In each iteration, number of collisions are calculate as a REAL variable
(collFrac) and stored in each cell. The number of collisions is
calculated as FLOOR(collFrac) and, if it is >1 collisions are computed
as usual. Per each collision calculated, 1.0 is removed from collFrac
Possibility to input initial species distributions (density, velocity
and temperature) via an input file for each species.
New moduleRandom includes function to generate random numbers in
different ways (still uses) the implicit RANDOM_NUMBER().
Now collisions can have a different time step.
Added species name to output names as it was starting to get confusing
in Gmsh for multiple species.
Output filenames adapted to match any number of iterations.
A boundary condition for each species must be indicated in the case
file.
This opens the door to use boundary conditions with different parameters
(for example, a wall temperature, coefficients for reflection or
absorption...)
The examples included with the code have been updated accordently.
into account (as in 1D Cartesian case).
The 1D Cathode example case has been modified, having now 2 input files:
- inputCart.json: Used for Cartesian coordinates
- inputRad.json: Used for Radial coordinates
Pusher is a Boris pusher but without z direction.
Maxwellian and Diract Delta distributions have been implemented.
The input for injection of particles should be rewritten to allow more
clear input file.
directional derivative depending on the definition of first node. Trying
to solve it with searching for the right first node but it is very
difficult. A solution is required to allow triangular meshes in charged
simulations.
Reading of this type of element needs to be implemented.
Fixed a bug in which the L_ref (reference length) was not correctly
being calculated for neutral solver.
Now, the solver needs to be an input parameter of the case, to select if
it is for charged or neutral particles.
Resolution of Poisson equation with Dirichlet boundary conditions is
possible. The source vector is the charge density. This resolution is
done in two steps to save computational time:
1. When reading the mesh, the PLU factorization of the K matrix is
computed.
2. In each iteration, the system K*u = f is solved, in which f is the
source vector (charge density) and u is the solution (potential) in
each node.
No case has been added to the repository. This will be done in next
commit.
The 'non-analog' scheme has been commented. It still needs to split
the particle to avoid 'overweight' particles.
- Injection is now performed in parallalel (an IF statement could be
required to avoid overhead when number of injected particles is
below a margin).
- Added the possibility for multiple injections.
assumed to be inside the domain but without cell assigned.
Now, particles are assumed to be outside the domain (n_in = .FALSE.)
until findCell assign them a cell.
Bugs fixed:
- Solved an issue with particles being injected with infinite velocity
resulting in Inf velocity in some cells of the output files.
- Particles are now equally distributed in cylindrical geometry along
the radial direction.
New features:
- Particles now have their own weight that is recalculated when the
particle moves to a new cell. This avoid the reduction of density at
r = 0.
Cases:
- Added a case of Argon flow around a cylinder to measure performance
and future improvements.
