def compute_bs():
from numpy import array
from pylada.escan import read_input, exec_input, ReducedBPoints
from pylada.vff import Vff
# reads input file.
input = read_input("input.py")
# creating unrelaxed structure.
structure = input.vff.lattice.to_structure()
structure.atoms[0].type = "Si"
structure.atoms[1].type = "Ge"
structure.scale = 5.65
# some kpoints + associated name
X = array( [1,0,0], dtype="float64" )
G = array( [0,0,0], dtype="float64" )
L = array( [0.5,0.5,0.5], dtype="float64" )
W = array( [0, 0.5,1], dtype="float64" )
# Each job is performed for a given kpoint (first argument), at a given
# reference energy (third argument). Results are stored in a specific directory
# (second arguement). The expected eigenvalues are given in the fourth argument.
input = read_input('input.py')
kescan = exec_input(repr(input.escan).replace('Escan', 'KEscan')).functional
kescan.fft_mesh = 14, 14, 14
kescan.kpoints = ReducedBPoints(density=20) + (X, G) + (G, L)
result = kescan( structure, outdir='results/projections',
nbstates = len(structure.atoms) * 4 + 4,
eref = None )
def gajobs(path, inputpath = "input.py"):
from copy import deepcopy
import IPython.ipapi
from numpy import array, all
from pylada.ga.escan.elemental.functional import Darwin as Functional
from pylada.ga.escan.elemental.evaluator import EffectiveMass as Evaluator
from pylada.ga.escan.elemental import Converter
from pylada.ga.functional import minimize
from pylada.jobs import JobFolder
from pylada.escan import read_input
# reads input file.
input = read_input(inputpath)
jobfolder = JobFolder()
for type in ["ee", "hh"]:
for direction in ["epi", "perp"]:
for scale in input.scales:
for range in input.ranges:
nmin, nmax = min(range), max(range)
for trial in input.trials:
escan = deepcopy(input.escan)
escan.vff.lattice.scale = scale
if direction == "perp": escan.direction = input.growth_direction
elif all(array(input.growth_direction) - (0,0,1) == (0,0,0)):
escan.direction = (1, 0, 0), (0, 1, 0)
elif all(array(input.growth_direction) - (1,1,0) == (0,0,0)):
escan.direction = (0, 0, 1), (-1, 1, 0)
else: raise ValueError("Unknown growth direction.")
if type == "ee": escan.type = "e"
elif type == "hh" or type == "lh": escan.type = "h"
converter = Converter(growth=input.growth_direction, lattice=escan.vff.lattice)
kwargs = { "popsize": input.population_size, "rate": input.offspring_rate,
"max_gen": input.max_generations, "pools": input.pools,
"crossover_rate": input.crossover_rate, "swap_rate": input.swap_rate,
"growth_rate": input.growth_rate, "nmin": nmin, "nmax": nmax,
"dosym": input.dosym, "rootworkdir": "$GSCRATCH", "comparison": minimize }
kwargs.update(getattr(input, 'kwargs', {}))
evaluator = Evaluator( converter = converter,
functional = escan,
**getattr(input, 'kwargs', {}) )
functional = Functional(evaluator, **kwargs)
functional.n = slice(2,4) if type == "lh" else slice(0,2)
name = "{0[0]}{0[1]}{0[2]}/{1}/{2}/"\
.format(input.growth_direction, direction, type)
name += "scale_{0:.2f}/{1}_{2}/trial_{3}".format(scale, nmin, nmax, trial)
gajob = jobfolder / name
gajob.functional = functional
ip = IPython.ipapi.get()
ip.user_ns["current_jobfolder"] = jobfolder
ip.magic("savejobs " + path)
def gajobs(path, inputpath="input.py"):
from copy import deepcopy
import IPython.ipapi
from pylada.ga.escan.elemental.functional import Darwin as Functional
from pylada.ga.escan.elemental.evaluator import Dipole as Evaluator
from pylada.ga.escan.elemental import LayeredConverter as Converter
from pylada.jobs import JobFolder
from pylada.escan import read_input
# reads input file.
input = read_input(inputpath)
jobfolder = JobFolder()
for scale in input.scales:
for p in input.periods:
for trial in input.trials:
supercell = input.supercell(p)
escan = deepcopy(input.escan)
escan.fft_mesh = input.fftmesh(supercell)
escan.vff.lattice.scale = scale
converter = Converter(supercell, lattice=escan.vff.lattice)
evaluator = Evaluator(converter=converter, escan=escan, **input.kwargs)
evaluator._outdir.envvar = "$SCRATCH"
functional = Functional(evaluator)
functional.popsize = input.population_size
functional.max_gen = input.max_generations
functional.rate = input.offspring_rate
functional.mean_conc = input.mean_conc
functional.stddev_conc = input.stddev_conc
functional.pools = input.pools
gajob = jobfolder / "scale_{0:.2f}/period_{1}/trial_{2}".format(scale, p, trial)
gajob.functional = functional
ip = IPython.ipapi.get()
ip.user_ns["current_jobfolder"] = jobfolder
ip.magic("savejobs " + path)
def gajobs(path, inputpath = "input.py"):
from copy import deepcopy
import IPython.ipapi
from pylada.ga.escan.elemental.functional import Darwin as Functional
from pylada.ga.escan.elemental.evaluator import Dipole as Evaluator
from pylada.ga.escan.elemental import Converter
from pylada.jobs import JobFolder
from pylada.escan import read_input
# reads input file.
input = read_input(inputpath)
jobfolder = JobFolder()
for scale in input.scales:
for range in input.ranges:
nmin, nmax = min(range), max(range)
for trial in input.trials:
escan = deepcopy(input.escan)
escan.vff.lattice.scale = scale
converter = Converter(growth=input.growth_direction, lattice=escan.vff.lattice)
kwargs = { "popsizse": input.population_size, "rate": input.offspring_rate,
"max_gen": input.max_generations, "pools": input.pools,
"crossover_rate": input.crossover_rate, "swap_rate": input.swap_rate,
"growth_rate": input.growth_rate, "nmin": nmin, "nmax": nmax,
"dosym": input.dosym, "rootworkdir": input.rootworkdir }
kwargs.update(getattr(input, 'kwargs', {}))
evaluator = Evaluator( converter = converter,
functional = escan,
**getattr(input, 'kwargs', {}) )
functional = Functional(evaluator, **kwargs)
gajob = jobfolder / "{4[0]}{4[1]}{4[2]}/scale_{0:.2f}/{1}_{2}/trial_{3}"\
.format(scale, nmin, nmax, trial, input.growth_direction)
gajob.functional = functional
ip = IPython.ipapi.get()
ip.user_ns["current_jobfolder"] = jobfolder
ip.magic("savejobs " + path)
# You should have received a copy of the GNU General Public License along with PyLaDa. If not, see
# <http://www.gnu.org/licenses/>.
###############################
from sys import exit
from shutil import rmtree
from os.path import exists, join
from math import ceil, sqrt
from numpy import dot, array, matrix
from numpy.linalg import norm
from boost.mpi import world
from pylada.vff import Vff
from pylada.crystal import Structure, Lattice, fill_structure
from pylada.escan import read_input
input = read_input("input.py")
structure = Structure()
structure.set_cell = (4, 0, 0.5),\
(0, 1, 0),\
(0, 0, 0.5)
structure = fill_structure(structure.cell)
for i, atom in enumerate(structure.atoms):
atom.type = "Si" if i < len(structure.atoms)/2 else "Ge"
result_str = Structure()
result_str.scale = 5.450000e+00
result_str.set_cell = (4.068890e+00, -4.235770e-18, 5.083297e-01),\
(-1.694308e-17, 1.016103e+00, 2.238072e-18),\
(-2.252168e-03, 8.711913e-18, 5.083297e-01)
def create_sl(path, direction, nmin, nmax, nstep, x=0.5, density=10e0, input='input.py'):
""" Creates dictionary of superlattices.
:Parameters:
path : str
Path to output dictionary.
direction : callable(int)->(3x3, 3x1)
A callable taking the number of layers on input and returning a tuple
consisting of a supercell with the correct number of layers and the
direction of growth of the superlattice. The supercell must be
have the correct periodicity: two vectors should be parallel to the
substrate such that layers can actually be defined. Otherwise, there
is not one-to-one correspondance between atoms and layers: the same
atom could belong to different layers. A number of examples are given
in this module: `direction001`, `direction011`, `direction111`.
nmin : int
Minimum number of layers.
nmax : int
Maximum number of layers (excluded).
nstep : int
Step between layers: [``nmin``, ``nmin``+``nstep``,
``nmin``+2*``nstep``, ..., ``nmax``[
x : float
Concentration in Si of the superlattice. Should be between 0 and 1.
density : float
Kpoint density for escan calculations,
input : str
Path to input file containing escan functional.
"""
from IPython.ipapi import get as get_ipy
from numpy.linalg import norm, inv
from pylada.jobs import JobFolder
from pylada.escan import read_input, exec_input, ReducedKDensity
from pylada.crystal.binary import zinc_blende
from pylada.crystal import layer_iterator
ip = get_ipy()
input = read_input(input)
kescan = exec_input(repr(input.escan).replace('Escan', 'KEscan')).functional
lattice = zinc_blende()
lattice.sites[0].type = 'Si', 'Ge'
lattice.sites[1].type = 'Si', 'Ge'
lattice.scale = 5.45
lattice.find_space_group()
density = 10e0 * max([1e0/norm(u) for u in inv(lattice.cell)])
rootjobs = ip.user_ns.get('current_jobfolder', JobFolder())
for n0 in range(nmin, nmax, nstep):
# create structure
cell, dir = direction(n0)
structure = lattice.to_structure(cell) # reduction not necessary if cell function done right.
N0 = int(len([0 for layer in layer_iterator(structure, dir)]) * x+1e-8)
for i, layer in enumerate(layer_iterator(structure, dir)):
for atom in layer: atom.type = 'Si' if i < N0 else 'Ge'
xp = float(len([0 for atom in structure.atoms if atom.type == 'Si']))
xp /= float(len(structure.atoms))
assert abs(xp - x) < 1e-12
# name and scale.
structure.name = "{0[0]}{0[1]}{0[2]}/x_{1:0<4.3}/n_{2}".format(dir, x, n0)
structure.scale = scale(structure)
# creates job-folder.
jobfolder = rootjobs / structure.name
jobfolder.jobparams['structure'] = structure.copy()
jobfolder.functional = kescan.copy()
jobfolder.functional.kpoints = ReducedKDensity(density, (0.5, 0.5, 0.5))
jobfolder.functional.reference = None
jobfolder.functional.fft_mesh = fftmesh(structure.cell)
jobfolder.functional.nbstates = int(len(structure.atoms) * 4 * 1.5+0.5)
if 'current_jobfolder' not in ip.user_ns: ip.user_ns["current_jobfolder"] = rootjobs
ip.magic("savejobs " + path)
return
from math import fabs as abs
from numpy import array
from pylada.crystal import fill_structure
from pylada.escan import read_input
from pylada.escan.emass import Functional
input = read_input("input.py", namespace = {"Escan": Functional})
G = array([0,0,0], dtype="float64")
X = array([0,0,1], dtype="float64")
structure = fill_structure(input.vff.lattice.cell, input.vff.lattice)
input.escan.nbstates = len(structure.atoms) * 4 + 4
input.escan.tolerance = 1e-12
orig = input.escan( structure, direction=(0,0,1), outdir="results/emass/100", \
do_relax_kpoint=False, type="e" )
assert abs(orig.mass[0] - 0.4381) < 0.01 # Gamma conduction emass
result = input.escan( structure, direction=((0,0,1), (1,1,1)), outdir="results/hmass", \
do_relax_kpoint=False, type="h", bandgap=orig.extract_bg )
assert abs(result.mass[0][0] - 0.2769) < 0.01 # Gamma conduction heavy hmass (100 direction)
assert abs(result.mass[0][2] - 0.2059) < 0.01 # Gamma conduction light hmass (100 direction)
assert abs(result.mass[1][0] - 0.6885) < 0.01 # Gamma conduction heavy hmass (111 direction)
assert abs(result.mass[1][2] - 0.1460) < 0.01 # Gamma conduction light hmass (111 direction)
# your option) any later version.
#
# PyLaDa is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even
# the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along with PyLaDa. If not, see
# <http://www.gnu.org/licenses/>.
###############################
from numpy import array
from pylada.escan import read_input, exec_input, ReducedBPoints
from pylada.crystal.binary import zinc_blende
X = array( [1,0,0], dtype="float64" )
G = array( [0,0,0], dtype="float64" )
L = array( [0.5,0.5,0.5], dtype="float64" )
W = array( [0, 0.5,1], dtype="float64" )
input = read_input('input.py')
kescan = exec_input(repr(input.escan).replace('Escan', 'KEscan')).functional
structure = zinc_blende().to_structure(subs={'A':'Si', 'B':'Si'})
structure.scale = 5.45
kescan.fft_mesh = 14, 14, 14
kescan.kpoints = ReducedBPoints(density=20) + (X, G) + (G, L)
result = kescan( structure, outdir='results/kescan',
nbstates = len(structure.atoms) * 4 + 4,
eref = None )
def create_start(path, nall = 3, nrand = 5, nmax=100, density=10e0, input='input.py'):
""" Creates dictionary with input structures for Si/Ge.
:Parameters:
path : str
Path to output dictionary.
nall : int
All structure with ``nall`` (excluded) unit-cells are included in the final dictionary.
nrand : int
Structures between ``nall`` (included) and ``nrand`` (excluded) unit-cells are
also considered for inclusion in the final dictionary. However, only
``nmax`` are randomly chosen in the end.
nmax : int
Structures between ``nall`` (included) and ``nrand`` (excluded) unit-cells are
also considered for inclusion in the final dictionary. However, only
``nmax`` are randomly chosen in the end.
density : float
Kpoint density for escan calculations,
input : str
Path to input file containing escan functional.
Creates a job-dictionary with a number of structures sampled from an
exhaustive list of structures to evaluate using escan.
"""
from random import shuffle
from itertools import chain
from IPython.ipapi import get as get_ipy
from numpy.linalg import norm, inv
from pylada.enumeration import Enum
from pylada.crystal.binary import zinc_blende
from pylada.jobs import JobFolder
from pylada.escan import read_input, exec_input, ReducedKDensity
from pylada.crystal.gruber import Reduction
input = read_input(input)
kescan = exec_input(repr(input.escan).replace('Escan', 'KEscan')).functional
enum = Enum(zinc_blende())
enum.sites[0].type = 'Si', 'Ge'
enum.sites[1].type = 'Si', 'Ge'
enum.scale = 5.45
enum.find_space_group()
density = density * max([1e0/norm(u) for u in inv(enum.cell * enum.scale).T])
strs = [u for n in range(nall, nrand) for u in enum.xn(n)]
shuffle(strs)
strs = [enum.as_structure(*u) for u in strs[:nmax]]
alls = [structure for n in range(nall) for structure in enum.structures(n)]
jobs = JobFolder()
for i, structure in enumerate(chain(alls, strs)):
structure.name = str(i)
nSi = len([a.type for a in structure.atoms if a.type == 'Si'])
structure.scale = float(nSi) / float(n) * enum.scale + float(n - nSi) / float(n) * 5.69
jobfolder = jobs / structure.name
jobfolder.jobparams['structure'] = structure.copy()
jobfolder.structure.cell = Reduction()(jobfolder.structure.cell)
jobfolder.functional = kescan.copy()
jobfolder.functional.kpoints = ReducedKDensity(density, (0.5, 0.5, 0.5))
jobfolder.functional.reference = None
jobfolder.functional.fft_mesh = fftmesh(structure.cell)
jobfolder.functional.nbstates = int(len(structure.atoms) * 4 * 1.5+0.5)
ip = get_ipy()
ip.user_ns["current_jobfolder"] = jobfolder.root
ip.magic("savejobs " + path)
return
