1: % ----------------------------------------------------------------------------
2: % file: ms.tex
3: %
4: % ApJ letter: Magneto-convection in a sunspot umbra
5: %
6: % last change: Feb 21, 2006
7: %-----------------------------------------------------------------------------
8:
9: %\documentclass[manuscript]{aastex}
10: \documentclass[12pt,preprint]{aastex}
11: %\documentclass[preprint2]{aastex}
12: \usepackage{natbib}
13:
14: %\shorttitle{}
15: %\shortauthors{}
16:
17: \begin{document}
18:
19: \title{ Magneto-convection in a sunspot umbra}
20: \author{M. Sch\"ussler and A. V\"ogler\altaffilmark{1}}
21: \affil{Max Planck Institut f\"ur Sonnensystemforschung,
22: Max-Planck-Str.~2,
23: 37191 Katlenburg-Lindau, Germany}
24:
25: \altaffiltext{1}{Current address: High Altitude Observatory,
26: NCAR, P.O. Box 3000, Boulder, Colorado 80307, USA}
27:
28: \email{schuessler@mps.mpg.de, voegler@mps.mpg.de}
29:
30:
31: \begin{abstract}
32: Results from a realistic simulation of 3D radiative magneto-convection
33: in a strong background magnetic field corresponding to the conditions in
34: sunspot umbrae are shown. The convective energy transport is dominated by
35: narrow upflow plumes with adjacent downflows, which become almost
36: field-free near the surface layers. The strong external magnetic field
37: forces the plumes to assume a cusp-like shape in their top parts, where
38: the upflowing plasma loses its buoyancy. The resulting bright features
39: in intensity images correspond well (in terms of brightness, size, and
40: lifetime) to the observed umbral dots in the central parts of sunspot
41: umbrae. Most of the simulated umbral dots have a horizontally elongated
42: form with a central dark lane. Above the cusp, most plumes show narrow
43: upflow jets, which are driven by the pressure of the piled-up plasma
44: below. The large velocities and low field strengths in the plumes are
45: effectively screened from spectroscopic observation because the surfaces
46: of equal optical depth are locally elevated, so that spectral lines are
47: largely formed above the cusp. Our simulations demonstrate that nearly
48: field-free upflow plumes and umbral dots are a natural result of
49: convection in a strong, initially monolithic magnetic field.
50: \end{abstract}
51:
52: \keywords{Sun: sunspots --- Sun: magnetic fields --- Sun: photosphere
53: --- methods: numerical --- MHD}
54:
55:
56: \section{Introduction}
57:
58: Sunspot umbrae appear dark because the convective flows are suppressed
59: by the strong magnetic field \citep{Biermann:1941, Cowling:1953},
60: reducing the emitted energy (per unit time and area) in the central
61: parts of a large umbra to about 10--20\% of the average value outside
62: sunspots. However, even this strongly diminished energy flux cannot be
63: carried by radiation alone below the umbral photosphere
64: \citep{Schlueter:Temesvary:1958}, so that some form of reduced
65: convective energy transport is required to construct a consistent
66: sunspot model \citep{Deinzer:1965}. The observation of umbral dots,
67: relatively bright features of sub-arcsecond size embedded in the dark
68: umbral background, has often been taken as a signature of such
69: convective energy transport, either in the form of overstable
70: oscillations (`elevator convection') in thin vertical columns
71: \citep[e.g.,][]{Savage:1969} or as intrusions from below of non-magnetic
72: plasma into a shallow cluster-type sunspot \citep{Parker:1979a}. For
73: monolithic sunspot models, linear stability analysis indicates that
74: oscillatory convection in slender columns is the preferred mode in the
75: first few Mm depth below the umbral photosphere, where the quantity
76: $\zeta=\eta/\kappa$, the ratio of the magnetic diffusivity, $\eta$, to
77: the thermal (radiative) diffusivity, $\kappa$, has values smaller than
78: unity \citep{Meyer:etal:1974}. Systematic studies based on numerical
79: simulations under idealized conditions extend this result into the
80: nonlinear regime \citep[e.g.,][]{Weiss:etal:1990,Weiss:etal:1996,
81: Weiss:etal:2002}. Here we report on the first realistic simulations of
82: convection in sunspot umbrae with full radiative transfer and partial
83: ionization effects. These simulations can be compared directly with
84: observations.
85:
86: \section{Simulation setup}
87:
88: We have carried out ab-initio numerical simulations of 3D radiative
89: magneto-con\-vect\-ion in the near-surface layers of a sunspot umbra
90: using the {\sl MURaM} code\footnote{
91: http://www.mps.mpg.de/projects/solar-mhd/muram\_site}
92: \citep{Voegler:etal:2003,Voegler:2003, Voegler:etal:2005}. The
93: computational box extends $5760\,$km$\times 5760\,$km in the horizontal
94: directions and $1600\,$km in depth, roughly covering the range from
95: $1200\,$km below to $400\,$km above the average level of Rosseland
96: optical depth unity, $\tau_{\mathrm R}=1$. The cell size of the
97: computational mesh is $20\,$km in both horizontal directions and
98: $10\,$km in the vertical. We assume a constant magnetic diffusivity of
99: 2.8$\cdot10^6$m$^2$s$^{-1}$ and hyperdiffusivities for the other
100: diffusive processes \citep[for details, see][]{Voegler:etal:2005}.
101: Since here we are mainly interested in studying the convective processes
102: around and below $\tau_{\mathrm R}=1$ without a detailed comparison with
103: spectro-polarimetric observations, we have included only the lower parts
104: of the umbral atmosphere and also restricted ourselves to grey radiative
105: transfer. The vertical magnetic flux through the box is fixed,
106: corresponding to a horizontally averaged vertical magnetic field of
107: $2500\,$G. The thermal energy density of the inflowing matter at the
108: (open) bottom boundary has been fixed at a value of
109: $3.5\cdot10^{12}\,$erg$\cdot$cm$^{-3}$, leading to an average radiative
110: energy output (per unit area and time) through the upper boundary around
111: 17--18\% of its value outside sunspots.
112:
113: The first phase of the thermal relaxation of the model was carried out
114: in 2D for larger computational efficiency. After about 30 hours of
115: computed solar time, we continued to evolve the model in 3D for about 3
116: hours until a statistically stationary state (i.e., no trends in
117: energy flux and in horizontally averaged profiles of temperature and
118: other quantities) had developed. The simulation was then continued for
119: another 2.3 hours of solar time.
120:
121:
122: \section{Results}
123:
124: The system develops a mode of convective energy transport which is
125: dominated by non-stationary narrow plumes of rising hot plasma. The
126: strong expansion of the rising matter with height leads to a drastic
127: decrease of the magnetic field strength in the upper layers of the
128: plumes. For a snapshot from the simulation run, Fig.~\ref{fig:slices}
129: shows the brightness (vertically emerging grey intensity) together with
130: the vertical velocity and the vertical magnetic field, the latter two at
131: a height of $z=1200\,$km, which roughly corresponds to the average level
132: of $\tau_{\mathrm R}=1$. The bright features in the intensity image have
133: a typical size of a 200--300 kilometers, a lifetime of the order of 30
134: minutes, and a broad distribution of brightness values reaching up to
135: 2.4 times the average brightness in the simulated umbra. These values
136: are consistent with the results of high-resolution photometry of umbral
137: dots in the darker central parts of sunspot umbrae
138: \citep{Sobotka:Hanslmeier:2005}. Most of the simulated umbral dots have
139: an elongated shape with a central dark lane and downflows concentrated
140: at the end points. Larger dots sometimes show a threefold dark lane with
141: three downflow patches. These shapes possibly result from the fluting
142: instability in the upper parts of the low-field-strength upflow channel
143: underlying the umbral dot: the elongated features correspond to
144: azimuthal wavenumber $m=2$ while the threefold structure results from
145: the mode $m=3$.
146:
147: We have studied in some detail the structure of the umbral dot indicated
148: by the arrow in the left panel of Fig.~\ref{fig:slices}. A cut through
149: the computational box in the direction of the arrow, roughly
150: perpendicular to the dark lane, is shown in Fig.~\ref{fig:cut_y}. The
151: cut covers the upflow plume underlying the umbral dot and its immediate
152: environment.
153: %Colours represent the magnetic field strength
154: %and the arrows indicate (projected) velocity vectors, the maximum upflow
155: %speed being about 3 km$\cdot$s$^{-1}$. The three black lines correspond
156: %to the levels of $\tau_{\mathrm R}=1.$, 0.1, and 0.01 (from bottom to
157: %top).
158: The magnetic field is strongly reduced down to values of a few hundred
159: Gauss in the near-surface layers of the plume. The corresponding lateral
160: expansion of the structure is mainly in the direction perpendicular to
161: the cut shown here, leading to the elongated shape of the visible umbral
162: dot. The upflow is strongly braked around $\tau_{\mathrm R}=1$, where
163: the plasma rapidly loses its buoyancy, mainly owing to cooling by
164: radiative losses and the mean stratification becoming subadiabatic. The
165: plasma piles up and the flow turns horizontal (mainly in the direction
166: perpendicular to the cut) in the region below the cusp until a
167: quasi-stationary state with locally enhanced density and pressure is
168: established. This is illustrated in Fig.~\ref{fig:blowup}, which shows
169: the density fluctuations with respect to the horizontal average at the
170: same height in the upper central part of the cut displayed in
171: Fig.~\ref{fig:cut_y}. As a result of the enhanced density below the
172: cusp, the level of $\tau_{\mathrm R}=1$ (full line) is so much elevated
173: in the central part that it cuts through regions of lower temperature
174: (indicated by the dashed isotherms) than in the more peripheral
175: parts. This leads to the dark lane in the intensity image; a similar
176: mechanism has been proposed to explain the dark lanes observed in bright
177: penumbral filaments \citep{Spruit:Scharmer:2006}. The flow turns
178: horizontal near the surface of $\tau_{\mathrm R}=1$ and proceeds in the
179: direction of the dark lane, finally descending in narrow downflow
180: channels, which are still within the volume that the flow has largely
181: cleared from magnetic field (cf. Fig.~\ref{fig:slices}, see also
182: Fig.~\ref{fig:timeseries}).
183:
184:
185: %Fig.~\ref{fig:cuts_x} shows a cut in the direction perpendicular to that
186: %fiven in Fig.~\ref{fig:cut_y} and \ref{fig:blowup}, i.e., roughly along
187: %the dark lane in the intensity image. In this case, colours represent
188: %the temperature and the black arrows are projected velocity vectors.
189:
190:
191: An interesting feature visible in Figs.~\ref{fig:cut_y} and
192: \ref{fig:blowup} is a narrow jet-like upflow above the cusp. The
193: pressure below the cusp builds up sufficiently strongly to drive matter
194: out along the magnetic field above the cusp with a velocity of about
195: $1\,$km$\cdot$s$^{-1}$, the cusp acting like a safety valve
196: \citep{Choudhuri:1986}. Although most of the upflowing matter turns over
197: and descends in the downflow channels and only a small part escapes
198: upward, such outflows occur above most upflow plumes (see also
199: Fig.~\ref{fig:timeseries}) and could possibly affect the chromospheric
200: dynamics above sunspots. However, since the simulation does not cover
201: the higher layers of the umbral atmosphere and has a closed top boundary,
202: we cannot predict the effects on the basis of the present results.
203:
204: The profiles of vertical components of velocity and magnetic field along
205: the horizontal direction of the cut shown in Fig.~\ref{fig:cut_y} are
206: given in Fig.~\ref{fig:cuts_by_vy}. There is a strong difference between
207: the values inside and outside the upflow plume at constant {\em
208: geometrical} level, but the elevation of the lines of constant {\em
209: optical} depth together with the strong decrease of the differences with
210: height leads to much smaller variations in the range $\tau_{\mathrm
211: R}\simeq0.01\dots0.1$, where most photospheric spectral lines are
212: formed. At $\tau_{\mathrm R}=0.01$ (0.1) we find only a maximum
213: reduction of the vertical field strength by 20\% (40\%) with respect to
214: the horizontal mean of $2500\,$G and a maximum upflow velocity of 0.3
215: (0.9) km$\cdot$s$^{-1}$. Consequently, spectroscopic observations would
216: reveal neither the strong field reduction nor the high velocities in the
217: upflow plumes, even if any instrumental or seeing-related smearing and
218: straylight effects were absent. This is consistent with existing
219: observational results \citep[e.g.,][]{Socas-Navarro:etal:2004}.
220:
221: Fig.~\ref{fig:timeseries} shows the time evolution of a typical upflow
222: plume developing into an umbral dot when hot plasma reaches the surface
223: ($\tau_{\mathrm R}=1$). The development starts as a feeble upflow below
224: the surface (left panels), which develops into a strong plume once the
225: hot gas reaches the surface, cools by radiative losses, and descends in
226: adjacent narrow downflow channels, thereby forming an overturning cell. The
227: expanding rising plasma creates a `gap' in the magnetic field which
228: develops into a cusp-like configuration outlined by the dark lane in the
229: brightness images. The jet- or sheet-like upflow above the cusp driven
230: by the steeper pressure gradient of the piled-up plasma below is clearly
231: visible in the velocity cuts. Since there is no sustained buoyancy
232: driving from below to supply the strong upflow in the upper layers
233: driven by radiative losses, density and pressure decrease and the
234: magnetic field strength grows in the deeper parts and,
235: eventually, the plume fades away.
236:
237:
238: \section{Discussion}
239:
240: Our simulations show that upflow plumes and umbral dots naturally appear
241: in strong-field magneto-convection with radiative transfer. The plumes
242: start off like oscillatory convection columns below the surface, but
243: turn into narrow overturning cells driven by the strong radiative
244: cooling around optical depth unity, which leads to descending
245: low-entropy fluid. Even though the quantity $\zeta=\eta/\kappa$ is much
246: smaller than unity in the upper third of our computational
247: box\footnote{Since our numerical scheme includes a spatio-temporarily
248: varying thermal `hyperdiffusivity' \citep{Voegler:etal:2005}, the value
249: of $\zeta$ is not readily determined in the lower part of the
250: computational box, where $\kappa$ is no longer dominated by the
251: radiative diffusivity. In the simulations discussed here, the effective
252: value of $\zeta$ is somewhat larger than unity in most of the lower two
253: thirds of the box.}, this mode dominates because the overturning flow is
254: largely confined to the gap in the magnetic field created by the flow,
255: so that no significant cross-field motion is involved. Extended
256: calculations in 2D have shown that (a) the character of the
257: magneto-convective pattern and energy transport as described above is
258: unaffected if we halve the magnetic diffusivity, so that the results can
259: considered to be robust in that sense, and (b) for a run with a four
260: times larger value of $\eta$, the magneto-convection is dominated by an
261: overturning mode with slowly evolving broad cells and cross-field flow,
262: similar to what is predicted by linear analysis.
263:
264:
265: Our simulations reproduce the general properties of a sunspot umbra with
266: umbral dots. The bright features resulting from the upflow plumes can be
267: identified with (central) umbral dots; their brightness, size, and
268: lifetime correspond well to the observations
269: \citep[e.g.,][]{Socas-Navarro:etal:2004}. The large velocities and small
270: field strengths in the upflow plumes are effectively hidden from
271: spectroscopic observations by the pile-up of plasma in the upper layers
272: of the plumes, raising the formation heights of spectral lines to layers
273: located largely above the plume. A similar underlying geometry of umbral
274: dots has been suggested by \citet{Degenhardt:Lites:1993a}; their
275: assumption of a stationary, unidirectional vertical flow is not
276: supported by our simulation results, however.
277:
278: The results show that no intrusion of field-free plasma from the
279: underlying convection zone into a cluster-type shallow sunspot is
280: required explain the existence of umbral dots. In fact, a simulation
281: with an about 15\% larger value of the internal energy of the inflows
282: resulted in flux separation, i.e., the formation of large field-free
283: patches with granule-like convection surrounded by very strong, largely
284: homogeneous magnetic field \citep[cf.][]{Weiss:etal:2002}, a
285: configuration very much different from that of an undisturbed inner
286: umbra. Therefore, a hypothetical fragmentation of the magnetic field
287: into a cluster of flux strands has to occur at a level significantly
288: below our bottom boundary (located about 1.6 Mm below the solar
289: surface). However, such intrusions on a larger scale could possibly be
290: relevant for the formation of light bridges as well as for peripheral
291: umbral dots connected to bright penumbral filaments.
292:
293: We have also run a simulation with a field of $3500\,$G and the original
294: value of the internal energy carried by the inflows. In this case,
295: umbral dots very rarely occur and the average energy output (per unit
296: area) is reduced to only about 9\% of its value in the quiet Sun. This
297: simulation could possibly represent the observed dark cores of sunspot
298: umbrae, which are largely devoid of umbral dots.
299:
300:
301: %\acknowledgments
302:
303: %\bibliographystyle{apj}
304: %\bibliography{apj-jour,papref}
305: %\bibliography{joushort,papref}
306: \bibliography{ms}
307:
308: \clearpage
309:
310:
311: \begin{figure*}[ht!]
312: %\epsscale{0.75}
313: %\plotone{f1a.eps}\hspace{-6mm}
314: %\plotone{f1b.eps}\hspace{-6mm}
315: %\plotone{f1c.eps}
316: \includegraphics[scale=0.57]{f1a.eps}\hspace{-3mm}
317: \includegraphics[scale=0.57]{f1b.eps}\hspace{-3.5mm}
318: \includegraphics[scale=0.57]{f1c.eps}
319: \caption{Vertically emerging grey intensity, normalized by its
320: horizontal average (left) and cuts of the vertical velocity (middle)
321: and vertical magnetic field (right) components at a height of
322: $1200\,$km above the bottom of the simulation box ($400\,$km below the
323: top), approximately corresponding to the average level of Rosseland
324: optical depth unity ($\tau_{\mathrm R}=1$), for a snapshot from the
325: simulation run. The length unit is Mm. The bright features in the
326: intensity image can be identified with umbral dots. They are caused by
327: strong upflows in regions of significantly reduced magnetic field. The
328: arrow indicates the umbral dot studied in more detail below.}
329: \label{fig:slices}
330: \end{figure*}
331:
332: \clearpage
333:
334: \begin{figure}[htb]
335: %\epsscale{0.6}
336: \includegraphics[scale=0.9]{f2.eps}
337: %\plotone{f2.eps}
338: \caption{Vertical cut through the simulation box in the direction
339: indicated by the arrow in Fig.~\ref{fig:slices}, nearly
340: perpendicular to the dark lane of the umbral dot. Colours
341: indicate magnetic field strength, the arrows represent
342: (projected) velocity vectors. The longest arrow corresponds to
343: a velocity of $2.7\,$km$\cdot$s$^{-1}$. The lines indicate the
344: levels of constant optical depth $\tau_{\mathrm R}=1.$, 0.1,
345: and 0.01, respectively (from bottom to top). The upper part of
346: the plume has developed a cusp-like shape with a strongly
347: decelerated uflow and a weak magnetic field. The flow turns
348: horizontal mainly in the direction perpendicular to the plane
349: of the cut.}
350: \label{fig:cut_y}
351: \end{figure}
352:
353:
354: \begin{figure}[htb]
355: %\epsscale{1.0}
356: \plotone{f3.eps}
357: \caption{Upper central part of the cut shown in Fig.~\ref{fig:cut_y}.
358: Colours represent the density fluctuation with respect to the horizontal
359: mean at the same height. The full line indicates the level of
360: $\tau_{\mathrm R}=1$ while the dashed lines are isotherms. The piling up
361: of matter raises the surfaces of constant optical depth into
362: low-temperature regions near the cusp, leading to the appearance of a
363: dark lane with a maximum contrast of about 15\% in the intensity
364: image. The longest velocity arrow corresponds to a speed of
365: $2.7\,$km$\cdot$s$^{-1}$.}
366: \label{fig:blowup}
367: \end{figure}
368:
369: \clearpage
370:
371: \begin{figure}[htb]
372: \epsscale{.7}
373: \plotone{f4a.eps}\\
374: \plotone{f4b.eps}
375: \caption{Profiles of the vertical magnetic field (upper panel) and
376: velocity (lower panel) components along the horizontal direction in
377: the cut shown in Fig.~\ref{fig:cut_y}. In both panels, the thick solid
378: curve corresponds to constant geometrical height $z=1200\,$km and the
379: other three lines indicate the levels of constant (Rosseland) optical
380: depth $\tau_{\mathrm R}=1$ (thin solid line), 0.1 (dotted line), and
381: 0.01 (dashed line). While the variation between the quantities inside
382: and outside the upflow plume (umbral dot) is very large at the same
383: geometrical level, it is strongly reduced at equal optical depth.}
384: \label{fig:cuts_by_vy}
385: \end{figure}
386:
387:
388: \begin{figure}[htb]
389: \epsscale{1.00}
390: \plotone{f5a.eps}\\
391: \plotone{f5b.eps}\\
392: \plotone{f5c.eps}
393: \caption{Time evolution of an upflow plume developing into an umbral
394: dot. The panels show, from left to right, the development of the
395: brightness (top row) and of vertical cuts roughly along the dark lane
396: in the brightness images. Shown are the vertical velocity (middle row,
397: blue: upflow, red: downflow) and the vertical magnetic field (bottom
398: row), the cuts covering the full depth of the simulation box and
399: $820\,$km in horizontal direction. The colour and greyscale schemes
400: are the same as those used in Fig.~\ref{fig:slices}, albeit with
401: different ranges: $I/\langle I\rangle = 0.7\dots2.25$, $v_z
402: =-1.2\dots4.1\,$km$\cdot$s$^{-1}$, $B_z = -0.3\dots3.8\,$kG. The
403: snapshots are taken at the times (in minutes, relative to the first
404: snapshot) indicated below the bottom panel. The umbral dot is clearly
405: detectable as a bright structure for at least 25 minutes. }
406: \label{fig:timeseries}
407: \end{figure}
408:
409: \end{document}
410: