A systematic study of radiative torque grain alignment in the diffuse interstellar medium

Context. Planck observations demonstrated that the grain alignment efficiency is almost constant in the diffuse ISM. Aims. We test if the Radiative Torque (RAT) theory is compatible with observational constraints on grain alignment. Methods. We combine a numerical simulation with the radiative transfer code POLARIS that incorporates a physical dust model and the detailed grain alignment physics of RATs. A dust model is designed to reproduce the spectral dependence of extinction of the ISM. From a RAMSES simulation of interstellar turbulence, we extract a cube representative of the diffuse ISM. We post-process the cube with POLARIS to get the grain temperature and RATs to simulate synthetic dust polarization maps. Results. In our simulation the grain alignment efficiency is correlated with gas pressure, but not with the RAT intensity. Because of the low dust extinction, the magnitude of RATs varies little, decreasing only for high column densities $N_H$. Comparing our maps with a uniform alignment efficiency, we find no systematic difference. The dependence of polarization fraction $p$ with $N_H$ or polarization dispersion $S$ is similar. The drop of RATs in dense regions barely affects the polarization pattern, the signal being dominated by the LOS and magnetic field geometry. If a star is inserted, the polarization increases, with no specific pattern around the star. The angle-dependence of RATs is not observed in the maps, and is weak using a uniform magnetic field. Conclusions. RATs are compatible with Planck data for the diffuse ISM such that both uniform alignment and RAT alignment lead to similar observations. To further test the predictions of RATs where an important drop of grain alignment is expected, polarization observations of dense regions must be confronted to numerical simulations sampling high column densities through dense clouds, with enough statistics.