calculations_example

import os
from pprint import pprint
from pathlib2 import Path

The aim of the STREAMM package is to design a project with numerous calculations, run them on local and remote resources and collect the output for analysis, thus facilitating high-throughput computational material design.

To accomplish this, the directory structure is contained within a resource as a dictionary. Resources, structures, and forcefields are contained within a calculation object. Sets of calculations are contained within a project

So let’s first create a resource object that we will use to set the directory locations of all the subsequent calculation objects

import streamm

We need some output from another example so check that it has been created first

need_files = ['ethane_struc.json']
for f in need_files:
    path = Path(f)
    if not path.is_file():
        print("Need to run buildingblocks_example.ipynb")
        os.system("jupyter nbconvert --to python  buildingblocks_example.ipynb")
        os.system("python buildingblocks_example.py")
need_files = ['oplsaa_param.json']
for f in need_files:
    path = Path(f)
    if not path.is_file():
        print("Need to run forcefields_example.ipynb")
        os.system("jupyter nbconvert --to python  forcefields_example.ipynb")
        os.system("python forcefields_example.py")
import streamm.calculations.resource as resource
res_i = resource.Resource('local')

This sets the current working directory as the root/home directory by default

pprint(res_i.dir)
EXAMPLE_DIR = res_i.dir['home']

However, we want to use structures from our previous structures and forcefields examples, so let’s set the materials directory to examples/

res_i.dir['materials'] = res_i.dir['home']

To write out input files we will use the templates provided in the streamm package

Set the template dir dictionary entry to the location of templates directory

res_i.dir['templates'] =  os.path.join(EXAMPLE_DIR,'..','templates','')
print(res_i.dir['templates'])

This also contains the properties dictionary, which can be used to write .pbs scripts on clusters

pprint(res_i.properties)

By default the resource type is ‘local’; however, setting type to ‘ssh’ will invoke an scp command when copying files

Okay create the directories we need for our calculation

res_i.make_dir()

Now we should have materials, scratch, scripts, storage and templates directories

We can create a gaussian calculation

import streamm.calculations.gaussian as gaussian
calc_i = gaussian.Gaussian('ethane_HF')

Set the resource and all the directories

calc_i.set_resource(res_i)
pprint(calc_i.dir)

Make the calculation directories

calc_i.make_dir()

Let’s assign a structure to this calculation

First copy the .xyz file from the materials directory to our scratch directory using the cp_file() function.

os.chdir(calc_i.dir['home'])

This takes an type and key to set the calc_i.files[type][key] dictionary

file_type = 'input'
file_key = 'xyz'
file_name = "ethane_struc.json"
from_dirkey = 'materials'
to_dirkey = 'scratch'
calc_i.cp_file(file_type,file_key,file_name,from_dirkey,to_dirkey)

Generally, the materials directory is thought to contain a repository of material files, and local versions in the scratch directory should be made in case modifications are necessary

Change to the scratch directory

pprint(calc_i.dir['scratch'])
os.chdir(calc_i.dir['scratch'])

Read in methane .json file from the structures example

calc_i.strucC.tag = 'ethane'
calc_i.strucC.import_json(read_file=True)
print(calc_i.strucC.print_properties())

Now that we have a structure and parameters for each interaction we can create an input file for a simulation

Get the bash run script for Gaussian. By setting the file_key to run, this will be the script that executed when the run() function is called

file_type = 'templates'
file_key = 'run'
file_name = "gaussian.sh"
from_dirkey = 'templates'
to_dirkey = 'scratch'
calc_i.cp_file(file_type,file_key,file_name,from_dirkey,to_dirkey)

Get the .com template

file_type = 'templates'
file_key = 'com'
file_name = "gaussian.com"
from_dirkey = 'templates'
to_dirkey = 'scratch'
calc_i.cp_file(file_type,file_key,file_name,from_dirkey,to_dirkey)

Make sure we are in the scratch directory

pprint(os.getcwd())

Load the template files into memory

calc_i.load_str('templates','com')
calc_i.load_str('templates','run')

Set the properties strings in the template files to have the values we want

calc_i.properties['commands'] = 'HF/3-21G SP'
calc_i.properties['charge'] = 0
calc_i.properties['spin_mult'] = 1
calc_i.properties['coord'] = calc_i.strucC.write_coord()

Replace the strings in the files[‘input’][‘com’]

calc_i.replacewrite_prop('com','input','com','%s.com'%(calc_i.tag))

Add the name of the .com file to the properties, and replace the strings in the files[‘input’][‘run’]

calc_i.properties['input_com'] = calc_i.files['input']['com']
calc_i.replacewrite_prop('run','scripts','run','%s.sh'%(calc_i.tag))

Save a .json file in the home directory

os.chdir(calc_i.dir['home'])
calc_json = calc_i.export_json()

Go to scratch directory and see if there is a completed output file for the calculation

os.chdir(calc_i.dir['scratch'])
calc_i.check()

Check the status

pprint("Calculation:{} has status:{}".format(calc_i.tag,calc_i.meta['status']))

If you have gaussian installed on your machine and g09 in your PATH you can run the bash script

calc_i.run()

You can read in the data from the log file

calc_i.add_file('output','log','{}.log'.format(calc_i.strucC.tag))
calc_i.check()
if(calc_i.meta['status'] == 'finished' ):
    calc_i.analysis()

Then compress the results and copy them to storage

calc_i.store()

Next, we can follow a similar procedure to run a LAMMPS MD simulation

import streamm.calculations.lammps as lammps
calc_j = lammps.LAMMPS('ethane_lmp')

Set the resource

calc_j.set_resource(res_i)

Make directories

calc_j.make_dir()
pprint(calc_j.dir)

This takes an type and key to set the calc_i.files[type][key] dictionary

file_type = 'input'
file_key = 'xyz'
file_name = "ethane_struc.json"
from_dirkey = 'materials'
to_dirkey = 'scratch'
calc_j.cp_file(file_type,file_key,file_name,from_dirkey,to_dirkey)
os.chdir(calc_j.dir['scratch'])

Read in the ethane .json file from the structures example

calc_j.strucC.tag = 'ethane'
calc_j.strucC.import_json(read_file=True)
print(calc_j.strucC.print_properties())

Set the forcefield particletypes

for pkey,p in calc_j.strucC.particles.items():
    if( p.symbol == 'C' ):
        p.paramkey = 'CT'
    elif( p.symbol == 'H' ):
        p.paramkey = 'HC'

Copy the forcefield parameter .json file to scratch and read it in

file_type = 'input'
file_key = 'param'
file_name = "oplsaa_param.json"
from_dirkey = 'materials'
to_dirkey = 'scratch'
calc_j.cp_file(file_type,file_key,file_name,from_dirkey,to_dirkey)
import streamm.forcefields.parameters as parameters
calc_j.paramC = parameters.Parameters('oplsaa')
calc_j.paramC.import_json()
print(calc_j.paramC)
for ptkey,pt in calc_j.paramC.particletypes.items():
    print(ptkey,pt,pt.unit_conf['energy'],pt.unit_conf['length'])
for btkey,bt in calc_j.paramC.bondtypes.items():
    print(btkey,bt,bt.unit_conf['harm_bond_coeff'],pt.unit_conf['length'])
for atkey,at in calc_j.paramC.angletypes.items():
    print(atkey,at,at.unit_conf['energy'],at.unit_conf['length'])

Use the set_ffparam() function to iterate through the structure container and set parameters based on paramkeys

calc_j.set_ffparam()

Now we have a structure that has forcefield parameters for each particle, bond and bond angle

Let’s get the input file template

file_type = 'templates'
file_key = 'in'
file_name = "lammps_sp.in"
from_dirkey = 'templates'
to_dirkey = 'scratch'
calc_j.cp_file(file_type,file_key,file_name,from_dirkey,to_dirkey)

Bash run file

file_type = 'templates'
file_key = 'run'
file_name = "lammps.sh"
from_dirkey = 'templates'
to_dirkey = 'scratch'
calc_j.cp_file(file_type,file_key,file_name,from_dirkey,to_dirkey)

Got to scratch dir

os.chdir(calc_j.dir['scratch'])

Read in template files

calc_j.load_str('templates','in')
calc_j.load_str('templates','run')

Write LAMMPS data file

calc_j.write_data()

Replace properties strings in template and write template

calc_j.replacewrite_prop('in','input','in','%s.in'%(calc_j.tag))

Set .in file in properties and write run script

calc_j.properties['input_in'] = calc_j.files['input']['in']
calc_j.replacewrite_prop('run','scripts','run','%s.sh'%(calc_j.tag))

Save a .json file in the home directory

os.chdir(calc_j.dir['home'])
calc_json = calc_j.export_json()

Go to scratch directory and see if there is a completed output file for the calculation

os.chdir(calc_j.dir['scratch'])
calc_j.check()
pprint("Calculation:{} has status:{}".format(calc_j.tag,calc_j.meta['status']))

So now we have two calculations, let’s put them in a project so we can operate on them both at the same time

import streamm.calculations.project as project
import copy
proj_i = streamm.Project('example_proj')
proj_i.add_calc(calc_i,deepcopy=True)
proj_i.add_calc(calc_j,deepcopy=True)

Now we can check the status of each calculation with a single command

proj_i.check()

We can run each simulation

proj_i.run()

We can tar up the results and copy the tar files to a storage location

proj_i.store()

And dump the details of the project to a json file

os.chdir(calc_i.dir['home'])
proj_i.export_json()
del proj_i
proj_i = streamm.Project('example_proj')
proj_i.import_json()
proj_i.check()

Neat-O!