Helium Emissions Observed in Ground-Based Spectra of Solar Prominences

The only prominent line of singly ionized helium in the visible spectral range, He ii 4686 Å, is observed together with the He i 5015 Å singlet and the He i 4471 Å triplet line in solar prominences. The Na D2 emission is used as a tracer for He ii emissions which are sufficiently bright to exceed the noise level near 10−6 of the disk-center intensity. The prominences thus selected are characterized by small non-thermal line broadening and almost absent velocity shifts, yielding narrow line profiles without wiggles. The reduced widths [ΔλD/λ] of He ii 4686 Å are 1.5 times broader than those of the He i 4471 Å triplet and 1.65 times broader than those of the He i 5015 Å singlet. This indicates that the He lines originate in a prominence–corona transition region with outwards increasing temperature.


Introduction
As the second most abundant element, helium plays an essential role in astrophysics. Nevertheless, its spectrum is poorly understood. The faint He II 4685.7 Å line is of particular interest, since it is the only important He II line that can be observed with ground-based telescopes allowing higher spectral resolution than currently achieved for EUV He II lines from space: Stellmacher, Wiehr, and Dammasch (2003) find that the SUMER spectrograph (designed for broad coronal lines) does not resolve the narrow emissions from cool prominences, even after application of a maximum instrumental profile.
He II 4685.7 Å has been observed in prominences during eclipses (Sotirovski, 1965;Poletto, 1967). Line-profile analyses, based on moderately resolved spectra from coronagraphs (Hirayama, 1972;Hirayama and Nakagomi, 1974) indicate that He II 4685.7 Å originates from the same (cool) prominence regions as the usually observed Balmer, He I, and metallic lines. Tandberg-Hanssen and Zirin (1959) observe in "flare-like loop" (i.e. highly active and "hot") prominences that the width of He II 4685.7 Å is larger than that of He I 4471.5 Å, however, they neither discuss quiescent ("cool") prominences nor the atomic fine-structure broadening of He II, which may explain most of that excess. A detailed analysis of the He I and He II lines in solar prominences requires a high spectral resolution, a high signal-to-noise ratio, and careful absolute calibration, which is difficult to achieve (see Illing, Landman, and Mickey, 1975), but it is possible with modern CCD techniques.
The high ionization and excitation energies of 25 and 48 eV suggest that He II 4685.7 Å may preferentially occur in hot prominences (T kin > 8000 K), which are observed to have a high He-to-Balmer emission ratio and to be highly structured (Stellmacher and Wiehr, 1994). These, however, usually show important velocity shifts, which disperse the line profiles. We therefore suppose that the opportunity to measure He II 4685.7 Å increases if the emitted photons are concentrated in wavelength, i.e. yield narrow line profiles free from spatial Doppler shifts. Such emissions should occur in prominences with negligible macro-velocities and low non-thermal line broadening, which are known to be bright in Hα, although with a low He-to-Balmer ratio (Engvold and Livingston, 1971;Stellmacher and Wiehr, 1995). Such prominences show significant Na D and Mg b emission and saturated Hα (and even Hβ) profiles, but so far did not allow a He II 4685.7 Å line profile analysis (Stellmacher and Wiehr, 2005).

Observations
In August 2011 we observed prominence emissions with the Gregory-Coudé telescope at the Locarno Solar Observatory (IRSOL: Ramelli et al., 2006). The faint He II 4685.7 Å emission is hardly visible in the raw spectra and can only be detected after careful subtraction of the superposed aureole spectrum. We therefore used the Na D 2 emission as a visual tracer for prominence candidates with "spectral photon concentration" and thus sufficient He II radiance. The emission regions so selected elevated only a few arcseconds over the limb and occurred at low solar latitudes: ϕ < 30 • . We oriented the spectrograph slit perpendicular to the solar limb by means of an image rotator, and pointed the telescope so that the light of the solar disk did not pass the aperture in the prime focus (cf. slit-jaw images in Figure 1). This avoids an illumination of the secondary and the two folding flat mirrors of the telescope, thus reducing the stray-light (cf. Figure 2 in Stellmacher and Wiehr, 1970) -in our case by almost a factor of two.
In order to obtain a sufficiently high signal-to-noise ratio for the faint He II line, we chose a slit width of two arcseconds and superposed spectra of (typically) ten seconds exposure, yielding total integration times between 50 and 300 seconds. We used an unmasked "e2v CCD 55-30" sensor mounted on a ZIMPOL camera (Ramelli et al., 2010) and operated the ZIMPOL system only in its intensity mode (Stokes I). The flat-field was deduced from attenuated disk-center spectra, which also served for the absolute calibration of the emissions. We took these spectra under identical conditions (exposure, slit width) as those for the prominence (and aureole), except for a defined attenuation of the high disk-center light the prominence displaced by a few arcseconds from the slit (see left panels); difference image (lower panel) with prominence emissions, color scale adapted to the faint Ti II 4468.4 Å line, the strong He I 4471.5 Å emission is thus over-saturated; the Hα slit-jaw images show the chromosphere just at their bottoms, the solar disk is outside the field-stop in the primary focus, seen in the left panels as the image boundary of 200 arcseconds diameter; the slit length (and thus the height of the spectra) amounts to 100 arcseconds. level using carefully calibrated neutral filters. This allows one to express the CCD counts of the prominence emission in terms of CCD counts at disk-center, which finally are converted into absolute radiance [erg s −1 cm −2 sterad −1 ] using the tables of Labs and Neckel (1970).
The prominence emissions are superposed by an absorption spectrum from the solar disk (the "aureole") which is due to Rayleigh scattering mostly by dust particles on the telescope mirrors and less in Earth's atmosphere (cf. Figure 2 in Stellmacher and Wiehr, 1970); it decreases in brightness with distance from the limb. We took such aureole spectra in the immediate vicinity of the respective prominence, normalized them to fit the spatial intensity distribution of the prominence spectra outside their emission lines, and subtracted them from the prominence spectral images (see Figure 1). The "zero level" of the resulting emission spectra is disturbed by spurious remnants of the aureole lines, arising from small (sub-pixel) spectral shifts between prominence and aureole exposure (spectrograph seeing). That "λoffset" produces signatures that are anti-symmetrically shaped and can be described by the derivative of the aureole absorption profiles. We largely removed these signals by either adding or subtracting a fraction of the derivative of the respective aureole spectral profile. In the corrected spectra we finally select spatial prominence regions with strong emission and negligible Doppler structures.
On 2 and 4 August we observed exclusively Na D 2 , He II 4685.7 Å, and the He I 4471.5 Å triplet line. After preliminary data reduction, we decided to include the He I 5015.7 Å singlet line as well. The first spectrum of that line (9 August) showed the emission of the neighboring Fe II 5018.4 line, drawing our attention also to the faint Ti II 4468.4 Å line in the vicinity of He I 4471.5 Å. The occurrence of such "chromospheric emissions", usually observed only during solar eclipses, establishes our selection criterion and demonstrates the high quality of our observations. On August 13 we added further metallic lines (see Table 3) with significant emission in the table of Sotirovski (1965).
Among dozens of prominences occurring in the first half of August 2011, we found only few with markedly bright and narrow Na D 2 profiles; however, only four of these showed measurable He II 4685.7 Å emission. The range of observed total He I 4771.5 Å emission (radiance), 745 < E(4471) < 5120 (see Table 1), largely exceeds the values E(4471) < 320 [erg s −1 cm −2 sterad −1 ] by Stellmacher and Wiehr (1997). This shows that prominences with high He I radiance also yield a sufficient He II emission above the noise level of few 10 −6 of the disk-center brightness (see Figure 3). Indeed, the chain of neighboring prominences occurring on 4 August at the east limb between 10°N and 35°N contained two with directly visible Na D 2 emission, but only one (14°N, marked in Figure 2) with Na D profiles narrow and bright enough to allow observable He II 4686 emission. Neither the Hα or the He II 304 Å appearance above the limb nor that of the corresponding filament on the disk and even not the cold prominence body, seen in the Fe XVI 335 Å image as Lyman and He I continuum absorption, gives an indication for a peculiarity of that prominence (E, 14°N) favoring the detectability of a He II 4685.7 Å emission. Figure 3 shows the strongest and the faintest He II emissions obtained at a noise level of ± 1 × 10 −6 of the disk-center radiance. In Table 1 we summarize the observed totalline emission and compare them with observations by Sotirovski (1965) and by Poletto (1967). Our low-latitude prominences (ϕ < 30 • ) show much stronger radiance of the helium lines than the eclipse observations. The highest observed Na D 2 radiance of 1010 [erg s −1 cm −2 sterad −1 ] (see Table 1) corresponds to a total Hα emission of 2.5 × 10 5 [erg s −1 cm −2 sterad −1 ], which, in turn, is related to a large optical thickness of τ 0 (Hα) ≈ 7.0 (according to observations by Wiehr, 1994, 2005); this proves that our criterion indeed selects thick prominences.

Line Radiance
We find mean radiance ratios E(4471)/E(5015) = 8.7 for triplet-to-singlet and E(4471)/ E(4686) = 51 for triplet-to-He II (cf. Table 1). These values of the ratios are comparable to those obtained by Sotirovski (1965) and by Poletto (1967) from eclipse observations, although our prominences are much brighter. Hence, the radiance ratios seem to be valid over a large range of He emission.

Line Widths
The spectrograph slit of 0.25 mm width (corresponding to two arcseconds) yields a typical instrumental line broadening of λ e /λ ≈ 8 × 10 −6 , which we apply to our observed line profiles. In addition, one has to consider the atomic fine-structure broadening: For He I 4471.5 five main components almost coincide ( λ = 19 mÅ); an additional faint component 210 mÅ red-wards (well visible in the lower panel of Figure 3) does not broaden the full width at half maximum [FWHM] of the composed profile. He II 4685.7 Å is composed of 13 atomic fine-structure components, leading to a significant line broadening. A corresponding deconvolution has to take into account that the various atomic fine-structure components have different intensity but a unique ("intrinsic") width. He II 4685.7 Å consists of two close main components ( λ = 0.53 mÅ, blue line in Figure 4); a third one, 100 mÅ red-wards (green line in Figure 4), broadens the FWHM of the composed profile (red line in Figure 4) by a factor of 1.25, which is independent of the intrinsic profile width. Two atomic fine-structure components (orange and yellow lines in Figure 4) produce a satellite at − 350 mÅ, which is hardly visible in the upper left panel of Figure 3 but does not influence the FWHM of the Gaussian fit to the upper part of the observed emission profile.
Considering the He II fine-structure broadening, we find that the He II 4685.7 Å line is 1.5 times broader than the He I 4471.5 Å triplet line, which, in turn, is 1.1 times broader than the He I 5015.7 Å singlet line (Table 2).

Kinetic Temperature
The Doppler widths [ λ D ] from atoms of different mass generally allow one to separate the thermal [T kin ] from the (Maxwellian) non-thermal [v nth ] line broadening: v 2 0 = (c λ D /λ) 2 = 2RT kin /μ + v 2 nth , if the lines are emitted in the same formation region. Associating [v 2 0 (1/μ)] Na D with [v 2 0 (1/μ)] He from the smallest He I (i.e. the singlet) line, we would obtain T kin ≈ 10 4 K; a temperature too high for our selected dense, Hα bright, and optically thick prominences. These typically show for hydrogen [ λ D /λ] H = 3.7 × 10 −5 (Stellmacher and Wiehr, 1994), which would yield T kin = 7000 K with the v 0 value observed for Na D 2 on 13 August (dashed line in Figure 5). The corresponding ordinate offset gives the small value of v nth = 2.8 km s −1 , confirming our criterion of "spectral photon concentration" (see the introduction).
The v 2 0 values of the singly ionized metallic lines are found above the dashed line in Figure 5, indicating excess broadening. This result disagrees with Landman (1985), who found Fe II 5169 to be smaller than Mg I b 2 and b 4 . Also, our observed v 2 0 values for He I

Figure 5
Velocity v 2 0 = (c λ D /λ) 2 versus inverse atomic weight 1/μ for the 13 August prominence; the dashed line connects v 2 0 (Na I) with a reasonable v 2 0 (H I) corresponding to [ λ e /λ] H = 3.7 × 10 −5 from Stellmacher and Wiehr (1994); the v 2 0 values of He II and of singly ionized metallic lines are located above that line; v 2 0 (He II) = 173 is outside the ordinate range.
singlet and triplet and He II are above the dashed line [T kin = 7000 K; v nth = 2.8 km s −1 ] in Figure 5. Keeping v nth = 2.8 km s −1 , the widths of these three helium lines observed on 13 August would correspond to T kin values of 9400 K, 10 800 K, and 40 000 K, respectively; for constant T kin = 7000 K they would yield v nth values of 7.3, 8.1, and 15 km s −1 .

Line Radiance
In order to correctly describe the helium spectrum in the prominence plasma, one has to consider the statistical equilibrium of ortho-, para-, and ionized helium and their interaction (e.g. Labrosse et al., 2010). Observations of spectrally well-resolved line profiles are required for the understanding of the helium spectrum. The optically thin He II 4685.7 Å line is poorly documented from observations. Its excitation is sensitive to radiation and collisions (Yakovkin and Zeldina, 1971;Labrosse et al., 2010), but also the influence of turbulence and of flows may enhance the He II emission (Jordan et al., 1997;Patsourakos and Vial, 2002). A blending with Ni I 4686.22 Å, mentioned by Worden, Beckers, and Hirayama (1973) in their study of chromospheric emissions, is not indicated in our prominence observations. Our radiance ratio, 7.2 < E(4471)/E(5015) < 9.8 (Table 1), is in good agreement with E(4471)/E(5015) = 9.5 from models with T = 8000 K and n H ≈ 10 10 cm −3 by Heasley, Mihalas, and Poland (1974). However, the radiance ratio 41 < E(4471)/E(4686) < 62, found in our study as well as in the eclipse data (Sotirovski, 1965;Poletto, 1967), is smaller than E(4471)/E(4686) = 6950 obtained from those models. The latter ratio would predict for our faintest prominence (13 August with E(4471) = 745) a He II 4685.7 Å radiance of E(4686) ≈ 0.1 [erg s −1 cm −2 sterad −1 ] far below the noise level -even for eclipse observations. Those models are isothermal; recent models, however, with a prominence-corona Table 1 Total line emission (radiance) [erg s −1 cm −2 sterad −1 ] in comparison with eclipse data from Sotirovski (1965) and from Poletto (1967

Line Widths
A further hint of the PCTR contribution to the He lines is given by the different line widths observed for the three atomic states: singly ionized, triplet, singlet. The reduced width [RW = λ D /λ] of the He II 4685.7 Å line (after deconvolution of the atomic fine-structure) is significantly larger than that of the He I triplet line: 1.18 < RW(4686)/RW(4471) < 1.8, and the triplet is broader than the singlet line: 1.07 < RW(4471)/RW(5015) < 1.09 (see Table 2). The latter result disagrees with observations by Heasley, Tandberg-Hanssen, and Wagner (1975), bearing in mind their lower spectral resolution. Since the He I lines are even visible in our raw spectra (see Figure 1; in contrast to the faint He II 4685.7 Å), the observed excess width RW(4471) > RW(5015) is outside the error bars of at most 2 % (see Table 2). The different widths of the He II and He I triplet and singlet lines may be due to their formation in prominence regions of different temperature. Indeed, Stellmacher, Wiehr, and Dammasch (2003) observe "hotter" lines to be more pronounced in such prominence regions which show less radiance in "cooler" lines. Labrosse (private communication, 2011) finds among 100 models with a PCTR of 10 5 K (cf. Labrosse and Gouttebroze, 2004) mean ratios of the reduced widths of 1.1 < RW(4686)/RW(4471) < 1.6 and 1.03 < RW(4471)/RW(5015) < 1.2, respectively, in good agreement with our results. The PCTR then seems to be essential for an explanation of the observed "hierarchy" of He line widths.
The difference between the singlet and the triplet emission is explained by the fact that the lowest He I energy level belongs to the singlet system, which can thus be populated directly by EUV radiation, while the triplet levels are populated from the (singlet) ground state mainly by ionization and recombination. Collisional excitation plays a minor role in cool prominence regions, which are clearly visible as dark absorption features in the Fe XVI 335 Å image (see Figure 2). These cool (and dense) prominence cores emit the narrow Na D lines, which correspond to low T kin and v nth values as, e.g. 7000 K and 2.8 km s −1 for 13 August in Figure 5. The broad He lines, however, originate in hotter prominence regions, where collisional excitation (and ionization) becomes effective. The observed "hierarchy" of excess broadening from He I singlet to He I triplet and He II lines may either be explained by differenttemperature threads or by a PCTR with outwards increasing temperature (or v nth ). But the width excess of He II 4685.7 Å corresponds to 40 000 K (cf. end of Section 3), which agrees with the formation temperature for He II lines calculated by Gouttebroze and Labrosse (2009). Hence, a temperature increase through the PCTR, as assumed in the models by Heinzel and Anzer (2001), may be more realistic than an increase of v nth . Also a combined outwards increase of both T kin and v nth might fit our observations.
The singly ionized metallic lines show a width excess similar to He I 4471 Å: In Figure 5 their v 0 exceed ≈ 1.2 times the values expected for T kin = 7000 K and v nth = 2.8 km s −1 . Hence, the singly ionized metallic lines (Table 3) may be emitted from similar layers as the He I triplet line. The rather tight relation of radiance and widths of Na D, He I 4471 Å, and He II 4686 Å (see Tables 1 and 2) favors a PCTR surrounding each individual thread rather than the prominence as a whole.