MAGNETIC FIELD PROPERTIES OF FLUX CANCELLATION SITES 1

Transcription

1 The Astrophysical Journal, 671:990Y1004, 2007 December 10 # The American Astronomical Society. All rights reserved. Printed in U.S.A. A MAGNETIC FIELD PROPERTIES OF FLUX CANCELLATION SITES 1 M. Kubo 2,3 and T. Shimizu 3 Received 2006 May 20; accepted 2007 September 4 ABSTRACT It has been frequently observed in longitudinal magnetograms that magnetic elements disappear when a magnetic polarity element approaches and collides with another polarity element. We examine 12 collision events observed with the Advanced Stokes Polarimeter. We find formation of new magnetic connection between the colliding opposite polarity elements both in the photosphere and in the corona. In some cases, the opposite polarity elements to be collided appear at different times and at widely separated positions. Magnetic fields horizontal to the solar surface are spontaneously formed on the polarity inversion line ( PIL) between such colliding elements, and transient bright X-ray loops connecting the opposite polarity elements appear. We suggest that formation of the coronal loops and the photospheric horizontal fields are due to magnetic reconnection between the colliding field lines, possibly at multiple locations with different heights. We also find that a global change in the direction of the photospheric horizontal fields between the colliding elements occurs in association with formation and disappearance of H dark filaments. Initial horizontal fields perpendicular to the PIL become parallel to the PIL, when dark filaments are observed along the PIL. They return to being perpendicular to the PIL at around the time of the disappearance of the dark filament. Subject headinggs: Sun: corona Sun: evolution Sun: magnetic fields Sun: photosphere Sun: prominences Online material: color figures 1. INTRODUCTION Total magnetic flux on the entire solar surface is estimated to be of the order Y10 24 Mx (Harvey 1994; Hagenaar 2001; Krivova & Solanki 2004), and a rate of magnetic flux emergence in ephemeral regions is estimated to be an order of Y10 22 Mx hr 1 (Harvey et al. 1975; Chae et al. 2001; Hagenaar 2001). This implies that magnetic fields in the quiet region are replaced with emerging magnetic flux in a short period lasting from several hours to several days (Schrijver et al. 1998; Hagenaar 2001). How are magnetic fields removed from the solar surface? A related observational signature for the field removal is magnetic flux cancellation, in which one magnetic polarity element collides with another polarity element, followed by disappearance of the magnetic elements in the longitudinal magnetograms ( Martinet al. 1985; Livi et al. 1985). The magnetic flux cancellation is observed notonlyinthequietsunbutalsoinactiveregions(martinetal. 1985; Chae et al. 2004). Some coronal activities have been reported at the flux cancellation sites. Most X-ray-bright points are associated with collisions of the opposite polarity flux (Webb et al. 1993; Harvey 1996). Microflares (Chae et al. 1999; Zhang et al. 2000; Shimizu et al. 2002) and surges (Chae et al. 1999; Zhang et al. 2000; Yoshimura et al. 2003) are also observed at the regions where preexisting magnetic flux collides with the opposite polarity of emerging flux. Dark filaments are observed in H above the boundary between the opposite polarity magnetic fields which are to cancel each other (e.g., Martin 1998). Major flares accompanied by filament eruptions are associated with collisions between the opposite polarity elements (e.g., Zhang et al. 2001; Somov et al. 2002). 1 This work was completed while the author was affiliated with the National Astronomical Observatory of Japan and University of Tokyo. 2 High Altitude Observatory, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency, Yoshinodai, Sagamihara, Kanagawa , Japan. 990 Zwaan (1987) illustrates three cases for removing magnetic flux from the photosphere with collision of the opposite polarity elements. When magnetic connection exists above the photosphere between colliding magnetic elements, the disappearance of the magnetic elements can be caused by retraction of magnetic fields to below the photosphere. When one bipole collides with another bipole, magnetic reconnection between the two bipoles is required to remove the magnetic flux from the photosphere. When reconnection occurs above the photosphere, an submerging (retracting) -shaped loop is formed ( Fig. 1a). On the other hand, an emerging U-shaped loop is formed when reconnection occurs below the photosphere ( Fig. 1b). Magnetic reconnection may not be needed for formation of the emerging U-shaped loop if the two different bipoles have connection below the photosphere (Parker 1984; Lites et al. 1995). When two different bipoles collide, new transverse magnetic fields connecting the canceling magnetic elements are formed in both cases of the -shaped loop and the U-shaped loop. Using vector magnetograms at Huairou Solar Observing Station, Wang & Shi (1993) find that transverse magnetic fields connecting canceling magnetic elements are formed for one or two cases of five cancellation sites. Chae et al. (2004) observe two cancellation sites with the Advanced Stokes Polarimeter (ASP; Elmore et al. 1992): magnetic fields are nearly horizontal to the solar surface at the polarity inversion line separating two canceling magnetic elements with field strength of about 800 G. Motion of horizontal magnetic field lines formed between colliding opposite polarity elements is important to reveal physical process for the magnetic flux cancellation. Downward motion of the horizontal field lines must be observed in the submerging -shaped loop ( Fig. 1a), while upward motion must be observed in the emerging -shaped loop (Fig. 1b). Chae et al. (2004) observe the downward motion of about 1 km s 1 in the two cancellation sites, and Yurchyshyn & Wang (2001) observe the upward motion of about 0.6 km s 1 in a cancellation site. Harvey et al. (1999) compare timing of 45 magnetic cancellations observed with chromospheric magnetograms and photospheric magnetograms.

2 MAGNETIC FIELD PROPERTIES OF FLUX CANCELLATION SITES 991 Fig. 1. Magnetic flux cancellation between two bipoles. The cross symbol represents a reconnection site. (a) -shaped loop submerges to below the photosphere. (b) U-shaped loop emerges from below the photosphere. (c) Magnetic reconnection takes place at multiple locations with different heights. They suggest that magnetic fields submerge for at least about half of the cancellations from the observations of magnetic flux disappearance in the chromosphere before in the photosphere. Do these observations show actual submerging and emerging motions of the horizontal field lines formed by the magnetic flux cancellation? We also focus on Doppler velocities around the polarity inversion line in the cancellation sites. There is lack of detailed observations telling how vector magnetic field evolves during the magnetic collision and cancellation. In this study, we describe the temporal change of the magnetic and velocity fields in the magnetic cancellation sites. Associated coronal activities are observed with the Transition Region and Coronal Explorer (TRACE; Handy et al. 1999) and the Soft X-Ray Telescope (SXT; Tsuneta et al. 1991) aboard Yohkoh (Ogawara et al. 1991). Dark filaments are observed in H line along the polarity inversion line between the colliding opposite polarity elements. We show that the formation and disappearance of dark filaments are intimately related to change of the photospheric magnetic fields. In x 5 the temporal evolution of one typical case is presented in detail. Properties of all the events are summarized in x 6. Finally, we discuss how the magnetic field evolves during the flux cancellation and then disappears from the photosphere in x OBSERVATIONS We observed 12 collision events of opposite polarity elements in four active regions with the ASP, Yohkoh SXT, TRACE, and Michelson Doppler Imager (MDI; Scherrer et al. 1995) (see Table 1). We excluded collisions of extremely small magnetic elements inside emerging flux regions or in moat regions, because their evolution was too complicated to identify colliding elements. The ASP observations provided the full Stokes profile of photospheric Zeeman-sensitive Fe i k and Fe i k lines with high accuracy. The spatial distribution of the magnetic field vector and thermodynamic parameters were derived from the full Stokes vector. The field of view along the slit was with the pixel size of A typical mapping to cover the active region took 20Y 30 minutes for 200Y300 steps with slit width of When the active region was large, two or three mappings were performed to cover the entire active region. SOHO MDI continued to provide full-disk longitudinal ( lineof-sight) magnetograms with pixel sampling of The cadence of MDI observations was 1 minute during most of the ASP observing period (15 : : 00 UT) and 96 minutes during the other period. TRACE obtained coronal images in Fe ix/x for active region NOAA 9231, and in Fe xii for active regions NOAA 9907 and NOAA The Fe ix/x line was formed with the temperature of about 1 MK and the Fe xii line was with about 1.5 MK. The field of view was ; with pixel size of The SXT was sensitive to hot plasma with temperatures above 2 MK. The active region NOAA 9231 was continuously tracked in partial frame image (PFI) mode with pixel size of A few PFI images were available for the observations of an active region NOAA The full-disk coronal images ( FFIs) with halfresolution ( pixel 1 ) and quarter-resolution ( pixel 1 ) were available for NOAA 8948, NOAA 9173, and NOAA The cadence varied from about a few minutes to an hour, depending on the observing plan. H images were only used to know the position of dark filaments. The second dual CCD system of the ASP was used simultaneously to obtain the full Stokes profiles of H line for active region NOAA 9231, and only Stokes I was used to generate H images. The universal birefringent filter ( UBF) with CCD camera,

3 992 KUBO & SHIMIZU Vol. 671 TABLE 1 Summary of Collision Events Location (deg) No. NOAA No. ASP Observation Date Latitude CMD TRACE (8) SXT H Magnetic Flux (10 20 Mx) L PIL (arcsec) Plus Minus Type Observed Step C Nov 19 S23 W PFI a ASP c I Nov 20 S23 W PFI ASP I Nov 21 S23 W PFI ASP f I 1, 2 C Sep 28 S11 E27... ( PFI) BBSO d I 2 C Apr 8 S14 E18... FFI b BBSO I Apr 9 S13 E04... FFI BBSO I 2 C Apr 8 S14 E21... FFI BBSO I Apr 9 S14 E08... FFI BBSO I 2 C Apr 9 S16 E04... FFI BBSO I 2 C Nov 19 S22 W PFI ASP I 2 C Nov 20 S22 W PFI ASP I 2, 3 C Nov 19 S23 W PFI ASP I 1, 2 C Nov 20 S22 W PFI ASP II 2 C Apr 15 S03 E UBF e III Apr 16 S03 E UBF III Apr 17 S03 W UBF III 2 C Apr 19 S03 W UBF III 2 C Nov 20 S20 W PFI ASP III 2, 3 Notes. L PIL represents length of the polarity inversion line. Magnetic flux is the value when colliding magnetic elements are isolated from the surrounding. When one polarity element collides with a part of the other ( bigger) polarity element, magnetic flux of the other polarity element is not estimated. See text for the type and observed step columns. a Partial frame image. b Full-frame image. c Second dual CCD system of Advanced Stokes Polarimeter. d Big Bear Solar Observatory. e Universal birefringent filter in ASP instruments. f The increase of negative flux is mainly cause by another negative elements denoted by B in Fig. 2. which was operated with the ASP, obtained H images at line center and in high time cadence for NOAA 9907 and NOAA The field of view of the UBF was ; with pixel size of about When the ASP and the UBF H data were not available, we used H images from Big Bear Solar Observatory (BBSO). 3. DATA ANALYSIS 3.1. Derivation of Magnetic Field Vector We investigated the orientation (inclination [] and azimuth [] angles) and strength (jbj) of the magnetic field, Doppler velocity (v los ), and continuum intensity (I c ) in 12 free parameters derived from the calibrated Stokes profiles by using a nonlinear least-squares fitting (Stokes inversion) code (Skumanich & Lites 1987; Lites & Skumanich 1990). The inversion code assumed a Milne-Eddington atmosphere and a two-component model atmosphere, in which the photospheric atmosphere was composed of magnetized and nonmagnetized atmospheres. The inclination and azimuth are shown in the local frame coordinates of the observing region. In this coordinate system, one views the region away from the normal direction on the solar surface. The zero velocity for the Doppler velocity was determined from averaging the center positions of the Fe i line over the entire field of view. The Doppler velocity represents the velocity relative to the mean rotational velocity at the position of the observation. Positive Doppler velocity corresponds to redshift. More detailed descriptions of r the analysis of the ASP data can be found in Kubo et al. (2003). The 180 ambiguity in the azimuth angle of photospheric magnetic field was resolved by using AZAM utility (Lites et al. 1995) with the following two steps: (1) The azimuth angle was selected to be close to the potential field in the photosphere for each pixel. The potential field was computed using the line-of-sight component of magnetic field as a boundary condition. (2) We interactively determined the azimuth to minimize discontinuities of the azimuth and inclination angles. We started the interactive resolution from the centers of sunspots where magnetic fields must diverge (converge) away from the centers for positive (negative) polarity. Magnetic fields in plage regions far from sunspots were determined to be more vertical to the surface. The resolution probably depended on the operators of the AZAM, but another person also obtained results similar to ours Image Co-Alignment Image co-alignments among magnetic field maps, H images, and coronal images are necessary to study their relationship. Image cross-correlation was used to match the light-of-sight magnetic flux map from the ASP data with the full-disk MDI longitudinal magnetogram taken at the time close to the time of the ASP observation. The position of ASP map was provided on the solar disk coordinate with the accuracy of 1 00 from the center and diameter of the full-disk MDI data. The limb position on full-disk images observed with the Extreme Ultraviolet Imaging Telescope ( EIT; Delaboudiniere et al. 1995) aboard SOHO provided the position of the TRACE and SXT field of view on the solar disk coordinate. When the TRACE images were aligned to the EIT full-disk images, The EIT images were used for the image-correlation rather than the EIT images, because the time cadence of EIT images (12 minutes) was much better than that of EIT images (6 hr). The accuracy of the image co-alignment between the TRACE and the EIT was 1 00 Y2 00. The SXT image was aligned to the EIT image with the accuracy of 3 00 Y4 00.

4 No. 1, 2007 MAGNETIC FIELD PROPERTIES OF FLUX CANCELLATION SITES 993 The H data simultaneously taken by another dual CCD system of the ASP were co-aligned to the photospheric magnetic field data with the accuracy of about 1 pixel ( ) by using the hair line markers in the both data. The position and shape of sunspots were used to have a rough alignment by visual inspection for the H images observed with the UBF and BBSO. The accuracy of the co-alignment for these images was about CLASSIFICATION OF COLLISION EVENTS We search when and where colliding opposite polarity elements appear by using the time series of MDI longitudinal magnetograms. The 12 events are then classified into three types as given in the column Type of Table 1. Type I events are those in which the opposite polarity elements to be collided appear at different times and at widely separated positions, while type II events are those in which the opposite polarity elements to be collided appear simultaneously in a neighborhood of collision. If it is not clear when and where both of the colliding opposite polarity elements appear, they are classified as type III. Magnetic flux cancellation proceeds with the following steps: (1) A magnetic polarity element approaches another polarity element. (2) They collide each other. (3) One or both of them disappear. ASP observations do not cover the entire duration of the collision process. The column Observed Step in Table 1 shows when ASP observations can be used in terms of three steps. 5. CASE STUDY We describe in detail one of the collision events (C1). Temporal changes in photospheric magnetic fields, formation of a dark filament, and coronal response are well observed for 3 days. Such temporal changes are clearly observed for event C1 because it has the largest size among the 12 events. Event C1 is a typical case for the collision of the opposite polarity elements without magnetic connection above the photospheric surface before the collision Magnetic Fields between the Colliding Opposite Polarity Elements Formation of Horizontal Magnetic Fields Event C1 is the collision of negative-polarity flux denoted by A and B with positive-polarity flux denoted by D and F (Fig. 2, left and right panels). The negative-polarity fluxes A and B exist from the beginning of the observations ( November 16). The positive D emerges with a negative-polarity element denoted by E on November 17. The positive F emerges with a negative polarity element indicated by the arrow in Figure 2 (l4) on November 19. The emerging negative elements merge into the leading sunspot, while the positive F merges with the positive D and forms a single positive region by November 20. Magnetic fields horizontal to the solar surface are newly formed in a narrow area along the polarity inversion line (PIL) between the negative-polarity flux A and the positive-polarity flux F (and D) after the negative A is in contact with the positive F (and D) on November 19 (see Figs. 3 and 4). The negative A and the center of the positive D have magnetic fields nearly vertical before the collision ( November 17), and magnetic fields are vertical for most of the flux A, D, and F islands during the collision (November 19). The horizontal fields in the narrow area along the PIL are directed from the positive F (and D) to the negative A (Fig. 5b). Magnetic fields in the negative B are nearly vertical until November 20. The degree of polarization is lower than the threshold (0.4%) in the region between the negative B and the positive D and F until November 20. When the negative B is located closer to the positive D and F on November 21, horizontal magnetic fields appear between them. The magnetic fields of the PIL are directed from the positive D and F to the negative B Temporal Change in Azimuth Angle of Horizontal Magnetic Fields A remarkable change of magnetic field azimuth on the PIL between the negative A and the positive D and F is observed during 3 days. Figure 5b shows that the magnetic fields are nearly perpendicular to the PIL for most of the PIL on November 19. The southern part of the PIL has magnetic fields parallel to the PIL. The magnetic fields are parallel to the PIL for most of the PIL on November 20, and return to be nearly perpendicular to the PIL on November 21. A dark filament is observed in H at the southern part of the PIL on November 19, and it is visible along the PIL on November 20, as shown in Figure 5c. The dark filament disappears at an unknown time between November 20 and 21. Thus, we note that the dark filament is seen when and where the photospheric field line is parallel to the PIL (Fig. 5d) Patchy High Doppler Velocity Areas along Polarity Inversion Line Spatial distributions of high Doppler velocity are compared with November 19Y21 continuum intensity maps (Fig. 6). Sunspot penumbrae have the Evershed flow, which is a horizontal outflow, around their outer boundaries (an example is shown by arrows in Fig. 6c). In addition to the sunspot penumbra, the area between colliding opposite polarity elements also has large Doppler motion. The patchy area with redshift is observed around the PIL during November 19Y21, and the blueshift is observed on November 21. The positive D and F have umbra and penumbra during 3 days of the observations, and the area with large blueshift is extended to the PIL from the positive D and F. Such large blueshift would be the Evershed flow. On the other hand, the patchy areas with large redshift are located at lower intensity areas beside the negative A without a pore and sunspot. Such lower intensity areas appear to be small penumbra without umbra Corona above the Colliding Magnetic Elements Active region NOAA 9231, which includes event C1, is observed with the SXT and the TRACE The SXT loops are brighter around their top than their footpoints, while the TRACE loops are bright around their footpoints. We here focus on steady SXT loops to investigate the change of magnetic connection between colliding magnetic elements. We define steady loops as the loops with lifetimes longer than several hours from the consideration of the radiative cooling time (about 10 4 s) and the conductive cooling time (about 10 3 s) for a loop with length l 1 ; cm, temperature T 3 ; 10 6 K, and density n 2 ; 10 9 cm 3. Multiple bundles of steady bright SXT loops (Fig. 2c1, black) connect the following positive-polarity region C with the leading negative-polarity regions A and B on November 16. When the regions D, E, F, and a part of the leading sunspot emerge, steady SXT loops appear to cover the emerging flux regions (Figs. 2c2 and 2c4, black arrows). Steady SXT loops connecting the positive D and F to the approaching negative B are formed on November 20 (Fig. 2c5, black arrow). A compact steady SXT loop is observed above the region between the negative A and the positive D and F (Fig. 2c5, gray arrow). Transient bright loops ( Fig. 2c4, gray arrow) appear between the negative-polarity regions (A and B) and the positive-polarity regions (D and F) from November 19, which is about 1 day before

5 Fig. 2. Day-by-day variation of collision event C1. The left panels are longitudinal magnetograms obtained with the MDI. White ( black) represents positive (negative) polarity. The center panels show X-ray images (reverse black and white) obtained by averaging several Yohkoh SXT images taken during one orbit (95 minutes). The right panels show schematic illustrations for variation of photospheric magnetic elements and coronal loops. The thick line represents existing SXT coronal loop, and the gray thick line represents SXT coronal loop formed on that day. The solid box on the MDI magnetograms defines the field of view of Fig. 3. The positions are given with respect to the center of the solar disk. See the text for arrows in the panels. [See the electronic edition of the Journal for a color version of this figure.] 994

6 Fig. 3. (a) Longitudinal magnetograms are obtained with the MDI. The location of the field of view is shown by the solid box in Fig. 2. The box indicates the field of view for Fig. 5. (b) Spatial distribution of the inclination of magnetic fields obtained with the ASP. The white areas (inclination of 0 and 180 ) have magnetic fields vertical to the solar surface and black areas (90 ) have magnetic fields with horizontal orientation. The degree of polarization is below 0.4% for the crosshatched area, so that no Stokes inversion is performed. There is no ASP observation on November 18 because of bad weather. 995

7 996 KUBO & SHIMIZU Fig. 4. Illustration of event C1 (a) before the collision of the opposite polarity elements and (b) during and (c) after the collision. The dash-dotted line represents the photospheric surface. The colliding elements in event C1 do not completely disappear during the ASP observations. Panel c is based on event C7. the formation of the steady loops. Their lifetime is from 30 minutes to 3 hr. The TRACE observations are better to identify transient loops in the narrow region along the PIL because of the higher spatial and temporal resolution than those of the SXT. The regions A, B, D, and F are bright in the TRACE images (Fig. 7), and they are identified as moss regions in Katsukawa & Tsuneta (2005). The corona above the PIL between the negative A and the positive D and F is darker (white in Fig. 8a) than the moss regions. In particular, it is dark around the southern part of the PIL on November 19 and 20. The dark area corresponds to the location of the dark filament (see Fig. 5c). Transient bright loops are frequently observed with the TRACE to connect the negative A and the positive F around the northern PIL (Fig. 8a). The intensity around the northern PIL increases owing to the appearance of such transient loops (Fig. 8b). Their lifetime is less than about 30 minutes. Transient bright loops ( Fig. 7, arrows) are also observed to connect the negative B to the positive D and F several times during the TRACE observations (15:25 23:38 UT) on November 19. Such bright loops are more frequently observed on November 20 and 21, when the position of negative B becomes closer to the positive D and F. The colliding opposite polarity elements in event C1 do not have magnetic connection above the photospheric surface before the collision. Thus, the appearance of coronal bright loops connecting the colliding elements indicates that new magnetic connections between the colliding elements are established in the corona above the PIL. 6. POLARITY INVERSION LINES FOR 12 COLLISION EVENTS We examine whether the results obtained for event C1 are common to the 12 collision events studied in this work Nature of Horizontal Magnetic Fields A polarity inversion line (PIL) is identified at 11 collision events. The exception is event C12, in which the degree of polarization is less than the threshold (0.4%) for the inversion between the colliding magnetic elements. Figure 9 shows magnetic field properties for all the PILs defined by the 11 collision events. Magnetic fields on the PILs are horizontal to the solar surface, regardless of the classification defined in x 4. Most of the magnetic fields have azimuth angle of between 0 and 90 with respect to azimuth angle of potential field in the photosphere. Some of the magnetic fields nearly parallel to the PIL have azimuth angle larger than 90 with respect to the potential field. The percentage of the area with the magnetic fields perpendicular to the PIL (shear angle <45 )is nearly same as that of the magnetic fields parallel to the PIL (shear angle >45 ). Magnetic field strength jbj has a peak around 500 G, with a gradual wing beyond 1000 G. The magnetic field strength for most of the PILs is weaker than the typical field strength with vertical orientation in the whole active region (1400 G). Figure 10a shows that the Doppler velocity of the PILs has similar amplitude for both red and blue directions, and maxima of the Doppler velocity become larger with increasing distance from the solar disk center. This tendency suggests that high Doppler velocities mainly consist of flows along the horizontal magnetic fields. Note that magnetic fields are horizontal to the solar surface on the PILs and these magnetic fields are almost parallel to the line of sight as they are located near the limb. Converging flows toward the PIL and shear flows along the PIL are observed in the collision site. Such flows also contribute to the Doppler shift as the observing region becomes close to the limb. However, the motion of colliding magnetic elements is estimated to be less than 0.5 km s 1 using MDI magnetogram, and it is slower than the observed high Doppler velocity (2kms 1 ) in the collision site. Small magnetic elements not resolved by MDI may move faster (e.g., Berger et al. 1998), but the size of the area with high Doppler velocity is larger than 1 00, as shown Figure 6. Therefore, we conclude that the origin of high Doppler velocity in the collision site is mainly the flow along the horizontal magnetic fields. The speed of the flow along the horizontal magnetic fields is estimated to be up to about 4kms 1 from the Doppler velocity and the angle between the magnetic field line and the line-of-sight direction. The areas with high Doppler velocity (>1 km s 1 ) have a continuum intensity of between 0.8 and 1.0, as shown in Figure 10b. The darkened intensity has a similarity to sunspot penumbra, and the Evershed flow is observed around the outer boundary of the sunspot penumbra. We point out that the inclination and the field strength of magnetic fields on the PILs are similar to those at the outer boundary of penumbra. When horizontal magnetic fields are newly formed between the colliding magnetic elements, the local magnetic structure may be similar to the outer boundary of sunspot penumbra, resulting in the appearance of gas flows along the magnetic field lines. Colliding magnetic elements are not always equal in size or in magnetic flux. Thus, when new connections are formed between the colliding magnetic elements, magnetic pressures would not balance. The horizontal gas flow along the field line is driven from the pressure imbalance Dark Filaments and Photospheric Magnetic Fields We investigate the relationship between the direction of photospheric magnetic fields and the formation or disappearance of dark filaments seen in H. We examine how much magnetic fields with shear angle larger than 45 is involved in the PIL (see Table 2). The shear angle represents whether magnetic field is more perpendicular or parallel to the PIL. Magnetic fields parallel to the PIL are dominant for all five events (C1, C3, C4, C5, and C10) with dark filaments. Change of

8 Fig. 5. (a) Polarity inversion line delineated on magnetic flux map, (b) transverse magnetic field, and (c) H image on November 19Y21. These images are made from ASP observations. H image on November 21 is observed at the Big Bear Solar Observatory. The field of view is identical to the box in Fig. 3. The white thick lines in panel a and the gray thick lines in panel b represent the polarity inversion line. Panel d shows illustrations for relationship between azimuth angle of photospheric magnetic field and a dark filament.

9 998 KUBO & SHIMIZU Fig. 6. (a) Doppler velocity map, (b) area with high Doppler velocity (>0.8 km s 1 ), and (c) continuum intensity map obtained with the ASP. The field of view is identical to Fig. 5. Positive Doppler velocity (white) corresponds to redshift in panels a and b. The arrows in panels a and c represent penumbra of the small sunspot observed on November 20. The contours in panel b represent vertical component of magnetic field of 600 G derived from the ASP. [See the electronic edition of the Journal for acolorversion of this figure.] the magnetic fields from perpendicular to parallel is only observed for the four events with dark filament (C1, C4, C5, and C10). The timescale of such change is about a day. The magnetic fields are already parallel to the PIL at the start of the ASP observations for the C3 event. Disappearance of a dark filament is only observed for event C1, and the magnetic fields are perpendicular to the PIL after the disappearance of the dark filament, as shown in x 5.1. Dark filaments are not observed for six events (C2, C6, C7, C8, C9, and C11). Magnetic fields perpendicular to the PIL are the dominant component in all these events except events C6 and C7. The length of the PIL is only 6 00 Y8 00 for events C6 and C7, and they are small-scale collisions. The relationship between the size of collisions and the formation of dark filaments still remains an open issue. Our observations clearly show that magnetic fields below dark filaments are parallel to the PIL when dark filaments are located along the PIL. Magnetic fields are essentially perpendicular to the PIL when dark filaments are not observed Coronal Structures Transient bright loops connecting colliding opposite polarity elements are observed with the TRACE and the SXT for event C1. We search such transient brightening loops for six events observed with both the TRACE and the SXT (see

10 Fig. 7. (a) Longitudinal magnetograms are obtained with the MDI. (b) Day-by-day change of the corona observed with TRACE at Coronal bright features are displayed in black. (c) Coronal loops are enhanced by the Laplacian filter ( Kano & Tsuneta 1995).

11 Fig. 8. (a) Short-term variation of the corona observed with the TRACE on November 19. Coronal bright features are displayed in black. The field of view is identical to that in Fig. 5. The white and gray lines in the first image represent the polarity inversion line obtained with the MDI longitudinal magnetogram taken at the nearest time of the TRACE data. (b) Time profiles of line intensity observed with the TRACE. The diamond shows the intensities at the northern side of the polarity inversion line and the cross shows those at the southern side of the line. The intensities are averaged along the polarity inversion line for each part. The plus shows the intensity averaged over the whole pixels in TRACE image for reference purposes. [See the electronic edition of the Journal for a color version of this figure.]

12 MAGNETIC FIELD PROPERTIES OF FLUX CANCELLATION SITES 1001 Fig. 9. Histograms of various quantities for all the polarity inversion lines defined by 11 flux colliding events observed with the ASP: (a) inclination of magnetic fields, (b) azimuth of magnetic fields with respect to the computed potential field (shear angle), and (c) magnetic field strength. The inclination of 0 and 180 represents that magnetic field is vertical to the solar surface and 90 corresponds to horizontal orientation. The inclination lower than 90 corresponds to the positive polarity. The potential field is calculated by using the MDI longitudinal magnetogram as a boundary condition ( Teuber et al. 1977). Light gray and black distributions show the properties for all the polarity inversion lines and for the whole active region NOAA 9231, respectively. The dark gray area represents overlaps between them. Table 3). All the six events are located in the active region NOAA Transient bright loops are observed with the SXTonly for event C1. The size of event C1 is the largest among the 12 collision events. The other events are overlaid with the steady SXT loops, as described in x 5.2. If there are small brightenings in these events, it would be difficult to identify such small brightenings. Moreover, the temporal cadence of the SXT is not high during the ASP observations. Therefore, we cannot conclude that there are no SXT brightenings during the collisions for these cases. Transient bright loops connecting colliding opposite polarity elements are observed with the TRACE for events C1, C6, and C8. These are the type I events, for which the opposite polarity elements appear at different times and positions. On the other hand, TRACE transient bright loops are not observed for other three events (C7, C9, and C12). The TRACE observations start just before magnetic flux completely disappears for the C7 and C12 events and their magnetic flux is much smaller than any other events (see Table 1). There are both collisions with and without formation of coronal loops. Such difference would depend on the size of the colliding elements or magnetic connection between the colliding elements before the collision. Nevertheless, our observations show that formation of coronal loops is essentially associated with the collisions. 7. DISCUSSION We analyzed 12 collision events of opposite polarity elements using simultaneous observations of photospheric vector magnetic field, X-ray/EUV coronal images, and H images. Our findings are summarized as follows: 1. Magnetic fields horizontal to the solar surface are newly formed on the polarity inversion line (PIL) when one polarity elements approaches and collides with the other polarity elements for 11 events. The opposite polarity elements have vertical magnetic fields before the collision. 2. The horizontal magnetic fields between the colliding elements are initially perpendicular to the PIL. They become parallel to the PIL when a dark filament is observed along the PIL. The magnetic fields return to be perpendicular with the PIL at around the disappearance of the dark filament for the C1 event. The photospheric magnetic fields below dark filaments are parallel to the PIL for all five events with dark filaments. 3. Large Doppler motions with positive and negative direction are observed around PIL. Our observations suggest that these Fig. 10. Doppler velocity as a function of (a) distance to the solar disk center and (b) continuum intensity. All the polarity inversion lines defined by flux colliding events are plotted in panel a. Only the polarity inversion lines located more than away from the disk center are plotted in panel b, because both large and small Doppler velocities are observed there.

13 1002 KUBO & SHIMIZU Vol. 671 TABLE 2 Azimuth Angle of Magnetic Fields and Dark Filaments No. Date Time L PIL a (arcsec) Shear Angle >45 b (%) Dark Filament C Nov 19 17: Observed 2000 Nov 20 15: Observed 2000 Nov 21 17: Not observed C Sep 28 19: Not observed 23: Not observed C Apr 8 15: Observed 2000 Apr 9 15: Observed C Apr 8 15: Observed 2000 Apr 9 15: Observed C Apr 9 14: Observed 19: Observed C Nov 19 15: Not observed 17: Not observed C Nov 20 15: Not observed 19: Not observed C Nov 19 16: Not observed C Nov 20 15: Not observed 20: Not observed C Apr 15 16: Observed 2002 Apr 16 23: Observed 2002 Apr 17 14: Observed C Apr 19 15: Not observed 20: Not observed a Length of the polarity inversion line ( PIL). b Percentage of magnetic fields with shear angle larger than 45 on the PIL, where shear angle is azimuth angle of magnetic field with respect to the computed potential field. motions are due to flows with a speed up to 4 km s 1 along horizontal field lines. 4. Bright loops with lifetime less than about 30 minutes connecting the colliding opposite polarity elements are newly observed with the TRACE for three of the six events. Transient brightening loops are also observed with the SXT only in one event Formation of Horizontal Magnetic Fields in Photosphere We often observe that opposite polarity elements to be collided appear at different times and positions. In this case, approaching one polarity element have no connection with another polarity element when they appear, as shown in Figure 1. Azimuth angle of the horizontal magnetic field is important for investigating configuration of photospheric magnetic fields. Submerging -shaped field lines (Fig. 1a) have the direction opposite to emerging U-shaped field lines ( Fig. 1b) around the PIL at the photospheric level. Our observation shows that magnetic fields of the PILs essentially have the shear angle less than 90 (see Fig. 9b). This shows that the horizontal magnetic fields of the PIL correspond to the -shaped loop. We probably rule out the possibility that magnetic fields have opposite direction due to the ambiguity of 180 for the azimuth angle, because we carefully deal with the azimuth ambiguity, as mentioned in x 3.1. We observe the large positive and negative Doppler velocities of horizontal magnetic fields between colliding magnetic elements. Both Doppler components in the cancellation sites have also been observed in previous works (Chae et al. 2004; Yurchyshyn & Wang 2001). We have found a clear tendency for the the Doppler velocity to increase as the horizontal magnetic fields become close to the limb (the line-of-site direction). This suggests that the observed velocities on the PILs are mainly due to the gas flows along No TABLE 3 Summary of TRACE and SXT Observations of Transient Bright Loops Date L PIL (arcsec) Type TRACE SXT C Nov 19Y Y90.4 I Observed Observed C Nov I Observed Not observed C Nov I Not observed Not observed C Nov I Observed Not observed C Nov II Not observed Not observed C Nov III Not observed Not observed the field lines rather than actual emerging or submerging motions of the magnetic field lines Coronal Loops Connecting Colliding Magnetic Elements We propose that transient coronal loops observed above the PILs are formed as a result of magnetic reconnection (Fig. 1a), because such coronal loops, which are not apparent before the collision, appear and connect the colliding opposite polarity elements. Our observations show that transient coronal loops frequently appear during the collision and decrease of photospheric magnetic flux. Thus, magnetic reconnection takes place in the corona during magnetic flux cancellation. If the reconnected magnetic field lines submerge to below the photosphere ( Fig. 1a, right panel), the formation of the photospheric horizontal fields between colliding magnetic elements and their disappearance will be explained. Harvey et al. (1999) observed that a coronal bright point, chromospheric magnetic flux, and photospheric magnetic flux disappear in sequence at the cancellation sites. Submergence of magnetic field lines from the corona to below the photosphere may not take place easily because magnetic buoyancy prevents the submergence of magnetic field lines owing to a significant increase in density as they move from the corona to the photosphere. Thus, magnetic reconnection may take place at multiple locations with different heights during the collision, as shown in Figure 1c.Notethatmagneticreconnectionshouldmoreeasily take place around the temperature minimum region, which is about 500 km above the photospheric surface, because a resistivity possesses a maximum at the temperature minimum region (Takeuchi & Shibata 2001a, 2001b) Formation of Dark Filaments Our results of photospheric magnetic fields below dark filaments are consistent with the previous observations: (1) Sheared magnetic fields are located below dark filaments (Hagyard et al. 1984, 1990; Wang & Li 1998; Tian et al. 2002; Lites 2005). (2) Shear angle of magnetic fields below dark filaments increases in time (Hagyard et al. 1984; Hagyard et al. 1990; Tian et al. 2002) (3) The shear angle decreases with an eruption of dark filament (Hagyard et al. 1984, 1990; Tian et al. 2002). Our observations show that such global reconfiguration from perpendicular to parallel (and then to perpendicular) with respect to the PIL takes place in1day. Measurements of magnetic fields in dark filaments using the Hanle effect show that a dominant component of the magnetic fields is along the axis of the dark filament (e.g., Leroy 1978, 1989, 1983, al. 1984; Bommier & Leroy 1998). There are the following two possibilities for the formation of the magnetic fields along the axis of dark filaments. One possibility is that magnetic field in a dark filament is generated above the photospheric surface (Pneuman 1983;

14 No. 1, 2007 MAGNETIC FIELD PROPERTIES OF FLUX CANCELLATION SITES 1003 van Ballegooijen & Martens 1989, see Fig. 1; Martens & Zwaan 2001). The footpoints of the magnetic field lines approach each other as a result of photospheric converging flows toward the PIL, and then their magnetic shear is enhanced. Multiple magnetic reconnection of the sheared field lines generates field lines along the axis of the dark filament. In this case, the photospheric magnetic field below the dark filament is not needed to be parallel to the PIL. Our observations show that magnetic fields are parallel to the PIL for all five events with dark filaments. Thus, our observations appear to be inconsistent with the model. Another possibility is that the magnetic fields parallel to the PIL are generated below the photospheric surface, and they become magnetic fields in a dark filament as a result of their emergence into the corona (van Ballegooijen & Martens 1990; Rust & Kumar 1994; Low & Hundhausen 1995; Lites et al. 1995). The photospheric magnetic field below the dark filament is always parallel to the PIL during the emergence. This is consistent with our observations. Lites (2005) obtained concave (U-shaped) photospheric magnetic fields around the PIL below two dark filaments (event C4 is one of them). However, we observe both concave (U-shaped) and convex (-shaped) geometries for magnetic fields nearly parallel to the PIL below dark filaments. This would rule out the simple emergence of concave magnetic fields as shown in Figure 1b. Considering the azimuth ambiguity of photospheric magnetic fields, it would be difficult to discuss the formation of dark filaments using only the azimuth angle. If the magnetic field lines parallel to the PIL emerge into the photosphere, they must appear prior to the formation of dark filaments. We cannot examine such temporal relationship because of the insufficient time resolution. Simultaneous and continuous observations of the horizontal magnetic fields with velocity information and the dark filaments are essential to understanding the relation between the magnetic flux cancellations and the formation of dark filaments. 8. CONCLUSIONS We investigated 12 collision events in which one magnetic polarity element approached (and canceled with) an opposite magnetic polarity element. The opposite polarity elements had vertical magnetic fields before the collision, and then horizontal fields were spontaneously formed between the colliding opposite polarity elements. Furthermore, the new magnetic connections are also found in the corona as transient coronal loops connecting the colliding opposite polarity elements. We suggest that the formation of such coronal loops and photospheric horizontal fields between the colliding elements are due to magnetic reconnection between colliding field lines, possibly at different (multiple) locations with different heights. We obtained a clear relationship between the direction of the photospheric magnetic fields at the collision site and the formation or disappearance of dark filaments. The photospheric magnetic fields below dark filaments were parallel to the polarity inversion line for all five events with dark filaments, while the magnetic fields were essentially perpendicular to the inversion line when dark filaments were not seen. Therefore, we suggest that the global change in the direction of the photospheric horizontal field lines between the colliding elements does occur in association with formation and disappearance of the dark filaments. Our observations showed that the cancellation site was one of the areas with patchy high Doppler velocity (>1 km s 1 ). The center-to-limb variation of the Doppler velocity clearly showed that the high Doppler velocity with both blue and red components was not due to the actual motion of the horizontal magnetic fields but rather to the gas flow along the field lines. We would like to express our sincere gratitude to S. Tsuneta for various helpful comments and suggestions on this paper. B. W. Lites, Y. Katsukawa, and K. Ichimoto are acknowledged for useful discussions. We are grateful to the US National Solar Observatory and the HAO for the setup and operation of the ASP observations. T. D. Tarbell is acknowledged for supporting the observations of TRACE and SOHO. We would like to thank the Big Bear Solar Observatory/New Jersey Institute of Technology for H data. This work was supported by the Professor S. Hayakawa Fund of the Astronomical Society of Japan for young scientists and the JSPS Japan-US collaborative science program Collaborative Research of Solar Explosive Phenomena with Yohkoh/SOHO/TRACE/Solar-B. Berger, T. E., Loefdahl, M. G., Shine, R. S., & Title, A. M. 1998, ApJ, 495, 973 Bommier, V., & Leroy, J. L. 1998, in IAU Colloq. 167, New Perspectives on Solar Prominences, ed. D. F. Webb et al. (ASP Conf. Ser. 150; San Francisco: ASP), 434 Chae, J., Martin, S. F., Yun, H. S., Kim, J., Lee, S., Goode, P. R., Spirock, T., & Wang, H. 2001, ApJ, 548, 497 Chae, J., Moon, Y., & Pevtsov, A. A. 2004, ApJ, 602, L65 Chae, J., Qiu, J., Wang, H., & Goode, P. R. 1999, ApJ, 513, L75 Delaboudiniere, J.-P., et al. 1995, Sol. Phys., 162, 291 Elmore, D. F., et al. 1992, Proc. SPIE, 1746, 22 Hagenaar, H. J. 2001, ApJ, 555, 448 Hagyard, M. J., Teuber, D., West, E. A., & Smith, J. B. 1984, Sol. Phys., 91, 115 Hagyard, M. J., Venkatakrishnan, P., & Smith, J. B. 1990, ApJS, 73, 159 Handy, B. N., et al. 1999, Sol. Phys., 187, 229 Harvey, K. L. 1994, in IAU Colloq. 143, The Sun as a Variable Star: Solar and Stellar Irradiance Variations, ed. J. M. Pap et al. (Cambridge: Cambridge Univ. Press), , in ASP Conf. Ser. 111, Magnetic Reconnection in the Solar Atmosphere, ed. R. D. Bentley & J. T. Mariska (San Francisco: ASP), 9 Harvey, K. L., Harvey, J. W., & Martin, S. F. 1975, Sol. Phys., 40, 87 Harvey,K.L.,Jones,H.P.,Schrijver,C.J.,&Penn,M.J.1999,Sol.Phys.,190,35 Kano, R., & Tsuneta, S. 1995, ApJ, 454, 934 Katsukawa, Y., & Tsuneta, S. 2005, ApJ, 621, 498 Krivova, N. A., & Solanki, S. K. 2004, A&A, 417, 1125 Kubo, M., Shimizu, T., & Lites, B. W. 2003, ApJ, 595, 465 Leroy, J. L. 1978, A&A, 64, , in Dynamics and Structre of Quiescent Solar Prominences, ed. E. R. Priest ( Dordrecht: Kluwer), 77 REFERENCES Leroy, J. L., Bommier, V., & Sahal-Brechot, S. 1983, Sol. Phys., 83, , A&A, 131, 33 Lites, B. W. 2005, ApJ, 622, 1275 Lites, B. W., Low, B. C., Martinez Pillet, V., Seagraves, P., Skumanich, A., Frank, Z. A., Shine, R. A., & Tsuneta, S. 1995, ApJ, 446, 877 Lites, B. W., & Skumanich, A. 1990, ApJ, 348, 747 Livi, S. H. B., Wang, J., & Martin, S. F. 1985, Australian J. Phys., 38, 855 Low, B. C., & Hundhausen, J. R. 1995, ApJ, 443, 818 Martens, P. C., & Zwaan, C. 2001, ApJ, 558, 872 Martin, S. F. 1998, Sol. Phys., 182, 107 Martin, S. F., Livi, S. H. B., & Wang, J. 1985, Australian J. Phys., 38, 929 Ogawara, Y., Takano, T., Kato, T., Kosugi, T., Tsuneta, S., Watanabe, T., Kondo, I., & Uchida, Y. 1991, Sol. Phys., 136, 1 Parker, E. N. 1984, ApJ, 280, 423 Pneuman, G. W. 1983, Sol. Phys., 88, 219 Rust, D. M., & Kumar, A. 1994, Sol. Phys., 155, 69 Scherrer, P. H., et al. 1995, Sol. Phys., 162, 129 Schrijver, C. J., et al. 1998, Nature, 394, 152 Shimizu, T., Shine, R. A., Title, A. M., Tarbell, T. D., & Frank, Z. 2002, ApJ, 574, 1074 Skumanich, A., & Lites, B. W. 1987, ApJ, 322, 473 Somov, B. V., Kosugi, T., Hudson, H. S., Sakao, T., & Masuda, S. 2002, ApJ, 579, 863 Takeuchi, A., & Shibata, K. 2001a, ApJ, 546, L b, Earth Planets Space, 53, 605 Teuber, D., Tandberg-Hanssen, E., & Hagyard, M. J. 1977, Sol. Phys., 53, 97 Tian, L., Wang, J., & Wu, D. 2002, Sol. Phys., 209, 375 Tsuneta, S., et al. 1991, Sol. Phys., 136, 37