Performance of the ATLAS Tile Calorimeter · 2019. 12. 1. · To maximise the use of radial space,...

Performance of the ATLAS Tile Calorimeter

Calorimetry for High Energy Frontier (CHEF 2019)

Pawel Klimek

on behalf of the ATLAS Collaboration

Northern Illinois University

November 27, 2019

Pawel Klimek (Northern Illinois University) CHEF 2019 November 27, 2019 1 / 21

The Tile Calorimeter

Central hadronic calorimeter(|η| < 1.7) in ATLAS detector

Used to measure the 4-vectors ofthe jets and the missing transverseenergy and in the ATLAS Level-1trigger

Sampling calorimeter:steel and scintillating plastic tiles

Double photomultiplier readoutusing wave length shifting fibers

9892 PMTs

2008 JINST 3 S08003

supplies which power the readout are mounted in an external steel box, which has the cross-sectionof the support girder and which also contains the external connections for power and other servicesfor the electronics (see section 5.6.3.1). Finally, the calorimeter is equipped with three calibrationsystems: charge injection, laser and a 137Cs radioactive source. These systems test the opticaland digitised signals at various stages and are used to set the PMT gains to a uniformity of ±3%(see section 5.6.2).

5.3.1.2 Mechanical structurePhotomultiplier

Wavelength-shifting fibre

Scintillator Steel

Source

tubes

Figure 5.9: Schematic showing how the mechan-ical assembly and the optical readout of the tilecalorimeter are integrated together. The vari-ous components of the optical readout, namelythe tiles, the fibres and the photomultipliers, areshown.

The mechanical structure of the tile calorime-ter is designed as a self-supporting, segmentedstructure comprising 64 modules, each sub-tending 5.625 degrees in azimuth, for each ofthe three sections of the calorimeter [112]. Themodule sub-assembly is shown in figure 5.10.Each module contains a precision-machinedstrong-back steel girder, the edges of whichare used to establish a module-to-module gapof 1.5 mm at the inner radius. To maximisethe use of radial space, the girder provides boththe volume in which the tile calorimeter read-out electronics are contained and the flux returnfor the solenoid field. The readout fibres, suit-ably bundled, penetrate the edges of the gird-ers through machined holes, into which plas-tic rings have been precisely mounted. Theserings are matched to the position of photomul-tipliers. The fundamental element of the ab-sorber structure consists of a 5 mm thick mas-ter plate, onto which 4 mm thick spacer platesare glued in a staggered fashion to form thepockets in which the scintillator tiles are lo-cated [113]. The master plate was fabricatedby high-precision die stamping to obtain the dimensional tolerances required to meet the specifica-tion for the module-to-module gap. At the module edges, the spacer plates are aligned into recessedslots, in which the readout fibres run. Holes in the master and spacer plates allow the insertion ofstainless-steel tubes for the radioactive source calibration system.

Each module is constructed by gluing the structures described above into sub-modules on acustom stacking fixture. These are then bolted onto the girder to form modules, with care beingtaken to ensure that the azimuthal alignment meets the specifications. The calorimeter is assembledby mounting and bolting modules to each other in sequence. Shims are inserted at the inner andouter radius load-bearing surfaces to control the overall geometry and yield a nominal module-to-module azimuthal gap of 1.5 mm and a radial envelope which is generally within 5 mm of thenominal one [112, 114].

– 122 –

48 Chapter 5: The Tile Calorimeter

to be close to zero. Nevertheless, this setting is accurate to 3 ns [55]. Iter-ative OFL takes the time of the maximum sample as an initial value of thephase. In next iterations, the phase is taken to be equal to tOFL calculated inprevious iteration. The algorithm converges to the actual phase value withaccuracy better than 0.5 ns in absence of pile-up pulses. At the end of re-construction procedure, so called quality factor QFOFL is calculated in orderto verify the quality of the estimation:

QFOFL =

vuut7X

i=1

(Si � AOFL · gi � POFL)2 (5.8)

where the gi are the values of the normalised pulse shape function computedat the time of the 7 samples Si. When the deviation between the true shapeand pulse shape function used in reconstruction is large, then QFOFL alsotakes large value. Therefore, quality factor can be used to detect problemsin the reconstruction procedure as developed in Chapter 6.

Non-iterative Optimal Filtering method

In the Non-iterative Optimal Filtering method the phase is taken to be equalto the time of the maximum sample and no further iterations are performed.Due to insu�cient processing time in the DSP the OFL reconstruction mustbe performed without iterations if the trigger rate is above 50 Hz. Therefore,the non-iterative Optimal Filtering method is now used online by the RODs.

In Paper IV comparison of iterative and non-iterative Optimal Filteringphase reconstruction with presence of out-of-time pile-up is presented. Non-iterative method shows better performance. This is due to the fact thatthe phase needed to compute coe�cients is adjusted to be approximatelyzero. The out-of-time pile up can lead to reconstructed phase values far fromnominal biasing the energy reconstruction when iterative method is used.The non-iterative Optimal Filtering method reconstructs better the phase ofin-time-pulse in presence of out-of-time pile-up. This method is also morerobust against the electronic noise for very low signals.

x

y

z



QFOFL =

vuut7X

i=1






x

y

z



QFOFL =

vuut7X

i=1






x

y

z

500 1000 1500 mm0

A3 A4 A5 A6 A7 A8 A9 A10A1 A2

BC1 BC2 BC3 BC5 BC6 BC7 BC8BC4

D0 D1 D2 D3

A13 A14 A15 A16

B9

B12 B14 B15

D5 D6

D4

C10

0,7 1,0 1,1

1,3

1,4

1,5

1,6

B11 B13

A12

E4

E3

E2

E1

beam axis

0,1 0,2 0,3 0,4 0,5 0,6 0,8 0,9 1,2

2280 mm

3865 mm=0,0η

~~Pawel Klimek (Northern Illinois University) CHEF 2019 November 27, 2019 2 / 21

Calibration Systems

E [GeV] = A [ADC] · CADC→pC · CpC→GeV · CTileSize · CCs · CLas

Systems used for calibration in Tile calorimeterCharge Injection System (CIS): Calibrates the response of ADCs: CADC→pC

Cesium system: Calibrates optical components and PMT gains: CCs

Laser System: Calibrates variations due to electronics and PMTs: CLas

Minimum Bias System (MB): Calibrates optical components and PMT gains

CpC→GeV EM scale constant measured during test beam campaignsCTileSize correction addressing the different size of tiles in different layers

Cell response is not constant in time due to the PMT gain variation andscintillator degradation due to the exposure to beam

137Cs source

CalorimeterTiles

PhotomultiplierTubes

Integrator Readout(Cs & Particles)

Charge injection (CIS)

Digital Readout(Laser & Particles)

Particles

Laser light


Cesium Calibration

A moveable radioactive source 137Cs passes throughthe calorimeter body, 2-3 times per year in Run-2

The source emits γ-rays with well known energy662 keV

It uses the independent integrator readout system(τ = 10 ms) during source movement

Calibration of the complete optical chain(scintillators, fibres, PMTs) and monitoring of thedetector response over time: CCs

Between Run I and Run II: Improvement of stabilityand safety of Cesium system and procedure (newwater storage system, lower pressure, precise waterlevel metering)


Cesium CalibrationPrecision of the system in a singletypical cell is approximately 0.3%

Deviation of the cell response in timeis caused by the PMT gain variationand scintillator degradation due to theexposure to beam

Maximal drift is observed for layer A,that is closest to the collision point

It allows to adjust PMT gain (changinghigh voltage) to restore EM scale andcalorimeter response uniformity

500 1000 1500 mm0

A3 A4 A5 A6 A7 A8 A9 A10A1 A2

BC1 BC2 BC3 BC5 BC6 BC7 BC8BC4

D0 D1 D2 D3

A13 A14 A15 A16

B9

B12 B14 B15

D5 D6

D4

C10

0,7 1,0 1,1

1,3

1,4

1,5

1,6

B11 B13

A12

E4

E3

E2

E1

beam axis

0,1 0,2 0,3 0,4 0,5 0,6 0,8 0,9 1,2

2280 mm

3865 mm=0,0η

~~

Date [month and year]

Jul 15 Jan 16 Jul 16 Dec 16 Jul 17 Dec 17 Jul 18 Dec 18

De

via

tio

n f

rom

exp

ecte

d C

s r

esp

on

se

[%

]

15−

10−

5−

0

5

10Layer A

Layer BC

Layer D

ATLAS PreliminaryTile CalorimeterRun 2

η

1.5− 1− 0.5− 0 0.5 1 1.5

drift [%

]

15−

10−

5−

0

5

10Layer A

Layer B

Layer D

ATLAS PreliminaryTile CalorimeterFeb 2015 Oct 2018


Laser Calibration

PMT gain drift affects the detector responseand calibration thus it is measured regularly

A controlled amount of light with wavelengthclose to the one of physical signals (532 nmlight) creating a pulse with similar shape andduration sent into each PMT

The gain variation is measured between twoCesium scans: Laser measures the drift seenin PMTs w.r.t the last Cesium scan. Also, itallows to detect the HV changes

Performed during dedicated calibration runs.Laser pulses also sent during collision runs(empty bunches), used to calibrate timing.

1"

BEAM EXPANDER

Figure 5: Scheme of the LaserII system

the photodiode box, and a charge injection splitter (CIS), a module dispatching a chargethat aims at monitoring the stability of the electronics,

• PHOCAL is an internal calibration system setup to monitor the stability of the ten pho-todiodes located in the photodiode box. It is composed of a LED emitting a signal in alight mixer that transmit the light to a set of eleven optics fibers, with ten connected tothe photodiode box, and one coupled to a reference photodiode. The stability of the tenphotodiodes is performed by monitoring the ratio of the LED signal seen by these photodi-odes to the magnitude measured by the reference photodiode. The stability of the referencephotodiode is estimated with a static radioactive source,

• PMT box: two PhotoMultiplier Tubes (PMTs) are used to trigger the LaserII acquisitionwhen the laser is flashing. The PMT box also contains two electronic cards to drive thePMT box (LicPMT) and the filter wheel (LicMot),

• Optical filters are used to attenuate the laser signal transmitted to the photodiodes andthe PMTs,

• VME crate: two boards located in the VME crate are used to drive the LaserII system:a VME Single Board Computer (SBC); LASCAR, that incorporates on one mezzanine acharge analog-to-digital converter (qADC), as well as LILAS, and HOLA card on anothermezzanine. LASCAR also includes the TTCrx chip,

10

Mean 0.01814< RMS 0.2101

Channel gain variation [%]10< 5< 0 5 10

Cha

nnel

s / b

in1

10

210

310

410

Mean 0.01814< RMS 0.2101

ATLAS PreliminaryTile Calorimeter July-August 2015


Laser Calibration

Precision better than 0.5%

Since 2016, updates of calibrationconstants are done weekly in orderto track changes in PMT responses

The maximal drift is observed in A-and E-cells which are the cells withhighest energy deposits

Deviations of any channel responsewith respect to nominal is translatedinto a calibration constant: CLas

Between Run I and Run II: upgradedelectronics and optical components,better control of the emitted light

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1

1.2

1.3

1.4

1.6

eta

ATLAS PreliminaryTile Calorimeter

-3.24A1

-3.34A2

-3.28A3

-3.39A4

-3.27A5

-3.82A6

-3.74A7

-3.80A8

-4.05A9

-4.06A10

-1.70B1

-1.73B2

-1.66B3

-1.81B4

-1.82B5

-1.72B6

-1.89B7

-2.52B8

-2.31B9

-1.70C1

-1.73C2

-1.66C3

-1.81C4

-1.82C5

-1.72C6

-1.89C7

-2.52C8

-.34D0

-.31D1

-.26D2

-1.26D3

-3.55A12

-4.40A13

-3.99A14

-2.80A15

-1.89A16

-2.87B11

-2.70B12

-1.97B13

-1.31B14

-.94B15

-2.54C10

-2.16D4

-2.00D5

-.78D6

-4.61E1

-5.93E2

-4.62E3

-3.92E4

<-5.0% -2.5% 0% 2.5% >5.0%

run 364147, 2018-10-22.

]-1

Inte

gra

ted

Del

iver

ed L

um

ino

sity

[fb

0

10

20

30

40

50

60

70

Time [dd/mm and year]201820/03

201819/04

201819/05

201818/06

201818/07

201817/08

201816/09

201816/10

Ave

rag

e L

aser

Dri

ft [

%]

4−

3−

2−

1−

0

1

2

3Layer ALayer BCLayer DLuminosity

PreliminaryATLASTile Calorimeter

= 13 TeVs2018 Data

Start of pp collisions


Charge Injection Calibration

Calibrates the response of ADCs (electronics)digital gains and linearities

Calibration performed weekly

The system injects a signal of known chargeand measures the electronic response

Spanning the full ADC range (0-800 pC)and saturate both LG and HG for all channels

Also used to calibrate analog L1 calo trigger

Extract the conversion factors from ADCcounts to pC: CADC→pC

Precision of 0.7%, stability over time of 0.03%Pawel Klimek (Northern Illinois University) CHEF 2019 November 27, 2019 8 / 21

Minimum Bias SystemHigh energy proton-proton collisions aredominated by soft parton interactions:Minimum Bias (MB) events

The same integrator readout as in Cs systemmeasures integrated PMT signals over alarge time (∼10 ms)

As the Cesium system, the Minimum Biassystem monitors the full optical chain. Alsocalibrates E-cells and MBTS.

Measured currents are linearly dependent onthe instantaneous luminosity

Monitors the instantaneous luminosity andprovides an independent measurement givenan initial calibration (luminosity coefficient)

cellη

1.5− 1− 0.5− 0 0.5 1 1.5

Cur

rent

[nA

]

1

10

210

310

410 A cells

BC cells

D cells

E cells

ATLAS PreliminaryTile CalorimeterData 2018

-1s-2 cm3310× = 13 TeV, L = 1.7s

EB

C,m

od

22

,ch

17

cu

rre

nt

[nA

]

28

30

32

34

36

38

40

42

44

=13 TeVsData 2015,

Pol1 fit: f(x)= 13.455x+1.032

Preliminary

Tile Calorimeter

ATLAS

]-1s-2cm33

Inst. Luminosity [10

2.2 2.4 2.6 2.8 3C

urr

en

t/f(

x)

0.99

1

1.01


Combined CalibrationComparison of cell response variation betweenCesium/MB and Laser measurements

Cesium and Minimum Bias access PMT gaindrift and scintillator agingLaser only monitor PMT gain drifts

Down drifts observed during collisions. Updrifts during maintenance periods

Differences between Cesium/MB and Lasermeasurements interpreted as a scintillatoraging due to irradiation

In 2015, a good agreement observed

In 2016 - 2018 this effect was clearly observedfor some of the most irradiated cells in the A-and E-cells

Laser drift [%]6− 4− 2− 0 2 4 6

Ces

ium

drif

t [%

]

6−

4−

2−

0

2

4

6

Layer ALayer BCLayer D

ATLAS PreliminaryTile Calorimeter17/07/2015 - 03/11/2015

-1Delivered luminosity: 3.9 fb


201629/05

201628/06

201628/07

201627/08

201626/09

201626/10

A13/D

6 A

vera

ge r

esponse [%

]

10−

8−

6−

4−

2−

0

]-1

To

tal

Inte

gra

ted

De

liv

ere

d L

um

ino

sit

y [

pb

0

5000

10000

15000

20000

25000

30000

35000

Laser

Minimum Bias

Cesium Scan

ATLAS Preliminary

Tile Calorimeter

= 13 TeVs2016 Data -1

Total Delivered: 39 fbA13 | < 1.3η1.2 < |

inner layer

D6 | < 1.3η1.1 < |outer layer

]-1

Tot

al In

tegr

ated

Del

iver

ed L

umin

osity

[fb

0

5

10

15

20

25

30

35

40

45


201723/07

201722/08

201721/09

201721/10

A13

/D6

Ave

rage

res

pons

e [%

]

6−

5−

4−

3−

2−

1−

0

1

A13

D6inner layer

< 1.3η1.2 <

outer layer < 1.3η1.1 <

Minimum Bias

Laser

ATLAS Preliminary

Tile Calorimeter = 13 TeVs2017 Data

-1Total Delivered: 46.5 fb


Time Calibration

A precise time calibration is important for thecell energy reconstruction

It adjusts a digitizer sampling clock to the peakof signal produced by the particle travelingfrom the interaction point at the speed of lightthrough the cell. Large phases result inunderestimated reconstructed amplitude.

Can also be exploited in TOF measurements,e.g. in search for heavy R-hadrons

Time calibration calculated using jets andmonitored during physics data taking with laser

Resolution is better than 1 ns forEcell > 4 GeV

Cell energy [GeV]

0 20 40 60 80 100

Cell

tim

e r

esolu

tion [ns]

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0.0004±p2 1.0380

0.0003±p1 1.6252

0.0003±p0 0.2715

high-gain

0.18±p2 13.32

0.04±p1 2.62

0.003±p0 0.316

low-gain

2

E2

p+

2

E

1p

+ 2

0p = σ

PreliminaryATLAS

Tile Calorimeter

= 13 TeV, 25 nss

Cell energy [GeV]

0 20 40 60 80 100

Mean c

ell

tim

e [ns]

0.5−

0

0.5

1

1.5

PreliminaryATLAS

Tile Calorimeter

= 13 TeV, 25 nss

Cell energy [GeV]

0 20 40 60 80 100M

ean c

ell

tim

e [ns]

0.5−

0

0.5

1

1.5

PreliminaryATLAS

Tile Calorimeter

= 13 TeV, 25 nss


Noise

The total noise per cell in the calorimetercomes from two sources:

Electronic noise - measured in dedicatedruns with no signal in the detectorPile-up contribution - originates frommultiple interactions occurring at thesame bunch crossing or from the eventsfrom previous/following bunch crossings

Electronics noise stays at the level below20 MeV for most of the cells. Noise ismeasured regularly with calibration runs.

Total noise is increasing with pile-up

The largest noise values are in the regionswith the highest exposure (A-cells, E-cells)

φcell 0 1 2 3 4 5 6

elec

tron

ic n

oise

[MeV

]

16

18

20

22

24

26

28

30

PreliminaryATLASTile Calorimeter

(September 2014)

HighGain-HighGainA cellsBC cellsD cells

>µ<

0 10 20 30 40 50

Noi

se [M

eV]

20

40

60

80

100

120

140

160

180ATLAS PreliminaryTile Calorimeter

=13 TeVsData MC

LBA

A5

BC5

D2


Detector Status and Data Quality

TileCal monitoring includes identifyingand masking problematic channelscorrecting for miscalibrations,monitoring data corruption or otherhardware issues

During maintenance periods there is acampaign to fix all issues, allowing fora good recovery of the system

Redundancy of cell readout systemreduces the impact of masked channels

Tile had 99.7% DQ efficiency in Run-2

2015: 100%, 2016: 99.3%,2017: 99.4%, 2018: 100%

TileCal status at the end of Run-2

month-yearDec-10 Dec-11 Dec-12 Dec-13 Dec-14 Dec-15 Dec-16 Dec-17 Dec-18

Evo

lutio

n of

Mas

ked

Cha

nnel

s an

d C

ells

[%]

0

1

2

3

4

5

6(1.07%)Masked Channels

(0.48%)Masked Cells

Maintenance


Evolution of Masked Channels and Cells: 2018-12-03

η1.5− 1− 0.5− 0 0.5 1 1.5

φ

2−

0

2

0

0.5

1

1.5

2

2.5

3


Amount of Tile Masked Cells 2018-12-03


Single Particle Response

The ratio of the calorimeter energyat EM scale to the trackmomentum 〈E/p〉 of single hadronsis used to evaluate uniformity andlinearity during data taking

Measured in Minimum Bias events

Expect 〈E/p〉 < 1 due to thesampling non-compensatingcalorimeter

Data and Monte Carlo simulation(Pythia8) agree within 5%

Jets are further calibrated to jetenergy scale

Nor

mal

ized

Dis

trib

utio

n

6−10

5−10

4−10

3−10

2−10

1−10

1

= 0.67⟩pE/⟨

pE/0 1 2 3 4 50.5

1.01.52.0

⟩µ⟨Data 2015, Low-Pythia8 MinBias

Dat

a/M

C



-1 = 13 TeV, 1.6 nbs

⟩p

E/

⟨

0.4

0.5

0.6

0.7

0.8

0.9

1.0

[GeV]p0 5 10 15 20 25 30

0.91.01.1


Dat

a/M

C



-1 = 13 TeV, 1.6 nbs


Muons

Muons from cosmic rays are used tostudy in situ the electromagneticenergy scale and intercalibration ofTile cells

A good energy response uniformitybetween calorimeter cells

< 5% response non-uniformity in ηwith cosmic muons

[rad]φTrack

1.9− 1.8− 1.7− 1.6− 1.5− 1.4− 1.3− 1.2−

Cel

l Res

pons

e [M

eV]

0

200

400

600

800

1000

1200

1400

1600 Tile Calorimeter

Cosmic data 2015

ATLAS Preliminary

η

1.5− 1− 0.5− 0 0.5 1 1.5

>dxdE

/<dxdE

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6Tile Calorimeter

Cosmic data 2015

All layersATLAS Preliminary


Jet Performance

A good description of the cell energydistribution and of the noise in thecalorimeter is crucial for the buildingof topoclusters which are used for jetand missing transverse energyreconstruction

Good agreement in Tile cell energydistribution

Aim for jet energy resolution:σE

E = 0.5√E⊕ 0.03

Jet energy resolution is better than 10%at pT > 100 GeV

Constant term is within expected 3%

Jetperformance

•  GoodagreementinTilecellenergydistribuGon•  Consistentoveralljetenergyscale•  JetenergyresoluGonisbelow10%atpT>100GeV•  Constanttermiswithinexpected3%02.07.17 O.Solovyanov-QFTHEP17 15

Tile Cell Energy [GeV]

Cel

ls /

0.06

GeV

/ Ev

ent

5−10

4−10

3−10

2−10

1−101

10210

310

410

510

Tile Cell Energy [GeV]1− 0 1 2 3 4 5

Dat

a / M

C

0.00.51.01.5

ATLAS Preliminary = 0.9 TeVsData,

= 13 TeVsData,

Random Filled Bunch Crossing

Random Empty Bunch Crossing

= 13 TeVsMinimum Bias MC,

[GeV]jet

Tp

20 30 210 210×2 310 310×2

Tp

) /

Tp(

σJe

t ene

rgy

reso

lutio

n,

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

in situDijet

Total uncertaintyMC prediction

= 0.4, EM+JESR tkanti-| < 0.7η |≤0.2

ATLAS Preliminary-1 = 13 TeV, 43.6 fbs


Conclusions

Tile Calorimeter is an important part of ATLAS detector at LHC

It is a key detector to measure the 4-vectors of the jets and missing energy

A set of calibration systems is used to calibrate and monitor the calorimeterresponse

Intercalibration and uniformity are monitored with isolated charged hadronsand cosmic muons

The stability of the absolute energy scale at the cell level was maintainedto be better than 1% during Run 2 data taking

In Run-3: New crack scintillators (E3-4), covering a larger |η| < 1.72 range,will result in a significant improvement in the energy resolution for electrons,photons and jets in this region


The Di-jet Event Produced in 2017Two jest with pT = 2.9 TeV and mjj = 9.3 TeV


Back-up

Back-up


Back-up

Calibration System


Performance of the ATLAS Tile Calorimeter · 2019. 12. 1. · To maximise the use of radial space,...

Documents

Transcript of Performance of the ATLAS Tile Calorimeter · 2019. 12. 1. · To maximise the use of radial space,...