Polarization leakage in epoch of reionization windows: II. Primary beam model and direction dependent calibration

Leakage of diffuse polarized emission into Stokes I caused by the polarized primary beam of the instrument might mimic the spectral structure of the 21-cm signal coming from the epoch of reionization (EoR) making their separation difficult. Therefore, understanding polarimetric performance of the antenna is crucial for a successful detection of the EoR signal. Here, we have calculated the accuracy of the nominal model beam of LOFAR in predicting the leakage from Stokes I to Q, U by comparing them with the corresponding leakage of compact sources actually observed in the 3C295 field. We have found that the model beam has errors of less than or equal to 10% on the predicted levels of leakage of ~1% within the field of view, i. e. if the leakage is taken out perfectly using this model the leakage will reduce to $10^{-3}$ of the Stokes I flux. If similar levels of accuracy can be obtained in removing leakage from Stokes Q, U to I, we can say, based on the results of our previous paper, that the removal of this leakage using this beam model would ensure that the leakage is well below the expected EoR signal in almost the whole instrumental k-space of the cylindrical power spectrum. We have also shown here that direction dependent calibration can remove instrumentally polarized compact sources, given an unpolarized sky model, very close to the local noise level.