Evidence for anisotropic polar nanoregions in relaxor Pb(Mg1/3Nb2/3)O-3: A neutron study of the elastic constants and anomalous TA phonon damping in PMN

We use neutron inelastic scattering to characterize the acoustic phonons in the relaxor Pb(Mg1/3Nb2/3)O-3 (PMN) and demonstrate the presence of a highly anisotropic damping mechanism that is directly related to short-range polar correlations. For a large range of temperatures above T-c similar to 210 K, where dynamic, short-range polar correlations are present, acoustic phonons propagating along [1 (1) over bar0] and polarized along [110] (TA(2) phonons) are overdamped and softened across most of the Brillouin zone. By contrast, acoustic phonons propagating along [100] and polarized along [001] (TA(1) phonons) are overdamped and softened for a more limited range of wave vectors q. The anisotropy and temperature dependence of the acoustic phonon energy linewidth Gamma are directly correlated with neutron diffuse scattering cross section, indicating that polar nanoregions are the cause of the anomalous behavior. The damping and softening vanish for q -> 0, i.e., for long-wavelength acoustic phonons near the zone center, which supports the notion that the anomalous damping is a result of the coupling between the relaxational component of the diffuse scattering and the harmonic TA phonons. Therefore, these effects are not due to large changes in the elastic constants with temperature because the elastic constants correspond to the long-wavelength limit. We compare the elastic constants we measure to those from Brillouin scattering experiments and to values reported for pure PbTiO3. We show that while the values of C-44 are quite similar, those for C-11 and C-12 are significantly less in PMN and result in a softening of (C-11 - C-12) over PbTiO3. The elastic constants also show an increased elastic anisotropy [2C(44)/(C-11 - C-12)] in PMN versus that in PbTiO3. These results are suggestive of an instability to TA(2) acoustic fluctuations in PMN and other relaxor ferroelectrics. We discuss our results in the context of the current debate over the "waterfall" effect and show that they are inconsistent with acoustic-optic phonon coupling or other models that invoke the presence of a second, low-energy optic mode.