1: \documentstyle[aps,prl,multicol,epsfig,array]{revtex}
2:
3: \newcommand \be{\begin{eqnarray}}
4: \newcommand \ee{\end{eqnarray}}
5: \newcommand \nn{\nonumber}
6:
7: \begin{document}
8:
9: \title{Period doubling in glow discharges:
10: local versus global differential conductivity}
11:
12: \author{Danijela D. \v{S}ija\v{c}i\'c$^1$, Ute Ebert$^{1,2}$
13: and Ismail Rafatov$^{1,3}$}
14:
15: \address{$^1$CWI, P.O.Box 94079, 1090 GB Amsterdam, The Netherlands,\\
16: $^2$Dept. Physics, Eindhoven Univ. Techn., The Netherlands,\\
17: $^3$American University in Central Asia, Bishkek, Kyrgyzstan}
18:
19: \date{\today}
20: \maketitle
21:
22: \begin{abstract}
23: Short planar glow discharges coupled to a resistive layer exhibit
24: a wealth of spontaneous spatio-temporal patterns. Several authors
25: have suggested effective reaction-diffusion-models to explore
26: similarities with other pattern forming systems. To test these
27: effective models, we here investigate the temporal oscillations
28: of a glow discharge layer coupled to a linear resistor.
29: We find an unexpected cascade of period doubling events.
30: This shows that the inner structure of the discharge is more
31: complex than can be described by a reaction-diffusion-model
32: with negative differential conductivity.
33: \end{abstract}
34:
35: \begin{multicols}{2}
36:
37: Glow discharges are part of our daily environment
38: in conventional and energy saving lamps, beamers, flat TV screens,
39: car and street lamps as well as in various industrial applications.
40: While applications typically try to avoid any instabilities,
41: experiments actually exhibit a realm of spontaneous pattern formation,
42: see, e.g.,\cite{Kogel}.
43:
44: An interesting series of experiments has been performed
45: on short planar dc driven glow discharges with wide lateral extension
46: \cite{Muenster0,Muenster1,Muenster2,StripeM,Muenster3,HexM,Zigstr,Str,filStr,Astr01,Str02,Muen03}
47: where the formed patterns were explored very systematically.
48: The observed structures resemble those observed in Rayleigh-Benard
49: convection in flat cells, in electroconvection in nematic liquid
50: crystals, or in various chemical or biological pattern forming systems.
51: All these systems show the formation of stripes, spots, spirals etc.
52: In comparison to the other systems, the glow discharge system has
53: the advantage of particular convenient experimental handling and
54: time scales \cite{Gwinn}. Besides structures known from other
55: physical systems, it continues to exhibit new structures
56: that might be specific for this system
57: \cite{Muenster0,Muenster1,Muenster2,StripeM,Muenster3,HexM,Zigstr,Str,filStr,Astr01,Str02,Muen03}.
58: We will focus on the experiment in \cite{Str},
59: where a complete phase diagram of different patterns
60: was identified: homogeneous stationary and
61: homogeneous oscillating modes, patterns with spatial and
62: spatio-temporal structures etc.
63:
64: The theory for these systems has largely focussed on effective
65: reaction-diffusion models in the two transversal directions,
66: and on the negative differential
67: conductivity of the glow discharge as the driving force of
68: pattern formation. Such models actually have been developed
69: independently by a number of authors
70: \cite{KGM,Rade89,Rade90,Rade92,Phelps93III,Petro97,Kolobov,Muenster2}.
71: On the other hand, the observation of unconventional patterns
72: like zigzag-destabilized spirals raises doubts whether
73: reaction-diffusion patterns are sufficient to understand
74: the observed patterns.
75:
76: In the present paper, we examine the concepts of reaction-diffusion
77: models and negative differential conductivity on the particular
78: case of a short DC driven glow discharge in a parameter range that exhibits
79: spontaneous temporal oscillations but no spatial structures transverse
80: to the current \cite{Str}.
81: In short, we find $~$ $(i)$ that a discharge on the transition from Townsend
82: to glow discharge can combine a positive local differential
83: conductivity with a negative global differential conductivity;
84: $~$ $(ii)$ that a glow discharge in a simple electric circuit shows
85: more complex behavior than can be expected from the proposed
86: reaction-diffusion models
87: \cite{KGM,Rade89,Rade90,Rade92,Phelps93III,Petro97,Kolobov,Muenster2}
88: for voltage $U$ and current $J$ with (global)
89: negative differential conductivity $dU/dJ<0$;
90: $~$ $(iii)$ in particular, that the system can show period
91: doubling bifurcations.
92: Period doubling actually has been observed experimentally in
93: glow discharges, but in more complex geometries and in longer systems
94: \cite{PD1,PD2}.
95: $~$ $(iv)$ Finally, we derive a new effective dynamical model in terms
96: of a parameter and a function by adiabatic elimination of the electrons.
97: There is no systematic way to reduce this model to a simpler one
98: \cite{KGM,Rade89,Rade90,Rade92,Phelps93III,Petro97,Kolobov,Muenster2}
99: with two scalar parameters like voltage $U$ and current $J$.
100: We draw this conclusion both from direct analysis and from
101: the occurence of period doubling in the numerical solutions.
102:
103: To be precise, in the experiments of
104: \cite{Muenster0,Muenster1,Muenster2,StripeM,Muenster3,HexM,Zigstr,Str,filStr,Astr01,Str02,Muen03},
105: a planar glow discharge layer
106: with short length in the forward direction and wide lateral dimensions
107: is coupled to a semiconductor layer with low conductivity.
108: The whole structure is sandwiched between
109: two planar electrodes to which a DC voltage $U_t$ is applied.
110: Theoretical predictions on how the different spatio-temporal
111: patterns depend on the parameters of the gas discharge, hardly exist.
112: In \cite{Rade89,Rade90,Rade92,Muenster2}, an effective
113: reaction-diffusion model in the two dimensions
114: transversal to the current is proposed. Roughly, it consists of two
115: nonlinear partial differential equations for the current $J$ and
116: the voltage $U$ of the form
117: \be
118: \label{RD}
119: \partial_tU(x,y,t)={\cal F}(U,J)~~~,~~~
120: \partial_tJ(x,y,t)={\cal G}(U,J)~,
121: \ee
122: where the nonlinear operators ${\cal F}$ and ${\cal G}$ contain spatial
123: derivatives $\partial_x$, $\partial_y$ and possibly also integral kernels.
124: The model is of
125: reactor-inhibitor form as studied extensively in the context of chemical
126: and biological systems in the past decades. If applicable to gas discharges,
127: this identification lays a connection to a realm of analytical and
128: numerical results on reaction-diffusion systems.
129:
130: To test whether a model like (\ref{RD}) is applicable to the gas discharge
131: system, we will focus on its temporal oscillations
132: that can occur in a spatially completely homogeneous mode \cite{Str};
133: hence a one-dimensional approximation is appropriate. Similar
134: oscillations have been observed in
135: \cite{Phelps93III,Petro97,Phelps93I,Pitch},
136: and similar effective models for current $J$ and voltage $U$
137: of the general form (\ref{RD}) have been proposed in
138: \cite{KGM,Phelps93III,Petro97,Kolobov}.
139:
140: Why have different authors come up with the same type of model?
141: The equation for $U$ directly results from the simplest form of
142: an external electric circuit: a semiconductor layer of thickness
143: $d_s$, linear conductivity $\sigma_s$ and dielectricity constant
144: $\epsilon_s$ will evolve as
145: \be
146: \label{U_t}
147: \partial_tU=\frac{U_t-U-R_sJ}{T_s}
148: \ee
149: where $U_t$ is the voltage on the total system, $J$ is the total current,
150: and $U=\int_0^{d_g}E\;dz$ is the voltage over the gas discharge which
151: is the electric field $E$ integrated in the $z$ direction over the
152: height $d_g$ of the discharge.
153: For the experiments in \cite{Str}, $R_s=d_s/\sigma_s$ is the resistance
154: of the whole semiconductor layer, and $T_s=\epsilon_s\epsilon_0/\sigma_s
155: =C_sR_s$ is the Maxwell relaxation time of the semiconductor.
156: In other experimental systems, the quantities $R_s$ and $T_s$
157: can have different realizations. Hence the form of the equation
158: for $U$ in a reaction-diffusion model (\ref{RD}) is clear.
159:
160: However, the equation for $J$ in a reaction-diffusion model
161: as (\ref{RD}) is based on guesses and plausibility. Different
162: choices have been suggested by different authors, but one
163: thing is clear: to be physically meaningful, the current-voltage
164: characteristics of the glow discharge has to be a stationary solution,
165: so ${\cal G}(U,J)=0$ on the characteristics. Beyond that,
166: there are different suggestions for the functional form
167: of ${\cal G}$ and the intrinsic time scale.
168:
169: If a model like (\ref{RD}) is applicable to oscillations
170: in glow discharge systems, then the following predictions apply:
171: $~$ 1) an oscillation can only occur in a region of negative
172: differential conductivity of the glow discharge characteristics, $~$
173: 2) only a single period can be formed, period doubling is not possible,
174: since this requires at least three independent parameters, $~$
175: 3) in a phase space plot in $U$ and $J$, the trajectory of an
176: oscillation can intersect the load line $U=U_t-R_sJ$ only parallel
177: to the $J$-axis (since $\partial_tU=0$ and $\partial_tJ\ne0$), and
178: it can intersect the characteristics of the glow discharge $U=U(J)$
179: only parallel to the $U$-axis (since $\partial_tU\ne0$ and $\partial_tJ=0$).
180:
181: We now introduce the simplest classical model for a glow discharge
182: \cite{Engel,Raizer,DanaStat},
183: solve it numerically and confront its results with the predictions above.
184:
185: A discharge between Townsend and glow regime consists of a gas
186: with Ohmic conductivity for the rare charged particles,
187: electrostatic space charge effects and two ionization mechanisms,
188: namely impact ionization by accelerated electrons in the bulk
189: of the discharge (the so-called $\alpha$-process) and secondary
190: emission from the cathode (the $\gamma$-process).
191: In its simplest form, it can be modelled by continuity equations
192: for electron particle density $n_e$ and ion particle density $n_+$
193: \be
194: \label{1}
195: \partial_t\;n_e \;+\; \nabla\cdot{\bf J}_e
196: ={\cal S} ~~~,~~~
197: \partial_t\;n_+ \;+\; \nabla\cdot{\bf J}_+
198: = {\cal S}~,
199: \ee
200: and the Poisson equation for the electric field {\bf E}
201: in electrostatic approximation,
202: \be
203: \label{3}
204: \nabla\cdot{\bf E} = {{\rm e}\over{\varepsilon_0}} \;(n_+ -n_e)
205: ~~~,~~~{\bf E}=-\nabla\Phi~.
206: \ee
207: The particle currents are approximated as purely Ohmic
208: \be
209: \label{4}
210: {\bf J}_e = - \mu_e \;n_e \;{\bf E}~~~,~~~
211: {\bf J}_+ = \mu_+\;n_+ \;{\bf E} ~.
212: \ee
213: The source of particles in the continuity equation (\ref{1}) is written as
214: a sum of generation by impact ionization in Townsend approximation
215: and recombination
216: \be
217: \label{5}
218: {\cal S} = |n_e \mu_e { E}| \;\alpha_0
219: \;\mbox{\large{e}}^{\textstyle -E_0/|{ E}|} - \beta n_e n_+ ~.
220: \ee
221: Finally, the secondary emission from the cathode enters as
222: a boundary condition at the position $d_g$ of the cathode
223: \be
224: \label{6}
225: \mu_en_e(d_g,t)=\gamma\mu_+n_+(d_g,t)~.
226: \ee
227: This is the classical glow discharge model \cite{Engel,Raizer,DanaStat}.
228:
229: We reduce the problem to one spatial dimension $z$
230: transverse to the layers which is an excellent approximation for
231: the experiment \cite{Str}. Furthermore, we introduce dimensionless
232: quantities as in \cite{DanaStat} by rescaling all parameters
233: and fields as $z=r_z/X_0$, $\tau=t/t_0$, $L=d_g/X_0$,
234: $\sigma(z,\tau)=n_e(r_z,t)/n_0$, $\rho=n_+/n_0$, ${\cal E}=E_z/E_0$
235: with the scales $X_0=\alpha_0^{-1}$, $t_0=(\alpha_0\mu_eE_0)^{-1}$
236: and $n_0=\epsilon_0\alpha_0E_0/{\rm e}$.
237: A key role is played by the small parameter $\mu=\mu_+/\mu_e$,
238: which is the mobility ratio of ions and electrons.
239:
240: The gas discharge layer is now modelled by
241: \be
242: \label{h01}
243: \partial_\tau\sigma&=&\partial_z({\cal E}\sigma)
244: +\sigma {\cal E}\alpha({\cal E})~~~,~~~\alpha({\cal E})=e^{-1/|{\cal E}|}~,\\
245: \label{h02}
246: \partial_\tau\rho&=&-\mu\partial_z({\cal E}\rho)
247: +\sigma {\cal E}\alpha({\cal E})~,\\
248: \label{h04}
249: &&\sigma(L,\tau)=\gamma\mu\rho(L,\tau)~,\\
250: \label{h05}
251: \rho-\sigma&=&\partial_z{\cal E}~,
252: \ee
253: where recombination was neglected [$\beta=0$ in (\ref{5}), a discussion of this
254: approximation follows below], while the external circuit is described by
255: \be
256: \label{h06}
257: \partial_\tau {\cal U}=
258: \frac{{\cal U}_t-{\cal U}-{\cal R}_s j}{\tau_s}~~,~~
259: \label{h07}
260: {\cal U}(\tau)=\int_0^L {\cal E}(z,\tau)\;dz
261: \ee
262: with the dimensionless voltage ${\cal U}=U/(E_0X_0)$, time scale
263: $\tau_s=T_s/t_0$ and resistance ${\cal R}_s=R_s/R_0$,
264: $R_0=X_0/({\rm e}\mu_en_0)$
265: and with a spatially conserved total current
266: \be
267: \label{h08}
268: &&j(\tau)=\partial_\tau {\cal E}+\mu\rho {\cal E}+\sigma {\cal E}
269: ~~,~~\partial_z j(\tau)=0~,
270: \ee
271: where $\partial_zj=0$ follows from
272: (\ref{h01}), (\ref{h02}) and (\ref{h05}) as usual.
273:
274: As a result, the gas discharge is parametrized by the three
275: dimensionless parameters of system length over ionization length $L$,
276: secondary emission coefficient $\gamma$ and mobility ratio $\mu$
277: (as discussed in \cite{DanaStat}), and the external circuit is
278: parametrized by relative resistance ${\cal R}_s$, ratio of time
279: scales $\tau_s$ and dimensionless applied voltage ${\cal U}_t$.
280:
281: For calculational purposes, the ion density $\rho$ can be
282: completely eliminated from the one-dimensional gas discharge equations
283: (\ref{h01})--(\ref{h05}) with the help of the Poisson equation (\ref{h05})
284: and the total current $j$, see \cite{DanaStat}. The result are two equations
285: of motion for $\partial_\tau\sigma$ and $\partial_\tau{\cal E}$.
286: In our numerical calculations, the system was implemented in
287: this form. Our choice of parameters was guided by
288: the experiments in \cite{Str}: we chose the secondary emission coefficient
289: $\gamma=0.08$, the mobility ratio $\mu=0.0035$ for nitrogen and
290: the dimensionless system size $L=50$ which amounts to 1.4 mm
291: at a pressure of 40 mbar. The external circuit has ${\cal R}_s=30597$,
292: $\tau_s=7435$ and a dimensionless total
293: voltage ${\cal U}_t$ in the range between 18 and 20.
294: This corresponds to a GaAs layer with $\epsilon_s=13.1$, conductivity
295: $\sigma_s=(2.6\cdot 10^5\Omega{\rm cm})^{-1}$ and thickness
296: $d_s=1.5 {\rm mm}$, and a voltage range between 513 and 570 V.
297:
298: \begin{figure}[h]
299: \centerline{
300: \psfig{figure=fig1.eps,width=8.5cm}}
301: \caption{Spontaneous oscillations of current $j$ and voltage ${\cal U}$
302: as a function of time $\tau$ for
303: $\gamma=0.08$, $\mu=0.0035$, $L=50$, ${\cal R}_s=30597$, $\tau_s=7435$,
304: and applied total voltage ${\cal U}_t=19$.}
305: \end{figure}
306:
307: Fig.~1 shows electric current $j$ and voltage on the gas discharge
308: ${\cal U}$ as a function of time for a total stationary voltage
309: ${\cal U}_t=19$ applied to the complete system of gas discharge and
310: semiconductor layer. The system exhibits spontaneous oscillations
311: with sharp current peaks: when the voltage ${\cal U}$ on the gas layer
312: becomes high enough, the discharge ignites. The conductivity of
313: the gas increases rapidly and produces a current pulse that deposits
314: a surface charge on the gas-semiconductor interface. Therefore the voltage
315: ${\cal U}$ over the gas layer breaks down. Due to the low conductivity of the
316: semiconductor, the voltage ${\cal U}$ recovers only slowly. Eventually
317: the gas dicharge ignites again, and the cycle is repeated.
318:
319: Note that the oscillations in Fig.~1 are not quite periodic.
320: This is not due initial transients since the system is observed
321: after the long relaxation time $\tau=4.745\cdot10^6$. The nature
322: of this temporal structure becomes clear when the trajectory
323: is plotted in the plane spanned by current $j$ and voltage ${\cal U}$
324: in Fig.~2(b). The figure contains the data of the time span
325: from $\tau=3\cdot10^6$ to $6\cdot10^6$ which amounts to approximately 90
326: current pulses.
327: The phase space plot shows that the system is actually periodic,
328: with a period of 8 current pulses. Fig.~1 shows precisely one period.
329:
330: This discovery raises the question whether our system actually
331: follows the well-known scenario of period doubling. Indeed, it does.
332: Fig.~2(a) for ${\cal U}_t=18$ shows an oscillation where one current pulse
333: is repeated periodically as observed experimentally in \cite{Str}.
334: For ${\cal U}_t=18.5$, a period consists of two current pulses (not shown).
335: For ${\cal U}_t=19$, the period is 8 pulses as in Fig.~1 and Fig.~2(b).
336: For ${\cal U}_t=20$, the systems seems to have reached the chaotic state
337: as can be seen in Fig.~2(c).
338:
339: A detailed comparison of the experiments in \cite{Str} with simple
340: oscillations as in Fig.~2(a) will be given elsewhere, and we only state
341: here that there is semi-quantitative agreement of several features.
342: Here we emphasize that period doubling events
343: in glow discharges have been observed experimentally in other systems
344: \cite{PD1,PD2}. However, this was always in systems with
345: more complicated geometries like long narrow tubes, and the authors allude
346: to general knowledge on nonlinear dynamics rather than to solutions of
347: explicit models. We state that period doubling can be a generic
348: feature of a simple, strictly one-dimensional glow discharge
349: when coupled to the simple circuit (\ref{U_t}). We propose to search
350: experimentally for a period doubling route to chaos in such simple
351: systems which would then allow quantitative comparison with theory.
352:
353: Let us return to
354: the initial question: is a 2-component reaction diffusion model
355: like (\ref{RD}) with negative differential conductivity appropriate
356: for the present system? Above Eq.~(\ref{1}), we gave a list of predictions
357: for the reaction diffusion model (\ref{RD}) to be applicable.
358: Prediction 2 is falsified by the observation of period doubling.
359: Prediction 3 is also falsified by a simple check of either of the
360: three figures in Fig.~2: the trajectories definitely do not intersect
361: with the characteristics or the load line with the angle prescribed
362: by (\ref{RD}), in particular not in the upper part of the figures
363: that represent the rapid current pulses.
364:
365: There rests prediction 1: is negative differential conductivity required
366: for the oscillations to occur? We have not found a numerical
367: counterexample where oscillations would occur while the current voltage
368: characteristics of the gas discharge shows a positive differential
369: conductivity, but we have found no reason to exclude its existence.
370: Furthermore we note that the characteristics is a global property of
371: the whole discharge layer with its boundary conditions \cite{DanaStat}
372: while the local differential conductivity in our model is always
373: positive: the field dependent stationary ionization is
374: $n_+=|\mu_eE|\alpha_0 e^{-E_0/|E|}/\beta$ according to (\ref{5}); hence
375: the local conductivity increases monotonically with the applied field $|E|$.
376: The global negative differential conductivity is due to electrode
377: effects being much stronger than bulk recombination $\beta$.
378:
379: Last but not least, we have derived an analytical
380: approximation of the model (\ref{h01})--(\ref{h08}) that can be
381: confronted with the suggested form (\ref{RD}). Electron and ion current
382: in the gas are of the same order of magnitude. Since the electrons
383: are much more mobile, their density is appropriately lower.
384: Rescaling this density like $s=\sigma/\mu$ and time like
385: $\bar\tau=\mu\tau$, the electrons can be eliminated adiabatically
386: in the limit of $\mu\to0$. Space charges in the gas discharge are
387: then due to the ions only $\rho=\partial_z{\cal E}$, and $\rho$
388: can be expressed by ${\cal E}$. Splitting the field
389: ${\cal E}(z,t)={\cal E}_L(t)+\epsilon(z,t)$ into the field on the cathode
390: ${\cal E}_L$ and a correction $\epsilon$ with $\epsilon(L,t)=0$,
391: the complete system for $\mu\to0$ can be expressed by two dynamic equations
392: \be
393: \label{RDus}
394: \partial_{\bar\tau}{\cal E}_L(t)=F({\cal E}_L,\epsilon)~~~,~~~
395: \partial_{\bar\tau}\epsilon(z,t)=G({\cal E}_L,\epsilon)~,
396: \ee
397: details will be given elsewhere.
398: While the equation for the time dependent parameter ${\cal E}_L$
399: corresponds to the equation for $U$ in (\ref{RD}), the space dependent
400: field correction $\epsilon$ within the gas layer cannot be reduced to
401: a single component like the current $J$ in (\ref{RD}). E.g., for the ion
402: density on the cathode $\rho_L=\partial_z\epsilon|_L$,
403: we can derive the equation of motion
404: \be
405: \label{adia}
406: \partial_{\bar\tau}\rho_L
407: =-\rho_L^2-{\cal E}_L(\partial_z\rho)|_L
408: +\gamma\rho_L{\cal E}_L\alpha({\cal E}_L)
409: \ee
410: A two component reaction-diffusion equation for $\rho_L$ and ${\cal E}_L$
411: could result from the completely unsystematic approximation
412: $(\partial_z\rho)|_L=0$. Rather the transport of ions $\rho$
413: from the bulk of the gas towards the cathode
414: is a central feature of the system. The field
415: $\epsilon(z,t)$ in (\ref{RDus}) indeed accounts for
416: the ion distribution within the gas gap with its intricate dynamics.
417:
418: We acknowledge support of D.S. by the Dutch physics
419: funding agency FOM and of I.R. by ERCIM.
420:
421: \begin{thebibliography}{99}
422:
423: \bibitem{Kogel} U. Kogelschatz, Plasma Chem.\ Pl.\ P.\ {\bf 23},
424: 1-46 (2003).
425:
426: \bibitem{Muenster0} Yu.A. Astrov {\it et al.},
427: %, E. Ammelt, S. Teperick, H.-G. Purwins,
428: Phys. Lett. A {\bf 211}, 184 (1996).
429:
430: \bibitem{Muenster1}
431: Yu.A. Astrov {\it et al.},
432: %, E. Ammelt and H.-G. Purwins,
433: Phys. Rev. Lett. {\bf 78}, 3129 (1997).
434:
435: \bibitem{Muenster2}
436: Yu.A. Astrov, Y.A. Logvin,
437: Phys. Rev. Lett. {\bf 79}, 2983 (1997).
438:
439: \bibitem{StripeM}
440: E. Ammelt {\it et al.},
441: %,Yu.A. Astrov and H.-G. Purwins,
442: Phys. Rev. E {\bf 55}, 6731 (1997).
443:
444: \bibitem{Muenster3}
445: Y.A. Astrov {\it et al.},
446: %, I. M\"uller, E. Ammelt and H.-G. Purwins,
447: Phys. Rev. Lett. {\bf 80}, 5341 (1998).
448:
449: \bibitem{HexM}
450: E. Ammelt {\it et al.},
451: %, Yu.A. Astrov and H.-G. Purwins,
452: Phys. Rev. E {\bf 58}, 7109 (1998).
453:
454: \bibitem{Zigstr}
455: C. Str\"umpel {\it et al.},
456: %, Y.A. Astrov, E.Ammelt and H.-G. Purwins,
457: Phys. Rev. E {\bf 61}, 4899 (2000).
458:
459: \bibitem{Str}
460: C. Str\"umpel {\it et al.},
461: %, Y.A. Astrov, and H.-G. Purwins,
462: Phys. Rev. E {\bf 62}, 4889-4897 (2000).
463:
464: \bibitem{filStr}
465: C. Str\"umpel {\it et al.},
466: %, H.-G. Purwins, and Y.A. Astrov,
467: Phys. Rev. E {\bf 63}, 026409 (2001).
468:
469: \bibitem{Astr01} Y.A. Astrov, H.-G. Purwins, Phys. Lett. A {\bf 283},
470: 349 (2001).
471:
472: \bibitem{Str02} C. Str\"umpel {\it et al.},
473: %, H.-G. Purwins, and Y.A. Astrov,
474: Phys. Rev. E {\bf 65}, 066210 (2002).
475:
476: \bibitem{Muen03} E.L. Gurevich {\it et al.},
477: %, A.S. Moskalenko, A.L. Zanin, Y.A. Astrov, H.-G. Purwins,
478: Phys. Lett. A {\bf 307}, 299 (2003).
479:
480: \bibitem{Gwinn} W. Breazeal, K.M. Flynn, E.G. Gwinn,
481: Phys. Rev. E {\bf 52}, 1503 (1995).
482:
483: \bibitem{KGM} K.G. M\"uller, Phys. Rev. A {\bf 37}, 4836 (1988).
484:
485: \bibitem{Rade89}
486: H.-G. Purwins {\it et al.},
487: %, C. Radehaus, T. Dirksmeyer, R. Dohmen, R. Schmeling, and H. Willebrand,
488: Phys. Lett. A {\bf 136}, 480 (1989).
489:
490: \bibitem{Rade90}
491: C. Radehaus {\it et al.},
492: %, R. Dohmen, H. Willebrand, and F.-J. Niedernostheide,
493: Phys. Rev. A {\bf 42}, 7426 (1990).
494:
495: \bibitem{Rade92} C. Radehaus {\it et al.},
496: %, H. Willebrand, R. Dohmen, F.-J. Niedernostheide,
497: Phys. Rev. A {\bf 45}, 2546 (1992).
498:
499: \bibitem{Phelps93III}
500: A.V. Phelps, Z.L. Petrovic, and B.M. Jelenkovic,
501: Phys. Rev. E {\bf 47}, 2825 (1993).
502:
503: \bibitem{Petro97}
504: Z.L. Petrovic, I. Stefanovic, S. Vrhovac, and J. Zivkovic,
505: J. Phys. IV France {\bf 7}, Colloque C4, 341-352 (1997).
506:
507: \bibitem{Kolobov}
508: V.I. Kolobov and A. Fiala,
509: Phys. Rev. E {\bf 50}, 3018 (1994).
510:
511: \bibitem{PD1} T. Braun {\it et al.}, Phys. Rev. Lett. {\bf 59}, 613 (1987).
512:
513: \bibitem{PD2} J. Qin {\it et al.}. Phys. Rev. Lett. {\bf 63}, 163 (1989).
514:
515: \bibitem{Phelps93I}
516: Z.L. Petrovic and A.V. Phelps,
517: Phys. Rev. E {\bf 47}, 2806 (1993).
518:
519: \bibitem{Pitch} I. P\'er\`es, L.C. Pitchford,
520: J. Appl. Phys. {\bf 78}, 774 (1995).
521:
522: \bibitem{Engel}
523: A. von Engel, M. Steenbeck, {\it Elektrische Gasentladungen},
524: Vol. II (Springer, Berlin 1934).
525:
526: \bibitem{Raizer}
527: Y.P. Raizer, {\it Gas Discharge Physics}
528: (Springer, Berlin, 2nd corrected printing, 1997).
529:
530: \bibitem{DanaStat} D.D. \v{S}ija\v{c}i\'c, U. Ebert,
531: Phys. Rev. E {\bf 66}, 066410 (2002).
532:
533: \end{thebibliography}
534:
535:
536:
537: \end{multicols}
538:
539: \begin{figure}[h]
540: \centerline{
541: \psfig{figure=fig2.eps,width=6cm}
542: \psfig{figure=fig3.eps,width=6cm}
543: \psfig{figure=fig4.eps,width=6cm}
544: }
545: \caption{Phase space plots of the trajectories of the oscillations
546: in the plane of current $j$ and voltage ${\cal U}$.
547: The time range is $3\cdot10^6\le\tau\le6\cdot10^6$ in all figures.
548: Shown are the orbits, the straight load line ${\cal U}={\cal U}_t-{\cal R}_sj$
549: and the curved current voltage characteristics ${\cal U}={\cal U}(j)$ of
550: the gas discharge [28].
551: The intersection of load line and characteristics
552: marks the stationary solution of the system.
553: (b) represents the data of Fig.~1 with total voltage ${\cal U}_t=19$,
554: (a) is for ${\cal U}_t=18$, (c) for ${\cal U}_t=20$.}
555: \end{figure}
556:
557: \end{document}
558:
559: