2007ApJ...669...10F

In this paper we analyze the number counts of false peaks resulting from intrinsic ellipticities in lensing convergence maps, including the alignment of source galaxies. The full noise variance in convergence κ-maps can be written as σ²₀=σ²_0ran+σ²_0corr, where σ²_0ran is the noise contributed from randomly oriented source galaxies and σ²_0corr denotes the additional noise from intrinsic alignments. However, it is observationally difficult to measure σ²_0corr, and usually only σ²_0ran can be estimated in weak-lensing observations. Thus, the observational signal-to-noise ratio is often defined with respect to σ_0ran, which we denote as ν_ran. The true signal-to-noise ratio ν in terms of σ₀is then ν=ν_ran/(1+σ²_0corr/σ²_0ran)^1/2. Given a detection threshold on ν_ran, a larger value of σ²_0corr/σ²_0ran leads to a lower threshold on ν and therefore a larger expected number of false peaks. With σ²_0corr/σ²_0ran∼10%, the average number of false peaks with ν_ran≥3.5 nearly doubles compared to that without considering the alignment, and for ν_ran≥5, the number is tripled. As a result, the efficiency of weak-lensing cluster detection declines significantly. The increase of the number of false peaks also affects the likelihood of the existence of dark clumps. On the other hand, if one can observationally distinguish false peaks and true peaks, the number of false peaks can be used to tightly constrain the level of intrinsic alignments. The CFHTLS Deep survey's 3.61 deg² lensing observations and follow-up studies find that 5 out of the 14 peaks with ν_ran>3.5 are likely to be false peaks, giving rise to a constraint of σ²_0corr≤1.6x10^–5 (1 σ) at the angular scale of 1' for galaxies at redshift z∼1. The corresponding limit on the correlation of intrinsic ellipticities is 3.2x10^–5.