WEBVTT

NOTE Narration transcript. Text is the script synthesized by minimax (speech-2.8-hd, voice English_expressive_narrator); each cue was synthesized as its own audio file and is timed by that file's measured duration. Verified at 106.0% of expected length, 136.9 wpm (median cue 137 wpm). Script source: public/videos/methane-and-refrigerants-cutting-the-non-co2-climate-forcers.vtt at commit 4641d96 (pre-overwrite narration prose)

1
00:00:00.000 --> 00:00:02.350
Carbon dioxide gets the headlines.

2
00:00:02.650 --> 00:00:10.530
But cutting methane thirty percent by 2030 could avoid about zero point three degrees of warming by 2040.

3
00:00:10.830 --> 00:00:17.520
Phasing down HFC refrigerants could avoid up to another zero point five degrees by 2100.

4
00:00:17.820 --> 00:00:20.330
That is huge near-term leverage.

5
00:00:20.630 --> 00:00:25.150
Here is the twist: both targets depend on materials we do not have yet.

6
00:00:25.450 --> 00:00:34.830
We need catalysts that convert stranded methane without burning the product, and refrigerants that are efficient, safe, and low-warming, all at once.

7
00:00:35.130 --> 00:00:38.370
Fast atomistic AI should help search the options.

8
00:00:38.670 --> 00:00:48.430
Except its biggest errors happen exactly where these materials work: surfaces, transition states, radical fragments, and phase boundaries.

9
00:00:48.730 --> 00:00:58.370
In those under-coordinated environments, universal models can soften the energy landscape by fifteen to sixty percent, and rank the wrong candidate first.

10
00:00:58.670 --> 00:01:02.210
Methane's carbon-hydrogen bond is brutally strong.

11
00:01:02.510 --> 00:01:09.880
A catalyst active enough to crack it often holds the methanol product too tightly and over-oxidizes it into carbon dioxide.

12
00:01:10.180 --> 00:01:17.520
Corrected barriers can expose rare sites that break this trap, and open lower-temperature methanol or cleaner hydrogen routes.

13
00:01:17.820 --> 00:01:21.060
Refrigerants face a different-looking version of the same problem.

14
00:01:21.360 --> 00:01:29.630
Millions of molecules must be screened for vapor pressure, efficiency, flammability, toxicity, and atmospheric lifetime.

15
00:01:29.930 --> 00:01:37.520
The safety-critical steps involve radicals and breaking bonds, the same low-coordination physics that raw models misread.

16
00:01:37.820 --> 00:01:46.610
The correction measures the error versus atomic coordination, anchors a correction field, and applies it at runtime, without retraining the base model.

17
00:01:46.910 --> 00:01:51.550
Then machine-checked proofs mark which predictions stay inside the measured domain.

18
00:01:51.850 --> 00:02:02.760
Across thirty-six blind model-and-material combinations, corrected predictions reached a correlation of zero point nine zero six, with zero adjustable parameters.

19
00:02:03.060 --> 00:02:07.880
Methane catalysts and next-generation cooling are not separate discovery stories.

20
00:02:08.180 --> 00:02:16.789
They share one hidden error geometry, and one chance to turn fast screening into evidence we can actually trust, before the policy clock runs out.