comment ͑GL͒ on our paper 1 "A 250 GHz ESR study of o-terphenyl: Dynamic cage effects above T C " by Earle and co-workers ͑EMPF͒ gives us an opportunity to clarify some issues related to the data analysis performed and conclusions drawn. The first point we wish to address is the role of the rotational correlation function ͑RCF͒ in calculating the mean rotational correlation time ͗͘. Most of GL's comment hinges on our analysis of the role of the nonexponential decay of the RCF in discussing our

more »

... esults in the context of other studies on o-terphenyl ͑OTP͒ as a glass-forming fluid. We observed that the probe rotational diffusion tensors ͑as distinct from ͗͘ Ϫ1 ) that we determined from our analysis of the ESR spectra followed an Arrhenius ͑i.e., activated͒ behavior above the crossover temperature T C , whereas the OTP selfdiffusion rates determined by other spectroscopies, e.g., nuclear magnetic resonance ͑NMR͒ and dynamic light scattering, which are related to ͗͘, were better described by a non-Arrhenius decay, which was typically fit to a Stokes-Einstein-Debye ͑SED͒ law above T C . The RCF's that result from our ESR analysis in terms of the cage model do not decay as a simple exponential, consistent with other studies. This is due to the dynamic coupling between probe and cage in our analysis. Thus we utilized ͗͘ in order to facilitate comparison of our work with results from other techniques ͑Ref. 2, and references therein͒. As an additional device for comparison of our results with those of other workers, we fit a stretched exponential of the form exp(Ϫ(t/ 0 ) ␤ ) to the RCF ͑i.e., G R (t)), and extracted 0 and ␤ from the fits. We then computed 1/6͗͘ by means of Eq. ͑4͒ of EMPF, which we reproduce here for convenience, viz. ͑1͒ Unfortunately, the graphics macros used to generate Fig. 10 in EMPF had an error in the term containing the Euler gamma function. This led to a spurious reduction in our values of 1/6͗͘, which GL rightly criticize. Figure 1 of GL shows the actual values of 1/6͗͘ and R 0 computed from our ͑corrected͒ Tables V and VI. 3 Clearly Fig. 1 of GL shows that the agreement between our results and the SED are only slightly improved by plotting 1/6͗͘. However, a more careful analysis of our RCF's revealed that our previous analysis had neglected a significant portion of the long-time tail that is associated with the slow relaxation of the cage in which the probe is diffusing. The longtime tail is only significant in the presence of a nonvanishing interaction potential, so it only affects the lower temperature data, where the RCF departs significantly from a simple exponential decay. When we computed the average correlation time directly, i.e., by numerically integrating the RCF ͓cf. Eq. ͑1͔͒ ͑instead of first fitting a stretched exponential and then using the analytical result in terms of the Euler gamma function͒ we found that the corrected ͗͘ values were rather close to those that we had originally computed with the faulty algorithm, as we show in our Fig. 1 . Thus, GL's criticisms, based on our previously published results, are no longer appropriate once this second correction is made. The G R (t) for our model, especially near T C , is closer to bi-exponential, where the fast process corresponds to relaxation of the probe molecule, and the slow process corresponds to relaxation of the cage. In support of this behavior, a recent molecular dynamics study 4 finds that the RCF of the "probe" system has biexponential character with the slow process interpreted in terms of a cage effect. On the other hand, an improved version of our cage model, 5 which incorporates a distribution of cage potentials, shows that this feature can smooth out the biexponential behavior of the G R (T) leading to an RCF that is closer in character to a stretched exponential. Above T C , the published values for ␤ vary from 0.6 to 1. Some light scattering experiments 6 show ␤ increasing monotonically from 0.6 near T C to unity near the melting temperature T M consistent with our analysis. Other light scattering experiments 7 suggest ␤ϭ0.78 in this region. Neutron scattering 8 and 2 HϪT 1 measurements 9 suggest that ␤ Ϸ0.6 in this temperature range. For TϾ330 K, i.e., above T M , the 2 HϪT 1 measurements do not exclude the possibility that ␤ may increase as the temperature increases. 9 Our estimates of 0 and ␤ are cutoff dependent when there is a significant cage contribution. We find that the estimate of 0 increases as the cutoff is moved to longer times, with smaller changes in ␤ and the uncertainties in both parameters increase. In light of the above, we offer the following summary. Our results are consistent with a temperature-dependent ␤ that is unity above T M and decreasing with decreasing temperature, for TϾT C consistent with the light-scattering results of Fischer and co-workers. 6 If there were a significant