Redshift estimation from low-resolution prism spectral energy distributions with a Next Generation Space Telescope multiobject spectrograph

We discuss the utility of a low-resolution prism as a component of a multiobject spectrometer for NASA's proposed Next Generation Space Telescope (NGST). Low-resolution prism spectroscopy permits simultaneous observation of the 0.6-5 mu m wavelength regime at R less than or similar to 50. Such data can take advantage of modern techniques in spectral energy distribution (SED) fitting to determine source redshifts, sometimes called "photometric redshifts." We compare simulated prism observations with filter imaging for this purpose with NGST. Low-resolution prism observations of galaxy SEDs provide a significant advantage over multifilter observations for any realistic observing strategy. For an ideal prism in sky background-limited observing, the prism has a signal-to-noise ratio advantage of the square root of the resolution over serial observations by filters with similar spatial and spectral resolution in equal integration time. For a realistic case the advantage is slightly less, and we have performed extensive simulations to quantify it. We define strict criteria for the recovery of input redshifts, such that to be considered a success, redshift residuals must be delta(z) < 0.03 + 0.1 log z. The simulations suggest that in 10(5) s, a realistic prism will recover (by our definition of success) the redshift of similar to 70% of measured objects (subject to multiobject spectrograph selection) at K(AB) < 32, compared to less than 45% of the objects with serial filter observations. The advantage of the prism is larger in the regime of faint (K(AB) > 30) objects at high redshift (z > 4), where the prism recovers 80% of redshifts, while the filters recover barely 35% to similar accuracy. The primary discovery space of NGST will be at the faintest magnitudes and the highest redshifts. Many important objects will be too faint for follow-up at higher spectral resolution, so prism observations are the optimal technique to study them. Prism observations also reduce the contamination of high-redshift samples by lower redshift interlopers.