SPARC-X-API: A Python API for the SPARC-X DFT Code

SPARC-X-API is a versatile Python API for the real-space density functional (DFT) package SPARC distributed under the GPLv3 license. SPARC-X-API leverages the powerful Atomic Simulation Environment (ASE) framework for manipulating input / output files, as well as running DFT calculations and analysis via the SPARC code written in C/C++. Key features include:

1. ASE-compatible I/O format for SPARC files
2. A JSON Schema interfacing with SPARC's C/C++ code for parameter validation and conversion
3. A comprehensive calculator interface for SPARC with file I/O and socket-communication support.

Overview

SPARC-X-API is part of the SPARC-X project, a collection of open-source packages aim to provide modern and efficient implementation of real space DFT simulations, led by the research groups of Phanish Suryanarayana and Andrew J. Medford from Georgia Tech. The name SPARC stands for Simulation Package for Ab-initio Real-Space Calculations, which comes in two variations:

• M-SPARC: self-consistent Matlab code for algorithm prototyping and testing
• SPARC: C/C++ implementation of efficient production code scaling up to millions of atoms

The SPARC-X project shares common input / output file formats, and parameter specification. SPARC-X-API serves as the interface that connects the core SPARC-X components with external workflows, as illustrated in the diagram below.

Quick start

SPARC-X-API is straightforward to install and use, adhering to the ASE standard for seamless integration into other computational workflows.

Installation

Install SPARC-X-API via conda.

conda install -c conda-forge sparc-x-api

Install the pre-compiled SPARC binary alongside SPARC-X-API (Linux only).

conda install -c conda-forge sparc-x

Note: the official SPARC binary in the conda-forge channel does not come with socket support yet. Please follow this instruction if you wish to install socket-compatible SPARC code

Setup SPARC-X-API

Preferences for SPARC-X-API and SPARC C/C++ code can be defined in ASE configuration file, located at ~/.config/ase/config.ini, such as following example:

[sparc]
; `command`: full shell command (include MPI directives) to run SPARC
command = srun -n 24 path/to/sparc

; `psp_path`: directory containing pseudopotential files (optional)
psp_path = path/to/SPARC/psps

; `doc_path`: directory for SPARC LaTeX documentation to build JSON schema on the fly (optional)
doc_path = path/to/SPARC/doc/.LaTeX/

Reading / Writing SPARC files

SPARC-X-API provides a file format sparc compatible with the ASE ioformat, which treats the calculation directory containing SPARC in-/output files as a bundle:

• Read from a SPARC bundle

# `format="sparc"` should be specified
from ase.io import read
atoms = read("sparc_calc_dir/", format="sparc")

• Write input files

# `format="sparc"` should be specified
from ase.build import bulk
atoms = bulk("Al") * [4, 4, 4]
atoms.write("sparc_calc_dir/", format="sparc")

Visualizing Atomic Structures in SPARC Files

You can use the ase gui commandline tool to visualize SPARC files:

ase gui sparc_calc_dir/*.ion

Parameter Validation with JSON Schema

SPARC-X-API allows user to validate SPARC parameters based on a JSON schema that is parsed from the LaTeX documentation of the SPARC-X project. To get help for a specific parameter:

from sparc.api import SparcAPI
print(SparcAPI().help_info("LATVEC"))

Running SPARC Calculations

SPARC-X-API provides two ways to run a DFT calculation via SPARC C/C++ code:

1. File I/O mode: classic way to drive SPARC calculation by running a standard SPARC process with input files. Suitable for things implemented internally in SPARC C/C++ codes:

   • Single point evaluation
   • Band structure calculations
   • Structural optimization (SPARC internal routines)
   • Ab-init molecular dynamics (AIMD)

2. Socket mode: run a background SPARC process while providing atomic positions and other data via socket communication. Suitable for:

   • Hundreds / thousands of single point DFT evaluations
   • Integration with complex algorithms / workflows
   • Combination with internal and external machine learning (ML) force fields

The calculator interface in SPARC-X-API is designed to be intuitive for users familiar with the ASE calculator interfaces for other DFT packages (e.g. VASP, Quantum ESPRESSO, GPAW, Abinit, etc):

File I/O mode

Run a single point DFT calculation with Dirichlet boundary conditions:

from sparc.calculator import SPARC
from ase.build import molecule
atoms = molecule("H2", cell=(10, 10, 10), pbc=False)
atoms.calc = SPARC(h=0.25, directory="run_sp") # 0.25 Å mesh spacing
atoms.get_potential_energy()
atoms.get_forces()

Socket mode

Note: A socket-compatible SPARC binary installation is required. Please check this instruction for more details.

Switching to the socket mode requires just a few parameters, ideal for workflows with hundreds or thousands of single point DFT calls with much less overhead and more flexibility. An example for optimization using socket mode and ASE optimizer:

from sparc.calculator import SPARC
from ase.build import molecule
from ase.optimize import BFGS
atoms = molecule("H2", cell=(10, 10, 10), pbc=False)
atoms.center()
atoms.calc = SPARC(h=0.25, directory="run_sp", use_socket=True) # 0.25 Å mesh spacing
opt = BFGS(atoms)
with atoms.calc:
    opt.run(fmax=0.01)

Documentation

Please check the full documentation for details regarding installation, usage, troubleshooting and contribution guidelines.

How to cite

If you find SPARC-X-API help, please consider cite the relevant publications below:

• The SPARC-X-API package itself: Tian et al. 2024
• The SPARC C/C++ code
  • v2.0 Zhang et al., 2024
  • v1.0 Xu et al., 2021
• The M-SPARC Matlab code
  • v2.0 Zhang et al., 2023
  • v1.0 Xu et al., 2020

For a full list of publications in the SPARC-X project please refer to:

• SPARC developement
• M-SPARC development
• Pseudopotentials

Acknowledgment

The development of SPARC-X-API is supported by the U.S. Department of Energy, Office of Science, under Grant No. DE-SC0019410 and DE-SC0023445.