Condensed Matter > Materials Science
[Submitted on 25 Mar 2024 (this version), latest version 7 Jun 2024 (v2)]
Title:Unified Differentiable Learning of the Electric Enthalpy and Dielectric Properties with Exact Physical Constraints
View PDF HTML (experimental)Abstract:The modern theory of polarization and electric enthalpy functionals have paved the way to first-principles studies of polarization in crystalline, disordered and liquid materials. However, calculating the polarization in large-scale ab-initio molecular dynamics remains prohibitive due to the high computational cost involved. In this work, we introduce a machine-learning formulation that simultaneously learns to predict electric enthalpy, atomic forces, polarization, Born charges, and polarizability for a given atomic configuration. These quantities are derived through differentiation of the electric enthalpy with respect to atomic positions and electric field within a local equivariant representation of the atomic environment. Our new method exactly constrains the polarization to be a conservative vector field, obeys the acoustic sum rule for Born charges, and ensures the conservation of electric enthalpy in machine-learning molecular dynamics. Importantly, the formulation is general and does not necessarily require equivariance. The locality of the model enables the efficient calculation of vibrational and dielectric properties of materials over large length-scale and long time-scale molecular dynamics at quantum-mechanical accuracy. We apply our approach to determine the infrared spectrum and frequency-dependent dielectric constant of $\alpha$-SiO$_2$ using molecular dynamics, finding excellent agreement with results obtained from density functional perturbation theory. Additionally, we show that our model accurately captures the electric-field contributions to the electric enthalpy, forces, and polarization in reference to density functional theory calculations. This demonstrates the capability of our formulation to perform large-scale molecular dynamics under electric fields, thus paving the way to efficient and accurate studies of dielectric response of complex materials.
Submission history
From: Stefano Falletta [view email][v1] Mon, 25 Mar 2024 21:35:59 UTC (262 KB)
[v2] Fri, 7 Jun 2024 20:08:25 UTC (3,373 KB)
Current browse context:
cond-mat.mtrl-sci
Change to browse by:
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
IArxiv Recommender
(What is IArxiv?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.