Restarting Simulations

Restarting a Widom Simulation

It is possible to restart a Widom simulation in two scenarions. First, if the previous simulation fail and you want to continue it from the last saved point. Second, if you believe that the number of steps that you have run is not enough to reach the convergence and you want to continue the simulation from the last saved point.

Since all moves in the Widom method are virtual moves, the only information that is loaded is the interaction energy for each step. The new simulation is completly independent of the previous one, and the previous data will only be used to calculate the average enthaly of adsorption and the Henry coefficient.

On the flames/examples/Basic/1-Restart_Widom folder you will find the cif file for the Mg-MOF-74. The simulation will be executed with the machine learning potential MACE, using the medium-0b2 pre-trained foundation model with D3(0) dispersion correction.

The script below will run the Widom simulation at 298 K (25°C).

import os

# Hide UserWarning and RuntimeWarning messages
import warnings

import ase
import torch
from ase.data import vdw_radii
from ase.io import read
from mace.calculators import mace_mp

from flames.widom import Widom

warnings.filterwarnings("ignore", category=UserWarning)
warnings.filterwarnings("ignore", category=RuntimeWarning)

os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"

device = "cuda" if torch.cuda.is_available() else "cpu"

FrameworkPath = "mg-mof-74.cif"
AdsorbatePath = "co2.xyz"

model = mace_mp(
    model="medium-0b2",
    dispersion=False,
    damping="zero",  # choices: ["zero", "bj", "zerom", "bjm"]
    dispersion_xc="pbe",
    default_dtype="float32",
    device=device,
)


# Load the framework structure
framework: ase.Atoms = read(FrameworkPath)  # type: ignore

# Load the adsorbate structure
adsorbate: ase.Atoms = read(AdsorbatePath)  # type: ignore

Temperature = 298.0

NSteps = 3000

widom = Widom(
    model=model,
    framework_atoms=framework,
    adsorbate_atoms=adsorbate,
    temperature=Temperature,
    device=device,
    vdw_radii=vdw_radii,
    debug=False,
    output_to_file=True,
    random_seed=42,
    cutoff_radius=6.0,
    automatic_supercell=True,
)

widom.logger.print_header()

widom.restart()

widom.run(NSteps)
widom.logger.print_summary()
widom.save_results()

The widom.restart() method will try to read the int_energy_0.00000.npy file, which is the last saved file from the previous simulation.

On the Output.out file will be printed the message:

Restarting simulation from step 171...

===========================================================================
Restart file requested.
Loaded state with 324 total atoms.
Current total energy: -2351.243 eV
Current number of adsorbates: 0
Current average binding energy: 0.000 kJ/mol
===========================================================================

---

Restarting a GCMC Simulation

It is possible to restart a existing GCMC simulation in two ways.

First, you can restart a failed simulation from the last saved point, continuing it from where it left off. This is useful if the simulation was interrupted or if you want to extend the simulation time.

Second, you start a new simulation using the specific configuration of a previous simulation as the initial configuration for the new one.

Restarting a failed simulation

In this example we will evaluate the MFI zeolite with the TraPPE-Zeo, a transferable potential for phase equilibria force field for all-silica zeolites [1].

On the flames/examples/Basic/4-Restart_fail folder you will find the cif file for the MFI [2].

Note

Since this model is a united-atom model, the hydrogen is not explicitly included in the structure, and are save on the Trajectory.traj file only for visualization purposes. Because of that, there is no point in using rotation moves in this simulation, so we will set the rotation move weight to zero.

The script below will run the simulation at 298 K (25°C) and 1 bar.

import json

import ase
from ase.data import vdw_radii
from ase.io import read

from flames.gcmc import GCMC
from flames.lennard_jones import CustomLennardJones
from flames.utilities import read_cif

with open("TraPPE_zeo.json", "r") as f:
    lj_params = json.loads(f.read())

calc = CustomLennardJones(lj_params, vdw_cutoff=12.8, shifted=True)

# Load the framework structure
framework: ase.Atoms = read_cif("MFI.cif")

# Load the adsorbate structure
adsorbate: ase.Atoms = read("ch4.xyz")

gcmc = GCMC(
    model=calc,
    framework_atoms=framework,
    adsorbate_atoms=adsorbate,
    temperature=298.15,
    pressure=1e5,
    device="cpu",
    vdw_radii=vdw_radii,
    debug=False,
    output_to_file=True,
    cutoff_radius=12.0,
    automatic_supercell=True,
    LLM=False,
    move_weights={
        "insertion": 0.25,
        "deletion": 0.25,
        "translation": 0.25,
        "rotation": 0,
        "reinsertion": 0.25,
    },
)

gcmc.logger.print_header()

gcmc.restart()

gcmc.run(10000 - gcmc.base_iteration)
gcmc.logger.print_summary()

gcmc.save_results()

---

Breaking down the input script

The gcmc.restart() method will try to read the total_ads_100000.00000.npy, total_energy_100000.00000.npy, uptake_100000.00000.npy, and Trajectory_100000.00000.traj files, which are the last saved files from the previous simulation. The gcmc.base_iteration variable will be set to the last iteration of the previous simulation, and the new simulation will continue from that point, ensuring that you obtain the desired number of steps.

In the Output.out file will be printed the message:

Restarting simulation from step 2100...

===========================================================================
Restart file requested.
Loaded state with 2304 total atoms.
Current total energy: 33166.818 eV
Current number of adsorbates: 0
Current average binding energy: 0.000 kJ/mol
===========================================================================


===========================================================================

Restarting GCMC simulation from previous configuration...

Loaded state with 2414 total atoms.

Current total energy: 33163.019 eV
Current number of adsorbates: 22
Current average binding energy: -16.661 kJ/mol

Current steps are: 2100

===========================================================================

If these files are not found, the code will print a warning message and will start a new simulation from scratch.

Restarting simulation from step 0...

===========================================================================
Restart file requested.
Loaded state with 2304 total atoms.
Current total energy: 33166.818 eV
Current number of adsorbates: 0
Current average binding energy: 0.000 kJ/mol
===========================================================================

============================================================================
WARNING: Error occurred while loading trajectory file:
This is not an ULM formatted file.
Cannot load the last state of the simulation.
This is likely due to empty or corrupted trajectory file.
Simulation will start from scratch.
============================================================================

Restarting from a previous configuration

On the flames/examples/Basic/5-Restart_state folder you will find the Trajectory.traj file from a previous simulation. You can use this file to restart a new simulation from the last configuration of the previous one.

The script below will read the last configuration from the Trajectory.traj file and will use it as the initial configuration for a new simulation.

import json

import ase
from ase.data import vdw_radii
from ase.io import read

from flames.gcmc import GCMC
from flames.lennard_jones import CustomLennardJones
from flames.utilities import read_cif

with open("TraPPE_zeo.json", "r") as f:
    lj_params = json.loads(f.read())

calc = CustomLennardJones(lj_params, vdw_cutoff=12.8, shifted=True)

# Load the framework structure
framework: ase.Atoms = read_cif("MFI.cif")

# Load the adsorbate structure
adsorbate: ase.Atoms = read("ch4.xyz")

gcmc = GCMC(
    model=calc,
    framework_atoms=framework,
    adsorbate_atoms=adsorbate,
    temperature=298.15,
    pressure=1e5,
    device="cpu",
    vdw_radii=vdw_radii,
    debug=False,
    output_to_file=True,
    cutoff_radius=12.0,
    automatic_supercell=True,
    LLM=False,
    move_weights={
        "insertion": 0.25,
        "deletion": 0.25,
        "translation": 0.25,
        "rotation": 0,
        "reinsertion": 0.25,
    },
)

gcmc.logger.print_header()

gcmc.load_state('Trajectory.traj')

gcmc.run(10000)
gcmc.logger.print_summary()

gcmc.save_results()

Note that the gcmc.load_state(‘Trajectory.traj’) method will read the last configuration from the Trajectory.traj file and will use it as the initial configuration for the new simulation. You are starting a new simulation, so the gcmc.base_iteration variable will be set to zero, and the new simulation will run for the specified number of steps. The state that you are loading does not need to be generated in the same conditions of the new simulation that you are running, but it is recommended that the conditions are similar to ensure a smooth transition between the two simulations. You can use this to start the simulation from a configuration that is closer to the equilibrium state, which can help to reduce the equilibration time of the new simulation. Or to start the simulation from the last state of the previous point in a isotherm simulation, which can help to reduce the equilibration time of the new simulation and ensure a smooth transition between the two points.

On the Output.out file will be printed the message:

===========================================================================

Restarting GCMC simulation from previous configuration...

Loaded state with 2394 total atoms.

Current total energy: 33163.932 eV
Current number of adsorbates: 18
Current average binding energy: -15.467 kJ/mol

Current steps are: 0

===========================================================================

Always check if that is what you want.

Note

The state file don’t necessarily need to be a Trajectory.traj file, it can be any ASE readable file format, such as xyz, cif, pdb, etc.

References

[1]

Bai, P., Tsapatsis, M. and Siepmann, J.I., 2013. TraPPE-zeo: Transferable potentials for phase equilibria force field for all-silica zeolites. The Journal of Physical Chemistry C, 117(46), pp.24375-24387. https://pubs.acs.org/doi/10.1021/jp4074224

[2]

Van Koningsveld, H., Van Bekkum, H. and Jansen, J.C., 1987. On the location and disorder of the tetrapropylammonium (TPA) ion in zeolite ZSM-5 with improved framework accuracy. Structural Science, 43(2), pp.127-132. https://journals.iucr.org/paper?S0108768187098173