Replace random steps with mean free path?

[jgostick]

@TomTranter Regarding the idea of doing Knudsen tortuosity with this package, how hard would it be to replace the random step with a predetermined step based on a mean free path value?

[TomTranter]

Not sure what you mean by predetermined. Do you mean the direction? Or are you talking about adjusting the length of the step?

[TomTranter]

If it's step length then that wouldn't be necessary. You could just change the time per step and I think that would simply mean scaling the msd so might be a bit redundant

[jgostick]

Sorry, I wrote my initial point quickly just to start the thread...let me explain:

If we could get each walking moving in a random direction, and keep it moving in that direction until either (a) it hit a wall or (b) it reached some predetermine distance (i.e. the mean free path at the given temperature and pressure), then we could probe Knudsen effects with no tricks (like weight step probability with local thickness).

The challenge is that the 'direction' needs to be more than just 6 grid directions. I "think" we'd need the walker direction vectors to be 'continuous'. The path would still follow voxels, but it would trace out a Bresenham line between it's start and end points. It would then either hit a wall and start a new line, or reach the end of the mean free path value and start a new line.

[TomTranter]

That would be a significant change. Lbm seems to work pretty well with discrete vectors so I'm not sure continuous ones are a good idea. Weighting probability of moving or changing direction would be simpler. However, there are plenty of things you could add to the algorithm but tying them to a physical process is more tricky. I'm interested to know more what you have in mind though. I'm planning to do something with the grey scale soon too