Vous êtes sur la page 1sur 5

Excerpt from the Proceedings of the COMSOL Conference 2007, Boston

Modeling of Low Pressure Magnetron Plasma Discharge


Francisco Jimenez, Samuel D. Ekpe* and Steven K. Dew
Department of Electrical and Computer Engineering, University of Alberta
Edmonton, AB, Canada T6G 2V4
*ekpe@ece.ualberta.ca

Abstract: The plasma characteristics of a low


pressure plasma discharge are dependent on the
process conditions in a complicated manner. In
order to understand this dependence, a
comprehensive model, which takes into
consideration the effects of the E X B field and
gas heating on the transport of plasma particles,
is required. In this work, a Monte Carlo fluid
Poisson hybrid model is used in describing the
plasma discharge at low pressure typical for
magnetron sputtering. The source term used in
the fluid model is derived from a Monte Carlo
code. The models are coupled to a gas heating
module.
Results
compare
well
with
experimentally measured data using a Langmuir
probe.
Keywords: Plasma,
hybrid model.

magnetron,

sputtering,

1. Introduction
Low pressure plasma discharges have several
technological applications. In a magnetron
sputtering process, it is widely used in the thin
film
deposition
of
elemental/compound
materials. The structures and properties of the
growing film are affected by the process
conditions, as well as the interactions, through
collisions, between energetic particles. These
particles include electrons, ions, neutrals, etc.
The interactions between these particles and
the process gas (such as argon) result in
momentum and energy exchange between the
colliding particles and the gas, and may result in
excitation, ionization and dissociation of the
process gas as well as other neutrals [1,2]. The
use of a magnetron plays an important role in
confining electrons around the vicinity of the
target (cathode), and in increasing the collision
probability of electrons with the process gas.
This in turn increases the creation rate of ions
within the plasma. As the target is biased
negatively with respect to the walls, the positive
ions are accelerated towards it.
The interactions also result in the heating of
the process gas and cause local rarefaction [3,4].

This will affect the ionization rate, as well as the


transport of the particles. Detailed understanding
of these plasma processes is essential in the
design and optimization of the system processes
and conditions for a typical application. To
achieve the above, an attempt is made in using
the COMSOL Multi-physics tool to solve a
fluid-Poisson model and in creating the
necessary magnetic field required for modeling a
magnetron operation.

2. Model
The plasma may be described using
continuity equations (for the charge species) and
Poisson equation (for the electrical potential
determination). The electrons may be grouped as
slow electrons (created within the main
discharge region) and fast electrons (created at
the cathode or in the sheath). The equations in
general form may be expressed as:
(1)
Slow electrons ne + J e = Re
t
ni
Ions
(2)
+ J i = Ri
t
(3)
Potentials 2 = q (n n n fast )
0 i e e
Where ni , ne , nefast are respectively, the
densities of ions, slow and fast electrons, is the
electrical potential, q is the elementary
discharge, and 0 is the permittivity of free space.
Ji and Je are the fluxes of ions and slow electrons
which are determined by the electric field and
the mobilities of the charged particles. The
source terms Ri and Re are ion and slow electron
creation rates. Since some of the fast electrons
reaching the bulk plasma may be converted to
the slow electrons, the electron creation rate may
be different from the creation rate of the ions.
These creation rates may be calculated from a
Monte Carlo simulation of fast electrons [1,5].
The electrons and ions fluxes may be
expressed as:
(4)
J e = e ne E Dene
J i = i ni E Di ni

(5)

Excerpt from the Proceedings of the COMSOL Conference 2007, Boston

Equations (4) and (5) describe the transport of


the charge particles as a drift-diffusion relation.
Where E is the electric field intensity given as:
E = , is the mobility and D is the diffusion
coefficient for the respective charges.
The mobility and diffusion coefficient in the
absence of magnetic field may be expressed as:
q
(6)
=
meff n
kT
(7)
D =
meff n
meff is the effective mass, and n = v n is the
collision frequency between particle and the
neutral species, n is the gas density (a function of
the gas temperature), is the collision crosssection of the particle with the gas, v is the
velocity of the particle, k is the Boltzmann
constant, and T is the temperature. In the
presence of a magnetic field as is the case in
magnetron sputtering, the mobility and diffusion
coefficient equations must be modified, and
hence equations (4) and (5). In addition, the
effect of the E X B field must also be included in
calculating the fluxes of the particles.
The modified particle flux becomes [2] :
J = zn ( // E // + E ) + d n [E h ]
(8)
D// (n ) // D (n ) + Dd [h n ]
z designates the particle charge, which is -1 for
electrons and 1 for the singly-charged ions, h =
B/|B| is the unit vector in the direction of the
magnetic field. The subscripts // and are used
for the directions parallel and perpendicular to
the magnetic field. Since the component of the
velocities parallel to the magnetic field are not
affected by the magnetic field, the parallel
components of the mobility and diffusion
coefficients are as expressed in equations (6) and
(7). The affected components may be expressed
as:
1
(9)
d =
2
B 1 + mn 2n / m2 2
kT
(10)
Dd =
z qB 1 + m2n 2n / m2 2
q
1
(11)
=
mn n 1 + m2 2 / m2n 2n
kT
1
(12)
D =
mn n 1 + m2 2 / m2n 2n

Where, = qB / m is the cyclotron frequency


of particle .

3. Numerical Method
The physical geometry consists of a
cylindrically
symmetric
magnetron
as
represented in a 2D sketch shown in figure 1.
The electrons are mainly confined in the trap
region above the target by magnetic field, where
ions are created. These ions are accelerated
towards the target, by an applied electric field,
where they impinge normally. Secondary
electrons are created in the process with an
emission coefficient, , such that at the cathode,
the electron current density may be expressed in
terms of the ion current density as:
je(0) = ji(0)
(13)
Equation (13) may be used as the boundary
condition for the ions.
Substrate (Anode)

Electron Trap region

E-field

B-field
Target (Cathode)
S
N
Soft Iron

y
x

Figure 1. Sketch of the physical geometry of a 2D


cylindrically symmetric magnetron used in the process
modeling. The electrons are confined by the magnetic
field above the target where ions are created.

To implement these physical assumptions,


and determine the density of fast electrons and
source terms, Monte Carlo simulation code is
used to establish the ionization sites, and follow
charged particles trajectories [1,5]. The
trajectories are defined by the electric and
magnetic fields, and are of course modified by
collisions at random intervals.
Gas temperature for the process conditions
used in this work was calculated by assuming
that the energies lost by the energetic particles in
collision with the process gas are used in part in
heating up the gas [4,6]. The gas heating causes
localized rarefaction of the process gas. The
density of the gas may be estimated for an ideal
gas as n = P/kT, where P is the gas pressure.

Excerpt from the Proceedings of the COMSOL Conference 2007, Boston

The source terms, density of fast electrons


and gas temperature are then defined as functions
in a COMSOL Multiphysics simulation tool
where the fluid-Poisson model is implemented.
In order to solve the continuity and Poisson
equations, suitable boundary conditions are set.
These boundary conditions are [7,8]:
Electrons: At the cathode and all walls, ne = 0
Ions: At the walls, ni = 0
At the cathode, eq (13)
Potential: At the walls, = 0 (ground)
At the cathode, = -Vapplied
To generate the necessary magnetic field,
magnetostatic module is used, with stationary
linear Direct UMFPACK solver, and with
surface current set to zero on all exterior walls.
The Poisson equation is solved using the
Electrostatic module of the COMSOL
Multiphysics tool, while continuity equations are
solved for the ions and electron densities by
adopting the conduction convection component
of the Chemical Engineering Module. Stationary
nonlinear solver Direct UMFPACK was
used. The equations were discretized into 2D.

4. Results and Discussion

3.1 Process and Model Parameters

Figure 2. Gas temperature profile of the discharge


space due to the sputtering of Aluminum at magnetron
power of 200W and pressure of 1 Pa

A magnetron power of 200 W was assumed


applied to a 7.6 cm aluminum (Al) disk target at
a potential of -385 V and discharge current of
0.52 A. The distribution of the discharge current
(ion current) over the surface of the target was
assumed to follow a Gaussian distribution
function, thus representing a typical erosion
profile. Pressure of the process gas (argon) was
taken as 1 Pa.
Ion temperature was assumed equal to the
gas temperature. The electron temperature was
estimated from [9], Te = qEre/k, where re = ve/ e
is the electron gyro radius. Collision crosssections for the electrons and ions were,
respectively, assumed as 3.0e-20 and 1.0e-19 m2.
A value of 0.02 was assumed for the secondary
electron emission coefficient.
For the diffusion component, effective
diffusion coefficient was evaluated from the
parallel and perpendicular components, and
isotropic diffusion assumed. To ensure stability,
a tuning parameter of 0.6 and 0.9 were used for
the electron and ion components respectively.

The calculated gas temperature is shown in


figure 2. This indicates that most of the heating
is localized at a region a short distance from the
target, and leads to an in-homogenous rarefaction
of the process gas within the discharge space.
Figure 3 shows electron source used in the
model. The source term for the ion follows a
similar trend.

Figure 3. Electron source calculated from a Monte


Carlo code [5]. The electron trap region is defined by
the applied magnetic field, where the charges are
created is highly localized above the target region.

The flux of ions and electrons consists of both


diffusive and convective as expressed in
equations (4) and (5). The calculated convective
component of the ion total flux is represented in

Excerpt from the Proceedings of the COMSOL Conference 2007, Boston

figure 4. The pattern of the flux on the target


surface defines the target erosion track.

Figure 4. Convective component of the total ion flux.


It clearly shows the pattern of the target erosion
profile.

Figure 5 shows the calculated ion density plotted


alongside generated magnetic field lines. The
result, as expected, indicates a high
concentration of ions just above the erosion track
of the targets surface. It further demonstrates the
suitability of using equation (13) as an
appropriate boundary condition for the ion at the
cathode. Outside the sheath, the ion density falls
sharply with distance.
The density profile for electrons is shown in
figure 6. The applied magnetic field, as shown in
the figure, clearly confines most of the electron
above the target. The electron density profile
rises from zero at the target and peaks within the
plasma bulk, and then falls off away from the
target toward the substrate region. The results
show that, in the cathode fall region, the
discharge current is mainly due to the ions which
are accelerated towards the cathode, in minimal
part due to the emitted secondary electrons from
the target.
The trend of these results compares
favourably with the experimental results reported
elsewhere [10]. This suggests that it was
reasonable to adopt the conduction-convection
component of the Chemical Engineering module
of COMSOL in solving for the electron and ion
densities.

Figure 5. Ion density profile plotted alongside


calculated magnetic field. The density peaks at just
above the target surface but falls off away sharply
away from the target.

Figure 6. Electron density profile, whose distribution


is greatly determine by the applied magnetic field.

The calculated electrical potential of the


discharge space is shown in figure 7. The
potential rises significantly from the applied
potential to almost 0 a short distance from the
target. At some distances from the target towards
the substrate region, the plasma potential was
obtained to vary between 5 V. However, at
distances close the side-walls, the potential as
high as about 31 V was noted. This high plasma
potential may attributed to some of the
assumptions made in the model as well as
associated instabilities, and may not be a true
physical
representation,
though
further
investigation is required to confirm the results
obtained. It is interesting to note the existence of
a slightly greater extension of the cathode fall

Excerpt from the Proceedings of the COMSOL Conference 2007, Boston

above the erosion track than at the central axis.


The effect of the magnetic field pattern is clearly
shown here.

Figure 7. Electrical Potential profile calculated for the


discharge space. The potential rises significantly in
front of the cathode to near zero at a few mm from the
target surface, and clearly defines the cathode fall. The
thickness of the cathode fall is slightly greater above
the erosion track than at the central axis.

5. Conclusions
In this study, we have modeled low pressure
plasma discharge typical for a dc magnetron
sputtering process using a hybrid Monte Carlo,
fluid-Poisson model. Gas heating effects are
included in the model. Densities of the plasma
species (electrons and ions) and electrical
potential of the discharge space are calculated
using the COMSOL Multiphysics modeling tool.
The trend of the results compare well with
experimental results reported elsewhere. The
effect of magnetic field on the profile of the
charge densities and fluxes, and electrical
potential are clearly illustrated in the results
obtained. The results of this model illustrate the
usefulness of adopting COMSOL tool in
modeling a plasma discharge. More work is
however required in redefining some of the
model assumptions / parameters, such as using a
more physically reasonable assumption for the
electrical conductivity of the process gas.

6. References
1. Sheridan, T.E., Goeckner, M.J. and Goree, J.,
Model of Energetic Transport in Magnetron
Discharges, J. Vac. Sci. Technol., A 8, 30-37
(1990)
2. Golant, V.E., Zhilinsky, A.P., and Sakharov,
I.E. Fundamentals of plasma Physics, John
Wiley & Sons, New York (1980)
3. Rossnagel, S.M., Gas Density Reduction
Effects in Magnetrons, J. Vac. Sci. Technol., A
6, 19-24 (1988)
4. Ekpe, S.D. and Dew, S.K., 3D Numerical
simulation of Gas Heating Effects in a
Magnetron Sputter Deposition System, J. Phys.
D, A 39, 1413-1421 (2006)
5. Field, D.J., Study of Electrons in Magnetron
Sputtering System for Novel Thin Film
Applications, PhD Thesis, University of Alberta,
Canada (2005)
6. Jimenez, F., Ekpe, S.D., and Dew, S.K.,
Inhomogeneous Rarefaction of the Process Gas
in a DC Magnetron sputtering System, J. Vac.
Sci. Technol., A 24, 1530-1534 (2006)
7. Bogaerts, A., Gijbels, R. and Goedheer, W. J.
Hybrid Monte Carlo-fluid Model of A Direct
Glow Discharge, J. Appl. Phys., 78, 22332241(1995)
8. Surendra, M, Graves, D.B., and Plano, L.S.,
Self-consistent DC Glow-discharge Simulations
Applied to Diamond Film Deposition Reactors,
J. Appl. Phys. 71, 5189-5198(1992)
9. Pekker, L., Longitudinal Distribution of
Plasma Density in the Low-pressure glow
Discharge with Transverse Magnetic Field,
Plasma Sources Sci. Technol., 4, 31-35 (1995)
9. Field, D.J., Dew, S.K. and Burrell, R.E.,
Spatial Survey of a Magnetron Plasma
Sputtering System Using a Langmuir Probe, J.
Vac. Sci. Technol., A 20, 2032-2041 (2002)

Vous aimerez peut-être aussi