Trajctory output and documentation

[GrigorySarnitsky]

I want to get raw particle trajectories for hard-sphere fluid at equilibrium, but the output of the Trajectory plugin is incomplete and not well documented.

I am interested in exact trajectory data, not fixed-interval snapshots. Later I want to analyze this data to extract various statistics on the free flight times (PDFs, conditional PDFs, etc). So I use --load-plugin Trajectory and get trajectory.out file.

The trajectory.out file consists of lines like

00071348, Source=Interaction, SourceID=0, Event Type=CORE, t=0.0481, dt=0.0000
   2PEvent: p1=2838937, p2=3250973, delP1=Vector{-1.897005143001594,-2.0577878813188395,-0.096265160431851543}, |r12|=1.0000, post-r12=Vector{0.67739874617502949,0.73481241515979434,0.034375183018056532}, post-v12=Vector{0.89103846739932724,2.918355977194599,1.5241686506073979}, post-rvdot=2.8004
00071349, Source=Interaction, SourceID=0, Event Type=CORE, t=0.0481, dt=0.0000
   2PEvent: p1=1233447, p2=1273417, delP1=Vector{0.12376906732442466,-0.70381774806314157,-1.0129318312133937}, |r12|=1.0000, post-r12=Vector{-0.099842693089243539,0.56775946470078509,0.81711726629620074}, post-v12=Vector{-1.0570535929212619,0.29568228747546499,1.1824804821708463}, post-rvdot=1.2396
00071350, Source=Interaction, SourceID=0, Event Type=CORE, t=0.0481, dt=0.0000
   2PEvent: p1=1443174, p2=1969684, delP1=Vector{0.55884805321490627,1.3833099246434661,2.2670967950890573}, |r12|=1.0000, post-r12=Vector{-0.20591600722677583,-0.50970143816572744,-0.83534606116242038}, post-v12=Vector{-1.929139595760371,-1.4950621032023974,-1.8611270717922062}, post-rvdot=2.7140

First, note that the time values are truncated, so that all three events are shown to happen at exactly the same time t=0.0481 with zero time separation between them dt=0.0000. Can this be tuned to show exact times? I think I can adjust the output code myself and recompile, just point me what file/place to change and give general recommendations. Also, I wonder if the event ID is correctly output when it overflows 100 million.

Second, I am not sure what type of data is being stored. I guess, DynamO stores the event number (e.g. 00071348) and the IDs of particles participating in the event (p1=2838937, p2=3250973). But what are delP1, post-r12, post-v12? Why do you store the inner product post-rvdot? How to obtain the sequence of velocities the particle has and the exact times it experienced collisions?

[toastedcrumpets]

Hi there,
The code lines you're looking for are here
https://github.com/dynamomd/DynamO/blob/master/src/dynamo/dynamo/outputplugins/trajectory.cpp#L43

You can see formatting commands there. I used to use the trajectory file for manual/hand debugging of new collisions, so it wasn't designed for fully accurate replays but to be easily readable by eye, so I used fixed width output to align everything. You can easily alter the precision statements to make it full precision though, and nowadays I would feed this into a python script for formatting anyway. So please feel free to increase the precision and make a pull request if you're motivated that way.

Regarding the data itself, delP1 is the change in momentum for particle 1. That is all you need to reconstruct most trajectories. The values of post-r12 and post-v12 are the post collision relative position and velocity. Finally post-rvdot is the dot product of those two. These are output for me to check certain things about the event, again these can probably be removed if you're looking to create a clean "trajectory" file.

Hope this helps you get started!

[GrigorySarnitsky]

Thank you, it seems that I am getting out what I want.

However there is something I do not understand. I am running a simple simulation of a hard sphere gas in a periodic cube. To my understanding I shall get only pair events. However, occasionally I get single particle events of type VIRTUAL. It seems that they occur only in less dense systems.

I encountered the issue with the simulation at rho = 0.01 and N = 4 * 128^3. I could provide you with more diagnostic information on them if needed. There are not a lot of them, 5976 out of 108 482 036 total events.

[toastedcrumpets]

Hi, those virtual events are probably from the PBC Sentinel. Because we only test for interactions with the nearest periodic image, whenever a particle travels more than half the periodic image length in any direction (less the interaction diameter) without having an event we introduce a virtual event so it tests against the new nearest images.

You'll only see these at low densities, and they can be ignored.