Brute-force Simulations Research Articles

Limits of a single surrogate model development methodology to represent housing stocks

Surrogate models for predicting housing stock thermal comfort, and operational cost, energy or emissions have increasingly gained popularity due to the potential to identify solutions to accelerate and optimise construction. In addition to the potential opportunity to progress towards climate goals and increase resource efficiencies by identifying areas of interest, they are much less time consuming than brute-force simulations of millions of scenarios. Although they are popular due to their promise, analysis of scalability limits for surrogate model development methodologies have been left out. Without this analysis, assurance that these may be used over differing housing stock scales cannot be provided. This work uses black-box model interpretation as well as common predictive performance metrics to assess the scalability of the development methodology presented.The novelty of this work stems from its comparison of the performance of a development methodology through training a learning algorithm with a progressively varied dataset to illustrate its scalability. This validates the use of a single surrogate model to represent numerous bottom-up archetypes used to represent a housing stock.

Phase behavior and crystal nucleation of hard triangular prisms.

The interplay between densification and positional ordering during the process of crystal nucleation is a greatly investigated topic. Even for the simplest colloidal model-hard spheres-there has been much debate regarding the potential foreshadowing of nucleation by significant fluctuations in either local density or local structure. Considering anisotropic particles instead of spheres adds a third degree of freedom to the self-organization process of crystal nucleation: orientational ordering. Here, we investigate the crystal nucleation of hard triangular prisms. Using Monte Carlo simulations, we first carefully determine the crystal-fluid coexistence values and calculate the nucleation barriers for two degrees of supersaturation. Next, we use brute force simulations to obtain a large set of spontaneous nucleation events. By studying the time evolution of the local density, positional ordering, and orientational ordering in the region in which the nucleus first arises, we demonstrate that all local order parameters increase simultaneously from the very start of the nucleation process. We thus conclude that we observe no precursor for the crystal nucleation of hard triangular prisms.

Finding high-performance MOFs for effective SF6/N2 separation through high-throughput computational screening and machine learning

Abstract Given the rapidly expanding pool of synthesized and hypothetical metal–organic frameworks (MOFs), testing every single material for SF6/N2 separation by iterative experimental methods or computationally demanding molecular simulations is not practical. In this study, we integrated high-throughput computational screening and machine learning (ML) approaches to evaluate SF6/N2 mixture adsorption and separation performances of over 25 000 different types of synthesized and hypothetical MOFs (hypoMOFs), representing the largest set of structures studied for SF6/N2 separation to date. SF6/N2 mixture adsorption data that we produced for synthesized MOFs using molecular simulations were utilized to develop ML models to accurately and quickly predict SF6 and N2 uptakes, SF6/N2 selectivities, SF6 working capacities, adsorbent performance scores, and regenerabilities of both synthesized and hypoMOFs. Results showed the MOF space that we studied exhibits very high SF6/N2 selectivities in the range of 1.8–4204 at 1 bar in addition to high SF6 working capacities between 0.04–5.68 mol kg−1 at an adsorption pressure of 1 bar and desorption pressure of 0.1 bar at room temperature. The top-performing MOF adsorbents for SF6/N2 mixture separation were identified to have Zn, Cu, Ni metals; terphenyl, pyridine, naphthalene linkers; and medium pore sizes. Our comprehensive computational approach offers a highly efficient alternative to brute-force computer simulations by enabling the rapid assessment of the MOF adsorbents for SF6/N2 separation and provides molecular insights into the key structural features of the most promising adsorbents.

Bridging classical nucleation theory and molecular dynamics simulation for homogeneous ice nucleation.

Water freezing, initiated by ice nucleation, occurs widely in nature, ranging from cellular to global phenomena. Ice nucleation has been experimentally proven to require the formation of a critical ice nucleus, consistent with classical nucleation theory (CNT). However, the accuracy of CNT quantitative predictions of critical cluster sizes and nucleation rates has never been verified experimentally. In this study, we circumvent this difficulty by using molecular dynamics (MD) simulation. The physical properties of water/ice for CNT predictions, including density, chemical potential difference, and diffusion coefficient, are independently obtained using MD simulation, whereas the calculation of interfacial free energy is based on thermodynamic assumptions of CNT, including capillarity approximation among others. The CNT predictions are compared to the MD evaluations of brute-force simulations and forward flux sampling methods. We find that the CNT and MD predicted critical cluster sizes are consistent, and the CNT predicted nucleation rates are higher than the MD predicted values within three orders of magnitude. We also find that the ice crystallized from supercooled water is stacking-disordered ice with a stacking of cubic and hexagonal ices in four representative types of stacking. The prediction discrepancies in nucleation rate mainly arise from the stacking-disordered ice structure, the asphericity of ice cluster, the uncertainty of ice-water interfacial free energy, and the kinetic attachment rate. Our study establishes a relation between CNT and MD to predict homogeneous ice nucleation.

Surface-induced water crystallisation driven by precursors formed in negative pressure regions

Ice nucleation is a crucial process in nature and industries; however, the role of the free surface of water in this process remains unclear. To address this, we investigate the microscopic freezing process using brute-force molecular dynamics simulations. We discover that the free surface assists ice nucleation through an unexpected mechanism. The surface-induced negative pressure enhances the formation of local structures with a ring topology characteristic of Ice 0-like symmetry, promoting ice nucleation despite the symmetry differing from ordinary ice crystals. Unlike substrate-induced nucleation via water-solid interactions that occurs directly on the surface, this negative-pressure-induced mechanism promotes ice nucleation slightly inward the surface. Our findings provide a molecular-level understanding of the mechanism and pathway behind free-surface-induced ice formation, resolving the longstanding debate. The implications of our discoveries are of substantial importance in areas such as cloud formation, food technology, and other fields where ice nucleation plays a pivotal role.

A methodology of quantifying membrane permeability based on returning probability theory and molecular dynamics simulation.

We propose a theoretical approach to estimate the permeability coefficients of substrates (permeants) for crossing membranes from donor (D) phase to acceptor (A) phase by means of molecular dynamics (MD) simulation. A fundamental aspect of our approach involves reformulating the returning probability (RP) theory, a rigorous bimolecular reaction theory, to describe permeation phenomena. This reformulation relies on the parallelism between permeation and bimolecular reaction processes. In the present method, the permeability coefficient is represented in terms of the thermodynamic and kinetic quantities for the reactive (R) phase that exists within the inner region of a membrane. One can evaluate these quantities using multiple MD trajectories starting from phase R. We apply the RP theory to the permeation of ethanol and methylamine at different concentrations (infinitely dilute and 1 mol % conditions of permeants). Under the 1 mol% condition, the present method yields a larger permeability coefficient for ethanol (0.12 ± 0.01cms-1) than for methylamine (0.069 ± 0.006cms-1), while the values of the permeability coefficient are satisfactorily close to those obtained from the brute-force MD simulations (0.18 ± 0.03 and 0.052 ± 0.005cms-1 for ethanol and methylamine, respectively). Moreover, upon analyzing the thermodynamic and kinetic contributions to the permeability, we clarify that a higher concentration dependency of permeability for ethanol, as compared to methylamine, arises from the sensitive nature of ethanol's free-energy barrier within the inner region of the membrane against ethanol concentration.

Comparison of probabilistic structural reliability methods for ultimate limit state assessment of wind turbines

The probabilistic design of offshore wind turbines aims to ensure structural safety in a cost-effective way. This involves conducting structural reliability assessments for different design options and considering different structural responses. There are several structural reliability methods, and this paper will apply and compare different approaches in some simplified case studies. In particular, the well known environmental contour method will be compared to a more novel approach based on sequential sampling and Gaussian processes regression for an ultimate limit state case study on the maximum flapwise blade root bending moment. For one of the case studies, results will also be compared to results from a brute force simulation approach. Interestingly, the comparison is very different from the two case studies. In one of the cases the environmental contours method agrees well with the sequential sampling method but in the other, results vary considerably. Probably, this can be explained by the violation of some of the assumptions associated with the environmental contour approach, i.e. that the short-term variability of the response is large compared to the long-term variability of the environmental conditions. Results from this simple comparison study suggests that the sequential sampling method can be a robust and computationally effective approach for structural reliability assessment.

Tipping points of evolving epidemiological networks: Machine learning-assisted, data-driven effective modeling.

We study the tipping point collective dynamics of an adaptive susceptible-infected-susceptible (SIS) epidemiological network in a data-driven, machine learning-assisted manner. We identify a parameter-dependent effective stochastic differential equation (eSDE) in terms of physically meaningful coarse mean-field variables through a deep-learning ResNet architecture inspired by numerical stochastic integrators. We construct an approximate effective bifurcation diagram based on the identified drift term of the eSDE and contrast it with the mean-field SIS model bifurcation diagram. We observe a subcritical Hopf bifurcation in the evolving network's effective SIS dynamics that causes the tipping point behavior; this takes the form of large amplitude collective oscillations that spontaneously-yet rarely-arise from the neighborhood of a (noisy) stationary state. We study the statistics of these rare events both through repeated brute force simulations and by using established mathematical/computational tools exploiting the right-hand side of the identified SDE. We demonstrate that such a collective SDE can also be identified (and the rare event computations also performed) in terms of data-driven coarse observables, obtained here via manifold learning techniques, in particular, Diffusion Maps. The workflow of our study is straightforwardly applicable to other complex dynamic problems exhibiting tipping point dynamics.

Computational models of compound nerve action potentials: Efficient filter-based methods to quantify effects of tissue conductivities, conduction distance, and nerve fiber parameters.

Peripheral nerve recordings can enhance the efficacy of neurostimulation therapies by providing a feedback signal to adjust stimulation settings for greater efficacy or reduced side effects. Computational models can accelerate the development of interfaces with high signal-to-noise ratio and selective recording. However, validation and tuning of model outputs against in vivo recordings remains computationally prohibitive due to the large number of fibers in a nerve. We designed and implemented highly efficient modeling methods for simulating electrically evoked compound nerve action potential (CNAP) signals. The method simulated a subset of fiber diameters present in the nerve using NEURON, interpolated action potential templates across fiber diameters, and filtered the templates with a weighting function derived from fiber-specific conduction velocity and electromagnetic reciprocity outputs of a volume conductor model. We applied the methods to simulate CNAPs from rat cervical vagus nerve. Brute force simulation of a rat vagal CNAP with all 1,759 myelinated and 13,283 unmyelinated fibers in NEURON required 286 and 15,860 CPU hours, respectively, while filtering interpolated templates required 30 and 38 seconds on a desktop computer while maintaining accuracy. Modeled CNAP amplitude could vary by over two orders of magnitude depending on tissue conductivities and cuff opening within experimentally relevant ranges. Conduction distance and fiber diameter distribution also strongly influenced the modeled CNAP amplitude, shape, and latency. Modeled and in vivo signals had comparable shape, amplitude, and latency for myelinated fibers but not for unmyelinated fibers. Highly efficient methods of modeling neural recordings quantified the large impact that tissue properties, conduction distance, and nerve fiber parameters have on CNAPs. These methods expand the computational accessibility of neural recording models, enable efficient model tuning for validation, and facilitate the design of novel recording interfaces for neurostimulation feedback and understanding physiological systems.

Collective Variable-Based Enhanced Sampling: From Human Learning to Machine Learning.

Enhanced-sampling algorithms relying on collective variables (CVs) are extensively employed to study complex (bio)chemical processes that are not amenable to brute-force molecular simulations. The selection of appropriate CVs characterizing the slow movement modes is of paramount importance for reliable and efficient enhanced-sampling simulations. In this Perspective, we first review the application and limitations of CVs obtained from chemical and geometrical intuition. We also introduce path-sampling algorithms, which can identify path-like CVs in a high-dimensional free-energy space. Machine-learning algorithms offer a viable approach to finding suitable CVs by analyzing trajectories from preliminary simulations. We discuss both the performance of machine-learning-derived CVs in enhanced-sampling simulations of experimental models and the challenges involved in applying these CVs to realistic, complex molecular assemblies. Moreover, we provide a prospective view of the potential advancements of machine-learning algorithms for the development of CVs in the field of enhanced-sampling simulations.

Kinetics of Oriented Attachment of Mica Crystals.

Oriented attachment (OA), that is, the coalescence of crystals through attachment on coaligned crystal faces, is a nonclassical crystal growth process. Before attachment, a mesocrystal consisting of coaligned parallel crystals but with liquid separating them was observed. Fundamental questions such as why OA is kinetically favored and whether a mesocrystal stage is a prerequisite for OA are raised. Through combining brute-force molecular dynamics simulations and path samplings based on extensive umbrella simulations, we address these questions with a case study on the OA of a mica nanocrystal onto a mica crystal substrate in water. Brute-force simulations show that if two mica crystals are attached but largely misaligned, coalignment hardly appears. Thus, if OA is possible, then coalignment must appear before the attachment between crystals. Electrophoresis of the nanocrystal toward the substrate surface is spontaneous, but mesocrystal formation is occasional, also shown by brute-force simulations. Free energies along different pathways show that OA is spontaneous and kinetically favored over non-OA, and a mesocrystal formation is just a bifurcation in the pathway. OA is through a pathway in which the nanocrystal is tilted with respect to the substrate. Part of the nanocrystal is attached to the substrate first, and then, OA is gradually completed. Once a mesocrystal is occasionally formed, then a jump event is needed for the nanocrystal to get back to the OA pathway. The sampling technique here can hopefully guide the design of nanostructured materials facilitated by OA.

Variational analysis and AI algorithm implementation of free energy landscapes of molecular system

Accurate description of the free energy landscape (FES) is the basis for understanding complex molecular systems, and for further realizing molecular design, manufacture and industrialization. Major challenges include multiple metastable states, which usually are separated by high potential barriers and are not linearly separable, and may exist at multiple levels of time and spatial scales. Consequently FES is not suitable for analytical analysis and brute force simulation. To address these challenges, many enhanced sampling methods have been developed. However, utility of them usually involves many empirical choices, which hinders research advancement, and also makes error control very unimportant. Although variational calculus has been widely applied and achieved great success in physics, engineering and statistics, its application in complex molecular systems has just begun with the development of neural networks. This brief review is to summarize the background, major developments, current limitations, and prospects of applying variation in this field. It is hoped to facilitate the AI algorithm development for complex molecular systems in general, and to promote the further methodological development in this line of research in particular.

The brute force simulation of the nutrient losses during aquaculture feed hydrothermal processing using Phyllospora comosa compositional data

The brute force simulation of the nutrient losses during aquaculture feed hydrothermal processing using Phyllospora comosa compositional data

Accelerated Discovery of Metal–Organic Frameworks for CO2 Capture by Artificial Intelligence

Accelerated Discovery of Metal–Organic Frameworks for CO₂ Capture by Artificial Intelligence

Unveiling the crystallization mechanism of cadmium selenide via molecular dynamics simulation with machine-learning-based deep potential

Cadmium selenide (CdSe) is an inorganic semiconductor with unique optical and electronic properties that make it useful in various applications, including solar cells, light-emitting diodes, and biofluorescent tagging. In order to synthesize high-quality crystals and subsequently integrate them into devices, it is crucial to understand the atomic scale crystallization mechanism of CdSe. Unfortunately, such studies are still absent in the literature. To overcome this limitation, we employed an enhanced sampling-accelerated active learning approach to construct a deep neural potential with ab initio accuracy for studying the crystallization of CdSe. Our brute-force molecular dynamics simulations revealed that a spherical-like nucleus formed spontaneously and stochastically, resulting in a stacking disordered structure where the competition between hexagonal wurtzite and cubic zinc blende polymorphs is temperature-dependent. We found that pure hexagonal crystal can only be obtained approximately above 1430 K, which is 35 K below its melting temperature. Furthermore, we observed that the solidification dynamics of Cd and Se atoms were distinct due to their different diffusion coefficients. The solidification process was initiated by lower mobile Se atoms forming tetrahedral frameworks, followed by Cd atoms occupying these tetrahedral centers and settling down until the third-shell neighbor of Se atoms sited on their lattice positions. Therefore, the medium-range ordering of Se atoms governs the crystallization process of CdSe. Our findings indicate that understanding the complex dynamical process is the key to comprehending the crystallization mechanism of compounds like CdSe, and can shed lights in the synthesis of high-quality crystals.

Leveraging Machine Learning for Metal-Organic Frameworks: A Perspective.

Metal-organic frameworks (MOFs) have attracted tremendous interest because of their tunable structures, functionalities, and physiochemical properties. The nearly infinite combinations of metal nodes and organic linkers have led to the synthesis of over 100,000 experimental MOFs and the construction of millions of hypothetical counterparts. It is intractable to identify the best candidates in the immense chemical space of MOFs for applications via conventional trial-to-error experiments or brute-force simulations. Over the past several years, machine learning (ML) has substantially transformed the way of MOF discovery, design, and synthesis. Driven by the abundant data from experiments or simulations, ML can not only efficiently and accurately predict MOF properties but also quantitatively derive structure-property relationships for rational design and screening. In this Perspective, we summarize recent achievements in leveraging ML for MOFs from the aspects of data acquisition, featurization, model training, and applications. Then, current challenges and new opportunities are discussed for the future exploration of ML to accelerate the development of new MOFs in this vibrant field.

Modeling of Finite Beams in Periodic FDTD Methods

This article presents a finite-difference time-domain (FDTD) framework that accelerates the modeling of the interaction of infinite periodic structures with practical sources (such as Gaussian or Bessel beams) by exploiting periodic boundary conditions (PBCs). The technique involves the introduction of arbitrary beam-generating sources in periodic simulations via the array-scanning method (ASM). The proposed method is both efficient and practical, being naturally broadband and highly parallelizable. We present examples of Gaussian beams in free space and interacting with grating geometries, efficiently replicating the results of multi-period brute-force simulations.

Brute-force nucleation rates of hard spheres compared with rare-event methods and classical nucleation theory.

We determine nucleation rates of hard spheres using brute-force molecular dynamics simulations. We overcome nucleation barriers of up to 28 kBT, leading to a rigorous test of nucleation rates obtained from rare-event methods and classical nucleation theory. Our brute-force nucleation rates show excellent agreement with umbrella sampling simulations by Filion et al. [J. Chem. Phys. 133, 244115 (2010)] and seeding simulations by Espinosa et al. [J. Chem. Phys. 144, 034501 (2016)].

Reinforcement learning based hybrid bond-order coarse-grained interatomic potentials for exploring mesoscale aggregation in liquid-liquid mixtures.

Exploring mesoscopic physical phenomena has always been a challenge for brute-force all-atom molecular dynamics simulations. Although recent advances in computing hardware have improved the accessible length scales, reaching mesoscopic timescales is still a significant bottleneck. Coarse-graining of all-atom models allows robust investigation of mesoscale physics with a reduced spatial and temporal resolution but preserves desired structural features of molecules, unlike continuum-based methods. Here, we present a hybrid bond-order coarse-grained forcefield (HyCG) for modeling mesoscale aggregation phenomena in liquid-liquid mixtures. The intuitive hybrid functional form of the potential offers interpretability to our model, unlike many machine learning based interatomic potentials. We parameterize the potential with the continuous action Monte Carlo Tree Search (cMCTS) algorithm, a reinforcement learning (RL) based global optimizing scheme, using training data from all-atom simulations. The resulting RL-HyCG correctly describes mesoscale critical fluctuations in binary liquid-liquid extraction systems. cMCTS, the RL algorithm, accurately captures the mean behavior of various geometrical properties of the molecule of interest, which were excluded from the training set. The developed potential model along with the RL-based training workflow could be applied to explore a variety of other mesoscale physical phenomena that are typically inaccessible to all-atom molecular dynamics simulations.

Spatially Dispersive Metasurfaces—Part II: IE-GSTC-SD Field Solver With Extended GSTCs

An integral equation (IE) based field solver to compute the scattered fields from spatially dispersive metasurfaces is proposed and numerically confirmed using various examples involving physical unit cells. The work is a continuation of Part-1 [1], which proposed the basic methodology of representing spatially dispersive metasurface structure in the spatial frequency domain, k||. By representing the angular dependence of the surface susceptibilities in k|| as a ratio of two polynomials, the standard generalized sheet transition conditions (GSTCs) have been extended to include the spatial derivatives of both the difference and average fields around the metasurface. These extended boundary conditions are successfully integrated here into a standard IE-GSTC solver, which leads to the new IE-GSTC-SD simulation framework. The proposed IE-GSTC-SD platform is applied to various uniform metasurfaces, including a practical short conducting wire unit cell, as a representative practical example, for various cases of finite-sized flat and curvilinear surfaces. In all cases, computed field distributions are successfully validated, either against the semi-analytical Fourier decomposition method or the brute-force full-wave simulation of volumetric metasurfaces in the commercial Ansys FEM-HFSS simulator.