How do I store the provenance of a material?

[scott-materials]

Let's say I'm using structures from OQMD as a jumping off point for advanced calculations. I'd like the OQMD ID to be stored in the results table, along with the specific data that resulted from the calculation. From my investigations, it seems like the "source" parameter is the proper way to store that information. This is what I've tried:

result = workflow.run_cloud( 
        structure=structure,
        command='mpirun vasp_std > vasp.out',
        tags="bg",
        source={
            "database_table": "OqmdStructure",
            "database_id": f"{test.id}"
            }

But my source info is being overwritten by the default source info.

Any suggestions on how to best link a calculation to the original materials ID?

Thanks!

[jacksund]

short answer

you're doing it correctly, but provenance will be difficult to interact with until #175 is closed

long answer

The provenance/history/lineage of calculations is something that's still at the experimental stages (as of v0.13.2). But I have outlined a strategy for how the source column should be used in #175. A guide will be added once I'm able to close that issue.

BUT you are using the source parameter correctly! If you set source on a single-step workflow (such as a static-energy calc), then you will see source show up in the database exactly as you set it. However, if you use a multi-stage workflow like electronic-structure.vasp.matproj-full, then this is probably where you're asking "why is it overwriting my source"? It's not!

Using a BS+DOS calc as an example, you can see that source is used in the static-energy calculation. Then the result of the static energy is used in the BS/DOS. So.... (i) source of static energy is whatever you set it as and (ii) source of the BS and DOS is the static energy calc. Even though BS/DOS aren't showing the source you expected, you can still track back through it's "lineage" to what you're looking for (e.g. BS --> static energy --> oqmd entry). I'm guessing you have the BS and wanted the OQMD entry listed as the source (skipping over the static energy). Your database already holds this info, but it's inconvenient to pull out.

Once I close #175, there won't be a source column, but instead source_direct, source_seed, and source_parent columns. Breaking source into 3 columns keeps lineage, let's you access your expected source, and also let's us track parent history in evolutionary searches. In your case, you'll use the source_seed attribute which will always be the OQMD entry no matter how many steps the workflow was (seed means go all the way back to the beginning).