Leginon

We do very little tomography here, but I will ask around if anyone has done a very long session. There are some people at Caltech who I think do long overnight or weekend sessions with the tomography application. I will check with them also.

The tomography application in Leginon behaves mostly like the standard MSI application in how it deals with correcting for drift. There are a few settings in various nodes that determine the behavior of drift detection and correction: Anytime a drift is known to have occurred, acquisition nodes will attempt to adjust each target to compensate for this drift. By acquisition node, I mean Grid, Square, Hole, Exposure, and also Tomography, which behaves much like standard acquisition nodes. Each of those nodes have a setting "Adjust target for drift" which enables this function. They will only make this target adjustment if a drift has been declared since the target was initially chosen. A drift is declared in one of the following ways:

- drift is directly measured as part of the autofocus routine

- drift is assumed to have occured whenever a Z stage correction is made by a focus node

- drift is assumed to have occured for any target that has been queued up (regardless of how long it has been sitting in the cue)

This final option is the most important for doing a long session with queued up targets. When you turn queuing on in the targeting node, be sure to also turn on the option "declare drift when queue submitted". This will ensure that each of those queued targets will be adjusted when they are actually acquired. For this adjustment to work, you also need the "adjust target for drift" option turned on in the node that will actually acquire those targets.

The way that a target is actually corrected following a drift is that the parent image where the target was chosen is reacquired and correlated with the original. Then the targets are adjusted by that distance.

Here is a chronological summary of how drift detection and correction works in Leginon:

- target is chosen on image A

- possibly several minutes or hours pass...

- drift is declared (because it was measured or assumed as described above)

- possibly several minutes or hours pass...

- target is finally going to be acquired, so it needs to be adjusted

- drift manager node reacquires image A and measures adjustment necessary

- updated target is used to acquire final image (or tilt series acquired)

- if more targets were chosen on image A, the same adjustment would apply to all of them, so image A only needs to be reacquired once.

We have successfully run overnight sessions with Leginon in the past. (We have done a bunch of hardware and software upgrades, so we don't have Legninon Tomography running smoothly at the moment.) There was even a 30-hour run at one point. The success of the long runs depends on the performance of the stage and to a lesser extent, the GIF. If the stage can reproducibly return to a saved position within one field of view of the "hole" preset, the long runs are very successful.

We "adjust for drift" by stage shift in the hole and z-focus nodes, and by image shift in focus, tomography, and tomography preview nodes. We also align the ZLP before every tilt series - there is a mechanism to move to a designated empty area between tilt series.

The take-home message is that it is definitely possible to do long (>24h) tomography runs with Leginon on a Polara with a GIF, as long as you have a good stage and a well-aligned scope.

Best,

Prabha