scipp development of radiation hard tracking detectors · scipp development of radiation hard...
TRANSCRIPT
1Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Development of Radiation Hard Tracking Detectors
Hartmut F.-W. Sadrozinski, SCIPP UC Santa Cruz, 1156 High Street, 95064 CA, USA
1) LHC Upgrade environment2) New tracker materials?3) Radiation damage in Si
a) Effectsb) Present microscopic interpretation
4) Bias Dependence of collected charge5) Thin sensors and Pixel sensors
2Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
1/9/2008 2
LHC vs time: a wild guess …
L=1035
?
3Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
1/9/2008 3
L=1035
Time to half Stat. Error
0
5
10
15
2008 2010 2012 2014 2016 2018 2020 2022Year
Year
s
LHCLHC + sLHC
Need to Upgrade
ID dies of radiation damage
4Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPFluence in Proposed sATLAS Tracker
sATLAS Fluences for 3000fb-1
1.E+12
1.E+13
1.E+14
1.E+15
1.E+16
1.E+17
0 20 40 60 80 100 120
Radius R [cm]
Flue
nce
neq/
cm2
All: RTF Formulan (5cm poly)pionproton
ATLAS Radiation Taskforce http://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/RADIATION/RadiationTF_document.html
5 - 10 x LHC Fluence
Mix of n, p, πdepending on radius R
Pixels
Radial Distributionof Sensors determined by Occupancy < 2%
ShortStrips
LongStrips
Design fluences for sensors (includes 2x safety factor) :Innermost Pixel Layer: 1*1016 neq/cm2
Outert Pixel Layers: 3*1015 neq/cm2
Short strips: 1*1015 neq/cm2
Long strips: 4*1014 neq/cm2
Strips damage largely due to neutrons
Pixels Damage due to neutrons+pions
5Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
42 European and Asian institutesBelarus (Minsk), Belgium (Louvain), Czech Republic (Prague
(3x)), Finland (Helsinki) , Laappeenranta), Germany (Dortmund, Erfurt, Freiburg, Hamburg, Karlsruhe, Munich), Israel (Tel Aviv),
Italy (Bari, Bologna, Florence, Padova, Perugia, Pisa, Torino, Trento), Lithuania (Vilnius), Netherlands (NIKHEF), Norway(Oslo (2x)), Poland (Warsaw(2x)), Romania (Bucharest (2x)),Russia (Moscow, St.Petersburg), Slovenia (Ljubljana), Spain
(Barcelona, Valencia), Switzerland (CERN, PSI), Ukraine (Kiev), United Kingdom (Exeter, Glasgow, Lancaster, Liverpool)
8 North-American institutesCanada (Montreal), USA (BNL, Fermilab, New Mexico, Purdue,
Rochester, UC Santa Cruz, Syracuse)
Started in 2002, now 261 Members from 50 Institutes
Detailed member list, Progress Reports, Workshop papers etc. :
http://cern.ch/rd50
RD50 – CERN R&D Collaboration “Development of Radiation Hard Semiconductor
Devices for High Luminosity Colliders”
6Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
RD50 started when the design of the LHC tracking detectors based on large-scale silicon strip and pixel sensor were frozen.
The performance of the LHC Si sensors (p-on-n FZ strips, n-on-noxigenated FZ DOFZ) was limited by radiation damage. The aim of RD50 is to develop Radiation-Hard Semiconductor Devices for High Luminosity Colliders -- like the LHC Upgrade, the Super-LHC (sLHC).
The program of RD50 was laid out and is progressing along these lines:
• Extend the radiation testing to the predicted sLHC fluences (1016, neutrons!)
• Search for alternative sensor materials to Si (more radiation-hard)
• Develop new experimental methods to help gaining insight into the radiation damage mechanism
• Understand on the microscopic level the observed radiation damage effects
• Optimize sensor geometry for radiation tolerance
• Start to transfer fabrication to commercial manufacturer in anticipation of large-scale production
RD50 : R&D Plan
7Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Property Diamond GaN 4H SiC Si Eg [eV] 5.5 3.39 3.26 1.12 Ebreakdown [V/cm] 107 4·106 2.2·106 3·105 μe [cm2/Vs] 1800 1000 800 1450 μh [cm2/Vs] 1200 30 115 450 vsat [cm/s] 2.2·107 - 2·107 0.8·107 Z 6 31/7 14/6 14 εr 5.7 9.6 9.7 11.9 e-h energy [eV] 13 8.9 7.6-8.4 3.6 Density [g/cm3] 3.515 6.15 3.22 2.33 Displacem. [eV] 43 20 25 13-20
New Materials: Diamond, SiC, GaN
• Wide bandgap (3.3eV) ⇒ lower leakage current
than silicon
• Signal:Diamond 36 e/μmSiC 51 e/μmSi 89 e/μm
⇒ more charge than diamond
• Higher displacement threshold than silicon
⇒ radiation harder than silicon (?)
R&D on diamond detectors:RD42 – Collaboration
http://cern.ch/rd42/
1013 1014 1015 101610-11
1x10-9
1x10-7
1x10-5
1x10-3
Leak
age
Cur
rent
Den
sity
[A/c
m2 ]
Fluence [1MeV n/cm2]
Si 300μm; Vfd = 600V -30°C SiC 55μm +20°C pCVD diam 300μm +20°C
Partial depletion
8Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
1014 1015 10160
5000
10000
15000
20000
25000
# el
ectro
ns
1 MeV n fluence [cm-2]
strips
pixels
p FZ Si 280μm; 25ns; -30°C [1] p-MCz Si 300μm;0.2-2.5μs; -30°C [2] n EPI Si 75μm; 25ns; -30°C [3] n EPI Si 150μm; 25ns; -30°C [3] sCVD Diam 770μm; 25ns; +20°C [4] pCVD Diam 300μm; 25ns; +20°C [4] n EPI SiC 55μm; 2.5μs; +20°C [5] 3D FZ Si 235μm [6]
[1] G. Casse et al. NIM A (2004)[2] M. Bruzzi et al. , STD6[3] G. Kramberger, RD50 Work. Prague 06[4] W: Adams et al. NIM A (2006)[5] F. Moscatelli RD50 Work.CERN 2005[6] C. Da Vià, STD6: Laser only
Not shown: GaN, other excotics
Plot (Sept. ‘06) outdated: much new Si data will be discussed
Comparison of measured collected charge on different radiation-hard materials and devices(M. Bruzzi, H. F.-W. Sadrozinski, A. Seiden, NIM A 579 (2007) 754–761)
New materials for Tracking Sensors?
Line to guide the eye for planar devices
At a fluence of ~1015 neq/cm2, all planar semiconductor sensors loose sensitivity:on-set of trapping!Solution for Pixels: 3-D sensors, which provide shorter drift distance
No “Wunder” material found, so investigate in depth best known and devloped material: Si.Erosion of collected charge @ Φ > ~1015 neq/cm2 : is S/N ok?
9Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Reverse Bias of junction: only thermal current generationScale : Band gap 1.12eV vs. kT = 1/40eV: huge Boltzmann factor
Cooling needed only in ultra-low noise applications.Wafer thickness 300um = 0.3%RL: 23k e-h pairs (small fluctuations)Depletion Voltage ~ thickness2 : < 100V Collection Time of e-h pairs: ~ 20nsPosition Resolution: 10-100umArea is given by wafer size: 4” & 6” => Ladders
Properties of Silicon Strip Detectors SSD
25-200 μm
300-400 μm
Alat ~ 100V
n+ implant
Al SiO2
p+ implantat ground
Depletion region. Charged particletraversing region produces ~80electron/hole pairs per micron.
Readout electronics(S/N typically > 20)
holes
Determine location of single particles on the micron scale
10Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPIf you still have Questions…
http://britneyspears.ac/lasers.htm
Visit your Favorite Web Site
SCIPPSCIPPRD50
SCIPPSCIPPRD50 Radiation Damage in Silicon Sensors
• Bulk (Crystal) damage due to Non Ionizing Energy Loss (NIEL)- displacement damage, built up of crystal defects, proportional to fluence
I. Change of effective doping concentration⇒ type inversion, higher depletion voltage, under-depletion
⇒ loss of active volume ⇒ decrease of signal, increase of noiseDifferent for neutron and proton irradiation
II. Increase of leakage current⇒ increase of shot noise, thermal runaway, power consumption…
⇒ need for cooling (Diamond has advantage, RD42)Universal if fluence is scaled according to NIEL
II. Increase of charge carrier trapping⇒ loss of charge
Different for electrons and holes
• Surface damage due to Ionizing Energy Loss (IEL- accumulation of +charges in oxide (SiO2) and the Si/SiO2 interface, saturates
⇒ interstrip capacitance, strip isolation, breakdown behavior, …
Influenced by impuritiesin Si – Defect Engineeringis possible!
~ Same for all tested Silicon materials!
• Important latent effects:Space charge sign inversion (SCSI)Annealing
Hartmut F.-W. Sadrozinski, SCIPP IEEE07
12Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
particle Sis Vacancy + Interstitial
Point Defects (V-V, V-O .. ) clusters
EK > 25 eV EK > 5 keV
Frenkel pairV
I
[Mika Huhtinen NIMA 491(2002) 194]
10 MeV protons 24 GeV/c protons 1 MeV neutrons
Initial distribution of vacancies after 1014 particles/cm2
Mainly clustersMore point defects
Ec
Ev
Ei
V-O
Ci-Oi
V2-
V2-O
Clusters
SD/TD
Main radiation induced defects in Si
Point -Defects
Radiation Damage Induced Defects in Bandgap
13Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPRadiation Damage caused by Defects
Trapping (e and h)⇒ CCE
shallow defects do not contribute at room
temperature due to fast detrapping
charged defects ⇒ Neff , Vdep
e.g. donors in upper and acceptors in lower half of band
gap
generation⇒ leakage current
Levels close to midgap
most effective
Trapping (e and h)⇒ CCE
shallow defects do not contribute at room
temperature due to fast detrapping
charged defects ⇒ Neff , Vdep
e.g. donors in upper and acceptors in lower half of band
gap
generation⇒ leakage current
Levels close to midgap
most effective
14Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPRadiation Damage – Leakage Current
1011 1012 1013 1014 1015
Φeq [cm-2]10-6
10-5
10-4
10-3
10-2
10-1
ΔI /
V
[A/c
m3 ]
n-type FZ - 7 to 25 KΩcmn-type FZ - 7 KΩcmn-type FZ - 4 KΩcmn-type FZ - 3 KΩcm
n-type FZ - 780 Ωcmn-type FZ - 410 Ωcmn-type FZ - 130 Ωcmn-type FZ - 110 Ωcmn-type CZ - 140 Ωcm
p-type EPI - 2 and 4 KΩcm
p-type EPI - 380 Ωcm
[M.Moll PhD Thesis][M.Moll PhD Thesis]
• Damage parameter α (slope in figure)
Leakage currentper unit volume
and particle fluence
• α is constant over several orders of fluenceand independent of impurity concentration in Si
can be used for fluence measurement
eqVI
Φ⋅Δ
=α
80 min 60°C
• Leakage current decreasing in time (depending on temperature)
• Strong temperature dependence:
Consequence:Cool detectors during operation!
1 10 100 1000 10000annealing time at 60oC [minutes]
0
1
2
3
4
5
6
α(t)
[10-1
7 A/c
m]
1
2
3
4
5
6
oxygen enriched silicon [O] = 2.1017 cm-3
parameterisation for standard silicon [M.Moll PhD Thesis]
80 min 60°C
annealing
cmAxC 1710)03.099.3(min)80,60( −±=°α ⎟⎟
⎠
⎞⎜⎜⎝
⎛−
KTE
TTI gexp)( 2α
15Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPRadiation Damage – Effective Doping
Change of Depletion Voltage Vdep in high resistivity n-type FZ Si
Full depletion voltage too high to fully depletedetectors at very high fluences (>1000V @ 1015cm-2
“Type inversion”: Neff changes from positive to negative (Space Charge Sign Inversion)
10-1 100 101 102 103
Φeq [ 1012 cm-2 ]
1
510
50100
5001000
5000
Ude
p [V
] (d
= 3
00μm
)
10-1
100
101
102
103
| Nef
f | [
1011
cm
-3 ]
≈ 600 V
1014cm-2
type inversion
n-type "p-type"
[M.Moll: Data: R. Wunstorf, PhD thesis 1992, Uni Hamburg]
…. annealing
Short term: “Beneficial annealing”Long term: “Reverse annealing”3x stable component
- time constant depends on temperature:~ 500 years (-10°C)~ 500 days ( 20°C)~ 21 hours ( 60°C)
- Consequence: Detectors must be cooledeven when the experiment is not running!
NC
NC0
gC Φeq
NYNA
1 10 100 1000 10000annealing time at 60oC [min]
0
2
4
6
8
10
Δ N
eff [
1011
cm-3
]
[M.Moll, PhD thesis 1999, Uni Hamburg]
2
2xNeffqV
ε= →2
2xNeffqV
ε= →
16Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Influence the defect kinetics by incorporation of impurities or defects: Oxygen
Initial idea: Incorporate Oxygen to getter radiation-induced vacancies⇒ prevent formation of Di-vacancy (V2) related deep acceptor levels
•Higher oxygen content ⇒ less negative space chargeOne possible mechanism: V2O is a deep acceptor
O VO (not harmful at RT)V
VO V2O (negative space charge)
V2O(?)
Ec
EV
VO
V2 in clusters
Defect Engineering of Silicon
0 1 2 3 4 5Φ24 GeV/c proton [1014 cm-2]
0
2
4
6
8
10
|Nef
f| [1
012cm
-3]
100
200
300
400
500
600
Vde
p [V
] (30
0 μm
)
Carbon-enriched (P503)Standard (P51)
O-diffusion 24 hours (P52)O-diffusion 48 hours (P54)O-diffusion 72 hours (P56)
Carbonated
Standard
Oxygenated
DOFZ (Diffusion Oxygenated Float Zone Silicon) RD48 NIM A465 (2001) 60 MCz (Magnetic Czochralski) has even higher O content NIM A568, 1, (2006), 83-88
17Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
0,0 5,0x1014 1,0x1015 1,5x10150
200400600800
100012001400
Vfd = 7E-13 fn - 18,818
24 GeV p - n-MCz Si [1] Reactor n n-MCz Si [2]
Vfd [V
]
Fluence 1 MeV neq [cm2]0 50 100 150 200 250 300
0
1x104
2x104
3x104
24 GeV protons 2.1x1015 cm-2
Reactor neutrons 5x1014 cm-2
Ele
ctric
Fie
ld [V
/cm
]
x [μm]
“Double Junction” develops mostly from back in neutron irradiated silicon, probably due to a high generation of cluster, behaving as a source of quasi-midgap acceptors.
n-type MCz Si after 24GeV p and after reactor neutrons:
Oxygen not effective for neutrons
Electric field distribution by TCT (Ioffe-St Petesburg on SMART n-
MCz Si single pad diodes)
M. Bruzzi 6th International "Hiroshima" Symposiumk STD06, Carmel Sept. 06
Difference in p and n irradiation in n-type Si
18Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPCharge Trapping
Collected Charge is limited by:
Collected Charge: trapεε Q Q depo ⋅⋅=
depbiasdep VVWx /==ε
t
c
etrapττ
ε−
=
W: Detector thicknessx: Active thicknessτc : Collection timeτt : Trapping time
Partial depletion Trapping at deep levels
1/τe,h = βe,h·[Φeq/1016 cm-2]βe,h = 3
τt = 3 ns for Φ = 1015 cm-2 ,
τt = 0.3 ns for Φ = 1016 cm-2
Trapping is the ultimate limitation for the active region in all semiconductors (even for Diamond!)
Annealing favorable for electrons not favorably for holes
Bias dependence of charge collection:increase depleted region & speed up collection J. Weber, XI RD50 Workshop, CERN, Nov ‘07
19Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
p-on-n silicon, under-depleted:
• Charge spread – degraded resolution
• Charge loss – reduced CCE
p+on-n
n-on-p silicon, under-depleted:
•Limited loss in CCE
•Less degradation with under-depletion
•Collect electrons (fast)
n+on-p
Be careful, this is a very schematic explanation,reality is more complex, e.g. DJ!
Device engineering: Collection at junctionp-on-n versus n-on-p (or n-on-n) SSD
n-type silicon after high fluences: p-type silicon after high fluences:
20Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPCharge collection efficiency CCE on p- and n-side after Inversion
Advantage of collecting on the n-side after inversion was established long time ago:
Double-sided n-type SSD
106 Ru telescope
(5.1013p/cm2)
This motivated the
n-on-n development for (ATLAS SCT), ATLAS & CMS Pixels, LHCb Velo
Paradigm Change for Upgrade Tracker:
Collect electrons, use n-on-p sensors (expect be much cheaper than n-on-n)
21Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPSSD Development for ATLAS Upgrade Tracker
Inst
itutio
n
P-ty
pe S
SD
Stri
p Is
olat
ion
HV
des
ign
Mod
ules
con
str.
Mec
h. S
truc
ture
s co
olin
g
Hyb
rids
,Rea
dout
Ele
ctr.
Cha
r.
Las
er C
CE
CC
E
Tra
ppin
g
Ope
r. P
aram
.
KEK x x x x x x x x x x Tsukuba x x x x x x Liverpool x x x x x x x x Lancaster x x x Glasgow x x x? x x? Sheffield x x x x Cambridge x x x QML x x Freiburg x x x x MPI x x x x Ljubljana x x x x Prague x x x Barcelona x x x x x x x Valencia x x x x x x Santa Cruz x x x BNL x x x PTI x x x x x
22Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Bias Voltage Dependence of Depletion after Neutron irradiation
The depleted sensor thickness x: can be measured two ways:
Capacitance-Voltage C-V measurement:
for un-irradiated sensor:
Collected Charge vs. Bias Q(CCE) – V measurements
for un-irradiated sensor
•All detectors studied are with n+ readout (n-on-p and n-on-n) – “electron collection”
Investigation of p-type SensorsLjubljana Univ., SCIPP - UC Santa Cruz, Univ. of Liverpool
x(V)AεC(V) = 2
2xNeffqV
ε= VC ~/1 2→
)(~)( VxVQ xVQ ~)(
– What collected charge can we expect at fixed bias 500 V for different materials?(ATLAS wants / needs to continue the use of present cables which are rated to 500V!)
– How much does charge collection depend on different readout (pads, strips, binary, analog)
– What is the most promising Si wafer material ?
23Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
RD50 MICRON 6” project
MCz (n-p) MCz(n-n) Fz (n-p) Fz (n-n)WaferResitivity 1.5 Ωcm 2 kΩcm 14 kΩcm ~20 kΩcmOrientation <100> <100> <100> <100>
2552-7 2553-11 2551-7 2535-11
Neutron and Proton and Pion (Aug. ’07 ) irradiation of SSD and Diodes•Liverpool: CCE with SSD and diodes•UCSC: CCE with SSD, both p-type and n-type, C-V, i-V on SSD and Diodes•Ljubljana: CCE with Diodes, C-V, i-V on SSD and Diodes
•36 processed , 20 received•Fz (topsil) and MCz (Okmetic) wafers of p&n type material•n-on-n, n-on-p, p-on-n structures (pixels, strips, diodes)
Strips: ATLAS strips geometry 80 μm pitch (w/p~1/3)Pads: 2.5 x 2.5 mm2 , multiple guard rings
MCz wafers have very high oxygen content
24Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
0
200
400
600
800
1000
1200
1400
0 2 4 6 8 10 12
Φ eq [1014 cm2]
Vfd
[V]
MCz n-n (2553-11)Fz n-n (2535-11)MCz n-p (2552-7)Fz n-p (2551-7)
C-V Measurements after Neutron Irradiation
•all detectors have negative space charge (decrease of Vfd during short term annealing)•Leakage current agrees with expectations (α~3.5-5.5∙10-17A/cm)
Defect Introduction Rate consistent with microscopic picture:•MCz (p and n type): 55 V/1014 cm-2 (gc~0.8 cm-2) – lower stable damage than seen before ?•Fz (p and n type): 125 V/1014 cm-2 (gc~1.8 cm-2) – in agreement with previous results
Charge collection (CCE) in all p-type wafers and n-type FZ should be similar at these fluences.Expect that n-type MCz should perform better – do we see this performance in CCE?Expect that MCz should be advantageous at very high fluences.
PADS
after 80 min @ 60oCG. KrambergerATLAS Tracker Upgrade Workshop, Valencia 12th-14th Dec. 2007
25Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
A. Affolder ATLAS Tracker Upgrade Workshop, Valencia 12th-14th Dec. 2007
Charge Collection CCE for p-type Sensors
5x1014 n cm-2
0
5
10
15
20
25
0 200 400 600 800 1000 1200Bias Voltage (V)
Cha
rge
Col
lect
ed (k
e- )
Micron FZ (30 kohm-cm)Micron FZ (14 kohm-cm)HPK FZ (5 kohm-cm)HPK MCz (2.1 kohm-cm)Micron MCz (1.5 kohm-cm)
Φ= 5×1014 n cm-2
fluence motivated by long strip (10 cm) region of straw-man design
No effect on CCE seen for different resistivities or FZ/MCz (as expected)
13-14 ke- collected at 500V
0
5
10
15
20
25
0 200 400 600 800 1000 1200
Charge Collected (ke-)
Bias Voltage (V)
1x1015 n cm-2
Micron FZ (14 kohm-cm)HPK FZ (5 kohm-cm)HPK FZ2 (5 kohm-cm)HPK MCz (2.1 kohm-cm)Micron MCz (1.5 kohm-cm)
Φ= 1×1015 n cm-2
fluence motivated by short strip (2.5 cm) region of straw-man design
8.5-10 ke- collected at 500V
26Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPBias Dependence of Charge collection CCE
NN--onon--NN NN--onon--PP
Open – FZ, Solid - MCZ
•Vfd from CV (denoted by arrows) agrees well with the kink in CCE•The slope of charge increase with voltage is directly related toVfd:
•increase of Vfd can be measured by the change of slope and vice versa•Similar Vfd = similar slope -> same E field
•High resistive non-depleted bulk is well reflected in linear increase of charge – different from un-irr.•N-type FZ and all P-type show similar behavior, and N-type MCz has best depletion characteristics, as predicted from C-V
G. Kramberger
ATLAS Tracker Upgrade Workshop,
Valencia 12th-14th Dec. 2007
27Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
1015cm-2
0.00E+00
2.00E+03
4.00E+03
6.00E+03
8.00E+03
1.00E+04
1.20E+04
1.40E+04
1.60E+04
1.80E+04
2.00E+04
0 200 400 600 800 1000 1200bias voltage [V]
Cha
rge
[e]
pad (peak)
strips (binary, mean, MCz n-n)
strips (binary, mean, FZ n-p)strips (analog, peak, Fz n-p)
pads (peak, Fz n-p)
Vfd~590 V measured by C-V
Big difference in the charge collection between MCz n-n and Fz n-p~14000 e vs. ~8000 e at 500V!
Fully depleted detector – one can see kink in Q-V plot
The difference in CCE is related to the difference in Vfd
Fz-p vs. MCz-n Difference due to Depletion Voltage
Vfd from CCE
80min@60oC
0
200
400
600
800
1000
1200
1400
0 2 4 6 8 10 12
Φ eq [1014 cm2]
Vfd
[V]
MCz n-n (2553-11)Fz n-n (2535-11)MCz n-p (2552-7)Fz n-p (2551-7)
G. Kramberger, ATLAS Tracker Upgrade Workshop, Valencia 12th-14th Dec. 2007
28Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Depletion Voltage from C-V vs. from Q(CCE)-VCorrelation of Vfd (CCE, CV)
0
200
400
600
800
1000
1200
0 200 400 600 800 1000 1200
Vfd (measured from CV)
Vfd
(mea
sure
d fro
m k
ink
in C
CE)
Pad detectorsstrips, binarystrips, analog
as irradiated (5e14)
5∙1014 cm-2
1∙1015 cm-2
1∙1014 cm-2
Vfd from CV underestimates the onset of saturation in CCE by max. 100-150 V!
•after Vfd the collected charge continues to increase due to shorter drift
•due to growth of depletion depth from electrode side the offset is smaller than in inverted p-on-n!
Advantage of charge collection on the junction0 100 200 300 400 500
0
100
200
300
400
500standard - oxygenated
Casse et al. Robinson et al. Buffini et al. Robinson et al. Casse et al. Lindstroem et al.
Vre
v 95%
Cha
rge
Col
l. [V
]
Vdep CV analysis [V]
Vfd from saturation in C-V and Q(CCE)-V curves( 80min @ 60oC –end of beneficial annealing )
The correlation holds for all investigated fluences in range of full depletion voltages up to 1000V!G. Kramberger
ATLAS Tracker Upgrade Workshop,
Valencia 12th-14th Dec. 2007
M. Bruzzi, H. F.-W. Sadrozinski
STD6 Carmel, Sept. 07
29Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Long term annealing of charge collection
The CCE for n-type MCz follows the annealing trend of the depletion Voltage Vfd:• initial rise (beneficial annealing) increase in collected charge by ~10% (@500V)•decrease by again 20-30% (@500V) during the reverse annealing 10000 min
Increases for p-type FZ-p by 30% and then slow decrease to the pre-anneal value
After depletion annealing of trapping visible
~30%
~10%
10000 min@60oC = 3 years RT
G. Kramberger, ATLAS Tracker Upgrade Workshop, Valencia 12th-14th Dec. 2007
30Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPAnnealing of p-type FZ and DOFZ sensors
(p irradiated)
• p-type strip detector (280μm) irradiated with 23 GeV p
(7.5 × 1015 p/cm2 )• expected from previous CV
measurement of Vdepon n FZ:- before reverse annealing:
Vdep~ 2800V- after reverse annealing
Vdep > 12000V• no reverse annealing visible in
the CCE measurement !
G.Casse et al.,10th European Symposium on Semiconductor Detectors, 12-16 June 2005
0 100 200 300 400 500time at 80oC[min]
0 500 1000 1500 2000 2500time [days at 20oC]
02468
101214161820
CCE
(103 e
lect
rons
)
800 V800 V
1.1 x 1015cm-2 1.1 x 1015cm-2 500 V500 V
3.5 x 1015cm-2 (500 V)3.5 x 1015cm-2 (500 V)
7.5 x 1015cm-2 (700 V)7.5 x 1015cm-2 (700 V)
M.MollM.Moll
[Data: G.Casse et al., to be published in NIMA][Data: G.Casse et al., to be published in NIMA]
31Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
A. Affolder ATLAS Tracker Upgrade Workshop, Valencia 12th-14th Dec. 2007
CCE vs. Annealing (n irradiated)
Stable plateau of increased signal between 20-100 days (30% for this piece)
(Current decreases by ~30% over same time period)
02468
101214161820
0 200 400 600 800 1000 1200
Col
lect
ed c
harg
e (k
e-)
Eq. days @20oC
HPK FZ-1x1015 n cm-2
500 V700 V900 V
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 200 400 600 800 1000 1200
Ratio
to p
re-a
nnea
ling
Time at 20oC eq. (Days)
HPK FZ-1x1015 n cm-2
500V
700V
900V
At sLHC fluences for p-type sensors, the entire annealing process is much less pronounced. than for n type FZ.It opens the possibility that sensors need to be cooled only during operations to control the leakage current, but not during beam-off time to prevent anti-annealing
32Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPMitigation of Radiation Damage in Si
• Increase depleted region at fixed bias: Wafer materials• Oxygen But neutrons• Epi But neutrons, protons• MCz But special wafer processing• -N type wafers: low resistivity , But SCSI, annealing• P-type wafers: high resistivity Need experience, good annealing
• Decrease drift distance to decrease trapping: Geometry• Thin But inferior at low / medium, no advantage at high fluences• 3-D But high capacitance, need experience
• Reduce trapping with beneficial annealing: Carrier• Collect at (main) junction n-on-p or n-on-n• Collect electrons less trapping, good trapping annealing
33Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
NEG. SPACE CHARGE
POS. SPACE CHARGE
Wafer Scorecard• Materials: Neff = Neff0+g*Φeq
• For p-type: need Neff0 low: high resistivity• For n-type, need Neff0 high: low resistivity
FZ and Mcz data verified for neutron irradiation, some proton data puzzling. Will soon have proton and pion data to check the advantage of MCz.
-1.00
-0.50
0.00
0.50
1.00
1.50
2.00
2.50
FZ-p,n DOFZ-p,n MCz - n? MCz - p EpiSi-p,n
g c[1
0-2cm
-1]
24 GeV Protons reactor neutrons
34Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Peak and Median Q @ 800V
0
5000
10000
15000
20000
25000
1.E+14 1.E+15 1.E+16Fluence [neq/cm2]
Col
lect
ed C
harg
e Q
[e-]
SCIPP n-on-p FZ 2551-7 n irr (prel
SCIPP n-on-p FZ n irr 1000 min ann (prel
Casse NIM A n-on-p FZ p irr
SCIPP n-on-n MCz 2553-11 n irr (prel
SCIPP n-on-n MCz n irr 1000min ann (prel
SCIPP p-on-n MCz 176-7 n irr (prel
SCIPP p-on-n MCz n irr 1000 min ann. (prel
.Casse IEEE07 n-on-p FZ n irr
SCIPP: Binary, Median, StripsCasse: Analog, Peak, Strips
N-on-p strip sensors are sufficiently radiation-hard for the sLHCNo obvious advantage for MCz over FZ. 7500 e after 1016 n/cm2!
Charge Collection in Irradiated SSD
p-on-n Mcz
Long StripsS/N > 10
Short StripsS/N > 10 n-on-p FZ
35Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
N-on-p strip sensors are sufficiently radiation-hard for the sLHC ?
Charge Collection in Upgrade Strips ATLAS bias voltage is constraint to < 500V (cables!).
Peak and Median Q @ 500V
0
5000
10000
15000
20000
25000
1.E+14 1.E+15 1.E+16Fluence [neq/cm2]
Col
lect
ed C
harg
e Q
[e-]
SCIPP n-on-p FZ 2551-7 n irr (prel
SCIPP n-on-p FZ n irr annealed (prel
Casse n-on-p FZ n-irr. IEEE07
Ljubljana n-on-p FZ pad 2551-7 n irr (prel
Ljubljana n-on-p Mcz pad 2552-7 n irr(prel
Casse NIM A n-on-p FZ p irr
SCIPP n-on-n MCz 2553-11 n irr (prel
SCIPP n-on-n MCz n irr ann. 80 min (prel
SCIPP n-on-n MCz n irr ann. 1000 min (prel
.Casse IEEE07 n-onp nirrLong StripsS/N = 10
Short StripsS/N = 10
SCIPP: Binary, Median, StripsCasse: Analog, Peak, StripsLjubljana: Analog, Peak, Pads
36Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
0
5000
10000
15000
20000
25000
0
0.2
0.4
0.6
0.8
1
1.2
0 200 400 600 800 1000
Median Pulse Height vs. Efficiency(n-on-p FZ, 5e14 n) D
G
B
E
Med
ian
Cha
rge
Col
lect
ed [e
-]
Efficiency at 1 fC Threshold
Bias Voltage [V]
Pre-rad
5e14 n
Efficiency Median Q Fluence
Long strips efficient at 1fC threshold
Efficiency vs. Collected Charge • For tracking sensors with
binary readout, the figure of merit is not the collected charge, but the efficiency.
• 100% efficiency is reached at a signal-to-noise ratio of S/N ≈ 10, S/Thr > 2.5
• For long strips (5e14 cm-2) with a signal of about 14ke, the usual threshold of 1fC = 6400 e can be used.
37Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
0
5000
10000
15000
20000
25000
0
0.2
0.4
0.6
0.8
1
1.2
0 200 400 600 800 1000
Median Pulse Height vs. Efficiency(n-on-p FZ, 1e15 n) D
J
B
H
Med
ian
Cha
rge
Col
lect
ed [e
-]
Efficiency at 1 fC Threshold
Bias Voltage [V]
Pre-rad
1e15 n
Efficiency Median Q Fluence
Short strips efficient if threshold can be lowered
Efficiency vs. Collected Charge
• For short strips (1e15 cm-
2) with a signal of about 8ke, the efficiency at 500V is only 70%.
• The threshold needs to be reduced to about 4500 e, i.e. electronics must be designed for a noise of ~700e.
38Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Thin and Pixel sensors
39Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPThin Sensors
300um + 200um 400V
0
5
10
15
20
25
1.E+14 1.E+15 1.E+16Fluence [neq/cm2]
Col
lect
ed C
harg
e
Casse et al 300um + 140um 400V
0
5
10
15
20
25
1.E+14 1.E+15 1.E+16Fluence [neq/cm2]
Col
lect
ed C
harg
e
Simulations 2005:(M. Bruzzi, H. F.-W. Sadrozinski, A. Seiden, NIM A 579 (2007) 754–761)
Measurements 2007(G. Casse, Affolder, P.P. Allport, IEEE 07, N07-06)
Benefits of thin sensors:Can be fully depleted at high fluences (important for p-on-n sensors ?):Thin detectors allow very high effective doping densities in n-type detectors and thus “delay” of inversion to very high fluences. Vbias = q*Neff/(2ε)*x2
Less material (sensor only a fraction of the total material budget)Disadvantage of thin sensors:Signal lower for much of the operationThick sensor can be operated under-depleted (current is lower)
40Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
• Planar electrodes replaced by columns through the wafer- diameter: 10mm, distance: 50 - 100mm
• De-couple deposit and charge collection
3D Detector - Concept n-columns p-columns wafer surface
P- or n-type substrateIntroduced by: S.I. Parker et al., NIMA 395 (1997) 328
p+
------
++++
++++
--
--
++
300
μm
n+
p+
50 μm
------
++ ++++++
----
++
3D PLANARp+
No proton irradiation
No radiation hardness with MIP (only laser)
• Lateral depletion assures Radiation Hardness– Low depletion voltage: Increase active thickness– Short drift distance and fast signal: Decrease
trapping
41Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
growth
substrate
Grain size: ~100-150μm
• large band gap and strong atomic bonds promise fantastic radiation hardness
• low leakage current and low capacitance both give low noise• Ionization energy is high: MIP≈ 2x less signal for same X0 of SI
–Diamond: ∼13.9ke- in 361 μm (140 enc; bare threshold ~1500e)–SI: ∼22.5 ke- in 282 μm
• Grain-boundaries, dislocations, and defects can influence carrier lifetime, mobility, charge collection distance and position resolution
• Available Size ~2 x 6 cm2 (12cm diameter wafer; ~2mm thick)
pCVD Diamond
ATLAS pixel module
hitmap
H.Kagan
M. Mikuz ATLAS Tracker Upgrade Workshop, Valencia 12th-14th Dec. 2007
42Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPPixel Sensor Efficiency: Signal/Threshold
Threshold “bare” threshold is set as low as the system permitsBare Threshold ~ 1500e for Diamond
~ 2500e for Silicon(why does it not scale with noise?)
In order to fit the hit into the beam crossing of 20 ns, the signal has to exceed the threshold by the “overdrive”.In-time Threshold = bare threshold + overdrive
Sensor Threshold [e] Bare Overdrive In-time
Required Signal [e]
Si planar 2500 1300 3800 7600
Si 3D 2500 (?) 1800 4300 8600
Diamond 1500 800 2300 4600
Signal needs to be > 2x In-time Threshold:
Overdrive scales with noise
Signal is Landau (here Diamond):
Efficiency requires Signal/Threshold > 2
O. Roehne ATLAS Tracker Upgrade Workshop, Valencia 12th-14th Dec. 2007
C. daVia ATLAS B-Layer Workshop, CERN Sep. 28-30. 2007
43Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Peak Q @ ~ 900V
0
5000
10000
15000
20000
25000
1.E+15 1.E+16 1.E+17Fluence [neq/cm2]
Col
lect
ed C
harg
e Q
[e-]
Casse NIM A n-on-p FZ p irr
.Casse IEEE07 n-on-p FZ n irr
Diamond
3D 2E
Pixel S/N -> Signal/Threshold
Need comparisons of CCE(V) with (m.i.p.) and S/N after 1016neq/cm2 of protonsPreliminary results show marginal performance for 3D.Need to optimize FEE?
Diamond4600e
Planar Si7600e
3D8600e
Planar n-on-p adequate for all but innermost Pixel Layer
Planar (protons!) and diamond are not adequate for innermost Pixel Layer
44Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
RD50 provided the bases for understanding the use of tracking detectors at the LHC upgrade. No magic new semiconductor tracking material found.
Paradigm change in Si at fluence of about 1015 neq/cm2:At lower fluences depletion of sensors important. MCz seems to be favored, at least for neutronsStrong annealing in n-type FZ, little annealing in p-type and n-type MCz.Collect electrons in n-on-p (cheaper than n-on-n) because of high mobility.Collected charge almost independent of wafer type except n-type FZ.At higher fluences trapping dominatesbut high resistivity MCz p-type could have depletion advantage at very high fluences.
A straw-man tracker layout at the sLHC looks like this: Planar n-on-p sensors with long and short strips at large and medium radius and the outer pixels.At small radii, the pixel sensors might need to be made from 3D sensors, but we need more high fluence proton data with planar p-type MCz.
Conclusions
45Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Thanks to
the foundries and institutes who supplied sensors,
RD50 collaborators in Ljubljana, Liverpool, Louvain, CERN, Karlsruhe, PSI, UCSC for carrying out the irradiations, and analyzing the data.
the many students who spend their evenings and nights taking and analyzing the data.
Acknowledgments
46Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
ATLAS Tracker Upgrade Workshop Valencia Dec 11-14 2007
http://indico.cern.ch/conferenceTimeTable.py?confId=21398
ATLAS Upgrade Website
http://indico.cern.ch/categoryDisplay.py?categId=350
RD50 Website
http://rd50.web.cern.ch/rd50/
LHC Machine Operations Website
http://lhc-operation.web.cern.ch/lhc-operation/
References
47Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPPTo keep ATLAS running more than 10 years the inner tracker will have to go ... (Current tracker designed to survive up to 730 fb-1 ≈ 10Mrad in strip detectors)For the luminosity-upgrade the new tracker will have to cope with:
• much higher occupancy levels 3000 fb-1 • much higher dose rates ~ 300Mrad• same performnace
To build a new tracker for 2015, major R&D program already needed.Steering group (lead Nigel Hessey) and several working groups. Formal proposal submittal structure.
Timescales:• R&D leading into a full tracker Technical Design Report (TDR) in 2010• Construction phase to start immediately TDR completed and approved.
The intermediate radius barrels are expected to consist of modules arranged in rows with common cooling, power, clocking and cooling.The TDR will require prototype super-modules/staves (complete module rows as an integrated structure) to be assembled and fully evaluatedAll components will need to demonstrate unprecedented radiation hardness
ATLAS Inner Detector Replacement
48Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
1/9/2008 48
many strategies to follow up, new techniques, many unknown parameters …. needs first beam
25 ns vrs 50 ns …. 12ns,75ns gone!
LHC nominal and ultimate L strategy -> new large aperture triplet
Early separation scheme, magnets inside the detector … small crossing angle
Beam-beam effects : unclear can be tolerated, wait for LHC beam
Luminosity leveling : new …. more fill lifetime, less peak luminosity
Crab cavities,… : new magic technology, need substantial R&D
ATLAS message is always the same:
we prefer more constant luminosity, less we prefer more constant luminosity, less pile up at the start of the run, higher luminosity at the pile up at the start of the run, higher luminosity at the endend
49Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Pixels (50 μm × 400 μm): 3 barrels, 2×3 disks 5cm < r < 15cm• Pattern recognition in high occupancy region• Impact parameter resolution (in 3d)Radiation hard technology: n+-in-n Silicon technology, operated at -6°CStrips (80 μm × 12 cm) (small stereo angle): “SCT” 4 barrels, 2×9 disks• pattern recognition 30cm < r < 51cm• momentum resolutionp-strips in n-type silicon, operated at -7°CTRT 4mm diameter straw drift tubes: barrel + wheels 55cm < r < 105cm• Additional pattern recognition by having many hits (~36)• Standalone electron id. from transition radiation
Current Inner Tracker Layout
r=30cm
0.61%
Mean Occupancy in Innermost Layer of Current SCT
Pixels: 1.8 m2, ~80M channels
SCT: 61 m2, ~6.3M channels
TRT straws: ~400k channels
ID TDR
50Hartmut F.-W. Sadrozinski, SLAC AIS Jan 9, 2008
SCIPPSCIPP
Pixels:Short (2.4 cm) μ-strips (stereo layers):Long (9.6 cm) μ-strips (stereo layers):
r=5cm, 12cm, 18cm, 27cmr=38cm, 49cm, 60cmr=75cm, 95cm
z=±40cmz=±100cmz=±190cm
Strawman 4+3+2New SLHC Layout Implications
Including disks this leads to:Pixels: 5 m2, ~300,000,000 channelsShort strips: 60 m2, ~28,000,000 channelsLong strips: 100 m2, ~15,000,000 channels
layer radius40 50 60 70 80 90 100
mea
n o
ccu
pan
cy
0.006
0.008
0.01
0.012
0.014
0.016
mean occupancy per layer (Muon+pileup)
1.6%
1.2%
0.8%
Short and Long Strip Occupancy Only LO MC (Pythia) . May
need to include ×2 safety factor?
J. Tseng
J. Mechnich