llzo neutrons · where the lithium sits from Zakaria Kharbouch's thesis to neutron and synchrotron beam time · Abdessamad Faik (UM6P), Irene Urcelay-Olabarria and Josu M. Igartua (EHU) · November 2026

diff-ref-struct in the project

diff-ref-struct in the project

The refinement application of the project is diff-ref-struct, a Rietveld program for the Mac in the line of FullProf and GSAS-II, written at the EHU by Josu M. Igartua: a core engine with a command line (drs), the app around it, and modules for figures and tables (Publish) and for the path of a refinement in configuration space (Explore). It is already in use here, and it is where the beam-time data will be refined.

What it has done already

  • Every structure factor of the contrast page: drs sf for the \(I\bar{4}3d\) garnet, X-rays at any wavelength or anode and neutrons with Sears's scattering lengths (the CrysFML table FullProf uses), 15 models \(\times 3\) probes.
  • The symmetry descent from \(Ia\bar{3}d\) to \(I\bar{4}3d\): drs site gave the Wyckoff positions (12a, 12b, 24d, 16c, 48e) and their site symmetries.

What it will do

  • Joint refinement of X-ray and neutron data of the same powder (the core's joint module): ID22 fixes the lattice, the profile and the heavy atoms; D2B the Li, Ga, Al and O; one model, one set of constraints (ICP composition, charge balance, site capacities), several phases (cubic and tetragonal garnet, \(\mathrm{Li_{2}ZrO_{3}}\), \(\mathrm{BaZrO_{3}}\), \(\mathrm{LiGaO_{2}}\)).
  • Sequential and parametric refinement of the 38 compositions and of the temperature ramps (P3), with the trends written straight into the fact base.
  • Explore: the correlations between lithium occupancies and displacement parameters, which decide what the data can and cannot say about lithium, shown as a path in configuration space rather than as a correlation matrix.
  • Publish: the figures and tables of the proposals and papers, typeset in LaTeX; the CIFs with esd for deposition.
  • A preliminary refinement of the two laboratory patterns of Ga–M–Sc garnets sent by Zakaria Kharbouch in 2023 (Ba 0.1 and Y 0.1), with the \(I\bar{4}3d\) model of this site: the "what the laboratory data can fix" figure of P1 and P2.

Bayesian probability refinement

The project's working spirit: each technique narrows what the structure can be, and the combination leaves little room. For the analysis it means one model per composition and every measurement a term in its probability: the ID22 and D2B patterns as likelihoods sharing the lattice and the heavy atoms; ICP-OES, TGA and EXAFS as priors carried with their uncertainties, not as hard constraints; the answer a distribution over the lithium per site, the gallium, the yttrium site fraction. For the design it means choosing the combination of probes by how much it narrows each answer, computed before the beam from the same structure factors. The structures page sets out both.

Why this application

The questions of this project are exactly where refinement programs are weakest and most opaque: occupancies of light atoms correlated with displacement parameters, several phases with overlapping lines, two radiations with different contrast, constraints from chemistry. An application that shows each step, records each choice and makes the correlations visible is the right instrument for them, and the garnets are a demanding real case for the application.