4. Implications for flare models
The finding that flare kernels are located on QSLs and that they are linked by field lines extending along them, together with the presence of a system of electric currents, brings new light on the flare mechanism. This section intends to put the results found with the QSLM within the framework of our present knowledge of this mechanism.
4.1. Flaring loop models
Flare loop models invoke a plasma micro-instability in a single magnetic loop when the electric current is too large (e.g. Alfvén & Carlqvist, 1967) or an ideal instability when the twist is too large (e.g. Miki et al. 1990). Since we show that more than one simple loop is involved in the studied flares, our results do not support single-loop flare models. This is consistent with the low-twist of the current loop.
The present results seem to be in conflict with the "standard" image of solar flares which has emerged from observations aboard Yohkoh: one soft X-ray loop with a pair of footpoints and a top source in hard X-rays (Kosugi, 1994; Masuda, 1994; Masuda et al., 1995). However, two interacting loops are also often observed in soft X-rays in flares (Inda-Koide, 1994; Dennis et al., 1994, Hanaoka, 1995) and interacting multiple loops are an important fraction ( %) of the transient brightenings observed in ARs (Shimizu et al., 1994). Besides multiple sources in hard X-rays are not a negligible fraction ( %) of the observed flares (Sakao et al., 1994) and some remote hard X-ray sources linked to the flare loops are not seen in soft X-rays, but appear bright in microwaves (e.g. Yaji et al., 1994). Interaction between loops is also observed in the corona in Fe lines (Smartt et al., 1993).
There are, at least, three reasons that can explain why only a single loop is seen in soft X-rays. First, as single-loop events are found to be smaller in extension (usually less than 20 Mm) than multiple-loop events (Shimizu et al., 1994), the observation of single-loop events can be due to a lack of spatial resolution (Shimizu et al., 1994). Second, in bipolar regions the two reconnected loops are close together and can be easily taken as one loop (see Figs. 3b, 4b for example). Third, even in quadrupolar regions where the X-ray loops are well separated, a large difference in size between them implies a large difference both in energy input per unit of volume and in density enhancement (e.g. due to evaporation). This implies that the longer loop, much less bright, can be overlooked, in particular if the time exposure is chosen for not overexposing the short bright loop. Such situation occurs when a small magnetic bipole impacts into a much larger one (see e.g. Mandrini et al., 1996). In hard X-rays there are also several reasons for missing some sources: the evolution of the hard X-ray sources during the flare from the feet to the top of loops (e.g. Yaji et al., 1994), the field strength difference between the feet of the loops implying different mirroring conditions of particles (Sakao et al., 1994), the anisotropy in the injection of particles, and finally only two X-ray footpoint sources are expected in bipolar regions.
While the spatial resolution is better for H observations, a direct analysis can also falsely support a single-loop flare model. Because H kernels are located inside the feet of X-ray loops, it is natural to find the counterpart of the three points listed above. First, for some small flares, the H observations may not have the resolution required to visualize the ribbons (for example kernel d in Fig. 1). Second, in bipolar regions, two ribbons are expected while four kinds of field-line linkages exist like in quadrupolar regions (e.g. Fig. 3). Third, some less intense kernels may be overlooked. In this sense the computation of QSLs is a useful tool to find the relevant features related to a given flare (for example, the faintest UV kernel in November 12, 1980 flare was found when looking for it in the zone close to the computed separatrices, see Bagalá et al., 1995).
4.2. Magnetic reconnection: a common feature to flares
In the set of events studied previously only one well-defined current loop associated to the flares was found. In AR 2372, a second current loop is present in the main bipole but its foot in the preceding spot is not located in the vicinity of the QSL (Fig. 1b). Moreover, this current evolves drastically from April 6-8 while homologous flares are observed (Mandrini et al., 1993). This main-bipole current seems then not to take part in the flaring process. Therefore, none of the studied flares appears to occur because two current loops attract as proposed in the current-loop coalescence models (e.g. Sakai & De Jager 1996, and references therein).
An emerging flux model for flares was proposed by Heyvaerts et al. (1977), being further investigated by Forbes & Priest (1984). The magnetic configurations of April 6-8 and June 13-15, 1980 are cases where this model can be applied. The eruption of a twisted flux-tube (e.g. Priest 1982, p. 367) is another candidate to induce flaring. The absence of a prominence eruption is not a problem for this model because the low plasma in the prominence has a negligible influence on the magnetic field. The main problem of the twisted flux-tube model, in the case of the confined flares studied, is rather the absence of a twisted configuration in the extrapolated coronal field. Of course this can be due to the intrinsic limitations of the extrapolation method (which cannot include the concentrate currents observed at the photosphere), but the fact that we find the flare kernels on QSLs shows that the extrapolated configuration is realistic. Moreover, the electric currents, measured directly from the transverse field, are not intense enough to form a twisted flux-tube and the flare ribbons have not the characteristic shape ("umbrella hand" shape) present in twisted configurations (Démoulin et al., 1996a).
For the studied flares, the locations of flare kernels in relation to QSLs point out clearly a reconnection mechanism. Due to the intrinsic difficulties of the problem, magnetic reconnection has been studied mainly in 2D and 2 D magnetic configurations (see Malherbe 1987 and Priest 1992 for reviews). In these models, energy release occurs on separatrices and, in particular, on their intersection; so the H kernels should be located where separatrices cut the chromosphere. Four brightenings are expected in quadrupolar regions and only two in bipolar ones involving a twisted magnetic structure. Therefore, with a symmetry of translation, the configuration associated with four and two-ribbon flares is quite different: the first kind of flares involve the interaction of two bipoles and the second one supposes the formation of a twisted flux-tube. The possibilities of field-line connections are much larger in 3D. In Paper I we have shown theoretically that we can pass continuously from a case with four ribbons to a case with two. The observations confirm this view: flares are present in configurations where two bipoles are antiparallel (Fig. 1), or when they are less antiparallel (Figs. 2, 3, 5), or even nearly parallel (Fig. 4). In 3D magnetic configurations, two-ribbon flares do not need the formation of a twisted flux-tube and may happen even in bipolar regions having low magnetic shear, so apparently in a simple magnetic arcade. All the studied flares strongly support a model based on magnetic reconnection taking place in different configurations. Theoretical research in 3D magnetic reconnection has started few years ago, and at this point solar observations seem to support this growing-up theory !
4.3. Formation and release of the electric currents
How are the current loops formed which are powering the flares? Based on the observations several ways have been proposed: by spot motions (e.g. Gesztelyi et al., 1986; Hanaoka, 1994), by photospheric twisting motions (e.g. Martres et al., 1970; Hénoux and Somov, 1996) or by emergence of twisted flux-tubes (Leka et al., 1996). In these cases a loop current is formed at some location in the AR and a flare occurs when the loop current reaches the QSLs. Another possibility, developed in Paper I, is that concentrated currents are naturally formed by any photospheric motion at QSLs. This is so because two neighboring field lines are subjected to different photospheric motions since their opposite footpoints are separated by a great distance and, therefore, electric currents with strong densities are created at QSLs. This gives a natural explanation to the results of Sect. 3.3; though only one current loop is detected above the noise level because the currents formed are stronger in the inner bipole where the photospheric velocities are larger. In fact, with the present data we cannot decide if the currents are transported to or formed in the QSLs; we need both a better time coverage with magnetograms and more accurate linear polarization measurements to follow the time evolution of the currents.
When QSLs are thin enough, magnetic energy release is possible at their location either because a current-density threshold is reached (Paper I) or because the field-line velocity becomes much larger than the plasma velocity (Priest & Démoulin 1995). This view is supported by the recent analysis of the temporal evolution of QSLs associated with an X-ray bright point: Mandrini et al. (1996) show that the QSL on the emerging bipole is very thin (typically less than 100 m) during the lifetime of the XBP, but becomes much thicker ( m) after the XBP has faded. In some of the flaring regions studied here (AR 2372, AR 2779, AR 2511), the QSLs are thin enough to allow reconnection; but in some regions (e.g. AR 2776) the QSL thickness can be as large as 1 Mm ! We can interpret these results as follows: in flares, storage of magnetic energy is a necessary first step, contrasting with XBPs where the energy is released as soon as it is available. For such storage, a QSL thickness of the order of 1 Mm is enough to create concentrated currents (as observed). Then at some point the quasi-static evolution of the configuration cannot go on, because QSLs are becoming too thin, and a flare occurs. The influence of growing electric currents on QSLs remains to be investigated with a nonlinear force-free field extrapolation, but there are some analytical results that show that the QSL thickness decreases exponentially with increasing twist (Démoulin et al., 1996a).
© European Southern Observatory (ESO) 1997
Online publication: May 5, 1998