SADS for SADs: Semi-Automatic Detection Software for Supra-Arcade Downflows Sabrina Savage 1, David E. McKenzie 1, Letisha McLaughlin 2 1 Montana State University, 2 University of North Carolina - Wilmington 1.Abstract Detectable signatures of magnetic reconnection, a possible source of solar coronal heating, aid in constraining flare energetics. Supra-arcade downflows (SADs), first detected during the Yohkoh mission, are an example of an observable consequence of magnetic flux tube reconnection. These sunward-traveling voids above arcade loops are consistent with outflows resulting from standard 3D reconnection models of solar flares. We have developed semi-automated detection software to detect downflows and analyze their trajectories, speeds, sizes, and magnetic flux in order to constrain parameters for flare modeling. We will present these measurements as observed primarily by SXT and TRACE and discuss their implications. We will also introduce detections from the newest solar X-ray telescope aboard Hinode. 5. Acknowledgements This work is supported by NASA Grant NNG04GB76G. 6.References “Downflow Motions Associated with Impulsive Nonthermal Emissions Observed in the 2002 July 23 Solar Flare”, by Asai, Ayumi; Yokoyama, Takaaki; Shimojo, Masumi; Shibata, Kazunari 2004, ApJ, 605, L77. “SUMER Spectral Observations of Postflare Supra-Arcade Inflows”, by D.E. Innes, D.E. McKenzie, & T. Wang 2003, Solar Physics, 217, 247. “A Model for Patchy Reconnection in Three Dimensions”, by M. Linton and D. Longcope 2006, ApJ, 642, “Observations of Separator Reconnection to an Emerging Active Region”, by D. Longcope, D. McKenzie, J. Cirtain, and J. Scott 2005, ApJ, 630, 596. “X-Ray Observations of Motions and Structure Above a Solar Flare Arcade”, by D. E. McKenzie and H. S. Hudson 1999, ApJ, 519, L93. “Supra-arcade Downflows in Long-Duration Solar Flare Events”, by D. E. McKenzie 2000, Solar Physics, 195, 381. “Signatures of Reconnection in Eruptive Flares”, Invited Review, by McKenzie, D.E., in Multi-Wavelength Observations of Coronal Structure and Dynamics, P.C.H. Martens and D.P. Cauffman, eds., COSPAR Colloquia Series, Elsevier Science Ltd. pub. (2002), 13, 155. “Characteristics of Coronal Inflows”, by Sheeley, N. R., Jr.; Wang, Y.-M. 2002, ApJ, 579, 874. “The Origin of Postflare Loops”, by Sheeley, N. R., Jr.; Warren, H. P.; Wang, Y.-M. 2004, ApJ, 616, “Initial features of an X-class flare observed with SUMER and TRACE”, by Wang, T. J.; Solanki, S. K.; Innes, D. E.; Curdt, W. 2002, in SOLMAG Proceedings of the Magnetic Coupling of the Solar Atmosphere Euroconference and IAU Colloquium 188, June 2002, Santorini, Greece. Ed. H. Sawaya-Lacoste. ESA SP-505. Noordwijk, Netherlands: ESA Publications Division, ISBN , 2002, p Introduction Supra-arcade downflows (SADs), as the name implies, are downward-moving features observed in the hot, low-density region above post-eruption flare arcades. Initially detected with Yohkoh/SXT during the 20 January 1999 flaring event (McKenzie & Hudson, 1999), these X-ray-dark, blob- shaped features have since been observed with TRACE (e.g., Innes et al., 2003; Asai et al., 2004), SOHO/SUMER (Innes et al., 2003), and SOHO/LASCO (Sheeley & Wang, 2002). The darkness in X-ray and EUV images is due to very low plasma densities, i.e., plasma voids (Innes et al., 2003). [As an aside, we note that many X-ray-emitting SADs are also known. McKenzie (2000) reported faint X-ray-emitting shrinking features in several flares; and in our catalog of 40 SAD flares observed by SXT and TRACE, approximately half display such bright shrinking features alongside dark SADs.] 3 Automated Detection and Analysis At Montana State University, we are developing semi-automated software for detection and measurement of SADs. Software development and testing are described in a paper submitted to the Astrophysical Journal. Here, we provide the latest results from the analysis of the famous 21 April 2002 flare as observed by TRACE (Figure 2), four west limb Yohkoh/SXT flares (including Figure 1), and one Hinode/XRT flare seen on the disk. Overview: As is suggested by the left panel of Figure 1, an important first step in performing our analysis is to sharpen the image sequences. Our techniques for extracting the SADs from the noisy, faint region above post-flare arcades include flattening, run-mean-differencing, and smoothing. The software then searches each image for voids using variable thresholds and then matches void paths between frames while allowing for small accelerations. Outputs from analysis on the subsequently accepted trajectories include initial and average velocities (head and centroid), accelerations (from a polynomial fit), initial and average areas, initial magnetic fields and fluxes for west limb flares (based on the potential-field source-surface (PFSS) approximation), initial height, and total displacement. TRACE: The 21 April 2002 flare was observed by TRACE (Figure 2), RHESSI, SOHO, and numerous other observatories (e.g., Wang et al., 2002; Innes et al., 2003). The TRACE 195Å images show SADs in high resolution and are therefore relatively easy to track with our automated software. Because the flare was observed on the west limb of the sun, there is the added bonus that we can implement the PFSS software to obtain magnetic flux estimates, we can estimate the height above the arcade, and the plane-of-sky velocities reported are also closest to true velocities. 4.Discussion Supra-arcade downflows are important signatures of reconnection in flares (McKenzie, 2002; Asai et al. 2004). As tracers of reconnection outflow, their characteristics are indicative of the parameters of 3D patchy reconnection, including the size of participating flux tubes, and, by extension, the characteristic size of the localized diffusion region. The application of automated software to real flares, as shown here, demonstrates that it is possible to derive quantitative data from images of these velocity fields. The TRACE histograms (Figure 3) demonstrate a range of sizes, with a smooth dropoff towards larger voids. This is directly relevant to models of 3D reconnection, by revealing the distribution of reconnection “patches”. It is worth noting that the areas observed in the 21-Apr flare are similar to the cross-sections of reconnecting loops observed in TRACE by Longcope et al (2005). Using the rough-estimate magnetic fields from the PFSS approximation, the flux in each shrinking loop is on the order of 2 x maxwells, (Table 1) which is on the same order as the per-loop flux estimated by Longcope et al. (4 x Mx). It should be noted, however, that most of the quantities derived from our routine are conservative underestimates. Similarly, the observed speeds indicate outflow that is slower than the nominal Alfvén speed: this is consistent with previous reports of downflow speeds, and with the simulated outflows of Linton & Longcope (2006), although no attempt has been made to estimate the drag forces necessary to produce these speeds. As a further example of the utility of quantitative measurement, consider that as a flux tube undergoes shrinkage by an amount ∆L, the energy lost is given by ∆W=B 2 A ∆L / 8 , where A is the cross-sectional area of the flux tube. Using the mean values shown in Table 1, the conservative estimate derived from our automated software yields shrinking energy on order of ergs per event. To date, SADs have been observed with SXT more often than in the TRACE data. This is no doubt due to the full-Sun field of view of SXT as well as its sensitivity to hotter plasmas--the supra-arcade region is very hot, and the dark SADs are easier to see against a bright background. However, most of the SADs were observed with SXT’s half-resolution (5 arcseconds per pixel), so that the smaller features were not detected. This is borne out by the histograms above--TRACE observes plasma voids much smaller than those seen by SXT. While TRACE offers much higher angular resolution, the cadence of TRACE images is slow enough to allow some flows--particularly the faster ones--to go undetected. In some cases, the cadence of SXT images was even too slow, so that motions faster than 700 km/s may have been unobservable (McKenzie, 2000). Moreover, TRACE’s smaller field of view means that some flares are not observed: in a study of 12 SAD flares observed by SXT, McKenzie (2000) found TRACE data for only two events. With increased solar activity, XRT and AIA will allow for more detailed SAD observations due to wider fields of view, high cadence capability, and high resolution: the TRACE-like angular resolution of XRT and AIA ensures that a wide spectrum of SAD sizes will be observable. Moreover, AIA’s high- temperature sensitivity--greater than TRACE’s--is expected to reveal the fan-like structure above eruptive flare arcades much more often than TRACE, so that SADs will be observed more often, and at greater heights. And AIA’s fast cadence of 1 image in each passband per 10 seconds is significantly faster than typical TRACE sequences, and faster even than most flare sequences in SXT, where 1 image per passband per resolution often required as much as 20 seconds. This higher rate of sampling may result in faster SADs being detected by AIA, contributing still further to the observational database. For instance, if AIA observes the same ∆L as found in the flares presented here, extending over four contiguous images in sequence, then SADs moving as fast as 1700 km/s should be detectable, if they exist. Figure 6. Cartoon depiction of supra-arcade downflows resulting from patchy reconnection. Discrete flux tubes are created, which then individually shrink, dipolarizing to form the post-eruption arcade. (From McKenzie, 2002) This figure has been updated to include quantifiable information as a result of using the semi- automated detection software. The downflows are traced by discrete X-ray features with a characteristic size. The present interpretation states that the downflows represent the outflow of magnetic flux from a reconnection site, in keeping with the standard reconnection model of eruptive flares (see Figure 6, from McKenzie, 2002; see also Sheeley et al., 2004). If they are reconnection outflows, then these tracers strongly suggest that the reconnection takes place between discrete collections of magnetic flux, i.e., flux tubes. This conclusion indicates “patchy” 3D reconnection. While the observed speeds of tens-to-hundreds of km/s were slower than initially expected (i.e., slower than the 1000 km/s which is often assumed to be the Alfvén speed), the recent model of 3D patchy reconnection by Linton & Longcope (2006) indicates the presence of drag forces working against the outflow. In their model, reconnection was allowed to happen in a localized region of slightly enhanced resistivity, and the evolution of the reconnected magnetic field was studied. As expected, the reconnected field retracted away from the reconnection site, accelerated by slow-mode shocks: “This accelerated field formed a pair of three-dimensional, arched flux tubes whose cross sections had a distinct teardrop shape. The velocities of the flux tubes was smaller than the reconnection Alfvén speed predicted by the theory, indicating that some drag force is slowing them down.” These drag forces, which seem to appear only in a genuine 3D simulation, result in outflow speeds at only a fraction of the Alfvén speed. Linton & Longcope considered field line tangling--a truly 3D effect--or added mass, perhaps due to snowplowing, as possible sources of drag. Quantitative measurements of downflows yield useful constraints for such models. For example, measurements of the characteristic sizes of SADs can be directly applied to the model as a means of limiting the duration of each magnetic reconnection episode, or the size of a resistive patch. Combined with estimates of the magnetic field in the supra-arcade region, measurements of the sizes of reconnected flux tubes yield estimates of the magnetic flux in individual flux tubes, and therefore the characteristic amount of flux that participates in a magnetic reconnection episode. Furthermore, combining the magnetic flux with the displacement yields an estimate of the energy released by the shrinkage. In the model of Linton & Longcope (2006), this shrinkage energy can account for as much as half the total energy converted by an individual reconnection episode. Figure 1. An example supra-arcade downflow (SAD) from the 12 July 2000 flare as seen by Yohkoh/SXT. The right panel displays the sharpened version of the prepped data (left panel). A total of 9 SADs were tracked and analyzed from this event. SXT: A total of 39 SADs were detected between the four flares shown in Figure 4. Due to the relatively low resolution and high noise in the SXT images, 19 of the tracks were only partially measured and are not included in the table below. The low resolution also means that small-area SADs, which account for the bulk of the TRACE detections, cannot be detected in the SXT images. 30-Sept Jul Dec Jul-2000 Figure 4. Four SXT, west-limb flares with relatively distinctive downflows have been analyzed: 30 September 1998; 25 July 1999; 07 December 1999; 12 July The above figure shows the resulting centroid tracks of the detected and verified downflows. XRT: The X-ray telescope aboard the new solar observatory Hinode provides the opportunity to observe SADs in environments similar to those observed by SXT but with resolution closer to that of TRACE (~1” versus ~5”). The current quiet state of the sun coupled with the recent launch of Hinode has not afforded many opportunities to ideally observe solar flares on the limb; however, downflows have been observed during flaring events on the disk. While magnetic flux estimates have yet to be applied using XRT, we do present detections and velocity/area measurements for the 13 December 2006 flare (Figure 5). Figure 5. Introducing SADs as seen by the new X-ray telescope (XRT) aboard Hinode in the 13 December 2006 flare. Because this flare was observed on the solar disk instead of the limb, one would expect slower plane-of-sky speeds due to the projected angle for the magnetic flux tube cross-section model. The measured speeds are, in fact, 2-5 times slower than the average speeds measured from the aforementioned TRACE and SXT flares. Instrument Initial Area (10 17 cm 2 ) Average Head Speed (km/s) Initial B Flux (10 18 Mx) Displacement (10 9 cm) Energy per SAD (ergs)TRACE 1.2 +/ / / / e27 SXT 4 +/ / / / e27 XRT 4 +/ /- 7 --(disk)----(disk)----(disk)-- Figure 2 (a,b,c). When applied to the TRACE data from 21-Apr-2002, the automated detection routine tracked a total of 54 verifiable downflows with 43 of them having complete or nearly complete tracks. The trajectories of all of the detected SADs are overlayed above the post-flare arcade in Figure 2c. In the north, downflows took the form of shrinking cusped loops; while in the south, primarily plasma voids were observed. Figure 3 (a,b,c). Output from the software includes average SAD area (10 6 km 2 ), average SAD head speeds (km/s), and initial magnetic flux (10 18 Mx). The TRACE flare results in a smooth distribution of these quantities. The largest errors in these values are due to inconsistencies in choosing footpoints, initial detection in the noise, and threshold usage (although error due to the latter has been greatly reduced from previous software versions). Table 1. Estimates of the mean values for initial area, average head speed, initial magnetic flux, loop shrinkage, and energy for each well-represented downflow detected by our automated routine. Except for the area, which is a result of resolution discrepancies, note the similarities in quantities between the instruments despite the large difference between temperature sensitivities. Downflows Area ~ 10 7 km 2 Speed ~ 10 2 km/s ∆L ~ 10 4 km ∆W ~ ergs Arcade B Flux ~ 2e10 18 Mx Magnetic Flux Tubes