1. CMS ECAL Cosmic Calibration
Jason Haupt University Of Minnesota
On behalf of the CMS ECAL collaboration

2. Compact Muon Soleniod (CMS)
CALORIMETERS
SUPERCONDUCTING
COIL
ECAL e 
Scintillating PbWO4 Crystals
HCAL
Plastic scintillator/ brass sampling
IRON YOKE
TRACKER
Silicon Micro Strips Pixels
CMS, near Cessy France. Talk outward from the center (interaction point). First within a 4T magnetic field
Silicon Tracker first… used for charge and momentum measurements in the mag field
ECAL, lead tungstate scintillating crystals. For photon and electron identification
HCAL, brass/scintillator sampling. Hadronic interactions
Muon (Chambers). 3 types. Charge determination, momentum measurement.
4 T magnetic Field
Total Weight 12,500 t
Overall Diameter 15m
Overall Length 21.6m
MUON BARREL
Drift Tube
Resistive Plate
Chambers ( ) DT
Chambers ( )
RPC
7/31/2019

3. ECAL 61200 14684 Module 1 Module 2 Module 3 Module 4
Note the 4 modules per SM. Note eta ranges of ECAL. Shape differences between the different modules. 61,200 barrel crystals. Endcap out to eta of 3.
7/31/2019

4. PbWO4 Crystals Lead Tungstate Crystals
Moliére radius 2.2cm
Radiation Length 0.89cm
Scintillation decay time 80% at 35ns
Been shown to be radiation resistant
Low LY compared to other commonly used crystals (CsI BGO …)
-1.9%/oC temp dependence
Lead Tungstate Crystals in CMS (Barrel)
“Average size”, 2.4x2.4cm2 and 23cm in length
34 Different crystal shapes
25.8 Xo
36 Supermodules make up the barrel calorimeter
Each Supermodule has 1700 crystals
Lead Tungstate Crystals: Good qualities… (Maybe more about radiation hardness/recovery)
Mention the reasonable LY and acceptable Temperature dependence. This is kept small by keeping the temperature to be known to within 0.1 degrees C.
Also mention the average size and makeup of the different crystals. Size of a super module
7/31/2019

5. ECAL Barrel Optical Readout
Two 5x5 mm2 APD’s/crystal
Gain – 50
QE – 420 nm
Temp sensitivity – -2.4%/oC
APD’s Nice QE. Great Linearity. Temperature sensitivity. Important that PMT’s cannot be used due to the high magnetic field.
≈ 4.5 photo-electrons/MeV
Total APD’s
Very Linear Devices
7/31/2019

6. ECAL Inter-calibration Goals
Energy Resolution
Goal : constant term “c”  0.5%  /E  0.5% (For High Energies)
In-situ Calibrations
Z  e+e- ~1 day 1% (with  ring inter-calibration)
W±  e± ~2 months E/p from Tracker
0 , 0  , etc.
Initial inter−calibrations
LY ~4%
Cosmics ~2%
Test Beam ~0.3% (Not available for all SM’s)
Reason for pre−calibration
Uniform detector response at startup
Can mention the other terms if needed, or include all of the energy resolution terms. Here I just want to point out that the dominate term at high energies is the constant term. In situ there are a few ways of accomplishing that 0.5% goal. Strict mass constraint on the Z decay and the W also provide excellent inter-calibrations.
Pre inter-calibrations are critical to ensure a uniform detector response at startup.
a = stochastic term (Shower containment fluctuations…)
b = noise term (Pileup and electronics noise)
c = contant term (Temp, uniformity, inter-calibration, …)
7/31/2019

7. Calibration Chain Crystal Energy → ADC count Crystal optical response
APD Gain
Amplifier Gain
ADC 12bit Out
Calibration Chain
x12 x6 x1
MGPA
Logic
12 bit ADC 2 1
12 bits
2 bits HV
APD/VPT
VFE architecture for single channel
Calibration Chain. Calibration Number includes all of the above components. Test pulse takes out the electronics effects. Laser can look at the whole chain.
7/31/2019

8. Performance Checks Cross Checks Test Pulse (after APD)
Compared to previous test pulses
Laser allows for self referencing
Compare one laser run to another
x12 x6 x1
MGPA
Logic
12 bit ADC 2 1
12 bits
2 bits HV
APD/VPT
VFE architecture for single channel
Laser
Calibration Chain. Calibration Number includes all of the above components. Test pulse takes out the electronics effects. Laser can look at the whole chain.
Test Pulse
7/31/2019

9. Laboratory Inter-Calibration
Two current methods to LY measurements (Basically Quality Checks) automated
1. Direct LY along crystal
~1.2 MeV source
2. Transmission through crystals longitudinally at 360nm
Combined Laboratory constants
Laboratory measurements are combined; LY, APD gain, the preamp.
Result of a ~4.0% agreement compared to testbeam calibration constants
Comparing ~1.2 MeV Source to 120 GeV testbeam!
60Co
Laboratory measurements were done on the elctronics and have shown to be well know and can be tested well with the test pulse. LY measurements then add to make a laboratory inter-calibration constant. LY refers to the direct number of pe’s per MeV. Mention just “briefly” that the transmission is also included in the measurement. (It has been shown that the LY and the Transmission at 360nm are well correlated. They are independent measurements and can be combined together to a single intercalibration constant)
The overall agreement is well. This was validated with test beam in There have been some recent improvements along these lines.
7/31/2019

10. Cosmic Ray Muons Cosmic ray muons deposit ~11 MeV/cm in PbWO4 crystals
A through going muon will deposit ~250 MeV in a 23cm crystal
APD gain was increased factor of 4 to get away from the ~38 MeV electronics noise. Actual gain ratio is found with laser. Also to improve the neighbor veto.
1 ADC count ~9 MeV
Goal
Improve measurement for all 61,200 barrel crystals with full readout chain
Excellent way to run each SM ~10 days as a final “burn-in” step
Maybe not call them “MIPs” they are on the relativistic rise of the dE/dx curve. Ave ~4 GeV (MIP ~0.5 GeV). Really note that the reason for the gain change is to keep the S/N as low as possible for this measurement. Also emphasize that the cosmic test is a great way to do the “burn-in” and electronics tests on a fully constructed Super Module. Really mention the Full Readout Chain.
7/31/2019

11. Cosmic Muon Trigger
The trigger is a coincidence of one of 6 plastic scintillator counters spread over the bottom of the Super-Module with another counter placed near the interaction point.
Trigger is designed as to select muons that are through-going. In this situation the amount of deposited energy is most well determined.
85x20 iηxi grid
One SM has more
than 34 Million Triggers
Most have 4-7 M
23 SM’s have been
calibrated with cosmics
iη (85)
Main reason of the trigger is to eliminate the need for a tracker. Doing a calorimetric selection really shows the ability of the crystals. The trigger was setup to collect data when a muon is through-going. Also some muons that are “mostly” through-going are selected as well. This was an effort to maximize the useful triggers
 i (20)
10o
24o
7/31/2019

12. Cosmic Trigger Setup Trigger Counters Top Frame 6 Bottom Counters 4 51 2 3
Just a reiteration of the trigger/cosmic setup.
Top Frame
7/31/2019

13. Crystal Selection E1 is the Energy of the Maximal crystal
E2 is the energy of the next highest crystal around the max
RED marks selection E1
> 10 ADC Counts
Single
Crystal E2 E1 E2
< 3 ADC Counts
2004 data SM10
G4 MC
Here I go over the selection for all 3 types of data. Mention that they are independent selections. All calorimetric selections.
7/31/2019

14. All 3 selection types are independent
Crystal Selection E3
EA is the Energy of the Maximal crystal
EB is the energy of the next highest crystal around the max
E3 refers to all other surrounding crystals
Magenta marks selection  EA EB
Double Crystal η
All 3 selection types are independent E3
Same  EA EA
& EB
> 3 ADC counts EB EA + EB
> 12 ADC counts
Same  E3
< 3 ADC counts
2004 data SM10
Here I go over the selection for all 3 types of data. Mention that they are independent selections. All calorimetric selections.
7/31/2019

15. Useful Events ~25% of the triggers combined! Of the triggers
Single ~10% Same eta ~6.8% Same phi ~8.6%
~500,000 Triggers/Day
~10 Days
~25% of the triggers combined!
The numbers here are currently relative numbers. And relate to a specific SM. They are close to the average. So the number of useful triggers is more than doubled.
7/31/2019

16. MC Reference distributions
Single Crystal Method
Pulse-height distributions of the crystal of highest energy deposited (E1) are made for each crystal
Pulse-height distributions are also generated from MC information.
17 different η dependent MC energy distributions are created.
The constant is then found by adjusting the pulse-height to MC with an unbinned maximum likelihood method.
Factor: Relative Scale = Calibration constant
MC Reference distributions
Soon, these E1 reference histograms will be made from a beam calibrated SM rather than relying on the Monte Carlo. But the number of reference distributions are expected to remain the same. Here are two fits from the CMS not quoted at the end of the paper.
7/31/2019
ADC
ADC

17. Two Crystal Method (Matrix Inversion)
Energy = Sum of both crystals
Define a 2
Minimize the 2 for each constant
Fill Matrix event by event
Invert
Only Input is the mean of the MC distributions
17 Monte Carlo (MC) E1+E2 reference distributions are made
The mean is extracted from each within the selection range
Reference distributions are created for special cases.
Module Borders
 edges
η edges
This sum of energies is only for the two crystals. E1 and E2. The chi sq’ed is minimized wrt to each crystal because we are taking the mean of a reference distribution and trying to equate it to the average of the data. Inputs are needed for the 17 different crystal types as well as for the special cases.
7/31/2019

18. Calibration Schema 36 SM’s 10 will have electron testbeam calibration
Use the beam inter-calibration constants to build reference distributions for the cosmic data
Use these cosmic reference distributions in place of MC
Validate method by comparing inter-calibration constants obtained with test beam and with cosmics for different SM’s.
Difference → precision
Then use cosmic ray data to obtain inter-calibration constants for remainder of SM's.
Use 1 beam calibrated SM to build the reference distributions for the other SM’s. We use the comparison of the cosmic to the beam for these other SM’s as a validation of the method. Therefore the Monte Carlo was only used as a tool to define the selection and validate that aspect. Then the reference distributions are valid on all the other SM’s
This method isn’t restricted to using a single SM as reference. Just to use it as a validation. All the beam SM’s can be combined to have an actual reference.
7/31/2019

19. Statistical Precisionvs
 ½
 ½
Each of the 3 data sets are divided in half
Odd vs. Even events
stat = odd vs. even/2 vs
 ½
 ½ vs
 ½
 ½
Single Crystal
Double Crystal
Each data set is split in half. Odd-even events. The sigma stat is quoted for the 4 different modules within a given SM. The single’s fit well to naïve expectations of peak width over root N. The double have a similar peak width, but they are larger due to the sharing of energy between more than one crystal.
7/31/2019

20. Test Beam data
ECAL setup employs full CMS geometry and electronics setup
Using electron beam with energy of 120 GeV
Preliminaries available for two supermodules
Used SM16 to build references
Validate them with SM18
Combine the datasets according to the statistical uncertainties
7/31/2019

21. Test-Beam Comparison Good to 1.5% in SM16 and 1.9% in SM18
Ignoring edges and module borders
This is for all inner-crystals. Neglecting module borders and edges of the SM.
7/31/2019

22. SM16 TB Comparison by module
This is a module by module showing of the comparison of SM16, Which was also the two crystal reference distribution.
7/31/2019

23. Future Combined results with all beam calibrated SM’s
Evaluate systematic uncertainties.
Including beam data to check for differences between muons and electrons signals
More than 23 SM’s have collected cosmic data. All are now expected to collect cosmic data.
SM installation starting soon
7/31/2019

24. Conclusion → Uniform energy response between all crystals
5 million triggers (10 days), provides a statistically accuracy of 2% or better
Testbeam comparisons show an average agreement over an entire SM of less than 2%
Cosmic ray muons will provide the most accurate inter-calibrations that will be available to all barrel crystals at CMS startup
Using the ~250 MeV cosmic signal we can predict to 2% the calibration constants found at 120 GeV.
→ Uniform energy response between all crystals
7/31/2019

25. References LY Cosmic General
L. M. Barone et al. “Correlation between Light Yield and Longitudinal Transmission in PbWO4 crystals and impact on the precision of the crystal intercalibration.” CMS RN
F. Cavallari et al. “Improvement on PbWO4 Crystal Intercalibration Precision from Light Yield Measurements at the INFN-ENEA Regional Center” CMS RN
F. Cavallari et al. “Relative Light Yield comparison between laboratory and testbeam data for CMS ECAL PbWO4 crystals” CMS RN
E. Auffray et al. “SPECIFICATIONS FOR LEAD TUNGSTATE CRYSTALS PREPRODUCTION” CMS Note
Cosmic
M. Bonesini et al. “Inter-calibration of the CMS electromagnetic calorimeter with cosmic rays before installation” CMS NOTE 2005/023
W. Bertl et al. “Feasibility of Intercalibration of CMS ECAL Supermodules with Cosmic Rays” CMS NOTE 2004/036
K. Deiters et al. “Test of the Feasibility of Pre-intercalibration of ECAL Supermodules with Cosmic Rays” CMS IN 2004/023
General
CMS Collaboration, CMS Technical Proposal, CERN/LHCC 94-38
CMS Collaboration, The Electromagnetic Calorimeter Project TDR, CERN/LHCC 97-33
Particle Data Handbook and references therein
7/31/2019

26. Extras
7/31/2019

27. ECAL: Higgs  
The reconstructed mass the the Higgs depends on the Energy of both photons as well as the angle between the two.
The error of the photon
energy is very important
Term
% expected
E for 100GeV H a
2.7%/E
270 MeV c
155 MeV in ET
MeV b
0.5%
500 MeV
7/31/2019

28. Selection MC
Selection from MC
7/31/2019