MODERN ENIAC WP2 Meeting
(WP2-T2.1)
WP2 and Tasks review
Milano Agrate, 2011 Oct. 05
Meeting hosted by Micron
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL
WP2: Relationship among work packages
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 1
T2.1 Task
Task T2.1: PV-aware process simulation
Process simulation tools will be extended to include the impact of variations in TCAD
simulations especially in etching and deposition processes; an interface to commercial
process and device simulation programs will be developed. Process simulations for the
extraction of behavioral models will be performed. In addition it is intended to build up
an interface between the process simulation environment and the semiconductor FAB
to obtain equipment parameters which affect variability.
Partners: ST-I, AMS, TUW
In the analysis and modelling of PV ST-I wants to link process information out of the
silicon manufacturing facility into TCAD environment in order to take into account
inevitable process variations and doping fluctuations with the objectives to create a
behavioural model of the process to be simulated and to perform statistical process
analysis and process optimization to improve parametric yield. AMS and TUW will
focus on TCAD process simulation to reflect major sources for PV in 0.35um, 0.18um
and 0.13um CMOS and HV technologies; main inline/equipment parameters will be
taken into account. TCAD based statements about pros and cons of emerging device
options will be given concerning variability. The methodology will be compared to the
one used in task 2.2.
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 2
Process simulation: T2.1 Deliverables
Partners: ST-I, AMS, TUW
Task 2.1 goal is to perform process simulation including treatment of PV. Application for discrete power devices, SiC,
AlGaN/GaN (ST-I) and HV-CMOS technologies (AMS)
Ref
Deliverable/ Contributors
Due date
D2.1.1
First process simulation including treatment of PV for Discrete Power
Device, HV-CMOS, SiC, GaN/AlGaN technologies, interfaced to
commercial TCAD tools
M15
DONE
(ST-I, AMS, TUW)
D2.1.2
Enhanced process simulation including treatment of PV for Discrete Power
Device, HV-CMOS, SiC, GaN/AlGaN technologies, interfaced to
commercial TCAD tools
(ST-I, AMS, TUW)
Task Leader: valeria.cinnera@st.com
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 3
M27
DONE
T2.1 Review Summary (ST-I)
Activity done so far:
– Interface between commercial process simulator and fab equipments: a web – based tool has
been realized. The tool accepts in input the product code used in manufacturing and produces in
output the process parameters that can be passed to the TCAD software tools (Silvaco or
Synopsys syntax) (Implementation activity is included in WP5)
– Process simulator setup & calibration
– Process simulation with Synopsys platform has been performed.
– Screening parameters: first device simulations have been executed in order to select the
process parameters that mainly affect the electrical behavior of the device:
– Design of Experiment has been arranged to build structures with a systematic variation of the
selected process parameters.
The activity done on a Silicon Power MOS (D2.1.1) has been extended to compound materials (SiC
and AlGaN/GaN devices) (D2.1.2)
– PCM extraction: polynomials function of process parameters have been extracted (T2.2)
– PCM validation: As a check for the robustness of the PCM, a new set of simulation data,
choosing a random combination of input parameters, have been generated and compared with
the prediction given by the PCM (T2.2)
Interactions
– T2.1 DOE results  T2.2 for device simulation and PCM extraction.
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 4
T2.1 Review Summary (AMS, TUW)
Activity done so far, with highlights on technical results, and dissemination
- HV-CMOS technologies (AMS, TUW):
•
•
•
•
•
•
•
•
Statistical fab data analysis (FEOL, BEOL): 0.35µm HV-CMOS Tech.  done
HV devices for DOE: NMOSI20T and PMOS20T
Process simulator setup & calibration  done
Critical PV selection: 6 ~ 8 variables (substrate resistivity, implant dose, CD & overlay,
GOX thickness, …)  done
PMOS20T process simulations with 7 PV variables  done (see Back-up slides)
286 (W/L=10/0.6, 10/10) structures from process simulations go to Minimos-NT for device
simulation.
Device simulations for statistical spice modeling: done (PMOS20T)
Script development for electrical parameter extraction  inline measurement
D2.1.2 deliverable: Done
– An interface between commercial process simulator and Minimos-NT (a two-dimensional device
simulator from TUW)
– Complete set of NMOS20T and PMOS20T process and device simulations  done
– Script update for electrical parameter extraction
– NMOS20T and PMOS20T sensitivity analysis with 7~8 process variations: done
Issues
Interaction need:
– T2.1 results (HV-CMOS)  T2.3 for Spice Monte-Carlo Model (D2.3.3)
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 5
T2.1 Back-up slides for ST-I
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 6
ST-I WP2 Activity
Process recipes
Process flow
Virtual device
High Level
factory
Specific
process
conditions
Mask Layout
FAB1
Technology
transferred to
FAB2 using PCM
PCM
PCM
Process
Compact model
derived from
TCAD
FAB2
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 7
WEB-Based Interface: High-level factory
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 8
T2.1
Process compact modeling approach incorporates statistical design of experiment methods
within the calibrated TCAD environment allowing to capture relationships between process
variations and device performance through a set of analytic functions (Response Surface
Method).
Input process
parameters
Output device
characteristics
Process
Compact Model
New process
conditions
Process
Compact Model
Predicted
device response
The flow of PCM extraction from TCAD simulation consists of the following steps:
- Definition of a calibrated TCAD flow (process and device);
- Analysis of process parameter sensitivity (parameter screening);
- Simulations of process splits and electrical variables extraction (design of experiment);
- PCM generation.
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 9
Application to Power-Mos cell (D2.1.1)
PCM STUDIO
Applied methodology flow.
EHD5 SEMICELL
SENTAURUS WORKBENCH
Synopsys platform:
Sentaurus and PCM Studio
DOE
PCM
Simulation of Power-Mos semi cell
with the nominal values of the process
input parameters
Parameter screening to identify the process parameters that
have an important impact on target electrical parameters.
Parameterized simulation setup (DOE) generating several simulation runs.
Device simulations of breakdown and I-V characteristic for each experiment.
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 10
Process variability
Input process parameters that have influence on the
device behavior.
Output electrical parameters: RDSon and BV
Variable
Description
epithick
Thickness of the epitaxial layer
epiconc
Doping concentration of the
epitaxial layer
gateox
Thickness of the gate oxide
bmask
CD of POLY gate mask
pmask
CD of JFET mask
Sensitivity index for process parameter variation
16
Selecting only those process parameters showing
the greatest impact on electrical performances.
Ron
14
Electrical parameter variation (%)
A large number of process variables requires a very
high computational cost.
BV
12
10
8
6
4
2
0
epithick
WP2 Review Meeting
Milan, October 05, 2011
epiconc
gateox
CONFIDENTIAL 11
bmask
pmask
Design of experiments
The DOE depends on the degree of the polynomial required for
a good data fit.
The output electrical parameter is a non-linear function of the
input process parameters, modeled by a second order
polynomial.
To fit a second-order model, a DOE with at least three levels
is required: 81 experiments are generated.
For each structure, device simulations are performed to extract Ron and BV.
Ron histogram
BV histogram
9
7
8
7
6
6
5
5
4
3
Ron
4
Ron 5%
3
BV
BV 5%
2
2
1
1
BV (V)
Ron (m Ohm *m m 2)
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 12
44
44.8
43.2
41.6
42.4
40.8
39.2
40
38.4
36.8
37.6
35.2
36
34.4
32
32.8
33.6
30.4
31.2
29.6
28
16.3
15.9
15.5
15.1
14.7
14.3
13.9
13.5
13.1
12.7
12.3
11.9
11.5
11.1
10.7
10.3
9.9
9.5
9.1
28.8
0
0
Application to 4H-SiC JBS diode (D2.1.2)
Calibration of TCAD
simulations
Process Step
Value
Epi Thickness
6 µm
Epi Doping
1e16 atm/cm3
Pwell mask
8 µm
PreImplant Oxide
Thickness
0.06 µm
Error Dose Factor
1 (nominal Value)
Activation percentage
35%
Schottky Barrier
1.14 eV
Resistance Anode
0.45e-3 Ω*µm
Resistance Schottky
1e-7 Ω*µm
Resistance Cathode
1.1e-3 Ω*µm
Forward and reverse
characteristics
comparison between
measure and simulation.
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 13
Process variability and DOE setup
Variable
Description
Nominal
Min
Max
Unit
Sp
Space among
p-wells
8
7.6
8.4
µm
EpiT
Epitaxy Thickness
6
5.7
6.3
µm
EpiD
Epitaxy Doping
1e16
9e15
1.1e16
Atm/cm
3
OxT
Pre-implant oxide Thickness
0.06
0.054
0.066
µm
errD
Dose Error
1
0.9
1.1
-
Att
Doping Activation
0.35
0.1
0.6
%
Barrier
Schottky Barrier
1.14
1.083
1.197
eV
Rc
Cathode Resistance
1.1e-3
9.9e-4
1.21e-3
Ω
Parameters screening
DOE setup in SWB
Input process parameters that have influence
on the device behavior:
-FW6 (anode voltage @6A)
-FW80 (anode voltage @80A)
-BV (breakdown Voltage)
-LK600 (leakage current @600V)
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 14
Application to AlGaN/GaN HEMTs (D2.1.2)
Schematic cross-section of the AlGaN/GaN HEMT under examination
Variable
Description
Nominal
Min
Max
Units
thAlGaN
Thickness of the first AlGaN layer
25
20
30
[nm]
GateFoot
Width of the gate foot
0.5
0.4
0.6
[µm]
GatePlate
Width of the gate plate
1.3
1.2
1.4
[µm]
Ldg
Drain-Gate distance
2.5
2.0
3.0
[µm]
Lsg
Source-Gate distance
2.0
1.5
2.5
[µm]
Hgate
Height of the gate foot
0.08
0.06
0.1
[µm]
GateRec
Erosion of the AlGaN layer under the gate contact
0
0
0.02
[µm]
MolFrac
Molar fraction of the first AlGaN layer
0.26
0.24
0.27
--
WP2 Review Meeting
Milan, October 05, 2011
Input process parameters that
have influence on the device
behavior:
- Id_MAX (maximum drain current
value at Vg=2V)
- Vth (threshold voltage)
- gm (maximum of
transconductance for a given Id–
Vg)
CONFIDENTIAL 15
Process variability and DOE setup
Sensitivity analysis
Experiments generated by DoE
Variable
Id_MAX
Vth
gm
thAlGaN
121.3%
32.9%
76.0%
GateFoot
0.50%
0.69%
1.0%
GatePlat
e
0.33%
0.32%
1.8%
Ldg
13.2%
0.56%
13.6%
Lsg
16.2%
1.0%
20.0%
Hgate
0.08%
0.05%
2.54%
GateRec
0.99%
87.5%
86.5%
MolFrac
47.4%
12.6%
32.1%
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 16
Conclusions
In D2.1.2 the methodology used to increase TCAD simulation efficiency, by deriving
Process Compact Models from systematic well-calibrated simulations, is described.
In order to demonstrate the general concepts of how to use the PCM approach, a 4HSiC JBS diode and an AlGaN/GaN HEMT device have been investigated. The Synopsys
platform (Sentaurus and PCM Studio) has been used.
First the standard cell of the device under examination has been simulated with the
nominal values of the process input parameters. Second the process parameters that
have an important impact on target electrical parameters have been identified performing
a parameter screening. Then, a parameterized simulation setup has been arranged.
To complete the analysis, device simulations have been performed, for each
experiment. Moreover post-processing scripts need to be introduced to automatically
extract the list of electrical outputs. In this way the RSM model of device characteristics
as function of process parameters will be generated by using PCM Studio. Details on this
work has been included on D2.2.4.
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 17
T2.1 Back-up slides for AMS and TUW
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 18
Process Variation
• Input Parameters (HV-PMOS)
SX
18
20
22
DN_DOSE
4.05E+012
4.10E+012
4.15E+012
DPOverlay
-0.1
0
0.1
SNOverlay
-0.1
0
0.1
0.1
10.05
20
2.65E+012
2.70E+012
2.75E+012
-2
0
2
PADOX_VthM
Vt_2p7e12
TOXTH
- 143 (7 PV for HV-PMOS) and 272 (8 PV for HV-NMOS)
Full Process and Device Simulations
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 19
Process Variation
• Variation Setup (RSM)
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 20
Process Variation
• TCAD Flow
Correlation
Parameters
Interface between commercial
Synopsys Process Simulator and
Minimos Device Simulator
Sentaurus
Work Bench
Parameter
Extraction
Minimos
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 21
Process Variation
• Minimos Device Simulation
•
•
•
•
Linked to commercial Synopsys Process Tools
Input from .grd and .dat output of Sentaurus Device
Geometry output of Minimos to .grd and .dat
Characteristics to .crv
• Parameter Extraction
• Inhouse Measurement Methods at AMS
• extract BSIM Parameters from Simulation
• Ron, Vth, Idsat, Sleak, Gamma, Leff
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 22
Process Variation
• Parameter Extraction
•
•
•
•
Original Matlab Methods at AMS
Rewritten to extract Parameters from Minimos
In Python with SciPy and Numpy
Nearly identical Syntax as in Matlab
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 23
Process Variation
• Vth Extraction
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 24
Process Variation
• Extracted Data
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 25
Process Variation
• Output Parameters <Gamma,Sleak>(SNOverlay)
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 26
Status of PV-aware Process & Device Simulations
• Status
Type
Parameters
Process
Device
Parameters
HV-PMOS(20/0.6)
finished
finished
finished
finished
HV-PMOS(20/10)
finished
finished
finished
finished
HV-NMOS(20/0.7)
finished
finished
finished
finished
HV-NMOS(20/10)
finished
finished
finished
finished
• Electrical parameter extraction (for Monte Carlo spice model) : done
(Vth_lin, Vth_sat, Idlin, Idsat, Ron, Sleak, Gamma, Leff, …)
D2.3.3
*Remarks: HV-PMOS = PMOS20T, HV-NMOS=NMOSI20T
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 27
Quadratic fitting from the simulation results
- H35 technology (0.35 µm HV-CMOS process)
- PMOS20T: 143 structures (7 PV)
- NMOSI20T: 272 structures (8 PV)
 TCAD Process & device simulations (sprocess & Minimos-NT)
 Quadratic fitting (all PV versus electrical data)
 Random 10000 PV-set generation by considering inline PV distributions
 Electrical parameter calculation from the quadratic formula
 Data analysis
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 28
PMOS20T(20/0.6): Ron and Ron-fit
16
0.5
15.5
0.4
15
0.3
Ron - Ron-fit
Ron, Ron-fit
14.5
14
13.5
13
0.2
0.1
0
12.5
-0.1
12
-0.2
11.5
11
0
50
100
150
-0.3
0
count
50
100
count
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 29
150
0.68
0.02
0.66
0.015
0.64
0.01
0.62
0.005
vthlin - vthlin-fit
vthlin, vtlin-fit
PMOS20T(20/0.6): Vthlin and Vthlin-fit
0.6
0.58
0
-0.005
0.56
-0.01
0.54
-0.015
0.52
-0.02
0.5
0
50
100
150
-0.025
0
count
50
100
count
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 30
150
16
16
15.5
15.5
15
15
14.5
14.5
14
14
Ron-fit
Ron
PMOS20T(20/0.6): Ron and Vthlin correlation
13.5
13.5
13
13
12.5
12.5
12
12
11.5
11.5
11
0.5
0.55
0.6
vthlin
0.65
0.7
11
0.5
WP2 Review Meeting
Milan, October 05, 2011
0.55
0.6
vthlin-fit
0.65
CONFIDENTIAL 31
0.7
PMOS20T(20/0.6): 7 PV distributions (10000 random values)
350
400
300
350
350
350
300
300
250
250
200
200
300
250
200
n
n
n
200
150
n
250
150
150
100
100
50
50
150
100
100
50
0
16
50
18
20
SX
22
0
4
24
4.05
4.1
DN dose
4.15
0
-0.2
4.2
350
300
300
250
250
250
200
200
200
150
100
100
100
50
50
50
20
30
0
-0.2
0.2
-0.1
n
150
10
Pad Ox
0.1
300
150
0
0
DP overlay
350
n
n
350
0
-10
-0.1
12
x 10
0
2.6
2.65
2.7
Vt implant
2.75
2.8
0
-4
-2
12
x 10
WP2 Review Meeting
Milan, October 05, 2011
0
GOX
2
4
CONFIDENTIAL 32
0
SN overlay
0.1
0.2
PMOS20T(20/0.6): Ron and Vthlin distribution
350
400
300
350
300
250
250
n
n
200
200
150
150
100
100
50
0
11
50
12
13
14
Ron random
15
0
0.5
WP2 Review Meeting
Milan, October 05, 2011
0.55
0.6
0.65
vthlin random
CONFIDENTIAL 33
0.7
PMOS20T(20/0.6) & NMOSI20T(20/0.7): Ron versus Vtlin
15.5
11
15
10.5
14.5
Ron random
Ron random
14
13.5
13
10
9.5
12.5
12
9
11.5
11
0.5
0.55
0.65
0.6
vthlin random
0.7
8.5
0.39
WP2 Review Meeting
Milan, October 05, 2011
0.4
0.41 0.42
vthlin random
0.43
CONFIDENTIAL 34
0.44
Conclusions
Statistical fab data analysis (FEOL, BEOL) for 0.35µm HV-CMOS Techhnology
was done for PV TCAD simulations.
TCAD environment construction:
- Process simulator setup & calibration
- Critical 7~8 PV selection
- Interface development between commercial process simulator and Minimos-NT (a twodimensional device simulator from TUW)
- Script development for electrical parameter extraction
TCAD simulations:
- A set of NMOS20T and PMOS20T process and device simulations
- NMOS20T and PMOS20T sensitivity analysis
- Quadratic curve fitting of simulation results, Random PV generation
- PV-aware statistical electrical parameters
T2.1 results (HV-CMOS)  T2.3 for Spice Monte-Carlo Model (D2.3.3)
WP2 Review Meeting
Milan, October 05, 2011
CONFIDENTIAL 35