Some questions from TOPAS ACADEMIC examples

john201004

Hi

When I am going through these example inps provided in Topas Academic (TOPAS_v4.1_Menus -> Examples -> Rietveld refinement) I found several different points difficult to understand. Such as in first li025.inp which demonstrates the use of "Absorption_With_Sample_Thickness_mm_Shape_Intensity", I got this result below which have enormous errors at most of the parameters. To get rid of these errors I had fixed occ of all the O-2 atoms to one and then it gave me reasonable errors on most of the parameters other than CS_L. So I had comment out the CS_L and then I got quite OK fit with acceptable errors. In the end, to check out whether I am at true minimum with parameters which close to the real value. I try to refine the isotropic temperature factor of all O as well. After refinement I got beq values range from 3 to 19 for these oxygen atoms in that specific phase . So I had obliged to restrain beq between 0-1, but after that beq didn't get good value instead I just get something very close to 0 or 1 with limit warnings. Because of that I had changed beq values to original values 0.25 and I fixed them. Then everything looked quite OK.

So my question is:
1. why in first place it gives such huge errors? What might be the possible cause?
2. After all parameters refined to something closer to real values, the beq should be between 0-1 but in this case I still trouble to get OK beq why is that? what will be the good strategy to improving the beq values? (BTW I refined beq in last step of the refinement and in my first try I fixed all other parameters other than cell parameters, in my second try I had refined almost everything ).
3. In other example files, when I looked at beq values of the the given atoms I found that some of the atoms have negative values and some of them have around 2. Why is that? How come it gets negative values and exceeds one? or these inp files are just examples which demonstrates how to use certain macros and commands which not finished yet or there is some reason for that?
To show these examples I had listed inp file title and the parts of the refinement in below.

4. In many examples, I found some calculated errors which are quite close to the refined values or in certain cases they are even bigger. Does it acceptable depending on what parameters we are talking about or for all we must have small errors? Is there any rules of thumb or guide lines for that?

li025.inp

STR(P_N_N_A) ' ICSD Collection Code 71946, Li (B2 O3 (O H)) (H2 O)
a @ 9.81374`_0.00102
b @ 8.28762`_0.00095
c @ 9.61229`_0.00087
site O x @ 0.84669`_0.00146 y @ 0.34700`_0.00165 z @ 0.14802`_0.00101 occ O-2 @ 1.3720`_365293.8226 beq 0.25
site O x @ 0.67015`_0.00141 y @ 0.12517`_0.00180 z @ 0.19524`_0.00122 occ O-2 @ 1.2271`_326722.3039 beq 0.25
site O x @ 0.85862`_0.00130 y @ 0.07572`_0.00161 z @ 0.03584`_0.00108 occ O-2 @ 1.5143`_403182.0622 beq 0.25
site O x @ 0.99152`_0.00141 y @ 0.32315`_0.00124 z @ 0.94401`_0.00101 occ O-2 @ 1.5983`_425546.4797 beq 0.25
site O x @ 0.11049`_0.00101 y @ 0.06481`_0.00149 z @ 0.24479`_0.00095 occ O-2 @ 1.3696`_364658.2325 beq 0.25
site B x 0.75 y 0. z 0.112 occ B @ 0.6629`_176510.1157 beq 0.25
site B x 0.903 y 0.229 z 0.035 occ B @ 1.3959`_371663.1750 beq 0.25
site B x 0.762 y 0.25 z 0.25 occ B @ 1.4559`_387623.7151 beq 0.25
site Li x 0.546 y 0.020 z 0.329 occ Li+1 @ 0.0054`_1441.9157_LIMIT_MIN_0 beq 0.25
scale @ 0.00162638624`_866
CS_L(@, 10000.00000`_33547.09850_LIMIT_MIN_0.3)
Strain_L(@, 0.19033`_0.01790)
MVW( 1119.045`_114525130.806, 781.792486`_0.139834038, 32.738`_11725787.657)

alvo4a.inp

site Al1 x @ 0.76166`_0.00078 y @ 0.69622`_0.00066 z @ 0.41573`_0.00059 occ Al+3 1 beq @ 1.3355`_0.2015
site Al2 x @ 0.46529`_0.00070 y @ 0.89253`_0.00054 z @ 0.21478`_0.00054 occ Al+3 1 beq @ 0.4475`_0.1672
site Al3 x @ 0.96647`_0.00077 y @ 0.30489`_0.00060 z @ 0.00779`_0.00059 occ Al+3 1 beq @ 0.6569`_0.1708
site V1 x @ 0.99808`_0.00059 y @ 0.00476`_0.00046 z @ 0.74552`_0.00040 occ V+5 1 beq @ 0.5915`_0.1151
site V2 x @ 0.19156`_0.00061 y @ 0.59911`_0.00045 z @ 0.34224`_0.00037 occ V+5 1 beq @ 0.8351`_0.1213
site V3 x @ 0.51551`_0.00059 y @ 0.29947`_0.00048 z @ 0.12832`_0.00041 occ V+5 1 beq @ 0.7379`_0.1084

site O1 x @ 0.64868`_0.00122 y @ 0.49358`_0.00096 z @ 0.26201`_0.00086 occ O-2 1 beq @ 1.6555`_0.3259
site O2 x @ 0.25731`_0.00118 y @ 0.43037`_0.00091 z @ 0.43384`_0.00071 occ O-2 1 beq @ 1.9556`_0.3529
site O3 x @ 0.04709`_0.00095 y @ 0.70195`_0.00095 z @ 0.43393`_0.00087 occ O-2 1 beq @ 1.6516`_0.3220
site O4 x @ 0.16964`_0.00132 y @ 0.09303`_0.00097 z @ 0.43606`_0.00086 occ O-2 1 beq @ 2.3165`_0.3648
site O5 x @ 0.46247`_0.00104 y @ 0.75081`_0.00095 z @ 0.35792`_0.00082 occ O-2 1 beq @ 1.4700`_0.3200
site O6 x @ 0.73783`_0.00086 y @ 0.86435`_0.00095 z @ 0.26409`_0.00080 occ O-2 1 beq @ 0.5560`_0.3211
site O7 x @ 0.51586`_0.00118 y @ 0.11668`_0.00063 z @ 0.21864`_0.00084 occ O-2 1 beq @ -0.0192`_0.2961
site O8 x @ 0.16863`_0.00088 y @ 0.86989`_0.00087 z @ 0.17100`_0.00073 occ O-2 1 beq @ 0.5146`_0.3105
site O9 x @ 0.35257`_0.00095 y @ 0.72678`_0.00088 z @ 0.03241`_0.00070 occ O-2 1 beq @ -1.3586`_0.2719
site O10 x @ 0.24841`_0.00095 y @ 0.29469`_0.00108 z @ 0.02910`_0.00091 occ O-2 1 beq @ 4.1780`_0.3726
site O11 x @ 0.94930`_0.00122 y @ 0.14905`_0.00095 z @ 0.13820`_0.00075 occ O-2 1 beq @ 2.7895`_0.3698
site O12 x @ 0.04424`_0.00111 y @ 0.51928`_0.00089 z @ 0.13686`_0.00054 occ O-2 1 beq @ -0.3250`_0.3089

quartz.inp
site O x @ 0.42791`_0.00073 y @ 0.28033`_0.00066 z @ 0.13438`_0.00046 occ O-2 1 beq @ 2.3071`_0.1105

pbso4a.inp
site Pb x @ 0.16716`_0.00011 y @ 0.18774`_0.00008 z 0.25 occ Pb+2 1 beq @ 1.4423`_0.0169
site S x @ 0.18448`_0.00059 y @ 0.43541`_0.00044 z 0.75 occ S 1 beq @ 0.8425`_0.0778
site O1 x @ 0.09525`_0.00109 y @ 0.59665`_0.00093 z 0.75 occ O-2 1 beq @ 0.9120`_0.2059
site O2 x @ 0.03728`_0.00139 y @ 0.31187`_0.00110 z 0.75 occ O-2 1 beq @ 1.5513`_0.2122
site O3 x @ 0.31532`_0.00096 y @ 0.42020`_0.00066 z @ 0.97554`_0.00092 occ O-2 1 beq @ 1.5191`_0.1470

brucite.inp
site Mg x 0 y 0 z 0 occ Mg 1 beq @ 1.2761`_0.0323

cpd-4.inp
site Fe x = 1/8; y = 1/8; z = 1/8; occ Fe+2 1 beq @ -0.1735`_0.1504
site Fe x 0.5 y 0.5 z 0.5 occ Fe+3 1 beq @ -0.0239`_0.1210
site O x 0.2553 y 0.2553 z 0.2553 occ O-2 1 beq @ -0.1420`_0.2812

zhu3a.inp

site CO x @ 0.81879`_0.00042 y 0.25000 z @ 0.43987`_0.00034 occ Co+2 1 beq @ 1.4199`_0.0593
site N1 x @ 0.25460`_0.00176 y @ 0.86960`_0.00162 z @ 0.79847`_0.00182 occ N 1 beq @ 1.3978`_0.3797
site N2 x @ 0.26205`_0.00203 y @ 0.89154`_0.00154 z @ 0.37880`_0.00216 occ N 1 beq @ 3.5365`_0.4515
site N3 x @ 0.90028`_0.00129 y @ 0.11483`_0.00104 z @ 0.28150`_0.00186 occ N 1 beq @ -1.0954`_0.2789
site N4 x @ 0.43100`_0.00151 y @ 0.66459`_0.00116 z @ 0.63405`_0.00151 occ N 1 beq @ -0.4824`_0.2859
site N5 x @ 0.89106`_0.00135 y @ 0.12453`_0.00116 z @ 0.66549`_0.00161 occ N 1 beq @ -0.9126`_0.2558
site O1 x @ 0.07224`_0.00216 y @ 0.36101`_0.00132 z @ 0.54562`_0.00204 occ O-2 1 beq @ 5.7334`_0.4804
site C x @ 0.19093`_0.00299 y @ 0.28323`_0.00233 z @ 0.56701`_0.00315 occ C 1 beq @ 3.2029`_0.5096
site O2 x @ 0.25713`_0.00140 y @ 0.15768`_0.00118 z @ 0.61701`_0.00133 occ O-2 1 beq @ 0.1328`_0.2597
site O3 x @ 0.34836`_0.00162 y @ 0.37592`_0.00108 z @ 0.53466`_0.00142 occ O-2 1 beq @ 2.4526`_0.3030
site N6 x @ 0.20294`_0.00186 y @ 0.29895`_0.00118 z @ 0.04826`_0.00196 occ N 1 beq @ 0.5458`_0.3401
site O x @ 0.34544`_0.00176 y @ 0.38347`_0.00131 z @ 0.01890`_0.00158 occ O-2 1 beq @ 5.5236`_0.4239
site O5 x @ 0.21286`_0.00197 y @ 0.16728`_0.00159 z @ 0.13774`_0.00183 occ O-2 1 beq @ 7.9792`_0.5150
site O6 x @ 0.06081`_0.00136 y @ 0.37125`_0.00103 z @ 0.02279`_0.00128 occ O-2 1 beq @ 1.3107`_0.2813
site OW x @ 0.58158`_0.00156 y @ 0.15157`_0.00133 z @ 0.88710`_0.00143 occ O-2 @ 0.8317`_0.0158 beq 1

mfisch

john201004 wrote 1. why in first place it gives such huge errors? What might be the possible cause?

You blew your refinement. Refining P-XRD data does not mean to refine everything...

john201004 wrote 2. After all parameters refined to something closer to real values, the beq should be between 0-1 but in this case I still trouble to get OK beq why is that? what will be the good strategy to improving the beq values? (BTW I refined beq in last step of the refinement and in my first try I fixed all other parameters other than cell parameters, in my second try I had refined almost everything ).

You blew your refinement. Refining P-XRD data does not mean to refine everything...

john201004 wrote 3. In other example files, when I looked at beq values of the the given atoms I found that some of the atoms have negative values and some of them have around 2. Why is that? How come it gets negative values and exceeds one? or these inp files are just examples which demonstrates how to use certain macros and commands which not finished yet or there is some reason for that?
To show these examples I had listed inp file title and the parts of the refinement in below.

There is no reason why B should not exceed one. However, negative B's are physically not meaningful.

john201004 wrote 4. In many examples, I found some calculated errors which are quite close to the refined values or in certain cases they are even bigger. Does it acceptable depending on what parameters we are talking about or for all we must have small errors? Is there any rules of thumb or guide lines for that?

Rule of thumb: If value < 3sigma, then it is not meaningful.

Please consider reading a book, as I have the impression that you are not really aware of what you are doing (see your questions 1, 2 and 3).

john201004

You blew your refinement. Refining P-XRD data does not mean to refine everything...

I did not refine everything.... I only run these inp files provided in TOPAS. Only thing I added was "do_errors". Check these inp files before comment.

There is no reason why B should not exceed one.

Depending on how well ordered is the structure, types of atom in question and depending on temperature the mean displacement is quite very. Higher B means higher displacement. In my observation, the <u> normally takes a value around 0.1 Å or less and it gives a B value between 0-1.

However, negative B's are physically not meaningful.

I know negative B's are physically meaningless, that is why I ask the question whether these inp is only for demonstration purpose or they are correct and published results.

Rule of thumb: If value < 3sigma, then it is not meaningful

Thanks for the tip, but why exactly 3sigma not 2sigma or something else? How did you derive that 3?

Please consider reading a book, as I have the impression that you are not really aware of what you are doing

I am not considering it, actually I am doing it.

(see your questions 1, 2 and 3).

Yes, please look these questions again. Looks like you have very limited patience. Try to finish reading whole question before answer the first part.

mfisch

Dear John201004,

I assume that the examples are just for what's written in the topmost comment line: to show how a specific macro is used. I have looked at some of them, so just some personal comments:

- I would not recommend to refine occupancy, B's together with a spherical harmonics PO correction. It all correlates with intensities.

- It is most likely not possible to get reasonnable B's from data with few high- and many low-intensity counts (quartz, brucite).

- B = 8 Pi^2 * U

- You find more on the 3-sigma rule on google.

john201004

Thank you mfisch, I am appreciated.

Just comment about formula of B. it should be:
- B = 8 Pi^2 * <u^2>.