Summary of two selected HCPSS 2012 courses R. M arki · Example: CMS Lead-Tungstate Calorimeter - response to high energy electrons 8/47 R. M arki Calorimetry and particle identi

Calorimetry and particle identificationSummary of two selected HCPSS 2012 courses

R. Märki

EPFL - LPHE

8 October 2012

1/47 R. Märki Calorimetry and particle identification

Outline

I attended the 7th Annual Fermilab-CERN HCPSS in August 2012and I summarize two selected topics in this presentation.

Calorimetry

Advantages

Calorimeter types

Calibration

Examples

Particle identification

Principle

Strategy and complete example of CMS

More techniques

Efficiency and purity


Calorimetry advantages

Who do calorimetry ?!

Pros:

Measure neutrals as well as charged particles

Resolution improves with particle energy

If hermetic, can be used to measure missing particles (eg.neutrinos)

Fast trigger

Cons:

Non-linear response

Must be BIG (hence expensive)

Needs non-trivial engineering for design, construction andsignal extraction


Calorimetry advantages

Combined with tracking, the energy resolution is highlyimproved

Example of CMS:


Material interactions

Particles interact in matter and deposit energy

Bethe-bloch for charged particles

Mean free path (or radiation length) important for calorimeterdesign


Shower developpement

Example here: electrons

Longitudinal shower: several radiation lengths are needed tocompletely stop a particle.

Lateral shower: the so called Moliere radiusRM ∼ X0(21MeV /Ec) contains 90% of the electromagneticcascade, though there are long tails.

longitudinal lateral


Sampling calorimeters

Only part of the deposited energy dE/dx is measured

The sampling fraction is defined as∑(dE/dx)active medium/

∑(dE/dx)absorber

The energy measurement is linear for an infinite detectorEparticle = k×(dE/dx)absorber/(dE/dx)active medium×

P(dE/dx)active medium

Number of particles in the shower is statistical but scales like:Nshower ∼ Eparticle/Ecritical

Energy deposition in the shower is a statistical processσE ∼ 1/

√Nshower → σE ∼ 1/

√Eparticle


Homogeneous calorimeters

Commonly

Scintillator (solid and liquid)Liquid Argon

Less commonly

Gas proportional tubesSilicon

Example: CMS Lead-Tungstate Calorimeter - response to highenergy electrons


Electromagnetic calorimeters

Both sampling and homogeneous are frequently used

Need fine granularity to distinguish photons from π0 forinstance, discussed more in PID part

Usually preshower with even finer granularity over 1-2 firstradiation lengths

In a sampling electromagnetic calorimeter, the samplingfraction changes in the shower


Hadronic calorimeters

Hadronic calorimeters are usually sampling calorimeters

Hadron showers have a complex composition:

EM energy (eg π0 → γγ) : O(50%)Visible non-EM energy (eg dE/dX) : O(25%)Invisible non-EM energy (eg nuclear breakup) :O(25%)Escaped energy (eg ν) :O(1%)

Therefore the simulation is complicated as well


Example of calorimeters - CDF

The sampling calorimeter of CDF

Lead absorber with sheets of plastic scintillator

The left picture was taken by myself this summer !


Example of calorimeters - D0

Homogeneous calorimeter at D0

Uranium metal bathed in liquefied argon


Example of calorimeters - ATLAS HCAL

ATLAS HCAL:

Sampling calorimeter

Absorber: steel

Scintillating tiles

8m diameter

12m long


Example of calorimeters - ATLAS Argon

ATLAS ECAL (and HCAL in theforward region):


Absorber: lead and stainless steel

Liquid argon to sample

Accordeon shaped electrodes


Example of calorimeters - CMS HCAL

CMS HCAL:


Absorber + plastic scintillator(scintillator plates 2m long)


Example of calorimeters - CMS ECAL

Very famous and compactCMS ECAL

Homogeneous calorimeter

Lead tungstate (PbWO4)crystal tiles

Before: pre-showerlead / silicon strips


Calorimeter calibration - ATLAS

Optical chain calibration:(in real time)

tiles with source (Cs137)

PMT with laser

readout electronics with test pulse

Aging effects can be measured and taken into account:


Calorimeter calibration - CMS

Very first calibration in test beam

ECAL calibrated with electrons (and photons)

HCAL calibrated with π0 (normal incidence, no working ECAL in front)→ need for correction

The energy calibration is parametrized withE = a + b(p, η)ECAL + c(p, η)HCAL

a, b and c are determined with “isolated” tracks in minimum bias events


Detector aging - CMS EM crystals

The aging can be monitored using the calibration methods seenbefore

Monitor using laser calibration system Response using E/p in W → eν



Already in calorimeters there are different shower responses forelectrons and hadrons



General detector response depending on particle


Particle identification - example in CMS



During reconstruction the event is “cleaned” up:

1 Find and “remove” muons (σtrack)

2 Find and “remove” electrons ( min[σtrack , σECAL] )

3 Find and “remove” charged hadrons (σtrack)

4 Find and “remove” converted photons ( min[σtrack , σECAL] )

5 Find and “remove” V0’s (σtrack)

6 Find and “remove” photons (σECAL)

7 Left with neutral hadrons (10%) (σHCAL + fake)



Link tracks to ECAL and HCAL

25 ECAL cells underneath each HCAL cell



Top view helps to see the links

The captions correspond to what was simulated

What is actually reconstructed: γ, γ, γ, π+, π−


Particle identification - example in ATLAS


Particle identification - when E � p

Few energy in calorimeter compared to measured momentumMainly due to muonsMuon ID very efficient, 98% in CMSThe resting 2% still contribute significantlyLooser muon cuts used but still many cases leftTrue origin: fake tracks and interactions in tracker material



Tracker acts like a pre-shower (silicon is heavy)

Has up to 2 radiation lengths for certain pseudo-rapidities

CMS ATLAS



Reduce hits progressively

Start from very pure track seeding

Remove used hits and start over with looser requirements

For charged hadrons: from 85% efficiency, 20% fake rate to93% efficiency, 1-2% fake rate


Particle identification - Time of Flight

Measure time difference between two detector plane crossingsDifferent times for different particles of same momentumβ = d/c∆t and p = γmcβ → ∆t = dp/γmDifference very small for relativistic particlesExample for a 12m distance:10 GeV/c K → 40.05 ns10 GeV/c π → 40.00 nsOne needs to measure 50 ps difference for a 12m distancewhich is already a large scale


Particle identification - Ionization

Particles lose energy according to the Bethe-Bloch formula

Energy loss depends on momentum and mass

If energy or momentum loss is measured, one candiscrimanate between particles

Hard to distinguish π and µ though, as their masses are veryclose


Particle identification - Transition radiation

A charged particle flying into a medium with different n (ordifferent dielectric constant, as n =

√ε) will have its relative

velocity with respect to c ′ changed

This change results in the emission of transition radiation(photons)

Emitted energy proportional to the boost (γ) of the particle

Hence also quite good for high energy particles


Particle identification - Cherenkov radiation

A charged particle having a velocity higher than c ′ emitsCherenkov radiation

The light is emitted in a cone, like the waves from amotorboat

The angle of the cone is proportional to the velocitycos(θc) = 1/βn

θC

(mra

d)

250

200

150

100

50

0

1 10 100

Momentum (GeV/c)

Aerogel

C4F10 gas

CF4 gas

eµ

p

K

π

242 mrad

53 mrad

32 mrad

θC max

Kπ


Particle identification - Electron ID

Electrons radiate around 70% of the energy in the track bybremsstrahlung

Photons have > 50% chance to convert into e+e− pair

Hence, energy spreads in ϕ (⊥ to B)

Standart Kalman Filterpattern recognition gives upquickly

Need to account forBethe-Heitler energy loss(bremsstrahlung)

Use sum of Kalman Filters(Gaussian Sum Filter) toapproximate non-gaussianpart



Very important to identify/recover bremsstrahlung photons

Otherwise their energy is counted twice:(in track + again in ECAL)

Holds also for electron pair converted bremsstrahlung photons

How it is done in CMS:

Check if tangent of trackpoints to ECAL cluster

Link cluster to track

Also test compatibilitybetween ECAL cluster E and∆p along GSF track



BUT ! Everything that we have just seen can be used todiscrimanate between e and π for instance

π radiate much less, so one can:

count the number of hits linked to the tracklook at ∆pcount the number of bremsstrahlung γ associated to the tracklook at Ebremsstrahlunglook at shower shape along ϕ and ηlook at linked HCAL energy

Everything put into a MVA gives 95% efficiency for isolatedelectrons and 70-80% efficiency in jets

KS → π+π−


Particle identification - Photon ID

If not converted, the only way to measure photons arecalorimeters

When looking at unconverted photons, everything known hasalready been “removed” from the event. Then:

Use fine segmentation to look at shower shapeUse isolation criteria

Clustering algorithm plays a big role (be sure that all energy islinked to the cluster !)


Particle identification - Efficiency and purity

Assume ID is uncorrelated with isolation

The true number of photons among NA is equal toN(γ) = NA − backgroundThis background is NB ×MA/MBHence the purity is P = 1− NB/NA ×MA/MB

Efficiency can also be simulated

Or use tag and probe method


Related physics results

The Higgs in H → γγ !


Thank you for your attention

Backup slides


CMS event

Massive Pile-up at CMS


Material interactions in tracking system at CMS

Interaction vertices in the CMS tracker


LHCb RICH PID

LHCb φ→ K+K− result from 2009 (√

s = 900GeV ) withoutand with RICH PID information


Tag and probe muons at CMS

Fit J/ψ mass for dimuons which pass or do not pass PID cut

Evaluate efficiency and purity


Documents

Summary of two selected HCPSS 2012 courses R. M arki · Example: CMS Lead-Tungstate Calorimeter - response to high energy electrons 8/47 R. M arki Calorimetry and particle identi