He i D3 OBSERVATIONS OF THE 1984 MAY 22 M6.3 SOLAR FLARE

Hα, the most commonly used line in solar observations, is broad and optically thick and responds to a wide range of features in the solar chromosphere. In contrast, the He i D3 line, located at 5876 Å near the Na i doublet D1 and D2, is narrow and mostly optically thin and is sensitive to nonthermal excitation. For the latter reason, the D3 line is considered better suited to the study of higher energy phenomena than Hα. The most unique property of D3 is that it appears in absorption in surges, eruptive filaments, flare ejecta, and weak flares, and turns into emission only in intense flares. Compact and transient bright cores or strands are often observed in D3 during the impulsive phase of major flares, displaying a close spatiotemporal association with footpoint-like white light (WL) and hard X-ray (HXR) emission (e.g., Labonte 1979; Zirin 1980; Zirin & Neidig 1981; Feldman et al. 1983; Tanaka & Zirin 1985; Zirin & Tang 1990; Wang et al. 1996). D3 emission can reach about twice the photospheric intensity during flares; this result means that the medium has a high temperature (∼2 × 10⁴ K) and high density (∼10¹³ cm⁻³) flaring plasma. Therefore, D3 may provide a diagnostic of the main flare energy source in the low atmosphere, complementary to WL and HXRs, as well as serving as a sensitive probe for tracing the flare development. The simultaneity of D3 and HXR footpoint emission could indicate that accelerated electrons penetrate down to at least the low level, as indicated by the above density (Zirin 1988). Wang et al. (1996) observed during the initial HXR peak of an M2.7 flare that D3 emission only occurred at one of the HXR footpoints with a harder spectrum. Another X12 flare studied by Tanaka & Zirin (1985) started with multiple compact brightenings in D3 and WL, two of which grew to the main flare footpoints when the HXR spectrum hardened with a sharply increasing flux. The footpoints further developed into two ribbons in D3 above sunspots around the soft X-ray (SXR) peak, mainly due to heating from the hot thermal plasma that is conducted down to dense layers. The D3 line has been dubbed "a gold mine of information" on flares (Zirin 1988); nonetheless, it has been largely ignored in recent flare observations.

As one of the strongest He lines in the visible spectrum, the D3 line results from transitions between the 2p ³P and 3d ³D terms of the He i triplet (ortho-helium), which involves a metastable level about 20 eV above the ground state. One mechanism to populate the metastable level is that the coronal extreme ultraviolet (EUV) radiation ionizes neutral He atoms primarily in the singlet ground state at typical chromospheric conditions, and subsequent recombinations produce a larger population of He in the triplet systems, hence strengthening the D3 feature (e.g., Mauas et al. 2005; Centeno et al. 2008). This photoionization–recombination (PR) scenario is supported by observations of the bright D3 band in a narrow height range of 1000–2000 km above the solar limb; these observations are, however, not reported over coronal hole regions (Zirin 1975). We note that EUV photons only help in populating the triplet levels, which then subsequently radiate by scattering the photospheric emission. Thus, in order for D3 to turn into emission against the disk, direct collisional excitation (CE) must exceed the photospheric radiative excitation. It can be deduced from this requirement that at T ≈ 2 × 10⁴ K, the transition from absorption to emission of D3 in a flare would occur as the density of the heated atmosphere reaches N ≳ 5 × 10¹² cm⁻³ (Zirin 1988). Thus far, only a single event exhibiting this implied reversal of D3 intensity has been reported. It was the SOL1978-07-10T17:35⁵ M8 flare, in which a dark ring (circular ribbon) in D3 around a satellite spot in the preflare state immediately preceded two bright D3 emission kernels, which supposedly lay below the ring (Zirin 1980).

Intriguingly, such an evolution from absorption to emission is very reminiscent of the so-called "negative flares" on stars, in which a preflare dip in the continuum level, i.e., a ∼20% decrement in amplitude in a time period of seconds to minutes, is seen right before the continuum brightening (e.g., Hawley et al. 1995, and references therein). Henoux et al. (1990) termed such a phenomenon "black-light flares" (BLFs) and proposed that continuum BLFs could also happen on the Sun for some 20 s before WL flares (WLFs). The underlying mechanism could be that the nonthermal ionization of hydrogen by the bombardment of electron beams on a cool atmosphere first produces an increase of H⁻ opacity, and its diminutive effect is overtaken by the enhanced emissivity from a fully heated atmosphere near the flare maximum phase (Grinin 1983; Henoux et al. 1990; Ding & Fang 2000; Ding et al. 2003). Observations and studies of BLFs would certainly be important for solar and stellar flare physics, as further information could be obtained on the heating of the low atmosphere in the flaring process. Nevertheless, solar BLFs in the continuum are extremely rare. van Driel-Gesztelyi et al. (1994) surveyed all nine Yohkoh WLFs greater than GOES-class M6 but found no unambiguous evidence for the BLF association. A possible example presented by Henoux et al. (1990) is a negative intensity contrast up to ∼5% in the 5500 Å continuum a few minutes before the SOL1981-07-26T13:53 X3 WLF. Another candidate is the SOL2001-03-10T04:05 M6.7 WLF observed near the Ca ii 8542 Å line, for which Ding et al. (2003) showed a dip of ∼1%–2% in its early phase before the continuum emission.

Nonthermal electron beams cannot only lead to BLFs in the continuum, but also are able to drastically change spectral lines. Using non-LTE calculations, Ding et al. (2005) demonstrated that in the nonthermal case, collisional ionization followed by recombinations (CR) may significantly contribute to populating the He triplet levels corresponding to the He i 10830 Å line; this contribution acts in addition to the PR and CE mechanisms. Importantly, the influence of CR on the 10830 Å line, when compared to the case without the nonthermal effects, was modeled by the authors. In their result, the nonthermal effect can cause stronger absorption early in a flare and stronger emission at the flare maximum. We consider that such a transition from absorption to emission of spectral lines, if observed, could be regarded as a variant of BLFs analogous to those in the continuum. In this context, the aforementioned event of a preflare dark ring followed by bright kernels (Zirin 1980) turns out to be a very rare instance of D3 BLFs. To our knowledge, no study has been carried out about the nonthermal effects on the He i D3 line.

In this paper, we present a detailed analysis of the SOL1984-05-22T15:03 M6.3/2B flare, taking advantage of the unique D3 observation made at the Big Bear Solar Observatory (BBSO). The most prominent characteristics of this event include (1) the main D3 flare source first darkens in intensity and then changes to emission at the time of the HXR peak, making it an excellent candidate for a BLF and (2) in the late phase, two dark ribbons form in D3, mirroring their bright counterparts in Hα as well as He i 10830 Å. We also compare the radiated energy in D3 with the energy carried by electron beams at the flare peak time and discuss the event progression using Hα images. The outline of the paper is as follows: in Section 2, we describe the data sets and the reduction procedure. In Section 3, we present the main results of data analysis and discuss their implications. Major findings are summarized in Section 4.

The BBSO has a long tradition of flare observations at Hα wavelengths. In addition, the BBSO started in 1973 to take high-resolution images in the D3 line using 25 or 65 cm telescopes (Feldman et al. 1983). All of the BBSO data obtained before 1995 were recorded on 35 mm films, a large portion of which have recently been digitized (see Wang & Liu 2012 for a more complete description). The film data have also been publicly released,⁶ which opens the door for studying a wealth of historical solar phenomena. The present event (SOL1984-05-22T15:03 M6.3) is selected from our survey on D3 flare observations.

We mainly used the digitized partial-disk D3 and Hα images taken by the east and west benches, respectively, of the 25 cm telescope, with a pixel size of ∼015 and a cadence of ∼15 s. These images are capable of offering subarcsecond resolution when the seeing conditions at the BBSO are excellent (as in the later phase of the flare event under study). A non-linear conversion table supplied by the digitizer, which transforms the original 12 bit digitized data to 8 bit, was applied to deal with the nonlinear intensity response of films. The image alignment was implemented with sub-pixel precision and the intensity was normalized to that outside the flaring region in a quiet-Sun area. A de-stretching algorithm (Shine et al. 1994) using running references was applied in order to reduce the atmospheric distortion. We define the image contrast as (I − I₀)/I₀, where I is the intensity of the feature of interest and I₀ is the intensity of the same feature in the preflare state or that of the undisturbed quiet-Sun background. For the Hα image time series, we computed the intensity-weighted centroid positions of the flare ribbons and used the relative distance of source centroids to infer the progression of the flare (e.g., Liu et al. 2010). Additional auxiliary data used in our study include the digitized full-disk (∼14 pixel⁻¹), one minute cadence Hα images taken by the National Solar Observatory (NSO) at Sacramento Peak (SP) and the daily full-disk spectroheliogram observation at 10830 Å made by the NSO at Kitt Peak (KP) using its 512-channel magnetograph (Livingston et al. 1976).

The daily full-disk magnetic field structure of the photosphere in the line of sight (LOS) was also observed by the NSO/KP in Fe i 8688 Å. We registered the D3 and Hα images with the NSO/KP magnetogram by matching sunspot and plage areas, with an alignment accuracy estimated to ≲5''. To reveal the magnetic structure of the flaring region, we resorted to the potential field extrapolation in Cartesian coordinates.

X-ray observations were used to study the temporal evolution of the flare and understand the nature of the D3 emission. SXR and HXR time profiles were obtained with the GOES 5 satellite and the Hard X-Ray Burst Spectrometer (HXRBS; Orwig et al. 1980), respectively, the latter of which was on board the Solar Maximum Mission (SMM). We also utilized HXR spectra fitted to all 15 channels of the HXRBS covering the energy range of 24–400 keV; these spectra were computed by an automatic spectral fitting routine using a single power-law model (Kiplinger 1995). The derived spectral parameters allowed us to evaluate the nonthermal thick-target electron power above a certain low-energy cutoff (Brown 1971). In addition, the SXR images were available from the X-Ray Polychromator (XRP; Acton et al. 1980) on board the SMM.

In this section, we first provide a brief account of the magnetic field structure of the flaring region in the preflare state. We then concentrate on describing the characteristic flare activities in D3, HXR, as well as Hα wavelengths throughout the event. More dynamic detail can be seen in the accompanying animations in the online journal.⁷

3.1. Magnetic Field Structure

The active region (AR) of interest NOAA AR 4492 started to appear in the βγδ magnetic configuration from 1984 May 19 and showed growth until May 21. The AR decayed in size from May 22, when the present M6.3/2B flare occurred in the leading portion of the AR with a β configuration (see Figure 1(a)). A prominent property of magnetic field structure of this flaring region is that the central spot of positive polarity seems to be encircled by a surrounding negative field, forming a quasi-circular magnetic polarity inversion line (PIL; the dotted line in Figure 1). This circular PIL is well delineated by the circle-like Hα filament, which lies closely along the PIL (Figure 1(b)). Different from classical eruptions, this filament does not show a preflare activation and largely survived after the flare, as is clearly seen in the time-lapse Hα movies. This feature may indicate that the flare-associated magnetic reconnection could occur above the filament (e.g., Liu et al. 2007).

Zoom In Zoom Out Reset image size

Remarkably, the result of the potential field extrapolation model (Figure 1(a)) reveals dome-shaped closed fields (red) overarching the filament. The closed dome is enveloped by open field lines (blue) stemming from the surrounding regions. The overall structure is highly indicative of a fan-spine configuration, which is usually associated with a coronal null point. It has been found that three-dimensional (3D) magnetic reconnection could occur at such a null point (Lau & Finn 1990; Török et al. 2009), especially when pronounced asymmetry is present (as for this AR; see Figure 1(a); Pariat et al. 2010). It is also found that sometimes flares associated with such 3D reconnection may exhibit a central compact kernel situated within an extended circular flare ribbon (e.g., Masson et al. 2009). In fact, observational signatures of the present flare bear much resemblance to those described in this scenario, and we will discuss these signatures in more detail in Section 3.3 below.

3.2. Event Evolution in D3

3.2.1. A BLF

According to the GOES and HXRBS X-ray fluxes, the event started at ∼14:51 UT and peaked at 15:03 UT on 1984 May 22, and had four conspicuous HXR spikes p1–p4 (see Figure 2(a)). The initiation of the flare in D3 images is marked by a gradual darkening of the elongated main source S1 above the weak positive magnetic field region (cf. Figures 3(a)–(c) and (f)). It is notable that within this darkening source S1, two D3 kernels begin to brighten from ∼14:54:30 UT (Figures 3(c) and (d)), cotemporal with the first HXR spike p1. The two bright kernels then appear to extend in a direction parallel to the PIL, forming an emission strand lying within the dark halo-like structure at 14:57:05 UT (Figure 3(e)), which is simultaneous with the most intense spike p2 in HXRs from 24–400 keV. An intriguing observation is that the strand apparently reaches its peak brightness a little later at 14:57:20 UT (Figure 3(f)), at the time of a slightly harder HXR spectrum (Figure 2(c)). From ∼14:57:50 UT, S1 broadens and enhances, particularly in its northern end, and during the time period of the HXR spike p3, S1 spreads eastward into the sunspot region (Figures 3(g)–(i)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To more quantitatively assess the temporal relationship between D3 emission and flare energy release, we measure the mean contrast of the D3 source S1 in a boxed region (see Figure 3(f)) and draw its time profile in Figure 2(b) (black line). It can be unambiguously seen that S1 first darkens and dips in intensity, reaching a contrast amplitude of −5% in about 4 minutes from 14:52 to 14:56 UT despite the brightening kernels. It then quickly turns to positive contrast after being overtaken by the bright emission strand and peaks with an enhancement of ∼5%–10% concurrently with the HXR spikes p2 and p3 with hardened spectra. It is worth mentioning that the highest D3 intensity of S1 actually occurs at the time of the strongest HXR peak p2, as expected. This peak D3 intensity shows a contrast up to 100%, as previously observed in other flares of similar magnitude (e.g., Zirin 1980). This result is clear in the time profile of averaged contrast that shows only contrasts greater than 60% (red line in Figure 2(b)). Here, the most significant result is the intensity reversal of S1 from absorption to emission, which qualifies this event as a rarely observed BLF. For the first time, this observation provides direct evidence for the unique property of BLFs in the D3 line.

We note that the absorption feature in D3 was generally interpreted as being due to heating of the dense layer by thermal conduction (Zirin 1988). This hypothesis could explain the darkening of S1 below the BLF level in the later thermal phase (after ∼15:02 UT). More importantly, D3 and HXR emission in the earlier impulsive phase are simultaneous with a good correlation, which implies electron beam heating, as suggested before (Feldman et al. 1983; Zirin 1988). In other words, D3 emitting sources in the flare impulsive phase could indicate the location of electron precipitation. As further evidence of this hypothesis, a compact and weak D3 bright kernel S2 is located at the umbral region of the western sunspot with negative polarity. S2 becomes visible after ∼14:56 UT, peaks at the HXR spike p1 around 14:57 UT with a contrast of ∼70% against the dark sunspot, and propagates southward after ∼14:58 UT (Figures 3(d)–(i)). We speculate that S1 and S2 are a pair of conjugate sources at the feet of reconnecting flare loops, which are indicated by the loop-like structure in SXRs crossing over the PIL (contours in Figure 3(m)). As S2 lies within the stronger umbral field region, its compactness in D3 could be due to the magnetic mirroring effect (e.g., Kundu et al. 1995). We also note that the last HXR spike p4 at 15:00 UT is reflected in D3 as a newly formed source S3 in the south (Figure 3(j)). Flaring loops could connect S2 and S3 at this stage, as imaged in SXRs at a later time (Figure 3(o)).

3.2.2. Bright Core–Dark Halo Structure

We placed two slits perpendicular to the elongated source S1, as shown in Figure 3; the intensity profiles along these slits are plotted in Figure 4 in order to more closely examine the dark absorption and bright emission features in D3 and their temporal evolution. Gaussian fits were applied to estimate the characteristic sizes and peak positions of the intensity profiles.

Zoom In Zoom Out Reset image size

Slit 1 crosses the maximum of the D3 kernel in Figure 3(d). The intensity profile along it already displays a discernible darkening about eight minutes before the flare (gray line), possibly implying the occurrence of preflare heating. After the flare onset at 14:52:20 UT, the darkening contrast deepens and shifts toward the PIL (green line). The later bright emission kernel at 14:56:35 (blue line) and 14:57:20 UT (red line) peaks ∼4'' to the east. However, the darkening contrasts on either side of the emission peak both deepen consecutively with a step of ∼0.05%. Furthermore, the western darkening keeps moving toward the PIL by ∼4''. Additionally, while the eastern darkening broadens, the western darkening narrows at the HXR peak time (red line) to reach the largest contrast of about −3%. It is reasonable to surmise that the preference of D3 darkenings toward the PIL is caused by the enhanced density near the filament lying along the PIL region (see Figure 3(k)), which provides a favorable condition for D3 line formation. These results thus clearly illustrate a D3 flare source structure comprising a low-lying bright emission core (due to nonthermal beam heating) encompassed by darkening halos (due to thermal conduction heating). Consistent with this picture, the Hα intensity profile along the slit appears, in contrast, as a broad bump (the central part is saturated) with about twice the FWHM width, straddling the D3 bright core and dark halos. It is also appealing to suggest that the bright core-dark halo source structure in D3 is analogous to the core-halo morphology of WLF sources in the photospheric level (Xu et al. 2006), where both the core and halo are bright and might be produced by direct electron heating and chromospheric back-warming, respectively.

Slit 2 crosses the maximum of the D3 strand S1 around the HXR peak time (Figure 3(f)). Its central region first darkens (green line), then reverts back (blue and red lines) to become a very intense emission core peaking at 100% contrast while with a similar FWHM compared to the case of slit 1. It does not seem that a dark halo developed in the east of the core; nevertheless, a darkening region is similarly built up in the western region close to the PIL/filament.

3.2.3. Flare Energy Release

We further chose flare peak p2 in D3 and HXRs, which is well defined both temporally and spatially, to roughly evaluate the energy budget. The underlying assumption is that the D3 line emission could be optically thick in flares (Svestka 1976; Zirin 1988), hence the gas radiates with a source function close to the Planck function for the corresponding temperature. In the canonical flare model (e.g., Wang et al. 1998), the intensity contrast enhancement of a flare is given by:

$$\begin{equation} \frac{\Delta I}{I}=\frac{B(T+\Delta T, \lambda)-B(T, \lambda)}{B(T, \lambda)} \, \end{equation} \tag{ 1 }$$

where B is the Planck function and ΔT is the temperature perturbation due to flare heating. The flare radiative loss flux can then be written as:

$$\begin{equation} \Delta L=4\sigma T^3 \Delta T \ \rm erg \ cm^{-2} \ s^{-1} \, \end{equation} \tag{ 2 }$$

where σ is the Stefan–Boltzmann constant (Najita & Orrall 1970). For the n ≈ 10¹³ cm⁻³ that is required for D3 emission, T is about 6400 K according to the VAL-F atmospheric model (Vernazza et al. 1981). We use this model because it was constructed for the magnetic network regions suitable for the present flare, hence is better than other models for the quiet-Sun conditions. If we integrate ΔL over the source S1 and adopt a duration of 30 s, the radiative energy in D3 during peak p2 is estimated to be about 1.3 × 10³⁰ erg. This energy is comparable to the nonthermal energy E_nonth ≈ 1.2 × 10³⁰ erg released by the thick-target electrons greater than 20 keV around peak p2 (Figure 2(c)). Obviously, such a comparison is meaningful provided that electrons with initial energies of 20 keV and above can penetrate to densities of ∼10¹³ cm⁻³. A simplified equation relating the observed photon energy to the column density N of a certain layer, into which the parent electrons of energy E can precipitate, is:

$$\begin{equation} N=10^{20} \ {\rm cm}^{-2} \left(\frac{\epsilon }{20 \ {\rm keV}}\right)^2 \end{equation} \tag{ 3 }$$

(Brown et al. 2002). Since E ≈ holds for a steep electron spectrum (δ ≈ 5 around peak p2), the above condition can be shown to be satisfied at a scale height of ∼200 km. Electrons with lower energies may be stopped higher up and produce emission in other lines (e.g., Hα). Therefore, the similarity between the amount of energy released in the D3 emission and that carried by nonthermal electrons bolsters the inference that the D3 line emission during the impulsive phase is capable of revealing the electron precipitation sites. Furthermore, the D3 line has the advantage of being a density diagnostic, as compared to HXRs.

3.3. Event Evolution in Hα

3.4. Dark and Bight Circular Ribbons

This flare is a long-duration event, since after the nonthermal impulsive phase, the GOES SXR flux decayed to its preflare level in about two hours until around 17:00 UT. The entire evolution of the event is clearly delineated by the distance-time profiles in D3 and Hα along slit 3 (Figures 5(e) and (f)), which lies perpendicular to the PIL and both the outer circular and inner compact ribbons (the white line in Figure 5). Multiwavelength profiles along the slit at a specific instance are also shown in Figure 6. It is obvious that after expanding across the network field during the impulsive phase (cf. Figures 5(f) and 6(d)), the outer ribbon stops and remains almost fixed. The inner ribbon, however, keeps propagating northwestward during the late gradual phase at an average speed of ∼4 km s⁻¹ (also see the movies). Strikingly, identical dynamics are also exhibited by the dark ribbons in D3 (Figure 5(e)). The continuous ribbon motion leads us to the conjecture that there might be persistent, moderate magnetic reconnection providing continuous injection and deposition of energy into the low atmosphere in the flare late phase. In such a stage, the energy transport mechanism is most probably thermal conduction (e.g., Czaykowska et al. 2001).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Around 15:16 UT, the NSO/KP obtained a spectroheliogram in He i 10830 Å. These data provide an opportunity to compare flare signatures at different heights spanning from the lower to the upper chromosphere using D3, Hα, and 10830 Å images (Figures 5(a)–(c)). Several features are described as follows. (1) The circular filament, especially its southern portion, appears dark in all wavelengths. This result is evidence of heating of the dense filament materials. The jet-like disruptive activity of the southern segment around this time may be associated with the enhanced darkening in D3. (2) Postflare loops, as indicated in Figure 5 by the arrow, are dark in D3 and 10830 Å while bright in Hα, consistent with the extrapolation model (thick lines in Figure 5(d)). This result may indicate plasma in the postflare arcades at a temperature of ∼2 × 10⁴ K and a density of 10¹⁰ cm⁻³ (Zirin 1980). (3) Most interestingly, the double flare ribbons are bright in Hα and 10830 Å while dark in D3, which is distinctly seen as the good correspondence between the peaks in Hα/10830 Å and valleys in D3 of their profiles along slit 3 (Figures 6(a)–(c)). Since both the 10830 Å and D3 lines are produced by ortho-helium, this result seems to imply that the 10830 Å line is more easily excited than D3 and that the 10830 Å line produces emission even by thermal conduction.

We explore further the dark D3 and bright Hα circular ribbons around 16:00 UT when the seeing turned excellent. It is evident that the dark ribbons in D3 and the bright ribbons in Hα demonstrate a nearly exact uniformity (cf. Figures 5(g) and (h)). Such a correlation shows up as a negative trend of ribbon pixels, as fit by the white-dashed line in the scatter plot of Hα/D3 contrast in Figure 5(i). Since the Hα emission is largely not saturated at this time, it points to a ∼20% D3 darkening at a ∼300% Hα enhancement level, consistent with the results in Figure 4. Dark flare ribbons in D3 are rarely observed. As a comparison, the aforementioned dark D3 ring observed by Zirin (1980) has a much smaller (∼$\frac{1}{7}$) spatial scale and exists for only a short period of time (∼5 minutes) in the early stage of a surge. Zirin (1980) interpreted the dark circular D3 ribbon as the projection of a shell of dense (∼10¹⁰ cm⁻³) and hot (∼2 × 10⁴ K) gas. Noticeably, this vision of the shell is well in line with the dome-like fan surface as we speculate for this event.

In this paper, we have presented a rare observation of a major flare in the He i D3 line, taking advantage of the recently-digitized, high spatiotemporal resolution BBSO film data. We also compared the radiative energy in D3 at the flare peak time with the nonthermal electron energy, and discussed the event evolution with the aid of Hα images and the potential field extrapolation model. Our results can be summarized as follows.

• 1.  
  The flare initiation is characterized by the development of an elongated darkening source S1 in D3, which reaches a negative contrast of ∼5% in about four minutes (Figures 2 and 3). Within this dark source, emission kernels are seen to brighten with the first minor HXR peak. Then, simultaneous with the main HXR peak p2, they evolve into a narrow bright strand lying within the dark halo, reverting its overall contrast to positive enhancement of ∼5%. This intensity reversal of S1 in D3 establishes this event as a rare BLF, which is highly reminiscent of those proposed in the continuum (Henoux et al. 1990).
• 2.  
  The main D3 source S1 displays a bright core surrounded by dark halos (Figures 3 and 4). The bright emission core and dark absorption halos are thought to be caused by nonthermal electron heating and thermal conduction heating, respectively. This bright core-dark halo structure might be a unique characteristic of flare sources in D3 and is analogous to the bright core-halo morphology of WLFs (Xu et al. 2006).
• 3.  
  The flare radiative energy loss in D3 during the ∼30 s duration of the main HXR peak is about 10³⁰ erg, which is comparable to the deposited energy of nonthermal electrons with initial energies higher than 20 keV. Together with the good correlation between the D3 and HXR time profiles (Figure 2), these results strongly suggest that D3 emission is an excellent tracer of electron precipitation in the low atmosphere during the flare impulsive phase.
• 4.  
  The flare disturbance generally propagates sequentially along the PIL and ends up with a dark circular ribbon in D3 mirroring the bright ribbons in Hα and He 10830 Å (Figures 3 and 5). The dark D3 ribbons in the late gradual phase indicates well the thermal conduction heating. The non-standard motions of flare ribbons and the dome-shaped magnetic structure of the flaring region (Figure 1) strongly suggest 3D magnetic reconnection in a fan-spine topology (e.g., Wang & Liu 2012).

The mechanism for the solar continuum BLFs is different from that for the solar BLFs observed at He i D3 and 10830 Å. In the former mechanism, the low atmosphere first responds to the precipitation of flare electrons with increased H⁻ opacity for a few tens of seconds at the start of the main HXR burst (Henoux et al. 1990). In the latter mechanism, the He line turns from absorption into emission when the He triplet levels are populated due to CE (for D3; Zirin 1988) and additional CR (for 10830 Å; Ding et al. 2005) processes. Unlike continuum BLFs that are proposed to occur before WLFs, the present D3 BLF coincides with the first small HXR peak possibly in a precursor stage before the main HXR bursts. To our knowledge, solar BLFs in the continuum and 10830 Å have not yet been unambiguously observed.

A comprehensive study of the current event is unavoidably hampered by the limited quality of the aging films. We recall that flare observations at the D3 wavelength had been recognized as an important probe of the flare heating mechanism. A full understanding of the D3 signatures may also help resolve the puzzle of WLF emission in the deep photosphere (Zirin 1988). Therefore, new D3 flare observations using modern instruments and additional modeling, especially of the nonthermal effects on the D3 line, are highly desired. These efforts will shed new light on the impacts of flares in the low atmosphere.

This study is dedicated to Professor Harold Zirin, the founder of the Big Bear Solar Observatory, who passed away on 2012 January 3. We are indebted to the BBSO staff for their tremendous efforts in obtaining the D3 and Hα observations on film. We thank the referee for helpful comments. Digitization of the BBSO as well as the NSO solar film images were carried out by the Space Weather Research Lab of the NJIT, with support from NSF grant AGS 0849453 and NASA grant NNX11AC05G. The LOS magnetogram and the He 10830 Å spectroheliogram used here were produced by the NSO/Kitt Peak. X-ray data were obtained by SMM, a NASA mission managed by GSFC, and by GOES, a NOAA/NASA project. C.L., Y.X., N.D., J.Z., and H.W. were also supported by NASA under grants NNX13AF76G, NNX13AG13G, and NNX11AO70G, and by NSF under grants AGS 1153424 and AGS 0839216. J.L. was supported by the international scholarship of Kyung Hee University and NASA grant NNX11AB49G. D.P.C. was partially supported by NSF grant AGS 0548260.