We study the impact of catastrophic errors occurring in the photometric redshifts of galaxies on cosmological parameter estimates with cosmic shear tomography. We consider a fiducial survey with nine-filter set and perform photo-z measurement simulations. It is found that a fraction of 1% galaxies at z spec ~ 0.4 is misidentified to be at z phot ~ 3.5. We then employ both χ2 fitting method and the extension of Fisher matrix formalism to evaluate the bias on the equation of state parameters of dark energy, w 0 and w a , induced by those catastrophic outliers. By comparing the results from both methods, we verify that the estimation of w 0 and w a from the fiducial five-bin tomographic analyses can be significantly biased. To minimize the impact of this bias, two strategies can be followed: (1) the cosmic shear analysis is restricted to 0.5 z 2.5, where catastrophic redshift errors are expected to be insignificant; (2) a spectroscopic survey is conducted for galaxies with 3 z phot 4. We find that the number of spectroscopic redshifts needed scales as N spec ∝ f cata × A, where f cata = 1% is the fraction of catastrophic redshift errors (assuming a nine-filter photometric survey) and A is the survey area. For A = 1000deg2, we find that N spec 320 and 860, respectively, in order to reduce the joint bias in (w 0, wa ) to be smaller than 2σ and 1σ. This spectroscopic survey (option 2) will improve the figure of merit of option 1 by a factor ×1.5 thus making such a survey strongly desirable.
展开▼