Latest / From First Principles / The Tech Elon Has Been Waiting For (EP 51)

Transcript

The Tech Elon Has Been Waiting For (EP 51) · From First Principles

• 0:00
  Industrial flash memory tops out at around 200 degrees Celsius.
• 0:05
  And above that, there's essentially nothing that works.
• 0:08
  It's a Memristor, which is a memory switching device.
• 0:11
  And it's made from graphene, hafnium oxide and tungsten.
• 0:14
  And it operates reliably at 700 degrees Celsius.
• 0:19
  Really cool.
• 0:19
  I mean, it's hotter than molten aluminum.
• 0:22
  Aluminum melts at 660 degrees Celsius.
• 0:25
  If you put this thing in a vat of aluminum, it will still retain memory.
• 0:29
  Hello, Internet.
• 0:30
  This is your captain speaking, Lester Nare, joined as always by my co-host and our resident PhD, Krishna Chowdhury.
• 0:41
  We have a special deep dive today for you all.
• 0:45
  Be sure to check out our rundown from last week.
• 0:48
  Be sure to heat up the future of electronics as we look at a breakthrough in high temperature memory.
• 0:56
  This is something that Elon has been waiting for for a long time that may enable electronics that can survive lava.
• 1:05
  I'm going to need the details.
• 1:07
  I'm going to need to understand what's going on here.
• 1:09
  This comes from a new research paper in the Journal of Science, shout out to science, from March 26, 2026.
• 1:14
  It was a collaboration between researchers at USC here in California, the Air Force Research Lab in Dayton, Ohio, and the Kumamoto Research Lab.
• 1:24
  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
• 2:05
  Today we're covering an invention that I think is going to be a game changer for the future of computing.
• 2:10
  You know, we live in the digital age of electronics.
• 2:14
  Chips are the bedrock of our economy.
• 2:16
  And yet all of those chips really have a Goldilocks zone when it comes to when they work.
• 2:22
  Goldilocks in the traditional sense of like not too hot, not too cold.
• 2:26
  This is...
• 2:28
  research is getting to that not too hot part.
• 2:31
  You know, the Papa Bear that was like too hot, the porridge was too hot.
• 2:35
  Now we're getting a chip that can withstand that porridge.
• 2:39
  Okay.
• 2:41
  These are memory chips that happen in our phone, in our laptop, in the cameras that we have here in the studio.
• 2:46
  They're all designed to work reliably roughly between 85 degrees Celsius, so 200 Fahrenheit for our Americans.
• 2:54
  Okay.
• 2:55
  to about 150 degrees Celsius, so 300 Fahrenheit, okay?
• 2:59
  That's where they sit.
• 3:02
  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.
• 3:14
  And above that, there's essentially nothing that works, okay?
• 3:19
  There's no computation happening above 300 degrees Celsius.
• 3:23
  But that doesn't mean that we don't want to have computation happening at that temperature.
• 3:28
  There are a lot of applications that even I can think of here, not being in the field, that would be really great.
• 3:35
  to have at 300 degrees or above.
• 3:37
  Okay, for example, oil and gas.
• 3:39
  Oil and gas drillers, they push deep into the Earth's crust.
• 3:42
  The deeper you get into the Earth's crust, the closer you're getting to the mantle where all of the magma is.
• 3:49
  So the temperature is going to go up.
• 3:51
  At 5 kilometers, you're flirting with like 150 degrees Celsius.
• 3:55
  That's already pushing it.
• 3:56
  At 10 kilometers, you're well past 300 degrees.
• 3:59
  And the sensor packages that we have today, when you send them down these boreholes...
• 4:05
  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.
• 4:13
  That's always not convenient.
• 4:16
  Not ideal.
• 4:17
  Yeah.
• 4:17
  And then they have to transmit their data to the surface in real time.
• 4:21
  They can't do any computation down there.
• 4:23
  Interesting.
• 4:24
  So then there's this like massive wiring problem that you have to deal with.
• 4:28
  The wires are going through this really hot process.
• 4:32
  The engineering is way more complex unnecessarily if you could just have it.
• 4:35
  Yes.
• 4:35
  If you could just have like a chip that's right on top of the drill, right?
• 4:40
  That's right there.
• 4:41
  Then you can have something like an autonomous downhole.
• 4:44
  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
• 5:05
  Yeah, well, I mean, but even the friction is going to create like a hot, yeah, some heat.
• 5:10
  You know, we're going to want to be able to do that.
• 5:11
  Yeah, we just look.
• 5:12
  Planetary defense.
• 5:13
  Yes, as always, you know, we are defense first here, not war, only defense.
• 5:19
  Now, you mentioned hypersonics.
• 5:22
  That's another big thing, right?
• 5:23
  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.
• 5:34
  Right.
• 5:35
  We've covered this in previous stories where the temperature gets so high that chemistry starts mattering.
• 5:40
  It gets a little weird.
• 5:40
  Yeah, and it starts getting a little weird.
• 5:43
  And we can't make any intelligent decisions on site because of the chip.
• 5:48
  The chip is just going to get fried.
• 5:50
  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.
• 5:57
  Right.
• 5:58
  This is the number one conversation in SF right now.
• 6:00
  Yeah.
• 6:01
  Yeah.
• 6:01
  I mean, Elon wants to put data centers in space.
• 6:03
  He's not the only one.
• 6:05
  SpaceX had an IPO of one point seven trillion dollars.
• 6:09
  And I think a lot of that hinges on Elon's promise that he can put AI data centers in space.
• 6:15
  We're going to get into the details there, but there's been a lot of pushback.
• 6:21
  Because of fundamental physics, right?
• 6:23
  They're saying, oh, like fundamental physics is telling you that you can't put data centers in space.
• 6:28
  Well, yes and no.
• 6:30
  Okay.
• 6:31
  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.
• 6:41
  But with this problem...
• 6:44
  You can have innovation like the one that we're going to talk about, and perhaps it's possible.
• 6:47
  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?
• 6:58
  That's because the space is a vacuum.
• 7:00
  There's no way to extract heat other than radiation.
• 7:04
  And radiation is not a great way to let go of heat because you're relying on photons.
• 7:10
  It's not – is it – it's like because it's not efficient, right?
• 7:13
  Yes, exactly.
• 7:14
  It's just not efficient.
• 7:15
  We're going to get into some of the physics of that later.
• 7:18
  And my favorite sort of conundrum that I think this is going to solve is planetary exploration.
• 7:25
  I think you're going to like this too.
• 7:28
  There's a reason why we have so many photos of Mars and not that many photos of Venus.
• 7:33
  Okay.
• 7:34
  Mars is pretty chill to go to.
• 7:36
  It's like kind of cold.
• 7:37
  Literally.
• 7:38
  It's like literally it's farther away from the sun and it's fine.
• 7:41
  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.
• 7:48
  Actually, the first guys to do it were the Soviets.
• 7:51
  Venus has a surface temperature of about 465 degrees Celsius and an atmospheric pressure that's 92 times that of the Earth.
• 8:01
  The Soviets put down the Soviet Venera landers, and they're the only spacecraft that have ever survived on the Venus surface.
• 8:10
  That's unbelievable.
• 8:10
  But guess how long they survived?
• 8:13
  Over or under an hour?
• 8:15
  It's got to be under an hour.
• 8:16
  It's under an hour, yeah.
• 8:18
  It's between like 23 minutes.
• 8:19
  I think their record was like two hours later on.
• 8:22
  But like two hours?
• 8:24
  Yeah, yeah.
• 8:25
  No.
• 8:25
  Right?
• 8:25
  Like the stuff that JPL puts on Mars has lasted like 20 years.
• 8:30
  Yeah, we're talking about two decades versus an hour.
• 8:33
  Yes, versus two hours.
• 8:34
  Right, right.
• 8:34
  Two hours.
• 8:35
  Is the best that they can do.
• 8:36
  I will note, this is, for those who are TV fans –
• 8:40
  Uh, we've talked about for all mankind on the show before, but Apple has a second.
• 8:45
  Apple TV has a second show called star city.
• 8:47
  Okay.
• 8:47
  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.
• 8:57
  Part of it relates to the Venus mission.
• 8:59
  Really?
• 8:59
  Dude, I got to watch that.
• 9:01
  It's, it's, it's alternative reality.
• 9:03
  So, you know, no spoilers.
• 9:05
  I love, I love that kind of stuff.
• 9:06
  But it is interesting to kind of understand the,
• 9:09
  The technical challenges, which they brushed on at a high level in the show.
• 9:14
  But recently I've been watching it and this brought it up as a concept.
• 9:18
  And the engineering problem is non-trivial.
• 9:21
  Yeah, no, completely, right?
• 9:23
  It's like going towards the center of the earth.
• 9:25
  It's the same drilling thing.
• 9:26
  High pressure, high temperature.
• 9:28
  Your electronics are going to get completely wrecked.
• 9:30
  Right.
• 9:31
  The photo we have from the Venera 9 in 1975.
• 9:34
  This is the best photo they got.
• 9:36
  It's the first image of a planet other than the Earth.
• 9:39
  Credit to the Soviets.
• 9:41
  The first image of a planet other than the Earth was not the United States and Mars.
• 9:45
  It was Venus.
• 9:47
  And that's all they got.
• 9:48
  But still, that's kind of crazy.
• 9:50
  But who got to the moon, though?
• 9:51
  Yeah, that's the real question.
• 9:53
  They were like, okay, let's go the other direction.
• 9:57
  The race is actually this way, guys.
• 9:59
  Later missions, they were actually able to get some color photos.
• 10:03
  This is the one that lasted two hours.
• 10:05
  And you could get some color photos.
• 10:07
  You can see the sort of yellow haze that we know that the Venus has if you point a telescope to it.
• 10:12
  But again, only two hours.
• 10:13
  And you can see it kind of looks like the surface of a volcanic planet.
• 10:18
  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
• 10:40
  Exactly.
• 10:40
  I mean, I'd like to know more about the surface of Venus, right?
• 10:44
  We have better surface images of Mars, Pluto.
• 10:49
  We have better surface images of Pluto than we do with large swaths of Venus, which is our nearest neighbor.
• 10:55
  Venus is closer to us than the Earth.
• 10:57
  And we still don't know what's under those clouds.
• 10:59
  That's so unbelievable.
• 11:00
  Right?
• 11:01
  And it's just because it's a completely...
• 11:05
  Messed up situation to get down there on the surface.
• 11:08
  Everything fails.
• 11:09
  We can't stand the heat, so we can't go in the kitchen.
• 11:11
  Yeah, exactly.
• 11:13
  Exactly.
• 11:14
  And into this landscape comes this paper that was published by USC.
• 11:19
  It's describing something I think genuinely new.
• 11:22
  Okay.
• 11:23
  It's a Memristor, which is a memory switching device.
• 11:25
  We're going to get into what that is.
• 11:27
  And it's made from graphene, hafnium oxide, and tungsten.
• 11:31
  And it operates reliably at 700 degrees Celsius.
• 11:35
  That's unreal.
• 11:37
  700 degrees.
• 11:37
  That's unreal.
• 11:38
  Yeah.
• 11:39
  They've raised the temperature by a factor of more than two.
• 11:43
  And so there's something fundamental.
• 11:45
  There's a real breakthrough here that's beyond an iterative upgrade.
• 11:50
  Yeah, I think it's really cool.
• 11:52
  I mean it's hotter than molten aluminum.
• 11:55
  Aluminum melts at 660 degrees Celsius.
• 11:58
  If you put this thing in a vat of aluminum, it will still retain memory.
• 12:02
  It's hotter than the surface of a lava flow on Earth.
• 12:06
  Not all lava flows.
• 12:07
  I know there's going to be people in the comments being like, no.
• 12:10
  The volcanologist?
• 12:10
  Yeah, yeah.
• 12:11
  The volcanologist is going to be like, no, lava is hotter than that.
• 12:13
  Actually, there is a volcano in Tanzania, the old Doinyo Lengai volcano, where the lava is 600 degrees Celsius.
• 12:23
  Okay?
• 12:24
  So there exists a lava...
• 12:26
  On Earth where I could put this chip and it would still work fine.
• 12:29
  So the tagline still kind of works, right?
• 12:32
  If you took it to Hawaii and put it in the lava in Hawaii, it would not.
• 12:37
  I think Hawaii's is like a thousand plus.
• 12:39
  But it perhaps might, and we'll get into that later.
• 12:42
  We're on the path.
• 12:43
  Yeah.
• 12:45
  The other crazy thing about this is the chip, the Memrister device, it maintained stable data retention for 50 hours.
• 12:54
  Really?
• 12:55
  50 hours.
• 12:55
  And it retained...
• 12:57
  That data retention after a billion switching cycles without failure.
• 13:02
  So this is both high temperature and long duration.
• 13:06
  Yes.
• 13:07
  Comparatively to where we were before.
• 13:08
  Yeah.
• 13:09
  And the physics is actually like it gives you both.
• 13:12
  Which is because both are different.
• 13:14
  Yeah.
• 13:15
  It's like a graph on different axes.
• 13:17
  And you can have high temperature.
• 13:18
  It's like, OK, great.
• 13:19
  But it's a microsecond.
• 13:20
  Yeah.
• 13:20
  Yeah.
• 13:20
  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.
• 13:27
  Yeah, that is very interesting.
• 13:28
  I think it's by a large margin the highest temperature at which any non-volatile memory has ever been demonstrated to work.
• 13:35
  This is phenomenal.
• 13:36
  Yeah.
• 13:36
  It's really, really cool.
• 13:37
  And I mean, why should normal people care, OK?
• 13:40
  Not scientists.
• 13:41
  Sure.
• 13:41
  Well, I think the border between environments where we can put intelligent electronics and environments where we cannot is about to shift dramatically.
• 13:51
  And then our understanding of everything about the world around us changes in a very big way.
• 13:55
  Yeah, I mean, it affects how we drill energy, how we explore planets, how we monitor volcanoes, how we put data centers in space.
• 14:03
  That has a lot of social and political implications for our society because then maybe we don't need to –
• 14:11
  Like take everyone's water.
• 14:13
  If you don't want a data center in your backyard, this story matters a lot because we can just put them in space.
• 14:19
  Yeah.
• 14:20
  Yeah.
• 14:20
  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.
• 14:28
  But I think it's going to expand the frontier of what our civilization can do.
• 14:33
  You're saying there's a chance.
• 14:34
  Yeah.
• 14:34
  And I think I think I think it's an incredible story.
• 14:37
  We are going to get into the details of this has been such a big conversation in the business and technology world.
• 14:48
  We are so over leveraged on AI in the US in terms of its impact on our GDP.
• 14:54
  Everyone's saying it's fine because we're going to put data centers in space.
• 14:58
  And it's been very contentious as a.
• 15:02
  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
• 15:17
  Welcome back.
• 15:18
  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.
• 15:40
  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.
• 15:51
  Our website has all of our episodes, all of the research papers and show notes that we cover.
• 15:57
  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.
• 16:08
  You can check out our research paper library.
• 16:10
  If you want to support the show, you can donate at ffppod.com backslash donate.
• 16:16
  And you can also check out the merch shop here.
• 16:19
  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
• 16:44
  And then it would be all rare.
• 16:46
  Gordon Ramsay would love that.
• 16:48
  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.
• 16:58
  Why I can't bring my iPhone and put it in a lava flow.
• 17:00
  Yeah, yeah, yeah.
• 17:01
  I'd like to.
• 17:02
  Yeah, I could put it in water.
• 17:03
  Yeah, but...
• 17:05
  Lava is something you can't do, okay?
• 17:07
  The physics of high temperature failure.
• 17:10
  We first have to understand the traditional CMOS architecture that is at the base of all of our technology.
• 17:16
  This is complementary metal oxide semiconductor.
• 17:19
  These devices fail at around 200 degrees Celsius.
• 17:22
  Here's how it usually works, okay?
• 17:25
  So at zero Kelvin, which means absolute zero, everything is stationary.
• 17:30
  Silicon is an insulator.
• 17:34
  What ends up happening is atoms have something called a conduction and a valence electron shell.
• 17:41
  Okay?
• 17:41
  They've got parts where valence electrons can occupy all of the shell.
• 17:48
  So there's nothing going on.
• 17:49
  All of the shells are occupied, right?
• 17:51
  Then there's a conduction shell.
• 17:53
  There's a conduction energy level where there's like one or two electrons hanging out.
• 17:57
  And those electrons are the ones that skip from one atom to the next to the next.
• 18:01
  And they create current, which is what we use.
• 18:03
  Okay?
• 18:04
  Now, at zero degrees Kelvin, all of silicon's valence band is completely filled.
• 18:13
  If you put a bunch of metal and a bunch of silicon together, the gap...
• 18:18
  between the conduction band and the valence band gets bigger just because of neighboring effects.
• 18:25
  It's like the neighboring atoms start pulling on your own energy gaps, and it expands the gap between these two energy bands.
• 18:35
  The wiggle craze starts getting more intense.
• 18:38
  Exactly.
• 18:38
  And so at finite temperature, what you can do is you can bump an electron –
• 18:42
  You get some wiggle, right?
• 18:43
  Because like there's jiggling happening.
• 18:45
  There's photons that are at some energy level.
• 18:48
  That's going to come in, strike an electron.
• 18:50
  The electron is going to move up to the conduction band.
• 18:52
  And then voila, you can have a little bit of conduction.
• 18:55
  Okay.
• 18:56
  Now let's get into exactly how this happens.
• 18:59
  We've got a video that shows exactly how this conduction and valence band works.
• 19:04
  So on the bottom, you've got your silicon.
• 19:07
  You've got all of the sites that are now occupied by electrons, which are in the
• 19:12
  In the blue.
• 19:12
  It's like the housing market.
• 19:13
  Yeah.
• 19:14
  There's nothing on the market.
• 19:14
  Yeah, yeah, yeah.
• 19:15
  There's no openings.
• 19:16
  Now, all of a sudden, somebody sells a house.
• 19:18
  So an electron from the valence band moves into the conduction band.
• 19:24
  Now that conduction band electron can move around and create a current.
• 19:28
  Notice at the same time, the hole that was left in the valence band can also move around.
• 19:34
  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.
• 19:44
  That's a positive thing.
• 19:46
  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
• 20:09
  Thing that matters in the valence band is that positive hole that's moving around, right?
• 20:16
  It unlocks the stationary state that exists because it's now moved up to.
• 20:22
  Yeah, yeah, yeah.
• 20:23
  Exactly.
• 20:23
  And so you can have two different types of semiconductors here.
• 20:26
  Engineers can make silicon switches using either phosphorus or boron.
• 20:31
  You can dope.
• 20:32
  the silicon with either phosphorus or boron.
• 20:34
  And then these artificially dictate how many charge carriers you're gonna get.
• 20:39
  Are you gonna get more electrons in the conduction band?
• 20:41
  Are you gonna get more holes?
• 20:43
  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.
• 20:53
  The P-type is the positive type semiconductor because the holes are the ones that are moving around.
• 20:57
  And when you combine the two, you get things like LEDs, transistors,
• 21:02
  But at too high of a temperature, you're going to get wrecked.
• 21:05
  Why?
• 21:06
  Because if you have too high of a temperature, everything is going to go into the conduction band because there's so much jiggle.
• 21:12
  And then I'm just going to get a short circuit.
• 21:13
  It's always going to be conducting.
• 21:15
  With a transistor, I'd like to control when it's conducting current for a 1.0.
• 21:20
  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.
• 21:31
  Everything moves.
• 21:32
  Okay.
• 21:32
  And so part of what we're saying is we're trying to engineer a level of control of how this movement happens.
• 21:38
  And at high temperatures currently in our traditional, uh, chips and systems, they are unable at high temperatures.
• 21:46
  You lose the ability to engineer and control.
• 21:49
  Yeah.
• 21:49
  The jiggle, the movement, how things move to the conduction layer and back accordingly.
• 21:55
  Yeah, you can't control the transistor going from a 1 to a 0.
• 21:58
  It's just always going to be a 1.
• 21:59
  Right, right.
• 22:00
  Because it's always – Yeah, and that's the bedrock of computation.
• 22:03
  Right, right.
• 22:04
  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.
• 22:14
  And then you can preserve this transistor property.
• 22:17
  But the authors of the paper that we're talking about, they cite this research from NASA Research.
• 22:25
  It shows that you can get all the way up to 800 degrees Celsius.
• 22:28
  But memory is still an issue.
• 22:32
  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,
• 22:42
  The stuff that's in our RAM.
• 22:44
  I turn off my computer.
• 22:45
  It's not like I'm going to lose everything.
• 22:48
  That kind of stuff is still not up to that scale.
• 22:53
  So through public funding, we discovered that it's possible to go above 800.
• 22:57
  But the issue still remains, memory does not, in this context, exist.
• 23:04
  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
• 23:22
  But like your circuit is doing fine.
• 23:24
  What's it going to do?
• 23:25
  There's no instructions.
• 23:26
  Right.
• 23:26
  Right.
• 23:27
  Right.
• 23:27
  So most of the applications that we have like in our computer, in the phone, they use something called flash memory.
• 23:34
  This also uses a transistor.
• 23:35
  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.
• 23:46
  Right.
• 23:47
  And if even at like a low temperature, like.
• 23:51
  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
• 24:01
  So they can't run forever.
• 24:04
  There's deterioration.
• 24:06
  Doing work deteriorates just like our own human body.
• 24:09
  Yeah.
• 24:10
  As an athlete, you can't be an athlete forever.
• 24:12
  Your body eventually gives up.
• 24:14
  Similarly, in this context, flash memory deteriorates at some time interval.
• 24:21
  Exactly.
• 24:22
  For use.
• 24:22
  Exactly.
• 24:23
  And so because of that, people have been slowly looking at other substrates of computation.
• 24:29
  Yeah.
• 24:29
  Specifically, something called resistive memory, R-RAM.
• 24:35
  This is resistive random access memory.
• 24:37
  And to understand that, we really have to understand something called the mem-rister.
• 24:43
  Let's go back to 1970.
• 24:45
  There are three types of fundamental circuit components.
• 24:49
  The resistor, the inductor, and the capacitor.
• 24:51
  This is the stuff of undergraduate physics.
• 24:54
  And if you're an electrical engineer, I'm really sorry for your loss.
• 24:59
  But you probably deal with this every single day, right?
• 25:04
  A resistor is something that resists electric current.
• 25:08
  So it drops a voltage across a resistor.
• 25:11
  And an inductor is something that takes a changing current and creates another current that's going in the opposite direction.
• 25:20
  So it kind of stores energy using a change in current.
• 25:23
  And a capacitor is something that stores energy using charge.
• 25:26
  Okay?
• 25:26
  Okay.
• 25:27
  Now, Leon Chua, in 1971, he was at UC Berkeley, he said, let's look at these three components that we've got.
• 25:36
  We've got a capacitor, right?
• 25:38
  A capacitor is something that has a relationship between charge and voltage.
• 25:44
  If I store a bunch of charge, that stores a voltage difference.
• 25:48
  I've got a resistor.
• 25:49
  That's a relationship between voltage and current, right?
• 25:52
  As current goes through, there's a voltage drop.
• 25:55
  Right?
• 25:56
  And I've got an inductor.
• 25:57
  That's something that has to do with current and magnetic flux.
• 26:02
  Okay?
• 26:05
  There's four variables, though.
• 26:07
  Right.
• 26:08
  And there's only three components.
• 26:10
  Right.
• 26:11
  So for those who may be audio listeners, we're looking at a square.
• 26:14
  It kind of looks like those, uh, those one, two, three, four.
• 26:18
  I don't know what you call it.
• 26:21
  Yeah.
• 26:21
  Yeah.
• 26:22
  The thing we used to do in elementary school.
• 26:24
  Yeah.
• 26:24
  That would tell me like what my fate was.
• 26:26
  Yeah.
• 26:26
  The paper thing.
• 26:27
  Yeah.
• 26:27
  You can let us know in the comments.
• 26:28
  Yeah.
• 26:29
  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,
• 26:38
  The point here being we've talked about every relationship but one side of the square.
• 26:42
  But one side of the square, right?
• 26:44
  There should be Leon Chua.
• 26:47
  decided there should be a relationship between flux and charge.
• 26:52
  The flux capacitor.
• 26:54
  In some sense, yeah.
• 26:56
  And he deemed this thing the memristor.
• 26:58
  He said there should be something out there, okay?
• 27:00
  This is a component where the resistance changes dynamically based on the charge flow.
• 27:05
  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.
• 27:13
  Okay?
• 27:14
  Okay.
• 27:15
  The telltale sign, he also showed, okay, if experimenters are looking at this thing, he was a theorist.
• 27:21
  And he said, if experimenters are looking at this thing, all you got to do is look at the IV curve.
• 27:26
  Okay?
• 27:29
  The IV curve is the current and the voltage.
• 27:32
  Current on the Y.
• 27:32
  On the two axes.
• 27:33
  Voltage on the X.
• 27:34
  Yes, okay?
• 27:35
  And if we look at the IV curves, they're all very...
• 27:39
  Unique.
• 27:40
  On the top left, that's a normal resistor.
• 27:42
  Okay.
• 27:43
  As I increase the voltage, I'm going to increase the current across the resistor.
• 27:49
  And there you've got a line.
• 27:50
  This is a nonlinear resistor.
• 27:52
  A linear resistor would be like the stuff that we do in undergrad physics is just like, oh, V equals IR.
• 27:57
  That's not entirely true, right?
• 27:59
  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.
• 28:08
  And then at high voltage, the resistance is way higher.
• 28:11
  Things like that.
• 28:12
  Yeah.
• 28:13
  No, it's the other way around.
• 28:14
  At a high voltage, the resistance is lower because there's more pull.
• 28:18
  So it's easier to shuttle electrons through.
• 28:21
  Okay, so that's what we're seeing in the top left corner.
• 28:24
  On the top right corner and the bottom left corner, we're seeing the IV curves for a capacitor and an inductor.
• 28:31
  Those things cycle around.
• 28:34
  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.
• 28:45
  If you combine a capacitor and an inductor, the capacitor gets charged and the inductor doesn't get charged.
• 28:51
  And then the inductor gets charged, the capacitor gets discharged.
• 28:53
  So there's a cycling that happens.
• 28:55
  And that is something that is characteristic of like sine and cosine.
• 28:59
  You got the unit circle.
• 29:00
  The x-axis is sine.
• 29:02
  The y-axis is cosine.
• 29:03
  Great.
• 29:04
  Leon Chua said, I should be looking for something called a memristor where charge and magnetic flux are related to one another.
• 29:16
  And what I would see is a curve that looks like a loop, a figure eight loop.
• 29:24
  And the challenge is there is no way to combine the other three curves to get that fourth curve.
• 29:33
  Interesting.
• 29:34
  Okay?
• 29:34
  It is an independent element in electronics.
• 29:38
  It's fundamental.
• 29:39
  It's not derivative.
• 29:40
  Yes, exactly.
• 29:41
  You know what?
• 29:42
  You see what I'm saying?
• 29:42
  Yeah.
• 29:43
  It's like you can't make an electron out of quarks.
• 29:47
  Right.
• 29:48
  Right.
• 29:48
  Similarly, you can't make a memristor by just cleverly placing resistors, capacitors, and inductors together.
• 29:55
  This is an independent component that is one of the four fundamental components of electronics.
• 30:01
  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.
• 30:12
  Yeah.
• 30:13
  And it's just a mathematical argument, which is so cool to think about.
• 30:17
  Right.
• 30:18
  Right.
• 30:18
  That it's just like there's four variables.
• 30:20
  We've only got three components.
• 30:22
  There should be a fourth one.
• 30:23
  Theorists have value, everybody, okay?
• 30:26
  It's very cool.
• 30:26
  And one more thing I want to say about the Memorister IV curve, right?
• 30:31
  Notice that for a capacitor and an inductor, at zero voltage, there is a non-zero current, okay?
• 30:40
  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?
• 30:50
  A resistor could store data
• 30:52
  Because at a zero voltage, there's zero current, but it only stores a one.
• 30:58
  It's whatever the resistance is.
• 30:59
  So it's not really storing anything.
• 31:01
  If we go back to Claude Shannon's information theory, I need two things to store data.
• 31:06
  I need a zero and a one.
• 31:07
  A memory store, on the other hand, at zero voltage, it can be in two states.
• 31:11
  It can be in a low resistance state.
• 31:13
  Right.
• 31:14
  or a high resistance state.
• 31:15
  This is so good.
• 31:16
  Right?
• 31:16
  This is so good.
• 31:17
  Now all of a sudden people start paying attention because this could be a substrate for non-volatile memory.
• 31:24
  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.
• 31:32
  If I can sense that, that's my zero and one.
• 31:35
  computers as a quick visual analogy here, because we're basically looking at this figure eight.
• 31:39
  That's kind of like a racetrack.
• 31:40
  You remember Hot Wheels as a kid?
• 31:42
  Yeah.
• 31:42
  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.
• 31:49
  And that's kind of what we're saying where it can be in two states at that center point.
• 31:52
  It can be on the top part of the Hot Wheels track or,
• 31:55
  or on the bottom part of the Hot Wheels track.
• 31:57
  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.
• 32:05
  Exactly, yeah.
• 32:05
  And for the Formula One fans, the only circuit that does this is Suzuka in Japan.
• 32:11
  There's a figure eight that's beautiful.
• 32:13
  One day we'll go.
• 32:15
  So the theorist has done his job.
• 32:17
  He's like, there should be this thing.
• 32:20
  And specifically, this is what you guys should look for.
• 32:24
  It takes 40 years.
• 32:26
  Oh, my God.
• 32:27
  Nearly 40 years.
• 32:29
  Stanley Williams at Hewlett Packard Labs.
• 32:32
  So that's HP Labs.
• 32:34
  I love these nature articles that are just like bombshell articles because the title is just a banger every single time.
• 32:42
  The missing membrists are found.
• 32:45
  It's nearly 40 years after Leon Chua postulated that this fourth passive element should exist.
• 32:51
  Yeah.
• 32:54
  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.
• 33:06
  I just want to note that we're talking about 2008.
• 33:08
  Yeah.
• 33:10
  Very soon.
• 33:11
  We were in high school.
• 33:12
  Mm-hmm.
• 33:13
  So it wasn't that crazy.
• 33:15
  It wasn't that long ago.
• 33:16
  It wasn't that long ago.
• 33:17
  No, we were in high school when the first Memrister, like when it went from theory to reality.
• 33:22
  Right.
• 33:23
  And it's still at a research stage level.
• 33:26
  We're not talking about in-end consumer products.
• 33:29
  Yeah.
• 33:30
  The iPhone came out a year before this.
• 33:32
  Yeah, that's crazy.
• 33:34
  Just to put it in time context.
• 33:36
  Wow.
• 33:37
  Which is... That's a good way to... Yeah, that's a good way to think about it.
• 33:40
  Yeah.
• 33:40
  I mean, we had an iPhone before we even knew that a Memorister was possible.
• 33:44
  Which is so crazy.
• 33:45
  That's crazy.
• 33:46
  Knowing what we use it for now.
• 33:47
  Yeah.
• 33:48
  Yeah.
• 33:48
  I mean, it's insane.
• 33:49
  Okay.
• 33:49
  So let's talk about this valence charge mechanism that I was telling you about, about how the Memorister works.
• 33:55
  Here's the idea.
• 33:56
  So I told you, right?
• 33:57
  He got a titanium electrode.
• 34:00
  A platinum electrode.
• 34:01
  And in the middle were oxygen vacancies.
• 34:05
  Now, vacancies is something that I alluded to earlier when I was talking about the holes.
• 34:09
  The holes are vacancies in the sense that I have taken something out.
• 34:14
  But now those holes where I took that thing out, those are mobile vacancies.
• 34:18
  just like the stuff that I took out.
• 34:21
  Those are mobile elements.
• 34:22
  And if those mobile elements can rearrange, they can create electronic components on their own, right?
• 34:29
  In the same way that a P-type
• 34:31
  Semiconductor can use the holes to do all of its electronics.
• 34:36
  The point being that vacancy is not stationary.
• 34:40
  It has the ability independent of the thing that left to move on its own.
• 34:44
  Yeah, exactly.
• 34:45
  Okay.
• 34:46
  So let's get into a little bit more about how these oxygen vacancies work because this is going to be the central –
• 34:53
  thing we need to understand about why the physics of this modern paper is working.
• 34:57
  Got it.
• 34:58
  Okay?
• 35:00
  When you have these two electrodes, you've got a positive electrode and a negative electrode, right?
• 35:08
  What you do is a switching process, okay?
• 35:11
  So this is effectively how it works.
• 35:15
  Let's say I've got a top electrode and a bottom electrode.
• 35:18
  In this case, it's like silver and titanium on the bottom.
• 35:23
  But it doesn't really matter what the identity of the two electrodes are yet.
• 35:27
  It's going to matter later.
• 35:29
  What happens is, in order to understand what this valence charge mechanism is, let's try to think of an analogy, a simple analogy.
• 35:36
  Okay?
• 35:36
  Imagine like a really thick, densely packed forest, like the forests of Endor in the Star Wars Episode VI Return of the Jedi.
• 35:47
  Okay.
• 35:47
  Okay?
• 35:48
  Okay.
• 35:48
  Now, if you remember, the Forest of Endor, they had those motorbike thingies.
• 35:54
  I actually looked up what they were called.
• 35:56
  They were called forest motorbikes.
• 35:58
  74Z speeder bikes.
• 36:00
  Okay?
• 36:01
  Remember, they zoomed through the forest at really, really high speed.
• 36:07
  Now, imagine a really, really dense forest.
• 36:09
  Okay?
• 36:10
  If you've got a speeder bike in a really, really dense forest, you're going to crash.
• 36:14
  You have to be a Lewis Hamilton.
• 36:16
  Yes, yes.
• 36:17
  Yes.
• 36:18
  And even then, you know, like I'm talking about like there's a tree literally everywhere.
• 36:24
  It's not feasible.
• 36:25
  It's not going to work.
• 36:26
  It's not going to work.
• 36:27
  So what you could do instead is have a bulldozer come in or let's say a really high wind.
• 36:33
  And that's going to take down some of the trees.
• 36:36
  Yeah.
• 36:36
  Yeah.
• 36:37
  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.
• 36:48
  These are transition metal oxides that are metals bonded to oxygen.
• 36:53
  That's in the middle, sandwiched in between the two electrodes.
• 36:55
  Right?
• 36:57
  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?
• 37:04
  This is why oxygen is so reactive and why it rusts everything.
• 37:10
  The oxygen is going to start moving.
• 37:12
  When it starts moving, it's going to leave behind holes.
• 37:15
  Okay?
• 37:15
  So it's kind of like the trees are coming down and now I've got a pathway.
• 37:19
  Now my motorbike...
• 37:21
  Can move really fast.
• 37:22
  The motorbike in this analogy represents the electron.
• 37:26
  The electron is piggybacking on these vacancies.
• 37:29
  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.
• 37:42
  Yeah.
• 37:43
  So that would be a high resistance state.
• 37:45
  Right.
• 37:46
  That's one of the states that we were thinking a memorister should have.
• 37:49
  It's got a memory, right?
• 37:50
  There was no wind.
• 37:51
  And so I am at a high resistance.
• 37:53
  Now, all of a sudden, I applied an electric field.
• 37:55
  I got a low resistance.
• 37:56
  Can I go backwards?
• 37:57
  Yeah.
• 37:58
  Just apply the electric field in the other direction.
• 38:00
  The trees stand up.
• 38:02
  Low resistance.
• 38:03
  Got it.
• 38:04
  Got it.
• 38:04
  Right?
• 38:05
  Yes.
• 38:05
  This is the mechanism behind the switching behavior, the set and the reset that we think of in memoristers.
• 38:12
  That's the memorister jargon.
• 38:13
  It's like you set it to the low resistance state, so then there's like conduction happening.
• 38:17
  So that's your one, let's say.
• 38:19
  Yes.
• 38:19
  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.
• 38:27
  There's no more vacancies, and the electrons don't have a highway to go through.
• 38:30
  Yes.
• 38:31
  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.
• 38:39
  Exactly.
• 38:39
  So now you've got a cycling.
• 38:40
  Right.
• 38:41
  And that's something that you need for non-volatile memory.
• 38:44
  Right.
• 38:44
  Okay.
• 38:45
  That's effectively what this vacancy mechanism is.
• 38:50
  Got it?
• 38:50
  Now, a lot of times now we're going to try and use hafnium oxide.
• 38:54
  The advantage here is that hafnium oxide is something that silicon chip makers have already been using in transistor gate stacks.
• 39:03
  It's already replaced the silicon dioxide.
• 39:07
  transition metal oxide in the famous high K dielectric constant metal gate transition that happened in like 2007.
• 39:15
  I wasn't aware of this, obviously, because I was in school.
• 39:17
  But in 2007, like there was sort of a plateau on Moore's law because we couldn't get stuff smaller.
• 39:25
  And then and then somebody decided, hey, actually, we can use halfnium oxide that reduces power leakage.
• 39:30
  And it allows these transitions.
• 39:33
  It allows these transistors to shrink even further.
• 39:37
  without having a significant decrease in performance.
• 39:39
  The smaller you get, right, the harder it is to keep the same level of performance.
• 39:43
  Yeah, because you're trying to dissipate heat.
• 39:45
  The smaller you get now, the heat is starting to mess with the components themselves.
• 39:49
  And so this hafnium oxide, it's CMOS compatible.
• 39:52
  It scales to very small dimensions.
• 39:54
  It switches fast.
• 39:56
  And it shows really good endurance, right?
• 39:58
  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.
• 40:09
  And they still do.
• 40:10
  And just to make a quick clarification point.
• 40:12
  So chip makers were already using this for other chips prior to using it in Memristors.
• 40:21
  And so part of it was like they had the industrial scale working with this particular material.
• 40:28
  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.
• 40:35
  Exactly.
• 40:36
  Is that fair?
• 40:36
  That's exactly right.
• 40:37
  And so the original paper that was by Stanley Williams, that one used titanium dioxide as the transition metal oxide.
• 40:44
  Yeah.
• 40:45
  People have rapidly switched now to hafnium oxide as the sort of sandwich, the thing in between the sandwich.
• 40:51
  Okay.
• 40:53
  Now, we've got hafnium oxide.
• 40:55
  Okay.
• 40:55
  That's the meat in between the sandwich.
• 40:58
  What do we want to use for the electrodes?
• 41:00
  Usually we use tungsten and platinum.
• 41:02
  Why tungsten?
• 41:03
  Well, tungsten has a very high affinity for oxygen.
• 41:07
  So that's good.
• 41:08
  Just trying to get the oxygen.
• 41:11
  And it's also really good at withstanding high temperatures.
• 41:16
  That's why the Thomas Edison light bulbs use a tungsten filament.
• 41:20
  And here we've got a zoom in of one of those filaments.
• 41:23
  It's made out of tungsten because it's not going to melt.
• 41:26
  Okay?
• 41:27
  So that makes sense.
• 41:29
  The basic idea now to make this memorister is the following.
• 41:32
  We've got a top electrode.
• 41:34
  We've got a bottom electrode.
• 41:35
  Tungsten, half-nium oxide, something else.
• 41:39
  And we use the oxygen vacancies in the half-nium oxide to transport electrons and turn that highway on or off.
• 41:50
  Okay?
• 41:50
  That's effectively the idea.
• 41:52
  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.
• 42:03
  And so there's fertile ground to optimize for different use cases.
• 42:08
  That is exactly right.
• 42:09
  And let's stay here.
• 42:10
  Okay?
• 42:11
  So...
• 42:12
  You called it.
• 42:13
  The top is tungsten.
• 42:15
  The meat in the middle is half neem oxide.
• 42:17
  What should go on the bottom?
• 42:19
  Okay.
• 42:19
  For a long time, people have been talking about platinum.
• 42:24
  Usually you want some kind of inert metal.
• 42:26
  It's a metal that doesn't like react with stuff, right?
• 42:29
  Because obviously if it reacts, that's going to cause performance degradation.
• 42:36
  So platinum is a pretty good metal.
• 42:38
  This is where the valence charge mechanism takes over, right?
• 42:41
  The problem is when you pump up the temperature, at high temperature, the tungsten starts diffusing.
• 42:51
  Right?
• 42:51
  If you get into a high enough temperature, the tungsten now starts jiggling around, and the tungsten starts exploring the forest.
• 43:00
  And we don't want that.
• 43:00
  We don't want that.
• 43:01
  We just want the oxygen.
• 43:02
  We want only the oxygen to move around, create these vacancies, and then there's a highway.
• 43:07
  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.
• 43:13
  And if those two meet, you've got a shorted circuit.
• 43:16
  Yeah.
• 43:18
  Now, there's no way to, I mean, even if you cut down all the trees, you've just paved a highway.
• 43:25
  For that, yeah, yeah, yeah.
• 43:26
  And there's no, it can no longer cycle.
• 43:28
  Yeah, yeah.
• 43:29
  Why would anyone take the trees?
• 43:30
  You just go through the highway.
• 43:32
  Right.
• 43:32
  Right.
• 43:32
  You go through the tungsten and the platinum filaments.
• 43:35
  And that's been the big problem.
• 43:36
  Okay.
• 43:37
  Right.
• 43:37
  They mix in ways that destroy this precise nanoscale structure.
• 43:41
  Platinum and tungsten is not the way.
• 43:44
  Okay?
• 43:44
  You're going to effectively short-circuit at high temperatures.
• 43:47
  At low temperatures, this thing is working.
• 43:48
  Great.
• 43:49
  Right?
• 43:50
  But you start pumping this thing up to 500, 600, 700 degrees Celsius, all of a sudden, the metal atoms...
• 43:58
  in the two electrodes, they start getting agitated.
• 44:01
  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.
• 44:12
  And so there's different regimes depending on the temperature band you're working in.
• 44:16
  Yeah.
• 44:17
  And, yeah,
• 44:18
  I'm just making the connection between these two things.
• 44:21
  Different physics starts mattering, right?
• 44:23
  It starts mattering.
• 44:24
  It's like the titanium to titanium bond is now smaller than the thermal noise.
• 44:30
  And so the titanium is like, all right, I'm going to dip out and let's start exploring.
• 44:33
  I don't know.
• 44:33
  I'm going to go this way.
• 44:35
  And the platinum is like, hey, hey, hey.
• 44:37
  You know?
• 44:37
  Yeah.
• 44:38
  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.
• 44:46
  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.
• 44:56
  But then as we got to higher temperatures, physics gets different.
• 45:00
  And the current regime of tungsten and platinum is...
• 45:05
  Doesn't work.
• 45:05
  Doesn't work.
• 45:06
  Yes, exactly.
• 45:07
  And now we get to the paper.
• 45:09
  Okay, this was out in science.
• 45:12
  Shout out, AAAS.
• 45:16
  High temperature memristors enabled by interfacial engineering.
• 45:20
  Okay.
• 45:21
  One of the co-authors actually is Stanley Williams, who's the original Memristors guy.
• 45:26
  The nature for – oh my god.
• 45:29
  What a legend.
• 45:30
  I thought that was a nice throughput.
• 45:32
  What a legend.
• 45:34
  So Joshua Yang, who's at USC, he worked with Stanley at HP Labs.
• 45:38
  And then now he's a professor of electrical engineering at USC.
• 45:43
  So this is a collaboration between, as you said, USC, Air Force Research, and Kumamoto.
• 45:47
  Yeah.
• 45:48
  But it's like the same lab that created the Memristors is now just upping the game.
• 45:53
  We're the best.
• 45:54
  What do you want us to do?
• 45:55
  Which is kind of cool.
• 45:56
  Their big idea is to replace that bottom inert electrode...
• 46:01
  Instead of platinum, we're going to use graphene.
• 46:04
  Okay.
• 46:04
  Okay.
• 46:04
  So this is the one that we need to be more stable.
• 46:06
  Yeah.
• 46:06
  Doesn't want to do stuff.
• 46:08
  Yeah.
• 46:08
  I want it to not talk to the titanium.
• 46:10
  Or sorry.
• 46:11
  I want it to not talk to the titanium.
• 46:13
  I want them to not be friends.
• 46:15
  Yes.
• 46:15
  Platinum and titanium, they're friends at high temperature.
• 46:18
  Turns out graphene, even at high temperature, it's fine on its own.
• 46:22
  I'm good.
• 46:22
  Yeah, it's good.
• 46:23
  And for very good reasons.
• 46:25
  So graphene, it won the Nobel Prize in 2010.
• 46:28
  It was isolated in 2004 by Andre Gaim and Konstantin Noyoso-Levonov.
• 46:36
  Noyo Salove.
• 46:37
  Noyo Salove.
• 46:38
  Okay?
• 46:39
  They were in the UK.
• 46:42
  It's a very funny story where they actually, they took graphite, which is the stuff in pencil lead.
• 46:50
  They put it on pieces of paper.
• 46:51
  And then they got scotch tape.
• 46:54
  Put it on the piece of paper and removed it.
• 46:56
  And the...
• 46:57
  Physics between the scotch tape and the graphite that's on the piece of paper would create single layers of graphite, right?
• 47:05
  Single layers of carbon hexagonal shells.
• 47:08
  Graphite is a bunch of layers stacked together.
• 47:12
  Graphene is a single layer, okay?
• 47:16
  It's an incredible story because it was scotch tape and a pencil, and it won them the Nobel Prize.
• 47:21
  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.
• 47:30
  Yeah, and I mean, since then, we've done loads of research on how to make graphene.
• 47:34
  Obviously, like, if you're trying to do industrial scale, you're not going to be doing scotch tape and pencil, right?
• 47:40
  But you wouldn't have known.
• 47:41
  But you wouldn't have known, right?
• 47:42
  And now people do chemical vapor deposition.
• 47:44
  Yeah.
• 47:45
  to make graphene.
• 47:47
  Graphene has extraordinary properties, okay?
• 47:50
  It's the strongest material ever measured.
• 47:54
  It's an excellent electrical conductor.
• 47:57
  It's transparent.
• 47:59
  It's chemically inert.
• 48:01
  It doesn't really mess around with other things.
• 48:04
  And that chemical inertness is the key property here.
• 48:08
  Side note, graphene for the longest time
• 48:14
  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
• 48:26
  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.
• 48:41
  It didn't really find a lot of chemical, like a lot of industrial applications.
• 48:47
  So it's really nice to see that it's actually doing something.
• 48:50
  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.
• 48:55
  It's just I haven't heard it, right?
• 48:58
  I've heard of silica chips.
• 48:59
  I've heard of all these other things.
• 49:00
  Carbon nanofibers, like in F1, they use like carbon nanofibers.
• 49:04
  But that's not graphene.
• 49:06
  And carbon nanofibers never won a Nobel Prize.
• 49:09
  So I'm just saying.
• 49:10
  Bad PR.
• 49:11
  Right?
• 49:12
  This is a totally new purpose, though.
• 49:15
  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.
• 49:24
  Here, we're exploiting the chemical inertness of graphene.
• 49:28
  And it may be finally getting its moment in the light because the implications of this –
• 49:38
  could be just so enormous.
• 49:41
  Yeah, it really could.
• 49:42
  It really could.
• 49:43
  So let's now get into the paper.
• 49:45
  Let's talk about figure one, the graphene solution.
• 49:48
  Figure 1A on the top left, that is their Memrister.
• 49:53
  Okay?
• 49:53
  You've got a little...
• 49:56
  Tungsten electrode up top, you've got hafnium oxide in the middle, and you've got a graphene electrode on the bottom.
• 50:02
  Okay?
• 50:03
  And that's effectively what's happening.
• 50:05
  The tungsten doesn't want to mess with the graphene.
• 50:08
  We're going to get into some of the cool physics about that later.
• 50:11
  On the right-hand side is what I want to show.
• 50:12
  You see that curve, that hysteresis curve?
• 50:15
  That's characteristic of your membrister.
• 50:19
  So that is where they're showing the membrister.
• 50:22
  Mm-hmm.
• 50:24
  All of the colors are from zero degrees Celsius all the way down to 700 degrees Celsius.
• 50:30
  And notice the curve maintains its shape.
• 50:34
  Yes.
• 50:34
  All the way down to 700 degrees Celsius.
• 50:38
  This is crazy.
• 50:39
  Yeah, yeah, yeah.
• 50:40
  Right?
• 50:41
  Because we're talking about 2.5, 3x.
• 50:44
  Yeah.
• 50:44
  The current level that we can... Yeah.
• 50:46
  And if you...
• 50:47
  I mean, usually in temperature, we always, as physicists, like to talk in Kelvin, right?
• 50:50
  Right.
• 50:51
  Zero degrees Celsius is going to be like 270, let's say 300, 300 Kelvin.
• 50:56
  700 degrees Celsius is 1,000 Kelvin.
• 50:59
  So this is three times the thermal energy that this thing is able to withstand.
• 51:04
  Because Kelvin is really how you scale how much kinetic energy is in the system.
• 51:09
  At 1,000 Kelvin or at 900 Kelvin, let's say,
• 51:14
  your atoms are moving with three times the kinetic energy as the atoms at 300 Kelvin.
• 51:21
  Right.
• 51:21
  Okay?
• 51:22
  At room temperature.
• 51:23
  So this is crazy.
• 51:24
  Yeah.
• 51:24
  Okay?
• 51:25
  Yeah.
• 51:26
  Yeah.
• 51:26
  On the bottom left plot, you're seeing the time axis on the x-axis, right?
• 51:33
  Yep.
• 51:34
  That's 10 to the 4 seconds...
• 51:38
  You're maintaining those two states of high resistance, low resistance.
• 51:42
  So the cycling is maintaining.
• 51:44
  Yeah.
• 51:45
  That we talked about.
• 51:46
  The forest is going up and down.
• 51:47
  Yeah.
• 51:48
  And it's fine.
• 51:48
  And it's no problem.
• 51:49
  And it's not even coming together.
• 51:52
  It's just flat.
• 51:53
  Which is...
• 51:54
  Right, right, right, right.
• 51:54
  Like there's not a... Like there's not a...
• 51:56
  Yeah, I don't see anything.
• 51:57
  It's not approaching the collapse.
• 51:58
  Yeah.
• 51:59
  It's not approaching the short circuit.
• 52:01
  No, it's not.
• 52:02
  Yeah, short circuit would exactly mean that like they're both coming to the same thing.
• 52:05
  We're not seeing that.
• 52:06
  We're not seeing that at all.
• 52:07
  At 700 degrees Celsius.
• 52:08
  Which is odd.
• 52:08
  This is cycling at 700 degrees Celsius.
• 52:10
  So it's stable.
• 52:11
  Yes.
• 52:12
  And on the right-hand side, we've got endurance of cycles.
• 52:15
  The x-axis there on the bottom right, that's in log scale.
• 52:19
  Oh.
• 52:19
  Okay, that's not a linear scale.
• 52:21
  Each of those tick marks is 10 endurance cycles, then 100, then 1,000, then 10,000, 100,000.
• 52:27
  It goes all the way up to a billion cycles.
• 52:31
  Before we start to see it.
• 52:32
  Yeah, and even then, that could just be noise because there's enough jitter in the beginning.
• 52:36
  Yeah, yeah, yeah.
• 52:38
  It's not clearly.
• 52:40
  This is incredible.
• 52:41
  Right?
• 52:41
  Right.
• 52:43
  Because what we're sort of saying, there's two things here.
• 52:46
  One, stability at high temperature, which is C. And then D is stability over time, which goes back to the two things.
• 52:54
  And I'm kind of getting what you were saying before about why the fundamentals of the structure are.
• 53:01
  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
• 53:31
  And creating a paved highway.
• 53:33
  So the electrons will just move through that.
• 53:35
  Yes.
• 53:36
  Because that's the path of least resistance.
• 53:37
  Yeah, no one's going to take the vacancies, the oxygen vacancies.
• 53:41
  Why would you do that when there's a highway right next door?
• 53:45
  It's like the Cars movie, you know, when the interstate was formed.
• 53:48
  Nobody went through Route 66, and then your town got wrecked, right?
• 53:52
  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.
• 54:00
  The graphene's like, I'm good.
• 54:02
  I'm all carbon, right?
• 54:04
  And I'm good.
• 54:05
  It's like California's high-speed train.
• 54:09
  Yeah.
• 54:10
  86 miles of track, but you can't really use it.
• 54:12
  Yeah, you can't.
• 54:12
  Yeah, no, where is it going?
• 54:13
  Where is it going?
• 54:14
  Nowhere.
• 54:14
  Nowhere, right?
• 54:16
  And so that's exactly right, yeah.
• 54:18
  And so the tungsten is sort of just like trying.
• 54:22
  It's still diffusing around.
• 54:23
  The atoms are still diffusing because of the high temperature, but it can't bind to the graphene.
• 54:29
  And so the only way through is with the vacancy centers that are created through that –
• 54:34
  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
• 54:59
  It's programmatic control, for lack of a better term.
• 55:01
  Exactly, exactly.
• 55:02
  And Professor Joshua Wang, who is the principal investigator, he described this interface interaction between the tungsten and the graphene as oil and water.
• 55:11
  It's just like they're just not going to mix, right?
• 55:13
  Makes sense.
• 55:14
  This is so good.
• 55:16
  It's pretty cool, right?
• 55:17
  So now let's – to get into a paper in science –
• 55:21
  You can't just have a cartoon.
• 55:22
  Yeah.
• 55:23
  Right?
• 55:24
  That's for me.
• 55:26
  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.
• 55:36
  So let's go to the first analytical verification.
• 55:38
  This is high-resolution transmission electron microscopy.
• 55:42
  Electron microscopes, we've covered this a lot.
• 55:44
  Instead of light, we're using high-energy electrons to really get down to, you know—
• 55:51
  Tiny, tiny nanometer level resolution.
• 55:55
  On the top is a traditional memristor.
• 55:59
  Okay?
• 56:00
  You can see after, like, not a lot of time, the curves collapse.
• 56:05
  Yeah.
• 56:05
  Right?
• 56:06
  After just, like, 2,000 seconds, which is, like, less than an hour.
• 56:09
  Yeah.
• 56:10
  there's no memorister.
• 56:12
  It's the slope of a mammoth.
• 56:14
  Yeah, yeah, yeah, exactly.
• 56:15
  Straight down.
• 56:16
  Yeah, yeah, chair 36, I believe.
• 56:19
  Correct me in the comments.
• 56:20
  I never go on those chairs, so that'll be you, lunatics.
• 56:23
  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?
• 56:32
  That's the platinum and the tungsten meeting up.
• 56:35
  Mm-hmm.
• 56:36
  And so instead of sort of maintaining the same substrate texture or orientation, there's just this disturbance in the force.
• 56:44
  Exactly.
• 56:44
  And you can visually see it.
• 56:45
  Yeah, and you can see the hafnium oxide in the middle, right?
• 56:47
  The meat.
• 56:48
  But like the platinum is just like going through.
• 56:51
  Why would you go through the hafnium oxide, right?
• 56:53
  On the bottom is our new memristor with graphene.
• 56:58
  And there you can see even at high temperature, even after all of that cycling,
• 57:04
  The graphene layer is inert.
• 57:06
  The tungsten layer is inert.
• 57:07
  Nothing is happening.
• 57:08
  This is incredible.
• 57:09
  There's no chemistry happening.
• 57:10
  And it's also so verifiable.
• 57:13
  Yeah.
• 57:13
  You can literally look at it under the microscope.
• 57:16
  I can see.
• 57:16
  And you can see that it's happening.
• 57:18
  That stupid analogy is like imagine you have a BLT sandwich from a deli and they put the toothpick through it.
• 57:26
  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.
• 57:34
  Yeah.
• 57:34
  And it looks like a very nice, undisturbed.
• 57:38
  But the point here is the –
• 57:41
  The high-resolution imagery that is able to validate that this graphene-based memorister actually sustains at high temperature is undeniable.
• 57:50
  Yeah, yeah, yeah.
• 57:51
  It's like now you don't need a cartoon.
• 57:53
  This is the thing that the cartoon is inspired from.
• 57:55
  This is literally what it is, right.
• 57:55
  Right?
• 57:56
  Like you're just seeing it.
• 57:58
  I can see it.
• 57:59
  Here's another way to see it.
• 58:00
  You can use energy dispersive x-ray spectroscopy.
• 58:03
  The idea here is that, like, so during, in the previous photo, we had transmission electron microscopy.
• 58:10
  So you're, like, sending these high-energy electron beams, right, that strike the sample.
• 58:14
  Now, when those electrons strike the sample, it can transfer energy and violently kick electrons out.
• 58:21
  And when the electrons come back, they're going to release x-rays.
• 58:25
  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.
• 58:37
  Like in the transmission electron microscopy, people could be like, oh, there's some weird thing happening.
• 58:41
  How do you know that's tungsten?
• 58:43
  How do you know that's platinum?
• 58:45
  It's just a gray image.
• 58:47
  Well, now we can tell that it is, in fact, tungsten.
• 58:51
  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.
• 59:02
  There's no tungsten spike.
• 59:04
  The W, by the way, for those watching, W is the chemical symbol for tungsten.
• 59:10
  Because of Latin.
• 59:11
  And so it's interesting because I think part of this is like you're doing both of these things almost simultaneously.
• 59:17
  Yeah, yeah, yeah.
• 59:17
  Because the electron microscopy is triggering an event that you track with the x-ray spectroscopy.
• 59:24
  And so you're getting both the visual and the signature.
• 59:28
  Yeah, exactly.
• 59:29
  That allows you to identify the composition of what you're looking at and can see visually.
• 59:35
  Exactly.
• 59:36
  And so if you look in the platinum electrode, there is tungsten there.
• 59:40
  There's no tungsten, right?
• 59:42
  Undeniable.
• 59:43
  There were also some calculations that were done from first principles.
• 59:47
  Yes.
• 59:47
  So you've got electron densities with density functional theory.
• 59:51
  Density functional theory is something that we covered in our America 250 episode.
• 59:55
  Yes.
• 59:55
  It won the Nobel Prize.
• 59:56
  This is what I mean by density functional theory, okay?
• 59:59
  It's like you don't have to worry about every single electron.
• 1:00:02
  You can just be like the electron clouds are like a potential, right?
• 1:00:06
  What happens when a tungsten atom goes through?
• 1:00:10
  Does it bind or not?
• 1:00:12
  On the right-hand side is what platinum looks like.
• 1:00:15
  These giant sort of honeycomb structures where the d-orbital, which is the stuff that...
• 1:00:22
  These are these large d-orbitals of metals, right?
• 1:00:25
  They're extremely large.
• 1:00:27
  The tungsten is fine, kind of just...
• 1:00:30
  binding to one of these d-orbitals because they're so big, right?
• 1:00:34
  And there's a lot of vacant spots where the tungsten can go and bind.
• 1:00:38
  On the left-hand side is graphene.
• 1:00:41
  Pristine hexagonal structure.
• 1:00:44
  I think it's only just like sp2 binding.
• 1:00:48
  Or maybe sp3.
• 1:00:49
  I forget my chemistry.
• 1:00:50
  But in any case, the bond lengths in between the carbons are smaller.
• 1:00:56
  The graphene electron clouds are a lot smaller.
• 1:01:00
  So the tungsten has nowhere to go.
• 1:01:02
  It's incredible.
• 1:01:04
  You see the fat little pimple?
• 1:01:06
  Yes.
• 1:01:06
  That's the tungsten atom.
• 1:01:08
  And it's trying to find somewhere to bind.
• 1:01:10
  And it can't.
• 1:01:12
  It's like the grill on your microphone or a speaker cover.
• 1:01:15
  Yeah.
• 1:01:16
  Versus like you described, a bunch of donuts that got blended.
• 1:01:20
  Or like Hawaiian rolls that got baked together.
• 1:01:24
  Yeah, exactly.
• 1:01:24
  With no clear, actual, like, structural, you know, like a well-defined small structural positioning.
• 1:01:30
  Exactly.
• 1:01:31
  And it's part of your point that the...
• 1:01:34
  The smaller size of the electron cloud is a material aspect of why.
• 1:01:40
  Because the tungsten is just not small enough to... Yeah, yeah.
• 1:01:42
  The tungsten is not small enough.
• 1:01:43
  The smaller size also means that you need a higher energy to get in there.
• 1:01:47
  That makes sense.
• 1:01:47
  Right?
• 1:01:47
  Yeah, yeah.
• 1:01:49
  And all of the stuff is taken.
• 1:01:51
  Yes.
• 1:01:51
  Right?
• 1:01:52
  With graphene, because of that hexagonal structure, it's just carbon on carbon on carbon.
• 1:01:55
  There's no... All of the carbon atoms are happy.
• 1:01:57
  They don't want anything else.
• 1:01:58
  They don't want anything.
• 1:01:59
  Right?
• 1:01:59
  Yeah.
• 1:01:59
  Yeah.
• 1:02:00
  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
• 1:02:18
  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.
• 1:02:26
  And then there's like a valley on the other side where I'm bound to the platinum.
• 1:02:32
  The height of the hill is 0.3 electron volts.
• 1:02:35
  And then there's a nice valley where I can just like sit.
• 1:02:39
  On the right-hand side is the same diagram but for graphene.
• 1:02:43
  The height of the hill is now 1 to 2 electron volts.
• 1:02:47
  And there's no nice valley, right?
• 1:02:50
  Like if I'm up there and I'm jiggling around, I'll just come right back to my original state.
• 1:02:55
  So again, same idea.
• 1:02:57
  It's really hard to bind to graphene.
• 1:03:00
  It's really good.
• 1:03:00
  Okay.
• 1:03:02
  Now, one thing that I was thinking about was, there's kind of a paradox in my head, right?
• 1:03:06
  Like the high resistance state makes sense.
• 1:03:09
  High resistance meaning there's no oxygen vacancy highway.
• 1:03:17
  And so, like my electrons aren't going through, but everything is kind of inert, right?
• 1:03:23
  Now, let's say I have oxygen vacancy highway.
• 1:03:27
  Okay.
• 1:03:28
  in that low resistance state, what prevents the oxygen vacancy highway from moving around?
• 1:03:34
  Yeah.
• 1:03:35
  At high temperature.
• 1:03:36
  Yeah, yeah.
• 1:03:36
  Right?
• 1:03:36
  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.
• 1:03:44
  As we talked about earlier, they're not stationary.
• 1:03:46
  They're not stationary.
• 1:03:47
  They're not fixed or whatever.
• 1:03:47
  Yeah, yeah, yeah.
• 1:03:48
  So what's preventing those guys from moving around at high temperature when there's all this jiggle?
• 1:03:53
  It has to do with something called phase separation.
• 1:03:55
  Right.
• 1:03:56
  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.
• 1:04:09
  and then an oxygen-poor conducting phase.
• 1:04:11
  And it's that same argument about oil and water.
• 1:04:14
  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.
• 1:04:21
  And those guys don't want to mix together.
• 1:04:22
  They don't want to mix.
• 1:04:23
  Because they become like oil and water.
• 1:04:25
  Now, how can we tell in our new memorister?
• 1:04:29
  Well...
• 1:04:30
  They did electron energy loss spectroscopy, which is similar to the x-ray stuff that I was talking about.
• 1:04:35
  Like the electrons are now coming out, and then we measure the energy of the electrons.
• 1:04:41
  The main thing that I want you to see is in the low resistance state,
• 1:04:46
  which is on the bottom.
• 1:04:47
  So the color tells you how much oxygen vacancy there is.
• 1:04:51
  The darker it is, the more oxygen vacancy there is.
• 1:04:55
  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.
• 1:05:01
  The key thing that I want you to notice is on the bottom, we're plotting...
• 1:05:06
  the number of oxygen vacancies as a function of space.
• 1:05:09
  Like we're going, we're moving along in nanometer position.
• 1:05:13
  We're moving along our memristor and we're plotting how much oxygen vacancy there is.
• 1:05:18
  If you, if you expect a lot of mixing, you're,
• 1:05:21
  it should be at equilibrium.
• 1:05:23
  But if there's not a lot of mixing, there should be clumps.
• 1:05:26
  Yeah, right.
• 1:05:26
  Where there's a lot of oxygen vacancy and then not a lot of oxygen vacancy.
• 1:05:29
  Because of this oil and water you just talked about.
• 1:05:31
  And that's what we're seeing.
• 1:05:32
  We're seeing a very bumpy profile, right?
• 1:05:35
  There's certain spots where there's a lot of oxygen vacancy and then other spots that there aren't.
• 1:05:39
  If this stuff mixed, it would look like the middle plot in that high resistance state.
• 1:05:44
  Mm-hmm.
• 1:05:45
  Right?
• 1:05:46
  Yes.
• 1:05:47
  But because it's not mixing, you get this valley and mountain landscape.
• 1:05:52
  And so part of what we're saying is there's two dynamics here.
• 1:05:55
  There's both the dynamic between how the tungsten and the graphene relate.
• 1:06:01
  And then the halphide oxygen?
• 1:06:04
  Yeah, hafnium.
• 1:06:05
  Hafnium oxide.
• 1:06:06
  Excuse me, hafnium oxide in the middle also has dynamics that...
• 1:06:11
  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.
• 1:06:27
  And so both of these things combined are meaningful for the impact in terms of from an industrial or performance perspective.
• 1:06:36
  And it's both things, not just the ends of the sandwich.
• 1:06:40
  Exactly.
• 1:06:41
  Yeah.
• 1:06:41
  So both the high resistance state, which is the one, or sorry, no, the high resistance is the zero.
• 1:06:46
  Yeah.
• 1:06:46
  Because that's like the electrons aren't moving.
• 1:06:49
  And then the low resistance state where there is current, that's the one, both of those are now stable.
• 1:06:53
  Right.
• 1:06:54
  Right.
• 1:06:54
  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?
• 1:07:03
  Yeah.
• 1:07:04
  Yeah.
• 1:07:04
  The middle stuff is also oil and water.
• 1:07:06
  Right, right, right.
• 1:07:07
  It's kind of cool.
• 1:07:08
  Yeah.
• 1:07:09
  And they did a lot of benchmarking of like, okay, what is all the state of the art?
• 1:07:13
  And then where is this?
• 1:07:14
  And it's kind of crazy.
• 1:07:17
  Retention, which is how long it takes and the temperature, that's part A.
• 1:07:22
  You want it to be high temperature, high retention.
• 1:07:25
  So you've got a desired corner.
• 1:07:27
  That's the shaded.
• 1:07:28
  And there's a star where the current work is.
• 1:07:30
  And all of the other riffraff is not even close.
• 1:07:35
  You've got the writing endurance in B versus temperature.
• 1:07:38
  Again, the only star in the desired corner is this work.
• 1:07:43
  C, the on-off ratio versus the temperature.
• 1:07:45
  Desired corner is us.
• 1:07:47
  The spiderweb diagram.
• 1:07:50
  The yellow is the current work.
• 1:07:52
  And it's always...
• 1:07:55
  Larger in temperature, in retention, in endurance, in the device size, all sorts of stuff.
• 1:08:02
  Now, there is a little bit of Photoshoppery happening here, right?
• 1:08:06
  Because I could easily make the desired corner like elsewhere.
• 1:08:11
  But in any case, by every metric, it's the most in that corner.
• 1:08:17
  So whatever desired corner you made, it would always be the most out there, right?
• 1:08:21
  Right.
• 1:08:22
  It's the closest to the optimal, regardless of whether you made the desired outcome smaller or larger or whatever.
• 1:08:28
  Yeah, yeah, yeah.
• 1:08:29
  But a game-recognized game because, like, I've done that.
• 1:08:32
  I know what you're up to.
• 1:08:33
  Yeah, I'm going to make this square right here so that all the others are out.
• 1:08:38
  But, like, why do you just mix it in?
• 1:08:40
  Yeah, but in this case, I think it's, like, fairly obvious.
• 1:08:43
  And I don't think they're doing any shenanigans.
• 1:08:44
  I just thought that was kind of funny as a plot.
• 1:08:47
  And finally, like, because I think they knew that people would call BS.
• 1:08:52
  Right.
• 1:08:52
  So they took a video of their lab.
• 1:08:56
  They're like, proof.
• 1:08:57
  They're like, this is proof.
• 1:08:58
  And the thingy shows, oh, 700 degrees Celsius.
• 1:09:01
  And they've got, like, a computer with the two electrodes on the two...
• 1:09:09
  on the titanium and the graphene, and they're measuring the on off ratio and they're measuring the hysteresis curve.
• 1:09:18
  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.
• 1:09:26
  I'm not making this stuff up.
• 1:09:28
  Like, I think it's really cool that like this is part of the supplement on the science website.
• 1:09:33
  Because like whenever you publish, right, there's the main figures and the main text.
• 1:09:37
  And then you provide supplementary data and supplementary figures.
• 1:09:40
  A lot of the figures that we've seen here in today's episode are from the supplement.
• 1:09:45
  Mm-hmm.
• 1:09:46
  And this is a supplementary video that they show.
• 1:09:49
  Okay, guys, here's my hysteresis curve.
• 1:09:51
  It's definitely a memristor.
• 1:09:52
  This is happening at 700 degrees Celsius, and it's definitely working.
• 1:09:57
  Pixar didn't happen.
• 1:09:58
  Yeah.
• 1:10:00
  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
• 1:10:27
  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.
• 1:10:47
  Mm-hmm.
• 1:10:48
  And it was a very – it was maybe subtle for folks who are material scientists and who work in the space.
• 1:10:54
  But if you're not in this lane, it can seem like, oh, this is a subtle point that has such huge implications.
• 1:11:02
  And they've used graphene as the other part of this layer here.
• 1:11:09
  Are there other things that have different performance characteristics?
• 1:11:14
  Yeah, maybe we can now start replacing the tungsten.
• 1:11:19
  I don't know.
• 1:11:20
  Right.
• 1:11:20
  And then what does that do?
• 1:11:22
  Yeah, it opens up all sorts of possibilities.
• 1:11:24
  Right.
• 1:11:25
  So now let's talk about the future.
• 1:11:27
  Okay.
• 1:11:28
  What would this current research have to do with how we move forward in computation as a society?
• 1:11:34
  Right.
• 1:11:35
  We've talked about the von Neumann bottleneck for a long time.
• 1:11:38
  This is the tyranny of...
• 1:11:41
  the processor having to be separate from our memory.
• 1:11:45
  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.
• 1:11:59
  Von Neumann came up with this, and it's the bedrock of almost all of the computation that we have in our society today.
• 1:12:06
  Now, with neural networks, this is especially bad because you've got to load the weights
• 1:12:11
  And then you've got to do the computation.
• 1:12:13
  Then you've got to load a new set of weights.
• 1:12:16
  So you've got to dump the old weights, right?
• 1:12:18
  And all of this takes a lot of time.
• 1:12:20
  It takes a lot of energy, dissipates heat, and...
• 1:12:25
  What exactly is the computation that happens in these AI systems?
• 1:12:29
  When you've got a neural network, what we're really doing fundamentally is we're relying on matrix times vector multiplication.
• 1:12:37
  The vector is your state and the matrix is the weights of your network.
• 1:12:41
  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.
• 1:12:48
  At the end of the whole thing, you get a giant vector.
• 1:12:51
  In the example of LLMs, you have a giant vector with all of the next possible tokens.
• 1:12:58
  And you pick usually the one with the highest probability.
• 1:13:02
  Okay.
• 1:13:03
  How does matrix multiplication work?
• 1:13:05
  How do you multiply a matrix by a vector?
• 1:13:08
  For those from linear algebra, you already know this.
• 1:13:10
  But let's just do a little bit of a review.
• 1:13:12
  Okay.
• 1:13:13
  Let's say I've got a 3x3 matrix.
• 1:13:14
  Okay.
• 1:13:15
  On the left there, I've got 1, 1, 2, 2, 1, 3, 1, 4, 2.
• 1:13:19
  And I want to multiply it by a vector 3, 1, 2.
• 1:13:21
  3, 1, 2 is the vector that's my state of like next tokens, let's say.
• 1:13:26
  And the matrix tells us how to change that vector to create a new probability vector with the new tokens.
• 1:13:34
  What you do is you go row by row and you take each row, you multiply it by the vector.
• 1:13:39
  So the 1 gets multiplied by the 3, the 1 gets multiplied by the 1.
• 1:13:42
  2 gets multiplied by the 2 and you add it all up.
• 1:13:45
  That's your first row.
• 1:13:46
  Then you go to the second row, add it all up.
• 1:13:48
  Third row, multiply, add it all up and you get your new vector.
• 1:13:52
  Okay.
• 1:13:52
  That's fundamentally what most of the AI computation that we all know and love, I guess.
• 1:14:02
  That's what that is.
• 1:14:03
  Okay.
• 1:14:04
  That's what's happening.
• 1:14:05
  Keyword, I guess.
• 1:14:07
  Yeah.
• 1:14:07
  That's what's happening.
• 1:14:09
  Now.
• 1:14:10
  We'd like to do better.
• 1:14:11
  We'd like to not load the weights every single time.
• 1:14:17
  in order to do this computation.
• 1:14:18
  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.
• 1:14:28
  Yeah, it has to store all of those matrices and it's got to do all of that computation.
• 1:14:33
  That's why GPUs are so important because GPUs are just insanely good at matrix multiplication.
• 1:14:39
  At that specific task.
• 1:14:40
  At that very specific task, okay?
• 1:14:42
  They're not good at doing normal emails,
• 1:14:47
  It turns out NVIDIA made the big gamble way back in the day that actually graphics, which is matrix multiplication...
• 1:14:57
  is going to have a use for later.
• 1:14:59
  And it turns out, yeah.
• 1:15:00
  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.
• 1:15:09
  Yeah, yeah, exactly.
• 1:15:10
  So instead of doing all this von Neumann nonsense, instead we can do something called in-memory computing.
• 1:15:17
  In-memory computing meaning I don't have to move anything, okay?
• 1:15:22
  We can use something like a resistor
• 1:15:28
  do the computation without changing anything.
• 1:15:32
  So again, let's talk about how that would work.
• 1:15:35
  I told you about how matrix multiplication works, right?
• 1:15:38
  Like matrices, you multiply it by a vector, and then you get a new vector.
• 1:15:41
  Well, now, what if we had the old vector come in like wires...
• 1:15:46
  Okay?
• 1:15:47
  And this is what we see on the right.
• 1:15:49
  And let me see if I can say this correctly.
• 1:15:51
  Okay?
• 1:15:52
  So I've got a matrix, and I've got an old vector, and I want to make a new vector.
• 1:15:56
  And the way that I do that is the rows get multiplied to each of them, and then I get added up, right?
• 1:16:01
  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.
• 1:16:12
  Here's what I do.
• 1:16:13
  My old vector is represented by the old... By currents that are going at the very top.
• 1:16:21
  The old voltages, I should say.
• 1:16:24
  Each of the wires at the very top is held at a certain voltage.
• 1:16:28
  I've got resistors that are attached to these voltages.
• 1:16:33
  And then I've got a new set of wires at the bottom.
• 1:16:37
  Because of electronics...
• 1:16:39
  The new current, it's a new set of currents at the bottom.
• 1:16:44
  Because of electronics, right?
• 1:16:46
  The new currents...
• 1:16:49
  are going to be related to the voltage at the top and the resistance in between.
• 1:16:55
  And all I have to do is add up the currents.
• 1:16:56
  I'm so mad.
• 1:16:57
  Right?
• 1:16:58
  In this case, it's not really resistance.
• 1:17:00
  It's one over resistance.
• 1:17:01
  It's the conductance.
• 1:17:03
  And then the current is just voltage multiplied by conductance.
• 1:17:06
  But it's the same thing.
• 1:17:07
  It's just V equals IR the other way around.
• 1:17:10
  And all we're doing is using the physics in place...
• 1:17:14
  We don't have to move weights around.
• 1:17:16
  As long as the weights are the same, I can use the physics to be like voltage multiplied by conductance gives me a current.
• 1:17:24
  Voltage multiplied by another conductance gives me another current.
• 1:17:28
  And so my new vector is going to be the sum of all of those currents.
• 1:17:32
  And there you go.
• 1:17:33
  Matrix multiplication using physics.
• 1:17:35
  I'm so...
• 1:17:36
  I'm so mad.
• 1:17:37
  It's so good.
• 1:17:38
  That's so good.
• 1:17:39
  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.
• 1:17:48
  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.
• 1:17:56
  I'm just using Ohm's law and Kirchhoff's junction rules, the stuff that you learn in intro physics, right?
• 1:18:03
  Now, are we capable of doing something like this?
• 1:18:05
  Well, this particular paper itself made a 32 by 32 analog resistance state
• 1:18:13
  And it works at 1300 degrees Fahrenheit.
• 1:18:16
  It works at low voltage.
• 1:18:17
  0.5 volts is all you need.
• 1:18:21
  30 nanoseconds switching in the speed.
• 1:18:25
  And it definitively proves that you can be fully capable of storing complex, multi-level
• 1:18:33
  analog neural weights for whatever AI interface that you're working with.
• 1:18:37
  It can do it in extreme environments, 1300 degrees Fahrenheit.
• 1:18:41
  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.
• 1:18:48
  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.
• 1:18:58
  God.
• 1:19:00
  It's so, it's so good.
• 1:19:02
  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.
• 1:19:22
  Yeah.
• 1:19:23
  Um,
• 1:19:24
  Because it's not just one or the other.
• 1:19:27
  Both things give you now the single chip –
• 1:19:31
  that can work in high temperature, that can run comparably for a long time.
• 1:19:38
  Faster.
• 1:19:39
  Faster.
• 1:19:41
  And can do both your compute and memory at the same time.
• 1:19:44
  Exactly.
• 1:19:45
  And of course, the authors have made a startup.
• 1:19:49
  Of course they did.
• 1:19:50
  There's a startup called TetraMem.
• 1:19:51
  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.
• 1:19:59
  Yeah.
• 1:20:00
  And it's already working to commercialize this crossbar Memristor device to replace power-hungry GPUs.
• 1:20:07
  Right?
• 1:20:08
  They're getting acquired immediately.
• 1:20:09
  Yeah.
• 1:20:10
  No, especially with this.
• 1:20:11
  Immediately.
• 1:20:12
  Now, let's talk about what this has to do with data centers in space.
• 1:20:16
  Okay.
• 1:20:16
  Okay?
• 1:20:17
  Elon wants to put data centers in space.
• 1:20:19
  Why?
• 1:20:20
  Because it's probably possible.
• 1:20:21
  Yeah.
• 1:20:22
  And, I mean, it helps SpaceX, obviously.
• 1:20:25
  And it's not just SpaceX that's trying to do this, right?
• 1:20:28
  Google is trying to do this.
• 1:20:29
  There's a bunch of startups that are trying to do this.
• 1:20:32
  And it's looked impossible for a very, very long time because of a sort of fundamental physics quandary.
• 1:20:40
  The reason why is because these GPUs are...
• 1:20:44
  Highly reliant on cooling mechanisms because the GPUs can't withstand a lot of heat.
• 1:20:50
  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.
• 1:21:00
  At the end of the day, that's most of the water being used.
• 1:21:02
  is just to extract heat from these GPUs and put it out into the environment.
• 1:21:08
  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.
• 1:21:15
  And then all the environmental people are like, okay, what's happening in the flora and fauna, which is completely reasonable.
• 1:21:21
  Right.
• 1:21:22
  I don't want to hurt like the planet and especially the planet that's right around those data centers.
• 1:21:27
  So yeah,
• 1:21:29
  Let's talk about how heat transfer happens.
• 1:21:31
  There's three types of heat transfer, okay?
• 1:21:33
  There's convection, conduction, and radiation.
• 1:21:36
  Convection is hot stuff moves around to the cold part, and then the cold part goes down to where the hot stuff is.
• 1:21:43
  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.
• 1:21:52
  So that's convection.
• 1:21:53
  Conduction is...
• 1:21:56
  The atoms themselves jiggling around to conduct heat from one spot to the other.
• 1:22:02
  That's why the handle gets hot.
• 1:22:04
  And then finally, the most inefficient part is radiation, which is literally photons coming out.
• 1:22:11
  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.
• 1:22:22
  And that's why a shade is so important because now the photons, the infrared photons from the sun are not hitting you.
• 1:22:29
  And then all of a sudden you feel like 10 or 20 degrees cooler.
• 1:22:33
  You're right next to the sun.
• 1:22:35
  The shade is only shading your part.
• 1:22:37
  But the reason why it's so effective is because the photons aren't getting to you.
• 1:22:40
  The convection and the conduction still is.
• 1:22:43
  The hot pockets of air are still moving towards you, but the photons are not.
• 1:22:47
  Okay, so those are the three types of heat transfer.
• 1:22:49
  Okay.
• 1:22:51
  Now, terrestrial data centers, they mostly use the first two.
• 1:22:56
  Mostly it's convection.
• 1:22:58
  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.
• 1:23:04
  It dumps the heat, and you cycle it back, and so on and so forth, right?
• 1:23:10
  Terrestrial data centers have issues, right?
• 1:23:13
  Where is the power going to come from?
• 1:23:14
  Where is the water going to come from?
• 1:23:16
  Who's going to build it?
• 1:23:17
  Yeah, who's going to pay for it?
• 1:23:19
  Yeah, there's a lot of nimbiness going on.
• 1:23:20
  And in this case, the nimbiness is a good thing, right?
• 1:23:23
  We don't want, I don't want data centers in my neighborhood.
• 1:23:26
  Yeah.
• 1:23:26
  Okay.
• 1:23:27
  I might want the AI, but I really don't want the data centers.
• 1:23:31
  Right.
• 1:23:31
  And so that's why people are looking into space.
• 1:23:33
  Now, people have already started looking.
• 1:23:35
  This is StarCloud.
• 1:23:36
  It's a startup that's backed by NVIDIA.
• 1:23:39
  Google is doing its own thing.
• 1:23:41
  And there's a lot of smart people betting on this thing.
• 1:23:43
  Here it shows a 5 gigawatt data center.
• 1:23:46
  That's crazy.
• 1:23:47
  But the radiator that you would need to dissipate all that heat with radiation is something like a 4 kilometer square.
• 1:23:55
  And you have real, you know, getting any payload into space is cost prohibitive.
• 1:24:02
  Yeah.
• 1:24:03
  And so getting a payload that's four kilometers by four kilometers wide will take a modular system, multiple launches.
• 1:24:09
  Yes, the cost of launches is going down.
• 1:24:11
  Yeah.
• 1:24:12
  Still, it's four kilometers worth of stuff.
• 1:24:15
  Yeah.
• 1:24:15
  worth of stuff, right?
• 1:24:16
  It's not a joke.
• 1:24:17
  It's not trivial, right?
• 1:24:18
  And there's several important papers that are coming out trying to plan for this kind of thing.
• 1:24:22
  Like, how do we develop carbon-neutral data centers in space?
• 1:24:26
  I want to highlight this because this is in Nature Electronics.
• 1:24:28
  This is a top journal.
• 1:24:31
  It's not impossible.
• 1:24:32
  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.
• 1:24:41
  It's not true, okay?
• 1:24:42
  It's hard, right?
• 1:24:45
  But that's a key distinction from it's impossible.
• 1:24:48
  Okay?
• 1:24:49
  I mean, we obviously do it, right?
• 1:24:51
  The International Space Station is already doing this.
• 1:24:54
  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.
• 1:25:02
  The flat ones, those are radiators.
• 1:25:05
  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.
• 1:25:14
  There are several computers up there.
• 1:25:16
  Those don't get cooked alive.
• 1:25:18
  The way that works is using both convection
• 1:25:21
  And radiation.
• 1:25:22
  So you've got convection in that they use ammonia instead of water because ammonia has a low freezing point.
• 1:25:27
  If you use water, then the water is just going to freeze when it gets to the radiator.
• 1:25:31
  And then like what?
• 1:25:33
  So that's why you use ammonia.
• 1:25:35
  The ammonia uses convection to take the heat to the radiator.
• 1:25:39
  And then the radiator takes that heat and radiates it out into space.
• 1:25:43
  Okay.
• 1:25:44
  Now, what we want to do is make this thing cheaper.
• 1:25:46
  Okay.
• 1:25:47
  And that is perhaps an engineering problem, especially in the context of the paper that we've just covered.
• 1:25:55
  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.
• 1:26:03
  The technology poses three advantages.
• 1:26:07
  One, let's talk about the physics of thermal radiation.
• 1:26:10
  The reason why everyone's saying that it's going to be impossible.
• 1:26:13
  Right?
• 1:26:14
  Right.
• 1:26:16
  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.
• 1:26:28
  This is why it's hard.
• 1:26:29
  Okay?
• 1:26:30
  It's not T squared.
• 1:26:31
  It's T to the fourth.
• 1:26:32
  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.
• 1:26:43
  Okay?
• 1:26:43
  Now, this is actually a good thing.
• 1:26:46
  Okay.
• 1:26:46
  In the context of this current paper.
• 1:26:49
  Now, standard silicon processors, they've got to be relatively cool to prevent failure, right?
• 1:26:53
  But you can only allow that chip with the standard silicon to run at like tops 100 degrees Celsius.
• 1:27:01
  At 100 degrees Celsius, right?
• 1:27:04
  Let's say 400 Kelvin is the tops that you can do.
• 1:27:09
  At 400 Kelvin...
• 1:27:12
  The amount of radiation that I'm letting off is going to be some amount.
• 1:27:15
  Some amount of photons are going off.
• 1:27:17
  Now, what if I could make that go up to 700 degrees Celsius?
• 1:27:24
  1,000 Kelvin.
• 1:27:26
  So I've effectively doubled the temperature that I'm working with.
• 1:27:29
  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.
• 1:27:35
  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.
• 1:27:48
  Yeah.
• 1:27:48
  Yeah.
• 1:27:49
  The size of the radiator goes down, the higher the temperature that you can operate.
• 1:27:54
  And this is crucial for putting the stuff into space.
• 1:27:59
  Because of what you said, it's hard to put stuff in space.
• 1:28:03
  It's hard.
• 1:28:04
  It's expensive.
• 1:28:05
  I like to put less stuff in space.
• 1:28:09
  This could be the key, right?
• 1:28:10
  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.
• 1:28:22
  This is so interesting.
• 1:28:23
  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...
• 1:28:37
  When you think about the concept of a data center in space.
• 1:28:40
  Yeah, exactly.
• 1:28:41
  And I did some back of the envelope calculations, right?
• 1:28:43
  So let's say I need a radiator for a one megawatt space server.
• 1:28:50
  Okay.
• 1:28:51
  That means one megawatt of electricity is coming in from my solar panels.
• 1:28:55
  Okay.
• 1:28:56
  Now at standard 30 degrees Celsius, so that's 300 Kelvin.
• 1:28:59
  If I want to dump one megawatt of heat, I would need 2,000 meters squared.
• 1:29:05
  That's eight tennis courts.
• 1:29:06
  Okay.
• 1:29:08
  And I want to dump just as much energy that's coming in, right?
• 1:29:11
  Because if I don't dump the energy that's coming in, then my temperature increases and then the electronics are not stable.
• 1:29:18
  My satellite keeps increasing in temperature.
• 1:29:20
  So whatever energy is coming in, I'd like to dump out.
• 1:29:22
  So one megawatt's in, one megawatt's out.
• 1:29:26
  What matters is the temperature that I'm operating at, okay?
• 1:29:29
  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.
• 1:29:37
  My area has gone down to 150 meters squared.
• 1:29:41
  I've gone from 2,000 meters squared to 150 meters squared.
• 1:29:44
  That's a reduction of 90%.
• 1:29:47
  Right.
• 1:29:47
  If I go even further, because this thing was at 700 degrees Celsius, right?
• 1:29:51
  If I go even further to 600 degrees Celsius now, meaning 900 Kelvin, now my radiators only need to be 30 meters squared.
• 1:30:01
  And that's totally doable.
• 1:30:03
  Yeah, that's very reasonable.
• 1:30:04
  Right.
• 1:30:05
  That's already in the ballpark of stuff that we've already put out there.
• 1:30:08
  Right, right, right.
• 1:30:10
  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.
• 1:30:20
  The heat, it doesn't feel – Yeah, as long as the stuff is not cooking.
• 1:30:25
  Right.
• 1:30:25
  Right?
• 1:30:26
  But it kind of makes sense because if you think about – you know those electric stoves with the coils?
• 1:30:32
  Yeah.
• 1:30:33
  Right?
• 1:30:34
  The higher the temperature, the more heat is coming out.
• 1:30:37
  That's the whole point of how it cooks stuff.
• 1:30:39
  Yeah.
• 1:30:40
  No, I guess that's not true because there's like contact.
• 1:30:42
  But you get what I'm saying, right?
• 1:30:43
  Like if you're around that thing, it's like there's a lot more photons that are coming out.
• 1:30:49
  So you're dumping a lot more energy.
• 1:30:50
  And as long as you can live –
• 1:30:53
  At 700 degrees Celsius, you're good to go.
• 1:30:55
  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.
• 1:31:00
  So it never was even a possibility.
• 1:31:02
  There were no chip components that could withstand that temperature.
• 1:31:04
  So everyone was doing these calculations at the lower temperature being like, oh, I need tennis court size, four kilometers squared size.
• 1:31:12
  But now I've reduced stuff down to like 95, 99 percent the area just because my components can withstand the higher temperature.
• 1:31:22
  My components, my computer can be hotter.
• 1:31:24
  And so the amount of radiator I need is a lot, lot lower by T to the fourth.
• 1:31:29
  This is so, I mean, you know, again, there is the industrial scale question and all of these other, you know, engineering.
• 1:31:37
  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.
• 1:31:51
  what we're doing at the frontier.
• 1:31:52
  Yeah.
• 1:31:53
  I mean, this paper came out in March.
• 1:31:54
  Yeah.
• 1:31:55
  By the way.
• 1:31:56
  And this discussion about data centers in space has really heated up as we've gotten into the summer.
• 1:32:00
  Yes.
• 1:32:00
  Which was after this paper had come out.
• 1:32:03
  And most of the folks who have been like, it's impossible, you know, certainly one from a fundamental physics perspective.
• 1:32:10
  It's not impossible.
• 1:32:11
  It's not impossible.
• 1:32:12
  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...
• 1:32:19
  You know, just talking about putting the same thing we have in a data center on the ground.
• 1:32:23
  Yeah.
• 1:32:23
  And just putting it in space.
• 1:32:24
  Exactly.
• 1:32:25
  And if you combine that with the fact that payloads to space are going to get cheaper, as you said...
• 1:32:30
  the amount of payload that I need is going to go down.
• 1:32:34
  This is totally a doable thing, right?
• 1:32:36
  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.
• 1:32:48
  Yeah, because you can have a lot more smaller, you can have much smaller space servers with high...
• 1:32:57
  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
• 1:33:23
  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
• 1:33:38
  Exactly.
• 1:33:39
  That exists that negates the need to have terrestrial observatories.
• 1:33:43
  Again, there's some caveats there.
• 1:33:45
  Yeah, yeah.
• 1:33:46
  There's a lot of caveats.
• 1:33:47
  A lot of caveats, right.
• 1:33:48
  But, you know, they're trade-offs and we'll have to continue to figure out.
• 1:33:50
  Yeah, maybe we can make the orbits, like, avoid Chile, which is where, like, half of everything is.
• 1:33:56
  Right, right.
• 1:33:57
  If you just avoid, like, Hawaii and Chile, like… No fly zones.
• 1:34:00
  Like, half of the astronomers are going to be happy.
• 1:34:02
  Yeah, right, right.
• 1:34:03
  You know, so I mean, that's going to be hard given that it's like high Earth orbit and whatever.
• 1:34:07
  But in any case, the other thing that I want to bring up is this is not just an advantage in terms of cooling.
• 1:34:14
  OK, there's two other things that this new Memristor technology would give us for data centers in space.
• 1:34:22
  One is the extreme power efficiency.
• 1:34:24
  I was talking about how it only takes like very little amount of voltage in order to do that computation with a Memristor.
• 1:34:31
  Now, satellites operate on extremely strict power budgets, you can imagine, right?
• 1:34:36
  Often it's under 20 watts.
• 1:34:38
  And in standard AI chips, nearly every single watt of electricity is consumed directly into heat.
• 1:34:45
  But Memristors use this in-memory analog computing.
• 1:34:48
  They don't have to do this von Neumann bottleneck, right?
• 1:34:51
  So if we're applying Memristors here, then that complex mathematics instantaneously happens without shuttling the data back and forth...
• 1:34:59
  Now you're consuming a fraction of the power of a traditional GPU.
• 1:35:03
  And that's going to bring down the amount of power that you need to dissipate again.
• 1:35:08
  So this is both you can have the radiators be smaller and you can have the solar panels be smaller.
• 1:35:12
  Exactly.
• 1:35:13
  Because you don't need as much power intake.
• 1:35:15
  Exactly.
• 1:35:15
  Or with the same solar panels, you can do a lot more.
• 1:35:18
  Right.
• 1:35:18
  You know, either way.
• 1:35:19
  Right.
• 1:35:20
  Efficiency is always great.
• 1:35:22
  Both in power dissipation and in terms of computing.
• 1:35:25
  And finally, the other big thing that people talk about with data centers in space is like normal GPUs, a cosmic ray comes in,
• 1:35:33
  I don't know if that bit was a zero or a one.
• 1:35:37
  Like, what?
• 1:35:39
  So a galaxy decided to fart, and then now my AI query is giving me nonsense.
• 1:35:46
  That's not great.
• 1:35:47
  You need electronics that is able to sustain performance in a high-radiation environment.
• 1:35:54
  Memristors are great for this because memristors store memory by physically moving atoms and creating structural bridges, right?
• 1:36:02
  This like valence charge mechanism rather than relying on these easily disrupted pockets of electrical charge.
• 1:36:09
  So now you can imagine a cosmic ray comes in.
• 1:36:12
  It's not really going to upset the valence charge mechanism all that much compared to if there was a single point of failure.
• 1:36:20
  Right.
• 1:36:22
  And so we have a thermal radiation benefit, a power efficiency benefit, and a resilience benefit.
• 1:36:32
  And it's hard for me to imagine that.
• 1:36:34
  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.
• 1:36:44
  We already can see now in the case of data centers in space why if you can figure out the industrial scaling of this –
• 1:36:55
  It makes the execution of it in three of the most important categories that you care about as an operator.
• 1:37:06
  significantly better in an order or orders of magnitude from what you would otherwise be doing with the current conventional options.
• 1:37:16
  Exactly, yeah.
• 1:37:17
  I mean, I think it's an incredibly exciting time for electronics at the cutting edge, right?
• 1:37:24
  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...
• 1:37:33
  increasingly important, perhaps even more important than traditional von Neumann architectures.
• 1:37:39
  We've still got a long way to go.
• 1:37:40
  So I do want to temper the brakes here and not dream too much because memory alone does not make a complete computer.
• 1:37:47
  You need high temperature logic circuits that also need to be developed and integrated alongside whatever chip that you're trying to put in.
• 1:37:54
  And current devices are built by hand, submicron scale in the lab.
• 1:38:00
  This is from their paper in the supplement.
• 1:38:04
  That length bar there says 5 micrometers.
• 1:38:08
  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.
• 1:38:19
  So we've got to make this thing smaller while still retaining the memorister characteristics, the graphene.
• 1:38:25
  That's going to be a problem,
• 1:38:27
  Right.
• 1:38:28
  Scale up is going to take time on the manufacturing side.
• 1:38:31
  Two of the three devices, two of the three materials in this device, namely the tungsten and the hafnium oxide.
• 1:38:37
  Those are already standard in semiconductor manufacturing.
• 1:38:41
  Right.
• 1:38:41
  Making the graphene smaller.
• 1:38:44
  That's going to be kind of tough, right?
• 1:38:45
  It's newer to the industry.
• 1:38:46
  But actually TSMC and Samsung both have that in their development roadmaps.
• 1:38:51
  If you go look, they have the shareholder pitch decks.
• 1:38:56
  They're like, we're still innovating, guys.
• 1:39:00
  So they have graphene.
• 1:39:03
  in their roadmap and it's already grown at wafer scale in research settings.
• 1:39:07
  We need to make it industrial if we want to have this thing grow.
• 1:39:11
  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.
• 1:39:20
  So we're, we're, we're heading for a new era in computation in a,
• 1:39:25
  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
• 1:39:44
  Large-scale computing is now taking precedence over personal small-scale computing.
• 1:39:49
  So we're entering into a new era of computation.
• 1:39:54
  We've lived through one in the 90s, and now this is a new one.
• 1:39:58
  As it relates to TSMC, it's interesting that both TSMC and Samsung are already working on integrating graphene into their industrial process.
• 1:40:08
  As if we needed another reason for Taiwan to be a geopolitical...
• 1:40:11
  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.
• 1:40:26
  Yeah.
• 1:40:26
  Um, yeah,
• 1:40:28
  This is going to be – and again, we talk about breaking science research papers on this story.
• 1:40:33
  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.
• 1:40:41
  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.
• 1:40:55
  And again, we talked about this in the context of data centers in space.
• 1:41:00
  But there are other things that you've mentioned, you know, that would be used.
• 1:41:07
  This will be useful for an extreme computing.
• 1:41:09
  That's right.
• 1:41:10
  Exploration and other things.
• 1:41:11
  So space exploration.
• 1:41:13
  I'd like to get better photos of Venus.
• 1:41:15
  Yeah.
• 1:41:15
  Yeah.
• 1:41:15
  That'd be really nice.
• 1:41:17
  That'd be great.
• 1:41:18
  I'd like to have a rover on Venus that's like running around and stuff.
• 1:41:21
  That would be cool.
• 1:41:23
  Heavy energy, like geothermal, nuclear monitoring.
• 1:41:26
  I'd like to have safer nuclear power.
• 1:41:31
  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.
• 1:41:43
  Yeah.
• 1:41:44
  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
• 1:42:10
  Who needs a room temperature semiconductor?
• 1:42:12
  Yeah.
• 1:42:14
  Well, we still want that.
• 1:42:14
  We still want that.
• 1:42:15
  We still want that.
• 1:42:17
  This was a really great electronics that can survive lava.
• 1:42:21
  Again, this was a recent science paper, March 2026, a collaboration led by researchers and
• 1:42:30
  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.
• 1:42:50
  Can't wait to see high-temperature immersers in a hypersonic vehicle near you.
• 1:42:58
  Because certainly they're going to be using it for command and control systems integrated with, you know, next generation military stuff.
• 1:43:07
  So before they do that, we would like to have it be used for the good of mankind.
• 1:43:13
  Do you have any last notes before we wrap up today?
• 1:43:15
  No, if you want to leave a comment, think of a really funny application for high temperature computing.
• 1:43:22
  Where would you want to put a chip?
• 1:43:25
  Run wild.
• 1:43:26
  Keep it PG, everybody.
• 1:43:27
  This is a family-friendly show.
• 1:43:30
  As always, my name is Lester Nari, joined by my co-host, our resident data centers in space, our possible PhD, Krishna Chowdhury.
• 1:43:43
  We are approaching our world.
• 1:43:46
  one year anniversary.
• 1:43:48
  So be sure to be tuned in for our anniversary episode.
• 1:43:52
  For our longtime listeners, we are extremely grateful.
• 1:43:56
  If you just happen to be listening to us for the first time, Catalog is a bunch of episodes that are just like this.
• 1:44:04
  50 of some of the best science entertainment on the planet.
• 1:44:09
  We will see you all next week.