Modeling Radiation Damage to Pixel Sensors in the ATLAS Detector M. - - PowerPoint PPT Presentation
PIXEL2018 - Academia Sinica, Taipei Modeling Radiation Damage to Pixel Sensors in the ATLAS Detector M. Bomben, LPNHE & UPD Paris on behalf of the ATLAS collaboration M. Bomben - Pixel 2018, 10-14 December, Academia Sinica, Taipei,
1
2
CERN LHC is the largest and most powerful particle accelerator ever built It provides proton-proton collisions at energies up to √s = 13 TeV LHC design luminosity was 1x1034 cm-2s-1 Design value has been widely exceeded!
3
CERN LHC is the largest and most powerful particle accelerator ever built It provides proton-proton collisions at energies up to √s = 13 TeV LHC design luminosity was 1x1034 cm-2s-1 Design value has been widely exceeded! Large dataset integrated over first 2 LHC Runs: > 180 fb-1
4
CERN LHC is the largest and most powerful particle accelerator ever built It provides proton-proton collisions at energies up to √s = 13 TeV LHC design luminosity was 1x1034 cm-2s-1 Design value has been widely exceeded! Large fluence integrated over first 2 LHC Runs: > 9x1014 neq/cm2 by the innermost pixel layer
5 200 400 600 800 1000 1200 Days Since Start of Run 2 2 4 6 8 10 ]
2
/ cm
eq
n
14
Total Fluence at z = 0 [10 50 100 150 ]
Run 2 Delivered Luminosity [fb
Preliminary ATLAS
Updated Oct. 25, 2018
IBL B-layer Layer 1 Layer 2
6
1 . 4 m 4 Pixel barrel layers 3 Outermost: 250 µm thick 50x400 µm2 pitch Innermost layer: IBL Inserted in Run2 200 µm thick 50x250 µm2 pitch
7
1 . 4 m 4 Pixel barrel layers 3 Outermost: 250 µm thick 50x400 µm2 pitch Innermost layer: IBL Inserted in Run2 200 µm thick 50x250 µm2 pitch P l a n a r p i x e l n
s e n s
s e v e r y w h e r e b u t a t h i g h η * i n I B L w h e r e n
e l 3 D n
a r e u s e d
*outside tracking volume
cluster size for IBL
bias voltage and adjust threshold to mitigate the negative trend
2 4 6 8 10 ]
2
/cm
eq
n
14
10 × [1 Φ 2 4 6 8 10 ]
2
/cm
eq
n
14
10 × [1 Φ
20 40 60 80 100 120 140 160 180
]
Run-2 Delivered Luminosity [fb
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
] or <cluster size> [pixels]
2
cm
<dE/dx> [MeV g
150 V → HV 80
Preliminary ATLAS Pixel Data2016 IBL Data 2017 Data 2018
<dE/dx> φ Cluster size Cluster size z HV=80(150) V
HV=350 V
HV=400 V
8
9
Include all this in ATLAS MonteCarlo
Charge carriers will drift toward the collecting electrode due to electric field, which is deformed by radiation damage. Their path will be deflected by magnetic field (Lorentz angle) and diffusion. Due to radiation damage they can be trapped and induce/screen a fraction of their charge (Ramo potential). Total induced charge is then digitized and clustered. *Digitization happens after simulated charge deposition and before space point reconstruction
10
Include all this in ATLAS MonteCarlo Now ready!
Charge carriers will drift toward the collecting electrode due to electric field, which is deformed by radiation damage. Their path will be deflected by magnetic field (Lorentz angle) and diffusion. Due to radiation damage they can be trapped and induce/screen a fraction of their charge (Ramo potential). Total induced charge is then digitized and clustered.
11
fluence trapping constant time travelled initial charge location E-field final depth Ramo potential induced charge Lorentz angle thermal diffusion final charge location
per e/h per geometry per condition
Start End
12
30 − 20 − 10 − 10 20 30 Distance along stave [cm] 1 2 3 4 5 6 7 ]
/fb
2
/cm
eq
Si 1 MeV n
12
Absolute fluence [10
Preliminary ATLAS
= 13 TeV s
20 40 60 80 100 120 (z=0) [%] Φ (z) / Φ Relative data fluence
Insertable B-layer (IBL) Predicted by Pythia (A2) + FLUKA Predicted by Pythia (A3) + FLUKA Predicted by Pythia (A3) + Geant4
π Predicted by Pythia (A3) + Geant4 (n + p + Extracted from Hamburg Model + Leakage Currents
Fluence prediction taken from FLUKA & Pythia FLUKA prediction validated with leakage current and Hamburg model* simulation Ø 15% difference in the central region Trapping constants from literature**: Ø βe = (4.5±1.5)x10-16 cm2/ns Ø βh = (6.5±1.5)x10-16 cm2/ns * M. Moll, DESY-THESIS-1999-040
** ATLAS pixel coll., JINST 3 (2008) P07007
13
Radiation damage induced defects deform the electric field distribution in the bulk We use TCAD simulation tools to make predictions of electric field in the bulk A 2 trap model due to CMS collaborators* has been used with Silvaco tools**
*V. Chiochia et al., Nucl. Instr. and Meth A 568 (2006) 51-55 ** https://www.silvaco.com/products/tcad.html
50 100 150 200 m] µ Bulk Depth [ 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
= 0 Φ
2
/cm
eq
n
14
10 × = 1 Φ
2
/cm
eq
n
14
10 × = 2 Φ
2
/cm
eq
n
14
10 × = 5 Φ
Vbias = 150 V
Model chosen because:
Main feature: double peak electric field
14
Fluence
(1014 neq/cm2)
EA
T (eV)
±0.4% ED
T (eV)
± 0.4% NA
(1014 cm−3)
± 10% ND
(1014 cm−3)
± 10% σA,D
e
& σD
h (10−15 cm2)
± 10% σA
h (10−15 cm2)
± 10% 1 EC-0.52eV EV +0.48eV 3.6 5 6.60 1.65 2 6.8 10 5 14 34
TCAD radiation damage model parameters come with no uncertainties So we had to explore the sensitivity of electric field on each defect parameter:
cross sections Trends are compatible with expectations
m] µ Bulk Depth [ 20 40 60 80 100 120 140 160 180 [V/cm]
zE 2000 4000 6000 8000 10000 12000 14000
Nominal + 10%
int Ag
g
2/cm
eqn
1410 × = 1 Φ = 80 V
biasV m] µ Bulk Depth [ 20 40 60 80 100 120 140 160 180 [V/cm]
zE 2000 4000 6000 8000 10000 12000 14000
Nominal energy + 0.4% energy - 0.4%
2/cm
eqn
1410 × = 1 Φ = 80 V
biasV m] µ Bulk Depth [ 20 40 60 80 100 120 140 160 180 [V/cm]
zE 2000 4000 6000 8000 10000 12000 14000
Nominal + 10%
eσ
σ
2/cm
eqn
1410 × = 1 Φ = 80 V
biasV m] µ Bulk Depth [ 20 40 60 80 100 120 140 160 180 [V/cm]
zE 2000 4000 6000 8000 10000 12000 14000
Nominal + 10%
hσ
σ
2/cm
eqn
1410 × = 1 Φ = 80 V
biasV
15
02/07/2015 01/01/2016 01/07/2016 31/12/2016
Date 20 40 60 80 100 120 140 [V]
depl
V
Preliminary ATLAS IBL
Hamburg Model Simulation Simulation Uncertainty Data with Bias Voltage Scan
Annealing not modeled in TCAD Effective correction to TCAD: rescale defects concentration in TCAD to match the average (constant) space charge concentration predicted by Hamburg Model Hamburg model predictions based on bias voltage scans
20 40 60 80 100 120 140 160 180 m] µ Bulk Depth [ 8000 − 6000 − 4000 − 2000 − 2000 4000
9
10 × ]
3
Space Charge [e/cm
2
/cm
eq
n
14
10 × = 2 Φ = 80 V
bias
V
Chiochia Hamburg TCAD with eff. annealing
IBL stayed cold most of the time è small correction More important effect for B-Layer
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Ramo Potential 0.05 0.1 0.15 0.2 0.25 0.3 0.35 m µ X / 200 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 m µ Depth / 200
+
m n µ 200
e-
16
Charge drift towards collecting electrode They induce larger and larger current the closer they get to the electrode If trapped only a fraction of the total charge will be induced Trapping position is stochastically determined, based on fluence and voltage conditions The final signal is calculated in a 3x3 pixels matrix thanks to the Ramo potential
2D slice of 3D Ramo potential calculated using TCAD simulations B
17
m] µ Starting Depth in z [ 50 100 150 Tangent Lorentz Angle 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 Electrons Holes Unirradiated
2
/cm
eq
n
14
10 × 1
2
/cm
eq
n
14
10 × 2
2
/cm
eq
n
14
10 × 5
Pixel Preliminary ATLAS
+
m n µ 200
(a)
Electric field profile no longer shows linear dependence on bulk depth
50 100 150 200 m] µ Bulk Depth [ 2000 4000 6000 8000 10000 12000 14000
= 0 Φ
2
/cm
eq
n
14
10 × = 1 Φ
2
/cm
eq
n
14
10 × = 2 Φ
2
/cm
eq
n
14
10 × = 5 Φ
It is now even more important to model the Lorentz angle depth-dependence
18
]
Integrated Luminosity [fb 10
2
10 Fraction of Charge 0.6 0.7 0.8 0.9 1
Data 400 V Data 350 V Data 150 V Data 80 V Standalone Simulation 400 V Standalone Simulation 350 V Standalone Simulation 150 V Standalone Simulation 80 V
ATLAS Preliminary
IBL planar modules
]
2
/cm
eq
n
14
Fluence [10 1 10
CCE for IBL across its lifetime Simulation uncertainties: Horizontal error bars include uncertainties on luminosity to fluence conversion (15%) Vertical error bars include uncertainties from the TCAD radiation damage model Data uncertainties Horizontal error bars include luminosity unc. (2%) Vertical error bars include calibration drift effects Good agreement with data but large uncertainties In the future collision data can be used to further constrain the radiation damage model
End 2016 End 2017 End 2018
19
]
Integrated Luminosity [fb 10
2
10 [mrad]
L
θ Δ 10 20 30 40 50 60 70 80 ]
2
/cm
eq
n
14
10 × fluence [
1 −
10 1
syst ⊕ stat σ Allpix Simulation 80 V - Allpix Simulation 80 V Petasecca Model Data 80 V
ATLAS Preliminary IBL planar modules
The trend of increase of Lorentz angle with luminosity is robust Models predicting no double peak in electric field fail at reproducing increase of L.A. with luminosity
0.1 − 0.1 0.2 0.3 0.4 0.5 Incidence Angle [rad] 1 1.2 1.4 1.6 1.8 2 Mean Transverse Cluster Size Standalone Simulation Insertable B-layer Data
Pixel Preliminary ATLAS
Bias Voltage 150 V | < 0.6 η |
2
/cm
eq
1 MeV n
14
10 × 2 ≈ Φ
Lorentz angle is extracted from a fit to the cluster size vs track incident angle
Petasecca et. al, IEEE TNS 53 (2006) 2971
20
Bias Voltage [V] 200 400 600 800 1000 ToT [BC] 1 2 3 4 5 6 7 8 9 10
data - end 2017 (end 2017)
2
/cm
eq
n
14
=6 10 φ Standalone Simulation: (end 2018)
2
/cm
eq
n
14
=8.7 10 φ Standalone Simulation:
ATLAS Preliminary IBL planar modules
Both data and simulation charge eements in both shape
point to avoid under depletion
End 2017 End 2018
Standalone simulation to predict MPV of the fitted Landau distribution
fluence Good agreements in both shape and plateau position This confirms that both the electric field and the trapping time are correctly reproduced in
Predictions now used to determine desired bias voltage during LHC Run3 for all pixel layers
21
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 20 40 60 80 100 120 140 160 180
Damage No Radiation Run 2 Beginning
~15 fb
Near End
Near End 2017 End of 2018 End of
0.6 0.8 1 1.2 1.4 1.6 / g]
2
Cluster dE/dx [MeV cm IBL B-layer Layer 1 Layer 2
Simulation Preliminary ATLAS
> 1 GeV
track T
p 2 4 6 8 10
]
2
/cm
eq
n
14
[1 MeV 10 Fluence
100 200 300 400 500
Bias Voltage [V]
Digitizer can be used to make predictions on fundamental observables Energy loss per layer for tracks with pT > 1 GeV Several scenarios considered, in terms of
N.B. other parameters (thr., tuning) fixed
22
23
24
25
26