SLIPPING MAGNETIC RECONNECTION, CHROMOSPHERIC EVAPORATION, IMPLOSION, AND PRECURSORS IN THE 2014 SEPTEMBER 10 X1.6-CLASS SOLAR FLARE

The Active Region NOAA 12158 (hereafter, AR 12158) was visible on the solar disk during 2014 September 03–16. During this time, the AR 12158 produced several flares, including 12 C-class ones, an M4.6-class flare on 2014 September 7, and an X1.6-class long-duration flare on 2014 September 10. It is this X-class flare that is studied in this paper.

The X-class flare itself started at about 16:47 UT, reached a maximum of its X-ray flux at 17:45, as measured in the 1–8 Å passband by the GOES-15 satellite at Earth, and exhibited a long gradual phase (Figure 1, top). Several specific aspects of the flare have already been reported on by different authors (Cheng et al. 2015; Li & Zhang 2015; Tian et al. 2015). The flare evolution is detailed in Section 2.2.

Zoom In Zoom Out Reset image size

The preflare state of the AR 12158 is shown in Figure 1. The leading positive-polarity sunspot is encircled by a wreath of several smaller ones of both polarities, with the negative-polarity pores located to the S, SW, and W of the spot. The magnetic configuration of this active region is peculiar since the leading sunspot is of positive and not negative polarity, as would be expected of an active region in the northern solar hemisphere during cycle 24.

The AR 12158 contains two filaments, F1 and F2, as shown by the chromospheric Hα observations obtained at the Big Bear Solar Observatory (BBSO); see Figure 1. The filaments are of the same chirality (see, e.g., Martin 1998; DeVore et al. 2005). Both these filaments are overlaid by a sigmoid visible only in the 94 Å channel (Cheng et al. 2015) of the Atmospheric Imaging Assembly (AIA, Lemen et al. 2012; Boerner et al. 2012) on board the SDO (Pesnell et al. 2012). This suggests that the temperature of the sigmoid corresponds to the formation of Fe xviii (i.e., about 7 MK; O'Dwyer et al. 2010; Del Zanna 2013). The X1.6-class flare occurs within this sigmoid (Figure 2).

Zoom In Zoom Out Reset image size

In this section, we describe the evolution of the flare, as observed by the SDO/AIA imager. AIA acquires full-Sun images in 10 EUV and UV filters with high spatial resolution (15, 06 pixel size) and high temporal resolution (12 s). The bandpasses of the AIA filters are centered on several strong lines in the solar EUV/UV spectrum. These emission lines originate at different plasma temperatures, with some filter bandpasses containing more than one strong emission line (e.g., O'Dwyer et al. 2010; Del Zanna et al. 2011b). This means that the AIA temperature responses are multithermal in general (see also, e.g., Del Zanna 2013; Schmelz et al. 2013). That is, the signal observed within a particular AIA filter can originate at several different temperatures. Using combinations of AIA filters, however, it is possible to identify the approximate temperature of the emitting plasma, as well as perform the differential emission measure reconstruction (Schmelz et al. 2011a, 2011b; Warren et al. 2012; Hannah & Kontar 2012, 2013; Del Zanna 2013; Schmelz et al. 2013; Dudík et al. 2014b, 2015). This makes the AIA instrument an excellent tool to study the morphology of plasma emission at different temperatures.

As an example of the multithermality of the AIA filters relevant to the present study, we point out that the 131 Å bandpass has two dominant contributions, from Fe viii and Fe xxi, arising at about 0.5 and 10 MK, respectively, in equilibrium conditions (see, e.g., Petkaki et al. 2012; Del Zanna 2013; Dudík et al. 2014b). The Fe viii emission can be discerned visually by comparison with the 171 Å bandpass, dominated by Fe ix formed at around 0.8 MK. This is because of the significant overlap of the peaks of the relative ion abundances of Fe viii and Fe ix under equilibrium conditions (Bryans et al. 2009; Dere et al. 2009). Under nonequilibrium conditions characterized by the presence of κ distributions with high-energy power-law tails, as is the case in flaring plasma (Kašparová & Karlický 2009; Oka et al. 2013, 2015), the AIA temperature responses become more multithermal, and the peaks of the temperature responses are shifted toward higher temperatures (Dzifčáková et al. 2015).

A timeline of individual events during the flare is given in Table 1. An overview of AIA observations of the flare is given in Figure 2 and in the online Movies 1, 2, and 3, corresponding to the filters 131, 171, and 304 Å, respectively. The 131 and 171 Å bandpasses are chosen since they allow us to distinguish the 10 MK flare emission from the warm coronal loops. The 304 Å bandpass is shown to depict the evolution of the filaments F1 and F2, as well as that of the flare ribbons. The ribbons are denoted PR and NR for the positive-polarity and negative-polarity ribbons, while PRH and NRH stand for the respective hooks of these ribbons. Note that the presence of hooks is a signature of the presence of a flux rope, as also evidenced by the sigmoid (Aulanier et al. 2012; Janvier et al. 2013, 2014; Dudík et al. 2014b), and F1 and F2 (see also Cheng et al. 2015).

Table 1.  Summary of Individual Events during the 2014 September 10 Flare

Approx. Time [UT]	Event	Location [Solar X, Y]	Notes
16:47 UT	Onset of the early flare phase in the sigmoid	$[-100^{\prime\prime} ,130^{\prime\prime} ]$	Figures 1, 2
16:50 UT onwards	Growing system of flare loops, precursors	$[-100^{\prime\prime} ,130^{\prime\prime} ]$	Figures 2 top, Section 4.3
16:50–17:24 UT	Expanding warm coronal loops on the AR periphery	$[60^{\prime\prime} ,200^{\prime\prime} ]$	Sections 2.2.1 and 4.1, Figures 2–3, 11
16:51–17:05 UT	Failed F2 eruption	$[0^{\prime\prime} ,170^{\prime\prime} ]$, F2	Figure 2 second row; Figure 3
16:58–17:26 UT	NR hook extension, squirming and slipping motions	$[-150^{\prime\prime} ,100^{\prime\prime} ]$, NRH	Section 2.3.1, Figures 2, 4– 6
17:00–17:30 UT	Slipping motion well visible in PR	$[-50^{\prime\prime} ,140^{\prime\prime} ]$, PR	Section 2.3.2, Figures 2, 7, and 8
17:03–17:14 UT	Weak blueshifts detected in Fe xxi	NR	Section 3.2, Figures 9–10
17:10–17:27 UT	Growing system of S-shaped loops, hot eruption	$[0^{\prime\prime} ,170^{\prime\prime} ]$	Section 2.2.2, Figure 2
17:24 UT	Impulsive phase onset, strong blueshifts in Fe xxi	NR	Figures 2, 9, and 10
17:24 UT	Onset of fast eruption (velocities $\gt \;270$ km s−1)		Figure 3
17:25–17:40 UT	Coronal loop oscillations following the hot eruption	$[60^{\prime\prime} ,200^{\prime\prime} ]$	Section 2.2.3, Figures 2–3
17:28, 17:32 UT	Peaks of the Callisto radio flux at 350 MHz		Figure 3 bottom, Section 2.2.3
17:45 UT	Peak of the GOES 1–8 Å X-ray flux, onset of gradual phase		Figure 1 top left
17:58 UT	End of IRIS sit-and-stare observations, Fe xxi nearly at rest	NR	Figure 9

Note. Times and locations given are approximate because the events are dynamic or have a spatial extension or temporal duration.

Download table as:  ASCIITypeset image

2.2.1. Early Flare Phase

The X1.6-class flare starts with a loop-like brightening within the sigmoid above F1. The brightening is observable in the 131 and 94 Å bandpasses and develops into a series of flare loops. This behavior is very similar to the one described for another X-class flare by Dudík et al. (2014b). At 16:50 UT, these loops are clearly visible (Figure 2, top left). At this time, F2 starts to rise and brighten in both 131 and 171 Å, indicating that it is heated to at least several times 10⁵ K. The F2 subsequently decelerates as it is stopped by the overlying coronal loops seen in AIA 171 and 193 Å (see Figure 2 at 17:00 UT and online Movie 2). To study the evolution of F2 and the neighboring coronal loops, we place an artificial cut across a direction of F2 rise (see Figure 2). The time–distance plots obtained along the direction of this cut are shown in Figure 3. Using these time–distance plots, we measure the deceleration of F2 to be about −14 m s⁻² by approximating a parabola to the profile on the time–distance plot (dashed black line in Figure 3). The stopping of F2 is denoted by the black horizontal dotted line in the 171 Å time–distance plot.

Zoom In Zoom Out Reset image size

The time–distance plots constructed along the artificial cut also reveal the rich dynamics of the overlying, warm coronal loops observed in the 171 Å bandpass as well as in 193 Å (not shown). These loops are highly likely to have been inclined with respect to the local vertical; note also the similar pattern of fibrils observed in Hα and AIA 304 Å (Figures 1, bottom left and 2, right). One of the coronal loops observed in 171 Å is contracting with a speed of about −2.9 ± 0.9 km s⁻¹. It is located at the approximate position of 100'' along the artificial cut and denoted by the white dotted line in the respective 171 Å time–distance plot (Figure 3). The contraction is discernible even before the onset of the F2 rise, an important fact discussed in terms of the coronal implosion mechanism in Section 2.2.3.

The rise of F2 is also accompanied by widespread dynamics of the coronal loops. A series of these loops start to rise at various times with different velocities, ranging from 5.8 ± 1.7 to 21.4 ± 2.1 km s⁻¹ (see Figure 3). The velocities are determined by calculating the slope of the respective line on a time–distance plot. The location of the line is determined by a trial-and-error method repeated many times. The uncertainty is then calculated by error propagation from the uncertainties of the line end points. We consider that the uncertainty in position is equal to half of the AIA resolution (075), while the uncertainty in time is half of the AIA cadence (0.6 s).

The rise of the warm coronal loops in our flare continues until the eruption and the associated disturbance during the impulsive phase. We, however, note that the occurrence of a contracting loop before the presence of the rising structures in our flare is contrary to the behavior reported for five different flares by Liu et al. (2012b).

During the rise of F2, the system of flare loops continues to widen and brighten (Figure 2), while individual flare loops exhibit an apparent slipping motion in both of their footpoints. The slipping motion is clearly visible in the online Movie 1 and is analyzed in Section 2.3.

2.2.2. Eruption of the Flux Rope

2.2.3. Apparent Implosion and Oscillations of the Warm Coronal Loops

The occurrence of the apparently slipping flare loops in this flare has already been reported by Li & Zhang (2015), who focused on the NR during the impulsive phase (from about 17:25 onward) in the vicinity of the IRIS slit. Upon reviewing the evolution of the AIA 131 Å observations, we found that the slipping reconnection is present during the entire flare. It is noticeable from the very beginning at about 16:50 UT through the impulsive phase, similar to that found by Dudík et al. (2014b). The following gradual phase is characterized by the strong-to-weak shear transition, which can also be explained by the standard solar flare model in 3D (see Aulanier et al. 2012). We also note that Zhao et al. (2016) calculated the photospheric traces of the QSLs in an extrapolated nonlinear force-free field (Gilchrist & Wheatland 2014) and found that these correspond well with the observed shape of the flare ribbons, thus indirectly supporting the idea of slipping reconnection occurring in QSLs.

Unlike the flare reported in Dudík et al. (2014b), however, we note that the early flare phase analyzed here exhibits several instances of apparent countermotions of slipping flare loops. That is, the sytem of flare loops exhibits apparent slippage in both directions toward both ends of the developing ribbons. These apparent countermotions are not easy to track, however, mainly due to the complicated evolution of the ribbons. For example, the NR exhibits local squirming motions, during which the slipping motion is seen to proceed even in an almost transverse direction with respect to the general direction of the ribbon extension during the next few minutes (see the online Movies 1 and 4). A similar but less pronounced evolution happens in the PR. Nevertheless, the apparent slipping motions of flare loops can be discerned during particular time intervals. Two such cases are reported on in the remainder of this section.

2.3.1. Slipping Reconnection along the NR

An example of the squirming nature of the evolution of the NR and the associated slipping reconnection can be seen in the online Movie 4 and the corresponding Figure 4, where the time interval of 17:08–17:17 UT is shown. Although the NR generally extends in the SW direction, which is the same direction as reported during the impulsive phase by Li & Zhang (2015), during the time interval shown in Figure 4, the slipping motion occurs predominantly in the north–south direction. To analyze these slipping motions, we place an artifical cut at Solar X = −150'' (see Figure 4, top left). This cut is used to produce the time–distance plots in AIA 131 and 304 Å shown in Figure 5. We chose these two AIA filters since the 131 Å shows the slipping loops emitting Fe xxi, while the corresponding footpoints are very bright in the 304 Å, making it a useful bandpass to study the evolution of the ribbon itself.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

At about 17:03 UT, the footpoints of the apparently slipping loops first reach the location of the cut. The dominant slipping motion is in the southern direction, with a speed of about −38 km s⁻¹. This velocity corresponds to the left-most dashed line in Figure 5. Several other slipping loops enter the cut later on. Some of them are highlighted by the dashed and dotted lines in Figure 5. After 17:09 UT, a prominent extension of the ribbon in the opposite (northern) direction occurs at the position of Solar Y = 110''. Following this time, a series of loops is seen to be apparently slipping in both directions along the cut. In Figure 5, the apparently slipping loops are denoted by dotted lines, while the apparently slipping bright knots in the ribbon are denoted by dashed lines. The distinction between the two is easily made by their presence in both the 131 Å and 304 Å time–distance plots. This is because the 304 Å passband shows only the flare loop footpoints, with higher portions of these flare loops being visible in 131 Å and not in 304 Å. The typical apparent slipping velocities found are 11–57 km s⁻¹, similar to that in Dudík et al. (2014b).

We next examined the relationship between the morphology of flare emission in the hot AIA bandpasses (131 and 94 Å) and the bandpasses registering the transition-region emission (304, 1600 Å, as well as 1700 Å). We note that the 304 Å bandpass is nearly cotemporal with the 131 Å, with only a 1 s difference. We find nearly a one-to-one correspondence (Figure 4) between the locations of the footpoints of the 131 Å loops and the 304 Å bright kernels within the evolving ribbon. Several conspicuous examples are pointed out by Arrows 1–5 in Figure 4. This relationship was already reported by Dudík et al. (2014b) and is confirmed here.

The nearest 1600 Å or 1700 Å image is usually taken at least several seconds earlier or later than the 131 Å (Figure 6). The close relationship between the 131 Å footpoints and the 304 Å bright kernels also holds for the kernels observed in AIA 1600 Å. An example is shown in Figure 6. This figure contains the AIA 1600 and 1700 Å observations, as well as their ratio, for the five last times shown in Figure 4. We see that the brightest kernels in 304 Å are also present in 1600 Å, with faint brightenings being present in the 1700 Å bandpass as well. The ribbon is, however, best seen in the 1600 Å/1700 Å ratio, the morphology of which appears similar to the 304 Å (compare Figures 4 and 6), although we note that the difference between acquiring the 1600 and 304 Å images is several seconds.

Zoom In Zoom Out Reset image size

We note that both the 1600 Å and 1700 Å bandpasses have a broad spectral response (see Figure 9 in Boerner et al. 2012). These bandpasses contain a strong contribution from the photospheric continuum being formed near the temperature-minimum region. The 1600 Å bandpass, however, also contains the prominent C iv 1548.19 and 1550.77 Å doublet. By taking the AIA 1600 Å/1700 Å ratio, we find that the locations of the bright flare kernels show strongly enhanced 1600 Å emission, up to a factor of $\approx 3.3$ compared to the average of a nearby plage and $\approx 3.8$ compared to a quiet Sun region containing both network and internetwork. This result points to a strong C iv emission being present in the flare kernels. A strong increase in transition-region line intensities is commonly observed in a solar flare (e.g., Cheng et al. 1981; Poland et al. 1982; Woodgate et al. 1983; Schmieder et al. 1996b; Del Zanna & Woods 2013), often associated with hard X-ray bursts. Note that a strong increase of C iv emission is expected for non-Maxwellian distributions (Dzifčáková et al. 2005; Dzifčáková & Karlický 2008), but without direct RHESSI observations, or modeling of the optically thick photospheric continuum, the presence of these non-Maxwellians cannot be unambiguously confirmed from the AIA 1600 Å observations, though high-energy tails are routinely observed in flares (e.g., Veronig et al. 2010; Fletcher et al. 2011; Battaglia & Kontar 2013; Oka et al. 2013, 2015; Simões & Kontar 2013; Milligan et al. 2014; Simões et al. 2015).

2.3.2. Slipping Reconnection along the PR

The slipping reconnection, including the apparent motion of flare loops in both directions, is best visible in the vicinity of the PR, located near the leading positive-polarity sunspot. An example of the evolution of the flare loops is given in the online Movie 5 and the corresponding Figure 7, where the interval of 17:16–17:20 UT is shown. The first part of this image shows the AIA 131 Å, while the second part shows the corresponding AIA 304 Å images. It can be seen that the one-to-one relationship between the footpoints of the flare loops as seen in AIA 131 Å and the bright kernels in 304 Å is confirmed for the PR as well.

Zoom In Zoom Out Reset image size

To identify the apparently slipping loops in both directions along the developing ribbon, we produced a time–distance plot along a curvilinear cut shown in Figure 7. This cut has the shape of an ellipse, centered on the location X = −635, Y = 152''. The elliptical shape of the cut reflects approximately the shape of the ribbon PR and was determined as the best fit to the manually placed knot points using a trial-and-error method. Using a curvilinear cut is necessary since a straight cut would only allow a measurement of one velocity component along the ribbon and could lead to significant underestimates depending on the position along a straight cut.

The time–distance plot along this cut is shown in Figure 8. The slipping motion spans nearly the entire preflare phase. Several well-defined slipping loops are denoted by the dotted lines. At several times, loops are observed to slip in both directions along the cut. We note especially that, in the vicinity of the position 60'' along the cut, a series of loops is observed to slip apparently in successively changing directions, creating a "crisscross" pattern. The apparent velocities of these loops are approximately 30 km s⁻¹. In particular, a pair of flare loops slipping in opposite directions is observable at about 17:17–17:19 UT and is denoted by white and orange arrows in Figure 7. These loops exhibit a converging motion until about 17:19 UT, after which time they are no longer visible. The corresponding velocities along the curvilinear cut (see Figure 8) are −37.5 ± 10.9 km s⁻¹ and 31.2 ± 5.0 km s⁻¹ for the loops denoted by the white and orange arrows, respectively. Another loop, denoted by a green arrow, is seen to be slipping from about 17:18 UT onward. Its velocity along the curvilinear cut, corresponding to the green dotted line in Figure 8, is 19.0 ± 3.7 km s⁻¹.

Zoom In Zoom Out Reset image size

Generally, the slipping velocities measured using the time–distance technique during the early flare phase (16:50–17:30 UT) are in the range of 18–44 km s⁻¹. This is consistent with the apparent slipping velocities determined for the NR in Section 2.3.1. We, however, note that the apparent slipping velocities measured here are only lower limits because of the changing angle of the loops with respect to the cut, which is due to the evolution of the ribbon itself.

Since its launch in 2013, IRIS (De Pontieu et al. 2014) has provided simultaneous imaging and spectroscopy of the solar atmosphere with unprecedented spatial resolution (033–04), cadence (up to 2 s), and spectral accuracy (allowing measurements of Doppler shifts of $\approx 3$ km s⁻¹). The IRIS Slit Jaw Imager (SJI) acquires high-resolution images in four different passbands (C ii 1330 Å, Si iv 1400 Å, Mg ii k 2796 Å, and Mg ii wing 2830 Å), allowing the study of plasma dynamics in great detail. Simultaneously, the IRIS spectrograph (SP) observes several emission lines formed over a broad range of temperatures ($\mathrm{log}(T)[K]\;=\;3.7-7$). Of particular interest is the Fe xxi 1354.08 Å line formed at ≈11 MK, which represents the only flare emission observed by IRIS. This spectral line was first identified by Doschek et al. (1975) in solar flare spectra obtained with the Naval Research Laboratory's S082B spectrometer on board Skylab.

The spatial and spectral characteristics of the Fe xxi line allow us to investigate the plasma response to the heating during flares and provide new insights into the chromospheric evaporation process. Spatially resolved, blueshifted asymmetric Fe xxi profiles indicative of plasma upflows ≈200 km s⁻¹ were first observed by Mason et al. (1986) during the impulsive phase of different flares with the UVSP instrument on board the Solar Maximum Mission. In contrast, 1D hydrodynamics simulations of a single flare loop (Emslie 1978) predict that entirely blueshifted profiles should be observed at the onset of the flare. However, these early observations lacked good spatial information, and the discrepancy can be explained by interpreting the asymmetric profiles as a superposition of different plasma upflows from different subresolution locations along the line of sight.

Recently, there has been a lot of interest in studying the Fe xxi emission during flares as observed with the unprecedented resolution of IRIS (e.g., Graham & Cauzzi 2015; Polito et al. 2015, 2016; Tian et al. 2015; Young et al. 2015). One of the important findings from these recent studies is that Fe xxi is observed to be entirely blueshifted during the impulsive phase of the flare, suggesting that the sites of evaporation are now likely to be resolved by IRIS.

On 2014 September 10, IRIS was running a flare watch observation of AR 12158 from 11:28 UT to 17:58 UT, so it captured the flare from onset well into the gradual phase. The observing mode was sit-and-stare with an exposure time of 8 s and a total cadence of 9.4 s for the FUV channel. The slit of the IRIS spectrograph crossed two locations along the ribbon NR during all of the impulsive and part of the gradual phase of the flare (see Figure 2). The SJI obtained 19 s cadence images in the 1400 Å and 2796 Å passbands over a field of view of 119'' × 119'' on the Sun. For each spectrograph exposure, one context SJI image was provided alternatively in one of the two filters. In this work, we focus on the O i and Si iv spectral windows included in the spectrograph FUVS and FUVL channels, respectively. We used IRIS level 2 data, obtained from level 0 after flat fielding, geometry calibration, and dark-current subtraction. The cosmic-ray removal was performed by using the SolarSoft routine despik.pro.

As a result of the temperature variation of the detectors during the IRIS satellite orbit, the FUV channel wavelength scale drifts by about 8 km s⁻¹ during one orbit. The orbital variation was corrected by using the strong O i 1355.60 Å neutral line included in the FUVS CCD. The Doppler shift of this line is often less than 1 km s⁻¹ and thus represents a suitable reference line for wavelength-calibration purposes. We measured the periodic variation of the O i line position over time and subtracted it from both FUVS and FUVL wavelength arrays, assuming that the same wavelength correction can be applied for different FUV CCD channels (IRIS TN20⁶ ). The absolute wavelength calibration can then be obtained by using the strongest photospheric lines in the same spectral range. For the FUVS channel, the absolute correction was given by the difference between the O i line position (after the orbital correction) and the expected rest wavelength 1355.5977 Å (Sandlin et al. 1986). For the FUVL CCD, the S i 1401.51 Å neutral line can be used. Even though this line is usually very weak, during this event it was visible at the ribbon location throughout the impulsive phase.

An overview of the flare evolution as observed by IRIS is shown in Figure 9. Section 3.2 provides a detailed description of the flaring plasma dynamics as observed in the Fe xxi emission. The O iv and Si iv lines are reported on in Section 3.3.

Zoom In Zoom Out Reset image size

The evolution of the flare as observed by IRIS is shown in Figure 9. In this figure, each row captures a particular time, showing (from left to right) the AIA 131 Å with the closest IRIS SJI 1400 Å images, as well as the corresponding O i and Si iv detector images. The IRIS SJI 1400 Å band is dominated by Si iv 1402.77 Å emission at log(T/K) ≈ 4.9 K. Therefore, the sequence of SJI images in Figure 9 shows the morphology of the low-temperature ribbon emission over time. The corresponding Fe xxi spectra are reported in Figure 10, as observed by the IRIS spectrograph slits. The line centroid and width given by the Gaussian fit are plotted in each spectrum.

Zoom In Zoom Out Reset image size

The expected width of the Fe xxi line observed by IRIS is ≈0.43 Å, given by the quadratic sum of the IRIS instrumental FWHM (0.026 Å; De Pontieu et al. 2014) and the line thermal width as estimated in CHIANTI v7.1 (Dere et al. 1997; Landi et al. 2013), assuming an ion formation temperature of 11 MK. However, the line width is typically observed to be significantly larger during the impulsive phase of flares (Mason et al. 1986; Polito et al. 2015). We estimate the nonthermal motions as a velocity parameter W_nth given by $\sqrt{{(4\mathrm{ln}2)}^{-1}{(\lambda /c)}^{-2}\cdot ({W}^{2}-{W}_{\mathrm{th}}^{2}-{W}_{{\rm{I}}}^{2})}$, where W is the line FWHM obtained from the fit, W_th is the line thermal width, W_I is the instrumental width, λ is the Fe xxi rest wavelength at 1354.08 Å, and c is the speed of light.

We note that the IRIS O i spectral window includes some cool temperature lines whose emission is usually enhanced during flares and can blend with the Fe xxi. Several authors, namely Young et al. (2015), Polito et al. (2015), Tian et al. (2015), and Graham & Cauzzi (2015), reported a detailed identification of these lines during different flare events. The most important blending is represented by the chromospheric C i line at 1534.3 Å (Mason et al. 1986). However, the profiles of these low-temperature lines are typically narrow, and in most of the cases they can be easily separated from the broad Fe xxi emission.

The first row of Figure 9 shows the flare plasma at around 17:00 UT, in the early flare phase. At this time, we observe flare loops connecting the two flare ribbons in the AIA 131 Å channel. These are not visible in the 171 Å AIA passband (Figure 2) and therefore are likely to be hot loops originating from the Fe xxi 128.75 Å emission contributing to the 131 Å band. We note that, at this time, IRIS does not detect any Fe xxi 1354.10 Å emission. When there is a flare, the Fe xxi 128.75 Å line dominates the 131 Å band (O'Dwyer et al. 2010; Petkaki et al. 2012), and the corresponding count rates detected in the IRIS Fe xxi 1354.07 Å line are about 100 times less than the count rates detected in the 131 Å band. This is despite the fact that the forbidden Fe xxi 1354.07 Å line emits 1.2 more photons than the resonance 128.75 Å line. For example, at 17:03 UT, the peak counts in the AIA 131 Å band are about 1100 DN s⁻¹, which are equivalent to about 12 DN s⁻¹ in the IRIS Fe xxi line. This estimate was obtained from a full Differential Emission Measure (DEM) analysis using the six AIA bands via the Hannah & Kontar (2013) method and the current understanding of the in-flight degradation of the IRIS channels. The IRIS study had an exposure time of 8 s, so 100 data numbers (DN) in the line correspond to about 5 DN in the peak value above the continuum, close to the limit of the line being observable, given that the line is normally very broad and blended (see Figure 10).

The Fe xxi line is first observed by IRIS only from ≈17:03 UT (second row of Figure 9) onward, during which time the intensity of the flare plasma in the 131 Å AIA band becomes more intense. At the location of the IRIS slit, the Fe xxi emission originates in the upper portions of the hot flare loops visible in the AIA image because the slit cuts these flare loops near their center. A spectrum of the IRIS Fe xxi line at ≈ 17:03:12 UT is shown in the first panel of Figure 10. The line profile has been obtained by averaging over the slit pixels between the horizontal blue lines indicated in Figure 9. At 17:03 UT, the intensity of the Fe xxi emission is very weak, around 10 DN. The line profile appears almost at rest (10 km s⁻¹) and is slightly broadened with a width of 0.56 Å, corresponding to a nonthermal width of around 48 km s⁻¹.

The red contours overlaid on the AIA and SJI images represent the intensity contours (70% and 90% of the maximum value) of the 6–12 keV sources observed by RHESSI (Lin et al. 2002) during 17:02:48–17:03:00 UT. These sources coincide with the footpoints of the hot loops rooted in the two flare ribbons, as expected from the thick-target flare heating model (Brown 1971). Unfortunately, we are not able to identify any X-ray sources at 17:15 UT onward because of the spacecraft night.

At 17:14 UT, we observe blueshifted (≈90 km s⁻¹) Fe xxi emission in the IRIS O i detector image, as shown in the third panel of Figure 9. At this time, the IRIS slit was crossing a bright portion of the ribbon NR visible in both the SJI 1400 Å filter and the AIA 131 Å bandpass. This portion of the ribbon corresponds to the footpoints of a series of hot flare loops connecting PR and NR. That is, the IRIS Fe xxi emission is now formed at a ribbon rather than at the loop top, meaning that the local magnetic field along which the Fe xxi emission originates is more vertical than at 17:03 UT. The corresponding Fe xxi spectrum, shown in the second panel of Figure 10, is indeed more strongly blueshifted. We note that the observation of the Fe xxi emission during the early phase of the flare (see Table 1) has not been reported in the previous studies of the same flare event by Tian et al. (2015) and Graham & Cauzzi (2015), which were concerned with the analysis of the strong blueshifts observed from around 17:25 UT onward.

From about 17:25 UT, that is, during the start of the impulsive phase characterized by the onset of the fast eruption (Figure 3), the IRIS slit crosses the NR ribbon at two different locations. Strong Fe xxi emission is detected only at the northern crossing with the NR ribbon, located at around Solar Y ≈ 120''. The line profile is plotted in the third panel of Figure 10. The Gaussian fit shows a strong blueshifted centroid position of about 200 km s⁻¹ and a very broad line width (≈0.81 Å or a nonthermal width of ≈90 km s⁻¹). A few minutes later, at about 17:31 UT, we observe a line profile that is even more blueshifted (≈270 km s⁻¹) and extends outside the O i spectral window, as reported by Tian et al. (2015) and Graham & Cauzzi (2015).

In summary, the Fe xxi line is at first weakly blueshifted and a few minutes later becomes more shifted as the ribbon NR enters the position of the IRIS slit. We interpret this as the consequence of the slipping reconnection along the NR that is itself developing. We note that although the 1D hydrodynamical simulations of a flare loop predict a rapidly increasing blueshift before the maximum blueshift velocity is reached (Fisher et al. 1985; Polito et al. 2016), this is strictly true only for one ribbon kernel. In the observations, this effect is compounded by the observed slipping motion of the flare loop footpoints, which are responsible for the changing inclination of the local magnetic field at the position of the IRIS slit. Therefore, the evaporating plasma would show different Doppler shifts along the line of sight. Unfortunately, in the present study we cannot reliably distinguish between the geometric effects and the evolution of the evaporation velocities itself. This is due to the lack of stereoscopic spectral observations.

In addition, we note that the Fe xxi blueshifted emission is located a few pixels just above the ribbon position where the intensity of the FUV continuum and the cooler emission lines is enhanced, as already observed by Young et al. (2015) in the study of another X-class flare. This can be best seen in the third and fourth panels of Figure 9, showing that the blue lines (indicating the location of the Fe xxi blueshifts) are clearly above the position where transition-region lines included in the Si iv window are more intense. We note, however, that such noncospatiality of only several IRIS pixels would not have been resolved by AIA, whose spatial resolution is 15. Therefore, the conclusion reached in Section 2.3 that the AIA ribbon as seen in 304 Å is cospatial with the footpoints of the 131 Å loops remains valid within the AIA resolution.

Later on during the flare, the Fe xxi line profile gradually moves to the rest position as the flare loops are being filled by the high-temperature evaporating plasma. The fourth panel in Figure 10 shows the Fe xxi spectrum being almost at rest, as observed by the IRIS slit at around 17:48 UT, after the onset of the gradual phase.

The O iv 1401.16 Å, O iv 1399.77 Å, and Si iv 1402.77 Å transition lines observed by IRIS provide various diagnostic opportunities but are affected by several complexities in their interpretation. The response of these transition-region lines in the ribbons during a flare is well known to be strongly dependent on nonequilibrium ionization effects (Bradshaw et al. 2004; Doyle et al. 2013; Olluri et al. 2013), as well as on the non-Maxwellian electron distributions (Dudík et al. 2014a). These lines can also be blended with unidentified photospheric or chromospheric transitions (Polito et al. 2016) during the impulsive phase.

The O iv emission is usually very low in active-region spectra but can be enhanced during the impulsive phase of flares. Here, we have estimated the ratio between the O iv 1401.16 Å and Si iv 1402.77 Å lines at particular times where the O iv emission was high enough to be reliably measured. The Si iv line is often saturated during the impulsive phase, and thus only an upper limit of the ratio can be obtained. We have integrated the line intensity after background subtraction and calibrated the values in physical units (erg s⁻¹ sr⁻¹ cm⁻² Å⁻¹). For instance, at 17:30:17 UT we find a ratio of O iv 1401.16 Å and Si iv 1402.77 Å equal to 0.03 at the slipping footpoint of the NR crossed by the IRIS slit at that time (see Movie 6). Similar values are found throughout the impulsive phase.

In addition, the ratio of the O iv 1401.16 Å and 1399.77 Å lines from IRIS is sensitive to the electron density of the plasma. Throughout the impulsive phase of the flare, we measure a O iv Å 1401.16/1399.77 Å ratio that is below the high-density limit of 2.5 reported by CHIANTI v7.1, assuming equilibrium conditions. In particular, the ratio is equal to 2.19 at 17:30:07 UT, which would indicate a density of at least 10¹² cm⁻³ in equilibrium. These line ratios are also consistent with lower densities and non-Maxwellian electron distributions in the flare plasma (Dudík et al. 2014a).

We first examined the standard solar flare model in 3D for signatures of loop expansion and contraction. Although this model is generic and its photospheric flux distribution does not represent the active region under study here, we found that both processes are present. In Figure 11, the flux rope is depicted by pink, green, and cyan field lines. The pink field lines represent the flux rope core, while the green and cyan field lines represent the S-shaped envelope that is created as a result of the slipping reconnection (Aulanier et al. 2012; Janvier et al. 2013; Dudík et al. 2014b). The loops overlying the flux rope are shown in red and white. The white loops are highly inclined, while the red ones are nearly vertical. Both of these loop systems are anchored in the leading positive-polarity spot in the model (see also Aulanier et al. 2012).

Zoom In Zoom Out Reset image size

During the course of the eruption, the unstable flux rope pushes the red overlying loops, causing them to expand and move sideways. This behavior persists for several tens of Alfvén times. Although the simulation is dimensionless (Aulanier et al. 2012; Janvier et al. 2013), taking an order-of-magnitude estimate of the Alfvén speed v_A ≈ 10³ km s⁻¹ and a loop length of 100 Mm, the duration of the expansion would be about 10³ s, in broad agreement with the observed duration of the loop expansion during the slow rise of the flux rope. In contrast to the red field lines, the highly inclined white field lines undergo a contraction. This is probably because the stretching of the legs of the flux rope during its eruption leads to a local decrease of magnetic pressure in the legs of the flux rope, leading to a contraction of the neighboring loops.

The expanding and contracting behavior that we have observed is therefore consistent with the predictions of the standard solar flare model in 3D. This also confirms our earlier conclusion (Section 2.2.3) that the coronal implosion does not occur during this flare.

Finally, the observed oscillatory behavior of the field lines (Section 2.2.3) after the fast eruption is not reflected in the simulation. The reason for this is the relative shortness of the calculation, which does not allow the flux rope to leave far enough for the overlying loops to have time to reach and bounce back from the central part of the AR. Additionally, the viscosity in the model may be too large in the locations of the loops, where the computational mesh is stretched.

The standard solar flare model in 3D also predicts that the apparent slipping velocities of the magnetic field lines v_slip and the outward velocity of the conjugate ribbon v_QSL, i.e., the speed of the ribbon movement perpendicular to the polarity inversion line, are related. The relation is linear, with a proportionality constant given to a first-order approximation by N, the local norm of the field line mapping (Janvier et al. 2013):

$$\begin{eqnarray}&&{v}_{\mathrm{slip}}\;=\;{{Nv}}_{\mathrm{QSL}}.\end{eqnarray} \tag{ 1 }$$

In principle, this equation permits measurement of N if v_slip and v_QSL are known, at least within a timescale where the QSLs are not evolving as much. In this work, we have measured the v_slip during the early and impulsive flare phase and found that this velocity is typically ≈20–40 km s⁻¹ in both ribbons (Section 2.3).

To estimate the v_QSL, we have first determined the evolution of the Solar Y position of the ribbon NR at the location of the IRIS slit (Figure 12). This is done during the time interval 17:25–17:40 UT, during the impulsive phase, when the ribbon is well defined. The locations of both the northern and southern branches of the NR are determined as the locations of maximum intensity of the FUV continuum as observed in the FUVS O i and FUVL Si iv spectral windows. Using the IRIS spectra at a given sit-and-stare slit position has the advantage of a very high cadence of about 9 s compared to the IRIS SJI or AIA images. The ribbon displacements determined using this method are not subject to confusion between the motion of the ribbon itself and the slipping motion, which would need to be separated if the ribbon position were determined from imaging data. The position of both branches of the NR ribbon shows a linear trend during the impulsive phase of the flare, as shown in Figure 12. The projected velocity ${v}_{\mathrm{QSL},\mathrm{proj}}$ along the IRIS slit is about 19 km s⁻¹ for the northern branch of the NR and has a similar value for the southern branch. The outward ribbon velocity is then obtained as v_QSL = ${v}_{\mathrm{QSL},\mathrm{proj}}$/sin $(\alpha )$, where α ≈ $41^\circ$ is the approximate angle between the northern branch of the NR and the IRIS slit at 17:30 UT (Figure 9). This yields v_QSL ≈ 29 km s⁻¹.

Zoom In Zoom Out Reset image size

Using these values, from Equation (1) we obtain N ≈ 0.7–1.4, which is very low for a QSL. Even using the highest reported value of v_slip = 200 km s⁻¹ for this flare (Tian et al. 2015) would yield only N ≈ 6.9. This poses a significant problem since the typically expected value for a QSL would be at least several tens or hundreds, due to the high squashing of the magnetic flux tubes in the QSL (Titov et al. 2002). The presence of a QSL and the associated current density enhancement (Masson et al. 2009; Wilmot-Smith et al. 2009) are necessary conditions for the occurrence of slipping reconnection.

Why do we then detect the apparent slipping motion of the flare loops and low values of N? In a strict sense, Equation (1) is valid only for a given field line in a given instant in time, with one footpoint anchored in one polarity-related QSL moving at speed v_QSL, while the opposite moving end moves in the conjugate ribbon at speed v_slip, with each quantity deduced on a short timescale, as the profile of the local norm N varies strongly along and across the QSL (see Figure 10(d) in Janvier et al. 2015). It is difficult to find a satisfactory relation between the analytical expression in its strict sense and the present data because both speed quantities are obtained only as averages over multiple locations, multiple times, and multiple field lines, which are themselves anchored in different sections of the QSL photospheric footprints. This is difficult to avoid because the remote-sensing observations generally rely on plasma emission, which in optically thin conditions is always dominated by dense(r) plasma. While Equation (1) is only strictly valid for a given field line in a given instant in time, in the observations a time delay also exists between the energy deposition in the ribbon and the filling of the flare loops by hot and dense evaporated plasma. Indeed, in Section 3.2, it was found that the Fe xxi blueshifted emission observed by IRIS is not exactly cospatial with the ribbon, as seen in transition-region Si iv emission or the FUV continuum (Figure 9). Therefore, we may be measuring the slipping speed slightly outside the QSL, which may have already moved to an adjacent location.

Furthermore, the v_slip determined from the apparent motion of a flare loop may not represent the slipping velocities of a single field-line footpoint, which could even be super-Alfvénic (Aulanier et al. 2006, 2012; Janvier et al. 2013). Rather, the observed slow, sub-Alfvénic slipping velocities may be an illusion resulting from the variable rate of the energy deposition into the chromosphere as a result of many slipping field lines along the ribbon. In any case, large, super-Alfvénic slipping velocities are out of the reach of current observations, which only have a cadence of the order of 10 s. Subsecond temporal resolution would be needed to distinguish a fast slipping motion. An additional complication is that a large v_slip would result in lower energy deposition per unit time and unit area of the ribbon, resulting in less evaporated plasma, that is, fainter loops. We note that faint, near-vertical stripes are present in the time–distance plots in Figures 5 and 8. For example, a short, near-vertical strip is located at 17:03:40 UT at the position Solar Y = 100''–103'' in Figure 5. Nevertheless, the 12 s cadence of AIA does not permit measurement of the velocity of such an intermittent stripe.

Chifor et al. (2007) studied the precursor activity before major flares and found that distinct, localized X-ray brightenings occur 2–50 minutes before the onset of the impulsive phase. These brightenings occured within 10'' of the polarity inversion line (PIL) and had both thermal and nonthermal components as observed by RHESSI. Typically, these X-ray brightenings also had an EUV counterpart observed either by SoHO/EIT (Delaboudinière et al. 1995) or TRACE (Handy et al. 1998). The main energy release in the impulsive phase occurred within 50'' of the locations of the preflare brightenings. Furthermore, the filament eruption began at the location of the preflare brightenings that also coincided with the presence of emerging or canceling flux. Chifor et al. (2007) interpreted these as the signatures of tether-cutting mechanisms (e.g., Moore & Roumeliotis 1992; Moore et al. 2001, 2011; Fan 2012) triggering the main flare and the eruption.

Inspecting the Helioseismic and Magnetic Imager (HMI) and AIA 304 Å data, we find that all these observational characteristics are present during the early flare phase at approximately 16:50–17:20 UT. A weak X-ray emission is detected in the 6–12 keV channel by RHESSI, an example of which is shown in Figure 9 at 17:03 UT. The advantage of IRIS and SDO/AIA is that they offer a good temperature coverage of the solar atmosphere from the chromosphere upward, including flaring plasmas. The SJI 1400 Å together with the AIA 1600, 1700, 304, and 171 Å images reveal that the EUV brightenings occur within the ribbons involved in the main flare. In fact, these EUV brightenings are the first instances of the ribbon development (like in Dudík et al. 2014b), with a one-to-one correspondence to footpoints to the flare loops exhibiting apparent slipping motion (Section 2.3). This leads us to conclude that, at least for the flare studied here, the precursors are in fact signatures of the flare itself, progressing from its early phase toward the impulsive phase. This is in line with the results of Fárník & Savy (1998), who concluded that for the preflare events (precursors) cospatial with the main flare, the soft X-ray emission is present with the same size, shape, and orientation at least 5 minutes before the onset of the impulsive phase.

Furthermore, the standard solar flare model in 3D of Aulanier et al. (2012) and Janvier et al. (2013) predicts explicitly that the slipping reconnection creating the flare loops and building the flux rope is the tether-cutting mechanism itself. An important distinction, however (see Section 5 of Aulanier et al. 2012), is that this tether-cutting does not trigger the eruption itself, as in the cartoon of Moore et al. (2001); rather, it contributes to the building of the flux rope, which then erupts via the torus instability. We note that the presence of tether-cutting reconnection was already suggested independently for this event by Cheng et al. (2015), who analyzed AIA and IRIS Si iv, C ii, and Mg ii observations.

The mechanism of the slipping reconnection and tether-cutting is represented in the cartoon shown in Figure 13. This is essentially a 3D representation of the magnetic field configuration in the standard solar flare model in 3D, including the dynamics during the eruption of a flux rope (see also Janvier et al. 2015, Figure 11 therein). The flux rope is represented by dash-dotted purple lines. It is anchored in each polarity in the hooked region of the flare ribbons (red), corresponding to the photospheric signature of the QSLs and the current density volume, represented in purple. The QSL wraps around the flux rope, with the hyperbolic flux tube, where the connectivity distortion is the highest, extending underneath the flux rope. Magnetic field lines entering this high current density volume reconnect successively in the slipping manner. Pairs of such field lines undergoing successive reconnections are represented by colored field lines in Figure 13: blue and yellow field lines have fixed footpoints in the positive and negative ribbons, respectively. Their reconnection counterparts are shown in green and orange, whose footpoints are again fixed in positive and negative ribbons, respectively. The blue field line slip-reconnected with the orange field line. The consequences are represented by the moving end of the blue field line in the negative polarity, away from the PIL and toward the legs of the flux rope. The slip-reconnected counterpart of this blue field line is the orange one, represented by a fixed footpoint anchored in the negative polarity in the left part of the cartoon. Its conjugate footpoint in the positive polarity is slipping and is accompanied by flare kernels moving toward the PIL. The other reconnecting pair is the yellow and green field lines, whose one footpoint is fixed in the negative and positive ribbons, respectively, while the conjugate footpoints move along the conjugate ribbons (Figure 13). Together, these four field lines create motions toward both ends of both ribbons. This important prediction of the model is vindicated by the observations reported in Section 2.3. These observations are in contrast to the first report of slipping motion of flare loops in another flare (Dudík et al. 2014b), where the slipping motion was observed to be predominantly in one direction, toward the hooks of ribbons.

Zoom In Zoom Out Reset image size

These successive reconnections represented in the cartoon in Figure 13 come to an end when the two field lines have gone across the current density volume. The orange field line becomes a low-lying flare loop, while the blue field now wraps around the erupting flux rope and becomes its envelope, likely observed as the hot S-shaped erupting loop. The slippage is likely accompanied by particle acceleration in the reconnecting region. These particles propagate along each successively connected field line. The deposition of energy in the chromosphere is seen as the kernel brightenings. These slipping brightenings (Figures 4 and 7), corresponding to the precursors in the early flare phase, are shown as bright patches appearing at different times at different positions along both ribbons.

In summary, the precursor activity as a result of the tether-cutting reconnection is fully consistent with the standard solar flare model in 3D. The presence of the apparent slipping motion during the precursor phase, clearly identified here as such for the first time, is the telltale signature uniting the precursors and the 3D model.