THE FORMATION OF IRIS DIAGNOSTICS. VIII. IRIS OBSERVATIONS IN THE C ii 133.5 nm MULTIPLET

We have used IRIS calibrated level 2 data, details of which are described in De Pontieu et al. (2014). To discriminate between different regions we used intensity maps taken at 280.0 nm (quasi-continuum in between the Mg ii h&k lines), with different thresholds to select regions such as umbra, penumbra, plage, and QS network. The QS data set from 2014 March 5 contains very little network and was therefore selected for sampling the internetwork, while the data set from 2014 July 30 contains more bright regions and was used for sampling the network. The selected regions are highlighted in Figures 1 (QS internetwork), 2 (QS network), and 3 (plage, penumbra, umbra). The two horizontal lines seen in the 280 nm images are IRIS alignment marks and were ignored in the region selection.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

HMI data were extracted and co-aligned with IRIS using the following procedure. First, a precise co-alignment between HMI and filtergrams from the Atmospheric Imaging Assembly (AIA, Lemen et al. 2012) was obtained by running the SolarSoft routine aia_prep on both HMI level 1.5 magnetograms and AIA level 1 filtergrams. The IRIS observations were then co-aligned with HMI and AIA using the IRIS 140 nm slit-jaw images with the AIA 30.4 nm channel (interpolated to the pixel scale of IRIS). The final HMI magnetogram maps were built by extracting the values from the HMI full disk images along the position of the IRIS slit, using the HMI image closest in time of each IRIS raster step.

One should keep in mind that each IRIS raster takes about 3.5 hr to complete and the Sun rotates about 32'' during this time. With the raster direction being solar east to west, the structures in all maps shown here (e.g., Figures 1–3) appear slightly compressed along the x-axis if one were to compare with a similar map taken at one instant in time.

Zoom In Zoom Out Reset image size

The IRIS data used here were processed with version 1.42 of the SolarSoft routine iris_prep, which includes an improved absolute wavelength calibration that ensures a negligible drift on reference neutral lines such as O i 135.56 nm, used here as a reference. There are also second-order terms included for the wavelength dispersion, which are important for the proper wavelength calibration at 133.5 nm. For obtaining the absolute wavelength calibration, we furthermore assume that the spatially averaged profile along the slit of the O i 135.56 nm line has a zero velocity shift, so all C ii velocity shifts are measured in relation to this line at each slit position.

An absolute intensity calibration factor for the FUV window was obtained by using the pre-flight spectrograph effective area (using the SolarSoft routine iris_get_response, version 2) and corrected by the sensitivity level as measured by observations of bright stars such as HD91316 (see IRIS Technical Note 24).

Earlier OSO 8 observations reported an integrated C ii multiplet intensity of 1.13 W m⁻² sr⁻¹, while from SUMER the total intensities for the multiplet in two internetwork data sets were 0.82 and 0.47 W m⁻² sr⁻¹ (Judge et al. 2003). Using IRIS spectra we measure a C ii multiplet integrated intensity of about 0.87 W m⁻² sr⁻¹ for the internetwork QS, which is consistent with earlier measurements.

To mitigate the effects of noise in the observed data, we have employed a weighted mean filter for each individual spectrogram (2D image with space and wavelength). For the original signal s we took the filtered signal s_filt as:

$$\begin{eqnarray}{s}_{\mathrm{filt}}=\left\{\begin{array}{ll}\displaystyle \frac{{\sigma }^{2}}{{\sigma }_{s}^{2}}{m}_{s}+\left(1-\displaystyle \frac{{\sigma }^{2}}{{\sigma }_{s}^{2}}\right)s & \mathrm{if}\ {\sigma }_{s}^{2}\geqslant {\sigma }^{2},\\ {m}_{s} & \mathrm{if}\ {\sigma }_{s}^{2}\lt {\sigma }^{2},\end{array}\right.\end{eqnarray} \tag{ 1 }$$

where m_s and σ_s² are the local mean and variance (from a 3 × 3 window) and σ² is an estimate of the noise power, which we took as the average of the local variances. The filtered intensities approach the original signal in regions with high S/N, and tend to the local mean for regions with low S/N (which are typically in the internetwork).

The noise filtering increases the number of good fits in the parameter estimations by about 10% in the internetwork data set and decreases the number of outliers. We note that the filtering does not change the overall trend of the results.

We have calculated the average spectra for each different region. The average C ii line profiles are shown in Figure 4. The line intensity increases from QS internetwork, network, penumbra to umbra, and it is highest in plage. The profiles show a velocity shift with respect to the position at laboratory wavelength (dotted line). Here, positive is redshift. This redshift is smallest in QS internetwork. The mean intensity profiles for the internetwork and network regions show a slight central reversal (double-peak profile) in both lines whereas, the umbral, penumbral and plage profiles are single-peak in the 133.4 nm line. The wavelength binning smoothens out the narrow double-peak profiles into single-peak profiles and this affects the number of peak statistics and shape of the mean intensity profiles.

Zoom In Zoom Out Reset image size

The statistics for the number of peaks in each line in different regions are shown in Figure 5. The 133.4 nm line has a larger proportion of single-peak profiles than the 133.5 nm line, which includes a blend. The 133.5 nm line shows about as many single-peak as double-peak profiles in both internetwork and network regions. There were some triple-peak intensity profiles as well, but their number was negligible and they were ignored.

Zoom In Zoom Out Reset image size

The umbrae of sunspots are a special case where the C ii lines (as well as the Si iv 139.4 nm and Mg ii h&k lines) are mostly single-peak.

Acoustic waves with periods of 2–3 minutes have been identified in sunspots for decades (Lites 1992; Thomas & Weiss 1992). The umbral oscillations are barely visible in the Si iv lines, whereas the C ii and Mg ii line intensities show very prominent oscillations. Given the 31.6 s raster step cadence, these umbral oscillations appear as bright/dark, mostly vertical stripes in the C ii intensity in Figure 3. These wave-like structures belong to magneto-acoustic waves generated below the photosphere, propagating upwards along the field, shocking in the chromosphere and creating umbral flashes.

Such flashes are similar in nature to the internetwork acoustic waves and shocks discussed by Carlsson & Stein (1995, 1997), and also to plage shocks and dynamic fibrils (Hansteen et al. 2006; De Pontieu et al. 2007a). These internetwork oscillations can also be seen in the intensities of Mg ii k and sometimes in C ii but they are even more clearly visible as vertical stripes in Doppler shift maps.

Coronal and transition region lines are typically formed under optically thin conditions, displaying single-peak Gaussian profiles. In these conditions, line ratios from some multiplets can be used to extract atmospheric properties such as density and temperature. However, this is normally not the case with lines that are optically thick. In such cases the line ratio depends on the source functions, hence on both temperature and opacities. In Paper I we studied the intensity ratio between the C ii lines and found that when the lines are optically thin, the ratio of the 133.5 nm line to the 133.4 nm line is 1.8 across the whole intensity profile, i.e., at all frequencies. For optically thick conditions the ratio depends on how the source function varies with height and on the ratio between the source functions. Double-peak profiles have a source function with a local maximum between the height where the continuum is formed and where the line core is formed. For a case where the source functions are equal for the two lines, this would lead to equal intensity peaks, a 133.5 nm line that is broader and with a deeper self-reversal than the 133.4 nm line and a low total intensity ratio. Single-peak profiles will have a larger total intensity ratio. However, the 133.4 nm line and the weak blend component share the same upper level, which reduces the source function of the 133.4 nm line via radiative rates, increasing the line ratio.

In Figure 6 we show histograms of the ratio of the total line intensity for the different regions considered. The line ratio is nearly always less than 1.8, meaning that the lines are not dominated by optically thin formation. The network regions show the lowest ratios, peaking around 1.25; the internetwork regions also have low ratios, peaking around 1.35. Both of these cases of low line ratios can be explained by the large proportion of double-peak profiles in these regions (Figure 5). The plage region also has ratios similar to the internetwork, although there are more single-peak profiles there. We interpret this as being caused by large densities where the plage profile maxima form, leading to larger collisional rates and therefore similar source functions in both lines. For the penumbral and umbral regions the proportion of single-peak profiles increases, and so do the line ratios.

Zoom In Zoom Out Reset image size

To extract the C ii line shifts and widths, we fit Gaussians to the line profiles. Despite several line profiles showing non-Gaussian shapes (e.g., double-peak profiles with a central reversal), this approach was favored as it allowed a reliable estimate of the shifts and widths over the whole field of view (in particular QS regions with lower S/N). Other fitting approaches (e.g., polynomial fitting to the wings of the line) were tested, but proved less robust and very sensitive to noisy line profiles. One should keep in mind that the Gaussian fits lead to an underestimation of the line width in the cases with double-peak profiles.

We have measured the peak (maximum) intensities, line shifts (Δλ, in velocity units) and the 1/e width of the Gaussian fit (Δν_D) for all points in the different regions. The line shifts were measured with respect to the O i 135.56 nm line.

Histograms of the line properties in different regions are shown in Figure 7. The peak intensities are highest in plage and smallest in the internetwork. In the umbra the peak intensities have a double peaked distribution, most likely caused by the upward propagating umbral flashes, where the hot shock front is about 1000 K hotter than the surroundings (de la Cruz Rodríguez et al. 2013) and increases the C ii ionization rates, leading to periodic high intensity spikes in the peak intensities.

Zoom In Zoom Out Reset image size

The histograms of line shifts are shown in Figure 7, panels (b) and (e) for the 133.4 and 133.5 nm lines, respectively. From the histograms of the line shift one can see that there is approximately a zero shift in the 133.4 nm line for the internetwork region, whereas the 133.5 nm line shows a small blueshift of about 2.1 km s⁻¹ because of the blend. In all the other regions the lines are redshifted with respect to the QS internetwork region (black curve). In the plage and network regions the lines are more redshifted compared to the umbra and penumbra (see also Table 2 which shows the median of the measured quantities). To avoid the effect of the blend on the shift, we focus on the 133.4 nm line. The zero shift in the internetwork region and redshifts for all the other regions might indicate that the C ii lines show no redshift when there is no magnetic connection to the corona. Also, it is interesting to note that the network region shows a broad distribution of velocity shifts, i.e., large range of velocities. This may suggest that the flows do not scale with the magnetic field strength, while the lack of mixed polarities or the stiffness of the field in plage may cause lower wave amplitudes (and thus a smaller line of sight motions).

The histograms of line widths are given in Figure 7, panels (c) and (f). Surprisingly, the plage and the QS internetwork show very similar width distributions. In the network one finds the broadest profiles among all regions, while sunspot umbrae show the smallest widths. The penumbrae show a similar peak in the width distribution as the umbrae, but with a broader distribution extending toward wider profiles. The mean widths of individual profiles for all regions are given in Table 2. In Paper 2, we discussed in detail the width of an optically thick emission line. There are basically two contributions: opacity broadening, which depends on how the source function varies through the atmosphere (and is larger for double-peak profiles than for single-peak profiles) and the broadening of the opacity profile itself (consisting of a thermal and a non-thermal component). It is difficult to separate the two effects without having an independent measure of one of them.

We find an inverse correlation between the line widths and the line ratios, which can be seen in the maps of Figure 8. The structures showing larger line widths (bright structures in panels (b), (e), and (h)) have small intensity ratios, whereas the structures with smaller widths have larger intensity ratios. In Paper I, we showed that double-peak profiles have more opacity broadening than single-peak profiles, and therefore a larger width. They also tend to have a lower intensity ratio.

Zoom In Zoom Out Reset image size

The line ratio in the network is typically lower than in other regions (Figure 6), while the widths are largest (Figure 7, panels (c) and (f)). This suggests that the increased width is partly caused by the larger opacity broadening for the double-peak profiles often found in the network. However, this cannot be the whole explanation since the internetwork has a similar fraction of double-peak profiles (Figure 5). This implies that the intrinsic broadening (thermal and non-thermal) must also be larger in the network. With the chromospheric fibrils that are the disk conterpart of spicules being prominent features in the chromospheric network, the increased line broadening could also be caused by their motions (Hansteen et al. 2006; de Pontieu et al. 2007; De Pontieu et al. 2007a, 2007b; McIntosh & De Pontieu 2012; Sekse et al. 2012, 2013a, 2013b; Pereira et al. 2014).

The plage shows similar widths as in the internetwork. We expect higher opacity broadening for the internetwork profiles because of a larger proportion of double-peak profiles. Therefore the similarity between the two distributions indicates that plage has larger non-thermal broadening than the internetwork.

We compared the line width and the ratio of the total line intensity for two data sets, the QS (from 2014 March 5) and the active region (2014 July 4). In Figure 9 we show the probability density function (PDF) of Δν_D as a function of the ratio of the total line intensities in the QS and the active region. We plot separate distributions for regions with single-peak (black) and double-peak (red) profiles. These PDFs show single-peak profiles typically having smaller widths and larger ratios, whereas the double-peak profiles have typically larger widths and smaller line ratios, consistent with the discussions above. Note that these panels include all the spectra in each raster, not just the ones selected for the different regions.

Zoom In Zoom Out Reset image size

The formation of the C ii lines has been described in detail in Paper 2. From the simulations it was found that the line forms somewhere between the Mg ii k line core height of formation and the transition region. However, in the IRIS observations the lines show many features similar to the Si iv 139.4 nm lines (under equilibrium conditions formed around 80,000 K) and sometimes features similar to the Mg ii k line core intensity.

In Figure 10 we compare intensity maps taken at the fixed wavelengths of the Mg ii k₃ feature, and the line cores of C ii 133.5 nm and Si iv 139.4 nm. It is noteworthy to see that in general most of the features in the C ii maps bear a strong resemblance to features in both the Mg ii and Si iv intensity maps. On one hand, this could indicate that the structures formed at high temperature, close to the transition region, as seen in Si iv are correlated with the upper chromosphere where the C ii lines are formed. On the other hand, there are also examples where the C ii line maps are very similar to the Mg ii k maps. Some examples are shown in Figure 10, in regions marked R1 and R2, enclosed in the yellow boxes at the top and bottom rows. In such regions the C ii intensity map resembles the Mg ii k intensity map more than the Si iv intensity map.

Zoom In Zoom Out Reset image size

In the QS map, shown in the top row, region R1 shows dark structures in both Mg ii and C ii, whereas these structures are brighter and smaller in the Si iv map. These structures resemble filaments, which typically have lower temperatures than mapped by Si iv. Most of the line intensity profiles in region R1 are double peaked, with deep central reversals, in both C ii and Mg ii k.

Another area in the active region maps where the C ii features resemble both Mg ii k and Si iv is marked as R2. This structure is bright in the Si iv image (panel (i)), shows mixed bright and dark features in the C ii image (panel (h)) and is dark in the Mg ii k image (panel (g)). We speculate that the dark features in the C ii and Mg ii k intensity maps in this region R2 are locations where both the Mg ii and C ii lines are formed higher at lower temperatures, where the source function decreases strongly with height, resulting in the line core intensities being lower for both lines. The brightness in Si iv suggests the presence of higher density material at the line's formation temperature in the transition region. The C ii Doppler shift maps, especially in the active region (panel (l) in Figure 8) clearly indicate a redshift (downfalling material) in region R2. This may be indicating a cool down-flow of coronal material. These downflows are expected to collect more and more material with an increase in depth which can result in larger dark structures in Mg ii and brighter ones in Si iv.

Both the peak intensity and the profile shape vary between solar regions. The highest intensities are found in plage and the lowest in internetwork and umbrae. The intensity profiles may have a single peak or two or more peaks. The number of peaks depends on how the source function varies with height. In the internetwork and network we find mostly double-peak profiles. The network shows slightly more double-peak profiles than the internetwork. In sunspot umbrae the profiles are mostly single-peak.

The ratio between the total line intensity of the 133.5 and 133.4 nm lines is almost always lower than 1.8, showing that the formation is not dominated by optically thin conditions. The ratio is lowest for the regions with the largest proportion of double-peak profiles (network) and highest for the region with predominantly single-peak profiles (umbra).

The comparison between the intensities of the Mg ii k, C ii 133.5 nm, and Si iv 139.4 nm lines gives the impression that the C ii lines are formed between the formation heights of Mg ii k and Si iv 139.4 nm. C ii line intensity maps show a mix of features from both the Mg ii k and the Si iv 139.4 nm maps. In most cases, the C ii 133.5 nm intensity is similar to that of Si iv 140.0 nm. However, some features in the C ii maps are similar to Mg ii k as well.

The absence of the blend makes the 133.4 nm line a cleaner diagnostic for velocity shifts. We found its velocity shift to be approximately zero in the internetwork and a redshift of 2.5 km s⁻¹ in the network. These numbers can be compared with those from SUMER time sequences by Hansteen et al. (2000). The authors found redshifts of 2.9 and 3.1 km s⁻¹ in the internetwork and 7.1 and 10 km s⁻¹ in the network, measured from two different data sets. An estimated error of ±2 km s⁻¹ in the mean velocity was reported for these observations. Betta et al. (1998), also using SUMER data, report mean redshifts of 2 and 4 km s⁻¹ in the internetwork and network, respectively. We believe that such differences could be explained by the lower spectral resolution of SUMER (8 km s⁻¹ pixel⁻¹ versus 2.7 km s⁻¹ pixel⁻¹ for IRIS), by difficulties in establishing an absolute wavelength scale (given a lack of appropriate neutral lines on both sides of the C ii lines), and by the limited spatial sampling of the SUMER observations. These SUMER time series observations only sample a 03 × 120'' slit on the Sun, while we have a wide raster sampling about 140'' × 180''. Our observations show a large variability, especially between different network patches. Sampling only very few network patches with the SUMER slit may thus lead to differences from our average, which is taken over a large area.

The lines are narrowest in the umbrae of sunspots and successively broader in penumbra, plage, internetwork and broadest in the network. We find average widths of 22 and 27 km s⁻¹ for the internetwork and network, respectively. This is comparable to the SUMER measurements from Hansteen et al. (2000), who find the widths of order 30 km s⁻¹ (after compensating for the instrumental width) for both internetwork and network. The difference is probably due to the same reasons as the differences in line shift: a combination of resolution and limited statistics for the SUMER observations.

The narrow profiles found in umbra and penumbra are consistent with the small opacity broadening expected in these regions due to their predominantly single-peak character (Paper I). The tail in the width distribution to large widths found in penumbrae may be due to a range of non-thermal velocities and a range of viewing angles to the penumbral filaments.

The network shows the widest profiles. This is partly explained by large opacity broadening expected for double-peak profiles. However, because the internetwork has the same proportion of double-peak profiles, but show narrower profiles, this is only part of the explanation. The network must in addition have large non-thermal broadening of the intrinsic opacity profile. We speculate that this may in part be due to motions in the disk counterpart of spicules that are frequently associated with the network.

Plage shows the same distribution of widths as internetwork. We expect less opacity broadening in plage than in the internetwork because of the smaller number of double-peak profiles with large peak separation. The equality in the width distribution, thus indicates larger intrinsic broadening (thermal and non-thermal) for plage.

The line widths for the C ii lines show an inverse correlation with the intensity ratios. The narrower single-peak intensity profiles are formed in the lower density upper atmosphere, whereas the broader lines with a central reversal have the peaks formed in the deeper atmosphere at higher densities. The double-peak lines are broader because of a larger opacity broadening factor (see Paper I).

Our results confirm that the C ii lines are good chromospheric diagnostics. The number of peaks, intensities, line ratios, shifts, and line widths show distinct differences between different types of solar regions. The structures observed in C ii intensity maps are well correlated with structures seen in Si iv maps and in fewer cases with structures seen in Mg ii k. The C ii lines can therefore be used as a tool to link the structures observed with other observables from the lower chromosphere to the transition region. In some cases, they can be distinctly different from other IRIS lines such as Mg ii h&k or Si iv 139.4 nm and, e.g., through their width maps, can provide a rich amount of information that is unique among the IRIS diagnostics.