From First Principles

From First Principles

The Tech Elon Has Been Waiting For (EP 51)

July 31, 2026 · transcript

ABOUT THIS EPISODE

The Tech Elon Has Been Waiting For (EP 51)

What happens when electronics can operate at temperatures hot enough to melt aluminum?

In this deep-dive episode, Lester Nare and Krishna Choudhary examine a new high-temperature memory device developed by researchers at USC, the Air Force Research Laboratory, Kumamoto University, and their collaborators.

Published in Science, the experimental memristor combines tungsten, hafnium oxide, and graphene. It operated reliably at 700°C—roughly 1,300°F—retained data for more than 50 hours, and survived more than one billion switching cycles.

We begin by explaining why conventional electronics and flash memory fail when temperatures rise. From deep-earth drilling and hypersonic aircraft to nuclear systems and the surface of Venus, many environments where intelligent electronics would be useful remain inaccessible to today’s hardware.

Krishna then builds the memristor from first principles. We explore the history of the “missing” fourth circuit element, how oxygen vacancies create low- and high-resistance memory states, why conventional platinum electrodes fail under extreme heat, and how graphene prevents tungsten atoms from diffusing through the device.

Finally, we examine the implications for artificial intelligence. Memristors can potentially store neural-network weights and perform matrix multiplication in the same physical location, reducing the energy wasted moving information between processors and memory.

Could that combination of heat tolerance and energy efficiency make AI data centers in space more practical? Lester and Krishna work through thermal radiation, radiator size, power consumption, radiation resilience, and the considerable engineering challenges that remain.

Support the showDonate: FFPod.com/donateFollow: @FFPod on X / Instagram / TikTok / Facebook

Research and Show Notes

High-temperature memristors enabled by interfacial engineering

USC: A memory device that operates at 700°C

The development of carbon-neutral data centres in space

NASA Venus facts

Science

English

United States

Use API to fetch this episode in JSON

IN THIS EPISODE

krishna choudhary

lester nare

Edit these tags

TRANSCRIPT 🔗

WEBVTT
1
00:00:00.431 --> 00:00:04.692
Industrial flash memory tops out at around 200 degrees Celsius.
2
00:00:05.293 --> 00:00:07.933
And above that, there's essentially nothing that works.
3
00:00:08.654 --> 00:00:11.014
It's a Memristor, which is a memory switching device.
4
00:00:11.194 --> 00:00:14.515
And it's made from graphene, hafnium oxide and tungsten.
5
00:00:14.936 --> 00:00:19.317
And it operates reliably at 700 degrees Celsius.
6
00:00:19.377 --> 00:00:19.757
Really cool.
7
00:00:19.797 --> 00:00:21.978
I mean, it's hotter than molten aluminum.
8
00:00:22.178 --> 00:00:25.159
Aluminum melts at 660 degrees Celsius.
9
00:00:25.719 --> 00:00:29.260
If you put this thing in a vat of aluminum, it will still retain memory.
10
00:00:29.740 --> 00:00:30.541
Hello, Internet.
11
00:00:30.561 --> 00:00:40.808
This is your captain speaking, Lester Nare, joined as always by my co-host and our resident PhD, Krishna Chowdhury.
12
00:00:41.388 --> 00:00:44.751
We have a special deep dive today for you all.
13
00:00:45.091 --> 00:00:47.132
Be sure to check out our rundown from last week.
14
00:00:48.677 --> 00:00:55.622
Be sure to heat up the future of electronics as we look at a breakthrough in high temperature memory.
15
00:00:56.082 --> 00:01:04.789
This is something that Elon has been waiting for for a long time that may enable electronics that can survive lava.
16
00:01:05.129 --> 00:01:06.250
I'm going to need the details.
17
00:01:07.110 --> 00:01:08.852
I'm going to need to understand what's going on here.
18
00:01:09.312 --> 00:01:14.698
This comes from a new research paper in the Journal of Science, shout out to science, from March 26, 2026.
19
00:01:14.838 --> 00:01:24.427
It was a collaboration between researchers at USC here in California, the Air Force Research Lab in Dayton, Ohio, and the Kumamoto Research Lab.
20
00:01:24.927 --> 00:01:49.710
university in japan please correct my pronunciation with funding for this research being provided by national science foundation the army research office and the air force research lab so clearly they think there is military applications i'm sure it relates to hypersonics as always we are going to learn about the science from the ground up today because this is from first principles
21
00:02:05.570 --> 00:02:10.811
Today we're covering an invention that I think is going to be a game changer for the future of computing.
22
00:02:10.991 --> 00:02:13.572
You know, we live in the digital age of electronics.
23
00:02:14.092 --> 00:02:16.553
Chips are the bedrock of our economy.
24
00:02:16.653 --> 00:02:21.694
And yet all of those chips really have a Goldilocks zone when it comes to when they work.
25
00:02:22.074 --> 00:02:25.575
Goldilocks in the traditional sense of like not too hot, not too cold.
26
00:02:26.255 --> 00:02:26.995
This is...
27
00:02:28.277 --> 00:02:30.998
research is getting to that not too hot part.
28
00:02:31.418 --> 00:02:35.240
You know, the Papa Bear that was like too hot, the porridge was too hot.
29
00:02:35.680 --> 00:02:38.981
Now we're getting a chip that can withstand that porridge.
30
00:02:39.301 --> 00:02:40.502
Okay.
31
00:02:41.102 --> 00:02:46.604
These are memory chips that happen in our phone, in our laptop, in the cameras that we have here in the studio.
32
00:02:46.964 --> 00:02:54.747
They're all designed to work reliably roughly between 85 degrees Celsius, so 200 Fahrenheit for our Americans.
33
00:02:54.767 --> 00:02:55.207
Okay.
34
00:02:55.367 --> 00:02:59.150
to about 150 degrees Celsius, so 300 Fahrenheit, okay?
35
00:02:59.450 --> 00:03:01.132
That's where they sit.
36
00:03:02.132 --> 00:03:14.302
And industrial flash memory tops out at around 200 degrees Celsius, even with a lot of the heroic packaging where you try to insulate temperature and things like that.
37
00:03:14.782 --> 00:03:18.345
And above that, there's essentially nothing that works, okay?
38
00:03:19.331 --> 00:03:23.134
There's no computation happening above 300 degrees Celsius.
39
00:03:23.754 --> 00:03:28.237
But that doesn't mean that we don't want to have computation happening at that temperature.
40
00:03:28.257 --> 00:03:34.562
There are a lot of applications that even I can think of here, not being in the field, that would be really great.
41
00:03:35.122 --> 00:03:37.123
to have at 300 degrees or above.
42
00:03:37.483 --> 00:03:39.064
Okay, for example, oil and gas.
43
00:03:39.544 --> 00:03:42.146
Oil and gas drillers, they push deep into the Earth's crust.
44
00:03:42.386 --> 00:03:48.669
The deeper you get into the Earth's crust, the closer you're getting to the mantle where all of the magma is.
45
00:03:49.069 --> 00:03:50.910
So the temperature is going to go up.
46
00:03:51.310 --> 00:03:55.232
At 5 kilometers, you're flirting with like 150 degrees Celsius.
47
00:03:55.252 --> 00:03:56.312
That's already pushing it.
48
00:03:56.652 --> 00:03:59.074
At 10 kilometers, you're well past 300 degrees.
49
00:03:59.494 --> 00:04:03.776
And the sensor packages that we have today, when you send them down these boreholes...
50
00:04:05.136 --> 00:04:12.959
They need to either operate with like an active cooling thing where like there's something with the drill that's trying to cool the sensors.
51
00:04:13.320 --> 00:04:16.141
That's always not convenient.
52
00:04:16.181 --> 00:04:16.861
Not ideal.
53
00:04:17.101 --> 00:04:17.661
Yeah.
54
00:04:17.761 --> 00:04:21.503
And then they have to transmit their data to the surface in real time.
55
00:04:21.523 --> 00:04:23.204
They can't do any computation down there.
56
00:04:23.704 --> 00:04:24.264
Interesting.
57
00:04:24.304 --> 00:04:28.586
So then there's this like massive wiring problem that you have to deal with.
58
00:04:28.866 --> 00:04:31.007
The wires are going through this really hot process.
59
00:04:32.079 --> 00:04:35.443
The engineering is way more complex unnecessarily if you could just have it.
60
00:04:35.563 --> 00:04:35.803
Yes.
61
00:04:35.963 --> 00:04:40.128
If you could just have like a chip that's right on top of the drill, right?
62
00:04:40.348 --> 00:04:41.149
That's right there.
63
00:04:41.189 --> 00:04:43.571
Then you can have something like an autonomous downhole.
64
00:04:44.333 --> 00:05:05.252
drill like electronics that decides where to go right there i mean we've got all this ai but we can't actually put that to use when we're going all the way down there right if i'm going to make another armageddon reference when we need to go to an asteroid and and and we need to i guess that's a bad example because it's not that hot
65
00:05:05.552 --> 00:05:09.734
Yeah, well, I mean, but even the friction is going to create like a hot, yeah, some heat.
66
00:05:10.274 --> 00:05:11.655
You know, we're going to want to be able to do that.
67
00:05:11.735 --> 00:05:12.415
Yeah, we just look.
68
00:05:12.435 --> 00:05:13.196
Planetary defense.
69
00:05:13.336 --> 00:05:19.179
Yes, as always, you know, we are defense first here, not war, only defense.
70
00:05:19.699 --> 00:05:21.840
Now, you mentioned hypersonics.
71
00:05:22.000 --> 00:05:23.781
That's another big thing, right?
72
00:05:23.921 --> 00:05:34.526
In aerospace engineering, the inside of a jet engine or the inside of a rocket engine or the nose of a hypersonic aircraft, these things are getting to really, really high temperatures.
73
00:05:34.606 --> 00:05:34.946
Right.
74
00:05:35.086 --> 00:05:40.150
We've covered this in previous stories where the temperature gets so high that chemistry starts mattering.
75
00:05:40.210 --> 00:05:40.951
It gets a little weird.
76
00:05:40.991 --> 00:05:42.592
Yeah, and it starts getting a little weird.
77
00:05:43.072 --> 00:05:48.016
And we can't make any intelligent decisions on site because of the chip.
78
00:05:48.377 --> 00:05:49.778
The chip is just going to get fried.
79
00:05:50.298 --> 00:05:57.424
And finally, the thing that Elon has been worried about and all of these tech bros are worried about is putting data centers in space.
80
00:05:57.884 --> 00:05:58.124
Right.
81
00:05:58.164 --> 00:06:00.726
This is the number one conversation in SF right now.
82
00:06:00.806 --> 00:06:01.026
Yeah.
83
00:06:01.086 --> 00:06:01.286
Yeah.
84
00:06:01.366 --> 00:06:03.768
I mean, Elon wants to put data centers in space.
85
00:06:03.808 --> 00:06:04.848
He's not the only one.
86
00:06:05.529 --> 00:06:09.391
SpaceX had an IPO of one point seven trillion dollars.
87
00:06:09.811 --> 00:06:15.435
And I think a lot of that hinges on Elon's promise that he can put AI data centers in space.
88
00:06:15.935 --> 00:06:20.378
We're going to get into the details there, but there's been a lot of pushback.
89
00:06:21.158 --> 00:06:23.099
Because of fundamental physics, right?
90
00:06:23.139 --> 00:06:28.282
They're saying, oh, like fundamental physics is telling you that you can't put data centers in space.
91
00:06:28.702 --> 00:06:30.583
Well, yes and no.
92
00:06:30.883 --> 00:06:31.163
Okay.
93
00:06:31.463 --> 00:06:40.948
Yes, in the sense that, yes, the physics and the electronics that we have today, it's probably not a good idea to put data centers in space.
94
00:06:41.469 --> 00:06:43.630
But with this problem...
95
00:06:44.379 --> 00:06:47.782
You can have innovation like the one that we're going to talk about, and perhaps it's possible.
96
00:06:47.822 --> 00:06:58.351
The main problem really is that to keep a one megawatt artificial intelligence server cool in the vacuum of space, you need a radiator that's roughly the size of four tennis courts, right?
97
00:06:58.531 --> 00:07:00.352
That's because the space is a vacuum.
98
00:07:00.633 --> 00:07:03.555
There's no way to extract heat other than radiation.
99
00:07:04.276 --> 00:07:10.201
And radiation is not a great way to let go of heat because you're relying on photons.
100
00:07:10.321 --> 00:07:13.023
It's not – is it – it's like because it's not efficient, right?
101
00:07:13.083 --> 00:07:14.084
Yes, exactly.
102
00:07:14.224 --> 00:07:15.124
It's just not efficient.
103
00:07:15.344 --> 00:07:17.305
We're going to get into some of the physics of that later.
104
00:07:18.726 --> 00:07:25.569
And my favorite sort of conundrum that I think this is going to solve is planetary exploration.
105
00:07:25.609 --> 00:07:27.770
I think you're going to like this too.
106
00:07:28.211 --> 00:07:33.433
There's a reason why we have so many photos of Mars and not that many photos of Venus.
107
00:07:33.793 --> 00:07:34.033
Okay.
108
00:07:34.334 --> 00:07:35.934
Mars is pretty chill to go to.
109
00:07:36.335 --> 00:07:37.515
It's like kind of cold.
110
00:07:37.835 --> 00:07:38.276
Literally.
111
00:07:38.336 --> 00:07:40.957
It's like literally it's farther away from the sun and it's fine.
112
00:07:41.297 --> 00:07:47.839
Venus is closer to the Earth, and it's got this massive greenhouse effect, which means that it's really hard to put stuff down there.
113
00:07:48.120 --> 00:07:51.181
Actually, the first guys to do it were the Soviets.
114
00:07:51.321 --> 00:08:01.465
Venus has a surface temperature of about 465 degrees Celsius and an atmospheric pressure that's 92 times that of the Earth.
115
00:08:01.785 --> 00:08:10.388
The Soviets put down the Soviet Venera landers, and they're the only spacecraft that have ever survived on the Venus surface.
116
00:08:10.408 --> 00:08:10.908
That's unbelievable.
117
00:08:10.968 --> 00:08:13.010
But guess how long they survived?
118
00:08:13.610 --> 00:08:14.411
Over or under an hour?
119
00:08:15.131 --> 00:08:15.972
It's got to be under an hour.
120
00:08:16.032 --> 00:08:17.653
It's under an hour, yeah.
121
00:08:18.273 --> 00:08:19.895
It's between like 23 minutes.
122
00:08:19.955 --> 00:08:22.256
I think their record was like two hours later on.
123
00:08:22.677 --> 00:08:24.578
But like two hours?
124
00:08:24.598 --> 00:08:25.018
Yeah, yeah.
125
00:08:25.079 --> 00:08:25.219
No.
126
00:08:25.259 --> 00:08:25.439
Right?
127
00:08:25.459 --> 00:08:30.883
Like the stuff that JPL puts on Mars has lasted like 20 years.
128
00:08:30.903 --> 00:08:33.205
Yeah, we're talking about two decades versus an hour.
129
00:08:33.245 --> 00:08:34.085
Yes, versus two hours.
130
00:08:34.105 --> 00:08:34.506
Right, right.
131
00:08:34.526 --> 00:08:34.906
Two hours.
132
00:08:35.026 --> 00:08:36.027
Is the best that they can do.
133
00:08:36.787 --> 00:08:39.809
I will note, this is, for those who are TV fans –
134
00:08:40.830 --> 00:08:45.113
Uh, we've talked about for all mankind on the show before, but Apple has a second.
135
00:08:45.173 --> 00:08:47.135
Apple TV has a second show called star city.
136
00:08:47.455 --> 00:08:47.635
Okay.
137
00:08:47.655 --> 00:08:56.562
It looks at the Russian space program in the same alternative history timeline where they got to the moon first and they do a whole, all of season one.
138
00:08:57.122 --> 00:08:59.344
Part of it relates to the Venus mission.
139
00:08:59.544 --> 00:08:59.904
Really?
140
00:08:59.944 --> 00:09:00.965
Dude, I got to watch that.
141
00:09:01.005 --> 00:09:03.006
It's, it's, it's alternative reality.
142
00:09:03.467 --> 00:09:05.668
So, you know, no spoilers.
143
00:09:05.728 --> 00:09:06.929
I love, I love that kind of stuff.
144
00:09:06.949 --> 00:09:09.211
But it is interesting to kind of understand the,
145
00:09:09.760 --> 00:09:13.864
The technical challenges, which they brushed on at a high level in the show.
146
00:09:14.285 --> 00:09:18.169
But recently I've been watching it and this brought it up as a concept.
147
00:09:18.549 --> 00:09:21.673
And the engineering problem is non-trivial.
148
00:09:21.933 --> 00:09:23.334
Yeah, no, completely, right?
149
00:09:23.414 --> 00:09:25.016
It's like going towards the center of the earth.
150
00:09:25.056 --> 00:09:26.077
It's the same drilling thing.
151
00:09:26.137 --> 00:09:27.739
High pressure, high temperature.
152
00:09:28.039 --> 00:09:30.362
Your electronics are going to get completely wrecked.
153
00:09:30.742 --> 00:09:30.942
Right.
154
00:09:31.302 --> 00:09:34.645
The photo we have from the Venera 9 in 1975.
155
00:09:34.905 --> 00:09:36.126
This is the best photo they got.
156
00:09:36.166 --> 00:09:39.228
It's the first image of a planet other than the Earth.
157
00:09:39.729 --> 00:09:41.350
Credit to the Soviets.
158
00:09:41.390 --> 00:09:45.453
The first image of a planet other than the Earth was not the United States and Mars.
159
00:09:45.633 --> 00:09:46.774
It was Venus.
160
00:09:47.074 --> 00:09:47.995
And that's all they got.
161
00:09:48.035 --> 00:09:49.756
But still, that's kind of crazy.
162
00:09:50.096 --> 00:09:51.157
But who got to the moon, though?
163
00:09:51.417 --> 00:09:53.939
Yeah, that's the real question.
164
00:09:53.979 --> 00:09:56.642
They were like, okay, let's go the other direction.
165
00:09:57.482 --> 00:09:59.644
The race is actually this way, guys.
166
00:09:59.744 --> 00:10:03.007
Later missions, they were actually able to get some color photos.
167
00:10:03.047 --> 00:10:04.748
This is the one that lasted two hours.
168
00:10:05.449 --> 00:10:07.090
And you could get some color photos.
169
00:10:07.651 --> 00:10:12.355
You can see the sort of yellow haze that we know that the Venus has if you point a telescope to it.
170
00:10:12.595 --> 00:10:13.896
But again, only two hours.
171
00:10:13.956 --> 00:10:18.480
And you can see it kind of looks like the surface of a volcanic planet.
172
00:10:18.640 --> 00:10:40.024
debris field because there's a bunch of volcanoes on venus um but that's that's it that's all we could get and we can imagine you know as we continue to expand in our own local solar system and otherwise having electronics that can survive high temperature environments is going to become yeah a valuable asset to expand where we can actually go
173
00:10:40.284 --> 00:10:40.684
Exactly.
174
00:10:40.724 --> 00:10:43.945
I mean, I'd like to know more about the surface of Venus, right?
175
00:10:44.246 --> 00:10:48.347
We have better surface images of Mars, Pluto.
176
00:10:49.107 --> 00:10:55.530
We have better surface images of Pluto than we do with large swaths of Venus, which is our nearest neighbor.
177
00:10:55.850 --> 00:10:57.510
Venus is closer to us than the Earth.
178
00:10:57.891 --> 00:10:59.951
And we still don't know what's under those clouds.
179
00:10:59.971 --> 00:11:00.652
That's so unbelievable.
180
00:11:00.892 --> 00:11:01.112
Right?
181
00:11:01.312 --> 00:11:04.173
And it's just because it's a completely...
182
00:11:05.677 --> 00:11:08.179
Messed up situation to get down there on the surface.
183
00:11:08.319 --> 00:11:09.059
Everything fails.
184
00:11:09.119 --> 00:11:11.681
We can't stand the heat, so we can't go in the kitchen.
185
00:11:11.761 --> 00:11:12.742
Yeah, exactly.
186
00:11:13.162 --> 00:11:13.583
Exactly.
187
00:11:14.343 --> 00:11:18.226
And into this landscape comes this paper that was published by USC.
188
00:11:19.347 --> 00:11:22.229
It's describing something I think genuinely new.
189
00:11:22.249 --> 00:11:22.849
Okay.
190
00:11:23.109 --> 00:11:25.451
It's a Memristor, which is a memory switching device.
191
00:11:25.471 --> 00:11:26.632
We're going to get into what that is.
192
00:11:27.525 --> 00:11:30.850
And it's made from graphene, hafnium oxide, and tungsten.
193
00:11:31.270 --> 00:11:35.676
And it operates reliably at 700 degrees Celsius.
194
00:11:35.857 --> 00:11:36.618
That's unreal.
195
00:11:37.038 --> 00:11:37.599
700 degrees.
196
00:11:37.619 --> 00:11:38.140
That's unreal.
197
00:11:38.160 --> 00:11:39.061
Yeah.
198
00:11:39.101 --> 00:11:43.007
They've raised the temperature by a factor of more than two.
199
00:11:43.927 --> 00:11:45.328
And so there's something fundamental.
200
00:11:45.348 --> 00:11:50.152
There's a real breakthrough here that's beyond an iterative upgrade.
201
00:11:50.192 --> 00:11:52.614
Yeah, I think it's really cool.
202
00:11:52.634 --> 00:11:54.795
I mean it's hotter than molten aluminum.
203
00:11:55.015 --> 00:11:57.998
Aluminum melts at 660 degrees Celsius.
204
00:11:58.558 --> 00:12:02.101
If you put this thing in a vat of aluminum, it will still retain memory.
205
00:12:02.501 --> 00:12:05.884
It's hotter than the surface of a lava flow on Earth.
206
00:12:06.305 --> 00:12:07.386
Not all lava flows.
207
00:12:07.466 --> 00:12:09.928
I know there's going to be people in the comments being like, no.
208
00:12:10.108 --> 00:12:10.949
The volcanologist?
209
00:12:10.969 --> 00:12:11.309
Yeah, yeah.
210
00:12:11.349 --> 00:12:13.711
The volcanologist is going to be like, no, lava is hotter than that.
211
00:12:13.872 --> 00:12:23.761
Actually, there is a volcano in Tanzania, the old Doinyo Lengai volcano, where the lava is 600 degrees Celsius.
212
00:12:23.881 --> 00:12:24.101
Okay?
213
00:12:24.321 --> 00:12:26.003
So there exists a lava...
214
00:12:26.703 --> 00:12:29.405
On Earth where I could put this chip and it would still work fine.
215
00:12:29.825 --> 00:12:32.346
So the tagline still kind of works, right?
216
00:12:32.626 --> 00:12:36.969
If you took it to Hawaii and put it in the lava in Hawaii, it would not.
217
00:12:37.069 --> 00:12:38.930
I think Hawaii's is like a thousand plus.
218
00:12:39.803 --> 00:12:42.425
But it perhaps might, and we'll get into that later.
219
00:12:42.465 --> 00:12:43.406
We're on the path.
220
00:12:43.646 --> 00:12:44.527
Yeah.
221
00:12:45.208 --> 00:12:53.635
The other crazy thing about this is the chip, the Memrister device, it maintained stable data retention for 50 hours.
222
00:12:54.615 --> 00:12:55.056
Really?
223
00:12:55.176 --> 00:12:55.756
50 hours.
224
00:12:55.876 --> 00:12:56.997
And it retained...
225
00:12:57.818 --> 00:13:02.720
That data retention after a billion switching cycles without failure.
226
00:13:02.960 --> 00:13:06.222
So this is both high temperature and long duration.
227
00:13:06.242 --> 00:13:06.502
Yes.
228
00:13:07.122 --> 00:13:08.743
Comparatively to where we were before.
229
00:13:08.783 --> 00:13:08.903
Yeah.
230
00:13:09.003 --> 00:13:11.824
And the physics is actually like it gives you both.
231
00:13:12.265 --> 00:13:14.346
Which is because both are different.
232
00:13:14.406 --> 00:13:14.686
Yeah.
233
00:13:15.046 --> 00:13:17.087
It's like a graph on different axes.
234
00:13:17.527 --> 00:13:18.568
And you can have high temperature.
235
00:13:18.608 --> 00:13:19.149
It's like, OK, great.
236
00:13:19.169 --> 00:13:20.029
But it's a microsecond.
237
00:13:20.049 --> 00:13:20.229
Yeah.
238
00:13:20.490 --> 00:13:20.810
Yeah.
239
00:13:20.870 --> 00:13:27.376
And here, what's cool is it's the same physics that's giving you both endurance limits, which I think is really, really cool.
240
00:13:27.396 --> 00:13:28.897
Yeah, that is very interesting.
241
00:13:28.977 --> 00:13:35.382
I think it's by a large margin the highest temperature at which any non-volatile memory has ever been demonstrated to work.
242
00:13:35.462 --> 00:13:36.804
This is phenomenal.
243
00:13:36.844 --> 00:13:36.964
Yeah.
244
00:13:36.984 --> 00:13:37.905
It's really, really cool.
245
00:13:37.965 --> 00:13:40.707
And I mean, why should normal people care, OK?
246
00:13:40.767 --> 00:13:41.568
Not scientists.
247
00:13:41.588 --> 00:13:41.768
Sure.
248
00:13:41.988 --> 00:13:51.310
Well, I think the border between environments where we can put intelligent electronics and environments where we cannot is about to shift dramatically.
249
00:13:51.550 --> 00:13:55.870
And then our understanding of everything about the world around us changes in a very big way.
250
00:13:55.990 --> 00:14:02.732
Yeah, I mean, it affects how we drill energy, how we explore planets, how we monitor volcanoes, how we put data centers in space.
251
00:14:03.132 --> 00:14:09.433
That has a lot of social and political implications for our society because then maybe we don't need to –
252
00:14:11.093 --> 00:14:12.713
Like take everyone's water.
253
00:14:13.314 --> 00:14:19.395
If you don't want a data center in your backyard, this story matters a lot because we can just put them in space.
254
00:14:19.555 --> 00:14:19.775
Yeah.
255
00:14:20.015 --> 00:14:20.215
Yeah.
256
00:14:20.315 --> 00:14:27.717
I mean, it's one of these like quiet material science results that it doesn't make the front page of the New York Times.
257
00:14:28.157 --> 00:14:33.258
But I think it's going to expand the frontier of what our civilization can do.
258
00:14:33.498 --> 00:14:34.458
You're saying there's a chance.
259
00:14:34.538 --> 00:14:34.778
Yeah.
260
00:14:34.898 --> 00:14:36.979
And I think I think I think it's an incredible story.
261
00:14:37.692 --> 00:14:48.117
We are going to get into the details of this has been such a big conversation in the business and technology world.
262
00:14:48.517 --> 00:14:54.540
We are so over leveraged on AI in the US in terms of its impact on our GDP.
263
00:14:54.960 --> 00:14:58.342
Everyone's saying it's fine because we're going to put data centers in space.
264
00:14:58.682 --> 00:15:01.083
And it's been very contentious as a.
265
00:15:02.647 --> 00:15:17.102
one aspect of that larger conversation this the potential for this to be an enabling layer is interesting but before we get into the details of that story a brief note of housekeeping for our long-time listeners in ffp nation
266
00:15:17.703 --> 00:15:18.364
Welcome back.
267
00:15:18.944 --> 00:15:39.785
If you happen to be catching this episode for the first time, it is the two of us here bringing you the latest breaking science news every week and any way that you can help to support the show so we can combat against the billionaire algorithms like a share a comment, bring it to Journal Club, put it into the group chat.
268
00:15:40.446 --> 00:15:50.115
All of those things help us reach more people with the way in which our world is changing and growing because of the advances we have at the frontier of science.
269
00:15:51.125 --> 00:15:56.630
Our website has all of our episodes, all of the research papers and show notes that we cover.
270
00:15:57.010 --> 00:16:08.179
So if you are someone who's a grad student, a postdoc, or someone who just really wants to get actually more into the weeds of any of these stories, go to ffppod.com.
271
00:16:08.519 --> 00:16:10.581
You can check out our research paper library.
272
00:16:10.961 --> 00:16:16.125
If you want to support the show, you can donate at ffppod.com backslash donate.
273
00:16:16.426 --> 00:16:19.168
And you can also check out the merch shop here.
274
00:16:19.548 --> 00:16:42.795
if you want to let the world know that you yourself are a member of ffp nation but enough of the riffraff let's get back to this very juicy story i do like my steaks medium rare but i i guess i can make it much quicker if i do it at 700 degrees celsius yes yeah it would it would only cook the outside like micron
275
00:16:44.186 --> 00:16:45.527
And then it would be all rare.
276
00:16:46.928 --> 00:16:48.128
Gordon Ramsay would love that.
277
00:16:48.809 --> 00:16:58.213
So let's first talk about why the conventional electronics that are in our phone, that are in the laptop, everywhere else, why they fail at high temperature.
278
00:16:58.334 --> 00:17:00.615
Why I can't bring my iPhone and put it in a lava flow.
279
00:17:00.735 --> 00:17:01.415
Yeah, yeah, yeah.
280
00:17:01.595 --> 00:17:02.276
I'd like to.
281
00:17:02.456 --> 00:17:03.336
Yeah, I could put it in water.
282
00:17:03.456 --> 00:17:04.137
Yeah, but...
283
00:17:05.157 --> 00:17:07.119
Lava is something you can't do, okay?
284
00:17:07.919 --> 00:17:09.641
The physics of high temperature failure.
285
00:17:10.141 --> 00:17:16.146
We first have to understand the traditional CMOS architecture that is at the base of all of our technology.
286
00:17:16.186 --> 00:17:19.509
This is complementary metal oxide semiconductor.
287
00:17:19.909 --> 00:17:22.431
These devices fail at around 200 degrees Celsius.
288
00:17:22.812 --> 00:17:25.214
Here's how it usually works, okay?
289
00:17:25.474 --> 00:17:30.258
So at zero Kelvin, which means absolute zero, everything is stationary.
290
00:17:30.938 --> 00:17:33.160
Silicon is an insulator.
291
00:17:34.441 --> 00:17:40.604
What ends up happening is atoms have something called a conduction and a valence electron shell.
292
00:17:41.164 --> 00:17:41.445
Okay?
293
00:17:41.725 --> 00:17:47.608
They've got parts where valence electrons can occupy all of the shell.
294
00:17:48.368 --> 00:17:49.709
So there's nothing going on.
295
00:17:49.849 --> 00:17:51.410
All of the shells are occupied, right?
296
00:17:51.710 --> 00:17:53.090
Then there's a conduction shell.
297
00:17:53.230 --> 00:17:56.812
There's a conduction energy level where there's like one or two electrons hanging out.
298
00:17:57.212 --> 00:18:00.714
And those electrons are the ones that skip from one atom to the next to the next.
299
00:18:01.054 --> 00:18:03.356
And they create current, which is what we use.
300
00:18:03.756 --> 00:18:03.976
Okay?
301
00:18:04.721 --> 00:18:12.184
Now, at zero degrees Kelvin, all of silicon's valence band is completely filled.
302
00:18:13.125 --> 00:18:18.207
If you put a bunch of metal and a bunch of silicon together, the gap...
303
00:18:18.954 --> 00:18:25.059
between the conduction band and the valence band gets bigger just because of neighboring effects.
304
00:18:25.379 --> 00:18:35.667
It's like the neighboring atoms start pulling on your own energy gaps, and it expands the gap between these two energy bands.
305
00:18:35.687 --> 00:18:37.848
The wiggle craze starts getting more intense.
306
00:18:38.089 --> 00:18:38.609
Exactly.
307
00:18:38.889 --> 00:18:42.272
And so at finite temperature, what you can do is you can bump an electron –
308
00:18:42.792 --> 00:18:43.793
You get some wiggle, right?
309
00:18:43.853 --> 00:18:45.414
Because like there's jiggling happening.
310
00:18:45.434 --> 00:18:47.796
There's photons that are at some energy level.
311
00:18:48.036 --> 00:18:50.077
That's going to come in, strike an electron.
312
00:18:50.097 --> 00:18:52.219
The electron is going to move up to the conduction band.
313
00:18:52.299 --> 00:18:55.361
And then voila, you can have a little bit of conduction.
314
00:18:55.881 --> 00:18:56.122
Okay.
315
00:18:56.542 --> 00:18:59.684
Now let's get into exactly how this happens.
316
00:18:59.944 --> 00:19:04.647
We've got a video that shows exactly how this conduction and valence band works.
317
00:19:04.668 --> 00:19:06.909
So on the bottom, you've got your silicon.
318
00:19:07.369 --> 00:19:10.972
You've got all of the sites that are now occupied by electrons, which are in the
319
00:19:12.201 --> 00:19:12.742
In the blue.
320
00:19:12.962 --> 00:19:13.923
It's like the housing market.
321
00:19:13.943 --> 00:19:14.044
Yeah.
322
00:19:14.064 --> 00:19:14.905
There's nothing on the market.
323
00:19:14.985 --> 00:19:15.505
Yeah, yeah, yeah.
324
00:19:15.526 --> 00:19:16.447
There's no openings.
325
00:19:16.507 --> 00:19:18.630
Now, all of a sudden, somebody sells a house.
326
00:19:18.970 --> 00:19:23.837
So an electron from the valence band moves into the conduction band.
327
00:19:24.577 --> 00:19:28.222
Now that conduction band electron can move around and create a current.
328
00:19:28.883 --> 00:19:34.447
Notice at the same time, the hole that was left in the valence band can also move around.
329
00:19:34.847 --> 00:19:44.394
It's like you have a bunch of packed theater seats and everybody moves one seat over, but then the empty seat can start moving around too.
330
00:19:44.734 --> 00:19:46.175
That's a positive thing.
331
00:19:46.295 --> 00:20:09.092
charge in some sense right because it's a hole the electron has gone away so it's left a positive hole and that hole can move around in the same sense that the electron can also move around because it's attracting because it's that empty space and it needs to be filled yeah yeah yeah it's the movement yeah the movement the actual movement the physical movement that's happening is always electrons right it's the negative charge carriers but the the the the sort of
332
00:20:09.953 --> 00:20:15.659
Thing that matters in the valence band is that positive hole that's moving around, right?
333
00:20:16.700 --> 00:20:21.906
It unlocks the stationary state that exists because it's now moved up to.
334
00:20:22.026 --> 00:20:22.547
Yeah, yeah, yeah.
335
00:20:23.007 --> 00:20:23.428
Exactly.
336
00:20:23.628 --> 00:20:25.830
And so you can have two different types of semiconductors here.
337
00:20:26.711 --> 00:20:31.176
Engineers can make silicon switches using either phosphorus or boron.
338
00:20:31.196 --> 00:20:31.857
You can dope.
339
00:20:32.377 --> 00:20:34.477
the silicon with either phosphorus or boron.
340
00:20:34.777 --> 00:20:38.938
And then these artificially dictate how many charge carriers you're gonna get.
341
00:20:39.038 --> 00:20:41.379
Are you gonna get more electrons in the conduction band?
342
00:20:41.639 --> 00:20:42.839
Are you gonna get more holes?
343
00:20:43.159 --> 00:20:52.840
So that's where you get, if you've ever heard of like P-type and N-type semiconductors, the N-type is the negative type semiconductor because the electrons are the ones that are moving around.
344
00:20:53.040 --> 00:20:57.561
The P-type is the positive type semiconductor because the holes are the ones that are moving around.
345
00:20:57.901 --> 00:21:01.422
And when you combine the two, you get things like LEDs, transistors,
346
00:21:02.122 --> 00:21:05.165
But at too high of a temperature, you're going to get wrecked.
347
00:21:05.485 --> 00:21:05.745
Why?
348
00:21:06.025 --> 00:21:11.630
Because if you have too high of a temperature, everything is going to go into the conduction band because there's so much jiggle.
349
00:21:12.130 --> 00:21:13.651
And then I'm just going to get a short circuit.
350
00:21:13.771 --> 00:21:15.153
It's always going to be conducting.
351
00:21:15.893 --> 00:21:20.177
With a transistor, I'd like to control when it's conducting current for a 1.0.
352
00:21:20.697 --> 00:21:31.165
And when it stops conducting current and becomes an insulator for a zero, but at a high enough temperature, it's always going to be conducting because there's so much jiggle that the electrons are just like, Oh, I'm just going to, I'm free to move around.
353
00:21:31.285 --> 00:21:31.945
Everything moves.
354
00:21:32.125 --> 00:21:32.566
Okay.
355
00:21:32.666 --> 00:21:37.829
And so part of what we're saying is we're trying to engineer a level of control of how this movement happens.
356
00:21:38.290 --> 00:21:46.216
And at high temperatures currently in our traditional, uh, chips and systems, they are unable at high temperatures.
357
00:21:46.256 --> 00:21:49.358
You lose the ability to engineer and control.
358
00:21:49.438 --> 00:21:49.598
Yeah.
359
00:21:49.638 --> 00:21:55.401
The jiggle, the movement, how things move to the conduction layer and back accordingly.
360
00:21:55.421 --> 00:21:58.323
Yeah, you can't control the transistor going from a 1 to a 0.
361
00:21:58.363 --> 00:21:59.464
It's just always going to be a 1.
362
00:21:59.624 --> 00:22:00.344
Right, right.
363
00:22:00.364 --> 00:22:03.566
Because it's always – Yeah, and that's the bedrock of computation.
364
00:22:03.586 --> 00:22:04.126
Right, right.
365
00:22:04.847 --> 00:22:13.512
Now, people have managed to make the band gap way bigger because if you make the band gap way bigger, then the temperature might not be enough to actually put it up.
366
00:22:14.412 --> 00:22:17.213
And then you can preserve this transistor property.
367
00:22:17.593 --> 00:22:23.614
But the authors of the paper that we're talking about, they cite this research from NASA Research.
368
00:22:25.354 --> 00:22:28.415
It shows that you can get all the way up to 800 degrees Celsius.
369
00:22:28.995 --> 00:22:32.296
But memory is still an issue.
370
00:22:32.536 --> 00:22:41.858
You've created a sort of circuit element here, digital integrated circuit, but non-volatile memory, something that retains data even when it's powered down,
371
00:22:42.439 --> 00:22:43.901
The stuff that's in our RAM.
372
00:22:44.361 --> 00:22:45.843
I turn off my computer.
373
00:22:45.883 --> 00:22:48.386
It's not like I'm going to lose everything.
374
00:22:48.846 --> 00:22:52.691
That kind of stuff is still not up to that scale.
375
00:22:53.371 --> 00:22:56.915
So through public funding, we discovered that it's possible to go above 800.
376
00:22:57.536 --> 00:23:03.943
But the issue still remains, memory does not, in this context, exist.
377
00:23:04.580 --> 00:23:21.528
have that high temperature limit exactly and if we want to do anything we got to store data right we got to store weights if we want to do a neural network we got to store what the algorithm is supposed to do if even if without the ai stuff right if it's just like instructions on what to do that is stored in your ram and if the ram fails
378
00:23:22.328 --> 00:23:24.369
But like your circuit is doing fine.
379
00:23:24.409 --> 00:23:25.109
What's it going to do?
380
00:23:25.149 --> 00:23:26.149
There's no instructions.
381
00:23:26.329 --> 00:23:26.510
Right.
382
00:23:26.690 --> 00:23:26.890
Right.
383
00:23:27.050 --> 00:23:27.590
Right.
384
00:23:27.790 --> 00:23:33.632
So most of the applications that we have like in our computer, in the phone, they use something called flash memory.
385
00:23:34.072 --> 00:23:35.593
This also uses a transistor.
386
00:23:35.613 --> 00:23:45.676
This is the MOSFET transistor where you combine a P and an N type and it runs into the same issue with the high temperature because you're using P and N types.
387
00:23:46.037 --> 00:23:46.857
Right.
388
00:23:47.417 --> 00:23:50.058
And if even at like a low temperature, like.
389
00:23:51.897 --> 00:24:01.054
room temperature right consumer electronics like the flash in our in our computers it wears off after about 10 000 to 100 000 read write cycles
390
00:24:01.933 --> 00:24:04.134
So they can't run forever.
391
00:24:04.574 --> 00:24:05.934
There's deterioration.
392
00:24:06.534 --> 00:24:09.615
Doing work deteriorates just like our own human body.
393
00:24:09.675 --> 00:24:09.915
Yeah.
394
00:24:10.215 --> 00:24:12.175
As an athlete, you can't be an athlete forever.
395
00:24:12.595 --> 00:24:13.776
Your body eventually gives up.
396
00:24:14.676 --> 00:24:21.517
Similarly, in this context, flash memory deteriorates at some time interval.
397
00:24:21.717 --> 00:24:22.137
Exactly.
398
00:24:22.177 --> 00:24:22.617
For use.
399
00:24:22.898 --> 00:24:23.298
Exactly.
400
00:24:23.578 --> 00:24:29.439
And so because of that, people have been slowly looking at other substrates of computation.
401
00:24:29.459 --> 00:24:29.739
Yeah.
402
00:24:29.919 --> 00:24:34.080
Specifically, something called resistive memory, R-RAM.
403
00:24:35.580 --> 00:24:37.441
This is resistive random access memory.
404
00:24:37.501 --> 00:24:42.622
And to understand that, we really have to understand something called the mem-rister.
405
00:24:43.642 --> 00:24:45.123
Let's go back to 1970.
406
00:24:45.563 --> 00:24:48.503
There are three types of fundamental circuit components.
407
00:24:49.044 --> 00:24:51.664
The resistor, the inductor, and the capacitor.
408
00:24:51.964 --> 00:24:54.485
This is the stuff of undergraduate physics.
409
00:24:54.945 --> 00:24:59.246
And if you're an electrical engineer, I'm really sorry for your loss.
410
00:24:59.906 --> 00:25:03.268
But you probably deal with this every single day, right?
411
00:25:04.009 --> 00:25:07.852
A resistor is something that resists electric current.
412
00:25:08.252 --> 00:25:10.754
So it drops a voltage across a resistor.
413
00:25:11.214 --> 00:25:19.880
And an inductor is something that takes a changing current and creates another current that's going in the opposite direction.
414
00:25:20.120 --> 00:25:23.142
So it kind of stores energy using a change in current.
415
00:25:23.663 --> 00:25:26.244
And a capacitor is something that stores energy using charge.
416
00:25:26.264 --> 00:25:26.765
Okay?
417
00:25:26.785 --> 00:25:27.005
Okay.
418
00:25:27.785 --> 00:25:36.629
Now, Leon Chua, in 1971, he was at UC Berkeley, he said, let's look at these three components that we've got.
419
00:25:36.710 --> 00:25:38.370
We've got a capacitor, right?
420
00:25:38.430 --> 00:25:43.953
A capacitor is something that has a relationship between charge and voltage.
421
00:25:44.774 --> 00:25:47.575
If I store a bunch of charge, that stores a voltage difference.
422
00:25:48.215 --> 00:25:49.156
I've got a resistor.
423
00:25:49.236 --> 00:25:52.477
That's a relationship between voltage and current, right?
424
00:25:52.658 --> 00:25:55.179
As current goes through, there's a voltage drop.
425
00:25:55.199 --> 00:25:55.499
Right?
426
00:25:56.688 --> 00:25:57.788
And I've got an inductor.
427
00:25:57.988 --> 00:26:01.990
That's something that has to do with current and magnetic flux.
428
00:26:02.790 --> 00:26:03.050
Okay?
429
00:26:05.031 --> 00:26:06.772
There's four variables, though.
430
00:26:07.272 --> 00:26:07.552
Right.
431
00:26:08.272 --> 00:26:09.833
And there's only three components.
432
00:26:10.597 --> 00:26:10.797
Right.
433
00:26:11.097 --> 00:26:14.879
So for those who may be audio listeners, we're looking at a square.
434
00:26:14.899 --> 00:26:18.902
It kind of looks like those, uh, those one, two, three, four.
435
00:26:18.922 --> 00:26:21.143
I don't know what you call it.
436
00:26:21.503 --> 00:26:21.883
Yeah.
437
00:26:21.943 --> 00:26:22.424
Yeah.
438
00:26:22.464 --> 00:26:24.205
The thing we used to do in elementary school.
439
00:26:24.245 --> 00:26:24.405
Yeah.
440
00:26:24.425 --> 00:26:26.026
That would tell me like what my fate was.
441
00:26:26.046 --> 00:26:26.326
Yeah.
442
00:26:26.346 --> 00:26:27.086
The paper thing.
443
00:26:27.126 --> 00:26:27.387
Yeah.
444
00:26:27.807 --> 00:26:28.767
You can let us know in the comments.
445
00:26:28.787 --> 00:26:29.028
Yeah.
446
00:26:29.188 --> 00:26:38.013
But on each of the four sides, you have exactly the components you mentioned and each of the four corners, you have the charge voltage current and flux and the,
447
00:26:38.713 --> 00:26:42.640
The point here being we've talked about every relationship but one side of the square.
448
00:26:42.660 --> 00:26:44.003
But one side of the square, right?
449
00:26:44.103 --> 00:26:46.146
There should be Leon Chua.
450
00:26:47.182 --> 00:26:52.406
decided there should be a relationship between flux and charge.
451
00:26:52.706 --> 00:26:53.767
The flux capacitor.
452
00:26:54.668 --> 00:26:56.309
In some sense, yeah.
453
00:26:56.569 --> 00:26:58.531
And he deemed this thing the memristor.
454
00:26:58.571 --> 00:27:00.372
He said there should be something out there, okay?
455
00:27:00.592 --> 00:27:04.875
This is a component where the resistance changes dynamically based on the charge flow.
456
00:27:05.556 --> 00:27:13.922
And he ruled that if there is such a relationship, what this thing is going to do is it's going to remember the state when the power is off.
457
00:27:13.942 --> 00:27:14.442
Okay?
458
00:27:14.462 --> 00:27:14.683
Okay.
459
00:27:15.523 --> 00:27:21.048
The telltale sign, he also showed, okay, if experimenters are looking at this thing, he was a theorist.
460
00:27:21.488 --> 00:27:26.132
And he said, if experimenters are looking at this thing, all you got to do is look at the IV curve.
461
00:27:26.672 --> 00:27:27.553
Okay?
462
00:27:29.715 --> 00:27:32.117
The IV curve is the current and the voltage.
463
00:27:32.277 --> 00:27:32.857
Current on the Y.
464
00:27:32.877 --> 00:27:33.458
On the two axes.
465
00:27:33.678 --> 00:27:34.518
Voltage on the X.
466
00:27:34.558 --> 00:27:35.119
Yes, okay?
467
00:27:35.299 --> 00:27:38.321
And if we look at the IV curves, they're all very...
468
00:27:39.243 --> 00:27:39.584
Unique.
469
00:27:40.184 --> 00:27:42.547
On the top left, that's a normal resistor.
470
00:27:42.828 --> 00:27:43.088
Okay.
471
00:27:43.348 --> 00:27:48.495
As I increase the voltage, I'm going to increase the current across the resistor.
472
00:27:49.015 --> 00:27:50.397
And there you've got a line.
473
00:27:50.457 --> 00:27:51.819
This is a nonlinear resistor.
474
00:27:52.079 --> 00:27:57.146
A linear resistor would be like the stuff that we do in undergrad physics is just like, oh, V equals IR.
475
00:27:57.426 --> 00:27:59.147
That's not entirely true, right?
476
00:27:59.187 --> 00:28:08.532
The R is a function of the V. So there's going to be some nonlinear resistance where at low voltage, the resistance is not that high.
477
00:28:08.552 --> 00:28:10.714
And then at high voltage, the resistance is way higher.
478
00:28:11.094 --> 00:28:11.594
Things like that.
479
00:28:12.175 --> 00:28:12.395
Yeah.
480
00:28:13.516 --> 00:28:14.437
No, it's the other way around.
481
00:28:14.537 --> 00:28:18.741
At a high voltage, the resistance is lower because there's more pull.
482
00:28:18.761 --> 00:28:21.744
So it's easier to shuttle electrons through.
483
00:28:21.945 --> 00:28:24.227
Okay, so that's what we're seeing in the top left corner.
484
00:28:24.627 --> 00:28:30.874
On the top right corner and the bottom left corner, we're seeing the IV curves for a capacitor and an inductor.
485
00:28:31.655 --> 00:28:32.956
Those things cycle around.
486
00:28:34.257 --> 00:28:45.483
And this is something that everyone knows from, you know, if you've ever done RC circuits, you'll know that there's an oscillation that happens in RC and RL circuits, especially.
487
00:28:45.523 --> 00:28:50.926
If you combine a capacitor and an inductor, the capacitor gets charged and the inductor doesn't get charged.
488
00:28:51.187 --> 00:28:53.708
And then the inductor gets charged, the capacitor gets discharged.
489
00:28:53.908 --> 00:28:55.428
So there's a cycling that happens.
490
00:28:55.669 --> 00:28:59.410
And that is something that is characteristic of like sine and cosine.
491
00:28:59.430 --> 00:29:00.650
You got the unit circle.
492
00:29:00.950 --> 00:29:02.110
The x-axis is sine.
493
00:29:02.130 --> 00:29:03.451
The y-axis is cosine.
494
00:29:03.571 --> 00:29:03.951
Great.
495
00:29:04.991 --> 00:29:15.774
Leon Chua said, I should be looking for something called a memristor where charge and magnetic flux are related to one another.
496
00:29:16.794 --> 00:29:23.496
And what I would see is a curve that looks like a loop, a figure eight loop.
497
00:29:24.940 --> 00:29:32.963
And the challenge is there is no way to combine the other three curves to get that fourth curve.
498
00:29:33.623 --> 00:29:34.043
Interesting.
499
00:29:34.063 --> 00:29:34.283
Okay?
500
00:29:34.603 --> 00:29:38.685
It is an independent element in electronics.
501
00:29:38.785 --> 00:29:39.665
It's fundamental.
502
00:29:39.705 --> 00:29:40.526
It's not derivative.
503
00:29:40.546 --> 00:29:41.346
Yes, exactly.
504
00:29:41.366 --> 00:29:41.686
You know what?
505
00:29:42.146 --> 00:29:42.866
You see what I'm saying?
506
00:29:42.886 --> 00:29:43.587
Yeah.
507
00:29:43.647 --> 00:29:47.128
It's like you can't make an electron out of quarks.
508
00:29:47.825 --> 00:29:48.025
Right.
509
00:29:48.065 --> 00:29:48.246
Right.
510
00:29:48.346 --> 00:29:55.035
Similarly, you can't make a memristor by just cleverly placing resistors, capacitors, and inductors together.
511
00:29:55.588 --> 00:30:00.831
This is an independent component that is one of the four fundamental components of electronics.
512
00:30:01.091 --> 00:30:12.877
And this sort of goes back to our previous visual, which, you know, in this context of those four parts, you needed, this is sort of the missing piece to the puzzle of the fundamental elements.
513
00:30:12.897 --> 00:30:13.117
Yeah.
514
00:30:13.397 --> 00:30:17.700
And it's just a mathematical argument, which is so cool to think about.
515
00:30:17.780 --> 00:30:18.000
Right.
516
00:30:18.140 --> 00:30:18.320
Right.
517
00:30:18.380 --> 00:30:20.421
That it's just like there's four variables.
518
00:30:20.681 --> 00:30:22.322
We've only got three components.
519
00:30:22.382 --> 00:30:23.403
There should be a fourth one.
520
00:30:23.903 --> 00:30:26.224
Theorists have value, everybody, okay?
521
00:30:26.304 --> 00:30:26.864
It's very cool.
522
00:30:26.904 --> 00:30:30.325
And one more thing I want to say about the Memorister IV curve, right?
523
00:30:31.825 --> 00:30:40.087
Notice that for a capacitor and an inductor, at zero voltage, there is a non-zero current, okay?
524
00:30:40.467 --> 00:30:49.909
That means it can't store data because if I turn the thing off, there's going to be current and that's going to dissipate whatever the stuff was happening, right?
525
00:30:50.329 --> 00:30:51.950
A resistor could store data
526
00:30:52.949 --> 00:30:57.132
Because at a zero voltage, there's zero current, but it only stores a one.
527
00:30:58.152 --> 00:30:59.473
It's whatever the resistance is.
528
00:30:59.533 --> 00:31:01.094
So it's not really storing anything.
529
00:31:01.354 --> 00:31:06.097
If we go back to Claude Shannon's information theory, I need two things to store data.
530
00:31:06.137 --> 00:31:07.318
I need a zero and a one.
531
00:31:07.678 --> 00:31:11.420
A memory store, on the other hand, at zero voltage, it can be in two states.
532
00:31:11.900 --> 00:31:13.842
It can be in a low resistance state.
533
00:31:13.862 --> 00:31:14.342
Right.
534
00:31:14.542 --> 00:31:15.623
or a high resistance state.
535
00:31:15.643 --> 00:31:16.143
This is so good.
536
00:31:16.163 --> 00:31:16.623
Right?
537
00:31:16.643 --> 00:31:17.564
This is so good.
538
00:31:17.804 --> 00:31:24.307
Now all of a sudden people start paying attention because this could be a substrate for non-volatile memory.
539
00:31:24.567 --> 00:31:32.411
Meaning I turned the thing off, but depending on the history of where I was on the curve, I could be in either a low resistance state or a high resistance state.
540
00:31:32.611 --> 00:31:34.552
If I can sense that, that's my zero and one.
541
00:31:35.312 --> 00:31:39.837
computers as a quick visual analogy here, because we're basically looking at this figure eight.
542
00:31:39.857 --> 00:31:40.838
That's kind of like a racetrack.
543
00:31:40.858 --> 00:31:42.500
You remember Hot Wheels as a kid?
544
00:31:42.540 --> 00:31:42.740
Yeah.
545
00:31:42.900 --> 00:31:49.487
So you remember how you could have the, you can have a figure eight like that, but one side would be going under and it would come back and it'd be going over.
546
00:31:49.827 --> 00:31:52.670
And that's kind of what we're saying where it can be in two states at that center point.
547
00:31:52.690 --> 00:31:54.953
It can be on the top part of the Hot Wheels track or,
548
00:31:55.173 --> 00:31:57.255
or on the bottom part of the Hot Wheels track.
549
00:31:57.315 --> 00:32:04.842
And that mechanism is how we're storing the zero and one, how we're able to have two states, which means we can have a zero and we can have a one.
550
00:32:05.002 --> 00:32:05.682
Exactly, yeah.
551
00:32:05.923 --> 00:32:11.347
And for the Formula One fans, the only circuit that does this is Suzuka in Japan.
552
00:32:11.848 --> 00:32:13.629
There's a figure eight that's beautiful.
553
00:32:13.850 --> 00:32:14.370
One day we'll go.
554
00:32:15.551 --> 00:32:17.893
So the theorist has done his job.
555
00:32:17.913 --> 00:32:19.995
He's like, there should be this thing.
556
00:32:20.415 --> 00:32:22.237
And specifically, this is what you guys should look for.
557
00:32:24.900 --> 00:32:26.000
It takes 40 years.
558
00:32:26.400 --> 00:32:26.941
Oh, my God.
559
00:32:27.421 --> 00:32:28.381
Nearly 40 years.
560
00:32:29.221 --> 00:32:31.842
Stanley Williams at Hewlett Packard Labs.
561
00:32:32.022 --> 00:32:33.162
So that's HP Labs.
562
00:32:34.642 --> 00:32:42.504
I love these nature articles that are just like bombshell articles because the title is just a banger every single time.
563
00:32:42.804 --> 00:32:45.204
The missing membrists are found.
564
00:32:45.405 --> 00:32:51.646
It's nearly 40 years after Leon Chua postulated that this fourth passive element should exist.
565
00:32:51.866 --> 00:32:52.166
Yeah.
566
00:32:54.073 --> 00:33:05.817
Stanley Williams announces it in Nature, and he shows that there is a memorister that he's created using titanium dioxide and platinum electrodes, and it uses something called oxygen vacancy migration.
567
00:33:06.177 --> 00:33:08.978
I just want to note that we're talking about 2008.
568
00:33:08.998 --> 00:33:09.378
Yeah.
569
00:33:10.738 --> 00:33:11.198
Very soon.
570
00:33:11.278 --> 00:33:12.479
We were in high school.
571
00:33:12.579 --> 00:33:12.859
Mm-hmm.
572
00:33:13.713 --> 00:33:15.033
So it wasn't that crazy.
573
00:33:15.093 --> 00:33:16.114
It wasn't that long ago.
574
00:33:16.174 --> 00:33:17.054
It wasn't that long ago.
575
00:33:17.134 --> 00:33:22.155
No, we were in high school when the first Memrister, like when it went from theory to reality.
576
00:33:22.215 --> 00:33:23.036
Right.
577
00:33:23.096 --> 00:33:26.336
And it's still at a research stage level.
578
00:33:26.356 --> 00:33:29.337
We're not talking about in-end consumer products.
579
00:33:29.357 --> 00:33:29.577
Yeah.
580
00:33:30.157 --> 00:33:31.858
The iPhone came out a year before this.
581
00:33:32.711 --> 00:33:34.191
Yeah, that's crazy.
582
00:33:34.251 --> 00:33:36.012
Just to put it in time context.
583
00:33:36.492 --> 00:33:37.132
Wow.
584
00:33:37.172 --> 00:33:40.052
Which is... That's a good way to... Yeah, that's a good way to think about it.
585
00:33:40.112 --> 00:33:40.292
Yeah.
586
00:33:40.632 --> 00:33:43.633
I mean, we had an iPhone before we even knew that a Memorister was possible.
587
00:33:44.033 --> 00:33:45.333
Which is so crazy.
588
00:33:45.653 --> 00:33:46.233
That's crazy.
589
00:33:46.293 --> 00:33:47.434
Knowing what we use it for now.
590
00:33:47.594 --> 00:33:47.834
Yeah.
591
00:33:48.254 --> 00:33:48.474
Yeah.
592
00:33:48.614 --> 00:33:49.434
I mean, it's insane.
593
00:33:49.494 --> 00:33:49.694
Okay.
594
00:33:49.954 --> 00:33:54.715
So let's talk about this valence charge mechanism that I was telling you about, about how the Memorister works.
595
00:33:55.035 --> 00:33:55.655
Here's the idea.
596
00:33:56.575 --> 00:33:57.556
So I told you, right?
597
00:33:57.596 --> 00:33:59.916
He got a titanium electrode.
598
00:34:00.256 --> 00:34:01.197
A platinum electrode.
599
00:34:01.618 --> 00:34:05.582
And in the middle were oxygen vacancies.
600
00:34:05.762 --> 00:34:09.426
Now, vacancies is something that I alluded to earlier when I was talking about the holes.
601
00:34:09.986 --> 00:34:13.970
The holes are vacancies in the sense that I have taken something out.
602
00:34:14.591 --> 00:34:18.575
But now those holes where I took that thing out, those are mobile vacancies.
603
00:34:18.925 --> 00:34:21.306
just like the stuff that I took out.
604
00:34:21.406 --> 00:34:22.466
Those are mobile elements.
605
00:34:22.767 --> 00:34:29.329
And if those mobile elements can rearrange, they can create electronic components on their own, right?
606
00:34:29.369 --> 00:34:30.510
In the same way that a P-type
607
00:34:31.500 --> 00:34:36.703
Semiconductor can use the holes to do all of its electronics.
608
00:34:36.763 --> 00:34:40.885
The point being that vacancy is not stationary.
609
00:34:40.965 --> 00:34:44.286
It has the ability independent of the thing that left to move on its own.
610
00:34:44.546 --> 00:34:45.467
Yeah, exactly.
611
00:34:45.607 --> 00:34:45.867
Okay.
612
00:34:46.227 --> 00:34:52.030
So let's get into a little bit more about how these oxygen vacancies work because this is going to be the central –
613
00:34:53.445 --> 00:34:57.589
thing we need to understand about why the physics of this modern paper is working.
614
00:34:57.829 --> 00:34:58.030
Got it.
615
00:34:58.050 --> 00:34:59.511
Okay?
616
00:35:00.632 --> 00:35:07.439
When you have these two electrodes, you've got a positive electrode and a negative electrode, right?
617
00:35:08.360 --> 00:35:10.702
What you do is a switching process, okay?
618
00:35:11.222 --> 00:35:14.606
So this is effectively how it works.
619
00:35:15.940 --> 00:35:18.243
Let's say I've got a top electrode and a bottom electrode.
620
00:35:18.343 --> 00:35:22.848
In this case, it's like silver and titanium on the bottom.
621
00:35:23.108 --> 00:35:26.452
But it doesn't really matter what the identity of the two electrodes are yet.
622
00:35:27.073 --> 00:35:27.914
It's going to matter later.
623
00:35:29.389 --> 00:35:35.875
What happens is, in order to understand what this valence charge mechanism is, let's try to think of an analogy, a simple analogy.
624
00:35:36.235 --> 00:35:36.476
Okay?
625
00:35:36.816 --> 00:35:46.985
Imagine like a really thick, densely packed forest, like the forests of Endor in the Star Wars Episode VI Return of the Jedi.
626
00:35:47.046 --> 00:35:47.506
Okay.
627
00:35:47.766 --> 00:35:48.046
Okay?
628
00:35:48.107 --> 00:35:48.587
Okay.
629
00:35:48.867 --> 00:35:53.309
Now, if you remember, the Forest of Endor, they had those motorbike thingies.
630
00:35:54.529 --> 00:35:56.150
I actually looked up what they were called.
631
00:35:56.170 --> 00:35:58.850
They were called forest motorbikes.
632
00:35:58.911 --> 00:36:00.251
74Z speeder bikes.
633
00:36:00.571 --> 00:36:00.851
Okay?
634
00:36:01.691 --> 00:36:06.493
Remember, they zoomed through the forest at really, really high speed.
635
00:36:07.073 --> 00:36:09.114
Now, imagine a really, really dense forest.
636
00:36:09.494 --> 00:36:09.774
Okay?
637
00:36:10.134 --> 00:36:14.356
If you've got a speeder bike in a really, really dense forest, you're going to crash.
638
00:36:14.776 --> 00:36:16.737
You have to be a Lewis Hamilton.
639
00:36:16.817 --> 00:36:17.477
Yes, yes.
640
00:36:17.497 --> 00:36:17.677
Yes.
641
00:36:18.177 --> 00:36:24.285
And even then, you know, like I'm talking about like there's a tree literally everywhere.
642
00:36:24.525 --> 00:36:25.626
It's not feasible.
643
00:36:25.666 --> 00:36:26.327
It's not going to work.
644
00:36:26.367 --> 00:36:26.948
It's not going to work.
645
00:36:27.328 --> 00:36:33.416
So what you could do instead is have a bulldozer come in or let's say a really high wind.
646
00:36:33.776 --> 00:36:35.499
And that's going to take down some of the trees.
647
00:36:36.300 --> 00:36:36.500
Yeah.
648
00:36:36.520 --> 00:36:36.660
Yeah.
649
00:36:37.599 --> 00:36:47.329
When you take down some of the trees, that's kind of like creating oxygen vacancies in something like a transition metal oxide like hafnium oxide or titanium dioxide.
650
00:36:48.050 --> 00:36:52.735
These are transition metal oxides that are metals bonded to oxygen.
651
00:36:53.356 --> 00:36:55.839
That's in the middle, sandwiched in between the two electrodes.
652
00:36:55.879 --> 00:36:56.059
Right?
653
00:36:57.230 --> 00:37:04.894
If I've got an electric field, the oxygen, which has these like, it's just got these properties where it wants to go towards the negative electrode, right?
654
00:37:04.914 --> 00:37:08.536
This is why oxygen is so reactive and why it rusts everything.
655
00:37:10.037 --> 00:37:11.518
The oxygen is going to start moving.
656
00:37:12.398 --> 00:37:14.419
When it starts moving, it's going to leave behind holes.
657
00:37:15.340 --> 00:37:15.540
Okay?
658
00:37:15.700 --> 00:37:18.702
So it's kind of like the trees are coming down and now I've got a pathway.
659
00:37:19.202 --> 00:37:20.343
Now my motorbike...
660
00:37:21.206 --> 00:37:22.207
Can move really fast.
661
00:37:22.247 --> 00:37:25.628
The motorbike in this analogy represents the electron.
662
00:37:26.169 --> 00:37:28.910
The electron is piggybacking on these vacancies.
663
00:37:29.610 --> 00:37:42.597
And part of your point is being until the wind or the bulldozer comes and knocks the trees over, which is the oxygen leaving, the electrons can't move through.
664
00:37:42.717 --> 00:37:42.917
Yeah.
665
00:37:43.137 --> 00:37:44.858
So that would be a high resistance state.
666
00:37:45.675 --> 00:37:45.875
Right.
667
00:37:46.096 --> 00:37:49.440
That's one of the states that we were thinking a memorister should have.
668
00:37:49.720 --> 00:37:50.621
It's got a memory, right?
669
00:37:50.641 --> 00:37:51.382
There was no wind.
670
00:37:51.843 --> 00:37:53.224
And so I am at a high resistance.
671
00:37:53.565 --> 00:37:55.367
Now, all of a sudden, I applied an electric field.
672
00:37:55.527 --> 00:37:56.589
I got a low resistance.
673
00:37:56.929 --> 00:37:57.770
Can I go backwards?
674
00:37:57.790 --> 00:37:57.970
Yeah.
675
00:37:58.211 --> 00:38:00.373
Just apply the electric field in the other direction.
676
00:38:00.954 --> 00:38:01.815
The trees stand up.
677
00:38:02.857 --> 00:38:03.577
Low resistance.
678
00:38:03.617 --> 00:38:03.877
Got it.
679
00:38:04.217 --> 00:38:04.497
Got it.
680
00:38:04.857 --> 00:38:05.218
Right?
681
00:38:05.258 --> 00:38:05.398
Yes.
682
00:38:05.418 --> 00:38:11.999
This is the mechanism behind the switching behavior, the set and the reset that we think of in memoristers.
683
00:38:12.019 --> 00:38:13.099
That's the memorister jargon.
684
00:38:13.139 --> 00:38:17.360
It's like you set it to the low resistance state, so then there's like conduction happening.
685
00:38:17.840 --> 00:38:19.200
So that's your one, let's say.
686
00:38:19.380 --> 00:38:19.541
Yes.
687
00:38:19.601 --> 00:38:26.802
And then you reset it back to the high resistance state, so then the trees are all up, the vacancies are back, the oxygen is back, I mean.
688
00:38:27.062 --> 00:38:30.883
There's no more vacancies, and the electrons don't have a highway to go through.
689
00:38:30.903 --> 00:38:31.303
Yes.
690
00:38:31.563 --> 00:38:39.033
And what's interesting about this is you can go, you have the ability to return back and forth in between each of these states.
691
00:38:39.053 --> 00:38:39.454
Exactly.
692
00:38:39.474 --> 00:38:40.495
So now you've got a cycling.
693
00:38:40.675 --> 00:38:40.896
Right.
694
00:38:41.497 --> 00:38:43.940
And that's something that you need for non-volatile memory.
695
00:38:44.040 --> 00:38:44.220
Right.
696
00:38:44.461 --> 00:38:44.701
Okay.
697
00:38:45.161 --> 00:38:48.947
That's effectively what this vacancy mechanism is.
698
00:38:50.048 --> 00:38:50.308
Got it?
699
00:38:50.848 --> 00:38:54.191
Now, a lot of times now we're going to try and use hafnium oxide.
700
00:38:54.792 --> 00:39:02.438
The advantage here is that hafnium oxide is something that silicon chip makers have already been using in transistor gate stacks.
701
00:39:03.139 --> 00:39:05.781
It's already replaced the silicon dioxide.
702
00:39:07.215 --> 00:39:13.957
transition metal oxide in the famous high K dielectric constant metal gate transition that happened in like 2007.
703
00:39:15.037 --> 00:39:17.237
I wasn't aware of this, obviously, because I was in school.
704
00:39:17.497 --> 00:39:25.059
But in 2007, like there was sort of a plateau on Moore's law because we couldn't get stuff smaller.
705
00:39:25.259 --> 00:39:30.240
And then and then somebody decided, hey, actually, we can use halfnium oxide that reduces power leakage.
706
00:39:30.921 --> 00:39:32.441
And it allows these transitions.
707
00:39:33.401 --> 00:39:36.502
It allows these transistors to shrink even further.
708
00:39:37.002 --> 00:39:39.265
without having a significant decrease in performance.
709
00:39:39.526 --> 00:39:43.111
The smaller you get, right, the harder it is to keep the same level of performance.
710
00:39:43.151 --> 00:39:45.374
Yeah, because you're trying to dissipate heat.
711
00:39:45.434 --> 00:39:49.079
The smaller you get now, the heat is starting to mess with the components themselves.
712
00:39:49.420 --> 00:39:52.384
And so this hafnium oxide, it's CMOS compatible.
713
00:39:52.804 --> 00:39:54.365
It scales to very small dimensions.
714
00:39:54.786 --> 00:39:55.686
It switches fast.
715
00:39:56.127 --> 00:39:57.988
And it shows really good endurance, right?
716
00:39:58.028 --> 00:40:08.977
And by the mid-2010s, all the major semiconductor companies, that's like Samsung, Micron, TSMC, they all have active Memristor R-RAM development programs.
717
00:40:09.037 --> 00:40:09.717
And they still do.
718
00:40:10.198 --> 00:40:12.760
And just to make a quick clarification point.
719
00:40:12.840 --> 00:40:20.746
So chip makers were already using this for other chips prior to using it in Memristors.
720
00:40:21.366 --> 00:40:28.091
And so part of it was like they had the industrial scale working with this particular material.
721
00:40:28.552 --> 00:40:35.717
And so they were like, it would be ideal if we could use this same thing we already do at scale in another chip type.
722
00:40:35.757 --> 00:40:36.318
Exactly.
723
00:40:36.358 --> 00:40:36.758
Is that fair?
724
00:40:36.778 --> 00:40:37.539
That's exactly right.
725
00:40:37.599 --> 00:40:44.744
And so the original paper that was by Stanley Williams, that one used titanium dioxide as the transition metal oxide.
726
00:40:44.764 --> 00:40:45.165
Yeah.
727
00:40:45.785 --> 00:40:51.814
People have rapidly switched now to hafnium oxide as the sort of sandwich, the thing in between the sandwich.
728
00:40:51.894 --> 00:40:52.114
Okay.
729
00:40:53.256 --> 00:40:55.459
Now, we've got hafnium oxide.
730
00:40:55.540 --> 00:40:55.780
Okay.
731
00:40:55.920 --> 00:40:57.623
That's the meat in between the sandwich.
732
00:40:58.063 --> 00:40:59.686
What do we want to use for the electrodes?
733
00:41:00.877 --> 00:41:02.659
Usually we use tungsten and platinum.
734
00:41:02.919 --> 00:41:03.559
Why tungsten?
735
00:41:03.800 --> 00:41:07.122
Well, tungsten has a very high affinity for oxygen.
736
00:41:07.723 --> 00:41:08.283
So that's good.
737
00:41:08.604 --> 00:41:10.045
Just trying to get the oxygen.
738
00:41:11.086 --> 00:41:15.790
And it's also really good at withstanding high temperatures.
739
00:41:16.090 --> 00:41:20.674
That's why the Thomas Edison light bulbs use a tungsten filament.
740
00:41:20.734 --> 00:41:23.596
And here we've got a zoom in of one of those filaments.
741
00:41:23.956 --> 00:41:25.778
It's made out of tungsten because it's not going to melt.
742
00:41:26.438 --> 00:41:26.679
Okay?
743
00:41:27.587 --> 00:41:28.547
So that makes sense.
744
00:41:29.167 --> 00:41:32.608
The basic idea now to make this memorister is the following.
745
00:41:32.648 --> 00:41:33.749
We've got a top electrode.
746
00:41:34.189 --> 00:41:35.329
We've got a bottom electrode.
747
00:41:35.949 --> 00:41:38.650
Tungsten, half-nium oxide, something else.
748
00:41:39.930 --> 00:41:49.333
And we use the oxygen vacancies in the half-nium oxide to transport electrons and turn that highway on or off.
749
00:41:50.102 --> 00:41:50.362
Okay?
750
00:41:50.883 --> 00:41:52.144
That's effectively the idea.
751
00:41:52.725 --> 00:42:03.475
And I'm guessing the idea then here is whatever is on the opposite side of the tungsten, depending on what it is, changes the dynamics of the back and forth.
752
00:42:03.675 --> 00:42:08.359
And so there's fertile ground to optimize for different use cases.
753
00:42:08.379 --> 00:42:09.440
That is exactly right.
754
00:42:09.520 --> 00:42:10.861
And let's stay here.
755
00:42:10.982 --> 00:42:11.262
Okay?
756
00:42:11.602 --> 00:42:11.882
So...
757
00:42:12.923 --> 00:42:13.504
You called it.
758
00:42:13.924 --> 00:42:14.865
The top is tungsten.
759
00:42:15.366 --> 00:42:17.628
The meat in the middle is half neem oxide.
760
00:42:17.968 --> 00:42:19.049
What should go on the bottom?
761
00:42:19.390 --> 00:42:19.590
Okay.
762
00:42:19.830 --> 00:42:23.353
For a long time, people have been talking about platinum.
763
00:42:24.374 --> 00:42:26.276
Usually you want some kind of inert metal.
764
00:42:26.476 --> 00:42:29.279
It's a metal that doesn't like react with stuff, right?
765
00:42:29.299 --> 00:42:35.565
Because obviously if it reacts, that's going to cause performance degradation.
766
00:42:36.045 --> 00:42:37.487
So platinum is a pretty good metal.
767
00:42:38.858 --> 00:42:41.839
This is where the valence charge mechanism takes over, right?
768
00:42:41.879 --> 00:42:50.163
The problem is when you pump up the temperature, at high temperature, the tungsten starts diffusing.
769
00:42:51.388 --> 00:42:51.609
Right?
770
00:42:51.669 --> 00:42:59.093
If you get into a high enough temperature, the tungsten now starts jiggling around, and the tungsten starts exploring the forest.
771
00:43:00.014 --> 00:43:00.754
And we don't want that.
772
00:43:00.795 --> 00:43:01.395
We don't want that.
773
00:43:01.435 --> 00:43:02.576
We just want the oxygen.
774
00:43:02.736 --> 00:43:06.959
We want only the oxygen to move around, create these vacancies, and then there's a highway.
775
00:43:07.239 --> 00:43:13.443
But if the tungsten starts moving around, all of a sudden maybe you've got filaments of tungsten and maybe filaments of platinum that are coming together.
776
00:43:13.703 --> 00:43:16.745
And if those two meet, you've got a shorted circuit.
777
00:43:16.765 --> 00:43:16.925
Yeah.
778
00:43:18.490 --> 00:43:24.794
Now, there's no way to, I mean, even if you cut down all the trees, you've just paved a highway.
779
00:43:25.074 --> 00:43:26.355
For that, yeah, yeah, yeah.
780
00:43:26.795 --> 00:43:28.376
And there's no, it can no longer cycle.
781
00:43:28.536 --> 00:43:29.036
Yeah, yeah.
782
00:43:29.156 --> 00:43:30.457
Why would anyone take the trees?
783
00:43:30.877 --> 00:43:31.998
You just go through the highway.
784
00:43:32.158 --> 00:43:32.318
Right.
785
00:43:32.558 --> 00:43:32.758
Right.
786
00:43:32.879 --> 00:43:34.840
You go through the tungsten and the platinum filaments.
787
00:43:35.420 --> 00:43:36.781
And that's been the big problem.
788
00:43:36.881 --> 00:43:37.161
Okay.
789
00:43:37.301 --> 00:43:37.481
Right.
790
00:43:37.521 --> 00:43:41.464
They mix in ways that destroy this precise nanoscale structure.
791
00:43:41.824 --> 00:43:43.625
Platinum and tungsten is not the way.
792
00:43:44.045 --> 00:43:44.305
Okay?
793
00:43:44.546 --> 00:43:47.069
You're going to effectively short-circuit at high temperatures.
794
00:43:47.369 --> 00:43:48.971
At low temperatures, this thing is working.
795
00:43:48.991 --> 00:43:49.151
Great.
796
00:43:49.712 --> 00:43:49.912
Right?
797
00:43:50.173 --> 00:43:58.082
But you start pumping this thing up to 500, 600, 700 degrees Celsius, all of a sudden, the metal atoms...
798
00:43:58.872 --> 00:44:01.134
in the two electrodes, they start getting agitated.
799
00:44:01.875 --> 00:44:12.784
This, this is giving me, uh, a feeling similar to the hypersonics episode that we talked about where Navier Stokes gets weird at higher temperatures.
800
00:44:12.884 --> 00:44:16.747
And so there's different regimes depending on the temperature band you're working in.
801
00:44:16.767 --> 00:44:16.967
Yeah.
802
00:44:17.327 --> 00:44:17.968
And, yeah,
803
00:44:18.588 --> 00:44:21.449
I'm just making the connection between these two things.
804
00:44:21.469 --> 00:44:23.170
Different physics starts mattering, right?
805
00:44:23.250 --> 00:44:24.211
It starts mattering.
806
00:44:24.231 --> 00:44:29.813
It's like the titanium to titanium bond is now smaller than the thermal noise.
807
00:44:30.273 --> 00:44:33.414
And so the titanium is like, all right, I'm going to dip out and let's start exploring.
808
00:44:33.594 --> 00:44:33.895
I don't know.
809
00:44:33.915 --> 00:44:34.815
I'm going to go this way.
810
00:44:35.435 --> 00:44:36.796
And the platinum is like, hey, hey, hey.
811
00:44:37.476 --> 00:44:37.856
You know?
812
00:44:37.917 --> 00:44:38.217
Yeah.
813
00:44:38.317 --> 00:44:46.163
And the structure, the way you've kind of laid it out now, that that makes sense in terms of, okay, we've discovered the fourth side, which is the memristor.
814
00:44:46.703 --> 00:44:55.590
We've figured out ways in which, because of the way the industry was at the time, what kind of works and is good for most of the use cases we thought we needed.
815
00:44:56.250 --> 00:44:59.413
But then as we got to higher temperatures, physics gets different.
816
00:45:00.033 --> 00:45:04.937
And the current regime of tungsten and platinum is...
817
00:45:05.357 --> 00:45:05.798
Doesn't work.
818
00:45:05.818 --> 00:45:06.358
Doesn't work.
819
00:45:06.418 --> 00:45:07.279
Yes, exactly.
820
00:45:07.580 --> 00:45:08.881
And now we get to the paper.
821
00:45:09.141 --> 00:45:10.463
Okay, this was out in science.
822
00:45:12.265 --> 00:45:15.068
Shout out, AAAS.
823
00:45:16.089 --> 00:45:20.033
High temperature memristors enabled by interfacial engineering.
824
00:45:20.053 --> 00:45:20.153
Okay.
825
00:45:21.464 --> 00:45:26.607
One of the co-authors actually is Stanley Williams, who's the original Memristors guy.
826
00:45:26.727 --> 00:45:29.428
The nature for – oh my god.
827
00:45:29.448 --> 00:45:30.108
What a legend.
828
00:45:30.248 --> 00:45:32.650
I thought that was a nice throughput.
829
00:45:32.870 --> 00:45:34.450
What a legend.
830
00:45:34.811 --> 00:45:38.552
So Joshua Yang, who's at USC, he worked with Stanley at HP Labs.
831
00:45:38.953 --> 00:45:42.694
And then now he's a professor of electrical engineering at USC.
832
00:45:43.055 --> 00:45:47.697
So this is a collaboration between, as you said, USC, Air Force Research, and Kumamoto.
833
00:45:47.717 --> 00:45:47.857
Yeah.
834
00:45:48.137 --> 00:45:53.681
But it's like the same lab that created the Memristors is now just upping the game.
835
00:45:53.721 --> 00:45:54.241
We're the best.
836
00:45:54.301 --> 00:45:55.022
What do you want us to do?
837
00:45:55.082 --> 00:45:55.782
Which is kind of cool.
838
00:45:56.062 --> 00:46:00.125
Their big idea is to replace that bottom inert electrode...
839
00:46:01.479 --> 00:46:03.800
Instead of platinum, we're going to use graphene.
840
00:46:04.081 --> 00:46:04.181
Okay.
841
00:46:04.201 --> 00:46:04.441
Okay.
842
00:46:04.561 --> 00:46:06.562
So this is the one that we need to be more stable.
843
00:46:06.622 --> 00:46:06.842
Yeah.
844
00:46:06.882 --> 00:46:07.963
Doesn't want to do stuff.
845
00:46:08.043 --> 00:46:08.283
Yeah.
846
00:46:08.383 --> 00:46:10.284
I want it to not talk to the titanium.
847
00:46:10.424 --> 00:46:10.784
Or sorry.
848
00:46:11.085 --> 00:46:13.366
I want it to not talk to the titanium.
849
00:46:13.826 --> 00:46:15.007
I want them to not be friends.
850
00:46:15.107 --> 00:46:15.307
Yes.
851
00:46:15.407 --> 00:46:17.929
Platinum and titanium, they're friends at high temperature.
852
00:46:18.549 --> 00:46:22.053
Turns out graphene, even at high temperature, it's fine on its own.
853
00:46:22.213 --> 00:46:22.593
I'm good.
854
00:46:22.713 --> 00:46:23.374
Yeah, it's good.
855
00:46:23.714 --> 00:46:25.456
And for very good reasons.
856
00:46:25.496 --> 00:46:27.999
So graphene, it won the Nobel Prize in 2010.
857
00:46:28.619 --> 00:46:34.986
It was isolated in 2004 by Andre Gaim and Konstantin Noyoso-Levonov.
858
00:46:36.457 --> 00:46:37.498
Noyo Salove.
859
00:46:37.778 --> 00:46:38.839
Noyo Salove.
860
00:46:38.859 --> 00:46:39.099
Okay?
861
00:46:39.460 --> 00:46:41.061
They were in the UK.
862
00:46:42.042 --> 00:46:49.689
It's a very funny story where they actually, they took graphite, which is the stuff in pencil lead.
863
00:46:50.029 --> 00:46:51.531
They put it on pieces of paper.
864
00:46:51.671 --> 00:46:52.872
And then they got scotch tape.
865
00:46:54.153 --> 00:46:55.794
Put it on the piece of paper and removed it.
866
00:46:56.455 --> 00:46:57.256
And the...
867
00:46:57.842 --> 00:47:05.507
Physics between the scotch tape and the graphite that's on the piece of paper would create single layers of graphite, right?
868
00:47:05.587 --> 00:47:08.708
Single layers of carbon hexagonal shells.
869
00:47:08.788 --> 00:47:11.510
Graphite is a bunch of layers stacked together.
870
00:47:12.471 --> 00:47:14.752
Graphene is a single layer, okay?
871
00:47:16.548 --> 00:47:21.029
It's an incredible story because it was scotch tape and a pencil, and it won them the Nobel Prize.
872
00:47:21.429 --> 00:47:30.291
Sometimes curiosity and just trying stuff is the key to breakthroughs that will last a lifetime, as we, again, have discussed multiple times on the show.
873
00:47:30.311 --> 00:47:34.692
Yeah, and I mean, since then, we've done loads of research on how to make graphene.
874
00:47:34.752 --> 00:47:40.433
Obviously, like, if you're trying to do industrial scale, you're not going to be doing scotch tape and pencil, right?
875
00:47:40.493 --> 00:47:41.133
But you wouldn't have known.
876
00:47:41.253 --> 00:47:42.193
But you wouldn't have known, right?
877
00:47:42.273 --> 00:47:44.534
And now people do chemical vapor deposition.
878
00:47:44.554 --> 00:47:45.094
Yeah.
879
00:47:45.434 --> 00:47:47.115
to make graphene.
880
00:47:47.516 --> 00:47:50.438
Graphene has extraordinary properties, okay?
881
00:47:50.878 --> 00:47:53.200
It's the strongest material ever measured.
882
00:47:54.921 --> 00:47:56.922
It's an excellent electrical conductor.
883
00:47:57.883 --> 00:47:58.784
It's transparent.
884
00:47:59.384 --> 00:48:00.825
It's chemically inert.
885
00:48:01.205 --> 00:48:03.287
It doesn't really mess around with other things.
886
00:48:04.588 --> 00:48:07.450
And that chemical inertness is the key property here.
887
00:48:08.350 --> 00:48:11.973
Side note, graphene for the longest time
888
00:48:14.052 --> 00:48:24.678
was overhyped it was in batman's suit isn't it doesn't know what he uses in his suit no it's in it's in all the movies all the movies and then when you go out you're like where is the graphy
889
00:48:26.810 --> 00:48:40.517
Okay, it won the Nobel Prize and everything, but like, I think if you talk to a lot of material scientists, they're going to be like, this thing was super overhyped because of those, it was the strongest material, it was a conductor, it was transparent, it was inert.
890
00:48:41.378 --> 00:48:46.680
It didn't really find a lot of chemical, like a lot of industrial applications.
891
00:48:47.241 --> 00:48:50.382
So it's really nice to see that it's actually doing something.
892
00:48:50.783 --> 00:48:55.285
I'm sure there's going to be people in the comments that tell me that there's all sorts of other stuff that graphene does.
893
00:48:55.746 --> 00:48:58.251
It's just I haven't heard it, right?
894
00:48:58.291 --> 00:48:59.614
I've heard of silica chips.
895
00:48:59.654 --> 00:49:00.556
I've heard of all these other things.
896
00:49:00.576 --> 00:49:04.625
Carbon nanofibers, like in F1, they use like carbon nanofibers.
897
00:49:04.926 --> 00:49:05.787
But that's not graphene.
898
00:49:06.972 --> 00:49:08.972
And carbon nanofibers never won a Nobel Prize.
899
00:49:09.493 --> 00:49:10.613
So I'm just saying.
900
00:49:10.833 --> 00:49:11.493
Bad PR.
901
00:49:11.633 --> 00:49:11.853
Right?
902
00:49:12.713 --> 00:49:14.634
This is a totally new purpose, though.
903
00:49:15.354 --> 00:49:24.276
Because a lot of times people like to use graphene for, as you said, the Mylar and Batman suit, like because it's strong or it's like electrically conducting.
904
00:49:24.396 --> 00:49:28.396
Here, we're exploiting the chemical inertness of graphene.
905
00:49:28.696 --> 00:49:35.958
And it may be finally getting its moment in the light because the implications of this –
906
00:49:38.548 --> 00:49:41.130
could be just so enormous.
907
00:49:41.150 --> 00:49:42.111
Yeah, it really could.
908
00:49:42.452 --> 00:49:42.932
It really could.
909
00:49:43.192 --> 00:49:45.154
So let's now get into the paper.
910
00:49:45.294 --> 00:49:48.136
Let's talk about figure one, the graphene solution.
911
00:49:48.997 --> 00:49:53.120
Figure 1A on the top left, that is their Memrister.
912
00:49:53.421 --> 00:49:53.641
Okay?
913
00:49:53.861 --> 00:49:54.662
You've got a little...
914
00:49:56.069 --> 00:50:02.452
Tungsten electrode up top, you've got hafnium oxide in the middle, and you've got a graphene electrode on the bottom.
915
00:50:02.852 --> 00:50:03.072
Okay?
916
00:50:03.493 --> 00:50:05.274
And that's effectively what's happening.
917
00:50:05.954 --> 00:50:08.235
The tungsten doesn't want to mess with the graphene.
918
00:50:08.255 --> 00:50:10.556
We're going to get into some of the cool physics about that later.
919
00:50:11.016 --> 00:50:12.637
On the right-hand side is what I want to show.
920
00:50:12.897 --> 00:50:14.818
You see that curve, that hysteresis curve?
921
00:50:15.198 --> 00:50:18.700
That's characteristic of your membrister.
922
00:50:19.200 --> 00:50:22.002
So that is where they're showing the membrister.
923
00:50:22.082 --> 00:50:22.402
Mm-hmm.
924
00:50:24.182 --> 00:50:29.746
All of the colors are from zero degrees Celsius all the way down to 700 degrees Celsius.
925
00:50:30.166 --> 00:50:34.049
And notice the curve maintains its shape.
926
00:50:34.129 --> 00:50:34.489
Yes.
927
00:50:34.969 --> 00:50:38.472
All the way down to 700 degrees Celsius.
928
00:50:38.872 --> 00:50:39.773
This is crazy.
929
00:50:39.913 --> 00:50:40.673
Yeah, yeah, yeah.
930
00:50:40.873 --> 00:50:41.454
Right?
931
00:50:41.474 --> 00:50:43.775
Because we're talking about 2.5, 3x.
932
00:50:44.195 --> 00:50:44.416
Yeah.
933
00:50:44.576 --> 00:50:46.477
The current level that we can... Yeah.
934
00:50:46.677 --> 00:50:47.017
And if you...
935
00:50:47.077 --> 00:50:50.540
I mean, usually in temperature, we always, as physicists, like to talk in Kelvin, right?
936
00:50:50.580 --> 00:50:50.760
Right.
937
00:50:51.100 --> 00:50:55.762
Zero degrees Celsius is going to be like 270, let's say 300, 300 Kelvin.
938
00:50:56.342 --> 00:50:59.063
700 degrees Celsius is 1,000 Kelvin.
939
00:50:59.264 --> 00:51:03.705
So this is three times the thermal energy that this thing is able to withstand.
940
00:51:04.026 --> 00:51:08.628
Because Kelvin is really how you scale how much kinetic energy is in the system.
941
00:51:09.108 --> 00:51:12.649
At 1,000 Kelvin or at 900 Kelvin, let's say,
942
00:51:14.598 --> 00:51:21.522
your atoms are moving with three times the kinetic energy as the atoms at 300 Kelvin.
943
00:51:21.562 --> 00:51:21.742
Right.
944
00:51:21.962 --> 00:51:22.202
Okay?
945
00:51:22.342 --> 00:51:23.043
At room temperature.
946
00:51:23.283 --> 00:51:24.523
So this is crazy.
947
00:51:24.543 --> 00:51:24.843
Yeah.
948
00:51:24.863 --> 00:51:25.444
Okay?
949
00:51:25.504 --> 00:51:25.764
Yeah.
950
00:51:26.024 --> 00:51:26.684
Yeah.
951
00:51:26.805 --> 00:51:33.408
On the bottom left plot, you're seeing the time axis on the x-axis, right?
952
00:51:33.688 --> 00:51:33.888
Yep.
953
00:51:34.709 --> 00:51:36.910
That's 10 to the 4 seconds...
954
00:51:38.014 --> 00:51:42.636
You're maintaining those two states of high resistance, low resistance.
955
00:51:42.896 --> 00:51:44.617
So the cycling is maintaining.
956
00:51:44.837 --> 00:51:45.057
Yeah.
957
00:51:45.277 --> 00:51:46.138
That we talked about.
958
00:51:46.198 --> 00:51:47.639
The forest is going up and down.
959
00:51:47.699 --> 00:51:47.879
Yeah.
960
00:51:48.159 --> 00:51:48.759
And it's fine.
961
00:51:48.839 --> 00:51:49.640
And it's no problem.
962
00:51:49.660 --> 00:51:51.320
And it's not even coming together.
963
00:51:52.021 --> 00:51:52.961
It's just flat.
964
00:51:53.021 --> 00:51:54.022
Which is...
965
00:51:54.042 --> 00:51:54.662
Right, right, right, right.
966
00:51:54.842 --> 00:51:56.563
Like there's not a... Like there's not a...
967
00:51:56.963 --> 00:51:57.764
Yeah, I don't see anything.
968
00:51:57.784 --> 00:51:58.725
It's not approaching the collapse.
969
00:51:58.745 --> 00:51:59.225
Yeah.
970
00:51:59.425 --> 00:52:01.367
It's not approaching the short circuit.
971
00:52:01.547 --> 00:52:02.008
No, it's not.
972
00:52:02.028 --> 00:52:05.651
Yeah, short circuit would exactly mean that like they're both coming to the same thing.
973
00:52:05.992 --> 00:52:06.592
We're not seeing that.
974
00:52:06.612 --> 00:52:07.533
We're not seeing that at all.
975
00:52:07.553 --> 00:52:08.334
At 700 degrees Celsius.
976
00:52:08.354 --> 00:52:08.694
Which is odd.
977
00:52:08.714 --> 00:52:10.596
This is cycling at 700 degrees Celsius.
978
00:52:10.736 --> 00:52:11.457
So it's stable.
979
00:52:11.657 --> 00:52:11.877
Yes.
980
00:52:12.177 --> 00:52:14.680
And on the right-hand side, we've got endurance of cycles.
981
00:52:15.080 --> 00:52:18.864
The x-axis there on the bottom right, that's in log scale.
982
00:52:19.184 --> 00:52:19.264
Oh.
983
00:52:19.684 --> 00:52:21.005
Okay, that's not a linear scale.
984
00:52:21.286 --> 00:52:27.450
Each of those tick marks is 10 endurance cycles, then 100, then 1,000, then 10,000, 100,000.
985
00:52:27.710 --> 00:52:30.453
It goes all the way up to a billion cycles.
986
00:52:31.093 --> 00:52:32.294
Before we start to see it.
987
00:52:32.314 --> 00:52:36.878
Yeah, and even then, that could just be noise because there's enough jitter in the beginning.
988
00:52:36.958 --> 00:52:38.479
Yeah, yeah, yeah.
989
00:52:38.619 --> 00:52:39.460
It's not clearly.
990
00:52:40.340 --> 00:52:41.521
This is incredible.
991
00:52:41.541 --> 00:52:41.741
Right?
992
00:52:41.761 --> 00:52:42.542
Right.
993
00:52:43.863 --> 00:52:46.684
Because what we're sort of saying, there's two things here.
994
00:52:46.804 --> 00:52:54.769
One, stability at high temperature, which is C. And then D is stability over time, which goes back to the two things.
995
00:52:54.969 --> 00:53:00.953
And I'm kind of getting what you were saying before about why the fundamentals of the structure are.
996
00:53:01.753 --> 00:53:28.884
impact both of these variables exactly it's impacting both the temperature side of things which is in the b and the time side of things because it's like stable the whole thing is stable and actually the next one which is in i think this is figure two in the paper this shows a schematic you always want a cartoon for all us stupid people like me we're trying to understand what's going on on the left is an old memristor okay tungsten half neem oxide platinum at high
997
00:53:31.603 --> 00:53:33.127
And creating a paved highway.
998
00:53:33.828 --> 00:53:35.492
So the electrons will just move through that.
999
00:53:35.813 --> 00:53:36.013
Yes.
1000
00:53:36.710 --> 00:53:37.651
Because that's the path of least resistance.
1001
00:53:37.671 --> 00:53:41.093
Yeah, no one's going to take the vacancies, the oxygen vacancies.
1002
00:53:41.113 --> 00:53:44.015
Why would you do that when there's a highway right next door?
1003
00:53:45.075 --> 00:53:47.977
It's like the Cars movie, you know, when the interstate was formed.
1004
00:53:48.317 --> 00:53:52.580
Nobody went through Route 66, and then your town got wrecked, right?
1005
00:53:52.760 --> 00:54:00.745
Okay, so on the right-hand side, now there's no highway because the tungsten is trying to diffuse, but it just bounces back from the graphene.
1006
00:54:00.765 --> 00:54:01.986
The graphene's like, I'm good.
1007
00:54:02.606 --> 00:54:03.587
I'm all carbon, right?
1008
00:54:04.363 --> 00:54:05.247
And I'm good.
1009
00:54:05.608 --> 00:54:09.263
It's like California's high-speed train.
1010
00:54:09.572 --> 00:54:10.072
Yeah.
1011
00:54:10.152 --> 00:54:12.154
86 miles of track, but you can't really use it.
1012
00:54:12.174 --> 00:54:12.735
Yeah, you can't.
1013
00:54:12.755 --> 00:54:13.755
Yeah, no, where is it going?
1014
00:54:13.775 --> 00:54:14.376
Where is it going?
1015
00:54:14.416 --> 00:54:14.676
Nowhere.
1016
00:54:14.976 --> 00:54:15.817
Nowhere, right?
1017
00:54:16.618 --> 00:54:18.579
And so that's exactly right, yeah.
1018
00:54:18.859 --> 00:54:22.242
And so the tungsten is sort of just like trying.
1019
00:54:22.582 --> 00:54:23.863
It's still diffusing around.
1020
00:54:23.903 --> 00:54:29.027
The atoms are still diffusing because of the high temperature, but it can't bind to the graphene.
1021
00:54:29.308 --> 00:54:34.212
And so the only way through is with the vacancy centers that are created through that –
1022
00:54:34.732 --> 00:54:59.027
hiker's paradise yes and that's where the electrons shuttle and so now we have exquisite control of the resistance state the the the one on the right here the the the hiker's path that you're talking about that structure which is the forest the the bike analogy where we have the ability to control when they're up or down that that's the point is that that's the level the lever that gives us uh
1023
00:54:59.247 --> 00:55:01.409
It's programmatic control, for lack of a better term.
1024
00:55:01.489 --> 00:55:02.490
Exactly, exactly.
1025
00:55:02.631 --> 00:55:10.579
And Professor Joshua Wang, who is the principal investigator, he described this interface interaction between the tungsten and the graphene as oil and water.
1026
00:55:11.059 --> 00:55:13.442
It's just like they're just not going to mix, right?
1027
00:55:13.662 --> 00:55:14.083
Makes sense.
1028
00:55:14.763 --> 00:55:16.525
This is so good.
1029
00:55:16.745 --> 00:55:17.466
It's pretty cool, right?
1030
00:55:17.706 --> 00:55:20.569
So now let's – to get into a paper in science –
1031
00:55:21.370 --> 00:55:22.530
You can't just have a cartoon.
1032
00:55:22.751 --> 00:55:22.971
Yeah.
1033
00:55:23.591 --> 00:55:24.231
Right?
1034
00:55:24.271 --> 00:55:24.891
That's for me.
1035
00:55:26.292 --> 00:55:36.195
But in order to publish in science and convince your competitors that this is legit and all that other kind of stuff, you need actual analytical verification.
1036
00:55:36.235 --> 00:55:38.196
So let's go to the first analytical verification.
1037
00:55:38.476 --> 00:55:42.257
This is high-resolution transmission electron microscopy.
1038
00:55:42.537 --> 00:55:44.518
Electron microscopes, we've covered this a lot.
1039
00:55:44.598 --> 00:55:50.480
Instead of light, we're using high-energy electrons to really get down to, you know—
1040
00:55:51.060 --> 00:55:54.441
Tiny, tiny nanometer level resolution.
1041
00:55:55.701 --> 00:55:58.722
On the top is a traditional memristor.
1042
00:55:59.302 --> 00:55:59.562
Okay?
1043
00:56:00.743 --> 00:56:05.324
You can see after, like, not a lot of time, the curves collapse.
1044
00:56:05.404 --> 00:56:05.644
Yeah.
1045
00:56:05.844 --> 00:56:06.044
Right?
1046
00:56:06.104 --> 00:56:09.045
After just, like, 2,000 seconds, which is, like, less than an hour.
1047
00:56:09.065 --> 00:56:09.165
Yeah.
1048
00:56:10.708 --> 00:56:11.769
there's no memorister.
1049
00:56:12.049 --> 00:56:14.131
It's the slope of a mammoth.
1050
00:56:14.371 --> 00:56:15.632
Yeah, yeah, yeah, exactly.
1051
00:56:15.652 --> 00:56:16.092
Straight down.
1052
00:56:16.292 --> 00:56:18.614
Yeah, yeah, chair 36, I believe.
1053
00:56:19.574 --> 00:56:20.475
Correct me in the comments.
1054
00:56:20.615 --> 00:56:23.317
I never go on those chairs, so that'll be you, lunatics.
1055
00:56:23.737 --> 00:56:31.862
But on the right-hand side is the transmission electron microscopes, and you can see at high temperature, you see this mushroom cloud that's forming on the right-hand side?
1056
00:56:32.383 --> 00:56:34.904
That's the platinum and the tungsten meeting up.
1057
00:56:35.144 --> 00:56:35.585
Mm-hmm.
1058
00:56:36.565 --> 00:56:43.990
And so instead of sort of maintaining the same substrate texture or orientation, there's just this disturbance in the force.
1059
00:56:44.050 --> 00:56:44.450
Exactly.
1060
00:56:44.470 --> 00:56:45.450
And you can visually see it.
1061
00:56:45.470 --> 00:56:47.852
Yeah, and you can see the hafnium oxide in the middle, right?
1062
00:56:47.892 --> 00:56:48.332
The meat.
1063
00:56:48.432 --> 00:56:50.633
But like the platinum is just like going through.
1064
00:56:51.014 --> 00:56:53.735
Why would you go through the hafnium oxide, right?
1065
00:56:53.995 --> 00:56:57.717
On the bottom is our new memristor with graphene.
1066
00:56:58.318 --> 00:57:01.740
And there you can see even at high temperature, even after all of that cycling,
1067
00:57:04.030 --> 00:57:05.713
The graphene layer is inert.
1068
00:57:06.113 --> 00:57:07.595
The tungsten layer is inert.
1069
00:57:07.795 --> 00:57:08.576
Nothing is happening.
1070
00:57:08.596 --> 00:57:09.197
This is incredible.
1071
00:57:09.217 --> 00:57:10.439
There's no chemistry happening.
1072
00:57:10.699 --> 00:57:12.622
And it's also so verifiable.
1073
00:57:13.002 --> 00:57:13.203
Yeah.
1074
00:57:13.443 --> 00:57:16.187
You can literally look at it under the microscope.
1075
00:57:16.287 --> 00:57:16.707
I can see.
1076
00:57:16.968 --> 00:57:18.249
And you can see that it's happening.
1077
00:57:18.838 --> 00:57:25.866
That stupid analogy is like imagine you have a BLT sandwich from a deli and they put the toothpick through it.
1078
00:57:26.746 --> 00:57:34.054
In the first case, it's like the toothpick through where you can see it disturbs the nice layers you have of your tomato, bacon, lettuce.
1079
00:57:34.315 --> 00:57:34.535
Yeah.
1080
00:57:34.755 --> 00:57:37.798
And it looks like a very nice, undisturbed.
1081
00:57:38.039 --> 00:57:40.241
But the point here is the –
1082
00:57:41.382 --> 00:57:50.911
The high-resolution imagery that is able to validate that this graphene-based memorister actually sustains at high temperature is undeniable.
1083
00:57:50.931 --> 00:57:51.471
Yeah, yeah, yeah.
1084
00:57:51.711 --> 00:57:53.033
It's like now you don't need a cartoon.
1085
00:57:53.073 --> 00:57:54.994
This is the thing that the cartoon is inspired from.
1086
00:57:55.014 --> 00:57:55.815
This is literally what it is, right.
1087
00:57:55.955 --> 00:57:56.155
Right?
1088
00:57:56.496 --> 00:57:57.417
Like you're just seeing it.
1089
00:57:58.231 --> 00:57:58.751
I can see it.
1090
00:57:59.431 --> 00:58:00.352
Here's another way to see it.
1091
00:58:00.392 --> 00:58:03.092
You can use energy dispersive x-ray spectroscopy.
1092
00:58:03.473 --> 00:58:10.135
The idea here is that, like, so during, in the previous photo, we had transmission electron microscopy.
1093
00:58:10.155 --> 00:58:14.116
So you're, like, sending these high-energy electron beams, right, that strike the sample.
1094
00:58:14.576 --> 00:58:20.878
Now, when those electrons strike the sample, it can transfer energy and violently kick electrons out.
1095
00:58:21.398 --> 00:58:24.519
And when the electrons come back, they're going to release x-rays.
1096
00:58:25.805 --> 00:58:37.293
And if I monitor the x-rays during that electron microscopy, because every atom is unique, the x-ray signature is going to be unique.
1097
00:58:37.333 --> 00:58:41.937
Like in the transmission electron microscopy, people could be like, oh, there's some weird thing happening.
1098
00:58:41.957 --> 00:58:43.057
How do you know that's tungsten?
1099
00:58:43.658 --> 00:58:45.919
How do you know that's platinum?
1100
00:58:45.999 --> 00:58:47.120
It's just a gray image.
1101
00:58:47.481 --> 00:58:50.523
Well, now we can tell that it is, in fact, tungsten.
1102
00:58:51.330 --> 00:59:02.838
That is leaking through because only a certain energy comes from tungsten, and we're seeing that energy spike in the old Memorister, but in the new Memorister, there's no tungsten spike.
1103
00:59:02.858 --> 00:59:03.999
There's no tungsten spike.
1104
00:59:04.519 --> 00:59:09.003
The W, by the way, for those watching, W is the chemical symbol for tungsten.
1105
00:59:10.172 --> 00:59:10.892
Because of Latin.
1106
00:59:11.632 --> 00:59:17.334
And so it's interesting because I think part of this is like you're doing both of these things almost simultaneously.
1107
00:59:17.374 --> 00:59:17.874
Yeah, yeah, yeah.
1108
00:59:17.894 --> 00:59:24.195
Because the electron microscopy is triggering an event that you track with the x-ray spectroscopy.
1109
00:59:24.555 --> 00:59:28.436
And so you're getting both the visual and the signature.
1110
00:59:28.676 --> 00:59:29.497
Yeah, exactly.
1111
00:59:29.537 --> 00:59:35.638
That allows you to identify the composition of what you're looking at and can see visually.
1112
00:59:35.738 --> 00:59:36.138
Exactly.
1113
00:59:36.178 --> 00:59:39.039
And so if you look in the platinum electrode, there is tungsten there.
1114
00:59:40.199 --> 00:59:41.800
There's no tungsten, right?
1115
00:59:42.180 --> 00:59:43.461
Undeniable.
1116
00:59:43.981 --> 00:59:47.322
There were also some calculations that were done from first principles.
1117
00:59:47.362 --> 00:59:47.562
Yes.
1118
00:59:47.823 --> 00:59:51.144
So you've got electron densities with density functional theory.
1119
00:59:51.164 --> 00:59:55.086
Density functional theory is something that we covered in our America 250 episode.
1120
00:59:55.106 --> 00:59:55.286
Yes.
1121
00:59:55.326 --> 00:59:56.547
It won the Nobel Prize.
1122
00:59:56.927 --> 00:59:59.428
This is what I mean by density functional theory, okay?
1123
00:59:59.688 --> 01:00:02.249
It's like you don't have to worry about every single electron.
1124
01:00:02.489 --> 01:00:06.131
You can just be like the electron clouds are like a potential, right?
1125
01:00:06.311 --> 01:00:10.433
What happens when a tungsten atom goes through?
1126
01:00:10.753 --> 01:00:11.594
Does it bind or not?
1127
01:00:12.014 --> 01:00:14.655
On the right-hand side is what platinum looks like.
1128
01:00:15.756 --> 01:00:21.479
These giant sort of honeycomb structures where the d-orbital, which is the stuff that...
1129
01:00:22.399 --> 01:00:24.821
These are these large d-orbitals of metals, right?
1130
01:00:25.141 --> 01:00:26.241
They're extremely large.
1131
01:00:27.362 --> 01:00:29.163
The tungsten is fine, kind of just...
1132
01:00:30.210 --> 01:00:34.033
binding to one of these d-orbitals because they're so big, right?
1133
01:00:34.553 --> 01:00:38.296
And there's a lot of vacant spots where the tungsten can go and bind.
1134
01:00:38.696 --> 01:00:40.357
On the left-hand side is graphene.
1135
01:00:41.018 --> 01:00:42.879
Pristine hexagonal structure.
1136
01:00:44.640 --> 01:00:46.882
I think it's only just like sp2 binding.
1137
01:00:48.075 --> 01:00:48.916
Or maybe sp3.
1138
01:00:49.536 --> 01:00:50.577
I forget my chemistry.
1139
01:00:50.597 --> 01:00:55.882
But in any case, the bond lengths in between the carbons are smaller.
1140
01:00:56.642 --> 01:01:00.085
The graphene electron clouds are a lot smaller.
1141
01:01:00.426 --> 01:01:02.587
So the tungsten has nowhere to go.
1142
01:01:02.608 --> 01:01:03.188
It's incredible.
1143
01:01:04.509 --> 01:01:06.450
You see the fat little pimple?
1144
01:01:06.731 --> 01:01:06.931
Yes.
1145
01:01:06.991 --> 01:01:08.092
That's the tungsten atom.
1146
01:01:08.132 --> 01:01:10.714
And it's trying to find somewhere to bind.
1147
01:01:10.954 --> 01:01:11.534
And it can't.
1148
01:01:12.195 --> 01:01:15.357
It's like the grill on your microphone or a speaker cover.
1149
01:01:15.417 --> 01:01:16.198
Yeah.
1150
01:01:16.258 --> 01:01:20.601
Versus like you described, a bunch of donuts that got blended.
1151
01:01:20.641 --> 01:01:23.964
Or like Hawaiian rolls that got baked together.
1152
01:01:24.284 --> 01:01:24.885
Yeah, exactly.
1153
01:01:24.925 --> 01:01:30.691
With no clear, actual, like, structural, you know, like a well-defined small structural positioning.
1154
01:01:30.751 --> 01:01:31.172
Exactly.
1155
01:01:31.272 --> 01:01:33.494
And it's part of your point that the...
1156
01:01:34.821 --> 01:01:39.465
The smaller size of the electron cloud is a material aspect of why.
1157
01:01:40.286 --> 01:01:42.568
Because the tungsten is just not small enough to... Yeah, yeah.
1158
01:01:42.608 --> 01:01:43.748
The tungsten is not small enough.
1159
01:01:43.768 --> 01:01:47.131
The smaller size also means that you need a higher energy to get in there.
1160
01:01:47.351 --> 01:01:47.712
That makes sense.
1161
01:01:47.732 --> 01:01:47.932
Right?
1162
01:01:47.952 --> 01:01:49.153
Yeah, yeah.
1163
01:01:49.273 --> 01:01:51.034
And all of the stuff is taken.
1164
01:01:51.274 --> 01:01:51.475
Yes.
1165
01:01:51.715 --> 01:01:51.915
Right?
1166
01:01:52.175 --> 01:01:55.638
With graphene, because of that hexagonal structure, it's just carbon on carbon on carbon.
1167
01:01:55.658 --> 01:01:57.539
There's no... All of the carbon atoms are happy.
1168
01:01:57.860 --> 01:01:58.760
They don't want anything else.
1169
01:01:58.780 --> 01:01:59.441
They don't want anything.
1170
01:01:59.541 --> 01:01:59.721
Right?
1171
01:01:59.741 --> 01:01:59.921
Yeah.
1172
01:01:59.941 --> 01:02:00.041
Yeah.
1173
01:02:00.782 --> 01:02:16.857
yeah oh that's so good so it's all coming together and the final thing that I want to talk about is the nudged elastic band this isn't the final thing but I just I'm going through their figures because they're just so cool okay so on the left hand side again that's the old memorister and
1174
01:02:18.984 --> 01:02:26.207
This is an activation energy diagram that shows if I want the tungsten to bind to my platinum, I need to go over this hill.
1175
01:02:26.847 --> 01:02:31.669
And then there's like a valley on the other side where I'm bound to the platinum.
1176
01:02:32.049 --> 01:02:34.750
The height of the hill is 0.3 electron volts.
1177
01:02:35.960 --> 01:02:39.461
And then there's a nice valley where I can just like sit.
1178
01:02:39.961 --> 01:02:42.701
On the right-hand side is the same diagram but for graphene.
1179
01:02:43.021 --> 01:02:46.622
The height of the hill is now 1 to 2 electron volts.
1180
01:02:47.462 --> 01:02:49.503
And there's no nice valley, right?
1181
01:02:50.143 --> 01:02:54.824
Like if I'm up there and I'm jiggling around, I'll just come right back to my original state.
1182
01:02:55.024 --> 01:02:56.744
So again, same idea.
1183
01:02:57.024 --> 01:02:59.365
It's really hard to bind to graphene.
1184
01:03:00.065 --> 01:03:00.806
It's really good.
1185
01:03:00.966 --> 01:03:01.226
Okay.
1186
01:03:02.227 --> 01:03:06.310
Now, one thing that I was thinking about was, there's kind of a paradox in my head, right?
1187
01:03:06.530 --> 01:03:08.572
Like the high resistance state makes sense.
1188
01:03:09.092 --> 01:03:15.818
High resistance meaning there's no oxygen vacancy highway.
1189
01:03:17.059 --> 01:03:22.983
And so, like my electrons aren't going through, but everything is kind of inert, right?
1190
01:03:23.644 --> 01:03:27.287
Now, let's say I have oxygen vacancy highway.
1191
01:03:27.307 --> 01:03:27.807
Okay.
1192
01:03:28.056 --> 01:03:33.799
in that low resistance state, what prevents the oxygen vacancy highway from moving around?
1193
01:03:34.719 --> 01:03:35.140
Yeah.
1194
01:03:35.240 --> 01:03:36.140
At high temperature.
1195
01:03:36.220 --> 01:03:36.640
Yeah, yeah.
1196
01:03:36.680 --> 01:03:36.860
Right?
1197
01:03:36.960 --> 01:03:44.684
At high temperature, if I've got these, like, vacancy lines where the electrons are going through, but those vacancies can move, as I was saying.
1198
01:03:44.764 --> 01:03:46.425
As we talked about earlier, they're not stationary.
1199
01:03:46.445 --> 01:03:47.105
They're not stationary.
1200
01:03:47.125 --> 01:03:47.806
They're not fixed or whatever.
1201
01:03:47.826 --> 01:03:48.386
Yeah, yeah, yeah.
1202
01:03:48.426 --> 01:03:53.228
So what's preventing those guys from moving around at high temperature when there's all this jiggle?
1203
01:03:53.629 --> 01:03:55.910
It has to do with something called phase separation.
1204
01:03:55.930 --> 01:03:56.270
Right.
1205
01:03:56.470 --> 01:04:08.537
There was a paper about this where the system reaches thermodynamic flux equilibrium where there's a net migration of zero, meaning that the material separates into an oxygen-rich insulating phase.
1206
01:04:09.363 --> 01:04:11.764
and then an oxygen-poor conducting phase.
1207
01:04:11.784 --> 01:04:14.005
And it's that same argument about oil and water.
1208
01:04:14.305 --> 01:04:21.368
There's going to be parts of the hafnium oxide that have a lot of these oxygen vacancies, and then there's going to be parts that don't.
1209
01:04:21.728 --> 01:04:22.949
And those guys don't want to mix together.
1210
01:04:22.969 --> 01:04:23.689
They don't want to mix.
1211
01:04:23.849 --> 01:04:25.470
Because they become like oil and water.
1212
01:04:25.730 --> 01:04:28.932
Now, how can we tell in our new memorister?
1213
01:04:29.752 --> 01:04:29.972
Well...
1214
01:04:30.288 --> 01:04:35.294
They did electron energy loss spectroscopy, which is similar to the x-ray stuff that I was talking about.
1215
01:04:35.555 --> 01:04:40.661
Like the electrons are now coming out, and then we measure the energy of the electrons.
1216
01:04:41.162 --> 01:04:45.447
The main thing that I want you to see is in the low resistance state,
1217
01:04:46.279 --> 01:04:47.600
which is on the bottom.
1218
01:04:47.620 --> 01:04:51.804
So the color tells you how much oxygen vacancy there is.
1219
01:04:51.924 --> 01:04:55.006
The darker it is, the more oxygen vacancy there is.
1220
01:04:55.046 --> 01:05:01.111
So that's why the low resistance state at the bottom, that has, it's kind of darker because there's a lot more oxygen vacancies.
1221
01:05:01.792 --> 01:05:05.294
The key thing that I want you to notice is on the bottom, we're plotting...
1222
01:05:06.035 --> 01:05:09.416
the number of oxygen vacancies as a function of space.
1223
01:05:09.716 --> 01:05:13.237
Like we're going, we're moving along in nanometer position.
1224
01:05:13.257 --> 01:05:17.619
We're moving along our memristor and we're plotting how much oxygen vacancy there is.
1225
01:05:18.419 --> 01:05:21.420
If you, if you expect a lot of mixing, you're,
1226
01:05:21.880 --> 01:05:23.161
it should be at equilibrium.
1227
01:05:23.721 --> 01:05:25.981
But if there's not a lot of mixing, there should be clumps.
1228
01:05:26.122 --> 01:05:26.402
Yeah, right.
1229
01:05:26.462 --> 01:05:29.843
Where there's a lot of oxygen vacancy and then not a lot of oxygen vacancy.
1230
01:05:29.863 --> 01:05:31.663
Because of this oil and water you just talked about.
1231
01:05:31.703 --> 01:05:32.464
And that's what we're seeing.
1232
01:05:32.484 --> 01:05:34.824
We're seeing a very bumpy profile, right?
1233
01:05:35.045 --> 01:05:39.186
There's certain spots where there's a lot of oxygen vacancy and then other spots that there aren't.
1234
01:05:39.926 --> 01:05:44.908
If this stuff mixed, it would look like the middle plot in that high resistance state.
1235
01:05:44.928 --> 01:05:45.448
Mm-hmm.
1236
01:05:45.922 --> 01:05:46.182
Right?
1237
01:05:46.863 --> 01:05:47.083
Yes.
1238
01:05:47.743 --> 01:05:52.146
But because it's not mixing, you get this valley and mountain landscape.
1239
01:05:52.606 --> 01:05:55.007
And so part of what we're saying is there's two dynamics here.
1240
01:05:55.047 --> 01:06:00.391
There's both the dynamic between how the tungsten and the graphene relate.
1241
01:06:01.051 --> 01:06:04.113
And then the halphide oxygen?
1242
01:06:04.233 --> 01:06:05.614
Yeah, hafnium.
1243
01:06:05.814 --> 01:06:06.394
Hafnium oxide.
1244
01:06:06.414 --> 01:06:10.537
Excuse me, hafnium oxide in the middle also has dynamics that...
1245
01:06:11.297 --> 01:06:27.680
that combine with the relationship between the tungsten and the graphene such that the graphene doesn't want to create the highway and the oxide in the middle also doesn't allow movement.
1246
01:06:27.920 --> 01:06:35.901
And so both of these things combined are meaningful for the impact in terms of from an industrial or performance perspective.
1247
01:06:36.001 --> 01:06:40.322
And it's both things, not just the ends of the sandwich.
1248
01:06:40.522 --> 01:06:40.982
Exactly.
1249
01:06:41.022 --> 01:06:41.242
Yeah.
1250
01:06:41.323 --> 01:06:46.806
So both the high resistance state, which is the one, or sorry, no, the high resistance is the zero.
1251
01:06:46.846 --> 01:06:46.966
Yeah.
1252
01:06:46.986 --> 01:06:48.767
Because that's like the electrons aren't moving.
1253
01:06:49.027 --> 01:06:53.189
And then the low resistance state where there is current, that's the one, both of those are now stable.
1254
01:06:53.449 --> 01:06:53.669
Right.
1255
01:06:54.030 --> 01:06:54.570
Right.
1256
01:06:54.650 --> 01:07:03.235
For dynamics that we just walked through, which is an important point because I was going to end up asking, but what about the middle stuff?
1257
01:07:03.475 --> 01:07:03.655
Yeah.
1258
01:07:04.214 --> 01:07:04.834
Yeah.
1259
01:07:04.854 --> 01:07:06.395
The middle stuff is also oil and water.
1260
01:07:06.415 --> 01:07:07.435
Right, right, right.
1261
01:07:07.515 --> 01:07:08.115
It's kind of cool.
1262
01:07:08.255 --> 01:07:08.895
Yeah.
1263
01:07:09.015 --> 01:07:13.356
And they did a lot of benchmarking of like, okay, what is all the state of the art?
1264
01:07:13.416 --> 01:07:14.356
And then where is this?
1265
01:07:14.976 --> 01:07:16.637
And it's kind of crazy.
1266
01:07:17.457 --> 01:07:21.818
Retention, which is how long it takes and the temperature, that's part A.
1267
01:07:22.558 --> 01:07:24.779
You want it to be high temperature, high retention.
1268
01:07:25.019 --> 01:07:26.339
So you've got a desired corner.
1269
01:07:27.320 --> 01:07:28.000
That's the shaded.
1270
01:07:28.160 --> 01:07:30.261
And there's a star where the current work is.
1271
01:07:30.661 --> 01:07:34.983
And all of the other riffraff is not even close.
1272
01:07:35.263 --> 01:07:38.664
You've got the writing endurance in B versus temperature.
1273
01:07:38.764 --> 01:07:42.886
Again, the only star in the desired corner is this work.
1274
01:07:43.166 --> 01:07:45.887
C, the on-off ratio versus the temperature.
1275
01:07:45.947 --> 01:07:47.288
Desired corner is us.
1276
01:07:47.668 --> 01:07:48.948
The spiderweb diagram.
1277
01:07:50.149 --> 01:07:51.949
The yellow is the current work.
1278
01:07:52.270 --> 01:07:53.250
And it's always...
1279
01:07:55.028 --> 01:08:02.112
Larger in temperature, in retention, in endurance, in the device size, all sorts of stuff.
1280
01:08:02.372 --> 01:08:06.854
Now, there is a little bit of Photoshoppery happening here, right?
1281
01:08:06.874 --> 01:08:11.177
Because I could easily make the desired corner like elsewhere.
1282
01:08:11.457 --> 01:08:17.080
But in any case, by every metric, it's the most in that corner.
1283
01:08:17.100 --> 01:08:21.923
So whatever desired corner you made, it would always be the most out there, right?
1284
01:08:21.963 --> 01:08:22.163
Right.
1285
01:08:22.703 --> 01:08:28.447
It's the closest to the optimal, regardless of whether you made the desired outcome smaller or larger or whatever.
1286
01:08:28.467 --> 01:08:29.007
Yeah, yeah, yeah.
1287
01:08:29.047 --> 01:08:31.789
But a game-recognized game because, like, I've done that.
1288
01:08:32.189 --> 01:08:33.510
I know what you're up to.
1289
01:08:33.590 --> 01:08:38.494
Yeah, I'm going to make this square right here so that all the others are out.
1290
01:08:38.514 --> 01:08:40.335
But, like, why do you just mix it in?
1291
01:08:40.615 --> 01:08:42.977
Yeah, but in this case, I think it's, like, fairly obvious.
1292
01:08:43.017 --> 01:08:44.558
And I don't think they're doing any shenanigans.
1293
01:08:44.618 --> 01:08:46.420
I just thought that was kind of funny as a plot.
1294
01:08:47.361 --> 01:08:52.165
And finally, like, because I think they knew that people would call BS.
1295
01:08:52.405 --> 01:08:52.605
Right.
1296
01:08:52.785 --> 01:08:54.747
So they took a video of their lab.
1297
01:08:56.108 --> 01:08:56.889
They're like, proof.
1298
01:08:57.029 --> 01:08:58.010
They're like, this is proof.
1299
01:08:58.270 --> 01:09:01.272
And the thingy shows, oh, 700 degrees Celsius.
1300
01:09:01.613 --> 01:09:07.377
And they've got, like, a computer with the two electrodes on the two...
1301
01:09:09.282 --> 01:09:18.070
on the titanium and the graphene, and they're measuring the on off ratio and they're measuring the hysteresis curve.
1302
01:09:18.230 --> 01:09:25.937
And here you can see the hysteresis curve, like, okay, I turned the voltage up, I turned the voltage down, I'm getting this hysteresis, guys, this is happening in real time.
1303
01:09:26.337 --> 01:09:28.698
I'm not making this stuff up.
1304
01:09:28.778 --> 01:09:33.039
Like, I think it's really cool that like this is part of the supplement on the science website.
1305
01:09:33.180 --> 01:09:37.121
Because like whenever you publish, right, there's the main figures and the main text.
1306
01:09:37.441 --> 01:09:40.282
And then you provide supplementary data and supplementary figures.
1307
01:09:40.602 --> 01:09:45.264
A lot of the figures that we've seen here in today's episode are from the supplement.
1308
01:09:45.384 --> 01:09:45.684
Mm-hmm.
1309
01:09:46.550 --> 01:09:48.793
And this is a supplementary video that they show.
1310
01:09:49.074 --> 01:09:50.896
Okay, guys, here's my hysteresis curve.
1311
01:09:51.256 --> 01:09:52.558
It's definitely a memristor.
1312
01:09:52.839 --> 01:09:57.384
This is happening at 700 degrees Celsius, and it's definitely working.
1313
01:09:57.465 --> 01:09:58.726
Pixar didn't happen.
1314
01:09:58.766 --> 01:09:59.027
Yeah.
1315
01:10:00.188 --> 01:10:25.244
uh in uh frontier science research but it's also kind of like i think one people one thing you know as we always talk about is scientists are some of the most skeptical people especially when it's other people that say they did something it's like why didn't i think of that i don't i don't believe you yeah but here yeah it's like okay here's 704 degrees celsius guys there's my thermometer yeah
1316
01:10:27.085 --> 01:10:47.603
God, this is so good because – and I think you've done a good job of walking us through to kind of understand that there's a fundamental insight that happens that then you take to the end of possibility or the end of kind of the story in order to get to something that's practical.
1317
01:10:47.683 --> 01:10:47.883
Mm-hmm.
1318
01:10:48.343 --> 01:10:54.048
And it was a very – it was maybe subtle for folks who are material scientists and who work in the space.
1319
01:10:54.408 --> 01:11:02.055
But if you're not in this lane, it can seem like, oh, this is a subtle point that has such huge implications.
1320
01:11:02.496 --> 01:11:07.460
And they've used graphene as the other part of this layer here.
1321
01:11:09.700 --> 01:11:14.664
Are there other things that have different performance characteristics?
1322
01:11:14.684 --> 01:11:18.727
Yeah, maybe we can now start replacing the tungsten.
1323
01:11:19.028 --> 01:11:19.408
I don't know.
1324
01:11:20.269 --> 01:11:20.489
Right.
1325
01:11:20.949 --> 01:11:22.430
And then what does that do?
1326
01:11:22.530 --> 01:11:24.932
Yeah, it opens up all sorts of possibilities.
1327
01:11:24.992 --> 01:11:25.172
Right.
1328
01:11:25.853 --> 01:11:27.194
So now let's talk about the future.
1329
01:11:27.534 --> 01:11:27.774
Okay.
1330
01:11:28.115 --> 01:11:34.580
What would this current research have to do with how we move forward in computation as a society?
1331
01:11:34.600 --> 01:11:35.140
Right.
1332
01:11:35.484 --> 01:11:38.368
We've talked about the von Neumann bottleneck for a long time.
1333
01:11:38.749 --> 01:11:40.491
This is the tyranny of...
1334
01:11:41.933 --> 01:11:45.695
the processor having to be separate from our memory.
1335
01:11:45.936 --> 01:11:58.524
Modern architectures waste a lot of power and time trying to move data between our CPU, which is where all of the computation is happening, and the RAM, which is where the data and the memory is stored.
1336
01:11:59.444 --> 01:12:06.089
Von Neumann came up with this, and it's the bedrock of almost all of the computation that we have in our society today.
1337
01:12:06.369 --> 01:12:10.612
Now, with neural networks, this is especially bad because you've got to load the weights
1338
01:12:11.867 --> 01:12:13.429
And then you've got to do the computation.
1339
01:12:13.870 --> 01:12:15.992
Then you've got to load a new set of weights.
1340
01:12:16.193 --> 01:12:17.775
So you've got to dump the old weights, right?
1341
01:12:18.095 --> 01:12:20.378
And all of this takes a lot of time.
1342
01:12:20.698 --> 01:12:24.183
It takes a lot of energy, dissipates heat, and...
1343
01:12:25.015 --> 01:12:28.878
What exactly is the computation that happens in these AI systems?
1344
01:12:29.278 --> 01:12:36.683
When you've got a neural network, what we're really doing fundamentally is we're relying on matrix times vector multiplication.
1345
01:12:37.003 --> 01:12:40.726
The vector is your state and the matrix is the weights of your network.
1346
01:12:41.126 --> 01:12:48.531
You multiply the matrix to the vector and then you get a new vector that tells you how the state is changing and how the state is changing.
1347
01:12:48.891 --> 01:12:51.012
At the end of the whole thing, you get a giant vector.
1348
01:12:51.212 --> 01:12:57.933
In the example of LLMs, you have a giant vector with all of the next possible tokens.
1349
01:12:58.774 --> 01:13:01.454
And you pick usually the one with the highest probability.
1350
01:13:02.054 --> 01:13:02.315
Okay.
1351
01:13:03.175 --> 01:13:05.595
How does matrix multiplication work?
1352
01:13:05.775 --> 01:13:07.916
How do you multiply a matrix by a vector?
1353
01:13:08.296 --> 01:13:10.597
For those from linear algebra, you already know this.
1354
01:13:10.897 --> 01:13:12.437
But let's just do a little bit of a review.
1355
01:13:12.677 --> 01:13:12.917
Okay.
1356
01:13:13.297 --> 01:13:14.518
Let's say I've got a 3x3 matrix.
1357
01:13:14.538 --> 01:13:14.738
Okay.
1358
01:13:15.897 --> 01:13:18.458
On the left there, I've got 1, 1, 2, 2, 1, 3, 1, 4, 2.
1359
01:13:19.079 --> 01:13:21.180
And I want to multiply it by a vector 3, 1, 2.
1360
01:13:21.260 --> 01:13:25.843
3, 1, 2 is the vector that's my state of like next tokens, let's say.
1361
01:13:26.243 --> 01:13:33.847
And the matrix tells us how to change that vector to create a new probability vector with the new tokens.
1362
01:13:34.287 --> 01:13:38.710
What you do is you go row by row and you take each row, you multiply it by the vector.
1363
01:13:39.030 --> 01:13:42.132
So the 1 gets multiplied by the 3, the 1 gets multiplied by the 1.
1364
01:13:42.612 --> 01:13:44.873
2 gets multiplied by the 2 and you add it all up.
1365
01:13:45.034 --> 01:13:45.894
That's your first row.
1366
01:13:46.194 --> 01:13:48.315
Then you go to the second row, add it all up.
1367
01:13:48.536 --> 01:13:51.878
Third row, multiply, add it all up and you get your new vector.
1368
01:13:52.338 --> 01:13:52.558
Okay.
1369
01:13:52.838 --> 01:14:02.084
That's fundamentally what most of the AI computation that we all know and love, I guess.
1370
01:14:02.784 --> 01:14:03.464
That's what that is.
1371
01:14:03.965 --> 01:14:04.245
Okay.
1372
01:14:04.965 --> 01:14:05.626
That's what's happening.
1373
01:14:05.746 --> 01:14:06.626
Keyword, I guess.
1374
01:14:07.086 --> 01:14:07.307
Yeah.
1375
01:14:07.847 --> 01:14:08.507
That's what's happening.
1376
01:14:09.128 --> 01:14:09.308
Now.
1377
01:14:10.825 --> 01:14:11.886
We'd like to do better.
1378
01:14:11.946 --> 01:14:16.152
We'd like to not load the weights every single time.
1379
01:14:17.251 --> 01:14:18.672
in order to do this computation.
1380
01:14:18.832 --> 01:14:28.880
And this is part of the reason why, for example, the cloud versus local with AI models and stuff, it's why if you want to do it locally, you need a very powerful machine because it has to do this.
1381
01:14:28.980 --> 01:14:33.424
Yeah, it has to store all of those matrices and it's got to do all of that computation.
1382
01:14:33.644 --> 01:14:39.429
That's why GPUs are so important because GPUs are just insanely good at matrix multiplication.
1383
01:14:39.469 --> 01:14:40.570
At that specific task.
1384
01:14:40.670 --> 01:14:42.531
At that very specific task, okay?
1385
01:14:42.712 --> 01:14:45.774
They're not good at doing normal emails,
1386
01:14:47.235 --> 01:14:55.850
It turns out NVIDIA made the big gamble way back in the day that actually graphics, which is matrix multiplication...
1387
01:14:57.167 --> 01:14:58.667
is going to have a use for later.
1388
01:14:59.028 --> 01:14:59.868
And it turns out, yeah.
1389
01:15:00.108 --> 01:15:09.310
In things other than video games, for those who are gamers, we've been very familiar with GPUs for a while, and now everyone is making custom PC builds unnecessarily expensive.
1390
01:15:09.390 --> 01:15:10.631
Yeah, yeah, exactly.
1391
01:15:10.871 --> 01:15:16.972
So instead of doing all this von Neumann nonsense, instead we can do something called in-memory computing.
1392
01:15:17.593 --> 01:15:21.814
In-memory computing meaning I don't have to move anything, okay?
1393
01:15:22.094 --> 01:15:24.875
We can use something like a resistor
1394
01:15:28.490 --> 01:15:31.571
do the computation without changing anything.
1395
01:15:32.231 --> 01:15:35.332
So again, let's talk about how that would work.
1396
01:15:35.812 --> 01:15:38.013
I told you about how matrix multiplication works, right?
1397
01:15:38.053 --> 01:15:41.534
Like matrices, you multiply it by a vector, and then you get a new vector.
1398
01:15:41.914 --> 01:15:45.895
Well, now, what if we had the old vector come in like wires...
1399
01:15:46.759 --> 01:15:47.019
Okay?
1400
01:15:47.480 --> 01:15:49.742
And this is what we see on the right.
1401
01:15:49.822 --> 01:15:51.764
And let me see if I can say this correctly.
1402
01:15:51.824 --> 01:15:52.064
Okay?
1403
01:15:52.604 --> 01:15:56.528
So I've got a matrix, and I've got an old vector, and I want to make a new vector.
1404
01:15:56.568 --> 01:16:00.832
And the way that I do that is the rows get multiplied to each of them, and then I get added up, right?
1405
01:16:01.873 --> 01:16:11.602
I can use the rules of electronics, V equals IR, and the summing of voltages and the summing of currents, to then just do that using physics.
1406
01:16:12.002 --> 01:16:12.562
Here's what I do.
1407
01:16:13.402 --> 01:16:21.429
My old vector is represented by the old... By currents that are going at the very top.
1408
01:16:21.689 --> 01:16:23.851
The old voltages, I should say.
1409
01:16:24.351 --> 01:16:28.175
Each of the wires at the very top is held at a certain voltage.
1410
01:16:28.495 --> 01:16:31.938
I've got resistors that are attached to these voltages.
1411
01:16:33.760 --> 01:16:36.482
And then I've got a new set of wires at the bottom.
1412
01:16:37.323 --> 01:16:38.684
Because of electronics...
1413
01:16:39.973 --> 01:16:44.445
The new current, it's a new set of currents at the bottom.
1414
01:16:44.966 --> 01:16:46.330
Because of electronics, right?
1415
01:16:46.952 --> 01:16:47.934
The new currents...
1416
01:16:49.112 --> 01:16:54.855
are going to be related to the voltage at the top and the resistance in between.
1417
01:16:55.156 --> 01:16:56.837
And all I have to do is add up the currents.
1418
01:16:56.977 --> 01:16:57.737
I'm so mad.
1419
01:16:57.917 --> 01:16:58.157
Right?
1420
01:16:58.438 --> 01:17:00.079
In this case, it's not really resistance.
1421
01:17:00.119 --> 01:17:01.499
It's one over resistance.
1422
01:17:01.780 --> 01:17:02.620
It's the conductance.
1423
01:17:03.100 --> 01:17:06.222
And then the current is just voltage multiplied by conductance.
1424
01:17:06.683 --> 01:17:07.443
But it's the same thing.
1425
01:17:07.463 --> 01:17:09.704
It's just V equals IR the other way around.
1426
01:17:10.285 --> 01:17:13.587
And all we're doing is using the physics in place...
1427
01:17:14.686 --> 01:17:16.727
We don't have to move weights around.
1428
01:17:16.847 --> 01:17:24.131
As long as the weights are the same, I can use the physics to be like voltage multiplied by conductance gives me a current.
1429
01:17:24.591 --> 01:17:27.713
Voltage multiplied by another conductance gives me another current.
1430
01:17:28.053 --> 01:17:32.256
And so my new vector is going to be the sum of all of those currents.
1431
01:17:32.596 --> 01:17:33.156
And there you go.
1432
01:17:33.496 --> 01:17:35.818
Matrix multiplication using physics.
1433
01:17:35.958 --> 01:17:36.418
I'm so...
1434
01:17:36.858 --> 01:17:37.479
I'm so mad.
1435
01:17:37.559 --> 01:17:38.660
It's so good.
1436
01:17:38.680 --> 01:17:39.320
That's so good.
1437
01:17:39.420 --> 01:17:48.688
And part of the point here is then you can get around the von Neumann bottleneck as a result of that because you're not having to transition between your compute layer and your memory layer.
1438
01:17:48.728 --> 01:17:56.395
Yeah, and all you have to do is the weights have to be programmed into the conductance, the input is the voltage, and the resulting multiplication comes out as current.
1439
01:17:56.675 --> 01:18:02.340
I'm just using Ohm's law and Kirchhoff's junction rules, the stuff that you learn in intro physics, right?
1440
01:18:03.080 --> 01:18:05.506
Now, are we capable of doing something like this?
1441
01:18:05.566 --> 01:18:12.761
Well, this particular paper itself made a 32 by 32 analog resistance state
1442
01:18:13.616 --> 01:18:15.998
And it works at 1300 degrees Fahrenheit.
1443
01:18:16.559 --> 01:18:17.740
It works at low voltage.
1444
01:18:17.940 --> 01:18:19.562
0.5 volts is all you need.
1445
01:18:21.403 --> 01:18:24.226
30 nanoseconds switching in the speed.
1446
01:18:25.768 --> 01:18:32.754
And it definitively proves that you can be fully capable of storing complex, multi-level
1447
01:18:33.622 --> 01:18:37.184
analog neural weights for whatever AI interface that you're working with.
1448
01:18:37.664 --> 01:18:40.706
It can do it in extreme environments, 1300 degrees Fahrenheit.
1449
01:18:41.466 --> 01:18:48.950
And the standard digital GPU, which relies on all these transistors and the CMOS and all that other kind of stuff, the traditional GPU is going to get fried.
1450
01:18:48.990 --> 01:18:58.855
But if I use this memristor where the conductances, the weights, are stored using this memristor, now all of a sudden I've got something that can actually work.
1451
01:18:58.875 --> 01:18:58.955
God.
1452
01:19:00.495 --> 01:19:02.356
It's so, it's so good.
1453
01:19:02.476 --> 01:19:22.484
And it's this interesting combination of both things, which is the, um, fundamental discovery around the memorister construction for high temperature, plus the concept of, uh, in memory computing together.
1454
01:19:22.684 --> 01:19:22.884
Yeah.
1455
01:19:23.344 --> 01:19:23.564
Um,
1456
01:19:24.951 --> 01:19:27.213
Because it's not just one or the other.
1457
01:19:27.634 --> 01:19:30.297
Both things give you now the single chip –
1458
01:19:31.558 --> 01:19:38.781
that can work in high temperature, that can run comparably for a long time.
1459
01:19:38.881 --> 01:19:39.441
Faster.
1460
01:19:39.681 --> 01:19:40.242
Faster.
1461
01:19:41.722 --> 01:19:44.744
And can do both your compute and memory at the same time.
1462
01:19:44.924 --> 01:19:45.364
Exactly.
1463
01:19:45.724 --> 01:19:48.845
And of course, the authors have made a startup.
1464
01:19:49.466 --> 01:19:49.966
Of course they did.
1465
01:19:50.006 --> 01:19:51.566
There's a startup called TetraMem.
1466
01:19:51.727 --> 01:19:58.769
It was co-founded by Joshua Yang, who's the PI of the paper that we're talking about, along with Chiang Fei Xia and the other authors.
1467
01:19:59.430 --> 01:19:59.550
Yeah.
1468
01:20:00.170 --> 01:20:06.932
And it's already working to commercialize this crossbar Memristor device to replace power-hungry GPUs.
1469
01:20:07.512 --> 01:20:07.712
Right?
1470
01:20:08.012 --> 01:20:09.613
They're getting acquired immediately.
1471
01:20:09.873 --> 01:20:10.133
Yeah.
1472
01:20:10.233 --> 01:20:11.373
No, especially with this.
1473
01:20:11.553 --> 01:20:12.233
Immediately.
1474
01:20:12.493 --> 01:20:16.314
Now, let's talk about what this has to do with data centers in space.
1475
01:20:16.494 --> 01:20:16.835
Okay.
1476
01:20:16.855 --> 01:20:17.055
Okay?
1477
01:20:17.395 --> 01:20:19.275
Elon wants to put data centers in space.
1478
01:20:19.355 --> 01:20:19.695
Why?
1479
01:20:20.235 --> 01:20:21.816
Because it's probably possible.
1480
01:20:21.836 --> 01:20:22.096
Yeah.
1481
01:20:22.836 --> 01:20:25.099
And, I mean, it helps SpaceX, obviously.
1482
01:20:25.479 --> 01:20:28.162
And it's not just SpaceX that's trying to do this, right?
1483
01:20:28.222 --> 01:20:29.563
Google is trying to do this.
1484
01:20:29.844 --> 01:20:31.946
There's a bunch of startups that are trying to do this.
1485
01:20:32.426 --> 01:20:39.914
And it's looked impossible for a very, very long time because of a sort of fundamental physics quandary.
1486
01:20:40.635 --> 01:20:43.338
The reason why is because these GPUs are...
1487
01:20:44.641 --> 01:20:50.546
Highly reliant on cooling mechanisms because the GPUs can't withstand a lot of heat.
1488
01:20:50.826 --> 01:20:59.534
That's why whenever you hear about all these data centers taking up local water, the water is being used to cool down the GPUs effectively.
1489
01:21:00.155 --> 01:21:02.237
At the end of the day, that's most of the water being used.
1490
01:21:02.877 --> 01:21:08.920
is just to extract heat from these GPUs and put it out into the environment.
1491
01:21:08.940 --> 01:21:15.624
That's why the data centers, the climate around the data centers, there's like a microclimate that's created and the temperature gets raised.
1492
01:21:15.904 --> 01:21:21.667
And then all the environmental people are like, okay, what's happening in the flora and fauna, which is completely reasonable.
1493
01:21:21.987 --> 01:21:22.247
Right.
1494
01:21:22.367 --> 01:21:27.550
I don't want to hurt like the planet and especially the planet that's right around those data centers.
1495
01:21:27.890 --> 01:21:28.090
So yeah,
1496
01:21:29.127 --> 01:21:31.209
Let's talk about how heat transfer happens.
1497
01:21:31.389 --> 01:21:33.370
There's three types of heat transfer, okay?
1498
01:21:33.691 --> 01:21:36.833
There's convection, conduction, and radiation.
1499
01:21:36.913 --> 01:21:43.058
Convection is hot stuff moves around to the cold part, and then the cold part goes down to where the hot stuff is.
1500
01:21:43.078 --> 01:21:52.285
That's how, like, if you put a pot on a stove, the water gets heated uniformly because the hot packets are moving and the cold packets are moving.
1501
01:21:52.365 --> 01:21:53.326
So that's convection.
1502
01:21:53.726 --> 01:21:55.128
Conduction is...
1503
01:21:56.322 --> 01:22:01.984
The atoms themselves jiggling around to conduct heat from one spot to the other.
1504
01:22:02.024 --> 01:22:03.824
That's why the handle gets hot.
1505
01:22:04.224 --> 01:22:10.286
And then finally, the most inefficient part is radiation, which is literally photons coming out.
1506
01:22:11.126 --> 01:22:21.149
Like the reason why a fire is hot, but if you put your hand in front of the fire, all of a sudden you don't feel that hot, is because you've stopped photons from getting to your face.
1507
01:22:22.009 --> 01:22:29.552
And that's why a shade is so important because now the photons, the infrared photons from the sun are not hitting you.
1508
01:22:29.832 --> 01:22:32.333
And then all of a sudden you feel like 10 or 20 degrees cooler.
1509
01:22:33.273 --> 01:22:34.594
You're right next to the sun.
1510
01:22:35.474 --> 01:22:37.355
The shade is only shading your part.
1511
01:22:37.735 --> 01:22:40.636
But the reason why it's so effective is because the photons aren't getting to you.
1512
01:22:40.896 --> 01:22:42.817
The convection and the conduction still is.
1513
01:22:43.237 --> 01:22:46.839
The hot pockets of air are still moving towards you, but the photons are not.
1514
01:22:47.099 --> 01:22:49.860
Okay, so those are the three types of heat transfer.
1515
01:22:49.880 --> 01:22:50.060
Okay.
1516
01:22:51.138 --> 01:22:56.280
Now, terrestrial data centers, they mostly use the first two.
1517
01:22:56.840 --> 01:22:57.901
Mostly it's convection.
1518
01:22:58.081 --> 01:23:04.683
That's why you've got water pipes going through, because the water gets hot, and then you move the water to a colder spot.
1519
01:23:04.844 --> 01:23:08.445
It dumps the heat, and you cycle it back, and so on and so forth, right?
1520
01:23:10.446 --> 01:23:12.987
Terrestrial data centers have issues, right?
1521
01:23:13.067 --> 01:23:14.267
Where is the power going to come from?
1522
01:23:14.747 --> 01:23:15.788
Where is the water going to come from?
1523
01:23:16.654 --> 01:23:17.374
Who's going to build it?
1524
01:23:17.514 --> 01:23:18.935
Yeah, who's going to pay for it?
1525
01:23:19.035 --> 01:23:20.856
Yeah, there's a lot of nimbiness going on.
1526
01:23:20.956 --> 01:23:23.157
And in this case, the nimbiness is a good thing, right?
1527
01:23:23.177 --> 01:23:26.559
We don't want, I don't want data centers in my neighborhood.
1528
01:23:26.599 --> 01:23:26.759
Yeah.
1529
01:23:26.959 --> 01:23:27.200
Okay.
1530
01:23:27.920 --> 01:23:30.381
I might want the AI, but I really don't want the data centers.
1531
01:23:31.022 --> 01:23:31.182
Right.
1532
01:23:31.442 --> 01:23:33.343
And so that's why people are looking into space.
1533
01:23:33.683 --> 01:23:35.324
Now, people have already started looking.
1534
01:23:35.864 --> 01:23:36.785
This is StarCloud.
1535
01:23:36.885 --> 01:23:39.246
It's a startup that's backed by NVIDIA.
1536
01:23:39.466 --> 01:23:40.687
Google is doing its own thing.
1537
01:23:41.527 --> 01:23:43.571
And there's a lot of smart people betting on this thing.
1538
01:23:43.611 --> 01:23:46.075
Here it shows a 5 gigawatt data center.
1539
01:23:46.095 --> 01:23:47.177
That's crazy.
1540
01:23:47.277 --> 01:23:54.709
But the radiator that you would need to dissipate all that heat with radiation is something like a 4 kilometer square.
1541
01:23:55.678 --> 01:24:02.682
And you have real, you know, getting any payload into space is cost prohibitive.
1542
01:24:02.822 --> 01:24:03.042
Yeah.
1543
01:24:03.082 --> 01:24:09.666
And so getting a payload that's four kilometers by four kilometers wide will take a modular system, multiple launches.
1544
01:24:09.706 --> 01:24:11.687
Yes, the cost of launches is going down.
1545
01:24:11.907 --> 01:24:12.747
Yeah.
1546
01:24:12.767 --> 01:24:15.349
Still, it's four kilometers worth of stuff.
1547
01:24:15.369 --> 01:24:15.629
Yeah.
1548
01:24:15.749 --> 01:24:16.609
worth of stuff, right?
1549
01:24:16.649 --> 01:24:17.670
It's not a joke.
1550
01:24:17.730 --> 01:24:18.850
It's not trivial, right?
1551
01:24:18.890 --> 01:24:22.652
And there's several important papers that are coming out trying to plan for this kind of thing.
1552
01:24:22.952 --> 01:24:26.033
Like, how do we develop carbon-neutral data centers in space?
1553
01:24:26.333 --> 01:24:28.834
I want to highlight this because this is in Nature Electronics.
1554
01:24:28.914 --> 01:24:30.595
This is a top journal.
1555
01:24:31.055 --> 01:24:32.275
It's not impossible.
1556
01:24:32.615 --> 01:24:41.139
There's a lot of rhetoric out there that's saying that data centers in space is impossible because of the vacuum of space and people need to take undergraduate physics.
1557
01:24:41.559 --> 01:24:42.619
It's not true, okay?
1558
01:24:42.639 --> 01:24:43.079
It's hard, right?
1559
01:24:45.567 --> 01:24:48.370
But that's a key distinction from it's impossible.
1560
01:24:48.830 --> 01:24:49.090
Okay?
1561
01:24:49.611 --> 01:24:51.292
I mean, we obviously do it, right?
1562
01:24:51.613 --> 01:24:54.515
The International Space Station is already doing this.
1563
01:24:54.575 --> 01:25:01.562
If we look at a photo of the International Space Station, the curved panels that we see, those are solar panels that are facing the sun.
1564
01:25:02.703 --> 01:25:04.485
The flat ones, those are radiators.
1565
01:25:05.125 --> 01:25:14.970
Those are radiating out heat into space so that the astronauts that are on board don't get cooked alive in all of the electronics that is doing computation.
1566
01:25:14.990 --> 01:25:16.311
There are several computers up there.
1567
01:25:16.711 --> 01:25:17.931
Those don't get cooked alive.
1568
01:25:18.192 --> 01:25:20.573
The way that works is using both convection
1569
01:25:21.713 --> 01:25:22.553
And radiation.
1570
01:25:22.593 --> 01:25:27.494
So you've got convection in that they use ammonia instead of water because ammonia has a low freezing point.
1571
01:25:27.774 --> 01:25:31.955
If you use water, then the water is just going to freeze when it gets to the radiator.
1572
01:25:31.995 --> 01:25:33.215
And then like what?
1573
01:25:33.495 --> 01:25:34.916
So that's why you use ammonia.
1574
01:25:35.236 --> 01:25:39.557
The ammonia uses convection to take the heat to the radiator.
1575
01:25:39.817 --> 01:25:43.177
And then the radiator takes that heat and radiates it out into space.
1576
01:25:43.697 --> 01:25:43.937
Okay.
1577
01:25:44.377 --> 01:25:46.498
Now, what we want to do is make this thing cheaper.
1578
01:25:46.518 --> 01:25:46.898
Okay.
1579
01:25:47.604 --> 01:25:54.909
And that is perhaps an engineering problem, especially in the context of the paper that we've just covered.
1580
01:25:55.209 --> 01:26:02.974
I'd like to pose that as a challenge and then see if we can work through with just some first principles, back-of-the-envelope thinking.
1581
01:26:03.794 --> 01:26:06.916
The technology poses three advantages.
1582
01:26:07.236 --> 01:26:10.699
One, let's talk about the physics of thermal radiation.
1583
01:26:10.939 --> 01:26:13.740
The reason why everyone's saying that it's going to be impossible.
1584
01:26:13.760 --> 01:26:14.141
Right?
1585
01:26:14.161 --> 01:26:14.301
Right.
1586
01:26:16.153 --> 01:26:27.897
Thermal radiation is governed by something called the Stefan-Boltzmann law, which states that the energy that is radiated by a surface at a certain temperature is proportional to the fourth power of that temperature.
1587
01:26:28.177 --> 01:26:29.197
This is why it's hard.
1588
01:26:29.577 --> 01:26:29.817
Okay?
1589
01:26:30.057 --> 01:26:30.958
It's not T squared.
1590
01:26:31.358 --> 01:26:32.398
It's T to the fourth.
1591
01:26:32.778 --> 01:26:42.701
So if I double the temperature, I'm increasing the amount of power output I need or the power output that's going out by a factor of 16.
1592
01:26:43.261 --> 01:26:43.501
Okay?
1593
01:26:43.922 --> 01:26:45.262
Now, this is actually a good thing.
1594
01:26:46.315 --> 01:26:46.575
Okay.
1595
01:26:46.975 --> 01:26:48.756
In the context of this current paper.
1596
01:26:49.257 --> 01:26:53.719
Now, standard silicon processors, they've got to be relatively cool to prevent failure, right?
1597
01:26:53.940 --> 01:27:01.324
But you can only allow that chip with the standard silicon to run at like tops 100 degrees Celsius.
1598
01:27:01.944 --> 01:27:03.905
At 100 degrees Celsius, right?
1599
01:27:04.286 --> 01:27:08.208
Let's say 400 Kelvin is the tops that you can do.
1600
01:27:09.769 --> 01:27:10.489
At 400 Kelvin...
1601
01:27:12.367 --> 01:27:15.150
The amount of radiation that I'm letting off is going to be some amount.
1602
01:27:15.391 --> 01:27:16.812
Some amount of photons are going off.
1603
01:27:17.673 --> 01:27:23.881
Now, what if I could make that go up to 700 degrees Celsius?
1604
01:27:24.972 --> 01:27:25.792
1,000 Kelvin.
1605
01:27:26.052 --> 01:27:28.633
So I've effectively doubled the temperature that I'm working with.
1606
01:27:29.073 --> 01:27:35.254
Well, the number of photons, the amount of power that I'm dissipating has now gone up by a 16, by a factor of 16.
1607
01:27:35.594 --> 01:27:47.937
If the amount of power per unit area has gone up by a factor of 16, then in order to dissipate the same amount of power, I only need a 16th the size of radiator.
1608
01:27:48.037 --> 01:27:48.157
Yeah.
1609
01:27:48.177 --> 01:27:48.317
Yeah.
1610
01:27:49.917 --> 01:27:54.879
The size of the radiator goes down, the higher the temperature that you can operate.
1611
01:27:54.939 --> 01:27:59.241
And this is crucial for putting the stuff into space.
1612
01:27:59.561 --> 01:28:02.783
Because of what you said, it's hard to put stuff in space.
1613
01:28:03.683 --> 01:28:04.183
It's hard.
1614
01:28:04.243 --> 01:28:05.023
It's expensive.
1615
01:28:05.404 --> 01:28:08.225
I like to put less stuff in space.
1616
01:28:09.067 --> 01:28:10.208
This could be the key, right?
1617
01:28:10.228 --> 01:28:21.774
Because now I can operate things at a much higher temperature, meaning that the amount of watts that are going out as radiation into space is higher, which means the amount of area that I need for my radiator is smaller.
1618
01:28:22.314 --> 01:28:23.395
This is so interesting.
1619
01:28:23.435 --> 01:28:37.363
And it is enabled by this high temperature mrister and the fundamentals that we just walked through that, again, now give you a different substrate, a whole different computing paradigm where...
1620
01:28:37.623 --> 01:28:40.565
When you think about the concept of a data center in space.
1621
01:28:40.706 --> 01:28:41.546
Yeah, exactly.
1622
01:28:41.586 --> 01:28:43.708
And I did some back of the envelope calculations, right?
1623
01:28:43.768 --> 01:28:49.593
So let's say I need a radiator for a one megawatt space server.
1624
01:28:50.293 --> 01:28:50.514
Okay.
1625
01:28:51.534 --> 01:28:54.897
That means one megawatt of electricity is coming in from my solar panels.
1626
01:28:55.758 --> 01:28:55.978
Okay.
1627
01:28:56.278 --> 01:28:59.521
Now at standard 30 degrees Celsius, so that's 300 Kelvin.
1628
01:28:59.861 --> 01:29:05.386
If I want to dump one megawatt of heat, I would need 2,000 meters squared.
1629
01:29:05.446 --> 01:29:06.547
That's eight tennis courts.
1630
01:29:06.567 --> 01:29:06.767
Okay.
1631
01:29:08.254 --> 01:29:11.775
And I want to dump just as much energy that's coming in, right?
1632
01:29:11.835 --> 01:29:18.076
Because if I don't dump the energy that's coming in, then my temperature increases and then the electronics are not stable.
1633
01:29:18.116 --> 01:29:20.136
My satellite keeps increasing in temperature.
1634
01:29:20.436 --> 01:29:22.716
So whatever energy is coming in, I'd like to dump out.
1635
01:29:22.977 --> 01:29:25.237
So one megawatt's in, one megawatt's out.
1636
01:29:26.217 --> 01:29:28.837
What matters is the temperature that I'm operating at, okay?
1637
01:29:29.278 --> 01:29:36.899
Now, if one megawatt is coming in, and now let's say I'm operating at 300 degrees Celsius, that's around 600 Kelvin-ish.
1638
01:29:37.899 --> 01:29:40.624
My area has gone down to 150 meters squared.
1639
01:29:41.064 --> 01:29:44.590
I've gone from 2,000 meters squared to 150 meters squared.
1640
01:29:44.610 --> 01:29:45.532
That's a reduction of 90%.
1641
01:29:47.515 --> 01:29:47.715
Right.
1642
01:29:47.995 --> 01:29:51.338
If I go even further, because this thing was at 700 degrees Celsius, right?
1643
01:29:51.838 --> 01:30:00.884
If I go even further to 600 degrees Celsius now, meaning 900 Kelvin, now my radiators only need to be 30 meters squared.
1644
01:30:01.485 --> 01:30:03.026
And that's totally doable.
1645
01:30:03.086 --> 01:30:04.286
Yeah, that's very reasonable.
1646
01:30:04.647 --> 01:30:04.847
Right.
1647
01:30:05.207 --> 01:30:08.129
That's already in the ballpark of stuff that we've already put out there.
1648
01:30:08.169 --> 01:30:09.610
Right, right, right.
1649
01:30:10.216 --> 01:30:20.082
Right, that's so – for me, it's a little counterintuitive where it's at the higher – before we started, the higher the temperature, the less surface area you need to dissipate.
1650
01:30:20.582 --> 01:30:25.185
The heat, it doesn't feel – Yeah, as long as the stuff is not cooking.
1651
01:30:25.505 --> 01:30:25.725
Right.
1652
01:30:25.965 --> 01:30:26.185
Right?
1653
01:30:26.485 --> 01:30:32.469
But it kind of makes sense because if you think about – you know those electric stoves with the coils?
1654
01:30:32.729 --> 01:30:32.949
Yeah.
1655
01:30:33.169 --> 01:30:33.369
Right?
1656
01:30:34.190 --> 01:30:37.592
The higher the temperature, the more heat is coming out.
1657
01:30:37.612 --> 01:30:39.473
That's the whole point of how it cooks stuff.
1658
01:30:39.653 --> 01:30:39.793
Yeah.
1659
01:30:40.193 --> 01:30:42.657
No, I guess that's not true because there's like contact.
1660
01:30:42.697 --> 01:30:43.758
But you get what I'm saying, right?
1661
01:30:43.778 --> 01:30:48.865
Like if you're around that thing, it's like there's a lot more photons that are coming out.
1662
01:30:49.085 --> 01:30:50.527
So you're dumping a lot more energy.
1663
01:30:50.808 --> 01:30:52.189
And as long as you can live –
1664
01:30:53.368 --> 01:30:55.629
At 700 degrees Celsius, you're good to go.
1665
01:30:55.969 --> 01:31:00.331
And it's just because we've never been able to think about it because we've never been able to operate at that temperature level.
1666
01:31:00.391 --> 01:31:01.991
So it never was even a possibility.
1667
01:31:02.191 --> 01:31:04.833
There were no chip components that could withstand that temperature.
1668
01:31:04.893 --> 01:31:11.795
So everyone was doing these calculations at the lower temperature being like, oh, I need tennis court size, four kilometers squared size.
1669
01:31:12.035 --> 01:31:21.799
But now I've reduced stuff down to like 95, 99 percent the area just because my components can withstand the higher temperature.
1670
01:31:22.119 --> 01:31:24.359
My components, my computer can be hotter.
1671
01:31:24.800 --> 01:31:29.020
And so the amount of radiator I need is a lot, lot lower by T to the fourth.
1672
01:31:29.060 --> 01:31:37.342
This is so, I mean, you know, again, there is the industrial scale question and all of these other, you know, engineering.
1673
01:31:37.822 --> 01:31:50.165
But to the point you brought up earlier, a lot of the narrative and rhetoric has been dismissive on its face in a way that doesn't even take into account that.
1674
01:31:51.465 --> 01:31:52.766
what we're doing at the frontier.
1675
01:31:52.926 --> 01:31:53.086
Yeah.
1676
01:31:53.126 --> 01:31:54.708
I mean, this paper came out in March.
1677
01:31:54.988 --> 01:31:55.228
Yeah.
1678
01:31:55.288 --> 01:31:55.728
By the way.
1679
01:31:56.089 --> 01:32:00.152
And this discussion about data centers in space has really heated up as we've gotten into the summer.
1680
01:32:00.372 --> 01:32:00.912
Yes.
1681
01:32:00.972 --> 01:32:03.014
Which was after this paper had come out.
1682
01:32:03.474 --> 01:32:09.939
And most of the folks who have been like, it's impossible, you know, certainly one from a fundamental physics perspective.
1683
01:32:10.279 --> 01:32:10.980
It's not impossible.
1684
01:32:11.000 --> 01:32:11.840
It's not impossible.
1685
01:32:12.261 --> 01:32:19.146
And then also from where we are on the frontier, both from a basic research and an engineering perspective, it's a very different paradigm than...
1686
01:32:19.566 --> 01:32:23.073
You know, just talking about putting the same thing we have in a data center on the ground.
1687
01:32:23.093 --> 01:32:23.293
Yeah.
1688
01:32:23.714 --> 01:32:24.796
And just putting it in space.
1689
01:32:24.856 --> 01:32:25.237
Exactly.
1690
01:32:25.497 --> 01:32:30.046
And if you combine that with the fact that payloads to space are going to get cheaper, as you said...
1691
01:32:30.667 --> 01:32:33.749
the amount of payload that I need is going to go down.
1692
01:32:34.089 --> 01:32:36.230
This is totally a doable thing, right?
1693
01:32:36.550 --> 01:32:48.557
As long as we keep innovating and we keep going through with high-temperature memisters, high-temperature memory, high-temperature computation, that's going to be the key, I think, to unlocking data centers in space.
1694
01:32:48.717 --> 01:32:57.182
Yeah, because you can have a lot more smaller, you can have much smaller space servers with high...
1695
01:32:57.902 --> 01:33:22.609
output capability um meaning we're not going to be you know looking up at the night sky and seeing you know yeah these giant floating radiators yeah and i i mean it perhaps this is is going to be bad for astronomy terrestrial astronomy but there's there's got to be ways around it i mean it's already bad with the starlink satellites that are up but does that mean that like we don't provide it
1696
01:33:23.309 --> 01:33:38.497
internet to like these remote areas like you know technology and innovation comes at a cost and we're just gonna have to figure out ways to battle and with the lower cost of payload you could argue we could have a much larger array of space-based instrumentation
1697
01:33:38.817 --> 01:33:39.237
Exactly.
1698
01:33:39.257 --> 01:33:43.801
That exists that negates the need to have terrestrial observatories.
1699
01:33:43.901 --> 01:33:45.623
Again, there's some caveats there.
1700
01:33:45.643 --> 01:33:46.083
Yeah, yeah.
1701
01:33:46.163 --> 01:33:47.024
There's a lot of caveats.
1702
01:33:47.044 --> 01:33:48.065
A lot of caveats, right.
1703
01:33:48.085 --> 01:33:50.907
But, you know, they're trade-offs and we'll have to continue to figure out.
1704
01:33:50.927 --> 01:33:56.071
Yeah, maybe we can make the orbits, like, avoid Chile, which is where, like, half of everything is.
1705
01:33:56.352 --> 01:33:56.792
Right, right.
1706
01:33:57.473 --> 01:34:00.235
If you just avoid, like, Hawaii and Chile, like… No fly zones.
1707
01:34:00.255 --> 01:34:02.117
Like, half of the astronomers are going to be happy.
1708
01:34:02.137 --> 01:34:02.637
Yeah, right, right.
1709
01:34:03.017 --> 01:34:07.519
You know, so I mean, that's going to be hard given that it's like high Earth orbit and whatever.
1710
01:34:07.780 --> 01:34:14.163
But in any case, the other thing that I want to bring up is this is not just an advantage in terms of cooling.
1711
01:34:14.923 --> 01:34:21.987
OK, there's two other things that this new Memristor technology would give us for data centers in space.
1712
01:34:22.167 --> 01:34:23.888
One is the extreme power efficiency.
1713
01:34:24.588 --> 01:34:30.271
I was talking about how it only takes like very little amount of voltage in order to do that computation with a Memristor.
1714
01:34:31.093 --> 01:34:36.415
Now, satellites operate on extremely strict power budgets, you can imagine, right?
1715
01:34:36.995 --> 01:34:38.295
Often it's under 20 watts.
1716
01:34:38.675 --> 01:34:45.317
And in standard AI chips, nearly every single watt of electricity is consumed directly into heat.
1717
01:34:45.738 --> 01:34:48.619
But Memristors use this in-memory analog computing.
1718
01:34:48.679 --> 01:34:51.219
They don't have to do this von Neumann bottleneck, right?
1719
01:34:51.259 --> 01:34:59.302
So if we're applying Memristors here, then that complex mathematics instantaneously happens without shuttling the data back and forth...
1720
01:34:59.902 --> 01:35:03.183
Now you're consuming a fraction of the power of a traditional GPU.
1721
01:35:03.503 --> 01:35:07.444
And that's going to bring down the amount of power that you need to dissipate again.
1722
01:35:08.345 --> 01:35:12.586
So this is both you can have the radiators be smaller and you can have the solar panels be smaller.
1723
01:35:12.786 --> 01:35:13.186
Exactly.
1724
01:35:13.206 --> 01:35:14.927
Because you don't need as much power intake.
1725
01:35:15.107 --> 01:35:15.527
Exactly.
1726
01:35:15.787 --> 01:35:17.948
Or with the same solar panels, you can do a lot more.
1727
01:35:18.128 --> 01:35:18.328
Right.
1728
01:35:18.668 --> 01:35:19.708
You know, either way.
1729
01:35:19.788 --> 01:35:20.368
Right.
1730
01:35:20.649 --> 01:35:21.829
Efficiency is always great.
1731
01:35:22.370 --> 01:35:24.813
Both in power dissipation and in terms of computing.
1732
01:35:25.114 --> 01:35:31.943
And finally, the other big thing that people talk about with data centers in space is like normal GPUs, a cosmic ray comes in,
1733
01:35:33.707 --> 01:35:37.270
I don't know if that bit was a zero or a one.
1734
01:35:37.970 --> 01:35:38.751
Like, what?
1735
01:35:39.091 --> 01:35:46.595
So a galaxy decided to fart, and then now my AI query is giving me nonsense.
1736
01:35:46.775 --> 01:35:47.956
That's not great.
1737
01:35:47.976 --> 01:35:54.060
You need electronics that is able to sustain performance in a high-radiation environment.
1738
01:35:54.620 --> 01:36:02.128
Memristors are great for this because memristors store memory by physically moving atoms and creating structural bridges, right?
1739
01:36:02.148 --> 01:36:08.856
This like valence charge mechanism rather than relying on these easily disrupted pockets of electrical charge.
1740
01:36:09.788 --> 01:36:11.929
So now you can imagine a cosmic ray comes in.
1741
01:36:12.509 --> 01:36:20.272
It's not really going to upset the valence charge mechanism all that much compared to if there was a single point of failure.
1742
01:36:20.652 --> 01:36:20.912
Right.
1743
01:36:22.153 --> 01:36:31.796
And so we have a thermal radiation benefit, a power efficiency benefit, and a resilience benefit.
1744
01:36:32.136 --> 01:36:34.197
And it's hard for me to imagine that.
1745
01:36:34.557 --> 01:36:44.080
Again, there is the engineering problem, but it's kind of like when NVIDIA started with GPUs before you could really understand what the impact would be.
1746
01:36:44.700 --> 01:36:53.003
We already can see now in the case of data centers in space why if you can figure out the industrial scaling of this –
1747
01:36:55.024 --> 01:37:04.412
It makes the execution of it in three of the most important categories that you care about as an operator.
1748
01:37:06.153 --> 01:37:16.296
significantly better in an order or orders of magnitude from what you would otherwise be doing with the current conventional options.
1749
01:37:16.456 --> 01:37:17.296
Exactly, yeah.
1750
01:37:17.676 --> 01:37:23.738
I mean, I think it's an incredibly exciting time for electronics at the cutting edge, right?
1751
01:37:24.058 --> 01:37:32.880
Because we're starting to see that things like GPUs and things that are capable of doing matrix multiplication and things like that are going to become...
1752
01:37:33.240 --> 01:37:38.341
increasingly important, perhaps even more important than traditional von Neumann architectures.
1753
01:37:39.141 --> 01:37:40.501
We've still got a long way to go.
1754
01:37:40.621 --> 01:37:46.983
So I do want to temper the brakes here and not dream too much because memory alone does not make a complete computer.
1755
01:37:47.323 --> 01:37:54.124
You need high temperature logic circuits that also need to be developed and integrated alongside whatever chip that you're trying to put in.
1756
01:37:54.464 --> 01:37:59.745
And current devices are built by hand, submicron scale in the lab.
1757
01:38:00.363 --> 01:38:03.864
This is from their paper in the supplement.
1758
01:38:04.704 --> 01:38:08.305
That length bar there says 5 micrometers.
1759
01:38:08.545 --> 01:38:18.666
So that's like a thousand times bigger than the chips that we have, like the transistors that we have in our chips in my computer, for example.
1760
01:38:19.087 --> 01:38:25.628
So we've got to make this thing smaller while still retaining the memorister characteristics, the graphene.
1761
01:38:25.908 --> 01:38:26.948
That's going to be a problem,
1762
01:38:27.228 --> 01:38:27.468
Right.
1763
01:38:28.089 --> 01:38:30.932
Scale up is going to take time on the manufacturing side.
1764
01:38:31.413 --> 01:38:37.519
Two of the three devices, two of the three materials in this device, namely the tungsten and the hafnium oxide.
1765
01:38:37.819 --> 01:38:41.483
Those are already standard in semiconductor manufacturing.
1766
01:38:41.523 --> 01:38:41.684
Right.
1767
01:38:41.984 --> 01:38:43.265
Making the graphene smaller.
1768
01:38:44.141 --> 01:38:45.482
That's going to be kind of tough, right?
1769
01:38:45.562 --> 01:38:46.623
It's newer to the industry.
1770
01:38:46.864 --> 01:38:51.788
But actually TSMC and Samsung both have that in their development roadmaps.
1771
01:38:51.928 --> 01:38:55.632
If you go look, they have the shareholder pitch decks.
1772
01:38:56.092 --> 01:38:59.536
They're like, we're still innovating, guys.
1773
01:39:00.036 --> 01:39:01.618
So they have graphene.
1774
01:39:03.047 --> 01:39:07.392
in their roadmap and it's already grown at wafer scale in research settings.
1775
01:39:07.692 --> 01:39:10.735
We need to make it industrial if we want to have this thing grow.
1776
01:39:11.015 --> 01:39:20.105
I mean, the startup that these USC scientists Tetra Mem co-founded by Joshua Yang and Xia, I mean, it's already like doing things.
1777
01:39:20.305 --> 01:39:24.630
So we're, we're, we're heading for a new era in computation in a,
1778
01:39:25.413 --> 01:39:43.681
era where computation is becoming the bread and butter for our society more than ever more so than ever i mean we used to think right ever since the advent of the 90s and the advent of the personal computer the computation would be big now it's like cloud computing and these large-scale computing
1779
01:39:44.485 --> 01:39:49.967
Large-scale computing is now taking precedence over personal small-scale computing.
1780
01:39:49.987 --> 01:39:53.668
So we're entering into a new era of computation.
1781
01:39:54.428 --> 01:39:57.569
We've lived through one in the 90s, and now this is a new one.
1782
01:39:58.650 --> 01:40:07.653
As it relates to TSMC, it's interesting that both TSMC and Samsung are already working on integrating graphene into their industrial process.
1783
01:40:08.133 --> 01:40:11.154
As if we needed another reason for Taiwan to be a geopolitical...
1784
01:40:11.594 --> 01:40:25.933
uh, yeah, uh, matchbox, um, particularly given the stakes of the perceived and real, uh, benefits of the race as it relates to AI writ large.
1785
01:40:26.273 --> 01:40:26.473
Yeah.
1786
01:40:26.754 --> 01:40:26.974
Um, yeah,
1787
01:40:28.495 --> 01:40:32.779
This is going to be – and again, we talk about breaking science research papers on this story.
1788
01:40:33.139 --> 01:40:41.065
There is a delta between or a distance between it working in a lab and it being in your phone at home or in a data center in space.
1789
01:40:41.525 --> 01:40:54.635
But it is a – we're trying to continue to build this roadmap for people to understand where we're going as a society and what some of the best and brightest among us are trying to bring to us.
1790
01:40:55.255 --> 01:40:59.498
And again, we talked about this in the context of data centers in space.
1791
01:41:00.058 --> 01:41:07.323
But there are other things that you've mentioned, you know, that would be used.
1792
01:41:07.383 --> 01:41:09.485
This will be useful for an extreme computing.
1793
01:41:09.565 --> 01:41:10.125
That's right.
1794
01:41:10.185 --> 01:41:11.526
Exploration and other things.
1795
01:41:11.726 --> 01:41:12.967
So space exploration.
1796
01:41:13.047 --> 01:41:14.988
I'd like to get better photos of Venus.
1797
01:41:15.008 --> 01:41:15.368
Yeah.
1798
01:41:15.609 --> 01:41:15.849
Yeah.
1799
01:41:15.989 --> 01:41:17.010
That'd be really nice.
1800
01:41:17.030 --> 01:41:17.570
That'd be great.
1801
01:41:18.170 --> 01:41:21.091
I'd like to have a rover on Venus that's like running around and stuff.
1802
01:41:21.111 --> 01:41:21.771
That would be cool.
1803
01:41:23.031 --> 01:41:26.832
Heavy energy, like geothermal, nuclear monitoring.
1804
01:41:26.872 --> 01:41:30.993
I'd like to have safer nuclear power.
1805
01:41:31.853 --> 01:41:43.415
If you're a green or clean energy person, I mean, this can have real, real impacts in that space because it can help us better both understand and also execute in terms of production.
1806
01:41:43.435 --> 01:41:43.535
Yeah.
1807
01:41:44.195 --> 01:42:09.454
maintenance operational aspects yeah really anything that requires high temperature computing right i mean with the advent of um fusion right if we're trying to make fusion a thing stuff is hot yep and so if we want um real-time monitoring and decision making using some ai algorithm about how to confine the plasma i'm just thinking out loud right like that's something that this would be used for i don't need like to cool it down to room temp i can just cool it down to 700 degrees
1808
01:42:10.630 --> 01:42:12.852
Who needs a room temperature semiconductor?
1809
01:42:12.972 --> 01:42:13.252
Yeah.
1810
01:42:14.153 --> 01:42:14.813
Well, we still want that.
1811
01:42:14.833 --> 01:42:15.414
We still want that.
1812
01:42:15.474 --> 01:42:17.115
We still want that.
1813
01:42:17.295 --> 01:42:21.799
This was a really great electronics that can survive lava.
1814
01:42:21.939 --> 01:42:30.526
Again, this was a recent science paper, March 2026, a collaboration led by researchers and
1815
01:42:30.958 --> 01:42:49.488
here in California Forever Goodbye at USC, alongside the Air Force Research Lab in Dayton, Ohio, and the Kumamoto University in Japan, with funding from the National Science Foundation, the Army Research Office, and the Air Force Research Lab.
1816
01:42:50.808 --> 01:42:56.491
Can't wait to see high-temperature immersers in a hypersonic vehicle near you.
1817
01:42:58.072 --> 01:43:07.917
Because certainly they're going to be using it for command and control systems integrated with, you know, next generation military stuff.
1818
01:43:07.937 --> 01:43:12.500
So before they do that, we would like to have it be used for the good of mankind.
1819
01:43:13.140 --> 01:43:15.581
Do you have any last notes before we wrap up today?
1820
01:43:15.802 --> 01:43:22.725
No, if you want to leave a comment, think of a really funny application for high temperature computing.
1821
01:43:22.745 --> 01:43:24.006
Where would you want to put a chip?
1822
01:43:25.507 --> 01:43:26.107
Run wild.
1823
01:43:26.167 --> 01:43:27.728
Keep it PG, everybody.
1824
01:43:27.788 --> 01:43:29.609
This is a family-friendly show.
1825
01:43:30.769 --> 01:43:43.434
As always, my name is Lester Nari, joined by my co-host, our resident data centers in space, our possible PhD, Krishna Chowdhury.
1826
01:43:43.934 --> 01:43:45.435
We are approaching our world.
1827
01:43:46.115 --> 01:43:47.977
one year anniversary.
1828
01:43:48.457 --> 01:43:52.581
So be sure to be tuned in for our anniversary episode.
1829
01:43:52.941 --> 01:43:55.744
For our longtime listeners, we are extremely grateful.
1830
01:43:56.324 --> 01:44:03.911
If you just happen to be listening to us for the first time, Catalog is a bunch of episodes that are just like this.
1831
01:44:04.111 --> 01:44:09.136
50 of some of the best science entertainment on the planet.
1832
01:44:09.616 --> 01:44:11.658
We will see you all next week.

Edit transcript

SEARCH PAST EPISODES

Search past episodes of From First Principles.

Embed this search bar to your website

Disclaimer: The podcast and artwork embedded on this page are from Krishna Choudhary and Lester Nare, which is the property of its owner and not affiliated with or endorsed by Listen Notes, Inc.

PROMOTED

We’re selling some of our podcast-related domain names, including breaker.fm, Podcast.mov, and Listen.Host. Place your bid on eBay.

We’re selling some of our podcast-related domain names, including breaker.fm, Podcast.mov, and Listen.Host. Place your bid on eBay.

ELSEWHERE

Listen on Spotify

Listen on Amazon Music

RECOMMENDATIONS

Use API to fetch this episode