You are on page 1of 199

N.B.

Balamurugan
Associate Professor
ECE Department
Thiagarajar College of Engineering
Madurai-15 (nbbalamurugan @tce.edu)
VLSI DEVICE MODELING
AND SIMULATION
UNIT II NOISE MODELING 9
Noise sources in MOSFET, Flicker noise modeling, Thermal noise modeling, model for
accurate distortion analysis, nonlinearities in CMOS devices and modeling, calculation
ofdistortion in analog CMOS circuit
UNIT III BSIM4 MOSFET MODELING 9
Gate dielectric model, Enhanced model for effective DC and AC channel length and
width, Threshold voltage model, Channel charge model, mobility model, Source/drain
resistance model, I-V model, gate tunneling current model, substrate current models,
Capacitance models, High speed model, RF model, noise model, junction diode
models, Layout-dependent parasitics model.
UNIT IV OTHER MOSFET MODELS 9
The EKV model, model features, long channel drain current model, modeling second
order effects of the drain current, modeling of charge storage effects, Non-quasi-static
modeling, noise model temperature effects, MOS model 9, MOSAI model)
UNIT V MODELLING OF PROCESS VARIATION AND QUALITY 9
ASSURANCE
Influence of process variation, modeling of device mismatch for Analog/RF
Applications, Benchmark circuits for quality assurance, Automation of the tests
OLD SYALLABUS OF SSDMAS
24 October 2013
2
OUTLINE
UNIT 2: MESH, NODE, MODIFIED NODAL ANALYSIS
Solving Linear Networks : Sparse Matrix
Solving Non linear Networks using NEWTON
RAPHSON Method.
UNIT 3 Stiffness and MULTISTEP Method
UNIT 4 Finite Difference Solution to
Poisson Equation, Continuity equation,
Schodinger Equation, drift diffusion Equation
HydroDynamic Equation
24 October 2013
3
Introduction
Circuit Simulations
Circuit Simulators SPICE
Device Equations
Network Equations
Elements
Passive and Active Elements
Equivalent circuit Model
Networks
Mesh, Nodal, Modified Nodal, Hybrid
Solution of Linear Circuit Equation
Solution of Nonlinear Circuit Equation
Solution of Differential Circuit Equation
Solution of Partial Differential Circuit Equation
24 October 2013
4
CIRCUIT SIMULATION
It is a technique for checking and verifying the design
of electrical and electronic circuits and systems.
Circuit simulation combines
Mathematical modeling of the circuit elements or devices.
Formulation of circuit / Network Equation.
Techniques for solution of these equations.
24 October 2013
5
Circuit Simulation
Simulator:
Solve numerically
Input and setup
Circuit
Output
Types of analysis:
DC Analysis
DC Transfer curves
Transient Analysis
AC Analysis, Noise, Distortions, Sensitivity
) ( ) (
) ( ) (
) (
) (
t FU t DX Y
t BU t GX
dt
t dX
C X f
+ =
+ = =
dt
t dX
C X f
) (
) ( =
24 October 2013
6
Program Structure (a closer look)
Numerical Techniques:
Formulation of circuit equations
Solution of ordinary differential equations
Solution of nonlinear equations
Solution of linear equations
Input and setup Models
Output
24 October 2013
7
CIRCUIT SIMULATORS
They use a Detailed (so called circuit level /
Transistor Level) description of the circuit and perform
relatively accurate simulation.
Typically a simulation uses physical models of the circuit
elements,solves the resuliting differential and algebraic
equations and generates time waveforms of node
Voltages and element Currents.
Early techniques for circuit simulation using computer
were introduced in 1950 and 1960s.
24 October 2013
8
SPICE
It is Universal standard simulator used to simulate
the operation of various electric circuits and devices.
It was developed by L.W.Nagel at the Univ. of California,
Berkeley in 1968.
The first simulator was named as CANCER (Computer Analysis
of Non-Linear Circuits Excluding Radiation).
It could not handle more components or circuit nodes.
1970 Improvements in CANCER continued
1971 an improved version of CANCER named SPICE 1 was
released.
1975 SPICE2 was introduced.
1983 SPICE 2G.6 version was released. All these version
were written in FORTRAN source code. Later it was
rewriten in C.
24 October 2013
9
SPICE 3 New C version of the program
Many SPICE like simulators in the market.
Meta-softwares HSPICE
Intusofts IS-SPICE
Spectrum Softwares MICRO-CAP
Microsims PSPICE
Texas instruments TINA SPICE
National Instruments LT SPICE
24 October 2013
10
24 October 2013
11
24 October 2013
12
DEVICE EQUATIONS
A device is a any simulation elements described by
means of a current voltage relationship.
Ex: Resistor
V = I R
I is found from V.
Linear Element
When the element equation contains no terms with
powers of 2 or higher, the element is said to be linear
Element circuit circuit.
Linear Circuit
A network of linear elements is said to be a linear circuit.
24 October 2013
13
Branch Constitutive Equations
(BCE)
Ideal elements
Element Branch Eqn Variable parameter
Resistor v = R i -
Capacitor i = C dv/dt -
Inductor v = L di/dt -
Voltage Source v = v
s
i = ?
Current Source i = i
s
v = ?
VCVS v
s
= A
V
v
c
i = ?
VCCS i
s
= G
T
v
c
v = ?
CCVS v
s
= R
T
i
c
i = ?
CCCS i
s
= A
I
i
c
v = ?
24 October 2013
14
NETWORK EQUATIONS
ELEMENTS and NETWORKS
An element is a two terminal electrical device.
An electrical Network or Circuit is a system consisting of
a set of elements and a set of nodes, where every element
terminal is unique node,and every node is identified
with atleast the element terminal.
A network is completely connected, i.e there is always
atleast one path from one node to another.
24 October 2013
15
Passive Elements
24 October 2013
16
24 October 2013
17
24 October 2013
18
24 October 2013
19
24 October 2013
20
24 October 2013
21
24 October 2013
22
24 October 2013
23
24 October 2013
24
NETWORK
They are two basic important techniques used in finding solutions
for a network
MESH ANALYSIS
If a network has a large
number of voltage sources, it is
useful to use MESH analysis.
It is only applicable only for
planar networks.
1. Check whether the circuit
is applicable.
2. To select mesh current
3. Writing Kirchhoffs voltage
law equations in terms of
unknowns and solving them to
the final solution
NODE ANALYSIS
If a network has more
current sources, Node
analysis is more useful.

24 October 2013
25
DC MESH Analysis
BRANCH AND MESH CURRENTS:
Applying KVL around the left loop
Applying KVL around the right loop
MATRICES AND MESH CURRENTS:
The n simultaneous equations of an n-mesh network can be written in
matrix form
When KVL is applied to the three mesh network
24 October 2013
26
Basic Circuits
Mesh Analysis: Example 7.1.
Write the mesh equations and solve for the currents I
1
, and I
2
.
+
_
10V
4 O 2 O
6 O
7 O
2V
20V
I
1
I
2
+
+
_
_
Figure 7.2: Circuit for Example 7.1.
Mesh 1
4I
1
+ 6(I
1
I
2
) = 10 - 2
Mesh 2
6(I
2
I
1
) + 2I
2
+ 7I
2
= 2 + 20
Eq (7.9)
Eq (7.10)
24 October 2013
27
Basic Circuits
Mesh Analysis: Example 7.1, continued.
Simplifying Eq (7.9) and (7.10) gives,
10I
1
6I
2
= 8
-6I
1
+ 15I
2
= 22
Eq (7.11)
Eq (7.12)
% A MATLAB Solution

R = [10 -6;-6 15];

V = [8;22];

I = inv(R)*V
I =
2.2105
2.3509
I
1
= 2.2105
I
2
= 2.3509
24 October 2013
28
Basic Circuits
Mesh Analysis: Example 7.2
+
_
6 O
10 O
9 O
11 O
3 O
4 O
20V 10V
8V
12V
I
1
I
2
I
3
+
+
_
_
_
_
+
+
_
Mesh 1: 6I
1
+ 10(I
1
I
3
) + 4(I
1
I
2
) = 20 + 10
Mesh 2: 4(I
2
I
1
) + 11(I
2
I
3
) + 3I
2
= - 10 - 8
Mesh 3: 9I
3
+ 11(I
3
I
2
) + 10(I
3
I
1
) = 12 + 8
Eq (1)
Eq (2)
Eq (3)
24 October 2013
29
Basic Circuits
Mesh Analysis:
Clearing Equations (1), (2) and (3) gives,
20I
1
4I
2
10I
3
= 30
-4I
1
+ 18I
2
11I
3
= -18
-10I
1
11I
2
+ 30I
3
= 20
In matrix form:
(
(
(

=
(
(
(

(
(
(




20
18
30
3
2
1
30 11 10
11 18 4
10 4 20
I
I
I
WE NOW MAKE AN IMPORTANT
OBSERVATION!!
Standard Equation form
24 October 2013
30
Basic Circuits
Mesh Analysis: Standard form for mesh equations
Consider the following:
R
11
=

of resistance around mesh 1, common to mesh 1 current I


1
.
R
22
=
of resistance around mesh 2, common to mesh 2 current I
2
.
R
33
= of resistance around mesh 3, common to mesh 3 current I
3
.
(
(
(

=
(
(
(

(
(
(

) 3 (
) 2 (
) 1 (
3
2
1
33 32 31
23 22 21
13 12 11
emfs
emfs
emfs
I
I
I
R R R
R R R
R R R
24 October 2013
31
NODE VOLTAGE METHOD
In a N node circuit one of the nodes is chosen as
reference, then it is possible to write N-1 nodal
equations by assuming N-1 node voltages.
Applying KCL at node 1,2,3
24 October 2013
32
Nodal Analysis
24 October 2013
33
STEADY STATE AC ANALYSIS
Two analysis have applied it only resistive circuits.
These concepts can also be used for sinusoidal steady
state condition.
In the sinusoidal steady state analysis,
we use voltage phasors, current Phasors,
impedances,admittances, to write branch
equations, KVL and KCL equations.
For ac circuits, the method of writing loop equation is
modified slightly. If the impedances are complex,
the sum of their voltages is found by vector addition.
24 October 2013
34
Modified Nodal Approach
It is another Method of writing circuit equations that is less
demanding of space and almost as flexible.
It allows a broad range of elements to be modeled.
we will consider networks with two terminals elements only.
We will label the elements as either being current controlled or
voltage controlled.
R , Inductor,and Independent voltage source will be considered
current controlled.
Conductors , Capacitors and independent currnet source will
be considered voltage controlled.
24 October 2013
35
If the branch relation is written
i = f(v)
The element is considered voltage controlled.
If the branch relation is written
v = g(i)
The element is considered current controlled.
This generates a set of equations with more
unknowns than equations
24 October 2013
36
24 October 2013
37
38
Advantages and problems of MNA
24 October 2013
39
Modified Nodal Analysis (MNA)
24 October 2013
40
Modified Nodal Analysis (2)
(
(
(
(
(
(
(
(

=
(
(
(
(
(
(
(
(

(
(
(
(
(
(
(
(
(
(
(
(

+
|
|
.
|

\
|
+ + +
0
6
0
0
0
0 0 1 0 7 7
0 0 0 1 1 0
1 0
1 1
0 0
0 1
1 1
0 0
0 1 0 0
1 1 1
0 0 0 0
1 1 1
5
7
6
4
3
2
1
8 8
8 8
4 3 3
3
2
3
2
1
ES
i
i
i
e
e
e
e
E E
R R
R R
R R R
R
G
R
G
R
s
24 October 2013
SOLVING OF NETWORK
EQUATIONS
USING SPARSE MATRICES
SPARSE MATRIX
A sparse matrix is a matrix in which the great majority of the
elements are zeros. Such matrices are used to solve network equations
and other engineering disciplines.
SPARSITY IN A CIRCUIT
Consider a complex linear circuit, containing large number of nodes.
By writing the node equations we came to know that the matrix have
large number of zeroes since each node is connected to only a few other
nodes. Sparsity is a key feature of large scale circuits such as VLSI digital
circuits or electric networks.
Sparse matrix
x x 0 0 0 0 0 0
0 x x 0 0 0 0 0
0 0 x x 0 0 0 0 An Sparse Matrix of order 8x8
0 0 0 x x 0 0 0 x-non-zero elements
0 0 0 0 x x 0 0
0 0 0 0 0 x x 0
0 0 0 0 0 0 x x
0 0 0 0 0 0 0 x
A circuits node equations when represented in matrix will also have same
matrix like the above one.
Storing of all entries need more space and harder to manipulate. Hence we
can store only the non zero elements and it will be easy to manipulate.
Solving sparse matrices
In network theory, big problem is solving a large system of linear
equations.
Ax=B
Sparse matrix can be solved using LU factorization.
Solution procedure
Rewrite A=LU,where L is Lower triangular matrix,U is upper triangular
matrix.
Solve Lz=b for z.
Solve Ux=z for b.
5 1 2 1 0 0 5 1 2
A= 1 4 1 = 0.2 1 0 0 3.8 0.6
2 2 5 0.4 0.42 1 0 0 3.94
L U
Fill-in
When a sparse matrix is factorize into L and U ,it contains non zero
elements in places whereas the same place in sparse matrix contains zeros.
This unwanted non zero elements are called as fill-ins.
Fill-in is a major problemin large matrices. Hence it should be reduced.
It can be either reduced by permuting or more easier elimination graphs. Also
in elimination graphs minimal degree ordering is the best way to reduce
fill-ins.
example
for the resistor circuit given ,we can
Apply sparsity since each node is
connected to few other nodes. While
Solving,a matrix of order 12x12 will be
obtained with large number of zeros,
Which is a sparse matrix.
Newtons Method,
Solving NonLinear Equation
Copyright The McGraw-Hill Companies, Inc. Permission required for reproduction or display.
History
Discovered by Isaac Newton and published in
his Method of Fluxions, 1736
Joseph Raphson described the method in
Analysis Aequationum in 1690
Method of Fluxions was written earlier in 1671
Today it is used in a wide variety of subjects,
including Computer Vision and Artificial
Intelligence
October 24, 2013
51
Review: Classification of
Equations
Linear: independent variable appears to the
first power only, either alone or multiplied by a
constant
Nonlinear:
Polynomial: independent variable appears raised to
powers of positive integers only
General non-linear: all other equations
October 24, 2013
52
Review: Solution Methods
Linear: Easily solved analytically
Polynomials: Some can be solved analytically
(such as by quadratic formula), but most will
require numerical solution
General non-linear: unless very simple, will
require numerical solution
October 24, 2013
53
Newtons Method
Newtons Method (also know as the Newton-
Rapshon Method) is another widely-used
algorithm for finding roots of equations
In this method, the slope (derivative) of the
function is calculated at the initial guess value
and projected to the x-axis
The corresponding x-value becomes the new
guess value
The steps are repeated until the answer is
obtained to a specified tolerance
October 24, 2013
54
Newtons Method
Initial guess x
i
y(x
i
)
Tangent
Line: Slope =
y(x
i
)
October 24, 2013
55
Newton Raphson
( )
( )
( )
( )
i
i
i i
i i
i
i
x f
x f
x x
rearrange
x x
x f
x f
f
dx
dy
gent tan
'
0
'
'
1
1
=


=
= =
+
+
f(x
i
)
x
i
tangent
x
i+1
If Initial guess at the root is xi,
a tangent can be extended from
the point [xi,f(xi)]. The point where
this tangent crosses the x axis usually
represents an improved estimate
of the root.
f(x
i
) -
0
Slope =
( )
i
x f
'
( )
i
x f
'
October 24, 2013
56
Newtons Method
x
i
y(x
i
)
y(x
i
)/y(x
i
)
New guess for x:
x
n
= x
i
- y(x
i
)/y(x
i
)
October 24, 2013
57
Newtons Method Example
Find a root of this equation:
The first derivative is:
Initial guess value: x = 10
October 24, 2013
58
Newtons Method Example
For x = 10:
This is the new value of x
October 24, 2013
59
Newtons Method Example
Initial
guess =
10
y = 1350
y = 580
New x-value =
7.6724
October 24, 2013
60
Newtons Method Example
For x = 7.6724:
October 24, 2013
61
Newtons Method Example
Initial
value =
7.6724
y = 368.494
y = 279.621
New x-value =
6.3546
October 24, 2013
62
Newtons Method Example
Continue iterations:
Method quickly converges to this root
October 24, 2013
63
Newtons Method - Comments
Usually converges to a root much faster than the
bisection method
In some cases, the method will not converge
(discontinuous derivative, initial slope = 0, etc.)
In most cases, however, if the initial guess is
relatively close to the actual root, the method will
converge
Dont necessarily need to calculate the derivative:
can approximate the slope from two points close
to the x-value. This is called the Secant Method
October 24, 2013
64
October 24, 2013
65
Need to Solve
( 1) 0
d
t
V
V
d s
I I e =
Nonlinear Problems - Example
0
1
I
r
I
1
I
d
0 ) 1 (
1
0
1 1
1
1
= +
= +
I e I e
R
I I I
t
V
e
s
d r
1 1
) ( I e g =
October 24, 2013
66
Nonlinear Equations
Given g(V)=I
It can be expressed as: f(V)=g(V)-I
Solve g(V)=I equivalent to solve f(V)=0
Hard to find analytical solution for f(x)=0
Solve iteratively
October 24, 2013
67
Nonlinear Equations Iterative Methods
Start from an initial value x
0
Generate a sequence of iterate x
n-1
, x
n
, x
n+1
which hopefully converges to the solution x*
Iterates are generated according to an iteration
function F: x
n+1
=F(x
n
)
October 24, 2013
68
Newton-Raphson (NR) Method
Consists of linearizing the system.
Want to solve f(x)=0 Replace f(x) with its linearized version
and solve.
Note: at each step need to evaluate f and f
function Iteration x f
dx
x df
x x
x x
dx
x df
x f x f
ies Taylor Ser x x
dx
x df
x f x f
k
k
k k
k k
k
k k
) (
) (
) (
) (
) ( ) (
... ) * (
) (
) ( *) (
1
1
1 1

+
+ +
(

=
+ =
+ + =
Solving Nonlinear Equations
nonlinear elements when simulating integrated
circuits.
For a simple nonlinear equation
f(x)=0
the usual method of solution is some variant on
the Newton-Raphson (NR) procedure.
October 24, 2013
69
In order to understand the NR procedure we
must realize that the entire procedure is based
upon a linear approximation to the function at
the current point, xc.
Thus, if we expand the function in a Taylor
series about the current point we have
if we further assume that the x
+
is the solution,
i.e.,
(which is the basic NR step.)
October 24, 2013
70
Newton-Raphson example.
October 24, 2013
71
Newton-Raphson Applied to Circuit
Simulation
The first difficulty is that nonlinear circuit equations are not just a
function of a single variable. Further, there is a set of equations and not
just a single one. This problem is easily handled by extending the 1D
NR Scheme to multidimensions using vector calculus.
The update formula now becomes
were f/x is known is the Jacobian of the set of circuit equations and
is the matrix of partial derivatives.
October 24, 2013
72
Rewriting this set of equations we have
This is precisely in the form
The final difficulty we face that is how to generate the equations in the first place
and how to take the partial derivative of this large set of equations.
First individual elements can be linearised and the equation written for this network
And we still have the same set of equations required for NR.
October 24, 2013
73
Example
We will assume a simple form for the diode equation
The linearized form of this becomes
If we rewrite this grouping the known and unknown quantities we have
We will consider the network in 8.49
October 24, 2013
74
Nonlinear network used for NR
October 24, 2013
75
October 24, 2013
76
October 24, 2013
77
Adv and disadv of NR method
ADV:
convergence rate for Newtons method is very high.
Error estimates are very good.
NR method can find complex roots.
DIS ADV:
If the local min/max is selected as an initial Guess,the slope does not interset
with X axis.
The formula for xi will lead to an infinite value.
In-Class Exercise
Draw a flow chart of Newtons Method
Write the MATLAB code to apply Newtons
Method to the previous example:
October 24, 2013
78
Define converge
tolerance tol
while abs(f(x)) > tol
Input initial guess x
Calculate f(x)
Calculate slope fpr(x)
x = x f(x)/fpr(x)
Calculate f(x)
Output root x
October 24, 2013
79
MATLAB Code
MATLAB functions defining the function and
its derivative:
October 24, 2013
80
MATLAB Code
October 24, 2013
81
MATLAB Results
>> Newton
Enter initial guess
10
Root found: 5.6577
>> Newton
Enter initial guess
0
Root found: 1.4187
>> Newton
Enter initial guess
-10
Root found: -2.0764
October 24, 2013
82
Excel and MATLAB Tools
General non-linear equations:
Excel: Goal Seek, Solver
MATLAB: fzero
Polynomials:
MATLAB: roots
Graphing tools are also important to locate
roots approximately
October 24, 2013
83
rootsExample
For polynomials, the MATLAB function roots
finds all of the roots, including complex roots
The argument of roots is an array containing
the coefficients of the equation
For example, for the equation
the coefficient array is [3, -15, -20, 50]
October 24, 2013
84
rootsExample
>> A = [3, -15, -20, 50];
>> roots(A)
ans =
5.6577
-2.0764
1.4187
October 24, 2013
85
rootsExample
Now find roots of
>> B = [3, -5, -20, 50];
>> roots(B)
ans =
-2.8120
2.2393 + 0.9553i
2.2393 - 0.9553i
Two complex roots
October 24, 2013
86
Summary
The bisection method and Newtons method
(or secant method) are widely-used algorithms
for finding the roots of equations
When using any tool to find the roots of non-
linear equations, remember that multiple roots
may exist
The initial guess value will affect which root is
found
October 24, 2013
87
88
Ordinary Differential Equations
Equations which are composed of an unknown
function and its derivatives are called differential
equations.
Differential equations play a fundamental role in
engineering because many physical phenomena are
best formulated mathematically in terms of their rate
of change.
v- dependent variable
t- independent variable
v
m
c
g
dt
dv
=
89
When a function involves one dependent variable, the
equation is called an ordinary differential equation
(or ODE). A partial differential equation (or PDE)
involves two or more independent variables.
Differential equations are also classified as to their
order.
A first order equation includes a first derivative as its
highest derivative.
A second order equation includes a second derivative.
Higher order equations can be reduced to a system of
first order equations, by redefining a variable.
90
Figure PT7.2
UNIT 3:Stiffness and
Multistep Methods
91
Stiff problem:
1. Natural time constants
2. Input time constants
3. Interval of interest
If these are widely separated, then the problem is
stiff
92
Stiff system are both individual and systems of ODEs
that have both fast and slow components to their
solution.
We introduce the idea of an implicit solution technique
as one commonly used remedy for this problem
93
Stiffness and Multistep Methods
Two areas are covered:
Stiff ODEs will be described - ODEs that have both
fast and slow components to their solution.
Implicit solution technique and multistep methods
will be described.
94
STIFFNESS
In many cases, the rapidly varying components are transient that die away
quickly,after which the solution becomes dominated by the slowly varying
components. Although the transient phenomena exist for only a short part of
the integration interval, they can dictate the time step for the entire solution.
95
STIFFNESS
Definition
Proof one problem
Solving the problem by Numerically
Explicit Eulers Method
Implicit Eulers Method
Step size
Literature survey
96
Stiffness
A stiff system is the one involving rapidly changing
components together with slowly changing ones.
Both individual and systems of ODEs can be stiff:
If y(0)=0, the analytical solution is developed as:
t
e y
dt
dy

+ = 2000 3000 1000
t t
e e y

= 002 . 2 998 . 0 3
1000
Suppose that lemda1 = 1000 and lemda2 =1 are widely
sepatared and giving a stiff system. Note that the first
exponential waveform dies out in 5 micro secs.
To get an accurate solution of the fast waveform we need
a small time step and to have an efficient simulatin of the
slow waveform we eed a large time step. The obvious
97
Figure 26.1
98
99
100
If Eulers method is used to solve the problem
numerically:
The stability of this formula depends on the step size
h:
) 1 ( or
1 1
1
ah y y h ay y y
h
dt
dy
y y
i i i i i
i
i i
= =
+ =
+ +
+

i y a h
ah
i
as / 2
1 1

101
Thus, for transient part of the equation, the step size
must be <2/1000=0.002 to maintain stability.
While this criterion maintains stability, an even
smaller step size would be required to obtain an
accurate solution.
Rather than using explicit approaches, implicit
methods offer an alternative remedy.
An implicit form of Eulers method can be developed
by evaluating the derivative at a future time.
102
103
Insight into the step size required for stability of such a
solution can be gained by examining the homogeneous
part of the ODE:
The solution starts at y(0)=y
0
and asymptotically
approaches zero.
at
e y y
ay
dt
dy

=
=
0
is the solution.
104
If Eulers method is used to solve the problem
numerically:
The stability of this formula depends on the step size
h:
) 1 ( or
1 1
1
ah y y h ay y y
h
dt
dy
y y
i i i i i
i
i i
= =
+ =
+ +
+

i y a h
ah
i
as / 2
1 1

105
ah
y
y
h ay y y
h
dt
dy
y y
i
i
i i i
i
i i
+
=
=
+ =
+
+ +
+
+
1
1
1 1
1
1
i as y
i
0
Backward or implicit Eulers method
The approach is called unconditionally stable.
Regardless of the step size:
ay
dt
dy
=
106
107
MULTI STEP METHODS
Definition
Non self staring Heun method
Eulers method as Predictor
Trapezoidals rule as a corrector
Slightly Modified Predictor and corrector
Derivation and Error analysis of Predictor- corrector Formulas
Error Estimates
Modifiers
One step methods give information at a single xi to predict a value of
the dependent variable yi+1 at a future point xi+1.
Multistep gives valuable information from previous points. The curvature of
the lines connecting these previous values that provides information regarding
the trajectory of the solution.
At each point, tn, we are going to compute an approximation, xn, to
the exact solution, x(tn). A large number of methods are available for
108
Graphical depiction of the fundamental difference between (a) one step and
(b) Multistep methods for solving ODEs.
109
Multistep Methods
The Non-Self-Starting Heun Method
It is characteristic of most multistep methods.
It use an open integration formula (the mid point method) to
make an initial estimate.
Huen method uses Eulers method as a predictor and
trapezoidal rule as a corrector.
Predictor is the weak link in the method because it has the
greatest error, O(h
2
).
One way to improve Heuns method is to develop a predictor
that has a local error of O(h
3
).
h y x f y y
i i i i
2 ) , (
1
0
1
+ =
+
110
Multistep Methods
Corrector formula
and you iterate until a maximum number of iteratios
is reached.
| |
h
y x f y x f
y y
j
i i i i
i
j
i
2
) , ( ) , (
1
1 1
1

+ +
+
+
+ =
111
112
113
114
115
116
117
Integration Formulas/
Newton-Cotes Formulas.
Open Formulas.
Closed Formulas.
dx x f y y
i
n i
x
x
n n i i
) (
1
1
}
+

+ =
+
f
n
(x) is an n
th
order interpolating
polynomial.
dx x f y y
i
n i
x
x
n n i i
) (
1
1
1 1
}
+
+
+ =
+ +
118
Adams Formulas (Adams-Bashforth).
Open Formulas.
The Adams formulas can be derived in a variety of
ways. One way is to write a forward Taylor series
expansion around x
i
. A second order open Adams
formula:
Closed Formulas.
A backward Taylor series around x
i+1
can be written:
) (
12
5
2
1
2
3
4 3
1 1
h O f h f f h y y
i i i i i
+
' '
+
|
.
|

\
|
+ =
+
) (
1
1
0
1 1
+

=
+ +
+ + =

n
n
k
k i k i i
h O f h y y |
Listed in Table
26.2
119
Higher-Order multistep Methods/
Milnes Method.
Uses the three point Newton-Cotes open
formula as a predictor and three point Newton-
Cotes closed formula as a corrector.
Fourth-Order Adams Method.
Based on the Adams integration formulas.
Uses the fourth-order Adams-Bashforth
formula as the predictor and fourth-order
Adams-Moulton formula as the corrector.
Finite Difference Method
Schrodinger equation
Need of Scaling (Moores Law )
Gordon Moore, a physical chemist working in electronics, made a prediction
in 1965, the number of transistors on an integrated chip, would double every
18 months.
Dr. Gordon Moore
15/3/2012
121
Karunya
Device Modeling Enable System
Transformations
15/3/2012
122
Karunya
Smaller
Faster
Cheaper
Miniaturisation: why?
15/3/2012
123
Karunya
1980 1985 1988 1991 1994 1999 2003
L
GATE
= 10 m 5 m 2.5 m 1.3 m 0.63 m 0.25 m 130 nm .
1980 1985 1988 1991 1994 1999 2003
L
GATE
= 10 m 5 m 2.5 m 1.3 m 0.63 m 0.25 m 130 nm .
1980 1985 1988 1991 1994 1999 2003
L
GATE
= 10 m 5 m 2.5 m 1.3 m 0.63 m 0.25 m 130 nm .
1980 1985 1988 1991 1994 1999 2003
L
GATE
= 10 m 5 m 2.5 m 1.3 m 0.63 m 0.25 m 130 nm .
1980 1985 1988 1991 1994 1999 2003
L
GATE
= 10 m 5 m 2.5 m 1.3 m 0.63 m 0.25 m 130 nm .
1980 1985 1988 1991 1994 1999 2003
L
GATE
= 10 m 5 m 2.5 m 1.3 m 0.63 m 0.25 m 130 nm .
1980 1985 1988 1991 1994 1999 2011
L
GATE
= 10 m 5 m 2.5 m 1.3 m 0.63 m 0.25 m 22 nm .
Scaling
Shrink dimensions maintaining aspect-ratio
Must shrink electrostatic features as well (depletion regions doping
level and profiles)
15/3/2012
124
Karunya
In Classical systems,
The particle will take a path between
the two positions.
In Quantum systems,
Instead, the particle is a
wave and it doesnt
take one path from the
initial position to the final
position, it takes all
possible paths.
DOS- Density of states
Describes the number of
states at each energy
levels that are available
to be occupied by charge
carriers.
is the
Momentum of the
charge carriers.
p =
Typically, near the silicon surface, the inversion layer charges are confined to a
potential well formed by:
Oxide barrier
bend Si-conduction band at the surface due to the applied voltage, V
g
Due to the confinement of inversion layer e
-
:
e
-
energy levels are grouped in discrete sub-bands of energy, E
j
Energy Wavefunction Particle Reduced
Potential
mass Plank
Constant
If the potential is dependent of x, then the solution is
based on numerical solution

( ) ( )
x
x f x x f
lim ) x ( ' f
x
V
V +
=
V 0
Continuity Equation
Derivation of Continuity
Equation
Consider carrier-flux into/out-of an infinitesimal volume:
J
n
(x) J
n
(x+dx)
dx
Area A, volume Adx
| | Adx
n
A dx x J A x J
q t
n
Adx
n
n n
t
A
+ =
|
.
|

\
|
c
c
) ( ) (
1
EE130/230M Spring 2013 Lecture 6, Slide 139
EE130/230M Spring 2013 Lecture 6, Slide 140
n
n
n
x
x J
q t
n
t
A

c
c
=
c
c
=>
) ( 1

L
p
p
G
p
x
x J
q t
p

) (
1
+
A

c
c
=
c
c
t
Continuity
Equations:
dx
x
x J
x J dx x J
n
n n
c
c
+ = +
) (
) ( ) (
L
n
n
G
n
x
x J
q t
n

) ( 1
+
A

c
c
=
c
c
t
Derivation of
Minority Carrier Diffusion
Equations
The minority carrier diffusion equations are derived from
the general continuity equations, and are applicable only
for minority carriers.
Simplifying assumptions:
1. The electric field is small, such that
in p-type
material
in n-type material
2. n
0
and p
0
are independent of x (i.e. uniform doping)
3. low-level injection conditions prevail
x
n
qD
x
n
qD n q J
n n n n
c
c
~
c
c
+ = c
x
p
qD
x
p
qD p q J
p p p p
c
c
~
c
c
= c
EE130/230M Spring 2013 Lecture 6, Slide 141
EE130/230M Spring 2013 Lecture 6, Slide 142
Starting with the continuity equation for electrons:
( ) ( )
L
n
n
G
n
x
n n
qD
x q t
n n

1
0 0
+
A

c
A + c
c
c
=
c
A + c
t
L
n
n
G
n
x
n
D
t
n

2
2
+
A

c
A c
=
c
A c
t
L
n
n
G
n
x
x J
q t
n

) ( 1
+
A

c
c
=
c
c
t
Carrier Concentration Notation
EE130/230M Spring 2013 Lecture 6, Slide 143
The subscript n or p is used to explicitly denote n-type or
p-type material, e.g.
p
n
is the hole (minority-carrier) concentration in n-type matl
n
p
is the electron (minority-carrier) concentration in n-type matl
L
p
n n
p
n
L
n
p p
n
p
G
p
x
p
D
t
p
G
n
x
n
D
t
n


2
2
2
2
+
A

c
A c
=
c
A c
+
A

c
A c
=
c
A c
t
t
Thus the minority carrier diffusion equations are
Summary
The continuity equations are established based on
conservation of carriers, and therefore hold generally:
The minority carrier diffusion equations are derived from
the continuity equations, specifically for minority carriers
under certain conditions (small E-field, low-level injection,
uniform doping profile):
L
p
n
L
n
n
G
p
x
x J
q t
p
G
n
x
x J
q t
n

) ( 1

) ( 1
+
A

c
c
=
c
c
+
A

c
c
=
c
c
t t
L
p
n n
P
n
L
n
p p
N
p
G
p
x
p
D
t
p
G
n
x
n
D
t
n

2
2
2
2
+
A

c
A c
=
c
A c
+
A

c
A c
=
c
A c
t t
EE130/230M Spring 2013 Lecture 6, Slide 144
Drift-Diffusion Modeling
Outline of the Lecture
Classification of PDEs
Why Numerical Analysis?
Numerical Solution Sequence
Flow-Chart of Equilibrium Poisson Equation Solver
Discretization of the Continuity Equation
Numerical Solution Techniques for Sparse Matrices
Flow-Chart of 1D Drift-Diffusion Simulator
Classification of PDEs
Classification of PDEs
Different mathematical and physical
behaviors:
Elliptic Type
Parabolic Type
Hyperbolic Type
System of coupled equations for several
variables:
Time : first-derivative (second-derivative for wave
equation)
Space: first- and second-derivatives
Classification of PDEs (cont.)
General form of second-order PDEs ( 2
variables)
PDE Model Problems
Hyperbolic (Propagation)
Advection equation (First-order linear)
Wave equation (Second-order linear )
PDE Model Problems (cont.)
Parabolic (Time- or space-marching)
Burgers equation (Second-order nonlinear)
Fourier equation (Second-order linear )
(Diffusion / dispersion)
PDE Model Problems (cont.)
Elliptic (Diffusion, equilibrium problems)
Laplace/Poisson (second-order linear)
Helmholtz equation
Boundary and Initial
Conditions
R
s
n
cR
Initial conditions: starting
point for propagation problems
Boundary conditions: specified
on domain boundaries to provide
the interior solution in
computational domain
Numerical Methods
Complex geometry
Complex equations (nonlinear, coupled)
Complex initial / boundary conditions
No analytic solutions
Numerical methods needed !!
Why Numerical Analysis?
Coupling of Transport Equations to
Poisson and Band-Structure Solvers
D. Vasileska and S.M. Goodnick, Computational
Electronics, published by Morgan & Claypool , 2006.
Drift-Diffusion Approach
Constitutive Equations
Poisson
Continuity Equations
Current Density Equations
1
1
J
J
n n
p p
n
U
t q
p
U
t q
c
= V +
c
c
= V +
c
( )
D A
V p n N N c
+
V V = +
( ) ( )
( ) ( )
n n n
p p p
dn
J qn x E x qD
dx
dn
J qp x E x qD
dx

= +
=
S. Selberherr: "Analysis and
Simulation of Semiconductor
Devices, Springer, 1984.
Poisson/Laplace Equation
Solution
Poisson/Laplace Equation
No knowledge of solving of PDEs
Method of images
With knowledge for solving of PDEs
Theoretical Approaches
Numerical Methods:
finite difference
finite elements
Poisson
Greens function method
Laplace
Method of separation of variables
(Fourier analysis)
Numerical Solution Sequence
Numerical Solution Details
Governing
Equations
ICS/BCS
Discretization
System of
Algebraic
Equations
Equation
(Matrix)
Solver
Approximate
Solution
Continu
ous
Solution
s
Finite-
Difference
Finite-
Volume
Finite-
Element
Spectral
Discret
e Nodal
Values
Tridiago
nal
SOR
Gauss-
Seidel
Krylov
Multigrid

i
(x,y,z,t)
p (x,y,z,t)
n (x,y,z,t)
D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes,
Arizona State University, Tempe, AZ.
What is next?
MESH
Finite Difference Discretization
Boundary Conditions
MESH TYPE
The course of action taken in three steps is dictated by the nature of the problem
being solved, the solution region, and the boundary conditions. The most commonly
used grid patterns for two-dimensional problems are
Common grid patterns: (a) rectangular grid, (b) skew grid, (c) triangular grid, (d)
circular grid.
Finite Difference Schemes
Before finding the finite difference solutions to specific PDEs, we will look at how
one constructs finite difference approximations from a given differential equation. This
essentially involves estimating derivatives numerically. Lets assume f(x) shown below:
Estimates for the derivative of f (x) at P using forward, backward, and central
differences.
Finite Difference Schemes
We can approximate derivative of f(x), slope or the tangent at P by the slope of the arc
PB, giving the forward-difference formula,
x
x f x x f
x f
A
A +
~ '
) ( ) (
) (
0 0
0
or the slope of the arc AP, yielding the backward-difference formula,
x
x x f x f
x f
A
A
~ '
) ( ) (
) (
0 0
0
or the slope of the arc AB, resulting in the central-difference formula,
x
x x f x x f
x f
A
A A +
~ '
2
) ( ) (
) (
0 0
0
Finite Difference Schemes
We can also estimate the second derivative of f (x) at P as
(

A
A

A
A +
A
=
A
A A + '
~ ' '
x
x x f x f
x
x f x x f
x
x
x x f x x f
x f
) ( ) ( ) ( ) ( 1
) 2 / ( ) 2 / (
) (
0 0 0 0
0 0
0
or
2
0 0 0
0
) ( ) ( 2 ) (
) (
x
x x f x f x x f
x f
A
A + A +
~ ' '
Any approximation of a derivative in terms of values at a discrete set of points is
called finite difference approximation.
Finite Difference Schemes
+ ' ' ' A + ' ' A + ' A + = A + ) ( ) (
! 3
1
) ( ) (
! 2
1
) ( ) ( ) (
0
3
0
2
0 0 0
x f x x f x x f x x f x x f
The approach used above in obtaining finite difference approximations is rather
intuitive. A more general approach is using Taylors series. According to the wellknown
expansion,
+ ' ' ' A ' ' A + ' A = A ) ( ) (
! 3
1
) ( ) (
! 2
1
) ( ) ( ) (
0
3
0
2
0 0 0
x f x x f x x f x x f x x f
and
Upon adding these expansions,
4
0
2
0 0 0
) ( ) ( ) ( ) ( 2 ) ( ) ( x x f x x f x x f x x f A O + ' ' A + = A + A +
where O(Ax)
4
is the error introduced by truncating the series. We say that this error is of the
order (Ax)
4
or simply O(Ax)
4
. Therefore, O(Ax)
4
represents terms that are not greater than (A
x)
4
. Assuming that these terms are negligible,
2
0 0 0
0
) ( ) ( 2 ) (
) (
x
x x f x f x x f
x f
A
A + A +
~ ' '
Finite Difference Schemes
+ ' ' ' A + ' ' A + ' A + = A + ) ( ) (
! 3
1
) ( ) (
! 2
1
) ( ) ( ) (
0
3
0
2
0 0 0
x f x x f x x f x x f x x f
+ ' ' ' A ' ' A + ' A = A ) ( ) (
! 3
1
) ( ) (
! 2
1
) ( ) ( ) (
0
3
0
2
0 0 0
x f x x f x x f x x f x x f
3
0 0 0
) ( ) ( ) ( 2 ) ( ) ( x x f x x x f x x f A O + ' A = A A +
Subtracting
from
We obtain
and neglecting terms of the order (Ax)
3
yields
x
x x f x x f
x f
A
A A +
~ '
2
) ( ) (
) (
0 0
0
This shows that the leading errors of the order (Ax)
2
. Similarly, the forward and backward
difference formula have truncation errors of O(Ax).
Poisson Equation
The Poisson equation is of the following general form:
It accounts for Coulomb carrier-carrier interactions in the
Hartree approximation
It is always coupled with some form of transport simulator
except when equilibrium conditions apply
It has to be frequently solved during the simulation procedure
to properly account for the fields driving the carriers in the
transport part
There are numerous ways to numerically solve this equation
that can be categorized into direct and iterative methods
) ( ) (
2
r f r = u V
Poisson Equation Linearization
The 1D Poisson equation is of the form:
( )
2
2
exp exp( / )
exp exp( / )
D A
F i
i i T
B
i F
i i T
B
d e
p n N N
dx
E E
n n n V
k T
E E
p n n V
k T

= +
| |

= =
|
\ .
| |

= =
|
\ .
Finite Difference Representation
1 1 1
1 1
2 2 2
1 2 1
( ) ( )
n n n
i i i i i
n
i i i i i i
n p
p n C p n

+ + +
+
| |
+ + + =
|
A A A
\ .
= + +
Equilibrium:
exp( ), exp( )
n n
i i i i
n p = =
Non-Equilibrium:
n calculated using PM coupling and p still calculated
as in equilibrium case (quasi-equilibrium approximation)
Boundary Conditions
There are three types of boundary conditions that
are specified during the discretization process of the
Poisson equation:
Dirichlet (this is a boundary condition on the potential)
Neumann (this is a boundary condition on the derivative of
the potential, i.e. the electric field)
Mixed boundary condition (combination of Dirichlet and
Neumann boundary conditions)
Note that when applying the boundary conditions for
a particular structure of interest, at least one point
MUST have Dirichlet boundary conditions specified
on it to get the connection to the real world.
1D Discretization
The resultant finite difference equations can be represented in a
matrix form Au= f, where:
x0 x1 x2 x3 x4
) (
2
;
2
;
) (
2
where
1 0
0
0
2
, ) , , , , , ( , ) , , , , , (
4 4 4
3 3 3
2 2 2
1 1 1
0 0
5 4 3 2 1 0 5 4 3 2 1 0
e
i
w
i
e
i
i
w
i
e
i
i
e
i
w
i
w
i
i
dx dx dx
c
dx dx
b
dx dx dx
a
c b a
c b a
c b a
c b a
c b
f f f f f
+
= =
+
=
|
|
|
|
|
|
.
|

\
|
=
u = u u u u u u =
A
f u
Neumann
Dirichlet
x
5
2D Discretization
In 2D, the finite-difference discretization of the Poisson equation
leads to a five point stencil:
N=5,M=4
Dirichlet: 0,4,5,9,10,14,15,19
Neuman: 1,2,3,16,17,18
xi-1 xi xi+1
xi-N
xi+N
w
i
dx
e
i
dx
s
i
dx
n
i
dx
xi-1 xi xi+1
xi-N
xi+N
w
i
dx
e
i
dx
s
i
dx
n
i
dx
0 1 2 3 4
5 6 7 8 9
10 11 12 13 14
15 16 17 18 19
Va
0 1 2 3 4
5 6 7 8 9 5 6 7 8 9
10 11 12 13 14 10 11 12 13 14
15 16 17 18 19 15 16 17 18 19
Va
2D Discretization (contd)
Dirichlet: 0,4,5,9,10,14,15,19
Neuman: 1,2,3,16,17,18
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
.
|

\
|
=
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
.
|

\
|
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
.
|

\
|
0
0
0
0
1
2
2
2
1
1
1
1
1
1
2
2
2
1
18
17
16
13
12
11
8
7
6
3
2
1
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
18 18 18 18
17 17 17 17
16 16 16 16
13 13 13 13 13
12 12 12 12 12
11 11 11 11 11
8 8 8 8 8
7 7 7 7 7
6 6 6 6 6
3 3 3 3
2 2 2 2
1 1 1 1
f
f
f
V
f
f
f
V
f
f
f
V
f
f
f
V
d c b a
d c b a
d c b a
e d c b a
e d c b a
e d c b a
e d c b a
e d c b a
e d c b a
e d c b
e d c b
e d c b
a
a
a
a
Flow-Chart of Equilibrium
Poisson Equation Solver
Initialize parameters:
-Mesh size
-Discretization coefficients
-Doping density
-Potential based on charge neutrality
Solve for the updated potential
given the forcing function using LU decomposition
Update:
- Central coefficient of the linearized Poisson Equation
- Forcing function
Test maximum
absolute error update
Equilibrium solver
V
A
= V
A
+AV
Calculate coefficients for:
- Electron continuity equation
- Hole continuity equation
- Update generation recombination rate
Solve electron continuity equation using LU decomposition
Solve hole continuity equation using LU decomposition
Update:
- Central coefficient of the linearized Poisson Equation
- Forcing function
Solve for the updated potential
given the forcing function using LU decomposition
Test maximum
absolute error update
Maximum voltage exceeded?
Calculate current
STOP
yes
no
> tolerance
< tolerance
Non-Equilibrium solver
> tolerance
< tolerance
AV is a fraction of the
thermal voltage V
T
Discretization of the Continuity
Equation
Sharfetter-Gummel Discretization Scheme
The discretization of the continuity equation in conservative
form requires the knowledge of the current densities
on the mid-points of the mesh lines connecting neighboring
grid nodes. Since solutions are available only on the grid
nodes, interpolation schemes are needed to determine the
solutions.
There are two schemes that one can use:
(a)Linearized scheme: V, n, p, and D vary linearly
between neighboring mesh points
(b) Scharfetter-Gummel scheme: electron and hole
densities follow exponential variation between mesh
points
p eD E x ep x J
n eD E x en x J
p p p
n n n
V =
V + =

) ( ) (
) ( ) (
Within the linearized scheme, one has that
This scheme can lead to substantial errors in regions of high
electric fields and highly doped devices.
2 / 1 2 / 1
1
2 / 1 2 / 1 2 / 1 + +
+
+ + +
V +

=
i i
i
i i
i i i
n eD
a
V V
en J
2
1 i i
n n +
+
i
i i
a
n n
+1
(

=
+ + +
+ + +
+ +
i
i
i
i i i
i
i
i
i
i i i
i i
a
eD
a
V V e
n
a
eD
a
V V e
n J
2 / 1 1 2 / 1
2 / 1 1 2 / 1
1 2 / 1
2
2

(a) Linearized Scheme


(b) Sharfetter-Gummel Scheme
One solves the electron current density equation
for n(V), subject to the boundary conditions
The solution of this first-order differential equation leads to
x
V
V
n
eD
a
V V
en
x
n
eD
a
V V
en J
i
i
i i
i
i
i
i i
i i
c
c
c
c
+

=
c
c
+

=
+
+
+
+
+
+ +
2 / 1
1
2 / 1
2 / 1
1
2 / 1 2 / 1


1 1
) ( and ) (
+ +
= =
i i i i
n V n n V n
| |
(

|
.
|

\
|

|
.
|

\
|

=

= + =
+ +
+
+
+

+
+
Vt
V V
B n
Vt
V V
B n
a
eD
J
e
e
V g V g n V g n V n
i i
i
i i
i
i
i
i
Vt V V
Vt V V
i i
i i
i
1 1
1
2 / 1
2 / 1
/ ) (
/ ) (
1
1
1
) ( ), ( ) ( 1 ) (
1

1
) (

=
x
e
x
x B
is the Bernouli function
Solution of the Coupled DD
Equations
There are two schemes that are used in solving the
coupled set of equations which comprises the Drift-
Diffusion model:
Gummels method
Newtons method
Gummels relaxation method, which solves the equations with
the decoupled procedure, is used in the case of weak coupling:
Low current densities (leakage currents, subthreshold regime),
where the concentration dependent diffusion term in the current
continuity equation is dominant
The electric field strength is lower than the avalanche threshold,
so that the generation term is independent of VV
The mobility is nearly independent of E
The computational cost of the Gummels iteration is one matrix
solution for each carrier type plus one iterative solution for the
linearization of the Poisson Equation
Gummels Method
Gummels Method (contd)
The solution strategy when using Gummels relaxation
scheme is the following one:
Find the equilibrium solution of the linearized Poisson
equation
After the solution in equilibrium is obtained, the applied
voltage is increased in steps AVs V
T
Now the scaled Poisson equation becomes:
( ) ( ) | |
( ) ( )
(


+ +
= o +
o
i
D A i
i
n
N N
V V
N
n
x d
V d
V V V
N
n
x d
V d
exp exp
exp exp
2
2
2
2
( ) ( )
(


+ | | =
i
D A
p n
i
n
N N
V V
N
n
x d
V d
exp ) exp( exp ) exp(
2
2
Gummels Method (contd)
The 1D discretized electron current continuity equation (as
long as Einsteins relations are valid) is:
For holes, one can obtain analogous equations by
substituting:
( ) ( ) | |
( ) ( ) | | ( )( ) 0
2
1
1
1 1 1 1
2 1
1 1 1
2 1
= + +

+ + +
+
i
i i i i i i i i
i
i
i i i i i i
i
i
R G a a V V B n V V B n
a
D
V V B n V V B n
a
D
/
/
p n V V ,
The decoupled iteration scheme goes as follows:
(1) Solve the Poisson equation with a guess for the quasi-
Fermi levels (use the applied voltage as initial guess)
(2) The potential is used to update the Bernouli functions
(3) The above equations are solved to provide an update for
the quasi-Fermi levels, that enter into the Poisson
equation
Gummels Method (contd)
The criterion for convergence is:
In the case of strong coupling, one can use the extended
Gummels scheme
|
|
.
|

\
|
(

o
+ o
(

o
=
|
.
|

\
|
(

o
+ o
(

o
=
o + =
+
+
+
k
T
T
k k
k
T
T
k k
k k
p
p
V V
V
V
p p
n
n
V V
V
V
n n
V V V
ln exp
ln exp
1
1
1
1
1
|
|
.
|

\
|
|
|
.
|

\
|

+ +
+
1 1
1
k
k
T
k
k
T
k k
p
p
V
n
n
V V V ln max , ln max , max
Gummels Method (contd)
initial guess
of the solution
solve
Poissons eq.
Solve electron eq.
Solve hole eq.
n
converged?
converged?
n
y
y
initial guess
of the solution
solve
Poissons eq.
Solve electron eq.
Solve hole eq.
n
converged?
converged?
n
y
y
initial guess
of the solution
Solve Poissons eq.
Electron eq.
Hole eq.
Update
generation rate
n
converged?
converged?
n
y
y
initial guess
of the solution
Solve Poissons eq.
Electron eq.
Hole eq.
Update
generation rate
n
converged?
converged?
n
y
y
Original Gummels scheme Modified Gummels scheme
The three equations that constitute the DD model, written in
residual form are:
Starting from an initial guess, the corrections are calculated by
solving:
0 ) , , ( 0 ) , , ( 0 ) , , ( = = = p n v F p n v F p n v F
p n v

(
(
(

=
(
(
(

A
A
A

(
(
(
(
(
(
(

c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
p
n
v
p p p
n n n
v v v
F
F
F
p
n
v
p
F
n
F
v
F
p
F
n
F
v
F
p
F
n
F
v
F
k k k
k k k
k k k
p p p
n n n
V V V
A + =
A + =
A + =
+
+
+
1
1
1
Newtons Method
Newtons Method (contd)
(
(

A
A
A

(
(
(
(
(
(

c
c
c
c
c
c

(
(

=
(
(

A
A
A

(
(
(
(
(
(

c
c
c
c
c
c
c
c
c
c
c
c
p
n
V
n
F
p
F
n
F
F
F
F
p
n
V
p
F
n
F
V
F
n
F
V
F
V
F
n
v v
p
n
v
p p p
n n
v
0 0 0
0 0
0
0
0 0
The method can be simplified by the following iterative
scheme:
1 1 1
1 1
1
+ + +
+ +
+
A
c
c
A
c
c
= A
c
c
A
c
c
A
c
c
= A
c
c
A
c
c
A
c
c
= A
c
c
k
p
k
p
p
k
p
k k
n
n
k
v
k
v
k
v
v
k
v
n
n
F
V
V
F
F p
p
F
p
p
Fn
V
V
F
F n
V
F
p
p
F
n
n
F
F V
V
F
k+1 k
Flow-Chart of 1D Drift-
Diffusion Simulator
Initialize parameters:
-Mesh size
-Discretization coefficients
-Doping density
-Potential based on charge neutrality
Solve for the updated potential
given the forcing function using LU decomposition
Update:
- Central coefficient of the linearized Poisson Equation
- Forcing function
Test maximum
absolute error update
Equilibrium solver
V
A
= V
A
+AV
Calculate coefficients for:
- Electron continuity equation
- Hole continuity equation
- Update generation recombination rate
Solve electron continuity equation using LU decomposition
Solve hole continuity equation using LU decomposition
Update:
- Central coefficient of the linearized Poisson Equation
- Forcing function
Solve for the updated potential
given the forcing function using LU decomposition
Test maximum
absolute error update
Maximum voltage exceeded?
Calculate current
STOP
yes
no
> tolerance
< tolerance
Non-Equilibrium solver
> tolerance
< tolerance
AV is a fraction of the
thermal voltage V
T
Hydrodynamic Modeling
In small devices there exists non-stationary
transport and carriers are moving through the
device with velocity larger than the saturation
velocity
In Si devices non-stationary transport occurs
because of the different order of magnitude of the
carrier momentum and energy relaxation times
In GaAs devices velocity overshoot occurs due to
intervalley transfer
T. Grasser (ed.): "Advanced Device Modeling and Simulation, World Scientific Publishing
Co., 2003, ISBN: 9-812-38607-6 M.
M. Lundstrom, Fundamentals of Carrier Transport, 1990.
Velocity Overshoot in Silicon
-5x10
6
0
5x10
6
1x10
7
1.5x10
7
2x10
7
2.5x10
7
0 0.5 1 1.5 2 2.5 3 3.5 4
1 kV/cm
5 kV/cm
10 kV/cm
50 kV/cm
time [ps]
D
r
i
f
t

v
e
l
o
c
i
t
y


[
c
m
/
s
]
0
0.05
0.1
0.15
0.2
0.25
0 0.5 1 1.5 2 2.5 3 3.5 4
1 kV/cm
5 kV/cm
10 kV/cm
50 kV/cm
E
n
e
r
g
y


[
e
V
]
time [ps]
Scattering mechanisms:
Acoustic deformation potential scattering
Zero-order intervalley scattering (f and g-
phonons)
First-order intervalley scattering (f and g-
phonons)
g
f
k
z
k
x
k
y
g
f
g
ff
k
z
k
x
k
y
X. He, MS Thesis, ASU, 2000.
How is the Velocity Overshoot Accounted
For?
In Hydrodynamic/Energy balance modeling
the velocity overshoot effect is accounted for
through the addition of an energy conservation
equation in addition to:
Particle Conservation (Continuity Equation)
Momentum (mass) Conservation Equation
Hydrodynamic Model due to
Blotakjer
Constitutive Equations: Poisson +
More convenient set of balance equations is in terms of n, v
d
and w:
( )
( )
( )
coll
d
d
B
d d
coll
d
d d
d d
coll
d
t
w
e
v m
w
k
n
n
w
t
w
t m
e
v nm nw
nm
m
m t
t
n
n
t
n
|
.
|

\
|
c
c
+
(
(

|
|
.
|

\
|

|
|
.
|

\
|
V
k
V V =
c
c
|
.
|

\
|
c
c
+
|
.
|

\
|
V V =
c
c
|
.
|

\
|
c
c
+ V =
c
c
) (
2
*
3
2
) (
*
*
2
1
* 3
2
) * (
*
) (
2
2
v E
v v
v
E

v
v v
v
Closure
To have a closed set of equations, one either:
(a) ignores the heat flux altogether
(b) uses a simple recipe for the calculation of the heat flux:
) ( * 2
5
,
2
w v m
nT k
T n
B
= k V k = q
Substituting T with the density of the carrier energy, the
momentum and energy balance equations become:
( )
( )
coll
d
d
B
d d
coll
d
d d d
d
t
nw
en
v m w
k
n w n
t
nw
t
n
en v nm nw n
t
n
|
.
|

\
|
c
c
+
(

|
.
|

\
|

|
|
.
|

\
|
V
k
V + V =
c
c
|
.
|

\
|
c
c
+
|
.
|

\
|
V V =
c
c
) (
*
2
1
3
2
) (
*
2
1
3
2
) (
2
2
v E
v v
p
E p v
p
Momentum Relaxation Rate
The momentum rate is determined by a steady-state MC
calculation in a bulk semiconductor under a uniform bias
electric field, for which:
d
p
d p
coll
d d
v m
eE
w
w
m
e
t m
e
t
*
) (
0 ) (
* *
= v
= v =
|
.
|

\
|
c
c
+ =
c
c
v
E
v
E
v
K. Tomizawa, Numerical Simulation Of
Submicron Semiconductor Devices.
Energy Relaxation Rate
The emsemble energy relaxation rate is also determined by a
steady-state MC calculation in a bulk semiconductor under a
uniform bias electric field, for which:
0
0
) (
0 ) (
w w
e
w
w w e
t
w
e
t
w
d
w
w d
coll
d

= v
= v =
|
.
|

\
|
c
c
+ =
c
c
v E
v E v E
K. Tomizawa, Numerical Simulation Of
Submicron Semiconductor Devices.

You might also like