def __init__(self, atoms,
temperature=100 * kB,
optimizer=FIRE,
fmax=0.1,
dr=0.1,
logfile='-',
trajectory='lowest.traj',
optimizer_logfile='-',
local_minima_trajectory='local_minima.traj',
adjust_cm=True):
Dynamics.__init__(self, atoms, logfile, trajectory)
self.kT = temperature
self.optimizer = optimizer
self.fmax = fmax
self.dr = dr
if adjust_cm:
self.cm = atoms.get_center_of_mass()
else:
self.cm = None
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
self.lm_trajectory = PickleTrajectory(local_minima_trajectory,
'w', atoms)
self.initialize()
def __init__(self, atoms, max_step = 0.5, T= 298.0, accept_ratio=None, one_d_coord =None):
self.max_step = max_step
self.beta = 1.0 / (T * units.kB)
self.accept_ratio = accept_ratio
self.one_d_coord = one_d_coord
Dynamics.__init__(self, atoms,logfile=None, trajectory=None)
def __init__(self,
atoms,
timestep,
integrator='verlet',
trajectory=None,
traj_interval=1000,
logfile=None,
loginterval=100):
Dynamics.__init__(self, atoms, None, None)
self.dt = timestep
if integrator == 'verlet':
self.run_style = 'verlet'
else:
raise RuntimeError('Unknown integrator: %s' % thermostat)
if trajectory:
if isinstance(trajectory, str):
trajectory = PickleTrajectory(trajectory, 'w', atoms)
self.attach(trajectory, interval=traj_interval)
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
self.fix = None
self.cell_relaxed = False
def __init__(self, atoms, timestep, trajectory, logfile=None,
loginterval=1, append_trajectory=False):
# dt as to be attached _before_ parent class is initialized
self.dt = timestep
Dynamics.__init__(self, atoms, logfile=None, trajectory=trajectory,
append_trajectory=append_trajectory)
self.masses = self.atoms.get_masses()
self.max_steps = None
if 0 in self.masses:
warnings.warn('Zero mass encountered in atoms; this will '
'likely lead to errors if the massless atoms '
'are unconstrained.')
self.masses.shape = (-1, 1)
if not self.atoms.has('momenta'):
self.atoms.set_momenta(np.zeros([len(self.atoms), 3]))
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
def __init__(self, atoms, timestep, trajectory, logfile=None,
loginterval=1):
Dynamics.__init__(self, atoms, logfile=None, trajectory=trajectory)
self.dt = timestep
# Store data locally except on parallel simulations
self._localdata = not getattr(atoms, "parallel", False)
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
def __init__(self,
atoms,
restart=None,
logfile=None,
trajectory=None,
algorithm='cg',
relax_cell=False):
Dynamics.__init__(self, atoms, logfile, trajectory)
self.algorithm = algorithm
self.relax_cell = relax_cell
def __init__(self,
atoms,
timestep,
trajectory,
logfile=None,
loginterval=1):
Dynamics.__init__(self, atoms, logfile=None, trajectory=trajectory)
self.dt = timestep
self.masses = self.atoms.get_masses()
self.masses.shape = (-1, 1)
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
def __init__(self, atoms, timestep, trajectory, logfile=None,
loginterval=1):
Dynamics.__init__(self, atoms, logfile=None, trajectory=trajectory)
self.dt = timestep
self.masses = self.atoms.get_masses()
if 0 in self.masses:
warnings.warn('Zero mass encountered in atoms; this will '
'likely lead to errors if the massless atoms '
'are unconstrained.')
self.masses.shape = (-1, 1)
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
def __init__(self, atoms, timestep, trajectory, logfile=None,
loginterval=1):
Dynamics.__init__(self, atoms, logfile=None, trajectory=trajectory)
self.dt = timestep
self.masses = self.atoms.get_masses()
if 0 in self.masses:
warnings.warn('Zero mass encountered in atoms; this will '
'likely lead to errors if the massless atoms '
'are unconstrained.')
self.masses.shape = (-1, 1)
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
def __init__(self, atoms,
temperature=100 * kB,
optimizer=SDLBFGS,
fmax=0.1,
dr=0.1,
logfile='-',
trajectory=None,
optimizer_logfile='-',
local_minima_trajectory='local_minima.con',
adjust_cm=True,
mss=0.2,
minenergy=None,
distribution='uniform',
adjust_step_size=None,
adjust_every = None,
target_ratio = 0.5,
adjust_fraction = 0.05,
significant_structure = False,
pushapart = 0.4,
jumpmax=None
):
Dynamics.__init__(self, atoms, logfile, trajectory)
self.kT = temperature
self.optimizer = optimizer
self.fmax = fmax
self.dr = dr
if adjust_cm:
self.cm = atoms.get_center_of_mass()
else:
self.cm = None
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
tsase.io.write_con(self.lm_trajectory,atoms,w='w')
tsase.io.write_con('best_energies.con',atoms,w='w')
self.minenergy = minenergy
self.distribution = distribution
self.adjust_step = adjust_step_size
self.adjust_every = adjust_every
self.target_ratio = target_ratio
self.adjust_fraction = adjust_fraction
self.significant_structure = True
self.pushapart = pushapart
self.jumpmax = jumpmax
self.mss = mss
self.initialize()
self.norm_dists = [None] * len(self.positions)
for i in range(len(self.norm_dists)):
self.norm_dists[i] = []
def __init__(self, atoms,
temperature=100 * kB,
optimizer=SDLBFGS,
fmax=0.1,
dr=0.1,
logfile='-',
trajectory=None,
optimizer_logfile='-',
local_minima_trajectory='local_minima.con',
mss=0.2,
minenergy=None,
adjust_step_size=None,
adjust_every = None,
target_ratio = 0.5,
adjust_fraction = 0.05,
pushapart = 0.4,
jumpmax=2,
inertia_weight=.4,
method='PSO'
):
Dynamics.__init__(self, atoms, logfile, trajectory)
self.local_optimizations = 0
self.bcms = 0
self.kT = temperature
self.optimizer = optimizer
self.fmax = fmax
self.dr = dr
self.method = method
self.cm = atoms.get_center_of_mass()
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
self.minenergy = minenergy
self.energy = 0
self.adjust_step = adjust_step_size
self.adjust_every = adjust_every
self.target_ratio = target_ratio
self.adjust_fraction = adjust_fraction
self.pushapart = pushapart
self.jumpmax = jumpmax
self.mss = mss
self.inertia_weight = inertia_weight
self.velocity = 0.0 * self.atoms.get_positions()
self.initialize()
def __init__(self,
atoms,
temperature=100 * kB,
optimizer=FIRE,
fmax=0.1,
dr=0.1,
logfile='-',
trajectory='lowest.traj',
optimizer_logfile='-',
local_minima_trajectory='local_minima.traj',
adjust_cm=True):
"""Parameters:
atoms: Atoms object
The Atoms object to operate on.
trajectory: string
Pickle file used to store trajectory of atomic movement.
logfile: file object or str
If *logfile* is a string, a file with that name will be opened.
Use '-' for stdout.
"""
self.kT = temperature
self.optimizer = optimizer
self.fmax = fmax
self.dr = dr
if adjust_cm:
self.cm = atoms.get_center_of_mass()
else:
self.cm = None
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
self.db = connect(local_minima_trajectory[:-4] + 'db')
self.lm_trajectory = Trajectory(local_minima_trajectory, 'a',
atoms)
Dynamics.__init__(self, atoms, logfile, trajectory)
self.initialize()
def __init__(self, atoms,
temperature=100 * kB,
optimizer=FIRE,
fmax=0.1,
dr=0.1,
logfile='-',
trajectory='lowest.traj',
optimizer_logfile='-',
local_minima_trajectory='local_minima.traj',
adjust_cm=True):
"""Parameters:
atoms: Atoms object
The Atoms object to operate on.
trajectory: string
Pickle file used to store trajectory of atomic movement.
logfile: file object or str
If *logfile* is a string, a file with that name will be opened.
Use '-' for stdout.
"""
self.kT = temperature
self.optimizer = optimizer
self.fmax = fmax
self.dr = dr
if adjust_cm:
self.cm = atoms.get_center_of_mass()
else:
self.cm = None
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
self.lm_trajectory = Trajectory(local_minima_trajectory,
'w', atoms)
Dynamics.__init__(self, atoms, logfile, trajectory)
self.initialize()
def __init__(self, atoms, timestep, integrator='verlet', trajectory=None,
traj_interval=1000, logfile=None, loginterval=100):
Dynamics.__init__(self, atoms, None, None)
self.dt = timestep
if integrator == 'verlet':
self.run_style = 'verlet'
else:
raise RuntimeError('Unknown integrator: %s' % thermostat)
if trajectory:
if isinstance(trajectory, str):
trajectory = PickleTrajectory(trajectory, 'w', atoms)
self.attach(trajectory, interval=traj_interval)
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
self.fix = None
self.cell_relaxed = False
def __init__(self,
atoms,
timestep,
trajectory,
logfile=None,
loginterval=1,
append_trajectory=False):
# dt as to be attached _before_ parent class is initialized
self.dt = timestep
Dynamics.__init__(self, atoms, logfile=None, trajectory=None)
self.masses = self.atoms.get_masses()
self.max_steps = None
if 0 in self.masses:
warnings.warn('Zero mass encountered in atoms; this will '
'likely lead to errors if the massless atoms '
'are unconstrained.')
self.masses.shape = (-1, 1)
if not self.atoms.has('momenta'):
self.atoms.set_momenta(np.zeros([len(self.atoms), 3]))
# Trajectory is attached here instead of in Dynamics.__init__
# to respect the loginterval argument.
if trajectory is not None:
if isinstance(trajectory, str):
mode = "a" if append_trajectory else "w"
trajectory = Trajectory(trajectory, mode=mode, atoms=atoms)
self.attach(trajectory, interval=loginterval)
if logfile:
self.attach(MDLogger(dyn=self, atoms=atoms, logfile=logfile),
interval=loginterval)
def __init__(self,
# General GO parameters
atoms, # ASE atoms object defining the PES
temperature = 500, # K, initial temperature
optimizer = SDLBFGS, # local optimizer
fmax = 0.01, # magnitude of the L2 norm used as the convergence criteria for local optimization
adjust_cm = True, # fix the center of mass (True or False)
mss = 0.1, # maximum step size for the local optimizer
minenergy = None, # the GO algorithm stops when a configuration is found with a lower potential energy than this value
pushapart = 0.1, # push atoms apart until all atoms are no closer than this distance
keep_minima_arrays = True, # create global_minima and local_minima arrays of length maximum number of Monte Carlo steps (True or False)
minima_threshold = 2, # round potential energies to how many decimal places
# Logging parameters
logfile = '-',
trajectory = None,
optimizer_logfile = None,
#local_minima_trajectory = 'local_minima.con',
local_minima_trajectory = None,
# Selecting move type
move_type = True, # True = basin hopping trial move; False = minima hopping
distribution = 'uniform', # The distribution to use in trial move. Make "molecular_dynamics" the distribution for MH trial move
# Random move parameters
dr = 0.45, # maximum displacement in each degree of freedom for Monte Carlo trial moves
adjust_step_size = None, # adjust dr after this many Monte Carlo steps (Default: None; does not adjust dr)
target_ratio = 0.5, # specified ratio of Monte Carlo steps
adjust_fraction = 0.05, # fraction by which to adjust dr by in order to meet target_ratio
significant_structure = True, # displace from minimum at each move (True or False)
# Dynamic move parameters (Molecular Dynamics)
timestep = 0.1, # fs, molecular dynamics time step
mdmin = 2, # number of minima to pass in MD before stopping
dimer_method = True, # uses an iterative dimer method before molecular dynamics (True or False)
dimer_a = 0.001, # scalar for forces in optimization
dimer_d = 0.01, # distance between two images in the dimer
dimer_steps = 14, # number of dimer iterations
# Occasional jumping parameters
jump_distribution = 'uniform', # The distribution to use in OJ move. Same options as distribution flag
jumpmax = None, # number of previously rejected MC steps before taking an OJ move; None = no OJ move
jmp = 7, # number of consecutive accepted moves taken in OJ
global_jump = None, # number of times to visit the same PE before we take an OJ; None = no global jump
global_reset = False, # True = reset all of the history counts after a jump (basically delete the history and start fresh)
# Select acceptance criteria
acceptance_criteria = True, # BH = Tue; MH = False
# BH acceptance parameters
accept_temp = 8000, # K; separate temperature to use for BH acceptance (None: use temperature parameter instead)
adjust_temp = False, # dynamically adjust the temperature in BH acceptance (True or False)
history_weight = 0.0, # the weight factor of BH history >= 0 (0.0: no history comparison in BH acceptance)
history_num = 0, # limit of previously accepted minima to keep track of for BH history (set to 0 to keep track of all minima)
# MH acceptance criteria
beta1 = 1.04, # temperature adjustment parameter
beta2 = 1.04, # temperature adjustment parameter
beta3 = 1.0/1.04, # temperature adjustment parameter
Ediff0 = 0.5, # eV, initial energy acceptance threshold
alpha1 = 0.98, # energy threshold adjustment parameter
alpha2 = 1./0.98, # energy threshold adjustment parameter
minimaHopping_history = False, # use history in MH trial move (True or False)
# Geometry comparison parameters
use_geometry = True, # True = compare geometry of systems when they have the same PE when determining if they are the same atoms configuration
eps_r = 0.1, # positional difference to consider atoms in the same location
use_get_mapping = True, # from atoms_operator.py use get_mapping if true or rot_match if false to compare geometry
neighbor_cutoff = 1.2, # parameter for get_mapping only
):
Dynamics.__init__(self, atoms, logfile, trajectory)
self.temperature = temperature
self.optimizer = optimizer
self.fmax = fmax
self.dr = dr
if adjust_cm:
self.cm = atoms.get_center_of_mass()
else:
self.cm = None
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
tsase.io.write_con(self.lm_trajectory,atoms,w='w')
self.minenergy = minenergy
self.move_type = move_type
self.distribution = distribution
self.adjust_step = adjust_step_size
self.target_ratio = target_ratio
self.adjust_fraction = adjust_fraction
self.significant_structure = significant_structure
self.pushapart = pushapart
self.jumpmax = jumpmax
self.jmp = jmp
self.jump_distribution = jump_distribution
self.global_jump = global_jump
self.global_reset = global_reset
self.dimer_method = dimer_method
self.dimer_a = dimer_a
self.dimer_d = dimer_d
self.dimer_steps = dimer_steps
self.timestep = timestep
self.mdmin = mdmin
self.w = history_weight
self.history_num = history_num
self.adjust_temp = adjust_temp
self.accept_temp = accept_temp
self.accept_criteria = acceptance_criteria
self.mh_history = minimaHopping_history
self.beta1 = beta1
self.beta2 = beta2
self.beta3 = beta3
self.Ediff = Ediff0
self.alpha1 = alpha1
self.alpha2 = alpha2
self.minima_threshold = minima_threshold
self.use_geo = use_geometry
self.eps_r = eps_r
self.use_get_mapping = use_get_mapping
self.neighbor_cutoff = neighbor_cutoff
self.keep_minima_arrays = keep_minima_arrays
self.global_minima = [] # an array of the current global minimum for every MC step
self.local_minima = [] # an array of the current local minimum for every MC step
self.allTemps = []
self.saveEdiff = []
self.trial_loop = []
self.num_localops = 0
self.num_geo_compare = []
# when a MD sim. has passed a local minimum:
self.passedminimum = PassedMinimum()
self.mss = mss
# for PE comparision
# dictionary for found local minima
# keys will be the potential energy rounded to self.minima_threshold digits left of the decimal
# values will be number of times the potential energy has been visited
self.minima = {}
# list with fixed size that will store history_num previous minima
self.temp_minima = [0] * self.history_num
self.temp_positions = [0] * self.history_num
# for geometry comparison
self.positionsMatrix = []
# count of unique geometries we have accepted so far
self.numPositions = 0
# dictionary for geometry comparison
# keys = approximate potential energy
# values = list of indexes in positionsMatrix with the same PE
self.geometries = {}
# dictionary with keys = index in positionsMatrix
# values = # accepted visits
self.geo_history = {}
self.current_index = None
self.last_index = None
# number of Monte Carlo moves that have been accepted in the run
self.num_accepted_moves = 0.0
self.initialize()
def __init__(self,
atoms,
maxstep=0.5,
parallel_drift=0.2,
energy_target=None,
angle_limit=None,
force_parallel_step_scale=None,
remove_translation=False,
use_FS=True,
initialize_old=True,
initialization_step_scale=1e-2,
use_target_shift=True,
target_shift_previous_steps=10,
seed=19460926,
verbose=False,
trajectory=None,
logfile=None,
use_tangent_curvature=False,
force_consistent=None,
append_trajectory=False,
loginterval=1):
'''Perpendicular drift breaks orbits like dimer form, so they spin on new axes'''
if force_parallel_step_scale is None:
# a hureistic guess since most systems will overshoot when there is drift
FPSS = 0.9 + 0.5 * parallel_drift
else:
FPSS = force_parallel_step_scale
self.verbose = verbose
self.seed = seed
self.remove_translation = remove_translation
self.use_FS = use_FS
self.force_consistent = force_consistent
self.kappa = 0.0 # initializing so logging can work
self.use_tangent_curvature = use_tangent_curvature
#### for FS taylor expansion
self.T = None
self.Told = None
self.N = None
self.Nold = None
self.r_old = None
self.r = None
###### initialize rng
self.rng = default_rng(seed)
#######
if energy_target is None:
self.energy_target = atoms.get_potential_energy(
force_consistent=self.force_consistent)
else:
self.energy_target = energy_target
# these need to be initialized before the initialize_old step so it doesn't crash
self.previous_energies = np.zeros(target_shift_previous_steps)
## initizing the previous steps at the target energy slows
## target shifting untill the system has had
## 'target_shift_previous_steps' steps to equilibrate
## this should prevent occilations
self.previous_energies.fill(self.energy_target)
# we need velocities or this won't run and will produce NaNs,
# if none are provided we make random ones
p = atoms.get_momenta()
masses = atoms.get_masses()[:, np.newaxis]
v = p / masses
from ase import units
if np.linalg.norm(v) < 1e-6:
# we have to pass dimension since atoms are not yet stored
v = self.rand_vect((len(atoms), 3))
atoms.set_momenta(v / masses)
print("No Velocities found, random velocities applied")
if initialize_old:
#self.step_size = maxstep*initialization_step_scale
self.maxstep = maxstep * initialization_step_scale
self.parallel_drift = 0.0
## should force_parallel_step_scale be 0.0 for a better initial curvature?
## Or at least smaller than 1.0? doesn't seem to matter much
self.force_parallel_step_scale = 1.0
self.use_target_shift = False
self.atoms = atoms
self.step()
#self.step_size = maxstep # this interfers with logging
self.maxstep = maxstep
self.angle_limit = angle_limit
self.parallel_drift = parallel_drift
self.force_parallel_step_scale = FPSS
self.use_target_shift = use_target_shift
## loginterval exists for the MolecularDynamics class but not the more
## general Dynamics class
Dynamics.__init__(
self,
atoms,
logfile,
trajectory, #loginterval,
append_trajectory=append_trajectory,
)
def __init__(self, atoms, timestep, trajectory, logfile=None,
loginterval=1, append_trajectory=False):
"""Molecular Dynamics object.
Parameters:
atoms: Atoms object
The Atoms object to operate on.
timestep: float
The time step in ASE time units.
trajectory: Trajectory object or str
Attach trajectory object. If *trajectory* is a string a
Trajectory will be constructed. Use *None* for no
trajectory.
logfile: file object or str (optional)
If *logfile* is a string, a file with that name will be opened.
Use '-' for stdout.
loginterval: int (optional)
Only write a log line for every *loginterval* time steps.
Default: 1
append_trajectory: boolean (optional)
Defaults to False, which causes the trajectory file to be
overwriten each time the dynamics is restarted from scratch.
If True, the new structures are appended to the trajectory
file instead.
"""
# dt as to be attached _before_ parent class is initialized
self.dt = timestep
Dynamics.__init__(self, atoms, logfile=None, trajectory=None)
self.masses = self.atoms.get_masses()
self.max_steps = None
if 0 in self.masses:
warnings.warn('Zero mass encountered in atoms; this will '
'likely lead to errors if the massless atoms '
'are unconstrained.')
self.masses.shape = (-1, 1)
if not self.atoms.has('momenta'):
self.atoms.set_momenta(np.zeros([len(self.atoms), 3]))
# Trajectory is attached here instead of in Dynamics.__init__
# to respect the loginterval argument.
try:
if trajectory is not None:
if isinstance(trajectory, str):
mode = "a" if append_trajectory else "w"
trajectory = self.ensureclose(
Trajectory(trajectory, mode=mode, atoms=atoms)
)
self.attach(trajectory, interval=loginterval)
if logfile:
logger = self.ensureclose(
MDLogger(dyn=self, atoms=atoms, logfile=logfile))
self.attach(logger, loginterval)
except BaseException:
self._closefiles()
raise
def __init__(self,
atoms,
maxstep=0.5,
parallel_drift=0.1,
energy_target=None,
angle_limit=20,
potentiostat_step_scale=None,
remove_translation=False,
use_frenet_serret=True,
initialization_step_scale=1e-2,
use_target_shift=True,
target_shift_previous_steps=10,
use_tangent_curvature=False,
rng=np.random,
force_consistent=None,
trajectory=None,
logfile=None,
append_trajectory=False,
loginterval=1):
"""Contour Exploration object.
Parameters:
atoms: Atoms object
The Atoms object to operate on. Atomic velocities are required for
the method. If the atoms object does not contain velocities,
random ones will be applied.
maxstep: float
Used to set the maximum distance an atom can move per
iteration (default value is 0.5 Å).
parallel_drift: float
The fraction of the update step that is parallel to the contour but
in a random direction. Used to break symmetries.
energy_target: float
The total system potential energy for that the potentiostat attepts
to maintain. (defaults the initial potential energy)
angle_limit: float or None
Limits the stepsize to a maximum change of direction angle using the
curvature. Gives a scale-free means of tuning the stepsize on the
fly. Typically less than 30 degrees gives reasonable results but
lower angle limits result in higher potentiostatic accuracy. Units
of degrees. (default 20°)
potentiostat_step_scale: float or None
Scales the size of the potentiostat step. The potentiostat step is
determined by linear extrapolation from the current potential energy
to the target_energy with the current forces. A
potentiostat_step_scale > 1.0 overcorrects and < 1.0
undercorrects. By default, a simple heuristic is used to selected
the valued based on the parallel_drift. (default None)
remove_translation: boolean
When True, the net momentum is removed at each step. Improves
potentiostatic accuracy slightly for bulk systems but should not be
used with constraints. (default False)
use_frenet_serret: Bool
Controls whether or not the Taylor expansion of the Frenet-Serret
formulas for curved path extrapolation are used. Required for using
angle_limit based step scalling. (default True)
initialization_step_scale: float
Controls the scale of the initial step as a multiple of maxstep.
(default 1e-2)
use_target_shift: boolean
Enables shifting of the potentiostat target to compensate for
systematic undercorrection or overcorrection by the potentiostat.
Uses the average of the *target_shift_previous_steps* to prevent
coupled occilations. (default True)
target_shift_previous_steps: int
The number of pevious steps to average when using use_target_shift.
(default 10)
use_tangent_curvature: boolean
Use the velocity unit tangent rather than the contour normals from
forces to compute the curvature. Usually not as accurate.
(default False)
rng: a random number generator
Lets users control the random number generator for the
parallel_drift vector. (default numpy.random)
force_consistent: boolean
(default None)
trajectory: Trajectory object or str (optional)
Attach trajectory object. If *trajectory* is a string a
Trajectory will be constructed. Default: None.
logfile: file object or str (optional)
If *logfile* is a string, a file with that name will be opened.
Use '-' for stdout. Default: None.
loginterval: int (optional)
Only write a log line for every *loginterval* time steps.
Default: 1
append_trajectory: boolean
Defaults to False, which causes the trajectory file to be
overwriten each time the dynamics is restarted from scratch.
If True, the new structures are appended to the trajectory
file instead.
"""
if potentiostat_step_scale is None:
# a heuristic guess since most systems will overshoot when there is
# drift
self.potentiostat_step_scale = 1.1 + 0.6 * parallel_drift
else:
self.potentiostat_step_scale = potentiostat_step_scale
self.rng = rng
self.remove_translation = remove_translation
self.use_frenet_serret = use_frenet_serret
self.force_consistent = force_consistent
self.use_tangent_curvature = use_tangent_curvature
self.initialization_step_scale = initialization_step_scale
self.maxstep = maxstep
self.angle_limit = angle_limit
self.parallel_drift = parallel_drift
self.use_target_shift = use_target_shift
# These will be populated once self.step() is called, but could be set
# after instantiating with ce = ContourExploration(...) like so:
# ce.Nold = Nold
# ce.r_old = atoms_old.get_positions()
# ce.Told = Told
# to resume a previous contour trajectory.
self.T = None
self.Told = None
self.N = None
self.Nold = None
self.r_old = None
self.r = None
if energy_target is None:
self.energy_target = atoms.get_potential_energy(
force_consistent=self.force_consistent)
else:
self.energy_target = energy_target
# Initizing the previous steps at the target energy slows
# target_shifting untill the system has had
# 'target_shift_previous_steps' steps to equilibrate and should prevent
# occilations. These need to be initialized before the initialize_old
# step to prevent a crash
self.previous_energies = np.full(target_shift_previous_steps,
self.energy_target)
# these first two are purely for logging,
# auto scaling will still occur
# and curvature will still be found if use_frenet_serret == True
self.step_size = 0.0
self.curvature = 0
# loginterval exists for the MolecularDynamics class but not for
# the more general Dynamics class
Dynamics.__init__(
self,
atoms,
logfile,
trajectory, # loginterval,
append_trajectory=append_trajectory,
)
# we need velocities or NaNs will be produced,
# if none are provided we make random ones
velocities = self.atoms.get_velocities()
if np.linalg.norm(velocities) < 1e-6:
# we have to pass dimension since atoms are not yet stored
atoms.set_velocities(self.rand_vect())
def __init__(
self,
atoms,
temperature=1500.0,
maximum_temp=None,
minimum_temp=None,
stop_steps=400,
logfile='grandcanonical.log',
trajectory='grandcanonical.traj',
local_minima_trajectory='local_minima.traj',
local_minima_trajecotry_db="local_minima.db",
adjust_cm=False,
restart=False,
chemical_potential=None,
bash_script="optimize.sh",
files_to_copied=None,
):
"""Parameters:
atoms: Atoms object
The Atoms object to operate on.
trajectory: string
Pickle file used to store trajectory of atomic movement.
logfile: file object or str
If *logfile* is a string, a file with that name will be opened.
Use '-' for stdout.
"""
self.T = temperature
if maximum_temp is None:
self.max_T = 1.0 / ((1.0 / self.T) / 1.5)
else:
self.max_T = max([maximum_temp, self.T])
if minimum_temp is None:
self.min_T = 1.0 / ((1.0 / self.T) * 1.5)
else:
self.min_T = min([minimum_temp, self.T])
self.stop_steps = stop_steps
self.restart = restart
self.bash_script = bash_script
self.copied_files = files_to_copied
# some file names and folders are hardcoded
self.fn_current_atoms = "Current_atoms.traj"
self.fn_status_file = "Current_Status.json"
self.opt_folder = "opt_folder"
self.structure_modifiers = {}
self.adjust_cm = adjust_cm
if 0:
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
self.lm_trajectory = Trajectory(local_minima_trajectory, 'a',
atoms)
self.lm_trajectory = connect(local_minima_trajecotry_db)
# Dynamics.__init__ simply set
# self.atoms and
# self.logfile and
# self.trajectory and
# self.nsteps
Dynamics.__init__(self, atoms, logfile, trajectory)
# print the program logo at the beginning of the output file
self.logfile.write("%s\n" % programlogo)
self.logfile.flush()
# setup the chemical potential for different elements
self.mu = {}
if chemical_potential is not None and os.path.isfile(
chemical_potential):
with open(chemical_potential, "r") as fp:
for i, istr in enumerate(fp):
if istr.strip() == "":
continue
k, v = istr.split()
self.mu[k] = float(v)
else:
raise RuntimeError("chemical potential file %s is not found" %
chemical_potential)
for k, v in self.mu.items():
self.dumplog("Chemical potential of %s is %.3f" % (k, v))
# try to read previous result
if self.restart:
if (not os.path.isfile(self.fn_status_file)) or (
not os.path.isfile(self.fn_current_atoms)):
self.dumplog("%s or %s no found, start from scratch\n" %
(self.fn_current_atoms, self.fn_status_file))
self.restart = False
elif os.path.getsize(self.fn_current_atoms) == 0:
self.dumplog("{} is empty, set self.restart=False".format(
self.fn_current_atoms))
self.restart = False
else:
try:
atoms = read(self.fn_current_atoms)
atoms.get_potential_energy()
except PropertyNotImplementedError:
self.dumplog(
"No energy found in {}, set self.restart=False".format(
self.fn_current_atoms))
self.restart = False
except RuntimeError as e:
self.dumplog(
"Error when read {}, set self.restart=False".format(e))
self.restart = False
self.energy = None
self.free_energy = None
self.energy_min = None
self.free_energy_min = None
self.no_improvement_step = 0
# negative value indicates no on-going structure optimization, otherwise it will be the on-going optimization
self.on_optimization = -1
# this is used for adjusting the temperature of Metropolis algorithm
self.accept_history = [
] # a series of 0 and 1, 0 stands for not accpeted, 1 stands for accepted
self.max_history = 25 # max length of self.accept_history is 25
if not self.restart:
self.initialize()
else:
self.reload_previous_results()
def __init__(self, atoms, restart=None, logfile=None, trajectory=None, algorithm='cg', relax_cell=False): Dynamics.__init__(self, atoms, logfile, trajectory) self.algorithm = algorithm self.relax_cell = relax_cell
def __init__(
self,
atoms,
temperature=100 * kB,
optimizer=FIRE,
optimizer2=FIRE, #which optimizer to use to fmax_mult*fmax
fmax_mult=15, #at what multiple of fmax do you want to use second optimizer
fmax=0.1,
dr=0.1,
logfile='-',
trajectory='lowest.traj',
optimizer_logfile='stdout.log',
local_minima_trajectory='temp_local_minima.traj', #local minima found
adjust_cm=True,
movemode=0, #Pick a way for displacement. 0->random cartesian, 1->delocalized internals, 2->Periodic
maxmoves=1000, #Should prevent an issue, where you get stuck in a structure, for which get_energy() fails
numdelocmodes=1, #should a LC of modes be applied for the displacement? How many should be combined?
adsorbmask=None, #mask that specifies where the adsorbate is located in the atoms object (list of lowest and highest pos)
cell_scale=[1.0, 1.0, 1.0], #used for translation in adsorbates
constrain=False): #constrain stretches?
Dynamics.__init__(self, atoms, logfile, trajectory)
if adsorbmask is None:
self.adsorbate = (0, len(atoms))
self.ads = False
else:
self.adsorbate = adsorbmask
self.ads = True
self.fmax_mult = fmax_mult
self.cell_scale = cell_scale
self.kT = temperature
if numdelocmodes < 1:
if self.ads:
self.numdelmodes = int(
np.round(numdelocmodes *
len(atoms[self.adsorbate[0]:self.adsorbate[1]]) *
3)) #3N dis in adsorbates
else:
self.numdelmodes = int(
np.round(numdelocmodes *
(len(atoms) * 3 - 6))) #3N-6 dis in gas phase
else:
self.numdelmodes = int(np.round(
numdelocmodes)) #round and int just for you trolls out there
self.optimizer = optimizer
self.optimizer2 = optimizer2
self.fmax = fmax
self.mm = maxmoves
self.dr = dr
self.lowstep = 0
self.movemode = movemode
self.movename = 'Random Cartesian'
self.minima = []
self.constr = constrain
if movemode == 1:
self.movename = 'Delocalized Internals, using %i modes' % self.numdelmodes
elif movemode == 2:
self.movename = 'Periodic DI'
if adjust_cm:
self.cm = atoms.get_center_of_mass()
else:
self.cm = None
self.lmfile = local_minima_trajectory
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
self.lm_trajectory = Trajectory(local_minima_trajectory, 'a',
atoms)
self.startT = datetime.now()
self.log(msg='STARTING BASINHOPPING at ' +
self.startT.strftime('%Y-%m-%d %H:%M:%S') +
':\n Displacements: ' + self.movename +
' Stepsize: %.3f fmax: %.3f T: %4.2f\n' %
(self.dr, self.fmax, self.kT / kB))
self.positions = 0.0 * self.atoms.get_positions()
self.Emin = self.get_energy(self.atoms.get_positions()) or 1.e15
self.rmin = self.atoms.get_positions()
self.call_observers()
self.log(-1, self.Emin, self.Emin)
def __init__(
self,
atoms,
temperature=100 * kB,
optimizer=SDLBFGS,
fmax=0.1,
move_atoms=True,
dr=0.1,
swap_atoms=False,
elements_lib=None, #elements which will be swapped, i.e., elements_lib = ['Au', 'Rh']
active_ratio=1.0, #define the number of atoms moved or swapped each time
visited_configs={}, # {'state_number': [energy, atoms, repeats], ...},
comp_eps_e=1.e-4, #criterion to determine if two configurations are identtical in energy
comp_eps_r=0.02, #criterion to determine if two configurations are identical in geometry
logfile='-',
trajectory=None,
optimizer_logfile='-',
local_minima_trajectory='local_minima.con',
adjust_cm=True,
mss=0.05,
minenergy=None,
distribution='uniform',
adjust_step_size=None,
target_ratio=0.5,
adjust_fraction=0.05,
significant_structure=False, # displace from minimum if accept
# significant_structure2 = False, # displace from global minimum found so far at each move
pushapart=0.4,
jumpmax=5000):
Dynamics.__init__(self, atoms, logfile, trajectory)
self.kT = temperature
self.optimizer = optimizer
self.fmax = fmax
self.move_atoms = move_atoms
self.dr = dr
self.swap_atoms = swap_atoms
self.elements_lib = elements_lib
self.active_ratio = active_ratio
self.visited_configs = visited_configs
self.comp_eps_e = comp_eps_e
self.comp_eps_r = comp_eps_r
if adjust_cm:
self.cm = atoms.get_center_of_mass()
else:
self.cm = None
self.optimizer_logfile = optimizer_logfile
self.lm_trajectory = local_minima_trajectory
if isinstance(local_minima_trajectory, str):
tsase.io.write_con(self.lm_trajectory, atoms, w='w')
self.minenergy = minenergy
self.distribution = distribution
self.adjust_step_size = adjust_step_size
self.target_ratio = target_ratio
self.adjust_fraction = adjust_fraction
self.significant_structure = significant_structure
# self.significant_structure2 = significant_structure2
self.pushapart = pushapart
self.jumpmax = jumpmax
self.mss = mss
self.initialize()