4

Plasma Reactors

4.1 PLASMA SYSTEMS

A block diagram of a cold plasma system used for material processing is presented
in Fig. 4-1. Generally, it consists of several subsystems providing different necess-
ary functions as following:
1. Gas handling system, which includes
a. Precursor gas supply. The source materials or precursors are in most
cases gases in high-pressure cylinders or liquids with sufficiently high
vapor pressures. Solids with reasonable vapor pressures are also used
sometimes as precursors.
b. Mass flow controllers. These are used to measure and control the flow
of the different gases fed to the reactor.
Power
Massftow supply
controllers . . ~~~~.,......-Reactor
r----e::::::J~h--.-_I
-<9 Pressure
gauge

Vacuum
pumps
Gas cylinders

Fig. 4-1 General outline of a system for cold plasma processing.

86
Sec. 4.1 Plasma Systems 87

c. Vacuum system comprising pumps and pressure controller. The plasma

reactors for materials processing operate at pressures between 10 - 4
and ~ 10 torr. However, lower background pressure is often required to
ensure the cleanliness of the process. Therefore, the whole range of vac-
uum pumps, from mechanical to cryogenic, are used; the types and sizes
of the pumps are determined by the required vacuum levels and gas flow
rates.
2. Plasma reactor. The plasma reactors will be described in detail in later
sections of this chapter.
3. Powers supplies. Different types of power supplies will be discussed
together with the corresponding reactors. The role of power supplies is
twofold: to sustain the plasma in the reactor and to provide a controlled
external substrate bias when such is needed for the process.
4. Safety devices for handling hazardous gases. Most of the precursors used
in plasma processing are hazardous, some even extremely so; these gases
can by corrosive, highly toxic, flammable; or explosive. Some gases, such
as the extensively used silane, are unpredictably pyrophoric. Germane, or
doping gases, such as diborane, phosphine, or arsine, are extremely toxic.
Some of the safety devices required when dealing with such hazardous
materials are
a. Flow limiters, mounted between the valve of the supply cylinder and
the pressure regulator to prevent excessive flow of the gas in case of
breakdown of the pressure regulator.
b. Flashback an-esters, required when using flammable or explosive gases
to prevent fire propagation to the gas in the cylinder; located in front of
the pressure regulator.
c. Cross-purge assemblies, used to purge the regulators and prevent re-
lease of the hazardous gas to atmosphere when exchanging the gas cylin-
ders; they are mounted between the cylinder and the pressure regulator.
d. Scrubbers or diluters of exhaust from the pumps.
e. Detectors for hazardous gases.

4.1.1 Process Parameters

For a given system, the outcome of the process is strongly dependent on its
parameters. A comprehensive list of the process parameters that can affect the end
result of plasma processing is given in Table 4-1 [1]. The classification of these
parameters largely complies with the first three subsystems described before. The
following list names the most important parameters among those in Table 4-1:

• Partial pressures of the feed gases, or flow rates of the different gases
• Total pressure in reactor
• Substrate temperature and bias
88 Plasma Reactors Chap. 4

TABLE 4-1 Parameters Controlling Materials Processing by Cold Plasmas

(adapted after [2], reprinted with permission)

Plasma Processing Parameters

Kinetic Electrical Surface

(gas system) (plasma system) (substrate system)

Precursor gases Frequency (DC to GHz) Material

Carrier gases Free fall Conducting
Mass flow rates Mobility Insulating
Pressure Diffusion Temperature
Gas delivery location Electrode geometry Position
Blectrodeless
Electrode
Discharge power
Field strength
Current density
Particle energy
Active neutrals
VIS to UV radiation
Electrode material

• Reactor geometry and material

• Electrode material and distance between electrodes
• Electric power applied to the plasma
These parameters have to be carefully controlled to define the plasma chemistry
and to achieve the desired results.
The complexity of the relationship between the macroscopic variables of a
cold plasma processing system and the microscopic plasma parameters is illus-
trated in Fig. 4-2. Changes in macroscopic plasma variables such as composition
of feed gas, flow rates, pumping speed, electrical discharge power, and frequency
will generally change the basic plasma conditions, but the precise manner of these
changes is in most cases unknown.
The RF power determines the current and voltage between the electrodes.
The composition of the gas mixture can affect the chemical reactions inside the
plasma reactor and the properties of the final product, for example the composition
of deposited film or the anisotropy and selectivity of etching. The composition of
the gas feed, the total pressure in the reactor, and the electrical power will each
affect the rates of the reactions. The reaction rates can be affected also by the
distance between electrodes and by substrate bias or temperature. These last two
plasma parameters can also affect the structure of a deposited film or the etching
Sec. 4.1 Plasma Systems 89

Kinetic Parameters Electrical Parameters

Precursors & diluents
Power
Flow rates
Frequency
Pumping speed
Electrodegeometry
Pressure

Plasma Parameters
Densities (charged and neutral particles)
Temperatures
Residence time

Surface Parameters
(substrate/electrode)
Material (conductingllnsulating)
Bias
Temperature

Fig. 4-2 Complexity of interaction between plasma variables (after [3], reprinted
with permission). The arrows indicate the possible interactions between pro­
cess parameters; they do not indicate that all the described interactions
always ta1ce place.

anisotropy. The flow rates, total pressure, and reactor geometry can determine the
uniformity of the process.
The gas flow rate, the pumping speed, and the pressure are interrelated. The
pressure can be changed in two ways: by changing the flow rate at constant
pumping speed or by changing the pumping speed at constant flow rate. While
the two methods can provide the same pressure, they result in different residence
times for the species in the reactor (see Sec. 4.1.2). The different residence times
can cause changes in the chemistry of the process [4].
The frequency of the electric field is also a system parameter that can affect
the plasma process or the properties of the product. A variation of the frequency
can affect the number and energy of ions that can follow the alternating field, thus
changing the flux and energy of the particles bombarding the plasma treated
surface.
90 Plasma Reactors Chap. 4

The process parameters must therefore be precisely controlled to achieve the

required results and process uniformity.
And last, but not least, the building materials of the reactors and electrodes
are an important factor to be considered for a specific process. Electrode and
reactor materials can change the chemistry occurring in glow discharge. The
materials should be chemically inert to the precursor gases and to the plasma
products. Because surfaces in contact with a plasma are always negative relative
to it, the electrodes and reactor walls will almost always be sputtered to a certain
degree. The sputtered atoms may contaminate the plasma and the processed
sample. The sputtered atoms may be incorporated in a deposited film or may mask
areas of an etched surface causing nonuniform etching. Such possible effects
should be taken into account when designing a reactor.

4.1.2 Scale-up

The transfer of a plasma process from a research laboratory apparatus to

commercial manufacturing is called scale-up. The scale-up requires the upgrading
of the working laboratory process to a technology that can process uniformly
larger samples in commercial quantities. The scale-up should take into account
economical considerations besides overcoming technological difficulties, some of
which will be discussed in the following paragraphs. Because commercial scaled-
up reactors are very expensive, it is of utmost importance to make a successful
transfer from a small-scale process into a large one.
As a plasma process can operate in a large parameter-space and due to the
complexity of interaction between the parameters (see Fig. 4-2), it is important
to understand, at least to first approximation, how to adjust the plasma parameters
according to changes in the size of the reactor.
The substrate temperature will normally be kept unchanged during the trans-
fer of the process from a small to a large reactor. The main question is how to
adjust gas flow rates, pressure, and electrical power in such a way that the scale-up
of the reactor will not significantly affect the plasma chemistry. For this purpose,
plasma processing can be conceptualized as being constituted of two subsequent
processes [5]:

1. Formation of reactive species

2. Mass transport of the reactive species to the processed surface by convec-
tive diffusion
For the first process, it has been shown theoretically that, if the dissociation
of the precursor molecules takes place through single collisions, the rate coeffi-
cients for the reactions caused by electrons depend only on the value of the ratio
E/p, because the electron temperature in the plasma is determined mainly by E/p
(E is electric field strength). One may therefore consider as a first approximation
the fact that molecular dissociation rates caused by electron impact are dependent
only on the E/p ratio. However, if dissociation is produced through multiple
Sec. 4.1 Plasma Systems 91

collisions, the dependency on E and p is more complex. It is therefore usually not

sufficient to keep only the E / p ratio constant when scaling-up the process.
Because the product D· p (D being the diffusion coefficient) is constant for
a gas, the mass transport through diffusion is also pressure dependent. Therefore,
to preserve both the plasma chemistry and the mass transport during scale-up, one
has to keep both p and E unchanged during scale-up. To maintain the same
electrical field strength across the discharge over the larger area of the manufac-
turing reactor, it is required to increase the total current or power supplied to the
discharge. In a parallel plate reactor, where the plasma is usually confined between
the electrodes, the power has to be scaled linearly with the electrode area to keep
the electric field constant. In other types of reactors, especially in. tubular ones,
the plasma volume does not coincide with the volume between electrodes and is
affected by both pressure and flow rate. The power of the scaled-up reactor has
to be therefore readjusted, often experimentally.
With the electric field, pressure, and substrate temperature kept constant
during scale-up, the only external adjustable parameter left is the gas flow rate.
The average gas flow velocity determines the convective mass transport of the
reactive species to the substrate. The flow rate should be scaled up in such a way
as to keep the average flow velocity identical in small and large reactors. It can
be shown that this requirement regarding the flow velocity scales the gas flow
rates proportional to the cross-sectional areas of the reaction zones, perpendicular
to the gas flow direction [5].
A different approach is to keep the same average gas residency time, t , in
both large and small reactors. The residency time is defined as
pVol,
r, = -Q- (4.1)

where V01, = volume of reaction zone

Q = total mass flow rate
The gas residency time is a measure of the average distance over which the
reactive species diffuse in the reaction zone, and thus determines the diffusional
mass transport. According to Eq. (4.1), if the pressure is kept constant, the
requirement of constant residence time scales the gas flow rates proportional to
the volumes of the reaction zone. If the interelectrode distances are the same in
the large and small parallel plate reactors, the requirements for the same gas
velocity and same gas residency time are equivalent. In other types of reactors,
the volume of the reaction zone can be dependent on the gas flow rate, and this
approach cannot be used.
In summary, the main process parameters have to be adjusted during scale-up
to a first approximation as following:

sample temperature unchanged

pressure unchanged
electric power proportional to area of electrodes in parallel plate reactors
92 Plasma Reactors Chap. 4

gas flow rates proportional to electrode area normal to the flow direction,
or proportional to volume of reaction zone.

It thus appears that the main changes in plasma parameters during scale-up
are the linear extrapolations of electric power and gas flow rates. However, these
changes alone may not be sufficient. In practice, a change in reactor geometry
may affect other important intrinsic plasma parameters, especially if the reactors
are not of the parallel plate type.
The foregoing description of scale-up principles did not take into consider-
ation how the active species are formed along the flow path of the gas from the
gas inlet to the treated substrate and their interaction before reaching the substrate.
The formation and transport of the reactive species that form the final product are
dependent on the flow patterns and on the time spent along the path to the
substrate. The diffusion of the precursor molecules takes place along most of the
distance from the gas inlet to the substrate, while diffusion of reactive species
takes place mainly from the boundary of the plasma zone to the substrate. The
reactor geometry and the geometry of the electrodes (size, distance between them)
can change the gas flow patterns, the relative volume of the plasma region, the
transport of the reactive species, and the rate and uniformity of the process.
Electric field patterns, especially at the edges, are also influenced by the
reactor geometry and may affect the uniformity of the process at the edges of the
substrate. This is especially true for tubular reactors or for bell jar reactors, often
used in plasma polymerization [6]. In such reactors the volume of the plasma is
not defined by the electrodes and is dependent on the interaction between different
plasma parameters. Finally, the wall temperature of the scaled-up reactor is gen-
erally different from that in the smaller one, causing different gas retention or
memory effects.
Because of these effects, the approximated parameters of the. scaled-up reac-
tor have to be generally readjusted either empirically or using theoretical
modelling of the process and reactor. The last approach would generally still need
empirical fine-tuning of the process parameters.
The ease with which a reactor can be cleaned has to be given careful
consideration during its design. This is especially important for deposition reac-
tors. Film material that deposits and accumulates on the surroundings of the
deposition zone and on the walls of the reactor has to be regularly removed from
any deposition system to avoid flaking and powder formation that may have
adverse effects on the quality of the film deposited on the substrates.
Another important aspect to be considered when scaling up a laboratory
system to a commercial manufacturing unit is the efficiency of precursor utiliz-
ation, or the material efficiency of the process. In a deposition process, the
material efficiency can be defined as the ratio between the weight of the deposited
film and the weight of the film components in the feed gas. The material effi-
ciency can be defined similarly for an etching process, in this case taking into
account the removed instead of the deposited material. The highest material
Sec. 4.3 RF Reactors 93

efficiency is generally obtained at high plasma density and low flow rates and, at
a first approximation, is independent of pressure for the processes described in
this book (based on gaseous precursors). The pressure, however, affects the
process efficiency in sputtering processes, where the sputtering yield is pressure
dependent. Because of the condition of low flow rate, high material efficiency is
usually associated with lower process rates in deposition and' etching. As a result,
high material efficiency does not always imply maximum process efficiency,
which demands high process rates [5].
The optimum trade-off between reactor design, material efficiency, and pro-
cess rate is determined by each individual manufacturing process. The develop-
ment of a manufacturing process is finally to a large extent empirical in defining
the optimal conditions for achieving either film deposition or etching rate uni-
formity.
Due to the diversity of plasma reactor designs and plasma parameters, it is
not within the scope of this chapter to describe in detail all designs and experi-
mental approaches used. Only the general principles of the most common used
reactor designs for producing the plasmas, based on the methods discussed in
Chapter 2, are described in the following.

4.2 DC REACTORS
DC plasma reactors operate as described in Sec. 2.1. The plasma is sustained
between two parallel plate electrodes as illustrated in Fig. 2-1, and the electric
power is supplied to excite the plasma in the abnormal discharge mode. The
distance between the electrodes and the pressure in the reactor have to satisfy
Paschen's law for the used gas mixture (see Sec. 2.1.1).
The power supplies used to sustain DC plasmas can generally be used at
constant voltage, constant current, or constant power mode. The power supplies
should be able to control the preset value in each mode. The excitation and
sustainment of a DC plasma require the use of electrically conductive electrodes
and samples. However, electrically insulating films may deposit locally on the
surface of the electrodes or samples due to contaminants formed in the plasma.
In such cases, local dielectric breakdown normally occurs, causing arcing that
results in spikes of high currents.
The power supply must be able to withstand these spikes and return to normal
operation without shutting off. However, it must also be able to distinguish
between the transient arcing currents and shorts that may occur between the
electrodes and grounded surfaces. This feature, called arc suppression, must be a
characteristic of a DC power supply of a plasma reactor.

4.3 RF REACTORS
The power supplies used to sustain commercial RF plasma systems are operated
at specific frequencies established by international agreements [7]. The most used
94 Plasma Reactors Chap. 4

frequency is 13.56 MHz, but sometimes the higher harmonics of 27.12 MHz and
40.68 MHz are also used. The RF generators are designed to operate at constant
output impedance of 50 O. The impedance of the glow discharge is, however,
much higher and varies with the process parameters. An impedance matching
network or matching unit is thus required between the RF power supply and the
reactor. Otherwise, most of the power will reflect to the power supply instead of
being absorbed into the plasma.
The matching unit is placed physically as close as possible to the powered
electrode. In parallel plate reactors the matching unit is usually mounted on the
flange supporting the powered electrode. Normally, variable automatic matching
units are used, and these adjust themselves to keep the output impedance of the
RF power supply constant at 50 0, independent from the changing plasma con-
ditions.

4.3.1 Electrodeless Discharges

When using high-frequency power supplies, it is possible to transfer the

energy to the plasma without inserting the electrodes inside the reactor, if a
suitable coupling between the power supply and the reactor is used. For such a
design, the reactor has to be built from dielectric materials such as quartz or Pyrex,
which are generally used for this purpose. Since the electrodes are not in direct
contact with the discharge, the plasma formed is called an electrodeless discharge.
Electrode designs for transferring the power to the plasma in electrodeless
discharges are illustrated in Fig. 4-3. These electrodes correspond to two types of
coupling the RF power to the electrodeless plasmas:

1. Inductive coupling, in which a coil connected to the power supply through

a matching unit is wound around a tubular reactor, as shown in Fig. 4-3
(a).
2. Capacitive coupling, in which the RF power is transferred between two
separated electrodes mounted outside the reactor. The electrodes can be
in the shape of two rings mounted around the reactor as shown in Fig. 4-3
(b) or two plates placed along the reactor as shown in Fig. 4-3 (c).

Although the terms of inductive and capacitive coupling are often used as just
defined, the inductive coupling is not purely inductive, and it always has a
capacitive component too, through the wall of the reactor. Thus, when an induc-
tive coupling is used, deposition on the walls is often observed to follow a pattern
matching the shape of the coil. This is an indication of localized stronger electric
fields on the walls, showing that the coupling is at least partly capacitive through
the walls of the reactor [2 l.
In a tubular reactor of one of the types shown in Fig. 4-3, and especially in
Fig. 4-3(a), the strength of the electric field is not uniform along the reactor.
Therefore the substrate position inside the reactor becomes a parameter that can
Sec. 4.3 RF Reactors 95

& ;o (la
RF

RF

of b
RF

(b)
~(C)
Fig. 4-3 RF coupling methods to electrodeless reactors.

affect the result of the process and can be controlled to achieve desirable proper-
ties [8, 9]. In tubular reactors the substrates are usually positioned with their
surfaces perpendicular to the axis of the reactor.
The types of RF reactors just described, also called barrel reactors, are used
sometimes for industrial applications, although they are more popular in research
laboratories, being relatively inexpensive and easy to set up. In industrial appli-
cations, the tubular reactors are normally used for ashing of photoresist or etching
in microelectronic fabrications and for noncritical deposition processes where
uniformity is of no great concern.
A barrel reactor used for etching often includes an etch tunnel. The metallic
perforated cylindrical etch tunnel, illustrated in Fig. 4-4, is inserted in the reactor

Gases in Gases in 7
/ Etch -
~DDD c..._
Quart z /4~~'"

reactor tunnel

Subs trates

Fig . 4-4 Diagram of a barrel reactor.

to improve temperature uniformity along the length of the sample holder and to
minimize particle bombardment of the substrates. This metallic cylinder acts as a
Faraday cage and confines the glow region to the annulus between the etch tunnel
and the wall of the reactor. Substrates are thereby shielded from direct contact
with the plasma and are subject to little, if any, ion or electron bombardment.
Neutral species diffuse, nevertheless, through the perforations and reach the sur-
face of the processed substrates.
96 Plasma Reactors Chap. 4

4.3.2 RF Discharges with Electrodes

4.3.2.1 Diode Reactors
The most common approach in the design of industrial RF plasma reactors
is to couple the RF power to two parallel electrodes inserted inside the reactor.
The coupling is done through special RF vacuum feedthroughs. One electrode is
often grounded together with the walls of the reactor, as shown schematically in
Fig. 4-5. This arrangement is also called the diode or parallel plate reactor.

Sho werhead

\
e /
Insulated
feedthrough
-!-

---
\.
-.. Elec trodes
/

/ Su bstrate

--
Fig. 4-5 Power coupling to a planar diode reactor.

For RF discharges, the surface of the electrodes in contact with the plasma
are no longer required to be electrically conductive since RF voltages can be
coupled through any kind of impedance. The surface of the electrode, substrate,
or deposited coating can be electrically insulating. The RF power is, nevertheless,
supplied in all cases through a conductive backing to the electrodes, to ensure
uniform electric potential across the electrode.
Although the discharge is mainly confined between the electrodes, plasma
species also diffuse to the side walls of the reactor. The interaction between
plasma and electrodes can result in contamination of the plasma with atoms, which
are either physically sputtered or chemically etched from the electrodes. The same
is true, though to a lesser extent, for contamination with atoms from the walls.
The right choice of materials for electrodes and reactor is therefore of utmost
importance, especially in microelectronic processing, for which prevention of
contamination is imperative. The electrodes are usually made of stainless steel and
the reactors are built from stainless steel, aluminum, or quartz.
Capacitive coupling to parallel electrodes inside the reactor enables the cre-
ation of uniform electric fields, making it possible to achieve high process uni-
fonnity over large areas, determined by the size of the electrodes. However, the
uniformity of the electric field alone is not sufficient to guarantee the unifonnity
of the results of PECVD processing when reactive gases are involved. Depending
on their concentration and consumption and on their flow pattern in the reactor,
Sec. 4.3 RF Reactors

the gases can be depleted of the reactive components along their path. This
depletion can cause nonuniform deposition or nonuniform etching across the
surface of the substrate.
The radial flow reactor, originally developed by Reinberg [10, II], was
designed to allow compensation for the depletion of the gases from the reactive
components, thus improving process uniformity. The precursor gases were intro-
duced at the periphery of the sample electrode and then flowed inward where they
were exhausted at the center as shown schematically in Fig. 4-6. The discharge
intensity is highest in the central region, promoting faster depositions in that area:

Electrodes

_ Substrate

To pump

Ag. 4-6 Diagram of a Reinberg reactor.

The higher velocity and consequent shorter residence time of the gases at the
center of the electrodes can counterbalance the higher discharge intensity in the
center, leading in many cases to a uniform deposition rate across the whole radius.
A reactor similar to the Reinberg reactor used for PECVD processing is
shown in Fig. 4-7 [12]. In this case, however, the gas enters through the center
of the lower electrode and flows out at the periphery. Another often-used approach
for achieving uniformity over a large area is by feeding the gas through a
showerhead that forms one of the electrodes in the parallel plate reactors. The
sample electrode can sometimes be heated or rotated to smooth out nonunifonnity
in processing.

4.3.2.2 Triode Reactors

If both electrodes in a parallel plate RF reactor are insulated from the walls,
the reactor is called a triode-type reactor. In the triode reactor, each electrode and
the reactor walls can be powered independently, and each can be biased, floating,
or grounded. A diagram of a triode reactor is shown in Fig. 4-8.
Another version of the triode reactor is the one in which the discharge is
sustained between two electrodes and the substrate is supported on a third elec-
98 Plasma Reactors Chap. 4

Shielded
RF Power input

!
Electrode

~~a·K·~~
""I~_I
Heater
11....---_
Heater

1 Rotating shaft
Out to Out to
VAC pump VACpump
Magnetic ~WUJ~
rotation
drive

1
Gases in

Fig. 4-7 Diagram of a radial flow reactor (from [ 12], reprinted with permission from
Thin Solid Films, vol. 113, p. 135, 1984).

trode, which can be biased separately [ 13 ]. The electrodes in a triode reactor can
be powered by combinations of power supplies, such as RF and microwave, RF
and DC, RF and AC ( < 50 kHz). While in the diode reactors, the RF power
density may be limited by the maximal bias allowable on the substrate electrode,
this limitation is eliminated in the triode reactors. The plasma density is controlled

Insulation

Substrate

AC

Fig. 4-8 Power coupling to a planar triode reactor.

Sec. 4.3 RF Reactors 99

by the power sustaining the plasma, while the substrate bias and thereby the
energy of the ions bombarding it is controlled directly by the generator supplying
the power to the substrate electrode.
Sometimes the discharge in a planar reactor may be assisted by thermionic
electrons emitted from a hot filament; this type of reactor is also called a triode
reactor.

4.3.3 High-Density RF Reactors

While different variations of the RF reactors described previously are widely

used in manufacturing, efforts are made to develop high-density RF plasma reac-
tors for achieving enhanced process rates. Two such high-density RF plasma
sources are the helical resonator and helicon or "whistler" source [ 14]. In the
helical resonator reactor shown in Fig. 4-9, the plasma is excited by a slow wave
structure that achieves quarter- or half-wave resonance in the source [14]. The
helicon source uses magnetic fields to obtain whistler mode waves that are
launched into the reactor chamber. The helicon reactor is illustrated in Fig. 4-10.
The helicon plasma reactor is composed of a glass plasma source, to which
electrical power is coupled inductively, and a metallic reaction chamber. The
helicon source comprises an external RF antenna consisting of two loops diametri-
cally placed on the outside of the glass tube and an electromagnet that produces
a constant axial field. This combination of the RF field with the magnetic field
excites a helicon wave in the source tube. Energy transferred from the helicon
wave sustains the plasma. The axial magnetic field restricts the radial motion of
the electrons and confines their movement along the magnetic field lines. If the

Gas inlet

RF power

Processing
chamber

Fig. 4-9 Diagram of plasma reactor with

helical source (from [14], re-
printed with permission).
100 Plasma Reactors Chap. 4

External
antenna

\ 13.56 MHz

Source solenoid

Chamber
solenoid

I"'--.. . . ..--S-.Ubstrate tabr-'e 1

Insulator ~ Vacuum
pumps

Substrate ---. c:::=~

water cooling

Fig. 4-10 High-density plasma reactor with helicon source (from [ 15], reprinted with
permission).

magnetic field is strong enough, such that the ion cyclotron radius is smaller than
the radius of the tube, it can also confine the ions by reducing their loss to the
walls. For a tube 15 em in diameter, a magnetic field of ~ 100 G can achieve
the confinement.
The helicon source is operated in the pressure range 10- 4 -10 - 2 torr, and
the plasma diffuses from the source into the reaction chamber. The diffusion is
controlled with additional magnets placed around the chamber. These can be
permanent magnets in a multipolar configuration or electromagnets [ 15 ]. A reac-
tion chamber 30 em long and 30 em in diameter can be operated with a helicon
source 22 em long and 15 em in diameter. The helicon source can be operated
over a large frequency range, the range 2-70 MHz appearing to be most suit-
able [ 15].
The RF reactors in the different configurations are the most versatile reactors
and are most used for plasma processing. The RF reactors can process electrically
conductive or insulating samples and are used for deposition of inorganic or
Sec. 4.4 Microwave Reactors 101

polymeric films, for surface treatment of organic or inorganic solids, and for
plasma and reactive ion etching.

4.4 MICROWAVE REACTORS

The microwave plasmas are sustained by coupling the microwave energy to the
plasma gas through applicators of various designs. Microwaves are easily absorbed
or reflected by most materials and cannot be transmitted via cables, like RF
energy, without significant losses. Special coaxial cables can be used to transmit
low-power ( < 200 W) microwaves. The coax consists of two concentric conduc-
tors separated by a dielectric. For higher-power levels, specially designed wave-
guides have to be used. The waveguides are hollow rectangular tubes of high
electrical conductivity. The cross sections of the conductors of the coax or the
waveguide are determined by the wavelength of the microwave [16].
A microwave power supply system used for microwave or ECR plasma
generation is illustrated in Fig. 4-11 and consists of
1. A filtered, low-ripple microwave power supply of constant frequency but
variable power.
2. A circulator, whose role is to protect the power supply from large reflected
power which may result from an impedance unmatched microwave appli-
cator.
3. Meters for monitoring both the incident and reflected power.
4. A variable, manual or automatic, impedance matching of the applicator
with plasma, at different and variable discharge conditions; an impedance
mismatch will cause the microwave to reflect instead of propagate into
the plasma.
Incident power
meter
Dlro<:l ional Applicator
, cou~
I per Plasma
,}-+ +-{;3--U:::3
,
<,
Reflected power
meter
Dummy load
(cooled)

Fig. 4-11 Diagram of microwave power system for cold plasma excitation .

The microwave applicators can be of several types: waveguide, cavity, and

coaxial.
The simplest way to transfer microwave power to the plasma is by inserting
a dielectric tube reactor through a rectangular waveguide at the center of its wide
face, as illustrated in Fig. 4-12. The axis of the reactor coincides with the position
102 Plasma Reactors Chap. 4

Quartz tube

MW cut-off shield

Fig. 4-12 Waveguide-tube microwave

coupling.

of the highest electric field in the waveguide, facilitating the acceleration of the
electrons by the microwave. To prevent leakage of microwave outside the reactor,
the discharge region is enclosed on both sides by metallic tubes with diameters
smaller than the cut-off diameter corresponding to the wavelength of the used
microwave [ 16 ].
The diameter of the reactor is limited in this case by the size of the wave-
guide, whose width for a frequency of 2.45 GHz is 7.21 em.
Fig. 4-13 illustrates a different way of microwave coupling to the plasma by
a quarter-wave microwave cavity or an Emerson antenna. It is essentially a
resonance cavity that is fed by a coaxial cable and is suitable for low powers
( < 200 W). The cavity has a removable cap that allows easy disconnection of the

... Tuning stub

1A
Discharge tube o,
Section B-8
,
1.0·
Scale

Fig. 4-13 Coaxial 1/4 cavity (from [17], reprinted with permission from Plasma
Chemistry in Electrical Discharges, p. 65, 1967).

reactor from the cavity. For microwaves of a frequency of 2.45 GHz, the diameter
of the reactor is generally limited to less than 2 em.
Microwave plasmas of larger volumes than those attainable by direct coupling
through the waveguide (Fig. 4-12) can be produced by coupling the microwaves
through a larger cavity. Special axisymmetric microwave couplers can convert the
Sec. 4.4 Microwave Reactors 103

microwaves from the rectangular waveguides to circular applicators. In a circular

applicator the plasma is symmetrical and can have a nearly uniform radial electric
field profile. An example of an axisymmetric applicator that produces plasmas for
processing wafers 3-4 inches in diameter [ 18] is shown in Fig. 4-14.

Axisymmetric coupler
Waveguide
•

+-Cavity
Quartz bell ja

.,...,I-~--Plasma

Substrate

Fig. 4-14
I ~aust
An axisymmetric microwave coupler (from [18]. reprinted with permission
from Research & Development Magazine. October. 1989. by Cabners Pub-
lishing Company).

Microwave reactors have been designed for plasma treatment of even larger
areas. A reactor designed for large-area etching of polymers [19, 20] combines
a long and narrow microwave applicator (42 x 7.5 em") with a translational
movement of the substrate, in a direction perpendicular to the applicator.

Water
!
External
coax

Flange to
Fig. 4-15 Coaxial microwave coupling to
reactor
a plasma reactor.
104 Plasma Reactors Chap. 4

Microwave power can be coupled into a plasma also by using coaxial appli-
cators. As illustrated in Fig. 4-15, the central conductor of the coax is in direct
contact with the plasma and has to be generally water cooled. The insulator that
separates the central conductor from the external one has also to provide a vacuum
seal at the insertion in the reactor.

4.5 ECR REACTORS

Several configurations have been developed for ECR sources for cold plasma
reactors. A typical system of the divergent magnetic field type [21] is illustrated
in Fig. 4-16. In this system, microwave power is fed by waveguide into the ECR
cavity, through a quartz or alumina window. Solenoidal electromagnets (see Fig.
4-16) create a magnetic field that reaches the ECR value of 875 gauss in the
source and diverges outside the source. Electrons diffuse faster than ions into the

2O ....-- - -r"\

10
:[
~ 0 I----t'----t

! · 10

-20 '--~---'
o0.2 0.4 0.6 0 .8 1.0
Magnetic Flu. ( kG I

--. DiAs

(a) (b)

Fig. 4-16 Diverging field-type ECR reactor: (a) plasma source and reactor; (b) dis-
tribution of magnetic field along the plasma.

processing chamber, along the diverging magnetic field lines, creating an electric
field which causes the extraction of the ions. A potential of 15-20 eV is com-
monly developed and is imparted to the ions extracted from the source. If higher
ion energy is needed, it can be achieved by placing extracting grids between the
ECR source and the processing chamber. The grids can also improve the uniform-
ity of the extracted plasma beam.
Additional magnets are sometimes added outside the source to permit shaping
of the magnetic field in the vicinity of the substrate and to achieve better control
and confinement of the plasma. The auxiliary magnets can be located behind the
substrate holder to improve the uniformity of ion current density across the
substrate and to ensure the arrival of ions perpendicular to the surface.
Sec. 4.5 ECR Reactors 105

The normal ion incidence on the substrate is especially important in etching

where ion bombardment contributes to etching anisotropy. In film deposition, the
normal arrival of the ion on the coated surface is essential for preventing void
formation when filling features of high-aspect ratio [22]. If additional substrate
bias is required, it can be achieved by connecting a RF power supply to the sample
holder as indicated in Fig. 4-16.
Another type of ECR plasma system is the multipolar ECR, or distributed
ECR, reactor [23] illustrated in Fig. 4-17. In this system, the microwave power
is fed through several internally cooled microwave antenna rods, located inside
the periphery of the reactor and close to its wall. Permanent samarium cobalt
magnets positioned outside the reactor and opposite the antennas create resonant
field cusps next to the antennas, as shown in Fig. 4-18. In this type of reactor the
plasma diffuses out of the ECR regions, creating a uniform plasma inside the
reactor. The internal plasma region is practically free of magnetic fields. The
configuration of the magnetic field also slows the loss of particles to the wall and
helps maintain a high plasma density. The type of reactor described here is also
named multicusp geometry reactor.
106 Plasma Reactors Chap. 4

Uniform plasma

Fig. 4-18 ECR region in multipolar ECR

MWan tonnas reactor.

In the diverging field ECR plasma system shown in Fig. 4-16, the ions flow
along the magnetic field lines, making it a line-of-sight process, while in the
multipolar ECR the plasma is essentially isotropic in the internal processing region
of the reactor.
For some applications, the substrates have to be treated while having a linear
translational movement through the plasma. This requires a linear plasma source
having a uniform density perpendicular to the direction of the substrate movement.
A special ECR source using a hom antenna expanding to an aperture of 60 x 9
em? has been developed for such applications [24]. The source is called by the
authors an elongated-mirror-geometry (EMG) source.
Because of the short wavelength of the microwaves, both regular microwave
and ECR plasmas can be confined to small volumes. As a result, processing in
microwave I ECR plasmas can be performed in several reactor arrangements:

• The substrate is placed entirely inside the microwave I ECR discharge .

• The substrate is placed in a processing chamber, located downstream from
the discharge.
• In a hybrid arrangement, the substrate is placed downstream from the
microwave discharge, with an additional RF or DC discharge sustained
adjacent to the substrate.

Processing performed using the last two configurations is also called processing
by remote plasma, which will be discussed later.
The ECR plasmas, characterized by high degrees of ionization, can produce
high-density plasmas over a broad range of pressures. The energies of the ions are
intrinsically low in the ECR plasmas, but can be increased using extraction grids.
Because of these features, ECR reactors produce large fluxes of low-energy ions,
are characterized by high processing rates, and are therefore used for single-wafer
processing, for etching of wafers, or for deposition of coatings. The large mean
free path of the particles at the low operating pressures of the ECR plasmas results
Sec. 4.6 Magnetically Enhanced Reactors 107

in directionality of the ion and neutral beams. When used for anisotropic etching ,
ECR reactors cause lower wafer damage than reactive ion etching performed in
RF reactors at much higher ion energies.

4.6 MAGNETICALLY ENHANCED REACTORS

The high-density plasmas obtained in the ECR reactors are a result of the inter-
action between the electric field at microwave frequencies and the superposed
magnetic field. The value of the magnetic field has to be adjusted to create
resonance between the frequency of the electric field and the electron cyclotron
frequency. Magnetic fields can be used to enhance plasma densities also at
nonresonant conditions. Two types of magnetically enhanced plasma reactors are
described in the following .
The magnetically confined reactor (MeR), is essentially a triode reactor that
uses additional magnetic fields to confine the plasma and to reduce the loss of

Top View

Side View

FIg. 4-19 Magnet arrangement in a magnetically confined reactor. The arrows indicate
the magnetic field lines (from [25], reprinted with permission) .
108 Plasma Reactors Chap. 4

charged species to the walls. The triode reactor, similar to the one shown in Fig.
4-8, uses two power supplies at different frequencies: one for exciting the plasma
and one for biasing the substrate electrode. A 13.56-MHz power supply is used
to sustain the plasma, while a 100-kHz power supply provides the bias to the
substrate. The magnetic confmement is produced by permanent magnets arranged
around the perimeter of the reactor and imbedded in the grounded upper electrode,
as illustrated in Fig. 4-19.
The power supplied by the high-frequency field controls the amount of
dissociation and generation of reactive species. The low-frequency field controls
the ion flux to the substrate and enhances the directionality of the process,
especially in etching.
The confmement of the plasma by the magnetic fields results in an enhanced
plasma density. The arrangement of the magnets keeps the magnetic filed lines
away from the processed wafer (see Fig. 4-19), preventing a direct influence of
the magnetic field on the process. The MCR is used at pressures of a few mtorr
[25, 26]. Being based on RF power supplies and using a magnetic field that is
independent from the frequency of the electric field, the equipment used for the
MeR is much simpler than that required for ECR reactors.
A different version of a magnetically enhanced plasma reactor used especially
for magnetically enhanced reactive ion etching (MERlE), is the rotating field
reactor. This is essentially a parallel plate RF reactor in which a time-varying
magnetic field is applied on the substrate electrode. As shown in Fig. 4-20, several
electromagnets are arranged around the substrate electrode, which in most cases
serves as the cathode of the discharge. The plasma is sustained by a 13.56-MHz
generator. The electromagnets produce a magnetic field that rotates slowly, at a

Magnetic field rotation

Wafer
(clamped
and cooled)

cathode

Fig. 4-10 Cathode and electromagnets in a rotating field reactor (from [27 ], reprinted
with permission).
Sec. 4.7 Remote PECVD Reactors 109

frequency of 0.5 Hz, in the plane of the electrode. This arrangement produces a
high flux of low-energy ions, resulting in high processing rates with reduced
substrate damage [27]. Both MCR and rotating field reactors are considered
alternatives to ECR reactors, especially for plasma etching.

4.7 REMOTE PECVD REACTORS

In the reactors previously discussed, the processed samples are inserted directly
in the plasma excited in the feed gases. The processing in such reactors is
sometimes called processing by direct PECVD. A different type of cold plasma
processing is by remote PECVD [28]. The remote PECVD is also called down-
stream processing.
The distinction between direct and remote PECVD is made primarily on the
basis of whether reactant or diluent gases are directly excited by the plasma. In
the direct processes described before, all process gases are exposed to the plasma.
In addition, the processed samples are immersed in the plasma discharge.
By contrast, in the remote PECVD processes, not all reactant gases are
excited in the plasma, and the substrate is removed from the discharge region of
the plasma. The gases can be introduced into both the discharge region and the
reaction chamber outside the discharge, often near the substrates. This arrange-
ment reduces the number of possible reaction pathways and can improve control
of the process or the stoichiometry of deposited films [28]. Physical effects of
the plasma on the processed substrate can be reduced in the remote reactors,
minimizing or completely eliminating radiation damage.
A remote plasma reactor is illustrated in Fig. 4-21. In such a system some
of the process gases (dilutants and/or reactants) are excited in the plasma sus-
tained in a quartz tube by inductive RF coupling, while other reactant gases are
introduced in the processing chamber through a dispersion ring. The processed
substrate is supported outside the plasma.
Gases In <,

/ Quartz tube

RFcoll
o
o
~ /GaSeS ln
o
o o
Proces s
Dispersion -M-- - -(( chamber
ring

-- Substrate

Fig. 4-21 Reactor chamber and RF plasma

""-To pump
source for remote PECVD.
110 Plasma Reactors Chap. 4

Hybrid systems combine characteristics of both direct and remote processes.

This is the case in some RF triode-type reactors in which the substrate sits below
a grid that replaces one of the electrodes and it is not exposed directly to the
discharge. A regular ECR reactor can also be considered a hybrid processing
system because the substrate is generally outside the plasma chamber while all
the gases may be excited in the plasma.
An ECR reactor can also be used as a remote PECVD reactor. The gases can
be introduced both in the ECR plasma chamber and in the processing chamber,
and as mentioned earlier, the substrate can be placed outside the plasma. For
example, to deposit silicon dioxide by the remote process, oxygen is introduced
into the system via the plasma chamber (the ECR source), while silane enters
directly in the reaction chamber. The oxide is formed by the reaction that takes
place on the substrate surface between a silane molecule and incident oxygen ions.
For deposition of silicon nitride the arrangement is similar, but in this case the
plasma gas is changed to nitrogen [22].
Remote plasma reactors are used when it is necessary to avoid direct contact
of the processed sample with the plasmas; to prevent damaging of sensitive
samples. They are used for deposition of inorganic or organic coatings or for
plasma treatment of polymers.

4.8 REACTOR CLUSTERS

Many manufacturing processes using PECVD consist of multiple steps, with one
or more plasma reactors. This is especially the case in manufacturing micro-
electronic components. For example, manufacturing of DRAM computer chips
involves hundreds of process steps and the processing of the 64-Mbit DRAM chips
will require up to 1000 steps [29].
For better control of the fabrication process and yield improvement, the
wafers have to be moved between process steps without exposure to environment
to avoid contamination. For this purpose, cluster tools that are modular, multi-
chamber, integrated process systems were developed in the last few years. A
cluster tool, as illustrated in Fig. 4-22, consists of a central handling station (the
core) and a number of peripheral modules also called process stations. Processing
modules, which include the reactors, load-locks, or inspection stations, can be
arranged in any configuration around the central core. PECVD reactors are often
included in the cluster tools. Each processing unit is mechanically and electrically
independent and has its own vacuum and control system.
The central handling platform contains transport mechanisms that move the
processed specimens from module to module. Several central platforms may be
interconnected for further process integration.
The modules of a cluster tool can be of two types:
• Single-process modules, which are usually batch modules for processing
multiple specimens in low rate processes.
Sec. 4.9 Questions 111

Procoss modu lo

Samplo
holdor
Contral
transtor modulo

Load-lock

Insortlon
modu lo

.'ig, 4-22 Diagram of a cluster 1001.

• Multiple-process modules, which consist of several fully separated internal

stations, each providing a single-wafer processing environment. A multi-
process module occupies only a single port of the central handling module .
An internal transport mechanism moves the wafers from one process sta-
tion to another within the module, while the central handler is free to serve
other ports of the cluster.

4.9 QUESTIONS
I. A certain plasma process has an optimum for a residence time of 0.4 sec, when
the process is performed in the pressure range of 100 mtorr to I torr. The
process will be performed in a parallel plate reactor having two electrodes, 8
in. in diameter and 3 in. apart. One electrode serves as a showerhead for
introduction of the gases in the system. What is the maximum range of mass
flow controllers to be used with this reactor? Explain.
2. The mass flow controller to be used with a plasma reactor requires a minimum
pressure difference of 10 psi between its ends for proper operation. The plasma
process performed at a pressure of 5 torr in the reactor has to use a precursor
that has a vapor pressure of 600 torr at room temperature. Describe an experi-
mental arrangement for supplying the precursor to the reactor at controlled flow
rate.
3. Another precursor has a vapor pressure of 50 torr at room temperature. De-
scribe the experimental arrangement for supplying this precursor at a controlled
flow rate. What is determining the flow rate?
4. One of the precursors for a plasma process has a very-low-vapor pressure at
room temperature. What experimental arrangement is required to supply a
controlled flow rate of this precursor to the reactor?
112 Plasma Reactors Chap. 4

5. A plasma process was developed in a cylindrical reactor, 10 em in diameter,

in which the plasma is inductively excited by a 7-cm-Iong coil. The coil is
powered with 50 watts by an RF generator at 13.56 MHz. The process has an
optimum at a pressure of 0.1 torr, a total mass flow rate of 10 seem and sample
temperature of 150 GC. The process has to be scaled up to a parallel plate reactor
with electrodes of 8 in. in diameter. What will be the process parameters in
the scaled-up reactor? Will these parameters ensure the reproduction of the
process developed in the small reactor, or will adjustments have to be made?
Explain why.
6. Describe the special equipment requirements for a plasma system to be used
for the deposition of polycrystalline films with extremely low oxygen content
employing pyrophoric precursors. Which process parameters will have to be
precisely controlled?
7. Which type of reactor appears to be most suitable for high rate reactive etching
with minimal induced substrate damage?
8. List all hardware components required to set up a diverging field ECR reactor.
What changes will be required for a multipolar reactor?
9. Design a parallel plate triode reactor and list the hardware components required
for this reactor. Describe different equipment options.

4.10 REFERENCES

[1] Hollahan, J. R., and R. S. Rosier, In Thin Film Processes, eds. J. L. Vossen
and W. Kern, p. 335. New York: Academic Press, 1978.
[2] Vossen, J. L., and W. Kern, eds., Thin Fi/m Processes. New York: Academic
Press, 1978.
[3] Catherine, Y., In Plasma Processing, eds. G. S. Mathad, G. C. Schwartz, and
G. Smolinsky, p. 317. Pennington, NJ: Electrochemical Society, 1985.
[4] Hess, D. W., and D. B. Graves, In Microelectronics Processing, eds. Dennis
W. Hess and Klavs F. Jensen, p. 377. Washington, DC: American Chemical
Society, 1989.
[5] Mort, J., and F. Jansen, eds. Plasma Deposited Thin Films. Boca Raton, FL:
CRe Press, 1986.
[6] Yasuda, H. K., Plasma Polymerization. New York: Academic Press, 1985.
[7] U.S. Federal Communications Commission Rules and Regulations. Washing-
ton, DC: U.S. Government Printing Office, 1964.
[8] Mayo, N., U. Carmi, I. Rosenthal, R. Avni, R. R. Manory, .and A. Grill, J.
Appl. Phys., 55: 4404 (1984).
[9] Grill, A., A. Raveh, and R. Avni, Surf. Coat. Techno/., 43/44: 745 (1990).
[10] Reinberg, A. R., U.S. Patent, 3, 757, 733 (1973).
Sec. 4.10 References 113

[11] Reinberg, A. R., In Proc. Electrochem. Soc. Meet., p. 30. San Francisco, May
12-75, 1975.
[12] Shennan, A., Thin Solid Films, 113: 135 (1984).
[13] Chapman, I. B., Glow Discharge Processes: Sputtering and Plasma Etching.
New York: J. Wiley & Sons, 1980.
[14] Flamm, D. L., Solid State Technol., 34(3): 47 (1991).
[15] Perry, A. J., D. Vender, and R. W. Boswell, J. Vac. Sci. Technol. B9, 310
(1991).
[16] Marcuwitz, N., Waveguide Handbook. London: Peter Peregrinus, 1986.
[17] McTaggart, F. K., Plasma Chemistry in Electrical Discharges. Amsterdam:
Elsevier, 1967.
[18] Post, R. S., J. R. Conrad, and D. K. Smith, Res. Dev. Mag., 31(10), 106
( 1989).
[19] Lamontagne, B., A. M. Wrobel, G. Jalbert, and M. R. Wertheimer, J. Phys.
D., 20: 844 (1987).
[20] Wrobel, A. M., B. Lamontagne, and M. R. Wertheimer, Plasma Chern.
Plasma Process. 8: 315 (1988).
[21] Matsuo,S., and M. Kiuchi, J. Appl. Phys. 22: L210 (1983).
[22] Denison, D. R., M. S. Chang, and T. Ebata, Microelectron. Manufact. Test.
22 (July 1990).
[23] Burke R. R., and C. Pornot, Solid State Technol. 31(2): 67 (1988).
[24] Geisler, M., J. Kieser, E. Rauche, and R. Wilhelm,.l Vac. Sci. Technol. A8:
908 (1990).
[25] Engelhardt, M., Semicond. Int., 52 (July 1991).
[26] Laporte, P., Y. Melaku, and M. Goethals, Solid State Technol. 34(4): 69
(1991).
[27] Nguyen, S. V., G. Christman, D. Dobuzinsky, and D. Hannon, Solid State
Technol., 33(10): 73 (1990).
[28] Lucovsky,. G., D. V. Tsu, and R. J. Markunas, In Handbook of Plasma
Processing Technology, eds. Stephen M. Rossnagel, Jerome J. Cuomo, and
William D. Westwood, p. 387. Park Ridge, NJ: Noyes Publications, 1990.
[29] Bader, M. E., R. P. Hall, and G. Strasser, Solid State Technol. 33(5): 149
(1990).