1: %$Id: ms.tex,v 1.5 2005/04/01 10:28:03 pgrigis Exp $
2: %
3: % Paolo Grigis
4: % submitted to ApJ Letters
5: %
6: %
7:
8:
9: %\documentclass[preprint2]{aastex}
10:
11: %\documentclass[12pt,preprint]{aastex}
12:
13: %% manuscript produces a one-column, double-spaced document:
14:
15: %\documentclass[manuscript]{aastex}
16:
17: %% preprint2 produces a double-column, single-spaced document:
18:
19: %\documentclass[preprint2]{aastex}
20:
21: \documentclass{emulateapj}
22:
23:
24: \begin{document}
25:
26:
27: \title{Evolution of reconnection along an arcade of magnetic loops}
28:
29: \author{Paolo C. Grigis and Arnold O. Benz}
30: \affil{Institute of Astronomy, ETH Z\"urich, 8092 Z\"urich, Switzerland}
31: \email{pgrigis@astro.phys.ethz.ch}
32:
33: \begin{abstract}
34:
35: RHESSI observations of a solar flare showing continuous motions of double
36: hard X-ray sources interpreted as footpoints of magnetic loops are presented.
37: The temporal evolution shows many distinct emission peaks of duration of some
38: tens of seconds ('elementary flare bursts').
39: Elementary flare bursts have been interpreted as instabilities or oscillations
40: of the reconnection process leading to an unsteady release of magnetic energy.
41: These interpretations based on two-dimensional concepts cannot explain these
42: observations, showing that the flare elements are displaced in a third
43: dimension along the arcade. Therefore, the observed flare elements are not
44: a modulation of the reconnection process, but originate as this process
45: progresses along an arcade of magnetic loops. Contrary to previous reports,
46: we find no correlation between footpoint motion and hard X-ray flux.
47: This flare apparently contradicts the predictions of the standard
48: translation invariant 2.5D reconnection models.
49: \end{abstract}
50:
51: \keywords{Sun: flares --- Sun: X-rays, gamma rays}
52:
53: \section{Introduction}
54: High-energetic electrons accelerated during solar flares emit
55: bremsstrahlung hard X-rays (HXR), whose evolution in time can
56: be followed in images, light curves, and spectra. This information is combined
57: here to infer characteristics of the unknown acceleration process of these particles.
58:
59:
60: As for light curves, quasi-periodic modulations of the HXR flux on
61: time scales of some tens of seconds (\emph{Elementary Flare Bursts}; EFB)
62: have long been known to observers \citep{Parks69,deJager78}.
63: Spectral studies show that EFBs preferentially follow a characteristic
64: \emph{soft-hard-soft} spectral behavior
65: \citep[][and references therein]{Grigis04}.
66: The modulations of the HXR and microwave flux have been interpreted
67: in 2D models of reconnection as fluctuations in the reconnection process due
68: to global oscillations of the loop \citep{Roberts84,Nakariakov03,Stepanov04}.
69:
70:
71: On the imaging front, early observational evidence from the
72: \emph{Solar Maximum Mission} Hard X-Ray Imaging Spectrometer
73: \citep{Hoyng81} showed that HXR sources are found at the location of
74: footpoints (FP) of magnetic loops.
75: Different types of HXR sources located higher in the solar
76: atmosphere were detected on some occasions:
77: the \emph{Yohkoh} Hard X-ray Telescope registered
78: fainter \emph{loop-top} sources above the soft X-ray loops
79: \citep{Masuda94} and the \emph{Reuven Ramaty High Energy Solar Spectrometric
80: Imager} \citep[RHESSI; ][]{Lin02} observed \emph{coronal} sources
81: \citep{Lin03}, as well as emission from collisionally thick coronal loops
82: \citep{Veronig04}. However, the bulk of non-thermal HXRs comes from loop FPs.
83:
84:
85: Two FPs on opposite sides of a magnetic neutral line are expected in the
86: standard model of eruptive flares
87: \citep[reviewed e.g. by][]{Priest02}: The rapid eruption of a
88: filament enables the magnetic field to reconnect, driving particle
89: acceleration in lower loops. Electrons precipitating
90: to a FP emit HXRs. In this scenario, one expects the observed
91: FP sources to drift apart as successive field lines
92: are reconnected at higher altitudes. This explanation fits the long-known outward
93: motion of $\mathrm{H}\alpha$ ribbons parallel to the neutral line.
94:
95:
96: Reports on both the morphology and the time evolution of the FPs
97: show a large range of behaviors: single \citep{Takakura83},
98: double \citep{Hoyng81} and multiple sources are seen, and many
99: authors observe FP motions of different kinds: decrease and increase
100: in the FP separation across the neutral line; parallel and antiparallel
101: movements along the arcade
102: \citep[for some recent observations see][]{Fletcher02,Liu04,Qiu04,Siarkowski04}.
103: This bewildering behavior demonstrates just how complex the flare phenomenon can be.
104:
105:
106: \cite{Krucker03} presented high-resolution observations of a particularly interesting
107: two-ribbon flare. One of the HXR FPs moved continuously along a ribbon, whereas the
108: other two FPs showed no systematic motion. No motion perpendicular to the ribbons are
109: noticeable, but the parallel motion correlated with the HXR flux. The observed behavior
110: allowed Krucker et al. to interpret the observations still in terms of the standard
111: reconnection model, where the motion is due to receding FPs. This requires a strongly sheared
112: arcade and a not-specified complex magnetic structure including
113: the 2 other FPs without systematic motion.
114:
115:
116: Here HXR source motions observed with RHESSI during
117: a flare are reported that do not allow for such interpretation by
118: the standard 2D reconnection model.
119: We study also the relation between the spatial motion and the
120: spectral evolution of EFBs in time.
121:
122:
123:
124:
125: \section{Observations}
126:
127:
128: RHESSI observed the Sun on November 9, 2002 from 12:23 to
129: 13:28 UT, when it entered the shadow of the Earth,
130: and it registered the HXR evolution of a solar flare
131: of soft X-ray (GOES) importance M4.9. RHESSI was in a configuration
132: well suited to the derivation of high-resolution HXR spectra and images:
133: No decimation of the data was active during the flare and
134: attenuation \citep{Smith02} was constantly in state 1
135: (thick attenuator in), thus ensuring that the detector dead time
136: was below about 5\% during the flare. Auxiliary data suggest that
137: this flare was an eruptive event, displaying a post-flare loop arcade
138: in the SOHO/EIT 195 \AA\ images (Fig.~\ref{fig1}), a moving Type IV radio burst
139: (Phoenix spectrometer), and an associated fast CME
140: \citep[listed in the catalog by][]{Yashiro04}.
141:
142:
143:
144: \begin{figure}
145: %\epsscale{0.65}
146: \plotone{f1.eps}
147: \caption{\label{fig1}
148: \emph{Top:} SOHO/EIT 195 {\AA} image of post-flare loops
149: with the RHESSI HXR source positions superimposed. The positions of the 20 - 50 keV
150: sources from the CLEAN images are represented by crosses with arm lengths
151: proportional to the errors, positions from the PIXON images are given by
152: circles. Simultaneous footpoints are connected and color coded according
153: to the time intervals defined in the bottom part. The neutral line is
154: shown in gray.
155: \emph{Bottom:} Time evolution of the flux and spectral index.}
156: \end{figure}
157:
158:
159:
160:
161: \begin{figure}
162: %\epsscale{0.5}
163: \plotone{f2.eps}
164: \caption{\label{fig2}
165: Time evolution of the source positions relative
166: to the trend lines. The color code is the same as in Fig.~\ref{fig1},
167: referring to the major subpeaks. Triangles and stars with error bars refer to values
168: derived using CLEAN, squares and circles using PIXON, for the western
169: and eastern FPs, respectively.
170: \emph{Top:} The upper curve displays the parallel coordinates of the western FPs,
171: the lower curve the same of the eastern FPs.
172: \emph{Bottom:} Time evolution of the coordinate perpendicular
173: to the regression lines. The upper curve refers to the western FP (scale on the right),
174: the lower curve to the eastern FP (scale on the left).
175: Both panels show in black the averaged smoothed motion for each FP (PIXON value),
176: defining a new reference for detailed analysis.}
177: \end{figure}
178:
179:
180:
181:
182: Spatially integrated HXR spectra for this event were obtained
183: at a cadence of one RHESSI rotation period ($\sim 4$ s)
184: and used in a previous study \citep{Grigis04} to analyze the
185: time evolution of the non-thermal part. Here we additionally
186: produce HXR images averaged over two rotation periods ($\sim 8$ s)
187: in the energy band 20--50~keV using the CLEAN \citep{Hurford02}
188: and the PIXON \citep{Metcalf96} reconstruction algorithms.
189: The images resulting from the two different methods were inspected
190: and compared. We discarded images of poor quality, obtaining
191: 43 CLEAN and 69 PIXON images. Most images show two sources
192: located at opposite sides of the magnetic neutral line, for some others only
193: one source is clearly defined. We computed the source
194: positions in each image by fitting a two-dimensional elliptical
195: Gaussian to each visible source separately. For the CLEAN images
196: we were able to estimate the statistical positional error by
197: dividing the $1\sigma$ source width (provided by the Gaussian fit)
198: by the signal to noise ratio (obtained dividing the peak flux
199: by the standard deviation of the fluctuations in the image outside
200: the sources).
201: The average error estimated in this way amounts to 1.4$\arcsec$.
202: This method cannot be applied for the PIXON images, since
203: the PIXON algorithm suppresses the noise in the image.
204:
205:
206:
207:
208:
209:
210: The evolution of the positions of the eastern and western
211: FPs are shown in the top panel of Fig.~\ref{fig1}
212: superimposed on a SOHO/EIT \citep{Delaboudiniere95} image taken at 13:48 UT
213: showing a post-flare loop arcade. The crosses represent the
214: CLEAN positions with their error bars, and the circles the PIXON
215: positions. We compensate for the effect of the solar rotation by
216: rotating each source to the position it would assume
217: at the time of the EIT image.
218:
219: Both FPs start from the northern part of the image and move
220: along the two ribbons visible to EIT in the south-east direction.
221: The northern part of the arcade is wider than its southern end,
222: and therefore the north-south movement along it effectively
223: causes a convergence of the opposite FPs.
224:
225: In the bottom panel of Fig.~\ref{fig1} the time
226: evolution of the non-thermal HXR flux at 35~keV and the spectral
227: index are presented. Emission peaks with a duration ranging from a few tens of seconds
228: to the observational limit at 8 s can be noted, each showing soft-hard-soft behavior.
229: The main peaks are drawn in a color different from their neighbors such that the source
230: positions in the top panel, having the same code, can be followed in their temporal evolution.
231:
232:
233:
234: To characterize the motion along the arcade, we define
235: an eastern and a western regression line obtained by two
236: independent least-squares fittings of all the positions of the eastern and western FPs.
237: The two straight lines go from SE to NW and are not shown in Fig.~\ref{fig1}. The lines
238: are inclined by $74\degr$ (eastern) and $36\degr$ (western) to the E--W
239: direction. From now on, every decomposition of a vector in its \emph{parallel}
240: and \emph{perpendicular} components will refer to the directions
241: given by the regression lines. The parallel coordinate increases from an arbitrary origin towards
242: NW, whereas the perpendicular coordinate is positive in the direction
243: which points away from the arcade.
244:
245: The motions parallel and perpendicular to the two regression lines are
246: presented in of Fig.~\ref{fig2}. The FPs move predominantly along the lines,
247: thus parallel to the ribbons. The parallel motion is quite smooth and continuous,
248: especially for the western FP. For both FPs, the speed diminishes after
249: about 13:17 UT. The only large discontinuity in the parallel motion
250: is a possible $20 \arcsec$ jump of the parallel eastern FP position after the strongest
251: HXR peak when the eastern source is not detectable. Afterwards, the eastern FP
252: move slower and get stationary after 13:20 UT. Contrary to previous reports, we find
253: no correlation between FP speed and HXR flux.
254:
255: Do subpeaks show motions perpendicularly outward from the ribbons as expected from the standard
256: reconnection model? In Fig.~\ref{fig2} (bottom) this is not obvious, although the two FPs are
257: apparently moving relative to the regression line. Note however, that the lines are converging, thus
258: the effective FP separation decreases. Moreover, the ribbons are not straight. To study the question
259: in more detail, we additionally define two smooth trend curves following the FP motions more closely.
260: A moving average of the PIXON positions of each FP branch was computed, using a boxcar smoothing
261: window of 15 elements, interpolating the missing points. The interval corresponds to a duration
262: of 120 seconds, longer than all impulsive subpeaks in the HXR light curve (Fig.~\ref{fig1}, bottom).
263: The smooth trends are shown in Fig.~\ref{fig2} as black continuous curves.
264:
265:
266: The standard reconnection model predicts outward FP motion at a given place in the arcade.
267: In order to look for such systematic trends within HXR subpeaks, we took the parallel
268: and perpendicular components of the difference vector from the smoothed source position
269: to the observed PIXON positions.
270: For each subpeak, we averaged the positions occurring during the first half and
271: the second half, and calculated the difference second minus first half,
272: $\Delta_\mathrm{POS}$, for both eastern and western sources.
273:
274:
275:
276:
277: \begin{figure}
278: \plotone{f3.eps}
279: \caption{\label{fig3}
280: Distribution of the average motions during a peak in
281: perpendicular and parallel directions relative to the time averaged trend curves.
282: Eastern FPs are shown with continuous lines, western FPs with dashed lines.}
283: \end{figure}
284:
285:
286:
287:
288:
289: For elementary flare bursts produced by standard reconnection, one would expect outward
290: moving sources, thus $\Delta^\perp_\mathrm{POS}$ being positive, at least on the average.
291: Furthermore, the motion along the ribbons should be stepwise and discontinuous with
292: $\Delta^\parallel_\mathrm{POS}$ being positive if each Elementary Flare Burst were a
293: localized event. Figure \ref{fig3}
294: demonstrates that these expectations are not satisfied during subpeaks of this flare.
295: The distribution of the average perpendicular motion during each peak shown in
296: Fig.~\ref{fig3} has a mean $\Delta^\perp_\mathrm{POS}$ value of $0.0\arcsec \pm 0.4 \arcsec$
297: for the eastern FP and $0.2\arcsec \pm 0.5 \arcsec$ for the western FP (the error is
298: the standard error of the average). The mean value of the relative parallel motion
299: during the peaks is $-1.0\arcsec \pm 1.0 \arcsec$ for the eastern FP and
300: $0.4 \pm 0.7 \arcsec$ for the western FP.
301:
302: The global motion along the arcade progresses with an average
303: velocity in the parallel direction of 63 km s$^{-1}$ for the
304: eastern FP and 55 km s$^{-1}$ for the western FP. The lower velocity of the
305: western FP is due to the fact that the latest data points have negative
306: parallel velocities since they move backwards (Fig.~\ref{fig2}).
307: Averaging the absolute values of the parallel component of the velocity, we get
308: 65 km s$^{-1}$ for the western FP.
309: A speed of about 110 km s$^{-1}$ is maintained for 2 minutes in the western FP
310: at the beginning of the flare, while the data gap and possible jump around 13:17
311: in the eastern FP position requires 180 km s$^{-1}$.
312:
313: The line connecting the two FPs is inclined with respect to the post-flare loops
314: seen in the EIT image (Fig.~\ref{fig1}).
315: The angle is in the range of $25\degr$ to $70\degr$,
316: the post-flare loops being nearly perpendicular to the neutral line. This indicates that
317: the HXR emitting loops are strongly sheared. RHESSI images at lower energy where thermal
318: emission dominates the spectrum show sources or loops between the FPs. They appear to be
319: coronal sources moving along the arcade with the FPs at higher energies.
320:
321:
322: \section{Conclusion}
323:
324: We have analyzed RHESSI HXR observations of the time evolution
325: of both images and spectra for a solar flare of GOES class
326: M4.9. Surprisingly, the footpoints move smoothly along the two ribbons
327: in contrast to the bursty evolution of the HXR flux. The observed
328: Elementary Flare Bursts have durations between 30 s and less than 8 s,
329: and show pronounced spectral soft-hard-soft behavior.
330: The parallel source motions exclude the generally held notion of
331: Elementary Flare Bursts being the modulation of a global reconnection process.
332: Instead, the temporal modulation of the HXR flux and spectral index appear
333: to be caused by a spatial displacement along the arcade.
334: This could be caused by some disturbance propagating smoothly along the arcade,
335: sequentially triggering a reconnection process in successive
336: loops of the arcade. The disturbance would have to propagate
337: with a speed in the range 50--150 km s$^{-1}$, much lower than the Alfv\'en velocity.
338:
339: In the impulsive phase of this flare, magnetic energy release appears not in the form of a
340: quasi-steady reconnection annihilating anti-parallel magnetic field and thus producing outward
341: moving FPs. The main flare energy release at a given position in the arcade seems to last only a
342: short time (order of a few seconds) and moves along the arcade in a systematic manner.
343: The observed modulation of the HXR flux and the related anti-correlation of the spectral index in
344: each Elementary Flare Burst appear to be caused by spatial variations of the acceleration efficiency.
345: The temporal variations thus seem to be the result of a continuously moving trigger propagating
346: through variable conditions in the arcade. The short lifetime of a FPs at a given position shows that
347: particle trapping is not effective over timescales larger than several tens of seconds.
348:
349: The observed simple and systematic motions set this event apart as a prototype for a type of HXR flare
350: evolving \emph{along the arcade}. The FP motions of this flare contradict clearly the expectations of
351: the standard 2D reconnection model. The fact that we do not observe a systematic increase
352: (up to the instrumental limits) of the
353: separation of the FPs, does challenge the idea that the reconnection points move upwards and particles
354: are accelerated in field lines successively farther out during the main HXR emitting phase of the
355: flare. A possible interpretation is that the trigger releases the main energy stored in a
356: two-dimensional loop structure within seconds, without noticeable FP motion, and moves on.
357: Reconnection in the given structure may still continue, but with HXR emission below RHESSI
358: sensitivity and at a much reduced energy release rate. Such secondary reconnection may be the
359: cause of decimetric radio emissions continuing for 6 minutes after 13:22 UT, the end of HXR
360: emission, and may produce the expansion of the two H$\alpha$ ribbons as observed in other flares.
361:
362:
363: We thus propose a scenario in which a disturbance, probably connected with the eruption of a filament,
364: propagates along the arcade like a burning fuse, sequentially triggering
365: reconnection and particle acceleration in the flare loops.
366: The main HXR emission from the FP reflects the propagation of this disturbance, not the
367: reconnection process at a given place in the arcade. If the dominating emission is strong
368: and short-lived, the local conditions cause the observed temporal modulation.
369:
370:
371: The global evolution may be compatible with the standard model of an eruptive flare, if one
372: allows the filament to erupt in such a way that one of its ends does not move while
373: the other starts to rise. In this scenario the reconnection process spreads along the arcade
374: until it reaches the end. The arcade erupts similar to the opening of a zipper, where the
375: lower side run across the arcade and the upper side is the filament. Future studies of HXR FPs
376: in a large number of flares may establish such a scenario and stimulate the development
377: of 3D reconnection models needed to understand these observations.
378:
379:
380:
381:
382: \acknowledgements
383: The analysis of RHESSI data at ETH Zurich is partially supported by the Swiss
384: National Science Foundation (grant nr. 200020-105366). We thank the many people
385: who have contributed to the successful operation of RHESSI and acknowledge
386: S. Krucker for helpful discussions.
387:
388:
389:
390: \begin{thebibliography}{}
391:
392:
393: \bibitem[Delaboudini\`ere et al.(1995)]{Delaboudiniere95} Delaboudini\`ere,
394: J.-P., et al.\ 1995, \solphys, 162, 291
395:
396: \bibitem[de Jager \& de Jonge(1978)]{deJager78} de Jager, C., \&
397: de Jonge, G.\ 1978, \solphys, 58, 127
398:
399: \bibitem[Fletcher \& Hudson(2002)]{Fletcher02} Fletcher, L., \&
400: Hudson, H.~S.\ 2002, \solphys, 210, 307
401:
402: \bibitem[Grigis \& Benz(2004)]{Grigis04} Grigis, P.~C., \& Benz,
403: A.~O.\ 2004, \aap, 426, 1093
404:
405: \bibitem[Hoyng et al.(1981)]{Hoyng81}
406: Hoyng, P., et al.\ 1981, \apjl, 246, L155
407:
408: \bibitem[Hurford et al.(2002)]{Hurford02} Hurford, G.~J., et al.\
409: 2002, \solphys, 210, 61
410:
411: \bibitem[Krucker et al.(2003)]{Krucker03} Krucker, S., Hurford,
412: G.~J., \& Lin, R.~P.\ 2003, \apjl, 595, L103
413:
414: \bibitem[Lin et al.(2002)]{Lin02} Lin, R.~P., et al.\ 2002,
415: \solphys, 210, 3
416:
417: \bibitem[Lin et al.(2003)]{Lin03} Lin, R.~P., et al.\ 2003,
418: \apjl, 595, L69
419:
420: \bibitem[Liu et al.(2004)]{Liu04} Liu, W., Jiang, Y.~W., Liu,
421: S., \& Petrosian, V.\ 2004, \apjl, 611, L53
422:
423: \bibitem[Masuda et al.(1994)]{Masuda94} Masuda, S., Kosugi, T.,
424: Hara, H., Tsuneta, S., \& Ogawara, Y.\ 1994, \nat, 371, 495
425:
426: \bibitem[Metcalf et al.(1996)]{Metcalf96} Metcalf, T.~R., Hudson,
427: H.~S., Kosugi, T., Puetter, R.~C., \& Pina, R.~K.\ 1996, \apj, 466, 585
428:
429: %sausage modes
430: \bibitem[Nakariakov et al.(2003)]{Nakariakov03} Nakariakov, V.~M.,
431: Melnikov, V.~F., \& Reznikova, V.~E.\ 2003, \aap, 412, L7
432:
433:
434: \bibitem[Parks \& Winckler(1969)]{Parks69} Parks, G.~K., \&
435: Winckler, J.~R.\ 1969, \apjl, 155, L117
436:
437: \bibitem[Priest \& Forbes(2002)]{Priest02} Priest, E.~R., \&
438: Forbes, T.~G.\ 2002, \aapr, 10, 313
439:
440: \bibitem[Qiu et al.(2004)]{Qiu04} Qiu, J., Lee, J., \& Gary,
441: D.~E.\ 2004, \apj, 603, 335
442:
443: \bibitem[Roberts et al.(1984)]{Roberts84} Roberts, B., Edwin,
444: P.~M., \& Benz, A.~O.\ 1984, \apj, 279, 857
445:
446: \bibitem[Siarkowski \& Falewicz(2004)]{Siarkowski04} Siarkowski, M.,
447: \& Falewicz, R.\ 2004, \aap, 428, 219
448:
449: \bibitem[Smith et al.(2002)]{Smith02} Smith, D.~M., et al.\
450: 2002, \solphys, 210, 33
451:
452: \bibitem[Stepanov et al.(2004)]{Stepanov04} Stepanov, A.~V., et al.\
453: 2004, Astronomy Letters, 30, 480
454:
455: \bibitem[Takakura et al.(1983)]{Takakura83} Takakura, T., Tsuneta,
456: S., Nitta, N., \& Ohki, K.\ 1983, \solphys, 86, 323
457:
458: \bibitem[Veronig \& Brown(2004)]{Veronig04} Veronig, A.~M., \&
459: Brown, J.~C.\ 2004, \apjl, 603, L117
460:
461: \bibitem[Yashiro et al.(2004)]{Yashiro04} Yashiro, S.,
462: Gopalswamy, N., Michalek, G., St.~Cyr, O.~C., Plunkett, S.~P., Rich, N.~B.,
463: \& Howard, R.~A.\ 2004, Journal of Geophysical Research (Space Physics),
464: 7105
465:
466: \end{thebibliography}
467:
468:
469: \end{document}