Imaging simulations of large-scale flux recovery at millimeter wavelengths

We present multiple-field imaging simulations that explore the issue of large-scale flux recovery using five current and future millimeter-wavelength interferometers. The simulations show that nonlinear deconvolutions routinely applied to interferometric maps interpolate and extrapolate to unsampled spatial frequencies and thereby reconstruct much larger scale structures than an analytical treatment of the flux recovery issue would suggest. We show that the fraction of flux recovered for a given observation is a function of the signal-to-noise ratio (S/N) of the map; however, even for S/Ns as low as 3, a deconvolved map reconstructs more flux than an un-deconvolved map. Both the noise-free and the noise-added simulations demonstrate that in order to make accurate maps of even moderately large (greater than or similar to20) sources at millimeter wavelengths, it is generally necessary to include single-dish (total-power) or very short spacing data. We demonstrate that for high-S/N data, how well an individual telescope recovers large-scale structure for a mosaiced observation is more closely related to the minimum distance between its dishes, S-min - D, than to the minimum center-to-center distance, S-min. For a source that is large in one dimension but small in another (e. g., an elliptical Gaussian), the simulations show that the flux recovery is more closely determined by the small dimension than by the long dimension. This may help with the real imaging of spiral galaxies, where the rotation tends to confine the emission in any given channel map to a relatively small region in one dimension. These simulations were motivated by our work with mosaics from the Berkeley-Illinois-Maryland Association Survey of Nearby Galaxies (BIMA SONG), but the results apply generally to millimeter-wavelength interferometric images.