Isaksen's algorithm of the photolysis rate calculation in Luther's modification (I-L method) is analyzed. The interpretation suggested by Thompson <1984> and Madronich <1982> are also considered. The accuracy of the I-L method is tested by (1) evaluation of the transmittance and reflectance coefficients for a single homogeneous Rayleigh scattering layer against delta Eddington and discrete ordinate results and (2) comparison of I-L and delta Eddington photolysis rates. These tests show that if the I-L method is applied for calculation of twilight UV photolysis rates, its inaccuracy can exceed 50% (compare 10% according to the earlier estimations). The possible effect of such inaccuracy on model predictions is illustrated by comparing the results obtained by the same one-dimensional PCM successively employing I-L and delta Eddington radiation codes. ¿ American Geophysical Union 1992