Rietveld Refinement with GSAS-II: A Practical Workflow for Powder XRD

A structured guide to preparing powder diffraction data, selecting a structural model, refining parameters in GSAS-II, and evaluating fit quality without overinterpreting the result.

COMPUTATIONAL TOOLS · PRACTICAL TUTORIAL 05

Rietveld refinement is a whole-pattern method used to compare an observed powder diffraction profile with a calculated profile derived from a crystallographic model. It can provide lattice parameters, profile information, phase fractions, and structural parameters when the data and model support them.

The method is not a button that converts an XRD scan into a definitive crystal structure. Refinement is an inverse problem: several parameters can influence the calculated pattern in similar ways. A stable and scientifically useful result therefore depends on data quality, a justified starting model, sensible parameter constraints, and critical inspection of the residuals.

This tutorial describes a general workflow in GSAS-II. Menu names and dialog layouts can vary between software versions; the scientific sequence is more important than a particular screen arrangement.

1. What Rietveld refinement actually minimizes

The calculated profile is compared with the observed intensities over the selected diffraction range. In simplified form, a weighted least-squares objective is:

\[ M=\sum_i w_i\left(y_i^{obs}-y_i^{calc}\right)^2 \]

where \(y_i^{obs}\) and \(y_i^{calc}\) are the observed and calculated intensities at data point \(i\), and \(w_i\) is the corresponding statistical weight.

The calculated pattern depends on the structural model, scale factor, background, peak positions, profile shape, instrument parameters, preferred orientation, and other terms selected for the refinement.

A low value of a residual statistic does not guarantee that every refined parameter is meaningful. Parameter correlations, model assumptions, counting statistics and unmodelled effects all matter.

2. Prepare the diffraction data first

Before opening GSAS-II:

• Confirm the radiation source and wavelength (for example, Cu Kα) and whether the data have been treated as Kα₁-only or contain the doublet.
• Preserve the original raw data and record any preprocessing.
• Check the scan range, step size, counting statistics and obvious artefacts.
• Verify the sample identity, composition, synthesis history and likely phases.
• Check for sample displacement, zero offset, fluorescence, preferred orientation and broad amorphous background.
• Use a calibration standard or validated instrument profile when accurate profile parameters are required.

Do not smooth, strip Kα₂, subtract a background or truncate data without documenting the operation. Preprocessing can alter peak shape and uncertainty.

3. Import data and inspect the pattern

Create a new GSAS-II project and import the powder diffraction histogram using the appropriate importer for the file format. Confirm that the instrument parameters and wavelength are correct.

Inspect the observed pattern before adding a structural phase:

1. Confirm that the x-axis is the expected 2θ or other measurement coordinate.
2. Check that the scan range contains the reflections needed to distinguish plausible phases.
3. Identify obvious outliers, holder peaks and fluorescence/background problems.
4. Confirm that the instrument profile corresponds to the measurement configuration.
5. Save the project under a clear, versioned filename.

The exact import options depend on the source file and instrument. If the file does not contain instrument metadata, enter or load the correct parameters from the instrument record rather than guessing.

4. Add a crystallographic phase

A phase model can be imported from a CIF or built from a known structural description. Before refinement, verify:

• Chemical formula and oxidation-state assumptions.
• Crystal system, space group and unit-cell setting.
• Lattice parameters and their units.
• Atomic species, fractional coordinates and occupancies.
• Site multiplicities and the resulting chemical composition.
• Whether the model is appropriate for the sample temperature and composition.

A CIF is a starting model, not an automatic validation. Check that the cell setting and atomic contents are consistent with the material under study. For doped or non-stoichiometric materials, do not refine occupancies freely unless the diffraction data have sufficient sensitivity and the model is chemically constrained.

5. Refine parameters in a controlled sequence

A conservative sequence helps reduce parameter correlation. The precise sequence may need adjustment for the instrument and material.

01 · PROFILEBackgroundModel the baseline without fitting away real peaks

02 · SCALEScale factorEstablish the overall calculated intensity

03 · POSITIONCell and zeroRefine justified lattice and position terms

04 · SHAPEPeak profileRefine width and shape with instrument awareness

05 · STRUCTUREAtomic termsOnly add parameters supported by data quality

Step A — Background and scale

Choose a background function that follows the measured baseline without absorbing broad diffraction features. Start with a modest number of terms. Refine the scale factor and inspect whether the calculated intensity is in the correct range.

Step B — Peak positions and lattice parameters

After confirming the wavelength and instrument calibration, refine justified peak-position terms and lattice parameters. A zero shift or specimen-displacement term should be based on the measurement geometry and standard/reference information. Do not allow multiple correlated position terms to compensate for an incorrect wavelength or instrument setup.

Step C — Peak profile

The profile function describes the peak shape and width. Instrumental broadening, specimen size, microstrain, axial divergence and wavelength distribution can all contribute. Use an appropriate instrument profile where available. Do not interpret refined broadening as crystallite size until instrumental contributions and model assumptions have been addressed.

Step D — Preferred orientation and asymmetry

Add preferred-orientation or asymmetry corrections only when the pattern and specimen geometry justify them. These terms can improve the fit but may also compensate for an incorrect structural model or sample preparation artefact.

Step E — Structural parameters

Atomic coordinates, displacement parameters, site occupancies and other structural variables are more demanding. Refine only parameters that the diffraction data can meaningfully determine. Apply chemically and crystallographically justified constraints, and report them.

6. How to evaluate the refinement

Inspect more than the numerical residual:

• Observed and calculated profiles: do the peak positions, widths and intensities agree across the full range?
• Difference curve: are there systematic positive/negative features around particular reflections?
• Background: does the model follow the baseline without distorting broad features?
• Parameter values: are they chemically and crystallographically plausible?
• Uncertainties and correlations: are key parameters well determined or strongly coupled?
• Residual indices: report Rwp, Rp, goodness-of-fit or other statistics used by the software, with definitions where needed.

A refinement can produce a visually low residual while retaining systematic errors or an inappropriate phase model. Conversely, a complex multiphase pattern may retain residual features even when the main structural model is reasonable. Interpret fit quality in the context of counting statistics, data resolution and the purpose of the analysis.

7. Common pitfalls

• Refining too many parameters at once: start with a stable, simple model and add complexity only when justified.
• Using an incorrect CIF setting: verify the cell, space group, composition and atomic contents.
• Ignoring the instrument profile: specimen broadening cannot be separated reliably without instrument information.
• Treating Rwp as a truth score: residual indices do not prove that the model is unique or correct.
• Refining occupancies without constraints: occupancy, scale, displacement and composition can be strongly correlated.
• Calling a refined cell expansion proof of doping: temperature, stress, oxygen non-stoichiometry, calibration and phase changes may also shift lattice parameters.
• Reporting crystallite size as particle size: coherent diffraction-domain size and microscopy particle size are different quantities.

8. Reporting a reproducible GSAS-II analysis

A publication-quality report should identify:

1. The instrument, radiation, wavelength, geometry and scan conditions.
2. Data preprocessing and instrument-parameter source.
3. Structural model/CIF source, space group, cell setting and phase composition.
4. Background and peak-profile functions.
5. Parameters refined, fixed or constrained, with the reason for each choice.
6. Preferred-orientation, absorption, displacement or other corrections used.
7. Rietveld plots showing observed, calculated and difference profiles.
8. Residual indices, key refined values, uncertainties and relevant correlations.
9. Any limitations, including unresolved impurity peaks or parameter non-uniqueness.

Further reading

• B. H. Toby and R. B. Von Dreele, “GSAS-II: the genesis of a modern open-source all purpose crystallography software package,” Journal of Applied Crystallography 46, 544–549 (2013). https://doi.org/10.1107/S0021889813003531
• H. M. Rietveld, “A profile refinement method for nuclear and magnetic structures,” Journal of Applied Crystallography 2, 65–71 (1969). https://doi.org/10.1107/S0021889869006558
• R. A. Young (ed.), The Rietveld Method, Oxford University Press (1993).

Related articles

• Reading an XRD pattern: what the peaks can and cannot tell you
• Electrochemical impedance spectroscopy (EIS)
• How to calculate relative density

GSAS-II is an analysis tool, not a substitute for crystallographic judgement. A defensible refinement is one whose model, parameter choices and uncertainty are consistent with the experiment and the chemistry.