Message by **Jérémie Hacquart**:

Hi,

I’m new to SPIS EP and I try to model a simple case in order to grasp the subtleties of SPIS-EP. I have several errors occurring during the run of my simulation.

Here is a summary of the simulation:

**Geometry GMSH**

-One solar panel: 2m*4m*0.02m - msh_coef (mesh element size) = 0.3

-One box: 2m*2m*2m - msh_coef = 0.3

-One thruster: a ring with ri=0.028m and ro = 0.05m - msh_coef = 0.08

-Plume : half ellipsoid with radius = 0.6m and height = 3m - msh_coef = 0.08

-Boundary: ellipsoid with msh_coef = 0.5

I create two volumes, one for the plume and one for the whole boundary without the plume, box and solar array. Then I create physical surfaces for the box, SA, etc. and only one physical volume with both volumes.

This gives 378246 tetrahedron with a quality of 0.7577 (0.08411->0.9878), after optimizing with netgen.

**SPIS**

I use the SPT100 model for the thruster in the group definition and a different node for every groups.

The electrical is circuit is the following:

R 0 1 100000 ~~> 0 = box and 1 = SA rear side~~> 2 = SA front size

V 1 2 0

For the global parameters, I remove photoemission, I model one electron population with PICVolDistrib. The duration of the simulation is 1E-4s with a time step of 1E-6s and a plasmaDuration of 1E-6s and PlasmaDt of 1E-6s. The others parameters are kept at their default value.

**SPIS NUM**

Here is my problem, the numerical kernel is initialized and I am able to launch the simulation. It runs but I have several error occurring like:

ReciprocalFunction warning: y out of range

ERROR in LDLt decomposition in assembPrecontNonLinear = 4.

If I understand correctly, the Poisson equation is responsible for the LDLt error. Is the quality of the mesh responsible for this? If I try to refine the mesh, the simulation seems to take a significant amount of time (with the same errors occurring) and this is without a realistic and more complex geometry.

Many thanks for the help.