The frequent presence of filaments on the disk, particularly during the high activity phase of the solar cycle, suggests that they are a common product of the solar atmosphere. Observations reveal that, after a filament eruption, the feature may reform in the same place (e.g., Bone et al., 2009*). This suggests that the conditions for filament formation may not be destroyed by the eruption or that they can be easily reproduced. An overview of our knowledge of the filament formation processes is presented below; more details can be found in Mackay et al. (2010*).
Filaments can be found above the PIL separating a simple bipolar region or more complex magnetic structures, including multiple bipolar regions. A classification in these terms was done in the past (Tandberg-Hanssen, 1995*), with the result that filaments favor the second (more complex) magnetic configuration (e.g., Tang, 1987*). A recent update of classification was done by Mackay et al. (2008*), who concentrated on long stable structures. They introduced new magnetic categories for the classification, as illustrated in Figure 24*, and we direct the reader to their work for further information. They showed that indeed 92% of filaments are found between multiple dipoles. As pointed out by these authors, the existence of a preferred magnetic configuration where filaments are located is indicative of the physical conditions suitable for their formation.
We have mentioned the different magnetic topologies that have been proposed for the filament support: sheared arcade (Antiochos et al., 1994; DeVore and Antiochos, 2000) and/or flux rope (e.g., Amari et al., 2000, 2010*). There are two possibilities for their formation: either they emerge from the photosphere, or they are the result of the modification of an existing magnetic structure due to photospheric flux cancellation and/or coronal reconnection. In the latter case we expect to find the properties of the magnetic envelope of the filament to be similar to those of the pre-existing magnetic region.
To interpret the observed filament formation in terms of these different proposed pictures we need to infer the magnetic field configuration of the filament and that of the pre-existing field. Considering the difficulty of their measurement, the information should be complemented or supplied by identifying all other possible observational signatures associated with filament formation. These various identified factors that influence or are associated with the filament formation and its magnetic structure are, in summary, the following: photospheric flow motion toward the PIL, photospheric magnetic topology and intensity, large scale plasma flows in the photosphere due to meridional circulation and differential rotation, magnetic flux emergence and cancellation, chromospheric small-scale flows, flows along the PIL.
In the literature, cases are found where one or more of these phenomena were recognized during filament formation (e.g., Mackay, 2003; Mackay and van Ballegooijen, 2006*). The presence of one process, or the combination of several of them, may lead to one or the other of the above mentioned filament magnetic-field topologies. Furthermore, the dominance of one or another of these mechanisms appears to be linked to the different activity levels on the Sun. For example, large-scale flux emergence is typical of AR areas and not of polar-crown regions (Mackay et al., 1998), so flux emergence will be more likely to be associated with active filament formation.
The first phase of filament formation, which lasts a few hours, is characterized by the Hα fibrils aligning along the PIL to form the filament channel, as shown in Figure 25*. Then the filament channel typically can be filled with plasma in a few days. The fibril alignment on both sides of the PIL can be parallel or anti-parallel. One proposed cause of such alignment is in the relative motion of the photospheric fibril footpoints in the two different polarities (e.g., Wang and Muglach, 2007*). Photospheric flux convergence is often caused by supergranular divergence motions at the sides and across the PIL (e.g., Rondi et al., 2007*). Both cases studied by Wang and Muglach (2007*) exhibited convergence toward the neutral line, but in the first case there was also parallel motion, while it was anti-parallel in the second case. However, Gaizauskas et al. (1997) noted that the change of the fibrils’ direction was too fast to originate in local reconnection in the photosphere driven by converging motions. They also associated the filament channel formation with the emergence of an activity complex. Hence, the question of the origin of the fibril realignment probably remains open.
The filling of the channel with plasma is not a homogeneous process. It can proceed by involving pack of threads or thread segments, each of them happening at different times. Thread segments may evolve into a unique longer filament. In addition, we recall that the channel of a well-formed prominence does not necessary overlap the full PIL length. It is not yet clear what governs this filling and the eventual filament channel evolution. The process is days and even weeks long, which means, to be properly monitored it needs to be observed continuously. This would also require observations in the far side of the Sun, which is yet to be accomplished by the available instrumentation.
Once a filament is formed, the dynamics around and beneath the filament continue, changing the filament shape and, in certain cases, possibly destabilizing it. One of the signatures of these dynamics is the continuous local convergence of magnetic flows toward the PIL. The presence of sporadic minor polarity concentrations around the PIL has also been considered important for filament formation. Rondi et al. (2007) claim that the converging motions may transport these small scale polarities across the PIL, thus forming the parasitic polarities in which filament barbs are thought to be rooted. They identified this process as causing the formation of the filament barbs, as supported by other observations (e.g., Schmieder et al., 2006) and modeling (e.g., Aulanier and Schmieder, 2002).
As mentioned, a pre-eruption flux rope structure in a filament channel has been identified only in a few cases. Okamoto et al. (2008, 2009*), using SOHO/MDI and HINODE/SOT Hα, Ca ii H and G bands images, claimed to have recorded the emergence of a flux rope inside an active region. Observationally this was associated with the temporal change of the horizontal magnetic-field region along the PIL, including its broadening and narrowing in time, and the change from normal-polarity to inverse-polarity, suggesting the lifting off of the structure. Okamoto et al. (2009*) also measured diverging motions in granular cells (3000 – 5000 km wide and 60 km deep). In a similar way, Lites et al. (2010*) interpreted their data as the result of the emergence of a pre-existing flux rope from below the photosphere. In both cases, the local granular motions dominate over the photospheric flows, and there is absence of magnetic flux converging toward the PIL. However, the interpretation of these observational elements are not straightforward: the magnetic field measurements are from photospheric data, which have the uncertainties we already discussed, and the lack of chromospheric and coronal field measurements prevents this study from clearly establishing whether the rope reaches coronal heights or not.
The region of emergence of the flux rope observed by Okamoto et al. (2009) was partially located in a region previously occupied by another filament. The authors could not definitively determine whether the mechanism of mass loading played any role; however, the presence of small reconnection sites suggested an interaction between the pre-existing and newly-emerged magnetic fields, causing an exchange of material. This is a case where the observations give no exhaustive answers to prominence formation.
The picture proposed by Lites et al. (2010) is different. Their data were interpreted as evidence for the emergence of a flux rope in an active region, with the filament plasma coming from the upper convection zone. The data suggested that the magnetic structure did not rise high in the corona. At the same time, several areas along the filament channel showed an absence of a strong vertical field, suggesting the absence of a coronal arcade that could prevent the rise of the filament structure. From this information they concluded that the filament was massive enough to keep the structure stable (see also Section 2.2.1).
As illustrated by the previous examples, the interpretation of the data is not straightforward and some conclusions remain controversial. At present there is no clear evidence that flux ropes can emerge fully and reach the coronal heights where filaments are found. Unfortunately, this is a limitation for properly testing and constraining filament formation models based on magnetic and plasma levitation.
The few examples presented here illustrate how the problems of filament magnetic-field and plasma formation are still far from being solved. One difficulty is the existence of only a few well-documented cases. For proper collection of useful information it is necessary to follow the target area for days, or even full disk passages, with multiple instruments. Today this may be possible with the SDO instruments, supported by ground-based instruments, but cannot be accomplished with the existing ground-based instrumentation alone.