Spectroscopy

A star’s spectrum is a measure of its light intensity as a function of wavelength. It is also known as the spectral energy distribution (SED) of the star. There are different types of stellar spectra found, because stars have different properties (temperature, radius, luminosity, chemical composition, etc). Along the main sequence, the main reason for different spectral appearance is the mass of the star, which translates into its effective temperature. In the spectra of stars, we often find absorption lines which are produced when photons travels through a cooler gas before reaching the observer (usually the star’s atmosphere). Photons are therefore absorbed by the atoms in the gas through what is known as bound-bound transitions. These are discrete and narrow energy electron transitions, which are reflected through lines in the spectra of stars.

In order to record a spectrum, an adjustable slit is placed at the focus of the telescope. The light from the source passes through the slit, and is then transformed into a thin beam of parallel rays by the use of a collimating lens or mirror. The light is then focused onto a diffraction or reflective grating, which is then sent to a detector, usually a CCD. The final raw data is a 2D image with a spatial and a spectral axis (see Figure 3.5). The basic steps to reduce spectroscopic raw data are the following: bias removal and flat fielding of the raw frame, extraction of the spectrum and wavelength and flux calibration of the extracted spectrum.

3.3.1 Bias removal

A bias frame is taken with no exposure time (zero seconds) and quantifies the electronic response of the CCD. It corresponds to an offset level introduced when the detector is read out. Also, the conversion from analog-to-digital units during the readout stage produces a small amount of noise. This noise is found in the

bias frame so one should acquire several bias ‘exposures’ and the average of all the frames should yield a good ‘master bias frame’ which will be used to de-bias all the other images. This averaging process should avoid introducing additional noise to the science frames, as well as eliminate spurious pixels in individual frames. Also, one can use the bias to calculate the readout noise, by taking its root-mean-square.

3.3.2 Flat Fielding

Each pixel in a CCD has a different sensitivity. This is explained by two main factors. The optics of the telescope can lead to nonuniform light transmission across the entire field of view which can be explained by either the presence of dust on the CCD or by vignetting. Also, the QE of individual pixels varies across the CCD, leading to variations in sensitivity. For these reasons, one must correct these effects by acquiring flat-field images. In our case, we use a Tungsten lamp located inside the spectrograph. Other observers choose to obtain dome or twilight flats as well. The smooth Tungsten spectrum is collapsed to a 1D spectrum along the spatial direction and fitted with a high order polynomial. The 2D master flat frame is then divided by the polynomial and applied to the science frames.

3.3.3 Spectrum extraction

Due to tilts introduced by the optics, the spectrum of the observed target can also be curved in the raw frames. For this reason, one must ‘track’ the spectrum by fitting it with a low order polynomial. Selecting two small regions either side of the spectrum defines the sky or background level (see Figure 3.5), which is required during the extraction of the spectrum. The sky regions are also fitted with polynomials, which will be subtracted from the target’s spectrum during its extraction. In our study, we use the optimal extraction method outlined by Marsh (1989) and made available through thePAMELAsoftware. Optimal extraction is designed to achieve the best possible signal-to-noise ratio with CCD spectral data.

3.3.4 Wavelength calibration

Once a 1D optimally extracted spectrum is obtained, the next step is to wavelength calibrate it. Arc lamp exposures, typically copper-argon, copper-neon or thorium- argon, are taken throughout the observations. These lamps have specific emission lines, where the expected wavelength positions of these lines are precisely known. An arc line map must be constructed from the observations and reference data, as it relates pixel position to wavelength, and then applied to the science spectra.

3.3.5 Flux calibration

Spectrophotometric flux standard stars must be observed in order to carry out the flux calibration stage. These stars have been carefully and precisely studied and have tabulated values of their fluxes as a function of wavelength. The difference between the template and the spectrum of the standard star is fitted with a spline,

which is then applied to the science spectra. It is important to note that flux calibra- tion is not usually very accurate, especially in spectral regions with many spectral lines. Also, the Earth’s atmosphere becomes less transparent at redder wavelengths, leading to forests of absorption lines, known as telluric lines. These can be approx- imately removed by creating a telluric-correction spectrum and subtracting it from the science spectra. A rather more important effect is the differential atmospheric refraction, which corresponds to the deviation of light from a straight line as it passes through the atmosphere due to the variation in air density as a function of altitude. Differential atmospheric refraction effects are more pronounced at shorter (i.e bluer) wavelengths.

3.4

Summary

We have briefly described the tools and methods we have used in the data analysis and studies described in the following Chapters. Achieving high quality observa- tional data is becoming standard in astronomy, yet all the truth lies within the quality of their reduction. Each step of the bias removal, flat fielding, extraction and calibration must be done with precision and care in order to trust the final results. The astronomical surveys used in my studies are presented in the following Chapter.