Experimental

We measured the spectra of SrB₆ and CaB₆ with either sweeping with the field or sweeping with the frequency. Sweeping with the field has the advantage of having always the same conditions for the detection (tank circuit, amplifier, ...) and sweeping with the frequency has the advantage of having a high resolution. In order to investigate general behaviour we measured first with sweeping the field. But it turned out that if there were any unexpected effects, they are too small to be seen with field sweeps and we had to make use of frequency sweeps. Since in our samples the lines are compareably narrow we had to take care of not having a wrong idea about their width because of irradiating with too short pulses. Some fine-tuning enabled us to measure with pulses of 40 s duration -- at least at temperatures below 20 K.

In order to determine a possible, but rather unlikely Knight shift we had to gauge the field because it may have shifted a bit since the last gauging. We did it by simply measuring the ⁶⁵Cu and ⁶³Cu resonances because we know the exact Knight shift of copper and we have a lot of Cu in the coil. It finally turned out that the field has shifted by 0.25%, a bit less than the quadrupole splitting in our samples. In addition to gauging the field we switched the magnet to persistent mode and measured SrB₆, CaB₆ and LaB₆ at exactly the same field.

We obtained the data again by integrating over the whole echo and plotting the echo intensity as a function of the frequency or the field respectively. In addition to this procedure we also used the fast fourier transform of a single echo obtained with pulses as short as possible. With that procedure we can obtain the structure of the whole central line at once. As a third technique we sweeped with the field, and, instead of integrating the single echos, we added up their fast fourier transforms, each of them shifted by the onset of the field at which it has been taken (private communication with W.G. Clark). We made a note of this last technique to implement it in a new data analysis program.

In order to interpret the spectra (and to learn for future applications) we wrote a fitting program in Metrowerks C++. Details about the implementation are shown in the appendix. With fits to spectra one can quickly check if and to what extent some specific effects influence the line shape and if the actually measured line could be explained. We tested for quadrupole splitting, Knight-shifts and inhomogeneous broadening.