In the context of redshift space distortions (RSDs), according to Mao et al. 2012, when the power spectrum is computed, the volume element is modified by the peculiar velocity gradient, see Eq. (56) and the text above in that paper. Since the observables in 21-cm interferometers, visibilities, are measured in redshift space rather than real space, it is therefore more appropriate to use the volume element in redshift space when the power spectrum is computed and we want to account for RSDs.
I think that in practice one should multiply the input box by the gradient term. Then, in the thin optical depth limit, since we already divide the brightness temperature in 21cmFAST by that exact term, this correction term cancels and we would be left only with the known effect of RSDs (of shifting the cells according to their peculiar velocities), as Eq. (57) in the paper predicts.
In order to have this effect in tuesday, I think that calculate_ps_coeval and calculate_ps_lc should be able to receive a new optional input called los_velocity. Then, if this input is not None, the gradient term should be computed similarly according to what we do here.
In the context of redshift space distortions (RSDs), according to Mao et al. 2012, when the power spectrum is computed, the volume element is modified by the peculiar velocity gradient, see Eq. (56) and the text above in that paper. Since the observables in 21-cm interferometers, visibilities, are measured in redshift space rather than real space, it is therefore more appropriate to use the volume element in redshift space when the power spectrum is computed and we want to account for RSDs.
I think that in practice one should multiply the input box by the gradient term. Then, in the thin optical depth limit, since we already divide the brightness temperature in
21cmFASTby that exact term, this correction term cancels and we would be left only with the known effect of RSDs (of shifting the cells according to their peculiar velocities), as Eq. (57) in the paper predicts.In order to have this effect in
tuesday, I think thatcalculate_ps_coevalandcalculate_ps_lcshould be able to receive a new optional input calledlos_velocity. Then, if this input is not None, the gradient term should be computed similarly according to what we do here.