Whole-pattern analysis of powder X-ray diffraction data, in a browser. Launch it — there is nothing to install, and your data never leaves your machine.
powder5 decomposes a powder pattern against a unit cell and a space group, without needing a structural model. From there it can go on to solve one: the extracted intensities feed a charge-flipping map or a direct-space Wyckoff search, both of which run in the browser too.
Pattern decomposition — refine the cell, zero-point, peak profile and background, and extract integrated intensities, by the Le Bail or Pawley method. Three profile functions: simple pseudo-Voigt, the Thompson–Cox–Hastings model with Stephens anisotropic broadening, and a split pseudo-Voigt for asymmetric peaks. Refinement runs on a background thread, by Levenberg–Marquardt or by parallel tempering with an LM polish.
Space-group determination — test the settings compatible with the observed extinctions against the Pawley intensities, and rank them.
Structure solution — charge flipping (WebGPU where available, CPU otherwise), or a Wyckoff-position swarm search that fits atoms directly to the Pawley intensities. Either produces coordinates, a structure plot, contacts and bond-valence sums, and a CIF.
Reporting — statistics, refined parameters with standard uncertainties, reflection tables, and export to text, CIF or PDF.
- Load data. Drop in a file, or use Open. Bruker (
.brml,.uxd,.raw), PANalytical (.xrdml), Rigaku (.ras,.rasx), Philips (.udf), GSAS (.gsa,.esd), CIF, and plain two-column ASCII are all read directly. Where the file carries the wavelength, it is filled in for you — check it. - Set the cell and space group on the Sample tab, and trim the 2θ range with the sliders.
- Place background anchors on the Background tab, or Ctrl-click the plot. The background is a spline through those anchors; it is interpolated, not refined.
- Choose a profile on the Profile tab. Start with the simple pseudo-Voigt.
- Refine. Le Bail first, to get the cell and profile settled — it is faster and far more forgiving of a poor starting point. Then Pawley, when you want individual intensities and their uncertainties.
- Read the difference curve, not just Rwp. Structure in the residual means the model is wrong somewhere, whatever the number says.
- Solve, if that is where you are going: Charge Flipping or Wyckoff, both of which need a converged Pawley fit.
A single status bar at the foot of the results panel reports whatever is running, from any tab. One report button, top right between the theme toggle and the help icon, exports whichever tab is in front as a PDF — including the plot before you have loaded any data, which gives you the theoretical reflection list.
The help file is the real documentation: the methods, what every control does, what the numbers mean, and where each one stops being trustworthy. It opens from the ? button in the application.
Two things worth knowing before you start:
- Peak overlap sets the limit, not the algorithm. Where two reflections are exactly coincident the data determines only their sum. powder5 says so explicitly rather than printing a split it cannot support — see Exactly Coincident Reflections in the report and the Overlap clusters section of the help.
- A structure solution from powder data is a hypothesis. Judge it on the contacts, the bond-valence sums and the chemistry, not on the correlation coefficient.
Any current browser. WebGPU accelerates charge flipping and the Wyckoff search where it is available (Chrome and Edge today); both fall back to the CPU otherwise. Everything runs locally — no upload, no account, no network needed after the page has loaded.
Developed by Nita Dragoe, Université Paris-Saclay. It is the successor to PowderV2 — Dragoe, N. (2001), J. Appl. Cryst. 34, 535.
Methods implemented here are due to their authors; the help file lists the references in full. The principal ones are Le Bail et al. (1988), Pawley (1981), Thompson, Cox & Hastings (1987), Stephens (1999), Oszlányi & Sütő (2004) for charge flipping, and Larson & Von Dreele (2004) for the GSAS profile parameterisation.
Disclaimer. Provided for research and teaching. The algorithms are standard and the implementation is tested, but this is not a substitute for GSAS-II, FullProf or TOPAS for work intended for publication. Check anything that matters against one of them.
Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International
