Photometric redshifts for weak lensing tomography from space: the role of optical and near infrared photometry

We study in detail the photometric redshift requirements needed for tomographic weak gravitational lensing in order to measure accurately the dark energy equation of state. In particular, we examine how ground-based photometry (u, g, r, i, z, y) can be complemented by space-based near-infrared (near-IR) photometry (J, H), e.g. onboard the planned DUNE satellite. Using realistic photometric redshift simulations and an artificial neural network photo-z method we evaluate the figure of merit for the dark energy parameters (w(0), w(a)). We consider a DUNE-like broad optical filter supplemented with ground-based multiband optical data from surveys like the Dark Energy Survey, Pan-STARRS and LSST. We show that the dark energy figure of merit would be improved by a factor of 1.3-1.7 if IR filters are added onboard DUNE. Furthermore we show that with IR data catastrophic photo-z outliers can be removed effectively. There is an interplay between the choice of filters, the magnitude limits and the removal of outliers. We draw attention to the dependence of the results on the galaxy formation scenarios encoded into the mock galaxies, e.g. the galaxy reddening. For example, very deep u-band data could be as effective as the IR. We also find that about 10(5)-10(6) spectroscopic redshifts are needed for calibration of the full survey.