skill-rheology

name	skill-rheology
description	MDOODZ rheological framework — flow laws, viscosity computation, dislocation creep, diffusion creep, Peierls mechanism, grain boundary sliding, elasticity (Maxwell model), plasticity (Drucker-Prager, strain softening), material properties, and the mat_prop structure.

Rheology and Material Properties

Viscosity Computation Pipeline

For each grid point, the effective viscosity is computed by:

1. Identify active mechanisms per phase from .txt switches: cstv, pwlv, linv, gbsv, expv
2. Compute each mechanism's viscosity from T, P, ε̇, grain size, composition
3. Combine via harmonic averaging: 1/η_eff = 1/η_pwl + 1/η_lin + 1/η_gbs + 1/η_exp + 1/η_cst (weaker mechanism dominates)
4. Apply elastic correction (if elastic = 1): η_ve = η·G·Δt / (η + G·Δt) (Maxwell model)
5. Apply plastic correction (if yield stress exceeded): Drucker-Prager yield
6. Clamp: η_final ∈ [min_eta, max_eta]

The main viscosity function is ViscosityConciseAniso in AnisotropyRoutines.c (or ViscosityConcise for isotropic cases in RheologyDensity.c).

Creep Mechanisms

Power Law — Dislocation Creep (pwlv)

Equation: η_pwl = A^(-1/n) · f^(-r/n) · exp((Q + P·V)/(n·R·T)) · ε̇_II^(1/n - 1)

Parameter	Symbol	Code field
Pre-exponential factor	A	Apwl[phase]
Stress exponent	n	npwl[phase]
Activation energy	Q	Qpwl[phase]
Activation volume	V	Vpwl[phase]
Grain size exponent	m	mpwl[phase]
Fugacity exponent	r	rpwl[phase]
Fugacity	f	fpwl[phase]

Flow Law Database (selected entries from FlowLaws.c):

pwlv	Name	Source	Typical use
1	User-defined (η₀, n, Q)	Custom	Any
2	User-defined (τ₀ constrained)	Custom	Any
3	User-defined (A, n, Q)	Custom	Any
10	Wet Quartzite	Ranalli 1995	Upper crust
11	Maryland Diabase	Ranalli 1995	Lower crust
12	Dry Granulite	Ranalli 1995	Lower crust
19	Westerly Granite	Hansen & Carter 1983	Upper crust
25	Plagioclase An75	Ranalli 1995	Crust
40	Dry Olivine	Hirth & Kohlstedt 2003	Lithospheric mantle
41	Wet Olivine	Hirth & Kohlstedt 2003	Asthenosphere

Linear — Diffusion Creep (linv)

Equation: η_lin = A^(-1/n) · d^(m/n) · f^(-r/n) · exp((Q + P·V)/(n·R·T)) · ε̇_II^(1/n - 1)

Grain-size sensitive (m > 0). Requires grain size initialisation via SetGrainSize callback.

linv	Name	Source
40	Dry Olivine diffusion	Hirth & Kohlstedt 2003

When both pwlv = 40 and linv = 40 are set (e.g., for mantle), both dislocation and diffusion creep are active. They combine harmonically — the weaker mechanism dominates at each P-T-ε̇ condition.

Exponential — Peierls Creep (expv)

Low-temperature, high-stress plasticity. Activated by expv switch. Only effective below T_max ≈ 1200°C (capped at TmaxPeierls = 1200 + 273.15 K).

expv	Name	Source
40	Dry Olivine Peierls	Hirth & Kohlstedt 2003

Grain Boundary Sliding (gbsv)

Intermediate mechanism between dislocation and diffusion creep. Grain-size and stress dependent.

Constant Viscosity (cstv)

When cstv > 0, uses a constant viscosity η = η₀ (set via eta0[phase] in the .txt file). Overrides all creep mechanisms.

Elasticity

Maxwell visco-elastic model (enabled by elastic = 1):

η_ve = η · G · Δt / (η + G · Δt)

Stress update (objective rate): τ_new = 2·η_ve · (ε̇ + τ⁰_rotated / (2·G·Δt))

where τ⁰_rotated is the old stress rotated according to stress_rotation mode.

.txt parameter	Description
elastic = 1	Enable elasticity globally
G (per phase)	Shear modulus [Pa], typically ~1e10
stress_rotation	Stress objectivity: 0=none, 1=Jaumann (default), 2=analytical rotation

The stress_rotation parameter controls how old deviatoric stresses are rotated before the visco-elastic update (RheologyParticles.c). Mode 0 disables rotation (valid only for irrotational flow). Mode 1 (Jaumann) uses the vorticity tensor. Mode 2 uses an analytical rotation matrix.

Plasticity

Plasticity Mode Switch (plast)

plast	Model	Description
0	None	Plasticity disabled
1	Drucker-Prager	Standard shear yield (default)
2	Combined mode-I/mode-II	Popov et al. (2025) — shear + tensile yield

Drucker-Prager Yield (plast = 1)

Yield criterion:

τ_II ≤ C·cos(φ) + P·sin(φ)

If yield is exceeded, viscosity is reduced to enforce the yield stress: η_pl = τ_yield / (2·ε̇_II)

Combined Mode-I/Mode-II Yield (plast = 2)

Introduced in commit 8aecb92 (PR #159, Popov et al. 2025). Combines a standard Drucker-Prager shear yield surface with a tensile (mode-I) cap.

Additional per-phase parameter:

Parameter	Field	Description	Typical value
Tensile strength	T_st	Mode-I tensile cutoff [Pa]	-10e6 (must be negative)

Requirements:

• eta_vp > 0 for all phases (numerical stability)
• T_st set per phase (code default: -10e6 Pa)
• Works with both Picard and Picard2Newton solvers
• Optional: tensile_line_search = 1 for additional solver stability in tensile-dominated problems

Example scenarios: Popov2025_Pureshear_VEVP, Popov2025_Tensile_VEVP, RiftingCombinedYield

Strain Softening

Parameter	Description
C	Initial cohesion [Pa]
C_end	Final cohesion after softening [Pa]
phi	Initial friction angle [degrees]
phi_end	Final friction angle [degrees]
psi	Dilation angle [degrees]
psi_end	Final dilation angle [degrees]
pls_start	Accumulated plastic strain at which softening begins
pls_end	Accumulated plastic strain at which softening is complete
Slim	Stress limiter [Pa] — maximum deviatoric stress
sig_tens	Tensile strength [Pa]
eta_vp	Viscoplastic regularisation viscosity [Pa·s] (see below)
n_vp	Viscoplastic stress exponent (must be 1.0 in MD7)

Viscoplastic Regularisation (eta_vp)

Purpose: eta_vp adds a rate-dependent (viscous) overstress to the plastic yield surface, preventing the stress from dropping to exactly the yield stress. This controls shear band width and improves numerical convergence.

Modified yield condition:

τ_II = C·cos(φ) + P·sin(φ) + η_vp · λ̇

where λ̇ is the plastic multiplier rate (related to the magnitude of plastic strain rate).

Return mapping (RheologyDensity.c, line ~630): A local Newton iteration finds λ̇ such that:

F = τ_trial - η_ve·λ̇ - τ_yield - η_vp·λ̇ = 0

The viscoplastic overstress is then OverS = η_vp · λ̇.

Physical effect: Higher eta_vp → wider shear bands, smoother localisation. Lower → thinner bands, more mesh-sensitive. Setting eta_vp = 0 recovers ideal (rate-independent) plasticity.

Typical values by model scale:

Application scale	Recommended eta_vp	Example scenario
Lithospheric rifting / collision	2.5e20 Pa·s	RiftingMelting, RiftingCombinedYield
Crustal shear zones	1e19–1e20 Pa·s	CrustalShearBandingAnisotropy
Localised magma chambers	1e19 Pa·s	PressurizedMagmaChamber, TCMagmaticSystem
General range	1e18–1e22 Pa·s	—

Stability warning: Using eta_vp = 1e19 in lithospheric-scale rifting models (especially with melting = 1) can cause instability after ~100 time steps — negative viscosities and values ~1e60 in the output. Use eta_vp ≥ 2.5e20 for lithospheric-scale models with melting.

Note: n_vp (viscoplastic exponent) must be 1.0 in MD7 — power-law viscoplasticity (n_vp > 1) is not implemented and will exit(1).

The mat_prop Structure

Holds per-phase material properties (up to 20 phases), populated from the PHASE PROPERTIES section of the .txt file. Arrays indexed by phase ID.

Key Fields

Field	Physical quantity	Units
rho[phase]	Reference density	kg/m³
G[phase]	Shear modulus	Pa
Cp[phase]	Heat capacity	J/kg/K
k[phase]	Thermal conductivity	W/m/K
Qr[phase]	Radiogenic heat production	W/m³
alp[phase]	Thermal expansivity α	K⁻¹
bet[phase]	Compressibility β	Pa⁻¹
C[phase]	Cohesion	Pa
phi[phase]	Friction angle	degrees (converted to radians internally)
psi[phase]	Dilation angle	degrees
eta0[phase]	Reference viscosity	Pa·s
n[phase]	Stress exponent	-
aniso_factor[phase]	Anisotropy factor	- (1.0 = isotropic)
aniso_angle[phase]	Initial anisotropy angle	degrees

Density Model

Equation of state:

ρ = ρ₀ · (1 - α(T - T₀)) · (1 + β(P - P₀))

Parameter	Symbol	Typical value
rho	ρ₀	2700 (crust), 3300 (mantle) kg/m³
alp	α	3.2e-5 K⁻¹
bet	β	1e-11 Pa⁻¹

The density_model switch allows alternative formulations including phase-diagram-based density.

Marker-to-Node Viscosity Averaging (eta_average)

When multiple material phases share a grid cell, their marker viscosities are averaged to the node. The eta_average .txt parameter selects the method:

Value	Method	Best for
0	Arithmetic (default)	Best convergence orders (Vx 1.02, P 0.75 in SolVi 41→81)
1	Harmonic	Lowest absolute Vx error (1.0e-3 at 51×51) but worst convergence orders (0.45)
2	Geometric	Best absolute P error at fixed resolution (18% lower than arithmetic)

This averaging operates on the phase mixture within each cell. It is distinct from cell-face viscosity interpolation in the FD stencil (which is not currently implemented as harmonic).

Key Source Files

• Flow law database: MDLIB/FlowLaws.c
• Viscosity evaluation: MDLIB/RheologyDensity.c, MDLIB/AnisotropyRoutines.c
• Rheology interface: MDLIB/RheologyDensity.h
• Particle rheology: MDLIB/RheologyParticles.c