Plasma Physics

Related content
- Nonlinear effect of large-amplitude ion
Dispersion of electron plasma waves acoustic waves
Y Kawai and H Ikegami

To cite this article: P J Barrett et al 1968 Plasma Phys. 10 911 - Transmission measurements of
electrostatic plasma waves in the GHz
frequency range
M Tutter

- Suppression of Kelvin-Helmholtz instability

View the article online for updates and enhancements. by electron plasma waves
H Sugai and N Sato

Recent citations
- Interpreting the behavior of a quarter-wave
transmission line resonator in a
magnetized plasma
G. S. Gogna et al

- Inverse mirror plasma experimental device

(IMPED) – a magnetized linear plasma
device for wave studies
Sayak Bose et al

- Computation of generalized and exact

dispersion relations for longitudinal plasma
waves in nonextensive statistics and the
effects of the nonextensivity on the
oscillation modes and damps
A. Esfandyari-Kalejahi and V. Ebrahimi

This content was downloaded from IP address 147.32.5.241 on 05/03/2018 at 15:23

 Plasma Physics. Vol. 10. pp. 911 t o 918. Pergamon Press 1968. Printed in Northern Ireland

DISPERSION OF ELECTRON PLASMA WAVES

P. J. BARRETT
and H. G. JONES*
U.K.A.E.A. Research Group, Culhani Laboratory, Abingdon, Berks.
R. N. FRANKLIN
Department of Engineering Science, Oxford University, Oxford

(Received 26 Febrirary 1968)

Abstract-The propagation of electron plasma waves along a cylindrical plasma column in a strong
magnetic field has been studied both theoretically and experimentally. The dispersion equation for
such waves has been solved numerically where the plasma density profile is assumed to be either
uniform or parabolic and the electron energy distribution Maxwellian. Dispersion curves have
been computed for the m = 0 azimuthal mode and compared with experimental points obtained
from a thermally ionized plasma in a strong magnetic field over the range of electron density IO7-
1Olo cn1r3.
1. INTRODUCTION
THISpaper is concerned with the propagation of electron plasma waves along a
cylindrical plasma column in a strong magnetic field; an earlier version was presented
at the Conference on Physics of Quiescent Plasmas, Frascati (BARRETT et al., 1967).
The term electron plasma wave refers to the longitudinal electrostatic wave in the
electrons of the plasma. Such waves are described by the well known BOHMand GROSS
(1949) dispersion relation
CL? = 092 -f- 3lc2t.,2

in the region where w/k 3 U, the electron thermal velocity and where boundary
effects are negligible. o, is the electron plasma frequency in rad/sec. A more com-
prehensive theoretical treatment for an unbounded plasma has been carried out by
DERFLER (1966) which is in good agreement with the dispersion observed by DERFLER
and SIMONEN (1966, 1967) even where the Landau damping is large (k,/k, > 0.1).
Propagation in a cylindrical plasma column of uniform density was treated by
TRIVELPIECE and GOULD(1959) who observed long wavelength waves (wavelength 2
column radius, w < up)whose dispersion was well explained by means of a cold-
plasma theory. This was extended by GOULD(1964) to include the effect of a non-zero
electron temperature, and other authors have since done similar work (DRUMMOND,
1964; LICHTENBERG and JAYSON, 1965; KUEHLand STEWART, 1966; DIAMENT et al.,
1966; GUNSHOR, 1966; LEE,1966).
Landau damping as well as the dispersion of these waves has been observed by
MALMBERG and WHARTCJN (1964, 1965, 1966), using a plasma column several Debye
lengths in radius. Their computed curves for both damping and dispersion are in
good agreement with the observed values.
Our experiments on electron plasma waves are being carried out in order to obtain
further data on the dispersion and damping, using a plasma column in which the
ratio of column radius to Debye length can be varied from Malmberg and Wharton's
value to a value approximately 30 times larger. We show here how the dispersion
curve changes with this ratio, and how it may be used to determine electron density
and temperature.
* Present address: Department of Physics, Loughborough University.
911
912 P. J. BARRETT,
H. G. JONES and R. N. FRANKLIN

2 . THEORY
The following theory is based on that of GOULD(1964). We shall consider the
propagation of waves along the magnetic field (k I ( B). The dielectric constant of the
plasma in this direction is
E =z 1 - (W,'/k'~U,2)Z'(O/kV,)
(1)
where the electron energy distribution is assumed to be Maxwellian with a thermal
velocity U, = (2kTe/n1)1'2,wDis the electron plasma frequency and Z' is the derivative
of the plasma dispersion function. Representing the wave as a potential perturbation
C(r, 0, z ) = +(r, 0) exp (ikz), one obtains from Maxwell's equations
V;,"C$ - k2E+= 0. (2)
If we consider the plasma to be uniform in density and bounded at r = a then (b is
of the form
4 = AJ,,,(i(Z/c)kr)exp (irn0) (3)
inside the plasma and
C$ = BK,(kr) exp ( i n 4 (4)
outside the plasma. By matching the electric induction D,and the electric field E, or
E, at the plasma boundary r = a we obtain

This is the equation derived by Gould. It has solutions corresponding to different

radial and azimuthal modes but we shall consider only the in = 0 azimuthal mode for
the lowest radial mode.
If on the other hand we wish to allow for a variation of density with r , we may, in
the m = 0 case, either choose the form of the radial density distribution n(r) so that
equation (3) has an analytic solution, or integrate this equation numerically for any
form of n(r). Rejecting the solution which is infinite at r = 0 one can compute
[(r/+)(d+/dr)lrGafor given values of (.\/c)ka and then find the value of ka for which

and solve for E .

Figure 1 shows curves of 0 2 / l w D 2 vs. ka which have been computed for n(r) =
n,(l - r2/a2)at different values of w,a/o,, where w , is the electron plasma freque,ncy
a t r = 0. The curves are dashed where the damping is too large for the theory to be
correct. The parameter w,a/v, is of the order of the ratio of column radius to Debye
length and determines the slope of the upper portion of the curve.
Figure 2 shows curves of d / w 2 vs. ka computed for n(r) uniform and n(r) para-
bolic at two values of w,a/v,. It should be noted that for large values of w,a/u, the
dispersion and damping at ka > 1 are almost independent of the shape of n(r)
because the electric fields are concentrated near the axis of the column (GOULD,
1964). At smaller values of this parameter, as in the Malmberg and Wharton experi-
ment, there is an appreciable dependence of the dispersion curves on the shape of n(r).
 Dispersion of electron plasma waves 913

cm
02 01 . .~ - I .:--

0 2 4 6 8 10 I2 14 16 I8 20 22 24
ka
FIG.1.-Dispersion diagram. Curves computed for n(r) parabolic at different values
of w,n/u, (from 74 to CO). Sets of points A, B, C, D from present experiment, and E
from Malmberg and Wharton. Inset: radial profiles A, B, C of ion current to probe
and radial profile E of electron density from Malmberg and Wharton.

3. E X P E R I M E N T A L P R O C E D U R E
The experiment was carried out with a thermally ionized caesium plasma column of
length 60 cm. A 2.5 cm diameter tantalum plate at each end was maintained at a
temperature in the range 1900-2200°K by means of a stabilized bombarding electron
current. The plasma density was monitored by measuring the ion current to a
cylindrical Langmuir probe (length 1.5 mm, diameter 0.35 mm) lying perpendicular
to the axis and biased at -20 V with respect to the end plates, The reading of this
current on a digital voltmeter showed fluctuations ranging from about 0.1 per cent to
several per cent of the average reading. The value of the magnetic field (300-4000 G)
was sufficient to make the electron cyclotron frequency at least an order of magnitude
larger than the plasma frequency. In the upper part of the magnetic field range, the
axial variation of density within the region of wave measurement was 0 3 %; at low
fields the variation was somewhat larger and it was then necessary to confine the
wavelength measurements of the electron plasma waves to a short length of the plasma
column.
The waves were observed by means of the interferometer system shown in Fig. 3.
A signal from a 10-470 or 470-1200 Mc/s generator, set at a level of typically 0.1-1 mW
and sine-wave modulated at 500 kc/s, was injected into the plasma by means of an
exciter probe lying along a diameter of the column. This was a bare tungsten wire
of diameter 0.4 mm at the end of a 50 fi coaxial transmission line. The wave signal
was received by a similar probe and mixed with part of the generator signal by means
of a single CV 2154 crystal through which was drawn a steady current of 150 PA.
Under this condition the crystal acted as a square-law detector, the output from which
 914 P. J. BARRETT,H. G. JONESand R . N. FRANKLIN

j..
l i l t e d parameters
up= I ~ l l x 1 0 9rad/sec
a UI'JOC~
opa/ve-47
1, = 3100'K
0.2 /I

I f , ! I I l l I I I I , ,
0 2 4 6 8 IO 12 14 16 IO 20 22 24
ka

fittrd parameters
up=1~17x109rad/src
a = 0.84 em
opa/v, -38
T. -2200'K

ka
FIG.2.-Dispersion diagram. Curves computed for n ( r ) parabolic (solid lines) and
n(r) uniform (dashed lines). Points B from Fig. 1 fitted to curve for (a) n(r) parabolic
and (b) n ( r ) uniform.
contained a 500 kc/s component proportional to V,V cos CI where Vn is the ampli-
tude of the reference signal and V and M are the amplitude and phase of the received
signal. This component was amplified by a tuned amplifier of bandwidth 20 kc/s and
gain lo6, rectified and fed to a constant-speed chart recorder. When the receiver
probe was driven axially a t constant speed, waveform recordings such as those in
Fig. 4 were obtained. On varying the amplitude of the signal to the exciter probe it
was found that the amplitude of the recorded waveform was proportional to this
signal over at least 20 dB. The cylinder shown in Fig. 3 acted as a wave-guide for
which the exciter frequencies were well below the cut-off value. However, it was
necessary to earth the outer conductor of the receiver probe near the probe tip in
order to eliminate a spurious signal propagating between the outer conductor and
the cylinder. An evanescent electromagnetic wave was observed close to the exciter
probe (see (b) in Fig. 4) which could be greatly reduced by varying the phase of the
reference signal.
It should be noted that wave damping cannot be determined from the waveform
recordings since at any given value of w the density fluctuations in the plasma produce
corresponding fluctuations in k and therefore fluctuations in the phase angle a. The
detection system averages the wave signal over this spread in a and thereby produces a
 Dispersion of electron plasma waves 915

Exciter probe
'I -

brook
Plasma
M e t a l cylinder

Atten.

Crystal
mixer
Signal gen.
10-470-l200

(500 kd5)
detector
Modula:or
500 kc/s

Chart
recorder

FIG.3.-Apparatus.

FIG.4.-Waveform recordings at different electron densities. (Note: as no decreases,

the percentage density fluctuation usually increases and the waveforms become more
heavily damped.)
 916 H. G. JONES and R. N. FRANKLIN
P. J. BARRETT,

damped waveform on the chart recorder (see Fig. 4), even when the plasma wave is
undamped.
4. EXPERIMENTAL RESULTS
Figure 1 shows the observed dispersion at four different plasma densities, together
with points obtained by MALMBERG and WHARTON(1966). The parameters w, and a
are chosen to provide the best fit with curves computed for a parabolic profile n(r) =
no(l - r2,’u2)at different values of o,a/u,. U, determines the vertical scale of each
dispersion curve, a the horizontal scale and w,a/u, the shape, i.e. the change of slope
from small to large values of ka. The table below contains the fitted values of me,
U and w,a/u, (from which no and T, are calculated) for the five sets of experimental
points in Fig. 1 ; also some values of no and a from readings of the probe ion current,
and the end-plate temperature T, from optical pyrometer measurements.
The latter values of no are calculated from the ion current lo+ to the probe at a
potential of -20 V. It should be noted that at these low densities where estimates of
no from probe data are unreliable, we have used the fitted values of no to interpret the
measurements of lo+. For the set of points A the estimated thickness of the ion sheath
at -20 V is 1.1 mm (i.e. less than the ion Larmor radius, 1.8 mm) and its surface area
A , = 0.21 cm2; tzo is obtained from the expression for the ion current to the probe,
Zo+ = A,noed,/4 = A,n,e(KT,/2m1,)~’~.
For the sets of points B and C, where the ion Larmor radius ~ ( 2 k T , m , ) ~ / ~is/ e2.4
B
and 2-2 mm respectively (i.e. less than the sheath thickness calculated for zero magnetic
field), the probe is assumed to collect those ions whose guiding centres move within an
ion Larmor radius of the probe. If the projected area thus defined is labelled A,, no is
obtained from the relation
IO+ 2Apn0e(KT,/n2,)1‘2
1

/ ~ r.m.s. value of the axial component of the ion velocity and

where ( K T , / V Z , )is~ the
where A , = 0.18 and 0.16 cm2 respectively.
The inset in Fig. 1 shows radial profiles of the ion current to the probe which
are approximately of the form (1 - r2/u2)for small values of r. The fact that the area
of ion collection is comparatively large introduces little error into the profile shape
where r is small and in this region the profiles correspond quite well with the fitted
values of a. The profile corresponding to the set of points D is not shown because the
low values of both plasma density and magnetic field (300 G) made the effective
collecting area of the probe almost as large as the cross-section of the plasma column.
The points labelled B in Fig. 1 are also plotted in Figs. 2(a) and (b). This illustrates
the fact that almost the same values of o,are obtained when the five sets of experimen-
tal points are fitted to curves for n ( r ) uniform, using smaller values of a (chosen to be
the radius at which n(r) = $no); see the third column of the table.
5 . DISCUSSION
Figures 1 and 2 show that for a given radial density distribution i z ( r ) , the disper-
sion curve is defined by the values of the electron density no at the axis and the electron
temperature T,. We may represent this dependence in the form
o ( k ) =.[((no, T,) for ka 1>
o ( k ) = g(n,, n(r)) for ka < 1.
 Dispersion of electron plasma waves 917

I f a ( r ) / a ois measured independently, it is clear that the dispersion of these waves offers
a means of determining the parameters no and T,,
One approach to this problem is to insert n(r)/nointo a computer programme for
determining w ( k ) , as did MALMBERG and WHARTON(1966), obtain no from the
lower portion of the dispersion curve and thence find T, from the upper part of the
curve. It appears from the third column of the table, however, that the estimated
value of no is not sensitive to the form of n(r) and that an algebraic expression for
~ ~ ( v ) /(defining
n, a parameter a) is adequate to determine no over most of the range of
~ , a / u ,to within 10 per cent. The errors quoted in the table represent the limits to
which w9 can be varied and still fit a computed curve reasonably well.
TABLE
1.-PLASMA COLUMN PARAMETERS

From dispersion From probe

Exptl 2 WD no a T, no a T,
points U, (10' rad/sec) (cm-9 (cm) (OK) (cm) ("K)
~.
A 2200 6.55 i0.20 1.35 X 10" 1.08 <4,100 1.2 x 10" 1.2 2,200
(6.50)*
B 47 1.1.1 i.0.05 3.9 X 10' 1.30 3,100 2.5 X 10' 1.3t 2,200
(38) (1.17) (2,200)
C 27 0.58 i 0.03 1.1 X 10' 1.12 1,900 1.1 x 10' 1.1 1.900
(0.58)
D io 0.19 i 0.02 1.1 x 107 1.33 2,100 2,200
(0.19)
E 7h 0.89 i 0.06 2.5 X 10' 1.57 114,000 1.4 111,000§
(0.84) 2.2 x 10':

* Figures in brackets are obtained from curves for n ( r ) uniform.

7 From the dashed profile in Fig. 1, obtained under similar values of density, magnetic field and
oven temperature.
Malmberg and Warton's fitted value.
§ Malmberg and Warton's measured value of T,.

The errors in T, are expected to be greater than 10 per cent because the estimate of
the thermal velocity U, from a dispersion curve is subject t o the combined errors in U,,,
a and w,a/u,, the latter being especially uncertain at large values ( 2 100). It is
evident in Fig. 1 for example that points A would almost fit the curve for w,a/u, = 260,
which would reduce the value of T, in Table 1 to 2200". The low values of w,a/u,
are also uncertain because in this region ( U / W , ) ~ vs. ka is almost a straight line and by
suitable adjustment of w,, a and U,, an equally good fit is obtained over a range of
w,a/u,. I n the case of the points D for example, if a is multiplied by 1.09, w, by 0.94
and T, by 1.3, so that o,a/u, becomes 9, the experimental points fit the curve for
oj,a/v, = 9 about as accurately as that for w,a/v, = 10. Similar adjustment of the
points C reduces w,a/u, to 25 and increases T, by a factor 1.3. The only conclusion
that can be drawn from the above data concerning the value of T, for the sets of
points A, B, C and D is that it is of the same order of magnitude as T,, probably
within the range from T, to 1.5T,. It is interesting to note that the condition T, > T,
has been reported in two experiments on thermally ionized plasmas (HIRSCHBERG
et al., 1966 and SATOet al., 1968).
We conclude that the dispersion of electron plasma waves offers the most accurate
means of determining the density of a steady-state, very low density plasma, and also
gives a rough value of the electron temperature.
918 H. G. JONES
P. J. BARRETT, and R. N. FRANKLIN

Ack~iowle~~~riierits-Tliesewaves were first observed here by Dr. R . W. MOTLEYof the Plasma

Physics Laboratory, Princeton University. The authors are much indebted to him and to Dr. P. F.
LITTLE for initiating and supporting this work, and also to Mr. C. R . MIDDLETONfor helping to
run the experiment.
REFERENCES
BARRETT P. J., FRANKLIN R. N. and JONESH. G. (1967) Proc. Coif. Physics Quiescent Plus~tins,
Frascati, Part 11, p. 463.
BOHM D. and GROSS E. P. (1949) P1iy.s. R w . 75, 1851.
DERFLER H. (1965) Pr,oc. 7/h l i l t . Cor?/: loriiiotioiz P/zeiioriieim irz Gnses, Be(yrude, 1965; p. 282.
(Gradjevinska Knjiga) (1966).
DERFLER H. and SIMONEN T. C . (1966) Pliys. Rei). Lett. 17, 172; J . uppf.Pliys. 38, 5014 (1967).
DIAMENT P. et crl. (1966) J. up/)/. Plip. 37, 1771.
DRUMMOND W. E. (1964). P/iysics F/uids 7, 816.
GOULD R. W. (1964) Propagation of electrostatic waves along a plasma column. Unpublished
article.
GUNSHOR R. L. (1966) J . appl. Pliys. 37, 3630.
HIRSCHBERG J. G., HESSER J. E., HINNOV E. and RYNNN. (1966) Princeton University Plasma
Physics Laboratory Annual Report MATT-Q-24, p. 148.
KUEHL H. H. and STEWART G. E. (1966) J . appl. Phys. 37, 1724.
LEEK . F. (1966) J . appl. Plrys. 37, 4367.
LICHTENBERG A. J. and JAYSONJ. S. (1965) J . uppl. Plrp. 36,449.
MALMBERC J . H. and WHARTON C . B. (1964) Phys. Reu. Lett. 13, 184, 17, 175 (1966); (with W. E.
DRUMMOND) Proc. h t . Coif. Coiftrolled Fusion, Culhani Laboratory (1969, paper CN 21/116.
SATO N., IKEZI H., YAMASHITA Y . and TAKAHASHI N. (1968) Phys. Rev. Lett. 20, 837.
TRIVELPIECE A. W. and GOULD R . W. (1959) J. uppl. Phys. 30, 1784.