1: \documentclass[preprint2]{aastex}
2: %\documentclass[12pt,preprint]{aastex}
3: %\documentclass[]{aastex}
4:
5: \newcommand{\lapprox} {\, \lower3pt\hbox{$\sim$}\llap{\raise2pt\hbox{$<$}}\,}
6: \newcommand{\gapprox} {\, \lower3pt\hbox{$\sim$}\llap{\raise2pt\hbox{$>$}}\,}
7:
8: %\shorttitle{Electron-Electron Bremsstrahlung Emission}
9: %\shortauthors{}
10:
11: \begin{document}
12:
13: \title{Electron-Electron Bremsstrahlung Emission and the Inference of Electron Flux Spectra in Solar Flares}
14:
15: \author{Eduard P. Kontar\altaffilmark{1}, A. Gordon Emslie\altaffilmark{2}, Anna Maria Massone\altaffilmark{3},
16: Michele Piana\altaffilmark{4}, John C. Brown\altaffilmark{1},
17: and Marco Prato\altaffilmark{5}}
18:
19: \altaffiltext{1}{Department of Physics \& Astronomy, University of
20: Glasgow G12 8QQ, UK \\ {eduard@astro.gla.ac.uk,
21: john@astro.gla.ac.uk}}
22:
23: \altaffiltext{2}{Department of Physics, Oklahoma State University, Stillwater, OK 74078 \\
24: gordon.emslie@okstate.edu}
25:
26: \altaffiltext{3}{CNR-INFM LAMIA, via Dodecaneso 33, I-16146
27: Genova, Italy(massone@ge.infm.it)}
28:
29: \altaffiltext{4}{Dipartimento di Informatica, Universit\`a di
30: Verona, Ca\`\ Vignal 2, Strada le Grazie 15, I-37134 Verona, Italy
31: (michele.piana@univr.it)}
32:
33: \altaffiltext{5}{Dipartimento di Matematica, Universit\`a di
34: Modena e Reggio Emilia, via Campi 213/b, I-41100 Modena, Italy
35: (marco.prato@unimo.it)}
36:
37: \begin{abstract}
38:
39: Although both electron-ion and electron-electron bremsstrahlung
40: contribute to the hard X-ray emission from solar flares, the
41: latter is normally ignored. Such an omission is not justified at
42: electron (and photon) energies above $\sim 300$~keV, and inclusion
43: of the additional electron-electron bremsstrahlung in general
44: makes the electron spectrum required to produce a given hard X-ray
45: spectrum steeper at high energies.
46:
47: Unlike electron-ion bremsstrahlung, electron-electron
48: bremsstrahlung cannot produce photons of all energies up to the
49: maximum electron energy involved. The maximum possible photon
50: energy depends on the angle between the direction of the emitting
51: electron and the emitted photon, and this suggests a diagnostic
52: for an upper cutoff energy and/or for the degree of beaming of the
53: accelerated electrons.
54:
55: We analyze the large event of January 17, 2005 and show that the
56: upward break around 400~keV in the observed hard X-ray spectrum is
57: naturally accounted for by the inclusion of electron-electron
58: bremsstrahlung. Indeed, the mean source electron spectrum
59: recovered through a regularized inversion of the hard X-ray
60: spectrum, using a cross-section that includes both electron-ion
61: and electron-electron terms, has a relatively constant spectral
62: index $\delta$ over the range from electron kinetic energy $E =
63: 200$~keV to $E = 1$~MeV. However, the level of detail discernible
64: in the recovered electron spectrum is not sufficient to determine
65: whether or not any upper cutoff energy exists.
66:
67: \end{abstract}
68:
69: \keywords{processes: radiation; Sun: flares; Sun: X-rays}
70:
71: \section{Introduction}
72:
73: The spatially integrated hard X-ray spectrum $I(\epsilon)$
74: (photons~cm$^{-2}$~s$^{-1}$~keV$^{-1}$ at the Earth) is produced
75: by bremsstrahlung of accelerated electrons, characterized (Brown,
76: Emslie, \& Kontar 2003) by a {\it mean electron flux spectrum}
77: ${\overline F}(E)$ (electrons~cm$^{-2}$~s$^{-1}$~keV$^{-1}$ at the
78: Sun) and related to $I(\epsilon)$ through
79:
80: \begin{equation}
81: I(\epsilon) = {1 \over 4\pi R^2} \, \, {\bar n} V
82: \int_\epsilon^\infty \overline F(E) \, Q(\epsilon,E) \, dE,
83: \label{def}
84: \end{equation}
85: where $Q(\epsilon,E)$ is the bremsstrahlung cross-section
86: (cm$^2$~keV$^{-1}$) differential in photon energy, $R = 1$~AU, and
87: the mean target density $\bar n$ (cm$^{-3}$) is defined by $\bar n
88: = V^{-1} \int n({\bf r}) \, dV$. Bremsstrahlung in the energy
89: range $\gapprox 10$~keV is produced by energetic electrons
90: scattering off both protons/ions and electrons (both free and bound in atoms); these
91: contributions are summed to give the total differential bremsstrahlung
92: cross-section $Q(\epsilon,E)$.
93:
94: For electron energies $\lapprox 300$~keV, the contribution from
95: electron-electron bremsstrahlung can be safely ignored (Haug
96: 1975). However, for higher energies this is no longer the case.
97: Generally, for a given electron spectrum, the additional
98: electron-electron bremsstrahlung yield acts to flatten (harden)
99: the photon spectrum in this energy range (see, e.g., Haug 1975).
100: Equivalently, the inclusion of electron-electron bremsstrahlung
101: requires, for a given photon yield, a softer (steeper) electron
102: spectrum than would be required assuming electron-ion
103: bremsstrahlung alone.
104:
105: The {\it Ramaty High Energy Solar Spectroscopic Imager} ({\em
106: RHESSI}) has opened a new era in the study of hard X-ray spectra
107: from solar flares. With the high-resolution hard X-ray spectra
108: made available by {\it RHESSI}, an investigation into the form of
109: the hard X-ray spectrum above $\sim 300$~keV with the full
110: (electron-ion + electron-electron) cross-section is now warranted.
111:
112: In this paper we therefore study the effect of adding the
113: electron-electron bremsstrahlung term on the mean
114: electron spectrum corresponding to a given hard X-ray spectrum. In
115: \S2 we discuss the forms of the electron-ion and electron-electron
116: cross-sections, and we show that the presence of an upper limit to
117: the photon energy in the electron-electron process can in
118: principle provide evidence for a high-energy cutoff in the
119: electron spectrum and/or evidence of strong anisotropy in the
120: injected electron distribution. In \S3 we discuss the sample
121: event (on January 17, 2005) chosen for analysis. In \S4 we
122: present the form of the electron spectrum corresponding to the
123: observed photon spectrum, using both forward-fitting (e.g, Holman
124: et al. 2003) and regularization (Piana et al. 2003) techniques in
125: conjunction with the full (electron-ion + electron-electron)
126: bremsstrahlung cross-section. In \S5 we discuss the results
127: obtained, and in particular we point out that certain features in
128: the electron spectrum inferred using the electron-ion
129: bremsstrahlung cross-section are artifacts that can vanish when
130: the full, correct, cross-section is employed.
131:
132: \section{Form of the Bremsstrahlung Cross-Section}
133:
134: The cross-section (e.g., Koch \& Motz 1959) for electron-ion
135: bremsstrahlung scales as $Z^2$, where $Z$ is the atomic number of
136: the ion. Further, in consideration of electron-electron
137: bremsstrahlung, the possible binding of target electrons to their
138: host ions in a neutral or partially-ionized medium is not
139: significant (E. Haug, personal communication). Hence, in a
140: quasi-neutral target of particles with atomic number $Z$, the
141: bremsstrahlung cross-section per atom for emission of a photon of
142: energy $\epsilon$ by an electron of energy $E$ is in general equal
143: to
144:
145: \begin{equation}
146: Q (\epsilon, E) = Z^2 Q_{e-p} (\epsilon, E) + Z Q_{e-e} (\epsilon,
147: E). \label{qtot}
148: \end{equation}
149: were $Q_{e-p} (\epsilon, E)$ and $Q_{e-e} (\epsilon, E)$ are the
150: cross-sections for bremsstrahlung in electron-proton, and
151: electron-electron collisions, respectively.
152:
153: The form of $Q_{e-e}(\epsilon,E)$, averaged over solid angle {\it
154: in the rest frame of the target electron} has been given\footnote
155: {Note that in the formula for $H(\epsilon,k,x)$ for the case $k >
156: {1 \over 2}$ (page 347 of that paper), the term $(\epsilon r /x +
157: s)$ on line 3 of the equation should be replaced by $(\epsilon r
158: /w + s)$ -- E. Haug, personal communication.} by Haug (1998),
159: while the solid-angle-averaged form in the {\it zero-momentum}
160: (``center-of mass'') frame has been given by Haug (1989). Neither
161: of these formulae are strictly appropriate to the case of an
162: electron beam incident on a warm plasma: the target electrons,
163: unlike the ions, have a velocity that may be comparable to the
164: velocity of the electrons in the impinging beam, so that a range
165: of injected particle/target particle relative velocities exist for
166: a given injected electron energy. However, as verified through
167: numerical simulation, for low electron energies (from $\sim 10$ to
168: $\sim 200$~keV), the form of ${\overline F}(E)$ corresponding to a
169: given hard X-ray spectrum, inferred using the
170: cross-section~(\ref{qtot}), does not differ, within statistical
171: uncertainties in the photon flux, from the form of ${\overline
172: F}(E)$ obtained using the electron-ion cross-section alone. Hence,
173: only at electron energies $\gapprox 200$~keV is the inclusion of
174: the electron-electron bremsstrahlung term necessary, and, for such
175: energies the velocity of the target particles {\it is} relatively
176: insignificant. The target particle rest frame is, therefore, a
177: better approximation to the observer frame than is the
178: zero-momentum frame. Hence, use of the electron-electron
179: cross-section in the {\it target particle} rest frame is more
180: appropriate.
181:
182: When the maximum electron energy is much larger than the photon
183: energies under consideration, the photon spectrum resulting from a
184: power-law spectrum of electrons ${\overline F}(E) \sim
185: E^{-\delta}$ is also close to power-law form $I(\epsilon) \sim
186: \epsilon^{-\gamma}$ (Haug 1989). However, while for pure
187: electron-ion bremsstrahlung $\gamma \simeq \delta+1$, for pure
188: electron-electron bremsstrahlung a significantly shallower photon
189: spectrum, with $\gamma \simeq \delta$, results. Thus, the
190: importance of the electron-electron bremsstrahlung contribution
191: increases with photon energy and the enhanced emission per
192: electron leads to a flattening of the photon spectrum
193: $I(\epsilon)$ produced by a given ${\overline F}(E)$, or,
194: equivalently, a steepening of the ${\overline F}(E)$ form required
195: to produce a given $I(\epsilon)$.
196:
197: \begin{figure}[pht]
198: \epsscale{0.99} \plotone{f1.eps} \caption{Maximum photon energy
199: $\epsilon_{\rm max}$ produced by electron-electron bremsstrahlung,
200: expressed as a fraction of the incident electron energy $E$ (in
201: units of the electron rest mass $mc^2$), for various values of
202: $\theta$, the angle between the incoming electron and the outgoing
203: photon trajectories For clarity, only curves for $\theta =
204: 0^\circ, 60^\circ, 120^\circ,$ and $180^\circ$ are labelled; the
205: curves for $\theta = 30^\circ, 90^\circ$ and $150^\circ$ lie
206: between these.} \label{epsmax}
207: \end{figure}
208:
209: It is also important to note that while the electron-ion
210: cross-section is finite for all $\epsilon < E$, the ``laboratory
211: frame'' cross-section for electron-electron bremsstrahlung
212: vanishes above a maximum photon energy, due to the necessarily
213: finite energy carried by the recoiling target electron.
214: Quantitatively (Haug 1975),
215: %\begin{equation} \epsilon _{\rm max} =mc^2 \, \frac{\psi}{1-\sqrt
216: %\psi \, \cos \theta}, \quad \psi=\frac E{E+2mc^2} \label{emax}
217: %\end{equation}
218:
219: \begin{equation} \epsilon _{\rm max} = \frac{E}{E + 2 -\sqrt{E(E+2)}
220: \, \cos \theta},\label{emax}
221: \end{equation}
222: where $E$ is the electron kinetic energy in the laboratory frame (in units
223: of the electron rest energy $mc^2$) and $\theta$ is the angle
224: between the incoming electron and the outgoing photon
225: trajectories. For highly {\it non}-relativistic electrons ($E \ll
226: 1$), $\epsilon_{\rm max} \rightarrow E/2$ for all values of
227: $\theta$. Only for highly relativistic electrons ($E \gg 1$) and
228: $\theta = 0$ (a singular case) does $\epsilon_{\rm max}
229: \rightarrow E$; for all other situations $\epsilon_{\rm max}$ is
230: less than $E$ and approaches $0$ as $E \rightarrow \infty$
231: (Figure~\ref{epsmax}).
232:
233: This result has important implications for the form of the photon
234: spectrum produced by electron-electron bremsstrahlung. If the
235: electron spectrum ${\overline F}(E)$ has a maximum energy $E_{\rm
236: max}$ that is not highly relativistic, then while electron-ion
237: bremsstrahlung will generate photons at all energies up to $E_{\rm
238: max}$, electron-electron bremsstrahlung will produce no photons at
239: all in the range $\epsilon_{\rm max} < \epsilon \le E_{\rm max}$.
240: The entire spectrum above $\epsilon_{\rm max}$ will therefore be
241: produced completely by electron-ion bremsstrahlung; the flattening
242: of the photon spectrum associated with the electron-electron
243: contribution disappears and the relationship between $I(\epsilon)$
244: and ${\bar F}(E)$ reverts to the form appropriate to electron-ion
245: bremsstrahlung alone.
246:
247: \begin{figure}[pht]
248: \epsscale{1.0} \plottwo{f2a.eps}{f2b.eps}\caption{{\it Left
249: panel:} Ratio of the total photon spectrum to that produced by
250: electron-ion bremsstrahlung alone, for four different viewing
251: angles $\theta$. The mean source electron spectrum is a power law
252: with spectral index $\delta =2$ and an upper cutoff energy at
253: 1~MeV; {\it Right panel:} Corresponding local spectral indices
254: $\gamma(\epsilon) = -d \log I(\epsilon)/d \log \epsilon$. Note
255: the sharp features in $\gamma$ caused by the absence of
256: electron-electron emission above a certain photon energy
257: $\epsilon_{\rm max}$ (see Figure~\ref{epsmax}). For clarity, only
258: select curves have been labelled in both panels.} \label{theta}
259: \end{figure}
260:
261: Note also that the maximum photon energy $\epsilon _{\rm max}$
262: depends significantly on the viewing angle $\theta$. Hence, if the
263: injected electron distribution is highly beamed, the strong
264: angular dependence of the maximum photon energy produced permits a
265: determination of the direction of the beam. For example, for
266: $E_{\rm max}=1$~MeV and $\theta =90^\circ$, there should be
267: evidence for such a high-energy cutoff in the photon spectrum
268: around 250~keV (Figure~\ref{epsmax}). Figure~\ref{theta} shows the
269: effect of such a 1~MeV upper energy cutoff on the total
270: (electron-ion + electron-electron) photon spectrum $I(\epsilon)$
271: and on its local spectral index $\gamma = -d \log I(\epsilon)/d
272: \log \epsilon$. There is an abrupt step in $\gamma$ at $\sim
273: 250$~keV; this step moves towards larger energies as $\theta$ is
274: reduced, so that the inferred value of $E_{\rm max}$ depends on
275: the value of $\theta$ appropriate.
276:
277: \section{Data analysis}
278:
279: In selecting suitable events for analysis, we searched for a clear
280: identification of high-energy photons in the flare light curve,
281: and specifically a count rate high enough to provide good count
282: statistics in energy channels above 300~keV. Quasi-logarithmic
283: energy binning was used in order to enhance the signal-to-noise
284: ratio in each energy channel and the time bins were chosen equal
285: to RHESSI's rotation period (as given for the time of the flare)
286: to ensure that there is no differential modulation of the light
287: curve from varying aspects of the imaging grids.
288:
289: The data were corrected for the following effects: decimation,
290: detector energy response, detector livetime, attenuator
291: transmission, imaging grid transmission, and pulse pile-up. These
292: steps were performed using standard software incorporating the
293: most up to date information on the instrumental calibration
294: (Schwartz et al. 2002). The background was then subtracted by
295: using the SPEX package to interpolate between two background time
296: intervals, one before and one after the flare. Data from detectors
297: 2 or 7 were not used, because their energy resolution is
298: significantly poorer than for the other detectors (Smith et~al.
299: 2002).
300:
301: \begin{figure}[pht]
302: \epsscale{0.99} \plotone{f3.eps} \caption{Photon spectrum for the
303: time interval 09:43:16 -- 09:44:24 UT in the 2005 January 17 event
304: with gamma-lines removed.} \label{photon}
305: \end{figure}
306:
307: Figure~\ref{photon} shows the photon spectrum for the time
308: interval 09:43:16 -- 09:44:24 UT (the time of approximate peak
309: flux) for the 2005 January 17 (GOES class X3.8) event. This event,
310: which produced several strong gamma-ray lines, was previously
311: studied by Kontar \& Brown (2006), who concluded that the pitch
312: angle distribution for electrons up to $\sim 300$~keV is close to
313: isotropic. We focus attention on the highest energy spectrum
314: ($\epsilon > 200$~keV) in this paper.
315:
316: This event was located at position $(x=380\arcsec, y=320\arcsec)$
317: on the solar disk, corresponding to a heliocentric angle $\sim
318: 30^\circ$. Consequently, the assumption of a downward-directed
319: electron beam leads to angles $\theta$ between the beam direction
320: and the observer in the second quadrant; this enhances the
321: possibility of observing the spectral features noted in \S2
322: associated with the upper limit to electron-electron
323: bremsstrahlung emission (see Figures~\ref{epsmax} and
324: \ref{theta}).
325:
326: \section{Determination of the Mean Source Electron Spectrum}
327:
328: Before attempting to determine the form of the mean source
329: electron spectrum responsible for the observed hard
330: X-ray/gamma-ray continuum, it is first necessary to subtract the
331: emission from strong gamma-ray spectral lines. In the energy
332: range under consideration, the two most significant ranges for
333: which this subtraction is necessary are (483-512)~keV and
334: (829-882)~keV. The corrected spectrum is presented in
335: Figure~\ref{photon}. The first of these corresponds to the
336: electron-positron annihilation line at 511~keV and its associated
337: positronium continuum at lower energies, the second to a variety
338: of strong emission lines from $^{27}$Al, $^{54}$Cr and $^{56}$Fe
339: (see Table~1 in Ramaty, Kozlovsky, \& Lingenfelter 1979; Table~1
340: in Kozlovsky, Murphy, \& Ramaty 2002). These lines were
341: ``removed'' by replacing the data in these ranges with a smooth
342: interpolation of the continuum spectrum on either side of each
343: feature.
344:
345: The residual photon spectra then represent principally
346: bremsstrahlung continuum, with an emissivity given by
347: equation~(\ref{def}). These continuum spectra were then used to
348: determine the mean electron flux spectrum ${\overline F}(E)$ in
349: the source, using two different, well-established, methodologies
350: for solution of equation~(\ref{def}).
351:
352: \subsection{Forward Fit}
353:
354: Here we follow the procedure of Holman et~al. (2003) and assume
355: that the mean electron spectrum is the sum of a low-energy
356: Maxwellian, plus a broken power law of the form
357:
358: \begin{equation}
359: {\overline F}(E) = \cases{A \, E^{-\delta_1}; \quad E < E_{\rm
360: brk} \cr A \, E_{\rm brk}^{\delta_2 - \delta_1} \, E^{-\delta_2};
361: \quad E \ge E_{\rm brk}.} \label{broken}
362: \end{equation}
363: Because the Maxwellian part of ${\overline F}(E)$ (with a
364: characteristic temperature $T \simeq 3$~keV) is utterly
365: insignificant at energies $E > 200$~keV, it is not necessary to
366: consider this component in our analysis.
367:
368: Calculation, using equation~(\ref{def}), of the photon spectrum
369: for an ${\overline F}(E)$ of the form~(\ref{broken}), and
370: comparison with the observed $I(\epsilon)$ above $\epsilon =
371: 200$~keV, permits determination of the best-fit values of the four
372: parameters $(A, E_{\rm brk}, \delta_1, \delta_2)$. We performed
373: such a forward fit for two forms of the bremsstrahlung
374: cross-section: $Q_{\rm e-i}(\epsilon, E) = Z^2 Q_{e-p} (\epsilon,
375: E)$ (which includes electron-ion bremsstrahlung only) and $Q_{\rm
376: tot}(\epsilon, E) = Z^2 Q_{e-p} (\epsilon, E) + Z Q_{e-e}
377: (\epsilon, E)$ (which includes both electron-ion and
378: electron-electron bremsstrahlung). Mean values $<Z> = 1.2$ and
379: $<Z^2> = 1.44$ (representative of mean solar abundances) were
380: assumed.
381:
382: \begin{figure}[pht]
383: \epsscale{0.99} \plotone{f4.eps} \caption{Forward-fit forms of
384: $\bar n \, V \, {\overline F}(E)$ for the selected flare. The
385: dashed curve assumes electron-ion bremsstrahlung only; the solid
386: curve includes the additional electron-electron term.
387: \label{ffit}}
388: \end{figure}
389:
390: Using the cross-section $Q_{\rm e-i}(\epsilon,E)$, representing
391: only electron-ion bremsstrahlung, results in best-fit values
392: $\delta_1 = 3.4,\; \delta_2 = 2.9$, and $E_{\rm brk} = 445 $~keV.
393: Using the more correct cross-section $Q_{\rm tot}(\epsilon,E)$
394: (which incorporates both electron-ion and electron-electron terms)
395: gives $\delta_1 = 3.5, \delta_2 = 3.1 $, and $E_{\rm brk} =
396: 431$~keV. The forms of both of these fits are shown in
397: Figure~\ref{ffit}. While inclusion of the electron-electron
398: bremsstrahlung term results in little change to the form of
399: ${\overline F}(E)$ at low energies, its inclusion does lead to the
400: break energy moving downward from $E \sim 450$~keV to $E \sim
401: 430$~keV, and to the spectral index for the high-energy component
402: steepening from $\delta \simeq 2.9$ to $\delta \simeq 3.1$
403: ($\Delta \delta \simeq 0.2$). Such a steepening of ${\overline
404: F}(E)$, and the energy above which it becomes significant, are in
405: accordance with the expectations expressed in \S1 and with earlier
406: quantitative estimates based on hardening of hard X-ray spectra
407: (e.g., Vestrand 1988).
408:
409: \subsection{Regularized Inversion}
410:
411: Piana et al. (2003) have demonstrated the construction of smooth, {\it
412: regularized}, forms for the mean electron flux spectrum
413: ${\overline F}(E)$ from high-resolution {\em RHESSI} photon spectra
414: $I(\epsilon)$. The advantage of this method is that it is not
415: necessary to assume an empirical form for the spectrum.
416: Additionally, as shown by Brown et~al. (2006), this method is
417: capable of revealing accurately the overall ``shape'' of the electron
418: spectrum and indicating the presence and approximate form of
419: small-scale features of sufficient amplitude, if present.
420:
421: \begin{figure}[pht]
422: \epsscale{0.99} \plotone{f5.eps} \caption{Recovered forms of the
423: quantity $\bar n \, V \, {\overline F}(E)$ (in units of
424: $10^{50}$~electrons~cm$^{-2}$~s$^{-1}$~keV$^{-1}$; see
425: equation~[\ref{def}]) using a zero-order regularization technique
426: and presented as a ``confidence strip,'' i.e., a series of
427: solutions, each based on a realization of the data consistent with
428: the size of the uncertainties. The dashed lines assume
429: electron-ion emission only; the solid lines include the additional
430: electron-electron emission term.} \label{reg_sol}
431: \end{figure}
432:
433: Figure \ref{reg_sol} shows the recovered ${\overline F}(E)$
434: solution for the same photon spectrum used in the forward fit
435: procedure of Figure~\ref{ffit}. The results are presented in the
436: form of a {\it confidence strip}, a set of different realizations
437: of ${\overline F}(E)$, each curve corresponding to a different
438: realization of the noisy data set $I(\epsilon)$.
439:
440: It is clear that the ${\overline F}(E)$ recovered using the full
441: cross-section~(\ref{qtot}), including electron-electron
442: bremsstrahlung, is, for $E \gapprox 300$~keV, steeper (spectral
443: index greater by $\sim 0.4$) than the ${\overline F}(E)$ recovered
444: assuming purely electron-ion emission. This result is consistent
445: not only with the forward-fit results of the previous subsection
446: but also with the physical expectations enunciated in \S1.
447: Moreover, the dashed confidence strip (corresponding to use of the
448: electron-ion cross-section alone) has an upward break near
449: $E=400$~keV (as can be verified visually by looking {\it along} --
450: rather than {\it at} -- the strip). However, the true form of
451: ${\overline F}(E)$, as exhibited by the solid confidence strip,
452: has a rather featureless power-law form over the energy range from
453: 200 - 1000~keV. Consequently, use of the full cross-section,
454: including the electron-electron term that becomes important at
455: energies $\gapprox 300$~keV, removes the need to account for the
456: $\sim 400$~keV energy that characterizes the (unphysical) upward
457: break in ${\overline F}(E)$ that appears when only the partial
458: (electron-ion) cross-section is used in the analysis.
459:
460: \section{Discussion and Conclusions}
461:
462: As expected, recognition of the growing importance of
463: electron-electron bremsstrahlung at high energies reduces, for a
464: given hard X-ray spectrum, the number of high-energy electrons
465: required to produce it; this leads to a steepening in the inferred
466: mean source electron spectrum ${\overline F}(E)$ above $\sim
467: 400$~keV. For the January 17, 2005 event studied, use of the
468: electron-ion cross-section alone leads, whether by forward fitting
469: or regularized inversion, to the inference of an upward break
470: ($\Delta \gamma \simeq 0.4$) in ${\overline F}(E)$ at $E \simeq
471: 400$~keV (Figure~\ref{reg_sol}). However, when both electron-ion
472: and electron-electron bremsstrahlung emission are considered, this
473: break disappears, resulting in an ${\overline F}(E)$ that has a
474: straightforward power-law form over the energy range from 100 -
475: 1000~keV. Single-power law suggests electron acceleration without
476: characteristic energy and corresponding characteristic scale.
477: Careful interpretation is therefore necessary when faced with
478: apparent hard X-ray spectral changes in this energy range.
479:
480: One process that can, for a sufficiently strong magnetic field,
481: operate strongly in the few 100 keV range and so affect this
482: argument is gyrosynchrotron emission. However, the presence of
483: this additional emission mechanism would cause the ${\overline
484: F}(E)$ to bend {\it downward} at higher energies. The fact that,
485: after inclusion of the electron-electron contribution, ${\overline
486: F}(E)$ has no such bend puts an upper limit on the importance of
487: gyrosynchrotron emission and so an upper limit ($\sim 10$~kG) on
488: the strength of the ambient magnetic field.
489:
490: Trottet et~al. (1998) report very significant upward breaks
491: ($\Delta \gamma \simeq 1.2 - 2; E_{\rm brk} \simeq 400$~keV) in
492: the hard X-ray spectrum for a series of intervals during an
493: electron-dominated gamma-ray event on 1990 June 11. We agree with
494: these authors that the inclusion of electron-electron
495: bremsstrahlung cannot account for such breaks. However, Vestrand
496: (1988) reports that ``most flares show a break $\simeq 0.5$''
497: occurring at an energy ``$\simeq 300 - 400$~keV'' and a similar
498: statement is made by Dennis (1985) (however, he also reports a
499: much larger spectral break [$\Delta \gamma \simeq 2$] in an event
500: observed on June 4, 1980). Such modest ($\Delta \gamma \simeq
501: 0.5$) upward breaks at a photon energies $\epsilon \simeq 300 -
502: 400$~keV are naturally accounted for by including the contribution
503: from electron-electron bremsstrahlung; other considerations, such
504: as energy-dependent anisotropy (Li 1995) or a separate
505: emission/acceleration process (e.g., Heristchi 1986) are in
506: general not required.
507:
508: Only features common to all (or at least nearly all) realizations
509: of ${\overline F}(E)$ can be considered real. Using this
510: criterion, one must concede that the recovered confidence strip
511: (Figure~\ref{reg_sol}) is sufficiently wide that no firm evidence
512: for a sudden change in the local spectral index $\gamma$ (cf.
513: Figure~\ref{theta}) can be claimed. Hence the data do not provide
514: compelling evidence for either strong beaming of the accelerated
515: electrons or an upper-energy cutoff $E_{\rm max}$ in the
516: accelerated electron energy distribution. Such an assessment is
517: bolstered by Kontar \& Brown's (2006) finding, using a comparison
518: of the brightness of the primary source with that of the
519: photospherically-backscattered albedo patch (Kontar et~al. 2006),
520: that the electron distribution at energies $E \lapprox 200$~keV in
521: the 2005 January 17 event was also consistent with isotropy.
522:
523: \acknowledgments EPK was supported by a PPARC Advanced Fellowship
524: and by a grant from the Royal Society; AGE was supported by NASA
525: Grant NNG04G063G and by subcontract SA4878-26308 from the
526: University of California, Berkeley. The overall effort has
527: greatly benefited from support by a grant from the International
528: Space Science Institute (ISSI) in Bern, Switzerland.
529:
530: %\clearpage %---------------------------------------------------------
531:
532: \begin{references}
533:
534: \reference{} Brown, J.C., Emslie, A.G., \& Kontar, E.P. 2003, \apjl, 595, L115
535:
536: \reference{} Brown, J.C., Emslie, A.G., Holman, G.D., Johns-Krull,
537: C.M., Kontar, E.P., Lin, R.P., Massone, A.M., \& Piana, M. 2006,
538: \apj, 643, 523
539:
540: \reference{} Dennis, B.R. 1985, \solphys, 118, 49
541:
542: \reference{} Haug, E. 1975, Z. Naturforsch, 30, 1099
543:
544: \reference{} Haug, E. 1989, \aap, 218, 330
545:
546: \reference{} Haug, E. 1998, \solphys, 178, 341
547:
548: \reference{} Heristchi, D. 1986, \apj, 311, 474
549:
550: \reference{} Holman, G.D., Sui, L., Schwartz, R.A., \& Emslie,
551: A.G. 2003, \apjl, 595, L97
552:
553: \reference{} Koch, H.W., \& Motz J.W., 1959, Phys. Rev., 31, 920
554:
555: \reference{} Kontar, E.P., \& Brown 2006, \apjl, 653, 149
556:
557: \reference{} Kontar, E.P., MacKinnon, A.L., Schwartz, R.A., and
558: Brown, J.C., 2006, Astronomy \& Astrophysics, 446, 1157.
559:
560: \reference{} Kozlovsky, B., Murphy, R.J., \& Ramaty, R. 2002, \apjs, 141, 523
561:
562: \reference{} Li, P. 1995, \apj, 443, 855
563:
564: \reference{} Piana, M. Massone, A. M., Kontar, E.P., Emslie, A.G.,
565: Brown, J.C., \& Schwartz, R.A. 2003,
566: \apjl, 595, L127
567:
568: \reference{} Ramaty, R., Kozlovsky, B., \& Lingenfelter, R.E. 1979, \apjs, 40, 487
569:
570: \reference{} Schwartz, R.A., et al. 2002, \solphys, 210, 165
571:
572: %\reference{} Smith, D. M., Share, G.H., Murphy, R. J., Schwartz, R.A., Shih, A.Y., \& Lin, R.P. 2003,
573: % \apjl, 595, L81
574:
575: \reference{} Smith, D. M., Lin, R. P., Turin, P. et al., 2002, {\it Solar Physics}, 210, 33
576:
577: \reference{}Trottet, G., Vilmer, N., Barat, C., Benz, A., Magun, A.,
578: Kuznetsov, A., Sunyaev, R., \& Terekhov, O. 1998, \aap, 334, 1099
579:
580: \reference{} Vestrand, W.T. 1988, \solphys, 118, 95
581:
582: \end{references}
583:
584: \end{document}
585: