Latest / The Quark Side - Quantum Physics Podcast / Quantum Computers Have a Hidden Flaw — Scientists Just Found It

Transcript

Quantum Computers Have a Hidden Flaw — Scientists Just Found It · The Quark Side - Quantum Physics Podcast

0:01
Speaker 1: Welcome to the qure Side Quantum Physics Podcast, an exploration
0:05
Speaker 1: of the fundamental structure of reality. We're quantum laws govern matter, energy,
0:11
Speaker 1: and information. Here, uncertainty is a feature, not a flaw,
0:16
Speaker 1: and understanding begins at the smallest scales.
0:27
Speaker 2: Welcome back today. We are going to try to wrap
0:29
Speaker 2: our heads around something that is simultaneously incredibly small and
0:34
Speaker 2: absolutely massive in terms of its implications.
0:37
Speaker 3: That's a good way to put it.
0:38
Speaker 2: We're talking about the future of computing, but you know,
0:40
Speaker 2: not the shiny marketing brochure version of it.
0:43
Speaker 3: No, we are getting into the messy reality, the plumbing
0:47
Speaker 3: behind the palace, so to speak.
0:49
Speaker 2: We always hear the grand promise, right We hear that
0:52
Speaker 2: quantum computers are going to solve these impossible problems revolutionized medicine,
0:56
Speaker 2: crack encryption codes that would take a normal computer a
0:59
Speaker 2: billion years to solve. It's painted is this sci fi
1:03
Speaker 2: utopia where the machine just knows the answer.
1:07
Speaker 3: It is the ultimate promise of computational power. It's elaked
1:10
Speaker 3: not just a step. But the gap between that promise
1:12
Speaker 3: and where we are today is well, it's filled with
1:15
Speaker 3: some of the most difficult engineering problems in human history.
1:18
Speaker 2: And today, we are going to zoom in on one
1:20
Speaker 2: of those problems.
1:21
Speaker 3: Way way in, right down to the component.
1:23
Speaker 2: Level, past the headlines, past the hype, and right down
1:27
Speaker 2: to the microscopic reality of what it actually takes to
1:30
Speaker 2: build one of these machines. Because it turns out building
1:32
Speaker 2: the future isn't just about big ideas. Wow, It's about
1:36
Speaker 2: controlling individual electrons in spaces that are only tens of
1:41
Speaker 2: nanometers wide.
1:42
Speaker 3: It is an engineering challenge as much as it is
1:44
Speaker 3: a physics challenge. You are you're literally wrestling with nature
1:48
Speaker 3: at its most fundamental level. Howso well, you're trying to
1:52
Speaker 3: force sub atomic particles to behave against their natural chaotic instincts,
1:57
Speaker 3: to hold them still and get them to do useful
1:59
Speaker 3: work for you.
2:00
Speaker 2: And we have a specific tension to look at today.
2:02
Speaker 2: We want to pack these things called quantum dots closer
2:05
Speaker 2: and closer together to build powerful computers, because a computer
2:09
Speaker 2: with two bits isn't very useful, not at all. But
2:11
Speaker 2: new research shows that these tiny neighbors are starting to
2:14
Speaker 2: well shout over each other.
2:15
Speaker 3: Shouting is a very vivid way to put it. In
2:18
Speaker 3: the quantum world, shouting is more about energy exchange and
2:20
Speaker 3: frequency shifts, but the effect is the same distraction, interference, noise.
2:26
Speaker 2: And we're basing today's discussion on a brand News study
2:30
Speaker 2: who was published just yesterday February twelve, twenty twenty six,
2:33
Speaker 2: and physical review applied.
2:35
Speaker 3: Yes, this is coming out of the Reichen Center for
2:37
Speaker 3: Quantum Computing. The research was led by Takashi Kobyashi and
2:40
Speaker 3: his team.
2:40
Speaker 2: And Reichen is a big deal in this space.
2:42
Speaker 3: Oh absolutely. Reraichen is a heavyweight in this field. So
2:45
Speaker 3: when they publish a paper on noise mechanisms, the community,
2:49
Speaker 3: you know it, sits up and listens.
2:50
Speaker 2: So what is our mission here? We aren't looking at
2:53
Speaker 2: a breakthrough where everything works perfectly. We are looking at
2:56
Speaker 2: a hurdle that was just discovered or at least just
3:00
Speaker 2: precisely measured.
3:01
Speaker 3: That's the key.
3:02
Speaker 2: The goal is to understand what happens when quantum bits
3:05
Speaker 2: or quibbits get too close for comfort.
3:08
Speaker 3: It's a story about the growing pains of technology. To
3:11
Speaker 3: solve a problem, you first have to measure it precisely.
3:14
Speaker 3: That is what Reichen has done. They haven't just found
3:17
Speaker 3: a problem, They've put a number on it, They've quantified it.
3:20
Speaker 2: So we're building the supercomputer of the future. But the
3:23
Speaker 2: parts are arguing with each other.
3:25
Speaker 3: In a manner of speaking. Yes, but understanding that argument
3:30
Speaker 3: is the key to silencing it, or perhaps, as we'll
3:33
Speaker 3: see later, even using it.
3:35
Speaker 2: Okay, let's get into the nuts and bolts. Then, before
3:37
Speaker 2: we can really talk about the noise, we need to
3:39
Speaker 2: understand the instrument. The study focuses on something called quantum dots.
3:44
Speaker 2: Now I see that term, and my first thought is
3:47
Speaker 2: those high NTV screens with the really vibrant colors. Are
3:50
Speaker 2: we talking about the same thing.
3:52
Speaker 3: That's a great question, and it's a common point of confusion.
3:55
Speaker 3: It's the same underlying physics, yes, but for a completely
3:58
Speaker 3: different application.
3:59
Speaker 2: Okay.
4:00
Speaker 3: In a TV, a quantum dot is a semiconductor nanocrystal
4:04
Speaker 3: that's designed to glow a very specific, pure color when
4:08
Speaker 3: you hit it with light or electricity. It's an emitter.
4:11
Speaker 2: It makes pretty colors, it makes very.
4:12
Speaker 3: Pretty, very precise colors. But in a quantum computer, a
4:16
Speaker 3: quantum dot is a trap.
4:18
Speaker 2: A trap.
4:18
Speaker 3: Think of it as a potential well, a tiny, tiny
4:22
Speaker 3: region in a semiconductor material where we can confine a
4:25
Speaker 3: single charged particle, in this case a single electron.
4:29
Speaker 2: So you're catching one electron and holding it in place.
4:31
Speaker 3: That's the entire goal.
4:32
Speaker 2: Okay, so why silicon. I feel like when people see
4:37
Speaker 2: pictures of quantum computers, they see those beautiful, big, golden chandeliers.
4:42
Speaker 2: The superconducting quantum computers from Google or IBM, those don't
4:46
Speaker 2: look like chips. But Reichen is using silicon.
4:50
Speaker 3: That is a crucial distinction, and it really defines the
4:52
Speaker 3: different competing approaches. The chandelier computers use superconducting circuits. Those
4:58
Speaker 3: are in electronic terms, quite lo large. They're macroscopic loops
5:02
Speaker 3: of metal cooled to extreme temperatures.
5:05
Speaker 2: And what Ryichan is doing well, Reichen.
5:07
Speaker 3: Is working on what's known as silicon's spin quibits. Looks
5:10
Speaker 3: much more like the chips you'd find in your laptop
5:12
Speaker 3: or your phone.
5:13
Speaker 2: And the advantage of that is, the hope.
5:15
Speaker 3: Is that because we already have, you know, fifty years
5:18
Speaker 3: of experience and trillions of dollars of infrastructure for manufacturing
5:22
Speaker 3: billions of transistors on a silicon wafer, right the whole
5:25
Speaker 3: semiconductor industry exactly, the hope is we can eventually leverage
5:28
Speaker 3: all of that to manufacture billions of kribbits the same way.
5:32
Speaker 2: So it's a bed on scalability. If we can make
5:34
Speaker 2: it work in silicon, we can make millions of them
5:37
Speaker 2: and maybe make them cheap.
5:38
Speaker 3: That is the long term vision. But making a standard
5:42
Speaker 3: transistor is very very different from making a quantum dot
5:45
Speaker 3: for a quibit. How so, well, in a transistor, you're
5:49
Speaker 3: essentially just creating a switch for a current. You're controlling
5:52
Speaker 3: a river of electrons flowing through a channel.
5:55
Speaker 2: On or off a whole crowd of them.
5:57
Speaker 3: A whole crowd. But in a quantum dot, you need
6:00
Speaker 3: to isolate one single electron from that river, just one,
6:05
Speaker 3: just one, and you need to hold it there very
6:07
Speaker 3: gently without disturbing its fragile quantum state.
6:11
Speaker 2: Okay, so how do you build this trap? You can't
6:13
Speaker 2: exactly use tiny tweezers.
6:15
Speaker 3: No, you use electric fields. It's actually quite clever. You
6:17
Speaker 3: fabricate tiny metal gates electrodes on top of the silicon.
6:21
Speaker 2: Wafer, like the gates in a normal transistor.
6:23
Speaker 3: Very similar, but the arrangement is different. When you apply
6:26
Speaker 3: a negative voltage to these metal gates, they create an
6:30
Speaker 3: electric field that repels the electrons and the silicon underneath.
6:34
Speaker 2: So you're pushing them away.
6:35
Speaker 3: You're pushing them away and by arranging these gates in
6:38
Speaker 3: a very specific pattern, like a little corral, you can
6:41
Speaker 3: create a small island, a box of low energy surrounded
6:45
Speaker 3: by high energy walls. An electron can get trapped in
6:48
Speaker 3: that box.
6:49
Speaker 2: So it's like using invisible force fields to build a
6:52
Speaker 2: jail cell for an electron.
6:53
Speaker 3: A very very small jail cell. We are talking about
6:56
Speaker 3: structures that are just tens of nanometers across.
6:59
Speaker 2: And once the electron is in there, what are we
7:01
Speaker 2: actually using? As the bit in my laptop, it's voltage
7:04
Speaker 2: being high or low that represents a one or a zero.
7:07
Speaker 2: What is it here?
7:08
Speaker 3: Here? We are using a fundamental quantum property of the electron.
7:12
Speaker 3: We're using its spin spin.
7:15
Speaker 2: This is one of those chronum terms that everyone uses,
7:17
Speaker 2: but it's always a bit fuzzy. Is the electron actually
7:20
Speaker 2: spinning like a tiny top?
7:22
Speaker 3: Not literally? No, that's just a helpful classical analogy. And
7:25
Speaker 3: electron is a point particle. It has no physical size
7:28
Speaker 3: to spin around.
7:28
Speaker 2: So what is it? Then?
7:29
Speaker 3: It's an intrinsic property like its charge or its mass.
7:33
Speaker 3: It has an intrinsic angular momentum and an associated magnetic
7:37
Speaker 3: moment It behaves as if it were a tiny spinning bar.
7:40
Speaker 4: Magnet a tiny magnet eject and.
7:42
Speaker 3: That tiny magnet can point in different directions. For our purposes,
7:46
Speaker 3: we care about whether it's pointing up or down relative
7:50
Speaker 3: to an external magnetic field we apply.
7:52
Speaker 2: Okay, so spin up is a one and spin down
7:54
Speaker 2: is a zero. Is that the basic idea.
7:57
Speaker 3: That is the basis of it. Yes, that's how you
7:59
Speaker 3: define your computationational lysis states. But because it's a.
8:02
Speaker 4: Quantum system, it's not that simple.
8:03
Speaker 3: It's not that simple. It can also exist in a
8:06
Speaker 3: superposition of both up and down at the same time
8:09
Speaker 3: until you measure it. That's where the real power comes from, right.
8:12
Speaker 2: The classic quantum weirdness. So, just to recap, we have
8:16
Speaker 2: a silicon chip. We have tiny electric gates creating a trap.
8:20
Speaker 2: We catch one single electron. We use its magnetic spin
8:24
Speaker 2: as our data storage.
8:25
Speaker 3: Correct. But here's the problem, the core vulnerability of this
8:28
Speaker 3: whole approach. That spin state is incredibly fragile. Well canness
8:32
Speaker 3: it up almost anything. The biggest enemy is thermal energy
8:35
Speaker 3: just heat. Heat is just atoms and electrons jiggling around.
8:39
Speaker 3: If the environment is too hot, the electron shakes, it
8:42
Speaker 3: interacts with the lattice of silicon atoms around it, and
8:44
Speaker 3: the spin state just flips randomly. Your data is lost instantly,
8:49
Speaker 3: your one becomes a zero, or your superposition just collapses.
8:52
Speaker 3: The information is gone. This is called decoherence.
8:56
Speaker 2: So you have to freeze it.
8:57
Speaker 3: You have to freeze it to near absolute zero. These
9:00
Speaker 3: chips don't just sit on a lab bench. They live
9:03
Speaker 3: inside a complex machine called a dilution refrigerator, operating at
9:06
Speaker 3: temperatures in the millikelvin range.
9:08
Speaker 2: How cold is that?
9:09
Speaker 3: It is significantly colder there than in deep space.
9:13
Speaker 2: So I should picture this a giant humming silver cylinder
9:17
Speaker 2: in a lab. Inside it, there are layers and layers
9:20
Speaker 2: of gold and copper plates that get progressively colder and colder,
9:25
Speaker 2: and at the very bottom, in the deepest cold sits
9:28
Speaker 2: this tiny silicon chip with these nanometer scale traps.
9:32
Speaker 3: That is the exact picture, and that is where the
9:34
Speaker 3: racing team is working. They aren't just looking at one dot.
9:37
Speaker 3: Though one quibbit is a science experiment, a computer needs many.
9:42
Speaker 2: And this is where we hit the scaling problem, the
9:43
Speaker 2: crowded house, as we called it. Right.
9:45
Speaker 3: To perform calculations, to run in algorithm, quibits need to
9:49
Speaker 3: interact with each other in a controlled way. You need
9:51
Speaker 3: to be able to entangle them.
9:53
Speaker 2: Which means they can't be miles apart.
9:55
Speaker 3: They need to be close, very close.
9:56
Speaker 2: How close were we're talking.
9:57
Speaker 3: In the device and this study, the which is the
10:01
Speaker 3: distance from the center of one quantum dot to the
10:03
Speaker 3: center of the next, is about one hundred to one
10:05
Speaker 3: hundred and fifty nanometers.
10:07
Speaker 2: That is incredibly tight.
10:08
Speaker 3: It is extremely tight. And when you pack charged particles
10:11
Speaker 3: that close together, you run into a fundamental force of nature,
10:15
Speaker 3: the kulum interaction.
10:17
Speaker 2: Remind us what that is, It's.
10:18
Speaker 3: The most basic electric force. Like charges repel, opposite charges attract.
10:23
Speaker 3: So two electrons, two electrons are both negatively charged. They
10:27
Speaker 3: hate each other. They are constantly trying to push each
10:30
Speaker 3: other away.
10:31
Speaker 2: So i have two of these quantum dots right next
10:33
Speaker 2: to each other, and I've got one electron in each one.
10:36
Speaker 2: They're constantly pushing against the walls of their little jail cells,
10:40
Speaker 2: trying to get away from their neighbor.
10:42
Speaker 3: Ideally, the electric field walls of the trap are strong
10:45
Speaker 3: enough to hold them in place, but they still feel
10:48
Speaker 3: each other's presence. The force is still there. Okay, So
10:51
Speaker 3: if the electron in Dot A shifts its position slightly,
10:54
Speaker 3: and it can. It has a little bit of wiggle room,
10:57
Speaker 3: maybe just by a nanometer.
10:58
Speaker 4: The electron in dot beef that the electron.
11:01
Speaker 3: In dot B feels a change in the repulsive force.
11:04
Speaker 3: The push gets a little stronger or a little weaker.
11:06
Speaker 2: It's like if you are standing in a small room
11:09
Speaker 2: with a powerful magnet strapped to your chest and someone
11:11
Speaker 2: in the next room had one too, even if there's
11:14
Speaker 2: a wall, if they take a step towards that wall,
11:16
Speaker 2: you feel a push.
11:17
Speaker 3: That is a perfect analogy, and this brings us right
11:21
Speaker 3: to the core of the Reichen discovery. Takashi Kobayashi and
11:25
Speaker 3: his team, we're investigating this exact scenario. They wanted to know,
11:29
Speaker 3: does that push from the neighbor mess up the information
11:33
Speaker 3: stored in the spin of our kubit.
11:35
Speaker 2: And the answer I'm going to assume is yes, otherwise
11:38
Speaker 2: we wouldn't be doing this whole conversation.
11:39
Speaker 3: The answer is a definitive yes. But the mechanism, the
11:43
Speaker 3: how is the really fascinating part because it's not a
11:46
Speaker 3: direct effect. It involves a component we haven't even mentioned yet. Oh,
11:49
Speaker 3: the micro magnet.
11:50
Speaker 2: The micromagnet. Okay, wait, let's back up. We have electric
11:54
Speaker 2: gates to trap the electron. We have the electron itself
11:57
Speaker 2: acting like a tiny magnet with its spin, And why
12:01
Speaker 2: on earth do we need another magnet.
12:03
Speaker 3: This is a great question, and it gets right to
12:05
Speaker 3: the heart of why silicon spin quibets are so hard
12:08
Speaker 3: to build and control. Remember, what we want to do
12:11
Speaker 3: is control the spin of the electron.
12:14
Speaker 2: Right flip it from up to down, or put it
12:16
Speaker 2: in this superposition exactly.
12:18
Speaker 3: Now, spin responds to magnetic fields, so the obvious way
12:22
Speaker 3: to control it would be to zap it with a
12:24
Speaker 3: little magnetic pulse.
12:25
Speaker 2: Seems simple enough.
12:26
Speaker 3: It is if you only have one quibbit. But remember
12:29
Speaker 3: we're in a crowded house. If you want to build
12:31
Speaker 3: a real computer, you have thousands, maybe millions, of these
12:34
Speaker 3: dots packed together, and if I.
12:36
Speaker 2: Turn on a magnet, everything nearby feels it. It's hard
12:39
Speaker 2: to target just one dot without disturbing all its neighbors.
12:42
Speaker 3: Precisely, it's called addressability, and it's a huge problem. You'd
12:46
Speaker 3: have magnetic cross stock everywhere. So the engineer's thought, is
12:50
Speaker 3: there a better way. We are very good at generating
12:53
Speaker 3: localized electric fields. We can switch voltages on those tiny
12:56
Speaker 3: gates incredibly fast and with pinpoint precision.
13:00
Speaker 2: So they wanted to control the spin using electricity instead
13:03
Speaker 2: of magnetism.
13:04
Speaker 3: That was the goal.
13:05
Speaker 2: But can you even do that? I thought you said
13:06
Speaker 2: spin doesn't talk to electricity.
13:08
Speaker 3: It doesn't, not directly. You need a translator. You need
13:12
Speaker 3: a physical mechanism that couples the electrons position which electricity controls,
13:16
Speaker 3: to its spin state. This is a real physical effect
13:20
Speaker 3: called spin orbit coupling. Okay, in some materials like gallium arsenide.
13:24
Speaker 3: This effect happens naturally and is quite strong. In silicon,
13:29
Speaker 3: which we want to use for manufacturing reasons, it's naturally
13:32
Speaker 3: very very weak.
13:33
Speaker 2: So we have to fake it.
13:34
Speaker 3: We have to engineer it. We create something called synthetic
13:36
Speaker 3: spin orbit coupling.
13:37
Speaker 2: Synthetic spin orbit coupling.
13:39
Speaker 3: That sounds expensive, it's clever engineering. This is why they
13:42
Speaker 3: install the micromagnet. They physically place a tiny strip of
13:46
Speaker 3: magnetic material like cobalt, right on top of the quantum dots.
13:50
Speaker 3: But this magnet isn't designed to create a uniform field.
13:53
Speaker 3: That's the key. It's designed to create a gradient.
13:56
Speaker 2: A gradient, what do you mean by that? So the
13:58
Speaker 2: magnetic field is stronger on one side of the day the.
14:00
Speaker 3: Other exactly that. Imagine the quantum dot is a small room.
14:04
Speaker 3: The magnetic field at the left wall is let's say,
14:06
Speaker 3: one hundred units strong. The magnetic field at the right
14:09
Speaker 3: wall is one hundred and five units strong. It's a slope.
14:12
Speaker 3: It's not flat.
14:14
Speaker 2: Okay, I have a magnetic slope across my tiny room.
14:17
Speaker 2: Why no?
14:18
Speaker 3: Remember the frequency at which the electrons spin processes the
14:22
Speaker 3: speed at which it wobbles like a top, depends directly
14:26
Speaker 3: on the strength of the magnetic field it is sitting in.
14:28
Speaker 3: This is called the Larmer frequency.
14:30
Speaker 2: So if the electron is sitting on the left side
14:32
Speaker 2: of the room, it spin wobbles it's speed one hundred.
14:35
Speaker 2: If a moving to the right side, it wobbles its
14:37
Speaker 2: speed one oh five.
14:38
Speaker 3: Precisely. Now you see the trick. We can use our
14:40
Speaker 3: electric gates to gently nudge the electrons position. If we
14:44
Speaker 3: apply a little voltage, we can push it from the
14:46
Speaker 3: left side of the dove to the right.
14:48
Speaker 2: Huh. And by pushing it we force it to move
14:50
Speaker 2: into a stronger part of the magnetic.
14:52
Speaker 3: Field, which changes its spin speed.
14:54
Speaker 2: Uh huh. So we are using electricity to move it,
14:56
Speaker 2: but it's the motion through that pre engineered magnetic gradient
15:00
Speaker 2: that actually changes the spin.
15:02
Speaker 3: You've got it. The micromagnet acts as a translator. It
15:05
Speaker 3: converts an electric signal, which is a change in position,
15:07
Speaker 3: into a magnetic control, which is a change in spin rotation.
15:11
Speaker 4: Skeeed.
15:12
Speaker 3: It acts as a lever. It makes the electrons spin
15:15
Speaker 3: highly sensitive to its physical location.
15:17
Speaker 2: That sounds brilliant. It's a really clever workaround problem solved right.
15:21
Speaker 3: It is brilliant. It's a technique called electric dipole spin
15:24
Speaker 3: resonance or EDSR, and it allows for very fast, localized
15:29
Speaker 3: control of individual quibbits.
15:31
Speaker 2: But there's always a butt.
15:32
Speaker 3: There is no free lunch in physics.
15:34
Speaker 2: I knew it. Here comes the bill.
15:36
Speaker 3: By making the spin exquisitely sensitive to its own position,
15:40
Speaker 3: you have also made it exquisitely sensitive to anything else
15:44
Speaker 3: that might shift.
15:45
Speaker 2: Its position, like the neighbor pushing on the wall.
15:47
Speaker 3: Like the neighbor. Let's go back to our crowded house.
15:50
Speaker 3: We have two dots, Dot A and Dot B side
15:52
Speaker 3: by side. Each one is sitting under this micromagnet gradient. Okay,
15:56
Speaker 3: Now suppose the electron in the neighboring dot Dot A
16:00
Speaker 3: move maybe we're performing an operation on it, or maybe
16:03
Speaker 3: it just jiggles because of some background charge noise.
16:05
Speaker 2: The neighbor moves the sofa.
16:07
Speaker 3: The neighbor moves the sofa. When that electron in Dot
16:09
Speaker 3: A moves, the kolum repulsion it exerts on our electron
16:13
Speaker 3: in Dot B changes, the.
16:14
Speaker 2: Push gets stronger or weaker.
16:16
Speaker 3: And because our electron in Dot B is sitting in
16:20
Speaker 3: this carefully balanced electric trap, that tiny change in push
16:24
Speaker 3: from the neighbor is enough to shift its positions slightly.
16:28
Speaker 2: And because of the micromagnet ingredient, when our electron.
16:30
Speaker 3: Moves, it moves into a different magnetic field strength, and
16:34
Speaker 3: therefore it spin frequency changes bingo.
16:37
Speaker 2: The neighbor's motion has changed the speed at which our
16:40
Speaker 2: quibod is spinning. This is the crosstock. This is what
16:44
Speaker 2: the Reichen paper calls a charge induced spin frequency shift.
16:48
Speaker 3: That's it in a nutshell.
16:49
Speaker 2: So the micromagnet is a complete double edged sword. It
16:52
Speaker 2: gives us the precise control we need, but at the
16:55
Speaker 2: same time it acts as an amplifier for the noise
16:58
Speaker 2: coming from the neighbors.
16:59
Speaker 3: That is the crux of the problem they investigated. You
17:01
Speaker 3: built a system that is hyper sensitive to motion, so
17:04
Speaker 3: you could control it with electricity, but now it's hyper
17:07
Speaker 3: sensitive to the unwonted motion of the guy next door.
17:10
Speaker 2: Okay, so this sounds bad in theory. But was it
17:13
Speaker 2: just a theoretical worry or is it actually messing things
17:16
Speaker 2: up in a real device.
17:17
Speaker 3: That is the billion dollar question, and it's what Kobayashi's
17:21
Speaker 3: team went to find out. They didn't just model this
17:24
Speaker 3: on a computer. They built the device, They cooled it
17:26
Speaker 3: down to one hundred millikelvin, and they measured it directly.
17:30
Speaker 2: Wow.
17:30
Speaker 3: And the way they measured it is fascinating in itself.
17:33
Speaker 2: I was going to ask, how do you measure a
17:34
Speaker 2: tiny shift in the wobble frequency of a single electron
17:38
Speaker 2: trapped in a piece of silicon.
17:40
Speaker 3: You use a very sensitive technique called Ramsey.
17:43
Speaker 2: Interferometry ramciator ferometry. Is that related to Ramsey kitchen nightmares?
17:47
Speaker 3: No relation, though it can be a nightmare if you
17:49
Speaker 3: don't get the calibration right. It works a bit like this.
17:53
Speaker 3: You start with the electron spin pointing up. Let's call
17:55
Speaker 3: that our starting line. Okay, you hit it with a
17:58
Speaker 3: precisely timed microwave pul ulse that puts it into a
18:02
Speaker 3: perfect superposition half up, half down. Now you can picture
18:06
Speaker 3: its state as a clockhand pointing, say at three o'clock.
18:09
Speaker 3: Got it, and then you just wait. You let it evolve.
18:12
Speaker 3: That clockhand starts sweeping around at its natural frequency, the
18:15
Speaker 3: Larmer frequency, ticktick. You let it tick for a very
18:19
Speaker 3: specific amount of time. Then you hit it with a
18:22
Speaker 3: second identical microwave pulse. If the frequency was exactly what
18:27
Speaker 3: you thought it was, this second pulse will push the
18:29
Speaker 3: clock hand all the way around to point down, so.
18:32
Speaker 2: You know exactly where it should end up. It's like
18:34
Speaker 2: checking your watch against a master clock exactly.
18:37
Speaker 3: But what if the frequency has shifted during that waiting time.
18:41
Speaker 3: What if the neighbor moved pushed our electron into a
18:44
Speaker 3: stronger magnetic field and made the clock tick a little
18:47
Speaker 3: faster than.
18:48
Speaker 2: The hand won't be where you expect it to be.
18:50
Speaker 3: It will have rotated too far when you hit it
18:52
Speaker 3: with that second pulse. It won't end up pointing down.
18:55
Speaker 3: It'll be somewhere else. This difference between where it is
18:58
Speaker 3: and where it should be is called a phase error.
19:00
Speaker 2: And in quantum computing, phases everything.
19:03
Speaker 3: Phase is where the information is stored in many algorithms,
19:06
Speaker 3: if the phase drifts uncontrollably, the calculation is wrong. It's
19:10
Speaker 3: a fatal error.
19:12
Speaker 2: So what did the Reichen team do with this technique?
19:14
Speaker 3: They performed this Ramsey experiment over and over again, but
19:18
Speaker 3: while our quibbet's clock was ticking, they were deliberately moving
19:21
Speaker 3: the neighboring electron back and forth using its control gate.
19:24
Speaker 2: They poked the bear just to see if it would growl,
19:26
Speaker 2: and it did.
19:27
Speaker 3: They observed a clear, perfectly correlated shift in the quibts
19:31
Speaker 3: frequency that matched the movement of the neighbor electron. When
19:35
Speaker 3: the neighbor moved closer, the frequency went up by a
19:38
Speaker 3: specific amount. When it moved away, it went down. They
19:41
Speaker 3: mapped it out precisely, so.
19:43
Speaker 2: It's not theoretical anymore.
19:44
Speaker 3: It's a hard number, a quantified, measured effect. This is
19:47
Speaker 3: the first time this specific mechanism has been directly measured
19:51
Speaker 3: and isolated.
19:52
Speaker 2: How bad was it? Was it a tiny blip, something
19:54
Speaker 2: that you could maybe ignore.
19:55
Speaker 3: According to the paper, the shift is large enough to
19:58
Speaker 3: cause a considerable quibit error rate.
20:01
Speaker 2: A considerable error rate. That doesn't sound good.
20:03
Speaker 3: It's not We measure the quality of our quantum operations.
20:06
Speaker 3: In terms of fidelity, a perfect operation has one hundred
20:10
Speaker 3: percent fidelity. To make a fault tolerant quantum computer work,
20:13
Speaker 3: you need operation fidelities of ninety nine point nine percent
20:17
Speaker 3: or really ninety nine point nine nine percent or higher.
20:19
Speaker 2: So you can only have one error in ten thousand
20:22
Speaker 2: operations at most.
20:24
Speaker 3: This cross stalk mechanism introduces errors that can easily drop
20:28
Speaker 3: you well below that threshold.
20:29
Speaker 2: So this isn't just background here so we can filter out. Yeah,
20:32
Speaker 2: this is a potential showstopper. If you can't fix this,
20:36
Speaker 2: you can't scale up to a large number of quibits.
20:39
Speaker 3: Show stopper might be too strong, but it is definitely
20:42
Speaker 3: a gatekeeper. It's a problem you cannot pass until you
20:46
Speaker 3: solve it. If you try to build a chip with
20:48
Speaker 3: a million quibbits and they are all shifting each other's
20:50
Speaker 3: frequencies every time.
20:51
Speaker 2: They move, you just have a million sources of noise.
20:54
Speaker 3: You don't have a computer. You have a very expensive,
20:56
Speaker 3: very cold random number generator.
20:58
Speaker 2: Which is useful for casinos maybe, but not for curing
21:01
Speaker 2: cancer exactly.
21:03
Speaker 3: This is what the paper refers to when it says
21:05
Speaker 3: this problem prevents the silicon quantum computing community from realizing
21:09
Speaker 3: a large scale device. It's a roadblock.
21:11
Speaker 2: So we have identified the villain. It's this unlucky combination
21:15
Speaker 2: of the crowded neighborhood, the Kulam interaction, and the very
21:19
Speaker 2: sensitive amplifier that we put in for control, the micromagnet.
21:22
Speaker 3: That's the mechanism.
21:23
Speaker 2: Now. Usually when engineers find a source of noise, they
21:27
Speaker 2: try to eliminate it. They add shielding, or they redesign
21:31
Speaker 2: the parts to be further away. Can we do that here?
21:33
Speaker 3: Well, moving them further away as a non starter. You
21:36
Speaker 3: lose the ability to entangle them, which you need for computation.
21:39
Speaker 2: Okay, so that's out. What about shielding? Can you put
21:42
Speaker 2: a little wall between them?
21:44
Speaker 3: Shielding electric fields at the Mano scale inside a solid
21:47
Speaker 3: piece of silicon is incredibly difficult. It's not really a
21:51
Speaker 3: practical solution right now. So Kobayashi and his team suggest
21:55
Speaker 3: a different approach in the paper what's the plan? The
21:57
Speaker 3: plan has two parts. The first is mitigate. The second
22:01
Speaker 3: and more exciting part is exploitation.
22:03
Speaker 2: Mitigation and exploitation. Okay, let's start with mitigostion first.
22:07
Speaker 3: The key discovery here is that we have quantified it
22:09
Speaker 3: we have measured it, we can predict it.
22:11
Speaker 2: Right. It's not random noise, it's correlated noise.
22:15
Speaker 3: That is the keyword. Random noise is a nightmare to fight.
22:19
Speaker 3: It's like trying to have a conversation in a room
22:21
Speaker 3: with a thousand people talking at once. But if the
22:24
Speaker 3: noise is correlated, if I know that every single time
22:27
Speaker 3: neighbor A moves to the left, my quibbit B is
22:30
Speaker 3: going to shift its frequency by exactly five megahertz, then
22:33
Speaker 3: I can do something about it.
22:34
Speaker 2: You can compensate for it exactly.
22:36
Speaker 3: This leads us to a strategy called feed forward correction.
22:40
Speaker 3: The classical computer that controls the quantum ship can be
22:43
Speaker 3: programmed with this knowledge.
22:45
Speaker 2: How would that work in practice?
22:47
Speaker 3: When the control system sends a command pulse to the
22:49
Speaker 3: gate of neighbor A telling it to move, it can,
22:52
Speaker 3: at the exact same time send a small corresponding correction
22:56
Speaker 3: pulse to the gate of quibbet B.
22:58
Speaker 2: It's like canceling headphones.
23:01
Speaker 3: That's a fantastic analogy noise. Canceling headphones have a microphone
23:04
Speaker 3: that listens to the outside sound, and then the electronics
23:07
Speaker 3: generate an inverse sound wave to cancel it out. In
23:10
Speaker 3: your ear. Here, we don't even need to listen. We
23:13
Speaker 3: know the sound the frequency shift is coming because we're
23:17
Speaker 3: the ones causing it by moving the neighbor. So we
23:19
Speaker 3: can preemptively adjust the dial the control voltage on quibit
23:23
Speaker 3: B to counteract the effect precisely as it happens.
23:27
Speaker 2: So you fix the error before it even has a
23:29
Speaker 2: chance to mess up the calculation.
23:31
Speaker 3: Ideally, yes, you can program these corrections right into the
23:34
Speaker 3: gate operations. It adds complexity to the control software, but
23:38
Speaker 3: it's a viable path forward.
23:39
Speaker 2: Okay, that makes sense. That's the mitigation was the exploitation part.
23:43
Speaker 3: This is the part that I find most exciting. Kobayashi's
23:45
Speaker 3: team suggests that instead of just canceling this effect, we might.
23:48
Speaker 4: Be able to use it, use the noise.
23:50
Speaker 2: How well.
23:51
Speaker 3: Think about what this crosstalk effect actually represents. It is
23:54
Speaker 3: a form of coupling. It is a connection between two
23:57
Speaker 3: quibbits that were not supposed to be directly interacting.
24:00
Speaker 2: Quibot A is talking to quibt B, even if we
24:02
Speaker 2: didn't want it to initially. It's an accidental phone line.
24:05
Speaker 3: And in quantum logic, what do we need to do?
24:08
Speaker 3: We need to perform two quibit gits. These are operations
24:11
Speaker 3: where the state of one quibbot conditionally changes the state
24:15
Speaker 3: of another. That is how you build logic. That is
24:18
Speaker 3: how you build an algorithm like Shores algorithm for factoring.
24:22
Speaker 2: So normally you have to work really hard to force
24:24
Speaker 2: them to talk to each other in a controlled way.
24:27
Speaker 3: Yes, a common way is to physically lower the electric
24:31
Speaker 3: potential barrier between them, let their electron wave functions overlap,
24:36
Speaker 3: and let them interact via something called the exchange interaction.
24:40
Speaker 3: It's tricky and has to be managed very carefully. But here,
24:43
Speaker 3: but here, the micromagnet has accidentally given us a new
24:47
Speaker 3: long range interaction. They are talking to each other via
24:50
Speaker 3: the magnetic gradient and the electric field even when their
24:52
Speaker 3: wave functions aren't touching.
24:54
Speaker 2: So the researchers are basically saying, don't mute the crosstalk.
24:57
Speaker 2: Let's try to turn it into a high fidelity phone line.
25:00
Speaker 3: That's it. The paper suggests this could lead to a
25:03
Speaker 3: new class of quibit operations. If you can control this
25:08
Speaker 3: energy shift precisely, if you can turn it on and
25:11
Speaker 3: off by moving the neighbor electron, you could potentially use
25:14
Speaker 3: it to perform logic gates between quibots that are not
25:17
Speaker 3: immediate neighbors.
25:18
Speaker 2: So you could do a calculation between quibot A and
25:20
Speaker 2: QUIBTC using quibotb's position as the switch.
25:24
Speaker 3: That's the kind of new capability it might open up.
25:26
Speaker 3: You can turn this bug into a feature.
25:28
Speaker 2: That is the ultimate engineering judo move, use the force
25:32
Speaker 2: of the problem itself to solve the problem.
25:34
Speaker 3: It connects to the broader history of developing these silicon quibots.
25:37
Speaker 3: Time and again we're finding that things we initially thought
25:39
Speaker 3: were bugs, things like subtle variations in the silicon crystal,
25:43
Speaker 3: the valley splitting, or roughness at the silicon oxide interface.
25:47
Speaker 3: And now this micromagnet cross stock can sometimes be understood, controlled,
25:51
Speaker 3: and engineered into resources if.
25:54
Speaker 2: You understand the physics deep enough.
25:56
Speaker 3: If you understand the physics deep enough, it's a testament
25:59
Speaker 3: to how mature the field is getting. We're moving from
26:02
Speaker 3: just trying to get one quibot to work to understanding
26:05
Speaker 3: the complex multi quibot interactions in a large system.
26:09
Speaker 2: It seems like a constant race between the problems we
26:12
Speaker 2: discover and the solutions we invent.
26:13
Speaker 3: It is the silicon quantum computing community, as the text
26:17
Speaker 3: calls it, is in a fierce but friendly race against
26:21
Speaker 3: the superconducting folks and the trapped ion folk.
26:23
Speaker 2: Different horses in the same race.
26:25
Speaker 3: Right, and silicon has this massive theoretical advantage of manufacturability,
26:31
Speaker 3: but it has these complex material physics challenges that the
26:34
Speaker 3: other platforms don't. This study is a major step in
26:37
Speaker 3: clearing the fog around one of those challenges.
26:39
Speaker 2: So let's just zoom back out for a second. We
26:42
Speaker 2: started with single electron shivering in the cold. We built
26:44
Speaker 2: an electric jail cell to trap it. We gave it
26:47
Speaker 2: a special magnet to control it. We then realized that
26:49
Speaker 2: magnet made it listen to its neighbors too much. We
26:52
Speaker 2: then figured out exactly how it was listening, measured the volume,
26:56
Speaker 2: and now we are planning to turn that unwonted listening
26:59
Speaker 2: into a controlled conversation.
27:01
Speaker 3: That is a beautiful summary of the scientific process. Right there,
27:04
Speaker 3: we went from identifying noise to potentially creating a new
27:08
Speaker 3: kind of signal.
27:09
Speaker 2: Why does this matter to the person listening right now?
27:11
Speaker 2: They might be driving to work or doing the dishes.
27:14
Speaker 2: Why should they care about the larmer frequency of a
27:17
Speaker 2: single electron spin?
27:18
Speaker 3: It matters because It peels back the curtain on how
27:21
Speaker 3: progress and technology actually happens. We often get this very clean,
27:26
Speaker 3: linear narrative of progress. We think of technology as just
27:29
Speaker 3: this inevitable march of better, faster, cheaper.
27:32
Speaker 2: But it's not like that at all.
27:34
Speaker 3: It's not. It's a messy, winding path. It's a series
27:37
Speaker 3: of roadblocks. It's discovering that your clever solution, the micromagnet,
27:42
Speaker 3: caused a brand new problem, the crosstalk. It's the painstaking,
27:45
Speaker 3: unglamorous work of measuring and characterizing that problem.
27:49
Speaker 2: It's less of a Eureka moment and more of a huh,
27:52
Speaker 2: that's weird moment, followed by a year of work.
27:55
Speaker 3: A year of very careful work in a very cold room,
27:57
Speaker 3: and it's the realization that you cannot just build it
28:00
Speaker 3: and hope it works. You have to understand the fundamental
28:03
Speaker 3: physics of the noise, of the imperfections.
28:05
Speaker 2: And it really grounds the hype around quantum computing. This
28:10
Speaker 2: isn't magic. It's plumbing. It's just extremely high tech, atomic
28:15
Speaker 2: scale plumbing with very leaky pipes.
28:17
Speaker 3: It is, and for anyone interested in the future of
28:20
Speaker 3: computing watching how the community solves these very specific hurdles
28:24
Speaker 3: like the crosstalk challenge that Reyken just mapped out is
28:27
Speaker 3: a much better preview of the future than any marketing material.
28:31
Speaker 3: This is where the real work is happening.
28:33
Speaker 2: If they do solve this, if they can either cancel
28:36
Speaker 2: this cross talk perfectly or turn it into a reliable tool,
28:39
Speaker 2: does Silicon win the race?
28:41
Speaker 3: Win is a strong word. There might be room for
28:43
Speaker 3: multiple technologies, but if they solve this, Silicon becomes a
28:47
Speaker 3: terrifyingly strong competitor. Why terrifyingly because if you can handle
28:50
Speaker 3: this cross talk, and you can solve a few other
28:52
Speaker 3: known materials issues, and you can make these things using
28:55
Speaker 3: standard semiconductor foundaries, the scaling potential is just massive. You
29:00
Speaker 3: could genuinely have millions of high quality quibbets on a
29:03
Speaker 3: single chip the size of your.
29:04
Speaker 2: Thumbail, whereas a golden chandeliers, as beautiful as they are,
29:07
Speaker 2: take up a whole room for a few hundred or
29:09
Speaker 2: a thousand quibbits.
29:10
Speaker 3: Exactly, density matters for building a truly large scale error
29:14
Speaker 3: corrected machine.
29:16
Speaker 2: So I want to leave the listener with the final thought,
29:17
Speaker 2: something to chew on, and it's about this idea of interconnectedness.
29:22
Speaker 2: One we tend to think of isolation as the ideal
29:24
Speaker 2: state right in our daily lives. We want peace and
29:27
Speaker 2: quiet to focus. In quantum computing, the mantra is always
29:32
Speaker 2: isolate your quibit so it doesn't decohere and lose its information.
29:37
Speaker 3: Isolation preserves the quantum state. That's rule number one.
29:40
Speaker 2: But a computer that is perfectly isolated does nothing. It's
29:42
Speaker 2: just a perfect box of secrets. To compute. To do
29:45
Speaker 2: anything useful, you must connect. You must have interactions.
29:49
Speaker 3: Information is physical, and physics is at its heart the
29:53
Speaker 3: study of interactions.
29:55
Speaker 2: So this struggle that the Reichen team is documenting, this
29:57
Speaker 2: push and pull between keeping the quibit safe from its
29:59
Speaker 2: ne neighbors but also needing it to talk to its neighbors.
30:03
Speaker 2: It's not just an engineering problem. It feels like the
30:05
Speaker 2: fundamental tension of the universe, doesn't.
30:07
Speaker 3: It in a way it is for any information processing system.
30:11
Speaker 3: It is the balance between coherence and coupling. Too much
30:15
Speaker 3: isolation and your system is useless. It can't process anything.
30:18
Speaker 3: Too much connection and it just becomes noise. Everything washes out.
30:23
Speaker 3: Finding that sweet spot, that perfectly controlled interaction is the
30:26
Speaker 3: whole game.
30:27
Speaker 2: And Reichen just showed us exactly where that line is
30:30
Speaker 2: drawn for this type of device.
30:32
Speaker 3: They gave us the map of a new part of
30:34
Speaker 3: that territory. Now the community has to figure out how
30:36
Speaker 3: to navigate it.
30:37
Speaker 2: Well, my brain is suitably entangled. Thank you for walking
30:41
Speaker 2: us through the nanoverse today. This was fascinating.
30:43
Speaker 3: It was a pleasure. It's always good to look at
30:45
Speaker 3: the small things that make the big things possible.
30:47
Speaker 2: We'll see you in the next one.
30:49
Speaker 3: Keep learning, stay curious,