CMS Inner Tracking

1
CMS Inner Tracking
2
High Luminosity Physics at the LHC
Golden Channel
pp High luminosity gt mess
Tracker Requirements

• Efficient robust Pattern Recognition algorithm
• Fine granularity to resolve nearby tracks
• Fast response time to resolve bunch crossings
• Ability to reconstruct narrow heavy object
• 12 Pt resolution at 100GeV
• Ability to tag b/t through secondary vertex
• Good impact parameter resolution
• This Review HLT, including b t jet tags

3
The CMS Pixel Vertex detectorSilicon strips
have become pixels
The region below 20cm is instrumented with
Silicon Pixel Vertex systems
The Pixel area is driven by FE chip The shape is
optimized for resolution CMS pixel 150
150 mm2
4 107 pixels Shaping time 25ns
With this cell size, and exploiting the large
Lorentz angle We obtain IPtrans. resolution 20
mm for tracks with Pt 10GeV
With this cell size occupancy is 10-4 This
makes Pixel seeding the fastest Starting point
for track reconstruction Despite the extremely
high track density
4
From strip Vertex to strip Tracking

• Single-sided, AC coupled, polysilicon biased
  sensors have become a mature technology
• Costs have decreased, and large scale production
  is now possible
• High level of expertise for FE IC design and
  system aspects of O(105) channels
• Move to detectors with a high level of
  independent tracking capability
• A few m2 CDF D0
• Several 101m2 ATLAS
• A couple 102 m2 CMS

5
The CMS Tracker 220m2 of silicon strip sensors
6
SST Module level Components
9648128 strips ? channels 75376 APV
chips 6136 Thin sensors 18192 Thick
sensors 440 m2 of silicon wafers 210 m2 of
silicon sensors 3112 21512 Thin
modules 5496 21800 Thick modules ss
dsb-to-b 17000 modules 25000000 Bonds
Silicon sensors
CF frame
Pitch adapter
FE hybrid with FE ASICS
7
The radiation hard P-on-N strip detector
Single-Sided Lithographic Processing ( AC,
Poly-Si biasing )
Radiation hardness recipe
P-on-N sensors work after bulk type
inversion, Provided they are biased well above
depletion Match sensor resistivity thickness
to fluence To optimize S/N over the full
life-time Follow simple design rules for guard
strip geometry Use Al layer as field plate to
remove high field _at_edges from Si bulk to Oxide
(much higher Vbreak) Strip width/pitch 0.25
reduce Ctot maintain Stable high bias voltage
operation Take care with process especially
implants Surface radiation damage can increase
strip capacitance noise Use lt100gt crystal
instead of lt111gt
8
Silicon Strip Sensor Properties
Strip capacitance 1.2pF/cm for w/p
0.25 Independent of pitch and thickness
Insensitive to irradiation for lt100gt crystal
lattice
Expected S/N after irradiation S/N 13 for thin
sensors, short strips S/N 15 for thick sensors,
long strips
Use 320mm thick Si for R lt 60cm, Strip
10cm Use 500mm thick Si for R gt 60cm, Strip
20cm
9
The CMS Silicon Strip Tracker from 4 to 6
The CMS SST exploits 6 technology Useful
surface/wafer 2.5 that of 4 wafers Large
scale high quality sensor production in
modern Industrial lines available from more than
one vendor Hamamatsu produces the thin 320m
sensors ST-Microelectronics the thick 500m
sensors
Production is well underway Hamamatsu
excellent quality, some concerns
regarding Selection of sufficiently low
resistivity raw material ST production quality
also good, problems due To inappropriate
manipulation of sensors during Testing are being
addressed
10
Front-End Hybrids
This has proven to be a major challenge, and has
defined the Critical path for module assembly. We
have finally converged on

• 4 layer Kapton flex circuit
• high resolution multi-layer Kapton circuits now
  available industrially in large volumes
• seamless integration of flexible pig-tail
• Laminated onto a ceramic substrate
• need rigidity for bonding chips to hybrid
• ceramic chosen since adequate thermally and
  mechanically (flatness!) and cheap
• provides support for pitch adapter
• can fully bond the hybrid to pitch-adapter
  assembly before gluing on module frame
• Pre-series production end of 2002 gt Design
  fine-tuned for efficient lamination and component
• assembly and wire-bonding. Hybrid production now
  ramping up, module assembly following

11
The Gantry in actionAssembly of 3 TOB Modules
12
Placement accuracy and reproducibilitywith
automatic pattern recognition
Sensors within a module are placed to better than
5m and 2mr Relative to each other
Miss-placements of up to 10m do not significantly
degrade the Ultimate muon Pt resolution even if
not corrected for in track reconstruction

• DX
• 3 mm

13
S/N performance in System Tests
S/N 25 (20 higher than b rays)
If muons (_at_ 500 mm) 40000 electrons noise (_at_
deco) 1600 electrons identical to predictions.
14
Basic design and performance considerationsfor
the CMS Tracker

• To set the scale for the momentum measurement,
  recall that
• The CMS B Field 4T and the TK Radius 110 cm
  result in
• 1.9mm sagitta for 100 GeV Pt tracks
• (190m sagitta for 1 TeV Pt tracks)
• To set the scale for speed and granularity,
  recall that
• At high luminosity expect 20 min. bias events
  every 25ns
• gt a very high charged particle flux (modified
  the B field)
• R 10cm 25cm 60cm
• Nch/(cm225ns) 1.0 0.10
  0.01

15
The CMS Tracking Strategy

• Rely on few measurement layers,
• each able to provide
• robust (clean) and precise coordinate
  determination

2 to 3 Silicon Pixel, and 10 to 14 Silicon Strip
Measurement Layers
Radius 110cm, Length 270cm
h1.7
6 layers TOB
h2.4
4 layers TIB
9 disks TEC
3 disks TID
16
Design considerations for CMS SST Cell size
strip pitch
Efficient clean track reconstruction is ensured
provided occupancy below few
DPt/ Pt 0.1Pt (Pt in TeV) allows to
reconstruct Z to mm- with DmZ lt 2GeV up to Pt
500GeV
Twelve layers with (pitch/ ? 12) spatial
resolution and 110cm radius give a momentum
resolution of
At small radii need cell size lt 1cm2 and fast
(25ns) shaping time This condition is relaxed at
large radii
A typical pitch of order 100mm is required in the
phi coordinate To achieve the required resolution
Strip length ranges from 10 cm in the inner
layers to 20 cm in the outer layers. Pitch
ranges from 80mm in the inner layers to near
200mm in the outer layers
17
Robust and clean hits
Hit contamination is 4 in the first Silicon
Strip layer Less than 2 elsewhere
4
2
18
Partial Track reconstruction

• Good track parameter resolution
• already with 4 or more hits

19
Robust Pattern Recognition
Well defined track parameters with 4 or more
hits gt Small uncertainties on the predicted
track state
20
Robust Pattern Recognition
Extrapolation to Pixel Layer 3 is matched to a
spurious hit in less than 5 of the cases
Includes empty hit
21
Robust Pattern Recognition
15-20 of track candidates Are matched to a
spurious hit
1 of track candidates Are matched to a
spurious hit
22
Robust Pattern Recognition
Even in the most crowded situations,lt10 of track
candidates extrapolated from Barrel Si Strip
layer 1 are matched to a spurious hit
23
Track reconstruction
24
Track reconstruction efficiency in jets
95
95
Efficiency for particles in a 0.4 cone around
jet axis No significant degradation compared to
single pions Loss of efficiency is dominated by
hadronic interactions in Tracker material
25
Impact Parameter resolution
For 10 GeV Pt tracks, s(d0) lt30m for hlt1.5
degrading to 40m for h2.4
For 10 GeV Pt tracks, s(Z0) lt50m for hlt1.5
degrading to 150m for h2.4
26
Pt Resolution For High Momentum Muons
The CMS Tracker provides 1 Pt resolution over
0.9 units of h, and 2 Pt resolution up to h
1.75, beyond which the lever arm is reduced
Even at 100 GeV muons are significantly affected
by multiple scattering a finer pitch, and higher
channel count Would therefore yield only
diminishing returns in improving the Pt resolution
27
Alignment tools and strategy
Tools implemented to introduce, And account for,
misalignments Following the hierarchical
organization Of the mechanical degrees of
freedom Inherent in the support structures
Pattern recognition works efficiently and cleanly
with misalignments of up to 1mm, for W-gtmn
events at 21033
This is the essential starting Point for
alignment with tracks
28
High quality track reconstruction code
Fully functional code released January
2001 Good efficiency and track quality
Well designed modular architecture Excellent
framework for systematic optimization
Timing analysis 85 in Trajectory
building Dominated by search for compatible
layers 5 times faster layer search gt Overall 3
times faster reconstruction (at least in the
barrel)
Example of technical improvements Previous track
propagator was good enough for government
work New track propagator perfect
29
The CMS Trigger and DAQ architecture Two level
data reduction Lvl-1 trigger HLT filter
Lvl-1 crude granularity and Pt
resolution Rate dominated by miss-measured jets
leptons
40 MHZ
HLT task reduce rate by 1000 Exploit much
better Granularity and Pt resolution to correctly
tag and retain only interesting physics events
50 KHz
100 Hz
On average 300ms available for HLT Decision on
any given event (Normalized to a 1GHz Pentium)
4 DAQ slices in 2007 gt 50 KHZ into HLT, 100 Hz
out
30
Examples of Tracker _at_ HLT tau tagging
Regional Tracking Look only in Jet-track
matching cone Loose Primary Vertex association
Conditional Tracking Stop track as soon as Pixel
seed found (PXL) / 6 hits found (Trk) If Ptlt1 GeV
with high C.L.
Reject event if no leading track found
Regional Tracking Look only inside Isolation
cone Loose Primary Vertex association
Conditional Tracking Stop track as soon as Pixel
seed found (PXL) / 6 hits found (Trk) If Ptlt1 GeV
with high C.L.
Reject event as soon as additional track found
31
Calo-PXL A0/H0-gt2t-gt2t jets
Optimization of the Calo-PXL signal efficiency as
a function of the Calo Tau Trigger suppression
factor For a fixed overall suppression factor 103
of the full path Calo rejection 3, pixel
rejection 330, time 175ms at high luminosity
Pile up tracks in isolation cone lead to some
loss of signal efficiency at high luminosity for
Calo-PXL
32
By adding a few Tracker hits, can measure track
momentum
Cut on leading track Pt (gt6,7 GeV) allows to
reduce isolation cone size gt higher signal
efficiency and less sensitivity to pile-up
Calo-Trk A0/H0-gt2t-gt2t-jet
0.44
Low lumi
0.43
High lumi
Trk tau fast enough at low luminosity for full L1
rate At high luminosity currently need a moderate
Calo pre-selection factor to reduce time
Trk tau timing _at_ low lumi QCD events
Trk tau timing _at_ low lumi signal events
33
Inclusive b tagging at HLT(exclusive Bs see
V.Ciullis talk)
Regional Tracking Look only in Jet-track
matching cone Loose Primary Vertex association
Conditional Tracking Stop track as soon as Pixel
seed found (PXL) / 6 hits found (Trk) If Ptlt1 GeV
with high C.L.
300ms low lumi 1s high lumi
Inclusive b tag at HLT possible (provided
alignment under control) Now considering how to
extend CMS (SUSY) physics reach using this
34
Radiation Length in the Tracker
As a result of the attention paid to controlling
the material budget in the design of the CMS
Tracker, nothing sticks out particularly. It
does, however, add up
35
Electron reconstructionwith the CMS Tracker
The design is frozen (the Tracker is under
construction!) and its heavy How to make the
best of it, also for electrons?
For electrons, using Bethe and Heitler formula
for energy loss (Yellow distribution) works
better than treating them as muons (White
distribution) Can one do better?
36
Electron reconstructionwith the CMS Tracker
In the standard treatment, a single Gaussian is
used to approximate the underlying probability
distribution
The energy loss of electrons in material is
manifestly not well described by this
37
Electron reconstructionwith the CMS Tracker
Gaussian Sum Filter (GSF) Approximate Bethe
Heitler with multiple Gaussians At each material
layer create and test new track hypotheses
corresponding to each of these Gaussians Retain
only the best ones (combinatorial reduction)
and continue
Residual and probability distributions for a
sample of 10 GeV electrons in the barrel
GSF significantly improves the resolution FWHM
is reduced by factor of 2 And provides a better
estimate of the errors
38
CMS Inner TrackingSummary and Conclusions

• The CMS Silicon Tracker has robust performance in
  a difficult environment
• The pixel vertex detector allows fast efficient
  track seed generation,
• As well as excellent 3-D secondary vertex
  identification
• The fine granularity of the pixel and strip
  sensors, together with the analyzing power of the
  CMS 4T magnet allow for a 2 or better Pt
  resolution for 100GeV muons over about 1.7 units
  of rapidity
• A good determination of track parameters with
  only a few hits (46) allows fast clean pattern
  recognition
• This makes possible the extensive use of track
  information at HLT level for essentially the full
  L1 output stream at both high and low luminosity
• We are now studying how this may be used to
  improve and extend the physics reach of the CMS
  experiment, in particular with jet flavor tagging