“…Contrarily to what happens for the Brewster angles this zero in the denominator is not compensated by the zero in the Fresnel coefficient. This implies that, in the critical region, θ cri + δ / k w 0 , the first order expansion of the (upper) transmitted beam (22) and, consequently, the cubic equation obtained for the angular deviation (15), with β pol instead of α pol , is valid for δ ≥ 5. For these values of δ the main contribution to the (upper) transmitted beam comes from the real part of the Fresnel coefficient and, in this case, the first order expansion represents a good approximation for the beam.…”