Machine learning accelerated nonadiabatic dynamics simulations of materials with excitonic effects

The Journal of Chemical Physics, Journal Year: 2025, Volume and Issue: 162(2)

Published: Jan. 8, 2025

This study presents an efficient methodology for simulating nonadiabatic dynamics of complex materials with excitonic effects by integrating machine learning (ML) models simplified Tamm–Dancoff approximation (sTDA) calculations. By leveraging ML models, we accurately predict ground-state wavefunctions using unconverged Kohn–Sham (KS) Hamiltonians. These ML-predicted KS Hamiltonians are then employed sTDA-based excited-state calculations (sTDA/ML). The results demonstrate that energies, time-derivative couplings, and absorption spectra from sTDA/ML accurate enough compared those conventional density functional theory based sTDA (sTDA/DFT) Furthermore, sTDA/ML-based molecular simulations on two different systems, namely chloro-substituted silicon quantum dot monolayer black phosphorus, achieve more than 100 times speedup the linear response time-dependent DFT simulations. work highlights potential ML-accelerated studying complicated photoinduced large offering significant computational savings without compromising accuracy.

Language: Английский

---

Detailed Complementary Consistency: Wave Function Tells Particle How to Hop, Particle Tells Wave Function How to Collapse

The Journal of Physical Chemistry Letters, Journal Year: 2024, Volume and Issue: 15(26), P. 6771 - 6781

Published: June 24, 2024

In mixed quantum-classical dynamics, the quantum subsystem can have both wave function and particle-like descriptions. However, they may yield inconsistent results for expectation value of same physical quantity. We here propose a novel detailed complementary consistency (DCC) method based on principle internal consistency. Namely, along each trajectory tells particle how to hop, while collapse active states in ensemble. As benchmarked diverse array representative models with localized nonadiabatic couplings, DCC not only achieves fully consistent (i.e., identical populations calculated functions states) but also closely reproduces exact results. Due high performance, our new has great potential give accurate description general dynamics after further development.

Language: Английский

---

Advancing nonadiabatic molecular dynamics simulations in solids with E(3) equivariant deep neural hamiltonians

Nature Communications, Journal Year: 2025, Volume and Issue: 16(1)

Published: Feb. 27, 2025

Abstract Non-adiabatic molecular dynamics (NAMD) simulations have become an indispensable tool for investigating excited-state in solids. In this work, we propose a general framework, N 2 AMD (Neural-Network Non-Adiabatic Molecular Dynamics), which employs E(3)-equivariant deep neural Hamiltonian to boost the accuracy and efficiency of NAMD simulations. Distinct from conventional machine learning methods that predict key quantities NAMD, computes these directly with Hamiltonian, ensuring excellent accuracy, efficiency, consistency. not only achieves impressive performing at hybrid functional level within framework classical path approximation (CPA), but also demonstrates great potential predicting non-adiabatic coupling vectors suggests method go beyond CPA. Furthermore, generalizability enables seamless integration advanced techniques infrastructures. Taking several extensively investigated semiconductors as prototypical system, successfully simulate carrier recombination both pristine defective systems large scales where often significantly underestimates or even qualitatively incorrectly predicts lifetimes. This offers reliable efficient approach conducting accurate across various condensed materials.

Language: Английский

---

Nonadiabatic Molecular Dynamics with Subsystem Density Functional Theory: Application to Crystalline Pentacene

Journal of Physics Condensed Matter, Journal Year: 2024, Volume and Issue: 36(38), P. 385901 - 385901

Published: June 12, 2024

In this work, we report the development and assessment of nonadiabatic molecular dynamics approach with electronic structure calculations based on linearly scaling subsystem density functional method. The is implemented in an open-source embedded Quantum Espresso/Libra software specially designed for simulations extended systems. As proof applicability method to large condensed-matter systems, examine nonradiative relaxation excess excitation energy pentacene crystals simulation supercells containing more than 600 atoms. We find that increased structural disorder observed larger supercell models induces couplings states accelerates excited states. conduct a comparative analysis several quantum-classical trajectory surface hopping schemes, including two new methods proposed work (revised decoherence-induced instantaneous decoherence at frustrated hops). Most tested schemes suggest fast occurring timescales 0.7-2.0 ps range, but they significantly overestimate ground state recovery rates. Only modified simplified decay mixing yields notably slower 8-14 ps, inhibited recovery.

Language: Английский

---

Unsupervised Machine Learning in the Analysis of Nonadiabatic Molecular Dynamics Simulation

The Journal of Physical Chemistry Letters, Journal Year: 2024, Volume and Issue: unknown, P. 9601 - 9619

Published: Sept. 13, 2024

The all-atomic full-dimensional-level simulations of nonadiabatic molecular dynamics (NAMD) in large realistic systems has received high research interest recent years. However, such NAMD normally generate an enormous amount time-dependent high-dimensional data, leading to a significant challenge result analyses. Based on unsupervised machine learning (ML) methods, considerable efforts were devoted developing novel and easy-to-use analysis tools for the identification photoinduced reaction channels comprehensive understanding complicated motions simulations. Here, we tried survey advances this field, particularly focus how use ML methods analyze trajectory-based simulation results. Our purpose is offer discussion several essential components protocol, including selection construction descriptors, establishment analytical frameworks, their advantages limitations, persistent challenges.

Language: Английский

---

State Tracking in Nonadiabatic Molecular Dynamics Using Only Forces and Energies

The Journal of Physical Chemistry Letters, Journal Year: 2024, Volume and Issue: unknown, P. 11944 - 11953

Published: Nov. 22, 2024

A new algorithm for the identification of unavoided (trivial) crossings in nonadiabatic molecular dynamics calculations is reported. The approach does not require knowledge wave functions or function time overlaps and uses only information on state energies gradients. In addition, a simple phase consistency correction time-derivative couplings proposed situations which are available. performance two algorithms demonstrated using several crossing models. approaches work best systems with localized coupling regions but may have difficulties those extended coupling. It found that tracking alone sufficient producing correct population required.

Language: Английский

---

Machine Learning Mapping Approach for Computing Spin Relaxation Dynamics

The Journal of Physical Chemistry Letters, Journal Year: 2024, Volume and Issue: unknown, P. 153 - 162

Published: Dec. 21, 2024

In this work, a machine learning mapping approach for predicting the properties of atomistic systems is reported. Within approach, atomic orbital overlap, density, or Kohn-Sham (KS) Fock matrix elements obtained at low level theory such as extended tight-binding have been used input features to predict electric field gradient (EFG) tensors higher those with hybrid functionals. It shown that machine-learning-predicted EFG can be compute spin relaxation rates several ions in aqueous solutions. From only fraction data direct calculation, one quadrupolar isotropic good accuracy, achieving relative errors between about 2–8% different ions.

Language: Английский

---

Detailed Complementary Consistency: Wave Function Tells Particle How to Hop, Particle Tells Wave Function How to Collapse

Published: May 7, 2024

In mixed quantum-classical dynamics, the quantum subsystem can have both wave function and particle-like descriptions. However, they may yield inconsistent results for expectation value of same physical quantity. We here propose a novel detailed complementary consistency (DCC) method based on principle internal consistency. Namely, along each trajectory tells particle how to hop, while collapse active states in ensemble. As benchmarked diverse array representative models, DCC not only achieves fully consistent (i.e., identical populations calculated functions states), but also closely reproduces exact results. Due high performance, our new is promising accurate description general nonadiabatic dynamics.

Language: Английский

---

Detailed Complementary Consistency: Wave Function Tells Particle How to Hop, Particle Tells Wave Function How to Collapse

Published: June 13, 2024

In mixed quantum-classical dynamics, the quantum subsystem can have both wave function and particle-like descriptions. However, they may yield inconsistent results for expectation value of same physical quantity. We here propose a novel detailed complementary consistency (DCC) method based on principle internal consistency. Namely, along each trajectory tells particle how to hop, while collapse active states in ensemble. As benchmarked diverse array representative models with localized nonadiabatic couplings, DCC not only achieves fully consistent (i.e., identical populations calculated functions states), but also closely reproduces exact results. Due high performance, our new has great potential give accurate description general dynamics after further development.

Language: Английский

---

Multi-channel machine learning based nonlocal kinetic energy density functional for semiconductors

Electronic Structure, Journal Year: 2024, Volume and Issue: 6(4), P. 045006 - 045006

Published: Oct. 25, 2024

Abstract The recently proposed machine learning-based physically-constrained nonlocal (MPN) kinetic energy density functional (KEDF) can be used for simple metals and their alloys (Sun Chen 2024 Phys. Rev. B 109 115135). However, the MPN KEDF does not perform well semiconductors. Here we propose a multi-channel (CPN) KEDF, which extends to semiconductors by integrating information collected from multiple channels, with each channel featuring specific length scale in real space. CPN is systematically tested on silicon binary We find that design beneficial machine-learning-based models capturing characteristics of semiconductors, particularly handling covalent bonds. In particular, 5 utilizes five demonstrates excellent accuracy across all systems. These results offer new path generating KEDFs

Language: Английский

---