In the previous section, we described how to perform grain indexing with GrainSpotter. After some time, you might realize that you are not satisfied with the results and need more advanced tuning of the process.

Below is an example on how to proceed to improve your indexing. It includes several steps

1. Check your expected crystal structures, the corresponding peaks, and your g-vectors input,
2. Run loops of GrainSpotter indexings,
3. Optimize results for your fitted grains with FitAllB.

As this point, you probably computed your list of G-vectors with ImageD11. The first line of the G-vectors file contains the unit cell parameters and lattice centering (one of P,A,B,C,I,F) and, below, you can find list of computed theoretical g-vectors for the expected unit cell.

This list is generated with crude calculations that do not know anything about atom types of positions in your structure. Chances are that many of those theoretical g-vectors do not exist for your structure or have a very low intensity.

At this point, you can use other softwares, such as VESTA to generate a list of the expected peaks as well as their relative intensities. You can then check your computed theoretical g-vectors in the G-vectors file and remove peaks that do not exist or with a low intensity.

You probably noticed that, after some time, GrainSpotter does not find any new grains. At this point, you should

• use the TIMEleSS tool to remove the G-vectors which have been assigned,
• re-run GrainSpotter with the same of less restrictive tolerances,
• merge the GrainSpotter output files into one.

This process can be combined into advanced indexing loops. Attached is an example of custom loop for olivine (restricted access to team members) that will run a series of 25 GrainSpotter indexing with various tolerances.

At the end, you will want to analyze and understand the reliability of your final indexings. Plotting grain orientations and comparing the indexings is a good start.

Next step…