ece 595, section 10 numerical simulations lecture 19: fem for …pbermel/ece595/lectures/ece... ·...

ECE 595, Section 10

Numerical Simulations

Lecture 19: FEM for Electronic

Transport

Prof. Peter Bermel

February 22, 2013

Outline

• Recap from Wednesday

• Physics-based device modeling

• Electronic transport theory

• FEM electronic transport model

• Numerical results

• Error Analysis

2/22/2013 ECE 595, Prof. Bermel

Recap from Wednesday

• Thermal transfer overview

– Convection

– Conduction

– Radiative transfer

• FEM Modeling Approach

• Numerical Results

• Error Evaluation

2/22/2013 ECE 595, Prof. Bermel

Physics-Based Device Modeling

2/22/2013

D. Vasileska and S.M. Goodnick, Computational Electronics, published by Morgan &

Claypool , 2006.

ECE 595, Prof. Bermel

Electronic Transport Theory• Will assume electronic bandstructures known,

and take a semiclassical approach

• Electrostatics modeled via Poissson’s equation:

• Charge conservation is required:

2/22/2013 ECE 595, Prof. Bermel

( )D AV p n N Nε + −∇ ⋅ ∇ = − − + −

1

1

J

J

n n

p p

nU

t q

pU

t q

∂= ∇ ⋅ +

∂

∂= − ∇ ⋅ +

∂

S. Selberherr: "Analysis and Simulation of

Semiconductor Devices“, Springer, 1984.

Electronic Transport Theory

• Both p-type and n-type currents given by a

sum of two terms:

– Drift term, derived from Ohm’s law

– Diffusion term, derived from Second Law of

Thermodynamics

2/22/2013 ECE 595, Prof. Bermel

( ) ( )

( ) ( )

n n n

p p p

dnJ qn x E x qD

dx

dnJ qp x E x qD

dx

µ

µ

= +

= −

S. Selberherr: "Analysis and Simulation of

Semiconductor Devices“, Springer, 1984.

FEM Electronic Transport Model

• Much like in earlier work, will employ the following strategy:

– Specify problem parameters, including bulk and boundary conditions

– Construct finite-element mesh over spatial domain

– Generate a linear algebra problem

– Solve for key field variables:

2/22/2013 ECE 595, Prof. Bermel

φi (x,y,z,t)

p (x,y,z,t)

n (x,y,z,t)

FEM Electronic Transport Model

2/22/2013 ECE 595, Prof. Bermel

• Regarding the grid set-up, there are several points that need to be

made:

� In critical device regions, where the charge density varies

very rapidly, the mesh spacing has to be smaller than the

extrinsic Debye length determined from the maximum doping concentration in that location of the device

� Cartesian grid is preferred for particle-based simulations

� It is always necessary to minimize the number of node points

to achieve faster convergence

� A regular grid (with small mesh aspect ratios) is needed for

faster convergence

2

maxeN

TkL B

D

ε=

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

Poisson Solver

2/22/2013 ECE 595, Prof. Bermel

• The 1D Poisson equation is of the form:

( )2

2

exp exp( / )

exp exp( / )

D A

F ii i T

B

i Fi i T

B

d ep n N N

dx

E En n n V

k T

E Ep n n V

k T

ϕ

ε

ϕ

ϕ

= − − + −

−= =

−= = −

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

Poisson Solver

• Perturbing potential by δ yields:

2/22/2013 ECE 595, Prof. Bermel

( )

( )

( ) ( )

( )

2/ /

2

/ /

2/ / / /

2

/ /

/

/

T T

T T

T T T T

T T

V Vii

V Vi

V V V Vnewi ii

V V oldi

new old

ende e C n

dx

ene e

en ende e e e C n

dx

ene e

ϕ ϕ

ϕ ϕ

ϕ ϕ ϕ ϕ

ϕ ϕ

ϕ

ε

δε

ϕϕ

ε ε

ϕε

δ ϕ ϕ

−

−

− −

−

= − − + +

+ +

− + = − − + −

− +

= −

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

Poisson Solver

• Renormalized form

2/22/2013 ECE 595, Prof. Bermel

( ) ( )

( ) ( ) ( )

2

2

2

2

new old

new old

dp n C p n

dx

dp n p n C p n

dx

ϕδ

ϕϕ ϕ

δ ϕ ϕ

= − − + + +

− + = − − + − +

= −

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

Poisson Solver

2/22/2013 ECE 595, Prof. Bermel

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

> tolerance

< tolerance

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

Current Discretization

2/22/2013 ECE 595, Prof. Bermel

• 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, µ and D vary linearly between neighboring mesh points

(b) Scharfetter-Gummel scheme: electron and hole densities follow exponential variation between mesh points

peDExepxJ

neDExenxJ

ppp

nnn∇−=

∇+=µµ)()(

)()(

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

Naïve Linearization Scheme

2/22/2013 ECE 595, Prof. Bermel

• Within the linearized scheme, one has that

• This scheme can lead to substantial errors in regions of high electric fields and highly doped devices.

2/12/11

2/12/12/1 +++

+++ ∇+−

−= iii

iiiii neD

a

VVenJ µ

2

1 ii nn ++i

ii

a

nn −+1

+

−−

+

−−=

+++

+++++

i

i

i

iiii

i

i

i

iiiii

a

eD

a

VVen

a

eD

a

VVenJ

2/112/1

2/112/112/1

2

2

µ

µ

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

Scharfetter-Gummel Scheme

2/22/2013 ECE 595, Prof. Bermel

• 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

neD

a

VVen

x

neD

a

VVenJ

ii

iii

ii

iiii

∂

∂

∂

∂+

−−=

∂

∂+

−−=

++

+

++

++

2/11

2/1

2/11

2/12/1

µ

µ

11

)(and)(++

==iiii

nVnnVn

[ ]

−−

−=

−

−=+−=

+++

++

−

−

++

Vt

VVBn

Vt

VVBn

a

eDJ

e

eVgVgnVgnVn

iii

iii

i

ii

VtVV

VtVV

iiii

i

111

2/12/1

/)(

/)(

11

1)(),()(1)(

1

1)(

−=

xe

xxB is the Bernouli function

D. Vasileska, EEE533 Semiconductor Device and Process Simulation Lecture Notes, Arizona State

ADEPT 2.0

• Available on nanoHUB from Prof. Gray’s team:

https://nanohub.org/tools/adeptnpt

2/22/2013 ECE 595, Prof. Bermel

ADEPT 2.0

• Can customize all the calculation details:

2/22/2013 ECE 595, Prof. Bermel

ADEPT 2.0

• Outputs include electrostatic (Poisson) solution:

2/22/2013 ECE 595, Prof. Bermel

ADEPT 2.0

• Energy band diagram

2/22/2013 ECE 595, Prof. Bermel

ADEPT 2.0

• Carrier concentrations:

2/22/2013 ECE 595, Prof. Bermel

ADEPT 2.0

• And finally, realistic I-V curves:

2/22/2013 ECE 595, Prof. Bermel

Next Class

• Is on Monday, Feb. 25

• Next time, we will cover electronic

bandstructures

2/22/2013 ECE 595, Prof. Bermel