chaotic Hamiltonian (Gaussian random matrix), Poisson for
      integrable (harmonic oscillator), OTOC growth rate bounded by MSS bound
      for thermal SYK — minimum 15 test cases
- [x] **114.12.1.6** Write optimal transport tests: verify W₂ distance between
      two Gaussians matches exact formula W₂² = |m₁-m₂|² + (√σ₁ - √σ₂)²,
      Sinkhorn converges to exact OT for small regularization — minimum 10 test
      cases
