Computer Physics Communications最新文献

Efficient simulation tools are crucial for studying complex systems such as reacting flows where computational costs of computing chemical reaction rates can vastly exceed the costs for the integration of the convective and diffusive transport terms. Load imbalance in parallel computing poses a significant challenge for massively parallel reacting flow simulations. In response, a novel load balancing library has been developed to enhance OpenFOAM's solver performance in parallel environments. This library seamlessly integrates with OpenFOAM, offering ease of use and applicability to any OpenFOAM reacting solver incorporating finite-rate chemistry. In addition, it supports the standard and the dynamic adaptive chemistry model (TDAC) of OpenFOAM.

The newly developed load-balanced standard and TDAC models address significant load imbalances by exchanging information between processes via MPI calls and tracking ODE solution times on a cell level. The TDAC model introduces dual tables on each core and enables immediate addition of computed solutions, enhancing computational efficiency. Validation on various test cases, including simulations on the HLRS Hawk supercomputer with up to 8000 cores, confirms identical results compared to the original unbalanced models, with notable speed-up factors of up to 6 for the standard and 5 for the TDAC model. Despite non-linear scaling at lower cell count per processor, load-balanced models consistently outperform unbalanced counterparts, making them the preferred choice for reacting flow simulations in OpenFOAM.

The paper focuses on the optimization of the FEM matrix kernels with respect to user-defined parameters such as materials, initial conditions, and boundary conditions that are known during run-time only. Adapting the kernels to specific parameters can save a significant amount of execution time and increase performance. Handling them efficiently is challenging due to the exponential number of potential combinations that the user can specify.

The paper presents an approach that combines (a) cross-element vectorization for the easy-to-write transformation of the original scalar code to vectorized one, (b) meta-programming for utilization of a compiler for building sub-kernels tailored for a particular set of parameters, (c) and dynamic polymorphism allowing run-time selection of sub-kernels.

We show that the above techniques allow (1) straightforward code modifications, (2) efficient handling of required dynamic behavior with a minor performance penalty for most kernels, and (3) achieving up to 8-fold speedups compared to non-adapted kernels.

α decay and proton emission were combined using symbolic regression to improve the accuracy of predicting their respective half-lives using two theoretical formulations: (1) adjustable parameters for the universal decay law and universal decay law for proton emission are obtained by regressions with fully-constrained symbolic regressions. (2) New theoretical formulas for calculating the half-life of proton emission and α decay are obtained using unconstrained symbolic regressions combined with nuclear data. Our computational analysis indicates that fully-constrained symbolic regressions and unconstrained symbolic regressions are reliable for specific and general nuclei, respectively, in terms of replicating experimental results, and are sufficiently robust to produce accurate half-life predictions. Unlike other machine learning methods that generate complex and opaque results, our approach integrates physics and machine learning to create interpretable formulas that provide intuitive parametric outcomes and transparent and dependable inferences of half-lives, even in areas with limited experimental data. The test results show that the equation yields accurate results, and can be easily applied to future α decay and proton emission studies.

The paper presents a new one-dimensional drift-kinetic electrostatic model based on the particle-in-cell method and capable of simulating the processes of plasma heating and confinement in mirror traps. The most of particle and energy losses in these traps occur along the magnetic field lines. The key role in limiting these losses is played by the ambipolar electric potential which creates a potential barrier for electrons and significantly reduces the heat flux that, without this barrier, would go to wall due to the classical electron thermal conductivity. However, modeling the formation of such a potential on real spatial and temporal scales of experiments is a challenging problem, since it requires a detailed description of not only ion, but also electron kinetics. In this work, we propose to solve the problem of taking into account electron kinetic effects on the time scale of plasma confinement in a mirror trap using the particle-in-cell method adapted to the approximate drift-kinetic equations of plasma motion. Unlike other electrostatic particle-in-cell models, which use fully implicit schemes to solve the nonlinear system of Vlasov-Poisson and Vlasov-Ampere equations, we propose a semi-implicit approach. By analogy with the Energy Conserving Semi-Implicit Method (ECSIM), it allows for precise conservation of energy and reduces the procedure for finding the electric field to inverting a tridiagonal matrix without multiple nonlinear iterations. Such a model will be useful for simulating not only collisional losses of hot plasma in fusion experiments, but also for studying the features of creating cold starting plasma in mirror traps using plasma or electron guns.

Predicting the phase diagram of solid solutions is crucial for understanding their physical behavior under various conditions and at different compositions. However, conventional sampling methods face challenges in efficiently addressing configurational and vibrational disorder for substitutional solid solutions. Additionally, a persistent obstacle has been the lack of a robust theoretical approach for determining the free energy of interstitial solid solutions characterized by the fluid-like diffusion of interstitial species. The method presented in this paper overcomes both hurdles by coupling thermodynamic integration with hybrid Monte Carlo algorithms. We validate the accuracy of the method by computing the free energies of iron alloys described by a Lennard-Jones potential. We also showcase its efficiency by determining the phase diagram of the MgO-CaO system, described by a machine-learning interatomic potential. The high efficiencies achieved with this method pave the way to the determination of the free energies of solid solutions with ab initio accuracy.

The dynamic mode decomposition (DMD) method has attracted widespread attention as a representative modal-decomposition method and can build a predictive model. However, DMD may give predicted results that deviate from physical reality in some scenarios, such as dealing with translation problems or noisy data. Here, we propose a physics-constrained DMD (PCDMD) method to address this issue. The proposed PCDMD method first employs a data-driven model using DMD, then calculates the residual of the physical equations, and finally corrects the predicted results using Kalman filter and gain coefficients. In this way, the PCDMD method can integrate the physics-informed equations with the data-driven model generated by DMD. Numerical experiments are conducted using PCDMD, including the Allen–Cahn, advection-diffusion, Burgers' equations and lid-driven cavity flow. The results demonstrate that the proposed PCDMD method can reduce the reconstruction and prediction errors by 1%-10% by incorporating physical constraints. Regarding noisy datasets and imperfect physical constraints, PCDMD can still ensure that the predicted results satisfy the physical constraints, thereby reducing errors.

Program summary

Program title: PCDMD

Dataset link: https://github.com/YinYuhuiTJU/PCDMD

Licensing provisions: MIT

Programming language: Python

Version 13 of XtalOpt, an evolutionary algorithm for crystal structure prediction, is now available for download from the CPC program library or the XtalOpt website, https://xtalopt.github.io. In the new version of the XtalOpt code, a general platform for multi-objective global optimization is implemented. This functionality is designed to facilitate the search for (meta)stable phases of functional materials through minimization of the enthalpy of a crystalline system coupled with the simultaneous optimization of any desired properties that are specified by the user. The code is also able to perform a constrained search by filtering the parent pool of structures based on a user-specified feature, while optimizing multiple objectives. Here, we present the implementation and various technical details, and we provide a brief overview of additional improvements that have been introduced in the new version of XtalOpt.

Program summary

Program Title: XtalOpt

CPC Library link to program files: https://doi.org/10.17632/jt5pvnnm39.4

Developer's repository link: https://github.com/xtalopt/XtalOpt

Licensing provisions: BSD 3-clause

Programming language: C++.

Journal reference of previous version: Comput. Phys. Commun. 237 (2019) 274–275.

Does the new version supersede the previous version?: Yes.

Reasons for the new version: Implementation of a multi-objective evolutionary search within the XtalOpt program package.

Summary of revisions: Implemented a general user-friendly multi-objective search capability, made various improvements to user interface and functionalities, performed bug fixes.

Nature of problem: The XtalOpt algorithm is designed to search for (meta)stable crystal structures, optionally with specific functionalities – a grand challenge in computational materials science, chemistry and physics.

Solution method: A generalized scalar fitness function, where a set of user-specified objectives contribute to the fitness value for candidate structures, is implemented within XtalOpt. This generalized fitness biases the search towards the discovery of (meta)stable phases with structural motifs that are key for the desired characteristics. As a result, the evolutionary search explores regions of the energy landscape of higher relevance in terms of target properties.

A new knowledge-based and machine learning hybrid modeling approach, called conditional Gaussian neural stochastic differential equation (CGNSDE), is developed to facilitate modeling complex dynamical systems and implementing analytic formulae of the associated data assimilation (DA). In contrast to the standard neural network predictive models, the CGNSDE is designed to effectively tackle both forward prediction tasks and inverse state estimation problems. The CGNSDE starts by exploiting a systematic causal inference via information theory to build a simple knowledge-based nonlinear model that nevertheless captures as much explainable physics as possible. Then, neural networks are supplemented to the knowledge-based model in a specific way, which not only characterizes the remaining features that are challenging to model with simple forms but also advances the use of analytic formulae to efficiently compute the nonlinear DA solution. These analytic formulae are used as an additional computationally affordable loss to train the neural networks that directly improve the DA accuracy. This DA loss function promotes the CGNSDE to capture the interactions between state variables and thus advances its modeling skills. With the DA loss, the CGNSDE is more capable of estimating extreme events and quantifying the associated uncertainty. Furthermore, crucial physical properties in many complex systems, such as the translate-invariant local dependence of state variables, can significantly simplify the neural network structures and facilitate the CGNSDE to be applied to high-dimensional systems. Numerical experiments based on chaotic systems with intermittency and strong non-Gaussian features indicate that the CGNSDE outperforms knowledge-based regression models, and the DA loss further enhances the modeling skills of the CGNSDE.

Astrophysical turbulent flows display an intrinsically multi-scale nature, making their numerical simulation and the subsequent analyses of simulated data a complex problem. In particular, two fundamental steps in the study of turbulent velocity fields are the Helmholtz-Hodge decomposition (compressive+solenoidal; HHD) and the Reynolds decomposition (bulk+turbulent; RD). These problems are relatively simple to perform numerically for uniformly-sampled data, such as the one emerging from Eulerian, fix-grid simulations; but their computation is remarkably more complex in the case of non-uniformly sampled data, such as the one stemming from particle-based or meshless simulations. In this paper, we describe, implement and test vortex-p, a publicly available tool evolved from the vortex code, to perform both these decompositions upon the velocity fields of particle-based simulations, either from smoothed particle hydrodynamics (SPH), moving-mesh or meshless codes. The algorithm relies on the creation of an ad-hoc adaptive mesh refinement (AMR) set of grids, on which the input velocity field is represented. HHD is then addressed by means of elliptic solvers, while for the RD we adapt an iterative, multi-scale filter. We perform a series of idealised tests to assess the accuracy, convergence and scaling of the code. Finally, we present some applications of the code to various SPH and meshless finite-mass (MFM) simulations of galaxy clusters performed with OpenGadget3, with different resolutions and physics, to showcase the capabilities of the code.

In this article, we propose a modified Allen–Cahn (AC) equation with a space-dependent interfacial parameter. When numerically solving the AC equation with a constant interfacial parameter over large domains, a substantial number of grid points are essential, which leads to significant computational costs. To effectively resolve this problem, numerous adaptive mesh techniques have been developed and implemented. These methods use locally refined meshes that adaptively track the interfacial positions of the phase field throughout the simulation. However, the data structures for adaptive algorithms are generally complex, and the problems to be solved may involve challenges at multiple scales. In this article, we present a modified AC equation with a mesh size-dependent interfacial parameter on a triangular mesh to efficiently solve multi-scale problems. In the proposed method, a triangular mesh is used, and the interfacial parameter value at a node point is defined as a function of the average length of the edges connected to the node point. The proposed algorithm effectively uses large and small values of the interfacial parameter on coarse and fine meshes, respectively. To demonstrate the efficiency and superior performance of the proposed method, we conduct several representative numerical experiments. The computational results indicate that the proposed interfacial function can adequately evolve the multi-scale phase interfaces without excessive relaxation or freezing of the interfaces. Finally, we provide the main source code for the methodology, including mesh generation as described in this paper, so that interested readers can use it.