First Detection of Prominence Material Embedded within a 2 × 10⁶ K CME Front Streaming away at 100–1500 km s⁻¹ in the Solar Corona

Prominences are considered to be one of the visual highlights of a total solar eclipse, with their characteristic bright red color due to Hα emission and their complex dense structures lying low in the corona. Prior to space-based measurements, catalogs of eruptive prominences produced from Hα observations showed that their speeds increased with heights to a maximum recorded value of 1280 km s⁻¹ at a heliocentric distance of 3 R_s (Kleczek 1964). (All distances given in this Letter are heliocentric.) In addition to Hα, emission from prominences is characteristic of neutrals and singly or doubly ionized elements, typical of ${10}^{4}\mbox{--}{10}^{5}\,{\rm{K}}$ chromospheric material (Tandberg-Hanssen 1995). Their ionization state is in stark contrast to that of the surrounding highly ionized coronal plasma at temperatures of a few 10⁶ K. Furthermore, multi-wavelength eclipse observations have amply demonstrated that prominences are found to be always enshrouded by $2\times {10}^{6}\,{\rm{K}}$ plasma in the corona (Habbal et al. 2010b).

With the advent of space-based observations, the occasional disruption of prominences was found to trigger gigantic eruptions in the overlying coronal material, in the form of bubble- or tornado-shaped fronts, known as coronal mass ejections (CMEs). The erupting prominences invariably formed their bright core (House et al. 1981; Illing & Hundhausen 1985; Gopalswamy et al. 2003; Chen 2011). Reported CME speeds, of a few hundred to 2000 km s⁻¹ within a few R_s (Yashiro et al. 2004; Yurchyshyn et al. 2005), were comparable to the earlier reported speeds of eruptive prominences (Kleczek 1964). Recently, Howard (2015a, 2015b) and Wood et al. (2016) found a few "rare" examples of very bright CME-associated prominences that were imaged out to 1 au with the heliospheric imagers on the two Solar Terrestrial Relations Observatory (STEREO) spacecraft (Kaiser et al. 2008).

The lack of spectroscopic capability in these coronagraphic data, however, is such that no information regarding the electronic and ionic temperature of CMEs, and the changes, if any, in the ionization state of any accompanying prominence material can be obtained. The first observations to yield any thermodynamic information were those from the extreme-ultraviolet spectroheliometer on the Apollo Telescope Mount on Skylab in the early 1970s (Reeves et al. 1977). Schmahl & Hildner (1977) reported an eruptive event whereby 10% of the filamentary prominence material maintained a temperature of $2\times {10}^{4}\,{\rm{K}}$ as it rose in the corona up to the 3 R_s limit of the field of view. The next opportunity materialized with the Ultraviolet Coronagraph Spectrometer (UVCS) on SOHO (Kohl et al. 1995). Since the lowest slit position of the spectrometer was at 1.5 R_s, these observations were often combined with contemporaneous extreme-ultraviolet observations, primarily from SOHO/EIT, to follow the origin and dynamics of an eruptive prominence–CME system starting from the solar surface outward. The main result to emerge from these studies was the helical nature of strands at temperatures ranging from $3\times {10}^{4}$ to $3\times {10}^{5}\,{\rm{K}}$, lending evidence for the survival of some chromospheric prominence material out to a few R_s (Ciaravella et al. 1997, 2000; Akmal et al. 2001; Raymond 2002; Raymond & Ciaravella 2004; Landi et al. 2010; Giordano et al. 2013; Lee & Raymond 2014).

The survival of a fraction of cool prominence material with its expansion into interplanetary space was demonstrated by in situ measurements. The most compelling evidence was the detection of enhanced singly ionized He (He⁺; Gosling et al. 1980), and unusually low charge states, e.g., O⁵⁺, Fe⁵⁺ (Burlaga et al. 1998); O³⁺, N³⁺, C²⁺ (Gloeckler et al. 1999); C²⁺, O²⁺, Fe⁴⁺ (Lepri & Zurbuchen 2010). These low charge states were found to be contemporaneous with measurements of high charge states of the multi-million degree CME plasma and have thus been interpreted as remnants of prominence material.

In this Letter, we report on the first comprehensive spectral observations obtained during the total solar eclipse of 2015 March 20, which led to the serendipitous discovery of a large number of spatially discrete Doppler-shifted fragments of Fe xiv 530.3 nm (i.e., Fe⁺¹³ ions) emission, scattered across a projected area of 2.5 $\times 1.3\,{R}_{{\rm{s}}}^{2}$ in the plane of the sky off the east limb of the Sun, with speeds ranging from under 100 to over 1500 km s⁻¹. Embedded within approximately 10% of them was Doppler-shifted emission from neutrals and singly ionized elements of Cr, He, Fe, and Mg. These observations are interpreted as evidence of filaments forming a large $2\times {10}^{6}\,{\rm{K}}$ CME front, enshrouding filaments of cool prominence material that had preserved its original ionic composition.

To capture the spectral composition of coronal structures, a dual-channel partially multiplexed two-dimensional imaging spectrometer (PaMIS) was designed and built for eclipse observations (see Figure 1). The two channels were centered on the Fe xi 789.2 nm and Fe xiv 530.3 nm lines, each with a 70 nm bandpass, separated by specially designed bandpass filters. The choice of these spectral lines was based on earlier multi-wavelength eclipse observations that showed that coronal structures were essentially dominated by two (electron) temperature regimes $1\times {10}^{6}\,{\rm{K}}$ and $2\times {10}^{6}\,{\rm{K}}$, corresponding to the peak ionization temperatures of Fe xi and Fe xiv, respectively (Habbal et al. 2010a).

Zoom In Zoom Out Reset image size

The entrance slit of the spectrometer was produced by laser ablation of a 30 μm × 2 cm rectangular part from an Al-coated mirror. The rest of the mirror surface reflected the image of the corona onto a video camera, such that the position of the slit, relative to the Moon-occulted Sun, could be monitored as a function of time. The optical design was such that the Sun diameter covered 25% of the usable 8 R_s length of the slit.

As the resolution of such an instrument is proportional to the grating order and inversely proportional to the slit width, PaMIS was operated in high (approximately 40th) order. The limited bandpass of 70 nm in each of the two channels enabled a relatively easy disentanglement of the overlapping orders.

The fact that several orders were recorded at each slit position during its operation yielded a valuable reference to check against any artifacts. For example, while He i in 37th order overlapped Fe xiv in 41st order, they were separated in the He i 38th order and the Fe xiv 42nd order. Furthermore, a small transmission beyond the long wavelength end enabled the detection of very strong lines such as He i 587.6 nm in the Fe xiv 530.3 nm channel and Fe xiii 1074.7 nm in the Fe xi 789.2 nm channel.

Using an effective slit width of 30 μm, a wavelength resolution, $\lambda /{\rm{\Delta }}\lambda \approx {\rm{15,000}}$, enabled the recording of the (non-relativistic) Doppler shift ${\rm{\Delta }}{\lambda }_{D}$, of an emission line at wavelength λ, with a velocity accuracy along the line of sight of ${v}_{{\rm{D}}}=({\rm{\Delta }}\lambda /\lambda )\times c=20\,\mathrm{km}\,{{\rm{s}}}^{-1}$ (where c is the speed of light).

The spectroscopic observations presented here were acquired during the total solar eclipse of 2015 March 20, from the old Auroral Station in Adventdalen, operated by the University Centre in Svalbard (UNIS), about 10 km away from the small town of Longyearbyen on the island of Svalbard, Norway. The coordinates were N 78°12' 091, E 15° 49' 445 at an altitude of 15 m. Totality, or second contact, started at 10:10:47 UT and ended at third contact at 10:13:14 UT, thus lasting 2 minutes 27 s.

Starting near central meridian at the beginning of totality, the motion of the tracking mount was disabled. The Sun thus drifted across the field of view, enabling the slit of the spectrometer to scan the corona eastward across a distance of 2.5 R_s throughout its 2.5 minute duration of totality. The download time from the detector was 2 s, thus extending the time difference between adjacent exposures. The exposure time at each slit position, starting at position 346 (1st position, almost at central meridian) to 383, was 1 s (or effectively 3 s including download time), yielding a spatial resolution of 45 arcsec. The exposure time was then increased to 2 s (or effectively 4 s) for the remainder of the time, when 14 spectra (slit positions 384–397) were obtained with a spatial resolution of 60 arcsec.

Emission from any spectral line, within the 70 nm bandpass of the Fe xiv channel, was detected in at least two orders, as shown in the example of Figure 2. Calibration of this channel was achieved by using the 532.6 nm emission from a doubled Nd:YAG laser, as well as a Fraunhofer absorption spectrum of the solar disk. The low number (approximately 30) of emission lines present in each channel facilitated their wavelength assignment.

Zoom In Zoom Out Reset image size

The fortuitous presence of prominences at the solar limb, when intercepted by the slit, and the fortuitous presence of chromospheric emission lines within the bandpass of both channels, served as an invaluable wavelength reference for any chromospheric emission in the corona. As shown in Figure 2, for two such slit positions, 354 and 368, the overlay of the coronal emission spectrum with the Fraunhofer absorption spectrum matched perfectly: the very narrow Cr i and Mg i triplets and Fe ii lines, and other weaker Fe i lines (not labeled), characteristic of chromospheric emission, could thus be readily identified in emission from the corona and absorption in the Fraunhofer sky spectrum. The coronal emission lines of Fe xiv 530.3 nm and Ni xiii 511.58 nm, not present in the Fraunhofer absorption spectrum, were identified by their larger spatial extent along the slit.

The spectral order number, n, of a line was—to first approximation—obtained by counting the orders of the spectrum of a green laser 532 nm line, starting from order 0 (reflection) to the order of maximum intensity. Accurate determination of the grating order was then achieved by comparing the apparent position of the Fe xiv line in relation to the position of the chromospheric Fe ii and Mg i triplet lines (λ = 516.9, 516.7, 517.3, 518.4 nm), taking into account the different order number. (Lines with reduced wavelength $\lambda * \,=\,\lambda$/n, show up at the same pixel position.) Only the assignment of 41st (left), 42nd (middle), and 43rd (right) order to the Cr and Mg triplet lines and 42nd (middle) and 43rd order (right) to Fe xiv gave the correct position of the emission lines with respect to each other.

In the Fe xi channel, emission from an O i triplet at 777.19, 777.42, and 777.54 nm also appeared at the two slit positions that intercepted the prominences at the limb. In their vicinity, the Fe xi intensity dropped significantly, while emission from Fe xiii 1074.7 nm appeared. This is identical to what had been observed in earlier multi-wavelength eclipse images (see Habbal et al. 2010b). A similar calibration procedure was followed for the Fe xi channel using the position of the O i triplet in relation to those of Fe xi and Fe xiii. Unfortunately, the choice of exposure times for the Fe xi channel proved to be too short to acquire meaningful data past 1.2 R_s because of the low quantum efficiency of the detector at that wavelength. In what follows, the results will therefore pertain only to the Fe xiv channel.

Shown in Figure 3 are examples of spectra taken in the Fe xiv channel at three different slit positions, 354 (A) when part of the slit intercepted the occulted solar disk, 378 (B) at 1.4 R_s, and 395 (C) at 2.3 R_s. The slit position in each panel is indicated by a dashed white line, with its position overlaying the corresponding white-light image of the corona taken during totality by M. Druckmüller, and processed by him, to highlight the finest details of coronal structures. To the left of the slit is the two-dimensional spectrum in 42nd order for Fe xiv. As noted earlier, and as can be clearly seen here, slit position 354 (A) intercepted a prominence at the limb, where emission from neutral Fe i, He i, and the Mg i and Cr i triplets was detected, in addition to a ubiquitous background Fe xiv emission. (Note that the overlaid spectrum in panel (A) has been shifted slightly from the slit position (dashed line), so as to expose the intercepted prominence.) Further out in the corona, such as in positions 378 (B) and 395 (C), Doppler redshifts appeared in Fe xiv in some sections along the slit. These sections will be referred to as plasmoids in what follows. They appeared superimposed on the white-light emission caused by electron scattering shown by the white bands in Figure 3. Some of these Fe xiv Doppler-shifted plasmoids were accompanied by He i, Fe ii, and Mg i triplet emission, indicative of the presence of chromospheric material from prominences embedded within them.

Zoom In Zoom Out Reset image size

Details of multiple Doppler-shifted plasmoids detected along any slit position are shown in the two-dimensional spectrum of Figure 4, for slit position 378 (the same as in Figure 3(B)), in the 41st (left) and 42nd (right) order for Fe xiv. Overlaid in the middle are five line spectra corresponding to the details of the spectra identified by the rectangular windows. B refers to the background Fe xiv emission. The other peaks refer to the Doppler-shifted ones. Emission from the Doppler-shifted Fe xiv (labeled Fe) appears in all plasmoids, and is labeled ${\mathrm{Fe}}_{C}$ when associated with cool inclusions. This is evident from Mg i, He i, and Fe ii emission, which appear in the top two panels. Note that these have higher speeds than the plasmoid without cool material. (The line width pointing to ${\mathrm{Fe}}_{C}$ is thicker than for Mg i to show that it was a stronger emission.) The designation Fe (2) in the third and fourth panels refers to an unresolved pair of Doppler-shifted parcels. In the lower panel, where the signal is much weaker, a double-peaked Doppler-shifted Fe xiv emission, compared to the top two spectra, is clearly visible. The blue spectrum superimposed on the top spectrum is from slit position 354 (see Figure 2) when it intercepted the prominence at the limb. It is used here for comparison and as a reference.

Zoom In Zoom Out Reset image size

The 51 spectra thus acquired throughout the 2.5 minutes of totality (and a couple past totality where the coronal emission could still be detected), were comparable in complexity to the example shown in Figure 4. They yielded approximately 200 Fe xiv Doppler-shifted plasmoids, with speeds ranging from under 100 to over $1500$ km s⁻¹, across the 2.5 × 1.5 ${R}_{{\rm{s}}}^{2}$ areal coverage spanned by the slit during that time period. Approximately 10% of them enshrouded chromospheric material, moving in tandem, and having the same spatial extent. Very few examples of faint Doppler-blueshifted lines were also found in these spectra.

There are several quantitative inferences that can be made from these spectra. We will focus here on the most reliable ones: the Doppler shifts and Doppler broadening.

5.1. Doppler Shifts

5.2. Doppler Broadening

The measured widths of the spectral lines can be used to infer an effective temperature, T_eff, for the different ions. Assuming a Boltzmann distribution, T_eff is given by (see, e.g., Esser et al. 1999)

$$\begin{eqnarray*}&&{T}_{\mathrm{eff}}=\displaystyle \frac{{m}_{i}}{2k}\displaystyle \frac{{c}^{2}{\rm{\Delta }}{\lambda }^{2}}{{\lambda }^{2}}={T}_{i}+\displaystyle \frac{{m}_{i}}{2k}\langle \delta {v}^{2}\rangle ,\end{eqnarray*}$$

where T_i is the ion temperature and $\sqrt{\langle \delta {v}^{2}\rangle }$ is the broadening due to any wave or turbulent motion along the line of sight, m_i is the ion mass, and k is the Boltzmann constant. The measured line widths of Fe xiv yielded an effective temperature ${T}_{\mathrm{eff}}=(3.7\pm 0.3)\times {10}^{6}\,{\rm{K}}$, with no significant change as a function of radial distance. We also found that the line width, ${\rm{\Delta }}\lambda$, of both the background and Doppler-shifted Fe xiv emission were comparable. We note that T_i should be larger than the peak ionization of Fe xiv, or electron temperature ${T}_{{\rm{e}}}=2\times {10}^{6}\,{\rm{K}}$. Since the contribution from wave motions increases with radial distance (see, e.g., Esser et al. 1999), it is then likely that T_i decreases as a function of distance; hence, the ions do not experience any heating with their expansion within the CME front. In fact, T_i is expected to decrease with supersonic expansion, as is the case with this fast CME front. While this argument applies to the CME plasma, it does not necessarily account for the fact that the Doppler broadening of the background, stationary Fe xiv was comparable, unless the contribution from non-thermal motions was the same inside and outside the CME front.

While chromospheric lines are also thermally broadened, analysis of the width of the Fe ii and Mg i spectral lines could not be achieved with reliable accuracy, since the resolution of the spectrometer is 0.03–0.04 nm. From their mass and their measured line width, and taking into account the broadening of the Mg i lines in the Fraunhofer spectrum, a maximum temperature of 10⁵ K has been derived. On the other hand, the He i line is much more intense, and exhibits a much larger Doppler broadening, because of its low mass, than atoms and ions with higher masses at the same temperature. The width of the He i line yields an average ionic temperature of $(1.8\pm 0.2)\,\times \,{10}^{5}\,{\rm{K}}$. This value can be considered to be the maximum temperature of the cool material embedded in the CME front. We note that this maximum temperature is larger than the published electron temperature of $6\times {10}^{4}\,{\rm{K}}$ (namely, the peak formation temperature of C iii 97.7 nm) reported by Akmal et al. (2001), but closer to the $1.2\times {10}^{5}\,{\rm{K}}$ value given by Landi et al. (2010), for the cool prominence cores they observed.

The observations presented here are not the first to show that eruptive prominence material escapes within a CME front while maintaining its ionic composition. As noted earlier, Howard (2015a, 2015b) and Wood et al. (2016) were able to follow the morphological evolution of a prominence from the Sun into interplanetary space, albeit without spectroscopic information. Spectroscopic observations with UVCS provided evidence of the presence of a cool core within a CME front (e.g., Ciaravella et al. 1997, 2000; Akmal et al. 2001; Raymond 2002; Raymond & Ciaravella 2004; Landi et al. 2010; Giordano et al. 2013; Lee & Raymond 2014). Interplanetary space measurements also reported the detection of singly ionized and low charge states of a number of ions, interpreted as the remnants of prominences within the CME front (e.g., Gosling et al. 1980; Burlaga et al. 1998; Gloeckler et al. 1999; Lepri & Zurbuchen 2010). While the observed 10% fraction of CME plasmoids accompanied by cool prominence material is higher than the 4% reported in interplanetary space measurements (Lepri & Zurbuchen 2010), it is not clear from these sole eclipse observations whether this is the norm, or whether this was an exception. Nevertheless, the relatively small, yet non-negligible, fraction of prominence material, entrained within a CME front, is consistent with the fractions reported from in situ measurements (e.g., Lepri & Zurbuchen) and from the remote sensing ultraviolet coronagraphic observations by Lee & Raymond (2014). On the other hand, one has to keep in mind that studies that suggest that eruptive prominence material can get heated up pertain only to the fate of the fraction of the prominence material that finds itself falling back toward the Sun (e.g., Filippov & Koutchmy 2002; Innes et al. 2016).

There are several novel and unique aspects of the spectral measurements in the visible wavelength range acquired with PAMIS: (1) the unambiguous identification of the emission spectrum of both hot (multi-10⁶ K) and cool (10⁴–10⁵ K) plasmas, (2) the inference of line of sight Doppler shifts, and (3) the very short exposure times of 1–2 s, which enabled a large areal coverage of the corona. These features are particularly important for following the fate of eruptive material as it expands away from the Sun. Earlier breakthrough spectroscopic observations of the corona had been acquired in the ultraviolet wavelength range with UVCS (Kohl et al. 1995). While resonant excitation, also referred to as "pumping," in the presence of large outflows, was a powerful diagnostic of their radial, rather than their line of sight component, it was applicable for only a few coincidental narrow wavelength ranges (see Raymond & Ciaravella 2004). In contrast, the design of PAMIS enabled the line of sight Doppler inference for both coronal (Fe xiv and Ni xiii) and chromospheric (Cr i, Fe i and Fe ii, He i, and Mg i) emission, with no instrumental limitations. From observations of the Doppler shifts, hot plasmoids with cool inclusions, traveling in tandem at speeds ranging from under 100 to over 1500 km s⁻¹, could be directly extracted from the spectra. From the thermal broadening of the spectral lines, the effective temperatures of the highly ionized species, e.g., Fe xiv and Ni xiii, and to a lesser extent those of the highly excited He triplet states, could also be determined. It is unfortunate that UVCS was no longer operational during the 2015 eclipse. With its diagnostic potential, it would have yielded the inference of the radial component of the outflow, thus complementing the line of sight observations in the visible and a diagnostic of the ionic temperature of the cool plasmas also complementing those reported here.

In conclusion, these first-ever spectroscopic observations in the visible wavelength range, acquired in 2.5 minutes during totality, spanning an uninterrupted projected area of $2.5\times 1.5\,{R}_{{\rm{s}}}^{2}$, starting from the solar surface, were enabled by the novel design of the spectrometer and its operation in higher order. They have contributed to the long-standing enigma of the fate of eruptive prominence material accompanying a CME. They showed, without a doubt, that cool and dense prominence material can survive the hot coronal environment, and its expansion into interplanetary space, enshrouded by segments of hot CME envelopes. Their presence and survival place serious new constraints on CME models (e.g., Landi et al. 2010): not only do models have to account for a source of energy to heat and accelerate a CME front, but they now have to also account for the survival of its cool core and the cool shreds embedded within it. These first-ever eclipse observations would not have been possible with any existing ground- or space-based observatories because of the absence of spectroscopic capabilities in the distance range covered by the eclipse observations. They underscore the unique opportunities provided by total solar eclipse observing conditions for in-depth exploration of the physical processes underlying the chemical evolution of the coronal plasma, and its dynamics, especially in conjunction with explosive events such as prominence eruptions and CMEs.

This work was supported by NASA grant NNX13AG11G and NSF grants AGS-1144913, AGS-1358239, and AGS-1255894 to the Institute for Astronomy of the University of Hawaii. These observations would not have been possible without help from Prof. Fred Sigernes from UNIS, who offered the use of the old Auroral Station in Adventdalen, Svalbard, Norway, and enabled the modification of the windows into glass doors. The authors are thankful to Judd Johnson, Paul Toyama, and Randy Chung for technical help with the spectrograph and its use during the total solar eclipse, to Chris Scharfenorth and Hans Eichler for providing the dichroic mirrors and filters used in the spectrometer, and to the late Heinrich Helfmeier who donated the entrance optics for the spectrometer. The authors also acknowledge help from Nathalia Alzate and Huw Morgan for access to the processed images of the ancillary SOHO-LASCO/C2 white-light data on CMEs, to Miloslav Druckmüller who produced the white-light eclipse image, and Enrico Landi for valuable discussions. The authors also thank Karen Teramura for her help with the final production of Figure 5 and the anonymous referee for valuable input.