please execute the cell below before starting the tutorial by selecting it and pressing Ctrl+Enter

%load_ext autoreload
%autoreload 2

from aiida import load_profile
load_profile()

from aiida.orm import Dict, load_node, load_code
from aiida.engine import submit

# First we will import a prepared dataset for this tutorial with some structures and simulations
# If this was already executed, it will add nothing to the database
!verdi archive import ~/4.AiiDA-FLEUR/files/fleur_tutorial_data.aiida

Input generator code

DISCLAIMER: In real life work, one will probably not execute inpgen or even fleur calculation like we show you below, but one rather runs directly higher workflows/workchains, which is part of the 6th Notebook. This is for demonstration purposes and to familiarize yourself with aiida and concepts of the plugin

You learned about the simplified input file for the FLEUR code. The inpgen code generates an inp.xml input-file for FLEUR using the simplified input. In AiiDA, a Code object representing input generator has to have a similar behaviour: it has to generate a FleurinpData from a list of simplified parameters. The code class wrapping inpgen code is called FleurinputgenCalculation. In this tutorial we are going to learn how to use it.

First things first - let us get familiar with inputs of the FleurinputgenCalculation.

Inputs of the input calculation

FleurinputgenCalculation has four main input nodes listed in the table below:

code is a Code representing inpgen inself in the database.

Simplified input file is represented as a pair of input nodes: structure and parameters. structure is a StructureData node to be used in the simplified input file (&lattice). parameters is a nested dictionary containing nested name: value pairs and representing all other parameters of the simplified input file. For instance, if one wants to add a line

&kpt
  div1=4   div2=5   div3=6 /

to the simplified input file, it is necessary to append

'kpt': {
    'div1': 4,
    'div2' : 5,
    'div3' : 6
    }

to the parameters dictionary.

In some cases the simplified input file does not have names for nested values. In this case you can refer to the list of all supported keys:

'input': ['film', 'cartesian', 'cal_symm', 'checkinp', 'symor', 'oldfleur']

'atom': ['id', 'z', 'rmt', 'dx', 'jri', 'lmax', 'lnonsph', 'ncst', 'econfig', 'bmu', 'lo', 'element', 'name']

'comp': ['jspins', 'frcor', 'ctail', 'kcrel', 'gmax', 'gmaxxc', 'kmax']

'exco': ['xctyp', 'relxc'],

'film': ['dvac', 'dtild'],

'soc': ['theta', 'phi'],

'qss': ['x', 'y', 'z'],

'kpt': ['nkpt', 'kpts', 'div1', 'div2', 'div3','tkb', 'tria'],

'title': {}

For example, you can find that SOC theta and phi values can be set appending

'soc': {
    'theta': 0.7,
    'phi' : 0.26
    }

to the parameters dictionary, resulting in:

&soc
  0.7 0.26 /

settings is used to modify a list of files to be copied to/from/between calculation folders and to add command options to the executable. We will not cover this in the tutorial.

Inpgen code submission

Inputs preparation

To submit an inpgen calculation, we need to import FleurinputgenCalculation class

from aiida_fleur.calculation.fleurinputgen import FleurinputgenCalculation

or use the CalculationFactory:

from aiida.plugins import CalculationFactory
FleurinputgenCalculation = CalculationFactory('fleur.inpgen')

and prepare all the inputs. Configured and ready-to-use inpgen code node is already stored in your database under PK=1. You can simply load it from the database:

inpgen_code = load_code('inpgen@localhost')

We are going to use a structure that you created in tutorial 1. Please, find a PK of the Fe monolayer structure and load it via the load_node() command:

# you need to modify this - replace Fe_PK
structure_Fe = load_node(<Fe_PK>)

# If you have installed iff_base via the iffdata command there are already some structures in your database
# or if you do not remember the pk you can execute
!verdi data core.structure list -A | grep Fe
# (there is also a precreated node '36b21c')

Next, we need to set up the parameters dictionary. For now we are going to specify k-mesh and kmax only:

parameters = Dict(dict={
    'comp': {
        'kmax': 3.85,
        },
    'kpt': {
        'div1': 4,
        'div2' : 4,
        'div3' : 4
        }})

Finally, we need to specify computational resources for the task. Since inpgen usually takes a few seconds to finish, we always run it in serial, passing False to 'withmpi' and setting 1 MPI task.

options = {
    'resources' : {"num_machines": 1, "num_mpiprocs_per_machine": 1},
    'queue_name': '',
    'withmpi': False
}

Assembling inputs in a single dictionary

All the input modes can be passed separately to the FleurinputgenCalculation. However, it much more convenient to assemble them in a single dictionary first, using get_builder() method:

inputs = FleurinputgenCalculation.get_builder()
inputs.parameters = parameters
inputs.code = inpgen_code
inputs.structure = structure_Fe
inputs.metadata.options = options  # note options not in inputs but inputs.metadata

Submission

To submit the input generator, run submit function passing the input dictionary:

inpgen_process = submit(inputs)
print('The PK of submitted inpgen job is {}'.format(inpgen_process))

inpgen_pk contains a PK of the input generator calculation node stored in the database. Later we can access all the information related to this run via this PK.

Results analysis

After inpgen code is submitted, you can check the status of all processes submitted not more than 24 hours ago:

!verdi process list -a -p 1

If you have to many process you can use the filter options of the verdi process command or grep in the output

!verdi process list -a | grep inputgen

You can also check the status of inpgen calculation directly. Try running next cell and examine the output:

# you need to modify this: insert an integer instead of INPGEN_PK
!verdi process status INPGEN_PK

Since input generator takes a few seconds to finish, you most probably will not be able to catch it in any other state except 'Finished'. Now let us explore the output nodes. Fleurinpgencalculation generates three output nodes:

You can get a list of their PKs running:

# you need to modify this: insert an integer instead of INPGEN_PK
!verdi process show INPGEN_PK

Now you can load any of the outputs via load_node() function. In addition, all the outputs can be found in calc_node.outputs:

# you need to modify this: insert an integer instead of INPGEN_PK
inpgen_calc_node = load_node(INPGEN_PK)
output_fleurinp = inpgen_calc_node.outputs.fleurinp
print('PK of the output FleurinpData is {}'.format(output_fleurinp.pk))

Please, remember the FleurinpData PK, we are going to use this in our next tutorial.

Scripting tasks

Jupyter notebooks are convinient way to learn and run some jobs. However, I personally prefer using and running python scripts in a terminal for my work. As a final task of each tutorial, beggining from this one, you will be asked to construct a small python script and run it in a terminal.

First, you need to start a new terminal

To execute a python-script, run

 $verdi run NAME_OF_THE_SCRIPT.py

if the the script contains the load_profile() line, you can also simply execute it with

 $python3 NAME_OF_THE_SCRIPT.py

1. Run an inpgen calculation for Si

Write a script that submits FleurinpgenCalculation for a Si structure given in the attached Si.cif file. Try to set up parameters dictionary to use parameters:

Try finding out the PK of the output FleurinpData and remember it - you will need it in the next tutorial!

2. Run an inpgen calculation for film structures

Write a script that submits FleurinpgenCalculation for a fcc Fe, Ni and Co mololayer film (110) structures with a perioud length = 4.16 Å. Try to set up parameters dictionary to use parameters:

Also make sure that you remember all structures and FleurinpData PKs - we will use it later!

Hint: Use StructureData methods. fcc (110) monolayer can be represented as a cell having a single atom at (0, 0, 0):

cell = [[0.7071068*a, 0.0, 0.0],
        [0.0, 1.0*a, 0.0],
        [0.0, 0.0, 0.7071068*a]]

where a is a period length. You also need to set up periodic boundary conditions via

structure.pbc = (True, True, False)

Try finding out PKs of the output FleurinpData and remember it - you will need it in the next tutorial!