1. CMS Pixel Simulation Vincenzo Chiochia
Quarter pixel: Electrostatic potential
Vincenzo Chiochia
Physik Institut der Universität Zürich-Irchel
CH-8057 Zürich (Switzerland)
PIXEL 2005 Workshop
Chuzenji Lake, Nikko (Japan)
November 7-11, 2005

2. Outline The CMS pixel detector
Radiation damage and impact on reconstruction
Beam test data and detailed sensor simulation
Physical modelling of radiation damage:
Models with a constant space charge density
Double-trap models
Fluence and temperature dependence
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

3. The CMS pixel detector 3-d tracking with about 66 million channels
Barrel layers at radii = 4.3cm, 7.2cm and 11.0cm
Pixel cell size = 100x150 µm2
704 barrel modules, 96 barrel half modules, 672 endcap modules
~15k front-end chips and ~1m2 of silicon
~1 m
0.3 m
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

4. The LHC radiation environment
sppinelastic = 80 mb
L = 1034 cm-2s-1
4 cm layer F=3x1014 n/cm2/yr
Fluence decreases quadratically with the radius
Pixel detectors = 4-15 cm mostly pion irradiation
Strip detectors = cm mostly neutron irradiation
What is the sensors response after
the first years of operation?
Fluence per year at full luminosity
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

5. Radiation effects r-f plane r-z plane  B field no B field
charge
width
charge
width
after irradiation
charge
charge
after irradiation E E
 B field
no B field
Irradiation modifies
the electric field profile:
varying Lorentz deflection
Irradiation causes
charge carrier trapping
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

6. Prototype sensors for beam tests
n-in-n type with moderated p-spray isolation, biasing grid and punch through structures (producer: CiS, Germany)
285 μm thick <111> DOFZ wafer
125x125 mm2 cell size, 22x32 pixel matrix
Samples irradiated with 21 GeV protons at the CERN PS facility
Fluences: Feq=(0.5, 2.0, 5.9)x1014 neq/cm2
Annealed for three days at +30º C
Bump bonded at room temperature to non irradiated front-end chips with non zero-suppressed readout, stored at -20ºC
125 mm2
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

7. Beam test setup Silicon strip beam telescope:
50 μm readout pitch,~1 μm resolution
Cooling circuit
T =-30 ºC or -10ºC
CERN Prevessin site
H2 area
3T Helmoltz magnet
beam:
150 GeV p
B field
pixel sensor
support
Data collected at CERN in
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

8. 2 years LHC low luminosity 2 years LHC high luminosity
Charge collection measurement
Charge collection was measured using cluster profiles in a row of pixels illuminated by a 15º beam and no magnetic field
Feq = 5×1013 n/cm2
2 years LHC low luminosity
n+side
p-side
2 years LHC high luminosity
AGEING
Feq = 2×1014 n/cm2
Feq = 6×1014 n/cm2
½ year LHC low luminosity
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

9. Detector simulation PIXELAV ISE TCAD 9.0
Double traps
models
(DESSIS)
3-D Electric
field mesh
ISE TCAD 9.0
Charge
deposit
Charge
transport
Trapping
Trapping times
from
literature
Electronic
response +
data formatting
ROC+ADC
response
ROOT
Analysis
PIXELAV
M.Swartz, Nucl.Instr. Meth. A511, 88 (2003);
V.Chiochia, M.Swartz et al., IEEE Trans.Nucl.Sci. 52-4, p.1067 (2005).
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

10. The classic picture Neff=ND-NA<0 Based on C-V measurements!
after type inversion
After irradiation the sensor bulk becomes more acceptor-like
The space charge density is constant and negative across the sensor thickness
The p-n junction moves to the pixel implants side
Sensors may be operated in “partial depletion”
Neff=ND-NA<0
Based on C-V measurements! -
…is all this really true?
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

11. Models with constant Neff
F = 6x1014 n/cm2
A model based on a type-inverted device with constant space charge density
across the bulk does not describe the measured charge collection profiles
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

12. (Shockley-Read-Hall statistics):
Two traps models
EConduction
acceptor
EC eV
Electron
traps
1.12 eV
donor
EV+0.48 eV
Hole
traps
EValence
EA/D = trap energy level fixed
NA/D = trap densities extracted from fit
se/h = trapping cross sections extracted from fit
Model parameters
(Shockley-Read-Hall statistics):
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

13. The double peak electric field
a) Current density
c) Space charge density
n+p
junction
np+
-HV
p-like
n-like
There are
P-N junctions at both
sides of the detector
detector depth
detector depth
b) Carrier concentration
d) Electric field
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

14. Model constraints The two-trap model is constrained by:
Comparison with the measured charge collection profiles
Signal trapping rates varied within uncertainties
Measured dark current
Typical fit iteration: (8-12h TCAD) + (8-16h PIXELAV)xVbias + ROOT analysis
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

15. Fit results Electric field and space charge density profiles Data
--- Simulation
F1=6x1014 n/cm2
NA/ND=0.40
sh/se=0.25
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

16. Scaling to lower fluences
F3=0.5x1014 n/cm2
NA/ND=0.75
sAh/sAe=0.25
sDh/sDe=1.00
Near the ‘type-invesion’ point: the double peak structure is still visible in the data!
Profiles are not described by thermodynamically ionized acceptors alone
At these low bias voltages the drift times are comparable to the preamp shaping time (simulation may be not reliable)
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

17. Space charge density Space charge density uniform before irradiation
Current conservation and non uniform carrier velocities produce a non linear space charge density after irradiation
The electric field peak at the p+ backplane increases with irradiation
n-type
p-type
sensor depth (mm)
V.Chiochia, M.Swartz,et al., physics/
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

18. Temperature dependence
F2=2x1014 n/cm2
T=-25º C
Comparison with data collected at lower temperature T=-25º C.
Use temperature dependent variables:
recombination in TCAD
variables in PIXELAV (m, D, G)
The double-trap model is predictive!
F1=6x1014 n/cm2
T=-25º C
V.Chiochia, M.Swartz,et al., physics/
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

19. Lorentz deflection tan() linear in the carrier mobility (E):
Switching on the magnetic field
tan() linear in the carrier mobility (E):
LHC startup
2 years LHC low luminosity
2 years LHC high luminosity
The Lorentz angle can vary a factor of 3 after heavy irradiation:
This introduces strong non-linearity in charge sharing
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

20. Conclusions and plans
After heavy irradiation trapping of the leakage current produces electric field profiles with two maxima at the detector implants. The space charge density across the sensor is not uniform.
What is the meaning of Vdep, depletion depth and type inversion? Measurements reflecting the electric field profile (e.g. TCT, CCE, long clusters etc.) are preferable to C-V characterization to understand radiation damage in running detectors
A physical model based on two defect levels can describe the charge collection profiles measured with irradiated pixel sensors in the whole range of irradiation fluences relevant to LHC operation
Our model is an effective theory: e.g. in reality there are several trap levels in the silicon band gap after irradiation. However, it is suited for applications related to silicon detector operation at LHC.
We are currently using the PIXELAV simulation to develop hit reconstruction algorithms optimized for irradiated pixel sensors.
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

21. References PIXELAV simulation: Double-trap model:
M.Swartz, “CMS Pixel simulations”, Nucl.Instr.Meth. A511, 88 (2003)
Double-trap model:
V.Chiochia, M.Swartz et al., “Simulation of Heavily Irradiated Silicon Pixel Sensors and Comparison with Test Beam Measurements”, IEEE Trans.Nucl.Sci. 52-4, p.1067 (2005), eprint:physics/
V. Eremin, E. Verbitskaya, and Z. Li, “The origin of double peak electric field distribution in heavily irradiated silicon detectors”, Nucl. Instr. Meth. A476, pp (2002)
Model fluence and temperature dependence:
V.Chiochia, M.Swartz et al., “A double junction model of irradiated pixel sensors for LHC”, submitted to Nucl. Instr. Meth., eprint:physics/
V.Chiochia, M.Swartz et al., “Observation, modeling, and temperature dependence of doubly-peaked electric fields in silicon pixel sensors”, submitted to Nucl. Instr. Meth., eprint:physics/
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

22. Backup slides

23. EVL models
F1=6x1014 n/cm2
100% observed
leakage current
s=1.5x10-15 cm2
30% observed
leakage current
s=0.5x10-15 cm2
The EVL model based on double traps can produce large tails
but description of the data is still unsatisfactory
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

24. E-field profiles Constant space charge density Double trap model
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

25. Scaling to lower fluences
Preserve linear scaling
of Ge/h and of the current
with Feq
F2=2x1014 n/cm2
NA/ND=0.68
sAh/sAe=0.25
sDh/sDe=1.00 !
Not shown: Linear scaling of trap densities does not describe the data!
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

26. Temperature dependence
F1=6x1014 n/cm2
T=-10ºC
T=-25ºC
Charge collection profiles depend on temperature
T-dependent recombination in TCAD and T-dependent variables in PIXELAV (me/h, Ge/h, ve/h)
The model can predict the variation of charge collection due to the temperature change
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

27. Lorentz deflection Lorentz angle Electric field
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

28. Lorentz angle vs bias
‘Effective’ Lorentz angle as function of the bias voltage
Strong dependence with the bias voltage (electric field)
Weak dependence on irradiation
This is a simplified picture!!
Magnetic field = 4 T
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

29. ISE TCAD simulation
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

30. PIXELAV simulation
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

31. SRH statistics
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

32. SRH generation current
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005

33. Test beam setup  Magnetic field = 3 T pixel sensor or
Four modules of silicon strip detectors
Beam telescope resolution ~ 1 m
Sensors enclosed in a water cooled box (down to -30ºC)
No zero suppression, unirradiated readout chip
Setup placed in a 3T Helmoltz magnet 
PIN diode trigger
V. Chiochia – CMS Pixel Simulation – Vertex Chuzenji Lake, Nikko, Japan , November 7-11, 2005