How to design, build and operate a huge experiment? 1

How to design, build and
operate a huge experiment?
9th
D. Froidevaux, CERN
Summer Student Lectures, CERN, 2006
and 11th August 2006 (D. Froidevaux, CERN)
1
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the challenge
Muon chambers
ν
µ
Hadronic calorimeter
γ
e
Electromagnetic calorimeter
n
π
Inner detector
D. Froidevaux, CERN
2
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the challenge
One word about neutrinos in hadron colliders:
• since most of the energy of the colliding protons
escapes down the beam pipe, one can only use
the energy-momentum balance in the transverse
plane
→ concepts such as ETmiss, missing transverse
momentum and mass are used everywhere
• the detector must therefore be quite hermetic
→ no neutrino escapes undetected
→ no human enters without major effort
(fast access to some parts of ATLAS difficult)
D. Froidevaux, CERN
3
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Main specific design choices of ATLAS/CMS
• Size of ATLAS/CMS directly related to energies of particles produced: need to
absorb energy of 1 TeV electrons (30 X0 or 18 cm of Pb), of 1 TeV pions
(11 λ or 2 m Fe) and to measure momenta of 1 TeV muons outside calorimeters
(BL2 is key factor to optimise)
• Choice of magnet system has shaped the experiments in a major way
• Magnet required to measure momenta and directions of charged particles
near vertex (solenoid provides bend in plane transverse to beams)
• Magnet also required to measure muon momenta (muons are the only
charged particles not absorbed in calorimeter absorbers)
• ATLAS choice: separate magnet systems (“small” 2 T solenoid for tracker
and huge toroids with large BL for muon spectrometer)
• Pros: large acceptance in polar angle for muons and excellent muon
momentum resolution without using inner tracker
• Cons: very expensive and large-scale toroid magnet system
• CMS choice: one large 4 T solenoid with instrumented return yoke
• Pros: excellent momentum resolution using inner tracker and more compact
experiment
• Cons: limited performance for stand-alone muon measurements (and
trigger) and limited space for calorimeter inside coil
D. Froidevaux, CERN
4
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Main specific design choices of ATLAS/CMS
• At the LHC, which is essentially a gluon-gluon collider, the unambiguous
identification and precise measurement of leptons is the key to many areas of
physics:
• electrons are relatively easy to measure precisely in EM calorimeters but
very hard to identify (imagine jet → leading π- with π- → leading π0 very early
in shower)
• muons in contrast are relatively easy to identify behind calorimeters but
very hard to measure accurately at high energies
→ This has also shaped to a large extent the global design and technology choices
of the two experiments
• EM calorimetry of ATLAS and CMS is based on very different technologies
• ATLAS uses LAr sampling calorimeter with good energy resolution and
excellent lateral and longitudinal segmentation (e/γ identification)
• CMS use PbWO4 scintillating crystals with excellent energy resolution and
lateral segmentation but no longitudinal segmentation
• Broadly speaking, signals from H → γγ or H → ZZ* → 4e should appear as
narrow peaks (intrinsically much narrower in CMS) above essentially pure
background from same final state (intrinsically background from fakes
smaller in ATLAS)
D. Froidevaux, CERN
5
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
ATLAS/CMS: from design to reality
R&D and construction for 15 years → excellent EM calo intrinsic performance
• Standalone performance measured in beams with electrons from 10 to 250 GeV
D. Froidevaux, CERN
6
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
ATLAS/CMS: from design to reality
Amount of material in ATLAS and CMS inner trackers
LEP
detectors
• Material increased by ~ factor 2 from 1994 (approval) to now (end constr.)
• Electrons lose between 25% and 70% of their energy before reaching EM calo
• Between 20% and 65% of photons convert into e+e- pair before EM calo
• Need to bring 70 kW power into tracker and to remove similar amount of heat
D. Froidevaux, CERN
7
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
ATLAS/CMS: from design to reality
Actual performance expected in real detector quite different
Photons at 100 GeV
ATLAS: 1-1.3%
energy resol. (all γ)
CMS: 0.8% energy
resol. (εγ ~ 70%)
Electrons at 50 GeV
ATLAS: 1.3-1.8%
energy resol.
(use EM calo only)
CMS: ~ 2.0%
energy resol.
(combine EM calo
and tracker)
D. Froidevaux, CERN
8
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the machine
What drives the luminosity at the LHC?
L(α=0) = 1.07 10-4 1/Δt N2 E / βe ε, where:
α is the crossing angle between the beams
Δt is the time between bunch crossings, Δt = 25 ns
N is the number of protons per bunch, N = 1011
E is the energy per beam, E = 7 TeV
βe is the β-function at the interaction point, βe = 0.5 m
ε is the normalised emittance, ε = 15π 10−6 m.rad
D. Froidevaux, CERN
9
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Extract number of inelastic collisions per bunch crossing
<n> = σinel x
L
x Δt / εbunch
LHC: <n> = 70 mb x 1034 cm-2s-1 x 25 ns / 0.8 = 23
Big change compared to recent and current machines:
LEP:
SppS:
HERA:
Tevatron:
D. Froidevaux, CERN
Δt = 22 μs
Δt = 3.3 μs
Δt = 96 ns
Δt = 0.4 μs
and
and
and
and
10
<n> << 1
<n> ≈ 3
<n> << 1
<n> ≈ 2
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Experimental environment ≡ Machine performance x Physics
Event rates in detectors:
number of charged tracks expected in inner tracking detectors
energy expected to be deposited in calorimeters
radiation doses expected (ionising and neutrons)
event pile-up issues (pile-up in time and in space)
Need to know the cross-section for uninteresting pp inelastic events:
simple trigger on these ≡ “minimum bias” trigger
D. Froidevaux, CERN
11
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
p
pT
θ
dσ/dpTdy is
Lorentz-invariant
η = y for m ≈ 0
Physics is ~ constant
versus η at fixed pT
D. Froidevaux, CERN
12
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Charged particle multiplicities
from different models
Charged particle multiplicity and energy
in pp inelastic events at √s = 14 TeV
<pT> ~ 500 MeV
Present models extrapolated from
Tevatron give sizeable differences
at the LHC
D. Froidevaux, CERN
13
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
What do we expect roughly speaking at L = 1034 cm-2s-1 ?
dncharged/dη ≈ 7.5 per Δη = 1
ncharged consists mostly of π+- with <pT> ≈ 0.6 GeV
dnneutral/dη ≈ 7.5, nneutral consists mostly of γ
from π0 decay with <nπ0> ≈ 4 and <pTγ> ≈ 0.3 GeV
Assume detector with coverage over –3 < η < 3 (θ = 5.7o)
for tracks and –5 < η < 5 (θ = 0.8o) for calorimetry:
• Most of the energy is not seen! (300000 GeV down the beam pipe)
• ~ 900 charged tracks every 25 ns through inner tracking
much
•… still
~ 1400
GeV more
transverse energy (~ 3000 particles) in calorimeters
complex
every 25 ns
than a LEP event
D. Froidevaux, CERN
14
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Radiation resistance of detectors
New aspect of detector R&D (from 1989 onwards)
→ for once make use of military applications!
The ionising radiation doses and the slow neutron fluences are
almost entirely due to the beam-beam interactions and can
therefore be predicted
→ was not and is not the case in recent and current machines
Use complex computer code developed over the past 30 years or
more for nuclear applications (in particular for reactors)
ATLAS neutron fluences
D. Froidevaux, CERN
15
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
1. Damage caused by ionising radiation
caused by the energy deposited by particles in the detector
material: ≈ 2 MeV g-1 cm-2 for a min. ion. particle
also caused by photons created in electromagnetic showers
the damage is proportional to the deposited energy or dose
measured in Gy (Gray):
• 1 Gy = 1 Joule / kg = 100 rads
• 1 Gy = 3 109 particles per cm2 of material with unit density
At LHC design luminosity, the ionising dose is:
≈ 2 106 Gy / rT2 / year,
where rT (cm) is the transverse distance to the beam
D. Froidevaux, CERN
16
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
2. Damage caused by neutrons
•
•
•
•
the neutrons are created in hadronic showers in the
calorimeters and even more so in the forward shielding of the
detectors and in the beam collimators themselves
these neutrons (with energies in the 0.1 to 20 MeV range)
bounce back and forth (like gas molecules) on the various
nuclei and fill up the whole detector
expected neutron fluence is about 3 1013 per cm2 per year in
the innermost part of the detectors (inner tracking systems)
these fluences are moderated by the presence of Hydrogen:
♦ σ(n,H) ~ 2 barns with elastic collisions
♦ mean free path of neutrons is ~ 5 cm in this energy range
♦ at each collision, neutron loses 50% of its energy
(this number would be e.g. only 2% for iron)
D. Froidevaux, CERN
17
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
•
the neutrons wreak havoc in semiconductors, independently
of the deposited energy, because they modify directly the
cristalline structure
→ need radiation-hard electronics (military applications only
in the early R&D days)
ƒ off-the-shelf electronics usually dies out for doses
above 100 Gy and fluences above 1013 neutrons/cm2
ƒ rad-hard electronics (especially deep-submicron) can
survive up to 105-106 Gy and 1015 neutrons/cm2
•
most organic materials survive easily to 105-106 Gy (beware!)
Material validation and quality control during production are
needed at the same level as for spatial applications!!
D. Froidevaux, CERN
18
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Pile-up effects at high luminosity
Pile-up is the name given to the impact of the 23 uninteresting
(usually) interactions occurring in the same bunch crossing as the
hard-scattering process which generally triggers the apparatus
Minimising the impact of pile-up on the detector performance has
been one of the driving requirements on the initial detector design:
• a precise (and if possible fast) detector response minimises
pile-up in time
→ very challenging for the electronics in particular
→ typical response times achieved are 20-50 ns (!)
• a highly granular detector minimises pile-up in space
→ large number of channels (100 million pixels, 200,000 cells
in electromagnetic calorimeter)
D. Froidevaux, CERN
19
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Pile-up effects at high luminosity
ATLAS
Photon converts at R = 40 cm
and electron pair is visible in
ATLAS TRT and EM calo
D. Froidevaux, CERN
20
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Pile-up effects at high luminosity
ATLAS barrel
H º ZZ* º eeμμ (mH = 130 GeV)
D. Froidevaux, CERN
21
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the environment
Interactions every 25 ns …
In 25 ns particles travel 7.5 m
Time-of-flight
‹
22
m
Weig
ht:
7000 t
44 m
Cable length ~100 meters …
In 25 ns signals travel 5 m
D. Froidevaux, CERN
‹
22
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Physics at the LHC: the challenge
Unprecedented scope and timescales for simulations
Many examples from ATLAS/CMS and also ALICE/LHCb:
• ≈ 30 million volumes simulated in GEANT
• Tbytes of simulated events over 10 years
• Full reconstruction of all benchmark Higgs-boson decays
• Unprecedented amount of material in Inner Detector leads
to significant losses of e.g. charged-pion tracks
(up to 20% at 1 GeV)
• b-tagging at low luminosity, for e.g. H → bb searches, yields
performance similar to that which has been achieved at LEP
D. Froidevaux, CERN
23
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Detector optimisation example: H→ γγ
Signal reconstruction
One wants to reconstruct:
Mγγ2 = 2 Eγ1Eγ2(1-cosθ12)
What contributes to resolution on mγγ?
1) Measurement of Eγ:
• Intrinsic resolution of calo
• Calibration/uniformity of calo
• Pile-up effects
2) Measurement of θ12
• Measurement of position and
direction of em showers
• Meas. of interaction vertex zv
D. Froidevaux, CERN
24
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Detector optimisation example: H→ γγ
Angular resolution and acceptance
• ATLAS calorimeter has
longitudinal segmentation
→ can measure γ direction
ATLAS, full simulation
Vertex resolution using EM
calo longitudinal segmentation
Photons from
H → γγ
θ
vertex spread
~ 5.6 cm
50 mrad
σ (θ ) ≈
E
CMS has no longitudinal segmentation (and no preshower in barrel)
→ vertex measured using secondary tracks from underlying event
→ often pick up the wrong vertex
→ smaller acceptance in the Higgs mass 25
window
D. Froidevaux, CERN
z
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Detector optimisation example: H→ γγ
D. Froidevaux, CERN
26
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Detector optimisation example: H→ γγ
Even assuming a negligible reducible
background, the signal-to-noise for
this channel is very low, around 1%.
The signal cannot be seen with the eye
alone
(see plot for crystal-calo resolution).
In this case, however, enough handles
to make a reliable statistical analysis:
1) sidebands around mH in the
mγγ distribution are smooth
and have high statistics
2) use nearby Z-boson resonance
to understand in detail
performance of em calo
D. Froidevaux, CERN
27
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Detector optimisation example: H→ γγ
~ 1000 events
in the peak
ATLAS
100 fb-1
CMS
100 fb-1
K=1.6
D. Froidevaux, CERN
28
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Oct. 2005
Barrel toroid: cool down ongoing, first tests of full field in Sept.
End-cap toroids: will be installed in the pit end 2006-beg 2007
D. Froidevaux, CERN
29
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Barrel calorimeter (EM liquid-argon + HAD Fe/scintillator Tilecal)
in final position at Z=0. Barrel cryostat cold and filled with Ar.
D. Froidevaux, CERN
30
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
TRT
In February, barrel Si
detector (SCT) was
inserted into barrel TRT
→ ready for installation in
the pit in August 2006
SCT
Three completed Pixel disks
(one end-cap) with 6.6 M channels
Barrel pixel detector on critical path (problems
with low-mass cables), but still scheduled for
installation in the pit in April 2007
D. Froidevaux, CERN
31
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Cosmics DATA taken in barrel SCT+TRT : ~ 450k events
ATLAS
preliminary
QuickTime™ and a
TIFF (LZW) decompressor
are needed to see this picture.
SCT efficiency
per layer
W. Hulsbergen
Δϕ (TRT track-SCT track)
QuickTime™ and a
TIFF (LZW) decompressor
are needed to see this picture.
Cabling error in
two modules
D. Froidevaux, CERN
32
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Muon Spectrometer : measurement chambers MDT, CSC (innermost forward)
trigger chambers RPC (barrel), TGC (end-caps)
~50% of barrel stations installed
(mostly complete end of Summer ‘06)
First sectors of TGC end-cap
“big-wheels” installed
D. Froidevaux, CERN
33
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
First cosmics have been registered in the underground cavern with
barrel Muon chambers (MDT and RPC) and Level-1 μ trigger
D. Froidevaux, CERN
34
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Towards Physics (1) : the 2004 ATLAS combined test beam
Full “vertical slice” of ATLAS tested on CERN H8 beam line May-November 2004
Geant4 simulation
of test-beam set-up
O(1%) of ATLAS
coverage
~ 90 million events collected
(~4.5 TB of data):
e±, π ±
1 → 250 GeV
±
±
μ , π , p up to 350 GeV
γ
20-100 GeV
B-field = 0 → 1.4 T
y
x
z
D. Froidevaux, CERN
All ATLAS sub-detectors (and LVL1 trigger)
integrated and run together with common
DAQ and monitoring, “final” electronics,
slow-control, etc. Data analyzed with common
ATLAS software. Gained lot of global
operation experience during ~ 6 month run.
35
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
At the surface, solenoid inserted on 14 Sept. 2005;
cooled down to 4.5 K in February 2006;
ramping up the current, now at 12.5 kA (2.5 T)
→ magnetic test/field map starting Aug./Sept. 2006 (MTCC)
Magnetic length
Diameter
Magnetic field
Nominal current
Stored energy
D. Froidevaux, CERN
36
12.5 m
6m
4T
20 kA
2.7 GJ
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Installation of Muon chambers : ~ 90% complete in end-cap (CSC+RPC),
~ 60% in barrel (DT+RPC)
Surface Hall
SX5
50% of RPCs installed
on YE disks.
ME2
RE2
YB+2
YE+2
YE+1
YE+2
YE+1
YB0
YB+1
Shaft
ME3
> 90% CSCs installed on
YE disks.
D. Froidevaux, CERN
ME1
RE1 37
3 out of
5 YB
wheels
done
(DTs,
RPCs)
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Cosmic muon bending in the CMS return yoke, collected yesterday
with 12.5kA in the solenoid (B=2.5T)
NEW
!
D. Froidevaux, CERN
38
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Electromagnetic calorimeter
Barrel : 36 SuperModules (SM), 1700 crystals each
Total of ~ 61000 barrel crystals (85% delivered)
27 bare SM assembled, 22 equipped with electronics
2 barrel SM installed
inside HCAL for MTCC
D. Froidevaux, CERN
39
Crystal delivery determines ECAL
schedule: last barrel (end-cap) crystal
delivered in Feb. 2007 (Jan. 2008).
Plan is to have barrel completed for
commissioning run in 2007 and end-caps
installed for 2008 physics run
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Inner tracker:
~ 220 m2 of Si sensors
10.6 million Si strips
65.9 million Pixels
Pixels
Outer Barrel TOB
End-cap TEC
Inner Disks TID
• Assembly of all 16000 modules completed
• Integration progressing well
• Installation at Point 5 in April 2007
Inner Barrel TIB
TIB layer 4
One TEC (9 disks) completed
D. Froidevaux, CERN
40
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Towards Physics (2) : the CMS Magnet Test and Cosmic Challenge (MTCC)
Cosmics run of a ~full detector slice (few percent of CMS coverage) inside 4T field.
Magnet being energized, detector closed, data taking started …
Test: detector installation and closing; magnet commissioning and field map;
combined operation of full chain detector-electronics-DAQ-trigger-DCS-software
identical to final experiment; timing, calibration, alignment procedures
ECAL
HCAL
Magnet
Tracker
Closing CMS …
Muon chambers
D. Froidevaux, CERN
41
High muon rate: 0.5-1 kHz
(run at surface)
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
What data samples in 2007 ?
ATLAS preliminary
30% data taking
efficiency included
(machine plus detector)
Trigger and analysis
efficiencies included
√s =900 GeV, L = 1029 cm-2 s-1
Jets pT > 15 GeV
(b-jets: ~1.5%)
Jets pT > 50 GeV
Jets pT > 70 GeV
Υ→ μμ
J/ψ→μμ
W → eν, μν
Z → ee, μμ
30 nb-1
100 nb-1
+ 1 million minimum-bias/day
Start to commission triggers and detectors with collision data (minimum bias, jets, ..)
in real LHC environment
J Maybe first physics measurements (minimum-bias, underlying event, QCD jets, …) ?
J Observe a few W→ lν, Υ → μμ, J/ψ → μμ ?
J
D. Froidevaux, CERN
42
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006
Example of initial measurement: understanding detector and physics with top events
Can we observe an early top signal with limited detector performance ?
And use it to understand detector and physics ?
ATLAS preliminary
σtt ≈ 250 pb for tt → bW bW → blν bjj
W.Verkerke
50 pb-1
4 jets pT> 40 GeV
3 jets with largest ∑ pT
2 jets M(jj) ~ M(W)
W+n jets (Alpgen) +
combinatorial background
Isolated lepton
pT> 20 GeV
NO b-tag !!
ETmiss > 20 GeV
Top signal observable in early days with no b-tagging and simple analysis
(100 ± 20 evts for 50 pb-1) → measure σtt to 20%, m to 10 GeV with ~100 pb-1 ?
In addition, excellent sample to:
• commission b-tagging, set jet E-scale using W → jj peak
• understand detector performance for e, μ, jets, b-jets, missing ET, …
• understand / constrain theory and MC 43
generators using e.g. pT spectra
D. Froidevaux, CERN
CERN Summer Student Lectures, 09/08/2006 and 11/08/2006