WEBVTT

1
00:00:00.000 --> 00:00:18.596
Hydrofluorocarbons were introduced to protect the ozone layer, but their warming potential is hundreds to thousands of times that of carbon dioxide. The Kigali Amendment now aims to cut HFC consumption roughly eighty percent by 2047, which could avoid up to half a degree Celsius of warming by 2100.

2
00:00:18.596 --> 00:00:39.555
Finding replacements is hard because the design space is constrained by thermophysical performance, flammability, toxicity, atmospheric lifetime, and lubricant compatibility all at once. Computational screening must predict vapor pressure, latent heat, transport properties, and decomposition pathways across millions of small molecules.

3
00:00:39.555 --> 00:00:58.959
Each of those properties depends on under-coordinated environments—radical transition states, surfaces, and phase boundaries—where generic force fields and universal machine-learning potentials systematically soften the energy surface. That softening can invert rankings and send labs chasing the wrong molecule.

4
00:00:58.959 --> 00:01:22.157
The new research payloads turn molecular trajectories into inspectable datasets. They carry full per-atom charges, velocities, forces, potential and kinetic energy, plus thermo tables and temperature-profile sidecars, all streamed in sync with the simulation. For refrigerants, that means watching the temperature profile evolve and correlating forces with local structure.

5
00:01:22.157 --> 00:01:41.748
Vector glyphs render per-atom force, velocity, and dipole arrows as camera-facing ribbon impostors, with p95 auto-scaling and magnitude colormaps. A fix ave/chunk parser replays spatial temperature, density, and velocity profiles in lockstep with the trajectory—the view transport-studies researchers actually need.

6
00:01:41.748 --> 00:02:04.447
Inspectable trajectories feed directly into the correction-and-verification layer. The correction recovers trustworthy barriers and lifetimes in under-coordinated environments, while the verification layer separates supported predictions from synthesis-dependent ones. Screening stops chasing false winners and starts sending laboratories candidates they can trust.