skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: SQERTSS: Dynamic rank based throttling of transition probabilities in kinetic Monte Carlo simulations

Abstract

Lattice based Kinetic Monte Carlo (KMC) simulations offer a powerful simulation technique for investigating large reaction networks while retaining spatial configuration information, unlike ordinary differential equations. However, large chemical reaction networks can contain reaction processes with rates spanning multiple orders of magnitude. This can lead to the problem of “KMC stiffness” (similar to stiffness in differential equations), where the computational expense has the potential to be overwhelmed by very short time-steps during KMC simulations, with the simulation spending an inordinate amount of KMC steps / cpu-time simulating fast frivolous processes (FFPs) without progressing the system (reaction network). In order to achieve simulation times that are experimentally relevant or desired for predictions, a dynamic throttling algorithm involving separation of the processes into speed-ranks based on event frequencies has been designed and implemented with the intent of decreasing the probability of FFP events, and increasing the probability of slow process events -- allowing rate limiting events to become more likely to be observed in KMC simulations. This Staggered Quasi-Equilibrium Rank-based Throttling for Steady-state (SQERTSS) algorithm designed for use in achieving and simulating steady-state conditions in KMC simulations. Lastly, as shown in this work, the SQERTSS algorithm also works for transient conditions: themore » correct configuration space and final state will still be achieved if the required assumptions are not violated, with the caveat that the sizes of the time-steps may be distorted during the transient period.« less

Authors:
 [1];  [2];  [3]; ORCiD logo [2]
  1. Virginia Polytechnic Inst. and State Univ. (Virginia Tech), Blacksburg, VA (United States). Dept. of Materials Science and Engineering
  2. Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States). Chemical Sciences Division
  3. Virginia Polytechnic Inst. and State Univ. (Virginia Tech), Blacksburg, VA (United States). Dept. of Materials Science and Engineering; Virginia Polytechnic Inst. and State Univ. (Virginia Tech), Blacksburg, VA (United States). Dept. of Mechanical Engineering
Publication Date:
Research Org.:
Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
Sponsoring Org.:
USDOE Laboratory Directed Research and Development (LDRD) Program
OSTI Identifier:
1376303
Grant/Contract Number:
AC05-00OR22725
Resource Type:
Journal Article: Accepted Manuscript
Journal Name:
Computer Physics Communications
Additional Journal Information:
Journal Volume: 219; Journal Issue: C; Journal ID: ISSN 0010-4655
Publisher:
Elsevier
Country of Publication:
United States
Language:
English
Subject:
71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; 97 MATHEMATICS AND COMPUTING; Kinetic Monte Carlo; Stiffness; Steady state; Non-equilibrium

Citation Formats

Danielson, Thomas, Sutton, Jonathan E., Hin, Céline, and Savara, Aditya. SQERTSS: Dynamic rank based throttling of transition probabilities in kinetic Monte Carlo simulations. United States: N. p., 2017. Web. doi:10.1016/j.cpc.2017.05.016.
Danielson, Thomas, Sutton, Jonathan E., Hin, Céline, & Savara, Aditya. SQERTSS: Dynamic rank based throttling of transition probabilities in kinetic Monte Carlo simulations. United States. doi:10.1016/j.cpc.2017.05.016.
Danielson, Thomas, Sutton, Jonathan E., Hin, Céline, and Savara, Aditya. Fri . "SQERTSS: Dynamic rank based throttling of transition probabilities in kinetic Monte Carlo simulations". United States. doi:10.1016/j.cpc.2017.05.016.
@article{osti_1376303,
title = {SQERTSS: Dynamic rank based throttling of transition probabilities in kinetic Monte Carlo simulations},
author = {Danielson, Thomas and Sutton, Jonathan E. and Hin, Céline and Savara, Aditya},
abstractNote = {Lattice based Kinetic Monte Carlo (KMC) simulations offer a powerful simulation technique for investigating large reaction networks while retaining spatial configuration information, unlike ordinary differential equations. However, large chemical reaction networks can contain reaction processes with rates spanning multiple orders of magnitude. This can lead to the problem of “KMC stiffness” (similar to stiffness in differential equations), where the computational expense has the potential to be overwhelmed by very short time-steps during KMC simulations, with the simulation spending an inordinate amount of KMC steps / cpu-time simulating fast frivolous processes (FFPs) without progressing the system (reaction network). In order to achieve simulation times that are experimentally relevant or desired for predictions, a dynamic throttling algorithm involving separation of the processes into speed-ranks based on event frequencies has been designed and implemented with the intent of decreasing the probability of FFP events, and increasing the probability of slow process events -- allowing rate limiting events to become more likely to be observed in KMC simulations. This Staggered Quasi-Equilibrium Rank-based Throttling for Steady-state (SQERTSS) algorithm designed for use in achieving and simulating steady-state conditions in KMC simulations. Lastly, as shown in this work, the SQERTSS algorithm also works for transient conditions: the correct configuration space and final state will still be achieved if the required assumptions are not violated, with the caveat that the sizes of the time-steps may be distorted during the transient period.},
doi = {10.1016/j.cpc.2017.05.016},
journal = {Computer Physics Communications},
number = C,
volume = 219,
place = {United States},
year = {Fri Jun 09 00:00:00 EDT 2017},
month = {Fri Jun 09 00:00:00 EDT 2017}
}

Journal Article:
Free Publicly Available Full Text
This content will become publicly available on June 9, 2018
Publisher's Version of Record

Citation Metrics:
Cited by: 2works
Citation information provided by
Web of Science

Save / Share:
  • Progress has been rapid in increasing the efficiency of energy conversion in nanoparticles. However, extraction of the photo-generated charge carriers remains challenging. Encouragingly, the charge mobility has been improved recently by driving nanoparticle (NP) films across the metal-insulator transition (MIT). To simulate MIT in NP films, we developed a hierarchical Kinetic Monte Carlo transport model. Electrons transfer between neighboring NPs via activated hopping when the NP energies differ by more than an overlap energy, but transfer by a non-activated quantum delocalization, if the NP energies are closer than the overlap energy. As the overlap energy increases, emerging percolating clusters supportmore » a metallic transport across the entire film. We simulated the evolution of the temperature-dependent electron mobility. We analyzed our data in terms of two candidate models of the MIT: (a) as a Quantum Critical Transition, signaled by an effective gap going to zero; and (b) as a Quantum Percolation Transition, where a sample-spanning metallic percolation path is formed as the fraction of the hopping bonds in the transport paths is going to zero. We found that the Quantum Percolation Transition theory provides a better description of the MIT. We also observed an anomalously low gap region next to the MIT. We discuss the relevance of our results in the light of recent experimental measurements.« less
  • The kinetics for the selective hydrogenation of acetylene-ethylene mixtures over model Pd(111) and bimetallic Pd-Ag alloy surfaces were examined using first principles based kinetic Monte Carlo (KMC) simulations to elucidate the effects of alloying as well as process conditions (temperature and hydrogen partial pressure). The mechanisms that control the selective and unselective routes which included hydrogenation, dehydrogenation and C-C bond breaking pathways were analyzed using first-principle density functional theory (DFT) calculations. The results were used to construct an intrinsic kinetic database that was used in a variable time step kinetic Monte Carlo simulation to follow the kinetics and the molecularmore » transformations in the selective hydrogenation of acetylene-ethylene feeds over Pd and Pd-Ag surfaces. The lateral interactions between coadsorbates that occur through-surface and through-space were estimated using DFT-parameterized bond order conservation and van der Waal interaction models respectively. The simulation results show that the rate of acetylene hydrogenation as well as the ethylene selectivity increase with temperature over both the Pd(111) and the Pd-Ag/Pd(111) alloy surfaces. The selective hydrogenation of acetylene to ethylene proceeds via the formation of a vinyl intermediate. The unselective formation of ethane is the result of the over-hydrogenation of ethylene as well as over-hydrogenation of vinyl to form ethylidene. Ethylidene further hydrogenates to form ethane and dehydrogenates to form ethylidyne. While ethylidyne is not reactive, it can block adsorption sites which limit the availability of hydrogen on the surface and thus act to enhance the selectivity. Alloying Ag into the Pd surface decreases the overall rated but increases the ethylene selectivity significantly by promoting the selective hydrogenation of vinyl to ethylene and concomitantly suppressing the unselective path involving the hydrogenation of vinyl to ethylidene and the dehydrogenation ethylidene to ethylidyne. This is consistent with experimental results which suggest only the predominant hydrogenation path involving the sequential addition of hydrogen to form vinyl and ethylene exists over the Pd-Ag alloys. Ag enhances the desorption of ethylene and hydrogen from the surface thus limiting their ability to undergo subsequent reactions. The simulated apparent activation barriers were calculated to be 32-44 kJ/mol on Pd(111) and 26-31 kJ/mol on Pd-Ag/Pd(111) respectively. The reaction was found to be essentially first order in hydrogen over Pd(111) and Pd-Ag/Pd(111) surfaces. The results reveal that increases in the hydrogen partial pressure increase the activity but decrease ethylene selectivity over both Pd and Pd-Ag/Pd(111) surfaces. Pacific Northwest National Laboratory is operated by Battelle for the US Department of Energy.« less
  • The present study aims at predicting the level of formation of the ultimate carcinogenic metabolite of methyleugenol, 1′-sulfooxymethyleugenol, in the human population by taking variability in key bioactivation and detoxification reactions into account using Monte Carlo simulations. Depending on the metabolic route, variation was simulated based on kinetic constants obtained from incubations with a range of individual human liver fractions or by combining kinetic constants obtained for specific isoenzymes with literature reported human variation in the activity of these enzymes. The results of the study indicate that formation of 1′-sulfooxymethyleugenol is predominantly affected by variation in i) P450 1A2-catalyzed bioactivationmore » of methyleugenol to 1′-hydroxymethyleugenol, ii) P450 2B6-catalyzed epoxidation of methyleugenol, iii) the apparent kinetic constants for oxidation of 1′-hydroxymethyleugenol, and iv) the apparent kinetic constants for sulfation of 1′-hydroxymethyleugenol. Based on the Monte Carlo simulations a so-called chemical-specific adjustment factor (CSAF) for intraspecies variation could be derived by dividing different percentiles by the 50th percentile of the predicted population distribution for 1′-sulfooxymethyleugenol formation. The obtained CSAF value at the 90th percentile was 3.2, indicating that the default uncertainty factor of 3.16 for human variability in kinetics may adequately cover the variation within 90% of the population. Covering 99% of the population requires a larger uncertainty factor of 6.4. In conclusion, the results showed that adequate predictions on interindividual human variation can be made with Monte Carlo-based PBK modeling. For methyleugenol this variation was observed to be in line with the default variation generally assumed in risk assessment. - Highlights: • Interindividual human differences in methyleugenol bioactivation were simulated. • This was done using in vitro incubations, PBK modeling and Monte Carlo simulations. • Major enzymatic reactions causing interindividual differences were identified. • A chemical-specific adjustment factor (CASF) for intraspecies variation was derived. • The default uncertainty factor for kinetic human variability covers 90% of population.« less
  • We present here results of continued efforts to understand the performance of microchannel plate (MCP)–based, high-speed, gated, x-ray detectors. This work involves the continued improvement of a Monte Carlo simulation code to describe MCP performance coupled with experimental efforts to better characterize such detectors. Our goal is a quantitative description of MCP saturation behavior in both static and pulsed modes. A new model of charge buildup on the walls of the MCP channels is briefly described. The simulation results agree favorably with experimental data obtained with a short-pulse, high-intensity ultraviolet (UV) laser. These results indicate that a weak saturation canmore » change the exponent of gain with voltage and that a strong saturation lead to a gain plateau. These results also demonstrate that the dynamic range of an MCP in pulsed mode has a value of between 10^2 and 10^3.« less
  • We present here results of continued efforts to understand the performance of microchannel plate (MCP)-based, high-speed, gated, x-ray detectors. This work involves the continued improvement of a Monte Carlo simulation code to describe MCP performance coupled with experimental efforts to better characterize such detectors. Our goal is a quantitative description of MCP saturation behavior in both static and pulsed modes. A new model of charge buildup on the walls of the MCP channels is briefly described. The simulation results are compared to experimental data obtained with a short-pulse, high-intensity ultraviolet laser, and good agreement is found. These results indicate thatmore » a weak saturation can change the exponent of gain with voltage and that a strong saturation leads to a gain plateau. These results also demonstrate that the dynamic range of a MCP in pulsed mode has a value of between 10{sup 2} and 10{sup 3}.« less