Adding a Level 1 Trigger to Calibrate the CMS ECAL with p0s - PowerPoint PPT Presentation

Title: Adding a Level 1 Trigger to Calibrate the CMS ECAL with p0s


1
Adding a Level 1 Trigger to Calibrate the CMS
ECAL with p0s

• Sean Simon, Elizabeth Dusinberre, Jim Branson

2
Outline

• The Parasitic p0 Calibration
• It works
• Some problems
• p0 Calibration with a L1 Trigger
• Less systematic problems
• Calibrate at high ? faster
• Use less of the HLT Farm
• CMS can attempt both parasitic and triggered
  calibration
• The basic L1 trigger and p0 selection
• Triggering for High ?
• Outlook

3
Parasitic Calibration

• Use all the 50 kHz of L1 triggers
• Called parasitic because it does not have its own
  trigger. All the p0s are volunteers.
• Search for p0s around all L1 isolated electron
  triggers
• About 2-5 of the L1 triggered events have a
  usable p0
• Reconstruct like HLT (High rate)
• Write minimal amount of info per event to a
  calibration stream.

1-2 days to calibrate
4
p0 Calibration Status

• It has been shown that the CMS ECAL can be
  calibrated with a statistical error of less than
  0.5 using the decay p0 ? ??
• Method is surprisingly insensitive to s/b ratio.

2-3000 p0s needed for a typical crystal
in the barrel several times more for high
? crystals.
Need 108 p0s total
5
Systematics of Parasitic Calibration
In QCD events there is a systematic shift in the
calibration due to extra energy from signal p0s
in jets
Mean 0.993
Mean 0.969
Calibration on out-of-jet p0s show no shift
Even without background, in-jet p0s show
systematic shift
6
High ? in Parasitic Calibration

• About 0.5-1 of L1 triggers have a p0 at ?gt1
• 3 days to calibrate
  ?gt1
• We want to improve on that to calibrate the whole
  barrel rather quickly.

7
p0 Calibration with L1 Trigger(s)

• Develop L1 triggers that have a high probability
  to yield usable p0s.
• Trigger will likely be pre-scaled to some fixed
  L1 rate.
• It is possible to drop all but ECAL regional data
  in the RU.
• Bandwidth limit is at ECAL FED output.
• Could take 10 kHz or more of this trigger.

8
Study Use of a L1 Trigger

• Analysis using CMSSW_1_3_1_HLT6
• Need p0s (ECAL SR) and L1 trigger emulator to
  work
• Probably still not true in any CMSSW version
  released
• 2M MinBias events using CRAB
• /MinBias/CMSSW_1_3_4-Spring07-1679/GEN-SIM-DIGI-R
  ECO
• Goal Formulate a L1 trigger decision optimizing
  the ratio of signal p0s reconstructed to number
  of triggered events (p0s/event) and signal to
  background (S/B)

9
ECAL Selective Readout

• We use the default SRO in CMSSW131
• Caltech noticed differences between ORCA and
  CMSSW131 due to SRO
• A new SRO is implemented in CMSSW 160

ORCA
CMSSW
V. Litvin, Pi0 Calibration and Selective
Readout, ECAL DPG, 29 May 2007
Implementation of SRO may not be fully understood
10
L1Trigger Information
Note ? asymmetry!

• L1 Extra Particles (ET, ?, ?, )
• Isolated and Non-isolated EM
• Forward, Central, and Tau Jets
• L1 ET Sums
• Total ET
• Hadronic ET
• Missing ET

11
Also found ?? Problem

• The ? of the L1 objects is shifted up by half the
  width of a supermodule (10? or 0.175 radian)

L1 emulator people found this error a week prior.
Fixed in CMSSW_1_6_0 onward
a)
b)
Signal Background
Signal Background
?? between p0 and the nearest L1isoEm a) original
b) fixed
12
p0 Selection

• Mass between 0.095 and 0.14 GeV
• wider window for background analysis
• Isolation No energy
• within the blue shaded
• region (?R 0.2 and ? ? 0.05)
• Reconstructed p0 within
• ?R lt 0.3 of nearest
• L1 isolated trigger

We have removed many cuts because larger number
of p0s is more important than good S/B.
13
p0 Selection
Note Ks??0?0 indication
Signal Background
0.095 lt mp0 lt 0.14
Signal Background
Signal Background
lt 0.05
?R lt 0.3
14
L1 Trigger Cuts
L1isoEm_Et gt 2.5 Gev
1 L1isoEms EVTS 29800
SGNL 6891 BKGD 3373 S/B 2.04
SGNL/EVT 0.231 2 L1isoEms EVTS 5506
SGNL 1853 BKGD 983 S/B 1.89
SGNL/EVT 0.337 3 L1isoEms EVTS 948
SGNL 371 BKGD 200 S/B 1.85
SGNL/EVT 0.391 4 L1isoEms EVTS 140
SGNL 55 BKGD 32 S/B 1.72
SGNL/EVT 0.393 CUMMULATIVE TABLE
1L1isoEms EVTS 36394 SGNL 9170
BKGD 4588 S/B 2 SGNL/EVT 0.252
2L1isoEms EVTS 6594 SGNL 2279 BKGD
1215 S/B 1.88 SGNL/EVT 0.346 3L1isoEms
EVTS 1088 SGNL 426 BKGD 232 S/B 1.84
SGNL/EVT 0.392 4L1isoEms EVTS 140
SGNL 55 BKGD 32 S/B1.72 SGNL/EVT
0.393

• Cuts studied include
• L1isoEm ET threshold
• nL1isoEm passing ET
• L1TotalEt

gt 2.5 GeV
gt1
40 - 140 GeV
15
L1 Trigger Cuts
40 lt L1TotalEt lt 140 GeV
Signal Background
16
Increased p0 Yield with L1 Trigger

• L1 Trigger Decision
• At least 2 L1isoEm objects with ET gt 2.5 GeV
• Require L1TotalEt to be in the range of 40 to 140
  GeV
• We see that a p0/event of 35 can be achieved
  while maintaining a S/B 1.9
• Can calibrate in about 6 days / kHz L1
  trigger
• less than a day using 10 kHz of L1
  trigger

17
L1 Trigger and High ? Calibration
Signal Background
?
?

• The p0s found are not uniformly distributed in ?
• We observe a reduced rate of p0s and worse S/B at
  high ?
• Propose a L1 trigger specifically aimed at
  optimizing p0s/event and S/B at high ?

?
18
L1 Trigger and High ? Calibration

• Cuts Consider only ? gt 1
• L1isoEm ET threshold
• nL1isoEm passing ET
• L1TotalEt

L1isoEm_Et gt 3.5 Gev
1 L1isoEms EVTS 9409
SGNL 756 BKGD 452 S/B 1.67 SGNL/EVT
0.0803 2 L1isoEms EVTS 318 SGNL 39
BKGD 36 S/B 1.08 SGNL/EVT 0.123 3
L1isoEms EVTS 11 SGNL 1 BKGD 4
S/B 0.25 SGNL/EVT 0.0909 4 L1isoEms
EVTS 0 SGNL 0 BKGD 0 S/B 0
SGNL/EVT 0 CUMMULATIVE
TABLE 1L1isoEms EVTS 9738
SGNL 796 BKGD 492 S/B 1.62 SGNL/EVT
0.0817 2L1isoEms EVTS 329 SGNL 40
BKGD 40 S/B 1 SGNL/EVT 0.122
3L1isoEms EVTS 11 SGNL 1
BKGD 4 S/B 0.25 SGNL/EVT 0.0909
4L1isoEms EVTS 0 SGNL 0
BKGD 0 S/B 0 SGNL/EVT 0


gt 0
gt 3.5 GeV
lt 130 GeV
?
19
L1 Trigger and High ? Calibration
L1TotalEt lt 130 GeV
Signal Background
20
Increased p0 Yield with High ? L1 Trigger

• L1 Trigger Decision
• At least 1 L1isoEm object at ?gt1 with ET gt 3.5
  GeV
• Require L1TotalEt to be less than 130 GeV
• We can acheive a p0/event 8 with a S/B 1.6

also calibrate in about 6 days / kHz L1 trigger
if used in conjunction with the basic L1 trigger
discussed above
21
Outlook

• Need fix for asymmetry in L1 ??
• (expect improvement)
• Bump in mass peak (KS?)
• Effects of Pile-up
• Check mass shift for trigger/no trigger
• Write proposal to implement L1 trigger

22
References

• CMS TDR Vol 1- Detector Performance and Software
• Gataullin, Marat, et al. Calibration of the CMS
  Barrel Electromagnetic Calorimeter Using p0 ? ??
  Decays, CMS Internal Note, 29 Dec, 2006
• Vladimir Litvin, Marat Gataullin Pi0 Calibration
  and Selective Readout, Talk given at ECAL DPG
  meeting, 29 May, 2007

23
Backup
24
L1 Total-ET Cut (p0s matched to any L1isoEm)
No Cut
40 lt L1 Total-ET lt 140 GeV
25
L1 Total-ET Cut (p0s matched only to triggered
L1isoEm)
No Cut
40 lt L1 Total-ET lt 140 GeV
26
High ? L1 Total-ET Cut (p0s matched to any
L1isoEm)
L1 Total-ETlt 130 GeV
No Cut
27
L1TotalEt in Various Triggers
1
1
2
2
3
4
3
4
No L1isoEm ET Cut
L1isoEm ET gt 2.5 GeV
Can see dependence of L1 Total ET on the L1 ET
threshold and the number of L1isoEms passing this
threshold
28
How long does it take to calibrate?
Calibration Time
1 day 2 days 3 days 6 days
Yield of our L1 trigger
29
Delta R between L1 isolated triggers
Signal Background
No useful cut here
30
Planned Calibrations

• Crystals will be pre-calibrated to about 4 using
  photons from a 60Co source, some calibrated more
  precisely with the test beam and cosmic rays
• Use ? independence to inter-calibrate to about
  1.5 precision for each phi ring in 10 hrs on 1
  kHz of jet triggers
• Use high energy electrons from Zs and Ws to
  calibrate in situ, but this will take months at
  start up
• Need tracker operational and well -
• aligned
• Electrons radiate in tracker
• Need 5fb-1 and that only calibrates
• the inner barrel to 0.5

5 fb-1 W ? e?
31
Planned Calibrations Continued

• Laser monitoring system to track the fastest
  changes due to radiation damage
• Use some of 3?s beam gap to inject pulse
• Takes about 30 minutes to do whole ECAL
• Possibility of p0 calibration suggested in 2004
  and, at first, dismissed due to low rate
• Later retried with improved selection and idea of
  using all L1 Triggers as source for calibration
  events