Cations distribution in synthetic ( MgFe2O4 and FeAl2O4 ) spinels by precession electron diffraction tomography

Abstract

Since recently, a method using precession electron diffraction tomography (PEDT) and dynamical calculations of the diffracted intensities has been developed, allowing the structure determination and refinement at the nanoscale in a TEM [1, 2]. We present the application of this method to the refinement of (MgFe 2 )O 4 and (FeAl 2 )O 4 spinels in order to determine the Fe, Mg and Al cations distributions on specific sites of the structure. This determination is essential for understanding the electrical or magnetic properties of spinels, their chemical reactivity or to retrieve their thermal history in the field of geosciences. Studied samples were obtained by the flux growth method and for different chemical compositions [3]. Some samples have also been heat‐treated (24 h at 1000°C followed by a rapid quench) in order to intentionally induce some structural disorder associated with Fe, Al and Mg occupancy variations on the tetrahedral and octahedral sites of the structures. Single crystals were large enough to be also studied and refined by X‐ray diffraction. The structures thus deduced serve as model for comparison with the much more local results obtained using PEDT. Thin sections for TEM observations are extracted from the crystals previously studied by XRD via two methods: Focused Ion Beam (FIB) thinning and simple mechanical grinding. The full refinement method is precisely described in [1]. It is based on the acquisition of a serie of precession (angle varying from 1 to 2°) diffraction patterns continuously acquired for various tilt angles of the sample (+/− 45 to 60°, by step of 1°) as currently done in tomography. The 3D reciprocal space of the structure is then reconstructed (Fig. 1) and experimental I hkl intensities are integrated using the softwares PETS and JANA2006. From the I hkl data set, the structure is solved, using conventional X‐ray methods based on the kinematical approximation implemented in JANA2006, in order to obtain a first reliable structure model (Fig.2). The model is finally accurately refined (atomic position and occupancies) using least‐squares methods based on the comparison of experimental intensities with calculated ones using the multi‐beam dynamical theory, taking into account interactions between diffracted beams for a given thickness and orientation of the sample. In this work, we will describe and discuss the influence of the various experimental and computational parameters on the accuracy and precision of the PEDT refinement results. These parameters are: i) thinning method (FIB / grinding), ii) precession angle, iii) refinement procedure (including thickness and/or orientation refinement). We will also discuss the need of constraining or not the chemical composition of the samples during the refinement procedure in order to obtain the most reliable results.