“…Although this diagnosis does not always correspond to the frequency fluctuation directly, it is simple and useful. 3,4 When the averaging time was 2 s, the best stability was obtained as 8 ϫ 10 Ϫ14 , which corresponds to a fluctuation of 30 Hz. For a longer term ͑Ͼ1 hour͒, however, drift of the zerosignal level, which is caused mainly by an additional polarization rotation independent of the atomic birefringence, becomes a major problem.…”