Additions to the Spectrum of Fe IX in the 110-200 {\AA} Region

The spectrum of eight-times ionized iron, Fe IX, was studied in the 110-200 {\AA} region. A low inductance vacuum spark and a 3-m grazing incidence spectrograph were used for the excitation and recording of the spectrum. Previous analyses of Fe IX have been greatly extended and partly revised. The numbers of known lines in the 3p^53d - 3p^54f and 3p^53d - 3p^43d^2 transition arrays are extended to 25 and 81, respectively. Most of the identifications of the Fe IX lines from the 3p^53d - 3p^43d^2 transition array in the solar spectrum have been confirmed and several new identifications are suggested.


INTRODUCTION
Eight-times ionized iron, Fe ix, belongs to the argon isoelectronic sequence with a fully-occupied 3p 6 ground configuration. The strongest line in the Fe ix spectrum is the 3p 6 1 S 0 -3p 5 3d 1 P 1 resonance transition at 171.073Å, which was identified independently by Gabriel & Fawcett (1965) and Alexander et al. (1965). The reference wavelength for the line listed in the NIST database (Kramida et al. 2020) comes from the solar spectral atlas of Behring et al. (1972). In the past four decades the 171Å line has been a popular choice for solar extreme ultraviolet (EUV) imaging instruments on account of its strength, relative isolation in the spectrum, and the proximity of an aluminium absorption edge at 170Å. EUV imaging instruments make use of multilayer coatings to yield narrow bandpasses that are typically 10% of the central wavelength. However, the high density of lines in the EUV means that multiple strong lines can contribute to the imaging channel emission. Aluminium filters are used to block visible radiation in these instruments, and the aluminium edge at 170Å strongly attenuates emission to the short wavelength side of Fe ix 171.07Å, yielding a relatively clean bandpass. Because of this a filter at or close to 171Å has been used for several telescopes on spacecraft, including the EUV Imaging Telescope (EIT: Delaboudinière et al. 1995), the Transition Region and Coronal Explorer (TRACE: Handy et al. 1999), the Extreme Ultraviolet Imager (EUVI: Howard et al. 2008), the Atmospheric Imaging Assembly (AIA: Lemen et al. 2012), and most recently the Extreme-Ultraviolet Imager (EUI: Rochus et al. 2020) on Solar Orbiter. The 171.07Å line has also been measured in spectra of the cool stars Procyon (Drake et al. 1995) and α Centauri (Drake et al. 1997).
Other Fe ix lines in the EUV are much weaker, but the ratio of the lines at 241.74 and 244.91Å (decays from the 3p 5 3d 3 P 2,1 states to the ground) has long been recognised as a useful density diagnostic of the solar atmosphere (Feldman et al. 1978).
The first excited configuration of Fe ix, 3s 2 3p 5 3d, has 12 fine-structure levels and yields a rich spectrum of forbidden lines between 1738Å and 2.9 µm. Many of these lines have been measured during solar eclipses, and a summary is provided by Del Zanna & DeLuca (2018).
At soft X-ray wavelengths, there are three groups of lines coming from n = 4 configurations. The 3p 6 -3p 5 4d and 3p 6 -3p 5 4s transition arrays give two pairs of lines at 82.43 and 83.46Å, and 103.57 and 105.21Å, respectively, and these lines were reported in a solar spectrum by Malinovsky & Heroux (1973). The 3p 5 3d-3p 5 4f array gives a group of lines between 111 and 120Å and an additional pair of lines at 133.92 and 136.57Å. None of these lines have been reported from solar spectra, but they have been studied with laboratory spectra. This array is the subject of study in the present article and is discussed in Section 3.1.
Recent advances in the spectral analysis of Fe ix have largely been motivated by the solar spectra obtained with the Extreme ultraviolet (EUV) Imaging Spectrometer (EIS: Culhane et al. 2007) on board the Hinode spacecraft, launched in 2006. EIS has a spectral resolution of 3000 and it covers the wavelength ranges 170-212 and 246-292Å, which contain a wealth of emission lines from iron ions, ranging from Fe vii up to Fe xxiv (Young et al. 2007). Brown et al. (2008) reported that about half of the lines in the EIS wavelength ranges were unidentified. Young (2009) was the first to report new Fe ix identifications in the EIS wavelength bands, and further studies of the Fe ix spectrum have been performed by Del Zanna (2009) Landi & Young (2009a,b) Young & Landi (2009), O'Dwyer et al. (2012) and Del Zanna et al. (2014. Laboratory studies of the Fe ix spectrum in the EUV have been performed by Liang et al. (2009), Beiersdorfer & Lepson (2012) and Beiersdorfer & Träbert (2018). These works have resulted in a number of line identifications from the 3p 5 3d-3p 4 3d 2 transitions and around two dozen lines suggested as due to Fe ix but without definite identifications.
The present work focuses on line identifications in the 110-200Å range from the 3p 5 3d-3p 5 4f and 3p 5 3d-3p 4 3d 2 transition arrays. Figure 1 gives an overview of the known (prior to the present work) Fe ix transitions in the 100-220Å as found in version 10 of the CHIANTI atomic database (Young et al. 2016;Del Zanna et al. 2021). The spectrum was generated using a temperature of 0.8 MK and an electron number density of 4.0 × 10 8 cm −3 , which are typical of conditions in the solar atmosphere. Emission line widths have been set to 3Å in order to group nearby lines together. Five transition arrays are highlighted: the two studied in this work, and the 3p 6 -3p 5 4s, 3p 5 3d-3p 5 4p and 3p 6 -3p 5 3d arrays. The latter yields the 171Å line, which towers over the other features to an intensity peak of 10 on the plot scale. The only known line from the 3p 5 3d-3p 5 4p array is at 197.86Å (Young 2009). Note that there are many more transitions in CHIANTI in this part of the spectrum that are not shown because they only have theoretical wavelengths.
Our new laboratory study uses spectra excited in a vacuum spark and recorded with a higher resolution than in all previous papers. The number of known lines in the 3p 5 3d-3p 5 4f transition array is extended to 25. Several previous identifications were corrected resulting in changes to the energies of the corresponding levels. Eighty-one lines of the 3p 5 3d-3p 4 3d 2 array have been identified. Most of the previous identifications from this array in the solar spectrum have been confirmed and several new ones have been made. Lines previously assigned to Fe ix but not identified have been classified. This article continues a series of publications on laboratory high resolution studies of iron ion spectra (Ryabtsev 2017;Young et al. 2021;Kramida et al. 2022) relevant for diagnostics of solar plasma.
Section 2 describes the experimental setup used to obtain the Fe ix spectra. Section 3 gives the line identification results for the two transition arrays, and we give our conclusions in Section 4.

EXPERIMENT
The procedures used in our earlier studies of the iron ions (see, for example, Kramida et al. 2022) were followed in the present measurements. In short, the spectra were taken with a 3 m grazing incidence (5 • ) spectrograph equipped with a 3600 line mm −1 grating. The spectra were excited in a three-electrode vacuum spark run with peak currents up to 100 kA. The high current tracks on the ORWO UV-2 photographic plates taken for an analysis of Fe viii (Ramonas & Ryabtsev 1980) were measured. The tracks were scanned on an EPSON EXPRESSION scanner and then digitized and measured using the Gfit code (Engström 1998).
Spectra were also recorded on phosphor imaging plates (Fuji BAS-TR) (Ryabtsev 2017). These spectra were scanned using a 0.01 mm scanning step with a Typhoon FLA 9500 reader and processed and analysed with the ImageQuant TL 7.0 image analysis software. For final reduction of the spectra the GFit code again was employed. Because of a high level of background, the imaging plate spectra taken at high currents were useful only for wavelengths above around 140Å. Figure 2 shows the photoplate spectrum between 110 and 140Å obtained for a current of 40 kA, and Figures 3 and 4 show the photoplate spectrum between 140 and 201Å obtained at 100 kA. The wavelength scales are not the final ones used to derive the Fe ix wavelengths as a grating formula with averaged parameters was used. The final Fe ix wavelengths, given in Tables 1 and 3, were obtained following a reduction with reference lines and a correction curve of the photoplate. Close-ups of selected portions of the spectra are shown in Figures 5-7, where spectra obtained for different currents are compared.
The iron line wavelengths were measured by first taking photoplate spectra with an iron anode and a titanium cathode of the spark, giving a mixed spectrum of iron and titanium lines. The known titanium wavelengths (Svensson & Ekberg 1969) were used to derive reference wavelengths for those iron lines that are not blended with titanium lines. These iron line wavelengths were then used as secondary standards when reducing the pure iron spectra obtained with both iron electrodes in the vacuum spark. The width of the Fe ix lines on the photoplate spectra in the region above 140Å was about 0.03Å whereas on the imaging plate spectra the width was larger by about 20%. The photoplate containing the spectrum below 140Å that was used for the analysis of the 3p 5 3d-3p 5 4f transitions had a better resolution, with line widths around 0.02Å. The photoplate spectra possessed two other advantages with respect to the imaging plate spectra that were important for the wavelength measurements. Firstly, the signal-to-noise was better due to a smaller background, and secondly, the line shape was smoother due to the approximately two times smaller scanning step of the EPSON EXPRESSION scanner in comparison with the Typhoon FLA 9500 reader. The root-mean-square deviation of the reference titanium, as well as the secondary iron, lines from the calibration curve was 0.002Å, and the deviations from the mean values of the wavelengths obtained from the measurements of different spectra recordings was below 0.001Å. Because Svensson & Ekberg (1969) claimed an uncertainty of 0.004Å for their wavelengths we adopt this value as our wavelength uncertainty for single, unperturbed lines. This estimation is supported by the deviations of the measured wavelengths from the Ritz values calculated in the present spectrum analysis. Where possible, the line intensities were obtained from the imaging plate spectra due to the greater linearity of the response. The intensities will be given on an arbitrary scale without taking into account the wavelength dependence of the spectrograph efficiency and recording media sensitivity.
Based on our past experience with the vacuum spark, the plasma is in approximate local thermodynamic equilibrium with an effective electron temperature. Rough estimates following the procedure of Kramida et al. (2022) for Fe vii give a value of 10 eV. It is not possible to estimate a density from the spectra, but we expect a value around 10 17 cm −3 based on measurements from similar plasma discharges. The density would be expected to vary as the current is varied to modify the charge balance.

RESULTS
3.1. The 3p 5 3d-3p 5 4f transitions Laboratory measurements of the 3p 5 3d-3p 5 4f transition array were performed in the 1970s, with Wagner & House (1971) listing 12 lines in the 111-117Å range. Nine of these were measured by Fawcett et al. (1972) with an accuracy of ±0.01Å, and further work was done by Swartz et al. (1976) who remeasured all 12 lines, yielding energies for 10 of the 12 3p 5 4f levels with an estimated uncertainty of ±200 cm −1 .
More recently, O'Dwyer et al. (2012) provided a list of 20 identified 3p 5 3d-3p 5 4f lines and all level energies of the 3p 5 4f configuration. The wavelengths include the 11 values from Fawcett et al. (1972), but the source of the other nine wavelengths is not given. We speculate that the authors had access to the original plate of Our results for this transition array are displayed in Table 1 for the wavelengths and in Table 2 for the energy levels of the 3p 5 4f configuration. Many weak lines were added to the previous analyses of these transitions extending to 25 the number of identified lines and resulting in the location of all the 3p 5 4f levels with uncertainties 30 cm −1 or less. The wavelengths of several lines marked by VII in Table 1 are close to the Fe vii wavelengths of Ekberg (1981). Indeed, these lines together with the other Fe vii lines are present in our tracks with "colder" iron spectra but their influence on the intensities and wavelengths of the Fe ix lines is negligible in our case. One line with Ritz value 112.463(3)Å is not measured, possibly being masked by the Fe viii 112.472Å line.
The relative intensities measured in the photographic plate spectrum are given on an arbitrary linear scale. The 250 value was adopted for the strongest line (113.789Å) for convenient comparison with the corresponding gA values. (g is the statistical weight of the transition upper level and A the radiative decay rate.) The wavelength responses of the grating and photoplate were not taken into account. Even for a narrow wavelength range, the intensities should be considered as more qualitative than quantitative because of a possible error in the model characteristic curve of the photoplate used for a transformation of the measured line densities to their intensities. Table 1 contains a comparison with the previous measurements by Swartz et al. (1976) and O'Dwyer et al. (2012). Good agreement of all measurements with some exceptions is seen. The two lines at 111.713 and 112.031Å were listed by O'Dwyer et al. (2012) as the transitions from the 3p 5 4f 3 D 1 level. A part of the iron spectrum in the region of these lines, taken at two modes of the spark operation, is shown in Figure 2. Fe ix lines are clearly distinguished from those of Fe vii and Fe viii by the change in intensity from "cold" to "hot" spark conditions. The Fe ix lines have about the same intensity in the displayed spectra, while the lower ionization states have reduced intensities in the "hot" spectrum. Both the 111.713 and 112.031Å lines identified by O'Dwyer et al. (2012) are seen to belong to lower ionization states than Fe ix. The feature at 112.031Å is a known line of Fe vii, measured at 112.030Å by Ekberg (1981). The lines at 111.692 and 112.011Å (marked with arrows on Figure 5) show behaviour similar to the other Fe ix lines and were adopted as the transitions from the 3p 5 4f 3 D 1 level. The wavelengths are in good agreement with those of Swartz et al. (1976).
The line of 118.27Å identified by Swartz et al. (1976) as the 3p 5 3d 1 F 3 -3p 5 4f 3 F 4 transition has a low transition probability and is absent in our spectrum. Its Ritz value is 118.220(3)Å.
Previous Fe ix studies did not provide definitive line identifications for the 3p 5 4f 1 D 2 and 3p 5 4f 3 F 2 levels. Swartz et al. tentatively suggested that the two lines at 134.743 and 115.46Å belong to decays from the 3p 5 4f 1 D 2 level to the 3p 5 3d 1 P 1 and 3 D 1 levels, resulting in a value of 1 326 700 cm −1 for the 1 D 2 level. It should be noted that both lines are not present in our "hot" spectrum, the second one being possibly the 115.472Å line of Fe vii (Ekberg 1981). Lepson et al. (2002) recorded an Fe ix spectrum at the Lawrence Livermore electron beam ion trap (EBIT). The spectral resolution was relatively low (around 300 at 100Å) and the authors measured two weak lines at 134.08 and 136.70Å that they assigned to 3d-4f transitions, although level assignments were not made. The same lines were identified in low-resolution (around 150 at 130Å) solar spectra from the Extreme Ultraviolet Variability Experiment (EVE: Woods et al. 2012) by Foster & Testa (2011), although level information was not provided.
Our high-resolution spectra reveal several lines near the location of the Lepson et al. (2002) 134.08Å line ( Figure 6). Of these, the 134.063 and 134.128Å lines belong to Fe vii (Ekberg 1981), while the 134.169Å line shows behavior consistent with the Fe ix lines shown in Figure 5. We identify this line with the 3p 5 3d 1 P 1 -3p 5 4f 1 D 2 transition, which is supported by the Ritz transition to the 3p 5 3d 3 D 1 level with a wavelength of 115.042Å. The energy of the 3p 5 4f 1 D 2 level of 1 329 872 cm −1 is consistent with the energy calculated from the Cowan code. O'Dwyer et al. (2012) Figure 5. Two iron spectra correspond to "hot" (red) and "cold" (black) excitation conditions. The lines are marked by the symbols: vii -Fe vii, viii -Fe viii and ix -Fe ix. Lines due to the three iron species are distinguished by their relative intensities in the two spectra. Arrows indicate the Fe ix lines at 111.692 and 112.011Å.
suggested the 3p 5 3d 1 P 1 -3p 5 4f 1 D 2 transition belonged to a line at 133.923Å. Our spectra have a line at 133.899Å that is due to Fe vii (Ekberg 1981 et al. (2012) to the transitions from the 3p 5 4f 1 D 2 level are not present in our spectra.
Our value of 136.674Å for the 3p 5 3d 1 P 1 -3p 5 4f 3 F 2 transition is in agreement with the Lepson et al. (2002) measurement. The corresponding energy of 1 316 205 cm −1 for the 3p 5 4f 3 F 2 level is in good agreement with the value computed from the Cowan code. It is also supported by a Ritz combination to the 3p 5 3d 3 D 1 level. The corresponding line at 116.881Å was the only remaining line in this region of the spectrum that had the properties of Fe ix. On the other hand, O'Dwyer et al. (2012) attributed a line at 136.572Å to the 3p 5 3d 1 P 1 -3p 5 4f 3 F 2 transition with Ritz support from lines at 113.258 and 116.803Å. No Fe ix lines can be identified in our "hot" spectrum at 136.572 and 113.258Å. The 116.803Å line, measured in our spectrum at 116.814Å, is assigned to the 3p 5 3d 3 D 3 -3p 5 4f 3 F 4 transition in Fe ix.
The 3p 5 4f level energies in Table 2 were derived from the identified 3p 5 3d-3p 5 4f lines using the program LOPT for least-squares optimization of energy levels (Kramida 2011). The 3p 5 3d levels were fixed to the values of Edlen & Smitt (1978) for the optimization. The uncertainties of our level energies range from 19 to 40 cm −1 . Column 2 shows the deviations of the Cowan code energies from the experimental ones after a fitting of 12 energy levels with four free parameters. The deviations are in good agreement with the estimated level uncertainties. Good agreement is also seen with the energy values found by Swartz et al. (1976) andO'Dwyer et al. (2012) except for the questionable cases noted earlier. The differences in the energies are within 60 cm −1 , which shows that the Swartz et al. (1976) uncertainties of ±200 cm −1 were overestimated. Table 2 shows the designations of the levels in the LS and JJ couplings. It is seen that the JJ coupling representation is generally better than the LS one. Nevertheless, in Table 1 we are using the more convenient LS designations because Figure 6. The iron spectrum in the 133.8-134.3Å range taken in the "cold" (black) and "hot" (red) modes of the spark operations (see Figure 5).
it is possible to give unambiguously the LS names to all levels although in several cases they are associated with the second component of their wavefunction.
The wide wavelength coverage and high resolution of our spectra have enabled us to make many new line identifications from the 3p 5 3d-3p 4 3d 2 and these are summarized in Table 3. Updated level energies are provided in Table 4. In total, eighty one lines in the range 151-200Å were identified, and five of them are doubly classified, i.e., two Fe ix transitions are assigned to the same observed line (marked with "db" in column 2 of Table 3). The iron ion spectrum in this region is rich in lines. Depending on excitation conditions in a spark, the lines of Fe vi to Fe xii can be present and they can blend, mask or perturb the Fe ix lines. Even at the high current modes the lines of lower stages of the ionization can be seen due to temporal and spatial inhomogeneity of the spark plasma. The spectra taken at the peak currents 50-100 kA were used for the Fe ix line measurements. The wavelengths were obtained from the spectra recordings on the photographic plates, while the intensities were taken from the imaging plates, where possible. About 15 Fe ix lines have wavelengths close to known Fe vi-Fe viii lines (Azarov et al. 1996;Kramida et al. 2022;Ramonas & Ryabtsev 1980) or to the unidentified "cold" lines that are seen with high intensity in the spectra taken with lower peak currents. It was estimated, after a study of the behavior of the line intensities of these ion species on the peak current, that in most cases their influence on the Fe ix lines is negligible. We found and marked in Table 3 only six lines that can have contribution to their intensities from "cold" lines (marked with bl(VII) or bl(VIII) in Table 3) The relative intensities are given on a linear scale without accounting for the wavelength dependence of the spectrograph and imaging plate effectivities. The resonance Fe ix line at 171.073Å has in our spectrum the intensity 6000 on this scale. The scale is not directly connected to that of the 3p 5 3d-3p 5 4f transitions because the spectra were taken on different recording media and with different excitation in the vacuum spark. According to rough estimates the scales can be different by a factor of two. Table 4 lists the 45 levels in the upper level range of the 3p 4 3d 2 configuration (energies above 930 000 cm −1 ). The 81 lines identified in the present work yield energies for 30 of these levels. Three of the level energies are uncertain (marked with a ? in the table) as they are based on a single questionable line identification. The second column of the table compares the observed energies with the energies obtained from Cowan's code using parametric calculations. The remaining 15 levels in Table 4 are listed for completeness and the energies are those from Cowan's code. The values should be valuable for future efforts to identify Fe ix transitions. Table 4 also gives the first three components of each level's eigenvector. The first component of the eigenvector together with the energy level value was used as a label for the transitions in Table 3. A difficulty in the level designation should be pointed out. The 3p 4 3d 2 configuration consists of two sub-shells with more than one electron. For unambiguous level designation all intermediate quantum numbers should be given: 3p 4 (L S )3d 2 (L S )LS. Shorthand designation by the L S -numbers together with the final LS is adopted in the Cowan code resulting in an ambiguity of the repeated final LS-numbers belonging to the same L S but to different L S . The letters a or b are added to distinguish between such cases. The full descriptions for the first components of the eigenvectors are shown in Table 4 in square brackets.
The energy parameters after the least-square fitting of the calculated to the experimental energy levels are shown in Table 5. The electrostatic parameters for the configurations with unknown levels as well as the configuration interaction parameters were scaled by a factor 0.85 with respect to the corresponding HFR values (see p. 464 of Cowan 1981). Only a parameter of interaction between the 3s 2 3p 4 3d 2 and 3s3p 6 3d configurations was fitted for better description of the levels with mixed eigenvectors. The spin-orbit parameters were not scaled, and all of them are omitted from Table 5. The average energies of these configurations were scaled so that their differences with those of the known configurations were approximately the same as in the HFR calculations. The parameters of Table 5 were used for the calculations of the energy levels and transition probabilities of the Fe ix spectrum. The branching ratios for the intensities of the lines from a particular level generally follow the calculated transition probabilities with three exceptions. The intercombination line 1 D 2 -( 3 P ) 3 Gb (J = 3) at 199.985Å is too intense and two other intercombination lines 1 D 2 -( 3 P ) 3 F a (J = 2) and 1 F 3 -( 3 P ) 3 P a (J = 2) at respectively 176.646 and 178.848Å are too weak in their transition arrays.
Finally, we give a few remarks about previous measurements and identifications of the Fe ix lines. Several laboratory studies of Fe ix were undertaken with the aid of the Heidelberg (Liang et al. 2009) or the Lawrence Livermore National Laboratory (Beiersdorfer & Lepson 2012;Beiersdorfer & Träbert 2018) electron beam ion traps (EBIT). The observed spectra showed the evolution of each ionic stage from Fe 5+ to at least Fe 15+ as a function of the electron energy, allowing to distinguish the emission lines from the neighboring ion charge states.
The spectra of Liang et al. (2009) were recorded in the 125-265Å range with 0.5-0.8Å resolution. Four lines (188.5, 189.9, 191.2 and 197.9Å) were suggested as belonging to Fe ix, and these were independently identified by Young (2009) from EIS spectra. The first three lines belong to the 3p 5 3d 3 F -3p 4 3d 2 ( 3 P ) 3 G multiplet, and the fourth is the 3p 5 3d 1 P 1 -3p 5 4p 1 S 0 transition. Beiersdorfer & Lepson (2012) measured with a resolution of about 0.3Å a dozen spectral features in the 170-200Å range, some of which could be attributed to Fe ix based on ionization energy and wavelength coincidences with a CHIANTI spectral model. Version 7.0 (Landi et al. 2012) of CHIANTI was used, which mostly had only theoretical wavelengths for the lines in this wavelength range and so definitive new identifications could not be made. The authors did provide supporting data for some of the identifications of Young (2009) and Young & Landi (2009), however.
Beiersdorfer & Träbert (2018) (hereafter BT18) studied emission in the wavelength region 165-175Å from various species excited in an EBIT, including Fe ix. The spectral resolution was 3000 and wavelengths were measured with uncertainties of 10-20 mÅ. The spectrum produced by a beam energy 300 eV yielded lines mainly due to Fe ix and Fe x. The authors modeled the Fe ix emission by employing atomic data computed with the relativistic Multi-Reference Møller-Plesset (MR-MP) perturbation theory (Vilkas et al. 1999) and the Flexible Atomic Code (FAC: Gu 2008). By comparing their predicted spectra with the EBIT measurements, BT18 were able to match features in their modeled spectra to features in the EBIT spectrum based on proximity in wavelength and intensity. Some of the lines were blended with other species.
The MR-MP method has previously been shown to yield level energies with a spectroscopic accuracy. For example, the n = 3 levels of Fe xiii (Vilkas & Ishikawa 2004) are reproduced to around 0.01%. However, earlier works did not address complex configurations of the form 3p k 3d m (k < 6, m > 1). Recently, Santana et al. (2020) performed MR-MP calculations for Fe viii and derived 3p 5 3d 2 energies with accuracies up to 0.8%, which is not high enough for classifying unidentified lines. Although not investigated by BT18, such uncertainties may also apply to their calculations for the 3p 4 3d 2 configuration of Fe ix.
BT18 did not publish their Fe ix atomic data and so it is not possible to compare radiative decay rates with the present or earlier calculations. However, they do comment that the 3p 4 3d 2 ( 1 D) 3 D 3 a level (using our level notation) has its strongest decay to 3p 5 3d 3 F 3 whereas the CHIANTI 8 atomic model has the strongest decay to 3p 5 3d 3 D 3 . The CHIANTI 8 model decay rates are from Del Zanna et al. (2014) and are comparable to the present decay rates. We compared with the calculations of Storey et al. (2002) and Tayal & Zatsarinny (2015), and these both confirmed that the strongest decay is to the 3 D 3 level. This particular discrepancy affects the identification of the lines at 167.478 and 174.03Å measured by BT18, as discussed below.
Of the 16 Fe ix lines identified in Table 5 of BT18, seven are present in our spectra and we also assign them to Fe ix. The BT18 line at 172.16Å was blended with O v in their spectrum but this is not the case in our spectra. BT18 noted that lines at 170.92 and and 173.90Å could be due to Fe ix. Both lines are present in our spectra with Fe ix properties. All nine of the BT18 measured wavelengths are given in the λ Pred column of Table 3.
The higher resolution of our spectra allow us to resolve the 170.11Å line of BT18 into two lines at 170.116 and 170.150Å (see Figure 7). The line at 174.03Å is also resolved into two lines at 174.024 and 174.043Å, and the former is a double-blend of two Fe ix transitions (Table 3).
Of the seven BT18 lines for which they assigned identifications, we confirm only one: the 3p 5 3d 3 D 3 -3p 4 3d 2 ( 1 D) 3 D 3 a transition at 174.03Å. This is the strongest of a pair of transitions that are double-blended in our spectra at 174.024Å (Table 3). Both identifications are supported by multiple Ritz combinations. Although we agree with the BT18 identification, we note that the BT18 atomic model predicted this transition to be the weakest of all the decays from the 3p 4 3d 2 3 D 3 a level, yet the measured line is quite strong in their spectra. This is a consequence of the problem with this level noted earlier.
Of the remaining six lines for which identifications disagree, we find two belong to the set of three 3p 5 3d 3 F J -3p 4 3d 2 ( 3 P ) 3 F J a transitions. The BT18 wavelengths 169.605 and 169.900Å match our wavelengths 169.614 and 169.914Å, corresponding to the transitions with J = 3 and J = 4, respectively. We identify the J = 2 transition with a line at 169.773Å (see discussion below). Each identification is supported by between four and six Ritz combinations. The 2-3 and 3-4 members of the same multiplet comprise the 170.92Å feature in the BT18 spectrum, with Ritz wavelengths of 170.918 and 170.927Å, respectively. The other Ritz combinations within the 165-175Å range are at 168.483 and 168.610Å (Table 3) and lie close to a strong Fe viii line in the BT18 spectrum and were not reported by BT18.
The identification of the 169.773Å line warrants further discussion as BT18 stated that this line must come from a spectrum lower than Fe ix. Figure 7 shows the spectra in the region of this line taken at different excitation conditions in the spark. The intensities in the spectra are scaled so that their comparison can help in the attribution of the lines to different ions. The 169.773Å line is clearly blended with a strong Fe vi line (Azarov et al. 1996). But a comparison of the changes of its intensity with those of the other Fe vi and Fe ix lines clearly shows that the main contribution to the intensity of this line in "hot" conditions comes from Fe ix. The identification is supported by the observation of an additional five Ritz lines in our spectra.
Two more of the six BT18 lines for which we have different identifications are 170.11 and 171.685Å. These are the two strongest decays of the ( 1 D) 3 P a (J = 2) level to 3p 5 3d 3 P 1 and 3 P 2 , respectively, that we measure at 170.116 and 171.681Å. Three additional Ritz combinations outside of the BT18 wavelength range are reported in Table 3. The remaining two BT18 lines are at 171.26 and 172.16Å and we identify them with decays from the ( 3 P ) 1 F 3 level. Our wavelengths are 171.279Å and 172.219Å and they correspond to decays to the 3p 5 3d 3 D 2 and 1 F 3 levels, respectively. Note the latter is blended with O v in the BT18 spectrum, hence the wavelength discrepancy. A third decay to the 1 D 2 level can be identified as an enhanced wing to the strong 169.614Å line, as indicated in Figure 7. BT18 listed two strong transitions at 167.478 and 167.654Å as blends of Fe ix with Fe viii. The former identification was made on account of the BT18 atomic model predicting the strongest transition from the 3p 4 3d 2 ( 1 D) 3 D 3 a level near this wavelength. As noted earlier this prediction is at odds with other atomic calculations, and our spectra do not suggest a contribution from Fe ix. A similar problem may affect the BT18 identification of the 167.654Å line. That is, if their atomic model predicts the 3p 5 3d 3 F 2 -3p 4 3d 2 3 D 2 a to be the strongest decay, then this disagrees with our atomic calculations and those in CHIANTI. As with the 167.478Å line, we do not find evidence of an Fe ix contribution.
The remaining seven Fe ix lines listed by BT18 that are not found in our spectra were weak in their spectrum (intensities ≤ 0.1 on the BT18 scale). They could belong to other stages of ionization, but the different excitation conditions in the EBIT compared to our spark spectrum could also be responsible for the differences. The EBIT plasma has a lower density that is more typical of the solar corona.
A breakthrough in the analyses of the Fe ix 3p 5 3d -3p 4 3d 2 lines in the Sun's spectrum came from observations by Hinode/EIS. The two wavelength bands 170-212 and 246-292Å are observed with a spectral resolution of around 0.06Å. Imaging capability is critical to discriminating between neighboring ionization stages, as demonstrated by Young (2009) who compared images of coronal loop structures in lines of Fe viii through Fe x. Landi & Young (2009a) created an atlas of the Sun's spectral lines of ions formed between 10 5 K and 10 6 K from a bright point related to the footpoint region of a coronal loop. The intensities of "cold" lines were enhanced over normal values in this atlas. It permitted to make the first identification of three lines from the 3p 5 3d-3p 4 3d 2 transitions, namely, the main lines from the 3p 5 3d 3 F -3p 4 3d 2 ( 3 P ) 3 G multiplet (Young 2009) mentioned above. In an extension of this result, Young & Landi (2009) found the Ritz combinations for the 3p 4 3d 2 ( 3 P ) 3 G 4,3 levels, thus confirming their identifications. Three lines (176.959, 177.594 and 199.986Å) were added as the transitions to the 3p 5 3d 3 F 3,4 levels from the 3p 4 3d 2 ( 1 D) 3 D 3,2 levels. It should be noted that the first line in the EIS spectrum is not fully resolved from the Fe vii 176.905Å line. The wavelength of this line is 176.978Å from our fully resolved spectrum. Two other lines at 178.699 and 178.985 A were tentatively suggested as the 3p 5 3d 3 F 2 -3p 4 3d 2 ( 1 D) 3 D 1 and 3p 5 3d 3 D 3 -3p 4 3d 2 ( 3 P ) 3 F a (J = 4) transitions, respectively. We have confirmed these suggestions. In summary, ten lines in the 3p 5 3d-3p 4 3d 2 array of Fe ix were identified with specific 3p 4 3d 2 levels. Young & Landi (2009) published a list of seven additional observed lines that were suggested as being due to Fe ix based on image morphology, but for which transition identifications could not be assigned. We have identified all but one of these lines and marked them with YLTW in Table 3. On a basis of our laboratory spectrum, we have identified three additional lines in the Landi & Young (2009a) atlas. The 182.158, 188.823 and 194.806Å EIS lines are identified respectively with 3p 3 3d 3 F 4 -3p 4 3d 2 ( 3 P ) 3 F b (J = 4), 3p 3 3d 1 F 3 -3p 4 3d 2 ( 1 S) 1 G 4 and 3p 3 3d 3 D 2 -3p 4 3d 2 ( 3 P ) 3 F b (J = 3) transitions. The latter is not fully resolved from the Fe vii 194.770Å line in the EIS spectrum, and our laboratory wavelength is 194.796Å. Note that the wavelengths of all the EIS Fe ix lines from the Landi & Young (2009a) atlas listed in Table 3 are corrected for the red shift with velocity 16 km s −1 as suggested by Young & Landi (2009).
Del Zanna (2009) published a spectral atlas from a coronal loop rooted in a sunspot that also exhibited strongly enhanced cool emission lines. He confirmed the Young (2009) identifications of the three main lines of the 3p 5 3d 3 F -3p 4 3d 2 ( 3 P ) 3 G multiplet and he tentatively identified three weak transitions from the 3p 4 3d 2 configuration levels that we reject. However, the line list contains many other lines left unidentified or considered as blends that are also present in the Landi & Young (2009a) atlas and are identified in our laboratory spectrum. These lines are marked with D09 in Table 3. Their wavelengths were corrected for a redshift of 10 km s −1 .
Large scale intermediate-coupling R-matrix scattering calculations for electron collisional excitation of Fe ix were performed by Del Zanna et al. (2014). The data were used to create a new atomic model for the ion, and intensities were computed and compared with observations. Good agreement with the known Fe ix lines in the EIS spectrum was obtained. Based on this agreement a few new weak Fe ix lines were tentatively identified (Del Zanna 2009). We confirmed the identification of one line at 192.630Å as the 3p 3 3d 3 D 3 -3p 4 3d 2 ( 3 P ) 3 F b (J = 4) transition. The identification of the 194.784Å line was changed. The other suggested lines are absent in our laboratory spectra, perhaps because of different excitation conditions in the Sun and our spark spectra or they belong to Fe vii (Ekberg 1981;Kramida et al. 2022).
The 3p 5 3d-3p 5 4p lines should be also located in the studied wavelength range. As mentioned above, a line at 197.862Å in the EIS spectrum was identified as the 3p 5 3d 1 P 1 -3p 5 4p 1 S 0 transition by Young (2009). This identification was supported by the observation of two lines at 717.661 and 803.422Å, corresponding to decays to the 3p 5 4s 1 P 1 and 3 P 1 levels (Landi & Young 2009b). These lines were measured with the Solar Ultraviolet Measurements of Emitted Radiation (SUMER: Wilhelm et al. 1995) instrument. The 197.862Å line is very weak in our spectrum. We did not succeed in finding any other lines of the 3p 5 3d -3p 5 4p transitions, which possess smaller transition probabilities than the 197.862Å line. Since only one upper level is known in the 3p 5 4p configuration, we considered this configuration as "unknown" in our calculations, scaling only its average energy by a predetermined factor (Table 5). In this approach, the calculated energy of the 3p 5 4p 1 S 0 level is lower than the experimental one by about 3900 cm −1 .

CONCLUSIONS
Using high-resolution laboratory spectra in the 110-200Å range the analysis of energy levels and spectral lines of Fe ix was greatly extended. Many weak lines were added to the previous analyses of the 3p 5 3d-3p 5 4f transitions extending to 25 the number of identified lines. Seventy-three lines of the 3p 5 3d-3p 4 3d 2 transition array were identified, bringing to 81 the number of known lines in this transition array. A number of lines assigned to Fe ix were identified in Hinode/EIS spectra. The data can be used for diagnostics of solar plasma and provide a benchmark for further development of atomic theory. P.R. Young acknowledges support from the NASA Heliophysics Data Environment Enhancements program and the NASA Individual Scientist Funding Model competitive work package program. CHIANTI is a collaborative project involving NASA Goddard Space Flight Center, George Mason University, the University of Michigan (USA), and the University of Cambridge (UK).
a Character of the observed line: db -intensity is shared by two transitions; bl -blended line (the blending species are given in parentheses where known): VII -Fe vii, VIII -Fe viii, XI -Fe xi; ? -identification is uncertain; p -perturbed by a stronger nearby line (both the wavelength and intensity may be affected); w -wide line.
b Relative intensities are given on an arbitrary linear scale (see text).
c Weighted transition probability (g is the statistical weight of the upper level) in 10 9 s −1 unit.
d Wavelength derived from the final level energies (Ritz wavelength).
e Difference between the observed and Ritz wavelengths (blank for lines that solely determine the upper level).
f Designation is restricted to a term of the 3p 4 sub-shell followed by a final term and a letter (a or b) distinguishing different terms of the 3d 2 configuration in the case the final term is repeated. For full designation see Table 4. h Lines are not fully resolved from, respectively, Fe vii 176.905 and 194.770Å lines in EIS spectra.
i Line is blended with O v line in EBIT spectrum. Table 4. Energy levels of the 3p 4 3d 2 configuration of Fe ix higher than 930 000 cm −1 .
c LS -composition of the level eigenvector. The 3p 4 configuration terms and the final terms are listed. The letters after the final terms distinguish different terms of the 3d 2 configuration. Full descriptions for the first component are shown in square brackets. * -( 2 S) 1 D and ( 2 S) 3 D terms belong to the 3s3p 6 3d configuration; ** -( 2 P ) 3 D stands for 3p 5 ( 2 P )4p( 2 P ) 3 D.
d Number of observed lines determining the level value in the least-squares optimization procedure. Table 5. Hartree-Fock with relativistic corrections (HFR) and least-square-fitted (LSF) parameter values (cm −1 ) with their uncertainties (Unc.) in Fe ix.