CONTENTS

Introduction ........................................................................................................... 3
Syllabus: Unit_III Basic Physical Models in Sentaurus Device ........................ 3
Abbreviations: ....................................................................................................... 4
1. Electrostatic Potential and Transport equations: ......................................... 5
1.1 Poisson Equation ...................................................................................... 5
1.2 Sample Solution of Poisson Equation for PN Junction Diode: ................ 6
1.3 Quasi Fermi Potential With Boltzmann Statistics ....................................... 6
1.4 Fermi Statistics ............................................................................................. 7
1.5 How Tcad Solves Poisson Equation? .......................................................... 8
1.6 Carrier Transport Models ............................................................................. 8
1.6 Drift Diffusion Model .............................................................................. 9
1.7 Thermodynamic Model for Current Densities ....................................... 12
1.8 Hydrodynamic Model............................................................................. 14
2 Boundary Conditions..................................................................................... 17
2.1 Unit Normal vector ( 𝒏) ......................................................................... 17
2.2 Three Types of Boundary Conditions .................................................... 17
3. Semiconductor Band Structure .................................................................. 22
3.1 Intrinsic carrier density (ni) .................................................................... 22
3.2 Effect of Eg vs T: ................................................................................... 23
3.3 Effect of Electron Affinity vs T: ............................................................ 23
3.4 Bandgap Narrowing Models: ................................................................. 24
3.4.1 Apparent BGN Models .................................................................... 26
3.4.2 Jain Roulston Model ............................................................................ 28
3.4.3 Schenk BGN ......................................................................................... 28
3.5 Additional Notes on Band structure simulation (Not for Exam) ........... 30
3.5.1 Multivalley Band structure .............................................................. 30
3.5.2 Band gap Widening .......................................................................... 30
3.6 DOS and Effective Mass ........................................................................ 30
4. Mobility Models ........................................................................................ 33
4.1 Introduction ................................................................................................ 33

1
 4.2 Mobility Models ......................................................................................... 34
4.2.1 Mobility Degradation due to Phonon scattering in the Bulk ............... 34
4.2.2 Mobility degradation due to impurity scattering ................................. 36
4.2.3 Carrier-Carrier Scattering Models........................................................ 40
4.2.4 Unified Mobility Model ....................................................................... 43
4.2.5 Interface Mobility Degradation ............................................................ 47
4.2.6 High Field Saturation ........................................................................... 48
5. Incomplete Ionization ................................................................................ 52
6. Recombination ........................................................................................... 53
7. Generation Models: ................................................................................... 56

1.1 Acknowledgment
Simplified lecture notes for M.Tech VLSI student for their preparation and
assignment. We tried to simplify the contents in Sentaurus Device Manual here.

2
 SYLLABUS

INTRODUCTION
Semiconductor device simulation is usually done by solving three fundamental equations to
find three unknowns (electrostatic potential, electron density and hole density):
1. Poisson
2. Electron Continuity
3. Hole Continuity
In this unit, modeling potential calculation (from Poisson), Transport equations (J n and Jp) ,
boundary conditions, intrinsic carrier density (ni = f(Nc, Nv, effective mass)=> Important term
in all recombination models and used to find the minority carrier density in non-degenerate
semiconductor), mobility (term in transport equation), generation (not due to external
illumination or radiation, it is discussed in next unit) and recombination models (G and R term
in continuity equation) will be studied in detail.

3.1 Syllabus: Unit_III Basic Physical Models in Sentaurus Device

Electrostatic potential – Quasi fermi potential with Boltzmann statistics – Fermi
statistics – Carrier transport – Drift Diffusion model – Thermodynamic model –
Hydrodynamic model – Boundary conditions – Ohmic contacts – Schottky contacts –
Resistive contacts – Floating contacts – Thermal boundary condition – Periodic
boundary conditions – Semiconductor band structure – Bandgap models – Effective
mass and effective density of states – Incomplete ionization model – Mobility models
– Mobility due to phonon scattering – Doping dependent mobility – Mobility models –
Carrier carrier scattering models – Mobility degradation at interfaces – High field
saturation – Generation and Recombination – Shockley Read Hall recombination –
Trap assisted recombination – Surface recombination – Avalanche generation – Impact
ionization.

3
3.2 Abbreviations:
Φ Electric Potential or Electrostatic Potential or Potential
BGN/bgn Band Gap Narrowing
D Diffusivity
𝜉 Electric Field
e Electron (neither charge nor density: Just the name)
n,p Electron Density and Hole Density
𝐸 Energy
h Hole (neither charge nor density: Just the name)
𝜇 Mobility
q Unit Charge

4
 ELECTROSTATIC POTENTIAL AND TRANSPORT EQUATIONS:
Electric Potential is the amount of electric potential energy that a unit charge
would have at the specified location [1]. Electric potential energy per electron is
𝐸
called potential (∵ 𝐸 = 𝑞𝜙 => 𝜙 = ). Electric potential energy of a charge is
𝑞
the amount of energy/force experienced by the charge in the force field depending
on its position. (Note: Mind the difference/analogy between electric potential
energy and gravitational potential energy. Here, we use potential to denote
electric potential).

Figure 1 Particle A has more potential energy than B if the above diagram is seen in the perspective of
Electron conduction band energy.

Potential depends on the amount of charge (Eg: heavily doped n-type has more
potential than lightly doped n-type for a PN junction) and in turn affect the
charges. Electrostatic => denotes potential is due to immobile or slowly moving
charges.
Charges in semiconductor devices can be due to
i) Mobile (free e-, free hole, exciton…)
ii) Immobile (Ionized Dopants, Traps and Defects)
How Potential is calculated in Numerical simulators?
It is calculated by solving Poisson equation. Graphical solution of Poisson
is a handy way to calculate potential manually.

1.1Poisson Equation:
∇. (𝜖∇Φ) = 𝜌 = −𝑞(𝑝 − 𝑛 + 𝑁𝐷 − 𝑁𝐴 ) − 𝜌𝑡𝑟𝑎𝑝

Note: Under equilibrium condition: all the voltages in all contacts are zero/same.

Band Diagram can be seen as solution of Poisson Equation since band diagram
can be constructed from potential.

5
4.1 Sample Solution of Poisson Equation for PN Junction Diode:

Figure 2 Simulated electrostatic potential for p-n junction diode. n is heavily doped. Observation: potential
is positive for n-type and negative for p-type and there is more positive potential (nearly +0.5V) than
negative potential (nearly −0.3V). Under applied forward bias, potential reduces and hence probability of
diffusion increases.

4.2 Quasi Fermi Potential With Boltzmann Statistics

Quasi Fermi Potential (or Chemical Potential or Fermi Level) can be used to
calculate n and p values. Considering Maxwell-Boltzmann approximation, it will
be further reduced to
𝐸 ′ −𝐸
− ( 𝐶 𝐹)
𝑘𝑇
𝑛 = 𝑁𝐶 𝑒
𝐸 ′ −𝐸
− ( 𝑉 𝐹)
𝑘𝑇
𝑝 = 𝑁𝑉 𝑒
𝑬′𝑪 = 𝑬𝑪 − 𝒒𝝓
Where 𝑬𝑪 = −𝝌
𝑬′𝑽 = 𝑬′𝑪 − 𝑬𝒈𝒆𝒇𝒇

where, Φ → 𝐸𝑙𝑒𝑐𝑡𝑟𝑜𝑠𝑡𝑎𝑡𝑖𝑐 𝑃𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 𝑢𝑠𝑖𝑛𝑔 𝑃𝑜𝑖𝑠𝑠𝑜𝑛 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛

𝜒 → 𝐸𝑙𝑒𝑐𝑡𝑟𝑜𝑛 𝐴𝑓𝑓𝑖𝑛𝑖𝑡𝑦
The above equation accounts how conduction and valence band energies are
constructed from potential and electron affinity in a numerical simulator.

6
Figure 3 Ec and Ev calculated from previously mentioned equation for the same potential (𝚽) shown in
figure 2 for 𝝌 = 𝟒. 𝟎𝟓 𝒆𝑽 and 𝚽𝒓𝒆𝒇 = 𝟒. 𝟔 𝑽.

Note: Energies are represented in eV and potential is represented in V.

4.3 Fermi Statistics
First simplification from general term lead to Fermi Statistics. Carrier densities
calculated using Fermi is more accurate than Boltzmann statistics. It accounts
Pauli blocking which is very important phenomena for heavily doped
semiconductor. It rectifies the overestimation of minority carrier density (and
hence pn product) by Boltzmann statistics for especially doping concentrations >
1019 𝑐𝑚−3 . Pauli blocking is considered in Fermi statistics.

7
4.4 How Tcad Solves Poisson Equation?
Initial guess for Electrostatic Potential and Quasi Fermi Potential is made
for each doping wells. Doping well => dividing the semiconductor device into
number of regions based on doping and contacted region.
4.5 Carrier Transport Models
The carrier transport model describes the transport behavior of carriers
(electrons/holes) in semiconductor devices. When a semiconductor is in non-
equilibrium (i.e) by applying voltage, temperature gradient and under
illumination, the carriers start to move from one region into other inside a material
or external to the circuit which leads to path of raising the current. In TCAD, the
sentaurus Device supports the several transport models for semiconductor. They
all can be written in the form of continuity equations, which describe charge
conservation. The continuity equations in TCAD is given in the form of
n p
.J n  qRnet , n  q  .J p  qRnet , p  q
t t
Where:
■ Rnet,n and Rnet,p are the electron and hole net recombination rate, respectively.
■ Jn is the electron current density.
■Jp is the hole current density.
■ n and p are the electron and hole density, respectively.
In general, the continuity equations are given in the form of:
n 1 Jn
 G  R
p 1tJ p q x
 G R
t q x

8
4.6 Drift Diffusion Model
The drift-diffusion model is the default carrier transport model in Sentaurus Device. For the drift-
diffusion equation, the current densities of electrons and holes are obtained which is given in the form of
For current density of holes: For current density of electrons:
JP=JDrift+JDiffusion JN=JDrift+JDiffusion
p n
J p  q p P  qD p Jn  q n P  qDn
x x

Drift current is due to the electric field  and diffusion is due to the p, n .
In TCAD, The drift-diffusion model is the default carrier transport model in Sentaurus Device. For the
drift-diffusion model, the current densities for electrons and holes are given by:

J n  n (nEc  1.5nkT ln mn )  Dn (n  n ln  n )

J p   p ( pEc  1.5 pkT ln m p )  D p (p  p ln  p )

The first term takes into account the contribution due to the spatial variations of the electrostatic
potential, the electron affinity, and the band gap. The remaining terms take into account the contribution
due to the gradient of concentration, and the spatial variation of the effective masses mn and mp.
For Fermi statistics  n and  p are given by euations

p p
p  exp( p ) n  exp(n )
NV ND

For Boltzmann statistics  n =  p =1.

The diffusivity terms Dn and Dp in diffusion equation can be obtained by using the Einstein relationship

which is given in the form of:

D n  kT n D p  kTp

When the Einstein relation holds, the current equations of electrons and holes can be simplified to:

J n  -nq n  n J p  -pqp p

Where  n and  p are the electron and hole quasi-Fermi potentials, respectively

https://personal.utdallas.edu/~frensley/technical/hetphys/node13.html#eqnNCde
f

9
4.7 Derivation of Drift-Diffusion for heterojunction devices:
Under thermal equilibrium:
𝑑𝐸𝐹
𝐽𝑑𝑖𝑓𝑓𝑢𝑠𝑖𝑜𝑛 + 𝐽𝑑𝑟𝑖𝑓𝑡 = 0; =0
𝑑𝑥
Let’s start from electron concentration:
(𝐸 ′ −𝐸 )
− 𝐶 𝐹
𝑛= 𝑁𝐶 𝑒 𝑘𝑇

𝑑𝑛
Let’s find :
𝑑𝑥
′ ′
𝑑𝑛 𝑑 (𝐸𝐶 −𝐸𝐹 ) (𝐸𝐶 −𝐸𝐹 ) 𝑑𝑁
𝐶
= 𝑁𝐶 (𝑒 − 𝑘𝑇 ) − 𝑒 − 𝑘𝑇
𝑑𝑥 𝑑𝑥 𝑑𝑥
′ ′
𝑑𝑛 (𝐸 −𝐸 ) 𝑑
− 𝐶 𝐹 (𝐸𝐶′ − 𝐸𝐹 ) − 𝐶 𝐹 𝑑𝑁𝐶
(𝐸 −𝐸 )
= 𝑁𝐶 (𝑒 𝑘𝑇 ) (− )−𝑒 𝑘𝑇
𝑑𝑥 𝑑𝑥 𝑘𝑇 𝑑𝑥
Let’s find the exponentials and replace with “n”:
𝑑𝑛 1 𝑑 1 𝑑𝐸𝐹 𝑛 𝑑𝑁𝐶
=𝑛 (−𝐸𝐶′ ) + −
𝑑𝑥 𝑘𝑇 𝑑𝑥 𝑘𝑇 𝑑𝑥 𝑁𝐶 𝑑𝑥
𝑑𝑛 1 𝑑 𝐸𝐶′ 1 𝑑𝑁𝐶
= −𝑛 ( + )
𝑑𝑥 𝑘𝑇 𝑑𝑥 𝑁𝐶 𝑑𝑥
3
2𝜋𝑚𝑛∗ 𝑘𝑇 3/2 2𝜋𝑘𝑇 2 3
∗2
𝑁𝐶 = 2( ) = 2( ) 𝑚𝑛
ℎ2 ℎ2
3 3
𝑑𝑁𝐶 2𝜋𝑘𝑇 2 3 ∗ 3−1 𝑑𝑚𝑛∗ 2𝜋𝑘𝑇 2 3 ∗ 1 𝑑𝑚𝑛∗
= 2( 2 ) 𝑚 2 = 2 ( 2 ) 𝑚𝑛 2
𝑑𝑥 ℎ 2 𝑛 𝑑𝑥 ℎ 2 𝑑𝑥
3
2𝜋𝑘𝑇 1 23
∗2
1 𝑑𝑁𝐶 2 ( ℎ2 ) 2 𝑚𝑛 𝑑𝑚𝑛∗ 3 ∗ 1−3 3 1 𝑑𝑚𝑛∗
= 3 = (𝑚𝑛 )2 2 =
𝑁𝐶 𝑑𝑥 3 𝑑𝑥 2 2 𝑚𝑛∗ 𝑑𝑥
2𝜋𝑘𝑇 2 ∗ 2
2 ( 2 ) 𝑚𝑛
ℎ
From derivative of ln 𝛼(𝑥)
𝑑 1 𝑑𝛼
(ln 𝛼) =
𝑑𝑥 𝛼 𝑑𝑥
1 𝑑𝑚𝑛∗ 𝑑
= (ln 𝑚𝑛∗ )
𝑚𝑛∗ 𝑑𝑥 𝑑𝑥

10
 1 𝑑𝑁𝐶 3 𝑑
= (ln 𝑚𝑛∗ )
𝑁𝐶 𝑑𝑥 2 𝑑𝑥

𝑑𝑛 1 𝑑 𝐸𝐶′ 3 𝑑
= −𝑛 ( + (ln 𝑚𝑛∗ ))
𝑑𝑥 𝑘𝑇 𝑑𝑥 2 𝑑𝑥
𝐽𝑑𝑖𝑓𝑓 = −𝐽𝑑𝑟𝑖𝑓𝑡
𝑑𝑛 1 𝑑 𝐸𝐶′ 3 𝑑
𝐽𝑑𝑖𝑓𝑓 = 𝑞 𝐷𝑛 = −𝑞 𝐷𝑛 𝑛 ( + 𝑘𝑇 (ln 𝑚𝑛∗ )) = −𝐽𝑑𝑟𝑖𝑓𝑡
𝑑𝑥 𝑘𝑇 𝑑𝑥 2 𝑑𝑥
𝑑𝑛 1 𝑑 𝐸𝐶′ 3 𝑑
𝐽𝑑𝑖𝑓𝑓 + 𝐽𝑑𝑟𝑖𝑓𝑡 = 𝑞 𝐷𝑛 + 𝑞𝐷𝑛 𝑛 ( + 𝑘𝑇 (ln 𝑚𝑛∗ )) = 𝐽𝑛
𝑑𝑥 𝑘𝑇 𝑑𝑥 2 𝑑𝑥
= 𝑇𝑜𝑡𝑎𝑙 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝐷𝑒𝑛𝑠𝑖𝑡𝑦
𝑑𝑛 1 𝑑 𝐸𝐶′ 3 𝑑
𝐽𝑛 = 𝑞 𝐷𝑛 + 𝑞𝐷𝑛 𝑛 ( + 𝑘𝑇 (ln 𝑚𝑛∗ ))
𝑑𝑥 𝑘𝑇 𝑑𝑥 2 𝑑𝑥
𝐷𝑛 𝑘𝑇 𝑞𝐷𝑛
Since = → = 𝜇𝑛
𝜇𝑛 𝑞 𝑘𝑇

𝑑𝑛 𝑑(𝐸𝑐 − 𝑞𝜙) 3 𝑑
𝐽𝑛 = 𝑞𝐷𝑛 + 𝜇𝑛 𝑛 ( + 𝑘𝑇 (ln 𝑚𝑛∗ ))
𝑑𝑥 𝑑𝑥 2 𝑑𝑥
𝑑𝑛 𝑑𝐸𝑐 𝑑𝜙 3 𝑑
𝐽𝑛 = 𝑞𝐷𝑛 + 𝜇𝑛 𝑛 − 𝑞 𝜇𝑛 𝑛 + 𝑘𝑇 𝜇𝑛 𝑛 (ln 𝑚𝑛∗ )
𝑑𝑥 𝑑𝑥 𝑑𝑥 2 𝑑𝑥
Thus the total current density depends on concentration gradient (diffusion),
conduction band gradient (drift) and variation in effective mass (difference in
material) and potential gradient as well.
Similarly for hole current density (𝑞 → −𝑞; 𝑛 → 𝑝; 𝐸𝑐 → 𝐸𝑣 ; 𝑚𝑛 → 𝑚𝑝 ):
𝑑𝑝 𝑑𝐸𝑉 𝑑𝜙 3 𝑑
𝐽𝑝 = −𝑞𝐷𝑝 + 𝜇𝑝 𝑝 + 𝑞 𝜇𝑝 𝑝 + 𝑘𝑇 𝜇𝑝 𝑝 (ln 𝑚𝑝∗ )
𝑑𝑥 𝑑𝑥 𝑑𝑥 2 𝑑𝑥
)

11
4.7.1 Drift Diffusion Models used in Sentaurus TCAD

4.8 Thermodynamic Model for Current Densities

The drift-diffusion carrier transport model consider all the things, which
are all contributes to carrier transport like electrostatic potential, the electron
affinity, gradient of concentration, effective masses and the band gap. But in drift-
diffusion model, temperature effect is not considered, hence the temperature
equations are used to calculate effect of temperature in current densities. There
are two types of models used to account the temperature effect.
1. Thermodynamic model
12
 2. Hydrodynamic model
Sentaurus Device can compute up to three different temperatures: lattice
temperature, electron temperature, and hole temperature. Lattice temperature
describes self-heating of devices. The electron and hole temperatures are required
to model non-equilibrium effects in deep submicron devices (in particular,
velocity overshoot and impact ionization). The following options are available:
■The lattice temperature can be computed from the total dissipated heat, assuming
the temperature is constant throughout the device.
■The lattice temperature can be computed nonuniformly using the default model,
or the thermodynamic model, or the hydrodynamic model.
■ One or both carrier temperatures can be computed using the hydrodynamic
model.
1. Thermodynamic model
The model differs from drift-diffusion when the lattice temperature equation is
solved. However, it is possible to solve the lattice temperature equation even
when using the drift-diffusion model. The equation is generalized to include the
temperature gradient in drift-diffusion model equation.

⃗⃗⃗
𝐽𝑝 = −nqµ𝑝 (∇𝜑 + 𝑃𝑛 ∇𝑇)
𝐽⃗⃗⃗𝑛 = −nqµ𝑛 (∇𝜑 + 𝑃𝑝 ∇𝑇)

Where 𝐽𝑛 and 𝐽𝑝 are the absolute thermoelectric powers and T is the lattice
temperature. The thermodynamic models is useful in high carrier density device
simulation like HEMT, HBT devices. (Lattice temperature gradient =>
Thermodynamic and Carrier Temp gradient => Hydrodynamic)

13
1.6 Hydrodynamic Model (Motion of compressible fluids – here gas
motions where e- an holes are considered as gases)

https://ieeexplore.ieee.org/document/1301241
The classical hydrodynamic model has become a standard industrial simulation
tool that incorporates important “hot electron” phenomena in submicron
semiconductor devices. Hot electron effects are missing in the simpler drift-
diffusion model, which assumes the electron gas is always at ambient
temperature. The hydrodynamic description should be valid for devices with
active regions greater than 0.05 microns
Hydrodynamic
Accounts for energy transport of the carriers. Suitable for devices with small
active regions.
 To get realistic device simulation results several different physical models
are needed and some of them have special needs concerning grid density.
The first thing to decide is which transport mechanism to choose. Thus, for
Note on Hydrodynamic Model
 Deep submicron devices (100 nm to 250nm size) simulation doesn’t
match experimental data using drift diffusion approach. Conventional
Drift Diffusion model cannot estimate:
a. Velocity overshoot effect. (Electron velocity is more than the
saturation velocity for smaller channel length, because of
electron temperature is higher than the lattice temperature.)
b. Overestimates Impact Ionization.
 Hence, energy based transport (you can see carrier temperature terms
in current density equation, Tn and Tp which denote energy transport)
is proposed and found good matching with experimental data after well
calibration (since approximations are used to derive those equations).
 To find lattice temperature (T), Tn and Tp, three additional equations
(called as energy balance equations which requires calibration) have to
be solved.

14
 this relatively long channel device it is sufficient with the basic drift-
diffusion equations but if the gate length is below 60 nm then it is often
suggested to use hydrodynamic (energy-balance) simulations.

Reason for velocity overshoot:

“It has been shown that an electric field step causes the electron velocity to
overshoot the value that corresponds to the higher field for a period shorter
than the energy relaxation time [6], [7] (the time needed by the electron to
once again reach equilibrium with the lattice [8]). Therefore, as the
longitudinal electric field increases, the electron gas starts to be in
disequilibrium with the lattice [9]. There is an insufficient number of phonon-
scattering events experienced by the electron during its flight, with the result
that electrons can be accelerated to velocities higher than the saturation
velocity, thus approaching ballistic transport conditions”

In the hydrodynamic case, current densities are defined as a sum of four

contributions:
4.9 Derivation of Hydrodynamic model from DD equation):

[Ref: Stratton PRB 1962]

15
The hydrodynamic model couples the basic semiconductor equations (Poisson
equation and continuity equations) with the additional energy balance equations
for electrons, holes and the lattice.

The electron current in a semiconductor at uniform lattice temperature T0, with a

nonuniform electric field distribution (e.g., a barrier layer), consists of terms
arising from conduction, diffusion, and thermal diffusion. The first two terms
involve the mobility and diffusion coefficient which are functions of the electron
temperature T or, more generally, depend on certain averages over the
nonequilibrium, field-dependent electron energy distribution function. The third
term is due to the electron temperature gradient and is analogous to conventional
thermal diffusion of a gas in a temperature gradient. In conventional theory,
which neglects electron heating or cooling, the mobility and diffusion coefficient
are material constants and thermal diffusion is absent. Contrary to the case of
uniform fields, T is not a unique function of the local field; it also depends on the
current and can only be determined by a simultaneous solution of the equations
for current flow and conservation of energy with boundary conditions for a
particular structure. As an example, a one carrier metal-semiconductor contact
rectifer has been analyzed in detail including a discussion of the Peltier effect. In
the barrier region T is greater than T0 (i.e., hot electrons) for a reverse bias but
less than T0 (i.e., cold electrons) for a forward bias. Computer solutions have been
obtained for a Schottky barrier and electron scattering due to acoustic phonons
only. (Stratton, PRB, 1962)
http://journals.aps.org/pr/abstract/10.1103/PhysRev.126.2002
Ref:
1. https://en.wikipedia.org/wiki/Electric_potential
2. +Sentaurus Device User Guide June 2015 version Pg: 217
3. Pierret, “Advanced Semiconductor Fundamentals” Pg:112

16
21-Sept-2015
2 BOUNDARY CONDITIONS
̂)
2.1Unit Normal vector ( 𝒏

Ref: http://mathworld.wolfram.com/NormalVector.html

Ref: https://en.wikipedia.org/wiki/Dot_product

̂ => Projected current component of the normal (or perpendicular) to the surface or
So 𝑱𝒏 . 𝒏
simply current coming out of the surface

2.2Three Types of Boundary Conditions

There are usually three types of boundary conditions for Φ, 𝑛, 𝑝 , 𝑇𝑛 , 𝑇𝑝 𝑎𝑛𝑑 𝑇𝐿 :
1. Hom Neumann (or Neumann) Boundary condition:
 Derivative of flux out of the domain is specified. For eg: mathematically
̂
represented as 𝑱𝒏 . 𝒏
 For non-contacting boundaries (Like side boundaries as shown below in
̂ = 0 Where normal
the figure), zero flux is assumed using Jn. 𝒏
component of current across the boundary is zero => 𝛁𝒏 = 𝟎
𝑑𝜙 𝑑
 . 𝑛̂(𝑥) = 𝑑𝑛̂ (𝜙)
𝑑𝑥

17
Figure 4 Boundary conditions at different part of an imaginary device. Usually side boundary
condition is assumed Neumann with zero flux and contacts are assumed Dirichlet with the voltage
applied as the potential value.
2. Dirichlet Boundary condition
 Fixed value for the solution variable (n, p, Φ, Tn , Tp and TL)
 For eg: Φ = 3𝑉, 𝑛 = 1010 𝑐𝑚−3 𝑎𝑡 𝑡ℎ𝑒 𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦
 For Gas phase diffusion, Dirichlet boundary condition is used to set the
Furnace concentration value in sprocess simulation.

3. Mixed Boundary Condition (Combination of both Neumann and Dirichlet

boundary condition)

4.10 Boundary Conditions in Ohmic Contacts:

Unknown variables are given directly (hence Dirichlet condition):

18
𝜙𝐹 equals the applied voltage if it is not a resistive contact.
If electron or hole recombination velocity is specified,

Assuming the flux of carriers which is Neumann condition.

(No need to mug-up those equations, just look at the form of boundary conditions)
4.11 Boundary conditions in Schottky contacts

Potential is applied using Dirichlet boundary condition and carrier concentration

is applied using Neumann condition.
(Mixed boundary condition).
(No need to mug-up those equations, just look at the form of boundary conditions)

4.12 Boundary conditions in Floating Contacts:

19
[Source: ScienceDirect]

How to give boundary condition for floating gate?

Total Charge
𝑑𝜉
𝜖 =𝜌
𝑑𝑥

𝜖 𝜉 = ∫ 𝜌 𝑑𝑥 = 𝑄

𝑑𝜙
𝜖 (− )=𝑄
𝑑𝑥
Along the boundary
dϕ
𝜖 𝑛̂. (− )=𝑄
dx

4.13 Thermal Boundary Condition

4.14 Periodic Boundary Condition

The use of periodic boundary conditions (PBCs) can be helpful for simulations of periodic
20
device structures, for example, lines or arrays of cells. Instead of building a
periodic simulation domain, you can supply periodic boundary conditions at the
lower and upper boundary planes of a single cell. The solution is continuous
across this interface, and currents (or fluxes in the general case) can flow from
one side to the other.

4.15 Ref:
1. http://www.iue.tuwien.ac.at/phd/palankovski/node30.html
2. http://www.nextnano.com/nextnano3/input_parser/keywords/poisson-
boundary-conditions.htm
3. https://en.wikipedia.org/wiki/Neumann_boundary_condition

21
24-Sept-2015

4. SEMICONDUCTOR BAND STRUCTURE

Simplified band diagram is simplified from E-k diagram which is obtained by
solving Schrodinger equation. This simplified band diagram is sufficient for most
of the device simulation. (Indirect band gap information which is one of property
of E-k diagram, might be hidden in absorption property of the material and also
in Radiative Recombination coefficient which will be discussed soon). Simplified
band diagram can be obtained manually by solving Poisson equation (𝐶ℎ𝑎𝑟𝑔𝑒 →
ξ → Φ → 𝐸𝑛𝑒𝑟𝑔𝑦). Band diagram what we usually consider is single electron
energy band diagram (because 𝐸𝑛𝑒𝑟𝑔𝑦, 𝐸 = −𝐪 𝜙, 𝑤ℎ𝑒𝑟𝑒 − 𝒒 →
𝑠𝑖𝑛𝑔𝑙𝑒 𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑛)
Simplified semiconductor band structure is quantified into 4 terms:
i) Conduction Band Energy OR i) Electron Affinity
ii) Valence Band Energy ii) Band Gap
iii) DOS effective mass for electron iii) Effective DOS for Conduction band
iv) DOS effective mass for hole iv) Effective DOS for Valence band
E-k Details in Simplified Band diagram:
a) Curvature of conduction and valence band is hidden in effective mass
b) Entire DOS is simplified to Band Edge or Effective DOS.
This band details enter the calculation of current through few terms:
i) Intrinsic Carrier Density (ni -> All the recombination rates (part of
continuity equation) depends on ni2 and minority carrier
concentration depends on ni2)
ii) Electric field (part of drift term in continuity equation) is derived
from Ec and Ev for electron and hole respectively which we already
discussed in Transport equations.
4.1Intrinsic carrier density (ni)
ni depends on effective DOS and Band gap.
𝑛𝑖 = 𝑓(𝑁𝑐 , 𝑁𝑣 , 𝐸𝑔 ) where 𝐸𝑔 = 𝑓(𝑇, 𝐷𝑜𝑝𝑖𝑛𝑔 𝑎𝑛𝑑 𝐶𝑎𝑟𝑟𝑖𝑒𝑟 𝐷𝑒𝑛𝑠𝑖𝑡𝑦)
Note: y=f(x) denotes the variable, y depends on (or function of) the variables inside the
parenthesis.
𝐸𝐶 −𝐸𝐹
𝑛 = 𝑁𝐶 𝑒 − ( )
𝑘𝑇

𝐸𝐹 −𝐸𝑉
𝑝 = 𝑁𝑉 𝑒 − ( )
𝑘𝑇

22
This below ni is calculated using Boltzmann approximation:

where, Eg(T) = Eg(0) - Δ𝐸𝑔 (𝑇) => BGN due to lattice temperature
(when T ↑ 𝒕𝒉𝒆𝒏 𝑬𝒈 ↓)

Ebgn => amount of BGN due to doping and carrier density (Doping ↑ and (n or
p)↑ then 𝐸𝑔 ↓ )
Note: 𝒏𝒊𝒆𝒇𝒇 ≥ 𝒏𝒊 since Ebgn is ≥ 𝟎 𝒆𝑽

Correct estimation of ni is important because:

i) In non-degenerate, minority carrier concentration depends on ni,eff
value.
ii) All recombination models depend on ni,eff2.
4.2Effect of Eg vs T:
Varshini’s Equation (1966):
𝛼𝑇 2
𝐸𝑔 (𝑇) = 𝐸𝑔 (0) −
𝑇+𝛽
This is the empirical model for band gap narrowing which obeys for many
materials (Si, Diamond, SiC and more). Where,
T -> Lattice Temperature;
Eg(0) -> Band gap for T = Tref (mostly Tref at 0K);
𝛼 𝑎𝑛𝑑 𝛽 are fitting parameters (material dependent).
4.3Effect of Electron Affinity vs T:
𝜒 = 𝑓(𝑇) 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑚𝑜𝑑𝑒𝑙𝑠 𝑒𝑥𝑐𝑒𝑝𝑡 𝐽𝑎𝑖𝑛 − 𝑅𝑜𝑢𝑠𝑡𝑜𝑛 𝑤ℎ𝑒𝑟𝑒 𝜒 = 𝑓(𝑇, 𝐷𝑜𝑝𝑖𝑛𝑔)
𝒘𝒉𝒆𝒏 𝑻 ↑, 𝒕𝒉𝒆𝒏 𝝌 ↑

23
Only Jain Roulston model, electron affinity is the function of doping
concentration where 𝜒 increases with doping concentration as shown in Figure 5.

Figure 5 BGN decreases with doping and 𝝌 increases with doping for Jain-Roulston model.
Similar trend is observed for temperature increase. Note: Vacuum level not necessarily be
constant and there can be slight band bending too.

4.4Bandgap Narrowing Models:

Physical Reason for BGN:
Bandgap is the minimum energy required to make free electron from the
atoms and hence creation of e-h pair. When there are more carriers (either due to
i) temperature increase, ii) doping or iii) excess carrier injection due to high
electric field or concentrated light), exciting the electron (making it free) from the
atoms becomes easier. That is, band gap is narrowed or shrinked.
Why it is easy to make electron free when carriers are more?
When there are lot of carriers, those carriers might interact with atom,
transfer energy to the atom and create free electron very easily.
Exact reason for band gap narrowing might be due to complex interactions:
i) Carrier – Dopant Interaction (e-Dopant, h-Dopant) or Ionic
Interaction
ii) Carrier – Carrier Interaction (e-e, h-h, e-h) or Exchange-Correlation
Interaction
iii) Band Tails at very heavy doping (acceptor and donor states form
bands at heavy doping)
These interactions are complex and involves deep quantum mechanical
calculations (many body interactions) to accurately model them. For eg:

24
 i) electrons of same spin will repel (Pauli/Fermi repulsion) and hence
even though electrons are crowded, this repulsion reduces their net
repulsive force.
ii) carriers might screen the dopant atoms (like electrons might
surround the ionized donor (positively charged) atom)
Hence modeling BGN as it is, is a complex task which is successfully accounted
in Schenk BGN model. Other models are empirical and approximate models
which will not suitable for all situations. Like delAlamo is suited only for n-type
materials.
Bandgap Narrowing is considered as sum of two terms:

where, Δ𝐸𝑔0 → Band gap narrowing due to doping or carrier

injection.
Δ𝐸𝑔𝐹𝑒𝑟𝑚𝑖 → Correction for using Fermi Dirac statistics when
apparent BGN models are used.
Bandgap narrowing (Older and Newer Models)

(apparent means: Looking correct but not necessarily)

Here apparent BGN is named because BGN used earlier is accumulation of many
effects like degeneracy in high doping levels (degeneracy: states with same
energy level), change in DOS due to impurity bands, asymmetry of band gap
shrinkage (Δ𝐸𝑐 ≠ Δ𝐸𝑣 𝑤ℎ𝑒𝑟𝑒 Figure 5 indicates asymmetry BGN because Ev
doesn’t change at all (Δ𝐸𝑉 = 0)). Apparent BGN doesn’t replicate the actual
bandgap narrowing.
Older model doesn’t match the device current-voltage characteristics for very
heavily doped devices (like BJT, solar cells).
Why it is so?
Boltzmann overestimates the minority carrier density and hence pn product
and hence to compensate it, BGN will be reduced than actual value and hence
band bending will be lower than actual and hence difference in experimental
device characteristics compared to simulation will be observed.

25
 4.4.1 Apparent BGN Models
Empirical Model based on Electrical measurement (pn product):
 Slotboom
2
𝑁𝑑𝑜𝑝 𝑁𝑑𝑜𝑝
Δ𝐸𝑔0 = 𝐸𝑟𝑒𝑓 [ln ( ) + √ln ( ) + 0.5 ]
𝑁𝑟𝑒𝑓 𝑁𝑟𝑒𝑓

2
𝑁𝑑𝑜𝑝 𝑁𝑑𝑜𝑝
Ebgn = 𝐸𝑟𝑒𝑓 [ln ( ) + √ln ( ) + 0.5 ]
𝑁𝑟𝑒𝑓 𝑁𝑟𝑒𝑓

Note: Slotboom fitting parameters are updated and old ones are called
OldSlotboom model.
 delAlamo (for n-type semiconductor).
𝑁𝑑𝑜𝑝
Δ𝐸𝑔0 = 𝐸𝑟𝑒𝑓 ln ( )
𝑁𝑟𝑒𝑓
In electrically measuring bandgap, they construct heavily doped diode or BJT and
measure the current and compare it with analytical (compact) model and calculate
ni,eff value. From ni,eff, BGN is calculated. They are based on transport equations
which in turn on mobility models and carrier density.

26
Empirical Model based on Optical absorption measurement:
 Bennett/Wilson model (for n-type material)
𝟐
0
𝑁𝑑𝑜𝑝
Δ𝐸𝑔 = 𝐸𝑟𝑒𝑓 [ln ( )]
𝑁𝑟𝑒𝑓
Sample is shined with light (using optical fiber) and transmitted light is
detected using spectrometer. Incident Light –(minus) Transmitted light
gives the absorbed wavelength of light (wavelength -> energy) and hence
bandgap is known. When there are free carriers, those carrier too absorb
energy which will be given back again as phonons (free carrier absorption)
which is ignored in the above calculation. Hence, it might be slightly
overestimated (square denotes it) compared to actual band gap.

27
 4.15.1 3.4.2 Jain Roulston Model
Doping dependent BGN model based on simplified quantum mechanical
approach. Jain et al has derived compact (or analytical) model based on all the
interactions discussed above and also considering the multi-valley nature of the
bands, anisotropic nature of the minima and maxima (direction dependent). Their
model works well for both n and p-type, Si, Ge, GaAs, GexSi1-x alloys and strained
layers.
𝑎
𝑁𝑑𝑜𝑝 1 1 1
Δ𝐸𝑔0 ∝ ( ) 𝑤ℎ𝑒𝑟𝑒 𝑎 𝑖𝑠 , ,
𝑁𝑟𝑒𝑓 2 3 4

4.15.2 3.4.3 Schenk BGN

All other models (Bennett, delAlamo, Slotboom, Jain Roulston) depends
only on doping. But, Schenk is only complete physics based analytical model
using quantum mechanical approach which accounts both doping and carrier
injection for BGN.
BGN => Doping Induced Narrowing +
28
 Plasma (Lot of free carriers) Induced Narrowing.
Large number of excess carriers (Might be large electric field or High
intensity optical beam) also induce bandgap narrowing called plasma induced
BGN. It is more common in BJT and concentrated solar cells.
Schenk BGN model doesn’t match with electrically measured Band gap
data because electrical measured data depends on the transport of carriers and
they use Boltzmann for extraction. But matches well with optical absorption data.
This model view the carrier densities as many body problem (i.e. their
interaction) rather ideal gas model (only elastic scattering between carriers).
There are two parts in the model which are treated equally:
i) Exchange correlation, Δ𝑥𝑐 = f(n, p, T)
Interaction between free energies of carriers
ii) Ionic correlation, Δ𝑖 = f(Ndop , n, p, T)
Interaction between dopants and carriers
Amount of band gap narrowing, Ebgn = Δ𝑥𝑐 + Δ𝑖
This model shows very good matching at low and moderate doping. But at very
high doping, little mismatch (10%) even though trend matches. This might be due
to non-consideration of band tails due to heavy doping.
Without Schenk, we will not able to simulate the realistic solar cell or BJT or any
other heavy doped devices. We are really standing on the shoulders of the Giants.

The revision in exp. Data is due to the update in the ni value

29
 4.5Additional Notes on Band structure simulation (Not for Exam)
4.5.1 Multivalley Band structure
Used for strained Si and for low band-gap materials.
4.5.2 Band gap Widening
When material dimension reduces, band-gap widens.
Also, when heavily doped, in some materials, band gap widens called Moss
Burstein effect.
Above said both phenomenon, can be accounted in numerical simulations.

4.6DOS and Effective Mass

If effective mass is given, effective DOS can be calculated and reverse calculation
(i.e. effective DOS from effective mass) is also possible.
Note: Effective mass values changes for DOS calculation and conductivity
calculation because the bands involved and nature of the calculation are different.
For further details: http://ecee.colorado.edu/~bart/book/effmass.htm
Effective mass for DOS calculation is taken as geometric mean of three masses:
(two transverse and one longitudinal)
Note: https://en.wikipedia.org/wiki/Geometric_mean for Geometric mean. If you
have to compare two (or more) numbers of different range in equal weightage,
then you must take Geometric mean.

30
Figure 6 E-k Diagram of Silicon along 100 direction within first Brillouin Zone showing two
anisotropic (direction dependent) minima. Other four minima lies in other directions.

These six anisotropic minima are characterized by longitudinal mass

𝑚𝑙 𝑎𝑛𝑑 𝑡𝑤𝑜 transverse mass 𝑚𝑡 . (Needs further understanding here). Effective
mass is calculated using below equation:

where, 6 in above equation denotes 6 anisotropic minima and other term

denotes the geometric mean of all three masses. This calculated effective mass is
used to calculate effective DOS as shown below:

Note: 𝑇𝑛 -> Electron temperature not lattice temperature. Depending on the field,
electron temperature might be more than lattice temperature.
Why DOS increases with carrier Temperature?
Since DOS is increasing as parabolic with energy, electron with higher
energy (higher Tn) would have more states than electron with lesser energy (lower
Tn).
Reverse Calculation:

31
If effective DOS states is given like below:

where, effective mass will be calculated from effective DOS from below
relation (Modified from previously specified effective DOS and effective
mass relation).

Ref:
1. For more information on effective mass:
http://ecee.colorado.edu/~bart/book/effmass.htm
2. References cited in the Sentaurus Device manual
3. Sentaurus Device manual

32
 28-Sept-2015

4. MOBILITY MODELS
4.16 4.1 Introduction
 Freedom of movement for carriers inside a material. (Material property)
o Mobility of GaAs is several times larger than Silicon
 It affects both the drift current ( 𝑞𝑛𝝁ξ) and diffusion current
𝑑𝑛 𝑘𝑇
( 𝑞𝐷 where, 𝐷 = 𝜇 for non-degenerate semiconductor. The
𝑑𝑥 𝑞
relation between D and 𝜇 is called Einstein relation.
 It depends on number of factors:
o Lattice Temperature (Lattice Scattering)
o Ionized Dopants (Impurity Scattering)
 Only Ionized Dopants not inactive dopants because of the
field exerted by ionized dopant affect the carrier’s mobility
by pulling or pushing the carriers.
o Carrier Density (or e & h density) (Carrier – Carrier Scattering)
o Interface of two materials (surface roughness and surface phonon
scattering)
o High Electric field (Mobility degradation)
 How all the effects are accounted?
o Matthiessen’s rule:
𝟏 𝟏 𝟏
= + +⋯ (4.1)
𝝁𝒆𝒇𝒇 𝝁𝑳𝒂𝒕𝒕𝒊𝒄𝒆 𝝁𝑰𝒎𝒑𝒖𝒓𝒊𝒕𝒚
Note:
 It is only approximate way of accounting. Actually
scattering of each carrier should be accounted by taking
average for all the carrier scattering. But since, it is
almost impractical for numerical simulation, the above
rule is used.
 Whichever lower mobility, dominate the overall
mobility.

o Effective mobility is the combination of all the factors (discussed

above).
o For high field effect:
o Low field mobility is computed using equation (4.1) and
for high field, mobility is computed (in two steps, along
with Eq. 4.1) using:
𝝁𝑯𝒊𝒈𝒉𝑭𝒊𝒆𝒍𝒅 = 𝒇(𝝁𝑳𝒐𝒘𝑭𝒊𝒆𝒍𝒅 , 𝝃𝒉𝒊𝒈𝒉 ) (4.2)
33
4.17 4.2 Mobility Models
Several mobility models are introduced for successful computer simulation
of semiconductor devices.
Need for accurate mobility model:
i) Correct mobility model at the inversion layer in MOSFET is
important for accurate estimation of terminal current calculation in
2D and 3D simulation.
ii) To get correct estimate of Sheet resistance (in process simulation).
Challenges in accurate mobility model:
i) Mobility model should be used for non-planar devices like shown
below:

Figure 7a Planar Device (MOSFET). Note: 2D device is different from 2D simulation where 2D
device represent the structure of the device is same in third dimension (into the paper for the
above device). 3D simulation can be done for 2D device and 2D simulation can be done for 3D
device.

Figure 8 Nonplanar device (FinFET : 3D device since gate is wrapped on three sides of the
channel)

ii) Separating the mobility contribution from temperature,

impurity, field is difficult.
4.17.1 4.2.1 Mobility Degradation due to Phonon scattering in the
Bulk
Due to random lattice vibrations (Phonons), electron movement (mobility)
will be hindered. Due to atomic core, carriers will get scattered randomly due to
the coulomb attraction/repulsion. If the lattice vibrates more, then more

34
degradation to mobility. Lattice vibration completely depends on lattice
temperature. Actually at low temperature, acoustic phonon (or Intra-valley
phonon) dominate and at higher temperature, higher energy phonons (optical or
Inter-valley phonon) will dominate the scattering. It is given by the empirical
relation following power law dependence:
300𝐾 +𝛾 𝑇 −𝛾
𝜇(𝑇) = 𝜇(300𝐾) ( ) = 𝜇(300𝐾) ( )
𝑇 300𝐾
5 𝑐𝑚/𝑠 𝑐𝑚/𝑠
𝑤ℎ𝑒𝑟𝑒 𝛾 ≈ 𝑎𝑛𝑑 𝜇𝑒 (300𝐾) = 1400 and 𝜇ℎ (300𝐾) = 470
2 𝑉/𝑐𝑚 𝑉/𝑐𝑚

when 𝑻 ↑ , 𝒕𝒉𝒆𝒏 𝝁(𝑻) ↓

Note: semi-empirical means some values are not physics based, i.e. those values
are measured from experiments and hence simplifying the physical derivation.
The above mentioned mobilities are valid for pure lattice (without impurity) and
hence, they represent the maximum mobility value.

Figure 9 This is taken from reference: (Jacoboni et al, SSE, 1977), where, the mobility is measured
from Hall effect measurement (filled circles) and Time of Flight measurement (filled triangles)
for high purity Si. The measured result shows good correspondence with power law dependence.

Note: Time of Flight (ToF): Experimental technique to measure the time duration
of moving (flying) carriers across the width of the sample generated by radiation
35
as shown in below Figure 10. Hall effect to measure mobility is discussed
elsewhere.

Figure 10 Time of Flight Experiment to measure mobility

Note: From current ( 𝐽 = 𝑞𝑛𝜇𝑛 𝜉 => 𝜉 𝑖𝑠 𝑘𝑛𝑜𝑤𝑛 𝑎𝑛𝑑 𝐽 𝑖𝑠 𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑. Hence,

n𝜇𝑛 product can be measured and hence only if n is known, 𝜇𝑛 can be measured.
4.17.2 4.2.2 Mobility degradation due to impurity scattering
- side effect of doping. (Doping is done to increase conductivity by
increasing number of carriers, but as a side effect, conductivity is slightly reduced
due to reduced mobility). Different models to account impurity scattering are:
i) Masetti (For Silicon)
ii) Arora (For GaAs)
Note: Narain Arora is an Indian who studied in IIT Delhi and currently
working as a Senior Vice President at Silterra Inc. USA
iii) University of Bologna (Italy)
4.17.2.1 4.2.2.1 Masetti Model
First empirical model which fits well with experimental data and simulated
sheet resistance data for wide range of doping concentration. It includes reduced
mobility in heavily doped Arsenic (As) compared to Phosphorus (P) (since As

36
atom is larger than P, scattering would be more in As) and also models mobility
degradation in heavily Boron (B) doped silicon.

Whether electrically inactive dopant (dopant which didn’t give away free
carrier) degrades mobility?
No, actually electrically inactive dopant scattering is very much negligible
compared to ionized dopant scattering due to the field around the active
dopants.

Note: Mobility degradation is given for dopant density greater than solid
solubility limit. How they achieved the active dopant density greater than solid
solubility limit? => By Laser annealing.
All the empirical measurements for mobility are taken from Hall effect
measurement.

Figure 11 Masetti model (empirical model) compared with experimental data where arrow
indicates the solid solubility limit at 1100°𝑪. This is taken from reference: Masetti et al, TED,
1983. Before Masetti model, all other models fit well for lower dopant concentration but rapidly
diverge for higher dopant concentration.

The Masetti empirical model with highlighted Masetti term to account good
convergence for n > 5 × 1019 𝑐𝑚−3 :

37
where, 𝑁1 ≈ 1016 𝑐𝑚−3 𝑎𝑛𝑑 𝑁2 ≈ 1020 𝑐𝑚−3 . The first two term is called max-
min equation. When n (since carrier concentration increases with doping
concentration) increases, 𝜇 decreases due to ionized impurity scattering. The first
term ensures mobility is not degraded too much by the model (since constants can
be varied) and hence, minimum mobility is maintained. The third term is the
Masetti’s addition to the existing model, where it accounts the scattering at the
higher carrier concentration (> 1020 𝑐𝑚−3 ). Second term is negligible for n <
1020 𝑐𝑚−3 since denominator will be very large. Figure 12 shows the comparison
of Masetti model and older model where third term contribution can be seen
clearly which fits well with experimental data as shown in Figure 11.

38
 Figure 12: Comparison of Masetti and Older models.

Requires correction in first term for boron doped silicon. Also it accounts
separately arsenic and phosphorus by different fitting parameters.
4.17.2.2 4.2.2.2 Arora Model
Actually Arora has proposed this model (one year before Masetti) for Silicon with
doping concentration below 1020 𝑐𝑚−3 . But it is unknown, why Sentaurus
recommends this model for GaAs. It follows max-min model which is the first
two term in Masetti model (see section 4.2.2.1 Masetti Model).
4.17.2.3 4.2.2.3 University of Bologna Model
This model is the extension of Masetti model. They made two extensions:
i) Extended temperature model
𝐓
𝑇 −𝛾+𝐜(𝟑𝟎𝟎𝐊)
𝜇(𝑇) = = 𝜇(300𝐾) ( )
300𝐾
Addition to previously discussed model is highlighted and it is done to
extend the validity for 25°𝐶 𝑡𝑜 973°𝐶
ii) Individual attraction and repulsion of ionized impurities are accounted
by:

39
 (𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) 𝝁𝟏
𝜇 = 𝜇𝑚𝑖𝑛 + − 𝜸
𝑁𝐴 𝛼 𝑁𝐷 𝛽 𝟏 + ( 𝑵𝟑 𝑵𝟒
1+( ) +( )
𝑁1 𝑁2 𝑵𝟑 𝑵𝑨 + 𝐍𝟒 𝑵𝑫 )
Here all the fitting parameters ( 𝑁1 , 𝑁2 , 𝑁3 , 𝑁4 ) are functions of
temperature. Once we compare with Masetti model, we can easily
observe the splitting of total doping concentration (and hence attraction
and repulsion) in the above model.
4.17.3 4.2.3 Carrier-Carrier Scattering Models
Note on Classical Mobility Model (Based on Ref: Li et al, SSE, 1977):
Carrier lose momentum when they hit on lattice atoms or impurity ions. Hence
velocity (and hence mobility) is reduced. But how they lose momentum by
carrier-carrier scattering? Will not the bombarded electron gains momentum?
Actually, when two electrons scatter, momentum exchange occurs in such a
way that, net transfer of momentum from electrons which dissipate momentum
(through lattice scattering) less efficiently to those which dissipate more
efficiently, resulting in momentum loss and hence velocity loss and reduction
in mobility.

Carrier-Carrier scattering is modeled as below (fortunately both classical

and quantum mechanical model agrees well in most cases):
𝑇 𝛼
( )
𝜇𝑒ℎ ∝ 300𝐾
√𝑛𝑝
Where, if carrier density is more, then mobility degrades quickly. But in high
temperature, mobility degradation due to carrier scattering is less and only lattice
scattering dominates and hence, to suppress the mobility degradation due to

40
carrier-carrier scattering, 𝜇𝑒ℎ increase with temperature. (Hence, this term
1
becomes negligible in Mattheisen’s rule, as ( => 𝑠𝑚𝑎𝑙𝑙))
𝑙𝑎𝑟𝑔𝑒

Hint on Carrier-carrier scattering modeling:

Actually this model is derived from well understood impurity-carrier scattering
model. Just replace impurity doping concentration by carrier density and include
models for screening.

There are two models to account scattering. They are discussed as follows:

i) Conwell and Weisskopf Model (CW model)

Classical based model (Ref: Fletcher 1957). This model is actually
derived for impurity carrier scattering. Since similar model is
derived for gas collisions by Chapman and Cowling, this model is
equally applicable to electron-hole scattering (as in classical
approximation, electron and holes can be seen as gas particles).

𝒌𝑻
Figure 13: Classical model that shows reduction of Diffusivity (Or mobility since 𝑫 = 𝒒
𝝁) with
17 -3
respect to carrier concentration. This model shows a threshold (p=10 cm , after which mobility
decays rapidly.
a. Problem in CW model:
i. Scattering is only assumed to happen at the center of the
impurity atom. Hence, small angle scattering is not included.

41
 ii. Screening effect is included in this model by randomly
reducing the potential to half the mean distance apart from
other scattering centers.

Figure 14 CW scattering model assumption (one of the assumption)

iii. Carriers are scattered only by one ion (but not the case
practically).
ii) Brooks-Herring Model (Quantum Model)
Quantum mechanical based model. It includes the screening effect
by reducing potential around the ion.
Screening is accounted by reducing the potential by using
𝛽𝑠 (𝑖𝑛𝑣𝑒𝑟𝑠𝑒 𝑠𝑐𝑟𝑒𝑒𝑛𝑖𝑛𝑔 𝑙𝑒𝑛𝑔𝑡ℎ) parameter.

If screening length is more (ion’s field can’t have any effect), then
𝛽𝑠 is less and hence V(r) is large
Both the above models agree well in most cases. Mobility degradation due to
carrier carrier scattering(𝜇𝑒ℎ ) is included through the below relation:
1 1 1
= +
𝜇 𝜇𝑜𝑡ℎ𝑒𝑟 𝜇𝑒ℎ
𝜇𝑜𝑡ℎ𝑒𝑟 is the mobility degradation to other dependencies (like temperature,
doping,..)

42
Note: Author Herring will not be found in Brooks paper. Actually his work is
unpublished but still he got his recognition.
4.17.4 4.2.4 Unified Mobility Model
It uses the Masetti model (see section 4.2.2.1 Masetti Model) as the base
model for the analytical calculation of all mobility degradation term in single
expression for both minority and majority carriers. Hence called unified mobility
model (since it unifies the temperature, impurity, carrier-carrier scattering,
screening for both majority and minority carrier) or Philips Unified mobility
model or Klaassen model.
Note: Klaassen is the scientist proposed this model while working at Philips
Research Laboratory in The Netherlands.

Usually electron mobility is greater than hole mobility, i.e., electron moves
faster than hole. But is it true of all situations?
Whether hole mobility will be greater than electron mobility?
Yes, in case of heavily doped n-type silicon, minority carrier mobility (hole)
will be greater than majority carrier (electron) by atleast three times.
(Hole pictured as tortoise and electron pictured as rabbit)

Importance of Klaassen Model:

 Earlier models use separate empirical expression for Minority carrier and
majority carrier mobility (as function of doping concentration) since they
behave differently at high doping. It causes convergence issue (simulation
will be failed) if majority and minority carriers have separate mobility
equations in numerical simulation.
 Certain empirical model proposed with single function of doping
concentration but they didn’t include electron-hole scattering and valid
only for room temperature. (electron-hole scattering can’t be ignored at
high doping)
Derivation for mobility degradation due to doping from Masetti’s Model
By subtracting out the lattice vibration term from Masetti’s model using
Matthiesen’s rule:

43
 1 1 1
= −
𝜇𝐼 𝜇𝑀𝑎𝑠𝑒𝑡𝑡𝑖 𝜇𝐿
where, lattice vibration degradation, 𝜇𝐿 = 𝜇𝑚𝑎𝑥 (Maximum limit for mobility)
1 1 1
= −
𝜇𝑖,𝐼 𝜇𝑚𝑖𝑛 + 𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛𝛼 −
𝜇1 𝜇𝑚𝑎𝑥
𝑁 𝛽
1+( 𝐼 ) 𝑁𝑟𝑒𝑓2
𝑁𝑟𝑒𝑓 1+( 𝑁 )
𝐼

Ignoring high doping term (or Masetti term) assuming lower doping
concentration:

𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛
𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 − ( )
𝑁𝐼 𝛼
1+(
1 1 1 𝑁𝑟𝑒𝑓 )
= − =
𝜇𝑖,𝐼 𝜇𝑚𝑖𝑛 + 𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛
𝛼
𝜇𝑚𝑎𝑥
𝑁 𝜇 − 𝜇𝑚𝑖𝑛
1+( 𝐼 ) 𝜇𝑚𝑎𝑥 (𝜇𝑚𝑖𝑛 + 𝑚𝑎𝑥 )
𝑁𝑟𝑒𝑓 𝑁𝐼 𝛼
1+(
𝑁 ) 𝑟𝑒𝑓

where, i => electron or hole and I => Acceptor Ion or Donor Ion and Nref is fitting
𝑁 𝑁
parameter (to avoid confusion from 𝐼 => 𝐼 is used)
𝑁1 𝑁𝑟𝑒𝑓

Taking 𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 as common term in Numerator:

1
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) (1 − )
𝑁𝐼 𝛼
1+(
1 𝑁𝑟𝑒𝑓 )
=
𝜇𝑖,𝐼 𝜇2 − 𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛
𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 + 𝑚𝑎𝑥
𝑁 𝛼
1+( 𝐼 )
𝑁𝑟𝑒𝑓
Taking 𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 as common term in denominator:

1
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) (1 − )
𝑁𝐼 𝛼
1+(
1 𝑁𝑟𝑒𝑓 )
=
𝜇𝑖,𝐼
1 2
𝜇𝑚𝑎𝑥
𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 (1 − )+
𝑁𝐼 𝛼 𝑁𝐼 𝛼
1+( 1+(
𝑁𝑟𝑒𝑓 ) 𝑁𝑟𝑒𝑓 )

44
Taking reciprocal:

2
1 𝜇𝑚𝑎𝑥
𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 (1 − )+
𝑁𝐼 𝛼 𝑁𝐼 𝛼
1+( 1+(
𝑁𝑟𝑒𝑓 ) 𝑁𝑟𝑒𝑓 )
𝜇𝑖,𝐼 =
1
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) (1 − )
𝑁𝐼 𝛼
1+(
𝑁𝑟𝑒𝑓 )
2
𝜇𝑚𝑎𝑥
𝑁𝐼 𝛼
1+(
𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 𝑁𝑟𝑒𝑓 )
= +
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 )
1
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) (1 − )
𝑁𝐼 𝛼
1+(
𝑁𝑟𝑒𝑓 )
2
𝜇𝑚𝑎𝑥
𝑁𝐼 𝛼
1+(
𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 𝑁𝑟𝑒𝑓 )
= +
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) 𝑁𝐼 𝛼
(𝑁 )
𝑟𝑒𝑓
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) ( )
𝑁𝐼 𝛼
1 + (𝑁 )
𝑟𝑒𝑓
2
𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 𝜇𝑚𝑎𝑥
= +
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) 𝑁𝐼 𝛼
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) (
𝑁𝑟𝑒𝑓 )
𝜇𝑚𝑎𝑥 𝜇𝑚𝑖𝑛 2
𝜇𝑚𝑎𝑥 𝑁𝑟𝑒𝑓 𝛼
= + ( )
(𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) (𝜇𝑚𝑎𝑥 − 𝜇𝑚𝑖𝑛 ) 𝑁𝐼
𝑵𝒓𝒆𝒇 𝜶
𝝁𝒊.𝑰 = 𝝁𝒊,𝒄 + 𝝁𝒊,𝑵 ( )
𝑵𝑰
𝜇 𝜇𝑚𝑖𝑛 2
𝜇𝑚𝑎𝑥
where, 𝜇𝑖,𝑐 = (𝜇 𝑚𝑎𝑥 )
and 𝜇𝑖,𝑁 = (𝜇
𝑚𝑎𝑥 −𝜇𝑚𝑖𝑛 𝑚𝑎𝑥 −𝜇𝑚𝑖𝑛 )

Now since carriers screen the ions and hence mobility degradation will be less at
high doping concentration. To account it, collision cross section that derived from
including screening is given as follows (From Brooks-Herring Theory):

45
 −1
−1 𝑵𝒓𝒆𝒇 𝜶
𝐶𝑜𝑙𝑙𝑖𝑠𝑖𝑜𝑛 𝑐𝑟𝑜𝑠𝑠 𝑠𝑒𝑐𝑡𝑖𝑜𝑛 ∝ {𝜇𝑖,𝐼 𝑁𝐼 } = {𝝁𝒊,𝒄 𝑵𝑰 + 𝝁𝒊,𝑵 ( ) 𝑵𝑰 }
𝑵𝑰
Since to find carrier screening effect, replace impurity concentration by carrier
concentration (For the second term in above equation, no replacement is done,
because, that term becomes smaller at high doping concentration and screening
becomes significant only at high doping concentration).
−1
−1 𝑁𝑟𝑒𝑓 𝛼
{𝜇𝑖,𝐼 𝑁𝐼 } = {𝜇𝑖,𝑐 𝑐 + 𝜇𝑖,𝑁 ( ) 𝑁𝐼 }
𝑁𝐼
where, 𝑁𝐼 𝑖𝑠 𝑟𝑒𝑝𝑙𝑎𝑐𝑒𝑑 𝑏𝑦 𝑐𝑎𝑟𝑟𝑖𝑒𝑟 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛, 𝑐 = 𝑛 + 𝑝.
𝒏+𝒑 𝑵𝒓𝒆𝒇 𝜶
𝝁𝒊,𝑰 = 𝝁𝒊,𝒄 + 𝝁𝒊,𝑵 ( )
𝑵𝑰 𝑵𝑰
Hence, when carrier density is more, screening will be more and hence scattering
will be less which results higher mobility value which is obvious from above
equation. The above equation gives the majority carrier impurity scattering
term.
For minority carrier impurity scattering,
𝜇𝑚𝑎𝑗𝑜𝑟𝑖𝑡𝑦
𝐹1 = => This is theoretically calculated function using quantum
𝜇𝑚𝑖𝑛𝑜𝑟𝑖𝑡𝑦
mechanical calculation and then analytically fitted to measure minority carrier
mobility. Hence majority carrier mobility is calculated first and from it, minority
carrier is calculated (No separate doping dependence for minority carrier mobility
and hence unified model).
for eg: Electron mobility in acceptor doped material (Minority carrier mobility)
𝜇
is given by, 𝜇𝑒,𝐴 = ℎ,𝐴
𝐹1

where, 𝜇ℎ,𝐴 is hole mobility in acceptor doped material (Majority carrier

mobility).
Note: Function given some random name, F1
For electron-hole scattering
Idea behind it: electron is seen as moving acceptor ion and hole is seen as moving
donor ion and scattering between two moving charged particles is solved quantum
mechanically. Contribution to electron mobility due to hole scattering (𝜇𝑒,ℎ ) and
(contribution to hole mobility due to electron scattering (𝜇ℎ,𝑒 )) is given by
function (F2) which is calculated similar to F1.

46
 𝜇𝑒,ℎ = 𝐹2 ∗ 𝜇𝑒,𝐷
Hence, this term also depends on majority carrier mobility term.

For dopant clustering

Physical meaning for third term (Masetti term) in Masetti model accounting
for high doping concentration?
Nearly after 9 years after Masetti model, Klaassen explains this as
follows: At very high doping concentration, carriers are no more scattered from
ion with one electronic charge. Dopants form cluster (groups) and hence their
ionic contribution is more than one electronic charge

It is accounted by replacing ionized dopant concentration by:

𝑁𝐼 = 𝐹3 ∗ 𝑁𝐼 𝑎𝑐𝑡𝑢𝑎𝑙
where F3 ≥ 1.
Effective mobility in unified model is the contribution from all the above
discussion (which is given by Matthiesen’s rule)
1 1 1 1 1
= + + +
𝜇𝑒𝑓𝑓 𝜇𝑖,𝐿 𝜇𝑖,𝐷 𝜇𝑖,𝐴 𝜇𝑖,𝑗
where, i and j can be e or h.
Note: The above expression is simplified well for quick computation in numerical
simulation which is not included here.
4.17.5 4.2.5 Interface Mobility Degradation
In the channel region of a MOSFET, the high transverse electric field
forces carriers to interact strongly with the semiconductor–insulator interface.

47
Carriers are subjected to scattering by acoustic surface phonons and surface
roughness and hence undergo mobility degradation.
1
𝜇∝
𝐹𝑖𝑒𝑙𝑑 𝑃𝑒𝑟𝑝𝑒𝑛𝑑𝑖𝑐𝑢𝑙𝑎𝑟 (𝐹⊥ )

4.17.6 4.2.6 High Field Saturation

When electric field increases, velocity of carriers increases. But at high
electric field, velocity becomes contant (saturated to higher value) and becomes
independent of field. This is called high field saturation.
𝑓(𝜉), 𝑓𝑜𝑟 𝑙𝑜𝑤 𝜉
𝐷𝑟𝑖𝑓𝑡 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦, 𝑣𝑑 = {
𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡, 𝑓𝑜𝑟 ℎ𝑖𝑔ℎ 𝜉
Time of flight technique (explained in Figure 10) is used to measure the drift
velocity, from the measured time required for the carriers to cross the sample
width.
𝑊𝑖𝑑𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑎𝑚𝑝𝑙𝑒
𝑣𝑑 =
𝑇𝑖𝑚𝑒 𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑑 𝑡𝑜 𝑐𝑟𝑜𝑠𝑠 𝑡ℎ𝑒 𝑠𝑎𝑚𝑝𝑙𝑒
Important challenge is how to segregate the self-heating effect at high electric
field. They have mounted the sample (silicon diode) on the hot finger whose
temperature is constantly maintained to 2°𝐶 so that any heat produced from the
sample will be coupled to the hot finger.
4.17.6.1 Models for high field saturation:
 Canali Model
 Extended Canali Model
4.17.6.2 4.2.6.1 Canali Model
Empirical model based on Time of Flight experiment done at large field and
extended temperature upto 473°𝐶. At low electric field, the experimental results
shows good match with the previously reported data (hence validating the
experimental technique). Then, they have extended their experiment for higher
electric field and found the following empirical fit:
𝜉
𝑣𝑑 = 𝜇 1
𝛽 𝛽
𝜉
(1 + (𝜉 ) )
𝑐

where, 𝜉𝑐 𝑎𝑛𝑑 𝛽 are fitting parameters where 𝜉 𝑖s the applied electric field.

48
 𝜉 𝛽
At high electric field, 𝜉 > 𝜉𝑐 => ( ) ≫ 1
𝜉 𝑐

𝜉
𝑣𝑑)𝑎𝑡 ℎ𝑖𝑔ℎ 𝜉 = 𝜇 1 = 𝜇 𝜉𝑐 = 𝑣𝑠𝑎𝑡 = 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡 (𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑜𝑓 𝜉)
𝛽 𝛽
𝜉
((𝜉 ) )
𝑐

The above expression is valid for fixed temperature where the fitting parameters
are measured. To include temperature dependence, then this model becomes little
bulky with power law dependence on temperature.
Note: Similar expression for field dependence is earlier reported by Caughey-
Thomas in 1967. Canali (1975) extended for higher temperature and found the
same dependence with differences in fitting parameters.

Figure 15 𝒗𝒅 𝒗𝒔 𝝃 𝒇𝒓𝒐𝒎 𝑹𝒆𝒇: Canali TED 1975. Dots represents experimental data and line represents the empirical
model. When temperature increases, velocity degrades further but following the same trend with respect to 𝝃.

This model is taken into simulation using the following mobility terms:
𝜉ℎ𝑓𝑠 𝜉ℎ𝑓𝑠
𝑣𝑑 = 𝜇(𝜉) 𝜉ℎ𝑓𝑠 = 𝜇𝑙𝑜𝑤 1 = 𝜇𝑙𝑜𝑤 1
𝜉ℎ𝑓𝑠 𝛽 𝛽
(1 + ( ) 𝛽) 𝜉ℎ𝑓𝑠
𝜉𝑐 (1 + ( 𝑣𝑠𝑎𝑡 ) 𝛽)
𝜇𝑙𝑜𝑤

49
where, 𝜇𝑙𝑜𝑤 is the mobility at low electric field and 𝜉ℎ𝑓𝑠 is the applied electric
field (hfs -> high field saturation) and 𝑣𝑠𝑎𝑡 is the constant which depends on
material and temperature.
𝝁𝒍𝒐𝒘
𝝁(𝝃) = 𝟏
𝝁𝒍𝒐𝒘 𝝃𝒉𝒇𝒔 𝜷 𝜷
(𝟏 + ( 𝒗 ) )
𝒔𝒂𝒕

The above one can be easily implemented in numerical simulation. Such a field
effect will be dominant for smaller size devices like MOS transistor (as size
𝑉 𝑉
reduces, field will be more dominant => for same 𝜉, ≫ , i.e., for same
𝑛𝑚 𝜇𝑚
applied voltage, nano-meter range devices have higher field than micro-meter
range devices). For such smaller devices, slight change (𝜇𝑙𝑜𝑤 → 2𝜇𝑙𝑜𝑤 𝑎𝑛𝑑 1 +
𝑡𝑒𝑟𝑚 𝑖𝑛 𝑑𝑒𝑛𝑜𝑚𝑖𝑛𝑎𝑡𝑜𝑟) in above equation is made (Hansch JAP 1986).
𝟐𝝁𝒍𝒐𝒘
𝝁(𝝃) = 𝟏
𝜷 𝜷
𝟐𝝁𝒍𝒐𝒘 𝝃𝒉𝒇𝒔
[𝟏 + (𝟏 + ( ]
𝒗𝒔𝒂𝒕 ) )

4.17.6.3 4.2.6.2 Transferred Electron Model

For GaAs devices, at high field, electron jump to higher energy valley where the
effective mass would be large and hence negative differential mobility. Such
cases can be modeled using Transferred electron model.

50
 𝜉 4
𝜇𝑙𝑜𝑤 𝜉 + 𝑣𝑠𝑎𝑡 ( )
𝜉0
𝑣𝑑 =
𝜉 4
1+( )
𝜉0
This model explains the negative differential mobility.
At low 𝜉: 𝜉 ≪ 𝜉0
𝑣𝑑 𝑎𝑡 𝑙𝑜𝑤 𝜉 ≈ 𝜇𝑙𝑜𝑤 𝜉
At very high 𝜉: 𝜉 ≫ 𝜉0
𝑣𝑑 𝑎𝑡 𝑣𝑒𝑟𝑦 ℎ𝑖𝑔ℎ 𝜉 ≈ 𝜇𝑙𝑜𝑤 𝜉
At 𝜉 at negative differential mobility regime: 𝜉 ≥ 𝜉0
𝑣𝑠𝑎𝑡
𝜇𝑙𝑜𝑤 𝜉 +
𝑣𝑑 𝑎𝑡 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑚𝑜𝑏𝑖𝑙𝑡𝑦 𝑟𝑒𝑔𝑖𝑚𝑒 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 < 𝑣 𝑎𝑡 𝑙𝑜𝑤 𝜉
𝑑
𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡
The above equation should be modified for mobility as we did earlier for Canali
model.

Ref:
1. Sentaurus Device Manual
2. All the models are referred with the original proposed article and the
key understandings are given here. For detailed understanding, readers
are encouraged to go through the original articles. (Brooks paper on
Scattering is not available for our access)

51
 5. INCOMPLETE IONIZATION

Incomplete ionization will be playing a role at low temperatures since

dopants will not be ionized at that temperatures and also it depends on the
ionization energy of dopant atom.
For eg: Indium dopant in Silicon will not completely ionized at room
temperature. (All other dopants like Boron, Phosphorus and Arsenic will be
ionized completely).
Model to account incomplete ionization is given by:

where, 𝑔𝐴 , 𝑔𝐷 is the degeneracy factors

𝑁𝐴,0 𝑎𝑛𝑑 𝑁𝐷,0 => 𝑇𝑜𝑡𝑎𝑙 𝑑𝑜𝑝𝑎𝑛𝑡 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛
𝑁𝐴 , 𝑁𝐷 => 𝐼𝑜𝑛𝑖𝑧𝑒𝑑 𝑑𝑜𝑝𝑎𝑛𝑡 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛
Hence, ionized dopant concentration depends on the ionization energy
(material property) and temperature.

52
6. RECOMBINATION
Recombination is a process whereby electron and hole are annihilated or destroyed. There
are two type of recombination, direct and indirect.

Direct recombination
In direct recombination method, an excess population of electrons and holes decays by
electrons falling from the conduction band to empty state in the valence band. Energy lost
by the electron in making the transition is given up as a photon. The net rate of change in
the conduction band electron concentration is the thermal rate minus the recombination rate
𝑑𝑛(𝑡)
= 𝛼𝑟 𝑛𝑖2 − 𝛼𝑟 𝑛(𝑡)𝑝(𝑡)
𝑑𝑡

Where,
𝛼𝑟 𝑛𝑖2 = Thermal generation rate
𝛼𝑟 𝑛(𝑡)𝑝(𝑡) = Recombination rate
The excess electron in the p-type semiconductor recombine with a decay constant
𝜏𝑛 = (𝛼𝑛 𝑝0 ) -1 called the minority carrier lifetime. More general expression for carrier
lifetime is
1
𝜏𝑛 = 𝛼𝑟 (𝑛0 +𝑝0 )

In case of direct recombination, the majority decay at exactly the same rate as the minority
carrier.
In TCAD,
Radiative model is used for account direct recombination method. The radiative (direct)
recombination model expresses the recombination rate as:
2
𝑅𝑛𝑒𝑡 = 𝐶 ∗ (𝑛𝑝 − 𝑛𝑖,𝑒𝑓𝑓 )
Where,
C= 2x10-10 cm3s-1 for GaAs and C=0 for other materials

Indirect recombination
The indirect recombination happens through intermediate energy level present within
the band gap as shown in fig. 1., and the resulting energy loss by recombination electron is
usually given up to the lattice as heat rather than by the emission of photons. The SRH
(Shockley-Read-Hall) recombination is the example for indirect recombination. The energy
level ET present between Eg is called trapping center. In silicon donor impurities provide
the energy level near the conduction band and acceptor impurities provide nearer to valence
band. Some impurity like Zn, produce multiple energy levels in the band gap Zn- level at
0.31 eV near above the valence band and Zn= at the middle of band gap.

53
 Fig. 1. Indirect recombination mechanism
In TCAD,
Recombination through deep defect levels in the gap is usually labeled Shockley–
Read–Hall
(SRH) recombination. In Sentaurus Device, the following form is implemented
2
𝑆𝑅𝐻 𝑛𝑝− 𝑛𝑖,𝑒𝑓𝑓
𝑅𝑛𝑒𝑡 =𝜏
𝑝 (𝑛+𝑛1)+ 𝜏𝑝 (𝑝+𝑝1)

Where,
n1 and p1 are trap energy
𝜏𝑛 𝑎𝑛𝑑 𝜏𝑝 are minority carrier lifetime
The minority carriers lifetime are modeled as doping dependent, temperature dependent
and field dependent.

Doping dependent
The doping dependence of the SRH lifetimes is modeled in Sentaurus Device with the
Scharfetter relation.
𝜏𝑚𝑎𝑥 −𝜏𝑚𝑖𝑛
𝜏𝑑𝑟𝑜𝑝 (𝑁𝐴 + 𝑁𝐷 ) = 𝜏𝑚𝑖𝑛 + 𝑁 +𝑁
1+( 𝐴 𝐷 )
𝑁𝑟𝑒𝑓

The minority carrier lifetime is depends on Nref, NA+ND below Nref minority carrier lifetime
is constant but below Nref minority lifetime starts to reduce.

Temperature dependent
The temperature dependence was modeled by a power law
𝑇
τ (T) = τ0[300𝐾]α
Minority carrier life time is increased with rising temperature.

Field dependent
Field enhancement reduces SRH recombination lifetimes in regions of strong
electric fields. It must not be neglected if the electric field exceeds a value of approximately
in 3x105 Vcm-1 certain regions of the device. For example, the I–V characteristics of
reverse-biased p-n junctions are extremely sensitive to defect-assisted tunneling, which
causes electron–hole pair generation before band-to-band tunneling or avalanche
generation.

54
Surface SRH recombination
The surface SRH recombination model can be activated at semiconductor–
semiconductor and
Semiconductor – insulator interfaces.
2
𝑆𝑅𝐻 𝑛𝑝− 𝑛𝑖,𝑒𝑓𝑓
𝑅𝑆𝑢𝑟𝑓,𝑛𝑒𝑡 = 𝑛+𝑛1 𝑝+𝑝1
+
𝑆𝑝 𝑆𝑛
Where,
Sp and Sn are surface recombination velocity.

Auger recombination
In the Auger recombination process, both direct and indirect recombination happens
with the collision between two like carrier. The energy releases by the recombination is
transferred during the collision to the surviving carrier. This highly energetic carrier
subsequently thermalizes losses energy in small step through heat-producing collision with
the semiconductor lattice.

In TCAD,
Auger recombination is typically important at high carrier densities. This effect is
explained as resulting from exciton decay: at lower carrier densities, excitons, which are
loosely bound electron–hole pairs, increase the probability for Auger recombination.
Excitons decay at high carrier densities, resulting in a decrease of recombination.

𝐴 2
𝑅𝑛𝑒𝑡 = (𝐶𝑛 𝑛 + 𝐶𝑝 𝑝)(𝑛𝑝 − 𝑛𝑖,𝑒𝑓𝑓 )

Where,
Cn and Cp are Auger coefficient.

55
 7. GENERATION MODELS:
Impact ionization is the inverse of Auger recombination. It is three particle
generation processes. Carriers can get high energies while traveling through
the high electric field undergoes scattering events with bonded electrons in the
valence band. The excess energy transferred to this electron which is lifted in
the conduction band creating a new electron hole pair.

Ref: http://www.iue.tuwien.ac.at/phd/triebl/node20.html

Electron–hole pair production due to avalanche generation (impact

ionization) requires certain threshold field strength and the possibility of
acceleration, that is, wide space charge regions. If the width of a space charge
region is greater than the mean free path between two ionizing impacts, charge
multiplication occurs, which can cause electrical breakdown. The reciprocal of
the mean free path is called the ionization coefficient. With these coefficients for
electrons and holes, the generation rate can be expressed as

Gii   n nvn   p pv p

vanOverstraeten–de Man

𝐾
𝛼𝑛 (𝐹𝑎𝑣𝑎 ) ∝ 𝑒𝑥𝑝 (− )
𝐹𝑎𝑣𝑎

𝛼(𝐹𝑎𝑣𝑎 ) = 𝑓(𝑇, 𝐹𝑎𝑣𝑎 )

56
 MODELS FOR DEVICES:
BJT:

 Mobility:
o Klaassen
 Band gap Narrowing:
o Schenk (Considers both doping and excess carrier)
 Recombination:
o Auger and SRH
 Generation:
o Impact Ionization (in CB junction)
 Transport Model:
o Hydrodynamic model
 Carrier Statistics:
o Fermi-Dirac
MOSFET:
 Mobility:
o Klaassen
o High Field Saturation
 Band gap Narrowing:
o Old Slot boom with fitted parameters (since only due to doping)
and time saver compared to Schenk
 Recombination:
o Auger and SRH
 Transport Model:
o Hydrodynamic model
 Carrier Statistics:
o Fermi-Dirac

57
Solar Cell:
 Mobility:
o Klaassen
 Band Gap Narrowing:
o Schenk (Due to optical induced excess carriers)
 Recombination:
o Radiative, Auger and SRH
 Transport Model:
o Drift-Diffusion Model (due to larger active area)
 Carrier Statistics:
o Fermi-Dirac
 Optical Generation:
o Raytracing

58
18-Feb-2016

FURTHER UPDATION IN MODELS

BGN

Auger
Fermi

59
 ADDITIONAL CURRENT IN SENTAURUS TCAD:

Interface conditions always include the displacement current in insulators,

conductive insulators, and semiconductors to ensure conservation of current.

No further details given in Sdevice manual. Probably this displacement current

should be due to the change in displacement vector (𝐷 = 𝜖 𝜉) across different
𝑑𝐷
materials. 𝐽𝐷 ∝
𝑑𝑡

“Displacement current density has the same units as electric current density.
However it is not an electric current of moving charges, but a time-varying electric
field. There is also a contribution from the slight motion of charges bound in atoms,
called dielectric polarization”

“The first term on the right hand side is present in material media and in free
space. It doesn't necessarily come from any actual movement of charge, but it
does have an associated magnetic field, just as a current does due to charge
motion. Some authors apply the name displacement current to the first term by
itself.[2]”
60
“The second term on the right hand side, called polarization current density,
comes from the change in polarization of the individual molecules of the
dielectric material. Polarization results when, under the influence of an
applied electric field, the charges in molecules have moved from a position of
exact cancellation. The positive and negative charges in molecules separate,
causing an increase in the state of polarization P. A changing state of polarization
corresponds to charge movement and so is equivalent to a current, hence the term
polarization current"
This displacement current that compensates for the conduction current in a
capacitor.

[Source: Wiki]

61