Pass 8 Overview
Transcript of Pass 8 Overview
![Page 1: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/1.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 1
Pass 8 Overview
Philippe BRUELLLR-CNRS/IN2P3-SLAC
on behalf of LAT collaboration
![Page 2: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/2.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 2
Overview
● Introduction● Event reconstruction● Data/simulation agreement● Event selection● Performance● Validation
![Page 3: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/3.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 3
What does Pass mean?● Each pass corresponds to a version of the Fermi LAT data● It implies a whole package:
● Intrument simulation● Reconstruction code● Event selection● Instrument Response Functions (IRFs)● Systematic uncertainties● Isotropic template (which includes the cosmic-ray residual background)● And sometimes more (Galactic diffuse model, Earth limb template,
Sun+Moon template)● It’s only when we have validated the whole package that we can release it to the public.
![Page 4: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/4.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 4
From Pass 6 to Pass 8
● Pass 6 (launch time)● Pass 6 reconstruction● Pass 6 selection● Based on pre-launch instrument simulation● First data revealed the issue of out-of-time pile-up (aka ghosts)● New: instrument simulation with ghosts -> correct IRFs
● Pass 7● Pass 6 reconstruction● New: Pass 7 selection optimized with simulations with ghosts
● Pass 8● New: improved instrument simulation● New: Pass 8 reconstruction, as ghost-proof as possible● New: Pass 8 selection
![Page 5: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/5.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 5
Pass 8 improvements● Ghost handling
● Tracker: ignoring ghost hits● ACD: partial deghosting● Calorimeter: clustering and cluster classification
● Improved direction measurement● Tree-based track finder
● Improved energy measurement● Extension of the energy range: from ~10 MeV to ~3 TeV
● Improved track/ACD matching information● Using the uncertainty of the tracker direction
● Improved event selection● Using the ROOT TMVA package (tmva.sourceforge.net)
● Additional sub-classes of events
![Page 6: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/6.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 6
Out-of-time pile-up● The slow signal characteristic time is: ACD=4μs, CAL=3.5μs, Tkr=10μs● The cosmic-ray rate up to 10kHz induces signals from out-of-time particles (aka ghosts)● This is taken into account in the simulation by overlaying on each simulated event a on-orbit periodic trigger (2Hz) event (event that has not been triggered by an incoming particle)
ghost
mainevent
![Page 7: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/7.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 7
De-ghostingACDRemove tiles with no trigger veto (based on fast signal) but with a signal that should have fired the veto
TrackerRemove hits from time-over-threshold information
CalorimeterPerform clustering and cluster classification and keep the most gamma-like
![Page 8: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/8.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 8
Tracking● Pass 7: track-by-track combinatorial pattern recognition
● Returns the optimum trajectory for each track
● Need a good seed: strong dependence on the CAL
● Track confusion and errors in high-multiplicity events
● Pass 8: global tree-based approach to track finding
● Reduce mistracking, improve the high energy PSF
● Provide additional information for background rejection
● No dependence on CAL
![Page 9: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/9.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 9
ACD
● The ACD is responsible for providing 0.9997% rejection power for charged particle entering the top or sides of the LAT● It must avoid self-vetoes from backsplash of high energy gamma rays.
● Pass 8: track and ACD tile association based on track covariant error propagation to the tiles
![Page 10: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/10.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 10
Calorimeter clustering● Pass 7 reconstruction assumes only one particle
● All crystals grouped in one cluster
● Small energy deposit by ghosts can have a large impact on cluster direction and transverse size.
● Pass 8 starts by performing a clustering, optimized such that:
● A gamma-ray event is not split● Ghosts are found as near as
possible to the main event cluster● Hadronic showers are not
subdivided in too many clusters (in order not to decrease rejection power)
![Page 11: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/11.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 11
Energy reconstruction● Pass 8 increases the LAT energy range:
● Down to E<50 MeV, by reconstructing events that do not reach the calorimeter. For these events the energy measurement is solely based on the number of hits in the tracker.
● Up to several TeV, by improving the shower profile fit:● New parameterization of shower development parameters as a
function of energy up to 10 TeV thanks to dedicated simulations● Phenomenological model of energy losses in gaps between towers● Crystal saturation handling thanks to a full 3D shower modeling
σ=15%
1<E<1.5 TeV, all cos(θ)One CAL modulePreliminary
![Page 12: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/12.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 12
Data/simulation agreement● The development of Pass 8 has been an opportunity to improve the data/simulation agreement.● The goal is:
● to have reliable variables that we can use for background rejection and quality subclasses selection
● to minimize the systematics uncertainties on the effective area● In Pass 7, we had to apply an ad hoc correction based on data/simulation comparison:
● Effective area correction in order to get a correct Front/Back ratio at low energy (energy dependent correction)
● In-flight PSF correction (the high energy prediction by the simulation was too small)
● Thanks to the improvement of the data/simulation agreement, there is no need of such ad-hoc corrections in Pass 8.
● The data/simulation comparison is performed using:● The Vela pulsar● The 30 brightest AGN● The Earth limb
![Page 13: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/13.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 13
Conversion point● The instrument simulation under-predicted the fraction of events converting in the first TKR plane. It implied that the amount of passive material between the ACD and TKR was slightly under-estimated.● We added to the simulation instrument model the amount of passive material (<1% X0) that gave the best data/simulation agreement.
P7REP P8
back front
100 MeV 1 GeV
100 GeV10 GeV
![Page 14: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/14.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 14
PSF at high energy
P7REP P8
● The PSF is very sensitive to the alignment of the planes in the tracker towers. The mis-alignments are corrected thanks to an alignment calibration derived from flight data.● Using slightly different (within measurement uncertainties) alignment calibrations during simulation and reconstruction allowed us to predict correctly the PSF at high energy
worst PSF best
100 MeV 1 GeV
10 GeV 100 GeV
![Page 15: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/15.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 15
Selection optimization (1)● The event selection is performed thanks to multivariate classification technique (using the TMVA package and Boosted Decision Trees)● This technique outputs a variable that is used to define the selection cuts● The power of this technique depends a lot on the data/simulation agreement
Gamma rays Background
SOURCE
TRANSIENT_R100
ULTRACLEAN
![Page 16: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/16.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 16
Selection optimization (2)● We use the simulation to produce templates for gammas and background● These templates are used to derive the background contamination in high-latitude data● The selection cuts are chosen such that the residual background contamination is a given fraction of the EGB (~60% for the SOURCE class) while keeping a smooth effective area as a function of energy● In Pass 8, we also developed a CT variable to reject back-entering photons
Preliminary
![Page 17: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/17.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 17
SOURCE class acceptance
![Page 18: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/18.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 18
Field of view
Field of view = Acceptance / on-axis effective area
![Page 19: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/19.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 19
PSF
![Page 20: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/20.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 20
Energy resolution
on-axis
![Page 21: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/21.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 21
Source sensitivity
Preliminary
P8 SOURCE
![Page 22: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/22.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 22
PSF event types● Split the data in 4 event types:
● with increasing quality of the PSF● with equal acceptance
● The optimal is to use the event types in a joint likelihood analysis(5-10% gain in sensitivity)
● Using >4 event types does not improve the sensitivity.
![Page 23: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/23.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 23
Energy disp. event types● Split the data in 4 event types:
● with increasing quality of the energy resolution● with equal acceptance
Pass 8 SOURCE100<E<177 GeV
![Page 24: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/24.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 24
Validation: acceptance● We measure the efficiency from a very loose selection to the SOURCE class selection for data (black) and the simulation (red)
VelaEarth limb
Preliminary
![Page 25: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/25.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 25
Validation: acceptance● We compare the data and simulation efficiencies:
Preliminary
VelaEarth limb
![Page 26: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/26.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 26
Validation: Front/Back● We measure the fraction of Front events in the SOURCE class selection for data (black) and the simulation (red)
VelaEarth limb
Preliminary
![Page 27: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/27.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 27
Validation: Front/Back● We compare the data and simulation fractions of Front events:
VelaEarth limb
Preliminary
![Page 28: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/28.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 28
Validation: PSF● Using Vela and the brightest AGN, we can compare the PSF measured from flight data to the one predicted by the simulation (and stored in the IRFs)
Preliminary Preliminary
![Page 29: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/29.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 29
Validation: PSF event types
Preliminary Preliminary
PreliminaryPreliminary
![Page 30: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/30.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 30
Pass 8 at low energy: Crab
![Page 31: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/31.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 31
Validation at low energy● In order to check how low in energy we can go, we perform a Vela ON-OFF spectral fit● It allows a validation of the IRFs >30 MeV● Taking into account energy dispersion allows us to:
● get a better fit● decrease the systematics
No energy dispersionχ2=117
With energy dispersionχ2=22
dataVelaoff
total
Preliminary
Preliminary
![Page 32: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/32.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 32
Fish-eye effect● Particles that scatter towards the center of the FoV are preferentially. triggered/reconstructed causing a bias in the PSF at low E and high theta. ● Transients may be observed with a narrow range of phi angles resulting in a large bias.● A correction is in development.
Vela ininstrumentcoordinates
Preliminary Preliminary
![Page 33: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/33.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 33
Pass 8 package● The Pass 8 data and a first reliable version of the IRFs have been validated
● We are now producing the rest of the Pass 8 package:● Diffuse model: for now we have rescaled the Pass 7 model
for energy dispersion distortion● Isotropic template● Earth limb
● The large Pass 8 field of view implies that we are more sensitive to the residual Earth limb
● Systematic uncertainties:● P7REP: 5% between 316 MeV and 10 GeV● Pass 8 goal: <2% between 100 MeV and 100 GeV
![Page 34: Pass 8 Overview](https://reader034.fdocuments.in/reader034/viewer/2022042611/588c69141a28ab30218b7f9d/html5/thumbnails/34.jpg)
Fermi LAT collaboration 5th Fermi Symposium Nagoya Oct 22 2014 34
Conclusions
● Pass 8 looks good and very promising
● First science results at presented at this Symposium
● We are currently finalizing the Pass 8 package
● Public release: ~mid 2015