You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 20 Next »

LAT Newsletter contributions from Beam Test Team

Issue 5 - submitted 25 july

Luca Latronico
July 25 2007

In the past two months the team performed much work behind the lines on the core software, to reorganize the BT sim/recon package (BTRelease), verify the consistency of the Geant4 simulation and better support the analysis. The main areas of activity have been:

  • synchronization with GlastRelease and updates of core packages (hadronic physics list management, G4Generator), which will help transferring BT results to the main LAT simulation
  • beam spot tuning from real data and transfer to the simulation
  • comparison with mars15 simulation and interaction with developers, which revealed inadequacy of the mars15 code wrt G4 for describing a simple EM cascade
  • study of the CAL response with different segmentations of the log, which did not show any dependance of the simulated CAL signal on the level of segmentation

The attempt to summarize the results and map the still existing discrepancies between data and MC, mostly in the TKR hit counts and the CAL energy deposit, continued and will be presented in the collaboration meeting talk:

  • summary of the TKR hit and cluster MC simulation for hadronic events and comparison with data
  • stability check of the CAL response in similar electron runs revealed dependancies of the relative CAL-TKR position and the effect of tkr converter gaps in the shower development; this confirms the importance of a good matching of the beam profile for comparison with the simulation
  • cal response to pion runs
  • angular resolution studies with high energy electrons and low energy tagged photons to complete the characterization of the TKR performance over the whole spectrum
  • PSF studies with covariance matrix
  • started collecting existing and new ideas towards a set of selected runs and plots to produce for each new MC production

In order to start transferring our results to the LAT simulation as soon as possible and without delaying the production of the one year data simulation of the Service Challenge, we plan to provide now the existing improvements of the simulation to GR, and keep on evaluating the impact of the discrepancies through the production of custom files for the LAT with built-in discrepancies and potential new fixes we may find, to assess the impact of such discrepancies to background rejection and the LAT performances.

Issue 4 - submitted 31 may

Luca Latronico
May 31 2007

The Calibration Unit worldwide tour was successfully concluded in the beginning of April, when the module was sent to SLAC and handed off to ISOC-SO that will turn it into a test bed to for studying flight hardware response to flight software.
This was the last step of a successfull series of CU operations carried on by the beam test team since its first integration just about one year ago .
We are happy to deliver such a valuable instrument to the ISOC, and completion of data analysis is now our only target.

On this ground, three main activites were pursued during the last month:

  • re-evaluation of data-MC discrepancies after the last data reprocess and updates to the simulation
  • completion of the standalone single-tower simulation and comparison with our standard Gleam MC
  • systematic exploration of available hadronic physics lists

The most recent changes to our simulation/reconstruction package (BTRelease) include the adjacent crystal cross-talk correction for the calorimeter energy measurement, two alternative Tracker digitization algorithms with description of charge sharing and large signal induced cross-talks, and a correct use of the tracker calibration database.
These last improvements did not significantly improve the data/MC agreement in the CAL energy and in the number of tracker hits, but provided a very accurate description of the measured ToT distribution.

The standalone Geant4 simulation was completed with a realistic tracker tray honeycomb description, as opposed to a density-averaged homogeneuous material. No effect was observed in the total number of hits and in the EM shower development, which is also very similar to the one produced by Gleam. Our conclusions are that Gleam and a standalone Geant4 simulation generate similar EM shower profiles, and a more realistic tray description does not solve either the CAL energy or TKR hits discrepancy.

A more complete survey of the available hadronic physics list in Geant4 was performed, and the currently best results are obtained with the Bertini model up tp 10GeV, and the QGSP libraries above 20GeV. Mor models are being studied now

Issue 3 - submitted 30 april

Luca Latronico, Philippe Bruel
April 30 2007

A very detailed status report of the beam test data analysis was presented at the last collaboration meeting, in a dedicated session which was very well received by the collaboration.
The CU response to irradiation was described in specific talks, as well as the changes suggested for the MonteCarlo simulation of the subsystems in order to get a better agreement with data.
The overall status was summarized in a final talk for the collaboration, where the main outcomes that we expect to deliver to the LAT simulation were presented. These include

  • a new TKR digitization algorithm, including charge sharing, and cross talk effects for large signals (heavy ions saturating input charge amplifier); this will partly cure the discrepancy
  • a new CAL modules calibration procedure that takes into account electronics cross talks and charge injection asimmetries
  • an optimized hadronic physics list that identifies the best Geant4 simulation engines for hadronic interactions in the LAT
  • a comprehensive material review extending to all subsystems

The schedule for these deliverables was later discussed inside our working group, and we agreed to release a first round of these changes by the end of april, to be able to evaluate the effects on data-MC comparison and support the implementation in GlastRelease.
As this issues goes in press, we are in fact going through a massive data reprocessing and MC generation with an updated version of our simulation/reconstruction package (BTRelease), which incorporates the latest calibrations, a new TKR digitization algorithm and the xtalk correction in the CAL.

Studies on hadronic phyisics are an on-going hot topic, as is a standalone G4 simulation with a realistic honeycomb structure that aims at testing the difference between homogeneous, distributed materials versus isolated material (like the TKR honeycomb). Preliminary results indicate no effect on the CAL energy deposition, while the impact on the TKR hit discrepancy is under evaluation.

Finally, the proceedings of the four contributions we submitted to the First Glast Symposium have been combined into a single paper and submitted for publication.

Issue 2 - Studying the ACD Backsplash with beam test Data

The capability of the LAT to measure photon energies up to 300 GeV with good energy resolution requires the presence of a heavy calorimeter to absorb enough of the electromagnetic cascade produced by the incident gamma-ray. Unfortunately, a small fraction of secondary particles in the shower can travel backwards from the calorimeter into the tracker and up to the Anti-Coincidence Detector (ACD). This backsplash radiation consist mostly of 100-1000 keV photons and represent a potential problem since it can generate a signal in the ACD that would cause the gamma-ray to be interpreted as background and therefore rejected.

For this reason, the LAT ACD was designed as a segmented detector, so that only the ACD segment intersected by the backwards projected path of the particle is used to veto the event. In this way, the ACD area that contributes to backsplash is relatively small. The ACD hit probability per unit area as a function of energy and distance backwards from the shower has been studied with past beam tests (Moiseev, A. A., et al. 2004, Astroparticle Physics, 22, 275) and used to optimize the level of segmentation in the ACD design. The backsplash probability was measured with the as-built detector in the Calibration Unit Beam Test campaign in summer 2006, and the capability of the LAT Monte Carlo simulations to reproduce backsplash effect has been verified.

A careful analysis conducted by Luis C. Reyes demonstrated that the current LAT simulations reproduce well the backsplash effect. The LAT simulations take into consideration the energy loss fluctuations in the ACD tile, the Poisson fluctuations in the number of photoelectrons created in the readout photo-multiplier, and the corrections due to the non-uniform light collection at the edge of each tile. The latter currently represents the largest source of uncertainty in the present simulation, as it has not been measured yet for the ACD tiles installed on the LAT calibration unit. The expected backsplash distribution is therefore bracketed in the analysis by considering the maximum and minimum light collection efficiency measured for the tiles in the LAT (Moiseev et al, ACD paper, in preparation).

The result is shown in figure 1, and a beautiful reassuring agreement can be seen.

Details of this analysis will be shown at the collaboration meeting at GSFC next week, in a dedicated plenary session from the Beam Test Team where the state of the art of Data-MonteCarlo comparison will be presented to the collaboration.

Figure 1 - see attachment jpg
Figure 1 explanation: Backsplash probability distribution for an ACD tile as obtained from beam test data (black points) and Monte Carlo expectations. In every case, backsplash is expressed as the fraction of events for which the signal in the tile is above a given threshold. The error bars in the data are statistical 1 sigma. Monte Carlo simulations consider two extreme scenarios of light collection efficiency through the tile edge. In the MIN collection efficiency scenario, the collection efficiency decreases linearly from 100% (3 cm away from the edge) to 70% at the tile edge. In the MAX case, the light collection decreases linearly from 100% (1 cm away from the edge) to 90% at the tile edge. Both scenarios are shown in the backsplash distribution as bands that bracket the expected backsplash distribution. The width of each band is given by twice the statistical error 2 _sigma_obtained from the simulation.

Issue 1 - Fine-tuning the LAT

See the released issue here

  • No labels