This repository contains the models and tests for various scintillator crystal geometries.
The models were created in FreeCad and then exported to STL for use in the pvtrace Python package.
Using pvtrace we ray trace photons in the scintillator to determine how many reach our detector.
There are STL files for various crystal geometries and accompanying ESR shells.
The ESR shells are constructed such that there is a 1 um gap between the crystal and the ESR, and the ESR is 0.1 mm thick.
See the "notebooks" folder for examples and the methodology used thus far.
The major downside to this is that pvtrace is inherently slow due to the ray tracing engine it uses.
It takes several minutes to run a simulation with a few thousand photons.
The world building functions that lay out the models assume that the SiPM readout is at the top of the geometries, where the +z axis is "up."
The STL files are in units of mm while the world building assumes units of centimeters (thus, the STL models are scaled when they're read in).
This inconsistency, while minor, should probably be corrected.
I tried making this change, but I was getting a lot of GeometryError throws... I am unsure why this is. It seems like scaling the STL files is the best approach for now.
There appears to be a rounding issue with importing meshes from the STL files.
The pvtrace functions use a threshold value hardcoded into the module, so I had to go in and multiply this threshold by 100 in order to prevent GeometryErrors (it's the EPS_ZERO variable in the pvtrace/geometry/utils.py file).
I slightly hesitate to do this since it's a bit hacky, but the threshold is on the order of 10^-10, and I have confirmed in FreeCad that the models themselves do not have any inherent geometry issues (e.g. overlaps).
It appears it may be a limitation of the trimesh package used by pvtrace, as there's even a comment in the pvtrace code that mentions this.
You can install the module using the following command while inside the "crystal-geometry" repository:
pip install .
for a static install or
pip install -e .
for an editable install.
BEWARE if you want to make custom models/worlds using the built-in functions and classes!
- The y-z plane is taken to be the plane in which the light grid points are made, so keep this in mind if you want to make a new crystal model.
- As mentioned above, there's an assumption in where the detector readout is in the geometry.
- The default absorption coefficient for LYSO is 1/16 cm-1. This was informed by research papers; however, this value has large error bars.
- The default optical pad material is assumed to be a perfect transmitter (absorption coefficient = 0), meaning it will not absorb any photons.
- The default SIPM material is assumed to be a perfect absorber. The default absorption coefficient is finite to allow the photons to propagate a small distance into the SiPM volume before being absorbed. This makes it easier to compute which photons were absorbed by the SiPM since we're a convex hull to determine which volume absorbed each photon.
- The ESR is assumed to be a perfect reflector.
- The world volume has a refractive index of 1 by default.
- Rod
- Rod with cut face ("wedge")
- Plate
- Plate with cut face ("chamfered")
Chamfered triangular prisms- Actually, there are some geometric issues with these. Removed until they can be fixed.
- Isosceles triangular prisms (the best!)
- 6 and 10 SiPM configurations
- Allow construction of a detector configuration using various crystal shapes.
- Allow testing light scintillation at various points within the crystal.
- Provide a nice baseline for additional or more complicated crystal shapes.
Lambertian scattering off the ESR surface.This is currently commented out since it greatly increases the run time and sometimes results in crashes due to geometry errors.
- Allow specification of depth where the scintillated photons are released. I think this is a relatively minor thing and would be easy to implement, but it's not currently configured to allow this.