VICTORY 3D Process Simulation - Silvaco

VICTORY 3D Process Simulation
Modeling of Deep Reactive Ion Etching
Presented at SISPAD 2007

Introduction
Motivation
Physical Process
VICTORY Simulation Environment
DRIE Model
Results
Conclusions
VICTORY 3D Process Simulation SISPAD 2007

Motivation
Deep Reactive Ion Etching (DRIE) is very popular in surface
micro-machined MEMS
Etches quickly nearly vertical trenches
High aspect ratios can be achieved
It is a very challenging process for simulation
Multiplexed sequence of processes
Thin passivation layers
Mutual influence of fast and slow processes components
Extremely demanding for 3D simulation analysis
VICTORY 3D Process Simulation SISPAD 2007

Physical Process
Multiply repeated sequence of an etching and passivation process
Number of cycles determines the trench depth
Each etching and passivation cycle lasts only a few seconds
Passivation :
Grow a thin layer of passivation material which cannot be attacked by the
chemical process
Etching :
Locally break the thin passivation layer by accelerated ions
Etch the underlying material by a chemical process
VICTORY 3D Process Simulation SISPAD 2007

Physical Process
Process shows three regimes when varying passivation / etching
cycle times
Passivation dominated regime :
The sputtering/etching time is not sufficient to remove the passivation
layer properly
Not more than only a very small part of the trench bottom near the corners
can be cleared from passivation layer material
With this conditions the DRIE process produces no proper trenches at all
VICTORY 3D Process Simulation SISPAD 2007

Physical Process
Process shows three regimes when varying passivation / etching
cycle times
Passivation influenced regime :
Shape of the passivation layer influences the shape of the bottom of the
trench
VICTORY 3D Process Simulation SISPAD 2007
Isotropic etching starts
near the trench corners,
because the sputtering is
homogenous at the bottom
The sputter/etching time is just not sufficient to remove the whole
passivation layer from the bottom of the trench
Effective etch rate reduces with trench width

Physical Process
Process shows three regimes when varying passivation / etching
cycle times
Etch dominated regime :
Shape of the passivation layer has no significant influence on the trench
bottom shape
Average etch rate increases by increasing the trench width
More than 50% of the etching cycle time is isotropic chemical etching
VICTORY 3D Process Simulation SISPAD 2007

VICTORY Simulation Environment
Capability to simulate full processes in 3D
Etching, Deposition, Oxidation, Implantation, Diffusion
Open modeling interface
Model parameters and modeling functions can be accessed and
modified
C-Interpreter is used to develop the model functions
Level-Set based framework
For structure representation (multi-layer structure concept)
For interface propagation
Hierarchical Cartesian mesh
VICTORY 3D Process Simulation SISPAD 2007

The DRIE Model
Models for both cycles are required
Passivation cycle
Deposition model with one reactant
Etching cycle
Etching model with two reactants
Interaction of the cycles only through geometrical modifications
Both cycles are characterized by
Ballistic transport of the reactive particles
Mean free particle path : several mm
Typical feature size : several µm
VICTORY 3D Process Simulation SISPAD 2007

The DRIE Model – Passivation Cycle
Assumptions:
Only one reactant is dominating the deposition reaction at the surface
The reactant has an isotropic momentum distribution above the wafer
surface
The sticking efficiency at the surface is very high
Secondary effects can be approximated
VICTORY 3D Process Simulation SISPAD 2007

The DRIE Model – Passivation Cycle
Implementation in VICTORY Process:
Use the primary ballistic mode of the deposition module to
calculate the transport towards each point of the surface
VICTORY 3D Process Simulation SISPAD 2007
1
( x ) = ⋅ f ( P , ... , P , D , ... , D ) ⋅cos(
( x , ϑ,
ϕ ) ⋅sin(
ϑ)
F s ∫∫ 1 n 1 m
s
N visible−space
F
f
P ,
1
N
( xs
)
( P , ... , P , D , ... , D )
1
... , P
D1
, ... , D
α
( x , ϑ,
ϕ)
s
n
m
n
1
m
....
.... Model parameters
....
....
....
Reactant distribution
( C - Interpreter
Model arguments (variables)
Angle to the local surface
Normalization
coefficient
F
α dϑ
dϕ
.... Reactant flux to a point on the surface
based modeling function)
( D ( x ) )
normal
( x ) = 1 for a flat surface point
s
function above the wafer surface
i
s

The DRIE Model – Passivation Cycle
Implementation in VICTORY Process:
Use the primary ballistic mode of the deposition module to calculate
the transport toward each point of the surface
Primary ballistic mode assumes a sticking efficiency of 100%
VICTORY 3D Process Simulation SISPAD 2007

The DRIE Model – Passivation Cycle
Implementation in VICTORY Process:
Use the primary ballistic mode of the deposition module to calculate
the transport toward each point of the surface
Primary ballistic mode assumes a sticking efficiency of 100%
To consider secondary effects (sticking efficiency < 100%) add a
very thin conformal deposition
Avoids solving the transport equation for secondary transport
F
VICTORY 3D Process Simulation SISPAD 2007
( x ) = ⋅ f ( P , ... , P , D , ... , D ) ⋅cos(
α(
x , ϑ,
ϕ ) ⋅sin(
ϑ)
+
s
1
N
∫∫
F
∫∫
visible−space
s
mutual−visible−surfacepoints
1
( y ) ⋅s(
y ) ⋅ε(
r , y ) ⋅
s
n
xy
1
s
cos
m
( α(
r , y ) ⋅cos(
α(
r , x )
xy
s
r
xy
s
2
yx
s
dϑ
dϕ
⋅ dσ
y

The DRIE Model – Passivation Cycle
Implementation in VICTORY Process:
Use the primary ballistic mode of the deposition module to calculate
the transport toward each point of the surface
Primary ballistic mode assumes a sticking efficiency of 100%
To consider secondary effects (sticking efficiency < 100%) add a very
thin conformal deposition
Avoids solving the transport equation for secondary transport
Assume isotropic momentum distribution of the reactant above
the wafer surface
f
VICTORY 3D Process Simulation SISPAD 2007
( P ... , P , D , ... , D ) = 1
1 , n 1
m

The DRIE Model – Passivation Cycle
Implementation in VICTORY Process:
Use a one particle surface reaction
Assume that local deposition rate only depends on the particle
flux
Influence of surface properties on the local rate is neglected
Surface reaction modeling function is C-Interpreter based
x = S P , ... , P , D , ... , D = F x ⋅ R
VICTORY 3D Process Simulation SISPAD 2007
( s)
( 1 n 1 m)
( ) 0
R s
R
S
R
F
( xs
)
( P , ... , P , D , ... , D )
( xs
)
( x )
0
1
s
n
1
m
.... Local deposition rate at a point of the surface
.... Surface reaction modeling function
R
are model variables
0
( x ) and F(
x )
s
s
.... Depositon rate at a plane wafer surface
.... Reactant flux towards
surface point x
s

The DRIE Model – Etching Cycle
Assumptions:
Two types of reactants have to be considered
Chemically reactive neutrals
Not reacting with passivation layer material
Isotropic momentum distribution above the wafer surface
Accelerated ions
Sputtering material from the wafer surface
Reaction is not selective
Sputtering reaction is much slower than chemical reaction
Momentum distribution with a preferential direction towards the wafer
surface
VICTORY 3D Process Simulation SISPAD 2007

The DRIE Model – Etching Cycle
Implementation in VICTORY Process:
Condense the contribution of both particles into a single particle
by using a surface material dependent transport model
Transport of only one particle has to be calculated
Saves a lot of calculation time
Effective plain wafer etch rate can be measured
Micro-loading effects are not modeled properly
VICTORY 3D Process Simulation SISPAD 2007

The DRIE Model – Etching Cycle
Implementation in VICTORY Process:
Use the primary ballistic mode of the etching module to calculate
the transport towards each point of the surface
Use a surface material dependent reactant distribution function
VICTORY 3D Process Simulation SISPAD 2007
1+
ε ( xs
) ⋅V
( κ , ϑ)
f ( P1
, ... , Pn
, D1
, ... , Dm)
=
1+
ε ( x )
V
ε
( κ
, ϑ)
( x )
s
....
Normalized Von - Mises distribution
.... Ratio of
ε
ε
ε
the contribution
( xs
) is large if
( xs
) is small if
( x ) is a calibration
parameter
s
surface material is
s
function
of neutrals - ions to the effective etchrate
surface material is passivation
silicon

The DRIE Model – Etching Cycle
Implementation in VICTORY Process:
Use a one particle surface reaction
Assume that local etch rate only depends on the particle flux
Influence of surface properties on the local rate is neglected
Surface reaction modeling function is C-Interpreter based
VICTORY 3D Process Simulation SISPAD 2007
( xs
) = S(
P1
, ... , Pn
, D1
, ... , Dm)
= F(
x ) ⋅ R0
R s
R
S
R
F
( xs
)
( P , ... , P , D , ... , D )
( xs
)
( x )
0
1
s
n
1
m
.... Local etch rate at a point of the surface
.... Surface reaction modeling function
R
are model variables
0
( x ) and F(
x )
s
s
.... Effective etch rate at a plane wafer surface
.... Reactant flux towards
surface point x
s

Results
VICTORY 3D Process Simulation SISPAD 2007
Regime dominated by isotropic etching
Experiment Simulation

Results
VICTORY 3D Process Simulation SISPAD 2007
Regime strongly influenced by the shape of the passivation layer
Experiment Simulation

Results
VICTORY 3D Process Simulation SISPAD 2007
Regime dominated by the passivation process
Experiment Simulation

Conclusion
An efficient model for the simulation of deep reactive ion etching
has been presented
It reproduces the main characteristics of the deep reactive ion
etching process
The model has been implemented in the three-dimensional
simulation environment VICTORY Process which properly
handles this numerically challenging simulation problem
The open modeling interface of VICTORY Process has been used
for model development purposes
VICTORY 3D Process Simulation SISPAD 2007