Introduction

The interstellar medium is the matrix within which the evolution of galaxies is played out -- governed largely by the processes that influence the cycles of star formation. The conditions of the ISM, its spatial, dynamical, thermal and chemical structure, reflect the evolutionary processes at work within the Galaxy. Observations over the last few decades have provided us with tantalizing glimpses of the complexity of the ISM. Far from a homogeneous and tranquil environment, the ISM displays large density and temperature variations. Velocity fields within the medium are turbulent, often supersonic. At the small end of the range of spatial scales, this highly disturbed state is maintained by point-like energy input from stars during both the formation and death stages. On the other end of the range of scales, energy input may take the form of global, large-scale phenomena, such as viscous dissipation or magnetic stress from Galactic rotation, the motion of spiral arm density waves, or gravitational infall from the halo. Despite the apparent flux of energy on all scales, there exist pockets of relative quiescence where dense, cold gas can become self-gravitating and the process of star formation begins.

Our knowledge of this system is still very much in the formative stage. Fundamental questions that remain unanswered are numerous.

• What is the detailed evolutionary relationship between the phases of the ISM?
• What role does the disk-halo interaction play in the star formation history of the Galaxy?
• How can we characterise the topology and structure of the ISM?
• How is it related to the local environment and/or the global Galactic context?
• What are the main transport mechanisms for mass and energy among the components?
• Where and how are steady-state conditions achieved?
• Are we dealing with a closed system or are fluxes of mass and energy into and out of the Galaxy important?

Because of our vantage point within it, the Milky Way is the only Galaxy for which we have the potential to observe the relevant interactions and structures in sufficient detail and over the required range of spatial scales. An understanding of the evolution of external galaxies out to cosmologically significant distances must rely on a detailed knowledge of these processes gleaned from our own Galaxy.

Among the tracers of the interstellar medium, the 21-cm HI line uniquely traces the diffuse medium. The HI is widespread, with filling factor of 25% to 50% throughout the Galactic disk, and exhibits structure on all observed spatial scales. In contrast, the molecular gas is confined to much smaller clouds which fill a tiny fraction of the volume of the disk. High resolution survey projects at millimetre (Heyer et al. 1998) and infrared wavelengths (Cao et al. 1997; Kerton and Martin 2000) are providing large-scale images of the molecules and dust in the Galaxy at arcminute scales. Due to the long wavelength of the HI emission line, HI surveys of the Galaxy have, until recently, lacked sufficient angular resolution to be very useful for ISM studies. Advances in interferometric techniques and computing power now permit the application of wide-field synthesis imaging to Galactic HI studies. The impact of high angular resolution is demonstrated in Figure 1, which shows one HI velocity channel of a region in the northern galactic plane from the Leiden/Dwingeloo survey of Burton & Hartmann (1994) (35' resolution) compared to the same region from the CGPS (1' resolution). In this segment of the Perseus spiral arm, 1' corresponds to 0.6 pc. At this resolution the power of HI gas as a tracer of ISM processes is strikingly apparent. Features such as the chimney above the star cluster OCl 352, cold, dark, filamentary, parsec-scale HI clouds, and bright, arc-like shock structures become visible.

Figure 1. A 17 ° segment of the Perseus spiral arm at VLSR =-42 km s-1 as observed by the Leiden/Dwingeloo single dish survey (top) and the CGPS (bottom).

The VLA survey, in combination with the CGPS and SGPS, will provide a 3D image of this detail and quality for over 90% of the stellar disk of the Galaxy. The global image of the Galaxy will be a unique resource for the astronomical community. The observing parameters for the VLA Galactic Plane Survey are summarized in Table 1. The VGPS sky coverage is shown in Figure 2.

Survey area	990 pointings, longitude 18° to 67°, latitude coverage between -1° to +1°, and -2° to +2° (Figure 2)
Integration per field	9 minutes, broken into 3-minute snapshots at various hour angles
Resolution	1 arcminute
Sensitivity	11 mJy/beam (2 K for 1' beam)
Velocity resolution	6.1 kHz = 1.3 km s-1, sampled at 3.05 kHz
Band width	1.866 MHz (two times 1.56 MHz overlapped by 1.259 MHz)

Figure 2: Outline of the VGPS survey area and the 990 pointing centers shown on top 21-cm line emission from the Leiden/Dwingeloo survey.

Spectral Parameters

Our solution to this problem is to offset the centers of the two bands of 1.56 MHz each by ± 153 kHz from v_c, offsetting one by just half a channel (6.1/2 = 3.05 kHz) further than the other. Thus we use bandwidth codes 5555 and correlator mode 2AD, and we offset the first band by -149.53 kHz and the second by +153.53 kHz. The two bands then overlap by 1.259 MHz centered on v_c, and each extends 303 kHz beyond the other on either side, giving total bandwidth of 1.866 MHz. The total usable bandwidth is thus 354 km/s, assuming 5% loss on each edge (2 × 78 kHz). In the overlap region (239 km/s wide, after dropping the edges) we can easily cover almost all the hydrogen emission. The half channel offset will allow us to obtain velocity sampling of half the channel spacing (two samples per resolution element). This scheme is illustrated in Figure 3, which shows the two offset bands in red and blue, with the edge channels crossed off. At the bottom is a representative spectrum (taken from the SGPS at longitude 330° with the velocity scale inverted to match the velocity range at longitude 30°) which shows how the Galactic HI will fit in the spectrometer bands.

At wavelength 21-cm the continuum can be significantly linearly polarized, but the line is unpolarized to better than 10^-3, so it will be no problem to combine the two circular polarizations after the baseline has been subtracted. In order to use UVLIN or a similar technique to subtract the baselines from the uv data on the two polarizations separately, we want to switch the offsets (± 153 kHz) between the two polarizations on a time shorter than the uv averaging time. This way after averaging two consecutive records we will end up with uv samples which cover 1.866 MHz in both polarizations, with resolution of 6.125 kHz and channel spacing of 3.05 kHz in the overlap region. Since we are working in D array at narrow band, averaging times of 20 sec or even much more will not diminish our field of view significantly. We hope to develop and test a special observing mode which allows us to switch, on consecutive 10 sec records, either the fluke synthesizers (between the two polarizations) or the transfer switch, without stopping the scan. We have requested a few hours of C array time to test this observing mode next spring. If worst comes to worst, we will split the 3 minute total snapshot dwell time into two scans of 1.5 minutes each with the offsets switched between them, and use the normal 2AD observing mode, but this will cost another 11% overhead, since it imposes another 20s gap between the two scans.

Figure 3. Outline of the spectrometer setup for the VLA observations. From top to bottom, the IFs are labeled as LA, RA, LB, and RB. A single visit to a VGPS field results in four separate files labeled accordingly. The IFs were staggered by half a channel to allow sampling at 0.64 km/s channels over the velocity range of Galactic HI.