You are on page 1of 31

LABORATORY PLASMA DEVICES

R. JONES Plasma Physics Research Institute, Physics Department, University of Natal, Durban, Rep. ofSouthAfrica

NORTH4IOLLAND PUBLISHING COMPANY - AMSTERDAM

PHYSICS REPORTS (Review Section of Physics Letters) 61, No. 5(1980)295-325. NORTh-HOLLAND PUBLISHING COMPANY

LABORATORY PLASMA DEVICES R. JONES*


Plasma Physics Research Institute, Physics Dept., University of Natal, Durban, Rep. of South Africa Received October 1979

Contents: I. Introduction 2. General selection criteria 3. Characteristic operating parameters of common laboratory plasma devices 297 297 309 4. Closing remarks and acknowledgements References 324 324

Abstract: We review the characteristics of the common types of laboratory plasma and the devices used to generate them. After discussing the relative importance of the various plasma parameters we give brief descriptions of the salient features of each device. This review is intended to aid the reader in the selection of a plasma source suitable for his own (unspecified) program of research.

Present address: Plasma Physics Research Laboratory, Box 44, Colonia, New Jersey 07067, U.S.A.

Single orders for this issue PHYSICS REPORTS (Review Section of Physics Letters) 61, No. 5 (1980) 295325. Copies of this issue may be obtained at the price given below. All orders should be sent directly to the Publisher. Orders must be accompanied by check. Single issue price Dfl. 16.00, postage included.

R. Jones, Laboratory plasma devices

297

1. Introduction Amongst the many hundreds of laboratory plasma generators which have been constructed throughout the world [1] one can discern various distinct family trees. In order to design a bench top plasma system for laboratory research purposes [2] one must first compare the needs imposed by his program objectives with the performance and critical parameters obtainable from each family of devices. The purpose of the present review is simply stated; it is to provide a starting point from which to make an intelligent selection in the design of a laboratory plasma device specially suited to the needs of a given researcher. In section 2 we will begin the review by listing and discussing some of the general selection criteria which should be considered. Some of these items directly involve the plasma properties which can be obtained while others are related to economic constraints or matters of reliability, versatility, etc. Keeping his own priorities firmly in mind the reader can then apply these general criteria to 17 common plasma sources which we catalog and describe in section 3. 2. General selection criteria It is useful to begin with a discussion of the various features which may be considered desirable in a laboratory plasma generator. Any list of such criteria [2] would certainly include: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. Range of plasma density obtainable. Degree (and range) of ionization. Range of electron temperature and degree to which the distribution is Maxweilian. Range of ion temperature and degree to which that distribution is Maxwellian. Range of magnetic field (if desired). Degree of plasma quiescence. Homogeneity obtainable. Range of gases employable. Reproducibility. Cost to build and maintain the source. Ease of engineering. Ease of operation, etc.

These criteria are, of course, strongly interrelated and their optimization is, often times, mutually exclusive. It is important at this point for the reader to stop, consider his research objectives, abilities, and resources, and then assign some sort of relative priority (weighting) to each of these ideal criteria. Later on, when faced with a choice between two desirable characteristics one should be in a position to say what he values most highly for his own purposes, and what other features he is willing to compromise. The required physics parameters and the available manpower and budget are usually the easiest to quantify and a good place from which to start. We will now proceed to discuss each of the basic criteria in its turn. 1. Range of plasma density The absolute plasma density,

~e,

will be important in any plasma physics experiment, at the very

298

R. Jones, Laboratory plasma devices

least, insofar as it affects the Debye length:


ADI

/ ____ T ~h/2 I 4lTnee \


2

where Te is the electron plasma temperature and e is the electronic charge. If one is to have a plasma at all n~must be large enough that, for given Te, AD is much smaller than the dimensions of the experimental chamber. Other constraints may be imposed by the diagnostics one hopes to employ. Standard microwave interferometry, for instance, places a lower bound on the product of plasma density and microwave path distance traversed, which is, commonly, iO~ cm2. Unless one is willing to generate a very long (or large) plasma this translates into a restriction on minimum density. Similarly, if one wishes to use standard Langmuir probes [3] the electron saturation current:

Ie.~t = neeAp

(21e)h/2

(2)
tie

where A~is the probe collection area and me is the electron mass, may set an upper bound on given Te) in order to avoid overheating and damaging the diagnostic. The electron and ion plasma frequencies:
=

(for

n e2 (~~)
1/2

(3)

and

(njZ2e2\1~12
irm~ 1
(4)
tie

where Z is the ionic charge state, may also be important to a researcher and place limits on

1. If fle is too large the wavelengths of any oscillations under investigation may then become too small (compared to the probe) to measure. Degree of ionization If one is interested, primarily, in the behaviour of a pure or fully ionized plasma one would like to minimize the effects of plasma-neutral particle interactions. Such criteria will place lower bounds on the degree of ionization which cart be tolerated. Since the various collision cross-sections depend upon the filling gas employed we must restrict our general comments to qualitative observations. The plasma frequency (and perhaps the cyclotron frequency as well) should, for instance, far exceed the neutral collisionfrequency. This criterion is obviously easier to observe for electrons and electron phenomena than for ions. Since neutral collision frequencies are proportional to n~, the neutral density, (and not n ~j2, analogous to the plasma frequency) it is, however, best to work at low plasma density if the degree of ionization, tie/Pin, is fixed for some reason. In many common gas discharge devices the degree of ionization may, in fact, be limited by the phenomena of arc starvation. Arc starvation [41 occurs when
2.

R. Jones, Laboratory plasma devices

299

the mean free path of a (refueling) neutral entering the plasma volume: A=
ne(oVe)

(5)

becomes less than the (desired) plasma dimension, L: A<L (6)

where V~ is the neutral influx velocity, n~ is the density of ionizing electrons, and (o-V~)is their rate coefficient for gas ionization. If this phenomenonoccurs the plasma will become hollow, extinguishitself, or develop instability. If the plasma is (desired to be) Maxwellian then arc starvation can be avoided only if: Vfl
fle(OVe)Te

>L

(7)

where ( )r, is an average taken over the Maxwellian distributed electrons of temperature Te. V~, may be taken to be the average neutral gas speed (nearly the room temperature value in many discharge devices). In the next section we will derive discharge equilibrium equations which allow us to recast (7) in the form:
VnTion loss

n~

where r10~ ~ is the average ion containment (or loss) time. For low gas pressure (low n~)unconfined discharges the Langmuir free fall model [5] allows us to write: (9) 2. For a reasonable choice of parameters it becomes where the ion acoustic fractions speed, Cs = (Te/mi) difficultC5 to is obtain ionization much above ne/nn 3 VJC. 0.1. (Note the temperature dependence.) This limitation on the degree of ionization obtainable in the common gas discharge plasma can be relaxed in several ways, however. In higher temperature plasmas (T, or Te) ions tend to be neutralized on the wall and return as relatively energetic neutrals [6]. These neutrals have large V~ and hence A. Furthermore, if T 1, the ion temperature, is raised then V~, can be increased by ionneutral collisions. Mechanical means [7] are also available with which to increase fle/fln. The plasma could be created in a high neutral pressure environment, under conditions of low fle/Pin, but be allowed to stream (along a magnetic field for instance) into a differentially vacuum pumped experimental volume (fig. 1). Differential pumping can be costly and it is advisable to use tubular apertures between the experimental volume and the source and plasma dump regions (fig. 1). (Note that the plasma current into the dump just balances the neutral current out; niviAa~pure = nnvnApumping duct.) The ionizing plasma present in the tube significantly enhances the vacuum pumping impedance of the aperture and so reduces the overall pumping requirements [71. (Such action is also useful in Tokamak divertors.)
loss =
-~

3L/C5

300
F A

R. Jones, Laboratory plasma devices

___

iip~_
1~
><
I

_ I ~ F>(

I ~_

5<

Fig. 1. Differentially pumped plasma device having: hot conical tantalum filament, F; anode, A; source chamber, S; experimental chamber, E; plasma dump chamber, D; solenoid magnets, M; and aperture tubes, T. Typical length: 12 meter, typical diameter of plasma column: 13 cm.

Such a device can be thought of as a variation on the hollow cathode discharge but in which one has separated the cathode and aperture elements. Since the aperture tube need not emit electrons any longer it can run cool and neednt be thin walled. If the magnetic field lines are allowed to diverge at the ends of the chamber a large cathode/aperture area ratio results which has improved heat loading characteristics. In general, of course, power is required in order to raise the degree of ionization for a given fill gas pressure (and physical damage may be associated with increased input power levels). Pulsed operation can be resorted to but only by complicating the diagnostics, increasing costs, and possibly degrading reproducibility, reliability, etc. Electron temperature and distribution function The electron temperature (along with n~) determines the classical plasma conductivity [8] and plays a role in many collective processes. Unfortunately, Te is not a truly independent parameter in a given laboratory plasma source. In a time independent steady state the ion loss rate must balance the volume ionization rate (surface ionization will be considered under a discussion of alkali metal plasma devices):
3.

T~

= n~n~(oV~).
0~ loss

(10)

Here we have assumed that n,. must be kept low and ionelectron recombination occurs solely at the walls. If we further require that the electron distribution be purely Maxwellian the gas ionization is

R. Jones, Laboratory plasma devices

301

solely attributable to the tail of the distribution and (10) becomes:


=
Piefln(t.TVe)T.

Tion loss

(11) n1: (12)

and, since by plasma charge neutrality, ne 1


Tj0fl loss = Pin(tTVe)T~

Using eq. (9): 12 (~Ve)~~ ( ) (T~/M1) (L V/A where V is the plasma volume and A is the ion loss area). As the rate coefficient is strongly dependent upon T~ the solutions of (13) have the general form of fig. 2, regardless of the gas assumed, and Te becomes (for moderate Te) an inverse function of the neutral filling pressure [9]. Eq. (13) is, of course, not universally valid [10, 11]. At low Te (less than about ~the gas ionization potential) plasma electrons lose their ionization ability. Similarly, at very low neutral pressure no discharge equilibrium exists at all [9] (fig. 2). Generally, eq. (13) is applicable only for highly Maxwellian discharge plasmas in the electron temperature range between about 5 and 50 eV. Discharges at low pressure (na) and low Te can be dominated by primary electron-gas ionization [911]. If primaries are numerous the ion loss rate, eq. (9), can be modified. Usually, however, one wishes to keep np/ne small and the plasma quasi-Maxwellian (n~is the primary density and we now have, rigorously, n~, + Pie n 1). In this limit the ion loss rate can still be governed by the (abundant) thermalized plasma electrons (according to eq. (9)) while gas ionization is dominated by the (few, but energetic) primaries. In that limit (10) becomes:
=

1
flfl

3L
(TeIM1)

p,

14

O~O

Fig. 2. Semi-log plot of the discharge equilibrium eq. (13), (for hydrogen) with VP/A in cm-torr and Te in electronvolts.

302

R. Jones, Laboratory plasma devices

where ( ) is now an average over the primary electron population only. Such an admixture of primaries may, of course, complicate analysis of an experiment. They may also stir up electron beam-plasma instabilities [12, 13] unless their distribution can be kept isotropic (such as by employment of numerous, spatially well distributed source filaments or surface magnetic scattering fields [11]). Very low temperature electrons can also be supplied to an ion rich plasma (e.g. synthesized plasma) if hot filaments are arrayed near the point of ion injection [14]. 4. Ion temperature and distribution function Most basic plasma physics processes depend only weakly on the ion temperature which is fortunate since many common sources generate plasmas having unknown or roughly room temperature ions. Because of the mass disperity (and momentum conservation) most heating mechanisms favour electrons. At the same time ions are more readily cooled by neutral collisions (such as charge exchange). In order to study processes which are dependent on ion temperature (cusp leak widths, for example [15])one must place more emphasis on low pressure operation or try to heat the ions directly. One way to do this would be to perform RF heating at some suitable ion resonance like the ion cyclotron resonance or a Buchsbaum resonance. Alternatively, one might use shocks, plasma beams, or electron space charge ion heating. (Electron space charge heating of ions is common in laboratory plasma devices [16].) One frequently finds that large scale gradients of plasma density exist in research plasma sources. These gradients may be due to inhomogeneous plasma generation or loss, for instance, and, in accordance with the Boltzmann relation:
V~(Z) -

V~(Z =0) =

n ln

[tie?~0)]

(15)

where V~(Z)is the plasma potential, such density gradients must be accompanied by plasma potential gradients (provided the electrons are in thermal equilibrium, i.e. Te is roughly homogeneous). (Assuming low gas pressure operation we will ignore collisional effects.) Such spatial gradients may be undesirable in themselves but they have long been known as a source of substantial space charge heating of diffusing ions [16].Briefly, since ions created by electron-gas ionization are born nearly at rest and at the local plasma potential [17] ions originating at various spatial locations in the discharge fall through different potential differences in traversing the device. This electron space charge heating substantially enhances the ion temperature; to a value which is only slightly less than T~[15, 18, 19]. Such T1 enhancement is generally not isotropic, of course. (Neither is ion cyclotron heating.) Space charge ion heating is especially convenient in synthesized plasma devices, wherein ions and electrons are supplied by external (independently controllable) sources and mixed in an experimental chamber [20].A spread ion beam [21](perhaps with rather low drift velocity), or several different beams, perhaps from tandem ion sources (fig. 3), can be used to form a plasma with widely variable (and high) T1. Many sources and injection angles could be employed in order to improve isotropy. 5. Range of magnetic field The existance and variability of a magnetic field is, of course, necessary for confinement, diffusion, and transport experiments. Furthermore, the presence of a magnetic field can reduce the plasma loss rate (such as by reducing A and increasing L = V/A in eqs. (13) and (14)) and, hence, the power required to sustain the discharge [10] (at adequate density, temperature, and percent ionization).

R. Jones, Laboratory plasma devices

303

III)

I,
A

FF
~

E Ti

+~4~-~

Fig. 3. Multiple plasma device electrically biased for tandem ion source operation. The target (experimental) region is marked T, filaments F, and

anodes A. Typical length: 1 meter.

As one approaches higher magnetic fields (1 kilogauss), however, one finds the economics of the solenoid itself becomes paramount. Magnets must be purchased, powered, and cooled. If discrete coils are used field ripple distorts the shape of the plasma and, hence, its homogenity; when a continuous solenoid is approached diagnostic access diminishes (and wide angle viewing windows are often useful). Small surface cusps, for primary electron containment, are readily supplied by permanent magnets. Large volume uniform fields are nearly always provided by solenoids. The magnet cost and power demand becomes a strong function of physical size and field strength; a rough estimate coming from the equationfor an infinite solenoid is:
B = 1.2566 x 102

NI/L1

(16)

where B is the field in Gauss, L1 is the length of the coil in meters, I is the current in the coil in amperes, and N is the number of turns of wire (or water cooled tubing) making up the coil. Eq. (16) significantly overestimates the field produced in short coils, however. A better estimate is given (for the field on axis) by:
B~ Z ~2~

11Z+~~ 1 Jr 2J12l L\ 2) n ~ri+[r~+(Z+Li/2)2]112J 2+[r~+(Z+Lj/2)

~ \

2)

1 Jr 2]12 ~tri+[r~+(ZLi/2)2]12 2+[r~+(ZL1/2)

( 17 )

where / is the current density in the coil, r

1 is the inner coil radius, and r2 is the outer coil radius. If the average coil diameter (not the inside diameter, 2r1) is d then the total length of wire required becomes Nird and the voltage needed to drive the solenoid is: v=IpNird/a
(18)

where p is the resistivity of the conductor (at operational temperature) and a is the conductors

304

R. Jones, Laboratory plasma devices

cross-sectional area. For high fields multilayer coils are used and the average coil diameter, d, greatly exceeds 2r1 and the plasma diameter. External magnets require (of course) that the vacuum chamber be non magnetic. Furthermore, (exterior) surface magnetic cusps only become efficient if the wall is unusually thin. Aside from economics, as the field is raised various new plasma instabilities may occur. These will reduce quiescence or degrade the plasma [22].
6. Plasma quiescence

Plasma quiescence is commonly defined by the value of dfle(t)Ifle dlsat(t)/Isat or d V~(t)I Te where Vf is the (probe) floating potential. Where transport is concerned one might, ideally, prefer that dne/ne be as close to zero as possible (i.e. perfect quiescence). When studying plasma waves one prefers that only a few oscillations be present; those of the type one wishes to investigate. Stray oscillations, then, are undesirable. This ideal is almost impossible to achieve. High gas pressure can damp low frequency oscillations preferentially but even high pressure discharges (and in the absence of magnetic fields) are unstable to so-called striations or ionization instabilities. Lower pressure discharges are subject to Volterra type ionization fluctuations [23], the nonlinear, oscillatory solutions to the time dependent ion and neutral gas particle balance equations: dn1
= flj72n(tTVe)i~ 3

C5n1A

(19)

and dn0 = n~,V~A fli~n(OVe~~s.


(20)

In this regard primary electron sustained discharges are more quiescent [11] than are those sustained by thermal electron ionization (eqs. (19) and (20)). For primary sustained plasmas eqs. (19) and (20) become [11]: dn1/dt = dn0/dt =0. Primary electrons, however, are difficult to confine (unless one invests in surface magnets) and the percentage of input power that ultimately goes into plasma thermal energy is lower (110% in the absence of surface magnets, versus 2550% for thermalized plasmas).
7. Plasma homogeneity

It is often difficult to obtain a homogeneous experimental plasma. Because of the advantages of low gas pressure operation electronion recombination may occur solely at the walls; similarly, the generation of the plasma may well be localized. Furthermore, as we have seen, non uniform plasma density also implies non uniform plasma potential. The plasma potential, itself, must satisfy the Poisson equation: 2V V 0(Z) = 4ire(n, Pie) (21)

R. Jones, Laboratory plasma devices

305

(in a one dimensional model) or, in dimensionless form [24]:


S

(~)

2v2q(s) =

exp [n(S)]

(22)

where lengths have been scaled in terms of


I

AD

and:

eIl1iJ

IAIsl/2
T~

23

n~ o~e

which is just the (low pressure) discharge equilibrium equation [10] obtained earlier assuming Maxwellian distributed electrons. The normalizations employed are:

n(S) = eVp(S)ITe
and
S=Z/L.

(24)

(25)

Equation (22) can be solved numerically, along with the Boltzmann equation, eq. (15), for boundary
conditions such as: =0, dq/dS =0 for S =0

in order to obtain [19]both fle =

compared with experimental results (fig. 4) obtained in a cylindrical axially magnetized plasma column [19]. Setting up and solving the Poisson equation may be difficult in arbitrary geometry (spherical
fle(Z)

and V~ =

V~(Z).This calculation is

geometry is also rather elementary, however [24]).Plots like fig. 4 can, of course, be used to obtain
(quantitatively) the ion space charge heating discussed earlier.

Fig. 4. Left-hand curves: Plasma electron density as a function of distance. Dashed curve theory. Solid curve experiment. Right-hand curves: Normalized plasma potential as a function of normalized distance. Solid curve theory. Dashed curve experiment.

306

R. Jones, Laboratory plasma devices

Clearly, the plasma in fig. 4 is only homogeneous near the center (S = 0) of the discharge and it is common for researchers to build a large plasma de~ice and then restrict experimentation to this small region. (The useful volume must, of course, be large compared to probes and any plasma wavelengths which may be of interest.) Alternatively, if one is willing to tolerate a non-Maxwellian plasma (perhaps by only a percent or two in tip/Pie) one could add the homogenizing effect of a uniform distribution of gas ionizing primary electrons. At higher gas pressures eq. (13) is invalid and collisions will modify the plasma-gradients. In a noisy (non-quiescent) discharge plasma waves can also contribute an homogenizing influence [25]. In fact, instabilities such as drift waves exist (in the presence of a magnetic field) specifically because of the gradients in plasma pressure, fleTe. Consequently, one can often expect to make trade-offs in plasma homogeneity versus quiescence [22]. We will, however, say no more with regard to cross-field plasma gradients. A detailed treatment would depend upon our model of the way in which plasmas transport across magnetic induction, a subject which is only now being attacked in earnest [25].

8. Range of gases
It may be desirable to create plasmas from a variety of elements in order to vary the ion mass and the electron-to-ion mass ratio. Many plasma devices are able to employ a wide range of filling gases, though metal or other impurities may be injected inadvertently (as well). Furthermore, one may obtain both atomic and molecular ions (for example H~, H~, H~) of varying ionization states (such as H~, ~ etc.). If a plasma source runs hot then such items as plastic insulators may outgas badly. In the case of RF driven discharges, for instance, the Teflon insulated coaxial cable that carries microwave power into the plasma chamber may outgas, break down, or even become a source of a Teflon contaminant plasma. Where surface ionization plasma sources are employed the range of working gases becomes dependent upon the work function of the electrodes as well as the gas ionization potential. This ultimately restricts operation to only a few elements, the alkali metals like cesium and potassium.
9.

Reproducibility

An ideal plasma would exhibit the same physical behaviour day after day, or pulse after pulse, dependent upon some externally controllable dial settings. Unfortunately, plasmas are nonlinear creatures, not restricted to a single unique operating mode, and, for their very existence, they must insulate themselves from their surroundings. As an experimenter one is, for the most part, confined to changing certain exterior boundary conditions which have an indirect (and complicated) influence on the plasma properties themselves. Furthermore, we have seen that many of our desired idealizations are, to a greater or lesser extent, mutually exclusive. We can help to improve reproducibility by application of our understanding; a lower magnetic field, for instance, may be perfectly acceptable and lead to more stable operation [22]. We can also improve the situation by proper servicing and maintenance; filaments and other components deteriorate, changing their characteristics with time, and should be periodically replaced. We should also design with great care; saving money on power supplies, by resorting to pulsed operation, might be no savings in the long run. 10, 11, 12. Cost, ease of engineering and operation Without resorting to the labor theory of value it should none-the-less be apparent that what is

R. Jones, Laboratory plasma devices

307

simple and easy to build and operate is also least expensive (unless, of course, one has inherited components from a forbear). Unfortunately, what is mechanically simple and reliable (and, hence, inexpensive) may not be what is simple from a physics viewpoint. Under such circumstances a compromise is called for based on the relative proportions of scientific expertise and coin-of-the-realm available. As this is no place to try to condense a philosophy of experimental science we will elaborate only on the material issues. Before building any device the reader should sit down, draw up a sketch embodying his ideas, identify the components required to turn it into hardware, and then look up the actual price of each item required. In some countries high quality surplus electronics, power supplies, and vacuum hardware are available and should be considered. In todays poor scientific climate some of us have even obtained plasma physics hardware (as gifts or at nominal cost) from various existing national laboratories. Usually this material was retired due to the vicissitudes of funding agencies and not due to obsolescence. Of course, improved reliability also reduces long-term costs and one should not resort to dangerous gambles on string-and-sealing-wax. In order to assess the cost and mechanical complexity of the various plasma sources in some very general way we have employed a crude ranking system in table 1. Cost and complexity escalate as one proceeds to higher vacuum and higher magnetic fields. Furthermore, if two (or more) independent plasma chambers (etc.) are called for, for instance, then the implied increase in complexity also adds to the cost and headaches. The vacuum requirements of a system needing only a mechanical pump (chamber pressure ~10_1 torr, e.g. glow discharges and some arcs) are clearly less than those of a device which requires both fore and diffusion pumping (pressure <102 torr). Similarly, devices requiring successively stronger magnetic fields will entail greater cost and complexity. Using an increasing cost-complexity scale of 0 to 4 we have ranked the various laboratory plasma devices on this basis (table 1). These assignments are, necessarily, somewhat subjective and apply to some sort of average device from each class. The power balance (and efficiency) of a given plasma device will also contribute to its cost (principally tied up in power supplies). We can not treat the power balance for each device but a zeroth order estimate for a thermalized gas discharge may prove enlightening. The power input to the plasma source, P, must balance that lost to the walls (in the free fall model) [10]: $PA~2Te+A~(Ei+eVp) where E1 is the gas ionization potential and V~, the plasma potential, is approximately:
=

(26)

ln (3

(27)

assuming Z = 1 and neglecting secondary emission and T~ 4 Te. Charge neutrality demands that:
Je = Jj ~ei~e(Te/Mj)h/2

(28)

and f3 is the efficiency of coupling power into the plasma; /3 ~for beam-plasma [12, 26] or RF driven sources [27,28]. The results are plotted in fig. 5 for a hydrogen plasma. Obviously, if radiative losses are significant [29] /3 must be decreased somewhat.

308

R. Jones, Laboratory plasma devices

~j

iV
:u ~ ~

~ L~i I
~

3 ~
~

.~

.~

,-~

fl~.

~
8

::: ;. : :;;: :H::; :


~ ~ ~

2
~ ,

a
~

I ~
o 1 ~

000

~
000

Q~

~i~i

0000

e
~1 0

:ii
000

a a
n

a a
n

aaao
nfl n 66~6

,.~ ,.. ,- aOOa

~i =a~

6
6

,~ 6~6

Th

~ ~
F-.

6...~ .7
I

6,766
I

66!~ ~~
I

6
~

.~

~
0 0

~aa

~1
o

o~~ ,.
~T 0

aa~o

a~oo

000

~ 6~6

!
~-

66~7 ~
~6

666

,7!~ 6aa

I itli i~i I

:~

R. Jones, Laboratory plasma devices

309

woo
Fig. 5. Semi-log plot of the discharge power balance eq. (26) (for hydrogen, $

~) with P/An, in watt-cm and T, in electronvolts.

3. Characteristic operating parameters of common laboratory plasma devices Table 1 summarizes the characteristic operating parameters of the laboratory plasma sources we intend to review. In a sense this table can be considered a highly condensed summary of the material contained in this section. The remainder of the review is intended to elaborate on the characteristics as tabulated and to provide a thumb-nail sketch of each of the different machines. The parameters given are believed to be representative but can never be exhaustive. It is especially important to recall again that important plasma parameters are usually interrelated and to improve one value it is necessary to degrade some other parameter. One can expect to obtain parameters listed as typical simultaneously, but one can not expect to obtain a high value for a given parameter without degrading some other characteristic (perhaps to its low value). Much of the data presented has been obtained from the literature. (The author, like most plasma researchers, has had personal experience working with only about half the devices named.) Some of these plasma systems constitute well developed technologies which are now fully matured. Others, however, are less well developed. Moreover, what one worker may be willing to accept as a typical operating point may, in fact, require hours of painstaking adjustments to obtain. Another man may on the other hand, quote typical parameters of a plasma which has not fully warmed up. We have sought to reduce this source of subjective error by consulting a variety of references. This will, however, tend to broaden the ranges of the parameters listed (beyond that likely to be available on one given experimental set up). Our next major caution, then, is that one should not expect to obtain the full range of (possible) parameters in a single experiment. Adjustments would be required in order to cover the full range; sometimes rather major ones.
1. The glow discharge

In the uniform positive column of a glow discharge [30]electron-gas ionization is balanced by radial plasma loss to the chamber wall:
d2fle ldPie M
+~+fflefln(OVe)T0

1p

(29)

310

R. Jones, Laboratory plasma devices

where i. is the ionneutral collision frequency. The experimental arrangement is shown in fig. 6. The electron temperature falls typically in the range of 1 to 5 eV and energy is supplied to the electron population by a longitudinal current flow driven by an ambient, roughly constant, axial electric field, fig. 7. The gas and column length are arbitrary, provided that the applied axial voltage drop is adjustable so as to maintain the required axial plasma potential gradient (i.e. electric field). The electron temperature decreases with increasing gas fill pressure (and column radius) and is dependent on the gas ionization potential: exp(eEi/Te) C 2 (eE1/T~)112 ( pr)

(30 )

where C is a tabulated constant [31]. The fill pressure typically ranges from atmospheric down to 10 ton and the degree of ionization is usually low (perhaps Pie/fln 10~). The peak plasma density (commonly 10 cm3) is limited by the external cooling available and various practical constraints on the peak axial current (10 amp/cm2). At low pressure the quiescence is often limited by the destabilization of ionization instabilities [32]. Such striations are the nonlinear oscillatory solutions to the coupled particle and energy balance equations. Striations can also be stationary, or quasi-stationary, in which case strongly inhomogeneous density and electron temperatures are observed. The presence of an axial magnetic field also reduces radial particle diffusion and enhances noise and inhomogeneity.

2. The arc
The glow discharge device just described employs a cold cathode. One can go over into the arc regime of discharge operation in various ways (including cathode self-heating). For our purposes we will not be concerned with historical terminology and simply suggest the introduction of an auxiliary cathode heating system (i.e. hot cathode operation). This usually involves either a wire filament placed behind the thin disc cathode assembly (fig. 6) or else the use of a biased filament, itself, as the cathode. Such arc-like operation can allow one to run the discharge at much lower pressures (10~to 10 torr) and somewhat higher plasma density, as indicated in table 1. The corresponding increase in percent ionization leads to some reduction in neutral collision effects and enhanced plasma noise. If the gas pressure is not too low (~ 10_2 ton) the arc again possesses a positive column and so we can relate the arc current density to the plasma density [31]:

j = eneKeE

(31)

Fig. 6. Diagram of the electrical circuit of a gas discharge tube havinganode and cathode discs D andballast resistor R and the characteristicdischarge current-voltage curve. (The scales are variable. The glow regime being the central plateau region of the characteristic.) Typical length: 0.510 meter. typical diameter: 0.1S cm.

R. Jones, Laboratory plasma devices

311

E\

N~~J~___

Fig. 7. Characteristic axial dependences of light output I, potential V, electric field E, positive charge density N+, electron density N, and electron current density J, for aglow discharge column. (The cathode ison theleft andthe central region isthe positive column. Thescales are variable,see table 1 and the text.)

through the electron mobility K0, and the axial electric field E. In the positive column the electron distribution is quasi-Maxwellian, possessing a significant axial drift in accordance with (31). Power balance is similar to our earlier discussion, the input power: 2jV~airr2jEl (32)
P=irr (where V is the anode-cathode potential difference neglecting electrode effects, r is the column radius,

and I the column length) is balanced by radial particle convected heat loss [31]:
P
2ITTITe fleDa/A (33)

(where Da is the ambipolar diffusion coefficient and A is the average of electron and ion mean-freepaths) plus radiation loss (usually roughly comparable in magnitude to (33)). At the lowest fill pressures obtainable the electron-neutral collisional mean-free-path exceeds the plasma dimensions and the positive column model is invalid. In such a limit one applies, instead, the particle and energy balance equations described in section 2.

312

R. Jones, Laboratory plasma devices

3. Hollow cathode discharge

Tne ordinary glow and arc discharges are usually limited to operation at low percent ionization [31] and (in fig. 1) we have illustrated means for the employment of a differential vacuum pumping system in order to reduce the neutral pressure in the experimental region [7].In order to avoid excessively large (and fast) vacuum pumps one wishes to employ the ionization enhanced gas impedance phenomena; a
simple way to do this is by use of a so-called hollow cathode, fig. 8.

A hollow cathode [33]is a section of thin wall metal tube (typically ~ 1 cm inside diameter and a few cm long) which electrically serves as the cathode and which, simultaneously, acts as the input tube for the sole (external) source of fresh fuel gas. The hollow cathode runs hot and is a source of thermionic electrons. These primaries are electrostatically accelerated into the discharge plasma which forms inside of the cathode tube [34].All gas emerging from the cathode must, thence, pass through a very intense
ionizing region. Furthermore, the gas pressure inside the tube is substantially higher than that in the

(vacuum pumped) experimental volume. Percent ionization in and near the cathode approaches 100% but neutral reflux from the walls can reduce this value elsewhere. A hollow cathode must run hot and is subject to damage (melting frequently occurs). Ignition can be by overvolting or Tesla coil or one can add an auxiliary filament and externally heat the cathode. Hollow anode operation has also been successfully achieved [35]. Large cathode diameters are prohibited due to mechanical problems though some success has been reported with a cluster of smaller tubes. The requirement of a hot and thin metal tube is one structural
difficulty. Such cathodes are usually made of tantalum or copper and while the active (inner) end must

run hot the outer end (leading through the vacuum chamber wall) needs to be cooled.
Gas ionization occurs both inside the cathode (primary dominated) and outside, in the experimental region [33] (thermal ionization). The admixture of these two modes of operation (at different pressures) complicates the overall hollow cathode discharge theory. Hollow cathode devices may be operated with,

or without an applied magnetic field.

Fig. 8. General arrangement of a hollow cathode discharge device having vacuum pump P, vacuum chamber V, electrical insulator I, gas inlet G,

hollow cathode C, and anode disc A. Typical length: 1 meter, typical diameter: 0.2510 cm.

R. Jones, Laboratory plasma devices

313

4. Multifilament devices

The hot cathode arc-like discharge can be operated at very low fill pressures (<10~ torr) thanks to primary electron-gas ionization. Such a discharge will not have a strictly Maxwellian electron distribution, of course. Furthermore, the primary electrons may constitute a beam which will excite electron beam-plasma instability [13] and degrade quiescence. This later instability is stabilized, however, if the primary distribution can be made isotropic, and several means exist for accomplishing this isotropization [11].Large numbers of hot cathodes (i.e. filaments) can be arrayed about the plasma volume so that even though primaries are lost after one transit of the plasma the average distribution is still isotropic. Alternatively, one could employ surface magnetic fields (see Multidipole devices) to scatter the electrons. Under temperature limited operation the filament emission (primary) current density is given by the
Richardson law:

j = CRT2 exp

WIRT)

(34)

where W is the filament work function, T is the filament temperature, and CR is the (tabulated) Richardson coefficient. Alternatively at high filament temperature, space charge limited operation takes over and the primary current density becomes:
j

en

0(T0/21Tm0)

127

(35)

where y = y(Vp/Te) is a parameter of order unity. Two or more such plasma sources can be joined together in the manner illustrated previously in fig. 3 in order to form a double (or n) plasma device [36].Briefly, two independent multifilament (or multidipole) devices are connected together, separated by an electrically biasable fine mesh grid. If this separator grid is biased sufficiently negatively electrons will be unable to pass freely back and forth between chambers. One then finds he is able to maintain each of the tandem plasmas at a different plasma potential simply by differentially biasing the different chamber walls (discharge anodes) [36]. Ions, of course, flow between the different plasmas according to the potential difference imposed and such devices then become especially useful for beam-plasma and flowing plasma (e.g. shock) experiments. In order to generate a homogeneous plasma one needs a uniform primary density across the plasma volume. Hence the mean-free-path for primary-gas collisions must exceed the dimensions of the device and many (most) primaries are lost to the walls without doing any ionization whatever. For this reason the energy efficiency is rather low (only 110% of the input power appears as plasma thermal energy). (In the absence of surface magnets.) Very small filament diameters imply a large ohmic voltage drop across each filament. This may be desirable for matching to common laboratory D.C. power supplies and contributes to a spread in primary electron energy. However, finer diameter filaments are more readily damaged by operation in an oxygen contaminated gas environment. Oxygen partial pressures of ~10~ torr are necessary for 0.005 inch diameter tantalum; whereas larger diameter ifiaments are less easily damaged. This difficulty may sometimes impose the most severe constraint on required system base pressure.
5. The backstreaming arc

Double (multiple) plasma devices, being sustained by primary ionization, can have relatively low

314

R. Jones, Laboratory plasma devices

electron temperature and a high degree of quiescence. It is possible to use one plasma chamber as a

source of ions which can then be extracted and mixed with electrons elsewhere, in an experimental chamber. If these neutralizing electrons are (all) supplied by hot filaments then one can hope to keep the final electron temperature to values which are just greater than the filament temperature. On the other hand, if the filaments are (few and) relatively cold then space charge fields will develop to confine electrons (and in so doing heat them). Depending upon the separator grid bias (and the ratio of grid spacing to local Debye length) some electrons may also escape the ion source chamber and be characterized by the electron temperature in that (source) discharge. Such devices [14,37] may take the form of the plasma source shown in fig. 9 and produce plasmas with 0.2 eV ~ T0 ~ 2 eV. Unless a large source chamber is used the density will often be rather low, 3. One can, however, hope to vary fle nearly independently of Te. 105_ba cm
Synthesized plasmas The backstreaming arc and double plasma devices are really examples of a more general class of plasma sources, the synthesized plasmas [20]. A synthesized plasma is one which is manufactured by actually mixing independently (and externally) supplied electrons and ions. One might, in fact, have independent electron and ion guns which fire into an experimental chamber. The desired advantages
6.

come from having independently controllable ion and electron source parameters (energy, density, energy distribution, etc.). Since a plasma is a good source ofelectrons, and one of the few reasonable sources of ions, a tandem 3 chamber device is suggested; all chambers contain plasmas, one as a source of ions, one a source of

_1~~1
H

L~
Fig. 9. Funnel-type back streaming plasma source with gas inlet I, anode A, grids G, funnel F, and filament cathode C. Plasma outlet, 0 is directed into a vacuum chamber of typical volume 12 cubic meter.

R. Jones, Laboratory plasma devices

315

electrons, and one synthesized from ion and electron beams. Of course the plasma parameters obtainable are not really as controllable as one might like. We can not, of course, violate the neutrality condition and fle n1. Still, ion and electron influx rates can be varied and, in fact, the plasma potential
adjusts itself accordingly so as to confine the rarer species. These devices have the disadvantage of being more complicated and having (ion and electron) species which are characterized by non-zero drift energy in the laboratory reference frame. This later

defect can be reduced by employing an increasing number of beam sources, but only by further complicating the design; Childs law applied to either beam source implies that high densities are
achieved only if high drift velocities are acceptable. This restriction is usually not much of a problem for electrons but does severely limit ion beam density.

To reduce overall system complexity most synthesized plasmas are of the double plasma type [36], wherein no separate electron source chamber is provided but electrons are emitted instead by hot filaments in the mixing (target) chamber itself.
7. Afterglows

Processes associated with creating, dispersing, and thermalizing a discharge may (all) contribute substantially to plasma fluctuations. By pulsing an arc or other plasma source one may hope to do (pulsed) measurements in the (transient) diffusing afterglow. Such plasmas will resemble their (D.C.) source discharge but, in general, will be more tenuous, cooler, quieter, and, of course, time varying. Being pulsed they will normally require less in the way of power supplies. Diagnostics become more complicated, of course, and aftergiows can still be noisy (due to fluctuations associated with loss processes, for instance). The penalties that must be incurred in density and temperature can, of course, be offset by resorting to the aftergiows of unusually intense sources such as sheet pinchs, shock driven discharges, or gun produced plasmas. (Shock tubes will be treated separately in subsection lb. Gun produced plasmas can

have the advantage of preferential ion heating provided that drifts are acceptable or can be thermalized.) Afterglow plasmas are also longer lived when they are formed in a proper confinement geometry. A very weak sustaining current has even been employed to prevent recombination in a stellarator confined afterglow. Unfortunately, complex confinement geometries complicate the interpretation of any laboratory experiment.
8. Photoionized discharges

Up until now we have considered discharges for which energy is input directly as energetic charged particles or by the action of D.C. electric fields on ambient plasma electrons and ions. Alternatively, one might irradiate a gas with photons [38,39] or microwaves. Photoionization depends upon a source of intense light having hw > E1 such as the spark source [40]shown in fig. 10. Such photons can cross any magnetic field present but must not be overly attenuated before reaching the target gas. For this reason the spark source is placed inside the vacuum container. Furthermore, only a small fraction of the power input to the source appears as light output so pulsed operation is employed. Intense vacuum sparks suffer considerable wear and reproducibility is degraded. The spark, itself, is a discharge and can mix with the photoplasma unless prevented from doing so by an intervening region of transverse magnetic screening fields. (Time-of-flight separation could also be used.) Photointensity also falls off as one moves away from the (poorly collimated) source, and plasma inhomogeneity will result

316

R. Jones, Laboratory plasma devices

High Voltage Peedthrough

Alumina Spacer

Molybdenum Cathode

Molybdenum Anode

Fig. 10. Spark-type UV source as developed at U.C.L.A. Designed for mounting in a 1 cubic meter vacuum chamber.

(unless one installs numerous spark sources, which would prove very difficult to pulse simultaneously and reliably).
9. Multidipole plasmas

From fig. 2 it is clear that low pressure gas discharges are only possible if sustained by primary ionization [9].In order to stabilize any primary electron beam-plasma instabilities, however, one must insure an isotropic primary distribution in the experimental volume. One way to do this was to array numerous source filaments about the plasma. Alternatively, surface magnetic (multidipole) cusps can be added to the chamber wall so as to scatter the primaries. This has the added advantage of confining the primaries and increasing the ionization that they are capable of [41].Such a device is illustrated in fig. 11. Because of the enhanced primary residence time [42]the energy efficiency is improved over the multifilament device and can rival that observed for Maxwellian electron sustained discharges (i.e. 50% of the input power appears as plasma thermal energy). Plasma electrons and ions are also confined but evidently this is of lesser importance [43]. At high neutral gas pressure cross field diffusion widens the cusp leak width and the multidipoles are relatively less efficient. At ~102torr the addition of surface cusps, at an expense of hundreds of
American dollars, may only double the ambient plasma density. At lower pressures (~10~ torr) density improvements of one or two orders of magnitude can be obtained. Since the cusp fields only penetrate a short distance into the plasma the core is left nearly field free [41]. The small volume occupied by magnetic field also makes permanent magnets more economical

than solenoids (under most conditions).


I

10. The Penning discharge

Penning suggested that ionizing electrons be confined to oscillate along magnetic field lines and between negatively biased electrodes. Plasma ions would be lost freely (but relatively slowly) to these (hot or cold) cathodes and electrons could only diffuse radially, across the magnetic field, to an exterior anode. Such a Penning, Phillips, Reflex or P.I.G. discharge [22,44, 45] is shown in fig. 12 and, like the multidipole device, can be said to have a magnetically insulated anode [22]. Once again, the magnetic insulation is particularly useful in confining any primary electrons present.

R. Jones, Laboratory plasma devices

317

N SN

Filamer*

9
Chamber/Anode Fig. 11. The multidipole source as developed by Limpaecher. (Only afew representative surface magnets are shown.)Typical plasmavolume: meter.

1 cubic

Unfortunately, the ion loss to the end electrodes, which is limited to the ion saturation value: I, = 2irr~n0ZeC~ may not automatically balance the electron diffusive loss across the magnetic field:
=

(36)

2irleD1n0

(37)

and instabilities may have to develop in order for the plasma to maintain itself neutral (22). Also, the
system will not work at all in the absence of an ambient magnetic field.

__

TA
H-

Fig. 12. A typical reflex plasma device with magnetically insulated central anode disc A, filament F, cathode C, magnets M, and vacuum chamber V. Typical length: 12 meter, typical plasma diameter: 25 cm.

318

R. Jones, Laboratory plasma devices

more quiescent discharge can be regained by decreasing the (relative) electron confinement in

some way (and with it the discharge density and/or efficiency). This can be done by reducing the electrical bias on one or both ends, increasing neutral ~,ressure,decreasing magnetic field, lengthening

the device, etc.


The presence of an oscillating but nonisotropic primary population also gives rise to beam-plasma instabilities and the reflex discharge (with or without strong electrostatic axial confinement) can also be called a beam-plasma discharge [46, 47]. Beam-plasma instability may not interfere with certain types of experiments or one may manufacture the plasma in one chamber of the device and then allow it to

flow into a second, experimental chamber (fig. 1) [13].In the later case one may hope to confine RF noise to the source chamber via ambient plasma RF damping or by use of a conducting grid placed at the aperture between chambers (to reflect RF waves).
-

In such beam-plasma devices [13]RF fields transfer energy from beam particles to the plasma proper [12]. Obviously a magnetic confining field may be useful but is no longer essential in this limit. The energy utilization efficiency of these devices can be quite good; as much as 5080% of the input power may go into plasma thermal energy [12, 26]. As with any cathode one can apply various chemical coatings in an attempt to lower work functions and increase current emission. A simple way of doing this is to use an aerosol spray gun to paint on a mixture of lanthanum hexaboride in methanol. The methanol evaporates leaving behind the chemical film covered cathode, ready for insertion into the plasma device. A slow run-up to cathode operating temperature and subsequent discharge operation activates the cathode. 11. Shock tubes Pulsed plasmas can also be generated in chemical or magnetic shock tubes [48]. In the chemical shock tube a low molecular weight gas at high pressure is separated from a higher molecular weight gas at low pressure by a diaphragm. The gas pressure difference is programmed so as to burst the diaphragm and the higher pressure component acts as a piston to drive a shock in the low pressure

target chamber. Since the escape speed of the driver piston can exceed the sound velocity in the target gas a strong shock forms which ionizes the shocked target material.
Higher temperature plasmas can be produced by using a magnetic shock tube. Typically a capacitor

bank is discharged between two electrodes placed at one end of the partially evacuated tube. This spark
may be simply struck between two point electrodes or, for example, the plasma focus geometry [49] can be used. In the later arrangement the electrodes are two concentric metal cylinders, coaxial with the surrounding glass vacuum tube and with each other. Gas breakdown occurs across the insulator that

supports the electrodes and the resulting radial current sheet is driven down the annulargap between the
electrodes by an azimuthal self magnetic field. Very intense plasmas can be formed in this way using quite inexpensive apparatus. A single one kilojoule (fast) capacitor charged to 10 kilovolt is quite adequate and vacuum requirements are minimal (1 ton). The plasma is very short lived, however, and difficult to diagnose. Various instabilities may also occur.

Plasmas having thermonuclear characteristics can be (briefly) formed in an intense focus near the end of the coaxial electrodes. For the purposes of the present review (and table 1), however, we concern
ourselves with the research plasma parameters usually generated in the larger volume target region. 12. RF discharge Time varying electric (and magnetic) fields can accelerate charged particles and input energy to

R. Jones, Laboratory plasma devices

319

sustain a discharge [50].These fields may be established by antennas emersed in the plasma itself or one may try to transmit RF power across an intervening spatial gap and into the plasma. A gas filled microwave cavity operated at some appropriate RF frequency is one common experimental arrangement. The discharge regime varies according to the RF frequency and amplitude, the presence of any plasma resonance processes, the antenna coupling method, and the plasma (and neutral gas) parameters. At very low frequency and high fill pressures one may simply excite transient glow discharges of alternating polarity. Where wall impact takes place ionization may occur or secondary
electrons may be produced. As one proceeds to higher frequency electric fields one eventually reaches a

point where the reflexing electrons can no longer oscillate between the chamber walls. The energy utilization efficiency of an RF discharge depends, of course, on the ability of the antenna structure to electronically couple to the plasma. That is, the antenna impedance must be matched to an effective plasma impedance [28].Furthermore, if the antenna is not physically embedded in the plasma
itself then one must first couple to waves which can propagate across the vacuum gap, reach the plasma, and ultimately couple in turn to actual plasma oscillations [51].

If the RF field amplitude (and by implication, the RF input power) is very high then one can of course, drive forced oscillations (which are not normal modes of the plasma dielectric medium). Energy economy, however, suggests lower input power and coupling to some one of various possible plasma waves. An antenna then, will be chosen so as to operate at the frequency of, and to have the spatial periodicity (wavelength) compatible with, some known normal mode of the (desired) plasma. Furthermore, unless one is interested in research on strong turbulence or the like, we restrict our attention to those plasma waves which will be damped in a distance which is shorter than the
dimensions of the discharge, thereby giving up their field energy to plasma thermal energy. This becomes more difficult in a collisionless plasma. Conversely, if the input RF waves are damped too quickly all the energy may heat the plasma edge, creating an inhomogeneous discharge.

By way of example, the total power coupled from the antenna to the plasma can be found by integrating the Poynting vector over the antenna area:
P=JS.dA.

(38)

For a magnetized plasma and electrostatic waves in the regime w~ w~ one can use Maxwells equations and the standard dielectric tensor to give the real time averaged components [28]:
.~ 1

5.

~-

0K11E~

(39)

S~=~rEOK~lE~

(413)

with k~/k~ = K~1/K~ (41)

where the notation is standard and z is the magnetic field direction. The antenna is taken to be a

320

R. Jones, Laboratory plasma devices

distance d away from the plasma and an additional evanescence layer of thickness a is assumed. For an antenna of characteristic wavelength 2~r/k~ we obtain:
E~ k~Ve~e~. (42)

Writing a in terms of the plasma gradient scale length dN/dx:


=

w20m0 2/R

43

a_e2dNIdx

and using the definition P ~V


R

we obtain the effective antenna impedance:

A~ exp[+4ird/A~] exp[+(4irw2 2 dN/dxA 0m0)/(e 5)] 4i~0A [w(w2/w2_ 1)~][(1 w~~/w2)Lf2]

44

(A factor of 2 appears to account for waves propagating in either direction.) This equation assumes cold plasma theory. A better agreement with experiment has required a warm plasma correction where the term

1)1~2

in (44) is replaced by:


2 Re(zz) (45)

with
~ where Z(u) is the FriedConte plasma dispersion function (46)

5XT0/m0) For a laboratory plasma (Te = 1 to 10 eV) of density 107_lOb cm3 and B = few hundred gauss (where there is adequate plasma-wave damping) experimental power balance measurements have established that at least 75% of the power input to the antenna is absorbed by the plasma provided [28] that the antenna length is at least: I= where: ZY=(a+j$)2
Z~Rohmic+~WL
Y~+j~

[52], with u = (w/k

2.

(47)

.1

R. Jones, Laboratory plasma devices

321

Fig. 13. Typical RF couplers. Left: Dipole and currentloop, and Right: Coaxial cavity, with coaxial feedsC, plasma P, electrical insulator I, and vacuum chamber V. Typical plasma diameter: 2Scm.

are calculated for the appropriate (FEM) antenna mode. Rohmjc accounts for the resistive loss in the antenna elements. Because of the dependence on R we find that 1 is strongly dependent on (increases with) d + a.
13. Electron cyclotron resonance discharge

In the special case of magnetized plasmas one can take advantage of the excellent coupling of external antenna structures to plasma waves occurring near the electron (or ion) gyrofrequency. One very popular antenna geometry is the helical Lisitano coil [53] but most any reasonable configuration seems to work adequately. Fig. 13 shows the half wavelength long resonant cavity employed by the
F M H
/

I
Fig. 14. An alkali metal 0 machine having solenoid magnet M, plasma column P, vacuum chamber V, hot plate H, oven outlet 0, and filament

heaters F. Typical length: 1 meter, typical plasma diameter: 25cm.

322

R. Jones, Laboratory plasma devices

author to generate an E.C.R.H. plasma at a frequency of 3 GHz [54]. A coaxial geometry was employed fed by semi-ridged coaxial cable. The frequency chosen took advantage of inexpensive surplus military magnetron C.W. power tubes but implies a fairly strong magnetic field at resonance (1070 gauss). Similar tubes are available at low cost in common commercial microwave ovens. Operation is also possible at lower magnetic fields if resonance is achieved with some multiple of the cyclotron frequency. Unfortunately, the coupling efficiency drops at these higher resonances and one is trading off magnetic field cost for a larger microwave power tube. The tapered center conductor (fig. 13) is intended to improve plasma uniformity since E.C.R.H. plasmas can tend to be hollow. (The hollow center conductor design was chosen solely to permit energy analyzer access to the plasma [54].) Coupling efficiencies for E.C.R.H. generated plasmas generally range between 10 and 80%; a substantial loss occuring in the coaxial cables (type N in our device) connecting the magnetron to the plasma device, and then leading the power on into the vacuum chamber (0.28 inch O.D. semi-ridged copper jacketed cable in our experiment). All cable connections should be carefully adjusted if input power is expected to exceed 50 watt. Cable reflections can be serious and lead to electrical breakdowns. Fortunately, one can maintain a reasonable research plasma (Te 5 eV, tie 1O~ cm3) with only 10 or 20 watt of microwave power. Since microwave source tubes operate over a rather limited range of frequencies only a narrow band of magnetic fields can be employed in these devices.
~

14. Ion cyclotron resonance discharge I.C.R.H. sustained plasmas [55] operate in a manner similar to E.C.R.H. devices but at lower RF frequencies, f~/f~ = m~/m 1.Antenna structures differ for this reason and, most important perhaps, ions are directly heated. Megahertz power oscillators are (again) readily available and coaxial cables are a suitable method of transferring the power. Stix coils (i.e. spatially periodic induction coils [51]), simple wire loops, or metal plates situated in close proximity to and parallel with the plasma have been used as antennae (fig. 13) and prove relatively efficient for generating (or heating) laboratory research plasmas. The efficiency of I.C.R.H. may be a bit below that of E.C.R.H. and again the frequency and magnetic field are not as widely variable as one might like. Rotating plasma devices Crossed electric and magnetic fields can accelerate charged particles by driving an E x B drift motion. This occurs to a greater or lesser degree in most magnetized discharges since the existence of a plasma density gradient transverse to B implies, by the Boltzmann relation (eq. (15)), a radial electric field. In Penning discharge geometry (fig. 12) it is possible to impose a relatively large radial electric field by simply making the anode inside diameter larger than the cathode (outside) diameter and so sustain a substantial E x B azimuthal plasma rotation (modified Penning discharge). This heating method has the advantage of driving ion as well as electron drift and, hence, heating ions directly [56]. In fact, T, can be made to exceed T~by a substantial amount (e.g. an order of magnitude or more) in such devices. Unfortunately, one may not fully thermalize this drift energy and the rotation is also a source of free energy that can drive plasma instabilities [57]. Rotating plasmas may also be inhomogeneous (appearing for example, as a rotating radial spoke or density maximum [56]) due either to the formation mechanisms (such as cathode/anode spots) or to macroscopic instabilities (such as exchange and KelvinHelmholtz instabilities).
15.

R. Jones, Laboratory plasma devices

323

16. Laser produced plasmas Laser-pellet irradiation makes possible the rapid creation of dense, energetic, and highly localized plasmoids. Such plasmoids can, for instance, be created~ in situ deep within a system of confinement fields or they can be allowed to expand and be studied in the afterglow [58]. Because of the very small initial size, the extreme density required for laser light absorption, and the inherently transient character (required by the overall low energy efficiency) laser initiated plasmas are rather difficult to diagnose. The plasma formation process and subsequent expansion mechanisms are, themselves, poorly known and worthy of extensive study in their own right [58].The ambipolar electric fields established during the plasma expansion phase may also be responsible for setting up convective cells in the confinement volume. Extremely pure and intense plasmas can be created by laser pellet heating but the composition is limited by the availability of solid state target materials. The experimental device itself must be quite complex, perhaps comparable to the alkali metal Q devices that will be described in the next subsection. Laser produced plasmas seem especially suited, then, to thermonuclear plasma confinement studies and other closely related experiments. 17. Alkali plasmas If a neutral atom impacts a hot metal surface it may be ionized upon contact and return as an ion, accompanied by a thermionically emitted electron [59, 60]. This process is energetically allowed provided that e4w > E~,where 4~ is the work function of the metal surface. Such surface ionized plasmas were among the first reasonably quiet plasmas (i.e. dne(t)/ne low) produced in the laboratory and they were given the name Q machines, the Q standing for quiescent. These devices (fig. 14) produce plasmas which are in fairly good thermal contact with the source hot plate (usually tungsten) and hence have T T1 Te 0.2 eV. This can be varied only by application of some auxiliary heating method. The fuel neutrals are often input through a nozzle as an atomic beam (originating from a specially designed, and complicated, source oven) and percent ionizations approach3). The electron ing 100% are obtainable at reasonable power levels and densities (up to ~b013cm thermionic emission rate must be adjusted so as to nearly match the atomic beam flow rate and the desired outgoing plasma flux. Relative influx rates can be used to adjust the plasma potential or percent ionization somewhat. Recombination is an important loss mechanism in Q devices and one defines the recombination coefficient by the relation:
(48)

For this cylindrical magnetically confined system [60] plasma flows from the source hot plate according to the usual fluid equation: m~n~[3V 1/3t+ ViVi/dZ]~Tit9fle/3Z. The equation of continuity for ions is: 3n1/t+ 3(n~V1)/3Z+ an~ =0
(50)

(49)

324

R. Jones, Laboratory plasma devices

where V1 is the ion drift velocity, which, for a Maxwellian distribution: 2/T~) n(V)=no(V=0)exp(mV gives:
(1 mV2/2T) V/8Z = an

(51)

2/4T) 0 exp(mV

(52)

and so (with m 1V~4T): V ano(L Z) (53)

where V = 0 at the end plate and L is the plasma column length. Equations (51) and (53) serve to model the degree of (axial) plasma homogeneity attainable in 0 devices. 0 plasmas [61] are limited to materials for which ediw > E~ and are generally magnetized. Since the appropriate working gases ate -high atomic number alkali metals the corresponding ion gyroradii are rather large. In order to obtain a plasma many gyroradii in diameter one must either operate at very high magnetic field or else employ large diameter devices. Either of these limits implies substantial solenoids and accompanying power supplies. Due to the low electron temperature the 0 plasma will tend to be fairly collisional even when nearly fully ionized. Finally, these devices are by far the most complex laboratory plasma sources we have described. Vacuum requirements can be especially exacting; if one desires 90% ionization at a plasma density of, say, 1010 cm~then the base pressure must be ~5 X iO~torr. 4. Closing remarks and acknowledgements The material summarized in this review was assembled by the author during most of calender year 1978. The author compiled this collection of data to assist him in the design of a new, general purpose, plasma research device which he hopes to build shortly. We have found it a useful asset during the present machine design phase of our work and hope that it may prove similarly valuable to the reader. The author acknowledges financial support from the South African Atomic Energy Board, Pelindaba, South Africa. We also acknowledge useful experiences working on the various plasma experiments at the Stevens Institute of Technology, Hoboken, New Jersey, U.S.A., the General Atomic Company, San Diego, California, U.S.A., and the University of Natal, Durban, Natal, R.S.A.

References
[1] See, for example, World Survey of Major Facilities in Controlled Fusion Research (I.A.E.A., Vienna, 1976). [2] F.F. Chen, Lecture Notes from the Plasma Physics Summer Institute (Princeton University, 1964) p. 35. 131 P.M. Chung, L. Talbot and K.J. Touryan, Electric Probes In Stationary And Flowing Plasmas (Springer-Verlag, Berlin, 1975). [4]G. Fuchs, I.E.E.E. Trans. Nuc. Sci. 19 (1971) 160. [5] L. Tonks and I. Langmuir, Phys. Rev. 34 (1929) 876. [6] R. Jones, Phys. Lett. 67A (1978) 194. [71R. Jones, Vacuum 29 (1979) 27. 18] L. Spitzer, Physics of Fully Ionized Gases (Wiley, New York, 1962). [9] R. Jones, J. md. Inst. Sd. 60 (1978) 71.

R. Jones, Laboratory plasma devices [10] R. Jones, Pramana 12 (1979)1. [11] R. Jones, Plasma Physics 21(1979) 399. [12] R. Jones, Nuovo Cimento 40B (1977) 261. [13] M. Seidl, W. Carr, D. Boyd and R. Jones, Phys. Flu. 19 (1976) 78. [141T. Honzawa, J. Phys. Soc. Japan 42 (1977) 1077. [151R. Jones, Plasma Physics 21(1979) 505. [16] I. Langmuir, The Collected Works, ed. C. G. Suits (Pergamon, New York, 1961)Vol. 5, p. 151. [17] D.A. Dunn and S.A. Self, J. AppI. Phys. 35 (1964) 113. [18] R. Jones, Plasma Physics 20 (1978) 717. [19] R. Jones, Plasma Physics 19 (1977) 259. [20] T. Honzawa, I. Phys. Soc. Japan 36 (1974)1674. [21] R. Jones, Pramana 13 (1979) 329. [22] R. Jones, J. Plasma Physics 20 (1978) 221. [23] R.J. Roth, Phys. Flu. 12 (1969) 260. [24] A. Caruso and A. Cavaliere, Nuovo Cimento 26 (1962) 1389. [25] R. Jones, Can. J. Phys. 55 (1977) 1356. [26] R. Jones, Plasma Phys. 19 (1977) 925. [27] R. Jones, W. Carr and M. Seidl, Phys. Flu. 20 (1977)791. 128] R. Jones, Plasma Phys. 20 (1978) 713. [29] R. Basile and J.M. Lagrange, Nuc. Instru. and Meth. 31 (1964) 195. [30] G. Francis, Handbook der Physik, Vol. 22 (Springer-Verlag, Berlin, 1956). [31] J.D. Cobine, Gaseous Conductors (Dover, New York, 1958). [32] 1. Grabec and S. Mikac, Plasma Phys. 16 (1974)1155. [33] L.M. Lidsky et al., J. AppI. Phys. 33 (1962) 2490. [34] D.J. Willins and R.L.F. Boyd, J. Phys. D 6 (1973) 1447. [35] R.A. Jacobsen and H.P. Eubank, Plasma Phys. 15 (1973) 243. [36] RJ. Taylor, K.R. MacKenzie and H. Ikezi, Rev. Sci. Instru. 43 (1972)1675. [37] K. Takayama, H. Ikegami and 5. Aihara, Proc. 8th Inter. Conference on Phen. in Ionized Gases, Vienna (1967) p. 552. [38]J.S. DeGroot and K.R. MacKenzie, Phys. Rev. Lett. 12 (1968) 907. [39]J. Robin and K.R. MacKenzie, Phys. Flu. 14 (1971)1171. [40] E.R. Ault, U.C.L.A. Physics Dept. report, PPG-116, April 1972. [41] R. Limpaecher and K.R. MacKenzie, Rev. Sc Instru. 44 (1973)726. [42] K.N. Leung, T.K. Samec and A. Lamm, Phys. Lett. 51A (1975) 490. [43] K.N. Leung, G.R. Taylor, J.M. Barrick, S.L. Paul and R.E. Kribel, Phys. Lett. 57A (1976)145. [44]J. Backus, J. AppI. Phys. 30 (1959) 1866. [45]F.F. Chen, R. Bingham and W.L. Harries, Phys. Rev. Lett. 8 (1962) 234. [46] 1. Alexeff et al., Phys. Rev. Lett. 10 (1963) 273. [47]W.D. Getty and L.D. Smullen, J. AppI. Phys. 34 (1963) 3421. [48] R.A. Alpher and D.R. White, Phys. Flu. 2 (1959) 153. [49]J.W. Mather, Phys. Flu. 8 (1965) 366. [50]G. Francis, Ionization Phenomena In Gases (Butterworths, London, 1960) Ch. 4. [51]T.H Stix, The Theory of Plasma Waves (McGraw-Hill, New York, 1962). [52] B.D. Fried and S. Conte, The Plasma Dispersion Function (Academic, New York, 1961). [53]G. Lisitano, Proc. 7th Inter. Conference on Phen. in Ionized Gases, Beograd (1966) Vol. 1, p. 464. [54] R. Jones, Ph.D. Thesis, Stevens Inst. of Tech., Hoboken, N.J., 1975, Xerox University Microfilms, 76-5026. [55] T.H. Stix and R.W. Palladino, Phys. Flu. 1(1958) 446. [56] J.R. Roth, Phys. Flu. 16 (1973)231. [57] B. Lehnert, Nuclear Fusion 11(1971) 485. [58] G.C. Goldenbaum and K.A. Gerber, Phys. Flu. 16 (1973) 1289. [59] M.A. Allen and G.S. Kino, Phys. Rev. Lett. 4 (1961)163. [60] N. Rynn, Rev. Sd. Instru. 35 (1964)40. [61] R.W. Motley, 0 Machines (Academic, New York, 1975).

325

You might also like