AMaRaNTA

⚠️ Note: This is a preliminary guide and will be updated in the near future. For a complete reference, please see the preprint: https://arxiv.org/html/2509.20226v1.

Overview

AMaRaNTA — Automating Magnetic paRAmeters iN a Tensorial Approach — is a computational package that systematically automates Density Functional Theory calculations (specifically, the “four-state” energy-mapping method) to extract exchange and anisotropy parameters in 2D magnets.
AMaRaNTA is implemented in Python as a workflow within AiiDA (Automated Interactive Infrastructure and Database for Computational Science) and interfaces with the Vienna Ab-initio Simulation Package (VASP) DFT code.

Package structure

The repository includes the following main components:

• runscript.py: a launch script where users specify the structure of the material’s unit cell, pseudopotentials to be used and, if needed, custom options for batch submission scripts
• a Workchain_AmarantaTemplate file: the body of the code, containing the implementation of the AMaRaNTA Workchain in the AiiDA format

Two versions of such file are currently available — Version 1.0 (Workchain_AmarantaTemplate_UniMiB.py) and Version 1.1 (Workchain_AmarantaTemplate_18dec_debugged.py) — that differ in the way atoms are chosen to construct the first-, second- and third-nearest neighbour pairs.
In both implementations, the choice is made inside a temporary 5x5 multiple of the input unit cell.
In Version 1.0, the first atom of each pair (the "central atom" hereafter) is chosen as the closest to the bottom left corner of the 5x5 supercell. This is the original implementation, used to generate most of the dataset reported in the preprint. However, depending on lattice geometry and unit cell choice, some atoms belonging to a certain neighbour shell with respect to the central atom may fall outside the 5×5 region, causing the neighbour selection algorithm to fail.
To overcome this issue, in Version 1.1 the central atom is chosen to be the closest to the centre of the 5x5 supercell. This approach avoids the selection issue above but introduces a new limitation: the central atom necessarily lies in the 3x3 repetition of the original unit cell, or nearby. As a result, the search algorithm will not consider supercells for the four-state below the 3x3 size, which can be larger than necessary. While this does not affect the accuracy of exchange parameters (and can even improve precision), it may reduce computational efficiency.

Future Development

A Version 1.2 is currently under development to optimize atom selection while maintaining accuracy and efficiency.