physics0604012/Br.tex
1: \documentclass[aps,pre,twocolumn,groupedaddress]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{amsmath, amssymb}
4: 
5: \newcommand \be{\begin{eqnarray}}
6: \newcommand \ee{\end{eqnarray}}
7: \newcommand \ba{\begin{eqnarray}}
8: \newcommand \ea{\end{eqnarray}}
9: \newcommand \bs{\boldsymbol}
10: \newcommand \mc{\mathcal}
11: \newcommand \E{{\bs{\mc E}}}
12: \newcommand \ns{\negthickspace}
13: \def\nn{\nonumber}
14: 
15: % You should use BibTeX and apsrev.bst for references
16: % Choosing a journal automatically selects the correct APS
17: % BibTeX style file (bst file), so only uncomment the line
18: % below if necessary.
19: %\bibliographystyle{apsrev}
20: 
21: \begin{document}
22: \title{Numerical convergence of the branching time of negative streamers}
23: \author{Carolynne Montijn$^1$, Ute Ebert$^{1,2}$, and Willem Hundsdorfer$^1$}
24: \affiliation{$^1$CWI, P.O.Box 94079, 1090 GB Amsterdam, The Netherlands,}
25: \affiliation{$^2$Dept.\ Physics, Eindhoven Univ.\ Techn., The Netherlands.}
26: 
27: \date{\today}
28: 
29: 
30: \begin{abstract}
31: In sufficiently large gaps and electric fields, discharge streamers 
32: do branch. In [Array\'as {\it et al.}, PRL {\bf 88}, 174502 (2002)], 
33: we observed streamer branching numerically within a deterministic 
34: particle density model and explained it as 
35: a Laplacian instability of a thin space charge layer. Our numerical
36: results were criticized in [Kulikovsky, PRL {\bf 89}, 229401 (2002)]. 
37: We here present an adaptive grid refinement method for streamer 
38: simulations, and we carry out the first conclusive investigation 
39: on the effect of the numerical grid on streamer branching in different 
40: fields. On stepwise finer grids the branching time converges, 
41: hence streamer branching is for the first time predicted quantitatively. 
42: \end{abstract}
43: 
44: \pacs{52.80.-s, 05.45.-a }
45: \maketitle
46: 
47: {\bf Problem setting and review.}
48: Streamers are transient weakly ionized plasma channels 
49: that rapidly grow into a non- or weakly ionized medium
50: under influence of the self-enhanced electric field at their tip.
51: They are widely used in technology~\cite{vel2000,ebe2006} and ubiquitous 
52: in nature, where they play a role in creating the path of sparks, 
53: lightning~\cite{baz2000} and of blue jets above thunderclouds.
54: Streamers are also directly observed as so-called 
55: sprites~\cite{sen1995,ger2000,pas2002}, which are very large discharge structures in the 
56: higher parts of the atmosphere that are composed of ten thousands of streamers.
57: Despite their high velocity, streamer evolution is now directly observable
58: in experiments; a further review can be found in~\cite{ebe2006}.
59: 
60: Streamers commonly branch in experiments if gap and applied voltage 
61: are large enough.
62: Recently a debate has risen about the proper physical concept for
63: this branching. In 1939, Raether~\cite{rae1939} proposed a mechanism 
64: for streamer propagation and Loeb and Meek~\cite{loe1940} 
65: developed it into a branching concept that
66: nowadays is found in many textbooks. The concept is based 
67: on a uniformly charged streamer head; ahead of it stochastic 
68: processes create secondary avalanches, that subsequently develop 
69: into different branches. However, the distribution of rare electrons
70: due to photo-ionization or background ionization ahead of the streamer 
71: has never been shown to agree with the conceptual pictures, and the
72: concept has never been demonstrated to work.
73: Furthermore, simulations in the past two 
74: decades~\cite{dha1987,vit1994,kul1997-2,pan2001} have shown that
75: the fully developed streamer head is not homogeneously charged, 
76: but rather neutral and surrounded by a thin space charge layer which enhances
77: the field ahead of it and screens it in the interior; this field
78: enhancement allows the streamer to penetrate
79: regions with a rather low background field. 
80: %The effect of stochastic mechanisms ahead of such 
81: %a space charge layer has not been investigated yet. 
82: Recent simulations also show that a streamer can branch within 
83: a fully deterministic model for charged particle densities, 
84: in a non-uniform background field~\cite{vit1993,hal2003,geo2005} as
85: well as in a uniform field~\cite{arr2002,roc2002,liu2004}, provided certain
86: requirements on the external parameters are met (e.g. a sufficiently
87: strong background electric field and a sufficiently long gap).
88: 
89: Some of the present authors have proposed~\cite{arr2002,roc2002} 
90: a physical explanation of these numerical observations that is directly 
91: related to the formation of the thin space charge layer: the layer
92: creates an almost equipotential streamer head that can undergo 
93: a Laplacian instability and branch in a manner similar 
94: to branching instabilities of fluid interfaces in viscous fingering. 
95: For a further discussion of the conceptual questions of streamer branching,
96: we refer to \cite{ebe2006}.
97: However, the numerical codes used in~\cite{vit1993,hal2003,geo2005,arr2002,
98: roc2002,liu2004} were not able to 
99: test the branching conditions on fine numerical grids. 
100: This lead some researchers
101: to question the physical nature of the instabilities~\cite{kul2002,pan2003,
102: hal2003, geo2005} despite the analytical arguments given 
103: in~\cite{arr2002,roc2002} and later in~\cite{meu2004,meu2005}.
104: 
105: To resolve the debate from the numerical side, we have developed
106: a code with comoving adaptive grids and we here present its results. 
107: The algorithm enables us to run the simulations on very fine grids;
108: therefore for the first time the effect of numerical grids 
109: on the branching process is investigated quantitatively. 
110: We here present its results: branching occurs both at very high fields 
111: like in Refs.~\cite{arr2002,roc2002} and also at fairly low 
112: background fields if the discharge has sufficient space to develop;
113: and the branching time saturates on sufficiently fine numerical grids.
114: This enables us to give the first quantitative predictions 
115: on streamer branching. 
116: \\
117: 
118: {\bf Model and multiscale structure of negative streamers.}
119: We investigate a minimal continuum model for streamers, 
120: which contains the essential physics for negative streamers 
121: in a non-attaching pure gas like N$_2$ or 
122: Ar~\cite{dha1987,vit1994,arr2002,roc2002}. The model is a two-fluid 
123: approximation for the charged particles, with a local field dependent 
124: impact ionization reaction coupled to the Poisson equation for 
125: electrostatic particle interactions. We investigate this 
126: model in a cylindrically symmetric geometry, reducing it to effectively
127: two dimensions. This constraint suppresses
128: one degree of freedom for the instability modes, and therefore the time of
129: branching in this cylindrical geometry is an upper bound for the branching 
130: time in a genuine three dimensional system~\cite{ebe2002-2,ebe2006}. 
131: In dimensionless units, the model reads
132: \ba
133: %\begin{array}{rcl}
134: \displaystyle\partial_\tau\sigma 
135:   & = & \nabla\cdot(\sigma{\bs{\mathcal E}}+D\nabla\sigma)
136:         + \sigma|{\bs{\mathcal E}}|\;\alpha(|{\bs{\mathcal E}}|), 
137: \label{sigmaeq}\\[0mm]
138: \displaystyle\partial_\tau\rho
139:   & = & \sigma|{\bs{\mathcal E}}|\;\alpha(|{\bs{\mathcal E}}|), 
140: ~~~\alpha(|{\bs{\mathcal E}}|) =e^{-1/|{\bs{\mathcal E}}|},
141: \label{rhoeq}\\[0mm]
142: \displaystyle - \nabla^2\phi& = &\rho-\sigma
143:   \, , \quad {\bs{\mathcal E}}=-\nabla\phi,
144: \label{phieq}
145: \ea
146: where $\sigma$ and $\rho$ are the electron and positive ion densities,
147: respectively. $\E$ and $\phi$ are respectively the electric field and 
148: potential, $D$ is the electron diffusion coefficient and $\tau$ is the
149: dimensionless time. The characteristic scales in this model depend
150: on the neutral gas density; therefore the simulation results 
151: can be applied to high altitude 
152: sprite discharges at low pressures as well as to high pressure laboratory 
153: experiments. We refer to~\cite{ebe1997,roc2002,ebe2006} for more details 
154: on the dimensional analysis. 
155: 
156: 
157: A planar cathode is placed at $z=0$ and a planar anode at $z=L_z$.
158: The potential at the electrodes is fixed, $\phi(r,z=0,\tau)=0$,
159: $\phi(r,z=L_z,\tau)=\phi_0>0$, generating a background electric
160: field with strength $|\E_b|=\phi_0/L_z$ along the negative $z$-direction.
161: The streamer is initiated by an electrically neutral Gaussian ionization 
162: seed on the axis of symmetry at the cathode ($r=z=0$). There is no
163: background ionization far from the initial seed.
164: 
165: We impose homogeneous Neumann conditions for the electron density at 
166: all boundaries. This results in a net inflow of electrons from the cathode 
167: if the streamer is attached to it~\cite{arr2002,mon2005-2}. 
168: In practice, the computational volume is restricted in the radial
169: direction by a boundary $L_r$ sufficiently far away not to
170: disturb the solution near and in the streamer. Moreover, we choose the
171: inter-electrode distance $L_z$ so large that the streamer does not
172: feel the anode proximity for the results shown. 
173: 
174: The generic spatial structure of the streamer is already 
175: discussed above and can be seen in the figures: it
176: contains a wide range of spatial scales,
177: from the very extended non-ionized medium on which the Poisson equation 
178: has to be solved through the length of the conducting channel
179: and its width up to the inner structure of the thin space charge layer
180: around the streamer head.
181: 
182: Moreover, the region just ahead of the streamer where the field
183: is substantially enhanced and the electron density is low, 
184: is highly unstable, in the sense 
185: that a small ionized perturbation will grow much more rapidly 
186: than in the mere background field. This unstable region ahead
187: of the streamer tip is commonly referred to as {\em leading 
188: edge}~\cite{ebe1997,ebe2000}. It requires special 
189: care when considering numerical methods~\cite{ebe2000,mon2005-2}. 
190: Accurate simulations of streamers therefore pose a great computational
191: challenge.
192: \\
193: 
194: {\bf Numerical algorithm.}
195: In order to deal efficiently with the numerical challenges posed by this
196: model, it has been implemented in a numerical code using adaptive grid 
197: refinements. We recall the essential features of this algorithm 
198: and refer to \cite{mon2005-2} for further details. 
199: The spatial discretizations are based on finite volumes, using a flux 
200: limiting scheme to prevent spurious oscillations in the results near steep 
201: gradients.  The time stepping is performed with a two-stage explicit 
202: Runge-Kutta method.  
203: 
204: \begin{figure}
205: \begin{center}
206: \includegraphics[width=8cm]{fig1a.eps}
207: \includegraphics[width=8cm]{fig1b.eps}
208: \caption{Electron density distribution before and just after
209: streamer branching in a background field $|\E_b|=0.15$,
210: computed on different finest mesh sizes $h_f=$ 8, 4 and 2 as indicated 
211: over the plots. The upper snapshots at $\tau$=10000 are taken
212: before branching and the lower ones after branching, at time $\tau=11250$. 
213: The contours correspond to the same density levels. In all three cases 
214: the same restricted part of the total computational domain with 
215: $z\le L_z = 32768$ and $r\le L_r = L_z/2$ is shown.}
216: \label{br015}
217: \end{center}
218: \end{figure}
219: 
220: Using an explicit time-stepping method allows us to decouple the 
221: computational grids for the continuity 
222: equations~(\ref{sigmaeq})-(\ref{rhoeq}) on the one hand 
223: from those for the Poisson equation~(\ref{phieq}) on the other 
224: hand.  The particle densities are first updated on a series of nested, 
225: stepwise refined grids. Then the Poisson equation, using the computed 
226: densities as an input, is solved on another series of nested, stepwise 
227: refined grids. The electric field on the grids for the continuity equations 
228: is then calculated from the potential computed on the grids for the Poisson 
229: equation using sufficiently accurate interpolations~\cite{wac2005}.
230: 
231: \begin{figure}
232: \begin{center}
233: \includegraphics[width=8cm]{fig2a.eps}\\~\\
234: \includegraphics[width=8cm]{fig2b.eps}
235: \includegraphics[width=8cm]{fig2c.eps}
236: \caption{Upper panel: Branching time in a background field $|\E_b|=0.5$ 
237: as function of the finest mesh size $h_f$ = 2, 1, 1/2, 1/4, 1/8.
238: Lower panels: the corresponding electron density distribution at $\tau$=275 
239: (middle row), and just after the respective branching time  (lower row), 
240: computed on different finest grids $h_f=$ 1, 1/2, 1/4 and 1/8. 
241: %These figures correspond to the branching times of Fig.~\ref{tsplits}.
242: The total computational domain is $z\le L_z= 2048$ and $r\le L_r = L_z/2$.}
243: \label{tsplits}
244: \label{sigma_tsplits}
245: \end{center}
246: \end{figure}
247: 
248: Adequate refinement criteria for the continuity and for the Poisson 
249: equation then lead to a grid distribution which is especially
250: designed to cope adequately and efficiently with the difficulties
251: inherent to both type of equations. More specifically, the refinement 
252: criterion for the grids for the Poisson equation is based on error estimate 
253: of the solution. The refinement criterion for grids for the continuity 
254: equations uses a curvature monitor of the solution. Moreover, it takes
255: explicitly into account the {\em leading edge},
256: % which is the unstable region just ahead of the streamer tip, 
257: where the densities are low but 
258: the electric field is greatly enhanced~\cite{ebe1997,ebe2000}.
259: 
260: The refinement criterion is computed at each  time step, in such a way that
261: the series of nested, consecutively refined grids move with the solution.
262: Special care has been taken for the discretizations as well as the mapping
263: of the solution from one grid to the other to be charge conserving. 
264: \\
265: 
266: \begin{figure}
267: \begin{center}
268: \includegraphics[width=8cm]{fig3a.eps}
269: \includegraphics[width=8cm]{fig3b.eps}
270: \caption{Zoom into the streamer head during branching. Upper plots:
271: $|\E_b|=$0.15 as in Fig.~\ref{br015}, $h_f$ = 2. Lower plots: 
272: $|\E_b|=$0.5 as in Fig.~\ref{tsplits}, $h_f$ = 1/8.
273: Contour lines (thick) of net charge density and equipotential lines 
274: (thin) are shown as a function of positive radius $r$ and appropriate 
275: $z$. 
276: The spacing of the charge contour levels is 0.004 for the low field case, 
277: and 0.16 in the high field case. The spacing of equipotential lines
278: is 5 in both cases.} %For clarity we only show the positive half plane
279: %on the $r$-axis. }
280: \label{chargephibr}
281: \end{center}
282: \end{figure}
283: 
284: {\bf Results.}
285: The adaptive grid refinement procedure enables us to resolve the streamer
286: with very high accuracy, and thus to investigate the dependence
287: of the branching process on the numerical grid.
288: The results are obtained
289: on increasingly finer grid sizes $h_f$, always taking the same coarsest
290: mesh width $h_c$ for both the continuity and the Poisson equations. If
291: the branching were of numerical nature, we would
292: expect that branching times on increasingly finer grids would not 
293: converge.
294: \\
295: 
296: {\it We first consider negative streamers evolving in a low background 
297: field of $\E_b=0.15$ corresponding to 30 kV/cm for N$_2$ at atmospheric 
298: pressure.} We use an electrically neutral, dense and relatively wide 
299: Gaussian ionization seed at the cathode, with a maximum of 1/4.8, and a 
300: characteristic radius 
301: of 10. This corresponds to a maximal electron and ion number density 
302: of 10$^{14}$ cm$^{-3}$ and an 1/e radius of 230 $\mu$m.
303: The gap length and width are set to $L_z=2L_r=2^{15}=32768$, which 
304: corresponds to an inter-electrode distance of approximately 7.5 cm.
305: 
306: The coarsest mesh width is set to $h_c=64$, and the finest
307: one to $h_f=$ 8, 4 and 2.   When a finest mesh of 8 is used, the electron
308: density in the streamer is lower, as can be seen in the
309: upper row in Fig.~\ref{br015}. This is due to the numerical
310: diffusion introduced on such a coarse grid by the flux limiter
311: that switches to the diffusive first order scheme in regions with large
312: gradients. This numerical diffusion smears the electrons out over the
313: streamer head, which in turn results in lower field enhancement
314: and lower ionization rates. The results on finer meshes of 4 and 2
315: on the other hand do agree with each other.
316: 
317: The branching in time is the same in all cases.
318: Fig.~\ref{br015} shows that the influence of the numerical grid on
319: the branching state rapidly decreases, and we thus 
320: can carry out not only qualitative but also quantitative numerical 
321: experiments of the streamer evolution up to branching. These results
322: show that branching is possible at lower electric fields than those
323: of~\cite{arr2002,roc2002}.
324: Branching was not observed in earlier simulations at lower 
325: fields~\cite{dha1987,vit1994} because the discharge gap was too short.
326: \\
327: 
328: {\it We now consider a negative streamer in a dimensionless background
329: field of $\E_b=0.5$ corresponding to 100 kV/cm in N$_2$ at atmospheric
330: pressure} in a gap of $L_z=2048$, or 4.6 mm. These external parameters 
331: are as in~\cite{arr2002,roc2002}. 
332: The initial seed is also taken as in~\cite{roc2002}, i.e., a Gaussian 
333: with amplitude 10$^{-4}$ and characteristic radius 10,
334: which corresponds to a maximal electron and ion number density of
335: $5\cdot10^{10}$ cm$^{-3}$ and an 1/e-radius of 23 $\mu$m for N$_2$ under
336: normal conditions. 
337: 
338: However, while \cite{arr2002} used a 
339: uniform grid of $h=2$ and \cite{roc2002} one of $h=1$, we now perform 
340: computations on a finest grid as small as $h_f=1/8$, i.e., 
341: more than a decade finer. More precisely,
342: the coarsest mesh width is set to $h_c=2$, and the finest
343: one to $h_f=2,1,\ldots,1/8$. Furthermore a better numerical scheme is used:
344: flux limiting~\cite{mon2005-2} rather than 3rd order 
345: upwind~\cite{arr2002,roc2002}.
346: 
347: Before branching, at $\tau$=275, Fig.~\ref{tsplits} shows that 
348: there is a quantitative difference between the results on a mesh with 
349: $h_f$=1 and the other three. As in the low field case, numerical 
350: diffusion spreads the space charge layer, which makes the field 
351: enhancement at the streamer tip and the field screening in the streamer 
352: body less efficient. Consequently, the ionization rate, 
353: and therefore the electron density, are higher in the 
354: streamer body. In the low field case we do not observe this because 
355: the background field is negligibly low, hence a less efficient screening 
356: will not affect significantly the ionization rate in the streamer body. 
357: It is clear that on meshes finer than 1/2, the results are the same 
358: during the stable streamer propagation. It is only after the branching 
359: that different states are observed on those very fine grids.
360: However, the time of branching converges within this range of mesh widths
361: $h_f$ as shown in the upper plot in Fig.~\ref{tsplits}. 
362: \\
363: 
364: {\bf Discussion, conclusion and outlook.}
365: We emphasize that the branching times converge on decreasing 
366: numerical grids in both cases. Therefore we here present the first 
367: conclusive and quantitative numerical predictions on streamer branching.
368: However, in contrast to the low field case, 
369: the lower plots in Fig.~\ref{sigma_tsplits} show that in 
370: the high field case different branched modes are reached after
371: approximately the same evolution time: in two cases,
372: the maximal electron density and field is on the axis 
373: of symmetry, and in two other cases, it is off axis. Apparently,
374: there are different branched states reachable at bifurcation
375: and tiny differences determine which one will be reached.
376: Such extreme sensitivity is well-known from deterministic chaos;
377: it is generic for nonlinear dynamics near bifurcation points.
378: On the other hand, the unstable state is reached in a
379: deterministic manner, and therefore the branching times converge.
380: 
381: %
382: %{\bf Summary and outlook.} 
383: But why is there once a unique branched state and once several?
384: The answer can be found in Fig.~\ref{chargephibr} showing the
385: two relevant spatial scales, namely
386: the thickness of the space charge layer and the radius of
387: the channel. In the high field case, the ratio of layer thickness
388: over radius is much smaller than in the low field case. Moreover,
389: the field screening and enhancement is much stronger 
390: and the equipotential lines follow the space charge layer much better.
391: Therefore the high field streamer is much closer to interfacial
392: models as discussed in~\cite{ebe1997,arr2002,meu2004,meu2005,ebe2006}
393: and can access more branching modes. This critical state
394: in future work will be characterized by the electric charge content
395: and electric field and potential at the streamer tip which would then 
396: allow us to relate branching to the external electric circuit. 
397: For sketches of such ideas as well as for a discussion 
398: of photo-ionization effects and of continuum versus particle models,
399: we refer to \cite{ebe2006}. 
400: 
401: {\bf Acknowledgment:} C.M.\ thanks the Netherlands 
402: Organization for Scientific Research (NWO) for a Ph.D.\ grant 
403: within the program on Computational Science.
404: 
405: %\bibliographystyle{aps}
406: %\bibliography{litbranching}
407: \begin{thebibliography}{27}
408: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
409: \expandafter\ifx\csname bibnamefont\endcsname\relax
410:   \def\bibnamefont#1{#1}\fi
411: \expandafter\ifx\csname bibfnamefont\endcsname\relax
412:   \def\bibfnamefont#1{#1}\fi
413: \expandafter\ifx\csname citenamefont\endcsname\relax
414:   \def\citenamefont#1{#1}\fi
415: \expandafter\ifx\csname url\endcsname\relax
416:   \def\url#1{\texttt{#1}}\fi
417: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
418: \providecommand{\bibinfo}[2]{#2}
419: \providecommand{\eprint}[2][]{\url{#2}}
420: 
421: \bibitem[{\citenamefont{van Veldhuizen}(2000)}]{vel2000}
422: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{van Veldhuizen}},
423:   \emph{\bibinfo{title}{Electrical Discharges for Environmental Purposes:
424:   Fundamentals and Applications}} (\bibinfo{publisher}{Nova Science
425:   Publishers}, \bibinfo{year}{2000}).
426: 
427: \bibitem[{\citenamefont{Ebert et~al.}(2006)\citenamefont{Ebert, Montijn,
428:   Briels, Hundsdorfer, Meulenbroek, Rocco, and van Veldhuizen}}]{ebe2006}
429: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}},
430:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Montijn}},
431:   \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Briels}},
432:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Hundsdorfer}},
433:   \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Meulenbroek}},
434:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Rocco}}, \bibnamefont{and}
435:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{van Veldhuizen}}
436:   (\bibinfo{year}{2006}),
437:   \bibinfo{note}{{http://www.cwi.nl/ftp/CWIreports/MAS/MAS-E0529.pdf, to appear
438:   in {\em Plasma Source Sci. Technol.}}}
439: 
440: \bibitem[{\citenamefont{Bazelyan and Raizer}(2000)}]{baz2000}
441: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Bazelyan}} \bibnamefont{and}
442:   \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Raizer}},
443:   \emph{\bibinfo{title}{Lightning Physics and Lightning Protection}}
444:   (\bibinfo{publisher}{IOP Publishing}, \bibinfo{address}{Bristol, U.K.},
445:   \bibinfo{year}{2000}).
446: 
447: \bibitem[{\citenamefont{Sentman et~al.}(1995)\citenamefont{Sentman, Wescott,
448:   Osborne, and Heavner}}]{sen1995}
449: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Sentman}},
450:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Wescott}},
451:   \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Osborne}}, \bibnamefont{and}
452:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Heavner}},
453:   \bibinfo{journal}{Geophys. Res. Lett.} \textbf{\bibinfo{volume}{22}},
454:   \bibinfo{pages}{1205} (\bibinfo{year}{1995}).
455: 
456: \bibitem[{\citenamefont{Pasko and Stenbaek-Nielsen}(2002)}]{pas2002}
457: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Pasko}} \bibnamefont{and}
458:   \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stenbaek-Nielsen}},
459:   \bibinfo{journal}{Geophys. Res. Lett.} \textbf{\bibinfo{volume}{29}},
460:   \bibinfo{pages}{82(1} (\bibinfo{year}{2002}).
461: 
462: \bibitem[{\citenamefont{Gerken et~al.}(2000)\citenamefont{Gerken, Inan, and
463:   Barrington-Leigh}}]{ger2000}
464: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Gerken}},
465:   \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Inan}}, \bibnamefont{and}
466:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Barrington-Leigh}},
467:   \bibinfo{journal}{Geophys. Res. Lett.} \textbf{\bibinfo{volume}{27}},
468:   \bibinfo{pages}{2637} (\bibinfo{year}{2000}).
469: 
470: \bibitem[{\citenamefont{Raether}(1939)}]{rae1939}
471: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Raether}}, \bibinfo{journal}{Z.
472:   Phys.} \textbf{\bibinfo{volume}{112}}, \bibinfo{pages}{464}
473:   (\bibinfo{year}{1939}).
474: 
475: \bibitem[{\citenamefont{Loeb and Meek}(1940)}]{loe1940}
476: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Loeb}} \bibnamefont{and}
477:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Meek}}, \bibinfo{journal}{J.
478:   Appl. Phys.} \textbf{\bibinfo{volume}{11}}, \bibinfo{pages}{438}
479:   (\bibinfo{year}{1940}).
480: 
481: \bibitem[{\citenamefont{Dhali and Williams}(1987)}]{dha1987}
482: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Dhali}} \bibnamefont{and}
483:   \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Williams}},
484:   \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{62}},
485:   \bibinfo{pages}{4696} (\bibinfo{year}{1987}).
486: 
487: \bibitem[{\citenamefont{Vitello et~al.}(1994)\citenamefont{Vitello, Penetrante,
488:   and Bardsley}}]{vit1994}
489: \bibinfo{author}{\bibfnamefont{P.A.}~\bibnamefont{Vitello}},
490:   \bibinfo{author}{\bibfnamefont{B.M.}~\bibnamefont{Penetrante}},
491:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.N.}~\bibnamefont{Bardsley}},
492:   \bibinfo{journal}{Phys.\ Rev.~E} \textbf{\bibinfo{volume}{49}},
493:   \bibinfo{pages}{5574} (\bibinfo{year}{1994}).
494: \bibitem[{\citenamefont{Kulikovsky}(1997)}]{kul1997-2}
495: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Kulikovsky}},
496:   \bibinfo{journal}{J.~Phys.~D: Appl.\ Phys.} \textbf{\bibinfo{volume}{30}},
497:   \bibinfo{pages}{441} (\bibinfo{year}{1997}).
498: 
499: \bibitem[{\citenamefont{Pancheshnyi et~al.}(2001)\citenamefont{Pancheshnyi,
500:   Starikovskaia, and Starikovskii}}]{pan2001}
501: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Pancheshnyi}},
502:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Starikovskaia}},
503:   \bibnamefont{and}
504:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Starikovskii}},
505:   \bibinfo{journal}{J.~Phys.~D: Appl.\ Phys.} \textbf{\bibinfo{volume}{34}},
506:   \bibinfo{pages}{105} (\bibinfo{year}{2001}).
507: 
508: \bibitem[{\citenamefont{Vitello et~al.}(1993)\citenamefont{Vitello, Penetrante,
509:   and Bardsley}}]{vit1993}
510: \bibinfo{author}{\bibfnamefont{P.A.}~\bibnamefont{Vitello}},
511:   \bibinfo{author}{\bibfnamefont{B.M.}~\bibnamefont{Penetrante}},
512:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Bardsley}},
513:   \bibinfo{journal}{Non-thermal Plasma Techniques for Pollution Control, NATO
514:   ASI Ser., Ser. G} \textbf{\bibinfo{volume}{34(A)}}, \bibinfo{pages}{249}
515:   (\bibinfo{year}{1993}).
516: 
517: \bibitem[{\citenamefont{Hallac et~al.}(2003)\citenamefont{Hallac, Georghiou,
518:   and Metaxas}}]{hal2003}
519: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hallac}},
520:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Georghiou}},
521:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Metaxas}},
522:   \bibinfo{journal}{J.~Phys.~D: Appl.\ Phys.} \textbf{\bibinfo{volume}{36}},
523:   \bibinfo{pages}{2498} (\bibinfo{year}{2003}).
524: 
525: \bibitem[{\citenamefont{Georghiou et~al.}(2005)\citenamefont{Georghiou,
526:   Papadakis, Morrow, and Metaxas}}]{geo2005}
527: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Georghiou}},
528:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Papadakis}},
529:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Morrow}}, \bibnamefont{and}
530:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Metaxas}},
531:   \bibinfo{journal}{J.~Phys.~D: Appl.\ Phys.} \textbf{\bibinfo{volume}{38}},
532:   \bibinfo{pages}{R303} (\bibinfo{year}{2005}).
533: 
534: \bibitem[{\citenamefont{Array{\'a}s et~al.}(2002)\citenamefont{Array{\'a}s,
535:   Ebert, and Hundsdorfer}}]{arr2002}
536: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Array{\'a}s}},
537:   \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}}, \bibnamefont{and}
538:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Hundsdorfer}},
539:   \bibinfo{journal}{Phys.\ Rev.\ Lett.} \textbf{\bibinfo{volume}{88}},
540:   \bibinfo{pages}{174502} (\bibinfo{year}{2002}).
541: 
542: \bibitem[{\citenamefont{Rocco et~al.}(2002)\citenamefont{Rocco, Ebert, and
543:   Hundsdorfer}}]{roc2002}
544: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Rocco}},
545:   \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}}, \bibnamefont{and}
546:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Hundsdorfer}},
547:   \bibinfo{journal}{Phys.\ Rev.~E} \textbf{\bibinfo{volume}{66}},
548:   \bibinfo{pages}{035102(R)} (\bibinfo{year}{2002}).
549: 
550: \bibitem[{\citenamefont{Liu and Pasko}(2004)}]{liu2004}
551: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Liu}} \bibnamefont{and}
552:   \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Pasko}},
553:   \bibinfo{journal}{J.\ Geophys.\ Res.} \textbf{\bibinfo{volume}{109}},
554:   \bibinfo{pages}{A04301} (\bibinfo{year}{2004}).
555: 
556: \bibitem[{\citenamefont{Kulikovsky}(2002)}]{kul2002}
557: \bibinfo{author}{\bibfnamefont{A.A.}~\bibnamefont{Kulikovsky}},
558:   \bibinfo{journal}{Phys.\ Rev.\ Lett.} \textbf{\bibinfo{volume}{89}},
559:   \bibinfo{pages}{229401} (\bibinfo{year}{2002}).
560: 
561: \bibitem[{\citenamefont{Pancheshnyi and Starikovskii}(2003)}]{pan2003}
562: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Pancheshnyi}} \bibnamefont{and}
563:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Starikovskii}},
564:   \bibinfo{journal}{J.~Phys.~D: Appl.\ Phys.} \textbf{\bibinfo{volume}{36}},
565:   \bibinfo{pages}{2683} (\bibinfo{year}{2003}).
566: 
567: \bibitem[{\citenamefont{Meulenbroek et~al.}(2004)\citenamefont{Meulenbroek,
568:   Rocco, and Ebert}}]{meu2004}
569: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Meulenbroek}},
570:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Rocco}}, \bibnamefont{and}
571:   \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}},
572:   \bibinfo{journal}{Phys.\ Rev.~E} \textbf{\bibinfo{volume}{69}},
573:   \bibinfo{pages}{067402} (\bibinfo{year}{2004}).
574: 
575: \bibitem[{\citenamefont{Meulenbroek et~al.}(2005)\citenamefont{Meulenbroek,
576:   Ebert, and Sch{\"a}fer}}]{meu2005}
577: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Meulenbroek}},
578:   \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}}, \bibnamefont{and}
579:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Sch{\"a}fer}},
580:   \bibinfo{journal}{Phys.\ Rev.\ Lett.} \textbf{\bibinfo{volume}{95}},
581:   \bibinfo{pages}{195004} (\bibinfo{year}{2005}).
582: 
583: \bibitem[{\citenamefont{Ebert and Hundsdorfer}(2002)}]{ebe2002-2}
584: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}} \bibnamefont{and}
585:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Hundsdorfer}},
586:   \bibinfo{journal}{Phys.\ Rev.\ Lett.} \textbf{\bibinfo{volume}{89}},
587:   \bibinfo{pages}{229402} (\bibinfo{year}{2002}).
588: 
589: 
590: \bibitem[{\citenamefont{Ebert et~al.}(1997)\citenamefont{Ebert, van Saarloos,
591:   and Caroli}}]{ebe1997}
592: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}},
593:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{van Saarloos}},
594:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Caroli}},
595:   \bibinfo{journal}{Phys.\ Rev.~E} \textbf{\bibinfo{volume}{55}},
596:   \bibinfo{pages}{1530} (\bibinfo{year}{1997}).
597: 
598: \bibitem[{\citenamefont{Montijn et~al.}(2006)\citenamefont{Montijn,
599:   Hundsdorfer, and Ebert}}]{mon2005-2}
600: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Montijn}},
601:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Hundsdorfer}},
602:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}}
603:   (\bibinfo{year}{2006}),
604:   \bibinfo{note}{{\\http://arxiv.org/abs/physics/0603070}}.
605: 
606: \bibitem[{\citenamefont{Ebert and van Saarloos}(2000)}]{ebe2000}
607: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Ebert}} \bibnamefont{and}
608:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{van Saarloos}},
609:   \bibinfo{journal}{Physica D} \textbf{\bibinfo{volume}{146}},
610:   \bibinfo{pages}{1} (\bibinfo{year}{2000}).
611: 
612: \bibitem[{\citenamefont{Wackers}(2005)}]{wac2005}
613: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Wackers}}, \bibinfo{journal}{J.
614:   Comp. Appl. Math.} \textbf{\bibinfo{volume}{180}}, \bibinfo{pages}{1}
615:   (\bibinfo{year}{2005}).
616: 
617: \end{thebibliography}
618: 
619: 
620: \raggedbottom
621: \end{document}
622: 
623: