VOLUME 56, NUMBER23 PHYSICAL REVIEW LETTERS 9 JUNE 1986

Experimental Measurements of Phase Space

Roger McWilliams and Daniel Sheehan
Department of Physics, University of California, Irvine, California 92717
(Received 28 February 1986)
Direct measurements of integrated phase-space densities, e.g., fix, vy,t), have been made in an
experiment. Using spectroscopically active ions, measurements in a plasma show the ion response,
fiix,v,t), to linear and nonlinear waves and phase-space particle bunching. Time-resolved mea-
surements show coherent and incoherent phase-space density changes in the presence of waves, in-
dicating that transitions to turbulence and chaos may be studied. The time, space, and velocity-
space resolution may allow experimental tests of predictions from the Boltzmann equation.

PACS numbers: 52.25.Kn, 05.20.-y, 05.45.-fb, 47.25.Ae

In this Letter we report direct experimental mea- acceptance angle is large). Theorists have predicted
surements of integrated phase-space densities. The wave-particle trapping,4 bunching, 5 routes to tur-
measurements allowed phase-space densities integrat- bulence, 6 " 8 and phase-space clumps. 9,10 An experi-
ed over two velocity components in physical-velocity mental test of a stochastic theory has been reported. 11
space to be determined. The measuring technique was Phase-space coordinates for Eq. (1) involve three
nonperturbing and did not require inference from the physical-space dimensions and three velocity-space
use of derivatives of the data. Resolution of about 2.0 dimensions. The experiments reported here are of
jxsec, 10~ 3 cm3 and 3 x l 0 1 0 cm 3 /sec 3 in time, space, measurements of integrated phase-space density
and velocity space were obtained. Measurements were presented in a planar view involving one physical-
made on spectroscopically active ions in a plasma. Ion space dimension and one velocity-space dimension,
responses to small- and large-amplitude waves were having been integrated over the two velocity com-
observed, showing coherent and incoherent time- ponents perpendicular to the velocity-space dimension
resolved responses, and indicating that transitions to displayed. For example, measurements were made of
turbulence and chaos may be studied. Additionally,
phase-space particle bunching was observed. fi(x>vy,t)=J fiixfytzfvXfvyfv2tt)dvxdvzt (2)
Many differential equations arising in physics have with y and z fixed. Each individual scan thus
solutions of the form i|> = t|>(x, v , / ) , that is, the solu- represents a partially integrated distribution function
tions are phase-space density functions. Moments of similar to that obtainable with a plane electrostatic
these functions yield information such as number den- probe. A set of these measurements throughout the
sity, momentum density, energy density, etc. Among physical- and velocity-space regions of interest along
important equations in statistical physics are the with electromagnetic field data may allow testing of
Boltzmann equations, which can be written for plas- features predicted by the Boltzmann equation. In par-
mas ticular, a set of scans at various angles in velocity space
allows the complete, nonintegrated / , ( x , v,f) to be ob-
ML + v - V / i + Sl- E + v x B ML tained by use of tomographic reconstruction tech-
St niques. 12
Phase-space densities were measured via laser-
dfj_
, (1) induced fluorescence techniques, 13 as shown schemati-
dt I collisions cally in Fig. 1. A tunable dye laser induces a transition
predicting the dependence of the jth-particle distribu- to an excited state in a target particle when the laser
tion function fjix,v,t) on charge, mass, space, veloci- frequency is tuned to the necessary Doppler-shifted
ty, collisions, and electromagnetic fields. resonance frequency for the moving target particle.
Experimentalists commonly can measure number The target particle then emits a photon which is gath-
density along with electromagnetic field quantities, but ered through collection optics. The spatial resolution
a direct measure of / is difficult to achieve. The of the collimated laser beam and collection optics is 1
motion of number-density "blobs" has been report- mm 3 . The velocity resolution is limited only by the
ed.1 Energy analyzers have been used with increasing natural linewidth of the absorption line, corresponding
success, 2,3 but still suffer from two drawbacks. Firstly, to an uncertainty of about 3 x l 0 1 0 cm 3 /sec 3 in three-
the analyzer usually is physically larger than many dis- dimensional velocity space or about 3x 103 cm/sec in
tance scales of interest and perturbs the measured sys- speed. For this experiment sufficient photons are col-
tem. Secondly, commonly in laboratory experiments lected (by means of a boxcar technique) for a time
the distribution function is obtained by differentiation resolution of 2.0 ^sec.
of an integrated flux measurement, a technique requir- The experiments reported here were performed in a
ing considerable caution (especially when the analyzer single-ended Q machine 14 (see Fig. 1) which provided
© 1986 The American Physical Society 2485
VOLUME56, NUMBER23 PHYSICAL REVIEW LETTERS 9 JUNE 1986

COLLECTION
OPTICS LASER
BEAM
FIG. 1. Schematic of experimental apparatus showing
cylindrical plasma, the channel through which an electron
current is drawn to the button (which creates instability),
and the movable laser beam with detection optics for phase-
space density measurements.

a low-density (n—-5xl09 c m " 3 ) , low-temperature

(F, = Te ==0.2 eV), nearly completely ionized barium
plasma 1.0 m long and 5 cm in diameter. The confin-
ing magnetic field was 4 kG. Plasma density was in-
ferred from use of a Langmuir probe in conjunction
with the angle of propagation of a lower hybrid test
vy ( I 0 5 cm/sec)
wave with respect to B 0 (O^co/aypg). Electron tem-
perature was estimated with a Langmuir probe. The
ion properties were measured with the laser-induced FIG. 2. (a) Phase-space density in two quadrants of the
fluorescence phase-space diagnostic. x-vy plane showing undisturbed distribution on the right and
The coordinate system used for measurements and perturbation from small-amplitude instability on the left.
display has B = BQz. The x-y plane is perpendicular to The bar below the horizontal axis indicates the extent of the
instability source driving function (button location) in x
the magnetic field, with the origin at r = 0 of the From top to bottom the contours represent the 5%, 10%,
cylindrical plasma. Thus, the ion Larmor orbit is such 15%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, and 90% of the
that the ion angular momentum projection on the z peak value of 2.8xl0 4 sec cm"4, (b) Phase-space density
axis is negative. All measurements shown in the ac- for vy 5* 0, x = const, obtained from a laser scan in the re-
companying figures were taken from scans along the x gion of plasma x ^ 0 . 5 cm showing the Maxwellian nature
axis, holding y and z constant, collecting independent of undisturbed ions.
information. This means that the displayed data are
fi(x,vytt). surements of two quadrants of the x-vy plane at a fixed
For circumstances of cylindrical symmetry, such as time when a small-amplitude (e<f>/T < 1) ion instabili-
the application measured here, these data may yield ty is generated by an electron current drawn to the
the ft at other x,y positions through rotation about the button. Because of the symmetry of the driving sys-
symmetry axis. Also, for periodic phenomena involv- tem and the ion Larmor orbit, the remaining two
ing ion Larmor motion, fi(x,vx,t) are obtained by ex- quadrants may be surmised by a mapping of
amination of fi(x9Vy,t + Zirl2ti>ci). Some carefully
resolved periodic phenomena involving ion Larmor ft(xfv t)~+ fti-x-l, vy,t),
motion then can yield fj(x,y,vx, vy,t). The assistance i.e., a flipping of each point through x = - 0 . 5 , vy = 0.
of symmetry is lost in the x-z plane and a more com- Contours of equal phase-space density are drawn in
plete set of measurements can be made and then un- Fig. 2(a) as a function of x and vy. The peak value of
folded to yield fj(x,y,z, vXf vyf vz,t).n the phase-space density in the plot occurs along the
An electrically conducting, variably biased button of vy = 0 axis outside of the button region and has a value
6 mm diam was placed in an x-y plane and centered at of 2.8 x 104 sec cm""4. The contours follow fractions of
x— — 0.5 cm, j> = 0 located as shown in Fig. 1. This this value and from large vy to vy = 0 (top to bottom in
button may be dc biased or ac driven to produce an the figures) occur progressively at 5%, 10%, 15%,
electrostatic ion cyclotron instability. Coherent ion 20%, 30%, 40%, 50%, 60%, 70%, 80%, and 90% of the
response to this instability has been observed. 15 This peak value of 2 . 8 x l 0 4 sec cm" 4 . For a distribution
instability creates perturbations in phase space, details function which is isotropic and homogeneous in physi-
of which are discussed below. Figure 2(a) shows mea- cal space, contours parallel to the x axis would result.
2486
VOLUME 56, NUMBER 23 PHYSICAL REVIEW LETTERS 9 JUNE 1986

This is seen on the right of Fig. 2(a). Figure 2(b) through a wave period. Displayed in Fig. 3 are four
shows a typical data scan of ft (x = const,v y ) taken at plots of phase space taken at different phases during a
x > 0.5 cm which corresponds to a vertical cut through wave period with 80 -~ TT/2 between consecutive plots.
the right-hand region in Fig. 2(a) and one can see how As expected, the ions show a coherent response tied to
a homogeneous, isotropic Maxwellian would give the the wave phase. There is a density reduction in the
parallel contour pattern displayed on the right-hand current channel corresponding with a density rise out-
side of Fig. 2(a). This figure also shows that ions side the channel which is most pronounced for
whose orbit passes through the button current channel 0 = 37r/2. There is a "tongue" of phase-space density
are affected by the instability for this experiment. oscillating at co and reaching out in xand vy. By calcu-
Hence, perturbations to the phase-space density are lating Larmor radii for ions passing through the button
observed only to the left in the figure. Ions whose channel one may conclude that the tongue consists of
paths may have x > x button will display phase-space particles which are near vx = 0 both when in the button
changes for vy < 0 only, consistent with the mapping channel and at the observation point outside the chan-
among quadrants in the x-vy plane (to see this, consid- nel. Additional observations 16 on the distribution
er the Larmor orbits of various ions which pass function measured parallel to B0 show a phase-
through the button current channel, where they cross resolved v2 bunching of the ions in the z direction. To
the x axis again, and the instantaneous vy at these generalize the concept of bunching, this tongue is a
points). measurement of phase-space bunching in four dimen-
Time evolution of phase-space density is shown in sions, three velocity dimensions and one spatial
Fig. 3. The ion wave (with e<f>/T < 1) frequency was dimension. We note that more dimensions can be ex-
about co = 3.2 x 105 sec" 1 . A sampling window in time amined when tomographic techniques are applied.12
corresponding to one-tenth of a wave period allowed As examples, particle bunching is thought to play a
measurements of phase-space density to be followed role in such wave-particle phenomena as the two-
stream instability and pulsar radiation mechanisms. 17
Additionally, there is some heating within the current
channel. As an aside, properties such as the energy
density u(xtt) and the velocity-dependent energy den-
sity u(x,vy,t), momentum density, etc., may be cal-
culated from this type of figure.
When the wave was driven nonlinearly to large am-
plitudes (e<t>/T» 1) the phase-space plot shown in
Fig. 4 was obtained. This phase-space density was

-I 0 -2 -] 0

x (cm) x (cm)
FIG. 3. Phase-space density at increasing phases during FIG. 4. Nonlinearly driven phase-space response, essen-
one period of a small-amplitude instability. The bar below tially independent of wave phase, i.e., time independent.
the horizontal axis indicates the extent of the instability The bar below the horizontal axis indicates the extent of the
source driving function (button location) in x. The contours instability source driving function (button location) in x
have the same definition as in Fig. 2(a). The contours have the same definition as in Fig. 2(a).

2487
VOLUME56, NUMBER23 PHYSICAL REVIEW LETTERS 9 JUNE 1986

found essentially to be independent of wave phase (in space, and speed resolution of 2x 10~ 6 sec, 10~ 3 cm 3 ,
contrast to the case e<f>/T< 1). The ion response in and 3 x l 0 3 cm/sec, respectively, have been obtained.
Fig. 3 was coherent and could be time resolved, Measurements were made of unperturbed phase space,
whereas for the conditions of Fig. 4 no coherent ion along with linear and nonlinear wave effects. Density
response could be found within the diagnostic time reductions and enhancements were observed as well as
resolution. Hence, at some intermediate wave ampli- particle acceleration. Coherent and incoherent re-
tude not studied in this experiment the ion response sponses showed linear particle responses and that a
changed from coherent to turbulent. For this large- transition to turbulence occurred. Phase-space particle
amplitude case, the waves in the plasma were bunching was seen.
comprised of electrostatic ion cyclotron waves above The authors thank Dr. Nathan Rynn for discussions
many of the ion cyclotron harmonics, but the role of and Mr. Stacy Roe for technical assistance. This work
the amplitude and spectral widths of these waves has was supported by National Science Foundation Grant
not been determined yet. The button bias does have a No. PHY-8306108.
strong effect on the parallel drift of the ions. The
average parallel drift speed of the ions is reduced in
the button channel 16 by the wave. The density reduc- *S. J. Zweben, Phys. Fluids 28, 974 (1985).
tion in the channel is due to a raising of plasma poten- 2
R. L. Stenzel, W. Gekelman, and N. Wild, Phys. Fluids
tial in the channel and to radial outflow of ions from 26, 1949 (1983).
3
the channel, as evidenced in the figure. Some parti- P. F. Mizera etal, J. Geophys. Res. 86, 2329 (1981).
4
cles are accelerated to large |v| by the potential oscilla- R. C. Davidson, Methods in Nonlinear Plasma Theory
tions in the channel. Figure 4 shows measurements of (Academic, New York, 1972).
5
vy acceleration creating a tail in f(vy) in the channel. S. P. Gary, M. F. Thomsen, and S. A. Fuselier, Phys.
These particles have a Larmor diameter of greater than Fluids 29, 531 (1986).
6
1.5 cm (bringing them near the plasma edge) and also M. J. Feigenbaum, J. Stat. Phys. 19, 25 (1978).
7
would be found in the phase-space quadrant of posi- S. Newhouse, D. Ruelle, and F. Takens, Commun.
Math. Phys. 64, 35 (1978).
tive x and negative vy. An enhanced density of fast »M. Kono, Phys. Fluids 28, 1494 (1985).
particles is seen for — 1.9 < x < —1.0 cm which cor- 9
T. H. Dupree, Phys. Fluids 15, 334 (1972).
responds to particles with negative vy in the channel l°T. H. Dupree, Phys. Fluids 25, 277 (1972).
orbiting to x values where they have a positive vy on ll
F. Doveil, Phys. Rev. Lett. 46, 532 (1981).
12
the x axis. For similar tail particles to those shown in R. Koslover and R. McWilliams, University of Califor-
the channel, but with negative vy in the channel, a nia, Irvine, Technical Report No. 86-18, 1986 (to be pub-
change in phase-space density would be expected near lished).
13
x = — 2.4 cm. Port restrictions on the vacuum vessel D. N. Hill, S. Fornaca, and M. G. Wickham, Rev. Sci. In-
restrict the laser diagnostic to U | ^ 2.0 cm, and so op- strum. 54, 309(1983).
14
tical information near the edge of the plasma is una- N. Rynn, Rev. Sci. Instrum. 35, 40 (1964).
15
vailable for this experiment. R. A. Stern, D. N. Hill, and N. Rynn, Phys. Rev. Lett.
47,792 (1981).
In summary, direct, nonperturbing measurements of 16
A. Lang, Ph.D. thesis, University of California, Irvine,
phase-space density integrated over two velocity com- 1984 (unpublished).
l7
ponents have been made in an experiment. Time, F. C. Michel, Rev. Mod. Phys. 54, 1 (1982).

2488