Computer simulations have been carried out to test the recently proposed model for the nanodomain structure of relaxor ferroelectrics such as lead zinc niobate (PZN). In this recent model it was supposed that the polar nanodomains are three-dimensional, that the observed diffuse rods of scattering originate from the boundaries between domains and that the Pb displacements may be directed along \langle {100} \rangle, \langle {111} \rangle or \langle {110} \rangle. This is in marked contrast to a previously published model, which described the polar domains as thin plates with Pb displacements confined to the \langle {110} \rangle directions within the essentially two-dimensional domains. The present results confirm that \langle {100} \rangle and \langle {111} \rangle types of Pb displacement are viable possibilities, but the number of domain boundaries required to produce sufficiently strong diffuse rods of scattering means that individual domains cannot be described as three-dimensional and must still be relatively thin. The current work has been carried out with no direct involvement of theB-site cation ordering, which many workers assume is necessary to understand the formation of the polar nanodomains. While it may be true that theB-site cation distribution could provide an underlying perturbation field that might ultimately limit the extent of any polar domain, it is certainly not necessary to produce the observed scattering effects.
