6 th Trento Workshop , March 3 rd , 2011 Advances in Development of 3D Silicon Detectors with CMS Pixel Readout Ozhan Koybasi 1 , E. Alagoz 1 , A. Krzywda 1 K. Arndt 1 , D. Bortoletto 1 , I. Shipsey 1 , G. Bolla 1 , T. E. Hansen 2 , A. Kok 2 , T. A. Hansen 2 , N. Lietaer 2 , G. U. Jensen 2 , R. Riviera 3 , L. Uplegger 3 , and S. W. L. Kwan 3 1 Physics Department, Purdue University, West Lafayette, IN 47907‐ 2036 USA 2 SINTEF, SINTEF MiNaLab, Blindern, 0314 Oslo, Norway 3 Fermilab, Batavia, IL 60510‐5011 USA 1
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
6th Trento Workshop , March 3rd , 2011
Advances in Development of 3D Silicon Detectors with CMS Pixel Readout
Ozhan Koybasi1, E. Alagoz1, A. Krzywda1 K. Arndt1, D. Bortoletto1, I. Shipsey1, G. Bolla1, T. E. Hansen2, A. Kok2, T. A. Hansen2, N. Lietaer2, G. U. Jensen2, R. Riviera3, L. Uplegger3, and S. W. L.
Kwan3
1Physics Department, Purdue University, West Lafayette, IN 47907 2036 USA‐2SINTEF, SINTEF MiNaLab, Blindern, 0314 Oslo, Norway
3Fermilab, Batavia, IL 60510 5011 USA‐
2
3D Silicon Detectors
ionizing particle
300µm
n+
p+
e
h
depletion
p-type
p+ n+
50µm
p-type
depletion
p+ and n+ electrodes are arrays of columns that penetrate into the bulk
Lateral depletion
Charge collection is sideways
Superior radiation hardness due to smaller electrode spacing: - smaller carrier drift distance - faster charge collection - less carrier trapping - lower depletion voltage
No guard rings required (active edge) Reduced dead volume
Higher noise
Complex, non-standard processing
PLANAR: 3D:
p+ active edge
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
3
3D CMS Pixel Layouts
2 readout electrodes per pixel (2E)4 readout electrodes per pixel (4E)
Lower depletion voltage
Faster response
Lower signal loss
100 µm
150 µm
p+
n+
Lower electronic noise
Less dead volume
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
n+: readout p+: bias
4
3D Sensor Wafer Layout
ATLAS FE-I4
ATLAS FE-I4
Medipix
Medipix
CMS
ATLAS FE-I3
4E
2E
2E4E 4E
4E
2E
2E 4E
4E 2E95 96 97 98 99 100
101 102 103 104 105 106
107 108 109 110 111 112
113 114 115 116 117 118
119 120 121 122 123
1E 1E5E
5E
2E
2E 2E
2E
3E
3E 3E
3E 3E
4E
4E
4E 4E 2E
3E4E 2E 2E4E 4E
2E 4E 2E 4E 2E
1
2 3
4 6
7 8 9
10 11
Includes ATLAS, CMS, and MediPix devices
CMS chips cover ~15% of wafer area
p-type silicon wafers with resistivity > 10 kΩ.cm
200 µm and 285 µm thick wafers processed in parallel
Well-performing sensors mostly located near the center of the wafer
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
5
Fabrication of 3D detectors at SINTEF
P-spray inter-electrode isolation
Wafer bonded to a support wafer by Si-Si fusion bonding
Columns and active edge etched by deep reactive ion etching (DRIE)
Holes and active edge trench filled with polysilicon and doped
Column diameter: 14 µm & active edge width: 5 µm
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
6
3D Detector Module Assembly
Sensor
Cooling tubes
Carbon fiber plate
Bias wire
VHDI
Fan-Out Opening blocked
Support waferSensor
ROCBump bonds
VHDIBase
plateAdhesive
Bias pad
Bias wire
Au plateCeramic plate
Wire
bond
3D sensors bump-bonded to CMS Pixel Readout Chip (PSI46v2) via Pb-Sn (IZM) or In bumps (SELEX)
Assembly of 3D modules was similar to the production FPIX modules except HV bias wiring
Small opening was made through carbon fiber plate at the end of the VHDI
Gold-ceramic piece was used as intermediate pad for HV wire-bonding between sensor and Fan-Out
Carbon-fiber plate was inverted on wire-bond machine to make HV bias wiring
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
7
I-V Characteristics
0 20 40 60 80 100 120 1401E-8
1E-7
1E-6
1E-5
1E-4
Cur
rent
(A)
Voltage (V)
On wafer After wire-bondingWB5_2E_2 WB5_4E_8 WB2-16_4E_5 WB2-16_2E_6
0 20 40 60 80 1000.0
0.5
1.0
1.5
2.0
2.5
3.0
Cur
rent
(µA
)
Voltage (V)
2E 4E
Measurement TCAD simulation
Measurements done at room temperature (no cooling)
Post-assembly leakage current is lower than wafer level leakage current
Predictions of simulations are in a reasonable agreement with measurement results
Wafer level measurements were done with a temporary metal layer connecting all n-type columns. This forms an MOS structure introducing extra surface leakage current.
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
8O. Koybasi et al., 16th RD50 Workshop ,Barcelona, June 2nd , 2010
Bump-Bond Quality
Tests done by “Modified External Calibration”
No radioactive or light source used
All sensors showed perfect bump-bond quality except one
2E Wafer B5 #2 4E Wafer B2 16 #5‐
9O. Koybasi et al., 16th RD50 Workshop ,Barcelona, June 2nd , 2010
Noise
Noise Distribution Entries : 4160 Mean : 4.52395 RMS : 0.546284
Sensor:
WB5_2E_2 (2E design, 285 µm thick)
Vbias=-40V
1 VCAL= 65.5 electrons
100µm. . .
.
.
.
.
.
.
300µm
200µm
150µm
10
Noise and Threshold
FPIX planar BPIX planar 3D with 2E configuration
3D with 4E configuration
110 155 ~450
FPIX planar BPIX planar 3D with 2E configuration
3D with 4E configuration
2870 2910 3200-5500 3200-5500
Noise:
Threshold:
250-300
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
11
Gain Calibration
Gain curve of a single pixel
Gain determined from slope and pedestal determined from the offset of linear region by fitting the curve with the function:
2E configuration, 285 µm thick
Mean : 493.63 RMS : 11.504
Mean : 1.49061 RMS : 0.125423
ADC = p3+p2*tanh(p0*VCAL-p1)
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
12
Efficiency Studies with Test Beam at FNAL
Charge Distribution Entries : 17834 Mean : 24542 RMS : 3887 Vbias=-40V
ADC to electron conversion: VCAL (DAC) = ADC x gain – pedestal Charge (e-) = VCAL x 65.5 – 410
Charge distribution does not have a very good convoluted Landau and Gaussian shape Gain calibration needs improvement
2E configuration, 285 µm thick
120 GeV protons
No magnetic field
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
13
Efficiency Simulations
100 101 102 103 1040.01
0.1
1
10
MIP through n+ (readout) MIP through p+ (bias)C
olle
cted
Cha
rge
(ke- )
Fluence (1012neq
/cm2)100 101 102 103
9
10
11
12
13
14
15
16
17
Col
lect
ed C
harg
e (k
e- )
Fluence (1013neq
/cm2)
Minimum Ionizing Particle (MIP):
Efficiency= ~100% before irradiationEfficiency= ~ 35% before irradiationEfficiency= ~ 55% before irradiation
Travels vertically through substrate thicknessTrack generates 80 electron hole pairs per micron Gaussian lateral profile with 1µm standard deviation > 99.99% of charge generated within a radius of ~2.1µm
MIP through midway between n+ and p+ columns
Average Efficiency = for before irradiation
x
x
x
p+
n+
2E 4E
97.8%95.6%
Substrate thickness = 200 µm
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
14
Efficiency Studies
0 10 20 30 40 50 60 70 80 90 10014
16
18
20
22
24
26
MPV Average PeakC
harg
e (k
e- )
Voltage (V)
2E configuration 285 µm thick
Test beam
Prediction of simulation:
Collected charge = if MIP charge is 80 electron-hole pairs per micron22.3 ke-
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
15
Efficiency Studies with Radioactive Source (90Sr)
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
16
Summary and Next Plans
O. Koybasi et al., 6th Trento Workshop , March 3rd , 2011
Promising results
Noise: 250-300 electrons for 2E (S/N=~90), ~450 electrons for 4E (S/N=~55)
Threshold: 3200-5500 electrons
Bump-bonding with In at SELEX is problematic
Gain calibration needs to be improved
Resolution studies in progress
3D sensors irradiated by 1.2x1014, 2x1015, 4.7x1015 (1MeV neq/cm2)