Transcript Pixels. - INFN Genova

What a Pixel detector should be doing there?
= the role
• Must precisely (~0.01mm) reconstruct tracks (and then vertices) in the vicinity
(4<r<15cm) of the p-p collision.
• Pixel is the only solution there because of the track density and the
expected dose.
• Pixel is of paramount importance to identify particles with ps lifetime
(like b-hadrons) which are important for Standard Model studies, and
messengers of (possible) BSM physics.
• If t~ps relativistic b-hadrons fly ~mm before decaying and their decay
and its decay products miss the interaction vertex by an impact
parameter d0 such that <d0>~ct (i.e.~0.3mm on average).
• Must determine both VTX1 and VTX2
with accuracies <<0.3mm
(VTX2)
Forse qua va messo + in
evidenza che il b-tagging si fa
misurando d0 o VTX1/2
26/03/16
(VTX1)
Leonardo Rossi - INFN Genova
1
• Accuracy of track extrapolation
to VTX1 & VTX2 depends on:
•
•
the lever arm extrapolation
distance (r1) and
the detector accuracy (s) and
thickness (X0) because of multiple
Manca def di r2, x
scattering
q
B~sqrt(x/X0)
• VTX1 is the origin for all physics
measurements (e.g. Hgg).
Mi sembra un po’ mischiata la
Invariant mass for massless
particles
risoluzione
e lo 2scopo del
body decay M2=2p1VTX1
p2(1-cosq)
con quello di VTX2.
Farei due slides
• VTX1 accuracy is important!
26/03/16
Leonardo Rossi - INFN Genova
2
• What are the main challenges a Pixel detector has to face to do its job
in the LHC environment?
Ti avrei detto che manca il requisito
• Granularity must be high to measure tracks
with the
necessary
sulla memoria
locale.
Ma forse si
precision (0.01mm) in the expected track
density
(~200
per frase,
h-unit)
puo’
riscrivere
questa
intanto
50x400mm2 pixel area (or equivalent)
•
•
•
non e’ che il pixel debba triggerare
must have fast read-out (<25ns time resolution), be able to compress information
(occupancy <10-4) and store information until a trigger arrives and select the
interesting events (react to trigger)
size of the electronics in front of pixel should match (i.e. ~103 transistors in 2 104 mm2)
must implement high density (and out-of-plane) connectivity (~5000/cm2)
• Radiation hardness should be high (100 kGy) to sustain the
particle flux over 10 years of operation at ~5 cm radius
•
•
all materials must (circuits, sensors, supports, glues, etc.) be qualified to 100kGy
detector must operate cold (-10C) to minimize radiation damage
• Detector must be “transparent” (~2% X0) over -2.5<h<2.5
•
•
not to “pollute” the information for downstream detectors
to better extrapolate tracks to VTX1&2
See how this is implemented taking ATLAS pixel as an example
26/03/16
Leonardo Rossi - INFN Genova
3
• A blow-up view of the ATLAS HPD [3 barrel layers + (3+3) disks], with
• the mechanical support in evidence and
• the mosaic arrangements of pixel modules to cover the full acceptance
Thin CRF (Carbon…boh, non so
neanche io l’acronimo)
mechanical support
Pixel modules
26/03/16
Vale la pena
mettere raggio e
Leonardo Rossi - INFN Genova
lunghezza?
2.5<h<2.5
4
8 107 channels
Pixel surface 50x400mm2
• Each tessera of the
mosaic is a replica (on a
surface of ~12cm2) of the
individual pixel unit
• The signals are first
elaborated and stored in
the Front-end chips,
waiting for the trigger
decision to transfer out
the information of
interest
26/03/16
80mm
thick
250mm thick
Paper dice che FE e’
200 um e flex 100.
per IBL il FE e’ 150
Leonardo Rossi - INFN Genova
150mm thick
5
• The signals are then sent
out through a two-step
connection system (while
control signals are sent in)
• First Al-wirebond from
the FE-electronics to the
flex circuit
• Then, after further coding
(MCC) a miniaturized
connector (and later
optical fibres)
wirebonds
Bond
foot
26/03/16
Leonardo Rossi - INFN Genova
6
• Front-end chip (FEI3) realized with
250nm IBM technology (intrinsically
rad hard as the oxide layers are thin and
charge is not trapped inside the oxides)
 ~no threshold shift in the
transistors (i.e. ~no change in
operating conditions).
• Sufficiently dense technology to
accommodate 103 transistor per
pixel (and therefore to include both
analogue and digital treatment of the
signal)
Part facing
pixel
Coding part
26/03/16
Leonardo Rossi - INFN Genova
7
• Thanks to very low Pixel capacitance (~200 fF) FE electronics can be
designed such that noise is ~200e-, while MIP signal is ~20ke(S/N~100).
• Threshold can be set as low as 2ke-, but parameter variations induce
threshold spread over a chip. This can be corrected by individual tuning
(otherwise it contributes to noise)
a
200
b
150
Tuned per chip
from Rossi et al
Tuned per pixel
150
100
100
50
50
0
0
0
1000
2000
3000
4000
5000
0
1000
Threshold [e -]
26/03/16
Leonardo Rossi - INFN Genova
2000
3000
4000
5000
Threshold [e -]
8
Operation performance : number of dead channels
Row
• Defects are at the level of 1% of the channels and due either to dead
chips or to bump-bonding failures
• Plus some dead modules
(https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PLOTS/PIX-2015009/)
300
ATLAS Preliminary
250
200
150
100
50
0
0
20
40
60
80
100
120
140
Column
26/03/16
Leonardo Rossi - INFN Genova
9
Operation performance : noisy channels
Noise occupancy, Hits/Pixel/BC
• Having S/N~100 and S/threshold~10 gives a very “silent” detector
which is important because of the large number of channels.
10-7
ATLASPreliminary
10-8
10-9
Better than 10-9
noise occupancy
10-10
Raw hits
First pass reconstruction
-11
10
Bulk reconstruction
10-12
153
1
1
1
1
1
1
1
1
1
1
1
1
134 53136 53159 53200 53565 53599 54465 54466 54471 54813 54815 54817 54822
Run number
• This is obtained after having rejected ~1000 channels (run dependent)
which are noisy (i.e. ~1 channel in 105)
26/03/16
Leonardo Rossi - INFN Genova
10
Operation performance : timing
• If the cluster has small charge, it
arrives a bit late (i.e. small % in next
1
BCO, it can be recuperated…)
ATLAS Preliminary
Pixel timing
0.8
Minimum Bias Stream
Solenoid on runs 141749 to 142154
Clusters off track
0.6
Clusters on track
0.4
0.2
0
Hit detection time [ns]
Fraction of clusters
• LHC has a 25ns bunch structure  resolution should be better than
25ns to associate an event to a given bunch crossing (BCO)
25
ATLAS Preliminary
s=7
s
=7 TeV
20
15
10
5
0
26/03/16
2
4
6
8
10 12 14
Bunch crossing
0
0
Leonardo Rossi - INFN Genova
5
10
15
20
25
30
Hit charge [ke]
11
Operation performance : efficiency
Association efficiency
• Typical efficiency is 99% (fully efficient neglecting dead channels)
1
0.995
0.99
0.985
0.98
0.975
0.97
0.965
0.96
ATLAS Preliminary
s=7 TeV
Disk
3
C
Disk
2C
Disk
1C
B-L
L
L
aye ayer-1 ayer-2 Disk1A Disk2A Disk3A
r
Pixel layer
26/03/16
Leonardo Rossi - INFN Genova
12
Operation performance : accuracy
s(d ) [m m]
0
Data 2015, s = 13 TeV
s(d ) [m m]
2015/2012
900
800 ATLAS Preliminary
0.4 < pT < 0.5 [GeV]
700
600
500
-1 -0.5
0
1.5
2
Data 2012, s = 8 TeV
1
Data 2015, s = 13 TeV
0.5
2.5
h
Leonardo Rossi - INFN Genova
400
300
-1.5
26/03/16
0
200
100
-2
innermost layer improves at low-momentum)
0
1
h
0.8
0.6
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
-2.5
• Position resolution (in the short pixel
direction) ~10mm (slightly worse in high pile-up).
• Extrapolation to VTX1 (IP=impact parameter)
~20mm at large momentum (adding an
Data 2012, s = 8 TeV
2 3 4 5 6 7 8910
20
pT [GeV]
4-layer pixel
400
1
350 ATLAS Preliminary
0.0 < h < 0.2
300
250
200
150
100
50
0
4´10-1
3-layer pixel
13
Operation performance : b-tagging
10
ATLAS Simulation Preliminary
2
Efficiency
c-jet rejection
• The capability to select jets containing b-quarks (b-tagging) depends
heavily on the IP resolution as tb~1ps.
ATLAS Simulation Preliminary
s=13 TeV, tt
1
MV2c20
MV2c00
10
10
-1
b jets
MV2c20
c jets
e =70%
s =13 TeV, t t
jet
10
-2
Light-flavour jets
b
jet
p >20 GeV, | h |<2.5
1
T
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
b-jet efficiency
26/03/16
10
-3
Leonardo Rossi - INFN Genova
100
200
300
400
500
600
Jet p [GeV]
T
14
Main lessons learned
• Hybrid Pixel Detectors requires a sophisticated know-how in many
fields (sensors, electronics, mechanics, etc.) therefore a large,
diversified and strongly motivated team.
• It was not easy to have HPDs accepted by the LHC collaborations in
the late 90’s (risk!), but it turned out they contributed very much to
LHC physics  now a lot of enthusiasm to use them in the LHC High
Intendi luminosita’ integrata? Allora
Luminosity Upgrade (2025?)
da 300 a 3000fb-1; se istantanea
• This is a project to extend the physics reachinvece
of LHC:
direi che si arrivera’ a 5-7.
• better study the Higgs (production andspecificherei
decay) peak o integrated.
se peak
, direi 7(BSM)
rispetto al
• search for new phenomena Beyond theE forse
Standard
Model
disegno.
with a 7-fold increase of the luminosity.
• It means up to 200 collisions in each BCO (now 25-40)  both
instantaneous and integrated dose push the other tracking detectors
to outer radius and give a larger role to Pixels.
26/03/16
Leonardo Rossi - INFN Genova
15
Events are complex at HL-LHC…
Rimosse figure per fare file piu’ piccolo
especially around the collision point…
26/03/16
Leonardo Rossi - INFN Genova
16
• Pixel detector extend to large radius (12.5cm  32cm) with more
layers, possible extension to h=4 and much larger surface covered.
• This is the challenge for the next 10 years: maintaining the same
performance of the current trackers with a much larger pile-up
26/03/16
Leonardo Rossi - INFN Genova
17
Another side of the LHC physics (and associated Pixel)
• The ALICE experiment is designed for
the study of the strong interactions
through Pb-Pb collisions (in particular the
Forse parlerei prima di Alice che di
Hl-LHC. cCosi’ uno puo’ dire come
funzionano bene gli attuali rivelatori,
con richieste diverse da ATLAS_CMS
properties of the q-g plasma, a status of matter
consisting of quasi-free quarks and gluons that can
be generated at very high energy density e.g.
colliding two heavy ions).
• This was the state of matter msec after big-bang (and QCD is the less
precisely tested part of the SM of particles).
• The study of q-g plasma is a complex issue and capability of
reconstructing short-lived particles (mostly s-quarks, i.e. ns lifetimes) is
part of the game  this is the role of the current pixel detector in
ALICE
• It works in a quite different environment than we have seen so far
26/03/16
Leonardo Rossi - INFN Genova
18
• Collision rate ~10-4 of ATLAS/CMS, but track multiplicity per event ~x 50
vs p-p and mostly low momentum tracks)  Pixel must not be fast, but
be small and thin.
• The onset of the quark-gluon
plasma should be better visible
in the central part of the hacceptance (|h|<…).
• That’s where the the ALICE pixel
detector is located (only barrel, no
disks).
• Also ALICE upgrades (in the next 4
years) and profits of the
development in electronics for
automotive applications where
CMOS is implanted in High
Resistivity silicon (HRCMOS).
26/03/16
Leonardo Rossi - INFN Genova
19
• As timing is less of a problem (e.g. Pb-Pb interactions in ALICE at ~4kHzUpgr: 50
kHz) charge may be collected by diffusion.
• E-field cannot be as uniform as in HPDs (charge collection by diffusion (mostly) +
drift), but complex CMOS circuit can be implanted around the collecting
electrode  electronics and sensor on the same substrate
e
h
e-h diffusion vs time
Charge generated in high-r epitaxial-layer 
diffuses laterally to collecting electrode
(doping concentration create potentials and charge
confined by potentials diffuse to electrode (n-well diode) )
and drifts (only when very close to it)
26/03/16
To arrive (diffusing) to the collecting
electrode charge would need hundreds
of ns (and no trapping due radiation damage)
Leonardo Rossi - INFN Genova
20