The H I interferometer observations were made with the Australia Telescope Compact Array (ATCA; Frater et al. 1992), a 6 km east-west synthesis array located near Narrabri, NSW, Australia. Over a 12 hr period on 1999 November 5, an area of approximately $2\hbox{$.\!\!^\circ$ }4 \times 2\hbox{$.\!\!^\circ$ }0$ was observed using the 210 m array configuration, which provides a well-distributed u-v coverage. The source was surveyed as a mosaic of 19 different pointings, following a hexagonal grid. The separation between grid-points was about 16 $^\prime$ , which satisfies the Nyquist sampling criterion for a primary beam of 33 $^\prime$ at $\lambda$ 21 cm. The observations in the HI line were centered at 1421 MHz, using 1024 channels over a bandwidth of 4 MHz. The synthesized beam of the data is $4\hbox{$.\mkern-4mu^\prime$ }7 \times 3\hbox{$.\mkern-4mu^\prime$ }0$ with PA = $14\hbox{$.\!\!^\circ$ }1$ and the channel separation 0.83 km s^-1 (velocity resolution 1 km s^-1). Flux density and bandpass calibration were carried out using observations of PKS B1934-638, with an assumed flux density of 14.9 Jy at 1.4 GHz. Antenna gains were calibrated using observations of the source PKS B1540-828.

The data were reduced using the MIRIAD package (Sault et al. 1995). After flagging and calibration, the contribution from the continuum was subtracted in the u-v plane (van Langevelde & Cotton 1990), and a cube was produced using natural weighting and discarding baselines longer than 1 k $\lambda$ . The cube was jointly deconvolved using the maximum entropy algorithm (Sault et al. 1996), resulting in a dataset with 800 velocity planes between -250 and +400 km s^-1 (LSR). The rms noise was determined from the flux density in the line-free channels of this final cube, resulting in 1 $\sigma$ level of 0.3 K per channel. The conversion between flux density in mJy beam^-1 and brightness temperature in K for the data is 0.013 K/(mJy/beam).

To recover structures at low spatial frequencies, the same area was observed with the single-dish, 30-m radiotelescope of the IAR, located in Villa Elisa, Argentina (Arnal et al. 2000). A 1008 channel correlator was used with a total effective bandwidth of 4 MHz, centered at 1420 MHz. The velocity resolution of the single-dish data is 1 km s^-1, and the rms noise per channel is 0.13 K in brightness temperature. The interferometer and single-dish data were normalized to a common temperature scale and identical velocity channel interval of 0.83 km s^-1. Both databases were then combined in the Fourier domain using the MIRIAD task IMMERGE. The rms noise of the combined databases was calculated in line-free channels to be 0.4 K per channel. A new image of SN 1006 was made in the radio continuum using the MOST telescope at 843 MHz. This observation, used to compare with the new H I observations, is part of a southern sky survey (Bock et al. 1999; Green 1999) and is a full 12 hr synthesis. The resolution of this image (included in Figs. 2 and 3) is 64§ $\times$ 43§ and the rms noise is $\sim$ 2 mJy beam^-1.

The CO observations were carried out on 1999 August 10 and 11 using the 15 m Swedish-ESO Submillimetre Telescope (SEST) in La Silla (Chile). We used SiS receivers to simultaneosly observe the ¹²CO J=1-0 (115 GHz) and ¹²CO J=2-1 (230 GHz) lines. The signals were fed into two Acousto-Optical spectrometers: a narrow band high-resolution spectrometer with 2000 channels, bandwidth 80 MHz, channel separation 41.7 kHz (0.054 km s^-1 at 230 GHz), and a wide band low-resolution instrument with 1440 channels, bandwidth 1000 MHz, 700 kHz channel separation (0.9 km s^-1 at 230 GHz, and 1.8 km s^-1 at 115 GHz). The angular resolution is 45§ and 23§ for the ¹²CO J=1-0 and J=2-1 transitions, respectively.

The pointing of the telescope was checked once during each observing run using the AH Sco and VX Sgr SiO maser sources at 86 GHz. The pointing errors were typically $\sim$ $3^{\prime\prime}$ . The system was calibrated at regular intervals to provide corrected antenna temperature. The beam efficiency is 0.7 and 0.5 for the J=1-0 and J=2-1 transitions, respectively.

A region of 5 $^{\prime} \times 6^{\prime}$ centered on the reference position 15 $^{\rm h}$ 04 $^{\rm m}$ 00 $^{\rm s}$ , -41 $^\circ$ 49 $^\prime$ 00§ (J2000) was surveyed in the two CO lines with low and high velocity resolutions, using 195 pointings spaced by 23§ in both spatial coordinates. The reference position was chosen to be near the TeV $\gamma$ -ray source (Tanimori et al. 1998). The data were taken in position-switched mode, with an off-source position selected to be relatively free of emission. The central velocity for these observations was -25 km s^-1.

In addition, 21 different profiles centered around the LSR velocity of -25 km s^-1, were obtained at positions distributed over the shell of SN 1006 and adjacent to it (Fig. 1).

$\begin{figure} \par\includegraphics[width=7cm,clip]{MS2223f1.ps} \end{figure}$

Figure 1: ROSAT HRI image of SN 1006 (from Winkler & Long 1997). The box and crosses indicate the positions of the ¹²CO J=1-0 and J=2-1 observations. The stars indicate the locations of possible detections.

There is a possible detection of a broad feature in the ¹²CO J=2-1 line between $V_{\rm LSR} \sim {{-}15}$ and $\sim$ -25 km s^-1 at a 2 $\sigma$ noise level in the two points identified by stars in Fig. 1 (near (J2000) 15 $^{\rm h}$ 04 $^{\rm m}$ 07 $^{\rm s}$ , -41 $^\circ$ 53$^$20§ and 15 $^{\rm h}$ 04 $^{\rm m}$ 00 $^{\rm s}$ , -41 $^\circ$ 44$^$30§). Future observations are required to confirm these possible detections.

2 Observations and data reduction

2.1 HI observations

2.2 CO observations