Workflow:

• Read equilibrium data
• Normalize data
• Build mesh
• Build the gather-scatter operator
• Build the (gyro-averaged) Poisson solver
• Initialize particles
• Execute the PIC cycle

The input files are stored in the “dat/” folder.

• g2c3.in (Input parameters)
• g_file.in (EFIT equilibrium data)
• profile.in (Temperature and density profile data)

In G2C3, all the read operations are performed in read_data.F90 by the Master processing element (mpe==0). Then the MPI_BCAST subroutine in mpi.F90 broadcasts the values to all the other processing elements. In case the read elements are an array, first the array dimension is broadcast and then the arrays are allocated in allocate_arrays.F90

The basic units and normalization used in G2C3 are summarized as follows:

• Mass: proton mass, $m_p$
• Charge: proton charge, e
• Magnetic field: on axis, $B_a$
• Length: Tokamak major radius, $R_0$
• Density: on axis electron density, $n_{e0}$
• Temperature: on axis electron temperature, $T_{e0}$
• Time: Inverse on axis cyclotron frequency of proton, $\omega_p^{-1}=m_pc/eB_a$

Mesh

We use a neural network in a Universal function approximation mode to map $\mathcal{N}_\Delta:(R,Z) \rightarrow \Delta$, where $\Delta$ is the triangle that encloses the point. (Learn more)

(Learn more)

We use the PETSc library in the field solver to solve MPI and OpenMP-based sparse matrix operations, such as inverse and multiplication.