2 Observations and data reduction

Spectroscopic observations of RR Tel were obtained using the 3.9m telescope at the Anglo-Australian Observatory on July 22 1996. The University College London Echelle Spectrograph was employed with the 31linesmm^-1 grating and 700mm camera, along with the Tek CCD as a detector. The result was a high resolution, high signal-to-noise (S/N) spectrum of RR Tel over the near-UV, optical and near-IR regions, 3100 - 9800Å. The resolution was $\lambda$/$\Delta\lambda$$\simeq$50000, a factor of 2 greater than for the previous data from the Cerro Tololo Inter-American Observatory (CTIO) (McKenna et al. [1997]), and the S/N was $\simeq$20pixel^-1 in the continuum, similar to the CTIO observations. This allowed a greater dynamic range than with earlier observations, and enabled us to measure reliable line intensities for much weaker features than was previously possible. A section of our high resolution data is shown in Fig.1.

  

Figure 1: The flux-calibrated high resolution optical spectrum of RR Tel in the wavelength range 4850 - 4980Å, where the flux is in units of 10-12ergscm-2s-1Å-1, featuring lines of the highly ionised species [FeVII] at 4893.90Å and 4942.30Å and [OIII] at 4931.00Å and 4958.91Å, as well as the strong H$\beta$ line at 4861.33Å

In order to absolutely flux-calibrate our data, it was necessary to obtain low resolution spectra of RR Tel. These were obtained on August 2 1996 with the Australian National University 2.3m telescope and double beam spectrograph, which gave a resolution of $\simeq$2Å pixel^-1. However, this instrumental setup only gave useful data for wavelengths in the range 3490$\leq$$\lambda$$\leq$5530Å. Line fluxes for wavelengths outside this range were derived from the HeII recombination line spectrum (see Sect. 4). The low resolution spectrum is shown in Fig.2, where, due to seeing of $\sim$2$^{\prime\prime}$, the accuracy of the flux calibration is estimated at $\pm$10%.

  

Figure 2: The flux-calibrated low resolution optical spectrum of RR Tel, where the flux is in units of 10-11 ergscm-2 s-1Å-1

  

Table 1: An illustration of the changes occurring in the intensities of lines as RR Tel progresses towards a higher degree of ionisation. All intensities are given relative to H$\beta$ = 100

All CCD images were reduced using the Image Reduction and Analysis Facility (IRAF) version 2.3 implemented at the Queen's University, Belfast STARLINK node. Standard procedures were followed in the reduction of the images to one dimensional spectra. Low resolution spectra were optimally extracted and flux-calibrated using standards from AAO lists. The densely-packed échelle spectra were extracted using cosmic ray rejection algorithms for sky regions only, as cosmic ray hits in the spectra themselves could be confused with real emission features. Multiple high resolution échelle images enabled us to median filter the extracted spectra to subsequently eliminate cosmic rays. Once reduced to one dimensional spectra, the data were input to the user-friendly spectrum analysis program DIPSO (Howarth et al. [1996]) for further analysis.