-
Notifications
You must be signed in to change notification settings - Fork 2
TMS Simulation
This TMS sim page was last revised 2024-12-02.
The TMS simulation uses as input the edep sim files used to simulate ND-LAr (which is also used in the 2x2 simulation). A similar simulation is also used in the FD simulation. More technical details about those steps are provided in the Genie and Edep Sim (Geant4) Stages section below. The first section is specifically about the dune-tms det sim
The TMS uses scintillator bars. Each bar contains a wavelength shifting fiber (WSF) which connects to the SiPM readout on one end. The other side is assumed to be mirrored with 95% reflection efficiency. Edep sim calculates how much energy was deposited in each scintillator bar. The scintillator converts that energy into photons which are then absorbed by the WSF, and then reemitted at a longer wavelength. They are then read out by the SiPMs and converted into a digital signal.
The purpose of the det sim is to add in the effects from the detector readout process. Broadly this has two main effects that need to be simulated: changes in hit energy and changes in hit time. On top of that, there are effects like noise, deadtime, and crosstalk that can be simulated.
To convert the amount of energy deposited in a scintillator bar to a hit, it undergoes a number of steps:
- Light yield
- Birk's suppression
- Poisson throws
- Fiber path adjustments
- Hit merging, pedestal subtraction, and readout simulation
The energy deposited in edep sim is converted to raw number of true Photo-Electrons (PE). The light yield is expected to be around 50 PE / MeV for this scintillator. So PE_{True} = E_{Dep} * 50 PE / MeV.
In the future, we want to add the Cherenkov component of the light as well. This is light that can also be absorbed by the WSF and contribute the total signal. It's expected to contribute a significant proportion (between 20-50%) of the energy observed.
Birk's suppression is an effect in scintillators where light yield is suppressed due to more energy being deposited in a small area. This has the largest effect on the relatively-high energy deposits from protons. The adjusted PE_{Suppressed} = PE_{True} / (1 + B*dedx) where the birk's parameter of 0.0905 mm / MeV is taken from Minerva, which has similar scintillator to what will be used.
The creation of photons from scintillation is a random process that follows poisson statistics. So this step takes the true PE after birk's suppression and does a poisson throw with that mean.
Next, the PE are split into two possible paths (long and short) using a binomial distribution with a 50% probability of each. The long path assumes the photon traveled to the side opposite the readout, got mirrored with 95% efficiency, and then got read out. This greatly increases total path length of the photon.
Using the respective path lengths of the photon, the energy deposited at the readout is calculated. This takes into account the exponential loss of energy given the attentuation length of 4.160m. An additional geometric correction of 1.8 is added to account for the additional path length the photon experiences as it bounces in the fiber. So A = exp(-1.8 d / 4.160m), where d is the distance traveled. d is calculated separately for the long and short paths.
We currently do not simulate (but plan to) the effect of short-range attentuation. This is an effect where additional light can be seen from energy deposits near the readout. This would be simulated with an additional attenuation term that has a smaller attentuation length.
Finally, all hits on a channel within a single readout window are merged to simulate the readout. The start of the readout window is defined as when the first photon hits the readout. The photon timing simulation takes into account the 2 scintillating processes (3ns for scintillator bars and 20ns for WSF) and the additional path lengths from the fibers. The fibers have both the geometrical factor but also an index of refraction of 1.5 which reduces light speed. The readout window is 120ns with no dead time but this is adjustable in the readout config file.
A gaussian random with mean 0 and signa 0.4 PE is added to simulate the readout noise. If the hit energy is not higher than the pedestal threshold (currently 3 PE), then the hit is removed to simulate pedestal subtraction. This is then converted into hit energy based on a rough calibration so that reco hit E ~ true hit E on average.
Slides about optical model. Slides about timing simulation.
The simulation contains code to simulate deadtime, but it was only minimally validated. Currently, the readout is expected to have little or no deadtime.
There currently is no noise or crosstalk simulation. The noise rate is expected to be low enough that on average we expect six noise hits per time slice on average. That is based on the mu2e readout which has 600 noise hits per spill, assuming around 100 time slices per spill.
There currently are two ways in which we can produce samples. The official "microprod" samples run on nersc, and the "old" technique that runs on the fermigrid.
The microprod sample runs simulates events on the ND-LAr bath using genie. Then it uses the rockbox technique to generate rock muons. This technique removes rock muon interactions whose muons do not have enough energy to hit the near detector. This saves a lot of time on the edep sim step. The two streams are then sent through edep sim and then combined in a spill building step to simulate pileup
The current up-to-date geometry is geometry v1.0.3 nd_hall_with_lar_tms_sand_TDR_Production_geometry_v_1.0.3.gdml. See the ND geometry here. We are working on a finalized geometry with the most up-to-date engineering specifications
The "old" production script is similar except we're not using it to make rock muons. Plus it simulates pileup using edep sim directly, rather than using the overlay script. This version of pileup does not match the spill structure perfectly, but is pretty close. Using this old technique has a quicker turnaround since it can be done by anyone, while the microprods currently require ND production.
Genie 3.04 AR23_20i_00_000 tune with edep sim
export ARCUBE_CONTAINER='mjkramer/sim2x2:genie_edep.3_04_00.20230620'
tune=AR23_20i_00_000
Genie v3.02 G1810a0211a tune with edep sim v3.2. By default, it uses the Flux/g4lbne/v3r5p4/QGSP_BERT/OptimizedEngineeredNov2017 flux