Professional Documents
Culture Documents
www.nanohub.org
ECE606: Solid State Devices
L t
Lecture 19: Numerical Solution of Transport Equations
19 N i l S l ti fT tE ti
Muhammad Ashraful Alam
alam@purdue.edu
Alam ECE‐606 S09 1
Outline
1) Basic Transport Equations
Basic Transport Equations
2) Gridding and finite differences
3) Discretizing equations and boundary conditions
4) Conclusion
REF. A primer on semiconductor device simulation
… M. Lundstrom
M L d (
(nanohub.org)
h b )
Alam ECE‐606 S09 2
Equations to solve … analytical/numerical approach
∇ • D = q ( p − n + N D+ − N A− ) Band‐diagram
∂n 1
= ∇ • J N − rN + g N
∂t q
J N = qnμ N E + qDN ∇n
∂p −1 Diffusion approximation,
= ∇ • J P − rP + g P Minority carrier transport,
∂t q
J P = qp μ P E − qDP ∇p Ambipolar transport
Alam ECE‐606 S09 3
1) The Semiconductor Equations
Conservation Laws: Constitutive Relations:
D = κε 0 E = −κε 0∇V
∇•D = ρ
ρ = q( p − n + N D+ − N A− )
( )
∇ • J n − q = ( g N − rN ) J n = nq μn E + qDn∇n
∇•(J p q) = ( g P − rP ) J p = pq μ p E − qD
q p ∇p
g N , P = f (n, p ) etc.
((steady‐state)
y )
Alam ECE‐606 S09 4
1) The Mathematical Problem
The “Semiconductor Equations”
∇•D = ρ 3 coupled, nonlinear,
second order PDE’ss
second order PDE
( )
∇ • J n − q = ( g N − rN ) for the 3 unknowns:
∇•(J p q) = ( g N − rN ) V (r ) n( r ) p(r )
Conservations laws: exact
T
Transport eqs. (drift‐diffusion): approximate
t (d ift diff i ) i t
Alam ECE‐606 S09 5
Outline
1) Basic Transport Equations
2) Gridding and finite differences
3) Making up the matrix and boundary conditions
Making up the matrix and boundary conditions
4) Conclusion
Alam ECE‐606 S09 6
2) The Grid
(ii) “exact” numerical solutions N nodes
3N unknowns
Vi
a
ni
pi
Alam ECE‐606 S09 7
Finite Difference Expression for Derivative
a a df a a df
f ( xo ) = f ( xo + ) − f ( xo + a ) = f ( x 0 + ) +
2 2 dx 2 2 dx
dx
f(x) df fi +1 − fi
=
dx ( xi+1/ 2 ) a
a
Alam ECE‐606 S09 8
The Second Derivative …
df a2 d 2 f
f ( x0 + a ) = f ( x0 ) + a + 2
+ ...
dx x0 = a 2 dx x0 = a
df a2 d 2 f
f ( x0 − a ) = f ( x0 ) − a + 2
− ...
dx x0 = a 2 dx x0 = a
2
d f
f ( x0 + a ) + f ( x0 − a ) − 2 f ( x0 ) = a 2 2
dx x0 = a
d f2
fi −1 − 2 fi + fi +1
2
= 2
dx i a
Alam ECE‐606 S09 9
Outline
1) Basic Transport Equations
2) Gridding and finite differences
3) Discretizing equations and boundary conditions
equations and boundary conditions
4) Conclusion
Alam ECE‐606 S09 10
2) Control Volume
3 k
3 unknowns at each node:
t h d x
(i ‐1) ((i)) (i +1)
Vi ,ni , pi
Need 3 equations
Need 3 equations
at each node
“control volume”
Alam ECE‐606 S09 11
Discretizing Poisson’s Equation
∇ 2V = − ρ / K sε o ∇ • D = ρ D = K s ε 0 E = - K s ε 0 ∇V
V ( i − 1) − V ( i ) + V ( i + 1) q + −
= − ( pi − ni + ND , i − NA , i )
a 2
K sε 0
F (Vi −1 , Vi , Vi +1 , ni , pi ) = 0
V
i
Alam ECE‐606 S09 12
Discretizing Continuity Equations
JnL ⎛ ni −1 + ni ⎞ ⎛ Vi − Vi −1 ⎞ ⎛ ni − ni −1 ⎞
= −⎜ ⎟ ⎜⎜ ⎟⎟ + ⎜ ⎟
kT μn ⎝ 2 ⎠ ⎝ a ( kT / q ) ⎠ ⎝ a ⎠
F (Vi −1 , Vi , ni , ni −1 , pi , pi −1 ) = 0
n
i
Alam ECE‐606 S09 13
Three Discretized Equations
F =0
V
i x
(i ‐1) ((i)) (i +1, j)
F =0
n
i
F =0
i
p
3 unknowns at each node
3 unknowns at each node
N nodes
3N unknowns and 3N equations (nonlinear!)
3N unknowns and 3N equations (nonlinear!)
Alam ECE‐606 S09 14
Numerical Solution – Poisson Equation Only
• Have a system of 3N nonlinear equations to solve
• Recall Poisson’s equation at node (i):
F (Vi −1 , Vi , Vi +1 , ni , pi ) = 0
V
i
n0 p0 = n 2
i
nN +1pN +1 = n
2
i
Dopant density
a
V = VA V =0
Alam ECE‐606 S09 16
Numerical Solution…
Alam ECE‐606 S09 17
3) Uncoupled Numerical Solution
The semiconductor equations are nonlinear!
(but they are linear individually)
(but they are linear individually)
Guess V,n,p
Uncoupled solution procedure Solve Poisson
for new V
for new V
repeat
Solve electron
until
cont for new n
satisfied
Solve hole
cont for new p
Alam ECE‐606 S09 18
Summary
1) Two methods to solve drift‐diffusion equation consistently –
analytical and numerical.
2) Analytical
Analytical solution provides great insight and the solution
solution provides great insight and the solution
methodology is similar to that of Schrodinger equations.
3) N
Numerical solution is more versatile. One begins with a set
i l l i i il O b i ih
of equations and boundary conditions, discretize the
equations on a grid with N nodes to obtain 3N nonlinear
equations in 3N unknowns, and solve the system of
nonlinear equations by iteration.
Alam ECE‐606 S09 19