0710.3223/ms.tex
1: %\\documentstyle[12pt,aasms4_fix]{article}
2: %\documentstyle[12pt,aasms4]{article}
3: %\documentstyle[11pt,aaspp4]{article}
4: %\documentstyle[aas2pp4]{article}
5: %\documentstyle[]{article}
6: \documentclass[preprint,12pt]{aastex}
7: %\documentclass[preprint2]{aastex}
8: %\documentclass[apjpt4]{aastex}
9: %\documentclass[letterpaper]{aastex}
10: 
11: \newcommand{\SOHO}{{\it SOHO}}
12: \newcommand{\TRACE}{{\it TRACE}}
13: \newcommand{\STEREO}{{\it STEREO}}
14: 
15: \begin{document}
16: %
17: \baselineskip=20pt
18: %
19: \title{Observational Test of Coronal Magnetic Field Models \\
20:        I. Comparison with Potential Field Model}
21: 
22: \author{Yu Liu \& Haosheng Lin}
23: \affil{Institute for Astronomy, University of Hawaii \\
24:        34 Ohia Ku Street, Pukalani, HI 96768, USA}
25: 
26: \begin{abstract}
27: Recent advances have made it possible to obtain two-dimensional 
28: line-of-sight magnetic field maps of the solar corona from 
29: spectropolarimetric observations of the \ion{Fe}{13} 1075 nm forbidden 
30: coronal emission line. Together with the linear polarization measurements 
31: that map the azimuthal direction of the coronal magnetic field projected 
32: in the plane of the sky containing Sun center, these coronal vector 
33: magnetograms allow for direct and quantitative observational testing of 
34: theoretical coronal magnetic field models. This paper presents 
35: a study testing the validity of potential-field coronal magnetic field 
36: models. We constructed a theoretical coronal magnetic field model of 
37: active region AR 10582 observed by the SOLARC coronagraph in 2004 by using 
38: a global potential field extrapolation of the synoptic map of Carrington 
39: Rotation 2014. Synthesized linear and circular polarization maps from 
40: thin layers of the coronal magnetic field model above the active 
41: region along the line of sight are compared with the observed maps. 
42: We found that the observed linear and circular polarization signals 
43: are consistent with the synthesized ones from layers located just above 
44: the sunspot of AR 10582 near the plane of the sky containing the Sun 
45: center. 
46: 
47: %This result provides the first observational evidence 
48: %that potential field extrapolation can yield a reasonable approximation 
49: %of the magnetic field configuration of the solar corona for simple and 
50: %stable active regions. 
51: \end{abstract}
52: 
53: \keywords{Sun: corona --- Sun: magnetic fields} 
54: 
55: \section{Introduction}
56: \label{sec:intro}
57: 
58: Understanding the static and dynamic properties of the solar corona 
59: is one of the great challenges of modern solar physics. Magnetic fields
60: are believed to play a dominant role in shaping the solar corona. 
61: Current theories also attribute reorganization of the coronal magnetic 
62: field and the release of magnetic energy in the process as the primary 
63: mechanism that drives energetic solar events. However, direct measurement 
64: of the coronal magnetic field is a very difficult observational problem.
65: Early experiments have demonstrated the feasibility of the measurement 
66: of the orientation of the coronal magnetic fields by observation of the 
67: linear polarization of forbidden coronal emission lines (CELs) in the 
68: visible and at IR wavelengths \citep{eddy1967,mickey1973,arnaud1982,
69: querfeld1982,tomczyk_et_al_2007}. Radio observations have also been 
70: successful in measuring the strength of the coronal magnetic field near 
71: the base of the solar corona \citep[e.g.,] [and references therein] 
72: {brosius_2006}. Direct measurement of the coronal magnetic field strength 
73: at a higher height by IR spectropolarimetry of the CELs was achieved 
74: only recently \citep{lin_et_al_2000, lkc_2004}. 
75: Without direct measurements, past studies involving coronal magnetic fields 
76: have relied on indirect modeling techniques to infer the coronal 
77: magnetic field configurations, including coronal intensity images observed in 
78: the EUV and X-ray wavelength ranges and numerical methods that reconstruct 
79: the three-dimensional coronal magnetic field structure by extrapolation 
80: and MHD (magnetohydrodynamics) simulation based on photospheric magnetic 
81: field measurements. Since experimental verification of theories and models 
82: is one of the cornerstones of modern science, the lack of observational 
83: verification of these indirect magnetic field inference methods that 
84: are in widespread use is a very unsatisfactory deficiency in our field. 
85: 
86: Our 2004 observations were obtained above active region AR 10582 right 
87: before its west limb transit. We have obtained the first measurement 
88: of the height dependence of the strength of the line-of-sight (LOS) 
89: component of the coronal magnetic field in this data, and it showed an 
90: intriguing reversal in the direction of the LOS magnetic field at a 
91: height of approximately $0.15\ R_{\odot}$ above the solar limb, as shown 
92: in Figures 4 and 5 of \cite{lkc_2004}. This feature and the observed 
93: linear polarization map have presented us with our first opportunity to 
94: carry out a comprehensive observational test of our coronal magnetic 
95: modeling methods in which the strength and direction of the  magnetic 
96: fields predicted by the models can be directly checked by the observations. 
97: 
98: Force-free extrapolation of photospheric magnetic fields is currently 
99: the primary tool for the modeling of coronal magnetic fields. However, 
100: it is not without limitations or uncertainties. For example, 
101: the force-free assumption does not hold true in the photosphere and low 
102: chromosphere, and possibly in the high corona above 2 $R_{\odot}$ 
103: \citep{gary_2001}. Moreover, a different assumption (current-free, linear and 
104: nonlinear force free) about the state of the electric current in the corona 
105: can lead to substantially different extrapolation results. Without direct 
106: magnetic field measurements, many fundamental questions concerning the 
107: basic assumptions and the validity of our tools cannot be addressed 
108: directly. To date, questions like ``Is a potential field approximation 
109: generally an acceptable approximation for coronal magnetic fields?" or 
110: ``Do linear or nonlinear force-free extrapolations provide a more accurate 
111: description of the coronal magnetic field?" can only be addressed by 
112: visual comparison between the morphology of selected field lines of the 
113: extrapolated magnetic field model and observational tracers of coronal 
114: magnetic fields such as the loops seen in EUV images. However, these 
115: visual tests are qualitative and subjective. Furthermore, they assume 
116: the coalignment between the magnetic field lines and the loops in EUV 
117: images, which has not been verified observationally. 
118: 
119: As our first test, we attempted to address the question ``Is the potential 
120: field extrapolation generally an acceptable approximation for the coronal 
121: magnetic field?" This test was conducted by comparing the observed polarization 
122: maps of AR 10582 to those derived from a coronal magnetic field model 
123: constructed from the potential field extrapolation of photospheric magnetic 
124: field data. However, before we present our study, we should point out 
125: that because of the nature of our modeling tools and observational data, 
126: the results and conclusions of this research are subject to certain 
127: limitations and uncertainties. One of the intrinsic limitations of the 
128: observational data used in this study is that because of our single sight 
129: line from Earth to the Sun, the photospheric and coronal magnetic field 
130: observations cannot be obtained simultaneously. In the case of global 
131: coronal magnetic field models, the whole-Sun photospheric magnetic field 
132: data used as the boundary condition of the extrapolation can be 
133: obtained only over an extended period of time. Although the large-scale 
134: magnetic structure of nonflaring active regions may appear stable over 
135: a long period of time, high-resolution EUV observations have shown that 
136: the small-scale coronal structures are constantly changing. Thus, 
137: studies such as ours that compare coronal magnetic field observations 
138: and models constructed from photospheric magnetic fields inevitably 
139: are subject to uncertainties due to the evolution of the small-scale 
140: structures in the regions, and we should not expect a precise match 
141: between the observed and synthesized polarization maps. 
142: 
143: Another deficiency due to the single sight line of our observations
144: is the lack of knowledge of the source regions of the coronal radiation. 
145: This is perhaps the most limiting deficiency of the observations and 
146: models of this research. The uncertainty of the location of the source
147: regions associated with coronal intensity observations due to the low 
148: optical density of the coronal plasma and the resulting long integration 
149: path length is familiar. In the case of the coronal magnetic 
150: observations, the LOS integration problem prevents us from performing 
151: an inversion of the polarization data to reconstruct the three-dimensional 
152: magnetic field structure of the corona for direct comparison with those 
153: derived from extrapolations or MHD simulations.  On the other hand, since 
154: extrapolation techniques do not include the thermodynamic properties of 
155: the plasma in the construction of the coronal magnetic field models, 
156: they do not include information about the location of the 
157: CEL source regions either. Therefore, they cannot predict the 
158: intensity and polarization distribution of the CEL projected on the 
159: plane of the sky that are needed for direct comparison with the 
160: polarimetric observations. 
161: 
162: Without the information about the location of the source regions 
163: from the observational data and the extrapolated models, we adopted 
164: a trial-and-error approach in which synthesized linear and circular 
165: polarization maps were derived using empirical source functions
166: and the extrapolated potential magnetic field model, and were 
167: compared directly with those obtained from observations. Obviously, 
168: if acceptable agreement can be achieved with any of the models tested,
169: then we can argue with a certain degree of confidence that these models
170: are plausible models of the observed corona and that the potential 
171: field approximation is a reasonable approximation of the coronal 
172: magnetic fields. Nevertheless, we should emphasize that this 
173: trial-and-error approach is not an exhaustive search of all the 
174: possible source functions and therefore cannot provide a clear-cut 
175: true-or-false answer. In other words, even if no acceptable 
176: agreement can be found with all the model source functions we have 
177: considered, the potential field approximation still cannot be 
178: dismissed completely. 
179: 
180: \section{Data Analysis and Results}
181: 
182: \subsection{Modeling the Coronal Magnetic Fields}
183: 
184: \subsubsection{Evolutionary History of AR 10581 and AR 10582}
185: 
186: Although the purpose of this study is to test the validity of potential 
187: field approximation for AR 10582, examination of the photospheric magnetic 
188: field configuration and evolution of the regions should be informative 
189: and helpful for assessing the results of this study. Figure 
190: \ref{fig:context_info} shows the \TRACE\ (Transition Region 
191: and Coronal Explorer) white-light image and the \SOHO/MDI 
192: (Solar and Heliospheric Observatory/Michelson Doppler Imager) 
193: magnetogram of these regions on 2004 April 1. Their activity history 
194: during their disk transit is shown in Figure \ref{fig:flare_counts}. 
195: AR 10581 and 10582 first appeared on the east limb of the Sun on 2004 
196: March 23. Images taken by \SOHO/EIT (Extreme-ultraviolet Imaging Telescope) and \TRACE\ 
197: data showed that AR 10582 was initially very active and produced several 
198: C- and M-class flares between March 23 and April 1. AR 10581, on the 
199: other hand, produced only two small flares in the first week.  No flare 
200: activities were observed from either region in the four days before the 
201: SOLARC observation. Close examination of EIT and \TRACE\ data also showed 
202: that the large-scale configuration of these active regions did not change 
203: significantly during this period. 
204: 
205: \subsubsection{Extrapolated Potential Field Model of AR 10582}
206: 
207: We employed a global potential field extrapolation program based on 
208: the Green's Function method developed by \cite{schatten_et_al_1969} 
209: and \cite{sakurai_1982} for the construction of the coronal magnetic 
210: field \citep{liu_2002}. The magnetic synoptic map of Carrington Rotation 
211: 2014 obtained by the \SOHO/MDI instrument was used as the boundary 
212: condition of the extrapolation. The first two panels, (a) and (b), of
213: Figure \ref{fig:extrap_B} show selected field lines from the 
214: extrapolated coronal magnetic field model of AR 10581 and AR 10582 
215: plotted over a \TRACE\ \ion{Fe}{9} 171 \AA\ image as they are viewed on 
216: the disk. Panels (c) and (d) show the same set of field lines overplotted on 
217: the \SOHO/EIT \ion{Fe}{9} 171 \AA\ images and \SOHO/MDI magnetograms 
218: when they transited the west limb. The field of view of the SOLARC LOS 
219: magnetogram observation is marked by the rectangular box in the figure. 
220: We found a general agreement between the orientation and distribution 
221: of the discernible EIT intensity loops during the time of the SOLARC 
222: observations, and the selected magnetic field lines from the extrapolated 
223: model are evident. The extrapolated magnetic field lines in AR 10582 are 
224: predominantly aligned along the east-west direction, while a set of field 
225: lines running along the north-south direction connecting the two active 
226: regions also coincide with a large north-south loop in the EIT images.  
227: Figure \ref{fig:trace_loops} shows a \TRACE\ high-resolution 
228: \ion{Fe}{9} 171 \AA\ image of AR 10582 taken about 3 hr after the 
229: limb spectropolarimetric observations had ended. We found that a 
230: subset of the extrapolated field lines appear to closely resemble 
231: the TRACE loops.  While the similarities in the morphology of the EUV 
232: intensity images and the extrapolated magnetic field lines seem to 
233: suggest that the extrapolated coronal magnetic field model is a fair 
234: representation of the magnetic field configuration of AR 10581 
235: and AR 10582, it is a subjective interpretation. 
236: 
237: \subsection{Observational Test of the Potential-Field 
238:             Coronal Magnetic Field Model}
239: 
240: \subsubsection{Synthesis of the Coronal Polarization Maps}
241: 
242: We have developed a program to calculate the LOS integrated linear 
243: and circular polarization signals at any point in the plane of the sky 
244: (POS) given the extrapolated coronal magnetic field model and the density 
245: and temperature distribution of the solar corona based on the classical 
246: theory of \cite{lin_casini_2000} for the forbidden CEL polarization. 
247: The formulae for the emergent Stokes parameter of the CEL are identical 
248: to those derived from a full-quantum mechanical formulation 
249: \citep{casini_judge_1999}, up to a proportional constant. However, 
250: as this classical formulation does not consider the effect of collisional 
251: depolarization, it may overestimate the degree of linear polarization at 
252: a lower height. The azimuth angle of the linear polarization predicted 
253: by our program should also be affected slightly when we integrate over 
254: a long path length, since the collisional depolarization effect may change 
255: the relative contribution of the sources along the LOS. 
256: \cite{judge_casini_2001} have also developed a CEL polarization synthesis 
257: program that includes the effect of collisions. We have implemented both 
258: programs to generate the linear and circular polarization signals from 
259: the extrapolated potential field model. 
260: We found that the synthesized polarization maps derived from these 
261: two programs are very similar, and analysis based on these two programs 
262: yielded the same results. The comparison of the synthesized polarization 
263: maps from these two programs will be presented when we present the 
264: study comparing the observed and synthesized polarization maps.
265: 
266: \subsubsection{Linear Polarization Maps}
267: 
268: The 2004 April 6 data include a linear polarization scan encompassing 
269: both AR 10581 and AR 10582. However, due to the long 
270: integration time required, only one circular polarization measurement 
271: was obtained above AR 10582.  Therefore, this study concentrates on 
272: the $320 \arcsec \times 160 \arcsec$ field with both circular and linear 
273: polarization measurements as marked by the rectangular area in Figures 
274: \ref{fig:extrap_B} and \ref{fig:trace_loops}. As it was mentioned in 
275: \S\ref{sec:intro}, the lack of knowledge of the coronal density and 
276: temperature distribution is the greatest uncertainty in our study. 
277: Nevertheless, based on decades of observations, we now know that strong 
278: coronal emissions are always associated with active regions. Furthermore, 
279: due to the small density scale height of the high-temperature emission 
280: lines, the contribution function of the CELs along the LOS is heavily 
281: weighted toward layers close to the POS containing Sun center. 
282: Therefore, we can expect that for isolated active regions, the 
283: forbidden coronal emission originates from a localized region near 
284: the POS containing Sun center during the active region's limb transit.  
285: According to the \SOHO/MDI white-light archive, there were no other 
286: active regions present on the solar disk on 2004 April 1, when AR 10581 
287: and AR 10582 were located approximately at disk center. Additionally, 
288: there was no evidence of new active regions emerging in the vicinity 
289: of these regions up to the day of the coronal polarization 
290: measurement.  Therefore, it is reasonable to assume that the polarized 
291: radiations we measured originated in the corona above these active 
292: regions near the POS containing Sun center. This expectation prompted 
293: us to test if an empirical source function constructed from a simple 
294: gravitationally stratified atmosphere and local magnetic field properties, 
295: such as the strength or the magnetic energy, can reproduce (if only 
296: qualitatively) the observed polarization maps.  
297: 
298: For our first test, we constructed a linear polarization map using a 
299: source function ${W_{n_e}}(\bf{r})$ that is proportional to the square of the density of 
300: a spherically symmetric, gravitationally stratified density distribution 
301: with a density scale height of $h_0=83$ Mm 
302: \citep[$0.11\ R_{\odot}$,][]{lang_1980}.  That is, 
303: %
304: \begin{eqnarray}
305: \label{eq:source_function_1}
306: W_{n_e}({\bf r}) = e^{-2h({\bf r})/h_0},
307: \end{eqnarray}
308: %
309: where ${\bf r}$ is the three-dimensional position vector in the heliocentric 
310: coordinate system and $h({\bf r})$ is the height at ${\bf r}$. 
311: The temperature of the corona was assumed to be constant and did not affect 
312: the source function. The magnetic properties of the corona are not included 
313: in this model either. This is similar to the atmospheric model used in the 
314: polarization synthesis performed by \cite{judge_et_al_2006}. The resulting 
315: normalized source function along the LOS in the center of the SOLARC field 
316: of view and the synthesized linear polarization map are shown in panel 
317: (b) of Figure \ref{fig:empirical_models}.  Note that because 
318: AR 10582 was located between W70 to W80  longitude, this density-only source 
319: function has its maximum located outside of the active region. Although 
320: $W_{n_e}({\bf r})$ is obviously a gross simplification of the magnetic coronal 
321: atmosphere, and we should not expect to see good agreement between the observed 
322: and synthesized linear polarization maps based on $W_{n_e}({\bf r})$ alone 
323: (as demonstrated in the top figure in panel (b) of Figure 
324: \ref{fig:empirical_models}),  this test is a necessary step in our systematic 
325: trial-and-error study. Furthermore, 
326: the importance of its inclusion in the estimate of the source function 
327: can be seen when we compare the linear polarization maps constructed with 
328: and without it (top figure in panel (a) of Figure \ref{fig:empirical_models}). 
329: 
330: In our next test, we examined if there is any simple relationship 
331: between the CEL source function and the magnetic properties of the corona.  
332: Because of the observed correlation between the strong  CEL radiation and 
333: active regions, it is only logical to test if a source function with its 
334: amplitude proportional to the local magnetic field strength or magnetic 
335: energy can reproduce the observed polarization signals. To test this idea,
336: we multiplied the source function that was constructed from the uniform 
337: temperature, gravitationally stratified atmosphere model by the local 
338: magnetic field strength $B$. Another model used the local magnetic field 
339: energy density $B^2$ as the additional magnetic weighting function. 
340: Accordingly, the two magnetic source functions are expressed by
341: %
342: \begin{eqnarray}
343: W_{B}({\bf r}) = e^{-2h({\bf r})/h_0} B
344: \end{eqnarray} 
345: and
346: \begin{eqnarray}
347: W_{B^2}({\bf r}) = e^{-2h({\bf r})/h_0} B^2.
348: \end{eqnarray} 
349: %
350: The additional magnetic constraints restrict the source function to a 
351: more localized region around the active region, and with a spatial 
352: scale comparable to that of the active region.  However, the linear 
353: polarization maps derived using these source functions still are not 
354: in good agreement with the observed one. These are demonstrated in 
355: panels (c) and (d) of Figure \ref{fig:empirical_models}. 
356: 
357: Although none of the linear polarization maps we have constructed 
358: so far can be considered to be in good agreement with the observation, 
359: it can be argued that the maps derived from the two source functions 
360: with magnetic constraints appear to better match the observed one, 
361: especially at the lower right-hand corner of the field, where the 
362: degree of linear polarization is the highest. Therefore, we suspect 
363: that the assumption about the magnetic dependence of the CEL source 
364: function is in general valid. However, the correlation may be occurring 
365: at a spatial scale smaller than that of the active regions.  Because 
366: space EUV observations have shown that radiation from the emission-line 
367: corona originates from loop-like structures with a characteristic 
368: size much smaller than the characteristic size of the active regions, 
369: it should not be surprising that the linear polarization maps produced 
370: by the broad source functions do not agree well with the observation.
371: This reasoning prompted us to experiment with source functions with
372: a spatial scale approximately equal to a few times the characteristic 
373: width of the coronal loops to test if better agreement between the 
374: observed and synthesized polarization maps can be achieved.  
375: 
376: Since we do not have information about the location of the source regions 
377: of the observed \ion{Fe}{13} 10747 \AA\ line emission at this spatial 
378: resolution, we constructed the linear polarization map of 205 layers 
379: along the LOS within the SOLARC FOV for comparison with the observed 
380: linear polarization map. The separation between the synthesized layers 
381: is 4.5 Mm, the resolution of the potential field extrapolation calculation. 
382: The first and last layers are located 720 Mm (or about 1000$\arcsec$) in 
383: front and 250 Mm (or about 340$\arcsec$) behind the POS 
384: containing Sun center, respectively. The locations of these layers with 
385: respect to the solar sphere, the observed active region, and the LOS 
386: of the observer are illustrated in Figure \ref{fig:obs_geometry}.  
387: %
388: We used a one-dimensional Gaussian function with a full width at 
389: half-maximum (FWHM) in the LOS direction equal to a few times that 
390: of the characteristic coronal loop size to model the source function. 
391: For this test, we adopted the 8 Mm characteristic loop width of the 
392: EUV \ion{Fe}{14} 28.4 nm line derived by \cite{aschwanden_et_al_2000}. 
393: This line is chosen because its ionization temperature of $2.2\times10^6$ 
394: K is closer to that of the IR \ion{Fe}{13} 10747 \AA\ line 
395: ($T_{\rm ion} = 1.7 \times 10^{6}$ K) than the other \SOHO/EIT EUV 
396: lines. We calculated and compared the polarization maps with the FWHM 
397: of the source function set from 1 to 14 times the characteristic loop 
398: width of the \ion{Fe}{14} 28.4 nm lines and found no significant 
399: difference between the polarization maps. This is expected, since the 
400: coronal magnetic field should vary slowly in space as it expands to 
401: fill the entire coronal volume. Finally, because of this lack of 
402: sensitivity in the synthesized polarization signals to the variations 
403: of the FWHM of the source function, we used a nominal 56 km FWHM 
404: source function (or 7 times the FWHM of the \ion{Fe}{13} 28.4 nm loops) 
405: for the calculation of the synthesized polarization maps in the rest 
406: of the paper. 
407: 
408: We evaluated the quality of the fit between the synthesized and observed 
409: polarization maps from the rms difference between the degree of polarization 
410: $p$ and the azimuthal angles $\chi$ of the linear polarization direction 
411: projected in the POS, $\sigma_p$ and $\sigma_\chi$, respectively. 
412: Figure \ref{fig:fit_errors} shows $\sigma_p$ and $\sigma_\chi$ as functions 
413: of $z$ along the LOS. Results derived from our own collisionless classical 
414: formulation are shown by the black lines, and those derived with Judge \& 
415: Casini's code (2001; hereafter referred to as the JC synthesis code) are shown 
416: in red. Figure \ref{fig:fit_errors} shows that the minimum of 
417: $\sigma_p$ and $\sigma_\chi$ both 
418: occur near the layer right above the sunspot of the active region. Figure 
419: \ref{fig:linear_polarization_maps} shows 15 synthesized linear polarization 
420: maps calculated with our classical synthesis code from layer 70 to layer 140, 
421: with an interval of 5 layers, plotted over the observed map. The best fit, 
422: determined from the total rms error $\sigma^2_{lp} = \sigma^2_p + 
423: (\sigma_{\chi}/\pi)^2$ of the linear polarization maps, occurs around layer 
424: 120, right above the sunspot. 
425: 
426: 
427: \subsubsection{Comparison with Judge \& Casini's Coronal Polarization 
428:                Synthesis Code}
429: 
430: Figure \ref{fig:judge_lp} shows the linear polarization map of layer 
431: 120 derived with the JC synthesis code compared with that derived from 
432: our collisionless code, and with observations. Apparently, the JC synthesis 
433: code consistently predicts a smaller linear polarization amplitude 
434: compared with our classical theory, as expected, and produced better 
435: agreement with the observed linear polarization map at the lower part 
436: of the field. However, there are still significant disagreements in the 
437: upper part of the field, where both the JC synthesis code and our own program 
438: overestimated the degree of linear polarization. This may be due to the 
439: larger measurement errors associated with the small amplitudes of the 
440: observed linear polarization, and their significance should not be 
441: overstated. Close examination of the data shows that the larger rms 
442: error shown in Figure \ref{fig:fit_errors} in the JC results is due to 
443: larger errors in this region. Therefore, we do not consider that our 
444: collisionless synthesis actually provides better agreement with the 
445: observations. 
446: 
447: In view of the inherent uncertainties in the modeling process discussed 
448: in \S\ref{sec:intro}, we do not feel that a meaningful quantitative 
449: comparison between these two methods can be justified at this point. 
450: Nevertheless, Judge \& Casini's program with collisional depolarization 
451: is definitely a more complete description of the physical processes in 
452: the atmosphere of the solar corona and should be preferred. Future 
453: observations at a lower height, where the collisional depolarization effect is 
454: expected to be more important, should allow us to clearly distinguish 
455: between the effect of collisional depolarization and uncertainties in 
456: the modeling process. 
457: 
458: \subsubsection{Strength and Reversal of Line-of-Sight Magnetic Fields}
459: 
460: Our analysis so far has demonstrated that the potential field 
461: model can reproduce the observed linear polarization maps. 
462: Is this a coincidence? To answer this question, it is interesting 
463: to first note that the layers of best fit for both the degree and 
464: azimuthal angle of the linear polarization occur at approximately 
465: the same location near the region of the strongest photospheric 
466: magnetic fields. Since $\sigma_p$ and $\sigma_{\chi}$ were determined 
467: independently, the probability that these two parameters reach minimum 
468: at approximately the same location and at the region with the strongest 
469: photospheric magnetic flux purely by chance should be very low. 
470: Therefore, we believe that our analysis of the linear polarization 
471: maps support the idea that the potential field model is a good 
472: approximation of the real coronal magnetic field of AR 10582. 
473: Furthermore, if an agreement between the modeled and observed height 
474: dependence of the LOS component of the coronal magnetic field can be 
475: found, then the validity of the potential field model, at least 
476: as a first order approximation for stable active regions, can be 
477: strongly argued. 
478: 
479: To check if the observed Stokes $V$ reversal can be reproduced, we 
480: derived the height of Stokes $V$ reversal $H_0$ for each of the 205 
481: 56-Mm-FWHM layers along the sight line. The LOS components of the 
482: magnetic field $B_z$ in the central $320 \arcsec \times 80 \arcsec$ 
483: region of each layer were averaged in the north-south direction (or 
484: the tangential direction with respect to the local solar limb) to 
485: simulate the spatial averaging performed on the observation data. 
486: The result is  shown in Figure \ref{fig:StokesV_Reversal}. 
487: Since the dominant magnetic structure around the active region in our 
488: FOV consists of magnetic loops oriented along the east-west direction, 
489: there were two locations with $H_0 = 0.15\ R_{\odot}$, one due to the 
490: front (closer to the observer) portion of the loops near layer 80, and 
491: one located in the back of the loops, at layer 130. Note that layer 130 
492: is much closer to the sunspot of AR 10582 and the maximum of the empirical 
493: source functions $W_B$ and $W_{B^2}$ shown in Figure 
494: \ref{fig:empirical_models}. In comparison, the amplitudes of 
495: $W_{n_e}$, $W_B$, and $W_{B^2}$ at layer 80 are only about 0.15, 0.1, 
496: and 0.05, respectively (Figure \ref{fig:empirical_models}). 
497: We calculated the net Stokes $V$ signals as a function of height above 
498: the limb for layer 130 and found that they agree well with the observed 
499: signals as shown in Figure \ref{fig:Synthesized_StokesV}. 
500: On the other hand, this is not the case for layer 80.
501: We conducted the same analysis using the JC synthesis code and 
502: produced virtually identical results. 
503: Therefore, it is more likely that the dominant source of the Stokes 
504: $V$ signals originates from around layer 130. 
505: 
506: The blue and red Gaussian curves in Figure \ref{fig:StokesV_Reversal} 
507: mark the locations where the best fits to the 
508: linear and circular polarization observation occur, respectively.  
509: As is clearly shown in this figure, layer 130 is only about 50 Mm 
510: away from the location with the best fit of the linear polarization 
511: maps. Given the proximity of the locations of the best fit of the three
512: parameters ($p$, $\chi$, and $H_0$) and the strongest magnetic feature
513: of the active region, we concluded that this comparative study demonstrates
514: the validity of potential field extrapolation as a tool for the modeling 
515: of the coronal magnetic field. 
516: 
517: \section{Summary, Discussions, and Conclusions}
518: 
519: This research examines observationally the validity of current-free,
520: force-free potential-field approximation for coronal magnetic fields. 
521: We conducted a study comparing observed and synthesized spatial 
522: variations of the linear and circular polarization maps in the corona 
523: above active region AR 10582, which after a week of extensive flaring 
524: activities should have settled into a minimum energy configuration that 
525: could be adequately modeled by a potential field model. The coronal 
526: magnetic field model used for this study was constructed from a global 
527: potential field extrapolation of the synoptic photospheric magnetogram 
528: of Carrington cycle 2014 obtained by the \SOHO/MDI instrument. 
529: Because the most important source of error of this type of study is the 
530: uncertainty of the location of the source function of the coronal 
531: radiation, we first tested three analytical, but empirically determined, 
532: source functions.  These simple source functions are based on a 
533: gravitationally stratified atmospheric density model with a uniform 
534: temperature in the entire modeled volume, supplemented by magnetic 
535: weighting functions based on the observational impression that, at 
536: least at the length scale of the typical active region size, CEL 
537: radiation seems to be correlated with the strength of the photospheric 
538: magnetic fields. We found that none of these empirical source functions 
539: can adequately reproduce the observed linear polarization maps, although 
540: it seems that the source functions that include both density and magnetic 
541: fields produced slightly better results. Based again on the observational 
542: impression that the coronal intensity structures have a spatial scale much 
543: smaller than that of the active regions, we then compared the observed 
544: polarization maps with those constructed from thin (56 Mm FWHM) layers along 
545: the LOS. In this analysis, we found that polarization maps originating 
546: from layers located near the sunspot of the region are in reasonable
547: agreement with the observed ones. However, the best fit for linear and
548: circular polarization did not occur at the same layer. They are separated 
549: by a distance of about 50 Mm. 
550: 
551: Does the small discrepancy between the best-fit locations of the linear 
552: and circular polarization weaken the support for the potential field 
553: extrapolation? As we have discussed in \S\ref{sec:intro}, many 
554: uncertainties conspire to limit the precision of this study. For example, 
555: the difference may be due to the evolution of the small-scale photospheric 
556: magnetic field of the active region, and we do not have any observational 
557: information that we can use to test this possibility.  The assumption of 
558: uniform temperature distribution in our source function is certainly not 
559: a physically realistic assumption. Therefore, it is not possible to assess 
560: the significance of the small difference in the location of the source 
561: regions of the linear and circular polarization.  Furthermore, in addition 
562: to density and temperature, the source functions of CEL linear and circular 
563: polarization depend on different components of the coronal magnetic field. 
564: So it is in fact physically reasonable that we would find the linear and 
565: circular polarization signals originate from slightly different locations. 
566: Finally, because the three parameters ($\sigma_p$, $\sigma_\chi$, and $H_0$) 
567: we used to evaluate the quality of the fit were obtained independently, 
568: the statistical significance that all three parameters reached minimum 
569: at approximately the same location near the strongest photospheric magnetic 
570: feature of the active region cannot be dismissed as pure coincidence. These 
571: considerations lead us to conclude that potential field extrapolation can 
572: be used to provide a zero-order approximation of the real solar corona 
573: if the active region is in a relatively simple and stable configuration.
574: Additionally, this study suggests that, at least for isolated active
575: regions, CEL radiation may originate from a region close to the strongest 
576: photospheric magnetic feature in the active region with a small spatial 
577: scale comparable to the characteristic size of the coronal loops seen in 
578: the intensity images. If this is confirmed, then a single-source inversion
579: to infer the magnetic field directly from the polarimetric observation such
580: as that proposed by \cite{judge_2007} may be justified.  
581: 
582: Our conclusion about the viability of potential field extrapolation 
583: as a coronal magnetic field modeling tool for stable active regions 
584: is supported by a study by \cite{riley_et_al_2006}, in which the 
585: coronal magnetic field configuration derived from a potential field 
586: model was found to closely match that derived from a MHD simulation 
587: in the case of untwisted fields. Nevertheless, we should emphasize 
588: that our conclusion is derived from a single observation of a simple and 
589: stable active region. Clearly, more observations and model comparison are 
590: needed for a more comprehensive test of this result. 
591: 
592: Can the potential field approximation be used to model more complicated 
593: active regions? Using radio observations, \cite{lee_et_al_1999} found 
594: that a force-free-field model yields better agreement between the 
595: temperatures of two isogauss surfaces connected by the modeled field 
596: lines of an active region with strong magnetic shear. This study thus
597: provides observational evidence against the use of potential field 
598: approximations for the modeling of complex active regions. Therefore,
599: linear and nonlinear force-free extrapolations should be employed in 
600: future testing of theoretical coronal magnetic field models using the 
601: IR spectropolarimetric observations to study if these models can offer 
602: a better description of the observed coronal fields. 
603: 
604: Can we distinguish the potential coronal magnetic field configurations 
605: from the non-potential ones with the spectropolarimetric observations 
606: of the coronal emission lines? In a numerical study, \cite{judge_2007} 
607: has demonstrated the sensitivity of LOS-integrated 
608: coronal polarization measurements to the electric current in the corona  
609: using theoretical coronal magnetic field models as input.  Therefore, we 
610: should expect to find better agreement between observed and synthesized 
611: polarization maps for more complex active regions with linear or nonlinear 
612: force-free magnetic field models. Work to model AR10582 using the force-free 
613: extrapolation method is already underway, and we should be able to address 
614: this question in the near future. Since all extrapolation methods are 
615: subject to the ambiguities problem of the source regions, we will also 
616: employ MHD simulations that include both the magnetic and thermodynamic 
617: properties of the corona in the calculation to help resolve this problem. 
618: These are research activities that we will be pursuing in the near future 
619: as the solar cycle evolves toward the next solar maximum and more coronal 
620: magnetic field data become available. 
621: 
622: The greatest difficulty of this study is the uncertainty of the location 
623: of the source function due to the long integration path along the LOS.
624: However, this is not a difficulty affecting only the interpretation of 
625: coronal magnetic field measurements. It affects the intensity observation 
626: as well, and is the primary reason that years into the operation of 
627: \SOHO/EIT and \TRACE, we still cannot deduce 3-D intensity and temperature 
628: structure of the corona using data from these instruments. Fortunately, 
629: this deficiency in our observing capability may finally be removed with 
630: the recent launch of the \STEREO\ mission (Solar TErrestrial RElations 
631: Observatory).  For the resolution of the LOS 
632: integration problem in polarimetric observations, \cite{kramar_et_al_2006} 
633: have demonstrated the promising potential of vector tomography techniques.  
634: While stereoscopic coronal magnetic field observations will not be 
635: realized any time soon, this method can be applied to observations
636: obtained over periods of several days during the limb transit of active 
637: regions, provided that the active regions are in a stable condition. 
638: This is perhaps the best observational tool available for the resolution 
639: of the LOS integration problem in the near future. 
640: 
641: \acknowledgements
642: The authors would like to thank Phil Judge for generously providing 
643: the polarization synthesis codes and for helping to implement them. 
644: Y. L. thanks Tongjiang Wang for many discussions on this work.
645: The authors also gratefully acknowledge helpful comments from the 
646: anonymous referee that greatly improved the presentation of the paper.
647: EIT data is courtesy of the \SOHO/EIT consortium. \SOHO\ is a 
648: project of international cooperation between ESA and NASA. This 
649: research is funded by NSF ATM-0421582 and NASA NNG06GE13G.
650: 
651: 
652: %\clearpage
653: 
654: \begin{thebibliography}
655: 
656: \bibitem[Arnaud, J.(1982)]{arnaud1982}
657:          Arnaud, J. 1982, 
658:          \aap, 112, 350
659: \bibitem[Aschwanden et al.(2000)]{aschwanden_et_al_2000}
660:          Aschwanden, M. J., Alexander, D., Hurlburt, N., 
661:          Newmark, J. S., Neupert, W. M., Klimchuk, J. A., \& Garry, G. A.
662:          2000, \apj, 531, 1129
663: \bibitem[Brosius \& White(2006)]{brosius_2006}
664:          Brosius, J. W., \& White, S. M. 2006, 
665:          \apj, 641, L69
666: \bibitem[Casini \& Judge(1999)]{casini_judge_1999} 
667:          Casini, R., \& Judge, P. G. 1999, \apj, 522, 524
668: \bibitem[Eddy \& Malville(1967)]{eddy1967}
669:          Eddy, J.A., \& Malville, M. J. 1967,
670:          \apj, 150, 289
671: \bibitem[Gary(2001)]{gary_2001}
672:          Gary, G. A. 2001, \solphys, 203, 71
673: \bibitem[Judge(2007)]{judge_2007}
674:          Judge, P. 2007, \apj, 662, 677
675: \bibitem[Judge \& Casini(2001)]{judge_casini_2001}
676:          Judge, P. G., \& Casini, R. 2001, in ASP Conf. Ser. 236,
677:          Advanced Solar Polarimetry: Theory, Observation, and 
678:          Instrumentation, ed. M. Sigwarth (San Francisco: ASP), 503 (JC)
679: \bibitem[Judge et al.(2006)]{judge_et_al_2006}
680:          Judge, P., Low, B. C., \& Casini, R. 2006, \apj, 651, 1229
681: \bibitem[Kramar et al.(2006)]{kramar_et_al_2006}
682:          Kramar, M. B., Inhester, B., \& Solanki, S. K. 2006,
683:          \aap, 456, 665
684: \bibitem[Lang(1984)]{lang_1980}
685:          Lang, K. 1980, Astrophysical Formulae (Berlin: Springer)
686: \bibitem[Lee et al.(1999)]{lee_et_al_1999}
687:          Lee, J., White, S. M., Kundu, M. R., McClymont, A. N., 
688:          \& Miki\'c, Z. 1999, 
689:          \apj, 510, 413
690: \bibitem[Lin \& Casini(2000)]{lin_casini_2000} 
691:          Lin, H., \& Casini, R. 2000, \apj, 542, 528
692: \bibitem[Lin et al.(2004)]{lkc_2004} 
693:          Lin, H., Kuhn, J. R., \& Coulter, R. 2004, \apj, 613, L177
694: \bibitem[Lin et al.(2000)]{lin_et_al_2000} 
695:          Lin, H., Penn, M. J., \& Tomczyk, S. 2000, \apj, 541, L83
696: \bibitem[Liu \& Zhang(2002)]{liu_2002} 
697:          Liu, Y., \& Zhang, H. 2002, Publ. Yunnan Obs., 92, 1
698: \bibitem[Mickey(1973)]{mickey1973}
699:          Mickey, D. L. 1973,
700:          \apj, 171, L19
701: %\bibitem[Mickey et al.(1996)]{mickey_et_al_1996} 
702: %         Mickey, D. L., Canfield, R. C., LaBonte, B. J., Leka, K. D., 
703: %         Waterson, M. F., \& Weber, H. M. 1996, \solphys, 168, 229
704: \bibitem[Querfeld(1982)]{querfeld1982}
705:          Querfeld, C.W. 1982,
706:          ApJ, 255, 764
707: \bibitem[Riley et al.(2006)]{riley_et_al_2006}
708:          Riley, P., Linker, J. A., Miki\'c, Z., \& Lionello, R. 2006,
709:          \apj, 653, 1510
710: \bibitem[Sakurai(1982)]{sakurai_1982} 
711:          Sakurai, T. 1982, \solphys, 76, 301
712: \bibitem[Schatten et al.(1969)]{schatten_et_al_1969} Schatten, K. H., 
713:          Wilcox, J. M., \& Ness, N. F. 1969, \solphys, 6, 442
714: \bibitem[Tomczyk et al.(2007)]{tomczyk_et_al_2007}
715:          Tomczyk, S., McIntosh, S. W., Keil, S. L., Judge, P. G.,
716:          Schad, T., Seeley, D. H., \& Edmondson, J. 2007,
717:          Science, 317, 1192
718: 
719: \end{thebibliography}{}
720: 
721: \clearpage
722: 
723: %\begin{center}
724: %FIGURE LEGENDS
725: %\end{center}
726: 
727: \begin{figure}[c] %Fig 1
728: \epsscale{1.0}
729: \plotone{f1.eps}
730: \figcaption{\TRACE\ white-light image (left) and \SOHO/MDI 
731: magnetogram (right) of AR 10581 and AR 10582 observed on the solar 
732: disk show the regions' photospheric intensity and magnetic field 
733: configuration. 
734: \label{fig:context_info}
735: }
736: \end{figure}
737: 
738: \clearpage
739: 
740: 
741: 
742: \begin{figure}[c] %Fig 2
743: \epsscale{0.75}
744: \plotone{f2.eps}
745: \figcaption{Flare counts of AR 10581 and AR 10582 from 2004 March 25 to 
746: 2004 April 07.
747: \label{fig:flare_counts}
748: }
749: \end{figure}
750: 
751: \clearpage
752: 
753: 
754: \begin{figure}[c] %Fig 3
755: \epsscale{1.0}
756: \plotone{f3.eps}
757: \figcaption{Selected extrapolated magnetic field lines plotted over 
758: \ion{Fe}{9} 171 \AA\ images and \SOHO/MDI magnetograms of AR 10581 
759: and AR 10582 when these regions are observed on the solar disk (a and b)
760: and at the west limb (c and d).  
761: \label{fig:extrap_B}
762: }
763: \end{figure}
764: 
765: \clearpage
766: 
767: 
768: \begin{figure}[c] %Fig 4
769: \epsscale{0.75}
770: \plotone{f4.eps}
771: \figcaption{top: \TRACE\ \ion{Fe}{9} 171 \AA\ image of AR 10582 about 3 
772: hr after the SOLARC magnetic field measurements were taken. The rectangle
773: marks the FOV of the SOLARC observations. The horizontal line within 
774: the SOLARC FOV indicates the location where a reversal in the direction 
775: of the longitudinal coronal magnetic field as a function of height above 
776: the solar limb was observed. bottom: Same as the top panel, with a subset 
777: of the magnetic field lines shown in Figure \ref{fig:extrap_B} plotted 
778: over the \TRACE\ image. 
779: \label{fig:trace_loops}
780: }
781: \end{figure}
782: 
783: \clearpage
784: 
785: 
786: \begin{figure*}[c] % Fig 5
787: \epsscale{1.0}
788: \plotone{f5.eps}
789: \figcaption{Comparison between observed and synthesized linear polarization
790: maps with four empirical source functions: (a) reference source function with 
791: uniform density and temperature distribution, (b) source function based on 
792: a gravitationally stratified density distribution and uniform temperature, 
793: (c) gravitationally stratified density source function in (b) weighted by 
794: the local magnetic field strength, and (d) gravitationally stratified density 
795: source function weighted by the local magnetic field energy density, $B^2$. 
796: The observed (black lines) and synthesized (red lines) linear polarization 
797: maps are shown in the top figure in each panel, and the source functions 
798: (thick grey lines) from the center of the observed field along the LOS are 
799: shown in the bottom figure in each panel. The thin black lines in the bottom
800: figures show the strength of the observed photospheric magnetic flux. The 
801: source function in (a) has an equal contribution from every point in 
802: space. It is not a physically realistic model and is shown to demonstrate 
803: how a simple gravitationally stratified density distribution can affect 
804: the outcome of the simulation.
805: \label{fig:empirical_models}
806: }
807: \end{figure*}
808: 
809: \clearpage
810: 
811: 
812: \begin{figure*}[c] %Fig 6
813: \epsscale{1.0}
814: \plotone{f6.eps}
815: \caption{
816: Three-dimensional plot to illustrate the locations of the thin layers along 
817: the LOS, and their relationship to the solar sphere and the sunspot 
818: of AR 10582. The Sun is represented by the shaded quarter 
819: sphere. The observing LOS is along the $Z$ axis. Layer 155 is located 
820: in the POS containing the Sun center. 
821: \label{fig:obs_geometry}
822: }
823: \end{figure*}
824: 
825: \clearpage
826: 
827: 
828: \begin{figure}[c] % Fig 7
829: \epsscale{1.0}
830: \plotone{f7.eps}
831: \figcaption{
832: top: The rms errors between the synthesized and observed linear 
833: polarization amplitude, $\sigma_p$ (dotted lines), azimuthal angle, 
834: $\sigma_\chi$ (dashed lines), and the combination of the two, 
835: $\sigma_{LP}$ (thick solid lines). Results derived with the coronal 
836: polarization synthesis program developed by the authors and 
837: \cite{judge_casini_2001} are shown in black and red, respectively.
838: Best-fit position for linear polarization occurs at approximately 
839: layer 120, right above the sunspot. 
840: bottom: The thin solid line shows the magnitude of the LOS magnetic 
841: field of the photosphere. The thick blue line shows the source 
842: function along the LOS for layer 120. 
843: \label{fig:fit_errors}
844: }
845: \end{figure}
846: 
847: \clearpage
848: 
849: 
850: \begin{figure*}[c] %Fig 8
851: \epsscale{1.0}
852: \plotone{f8.eps}
853: \figcaption{
854: The observed (black lines) and synthesized (red lines) linear polarization
855: maps derived with our classical synthesis program from 15 layers along the 
856: LOS near the active region. Layer 120 is the layer with the smallest rms 
857: error. 
858: \label{fig:linear_polarization_maps}
859: }
860: \end{figure*}
861: 
862: \clearpage
863: 
864: 
865: \begin{figure}[c] %Fig 9
866: \epsscale{0.75}
867: \plotone{f9.eps}
868: \figcaption{
869: top: The linear polarization maps derived from Judge \& Casini's
870: synthesis program (red lines), and from Liu and Lin's classical 
871: synthesis program (black lines). bottom: Comparison of the linear 
872: polarization map derived from Judge \& Casini's program and that 
873: observed by SOLARC. 
874: \label{fig:judge_lp}
875: }
876: \end{figure}
877: 
878: \clearpage
879: 
880: 
881: \begin{figure*}[c] %Fig 10
882: \epsscale{1.0}
883: \plotone{f10.eps}
884: \caption{
885: top: The calculated height of Stokes $V$ reversal as a function of height 
886: from the limb derived from the potential field model. 
887: bottom: Similar to Figure \ref{fig:fit_errors}, the thin solid line shows 
888: the magnitude of the LOS magnetic field of the photosphere, and the thick 
889: blue line shows the source function of layer 120 where the best fit for the 
890: linear polarization map occurs. The thick red 
891: line shows the source function along the LOS for layer 130. 
892: \label{fig:StokesV_Reversal}
893: }
894: \end{figure*}
895: 
896: \clearpage
897: 
898: 
899: \begin{figure}[c] %Fig 11
900: \epsscale{1.0}
901: \plotone{f11.eps}
902: \figcaption{
903: The observed (solid line with star and error bars) and synthesized
904: net circular polarization signals from the source layer 130 (dotted line) 
905: and layer 80 (dashed line) are plotted as a function of distance from 
906: the solar limb.  
907: \label{fig:Synthesized_StokesV}
908: }
909: \end{figure}
910: 
911: \clearpage
912: 
913: 
914: \end{document}
915: