Most easiest solution is to set appropriate starting parameters.

I would try to set constant background (0th in polynomial) to ~2000 and scale factor to ~100 and refine scale factor only to start.

If this doesn't work look at the out-file: are the structure factors calculated correctly?

You appear to have sent the calculated background (BG) to zero initially. Start with a calculated background that gives a reasonable fit to the BG in the measured pattern. Also, start with a phase scale factor that gives a reasonable fit to the measured pattern. And also, get the peak widths about right initially by trial and error.

Then refine/optimise these paremeters before 'playing' with other parameters.

I would next refine the 2-theta zero (unless you have determined this by measurement - which is highly preferable). And then the lattice parameters.

Remember that Rietveld modelling is a very fragile, albeit powerful, technique. It is easy to get nonsense results unless we are aware of the limitations of the technique. In particular, parameter correlations can cause a poor initial model to go right off track and so provide nonsensical answers. Use of non-linear least squares in other applications has this same limitation, e.g. in climate change modelling.

Hi, Shruti,

At the first sight of your attached image, one can easily say that there is mismatch between the observed peaks and modeled peaks. Modeled peaks is shifted towards lower angle side. So, to match the peaks you have to reduce lattice constants. You can use input data file in free format. It will be better if you perform first profile matching and then Rietveld refinement. It will be better to find out the mistakes which are going on if you attach pcr file.

Regards

Dinesh

Many thanks for all your responses!!!

Dear Matthias, I am already using a .bgr file generated using WinPlotR as background. But using a scale factor of the magnitude you suggested also did not help and resulted in the same result. I had a look at the structure factors reported in the .out file and calculated them manually for comparison. The values do not match up.

Dear Brian, I am using an independently generated .bgr file as an input for background model which fits nicely with the observed data. I also have already done several hit and trials (manually) for finding out a reasonable scale value. A value of about 4*10^-5 seems to be the closest to the observed pattern but it too, almost instantaneously, dies out in about 5-6 refinement cycles (once I vary the scale), leaving me with a situation like the one I've shown in the image. The scale value becomes 10^-10 here.

I completely agree with you over the sensitivity of the Rietveld code to initial parameters and I hence only vary the scale when I begin refining. But this is the first time I have encountered an error like this and I am unable to pinpoint the source.

Dear Dinesh, I have attempted profile matching (LeBail Method) too. Since FullProf doesn't allow scale variation with LeBail, the intensities do sustain, but the zero and displacement keep fluctuating about the initial values on refining. And the shape parameters give unphysical and values on refinement trials. As per your suggestion, I will now try to regenerate the dat file in free format and get back to you.

Please share data and pcr file - this will allow us to answer your question.

I thank you all for your valuable suggestions! My problem was fixed by converting my dat file into free format. I still don't understand why EdPCR was refusing to accept the XY Format of my dat file, but the refinement works now!