🔬 materials science

Nonadiabatic molecular dynamics simulation with external field

This paper introduces CPA-NAMD-extF, a first-principles density-matrix framework that extends nonadiabatic molecular dynamics to simulate field-driven electronic responses in large, complex systems by combining explicit external-field coupling with self-consistent density updates, demonstrating its accuracy and utility in analyzing opto-currents and vortex currents in materials like SiO $_2$ and hybrid perovskites.

Original authors: Chengji Li, Fan Zheng

Published 2026-11-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Chengji Li, Fan Zheng

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Light does more than just illuminate the world; when it strikes certain materials, it can push electrons into motion, creating electric currents without the need for batteries or wires. This phenomenon, known as the photovoltaic effect, is the foundation of solar panels, but scientists are also hunting for a more exotic version called the "shift current." In this effect, the very shape of the material's internal structure acts like a ratchet, shoving electrons in a specific direction when hit by light. To understand how this works, researchers must track the frantic, quantum-mechanical dance of electrons as they respond to light, all while the atoms they are bound to vibrate with heat. The challenge has been that simulating this dance for large, complex materials is incredibly difficult, especially when trying to account for the jiggling of atoms at room temperature.

A team of researchers at ShanghaiTech University has developed a new computational tool to solve this problem, allowing them to watch these electron currents flow in real time within massive, complex structures. Their method, a refined version of a technique called nonadiabatic molecular dynamics, bridges a gap that previously existed between two powerful but limited approaches. One approach, which tracks electrons moving along a pre-calculated path of vibrating atoms, is fast but could not handle the push of an external electric field. The other, which simulates the field's effect directly, is accurate but becomes impossibly slow and expensive for systems containing hundreds of atoms. The new method combines the speed of the first with the field-handling ability of the second, creating a framework that can simulate how light drives currents in materials with over 500 atoms, a scale that was previously out of reach for such detailed, field-driven calculations.

The researchers tested their new tool on two very different types of materials to prove it worked. First, they looked at silicon dioxide, the main component of glass and quartz, in both its ordered crystalline form and its disordered amorphous form. They simulated a block of this material containing 576 atoms, a size that would typically require days of supercomputer time to analyze with traditional methods. Their new approach produced the results in just a few hours. The simulations confirmed that the disordered glass absorbed light differently than the ordered crystal, showing a smoother, broader absorption spectrum with a distinct tail at lower energies, matching what experimentalists have observed in the real world. This success demonstrated that the tool could handle the structural chaos of real-world materials while accurately predicting how they interact with light.

The team then turned their attention to a more complex material: a two-dimensional hybrid perovskite, a type of crystal made of organic and inorganic parts that is being studied for next-generation electronics. This material is interesting because it can switch between a polar, ferroelectric state and a non-polar, paraelectric state depending on the temperature. The researchers simulated the material at two temperatures: 300 Kelvin, which is just below the point where it switches states, and 450 Kelvin, which is just above it. They found that the ability of the material to generate a shift current was not a fixed property but was deeply sensitive to the thermal vibrations of its atoms. At the lower temperature, the material consistently generated a current in one direction. However, at the higher temperature, the current fluctuated wildly, sometimes pointing one way and sometimes the other, eventually averaging out to nearly zero. This revealed that a single, static picture of the material's structure is insufficient to predict its behavior; one must account for the constant, thermal shuffling of atoms to understand how the material will perform in a real device.

Digging deeper into the microscopic origins of these currents, the researchers mapped the flow of electrons in real space, visualizing exactly where the current moved within the crystal lattice. They discovered a surprising pattern: the current did not flow in a straight line but formed swirling vortices centered around bromine atoms. These tiny, localized whirlpools of electricity were present in both the ordered and disordered phases of the material. By breaking down the simulation data, they traced the cause of these vortices to a specific interaction between the electron orbitals of the bromine atoms. It turns out that the electrons are moving in circles because of a coupling between the atom's spherical inner electron shells and its dumbbell-shaped outer shells. This specific orbital arrangement acts as a microscopic engine, generating the swirling currents that were previously difficult to pinpoint.

The implications of this work extend beyond just understanding one specific material. By providing a way to simulate how light drives currents in large, hot, and structurally complex systems, this new method opens the door to designing better materials for solar energy and ultra-fast electronics. It allows scientists to move beyond idealized, perfect crystals and study materials as they truly exist: jiggling, disordered, and full of thermal noise. The researchers showed that these thermal fluctuations are not just background noise to be ignored; they are active participants that can enhance, suppress, or even cancel out the very currents we hope to harness. With this tool, the path forward involves simulating these complex, real-world conditions to find materials that can reliably generate electricity from light, even as their atoms dance with heat.

Technical Summary: CPA-NAMD-extF for Nonadiabatic Molecular Dynamics with External Fields

Problem Statement
Nonadiabatic molecular dynamics (NAMD) simulations utilizing the classical path approximation (CPA-NAMD) are widely used to study excited carrier dynamics in extended systems. However, standard CPA-NAMD implementations rely on Hamiltonians derived from precomputed ground-state calculations, explicitly neglecting the modification of the electronic potential induced by external fields. Consequently, these methods cannot simulate carrier dynamics driven by external fields, such as field-induced currents or optical responses. While real-time time-dependent density-functional theory (rt-TDDFT) offers a direct first-principles approach for such problems, it suffers from prohibitive computational costs for large systems (hundreds of atoms) and long propagation times, particularly when finite-temperature sampling via multiple Born–Oppenheimer molecular dynamics (BOMD) trajectories is required.

Methodology: CPA-NAMD-extF
To address these limitations, the authors develop CPA-NAMD-extF, a first-principles density-matrix framework designed to calculate field-driven electronic responses along precomputed finite-temperature ionic trajectories. The method retains the computational efficiency of the classical-path approximation while explicitly incorporating external-field coupling.

Key technical features include:

Self-Consistent Propagation: The electronic density matrix is propagated under an external field (introduced in the velocity gauge via a time-dependent vector potential $\mathbf{A}(t)$ ). The Hartree and exchange–correlation potentials are updated self-consistently at each electronic time step based on the propagated density, replacing the static ground-state contribution.
Active-Inactive Subspace Truncation: To reduce cost, the adiabatic state space is divided into an active subspace (where population changes occur) and an inactive subspace (deep valence states). The density matrix in the active subspace is propagated explicitly, while the inactive subspace occupations remain fixed to their ground-state values.
Coherence Handling: Field-induced coherences between active and inactive subspaces are evaluated instantaneously using a first-order adiabatic approximation at each electronic step. These coherences contribute to the current but are excluded from the self-consistent density evolution.
Finite-Band Correction: To mitigate errors arising from the truncation of high-energy unoccupied states in the velocity gauge, a first-order adiabatic correction is applied to the current calculation.
Real-Space Current Density: The framework allows for the construction of real-space current density $\mathbf{j}(\mathbf{r}, t)$ by combining paramagnetic, diamagnetic, and finite-band correction contributions on the DFT FFT grid.
Implementation: The method is implemented as a post-processing module interfacing with the Quantum ESPRESSO (QE) package, utilizing hybrid MPI/OpenMP parallelization. It reuses ground-state BOMD data (Hamiltonians, overlaps, velocities) for various field amplitudes and polarizations.

Key Results
The authors validate and apply CPA-NAMD-extF to several systems:

Benchmarks: The method was benchmarked against existing rt-TDDFT codes (PWmat and SALMON). Results for time-dependent dipoles in small molecules and 2D materials, finite-field currents in monolayer MoS $_2$ , and linear optical absorption in crystalline silicon showed close agreement with reference calculations.
Large-Scale Linear Response: The method was applied to calculate the linear absorption spectra of crystalline and amorphous SiO $_2$ (576 atoms, 3072 valence electrons) at 300 K. The results reproduced experimental trends (e.g., broader features in amorphous structures) and were obtained in hours on CPU cores, a task that would require days on multiple GPUs using conventional rt-TDDFT.
Nonlinear Response in Perovskites: The framework was used to investigate the second-order opto-current (shift current) response of the 2D hybrid perovskite (BA) $_2$ $_{2}$ CsPb $_2$ $_{2}$ Br $_7$ $_{7}$ at 300 K (ferroelectric) and 450 K (paraelectric).
- Finite-temperature calculations revealed that the second-order response is strongly affected by structural fluctuations. While the 300 K trajectory showed a stable non-zero mean response, the 450 K trajectory exhibited frequent sign changes, leading to a vanishing macroscopic average.
- Real-space analysis of the DC current density revealed pronounced Br-centered vortex currents in both phases.
- Microscopic decomposition identified that these vortices arise primarily from the coupling between Br $p$ -orbital character states and Br $s$ -orbital character intermediate states. A minimal hydrogenic model confirmed that such $s$ – $p$ orbital coupling can generate atom-centered circulating current patterns.

Significance and Claims
The paper claims that CPA-NAMD-extF provides an efficient framework for simulating charge density evolution under external fields within the CPA-NAMD paradigm. Its primary significance lies in enabling the study of field-induced optical and nonlinear responses in large and structurally complex systems (hundreds of atoms) at finite temperatures, which are computationally prohibitive for standard rt-TDDFT. The method successfully captures the critical role of thermal fluctuations in nonlinear responses and offers a pathway to investigate microscopic current mechanisms, such as the Br-centered vortex currents identified in the perovskite study, without the extreme cost of full real-time propagation of the entire system. The authors note that the method is most reliable in the relatively weak field regime, as strong fields may require basis sets (like the Houston basis) that are computationally expensive for large systems.

More like this