! #"%$'&)(+*-,./#01#2!354638&7 /69:#2;,$=<-.),$=(+*&>4?,.)@,.A/B2C#DFE&)$=<#3-/
0G*HGIJ$'<KEL.)MN,./01#0G$=(+* ,./2C#0PE5O QSR /6!UTB3 3-/-V:UTW(+*% O QSRL/
T??T8, 4U,.S*-,.).A/6XY?TZQ[$\MQ[$'<-@,.A/-X#]^* ._<-,.
Contents
1 Introduction to MINIMOS
Hj jkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.2 MINIMOS 2D Model Structure jljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.3 MINIMOS 3D Model Structure jljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.4 Transient Simulation jHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.5 Monte Carlo Simulation of Carrier Transport jHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.6 Small-Signal AC Analysis jHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.7 Nonplanar Geometries jkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.8 Three-Dimensional Simulation jHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.9 MESFET Simulation jHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.10 Gate Charge and Capacitance Calculation jHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.11 Monitoring Program Execution jljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
1.1
jHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
Mobility Model for Gallium Arsenide jljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
Generation-Recombination Modeling jljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
The Local Carrier Heating Model jHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
Modeling of Band-Structure Parameters jHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
2.5.1 Models for Silicon jHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj
Mobility Model for Silicon
1
1
2
4
5
8
9
10
11
12
13
14
17
17
20
21
23
24
24
jHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.6 The Band-to-Band Tunneling Model jljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.7 The Dynamic Interface and Bulk Trap Model jHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.7.1 The Physical Model and Numerical Considerations jHjHjljHjkjljHjHjHjljHjkjljHjHj
2.7.2 Interface Traps jljHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.7.3 Bulk Traps jHjHjljHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.7.4 Examples jHjHjljHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.8 Gate Depletion Analysis jHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.8.1 The Full Two-Dimensional Model (MPOLY=1) jljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.8.2 The Approximate One-Dimensional Model (MPOLY=0) jljHjkjljHjHjHjljHjkjljHjHj
2.8.3 Fixed Charge and Traped Charge at the Gate/Oxide Interface jljHjHjHjljHjkjljHjHj
2.8.4 Examples jHjHjljHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.9 Power VDMOSFET Simulation jHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10 The Monte Carlo Transport Module jljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.1 Band Structure jljHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.2 Scattering Rates jHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.3 Impact Ionization jkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.4 Material Characteristics jHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.5 The Monte-CarloDrift-Diffusion Coupling Technique jljHjkjljHjHjHjljHjkjljHjHj
2.10.6 The Monte Carlo Windows jljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.7 The Self-Consistent Iteration Scheme jHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.8 Calculation of Averages jHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.9 The Particle Weighting Algorithm jHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.10.10 The Self-Scattering Algorithm jHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
2.5.2
jkjHjljHjHjljHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
DEVICE jljHjHjljHjkjHjljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj
BIAS
ii
25
26
31
32
33
35
37
40
41
42
44
44
47
50
50
51
54
55
55
56
58
58
59
60
61
62
64
jHjHjHjljHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 68
FIMPLANT jHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 71
GEOMETRY jkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 73
GRID jHjHjljHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 77
IMPLANT jljHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 83
INTERFACE jHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 85
MOBILITY jHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 94
MONTE-CARLO jHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 96
OPTION jHjljHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 103
OUTPUT jljHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 109
PROFILE jljHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 112
RECOMBINATION jHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 120
STEP jHjHjljHjkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 125
TRANSIENT jkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 128
END
4 Files
4.1
129
jljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 129
The Input Files jkjljHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 129
One-dimensional Profiles (FILE=1-D) jHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 130
Twodimensional Profiles (FILE=2-D) jHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 131
The Main Ouput Files jHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 134
The Formatted Output File jHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 134
The Reference File jHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 135
The Binary Output File jljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 135
The Formatted PIF Output File jHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 138
Auxiliary Output Files jHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 139
Table of I-V Data jHjHjHjljHjkjljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 139
The Monte Carlo Output File jljHjHjljHjHjkjljHjHjljHjHjkjljHjHjljHjHjljkjHjHjljHj 139
4.1.2
4.1.3
iii
jHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 140
The Gate Charge File jljHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 141
Poly-Gate Voltage Drop jHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 141
Charge-Pumping Output jHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 142
Band-to-Band Tunneling Currents jHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 143
The MINIMOS 3D Files jHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 143
The Binary Data Link File jHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 143
The Formatted Output and the Reference Files jljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 143
The Binary Output File jHjHjljHjkjHjljHjHjljHjkjljHjHjHjljHjkjljHjHjHjljHjkjljHjHj 144
The Transient Output File
4.1.4
Bibliography
149
Index
155
iv
1
Introduction to MINIMOS
MINIMOS is a software tool for the numerical simulation of semiconductor field-effect transistors. The
first version was released over 10 years ago, and since then the code has undergone continuous extensions and improvements, culminating in 1991s Version 5.2, which is capable of modelling silicon
MOSFETs and SOI transistors as well as gallium arsenide MESFETs. The fundamental semiconductor
equations wich are solved in MINIMOS numerically comprise Poissons equation and two carrier continuity equations for electrons and holes. For the various coefficients such as carrier mobility, carrier
generation/recombination and impact ionization rates carefully chosen models are implemented. The
numerical schemes of MINIMOS are designed to handle planar and non-planar device structures in both
two and three space dimensions.
m
m
Transient analysis.
Self-consistent inclusion of the trap-rate equations for dynamic of interface and bulk traps.
A Monte Carlo Module which couples self-consistently to the Poisson equation and replaces the
drift-diffusion approximation in critical device areas.
For instance, the transient facility in conjunction with the trap rate equations allow for an accurate
numerical analysis of the charge pumping-phenomenon.
The Monte Carlo Module available in this version makes MINIMOS 6 to an easy to use Monte Carlo
device simulator which describes hot electron transport and impact ionization in a fully non-local manner.
In the following some minor extensions are summarized:
P-i-n diodes can be handled (flag for p-i-n diode simulation is set if an NPP+ or PNN+ structure
is specified by appropriate profile statements).
If a STEP directive is specified the resulting I-V data are written automatically to a file
(unless OUTPUT IV=NO).
Coupled solution of the AC small-signal device equations for more robustness at higher
frequencies (OPTION CP=YES)
builds its grid through an adaptive grid refinement loop. Each refinement step consists of a grid update
followed by a Gummel iteration loop, which terminates when all error norms lie beyond the threshold set
by the key ERROR on the END directive. After completing the grid loop, that is when the mesh weighting
function (BWF) is smaller than unity for all mesh intervals, MINIMOS switches its internal mode to
2-D. Now the minority continuity equation is solved in the full simulation domain, for the majorities
there is still a constant quasi fermi level assumed. No grid updates are carried out. Since the deviation
from thermal equilibrium in the deep bulk is usually small, only few Gummel iteration steps should be
necessary.
Finally, MINIMOS executes in AVAL mode. The Poisson equation and both carrier continuity equations
are solved in the full domain. The recombination rate is computed at every Gummel step, the impactionization and tunneling generation rate, however, only after a sufficient decrease in the error norms. This
leads to an avalanche subcycle. MINIMOS completes its execution when the last carrier generation update
does not increase the error norms over the threshold set by ERROR.
For a MESFET MINIMOS proceeds slightly different. In the 1-D mode both carrier continuity
equations are solved, however generation/recombination is neglected. Then in 2-D mode the generation/recombination mechanisms, impact ionization excluded, are switched on. Finally, the AVAL mode
for the MESFET has the same meaning as the AVAL mode for the MOSFET.
In model HOT another cycle is superimposed on the previous model AVAL. To account for carrier heating
effects local carrier temperatures are introduced. After the completion of each AVAL cycle a carrier
temperature update is performed. MINIMOS in HOT mode terminates when the error norms in both
carrier temperatures have decreased beyond the error measure TCERR, specified within the END directive.
In model MC MINIMOS generates an initial solution for the subsequent Monte Carlo iteration by processing
the internal models 1-D through AVAL. The resulting distribution of the electrostatic potential serves as
input for the Monte Carlo simulation. As output the distributions of electron mobility, electron temperature
and ionization coefficient are obtained in critical device regions. With the updated mobility and temperature
distributions a gummel iteration is performed. After its termination an avalanche subcycle is imposed to
account for the altered ionization coefficients. This procedure yields the potential for the next Monte Carlo
iteration. All the above described steps which are needed to obtain the new potential from the old one
can be considered as one Monte Carlo Poisson iteration step. MINIMOS terminates after a user-defined
number of such iteration steps has been executed.
In the following we summarize the model hierarchy of MINIMOS. A model can be specified by the
MODEL key on the OPTION directive.
MODEL
1-D
THRES
2-D
AVAL
HOT
MC
kn oqp
nkoqp
nkoqp
nJoqp
nkoqp
nJoqp
Description
Bulk-MOSFET: Poisson, minorities, partial domain, no
SOI-MOSFET : Poisson, minorities, full domain, no
MESFET : Poisson, both carriers, full domain, no
Loop for threshold voltage calculation by model 1-D
MOSFET : Poisson, minorities, full domain, no
MESFET : Poisson, both carriers, full domain,
, no impact ionization
MOSFET, MESFET: Full equation set, full domain,
including impact ionization.
Local carrier tempearture model in addition to model AVAL.
Full set of equations with electron temperature, electron mobility and ionization
coefficient from a Monte Carlo simulation.
M3MODE=2O, rename the binary outputfile of the previous MINIMOS 3D run appropriately and restart
MINIMOS 2D. See Chapter 4. MINIMOS 2D is terminated after having loaded the changed commonblock data. Now invoke MINIMOS 3D. If MINIMOS 3D finds a file with the old three-dimensional
solution of M3MODE=1 it continues immediately with M3MODE=2. See also the key M3MODE on the
option directive Page 103 to obtain further information.
1N
1O
2N
2O
n:r;p
sAt
After each time step a summary of terminal currents is printed in the reference file. These currents are
split up into the electron (ele), hole (hol) and displacement (dis) current, each of these components being
printed on a separate line. The last line of this current information record prints the sum of electron, hole
and displacement current, i.e. the total electrodynamic current for each individual contact. The last entry
in this line is the sum of all terminal currents. Due to the conservation law of charge this must be a small
number, which depends on the solution error of the semiconductor equations. After the last time step
average values of all terminal currents across the total simulation time are printed in the reference file.
For the calculation of the terminal currents in transient conditions we employed an accurate weighting
functions-based technique.
The evolution of the terminal voltages and currents are written to file unit NUMTRA.
u v
Wi=2 denotes a periodic trapezoidal voltage with the starting point at t=0 located at the beginning of
the falling edge of the trapeze (see Fig. 1.1). To specify parameters for this waveform the Tij and Uij
parameters are used, where i denotes the contact as outlined above and j enumerates the parameters of the
individual waveform. All T parameters are specified in seconds, all U parameters in volts.
{}|
wAxgy
{}~
{'
2 are the following: Ti1 is the period length of the trapeze pulse ( ),
The waveform parameters for
Ti2 is the fall time ( ), Ti3 the lower-plateau time ( ), Ti4 the rise time ( ) and Ti5 is the simulation end
time. Note that the upper-plateau time ( ) of the waveform needs not be specified. If Ti5 is larger then
Ti1, more than one period will be simulated. The parameter Ui1 is by default equal to the corresponding
voltage given in the bias directive and needs not be specified. Ui2 specifies the amplitude of the trapezoidal
waveform. The given value for the amplitude may be positive or negative. If a negative value is specified
for Ui2, the waveform is mirrored with respect to the voltage axis of the initial steady-state bias. In this
case the specifications falling edge and rising edge interchange their meaning and the transient analysis
starts at
0 with a rising edge. Clearly, Ti2 denotes the rising edge and Ti3 denotes the falling edge
when 2 0.
{}
{6y
The parameters Ui3 and Ui4 control the time-step lengths when a rising or falling voltage ramp is stepped
through during a transient simulation. Specifically, Ui3 denotes the voltage increment to be used for the
contact i for the falling edge (if 2 0, otherwise the rising edge), Ui4 denotes the voltage increment to
be used for the contact i for the rising edge (if 2 0, otherwise the falling edge).
wxy
{=|
With
3 three-level voltage pulses can be constructed (see Fig. 1.2). The following parameters
specify all required values: Ti1 is the waveform period ( ), Ti2 the fall time prime ( ), Ti3 the mid-level
duration ( ), Ti4 the fall time double prime ( ), Ti5 the rise time ( ), Ti6 the top-level duration ( ) and
Ti7 the end of the simulation. Specification of the low-level duration ( ) is not necessary. The waveform
is periodic with the period length Ti1. Ui1 equals the steady-state bias and needs not be specified, Ui2
is the waveform total amplitude and can be positive or negative. In the negative case the waveform is
mirrored with respect to the axis placed on the top level. Both falling edges change to rising edges, the
{'
{ |
{=
{}~
{ |
rising edge changes to the falling edge and the simulation starts at the first rising edge ( ). Ui3 specifies
the mid-level voltage, which must range within [Ui1,Ui1-Ui2].
The parameters Ui4, Ui5 and Ui6 control the time step lengths for the three voltage ramps: Ui4 is the ramp
step voltage increment of fall time prime, Ui5 is the ramp step voltage increment of fall time double prime
and Ui6 is the ramp step voltage increment of the rising voltage ramp. For a negative amplitude Ui2 these
parameters change their meaning accordingly.
waveforms. The waveform parameters for each specified contact will be checked for completeness and
physical consistency. End time can be any positive number, also less than the period. Therefore, one can
focus on a part of the specified waveform when analyzing only e.g. turn-on and turn-off of device.
The actual time stepping is controlled by waveforms on all contacts simultaneously, fulfilling the requirements specified by user for the particular waveforms.
During a Monte Carlo simulation statistical data, sometimes also referred to as raw-data, are recorded
on an auxiliary grid which is independent of the MINIMOS grid. After a Monte Carlo simulation
the raw-data are assigned to the grid points of the MINIMOS grid by means of a convolution method.
The parameters for the weighting function used for this step can either be specified in the input file or
determined automatically by MINIMOS.
The result of the AC-analysis is one row of the admittance matrix, if just one terminal bias was specified
by the JCAP key, or the whole 4 4 matrix if JCAP=A was declared. The admittance matrix data are
printed (together with the other output data) in the formatted output file. The left item is the real part in
Siemens ( ), the right item is the capacity in Farad ( ).
If convergence problems arise with the default settings one should invoke the coupled solution algorithm
with CP=YES on the the OPTION directive. To obtain accurate results it is recommended to explicitly
specify the length of the source and drain terminals by the LSOURCE and LDRAIN keys on the OPTION
directive.
10
Key
JCAP
VI
VR
CAPERR
NAC
CP
LDRAIN
LSOURC
OMEGA
Description
ACbias terminal(s)
Imaginary part of applied voltage
Real part of applied voltage
Final Error
Iteration limit
switch to coupled solution method
Drain contact length
Source contact length
ACbias angular frequency
Directive
BIAS
BIAS
BIAS
END
END
OPTION
OPTION
OPTION
OPTION
1yr
Gy ~
MINIMOS 2D automatically switches to nonplanar mode if the GEOMETRY directive is present within
the input deck. MINIMOS 3D calculates always nonplanarly in the zdirection. It runs nonplanar in all
spatial directions only if the GEOMETRY directive is given on the input deck.
The nonplanarity is handled internally by a point characteristics and a box integration concept. Due to the
necessity of resolving the nonplanarities sufficiently accurate there will generally be increased mesh sizes
and thus higher computer resources will be required. This is particularly true for full twodimensional
and threedimensional nonplanar simulation.
11
G1CGRD
G2GRD
SGAP
DCGRD
OXIDEF/2
+TINS/2
SDREOX
DGAP
DDREOX
RECESS
TINS
Y
S1GAP
D1GAP
D2GAP
S2GAP
SRGRD
R1GRD
R2GRD
DRGRD
12
W/2
W/4
OXIDEF
Z
RECESS
TINS
OXIDEF/2
-TINS/2
13
PROFILE and RECOMBINATION. For the foundation of the mobility model for GaAs in MINIMOS see
Section 2.2. Both, steady-state and transient simulation, self-consistently including transient rate equations
for (deep) bulk traps, can be normally performed with MESFETs, Section 2.7.
^y H
1 j 1
where H is the total charge in the gate. Performing the simulation at two gate biases r o 2 and
gate capacitance may be calculated by
o 2 the quasi-static
ZZ k o 2Lr! H r Zo 2
1 j 2
The differentiation should be done externally after a MINIMOS run. Proper results can be obtained
assuming i 25 . The value of can be specified on the STEP directive by the VG key.
the discretization error is quite reduced by using the same grid for both bias points. In Section 2.8 the
accuracy of this technique is demonstrated.
Key
GCHC
GRidf
VG
Description
Invoke calculation of total gate charge
freeze the discretization mesh
for each gate bias UG two DC-calculations are performed
at UG-VG/2 and UG+VG/2
The calculation of the gate charge is possible only for planar device structures.
Directive
OPTION
OPTION
STEP
14
m
m
Gummel iterations.
Grid updates.
The displayed data are grouped according to the active model 1-D, 2-D, THRES, AVAL or HOT for
MINIMOS 2D and 1, 3 for MINIMOS 3D. Grid information is given selectively when grid adaption is
carried out i.e. when MODEL=1-D and/or M3MODE=1.
A grid adaption record contains the following information:
m
m
The error norms for the carrier temperatures if the HOT model is active.
15
The inner iteration count for the majorities (MA), MINIMOS 3D only.
The inner iteration count for the minorities (MI), MINIMOS 3D only.
Note: If the Gummel iteration count is preceded by a minus sign, MINIMOS tries to accelerate convergence.
When MODEL=HOT is active a carrier temperature subiteration is performed. At each carrier temperature
update an informational record is issued, which contains:
m
m
a item.
m The carrier temperature error norm for the minorities, 3 item.
The carrier temperature error norm for the majorities, 2
MINIMOS controls convergence by the values of certain error norms. A Gummel iteration or a HOT
subiteration loop is terminated if all required error norms are smaller than the threshold set by the key
ERROR on the END directive. All error norms are dimensionless real numbers.
16
2
The Physical Models of MINIMOS
~ y
300K
z
~ y 460 cmVs 300K
1430
cm2
Vs
2 j 1
2 j 2
2 18
To account for mobility reduction due to ionized impurity scattering the formulae of Caughey and Thomas
[9] is used in conjunction with temperature dependent coefficients.
~ 8
S~_ Cy W_ _ r S
1
! 8
z z 200 K
8 y 80 cmVs 300K
z z
8 y 80 cmVs 200K
200K
300K
z z 200 K
8 y 45 cmVs 300K
0 45
0 15
0 45
2 j 3
0 45
17
200 K
2 j 4
2 j 5
2 j 6
18
W y
45
5 # 200K z
| y 1 j 12 10 cm z
300K
z
| y 2 j 23 10 cm 300K
z
_ y 0 j 72 300K
cm2
Vs
200K
300K
0 45
0 15
200 K
32
17
17
32
0 065
2 j 7
2 j 8
2 j 9
2 j 10
| ~[ } |
_~[N y _ 1 b _ r; S o> 1 | rP\
_ S
z
| y 638 cmVs 300K
z
| y 240 cmVs 300K
|
exp r
Zy 1 2 exp
r 2 o o } |
2
1 19
1 09
2 j 11
2 j 12
2 j 13
2 j 14
~_N
2
[ ~ N S y
S
_ 1
1 N } Z
y grad Pr 1 grad (5 )
2 j 15
2 j 16
2 j 17
19
Here
represent the driving forces for electrons and holes and
voltages.
\N y 1 j 45 10 cms S
\N y 9 j 05 10 cms
7
tanh
2 j 18
tanh
2 j 19
155K
312K
Several parameters in the MINIMOS mobility model can be weighted. The weight coefficients which can
be specified on the MOBILITY directive have the following meaning:
Key
ML
MI
MC
MR
MS
MT
MV
MB
Symbol
Lattice and Impurity Scattering
and
in Equation 2.1 and Equation 2.2, respectively.
weight for
weight for
and
in Equation 2.3.
Surface Scattering
weight for
and in Equation 2.11.
weight for
and
in Equation 2.11.
and
in Equation 2.14.
weight for
weight for
and
in Equation 2.11.
High Field Mobility
weight for
and
in Equation 2.15.
weight for the exponents
and in Equation 2.15.
~ |
|
|
|
'
~ |
_
} |
|
|
'
|
_
All coefficients multiply the coresponding parameters, except MS, which divides the default
. Those
weights should only be used for fine tuning, since the default mobility parameters in MINIMOS are chosen
carefully and agree rather well to experiments. The default parameters in MINIMOS are:
20
W ~
|
} ||
} |
'
o>t
o>t
j
j
o>t
j Boq
j
j oat
=
=
=
=
=
=
=
=
=
=
1430
80
1 12 1017
0 72
638
7 0 105
1 69
10 10 7
1 45 107
2
2
2
8 ~
} |
||
|
'
=
=
=
=
=
=
=
=
=
=
o>kt
o>kt
j
j
o>kt
j Boq
j
j oat
2
460
45
2 23 1017
0 72
240
2 7 105
1 00
10 10 7
9 05 106
1
z
_~ Cy _ 5 300K
#
2 j 20
Impurity scattering is modeleled by the Caughey and Thomas [9] expression Equation 2.3. For gallium
arsenide the asymptotic mobilities at high impurity concentrations,
and
, can be weighted. The
coresponding key is MB key on the MOBILITY directive. The default value of MB is zero in case of gallium
arsenide, so that the influence of
is suppressed. Note that the MB key acts differently for gallium
arsenide and silicon (see key description Page 94).
W_
~ _ y ~[ 1 F A ~[ N
1F
A ~[ N
' y 6 10 cms 8 1 ~_ r ~_
10
5 j 477 10
~[
~_ #
: y 10 0 j 6 exp 10 r 2 exp 7 r 285 j 7 2
2 j 21
2 j 22
4 cm2
Vs
6 cm2
Vs
3 cm2
Vs
cm2
Vs
2 j 23
{Zy 4
1280 sinh
21
~_
250
2 j 24
cm2
Vs
~_
~_ y 1 q N
' y 1 j 5 10 cms
2 j 25
2 j 26
and
represent the driving forces for electrons and holes according to Equation 2.16 and Equation
2.17, respectively.
For gallium arsenide several parameters can be weighted in the mobility model. The weights which can
be specified on the MOBILITY directive have the following meaning:
Key
MB
MI
ML
MV
Symbol
weight for
weight for
weight for
weight for
All weights multiply the corresponding parameters. The default settings of the various parameters in the
gallium arsenide mobility model are:
8 ~
|
=
=
=
=
=
=
>o Jt
>o Jt
>o Jt
8000
1
1500
8000
1017
05
2
2
8 ~
|
=
=
=
=
=
=
380
2
50
380
3 232 1017
0 4956
j
j
>o Jt
>o Jt
>o Jt
2
2
22
bulk traps is described below in Section 2.7. A description of generation-recombination due to interband
tunneling as it is modeled in MINIMOS can be found in Section 2.6.
The parameters in the following expressions can be set on the RECOMBINATION directive.
p S y _ ( )s r [ (s )
2 j 27
where , s and are the local concentrations of electrons, holes and the intrinsic concentration, respec2
plJy s r s
2 j 28
specification and are superposed on the generation-recombination calculated for bulk traps.
p | y s r s%
2 j 29
with SN and SP being the surface recombination velocity for electrons and holes, respectively at the
2
interface between substrate and gate insulator. This surface recombination is independent of terminal bias
and interface trap density. It is superposed to the surface recombination rate calculated for interface traps
as explained in the previous section. If
0 or
0 is specified this surface generation-recombination
vanishes. Note, default values are not 0.
pH | y s r s
2 j 30
with VN and VP being the surface recombination velocity for electrons and holes, respectively. This
2
surface recombination is independent of terminal bias and interface trap density. It is superposed to
the surface recombination rate calculated for interface traps placed at the back interface. This surface
generation-recombination can be suppressed by
0.
23
Impact Ionization
MINIMOS has been enhanced to include a depth dependent ionization model that has been verified against
n-channel substrate current measurements. The ionization coefficients are given by
Zy: exp r Z
2 j 31
Zy: exp r
2 j 32
with denoting the local electric field and the depth from the Si-SiO interface. The existence of a
The form of the depth dependence is similar to that used in the above references. The effect of the
depth dependence is to increase the
factor near the surface, which is similar to decreasing the
S settings
mean free path near the surface. Default
for the electron coefficients have been derived from
experimental measurements of the substrate current over a number of n-channel devices with channel
length in micrometer and 3 o 4 micrometer range. Hole parameters have not been calibrated against local
experimental data and rather use the so-called dark space coefficients from MINIMOS 4. Since no
experimental data were available for checking a depth dependent ionization model for holes, the default
is to apply the electron depth dependence to holes analogously.
All the above described generation-recombination mechanisms are active in steady-state as well as in
transient conditions.
z S y 2 \N ~[1N) r ~[1N
S
5 3 S _ _ _
2 j 35
The energy relaxation times
and are constant with respect to the carrier temperatures. They are
modeled by
24
3
y 2 a5
y 32 )
~_N
'
~[N
'
2 j 36
2 j 37
The energy relaxation time weights UN and UP have a default value of 0 8 and can be specified on the
OPTION directive.
Cy 1 j 045 4 j 5 10 Cz Z
yC 0 j 523 1 j 4 10 z r 1 j 48 10 z
2 j 38
2 j 39
The intrinsic concentration, if not explicitly specified by the INTRINSIC key on the OPTION directive,
is modeled by
E
exp
2 40
2
r z
where E is the energy bandgap for silicon E r E according to
z r 6 j 05 10 z
E y 1 j 1700 1 j 059 10
C
C z
z r 3 j 05 10 z
EUy 1 j 1785 r 9 j 025 10
z
with E being obtained in q .
170
170
2 j 41
2 j 42
2 j 43
y 2 j 51 10
19
25
3 2
z6oq^ 204
#z oq
E y 1 j 981 r 4 j 600 10
z6oq 204 q
The electron effective mass is modeled by
yC exp r z exp r z
with
y 0 j 067
y 0 j 55
y 0 j 85
y E r E
y E r E
3 2
3 2
2 j 44
2 j 45
2 j 46
2 j 47
22 jj 4489
2 j 50
2 j 51
2 j 52
2 3
3 2
!
with
y 0 j 076
y 0 j 50
The density of states of the conduction and the valence band is calculated by
y 8 j 63 10 z
z
z
1 r 1 j 93 10 C r 4 j 19 10 C 21 exp r zG 44 exp r z
y
"!$#
"
13
2 j 53
22 jj 5545
3 2 2 3
3 2
3 2
26
y 2 2 z
%
2 j 56
3 2
& 2
'(
exp r 2E%z j
2 j 57
The first integer number in the key MTUN in the RECOMB directive specifies which of these two approaches
is used, as explained bellow.
y o
To calculate the generation rate associated with an individual path, models for direct and indirect tunneling
)
E * and the field
as function of the starting field , an average field along the tunneling path
variable from
to
(endfield) are implemented. Which field-model is applied in the calculation can
be specified by the third integer number in the key MTUN.
The calculated generation rates of electron and holes are separated in the position space. They are coupled
with the continuity equations via the generation term in a selfconsistent manner after filtering. The total
charge in device is strictly conserved. Filtering is used for smoothing the distributions only, to allow an
efficient and automatic grid adaption in the critical areas. Standard deviations for filtering associated with
and directions can be scaled by the keys SXT and SYT, respectively in the GRID directive. The default
27
values of the standard deviations are chosen carefully, being appropriate for the grid adaption when the
tunneling is localized in the gate/drain overlap region. Therefore, it is not recommended to change these
keys when analyzing the standard gate-induced-drain-leakage (GIDL) problem ( [17, 18, 23, 25]). It
is, however, necessary to increase SXT up to 5 when tunneling takes place along the whole MOSFET
channel, as in the BBISHE injection [22].
Since the tunneling rate is strongly dependent on the electric field, the tunneling analysis is done on the
grid which is MXT times finer for the -mesh and MYT times finer for the -mesh than the MINIMOS
mesh for solving the basic semiconductor equations in the bulk. Increasing MXT and MYT quite reduce
the discretization error in the terminal currents, but can prolong the computation time. The default values
of 2 should represent a proper choice.
Terminal currents are calculated by a very accurate technique based on the local concentration-dependent
weighting functions (a similar method is proposed in [33]).
After an individual bias point is calculated, some quantities characteristic for the band-to-band tunneling
analysis are printed in the reference file. These cover the number of active starting points in the fine
discretization mesh, the actual fine mesh for the analysis in the bulk, the minimal and an average tunneling
distance, the maximum electric field involved in the calculation of the tunneling rates (either starting
or an average field, dependent on the specification) and an average electric field across all paths found.
If the variable-field model is specified, the field variation along the shortest tunneling distance and an
average field variation across all paths are printed, as well. In addition, the total tunneling generation
rates of electrons and holes in the whole device are printed; they should be equal. Particularly, the carriers
0) and the drain-half of device are presented. Note that
generated in the source-half ( ,+ 2,
the tunneling can occur in both junctions at some bias conditions (sufficiently large bulk reverse bias).
After the tunneling rates the total avalanche generation rate in device, as well as the sum of the both, total
avalanche and tunneling rates are printed ( [29]). These rates should be compared with the electron and
hole components of the terminal currents, printed in the current-information record. Note that additional
generation-recombination processes also contribute to the terminal currents.
-o
E
exp r
2 j 58
18 E
The values of the physical parameters are given below. This model is derived for the internal-field emission
in an infinite material, assuming a constant field and two-band interacting over
perturbation [35].
Because of the later, it is convenient for narrow-gap semiconductors, but not for silicon. Regarding to
.-0/
2
%
1 2
& 2
1 2
1 2
&
3 2
2143
silicon MOS devices, the second restriction is the assumption of an infinite medium. In MOSFETs,
the tunneling occurs very close to the interface, where the movement becomes quantized in the normal
direction. Third important restriction is a variation of the electric field along the individual tunneling
paths, an effect normally taking place in devices.
28
In Equation 2.58 stands either for an average field E * , being the tunneling length, or for the field
at the starting point . For the gap E the user can choose both, indirect and direct silicon band gap, in
spite of the former being physically unreasonable. TLIF and TEXF are fitting parameters with a value 1
as default.
We observed significant variations of the electric field along the tunneling paths, from the starting field
to the endfield .
ranges from 1 5 to 5 for GIDL problem in common cases. This ratio can also
be smaller than 1, as for tunneling towards the interface in the BBISHE injection in n-channel devices and
for the GIDL problem in p-channel devices. To analyze the impact of the field variation on the tunneling
)
) )
current, we develop a model for linearly variable field:
. It is based on WKBJ
6
and the two-band 5 5 dispersion relation. Assuming that
is independent of the energy associated with
the momentum perpendicular to the tunneling direction, after several approximations we arrive at [31]
8 y r r o
|y 181 E 1 3 o 16 exp r 2 E 1 3 o 32
2 j 59
variations less then four times are assumed. The WKBJ part was multiplied with o 9 1 j 1 in the
transmission coefficient in a non-constant field (cf.
[36]). Because of the later, this model reduces
exactly to Equation 2.58 in a constant field case. For moderate field variations | becomes close to the
constant-field model
which employs an average field y E o y o 2, Fig. 2.1.
Indirect phonon-assisted tunneling employs the model by Keldysh derived for a constant field [36, 37]
y 32 2 E 1
2 j 60
with phonon-absorption and emission related factors:
y exp r 43 2 E6r
2 j 61
-y exp r 43 2 EW
2 j 62
2-7/
:9
2
& 2
1 2
&
3 2
2143
% 2
.-0/
*
5 2
2
";
& 9 22< = 7 4<
3 4 % 5 2
7 4
@
&
3 2
&4C
3 2
2143
2143
BA
@
and
&4C
1 2
@
?>
&
@
5 2
1 2
r 1
z
1
&C
exp
2 j 63
y o
If the indirect tunneling with the variable-field model is specified, the Equation 2.60 is used assuming an
average field
2. After extensive comparisons between Equation 2.58 and Equation 2.59
we found an average-field approach to be satisfactory for direct tunneling problems if the field variations
are not too large. We may expect an average-field approach together with Equation 2.60 Equation 2.62
may be sufficient for engineering use regarding to phonon-assisted tunneling, since other effects are so
and so not accounted for, as the realistic band structure, manifoldness of the phonon branches and the
29
interface-vicinity effect.
An anisotropy in silicon has not been accounted for in the models implemented. The anisotropy effects
seem to be small at higher fields which are of interest [38].
Note that the starting-field model and the 1D approach for the tunneling direction are relevant for the
investigation purposes only.
y j
E
&
elementary charge;
1 602 10 19
for indirect band gap of silicon see Section 2.5;
for direct tunneling band gap in silicon we assume E
3 42
&
&
reduced Planks constant (Diracs constant);
2%
1 0544 10 34 D
Boltzmanns constant;
1 3804 10 23 D
2E
free-electron mass in vacuum;
9 108 10 31
1
;
reduced mass in tunneling process;
1
1
F
! ;
0 26 ,
0 1 , is used for both direct and indirect tunneling.
0 588
The effective masses in silicon are:
0 98 , conduction band electron longitudinal mass;
0 19 , conduction band electron transversal mass;
F
!
0 16 , valence band light hole mass;
!"
!
0 49 , valence band heavy hole mass.
For a possible impact of the field on the hole mass see discussion in [38].
number of valence band electrons per unit cell, taking place in tunneling;
2
;
product of the volume of unit cell and the matrix element for the character
2
istic interband phonon scattering in indirect tunneling process;
<
=
<
1 0 10 62 D 2 3 is roughly estimated according to electron-phonon scattering coupling constants extracted by MC simulation of high-field transport in
silicon [39].
&4C
optical-phonon energy;
60
y
y
y
y
y
y
&4C
.< =
<
y j q
y o y j t
y j oq
#y j
#y o o
j Cy
j
j
j
j
j
j
y q
Model description
The model is specified by the key MTUN in the RECOMB directive. The input data consists of three integer
numbers GIH
with the meaning:
30
G
H
As example we present the calculation of a characteristics which can serve as a good test for each bandto-band tunneling model in comparison with experiment [28]: generation rate, drain or bulk currents are
computed versus drain-bulk voltage, while the drain-gate bias is kept constant for individual curves and
bulk-source bias is set zero, Fig. 2.1.
A roll-off in the family of the characteristics reflects well the
10-5
U DG =6.5V
-6
10
10-7
10-8
10-9
U DG =5V
-10
10
-11
10
-12
10
tox=8nm
average field
starting field
variable field
10-13
0
Figure 2.1: MINIMOS numerical calculation of the GIDL current due to the band-to-band tunneling
versus drain bias. Gate-drain bias is held constant. Impact ionization increases the terminal currents less
than 5% in this case.
two-dimensional nature of the potential and field distributions in the critical area and the impact of the
field variations along the tunneling paths on the tunneling current [31]. One of the input files used in the
calculation is
BBT rate;
*
DEVICE
+
*
BIAS
STEP
*
PROFILE
INTERF
*
GRID
+
+
+
*
RECOMB
+
*
OPTION
OUTPUT
END
31
SGAP=0.40E-4
TN=1E9 TP=1E9
MODEL=AVAL TEMP=298.0
NONE=YES
BIN=NO ERR=1E-3
MYT=2
CN=0 CP=0
PHYS=NO
Several keys deserve attention here: LUDG which determines that both, gate and drain bias are changed
with DD simultaneously, MXT and MYT and MTUNN. PHYS=NO enables the negative gate bias to be applied
on n-channel device. With SN=0, SP=0, CN=0, CP=0 and large TN and TP we suppressed other
generation-recombination processes, except the avalanche generation, in order to check the accuracy of
the terminal current calculation.
32
NPO?QIRBSUT
div M
R
V
Z[ eff
O7QIRYS]\
QIRYS
N X
_ QIRYSUT
_XQIRBSc\
V `
a
Z [ eff
b
Q
S div M
R
Z
V
J
QedcK
QIRYSBSUT
K7f QIRYS
IQ RYS
QIRBSik
h jml
n
Z `
Z ?
div
grad
W
W g
_
Q
jUo
prq
Z
2 ^ 64
1J
J
[
where
QIRYS
VXW
2 ^ 65
S
2 ^ 66
and [ eff are the effective (net) generation rates for electrons and holes, respectively, and
is the total trapped charge. Trap-related quantities do not appear in the total-current-continuity
equation which is solved at the beginning of each time step (see Section 1.4):
K
W?g
QIRYS eff
div s
d]K
N O h
N _
V T
grad u
div M
div M ^
R t
Q
2 ^ 67
Our model allows arbitrary interface and bulk trap distributions in both the energy and position space.
jwv
For an individual single-level interface (or bulk) trap with density , located at the position xzy along the
v
channel, at energy level E , the trap dynamics is described by the equations
{ O T
jwv
Z W
| }O
~
v
\
{_wT
{ O
W
| }_
~
v
\
O T
[
W
| O
~
v
\
{_
_T
j'v
Z W
| _
~
v
\
2 ^ 68
_
where
is the electron capture rate,
the hole capture rate, [
the electron emission rate and [ is
v
the hole emission rate. The occupied trap density is denoted by W . The electron and hole capture and
emission time constants are given by
\
T v Y_ K"2_K
| }l_ 1
| } l O 1 T v O K O K i
W
`7
v
v
\
T v Y_
K _K"j'K
E} Z E
E Z E
| l O 1 T v O K O K"j } K
| l_ 1
exp s Z K~ t
exp s Z ~K t
v O
v_
2 ^ 69
S
2 ^ 70
2_
,
are the average thermal velocities towards the interface and ,
the average capture crossj
j'
T{rO
Z[
O
Z
{ _ h
[
_ T
Z
W
|
h*
K"j'v
|
2 ^ 71
where * being the steady-state occupancy function and | an effective time constant which determines the
actual trap dynamics
| l 1 T
| }lO 1 h
| l O 1 h
| }l_ 1 h
| l_ 1 \
| KQ | } l O 1 h
| l _ 1S ^
2 ^ 72
33
The trap-trap transitions, when more trap levels exist in the forbidden gap, are not accounted for in
Equation 2.71 in the present implementation. Such an approach is proper for interface traps. In spite
of their possible close location in the energy space, interface traps are separated in the position space.
The electron wave functions are very localized at the trap sites. It is believed that the wave functions of
neighbour interface traps do not overlap. However, trap-trap transitions may take place in case of bulk
traps.
The trap-dynamics equations are discretized by a strongly implicit method. The net generation during an
J
J
individual time
step, associated with the trap,
is calculated consistently with the discretization. The charge
K v
h
KaQj'v
vS
Z W
becomes Z
for donor-like traps. To solve the rate equations the
W for acceptor-like and
R
time steps (say ) for solving the basic semiconductor equations (Equation 2.64, Equation 2.65, Equation
2.66 and Equation 2.67) are divided into NTSI subintervals. The key NTSI belongs to TRANS directive.
R
The carrier concentrations (W , ` ) are exponentially interpolated in the time interval . By this approach
we can reduce the time-discretization error in the rate equations for the capture-related processes. Of
course, this refinement is irrelevant for the non-steady-state emission. Finding a compromise between the
computation time and the accuracy, controlled by the time stepping (see Section 1.4) and the parameter
NTSI, requires experience and physical reasoning for each specific class of problems. Some directions
are given in [2]. In addition to the net generation and the trapped charge associated with the individual
traps we calculate the corresponding derivations of these quantities with respect to the local carrier
concentrations and the potential.
Both interface and bulk traps are discretized in the energy space as explained below. To calculate the
total contribution from traps at the particular position the contributions from all individual trap levels are
R
superposed. The total net generation during , the trapped charge and their derivatives at the particular
position are coupled with the basic semiconductor equations. At each time step the complete system is
solved self-consistently in the whole device.
To accelerate the convergence of the discretized basic equations the derivatives of the net generation
rates and the trapped charge are taken into account. This is indispensible to achieve convergence of the
iterative algorithm when high trap-densities are assumed in device. Within the limits specified on the
INTERFACE and PROFILE directives the convergence properities are quit satisfactorily in both steadystate and transient conditions and for donor-like as well as acceptor-like interface and bulk traps at room
temperature. These limits are much higher than densities commonly taking place in practice, but the
conditions become worse at low temperature.
A more detailed discussion about the time-discretization of the trap-dynamics equations and the basic
semiconductor equations, the time-discretization error and the numerical methods can be found in [2].
In the steady-state the implemented equations reduce exactly to the well known Shockley-Read-Hall
trap-assisted recombination model.
34
AUUD, DUUD: Traps uniform in both the position and energy space.
vQ
x
ScT
exp Z
.a4
x Z
s
.aX?
2P0z
2 ^ 73
In MINIMOS the origin of the x -axis is the source-sided gate edge. Note that, although we called
AGUS the standard deviation, it is 2 times larger than this quantity in the mathematical sense.
AULD, DULD: Traps uniform in the position space and linearly distributed in the energy space.
The traps are distributed in the forbidden gap from 0 at the valence band to AULD or DULD at the
conduction band.
AUGD,AUGS,AUGX and DUGD,DUGS,DUGX: Traps uniform in the position and gaussian distributed in the energy space.
vQ
ScT
XP
E Z 2.
exp 7Zs
2
vYQ
2 ^ 74
independent of the energy discretization mesh. The location of the peak should be given with
respect to the valence band edge for both type of traps.
AGLD,AGLS,AGLX and DGLD,DGLS,DGLX: Gaussian distributed traps in the position space and
linearly in the energy space. The traps are distributed in the forbidden gap from 0 at the valence
band to AGLD or DGLD at the conduction band (for the peak in the position space) and analogous
to Equation 2.73 in the position space.
For SOI devices interface traps can be specified at the silicon-film/bulk-oxide interface.
BAUUD, BDUUD: Traps distributed uniformly along the back-interface in SOI devices, uniform in
the energy space.
For all traps the first letter determines the nature of the traps, A standing for acceptor-like and D for donorlike. The interface traps may be arbitrarily combined. The total trap density becomes a superposition of
the individual densities specified. Negative values are also allowed on input (they will be subtracted), but
the key PHYSCK=NO should be set on the OPTION directive.
35
Ja
An equidistant discretization of the energy space has been implemented
for a continuum of interface
v
g between two adjacent energy
states from the valence to the conduction band, with spacing E
levels. With this choice the equilibrium Fermi-Dirac occupancy function is well resolved. Regarding
the non-steady-state emission occupancy function , in each time decade about 8 trap levels effectively
T
ve
contribute to the signal. The number of energy levels is E
E* E . Note that E is temperature
dependent, becoming large at low temperature. By the key MENErg on the INTERFACE directive the
user can specify E externally.
CSEA, CSHA cross-section for all acceptor-like interface traps for capture of electrons and holes,
respectively
CSED, CSHD cross-section for all donor-like interface traps for capture of electrons and holes,
respectively
yv
107
QIO
vY_ T yv K
where
3 ^ 35
' , g is the thermal voltage and
models of which can be found in Section 2.5.
QI_
O
2 ^ 75
Other interface traps which can be specified by the keys FSS, PFSS, DA0 and DD0 on the INTERFACE
directive are populated in the transient conditions as in the steady-state.
36
| O p
| O n
| _ p
K~ v O K
| _ n
K v O K
K~ vY_ K
K" vY_ K
2 ^ 76
S
2 ^ 77
with capture cross-sections introduced below. Typical values for deep-donors and acceptors in GaAs are:
2
T
4
EL2
Cr
0 ^ 76 J
0 ^ 76
4
T
5 10 16
K
l
1 ^ 2 10 19
.
T
l 18 2
17 2
2 10
K
l
5 10
Traps distributed uniformly in the energy are discretized equidistantly in the energy space according to
K
T
J
the key VDEN, which belongs
to PROFILE directive. We set ..
2 g as default. The number of
Gaussian distributed bulk traps are discretized with an equidistant energy-mesh placed symmetrically
around the peak position in the 3 interval. For the acceptor-like traps the difference between adjacent
energy-levels becomes
T min
2.
S
2^ 5
S
Z
4.
max 0 2. Z 2 ^ 5
2 ^ 78
g
.42 2 as default. In common cases traps are discretized with 10 levels.
with 42
min 2
For donor-like traps an analogous formula follows. The total amount of traps per unit volume is given
exactly by
E
0
.P
exp s Z
f E l~~ 2
E for all values of VAGX and VAGS, independent of the energy
~~~ q t
discretization mesh.
The capture cross-sections of bulk traps are specified in the RECOMB directive:
37
CAN, CAP cross-section for all acceptor-like bulk traps for capture of electrons and holes, respectively
CDN, CDP cross-section for all donor-like bulk traps for capture of electrons and holes, respectively
v O T v K
with
v T
13 ^ 4 107
QI
vY_ T v K
QI _
2 ^ 79
.
The bulk traps are resolved in time as the basic semiconductor equations, with other words the key NTSI
is set to 1 internally.
2.7.4 Examples
The present model is verified against several experimental data for the charge-pumping effect. Some
comparisons with experiment can be found in [40, 41, 42].
As an example for interface traps we analyze the numerical simulation of the charge-pumping experiment
used to extract the spatial distribution of interface states [43]. MINIMOS is employed to calculate the
family of the characteristics for an assumed trap distribution J localized in the drain junction, Fig. 2.2. The
l 2
l 1
input file for the curve related to the peak density 1012
is:
38
RECOMB
*
INTERF
+
*
OPTION
+
OUTPUT
END
GRIDFR=YES
LCPUMP=YES
Input file used to calculate the characteristics presented in Fig. 2.2 (from [43]).
The input file for the curve denoted as uniform in Fig. 2.2 is the same, but without the second line in the
INTERFACE directive.
The interesting keys in this example are:
LUGB
WG=2
TG1-TG5
UG2-UG4
NTSI=2
XS1,XE1,DX1
XS2,XE2,DX2
AN,AP,SN,SP
CN,CP,TN,TP
AUUD
AGUD1
AGUS1,AGUX1
MENER
MODEL
PHYS
GRIDF
LCPUMP
PACC
The next example assumes bulk traps in an SOI pin-diode made in polycrystalline thin-film technology.
In general, using the pin-diode structure we can avoid the convergence problems which may occur when
the Gummel successive algorithm is applied to floating-body SOI devices. Pin-diode as a test structure
is also very useful for measurements in SOI technology [44]. In this example we omitted interface
traps, which normally occur in real devices. Assuming surface-trap density at the grain boundaries in the
j v
T
l 2
T
polysilicon
1 ^ 5 1012
20W , it follows the equivalent
and the average grain size c
39
1012
f=1MHz
t r =t f =0.1 s
UG =9V
UGL =-3.1V
LG=0.8 m
t ox =12nm
W=100 m
11
10
uniform
1
0
0.5
1.0
1.5
2.0
2.5
3.0
Tj
}_
3.5
for assumed
40
20
15
Ur =0.1V
UB =+20V
UG =8V
10
f=10kHz
tf =1s
tr=1s
d.cycle 50%
0
-14
-12
-10
-8
-6
-4
-2
LSOURCE=1E-4 LDRAIN=1E-4
AUUD=0 BAUUD=0 PSS=4E11
NONE=YES
TN=1E9 TP=1E9 SN=0 SP=0 VN=0 VP=0 CN=0 CP=0 AN=0 AP=0
CAN=1E-15 CAP=1E-15 CDN=1E-16 CDP=1E-16
BIN=N ERR=1E-3 PACC=1E-6
Input file used to calculate the charge-pumping curve presented in Fig. 2.3.
The key ASYM=Y defines that the second PROFILE directive is used for the drain junction. On the
second PROFILE directive ELEM=B is specified, while on the first ELEM=P is specified for the drain and
the source, respectively. To suppress the MINIMOS profile-clash error we use PHYSCK=NO. The total
simulation time covers 1 ^ 5 signal periods. We calculate the charge-pumping currents in the time interval
K~\
K~
0^ 5
1^ 5
in a post-processing step from the data written to the file units NUMCP1 and NUMCP2.
Several examples for the transient simulation including interface traps are given in [2, 40, 41, 42, 43, 45].
41
nondegenerately doped. As a consequence of the band-bending, caused by the penetration of the electric
field, these gates can no longer be assumed to be an equipotential area. The effect may have a profound
impact on the device characteristics and performance.
Two approaches to analyze the gate depletion effect are implemented in MINIMOS. The first is a full
two-dimensional solution to the problem, which can be specified by the key MPOLY=1 in the OPTION
directive. The second is an approximative one-dimensional approach based on the gradual channel
approximation in the gate, specified by MPOLY=0. In both models the activated impurity concentration
in the gate is given by the key PDOP in the OPTION directive, which also activates the gate depletion
analysis. The gate is assumed to be uniformly doped with a level specified by PDOP.
where
STj } K
1 2 s
and `
STj'K
1 2 s
2 ^ 80
j } jw
,
are the effective density of states of the conduction and valence bands, respectively. It holds
J
h
Q
S
S
\
Q
T
}
Z EuZ E
2 ^ 81
where E is the ideal band gap, discussed in Section 2.5 and E is the total band-gap narrowing. The
K" l 1 j
j }Y \
K l 1 j
jw
Q
S
T
2 ^ 82
1 2
1 2
j
for N-type and P-type gates, respectively. is the activated impurity concentration in the gate, specified
T
}
by the key PDOP. At the gate contact neutral majority carrier concentration is assumed, which equals
the activated impurity concentration. The top gate potential with respect to the Fermi level in the source
which is imposed as boundary condition is given by
} Z
E}
hQ
E} Z
Ey
2 ^ 83
with E being the band-gap narrowing at the conduction band. is the terminal voltage and E} , Ey
denote the conduction band edge and the intrinsic level in the ideal silicon band. The Fermi integral 1 2
and its inverse are calculated by analytical approximations proposed in [49, 50].
42
Figure 2.4: Schematic presentation of the simulation area for the two-dimensional model. The dot-dashed
curve is the contour referred to the field-flux calculation.
The band-gap narrowing in the gate is modeled by an empirical expression
E
j
32 ^ 0 ln 1
h
s
j
K
4 1018
h
s
j
K
9 1018
QI
S\
2 ^ 84
3 . Equation 2.84 is in agreement with rather scattered experimental data for the
where is given in
gap narrowing in the quasi-neutral regions from the literature. No temperature dependence of E has been
T
implemented. In absence of any experimental data a symmetrical gap narrowing is assumed: E }
E 2.
A criticism regarding the application of the model Equation 2.80 to heavy doped space-charge regions can
be found in [2, 51].
The height of the gate electrode can be specified by the key GCTHE in the GEOMETRY directive.
While adapting the mesh to the doping profile, before the model 1D is executed, MINIMOS will adapt
Qj S
. Additional mesh refinement in the gate
the y-mesh in the gate according to the Debye length
area takes place in the model 1D and is controlled by the maximum potential difference allowed between
adjacent grid lines, specified by the keys DPXP and DPYP in the GRID directive.
can range from roughly Z 200 in the strong accumulation to 1 ^ 2 in the strong inversion, assuming
an N-type poly-gate. The maximum difference between at the source side and the drain side is
limited on about 1 ^ 4 , whereas the corresponding difference in the channel is roughly p . Therefore,
the poly-gate may be considered as a MOS device with a small drain-source bias. In such devices
the gradual-channel-approximation (G.C.A.) and the capacitance relation ( 0 ) are reasonable
assumptions near the interface. Let us assume that the MOSFET is in saturation. The G.C.A. is valid
then in the so-called source channel-part. We may expect that it holds in the corresponding part of the
poly-gate as well (just on the other side of the oxide). The G.C.A. does not hold in the drain channel-part
and probably vis-`a-vis in the poly-gate too, however the gate has a small influence on the bulk in this
43
region. Consequently, where the gate controls the channel, the G.C.A. holds in the gate. Unlike in the
bulk, two-dimensional effects are of a second order in the poly-gate, except just at the corners and we can
assume the problem to be one-dimensional. The surface potential depends only on the local field in
the polysilicon, perpendicular to the gate/oxide interface (Fig. 2.5). Accounting for Fermi-Dirac
statistics, after one integration of the one-dimensional Poisson equation, we obtain
j }
h
} Z
}
t Z
t
3 2 s
3 2 s
g
g
T:
d_
j'
h
j
3 2 s
T
t Z
t
3 2 s
g
g
s
c
g
2 ^ 85
d_
with c
g
and } ,
defined previously [52]. For the permittivity of the single
crystal undoped silicon is assumed. The transcendental Equation 2.85 is solved by Newtons iterative
method, applying limiting and damping on the potential increment, which should converge absolutely.
Q S
For the Fermi integral 3 2 x we employed the approximation proposed in [50].
The relation between
d_
v
y
d _
v
y
\
2 ^ 86
where we accounted for the fixed charge and the trapped charge of both, donor y and acceptor
v
nature y at the gate/oxide interface. Assuming we have no interest in the gate-side effects (like
sidewall capacitances) we can neglect the actual gate thickness. The model reduces to the conventional
Q S]T
planar MOSFET simulation, but with the gate-contact potential variable along the x coordinate: x
Z . In MINIMOS, this is implemented by updating the gate potential boundary condition after each
Q S
Q S
Poisson/continuity equation solution-step by calculating x for a given local x . This calculation
is very fast compared to the two-dimensional solution of the Poisson equation in the gate. Moreover, the
additional x and grid lines necessary to resolve the gate/oxide interface, as well as grid adaptation are
avoided. This is an important savings, since the depletion region in the poly-gate is usually quite small,
due to the relatively large doping levels and requires a fine grid to resolve it. Damping of the potential
changes applied to the gate may employed to speed convergence of the simulation.
GO (x)
EGO
x
y
Figure 2.5: Simulation area for the one-dimensional model is the same as in the conventional calculation.
Q S
The boundary condition for the gate potential x is variable from 0 to . It depends on the local
Q S
normal field x in a selfconsistent manner. The dot-dashed curve refers to the field-flux calculation.
For both models MPOLY=0 and MPOLY=1, the type of the gate electrode is determined by GATE in the
DEVICE directive, while PDOP should be a positive number. All four cases, n- and p-channel devices
with N- and P-type of gate are supported.
44
fixed charge
traps uniformly distributed in the forbidden gap. Positive values implicitly
denote donor-like and negative values acceptor-like traps.
acceptor-like traps parabolically distributed in the energy space from 0 at the
For example, the density distribution for the latter trap type is given by
v Q
ST
y
v0 K
s
E} Z E
t
E} Z E
2 ^ 87
After considering several data sets, we have introduced the parabolic density-distributions to enable better
fitting of experimental data. An example is given in Fig. 2.8 (cf. in [51]).
The trap population in determined by the local Fermi level at the gate/oxide interface. Fast and stable
Gummels iterative procedure, when high trap densities are presented, is achieved by employing the
derivatives of the trapped charge with respect to the local potential to predict the new potential increment.
All charges are taken into account in both MPOLY=0 and MPOLY=1 model. For the one-dimensional
model the charge is uniformly distributed along the gate, from 0 to (Fig. 2.5). In the two-dimensional
model the charge is uniformly placed along the whole non-planar poly-gate/oxide interface (Fig. 2.4).
The traps at the grain boundaries in the polysilicon have not been taken into account in the present model.
A reason is given in [51].
2.8.4 Examples
In the following example the accuracy of the numerical technique for calculating the quasi-static gatecapacitance is demonstrated. The result obtained by MINIMOS is compared with an analytical (exact)
solution to the gate-depletion problem, Fig. 2.6. The analytical model is explained in [52]. This
comparison, of course, does not include any fitting. The input file used for this calculation reads:
45
PDop
MPoly
DG,NG
VG
GRidf
The voltage drop in the poly-gate is shown in Fig. 2.7. Because the model MPOLY=0 does agree well with
the analytical result, the discretization, as well as the numerical error are judged as very small in the oxide
and in the bulk. Differences between MPOLY=1 and the exact solution occur due to the discretization
error caused by a too coarse -mesh in the gate area. By reducing DPYP parameter the error at inversion
and accumulation in the gate can be suppressed.
In the next example the inversion part of a quasi-static gate-capacitance curve is calculated.
P-gate/P-channel device; inversion part
*
DEVICE
CHANNEL=P GATE=PPOLY TOX=12.5E-7 W=100E-4 L=99.5E-4
+
DGAP=0.5E-4 SGAP=0.5E-4
BIAS
US=0 UB=0 UD=-0.001 UG=-10.0
STEP
NG=20 DG=0.400 VG=0.025
PROFILE
FILE=2D
INTERF
NSS=1E10 PFSS=0 PNSS=1.0E12 DA0P=6.0E12 DD0P=0
MOBIL
MT=1E4 MC=2
46
2
gate capacitance (fF/ m )
tox =7nm
17
ideal
exact
mpoly=0
mpoly=1
N B =10 cm
+
18
-3
N G =8 10 cm
-3
0
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2
0.2
0.4
0.6
0.8
1.0
Figure 2.6: Comparison between numerically calculated quasi-static gate capacitance and analytical
j
T
K
l 3
8 1018
. Bulk is uniformly doped
(exact) result. Gate depletion is analyzed. Gate doping is
j T
l 3
. The ideal characteristics is the analytical results for a negligible gate depletion.
1017
with
numerical calculation versus analytical solution
1.2
tox =7nm
17
1.0
-3
N B =10 cm
+
18
-3
N G =8 10 cm
0.8
0.6
exact
mpoly=0
mpoly=1
0.4
0.2
0
-0.2
-5
-4
-3
-2
-1
Figure 2.7: Voltage drop in the gate for the same device as in the previous figure. Note that the differences
between the solid-line (analytical results) and the dashed line (MPOLY=0 MINIMOS model) are negligible.
OPTION
+
GEOM
GRID
+
OUTPUT
END
47
Input file used to calculate the inversion part of the quasi-static gate-capacitance presented in Fig. 2.8
(solid-curve).
30
25
20
15
10
experiment
calculation
tox=12.5nm
0
-10
-8
-6
-4
-2
Figure 2.8: Quasi-static gate capacitance calculated by MINIMOS using MPOLY=0 compared with
experimental data.
The keys particularly interesting in this example are: Gate, PDop, MPoly, PFss, PNss, DA0p, DA0p
and GCthe.
Both the one-dimensional and two-dimensional models of the gate depletion effect can be combined with all
other MINIMOS features, as for instance transient simulation and band-to-band tunneling.
48
the parasitic JFET between the P junctions and the epi-region, the rest of the epi-region which acts
o
as a resistor and the N (buried) layer contacted by the drain. This complex structure posses severe
difficulties to analytical modeling approaches. In particular, the channel region alone determines the
threshold voltage of VDMOSFET and the saturation transfer characteristics at low gate biases. Unlike
the conventional MOSFETs, the channel doping decreases sharply along the interface towards the drain
in VDMOS transistors. Moreover, the channel of modern power VDMOSFETs is sufficiently short that
the mobility degradation and the drift velocity saturation due to high lateral fields must be accounted for
in modeling in addition to the surface mobility variation with the perpendicular field [57, 58]. The drift
velocity saturation also occurs within the parasitic JFET [59, 60] which has a non-uniform geometry
and the length/depth ratio so low that the conventional gradual-channel-approximation becomes invalid.
It has been proposed that the carrier velocity saturation is responsible for the quasi saturation effect in
the output characteristics of VDMOSFETs [60]. Recent studies have proposed that the quasi-saturation
effect even involves more complex phenomena [57, 61]. The current flow in the rest of the epi-layer
is strongly twodimensional [62, 63]. The distributed epi-layer close to the interface where the current
flow strongly changes in direction can be either accumulated by majority carriers (in the linear and the
quasi-saturation region) or free of them (in the saturation region), which depends on the bias conditions
[64, 57, 60, 65]. The rigorous numerical approach is the only way to account for these various effects
with sufficient accuracy.
In MINIMOS model of the VDMOS structure the basic semiconductor equations
1J
N O QIRYST
div M
div
Qed
QIRYS
VXW
V
grad
O QIRYS
Z[ eff
J
QIRBSBST
f
;
QIRYS
Z
`
1J
QIRYS
N X
_ QIRYScT
div M
Z
W?g
QIRYShkj
QIRYS
VP`
V
n
Z[ eff
R
l
Z
o
p q
_XQIRYS
2 ^ 88
2 ^ 89
are solved in the rectangular domain denoted by p in Figure Fig. 2.9 (dashed line). The solution of
the complete system of equations is obtained selfconsistently applying the successive Gummel-algorithm
for the steady-state problems [66]. In the transient simulations, the decoupling Mock-scheme [67] is
employed at the beginning of each time step, followed by the time-dependent Gummel-algorithm. These
solution approaches have shown absolute stability in all simulations of low-voltage VDMOS structures
49
Figure 2.10: Doping profile of a low-voltage power VDMOSFET as interpolated onto MINIMOS grid.
we
regions,
including those of high terminal currents (large and
p ) and deep saturation biases ( p
100 ) [68]. Contrary to a general opinion that the decoupled
algorithms cannot be applied to simulate the power devices due to the strong coupling between the
equations (high current densities) see discussions in [65, 66] we have demonstrated the applicability
of the decoupled schemes on the low-voltage power VDMOSFETs in both, steady-state and transient
studies. A slow convergency in the very high-current conditions and divergency of the scheme for very
high drain biases (several hundreds volts) in the OFF-state conditions have still been observed.
complete domain
is simulated, but not the half-domain denoted by
. Since the used x -mesh is not
T
symmetrical with respect to x
u 2, differences occur between the terminal currents computed for the
left and the right source contact as a consequence of the discretization error. These differences lay under
a few percent in common cases.
For the VDMOS structure a device-independent initial guess for the potential is applied, which is
Q
h
S
Q
ST
50
values at the contacts given by the external biases. The W and ` are the local neutral
carrier concentrations
T
Q
S
T
2
g ln W~Wy is the built-in
at equilibrium (W`0 W y ), which are determined by the net doping and y
potential. This equation follows from the steady-state carrier continuity equations assuming vanishing
W and `
`7 . For the initial
recombination rates and drift-diffusion transport model, after setting W
electron and hole concentration the neutral equilibrium values are assumed. We found this initial guess
very suitable for VDMOSFETs. Other an approach has been proposed in [65], which relies on solving
the Laplace equation for the initial potential.
Starting from the initial solution, MINIMOS solves the coupled Poisson and the carrier continuity equation
for the minority carriers in the complete domain p , while the concentration of the majority carriers is set
according to the local potential assuming a constant quasi-Fermi level. For VDMOSFETs this quasi-Fermi
level is governed by the bias at the source contact, but not by the bulk-contact bias as in simulation of bulk
MOSFETs. Note that the width of the analyzed domain p must be sufficiently large so that the source
o
o
contact which connects the N area also extends over the P region, as shown in Figure Fig. 2.9. After
a given accuracy is achieved, both carrier continuity equations are solved selfconsistently along with the
Poisson equation including generation-recombination processes. Since the majority carrier currents are
small in VDMOSFETs when the impact-ionization rate is low, the later solution step only requires a few
iteration cycles.
All physical models and features implemented in MINIMOS are normally applicable for VDMOSFET
simulation (e.g. AC analysis).
The MINIMOS approach provides a rigorous model for the terminal electrical characteristics of VDMOSFETs. It is convenient to study the resistivity in ON-state and its temperature dependence, the
quasi-saturation effect, the transfer characteristics including the short-channel effects, the switching characteristics and the dynamics effects of slow body-drain diode.
The VDMOSFET simulation feature is activated by key LVDM=YES on the OPTION directive.
2 90
51
ec
2 91
2
For instance, if we consider a valley located on the 100 axis the effective masses are
which has the non-parabolicity factor as the only parameter. The band-form function is related to the
, with
and
as the longitudinal and transversal effective masses, respectively. In
MINIMOS, the Herring and Vogt transformation is introduced, which reduces the ellipsoidal equienergetic
surfaces to spheres:
2 92
2 93
/.
+
0 0
0
0
1
0
0 1
+
$ &6"
Three transformation matrices
are required to account for the three possible orientations of the
ellipsoidal valleys. Consequently, the band-form function in the starred space becomes of spherical type,
2 94
2
1
Note, that one is free in the choice of the reference mass occurring in the denominator of Equation 2.94.
such that in the transformation matrices (Equation 2.93) two
We have chosen this mass equal to
diagonal elements are always unity.
The parameters of the band-structure can be specified on the MONTE-CARLO directive by the following
keys:
Key
MLONG
MTRANS
NONPAR
Symbol
Description
Longitudinal effective electron mass
Transversal effective electron mass
Non-parabolicity factor
Default Setting
0 9163
0 1905
1
05
798;:
52
Acoustic deformation potential scattering (ADP): For ADP scattering, which introduces intravalley
transitions only, the elastic approximation is assumed. The integrated scattering probability reads
ACBEDGF
>= 2 ?I @ J KML H
+
<
N
2 95
? O @ denotes the density-of-states effective mass, F H the acoustic deformation
where
potential, J K the longitudinal sound velocity and L the semiconductor mass density. In the elastic
3 2
ec
2 1 3
F H
JK
L
Symbol
BED
Description
Acoustic deformation potential
Longitudinal sound velocity
Semiconductor mass density
Default Setting
90
90 33 104
23 28 10 4
7 8 Q 5RTS
P V U R Q
P :
Intervalley phonon scattering: Electron transitions between states in two different equivalent valleys
can be induced by both acoustic and optical mode phonon scattering. In the following
denotes the
number of possible final equivalent valleys. For so-called g-type scattering this number equals to one
WX
WNX
<
ec
? @ F Ym ZM[ ]
NW X 2 I L Z\[ ^] + X 1 X 1 2 X
=
ZM[
X
3 2
WX
2 96
2 97
? @ F _ Ym ZM[
NW X 2 I L ` ZM[ ]
X 1 X 1 2 X a
1
2 98
+
=
Nb ZM[
X
2 99
F
`Z\[
In these equations, Y is the coupling constant, the phonon energy and X the electron energy after
\
Z
[
cZ\[
scattering. ] is the average phonon occupation number for the phonon ,
Z [ b :
Z\[ ed
d
AMBED/f 1f
exp
2 100
]
<
e]
3 2
F _Ym
F Y
_m
F Y
m
F Y
_m
F Y
m
F Y
_m
Symbol
1
2
3
4
5
6
798 RR QQ
798 R Q
798 R Q
798 R Q
798 R Q
798
Default Setting
0 5 108
0 8 108
11 0 108
0 3 108
2 0 108
2 0 108
P
P
P
P
P
P
53
cZ
cZ
cZ
cZ
cZ
c Z
Key
EPH1
EPH2
EPH3
EPH4
EPH5
EPH6
Symbol
1
2
3
4
5
6
Default Setting
12 1 10 3
18 5 10 3
62 0 10 3
19 0 10 3
47 4 10 3
59 0 10 3
P :
P :
P :
P :
P :
P :
798
798
798
798
798
97 8
Mode
TA
LA
LO
TA
LA
TO
WNX
1
1
1
4
4
4
Ionized Impurity Scattering: Ionized Impurity Scattering is basically treated by the Brooks and Herring
approach. Additionally, the influence of scattering events having a large impact parameter is reduced
according to the reasoning of Ridley [69]. The starting point is the total scattering rate after Brooks and
Herring,
ih
jVk @ I e] [ l"7 ? m 1n 1 1 2
2 101
4 1m
+
2
=
1
n
m n ?po
2 102
q
2
[
m
Here, ] is the concentration of charged impurities, h
jVk is a dimensionless weighting factor and n is
defined through the screening length o q . The screening length equals to the Debye length weighted by a
dimensionless factor hsrVtvuwVw
x ,
VAMB
X
@
o q yhcrVtzu
w
wVx|~ {}} K
2 103
7
@
:
], @ X
2 104
B
from the electron concentration by solving
where
is the electron temperature, and is calculated
< Ag
ec
3 2
1 2
1 2
1 2
Equation 2.104.
1 b exp d`b < Ag e fN
2 105
yhVzrv 2I ] [ : @
2 106
In these equations, denotes the electron group velocity and the average distance between ions. The
latter can again be weighted by a dimensionless factor h
vrz .
Ionized impurity scattering behaves strongly anisotropic. In MINIMOS, the scattering angle is calculated
from a random number evenly distributed between 0 and 1 according to the following equations,
]b ln ^ 1 b exp d b < Ag e fa
2 107
Ag eb
<
cos 1 b 2 < Ag
2 108
e
4 1m n
<
ec
1 3
54
I ?P o 27
2
K ] [ ] ?
Surface Roughness Scattering: Surface roughness scattering is currently modeled after [71]
<
2 109
] ?
where is the root-mean-square deviation of the surface from flatness and is the correlation length of
roughness. In MINIMOS the gradual channel approximation is employed to evaluate
in Equation
2.109, and additional fitting parameters are introduced such that we end up with the following expression,
I ?P o 7 d K_ f m d m f n
2 110
m
K
m max
b m P 0
2 111
Equation 2.110 admits a general power law to be specified for the normal-field dependence. As a
consequence of the introduction of a general exponent one has to introduce some reference field strength
m . Note that for 2 the value assigned to m has no influence on the scattering rate. In Equation
2.111 is defined as unity vector normal to the semiconductor surface.
The lateral dimension of the region in which surface roughness scattering may occur is determined by the
m be positive. The vertical dimension of this region is a fixed distance
condition that the pressing force
<
2 2
ref
2
ref
ref
ref
Symbol
Description
Depth to which surface scattering may occur
Standard deviation times correlation length
Exponent of normal-field dependence
Reference normal-field for 2
P o
m
ref
P : QQ
P :
RQ
8
Default Setting
25 10 7
3 10 15 2
20
105
2 112
yh
w
m
2 113
<z e is the
Here, WBII is a constant prefactor, is the threshold energy for impact ionization, and
m denotes the band-gap energy. The parameters of
phonon scattering rate at the threshold energy .
impact ionization can be specified on the MONTE-CARLO directive by the following keys:
Key
EXII
WBII
WEII
Description
Exponent in Keldysh formula
Prefactor in Keldysh formula
Weight factor for threshold energy
Default Setting
20
3 9 10 2
10
P :
55
8 RQ
*
OPTION
MONTE-CARLO
STEP
END
8 RQ
10
Monte Carlo
Chynoweth
Ionization Coefficient (1/cm)
10
10
10
10
1.0
2.0
3.0
4.0
5.0
6.0
-6
Figure 2.11: Electron ionization coefficient from Monte Carlo simulation in comparison with Chynoweths
law.
P Q :
P 8 RQ
Fig. 2.11 compares the results from a Monte Carlo simulation with the analytic Chynoweth model. The
1
7 03 105
1 231 106
[73].
parameters of the latter model are
div e grad u
7 b b ]
2 114
56
b^ 2
b^
1
7 div
7 7
7 E d`b grad 1 grad VAMB f
7B
M
A
2 115
2 116
2 117
2 118
2 119
div
Three coefficients of this equation set, namely electron mobility, electron temperature and electron
ionization coefficient, are extracted from Monte Carlo simulation. Monte Carlo generated averages
are converted into these coefficients as follows:
7 E R9 z Y
B V A 1 [ [
] <c[O [[
e
2 120
2 121
2 122
In Equation 2.122
2).
The remaining coefficients in Equation 2.114 to Equation 2.119 are modeled analytically (Section Section
2.1). In particular, holes are in any case described by the conventional drift-diffusion current relation
(
).
B BD
According to Fig. 2.12 the whole simulation domain is labeled 0 . The subdomain 1 is called the
Monte Carlo window. Within this window particle trajectories are calculated and averages are recorded.
The subdomain 2 is called coefficient-window, because only inside this window the Monte Carlo and
drift-diffusion models are actually coupled by means of the coefficients given by Equation 2.120 to
Equation 2.122.
The boundaries of the simulation domain
keys on the DEVICE directive.
To impose safe boundary conditions the Monte Carlo window 1 has to be chosen such as to include
parts of the heavily doped source and drain areas. Electrons are randomly injected at the left boundary of
57
G
S
D2
D1
D0
1 according to an equilibrium distribution function. Electrons arriving at the right boundary are either
reflected or absorbed, depending on the BDR (back diffusion ratio) key on the MONTE-CARLO directive.
The distance between the left boundary of 1 and the vertical source junction can be specified by the
SOVL (source overlap) key. Analogously, the distance between the vertical drain junction and the right
boundary of 1 can be specified by the DOVL (drain overlap) key. Note that an increased overlap of the
Monte Carlo window with the source and drain areas increases physical accuracy but decreases efficiency,
and vice versa.
The coefficient window 2 is determined by certain criteria. Since for both mobility and temperature
different criteria are used, different windows are obtained. In what follows we shall call them mobilty
and temperature windows, respectively. A mesh point lies inside the mobility window if the average
momentum loss rate exceeds a given minimum value.
d p f
Vx
^
7
P
2 123
coll
The default setting of the key DFMIN is 2000 8 R Q . A point lies inside the temperature window if the
carrier temperature exceeds the lattice temperature by a certain amount.
B * B D P Vx
2 124
UTMIN is a factor close to unity. Default setting of this key is 1 05. Additionally, only those points
identified as carrying statistically reliable averages can contribute to a coefficient window. If the number
of scattering events assigned to a certain mesh point is less than CNTMIN times the number of total
scattering events (given by NUM or NUM1,NUM2 ) then this point is excluded from any coefficient
window.
pp
Note that in any case the Monte Carlo window should be larger than the coefficient windows. In particular
the right boundary should lie deep enough in the drain such that the Monte Carlo window covers the whole
area in which according to Equation 2.124 elevated carrier temperatures exist.
58
pp
The iterative procedure can by tuned by the keys NUM1 NUM5. Note that the very first Monte Carlo
iteration uses an initial potential profile which results from a standard drift-diffusion simulation. As the
initial potential may deviate from the self-consistent one the Monte Carlo results of this iteration may
deviate from the final results as well. Therefore, the results of the very first Monte Carlo iterations need
not be calculated as accurately as the results of the final iterations. Assume one specifies MCPIT 4 (four
self-consistent iterations) and NUMB 40 106. The total number of scattering events to be processed for
this bias point would be 160 106 . A similar accuracy can be achieved by processing 100 106 scattering
events if one assigns the numbers 15 106 15 106 30 106 40 106 to the keys NUM1 through NUM5,
respectively.
P P P P
are the
the following denotes a generic physical quantity, the total scattering rate, and
electrons position and momentum immediately before the -th scattering event. If is the volume of the
-th cell of the auxiliary grid the average within this cell is defined by
<
[ < : [ [
< [
l v : [
[ [
2 125
59
Once the averages have been evaluated on the auxiliary grid they are assigned to the MINIMOS discretization grid by means of a convolution method. The assignment function used for this step is
Xc No T1oN P 1
b oN oN
2
2
I
2I
P
1
cos
o
o;
b oN o
2
2
cos
2 126
o of
Outside the interval given by Equation 2.127 the assignment function vanishes. The lateral width
the assignment function can be specified by the XWIDTH key, the vertical width oN by the YWIDTH key.
2 127
In order not to introduce an additional error due to charge assignment the widths of the assignment
function needs to be chosen carefully. If a Monte Carlo generated profile is very smooth the filter widths
are probably too large. If there is much noise on such a profile the filter widths are too small with respect
to the number of processed scattering events. In this case one can either increase the filter widths, which
does not increase accuracy, or one can increase the number of scattering events.
The number of points of the auxiliary grid is predefined. The respective parameters are MAXNFX and
MAXNFY in the file mccom.inc. The left and right boundaries of the auxiliary grid can be specified by
the AUXS and AUXD keys, respectively. The auxiliary grid can be smaller than the Monte Carlo window.
However, one must be aware that no averages are recorded outside this grid. Therefore, the auxiliary grid
has to be larger than any of the coefficient windows. Otherwise, the latter are clipped by the boundaries
of the grid.
R1
R2
R1
Figure 2.13: Decomposition of the Monte Carlo window for the purpose of trajectory multiplication
the Monte Carlo window divided into three regions. Following the notation of [74] denotes the region
of common events, in 1 and 2 rare events occur. Note that regions and 1 are not contiguous.
60
algorithm in MINIMOS ensures that in a certain region all particles have the same weight, regardless of
their history.
The user can specify the number of split boundaries by the NSB key. For example, in Fig. 2.13 the Monte
Carlo window is divided into three regions, hence the number of split boundaries is NSB=2. NSB can be
interpreted as that number of boundaries a particle crosses on its way from the outermost to the innermost
region. Setting NSB=0 switches off particle splitting in a simulation.
Once the number of split boundaries is given MINIMOS tries to find optimized locations for these
boundaries both on the source and drains sides of a device. The splitting or multiplication factors are
chosen by MINIMOS automatically, as well.
In energy domain a carrier trajectory is multiplied if the energy threshold for impact ionization is exceeded.
The multiplication factor in that case can be specified by the EMULT key. Setting EMULT=1 disables
particle splitting in the energy domain.
Generally speaking, Monte Carlo simulation is feasible only for devices in the on-state. The larger the
currents the better the Monte Carlo technique will perform. The closer a bias point lies to turn-off, e.g.
the closer gate voltage lies to threshold voltage, the more important is the role of the particle weighting
algorithm. This algorithm can be thought to increase the injection probability of electrons over the built-in
barrier present between source and channel. The algorithm can furthermore be thought to reduce the
fraction of time spent for the simulation of cold electrons in the heavily doped source and drain regions,
and to increase the fraction of simulation time spent for the hot electrons in the pinch-off region.
An upper bounding function to the physical scattering rate is constructed by a piece-wise linear function
on the 2 axis. In MINIMOS, the bounding function comprises three linear segments of the form
2 128
Due to impurity and surface roughness scattering the parameters and are chosen spatially dependent.
The choice of these parameters can be controlled by the E1SELF and E2SELF keys. At each mesh point
the parameters and for the first linear segment are determined from the physical scattering rate in the
energy interval 0
E1SELF. To set up the parameters for the third linear segment the energy interval
E1SELF E2SELF is considered.
3
Reference of Input Directives
The input parameters for MINIMOS are specified by a sequence of lines in a formatted data file. The first
72 characters are used to parse information and 80 characters are read and written to the output file to
offer the user the possibility of sequencing his/her statements. The first line in the data file is interpreted
as a title statement. All text on this line is used to identify the printout pages. This title line must not be
omitted. All the other statements conform to a simple syntax, namely:
DIRECTIVE1
DIRECTIVE2
..........
DIRECTIVEj
Abbreviations are allowed. Valid delimiters for the keys within the directive are blanks, commas, or
asterisks. Any directive can be continued in the next line. In this case the continuation line has to start
with a plus in the first column. Comment information can be placed anywhere in the input stream.
Start of comment is indicated by an asterisk. Note that comments cannot be continued by a + sign on the
next line. Thus, a line starting with an asterisk is totally interpreted as a comment line and an asterisk in
any column of a directive indicates that the remainder of this line is comment. Comment is allowed within
continuation lines. All blank lines are ignored but not allowed within a continued directive. Blanks in
front of a directive are ignored.
The BIAS, DEVICE and PROFILE directive with appropriate keys are required in any MINIMOS input
file. All other directives are optional. The directives may appear in arbitrary order in the input file, except
the END directive which has to be the last statement to indicate the end of the input stream. One excetpion
is when material characteristics are calculated by the Monte-Carlo model (MONTE-CARLO MCMOD=2).
61
62
Then no device,no bias and no profile have to be specified since uniform conditions are assumed. In that
case the OPTION and the STEP directive are sufficient for a valid input file.
The following sections are a comprehensive description of all possible directives and keys that may be
specified on the MINIMOS input deck.
The significant characters of the individual keys are printed enlarged. Variable types, default values,
ranges, units etc. are given in tables.
The variable types are
<C> Single character
<I> Integer Number
Note that MINIMOS reads all numbers integer or float as real numbers. Thus, an <I>
specification in the keydescription table is just a meaningful recommendation.
<L> Logical
Valid specifications are the character strings YES and NO only, of which only the first character is
significant.
<R> Real Number
<S> Character string
Directive BIAS
b r
rGb j
The BIAS directive is used to specify the operating point for which the simulation should be performed.
and
must not be
The input data will be checked for physical plausibility. Therefore
negative for nchannel devices and not positive for pchannel devices. This plausibility check can be
suppressed by setting PHYSCK=NO on the OPTION directive.
BIAS
63
<R>
US
8b
20
UB
20
<R>
(obligatory key)
8b
20
UD
20
8b
<R>
0
8b
20
US
20
The following two keys refer to AC-analysis calculations. They specify the real and imaginary parts of
the voltage applied to the terminal in question. The magnitude should be unimportant due to the linear
dependence, but the angle between VR and VI may influence the iteration history. If both keys are omitted
no AC-analysis is carried out. At least one of the keys VR and VI must be specified, if AC-analysis is to
be carried out.
<R>
0
VI
10
<R>
0
VR
10
64
Directive DEVICE
The DEVICE directive is used to specify the principal physical device parameters.
Units:
Range:
10 5 BULK 2 10 3
Q
P :
Q
<R>
0
Q
10
:
6
10
10
E1
<R>
0
Q
10
:
6
E2
Units:
Range:
2 10 6 FILM 5 10 4
Q
P :
P :
DEVICE
65
b 8
b 8
b
7 8
Range:
1 KBULK 30
L Gate Length
Q
66
Q P
:
Q
P :
W Channel Width
Type:
Default:
Units:
Range:
<R>
(obligatory key)
Q
P :
5 10
Q
b :
Q
DEVICE
67
Units:
Range:
10 5 Y 10 2
Q
68
Directive END
The END directive is used to signal the end of the input file. This directive may be omitted. If present,
however, it must be the last directive.
Range:
10 6 CAPERR 10 2
<R>
10 8
10 10
CACCREL
10
This key specifies the error norm threshold of the simulation. See Section 1.11 for information about the
error norms in MINIMOS. To terminate a Gummel iteration loop the error norms under consideration have
to be smaller than ERROR. Error norms are computed from the increments of the Poisson equation, the
minority carrier continuity equation and additionally for the majority carrier equation if MODEL=AVAL
or MODEL=HOT is specified on the OPTION directive. Note that all above mentioned error norms are
dimensionless real numbers and are an integral error measure. Beside those norms the F-norm is used,
an error measure, which is computed as the local maximum of the relative space charge increment. If
MODEL=THRES is specified ERROR additionally specifies:
_ :
2. The limit for the threshold voltage increment in 8 . The threshold iteration is terminated if any of
Type:
Default:
Units:
Range:
<R>
10 2
10 5
ERROR
0 5
END
69
<I>
20
1 NUPDATES
100
<R>
10 3
10 10
PACCREL
10
70
<I>
20
1 3H
500
100
<I>
20
1 3U
FIMPLANT
71
Directive FIMPLANT
The FIMPLANT directive is used to specify a field implantation at the channel edge in the channel-width
direction. The edge is assumed at the zcoordinate zero. The available keys are handled analogously to
the keys within the IMPLANT directive.
Note: This directive is meaningful only for threedimensional simulation.
Unit:
Range:
10 AKEV 300
78
2
Unit:
9
Range:
10
DOSE 1018
Q :
Range:
AS Arsenic
B Boron
BE Beryllium
MG Magnesium
O Oxygen
P Phosphorus
SB Antimony
SI Silicon
Unit:
Range:
800 TEMP 1300
72
Unit:
Range:
60 TIME 48 3600
GEOMETRY
73
Directive GEOMETRY
The GEOMETRY directive is used to specify various geometric parameters. This directive can be omitted.
Units:
Range:
0 BEAKL 10 2
Q
<R>
Q
DCGRD
10
<R>
Q
DDREOX
10
<R>
Q
DRGRD
10
Units:
Range:
0 D1GAP 10 2
Q
Units:
Range:
0 D2GAP 10 2
Q
74
This key specifies the thickness of the gate contact. The default value for MOSFETs is TINS/4, whereas
for MESFETs 0 1 .
Type:
<R>
Default:
Built-in Model
Units:
Range:
0 GC 10 3
<R>
G1CGRD
10
<R>
G2CGRD
10
<R>
10
LD
10
<R>
10
LS
10
<R>
OXIDEF
2 10
<R>
RECESS
10
<R>
R1GRD
10
GEOMETRY
75
<R>
R2GRD
10
<R>
SCGRD
10
<R>
SDREOX
10
<R>
SRGRD
10
Units:
Range:
0 S1GAP 10 2
Units:
Range:
0 S2GAP 10 2
<R>
TD
10
<R>
TGATE
10
76
<R>
TS
10
<R>
T1GRD
10
<R>
T2GRD
10
<R>
10
YMDEP
10
GRID
77
Directive GRID
The GRID directive is used to modify parameters for the initial grid and for the grid refinement. This
directive may be omitted.
78
GRID
79
80
<R>
10 7
10
GGMI
10
<R>
0 25 10
2 10
9
PGMI
10
GRID
81
for the y-direction used to filter the band-to-band tunneling generation rates is
<R>
1
0 5 SYT
The following keys can be used for defining supplementary gridlines in certain areas of the device.
Units:
Range:
10 10 4 XE1 10 10 4
Units:
Range:
10 10 4 XE2 10 10 4
Units:
Range:
10 10 4 XE3 10 10 4
Units:
Range:
10 10 4 XS1 10 10 4
Units:
Range:
10 10 4 XS2 10 10 4
82
Units:
Range:
10 10 4 XS3 10 10 4
Units:
Range:
5 10 4 YE 10 3
Units:
Range:
5 10 4 YS 10 3
Units:
Range:
2 10 3 ZE 10 3
Units:
Range:
2 10 3 ZS 10 3
IMPLANT
83
Directive IMPLANT
The IMPLANT directive is used to specify a channel implantation. This directive can be omitted if a
homogeneous doped channel area is desired. One IMPLANT directive is necessary for a single channel
implanted device, and two IMPLANT statements are required for a double channelimplanted device.
However, only one set of anneal parameters is allowed, as both implants are always annealed together.
The channel and the sourcedrain temperature treatments are totally independent and do not at all affect
each other. The channel anneal parameters can be specified on the first or the second IMPLANT directive.
If FILE=2-D or FILE=1-D is specified on the PROFILE directive, no IMPLANT directive at all is
allowed.
The channel implantation is assumed to be performed through an insulator of material INS with thickness
TINS (cf. DEVICE directive). If a permittivity number is explicitly given with the INS key,
2 is
assumed for the insulator if INS=3 9 and 3 4 is assumed for any other value of INS.
#2\
1
1 1018
Units:
Range:
0 5 AC 1
Units:
Range:
10 AKEV 300
2
Units:
Range:
109 DOSE 1018
84
Range:
A Arsenic
B Boron
BE Beryllium
M Magnesium
O Oxygen
P Phosphorus
SB Antimony
S Silicon
LD Diffusion Length
The LD key can be used to determine the extent of profile broadening due to annealing. If this key
is specified the TEMP and the TIME key are not effective. This key effects each IMPLANT statement
separately.
Type:
<R>
Default:
Units:
Range:
0 LD 10 4
Units:
Range:
500 TEMP 1300
Units:
Range:
60 TIME 48 3600
INTERFACE
85
Directive INTERFACE
The INTERFACE directive is used to specify the charge and traps at the semiconductor/insulator interfaces.
MINIMOS offers different possibilities to specify interface states and fixed oxide charge. All charge
profiles are independent, in other words the total charge density at the interface is given by superposition
of all specified densities. One elementary charge is assumed per state. The iterative procedure should
converge when the total trap density lies in the proposed ranges in common cases at room temperature. For
traps which are accounted for in both, steady-state and transient analysis it is possible to specify negative
values for densities, but PHYSCK=NO should be used. These densities will be subtracted from the total
density. More explanation is given in Section 2.7.
The INTERFACE directive can be omitted.
<R>
0
10
2
13
NRD
1013
<R>
0
10
3
NRE
10
<R>
0
10
3
NRS
10
86
Units:
Range:
10 3 XISS 10 2
INTERFACE
87
88
INTERFACE
89
Interface traps at the gate-oxide/bulk interface accounted for in the transient analysis:
Range:
1 MENE 350
<R>
0
0 AUUD
2
1014
ARUD1 Density of Acceptor-like Traps; Space Rectangular; Energy Uniform (first set)
Type:
Default:
Units:
Range:
<R>
0
0 ARUD1 2 10
2
13
ARUS1 Starting Point for Acceptor-like Traps; Space Rectangular; Energy Uniform (first set)
Type:
Default:
Units:
Range:
<R>
10
3
ARUS1
10
ARUE1 Endpoint for Acceptor-like Traps; Space Rectangular; Energy Uniform (first set)
Type:
Default:
Units:
Range:
<R>
10
3
ARUE1
10
ARUD2 Density of Acceptor-like Traps; Space Rectangular; Energy Uniform (second set)
Type:
Default:
Units:
Range:
<R>
0
0 ARUD2 2 10
2
13
ARUS2 Starting Point for Acceptor-like Traps; Space Rectangular; Energy Uniform (second set)
Type:
Default:
Units:
Range:
<R>
10
3
ARUS2
10
ARUE2 Endpoint for Acceptor-like Traps; Space Rectangular; Energy Uniform (second set)
Type:
Default:
Units:
Range:
<R>
10
3
ARUE2
10
90
AGUD1 Peak Density of Acceptor-like Traps; Space Gaussian; Energy Uniform (first set)
Type:
Default:
Units:
Range:
<R>
0
0 AGUD1 2 10
2
13
AGUS1 Standard Deviation for Acceptor-like Traps; Space Gaussian; Energy Uniform (first set)
Type:
Default:
Units:
Range:
<R>
10
AGUS1
10
AGUX1 Peak Position for Acceptor-like Traps; Space Gaussian; Energy Uniform (first set)
Type:
Default:
Units:
Range:
<R>
10
3
AGUX1
10
AGUD2 Peak Density of Acceptor-like Traps; Space Gaussian; Energy Uniform (second set)
Type:
Default:
Units:
Range:
<R>
0
0 AGUD2 2 10
2
13
AGUS2 Standard Deviation for Acceptor-like Traps; Space Gaussian; Energy Uniform (second set)
Type:
Default:
Units:
Range:
<R>
10
AGUS2
10
AGUX2 Peak Position for Acceptor-like Traps; Space Gaussian; Energy Uniform (second set)
Type:
Default:
Units:
Range:
<R>
10
3
AGUX2
10
<R>
0
0 AUGD
2
1014
INTERFACE
91
AUGS Standard Deviation in Energy-space for Acceptor-like Traps; Space Uniform; Energy Gaussian
Type:
Default:
Units:
Range:
<R>
10
AUGS
AUGX Peak Position in Energy Space for Acceptor-like Traps: Space Uniform; Energy Gaussian
This enery has to be specified relatively to the valence band edge.
Type:
<R>
Default:
Units:
Range:
1 AUGX 2
AGLS Standard Deviation for Acceptor-like Traps; Space Gaussian; Energy Linear
Type:
Default:
Units:
Range:
<R>
10
AGLS
10
AGLX Peak Position for Acceptor-like Traps; Space Gaussian; Energy Linear
Type:
Default:
Units:
Range:
<R>
10
3
AGLX
10
92
<R>
0
0 DUUD
2
1014
<R>
0
0 DGUD 2 10
2
13
DGUS Standard Deviation for Donor-like Traps; Space Gaussian; Energy Uniform
Type:
Default:
Units:
Range:
<R>
10
DGUS
10
DGUX Peak Position for Donor-like Traps; Space Gaussian; Energy Uniform
Type:
Default:
Units:
Range:
<R>
10
DGUX
10
<R>
0
0 DUGD
2
1014
DUGS Standard Deviation in Energy Space for Donor-like Traps; SpaceUniform; Energy Gaussian
Type:
Default:
Units:
Range:
<R>
9
10
DUGS
DUGX Peak Position in Energy Space for Donor-like Traps: Space Uniform; Energy Gaussian
This energy has to be specified with respect to the valence band edge.
Type:
<R>
Default:
Units:
Range:
1 DUGX 2
INTERFACE
93
<R>
0
0 DGLD 2 10
2
13
DGLS Standard Deviation for Donor-like Traps; Space Gaussian; Energy Linear
Type:
Default:
Units:
Range:
<R>
10
DGLS
10
DGLX Peak Position for Donor-like Traps; Space Gaussian; Energy Linear
Type:
Default:
Units:
Range:
<R>
10
DGLX
10
At present the traps are uniform in energy, acceptor-like and will be placed at the gate-oxide/bulk interface.
&
The data are read from an external file (file unit NUMETP), which has two columns containing !#"%$ (' .
1 , respectively.
The units of the data have to be and 2
Type:
<I>
Default:
0
Range:
0 no external traps
1 acceptor-like traps; energy uniform
G
94
Directive MOBILITY
The MOBILITY directive is used to specify parameters that selectively stress or attenuate the various
scattering mechanisms. The physical meaning of the parameters given below are explained in detail in
Section 2.1 and Section 2.2. We recommend to use those weights only for fine tuning, since the default
values in the code are chosen carefully and agree rather well to experiments. This is particularly true for
silicon devices. For galliumarsenide devices only the MB, MI, ML, MT and MV keys are meaningful due
to a different mobility model, Section 2.2.
This directive can be omitted. No key of this directive is required. All values are dimensionless real
numbers.
MB Generic Weight
For Silicon, this key specifies the weight for the exponent in the relation between the drift velocity and the
electric field. A higher value leads to a shorter transition from ohmic mobility to saturation velocity. For
gallium arsenide this key is used to weight the minimum mobility in the mobility impurity concentration
relation.
Type:
<R>
Default:
1 for Si devices; 0 for GaAs devices
Range:
0 MB 3
MOBILITY
95
96
Directive MONTE-CARLO
AUXS Source Sided Extension of Auxiliary Grid
Type:
Default:
Units:
Range:
<R>
0 1 10
AUXS
0 5 10
<R>
0 2 10
AUXD
0 5 10
9 9
MONTE-CARLO
97
<R>
90
DEFPOT
102
<R>
25 10
DEPSS
0 1 10
<R>
2000
9
0
1
DFMIN
105
<R>
0 3 10
0 DOVL 2 10
4
<R>
01
10
E1SELF
1.0
<R>
50
E2SELF
10
<I>
10
1
1 EMULT
104
98
9
<R>
105
2
EREF
106
<R>
2
1
0 2 EXII
5 0
MCMOD
1
2
3
4
5
In MCMOD=2 material characteristics are calculated for a range of electric field strengths and for a range
of doping levels. These ranges can be specified on the STEP directive.
Type:
<I>
Default:
1
Range:
1 MCMOD 5
MONTE-CARLO
99
<I>
20 106 , MCMOD=1
106 , MCMOD=2
105 NUMB 5 108
100
MCPIT
Range:
1 SAV
<I>
78
1 SEED
168
<R>
0 3 10
0 SOVL 2 10
4
<R>
1 05
1
1 UTMIN
10
MONTE-CARLO
101
<R>
90 33 104
10 VSOUND
5
106
<R>
3 9 10 2
1
0 WBII
103
<R>
1
1
10 2
WEII
102
102
<R>
Minimum of <1 10 and 50 10
10 XWIDTH 0 5 10
<R>
Minimum of <
10 YWIDTH 0 5 10
50 and 10 10
OPTION
103
Directive OPTION
The OPTION directive serves as a directory for general purpose keys. Some physical device parameters
are collected herein as well as parameters to control MINIMOS execution. This directive can be omitted.
No key of this directive is required.
3
Units:
5
Range:
10
CCLP 1020
<L>
NO
104
If GCHCALC is specified, the total charge in the gate will be calculated by evaluating the electric field flux
, the source-sided charge U DVWV ,
through a contour around the gate. The results are the total charge U
the bottom charge U (X and the drain-sided charge U (YZV . The result is written to file unit NUMCG1 in the
&
form =>?A@B?2?AC(U[
U[DV\V8U (X U (YZV ' .
Type:
<L>
Default:
NO
9 @
1
&`_ ]Qa
^
'
cb
3
This key is only effective in case of MESFET simulation. If this key is omitted an ideal gate diode will be
assumed.
Type:
<R>
Default:
1
Units:
Range:
1 ID 5
OPTION
105
Units:
Range:
10 6 LDRAIN 0 1
Units:
Range:
10 6 LSOURCE 0 1
106
Range:
1N
1O
2N
2O
3
Units:
Range:
1014 PDOP 1021
OPTION
107
Range:
0 RBULK 105
Range:
0 RDRAIN 100
This key specifies an effective source resistor. When this key is given, the effective source bias is calculated
xw
as eqt ugqt4veg
sj8C . For US see description of the BIAS directive. As a large value of RSOURCE can
affect the computation time in an undesirable manner it is recommended to determine the effective source
bias by the US key directly, provided that j8C is known. This is not applicable if the STEP directive is
given.
Type:
<R>
Default:
0
Units:
Range:
0 RSOURCE 100
<R>
300
y
77
TEMP
450
108
Range:
0 UN 1
Range:
0 UP 1
OUTPUT
109
Directive OUTPUT
The OUTPUT directive is used to specify the physical quantities which are to be printed after the simulation
has terminated successfully. All keys possible on this directive are logical items. Thus, the only decision
requested is if some physical quantity is to be printed or not.
<L>
YES
<L>
NO
<L>
YES
<L>
NO
<L>
YES
<L>
YES
<L>
YES
110
<L>
NO
<L>
NO
<L>
YES
<L>
NO
<L>
YES
OUTPUT
111
<L>
NO
112
Directive PROFILE
The PROFILE directive is used to specify the source and drain doping profiles, the bulk doping and the
bulk trap distribution. Either an external file, from which the doping profile for sourcedrain and the
channel is to be read, can be specified, or parameters for an internal process simulation that generates the
doping profile can be given. Both possibilities are mutually exclusive.
In case of FILE=NO or FILE=1-D the mask edge for source or drain profiles is assumed to be at the gate
contact edges for MOSFETs. XOFF specifies the offset from these mask edges. For MESFETs the mask
edges are assumed to be the source contact edge for the source profile and the drain contact edge for the
drain profile. These mask edges can be shifted with the DMOFF and the SMOFF key. To specify a constant
doping of either n or p type, e.g. for the simulation of a MESFET with constant substrate doping, the only
permitted keys on the PROFILE directive are the NB key together with the SUBS key. For MINIMOS 3D
FILE=NO must be in effect, which means that presently no external doping file for threedimensional
simulation can be used.
# \
1
l
1 1018
Units:
Range:
0 1 AC 1
PROFILE
113
Units:
Range:
10 AKEV 300
114
2
Units:
10
Range:
10
DOSE 1017
Units:
Range:
0 EA 1 2
Units:
Range:
0 ED 1 2
PROFILE
115
Range:
AS Arsenic
B Boron
BE Beryllium
MG Magnesium
O Oxygen
P Phosphorus
SB Antimony
SI Silicon
LD Diffusion Length
The LD key can be used to determine the extent of profile broadening due to annealing. If this key is
specified the TEMP and the TIME key are not effective. This key effects all PROFILE statements.
Type:
<R>
Default:
Units:
Range:
0 LD 10 4
116
NB Bulk Doping
The NB key has to be specified unless FILE=2-D or FILE=1-D has been given. In this case the NB key
must be omitted.
Type:
<R>
Default:
0
3
Units:
Range:
0 NB 5 1018
NS Predeposition Concentration
This key specifies the surface concentration for a sourcedrain predeposition. If the NS key is specified with
a particular PROFILE directive, the ELEM, DOSE, AKEV and TOX keys must not be specified. Phosphorus
and boron dopants are assumed for n-channel and p-channel devices, respectively. The specified value
must be at least 100 times larger than the bulk doping. The NS key must be omitted, if FILE=2-D or
FILE=1-D is specified or the ELEM, DOSE and AKEV keys are given. Note that predeposition is not
implemented for GaAs as substrate material.
Type:
<R>
Default:
3
Units:
Range:
1015 NS 5 1022
PROFILE
117
Units:
Range:
100 TEMP 1300
Units:
Range:
60 TIME 48 3600
118
<R>
9
10
VAGS
Units:
Range:
1 VAGX 2
Units:
Range:
2 5 10 4 VDEN 0 5
PROFILE
119
G
<R>
10
VAGS
Units:
Range:
1 VDGX 2
120
Directive RECOMBINATION
The RECOMBINATION directive is used to specify generation and recombination parameters. For a more
detailed explanation see Section 2.7. This directive can be omitted. No key of this directive is required.
Any of the following keys is active only if MODEL=AVAL or MODEL=HOT is specified (cf. OPTION
directive).
<R>
9 20 105 in silicon
1 34 106 in galliumarsenide
0 AN
1
109
<R>
1 70 107 in silicon
1 34 106 in galliumarsenide
Units:
Range:
0 AP
1
109
<R>
2 02 106 in silicon
2 03 106 in galliumarsenide
9
5 10
2
BN
107
<R>
3 20 106 in silicon
2 03 106 in gallium arsenide
Units:
Range:
2
9
5 10
5
BP
107
RECOMBINATION
121
<R>
2 8 10
5 0 10
0 CN
6
31
31
in silicon
in gallium arsenide
10
28
<R>
9 9 10
5 0 10
z
0 CP
6
32
31
in silicon
in galliumarsenide
10
28
122
<R>
1 5 10 6 in silicon
not used for galliumarsenide
10
DN
10
<R>
1 5 10 6 in silicon
not used for galliumarsenide
10
DP
10
<R>
1 in silicon
2 in galliumarsenide
1 EN 2
<R>
1 in silicon
2 in galliumarsenide
1 EP 2
RECOMBINATION
123
<R>
2 2 in silicon
0 in galliumarsenide
0 FN 10
<R>
2 2 in silicon
0 in galliumarsenide
0 FP 10
Model for tunneling can be specified with: {zx . For the explanation refer to Section 2.6.
Type:
<I>
Default:
0
Range:
0 MTUN 1000
Range:
0 1 TEXF 10
124
Range:
10 6 TLIF 106
TN Electron Lifetime
Type:
Default:
Units:
Range:
<R>
10 6
10
TN
109
TP Hole Lifetime
Type:
Default:
Units:
Range:
<R>
10 6
10
TP
109
STEP
125
Directive STEP
The STEP directive is used to specify a step sequence for any applied terminal voltage. The input data
will be checked for physical plausibility unless PHYSCK=NO is set on the OPTION directive. Therefore
ePn
eqt
?nPn
?nf
oeqt~}(s
oef}(s
1
1
r
n
r
Units:
Range:
5 DB
0 00999
0 00999 DB 5
Units:
Range:
5 DD
0 00999
0 00999 DD 5
Units:
Range:
5 DG
0 00999
0 00999 DG 5
126
This key specifies the lower value of the electric field range, for which material characteristics are
calcualted by the Monte Carlo method. If E1 0, a special algorithm for zero field strength is invoked.
This key applies only if MONTE-CARLO MCMOD=2 is set. See also the keydescriptions of NE and ELOG.
Type:
<R>
Default:
0
Units:
2
Range:
0 E1 107
STEP
127
Units:
Range:
2 10 3 VG 1
128
Directive TRANSIENT
The TRANS directive is used to specify the parameters for the transient simulation. The specification
includes type of terminal and waveform applied to this terminal, as well as time and voltage parameters
of the waveform. Different waveforms can be applied on different terminals simultaneously. The detailed
description of the waveforms is given in Section 1.4. If not specified, a constant bias which equals to
steady-state value (cf. BIAS directive) is assumed at the particular terminal.
The meaning of Uz ,
Type:
Units:
Range:
1 2 3
<R>
zu
20
U
{z
for terminal
08f,n,#t
20
W Waveform Specification
0u8fn(#qt
4
Files
NUMINP
formatted
obligatory
MINIMOS directives
Another input file is the doping file which may optionally be given.
Unit:
Format:
Usage :
Contents:
NUMDOP
binary
optional
Doping profile of a device
For a complete overview how to specify an appropriate doping profile see the PROFILE directive
(Page 112). The format of the doping file depends on the setting of the keys ASYM, FILE and XOFF on
the PROFILE directive. One- or two-dimensional doping data may be read depending on the setting of
the key FILE on the PROFILE directive. MINIMOS reads an external doping file in FORTRAN binary
mode.
129
CHAPTER 4 FILES
130
YINC is the distance increment for the doping table. The next MAXDO1 records of the doping file are
interpreted as the acceptor and donor concentration and their derivatives (one slice):
DO 1 I=1,MAXDO1
READ (NUMDOP) (AC(I,J),DACDY(I,J),DC(I,J),DDCDY(I,J),J=1,NSLICE)
DACDY(I,J)
Spatial derivative of the acceptor concentration in
DC(I,J)
Donor concentration in
DDCDY(I,J)
Spatial derivative of the donor concentration in
The index I denotes the location, where I=1 is the surface and I=MAXDO1 denotes the depth
(MAXDO1-1) YINC. If MINIMOS needs the doping at a deeper distance the bulk value will be assumed.
The index J denotes the specific slice:
J = 1
first source and drain doping slice.
J = 2
channel profile extended homogeneously all over the device.
J = 3
second source and drain profile slice.
131
**
*
**
CHAPTER 4 FILES
132
*
*
1
*
133
(NHS,NVS) MAXDO2. The parameter MAXDO2 can be found in the include file cmos02x.inc.
PL(I) and PV(J) are the coordinates of the point (I,J) in centimeters, PL(I) in lateral and PV(J)
in the vertical direction.
There has to be an index K between 1 and NL such that PL(K)=0. This location is interpreted as the mask
edge of the two-dimensional profile. PL(I) less than zero is the free surface and PL(I) greater than
zero denotes the masked surface. There must also be an index L between 1 and NV such that PV(L)=0.
This location is interpreted as the gate-insulator/semiconductor interface. Positive PV(L) denote the
semiconductor area, values at negative PV(L) are ignored.
The following arrays denote the two-dimensional source doping data. The derivatives are only required if
NPIECE > 0 :
CDS(I,J)
The donor concentration for the source profile in 3 .
DCDSDH(I,J)
The partial spatial derivative of the donor concentration with respect to the horizontal direction for
the source profile in 4 .
DCDSDV(I,J)
The partial spatial derivative of the donor concentration with respect to the vertical direction for the
source profile in 4 .
CAS(I,J)
The acceptor concentration for the source profile in 3 .
DCASDH(I,J)
The partial spatial derivative of the acceptor concentration with respect to the horizontal direction
for the source profile in 4 .
CHAPTER 4 FILES
134
DCASDV(I,J)
The partial spatial derivative of the acceptor concentration with respect to the vertical direction for
the source profile in 4 .
The following arrays denote the twodimensional drain doping data. The derivatives are only required if
NPIECE > 0 :
CDD(I,J)
The donor concentration for the drain profile in 3 .
DCDDDH(I,J)
The partial spatial derivative of the donor concentration with respect to the horizontal direction for
the drain profile in 4 .
DCDDDV(I,J)
The partial spatial derivative of the donor concentration with respect to the vertical direction for the
drain profile in 4 .
CAD(I,J)
The acceptor concentration for the drain profile in 3 .
DCADDH(I,J)
The partial spatial derivative of the acceptor concentration with respect to the horizontal direction
for the drain profile in 4 .
DCADDV(I,J)
The partial spatial derivative of the acceptor concentration with respect to the vertical direction for
the drain profile in 4 .
If ABS(NPIECE) > 1, which means that two separate profile pieces for source and drain are given, the
profiles have to match for the middle of the channel. If the simulation domain generated by MINIMOS
is larger than the domains described by the external doping arrays, the profile is simply extended in the
respective direction.
Note: The doping data read from the doping file are not physically verified.
NUMOUT
formatted
MINIMOS 2D results
135
NUMREF
formatted
MINIMOS 2D execution history
NUMBIN
binary
optional
MINIMOS 2D binary results
The binary output file is written in FORTRAN binary mode. The main purpose of this file is to store the
output data for subsequent postprocessing e.g. plotting. The length of this file depends on the specification
of the MODEL key (see Page 103). The various distributed quantities are given in modified j units
( , , , T ). To enable access to these data by user-written programs at first the successive WRITE
statements of the FORTRAN routine BINOUT are given, to highlight the exact order and type of the data
written.
M s
136
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
CHAPTER 4 FILES
HTITLE
CHCOD
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
UFB
TINS
XLGEO
XWGEO
US
UG
UD
UB
ID
IS
IB
LI
UT
NI
NX
NT
NY
NYOX-1
NYTOT
NXS+1
23
NXD-1
24
25
X
Y
137
string of 80 characters. The first line of the input file. The last 4 characters of HTITLE
always contain the word -2D. If MINIMOS has calculated in nonplanar mode
the character substring HTITLE(68) to HTITLE(76) equals to NONPLANAR.
Thus, it is easy to determine if MINIMOS calculated planarly or not. Note that the
distinction between both calculation modes is essential, because the structure of the
binary file varies with both modes.
1 character. This character may be N, or P and specifies the channel type of the
simulated transistor (see Page 64).
1 floating point word. The flatband voltage (see Page 103).
1 floating point word. The thickness of the gate insulator (see Page 64).
1 floating point word. The gate length < (see Page 64).
1 floating point word. The channel width (see Page 64).
1 floating point word. Source terminal voltage (see Page 62).
1 floating point word. Gate terminal voltage (see Page 62).
1 floating point word. Drain terminal voltage (see Page 62).
1 floating point word. Bulk terminal voltage (see Page 62).
1 floating point word. Drain terminal current.
1 floating point word. Source terminal current.
1 floating point word. Bulk terminal current.
1 floating point word. Intrinsic Debye length.
1 floating point word. Thermal voltage.
1 floating point word. Intrinsic concentration.
1 integer word. Number of mesh lines in x-direction.
1 integer word. Number of mesh lines of the active semiconductor in y-direction.
1 integer word. Number of mesh lines of the semiconductor in y-direction.
1 integer word. Number of mesh lines in the gate insulator in y-direction.
1 integer word. Total number of mesh lines in y-direction.
1 integer word. NXS denotes the x-index at the end of the source contact (planar
mode).
1 integer word. NXD denotes the x-index at the beginning of the drain contact (planar
mode).
NX floating point words. x-coordinates of the mesh.
NYTOT floating point words. y-coordinates of the mesh. Note that the mesh lines
in y-direction are in the oxide from index 1 to NYOX and in the semiconductor from
index NYOX+1 to NYTOT.
In case of nonplanar calculation the following geometry data will be included:
26
NINT
27
28
NXR
1
29
IMDINT
1 integer word. Number of interfaces stored on this file. Currently 2 interfaces, FOXO
and FOXU, are available.
1 integer word. Number of interface points in lateral (x) direction.
1 integer word. Number of interface points in transversal (z) direction. (1 in
MINIMOS 2D).
1 integer word. The mode how the interfaces are stored on this file. IMDINT=1
means that the interface information is stored on the original mesh, IMDINT=2
means that the interface information is stored on an equidistant mesh. IMDINT=3
covers all other cases. IMDINT=1 is default.
CHAPTER 4 FILES
138
30
31
32
33
HTITLE
FOXO
HTITLE
FOXU
34
35
36
37
CDOP
PSI
CMIN
MOMINL
38
MOMINT
39
40
JLMIN
JTMIN
41
42
CMAJ
MOMAJL
43
MOMAJT
44
45
46
JLMAJ
JTMAJ
AVAL
47
48
TCMIN
TCMAX
NUMPIF
formatted
If PIF=YES is set on the OUTPUT directive an ASCII-PIF file is written, but no formatted output written
to unit NUMOUT.
139
NUMIVC
formatted
Symbol
A=
A@
A
AC
j8=
j8@
j
j8C
Description
Bulk voltage (V)
Drain voltage (V)
Gate voltage (V)
Source voltage (V)
Bulk current (A)
Drain current (A)
Gate current (A)
Source current (A)
NUMUAI
binary
TITLE
NY,NX
(Y(I)*YSCAL,I=1,NY)
(X(I)*XSCAL,I=1,NX)
((A(I+J*NX)*ASCAL,I=1,NX),J=0,NY-1)
END
Quantities can not only be written after termination of the self-consistent iteration loop but also after completion of intermediate iterations. Up to five iteration numbers can be specified by the keys SAV1 SAV5.
For each specified iteration number a set of quantities is appended to the file:
pp
CHAPTER 4 FILES
140
Quantity
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Symbol
]
s @@
X`$
s
q
q
&
6qxs6N#'
&
s6N#'
Hb
H
&
9#9
9#9
#
/
6Ds6N#'
WZ
\
6s6x
6!As6N
window
window
Description
&
Electron concentration from solution of PDEs 3 '
&
Xvelocity from solution of PDEs 5s'
&
Potential update %'
&
Negative xelectric field 1s5'
&
Yelectric field 2s5'
Number of assigned scattering events
Normalization quantity (= carrier conc. in arbitrary units)
&
Xvelocity 5'
&
Yvelocity 5'
&
XX-component of thermal voltage tensor %'
&
YY-component of thermal voltage tensor %'
&
Xmomentum loss rate 1s5'
&
Ymomentum loss rate 1s5'
&
Average energy s%'
&
Impact ionization rate 1 s'
&
Ionization coefficient 1 s5'
&
Magnitude of average velocity 5s'
&
Magnitude of average momentum loss rate 1s5'
&
Mobility from MC 2 m:'
&
Temperature voltage for transport from MC :'
&
Overlay of MC and analytical mobilities 2 m'
&
Overlay of MC and equilibrium %'
&
Xenergy gradient field 2s5'
&
Yenergy gradient field 2s5'
&
Magnitude of energy gradient field 1s5'
&
Mobility window 1 '
&
Temperature window 1 '
NUMTRA
formatted
Symbol
t
A=
j =
@
j8@
A
j8
C
j8C
Description
Time (s)
Bulk voltage (V)
Bulk current (A)
Drain voltage (V)
Drain current (A)
Gate voltage (V)
Gate current (A)
Source voltage (V)
Source current (A)
141
NUMCG1
formatted
The two-dimensional cross-section of the gate contact is assumed to be of rectangular shape. Not only
the total gate charge is written to the file but also the components located a the source-sided edge, at the
interface to the gate-oxide and at the drain-sided edge. The meaning of the columns is:
Column
1
2
3
4
5
6
7
8
Symbol
=
A@
A
AC
U
U
U
/
dW
U $Z
Description
Bulk voltage (V)
Drain voltage (V)
Gate voltage (V)
Source voltage (V)
Total gate charge (As)
Part of gate charge at the source-sided edge (As)
Part of gate charge at the lower edge (As)
Part of gate charge at the drain-sided edge (As)
NUMCG2
formatted
The voltage drop is evaluated at a coordinate half-way between the source and drain edges of the gate.
The meaning of the columns is:
Column
1
2
3
4
5
Symbol
A=
A@
AC
A
Description
Bulk voltage (V)
Drain voltage (V)
Gate voltage (V)
Source voltage (V)
Voltage drop in the gate (V)
CHAPTER 4 FILES
142
Charge-Pumping Output
In charge-pumping mode additional data regarding the components of the terminal currents is written to
a file.
Unit:
Format:
Column
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
NUMCP3
binary
Symbol
t
A=
j =
K
j8=A
G
j = Y
@
j8@1
K
j @1
G
j @1 Y
A
j8!
K
j8!
G
j ! Y
AC
j CN
j8Cm
G
j Cm Y
:
64T
K
64T
Description
Time (s)
Bulk voltage (V)
Electron bulk current (A)
Hole bulk current (A)
Displacement bulk current (A)
Drain voltage (V)
Electron drain current (A)
Hole drain current (A)
Displacement drain current (A)
Gate voltage (V)
Electron gate current (A)
Hole gate current (A)
Displacement gate current (A)
Source voltage (V)
Electron source current (A)
Hole source current (A)
Displacement source current (A)
Integrated electron generation rate at front interface (A)
Integrated hole generation rate at front interface (A)
Information on the area actively contributing to the charge-pumping effect is written to a file as well.
Unit:
Format:
Column
1
2
3
4
5
6
7
8
9
10
11
NUMCP2
formatted
Symbol
=
@
A
AC
4G Y
GV
u G Y
<2
KY
5
qK V
<1
u
$
KY
Description
&
Bulk voltage '
&
Drain voltage %'
&
Gate voltage '
&
Source voltage '
&
Drain sided location where the critical hole concentration occurs 5'
&
Source sided location where the critical hole concentration occurs 5'
&
Distance between location G Y and drain edge 5'
&
Drain sided location where the critical electron concentration occurs 5'
&
Source sided location where the critical electron concentration occurs 5'
&
Distance between location KY and drain edge 5'
&
Critical carrier concentration 1 s 3 '
143
NUMBBT
formatted
Symbol
A
j = XX
j = XX
j Cm XX
j @1 XX
j = 9#9
G
K
K
Description
Gate voltage (V)
Drain voltage (V)
BBT electron bulk current (A)
BBT hole bulk current (A)
BBT electron source current (A)
BBT electron drain current (A)
Bulk current due to impact ionization (A)
NUMBI3
binary
MINIMOS 3D has no specific formatted input file as MINIMOS 2D. The input deck of the corresponding
two-dimensional example is fully valid. MINIMOS 3D requires a binary input file that serves as a link to
the corresponding two-dimensional solution.
Instead of a binary link file to MINIMOS 2D, MINIMOS 3D allows the use of a previous three-dimensional
solution as initial input. This may avoid repeated calculation of costly three-dimensional Gummel
iterations. The user has to specify an O as second character for the M3MODE key to direct MINIMOS 3D
to read such a file when invoked. Note that MINIMOS 3D does not check the existence of such a file and
crashes if the previous solution file is not found. Note also that MINIMOS 3D produces a binary output
file only if BIN=YES is set on the END directive. Furthermore, MINIMOS 2D has to be invoked first to
preset various common block variables.
CHAPTER 4 FILES
144
Unit:
Format:
Contents:
NUMOUT
formatted
MINIMOS 3D results
The reference file for MINIMOS 3D is fully analogous to the reference file of MINIMOS 2D.
Unit:
Format:
Contents:
NUMREF
formatted
MINIMOS 3D execution history
NUMBIN
binary
The binary output file is written in FORTRAN binary mode. The main purpose of this file is to store the
output data for subsequent postprocessing. The length of this file depends on the setting of the MODEL
and the M3MODE keys (see Page 103). The various distributed quantities are given in modified j units
( , , , T ). The exact order and type of the written data can be found by inspection of the FORTRAN
subroutine BINAT3.
M c
To enable access to these data by user-written programs at first the successive WRITE statements of the
FORTRAN routine BINAT3 are given, to highlight the exact order and type of the data written:
**
**
**
**
**
**
**
**
**
**
**
1
**
2
**
3
**
4
**
6
**
8
*
9
**
10
11
**
12
145
CHAPTER 4 FILES
146
13
14
**
15
**
16
17
147
TITLE
CHCOD
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
UFB
TINS
XLGEO
XWGEO
US
UG
UD
UB
ID
IS
IB
LI
UT
NI
NX
NT
NY
NYOX
NYTOT
NXS
NXG1
NXG2
NXD
NZ
NZSD
28
29
30
31
NZA
NYFOX
X
Y
32
33
Z
NINT
34
35
36
NX
NZ
IMDINT
string of 80 characters. The first line of the input file. The last 4 characters of TITLE
always contain the word -3D.
1 character. This character may be N, or P and specifies the channel type of the
simulated transistor (see section Page 64)
1 floating point word. The flatband voltage (see section Page 103)
1 floating point word. The thickness of the gate insulator (see section Page 64)
1 floating point word. The gate length L (see section Page 64)
1 floating point word. The channel width W (see section Page 64)
1 floating point word. Source terminal voltage (see section Page 62)
1 floating point word. Gate terminal voltage (see section Page 62)
1 floating point word. Drain terminal voltage (see section Page 62)
1 floating point word. Bulk terminal voltage (see section Page 62)
1 floating point word. Drain terminal current.
1 floating point word. Source terminal current
1 floating point word. Bulk terminal current.
1 floating point word. Intrinsic Debye length.
1 floating point word. Thermal voltage.
1 floating point word. Intrinsic concentration.
1 integer word. Number of mesh lines in xdirection.
1 integer word. Number of mesh lines of the active semiconductor in ydirection.
1 integer word. Number of mesh lines of the semiconductor in ydirection.
1 integer word. Number of mesh lines in the gate insulator in ydirection.
1 integer word. Total number of mesh lines in ydirection.
1 integer word. Xindex of the last source point
1 integer word. X-index of the first gate point (drain side)
1 integer word. X-index of the last gate point (drain side)
1 integer word. Xindex of the first drain point
1 integer word. Number of meshlines in zdirection
1 integer word. Number of meshlines intersecting the source and drain contacts in
zdirection
1 integer word. Number of meshlines od the active semiconductor in zdirection
1 integer word. Yindex of the mesh line inserted at depth OXIDEF
NX floating point words. X-coordinates of the mesh
NYTOT floating point words. Y-coordinates of the mesh. Note that the mesh lines in
ydirection are in the oxide from index 1 to NYOX and in the semiconductor from
index NYOX+1 to NYTOT.
NZ floating point words. Z-coordinates of the mesh
1 integer word. Number of interfaces stored on this file. Currently 2 interfaces, FOXO
and FOXU, are available.
1 integer word. Number of interface points in lateral (x) direction.
1 integer word. Number of interface points in transversal (z) direction
1 integer word. The mode how the interfaces are stored on this file. IMDINT=1
means that the interface information is stored on the original mesh, IMDINT=2
means that the interface information is stored on an equidistant mesh. IMDINT=3
covers all other cases. IMDINT=1 is default.
CHAPTER 4 FILES
148
37
38
39
40
41
42
43
44
45
TITLE
FOXO
TITLE
FOXU
CDOP
PSI
CMIN
CMAJ
MOMINL
46
MOMINT
47
48
49
JLMIN
JTMIN
JZMIN
50
MOMAJL
51
MOMAJT
52
53
54
55
JXMAJ
JYMAJ
JZMAJ
AVAL
NX NT NZ floating point words. Majority carrier mobility parallel to the gate insulator
interface
NX NT NZ floating point words. Majority carrier mobility perpendicular to the gate
insulator interface
NX NT NZ floating point words. Majority carrier current density (x-component)
NX NT NZ floating point words. Majority carrier current density (y-component)
NX NT NZ floating point words. Majority carrier current density (z-component)
NX NT NZ floating point words. Avalanche (impact ionization) generation rate.
EOF for MODEL=AVAL
56
57
TCMIN
TCMAX
Bibliography
[1] O. Heinreichsberger. Transiente Simulation von Silizium-MOSFETs. Dissertation, Technische Universitat Wien, 1992.
[2] P. Habas. Analysis of Physical Effects in Small Silicon MOS Devices. Dissertation, Technische
Universitat Wien, 1993.
[3] S. Bandyopadhyay, M.E. Klausmeier-Brown, C.M. Maziar, S. Datta, and M.S. Lundstrom. A Rigorous Technique to Couple Monte Carlo and Drift-Diffusion Models for Computationally Efficient
Device Simulation. IEEE Trans.Electron Devices, ED-34(2):392399, 1987.
[4] H. Kosina and S. Selberherr. A Hybrid Device Simulator that Combines Monte Carlo and DriftDiffusion Analysis. IEEE Trans.Computer-Aided Design, 13(2):201210, 1994.
[5] J.O. Nylander, F. Masszi, S. Selberherr, and S. Berg. Computer Simulations of Schottky Contacts
with a Non-Constant Recombination Velocity. Solid-State Electronics, 32(5):363367, 1989.
[6] Ph. Lindorfer and S. Selberherr. MESFET Analysis with MINIMOS. In Heuberger et al. Your
command was ignored. Type I <command> <return> to replace it with another command,or
<return> to continue without it. [75], pages 9296.
[7] S. Selberherr. MOS Device Modeling at 77K. IEEE Trans.Electron Devices, 36(8):14641474,
1989.
[8] S. Selberherr, W. Hansch, M. Seavey, and J. Slotboom. The Evolution of the MINIMOS Mobility
Model. Solid-State Electronics, 33(11):14251436, 1990.
[9] D.M. Caughey and R.E. Thomas. Carrier Mobilities in Silicon Empirically Related to Doping and
Field. Proc.IEEE, 52:21922193, 1967.
[10] J. Xu and M. Shur. Velocity-Field Dependence in GaAs. IEEE Trans.Electron Devices, ED34(8):18311832, 1987.
[11] M. Hirose, J. Yoshida, and N. Toyoda. An Improved Two-Dimensional Simulation Model (MEGA)
for GaAs MESFET Applicable to LSI Design. IEEE Trans.Computer-Aided Design, 7(2):225230,
1988.
[12] J.W. Slotboom, G. Streutker, G.J.T. Davids, and P.B. Hartog. Surface Impact Ionization in Silicon
Devices. In Proc.Int.Electron Devices Meeting, pages 494497, 1987.
149
150
BIBLIOGRAPHY
[13] T. Thurgate and N. Chan. An Impact Ionization Model for Two-Dimensional Device Simulation.
IEEE J.Solid-State Circuits, SC-20(1):320324, 1985.
[14] W. Hansch, M. Orlowski, and W. Weber. The Hot-Electron Problem in Submicron MOSFET. In
Nougier and Gasquet Your command was ignored. Type I <command> <return> to replace it
with another command,or <return> to continue without it. [76], pages 597606.
[15] S. Banerjee, D. Coleman, W. Richardson, and A. Shah. Leakage Mechanisms in the Trench Transistor
DRAM Cell. IEEE Trans.Electron Devices, 35(1):108116, 1988.
[16] T. Ozaki, A. Nitayama, T. Hamamoto, K. Sunouchi, and F. Horiguchi. Analysis of Band-toBand Tunneling Leakage Current in Trench-Capacitor DRAM Cells. IEEE Electron Device Lett.,
12(3):9597, 1991.
[17] T. Hori. Drain-Structure Design for Reduced Band-to-Band and Band-to-Defect Tunneling Leakage.
In VLSITECH90 Your command was ignored. Type I <command> <return> to replace it with
another command,or <return> to continue without it. [77], pages 6970.
[18] C. Chang and J. Lien. Corner-Field Induced Drain Leakage in Thin Oxide MOSFET.
Proc.Int.Electron Devices Meeting, pages 714717, 1987.
In
[19] Y. Igura, H. Matsuoka, and E. Takeda. New Device Degradation Due to cold Carriers Created by
Band-to-Band Tunneling. IEEE Electron Device Lett., 10(5):227229, 1989.
[20] G.Q. Lo, A.B. Joshi, and D.-L. Kwong. Hot-Carrier-Stress Effects on Gate-Induced Drain Leakage
Current in n-channel MOSFETs. IEEE Electron Device Lett., 12(1):57, January 1991.
[21] T. Tsuchiya. A New Enhanced Degradation Phenomenon in MOSFETs Under AC Stress: The Effect
of Band-to-Band Tunneling. In Symposium on VLSI Technology, pages 7980, Kyoto, Japan, 1989.
[22] I.-C. Chen and C.W. Teng. A Quantitative Physical Model for the Band-to-Band Tunneling-Induced
Substrate Hot Electron Injection in MOS Devices. IEEE Trans.Electron Devices, 39(7):16461651,
1992.
[23] Y. Tada, C. Nagata, and M. Iwahashi. Band-to-Band Tunneling Current as a Probe for Hot Carrier
Effects. In Ext. Abstr. of the 22nd Conf. on Solid State Devices and Materials, pages 315318,
Sendai, 1990.
[24] C. Duvvury, D.J. Redwine, and H.J. Stiegler. Leakage Current Degradation in N-MOSFETs Due to
Hot-Electron Stress. IEEE Electron Device Lett., 9(11):579581, 1988.
[25] A. Acovic, M. Dutoit, and M. Ilegems. Characterization of Hot-Electron-Stressed MOSFETs
by Low-Temperature Measurements of the Drain Tunnel Current. IEEE Trans.Electron Devices,
37(6):14671476, 1979.
[26] H.S. Momose, S. Kambayashi, H. Kawaguchi, I. Mizushima, S. Onga, and J. Matsunaga. A Bandto-Band Tunneling MOSFET using a Thin Film Transistor. In Ext. Abstr. of the 22nd Conf. on Solid
State Devices and Materials, pages 365368, Sendai, 1990.
[27] E. Takeda, H. Matsuoka, Y. Igura, and S. Asai. A Band-to-band Tunneling MOS Device (B2 TMOSFET) A Kind of Si Quantum Device . In Proc.Int.Electron Devices Meeting, pages
402405, 1988.
BIBLIOGRAPHY
151
[28] Y. Odake, K. Kurimota, and S. Odanaka. Three-Dimensional Numerical Modeling of the Indirect
Band-to-Band Tunneling in MOSFETs. In Ext. Abstr. of the 22nd Conf. on Solid State Devices and
Materials, pages 131134, Sendai, 1990.
[29] M. Orlowski, S.W. Sun, P. Blakey, and R. Subrahmanyan. Two-Dimensional Simulation of Band-toBand Tunneling in an LDD-MOSFET: Explanation of Experimental Results and Prediction of New
Phenomena. In VLSITECH90 Your command was ignored. Type I <command> <return> to
replace it with another command,or <return> to continue without it. [77], pages 6768.
[30] I. Nedev, A. Asenov, and E. Stefanov. Experimental Study and Modeling of Band-to-Band Tunneling
Leakage Current in Thin-Oxide MOSFETs. Solid-State Electronics, 34(12):14011408, 1991.
[31] P. Habas, A. Lugbauer, and S. Selberherr. Two-Dimensional Numerical Modeling of Interband
Tunneling Accounting for Nonuniform Electric Field. In Workshop on Numerical Modeling of
Processes and Devices for Integrated Circuits NUPAD IV, pages 135140, Seattle, 1992.
[32] T. Endoh, R. Shirota, M. Momodomi, and F. Masuoka. An Accurate Model of Subbreakdown Due
to Band-to-Band Tunneling and Some Applications. IEEE Trans.Electron Devices, 37(1):290296,
1990.
[33] G. Nanz, P. Dickinger, and S. Selberherr. Calculation of Contact Currents in Device Simulation.
IEEE Trans.Computer-Aided Design, 11(1):128136, 1992.
[34] L.V. Keldysh. Behavior of Non-Metallic Crystals in Strong Electric Field.
6(33)(4):763770, 1958.
Sov.Phys.JETP,
BIBLIOGRAPHY
152
[43] P. Habas and S. Selberherr. A Closed-Loop Extraction of the Spatial Distribution of Interface Traps
Based on Numerical Model of the Charge-Pumping Experiment. In Ext. Abstr. of the Int. Conf. on
Solid State Devices and Materials, pages 170172, Tsukuba, 1992.
[44] T. Ouisse, S. Cristoloveanu, T. Elewa, H. Haddara, G. Borel, and D.E. Ioannou. Adaptation of
the Charge Pumping Technique to Gated p-i-n Diodes Fabricated on Silicon on Insulator. IEEE
Trans.Electron Devices, 38(6):14321444, 1991.
[45] P. Habas, O. Heinreichsberger, and S. Selberherr. Transient Two-Dimensional Numerical Analysis of
the Charge-Pumping Experiment. In Maes et al. Your command was ignored. Type I <command>
<return> to replace it with another command,or <return> to continue without it. [78], pages
687690.
[46] M.M. Mandurah, K.C. Saraswat, and C.R. Helms. Dopant Segregation in Polycrystalline Silicon.
Journal of Applied Physics, 51(11):57555763, 1980.
[47] J.Y.-C. Sun, R. Angelucci, C.Y. Wong, G. Scilla, and E. Landi. Rapid Thermal Processing of
Arsenic-Implanted Polysilicon on Very Thin Oxide. In Nougier and Gasquet Your command was
ignored. Type I <command> <return> to replace it with another command,or <return> to
continue without it. [76], pages 401404.
[48] C.L. Chu, G. Chin, K.C. Saraswat, S.S. Wong, and R. Dutton. Technology Limitations for N /P
Polycide Gate CMOS due to Lateral Dopant Diffusion in Silicide/Polysilicon Layers. IEEE Electron
Device Lett., 12(12):696698, 1991.
[49] N.G. Nilsson. Empirical Approximations for the Fermi Energy in a Semiconductor with Parabolic
Bands. Appl. Phys. Lett., 33(7):653654, 1978.
[50] X. Aymerich-Humett, F. Serra-Mestres, and J. Millan. An Analytical Approximation for the Fermi&
Dirac Integral 3 ; 2 ' . Solid-State Electronics, 24(10):981982, 1981.
[51] P. Habas and J.V. Faricelli. Investigation of the Physical Modeling of the Gate-Depletion Effect.
IEEE Trans.Electron Devices, 39(6):14961500, 1992.
[52] P. Habas and S. Selberherr. On the Effect of Non-Degenerate Doping of Polysilicon Gate in Thin
Oxide MOS-Devices Analytical Modeling. Solid-State Electronics, 33(12):15391544, 1990.
[53] P. Habas and S. Selberherr. Numerical Simulation of MOS-Devices with Non-Degenerate Gate. In
Heuberger et al. Your command was ignored. Type I <command> <return> to replace it with
another command,or <return> to continue without it. [75], pages 161164.
[54] G. Yaron and D. Frohman-Bentchkowsky. Capacitance Voltage Characterization of Poly
Structures. Solid-State Electronics, 23:433439, 1980.
M
2
[55] N. Lifshitz and S. Luryi. Influence of a Resistive Sublayer at the Polysilicon/Silicon Dioxide Interface
on MOS Properties. IEEE Trans.Electron Devices, ED-30(7):833836, 1983.
[56] T.W. Hickmott and R.D. Isaac. Barrier Heights at the Polycrystalline Silicon-SiO2 Interface. Journal
of Applied Physics, 52(5):34643475, 1981.
[57] Y.-S. Kim and J.G. Fossum. Physical DMOST Modeling for High-Voltage IC CAD. IEEE
Trans.Electron Devices, 37(3):797803, 1990.
BIBLIOGRAPHY
153
[58] P. Habas. A Physically Based Analytical MOSFET Model with Accurate Field Dependent Mobility.
Solid-State Electronics, 33(7):923933, 1990.
[59] J. Rebollo, E. Figueras, J. Millan, E. Lora-Tamayo, and F. Serra-Mestres. Analysis of the QuasiSaturation Region of High Voltage VDMOS Devices. Solid-State Electronics, 30(2):177180, 1987.
[60] M.N. Darwish. Study of the Quasi-Saturation Effect in VDMOS Transistors. IEEE Trans.Electron
Devices, ED-33(11):17101716, 1986.
[61] J. Paredes, S. Hidalgo, F. Berta, J. Fernandez, J. Rebollo, and J. Millan. A Steady-State VDMOS
Transistor Model. IEEE Trans.Electron Devices, 39(3):712719, 1992.
[62] C. Hu, M.-H. Chi, and V.M. Patel. Optimum Design of Power MOSFETs. IEEE Trans.Electron
Devices, ED-31(12):16931700, 1984.
[63] K. Board, D.J. Byrne, and M.S. Towers. The Optimization of On-Resistance in Vertical DMOS
Power Devices with Linear and Hexagonal Surface Geometries. IEEE Trans.Electron Devices,
ED-31(1):7580, 1984.
[64] J. Fernandez, S. Hidalgo, J. Paredes, F. Berta, J. Rebollo, J. Millan, and F. Serra-Mestres. An
ON-Resistance Closed Form for VDMOS Devices. IEEE Electron Device Lett., 10(5):212215,
1989.
[65] A.F. Franz and G.A. Franz. BAMBIA Design Model for Power MOSFETs. IEEE Trans.ComputerAided Design, CAD-4(3):177189, 1985.
[66] S. Selberherr. Analysis and Simulation of Semiconductor Devices. Springer, Wien, 1984.
[67] M.S. Mock. Analysis of Mathematical Models of Semiconductor Devices. Boole Press, Dublin,
1983.
[68] P. Habas, D. Pantic, Z. Prijic, and N. Stojadinovic. Technology CAD of Low-Voltage Power
VDMOSFET. In Proc. CAS94, October 1994.
[69] B.K. Ridley. Reconciliation of the Conwell-Weisskopf and Brooks-Herring Formulae for ChargedImpurity Scattering in Semiconductors: Third-Body Interference. J.Phys.C, 10:15891593, 1977.
[70] T.G. Van de Roer and F.P. Widdershoven. Ionized Impurity Scattering in Monte Carlo Calculations.
Journal of Applied Physics, 59(3):813815, 1986.
[71] C. Hao, J. Zimmermann, M. Charef, R. Fauquembergue, and E. Constant. Monte Carlo Study of TwoDimensional Electron Gas Transport in Si-MOS Devices. Solid-State Electronics, 28(8):733740,
1985.
[72] L.V. Keldysh. Concerning the Theory of Impact Ionization in Semiconductors. Sov.Phys.JETP,
21:11351144, 1965.
[73] R. VanOverstraeten and H.J. DeMan. Measurement of the Ionization Rates in Diffused Silicon p-n
Junctions. Solid-State Electronics, 13:583608, 1970.
[74] A. Phillips and P.J. Price. Monte Carlo Calculations on Hot Electron Energy Tails. Appl. Phys. Lett.,
30(10):528530, 1977.
154
BIBLIOGRAPHY
[75] A. Heuberger, H. Ryssel, and P. Lange, editors. Proc. 19th European Solid State Device Research
Conference - ESSDERC89, Berlin, 1989. Springer.
[76] J.-P. Nougier and D. Gasquet, editors. Proc. 18th European Solid State Device Research Conference
- ESSDERC 88, volume 49 of Journal de Physique, Les Ulis Cedex, France, 1988. les e ditions de
physique.
[77] Proc. Symposium on VLSI Technology, Honolulu, 1990.
[78] H.E. Maes, R.P. Mertens, and R.J. Van Overstraeten, editors. Proc. 22nd European Solid State
Device Research Conference - ESSDERC92, volume 19 of Microelectr.Eng., Leuven, Belgium,
1992. Elsevier.
Index
Abbreviation
Simulator Key
3C, 69
3G, 70
3H, 70
3U, 70
AC, 83, 103, 112
AK, 71, 83, 113
AL, 109
AS, 113
AV, 109
B, 64, 68
BE, 73
C, 64, 68
CA, 68
CI, 85
CNT, 96
CU, 103
D, 64
D1, 73
D2, 73
DC, 73
DD, 73
DEF, 97
DEPS, 97
DFM, 97
DI, 109
DLS, 97
DM, 114
DO, 71, 83, 97, 114
DR, 73
DTK , 96
E, 68
E1S, 97
E2S, 97
EL, 71, 84, 109, 115
EMU, 97
EPH , 98
ERE, 98
ET, 109
EXI, 98
EXS, 98
F, 64
FI, 115
FS, 87
G, 65
G1, 74
G2, 74
GC, 74
GCHC, 104
GFIL, 80
GR, 104
I, 65
ID, 104
IN, 104
JC, 62
JL, 110
JT, 110
K, 65
LCP, 105
LD, 105
LF, 115
LS, 105
LVDM, 105
M, 66
M3, 106
MA, 110
MCMO, 98
MCP, 99
MENE, 89
METR, 93
MI, 110
155
INDEX
156
ML, 99
MO, 105, 110
MP, 105
MT, 99, 110
MXT, 80
MYT, 80
NA, 69
NC, 69
NG, 69
NH, 69
NO, 110
NONP, 99
NS, 86
NU, 69
NUM , 100
OM, 106
OX, 74, 110
PA, 69
PD, 106
PF, 88
PH, 106, 110
PI, 110
PN, 87
PS, 86, 110
R1, 74
R2, 75
RB, 107
RD, 107
RE, 74
RH, 111
RS, 107
S, 66
S1, 75
S2, 75
SAV , 100
SC, 75
SD, 75
SI, 86
SM, 116
SO, 100
SR, 75
ST, 111
SU, 116
SXT, 80
SYT, 81
T, 66, 69
T1, 76
T2, 76
TE, 71, 84, 107, 117
TG, 75
TI, 72, 84, 117
TO, 117
TRAC, 100
UF, 107
UTM, 100
VSO, 101
WBI, 101
WD, 101
WEI, 101
WI, 101
WS, 101
XI, 86
XL, 66
XO, 117
XR, 66
XW, 102
Y, 67
YM, 76
YO, 118
YW, 102
Simulator Directive
BIAS, 62
DEVICE, 64
END, 68
FIMPLANT, 71
GEOMETRY, 73
GRID, 77
IMPLANT, 83
INTERFACE, 85
MOBILITY, 94
MONTE-CARLO, 96
OPTION, 103
OUTPUT, 109
PROFILE, 112
RECOMBINATION, 120
STEP, 125
TRANSIENT, 128
Simulator Key
3CYCLES, 69
3GEN, 70
3HOT, 70
3UPDATES, 70
Abbreviation
INDEX
157
3C, 69
3G, 70
3H, 70
3U, 70
AC, 83, 103, 112
AK, 71, 83, 113
AL, 109
AS, 113
AV, 109
B, 64, 68
BE, 73
C, 64, 68
CA, 68
CI, 85
CNT, 96
CU, 103
D, 64
D1, 73
D2, 73
DC, 73
DD, 73
DEF, 97
DEPS, 97
DFM, 97
DI, 109
DLS, 97
DM, 114
DO, 71, 83, 97, 114
DR, 73
DTK , 96
E, 68
E1S, 97
E2S, 97
EL, 71, 84, 109, 115
EMU, 97
EPH , 98
ERE, 98
ET, 109
EXI, 98
EXS, 98
F, 64
FI, 115
FS, 87
G, 65
G1, 74
G2, 74
GC, 74
GCHC, 104
GFIL, 80
GR, 104
I, 65
ID, 104
IN, 104
JC, 62
JL, 110
JT, 110
K, 65
LCP, 105
LD, 105
LF, 115
LS, 105
LVDM, 105
M, 66
M3, 106
MA, 110
MCMO, 98
MCP, 99
MENE, 89
METR, 93
MI, 110
ML, 99
MO, 105, 110
MP, 105
MT, 99, 110
MXT, 80
MYT, 80
NA, 69
NC, 69
NG, 69
NH, 69
NO, 110
NONP, 99
NS, 86
NU, 69
NUM , 100
OM, 106
OX, 74, 110
PA, 69
PD, 106
PF, 88
PH, 106, 110
PI, 110
PN, 87
PS, 86, 110
INDEX
158
R1, 74
R2, 75
RB, 107
RD, 107
RE, 74
RH, 111
RS, 107
S, 66
S1, 75
S2, 75
SAV , 100
SC, 75
SD, 75
SI, 86
SM, 116
SO, 100
SR, 75
ST, 111
SU, 116
SXT, 80
SYT, 81
T, 66, 69
T1, 76
T2, 76
TE, 71, 84, 107, 117
TG, 75
TI, 72, 84, 117
TO, 117
TRAC, 100
UF, 107
UTM, 100
VSO, 101
WBI, 101
WD, 101
WEI, 101
WI, 101
WS, 101
XI, 86
XL, 66
XO, 117
XR, 66
XW, 102
Y, 67
YM, 76
YO, 118
YW, 102
ACCELERATION, 103
INDEX
159
W, 66
XLEFT, 66
XRIGHT, 66
YDEPTH, 67
DFMIN, 97
DG, 125
DGAP, 64
DGLD, 93
DGLS, 93
DGLX, 93
DGMI, 77
DGUD, 92
DGUS, 92
DGUX, 92
DIGITS, 109
DJRX, 77
DJRY, 77
DLSS, 97
DMOFF, 114
DN, 122
DOPX, 77
DOPY, 77
DOPZ, 77
DOSE, 71, 83, 114
DOVL, 97
DP, 122
DP1D, 79
DPXC, 78
DPXD, 78
DPXP, 78
DPYC, 78
DPYD, 78
DPYP, 78
DPZ, 79
DRGRD, 73
1 2 3 4 5 6, 96
DTK ,
DUGD, 92
DUGS, 92
DUGX, 92
DULD, 92
DUUD, 92
DX1, 79
DX2, 79
DX3, 79
DY, 79
DZ, 80
E1, 64, 126
INDEX
160
E1SELF, 97
E2, 64, 126
E2SELF, 97
EA, 103, 114
ED, 104, 114
ELAT, 109
ELEM, 71, 84, 115
ELOG, 126
EMULT, 97
EN, 122
END
3CYCLES, 69
3GEN, 70
3HOT, 70
3UPDATES, 70
BIN, 68
CACCREL, 68
CAPERR, 68
ERROR, 68
NAC, 69
NCYCLES, 69
NGEN, 69
NHOT, 69
NUPDATES, 69
PACCREL, 69
TCERR, 69
EP, 122
1 2 3 4 5 6, 98
EPH ,
EREF, 98
ERROR, 68
ETRAN, 109
EXII, 98
EXSS, 98
FILE, 115
FILM, 64
FIMPLANT
AKEV, 71
DOSE, 71
ELEM, 71
TEMP, 71
TIME, 72
FN, 123
FP, 123
FSS, 87
G1CGRD, 74
G2CGRD, 74
GATE, 65
GCHCALC, 104
GCTHE, 74
GEOMETRY
BEAKL, 73
D1GAP, 73
D2GAP, 73
DCGRD, 73
DDREOX, 73
DRGRD, 73
G1CGRD, 74
G2CGRD, 74
GCTHE, 74
LD, 74
LS, 74
OXIDEF, 74
R1GRD, 74
R2GRD, 75
RECESS, 74
S1GAP, 75
S2GAP, 75
SCGRD, 75
SDREOX, 75
SRGRD, 75
T1GRD, 76
T2GRD, 76
TD, 75
TGATE, 75
TS, 76
YMDEP, 76
GFILL, 80
GGMI, 80
GRID
DGMI, 77
DJRX, 77
DJRY, 77
DOPX, 77
DOPY, 77
DOPZ, 77
DP1D, 79
DPXC, 78
DPXD, 78
DPXP, 78
DPYC, 78
DPYD, 78
DPYP, 78
DPZ, 79
DX1, 79
INDEX
DX2, 79
DX3, 79
DY, 79
DZ, 80
GFILL, 80
GGMI, 80
MXTun, 80
MYTun, 80
PGMI, 80
SXTun, 80
SYTun, 81
XE1, 81
XE2, 81
XE3, 81
XS1, 81
XS2, 81
XS3, 82
YE, 82
YS, 82
ZE, 82
ZS, 82
GRIDFREEZE, 104
IDEALITY, 104
IMPLANT
ACTIVATION, 83
AKEV, 83
DOSE, 83
ELEM, 84
LD, 84
TEMP, 84
TIME, 84
INS, 65
INTERFACE
AGLD, 91
AGLS, 91
AGLX, 91
AGUD1, 90
AGUD2, 90
AGUS1, 90
AGUS2, 90
AGUX1, 90
AGUX2, 90
ARUD1, 89
ARUD2, 89
ARUE1, 89
ARUE2, 89
ARUS1, 89
161
ARUS2, 89
AUGD, 90
AUGS, 91
AUGX, 91
AULD, 90
AUUD, 89
BAUUD, 91
BDUUD, 91
CISS, 85
DA0, 88
DD0, 88
DGLD, 93
DGLS, 93
DGLX, 93
DGUD, 92
DGUS, 92
DGUX, 92
DUGD, 92
DUGS, 92
DUGX, 92
DULD, 92
DUUD, 92
FSS, 87
MENERG, 89
METRAP, 93
NRD, 85
NRE, 85
NRS, 85
NSS, 86
PFSS, 88
PNSS, 87
PSS, 86
SISS, 86
XISS, 86
INTRINSIC, 104
IV, 109
JCAP, 62
JLAT, 110
JTRAN, 110
KBULK, 65
L, 65
LCPUMP, 105
LD, 74, 84, 115
LDRAIN, 105
LFIT, 115
LS, 74
LSOURCE, 105
INDEX
162
LUDG, 126
LUGB, 126
LVDMOSFET, 105
M3MODE, 106
MAJ, 110
MAT, 66
MB, 94
MC, 94
MCMOD, 98
MCPIT, 99
MENERG, 89
METRAP, 93
MI, 94
MIN, 110
ML, 94
MLONG, 99
MOB, 110
MOBILITY
MB, 94
MC, 94
MI, 94
ML, 94
MQ, 95
MR, 95
MS, 95
MT, 95
MV, 95
MODEL, 105
MONTE-CARLO
AUXD, 96
AUXS, 96
BDR, 96
CNTMIN, 96
DEFPOT, 97
DEPSS, 97
DFMIN, 97
DLSS, 97
DOVL, 97
1 2 3 4 5 6, 96
DTK ,
E1SELF, 97
E2SELF, 97
EMULT, 97
EPH ,
1 2 3 4 5 6, 98
EREF, 98
EXII, 98
EXSS, 98
MCMOD, 98
MCPIT, 99
MLONG, 99
MTRAN, 99
NONPAR, 99
NSB, 99
NUM , =1,2,3,4,5, 100
NUMB, 99
RHO, 100
SAV ,
1 2 3 4 5, 100
SEED, 100
SOVL, 100
TRACE, 100
UTMIN, 100
VSOUND, 101
WBH, 101
WBII, 101
WDIST, 101
WEII, 101
WINFREEZE, 101
WSCREEN, 101
XWIDTH, 102
YWIDTH, 102
MPOLY, 105
MQ, 95
MR, 95
MS, 95
MT, 95
MTRAN, 99
MTUN, 123
MTV, 110
MV, 95
MXTun, 80
MYTun, 80
NAC, 69
NB, 116, 127
NC, 127
NCYCLES, 69
ND, 127
NE, 127
NG, 127
NGEN, 69
NHOT, 69
NONE, 110
NONPAR, 99
NRD, 85
NRE, 85
NRS, 85
INDEX
163
NS, 116
NSB, 99
NSS, 86
NTSI, 128
NUM , =1,2,3,4,5, 100
NUMB, 99
NUPDATES, 69
OMEGA, 106
OPTION
ACCELERATION, 103
CCLP, 103
CP, 103
CURRENT, 103
EA, 103
ED, 104
GCHCALC, 104
GRIDFREEZE, 104
IDEALITY, 104
INTRINSIC, 104
LCPUMP, 105
LDRAIN, 105
LSOURCE, 105
LVDMOSFET, 105
M3MODE, 106
MODEL, 105
MPOLY, 105
OMEGA, 106
PDOP, 106
PHYSCK, 106
RBULK, 107
RDRAIN, 107
RSOURCE, 107
TEMP, 107
UFB, 107
UN, 108
UP, 108
OUTPUT
ALL, 109
AS, 109
AVAL, 109
BI, 109
CC, 109
DC, 109
DIGITS, 109
ELAT, 109
ETRAN, 109
IV, 109
JLAT, 110
JTRAN, 110
MAJ, 110
MIN, 110
MOB, 110
MTV, 110
NONE, 110
OXIDE, 110
PHI, 110
PIF, 110
PSI, 110
RHO, 111
STRIPS, 111
TC, 111
OXIDE, 110
OXIDEF, 74
PACCREL, 69
PDOP, 106
PFSS, 88
PGMI, 80
PHI, 110
PHYSCK, 106
PIF, 110
PNSS, 87
PROFILE
ACTIVATION, 112
AKEV, 113
ASYM, 113
C1, 114
C2, 114
CA, 113
CD, 113
DMOFF, 114
DOSE, 114
EA, 114
ED, 114
ELEM, 115
FILE, 115
LD, 115
LFIT, 115
NB, 116
NS, 116
SMOFF, 116
SUBS, 116
TEMP, 117
TIME, 117
TOX, 117
INDEX
164
VAGD, 118
VAGS, 118
VAGX, 118
VAUD, 118
VDEN, 118
VDGD, 119
VDGS, 119
VDGX, 119
VDUD, 119
XOFF, 117
YOFF, 118
PSI, 110
PSS, 86
R1GRD, 74
R2GRD, 75
RBULK, 107
RDRAIN, 107
RECESS, 74
RECOMBINATION
AN, 120
AP, 120
BN, 120
BP, 120
CAN, 120
CAP, 121
CDN, 121
CDP, 121
CN, 121
CP, 121
CSEA, 121
CSED, 122
CSHA, 122
CSHD, 122
DN, 122
DP, 122
EN, 122
EP, 122
FN, 123
FP, 123
MTUN, 123
SN, 123
SP, 123
TEXF, 123
TLIF, 124
TN, 124
TP, 124
VN, 124
VP, 124
RHO, 100, 111
RSOURCE, 107
S1GAP, 75
S2GAP, 75
1 2 3 4 5, 100
SAV ,
SCGRD, 75
SDREOX, 75
SEED, 100
SGAP, 66
SISS, 86
SMOFF, 116
SN, 123
SOVL, 100
SP, 123
SRGRD, 75
STEP
C1, 125
C2, 126
DB, 125
DD, 125
DG, 125
E1, 126
E2, 126
ELOG, 126
LUDG, 126
LUGB, 126
NB, 127
NC, 127
ND, 127
NE, 127
NG, 127
VG, 127
STRIPS, 111
SUBS, 116
SXTun, 80
SYTun, 81
T, 128
T1GRD, 76
T2GRD, 76
TC, 111
TCERR, 69
TD, 75
TEMP, 71, 84, 107, 117
TEXF, 123
TGATE, 75
TIME, 72, 84, 117
INDEX
TINS, 66
TLIF, 124
TN, 124
TOX, 117
TP, 124
TRACE, 100
TRANSIENT
NTSI, 128
T, 128
U, 128
W, 128
TS, 76
U, 128
UB, 63
UD, 63
UFB, 107
UG, 63
UN, 108
UP, 108
US, 63
UTMIN, 100
VAGD, 118
VAGS, 118
VAGX, 118
VAUD, 118
VDEN, 118
VDGD, 119
VDGS, 119
VDGX, 119
VDUD, 119
VG, 127
VI, 63
VN, 124
VP, 124
VR, 63
VSOUND, 101
W, 66, 128
WBH, 101
WBII, 101
WDIST, 101
WEII, 101
WINFREEZE, 101
WSCREEN, 101
XE1, 81
XE2, 81
XE3, 81
XISS, 86
165
XLEFT, 66
XOFF, 117
XRIGHT, 66
XS1, 81
XS2, 81
XS3, 82
XWIDTH, 102
YDEPTH, 67
YE, 82
YMDEP, 76
YOFF, 118
YS, 82
YWIDTH, 102
ZE, 82
ZS, 82
166
INDEX