The observation of canonical spin-glass behavior in the pyrochlore oxide ${\text{Y}}_2{\text{Mo}}_2{\text{O}}_7$ has been a subject of considerable interest as the original structural studies were interpreted in terms of a well-ordered crystallographic model. It is widely held that the stabilization of the spin-glass state requires some level of positional disorder along with frustration. Recent reports from local probe measurements, extended x-ray-absorption fine structure (EXAFS) and ${}^{89}\text{Y}$ NMR, have been interpreted in terms of disorder involving the Mo-Mo distances (EXAFS) and multiple Y sites (NMR). This work reports results from temperature-dependent (15–300 K) neutron diffraction (ND) and neutron pair distribution function studies which can provide from the same data set information on both the average and local structures. The principal findings are that: (1) there is no crystallographic phase transition over the temperature region studied within the resolution of the ND data; (2) the diffraction data are well fitted using a fully ordered model but with large and anisotropic displacement parameters for three of the four atomic sites; (3) the pairwise real-space correlation function $G\left(r\right)$ shows clear evidence that the principal source of disorder is associated with the Y-O1 atom pairs rather than the Mo-Mo pairs, in disagreement with the interpretation of the EXAFS results; (4) fits to the $G\left(r\right)$ improve significantly when anisotropic displacements for all sites are included; (5) inclusion of a split-site position parameter for O1 improves, slightly, both the $G\left(r\right)$ fits and the Rietveld fits to the ND data; and (6) for all models the fits become worse as the temperature decreases and as the fitting range decreases. These results are qualitatively consistent with the ${}^{89}\text{Y}$ NMR observations and perhaps recent muon-spin-relaxation studies. The issue of static versus dynamic disorder is not resolved, definitively. An estimate of the distribution of exchange constants due to the disorder is made using spin-dimer analysis and compared with the Saunders-Chalker model for the generation of spin-glass behavior from “weak” disorder on geometrically frustrated lattices.

• Received 18 November 2008

DOI:https://doi.org/10.1103/PhysRevB.79.014427