1. Simulation studies of isolation techniques for n-in-p silicon sensors
HPK Workshop , Jan – Tracker Upgrade Sensor
Simulation studies of isolation techniques
for n-in-p silicon sensors
on behalf of the Simulation Working Group
INSTITUT FÜR EXPERIMENTELLE KERNPHYSIK

2. p-stop and p-spray strip isolation
Needed to isolate adjacent n-strips due to fixed and trap charges in
Si/SiO2 interface  create an inversion layer on the n-side
Impact on sensor performance !?: CCE, Ccoup, Cint, max E-Field, breakdown Voltage…
We consider 2 p-stop configurations: Atoll und Common
 impact on Signal-to-Noise (see Manfred Valentan, HEPHY)
Fig.: detail of Layout GDS file for masking shop showing the two p-stop
techniques;
left: p-stop atoll, and right: p-stop common
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

3. P-spray isolation technique
Uniform layer, covering whole wafer; no additional mask
High electric fields at n+ implant edges, as the p-spray layer is in direct contact to the n+ strips!
Simulation studies consider electrical fields depending on doping concentration, fixed oxide charge at the bulk/oxide interface, interstrip capacitance…
P-spray
N+ strips
Fig.: Electric Field strenght V/cm-1
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

4. P-stop isolation technique
P-stop structure cuts the accumulation layer
Electric fields depend on placement, width, doping conc., structure…
Additional photolitography mask is needed
Fig.: eDensity on sensor front side
Fig.: eDensity on sensor front side; slice 100nm underneath
the bulk/oxide interface
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

5. Validation of Simulation, University of Dehli, Silvaco
Comparison with Review article:
“Device Simulations of Isolation Techniques for Silicon Microstrip Detectors Made on p-Type Substrates” by Claudio Piemonte, IEEE TRANS. NUCL. SCI., VOL. 53, NO. 3, JUNE 2006.
General Trends were same (although we don’t have exact simulation parameters as used by Claudio). Validation with some other papers are also carried out.
Paper
Silvaco TCAD Sim.
Silvaco TCAD Sim.
Paper
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

6. 12 Configurations of Multi-SSD with P-stop in HPK
Experimental results of Cint taken from Lyon Tracker Upgrade Database (thanks to Robert)
n+-p--p+ configuration chosen for these results.
For each design, two separate results for non-irradiated sensors measured by FIRENZE probe station and one result obtained by simulation have been compared
Strip length = cm
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

7. Simulation of Cint for Multi-SSD with Double P-stops
Simulated Structure – zoomed region
Double P-stops
4µm wide separated by 6µm
Instead of two adjacent half- P+ neighbouring strips, we have considered five strips in which Cint is evaluated by sending AC signal to central electrode and measuring it w.r.t. two adjacent strips (which are shorted).
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

8. Simulation Parameters – Same for all 12 configurations
1. Substrate Doping Conc. (NB) = 3.4x1012 cm Surface Charge Density  (QF) = ?
3. Temp = 21o C corresponding to 294 K 4. p+ implant of 30 micron from the back side 5. n+ strip junction depth of 1.5 mm 6. Strip length for normalization = cm 7. Total device depth is 320 mm. 8. Frequency = 1MHz
9. P-stop Junction width = 4 mm, P-stop separation = 6 mm 10. P-stop junction depth = ?
11. P-stop peak doping conc. (Npst) = ?
Other than Width and Pitch, nothing else is changed in 12 Configurations. MO = 6.5 mm on either side is considered on each strip as per MSSD design.
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

9. Simulation of Cint for Multi-SSD with P-stops, Silvaco T-CAD
We performed “sensitivity studies” for peak P-stop Doping Density (Npst) , P-stop Junction Depth (XJ), Surface charge density (Qf) for structure no-1
P-stop peak doping density (Npst) and P-stop junction depth do not affect Cint in a significant manner ensuring desired isolation.
Structure # 1
P-stop Xj = 0.5,1,1.5µm
Structure # 1
NPST = 5e16, 1e17, 2e17, 1e18 cm-3
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

10. Simulation of Cint for Multi-SSD for Structure #1, Silvaco T-CAD
QF effect on Cint is quite significant!
Comparison with measurement shows QF = 5x10cm-2 matches closely with measurement
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

11. MSSD: Measurement vs. Simulation, Silvaco T-CAD
For FZ320P, Double P-stop (each 4µm wide separated by 6µm), Optimized Parameters:
peak doping density = 5x1016cm-3
P-stop doping depth = 1µm
QF= 5x1010 cm-2
Observations:
Initial drop in Cint is not reproduced by Simulations
Cint plateau values are within 20% of the experimental values
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

12. For FZ 320Y: Cint vs. Bias Voltage, Silvaco T-CAD
Simulation vs. Measurement for FZ320Y
All 12 N+-P--P+ configurations (non-irradiated)
For FZ320Y, Peak P-spray doping density =
4x1016cm-3, P-spray doping depth = 0.2µm
Only FIRENZE Probe station measurements are used for comparison
January 22nd, 2013
HPK Workshop CERN Studies of isolation techniques for n-in-p silicon sensors

13. Synopsys T-CAD Simulations of irradiated MSSDs with p-stop, HIP,
5-strip structure for 7-80 and regions
Two-level radiation damage model [1]
Parameter
Value
n+ / p+ doping [cm-3]
5e18
Neff (320P) [cm-3]
3.4e12
Neff (200P) [cm-3]
3e12
Neff (120P) [cm-3]
1.5e12
implant thickness [µm]
1.5
backplane diffusion depth (320P) [µm] 33
backplane diffusion depth (200P) [µm]
115
backplane diffusion depth (FTH200P) [µm] 10
backplane diffusion depth (120P) [µm]
215
Np [cm-3]
3e16, 4e16, 5e16
p-stop thickness [µm]
0.5, 1.0, 1.5
SiO2 thickness [µm]
0.5
T [K]
253, 273, 293
area factor
30490
F (Cint)
1.0 MHz
F (Cback)
1.0 kHz
Qf (irr.) [cm-2]
1e12 [1]
Qf (unirr.) [cm-2]
1e10, 5e10, 4e11
Level
[eV] σn
[cm-2] σp
Introduction rate [cm-1] EC
2e-15
2e-14
1.613 EC
5e-15
5e-14
0.9
1 MeV neutron equivalent doses for the simulations (from CMS irr. campaign)
Device thickness
[μm]
Proton 1 MeV neqv fluence
[cm-2]
1 MeV n fluence
1 MeV neqv
summed fluence
≥ 200
3.0e14
4.0e14
7.0e14
1.0e15
5.0e14
1.5e15
6.0e14
2.1e15
≤ 200
3.0e15
3.7e15
< 200
1.3e16
1.4e16
p-stop configurations (preliminary)
p-stop count
p-stop width
[µm]
spacing 2 10 4 6 1 14
Total p-stop width = constant
[1] M. Petasecca et al. NIM A 563 (2006)
January 22nd, 2013
HPK Workshop CERN Studies of isolation techniques for n-in-p silicon sensors

14. Interstrip capacitances of measured and simulated nonirradiated 200P detectors @ +20 °C , Synopsys
Simulation dp = 1.5 μm
Qf = 5e10 cm-2
Simulation dp = 1.5 μm
Qf = 5e10 cm-2
region 5
region 7
Initial dip not produced by simulation
TEST ID: 15016
TEST ID: 15010
By lowering the value of Qf from 5e10 cm-2 to 2.7e10 cm-2 the initial dip of measured curve is produced by simulation. At higher voltages simulated curve is within ~2 % of measurement.
At higher value of non-irradiated inversion layer oxide charge Qf = 4e11 cm-2 simulation does not match the measurement and fails at ~600 V due to high E at p-stop edges.
For the irradiation simulations, p-stop depth was varied by dp = 1.0 and 1.5 μm
Irradiated inversion layer oxide charge Qf = 1e12 cm-2
region 7
Simulation dp = 1.5 μm
Qf = 2.7e10 cm-2
TEST ID: 15016
January 22nd, 2013
HPK Workshop CERN Studies of isolation techniques for n-in-p silicon sensors

15. Irradiated 200P 7-80 region for varying Np and p-stop count/width @ -20 °C, Synopsys T-CAD
Radiation damage removes electrons
from the inversion layer → isolation
reached at lower Vd?
Widest wp has the lowest Cint and Vd
where minimum Cint is reached
for Np = 3e16
p-stop thickness = 1.0 μm
Cint
Cback
f = 1.0 MHz
f = 1.0 kHz
Are the strips short-circuited for Np = 3e16
until Vd has removed electrons from the inversion
layer?
e- density between two strips and p-stops for
Φ = 7e14 cm-2 and wp = 2μm. Isolation fails for
smaller low Vd?
Qf = 1e12 cm-2
January 22nd, 2013
HPK Workshop CERN Studies of isolation techniques for n-in-p silicon sensors

16. Irradiated 200P 7-80 region for varying Np and p-stop count/width @ +20 °C, Synopsys T-CAD
p-stop thickness = 1.0 μm
Cback
Cint
At higher T electrons are removed
from inversion layer more quickly
→ min. Cint reached at lower Vd
350 V (-20 °C) → 290 V (+20 °C)
e- density between two strips and p-stops for
Φ = 7e14 cm-2 and wp = 2μm
January 22nd, 2013
HPK Workshop CERN Studies of isolation techniques for n-in-p silicon sensors

17. Irradiated 200P 7-80 region for varying Np and
p-stop -/+20 °C, Synopsys T-CAD
p-stop thickness = 1.5 μm
e- density between two strips and
p-stops for Φ = 7e14 cm-2 and wp = 2μm
-20 °C
-20 °C
No step in leakage current
observed for both Np
-20 °C
+20 °C
No high values of Cint observed
after ~30 V
+20 °C
+20 °C
Low e- density even at
low voltage for both Np
Superior configuration to dp = 1.0 μm
January 22nd, 2013
HPK Workshop CERN Studies of isolation techniques for n-in-p silicon sensors

18. P-spray: Electric fields on p+ doping concentration, nonirradiated, Synopsys T-CAD
p=90um, w=20um, w/p=0.22
With increasing doping conc of p+ dopants (boron), max. electric fields increase too!
Si breakdown field about 3x105 V/cm !!!!
max. E-fields raise exponentially with increasing doping conc. !!!
Hence, in order to achieve a sufficient strip isolation and simultanously low electric fields at the strips, the doping conc. of p- spray layer must be calculated carefully !
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

19. P-spray isolation and interstrip capacitance Cint, Synopsys T-CAD
Cint is almost! independent of p+ doping conc. (until 1e16cm-3) for a given oxide charge but then increases with increasing doping conc.
->higher acceptor concentration N_p determines a narrower lateral depletion region and therefore a tighter coupling!?
Ok, but why not observed for concentrations up to 8e15cm^-3 ?
For a given doping conc Cint decreases slightly with increasing oxide charge
-> are e and h removed from surface region? > the p-spray layer gets partialy depleted and less conductive…
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

20. P-stop: electric field on p+ doping concentration, Synopsys T-CAD
Fig.: Electric field strength
N+ strip
P-stop atoll
Max. electric field strengths saturate with increasing doping conc. !!!
Placement of p-stop affects more significantly the e-fields than p-stop doping conc.!!!
Calculation of p-stop doping conc. much more easier compared to p-spray technique…
Electric field strength with P-spray!!
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

21. max. electric field strength on p-stop distance, Synopsys T-CAD
Regarding E-field strenght, the p- stops should be implanted in the center of adjacent strips
No constraint through process technology (but lateral diffusion)
Experimental studies also prefer p-stop distances of at least 15um for pitch 90um.
SNR, CCE. ( M.Valentan,HEPHY)
P-stop distance [um]
P-stop
P-stop
Fig.: strip sensor (two half electrodes) with p-stop atoll config
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

22. E-field and electrostatic potential, Synopsys T-CAD
dist width 8 um
dist width 4 um
dist width 8 um
dist width 4 um
Edge n+ strip
Edge p-stop
E-Field [Vcm^-1]
Potential defines electric field and max
electric field strength at the edges of n+
and p+ implants
Like on the slide before, wider p-stop dist-
ance and lower p-stop width ensures high-
voltage operation
Optimization of p-stop/p-spray dose profile
to ensure good isolation and satisfying the
breakdown performance
potential is higher with wider p-stop
electric field at p-stop edges very high
Potential [V]
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

23. P-stop and interstrip capacitance, Synopsys T-CAD
Gap between strip and p-stop pattern doesn´t affect Cint
but is significantly influenced by surface damage?! Just interface charge, we also have to consider oxide charge growth models…
Experimental studies necessary to confirm simulation outcomes
P-stop conc. 5e16cm-3
Oxide charge. 1e11cm-2
It seems that the accumulation layer „extends“ the n+ strips
-> strip geometry „changes“ and significantly affects Cint…
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors

24. Summary
We are able to reproduce HPK MSSD measurements like depletion voltages, currents (see Ranjeet´s and Robert´s talk) and interstrip capacitances within 20 % of experimental values.
Parameters taken for correct simulation of sensors agree with values from spreading resistance measurements (example: p+, n+ doping concentrations, see Wolfgang´s talk on on processing!?)
Comparison of p-stop and p-spray techniques with the goal to determine the perfect pattern or concentrations etc. is difficult without knowledge of real process parameters as some variables we assume from calculations.
Nevertheless, simulations indicate the p-stop technique to be more convenient compared to p-spray.
P-spray doping conc. is difficult to calculate and electric fields raise exponentially with conc.! But interstrip capacitance is less affected compared to p-stop technique.
P-stop doping conc. can be chosen higher than necessary because of no significant effect on electric field strenghts. Cint is strongly affected by surface damage!
Combination of both has still to be studied.
Simulations exclude some p-stop patterns or distances in agreement with first exp. studies
Radiation damage of surface is used but models for oxide charge growth or interface traps have to be considered.
Next step could be the implemantation of trap models in order to investigate bulk defects and their impact on p+ isolation.
See Robert´s talk
HPK Workshop CERN
Studies of isolation techniques for n-in-p silicon sensors