A First Look At VERITAS Data
Stephen FeganVladimir Vassiliev
UCLA
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Data Analysis Flowchart
Image conditioning.Geometry reconstruction.Event parameterization.Energy reconstruction.Data Selection.
Independent telescope parameters. Array.
Results.
Standard approach. Picture/boundary cleaning. Has not been optimized.
Has not been investigated.
Array cuts, based on combining the various parameters from all images have not been investigated. To be investigated…
Methods 1&2 from reconstruction memo of December. Parameterization from memo also.
Cuts based on individual telescope images using parameters from memo. Simple “box” shaped cuts and multi-dimensional cuts using “NSpace”
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Geometry reconstruction
Method1: fit arrival direction and core from individual image axes, weighted “appropriately”.
Method 2: Fit single shower axis to all images simultaneously. This axis is projected onto focal plane of each telescope.
Method 3: Fit single shower axis to all images simultaneously. This axis is defined in real 3D space.
Methods 1 and 2 give comparable results with two telescopes. Method 3 has not been investigated.
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Event parameterizationOnce the shower axis has been reconstructed each image can be mapped into “physical space”, giving:
1) Mean emission height and track length.
2) Depths of mean emission in atmosphere in g/cm2.
3) Mean angle of emission with respect to trajectory.
4) Physical width of emission region.
5) Dispersions in photon arrival times assuming emission along primary trajectory.
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Data selection
Simple cuts on parameters in Dec. memo: Log10(size) > 2.5 Emission region width < 15m Emission height > 5000m Emission length > 750m
Simulations suggest cuts should be energy dependent. Studies using n-dimensional cutting system NSpace1 is ongoing. Preliminary results discussed here.1 See T. Nagai’s thesis for write up of NSpace methodology
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: Mrk 421 – 9 pairs
Background rate of 16.4/min/deg2 based on θ<0.2°
(SIMPLE CUTS)
Significance 29σ, rate 5.76/min
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: table of Mrk421 pairs
Run Duration ON OFF Rate Significance
30328 28.0 223 51 6.15 10.39
30330 28.0 253 64 6.75 10.62
30358 28.0 224 56 6.00 10.04
30394 28.0 192 46 5.22 9.46
30396 28.0 244 56 6.71 10.85
30398 28.0 256 62 6.93 10.88
30400 28.0 236 69 5.96 9.56
30476 28.0 167 65 3.64 6.70
30478 28.0 174 50 4.43 8.29
TOTAL 252.0 1969 519 5.76 29.07
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: Mrk 421 – 10 wobble at 0.3°
Theta2 distribution with respect to source position
Theta2 distribution with respect to symmetric wobble position
Source counts as they appear from off source location. No obvious spill over to theta=0 area.Wobble offset of 0.3º seems to be sufficient.
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: table of Mrk421 wobble runsRun Duration ON OFF Rate Significance
30480 15.0 63 30 2.20 3.42
30490 28.0 154 38 4.14 8.37
30491 28.0 175 57 4.21 7.75
30492 28.0 182 73 3.89 6.83
30493 28.0 164 70 3.36 6.14
30494 28.0 176 64 4.00 7.23
30495 28.0 176 58 4.21 7.71
30496 28.0 157 73 3.00 5.54
30497 28.0 140 71 2.46 4.75
30572 28.0 48 30 0.64 2.04
TOTAL 266.95 1435 564 3.26 19.48
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: strong source optimization
Photon dominated regime:For strong source, optimize theta cut for significance. Peak is quite broad. Theta cut from 0.15° to 0.25° gives high significance. We chose 0.2° for high rate.
9 Mrk421 pairs
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: weak source optimization
Background dominated regime:For weak source optimize theta cut for excess/sqrt(background) – Q-value. Shown is Q-value, normalized to 1 at theta=1. Peak is much narrower. Theta cut from 0.12° to 0.16° gives best significance on weak sources.
9 Mrk421 pairs
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: Mrk421comparison with Whipple
VERITASRun
VERITASRate
VERITASSignificance
Whipple Run
WhippleRate
WhippleSignificance1
30358 6.00 10.0 313062 1.77 6.2
30398 6.93 10.9 31323 2.72 8.95
31324 2.97 9.20
30400 5.96 9.6 31325 2.86 9.12
30476 3.64 6.7 31369 1.81 6.79
30478 4.43 8.3 31372 1.55 5.45
1 Whipple results from J. Kildea2 VERITAS and Whipple run numbers are similar after 40 years of Whipple operation
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: Mrk501comparison with Whipple
VERITASRun
VERITASRate
VERITASSignificance
Whipple Run
WhippleRate
WhippleSignificance1
30483 0.97 2.30 31377 0.11 0.53
30498 0.29 0.71 31400 0.33 1.66
30600 0.71 1.78 31478 0.22 1.20
30844 0.18 0.43 31635 -0.04 -0.19
30863 0.40 0.88 31659 0.19 0.82
30878 0.83 1.98 31677 0.22 1.09
30897 -0.04 -0.08 31694 0.15 0.88
30898 0.18 0.42 31696 0.04 0.191 Whipple results from J. Kildea
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Results: aperture function
Cosmic-ray acceptance as a function of off-axis angle, after cuts.
These events are the most gamma-ray like CRs. Gamma-ray response will likely be somewhat similar.
Quite flat to 0.8 degrees.
Fast decline at 1.75 degrees
(SIMPLE CUTS)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: basic ideas Form multi-dimensional histogram in space defined by
event parameters. Fill two histograms: from ON & OFF runs or from SIM &
OFF data. Order cells “according to significance (likelihood,…)”
(many ordering schemes possible). Make “filter” by picking cells from ordering to maximize
total significance (likelihood,…). Repeat all steps above varying bin sizes of the
histogram until significance (likelihood,…) is maximized. Bin sizes and histogram dimension are your optimization parameters. Final filter is an effective volume in multidimensional space.
Cut ON events according to whether they are inside or outside the filter.
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: example in two dimensions
log(size)
width
2.0-3.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-4.5
0.00-0.02 6747
6852
10044
7645
6547
0.02-0.04 6960
9944
9458
7950
6758
0.04-0.06 6855
8646
10155
9853
6149
0.06-0.08 6150
5943
6842
8044
5059
0.08-0.10 4949
5760
4551
5559
4865
0.10-0.12 5840
5248
6250
5445
4343
ON countsOFF counts
Highest single cell significance(4.7 sigma)Two cell set with highest overall significance(6.5 sigma)Seven cell set with highest
overall significance (9.9 sigma). Thereafter adding any more cells decreases significance.
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: single telescope spaceSpace based on three single telescope
parameters: Log10(λc)* 12 binsno 4 – 8 Emission width 20 binsno 0 – 40 m Emission depth 10 binsno 0 – 1000 g/cm2
ON histogram: 9 Mrk421 runs – theta<0.2OFF histogram: 9 Mrk421 runs – theta<1.0
Optimize filter on T1 parametersApply to both T1 & T2 when cutting
* λc : density of emitters in coherent regime [m-1], see December memo for details
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: optimization results
Filter with 70 cells gives close to the peak Q-value and gives good rate
For comparison, straight cuts have Qmax=6
(with theta cut of 0.16°)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: optimization results
Filter with 70 cells gives close to the peak Q-value and gives good rate
For comparison, straight cuts have Qmax=6
(with theta cut of 0.16°)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Straight Cuts NSpace Cuts
Data Set ON OFF Rate Sigma ON OFF Rate Sigma
9 Mrk421 Pairs1
1436 291 4.54 27.55 1008 102 3.60 27.19
8.5 Mrk421 Wobble
1037 323 2.67 19.36 596 81 1.93 19.79
19 Mrk501Wobble
863 582 0.52 7.39 567 299 0.50 9.11
Faint source improvement not as large as indicated from optimization (1.332=1.76 vs 1.232=1.51). This reflects some degree of “tuning” of space to specific runs used in optimization.
NSpace: comparison with straight cuts
(NSpace with theta cut of 0.15°)(Straight with theta cut of 0.14°)
1 Optimization done on these runs
Bright source significance is approximately equal with this NSpace filter but rate is lower. A theta cut of 0.2 gives better significance and rate for both sets of cuts.
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: 10 Mrk421 wobble runs
Background rate of 5.0/min/deg2
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
Run Duration ON OFF Rate Significance1
30483 29.8 43 16 0.91 3.52
30484 28.0 44 11 1.18 4.45
30486 27.7 43 16 0.98 3.52
30487 16.7 16 6 0.60 2.13
30498 28.0 21 13 0.29 1.37
30499 28.0 33 15 0.64 2.60
30500 28.0 25 12 0.46 2.14
30501 18.9 11 13 -0.11 -0.41
30602 28.0 20 8 0.43 2.27
30603 28.0 30 14 0.57 2.41
30822 17.7 23 10 0.73 2.26
30823 28.0 31 15 0.57 2.36
30863 30.0 27 23 0.13 0.57
30864 30.0 32 13 0.63 2.83
30878 30.0 36 23 0.43 1.69
30879 30.0 29 20 0.30 1.29
30897 28.0 20 16 0.14 0.67
30898 28.0 19 21 -0.07 -0.32
30899 28.0 23 17 0.21 0.96
TOTAL 538.7 567 299 0.50 9.11
NSpace: results for Mrk501 wobble
1 Using traditional Gaussian significance formula (somewhat inaccurately)
If Crab is 5/min then4% Crab in 20hours/wobble @ 5
or 18% Crab in 1hour/wobble @ 5
HESS claimed11% Crab in 1hour/wobble @ 5
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: 501 comparison with Whipple
VERITASRun
VERITASSignificance
VERITASRate
Whipple Run
WhippleSignificance1
WhippleRate
30483 3.52 0.91 31377 0.53 0.11
30498 1.37 0.29 31400 1.66 0.33
30600 3.18 0.64 31478 1.20 0.22
30844 0.80 0.18 31635 -0.19 -0.04
30863 0.57 0.13 31659 0.82 0.19
30878 1.69 0.43 31677 1.09 0.22
30897 0.67 0.14 31694 0.88 0.15
30898 -0.32 -0.07 31696 0.19 0.041 Whipple results from J. Kildea
(NSpace)
VERITAS Collaboration Meeting June 29, 2006, Leeds, UK
A First Look At VERITAS Data Stephen Fegan & Vladimir Vassiliev
NSpace: future work Preliminary optimization done quickly for
meeting, need to look in detail at: Ordering strategy Array/Energy parameters Bin size optimization Parameter choice/Dimension Optimization
9 runs are not enough to do filter selection properly. Need lots of CRAB data and/or simulations.
NSpace provides flexible and robust frame of work for data analysis including detailed spectrum analysis.
Optimistic about success of this method!