Spectroscopy in the near infrared is in some way a tricky subject, the following text has been written as a guideline for the observations. If you are about to start an observing run with LonGSp we strongly recommend you to carefully read what follows.
The procedures for LonGSp observations are those commonly used in NIR spectroscopy, optimized for the characteristics of the instrument.
The J, H and K bands are characterized by a bright background, such a background is mainly due to the emission of atmospheric lines and of thermal radiation. In particular the atmospheric emission lines are the dominant component in the J and H bands while the thermal radiation mostly affect the wavelengths beyond 2.2 um. Since the background is usually much brighter than the scientific target it must be subtracted from the observed spectra to make faint spectroscopic features in the target observable.
Spectra with the source in different positions are then subtracted to remove sky lines and thermal background. If the source cover a large fraction of the slit, spectra off-source must be observed to adequately sample the sky background.
When you observe with LonGSp you need to nod the source along the slit using one of the mosaics available. We suggest two positions on the slit +/- 20 arcsec from the center.
The on chip integration time (time spent in each position of the slit) must not exceed 40 seconds in J and H bands and 60 seconds in K band. With these times there are no problems of saturation for the sky emission, but you still need to check for saturation of bright sources. If the source is bright, integration time will be reduced until to have less than 7000 counts/pixel, to stay within the linear range of the array. In the latter case, at a given position along the slit, several frames can be coadded but the total integration time must not exceed the above limits. The short integration time has as a drawback the fact that background limited conditions are usually not reached.
When you use a mosaic to nod ALWAYS CHECK that the tar get do not follow out of the slit. A straight subtraction is not always enought. Due to the variation in brightness the of the background multiplying by a factor might be necessary before subtracting. As drawback the dark pattern is not subtracted properly anymore. It is therefore necessary to subtract the dark before multiplying.
A sequence of darks must be acquired for each integration time used during the night. That include scientific exposures as well as flat frames. Since the detector presents memory effects the dark frames MUST NOT be observed after a bright source. The S/N of the darks must be high enought not to compromise the quality of the observations. Darks will also be used to define a dead-pixel-mask. Moreover to have a good subtraction of the sky lines it is necessary to use not too long integration times.
Flat-fielding is of course necessary. For each band observed a spectroscopic flat frames are obtained observing the dome illuminated by an halogen lamp. The illumination must be chosen to give 3000-8000 counts/pixel as average of the frame (only the subsection of the frame used for spectroscopy). The S/N of the flats must be high enought not to compromise the quality of the observations. Flats will also be used to define a hot-pixel-mask.
The atmosphere is not uniformly transparent in the NIR, atmospheric absorption bands will therefore be present in the spectrum of the scientific source. To remove these bands it is necessary to observe an astronomical source whose spectrum is well known, this observation will be used to build a reference spectrum. The spectrum of the source will be divided by this reference spectrum to correct for telluric features.
Reference spectra of stars must be observed quite often, every 30-90 minutes of integration (according to the quality of the night) and every time the instrumental configuration is changed (BEFORE CHANGING IT).
You could use an early type, featureless star (preferably an O star), to correct for telluric absorption and differential efficency of the system and a photometric standard star if you want to flux calibrate the final spectrum (only one grating position in each band is required). An alternative technique, proposed by Maiolino et al (1996; AJ, 111, 537), consist of using a G star corrected through data of solar spectrum. Both ways are succesfully tested.
Some more points must be kept in mind when observing: the wavelength calibration of LonGSp is good but not perfect and it is subject to change with time (it is a good idea then to check it before starting the observations using sky spectra) and the mechanical mounting of LonGSp is subject to flections; the position of the target in the slit during the mosaic must be checked every time the telescope is moved to a new position.
What has been told so far can be summarized in the following check list for the observations.