Rietveld
Fit the whole pattern; watch the difference flatten.
Rietveld refinement fits a calculated diffraction pattern to the whole measured one, every peak, every overlap, simultaneously, and minimizing the difference curve yields lattice parameters, phase fractions, and structural detail no single-peak analysis can give.
What it measures
Single-peak XRD analysis reads positions (Bragg’s law, nλ = 2d·sinθ) and widths (Scherrer) one reflection at a time. Rietveld instead refines a full structural model against the entire pattern, minimizing Σwᵢ(yᵢ,obs − yᵢ,calc)² over every measured point. Because all peaks constrain the model simultaneously, it extracts:
- Lattice parameters to high precision, in layered cathodes, the c/a ratio and site occupancies track lithiation state and cation mixing (Ni²⁺ on Li sites).
- Quantitative phase fractions: weight percentages of each phase in a mixture, from full-pattern intensities rather than one hand-picked peak pair.
- Fit-quality metrics: R_wp (weighted profile residual) and GoF = R_wp/R_exp; the refinement is credible when the difference curve is flat and GoF approaches ~1.
How to read the output
The difference curve is the lie detector. Residual structure under specific peaks means a phase is missing from the model or a preferred-orientation correction is absorbing what it shouldn’t; a wavy residual everywhere means the background model is wrong. Read the numbers with their errors: a lattice parameter quoted to five decimals with a bumpy difference curve is precision without accuracy. And watch the refinement strategy, parameters released in a sensible order (background and scale before atomic positions and occupancies) converge to physics; everything released at once converges to whatever minimizes the math.
A real use case
A cathode team shortens the calcination hold to lift furnace throughput and needs to know whether the material paid for it. Plain pattern inspection says the lots are identical, same peaks, same phases. Refinement of the layered structure says otherwise: the shortened-hold lot shows the (003)/(104) intensity relationship and refined site occupancies drifting toward more Ni on lithium sites, cation mixing creeping up, exactly the defect that later reads as first-cycle capacity loss and sluggish kinetics. The hold time goes back, and refined occupancy joins the release-criteria list for calcination changes.
Common mistakes
- Refining everything at once. Release parameters in stages; simultaneous release of background, profile, and occupancies finds numerical minima, not structures.
- Trusting R_wp alone. A low residual with unphysical thermal parameters or occupancies outside [0, 1] is a bad fit wearing a good number.
- Ignoring preferred orientation, plate- and rod-shaped particles bias intensities, and uncorrected texture flows straight into wrong phase fractions.
- Quantifying a phase the pattern barely shows: trace phases near the detection limit need longer counting, not braver refinement.
- Refining against data collected too fast. Counting statistics set the information content; no model fixes a two-minute scan.
The XRD pipeline in depth; full Rietveld, scoped honestly
Niobia runs the X-ray diffraction workflow end to end: baseline correction, peak detection, d-spacing calculation through Bragg’s law, phase identification against reference patterns, crystallite size by Scherrer with instrument-broadening correction, Williamson-Hall strain analysis, preferred-orientation detection, and, in the refinement direction, lattice-parameter refinement and multi-phase fraction estimation from the pattern. Full whole-pattern Rietveld with complete structural models is partially supported today: Niobia handles the surrounding workflow and the peak-indexed refinements, and where a question demands a full structural refinement beyond that scope, it says so explicitly rather than dressing a partial fit as one.
Frequently asked
When do I need Rietveld instead of simple peak fitting?
When peaks overlap (almost always in multi-phase battery materials), when you need quantitative phase fractions, or when the question is structural, occupancies, cation mixing, precise lattice parameters. Single-peak methods remain fine for quick phase ID and Scherrer size estimates.
What is an acceptable R_wp?
There is no universal threshold, R_wp depends on counting statistics and background. The honest checks are a flat difference curve, GoF near 1, and physically sensible parameters. A flat residual at R_wp of 8% beats a structured residual at 5%.
Why does cation mixing in layered cathodes show up in a refinement?
Ni²⁺ sitting on lithium sites changes the scattering on those sites, which shifts characteristic intensity relationships like (003)/(104) in the pattern. Refined site occupancies quantify the disorder, one of the few non-destructive ways to catch it at the powder stage.