multi-species architecture

[stonecoldhughes]

The way GITR models surface interactions needs to be re-designed to enable multi-species functionality. The foundations of the new design include:

• separation of particle transport modeling and surface interaction modeling into coupled modules
• performant collaboration between those modules

Community input is welcome for planning this re-design.

[stonecoldhughes]

example idea for the new time loop:

initialize all particles (species, position, velocity)

for each particle:

transport (until surface impact occurs)

~ surface impact occurs ~

Freeze particle (gather data): {  impact location, impact energy, impact angle, species involved }

Reinitialize (new species, previous impact location, new velocity)

Terminology:

• particle_state: energy, angle, species, position

• impact_event: a state transition from transporting through the plasma to interacting with a surface mesh triangle. Described by location, species involved, energy, angle.

• sample_outcome: a new particle_state. The sample outcome represents a possible result of impact_event randomly sampled from a probability distribution. Examples: the sample outcome of a reflection type impact_event would include a new energy, a new angle, the same location as the impact event. The sample outcome of a sputtering type impact_event would include new energy, new angle, new species, new location.

Data structures for this functionality:

• impact_queue: a bucket of particles that have impacted surface triangles and are awaiting further instructions
• transport_queue: a bucket of particles that are moving around in the plasma until they hit a surface

Functional modules that will carry out this functionality:

GITR_transport:

This module will transport the particles from their initial positions to a surface:

• transport_queue ---> impact_queue
• when particles are in the bulk plasma and far from any surface, there is no need to check mesh triangles for surface impact events
• when particles are close to a surface, a small subset of mesh triangles near the particle must be examined to detect an impact_event
• when an impact_event is detected, the transport module stops updating the particles position and puts them in a queue to be processed by GITR_surface later

GITR_surface:

This module determines the result of impact events.

• impact_queue ---> transport_queue
• The outcome of an impact event is probabilistic and follows probability distributions.
• By sampling the distributions of outcomes for many events, realistic behavior is modeled
• characteristics about the impact event inform which probability distribution must be sampled
• select a sample_outcome for each impact_event. Re-initialize a new particle according to the sample_outcome and put it in the transport_queue
• periodically update the probability distributions to account for the effects of surface evolution over time

GITR_data:

• given the location of a particle, provide the relevant small subset of mesh triangles that need to be checked for surface impact.
• provide access to probability distributions
• manage the impact_queue and the transport_queue