Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 348, 768-782 (1999)

Previous Section Next Section Title Page Table of Contents

Appendix: NIR Observing and data processing techniques

We are in the process of mapping the inner part of the Nuclear Bulge in the K, H and J bands with the IRAC2B camera in the ESO/MPG 2.2m-telescope. IRAC2B uses a NICMOS3 256[FORMULA]256 pixel detector array. We use the pixel scale of [FORMULA]; the field-of-view covered by a single exposure is thus [FORMULA].

Appendix A: observing strategy

All data presented in this paper were obtained in June 1996. The "sky" reference (OFF position) was chosen in the direction of a local Dark Cloud which is offset from Sgr A* by [FORMULA], [FORMULA] corresponding to a projected distance of [FORMULA] pc from Sgr A* (see Table 2). I.S. Glass (1987) estimates for this cloud a visual extinction of [FORMULA] [FORMULA] 60 mag which means that H and K band emission from all stars located behind this Dark Cloud will be blocked. Stars located between Dark Cloud and Sun would appear as negative sources in the mosaic and therefore were readded to the mosaic (see Sect. 4.1).

We observed in sequences of 12 exposures, with each sequence taking about 10-12 minutes. An observing sequence for the production of four different individual mosaic images consists of: central image (15 s integration time) - Dark Cloud (30 s) - [FORMULA] mosaic images (60 s) - Dark Cloud (30 s) - [FORMULA] mosaic images (60 s) - Dark Cloud (30 s). Although the usage of a fixed "sky" position is quite time-consuming, because the telescope slewing-time is longer than the exposure-time, such a procedure is necessary to relate the intensity scale within the mosaic to the same relative zero-level. The central image is used as calibrator to eliminate atmospheric absorption.

No straight-forward measurements of the continuum due to unresolved stars are possible. Hence, all integrated surface brightnesses should be considered as (in most cases rather reliable) lower limits (see Sect. 4.3 where we estimate an upper limit of [FORMULA] Jy for the integrated continuum in the direction of the Dark Cloud).

Appendix B: data processing

The mosaic presented here was analyzed with the reduction program MOPSI 9 developed by R. Zylka. The regridding algorithm uses a general 3-dimensional rotation and is based on a method developed by C.G.T. Haslam for the NOD2 software package.

B.1. Coordinate calculation and image alignment

Datasets obtained with radio- and (sub)mm-telescopes usually do not need any coordinate-corrections because the pointing-accuracy of these telescopes is clearly better than the angular resolution (at the IRAM 30-m telescope the pointing accuracy is 1-2" compared to an angular resolution of[FORMULA]). In the case of infrared and optical telescopes the effective resolution (limited by seeing ) can be better than the pointing-accuracy (e.g. 1" compared to 5" at the ESO/MPG 2.2m telescope). Hence, accurate positioning of the individual images is one of the main tasks in the construction of a mosaic image. We used two different methods to adjust the coordinates of overlapping images. In the first approach we calculate the correlation coefficient between the two images. In the second approach we use the positions of individual stars as determined by fitting a Lorentzian distribution to the intensity distribution (see below) and solved the overdetermined set of linear equations for the angular shift between two overlapping images. These two methods give similar accuracy of [FORMULA] pixel size. We started from the image on the central stellar cluster. All coordinates are calculated as angular offsets relative to IRS 7. The positional errors at the edges of the current mosaic might thus reach [FORMULA]. The mosaic shown here in Fig. 1a consist of less than 500 images. The final H band mosaic contains [FORMULA]1500 images and thus is roughly 3 times larger. This part of the data reduction is the most computing-time expensive process.

B.2. Calibration

The mosaicing was performed under varying atmospheric conditions. We did not use point sources within the central cluster to avoid seeing effects. Rather, we used the integral over all stars in the overlapping region of all central images (about [FORMULA]) were used as a substandard. The time variation of the flux in this overlapping region is due to different atmospheric extinction and in elevation which were corrected for. During very stable atmospheric conditions we derive from the variation of the flux density with the elevation an atmospheric K band zenith extinction [FORMULA].

This calibration procedure using the central stellar cluster as subcalibrator allowed us to achieve a relative calibration of the surface brightness within the mosaic of [FORMULA] 15% - almost independent of weather conditions. In addition to the variable extinction of the atmosphere, its variable emission also causes problems. Because the slewing of the telescope between the source and the sky-positions takes more than 1 minute we use the time-weighted average of two sky-exposures to subtract the sky emission. This minimizes the effects of variable atmospheric emission. The remaining much smaller intensity steps are automatically corrected, during construction of the mosaic from the center to the outer regions.

Appendix C: source decomposition

To separate sources from the unresolved background we fitted Lorentzian distributions to the sources after having subtracted a background from the mosaic, which was determined in areas away from bright stars. The Lorentzian distributions which we fitted are generalizations of the ones given by Diego (1985). The Intensity I at a point (x,y) is given by:








while [FORMULA] and [FORMULA] are defined through


b is the pedestal and h the maximum height of the distribution, [FORMULA] and [FORMULA] refer to major and minor axis of the ellipse and [FORMULA] is linked to the position angle [FORMULA] of the ellipse. The values of [FORMULA] and [FORMULA] determine the width of the peak area, which are modified by the exponent power p. For a circular source [FORMULA] [FORMULA] and [FORMULA]. The result of fits using Lorentzian and Gaussian distributions, respectively, is shown in Fig. C1. Obviously, the Lorentzian distribution fit the seeing-broadened point-spread-function of the stars much better.

[FIGURE] Fig. C1a-c. Comparison between the fitting of Gaussian and Lorentzian distributions. a  A K band image showing two stronger and a few weaker stars well above the detection limit of [FORMULA]Jy. b  The residual image after subtraction of Lorentzian fits to the stars. c  The residual image after subtraction of Gaussian fits to the stars.

Appendix D: spectral types and K band flux densities

To estimate the K band flux densities of stars of different spectral types and luminosity classes we use the stellar parameters given in Lang (1992) with Eq(5.4) in MDZ96 valid for [FORMULA] and calculate the flux density [FORMULA] of a single star of effective temperature [FORMULA] and [FORMULA] with [FORMULA] the stellar radius. This flux density is given for [FORMULA]m:


with [FORMULA] K/[FORMULA]. The flux densities shown in Fig. D1 have been reddened for a K band extinction of [FORMULA] = 0.122 (Mathis et al., 1983).

[FIGURE] Fig. D1. Reddened flux densities for stars of the luminosity classes MS, Giants and Supergiants with different spectral types as well as for different types of Wolf-Rayet stars. Indicated as light dashed line is the flux density corresponding to our approximate completeness limit and as heavy dashed line the flux density corresponding to our detection limit.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1999

Online publication: August 13, 199