qeflow Quantum Espresso workflow tool

QE-flow is a python tool which helps to automate quantum espresso workflows "without scassamento di maroni", i.e. easily. Out in the internet there are many other options that perform this very same tasks and in a much more efficient way. The most notable are pwtk and aiida. However, in my opinion, these tools have a very steap learning curve. So I decided to just write my own.

Installation

Simply download the latest version and run

pip install qeflow-X.Y.tar.gz

where you must replace X.Y with the version you have downloaded. In the future, if this project keeps growing, I'll try to upload qeflow to PyPI.

Quick start

QE-flow is meant to be easy. We only need to setup 2 files:

1. a very small configuration file (once and for all),
2. and an input file with the workflow instructions.

Note that the input file is very easy to setup: it is mostly a collection of the usual well known QE flags plus the workflow instructions.

1. Configuration file

The first thing to do is to create a configuration file. QE-flow will try to read the configuration file in the following order:

1. from ./qeflow.cfg (in the current directory)
2. or from $HOME/.qeflow.cfg (notice the dot, hidden file)

A configuration file skeleton is found in qeflow.cfg. In the following we show 2 examples.

1. Local configuration: useful we running simulation on a personal computer.

   # $HOME/.qeflow.cfg
   cluster: local
   pwx: path/to/pw.x
   mpix: path/to/mpirun

2. Slurm configuration: this is necessary when we are logged in a HPC facility with a slurm batch system.

   # $HOME/.qeflow.cfg
   cluster: slurm
   cores_per_node: 128
   account: account-name
   pyenv: /path/to/python/env/bin/activate
   modules:
       - quantum-espresso-7.1
   pwx: pw.x
   mpix: mpirun

In the future, I might implement also the PBS batch system.

2. Input file

The input file skeleton with the currently supported flags and relative description is found in qeflow.in. In the following I'll show only 3 minimal examples:

1. a simple scf calculation with different values of ecutwfc;
2. a scf followed by nscf with respect to both ecutwfc and kpoints
3. a vc-relax followed by a nscf at the relaxed structure.

2.1 Convergence of scf workflow

We want to perform a simple scf calculation with different values of ecutwfc. The input file is a yaml file, name it ecutSi.in (can be any name, actually) and type the following:

# ecutSi.in
---
name: ecutSi
workflow:
    - scf:
withrespectto:
    - ecutwfc: [10, 20, 30, 40, 50, 60, 70, 80, 90, 100]

nprocs: 128
time: 0:20:0
partition: standard
qos: short

pseudo_pots    : [Si.pbe-n-rrkjus_psl.1.0.0.UPF]
atoms          : [Si]
masses         : [28.085]
positions      : [[ Si,    0.0,    0.0,     0.0 ],
                  [ Si, 1.3575, 1.3575,  1.3575 ]]
unit_cell      : [[ 2.715, 2.715,   0.0 ],
                  [   0.0, 2.715, 2.715 ],
                  [ 2.715,   0.0, 2.715 ]]

kpoints        : [8, 8, 8]
prefix         : silicon
pseudo_dir     : /path/to/SSSP_efficiency

The first three flags are the workflow instructions:

• name: a unique name to identify this workflow, a folder with the same name will be created, containing the results of the calculations;
• workflow: a list of tasks. In this case we only have one task, the scf calcualtion;
• withrespectto: a list of parameters, where each of them is a also a list. In this case we only have one parameter ecutwfc spanning values from 10 to 100 Ry.

The second block of instruction applies only when running on a HPC. In this case we have the usual slurm specifications for number of processors, allocation time, partition and quality of service.

Finally, the third block of flags contains all the usual quantum espresso keys necessary for this calculation. The order in which they appear does not matter, just dump them in there (also remember to look at the definitions in qeflow.in).

Now it's time to run the workflow: the executable to call is simply qeflow. You can type qeflow -h for a short help menu showing other options such as dry-run and verbosity. Now type:

qeflow ecutSi.in

And that's it! After the calculation is done you should have all your calculations and results collected into a folder named ecutSi.

Most importantly, QE-flow parses automatically the most important properties of the system and collects them into a file in the results folder: ./ecutSi/results/silicon.txt. You can inspect this file using yaml and pandas:

import yaml
import pandas
with open('./ecutSi/results/silicon.txt', 'r') as f:
    data = yaml.safe_load(f)
df = pandas.DataFrame(data)
print(df[['ecutwfc','etot']])

The output is the following:

   ecutwfc        etot
0       10 -310.790657
1       20 -310.720741
2       30 -310.744642
3       40 -310.746610
4       50 -310.747418
5       60 -310.747733
6       70 -310.747919
7       80 -310.748074
8       90 -310.748119
9      100 -310.748170

2.2 A scf followed by nscf workflow, as function of ecutwfc and kpoint

In this example we show a more complex workflow, where we converge a calculation with respect to 2 parameters. The workflow instructions are the following

# ecutKpointSi.in
---
name: ecutKpointSi
workflow:
    - scf:
    - nscf:
        kpoints: [12,12,12]
withrespectto:
    - ecutwfc: [10, 20, 30, 40, 50, 60, 70, 80, 90, 100]
    - kpoints: [[4,4,4], [8,8,8], [12,12,12]]

# below this point is the same as in the previous example

The are some differences compared to the previous case:

• the workflow flag is now a list of 2 tasks: scf and nscf;
• the second nscf task declares kpoints: [12,12,12]: this overwrites any other kpoint declaration defined later in the input file;
• the withrespectto flag also has 2 entries: the final workflow will be the cartesian product of ecutwfc and kpoints values.

This means that the scf + nscf workflow will be performed 30 times, each time with a different ecutwfc and kpoints, but keeping the nscf kpoints fixed.

Everythong else remains the same. Data are collected in ./ecutKpointSi/results/silicon.txt and can be inspected with yaml and pandas.

2.3 A vc-relax followed by nscf workflow

Finally, last example is a vc-relax plus a nscf at the relaxed structure with no iterations to check convergence. The workflow instructions are

# relaxSi.in
---
name: relaxSi
workflow:
    - vc-relax:
        ecutwfc: 80
        mixing_beta: 0.6
    - nscf:
        kpoints: [12,12,12]
        mixing_beta: 0.7

# below this point is the same as the first case

The withrespectto flag is missing. No variation over any parameter will be performed. However, we can always fine tune each task by defining a specific flag that overwrites the default ones (given in the bottom section, not shown).

That's all for now.

Support

For any problems, questions or suggestions, please contact me at [email protected].

Roadmap

Currently this is a very basic python script. But I'll try to keep working on it, as long as I keep doing science, which is definitely not guaranteed LOL!
Anyway, future developments will include:

• pp.x, dos.x, bands.x, projwfc.x, plotband.x, wannier90.x, pw2w90.x support;
• more QE flags, with priority to:
  • ibrav different from 0;
  • kpoints for BZ path and band structure;
• more parsing options;
• intelligent parsing for dependent calculations.

Authors and acknowledgment

The development of QE-flow is proudly powered by me.
Quantum Espresso is licensed under GPL3 and citation guidelines can be found here and here.

License

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program 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 this program. If not, see https://www.gnu.org/licenses/.