1: %% 2007 December 11: final revision
2: %% 2007 October 12: final (?) GF version; appendix removed; typos fixed
3: %% Beginning of file 'march10.tex'
4: %%
5: %% Modified 2007 August 18
6: %%
7: %% This is a draft of paper using the
8: %% AASTeX v5.x LaTeX 2e macros.
9:
10: \documentclass[12pt,preprint]{aastex}
11:
12: \newcommand{\vdag}{(v)^\dagger}
13: \newcommand{\myemail}{skywalker@galaxy.far.far.away}
14:
15: \newcommand{\tot}{_{\rm tot}}
16: \newcommand{\keV}{\rm keV}
17:
18:
19: %\slugcomment{18 August 2007}
20:
21: \shorttitle{Beam gyrosynchrotron emission} \shortauthors{Altyntsev et al.}
22:
23: \begin{document}
24:
25: %% LaTeX will automatically break titles if they run longer than
26: %% one line. However, you may use \\ to force a line break if
27: %% you desire.
28:
29: \title{Broadband microwave burst produced by electron beams}
30:
31: %pulse on 10 March 2002 - gyrosynchrotron emission
32: %generated by electron flux with anisotropic velocity distribution?}
33:
34: %[today]
35:
36:
37: \author{A.T. Altyntsev\altaffilmark{1,2}, G.D. Fleishman\altaffilmark{3,4},
38: G.-L. Huang\altaffilmark{1}, and V.F. Melnikov\altaffilmark{5}
39: }
40:
41: \altaffiltext{1}{Purple Mountain Observatory, Nanjing 210008, China}
42: \altaffiltext{2}{ISZF, Institute of Solar Terrestrial Physics,
43: Irkutsk 664033, Russia} \altaffiltext{3}{New Jersey Institute of
44: Technology, Newark, NJ 07102} \altaffiltext{4}{Ioffe
45: Physico-Technical Institute, St. Petersburg 194021, Russia}
46: \altaffiltext{5}{Radiophysical Research
47: Institute, Nizhny Novgorod 603950, Russia}
48:
49: \date{} %\today}
50:
51:
52: \begin{abstract}
53: Theoretical and experimental study of fast electron beams attracts a
54: lot of attention in the astrophysics and laboratory. In the case of
55: solar flares the problem of reliable beam detection and diagnostics
56: is of exceptional importance. This paper explores the fact that the
57: electron beams moving oblique to the magnetic field or along the
58: field with some angular scatter around the beam propagation
59: direction can generate microwave continuum bursts via
60: gyrosynchrotron mechanism. The characteristics of the microwave
61: bursts produced by beams differ from those in case of isotropic or
62: loss-cone distributions, which suggests a new tool for quantitative
63: diagnostics of the beams in the solar corona. To demonstrate the
64: potentiality of this tool, we analyze here a radio burst occurred
65: during an impulsive flare 1B/M6.7 on 10 March 2001 (AR 9368,
66: N27W42). The burst is remarkably suitable for this goal because of
67: its very short duration, wide frequency band and unusual
68: polarization being in the ordinary wave mode in the optically thin
69: range of the spectrum. Based on detailed analysis of the spectral,
70: temporal, and spatial relationships, we obtained firm evidence that
71: the microwave continuum burst is produced by electron beams. For the
72: first time we developed and applied a new forward fitting algorithm
73: based on exact gyrosynchrotron formulae and employing both the total
74: power and polarization measurements to solve the inverse problem of
75: the beam diagnostics. We found that the burst is generated by a
76: oblique beam in a region of reasonably strong magnetic field ($\sim
77: 200-300$ G) and the burst is observed at a quasi-transverse viewing
78: angle. We found that the life time of the emitting electrons in the
79: radio source is relatively short, $\tau_l \approx 0.5$ s, consistent
80: with a single reflection of the electrons from a magnetic mirror at
81: the foot point with the stronger magnetic field. We discuss the
82: implications of these findings for the electron acceleration in
83: flares and for beam diagnostics.
84:
85:
86: \end{abstract}
87:
88: \keywords{acceleration of particles --- Sun: flares --- Sun: radio
89: radiation
90: --- Sun: X-rays}
91:
92:
93:
94: \section{Introduction}
95:
96: Electron beams are believed to represent one of important
97: elementary ingredients of the solar activity \citep[][and references
98: therein]{Aschw_2002}. They ionize and excite hydrogen atoms in the
99: chromosphere giving rise to optical H$_\alpha$ flares. They are
100: capable of producing nonthermal hard X-ray (HXR) and gamma-ray
101: radiation via the bremsstrahlung mechanism as well as of driving the
102: chromospheric evaporation. Then, they can drive various kinetic
103: instabilities in the corona giving rise to a variety of coherent
104: radio emission types widely observed throughout the entire radio
105: band \citep{Aschw_2005}.
106:
107: Nevertheless, there is apparent lack of the observational tools now
108: for quantitative diagnostics of the electron beams in the solar
109: atmosphere. Currently, H$_\alpha$, HXR, and gamma emissions, as well
110: as type III bursts in the radio range are considered to represent
111: signatures of electron beams. However, these processes do not offer
112: any reliable straightforward diagnostics of the angular
113: distributions of electron beams. Even though a valuable information
114: of the electron beam properties might in principle be derived from
115: the linear polarization of the H$_\alpha$ and H$_\beta$ lines in the
116: chromospheric flares \citep{Henoux_etal_2003}, the measurement of
117: linear polarization in these spectral lines is extremely difficult,
118: so the corresponding beam diagnostic has not been established yet
119: \citep{Bianda_etal_2005}. In case of HXR and gamma emission, no
120: method has been proposed yet to get the angular distribution of fast
121: electrons from the data. Moreover, many HXR bursts originate from
122: coronal rather than chromospheric sources
123: \citep{Veronig_Brown_2004}, although even in case of chromospheric
124: sources the HXR emission can originate from a precipitating fraction
125: of approximately isotropic electron distributions rather than from
126: beams. Finally, any diagnostics based on the type III bursts is
127: difficult because for a coherent process the dependence of the
128: output radiation on the electron beam properties is highly nonlinear
129: and difficult to disentangle \citep{Aschw_2002}. On the other hand,
130: the beam can be invisible via the type III emission if it propagates
131: in a dense plasma, where high collisional damping rate quenches the
132: beam instability and no coherent emission is generated. On top of
133: this, the observed fast drift of radio fine structures is not
134: necessarily related to the beam propagation: it can be provided by
135: dynamics of MHD and/or reconnection processes as well
136: \citep{Altyntsev_etal_2007}, while lower values of the drift rates
137: can be ascribed to emissions originating at thermal conduction
138: fronts \citep{Farnik_Karlicky_2007}. We can conclude that available
139: tools are currently insufficient for reliable detection and detailed
140: diagnostics of the beams in the solar corona.
141:
142: Curiously, one of the most promising methods of the beam study,
143: namely, analysis of the beam-produced microwave gyrosynchrotron
144: radiation remains entirely unexplored yet, although the synchrotron
145: radiation by nonthermal electrons has long ago been recognized to be
146: the main mechanism producing the microwave emission in solar flares
147: \citep{Ramaty_1969, Ramaty_Petrosian_1972, Benka_Holman1992, BBG,
148: Nindos_etal_2000, Kundu_etal_2001, Trottet_etal_2002, Bastian_2006,
149: Gary_2006, Nindos_2006}. Exact expressions for the gyrosynchrotron
150: emission and absorption coefficient \citep{Eidman1958,
151: Eidman1959,Ramaty_1969,Ramaty94} are cumbersome and difficult for
152: direct use. Thus, significant efforts have been made to find simple
153: analytical approximations for both the synchrotron emission in the
154: ultrarelativistic case
155: \citep{Ginzburg1951,KorchakTerletsky1952,Getmantsev1952,Ginzburg1953,Korchak,GinzburgSyrovatsky1964}
156: and the gyrosynchrotron emission generated by nonrelativistic and
157: moderately relativistic electrons \citep{Dulk82, Dulk_1985, Zhou99}.
158: Also, simplified numerically fast computation schemes were developed
159: \citep{Petrosian1981,Klein1987}. All these studies, however, assumed
160: isotropic electron distributions or only weakly anisotropic in some
161: cases. These approximations are evidently insufficient to describe
162: and analyze the beam-produced gyrosynchrotron emission, where the
163: pitch-angle anisotropy is expected to be strong.
164:
165:
166:
167: Recently, \cite{Fleish03} discovered significant effect of the
168: pitch-angle anisotropy on the gyrosynchrotron spectrum and
169: polarization. So far \citep[see for a review][]{Fl_Nobe_2006},
170: analysis of the microwave continuum bursts provided ample evidence
171: for the loss-cone particle distribution formed due to trapping of
172: the accelerated electrons in the coronal magnetic loops
173: \citep{Melnikovetal2002a,Melnikov2006,FlGaryNita2003,Fl_etal_2007}.
174: In contrast, here we present a different class of events, when there
175: is a beam-like anisotropy of the particle distribution.
176:
177: In case of the beam-like angular distributions of the fast
178: electrons, in particular, the high-frequency spectral index depends
179: noticeably on the anisotropy and the angle of view in addition to
180: the standard dependence on the energy distribution. The degree of
181: polarization can differ strongly from that in the isotropic case.
182: Remarkably, the sense of polarization can correspond to the
183: \emph{ordinary} wave mode (O-mode) in the \emph{optically thin}
184: range of the spectrum in contrast to X-mode polarization in the
185: isotropic case \citep{Fleish03}.
186:
187:
188:
189:
190:
191: Thus, the beam-like anisotropy when present must be properly taken
192: into account for correct modeling the solar microwave continuum
193: bursts, which is especially important to interpret the polarization
194: spectra. The goal of our study is to identify an example of the
195: microwave burst in which the presence of electron beams is likely
196: and then evaluate the properties of the pitch-angle distribution and
197: possibly other relevant parameters of the source from the forward
198: fitting of the gyrosynchrotron formulae to the observed radio data.
199:
200:
201:
202:
203: The guess criteria for the candidates to the beam-produced microwave
204: bursts are the presence of either short (of the order of second)
205: broadband pulses or type III like drifting bursts at lower
206: frequencies or both. Among many burst-candidates we eventually
207: selected the event of 10 March, 2001. This selection is based
208: on fortuitous combination of radio observations of this burst made
209: by different observatories as well as availability of other
210: important context observations for this event.
211:
212:
213: Below we describe the key observational characteristics of the
214: event, then suggest semi-quantitative interpretation of the data,
215: then describe a specially developed nonlinear chi-square
216: minimization fit and apply it simultaneously to total power and
217: polarization spectra. Eventually, the fitting yields the parameters
218: of the angular distribution of the radiating electrons, the viewing
219: angle of the emission, and the characteristic life time of the
220: electrons at the radio source.
221:
222:
223:
224:
225: \section{Instrumentation}
226:
227: The event was observed by a number of radio instruments.
228:
229: The \emph{Nobeyama Radio Polarimeters} (NoRP) \citep{Polar,
230: Nakajima85} measure the total power and circularly polarized
231: intensities (Stokes parameters I and V) at 1, 2, 3.75, 9.4, 17, 35,
232: and 80 (Stokes I only) GHz with time resolution as high as 0.1~s. We
233: applied corrections provided by the Nobeyama team for the
234: polarization data at 1 and 2 GHz and the intensity data at 80 GHz
235: available at the Nobeyama Observatory Internet archives.
236:
237: Chinese \emph{Solar broadband Radio Spectrometers}
238: \citep[SRS,][]{Fu04} measure total flux and polarized flux at
239: frequencies 5.2-7.6~GHz with 20~MHz spectral and 5~ms temporal
240: resolution (NAOC, Huairou station), and the total flux at
241: frequencies 4.5-7.5~GHz with 10~MHz spectral and 5~ms temporal
242: resolution (Purple Mountain Observatory, PMO).
243:
244: The \emph{Nobeyama Radioheliograph} (NoRH) \citep{Helio} obtains
245: images of the Sun at 17~GHz (Stokes I and V) and 34~GHz (Stokes I)
246: with 0.1~s temporal resolution. At the time of the burst the angular
247: resolution of the NoRH was 17~arc~sec at 17~GHz and 10~arc~sec at
248: 34~GHz.
249:
250: Fortuitously, the data from another big imaging \emph{Siberian Solar
251: Radio Telescope} (SSRT) \citep{Smolkov, Grechnev03} are available
252: for this burst. SSRT yields 1D brightness distributions at 5.7~GHz
253: with 14~ms temporal resolution. During observation of this burst,
254: the knife-edge beam of north-south linear array (SSRT/NS) was
255: directed at $43.53$~degree from the central solar meridian (angular
256: resolution of 24 arc sec). The east-west array (SSRT/EW) knife-edge
257: beam was directed at $-23.61$ degree angle from the central solar
258: meridian with the angular width of 16.5 arc sec.
259:
260: HXR observations made at \emph{Yohkoh} satellite by Hard X-ray
261: Telescope (HXT) at four energy bands (L-band, 14-23~keV; M1-band,
262: 23-33~keV; M2-band, 33-53~keV; H-band, 53-93~keV) and Wide-Band
263: Spectrometer (WBS) \citep{Yosh92} at the high-energy band
264: (80-600~keV) and low-energy band (20-80~keV) are available.
265:
266: Additionally, context data of the SOHO/MDI magnetic field and
267: SOHO/EIT are available for the flare.
268:
269:
270:
271: \section{Observations}
272:
273:
274:
275: Impulsive flare (10 March 2001, M6.7/1B) has occurred in active
276: region AR9368 (N27W42). NoRP recorded an intense microwave burst
277: at all frequency channels (Figure~\ref{norp}, left panel). The
278: light curves throughout all those frequencies except 1~GHz
279: display a very short (about 3 s duration) broadband peak at
280: 04:03:39.6 UT. The pulse magnitudes were exceptionally strong in
281: the range 9.4 -- 35 GHz ($> 1000$ sfu).
282:
283: Analysis of this flare at different wavelengths, viz., $H_\alpha$,
284: HXR, SXR and radio waves was published in a number of papers.
285: \citet{Dingetal03} and \citet{Ding03} have classified this flare as
286: a white-light flare. From a good time correlation of $H_\alpha$ and
287: Ca II 8542 \AA\ brightenings with the peak of the microwave radio
288: flux, they concluded that the response in optical emission is due to
289: chromosphere heating by an electron beam. \citet{Uddin04} and
290: \citet{Chandra06} studied evolution of the flare active region and
291: associated this flare with small positive polarity region emerging
292: near the following negative sunspot. They distinguished two bright
293: $H_\alpha$ kernels connected by a dense plasma loop with distance
294: of about $10^4$ km between the footpoints.
295:
296: \subsection{Temporal characteristics}
297:
298: The entire burst duration is rather short ($\sim 40$ s) at
299: frequencies above 3.75 GHz. The time profiles are remarkably
300: similar to each other and the duration of the prominent peak was
301: the same throughout the entire spectral range. The light curves at
302: 9.4--80~GHz peaks at the same time, while that at 3.75~GHz is
303: delayed by a fraction of second and that at 2~GHz is delayed even
304: stronger (by half of second). Interestingly that the decay time
305: becomes slightly larger at low frequencies.
306:
307:
308: The NoRP time profiles of the circularly polarized radiation (Stokes
309: V) are shown in Figure~\ref{norp} (right panel). Except 35~GHz, the
310: profiles of the polarized fluxes differ essentially from the
311: profiles in intensity. The degree of polarization varies from
312: $+10\%$ at 1~GHz down to $-10\%$ at 35~GHz. The main peak is clearly
313: distinguished at 3.75, 9.4, and 35~GHz. Note that across the NoRP
314: spectrum, the right-handed circular polarization gives way to the
315: left-handed one somewhere between 3.75~GHz and 9.4~GHz. Analysis of
316: the SRS data at the range 5.2-7.6 GHz shows that this polarization
317: reversal occurs around 6.5 GHz.
318:
319: In the dynamic spectrum (Figure~\ref{fig2}, top) the NoRP data are
320: complemented by the SRS profiles at 5.4~GHz and 7.4~GHz to fill
321: large gap between the NoRP receiving frequencies 3.75~GHz and
322: 9.4~GHz. The shape of the microwave spectrum does not change much
323: during the burst that is visualized by similarity of the contours of
324: equal intensity shown in the dynamic spectrum (Figure~\ref{fig2},
325: top). The time profiles in Figure~\ref{norp} show a number of pulses
326: besides the main peak. They are clearly seen in the time derivative
327: of the dynamic spectrum (Figure~\ref{fig2}, bottom).
328:
329: Time derivative of dynamic spectrum (Figure~\ref{fig2}, bottom)
330: shows wide band fine structures during the entire bursts. Bright
331: strips correspond to increase of the emission, while the dark ones
332: show emission decrease. The duration of the shortest strips is
333: comparable with the NoRP temporal resolution (0.1~s), and there
334: bandwidth covers the entire spectral range from 1 GHz to 80 GHz in
335: some cases. The strips are not noise or interference since they are
336: wider than several frequency channels including data from different
337: observatories. Note that most of the broadband pulses display no
338: frequency drifts, although any drift value within $\pm800$ GHz/s is
339: measurable with given time resolution (0.1 s) and spectral bandwidth
340: (80 GHz) of the pulses. The duration of decay phases (the dark
341: strips after the bright ones) does not depend on the frequency.
342:
343: At millisecond time scale there are pulses with \emph{reverse} and
344: \emph{normal} frequency drift observed in the dynamic spectrum in
345: interval within 04:03:37 -- 04:03:56 UT, some of them are shown in
346: Figure~\ref{fine}. The total bandwidth of the fine structures does
347: not exceed 0.5 GHz and the life time is about 50 ms. The
348: corresponding drift rates are within $10-15$ GHz/s in this event.
349: The characteristic frequency of the fine structures rises from
350: 4.5~GHz up to 7~GHz during this interval.
351:
352: Time profiles of the hard X-ray emission are remarkably similar to
353: the microwave profiles (Figure~\ref{hxtt}). The same profiles were
354: observed with the Wide-Band Spectrometer at the high-energy band
355: (80-600~keV). Note, that HXR emission is delayed relative to the
356: radio emission. The overall L signal delays relative to various
357: radio frequencies by about 1-2~s, although the delay between the
358: impulsive peaks occurring at about 04:03:40 UT is shorter than 1 s.
359: At higher energies the delays are shorter being a fraction of
360: second. In particular, the cross correlation between the H channel
361: and 35~GHz light curve yields the delay of 0.5~s with the
362: correlation coefficient equal to $0.8$.
363:
364:
365:
366:
367: \subsection{Spatial characteristics}
368:
369: The flare spatial structure in $H_\alpha$ emission was studied by
370: \citet{Uddin04} and \citet{Chandra06}. Before the flare, emerging
371: flux of N polarity had penetrated into the S polarity region and
372: triggered the flare. From the spatial correlation between different
373: waveband sources it was concluded that the flare had \emph{the
374: three-legged} structure, i.e., it may be considered to be one of the
375: typical configuration of loops as determined earlier by Hanaoka
376: (1996). The $H_\alpha$ flare began (around 04:01 UT) as two bright
377: kernels (K1 and K2), which rapidly transformed during the flare peak
378: into a bright region expanding in the southwest direction. A short
379: loop connecting magnetic regions of opposite polarity is seen in 195
380: \AA\ emission (Figure~\ref{Fig2_3}); its length is about $10^4$ km.
381: One more source (so called remote source, RS) appeared after
382: 04:03:51~UT at 17~GHz. The RS was located 150 arcsec southwest from
383: the main flare site (Figure~\ref{Fig2_3}) and had right handed
384: polarization; the presence of the RS is confirmed by a secondary
385: peak at 280 arc sec in the 5.7~GHz SSRT/NS scan at 04:04:21~UT
386: (Figure~\ref{1dssrt}, middle), which is unpolarized, however, at
387: this relatively low frequency.
388:
389: The footpoints of the loop are seen in the 17~GHz polarization
390: images during the late decay phase of the burst
391: (Figure~\ref{pol_mag}, right panel). The shown maps are the result
392: of averaging of the emission over more than a minute. The averaging
393: is made since the brightness temperature of right handed emission
394: was rather weak. In the main phase of the burst the bulk of
395: microwave emission is generated in the source K2 with the north
396: polarity of magnetic field, and it is left hand polarized. This
397: means that the microwave source K2 is polarized in the ordinary wave
398: mode. The magnetic field near the footpoints can be estimated from
399: the MDI magnetogram as $-170$~G (K1) and 340~G (K2).
400:
401: The spatial behavior of the HXT emission is studied by
402: \citet{Chandra06} in detail. In all energy bands only single source
403: was observed. The estimated size of the HXR source was $7.0 \times
404: 3.8$, $6.7 \times 3.3$, $6.5 \times 3.3$ and $6.2 \times 3.2$ arc
405: sec in L, M1, M2, and H energy bands respectively. According to
406: Figure ~6 in \cite{Chandra06} the HXR source is located near source
407: K1. Note that the situation when the brightness peaks in the hard
408: X-ray and the microwaves are located in the opposite ends of a
409: single loop is typical for asymmetric magnetic loops.
410:
411: The SSRT 1d images are shown in Figure~\ref{1dssrt}. The vertical
412: lines correspond to the integration paths crossing the centers of
413: the two-dimensional brightness distribution at 17 GHz; these
414: integration paths are shown by dash-dotted lines for both EW and NS
415: linear arrays in Figure~\ref{Fig2_3}. The microwave source position
416: did not change in the NS scans, while slightly shifts to the West in
417: the EW scans at the burst peak. The sources K1 and K2 are not
418: resolved in the SSRT scans. The size of the 5.7~GHz microwave source
419: %scans
420: is about 50~arc sec in the NS scans and 25~arc sec in the EW scans.
421: On the contrary the polarization distributions are radically
422: changing with time. At the burst peak the polarization sense changes
423: from the left handed to the right handed polarization and then back
424: in the NS scans. In the EW scans a two-polarity structure appeared
425: in the time interval around the burst peak.
426:
427:
428:
429: \subsection{Spectrum description}
430:
431: The microwave spectra are shown in Figure~\ref{gyro} at three
432: moments corresponding to the main peak and neighboring peaks. The
433: spectra are similar to each other. At low frequencies the spectra
434: increase with index $\gamma \simeq 1.9$. Spectral peak is at about
435: 17~GHz. At high frequencies the spectrum decreases rather quickly
436: with the high-frequency spectral index $\gamma \simeq -2.1$.
437:
438: The hard X-ray spectrum $I(E)$ was studied by \citet{Chandra06}
439: using the Yohkoh$\backslash$HXT data. At the peak time they obtained
440: $I(E)=F_oE_{ph}^{-\gamma}=2.02\times 10^5 E_{ph}^{-2.4}$~photons
441: $\cdot$ cm$^{-2} \cdot$ s$^{-1} \cdot$ keV$^{-1}$. Assuming the
442: electron spectrum in the form $F_{el}(E)=AE^{-\delta}$ electrons/s
443: we get the parameters of electron flux under the thick-target
444: assumption as follows, by making necessary corrections for the
445: equation given by \cite{Brown}:
446: \begin{equation}\label{brown}
447: A=5.2 \times 10^{33}\gamma^2(\gamma-1)^2B(\gamma-0.5,1.5)F_o=5.1
448: \times 10^{39}
449: \end{equation}
450: and $\delta=\gamma+1 =3.4$, where $E$ is expressed in keV.
451:
452:
453: \subsection{Summary of observations}
454:
455: Here we summarize main observational characteristics of the event
456: important for further analysis:
457:
458: \begin{enumerate}
459:
460: \item The microwave emission in the 10 March 2001 event consists of many
461: short broadband pulses.
462:
463: \item
464: Type III like features are observed around 5 GHz.
465:
466: \item
467: The microwave emission is O-mode polarized at 17 GHz.
468:
469: \item
470: The polarization at 5.7 GHz displays high variability and
471: corresponds to X-mode at the peak time.
472:
473: \item
474: HXR light curves are remarkably similar to the microwave light
475: curves.
476:
477: \item
478: HXR is delayed by a fraction of second compared with the microwave
479: emission.
480:
481: \end{enumerate}
482:
483: \section{Model}
484:
485: \subsection{General trends and model dependences}
486:
487: Our goal in the analysis of the microwave emission is to derive
488: important source parameters from forward fitting of the observed
489: radio spectra by the gyrosynchrotron formulae. However, as is widely
490: known \citep[e.g.,][]{Bastian_etal_2007} the gyrosynchrotron
491: emission depends on too many physical effects and parameters even if
492: angular distribution of fast electrons is isotropic. Expected
493: presence of the beam-like anisotropy of fast electrons adds a few
494: new free parameters, which complicates further the procedure of the
495: fitting. Therefore, before developing a specific forward fitting
496: model dealing with gyrosynchrotron emission produced by electron
497: beams, we critically evaluate the observed properties of the burst
498: to restrict or fix as many parameters as possible.
499:
500: First of all, we make use of the close similarity of the radio and
501: HXR light curves. This similarity suggests that no trapping effect
502: is important for this event and both HXR and microwave emissions are
503: produced by the same electron distribution. The peak injection rate
504: of the electrons above 10 keV derived from the HXR spectrum is
505: $J(>10~\keV) \simeq 8.5\cdot10^{36}$ electrons/sec. We adopt that
506: the total number of the emitting electrons in the radio source is
507: $N\tot = \tau_l J(>10~\keV)$, where $\tau_l$ is the characteristic
508: life time of the emitting electrons in the radio source. Since no
509: trapping is important, we adopt that $\tau_l$ is a single,
510: energy-independent, free parameter, which will be determined later
511: from the forward fitting model. It is clear, however, that $\tau_l$
512: must not exceed a few seconds, the typical duration of the single
513: pulses composing the burst. Then, regarding the energy dependence of
514: the electron distribution, we adopt the simplest assumption of a
515: single power-law over the momentum modulus with the spectral index
516: determined from the HXR spectrum (note, the energetic spectral index
517: of 3.4 corresponds to the index of 7.8 in the distribution over
518: momentum).
519:
520: Second, address the question what can cause the delay of the HXRs
521: relative to microwave emission. If the electron beam is injected
522: somewhere at the top of the loop towards the foot points, then
523: directly precipitating electrons will first produce the HXR and a
524: fraction of the electrons reflected back into the loop will later
525: produce the radio emission. Thus, the model involving directly
526: precipitating beam predicts opposite delay (HXR leads radio) to what
527: is actually observed. Note, that models with electron trapping also
528: predict a delay of radio emission relative to HXR emission \citep{Melnikov1994}.
529:
530: The only transport model allowing radio to lead HXR is a 'single
531: reflection' model, which is adopted below. Specifically, if a
532: particle beam with some angular scatter is injected at a asymmetric
533: magnetic trap towards a foot-point with stronger magnetic field,
534: most of the electrons can be reflected back to form a hollow beam,
535: produce gyrosynchrotron emission in the region of relatively strong
536: magnetic field, and then after corresponding travel time over the
537: loop reach the other foot point with weaker magnetic field to
538: penetrate deeper into the chromosphere and produce HXRs. In this
539: case the observed delay between radio and HXR originates naturally.
540: In addition, the presence of initially downward injected beam is
541: confirmed by the reverse drifting coherent subbursts (Figure
542: \ref{fine}) leading both microwave and HXR peaks by a fraction of
543: second.
544:
545: In our event the formation of a asymmetric loop is likely because
546: the photospheric magnetic field has the extremes of $-170$ G and
547: +340 G at the flare kernels K1 and K2 respectively. This means that
548: the magnetic field in the radio source should belong to the range
549: +170~G~$ < B <$ +340~G. Indeed, if the regions of weaker magnetic
550: field provided noticeable radio emission, then the trapping of the
551: particles between $-170$ G and 170 G loop layers were important,
552: which is not observed.
553:
554: Third, the information about the largest possible value of the
555: magnetic field at the source allows to make a firm conclusion about
556: possible role of the gyrosynchrotron self-absorption in the event.
557: This question is important because the low-frequency slope of the
558: spectrum ($\gamma= 1.9$) is consistent with that expected for the
559: optically thick gyrosynchtotron radiation \citep{Dulk_1985}.
560: However, with the given magnetic field range and the given electron
561: distribution it is impossible to ensure the spectral peak at about
562: 17 GHz (as observed) by the self-absorption effect unless the source
563: is extremely compact, with the linear scale less than $700$ km. In
564: this case, however, the number density of the fast electrons will
565: exceed $10^{12}$ cm$^{-3}$, which we believe is not realistic. Thus,
566: we adopted the typical angular scale of the source to be 6''
567: consistent with imaging observations of the radio source.
568:
569: Alternatively, a large value of the microwave spectral peak
570: frequency can be provided by Razin effect, which requires high
571: plasma density at the radio source. Indeed, the presence of high
572: plasma density at the source is likely because we observe type III
573: like drifting bursts around 5 GHz (Figure~\ref{fine}).
574: Accordingly, we adopt the background plasma density to be
575: \begin{equation}\label{density_fixed}
576: n_e = 3\cdot 10^{11} {\rm cm}^{-3}.
577: \end{equation}
578: This estimate agrees with the value of the emission measure
579: determined from the soft X-ray emission \citep{Uddin04}.
580:
581: In such a dense plasma the Razin-effect is strong for the entire
582: range of the magnetic field $170-340$ G. In conditions of strong
583: Razin effect the low-frequency slope of the gyrosynchrotron spectrum
584: in the uniform source is much steeper than one observed. We must,
585: therefore, conclude that the low-frequency slope of the spectrum is
586: eventually formed by the source inhomogeneity, i.e., by different
587: layers with the magnetic field ranging from 170 G to 340~G in the
588: region of kernel K2. In such a dense plasma the free-free absorption
589: is typically important throughout the microwave range
590: \citep{Bastian_etal_2007}. However, high plasma temperature of about
591: 3$\cdot10^7$K was determined for this event from the soft X-ray data
592: \citep{Uddin04}, therefore, the free-free optical depth is less than
593: unity at $f>7$ GHz. Thus, we will not take into account the
594: free-free absorption in our analysis.
595:
596:
597: We note that the spectral peak provided by the Razin effect
598: increases as the magnetic field decreases. This means, in
599: particular, that \emph{lower-frequency} emission should arise
600: \emph{lower} in the loop, in contrast to usual situation when
601: \emph{higher-frequency} sources are located lower in the corona. We
602: checked that the position of the brightness peak at 17~GHz is indeed
603: displaced southwest by about 5'' relative to the 34~GHz brightness
604: peak in agreement with the prediction made. Therefore, the
605: high-frequency emission should arise from the region of the lowest
606: possible magnetic field, thus, we adopt
607: \begin{equation}\label{B_fixed}
608: B = 180~G
609: \end{equation}
610: consistent with the requirement $B > 170$ G. Now, when most of the
611: source parameters are fixed based on straightforward use of various
612: observational indicators, we can turn to formulating the forward
613: fitting model.
614:
615:
616:
617:
618:
619: \subsection{Forward fitting scheme}
620:
621: Even though there are many individual measurements of the radio
622: emission produced by the considered event, we can make use of only a
623: minor fraction of them. Indeed, since the low-frequency part of the
624: spectrum is related to the inhomogeneity of the source, which cannot
625: be reliably constrained by the observations, we can only model the
626: high-frequency part of the radio spectrum by the uniform source
627: (which we refer to as 'high-frequency source').
628:
629: Specifically, we adopt that the high-frequency source is entirely
630: responsible for the emission at 80~GHz and 35~GHz and for a
631: significant fraction of the emission at 17~GHz. Therefore, we have
632: at best five different observational data points (Stokes I and V at
633: 17~GHz and 35~GHz and Stokes I at 80~GHz), which may allow for
634: finding four free parameters or less. It is clear, that the weights
635: of the measurements are different from each other: the highest
636: weight is given to measurements at 35~GHz (5$\%$ error in the
637: intensity, and 10$\%$ error in polarization) as they have a small
638: experimental error and should be well described by the uniform
639: source model. The 80~GHz intensity has lower weight (while it should
640: be described by the uniform model even better that the 35~GHz data,
641: the 80~GHz intensity is measured with 40$\%$ error). The
642: experimental error at 17~GHz is small, however, the effect of source
643: inhomogeneity becomes important. Thus, we adopt that the
644: high-frequency source provides 70$\%$ of the observed flux at
645: 17~GHz with error of 40$\%$; the error of 100$\%$ in the degree of
646: polarization at 17~GHz is adopted.
647:
648: In previous subsection we mentioned already that one of the free
649: parameters we determine from the fitting is the characteristic life
650: time $\tau_l$ of the emitting electrons at the radio source. Another
651: important parameter, which is not known from the observations, is
652: the viewing angle $\theta$ between the line of sight and the
653: direction of the magnetic field at the source, so the viewing angle
654: is the second free parameter in the forward fitting scheme. Thus, we
655: have to use a test function for the angular part of the electron
656: distribution, which depends on only one or two free parameters. Our
657: model involving one reflection of the electrons from the magnetic
658: mirror cannot be described by a function with a single free
659: parameter: it must include both the direction where the angular
660: distribution reaches the maximum, which differs from the direction
661: along the field lines after the reflection, and typical angular
662: scatter of the distribution. Thus, a test function with two free
663: parameters is necessary. As a simplest approximation we adopt a
664: normalized gaussian angular distribution over the cosine of the
665: pitch-angle with unknown mean and dispersion:
666: \begin{equation}
667: \label{beam_mu_gauss}
668: f_2(\mu) \propto
669: \exp\left(- {(\mu-\mu_0)^2 \over \Delta \mu^2}\right) .
670: \end{equation}
671:
672: Then, we apply a nonlinear code that adjusts model free parameters
673: to minimize the $\chi^2$ statistics using the downhill simplex
674: method \citep{Press_etal_1986}. The $\chi^2$ statistics is
675: calculated as
676: \begin{equation}
677: \chi^2 = %{1\over {N-n}}
678: \sum_{i=1}^N {{(S_i^{obs}-S_i^{mod})^2}\over{\sigma_i^2}},
679: \end{equation}
680: where $S_i^{obs}$ are the observational data of either Stokes I or V
681: for the selected three frequencies, $\sigma_i$ are defined by the
682: errors introduced above, $S_i^{mod}$ are the model values of the
683: intensity and polarization.
684:
685: Our current forward fitting scheme is different from the scheme
686: applied previously \citep{Bastian_etal_2007} in two instances.
687: First of all, minimizing the $\chi^2$ statistics we employ both
688: intensity and polarization data simultaneously in the same run,
689: which is the first example when the polarization data is used for
690: quantitative diagnostics of the fast electron distribution. And
691: then, our model function is the exact expression of the
692: gyrosynchrotron emission including the summation over the series of
693: the Bessel functions and their derivatives. Because the magnetic
694: field at the source is somewhat low, the contribution of large
695: harmonics is important at high frequencies, therefore, we had to use
696: the sum over 1600 terms of the series to describe the emission
697: correctly up to 80~GHz. This made our scheme computationally
698: expensive: the full run with one spectrum took about 50 hours of the
699: PC with 2.1 GHz processor.
700:
701: \subsection{Forward fitting results}
702:
703:
704: Given that the observed spectrum does not evolve much during the
705: burst, while the forward fitting scheme employing exact
706: gyrosynchrotron equations is very time consuming, we concentrated on
707: study of the emission at the peak time of the burst (04:03:40~UT)
708: only. Result of the fitting is shown in Figure~\ref{gyro}. A number
709: of important things should be noticed in the figure. First, the
710: model radio spectra (top left panel) obtained for the electron
711: energy spectrum derived from HXR data are good match for the
712: observed high-frequency part of the radio spectra. Thus, the data is
713: consistent with the model assumption of a single power-law electron
714: spectrum in the event. Second, the curves for the degree of
715: polarization are very sensitive to the details of the electron
716: angular distribution. In particular, the polarization data is
717: entirely inconsistent with isotropic angular distribution of the
718: fast electrons, since the isotropic distribution would produce
719: X-mode polarized emission, while the observed polarization
720: corresponds to O-mode emission. The gyrosynchrotron radiation
721: produced by oblique beam observed by quasitransverse direction is
722: O-mode polarized as needed. The exact value of the degree of
723: polarization depends strongly on the details of the pitch-angle
724: distribution, thus, the joint use of the intensity and polarization
725: measurements is indeed a key to constrain the angular distribution
726: of the electron beam. For the peak time of the burst the following
727: parameters provide the best fit to the observed spectrum and
728: polarization:
729: \begin{equation}\label{fit_res}
730: \tau_l=0.45~s, \qquad \theta=80^o, \qquad \mu_0=0.5 , \qquad
731: \Delta\mu=0.35.
732: \end{equation}
733:
734:
735: All these numbers look reasonable against observations and theory of
736: gyrosynchrotron radiation from anisotropic electron distribution.
737: Indeed, the life-time of the fast electrons is small enough, $\tau_l
738: \approx 0.45$~s, i.e., less than the radio peak duration as
739: required. Then, as is known from the gyrosynchrotron theory, the
740: O-mode polarization of the optically thin source is only possible
741: for beam-like electron distributions and for viewing angles larger
742: than the peak angle in the pitch-angle distribution. The obtained
743: values of $\theta$ and $\mu_0$ obey these requirements since
744: $\cos\theta=0.16 < \mu_0$.
745:
746: It is tempting now to subtract the model contribution of the
747: high-frequency source from the observational data points and repeat
748: the forward fitting procedure. Indeed, we can perform a reasonable
749: fit to the total power data. As an example, a contribution of a
750: "lower-frequency source" with the same pitch-angle distribution but
751: with stronger magnetic field ($B \approx 300$ G), while smaller
752: life-time of the fast electrons ($\tau_l \approx 0.05$~s) is shown
753: in the figure.
754:
755: However, it is not possible to perform a consistent fit to the
756: polarization measurements, which is the key to constrain the
757: electron angular distribution, because the observed degree of
758: polarization is essentially the result of averaging of various
759: contributions along the non-uniform source. This conclusion is in
760: agreement with high spatial and temporal variability of the
761: polarization patterns observed at 5.7~GHz, which is most probably a
762: result of changing relative contributions from different parts of an
763: inhomogeneous source (note a very strong frequency dependence of the
764: degree of polarization in the model curves below 10 GHz in
765: Figure~\ref{gyro}). Therefore, the low-frequency observations cannot
766: be conclusively fitted by the uniform source model.
767:
768:
769: \section{Discussion}
770:
771: In this paper we presented a new tool of studying electron beams
772: accelerated during solar flares by analysis of the gyrosynchrotron
773: emission produced by the beams. Methodologically, this result is
774: achieved by quantitative use of the polarization measurements of the
775: microwave gyrosynchrotron emission. Specifically, to obtain the
776: information of the beam-like angular distribution of the
777: accelerated electrons we developed a nonlinear $\chi^2$ fit
778: employing Stokes I $\&$ V measurements simultaneously and exact
779: gyrosynchrotron formulae.
780:
781: This approach allows for unambiguous detection of oblique beams at
782: the radio source. In addition, our scheme yields a number of
783: important physical parameters of the radio burst such as the viewing
784: angle of the radio emission relative to the magnetic field at the
785: source, characteristic parameters of the electron distribution over
786: pitch-angle, and typical life time of the electrons in the radio
787: source, see Eq. (\ref{fit_res}).
788:
789: The diagnostics of the electron beam obtained from the fit of the
790: gyrosynchrotron data within 'one-reflection model' is consistent
791: with all other available observations. In particular, the time
792: delays of the hard X-rays relative to the microwaves (a fraction of
793: second) is consistent with the transit time of the electrons with
794: relatively large pitch-angles (found from the forward fitting
795: technique) through the flaring loop of about 10$^4$ km in
796: projection. In addition, the fine structures consisted of the
797: reverse drift bursts followed by the normal drift bursts detected at
798: about 5~GHz are in agreement with the 'one-reflection model', which
799: implies the downward beam propagation followed by its reflection at
800: the magnetic mirror and consequent upward motion of the beam. The
801: radio data suggest that the main radio source is filled by a rather
802: dense plasma, Eq. (\ref{density_fixed}), with the plasma frequency
803: of about 5 GHz. The radio emission observed at lower frequencies
804: (3.75 and 2 GHz) cannot be produced at this main source. The most
805: straightforward interpretation for this low-frequency component is
806: to postulate an adjacent more tenuous loop, where a minor fraction
807: of the accelerated electrons produce lower-frequency radiation. Such
808: a larger loop is likely and confirmed by the presence of a remote
809: source magnetically connected with the main flare site. If so,
810: larger decay constant and the observed delay of the low-frequency
811: emission relative to higher-frequency emission (see \S 3.1) receives
812: a natural interpretation because it simply relates to larger source
813: size implying longer transit time, where, in addition, some trapping
814: of the electrons can be important.
815:
816:
817: Another possible effect of the dense plasma is an enhanced role of
818: the Coulomb collisions on the radiating electrons. Given relatively
819: low value of the magnetic field at the source, it is easy to
820: estimate the typical energies of the radio emitting electrons as
821: $1-10$ MeV, which have the Coulomb energy decay time $\tau_c > 15$
822: s, which is much longer than the transit time. The time of
823: isotropization of these relativistic electrons due to Coulomb
824: collisions is even longer than the energy decay time
825: \citep{Petrosian_1985}. We, thus, conclude that the Coulomb
826: collisions have very little effect on the electron distribution at
827: the time scale of the electron precipitation ($\sim 1$ s) in this
828: event.
829:
830: Although we quantitatively describe the emission at the burst peak
831: only, we can reliably extrapolate the main findings, such as
832: acceleration of the fast electrons in the form of beams and the
833: one-reflection transport model to the entire burst duration. It
834: follows from the close similarity between the radio and the hard
835: X-ray light curves and from the constancy of the sense of
836: polarization in the main source at 17 GHz. This means that some of
837: the flares, such as one considered here, can predominantly
838: accelerate electrons along the magnetic field lines even though the
839: pitch-angle distribution of the beams has some angular scatter.
840:
841: The potentiality of the developed method is very strong and will be
842: especially helpful when imaging spectroscopy data is availably.
843: However, the routine use of this method requires optimization of the
844: computation scheme, which is not fast enough at present. Therefore,
845: deducing simplified, although precise enough, gyrosynchrotron
846: formulae for anisotropic electron distributions can be very helpful
847: here.
848:
849:
850:
851:
852: \section{Conclusion}
853:
854: Electron beams with some angular scatter can efficiently produce
855: microwave continuum bursts via gyrosynchrotron mechanism. As was
856: shown theoretically by \cite{Fleish03}, the gyrosynchrotron emission
857: produced by electron beams can be distinguished from that produced
858: by isotropic electron populations by analysis of the degree of
859: polarization of the microwave emission. Here we presented a
860: compelling example of such event, where the optically thin
861: gyrosynchrotron radiation is indeed O-mode polarized as expected for
862: the beam-like distributions. Remarkably, the presence of the beam
863: is confirmed by the whole set of the available data for this event.
864:
865: In case of the radio data with good enough quality (including both
866: intensity and polarization) the use of forward fitting inversion
867: methods allows for quantitative diagnostics of the fast electron
868: angular distribution as well as a number of other important physical
869: parameters of the flaring source. These methods will become much
870: more useful when the imaging spectroscopy data is available.
871:
872: \acknowledgments The authors express appreciation to Prof. Kiyoto
873: Shibasaki and Prof. Hiroshi Nakajima for the NoRP calibration
874: corrections they provided. This work was supported in part by NSF
875: grants ATM-0607544 and ATM-0707319 to New Jersey Institute of
876: Technology, by the Russian Foundation for Basic Research, grants
877: No. 06-02-16295, 06-02-16859, 06-02-39029, 07-02-01066, NFSC project
878: No. 10333030 and ‘‘973’’ program with No. 2006CB806302. We have made
879: use of NASA's Astrophysics Data System Abstract Service.
880:
881:
882: \begin{thebibliography}{}
883:
884: \bibitem[Altyntsev et al.(2007)]{Altyntsev_etal_2007}
885: Altyntsev, A. T., Grechnev, V. V., Meshalkina, N. S., \& Yan, Y.
886: 2007, \solphys, 242, 111
887:
888: \bibitem[Aschwanden (2002)]{Aschw_2002}
889: Aschwanden, M.~J. 2002, Particle Acceleration and Kinematics in Solar
890: Flares (Dordrecht: Kluwer)
891:
892: \bibitem[Aschwanden (2005)]{Aschw_2005}
893: Aschwanden, M.~J. 2005, Physics of the Solar Corona. An Introduction
894: with Problems and Solutions (New York: Springer)
895:
896:
897: \bibitem[Bastian (2006)]{Bastian_2006}
898: Bastian, T. B. 'Solar Physics with the Nobeyama Radioheliograph',
899: Proceedings of Nobeyama Symposium 2004, held in Kiyosato, Japan, October 26-29, 2004,
900: Nobeyama Solar Radio Observatory, 2006, pp.1-8
901:
902: \bibitem[Bastian, Benz \& Gary(1998)]{BBG}
903: Bastian, T. S., Benz, A. O., \& Gary, D. E. 1998,
904: \araa, 36, 131
905:
906: \bibitem[Bastian et al.(2007)]{Bastian_etal_2007}
907: Bastian, T. S., Fleishman, G. D., \& Gary, D. E. 2007, \apj,
908: 666, 1256
909:
910: \bibitem[Benka \& Holman(1992)]{Benka_Holman1992} Benka, S.~G., \&
911: Holman, G.~D.\ 1992, \apj, 391, 854
912:
913: \bibitem[Bianda et al.(2005)]{Bianda_etal_2005}
914: Bianda, M., Benz, A.~O., Stenflo, J.~O., Kuveler, G.,\& Ramelli, R.\
915: 2005, \aap, 434, 1183
916:
917: \bibitem[Brown(1971)]{Brown}
918: Brown, J. C. 1971, \solphys, 18, 489
919:
920: \bibitem[Chandra et al.(2006)]{Chandra06}
921: Chandra, R., Jain, R., Uddin, W., Shimura, K., Kosugi, T., Sakao,
922: N., Joshi, A., \& Despandey, M. R. 2006, \solphys, 239, 239
923: %arXiv:astro-ph/0602611 v1 28 Feb 2006
924:
925: \bibitem[Ding (2003)]{Ding03} Ding, M. D. 2003,
926: Journal of the Korean Astronomical Society, 36, 49
927:
928: \bibitem[Ding et al.(2003)]{Dingetal03} Ding, M. D., Liu, Y., Yeh, C.-T. \& Li, P. I. 2003,
929: \aap, 403, 1151
930:
931: \bibitem[Dulk (1985)]{Dulk_1985}
932: Dulk, G. A. 1985, \araa, 23, 169
933:
934: \bibitem[Dulk \& Mursh (1982)]{Dulk82} Dulk, G. A. \& Mursh, K. A. 1982,
935: \apj, 259, 350
936:
937: \bibitem[Eidman (1958)]{Eidman1958}
938: Eidman, V. Ya. 1958, Sov. Phys. JETP, 7, 91
939: \bibitem[Eidman (1959)]{Eidman1959}
940: Eidman, V. Ya. 1959, Sov. Phys. JETP, 9, 947
941:
942: \bibitem[Farnik \& Karlicky (2007)]{Farnik_Karlicky_2007}
943: Farnik \& Karlicky, 2007, \solphys, 240, 121
944:
945:
946: \bibitem[Fleishman (2006)]{Fl_Nobe_2006}
947: Fleishman, G. D. 'Solar Physics with the Nobeyama Radioheliograph',
948: Proceedings of Nobeyama Symposium 2004, held in Kiyosato, Japan, October 26-29, 2004,
949: Nobeyama Solar Radio Observatory, 2006, pp.51-62
950:
951:
952: \bibitem[Fleishman et al. (2007)]{Fl_etal_2007}
953: Fleishman, G. D., Bastin, T. S., \& Gary, D. E. 2007, \apj,
954: to be submitted
955:
956: \bibitem[Fleishman et al. (2003)]{FlGaryNita2003}
957: Fleishman, G. D., Gary, D. E., \& Nita, G. M. 2003, \apj, 593,
958: 571
959:
960: \bibitem[Fleishman and Melnikov(2003)]{Fleish03} Fleishman G. D. \& Melnikov V. F. 2003,
961: \apj, 587, 823
962:
963: \bibitem[Fu et al.(2004)]{Fu04}
964: Fu, Q., Ji, H., Qin, Z., Xu, Z., Xia, Z., Wu, H., Liu, Y., Yan,
965: Y., Huang, G., Chen, Z., Jin, Z., Yao, Q., Cheng, C., Xu, F.,
966: Wang, M., Pei, L., Chen, S., Yang, G., Tan, C., Shi, S. 2004,
967: \solphys, 222, 167
968:
969: \bibitem[Gary (2006)]{Gary_2006}
970: Gary, D. E. 'Solar Physics with the Nobeyama Radioheliograph',
971: Proceedings of Nobeyama Symposium 2004, held in Kiyosato, Japan, October 26-29, 2004,
972: Nobeyama Solar Radio Observatory, 2006, pp.119-129
973:
974: \bibitem[Getmantsev (1952)]{Getmantsev1952}
975: Getmantsev, G. G. 1952, DAN SSSR, 83, 557
976: \bibitem[Ginzburg (1951)]{Ginzburg1951}
977: Ginzburg, V. L. 1951, DAN SSSR, 76, 377
978: \bibitem[Ginzburg (1953)]{Ginzburg1953}
979: Ginzburg, V. L. 1953, Uspekhi Fiz. Nauk, 51, 343
980: \bibitem[Ginzburg and Syrovatsky (1964)]{GinzburgSyrovatsky1964}
981: Ginzburg, V. L., \& Syrovatsky, S. I. 1964, The Origin of Cosmic Rays
982: (New York: Macmillan Co.)
983:
984: \bibitem[Grechnev et al.(2003)]{Grechnev03}
985: Grechnev, V. V. Lesovoi, S. V., Smolkov, G. Ya., Krissinel, B. B.,
986: Zandanov, V. G., Altyntsev, A. T., Kardapolova, N. N., Sergeev, R. Y.,
987: Uralov, A. M., Maksimov, V. P., \& Lubyshev, B. I. 2003, \solphys, 216, 239
988:
989: \bibitem[Henoux et al.(2003)]{Henoux_etal_2003}
990: Henoux, J. C., Vogt, E., Sahal-Brechot, S., Karlichky, M., Feautrier, N., Fa'rni'k, F., Chambe, G., Balanca,
991: C. 2003, ASP Conf. Proc., 307, 480
992:
993:
994: \bibitem[Klein (1987)]{Klein1987}
995: Klein, K.-L. 1987, \aap, 183, 341
996: \bibitem[Korchak and Terletsky (1952)]{KorchakTerletsky1952}
997: Korchak, A. A. and Terletsky, Ya. P. 1952, Zh. Eks. Teor. Phys., 22, 507
998: \bibitem[Korchak (1957)]{Korchak}
999: Korchak, A. A. 1957, Soviet Astronomy, 1, 360
1000:
1001:
1002:
1003:
1004: \bibitem[Kundu et al. (2001)]{Kundu_etal_2001}
1005: Kundu, M. R., White, S. M., Shibasaki, K., Sakurai, T., \& Grechnev,
1006: V. V. 2001, \apj, 547, 1090
1007:
1008: \bibitem[Lui, Ding\&Fang (2001)]{Lui01}
1009: Liu, Y., Ding, M. D. and Fang, C. 2001, \apjl, 563, L169
1010:
1011: \bibitem[Melnikov (1994)]{Melnikov1994}
1012: Melnikov, V. F. 1994, Radiophysics and Quantum Electronics, 37, 557
1013:
1014: \bibitem[Melnikov et al. (2002a)]{Melnikovetal2002a}
1015: Melnikov, V. F., Shibasaki, K., \& Reznikova, V. E. 2002a, \apjl, 580,
1016: L185
1017:
1018: \bibitem[Melnikov (2006)]{Melnikov2006}
1019: Melnikov, V. F. 'Solar Physics with the Nobeyama Radioheliograph',
1020: Proceedings of Nobeyama Symposium 2004, held in Kiyosato, Japan, October 26-29, 2004,
1021: Nobeyama Solar Radio Observatory, 2006, pp.9-20
1022:
1023: \bibitem[Nakajima et al.(1985)]{Nakajima85}
1024: Nakajima, H., Sekiguchi, H., Sawa, M., Kai, K., \& Kawashima, S.
1025: 1985, PASJ, 37, 163
1026:
1027: \bibitem[Nakajima et al.(1994)]{Helio}
1028: Nakajima, H., Nishio, M., Enome, S., Shibasaki, K., Takano, T., Hanaoka, Y.,
1029: Torii, C., Sekiguchi, H., Bushimata, T., Kawashima, S., Shinohara, N.,
1030: Irimajiri, Y., Koshiishi, H., Kosugi, T., Shiomi, Y., Sawa, M., \& Kai, K.
1031: 1994, Proc. of the IEEE, 82, 705
1032:
1033: \bibitem[Nindos (2006)]{Nindos_2006}
1034: Nindos, A. 'Solar Physics with the Nobeyama Radioheliograph',
1035: Proceedings of Nobeyama Symposium 2004, held in Kiyosato, Japan, October 26-29, 2004,
1036: Nobeyama Solar Radio Observatory, 2006, pp.37-48
1037:
1038:
1039: \bibitem[Nindos et al. (2000)]{Nindos_etal_2000}
1040: Nindos, A., White, S. M., Kundu, M. R., \& Gary, D. E. 2000,
1041: \apj, 533, 1053
1042:
1043: \bibitem[Petrosian (1981)]{Petrosian1981}
1044: Petrosian, V. 1981, \apj, 251, 727
1045:
1046: \bibitem[Petrosian (1985)]{Petrosian_1985}
1047: Petrosian, V. 1985, \apj, 299, 987
1048:
1049: \bibitem[Press et al.(1986)]{Press_etal_1986} Press, W.~H., Flannery,
1050: B.~P., \& Teukolsky, S.~A.\ 1986, Numerical Recipes: The Art of
1051: Scientific Computing (Cambridge: Univ. Press) %, 1986
1052:
1053: \bibitem[Ramaty(1969)]{Ramaty_1969}
1054: Ramaty, R. 1969, \apj, 158, 753
1055: \bibitem[Ramaty \& Petrosian(1972)]{Ramaty_Petrosian_1972} Ramaty, R., \&
1056: Petrosian, V.\ 1972, \apj, 178, 241
1057:
1058: \bibitem[Ramaty et al.(1994)]{Ramaty94} Ramaty, R., Schwartz,
1059: R. A., Enome, S., \& Nakajima, H. 1994, \apj, 436, 941
1060:
1061:
1062: \bibitem[Smolkov et al.(1986)]{Smolkov}
1063: Smolkov, G. Ia., Pistolkors, A. A., Treskov, T. A., Krissinel, B.
1064: B., \& Putilov, V. A. 1986, \apss, 119, 1
1065:
1066: \bibitem[Torii et al. (1979)]{Polar}
1067: Torii, C., Tsukiji, Y., Kobayashi, S., Yoshimi, N., Tanaka, H., \& Enome, S.
1068: 1979, Proc. of the Res. Ist. of Atmospherics, Nagoya Univ., 26, 129
1069:
1070: \bibitem[Trottet et al. (2002)]{Trottet_etal_2002}
1071: Trottet, G., Raulin, J.-P., Kaufmann, P., Siarkowski, M., Klein,
1072: K.-L., \& Gary, D. E. 2002, \aap, 381, 694
1073:
1074:
1075: \bibitem[Uddin et al.(2004)]{Uddin04} Uddin, W., Jain, R.,
1076: Yoshimura, K., Chandra, R., Sakao, T., Kosugi, T., Joshi, A., \&
1077: Despande, M. R. 2004, \solphys, 225, 325
1078:
1079: \bibitem[Veronig \& Brown(2004)]{Veronig_Brown_2004}
1080: Veronig, A. M., \& Brown, J. C. \ 2004, \apj, 603, L117
1081:
1082: \bibitem[Yoshimori et al.(1992)]{Yosh92} Yoshimori, M., Takai, Y., Morimoto, K., Suga, K.,
1083: Ohki, K., Watanabe, T., Yamagami, T., Kondo, I., \& Nishimura, J.
1084: 1992, PASJ, 44, L51
1085:
1086: \bibitem[Zhou, Huang \& Wang (1999)]{Zhou99}
1087: Zhou, A.-H., Huang, G.-L. \& Wang, X.-D. 1999, \solphys, 189, 345
1088:
1089: \end{thebibliography}
1090:
1091:
1092:
1093: \begin{figure}
1094: \epsscale{1.20} \plottwo{f1I.eps}{f1V.eps} \caption{Microwave
1095: fluxes, recorded by the NoRP polarimeters. Left: total intensity
1096: (R+L), right: polarized intensity (R-L). Magnitudes of fluxes are
1097: in sfu. \label{norp}}
1098: \end{figure}
1099:
1100: \begin{figure}
1101: \epsscale{1.0} \plotone{f2.eps} \caption{Top: Dynamic spectrum of
1102: the 10 March 2001 flare burst. The NoRP measurements are
1103: complemented by SRS measurements at 5.4 and 7.4~GHz. Contours: $2000
1104: \times (0.1,0.3,0.5,0.8)$ sfu. Bottom: The derivatives of the time
1105: profiles. \label{fig2}}
1106: \end{figure}
1107:
1108: \begin{figure}
1109: \epsscale{.70} \plotone{f3.eps} \caption{Radio vs Yohkoh/HXT light
1110: curves. Energy ranges are L(14 –-- 23~keV), M1(23 –-- 33~keV), M2(33
1111: --- 53~keV), H(53 –-- 93~keV). HXR light curves (dashed) are normalized such as to match
1112: the peak value of the corresponding radio light curves (solid).
1113: Available at
1114: $http://gedas22.stelab.nagoya-u.ac.jp/HXT/catalogue/image\_html/eid\_html/eid\_25110.html$
1115: \label{hxtt}}
1116: \end{figure}
1117:
1118: \begin{figure}
1119: \epsscale{.70} \plotone{f4.eps}%{fine.eps}
1120: \caption{Drifting fine structures recorded slightly before the main
1121: flare peak with the PMO spectrometer. \label{fine}}
1122: \end{figure}
1123:
1124:
1125:
1126:
1127:
1128: \begin{figure}
1129: \epsscale{1.0} \plottwo {f5_1.eps}{f5_2.eps} \caption{ Left: Flare
1130: sources in UV and microwaves (right and left circular polarization).
1131: Dash-dotted lines show directions of the EW and NS fan strips.
1132: Right: Extended EIT partial frame with MDI magnetogram background.
1133: Dashed contour shows the magnetic neutral line. \label{Fig2_3}}
1134: \end{figure}
1135:
1136: \begin{figure}
1137: \epsscale{0.7} \plotone{f6.eps} \caption{One-dimensional brightness
1138: distributions (scans) recorded with the EW and NS arrays at 5.7~GHz
1139: in intensity (left) and polarization (right). Each point in these
1140: brightness distributions is a result of integration of the true
1141: brightness along a line parallel to the dash-dotted EW (EW-scans) or
1142: NS (NS-scans) lines shown in Figure~\ref{Fig2_3}. Solid profiles
1143: correspond to the burst peak. Profile magnitudes are in arbitrary
1144: units. Vertical dash-dotted lines correspond to positions marked in
1145: the NoRH map in Figure~\ref{Fig2_3} (left). Middle panels show the
1146: appearance of the unpolarized RS in the NS scans. \label{1dssrt}}
1147: \end{figure}
1148:
1149: \begin{figure}
1150: \epsscale{0.9} \plotone{f7.eps} \caption{Brightness distributions in
1151: intensity and polarization at 17~GHz. Average values of contours are
1152: $4\times 10^7$~K in intensity, $1.5\times 10^4$~K in right handed
1153: polarization, and $-6\times 10^6$~K in left handed polarization in
1154: the left panel, and $ 2\times 10^6$~K in intensity, $ 1.35\times
1155: 10^4$~K in right handed polarization, and $-10^5$~K in the right
1156: panel. \label{pol_mag}}
1157: \end{figure}
1158:
1159:
1160: \begin{figure}
1161: \epsscale{1.0} \plotone{f8.eps} %{Gyro_fit_full.eps}
1162: \caption{Observed spectra at various time frames (symbols) and
1163: fitting results (curves). Left column presents the best fit curves
1164: for the total intensity and degree of polarization for the
1165: high-frequency source, green/thick lines as well as curves
1166: obtained by small variations of the dispersion in the fast
1167: electron angular distribution (blue and red curves). Black curve
1168: show the results for the isotropic electron distribution all other
1169: conditions being equal. Right column displays the best fit curves
1170: for the high-frequency source (green/solid) supplemented by curves
1171: for the emission from a secondary source with higher magnetic
1172: field and smaller amount of the fast electrons (blue/dash-dotted
1173: curves) all other conditions being equal. This simple two-source
1174: model is good enough to describe the total intensity, although
1175: insufficient to fit the polarization data. \label{gyro}}
1176: \end{figure}
1177:
1178:
1179: \end{document}
1180: