Glossary#

In the statistics below, the sums run over the \(N\) data points inside the fit window with a positive weight \(w_i\). \(P\) is the number of parameters the engine varied.

ADP, Uiso, Uaniso: Atomic displacement parameter: the mean-square displacement of an atom about its site, from vibration or static disorder. Uiso is one isotropic value; Uaniso is the six values U11 U22 U33 U12 U13 U23. Both are in Å². An atom’s ADP field says which one it states.
Bound: The min or max of a [value, refine_flag, min, max] parameter, or null for an open side. What a bound means depends on the engine; see Bounds.
Calibrant: A standard reference material, such as LaB6, Si or CeO2, measured with the same wavelength and geometry as your sample. Refining it gives the instrument’s peak shape and the corrections that apply.
Canonical space-group name: The one spelling gemmi gives each space-group setting, such as "P m -3 m" or "F d -3 m:2". Recipes must use it exactly.
Chebyshev background: A smooth background written as a sum of Chebyshev polynomials over the fit window. It is the background model every engine shares.
Core block: A part of the payload that is the same in every engine: xrd_data, fit_range, the Chebyshev background and a phase’s space_group, unit_cell and atoms.
Disposition: How a value of the output recipe came about: default_applied, held_by_symmetry, held_by_user, requested_not_varied or varied.
Engine: The refinement program that does the calculation: GSAS-II, TOPAS or easydiffraction.
Envelope: The fixed shape of a run’s result, shared by successful, generated and failed runs. Its version is envelope_version.
ESD, s.u.: Estimated standard deviation, the uncertainty of a refined value. The IUCr calls the same quantity the standard uncertainty (s.u.). PowderLine’s columns are named esd, as in the engines.
Execution mode: Where a GSAS-II recipe runs: auto (the default: try the GSAS-II server, else run in-process), server (the server only) or subprocess. Despite its name, subprocess runs the refinement in PowderLine’s own process (in-process). See Run at a beamline.
Fit window: The 2θ range that is fitted, set by fit_range. Every engine fits exactly the points inside it with a positive weight.
Gateway: The adapter between PowderLine and one engine. It knows the engine’s recipe schema, how to run the engine and how to read its output.
Generate-only: A TOPAS run where no tc.exe is found. PowderLine writes the TOPAS input files and returns status: "generated".
GoF: Goodness of fit, \(R_\mathrm{wp} / R_\mathrm{exp}\). Close to 1 when the model fits the data to within its weights.
GSAS-II server: A local process, one per user, that keeps GSAS-II loaded so that repeated runs skip its start-up time.
Metadata: The optional free-form metadata block of a recipe. PowderLine never reads it and carries it into the output recipe.
Multiplicity: The number of equivalent positions of an atom’s site in the unit cell. It follows from the space group and the site.
Output recipe: The recipe a run returns: your recipe with the refined values written back. It is ready to be the input of the next run.
Payload: The payload block of a recipe: the refinement itself, written in the engine’s terms.
Provenance: The record of where a result came from. PowderLine writes it into the output recipe as metadata.powderline_provenance: the parent run ID, the engine and its version, and the PowderLine version.
Recipe: A complete refinement as structured data: a dictionary in Python, serialized as JSON when it is stored or sent.
Refine flag: The second entry of a parameter: true to refine the value, false to hold it.
Rexp: Expected R factor: the Rwp a correct model would give if the only misfit were the noise the weights describe, \(R_\mathrm{exp} = 100 \sqrt{(N - P) / \sum w_i y_{\mathrm{obs},i}^2}\), in percent. It is not a lower bound: Rwp can fall below it when the weights overstate the noise.
Run ID: A UUID that identifies one run (result["run_id"]). It becomes the parent_run_id of the output recipe.
Rwp: Weighted-profile R factor: \(R_\mathrm{wp} = 100 \sqrt{\sum w_i (y_{\mathrm{obs},i} - y_{\mathrm{calc},i})^2 / \sum w_i y_{\mathrm{obs},i}^2}\), in percent.
Schema: The exact definition of what a recipe may contain. A recipe is checked against the core schema and one engine schema, each with its own version.
Setting, origin choice: Different ways to describe the same space group: origin choice 1 or 2 (:1, :2), hexagonal or rhombohedral axes (:H, :R), and the unique axis and cell choice of a monoclinic group. The canonical name states the setting.
Simulation: A run that refines nothing and only calculates the pattern from the stated values. A recipe is a simulation when it sets refinement_cycles: 1 (GSAS-II), iters: 0 (TOPAS), or no refine flag (easydiffraction); the result then has simulation_mode: True.
Single-peak fitting (SPF): Fitting individual peaks, each with its own position, intensity and widths, without a crystal structure. The gsasii.spf and topas.spf workflows.
Special position: A site that lies on a symmetry element, so some of its coordinates are fixed or tied to each other, such as (x, x, x) or (1/2, 0, 0).
Structured warning: A warning as data: {code, message, field_path[, level]}. The codes are on Messages.
Tie group: Parameters that symmetry makes one parameter, such as the cubic a, b and c. Every member is stated, with the same refine flag.
Validation: Checking a recipe against its schemas and its crystallographic symmetry before anything runs. It needs no engine.
Workflow: What a recipe does with its engine: Rietveld refinement (rietveld) or single-peak fitting (spf). The second part of schema_name.
χ²_red: Reduced chi-squared, \(\chi^2_\mathrm{red} = \sum w_i (y_{\mathrm{obs},i} - y_{\mathrm{calc},i})^2 / (N - P)\). It equals GoF².