Imitating intrinsic alignments: a bias to the CMB lensing-galaxy shape cross-correlation power spectrum induced by the large-scale structure bispectrum

Cross-correlating the lensing signals of galaxies and comic microwave background (CMB) fluctuations is expected to provide valuable cosmological information. In particular, it may help tighten constraints on parameters describing the properties of intrinsically aligned galaxies at high redshift. To access the information conveyed by the cross-correlation signal, its accurate theoretical description is required. We compute the bias toCMBlensing-galaxy shape cross-correlation measurements induced by non-linear structure growth. Using tree-level perturbation theory for the large-scale structure bispectrum, we find that the bias is negative on most angular scales, therefore mimicking the signal of intrinsic alignments. Combining Euclid-like galaxy lensing data with a CMB experiment comparable to the Planck satellite mission, the bias becomes significant only on smallest scales (l greater than or similar to 2500). For improved CMB observations, however, the corrections amount to 10-15 per cent of the CMB lensing-intrinsic alignment signal over a wide multipole range (10 less than or similar to l less than or similar to 2000). Accordingly, the power spectrum bias, if uncorrected, translates into 2 sigma and 3 sigma errors in the determination of the intrinsic alignment amplitude in the case of CMB stage III and stage IV experiments, respectively.