Search NASA⌕ Search

Engineering topics

Goswami, Anshuman

Publications and source records attributed to Goswami, Anshuman.

Mean field model parameterization to recover coverage-dependent kinetics

Lateral interactions between adsorbates introduce coverage dependence into adsorption energies and activation barriers of surface reactions. Lattice-based kinetic Monte Carlo (kMC) simulations can capture these interactions quantitatively but are laborious to parameterize and solve. Mean field models are more tractable, but protocols to construct and parameterize them are unclear. Here we explore the ability of a coverage-aware mean-field model to map to a lattice-kMC model of a generic two-step reaction network, including quasi-equilibrated adsorption and rate-limiting dissociation steps. We derive expressions for mean-field and coverage-dependent adsorption energies and dissociation barriers and parameterize against lattice-kMC predictions. We show that the parameterized mean-field rates correlate with ground truth lattice-kMC results across a wide range of reaction conditions and identify regions where the mean field fails. The mean field model similarly captures kMC-derived rate-order, Arrhenius and Sabatier plots at a greatly reduced computational cost. Further, the results provide guidance for parameterizing mean-field models, benchmarked against explicit lattice-based approaches for incorporating the influence of coverage effects.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A redox model for NO oxidation, NH 3 oxidation and high temperature standard SCR over Cu-SSZ-13

A kinetic model is developed to predict the influence of temperature and hydrothermal aging on the redox of active Cu sites under standard SCR, NO oxidation and NH 3 oxidation conditions over a practically relevant fully-formulated Cu-SSZ-13 catalyst. NO 2 /N 2 production during NO/NH 3 titration of Cu II sites is utilized to identify rate parameters associated with NO-only RHC (reduction half cycle) and NH 3 -only RHC respectively. Integral N 2 formation during subsequent NO + NH 3 titration is consistent with the production of one NO 2 per two CuII sites reduced during NO-only RHC and one N 2 per six Cu II sites reduced during NH 3 -only RHC. Decreased reduction of Cu II sites by NO/NH 3 upon hydrothermal aging, along with the production of one NO 2 per two CuII sites during NO-only RHC, is accordant with the involvement of proximal ZCuOH and oxygen-bridged dimeric Cu II sites. Oxidation of partially solvated and framework coordinated ZCu (Cu I ) sites occurs in presence of O 2 , does not produce N 2 and can lead to the consumption of Brønsted acid sites. A global OHC kinetic model is developed to predict transient and integral N 2 formation during exposure of CuI sites to a mixture of NO and O 2 . The resulting redox kinetic model quantitatively predicts NO and NH 3 consumption during isothermal transient response Cu redox (TRCR) protocols, along with temperature and age dependent steady-state standard SCR and oxidation conditions. The redox model presented in this work synthesizes recent kinetic, spectroscopic and computational findings to provide a foundational description of active site redox during standard SCR, NO oxidation and NH3 oxidation over Cu-SSZ-13.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Plasma-Catalyst Reactivity Control of Surface Nitrogen Species through Plasma-Temperature-Programmed Hydrogenation to Ammonia

Nonthermal plasma activation of N 2 can facilitate nitrogen adsorption on metal catalysts at low bulk temperatures and atmospheric pressure. Here, we apply a plasma-assisted temperature-programmed reaction (plasma-TPRxn) for ammonia (NH 3 ) synthesis using sequential exposure of a silica-supported metal catalyst to N 2 plasma followed by thermal hydrogen treatment while ramping the temperature to decouple the plasma activation of N 2 from surface catalyzed hydrogenation steps. This approach eliminates the effects from bulk plasma phase reactions, thereby allowing for direct interrogation of plasma activated nitrogen on the active metal surfaces. We confirm previously reported spectroscopic observations that show plasma-generated surface nitrogen can be converted to NH 3 through surface catalyzed pathways. Further, we demonstrate that the ammonia desorption peak temperature is sensitive to metal, with Pt desorbing NH 3 at the lowest temperature. Unsteady state microkinetic models of desorption kinetics as a function of initial N coverage and metal recover observed trends in NH 3 desorption temperatures and confirm that observed results reflect hydrogenation of plasma-induced N accommodation at each surface. In total, we show that the hydrogenation ability of the catalyst after plasma activation of N 2 is responsible for the reactivity trends observed in plasma-assisted NH 3 synthesis.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Consequences of adsorbate-adsorbate interactions for apparent kinetics of surface catalytic reactions

Lateral adsorbate interactions at catalyst surfaces are known to influence adsorption energies and reaction rates. Lattice-based kinetic Monte Carlo (kMC) simulations are able to capture these influences, but such models are typically parameterized for a specific reaction network and catalyst surface. Here we report kMC simulations to probe the influence of lateral interactions on simulated rates, rate orders, apparent activation energies, and Sabatier plots. We construct a simple, two-step reaction network involving a single adsorbate and rate- limiting diatomic dissociation, employ a generic repulsive lateral interaction model consistent with known adsorbate-adsorbate interactions on metal surfaces, and a rate model consistent with known Brønsted-Evans-Polyani relationships for diatomic dissocations. Furthermore, we juxtapose reaction kinetics over a wide range of reaction conditions and catalyst binding energies, as a function of interaction strength. We find that at a given zero-coverage binding energy and external conditions, adsorbate coverage decreases monotonically with increasing interaction strength, but absolute rates can vary linearly or nonlinearly. Interactions flatten the Sabatier volcano and shift the maximum towards stronger binding. Influences on apparent rate orders and activation energies are modest and are sensitive to interaction-induced adsorbate ordering. Model predictions are sensitive to lattice size effects at high coverages. The results, modelled for a simple reaction system, highlight the generic consequences of lateral interactions and guidance for identifying their signatures in observed kinetics.

kinetic Monte Carlo↗

GPU-Centric Communication on NVIDIA GPU Clusters with InfiniBand: A Case Study with OpenSHMEM

GPUs have become an essential component for building compute clusters with high compute density and high performance per watt. As such clusters scale to have 1000s of GPUs, efficiently moving data between the GPUs becomes imperative to get maximum performance. NVSHMEM is an implementation of the OpenSHMEM standard for NVIDIA GPU clusters which allows communication to be issued from inside GPU kernels. In earlier work, we have shown how NVSHMEM can be used to achieve better application performance on GPUs connected through PCIe or NVLink. As part of this effort, we implement IB verbs for Mellanox InfiniBand adapters in CUDA. We evaluate different design alternatives, taking into consideration the relaxed memory model, automatic memory access coalescing and thread hierarchy on the GPU. We also consider correctness issues that arise in these designs. We take advantage of these designs transparently or through API extensions in NVSHMEM. With micro-benchmarks, we show that a Nvidia Pascal P100 GPU is able saturate the network bandwidth using only one or two of its 56 available streaming multiprocessors (SM). On a single GPU using a single IB EDR adapter, we achieve a throughput of around 90 million messages per second. In addition, we implement a 2dstencil application kernel using NVSHMEM and compare its performance with a CUDA-aware MPI-based implementation that uses GPUDirect RDMA. Speedups in the range of 23% to 42% are seen for input sizes large enough to fill the occupancy of Nvidia Pascal P100 GPUs on 2 to 4 nodes indicating that there are gains to be had by eliminating the CPU from the communication path when all computation runs on the GPU.

Potluri, Sreeram↗

Efficient Breadth First Search on Multi-GPU Systems Using GPU-Centric OpenSHMEM

NVSHMEM is an implementation of OpenSHMEM for NVIDIA GPUs which allows communication to be issued from inside CUDA kernels. In this work, we present an implementation of Breadth First Search for multi-GPU systems using NVSHMEM. We analyze the benefits and bottlenecks of moving fine-grained communication into CUDA kernels. Using our implementation of BFS, we achieve up to 75% improvement in performance compared to a CUDA-aware MPI-based implementation, in the best case.

97 MATHEMATICS AND COMPUTING↗

Landrush: Rethinking In-Situ Analysis for GPGPU Workflows

In-situ analysis on the output data of scientific simulations has been made necessary by ever-growing output data volumes and increasing costs of data movement as supercomputing is moving towards exascale. With hardware accelerators like GPUs becoming increasingly common in high end machines, new opportunities arise to co-locate scientific simulations and online analysis performed on the scientific data generated by the simulations. However, the asynchronous nature of GPGPU programming models and the limited context-switching capabilities on the GPU pose challenges to co-locating the scientific simulation and analysis on the same GPU. This paper dives deeper into these challenges to understand how best to co-locate analysis with scientific simulations on the GPUs in HPC clusters. Specifically, our 'Landrush' approach to GPU sharing proposes a solution that utilizes idle cycles on the GPU to provide an improved time-to-answer, that is, the total time to run the scientific simulation and analysis of the generated data. Landrush is demonstrated with experimental results obtained from leadership high-end applications on ORNL's Titan supercomputer, which show that (i) GPU-based scientific simulations have varying degrees of idle cycles to afford useful analysis task co-location, and (ii) the inability to context switch on the GPU at instruction granularity can be overcome by careful control of the analysis kernel launches and software-controlled early completion of analysis kernel executions. Results show that Landrush is superior in terms of time-to-answer compared to serially running simulations followed by analysis or by relying on the GPU driver and hardwired thread dispatcher to run analysis concurrently on a single GPU.

Computer Science↗