Abstract
The linewidth in level-crossing experiments as normally carried out is the sum of the natural widths of the levels concerned. We have reduced the linewidth by taking the fluorescent light, not from all atoms in the assembly, but from a sample biased in favour of those which have survived as excited atoms for times longer than the average. This was achieved by pulse techniques. Improved resolution was obtained of the components on the wing of the zero-field level-crossing curve for the level 3 2P3/2 in 23Na.
The values obtained for the hyperfine-coupling constants are a = 18·5 ± 0·40 MHz, b = 3·0 ± 0·6 MHz, in agreement with values found in double-resonance experiments. The present values do not represent the limit of the method, since the experiment was designed to investigate the technique, not to make precision measurements.
A theoretical analysis of the method reveals that level-crossing curves obtained by this technique may, in certain circumstances, have undesirable satellites. It is shown how the line profile may be controlled and improved without too great a sacrifice in statistical accuracy.