-
Notifications
You must be signed in to change notification settings - Fork 0
Physics Formulations
GRANITE v0.6.8 | ← Parameter Reference | Roadmap →
Complete governing equations with references. For implementation details see
docs/DEVELOPER_GUIDE.md §5.
GRANITE evolves the Einstein field equations G_μν = 8π T_μν using the CCZ4 (Conformal and Covariant Z4) formulation of Alic et al. (2012). CCZ4 extends BSSN by promoting the algebraic constraint Z_μ = 0 to a dynamical field, enabling exponential damping.
Conformal decomposition:
γ̃_ij = χ · γ_ij, χ = (det γ_ij)^{-1/3}
Evolution equations:
∂_t χ = (2/3) χ (α K − ∂_i β^i) + β^i ∂_i χ
∂_t γ̃_ij = −2α Ã_ij + γ̃_ik ∂_j β^k + γ̃_jk ∂_i β^k
− (2/3) γ̃_ij ∂_k β^k + β^k ∂_k γ̃_ij
∂_t K = −D_i D^i α + α(Ã_ij Ã^ij + K²/3)
− κ₁(1+κ₂)αΘ + 4πα(ρ + S) + β^i ∂_i K
∂_t Θ = (1/2)α(R − Ã_ij Ã^ij + (2/3)K²)
− α Γ̂^μ_{μν} Z^ν − κ₁(2+κ₂)αΘ + β^i ∂_i Θ
Production defaults: κ₁ = 0.02, κ₂ = 0, η = 2.0
∂_t α = β^i ∂_i α − 2α K
Singularity-avoiding: α → 0 near punctures, preventing evolution from reaching physical singularity.
∂_t β^i = (3/4) B^i
∂_t B^i = ∂_t Γ̃^i − η B^i
(Campanelli et al. 2006; Baker et al. 2006)
Trumpet geometry: At sufficient resolution (dx < 0.2M) and with ko_sigma = 0.1, the lapse profile forms a trumpet geometry near each puncture. At coarser resolution (dx > 0.5M), the trumpet is unresolved — this is expected behavior, not a bug.
The matter sector evolves conserved variables via:
∂_t U + ∂_i F^i(U) = S(U, g_μν)
Conserved variables:
U = √γ · (D, S_j, τ, B^i, D·Y_e)^T
D = ρ W (baryon density)
S_j = ρ h W² v_j + ε_jkl (v^k − β^k/α) B^l √γ/α (momentum)
τ = ρ h W² − P − D (energy)
Geometric source terms S couple to spacetime via Christoffel symbols evaluated from the CCZ4 GRMetric3 output.
Evolves via the induction equation:
∂_t B^i + ∂_j (v^j B^i − v^i B^j) = 0
subject to: ∂_i B^i = 0 (no magnetic monopoles)
GRANITE enforces ∇·B = 0 to machine precision at all times via Constrained Transport (Evans & Hawley 1988). The magnetic field is stored on cell faces (staggered grid); the CT update guarantees exact discrete divergence = 0.
| EOS Type | Implementation | Use Case |
|---|---|---|
| Ideal gas (Γ-law) | IdealGasEOS(Γ) |
SMBH mergers (Γ=4/3), validation tests |
| Tabulated nuclear |
TabulatedEOS with tri-linear interpolation in (ρ, T, Y_e) |
Neutron star mergers (v0.8+) |
Sound speed is always computed exactly from the EOS — never from a hardcoded Γ (fixed bug C1).
Energy equation:
∂_t E + ∂_i F^i = −κ_a (E − E_eq) + Q_leakage
Flux equation:
∂_t F^i + ∂_j P^ij = −(κ_a + κ_s) F^i
M1 Eddington tensor (Minerbo 1978, Levermore 1984):
P^ij = [(1−ξ)/2] δ^ij E + [(3ξ−1)/2] n^i n^j E
ξ = ξ(f), f = |F|/(cE) (radiation flux factor)
⚠️ Status: The M1 module is implemented and tested but is not called in the main RK3 evolution loop in v0.6.8. Radiation physics is completely inactive. Planned for v0.7.
(∂_x f)_i = (−f_{i+2} + 8f_{i+1} − 8f_{i−1} + f_{i−2}) / (12 Δx) + O(Δx⁴)
Requires nghost = 4.
(∂_t u)_KO = −σ (Δx)^{2p-1} / 2^{2p} · D^{2p} u, p = 3 (6th-order)
σ = 0.1 is the safe default. Never exceed 0.15.
U^(1) = U^n + Δt L(U^n)
U^(2) = (3/4) U^n + (1/4) [U^(1) + Δt L(U^(1))]
U^(n+1) = (1/3) U^n + (2/3) [U^(2) + Δt L(U^(2))]
| Scheme | Order | Stencil | Use case |
|---|---|---|---|
| PLM (Piecewise Linear Method) | 2nd | 3 cells | Quick tests |
| PPM (Colella & Woodward 1984) | 3rd | 4 cells | Contact discontinuities |
| MP5 (Suresh & Huynh 1997) | 5th | 6 cells | Production default |
| Solver | Physics | Notes |
|---|---|---|
| HLLE | GR-aware (uses actual GRMetric3) | Fixed C3; fast but diffusive |
| HLLD (Miyoshi & Kusano 2005) | All 7 MHD wave families | Preferred for MHD accuracy |
| Reference | Used in GRANITE |
|---|---|
| Alic et al. 2012, PRD 85, 064040 | CCZ4 formulation |
| Baker et al. 2006, PRL 96, 111102 | Moving punctures |
| Berger & Oliger 1984, J. Comput. Phys. 53, 484 | AMR + subcycling |
| Bona et al. 1995, PRL 75, 600 | 1+log slicing |
| Brandt & Brügmann 1997, PRL 78, 3606 | Two-Punctures ID |
| Campanelli et al. 2006, PRL 96, 111101 | Moving punctures / Gamma-driver |
| Colella & Woodward 1984, J. Comput. Phys. 54, 174 | PPM reconstruction |
| Evans & Hawley 1988, ApJ 332, 659 | Constrained transport |
| Levermore 1984, J. Quant. Spectrosc. Radiat. Transfer 31, 149 | M1 closure |
| Minerbo 1978, J. Quant. Spectrosc. Radiat. Transfer 20, 541 | Eddington factor |
| Miyoshi & Kusano 2005, J. Comput. Phys. 208, 315 | HLLD Riemann solver |
| Nakano et al. 2015, PRD 91, 104022 | Ψ₄ extrapolation |
| Reisswig & Pollney 2011, CQG 28, 195015 | Fixed-frequency GW integration |
| Shu & Osher 1988, J. Comput. Phys. 77, 439 | SSP-RK3 |
| Suresh & Huynh 1997, J. Comput. Phys. 136, 83 | MP5 reconstruction |
See also: Architecture Overview | Benchmarks & Validation | Scientific Context
v0.6.8 · Repository · Issues
- 🩺 Simulation Health & Debugging
- 📊 Benchmarks & Validation
- 🗂️ AMR Design
- 🖥️ HPC Deployment
- 🌀 VORTEX Engine
"Simulate the unimaginable."