Are All Flare Ribbons Simply Connected to the Corona?^*

We refer to T2014 for the circumstances of the IRIS and other observations of active region (AR) 11890 on 2013 November 9. Figure 2 shows images of interest from IRIS and the AIA instrument. The ribbon of central interest here contains the point "D" analyzed by T2014, lying 252''  SW of Sun center. The local vertical is $\vartheta =15^\circ$ from the line of sight, with $\mu =\cos \vartheta =0.96$. We will refer to this ribbon as the "ROI."

Zoom In Zoom Out Reset image size

During the IRIS observations, the chromospheric ribbon brightened twice, of which only the first, weaker event was discussed by T2014. Here we concentrate on the second brightening, which is similar in morphology. This makes no difference to our conclusions because we examine the magnetic connectivity between the corona and ribbon prior to both brightenings. The bright ribbon is clearly seen at its brightest phase (12:15:41 UT) in the lower left panel of Figure 2. This ribbon's locus is traced in the other images using white lines.

The ribbon is barely seen in the 1600 Å AIA channel, which admits contributions from the UV continuum, the resonance lines of C iv, and the Balmer-α line of He ii (1640 Å). It is clearly seen in the Mg ii slit-jaw image (Figure 2), which forms across the entire chromosphere. The AIA image at 94 Å appears superficially to overlap with the ribbon. However, in movies S1 and S2 published by T2014, the 94 Å emission clearly originates from closer to the larger UV flaring patch seen near (X, Y) = (167, −192), highlighted with a black ellipse in our figures. The 94 Å coronal emission appears close to but over the ROI in a loop-like structure connected to the region of the 150 G line-of-sight field, near (167, −192). We have examined all other channels from the AIA instrument between 12:15 and 12:16 UT. The clearest indicator of the ROI seen in AIA data is in the 304 Å channel (mostly He ii). It is not detected in the 1700 Å UV continuum channel, but during the ribbon's brightening there are hints of weak emission at 171, 193, 211, and 335 Å, all coronal bands.

In conclusion, the near absence of 1600 Å ribbon emission and its presence in lines from Mg ii to He ii indicate that the ROI's emission originates from the mid- to upper chromosphere and the transition region.

Figure 3 shows magnetic parameters inverted from raster scans with the SP instrument on the Hinode spacecraft (Lites et al. 2008). We used data from the CSAC at HAO based on the code MERLIN. The coordinates are all rotated to a reference frame close in time to the two ribbon flaring events obtained using the HMI instrument (Schou et al. 2012) on SDO. Given the small size of the ribbons, however, we chose to base our further analysis on the Hinode raster scans, due to the higher spatial resolution and magnetic sensitivity. In spite of the evolution of solar features during the Hinode scans (the 30'' widths of the images shown took ≈6 minutes, comparable to a granule lifetime), the Hinode images shown have been successfully co-aligned to those of HMI to ≈± 03. The magnetic data are striking. There is essentially just noise in the line-of-sight (longitudinal) component of the field B_L near the ROI. The Hinode data give a mean value of 11 ± 10 G for the longitudinal field calculated from areas extending 2'' around the ribbon loci shown. The noise in longitudinal field acquired with the SP corresponds to just 3 Mx cm⁻² in each spatial pixel (Lites et al. 2008). The average Hinode tangential (near-horizontal) field component B_T is 73 ± 24 G.

If similar connections existed from the ROI, it would have several such origins distributed along its locus in the 94 Å or other coronal images.

The raster scans from the Hinode spectropolarimeter shown in Figure 3 were obtained about 1 hr before and 6 hr after the flaring activity. An intermediate scan was started but failed during the flaring period; it contains no useful data. But from the data shown, three characteristics caught our attention. First, the 200 G horizontal flux seen near the ROI before the flaring is replaced by 200 G vertical field components hours after the flare. Second, the azimuth of the field in the neighborhood of the ROI before the flare has significant spatial coherence, with a mean value close to ${44^\circ }_{-27}^{+31}$ relative to the N–S direction. This aspect suggests, albeit weakly, the presence of an emerging magnetic field roughly along the same direction as the long axis of the ribbon. An HMI time-series animation of the field evolution in between the two Hinode raster scans is provided in the online animated version of Figure 3. HMI unfortunately does not have the sensitivity to reveal horizontal fields in the upper photosphere of the ROI during this period.

The question of connection of the ribbon to the overlying corona concerns only the connection between chromosphere and corona (such small flares generate undetectable changes in deeper layers). The magnetic field threading the chromosphere must span several scale heights and must therefore be significantly different than measured in the photosphere, which therefore presents a difficulty. But given the measured photospheric fields, it is hard to see from where a chromospheric field connected upward might originate. The entire stratified chromosphere spans about 1500 km, which corresponds to ≈2''. In Figure 3 the ribbon is encased in a void of field with values ≲10 G extending at least 2''  all around the ribbon. We already noted the absence of moss emission in this area (Figure 1 of T2014).

Some weak longitudinal field may suffice to provide unseen connections to the corona along the ribbon. Indeed, a field is present to the south of both ends of the ROI, of the same polarity, 30–40 G in strength and covering just a couple of square arcseconds (see the Hinode panel B_L in the top left of Figure 3). These fields are not visible in HMI by the time of the flare, 1^h22^m later (second row of the figure). These small concentrations are some 5''  from the center of the ROI. Could longitudinal fields such as these have been advected to the locus of the ROI by known surface motions in time for a coronal connection to be established? Berger et al. (1998) studied surface diffusion rates around ARs, finding a coefficient of κ ≈ 80 km² s⁻¹ for several tens of minutes. In the t = 100 minutes after the first Hinode scan, elements would diffuse a distance of $\tfrac{1}{3}\sqrt{\kappa t}=\tfrac{1}{3}\sqrt{80\times 6000}\approx 200$ km, which is an angular distance of 03. Supergranular diffusion in regions far from strong fields is larger, with $\kappa \lesssim 800$ km² s⁻¹. Even with such a high rate of diffusion, the surface diffusion would have to accumulate from within a distance of 1'' of the ribbon and produce vertical components along the ROI in the beam-heating scenario. This seems unlikely given the quasi-random nature of this diffusion implicit in the work of Berger et al. (1998).

In the 6 hr between the flare and the second Hinode scan, relatively strong post-flare longitudinal "network" fields appear in the lowest panel of Figure 3. It is seen meandering underneath the position of the ROI, with a length ${\ell }\approx 10^{\prime\prime}$. Emergence and/or surface diffusion of field must be responsible for this new network flux. The new network flux amounts to ${\rm{\Phi }}\approx 200\,{\rm{G}}$ times the apparent area of the network, ≈1'' × ℓ''  $\equiv \,5\times {10}^{16}$ cm², so that Φ ≈ 10¹⁹ Mx. To build this flux from the mean pre-ribbon longitudinal field strength of 10 G, this flux would have to have collected, merged, and strengthened from an area of about 200 square arcseconds, about 100 Mm². This area is far larger than the area available for diffusion in the time available, which we write as $L{\ell }\approx \tfrac{1}{3}\kappa \,21600$ ≈ 6 Mm², even using κ ≈ 800 km² s⁻¹.

It seems that some flux emergence into the chromosphere occurred between the first and second Hinode scans. Such flux emergence is below the detection limit for HMI (row 2, column 2 of Figure 3) but is hinted at by the coherent horizontal field seen in the first Hinode scan (row 1, column 2 of Figure 3). We suggest that this flux emergence might help explain the presence of the ROI, in particular, the horizontal fields that appear somewhat close to the ribbon in the pre-ribbon Hinode scan strongly indicate horizontal flux. The series of HMI vector field maps was examined between 09:58 and 20:10 UT; from 11 to 19 UT is shown in the online animated version of Figure 3. From this movie, any horizontal fields are too noisy to be seen as any coherent structure rising through the photosphere. Beginning close to 18:00 UT, the longitudinal field patch close to (−165, −201) migrated quickly toward the position of the clump of flux seen underlying the ROI in the last Hinode (18:09–19:14 UT) scan.

Even if we cannot conclusively detect horizontal emerging flux, we can conclude that the magnetic conditions around the ROI are qualitatively different from the other ribbons discussed by T2014.

But can we discount a coronal origin for the ROI? Let us consider the energy requirements; for this we can use Figure 3 of T2014. From this figure we see that the Mg ii k line has excess emission averaging 400 DN/s/spatial pixel for a period of 200 s. Using calibration factors from SolarSoft routine iris_get_response.pro, we find that the intensity of Mg ii k is ${I}_{k}=1.5\times {10}^{5}$ erg cm⁻² s⁻¹ sr⁻¹. The k line contributes ≈10% of the total chromospheric radiative losses known in 1981 (Table 29 of Vernazza et al. 1981), and allowing for a doubling of the losses due to the inclusion of Fe ii lines (Anderson & Athay 1989b), we find that the excess emission is $\approx 20\times 4\pi {I}_{k}\approx 4\times {10}^{7}$ erg cm⁻² s⁻¹. The same calculation for the Si iv λ1403 line gives an excess intensity of 6 × 10³ erg cm⁻² s⁻¹ sr⁻¹, 25 times smaller than seen in the k line.

We assume that we have a field strength of 10 G, corresponding to the 3σ upper limit, emerging from the region of the ROI into the corona. (Below we relax this assumption.) The total energy per unit area in a straight tube of length L is ${{LB}}^{2}/8\pi$, which for L ≈ 10¹⁰ cm (100 Mm, as used by T2014) and $B\lesssim 10$ G gives $\lesssim 4\times {10}^{10}$ erg cm⁻². The 304 Å frame in movie "S2" of T2014 shows that the heating of the ROI continues beyond the time span of the IRIS sequence, which ends at 12:17:16 UT. Thus, we can set a time $\gtrsim {10}^{2}$ s for the duration of the heating. Therefore, the overlying corona can supply $\lesssim 4\times {10}^{8}$ erg cm⁻² s⁻¹ of magnetic power per unit area into the chromosphere below. If, optimistically, we assume that 10% of the magnetic energy along this tube is free energy available, then just $\lesssim 4\times {10}^{7}$ erg cm⁻² s⁻¹ is available. This is close to the energy requirements of the quiet-Sun chromosphere (Anderson & Athay 1989a) and our excess estimate above. An alternative limit can be derived using the Alfvén crossing time for the tube, which is $L/{V}_{{\rm{A}}}$. With a pre-flare coronal gas pressure p ≈ 0.1 dynes cm⁻² (Jordan 1992), T = 10⁶ K, we have a density $\rho \approx 2\times {10}^{-15}$ g cm⁻³ and ${V}_{{\rm{A}}}\lesssim 700$ km s⁻¹. Then $L/{V}_{{\rm{A}}}\gtrsim 140\,{\rm{s}}$.

The energy available in a coronal tube is formally enough to account for the radiation losses from the chromosphere below. But our estimate is a strict upper limit for the following reasons: (1) We have used the 3σ "detection" of longitudinal field from the Hinode scan 1 hr before the flare, and the energy is $\propto {B}_{L}^{2}$. The field strength in the overlying corona is surely lower than the 10 G upper limit in the photosphere. (2) The energy is $\propto L$, and L has been taken from neighboring coronal structures seen clearly in the 94 Å channel that are rooted in far stronger concentrations. (3) We have assumed that 10% of all the magnetic energy is available for release within the tube. (4) We have assumed that sufficient time exists for this energy to be released. Reconnection typically takes place at $0.1{V}_{{\rm{A}}}$, through complex dynamical mechanisms such as the plasmoid instability (e.g., Bhattacharjee et al. 2009). Thus, the dynamical time for energy release becomes too long, $L/0.1{V}_{{\rm{A}}}\approx 1400\,{\rm{s}}$.

It appears that the ROI cannot be reconciled by heating from the overlying corona. Either the dynamical time is too long (L = 100 Mm) or, if L is small enough to account for the dynamical time, the free energy $\propto L$ is insufficient to account for the observed losses.

Now let us permit the coronal connection to include the transverse component of the field, as well as the longitudinal component. Then we expect the connection to the corona to be highly oblique to the line of sight, as is seen in the 94 Å image (Figure 2). In this case the energy available is some 50 times bigger and the Alfvén speeds seven times larger. In this case, the observations are reconciled with a beam-heating picture: the energy release rate from the corona is $\lesssim 2\times {10}^{9}$ erg cm⁻² s⁻¹ and the dynamical time is about 200 s, compared with about 2 × 10⁷ erg cm⁻² s⁻¹ and 100 s, respectively.

But is this scenario credible? The morphology of the horizontal field and the absence of moss coronal emission simply do not support the idea that the fields detected are connected to the corona.

We are forced to consider local sources of heating for the ROI. The tangential field in the Hinode scan is around 75 G. Let us consider the emergence of a near-horizontal tube of flux through the chromosphere in a quasi-steady state prior to some energy release process. It emerges into a preexisting field in the upper chromosphere in the AR where a preexisting canopy field exists, fields rooted elsewhere in stronger concentrations that have turned horizontal to ensure pressure balance as the chromospheric gas pressure drops well below the magnetic pressure. The total energy density available in an emerging tube with this field strength is ${B}_{T}^{2}/8\pi$ ≈ 220 erg cm⁻³. The Alfvén speed is roughly ${B}_{T}/\sqrt{4\pi {\rho }_{u}}\approx 200$ km s⁻¹, where ${\rho }_{u}\approx {10}^{-12}$ g cm⁻³ is the density in the upper chromosphere. Let us assume that the tube emerges into canopy fields, rooted in more distant network patches, of a lower strength. To estimate this strength, we note that in the neighborhood of the ribbon the longitudinal flux is all of the same sign, and the average longitudinal flux density over a region centered at the ribbon is about 55–80 Mx cm⁻², depending on the local area chosen. For the surrounding 20'' × 20''  area (of order one supergranule in size) the average is 80 Mx cm⁻². We envisage that the locally emerging flux begins to reconnect with the canopy field and generate the extra emission in the ROI. To account for the ribbon, we require, as above, 4 × 10⁷ erg cm⁻² s⁻¹ for about 200 s. Once the reconnection starts, it will proceed at a rate close to $0.1{V}_{{\rm{A}}}^{\prime }$, where ${V}_{{\rm{A}}}^{\prime }$ is the Alfvén speed in the reconnecting component of the field. If the angle between the emerging tube and the canopy fields is ϑ, then ${V}_{{\rm{A}}}^{\prime }={V}_{{\rm{A}}}\sin \vartheta$ Thus, the reconnection will convert magnetic into thermal energy at the rate

$$\begin{eqnarray}&&{ \mathcal F }\lesssim 0.1{\sin }^{3}\vartheta \displaystyle \frac{{B}_{T}}{\sqrt{4\pi {\rho }_{u}}}\displaystyle \frac{{B}_{T}^{2}}{8\pi }\lesssim 5\times {10}^{7}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}\,{{\rm{s}}}^{-1},\end{eqnarray} \tag{ 1 }$$

where we have used $\sin \vartheta =0.5$ as a rough estimate of the mismatch in directions of the horizontal and canopy fields. It seems that there is power available from chromospheric reconnection to drive the radiation losses seen in the ROI. With a (vertical) inflow speed into the reconnection layer of $0.1\sin \vartheta {V}_{{\rm{A}}}\approx 25$ km s⁻¹, lasting for ≈200 s, the reconnection advects a total mass per unit area of ${m}_{R}=2.5\times {10}^{6}\,\times 200{\rho }_{u}=5\times {10}^{-4}$ g cm⁻² from the chromosphere into the reconnection layer. If this partially ionized plasma is heated via dynamical instabilities (e.g., the plasmoid instability; Bhattacharjee et al. 2009) or kinetic processes (ion–neutral collisions, for instance), then this emerging flux effectively leads to heating in the upper chromosphere for column masses above ${m}_{R}=5\times {10}^{-4}$ g cm⁻². Interestingly, this corresponds roughly to the range of heights between which the proposed beam dissipation arises in the models of T2014 (Figure S4). Therefore, we would expect similar dynamical signatures from this type of heating to those from beam heating when the reconnected field has access to the overlying corona.

One might argue that the chromosphere remains "closed" rather than open to the overlying corona, and therefore that lines formed are trapped in closed field lines under a low-β regime in which significant line shifts observed in Si iv cannot be observed. However, in this picture we envisage plasma mass advected into the canopy fields, which are themselves open to the overlying corona, a process commonly called "interchange reconnection." We would naturally expect the same kind of dynamics along one of the canopy field lines as computed in the beam heating scenario, since both result from deposition of energy in plasma of similar column masses, which later enters via reconnection a tube of field that is connected to the corona. This ambiguity reminds us of the sobering reality that lines such as Si iv are like a dog with two masters—both the corona and chromosphere have significant effects on such transition-region lines. Also, it must be remembered that Si iv intensities are a factor of 25 weaker than the Mg ii lines. Thus, these lines reflect only a small fraction of the energy release and its properties in the evolving atmosphere.

A common reason for discounting chromospheric reconnection is that the ribbons "light up" between observations separated by tens of seconds, and that there is nothing that can physically communicate one part of the chromosphere to reconnect on these timescales. But this is not correct near ARs with transverse fields of order 70 G. The ROI has a horizontal extent of about 7 Mm (Figure 2), so it would take just 14 s for magnetic perturbations, such as a tearing mode, to propagate horizontally across the surface. Further, an emerging flux rope defines a ribbon-line morphology as it interacts with the overlying fields. Therefore, we respectfully disagree with the dismissal of chromospheric reconnection discussed in Section S3 of T2014, in the case of the ROI.

Finally, a widely held belief that the chromosphere is highly dynamic has arisen from studies seeking dynamical phenomena associated with the chromosphere (e.g., de Pontieu et al. 2007). The question would then arise whether the upper chromosphere would shred a rope of horizontal flux before it could interact with preexisting fibrils extending across supergranular cells. But this is a nonissue since simple estimates of such things as spicules and related phenomena show them to cover no more than 0.1% of the solar surface area and to originate at supergranular vertices. Spectral observations of the solar disk with HRTS by Dere et al. (1983) show that both line widths and shifts very rarely approach those of the dynamic type II spicules reported by de Pontieu and colleagues. It is also clear from narrowband imaging observations of spectral lines formed in the upper chromosphere, such as the Ca ii infrared triplet lines, that fibril structures are stable on timescales of hours or more (e.g., Cauzzi et al. 2008).