THE SHAPES OF CROSS-CORRELATION INTERFEROMETERS

Cross-correlation imaging interferometers designed in the shape of a c urve of constant width offer better sensitivity and imaging characteri stics than other designs because they sample the Fourier space of the image better than other shapes, for example, ''T's'' or ''Y's.'' In a doss-correlation interferometer each pair of antennas measures one Fou rier component with a spatial wavenumber proportional to the separatio n of the pair. Placing the individual antennas of the interferometer a long a curve of constant width, a curve that has the same diameter in all directions, guarantees that the spatial resolution of the instrume nt will be independent of direction because the measured Fourier compo nents will have the same maximum spatial wavenumber in all directions. The most uniform sampling within this circular region in Fourier spac e will be created by the particular symmetric curve of constant width that has the lowest degree of rotational symmetry or fewest number of sides, which is the Reuleaux triangle. The constant width curve with t he highest symmetry, the circle is the least satisfactory although sti ll considerably better than T's or Y's. In all cases, the sampling can be further improved by perturbing the antenna locations slightly off a perfect curve to break down symmetries in the antenna pattern which cause symmetries and hence nonuniformities in the sampling pattern in Fourier space. Appropriate patterns of perturbations can be determined numerically. As a numerical problem, optimizing the sampling in Fouri er space can be thought of as a generalization of the traveling salesm an problem to a continuous two-dimensional space. Self-organizing neur al networks which are effective in solving the traveling salesman prob lem are also effective in generating optimal interferometer shapes. Th e Smithsonian Astrophysical Observatory's Submillimeter Array, a cross -correlation imaging interferometer for astronomy, will be constructed with a design based on the Reuleaux triangle.