1: %\documentclass{aastex}
2: \documentclass[12pt,preprint]{aastex}
3:
4: \newcommand{\bb}{\begin{equation}}
5: \newcommand{\ee}{\end{equation}}
6:
7: \shorttitle{Sunspot Oscillations}
8: \shortauthors{Rajaguru et al.}
9:
10: \begin{document}
11:
12: \title{Radiative transfer effects on Doppler measurements as sources of surface effects in
13: sunspot seismology}
14: \author{S.P. Rajaguru\altaffilmark{1}, K. Sankarasubramanian\altaffilmark{2},
15: R. Wachter\altaffilmark{1} and P.H. Scherrer\altaffilmark{1}}
16: \altaffiltext{1}{W.W. Hansen Experimental Physics Laboratory, Stanford University, Stanford CA 94305}
17: \altaffiltext{2}{Space Astronomy and Instrumentation Division, ISRO Satellite Centre, Bangalore, India}
18:
19: \begin{abstract}
20: %
21: We show that the use of Doppler shifts of Zeeman sensitive spectral lines to observe waves
22: in sunspots is subject to measurement specific phase shifts arising from,
23: (i) altered height range of spectral line formation and the propagating
24: character of p mode waves in penumbrae, and (ii) Zeeman broadening and splitting.
25: We also show that these phase shifts depend on wave frequencies, strengths and line of sight
26: inclination of magnetic field, and the polarization state used for Doppler measurements.
27: We discuss how these phase shifts could contribute to local helioseismic measurements of
28: 'surface effects' in sunspot seismology.
29: %
30: \end{abstract}
31: %
32: \keywords{Sun: helioseismology --- Sun: magnetic fields --- Sun: oscillations --- radiative transfer --- sunspots}
33:
34: \section{Introduction}
35: \label{sec:intro}
36:
37: The solar oscillations are measured with better signal-to-noise from Doppler shifts of a spectral line than
38: intensity fluctuations. Existence of reliable algorithms to estimate Doppler shifts
39: from line intensities (filtergrams) scanned over a few locations across a line (MDI; \citet{scherreretal95}),
40: or a fast Fourier tachometric scanning of a line (GONG; \citet{harveyetal94}), makes possible deriving velocity
41: images of solar surface frequent enough to study the solar oscillations.
42: Above methods of estimating line of sight velocities from Doppler shifts assume
43: rigid displacements of a spectral line in wavelength in response to material motions.
44: However, gradients along line of sight in solar velocity and other physical quantities
45: introduce asymmetries in line profile leading to Doppler cross-talk of signals from
46: different layers in the atmosphere. In quiet Sun, the five minute band p mode oscillation
47: signals are evanescent at observation heights and hence oscillation phases are only minimally
48: affected by height gradients.
49: In active regions, not only the vastly different thermal
50: and magnetic structure complicate Doppler shift measurements through changes in line profile shapes but also
51: altered character of waves in line forming layers.
52:
53: In this Letter, using a temporal sequence of spectropolarimetric observations of a sunspot
54: we demonstrate: (i) propagating character of p modes in the penumbra and
55: extended height range of spectral line formation cause phases of waves to depend strongly
56: on intensity level within a line from where Doppler velocities are measured,
57: (ii) sampling Zeeman split portions near the line core causes
58: filtergraph (MDI/SOHO) or bisector methods
59: for Doppler velocity estimation to introduce spurious phase shifts in the waves measured.
60: Wave propagation signatures, in the first case above, which have recently been studied
61: in chromospheric lines \citep{finsterleetal04}, are also shown to differ
62: in the Doppler shifts of Stokes I and polarized components (circular and linear).
63: We discuss how the above findings are related to recent
64: local helioseismic measurements of 'surface effects' \citep{schunkeretal05,zhaoetal06}.
65:
66: \section{Data and Analysis Methods}
67: The observations were carried out using the Advanced Stokes Polarimeter (ASP) at the
68: Dunn Solar Telescope of the National Solar Observatory at Sacramento Peak, Sunspot, New Mexico.
69: A medium sized sunspot (NOAA AR0750, diameter $\approx$ 16 Mm) located close to disk center
70: was observed on April 12, 2005, in mid-photospheric spectral line Ni {\sc i} ($\lambda$=6768 \AA,
71: used by MDI/SOHO and GONG).
72: We have made a temporal sequence of Stokes vectors of this line by scanning
73: the sunspot over consecutive (in $x$-direction) 14 slit positions oriented N-S ($y$-direction)
74: in one minute and repeating it for about 100 minutes (with a solar image tracker fixing the field of view).
75: The N-S extent of slit covers about 64 Mm with a resolution of 0.375$^{"}$ with the spot at the center
76: and, with a slit width of 0.6$^{"}$, the 14 slit positions (x-direction) cover mainly umbral region.
77:
78: We do Milne-Eddington (M-E) inversions of Stokes vectors ($I,Q,U,V$'s) to obtain line of sight velocities
79: $v(x,y,t)$, total magnetic field strengths $B(x,y,t)$, line of sight inclination $\gamma(x,y,t)$ and
80: azimuth $\psi(x,y,t)$ of magnetic field among other standard inverted quantities \citep{sku-lites87}.
81: The M-E inversion procedure is weighted towards polarized component within magnetic
82: field and provides velocity estimates from the line core position which is used as a free parameter
83: \citep{sku-lites87,jefferiesetal89}.
84: Since we want to examine how wave phases change with height within line
85: formation layers, how they are seen in polarized and unpolarized light, and how they contribute to
86: local helioseismology measurements of 'surface magnetic effects', we also obtain line of sight velocities
87: from the following methods: (i) Doppler shifts of bisector points at different intensity levels within $I$
88: \citep{keil-yackovich81,cavallinietal85,rimmele95},
89: and those within circular (average of $I+V$ and $I-V$ velocities, hereafter denoted CP) and linear
90: ($I+Q$, denoted LP1) polarized components \citep{sankar-rimmele02,dtiniesta03},
91: and (ii) the two Doppler filtergram measurements in circular (CP) and linear (LP1) polarized
92: components employed by MDI/SOHO \citep{scherreretal95}.
93:
94: Phase shifts between any two velocity signals, $v_{1}(t)$ and $v_{2}(t)$, are calculated according to,
95: \begin{equation}
96: \phi_{1,2}(\omega)=tan^{-1}\left (\frac{Im[v_{1}(\omega)v^*_{2}(\omega)]}{Re[v_{1}(\omega)v^*_{2}(\omega)]}
97: \right) ,
98: \label{eq:phase}
99: \end{equation}
100: where $v_{1}(\omega)$ and $v_{2}(\omega)$ are the Fourier transforms of the two time series. In the
101: above convention, positive values for $\phi_{1,2}$ at positive $\omega$ mean $v_{1}(t)$ is
102: advanced in phase with respect to $v_{2}(t)$.
103: Our aim here is to study how $\phi$ between velocities from different bisector levels as well as that between
104: different measurements vary over the sunspot as a function of position and $\omega$, i.e. $\phi(x,y,z,\omega)$,
105: due to propagating waves and altered line profile shapes (Zeeman splitting). The vertical
106: co-ordinate $z$ represents height within line forming layers and we use intensity levels
107: (i.e., bisector points) within the line to scan it. Based on the plane-parallel VAL-C model of quiet solar
108: atmosphere, the Ni {\sc i} line formation is reported to span a height range of
109: 18 km (continuum) -- 288 km (line core) ($z$=0 being referenced to continuum optical
110: depth $\tau_{c}$=1) \citep{nortonetal06,bruls93}. We do not attempt full Stokes profiles inversions
111: to determine $z$ dependences of $\phi$;
112: this is not only computationaly expensive (as we have too many profiles from the $x,y,t$
113: scans of Stokes vector) but also lead to noisier signals than those from direct
114: bisector analyses. In quiet-Sun region
115: the main p mode band acoustic waves ($\omega < \omega_{ac} \approx $5.3 mHz, the acoustic
116: cut-off frequency) are largely reflected down into the solar interior, and hence
117: $\phi(x,y,z,\omega) \approx$ constant = 0; at $\omega > \omega_{ac}$, however, because
118: of wave propagation, $\phi(x,y,z,\omega) \approx \phi(z,\omega) \neq$ 0.
119: Within the sunspot, we study $x$ and $y$ dependences of $\phi$ using
120: temporal averages of M-E inverted $B(x,y,t)$ and $\gamma(x,y,t)$ shown in Figure 1.
121:
122: \section{Surface magnetism effects due to propagating waves and Zeeman splitting}
123:
124: In order to analyse phase shifts due to the propagating nature of waves,
125: we calculate bisector velocities from 9 equally spaced intensity levels within $I$
126: and CP profiles, $I_{1}$=0.1$I_{c}$, $I_{2}$=0.2$I_{c}$,..., $I_{9}$=0.9$I_{c}$, where $I_{c}$ is
127: the continuum level. Using the time series of velocities from different
128: bisector levels, we calculate phase shifts between them using Equation \ref{eq:phase}.
129: When the line is Zeeman split, formation heights
130: of bisector levels of $I$ profiles near the core are significantly altered and any asymmetries in
131: depths of Zeeman components further corrupt the velocity signals. Bisector points of individual
132: CP components are largely free of the above uncertainties. In view of the
133: above, we use M-E inverted velocities as representative of the highest layer within line formation
134: region, and we show that this is indeed the case by comparing them with CP bisector velocities
135: from near the core for the Ni {\sc i} line.
136:
137: The phase shifts $\phi(x,y,z,\omega)$
138: exhibit random fluctuations over space as well as in $\omega$ due to the inherent stochastic
139: nature of the solar velocity field and due to noise from various sources. Fluctuations over $\omega$
140: are reduced by taking median values of $\phi$ over the p mode band (2 - 5 mHz) or
141: over bands of 1 mHz width centered at 3, 4 and 5 mHz (to study any frequency dependence). Signals
142: over space are studied using the temporal averages $B(x,y)$ and $\gamma(x,y)$ (Figure 1):
143: we average the phase shifts over all pixels that fall within a 25 G window
144: of $B$ or 2$^{\circ}$ of $\gamma$ and assign the average to all those pixels (spatially) and to the
145: central values of $B$ and $\gamma$. The resulting spatial maps of $\phi$ are essentially one dimensional
146: functions of $B$ or $\gamma$ with any azimuthal variations due to such variations in $B$ or $\gamma$
147: averaged out. In Figure 2, we plot $\phi$ between mid-level
148: ($I_{5}=0.5I_{c}$) bisector velocities and those from other levels: panels $a$ and $b$
149: show median $\phi^{CP}_{5,i}$(i=1,3,7,9) (superscripts identify the profiles used and subscripts
150: the pair of levels that the phase shifts come from) over the full p mode band as a function of
151: $B$ and $\gamma$, respectively,
152: panel $c$ shows $\phi^{CP}_{5,1}$ and $\phi^{CP}_{5,9}$ as a function of $B$ for different frequency bands,
153: and panel $d$ compares $\phi^{CP}$ and $\phi^{I}$ between levels (5,1) and (5,9) as a function of $B$ for
154: the full p mode band. For clarity, we have plotted error bars only at representative values of
155: $B$ or $\gamma$. The error bars correspond to standard deviations of $\phi$ over the pixels that fall within
156: the bin sizes of $B$ or $\gamma$. The spatial maps that are obtained after binning over 25 G windows in line
157: of sight magnetic field $B_{z}=B cos(\gamma)$ are shown in Figure 3:
158: in panels $a$ and $b$, respectively, are $\phi^{I}_{5,i}$ and $\phi^{CP}_{5,i}$, where $i=$M-E,1,2 and 9,
159: in panel $c$ are $\phi^{I,CP}$ that are between same level bisector velocities from CP and $I$ profiles,
160: and in panel $d$ are $\phi^{CP}_{1,9}$ and $\phi^{I}_{1,9}$ in the 2.5 - 3.5 mHz and 3.5 - 4.5 mHz frequency bands.
161: Changes in $\phi$ against $B$ or $B_{z}$ and the bisector intensity levels, in Figures 2 and 3,
162: clearly show that these are due to propagating waves, even within the main p mode band, and that propagation is
163: mainly confined to penumbral region.
164: In quiet Sun pixels ($B <$ 200 G), $\phi$ due to wave propagation into higher layers at $\omega > \omega_{ac}$,
165: though much smaller than that seen in the penumbra, is noticeable in panels $c$ and $d$ of Figure 2.
166: Both Stokes $I$ and CP profiles' velocities show propagating waves but the signals are larger in CP
167: bisector velocities indicating that propagation is mostly in magnetized channels (panel $d$ of Figure 2 and
168: panels $a$ and $b$ of Figure 3). Phase shifts between
169: same level bisector velocities from $I$ and CP profiles, shown in panel $c$ of Figure 3, further substantiates
170: the above inference.
171: The $\phi^{I,CP}$ from wing level (7 and 9, the right two panels in panel $c$ of Figure 3)
172: bisector velocities are very small, suggesting that differences in formation heights of wings of $I$ and
173: CP profiles is small.
174:
175: The above results show that if we have to
176: minimize contributions from propagating waves we should restrict Doppler measurements to wings.
177: In panel $d$ of Figure 2 and panel $a$ of Figure 3, it is seen that the Stokes $I$ velocities introduce spurious
178: advanced phases within the umbral and umbral - penumbral boundary region, when bisector velocities from
179: near the core are involved: the fallen or collapsed $\phi^{I}_{5,1}$ plotted as red triangles in panel $d$ of Figure 2,
180: and the scrambled phase signals seen in the middle two maps of panel $a$ in Figure 3. We attribute the above
181: to changes in $I$ profile shapes due to Zeeman splitting, as these signals are not present or weak in CP
182: measurements. We confirm this using the M-E inverted velocities, which model the Zeeman splitting and retrieve
183: velocities; the left most map in panel $a$ of Figure 3 show that $\phi^{I}_{5,ME}$ is free of the above discussed
184: artifacts, and also matches well with map of $\phi^{CP}_{5,ME}$ in panel $b$ of Figure 3.
185: The similar values and structure of $\phi^{CP}_{5,1}$ and $\phi^{CP}_{5,2}$ to that of $\phi^{CP}_{5,ME}$ further
186: clarifies that wave propagation is confined to magnetized medium. Interestingly, the M-E velocities
187: show negative changes for $\phi^{I}_{5,ME}$ and $\phi^{CP}_{5,ME}$ in the outer penumbra (panels
188: $a$, $b$ of Figure 3), which we identify as a possible signature of downward propagating waves along
189: field lines that return back to the photosphere.
190:
191: The MDI measurements \citep{scherreretal95} use CP and LP1 profiles:
192: most full-disk observations are based on LP1
193: and almost all helioseismic holography results \citep{braun-lindsey99,braunetal04,schunkeretal05,
194: lindseyandbraun05a,lindseyandbraun05b} on
195: surface magnetism effects are based on these data; on the other hand, most time-distance helioseismology
196: studies of sub-surface structure and flows \citep{sashaetal00,zhaoetal01} and the surface
197: magnetism effects \citep{zhaoetal06} have used the MDI CP hi-res data. Here, we have derived
198: Doppler velocities from both of these MDI measurements by executing the MDI algorithms (which the
199: onboard processor of MDI uses). In quiet Sun, i.e. when there are no strong magnetic fields to
200: modify the line profile through Zeeman effect, both of these MDI measurements reduce to deriving
201: Doppler velocities from unpolarized Stokes $I$ and hence are the same. However, as demonstrated in Figures 2 and 3,
202: propagating waves introduce phase shifts, and to estimate how much of these are present in MDI measurements
203: we have calculated phase shifts between MDI CP and LP1 velocities and wing level (level 7) bisector
204: velocities from $I$, denoted respectively as $\phi^{CP}_{MDI}$ and $\phi^{LP1}_{MDI}$.
205: The results are in Figure 4: panel $a$ shows $\phi^{LP1}_{MDI}$
206: and panel $b$ shows $\phi^{CP}_{MDI}$ over bands of 1 mHz width centered at frequencies of
207: 3, 4 and 5 mHz and also the full p mode band (2 - 5 mHz). We found from bisector velocities from near the
208: core that Zeeman splitting causes an advancement of phases within umbra (panel $a$ and $d$ in Figure 3);
209: since LP1 profile within umbra is the same as $I$ (because $\gamma$ is close to zero), similar advanced
210: phases are expected from MDI LP1 measurements. This is clear in panel $a$ of Figure 4.
211: In summary, for the medium size spot observed, surface effects due to wave propagation in the penumbra
212: cause a phase lag in the range of 5 - 10 seconds in both MDI LP1 and CP measurements and the Zeeman
213: split profiles cause positive phase shifts of the order of 8 seconds in umbra in MDI LP1 measurements.
214:
215: \section{Discussions and conclusion}
216: Recent studies in sunspot seismology \citep{schunkeretal05,lindseyandbraun05a,lindseyandbraun05b,zhaoetal06}
217: have recognised the role of 'surface magnetic effects' that contribute to seismically measured
218: phases or travel times of acoustic waves.
219: We have shown that phases of acoustic waves observed within sunspots using Doppler shifts
220: of a spectral line have signatures of physical changes that waves undergo within line forming
221: layers as well as of systematics in Doppler measurements induced by Zeeman split profiles.
222: We have also shown that wave propagation effects and systematics from
223: split portions of the line profiles are minimal when Doppler velocity measurements are restricted
224: to wings of a spectral line.
225: Since a seismologically correct determination of phase shifts
226: require separating the above surface effects in the measurements, our latter results
227: above show that such a correction can be carried out in the observation procedure itself: within sunspots,
228: the Doppler shift measurements could be restricted to wings of spectral lines. In the case of MDI/SOHO,
229: the onboard algorithm combines all the filtergrams
230: to determine Doppler velocities. Since the filter positions
231: are fixed with respect to the central wavelength, when
232: the line is wider due to Zeeman splitting, there is a systematic sampling of split
233: core region of the profiles and hence higher atmospheric layers.
234: This causes phase lags in waves from penumbral region and spurious advanced
235: phases within umbra when LP1 profile is used (most full-disk measurements).
236: Correcting the MDI measurements for such surface effects,
237: hence, requires maps such as the ones derived here (Figure 4).
238: A detailed 'calibration' of phase shifts due to surface effects against
239: $B$ or $\gamma$ require similar studies involving spots of various sizes.
240:
241: We have not assessed how the surface effects studied here would depend on positions across the solar disk.
242: However, the present finding that propagating
243: waves in penumbra cause a phase shift depending on height of observation
244: points to similar effects when there are line of sight changes in optical depth. Such changes
245: are expected for sunspots located away from disk center, due to the Wilson depression, and
246: this combined with the field line alligned wave propagation can introduce opposite changes,
247: in measured phases, in the limb side (deeper level of observation) and center side (higher level
248: of observation) penumbrae. We suggest that such effects on the
249: phase shifts are a likely contributor to the azimuthally varying 'inclined magnetic field
250: effect' of \citet{schunkeretal05} and \citet{zhaoetal06}. The magnitude of surface effect signal
251: that we have, for the medium sized spot, falls in the range of 5 - 15 seconds (Figures 2, 3 and 4)
252: and it points to larger phase shifts for larger spots. We note
253: that, for fairly large sunspots, the magnitude of travel time perturbations measured in
254: time-distance helioseismic measurements \citep{duvalletal96,sashaetal00,zhaoetal01},
255: as well as the vantage dependent control correlation phases \citep{schunkeretal05},
256: fall in the range of 30 - 60 seconds. Our results
257: on surface effect signal due to Zeeman splitting in umbral region (Figures 3 and 4) call
258: for a careful study of active regions in one or more magnetically insensitive photospheric lines.
259:
260: \acknowledgments
261: This work is supported by NASA grants NNG05GH14G to $SOHO$ MDI project and NNG05GM85G to Living With a Star
262: (LWS) program at Stanford University.
263:
264: \begin{thebibliography}{}
265: \bibitem[Braun \& Lindsey(1999)]{braun-lindsey99} Braun, D.C., \& Lindsey, C.
266: 1999, \apj, 513, L79
267: \bibitem[Braun et al.(2004)]{braunetal04} Braun, D.C., Lindsey, C., \& Birch, A.C. 2004, BAAS, 204, 530
268: \bibitem[Bruls(1993)]{bruls93} Bruls, J.H.M.J. 1993, \aap, 269, 509
269: \bibitem[Cavallini et al.(1985)]{cavallinietal85} Cavallini, F., Cepatelli, G., \& Righini, A. 1985,
270: \aap, 143, 116
271: \bibitem[del Toro Iniesta (2003)]{dtiniesta03} del Toro Iniesta, J. C. 2003, Introduction to
272: Spectropolarimetry, Cambridge Univ. Press, p.149-164, Cambridge
273: \bibitem[Duvall et al.(1996)]{duvalletal96} Duvall, T. L., Jr., D'Silva, S., Jefferies, S. M., Harvey, J. W.,
274: \& Schou, J. 1996, \nat, 379, 235
275: \bibitem[Finsterle et al.(2004)]{finsterleetal04} Finsterle, W., et al. 2004, \solphys, 220, 317
276: \bibitem[Harvey et al.(1994)]{harveyetal94} Harvey, J., et al. 1988, in ESA proc. of 'Seismology
277: of the Sun and sun-like stars', p203
278: \bibitem[Jefferies et al.(1989)]{jefferiesetal89} Jefferies, J., Lites, B. W., \& Skumanich, A. 1989,
279: \apj, 343, 920
280: \bibitem[Keil \& Yackovich(1981)]{keil-yackovich81} Keil, S. L., \& Yackovich, F. H. 1981, \solphys, 69, 213
281: \bibitem[Kosovichev et al.(2000)]{sashaetal00} Kosovichev, A.G., Duvall, T.L., Jr.,
282: Scherrer, P.H. 2000, \solphys, 192, 159
283: \bibitem[Lindsey \& Braun(2005a)]{lindseyandbraun05a} Lindsey, C., \& Braun, D.C. 2005a, \apj, 620, 1107
284: \bibitem[Lindsey \& Braun(2005b)]{lindseyandbraun05b} Lindsey, C., \& Braun, D.C. 2005b, \apj, 620, 1118
285: \bibitem[Norton et al.(2006)]{nortonetal06} Norton, A.A. et al. 2006, arXiv:astro-ph 0608124
286: \bibitem[Rimmele(1995)]{rimmele95} Rimmele, T. 1995, \aap, 298, 260
287: \bibitem[Sankarasubramanian \& Rimmele(2002)]{sankar-rimmele02} Sankarasubramanian, K. \& Rimmele, T. 2002,
288: \apj, 576, 1048
289: \bibitem[Scherrer et al.(1995)]{scherreretal95} Scherrer, P.H., et al. 1995, \solphys, 162, 219
290: \bibitem[Schunker et al.(2005)]{schunkeretal05} Schunker, H., Braun, D.C., Cally, P.S., \& Lindsey, C. 2005,
291: \apj, 621, L149
292: \bibitem[Skumanich \& Lites(1987)]{sku-lites87} Skumanich, A., \& Lites, B. W. 1987, \apj, 322, 473
293: \bibitem[Zhao et al.(2001)]{zhaoetal01} Zhao, J., Kosovichev, A.G., \& Duvall, T.L.,Jr.
294: 2001, \apj, 557, 384
295: \bibitem[Zhao et al.(2006)]{zhaoetal06} Zhao, J., \& Kosovichev, A.G. 2006, \apj, 643, 1317
296: \end{thebibliography}
297:
298: \clearpage
299:
300: \begin{figure}
301: \epsscale{0.8}
302: \plotone{f1.eps}
303: \caption{Maps of continuum intensity $I_{c}$ and temporal averages of M-E inverted $B$ and $\gamma$
304: over the sunspot from Ni {\sc i} observations.}
305: \label{fig:1}
306: \end{figure}
307:
308: \clearpage
309:
310: \begin{figure}
311: \epsscale{1.0}
312: \plottwo{f2a.eps}{f2b.eps}
313: \caption{Phase shifts between velocities from different bisector levels of $CP$ and $I$ profiles of
314: Ni {\sc i} line: panels $a$ and $b$ show $\phi^{CP}$ against $B$ and $\gamma$, respectively,
315: for the full p mode band (2-5 mHz), panel $c$ shows $\phi^{CP}_{5,1}$ and $\phi^{CP}_{5,9}$ over different
316: frequency bands, and panel $d$ compares $\phi^{CP}$ and $\phi^{I}$ from the bisector levels (5,1) and (5,9)
317: for the full p mode band; see text for further details}
318: \label{fig:2}
319: \end{figure}
320:
321: \clearpage
322:
323: \begin{figure}
324: \figurenum{3}
325: \plottwo{f3a.eps}{f3b.eps}
326: \caption{Maps of phase shifts of waves observed using Ni {\sc i} line; see text for details}
327: \label{fig:3}
328: \end{figure}
329:
330: \clearpage
331:
332: \begin{figure}
333: \figurenum{4}
334: \epsscale{0.9}
335: \plotone{f4.eps}
336: \caption{Maps of phase shifts due to 'surface magnetic effects' in MDI linear (LP1, full disk) and circular polarization
337: (CP, hi-res) measurements.}
338: \label{fig:4}
339: \end{figure}
340:
341: \end{document}
342:
343: