Not yet reviewed by Pith; the record is open.

Detection of the 21-cm signal coming from the epoch of reionization (EoR) is challenging especially because, even after removing the foregrounds, the residual Stokes $I$ maps contain leakage from polarized emission that can mimic the signal. Here, we discuss the instrumental polarization of LOFAR and present realistic simulations of the leakages between Stokes parameters. From the LOFAR observations of polarized emission in the 3C196 field, we have quantified the level of polarization leakage caused by the nominal model beam of LOFAR, and compared it with the EoR signal using power spectrum analysis. We found that at 134--166 MHz, within the central 4$^\circ$ of the field the $(Q,U)\rightarrow I$ leakage power is lower than the EoR signal at $k<0.3$ Mpc$^{-1}$. The leakage was found to be localized around a Faraday depth of 0, and the rms of the leakage as a fraction of the rms of the polarized emission was shown to vary between 0.2-0.3\%, both of which could be utilized in the removal of leakage. Moreover, we could define an `EoR window' in terms of the polarization leakage in the cylindrical power spectrum above the PSF-induced wedge and below $k_\parallel\sim 0.5$ Mpc$^{-1}$, and the window extended up to $k_\parallel\sim 1$ Mpc$^{-1}$ at all $k_\perp$ when 70\% of the leakage had been removed. These LOFAR results show that even a modest polarimetric calibration over a field of view of $\lesssim 4^\circ$ in the future arrays like SKA will ensure that the polarization leakage remains well below the expected EoR signal at the scales of 0.02-1 Mpc$^{-1}$.