FORBIDDEN AND INTERCOMBINATION LINES OF RR TELESCOPII: WAVELENGTH MEASUREMENTS AND ENERGY LEVELS

The 2s²2p³(⁴S)3s³S₁ level in O i is believed to be excited in giant stars through radiative pumping of the 2s²2p³(⁴S)3d³D_J levels by H i Lyβ: the ³D levels decay to 2s²2p³(⁴S)3p³P_J, which in turn decay to the ³S₁ level, strongly enhancing the level's population (Haisch et al. 1977). ³S₁ decays to the three ³P_J levels in the ground configuration, giving rise to lines at 1302.2, 1304.9, and 1306.0 Å. Since λ1302.2 is a decay to the ground level of O i, interstellar neutral oxygen strongly absorbs the stellar spectrum at this wavelength, but the high radial velocity of the star leaves a portion of the stellar emission line which, however, is not useful for the present study. λ1304.9 is completely removed from the spectrum by interstellar Si ii λ1304.370. The remaining line, λ1306.0, is unaffected by interstellar absorption.

A further decay from ³S₁ occurs to the ¹D₂ level in the ground configuration giving a line at 1641.305 Å, and this is found in the long wavelength wing of the very strong He ii λ1640.4. The line thus sits on a sloping background and was fit here with a single Gaussian rather than by attempting a double-Gaussian fit to the two lines together. The width of λ1641.305 is very narrow at 11.4 km s⁻¹, compared to 41.5 km s⁻¹ for λ1306.0. This implies that λ1306.0 is optically thick, which is confirmed by the flux ratio of 1.6 (λ1641.3/λ1306.0) which is much larger than the theoretical branching ratio of 2 × 10⁻⁵. For this reason λ1306.0 is not included in Table 1.

Four lines are available for C iii with the strongest line, λ1908.7, showing a clearly asymmetric line profile with an extended long wavelength wing. The wing is likely related to the redshifted, low density plasma components discussed in Section 3. The line profile was fitted with two Gaussians, each forced to have the same width, and the stronger, short wavelength Gaussian was assumed to correspond to the radial velocity of the system and the wavelength is given in Table 1. The weak forbidden line λ1906.7 is found to be redshifted relative to the expected wavelength by around 20 km s⁻¹ and the Appendix demonstrates that this is consistent with the line being formed in the redshifted, low density plasma component. It is therefore not useful as a wavelength fiducial here.

The remaining two lines, λ1247.4 and λ2297.6, are significantly weaker than λ1908.7 but have very similar line widths to this line and so appear to be unblended. Both lines are listed in Table 1.

The 2s2p³P_J–2p² $^3P_{J^\prime }$ multiplet is found at around 1175 Å but the spectrum is noisy in this region and ratios of the multiplet's lines vary significantly from the optically thin case suggesting the lines are optically thick and possibly absorbed by the interstellar medium so they are not used as velocity references.

The intercombination lines at 1660.8 and 1666.2 Å are very strong in RR Tel and show asymmetric profiles similar to C iii λ1908.7 discussed in the previous section. The centroids were derived in the same manner as the C iii line using two Gaussians of equal width. The O iii lines are found in both the SW and MW spectra but for both they are found to be around 3 km s⁻¹ blueshifted relative to the other wavelength reference lines. For this reason they have not been used as reference lines.

The forbidden ³P₁–¹S₀ transition at 2321.7 Å is strong in the RR Tel spectrum and has an asymmetric profile, with the long wavelength wing being significantly stronger than for the intercombination lines. Since the O iii forbidden lines in the visible have unusual profiles (Section 3) it was decided not to include the λ2321.7 line in the present analysis.

Ten Bowen fluorescence lines of O iii are found between 2800 and 3060 Å. The strongest lines, λ2837 and λ3048, clearly show asymmetric profiles similar to λλ1660.8, 1666.2. The three lines at 2819.5, 3036.3, and 3060.2 Å were chosen as velocity references as they are each isolated in the spectrum and so unaffected by blending.

The strong λ2796.4, 2803.5 resonance lines are not suitable as wavelength references as they are redshifted relative to other species, suggesting they are affected by P Cygni like absorption on the short wavelength side of their profiles. In addition, both lines have strong interstellar absorption features on their long wavelength sides. The much weaker 3p²P_3/2–3d²D_5/2 transition lies between the two strong resonance lines at 2798.82 Å and is suitable as a wavelength reference. A further 3p–3d transition occurs at 2791.60 Å but is blended, probably with an Fe v transition. 3p²P_3/2–4s²S_1/2 is also found in the STIS spectrum at 2937.37 Å but it is very weak and is not used as a wavelength reference here.

The intercombination line at 2669.9 Å is unblended and does not show an asymmetric profile. Another Al ii line is the strong resonance transition at 1670.8 Å, however this clearly shows interstellar absorption on the long wavelength side of the profile and so is not suitable as a velocity reference.

The intercombination lines, 3s²3p²P_J–3s3p² $^4P_{J^\prime }$, are found between 2329 and 2351 Å. The weakest line, λ2329.24, is too faint to be observed while λ2344.92 is coincident with Fe ii interstellar absorption and is not seen. The line at 2350.89 Å is close to Si vii λ2350.73, but this line will be very weak based on the strength of Si vii λ2147.40 and so we use λ2350.89 as a reference. The strongest line of the Si ii multiplet is λ2335.32 however this is notably broader than λ2350 which is likely due to blends from Si ii λ2335.12 (part of the same multiplet) and Fe ii λ2335.18. It is not possible to separate these components and so we do not use λ2335.32 as a wavelength reference.

As with other strong resonance lines, Si iii λ1206.5 is redshifted relative to other lines in the spectrum, and it also shows interstellar absorption on the long wavelength side of the profile. We thus do not use it as a wavelength reference. The intercombination line, λ1892.0, is actually much stronger than λ1206.5 and, like other strong intercombination lines in the spectrum, has an asymmetric line profile. The line has been fit with two Gaussians of equal width and the shorter wavelength component is taken to be at the radial velocity of the star. λ2542.6 is a much weaker line, but is narrow and unblended and we list it in Table 1.

The 3s²3p^{2 3}P_{1, 2}–3s3p^{3 5}S₂ intercombination lines at 1713.1 and 1728.9 Å are narrow, although fairly weak, and suitable as wavelength references.

Fe ii gives rise to more lines than any other species in the RR Tel spectrum and accurate wavelengths are known for many of them. However, many of the Fe ii lines are weak and/or blended which means care has to be taken in selecting lines as wavelength references. In addition, it is noticeable that a number of the stronger Fe ii transitions are broad and redshifted relative to the other transitions. Examples include a⁶D_7/2–z⁶D_9/2 (λ2626.451), a⁶G_9/2–z⁶H_11/2 (λ2459.528), and a⁶D_9/2–z⁶F_11/2 (λ2382.765), which have velocities of −60 to −62 km s⁻¹ and widths of 30–45 km s⁻¹, compared to −66 to −70 km s⁻¹ and 15 to 25 km s⁻¹ for more typical lines.

There are few Fe ii lines in the STIS SW spectrum and we have used three lines between 1360 and 1414 Å. λ1360.2 is a narrow, well-observed line which was first identified as being fluoresced by H i Lyα by Johansson & Carpenter (1988), although O v λ1218.3 may also contribute to the radiative pumping (Hartman & Johansson 2000). λ1392.1 and λ1413.7 are both excited through radiative pumping by He ii λ1084.9 and were identified by Hartman & Johansson (2000) and Jordan & Harper (1998), respectively.

Only two Fe ii lines in the MW spectrum are used as wavelength fiducials, and both arise from the z⁴H_11/2 level which is radiatively pumped by C iv λ1548.2 (Johansson 1983) giving rise to 10 lines in all that are very prominent in the RR Tel Fe ii spectrum. Of the 10 lines three are anomalously broad, indicating blending, and a fourth (λ2168.105) shows an anomalous blueshift. The six remaining lines have narrow widths between 20 and 25 km s⁻¹ and velocity shifts between −67 and −70 km s⁻¹ and have been used as reference lines.

The remaining Fe ii lines used for the wavelength calibration are all found in the LW spectrum and arise from four multiplets that are excited through radiative pumping by H i Lyα, either directly or by cascading from fluoresced levels. There are four groups of lines in all, three corresponding to UV multiplets 78, 391, and 399, and a fourth corresponding to the unnumbered multiplet z⁴P–e⁴D.

UV multiplet 78 (a⁴P–z⁴P) gives rise to seven lines between 2944 and 3003 Å, which is a region less crowded than other parts of the STIS RR Tel spectrum. One line (λ2965.489) is a known blend with another Fe ii transition, but the remaining six lines have narrow widths between 20 and 23 km s⁻¹ and velocities ranging from −67 to −71 km s⁻¹, with an average of −68.9 km s⁻¹.

Five lines from UV multiplet 391 (z⁴F–e⁴D) are found in the RR Tel spectrum between 2840 and 2867 Å. Each line is unblended and the line widths range from 13 to 19 km s⁻¹, and the velocities from −70 to −71 km s⁻¹.

UV multiplet 399 (z⁴D–e⁴D) is also emitted from the e⁴D levels, and seven lines are found in the RR Tel spectrum between 2845 and 2886 Å. One line (λ2885.611) is significantly broader than the others, indicating it is blended. The remaining lines have narrow widths between −15 and −18 km s⁻¹ and velocities between −70 and −72 km s⁻¹.

Six lines are observed from the unnumbered UV multiplet z⁴P–e⁴D between 3037 and 3080 Å. Five of these lines have never previously been reported in the RR Tel spectrum, with the remaining line (λ3079.574) first being reported by Jordan & Harper (1998). All of the lines appear to be unblended, with narrow line widths of between 16 and 25 km s⁻¹ and velocities between −67 and −73 km s⁻¹.

The intercombination line, λ1486.5, is strong and lies at the edge of two spectral orders in the SW spectrum. The line has an asymmetric line profile with an extended long wavelength wing and has been fit with two Gaussians forced to have the same width. The stronger, short wavelength component is assumed to be at the rest wavelength of the system and the average wavelength from the two spectral orders was used to derive the wavelength given in Table 3. The wavelength is in good agreement with the NIST wavelength, the previous RR Tel measurement of Penston et al. (1983), and the solar measurements of Doschek et al. (1976b) and Sandlin et al. (1977).

The nearby forbidden line, λ1483.3, is measured but shows a significant redshift relative to the NIST wavelength: the measured wavelength is 1483.385 ± 0.031 Å compared to the NIST value of 1483.321 Å, equating to a velocity shift of 13 km s⁻¹. This is consistent with the corresponding forbidden line of C iii (Section 5.1.2) and suggests that it too arises from the redshifted plasma component.

The forbidden and intercombination lines lie either side of the very broad interstellar absorption line of H i Lyα at 1213.8 and 1218.3 Å, respectively. The forbidden line was not reported by Penston et al. (1983) from IUE spectra of RR Tel, but is clearly seen in the STIS spectrum. It lies on the sloping wing of the Lyα absorption feature and thus the long wavelength wing will be more absorbed than the short wavelength wing which may create a false blueshift for the line profile. Due to the higher IP of O v we do not expect the forbidden line to be emitted from the low density plasma component as was found for C iii and N iv. The measured centroid lies blueward of the only previous measurement of the λ1213.8 line (Sandlin et al. 1977), but the wavelength is in excellent agreement with the NIST wavelength.

As with other intercombination lines, λ1218.3 shows an asymmetric profile with an extended long wavelength wing. The effect is more pronounced in this case, however, because of the Lyα absorption on the short wavelength side of the profile. The line has been fit with two Gaussians forced to have equal width and with the short wavelength side of the profile partly masked off to prevent the Lyα absorption distorting the fit. The strongest of the fitted Gaussians is taken to correspond to the rest component of the system and yields the wavelength given in Table 3 which is in good agreement with the previous measurements of Doschek et al. (1976b), Sandlin et al. (1977), and Penston et al. (1983).

All five of the C ii intercombination lines are found in the RR Tel spectrum. The four strongest lines all show asymmetric profiles with enhanced long wavelength wings, similar to other intercombination lines in the spectrum, and they were fit with two Gaussians forced to have the same width. The wavelength of the stronger, short wavelength component was used to generate the rest wavelengths in Table 3. For the weakest transition, ²P_1/2–⁴P_3/2, the asymmetry is not pronounced due to the lower signal to noise and a single Gaussian fit was used. The 3/2–3/2 transition at 2327.6 Å has a much more extended long wavelength wing than the other lines which is likely due to the Fe ii a⁶D_5/2–z⁶P_3/2 transition at 2328.111 Å. This part of the profile was thus not included in the two Gaussian fit to the C ii line.

The C ii intercombination lines have been previously measured in off-limb solar spectra by Doschek et al. (1977), and in the IUE spectrum of RR Tel by Penston et al. (1983) and the comparison in Table 3 shows good agreement except for the λ2324.3 and λ2328.9 lines, where the Penston et al. (1983) wavelengths are significantly shorter than the current wavelengths. (They are also significantly shorter than the Doschek et al. 1977 wavelengths.)

One discrepancy that is present for all three sets of measurements comes from deriving the ground ²P_J level splitting by using the two pairs of transitions ²P_{1/2, 3/2}–⁴P_1/2 and ²P_{1/2, 3/2}–⁴P_3/2. For the current wavelength measurements⁵ the latter transition pair yields a splitting of 62.79 ± 0.16 cm⁻¹, while the former yields 63.89 ± 0.10 cm⁻¹, neither of which is consistent with the very accurate value of 63.39 cm⁻¹ from Cooksy et al. (1986). The STIS Data Handbook (Bostroem & Proffitt 2011) states that relative wavelength accuracy within the echelle orders is good to within 0.2 pixels, which translates to 0.008 Å in the region of the C ii lines, but this is not sufficient to explain the level splitting discrepancy. This discrepancy is a puzzle and perhaps is related to the non-Gaussian shapes of the line profiles.

The five intercombination lines are found within 8 Å of each other and all have very similar line widths of around 30 km s⁻¹, suggesting they are unblended. The two strongest transitions, λ1749.7 and λ1752.1, clearly show asymmetric profiles similar to other intercombination lines in the spectrum and have been fit with two Gaussians forced to have the same width. The stronger, shorter wavelength Gaussians are used to derive the rest wavelengths in Table 3. Single Gaussian fits were used for the remaining transitions. The measured wavelengths are in excellent agreement with the solar measurements of Doschek et al. (1976b) and Sandlin et al. (1977). The agreement with the wavelengths of Penston et al. (1983) from IUE spectra of RR Tel is slightly less good, but consistent within the uncertainties.

The O iv intercombination lines are very strong in RR Tel and have been studied in some detail by previous authors (Harper et al. 1999; Keenan et al. 2002). As with other strong intercombination lines such as C iii λ1908.7, O iii λλ1660.8, 1666.2, and Si iii λ1892.0 the O iv lines have asymmetric profiles with the long wavelength wings of the lines being more extended than the short wavelength wings. They have been fit with two Gaussians forced to have the same width, with the stronger component being taken as that of the rest component of the nebula.

Three previous measurements of the lines' wavelengths have been made from RR Tel spectra by Penston et al. (1983), Harper et al. (1999), and Keenan et al. (2002) from IUE, HST/GHRS, and HST/STIS spectra, respectively, the latter work using the same spectra as used here. In addition, the lines have also been measured from solar spectra by Doschek et al. (1976b) and Sandlin et al. (1977). Comparing the results in Table 3 it is clear that the Keenan et al. (2002) wavelengths are blueshifted relative to the other results by around 0.02–0.06 Å. We believe this is because the authors used the nearby Si iv resonance lines to determine a rest wavelength scale. We find that these lines, like other resonance lines in the spectrum, are redshifted relative to the system's radial velocity by 10 km s⁻¹ and so lead to the wavelength offset for the O iv lines.

The λ1404.8 line is known to blend with an S iv transition, but Harper et al. (1999) demonstrated that this line contributes only 4% to the measured feature's flux and so the measured line centroid is a reliable measure of the O iv line's wavelength.

The present wavelengths tend to be midway between those of Penston et al. (1983) and Harper et al. (1999), the latter's results being close to those of Sandlin et al. (1977). The wavelength separations of the lines are consistent between all of the measurements to within a few mÅ. The NIST wavelengths show significant discrepancies with the astrophysical wavelengths and should be revised.

The model presented in the Appendix suggests that N ii λ6585.3, which is emitted from the ¹D₂ level, principally comes from the low density plasma component, whereas λ5756.2, which is emitted from the higher energy ¹S₀ level, principally comes from the high density plasma component. The λ6585.3 line profile shows three distinct components, the strongest of which corresponds to the redshifted component seen in O ii λ3729.9, and so is consistent with the model. The additional components are at the rest wavelength of the system and at a velocity of around −30 km s⁻¹ (this is a broad component). The profile thus seems to be midway between O ii λ3729.9 and O iii λ5008.2. The weaker ³P₂–¹D₂ transition at 6549.9 Å shows a similar structure (as expected since the lines share the same upper level) but the line seems to be affected by a blend in the short wavelength wing. We do not include these two transitions in Table 3 on account of the complexity of the line profiles.

As expected from the emission model of the Appendix, the ¹D₂–¹S₀ emission line at 5756.2 Å has a much simpler profile than λ6585.3. Gone is the broad, blueshifted component at −30 km s⁻¹, while the redshifted component is weaker than the rest component. Fitting the line with two Gaussians yields the wavelength for the rest component in Table 3, which is in good agreement with the NIST wavelength and the value of Bowen (1960).

The ³P₁–¹S₀ transition at 3063.7 Å is partly blended with an O iv recombination line at 3064.3 Å but a weak component in the short wavelength wing of this line can be identified as the N ii transition. A two Gaussian fit was performed and the wavelength of the weak component is given in Table 3. Since λ3063.7 is emitted from the same upper level as λ5756.2 then it will be expected to show the same two-component structure as this line. It is not possible to resolve the redshifted component on account of the blending O iv line. For this reason the wavelength given in Table 3 should be treated with caution and the more accurate value of Bowen (1960) is preferred.

The intercombination lines occur at 2139.7 and 2143.5 Å for N ii, but the latter is blended with an Fe vii transition (see the discussion in Young et al. 2005a) and the wavelength cannot be accurately estimated. λ2139.7 shows an extended long wavelength wing like other intercombination lines in the RR Tel spectrum and the feature has been fit with two Gaussians forced to have the same width. The wavelength of the stronger, short wavelength component is given in Table 3 and is found to be in excellent agreement with the NIST wavelength but discrepant with the measurement of Penston et al. (1983) from IUE spectra of RR Tel.

Six Ne v forbidden lines and one intercombination line are found in the STIS and UVES spectra. The longest wavelength lines are the decays of the ¹D₂ level to the ³P₁ and ³P₂ levels, giving two strong lines at 3347.0 and 3427.0 Å, respectively. The UVES wavelengths agree with those of Bowen (1960) within the uncertainties, while the separation of the two UVES lines implies a ³P₁–³P₂ separation of 698.3 ± 0.8 cm⁻¹ in good agreement with the measurement of 698.242 ± 0.010 of Feuchtgruber et al. (1997) from infrared spectra. The ¹D₂ level also yields a weak decay to ³P₀ at 3301.3 Å but this is blended with a much stronger O iii line at 3300.3 Å and cannot be measured.

The ¹S₀ level also decays to the ³P₁ and ³P₂ levels, giving two lines at 1574.7 and 1592.2 Å. The latter is a weak line and was not measured in IUE spectra of RR Tel (Penston et al. 1983). The separation of the two lines implies a ³P₁–³P₂ separation of 698.6 ± 2.2 which is in good agreement with the measurement of 698.242 ± 0.010 of Feuchtgruber et al. (1997). The λ1574.7 wavelength is in very good agreement with that of Penston et al. (1983).

The ¹D₂–¹S₀ transition is found in the STIS spectra at 2974.0 Å and agreement within the uncertainties is found with the previous measurement of Penston et al. (1983). Combining the measured wavelength of this line with that of λ3346.8 yields a predicted wavelength of 1574.699 Å for the ³P₁–¹S₀ which is 5.3 km s⁻¹ longward of the measured wavelength for this transition (see Section 5.3).

The intercombination transition, 2s²2p^{2 3}P₂–2s2p^3 5S₂, occurs at 1145.6 Å and is the shortest wavelength line found in the STIS spectrum. The measured wavelength is a little shorter than the solar measurements presented by Sandlin et al. (1977) and Young et al. (2005b), but consistent within the uncertainties.

Measurements of the Na vi forbidden lines have not previously been reported in the literature, and the estimated wavelengths of Bowen (1960) have large uncertainties. However, Edlén (1972) provided calculated energy levels that yield accurate wavelengths. Note that these energies are found to be significantly more accurate than those contained in the NIST database. Four Na vi forbidden lines are found in the RR Tel STIS spectra, and the two strongest are the ³P_{1, 2}–¹D₂ transitions at 2872.65 and 2971.79 Å, respectively, which are close to the predicted wavelengths of Edlén (1972): 2872.59 and 2971.65 Å, respectively. The separation of the lines implies a ³P₁–³P₂ splitting of 1161.3 ± 1.0 cm⁻¹, in good agreement with the infrared measurement of Feuchtgruber et al. (1997) who found 1161.36 ± 0.12 cm⁻¹.

The ¹D₂–¹S₀ and ³P₁–¹S₀ transitions are both weak and Edlén (1972) predicts wavelengths of 2569.71 and 1356.36 Å, respectively. The former is close to a broad, weak line in the STIS spectrum at 2569.59 Å that we identify with the Na vi transition. A clump of five emission lines is found in the STIS spectrum at 1356 Å, and one at 1356.32 Å is close to the Edlén (1972) wavelength and has a line width consistent with the other Na vi lines. Another line at 1355.94 Å has a similar flux and width, and thus is another potential candidate for the Na vi transition. However, by using the measured wavelengths of the ³P₁–¹D₂ and ¹D₂–¹S₀ transitions we can predict a wavelength of 1356.34 ± 0.03 Å for the ³P₁–¹S₀ which is consistent with the observed 1356.32 Å line.

The ²D–²P transitions lie between 7321 and 7333 Å and so outside the UVES wavelength range, while the ⁴S–²D transitions were discussed in Section 3 where they were found to be emitted from a redshifted plasma component.

The ⁴S_3/2–²P_{1/2, 3/2} lines occur at 2471.0 and 2471.1 Å and are found blended in a single spectral feature in the STIS spectrum. The CHIANTI atomic model predicts that λ2471.1 should be around four times stronger than the companion line, and also that both lines are significantly more sensitive to high densities than the λλ3727.1,3729.9 line pair. The latter point means that the ⁴S_3/2–²P_{1/2, 3/2} transitions may have a significant component from the rest component of the plasma (see also the Appendix), unlike the ⁴S–²D transitions. The observed line profile does not show any clear asymmetry nor any evidence of extended wings. Fitting it with a single Gaussian and assuming it is entirely due to the ⁴S_3/2–²P_3/2 transition yields the wavelength shown in Table 3, which is discrepant with both the NIST wavelength and the Bowen (1960) wavelength. Given the uncertainty over the contribution from the redshifted, low density plasma and degree of blending we advise the reader to treat the present λ2471 wavelength measurement with caution.

Wavelengths and energy levels of Ne iv were assessed by Kramida et al. (1999). The ²D–²P forbidden transitions lie between 4715 and 4727 Å and so are not found in the present UVES spectra, thus the wavelengths of Bowen (1960) still represent the best measurements of these lines. The ⁴S_3/2–²P_{1/2, 3/2} transitions at 1601.5 and 1601.7 Å are close in wavelength and have not been resolved in the previous measurements of Sandlin et al. (1977) and Penston et al. (1983). The line in the STIS spectrum is clearly asymmetric, suggesting two lines with the weaker lying in the long wavelength wing of the stronger line. Fitting the feature with two Gaussians forced to have the same width yields the wavelengths listed in Table 3.

The ⁴S_3/2–²D_{3/2, 5/2} transitions are separated by around 3 Å, but both are partly blended with other, narrow lines. Simultaneous two Gaussian fits were performed to each feature to resolve the components. The stronger of the two Ne iv lines, λ2422.4, has an unidentified narrow line in the short wavelength wing, which is likely an Fe ii transition. The weaker Ne iv line, λ2425.0, also has a narrow line in the short wavelength wing that can be identified with the Fe ii b⁴F_9/2–y⁴G_11/2 transition (λ2424.883). The widths of the two Ne iv lines are 43 and 41 km s⁻¹, respectively, which are in very good agreement with the width of 41 km s⁻¹ found for the ⁴S_3/2–²P_{1/2, 3/2} transitions, giving confidence in the two Gaussian fit employed for these lines.

The analysis presented in the Appendix suggests that the Ne iv ⁴S_3/2–²D_{3/2, 5/2} transitions may be principally formed in the redshifted, low density plasma component of the nebula, unlike the ⁴S_3/2–²P_{1/2, 3/2} transitions. This may explain the wavelength differences compared to Penston et al. (1983) and Jordan & Harper (1998), although this would imply that the low density plasma component was not present at the time of these earlier RR Tel observations.

Since the four decays to the ground level described above directly yield the energies of the four excited levels in the ground configuration, one can then derive wavelengths for the four ²D–²P transitions and compare with the wavelengths presented by Bowen (1960). The uncertainties on the derived wavelengths are around 0.35 Å, significantly larger than those of Bowen (1960) which are 0.04 Å. We find agreement within these uncertainties for all of the transitions except ²D_3/2–²P_3/2 for which the derived air wavelength is 4723.75 Å and the Bowen (1960) wavelength is 4724.15 Å. Note that this discrepancy could be explained if the ⁴S_3/2–²D_{3/2, 5/2} transitions are formed in the redshifted, low density plasma component.

The four ²D–²P transitions are expected to lie between 4012 and 4026 Å and so are not found in the UVES spectra. The ⁴S_3/2–²D_{3/2, 5/2} and ⁴S_3/2–²P_{1/2, 3/2} transitions occur at ultraviolet wavelengths and three of the transitions can be identified in the STIS spectra. Edlén (1972) provided calculated energies for the ground levels of Na v and these yield predicted wavelengths for the ⁴S_3/2–²D_{3/2, 5/2} transitions of 2067.85 and 2069.81 Å. A line is found at the former wavelength but it is blended with the n = 33 member of the He ii Fowler series. However, by comparison with other members of the Fowler series it is clear that Na v provides the dominant contribution to the blend and so we associate the measured wavelength with Na v. The ⁴S_3/2–²D_3/2 Na v transition is also blended with a Fowler series line, in this case the n = 31 member, and it lies in the short wavelength wing of a much stronger line that we believe is due to O vi. The strength of the Na v–He ii blend is consistent with the line predominantly arising from He ii and so we do not associate the measured wavelength with Na v.

As for Ne iv, the Appendix suggests that the ⁴S–²D transitions may be predominantly formed in the redshifted, low density plasma component, therefore readers are recommended to treat the rest wavelength for λ2069.9 in Table 3 with caution.

Penston et al. (1983) identified a line at 1365.37 Å that the authors identified with both of the ⁴S_3/2–²P_{1/2, 3/2} transitions. The STIS spectra clearly resolve both components at 1365.39 and 1366.08 Å, with the former being stronger by a factor of three thus the Penston et al. (1983) wavelength seems to correspond only to the ⁴S_3/2–²P_3/2 transition. The calculated energy values of Edlén (1972) yield predictions of 1365.44 and 1366.08 Å for the two transitions, in good agreement with the STIS measurements.

The four ²D_{3/2, 5/2}–²P_{3/2, 1/2} transitions are found in the UVES spectra between 3488 and 3504 Å. The 5/2–1/2 transition is weak and difficult to measure so is not listed in Table 3. Wavelengths were given for all four lines by Bowen (1960) but these were obtained by calculation and are only accurate to ±3 Å. The three lines observed in the UVES spectra allow the splittings of the ²P and ²D terms to be derived and we obtain values of 14.7 ± 0.8 and 108.7 ± 0.8 cm⁻¹ for the ²D and ²P terms, respectively.

The ⁴S_3/2–²P_{3/2, 1/2} transitions give rise to two strong lines at 1190.0 and 1191.6 Å, respectively, with widths of 69 and 75 km s⁻¹. λ1190.0 is partly blended with Mg vii λ1189.9, but based on the flux of the unblended Mg vii λ2629 line we estimate this contributes less than 1%. The separation of the λ1190.0 and 1191.6 lines yields a separation for the ²P levels of 109.2 ± 2.4 cm⁻¹, in good agreement with the previous determination.

The ⁴S_3/2–²D_{3/2, 5/2} lines are close in wavelength and the observed feature in RR Tel shows a strong line with an extended long wavelength wing (Figure 3). A two Gaussian fit was performed, forcing the two lines to have the same width, however the resulting wavelengths are not consistent with the separation of the ²D_{3/2, 5/2} levels obtained above: the implied energy separation is 19.2 ± 0.6 cm⁻¹. The λ1190.0 and λ1191.6 lines both show some structure in their line profiles beyond a simple Gaussian shape and this is also seen in λ1806, it thus seems that the weak ⁴S_3/2–²D_5/2 transition (which is expected to be around a factor of 10 weaker than its neighbor) is not correctly extracted by assuming a two Gaussian fit. In Table 3 we list only the ⁴S_3/2–²D_3/2 transition.

As discussed in Section 5.3, the Mg vi lines are useful for checking the wavelength scales of the UVES and STIS spectra, however a significant discrepancy was found that led to an additional source of uncertainty in the present analysis. The measured wavelengths of λλ3503.2, 3489.9, and 1805.9 yield predictions for the short wavelength lines of 1190.068 and 1191.610 Å that are up to 0.028 Å different from the measurements—a difference of 7.1 km s⁻¹.

The strongest line from Ne iii is the ³P₂–¹D₂ transition at 3869.8 Å and we find that it has a distinctive "shoulder" on the long wavelength side of the profile that is around half the strength of the main line. We believe this is emission from the redshifted plasma component that is prominent in the cooler ions (Section 3 and the Appendix). The line has thus been fit with two Gaussians and the shorter wavelength Gaussian gives the wavelength quoted in Table 3, which is in good agreement with the value of Bowen (1960).

The ¹D₂–¹S₀ transition, λ3343.4, is much weaker than λ3869.8 and shows an extended long wavelength wing that is relatively less intense than the shoulder of the λ3869.8 line. λ3343.4 is more sensitive to high densities than λ3869.8 and so the weaker wing found for this line is due to the lower density of the redshifted plasma component.

The ³P₁–¹S₀ transition is found at 1814.65 Å and there is no evidence of a long wavelength wing to the profile, although the signal in the line is weaker than the longer wavelength lines.

The three Ne iii lines allow a consistency check to be performed on the UVES and STIS wavelength scales (Section 5.3), although it is necessary to use the ³P₂–³P₁ separation of Feuchtgruber et al. (1997) to complete the calculation. We find a predicted wavelength for the ³P₁–¹S₀ transition of 1814.636, within 0.01 Å of the STIS measurement.

The Na iv ³P_{1, 2}–¹D₂ transitions are found in the UVES spectra at 3242.7 and 3363.3 Å and the measured wavelengths are consistent with those measured by Bowen (1960), although the UVES spectra have smaller uncertainties. Transitions from the ¹S₀ level are expected in the STIS wavelength range, but none can be identified.

The Mg v forbidden lines all lie in the ultraviolet part of the spectrum and so Bowen (1960) was only able to give approximate wavelengths based on calculations and extrapolation along the isoelectronic sequence. Four lines are found in the STIS spectrum, one of which has not previously been reported.

Penston et al. (1983) reported the ³P_{1, 2}–¹D₂ transitions from the IUE spectrum of RR Tel and the STIS wavelengths are in excellent agreement (Table 3). The separation of the two lines implies a splitting of the ³P_{1, 2} levels of 1782.7 ± 1.0 cm⁻¹, in good agreement with the measured infrared value of 1782.58 ± 0.20 cm⁻¹ of Feuchtgruber et al. (1997). The infrared measurement of the ³P₁–³P₀ splitting (also Feuchtgruber et al. 1997) can be used to predict a wavelength for the ³P₀–¹D₂ transition of 2993.84 Å. This is close to a line at 2993.73 Å that has a width consistent with the other members of the multiplet, but the wavelength discrepancy is outside the uncertainties and the strength of the observed line is also larger than expected so we do not make the identification.

The ¹D₂–¹S₀ transition at 2417.6 Å has not previously been reported in the literature, but is clearly seen in the STIS spectrum and the wavelength is given in Table 3. The ³P₁–¹D₂ transition at 1324.4 Å has been measured in solar spectra by Sandlin et al. (1977) and in IUE spectra of RR Tel by Penston et al. (1983) and both measurements are in good agreement with the STIS wavelength (Table 3).

Combining the ¹D₂–¹S₀ and ³P₁–¹D₂ wavelengths yields a predicted wavelength of 1324.427 Å which is within 2 km s⁻¹ of the measured position, and confirms the identification of the λ2417.6 line.

Finally, we note that there are significant discrepancies between the wavelengths derived from the NIST energy levels and the present measurements, suggesting the NIST database needs to be updated.

Two Al vi lines are found in the STIS spectrum: the ³P_{2, 1}–¹D₂ transitions at 2430.2 and 2603.1 Å, respectively. The former was previously identified by Jordan & Harper (1998) and the present wavelength is in good agreement with their value. λ2603.1 has not previously been reported.

The P iv intercombination line was first measured in the laboratory by Robinson (1937) and later by Zetterberg & Magnusson (1977), and their values of 1467.424 and 1467.427 Å, respectively, are in good agreement with the present measurement (Table 3). The line has also been measured from solar spectra by Sandlin et al. (1977) and from IUE spectra of RR Tel by Penston et al. (1983), and agreement is again good.

The S v intercombination line is close to a strong interstellar absorption line of N i λ1199.55, but the line profile is not affected and a good measurement of the line centroid can be made. The wavelength agrees with the solar measurement of Sandlin et al. (1977) and the Penston et al. (1983) value from the IUE spectra of RR Tel.

The S iv transitions lie close in wavelength to the stronger intercombination transitions of O iv (Section 6.2.3), and the STIS RR Tel lines have previously been studied by Keenan et al. (2002). The 1/2–1/2 transition at 1404.8 Å is blended with one of the O iv transitions and Keenan et al. (2002) found that the S iv transition contributes less than 2% to the observed line's intensity so the line cannot be used to determine a rest wavelength.

The strongest transitions, λ1406.0 and λ1416.9, are both unblended and well-observed in the spectrum. As with other intercombination lines in the RR Tel spectrum, they show asymmetric line profiles and have been fitted with two Gaussians forced to have the same width. The stronger, shorter wavelength component is used to derive the rest wavelengths, which are in good agreement with previous measurements (Table 3) except for Keenan et al. (2002). As mentioned in Section 6.2.3, this is probably due to the use of the Si iv resonance lines as wavelength fiducials by these authors.

The 1/2–3/2 transition at 1398.0 Å is extremely weak but can be measured, although only a single Gaussian fit was used due to the low signal to noise. The λ1398.0 line has only been measured previously from RR Tel spectra, and there is a significant discrepancy between the present wavelength and Penston et al. (1983) value. It is possible that the Penston et al. (1983) measurement was affected by the nearby Fe ii λ1397.845 line which is clearly resolved in the STIS spectrum, but may have blended with the S iv line in the IUE spectrum.

The 3/2–1/2 transition at 1423.8 Å is blended with a broad spectral feature, and there is also a narrow Fe ii line nearby. Performing a three Gaussian fit yields the wavelength for the S iv line given in Table 3. Agreement is found with Sandlin et al. (1977) and Harper et al. (1999).

The three decays from the ¹D₂ level to ³P_J lie outside the wavelength range of UVES, between 7260 and 8050 Å. The ³P₂–¹S₀ transition is blended with the n = 5 member of the He ii Fowler series at 3204 Å and cannot be resolved. The remaining transitions, ³P₁–¹S₀ and ¹D₂–¹S₀, are observed in the UVES spectrum and the wavelengths are given in Table 3, where good agreement is found with the values of Bowen (1960).

The ³P₂–¹D₂ and ¹D₂–¹S₀ transitions lie outside the UVES wavelength ranges, at 7005.7 and 4625.5 Å, respectively. The weak ³P₀–¹D₂ transition (6133.1 Å) cannot be found in the spectrum, but ³P₁–¹D₂ is clearly seen and the wavelength is given in Table 3 where good agreement is found with the Bowen (1960) measurement. The ³P₁–¹S₀ transition is found in the STIS spectrum at 2691.85 Å but the wavelength is 0.11 Å longer than the IUE measurement of Penston et al. (1983), which is outside the uncertainties of the two measurements. However, the STIS wavelength is consistent with the Bowen (1960) wavelength which was derived from longer wavelength lines. The ¹S₀ level also decays to ³P₁ with an expected wavelength of 2786.8 Å, but it is predicted to be about a factor of 100 weaker than λ2691.85 and it cannot be found in the STIS spectrum.

The ³P_{1, 2}–¹D₂ transitions are both found in the UVES spectra at 5603.8 and 6230.1 Å, respectively, and their wavelengths are consistent with the values of Bowen (1960) although the new wavelengths have a significantly improved accuracy. The ¹D₂–¹S₀ line with expected wavelength 4101.6 Å lies between the two wavelength ranges of UVES and is not observed.

The UV transition, ³P₁–¹S₀, has not previously been identified and we believe it is the line at 2368.26 Å in the STIS spectrum. The K vi energy levels determined by Smitt et al. (1976) from allowed transitions in the extreme ultraviolet yield a predicted wavelength of 2368.24 Å with an accuracy of around 0.11 Å, in excellent agreement with the STIS measurement. Note, however, that the STIS line is very weak and partly blended with an Fe ii line at 2367.41 Å thus there is some uncertainty over this identification.

The ³P_{1, 2}–¹D₂ transitions are well-observed in the UVES spectrum at 4940.6 and 5620.1 Å, respectively, and the wavelengths are in agreement with the values reported for RR Tel by Thackeray (1974).

The ³P₁–¹S₀ transition is found at 2111.49 Å and is a strong line in the STIS spectrum. It was not measured by Penston et al. (1983) probably due to the low instrument sensitivity at this wavelength. Smitt et al. (1976) derived energy levels for Ca vii using allowed transitions at EUV wavelengths, and the two forbidden line measurements of Thackeray (1974) combined with these energies yield a predicted wavelength of 2111.64 Å, close to the measured STIS wavelength. λ2111.5 has previously been measured from a STIS spectrum of another symbiotic star, AG Draconis, by Young et al. (2006). We note that if the new reference wavelength from RR Tel is used then the velocity of the AG Dra line becomes −144 km s⁻¹ in very good agreement with other emission lines from that system.

The ¹D₂–¹S₀ transition is blended with a much stronger transition due to the n = 19 member of the H i Balmer series (λ3687.91) and an accurate measurement of the Ca vii line cannot be made. However, combining the measured wavelengths of the ³P₁–¹S₀ and ³P₁–¹D₂ transitions yields a predicted wavelength of 3687.37 Å (vacuum), placing it on the short wavelength side of the blending H i line (3687.91 Å).

Only two of the Ar iv forbidden transitions are found in the RR Tel spectra: the ⁴S_3/2–²P_{1/2, 3/2} transitions at 2869.1 and 2854.6 Å, respectively. Both lines were reported by Penston et al. (1983) from IUE spectra and the STIS measurements agree with these within the uncertainties. The Bowen (1960) wavelengths reported in Table 3 were derived indirectly from transitions measured at visible wavelengths.

The NIST energies for K v are from Smitt et al. (1976) who used the forbidden wavelengths of Bowen (1960) together with EUV measurements of allowed transitions to determine the ground configuration energies. Bowen (1960) only measured the ⁴S_3/2–²D_{3/2, 5/2} transitions which, at 4163.30 and 4122.63 Å (air), are not available in the UVES spectra. Thackeray (1977) measured both of these transitions in optical spectra of RR Tel and found wavelengths of 4163.55 and 4122.75 Å (air).

The energies provided by Smitt et al. (1976) put the ⁴S_3/2–²P_{1/2, 3/2} transitions at 2515.21 and 2495.00 Å, respectively. The latter is an excellent wavelength match for a line in the STIS spectrum at 2494.97 Å, and the width of the line is also consistent with lines of similar IP (e.g., O iv). The λ2515.2 line is expected to be 2–4 times weaker based on the CHIANTI atomic model, and a good match is the observed line at 2515.35 Å, which also has a width consistent with λ2494.97. The line is partly blended with a weak Fe ii transition and the components have been resolved by fitting two Gaussians.

The ²D_{3/2, 5/2}–²P_{1/2, 3/2} transitions occur between 6223 and 6351 Å, with the strongest expected to be 3/2–3/2 at 6223.7 Å. A line is found at this wavelength in the UVES spectrum although it is blended with another line in the short wavelength wing. A Gaussian was fit to the K v line by ignoring the wing in the fitting process. The wavelength is given in Table 3 and is in excellent agreement with that predicted from the Smitt et al. (1976) energy levels.

The four ²D–²P transitions are found between 5462 and 5767 Å, and two of the lines have previously been measured in RR Tel by Thackeray (1974; see also Thackeray 1977): the 3/2–3/2 and 5/2–3/2 transitions at 5460.7 and 5586.3 Å (air wavelengths). Both lines are also found in the UVES spectrum and the wavelengths are in good agreement. Note that the wavelengths of Bowen (1960) quoted in Table 3 were derived through calculation or interpolation along the isoelectronic sequence and so are of limited accuracy. The 5/2–1/2 transition (5767.0 Å) is expected to be very weak and is not observed, while the 3/2–1/2 transition (5633.3 Å) is blended with Fe vi λ5632.6.

The UVES spectra contain the ⁴S_3/2–²D_{3/2, 5/2} transitions at 3670.2 and 3726.5 Å. The latter is the stronger line and has a weaker O ii line in the long wavelength wing. The Ca vi centroid was estimated by performing a single Gaussian fit, but ignoring the line wing. We note that Thackeray (1974) suggested there was a strong O iv contribution at this wavelength, but this is not the case in the UVES spectrum. λ3670.2 is blended with the n = 25 member of the H i Balmer series which provides the dominant contribution. An accurate measurement of the line's centroid could not be made and so the transition is not listed in Table 3.

The ⁴S_3/2–²P_{3/2, 1/2} transitions are found at 2215.2 and 2242.8 Å, respectively, in the STIS spectra. Both are strong, well-observed lines, however λ2215.2 is blended with the n = 11 member of the He ii Fowler series. Comparisons with other members of the Fowler series suggest that He ii contributes around one third to the blend, while the wavelength of the blended feature is shorter than expected for the He ii line, suggesting Ca vi is on the short wavelength side of the profile and He ii on the long wavelength side. Despite this a two Gaussian fit to the feature fails to yield an He ii fit consistent with the expected flux and wavelength. To estimate the Ca vi wavelength, we fit a single Gaussian to the blended feature and apply a −5 km s⁻¹ shift to the resulting centroid to indicate that Ca vi is on the short wavelength side of the fitted line. An uncertainty of 5 km s⁻¹ has been added to the other sources of error for this centroid to yield the wavelength given in Table 3. Penston et al. (1983) measured both Ca vi lines in IUE spectra of RR Tel, but did not note the blend of λ2215.2 with He ii. This perhaps explains the 4 km s⁻¹ discrepancy for this line; the wavelengths for λ2242.8 are in good agreement.

A consistency check on the Ca vi levels can be performed using the λλ2215.2, 3726.5, and 5462.2 lines. We find a predicted wavelength for the ⁴S_3/2–²P_1/2 transition of 2215.158 ± 0.032 Å which is in good agreement with the measured wavelength. We can also use the measured wavelengths for the λλ2215.2 and 5587.8 lines to predict a wavelength for the ⁴S_3/2–²D_3/2 transition which we find to be 3670.15 ± 0.16 Å, placing it on the short wavelength side of the stronger H i line (3670.51 Å).

The energies for the Ar iii ground configuration levels in the NIST database are given to three decimal places, suggesting a very high accuracy. The levels were derived from unpublished work so the measurement sources are unknown, although the wavelengths derived from the level energies suggest that they are based on the wavelengths of Bowen (1960) who gave quite precise wavelengths for the four strong forbidden transitions. The ³P₁–¹S₀ and ¹D₂–¹S₀ transitions have been reported by Thackeray (1977) from optical spectra of RR Tel, and ³P₁–¹S₀ has been reported by Penston et al. (1983) from ultraviolet spectra. The ³P_{1, 2}–¹D₂ transitions at 7751.1 and 7135.8 Å lie outside of the UVES wavelength range and so cannot be measured here.

The ³P₁–¹S₀ transition at 3110.1 Å is the longest wavelength line seen in the STIS UV spectrum and is also one of only two lines that are also observed by UVES (the other is Ni vii λ3106.2). The STIS line profile is symmetric, while the UVES profile has an enhanced long wavelength wing. Fitting the UVES profile with a single Gaussian yields a wavelength within 0.005 Å of the STIS line. The wavelength given in Table 3 is from the STIS spectrum and agrees well with the Bowen (1960) and NIST values but there is a significant discrepancy with the Thackeray (1977) and Penston et al. (1983) wavelengths.

The ¹D₂–¹S₀ transition at 5193.1 Å has a very extended long wavelength wing that may partly include another emission line. This wing has been masked out when performing the fit so that only the central portion of the profile and the short wavelength wing are included. The derived wavelength agrees within the uncertainties with the measurements of Bowen (1960) and Thackeray (1977).

The ¹D₂–¹S₀ transition (4512.2 Å) lies outside of the UVES wavelength ranges and so cannot be measured. The strongest transition, ³P₂–¹D₂, however is clearly identified at 6103.5 Å and the wavelength is in agreement with the measurements of Bowen (1960) and Thackeray (1977). The latter author also identified the ³P₁–¹D₂ transition at 6792.5 Å (air) in the RR Tel optical spectrum and the same line can be seen in the present UVES spectra. However, Feuchtgruber et al. (1997) provided an accurate measurement of the ³P₂–³P₁ splitting which, using the measured λ6103.5 wavelength, implies a wavelength for ³P₁–¹D₂ of 6797.04 Å. A group of four weak lines can be found near this wavelength, one of which has a wavelength of 6797.05 Å. Therefore, we tentatively make this identification.

The ³P₁–¹S₀ transition is expected in the ultraviolet at 2711.9 Å, but cannot be found in the present STIS spectrum.

Of the four potentially observable Ca v lines, one line, λ3999.0, is outside of the UVES wavelength ranges, while another, λ6088.1, is blended with an Fe vii line. Young et al. (2005a) estimated that the Ca v line contributes <2% to this blend and so it cannot be used for determining the rest wavelength of the line.

The ³P₂–¹D₂ transition at 5310.6 Å was previously measured by Bowen (1960) and Thackeray (1977) and the present measurement is in good agreement with the latter, but marginally discrepant with the former. In the STIS spectrum the ³P₁–¹S₀ transition is found at 2413.6 Å and the measured wavelength is in good agreement with the previous value of Penston et al. (1983).