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 103.7% of expected length, 139.9 wpm (median cue 135.5 wpm). Script source: public/videos/five-materials-for-5-to-12-gtco2-year.vtt at commit 4641d96 (pre-overwrite narration prose)

1
00:00:00.000 --> 00:00:05.200
The gap between a climate target and an actual climate solution is usually a material.

2
00:00:05.500 --> 00:00:13.220
And 5 missing materials could unlock between 5 and 12 gigatons of carbon dioxide reductions every year.

3
00:00:13.520 --> 00:00:15.350
So, what are they?

4
00:00:15.650 --> 00:00:21.620
First, cobalt-free cathodes that push past a thousand charge cycles without voltage fade.

5
00:00:21.920 --> 00:00:34.510
Second, solid electrolytes made from abundant elements, so lithium metal batteries can pass 400 Wh per kilogram without flammable liquid or critical material pricing.

6
00:00:34.810 --> 00:00:38.740
Third, metal organic frameworks for direct air capture.

7
00:00:39.040 --> 00:00:49.410
They need to grab carbon dioxide at just 400 PPM, ignore all that competing water, survive humidity, and stay cheap enough to manufacture by the ton.

8
00:00:49.710 --> 00:00:57.180
Fourth, an ammonia catalyst that works near room conditions instead of inside the energy-hungry Haber-Bosch process.

9
00:00:57.480 --> 00:01:06.780
And fifth, a lead-free perovskite solar absorber that can clear 20% efficiency and last 25 years without tin oxidizing in minutes.

10
00:01:07.080 --> 00:01:10.020
Different industries, one shared failure.

11
00:01:10.320 --> 00:01:17.560
Performance depends on defects, surfaces, and atomic transition states where atoms have fewer neighbors.

12
00:01:17.860 --> 00:01:23.530
That is exactly where fast universal models soften the energy landscape and get the ranking wrong.

13
00:01:23.830 --> 00:01:31.420
Brute force quantum calculations can recover the answer, but they are far too slow for millions or trillions of candidates.

14
00:01:31.720 --> 00:01:36.620
A 100 millielectron volt error can change a predicted hopping rate 50-fold.

15
00:01:36.920 --> 00:01:45.450
The correction measures that systematic error as a physical field corrects it while the model runs and checks the result with machine-verified proofs.

16
00:01:45.750 --> 00:01:47.240
Good candidates move up.

17
00:01:47.540 --> 00:01:49.030
Bad ones fall away.

18
00:01:49.330 --> 00:01:54.400
And when a prediction leaves the field's measured domain, the system can refuse to pretend.

19
00:01:54.700 --> 00:02:03.382
Five materials, one correction layer, and a path from plausible crystals to climate hardware that can actually scale.