In general the sequence of parameters to be refined should match their importance regarding their contribution to the pattern. Scale and zero shift, while rather unimportant from a structural point of view nevertheless contribute the most for the rough description of the pattern in the y and x direction. Thus the strategy should aim for getting the rough pattern right before focusing on ever smaller details (and using ever more parameters to fit these).

In thinking about how refinement parameters affect each other it is helpful to consider the physical background of parameters. Reflection position on the x axis is due to lattice spacing in reciprocal space while reflection intensity on the y axis is (mainly/partly) due to atomic distribution in real space, thus changing a lattice parameter does not change reflection intensity and shifting an atom does not change reflection position, and both parameters cannot interfere with each other. A qualitative picture of which parameter is possibly interfering with which other parameter can therefore get by understanding their origin in terms of diffraction theory. As a second example consider any peak width parameter. These 'live' in the '1D x direction space' of the 2D diffraction pattern and therefore could correlate with any other parameter affecting the x direction.

The tool to analyze this quantitatively is the correlation matrix of the refinement parameters which usually can be obtained by any decent refinement program (possibly/depending on the program only after some search, unfortunately). Correlation of parameters mean that the effect of the change of one parameter could to some extent (as described by the amount of correlation) be modeled by the change of another parameter, thus rendering both parameters less informative.

The other question is, if this really helps you with regard to your chi square value? The reason why this deviates from the value you expect could be difficult to trace back to a single refinement parameter and the reason why it should change with a variation of the refinement strategy is not really obvious for me, except for getting stuck in a false minimum perhaps. In any case one should judge any refinement not only on the basis of the numerical values of single parameters but in all its glory (especially including the visual appearance of the fit, which could give a hint at the kind of any occurring problems). There are perfect valid reasons why sometimes one or the other parameter does not refine to an 'expected' value and one has to know the data very well and have some general experience to give a judgement if this is ok or an issue to think about.

For general guidelines on how to perform the refinement, I can recommend this article: http://scripts.iucr.org/cgi-bin/paper?gl0561

I would avoid smoothing the data, it might look a bit nicer afterwards but you also risk to add falsifying manipulations...

In general the sequence of parameters to be refined should match their importance regarding their contribution to the pattern. Scale and zero shift, while rather unimportant from a structural point of view nevertheless contribute the most for the rough description of the pattern in the y and x direction. Thus the strategy should aim for getting the rough pattern right before focusing on ever smaller details (and using ever more parameters to fit these).

In thinking about how refinement parameters affect each other it is helpful to consider the physical background of parameters. Reflection position on the x axis is due to lattice spacing in reciprocal space while reflection intensity on the y axis is (mainly/partly) due to atomic distribution in real space, thus changing a lattice parameter does not change reflection intensity and shifting an atom does not change reflection position, and both parameters cannot interfere with each other. A qualitative picture of which parameter is possibly interfering with which other parameter can therefore get by understanding their origin in terms of diffraction theory. As a second example consider any peak width parameter. These 'live' in the '1D x direction space' of the 2D diffraction pattern and therefore could correlate with any other parameter affecting the x direction.

The tool to analyze this quantitatively is the correlation matrix of the refinement parameters which usually can be obtained by any decent refinement program (possibly/depending on the program only after some search, unfortunately). Correlation of parameters mean that the effect of the change of one parameter could to some extent (as described by the amount of correlation) be modeled by the change of another parameter, thus rendering both parameters less informative.

The other question is, if this really helps you with regard to your chi square value? The reason why this deviates from the value you expect could be difficult to trace back to a single refinement parameter and the reason why it should change with a variation of the refinement strategy is not really obvious for me, except for getting stuck in a false minimum perhaps. In any case one should judge any refinement not only on the basis of the numerical values of single parameters but in all its glory (especially including the visual appearance of the fit, which could give a hint at the kind of any occurring problems). There are perfect valid reasons why sometimes one or the other parameter does not refine to an 'expected' value and one has to know the data very well and have some general experience to give a judgement if this is ok or an issue to think about.

Dear

Dr. Singh

I have just look at your XRD data. As Dr. Breternitz have said, your chi2 is below 1 because of low XRD intensity of your sample. In many positions there zero ! .Try to avoid any type of smoothing or background subtraction. Use your original data and start refinement. If this data was provided by operator the tell him to take XRD again.

The below paper described the sequence of Rietveld refinement of XRD data to describe the structure and crystallographic preferred orientation:

"Comparative evaluation of the March and generalized spherical harmonic preferred orientation models using X-ray diffraction data for molybdite and calcite powders" - February 2005, Journal of Applied Crystallography 38(1):158-167 Article Comparative evaluation of the March and generalized spherica...

DOI 10.1107/S0021889804031231

The paper can be also downloaded from the link at - Article Comparative evaluation of the March and generalized spherica...

try to avoid smoothning....usally occur in smoothed data

Dear Rajan Singh

Please refer Young's book.

You can use the following sequence

Scale

Profile-a,b,c

Instrumental

Background

U,V,W

Shape parameters

Profile-Asymmetry

Atomic positions-X,Y,Z etc

During my using the Rietveld Refinement i can tell you to do the following 1st step you click on scale, a, b,c , Instrumental error, and d1, in background and then do run u will see that the Ch2 value will decrease and once again do run and continue till the value of Ch2 become constant.

2nd step go to background click on d2 and run, d3 and run, you will see Ch2 is decreased.

3rd step fix the value of background and instrumental error and go to profile and select U and run, then V and run, then W and run. you will see that the V value become negative. click on fix values and then run .

last step go to Atoms and do refinement for atomic positions, Bio and Occ one by one and continue same the above steps till you get good fitting