COMPLEX FLARE DYNAMICS INITIATED BY A FILAMENT–FILAMENT INTERACTION

The filaments under study were located in NOAA AR 11777 (E71S16, Figure 1), and appeared near the eastern limb on 2013 June 21, as viewed from the perspective of the Solar Dynamics Observatory (SDO). The Atmospheric Imaging Assembly (AIA; Lemen et al. 2012) on board SDO takes full-disk images of the Sun in seven extreme-ultraviolet (EUV) channels ($\mathrm{log}T$ ranges 5.6–7.3) and three UV to visible channels ($\mathrm{log}T$ ranges 3.7–5.0), with a pixel scale of 06 pixel⁻¹ and a cadence of 12 s. The Helioseismic and Magnetic Imager (HMI; Schou et al. 2012) on board SDO provides full-disk vector magnetograms (Hoeksema et al. 2014), with a pixel scale of 05 pixel⁻¹ and a cadence of 12 minutes.

Zoom In Zoom Out Reset image size

The evolution of the filaments was also observed near the western limb from the perspective of the Solar Terrestrial Relations Observatory Behind spacecraft (STEREO-B), with a separation angle of approximately 140° away from SDO. The Extreme-ultraviolet Imager (EUVI; Wuelser et al. 2004) on board the STEREO took images at four bandpasses centered at 171, 195, 284, and 304 Å, with a spatial resolution of 16 pixel⁻¹. The STEREO/EUVI had an image cadence of 5 minutes in the 195 Å filter and a 10 minute cadence at 304 Å for the data used in this study.

In order to reduce the impact of projection effects on the line of sight (LOS) magnetogram due to the location of AR 11777 near the solar limb (see the small panel in Figure 1(c)), we utilize an LOS magnetogram obtained four days later, when the projection effects were minimal and differentially rotated to the same time as that of Figure 1(a) and displayed in Figure 1(c). The related PIL in this active region is determined based on the smoothed LOS magnetogram and indicated in the figure with a thick line. Two filaments were suspended above the same segment of the PIL: one filament was lower and shorter, and the other was higher and longer, as seen in Figure 1(a). These two filaments are the focus of the present study. In addition, a related thin loop-like structure (LLS) is observed to lie above the upper filament.

The observations during one day before the eruption reveal the formation of this complex configuration. The upper filament began to rise to a higher altitude after 11:00 UT on 2013 June 20. The lower filament started to ascend slowly about 15 hr later. During this interval, several strands of material were observed to transfer from the lower filament to the upper one, similar to previous studies on double-decker filaments (Liu et al. 2012; Zhu & Alexander 2014). A height-time stack plot (Figure 2(b)) was generated from a series of SDO/AIA 193 Å images along a slit marked in Figure 2(a). Each material transfer is indicated by an upward arrow. A few negative slopes, denoted by the downward arrows, correspond to mass drainage (see Figure 2). These observations suggest that both filaments were magnetically connected and associated with the same filament channel.

The LLS was observed to be close to but separated from the upper filament shortly following a material transfer episode that occurred at around 23:40 UT on 2013 June 20, as seen in animation a of Figure 2, indicating that the LLS was closely related to the upper filament. After 23:50 UT, the upper filament became very faint, possibly due to mass drainage and/or heating, until it reappeared in the AIA 193 and 304 Å about two hours later (see Figure 2(b)).

With the data from both SDO and STEREO-B (Figures 1(d) and (e)), the three-dimensional positions of both filaments before the eruption were measured by the SolarSoft routine scc_measure (Li et al. 2011; Thompson et al. 2012). However, this analysis was not applied to the LLS or the two ends of the upper filament, which were indistinguishable against the background emission when seen from STEREO-B. The reconstructed locations are rotated to the center of the solar disk and shown in Figure 1(f). When seen from above as in Figure 1(g), the upper filament appeared as a "C" shape. There was a crossing angle of around 20° between the projections of the two filaments onto the solar surface.

The lower filament began to rise at around 02:20 UT on 2013 June 21. The evolution of this filament eruption is displayed in Figure 3. A "slit" is placed along the east–west direction (the dash–dotted line in Figure 3(b)) in a series of AIA 304 Å images to generate a stack plot as shown in Figure 4(a), in which the height of the lower filament is tracked by the dotted curve. From the time derivative of this height-time evolution, the velocity of the filament ascent is determined and is found to exhibit four distinct phases (Figure 4(c)): (1) a slow-rise phase from 02:20–02:30 UT, with the velocity below 10 km s⁻¹, (2) a short fast-rise phase, from 02:30–02:35 UT, with the velocity changing from ∼10 to 90 km s⁻¹, (3) a minor relaxation phase with the velocity decreasing from 90 to 60 km s⁻¹, between 02:35–02:43 UT, and (4) another fast-rise phase starting from 02:43 UT, with the velocity increasing to ∼400 km s⁻¹ by around 02:53 UT.

A bright front ahead of the erupting filaments was observed to form at ∼02:38 UT, as seen in AIA 171 Å (Figures 3(k)–(m) and the animation). With a stack plot (Figure 4(b)) along the same location in Figure 3(b), the velocity of the bright front is determined and shown in Figure 4(c). Its value increased from 40 km s⁻¹ at 02:41 UT to 340 km s⁻¹ at 02:53 UT. This velocity is slightly slower than that of the erupting filament, indicating that it is a compression front driven by the erupting filaments. The successful eruption was associated with a partial halo CME (Figure 3(n)) reported by Solar and Heliospheric Observatory.⁶ A similar bright front was reported by Zhang et al. (2012), and was interpreted as the enveloping front of the CME associated with a filament eruption.

Solar flaring that was related to this filament eruption displayed two peaks in the GOES X-ray flux, as shown in Figure 5(a), a C3.0 peak followed by a transition to a second, much larger, M2.9 peak. As both of them show distinct peaks in soft and hard X-rays, and the hard X-ray light curves roughly returned to the background level between them (Figure 5(a)), we treat them as two distinct solar flares overlapping in time. In the remainder of this section, we investigate in detail the interaction between the two filaments and the evolution of their eruptions.

Zoom In Zoom Out Reset image size

Plasma heating. Prior to the eruption, the upper filament, which appeared in emission in AIA 304 Å (Figure 3(a)), was barely identifiable in the hot EUV channels such as AIA 131 Å (Figure 3(f)), possibly due to its lower temperature and density. However, at the time of the filament–filament interaction, ∼02:32 UT on June 21, a complex region of hot plasma appeared in the vicinity of the upper filament and the LLS, as seen in AIA 131 Å (Figures 3(g)–(i)). The observation in the AIA 1600 Å channel shows an extended chromospheric ribbon (Figure 3(e)), which is composed of the footpoints of numerous bright loops (Figures 3(g) and (h)) connected through this hot plasma region. At the same time, all four filament footpoints (marked by diamonds and squares in Figure 3(e)) are found to be brightening in AIA 1600 Å. The enhancement in the hot plasma was most evident at around 02:38 UT (Figure 3(h)). As AIA 131 Å is sensitive to both hot and cold plasmas at 10 MK and 0.4 MK, respectively, we reconstructed the differential emission measure (DEM) of the region using the code developed by Plowman et al. (2013) and calculated the DEM-weighted average temperature $\langle T\rangle$ (e.g., Cheng et al. 2012; Gou et al. 2015; Guidoni 2015), as shown in Figure 3(j). The value of $\langle T\rangle$ in the vicinity of the upper filament ranges from 7 to 12 MK.

Converging motion. The motions of the two filaments are determined by tracking the moving features evident in the generated stack plot (Figure 4(a)), with derived velocities shown in Figure 4(d). Immediately prior to the filament interaction at 02:32 UT, the descending upper filament structure reached a speed of 10 km s⁻¹. At this time, the lower filament was rising at ∼40 km s⁻¹. This observation indicates that the filaments converged with a relatively rapid rise of the lower filament and a slow descent of the upper filament.

It is noticeable that the eastern leg of the LLS was brightening in AIA 304 Å as the interaction was in progress (Figures 4(e)–(g)). This leg, lying behind the upper filament before the interaction, appeared to be crossing through the upper filament to the foreground after ∼02:32 UT (Figure 4(f) and the animation). The brightening and the apparent crossing of the LLS indicate its direct involvement in the interaction with the rising lower filament.

The GOES C3.0 flare. Coincident with the filament interaction, the GOES satellite detected the onset of a C3.0 flare. A hard X-ray burst observed by the Fermi Gamma-ray Burst Monitor (Meegan et al. 2009) initiated at 02:32 UT and lasted for about 7 minutes (Figure 5(a)). Three intervals were chosen to study the spatial evolution of the RHESSI X-ray sources (denoted by the first three dashed lines in Figure 5(a), and corresponding to the early, middle, and late intervals of the burst, respectively). The results are displayed in Figures 5(b)–(d) and (f)–(h). At 02:33 UT (Figures 5(b) and (f)), a coronal X-ray source was detected near the interface of the two filaments, emitting in both soft and hard X-rays. Three footpoint sources are identified: two of them were located near the footpoints of the lower filament (indicated by two diamond symbols), and a third one (marked by the southern square symbol), which was relatively weak, lay near the southern footpoint of the upper filament. Beginning at 02:34 UT, the northmost footpoint source became dominant in the hard X-ray emission (Figures 5(c) and (g)), indicating that this flare is asymmetric (Alexander & Coyner 2006; Coyner & Alexander 2009). At 02:38 UT (Figures 5(d) and (h)), with the decay of the hard X-ray flux (Figure 5(a)), the nature of the X-ray sources is difficult to determine due to the complexity of the configuration and possible line of sight confusion of the various features.

An M2.9 solar flare began at around 02:45 UT, five minutes after the decay phase of the C3.0 flare (Figure 5(a)), and lasted for more than 3 hr, until ∼06:00 UT. RHESSI observations taken near this flare peak are shown in Figures 5(e) and (i). A loop-top source was located above the brightening arcade, with a projection of a hard X-ray source near the northern footpoint of the lower filament.

Field Line shrinkage (or loop retraction, e.g., Forbes & Acton 1996; McKenzie & Hudson 1999) was observed during this M flare. Specifically, two locations are identified at the northern and southern ends of the post-flare arcade at 03:28 UT (Figure 6(a) and the animation). In both regions, shrinking loops are observed to retract toward the post-flare loops. The velocity of the retracting loops is shown via a stack plot from a cut along the east–west direction (along the arrow in Figure 6(a), with the arrow indicating the direction of motion). Two moving features, denoted by the dashed lines in Figure 6(b), correspond to two distinct retracting loops. The velocities of both loops were observed to be decreasing as they approached the post-flare loops, changing from a few hundreds km s⁻¹ to rest (Figure 6(c)).

Zoom In Zoom Out Reset image size

It is also interesting to note that several tadpole-like dark voids were evident in AIA 131 Å (the animation of Figure 6), successively travelling downward, in this flare. These tadpole-like voids are also called supra-arcade downflows (SADs; e.g., McKenzie & Hudson 1999; Liu 2013). The box region in Figure 6(d) shows one example of these SADs. The location of the SADs, directly above the center of the flare loop arcade, was distinct from the region of the observed pronounced field line shrinkage. The trajectory of the SAD is determined, again, from the stack plot method (Figure 6(e)) and the velocity derived is shown in Figure 6(f). The SAD structure descended with an initial velocity of 240 km s⁻¹ at 03:57 UT until it came to rest at the flare loop region 10 minutes later. Additional SADs were still observable by the end of this flare, at ∼06:00 UT.

Kumar et al. (2010) reported a converging motion of two filaments at a speed of ∼10 km s⁻¹ during their slingshot reconnection. This value is typical for reconnection inflows detected at EUV wavelengths (Yokoyama et al. 2001; Liu et al. 2010), and is consistent with the theoretical expectations of Petschek (1964). In the present study, immediately prior to the direct interaction of the two filaments, the upper filament was observed to descend at ∼10 km s⁻¹, while the eruptive lower filament rose up with a velocity of ∼40 km s⁻¹. The velocities of the observed converging motions are comparable with previous studies.

While the dynamical motions are similar to those observed in previous observations, it would appear that in this case the driving mechanisms for the converging motion are different. In the study of Kumar et al. (2010), the approaching filaments are thought to be driven by slow photospheric motions. The filament–filament interaction reported by Jiang et al. (2013) is initiated by one of the filaments that erupted. In the event reported here, the lower filament is eruptive, similar to Jiang et al. (2013), but the descending motion of the upper filament requires different explanations. A possible candidate is the ${\boldsymbol{J}}\times {\boldsymbol{B}}$ force between the two filament current systems. This scenario is expected from the simulation of a full eruption of an unstable double flux rope (see Figure 3 in Kliem et al. 2014) and the laboratory experiment of two parallel current channels by Intrator et al. (2009). An alternative explanation calls for an increase in the magnetic tension of the strapping fields that regulate the height of the upper filament, though this scenario lacks observational evidence, as the evolution of the strapping field is difficult to diagnose in this event. Furthermore, the eastern leg of the LLS, which appeared to pass through the upper filament during the interaction, might serve to depress the upper filament and contribute to its observed decent. To our knowledge, this is the first time that a converging motion has been observed between filaments in a double-decker configuration.

The newly formed hot region of plasma in the vicinity of the upper filament consists of numerous hot loops (Figure 3(h)). This hot plasma may be heated in a quasi-separatrix layer (Démoulin et al. 1996), which wraps around the upper filament, separating its twisted magnetic fields from the outer, untwisted fields. As the two filaments interact, accelerated electrons stream along the loops in this layer and deposit energy into the footpoints of these loops as the electrons are stopped by the dense chromosphere, producing the extended ribbon observed in AIA 1600 Å (Figure 3(e)).

A coronal hard X-ray source is identifiable near the interaction interface of the filaments, as seen in Figure 5(b). Due to the limited resolution of RHESSI, this coronal source could have two origins: ongoing reconnection between the interacting fields of the two filaments, and/or a loop-top source located at or above the apex of the flare loops (e.g., Masuda et al. 1994; Liu et al. 2013). However, the coronal source dominated the hard X-ray emission at the onset of this C flare, unlike a typical loop-top source, which is usually fainter than its associated footpoint sources (Petrosian et al. 2002). Thus, we suggest that the observed coronal source at least partially resulted from ongoing reconnection between the magnetic fields around both filaments. It is worth noting that the coronal hard X-ray source is barely detected by RHESSI in the later phase of the C3.0 flare (Figures 5(c) and (d)). This may be because the footpoint source had grown in intensity to dominate over the coronal source and the limited RHESSI dynamic range would make it extremely difficult to detect any weak coronal emission (e.g., Sui et al. 2004). Thus, it is difficult to determine whether the two filaments were still interacting or not in the later stage of the eruption.

Both the upper and lower filaments are suspended above the same region of AR 11777 and oriented with a small projected angle of around 20° between them. In the simulations of Linton et al. (2001) and Kliem et al. (2014), such a configuration would result in the merger of the two filaments following the initial interaction. In our study, there are no evident signatures for the other types of interactions discussed by Linton et al. (2001) (i.e., bounce, tunnel, and slingshot). As the lower eruptive filament interacts with the upper filament, the combined system continues to rise until both finally erupt away. From this point of view, both filaments are assumed to have merged, at least partially, to form a complicated eruptive structure with four distinct footpoints rooted in the chromosphere.

Both the observed field line shrinkage and the SADs are generally thought to be consequences of the newly reconnected evacuating flux tubes retracting from the reconnection site above the bright arcade of loops during solar flares (Švestka et al. 1987; Forbes & Acton 1996; McKenzie & Hudson 1999; McKenzie 2000). Whether the retractions appear as loop shrinkage or sunward dark voids depends on the different viewpoints (Figure 2 in Savage & McKenzie 2011): when the retracting flux tubes are viewed along the axis of the flare arcade (i.e., face-on), they appear as shrinkage; when viewed perpendicular to the axis (i.e., side-on) they are observed as dark voids (Savage & McKenzie 2011; Savage et al. 2012). Warren et al. (2011) demonstrated this speculation using STEREO and SDO observations. In our study, the post-flare loop arcade appears C-shaped (see Figure 6(d)), and fans out toward the dispersed positive polarities in the east, with the footpoints of individual loops concentrated near the compact sunspot in the west. Due to the particular shape of this arcade, our view toward the loops changes from face-on in the south to side-on at the center, and becomes face-on again in the north of the arcade. As a result, field line shrinkages at the two ends of the arcade were observed in a face-on view, and SADs in the center were observed in a side-on view. Warren et al. (2011) also reported a similar result in a Γ-shaped post-flare loop arcade, though the configuration here may be more complicated. The initial velocities of the two loop shrinkage and single SAD events are 260, 120, and 240 km s⁻¹, respectively. These values are typical of SADs (e.g., Savage & McKenzie 2011).

The observation of loop shrinkage and SADs indicates that a large-scale vertical current sheet existed above the flare loop arcade (McKenzie & Hudson 1999; Liu et al. 2013; Liu 2013). In our study, the overlying magnetic field lines might become highly stretched due to the filament eruption, forming a vertical current sheet underneath it. This scenario is expected in the standard flare model (Kopp & Pneuman 1976). Thus, we suggest that the process for this M2.9 flare follows the standard model.