Errors in Refined Parameters

[04anuvra...@gmail.com]

Hello all, 

I am refining PDF data of crystalline sample. I had few questions:

(a) Can we refine PDF data for small r values (1.7<r<6)?

(b) The refined values have large errors like one of the parameter (angle) is 89.6834 (error: 0.72). Is this right? The errors are coming out to be too large for some of the parameters.

(c) Can we reduce such large errors? If any possible way please help me.

Regards,

Anuvrat

[04anuvra...@gmail.com]

If anyone can help. Please do. 

Regards,

Anuvrat

[Mikkel Juelsholt]

Hi Anuvrat

1. You can, but you have to ask why do that? If you have crystalline material you should refine over a longer r-range. From 1.7 to 6 Å you usually have so few peaks that any model will give a good fit. I would always do at least 1-10 if I am trying to refine the local structure of a crystalline material.

2.  I would not agree that 89.6834 (error: 0.72) is a large error. However generally errors in PDF fitting is tricky. You have to be very careful with error propogaton from your initial integration of the raw scattering data, then following corrections and the Fourier transform. I do not know how you made your PDFs, but in general your errors are likely not that trustworthy anyway. Rather make sure your fit is visually good, the structure is physically reasonable and redo the fit multiple times to see how much the different parameters change.

3. The only real way to reduce experimental errors is to get better data. However as mentioned in 2 your error likely stem from wrong error propagation.

Cheers Mikkel

[04anuvra...@gmail.com]

Thank you sir, for your help. 

Thanks and regards,

Anuvrat