Modelling hypothetical XRD patterns. Is this even a thing?

May 6, 2012 5:05 AM

I'm looking for (preferably free) software that will model X-ray diffractograms for mineral mixes of known quantities. Does such a thing even exist?

I have software for displaying and analysing collected patterns, so that's not my problem. I also know about various websites that can provide diffractograms for individual minerals, but I need to create a model of what a diffraction pattern would look like for a mixture of minerals in known quantities.

I've done this kind of thing specifically for clay minerals with NEWMOD, but I'm looking for something that can handle other common minerals. My internet searching has turned up nothing useful. Any suggestions for specific software packages, online resources, or even workarounds for typical diffractogram display/analysis software? I have PowderX - maybe there's someway to do it with this that I've not been able to figure out?

posted by Eumachia L F to Science & Nature (3 answers total)

I've had to do this before manually, so that's all I can offer. You analyze a series of known mixtures (I was doing calcite/aragonite) and did it with 10% differences (10/90, 20/80, etc.). It did not allow for an exact percentage match, but I was able to narrow it down significantly. Building a model to cover the stretches between the known points might work for you, but I don't know what degree of accuracy you are going for.

I hope that helps!
posted by Gneisskate at 6:33 AM on May 6, 2012

If it's a mix of grains of different minerals (i.e. no intermixing between the two phases), then the diffraction pattern is just the sum of both patterns to first order. (i.e. kinematic diffraction, and ignoring dynamical effects from multiple-order diffraction). You should be able to just superimpose the two patterns. If you have mixed phases, things get more complex, as different grains might have different compositions. If the phases have the same structure but different lattice constants, a good first order approximation (which will give correct shapes but incorrect intensities) would be to superimpose the patterns and then connect corresponding points. If the structures and lattice constants are different, all bets are off.
posted by JMOZ at 12:54 PM on May 6, 2012

Response by poster: I must admit, I'm working with a nightmare scenario. What I have is a synthetic physical mixture of natural materials, which are themselves quite heterogeneous. The sample has been separated as much as possible, but it's still quite mixed. Think concrete with a bit of microscopic aggregate polluting the cement binder. I sent the samples for analysis to an outside lab who gave me some preliminary abundances from Reitveld transformations. The data they gave me account for 90% of the peaks in each pattern. Naturally, I'm actually more interested in the remaining 10% and I have to ID those myself. I can't begin to describe how ugly these patterns are, with multiple reflections and overlapping and amorphous silicate-driven peak broadening and whatnot. I was hoping to calculate an approximation (no matter how approximate) of what the diffractogram would look like for just the 90% to compare to the raw data so I can get an idea of how last 10% impacts things. NEWMOD did exactly this for clays, and I was hoping there was a similar option for what I'm doing. I know it's pure laziness and no geologist worth her lab coat would admit this kind of defeat, but I don't really want to spend the rest of my life brute forcing my way through this for Each. Sample. Looks like that is exactly my future. Thanks for taking the time read my question.
posted by Eumachia L F at 4:41 PM on May 7, 2012