Source code for jaxdem.domain

# SPDX-License-Identifier: BSD-3-Clause
# Part of the JaxDEM project – https://github.com/cdelv/JaxDEM
"""
Interface for defining domains. The domain performs boundary conditions coordinate transformation and computes the displacement vector according to the boundary conditions.
"""

import jax
import jax.numpy as jnp

from typing import ClassVar, Tuple
from abc import ABC, abstractmethod
from dataclasses import dataclass

from .factory import Factory
from typing import TYPE_CHECKING

if TYPE_CHECKING:
    from .state import State
    from .system import System


[docs] @jax.tree_util.register_dataclass @dataclass(slots=True) class Domain(Factory["Domain"], ABC): """ The base interface for defining the simulation domain and the effect of its boundary conditions. The `Domain` class defines how: - Relative displacement vectors between particles are calculated. - Particles' positions are "shifted" or constrained to remain within the defined simulation boundaries based on the boundary condition type. Notes ----- All concrete `Domain` implementations must support both 2D and 3D simulations. All methods must be JIT-compatible. Example ------- To define a custom domain, inherit from `Domain` and implement its abstract methods: >>> @Domain.register("my_custom_domain") >>> @jax.tree_util.register_dataclass >>> @dataclass(slots=True) >>> class MyCustomDomain(Domain): ... """ box_size: jax.Array """Length of the simulation domain along each dimension. Defines the size of the simulation box.""" anchor: jax.Array """Anchor position of the simulation domain. This represents the minimum coordinate (e.g., the "left-down corner") of the domain in each dimension.""" periodic: ClassVar[bool] = False """ Whether or not the domain is periodic. This is a class-level attribute that should be set to `True` for periodic boundary condition implementations. """
[docs] @staticmethod @abstractmethod @jax.jit def displacement(ri: jax.Array, rj: jax.Array, system: "System") -> jax.Array: r""" Computes the displacement vector between two particles :math:`r_i` and :math:`r_j`, considering the domain's boundary conditions. Parameters ---------- ri : jax.Array Position vector of the first particle :math:`r_i`. Shape `(dim,)`. rj : jax.Array Position vector of the second particle :math:`r_j`. Shape `(dim,)`. system : System The configuration of the simulation, containing the `domain` instance. Returns ------- jax.Array The displacement vector :math:`r_{ij} = r_i - r_j`, adjusted for boundary conditions. Shape `(dim,)`. Raises ------ NotImplementedError This is an abstract method and must be implemented by subclasses. Example ------- """ raise NotImplemented
[docs] @staticmethod @abstractmethod @jax.jit def shift(state: "State", system: "System") -> Tuple["State", "System"]: """ Applies boundary conditions to particles state. This method updates the `state` based on the domain's rules, ensuring particles remain within the simulation box or handle interactions at boundaries appropriately (e.g., reflection, wrapping). Parameters ---------- state : State The current state of the simulation. system : System The configuration of the simulation. Returns ------- Tuple[State, System] A tuple containing the updated `State` object adjusted by the boundary conditions and the `System` object. Raises ------ NotImplementedError This is an abstract method and must be implemented by subclasses. Example ------- """ raise NotImplemented
[docs] @Domain.register("free") @jax.tree_util.register_dataclass @dataclass(slots=True) class FreeDomain(Domain): """ A `Domain` implementation representing an unbounded, "free" space. In a `FreeDomain`, there are no explicit boundary conditions applied to particles. Particles can move indefinitely in any direction, and the concept of a "simulation box" is only used to define the bounding box of the system. Notes ----- - The `box_size` and `anchor` attributes are dynamically updated in the `shift` method to encompass all particles. Some hashing tools require the domain size. Example ------- >>> system = jaxdem.System.create(dim=2, domain_type="free", domain_kw=dict(box_size=jnp.array([10., 10.]), anchor=jnp.array([0., 0.]))) >>> >>> # After a step, particle moves, and the domain's effective box_size and anchor update >>> state, system = sim_system.domain.shift(state, system) >>> print("Updated Domain Anchor:", system.domain.anchor) >>> print("Updated Domain Box Size:", system.domain.box_size) """
[docs] @staticmethod @jax.jit def displacement(ri: jax.Array, rj: jax.Array, _: "System") -> jax.Array: r""" Computes the displacement vector between two particles. In a free domain, the displacement is simply the direct vector difference between the particle positions. Parameters ---------- ri : jax.Array Position vector of the first particle :math:`r_i`. rj : jax.Array Position vector of the second particle :math:`r_j`. _ : System The system object. Returns ------- jax.Array The direct displacement vector :math:`r_i - r_j`. """ return ri - rj
[docs] @staticmethod @jax.jit def shift(state: "State", system: "System") -> Tuple["State", "System"]: """ Updates the `System`'s domain `anchor` and `box_size` to encompass all particles. Does not apply any transformations to the state. Parameters ---------- state : State The current state of the simulation. system : System The current system configuration. Returns ------- Tuple[State, System] The original `State` object (unchanged) and the `System` object with updated `domain.anchor` and `domain.box_size`. """ p_min = jnp.min(state.pos - state.rad[..., None], axis=-2) p_max = jnp.max(state.pos + state.rad[..., None], axis=-2) system.domain.anchor = p_min system.domain.box_size = p_max - p_min return state, system
[docs] @Domain.register("reflect") @jax.tree_util.register_dataclass @dataclass(slots=True) class ReflectDomain(Domain): """ A `Domain` implementation that enforces reflective boundary conditions. Particles that attempt to move beyond the defined `box_size` will have their positions reflected back into the box and their velocities reversed in the direction normal to the boundary. Notes ----- - The reflection occurs at the boundaries defined by `anchor` and `anchor + box_size`. Example ------- >>> system = jaxdem.System.create(dim=2, domain_type="reflect", domain_kw=dict(box_size=jnp.array([10., 7.]), anchor=jnp.array([1., 0.]))) """
[docs] @staticmethod @jax.jit def displacement(ri: jax.Array, rj: jax.Array, _: "System") -> jax.Array: r""" Computes the displacement vector between two particles. In a reflective domain, the displacement is simply the direct vector difference. Parameters ---------- ri : jax.Array Position vector of the first particle :math:`r_i`. rj : jax.Array Position vector of the second particle :math:`r_j`. _ : System The system object. Returns ------- jax.Array The direct displacement vector :math:`r_i - r_j`. """ return ri - rj
[docs] @staticmethod @jax.jit def shift(state: "State", system: "System") -> Tuple["State", "System"]: """ Applies reflective boundary conditions to particles. Particles are checked against the domain boundaries. If a particle attempts to move beyond a boundary, its position is reflected back into the box, and its velocity component normal to that boundary is reversed. .. math:: l &= a + R \\\\ u &= a + B - R \\\\ v' &= \\begin{cases} -v & \\text{if } r < l \\text{ or } r > u \\\\ v & \\text{otherwise} \\end{cases} \\\\ r' &= \\begin{cases} 2l - r & \\text{if } r < l \\\\ r & \\text{otherwise} \\end{cases} \\\\ r'' &= \\begin{cases} 2u - r' & \\text{if } r' > u \\\\ r' & \\text{otherwise} \\end{cases} r = r'' where: - :math:`r` is the current particle position (:attr:`state.pos`) - :math:`v` is the current particle velocity (:attr:`state.vel`) - :math:`a` is the domain anchor (:attr:`system.domain.anchor`) - :math:`B` is the domain box size (:attr:`system.domain.box_size`) - :math:`R` is the particle radius (:attr:`state.rad`) - :math:`l` is the lower boundary for the particle center - :math:`u` is the upper boundary for the particle center TO DO: Ensure correctness when adding different types of shapes and angular vel Parameters ---------- state : State The current state of the simulation. system : System The configuration of the simulation. Returns ------- Tuple[State, System] The updated `State` object with reflected positions and velocities, and the `System` object. """ lower_bound = system.domain.anchor + state.rad[:, None] upper_bound = system.domain.anchor + system.domain.box_size - state.rad[:, None] outside_lower = state.pos < lower_bound outside_upper = state.pos > upper_bound state.vel = jnp.where(outside_lower + outside_upper, -state.vel, state.vel) reflected_pos = jnp.where( outside_lower, 2.0 * lower_bound - state.pos, state.pos ) reflected_pos = jnp.where( outside_upper, 2.0 * upper_bound - reflected_pos, reflected_pos ) state.pos = reflected_pos return state, system
[docs] @Domain.register("periodic") @jax.tree_util.register_dataclass @dataclass(slots=True) class PeriodicDomain(Domain): """ A `Domain` implementation that enforces periodic boundary conditions. Particles that move out of one side of the simulation box re-enter from the opposite side. The displacement vector between particles is computed using the minimum image convention. Notes ----- - This domain type is periodic (`periodic = True`). - The `shift` method wraps particles back into the primary simulation box. Example ------- >>> system = jaxdem.System.create(dim=2, domain_type="periodic", domain_kw=dict(box_size=jnp.array([10., 7.]), anchor=jnp.array([1., 0.]))) """ periodic: ClassVar[bool] = True
[docs] @staticmethod @jax.jit def displacement(ri: jax.Array, rj: jax.Array, system: "System") -> jax.Array: """ Computes the minimum image displacement vector between two particles :math:`r_i` and :math:`r_j`. For periodic boundary conditions, the displacement is calculated as the shortest vector that connects :math:`r_j` to :math:`r_i`, potentially by crossing periodic boundaries. Parameters ---------- ri : jax.Array Position vector of the first particle :math:`r_i`. rj : jax.Array Position vector of the second particle :math:`r_j`. system : System The configuration of the simulation, containing the `domain` instance with `anchor` and `box_size` for periodicity. Returns ------- jax.Array The minimum image displacement vector: .. math:: & r_{ij} = (r_i - a) - (r_j - a) \\\\ & r_{ij} = r_{ij} - B \\cdot \\text{round}(r_{ij}/B) where: - :math:`a` is the domain anchor (:attr:`system.domain.anchor`) - :math:`B` is the domain box size (:attr:`system.domain.box_size`) """ rij = ri - rj return rij - system.domain.box_size * jnp.round(rij / system.domain.box_size)
[docs] @staticmethod @jax.jit def shift(state: "State", system: "System") -> Tuple["State", "System"]: """ Wraps particles back into the primary simulation box. .. math:: r = r - B \\cdot \\text{floor}((r - a)/B) \\\\ where: - :math:`a` is the domain anchor (:attr:`system.domain.anchor`) - :math:`B` is the domain box size (:attr:`system.domain.box_size`) Parameters ---------- state : State The current state of the simulation. system : System The configuration of the simulation. Returns ------- Tuple[State, System] The updated `State` object with wrapped particle positions, and the `System` object. """ state.pos -= system.domain.box_size * jnp.floor( (state.pos - system.domain.anchor) / system.domain.box_size ) return state, system