Minutes of the February 20 Liquid Argon Simulations meeting
Present:
Tom Junk, Eric Church, Brian Rebel, Herb Greenlee, Mike Kirby, Qizhong Li,
Kate Scholberg, Zepeng Li, Matt Szydagis, Jonathan Insler, Matt Toups,
Denver Whittington, Thomas Kutter, Dave McKee, Stan Seibert, Mark Convery
Apologies to those omitted.
The LBNE geometry is nearly operational. Tom investigated the event
he pushed through the simulation last week only to find that no
charge was digitized. The problem was traced to the voxelization
not being set up for more than one TPC -- LArVoxelGeometry was built
for MicroBooNE and ArgoNeuT, and did not handle multi-TPC configurations.
Brian fixed this yesterday and checked in a version that voxelizes
each TPC one at a time. One consquence of this is that LArVoxelReadout
will now make a SimChannel instance for each hit channel in each TPC,
and this will result in some duplicates, as in LBNE channels will
record charge passing by the induction planes in more than one TPC.
SimWireXXX adds the charges together from all the SimChannels and
applies signal shaping and noise, performing the addition of
signals from multiple G4 steps. This can remain the same, though
the SimChannels will have to be merged so there are no duplicates.
Brian proposes this as a step as part of LArG4, before the wire simulation.
Tom is concerned that charge deposited in the gap between APA's may be
collected on induction wires, adding unipolar signals to what would
otherwise be expected to be only bipolar signals.
Jae Kim has been able to get muon events through a simulation using
a one-TPC version of the LBNE geometry, using a special GDML file
with just one TPC, and the event display works out of the box with it.
Herb suggests moving the noise simulation out of SimWireXXX and
into the signal shaping service to make it visible to other utilities.
This has the benefit of allowing the Wiener filter to be computed
from signal and background models, for use in noise filtering and hit
finding.
Tyler has volunteered to start work on the 35T GDML soon.
Jonathan has read a paper on stochastic noise generation by N. Jeremy
Kasdin, in IEEE proceedings, Vo. 83, No. 5 (1995). This paper provides
a review of several ways to approximate noise spectra digitally.
The existing noise model isn't so bad, but it could run faster and
take less memory, in the case of LBNE. Jonathan will code up the
methods suggested in the paper, as well as collect the noise model
from Andrzej.
Dave Muller reported on a fast simulation of light in a rectangular
acrylic bar immersed in LAr. He introduced bulk absorption and surface
reflection efficiency; both have strong effects on the attenuation of
light propagating in the bar. With available acrylic, however, it looks
as if more light is lost on the reflections than in the bulk. Dave
assumes a mirror on the other side of the acrylic bar, which has an
impact on the light collection as half the light initially bounces
away from the SiPM. Thicker bars reduce the number of bounces and
thus allow for more light collection. The effect is biggest for
cases in which the surface reflectivity is lower (no surprise). If
the surface reflectivity is 0.97 per bounce, then doubling the thickness
roughly doubles the light output.
The total of all of the effects gives an attenuation
curve that deviates from exponential largely due to geometric effects near
the end with the SiPM.
Ben Jones presented a study of light collection in acrylic bars in liquid
argon done at MIT. Measurements are detailed in arXiv:1210.3793.
The attenuation has the same characteristic shape Dave showed -- it falls
faster than exponential due to geometric effects of isotropically radiated
light from the wavelength-shifting coating getting to the detector more
easily at the detector end than if it is required to reflect. It is
also a sum of many exponentials given that light emitted at some angles
reflects more just to cover the same z distance and thus suffers more
absorption.
Ben found that for realistic lightguides, the bulk attenuation was
much less than surface reflection attenuation, and thus the bulk
attenuation could be neglectd. Losses per bounce on the surface
in the samples studied seemed higher -- about 6% per bounce.
The measured and predicted attenuation given this loss per bounce
are in good agreemnent.
Surface roughness due to macroscopic changes in the direction of the
reflecting plane have negligibly small effects for realistic surface
distortions. Mark Convery pointed out that features on the surface that
are comparable to the wavelength of the light traveling in the bar
would have to be modeled using more realistic techniques including
both specular and diffuse reflections and absorption. It may be possible
just to parameterize the effects as absorption. Dave suggests that
the wavelength shifting coating could fill in the rough gaps, though
the production details have to be controlled well to make sure the
surface of the WLS coating is itself highly smooth.
Ben put in wavelength-dependent effects. New data from Indiana
show wavelength-dependent attenuation effects. There is little
impact on the calculated absorption coefficients due to competing
effects.
There are minutes attached to this event.
Show them.