The overall quality of IUE high resolution spectra processed with NEWSIPS has been evaluated by studying the accuracy and the stability of wavelength determinations, the accuracy of equivalent width measurements, the flux repeatability and the residual camera non-linearities.
The analysis is based on a large number of spectra, mainly of the IUE standard stars. The spectra have been corrected for the echelle blaze function and calibrated in terms of absolute fluxes according to the procedure described in Paper II. Hereinafter, wavelengths are assumed to be in the heliocentric reference frame and in vacuum.
To assess the wavelength accuracy three aspects have been considered separately:
a) the accuracy and repeatability of wavelength measurements of a given spectral feature in several spectra of the same star;
b) the stability of the wavelength scale along the full spectral range;
c) the consistency of radial velocity determinations obtained from the SWP, LWP and LWR cameras.
One of the most important limitations to the wavelength accuracy of IUE high resolution spectra are the target acquisition errors at the nominal center of the spectrographs entrance apertures. These, if large enough, can also affect the quality of the ripple correction (see Paper II).
From the NEWSIPS dispersion constants and the central
wavelengths of the spectral orders it can be readily deduced
that the velocity dispersion corresponding to one pixel on the
image is practically constant all through the range covered by
the cameras, and namely:
0.05 km s-1 for SWP
0.09 km s-1 for LWP
0.03 km s-1 for LWR.
Taking into account that the plate scales are 1.530, 1.564 and 1.553 arcsec/pix for SWP, LWP and LWR, respectively (Garhart et al. 1997), an acquisition error of 1 arcsec along the high resolution dispersion direction would lead to a constant velocity offset of 5.1 km s-1 for SWP, 4.6 km s-1 for LWP and 4.7 km s-1 for LWR. Since the pointing/tracking accuracy is usually better than 1 arcsec, we can consider 5 km s-1 as a reasonable upper limit to the expected wavelength accuracy. Wavelength errors substantially larger might arise internally in the data extraction procedures.
To obtain a reliable information on the self-consistency of wavelength determinations, we have attempted to reduce the effects of spectral noise by performing a large number of measurements of selected narrow and symmetric absorption lines from the interstellar medium, which are strong in some of the IUE calibration stars, as well as of many emission lines in RR Tel. Out of the IUE standards, we have selected BD+28 4211, HD 60753, HD 93521, BD+75 325, Lep (HD 34816), Cas (HD 3360) and UMa (HD 120315). In addition, we have also used spectra of the star Oph (HD 149757). The present measurements refer only to large aperture spectra.
The interstellar lines selected for the SWP range were: SII 1259.520 Å, SiII 1260.412 Å, OI 1302.168 Å, SiII 1304.372 Å and CII 1334.532 Å, 1335.703 Å. For the long wavelength range we have used the K and H components of the MgII doublet at 2796.325 Å and 2803.530 Å and MgI 2852.965 Å. In the case of Oph we have also measured MnII 2576.877 Å, 2594.507 Å and several FeII lines. The cases in which the distortion of the profile due to close-by reseau marks (in particular for OI 1302 and CII 1334) precluded the accurate determination of the line position have not been taken into account in the final statistics. Laboratory wavelengths have been taken form Morton (1991).
The mean values of the radial velocities, the corresponding rms deviation and the number of independent measurements are reported in Table 1 for each target and camera. According to this table, the rms repeatability error on radial velocities, averaged over the three cameras is 4.6 1.5 km s-1. This value is smaller than the upper limit of about 5 km s-1 expected from acquisition errors. Considering the presence of spectral noise, we can safely conclude that the repeatability of wavelength (or velocity) determinations is satisfactory.
The results in Table 1 indicate that the radial velocities derived from the two long wavelength cameras are consistent while, on the contrary, the velocities derived from SWP spectra are systematically more negative. This, and other considerations about the consistency of radial velocity determinations from the three cameras will be discussed in Sect. 4.
We have studied the accuracy of the wavelength scale over a wide spectral range to look for possible time-dependent distortions across the camera faceplate. To this purpose, we have selected 6 SWP, 11 LWP and 3 LWR spectra of the emission line object RR Tel obtained at different epochs. For each spectrum we have measured the peak wavelengths of several emission lines chosen among those reasonably well exposed and with the cleanest profiles, covering the full spectral range. The highest excitation lines, such as those from [MgV], were purposely excluded because they provided systematically higher negative radial velocities probably due to stratification effects within the nebular region. The mean radial velocities of RR Tel are -69.5 6.5 km s-1, (SWP), -49.3 3.0 km s-1 (LWP), and -51.0 4.1 km s-1 (LWR). The total number of measurements are 106, 170 and 132 for SWP, LWP and LWR, respectively. Since the errors are of the same order as the repeatability errors quoted in the previous section, we conclude that the wavelength scales do no present appreciable distortions over the wavelength range covered and, within the observational errors, are stable over the period of time considered (1983-1994, 1985-1995 and 1978-1983 for SWP, LWP and LWR, respectively).
The present analysis has revealed the existence of an inconsistency in the radial velocities derived from the MgII K(2796.32 Å) and H (2803.53 Å) lines as measured in the LWP camera, where the two lines are present in both orders 82 and 83. To quantify this discrepancy, we have measured the velocities of the Mg II interstellar lines in 89 spectra of five IUE standard stars.
We find that, in the LWP camera, the radial velocity difference is -10.8 2.5 km s-1 when measured in order 83, and 11.7 1.5 km s-1 when measured in order 82. We find also that there is a discrepancy in the velocity of the K line measured in the two orders ( 0.9 km s-1), while the velocities of the H line measured in the two orders are consistent within 2 km s-1, on average. Since the velocity derived from other interstellar lines (e.g. MgI 2852.965 Å) is fully consistent with the measurements from the K line in order 83, we conclude that only this line provides correct radial velocity values. In INES concatenated spectra (Sect. 3.1) the K line comes from order 83 and the H line from order 82. Therefore, there is a systematic difference 1.3 km s-1 between the velocities determined from the two lines, being the correct value only that given by the K line.
A similar study performed in LWR spectra, where the K line appears only in order 83, shows that this problem is not present in this camera, where the two Mg II lines provide consistent velocity values: 1.2 km s-1.
|Figure 1: A portion of the longest available SWP exposure of RR Tel (SWP20246: 820 minutes). The two Lyman lines arise from geocoronal emission filling the large and the small apertures. The emission features labeled with an asterisk are due to overspilling of the strong NV emission into the adjacent order|
It has been repeatedly pointed out that the background extraction for high resolution spectra processed with IUESIPS was not accurate enough especially shortward of 1400 Å in the SWP camera and 2400 Å in the long wavelength cameras, as denoted by the negative fluxes assigned to the wings of the strongest emission lines and to the core of the saturated absorption lines. As shown below, this effect is not present anymore in spectra processed with NEWSIPS, which makes use of an upgraded background determination procedure (Smith 1999).
Overestimating or underestimating the background level leads to underestimating or overestimating the fluxes of the emission lines and the equivalent widths of the absorption lines. In the following, we report the tests done on the accuracy of the equivalent widths to verify the correctness of the background extraction. In addition, we have used the repeatability of the equivalent widths determinations as an indirect test of the stability of the background levels.
Figure 1 shows the profiles of the NV doublet emission and of the broad Lyman feature in the longest exposure available of RR Tel. It is clearly seen that the wings of these lines are not assigned negative values. Particularly interesting are the NV "ghost'' lines marked with an asterisk, which are still present in NEWSIPS data, but sensibly fainter than in the IUESIPS spectra, most likely due to the optimized extraction slit. The presence of such spurious lines has recently been reported by Zuccolo et al. (1997) and ascribed to overspilling of the strong NV doublet into adjacent orders.
To verify the accuracy of the background subtraction, we have compared the equivalent widths of the strongest interstellar lines in four spectra of Oph with those reported by Morton (1975) obtained from Copernicus data. These latter determinations are presumably not affected by background determination problems, unlike the IUE echelle spectra near the short wavelength end of the cameras. The results of the comparison are given in Fig. 2 and in Table 2, which provides the mean and the standard deviation of the four measurements. As it appears clearly from the figure, the NEWSIPS measurements are consistent, within the errors, with the values from Copernicus data, suggesting that the background evaluation for SWP spectra is essentially correct.
|Figure 2: Comparison of the equivalent widths of interstellar lines in the spectrum of Oph as measured in Copernicus and in NEWSIPS spectra. Copernicus measurements have been taken from Morton (1975). The dashed line in each panel indicates the 1:1 relation|
|SII 1250.59||81 12||100 3|
|SII 1253.81||92 11||106 2|
|SII 1259.52||93 18||112 1|
|SiII 1260.42||167 26||70|
|CI 1277.24||62 13||74|
|OI 1302.17||221 20||201 7|
|SiII 1304.37||115 13||132 1|
|CI 1328.83||49 7||52 6|
|CII 1334.53||193 17||189 2|
|CII 1335.70||147 21||140 4|
|SiII 1526.71||170 15||190:|
|FeII 1608.46||120 15||85:|
|SiII 1808.01||54 8||87 11|
|AlIII 1854.72||49 5||57:|
|AlIII 1862.79||28 2||34:|
: Means uncertain value.
The stability of the background subtraction has been evaluated by measuring the equivalent width of several strong interstellar lines in a large sample of spectra of two standard stars. The repeatability of the equivalent widths ranges from 10% for the strongest lines to 30% for the faintest ones (Table 4).
|ZnII 2026.16||111 26||117 32||100:|
|FeII 2382.76||251 6||246 11||238 20|
|MnII 2576.88||133 12||124 20||138 21|
|FeII 2586.65||225 10||200 8||214 19|
|MnII 2594.51||112 8||125 10||116 5|
|FeII 2600.17||251 5||253 6||226 16|
|MnII 2606.48||104 14||120 8||111 6|
: Means uncertain value.
|Line||BD+28 4211||BD+75 325|
|SII 1250.59||32 9 (29)||34 8 (26)|
|SII 1259.52||48 10 (28)||70 12 (35)|
|SII 1260.42||92 12 (37)||128 14 (25)|
|CII 1334.53||115 14 (36)||125 19 (35)|
|Line||BD+28 4211||BD+75 325|
|MgII 2796.53||257 20 (37)||369 50 (33)|
|MgI 2852.97||276 40 (11)||45 16 (10)|
|MgII 2796.35 Å||65 11 (41)||354 22 (36)|
|MgII 2803.53 Å||75 23 (41)||345 23 (36)|
Equivalent widths in mÅ. The number of measurements is given in brackets.
We tested the accuracy of the background extraction from the cores of strongly saturated absorption lines and the wings of strongly saturated emission lines. Figure 3 shows a portion of a spectrum of SN 1987A centered around the Mg II doublet at 2800 Å. It appears from the figure that the cores of the absorption lines do not become systematically negative, as expected for a correct background extraction. Shown in the same figure is another example, that of the strongly saturated Mg II emission doublet in the longest LWP exposure available of RR Tel (LWP25954): the lines wings do not reach negative fluxes, as required.
A second test has been performed in four spectra of Oph, in which we have measured the equivalent widths of some strong interstellar lines, and compared them with the Copernicus values given by Morton (1975). The results are summarized in Fig. 2 and Table 3, where are given the mean and the standard deviation of the four spectra.
Finally, we have measured the equivalent widths of the MgII and MgI lines in a large sample of spectra of BD+28 4211 and BD+75 325. The repeatability errors range from 35% for the faint MgI line in BD+75 325, to less than than 10% for the strong lines. Results are shown in Table 4.
|Figure 3: Top panel: The region of the MgII doublet in the spectrum of SN 1987A (LWP10194). Bottom panel: The same region in the deepest LWP exposure of RR Tel (LWP25954)|
|Band||Order||set A||set B|
Set A: BD+28 4211: a) 4 spectra (Jun. 86-Dec. 86); b) 6 spectra (Nov. 90-Jul. 91), BD+75 325: 8 spectra (Dec. 88-May 89), HD 60753: a) 4 spectra (Apr. 80-May 80); b) 8 spectra (Feb. 88-May 89), HD 93521: 11 spectra (March 1987).
Set B: BD+28 4211: 45 spectra (Dec. 82-Aug. 95), BD+75 425: 53 spectra (Sep. 85-Feb. 95), HD 60753: 38 spectra (Jul. 79-March 95).
|Band||Order||set A||set B|
BD+75 325: a) 11 spectra (March 85-Dec. 85); b) 9 spectra (Jan. 89-Dec. 89);
c) 8 spectra (Feb. 92-Nov. 92).
Set B: BD+28 4211: 34 spectra (Dec. 82-Aug. 95), BD+75 325: 37 spectra (Sep. 85-Feb. 95), HD 60753: 24 spectra (Jul. 86-Aug. 95).
|Band||Order||Set A||Set B|
Cas: a) 4 spectra (Dec. 81-Dec. 82); b) 5 spectra (Sep. 83-Jul. 84),
UMa: a) 6 spectra (March 81-Apr. 82); b) 5 spectra (Aug. 82-Jun. 86).
Set B: Cas: 30 spectra (Feb. 81-Feb. 87), UMa: 42 spectra (Sep. 78-Jul. 90).
As for the LWR camera, we have verified the accuracy of the background subtraction by measuring the equivalent widths of six strong FeII and MnII interstellar lines in four spectra of Oph, and compared them with measurements based on Copernicus data. As shown in Fig. 2, there is a good agreement between the two sets of equivalent widths, and no systematic departures are found.
The repeatability of equivalent width determinations has been determined measuring the equivalent widths of the MgII doublet in a large sample of spectra of UMa and Cas and Lep. The repeatability error ranges from 30% for the weak lines to less than 5% for the strongest ones (Table 4).
Two different tests have been performed to assess the flux repeatability of high resolution spectra. In the first one we selected restricted samples of spectra obtained close in time, and measured the ripple corrected net fluxes, i.e. without applying the time sensitivity degradation and the temperature dependence corrections. The second test has been performed on larger samples covering extended periods of time, measuring the absolutely calibrated fluxes, which include time and temperature corrections. In all case we have averaged the flux over a narrow wavelength interval free of lines. The flux repeatibility was defined as the percent rms deviation from the mean value. The results are summarized in Table 5.
For the first test we have used 41 high resolution spectra of IUE calibration standards grouped into sets of data with a similar exposure level and obtained close enough in time. The test was done in six bands 5 Å wide. In Table 5 (under "set A'') we report the percent rms deviation. The repeatability of spectra obtained sufficiently close in time is about 2%. The second test was made on a larger number of spectra with similar exposure time, without restricting the date of observation ("set B''). This test provides repeatability errors ranging from 3 to 4%. These somewhat larger errors are due to the intrinsic uncertainties of the sensitivity degradation correction algorithm.
The tests performed are similar to those described above for the SWP camera. The spectra in "Set A'' consist of three groups each containing images obtained in a restricted period of time. The flux repeatability was evaluated in four wavelength bands 5 Å wide. In this camera, the repeatability errors can reach the 5% level near the short wavelength end of the camera, but are a factor of two lower in the central bands. A similar test performed on a larger set of spectra needing correction for the time-dependent sensitivity degradation ("Set B'') provides errors slightly larger, confirming that the sensitivity degradation algorithm adopted for the LWP camera is essentially correct.
The flux repeatability was evaluated in four wavelength bands 5 Å wide. The test performed on spectra taken close in time, "Set A'', shows that the repeatability errors reach 4% near the short wavelength end of the camera, decreasing in the region of maximum sensitivity and increasing again at the longest wavelengths. In the "Set B'' spectra the repeatability is worse, reflecting the uncertainties in the time degradation correction and also the instability of the camera after it ceased to be routinely used.
|Image||t/t(opt)||1185 Å||1285 Å||1485 Å||1785 Å|
|Group||t/t(opt)||2120 Å||2460 Å||2925 Å||3132 Å|
a: 2 spectra of BD+28 4211,
b: 1 spectrum of BD+28 4211,
c: 2 spectra of BD+28 4211,
d: 2 spectra of BD+75 325,
e: reference group: 4 spectra of BD+75 325 and 4 spectra of BD+28 4211,
f: 1 spectrum of BD+75 325,
g: 1 spectrum of BD+75 325.
|Image||t/t(opt)||2120 Å||2460 Å||2925 Å||3080 Å|
Despite the linearity correction applied during the processing, residual non linearities are still present in IUE data. This effect has been evaluated in Paper I for low resolution data. In what follows we discuss this effect in high resolution spectra. The method followed consists on studying a set of spectra of the same star with different exposure times obtained, whenever possible, very close in time (preferably on the same observing shift) and with similar camera temperatures. The variation of the flux with the level of exposure (or the exposure time) defines the flux linearity. Unfortunately there exist few sets of high resolution data suitable for this study, and a slightly different approach has been taken here. The results are summarized in Table 6. In general, these results are in good agreement with those derived in Paper I.
The most complete set of SWP spectra appropriate to assess the linearity consists of seven images of the white dwarf CD-38 10980 obtained in the period April-August 1991, with exposure times ranging from 50% to 180% of the optimum value (200 min). For these images we have measured the mean flux in five 5 Å wide bands. The fluxes in each band have been averaged together and divided by the mean flux in the two 100% exposures. The ratios so obtained indicate departures from linearity ranging from -6% at 1185 Å to +4% at 1785 Å for the 50% exposure and up to -5% for the 70% exposure.
There is not any complete set of high resolution data of the same star which allows to study the LWP camera linearity. We have constructed average spectra of different exposure levels of the two standard stars BD+28 4211 and BD+75 325, and divided them by the corresponding 100% spectrum. The exposure levels covered range from 27% to 207% of the optimum exposure time. The test was performed in four wavelength bands 5 Å wide, selected for being relatively free from strong absorption lines. The maximum departures from linearity, reaching 8%, are found for the 133% level at 2120 Å, and for the 207% level near the regions of maximum sensitivity of the camera. The latter deviation can be easily understood in terms of saturation.
The LWR high resolution linearity test has been performed with five images of the standard star HD 93521 obtained in the period July 1980 to February 1981, covering the range of exposure times from 50% to 250% of the optimum value. The test was performed in four bands 5 Å wide. The maximum departures form linearity (up to 25%) are found at the shortest wavelengths, where fluxes are underestimated by 25% for the 50% exposure and overestimated by 12% for the 250% exposure.
Copyright The European Southern Observatory (ESO)