Results

The plots show the Paranal atmospheric night-sky spectrum as seen by UVES. These are the deepest-ever high-resolution observations of the sky emission.

The vertical scale is physical flux in 1e-16 erg / (s*A*cm^2*arcs^2) and includes the continuum sky background. The data are extinction-free.

In the red spectra, the plotted spectrum is repeated on a compressed scale (factor 5 or 10 as indicated in the plots, red curve).

The blue bars denote identifications from this data set (CENTER), see below.

Emission line measurements

The fluxed sky spectra are dominated by faint emission lines in the blue, and a rich emission line spectrum with both strong and faint lines in the red. The line positions and strengths have been measured interactively. For that purpose, both the 2D spectra, the extracted SKY1 and SKY2 spectra, and their error (expressed as sqrt[difference**2]), have been compared.

Fig. 1: Extraction and line identification. The upper display shows the pipeline-delivered 2D spectrum with slit coordinate in vertical direction. Blue and red frames mark the lower and upper sky windows. Bright and faint sky emission lines are readily visible. The lower graph has been used to identify emission lines. The black and red lines mark the extracted SKY windows. There is also a SKY plot at reduced intensity (factor 10) to show the brightest line profiles. The lowermost graph has the combined signal (SKY1+SKY2) and the difference spectrum (abs[SKY1-SKY2]). The difference (error) plot has been used to discriminate the extremely bright feature at 6947 (being due to a residual cosmic).

The following criteria for a line to be accepted for the catalogue had to be satisfied:

1. Ideally the emission line should be visible as vertical bar in the 2D spectrum.This criterion is powerful for the distinction between real emission and noise.
2. The error file should not increase around a line candidate. This criterion effectively means that a candidate line should be visible in both extracted SKY spectra with about the same strength. This criterion protects against misidentifications of residual cosmics and also in noisy spectral regions. These are common at the beginning of each echelle order. This criterion is very powerful, in particular when the first criterion starts failing for faint features.

These two criteria in combination are very efficient for the detection of sky emission lines which essentially should be monochromatic images of the slit. E.g., those many peaks in the blue spectrum which do not have an identification bar have clearly failed at least one criterion.

While the two above criteria were used to identify a line candidate, technically the lines were then identified with the MIDAS command center/gauss. Two cursor inputs per identifications were needed to define the range within which MIDAS then fitted a Gaussian to the sky emission line. Output parameters were the central wavelength from the Gauss fit, the central intensity from the measured spectrum, and the FWHM of the fit. These three parameters (called CENTER, INT_PEAK, and FWHM) are listed in the result tables. Central intensity is in counts (see above for correction into counts per hour) and includes the local background. The approximate level of the background is indicated in the line intensity plot (see below).

The REDU (red upper) CCD suffered from a leaking column causing broad quasi-emission in the echelle order. These features have been easily recognized due to their known position and atypical width. They are marked by squared crosses in the plots.

Line blends which would be considered as resolved with a simple cursor-marked approach are still a blend for the Gaussian fit procedure unless two peaks are clearly separated. Hence some of the lines catalogued here are indeed marginally resolved blends with a FWHM in excess of the resolution limit.

The list of measured lines is almost complete, at the level of the resoltuion and SNR of the data. Those spectral regions corresponding to the beginning of a new echelle order may still hide some features in noise. There are also small spectral gaps between the redmost orders in the 860U settings.

Line tables. The line measurements are available as ASCII files, listing measured wavelength (in air, CENTER), FWHM, and peak flux.

These measurements are solely based on the above-mentioned criteria, i.e. they mark those emission lines which are considered as trustworthy identifications. No attempt has been made to associate parent atoms or molecules, nor transition designations.

Such are available from two external sources:

• the list of computed OH line transitions as provided by Rousselot et al., 2000 (A&A, 354, 1134) and extended by Lidman et al (2000). It has theoretical OH line positions and intensities longward of 6138 A, and a compilation to sky emission lines as observed with ISAAC@VLT and SOFI@NTT.
• the list of observed sky emission lines as provided by Osterbrock et al. (PASP 108, 277 [1996] and PASP 109, 614 [1997]). These data are based on HIRES@Keck spectra and cover the range redwards of 5200 A.

The computed OH line positions are marked in the plots by a blue cross. The wavelengths have been transformed from vacuum to air using the Edlen formula.

The observed Keck line positions are indicated as red filled circles.

Evaluation

The UVES sky emission spectrum starts at 3140 A and extends down to 10,400 A. There is a plethora of faint emission lines up to about 4400 A, up to 100 lines per 100A interval. According to Meinel (1961), they are mostly due to O2. There is a range between about 5000 and 5600 A with very little lines showing up. Above about 5800 A, the spectrum gets richer and richer again and has a second population maximum between 7000 and 8200 A, with typically 70 lines per 100 A interval. The red spectra are dominated by OH emission lines, with telluric O2 absorption showing up as well. A total of 2810 emission lines has been found.

The resolving power plot (ps version) has been obtained by dividing the wavelength of the line position by the FWHM data. The large scatter in the blue part is due to the intrinsic uncertainties of the Gaussian fit to faint lines. The red lines are better defined in shape. Their resolving power numbers cluster around 42,000, very close to the nominal value for a 1" slit (43,000). The histogram has line number counts per 100A interval and clearly shows the two distribution peaks in the blue and the red parts of the spectrum.

The line flux plot (ps version) shows all line fluxes on a logarithmic scale. The continuous line indicates the continuum emission. All blue sky emission lines are faint, while among the red emission lines many strong ones are found.

Look here for a comparison of the UVES spectra and the Keck data: UVES | Keck

Light pollution.  Osterbrock et al. (1996) list emission lines tracing light pollution, as HgI 5461, 5770, 5791 and components of the NaI 5890, 5896 lines indicative of both high-pressure and low-pressure sodium lamps. As expected for a dark site like Paranal, there is definitely no trace of 5461, 5770, or 5791 A pollution in the spectra. The same is true for the sodium resonance lines.

Downloads

Further links:

ISAAC OH atlas: under calibration

KECK HIRES SKY line atlas (PASP): Night-Sky High-Resolution Spectral Atlas of OH and O2 Emission Lines for Echelle Spectrograph Wavelength Calibration (PASP)

KECK HIRES SKY line atlas Catalog: VizieR Online Data Catalog: Keck/HIRES Sky Line Atlas (Osterbrock+ 1997)