Home Search Collections Journals About Contact us My IOPscience

Helicon modes in a cylindrical plasma source

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

1999 Plasma Sources Sci. Technol. 8 79

(http://iopscience.iop.org/0963-0252/8/1/010)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 134.68.190.47
The article was downloaded on 30/09/2012 at 10:20

Please note that terms and conditions apply.

Plasma Sources Sci. Technol. 8 (1999) 79–87. Printed in the UK PII: S0963-0252(99)99768-2

Helicon modes in a cylindrical plasma

source
J P Rayner† and A D Cheetham‡
Plasma Instrumentation Laboratory, University of Canberra, Canberra ACT 2600, Australia

Received 7 August 1998, in final form 5 November 1998

Abstract. The axial wavelengths and azimuthal mode structures of a cylindrical helicon
plasma source have been investigated over a broad range of input conditions (pressure
1–16 µbar, field 0–450 G, rf power 0–2500 W @ 13.56 MHz) in an argon plasma. Several
distinct modes of operation, separated by discontinuous jumps, have been identified:
an electrostatic mode, an m = 0 wave mode and three m = 1 helicon wave modes.
Two different types of mode jump have also been identified. A cavity mode transition has
been identified in which a mode jump occurs when the wavelength of standing waves
governed by the length of the plasma vessel leads to a node in the wave field at the centre of
the antenna. A radial mode transition has also been identified in which the favourable
condition that the axial wavelength should be equal to twice the length of the antenna for
efficient coupling of the rf power is maintained at high powers by the plasma making a
transition from the first to the second radial mode. A procedure based on these observations
is suggested for predicting where mode transitions are likely to occur within the parameter
space of a helicon source.

1. Introduction

The helicon plasma source is finding increasing application

in industrial situations involving the micro-etching and
deposition of materials such as silicon, silicon oxides and
metals as required for example in the manufacture of
integrated circuits [1, 2]. The PR-1 Helicon plasma source
was commissioned in 1995. Since then the operating space
of the source has been explored by using a Langmuir
probe to measure the plasma density as a function of the
input parameters of magnetic field (0 to 450 G), rf input
power (0 to 2500 W) and filling pressure (1–100 µbar). A
detailed description of the PR-1 source and the results of
this investigation have been reported elsewhere [3]. This
exploration has mapped out three major regions of operation,
or modes, that tentatively have been identified as: an
electrostatic mode (E mode); an inductive mode (H0 mode)
and several helicon modes (HI, HII and HIII modes)
Some of these modes are clearly identified in figure 1
which shows the ion current to a probe biased at −100 V
as a function of the rf power (<1.2 kW) for two different Figure 1. Modes observed in the helicon plasma source PR-1.
magnetic fields and filling pressures. Similar modes to E, H0
and HI have been identified in other Helicon sources [1, 4]. supplying the rf power and the plasma [4]. Other jumps have
Figure 2 shows that further modes are observed in addition to been associated with resonances between the wavelength of
the earlier modes if the input power increases above 1.2 kW. the helicon waves and the length of the antenna [1] or to
Discontinuous jumps separate the modes from each the possibility of standing waves in the plasma cavity [5].
other. Various explanations have been offered for the However in any given experimental situation the mechanism
existence of these mode jumps. Some jumps, such as that leading to a particular jump is often not clear. In addition it
from the electrostatic to the inductive mode, occur because is often difficult to predict where the jumps will occur within
of changes in the coupling mechanism between the antenna the parameter space of the source.
† E-mail address: [email protected]. The primary aim of the Plasma Instrumentation
‡ E-mail address: [email protected]. Laboratory is to carry out research into computer-controlled

0963-0252/99/010079+09$19.50 © 1999 IOP Publishing Ltd 79

J P Rayner and A D Cheetham

Figure 2. Probe current as a function of net power, p = 2 and

8 µbar, B = 400 G.
Figure 3. Sketch of the helicon plasma source PR-1.

automatic instrumentation and control systems for plasma

processing sources. However, the existence of mode jumps
makes it extremely difficult to set up an automatic control
system for the plasma source. Conventional control systems
normally operate on smoothly varying linear systems. Highly
non-linear systems, such as this one, pose very significant
problems for the controller. To solve these problems it is
important that a better understanding is obtained of the mode
jumps and where they occur. This will allow the control
system either to avoid those regions where jumps are likely
to occur or to take pre-emptive action in anticipation of a
jump.
This article describes the results of a preliminary
investigation of the rf wave fields associated with the different
modes in an attempt to determine the conditions under which
the jumps occur and to better characterize the different modes.
It presents evidence that the transition from the HI to the HII
helicon mode is associated with the existence of standing
waves determined by the geometry of the plasma cavity and
Figure 4. Location of the Bθ rf magnetic field coils around the
is therefore referred to as a ‘cavity mode transition’. The vessel (A to DD) and orientation of the rf magnetic field launched
jump from the HII to the HIII mode will be shown to be by the antenna straps (1 to 4). The current in straps 1 and 2
due to a transition to a higher order radial mode controlled opposes that in 3 and 4. The antenna is about 0.23 m down-stream
by a tendency to maintain the axial wavelength at twice the from the pick-up coils.
length of the antenna and is referred to as a ‘radial mode
transition’. thus largely cancelling the axial component of the imposed
rf field. The antenna has an axial length of 0.11 m and its
2. Experimental arrangement centre is located 0.20 m from the end plate.
Four coils mounted at 90◦ intervals around the outside
Figure 3 is a sketch of the PR-1 source showing its major of the plasma vessel measure the azimuthal magnetic field
dimensions. It consists of a 0.10 m diameter Pyrex vacuum components of the rf wave fields. Each coil consists of two
vessel surrounded by a large Helmholtz pair to provide the windings of five turns each wound in opposite directions.
magnetic field. At the left hand end of the vessel there is a A balanced rf transformer subtracts the outputs from the
grounded metal end plate that serves as the ground for the windings. This arrangement cancels the common mode pick-
Langmuir probe. The probe tip is located under the centre up from the electric fields while summing the magnetic fields.
of the left hand coil and adjacent to one end of the antenna. A four-channel digital oscilloscope monitors the outputs
The right hand end of the vessel has a glass end cap located from the coils with the traces from the oscilloscope being
0.60 m from the centre of the antenna. The plasma is excited down-loaded to a computer for analysis. A Pearson current
by an m = 1 Boswell type helicon antenna [1] and is fed transformer that monitors the antenna current provides the
via a balanced feed from an industrial rf supply operating at trigger signal for the oscilloscope, and therefore acts as a
13.56 MHz [6]. The antenna consists of two single-turn loops stable reference for all phase measurements. The coils are
in series with each other located on each side of the vessel. normally located at position A in figure 3, 0.23 m from the
The currents in both sets of end straps oppose each other centre of the antenna, but can be moved to position B, 0.305 m

80
 Helicon wave modes

Figure 5. (a) Theoretical polar plots showing the amplitude of the azimuthal field as a function of the azimuthal angle. (b) Theoretical plots
of the helicon wave field structure for n = 1 and n = 2 radial modes.

from the antenna so that wavelength determinations can be The radial profiles of the wave magnetic field components
made on the travelling waves by noting the change in phase of of a helicon wave in a uniform plasma for n = 1 and 2 are
the signals. The coils normally are mounted with two coils shown in figure 5(b) [7].
horizontal and two vertical. However, they can be rotated
by 45◦ as required to give further data points around the 3.1. Inductive mode, H0
circumference of the vessel. Figure 4 shows the location of
the coils and also the direction of the initial rf field imposed The inductive mode, H0, has been explored experimentally
by the antenna which is a maximum at approximately 45◦ to in argon over the following range of conditions:
the vertical. pressure: 1–100 µbar
RF power: 300–1200 W
3. Investigation of the structure of the wave fields magnetic field: <150 G.
In this mode the plasma does not extend significantly beyond
Various mode structures, (m, n), can be identified depending the field coils and visually appears to have a rather flat to
upon which of the spatial co-ordinates (r, θ, z) is being slightly hollow radial structure with its highest luminosity in
considered and are characterized by an azimuthal mode the immediate vicinity of the antenna. The number density is
number m and radial mode number n. Different azimuthal ∼1018 m−3 and increases linearly with applied power. The
mode patterns can be represented on a polar plot, as in signals picked up by the Bθ coils show that at any time the
figure 5(a), showing the wave amplitude as a function of θ. signals from all four coils are essentially in phase with each
This figure shows a set of theoretical polar plots for different other with an almost circular radial pattern. This behaviour
values of m in which the amplitude Aθ of the azimuthal is characteristic of an m = 0 wave mode. A spatial Fourier
magnetic field Bθ is plotted radially as a function of the transform of the data confirms this observation and shows
azimuthal angle θ, with Aθ = 0 placed at half the radius. that the m = 0 mode is dominant although a significant
Each plot is made at time t = 0. For travelling waves at amount of m = 1 mode is also present (∼25% of the m = 0
other times the plot rotates at the frequency of the applied rf mode). The dominance of the m = 0 mode is somewhat
signal. surprising as it is difficult to see how such a mode could be
The radial modes of the helicon wave are defined by the launched by the type of antenna used in these experiments.
boundary condition that the radial field Br = 0 at the wall. However, m = 0 modes have also been observed in other

81
J P Rayner and A D Cheetham

experiments employing the m = ±1 symmetric Nagoya type The mode exhibits an intense blue core with a radius ∼60%
III antenna [8]. It therefore appears that some other coupling of the vessel radius. Number densities are typically a factor
mechanism may be in operation which can excite the m = 0 of two or three higher than for the HII mode. Wave field
mode from antennas whose geometry should preclude this measurements show that the m = 1 component of the
possibility. travelling wave is still dominant.
Movement of the pick-up coils by 75 mm from position A
to B in figure 3 introduced no discernible change in the phase 4. Wavelength theory
of the signals with respect to the antenna current reference,
as is consistent with the existence of an m = 0 wave. This The many modes of operation identified above, the existence
mode, however, is not fully understood and is the subject of of the three input parameters of pressure, field and power
continuing work. and a variety of geometries make it difficult to make
generalizations about the behaviour of the source. This
3.2. Helicon mode, HI section, therefore, seeks to cast the dispersion relation for
helicon waves into a non-dimensional form in order to obtain
This mode has been explored over the following range of appropriate non-dimensional numbers that can be used to
conditions: characterize the different regimes of operation.
pressure: 1–3 µbar Helicon waves in a uniform cylindrical plasma of radius
RF power: 400–1200 W a subjected to an axial magnetic field B0 , can be represented
magnetic field: 150–450 G. by a dispersion relation of the form ([9] p 437):
In this mode the plasma has a well defined blue core
eµ0 n0 ω
indicative of a significant population of Ar II, and extends kkz = (1)
over the length of the plasma vessel. Number densities B0
are typically a factor of two higher than for the H0 mode where n0 is the electron number density, ω is the applied
and increase linearly with power. However, as figure 1 rf angular frequency, kz is the axial wavenumber, k is the
shows, there is evidence of instability at an input power of magnitude of the total wave-vector given by
about 900 W, possibly indicative of an attempted transition
1
to another mode. The pick-up coils show that the mode k = (k⊥
2
+ kz2 ) 2 (2)
pattern is rotating in time consistent with an m = +1
wave. The spatial Fourier transform shows that the m = and k⊥ is the radial wavenumber.
1 mode is dominant with some evidence of an m = 3 If (1) is non-dimensionalized using the plasma vessel
mode (∼20% of the m = 1 mode). Movement of radius a then:
the pick-up coils from A to B led to significant phase (ka)(kz a) = Q (3)
changes, indicating a travelling wave, and from which the where
wavelength measurements to be presented in section 5 were eµ0 n0 ωa 2
Q= . (4)
made. B0
It is found that Q is a useful dimensionless number for
3.3. Helicon mode, HII characterizing helicon waves.
If k⊥ a = β is the non-dimensional radial wavevector,
This mode exists above pressures of 5 µbar for an input power
then from (2) and (3):
of 0.5 kW increasing to 1.0 kW as the pressure decreases to
2 µbar. This mode exhibits a more diffuse blue core than the

  1/2
β Q2 1/2
HI mode, and number densities ∼4 × 1018 m−3 . This mode kz a = 1+4 4 −1 (5)
2 β
has been explored over the following range of conditions:
pressure: 2–16 µbar and
 1/2 
RF power: 500–2200 W kz 1 β4 β2
= +4 − . (6)
magnetic field: 350–450 G. k 2 Q2 Q
Phase measurements and the Fourier transform again indicate The wavenumbers k, kz and k⊥ are linked by an equation of
that a travelling wave dominated by the m = 1 mode is the form:
present although its amplitude is somewhat smaller than for kz
mJm (k⊥ a) + aJm (k⊥ a) = 0 (7)
the HI mode. k
where Jm is a Bessel function of order m and Jm (k⊥ a) is the
3.4. Helicon mode HIII derivative with respect to r of Jm (k⊥ r) evaluated at r = a.
Substitution of (6) into (7) yields an equation for the non-
This mode exists above rf power levels of 1.1 kW for a dimensional radial wavevector, β, of the form:
pressure of 8 µbar, increasing to 2.1 kW as the pressure
decreases to 2 µbar. Typical densities are ∼(8–10) ×
  21 
β β4 β2 d
1018 m−3 . The mode has been explored over the range: mJm (β) + +4 − Jm (β) = 0. (8)
2 Q2 Q dβ
pressure: 2–8 µbar
RF power: 1.1–2.5 kW For a given value of m, equation (8) can be solved numerically
magnetic field: 200–400 G. for β as a function of Q. These results can then be substituted

82
 Helicon wave modes

where the electron temperature Te is in eV, and typically has

a value ∼3.0 eV for these experiments.
For comparison, figure 6 also shows the theoretical
wavelength predicted by equation (9) using a value of ka =
3.1 and (10) for the ion current. The error bars on the
wavelengths represent uncertainties in the measurement of
λz , while those on the ion current allow for uncertainties
in the assumed temperature (8%), and in the dimensions of
the probe (6%). The good agreement between theory and
experiment confirms the adequacy of the simple dispersion
relation given by equation (9) provided that an appropriate
value of ka is employed.
The results presented here, and their apparently good
agreement with the theory, need to be qualified in several
respects.
Figure 6. Axial wavelength as a function of the ion current to the
Langmuir probe at p = 2 µbar and B = 400 G for the HI and HII The first qualification is that the measurement baseline
modes. of 7.5 cm is small compared with the observed wavelengths
particularly in the HI mode. This may lead to the
into (5) to give kz a as a function of Q. Different radial modes, possibility that higher order axial modes [8, 10] that would
of radial mode number n, are obtained by choosing different not be observed using this baseline will increase the true
starting values for β. wavelengths above the reported values. A simple calculation
For the (m, n) = (0, 1) mode β corresponds to the first indicates that if a second axial harmonic were present with
zero of J0 i.e. 3.83, and is independent of the value of Q. For an amplitude ∼0.5 of the first harmonic then the observed
the next radial mode, (0, 2), β = 7.016. values of the wavelength should be increased by ∼30%.
Solutions for kz a as a function of Q for the (m, n) = The second qualification is that the measurements were
(1, 1) and (1, 2) modes are obtained by assuming an initial made in a region where the magnetic field B0 is decreasing
value of β = 3 in the numerical solution of (8) for the n = 1 fairly rapidly in the axial direction. However the number
mode and β = 6 for the n = 2 mode. Solutions were obtained density n0 is also decreasing with distance. As will be shown
for values of Q in the range from 0 to 40. These values in connection with figures 8 and 9 the conditions in PR-1 are
cover most helicon sources of practical interest from small such that the ratio n0 /B0 remains almost constant for values
diameter high field sources to large diameter (∼0.2 m), low of B0 >150 G. Since the wavelength is largely determined by
field (∼25 G) facilities. For the PR-1 source Q typically lies this ratio and B0 remains >150 G between positions A and B it
in the range: 1 < Q < 10. The solutions show for Q < 8 in is probable that the measured wavelengths are representative
the (1, 1) mode, and Q < 20 in the (1, 2) and (1, 3) modes, of the entire region from the end plate to position B.
that kz a increases in an approximately linear fashion with The third qualification is that the apparatus is not
Q. This implies from equation (3) that ka is approximately symmetrical about the centre line of the antenna. The
constant so that for engineering design purposes: conditions to the left of the antenna where the plasma is
terminated by a metal end plate may well be different to
 
1 those to the right where the plasma vessel is terminated
kz a ≈ Q (9) by a glass end cap. The effect of this asymmetry is not
ka
known. In addition the geometrical constraints of the present
where ka = 3.1 for the (1, 1) mode, 6.4 for the (1, 2) experiment meant that the number density was measured on
mode and 9.1 for the (1, 3) mode. Within this linear region the opposite side of the antenna to that where the wavelength
kz a  β, so that from (2), ka ≈ β. measurements were made. Measurements however of the
axial field showed that it was symmetrical about the plane
5. Cavity mode transitions of the antenna. Also the plasma visually appeared to be
symmetrical with the blue core of the plasma extending for
This section describes measurements of the axial wavelength a similar distance on each side of the antenna. There is
λz of the travelling waves observed in the helicon HI and HII therefore some evidence that the experiment is symmetrical
modes. The wavelengths were measured by recording the about the antenna. In the theoretical calculations the field
phase differences between the signals from the pick-up coils used was that at the location of the probe tip in order to obtain
at positions A and B in figure 3. a representative value for the ratio n0 /B0 .
Figure 6 shows the wavelength measured for the m = 1 Figure 6 also shows the measured axial wavelengths for
helicon HI mode at a pressure of 2 µbar and a field of 400 G the HII mode compared with the HI mode. Wavelengths
as a function of the ion current Ii recorded at the same time in the HI mode cover a range from about 0.90 m down to
by the Langmuir probe. For the probe Ii is related to the about 0.45 m with a region of instability becoming apparent
number density n0 by the expression at a wavelength of around 0.40 m. Wavelengths in the HII
√ mode appear to be locked into the range between 0.33 m and
n0 × 10−18 Te 0.29 m with no evidence of stable wavelengths >0.33 m.
Ii = mA (10)
1.25 These results indicate that a significant problem exists with

83
J P Rayner and A D Cheetham

expected throughout the parameter space for the (1, 1) mode.

Figure 8 is a contour plot showing λz as a function of power
and magnetic field at a pressure of 2 µbar. The plot shows that
at constant pressure and power input, the wavelength remains
fairly constant as the magnetic field increases above about
150 G. As the magnetic field increases, it appears that the
increase in the number density due to the radial confinement
provided by the field is just sufficient to keep the ratio n0 /B0
constant. Thus, from equations (1) and (9) Q, and hence λz ,
remains constant. Increasing the pressure increases n0 and
hence decreases the wavelength for given values of the power
and field.
Figure 8 also shows several distinct regions. Regions
Figure 7. Geometry of PR-1 showing some possible standing HI and HII correspond to allowed or preferred regions
wave patterns. of the parameter space in which helicon waves are easily
established. These regions are separated by unfavourable
generating a stable mode when the wavelength is in the regions where it is difficult to excite stable waves as such
vicinity of 0.35–0.40 m. Further evidence of this problem waves require a node in the vicinity of the antenna.
is evident in figure 2 where the instability in the HI mode for Figure 9, which shows the Langmuir probe current as
an input power of 900 W occurs at a wavelength ∼0.42 m. a function of magnetic field at a constant input power of
Figure 7 is a sketch of the PR-1 source showing the 800 W and pressure of 2 µbar, provides further evidence
end plate, antenna and some possible standing wave patterns. for the picture presented in figure 8. These results show
Although the axial geometry of the plasma to the right of the that above 125 G the ion current increases linearly with
antenna is rather poorly defined it is significant that, to the field which implies that n0 /B0 , and hence Q, is constant.
left, the distance, L0 , between the grounded metal end plate From equation (9) the associated constant wavelength is
and the centre of the antenna has a value of 0.20 m. Due to 0.48 ± 0.02 m. These results correspond to the horizontal
the reflection of waves from the end plate, standing waves region of figure 8 which predicts a wavelength of 0.50 m for
can be set up in the region between the antenna and the end a horizontal cut taken across the plot at 800 W.
plate. Figure 7 shows that standing waves with wavelengths Figure 9 also shows two dips at around 125 G and
either greater than 0.40 m or in a range around 0.27 m are 50 G with small peaks at 25 G and 75 G. The dip at
easily excited as they have a node at the end plate and an 125 G corresponds to the unfavourable excitation boundary at
anti-node in the vicinity of the antenna. On the other hand λz = 0.4 m while that at 50 G may represent the λz = 0.2 m
wavelengths of 0.20 m or 0.40 m are difficult to excite as they boundary.
require a node at the antenna. However, since the antenna
has an axial length of 0.11 m some excitation may still be 6. Radial mode transitions
possible. Based on these observations, and the experimental
results presented in figure 6, it therefore appears that the Figure 10 shows experimental wavelengths as a function
transition from the HI to the HII mode is associated with of the ion current to the probe for the HI, HII and HIII
a jump from one stable standing wave pattern to another and modes. The figure also shows the corresponding theoretical
that wavelengths in the vicinity of 0.40 m are excluded as wavelengths based on equation (9) for the first and second
they would require a node at the antenna. As the standing radial modes when m = 1. Spatial Fourier transforms for
wave patterns are determined by the axial dimensions of the these modes confirm that the dominant azimuthal mode is
plasma, these observations provide evidence for a ‘cavity m = 1 although in both cases there is also a significant
mode transition’ when λz ∼2L0 . component of the m = 3 mode. Figure 10 shows that the
The HII wavelengths ∼0.29 m, and their lack of variation wavelengths for the HII and HIII modes are locked at a value
with ion current, are consistent with the idea ([9] p 439) that approximately equal to twice the antenna length i.e. λZ =
efficient coupling between the antenna and the plasma occurs 2LA . It appears that as the input power and density increase,
when the wavelength λz ∼2LA where LA is the length of the and λZ decreases, in the HII mode a stage is reached where
antenna and LA = 0.11 m in PR-1. a further increase in density would lead to a wavelength that
A global model of the plasma developed by Lieberman becomes <2LA . Rather than this occurring, the effect of the
[9, 11] has been adapted for the PR-1 source by the inclusion antenna length imposing a defined wavelength is sufficiently
of an additional term that takes into account the radial strong that the plasma makes a ‘radial mode transition’ to the
confinement of the plasma due to the axial magnetic field HIII mode. In this mode a similar wavelength is maintained
[3]. For the parameter space considered here the model but at a much higher density. The same transition is observed
predicts a typical electron temperature of 3.0 ± 0.5 eV for as the input power reduces and the plasma reverts from the
an argon plasma, and plasma densities as a function of the HIII mode back to the HII mode although some hysteresis is
filling pressure, field and input power that are consistent with apparent. Boswell has suggested a similar radial transition
observations to within ∼15%. based on observations of the change in the radial density
The densities predicted by the model have been used distribution [1]. Radial mode transitions governed by the
in equation (9) to calculate the axial wavelengths λz to be antenna length, however, would not be expected for antennas

84
 Helicon wave modes

Figure 8. Allowed and unfavourable regions of operation of the PR-1 source for (m, n) = (1, 1) helicon waves.

Figure 9. Langmuir probe current against magnetic field for Figure 10. Experimental wavelengths for the HI, HII and HIII
800 W net power to the plasma. The line represents the best fit to modes as a function of ion current to the Langmuir probe compared
the data in the linear region. with theoretical wavelengths for the (1, 1) and (1, 2) modes.

separated so that the cavity mode transitions and the radial

with essentially zero axial length, such as a double half turn mode transitions can be studied in isolation.
antenna [4].
As noted earlier the HIII mode visually exhibits a highly 7. Discussion
ionized central core. Theoretical radial distributions of the rf
magnetic wave fields indicate that the Br and Bθ components The results presented in sections 5 and 6 provide evidence
are highly peaked in the (m, n) = (1, 2) HIII mode with of two different types of transition, namely cavity mode
significant values only over the central 60% of the plasma. and radial mode transitions. Cavity mode transitions are
This suggests that the rf power coupled into the plasma in the associated with the wavelengths of standing waves that lead
HIII mode is likely to be strongly concentrated in the inner to a node in the vicinity of the centre of the antenna. Radial
core thus leading to the high densities observed in this region. mode transitions take place in order to maintain a wavelength
It should be noted that in the apparatus used for approximately equal to twice the length of the antenna and
these experiments the antenna wavelength (22 cm) and the a favourable cavity mode. Also section 4 showed that
favourable standing wavelength (27 cm) are very close in the parameter Q = µ0 eωn0 a 2 /B0 is a useful number for
value and that wavelengths in this range are therefore likely characterizing the helicon dispersion relation and that for
to be strongly preferred for two reasons. The wavelength at values of Q
8 the axial wavenumber increases almost
which the radial mode transition occurs is not exactly that of linearly with Q.
the antenna wavelength, but it is within experimental error. It These ideas are brought together in figure 11 which is a
is not clear which mechanism may be dominant in the present composite graph showing the results presented in the previous
experiment. However, the modifications to the experiment, sections plotted as wave numbers against the corresponding
to be described in section 8, will allow these values to be experimental values of Q. The horizontal lines correspond to

85
J P Rayner and A D Cheetham

that the transition corresponding to Q0 = 2.5 should occur

at an input power ∼840 W, whereas the observed transition
occurs at ∼750 W. Similarly for p = 2 µbar and B0 = 400 G
the model predicts that Q0 = 2.5 for an input power of
∼1000 W compared with an observed value of ∼950 W.
This reasonable level of agreement should be sufficient for
engineering applications of the source.

8. Conclusions

The analysis given here shows that when characterizing or

designing helicon sources, and the control systems required
for their automation, the non-dimensional Helicon dispersion
relation, the geometry of the antenna and the source and
Figure 11. Composite plot showing experimental the global model can be used to predict the location of
non-dimensional wavenumbers kz a as a function of Q compared the mode jumps within the parameter space of the source.
with the theoretical wavenumbers for the (1, 1) and (1, 2) modes. Appropriate action can then be taken either to avoid these
Also shown are the boundaries for cavity mode transitions at
0.40 m and 0.20 m and the antenna wavenumber for which regions or to devise suitable control strategies to cope with the
λZ = 2LA . transitions. Similarly, preferred, stable regions of operation
can be identified.
As was pointed out in section 5 a number of qualifications
wave numbers for the cavity mode boundaries at λZ = 2L0
need to be considered when interpreting the present
and λZ = L0 and the antenna condition at λZ = 2LA .
preliminary observations and conclusions. Consequently the
The Helicon modes may be identified from this graph as
experiment is being re-designed and further experiments are
follows:
proposed. In particular, a uniform field has been established
HI: (1, 1) mode with λZ > 2L0
over an extended length of plasma column and two movable
HII: (1, 1) mode with λZ < 2L0 and λZ ∼2LA and
metal end plates have been installed within the plasma vessel.
∼4L0 /3
These plungers will provide a well defined length for the
HIII: (1, 2) mode with λZ < 2L0 and λZ ∼2LA and
plasma column that can be varied over a variety of possible
∼4L0 /3.
lengths producing different associated standing wave modes.
The transitions can be summarized as: Finally an extended array of pick-up coils is being installed so
HI to HII: cavity mode transition that wavelength measurements can be made over a significant
HII to HIII: radial mode transition governed by the length of the vessel. Using this flexible geometry it should
antenna length and favourable antenna length. be possible to sort out which modes are determined by the
From these observations a possible strategy for finding the plasma cavity and which may be determined by the antenna.
location of helicon mode transitions is to determine the Some problems still remain with respect to the H0 mode
wavenumber of any standing waves, based on the dimensions and in particular the way in which the m = 0 mode can be
of the plasma vessel, that lead to a node in the plane of launched from the present antenna geometry. Also further
the antenna. Also, for an antenna of finite length, the work needs to be completed on a fuller examination of the
wavenumber of those waves with a wavelength equal to cavity modes, and the effect of highly non-uniform radial
twice the antenna length can be found. These two sets density distributions on the dispersion relations. Finally a
of wavenumbers are then plotted on the universal curves continuing problem with the operation and control of helicon
of normalized wavenumbers for different modes (m, n) sources relates to the matching network linking the rf source
expressed as functions of Q. The intersection of the curves to the antenna and the automation of the tuning process to
with the cavity mode wavenumbers yield values of Q (=Q0 ) allow for the significant mismatches that occur in the vicinity
where cavity mode transitions are likely to occur. Similarly of the mode jumps. This problem may be resolved with
Q values (= QA ) for radial mode transitions governed by increased understanding of the behaviour of the waves in the
the antenna length can be found. Since Q = µ0 eωn0 a 2 /B0 , vicinity of these jumps and the consequent changes in the
and in a given situation ω and a normally are fixed, it only impedance of the antenna.
remains to determine those values of n0 and B0 that lead
to the critical values of Q0 and QA . This can be achieved Acknowledgments
using the global model [3] of the source which predicts the
density with adequate precision for the lowest order radial This work is supported by a University of Canberra Research
mode (n = 1). The model, however, may require further Grant and an Australian Research Council Small Grant.
modification to take into account the highly peaked radial
plasma distributions encountered in the higher order radial References
mode.
By way of example, figure 11 shows that for the PR-1 [1] Boswell R W 1984 Plasma Phys. Control. Fusion 26
1147–62
source a cavity mode transition should exist for Q0 = 2.5. [2] Perry A J, Vender D and Boswell R W 1991 J. Vac. Sci.
The global model predicts for p = 5 µbar and B0 = 400 G, Technol. B 9 310–16

86
 Helicon wave modes

[3] Cheetham A D and Rayner J P 1998 J. Vac. Sci. Technol. A [8] Chen R T S, Bruen R A, Gross S, Herskowitz N, Hsieh M-K J
16 2777–84 and Jacobs J 1995 Plasma Sources Sci. Technol. 4 337–44
[4] Ellingboe A R and Boswell R W 1996 Phys. Plasmas 3 [9] Lieberman M A and Lichtenberg A J 1994 Principles of
2797–804 Plasma Discharges and Materials Processing (New York:
[5] Shamrai K P, Pavlenko V P and Taranov V B 1997 Plasma Wiley)
Phys. Control. Fusion 39 505–29 [10] Light M, Sudit D, Chen F F and Arnush D 1995 Phys.
[6] Rayner J P, Cheetham A D and French G N 1996 J. Vac. Sci. Plasmas 2 4094–103
Technol. A 14 2048–55 [11] Lieberman M A and Ashida S 1996 Plasma Sources Sci.
[7] Chen F F 1991 Plasma Phys. Control. Fusion 33 339–64 Technol. 5 145–58

87