DeePMD-kit v2: A software package for deep potential models

The Journal of Chemical Physics, Год журнала: 2023, Номер 159(5)

Опубликована: Авг. 1, 2023

DeePMD-kit is a powerful open-source software package that facilitates molecular dynamics simulations using machine learning potentials known as Deep Potential (DP) models. This package, which was released in 2017, has been widely used the fields of physics, chemistry, biology, and material science for studying atomistic systems. The current version offers numerous advanced features, such DeepPot-SE, attention-based hybrid descriptors, ability to fit tensile properties, type embedding, model deviation, DP-range correction, DP long range, graphics processing unit support customized operators, compression, non-von Neumann dynamics, improved usability, including documentation, compiled binary packages, graphical user interfaces, application programming interfaces. article presents an overview major highlighting its features technical details. Additionally, this comprehensive procedure conducting representative application, benchmarks accuracy efficiency different models, discusses ongoing developments.

Язык: Английский

---

Classical and reactive molecular dynamics: Principles and applications in combustion and energy systems

Progress in Energy and Combustion Science, Год журнала: 2023, Номер 97, С. 101084 - 101084

Опубликована: Апрель 29, 2023

Molecular dynamics (MD) has evolved into a ubiquitous, versatile and powerful computational method for fundamental research in science branches such as biology, chemistry, biomedicine physics over the past 60 years. Powered by rapidly advanced supercomputing technologies recent decades, MD entered engineering domain first-principle predictive material properties, physicochemical processes, even design tool. Such developments have far-reaching consequences, are covered first time present paper, with focus on combustion energy systems encompassing topics like gas/liquid/solid fuel oxidation, pyrolysis, catalytic combustion, heterogeneous electrochemistry, nanoparticle synthesis, heat transfer, phase change, fluid mechanics. First, theoretical framework of methodology is described systemically, covering both classical reactive MD. The emphasis development force field (ReaxFF) MD, which enables chemical reactions to be simulated within framework, utilizing quantum chemistry calculations and/or experimental data training. Second, details numerical methods, boundary conditions, post-processing costs simulations provided. This followed critical review selected applications methods systems. It demonstrated that ReaxFF been successfully deployed gain insights pyrolysis oxidation fuels, revealing detailed changes pathways. Moreover, complex physico-chemical dynamic processes reactions, soot formation, flame synthesis nanoparticles made plainly visible from an atomistic perspective. Flow, transfer change phenomena also scrutinized simulations. Unprecedented nanoscale droplet collision, evaporation, CO2 capture storage under subcritical supercritical conditions examined at atomic level. Finally, outlook discussed context emerging computing platforms, machine learning multiscale modelling.

Язык: Английский

---

Neural network potentials for chemistry: concepts, applications and prospects

Digital Discovery, Год журнала: 2022, Номер 2(1), С. 28 - 58

Опубликована: Дек. 21, 2022

Artificial Neural Networks (NN) are already heavily involved in methods and applications for frequent tasks the field of computational chemistry such as representation potential energy surfaces (PES) spectroscopic predictions. This perspective provides an overview foundations neural network-based full-dimensional surfaces, their architectures, underlying concepts, to chemical systems. Methods data generation training procedures PES construction discussed means error assessment refinement through transfer learning presented. A selection recent results illustrates latest improvements regarding accuracy representations system size limitations dynamics simulations, but also NN application enabling direct prediction physical without simulations. The aim is provide current state-of-the-art approaches point out challenges enhancing reliability applicability on a larger scale.

Язык: Английский

---

Determining the mechanical and decomposition properties of high energetic materials (α-RDX, β-HMX, and ε-CL-20) using a neural network potential

Physical Chemistry Chemical Physics, Год журнала: 2024, Номер 26(13), С. 9984 - 9997

Опубликована: Янв. 1, 2024

NNP models covering three typical C/H/N/O element HEMs were developed to capture the mechanical and decomposition properties of RDX, HMX CL-20. The trajectory is mainly divided into two stages: pyrolysis oxidation.

Язык: Английский

---

Reaction Network of Ammonium Perchlorate (AP) Decomposition: The Missing Piece from Atomic Simulations

The Journal of Physical Chemistry C, Год журнала: 2023, Номер 127(27), С. 12976 - 12982

Опубликована: Июнь 29, 2023

The decomposition network of ammonium perchlorate (AP) is essential for combustion performance and safety solid propellants, while the detailed reaction pathway during thermolysis far from clear due to ultrafast complex reactions involved. Herein, we present direct atomic simulations AP thermal propose a fill missing piece in kinetic models by using neural model derived ab initio calculations. proton transfer dominant channel (NH4 + ClO4 → NH3 HClO4), which also observed previous mass spectra experiments. In addition, gas products play critical role promoting AP. For example, H abstraction OH found be decomposition. These provide insights into dynamics can extended investigate mechanism novel energetic materials.

Язык: Английский

---

The thermal decomposition mechanism of RDX/AP composites: ab initio neural network MD simulations

Physical Chemistry Chemical Physics, Год журнала: 2024, Номер 26(15), С. 11545 - 11557

Опубликована: Янв. 1, 2024

A neural network potential (NNP) is developed to investigate the decomposition mechanism of RDX, AP, and their composites.

Язык: Английский

---

Revealing the thermal decomposition mechanism of RDX crystals by a neural network potential

Physical Chemistry Chemical Physics, Год журнала: 2022, Номер 24(42), С. 25885 - 25894

Опубликована: Янв. 1, 2022

A neural network potential (NNP) is developed to investigate the complex reaction dynamics of 1,3,5-trinitro-1,3,5-triazine (RDX) thermal decomposition.

Язык: Английский

---

Unraveling pyrolysis mechanisms of lignin dimer model compounds: Neural network-based molecular dynamics simulation investigations

Fuel, Год журнала: 2023, Номер 357, С. 129909 - 129909

Опубликована: Сен. 23, 2023

Язык: Английский

---

Benchmark of general-purpose machine learning-based quantum mechanical method AIQM1 on reaction barrier heights

The Journal of Chemical Physics, Год журнала: 2023, Номер 158(7)

Опубликована: Янв. 30, 2023

Artificial intelligence-enhanced quantum mechanical method 1 (AIQM1) is a general-purpose that was shown to achieve high accuracy for many applications with speed close its baseline semiempirical (SQM) ODM2*. Here, we evaluate the hitherto unknown performance of out-of-the-box AIQM1 without any refitting reaction barrier heights on eight datasets, including total ∼24 thousand reactions. This evaluation shows AIQM1's strongly depends type transition state and ranges from excellent rotation barriers poor for, e.g., pericyclic clearly outperforms ODM2* and, even more so, popular universal potential, ANI-1ccx. Overall, however, largely remains similar SQM methods (and B3LYP/6-31G* most types) suggesting it desirable focus improving in future. We also show built-in uncertainty quantification helps identifying confident predictions. The predictions approaching level density functional theory types. Encouragingly, rather robust optimizations, reactions struggles most. Single-point calculations high-level AIQM1-optimized geometries can be used significantly improve heights, which cannot said method.

Язык: Английский

---

Challenges for Kinetics Predictions via Neural Network Potentials: A Wilkinson’s Catalyst Case

Molecules, Год журнала: 2023, Номер 28(11), С. 4477 - 4477

Опубликована: Май 31, 2023

Ab initio kinetic studies are important to understand and design novel chemical reactions. While the Artificial Force Induced Reaction (AFIR) method provides a convenient efficient framework for studies, accurate explorations of reaction path networks incur high computational costs. In this article, we investigating applicability Neural Network Potentials (NNP) accelerate such studies. For purpose, reporting theoretical study ethylene hydrogenation with transition metal complex inspired by Wilkinson's catalyst, using AFIR method. The resulting network was analyzed Generative Topographic Mapping network's geometries were then used train state-of-the-art NNP model, replace expensive ab calculations fast predictions during search. This procedure applied run first NNP-powered exploration We discovered that particularly challenging general purpose models, identified underlying limitations. addition, proposing overcome these challenges complementing models semiempirical predictions. proposed solution offers generally applicable framework, laying foundations further Machine Learning Fields, ultimately explore larger systems currently inaccessible.

Язык: Английский

---