The first rule of the Rietveld refinement: never trust the results of the Rietveld refinement. The modern software can be very user-friendly, it will allow you to refine everything you want, but it knows nothing about chemistry or physics. It only compares two sets of data and tries to make them as similar as possible. If you see something strange in the results, it is your responsibility to check all possible sources of errors. Some typical problems should be checked in your case before going any further:

1. Is the quality of your data enough to refine the occupancies? If Chi**2 is too low or its value is not affected by the change of the occupancies, you cannot refine them.

2. Is the difference between the calculated occupancies and the expected ones significant? What is the value of the standard deviation given by the program? Obviously, you can never obtain 6 significant digits of the occupancy.

3. Do you have any kind of strong correlations between the refined parameters? Occupancies, thermal parameters (B or U) and sample roughness usually correlate strongly and an incorrect shift of one parameter can be mathematically compensated by an incorrect shift of another parameter to the opposite direction. The program will show you something pretty, but absolutely meaningless.

The first rule of the Rietveld refinement: never trust the results of the Rietveld refinement. The modern software can be very user-friendly, it will allow you to refine everything you want, but it knows nothing about chemistry or physics. It only compares two sets of data and tries to make them as similar as possible. If you see something strange in the results, it is your responsibility to check all possible sources of errors. Some typical problems should be checked in your case before going any further:

1. Is the quality of your data enough to refine the occupancies? If Chi**2 is too low or its value is not affected by the change of the occupancies, you cannot refine them.

2. Is the difference between the calculated occupancies and the expected ones significant? What is the value of the standard deviation given by the program? Obviously, you can never obtain 6 significant digits of the occupancy.

3. Do you have any kind of strong correlations between the refined parameters? Occupancies, thermal parameters (B or U) and sample roughness usually correlate strongly and an incorrect shift of one parameter can be mathematically compensated by an incorrect shift of another parameter to the opposite direction. The program will show you something pretty, but absolutely meaningless.

Alex, unfortunately you are a lone voice in the wilderness... These questions/ expectations/ wishes pop up again and again. This very uncritical thinking is not only common nowadays in XRD but practically everywhere. Only who wrote software himself realized the potential of errors, typing mistakes, sometimes also mean bugs and is therefore very careful. Only because a software does not crash does not mean that it is free of erros. And, as you metioned, it is finally the believe of the user what he is thinking about the shown result. However, without any experience/knowledge in mathematic and understanding of the technique used and statistics (!)  this is practically impossible. From my opinion this complain is more directed to the supervisors than to the students since mainly the first decide to use these programs (without any fundamental knowledge and experience). The results are publications of questionable quality which we have to fight with during discussion with other people who think they are fantastic...

can anybody send the CIF for CaMoO4

I will stand up for Alex's points, I think they are all valid. Gert I do not see the relevance of the first part of your answer to either the OP or to Alex. The second part you are agreeing with him. I'm just a bit puzzled by this.

Nancy - please delete you comment as it is irrelevant on this thread and will not reach the correct audience. Either post it as a separate question or try doing some some research to find it for yourself.

(sorry people - feeling a little crabby today, I've just returned home from my 4th round of radiotherapy)

Gert,

I prefer to be more optimistic about that. At least some of those, who ask naive questions here, will learn from the answers. In fact, I do not have a systematic crystallographic background myself and had to pass the same way. Of course, it took much longer than it could and sometimes I have to return to the very basic points that I missed, but nothing is impossible if you really want.

Another point is that the same questions appear again and again. But this is very common for modern social networks: most people believe that they have a specific problem while in fact it is an individual case of something typical. This happens everywhere from crystallography to choosing of a better restaurant, and we can only decide for ourselves, to which group we will belong: to those who have enough patience to repeat the same answers or those who come into the topic to say "Just google it".

Well, I haven't seen anyone actually answer how to interpret the correct method for sites with multiple occupancies. If you don't know how to then don't respond at all. Trying to understand how you might understand the results. Even if my numbers make no sense nobody seems to know how to interpret vacancies with multiple atomic sites.