mdsuite/transformations/thermal_flux.py
"""
MDSuite: A Zincwarecode package.
License
-------
This program and the accompanying materials are made available under the terms
of the Eclipse Public License v2.0 which accompanies this distribution, and is
available at https://www.eclipse.org/legal/epl-v20.html
SPDX-License-Identifier: EPL-2.0
Copyright Contributors to the Zincwarecode Project.
Contact Information
-------------------
email: zincwarecode@gmail.com
github: https://github.com/zincware
web: https://zincwarecode.com/
Citation
--------
If you use this module please cite us with:
Summary
-------
"""
import typing
import numpy as np
import tensorflow as tf
from mdsuite.database.mdsuite_properties import mdsuite_properties
from mdsuite.transformations.transformations import MultiSpeciesTrafo
class ThermalFlux(MultiSpeciesTrafo):
"""Transformation to calculate the integrated heat current (positions * energies)."""
def __init__(self):
super(ThermalFlux, self).__init__(
input_properties=[
mdsuite_properties.stress,
mdsuite_properties.velocities,
mdsuite_properties.kinetic_energy,
mdsuite_properties.potential_energy,
],
output_property=mdsuite_properties.thermal_flux,
scale_function={"linear": {"scale_factor": 5}},
)
def transform_batch(
self,
batch: typing.Dict[str, typing.Dict[str, tf.Tensor]],
carryover: typing.Any = None,
) -> tf.Tensor:
fluxes = []
for properties in batch.values():
stress = properties[mdsuite_properties.stress.name]
vel = properties[mdsuite_properties.velocities.name]
ke = properties[mdsuite_properties.kinetic_energy.name]
pe = properties[mdsuite_properties.potential_energy.name]
phi_x = (
stress[:, :, 0] * vel[:, :, 0]
+ stress[:, :, 3] * vel[:, :, 1]
+ stress[:, :, 4] * vel[:, :, 2]
)
phi_y = (
stress[:, :, 3] * vel[:, :, 0]
+ stress[:, :, 1] * vel[:, :, 1]
+ stress[:, :, 5] * vel[:, :, 2]
)
phi_z = (
stress[:, :, 4] * vel[:, :, 0]
+ stress[:, :, 5] * vel[:, :, 1]
+ stress[:, :, 2] * vel[:, :, 2]
)
phi = np.dstack([phi_x, phi_y, phi_z])
phi_sum_atoms = phi.sum(axis=0)
# phi_sum_atoms = (
# phi_sum_atoms / self.experiment.units["NkTV2p"]
# ) # factor for units lammps nktv2p
# TODO why is there a unit conversion in the transformation?
energy = ke + pe
energy_velocity = energy * vel
energy_velocity_atoms = tf.reduce_sum(energy_velocity, axis=0)
fluxes.append(energy_velocity_atoms - phi_sum_atoms)
return tf.add_n(fluxes)