Level set tools

level_set_tools

This module provides a set of classes and functions enabling multi-material capabilities. Users initialise materials by instantiating the Material class and define the physical properties of material interfaces using field_interface. They instantiate the LevelSetSolver class by providing relevant parameters and call the solve method to request a solver update. Finally, they may call the entrainment function to calculate material entrainment in the simulation.

Material(*, density=None, B=None, RaB=None) dataclass

A material with physical properties for the level-set approach.

Expects material buoyancy to be defined using a value for either the reference density, buoyancy number, or compositional Rayleigh number.

Contains static methods to calculate the physical properties of a material. Methods implemented here describe properties in the simplest non-dimensional simulation setup and must be overriden for more complex scenarios.

Attributes:

Name	Type	Description
density	Optional[Number]	An integer or a float representing the reference density.
B	Optional[Number]	An integer or a float representing the buoyancy number.
RaB	Optional[Number]	An integer or a float representing the compositional Rayleigh number.
density_B_RaB	Optional[Number]	A string to notify how the buoyancy term is calculated.

__post_init__()

Checks instance field values.

Raises:

Type	Description
ValueError	Incorrect field types.

Source code in g-adopt/gadopt/level_set_tools.py

def __post_init__(self):
    """Checks instance field values.

    Raises:
        ValueError:
          Incorrect field types.
    """
    count_None = 0
    for field_var in fields(self):
        field_var_value = getattr(self, field_var.name)
        if isinstance(field_var_value, Number):
            self.density_B_RaB = field_var.name
        elif field_var_value is None:
            count_None += 1
        else:
            raise ValueError(
                "When provided, density, B, and RaB must have type int or float."
            )
    if count_None != 2:
        raise ValueError(
            "One, and only one, of density, B, and RaB must be provided, and it "
            "must be an integer or a float."
        )

viscosity(*args, **kwargs) staticmethod

Calculates dynamic viscosity (Pa s).

Source code in g-adopt/gadopt/level_set_tools.py

@staticmethod
def viscosity(*args, **kwargs):
    """Calculates dynamic viscosity (Pa s)."""
    return 1.0

thermal_expansion(*args, **kwargs) staticmethod

Calculates volumetric thermal expansion coefficient (K^-1).

Source code in g-adopt/gadopt/level_set_tools.py

@staticmethod
def thermal_expansion(*args, **kwargs):
    """Calculates volumetric thermal expansion coefficient (K^-1)."""
    return 1.0

thermal_conductivity(*args, **kwargs) staticmethod

Calculates thermal conductivity (W m^-1 K^-1).

Source code in g-adopt/gadopt/level_set_tools.py

@staticmethod
def thermal_conductivity(*args, **kwargs):
    """Calculates thermal conductivity (W m^-1 K^-1)."""
    return 1.0

specific_heat_capacity(*args, **kwargs) staticmethod

Calculates specific heat capacity at constant pressure (J kg^-1 K^-1).

Source code in g-adopt/gadopt/level_set_tools.py

@staticmethod
def specific_heat_capacity(*args, **kwargs):
    """Calculates specific heat capacity at constant pressure (J kg^-1 K^-1)."""
    return 1.0

internal_heating_rate(*args, **kwargs) staticmethod

Calculates internal heating rate per unit mass (W kg^-1).

Source code in g-adopt/gadopt/level_set_tools.py

@staticmethod
def internal_heating_rate(*args, **kwargs):
    """Calculates internal heating rate per unit mass (W kg^-1)."""
    return 0.0

thermal_diffusivity(*args, **kwargs) classmethod

Calculates thermal diffusivity (m^2 s^-1).

Source code in g-adopt/gadopt/level_set_tools.py

@classmethod
def thermal_diffusivity(cls, *args, **kwargs):
    """Calculates thermal diffusivity (m^2 s^-1)."""
    return cls.thermal_conductivity() / cls.density() / cls.specific_heat_capacity()

LevelSetSolver(level_set, velocity, tstep, tstep_alg, subcycles, epsilon, reini_params=None, solver_params=None)

Solver for the conservative level-set approach.

Solves the advection and reinitialisation equations for a level set function.

Attributes:

Name	Type	Description
mesh		The UFL mesh where values of the level set function exist.
level_set		The Firedrake function for the level set.
func_space_lsgp		The UFL function space where values of the projected level-set gradient are calculated.
level_set_grad_proj		The Firedrake function for the projected level-set gradient.
proj_solver		An integer or a float representing the reference density.
reini_params		A dictionary containing parameters used in the reinitialisation approach.
ls_ts		The G-ADOPT timestepper object for the advection equation.
reini_ts		The G-ADOPT timestepper object for the reinitialisation equation.
subcycles		An integer specifying the number of advection solves to perform.

Initialises the solver instance.

Parameters:

Name	Type	Description	Default
level_set	Function	The Firedrake function for the level set.	required
velocity	ListTensor	The UFL expression for the velocity.	required
tstep	Constant	The Firedrake function over the Real space for the simulation time step.	required
tstep_alg	ABCMeta	The class for the timestepping algorithm used in the advection solver.	required
subcycles	int	An integer specifying the number of advection solves to perform.	required
epsilon	Constant	A UFL constant denoting the thickness of the hyperbolic tangent profile.	required
reini_params	Optional[dict]	A dictionary containing parameters used in the reinitialisation approach.	None
solver_params	Optional[dict]	A dictionary containing solver parameters used in the advection and reinitialisation approaches.	None

Source code in g-adopt/gadopt/level_set_tools.py

def __init__(
    self,
    level_set: fd.Function,
    velocity: fd.ufl.tensors.ListTensor,
    tstep: fd.Constant,
    tstep_alg: abc.ABCMeta,
    subcycles: int,
    epsilon: fd.Constant,
    reini_params: Optional[dict] = None,
    solver_params: Optional[dict] = None,
):
    """Initialises the solver instance.

    Args:
        level_set:
          The Firedrake function for the level set.
        velocity:
          The UFL expression for the velocity.
        tstep:
          The Firedrake function over the Real space for the simulation time step.
        tstep_alg:
          The class for the timestepping algorithm used in the advection solver.
        subcycles:
          An integer specifying the number of advection solves to perform.
        epsilon:
          A UFL constant denoting the thickness of the hyperbolic tangent profile.
        reini_params:
          A dictionary containing parameters used in the reinitialisation approach.
        solver_params:
          A dictionary containing solver parameters used in the advection and
          reinitialisation approaches.
    """
    self.level_set = level_set
    self.tstep = tstep
    self.tstep_alg = tstep_alg
    self.subcycles = subcycles

    self.reini_params = reini_params or reini_params_default
    self.solver_params = solver_params or solver_params_default

    self.mesh = extract_unique_domain(level_set)
    self.func_space = level_set.function_space()

    self.func_space_lsgp = fd.VectorFunctionSpace(
        self.mesh, "CG", self.func_space.finat_element.degree
    )
    self.level_set_grad_proj = fd.Function(
        self.func_space_lsgp,
        name=f"Level-set gradient (projection) #{level_set.name()[-1]}",
    )

    self.proj_solver = self.gradient_L2_proj()

    self.ls_eq_attrs = {"u": velocity}
    self.reini_eq_attrs = {
        "level_set_grad": self.level_set_grad_proj,
        "epsilon": epsilon,
    }

    self.solvers_ready = False

gradient_L2_proj()

Constructs a projection solver.

Projects the level-set gradient from a discontinuous function space to the equivalent continuous one.

Returns:

Type	Description
LinearVariationalSolver	A Firedrake solver capable of projecting a discontinuous gradient field on
LinearVariationalSolver	a continuous function space.

Source code in g-adopt/gadopt/level_set_tools.py

def gradient_L2_proj(self) -> fd.variational_solver.LinearVariationalSolver:
    """Constructs a projection solver.

    Projects the level-set gradient from a discontinuous function space to the
    equivalent continuous one.

    Returns:
        A Firedrake solver capable of projecting a discontinuous gradient field on
        a continuous function space.
    """
    test_function = fd.TestFunction(self.func_space_lsgp)
    trial_function = fd.TrialFunction(self.func_space_lsgp)

    mass_term = fd.inner(test_function, trial_function) * fd.dx(domain=self.mesh)
    residual_element = (
        self.level_set * fd.div(test_function) * fd.dx(domain=self.mesh)
    )
    residual_boundary = (
        self.level_set
        * fd.dot(test_function, fd.FacetNormal(self.mesh))
        * fd.ds(domain=self.mesh)
    )
    residual = -residual_element + residual_boundary
    problem = fd.LinearVariationalProblem(
        mass_term, residual, self.level_set_grad_proj
    )

    solver_params = {
        "mat_type": "aij",
        "ksp_type": "preonly",
        "pc_type": "lu",
        "pc_factor_mat_solver_type": "mumps",
    }

    return fd.LinearVariationalSolver(problem, solver_parameters=solver_params)

update_level_set_gradient(*args, **kwargs)

Calls the gradient projection solver.

Can be used as a callback.

Source code in g-adopt/gadopt/level_set_tools.py

def update_level_set_gradient(self, *args, **kwargs):
    """Calls the gradient projection solver.

    Can be used as a callback.
    """
    self.proj_solver.solve()

set_up_solvers()

Sets up the time steppers for advection and reinitialisation.

Source code in g-adopt/gadopt/level_set_tools.py

def set_up_solvers(self):
    """Sets up the time steppers for advection and reinitialisation."""
    test = fd.TestFunction(self.func_space)

    advection_equation = Equation(
        test,
        self.func_space,
        scalar_eq.advection_term,
        mass_term=scalar_eq.mass_term,
        eq_attrs=self.ls_eq_attrs,
    )
    self.ls_ts = self.tstep_alg(
        advection_equation,
        self.level_set,
        self.tstep / self.subcycles,
        solver_parameters=self.solver_params,
    )

    reinitialisation_equation = Equation(
        test,
        self.func_space,
        reinitialisation_term,
        mass_term=scalar_eq.mass_term,
        eq_attrs=self.reini_eq_attrs,
    )
    self.reini_ts = self.reini_params["tstep_alg"](
        reinitialisation_equation,
        self.level_set,
        self.reini_params["tstep"],
        solver_parameters=self.solver_params,
    )

solve(step)

Updates the level-set function.

Calls advection and reinitialisation solvers within a subcycling loop. The reinitialisation solver can be iterated and might not be called at each simulation time step.

Parameters:

Name	Type	Description	Default
step	int	An integer representing the current simulation step.	required

Source code in g-adopt/gadopt/level_set_tools.py

def solve(self, step: int):
    """Updates the level-set function.

    Calls advection and reinitialisation solvers within a subcycling loop.
    The reinitialisation solver can be iterated and might not be called at each
    simulation time step.

    Args:
        step:
          An integer representing the current simulation step.
    """
    if not self.solvers_ready:
        self.set_up_solvers()
        self.solvers_ready = True

    for subcycle in range(self.subcycles):
        self.ls_ts.advance(0)

        if step >= self.reini_params["frequency"]:
            self.reini_ts.solution_old.assign(self.level_set)

        if step % self.reini_params["frequency"] == 0:
            for reini_step in range(self.reini_params["iterations"]):
                self.reini_ts.advance(
                    0, update_forcings=self.update_level_set_gradient
                )

            self.ls_ts.solution_old.assign(self.level_set)

reinitialisation_term(eq, trial)

Term for the conservative level set reinitialisation equation.

Implements terms on the right-hand side of Equation 17 from Parameswaran, S., & Mandal, J. C. (2023). A stable interface-preserving reinitialization equation for conservative level set method. European Journal of Mechanics-B/Fluids, 98, 40-63.

Source code in g-adopt/gadopt/level_set_tools.py

def reinitialisation_term(
    eq: Equation, trial: fd.Argument | fd.ufl.indexed.Indexed | fd.Function
) -> fd.Form:
    """Term for the conservative level set reinitialisation equation.

    Implements terms on the right-hand side of Equation 17 from
    Parameswaran, S., & Mandal, J. C. (2023).
    A stable interface-preserving reinitialization equation for conservative level set
    method.
    European Journal of Mechanics-B/Fluids, 98, 40-63.
    """
    sharpen_term = -trial * (1 - trial) * (1 - 2 * trial) * eq.test * eq.dx
    balance_term = (
        eq.epsilon
        * (1 - 2 * trial)
        * fd.sqrt(fd.inner(eq.level_set_grad, eq.level_set_grad))
        * eq.test
        * eq.dx
    )

    return sharpen_term + balance_term

field_interface_recursive(level_set, material_value, method)

Sets physical property expressions for each material.

Ensures that the correct expression is assigned to each material based on the level-set functions. Property transition across material interfaces are expressed according to the provided method.

Parameters:

Name	Type	Description	Default
level_set	list	A list of level-set UFL functions.	required
material_value	list	A list of physical property values applicable to each material.	required
method	str	A string specifying the nature of property transitions between materials.	required

Returns:

Type	Description
Expr	A UFL expression to calculate physical property values throughout the domain.

Raises:

Type	Description
ValueError	Incorrect method name supplied.

Source code in g-adopt/gadopt/level_set_tools.py

def field_interface_recursive(
    level_set: list, material_value: list, method: str
) -> fd.ufl.core.expr.Expr:
    """Sets physical property expressions for each material.

    Ensures that the correct expression is assigned to each material based on the
    level-set functions.
    Property transition across material interfaces are expressed according to the
    provided method.

    Args:
        level_set:
          A list of level-set UFL functions.
        material_value:
          A list of physical property values applicable to each material.
        method:
          A string specifying the nature of property transitions between materials.

    Returns:
        A UFL expression to calculate physical property values throughout the domain.

    Raises:
      ValueError: Incorrect method name supplied.

    """
    ls = fd.max_value(fd.min_value(level_set.pop(), 1), 0)

    if level_set:  # Directly specify material value on only one side of the interface
        match method:
            case "sharp":
                return fd.conditional(
                    ls > 0.5,
                    material_value.pop(),
                    field_interface_recursive(level_set, material_value, method),
                )
            case "arithmetic":
                return material_value.pop() * ls + field_interface_recursive(
                    level_set, material_value, method
                ) * (1 - ls)
            case "geometric":
                return material_value.pop() ** ls * field_interface_recursive(
                    level_set, material_value, method
                ) ** (1 - ls)
            case "harmonic":
                return 1 / (
                    ls / material_value.pop()
                    + (1 - ls)
                    / field_interface_recursive(level_set, material_value, method)
                )
            case _:
                raise ValueError(
                    "Method must be sharp, arithmetic, geometric, or harmonic."
                )
    else:  # Final level set; specify values for both sides of the interface
        match method:
            case "sharp":
                return fd.conditional(ls < 0.5, *material_value)
            case "arithmetic":
                return material_value[0] * (1 - ls) + material_value[1] * ls
            case "geometric":
                return material_value[0] ** (1 - ls) * material_value[1] ** ls
            case "harmonic":
                return 1 / ((1 - ls) / material_value[0] + ls / material_value[1])
            case _:
                raise ValueError(
                    "Method must be sharp, arithmetic, geometric, or harmonic."
                )

field_interface(level_set, material_value, method)

Executes field_interface_recursive with a modified argument.

Calls field_interface_recursive using a copy of the level-set list to ensure the original one is not consumed by the function call.

Parameters:

Name	Type	Description	Default
level_set	list	A list of level-set UFL functions.	required
material_value	list	A list of physical property values applicable to each material.	required
method	str	A string specifying the nature of property transitions between materials.	required

Returns:

Type	Description
Expr	A UFL expression to calculate physical property values throughout the domain.

Source code in g-adopt/gadopt/level_set_tools.py

def field_interface(
    level_set: list, material_value: list, method: str
) -> fd.ufl.core.expr.Expr:
    """Executes field_interface_recursive with a modified argument.

    Calls field_interface_recursive using a copy of the level-set list to ensure the
    original one is not consumed by the function call.

    Args:
        level_set:
          A list of level-set UFL functions.
        material_value:
          A list of physical property values applicable to each material.
        method:
          A string specifying the nature of property transitions between materials.

    Returns:
        A UFL expression to calculate physical property values throughout the domain.
    """
    return field_interface_recursive(level_set.copy(), material_value, method)

density_RaB(Simulation, level_set, func_space_interp, method='sharp')

Sets up buoyancy-related fields.

Assigns UFL expressions to buoyancy-related fields based on the way the Material class was initialised.

Parameters:

Name	Type	Description	Default
Simulation		A class representing the current simulation.	required
level_set	list	A list of level-set UFL functions.	required
func_space_interp	WithGeometry	A continuous UFL function space where material fields are calculated.	required
method	Optional[str]	An optional string specifying the nature of property transitions between materials.	'sharp'

Returns:

Type	Description
Constant	A tuple containing the reference density field, the density difference field,
Constant | Expr	the density field, the UFL expression for the compositional Rayleigh number,
Function	the compositional Rayleigh number field, and a boolean indicating if the
Constant | Expr	simulation is expressed in dimensionless form.

Raises:

Type	Description
ValueError	Inconsistent buoyancy-related field across materials.

Source code in g-adopt/gadopt/level_set_tools.py

def density_RaB(
    Simulation,
    level_set: list,
    func_space_interp: fd.functionspaceimpl.WithGeometry,
    method: Optional[str] = "sharp",
) -> tuple[
    fd.Constant,
    fd.Constant | fd.ufl.core.expr.Expr,
    fd.Function,
    fd.Constant | fd.ufl.core.expr.Expr,
    fd.Function,
    bool,
]:
    """Sets up buoyancy-related fields.

    Assigns UFL expressions to buoyancy-related fields based on the way the Material
    class was initialised.

    Args:
        Simulation:
          A class representing the current simulation.
        level_set:
          A list of level-set UFL functions.
        func_space_interp:
          A continuous UFL function space where material fields are calculated.
        method:
          An optional string specifying the nature of property transitions between
          materials.

    Returns:
        A tuple containing the reference density field, the density difference field,
        the density field, the UFL expression for the compositional Rayleigh number,
        the compositional Rayleigh number field, and a boolean indicating if the
        simulation is expressed in dimensionless form.

    Raises:
        ValueError: Inconsistent buoyancy-related field across materials.
    """
    density = fd.Function(func_space_interp, name="Density")
    RaB = fd.Function(func_space_interp, name="RaB")
    # Identify if the governing equations are written in dimensional form or not and
    # define accordingly relevant variables for the buoyancy term
    if all(material.density_B_RaB == "density" for material in Simulation.materials):
        dimensionless = False
        RaB_ufl = fd.Constant(1)
        ref_dens = fd.Constant(Simulation.reference_material.density)
        dens_diff = field_interface(
            level_set,
            [material.density - ref_dens for material in Simulation.materials],
            method=method,
        )
        density.interpolate(dens_diff + ref_dens)
    else:
        dimensionless = True
        ref_dens = fd.Constant(1)
        dens_diff = fd.Constant(1)
        if all(material.density_B_RaB == "B" for material in Simulation.materials):
            RaB_ufl = field_interface(
                level_set,
                [Simulation.Ra * material.B for material in Simulation.materials],
                method=method,
            )
        elif all(material.density_B_RaB == "RaB" for material in Simulation.materials):
            RaB_ufl = field_interface(
                level_set,
                [material.RaB for material in Simulation.materials],
                method=method,
            )
        else:
            raise ValueError(
                "All materials must share a common buoyancy-defining parameter."
            )
        RaB.interpolate(RaB_ufl)

    return ref_dens, dens_diff, density, RaB_ufl, RaB, dimensionless

entrainment(level_set, material_area, entrainment_height)

Calculates entrainment diagnostic.

Determines the proportion of a material that is located above a given height.

Parameters:

Name	Type	Description	Default
level_set	Function	A level-set Firedrake function.	required
material_area	Number	An integer or a float representing the total area occupied by a material.	required
entrainment_height	Number	An integer or a float representing the height above which the entrainment diagnostic is determined.	required

Returns:

Type	Description
	A float corresponding to the calculated entrainment diagnostic.

Source code in g-adopt/gadopt/level_set_tools.py

def entrainment(
    level_set: fd.Function, material_area: Number, entrainment_height: Number
):
    """Calculates entrainment diagnostic.

    Determines the proportion of a material that is located above a given height.

    Args:
        level_set:
          A level-set Firedrake function.
        material_area:
          An integer or a float representing the total area occupied by a material.
        entrainment_height:
          An integer or a float representing the height above which the entrainment
          diagnostic is determined.

    Returns:
        A float corresponding to the calculated entrainment diagnostic.
    """
    mesh_coords = fd.SpatialCoordinate(level_set.function_space().mesh())
    target_region = mesh_coords[1] >= entrainment_height
    material_entrained = fd.conditional(level_set < 0.5, 1, 0)

    return (
        fd.assemble(fd.conditional(target_region, material_entrained, 0) * fd.dx)
        / material_area
    )