Coarse-grained Techniques Research Articles

Accurate nuclear quantum statistics on machine-learned classical effective potentials.

The contribution of nuclear quantum effects (NQEs) to the properties of various hydrogen-bound systems, including biomolecules, is increasingly recognized. Despite the development of many acceleration techniques, the computational overhead of incorporating NQEs in complex systems is sizable, particularly at low temperatures. In this work, we leverage deep learning and multiscale coarse-graining techniques to mitigate the computational burden of path integral molecular dynamics (PIMD). In particular, we employ a machine-learned potential to accurately represent corrections to classical potentials, thereby significantly reducing the computational cost of simulating NQEs. We validate our approach using four distinct systems: Morse potential, Zundel cation, single water molecule, and bulk water. Our framework allows us to accurately compute position-dependent static properties, as demonstrated by the excellent agreement obtained between the machine-learned potential and computationally intensive PIMD calculations, even in the presence of strong NQEs. This approach opens the way to the development of transferable machine-learned potentials capable of accurately reproducing NQEs in a wide range of molecular systems.

Creating dynamic leakage-free paths using coarse-graining techniques in the presence of decoherence

Creating dynamic leakage-free paths using coarse-graining techniques in the presence of decoherence

DELTA: Memory-Efficient Training via Dynamic Fine-Grained Recomputation and Swapping

To accommodate the increasingly large-scale models within limited-capacity GPU memory, various coarse-grained techniques, such as recomputation and swapping, have been proposed to optimize memory usage. However, these methods have encountered limitations, either in terms of inefficient memory reduction or diminished training performance. In response to this, our paper introduces DELTA, an innovative approach for memory-efficient large-scale model training that combines fine-grained memory optimization and prefetching technology to reduce memory usage while maintaining high training throughput concurrently. Initially, we formulate the problem of memory-throughput joint optimization as an easy-solving 0/1 Knapsack problem. Leveraging this formalization, we use an improving polynomial complexity heuristic algorithm to address the problem effectively. Furthermore, we introduce a novel bidirectional prefetching technology into dynamic memory management, which significantly accelerates the model training when compared to relying solely on recomputation or swapping. Finally, DELTA offers users an automated training execution library, eliminating the need for manual configuration or specialized expertise. Experimental results demonstrate the effectiveness of DELTA in reducing GPU memory consumption. Compared to state-of-the-art methods, DELTA achieves substantial memory savings ranging from 40% to 72%, while maintaining comparable convergence performance for various models, including ResNet-50, ResNet-101, and BERT-Large. Notably, DELTA enables the training of GPT2-Large and GPT2-XL with batch sizes increased by 5.5 × and 6 ×, respectively, showcasing its versatility and practicality in enabling large-scale model training on GPU hardware.

Crystalline and Amorphous Interface Simulations of Donor-Acceptor Blends.

In an organic solar cell, exciton dissociation and charge transport to generate current depend on interface and bulk morphology, respectively, and their rates dictate device performance. Blend miscibility and processing determine the final morphology. We investigate the blend miscibility of P3HT and O-IDTBR, employing our recently developed "push-pull" computational technique, and explore its effect on nanoscale morphology. The resulting Flory-Huggins χ parameter is primarily enthalpic, and the blend exhibits UCST behavior. To directly simulate an equilibrated morphology in such a blend requires large (30 nm) systems and long (10 μs) simulations, which we attain with our virtual site coarse graining technique. In the resulting amorphous phase-separated state, O-IDTBR swells P3HT by about 15%, but does not percolate within the P3HT-rich phase. The interfacial profile depends on crystallinity: the amorphous interface is several nanometers wide, whereas an acceptor crystal induces order in the adjacent donor polymer, forming a sharp interface quite similar to a crystal-crystal interface. The wide amorphous interface promotes more donor-acceptor contacts with a wide range of relative orientations; however, the acceptors do not form a percolating network, which will likely lead to charge recombination. A crystal interface exhibits fewer donor-acceptor contacts with primarily face-on orientation that are more likely to charge-separate because of instant access to the preferred domain.

Wurster fluidised-bed coating: Coarse-graining technique within CFD-DEM in conjunction with heat and mass transfer

In this study, we use a combined CFD-Discrete Element Method to assess predictive capabilities and numerical implications of a coarse-graining technique in Wurster fluidised-bed coaters. We investigated both hydrodynamics and heat and mass transfer and conducted simulations of a full three-phase system for the original and three coarse-grained systems, analysing velocity distributions, macroscopic solid stresses, moisture content and phase temperatures. We achieved a logarithmic simulation speed-up by aggregating up to 64 original particles into each coarse grain. This was accomplished while maintaining fidelity to the original CFD-DEM system in terms of reproducing with high accuracy macroscopic granular flow properties in different regions of a coater (drying, tube and bed regions). By integrating a liquid spray and humid air, we demonstrated that the phase temperatures were accurately predicted within the coarse-grained system, with a high capability of delivering liquid spray distributions with the same uniformity and drying. We also give arguments for choosing a certain degree of coarse-graining as a compromise between a desired reduction of computational costs and a trustworthy reproduction of granular-flow physics encountered in different regions of a Wurster bed. Our findings pave the way to using CFD-DEM to industrially-scaled Wurster-bed systems, which is currently unfeasible due to prohibitive computational costs.

Simulations of Functional Motions of Super Large Biomolecules with a Mixed-Resolution Model.

Many large protein machines function through an interplay between large-scale movements and intricate conformational changes. Understanding functional motions of these proteins through simulations becomes challenging for both all-atom and coarse-grained (CG) modeling techniques because neither approach alone can readily capture the full details of these motions. In this study, we develop a multiscale model by employing the popular MARTINI CG model to represent a heterogeneous environment and structurally stable proteins and using the united-atom (UA) model PACE to describe proteins undergoing subtle conformational changes. PACE was previously developed to be compatible with the MARTINI solvent and membrane. Here, we couple the protein descriptions of the two models by directly mixing UA and CG interaction parameters to greatly simplify parameter determination. Through extensive validations with diverse protein systems in solution or membrane, we demonstrate that only additional parameter rescaling is needed to enable the resulting model to recover the stability of native structures of proteins under mixed representation. Moreover, we identify the optimal scaling factors that can be applied to various protein systems, rendering the model potentially transferable. To further demonstrate its applicability for realistic systems, we apply the model to a mechanosensitive ion channel Piezo1 that has peripheral arms for sensing membrane tension and a central pore for ion conductance. The model can reproduce the coupling between Piezo1's large-scale arm movement and subtle pore opening in response to membrane stress while consuming much less computational costs than all-atom models. Therefore, our model shows promise for studying functional motions of large protein machines.

Uncertainty driven active learning of coarse grained free energy models

Coarse graining techniques play an essential role in accelerating molecular simulations of systems with large length and time scales. Theoretically grounded bottom-up models are appealing due to their thermodynamic consistency with the underlying all-atom models. In this direction, machine learning approaches hold great promise to fitting complex many-body data. However, training models may require collection of large amounts of expensive data. Moreover, quantifying trained model accuracy is challenging, especially in cases of non-trivial free energy configurations, where training data may be sparse. We demonstrate a path towards uncertainty-aware models of coarse grained free energy surfaces. Specifically, we show that principled Bayesian model uncertainty allows for efficient data collection through an on-the-fly active learning framework and opens the possibility of adaptive transfer of models across different chemical systems. Uncertainties also characterize models’ accuracy of free energy predictions, even when training is performed only on forces. This work helps pave the way towards efficient autonomous training of reliable and uncertainty aware many-body machine learned coarse grain models.

Influence of ion and hydration atmospheres on RNA structure and dynamics: insights from advanced theoretical and computational methods.

RNA, a highly charged biopolymer composed of negatively charged phosphate groups, defies electrostatic repulsion to adopt well-defined, compact structures. Hence, the presence of positively charged metal ions is crucial not only for RNA's charge neutralization, but they also coherently decorate the ion atmosphere of RNA to stabilize its compact fold. This feature article elucidates various modes of close RNA-ion interactions, with a special emphasis on Mg2+ as an outer-sphere and inner-sphere ion. Through examples, we highlight how inner-sphere chelated Mg2+ stabilizes RNA pseudoknots, while outer-sphere ions can also exert long-range electrostatic interactions, inducing groove narrowing, coaxial helical stacking, and RNA ring formation. In addition to investigating the RNA's ion environment, we note that the RNA's hydration environment is relatively underexplored. Our study delves into its profound interplay with the structural dynamics of RNA, employing state-of-the-art atomistic simulation techniques. Through examples, we illustrate how specific ions and water molecules are associated with RNA functions, leveraging atomistic simulations to identify preferential ion binding and hydration sites. However, understanding their impact(s) on the RNA structure remains challenging due to the involvement of large length and long time scales associated with RNA's dynamic nature. Nevertheless, our contributions and recent advances in coarse-grained simulation techniques offer insights into large-scale structural changes dynamically linked to the RNA ion atmosphere. In this connection, we also review how different cutting-edge computational simulation methods provide a microscopic lens into the influence of ions and hydration on RNA structure and dynamics, elucidating distinct ion atmospheric components and specific hydration layers and their individual and collective impacts.

Accelerated DEM simulation of the hopper-screw feeder and tablet-press feeder using the multi-level coarse-graining technique

A recently proposed multi-level coarse-graining method (MCG) for DEM simulation [1] is demonstrated in this work for two additional powder handling systems, namely, a hopper screw feeder and a tablet-press feeder. In contrast to the original demonstration by De et al., these two systems are characterized by moving geometry, which generates complex flow patterns consisting of rapid particle movement and recirculation. Despite these complexities, the proposed MCG technique successfully captures these systems' critical features of the powder flow. It substantially reduces the computation time compared to the fully resolved DEM simulations. The detailed time analysis reveals that the moving mesh consumes significant computation time and reduces the degree of speed-up achieved by MCG for some cases. To alleviate this situation, the transition planes in MCG should be positioned appropriately, along with an efficient neighbor-searching algorithm. Despite these constraints, significant speed-up could be achieved for both the systems.

MODELING A RED BLOOD CELL CYTOSKELETON: INSIGHTS AND TIPS

The aim of this work is to provide a clear and ready-to-implement mathematical model to describe the red blood cell (RBC) cytoskeleton. The motivation for this methodo- logy work lies in the lack of available sources facilitating the implementation of computational (in silico) modelling of RBC mechanical properties. Our approach adopts a worm-like chain force model to mimic an individual spectrin deformation and the virtual work principle to extrapolate the spectrin network dynamics. We report modelling tips using coarse-grained scaling techniques and optical tweezers experiments. We validate the model in virtual expe- riments gaining insights on single spectrin equilibrium, the impact of cytoskeletal symmetry defects on system behaviour, and stress transmission and distribution within the network. By performing these virtual experiments, we elucidate the role that the cytoskeleton plays in de- termining the remarkable RBC mechanical behaviour. These findings emphasise the need to consider the cytoskeleton component in the latest and most advanced single-cell mathematical models.

Single-layer MoS2 solid-state nanopores for coarse-grained sequencing of proteins

Proteins are essential biological molecules to use as biomarkers for early disease diagnosis. Therefore, their detection is crucial. In recent years, protein sequencing has become one of the most promising techniques. In particular, solid-state nanopores (SSNs) are powerful platforms for single biological molecule sensing without any labeling and with high sensitivity. Atomically thin two-dimensional (2D) materials with nanometer-sized pores, such as single-layer MoS2, represent the ideal SSN because of their ultimate thinness. Despite the benefits they offer, their use for protein sequencing applications remains very challenging since the fast translocation speed provides a short observation time per single molecule. In this work, we performed extensive molecular dynamics simulations of the translocation of the 20 proteinogenic amino acids through single-layer MoS2 nanopores. From ionic current traces, we characterized peptide-induced blockade levels of current and duration for each of the 20 natural amino acids. Using clustering techniques, we demonstrate that positively and negatively charged amino acids present singular fingerprints and can be visually distinguished from neutral amino acids. Furthermore, we demonstrate that this information would be sufficient to identify proteins using the coarse-grained sequencing technique made of only three amino acid categories depending on their charge. Therefore, single-layer MoS2 nanopores have great potential as sensors for the identification of biomarkers.

Ocean 2D eddy energy fluxes from small mesoscale processes with SWOT

Abstract. We investigate ocean dynamics at different scales in the Agulhas Current system, a region of important interocean exchange of heat and energy. While ocean observations and some of the most advanced climate models capture the larger mesoscale dynamics (> 100 km), the smaller-scale fronts and eddies are underrepresented. The recently launched NASA–CNES Surface Water and Ocean Topography (SWOT) wide-swath altimeter mission observes the smaller ocean geostrophic scales down to 15 km in wavelength globally. Here we will analyse different eddy diagnostics in the Agulhas Current region and quantify the contributions from the larger mesoscales observable today and the smaller scales to be observed with SWOT. Surface geostrophic diagnostics of eddy kinetic energy, strain, and energy cascades are estimated from modelled sea surface height (SSH) fields of the Massachusetts Institute of Technology general circulation model (MITgcm) latitude–longitude polar cap (LLC4320) simulation subsampled at 1/10∘. In this region, the smaller scales (<150 km) have a strong signature on the horizontal geostrophic strain rate and for all eddy diagnostics in the Western Boundary Current and along the meandering Agulhas Extension. We investigate the horizontal cascade of energy using a coarse-graining technique, and we observe that the wavelength range where the inverse cascade occurs is biased towards larger mesoscale wavelengths with today’s altimetric sampling. We also calculate the projected sampling of the eddy diagnostics under the SWOT swaths built with the NASA–CNES simulator to include the satellite position and realistic noise. For the swaths, a neural network noise mitigation method is implemented to reduce the residual SWOT random error before calculating eddy diagnostics. In terms of SSH, observable wavelengths of 15 to 20 km are retrieved after neural network noise mitigation, as opposed to wavelengths larger than 40 km before the noise reduction.

Enhancing efficiency in particle aggregation simulations: Coarse-grained particle modeling in the DEM-PBM coupled framework

The computational cost of the discrete element method (DEM)-population balance model (PBM) coupled framework is predominantly attributed to DEM simulations. To overcome this challenge, coarse-grained (CG) particles have been introduced in the DEM-PBM coupled framework. In this study, we proposed a new CG-enabled DEM-PBM coupled framework that builds upon the previous work of Das et al. (Proc. R. Soc. Lond. Ser. A Math. Phys. Eng. Sci. 478 (2261) (2022) 20220076). By incorporating the CG technique, the particle number density is reduced, resulting in fewer collisions compared to the resolved system. To address this issue, a scaling law has been developed to derive the collision frequency of the resolved system from the CG system. The verification of the new scaling law has been demonstrated through various simulation studies. Furthermore, the entire DEM-PBM coupled framework has been modified using the proposed methodology. The efficiency of the CG–DEM–PBM coupled simulation method has been successfully demonstrated through simulations of rotating drum and continuous mixing technology (CMT). Compared to the resolved simulation approach, the newly proposed CG-enabled DEM-PBM coupled framework maintains accuracy in terms of particle size distribution and other essential findings while significantly reducing simulation time.

Numerical assessment of transition in cutting mode and its effect on roughness creation

Ductile-to-brittle transition in cutting mode is an important phenomenon in the ultra-precision machining and surface finish of brittle materials. Using a newly-developed coarse-grained molecular dynamics technique, we systematically explore the effect of depth of cut on the cutting mode and the evolution of roughness. Our simulations demonstrate that the transitional depth of cut (TDoC) model accurately predicts the material removal mechanisms, which are affected by both the material’s properties and the cutting depth. When the cutting depth increases, the operating mechanism shifts from shearing-induced serrated chip formation to cleavage-cracking chip breakage. The RMS roughness study demonstrates that the roughness is a direct outcome of the chip formation mechanism and is, in general, higher when chip breakage occurs.

Fluidized bed gas-solid heat transfer using a CFD-DEM coarse-graining technique

Computational Fluid Dynamics - Discrete Element Method (CFD-DEM) is extensively used for modeling heat transfer in gas-solid fluidized beds. However, CFD-DEM is computationally expensive, leading to a restriction regarding the number of simulated particles. Coarse-grained CFD-DEM is a technique to circumvent this constraint, allowing one to simulate larger fluidized beds. In this work, a scaling law used for coarse-graining hydrodynamics is generalized to gas-solid heat-transfer. This approach for coarse-graining heat transfer is tested using three different superficial gas velocities where the coarse-grained particle temperatures and Nusselt numbers are obtained. The particle temperature shows good correspondence with the original system for all cases and the Nusselt number is accurately predicted by the coarse-graining scaling law.

Prediction of Electronic Properties of Radical-Containing Polymers at Coarse-Grained Resolutions

The properties of soft electronic materials depend on the coupling of electronic and conformational degrees of freedom over a wide range of spatiotemporal scales. The description of such properties requires multiscale approaches capable of, at the same time, accessing electronic properties and sampling the conformational space of soft materials. This could in principle be realized by connecting the coarse-grained (CG) methodologies required for adequate conformational sampling to conformationally averaged electronic property distributions via backmapping to atomistic-resolution level models and repeated quantum-chemical calculations. Computational demands of such approaches, however, have hindered their application in high-throughput computer-aided soft materials discovery. Here, we present a method that, combining machine learning and CG techniques, can replace traditional backmapping-based approaches without sacrificing accuracy. We illustrate the method for an emerging class of soft electronic materials, namely, nonconjugated, radical-containing polymers, promising materials for all-organic energy storage. Supervised machine learning models are trained to learn the dependence of electronic properties on polymer conformation at CG resolutions. We then parametrize CG models that retain electronic structure information, simulate CG condensed phases, and predict the electronic properties of such phases solely from the CG degrees of freedom. We validate our method by comparing it against a full backmapping-based approach and find good agreement between both methods. This work demonstrates the potential of the proposed method to accelerate multiscale workflows and provides a framework for the development of CG models that retain electronic structure information.

Efficiency of energy and enstrophy transfers in periodical flows

We apply a coarse-graining technique to understand the efficiency of scale-to-scale transport of energy and enstrophy in a quasi-two-dimensional weakly turbulent periodic flow. The investigated periodic flow resembles the propagation of a monochromatic tide in a tidal channel, connected to open sea through an inlet. The interaction of the periodic flow with the inlet mouth generates vortical structures in a wide spectrum of scales, and recently, how the corresponding energy and enstrophy fluxes change their signs depending on the tidal phase has been shown. In the present study, we are interested to extend the analysis to the efficiency of the nonlinear transfer rates by analyzing the geometric alignment between the turbulent stresses and the strain rates for the energy, and the vorticity stress and large-scale vorticity gradient for the enstrophy. Our results suggest that, depending on the phase of the period, energy is efficiently transferred to larger scales (inverse cascade) in a finite range of scales, whereas the observed direct energy cascade for very small and very large scales is much less efficient. Enstrophy shows similar behaviors in terms of transitions between direct and inverse cascading; however, all transfers seem to be relatively inefficient.

Intelligent fault detection scheme for constant-speed wind turbines based on improved multiscale fuzzy entropy and adaptive chaotic Aquila optimization-based support vector machine

Timely and effective fault detection is essential to ensure the safe and reliable operation of wind turbines. However, due to the complex kinematic mechanisms and harsh working environments of wind turbine equipment, it is difficult to extract sensitive features and detect faults from acquired wind turbine signals. To address this challenge, a novel intelligent fault detection scheme for constant-speed wind turbines based on refined time-shifted multiscale fuzzy entropy (RTSMFE), supervised isometric mapping (SI), and adaptive chaotic Aquila optimization-based support vector machine (ACAOSVM) is proposed. In the first step, the RTSMFE method is used to fully extract features of the wind turbine system. The time-shifted coarse-grained construction technique and a refined computing technique are adopted in the RTSMFE method to enhance the capability of traditional multiscale fuzzy entropy for measuring the complexity of signals. Subsequently, an effective manifold learning approach, SI, is applied to obtain the important and low-dimensional feature set from the high-dimensional feature set. Finally, sensitive features are fed into the ACAOSVM classifier to identify faults. The proposed ACAO algorithm is used to optimize important parameters of the SVM, thereby improving its detection performance. Simulations and wind turbine experiments verified that the proposed RTSMFE outperforms existing entropy techniques in terms of complexity measurement and feature extraction. Furthermore, the proposed ACAOSVM classifier is superior to existing advanced classifiers for fault pattern recognition. Finally, the proposed intelligent fault detection scheme can more correctly and efficiently detect wind turbine single/hybrid faults than other recently published schemes.

A detailed gas-solid fluidized bed comparison study on CFD-DEM coarse-graining techniques

Computational Fluid Dynamics - Discrete Element Method (CFD-DEM) is a numerical tool used for detailed fluidized bed studies. However, CFD-DEM is computationally expensive, leading to a restriction regarding the number of simulated particles. Therefore, coarse-graining techniques have been developed to increase the simulation scale or reduce computational requirements for CFD-DEM simulations in fluidized beds. In this work, we critically compared the coarse-graining scaling laws of Mu et al. [2020, Chemical Engineering Science: X 6.] and Sakai and Koshizuka [2009, Chemical Engineering Science: 64, 533–539.] for their effectiveness in characterizing the original system. The first mentioned approach was not able to accurately characterize the original system, while the latter methodology showed good correspondence. We also demonstrated that applying a continuous two-way smoothing function to the coarse-graining process can yield grid-independent solutions, particularly when the particle diameter is much larger than the grid cell size, which is often the case in CFD-DEM simulations.

FFT-based homogenization at finite strains using composite boxels (ComBo)

Computational homogenization is the gold standard for concurrent multi-scale simulations (e.g., FE2) in scale-bridging applications. Often the simulations are based on experimental and synthetic material microstructures represented by high-resolution 3D image data. The computational complexity of simulations operating on such voxel data is distinct. The inability of voxelized 3D geometries to capture smooth material interfaces accurately, along with the necessity for complexity reduction, has motivated a special local coarse-graining technique called composite voxels (Kabel et al. Comput Methods Appl Mech Eng 294: 168–188, 2015). They condense multiple fine-scale voxels into a single voxel, whose constitutive model is derived from the laminate theory. Our contribution generalizes composite voxels towards composite boxels (ComBo) that are non-equiaxed, a feature that can pay off for materials with a preferred direction such as pseudo-uni-directional fiber composites. A novel image-based normal detection algorithm is devised which (i) allows for boxels in the firsts place and (ii) reduces the error in the phase-averaged stresses by around 30% against the orientation cf. Kabel et al. (Comput Methods Appl Mech Eng 294: 168–188, 2015) even for equiaxed voxels. Further, the use of ComBo for finite strain simulations is studied in detail. An efficient and robust implementation is proposed, featuring an essential selective back-projection algorithm preventing physically inadmissible states. Various examples show the efficiency of ComBo against the original proposal by Kabel et al. (Comput Methods Appl Mech Eng 294: 168–188, 2015) and the proposed algorithmic enhancements for nonlinear mechanical problems. The general usability is emphasized by examining various Fast Fourier Transform (FFT) based solvers, including a detailed description of the Doubly-Fine Material Grid (DFMG) for finite strains. All of the studied schemes benefit from the ComBo discretization.