Engineering Papers⌕ Search

Engineering topics

Li, Dong

Publications and source records attributed to Li, Dong.

At least 19 records

Anharmonic strong-coupling effects at the origin of the charge density wave in CsV 3 Sb 5

The formation of charge density waves is a long-standing open problem, particularly in dimensions higher than one. Various observations in the vanadium antimonides discovered recently further underpin this notion. Here, we study the Kagome metal CsV3Sb5 using polarized inelastic light scattering and density functional theory calculations. We observe a significant gap anisotropy with 2Δ max /κ B T CDW ≈ 20, far beyond the prediction of mean-field theory. The analysis of the A 1g and E 2g phonons, including those emerging below T CDW , indicates strong phonon-phonon coupling, presumably mediated by a strong electron-phonon interaction. Similarly, the asymmetric Fano-type lineshape of the A 1g amplitude mode suggests strong electron-phonon coupling below T CDW . The large electronic gap, the enhanced anharmonic phonon-phonon coupling, and the Fano shape of the amplitude mode combined are more supportive of a strong-coupling phonon-driven charge density wave transition than of a Fermi surface instability or an exotic mechanism in CsV 3 Sb 5 .

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Auto-HPCnet: An Automatic Framework to Build Neural Network-based Surrogate for High-Performance Computing Applications

High-performance computing communities are increasingly adopt- ing Neural Networks (NN) as surrogate models in their applications to generate scientific insights. Replacing an execution phase in the application with NN models can bring significant performance im- provement. However, there is a lack of tools that can help domain scientists automatically apply NN-based surrogate models to HPC applications. We introduce a framework, named Auto-HPCnet, to democratize the usage of NN-based surrogates. Auto-HPCnet is the first end-to-end framework that makes past proposals for the NN-based surrogate model practical and disciplined. Auto-HPCnet introduces a workflow to address unique challenges when apply- ing the approximation, such as feature acquisition and meeting the application-specific constraint on the quality of final computation outcome. We show that Auto-HPCnet can leverage NN for a set of HPC applications and achieve 5.50× speedup on average (up to 16.8× speedup and with data preparation cost included) while meeting the application-specific constraint on the final computation quality.

Dong, Wenqian↗

Emergent and robust ferromagnetic-insulating state in highly strained ferroelastic LaCoO 3 thin films

Transition metal oxides are promising candidates for the next generation of spintronic devices due to their fascinating properties that can be effectively engineered by strain, defects, and microstructure. An excellent example can be found in ferroelastic LaCoO 3 with paramagnetism in bulk. In contrast, unexpected ferromagnetism is observed in tensile-strained LaCoO 3 films, however, its origin remains controversial. Here we simultaneously reveal the formation of ordered oxygen vacancies and previously unreported long-range suppression of CoO 6 octahedral rotations throughout LaCoO 3 films. Supported by density functional theory calculations, we find that the strong modification of Co 3d-O 2p hybridization associated with the increase of both Co-O-Co bond angle and Co-O bond length weakens the crystal-field splitting and facilitates an ordered high-spin state of Co ions, inducing an emergent ferromagnetic-insulating state. Our work provides unique insights into underlying mechanisms driving the ferromagnetic-insulating state in tensile-strained ferroelastic LaCoO 3 films while suggesting potential applications toward low-power spintronic devices.

36 MATERIALS SCIENCE↗

Face mask integrated with flexible and wearable manganite oxide respiration sensor

Face masks are key personal protective equipment for reducing exposure to viruses and other environmental hazards such as air pollution. Integrating flexible and wearable sensors into face masks can provide valuable insights into personal and public health. The advantages that a breath-monitoring face mask requires, including multi-functional sensing ability and continuous, long-term dynamic breathing process monitoring, have been underdeveloped to date. Here, we design an effective human breath monitoring face mask based on a flexible La 0.7 Sr 0.3 MnO 3 (LSMO)/Mica respiration sensor. The sensor’s capabilities and systematic measurements are investigated under two application scenes, namely clinical monitoring mode and daily monitoring mode, to monitor, recognise, and analyse different human breath status, i.e., cough, normal breath, and deep breath. This sensing system exhibits super-stability and multi-modal capabilities in continuous and long-time monitoring of the human breath. We determine that during monitoring human breath, thermal diffusion in LSMO is responsible for the change of resistance in flexible LSMO/Mica sensor. Both simulated and experimental results demonstrate good discernibility of the flexible LSMO/Mica sensor operating at different breath status. Our work opens a route for the design of novel flexible and wearable electronic devices.

36 MATERIALS SCIENCE↗

A de novo matrix for macroscopic living materials from bacteria

Engineered living materials (ELMs) embed living cells in a biopolymer matrix to create materials with tailored functions. While bottom-up assembly of macroscopic ELMs with a de novo matrix would offer the greatest control over material properties, we lack the ability to genetically encode a protein matrix that leads to collective self-organization. Here we report growth of ELMs from Caulobacter crescentus cells that display and secrete a self-interacting protein. This protein formed a de novo matrix and assembled cells into centimeter-scale ELMs. Discovery of design and assembly principles allowed us to tune the composition, mechanical properties, and catalytic function of these ELMs. This work provides genetic tools, design and assembly rules, and a platform for growing ELMs with control over both matrix and cellular structure and function.

59 BASIC BIOLOGICAL SCIENCES↗

In vitro investigation of protein assembly by combined microscopy and infrared spectroscopy at the nanometer scale

The nanoscale structure and dynamics of proteins on surfaces has been extensively studied using various imaging techniques, such as transmission electron microscopy and atomic force microscopy (AFM) in liquid environments. These powerful imaging techniques, however, can potentially damage or perturb delicate biological material and do not provide chemical information, which prevents a fundamental understanding of the dynamic processes underlying their evolution under physiological conditions. Here, we use a platform developed in our laboratory that enables acquisition of infrared (IR) spectroscopy and AFM images of biological material in physiological liquids with nanometer resolution in a cell closed by atomically thin graphene membranes transparent to IR photons. In this work, we studied the self-assembly process of S-layer proteins at the graphene-aqueous solution interface. The graphene acts also as the membrane separating the solution containing the proteins and Ca 2+ ions from the AFM tip, thus eliminating sample damage and contamination effects. The formation of S-layer protein lattices and their structural evolution was monitored by AFM and by recording the amide I and II IR absorption bands, which reveal the noncovalent interaction between proteins and their response to the environment, including ionic strength and solvation. Our measurement platform opens unique opportunities to study biological material and soft materials in general.

59 BASIC BIOLOGICAL SCIENCES↗

Large Scale Caching and Streaming of Training Data for Online Deep Learning

The training of deep neural network models on large data remains a difficult problem, despite progress towards scalable techniques. In particular, there is a mismatch between the random but predetermined order in which AI flows select training samples and the streaming I/O patterns for which traditional HPC data storage (e.g., parallel file systems) are designed. In addition, as more data are obtained, it is feasible neither simply to train learning models incrementally, due to catastrophic forgetting (i.e., bias towards new samples), nor to train frequently from scratch, due to prohibitive time and/or resource constraints. In this paper, we study data management techniques that combine caching and streaming with rehearsal support in order to enable efficient access to training samples in both offline training and continual learning. We revisit state-of-art streaming approaches based on data pipelines that transparently handle prefetching, caching, shuffling, and data augmentation, and discuss the challenges and opportunities that arise when combining these methods with data-parallel training techniques. We also report on preliminary experiments that evaluate the I/O overheads involved in accessing the training samples from a parallel file system (PFS) under several concurrency scenarios, highlighting the impact of the PFS on the design of the data pipelines.

data pipelines↗

Manipulating the metal-to-insulator transition and magnetic properties in manganite thin films via epitaxial strain

Strain engineering of epitaxial transition metal oxide heterostructures offers an intriguing opportunity to control electronic structures by modifying the interplay between spin, charge, orbital, and lattice degrees of freedom. Here, we demonstrate that the electronic structure, magnetic and transport properties of La 0.9 Ba 0.1 MnO 3 thin films can be effectively controlled by epitaxial strain. Spectroscopic studies and first-principles calculations reveal that the orbital occupancy in Mn eg orbitals can be switched from the d 3z 2 -r 2 orbital to the d x 2 -y 2 orbital by varying the strain from compressive to tensile. Further, the change of orbital occupancy associated with Mn 3d-O 2p hybridization leads to dramatic modulation of the magnetic and electronic properties of strained La 0.9 Ba 0.1 MnO 3 thin films. Under moderate tensile strain, an emergent ferromagnetic insulating state with an enhanced ferromagnetic Curie temperature of 215 K is achieved. These findings not only deepen our understanding of electronic structures, magnetic and transport properties in the La 0.9 Ba 0.1 MnO 3 system, but also demonstrate the use of epitaxial strain as an effective knob to tune the electronic structures and related physical properties for potential spintronic device applications.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Dissecting the Shared Genetic Architecture of Suicide Attempt, Psychiatric Disorders, and Known Risk Factors

Suicide is a leading cause of death worldwide, and non-fatal suicide attempts, which occur far more frequently, are a major source of disability and social and economic burden. Both have substantial genetic etiology, which is partially shared and partially distinct from that of related psychiatric disorders. Methods: We conducted a genome-wide association study (GWAS) of 29,782 suicide attempt (SA) cases and 519,961 controls in the International Suicide Genetics Consortium. The GWAS of SA was conditioned on psychiatric disorders using GWAS summary statistics via mtCOJO, to remove genetic effects on SA mediated by psychiatric disorders. We investigated the shared and divergent genetic architectures of SA, psychiatric disorders and other known risk factors. Results: Two loci reached genome-wide significance for SA: the major histocompatibility complex and an intergenic locus on chromosome 7, which remained associated with SA after conditioning on psychiatric disorders and replicated in an independent cohort from the Million Veteran Program. This locus has been implicated in risk-taking, smoking, and insomnia. SA showed strong genetic correlation with psychiatric disorders, particularly major depression, and also with smoking, pain, risk-taking, sleep disturbances, lower educational attainment, reproductive traits, lower socioeconomic status and poorer general health. After conditioning on psychiatric disorders, the genetic correlations between SA and psychiatric disorders decreased, whereas those with non-psychiatric traits remained largely unchanged. Conclusions: Our results identify a risk locus that contributes more strongly to SA than other phenotypes and suggest a shared underlying biology between SA and known risk factors that is not mediated by psychiatric disorders.

60 APPLIED LIFE SCIENCES↗

Athena: High-Performance Sparse Tensor Contraction Sequence on Heterogeneous Memory

Sparse tensor contraction sequence has been widely employed in many fields, such as chemistry and physics. However, how to efficiently implement the sequence faces multiple challenges, such as redundant computations and memory operations, massive memory consumption, and inefficient utilization of hardware. To address the above challenges, we introduce Athena, a high-performance framework for SpTC sequences. Athena introduces new data structures, leverages emerging Optane-based heterogeneous memory (HM) architecture, and stage parallelism. In particular, Athena introduces shared hash table-represented sparse accumulator to eliminate unnecessary input processing and data migration; Athena uses a novel data-semantic guided dynamic migration solution to make the best use of the Optane-based HM for high performance; Athena also co-runs execution phases with different characteristics to enable high hardware utilization. Evaluating with 12 datasets, we show that Athena brings 327-7362× speedup over the state-of-the-art SpTC algorithm. With the dynamic data placement guided by data semantics, Athena brings performance improvement on Optane-based HM over a state-of-the-art software-based data management solution, a hardware-based data management solution, and PMM-only by 1.58×, 1.82×, and 2.34× respectively.

Liu, Jiawen↗

Trust: Triangle Counting Reloaded on GPUs

Triangle counting is a building block for a wide range of graph applications. Here, traditional wisdom suggests that i) hashing is not suitable for triangle counting, ii) edge-centric triangle counting beats vertex-centric design, and iii) communication-free and workload balanced graph partitioning is a grand challenge for triangle counting. On the contrary, we advocate that i) hashing can help the key operations for scalable triangle counting on Graphics Processing Units (GPUs), i.e., list intersection and graph partitioning, ii) vertex-centric option reduces both hash table construction cost and memory consumption, which is limited on GPUs. In addition, iii) we exploit graph and workload collaborative, and hash-based 2D partitioning to scale vertex-centric triangle counting over 1,000 GPUs with sustained scalability. In this work, we present TRUST, which performs triangle counting with the hash operation and vertex-centric paradigm. To the best of our knowledge, TRUST is the first work that achieves over one trillion Traversed Edges Per Second (TEPS) rate for triangle counting.

97 MATHEMATICS AND COMPUTING↗

PARIS: Predicting application resilience using machine learning

The traditional method to study application resilience to errors in HPC applications uses fault injection (FI), a time-consuming approach. Furthermore, while analytical models have been built to overcome the inefficiencies of FI, they lack accuracy. In this paper, we present PARIS, a machine-learning method to predict application resilience that avoids the time-consuming process of random FI and provides higher prediction accuracy than analytical models. PARIS captures the implicit relationship between application characteristics and application resilience, which is difficult to capture using most analytical models. We overcome many technical challenges for feature construction, extraction, and selection to use machine learning in our prediction approach. Our evaluation on 16 HPC benchmarks shows that PARIS achieves high prediction accuracy. PARIS is up to 450x faster than random FI (49x on average). Compared to the state-of-the-art analytical model, PARIS is at least 63% better in terms of accuracy and has comparable execution time on average.

97 MATHEMATICS AND COMPUTING↗

Smart-PGSim: Using Neural Network to Accelerate AC-OPF Power Grid Simulation

In this work we address the problem of accelerating complex power-grid simulation through machine learning ( ML). Specifically, we develop a framework, Smart-PGSim,which generates multitask-learning (MTL) neural network (NN)models to predict the initial values of variables critical to the problem convergence. MTL models allow information sharing when predicting multiple dependent variables while including customized layers to predict individual variables. We show that,to achieve the required accuracy, it is paramount to embed domain-specific constraints derived from the specific power-grid components in the MTL model. Smart-PGSim then employs the predicted initial values as a high-quality initial condition for the power-grid numerical solver (warm start), resulting in both higher performance compared to state-of-the-art solutions while maintaining the required accuracy. Smart-PGSim brings 2.60×speedup on average (up to 3.28×) computed over 10,000 problems, without losing solution optimality.

machine learning, neural networks↗

Engineering the electronic and strained interface for high activity of PdM core @Pt monolayer electrocatalysts for oxygen reduction reaction

Alloyed nanoparticles with core-shell structures provide a favorable model to modulate interfacial interaction and surface structures at the atomic level, which is important for designing electrocatalysts with high activity and durability. Herein, core-shell structured Pd 3 M@Pt/C nanoparticles with binary PdM alloy cores (M = Fe, Ni, and Co) and a monolayer Pt shell were successfully synthesized with diverse interfaces. Among these, Pd 3 Fe@Pt/C exhibited the best oxygen reduction reaction catalytic performance, roughly 5.4 times more than that of the commercial Pt/C catalyst used as reference. The significantly enhanced activity is attributed to the combined effects of strain engineering, interfacial electron transfer, and improved Pt utilization. Density functional theory simulations and extended X-ray absorption fine structure analysis revealed that engineering the alloy core with moderate lattice mismatch and alloy composition (Pd 3 Fe) optimizes the surface oxygen adsorption energy, thereby rendering excellent electrocatalytic activity. Finally, future researches may use this study as a guide on the construction of highly effective core-shell electrocatalysts for various energy conversions and other applications.

25 ENERGY STORAGE↗