The transport and tally capabilities based on unstructured mesh (UM) offer significant advantages in representing complex geometry and producing precise tally results over Constructive Solid Geometry(CSG) and Direct Accelerated Geometry Monte Carlo (DAGMC), which can enhance Monte Carlo applications in high-fidelity multi-physics coupling. In this paper, two new capabilities of tally and particle transport based on UM have been studied and implemented in the Reactor Monte Carlo code (RMC). First, a tally capability based on UM under CSG or DAGMC transport framework has been researched, implemented, and verified in RMC, demonstrating considerable tally efficiency in the KRUSTY application. Next, particle transport and tally directly on UM have also been implemented and verified in RMC. In comparison to DAGMC for the irregular Stanford Bunny model, the calculation speed is 8.63 times faster. Additionally, a practical Molten Salt Fast Reactor (MSFR) model has been calculated with RMC’s new transport and tally capabilities, showcasing RMC's ability to handle complex geometries and produce precise tally results in real irregular reactor application.

Video moment retrieval locates a specified moment by a sentence query. Recent approaches have made remarkable advancements with large-scale video-sentence annotations. These annotations require extensive human labor and expertise, leading to the need for unsupervised fashion. Generating pseudo-supervision from videos is an effective strategy. With the power of the large-scale pre-trained model, we introduce knowledge into constructing pseudo-supervision. The main technical challenge is improving pseudo-supervision diversity and alleviating noise brought by external knowledge. To address these problems, we propose two Knowledge-based Pseudo Supervision Construction (KPSC) strategies: KPSC-P and KPSC-F. They all follow two steps: generating diverse samples and alleviating knowledge chaos. The main difference is that the former first learns a representation space with prompt tuning, while the latter directly utilizes data information. KPSC-P has two modules: 1) Proposal Prompt (PP): Generate temporal proposals; 2) Verb Prompt (VP): Generate pseudo-queries with noun-verb patterns. KPSC-F also has two modules: 1) Captioner: Generating candidate queries; 2) Filter: Alleviating knowledge chaos. Thus, our KPSC involves two attempts to extract knowledge from pre-trained models. Extensive experiments show that our attempts outperform the existing unsupervised methods on two public datasets (Charades-STA and ActivityNet-Captions) and perform on par with several methods using stronger supervision.

Nowadays, convolutional neural networks are representative of deep learning algorithms. With the development of convolutional neural networks, their network structures become more complex, and the number of parameters for training becomes larger and larger. The parallelization of convolutional neural network algorithms on multicore or many-core processors is essential for training convolutional neural networks. In this paper, we propose a parallel algorithm of LeNet-1 based on the Sunway Bluelight II supercomputer, named SW-LeNet. Moreover, we propose a two-level parallelization scheme, including thread-level optimization and process-level optimization. In thread-level optimization, the following optimization methods are used, including CPEs parallelism, hybrid scheme and DMA optimization, register optimization, and SIMD data parallelism. Data parallelism optimization and parameter packing optimization are used in process-level optimization. Compared with the original LeNet, SW-LeNet can achieve 4.94x speedups in a single core group. Moreover, SW-LeNet can be scaled up to 2,048 processes, with 133,120 cores, and achieves 84.93% parallel efficiency.

In multi-group deterministic neutron transport simulations for nuclear reactors, anisotropic scattering between neutrons and light nuclides is a common issue causing errors in eigenvalue and power distributions. Transport correction is the main approach for mitigating this error and it has advantages over high-order scattering due to higher efficiency in computational time and storage. The research on transport correction for light water reactors, in which hydrogen is the main nuclide for neutron slowing down, has been underway for more than 40 years. However, there is a lack of effort on transport correction for nuclear reactors using graphite as moderators, in which carbon is the main nuclide for neutron slowing down. Although neutron scattering with carbon is less anisotropic compared to hydrogen, previous tests have demonstrated significant errors resulting from anisotropic scattering. In this article, we focus on the transport correction for graphite-moderated reactors using the Cumulative Migration Method (CMM). Anisotropy in neutron scattering with graphite is examined and multi-group transport correction ratios for C-12 are evaluated using CMM with Monte Carlo code OpenMC. Numerical simulations are performed on the MIT Graphite Exponential Pile (MGEP) and HTR-10 models. The test results demonstrate that CMM achieves higher accuracy in computing the effective multiplication factor for transport cross sections on nuclear reactors involving graphite.

The 2D/1D coupling method is recognized as one of the preferred high-fidelity calculation methods for reactors featured with well homogeneity in the axial direction. However, the choices of the technicalities, such as transverse leakage, axial solver, and transverse leakage splitting method, will lead to different calculational performance. This paper outlines the theory of the 2D/1D coupling method and describes the detailed implementations of the key technicalities. Based on numerical tests, the choices of the related calculational parameters are analyzed, such as the order of the quadrature set, ray spacing, and the axial mesh size. Then, the computational performance of six typical 2D/1D technicalities is compared and assessed. A comparison between 2D/1D coupling method and direct 3D MOC calculation is also made. For the transverse leakage, the Fourier expansion technique could significantly reduce the memory burden and computational cost, but with the similar accuracy as the anisotropic leakage term. It is recommended to use the axial DGFEM SN solver, which has better consistency and leads to higher computational efficiency. Additionally, the results also show that the 2D/1D coupling method can appropriately increase the axial mesh size, which only has slight effect on the accuracy. For nuclear reactors featured with nonstrong heterogeneity in axial direction, the 2D/1D coupling method has significant advantages than the direct 3D MOC calculation.

This study developed a new structured nuclear database to improve the readability and extensibility of the ACE (A Compact ENDF) nuclear database, save disk space, reduce memory usage, and enhance the computational efficiency of the Reactor Monte Carlo (RMC) code. A Python package was developed to store nuclear data in HDF5 format. Compared to the ACE database, the HDF5-format database shows significant improvements: an 80% reduction in disk space for continuous energy neutron data and a 60% reduction for neutron thermal scattering data. The HDF5-format database was implemented in RMC’s criticality calculation mode and validated through VERA benchmark problem 2B, demonstrating perfect agreement with the ACE results in keff and neutron flux counts. The Computational results indicate a 7.7% reduction in memory usage and a 20.1% improvement in computational efficiency with the HDF5-format database. Additional tests show that using the database at a single temperature point reduces memory usage by 5.4% and running time by 13.2%. At two temperature points, memory usage decreases by 35.2% and running time by 18.0%. The new data structure reduces temperature-independent redundant data and improves indexing efficiency, leading to greater savings with more temperature points. This development enhances performance of criticality calculation of RMC and addresses ACE database limitations.

Based on academic research and industrial applications over more than 20 years, the Reactor Monte Carlo code (RMC) developed by the REAL (Reactor Engineering Analysis Laboratory) team at Tsinghua University since 2000 has become a powerful, innovative, and versatile simulation platform for nuclear reactor analysis, shielding simulations, criticality safety calculations, fusion neutronics analysis and beyond. Utilizing collaborative and agile development technology, advanced methods and the most cutting-edge algorithms can be tested and implemented in RMC quickly and efficiently. RMC has been deployed on many world-class supercomputers in China and played an irreplaceable role in the design and analysis of commercial nuclear power plants and newly designed types of advanced nuclear reactors. This paper reviews the state-of-the-art technologies developed in RMC in recent years, such as stochastic and continuous-varying media modeling, advanced transient simulation capability, more accurate energy deposition model, etc. Parallel acceleration on heterogeneous architecture supercomputers and machine learning algorithms would be incorporated in ongoing research and future development plans.

The randomly dispersed Tristructural-isotropic (TRISO) particle fuel is an attractive nuclear fuel type widely used in advanced reactor designs. However, the massive randomly-distributed fuel elements and the prohibitive computational costs pose challenges for high-fidelity modeling and simulation of dispersion fuels. To address this, we present a virtual lattice optimization method for accelerating the Monte Carlo particle transport simulation of nuclear systems with dispersion fuels. The new method involves using a regular overlaid mesh on the dispersed fuel region to improve geometry processing in Monte Carlo simulations. We describe the optimization methodology in our paper and construct a theoretical performance model to determine the optimal lattice pitch size that minimizes calculation cost. We then develop and implement the algorithm of the virtual method in the open-source Monte Carlo code OpenMC. To demonstrate the efficacy of the new algorithm, high-fidelity HTR-PM full core models based on the Shidao-Bay nuclear power plant are constructed, including an explicit representation of up to 420,000 fuel pebbles. Criticality and depletion simulations are performed on large-scale models. Results show that the k-effective derived by the optimized code agrees well with both the experiment and the original code. Further performance comparison shows that the virtual lattice method exhibits lower time consumption, higher parallel efficiency, and lower storage requirement in comparison to the conventional physical lattice method. The study establishes the effectiveness of the proposed methodology in facilitating efficient and accurate high-fidelity simulations of large-scale dispersion fuel models.

The structural evolution of dielectric elastomer induced by pre-strain is a complex, multi-scale process that poses a significant challenge to a deep understanding of the effect of pre-strain. Through simulation results, we identify the variation in the dielectric constant and multi-scale (electronic structure, molecular chain conformation, and aggregation structure) response of poly(methyl acrylate). As the pre-strain increases, the dielectric constant initially rises (below 200% pre-strain) and then declines (above 200% pre-strain). Analysis of the charge distribution, surface electrostatic potential, HOMO-LUMO bandgap, and electron density differences reveal that adjusting chain conformation appropriately could enhance polarity domain and electron polarization. The correlation between permittivity and segment dynamics of deformed molecules is explored, encompassing segment orientation, mean shift displacement, and diffusion coefficient. Following molecular chain orientation, the kinematic capability of the chain segment improves, which leads to an increase in the number and activity of effective dipoles and the enhancement of orientation polarization. Excessive stretching restricts the polymer molecular chain mechanically, reducing the number and activity of effective dipoles and negatively impacting electron polarization. The permittivity transitions from isotropic to anisotropic behavior when the system is subjected to strain. This study provides an interesting solution for research on multiscale responses and intrinsic mechanisms of pre-strain.

The Random Ray Method (TRRM) is a new approach to solving partial differential equations (PDEs) based on the method of characteristics (MOC). It employs stochastic rather than deterministic discretization of characteristic tracks and can be used for the numerical simulation of nuclear reactors. In this paper, we propose SW-TRRM, a parallel optimization program for TRRM based on the Sunway Bluelight II Supercomputer for the first time. We present a two-level parallelization scheme that consists of thread-level and process-level optimization. At the thread-level, we introduce three schemes for speeding up within a single core group, including direct parallelization, parallelization by energy groups, and loop structure optimization. At the process-level, we implement task parallelization among multiple processes using domain replication. Moreover, we devise an algorithm to optimize the MPI collective communication across super-nodes. Experimental results show that SW-TRRM achieves a 17.40 × speedup within a single core group compared to the original TRRM program. When scaled up to 2,048 processes and 133,120 cores, SW-TRRM maintains good strong and weak scalability.

A novel experimental method for measuring the coefficient of thermal expansion of supercooled liquids is invented, as well as a new analytical methodology for determining the atomic-level thermal expansion of metallic glasses. We find a general rule of 1.7 times thermal expansion from glass to liquid for twenty bulk metallic glasses. This rule does not apply to other types of glasses or pure metals but only works for bulk metallic glasses containing copper or nickel. We revealed using molecular dynamics simulation that atomic shells of Cu (or Ni) with interatomic distances of 6.5–7.6 Å (or 6.4–7.6 Å for Ni) changed from 32 to 46 (or to 48 for Ni), when Cu50Zr50 (or Ni50Zr50) metallic glass enters its supercooled liquid, resulting in the sample’s 1.7 times thermal expansion at the glass transition. This research provides insights on how bulk metallic glasses form and highlights the relevance of chemical order at an interatomic distance of 6.5–7.6 Å in driving glass transition.

Collisions between charged particles can be described and solved using Monte Carlo methods in the framework of stochastic differential equations (SDEs). In this paper, we start from an SDE including the extended Lorentz collision operator, which can recover the collisions between a sampling electron and background ions and electrons. On this basis, we construct a second order weakly convergent algorithm (WCA2) to simulate collisional effects of electrons in plasmas. Superseding the Weiner process by a three-point distribution, WCA2 possesses high weakly convergent accuracy as well as low computational costs. The definition and properties of weak convergence are discussed in detail. The weakly convergent order of WCA2 is verified both theoretically and numerically. Through two trial moment functions, we carefully analyze the numerical solutions of the SDE using rigorous statistical tests in the sense of weak convergence. The criteria and practical operations of finding the benchmark solution of SDEs are introduced at length. In order to illustrate the power of WCA2, we apply it to simulate the backward runaways in plasmas, which is a dramatic physical phenomenon. By comparison with the Euler-Maruyama method and the Cadjan-Ivanov method, the advantage and efficiency of WCA2 is exhibited. The backward runaway probability and its dependence on initial conditions are accurately studied using WCA2.

The formation of bulk metallic glass requires the constituent elements to have a negative heat of mixing but has no restrictions on its magnitude. An understanding of this issue is lacking due to the absence of a valid method for describing chemical ordering of metallic glasses. For example, the radial distribution function is ineffective in identifying the elemental preferences of packed atoms. Here, we show that using molecular-dynamics simulation, the chemical medium-range ordering of liquid alloys can be evaluated from persistent homology. This inherently arising chemical medium-range order in metallic glasses is exclusively regulated by the activation and inhibition of the constituent components, making the topology of metallic glasses a Turing pattern. The connecting schemes of atoms of the same species form three distinct regions, reflecting different correlations at the short and medium length scales, while the difference in the schemes corresponds to chemical ordering. By changing the elemental types, it is demonstrated that the chemical medium-range order strongly depends on the relative depth of the interatomic-potential wells. The study separates metallic glasses from crystals under the condition of negative heat of mixing by emphasizing their fundamental difference in interatomic potentials.

At present, China has reached the world-class level in the research and development of high-performance processors and supercomputers, such as the sunway many-core processor, which is the representative of Chinese high-performance processors. However, the hardware upgrade is difficult to bring about the improvement on the software performance in proportion, and its powerful computing resources cannot be used well at present. In general, sunway many-core processor uses MPE-CPE parallel optimization for hotspot. This paper improves the MPE-CPE parallel optimization, and research a set of new optimization methods, which includes a synchronous mechanism between the MPE (Management Processing Element) and CPE (Computing Processing Element) cluster and an acceleration method of multi-data partition in the CPE cluster. These methods solve the problem of insufficient utilization of CPEs' computing resources and insufficient coordination between MPE and CPEs to a certain extent. By employing the Sunway Taihulight supercomputer, the evaluation measures the execution times and speedups of the comprehensive parallel optimization and MPE-CPE parallel optimization. The result shows that, with different number of processes, the speedups of comprehensive parallel optimization are much higher than those of MPE-CPE parallel optimization, from 23.42% to 36.41%.

With the development of nuclear energy technology, reactor physical calculations have higher requirements for calculation accuracy and speed, and it has become an inevitable trend to use high-performance computers for reactor simulation calculations. The method of characteristics (MOC) is currently recognized as the preferred method for simulating neutron transport in the nuclear reactor core. Based on the architecture of Sunway many-core processor and Sunway Bluelight II supercomputer, this paper proposes a fine grained and universal two-level parallelization, including thread-level parallelization and process-level parallelization. In the thread-level parallelization, the methods such as job pipeline optimization, load balancing across CPEs, and I/O optimization are proposed for acceleration. In the process-level parallelization, a mapping method from software to hardware is proposed. This method can make full use of the hardware of Sunway supercomputers and improve the computing efficiency and data transmission efficiency. For the first time, the OpenMOC program is transplanted and parallelly optimized on the Sunway supercomputers, which enriched the application ecology of Sunway supercomputers. Compared with the original program, the two-level parallelization can achieve up to 18.6x speedup. Moreover, our parallelization is capable to run on more than 3750 processes of Sunway Bluelight II supercomputer with good strong and weak scalability.

Fourier moment expanded axial DGFEM (Discontinuous Galerkin Finite Element Method) SN based 2D/1D coupling method is a promising variant for the pin-by-pin neutron transport calculation, due to its excellent computational efficiency and stability. In this study, the incorporation of anisotropic scattering based on spherical harmonic expansion is developed for the axial DGFEM SN based 2D/1D transport approach, to further capture the anisotropic scattering accurately for the heterogeneous fuel rod or strong absorber. Several wellknown benchmarks are used to evaluate its computational performance. Numerical results shows that the implementation of anisotropic scattering can significantly improve the accuracy of 2D/1D coupling method. Moreover, compared with the traditional 1D SN based 2D/1D method, the newly developed method could achieve high computational efficiency and convergence rate, as well as a relatively low storage requirement.

Solar flares can release coronal magnetic energy explosively and may impact the safety of near-Earth space environments. Their structures and properties on the macroscale have been interpreted successfully by the generally accepted 2D standard model, invoking magnetic reconnection theory as the key energy conversion mechanism. Nevertheless, some momentous dynamical features as discovered by recent high-resolution observations remain elusive. Here, we report a self-consistent high-resolution 3D magnetohydrodynamical simulation of turbulent magnetic reconnection within a flare current sheet. It is found that fragmented current patches of different scales are spontaneously generated with a well-developed turbulence spectrum at the current sheet, as well as at the flare loop-top region. The close coupling of tearing mode and Kelvin–Helmholtz instabilities plays a critical role in developing turbulent reconnection and in forming dynamical structures with synthetic observables in good agreement with realistic observations. The sophisticated modeling makes a paradigm shift from the traditional to a 3D turbulent reconnection model unifying flare dynamical structures of different scales.

Tetrahedral stacking of atoms has great importance for melt stability, crystallization, and glass formation. For example, the spiral chain of close-packed tetrahedra was considered as a pseudonucleus for crystallization. Tetrahedral stacking can yield icosahedral or helical structures, which are seen in the self-organized superstructures of nano- or colloidal particles. Previous studies have mostly focused on tetrahedral packings that lead to a form with high-order symmetry, but imperfect tetrahedral stacking, which is common in practice, has rarely been studied. Here, we annealed pure copper in order to analyze the growth of a seeded 5-fold twinning structure made up of three tetrahedra over the incubation period. To do this, we used a superlarge system made up of 55 million atoms in molecular-dynamics simulations. The tetrahedral stacking gives rise to a chainlike structure that is capable of being described by a three-element coding scheme. The findings imply that the melt stability can be improved, even if the stacking structure of tetrahedra lacks high-order symmetry.

In recent years, learnable prompts have emerged as a major prompt learning paradigm, enhancing the performance of large-scale vision-language pre-trained models in few-shot image classification. However, enhancing methods are often time-consuming and inflexible because 1) class-specific prompts are inefficient in certain situations; 2) instance-specific prompts are put in a fixed position. To address these issues, inspired by the coarse-to-fine decision-making paradigm of human, we propose an Instance-Aware Hierarchical-Structured Policy (IAHSP) that integrates instance-specific prompt selection and appropriate position selection using a reinforcement learning fashion. Specifically, IAHSP consists of two sub-policies: 1) the root policy selects the most suitable prompt from the prompts pool, and 2) the leaf policy identifies the optimal position for inserting the selected prompt. We train these two policies iteratively with rewards constraining the prompts while maintaining their diversity. Extensive experiments on 11 public benchmarks demonstrate that our IAHSP significantly boosts the few-shot image classification performance of vision-language pre-trained models, while also exhibiting superior generalization performance.

Due to the strong geometric adaptability, the Three-Dimensional (3-D) Method Of Characteristics (MOC) is a promising candidate for 3-D whole-core high-fidelity neutron transport calculations. However, the 3-D MOC can hardly be applied to large-scale full core problems on account of the huge computational costs in memory and time. In the past, the 3-D MOC was mainly used in traditional Light Water Reactor (LWR) calculations, but rarely used for more complex core calculations. For pebble-bed High Temperature gas-cooled Reactors (HTRs), most 3D MOC codes are out of ability to construct complex pebble beds, let alone efficient acceleration methods. The 3D MOC code ARCHER can efficiently simulate the large-scale pebble-bed HTRs through using the Linear Source Approximation (LSA), the Coarse Mesh Finite Difference (CMFD) and the hybrid MPI-OpenMP parallel. In this work, the two-level CMFD acceleration and efficient preconditioners on Krylov subspace linear solvers are implemented in the ARCHER code in order to achieve better performance. In addition, the practical HTR-PM criticality problem is calculated by ARCHER, which is the world's first high-fidelity neutron transport solution of large-scale pebble-bed HTRs by deterministic numerical method. It also shows that the 3-D MOC has great application potential in other complex reactors.