Search NASA⌕ Search

SEARCH · Search NASA

Results for “bottleneck structure”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 379 records · Page 21

Thermophilic Chassis-Enabled High-Throughput Selection of a Thermostable Fluorogenic Reporter

Thermostable proteins show increased shelf life and performance at elevated temperatures and under harsh conditions, resulting in lower costs for various industrial and biotechnological applications. However, due to a limited understanding of the relationship between stability and function, protein stabilization remains primarily a trial-and-error approach. Therefore, building a combinatorial library of mutations predicted to improve stability, followed by experimental testing, represents a markedly improved methodology. However, the lack of high-throughput approaches to screen even a moderately sized library presents a major bottleneck in the field. Here, in this study, we use a thermophile, Parageobacillus thermoglucosidasius (Ptherm) to rapidly screen combinatorial libraries consisting of rationally designed thermostabilizing mutations (∼10 3 –10 4 ) of a mesophilic fluorescent reporter, Y-FAST. On a Petri dish, microbial growth at an elevated temperature and exposure to fluorogen yielded several colonies of Ptherm that showed distinct fluorescence at 55 and 68 °C in our two sequentially generated libraries using Rosetta and ProteinMPNN, respectively. The Y-FAST variants isolated from fluorescent colonies were brighter than Y-FAST and showed higher resistance to thermal and chemical denaturation. AlphaFold-predicted structures and MD simulations revealed stability-enhancing salt bridges and hydrogen bond networks in the isolated FAST variants. The moderately thermostable FAST (tsFAST) and hyperstable FAST (hsFAST) were then demonstrated as translation reporters for protein expression and folding at elevated temperatures, such as 55 and 68 °C. Our approach of combinatorial library generation and high-throughput screening in a thermophilic chassis could, in principle, be extended to other proteins fused to these translation reporters. Furthermore, the hsFAST protein is small─half the size of the green fluorescent protein─and does not require oxygen for maturation, making it ideal for engineering extremophilic anaerobes for biosensing and bioconversion.

59 BASIC BIOLOGICAL SCIENCES↗

A combined electron beam/optical lithography process step for the fabrication of sub-half-micron-gate-length MMIC chips

Advanced radar and communication systems rely heavily on state-of-the-art microelectronics. Systems such as the phased-array radar require many transmit/receive (T/R) modules which are made up of many millimeter wave - microwave integrated circuits (MMIC's). The heart of a MMIC chip is the Gallium Arsenide (GaAs) field-effect transistor (FET). The transistor gate length is the critical feature that determines the operating frequency of the radar system. A smaller gate length will typically result in a higher frequency. In order to make a phased array radar system economically feasible, manufacturers must be capable of producing very large quantities of small-gate-length MMIC chips at a relatively low cost per chip. This requires the processing of a large number of wafers with a large number of chips per wafer, minimum processing time, and a very high chip yield. One of the bottlenecks in the fabrication of MIMIC chips is the transistor gate definition. The definition of sub-half-micron gates for GaAs-based field-effect transistors is generally performed by direct-write electron beam lithography (EBL). Because of the throughput limitations of EBL, the gate-layer fabrication is conventionally divided into two lithographic processes where EBL is used to generate the gate fingers and optical lithography is used to generate the large-area gate pads and interconnects. As a result, two complete sequences of resist application, exposure, development, metallization and lift-off are required for the entire gate structure. We have baselined a hybrid process, referred to as EBOL (electron beam/optical lithography), in which a single application of a multi-level resist is used for both exposures. The entire gate structure, (gate fingers, interconnects and pads), is then formed with a single metallization and lift-off process. The EBOL process thus retains the advantages of the high-resolution E-beam lithography and the high throughput of optical lithography while essentially eliminating an entire lithography/metallization/lift-off process sequence. This technique has been proven to be reliable for both trapezoidal and mushroom gates and has been successfully applied to metal-semiconductor and high-electron-mobility field-effect transistor (MESFET and HEMT) wafers containing devices with gate lengths down to 0.10 micron and 75 x 75 micron gate pads. The yields and throughput of these wafers have been very high with no loss in device performance. We will discuss the entire EBOL process technology including the multilayer resist structure, exposure conditions, process sensitivities, metal edge definition, device results, comparison to the standard gate-layer process, and its suitability for manufacturing.

Sewell, James S.↗

Machine Learning-enabled Scalable Performance Prediction of Scientific Codes

Hardware architectures become increasingly complex as the compute capabilities grow to exascale. Here, we present the Analytical Memory Model with Pipelines (AMMP) of the Performance Prediction Toolkit (PPT). PPT-AMMP takes high-level source code and hardware architecture parameters as input and predicts runtime of that code on the target hardware platform, which is defined in the input parameters. PPT-AMMP transforms the code to an (architecture-independent) intermediate representation, then (i) analyzes the basic block structure of the code, (ii) processes architecture-independent virtual memory access patterns that it uses to build memory reuse distance distribution models for each basic block, and (iii) runs detailed basic-block level simulations to determine hardware pipeline usage. PPT-AMMP uses machine learning and regression techniques to build the prediction models based on small instances of the input code, then integrates into a higher-order discrete-event simulation model of PPT running on Simian PDES engine. We validate PPT-AMMP on four standard computational physics benchmarks and present a use case of hardware parameter sensitivity analysis to identify bottleneck hardware resources on different code inputs. We further extend PPT-AMMP to predict the performance of a scientific application code, namely, the radiation transport mini-app SNAP. To this end, we analyze multi-variate regression models that accurately predict the reuse profiles and the basic block counts. We validate predicted SNAP runtimes against actual measured times.

97 MATHEMATICS AND COMPUTING↗

Domain-aware Control-oriented Neural Models for Autonomous Underwater Vehicles

Conventional physics-based modeling is a time-consuming bottleneck in control design for complex nonlinear systems like autonomous underwater vehicles (AUVs). In contrast, purely data-driven models, require a large number of observations and lack operational guarantees for safety-critical systems. Data-driven models leveraging available partially characterized dynamics have potential to provide reliable systems models in a typical data-limited scenario for high value complex systems, thereby avoiding months of expensive expert modeling time. In this work we explore this middle-ground between expert-modeled and pure data-driven modeling. We present control-oriented parametric models with varying levels of domain-awareness that exploit known system structure and prior physics knowledge to create constrained deep neural dynamical system models. We employ universal differential equations to construct data-driven blackbox and graybox representations of the AUV dynamics. In addition, we explore a hybrid formulation that explicitly models the residual error related to imperfect graybox models. We compare the prediction performance of the learned models for different distributions of initial conditions and control inputs to assess their suitability for control.

Shaw Cortez, Wenceslao E.↗

AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing

The recent development of deep learning has been mostly focusing on Euclidean data, such as images, videos, audios, etc. However, most real-world information and relation are often expressed as graphs. To efficiently learn from graph data, graph convolutional networks (GCNs) emerge as a promising approach, showing advantages in several practical applications such as social network analysis, knowledge discovery, 3D modeling, motion capturing, etc. Real-world graphs are usually extremely large and imbalanced, posting significant performance demand and design challenges on the hardware dedicated for GCN inference. In this paper, we propose an architecture design called UW-GCN to accelerate graph convolutional network inference. To tackle the major performance bottleneck from workload imbalance, we propose dynamic neighborhood stealing and remote chunk shuffling techniques, relying on hardware flexibility to achieve hardware auto-tuning under negligible area or delay overhead. Specifically, UW-GCN is able to smartly profile the sparse graph pattern while continuously adjusting the workload distribution via routing reconfiguration among parallel processing elements (PEs). The ideal configuration is then reused in the remaining iterations. To the best of our knowledge, this is the first accelerator design particularly for GCN and the first work relying on hardware auto-tuning, which is normally based on software, to achieve near-optimal workload balance in processing sparse structures.

Geng, Tong↗

Origin of rate limitations in solid-state polymer batteries from constrained segmental dynamics within the cathode

Currently, one of the key challenges preventing the wide commercialization of polymer-based solid-state batteries is achieving high-rate capabilities. This work unravels chain structure and segmental dynamics in a polymer-based composite cathode consisting of LiFePO 4 (LFP), carbon, and poly(ethylene oxide) (PEO) with lithium bis(trifluoromethanesulfonyl)imide. Small-angle neutron scattering (SANS) data reveal that PEO chains are adsorbed on the surface of LFP particles during slurry processing to make the composite electrode. The strong interaction between LFP and PEO chains leads to greatly reduced segmental dynamics of PEO, as discovered by quasi-elastic neutron scattering (QENS). The reduced segmental dynamics results in 70% decreased Li + mobility of the polymer electrolyte in the composite cathode. The combined SANS and QENS study indicates that one of the key bottlenecks that limits the rate performance of PEO-based polymer batteries originates from molecular interactions within the cathode.

25 ENERGY STORAGE↗

Performance Models for the Spike Banded Linear System Solver

With availability of large-scale parallel platforms comprised of tens-of-thousands of processors and beyond, there is significant impetus for the development of scalable parallel sparse linear system solvers and preconditioners. An integral part of this design process is the development of performance models capable of predicting performance and providing accurate cost models for the solvers and preconditioners. There has been some work in the past on characterizing performance of the iterative solvers themselves. In this paper, we investigate the problem of characterizing performance and scalability of banded preconditioners. Recent work has demonstrated the superior convergence properties and robustness of banded preconditioners, compared to state-of-the-art ILU family of preconditioners as well as algebraic multigrid preconditioners. Furthermore, when used in conjunction with efficient banded solvers, banded preconditioners are capable of significantly faster time-to-solution. Our banded solver, the Truncated Spike algorithm is specifically designed for parallel performance and tolerance to deep memory hierarchies. Its regular structure is also highly amenable to accurate performance characterization. Using these characteristics, we derive the following results in this paper: (i) we develop parallel formulations of the Truncated Spike solver, (ii) we develop a highly accurate pseudo-analytical parallel performance model for our solver, (iii) we show excellent predication capabilities of our model – based on which we argue the high scalability of our solver. Our pseudo-analytical performance model is based on analytical performance characterization of each phase of our solver. These analytical models are then parameterized using actual runtime information on target platforms. An important consequence of our performance models is that they reveal underlying performance bottlenecks in both serial and parallel formulations. All of our results are validated on diverse heterogeneous multiclusters – platforms for which performance prediction is particularly challenging. Finally, we provide predict the scalability of the Spike algorithm using up to 65,536 cores with our model. In this paper we extend the results presented in the Ninth International Symposium on Parallel and Distributed Computing.

Manguoglu, Murat↗

Cross-scale Efficient Tensor Contractions for Coupled Cluster Computations Through Multiple Programming Model Backends

Coupled-cluster methods provide highly accurate models of molecular structure by explicit numerical calculation of tensors representing the correlation between electrons. These calculations are dominated by a sequence of tensor contractions, motivating the development of numerical libraries for such operations. While based on matrix-matrix multiplication, these libraries are specialized to exploit symmetries in the molecular structure and in electronic interactions, and thus reduce the size of the tensor representation and the complexity of contractions. The resulting algorithms are irregular and their parallelization has been previously achieved via the use of dynamic scheduling or specialized data decompositions. We introduce our efforts to extend the Libtensor framework to work in the distributed memory environment in a scalable and energy efficient manner. We achieve up to 240 speedup compared with the best optimized shared memory implementation. We attain scalability to hundreds of thousands of compute cores on three distributed-memory architectures, (Cray XC30&XC40, BlueGene/Q), and on a heterogeneous GPU-CPU system (Cray XK7). As the bottlenecks shift from being compute-bound DGEMM's to communication-bound collectives as the size of the molecular system scales, we adopt two radically different parallelization approaches for handling load-imbalance. Nevertheless, we preserve a uni ed interface to both programming models to maintain the productivity of computational quantum chemists.

74 ATOMIC AND MOLECULAR PHYSICS↗

Halide Perovskite High- k Field Effect Transistors with Dynamically Reconfigurable Ambipolarity

Despite the remarkable optoelectronic properties of halide perovskites, achieving reproducible field effect transistor (FET) action in polycrystalline films at room temperature has been challenging and represents a fundamental bottleneck for understanding electronic charge transport in these materials. In this work, we report halide perovskite-based FET operation at room temperature with negligible hysteresis. Extensive measurements and device modeling reveal that incorporating high-k dielectrics enables modulation of the channel conductance. Furthermore, continuous bias cycling or resting allows dynamical reconfiguration of the FETs between p-type behavior and ambipolar FET with balanced electron and hole transport and an ON/OFF ratio up to 104 and negligible degradation in transport characteristics over 100 cycles. Furthermore, these results elucidate the path for achieving gate modulation in perovskite thin films and provide a platform to understand the interplay between the perovskite structure and external stimuli such as photons, fields, and functional substrates, which will lead to novel and emergent properties.

36 MATERIALS SCIENCE↗

Exploring a new synthesis route to lithium-excess disordered rock salt (DRX) cathode materials

Lithium-excess disordered rock salt (DRX) materials are promising candidates for Co/Ni-free Li-ion cathodes due to their high specific energy (800+ W h kg −1 ) and compositional flexibility. DRX cathodes are typically synthesized using solid-state reactions, which are difficult to scale and provide little-to-no control over particle morphology. Here, to address this bottleneck, the present study reports a two-step, solution-based reaction route to prepare Mn/Ti-based DRX oxyfluoride cathodes with nominal compositions of Li 1.25 Mn 0.5 Ti 0.3 O 1.95 F 0.05 and Li 1.35 Mn 0.7 Ti 0.1 O 1.85 F 0.15 . More specifically, a glycine–nitrate combustion reaction is used to produce a lithiated transition metal oxide, which is further reacted with LiF to produce high-purity DRX powders. Remarkably, this route yields 80–90% pure DRX after annealing for 1 h at 800–1000 °C, and 19 F solid-state nuclear magnetic resonance (ssNMR) spectra demonstrate that F− anions are successfully incorporated into the DRX structure. Cathodes prepared using this approach exhibit promising electrochemical performance, with Li 1.35 Mn 0.7 Ti 0.1 O 1.85 F 0.15 attaining reversible capacities ∼210 mA h g −1 and moderate cycling stability in half cells (65% capacity retention over 150 cycles). Overall, these results demonstrate that utilizing novel metal oxide precursors presents a viable and largely unexplored method to produce high-performance Co/Ni-free DRX cathodes.

25 ENERGY STORAGE↗

Electronic heat generation in semiconductors: Non-equilibrium excitation and evolution of zone-edge phonons via electron–phonon scattering in photo-excited germanium

We investigate experimentally and using first-principles theory the generation of phonons and the relaxation of carriers on picosecond timescales across the Brillouin zone of photo-excited Ge by inter-valley electron–phonon scattering. The phonons generated are typical of those generated in semiconductor devices, contributing to the accumulation of heat within the material. We simulate the time-evolution of phonon populations, based on first-principles band structure and electron–phonon and phonon–phonon matrix elements, and compare them to data from time-resolved x-ray diffuse scattering experiments, performed at the Linac Coherent Light Source x-ray free-electron laser facility, following photo-excitation by a 50 fs near-infrared optical pulse. We show that the intensity of the non-thermal x-ray diffuse scattering signal, which is observed to grow substantially near the L-point of the Brillouin zone over 3–5 ps, is due to phonons generated by scattering of carriers between the Δ and L valleys. These phonons have low group velocities, resulting in a heat bottleneck. With the inclusion of phonon decay through 3-phonon processes, the simulations also account for other non-thermal features observed in the x-ray diffuse scattering intensity, which are due to anharmonic phonon–phonon scattering of the phonons initially generated by electron–phonon scattering.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Multitask methods for predicting molecular properties from heterogeneous data

Data generation remains a bottleneck in training surrogate models to predict molecular properties. We demonstrate that multitask Gaussian process regression overcomes this limitation by leveraging both expensive and cheap data sources. In particular, we consider training sets constructed from coupled-cluster (CC) and density functional theory (DFT) data. We report that multitask surrogates can predict at CC-level accuracy with a reduction in data generation cost by over an order of magnitude. Of note, our approach allows the training set to include DFT data generated by a heterogeneous mix of exchange–correlation functionals without imposing any artificial hierarchy on functional accuracy. More generally, the multitask framework can accommodate a wider range of training set structures—including the full disparity between the different levels of fidelity—than existing kernel approaches based on Δ-learning although we show that the accuracy of the two approaches can be similar. Consequently, multitask regression can be a tool for reducing data generation costs even further by opportunistically exploiting existing data sources.

Chemistry↗

Growing Larger Crystals for Neutron Diffraction

Obtaining crystals of suitable size and high quality has been a major bottleneck in macromolecular crystallography. With the advent of advanced X-ray sources and methods the question of size has rapidly dwindled, almost to the point where if one can see the crystal then it was big enough. Quality is another issue, and major national and commercial efforts were established to take advantage of the microgravity environment in an effort to obtain higher quality crystals. Studies of the macromolecule crystallization process were carried out in many labs in an effort to understand what affected the resultant crystal quality on Earth, and how microgravity improved the process. While technological improvements are resulting in a diminishing of the minimum crystal size required, neutron diffraction structural studies still require considerably larger crystals, by several orders of magnitude, than X-ray studies. From a crystal growth physics perspective there is no reason why these 'large' crystals cannot be obtained: the question is generally more one of supply than limitations mechanism. This talk will discuss our laboratory s current model for macromolecule crystal growth, with highlights pertaining to the growth of crystals suitable for neutron diffraction studies.

Pusey, Marc↗

Extended cycle life implications of fast charging for lithium-ion battery cathode

Enabling extreme fast charging (XFC, =10–15 min charging) requires a comprehensive understanding of its implications. While lithium plating is a key bottleneck for the anode, the full extent of limitations for the cathode are not well-understood, particularly in extended-cycle settings with well-defined battery designs and conditions. This article presents cycle-life implications of XFC on cathodes at multiple length scales, combining electrochemical analyses, degradation modeling, and post-test characterizations. The comprehensive test matrix includes 41 well-defined gr/NMC pouch cells under varied fast-charge rates (1–9C) and state-of-charges cycled up to 1000 times. Cathode issues remain minimal in early cycling, but begin to accelerate in later life, when distinct cracking is found and identified as a fatigue mechanism. The bulk structure of cathodes remains intact, but distinct particle surface reconstruction is observed; however, this shows less pronounced effect on cathode aging than does cracking.

25 ENERGY STORAGE↗

Emerging CFD technologies and aerospace vehicle design

With the recent focus on the needs of design and applications CFD, research groups have begun to address the traditional bottlenecks of grid generation and surface modeling. Now, a host of emerging technologies promise to shortcut or dramatically simplify the simulation process. This paper discusses the current status of these emerging technologies. It will argue that some tools are already available which can have positive impact on portions of the design cycle. However, in most cases, these tools need to be integrated into specific engineering systems and process cycles to be used effectively. The rapidly maturing status of unstructured and Cartesian approaches for inviscid simulations makes suggests the possibility of highly automated Euler-boundary layer simulations with application to loads estimation and even preliminary design. Similarly, technology is available to link block structured mesh generation algorithms with topology libraries to avoid tedious re-meshing of topologically similar configurations. Work in algorithmic based auto-blocking suggests that domain decomposition and point placement operations in multi-block mesh generation may be properly posed as problems in Computational Geometry, and following this approach may lead to robust algorithmic processes for automatic mesh generation.

Aftosmis, Michael J.↗

Vertex protein PduN tunes encapsulated pathway performance by dictating bacterial metabolosome morphology

Engineering subcellular organization in microbes shows great promise in addressing bottlenecks in metabolic engineering efforts; however, rules guiding selection of an organization strategy or platform are lacking. Here, we study compartment morphology as a factor in mediating encapsulated pathway performance. Using the 1,2-propanediol utilization microcompartment (Pdu MCP) system from Salmonella enterica serovar Typhimurium LT2, we find that we can shift the morphology of this protein nanoreactor from polyhedral to tubular by removing vertex protein PduN. Analysis of the metabolic function between these Pdu microtubes (MTs) shows that they provide a diffusional barrier capable of shielding the cytosol from a toxic pathway intermediate, similar to native MCPs. However, kinetic modeling suggests that the different surface area to volume ratios of MCP and MT structures alters encapsulated pathway performance. Finally, we report a microscopy-based assay that permits rapid assessment of Pdu MT formation to enable future engineering efforts on these structures.

59 BASIC BIOLOGICAL SCIENCES↗

Computationally Efficient Multiscale Neural Networks Applied to Fluid Flow in Complex 3D Porous Media

Abstract The permeability of complex porous materials is of interest to many engineering disciplines. This quantity can be obtained via direct flow simulation, which provides the most accurate results, but is very computationally expensive. In particular, the simulation convergence time scales poorly as the simulation domains become less porous or more heterogeneous. Semi-analytical models that rely on averaged structural properties (i.e., porosity and tortuosity) have been proposed, but these features only partly summarize the domain, resulting in limited applicability. On the other hand, data-driven machine learning approaches have shown great promise for building more general models by virtue of accounting for the spatial arrangement of the domains’ solid boundaries. However, prior approaches building on the convolutional neural network (ConvNet) literature concerning 2D image recognition problems do not scale well to the large 3D domains required to obtain a representative elementary volume (REV). As such, most prior work focused on homogeneous samples, where a small REV entails that the global nature of fluid flow could be mostly neglected, and accordingly, the memory bottleneck of addressing 3D domains with ConvNets was side-stepped. Therefore, important geometries such as fractures and vuggy domains could not be modeled properly. In this work, we address this limitation with a general multiscale deep learning model that is able to learn from porous media simulation data. By using a coupled set of neural networks that view the domain on different scales, we enable the evaluation of large ( $$>512^3$$ > 512 3 ) images in approximately one second on a single graphics processing unit. This model architecture opens up the possibility of modeling domain sizes that would not be feasible using traditional direct simulation tools on a desktop computer. We validate our method with a laminar fluid flow case using vuggy samples and fractures. As a result of viewing the entire domain at once, our model is able to perform accurate prediction on domains exhibiting a large degree of heterogeneity. We expect the methodology to be applicable to many other transport problems where complex geometries play a central role.

36 MATERIALS SCIENCE↗