1: %\documentclass{emulateapj}
2: %\documentclass[manuscript]{aastex}
3: \documentclass[12pt,preprint]{aastex}
4: %\documentclass[preprint]{aastex}
5:
6: \slugcomment{}
7:
8: \shorttitle{Large-scale Alfv\'en waves in solar flares}
9: \shortauthors{Fletcher \& Hudson}
10: \begin{document}
11: \title{Impulsive phase flare energy transport by large-scale Alfv\'en waves and the electron acceleration problem}
12:
13: \author{L.~Fletcher\altaffilmark{1}\email{lyndsay@astro.gla.ac.uk}}
14: \affil{Department of Physics and Astronomy, University of Glasgow, Glasgow G12 8QQ}
15: \author{H.~S. Hudson\email{hhudson@ssl.berkeley.edu}}
16: \affil{Space Sciences Laboratory, University of California, Berkeley, CA 94720}
17:
18: \altaffiltext{1}{Carried out while a Visiting Researcher, Space Sciences Laboratory, University of
19: California, Berkeley, CA}
20:
21: \begin{abstract}
22: The impulsive phase of a solar flare marks the epoch
23: of rapid conversion of energy stored in the pre-flare coronal magnetic
24: field. Hard X-ray observations imply that a substantial fraction of
25: flare energy released during the impulsive phase is converted to the
26: kinetic energy of mildly relativistic electrons (10-100~keV). The
27: liberation of the magnetic free energy can occur as the coronal
28: magnetic field reconfigures and relaxes following reconnection. We
29: investigate a scenario in which products of the reconfiguration --
30: large-scale Alfv\'en wave pulses -- transport the energy and
31: magnetic-field changes rapidly through the corona to the lower
32: atmosphere. This offers two possibilities for electron
33: acceleration. Firstly, in a coronal plasma with $\beta < m_e/m_p$, the
34: waves propagate as inertial Alfv\'en waves. In the presence of strong
35: spatial gradients, these generate field-aligned electric fields that
36: can accelerate electrons to energies on the order of 10~keV and above,
37: including by repeated interactions between electrons and
38: wavefronts. Secondly, when they reflect and mode-convert in the
39: chromosphere, a cascade to high wavenumbers may develop. This will
40: also accelerate electrons by turbulence, in a medium with a locally
41: high electron number density. This concept, which bridges MHD-based
42: and particle-based views of a flare, provides an interpretation of the
43: recently-observed rapid variations of the line-of-sight component of
44: the photospheric magnetic field across the flare impulsive phase, and
45: offers solutions to some perplexing flare problems, such as the flare
46: ``number problem'' of finding and resupplying sufficient electrons to
47: explain the impulsive-phase hard X-ray emission.
48: \end{abstract}
49:
50: \keywords{Sun:flares,corona; waves; acceleration of particles}
51:
52: \section{Introduction}\label{sect:intro}
53:
54: Strong chromospheric hard X-ray emission and strong UV and white-light
55: emission mark the impulsive phase of a solar flare. These signatures
56: are usually interpreted in terms of the well-known ``thick-target
57: model'' \citep{Brown71,1972SoPh...24..414H} in which fast electrons
58: lose energy in Coulomb collisions and ionizing collisions in the
59: chromosphere, heating and producing bremsstrahlung {\it en route}. The
60: inefficiency of the bremsstrahlung process in a cold thick target
61: implies that a large fraction of flare energy resides in these
62: electrons \citep{Kane71,Lin76,Holman03}, and calculations under the
63: assumptions of the thick-target model yield numbers on the order of
64: $10^{34} - 10^{37}$ electrons accelerated per second
65: \citep[e.g.][]{1997ApJ...491..939M,Holman03}. Various strands of
66: evidence have led to the commonly-accepted idea that the particle
67: acceleration takes place in the solar corona, following which the
68: electrons propagate into the lower atmosphere where they heat, and
69: generate the observed hard X-ray radiation. Extensive theoretical work
70: over four decades (which we will not attempt to summarize here) has
71: elucidated the basics and the specifics of numerous different coronal
72: acceleration mechanisms, in the electric fields present in current-sheets
73: and X-lines/points generated by reconnection, in large- and small-scale
74: plasma waves and turbulence, and at shocks. Recent reviews can be
75: found in \cite{2002SSRv1011A} or \cite{2003LNP...612..213L}, for example.
76: However, a coronal acceleration site still
77: presents some problems for theory. The primary difficulty, especially
78: in the context of the high intensity of the energy deposition implied
79: not only by hard X-rays but also by UV and white-light continuum
80: observations \citep[e.g.][]{2007ApJ...656.1187F}, is the so-called
81: ``number problem'' - the high total number of electrons required
82: compared to that available in the corona - and the associated (and in
83: fact more problematical) supply and re-supply problems.
84:
85: The thick-target model as normally understood requires intense
86: electron beams to transport the flare energy. We propose instead that
87: flare energy is transported by the Poynting flux of Alfv{\' e}n
88: waves. Since flare energy release implies large-scale restructuring of
89: the coronal magnetic field (e.g. via reconnection) it is natural to
90: expect the excitation of such waves \citep{1982SoPh...80...99E}. The
91: electron acceleration can then take place where the waves dissipate,
92: in the legs of the coronal loops or in the chromosphere itself.
93:
94: The possibility of flare energy transport by Alfv{\'e}n waves has been
95: discussed before, for example by \cite{1982SoPh...80...99E} in the
96: context of heating the temperature minimum region, and more generally
97: by \cite{1992ApJ...387..403M} and \cite{1994PASAu..11...25W} who
98: examined the propagation of twist in a flare loop. The present paper
99: seeks to restart the discussion of flare wave energy transport, in
100: the light of recent solar observations and recent developments in
101: magnetospheric physics, as well as because of the outstanding
102: theoretical issues with coronal electron acceleration, which have
103: been exacerbated by RHESSI, TRACE and other observations.
104:
105: The main solar physics drivers for revisiting this idea are as
106: follows. Firstly, recent microwave (gyrosynchrotron) observations of
107: the corona above active regions demonstrate conclusively that
108: magnetic field strengths of several hundredths up to more than a
109: tenth of a Tesla (i.e. several 100s of Gauss to kG) exist in the
110: cores of active regions, measured at heights up to 10,000 - 15,000~km
111: above the photosphere. Coupled with reasonable coronal densities of
112: $10^{15}{\rm m}^{-3}$ these fields imply Alfv{\'en} wave speeds well
113: above $10^4{\rm~km~s}^{-1}$, and correspondingly high Poynting fluxes. The
114: observational basis for these physical parameters described in some
115: detail in Section~\ref{sect:phys}. Secondly, there is clear evidence
116: that substantial perturbations to the photospheric magnetic field (on
117: the order of 0.01 to 0.02~T) occur during solar flares. Field changes
118: in the low corona, on height scales comparable to the horizontal
119: dimensions of active regions, must be of similar magnitude. This
120: strongly suggests a violent perturbation to the magnetic field, at a
121: low level in the atmosphere, which is at least qualitatively
122: consistent with a very energetic magnetic disturbance
123:
124: In magnetospheric physics, electron acceleration in the parallel electric field that results from the propagation of
125: large-scale Alfv\'en waves and wave pulses in a non-idealised MHD fluid is a promising prospect
126: for auroral electron acceleration, and also motivates us in this work.
127: In the magnetospheric/ionospheric context it was pointed out early on
128: that non-ideal effects arise from considering both the two-fluid nature of the plasma (i.e. treating electrons as a separate fluid, and including their inertia and thermal pressure) and also the particle aspects of the problem (e.g., the finite ion
129: gyroradius). These lead to field-aligned electric fields, and the presence of such
130: dispersive Alfv\'en waves and their link to electron acceleration is
131: now well-established observationally
132: \citep[e.g.][]{2002JGRA.107hSMP24W}. \cite{2002GeoRL..29k..30C} have
133: demonstrated that the value of the energy flux carried by auroral
134: electrons is similar to the Poynting flux of low frequency Alfv\'en
135: oscillations of the magnetospheric field. Debates persist about the
136: precise mechanism for generating the electric fields that accelerate
137: auroral electrons \citep[e.g.,][]{2000SSRv...92..423S}, but the
138: inertial Alfv\'en wave (see Section~\ref{sect:iaw}) is a strong
139: candidate. This may also have a role to play in the case of flares,
140: although the solar and magnetospheric cases of course represent very
141: different parameter regimes. We demonstrate in Section~\ref{sect:iaw} that the inertial Alfv\'en wave mode is also the appropriate one to consider for flare parameters. The critical factor in determining the parallel electric field that can be generated is the spatial scale of perpendicular structuring of the magnetic field compared to the electron inertial length, and - as we describe - observations at ever higher resolution are showing finer and finer magnetic field stucturing.
142:
143: We note also that electron acceleration
144: by non-ideal MHD waves is also making its way into the discussion of
145: coronal heating. \cite{2006PhRvL..96q5003S} and
146: \cite{2007PhRvL..98d9502S} claim that dispersive Alfv\'en waves driven
147: by photospheric turbulence lead to parallel electric fields and
148: electron heating, and \cite{2006A&A...455.1073T} finds the generation
149: of a parallel electric field and runaway electron heating when an
150: initially ideal (non-dispersive) but non-linear Alfv\'en wave couples
151: to dissipative modes when it is launched into a corona with transverse
152: density structure. Our considerations are somewhat different from this idea,
153: in that our inertial Alfv\'en wave is dispersive from the start. This does not preclude also the kind of mode coupling discussed by \cite{2006A&A...455.1073T}; instead this would be an additional energy loss term which will require further study in the future.
154:
155: We first describe the proposed mechanism in
156: Section~\ref{sect:mechanism}, including a detailed description of the
157: observations that motivate us. The hard X-ray observations, as
158: confirmed by RHESSI, require powerful electron acceleration, and in
159: Section~\ref{sect:acceleration} we discuss possibilities for this in
160: the framework of the wave transport model.
161: Section~\ref{sect:energetics} then considers the overall implications for
162: flare energetics.
163:
164: \section{The proposed mechanism}\label{sect:mechanism}
165:
166: \subsection{The waves}
167:
168: \subsubsection{Wave source}
169: The release of stored magnetic energy
170: requires a re-structuring of the field, for example as envisioned in
171: large-scale magnetic reconnection, However the amount of magnetic
172: free energy that can be {\it dissipated} within the reconnection
173: region itself -- current sheet, X-point or 3-D null -- is restricted,
174: given its small dimensions and the short flare time-scale. The more
175: important release of free energy occurs in the large-scale
176: `convulsion' as the newly--reconnected magnetic field relaxes from its
177: pre-flare stressed state. Where they detach from the coronal
178: current-sheet or null structure but are still stressed, these magnetic
179: field lines will be highly distorted from a potential configuration,
180: with a locally high tension force. We know observationally that the
181: impulsive energy release occurs in a highly-stressed magnetic field,
182: with large fluctuations on time scales ranging down to a fraction of
183: a second \citep[e.g.,][]{1985SoPh..100..465D}. This implies irregular and time-varying structures in a
184: three-dimensional reconnection flow. Thus Petschek
185: reconnection, which is essentially steady-state, cannot properly
186: describe it.
187:
188: The rapid restructuring of the field implies an energy flow describable
189: in terms of MHD wave propagation, and we infer that it will create a
190: complicated mixture of fast-mode, slow-mode, and Alfv\'en-mode propagating wave
191: pulses in the adjacent plasma. For example, flare loop `shrinkage'
192: \citep[e.g.,][]{1996ApJ...459..330F} identifiable with the MHD fast
193: mode is a simple and well-known illustration of this idea, as are the
194: slow-mode shocks of Petschek fast reconnection. MHD modeling of
195: three-dimensional reconnection is at an early stage, but in three
196: dimensions a torsional component will in general also exist,
197: particularly in a reconnecting twisted field
198: \citep{1982SoPh...80...99E}. Indeed, in-situ observations of
199: reconnection in the solar wind \citep {2005JGRA..11001107G} show
200: Alfv\'en waves propagating along just-reconnected field lines, and
201: the MHD simulations of \cite{2006ApJ...642.1177L} demonstrate a
202: post-reconnection state of initially untwisted flux tubes in which
203: field-line kinks propagate away at close to the Alfv\'en speed. Since we require to deposit flare energy in the flare footpoints, we require a wave mode that propagates along the magnetic field - either the Alfv\'en mode or the slow mode. However, the slow mode speed is too low to explain the observed footpoint simultaneity unless we have extremely symmetric propagation from exactly half way between the footpoints. For the same reason of low speed, neither can it explain the require high energy flux (see Section 4). Thus, we work under the assumption of an Alfv\'enic disturbance carrying energy
204: along the post-reconnection field.
205:
206: \subsubsection{Wave development}\label{sect:develop} We sketch our
207: overall view of a post-reconnection loop and the processes taking
208: place in it in Figure~\ref{fig:cartoon}. The perturbation in 3D takes
209: the form of fast-mode and Alfv{\' e}n-mode wave pulses
210: \citep{1982SoPh...80...99E}; the group velocity of the Alfv{\' e}n
211: mode is parallel to the magnetic field {\bf B}, so this component of
212: the energy propagates directly to the footpoints as shown in the
213: cartoon. In the MHD view the propagation speed is just the Alfv\'en
214: speed \citep[in a kinetic treatment][also
215: recovered this result]{1979JGR....84.7239G}.
216:
217: The wave spectrum will be determined by the largely-unknown geometry
218: of the energy release. It is likely that the Alfv\'en wave will take
219: the form of a short-wavelength propagating pulse - a wavefront - with
220: parallel wavelength much smaller than the length $L$ of a
221: just-reconnected loop. The perpendicular wavelengths would be much
222: smaller than the loop length, as dictated by the reconnection rate
223: and its fluctuations.
224:
225: Under appropriate conditions (Section~\ref{sect:phys}) the Alfv\'enic
226: perturbation will propagate rapidly through the coronal field to the
227: chromosphere without significantly cascading to smaller scales. This
228: is different from (but complementary to) the view of
229: \cite{1994ApJ...425..856L} and \cite{1997ApJ...491..939M}, in which
230: the large-scale fast-mode waves formed by reconnection are assumed to
231: cascade rapidly to short-wavelength turbulence within the coronal
232: loop, leading eventually to the Fermi acceleration of electrons in
233: high-frequency turbulence directly in the corona. For the ducted
234: Alfv{\' e}n mode it has been shown \citep[e.g][]{PhysRevE.57.7111,
235: 2005ApJS..156..265C} that a cascade will not develop immediately.
236: Therefore in the situation we envisage, the Alfv\'enic perturbation
237: will move from corona to chromosphere along a strong guide field
238: without driving a cascade, at least in the initial pass. The wave
239: energy will be strongly ducted towards the chromosphere.
240:
241: If some fraction of the wave energy is reflected at the chromosphere,
242: so that counter-moving waves are present in the corona, then a cascade
243: may occur. However, even then, \cite{PhysRevE.57.7111} demonstrate
244: using reduced MHD simulations that the cascade to high parallel
245: wavenumbers is inhibited, and an exponentially-decaying rather than a
246: power-law spectrum will be formed, while the cascade to high {\it
247: perpendicular} wave numbers proceeds independently.
248:
249: On arriving at the chromosphere and photosphere the wave propagation
250: will become more complicated, with transmission, reflection and
251: damping all playing a role. The waves will undergo different kinds of
252: damping, including -- in the temperature-minimum region -- significant
253: ion-neutral damping. The line-tied boundary conditions at the
254: photosphere mean that the purely Alfv\'en disturbance will not survive
255: as such but instead, as demonstrated by \cite{1994A&A...286..275G}, a
256: reflected wave spectrum with hybrid characteristics will be generated,
257: and some fast-mode-like components will arise, particularly in the
258: presence of chromospheric small-scale structuring and flows. Being
259: compressional, these fast-mode-like waves can be locally damped by
260: other mechanisms, and offer also the possibility for a turbulent
261: cascade development in the chromosphere, analogous to that proposed
262: by \cite{1994ApJ...425..856L} for coronal acceleration. The
263: analysis of \cite{1994A&A...286..275G} suggested that any reflected
264: waves that do re-emerge into the corona would have a mostly torsional
265: (Alfv\'enic) character.
266:
267: \begin{figure}
268: \plotone{f1.eps}
269: \caption{The reconfiguring coronal field
270: launches a torsional Alfv\'en wave pulse through the corona
271: and into the chromosphere, as well as a fast-mode wave pulse. The Alfv\'en
272: wave, which propagates in the inertial regime, can lead to electron
273: acceleration in the corona. That fraction of the Alfv\'en wave energy that
274: survives into the chromosphere can also lead to stochastic
275: acceleration there. The wave will be partially reflected from the
276: steep gradients in the chromosphere (not shown) and re-enter the
277: corona.}
278: \label{fig:cartoon}
279: \end{figure}
280:
281: \subsection{The particles}
282:
283: The hard X-ray observations unambiguously require powerful electron
284: acceleration. How can this arise from energy transported in the
285: Poynting flux of Alfv{\' e}n waves? We discuss possible mechanisms in
286: Section~\ref{sect:acceleration} and briefly comment here on the
287: particle behavior in the context of Figure~\ref{fig:cartoon}. In the
288: new scenario the acceleration of the energetically important
289: 10-100~keV electrons either takes place in the legs of the flaring
290: loops, or actually in their footpoint regions.
291:
292: Alfv\'enic perturbations propagating in the limit $\beta <
293: \rm{m_e/m_p} $ (the inertial Alfv\'en wave limit) lead to a parallel
294: electric field $E_{\parallel}$. For a wave traveling downwards,
295: electron inertia produces an upwards $E_{\parallel}$. A fraction of
296: the electrons are resonantly accelerated in this field, in a process
297: that can be thought of as an encounter with a moving mirror
298: \citep{1994JGR....9911095K,chaston06}, with the electrons reflecting
299: from the traveling perturbation front and accelerating to twice the
300: Alfv\'en speed $v_A$. In the conditions we envisage, where the
301: Alfv\'en speed is on the order of 0.1-0.3~$c$ (see
302: Section~\ref{sect:phys}), this corresponds to an `Alfv\'en energy'
303: $( = {1\over 2} m_e v_A^2)$ in the few to tens of keV range. Multiple
304: reflections of the electron between the wave front and magnetic mirror
305: formed by the converging chromospheric magnetic field may occur, each
306: reflection from the wave front increasing the electron speed by $2
307: v_{A}$ in first-order Fermi acceleration.
308:
309: As mentioned in Section~\ref{sect:develop} a turbulent wave spectrum
310: may be generated in the footpoint regions. In the chromosphere, the
311: damping of this spectrum will broadly-speaking result in plasma
312: heating, since the electron-electron thermalization times are very
313: short. However, an essentially collisionless tail of fast electrons
314: can be accelerated by Fermi processes, as in the case of coronal
315: stochastic acceleration. The question is how large that tail may be.
316: We
317: discuss this in Section~\ref{sect:turb} but note here that stochastic
318: acceleration can take place in a collisional environment
319: \citep[e.g.,][]{1992ApJ...398..350H}. The particular advantages
320: offered by chromospheric acceleration are firstly a high ambient
321: electron density (compared to the corona), possibly easing the number
322: and resupply problems, and secondly - as pointed out by Brown (2006,
323: private communication) and \cite{2006GMS...165..157M} - , the
324: requirement on the total number of accelerated electrons implied by
325: their hard X-ray signature is reduced if the accelerator acts on them
326: at the same time as they radiate bremsstrahlung emission, which would
327: be satisfied in a chromospheric accelerator. (This advantage is
328: analogous to the increased bremsstrahlung efficiency that pertains in
329: a thermal model for flare hard X-rays, where the radiating electrons
330: are continually re-boosted by interactions with a hot rather than a
331: cold target \cite[e.g.][]{1979ApJ...232..582S}.)
332:
333: This overall scenario also provides a mechanisms for some accelerated
334: electrons to appear in the corona. This is important because of the
335: extensive observational evidence for coronal non-thermal electrons,
336: e.g. via the microwave spectrum, or low energy hard X-rays. Any
337: reflected component of the inertial Alfv\'en wave pulse produces a reversed
338: electric field, which can draw chromospheric electrons back into the
339: corona. Furthermore, coronal electron acceleration by the cascade of
340: fast-mode turbulence - as proposed by \cite{1994ApJ...425..856L} - may
341: operate alongside the Alfv\'enic transport, as both wave types will be
342: generated by the reconnection process.
343:
344: \subsection{Physical Parameters\label{sect:phys}}
345:
346: The properties of the Alfv\'en waves, and the magnitude of the
347: parallel electric fields they generate, depend critically on the plasma
348: parameters; density, electron and ion temperatures, magnetic field
349: strength and length scales. We review the relevant observations here.
350:
351: \bigskip
352: \noindent {\bf Magnetic field strength:}
353:
354: It is notoriously difficult to determine the strength of the coronal
355: magnetic field, or to calculate it by extrapolations from a given
356: boundary. However, in solar flares and in the cores of active regions,
357: where the magnetic field is strong, simple geometrical arguments
358: point to intense fields in the low corona. A large sunspot may have a
359: size scale of some 3~$\times$~10$^4$~km, an umbral field of a few
360: $\times$~0.1~T, and an outer penumbral field of 0.08 to 0.17~T
361: \citep{2003A&ARv..11..153S}. For the dominant dipole term of a
362: multipole expansion of this photospheric source structure, we would
363: expect comparable coronal field intensities, at heights in the
364: vicinity of the spot comparable to the spot extent.
365:
366: Direct measurement of the strength of (strong) coronal magnetic fields
367: is also possible via the microwave gyrosynchrotron spectrum generated
368: by fast electrons. Very Large Array radio observations of active
369: regions show emission consistent with {\it average} active region
370: coronal field strengths of a few~$\times$~0.01~T
371: \citep{1998ApJ...501..853L} at a height of 10,000~km above the
372: photosphere. In the corona above sunspots, even stronger fields
373: have been measured
374: \citep{1991ApJ...366L..43W,1994PASJ...46L..17S,2002ApJ...574..453B,
375: 2006PASJ...58...11V,2006ApJ...641L..69B}. For example, using VLA and
376: SOHO data, \cite{2002ApJ...574..453B} deduce field strengths in excess
377: of 0.1~T at heights of 10,000~km above the photosphere over a sunspot
378: on the disk, and for a substantial area around it. Limb observations,
379: with less confusion in the dependence of the field strength on
380: altitude \citep{2006ApJ...641L..69B}, also give these values. Based
381: upon these observations, we can reasonably expect field strengths of a
382: few~$\times$~0.01~T at heights of 10,000~km above sunspot or strong
383: plage regions, and since flare ribbons also penetrate into sunspot
384: umbrae, low-coronal fields $> 0.1\ T$ are certainly not out of the
385: question. These magnetic field strengths are substantially higher than
386: the values inferred from coronal seismology, however the coronal
387: seismology technique has only been applied so far to large active-region
388: loops \citep[e.g.,][]{2001A&A...372L..53N}.
389:
390: The height of 10,000~km at which these strong fields are observed is
391: also consistent with the height of loops
392: involved in flares, based on their typical HXR footpoint separations of
393: typically a few tens of arcseconds. There are not to our knowledge
394: any statistical studies of this, but numerous examples can be seen in e.g.
395: \cite{1994PhDT.......335S}, \cite{2005ApJ...630..561B}, \cite{2006A&A...456..751B}, \cite{2007ApJ...656.1187F}. A
396: typical separation value of 30" or 20,000~km corresponds to a
397: semicircular loop with apex height of 10,000~km.
398:
399: \bigskip
400: \noindent {\bf Density:} With the exception of coronal soft X-ray
401: `knots' \citep[e.g.][]{1995ApJ...440..370D} and rare observations of
402: dense coronal loop flares which show negligible footpoint emission
403: \citep[e.g.][]{2004ApJ...603L.117V}, the coronal density before and
404: early in a flare is fairly low. Several studies have sought pre-flare
405: signatures of the bright flare loops but the general result is that in
406: most cases no feature visible in soft X-rays matches the flare loops
407: that form after the impulsive phase
408: \citep{1996SoPh..165..169F,1998SoPh..183..339F}. This suggests that
409: the energy release takes place in regions of yet lower density than
410: the average active-region corona. Normal active-region loop densities
411: are on the order of $1-3 \times 10^{15}{\rm cm}^{-3}$
412: \citep{2003A&A...406.1089D} and even post-flare arcade loop
413: measurements \citep{2000A&A...355..769V,2003ApJ...582..506L} are a
414: few $\times 10^{15}{\rm m^{-3}}$, which might reasonably be taken as
415: an upper limit for the pre-flare density in the flare region. In the
416: study of a sunspot magnetic field mentioned above,
417: \cite{2002ApJ...574..453B} estimated plasma densities at a few
418: $\times 10^{14}{\rm m}^{-3}$ to $10^{15}{\rm m}^{-3}$ in the
419: essentially `empty' corona above a sunspot. Finally,
420: \cite{1999ApJ...520L.135F} find upper-transition region densities of
421: $2-5 \times 10^{15}{\rm m}^{-3}$ in the cores of active regions, again implying a lower density for the overlying hotter
422: corona.
423: %It should also be noted that line-ratio diagnostics applied in
424: %an inhomogeneous plasma are biased towards the higher densities
425: %\citep[e.g.,][]{1989A&A...224..328A}.
426: Taken together, these various
427: strands of evidence imply that pre-flare coronal densities on the
428: order of $10^{15}{\rm m}^{-3}$ or possibly smaller are common, and in
429: many cases we have only upper limits.
430:
431: \bigskip
432: \noindent {\bf Alfv{\' e}n speed:}
433: If we take a magnetic field strength of 0.05~T and a proton number
434: density $n_p = 10^{15}$~m$^{-3}$, in a fully-ionized hydrogen plasma
435: the Alfv\'en speed is $3.5~\times$~10$^{4}$~km~s$^{-1}$. Higher values
436: of $\left| \mathbf{B} \right|$ or lower values of $n_p$ are also
437: possible, so $v_A$ could thus be a few~$\times$~0.1~c. These may seem
438: like extreme values given that the `canonical' coronal value often
439: discussed is on the order of 10$^{3}$~km~s$^{-1}$, and that fast
440: coronal mass ejections -- presumably ejected at some fraction of the
441: local Alfv\'en speed -- travel at around 3,000~${\rm km\;s^{-1}}$
442: above a couple of solar radii. However the measurements, and our
443: considerations, refer to the low corona, where the bulk of the
444: magnetic energy resides, in highly-stressed, compact fields. Note
445: that since $(v_A/c)^2 << 1$ the wave can still be described
446: non-relativistically, and the displacement current may still be
447: neglected, allowing an MHD description.
448:
449: Assuming a loop half-length of $10^7$~m, the propagation time of such a
450: wave into the chromosphere from a coronal launch site is a few tenths
451: of a second at most. This is shown in Figure~\ref{fig:traveltime} for
452: a hydrostatic corona at $T=10^6$~K matched to the top of the of the
453: VAL-C chromospheric model \citep{1981ApJS...45..635V}, and using the
454: chromospheric magnetic scaling of \cite{1983ApJ...264..648Z}, i,e.,
455: $\left| \mathbf{B} \right| \propto P_{g}^{\alpha}$, where $P_{g}$ is
456: the gas pressure. The parameter $\alpha$ has been chosen to give a
457: field strength at the photosphere of 0.2~T. The propagation time
458: obtained is adequate to explain the observed timescales of hard X-ray
459: emission as well as the simultaneity of hard X-ray footpoints
460: \citep{1994PhDT.......335S}, an argument often advanced in favor of
461: energy transport by energetic electrons accelerated in the corona and
462: precipitating at the footpoints. The commonly-observed pattern of
463: slower non-thermal variations in the later phase of a solar flare may
464: result from the increase of coronal densities and decrease in the
465: strength of the reconnecting fields in this phase, and thus reduced
466: Alfv\'en speeds.
467:
468: \begin{figure}
469: \plotone{f2.eps}
470: \caption{Propagation time at the Alfv\'en speed from loop top to a
471: given height. The vertical dashed line indicates the temperature
472: minimum, and the vertical dotted the top, of the VAL-C atmospheric
473: model. This model is extended into the corona with a semicircular
474: loop of coronal half-length 10,000~km and a density scale height
475: given by the temperature assumed for the base of the corona, $10^6$~K.
476: The dashed curve shows a coronal field of 0.05 T, extended
477: through the atmosphere with $\alpha$ = 0.052 (see text); the solid
478: curve the more extreme case of 0.1~T with $\alpha$ = 0.0. }
479: \label{fig:traveltime}
480: \end{figure}
481:
482: \bigskip
483: \noindent {\bf Photospheric magnetic perturbations:}
484: The observations of non-reversible changes to the line-of-sight
485: magnetic field at the photospheric level mentioned in
486: Section~\ref{sect:intro} lend credence to our supposition that strong
487: perturbations to the magnetic field are present throughout the
488: atmosphere. For example \cite{1999ApJ...525L..61C} and
489: \cite{2001ApJ...550L.105K} observed such changes in ground-based and
490: SOHO/MDI data respectively. \cite{2005ApJ...635..647S}, using
491: simultaneous SOHO/MDI and GONG magnetogram data, observe permanent
492: line-of-sight photospheric magnetic changes (0.01-0.02 T) to be
493: ``ubiquitous features'' of X-class flares at least. The changes are observed
494: to be roughly co-spatial with the flare
495: ribbons and occur rapidly, on timescales of minutes. They are
496: therefore too fast to be due to Alfv\'enic perturbations propagating
497: upwards from the sub-photospheric region. Rather, it is as if the
498: magnetic field at the photospheric level is `jerked' by the overlying
499: magnetic field as it restructures in the corona, with both a twisting
500: component and a loop retraction.
501: The fact that we see a distortion to the photospheric magnetic field indicates
502: that there is substantial wave energy transmitted to low levels in the
503: atmosphere, although with present line-of-sight observations we cannot
504: distinguish between components corresponding to twisting and retracting.
505:
506: \bigskip
507: \noindent {\bf Transverse magnetic structuring:}
508: As will become apparent in Section~\ref{sect:damping}, the transverse
509: scale of magnetic structure is a vital parameter in our calculations,
510: but observations are strongly limited by instrumentation. We do have
511: observed upper limits to the transverse structuring of the
512: chromospheric magnetic field in the quiet sun: in recent observations
513: using the Swedish Vacuum Solar Telescope, \cite{2004A&A...428..613B}
514: report that magnetic elements seen in the G-band (the photosphere)
515: appear unresolved at 70~km spatial resolution. We may expect that
516: transverse photospheric structuring on still smaller scales may be
517: present. A lower limit to the transverse scales would be the
518: ion inertial length, in the range 10$^{-2}$--1~km at the
519: transition region interface.
520:
521:
522: \begin{figure}
523: \centering
524: \begin{vbox}{
525: \plottwo{f3a.eps}{f3b.eps}
526: \plottwo{f3c.eps}{f3d.eps}
527: }
528: \end{vbox}
529: \caption{The ratio of the Alfv\'en propagation time along a 10$^{4}$~km
530: loop to the damping time by phase mixing, for a different wavelength
531: perturbations in a coronal field of 0.05~T, a temperature of
532: $10^6$~K (left hand panel) and $10^5$~K (right hand panel). The
533: perpendicular wavelength of the perturbation is 10$^{2}$~km (upper row)
534: and 1~km (lower row).}
535: \label{fig:phase}
536: \end{figure}
537:
538: \subsection{Wave passage through the corona}\label{sect:damping}
539:
540: We establish here that the coronal Alfv\'en wave pulses can traverse
541: the corona and arrive at the chromosphere without significant viscous
542: or resistive damping. In a corona with strong non-uniformities
543: perpendicular to the field, the damping of Alfv\'en waves is by phase
544: mixing \citep[e.g.,][]{2000SoPh..193..139R}. The damping time is
545: given by his Equation~22, expressed here in terms of the wavelength:
546: \begin{equation}\label{eq:phasem}
547: \tau_{pm}= \left({6 \lambda_\parallel^2 \lambda_\perp^2}\over{4 \nu
548: \pi^2 v_a^2}\right)^{1/3}
549: \end{equation}
550: where $\lambda_\parallel$ and $\lambda_\perp$ are the parallel and
551: perpendicular wavelengths respectively. Under most conditions the
552: viscosity $\nu$ is the plasma shear viscosity, $\nu_s$, which is the kinematic
553: viscosity multiplied by $(\omega_i
554: \tau_i)^{-2}$ \citep{1983A&A...117..220H} where $\omega_i$ is the ion
555: gyrofrequency and $\tau_i$ the ion collision time. In circumstances
556: where this factor is much less than unity, Joule dissipation will
557: dominate, and the viscosity will be given by the magnetic diffusivity,
558: \begin{equation}
559: \nu_m = {1\over(\mu_o \sigma)},
560: \end{equation}
561: with $\sigma$ the Spitzer conductivity. The total viscosity we use in
562: Eq.~\ref{eq:phasem} is the sum of the shear and the Joule viscosity.
563: Figure~\ref{fig:phase} compares the phase-mixing time scale to
564: the Alfv\'en propagation time $\tau_A$ along the coronal part of the
565: loop, This shows that
566: $\tau_A/\tau_{pm} < 1$ for perturbations with parallel wavelengths of
567: more than a few tens of~km propagating in a coronal density of $\sim
568: 10^{15}{\rm m}^{-3}$. However, wave energy may be lost in accelerating
569: particles, as we describe in the next section.
570:
571: \section{Electron acceleration in the context of energy transport by Alfv\'en wave pulses}
572: \label{sect:acceleration}
573:
574: If the wave energy is transported by ducted Alfv\'en wave pulses as we
575: suggest, there are several possibilities for electron acceleration; we
576: consider three, most closely related to the wave nature of the
577: transport mechanism. Firstly, in a hot, tenuous, strongly magnetized
578: coronal plasma, it may be possible to accelerate electrons directly in
579: the corona, in the parallel electric field generated by a dispersive
580: Alfv\'en wave pulse (Sections~\ref{sect:iaw} and ~\ref{sect:field}).
581: Secondly, associated with this is the possibility that the electrons,
582: accelerated ahead of the wavefront, mirror in the converging solar
583: magnetic field and return for repeated interactions with the wave
584: (Section~\ref{sect:fermi1}). This comprises a first-order Fermi
585: acceleration process. Thirdly, the wave energy can be dissipated in
586: or near the chromosphere in a turbulent cascade which accelerates
587: electrons stochastically (Section~\ref{sect:turb}) and we discuss
588: separately the two primary models for turbulent electron acceleration;
589: stochastic resonant acceleration in high frequency whistler
590: turbulence, and transit-time acceleration in lower-frequency MHD
591: turbulence. We consider first the acceleration by inertial Alfv\'en
592: waves.
593:
594: \subsection{Inertial Alfv\'en waves}\label{sect:iaw}
595: In ideal MHD, the (massless) negative charge
596: carriers respond instantaneously to any parallel electric field
597: generated by the Alfv\'enic perturbation, shorting it out so that no
598: $E_\parallel$ exists. An ideal MHD wave includes an $E_{\perp}$, but
599: this does not accelerate particles. However, in a real plasma,
600: the electrons have (i) a finite mass and therefore inertia, and (ii) a
601: finite thermal speed and therefore a pressure. Both of these
602: properties make parallel electric fields possible, which lead to the
603: dissipation of the wave energy by electron energization. The
604: importance of electron inertia in generating parallel electric
605: fields in the magnetosphere/ionosphere was first discussed by
606: \cite{1979JGR....84.7239G}.
607:
608: We follow here the definitions of \cite{2000SSRv...92..423S}, who give
609: an overview of dispersive Alfv\'en waves. An inertial Alfv\'en wave
610: (IAW) results if the electron thermal speed is smaller than or
611: comparable to the Alfv\'en speed. The electric field is due to the
612: finite inertia of the electrons, which cannot respond instantaneously
613: to the wave perturbation. (If the electron thermal speed exceeds the
614: Alfv\'en speed, but the electron pressure gradient is important, then
615: wave is termed a kinetic Alfv\'en wave, or KAW). Alternatively, the
616: conditions correspond to an IAW if $\beta \le m_e/m_p$ (and a KAW if
617: $m_e/m_p \le \beta \le 1$).
618:
619: The plasma $\beta$ is:
620: \begin{equation}\label{eq:beta}
621: \beta={2 \mu (n_p+n_h) k_B T \mu_o \over{\left|\mathbf{B}\right|^2}}
622: \end{equation}
623: where $k_B$ is Boltzmann's constant, $\mu_o$ the permeability of the
624: vacuum, $T$ is the temperature (we assume that the electron and ion
625: temperatures are equal), $\mu$ the mean molecular weight, $n_p$ the
626: proton number density and $n_h$ the neutral hydrogen number density.
627: Although the neutral hydrogen does not respond directly to the
628: Alfv\'enic disturbance, it is strongly collisionally coupled to the ion
629: component \citep[e.g.][]{2001ApJ...558..859D} and thus modifies the Alfv\'en
630: speed in the lower atmosphere. It also provides a mechanism for damping
631: the wave in the lower atmosphere, which will locally heat the
632: chromospheric plasma. Taking a mean molecular weight of 0.6, and
633: assuming a completely ionized target of density $n_e = n_{15}\times
634: 10^{15}{\rm m^{-3}}$ and temperature $T = T_6\times 10^6$~K, we have
635: \begin{equation}\label{eq:beta2}
636: \beta={2\times 10^{-8} {n_{15} T_6}\over{\left|\mathbf{B}\right| ^2}},
637: \end{equation}
638: for $\left|\mathbf{B}\right|$ in~T.
639: So, for example, if
640: $\left|\mathbf{B}\right| = 0.05$, $n_{15}=1$, $T_6 = 1$, $\beta = 8
641: \times 10^{-6}$, and the waves are inertial.
642: The inertial regime pertains
643: for substantial distances into the chromosphere (down to about
644: 1500~km above the photosphere in the VAL-C semi-empirical model).
645: Note that in other regions of the solar atmosphere, such as in long active region loops with a relatively small magnetic field, the KAW is appropriate, but not in the high magnetic field strength relevant to a flare.
646:
647: We have in mind an IAW disturbance with the form of a wave pulse or
648: simple wave, a case considered by
649: \cite{1994JGR....9911095K} and \cite{2007JGRA..11204214W}. However, acceleration
650: in IAWs is also discussed in the context of global resonances of the
651: magnetospheric field
652: \citep[e.g.,][]{2002JGRA.107gSMP19W,2003JGRA.108cSMP22W,2003GeoRL..30pSSC2W},
653: which could be established by repeated partial reflections of the IAW
654: from the photospheric or low chromospheric boundary. Evidently, the
655: exact nature of the oscillation will have to be determined in a
656: self-consistent way along with the particle acceleration.
657:
658: \subsection{The electric field strength and electron energy}\label{sect:field}
659:
660: Described in two-fluid MHD, a large-scale Alfv\'enic perturbation
661: causes particle cross-field drifts; an $\bf{E}\times\bf{B}$ drift
662: equal for both species, and a polarization drift. The ion
663: polarization drift is a factor $m_i/m_e$ faster than that of the
664: electrons, constituting a net cross-field current, the magnitude of
665: which depends on the wave amplitude at a given position. A
666: field-aligned current of electrons flows to maintain plasma
667: quasi-neutrality.
668:
669: From \cite{2000SSRv...92..423S}, their Equation 47, the relationship
670: between the perpendicular electric field $E_\perp$ and the change in
671: the perpendicular magnetic field $b_\perp$ is
672: \begin{equation}\label{eq:eperp}
673: E_\perp = v_A b_\perp \left(1+k_\perp^2\lambda_e^2\right)^{1/2},
674: \end{equation}
675: which is a modification of the ideal MHD relationship.
676: Here $k_\perp = 2\pi/\lambda_\perp$ is the
677: perpendicular wavenumber of the magnetic disturbance and $\lambda_e$
678: is the electron skin depth ($=c/\omega_{pe}$; $\omega_{pe}$ being the
679: electron plasma frequency).
680: In Eq.~\ref{eq:eperp} we have also used the fact that the
681: perpendicular scale of the magnetic disturbance is much larger than
682: the ion Larmor radius (see also Chaston et al. 2002).
683:
684: The relationship between the parallel and perpendicular components of
685: the wave is given by \cite{2000SSRv...92..423S}, Equation 43:
686: \begin{equation}\label{eq:epar}
687: E_\parallel = \left({k_\parallel k_\perp \lambda_e^2 }\over{1+k_\perp^2\lambda_e^2} \right) E_\perp
688: \end{equation}
689: where $k_\parallel$ is the parallel wavenumber. While in the magnetosphere, the ratio $ k_\perp^2\lambda_e^2 $ can be comparable to unity, in the solar atmosphere it is typically much less than unity. Therefore the ratio between parallel and perpendicular electric field in the solar atmosphere is going to be small in the solar atmosphere. However, since the perpendicular electric field calculated from Eq.~\ref{eq:eperp} is large, this small fraction can still result in a parallel field large enough to be interesting.
690: In the absence of precise knowledge about these scales we investigate
691: the parameter regimes in which substantial field-aligned electric fields might
692: be obtained.
693:
694: To be effective in accelerating electrons, $E_\parallel$ must exceed
695: the local Dreicer field, $E_D$
696: \citep{1959PhRv..115..238D,1982SSRv...31..351S}, above which the bulk
697: of the thermal electron distribution will be freely accelerated
698: (`runaway'). The Dreicer field, $E_D$, is
699: \begin{equation}\label{eq:dreicer}
700: E_D = {{e \ln \Lambda}\over{4\pi\epsilon_o\lambda_{D}}^2} = {{{e \ln \Lambda}\over{{4\pi\epsilon_o}^2}}
701: {{n e^2}\over{\epsilon_o k_B T}}},
702: \end{equation}
703: where $\ln \Lambda$ is the Coulomb logarithm, $\epsilon_o$ the
704: permittivity of free space and $\lambda_D$ is the Debye length. $\ln
705: \Lambda$ is usually taken to be between 20 and 25 for the corona.
706: In the partially ionized plasma of the lower chromosphere,
707: $\ln \Lambda$ is modified to $x \ln \Lambda + (1-x)\ln
708: \Lambda'$ where $x$ is the ionization fraction and $\ln \Lambda'$ the
709: `effective Coulomb logarithm' describing the interaction of charged
710: and neutral particles \citep[see e.g.,][]{1973SoPh...28..151B}.
711:
712: Neglecting the temperature-dependence of the Coulomb logarithm, the
713: ratio of the parallel electric field to the Dreicer field in the
714: corona is
715: \begin{equation} {E_\parallel \over E_D} = {10^5{T_6\over
716: n_{15}^{5/2}}{\left|\mathbf{B}\right| b_\perp \over{l_\parallel
717: l_\perp}}}
718: \end{equation}
719: where $l_\parallel,\; l_\perp$ the parallel and perpendicular
720: wavelengths in kilometers and $\left|\mathbf{B}\right|, b_\perp$ are
721: in T. Evidently, only in hot, tenuous, strongly-magnetized plasmas
722: will this ratio exceed unity; the ratio is
723: plotted in Figures~\ref{fig:eratio1} and~\ref{fig:eratio2}, where it
724: can be seen that at a coronal density of $10^{15}{\rm m^{-3}}$, ${E_\parallel \over E_D} $
725: exceeds unity only for scales $l_\parallel \sim 10 - 100$~km and
726: $l_\perp \leq 5$~km. Increasing the temperature, field strength, or
727: the perturbation amplitude, or decreasing the length-scale of the
728: perturbation gives a higher value for ${E_\parallel \over E_D} $. However,
729: wave-generated super-Dreicer fields are not possible in the chromosphere for
730: realistic parameters of the ambient medium or perturbation.
731:
732: \begin{figure}
733: \plotone{f4.eps}
734: \caption{The ratio of parallel electric field to Dreicer field for
735: field and perturbation values given in the top right corner of each
736: panel. The local coronal electron density is $10^{15}{\rm m}^{-3}$
737: and the temperature is $10^6$~K. The lines correspond to
738: $\lambda_\perp$=0.5~km (solid), 5~km (dotted), 50~km (dashed).}
739: \label{fig:eratio1}
740: \end{figure}
741:
742: \begin{figure}
743: \plotone{f5.eps}
744: \caption{As in Figure~\ref{fig:eratio1}, but with an electron density of
745: $5\times 10^{14}{\rm m}^{-3}$. }
746: \label{fig:eratio2}
747: \end{figure}
748:
749:
750: A full calculation of the electron energy spectrum accelerated must be
751: left for future investigations, as it requires a simulation capable of
752: following the non-linear evolution of the wave and of the electron
753: distribution function
754: \citep[e.g.,][]{2004PhPl...11.1277W,2005JGRA..11001201D}. But we can
755: observe that, in a corona of density $5\times 10^{14}{\rm m}^{-3}$,
756: super-Dreicer fields are produced in strong fields, by propagating
757: wave pulses having parallel wavelengths of around 100~km and
758: perpendicular wavelengths of around 5~km. Electrons with a thermal
759: speed similar to the wave phase speed can be accelerated, via a single
760: interaction with the traveling wave front, up to twice the Alfv\'en
761: speed \citep{chaston06} thus gaining 4 times the `Alfv\'en energy',
762: ${1\over 2} m_e v_A^2$, corresponding to 27~keV for $B = 0.05$~T, and
763: $n = 5\times 10^{14}{\rm m}^{-3}$. The maximum instantaneous electron
764: flux from a single interaction of electrons with the wave field is $n
765: v \sim 5\times 10^{14}\ {\rm m}^{-3} \times 2 \times 4.9 \times
766: 10^7{\rm m s^{-1}} = 4.9 \times 10^{22}{\rm m^{-2} s^{-1}}$ (this is
767: comparable with typical electron fluxes inferred from hard X-rays of
768: $10^{36}\ {\rm s}^{-1}$ over an area of perhaps $10^{13} - 10^{14}\
769: {\rm m^{2}}$). However, this flux will only be achieved if all
770: electrons are accelerated, which will not happen because of the
771: required velocity resonance condition of the electrons with the
772: wavefront. Thus, waves with scales of tens to hundreds of kilometers
773: may be capable of providing a modest flux of coronal electrons at 10
774: -- 30~keV, running ahead of the wave front.
775:
776: \subsection{First-order Fermi acceleration in a moving mirror}~\label{sect:fermi1}
777:
778: Further acceleration can occur via a first-order Fermi process as the
779: Alfv\'enic wave front, itself a moving mirror, approaches a magnetic
780: mirror in the lower corona and chromosphere. In repeated reflections,
781: the parallel electron speed would be increased by $2v_A$ at each
782: interaction, until the resulting decrease of pitch angle allows the
783: electron to penetrate the mirror. There
784: is thus the possibility to accelerate a fraction of the injected
785: electrons up to significantly higher energies.
786:
787: For repeated reflections, the mirroring electrons must not be
788: collisionally stopped between one interaction with the wavefront and
789: the next. So the separation in column depth between wave-front and
790: mirror must be less than half of the collisional stopping column depth
791: of the electrons at 2$v_A$ (neglecting the decreasing distance between
792: wave-front and mirror as the pulse approaches the
793: chromosphere). Using the expression from \cite{1978ApJ...224..241E},
794: the collisional stopping column depth of an electron of energy $E$ (in
795: keV) is:
796: \begin{equation}
797: N_c = 10^{21}\mu_e E^2{\;\rm m^{-2}}.
798: \end{equation}
799: where $\mu_e$ is the electron pitch-angle cosine. So for an electron
800: at 20~keV (i.e. following its first encounter with the wavefront),
801: with a pitch angle of 45$^o$, $N = \int n dl = 2.8\times 10^{23}{\rm
802: m^{-2}}$. The electron must therefore mirror within $1.4\times 10^{23}{\rm
803: m^{-2}}$. An underdense corona, of $n < 10^{15}{
804: \ \rm m^{-2}}$ with a loop half-length of $10^7$~m, has $N < 10^{22}\ {\rm
805: m^{-2}}$, so a 20~keV electron could penetrate some way into the
806: chromosphere - to a depth of around 1700~km above the photosphere in
807: the VAL-C chromospheric model (a column mass of
808: $2.3\times 10^{-4}\ {\rm kg\;m^{-2}}$). Thus, an electron could cross
809: the corona and chromosphere, mirror quite deep down and return for further
810: acceleration -- producing bremsstrahlung emission en route. The
811: details of this should be worked out in future.
812:
813: \subsection{Turbulent Acceleration and Heating in the
814: Chromosphere}\label{sect:turb}
815: We have seen that - with the possible exception of electron acceleration in their parallel electric field - Alfv\'en wave pulses will not dissipate significantly in the corona. This
816: leads us to consider the consequences when the wave reaches the chromosphere, and to discuss ways in which the wave energy could be damped there. It is well known that, in a strongly-magnetized atmosphere, it is not easy to damp Alfv\'en waves by straightforward collisional means, either by ion-electron (Joule) or ion-ion (viscous) collisions \citep[e.g.][]{1961ApJ...134..347O}. For this reason the dissipation of wave energy is normally thought to happen via a cascade process, with the energy ending up in wavelengths small enough for the Joule and viscous processes to be significant. (Note, ion-neutral damping probably is significant in the chromosphere and we return to this later). If such a cascade can develop, it will result in chromospheric heating, but possibly also electron acceleration. The
817: theory of stochastic electron acceleration
818: \citep[e.g.][]{1992ApJ...398..350H,1994ApJ...425..856L,1996ApJ...461..445M,1997ApJ...482..774P,2004ApJ...610..550P,2004ApJ...614..757Y,2006ApJ...644..603P}
819: provides a possible mechanism for the acceleration of electrons into a
820: broad spectrum extending to the high energies that are observed. There
821: is an extensive literature on such acceleration processes; the reader
822: is directed to \cite{2002SSRv1011A} (Section 5.2) for an
823: overview of the process. Here we will mention only some aspects pertinent to the
824: application of ideas of stochastic electron acceleration in this wave model
825: within the collisional environment of the chromosphere. Firstly we discuss briefly the generation of the cascade itself.
826:
827: As discussed in Section~\ref{sect:mechanism} it is reasonable to
828: expect that some fraction of the Alfv\'en mode energy that arrives at
829: the chromosphere will be reflected at the steep gradients within the
830: chromosphere or from the photosphere, and allow the development of a
831: turbulent spectrum in the counter-streaming wave field, with fast-mode
832: and Alfv\'en components. To be viable, this should happen quickly -
833: in less than the wave crossing-time of the chromosphere. There is a
834: vast literature on the development of magnetic turbulence, but
835: \cite{2004ApJ...614..757Y} provide useful expressions for the
836: relevant timescales. The Alfv\'en spectrum develops within the
837: turnover time of the longest wavelength present, $\lambda_{max}$,
838: i.e. $t =\lambda_{max}/\delta v$ \citep[see
839: also][]{1996ApJ...461..445M} where $\delta v/v_A = b_\perp/B$,
840: $\delta v$ being the velocity perturbation. So a (perpendicular)
841: cascade with energy injected at wavelengths less than $\lambda =
842: (b_\perp/B)$ times the height of the chromosphere can develop as the
843: Alfv\'en waves cross the chromosphere. The development of the
844: (isotropic) fast-mode spectrum, driven by reflected fast mode waves
845: \citep{1994A&A...286..275G} or fed by the Alfv\'en spectrum, depends
846: on the plasma $\beta$. In the high-$\beta$ medium of the low
847: chromosphere it develops in approximately
848: $t={(\lambda_{max}/v_A)(v_A/\delta_V)^2}$, so that only energy
849: injected at relatively short wavelengths will cascade quickly
850: enough. In the low-$\beta$ upper chromosphere the development is yet
851: slower.
852:
853: Damping by Fermi acceleration will dominate in the chromosphere, compared
854: to ion-viscous damping which may be significant in the corona. We demonstrate this by considering the
855: ratio of the ion-viscous damping rate to the Fermi damping rate, given by \citep[e.g.][]{2000SoPh..194..131T}:
856: \begin{equation}
857: {{\gamma_v}\over{\gamma_F}} ={{\tau_F}\over{\tau_{v}}} = 6 \times
858: 10^{11} {{kT_i^{5/2}}\over{vn_a}}
859: \end{equation} (converted into S.I. units) where $k$~is the
860: wavenumber, $T_i$~the ion temperature, $v$~the velocity, and $n_a$~the
861: density of particles accelerated by the Fermi mechanism. This ratio,
862: plotted in Figure~\ref{fig:tauftauiv} for a range of different
863: wavelengths of magnetoacoustic waves, is much less than one in the low
864: temperature chromosphere, primarily because of the strong temperature
865: dependence of the ion-viscous damping time. Therefore, chromosphere wavelengths
866: longer than about 1 meter will be preferentially damped by Fermi
867: acceleration \citep[see also][]{2006ApJ...644..603P} (however in the
868: corona ion-viscous damping, though weak, can still be dominant).
869:
870: \begin{figure}
871: \plotone{f6.eps}
872: \caption{Ratio of the Fermi and ion-viscous damping times in the
873: chromosphere, using a VAL-C model atmosphere. While this ratio may
874: be much larger than unity in the corona, implying that ion-viscous
875: damping dominates, it is less than unity throughout the
876: chromosphere.}
877: \label{fig:tauftauiv}
878: \end{figure}
879:
880: Electron acceleration by a turbulent wave spectrum has been mostly studied
881: in two main cases; `transit-time' acceleration by low-frequency fast
882: mode waves
883: \citep[e.g.][]{1997ApJ...491..939M,1998ApJ...493..451L,2002PhRvL..89B1102Y},
884: and gyroresonant interaction with a whistler spectrum - the high
885: frequency end of the Alfv\'en spectrum with $\omega > \Omega_i$
886: \citep[e.g.][]{1987SoPh..113..195M,1992ApJ...398..350H,2002PhRvL..89B1102Y,2004ApJ...610..550P}.
887: Of particular importance to us is the effect in these models of
888: Coulomb collisions: the dense chromosphere might be thought of as
889: unfavorable for any particle acceleration to exist since energy gained
890: can be quickly lost again. Some modeling has considered Coulomb
891: energy losses and isotropization \citep[e.g.,][]{1992ApJ...398..350H,
892: 1998ApJ...493..451L,2004ApJ...614..757Y}. In general, one finds that
893: below the electron energy at which the acceleration timescale exceeds
894: the collisional loss timescale, the electron distribution is
895: quasi-thermal. Above this critical energy the distribution can have a
896: non-thermal character as the electrons become increasingly
897: collisionless at high energy. For whistler-mode acceleration, the
898: critical energy is
899: \begin{equation}
900: E_c \sim 3.8{n_{16}}^{3/2}\left({0.01T\over {\left|\mathbf{B}\right|}}\right)^2
901: \left({10^{-4}\over R} \right){\rm keV}
902: \end{equation}
903: where $R$ is the ratio of turbulent magnetic energy density to total
904: magnetic energy density \citep[][Eq. 20]{1992ApJ...398..350H}. For
905: chromospheric parameters of $n_{16} = 100,\; \left|\mathbf{B}\right| =
906: 0.05$ then $E_c = 0.015/R$~keV. If the turbulent energy density fraction
907: contained in whistlers is $R \sim 10^{-3}$ then the electron
908: distribution will be non-thermal above 15~keV. It remains to be seen
909: whether this level of whistler turbulence is plausible.
910:
911: In the case of transit-time acceleration, which operates at much lower
912: wave frequencies, \cite{1998ApJ...493..451L} find that energy exchange
913: between waves and particles is in fact made significantly more
914: efficient in the presence of Coulomb interactions. This is because
915: Coulomb collisions (i) exchange energy between accelerated and
916: non-accelerated electrons, raising the slower electrons up to resonant
917: energies, and (ii) redistribute the energy gained between parallel and
918: perpendicular components of momentum, increasing the magnetic moment
919: of the electrons and thus the rate of the transit-time process.
920: Transit-time damping by electrons requires that the local electron
921: thermal speed be comparable to the Alfv\'en speed, equivalent to
922: $\beta \sim{m_e}/{m_p}$. Using the VAL-C model, this occurs at around
923: 1500~km above the photosphere, where the density is $\sim 10^{18}\
924: {\rm m^{-3}}$. It also requires that the wave spectrum be continuous
925: (as in a turbulent spectrum), or at least have discrete overlapping
926: modes to allow electrons to stay in resonance as they accelerate.
927:
928: It should be noted that the simulations of \cite{1998ApJ...493..451L}
929: are done for a temperature of $3\times 10^6$~K and a density of
930: $10^{16}\ {\rm m^{-2}}$. It remains to be seen whether the beneficial
931: trade-off between energy loss and scattering will occur at higher
932: densities, though since both scattering and loss terms in the
933: Fokker-Planck equation describing the evolution of the particle
934: distribution function have the same density dependence (see e.g.
935: Lenters \& Miller 1998 Eq. 4) we expect that it will.
936:
937: However, even with the enhanced efficiency provided by Coulomb
938: collisions, transit-time acceleration does not yield a power-law
939: distribution as is observed from hard X-rays -- instead it produces
940: `bulk heating' of electrons, albeit to energies of 10s of keV. Conceivably,
941: a low level of whistler turbulence could provide the necessary pitch-angle
942: scattering (but without energy redistribution) leading to the formation of an
943: accelerated non-thermal tail.
944:
945: \section{Overall Energetics and Open Questions}\label{sect:energetics}
946: We have introduced the idea of impulsive-phase transport of flare energy
947: from its initial site of energy release via Alfv{\'e}n wave pulses,
948: and in the previous section have shown how this may
949: lead to the electron acceleration needed to explain the
950: hard X-ray observations. A complete theory should also address the
951: generation of the wave energy in the first place, discuss the
952: efficiency of the conversion, and describe the regulation mechanisms
953: that allow the hard X-ray signatures to be so universal.
954:
955: The partition of energy at its original source poses the first
956: important question: what fraction goes into the Alfv{\' e}n mode and
957: what fraction goes into other wave modes? Emslie and Sturrock (1982)
958: deal with this question qualitatively and suppose that half of the
959: energy winds up in the Alfv{\' e}n mode and the other half in the fast
960: mode, with the slow mode getting a negligible amount because of the
961: mismatch between the sound speed and the Alfv{\' e}n speed. To obtain
962: a better understanding of this energy partition would require a full
963: understanding of the non-linear development of the energy release,
964: thus determining the flow fields involved in the deformation of the
965: magnetic field. In a low-$\beta$ plasma one would expect this
966: deformation to proceed at or near the Alfv{\' e}n speed.
967:
968: The next step in the flow of energy consists of the Poynting flux $S$
969: of the resulting waves, with $S \sim v_A \times b_\perp^2 / \mu_o$.
970: The magnitude of the wave field $b_\perp$ can be crudely estimated
971: from the requirement that this Poynting flux supply the flare energy.
972: \cite{2007ApJ...656.1187F} show that the broadband flare output in
973: moderate white-light and UV events, occurring in small footpoint
974: areas, corresponds to an energy input in excess of $ S \sim 10^{7}\
975: {\rm J m^{-2} s^{-1}}$. For an X-class flare energy dissipation of
976: $10^{25}$ Joules in $10^3$ seconds, over a spatial footpoint scale of
977: $(10^4~\rm{km})^2$, we need $S \sim 10^8 {\rm J\ m^{-2} s^{-1}}$. For
978: $v_{A} \lesssim 1 \times 10^{4}\;{\rm km\ s^{-1}}$ at the
979: chromospheric formation depth of the broad-band emission, then $|{\bf
980: b_\perp}| \gtrsim$~=0.003~T. This is well within the upper limit to
981: plausible field variations, given by the permanent line-of-sight field
982: changes observed at the photosphere in large flares.
983:
984: Other areas of theoretical uncertainty involve the degree of
985: reflection of the wave energy on the gradients at and below the
986: transition region, and the related question regarding the growth rate
987: of the turbulent cascade. In the lower atmosphere the Alfv\'en speed
988: varies over a scale short compared to the wavelength of the
989: disturbance, so the disturbance will be partially reflected and
990: partially transmitted (though the fact that stepwise photospheric
991: field changes of order ten percent are seen suggests that a
992: considerable fraction of wave energy is transmitted to the
993: photospheric level). Emslie \& Sturrock (1982) discuss wave
994: transport and dissipation in the context of a normal solar atmospheric
995: model, in which thermal conduction creates a sharp transition
996: layer. In this case substantial wave reflection will occur, launching a
997: propagating wave towards the conjugate footpoint. The a
998: coefficient of reflection is given by $R_{E} =
999: {{(\theta}^{1/2}-1)^{2}/{(\theta^{1/2} + 1)^{2}}}$ where $\theta$
1000: represents the temperature ratio between corona and chromosphere. For
1001: a quiet solar atmosphere we might have $\theta$ = 200 and $R_{E} \sim
1002: $75\%, but clearly in a flaring atmosphere this estimate will have to
1003: be modified and will affect the wave energy reaching the chromosphere.
1004: Strong heating should increase the scale height, soften the transition region
1005: and reduce the reflected component. In the radiative hydrodynamic models of \cite{2005ApJ...630..573A}, the density and temperature gradients between chromosphere and corona are indeed
1006: at first on average smoothed out by atmospheric heating in the impulsive phase,
1007: but then steeper temperature gradients occur as the corona heats.
1008: But the behavior also varies with the intensity, and location of heat input, which
1009: depends of course on the energy transport model and atmospheric structure, and will need to be examined in detail.
1010:
1011: The energy of the transmitted fraction will be dissipated in the
1012: chromosphere. Alfv\'enic
1013: disturbances can damp resistively, if on small enough scales, or by
1014: other means such as ion-neutral coupling which may be particularly
1015: important in the lower chromosphere. De Pontieu et al. (2001)
1016: considered the damping by ion-neutral coupling in the lower
1017: chromosphere of large-scale coronal oscillations, observed in TRACE to
1018: be excited by flares and filament eruptions \citep{2002SoPh..206...69S}.
1019: Although these waves are kink (fast mode) waves in flux tubes with
1020: relatively low fields, analogous damping may occur for our Alfv{\' e}n
1021: mode waves in strong field regions. The Joule dissipation as
1022: calculated by Emslie \& Sturrock is enhanced by a factor $(1+s)$ where
1023: $s$ is the ``ion slip'' term;
1024: \begin{equation}
1025: s = \left({\rho_n\over
1026: \rho_t}\right)^2 {{\Omega_e \Omega_i}\over{\nu_{\it eff}
1027: \nu_{in}}};
1028: \end{equation}
1029: here $\Omega_e$ and $\Omega_i$ are the electron and ion
1030: gyrofrequencies, $\nu_{\it eff} = \nu_{ei}+\nu_{en}$, the collision
1031: frequencies of electrons on ions and neutrals, respectively, and
1032: $\nu_{in}$ is the ion-neutral collision frequency. De Pontieu et al.
1033: found the slip $s$ to be large throughout the chromosphere, resulting
1034: in Joule heating that peaks between around 300~km and 1000~km above
1035: the photosphere. This is close to the temperature minimum region where
1036: localized energy input is required to generate the observed
1037: white-light flare continuum excess.
1038:
1039: Finally, any remaining undamped waves, once reflected at the photosphere
1040: or at strong chromospheric gradients may lead to the development of a turbulent
1041: cascade which, as we have noted, provides one of the major
1042: possibilities for chromospheric electron acceleration. Again a quantitative
1043: description of this partitioning is beyond the scope of this paper.
1044:
1045: \section{Conclusions}
1046:
1047: Energy transport by Alfv{\' e}n waves has a well-developed literature
1048: in the context of the terrestrial aurora, and we have applied similar
1049: ideas here to the problem of flare effects in the solar atmosphere.
1050: Our new understanding of active-region magnetic fields, based on
1051: microwave observations, now convinces us that the transport time for
1052: these waves is very short -- short enough to explain the rapid time
1053: variations and tight conjugacy of double-footpoint hard X-ray sources
1054: -- and also that the energy flux can be very large. From this point
1055: of view, Alfv{\' e}n waves therefore provide an alternative to energy
1056: transport by electron beams. Emslie \& Sturrock aimed at explaining a
1057: relatively weak warming of the temperature-minimum region late in the
1058: flare, as required by Ca~K line observations of Machado et
1059: al. (1978). We instead wish to explain the entire energy of the flare
1060: impulsive phase in this manner.
1061:
1062: Replacing the electron beam of the standard thick-target model with an
1063: Alfv{\' e}n-wave Poynting flux implies particle acceleration in the
1064: chromosphere or at the base of the coronal loop carrying the wave.
1065: Because of the dominance of fast electrons in the flare energy budget,
1066: we have discussed mechanisms for electron acceleration in this
1067: scenario at length. Our analysis establishes the feasibility of these
1068: ideas without pinpointing which of the possible acceleration modes
1069: dominates.
1070:
1071: Finally, we note that the ideas we present are novel in the solar
1072: context but are well-established in the Earth's magnetosphere. These
1073: ideas should be considered not only for solar flares, but elsewhere
1074: in the Universe where magnetic reconnection is invoked.
1075:
1076: \acknowledgments
1077:
1078: We would like to acknowledge many useful discussions with numerous
1079: colleagues, including J.~C.~Brown, P.~Cargill, C.~Chaston, P.~Damiano,
1080: N.~Gopalswamy, G.~Haerendel, J.~I.~Khan, E.~P. Kontar,
1081: A.~L.~MacKinnon, J.~P.~McFadden and K.~Shibata. We would also like to
1082: thank an anonymous referee for vigorous discussion which led to
1083: significant improvements in the overall structure and clarity of the
1084: paper. This work was supported by NASA under grants NNG05GG17G
1085: (L. F. and H. S. H) and NAS 5-98033 (H. S. H.), and by PPARC under
1086: Rolling Grant PP/C000234/1 (L.F.). Financial support by the European
1087: Commission through the SOLAIRE Network (MTRN-CT-2006-035484) is
1088: gratefully acknowledged.
1089:
1090: \bibliographystyle{apj}
1091: \bibliography{ms}
1092:
1093: \end{document}
1094:
1095: