(443a) An Automated Job Management Framework for High Fidelity Quantum Chemistry Calculations | AIChE

(443a) An Automated Job Management Framework for High Fidelity Quantum Chemistry Calculations


Sirumalla, S. K. - Presenter, Northeastern University
Farina, D. Jr., Northeastern University
Harms, N., Northeastern University
West, R. H., Northeastern University
Recent advancements in quantum chemistry methods allow chemical properties to be calculated ab-initio within chemical accuracy (1 kcal/mol). With the advance of High Performance Computing the bottleneck is no longer the computation, but the human interaction required to set up calculations, curate, and interpret the results. Calculating thermochemical kinetics accurately requires multiple interdependent calculations, for example to search for conformers, optimize geometry, calculate electronic correlation, calculate valence and core electrons separately, perform hindered rotor scans, etc. Automation of these calculations with minimal human interaction is needed to realize the full potential of ever-increasing computational power in many research fields. Several workflows and protocols have been proposed to automate these calculations, but not for actual job spawning, managing dependencies between these jobs, and post processing results automatically.

In this talk, we present an automated framework that automatically spawns jobs, and handles dependencies between jobs, on multiple supercomputing clusters, with minimal user interaction. Specifically, we use FireWorks [1], an open source automated framework written in Python for defining, managing, and executing complex workflows. Jobs along with their dependencies are stored in MongoDB. FireWorks consists of two components, a FireServer that manages workflows and a FireWorker that executes these workflows. We demonstrate the efficacy of FireWorks through automated scheduling of high-fidelity quantum chemistry calculations for the construction of detailed kinetic models. This will enable accurate thermodynamic and kinetic parameters to be calculated in a high-throughput manner that makes the most of computational resources available to a user. In addition, this framework allows for the creation of a standardized dataset that can be used as training data for machine-learning based estimation methods.


1. Jain, A., Ong, S. P., Chen, W., Medasani, B., Qu, X., Kocher, M., Brafman, M., Petretto, G., Rignanese, G.-M., Hautier, G., Gunter, D., and Persson, K. A. (2015) FireWorks: a dynamic workflow system designed for high-throughput applications. Concurrency Computat.: Pract. Exper., 27: 5037–5059. doi: 10.1002/cpe.3505.