Discover ways to transform this post
POST /speech/from-post
How this post is connected to other assets
Finished a couple basic simulations:
Temperature ramping AIMD simulation of H2O (mp-697111), taken from 0 K to 300 K over 10ps.
https://next-gen.materialsproject.org/materials/mp-697111
Note on .cif visual above: bonding algo for our material visualizations are not so good. Will work on something better, or at least more configurable.
MLIP model: M3GNet
Friction: 0.1 / units.fs
Time step: 1.0 * units.fs
Steps: 15000 -> 15ps
Temperature range: 0 K to 300 K
Temperature ramp duration: 10ps
Supercell: 3x3x3
Total atoms: 324
Changed from previous simulation (NaCl):
Increased friction from 0.01 / units.fs to 0.1 / units.fs
Increased trajectory sampling rate, from 1 frame per 1000 steps to 1 frame per 100 steps
A little more action this time! Very cool to see the lattice pick up more energy and really start falling out of order. Next time, I'll just simulate right below freezing to right above freezing so we can watch the phase transition, if that works. By the end of the simulation here at 300 K, any resemblance of a crystal structure is gone and it's just free floating H20.
Also, I increased the trajectory sampling rate here so now each frame is 100 fs. Theoretically we can get 1 frame per femtosecond since that's our simulation step size, but we'll see how useful that much data actually is.
Watching the hydrogens on the oxygen is interesting too. They don't perfectly keep the 104.5 degree bond angle but instead move around quite a bit.
Simulating ice into water
Steadier this time. It could be due to the increased friction which helped coupling between the environment and the material, but it could also be that this supercell was much larger. Around 300 atoms compared to around 50 for the NaCl simulation.
Trajectories file
Molecular dynamics simulation temperature ramping H2O 3x3x3 supercell from 0 K to 300 K
Discover other posts like this one
After wrestling with Mattergen finetuning for longer than I would've liked to, I pivoted back to simple property conditioned generation on Zn-Mg-H systems per 's recommendation. Each generated system
Over the last week or so, I've been working on making some upgrades to the superconducting state classifier model. See the first attempt here:
In this next experiment, I'm going to try to build more of an intuition and physical understanding of some of our most important latent features, as they relate to importance of predicting the superc