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

Table of Contents

• Installation
• Setting Up Environment
• Basic Usage
• Advanced Usage: Socket Interface
• Troubleshooting
• Advanced Topics
• Support and Contribution

SPARC-X-API is an ASE-compatible Python API for the density functional theory (DFT) code SPARC. It offers:

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

Fig. 1 provides an overlook of the components of SPARC-X-API and its relation with the SPARC C-code.

Fig. 1 Schematic drawing for the architecture of the SPARC-X-API package

Installation:

The Python API may be installed via either of the following approaches:

1. Via anaconda or miniconda (recommended)

Set up a conda environment and install the Python API, which includes the pseudopotential files:

# Change 'sparc-env' to your desired name if needed
conda create -n sparc-env
conda activate sparc-env
conda install -c conda-forge sparc-x-api

On Linux platforms (x86_64, aarch64), you can also install the precompiled sparc DFT binaries alongside the API:

conda install -c conda-forge sparc-x
conda activate sparc-env   # Re-activate to have the env variables effective

2. Manual installation from source with pip

python -m pip install git+https://github.com/SPARC-X/SPARC-X-API

Optionally, you can download the latest SPMS pseudopotentials and unpacks the pseudopotential files into <python-lib-root>/site-packages/sparc/psp:

python -m sparc.download_data

To utilize the API for drive SPARC calculations, please following the SPARC manual for compilation and installation of the SPARC DFT code itself.

Post-installation check

We recommend the users to run a simple test after installation and setup:

python -m sparc.quicktest

A proper setup will display the following sections at the output's conclusion:

For using the API to parse SPARC input and output files, it's essential that the "Import" and "JSON API" tests are successful. For run SPARC calculations, all tests must pass.

Please refer to the Setting Up the Environment or guidance on correctly configuring the environment variables. If you run into further problems, consult our Trouble Shooting.

Setting up the environment

SPARC-X-API is designed to automate the discovery of pseudopotential files, the JSON API, and the SPARC binary. However, you can exert fine-grained control over their setup:

A) Pseudopotential files

Pseudopotential files (in Abinit psp8 format) are loaded in the following order:

1. Via the psp_dir argument passed to the sparc.SPARC calculator.
2. Through the environment variables $SPARC_PSP_PATH or $SPARC_PP_PATH (this is the method employed by conda installation).
3. By using psp8 files bundled with the SPARC-X-API installation (see the manual installation).

To specify a custom path for your psp8 files, set the $SPARC_PSP_PATH or $SPARC_PP_PATH variable as follows:

export SPARC_PSP_PATH="/path/to/your/psp8/directory"

To determine the default location of psp8 files (as per option 3), run the following code:

python -c "from sparc.common import psp_dir; print(psp_dir)"

B) JSON schema

SPARC-X-API is engineered for compatibility with the SPARC C-code. It achieves this by loading a JSON schema for parameter validation and unit conversion. You can review the default schema used by the API at sparc.sparc_json_api.default_json_api

"FD_GRID": {
   "symbol": "FD_GRID",
   "label": "FD_GRID",
   "type": "integer array",
   "default": null,
   "unit": "No unit",
   "example": "FD_GRID: 26 26 30",
   "description": "#<Some description...>",
   "allow_bool_input": false,
   "category": "system"
  },

The schema file is generated from SPARC's LaTeX documentation. In upcoming releases of SPARC-X-API, we're aiming to provide users the flexibility to use their own custom schema files. This would be particularly useful for those who might be testing a development branch of SPARC. By default, the JSON schema is packaged under sparc/sparc_json_api directory. If you have another version of SPARC source code, you can set the environment variable $SPARC_DOC_PATH to the directory containing the LaTeX codes for the documentation, such as <SPARC-source-code-root>/doc/.LaTeX. If you obtain sparc-x from the conda method as mentioned above, By default, the JSON schema is packaged under sparc/sparc_json_api directory. If you have another version of SPARC source code, you can set the environment variable $SPARC_DOC_PATH is automatically set to <conda-env-root>/share/doc/sparc/.LaTeX. Setting up the environment variable $SPARC_DOC_PATH helps loading the correct JSON schame that is compatible with your SPARC binary code.

C) SPARC Command Configuration

The command to execute SPARC calculations is determined based on the following priority:

1. The command argument provided directly to the sparc.SPARC calculator.
2. The environment variable $ASE_SPARC_COMMAND
3. If neither of the above is defined, SPARC-X-API looks for the SPARC binary under current $PATH and combine with the suitable mpi command prefix.

Example:

export ASE_SPARC_COMMAND="mpirun -n 8 /path/to/sparc -name PREFIX"

Note: the -name PREFIX part can be omitted the label property of the sparc.SPARC calculator is set (which is the default behavior). Any extra features of the SPARC code (e.g. GPU acceleration) should be specified in the command.

Basic usage of the Python API

1. Read / write SPARC files

In contrast to many other DFT codes, where the ASE I/O formats refer to a single file, SPARC-X-API operates on the whole calculation directory, also known as a "SPARC bundle". This API integrates seamlessly with ASE, allowing for the automatic detection of the SPARC file format:

• Reading from a SPARC bundle

import sparc
from ase.io import read, write
atoms = read("test.sparc", index=-1)

Note: To read multiple output files from the same directory, e.g., SPARC.aimd, SPARC.aimd_01, pass the keyword argument include_all_files=True to read()

• Writing a minimal SPARC bundle from atoms

import sparc
from ase.io import read, write
from ase.build import Bulk
atoms = Bulk("Al") * [4, 4, 4]
atoms.write("test.sparc")

For a deeper dive into the bundle I/O format, see Advanced Topics.

2. JSON Schema for SPARC calculator

A recurring challenge of Python interfaces to DFT codes it the inconsistencies between low-level codes (Fortran/C/C++) and outdated upper-level APIs regarding parameter sets and default values. To address this issue, SPARC-X-API handles DFT parameters through a JSON schema translated from SPARC's LaTeX documentation. Each release of SPARC-X-API is linked with a specific version of the SPARC source code, ensuring compatibility and consistency with the default parameter set. The main driver of this feature is the sparc.api.SparcAPI class.

If you've obtained the full SPARC source code, you can generate a copy of the schema by the following code:

python -m sparc.docparser <sparc-source-code-root>/doc/.LaTeX

which produces a parameters.json file.

To learn more about the JSON schema design, please refer to Advanced Topics.

3. Calculator interface (File-IO mode)

SPARC-X-API offers a calculator interface based on file I/O that aligns with many other ASE calculators. If you've worked with ASE modules like Vasp, QuantumEspresso, or GPAW, you'll find this package intuitive, as shown in the following examples:

1. Single point calculation

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

This example sets up a calculation for H2 atoms in a 10 Å x 10 Å x 10 Å PBC cell with default parameters (PBE exchange correlation functional, and a grid spacing (h) of 0.25 Å). Note by calling atoms.get_forces, the calculator will automatically sets the flags for printing the forces.

2. Geometric optimization (using SPARC's internal routines)

from sparc.calculator import SPARC
from ase.build import bulk
atoms = bulk("Al", cubic=True)
atoms.rattle(0.05)
atoms.calc = SPARC(h=0.25, kpts=(3, 3, 3), relax_flag=True, directory="run_opt")
atoms.get_potential_energy()
atoms.get_forces()

This example sets up a calculation for a rattled Aluminum primitive unit cell, calculate with PBE functional, grid spacing of 0.25 Å, and 3 x 3 x 3 k-point grid. Optimization of ionic positions is handled with SPARC's default LBFGS routine.

3. AIMD in SPARC

from sparc.calculator import SPARC
from ase.build import bulk
md_params = dict(md_flag=True, ion_temp=800, md_method="NVE", md_timestep=0.6, md_nstep=5)
atoms = bulk("Al") * (3, 3, 3)
atoms.rattle()
atoms.calc = SPARC(h=0.25, kpts=(1, 1, 1), directory="run_aimd", **md_params)
atoms.get_potential_energy()

This example runs a short NVE MD simulation (5 steps) at 800 K for 27 Al atoms. If you want to extract more information about the MD simulation steps, take a look at SPARC.raw_results.

4. Geometric optimization using ASE's optimizers

The power of SPARC-X-API is to combine single point SPARC calculations with advanced ASE optimizers, such as BFGS, FIRE or GPMin. Example 2 can be re-written as:

from sparc.calculator import SPARC
from ase.build import bulk
from ase.optimize import LBFGS
atoms = bulk("Al", cubic=True)
atoms.rattle(0.05)
atoms.calc = SPARC(h=0.25, kpts=(3, 3, 3), directory="run_opt_ase")
opt = LBFGS(atoms, alpha=90)
opt.run(fmax=0.02)

4. Command-line tools

A simple command wrapper sparc-ase is provided to add support of SPARC file formats to the ase cli tools. Simple replace ase [subcommand] [args] with sparc-ase [subcommand] [args] to access your SPARC bundle files as you would use for other file formats. As an example, use sparc-ase gui path/to/your/bundle.sparc for the visualization of atomistic structures. Depending on the bundle's contents, this could display individual atoms or multiple images.

Fig. 2 is a screenshot showing the usage of sparc-ase gui to visualize a short MD trajectory.

Fig 2. A screenshot of the sparc-ase program

5. Parameters and units used in SPARC-X-API

In the SPARC DFT code, all input parameters conventionally employ atomic units, such as Hartree and Bohr. Conversely, ASE objects (like Atoms.positions, Atoms.cell, Atoms.get_potential_energy()) utilize eV/Angstrom units.

When you set up a calculator as below:

atoms.calc = SPARC(h=0.25, REFERENCE_CUTOFF=0.5, EXX_RANGE_PBE=0.16, **params)

inputs following ASE's convention (e.g., h) adopt eV/Angstrom units (thus the same setting can be applied to other DFT calculators), On the other hand, all SPARC-specific parameters, which can often be recognized by their capitalized format (like REFERENCE_CUTOFF, EXX_RANGE_PBE), retain their original values consistent with their representation in the .inpt files. The reasoning and details about unit conversion can be found in the Rules for Input Parameters in Advanced Topics.

In order for SPARC-X-API to be compatible with other ASE-based DFT calculators, there is a list of special parameters consistent with the ASE convention and uses Å / eV / GPa / fs unit system:

parameter name	meaning	example	equivalent SPARC input
xc	Exchange-correlation functional	xc=pbe	EXCHANGE_CORRELATION: GGA_PBE
		xc=lda	EXCHANGE_CORRELATION: LDA_PZ
		xc=rpbe	EXCHANGE_CORRELATION: GGA_RPBE
		xc=pbesol	EXCHANGE_CORRELATION: GGA_PBEsol
		xc=pbe0	EXCHANGE_CORRELATION: PBE0
		xc=hf	EXCHANGE_CORRELATION: HF
		xc=hse or xc=hse03	EXCHANGE_CORRELATION: HSE
		xc=vdwdf1 or xc=vdw-df	EXCHANGE_CORRELATION: vdWDF1
		xc=vdwdf2 or xc=vdw-df2	EXCHANGE_CORRELATION: vdWDF2
		xc=scan	EXCHANGE_CORRELATION: SCAN
h	Real grid spacing (Å)	h=0.2	MESH_GRID: 0.38 (in Bohr)
gpts	Explicit grid points	gpts=[10, 10, 10]	FD_GRID: 10 10 10
kpts	Kpoint mesh	kpts=[3, 3, 3]	KPOINT_GRID: 3 3 3
convergence	Dict of convergence criteria (see below)
	energy eV/atom	convergence={"energy": 1e-4}	TOL_SCF: 3e-6
	relax (forces) eV/Å	convergence={"relax": 1e-2}	TOL_RELAX: 2e-4
	density e/atom	convergence={density: 1e-6}	TOL_PSEUDOCHARGE: 1e-6

Users from other DFT codes can easily port their ASE codes to SPARC-X-API using the special parameters with minimal modification:

Example 1: VASP vs SPARC

# Using VASP
from ase.calculators.vasp import Vasp
calc = Vasp(xc="rpbe", kpts=(9, 9, 9), directory="vasp-calc")

vs

# Using SPARC
from sparc.calculator import SPARC
calc = SPARC(xc="rpbe", kpts=(9, 9, 9), directory="sparc-calc.sparc")

Example 2: GPAW (another real-space DFT code) vs SPARC

# Using GPAW
from gpaw import GPAW
calc = GPAW(xc="PBE", kpts=(9, 9, 9), h=0.25, directory="vasp-calc", convergence={"energy": 1.e-4})

vs

# Using SPARC
from sparc.calculator import SPARC
calc = SPARC(xc="PBE", kpts=(9, 9, 9), h=0.25, directory="sparc-calc.sparc", convergence={"energy": 1.e-4})

Advanced Usage: SPARC-X-API as a Socket Interface

Disclaimer: The socket communication feature in SPARC and SPARC-X-API are experimental and subject to changes until the release of v2.0 SPARC-X-API.

Overview

Experienced users can harness the power of SPARC and SPARC-X-API's socket communication layer to build efficient and flexible computational workflows. By integrating a socket communication interface directly into SPARC, users can significantly reduce the overhead typically associated with file I/O during calculation restarts. This feature is particularly beneficial for tasks involving repetitive operations like structural optimization and saddle point searches, where traditional file-based communication can become a bottleneck. The underlying software architecture is shown in Fig. 3:

Fig. 3. SPARC electronic calculations with socket communication across hybrid computing platforms

Requirements: the SPARC binary must be manually compiled from the source code with socket support and with the USE_SOCKET=1 flag enabled (see the installation instructions.

Usage

The socket communication layer in SPARC and SPARC-X-API are designed for:

• Efficiency: Eliminates the need for intermediate file I/O, directly streaming data between processes.
• Speed: Enhances the performance of iterative calculations, crucial for large-scale simulations.
• Flexibility: Allows dynamic modification of calculation parameters without the need to restart the process.

The communication protocol implemented in SPARC and SPARC-X-API adheres to the i-PI protocol standard. Specifically, we implement the original i-PI protocol within the SPARC C-source code, while the python SPARC-X-API uses a backward-compatible protocol based on i-PI. The dual-mode design is aimed for both low-level and high-level interfacing of the DFT codes, providing the following features as shown in Fig. 4:

Fig. 4. Overview of the SPARC protocol as an extension to the standard i-PI protocol.

Based on the scenarios, the socket communication layer can be accessed via the following approaches as shown in Fig. 5:

Fig. 5. Different ways of using SPARC's socket mode.

1. SPARC binary only (Fig. 5 a)

   SPARC binary with socket support can be readily coupled with any i-PI compatible socker server, such as ase.calculators.socketio.SocketIOCalculator, for example

   from ase.calculators.socketio import SocketIOCalculator
   from subprocess import Popen
   calc = SocketIOCalculator(port=31415)
   with calc:
       # Start sparc at background
       process = Popen("mpirun -n 8 sparc -name SPARC -socket localhost:31415", shell=True)
       # Single point calculations
       process.kill()

   The end user is responsible for generating the input files and making sure the same atoms structures are used by SocketIOCalculator and the SPARC binary. The mode is also limited to be run on a single computer system.

2. Local-only Mode (Fig. 5 b)

   Ideal for standalone calculations, this mode simulates a conventional calculator while benefiting from socket-based efficiency.

   with SPARC(use_socket=True, **normal_parameters) as calc:
       # Execute single-point calculations

   For most users we recommend using this mode when performing a calculation on a single HPC node.

3. Client (Relay) Mode (Fig. 5 c)

   In this mode, the sparc.SPARC calculator servers as a passive client which listens to a remote i-PI-compatible server. When messages are received on the client side, it relays the relevant message to a local SPARC binary and send results back through the socket pipe. The server side can either be a normal i-PI compatible server (such as SocketIOCalculator) or server-mode sparc.SPARC (see 4).

   Start the client by:

   client = SPARC(use_socket=True,
                  socket_params=dict(host="host.address.com", port=31415))
   with client:
       client.run()

   Or via Command-Line:

   python -m sparc.client -s host:port

   Note: when running SPARC-X-API as a socket client, the atoms object can be ommitted (is the server also runs the SPARC protocol). When new atoms positions and parameters arrive, the client will automatically determine if it is necessary to restart the SPARC subprocess.

4. Server Mode (Fig. 5 d)

   Paired with the client mode in (3), SPARC-X-API can be run as a socket server, isolated from the node that performs the computation. This can be useful for highly-distributed computational workflows.

   On the server node, run:

   server_calc = SPARC(use_socket=True, socket_params=dict(port=31415, server_only=True), **normal_parameters)
   with server_calc:
       # Execute single point calculations for atoms_1
       # Execute single point calculations for atoms_2

   In this case, the server will opens 0.0.0.0:31415 for connection. Make sure your server is directly accessible from the clients and the port is not occupied. The socker server is capable of receiving raw_results directly from the clients, making it possible to access server_calc.raw_results without access to the file systems on the client side.

(In-progress) Controlling SPARC routines from socket interface

As shown in Fig. 4, the SPARC socket protocol designs allows bidirectional control of internal SPARC routines. Local- or server-mode sparc.SPARC calculators can communicate with the SPARC binary via functions like set_params. This can be useful for applications like on-the-fly force field learning, electron density fitting, setting up boundary conditions etc. Applications will be updated in both SPARC and SPARC-X-API repositories.

Troubleshooting

Please refer to the troubleshooting guidelines

Advanced topics

A detailed description about how the API works can be found here

API changes

The API changes compared to the older release (v0.1) are summarized here

How to contribute

Please refer to our guidelines for contributors