Last week I ask about the goodness of the fit of XRD with Fullprof. I received than many suggestions about the refinement process.
Unfortunately, I still become a chi sqare smaller than one (what should non be possible and means that the standard deviation od the data is overestimated).
Example:
I take XRD for Sr3Cr2O8. I choose for the Pattern calculatio the Pseudo-Voigt function. The reflections list are automatically generated from the space group R-3m. After adding the atomic coordinates and occupancies, as well as the background (24 points) I started the refinement:
I do not use FullProf, so can't add very much specifically, but I would like to make a couple of observations.
Your Rexp value at 31 is very high. I would expect this to be an order of magnitude lower. This may be due to poor initial data. It is difficult to see, as you have blobs for data points, but the background looks very noisy, and you only have 2000 counts at max intensity, so longer counts required. Again, I can't quite see it, but you may want to decrease step size slightly to define the peak shapes a little better.
Are you able to post your original raw data so that we may have a look at it? Details of instrumentation will also help.
Trying to refine temperature factors on powder data is fraught with problems, especially with poor data, so I am not surprised that you have impossible results. Also a starting value of zero is going to cause problems. The values for Sr and Cr should be small, while O will be larger. - What sort of values did you get? There are a number of discussions on here about refinements and temperature factors, go and search for them. I would not try and do it at all unless I had exceptionally good data up to very high angles.
In the Rietveld refinement the correlation factor among refining parameters is often large, affecting after all chi square. To reduce the effects of this commonly observed situations , to my experience , is convenient to attribute very small values to the atomic thermal factors (in some case found negative), moreover PseudoVoigt factor should be controlled to be reliable , finally the refinement once inserted atomic positions should be re-started step by step in order to allow profile parameters to settle the value proper of the structure. All in all it should be reminded that the minimum of disagreement factor as a function of each parameter is always flat , i.e. hugely large variability without diverging . Best wishes
Dear Michele,
Please also take into account that the pattern you use could be corrected by the software (e.g. for polarization, average by some number of pixels, etc), i.e. do the point-by-point intensities of your pattern correspond to real counts obeying the Poisson count statistics or not.
Best regards
I do not use FullProf, so can't add very much specifically, but I would like to make a couple of observations.
Your Rexp value at 31 is very high. I would expect this to be an order of magnitude lower. This may be due to poor initial data. It is difficult to see, as you have blobs for data points, but the background looks very noisy, and you only have 2000 counts at max intensity, so longer counts required. Again, I can't quite see it, but you may want to decrease step size slightly to define the peak shapes a little better.
Are you able to post your original raw data so that we may have a look at it? Details of instrumentation will also help.
Trying to refine temperature factors on powder data is fraught with problems, especially with poor data, so I am not surprised that you have impossible results. Also a starting value of zero is going to cause problems. The values for Sr and Cr should be small, while O will be larger. - What sort of values did you get? There are a number of discussions on here about refinements and temperature factors, go and search for them. I would not try and do it at all unless I had exceptionally good data up to very high angles.
- Badges
- Science method
- Similar topics
- Methodology
- Crystallography
- X-ray Diffraction