The estimate of redshifts through photometric techniques is one of the most promising techniques in deep universe studies, and certainly a key point to optimize field surveys with Megacam. We briefly discuss the limits and the expected accuracy of such a method through simulations. These results determine the observational strategy for some suggested programs.

The technique used to compute photometric redshifts (hereafter *z*_{phot}) is a standard ^{2} minimization procedure. The observed spectral energy distribution (SED) of each galaxy is compared to a set of template spectra

(5) |

where

We have studied the accuracy of *z*_{phot} as a function of the relevant parameters, namely the photometric errors, the filter bands available and the type of galaxy. For each test galaxy, a value of *z*_{phot} has been computed as well as the z corresponding to different confidence levels, which give an estimate of individual errors. Different sets of simulated catalogues were created for this exercise, basically reproducing the photometric properties of two extreme spectrophotometric types of galaxies, taken at different ages and redshifts, with and without evolution: E/S0 galaxy (evolving 0.1 Gyr burst, *z*_{form} = 5.3, aged 15 Gyr at z = 0, with H_{0 }= 50 km s^{-1} Mpc^{-1} and q_{0} = 0.1), and a constant star forming system. Photometric errors were introduced as gaussian noise distributions of fixed FWHM for each filter band, and they are uncorrelated for different filters.

Figure 10 shows the behaviour of a set of simulated objects when the *z*_{phot}, computed from UBVRI photometry, is compared to the true *z* (model). BVRI filter responses are similar to Focam, and U is a ``red'' Johnson (_{c} ~ 3750 Å). As expected, the errors on individual galaxies (at 90% confidence level) become huge at ~1.2 *z* ~2.2 because of the lack of strong spectral features in the visible band. This problem is solved when near-IR photometry is included.

The dispersion in the estimate of *z*_{phot} is strongly dependent on photometric uncertainties. There is no significant gain for *m* 0.1, but the dispersion and the number of multiple solutions with similar weight increases rapidly up to *m* ~ 0.3, which is probably a limiting value for individual objects. Including near-IR J and K photometry does not improve significantly the uncertainties in *z*_{phot} outside the ~1.2 *z* ~2.2 range, where it is absolutely needed, and where the individual errors are strongly reduced. When the number of filters in the visible band is reduced to UBRI, the dispersion in *z*_{phot} increases roughly to 0.09 for *z* ~1.5 and 0.3 for *z* ~2.2. The dispersion in *z*_{phot} is quite similar to the values found in the literature, even when the techniques used are appreciably different (Brunner et al. 1997, Conolly et al. 1995), but it is extremely difficult to compare the accuracy of *z*_{phot} as a function of the relevant parameters (photometric errors in particular).

*Connolly A.J., Csabai I., Szalay A.S., 1995, AJ 110, 2655*

*Gwyn S.D.J., Hartwick F.D.A. 1996, ApJ 468, L77*

*Sawicki M.J., Lin H., Yee H.K.C., 1997, AJ 113, 1*

tmca@cfht.hawaii.edu