There have been a number of investigations on the absorption-line spectrum of 0237-233. We summarize these investigations here.
A spectrum of 195 Å per mm obtained by Arp et al. (1967) at the prime focus of the 200-inch telescope showed the presence of many absorption lines. These authors listed 33 absorption lines in the interval 3526-5096 Å and attempted to identify them on the redshift hypothesis with z=2.20. Burbidge (1967) obtained a spectrum of 0237-233 on baked IIaO emulsion with the conventional prime-focus spectrograph on the Lick 120-inch telescope, with the camera-grating combination giving a dispersion of 370 Å per mm. She listed 20 lines in the interval 3444-4613 Å and attempted to identify some of the lines - at z about 1.95. Greenstein (cf. Greenstein and Schmidt, 1967) obtained two spectra of 0237-233 at a higher resolution. They were obtained at 89 Å per mm in the first order with the Palomar prime-focus spectrograph. About 150 possible absorption features were measured. Of these, as described by Greenstein and Schmidt (1967), a highly selected list of 40 lines was published in their Table I. It is not clear how many strong lines were left unmentioned. Greenstein and Schmidt (1967) assigned most of these 40 lines to two redshift systems, z1=2.2020 and z2=1.9555. Thus was born the multiple-red shift hypothesis for the absorption lines of quasars.
Meanwhile several more spectrograms were obtained by Greenstein and Sargent (BGS) and a list of 49 absorption lines in the range 3346-4493 Å was analyzed in collaboration with Bahcall using the multiple-redshift hypothesis. Twenty-eight of the forty-nine lines were claimed to have been identified by five redshifts. These were z=2.2015, 1.6704, 1.6560, 1.5132, and 1.3642. However, many of the strong lines were unidentified, even with five redshift systems. It is not without significance that after this work on 0237-233, Greenstein, a very experienced stellar spectroscopist, quit quasars altogether and years later wrote:
Burbidge et al. (1968; hereafter BLS) obtained spectra of 0237-233 at Lick and Kitt Peak observatories and they list a total of 92 absorption lines in the interval 3205-6630 Å. Comparing their results with those of Greenstein and Schmidt (1967), Burbidge et al. commented :
The investigations of BLS and BGS were based on conventional image-tube and photographic spectrograms which were near the limits of the resolution, about 2.3 Å, then available on large telescopes.
Boksenberg and Sargent (1975) used the University College London image photon counting system (IPCS) at the Coude focus of the Palomar telescope to obtain the spectrum of 0237-233 at a resolution of 0.71 Å. They give a list of 75 absorption lines in the wavelength range 3737-4270 Å. Boroson et al. (1978, hereafter BSBC) carried out further observations on 0233-233 with the IPCS and these observations were combined with the older data of Boksenberg and Sargent (1975). These authors tabulated wavelengths and equivalent widths of 193 absorption lines in the special range 3716-4289 Å. BSBC claimed that many ofthese lines are due to C IV 1548.20, 1550.77 Å at various redshifts. A total of 45 redshifts were claimed. Varshni (1981) examined these supposed identifications and showed that the number of redshift systems based on C IV doublets differs insignificantly from what would be expected from chance coincidences. Consequently, these systems and their z-values seem devoid of any physical significance.
Varshalovich and Levshakov (1981) analyzed the absorption line data reported by BSBC and claimed to have found 12 redshift systems containing CO lines and they tabulate five of these. Varshni (1983) examined this claim also and showed that the number of redshifted CO absorption-line systems that are found in the spectrum of 0237-233, using the rules of Varshalovich and Levshakov, differs insignificantly from what would be expected from chance coincidences. Consequently, the CO systems proposed by Varshalovich and Levshakov also seem to have no physical reality.
It is obvious from the above summary that the redshift hypothesis has been singularly unsuccessful in explaining the absorption-line spectrum of 0237-233. Before coming to the identifications on the basis of our theory, we shall first consider the quality of the available data. Certain problems which were encountered in the case of 0805+046 (Varshni, 1985) are also present here.
With the improvement in observational techniques, the number of observed lines has been rapidly increasing. This can be seen in the number density oflines observed. Most of the reported observations of 0237-233 are shortward of 5100 Å, and in this region the number of lines per 100 Å for the various investigators are as follows : Arp et al. (1967) - 2.10, BLS (1968) - 4.51 (Kitt Peak spectrum) and 4.45 (Lick spectrum), BGS (1968) - 4.27, Boksenberg and Sargent (1978) - 13.70, BSBC (1978) - 33.68. The resolution obtained by BSBC was only 0.7 Å (corresponds to 52 km/s). Now it is possible to obtain much better resolution in quasar spectra (Chaffee et al., 1983, 12 km/s. York et al., 1984, 10 km/s ). If 0237-233 is studied with a resolution of 10 km/s, we expect the number density of lines to increase by a factor of about 2.5. The incompleteness of the BSBC data has to be borne in mind while considering the number oflines of any element which are identified as against the expected number of lines, and also as regards the completeness of multiplets in the identifications.
As noted earlier, the estimated resolution in the BSBC data is 0.71 Å. However, a close inspection of enlargements of Figures 1 and 2 of BSBC shows that many lines are wider than 0.7 Å and many of them are blends. BSBC state "The wavelengths of the lines are probably accurate to 0.3 Å except in the case of blended lines, where the error may be twice as large". Allowing for these various factors a value of 1 Å appears to be a reasonable guiding value for the tolerance : i.e., the discrepancy between the observed and identified wavelengths. When the line is very wide (blended) clearly a larger value is acceptable.
As regards the relative intensities of lines, again the situation is similar to that of 0805+046. We would give here only two examples. For the lines 3886.70 and 3891.8 Å, Boksenberg and Sargent (1975) give the intensities as 3 and 5, respectively, while BSBC give 1.9 and 0.9, respectively. Similarly for 3905.5 and 3909.8 Å, Boksenberg and Sargent's (1975) values are 3 and 2, while BSBC give 0.9 and 0.1, respectively. Clearly the recorded equivalent widths can only be taken as a qualitative guide and no more.
Identifications for most of the absorption lines in the range 3716-4290 Å reported by BSBC are presented in Table II. The reported equivalent widths of lines by BSBC appear to indicate that in some cases, besides the proposed identifications, there is blending due to unknown components. The average of the absolute difference between observed lines and their identifications is 0.49 Å which is quite satisfactory. Next we consider the evidence bearing on the presence of some of the atoms and ions.
Clearly the quasar is highly deficient in hydrogen. It will be noticed that most of the lines which we have not been able to identify are common with unidentified lines in the star upsilon Sgr.
For lines shortward of 3716 Å or longward of 4290 Å, the available data are very poor, positions being accurate to only about 3 Å. Consequently, we have not attempted to identify these highly blended lines, but instead we predict the expected spectrum in the visible region outside of 3716-4290 Å on the basis of our identifications of Table II. He I. Many lines should be present.
TiII, CrII, FeI, and FeII. Many multiplets which arise from low-lying terms in these ions should be present, especially those shown on the Grotrian diagrams of Moore and Merrill (1968). This prediction is based on the assumption that there is no significant change in the spectrum in the region 3716-4290 Å. It is known that shell stars change with time. Should a significant change have occurred in the spectrum in the region 3716-4290 Å when future observations are carried out, obviously a corresponding modification will have to be made in the prediction.
It would be most desirable to obtain the spectrum of 0237-233 over a wide wavelength interval (3200-5000 Å) at a high resolution c/R of about 10 km/s) with instrumentation which can detect weaker lines.