MULTIWAVELENGTH OBSERVATIONS OF A PARTIALLY ERUPTIVE FILAMENT ON 2011 SEPTEMBER 8

Solar prominences or filaments are cool, dense plasmas embedded in the million-Kelvin corona (Mackay et al. 2010). The plasmas originate from the direct injection of chromospheric materials into a pre-existing filament channel, levitation of chromospheric mass into the corona, or condensation of hot plasmas from the chromospheric evaporation due to the thermal instability (Xia et al. 2011, 2012; Keppens & Xia 2014; Zhou et al. 2014). Prominences are generally believed to be supported by the magnetic tension force of the dips in sheared arcades (Guo et al. 2010b; Terradas et al. 2015) or twisted magnetic flux ropes (MFRs; Cheng et al. 2012, 2014a; Su & van Ballegooijen 2012; Sun et al. 2012a; Zhang et al. 2012a; Xia et al. 2014a, 2014b). They can keep stable for several weeks or even months, but may get unstable after being disturbed. Large-amplitude and long-term filament oscillations before eruption have been observed by spaceborne telescopes (Chen et al. 2008; Li & Zhang 2012; Zhang et al. 2012b; Bi et al. 2014; Shen et al. 2014) and reproduced by numerical simulations (Zhang et al. 2013), which makes filament oscillation another precursor for coronal mass ejections (CMEs; Chen 2011) and the accompanying flares. When the twist of a flux rope supporting a filament exceeds the threshold value (2.5π–3.5π), it will also become unstable and erupt due to the ideal kink instability (KI; Hood & Priest 1981; Kliem et al. 2004; Török et al. 2004, 2010; Fan 2005; Srivastava et al. 2010; Aschwanden 2011; Kumar et al. 2012). However, whether the eruption of the kink-unstable flux rope is failed or ejective depends on how fast the overlying magnetic field declines with height (Török & Kliem 2005; Liu 2008; Kumar et al. 2010). When the decay rate of the background field exceeds a critical value, the flux rope loses equilibrium and erupts via the so-called torus instability (TI; Kliem & Török 2006; Amari et al. 2014; Jiang et al. 2014). On the other hand, if the confinement from the background field is strong enough, the filament will decelerate to reach the maximum height before falling back to the solar surface, which means the eruption has failed (Ji et al. 2003; Liu et al. 2009; Guo et al. 2010a; Kumar et al. 2011; Joshi et al. 2013, 2014; Song et al. 2014).

In addition to successful and failed eruptions, there are also partial filament eruptions (Gilbert et al. 2007; Liu et al. 2007). After examining 54 Hα prominence activities, Gilbert et al. (2000) found that a majority of the eruptive prominences show a separation of escaping material from the bulk of the prominence; the latter initially lifted away from and then fell back to the solar surface. To explain the partial filament eruptions, the authors proposed a cartoon model in which magnetic reconnection occurs inside an inverse-polarity flux rope, leading to the separation of the escaping portion of the prominence and the formation of a second X-type neutral line in the upper portion of the prominence. The inner splitting and subsequent partial prominence eruption was also observed by Shen et al. (2012). Gilbert et al. (2001) interpreted an active prominence with the process of vertical reconnection between an inverse-polarity flux rope and an underlying magnetic arcade. Liu et al. (2008) reported a partial filament eruption characterized by a quasi-static, slow phase and a rapid kinking phase showing a bifurcation of the filament. The separation of the filament, the extreme-ultraviolet (EUV) brightening at the separation location, and the surviving sigmoidal structure provide convincing evidence that magnetic reconnection occurs within the body of a filament (Tripathi et al. 2013). Gibson & Fan (2006a, 2006b) carried out three-dimensional (3D) numerical simulations to model the partial expulsion of an MFR. After multiple reconnections at current sheets that form during the eruption, the rope breaks in an upper, escaping rope and a lower, surviving rope. The "partially expelled flux rope" model has been justified observationally (Tripathi et al. 2009). Tripathi et al. (2006) observed a distinct coronal downflow following a curved path at the speed of <150 km s⁻¹ during a CME-associated prominence eruption. Their observation provides support for the pinching off of field lines drawn-out by the erupting prominences, and the contraction of the arcade formed by the reconnection. Tripathi et al. (2007) reported similar multithermal downflow at the speed of ∼380 km s⁻¹, starting at the cusp-shaped structures where magnetic reconnection occurred inside the erupting flux rope that led to its bifurcation. Liu et al. (2012) studied a flare-associated partial eruption of a double-decker filament. Cheng et al. (2014b) found that a stable double-decker MFR system existed for hours prior to the eruption on 12 July 2012. After entering the domain of instability, the high-lying MFR impulsively erupted to generate a fast CME and GOES X1.4 class flare, whereas the low-lying MFR remained behind and continuously maintained the sigmoidicity of the active region (AR). From previous literature, we can conclude that magnetic reconnection and the release of free energy is involved in most of the partial filament eruptions. However, the exact mechanism of partial eruptions, which is of great importance to understanding the origin of solar eruptions and forecasting space weather, remains unclear and controversial.

In this paper, we report multiwavelength observations of a partial filament eruption and the associated CME and M6.7 flare in NOAA AR 11283 on 8 September 2011. The AR emerged from the eastern solar limb on 30 August 2011 and lasted for 14 days. Owing to its extreme complexity, it produced several giant flares and CMEs during its lifetime (Dai et al. 2013; Feng et al. 2013; Jiang et al. 2014; Li et al. 2014; Liu et al. 2014; Ruan et al. 2014). In Section 2, we describe the data analysis using observations from the Big Bear Solar Observatory (BBSO), the Solar and Heliospheric Observatory (SOHO), the Solar Dynamics Observatory (SDO), the Solar Terrestrial Relation Observatory (STEREO; Kaiser 2005), GOES, RHESSI (Lin et al. 2002), and WIND. Results and discussions are presented in Sections 3 and 4, respectively. Finally, we draw our conclusion in Section 5.

2.1. BBSO and SOHO Observations

2.2. SDO Observations

2.3. STEREO and WIND Observations

2.4. GOES and RHESSI Observations

Figure 1 shows eight snapshots of the Hα images to illustrate the whole evolution of the filament (see also the online movie Animation1.mpg). Figure 1(a) displays the Hα image at 15:30:54 UT before eruption. It is overlaid with the contours of the LOS magnetic field, where green (blue) lines stand for positive (negative) polarities. The dark filament is ∼39 Mm long and resides along the polarity inversion line (PIL). The top panels of Figure 2 demonstrate the top-view of the 3D magnetic configuration above the AR at the beginning and after eruption, with the LOS magnetograms located at the bottom boundary. Using the same method described in Zhang et al. (2012c), we found a magnetic null point and the corresponding spine and separatrix surface. The normal magnetic field lines are denoted by green lines. The magnetic field lines around the outer and inner spine and the separatrix surface (or arcade) are represented by red and blue lines. Beneath the null point, the sheared arcades supporting the filament are represented by orange lines. The spine is rooted in the positive polarity (P1) that is surrounded by the negative polarities (N1 and PB). It extends in the northeast direction and connects the null point with a remote place on the solar surface. Such magnetic configuration is quite similar to those reported by Sun et al. (2012b), Jiang et al. (2013), and Mandrini et al. (2014).

Zoom In Zoom Out Reset image size

As time passed, the filament rose and expanded slowly (Figure 1(b)). The initiation process is clearly revealed by the AIA 304 Å observation (see the online movie Animation2.mpg). Figure 3 shows eight snapshots of the 304 Å images. Initial brigtenings (IB1, IB2, and IB3) appeared near the ends and center of the sigmoidal filament, implying that magnetic reconnection took place and the filament got unstable (Figures 3(b)–(d)). Such initial brightenings were evident in all the EUV wavelengths. With the intensities of the brightenings increasing, the dark filament rose and expanded slowly, squeezing the overlying arcade field lines. Null-point magnetic reconnection may have been triggered when the filament reached the initial height of the null point (∼15 Mm), leading to impulsive brightenings in Hα (Figures 1(c)–(d)) and EUV (Figures 3(e)–(h)) wavelengths, and increases in SXR and HXR fluxes (Figure 4). The M6.7 flare then entered the impulsive phase. The bright and compact flare kernel, indicated by the white arrow in Figure 1(c), extended first westward and then northward, forming a quasi-circular ribbon at ∼15:42 UT (Figure 1(d)), with the intensity contours of the HXR emissions at 12–25 keV superposed. There was only one HXR source associated with the flare, and the source was located along the flare ribbon with the strongest Hα emission, which is compatible with the fact that the footpoint HXR emissions come from the nonthermal bremsstrahlung of the accelerated high-energy electrons after penetrating into the chromosphere. The flare demonstrates itself not only around the filament, but also at the point-like brightening (hereafter PB) and the V-shape ribbon to the left of the quasi-circular ribbon. Since the separatrix surface intersects with the photosphere at PB to the north and the outer spine intersects with the photosphere to the east (Figure 2(a)), it is believed that nonthermal electrons accelerated by the null-point magnetic reconnection penetrated into the lower atmosphere not only at the quasi-circular ribbon, but also at PB and the V-shape ribbon.

Zoom In Zoom Out Reset image size

Figure 4 shows the SXR (black solid and dashed lines) and HXR (colored solid lines) light curves of the flare. The SXR fluxes started to rise rapidly at ∼15:32 UT and peaked at 15:45:53 UT for 1–8 Å and 15:44:21 UT for 0.5–4.0 Å. The HXR fluxes below 25 keV varied smoothly like the SXR fluxes, except for earlier peak times at ∼15:43:10 UT. The HXR fluxes above 25 keV, however, experienced two small peaks that imply a precursor release of magnetic energy and particle acceleration at ∼15:38:36 UT and ∼15:41:24 UT and a major peak at ∼15:43:10 UT. The time delay between the SXR and HXR peak times implies the possible Neupert effect for this event (Ning & Cao 2010). The main phase of the flare sustained until ∼17:00 UT, indicating that the flare is a long-duration event.

During the flare, the filament continued to rise and split into two branches at the eastern leg around 15:46 UT (Figure 1(e)), the right side of which is thicker and darker than the left. This process is most clearly revealed by the AIA 335 Å observation (see the online movie Animation3.mpg). Figure 5 displays eight snapshots of the 335 Å images. It is seen that the dark filament broadened from ∼15:42:30 UT and completely split into two branches around 15:45:51 UT. We define the left and right branches as the runaway part and major part of the filament. The two interwinding parts also underwent rotation (panels (d)–(h)). Meanwhile, the plasma of the runaway part moved in the northwest direction and escaped. To illustrate the rotation, we derived the time-slice diagrams of the two slices (S4 and S5 in panel (f)), which are plotted in Figure 6. The upper (lower) panels represent the diagrams of S4 (S5), and the left (right) panels represent the diagrams for 211 Å (335 Å). s = 0 in the diagrams stands for the southwest endpoints of the slices. The filament began to split into two parts around 15:42:30 UT, with the runaway part rotating round the eastern leg of the major part for ∼1 turn until ∼15:55 UT.

Zoom In Zoom Out Reset image size

During the eruption, the runaway branch of the filament disappeared (Figure 1(f)). The major part, however, fell back to the solar surface after reaching the maximum height around 15:51 UT, suggesting that the eruption of the major part of the filament failed. The remaining filament after the flare was evident in the Hα image (Figure 1(h)). NLFFF modeling shows that the magnetic topology was analogous to that before the flare, with the height of the null point slightly increased by 0.4 Mm (Figure 2(b)).

Figure 7 shows six snapshots of the 171 Å images. The rising and expanding filament triggered the M-class flare and the kink-mode oscillation of the adjacent large-scale coronal loops within the same AR (see the online movie Animation4.mpg). With the filament increasing in height, part of its material was ejected in the northwest direction represented by S1 in panel (c). After reaching the maximum height at ∼15:51:12 UT, the major part of the filament returned to the solar surface. The bright cusp-like post-flare loops (PFLs) in the main phase of the flare are clearly observed in all EUV filters, see also Figure 7(f).

To illustrate the eruption and loop oscillation more clearly, we extracted four slices. The first slice, S0 in Figures 1(f) and 7(d), is 170 Mm in length. It starts from the flare site and passes through the apex of the major part of the filament. The time-slice diagram of S0 in Hα is displayed in Figure 8(a). The filament started to rise rapidly at ∼15:34:30 UT with a constant speed of ∼117 km s⁻¹. After reaching the peak height (z_max) of ∼115 Mm at ∼15:51 UT, it fell back to the solar surface in a bumpy way until ∼16:30 UT. Using a linear fitting, we derived the average falling speed (∼22 km s⁻¹) of the filament in the Hα wavelength. The time-slice diagrams of S0 in the UV and EUV passbands are presented in Figure 9. We selected two relatively hot filters (335 and 211 Å in the top panels), two warm filters (171 and 304 Å in the middle panels), and two cool filters (1600 and 1700 Å in the bottom panels). Similar to the time-slice diagram in Hα (Figure 8(a)), the filament rose at apparent speeds of 92–151 km s⁻¹ before felling back in an oscillatory way at speeds of 34–46 km s⁻¹ during 15:51–16:10 UT and ∼71 km s⁻¹ during 16:18–16:30 UT (Figures 9(a)–(d)). The falling speeds in UV wavelengths are ∼78 km s⁻¹ during 15:51–16:10 UT (Figures 9(e)–(f)). The times when the major part of the filament reached maximum height in UV and EUV passbands, ∼15:51 UT, are consistent with that in Hα. The later falling phase during 16:18–16:30 UT is most obvious in the warm filters. The downflow of the surviving filament in an oscillatory way was also observed during a sympathetic filament eruption (Shen et al. 2012).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Owing to the lower time cadence of BBSO than AIA, the escaping process of the runaway part of the filament in Figures 1(e)–(f) was detected by AIA. We extracted another slice, S1, that is 177 Mm in length along the direction of ejection (Figure 7(c)). s = 0 Mm and s = 177 Mm represent the southeast and northwest endpoints of the slice. The time-slice diagram of S1 in 171 Å is displayed in Figure 8(b). Contrary to the major part, the runaway part of the filament escaped successfully from the corona at speeds of 125–255 km s⁻¹ without returning to the solar surface. The intermittent runaway process during 15:45–16:05 UT was obviously observed in most of the EUV filters. We extracted another slice, S2, which also starts from the flare site and passes through both parts of the filament (Figure 7(d)). The time-slice diagram of S2 in 171 Å is drawn in Figure 8(c). As expected, the diagram features the bifurcation of the filament as noted the white arrow (i.e., the runaway part escaped forever while the major part moved on after bifurcation and finally fell back).

The eruption of the filament triggered a transverse kink oscillation of the adjacent coronal loops (OL in Figure 7(a)). The direction of oscillation is perpendicular to the initial loop plane (see the online movie Animation4.mpg). We extracted another slice, S3, which is 80 Mm in length across the oscillating loops (Figure 7(b)). s = 0 Mm and s = 80 Mm represent the northwest and southeast endpoints of the slice. The time-slice diagram of S3 in 171 Å is shown in Figure 8(d), where the oscillation pattern during 15:38–15:47 UT is evidently demonstrated. The OL moved away from the flare site during 15:38–15:41 UT before returning to the initial position and oscillating back and forth for ∼2 cycles. By fitting the pattern with a sinusoidal function, as marked by the white dashed line, the resulting amplitude and period of the kink oscillation were ∼1.6 Mm and ∼225 s. We also extracted several slices across the OL and derived the time-slice diagrams, finding that the coronal loops oscillated in phase and the mode was fundamental. The initial velocity amplitude of the oscillation was ∼44.7 km s⁻¹. The speed of propagation of the mode is ${{C}_{K}}=2\;L/P=\sqrt{2/(1+{{\rho }_{o}}/{{\rho }_{i}})}{{v}_{A}}$, where L is the loop length, P is the period, v_A is the Alfvén wave speed, and ${{\rho }_{i}}$ and ${{\rho }_{o}}$ are the plasma densities inside and outside the loop (Nakariakov et al. 1999; Nakariakov & Ofman 2001; White & Verwichte 2012). In Figure 7(a), we denote the footpoints of the OL with black crosses that are 106.1 Mm away. Assuming a semi-circular shape, the length of the loop L = 166.7 Mm and ${{C}_{K}}=1482$ km s⁻¹. Using the same value of ${{\rho }_{o}}/{{\rho }_{i}}=0.1$, we derived ${{v}_{A}}=1100$ km s⁻¹. In addition, we estimated the electron number density of the OL to be ∼2.5 × 10¹⁰ cm⁻³ based on the results of NLFFF extrapolation in Figure 2(a). The kink-mode oscillation of the loops was best observed in 171 Å, indicating that the temperature of the loops was ∼0.8 MK.

The escaping part of the filament was also clearly observed by STA/EUVI. Figure 10 shows six snapshots of the 304 Å images, where the white arrows point to the escaping filament. During 15:46–16:30 UT, the material moved outward in the northeast direction without returning to the solar surface. The bright M6.7 flare noted by the black arrows is also quite clear.

Zoom In Zoom Out Reset image size

The runaway part of the filament resulted in a very faint CME observed by the WL coronagraphs. Figures 11(a)–(d) show the running-difference images of STA/COR1 during 16:00–16:15 UT. As noted by the arrows, the CME first appeared in the FOV of STA/COR1 at ∼16:00 UT and propagated outward at a nearly constant speed, with the contrast between CME and the background decreasing as time goes on. The propagation direction of the CME is consistent with that of the runaway filament in Figure 10. Figures 11(e) and (f) show the running-difference images of LASCO/C2 during 16:36–16:48 UT. The faint blob-like CME first appeared in the FOV of C2 at ∼16:36 UT and propagated in the same direction as that of the escaping filament observed by AIA in Figure 7(c). The central position angle and angular width of the CME observed by C2 are 311° and 37°, respectively. The linear velocity of the CME is ∼214 km s⁻¹. The time-height profiles of the runaway filament observed by STA/EUVI (boxes) and the corresponding CME observed by STA/COR1 (diamonds) and LASCO/C2 (stars) are displayed in Figure 12. The apparent propagating velocities represented by the slopes of the lines are 60, 358, and 214 km s⁻¹, respectively. Taking the projection effect into account, the start times of the filament eruption and the CME observed by LASCO/C2 and STA/COR1 from the lower corona ($\approx 1.0\;{{R}_{\codt }}$) are approximately coincident with each other. In the CDAW catalog, the preceding and succeeding CMEs occurred at 06:12 UT and 18:36 UT on 8 September. In the COR1 CME catalog, the preceding and succeeding CMEs occurred slightly earlier at 05:45 UT and 18:05 UT on the same day, which is due to the smaller FOV of COR1 than LASCO/C2. Therefore, the runaway part of the filament was uniquely associated with the CME during 16:00–18:00 UT.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Then, a question is raised: How can the runaway part of the filament successfully escape from the corona and give rise to a CME? We speculate that open magnetic field lines provide a channel. In order to justify this speculation, we turn to the large-scale magnetic field calculated by the potential-field source surface (PFSS; Schatten et al. 1969; Schrijver & De Rosa 2003) modeling and the radio dynamic spectra from the S/WAVES and WAVES instruments. In Figure 13, we show the magnetic field lines whose footpoints are located in AR 11283 at 12:04 UT before the onset of the flare/CME event. The open and closed field lines are represented by the purple and white lines. It is clear that open field lines do exist in the AR and their configuration corresponds to the directions of the escaping part of filament observed by AIA and the CME observed by C2. The radio dynamic spectra from S/WAVES and WAVES are displayed in panels (a)–(b) and (c)–(d) of Figure 14, respectively. There are clear signatures of type-III radio burst in the spectra. For STA, the burst started at ∼15:38:30 UT and ended at ∼16:00 UT, during which the frequency drifted rapidly from 16 to ∼0.3 MHz. For STB, which was ∼0.07 AU further than STA from the Sun, the burst started slightly later, ∼2 minutes, with the frequency drifting from ∼4.1 to ∼0.3 MHz because the early propagation of the filament was blocked by the Sun. For WAVES, the burst started at ∼15:39:30 UT and ended at ∼16:00 UT, with the frequency drifting from 13.8 to ∼0.03 MHz. The starting times of the radio burst were consistent with the HXR peak times of the flare. Because the type-III radio emissions result from the cyclotron maser instability of the nonthermal electron beams that are accelerated and ejected into the interplanetary space along the open magnetic field lines during the flare (Tang et al. 2013), the type-III radio burst observed by STEREO and WIND provides indirect and supplementary evidence that open magnetic field lines exist near the flare site.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

4.1. How is the Energy Accumulated?

4.2. How is the Eruption Triggered?

Once the free energy of the AR is accumulated to a critical value, chances are that the filament constrained by the overlying magnetic field lines undergoes an eruption. Several types of triggering mechanism have been proposed. Processes where magnetic reconnection is involved include the flux emergence model (Chen & Shibata 2000), catastrophic model (Lin & Forbes 2000), tether-cutting model (Moore et al. 2001; Chen et al. 2014), and breakout model (Antiochos et al. 1999), to name a few. Another type is the ideal MHD process, which is a result of KI (Kliem et al. 2004) and/or TI (Kliem & Török 2006). From Figure 15 and the movie (Animation5.mpg), we can see that before the flare there was continuous magnetic flux emergence (P2, P3, and P4) and subsequent magnetic cancellation along the fragmented PIL. We extracted a large region within the white dashed box of Figure 15(d) and calculated the total positive (${{{\Phi }}_{P}}$) and negative (${{{\Phi }}_{N}}$) magnetic fluxes within the box. In Figure 17, the temporal evolutions of the fluxes during 11:00–16:30 UT are plotted, with the evolution of ${{{\Phi }}_{P}}$ divided into five phases (I–V) separated by the dotted lines. The first four phases before the onset of the flare at 15:32 UT are characterized by quasi-periodic and small-amplitude magnetic flux emergence and cancellation, implying that the large-scale magnetic field was undergoing rearrangement before the flare. The intensity contours of the 304 Å images in Figures 3(b) and (d) are overlaid on the magnetograms in Figures 15(b) and (c), respectively. It is clear that the initial brightenings IB1 and IB2 are very close to the small positive polarities P4 and P3. There is no significant magnetic flux emergence around IB3. In the emerging-flux-induced-eruption model (Chen & Shibata 2000), when the reconnection-favorable magnetic bipole emerges from beneath the photosphere into the filament channel, it reconnects with the pre-existing magnetic field lines that compress the inverse-polarity MFR. The small-scale magnetic reconnection and flux cancellation serve as the precursor for the upcoming filament eruption and flare. During the flare, when magnetic reconnection occurred between 15:32 UT and 16:10 UT, both the positive and negative magnetic fields experienced impulsive and irreversible changes.

Zoom In Zoom Out Reset image size

Flux emergences are one plausible way to interpret the triggering mechanism, however there is another possibility. In the tether-cutting model (Moore et al. 2001), a pair of J-shape sheared arcades that comprise a sigmoid reconnect when the two elbows come into contact, forming a short loop and a long MFR. Whether the MFR experiences a failed or ejective eruption depends on the strength of compression from the large-scale background field. The initial brightenings (IB1, IB2, and IB3) around the sigmoidal filament might be the precursor brightenings as a result of internal tether-cutting reconnection due to the continuous shearing motion along the PIL. After onset, the whole flux system erupted and produced the M-class flare. Considering that the magnetic configuration could not be modeled during the flare, we are not sure whether a coherent MFR was formed after the initiation (Chen et al. 2014). Compared to the flux emergences, the internal tether-cutting seems more believable for interpreting how the filament eruption was triggered for the following reasons. First, the filament was supported by sheared arcade. Second, there were continuous shearing motions along the PIL, and the directions were favorable for the tether-cutting reconnection. Finally, the initial brightenings (IB1, IB2, and IB3) around the filament in Figure 3 fairly match the internal tether-cutting reconnection with the presence of multiple bright patches of flare emission in the chromosphere at the feet of reconnected field lines, whereas there was no flux emergence around IB3. NLFFF modeling shows that the twist number (∼1) of the sheared arcades supporting the filament is less than the threshold value (∼1.5), implying that the filament eruption may not be triggered by ideal KI. The photospheric magnetic field of the AR features a bipole (P1 and N1) and a couple of mini-polarities (e.g., P2, P3, P4, and N2). Therefore, the filament eruption could not be explained by the breakout model that requires quadrupolar magnetic field, although null-point magnetic reconnection took place above the filament during the eruption.

After the onset of the eruption, the filament split into two parts as described in Section 3. How the filament split is still unclear. In the previous literature, magnetic reconnection is involved in most cases (Gilbert et al. 2001; Gibson & Fan 2006a; Liu et al. 2008). In this study, however, the split occurred during the impulsive phase of the flare at the eastern leg, which was closer to the flare site than the western one, implying that the split was associated with the release of magnetic energy. The subsequent rotation or unwinding motion implies the release of magnetic helicity stored in the filament before the flare, presumably due to the shearing motion in the photosphere. Nevertheless, it is still elusive whether the filament existed as a whole or was composed of two interwinding parts before splitting. It seems difficult to explain the way of splitting by any previous model and requires an in-depth investigation.

Although the runaway part escaped out of the corona, the major part failed. It returned to the solar surface after reaching the apex. Such failed eruptions have been frequently observed and explained by the strapping effect of the overlying arcade (Ji et al. 2003; Guo et al. 2010a; Song et al. 2014; Joshi et al. 2014) or asymmetry of the background magnetic fields with respect to the location of the filament (Liu et al. 2009). In order to determine the cause of failed eruption of the major part, we turn to the large-scale magnetic configurations displayed in the bottom panels of Figure 2. It is shown that the overlying magnetic arcades above AR 11283 are asymmetric to a great extent (i.e., the magnetic field to the west of AR is much stronger than that to the east, which is similar to the case of Liu et al. 2009). According to the analysis of Liu et al. (2009), the confinements of the large-scale arcade acted on the filament are strong enough to prevent it from escaping. We also performed magnetic potential-field extrapolation using the same boundary, and derived the distributions of $|{\boldsymbol{B}} |$ above the PIL. We found that the maximum height of the major part considerably exceeds the critical height (∼80'') of TI where the decay index ($-d{\rm ln} |{\boldsymbol{B}} |/d{\rm ln} z$) of the background potential field reaches ∼1.5. If TI had worked, the major part would have successfully escaped from the corona after entering the instability domain. Therefore, the asymmetry with respect to the filament location, rather than TI of the overlying arcades, seems reasonable and convincing to explain why the major part of the filament underwent a failed eruption. In this study, both successful and failed eruptions occurred in a partially eruptive event, which provides more constraints to the theoretical models of solar eruptions.

4.3. How is the Coronal Loop Oscillation Triggered?

4.4. Significance for Space Weather Prediction

Using the multiwavelength observations from both spaceborne and ground-based telescopes, we studied in detail a partial filament eruption event in AR 11283 on 8 September 2011. The main results are summarized as follows:

• 1.  
  A magnetic null point was found above the pre-existing positive polarity surrounded by negative polarities in the AR. A spine passed through the null and intersected with the photosphere to the left. A weakly twisted sheared arcade supporting the filament was located under the null point, whose height increased slightly by ∼0.4 Mm after the eruption.
• 2.  
  The filament rose and expanded, which was probably triggered by the internal tether-cutting reconnection, or by continuous magnetic flux emergence and cancellation along the highly complex and fragmented PIL, the former of which seems more convincing. During its eruption, it triggered the null-point magnetic reconnection and the M6.7 flare with a single HXR source at different energy bands. The flare produced a quasi-circular ribbon and a V-shape ribbon where the outer spine intersected with the photosphere.
• 3.  
  During the expansion, the filament split into two parts at the eastern leg that was closer to the flare site. The major part of the filament rose at speeds of 90–150 km s⁻¹ before reaching the maximum apparent height of ∼115 Mm. Afterward, it returned to the solar surface staggeringly at speeds of 20–80 km s⁻¹. The rising and falling motions of the filament were clearly observed in the UV, EUV, and Hα wavelengths. The failed eruption of the major part was probably caused by the asymmetry of the overlying magnetic arcades with respect to the filament location.
• 4.  
  The runaway part, however, separated from and rotated around the major part for ∼1 turn before escaping outward from the corona at speeds of 125–255 km s⁻¹, probably along the large-scale open magnetic field lines, as evidenced by the PFSS modeling and the type-III radio burst. The ejected part of the filament led to a faint CME. The angular width and apparent speed of the CME in the FOV of C2 are 37° and 214 km s⁻¹. The propagation directions of the escaping filament observed by SDO/AIA and STA/EUVI are consistent with those of the CME observed by LASCO/C2 and STA/COR1, respectively.
• 5.  
  The partial filament eruption also triggered transverse oscillation of the neighboring coronal loops in the same AR. The amplitude and period of the kink-mode oscillation were 1.6 Mm and 225 s. We also performed diagnostics of the plasma density and temperature of the oscillating loops.

The authors thank the referee for valuable suggestions and comments to improve the quality of this article. We gratefully acknowledge Y. N. Su, P. F. Chen, J. Zhang, B. Kliem, R. Liu, S. Gibson, H. Gilbert, M. D. Ding, and H. N. Wang for inspiring and constructive discussions. SDO is a mission of NASA's Living With a Star Program. AIA and HMI data are courtesy of the NASA/SDO science teams. STEREO/SECCHI data are provided by a consortium of US, UK, Germany, Belgium, and France. QMZ is supported by Youth Fund of JiangSu BK20141043, by 973 program under grant 2011CB811402, and by NSFC 11303101, 11333009, 11173062, 11473071, and 11221063. H. Ji is supported by the Strategic Priority Research Program—The Emergence of Cosmological Structures of the Chinese Academy of Sciences, grant No. XDB09000000. Y.G. is supported by NSFC 11203014. Li Feng is supported by the NFSC grant 11473070, 11233008, and by grant BK2012889. Li Feng also thanks the Youth Innovation Promotion Association, CAS, for the financial support.