[Paper Review] CRANE: A MOOSE-based Open Source Tool for Plasma Chemistry Applications

Shane Keniley, Davide Curreli|arXiv (Cornell University)|May 24, 2019

Plasma Diagnostics and Applications14 references4 citations

TL;DR

CRANE is an open-source, MOOSE-based finite element tool for simulating plasma chemistry via coupled ODEs for species densities and reaction networks. It enables standalone 0D modeling and seamless integration into multidimensional plasma simulations (e.g., with Zapdos), demonstrating accurate coupling and scalability for complex, multiscale plasma-chemistry problems with full support for external solvers like Bolsig+.

ABSTRACT

Numerical simulation of plasma discharges is often performed by models developed in-house and coupling externally and separately written codes. The MOOSE (Multiphysics Object Oriented Simulation Environment) framework provides tools for quickly developing and coupling together software in a scalable framework, making it well-suited for plasma simulations. A new MOOSE application, CRANE (Chemical ReAction NEtwork), was developed in order to add an independent chemical kinetics application to the framework. In this work the capabilities of CRANE are detailed, including its use as a standalone solver for global plasma chemistry models. The capability of fully coupling applications in the MOOSE framework is also shown by compiling CRANE directly into the MOOSE-based plasma transport software Zapdos, showing the possibility to solve fully coupled drift-diffusion-reaction plasma chemistry problems. The code is open-source and available at the following url, https://github.com/lcpp-org/crane.

Motivation & Objective

• To develop a flexible, open-source plasma chemistry solver that integrates natively into the MOOSE finite element framework for scalable, multiphysics plasma simulations.
• To enable accurate and extensible modeling of large, nonlinear, nonequilibrium reaction networks in low-temperature plasmas without requiring C++ coding.
• To demonstrate seamless coupling between plasma chemistry (CRANE) and transport (Zapdos) via direct compilation into a single MOOSE application.
• To support dynamic integration of external physics solvers such as Bolsig+ for electron-impact cross-section calculations.
• To provide a modular, extensible platform for future development of coupled plasma-chemistry-transport models in diverse applications.

Proposed method

• CRANE uses the MOOSE Actions system to parse user-defined reaction networks from human-readable input files, automatically constructing the ODE system for species concentrations.
• It solves a system of coupled ODEs for species densities using implicit time integration (implicit Euler) and a Newton-Krylov nonlinear solver with relative tolerance of 1e-4.
• The code models reaction rates via elementary bimolecular and termolecular reactions with rate coefficients derived from Arrhenius or other kinetic models.
• Gas temperature evolution is modeled via energy balance, accounting for enthalpy changes in each reaction using the adiabatic isochoric approximation.
• CRANE supports external coupling via the MOOSE MultiApp system, allowing it to be compiled into other MOOSE applications like Zapdos for fully coupled drift-diffusion-reaction simulations.
• It leverages the MOOSE mesh and linear algebra infrastructure, enabling native 1D–3D, parallel, and scalable simulations from input files alone.

Experimental results

Research questions

• RQ1 Can CRANE accurately simulate complex, large-scale plasma chemistry networks in a fully extensible, open-source framework?
• RQ2 How efficiently can CRANE couple with transport solvers like Zapdos to enable fully coupled plasma-chemistry simulations?
• RQ3 What is the computational overhead of runtime parsing of reaction networks compared to compiled Fortran solvers like ZDPlasKin?
• RQ4 How does CRANE’s performance scale with increasing numbers of species and reactions in realistic plasma systems?
• RQ5 To what extent can CRANE support dynamic integration of external physics codes such as Bolsig+ for electron-impact processes?

Key findings

• CRANE successfully simulated a 13-reaction global argon plasma model in 1D, achieving steady state in 53 μs with a single-core runtime of 12 seconds on a 2015 MacBook Pro.
• The code demonstrated good agreement with reference solvers for species density evolution, including strong cathode-peak behavior in excited argon states and electron depletion near electrodes.
• Benchmarking showed CRANE’s runtime was slower than ZDPlasKin (e.g., 66.27s vs. 32.27s for 8 species, 13 reactions) due to runtime parsing and lack of compiled code and caching.
• The performance gap was significantly reduced when accounting for external solver overhead: CRANE’s dynamic Bolsig+ calls added substantial latency compared to ZDPlasKin’s precompiled library.
• CRANE achieved full integration with Zapdos by compiling it as a submodule, enabling a single executable to solve coupled drift-diffusion-reaction equations in 1D with 13 reactions.
• The framework’s ability to scale to multiple dimensions and parallel execution was confirmed via MOOSE’s internal mesh and linear algebra management, requiring no code changes beyond Makefile and two C++ lines.