Skip to main content
Version: Latest

Poromechanics

DES3D can couple the flow of one pore fluid with deformation of the solid skeleton. Pore pressure changes the effective stress and therefore the strength, and deformation changes the pore pressure. Optional extensions cover partly saturated ground, groundwater heat transport, thermal pressurization, evolving porosity and permeability, and an immobile matrix reservoir. Fluid and heat transport use explicit substeps. The mechanical solver can use ordinary explicit stepping, GVS or PT relaxation; local pressure/content conversions do not introduce a global implicit flow solve.

Hydraulics is enabled with control.has_hydraulic_diffusion = yes.

Conventions​

Stress and strain are positive in tension, as elsewhere in DES3D. Pore pressure is positive in compression, so suction is . The vertical coordinate points up, and gravity acts in . The volumetric strain is and the mean stress .

Governing equations​

Effective stress​

The skeleton responds to the effective stress (Biot, 1941):

where is the total stress and the Biot–Willis coefficient (mat.biot_coeff). With mat.derive_biot_coeff_from_bulk_moduli = yes, DES3D sets from the drained bulk modulus (mat.bulk_modulus) and the grain modulus (mat.bulk_modulus_s) (Biot and Willis, 1957).

The plastic models use the effective stress in their yield functions. A rise in pore pressure moves the Mohr circle toward failure without changing the deviatoric stress. The stored stress remains the total stress, which balances the external loads.

Fluid content​

The default control.hydraulic_content_model = 0 uses the linear excess fluid volume per unit bulk volume (Rice and Cleary, 1976)

with porosity (mat.porosity), fluid bulk modulus (mat.fluid_bulk_modulus) and Biot modulus . The thermal term applies only with thermal pressurization (below). Valid materials satisfy .

Darcy flow​

The Darcy flux is

where is the fluid viscosity (mat.fluid_visc), the reference fluid density (mat.fluid_rho0) and the relative permeability ( when saturated). The permeability tensor is the scalar intrinsic permeability (mat.hydraulic_perm, m²) times a dimensionless symmetric positive-definite tensor (mat.permeability_tensor = [xx,yy,zz,xy,xz,yz], isotropic by default). In 2D its x–z part is used. Fluid mass balance closes the system:

where is a volumetric source from wells or point sources.

Weight of the saturated rock​

For the default hydraulic model, gravity uses the mixture density

with the temperature-dependent solid density and saturation ( when saturated). The same density loads a wet Winkler foundation. The reference solid density is mat.rho0. Without hydraulics, gravity uses the solid density alone.

This gravity density must be distinguished from inertial mass. With control.is_quasi_static = yes, ordinary explicit mechanics and GVS use fictitious masses set by mass scaling. The default hydraulic model's unscaled mass and wave estimates use the saturated mixture (), even when gravity uses a lower saturation. The finite-density model below has a separate inventory-based weight.

Coupling in time​

Each physical step advances mechanics once and pressure over the same interval. Which quantity drives the pressure depends on the mechanical solver.

Strain-driven form (default content model). With ordinary explicit mechanics or GVS and pressure feedback enabled, pressure advances first from the strain rate projected from the current velocity:

The stress and nodal forces then use the new pressure. Unscaled physical inertia also uses this ordering. The sealed elastic limit gives the undrained longitudinal wave; relative-fluid inertia is not integrated as a separate momentum equation.

Fixed-stress form (PT mechanics). Mechanics first reaches equilibrium for the step. Pressure then follows from the accepted change in mean total stress, with the fixed-stress storage (Kim et al., 2011):

in 3D. In 2D, is the in-plane mean stress and . The pressure change enters the effective stress in the next mechanical update. Saving it in checkpoints makes restarts exact. Because mechanics sees the pressure one step late, the coupling error is first order in .

control.has_poroelastic_pressure_feedback = no removes the mechanical source term (the or term). Pressure still diffuses and still acts on the effective stress. In quasistatic explicit/GVS runs with feedback disabled, storage is , as in the reduced fixed-stress flow model; unscaled physical inertia uses . Removing feedback therefore does not mean keeping the same diffusivity as every coupled explicit case. It does not disable the pressure's effect on skeleton strength.

Pressure sampled by a plastic return​

The fixed-stress mechanical return samples , while a completed saved frame pairs its stress with . For a saturated Mohr–Coulomb material, let and . The admissible slack at the sampled pressure and the saved-frame diagnostic obey

A negative saved-frame slack during a pressure rise can therefore measure coupling lag. In the recorded pressure-driven slope, doubling 404 to 808 steps reduces the minimum in-plane saved-pressure slack from −205.617 to −102.843 Pa; using the sampled pressure gives −0.000744 and −0.000226 Pa. This timing check does not replace the three-principal material return checks. Hydraulic subcycling does not remove the mechanical lag.

After plastic yielding, the pressure source uses accepted total volumetric strain, rather than converting stress through elastic moduli. The sealed plastic cell checks the matching fixed-stress recurrence. With retention, the Bishop cell checks its own rate equation and matching stabilization.

Hydraulic substeps​

Pressure diffusion often needs smaller steps than mechanics. Within each physical step DES3D advances pressure in explicit substeps. Their size is set by the assembled lumped storage and conductance. A long physical step, including control.fixed_dt, can therefore contain many stable transport substeps. This does not make an oversized mechanical step stable or remove the coupling lag. Mechanics, porosity and other histories advance once per physical step. Scheduled flux and source amounts use their exact schedule integrals; prescribed pressures are imposed at substep endpoints.

Boundary and initial conditions​

Each face has a hydraulic type, bc.hbc_x0 to bc.hbc_z1 (y faces in 3D only):

TypeCondition
0No normal flow (default)
1Hold the initial nodal pressure
2Prescribed pressure from bc.hbc_pressure (Pa)
3Prescribed outward Darcy flux from bc.hbc_flux (m/s)

bc.hbc_pressure and bc.hbc_flux list six values in the order x0 x1 y0 y1 z0 z1. A shared piecewise-linear multiplier, bc.hbc_schedule_time_in_yr and bc.hbc_schedule_ratio with bc.num_hbc_schedule entries, scales types 2 and 3 in time. A repeated time makes a step; the ratio is held outside the table. Two faces that meet must prescribe the same pressure at their shared nodes.

ic.pore_pressure_option sets the initial pressure:

ValueInitial pressure
0 (default)Hydrostatic, with water table = ic.water_table_z, plus ic.excess_pore_pressure away from the boundaries
1Linear in between the pressures prescribed on the x0 and x1 faces
2Uniform, ic.pore_pressure_value

Above the water table the hydrostatic pressure is negative (suction).

Wells and point sources​

Wells are vertical screens with a prescribed volume rate: bc.num_wells, bc.well_x, bc.well_y (3D), bc.well_z_top, bc.well_z_bottom and bc.well_rate. The rate is m³/s in 3D and m²/s per unit out-of-plane length in 2D; positive extracts. The rate is spread uniformly along the screen and integrated exactly against the element shape functions. All wells share one schedule, bc.well_schedule_time_in_yr and bc.well_schedule_ratio. Each step applies the exact time average of the schedule, so the pumped volume does not depend on the step size.

Point sources ([injection] section) supply or extract fluid at chosen points, with a constant rate or a total amount over a time window. Positive rates supply fluid. With fluid heat advection, an injected temperature can be set. A parallel [heat_injection] section supplies heat at points.

Partly saturated flow​

Setting mat.retention_model to a nonzero value per material enables Richards flow (Richards, 1931) with the gas phase at zero gauge pressure. With suction , effective saturation and residual saturation (mat.residual_saturation), the saturation is :

retention_model
01 (saturated)1
1 van Genuchten–Mualem,
2 Brooks–Corey–Burdine for , else 1
3 Gardner

Here is mat.retention_alpha (1/Pa) and is mat.retention_n (van Genuchten, 1980; Mualem, 1976). For Brooks–Corey, is the air-entry pressure and the pore-size index.

The pressure-dependent nodal content is , where is in the strain-driven form or in the fixed-stress form. Mechanical source increments are assembled separately. DES3D updates this content explicitly and inverts it for pressure node by node. Its positive capacity gives one local root. Effective stress uses Bishop's form with (Lu et al., 2010):

Suction therefore adds apparent cohesion to a partly saturated slope. dppressure remains the raw . Each element evaluates its own Bishop increment with its own retention law, averaging the nodal products over the element. This differs from multiplying an averaged saturation by an averaged pressure. Matching fixed-stress terms use the same element law. Elements saturated at both states keep the saturated arithmetic. Retention cannot yet be combined with evolving porosity or permeability, thermal pressurization, fluid heat advection or matrix exchange.

Evolving porosity and permeability​

With mat.porosity_model = 1, porosity follows the deformation through an element history, the pore strain . Each step adds

where is the solid thermal expansivity (mat.alpha, used only with thermal pressurization), the material porosity and recovery of previously accepted plastic pore dilation (zero unless pore healing is enabled). A step that would leave stops the run with a diagnostic.

mat.permeability_model lets permeability follow porosity:

ValuePermeability
0 (default)Fixed,
1 Kozeny–Carman
2 power law, = mat.permeability_exponent

(Costa, 2006; Yang et al., 2021).

Recovery of accepted plastic opening​

Recovery uses signed plastic volume, including the out-of-plane stress in plane strain. Reversible pressure and thermal increments are excluded from the elastic stress change:

For remaining recoverable opening , each accepted interval uses

Closure precedes the new plastic increment; compaction reduces opening and is not itself healed. Elastic dilation does not enter . With mat.pore_healing_time ,

where mat.pore_healing_temperature is in K, mat.pore_healing_energy is in J/mol and J/(mol K). Mixed elements average recovery rates by marker count. Frozen pressure, strain and temperature give expelled pore volume ; its negative increment enters the hydraulic source. Changing permeability alone would omit it.

During a drained hold starting at the accepted state ,

Together with the chosen permeability law, this is the independent hold reference in check_recovery.py. The recorded 2D/3D controls use finite grain stiffness, signed dilation/compaction, elastic holds, restart, changed timestep and remeshing. PT trials freeze recovery; acceptance commits it once. Checkpoints require its history when enabling recovery. The qualification covers this property law and tangent Biot coupling; finite-density recovery energy and retentive remapping are outside it.

Matrix storage​

mat.matrix_storage (1/Pa per bulk volume) and mat.matrix_exchange (1/(Pa s)) add an immobile reservoir to each element, as in a double-porosity medium (Warren and Root, 1963). The reservoir exchanges fluid with the mobile pressure at the rate . It starts at the mobile pressure, and the volume it gains is removed from the flow exactly. Its pressure is saved as matrix pressure.

Heat transport and thermal pressurization​

Advection by groundwater​

With control.has_fluid_thermal_advection = yes, temperature is carried by the Darcy flux:

with fluid heat capacity (mat.fluid_heat_capacity). This is the default heat-capacity model; evolving storage requires the separate finite-density inventory below for retained sensible-energy accounting. The fluid density in Darcy gravity varies with temperature (Boussinesq), , with = mat.fluid_alpha and the surface temperature. This drives free convection. Heat diffusion also runs in explicit substeps within each physical step. Thermal boundary types bc.tbc_x0 to bc.tbc_z1 select insulated (0) or prescribed-temperature (1) faces, with values in bc.tbc_val_* (K).

Thermal pressurization​

With control.has_thermal_pressurization = yes, heating expands both the solid and the pore fluid (McTigue, 1986; Rice, 2006). The solid adds a thermal stress to each normal component, and the fluid content gains the term with

In a sealed or poorly drained region, heating therefore raises pore pressure and lowers effective stress. It requires hydraulic and thermal diffusion and an elastic bulk response.

Remeshing​

After a remesh, nodal pressure is interpolated onto the new mesh like other nodal fields. For saturated material with fixed properties, DES3D then adjusts the pressure on free nodes by a common offset, so the total linear Biot content is the same before and after the remesh. Prescribed-pressure nodes keep their values; domain changes and boundary reservoir reactions are accounted for separately. This is a global constraint, not conservation element by element. Partly saturated material and evolving porosity in the default model are interpolated without this adjustment. The optional inventory model has its own remap, below.

Finite-density water and sensible heat​

control.hydraulic_content_model = 1 selects a saturated inventory model. It requires pressure feedback, mat.porosity_model = 1, an elastic bulk response and control.is_quasi_static = yes; ordinary explicit mechanics, GVS and PT are available, but unscaled physical inertia is excluded. Retention, matrix exchange, surface processes and material phase conversion are outside this mode. All solid materials must share one fluid law.

Constant compressibility and expansivity give

is the saved EOS temperature origin and the reference pressure is zero. This is a constitutive EOS, not a tabulated water-property model. The porosity law above is integrated over the accepted path. A moving mesh supplies ; a fixed mesh uses the virtual volume . That convention does not establish finite-strain mechanical accuracy.

The forward transport step uses the starting substep's Darcy mass and sensible-heat fluxes. Converting the updated local inventories to pressure and temperature requires a local root, without a global implicit transport solve. A fixed-stress numerical mass buffer is recorded separately: the split closes the budget, while physical mass alone retains a coupling defect that must decrease with physical-step refinement. is neither fluid density nor heat capacity.

For heat transport this mode requires both thermal pressurization and fluid thermal advection. Retained sensible energy is

where is the retained solid heat-capacity inventory. Injection uses inlet temperature and density; outflow uses the local state. Remapping transfers physical water, the numerical buffer, solid mass, , sensible energy and accepted pending shear energy, with domain and reservoir exchanges recorded separately. The constraint is global, not exact local inventory matching. Checkpoints preserve these inventories and their time levels when the next timestep changes.

Full pressure work and thermoelastic caloric energy are outside this sensible-heat model. A closed ledger alone does not qualify a general thermal plume or complete earthquake-cycle energy balance.

Output​

Saved frames contain pore pressure and, when the corresponding model is active, porosity, permeability, saturation, effective saturation, relative permeability and matrix pressure. The field fluid balance lists, for the last step, the change in stored fluid, the mechanical and thermal source, well extraction, net boundary inflow and their residual. The residual measures how well the discrete update balances its own fluid volume.

With content model 1, fluid mass and hydraulic split buffer separate physical mass from the numerical reservoir. The six entries of fluid mass balance are physical stored-mass change, buffer change, extraction, boundary inflow, reservoir reset and residual. With heat transport, sensible energy balance separates retained-energy change, supplied heat, well enthalpy, carried boundary heat, thermal reaction and residual. These mass and energy diagnostics are distinct from the default five-entry fluid balance.

Parameters​

ParameterDefaultMeaning
control.has_hydraulic_diffusionnoEnable pore-pressure transport
control.has_poroelastic_pressure_feedbackyesLet deformation drive pore pressure
control.hydraulic_content_model00 linear Biot/retention, 1 saturated finite-density inventory
mat.porosity[0.0]Porosity
mat.hydraulic_perm[1e-14]Intrinsic permeability (m²)
mat.permeability_tensor[1,1,1,0,0,0]Dimensionless anisotropy tensor
mat.fluid_visc[1.002e-3]Fluid viscosity (Pa s)
mat.fluid_rho0[1000]Reference fluid density (kg/m³)
mat.fluid_bulk_modulus[2.17e9]Fluid bulk modulus (Pa)
mat.fluid_alpha[2.07e-4]Fluid thermal expansivity (1/K)
mat.fluid_heat_capacity[4180]Fluid heat capacity (J/(kg K))
mat.biot_coeff[1.0]Biot–Willis coefficient
mat.bulk_modulus_s[37e9]Grain bulk modulus (Pa)
mat.derive_biot_coeff_from_bulk_modulinoUse
mat.retention_model[0]0 saturated, 1 van Genuchten–Mualem, 2 Brooks–Corey, 3 Gardner
mat.retention_alpha[1e-4]Retention parameter (1/Pa)
mat.retention_n[2]Retention exponent
mat.residual_saturation[0]Residual saturation
mat.porosity_model00 fixed, 1 evolving with deformation
mat.permeability_model00 fixed, 1 Kozeny–Carman, 2 power law
mat.permeability_exponent[3]Power-law exponent
mat.pore_healing_time[0]Recovery timescale of accepted plastic pore dilation; zero disables recovery
mat.matrix_storage, mat.matrix_exchange[0], [0]Immobile reservoir storage (1/Pa) and exchange coefficient (1/(Pa s))
control.has_fluid_thermal_advectionnoHeat transport by Darcy flow
control.has_thermal_pressurizationnoThermal stress and thermal fluid content
bc.hbc_x0 … bc.hbc_z10Hydraulic boundary type per face
bc.hbc_pressure, bc.hbc_fluxzerosFace values for types 2 and 3
ic.pore_pressure_option0Initial pressure field
ic.water_table_z0Initial water-table height (m)

References​

  • Biot, M. A. (1941). General theory of three-dimensional consolidation. Journal of Applied Physics, 12(2), 155–164. https://doi.org/10.1063/1.1712886
  • Biot, M. A., and Willis, D. G. (1957). The elastic coefficients of the theory of consolidation. Journal of Applied Mechanics, 24, 594–601. https://doi.org/10.1115/1.4011606
  • Costa, A. (2006). Permeability-porosity relationship: A reexamination of the Kozeny-Carman equation based on a fractal pore-space geometry assumption. Geophysical Research Letters, 33, L02318. https://doi.org/10.1029/2005GL025134
  • Kim, J., Tchelepi, H. A., and Juanes, R. (2011). Stability and convergence of sequential methods for coupled flow and geomechanics: Fixed-stress and fixed-strain splits. Computer Methods in Applied Mechanics and Engineering, 200, 1591–1606. https://doi.org/10.1016/j.cma.2010.12.022
  • Lu, N., Godt, J. W., and Wu, D. T. (2010). A closed-form equation for effective stress in unsaturated soil. Water Resources Research, 46, W05515. https://doi.org/10.1029/2009WR008646
  • McTigue, D. F. (1986). Thermoelastic response of fluid-saturated porous rock. Journal of Geophysical Research, 91(B9), 9533–9542. https://doi.org/10.1029/JB091iB09p09533
  • Mualem, Y. (1976). A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research, 12(3), 513–522. https://doi.org/10.1029/WR012i003p00513
  • Rice, J. R. (2006). Heating and weakening of faults during earthquake slip. Journal of Geophysical Research, 111, B05311. https://doi.org/10.1029/2005JB004006
  • Rice, J. R., and Cleary, M. P. (1976). Some basic stress diffusion solutions for fluid-saturated elastic porous media with compressible constituents. Reviews of Geophysics, 14(2), 227–241. https://doi.org/10.1029/RG014i002p00227
  • Richards, L. A. (1931). Capillary conduction of liquids through porous mediums. Physics, 1, 318–333. https://doi.org/10.1063/1.1745010
  • van Genuchten, M. T. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal, 44, 892–898. https://doi.org/10.2136/sssaj1980.03615995004400050002x
  • Warren, J. E., and Root, P. J. (1963). The behavior of naturally fractured reservoirs. SPE Journal, 3(3), 245–255. https://doi.org/10.2118/426-PA
  • Yang, Z., Yehya, A., Iwalewa, T. M., and Rice, J. R. (2021). Effect of permeability evolution in fault damage zones on earthquake recurrence. Journal of Geophysical Research: Solid Earth, 126, e2021JB021787. https://doi.org/10.1029/2021JB021787