ZrW2O8 Constant Wavelength Neutron Data

Aim/outcomes of tutorial: this tutorial will show
you how to perform a structure independent (Le Bail) fit of a data set
to determine optimal unit cell and peak shape parameters for a neutron
diffraction pattern of ZrW2O8.
Le Bail/Pawley fits are often a good test of e.g. data quality/unit cell
correctness/the presence of impurity phases. By comparing a structure-independent
fit with a final structural model you can also get a good idea of the
model quality - your final model should approach the quality of the Le
Bail fit.

The cell parameter and peak shape will then be fixed at the Le Bail values
for initial structural refinements of a guessed partial model. This approach
can be very useful for unknown/difficult structures. Fourier maps will
be used to find the missing O atom.

The same data are used in one of the topas based tutorials - compare your answers!

Files needed: zrwneut.gsas and neutron.inst.

Instructions

1. Save the.gsas data file in your area and import the data into gsas
using expgui.

2. The space group is P213 and cell parameter ~9.16 Å. You will need to enter a "dummy" atom to perform a Le Bail refinement (e.g. enter Zr at 0, 0, 0). The neutron diffractometer had a significant zero point error of 0.97 . Enter this as 97 in gsas (which uses centidegrees). Use a Chebychev background (3 terms). Refine initially from 10-70 degrees only. In the "Scale" tab, turn off the refinement of the scale factor. In the "LS Control" tab, click on "Equally Weighted" to perform a structure independent fit to the data. Use a Le Bail damping of 5. Start with a Marquadt damping factor of 5. Use a peak cut off of 0.01.

3. After a few cycles of refinement refine the cell parameter and zero point. Save the .exp file here in case all goes wrong! Refine the profile parameters. You should get values of 0/ 32/87/2.6/1.7/14.4 for GU, GV, GW, LX, LY, asym. Chi**2 ~ 0.85.

4. Fix the cell, zero and profile. Fit the whole data range. Then re-refine cell/zero/profile. You should achieve Chi**2 around 1.1.

5. Save the .exp as a new filename. Fix the cell, profile, and zero. Run a few cycles of genles then fix background.

6. Enter the atomic coordinates of ZrW2O8 below. These were guessed from
related structures. Turn the scale factor back on, switch to Rietveld
refinement and refine the structure. You should be able to get Chi**2
~ 5.2, wRp=24%.

Zr1 0.00 0.00 0.00 O1 0.20 0.44 0.45

W1 0.33 0.33 0.33 O2 0.80 0.57 0.55

W2 0.60 0.60 0.60 O3 0.50 0.50 0.50

7. You can find the final O atom by calculating a Fourier difference map. This is best done using pc-gsas. Quit expgui. Find your .exp file and launch it in pc-gsas by right clicking on the icon. In pc-gsas run Setup/expedt and answer "Y" to the first question.

8. Enter the Fourier menu by typing "F". Compute a DELF map. Section X. "N" to individual map steps. Use a step of 0.2. Calculate fractional coordinates 0 to 1 in each of x, y and z directions. Include histogram 1. Use x, x, x to exit expedt.

9. Go to Compute/fourier to calculate the map. Go to Compute/forsrh to search the fourier map for peaks. Search for the top 10 positive peaks. Don't save them in the .exp file. Note down the strongest peak.

10. Return to expgui and try entering the peak from the Fourier map as an extra oxygen.

11. Refine the fractional coordinates/temperature factors of all atoms.

12. You could now try re-refining the profile parameters and cell constants.

13. If the refinement doesn't converge use lstview to see which parameters are poorly behaved.

14. Try going anisotropic on all atoms (click on the Xform Atoms box to do this).

Outcome

Refining 44 variables in total you should be able to get Chi**2 around
0.96, wRp=10.2%, R(F**2)=5.33%.

Additional Work

You can use the forplot routines in expgui to view the difference Fourier
map.

Export a .cif file and use Mercury to view the structure.

Compare your refinement results with those obtained in topas.

Use atoms to view the structure/adps/polyhedra.

[Modified 31-Jul-2017 by John S.O. Evans. Pages checked for Google Chrome.]