Neural network assisted electrostatic global gyrokinetic toroidal code using cylindrical coordinates
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866929471452348416 |
|---|---|
| author | Alageshan, Jaya Kumar Das, Joydeep Singh, Tajinder Sharma, Sarveshwar Kuley, Animesh |
| author_facet | Alageshan, Jaya Kumar Das, Joydeep Singh, Tajinder Sharma, Sarveshwar Kuley, Animesh |
| contents | Gyrokinetic simulation codes are used to understand the microturbulence in the linear and nonlinear regimes of the tokamak and stellarator core. The codes that use flux coordinates to reduce computational complexities introduced by the anisotropy due to the presence of confinement magnetic fields encounter a mathematical singularity of the metric on the magnetic separatrix surface. To overcome this constraint, we develop a neural network-assisted Global Gyrokinetic Code using Cylindrical Coordinates (G2C3) to study the electrostatic microturbulence in realistic tokamak geometries. In particular, G2C3 uses a cylindrical coordinate system for particle dynamics, which allows particle motion in arbitrarily shaped flux surfaces, including the magnetic separatrix of the tokamak. We use an efficient particle locating hybrid scheme, which uses a neural network and iterative local search algorithm, for the charge deposition and field interpolation. G2C3 uses the field lines estimated by numerical integration to train the neural network in universal function approximator mode to speed up the subroutines related to gathering and scattering operations of gyrokinetic simulation. Finally, as verification of the capability of the new code, we present results from self-consistent simulations of linear ion temperature gradient modes in the core region of the DIII-D tokamak. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_12851 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Neural network assisted electrostatic global gyrokinetic toroidal code using cylindrical coordinates Alageshan, Jaya Kumar Das, Joydeep Singh, Tajinder Sharma, Sarveshwar Kuley, Animesh Plasma Physics Computational Physics Gyrokinetic simulation codes are used to understand the microturbulence in the linear and nonlinear regimes of the tokamak and stellarator core. The codes that use flux coordinates to reduce computational complexities introduced by the anisotropy due to the presence of confinement magnetic fields encounter a mathematical singularity of the metric on the magnetic separatrix surface. To overcome this constraint, we develop a neural network-assisted Global Gyrokinetic Code using Cylindrical Coordinates (G2C3) to study the electrostatic microturbulence in realistic tokamak geometries. In particular, G2C3 uses a cylindrical coordinate system for particle dynamics, which allows particle motion in arbitrarily shaped flux surfaces, including the magnetic separatrix of the tokamak. We use an efficient particle locating hybrid scheme, which uses a neural network and iterative local search algorithm, for the charge deposition and field interpolation. G2C3 uses the field lines estimated by numerical integration to train the neural network in universal function approximator mode to speed up the subroutines related to gathering and scattering operations of gyrokinetic simulation. Finally, as verification of the capability of the new code, we present results from self-consistent simulations of linear ion temperature gradient modes in the core region of the DIII-D tokamak. |
| title | Neural network assisted electrostatic global gyrokinetic toroidal code using cylindrical coordinates |
| topic | Plasma Physics Computational Physics |
| url | https://arxiv.org/abs/2408.12851 |