New example app: visualize structure relaxation

[janosh]

Might even be useful on the MP website for investigating unusual relaxations like mp-1033715 in this screen recording. Curious to know your thoughts @mkhorton.

Screen.Recording.2023-03-18.at.19.04.03.mp4

I originally built this to inspect relaxations from ML IAPs like M3GNet and CHGNet (related PR CederGroupHub/chgnet#14) but then thought this would make a good example app too.

[mkhorton]

I agree, this is useful. Could be added to the TaskDocument component in #313 even.

I will say that the longer-term goal was to support this via a Trajectory component which would render the Trajectory object obtainable from e.g. a vasprun or the API, and a lot of the groundwork is in place for this -- namely, animation support in the Scene renderer via Three.js.

This would be a lot faster, since geometries etc. are generated just once, and then it loops through a set of displacements, so it becomes much more responsive.

However, in the interim, I think this solution is great--it works today! so in that sense it is the best solution :)

[janosh]

I agree, this is useful. Could be added to the TaskDocument component in #313 even.

Good idea! Are there some preliminary screenshots of the TaskDocument UI somewhere to get a sense of what it will look like?

This would be a lot faster, since geometries etc. are generated just once, and then it loops through a set of displacements, so it becomes much more responsive.

I was thinking about that as well. Definitely makes sense. This PR constructs a list of structures one for each step which is needlessly expensive.

[janosh]

I added the normed force averaged over atoms to the relaxation plot as a secondary y-axis and cleaned up the code. The app now looks like this:

[mkhorton]

Good idea! Are there some preliminary screenshots of the TaskDocument UI somewhere to get a sense of what it will look like?

@janosh yes! The Materials Project "task detail" pages were intended as a proof-of-concept for this component. Changing it to a general TaskDocument component was considered beneficial since it would allow re-use of the component in e.g. an atomate2 UI or in other contexts.

[janosh]

Oh cool! I wasn't even aware these exist. For reference: https://materialsproject.org/materials/mp-1033715/tasks/mp-1033715

Perhaps the convergence graph should have the option to toggle between showing electronic and ionic steps as it can be hard to read for long trajectories. @mkhorton I'll create a tracking issue for that.