Simulation of a PIII reactor with a magnetized remote source
Mathieu MAURY1, Khaled HASSOUNI1, Armelle MICHAU1
Frank TORREGROSA2, Gaël BORVON2
1
: LSPM (UPR3407), Paris 13 University, 99 avenue Jean-Baptiste Clement, 93430 Villetaneuse, FRANCE
2
: Ion Beam Services S.A, avenue Gaston Imbert Prolongée, 13790 Peynier, FRANCE
Abstract:
Plasma Immersion Ion Implantation (PIII) is reaching the industrial phase, but
fine tuning of PIII reactors is still difficult without a correct understanding of the
phenomena involved. This is critical in the case of an ICP remote source
connected to a large implantation chamber, because plasma coupling mechanisms
scale differently in each part of the device.
A modular simulation has been developed to optimize the PULSION
reactor made by IBS. First a global ICP model, coupled to a chemical kinetics
module computing the relevant reactions for BF3, is used to assess the electronimpact fragmentation of the doping precursor. Then the magnetically expanding
plasma is treated with a simple flux conservation model. Finally, a Particle In
Cell sheath model including all the fragments previously determined is used to
determine the sheath dynamics and ion energy distribution on the wafer.
Parametric studies have been done on the complete model. The plasma
chemical composition and ion energy distribution can readily be obtained from
the reactor tuning parameters accessible to the operators.
Keywords: ion implantation ; numerical simulation ; fluid model
1
Introduction
The Plasma Immersion Ion Implantation (PIII)
is a emerging technology in the plasma processing
flied that can greatly simplify ion implantation in
comparison to traditional beam-line implantation.
2. An implantation chamber where large substrates
can be hold and polarized with a pulsed negative
high voltage bias.
The principle of operation is simple. A plasma
is generated in a source with the suitable doping gas
and then transferred near to the wafer sample to be
implanted. The latter is placed on a chuck (sample
holder) connected to a pulsed high voltage power
supply. When the substrate is biased to a negative
voltage of a few kV, the resultant electric field
accelerates the doping ions into the substrate with an
implantation energy controlled by the substrate bias.
The industrial reactor investigated [1] is
composed of two main parts :
1. A remote source where an inductively coupled
plasma (ICP) is generated through a RF solenoid
antenna, with magnetic coils to reduce ion losses
at the lateral wall.
Figure 1. Schematic diagram of the simulated PIII reactor. All
dimensions are in millimeters.
But fine tuning of this type of reactors is still
difficult because of a strong coupling between
relevant plasma mechanisms (power deposition,
doping gas fragmentation, wall losses) and that said
mechanisms scale differently in each part of the tool.
In consequence, a correct understanding of the
plasma phenomena involved is needed to implement
efficient changes in the reactor design without a
lengthy trial-and-error process.
2
Description of the simulation
The code developed to simulate this PIII reactor
has been divided into several modules in order to
simplify the numerical treatment of the different
equation sets involved and to ease troubleshooting.
As of today, it is comprised of :
1. A global model derived from the argon ICP
plasma model developed by Lieberman [2],
coupled to a chemical kinetics module
computing the relevant doping gas reactions [3].
2. A 1D Particle In Cell (PIC) sheath model
developed in our laboratory and adapted to PIII,
connected to the global model trough a simple
flux conservation model.
The preliminary results presented in this paper
have been obtained for a chamber pressure of 0.1 Pa
and an absorbed RF power of 200 W. The current
doping gas studied is boron trifluoride BF3, but
phosphine PH3, arsine AsH3 and diborane B2H6 will
be investigated in the future.
2.1
Global source module
A preliminary study was done with the global
source model described above. The goal here was to
determine the order of magnitude for characteristic
electron parameters and get a global view of the
space-time behavior of the discharge.
Although the Maxwellian EEDF hypothesis
was kept, power deposition equations were modified
for the case of a magnetized source and the reaction
cross-sections for argon were with ones for BF3
taken from Biagi’s work [4].
2.2
PIC sheath module
The sheath model is used to determine the
sheath dynamics and ion energy distribution on the
wafer, including all the fragments previously
determined by the chemical kinetics routine.
The starting concentration of each species,
maintained constant at the unbiased side of the
sheath, is calculated by an analytic model describing
the expansion along a diverging magnetic field of an
uniform plasma [2]. This model assumes a collisionless environment with no radial diffusion and no ion
generation outside of the source.
3
Results and discussion
3.1
Global source module
The discharge is first implemented as stationary
with a continuous power deposition and neither
residence time nor gas flow are considered.
The electron density can range between
1.509x1011 and 6.217x1012 cm-3, the variation being
proportional to the absorbed power and to the square
root of pressure. The magnetic field helps to confine
the plasma, increasing the electron density by one
order of magnitude, and considerably reduces the
wall losses.
We observe that the plasma conditions
concerning interaction scale (Debye length of 3x10-5
m, much smaller than the source dimensions) and
charge separation (plasma frequency of 8x109 Hz,
larger than the electron frequency) allow a fluid
approach to be employed for further improvements
of the simulation code.
The plasma composition in the discharge
(figure 2) varies in three distinct phases :
1. Between 100 ns and 1 ms, a steady
fragmentation of the doping gas begins with a
relative density of each species increasing as a
function of the power of time. The resulting
plasma is electropositive with a majority of
electrons and BF2+ ions.
2. Between 1 ms and 10 ms, we observe a
temporary stabilization of the ion plasma
accompanied by a major change in the nature of
the plasma. Negative particles like F-, coupled
with BF2+, constitute now the majority of the
ions and the discharge becomes electronegative.
3. After 10 ms, the doping gas is subjected to an
intensive fragmentation into neutral atoms. The
plasma is electropositive again and composed of
atomic neutrals B/F and B+ ions.
For a continuous power deposition, the precursor
fragmentation depends of the gas flow rate. So a
strict control of this parameter is necessary to limit
the fluctuations of the residence time in the source,
as it can lead to great differences in plasma
composition. For a pulsed power deposition, the
fragmentation is controlled by the pulse duration if
the latter is shorter than the residence time.
Figure 3 : Evolution of the electron temperature in function of
time.
The figure 4 represents the repartition of the
power loss mechanisms in function of time. The
initial broad peak in collisional losses around 5ms
corresponds to the electronegative phase of the
discharge, when negative ions are created though
low-threshold dissociative attachment processes [5].
After 0.14 s of constant power deposition, the
dissipation mode switches from collisions to wall
losses because the available gas has been
significantly depleted and ion production inside the
plasma cannot compensate for the wall losses due to
ambipolar diffusion.
Figure 2 : Variation in time of the plasma ion composition.
Dotted lines correspond to neutrals, both boron-based (blue)
and fluoride-based (green), while thick lines correspond to ions
(same color conventions). Electrons are in red.
The variation of the electron temperature
(figure 3) follow three steps, directly linked to the
fragmentation phases previously determined :
1. First, germ electrons are created in an interval
lasting 650 ns that establishes the delay before
any significant electron-impact fragmentation
can occur.
2. This temperature peak at 15 eV is followed by
an electronic avalanche triggered by the
ionization of the doping gas, during the steady
fragmentation phase between 100ns and 1µs.
3. Finally the electron temperature is controlled by
the ionization kinetics, the small spike at 450 ms
being linked to the final phase of intensive
fragmentation into neutral atoms.
Figure 4 : Repartition of the power dissipation modes in
function of time.
3.2
PIC sheath module
The following figures correspond to the case
where a substrate bias of –500V is applied during
25µs with the rise time and fall time both at 100ns.
The ratio between the electron and ion
densities in the source and the same parameters
nearby the substrate is determined to be about 20.
This density ratio is directly related to the area ratio
between the chamber and the source because of the
way the equations of the flux conservation model are
established.
The figure 5 shows well that lighter ions
reaches the maximum value for implantation energy
sooner than the heavier ones. The time needed to
reach the full implantation energy increases with
substrate bias, from 2.5µs at 500V to 4.65µs at
5000V, with a higher bias resulting in a broader ion
implantation energy distribution.
4
Summary
Parametric studies have been done on this first
version of the complete model. The plasma chemical
composition and the ion energy distribution can
readily be obtained from the reactor tuning
parameters accessible to the operators. Further
efforts are still required to streamline the data
exchanges between modules and simplify use.
An ambipolar 2D fluid model [6] coupled to an
analytical ion implantation module [7] is currently
developed in order to replace the current global
model-PIC tandem.
5
Acknowledgements
I would like to thank the OSEO public holding for
providing the financial support to the OSEO-ISI
EQUIPE project from which this work is derived.
6
Figure 5 : Ion implantation energy at the beginning of the pulse,
in function of time. Boron-based ions are in blue, fluoride-based
ions in green.
The mean ion density (figure 6) is stable once
the sheath is full expanded after 2.5µs, as an
equilibrium is reached between the implantation ion
flux exiting the sheath and the ion flux entering from
the other side. When the pulse ends, the steady
increase in density can be explained by the boundary
conditions for the PIC model that result in a net ion
production in the simulation volume in absence of
implantation.
References
[1] Kaeppelin V. et al., “Characterization of an
industrial PIII reactor with Langmuir probe and
energy selective MS.” Surface and Coatings
Technology, 2002. 156: p. 119-124.
[2] Lieberman M. A. and Lichtenberg A. J. (2005)
“Principles of Plasma Discharges and Materials
Processing”, 2nd edition (New York: Wiley)
[3] G. LOMBARDI, K. HASSOUNI, et al. (2005).
"Numerical investigation of chemistry and transport
in H2/CH4 microwave plasmas under several
discharge conditions typical for diamond deposition"
J. Appl. Phys., 98 053303
[4] Biagi S., Cross sections extracted from the
program MAGBOLTZs in its version 7.1,
http://rjd.web.cern.ch/rjd/cgi-bin/cross
[5] Harland, P.W. and J.L. Franklin, “Partitioning of
excess energy in dissociative resonance capture
processes.” Journal of Chemical Physics, 1974.
61(5): p. 1621-1636.
Figure 6 : Global density for each ion in function of time.
Boron-based ions are in blue, fluoride-based ions in green,
electrons in red.
[6] Bose, D., T. R. Govindan, et al. (1999). "A
Continuum Model for the Inductively Coupled
Plasma Reactor in Semiconductor Processing."
Journal of The Electrochemical Society 146(7):
2705-2711