in Fig. 5. The fitting curve (line) agrees well with the absorption lineshape function (circles), where the curves due to the fast component, the slow component, and the baseline are also displayed. The time constants of the fast and slow components were obtained from the fitting. Moreover, we acquired the time constants using two procedures. The first was the calculation of the absorption lineshape function from an average over eight sets of data obtained from the repeated THz measurement before the curve fitting; the second was the calculation of the lineshape function from each set of data and the averaging of the time constants obtained by the curve fitting. The time constants acquired from the two procedures were found to be almost the same. The results obtained from averaging the time constants with the standard deviation of the mean are presented below. There is a question concerning the uniqueness of the time constants derived from the curve fitting, although multi-parameter fitting is a common method for spectral analysis [32–34,41–43]; in fact, as shown for liquid water in Fig. 6(a), there is a small difference between the time constants due to the measurements in the different frequency ranges [32,40]. However, a systematic deviation due to the frequency range examined and multiparameter fitting should not influence the conclusion, because our discussion is based on the temperature dependence of the values derived from the same procedure.
Fig. 6(a) shows the temperature dependence of the time constant of the slow component (circles) with the results for liquid water ( triangles [32] and squares [41]) from literature. Three features of note are observed as the temperature is lowered. Firstly, considerable slowing down occurs above Tc in comparison to the case of liquid water. Secondly, the time constant dramatically decreases at Tc; this is ascribed to a change in the THz waveform of the sample solution at Tc, as seen in Fig. 4(b). Lastly, the temperature-dependent behavior of the time constant is similar to that of liquid water below Tc. We found that the results obtained by another series of experiments including the sample preparation show the same three features; however, the value of Tc was 285 K. This is believed to be attributed to the low accuracy and stability of the temperature control by the cryostat. In Fig. 6(a), large standard deviations of the mean are seen in the temperature range wherein considerable slowing down is observed above Tc. This is reasonable because the time constant fluctuates more due to the temperature uncertainty in the temperature range where it responds more sensitively to the temperature. As for the fast component, the temperature dependence of the time constant is depicted in Fig. 6(b) and the value is a few ps. An abrupt decrease in the time constant is also seen at Tc. In Fig. 6(c), the amplitudes of the fast (squares) and slow (circles) relaxation components together with the constant baseline (triangles) are plotted as a function of temperature. All the amplitudes somewhat drop at Tc.