An adaptive domain-hybridized plasma fluid model
POSTER
Abstract
Plasma models have regimes of validity that depend on the local parameters.
In some problems a computationally expensive model is required in a small subset of the domain while faster reduced models can adequately describe the plasma behavior everywhere else.
A model which adaptively switches between a two-fluid and MHD model with per-element resolution is presented. This model reduces the overall degrees of freedom and associated computational costs required while maintaining two-fluid physical fidelity in the entire domain.
To handle the disparate timescales of the constituent models, a fully implicit high order accurate hybridizable discontinuous Galerkin (HDG) finite element method is utilized.
HDG methods provide an effective method for reducing the global system solve size and improves the system condition number by designing a system where a Schur complement is relatively cheap to construct in a highly parallelizable fashion.
This allows timesteps on the order of 60x-500x the CFL timestep stability limits, with limits dictated by truncation error on the ion dynamics timescale.
A tool capable of generating the required HDG code and analytical Schur complement is also presented.
In some problems a computationally expensive model is required in a small subset of the domain while faster reduced models can adequately describe the plasma behavior everywhere else.
A model which adaptively switches between a two-fluid and MHD model with per-element resolution is presented. This model reduces the overall degrees of freedom and associated computational costs required while maintaining two-fluid physical fidelity in the entire domain.
To handle the disparate timescales of the constituent models, a fully implicit high order accurate hybridizable discontinuous Galerkin (HDG) finite element method is utilized.
HDG methods provide an effective method for reducing the global system solve size and improves the system condition number by designing a system where a Schur complement is relatively cheap to construct in a highly parallelizable fashion.
This allows timesteps on the order of 60x-500x the CFL timestep stability limits, with limits dictated by truncation error on the ion dynamics timescale.
A tool capable of generating the required HDG code and analytical Schur complement is also presented.
Presenters
-
Andrew Ho
University of Washington
Authors
-
Andrew Ho
University of Washington
-
Uri Shumlak
University of Washington