0808.3195/ms.tex
1:  \documentclass[12pt,preprint]{aastex}
2: 
3: \slugcomment{Version \today}
4: 
5: \shorttitle{Oblique Ion Two-Stream Instability}
6: \shortauthors{Ohira and Takahara}
7: 
8: \begin{document}
9: 
10: \title{Oblique Ion Two-Stream Instability in the Foot Region of 
11: a Collisionless Shock}
12: 
13: \author{Yutaka Ohira and Fumio Takahara}
14: \affil{Department of Earth and Space Science, Graduate School
15: of Science, Osaka University, Machikaneyama 1-1, Toyonaka, 
16: Osaka 560-0043, Japan \\
17: {\rm yutaka@vega.ess.sci.osaka-u.ac.jp,
18:  takahara@vega.ess.sci.osaka-u.ac.jp}}
19: 
20: \begin{abstract}
21: Electrostatic behavior of a collisionless plasma in the foot region 
22: of high Mach number perpendicular shocks is investigated through 
23: the two-dimensional linear analysis and electrostatic 
24: particle-in-cell (PIC) simulation. The simulations are double 
25: periodic and taken as a proxy for the situation in the foot. 
26: The linear analysis for relatively cold unmagnetized plasmas with 
27: a reflected proton beam shows that obliquely propagating Buneman 
28: instability is strongly excited. We also found that when the electron 
29: temperature is much higher than the proton temperature, the most 
30: unstable mode is the highly obliquely propagating ion two-stream 
31: instability excited through the resonance between ion plasma 
32: oscillations of the background protons and of the beam protons, 
33: rather than the ion acoustic instability that is dominant for 
34: parallel propagation. 
35:  
36: To investigate nonlinear behavior of the ion two-stream instability, 
37: we have made PIC simulations for the shock foot region in which the 
38: initial state satisfies the Buneman instability condition. 
39: In the first phase, electrostatic waves grow two-dimensionally by the 
40: Buneman instability to heat electrons. In the second phase, 
41: highly oblique ion two-stream instability grows to heat mainly ions. 
42: This result is in contrast to previous studies based on one-dimensional 
43: simulations, for which ion acoustic instability further heats electrons.  
44:  
45: The present result implies that overheating problem of electrons for 
46: shocks in supernova remnants is resolved by considering ion two-stream 
47: instability propagating highly obliquely to the shock normal and that 
48: multi-dimensional analysis is crucial to understand the particle heating 
49: and acceleration processes in shocks.
50: \end{abstract}
51: 
52: \keywords{supernova remnants -- shock waves -- plasmas -- instabililes --
53: cosmic rays -- acceleration of particle}
54: 
55: \section{Introduction}
56: The discovery of thermal and synchrotron X-rays from young supernova 
57: remnants (SNRs) provides the evidence that electrons are heated up to 
58: a few keV and that a portion of them are accelerated to highly 
59: relativistic energy in SNR shocks \citep{koy95}. Because SNR shocks 
60: are collisionless, not only particle acceleration mechanisms but also 
61: electron heating mechanisms in SNR shocks are not so simple. Previous 
62: studies have given an important key to the formation mechanism of 
63: perpendicular collisionless shocks. When the Alfv${\rm \acute{e}}$n Mach 
64: number $M_{\rm A}$ is larger than the critical Mach number, about 3, a 
65: perpendicular shock reflects some of the incident ions to the upstream, 
66: where a foot region forms on a spatial scale of the ion gyroradius 
67: \citep{ler83}. The plasma in the foot region consists of incident ions 
68: and electrons and reflected ions and returning ions which are made from 
69: reflected ions and move to the shock after a gyration. As for the 
70: electron heating machanism, \citet{pap88} proposed that when the Mach 
71: number is larger than $0.5(m_{\rm p}/m_{\rm e})^{1/2}\sim 20$, incident 
72: electrons and reflected ions excite electrostatic waves by the Buneman 
73: instability \citep{bun58} because the relative velocity between them is 
74: large compared with the electron thermal velocity. They also suggest 
75: that after electrons are  heated by electrostatic waves induced by the 
76: Buneman instability, ion acoustic instability is triggered because the 
77: electron temperature becomes much higher than the proton temperature. 
78: As a result, electrons are strongly heated by the Buneman instability 
79: and the ion acoustic instability. \citet{car88} performed a 
80: one-dimensional hybrid simulation and demonstrated that strong electron 
81: heating actually occurs. They concluded that an $M_{\rm A}=500$ shock 
82: heats electrons by a factor of $10^5$ across the shock. This means that 
83: if the upstream electron temperature is 1 eV, the downstream electron 
84: temperature becomes 100 keV. This value of the downstream temperature 
85: is much larger than the recent observational one for SNRs, a few keV 
86: \citep{sta06}. This discrepancy has been an open issue to be resolved 
87: for a long time.
88: 
89: On the other hand, \citet{shi00} and \cite{hos02} performed 
90: one-dimensional full particle-in-cell (PIC) simulations to investigate 
91: the electron acceleration at perpendicular shocks. Their simulation 
92: solves a whole region of a collisionless perpendicular shock and makes 
93: reflected ions self-consistently by employing a small proton to 
94: electron mass ratio. Their results showed that electrons are not only 
95: heated at the foot region but also significantly accelerated by surfing 
96: acceleration mechanism. However, this acceleration is valid only for 
97: the one-dimensional case because the surfing acceleration strongly 
98: depends on the structure of the electrostatic potential. In our first 
99: paper \citep{ohi07}, we performed two-dimensional electrostatic PIC 
100: simulations to solve for the two-dimensinal structure of the 
101: electrostatic potential excited by the Buneman instability. We 
102: employed the real mass ratio but the simulation region is limited to 
103: the foot region. Our results showed that oblique modes grow as strongly 
104: as the modes parallel to the beam direction, that the potential 
105: structure becomes two-dimensional and that no efficient surfing 
106: acceleration occurs, while electron heating occurs. Thus, the problem 
107: of electron acceleration has been back to the start again. In that 
108: paper, we concentrated on the stage of the Buneman instability and did 
109: not follow the long time scale evolution after the Buneman nstability 
110: has saturated.
111: 
112: In this paper, we study the time evolution of electrostatic collisionless 
113: plasma instabilities in the foot region by making linear analysis and 
114: by performing two-dimensional electrostatic PIC simulation. We perform 
115: simulations with a higher resolution, a larger simulation box and a 
116: longer simulation time than in \citet{ohi07}. Especially, we focus on 
117: the evolution after electrostatic waves excited by the Buneman 
118: instability have decayed. Our simulation substantially improves most 
119: previous works that are one-dimensional, employ an artifically small 
120: proton electron mass ratio or impose rather strong magnetic field. It 
121: is obvious that a multi-dimensional analysis is necessary as discussed 
122: above \citep{blu60, lam74, ohi07}. Our motivation for employing the real 
123: mass ratio is as follows. For a small mass ratio, the foot region in 
124: the simulation is shorter than the realistic one and the time scale on 
125: which electrons stay in the foot region in the simulation is also 
126: shorter than the realistic one because the size of the foot region is 
127: about the ion gyroradius $m_{\rm p}v_{\rm d}c/eB$, where $v_{\rm d}$ 
128: and $B$ are the drift velocity of reflected protons and the magnetic 
129: field, respectively. Because reflected ions have a large free energy, 
130: we expect that more energy is transported to electrons through 
131: collective instabilities with the realistic mass ratio in the foot region. 
132: The drift velocity is not large enough to excite electromagnetic 
133: waves, so that electrostatic waves are more important.
134: 
135: In \S 2 we perform linear analysis for two-dimensional electrostatic 
136: modes. In \S 3 we describe the initial setting of the PIC simulations 
137: and numerical results, followed by a discussion in \S 4.  
138: 
139: 
140: \section{Linear analysis}
141: In this section, we perform linear analysis for two-dimensional 
142: electrostatic modes in unmagnetized plasmas with beams. In the foot 
143: region of perpendicular shocks of SNRs, we regard that there are several 
144: beams with a finite temperature and that their relative velocities are 
145: much smaller than the light speed. Therefore, the fastest growing modes 
146: are electrostatic modes. Then, we here concentrate on the electrostatic 
147: modes. For typical interstellar medium, the ratio 
148: %
149: \begin{equation}
150: \frac{ \Omega_{\rm ce} }{ \omega_{\rm pe} } 
151: \simeq 10^{-3} \left(\frac{B}{3\rm{\mu G}}\right) 
152: \left(\frac{n_{\rm e}}{1\rm{cm}^{-3}}\right)^{-1/2}
153: \end{equation}
154: %
155: is relatively small, where $\Omega_{\rm ce}, \omega_{\rm pe}$ and 
156: $n_{\rm e}$ are electron cyclotron frequency, electron plasma frequency, 
157: and electron number density, respectively. So plasma oscillations are 
158: hardly changed by the magnetic field in the foot region of shocks of 
159: SNRs. When we consider spatial scale smaller than the gyroradius, 
160: we may neglect the effects of magnetic fields. Thus, we concentrate here 
161: on unmagnetized plasmas. 
162: 
163: We define such that the $x$-direction is shock normal direction and 
164: the $y$-direction is the direction that is perpendicular to shock normal 
165: and wave vectors are on the $x-y$ plane. For unmagnetized collisionless 
166: plasmas, the electrostatic dispersion relation reads as
167: %
168: \begin{eqnarray}
169: &&1+\sum_{\rm s} \frac{\omega_{\rm ps}^2}{k^2}
170: \int d^2 v \frac{ {\bf k} \cdot {\bf \nabla}_v f_{\rm s0}}
171: {\omega - {\bf k} \cdot {\bf v}} = 0,  \\
172: &&k=\sqrt{k_{x}^2+k_{y}^2},
173: \end{eqnarray} 
174: %
175: where the subscript s represents particle species, here electrons, ions 
176: and beam ions, $\omega_{\rm ps}=(4\pi n_{\rm s}e^2/m_{\rm s})^{1/2}$ is 
177: the plasma frequency of the particle species s and $f_{\rm s0}$ is the 
178: normalized distribution function of the particle species s,
179: %
180: \begin{equation} 
181: f_{\rm s0}
182: =\frac{1}{\pi v_{\rm th,s}^2} 
183: \exp \left[-\frac{(v_{x}-v_{\rm d,s})^2+v_{y}^2}
184: {v_{\rm th,s}^2}\right]
185: \end{equation} 
186: %
187: where $v_{\rm th,s}=(2kT_{\rm s}/m_{\rm s})^{1/2}$ and $v_{\rm d,s}$ is 
188: the thermal velocity and drift velocity of the particle species s, 
189: respectively. 
190: 
191: To make equation (2) simpler, we use new coordinates $x'$ and $y'$ as 
192: %
193: \begin{eqnarray}
194: && x'=x \cos \theta+y \sin \theta , \nonumber \\
195: && y'=-x \sin \theta +y \cos \theta ,\\
196: && \cos \theta = \frac{k_{x}}{\sqrt{k_{x}^2 + k_{y}^2}}. 
197: \nonumber
198: \end{eqnarray}
199: %
200: Then, equations (2) and (3) become 
201: %
202: \begin{equation}
203: 1+\sum_{\rm s} \frac{\omega_{\rm ps}^2}{k}
204: \int d^2 v' \frac{\partial f_{\rm s0}/\partial v'_{x}}
205: {\omega - k v'_{x}} = 0,
206: \end{equation}
207: %
208: and
209: %
210: \begin{equation}
211: f_{\rm s0}=\frac{1}{\pi v_{\rm th,s}^2} 
212: \exp \left[-\frac{(v'_{x}-v_{\rm d,s}\cos \theta)^2
213: +(v'_{y}+v_{\rm d,s}\sin \theta)^2}{v_{\rm th,s}^2}\right].
214: \end{equation}
215: %
216: Finally, we substitute equation (7) into  equation (6), and we obtain
217: %
218: \begin{eqnarray}
219: &&1+\sum_{\rm s} \frac{2\omega_{\rm ps}^2}{k^2v_{\rm th,s}^2}
220: \left[1+\xi_{\rm s} Z(\xi_{\rm s})\right]=0, \nonumber \\
221: &&Z(\xi_{\rm s}) = \frac{1}{\sqrt{\pi}}
222: \int_{-\infty}^{\infty}\frac{e^{-z^2}}{z-\xi_{\rm s}}dz, \\
223: &&\xi_{\rm s} = \frac{\omega-k_{x}v_{\rm d,s}}{kv_{\rm th,s}}, \nonumber
224: \end{eqnarray}
225: where $Z(\xi_{\rm s})$ is the plasma dispersion function and it can be 
226: numerically solved \citep{wata91}. 
227: 
228: We here present results of the linear analysis about two cases of plasma 
229: conditions and discuss on three kinds of plasma instabilities. 
230: 
231: \subsection{Buneman instability}
232: 
233: We first consider the situation in which there are three beams, incident 
234: protons, incident electrons and reflected protons and the temperatures 
235: of all plasma beams are low, typically around 1 eV. We thus we neglect 
236: the contribution of returning ions in the dispersion relation, for 
237: simplicity. As is easily understood, the returning component plays a role 
238: in assuring the vanishing net current in the unperturbed state. In the 
239: dispersion relation, they play a symmetrical role to the reflected ions 
240: and their effects are starightforwardly understood when we make clear 
241: the role of reflected ions. We make analyses in the upstream rest frame 
242: in which only reflected ions have a drift velocity, typically 
243: $v_{\rm d}=v_{\rm d,ref}=0.02c=2v_{\rm sh}$, where $v_{\rm sh}$ is the 
244: shock velocity. Hence, a typical velocity ratio is 
245: $v_{\rm d}/v_{\rm th,e}=10$. We assume that the proton reflection ratio 
246: is $n_{\rm ref}/n_{\rm p}=0.25$. 
247: 
248: The growth rate obtained by solving the linear dispersion relation is 
249: displayed in Figure 1(a). In this condition, the most unstable mode is 
250: the Buneman instability. The Buneman instability is caused by the 
251: resonance between the electron plasma oscillation of the upstream 
252: electrons and proton plasma oscillation of the reflected proton beam. 
253: In Figure 1, $k_{x}$ and $k_{y}$ are wavenumbers normalized by 
254: $\omega_{\rm pe}/v_{\rm d}$ and the color contours show the growth rate 
255: normalized by $\omega_{\rm pe}$, where only the growth rate of growing 
256: modes is shown. As is seen, the growth rate of obliquely propagating 
257: modes is as large as that of modes parallel to the beam direction.  
258: This feature of the Buneman instability can be well understood in the 
259: cold limit. We present the results of the cold limit for which all 
260: temperatures are set to zero in Figure 1(b). In the cold limit, the 
261: distribution function (4) becomes 
262: %
263: \begin{equation}
264: f_{\rm s0}=\delta(v_{x}-v_{\rm d,s})\delta(v_{y}),
265: \end{equation}
266: %
267: and the dispersion relation (2) is reduced to
268: %
269: \begin{equation}
270: 1-\sum_{\rm s} \left( \frac{\omega_{\rm ps}}
271: {\omega-k_{x} v_{\rm d,s}} \right)^2 = 0.
272: \end{equation} 
273: %
274: Because $k_{y}$ does not appear in equation (10), the growth rate of 
275: electrostatic instabilities in the cold limit does not depend on 
276: $k_{y}$. Therefore, in the cold limit, excited waves have any $k_{y}$ 
277: and the structure of electrostatic potential to the $y$-direction is 
278: strongly disordered and loses coherence to the $y$ direction. This 
279: feature is very important to negate the electron surfing acceleration 
280: mechanism \citep{ohi07}. The maximum growth rate of the Buneman 
281: instability in the cold limit is \citep{bun58}
282: %
283: \begin{equation}
284: \gamma_{\rm max} 
285: = \left[\frac{3\sqrt{3}}{16} 
286: \left(\frac{m_{\rm e}}{m_{\rm p}}\right) 
287: \left(\frac{n_{\rm ref}}{n_{\rm e}}\right)\right]^{1/3}
288: \omega_{\rm pe}  \ \ 
289: {\rm at} \ \ k_x =\frac{\omega_{\rm pe}}{v_{\rm d}}
290: =k_{\rm Bun}. \nonumber 
291: \end{equation}
292: 
293: In reality, because of a finite temperature, the modes with large 
294: wavenumbers are suppressed to grow. This is seen in Figure 1(a); 
295: the growth rate for large $k_y$ decreases. The dispersion relation of 
296: the Buneman instability depends on $v_{\rm th,e}/v_{\rm d}$ and the 
297: number density ratio. When $v_{\rm th,e}/v_{\rm d}$ is small, modes 
298: that have a large $k_{y}$ can grow as long as the wavelength is larger 
299: than the  electron Debye length. The wavenumber corresponding to the 
300: electron Debye length is 
301: $k_{\rm D,e}= \omega_{\rm pe}/v_{\rm th,e}=10k_{\rm Bun}$ 
302: in the present plasma condition and the boundary between 
303: growing and damping region in Figure 1(a) is about 
304: $k_{y} =7k_{\rm Bun}\sim k_{\rm D,e}$, as is consistent with the present 
305: consideration. This result implies that in SNRs condition, oblique 
306: modes propagating to the beam direction, i.e., to the shock normal, 
307: can grow as strongly as the modes to the parallel direction. 
308: Because these modes are electrostatic, the direction of the excited 
309: electric fields is  to the wave vector and the energy density of the 
310: electric field of the $y$-component can be larger than that of 
311: the $x$-component. 
312: 
313: \subsection{Ion two-stream instability and ion acoustic instability}
314: After the upstream electrons are heated by the Buneman instability, 
315: the thermal velocity of electrons increases up to 
316: $v_{\rm th,e} \sim v_{\rm d}$ while the thermal velocity of protons 
317: is roughly the same as the initial one, so that the situation of 
318: $T_{\rm e} \gg T_{\rm p}$ is realized. The other parameters are the 
319: same as in the low temperature case described in the previous 
320: subsection. In this condition, while the Buneman instability is 
321: stabilized, other types of instabilities can occur. We found that 
322: in addition to the ion acoustic instability discussed previously 
323: \citep{car88}, the ion two-stream instability that has not been 
324: well noticed in the literature becomes unstable. The ion two-stream 
325: instability is caused by the resonance of the ion plasma oscillation 
326: of the upstream plasma and that of reflected ions in the situation 
327: $T_{\rm e} \gg T_{\rm p}$. Thus, it occurs concurrently with the ion 
328: acoustic instability which is caused by the resonance between ion 
329: acoustic waves of the reflected protons and electron plasma oscillation 
330: of the upstream plasma, where the former modes are mediated by the 
331: presence of the hot upstream electrons. 
332: 
333: The numerical results of the growth rate for ion two-stream instability 
334: and ion acoustic instability are shown in Figures 2(a) and (b), 
335: respectively. In Figure 2, $k_{x}$, $k_{y}$ and the growth rate are 
336: normalized in the same way as in Figure 1, and the color contours show 
337: the growth rate of the growing modes. Note that the difference of the 
338: range of the wavenumber in the $x$-direction $k_x$ between (a) and (b). 
339: While ion acoustic instability grows at around 
340: $k_x\approx \omega_{\rm pe}/v_{\rm d}$, ion two-stream instabily grows 
341: at much smaller $k_y$. It is also noted that ion two-stream instability 
342: is seen for low but finite values of $k_y$, In the present conditions, 
343: the wavenumber corresponding to the ion Debye length is 
344: $k_{\rm D,p} = \omega_{\rm pi}/v_{\rm th,i}=10k_{\rm Bun}$ while that to the 
345: electron Debye length is $k_{\rm Bun}$. Ion plasma oscillation exists 
346: between these two wavenumbers. For wavenumbers lower than $k_{\rm Bun}$, 
347: it reduces to the ion acoustic mode while for wavenumbers higher than 
348: $k_{\rm D,p}$ it damps by thermal motions of ions. It should be noted 
349: that in Figure 2(a), ion two-stream modes of parallel propagation to the 
350: beam direction are only weakly growing, but that highly oblique modes grow 
351: very fast and the maximum growth rate is larger than that of the ion 
352: acoustic instability by a factor of a few. The reason is explained as 
353: follows. To excite the ion two-stream instability, the resonance 
354: condition, $k_{x} v_{\rm d} \sim \omega_{\rm pi}$ must be satisfied in 
355: addition to the wavenumber condition mentioned above. The resonance 
356: condition requires a small 
357: $k_x \approx (m_{\rm e}/m_{\rm p})^{1/2}k_{\rm Bun}$. For propagation 
358: parallel to the beam direction, this is imcompatible with the wavenumber 
359: condition and the growth rate is very small. In contrast, when the wave 
360: has a large $k_{y}$, both the wavenumber condition and the resonance 
361: condition are fulfilled simultaneously and a larger growth rate is obtained. 
362: The results shown in Figure 2 (a) are fully consistent with this picture 
363: of the ion two-stream instability.
364: 
365: 
366: To understand the ion two-stream instability through the dispersion 
367: relation, we consider a situation where the proton temperature 
368: is zero and the electron temperature is very high and $k_y \gg k_x$ 
369: (if the electron drift velocity is much smaller than thermal velocity, 
370: we do not need the final condition). Then, we can approximate as  
371: $|\xi_{\rm p}|, |\xi_{\rm ref}| \gg 1$ and $|\xi_{\rm e}| \ll 1$, 
372: and the dispersion relation becomes
373: %
374: \begin{equation}
375: 1+2\left(\frac{k_{\rm D,e}}{k}\right)^2- \left( \frac{\omega_{\rm pp}}
376: {\omega%-k_{x} v_{\rm d,p}
377: } \right)^2-\left( \frac{\omega_{\rm ref}}
378: {\omega-k_{x} v_{\rm d,ref}} \right)^2= 0.
379: \end{equation} 
380: %
381: The second term represents the Debye shielding effect of hot electrons 
382: and becomes small for $k \gg k_{\rm D,e}$ as discussed above. 
383: It is seen that for $k \gg k_{\rm D,e}$, the dispersion relation has 
384: the same form as equation (10) and we obtain the maximum growth rate of 
385: the ion two-stream instability when $n_{\rm ref} \ll n_{\rm p}$, 
386: by replacing $\omega_{\rm pe}$ with $\omega_{\rm pp}$ in equation (11), as 
387: %
388: \begin{equation}
389: \gamma_{\rm max} 
390: = \left[\frac{3\sqrt{3}}{16} 
391: \left(\frac{n_{\rm ref}}{n_{\rm p}}\right)\right]^{1/3}
392: \omega_{\rm pp}  \\ 
393: {\rm at} \ \ k_x =\frac{\omega_{\rm pp}}{v_{\rm d}}. \nonumber 
394: \end{equation}
395: %
396: Because the upstream plasma stays in foot region by about the proton 
397: gyro-period $\Omega_{\rm cp}^{-1}$ and because 
398: $\Omega_{\rm cp}/\omega_{\rm pp}\sim 4.3 \times 10^{-5}$, ion two-stream 
399: instability can grow enough in the foot region. 
400: 
401: As far as we are aware, this oblique unstable mode has not been 
402: considered up to now. We expect that this instability heats ions and that 
403: much affects subsequent electron heating processes. 
404: 
405: \section{Simulation}
406: To perform two-dimensional simulations with real proton electron mass 
407: ratio, we confine our attention to the foot region through a proper 
408: modeling instead of solving the whole shock structure. Our simulation 
409: box is taken to be at rest in the upstream frame of reference, i.e., 
410: that of incident protons and electrons. We do not solve electromagnetic 
411: waves and concentrate on electrostatic waves.
412: 
413: \subsection{Setting}
414: We define the $x$-direction as the shock normal pointing to the shock 
415: front, and thus the reflected protons move in the $-x$-direction and 
416: returning protons move in the $x$-direction. The magnetic field is 
417: taken to be spatially homogeneous pointing in the $z$-direction and 
418: we solve the particle motion and electric field in the $x-y$ plane. 
419: As the initial condition, we prepare upstream electrons, upstream 
420: protons, reflected protons and returning protons. Each population is 
421: uniformly distributed in the $x-y$ plane and their momentum 
422: distribution is given by a Maxwellian at the same temperatures 
423: $T=T_{\rm e}=T_{\rm p}=T_{\rm ref}=T_{\rm ret}=1.75{\rm eV}$. In 
424: addition, reflected and returning protons have an extra drift velocity 
425: in the $x$-direction of $v_{\rm d}=\pm0.04c$ ($v_{\rm d}=2v_{\rm sh}$). 
426: The number densities of each population are taken as 
427: $n_{\rm e}=1.5n_{\rm p}=1\rm{cm}^{-3}$ and $n_{\rm ref}
428: =n_{\rm ret}=0.25n_{\rm p}$, where subscripts e, p, ref and ret 
429: represent upstream electrons, upstream protons, reflected protons and 
430: returning protons, respectively (see Figure 3). These parameters are 
431: typical of young SNRs and satisfy the charge neutrality and a vanishing 
432: current. 
433:  
434: We employ the periodic boundary condition both in the $x$- and 
435: $y$-directions. The electric field is solved by the Poisson equation. We 
436: have examined two cases of the background magnetic field,  0 and 
437: 90 $\mu $G ($\Omega_{\rm ce}/\omega_{\rm pe}=0$ and 0.03), where 
438: $\Omega_{\rm ce}=eB/m_{\rm e} c$ is the electron cyclotron frequency. 
439: We sometimes refer the former and latter cases to the unmagnetized and 
440: magnetized cases, respectively. 
441:  
442: The size of the simulation box to the $x$- and $y$-directions is taken 
443: to be $L_{x}=64\lambda_{\rm Bun}$ {\rm and} $L_{y}=16\lambda_{\rm Bun}$, 
444: with a total of 2048 $\times$ 512 cells, where 
445: $\lambda_{\rm Bun}=2\pi \omega_{\rm pe}/v_{\rm d}$ is the wavelength 
446: of the most unstable mode of the Buneman instability.  
447: Thus, the length of each cell $\Delta x=\Delta y$ is 3 times the 
448: initial electron Debye length. The number of macroparticles is taken 
449: so that initially each cell includes 96 electrons and 96 total protons. 
450: The time step $\Delta t$ is taken as $5\times 10^{-3}\omega_{\rm pe}^{-1}$ 
451: and the simulation is followed until $3 \times10^3\omega_{\rm pe}^{-1}$ 
452: or $1.6 \times 10^{-3} \Omega_{\rm cp}^{-1}$ where $\Omega_{\rm cp}^{-1}$ 
453: corresponds to the time scale the upstream plasma stays in the foot region.
454: 
455: The differences from previous simulations \citep{ohi07} are as follows. 
456: First, the initial temperature is lower than the previous one $7{\rm eV}$. 
457: This is a more realistic one because the typical temperature of the 
458: interstellar matter is about 1eV. Secondly, we add returning proton beam 
459: in order that the total current vanishes, although in the electrostatic 
460: simulation it is not so critical. Thirdly, simulation time and simulation 
461: box are larger than the previous values so that we can investigate the 
462: ion two-stream instability. 
463: 
464: \subsection{Results}
465: 
466: Although we have performed simulations for two cases of the magnetic 
467: field strength ($0, 90\mu G$), the results turn out to be almost the 
468: same. Hence, we present the results of the unmagnetized case and add 
469: those of the magnetized case when necessary. 
470: 
471: First, we discuss the time development of the electric field. The 
472: evolution of the spatially averaged energy density of the electric 
473: field is shown in Figure 4. The solid and dashed curves show the 
474: $x$- and $y$-components, respectively, and bold and thin curves are 
475: unmagnetized and magnetized cases, respectively. In the first stage 
476: for $t<250\omega_{\rm pe}^{-1}$, the Buneman instability occurs and 
477: the electric field to both directions grow. After they attain peak 
478: values, they continue to decay till around 
479: $t\approx 10^3\omega_{\rm pe}^{-1}$. Then, after 
480: $t>10^3\omega_{\rm pe}^{-1}$, the $y$-component of the electric field 
481: starts to grow again while the $x$-component continues to decay. 
482: This feature is due to the ion two-stream instability as discussed below.
483: 
484: It should be noted that in the first stage of the Buneman instability 
485: ($0<t<250\omega_{\rm pe}^{-1}$), the $y$-component of electric field 
486: is larger than the $x$-component. This is different from our previous 
487: result. As discussed in \S 2, this is because the temperature is lower 
488: and the waves with a larger obliqueness grow faster compared with our 
489: previous simulation \citep{ohi07}. In this stage, only electron 
490: temperature has risen up to about $m_{\rm e} v_{\rm d}^2$, but the 
491: ion temperature little changes (see Figure 5). Here, we define the 
492: temperature by the velocity dispersion, 
493: $T_{\rm s}\equiv m_{\rm s}\langle(v-\langle v\rangle)^2\rangle/2k_{\rm B}$.
494: 
495: In contrast to \citet{car88}, after the electrostatic waves caused by 
496: the Buneman instability decay at about $t=10^3 \omega_{\rm pe}^{-1}$, 
497: only the $y$-component of the electric field grows and oscillates 
498: after the amplitude saturates. In contrast, the $x$-component continues 
499: to decay. Of course, this feature can not be seen in one-dimensional 
500: simulations. The growth of the $y$-component of electric field is 
501: caused by highly oblique ion two-stream instability. If we consider 
502: only the parallelly propagating modes as in the one-dimesional simulations, 
503: ion acoustic instability has the largest growth rate. In the 
504: two-dimensional simulations we can take obliquely propagating modes 
505: with $k_{y}\neq 0$ into account, and the growth rate of the ion 
506: two-stream instability is larger than that of the ion acoustic 
507: instability as mentioned \S 2. A snapshot of the electrostatic 
508: potential at the saturation phase of the ion two-stream instabilities 
509: is shown in Figure 6. In Figure 6, spatial coordinates are normalized 
510: by $v_{\rm d}/\omega_{\rm pe}$, and the color contours show the 
511: electrostatic potential normalized by $m_{\rm e}v_{\rm d}^2/2$. 
512: Typical wave length scale to the $x$-direction is the order of the 
513: resonance scale $2\pi v_{\rm d} / \omega_{\rm pi}$ and that to 
514: the $y$-direction is comparable to $2\pi/k_{\rm D,e}$ as the linear 
515: analysis predicts. Before the saturation stage, the wavelength is 
516: smaller than $2\pi$ electron Debye length for which the linear growth 
517: rate is higher. 
518: 
519: Time development of the temperatures is shown in Figure 5. The 
520: electron temperature rises rapidly up to about $m_{\rm e} v_{\rm d}^2$ 
521: when the Buneman instability grows, while at this phase the proton 
522: temperature does not rise so much. At about 
523: $t=10^3 \omega_{\rm pe}^{-1}$, the proton temperature begins to rise 
524: by the growth of the ion two-stream instability but the electron 
525: temperature is kept at almost the same because this instability occurs 
526: between two ion beams and the energy density of the electric field is 
527: nearly 100 times smaller than the thermal energy density of electrons 
528: at this stage. At the end of the simulation, upstream proton 
529: temperature becomes about 100 times larger than the initial value 
530: and the electron to proton temperature ratio $T_{\rm e} / T_{\rm p}$ 
531: becomes about 10. For the magnetized case, $T_{\rm e} / T_{\rm p}$ 
532: becomes about 7. Although it is not explicitly shown in this paper, 
533: the proton distribution has a large anisotropy. Only the proton 
534: temperature of the $y$-direction rises up, but that of the 
535: $x$-direction is kept to be almost constant.
536:  
537: Figure 7 shows the energy distribution of electrons at the end of the 
538: simulation. The bold and thin curves represent non-mangetized and 
539: magnetized cases, respectively. It is noted that no high energy tail 
540: is seen. Even when there exists a magnetic field, electrons are not 
541: accelerated, that is, no surfing acceleration occurs in two-dimensional 
542: simulations. Both curves are of a flat top shape and can not be fitted 
543: by a Maxwellian distribution. If one draws Maxwellian distribution 
544: in Figure 7, it is a straight line with an inclination of 
545: $1/k_{\rm B}T_{\rm e}$. As discussed in \citet{ohi07}, at the end 
546: of simulation, the electron temperature becomes 
547: $T_{\rm e}\sim 0.5 m_{\rm e}v_{\rm d}^2$. These results are consistent 
548: with our previous result \citep{ohi07}. 
549: 
550: \section{Discussion}
551: Now we discuss the final outcome of the ion two-stream instability. 
552: It becomes unstable when two conditions are satisfied; one is the 
553: resonance condition, $k_{x} v_{\rm d} \sim \omega_{\rm pi}$ and the 
554: other is that the wave length be between the ion Debye length and 
555: the electron Debye length. Thus, the ion two-stream instability 
556: becomes stabilized when $T_{\rm p} \sim T_{\rm e}$. At the end of 
557: our simulation, the $y$-component of electric field has not decayed 
558: still completely. So, we expect $T_{\rm e}/T_{\rm p} < 10$ in the 
559: final stage, and probably ions will be heated up to 
560: $T_{\rm i} \sim m_{\rm e} v_{\rm d} ^2 \sim 4m_{\rm e} v_{\rm sh}^2$ 
561: at the foot region in high Mach number perpendicular shocks. 
562: Of course, because the proton temperature of the drift direction is 
563: still cold in the present simulation, we must check the isotropilazation 
564: process of ion velocity distribution by doing longtime full PIC-simulations.  
565: 
566: As for the electrons, the electron temperature in the foot region is 
567: also about $m_{\rm e} v_{\rm sh}^2$. In the later stage, the growth of 
568: the two-stream instability dominates over the ion acoustic instability 
569: and little electron heating occurs. Hence, as mentioned in \citet{ohi07}, 
570: if other electron heating mechanisms do not exist, after passing the 
571: shock front, electrons will undergo the adiabatic heating and finally, 
572: the electron temperature in the downstream becomes
573: %
574: \begin{equation}
575: T_{\rm e}\sim 4\times \frac{1}{2}m_{\rm e}(2v_{\rm sh})^2
576: = 0.41{\rm keV} \left(\frac{v_{\rm sh}}{0.01c}\right)^2, 
577: \end{equation}
578: %
579: where we assume that the compression ratio is 4. This has a very 
580: important implication for the electron heating process of the SNR shocks. 
581: The proton temperature in the downstream is 
582: $T_{\rm p}=3m_{\rm p}v_{\rm sh}^2/16$, hence the ratio of two temperatures 
583: is
584: %
585: \begin{equation}
586: T_{\rm e}/T_{\rm p} \sim \frac{128}{3}\frac{m_{\rm e}}{m_{\rm p}} \sim 0.023. 
587: \end{equation}
588: %
589: This value is close to the observed value as long as the shock velocity 
590: $v_{\rm sh}$ is larger than $1500{\rm km/s}$ \citep{ade08}. Namely, 
591: we expect that the overheating problem of electrons raised by 
592: \citet{car88} can be solved by the ion two-stream instability. 
593:  
594: 
595: In this paper, we prepare three ion beams as the initial condition. We 
596: have also performed simulations for other initial conditions such that 
597: there exist upstream electrons, upstream and reflected protons. Electrons 
598: and reflected protons have drift velocities to satisfy the vanishing 
599: current condition. This situation corresponds to the initial phase of the 
600: shock reformation phenomena. 
601: The results turn out to be basically the same as those in the results 
602: presented in this paper.
603: 
604: 
605: Our simulation does not include electromagnetic modes. In the linear 
606: stage of the Buneman and ion two-tream instabilities, the effects are 
607: negligible because the growth rates of  electromagnetic modes are 
608: smaller than those of electrostatic modes for $v_{\rm d} \sim 0.01c$. 
609: In the nonlinear stage, electromagnetic modes may be important. One of 
610: the reasons is that the electrostatic waves with a wave vector almost 
611: perpendicular to the drift direction may make currents because the 
612: charge fluctuation can not be shielded by electrons in this situation, 
613: where the fluctuation scale is smaller than the electron Debye length. 
614: Consequently the current may make the magnetic field. It is an interesting 
615: speculation that the magnetic field might be amplified more rapidly by 
616: the ion two-stream instability than the ion Weibel instability. The 
617: other reason is that anisotropic ion heating caused by the highly 
618: oblique ion two-stream instability excites the Weibel instability due 
619: to ion temperature anisotropy. At the end of our simulations, the ion 
620: temperature of $x$-direction is almost the same as initial one and the 
621: ratio of the ion temperature of $y$-direction to that of $x$-diretion 
622: $T_{\rm iy}/T_{\rm ix}$ is about 100. These two features may lead to 
623: magnetic field amplification and accompanying particle acceleration and 
624: heating in the shock foot region. We will make full-PIC simulations in 
625: future work to investigate these issues.
626:  
627: \section{Summary}
628: We performed linear analysis of two-dimensional electrostatic modes 
629: and electrostatic two-dimensional PIC simulations with the real proton 
630: electron mass ratio to investigate the time evolution of electrostatic 
631: waves in the foot region of collisionless shocks with a high Mach number. 
632: We consider only the foot region by properly modeling the effects of 
633: reflected and returning protons. Performing the linear analysis, we 
634: have shown that after electrons are heated by the Buneman instability, 
635: the fastest growing mode is not the ion acoustic instability but the 
636: highly oblique ion two-stream instability. The latter mode, which has 
637: not been noticed previously, is excited by the resonace between 
638: ion plasma oscillations of the two proton beams when the electron 
639: temperature is much higher than the ion temperature. The PIC simulation 
640: confirms that the excitaion of the ion two-stream instability occurs 
641: faster than the ion acoustic instability and that protons are heated 
642: preferentially to the perpendicular direction to the shock normal 
643: direction. As a result, electron heating basically stops at the stage of 
644: the Buneman instability and the expected electron temperature in 
645: supernova remnants is fully compatible with the observation, avoiding 
646: the overheating problem raised by \citet{car88}. 
647: 
648: \acknowledgments
649: We are grateful to T. Tsuribe, T. Umeda, R. Yamazaki, T. Kato, 
650: M. Hoshino, T. Terasawa, S. Matsukiyo for discussions and suggestions 
651: especially for providing useful guidance and information in doing 
652: numerical simulations and for the nature of the ion two-stream instability. 
653: This work is partly supported by Scientific Research Grants (F.T.: 18542390
654: and 20540231) by the Ministry of Education, Culture, Sports, Science and Technology 
655: of Japan. Y. O. is supported by a Grant-in-Aid for JSPS Research Fellowships for Young Scientists. Numerical computations were carried out on fpc-cluster system at Center for Computational Astrophysics, CfCA, of National Astronomical Observatory of Japan.
656: 
657: \begin{thebibliography}{}
658: 
659: \bibitem[Adelsberg et al.(2008)]{ade08} Adelsberg, M., Heng, K., McCray, R., \& Raymond, J. C. 2008, astro-ph/0803.2521 
660: 
661: 
662: \bibitem[Bludman et al(1960)Bludman, Watson, \& Rosenbluth]{blu60} 
663: Bludman, S. A., Watson, K. M., \& Rosenbluth, M. N. 1960, Phys. Fluids, 3, 747 
664: 
665: \bibitem[Buneman(1958)]{bun58} Buneman, O. 1958, \prl, 1, 8
666: 
667: \bibitem[Cargill \& Papadopoulos(1988)]{car88} 
668: Cargill, P. J., \& Papadopoulos, K. 1988, \apjl, 329, L29
669: 
670: \bibitem[Hoshino \& Shimada(2002)]{hos02} 
671: Hoshino, M \& Shimada, N. 2002, \apj, 572, 880
672: 
673: \bibitem[Koyama et al.(1995)]{koy95} 
674: Koyama, K., Petre, R., Gotthelf, E. V., Hwang, U., Matsuura, M., 
675: Ozaki, M., \& Holt, S. S. 1995, \nat, 378, 225
676: 
677: \bibitem[Lampe et al.(1974)]{lam74} 
678: Lampe, M., Haber, I., Orens, J. H., \& Boris, P. 1974, Phys. Fluids, 17, 428
679: 
680: \bibitem[Leroy(1983)]{ler83} Leroy, M. M. 1983, Phys. Fluids, 26, 2742
681: 
682: \bibitem[Ohira \& Takahara(2007)]{ohi07} 
683: Ohira, Y., \& Takahara, F. 2007, \apjl, 661, L171
684: 
685: \bibitem[Papadopoulos(1988)]{pap88} 
686: Papadopoulos, K. 1988, \apss, 144, 535
687: 
688: \bibitem[Shimada \& Hoshino(2000)]{shi00} 
689: Shimada, N., \& Hoshino, M. 2000, \apjl, 543, L67
690: 
691: \bibitem[Stage et al.(2006)]{sta06} 
692: Stage, M. D., Allen, F. E., Houck, J. C., \& Davis, J. E. Nature Phys., 2, 614
693: 
694: \bibitem[Watanabe(1991)]{wata91} 
695: Watanabe, T. 1991, National Institute for Fusion Science, Nagoya, 65, 556
696: 
697: \end{thebibliography}
698: 
699: \clearpage
700: 
701: 
702: \begin{figure}
703: \plottwo{f1a.eps}{f1b.eps}
704: \caption{The color contour plot of the growth rate of the Buneman instabilitiy. The left panel (a) is for $v_{\rm d}/v_{\rm th,e}=10, T_{\rm e} = T_{\rm p}$, while the right panel (b) is for the  cold limit. \label{fig1}}
705: \end{figure}
706: 
707: %\clearpage
708: 
709: \begin{figure}
710: \plottwo{f2a.eps}{f2b.eps} 
711: \caption{The color contour plot of the growth rate of the ion two-stream instability (the left panel (a)) and that of the ion acoustic instability, for $v_{\rm d}/v_{\rm th,e}=1, T_{\rm e} = 100T_{\rm p}$. Note the diffenence in the scale of the abscissa between (a) and (b).
712: \label{fig2}}
713: \end{figure}
714: 
715: %\clearpage
716: 
717: \begin{figure}
718: \plotone{f3.eps} \caption{The particle distribution in the initial state for the simulation. Solid and dashed curves represent electron and ion distribution functions of $v_x$, respectively. 
719: \label{fig3}}
720: \end{figure}
721: 
722: %\clearpage
723: 
724: \begin{figure}
725: \plotone{f4.eps} 
726: \caption{The development of the energy density of electric field. 
727: The bold and thin lines represent non-magnetized and magnetized cases, respectively. Solid and dashed curves represent $E_x$ and $E_y$, respectively. 
728: \label{fig4}}
729: \end{figure}
730: 
731: %\clearpage
732: 
733: \begin{figure}
734: \plotone{f5.eps} 
735: \caption{Time development of the temperatures. 
736: The bold and thin lines represent non-magnetized and magnetized cases, respectively. Solid and dashed curves represent electron and proton temperatures, respectively. Dotted curves represent electron proton temperature ratio. 
737: \label{fig5}}
738: \end{figure}
739: 
740: %\clearpage
741: 
742: \begin{figure}
743: \plotone{f6.eps} 
744: \caption{The contour of the electrostatic potential at $t=1740 \omega_{\rm pe}^{-1}$.The potential is normalized by $m_{\rm e}v_{\rm d}^2/2$. Note that the structure is filamentary so that the electric field is almost perpendicular to the shock normal direction.
745: \label{fig6}}
746: \end{figure}
747: 
748: %\clearpage
749: 
750: \begin{figure}
751: \plotone{f7.eps} 
752: \caption{Energy distribution function of electrons at the end of simulations. Bold and thin curves represent non-magnetized and magnetized cases, respectively. 
753: \label{fig7}}
754: \end{figure}
755: 
756: \end{document}
757: 
758: 
759: