Accurate and efficient protocols for high-throughput first-principles materials simulations

Gabriel de Miranda Nascimento, Flaviano José dos Santos, Marnik Bercx, Davide Grassano, Giovanni Pizzi, Nicola Marzari · 2025

Read Accurate and efficient protocols for high-throughput first-principles materials simulations by Gabriel de Miranda Nascimento, Flaviano José dos Santos, Marnik Bercx, Davide Grassano, Giovanni Pizzi, Nicola Marzari (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-03962.txt.

Introduction

Gabriel de Miranda Nascimento and a team of co-authors set protocols for large-scale first-principles materials simulations, in two thousand twenty-five. Such simulations can discover new materials, but each run needs settings that balance accuracy against computing cost. Picking those settings by hand does not scale to thousands of materials.

The team built strict criteria to estimate errors in energies, forces, and other outputs as settings vary. Their checks weigh the trade-off between precision and speed across many crystal types. The resulting standard protocols pick optimized settings for a chosen balance. They ship as open tools, from input generators to full workflow systems, for use by other groups.

Catalog record

CREATORGabriel de Miranda Nascimento, Flaviano José dos Santos, Marnik Bercx, Davide Grassano, Giovanni Pizzi, Nicola Marzari
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE54 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-03962

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Accurate and efficient protocols for high-throughput first-principles materials simulations Gabriel de Miranda Nascimento, Flaviano José dos Santos, Marnik Bercx, Davide Grassano, Giovanni Pizzi, Nicola Marzari Abstract Advancements in theoretical and algorithmic approaches, workflow engines, and an ever-increasing computational power have enabled a novel paradigm for materials discovery through first-principles high-throughput simulations. A major challenge in these efforts is to automate the selection of parameters used by simulation codes to deliver numerical precision and computational efficiency. Here, we propose a rigorous methodology to assess the quality of self-consistent DFT calculations with respect to smearing and k k -point sampling across a wide range of crystalline materials. For this goal, we develop criteria to reliably estimate average errors on total energies, forces, and other properties as a function of the desired computational efficiency, while consistently controlling k k -point sampling errors. The present results provide automated protocols (named standard solid-state protocols or SSSP) for selecting optimized parameters based on different choices of precision and efficiency tradeoffs. These are available through open-source tools that range from interactive input generators for DFT codes to high-throughput workflows. Introduction Density-functional theory (DFT) has been a very successful approach to address the ground state of many-electron systems and has evolved into a predictive tool for materials discovery and design Giustino 2014 ; Marzari et al. 2021 . The predictive power of the theory, the robust implementations provided by DFT codes, and increasing available computational power in the past years have unlocked the potential of high-throughput approaches for materials discovery Jain et al. 2011 ; Mounet et al. 2018 ; Merchant et al. 2023 , where automated computational workflows are leveraged for the study and simulation of a large number of compounds. Starting from experimentally known crystal structures, these efforts are responsible for the construction and curation of numerous computational databases of materials Jain et al. 2013 ; Saal et al. 2013 ; Curtarolo et al. 2012 ; Haastrup et al. 2018 ; Schmidt et al. , where several properties of many known compounds are addressed with various levels of accuracy. Moreover, data for these approaches serve as a starting point for strategies that aim at generating completely novel crystal structures, with approaches that range from random structure searching Pickard and Needs 2011 , structural prototyping Haastrup et al. 2018 to genetic algorithms d’Avezac et al. 2012 ; Jennings et al. 2019 and machine-learning-powered generative models Xie et al. 2021 ; Zeni et al. 2023 . To properly manage these efforts on high-throughput calculations and support their scalability needs, workflow engines have been developed and coevolved with DFT codes. They provide different levels of functionality that are suited for multiple use cases, ranging from input generation/output parsing and managing input/output operations, to interfacing with computational facilities and schedulers. Examples include materials APIs such as pymatgen Ong et al. 2013 and ASE Larsen et al. 2017 , and workflow managers such as FireWorks Jain et al. 2015 , Pyiron Janssen et al. , Atomic Simulation Recipes Gjerding et al. 2021 and AiiDA Pizzi et al. 2016 ; Huber et al. 2020 ; Uhrin et al. 2021a . In such frameworks, one of the main challenges in designing robust and efficient high-throughput workflows is achieving a reliable balance between precision and computational cost. While carefully controlling the inner parameters of each method to achieve the most precise results for a given level of theory has been a standard practice in computational materials modeling since its origins, large-scale computational workflows also demand a stronger emphasis on computational efficiency, since they involve workloads that may scale to tens of thousands of materials and hundreds of thousands of DFT calculations. Optimizing the methods and inner parameters in these calculations such that they run as efficiently as possible, and ensuring that no computational resource is wasted and not converted into precision gains is, therefore, an essential part of designing robust computational workflows for first-principles materials simulations. Remarkable efforts in this direction include the creation of standard solid-state pseudopotential libraries (SSSP) Prandini et al. 2018 , that established a testing protocol based on extensive DFT and density-functional perturbation theory (DFPT) calculations using the main pseudopotential libraries available Kucukbenli et al. 2014 ; Dal Corso 2014 ; Topsakal and Wentzcovitch 2014 ; Garrity et al. 2014 ; Schlipf and Gygi 2015 ; Van Setten et al. 2018 ; Willand et al. 2013 . Using these protocols, SSSP curates an exhaustive and open collection of extensively tested pseudopotentials and related parameters, suitable for different use cases of high-throughput and high-precision materials modeling. This work contributes to solutions to these challenges by introducing the concept of standard solid-state protocols (SSSP) as an extended framework to the SSSP pseudopotentials Prandini et al. 2018 . They are defined as a collection of parameters for specific codes and calculations, optimized for different levels of precision and computational efficiency and extensively tested for a wide range of materials. These operate within computational workflows, acting as an interface between their higher-level logic of operations and the inner-level details of parameters for a given calculation called by the workflow. More specifically, in this work we build upon the AiiDA workflow framework and its plugin and workflows for Quantum ESPRESSO Giannozzi et al. 2009 ; Giannozzi et al. 2017 ; Giannozzi et al. 2020 ; aii , a plane-wave pseudopotential DFT code, but the results are general and applicable to any DFT code. In these protocols, we start from the results obtained by Prandini et al. 2018 Prandini et al. 2018 for pseudopotentials and extend these to further benchmarks for the additional parameters that most strongly influence the precision and computational cost of DFT calculations for crystals: namely, k -point sampling and smearing techniques. For this reason, the SSSP acronym is extended to refer to protocols (and the pseudopotentials that they encompass). Similar work in this direction has been most notably performed by Choudhary and Tavazza 2019 Choudhary and Tavazza 2019 , where they employed a systematic procedure for benchmarking k -point grids and plane-wave energy cutoffs in calculations run with VASP Kresse and Furthmüller 1996a ; Kresse and Furthmüller 1996b , and has also been developed along the construction of materials databases, such as the Materials Project Jain et al. 2011 ; Jain et al. 2013 . However, we note that these works do not use smearing as a variable for optimization (employing a fixed value instead), missing out on the benefits that smearing techniques may bring to the convergence of Brillouin zone sampling, either by reducing the required computational cost for the calculations or improving their precision. In our work, we simultaneously consider k -point sampling and smearing temperature in the optimization process and combine these results with the guidelines from Prandini et al. 2018 Prandini et al. 2018 to create reliable sets of parameters that are suited for different tradeoffs of precision and computational time. These parameters are then extensively tested for a wide distribution of materials. In this work, we start by summarizing the general interplay between k -point sampling and smearing techniques targeted at improving the convergence and efficiency of DFT calculations. We then describe the automated workflows developed for benchmarking the choice of parameters and their effect on relevant

Read the complete text