Why there is a sharp peak in low r when I refine a structure using fullRMC?

[Leon Wang]

I have XRD measurements for nano materials. The structures are refined by fullRMC. The total agreement is pretty good (see Fig 1 and 2 which are errors for sample #1 and #2, respectively ). However, I noticed that there are sharp peaks at low r (~2 Å). In the figure of sample #1, the sharp peak is contributed by Al-O, while the sharp one is contributed by Ti-O in the figure of sample #2. I adjusted the related cutoffs to larger or smaller, however, the peak cannot be eliminated.

I do not understand which factor(s) results such errors. Any suggestion and help would be highly appreciated.

[Bachir Aoun]

Hello Leon,

Normally I refrain from giving pure scientific advises because it takes more than a plot to answer reasonably. 

Nevertheless, from what i see here your fit is not good even at high R values. fullrmc's stochastic engine knows how to torture the atomic system to get something reasonable out but be careful this is not always good.

your system density is obviously way off and nothing will compensate for that unless you fix the issue. It can be even your simulation atomic system or your experimental data.

Always remember that experimental data preparation is a bit empirical and many things can be off. In this particular case I think (I leave it up to you to judge and reverify) it's rescaled and damped too harshly.

in fullrmc you can account for that by letting the engine fit the rescale factor. 

constraint.set_adjust_scale_factor(...)

still i recommend you verifying your experimental data preparation 

Hope that helps

[Leon Wang]

Hi, Bachir,

My apologies for the non-related questions.I did some literature researches. As lack of experience and knowledge of XRD refinement and fullRMC, I have not figured out where to start to fix those issues. Please forgive my ignorance of fullRMC. I hope that reading the original code of fullRMC could help me to understand more and figure out issues in the future.

 

I have following questions.

1. How do you tell my fit not good even at high R values?

2. "your system density is obviously way off and nothing will compensate for that unless you fix the issue". Would you please instruct me how to tell the system density is way off?

3. Originally, I have set "PDF_CONSTRAINT.set_adjust_scale_factor((10,0.7,1.3))". Owing to the difference (errors) between experimental and simulating data, my colleague suggested me comment it. The difference (I mean the experimental and simulating curves are almost same) became much small. How you tell that I did not count in the engine "PDF_CONSTRAINT.set_adjust_scale_factor((10,0.7,1.3))" from these two figures?

Any further suggestion would be highly appreciated.

Appreciatively,

Leon

 

I highly appreciate it if you could conduct me to read some

[Bachir Aoun]

Leon,
I you zoom in you will see that your model curve is always above the experimental data especially at high r
Regarding your density, the low r experimental data slope doesn't match your model ...