CORONAL MAGNETIC RECONNECTION DRIVEN BY CME EXPANSION—THE 2011 JUNE 7 EVENT

For this work we use AIA (Lemen et al. 2012) on board SDO (Pesnell et al. 2012), which takes images of the full solar disk with a resolution of about 06 pixel⁻¹ through 10 filters, selected to include specific strong lines in the corona and continuum emission from the lower chromosphere. We investigate cold absorbing and hot emitting structures. Although we checked images in all 10 channels, for this analysis data from the 304, 171, 193, and 211 Å channels were the most suitable. The cadence was on average 12 s in the 171, 193, and 211 Å channels and 60 s in the 304 Å channel. The 304, 171, 193, and 211 Å passbands are dominated by ions formed at temperatures around 5 × 10⁴, 6.3 × 10⁵, a combination of 1.2 × 10⁶ and 2 × 10⁷ K, and 2 ×10⁶ K, respectively.

Additionally, 195 Å EUVI images are used from SECCHI (Howard et al. 2008) on STEREO-A, which was 949 ahead of Earth, and observed the eruption in the SE quadrant close to the limb.

We also use line-of-sight (LOS) magnetic field data from the Helioseismic and Magnetic Imager (HMI; Schou et al. 2012; Scherrer et al. 2012) on board SDO. The LOS magnetograms are taken in the 6173.3 ± 0.1 Å Fe i line, with an angular resolution 05 pixel⁻¹, a cadence of 45 s, and a noise level of 10 G.

As the CME eruption occurred close to the western limb (S22, W53), we used SDO/HMI data 4 days prior to the CME on 2011 June 3 at 20 UT for the analysis of the magnetic connectivities and topology in the three-AR complex including the CME source region AR 11226 and two other ARs (ARs 11227 and 11233) on its east side (Figure 1(a)). During the 4 days leading up to the eruption, a significant amount of magnetic cancellation took place within AR 11226, accompanied by weaker flux cancellation in the other two ARs (see Figure 4 and the SDO/HMI movie in the online version [movie2.mov]). Since all three ARs were in their decaying phase, their overall photospheric flux distribution did not change substantially, and their relative positions remained fairly constant, it is reasonable to assume that the overall topology of the magnetic connections between the ARs did not significantly change in these 4 days. We carried out a potential-field extrapolation (Figure 1(d)) of the SDO/HMI LOS magnetogram (Figure 1(c)), applying the method of Schmidt (1964).

Zoom In Zoom Out Reset image size

We also computed the location of quasi-separatrix layers (QSLs; Démoulin et al. 1996) in the three-AR complex on 2011 June 3. The location of the QSLs is determined by computing the values of the squashing degree, Q (Titov et al. 2002; Titov 2007), in the domain using the method described in Pariat & Démoulin (2012). The horizontal plane 1 pixel above the boundary (at z ∼ 4.5 Mm) is taken as the reference plane. This enables us to partly remove the QSLs related to small magnetic polarities and localized bald patches (as in Savcheva et al. 2012). We compute and plot the QSLs on this plane using an initial grid with a mesh size of 07 (∼0.5 Mm) with the Q-values eventually computed on an adapted mesh of length 5.5 × 10⁻³''. Regions where one of the footpoints of the field lines has a magnetic field intensity smaller than 10 G are also not considered. As a result, the large-scale QSLs separating the different connectivity domains are better highlighted. We also performed computation along a vertical cut of the domain, in an xz plane at y = −430''. The initial grid for that vertical computation has a mesh size of ∼15 (∼1 Mm) and reaches 1.1 × 10⁻²'' at the maximum resolution of the Q computation.

To further investigate the interaction between the CME and the neighboring AR 11227, we perform an idealized, three-dimensional MHD simulation. Although the parameters of the initial configuration are guided by the observed values, we do not attempt to match all numbers closely, as the simulation is intended to serve as a proof of concept, not to reproduce the observed eruption as closely as possible. We employ the coronal flux rope model of Titov & Démoulin (1999), hereafter TD, to construct the initial magnetic field in the simulation. The main ingredient of the TD model is a toroidal current channel of major radius R and minor radius a, positioned such that the symmetry axis of the torus is located at a depth d below the photospheric plane. The outward directed Lorentz self-force (or "hoop force") of the current ring is balanced by the force provided by the external field. This field is created by a pair of sub-photospheric point sources, ±q, which are placed at the symmetry axis, at distances ±L from the torus center (see Figure 2 of TD). The resulting force-free coronal equilibrium consists of an arched and line-tied flux rope embedded in the bipolar potential field.

We build an initial magnetic field that consists of three bipolar ARs, of which the western one (representing the CME source AR 11226) contains a TD flux rope, while the other two ARs contain purely potential fields. The chosen parameter values ensure that the rope is initially stable with respect to the helical kink (Török et al. 2004) and torus instabilities (Kliem & Török 2006). The AR centers are placed such that they correspond to the relative positions of the observed ARs. The distance between the two sub-photospheric point sources for each AR is also guided by the observations, and the signs of the point sources are chosen according to the observed polarities. In order to reflect that AR 11226 exhibited the largest flux, the point sources of the western AR are 25% stronger than those in the other ARs. Figure 1(b) shows the resulting magnetic configuration, which is similar to the potential field extrapolation of the three-AR complex in Figure 1(d).

We integrate the zero-β compressible ideal MHD equations, neglecting thermal pressure and gravity, which is justified in the low corona where the Lorentz force dominates. The MHD variables are normalized by the initial TD torus axis apex height, R−d, the maximum initial magnetic field strength, B_{0, max}, the maximum Alfvén velocity, v_{a0, max}, and derived quantities. The Alfvén time is τ_a = (R − d)/v_{a0, max}. We use a nonuniform Cartesian grid of size [ − 25, 25] × [ − 25, 25] × [0, 50] with a resolution of ≃ 0.04 in the flux rope area. The initial density distribution is ρ₀(x) = |B₀(x)|^3/2, such that v_a(x) decreases slowly with distance from the flux concentrations. For further numerical details we refer to Török & Kliem (2003).

We first relax the system for 58 τ_a and reset the time to zero. Then the eruption of the flux rope is triggered by imposing localized, strongly sub-Alfvénic converging flows at the bottom plane, similar to the flows applied in Török et al. (2011a). The flows slowly drive the polarities surrounding the flux rope toward the local inversion line, yielding a quasi-static expansion of the flux rope's ambient field. As a result, the flux rope slowly rises until it reaches the critical height for the onset of the torus instability, which then produces an eruption (see also Török et al. 2011b). Such converging motions, which were indeed observed to be persistent and strong, provide an efficient way to destabilize the flux rope and produce an eruption. As the SDO/HMI movie [movie2.mov] shows, the typical converging motions of opposite polarities induced by flux dispersion were enhanced by the fragmentation of the leading positive spot during the 7 days prior to the eruption and the highly unusual motion toward the main magnetic inversion line of one of the large spot fragments, which gradually canceled there, removing an equal amount of negative polarity.

Although we solve the ideal MHD equations, numerical diffusion occurs in regions of high current densities, enabling magnetic reconnection of field lines. As the diffusion in the simulation is much higher than expected in the solar corona, we refrain from a quantitative investigation of the reconnection scaling properties. In spite of this limitation, the simulation produces solutions that show an excellent match with the observed event (see Section  3.5).

The unusually massive filament considered in this paper formed in AR 11226, which was part of a complex of three adjacent decaying bipolar regions. It erupted at about 06:15 UT on 2011 June 7. The accompanying M2.5 flare had a start time of 06:16 UT and peaked at 06:30 UT. A dome-like bright front (Figure 3) formed by about 06:26 UT, which expanded laterally, reaching its maximum speed of ∼450 km s⁻¹ by 6:36 UT, after which the radial expansion started slowing down (for details see Cheng et al. 2012). The erupting filament in the core of this expanding dome-like shell, well behind the bright front, also expanded in the lateral direction while erupting (Figures 2 and 3). This lateral expansion brought it in contact with neighboring magnetic structures, in particular, with magnetic field lines/loops of AR 11227 (Figure 1). While the erupting flux rope was lifting and accelerating the bulk of the filament, many plasma blobs formed via the Rayleigh–Taylor (RT) instability (Innes et al. 2012; Carlyle et al. 2014) and began to detach from the filament. This occurred in the particularly massive parts of the filament, i.e., in its top part and in the northwest (NW) leg. These blobs obviously did not follow the field lines threading the filament in their return to the solar surface. However, their descent appeared to be influenced/guided by the magnetic field, implying that they were dragging field lines with them. On the other hand, along the straighter SE leg of the filament cool material was mainly flowing back along the field lines toward the SE footpoint (Figure 3), developing only minor indications of the RT instability. An interaction of the SE flank with external magnetic structures, via magnetic reconnection, became apparent in AIA images when the formation of a new branch of downward-flowing filament material was observed (see a snapshot of the well-developed process at 07:09 UT in Figure 2 and full detail in Figures 5–8, as well as the online multiwavelength movie [movie1.mov]). Dense, cool filament plasma originally contained within the erupting flux rope was redirected toward remote magnetic footpoints in the neighboring AR 11227, thus unambiguously tracing the change of large-scale magnetic connectivity.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The back-flowing material from the ejected filament is optically thick, in that it obscures emission from the corona behind it through photoionization (i.e., photoelectric) opacity. The strength of this absorption allows the column density of the absorbing species—H i, He i, and He ii—to be determined. In principle, the optical depth of this absorption may be measured by comparing the intensity of pixels containing this prominence matter with pixels on either side; this is known as a spatial-interpolative approach and has been exploited to great effect by Kucera et al. (1998), Golub et al. (1999), and Gilbert et al. (2005, 2006), and it is appropriate in cases where the background emission has to be modeled because it is not seen directly. Here, however, we take advantage of the fact that the low-ionization material sweeps through the AIA image, and we use images just before its arrival to model the background radiation field: this constitutes a temporal-interpolative approach. In doing so, we can more accurately estimate the column depth of material contained in the back-flowing material just before it enters the region of interest. Under the assumptions that the foreground EUV emission is negligible and that the depth of the plasma blob/thread is equal to its FWHM in the plane of the sky (approximately 4''), a lower limit of the electron density in the redirected filament plasma in the vicinity of the reconnection region was estimated to be 10¹⁰ cm⁻³ (Figure 5; for details see Williams et al. 2013). This is more than one order of magnitude larger than the typical coronal density of bright loops.

The redirection of back-flowing filament plasma started at around 06:57 UT, and by 07:00 UT plasma was flowing toward the negative-polarity footpoints of AR 11227 in well-developed threads (see the AIA 171 Å data in Figure 6 and the accompanying movie [movie1.mov]). Around the location of the sharp change in plasma flow direction, some sporadic brightenings have appeared by 07:00 UT, which intensified with time. The brightest emission was observed in the 171 Å channel between 07:04 and 07:06 UT. The appearance of emission in EUV around the redirection/reconnection site clearly indicates that in situ heating was occurring there. Particles accelerated during the reconnection process were in an unusually dense coronal environment and transferred their kinetic energy through Coulomb collisions to the filament plasma. Heating by slow-mode shocks (i.e., Petschek shocks) is also expected to take place.

Stack-plots of two series of narrow cuts approximately along (A-cuts) and perpendicular to (B-cuts) the threads of the down-flowing plasma redirected by reconnection are shown in Figure 7, revealing the evolution of plasma behavior around the reconnection region. In the top panel of Figure 7 the location of the cuts is indicated over an AIA 171 Å base-difference image. The middle panels show stack-plots along the series of A-cuts, the bottom panel along the B-cuts.

Within the SE leg of the erupting and laterally expanding filament, plasma blobs were swirling in a clockwise direction toward the SE (negative) magnetic footpoint in AR 11226. As the motions were quite complex and formed part of a complex evolution with blobs/threads of different heights overlapping in the images, the interpretation of the stack-plots is not straightforward.

The B-cuts run roughly parallel to the original (projected) flow direction along the SE leg of the erupting filament (indicated by a long curved arrow in Figure 7 top panel), and the B3–B4 cuts capture the proper motion of some of the blobs. However, what we see in the B0–B2 cuts is not real plasma motion, as different blobs, presumably on different field lines, enter and leave the cut directions in succession, forming an apparent continuous thread. Eventually, some of the blobs enter the reconnection region and get heated. As the reconnection region is shifting along the B-cut direction, the stack-plots capture that motion as a bright track.

The cuts A0–A1 capture the entire path of the redirected filament plasma, which is moving along reconnected field lines. These cuts show many thin threads, which were formed at different times and give the impression of torsional motion around one another (A1). The cold plasma flowing toward the footpoints became heated in lower-height, higher-density layers. As there are no bright V-shaped features around the footpoints in the stack-plots of the cuts A0–A1, chromospheric evaporation was apparently not strong.

In Figure 6 the bright reconnection region (1) extends in length along the A-cut direction and (2) between 07:04:24 and 07:06:48 UT presents a significant proper motion along the A-cut direction. These both are manifested in the A-cut stack-plots. In the upper (filament-side) part of the A-cut stack-plots, seen above the bright reconnection region in Figures 7 and 8, the most striking features are V-shaped threads. The downward branch of the V shows expanding (southeastward) motion of the filament leg before reconnection. The upward branch of the V shows northwestward-directed downflow along the non-reconnected part of the flux in the filament leg accelerating due to gravity into AR 11226.

A similar V-pattern appears in the stack-plots presented by Su et al. (2013) for the 2011 August 17 limb flare, interpreted as reconnection inflow and outflow in the flare using coronal loops as tracers. However, in our case the picture was made more complex by the presence of cold plasma blobs as tracers. These plasma blobs populated only segments of a field line and even moved along the magnetic field lines. Therefore, the stack-plots of the reconnection region presented by Su et al. (2013) and our Figures 7 and 8 are not directly comparable.

The multiwavelength capability of SDO allows a rough diagnostic of temperature dependence of plasma heating in the vicinity of the reconnection region and the footpoints. The most prominent brightening is seen in the 171 Å waveband, indicating that the bulk of the plasma has been heated to ≈6 × 10⁵ K. This is the case for both the reconnection region and the footpoints (see Figure 8). The emission of the reconnection and footpoint regions is weaker in the 193 Å waveband, which images plasma at ≈1.2 × 10⁶ K. The weakest emission is seen in the 211 Å waveband, which images plasma at ≈2 × 10⁶ K. This indicates that these regions are only heated to the lower end of the coronal temperature range.

The emission region seen in Figure 6 is shifting toward the west, which is also seen in Figure 7 (middle panels, A2–A3) as an upward-propagating branch of the emission region. This westward shift of the bright emission region might result from successive reconnections progressing toward the core of the filament flux rope, but the relatively high displacement velocity of ∼90 km s⁻¹ makes it more likely that the reconnection region shifts westward. Such a shift could be the result of the large-scale flows from the encounter of the filament flux with the flux of AR 11227. These flows have a downward component inside the leg of the filament, which drives the interaction. This represents a flow along the current sheet and may cause a shift of the reconnection X-line in the direction of the flow, westward in projection. Such a flow is not normally included in reconnection models.

How can we put further constraints on the identity of the magnetic loops that reconnect with the filament's expanding flux rope? We use the tools of magnetic topology analysis: a magnetic extrapolation and a map of QSL locations for this.

As the location of footpoint brightenings (Figures 9(a) and (c)) provides insight into the reconnecting loops' identity, we use a magnetic extrapolation (Figure 9(b)) and a map of QSLs (Figure 9(d)) in the three-AR complex to explore this question. QSLs are locations where there is a large gradient (i.e., drastic change) in magnetic connectivity (for the method, see Section 2.2). Comparing the four panels in Figure 9, it is apparent that the impact-brightened footpoints are over the negative polarity of AR 11227. This is where two sets of loops originate: one set of medium-long loops (in red) connecting to the leading positive polarity of AR 11227, and another set of long loops (in blue) connecting to the leading positive polarity of AR 11226 (Figure 9(c)). Between these two sets of loops, a QSL is present (Figure 9(d)). The impact-brightened footpoints are on both sides of this QSL, implying that magnetic field lines of the erupting filament reconnected with both sets of these magnetic loops.

Zoom In Zoom Out Reset image size

We found a group of null points over a quiet-Sun area outside of the complex, at a height of 4 Mm at around y = −480'' along the QSL extending southward from the leading polarity of AR 11227 (Figure 9(d)). These quiet-Sun nulls did not seem to play any significant role in the reconnection process described here, although at the start of the reconnection event a narrow channel of southward-flowing filament material formed in the direction of these nulls.

The topology analysis revealed the existence of a hyperbolic flux tube (HFT; Titov et al. 2002) at a height of ≈80 Mm (110''), as shown in Figure 10(d), between the connectivity domains of ARs 11226 and 11227 (cf. Figures 9 and 10), at the location where the reconnection is observed to take place. It is most likely that the expanding flux rope of the CME/filament provided a forcing along this HFT, which led to the observed reconnection.

Zoom In Zoom Out Reset image size

A weak guide field points in the direction of the HFT, so that there is no null point (i.e., no fan-spine topology) at the location of the HFT. It is also noteworthy that the ratio of total flux in AR 11226 to the total flux in the two other ARs decreased considerably from June 3 through 7, due to particularly strong flux cancellation in AR 11226. Therefore, the flux lobes of AR 11226 and AR 11227 are likely more symmetric at the time just before the eruption, i.e., somewhat closer to the simulation configuration (cf. Figures 10(b) and (c).

In support of our conjecture that the observed redirection of filament material and the associated coronal brightening were signatures of magnetic reconnection between the CME and AR 11227, we performed an MHD simulation as described in Section 2.3. The dynamic evolution of the simulated eruption is shown in Figure 11 and can be appreciated in the animation accompanying Figure 12, which contains co-temporal AIA 171 Å and 304 Å observations and corresponding simulation frames. The simulation reproduces to a very good approximation the huge expansion and global structure of the erupting flux rope, as well as the formation of a current layer consistent with the location of the coronal brightening. The formation of such a layer and subsequent reconnection are expected in a magnetic topology like the one studied here, if the system is subject to a sufficiently strong perturbation, as in a CME. In the following, we describe the main field line connectivity changes that occur as a result of reconnection in the simulation.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 11(a) shows the configuration after the initial relaxation, before the converging flows used to trigger the eruption are applied. The similarity of the field line connectivities with the extrapolated fields shown in Figures 9 and 10 is obvious (see also Figure 10(c); the purple field lines shown in that figure are omitted here for clarity). The simulation field model also contains an HFT, running predominantly along the y direction. As for the extrapolated fields, the presence of a weak guide field precludes the presence of a null point in the HFT.

The flux rope is visualized by two flux bundles (golden field lines), both of which are rooted in the rope's positive-polarity footpoint. The outer, flat bundle is located at the rope's periphery, while the inner, arched bundle is located close to the flux rope center. The remaining field lines represent initially potential flux. An arched current layer (red iso-surface) has formed at the HFT (cf. Figure 10(c)), as a result of the dynamics present during the relaxation (see Figure 4 in Török et al. 2011b for a similar case).

As the flux rope erupts and expands (Figure 11(b)), the green field lines located directly above it are pushed against the current layer from the top right direction, leading to a tilt of the layer that is favorable for reconnection between the green and magenta field lines. Once the current density has sufficiently steepened, these two flux systems indeed reconnect, producing new, low-lying connections between AR 11226 and AR 11227 (not shown in the figure, but see Figure 3 in Török et al. 2011b for a similar case); additionally, connections are formed that arch over the whole AR complex. However, this process is not expected to produce the observed redirection of filament material, since flux rope field lines have not yet reconnected. The eruption of the flux rope leads to the formation of a second, vertical, current layer between the rope's legs (the canonical "flare current sheet"; cyan iso-surface).

Following the eruption further, connections between the flux rope and AR 11227 develop in two ways (Figures 11(c) and (d)). These correspond to the observation of brightenings on either side of the QSL over the negative polarity in AR 11227, as shown in Figure 9. First, as the lower part of the expanding southern flux rope leg approaches the red current layer, field lines at its periphery are dragged into the reconnection region and reconnect with flux rooted in the negative polarity of AR 11227 (see the zoom in Figure 11(d)). The location of this current layer and the bottom part of the reconnected field lines correspond to the location of the EUV brightening at the reconnection region (Figures 9 and 12) and to the path of the redirected filament material, respectively, supporting our interpretation of the data. We can tentatively deduce from the simulations that the long (eastward-pointing) arm of the Y-shaped brightening corresponds to the bottom part of the reconnected field lines. The westward-pointing arm of the brightening corresponds to flux rope field lines that did not (yet) reconnect—they represent the inflow region of the reconnection process. The short southwestward-pointing third arm illuminates the segments of these two sets of field lines above the reconnection region, which lie next to each other in the downward-pointing flux of the southern flux rope leg. The remaining element of the reconnection process is flux originating in the positive polarity of AR 11227 (similar to the magenta field lines in panel (a) before reconnection), connecting to the negative polarity of AR 11226 after reconnection. These do not carry filament material and are not included in the figure, to avoid a crowded appearance. In this interpretation, the current sheet is located between the leftward- and rightward-pointing arms of the brightening and passing through the short third arm.

Second, field lines initially or temporarily arching over the expanding flux rope reconnect in the cyan current layer with flux rope field lines and become part of the CME. The latter include blue field lines and magenta field lines that have previously reconnected at the red current layer and temporarily overarch the flux rope as well. Both reconnection processes occur more or less simultaneously and result in new connections between the erupting flux rope and the negative polarity of AR 11227. However, the reconnection occurring at the cyan current layer cannot be expected to be responsible for the observations of the deviating filament material, since (1) it will not produce a brightening at the observed location in the corona (Figures 12(a)–(c)), which we associate with the red current sheet (Figure 11(d)), and (2) the resulting field line connections between AR 11226 and AR 11227 are made up for most of their length from initially potential field that cannot carry filament material.

In order to clarify the role of the large lateral expansion of the erupting filament and investigate whether or not it is indeed the driver of the reconnection between the erupting flux rope and field lines of AR 11227, we computed the maximum current density in the red current layer for the times shown in Figures 11(a)–(d) and found a strong (15- to 20-fold) enhancement of J, before reconnection begins in this (red) current sheet in the simulation. Reconnection is much more likely to occur, or to become faster, when the current sheet is getting thinner and, correspondingly, J is becoming higher. Titov et al. (2003) and Galsgaard et al. (2003) found that an HFT can be pinched into a current sheet in the presence of appropriate external driving motions. In this case, it is most likely that such driving motions were mainly provided by the lateral expansion of the unstable flux of AR 11226. Therefore, we conclude that the lateral expansion of the flux rope erupting in AR 11226 was the major driver of its interaction/reconnection with AR 11227.

We note that the overall reconnection in the simulation is more complex than just described. For example, a significant number of field lines reconnect subsequently in both current layers, and some appear to do so several times, repeatedly changing their connectivity. Also, there is reconnection occurring at a current layer that forms between AR 11233 and AR 11226 (not shown in Figure 11 for clarity). As a consequence, magenta and blue field lines, starting in AR 11227 and arching over the erupting flux, change their footpoints a second time to connect to AR 11233 (Figure 11(d)). A detailed investigation of the evolution and interplay of all reconnection processes is beyond the scope of this paper. The reconnection primarily relevant in the present context is the one occurring in the red current layer between the southern flux rope leg and the flux initially connecting the polarities within AR 11227, as described above.