2 Observations and data reduction

2.1 IRAS data acquisition and image processing

IRAS co-added images at 12, 25, 60 and 100 $\mu$m were obtained from the Infrared Processing and Analysis Center (IPAC). The images were interpolated to a pixel size of $15^{\prime\prime}\times 15^{\prime\prime}$. Our image processing procedure, which is similar to the procedures developed by Langer et al. (1989), Snell et al. (1989), and Wood et al. (1994) has two steps. First, we remove the emission from zodiacal dust in the 60 and 100 $\mu$m IRAS images of each field. Next, we subtract a background from all images in order to ensure that empty sky has zero surface brightness.

The images in our study exhibit only marginal smooth large-scale brightness gradients produced by zodiacal dust. The 60 $\mu$m images are generally more affected by zodiacal dust than the 100 $\mu$m images. Before we calculate dust temperature and opacity images, the contribution due to zodiacal and other background emission unrelated to GF 17 and GF 20 should be removed. Fortunately, the temperature of zodiacal dust, ${\sim} 150$ K, is very different from the ${\sim} 30$ K dust found in dark clouds, and unlike most dark clouds, zodiacal dust emission is generally smooth and very large-scale (i.e., structures that span several degrees). However, since GF 17 and GF 20 lie more than $20^{\circ}$ away from the plane of the ecliptic, the zodiacal light contribution to our images is minimal and can safely be ignored. To test this assumption, we selected 3 points in each 100 $\mu$m image that appeared to have no dust emission. The mean value of the image in a $10^{\prime}\times 10^{\prime}$ box at each of the 3 positions was determined. We then calculated the image of a plane that passes through these points and subtracted this "flat-field'' from the 100 $\mu$m image. For the 60 $\mu$m images we calculated the plane at the same 3 positions used for the 100 $\mu$m image, and subtracted this "flat-field'' from the corresponding 60 $\mu$m image. We find that the calculated dust temperatures of the clouds vary by less than 2 K (or 5$\%$). Thus, the errors are negligible and zodical light can be ignored.

The zero level of the IRAS images is somewhat arbitrary. In order to obtain images in which apparently empty sky has a surface brightness of zero, we offset the background level in each field. Using the $100~\mu$m image as a reference (because it shows cold cloud structures better than the $60~\mu$m image), we calculated the minimum flux values in the 12, 25, 60 and $100~\mu$m images. The minimum flux value was then subtracted from each image to produce the final 12, 25, 60 and $100~\mu$m images. To insure that the images were treated the same, we used the same positions in all the images of a field when evaluating the background level.

2.2 Millimeter data

Observations of GF 17 and GF 20 were carried out in four different periods, during 1996 January 20-25 and July 10-13, 1997 January 25-28, and 1999 January 19-24 at La Silla, ESO, Chile. Observations of the CO (J=1-0), ¹³CO(J=1-0), and C¹⁸O(J=1-0) rotational transitions (hereafter, CO, ¹³CO, and C¹⁸O) were performed with the 15 m SEST telescope. The SEST and its instrumentation have been described in detail by Booth et al. (1989) and an update was given by Nyman & Booth (1990). A high-resolution 2000-channel acousto-optical spectrometer (AOS) was used as a back end, with a total bandwidth of 86 MHz and a resolution of 43 kHz per channel. SIS receivers were used in single side-band (SSB) mode. Typical SSB system temperatures were found to be in the range 160-400 K during the observations. The antenna half-power beamwidth is $44^{\prime\prime}$ for CO and $46^{\prime\prime}$ for ¹³CO and C¹⁸O. Typical signal-to-noise ratios were 60, 50, and 20 for CO, ¹³CO, and C¹⁸O, respectively. The main beam efficiency ( $\eta_{\rm mb}$) was 0.7 at the frequencies of the J=1-0 transition of CO, ¹³CO, and C¹⁸O.

We have obtained CO spectra toward 316 and 522 positions in GF 17 and GF 20, respectively. For ¹³CO, a total of 356 spectra were taken in GF 17, and 391 spectra were obtained for GF 20. C¹⁸O emission was searched in 139 positions in GF 17 and in 274 positions toward GF 20. The clouds were mapped with $46^{\prime\prime}$ spacing, corresponding approximately to full-beam spacing at the frequencies of the CO, ¹³CO, and C¹⁸O line transitions. In GF 17, two regions (delimited by the boxes in Fig. 1) were searched for CO emission, while GF 20 was virtually completely mapped. The spectra were taken in frequency-switching mode (with frequency throws in the range 10-16 MHz). The spectral line intensities were calibrated and corrected for atmospheric losses using the standard chopper wheel method (Kutner & Ülich 1981). The observed line intensities are expressed as antenna temperature $T_{\rm A}^{\ast}$. Division of $T_{\rm A}^{\ast}$ by $\eta_{\rm mb}$ yields the main beam radiation temperature, $T_{\rm mb}$. Pointing was checked every 2-3 hours by means of repeated spectral line observations of the SiO (v=1, J=2-1) maser sources W Hya, R Dor, L2 Pup, and Vx Sgr, and was found to be accurate to within 6 $^{\prime\prime}$. The data were processed using standard procedures of the Continuum and Line Analysis Single-dish Software (CLASS) package developed at Observatoire de Grenoble and IRAM Institute. The spectra were folded, and baselines of order ${\leq} 3$ were fitted and removed. Upon fitting a baseline, each observed spectral line was fitted by a gaussian profile using a non-linear least chi-square ($\chi^{2}$) minimization, yielding the antenna temperature, central velocity, and linewidth of the line.