Michelle Simmons: quantum machines at the atomic limit | 91TV
Transcript
- Michelle Simmons: Well, look, it's absolutely wonderful to be here and to be here in person
- with so many of my colleagues and friends. So thank you for coming to this lecture tonight. I'm
- particularly honoured to receive this lecture and to give this lecture because I really do relate
- to Henry Baker. I read about him and I realised that giving an award so that the fellows of the
- Society could get together to discuss natural science was one thing. If you look in detail
- at his philosophy of life, it was about trying to understand the world at its very smallest scale.
- So even back then he saw that the real truth in the world came from understanding nature and the
- way that nature worked. In particular, he was somebody that looked at optical microscopes and
- he saw that, when you look down an optical microscope, you could see things that you
- could see beyond what the naked eye could see. He always thought, this is the way that we're
- going to really understand the way that nature works by getting down smaller and smaller scales.
- That was where the fundamental truth lay. That's something that I inherently believe.
- One of the things that humans have done over the last two to 300 years is develop
- ever increasing tools to be able to see nature at a length scale that we simply couldn't before. The
- more we do that, the more we start to understand how it's composed, how it comes together. That
- miniaturisation for me goes down to the atomic scale. It's something that I fundamentally believe
- new worlds will open up, new machines will open up if we understand nature at that scale.
- Now going back in time. It's fascinating to see how some of the things existed 200 years ago.
- So you can look at in the case of electric motors, it's not so long ago, but electric motors,
- when they first designed them, they were designed to pump water out of mines. They were fairly big,
- bulky machines that existed in a fixed state near the mine. If you look at time, we used to have
- pendulum clocks, and those clocks would only exist in the side of a town hall or a church or in a
- station, so that people could start to plan their lives. The mere presence of time and the fact that
- people could synchronise with each other across the world was something that was quite, quite new
- in the way that the world lived. Then printers obviously go back to the very Stone Age where
- people tapped into stones, but eventually they found out how to create printing machines that
- could get the word out to educate everybody in the world. Then computers, obviously computers allow
- us to calculate things that we've never simply been able to do before. That whole journey that
- we've been on with computers has been fascinating. If you fast forward now, the mere process of
- miniaturisation has put those motors into just about every electronic device that you have,
- whether it's a fridge, whether it's a dishwasher, a hoover a camera, a video camera, those little
- motors, the smaller they get, they basically mean that individuals can now hold on to machines to
- do things that they simply couldn't imagine in the past. If we look at clocks and times, we now have
- watches the smallest feature sizes down to about a millimetre. Everyone can tell time. Timing is
- synchronised across the world. So travelling coming from Australia, everywhere, the planes,
- the boats, the trains they all are synchronised so that now the world communicates in a much
- more different way. With printers, the advent of printers, everyone can now start to print
- things real time. So get information out to the youngest people. They can actually
- print off ideas. They can discuss things. Obviously computers have allowed us to go
- to the moon. They've allowed us to communicate across the world. They've allowed us to solve the
- way the world works in a completely unique way. Each time we've made things smaller and smaller,
- we have transformed the way that society behaves and transformed our understanding of the world.
- If we go back to 1959, another hero of mine is Richard Feynman. Richard Feynman really had this
- lecture. If you read that lecture, I read it many years ago and I've read it again recently,
- I suddenly realised for me it literally has mapped out all the things that I've been doing
- in my life for the last 20 years. He said he's not too afraid to consider the final question.
- Back in 1959, ultimately in the great future, we could arrange the atoms the way we want the
- very atoms all the way down at that time at the scanning tunnelling microscope didn't exist.
- We had optical microscopes. So he was already looking beyond the technology that he had.
- He said what he wants to talk about is controlling the world at the small scale.
- Then he had this quote, 'We must always accept some atomic arrangements. Must we always accept
- some atomic arrangement that nature gives us.' So he realised that the world is made of atoms.
- We see nature forms atoms naturally into different components, into different biological systems.
- What happens if we could control where atoms were? We could actually make things that never
- existed before we could actually create things beyond what humans had thought of.
- Then he had this idea that as far as he could tell, there was no reason physically,
- the principles of physics didn't speak against the ability to manoeuvre the world atom-by-atom.
- So if you just go back and look at the computing industry, it's changed dramatically over the
- last 70 years. So we can see those original valve computers. They were literally the size of a hand.
- They were over 19,000 of them in the 1946 ENIAC computer. If a valve went wrong, it would take
- many hours to find where that valve was. So you could fix the computer to get it work, and
- it took up the size of a room, Obviously with the first transistor in 1947 and Germanium we started
- to integrate the ability to switch between ones and zero states by putting it inside a Germanium
- semiconductor. That was a fairly clunky tool. It's a couple of inches big in Germanium, very Heath
- Robinson experiment. Then in 1958 they started to integrate all those components onto a chip.
- Now we're starting to see how that Miniaturisation is giving a dramatically big change in the
- way that the world operates. We start to get integrated circuits so that by 1981 we start to
- have our first personal computers. That's quite a dramatic change in our ability to take these
- small components, integrate them in a monolithic chip and create devices that didn't exist before.
- Nowadays you take a chip outside of a PC and you see it's about two inches square. It has about
- 14 billion transistors on it. So imagine chopping up 14 billion times to the smallest feature size.
- They put all of these together to get that massively parallel processing ability.
- They put them in server farms that generate so much heat that they have to move them towards
- the Arctic Circle. So this is out in Sweden where they literally put them out there because of the
- heat generation. It's enough heat, enough power generated there to run a small city from the heat,
- and then they just leave the windows open so they don't have to air condition it.
- If you look at the plants that make them, they're now about $15 to $16 billion to
- make a semiconductor manufacturing plant that allows you to make these very small transistors.
- That's quite amazing. It's like 16 football pitches side-by-side
- to make these plants. They're all operated by robots. So it's quite amazing to go
- and see one of those plants globally. Well what's dominated that whole revolution
- in making things smaller and smaller is Moore's Law. So Gordon Moore was the co-founder of Intel,
- and he saw that the feature size roughly doubled every 18 months to two years. Sorry the feature
- size decreased every 18 months, to two years. That's the number of transistors increased doubled
- every 18 months to two years. So you can actually go in here and see the smallest feature size as a
- function of time. What's amazing about this law that Gordon Moore came up with was he basically
- said in order to keep this going, you have to put out a roadmap ahead of time so that you can write
- down all the technical problems that you have to solve in order to keep reducing the feature size
- to make them smaller and smaller. Obviously the smaller they are, the faster they are.
- So he put that law out on the web and it became the ITRS roadmap, the International
- Technology Roadmap for Semiconductors, and it became something that everybody across the world
- would look at all the different technical problems that had to be solved so that you could make those
- feature sizes smaller and smaller every year. It became a self-fulfilling prophecy. So if you look
- at a kind of a single transistor here, it's called a silicon FinFET. It's literally like a shark fin
- that sits out. So you've got your silicon wafer, a very thin, that bright region is an oxide. Then
- over the top of that you've got what we call a wraparound gate. If you put a positive voltage on
- that gate, you basically repel all the electrons inside the fin and you turn the transistor off…
- The positive voltage gate, you attract the electrons into the fin
- and you turn the transistor on. If you put a negative voltage on the gate, you repel the
- electrons and you turn the transistor off. So in such a way you can turn this on and
- off just by applying a voltage. There are 14 billion of these on a transistor chip.
- What we wanted to do is to see where this would go. You can literally plot as a function of time.
- By 2020, theoretically you'd get to the stage where the smallest feature size is an individual
- atom. So back in 2000, in Australia, we decided we would take the technologies that existed out
- there, which were microscopes that allow you to visualise individual atoms and we would adapt
- those microscopes so that you could actually manipulate the atoms and put them in place.
- This is some examples of some lithography in our labs. These are silicon, rows of silicon
- atoms which have been terminated with hydrogen. We come along with a very fine metal tip and we
- remove the hydrogen to give us this atomic scale lithography. What we wanted to do was
- to see can we actually now make electronic devices where the smallest feature size, the
- active component of the device is a single atom. So what I want to do now is just go through and
- look at some of the quotes that Richard Feynman had. He basically said he wasn't afraid to
- consider could you build things out of atoms? So what I wanted to do now is just to show the kind
- of evolution in the microscopes that we have. So obviously light microscopes allow you to see about
- 0.15 millimetres. It's just below what the eye can see if you take an electron microscope, this is an
- onion skin on the left hand side here and you look at the same onion skin. You can now start
- to see the nucleus inside that. You can also look at human hairs which are about 50 microns long.
- If you take a scanning tunnelling microscope now you have an ultrahigh vacuum system.
- You pump everything out, you put your sample in and then you heat it up to high temperature and
- cool it down. You can actually get down now to see the individual atoms. So it's 50000 times smaller
- than the width of a human hair. So really now what we're seeing is the world at the
- atomic scales. These are actually pairs of silicon atoms that sit on the silicon surface.
- So that's quite an amazing, 1984, the Nobel Prize was won the invention of the microscope.
- It really allowed us for the first time to know that atoms did actually exist.
- So how does the microscope work? It has a very fine metal tip that you bring down to the surface
- of your material under voltage control. When you get close enough inside the vacuum system, you get
- a constant current. The closer you get, the higher the current you get and you keep the current
- constant and you scan the tip over the atoms on the surface and as they go over the atoms they
- deflect in height. So the raster scan to form a topographic image of what the surface looks like.
- It's all done in ultrahigh vacuum so that you get rid of everything except for the surface
- that you want. So here is the image of the silicon atoms on the surface. This is really
- the first time that we knew that atoms existed. Well IBM came along and they said, well look,
- let's not just look at that atom on the surface. Let's actually see if we can actually pick those
- atoms up with the tip of the microscope and move them around. So they formed the world's smallest
- logo, IBM, out of picking up metal atoms on a surface by applying a voltage to the tip,
- picking the atom up, moving along and then pulsing it off so that it dropped down. So the hope was
- that we could actually start to build things in semiconductors with that kind of precision.
- The challenge was, however, it's very easy to pick up metal atoms on a metal surface,
- but it's not easy to pick up atoms inside a semiconductor. The bonds are just too strong.
- So we have to come up with another way to actually make devices in silicon. What we
- do is we mimic the kind of lithographic process that you would find in a conventional clean room,
- like an Intel clean room. So the important thing is when you start, you put your silicon
- wafer inside the microscope and you take it out because it's so small you can't see anything. So
- you have to make marks on the surface that you can then see when you put it inside the microscope,
- and when you bring it out. So typically what we do is we etch registration markers in the surface.
- We then load it into the ultrahigh vacuum system. We heat it up to high temperature and cool it down
- so we can reconstruct the surface. Then we basically terminate with atomic scale
- with atomic hydrogen. So we have a cracker source that takes hydrogen and breaks it up.
- That basically covers the whole of the surface. We then come along with a very fine metal tip for the
- scanning tunnelling microscope, and we literally write into the surface. We dissolve the hydrogen
- atoms, exposing the underlying silicon underneath. We then want to bring in the atoms that we're
- going to make the active components of the device. These are our phosphorus atoms. They
- have one extra electron and the phosphorus will as phosphine gas will stick to where the
- silicon is exposed but not to the surrounding hydrogen. So then when we heat the surface up,
- the phosphorus actually go into the top layer of silicon and it will kick the silicon atoms out,
- but only in the regions where the resist is maintained, where the resist is exposed.
- We then take it across to silicon crystal growth system, and we encapsulate the whole device with
- silicon at low temperature to make sure the atoms don't move. Then we basically bring the
- tip a little bit back and we can image where the atoms are beneath the surface to show that they're
- still there. Then using those registration markers we take it out of the system the
- ultrahigh vacuum microscope through to a clean room where we make contacts to the very device.
- Well, to do that basically brought together two different types of technology.
- On the left hand side here, we've got a crystal growth system that's called molecular beam
- epitaxy that allows us to grow layer-by-layer of silicon in the machine just by heating it
- up in an evaporation mode. On the right hand side here we have the scanning tunnelling microscope.
- So we can manipulate the atoms in one system transfer under ultrahigh vacuum and then grow.
- Those two technologies at the time we started this project were actually pretty incompatible.
- To get a very high quality vacuum, you typically have pumps that vibrate
- and at the same time you're trying to manipulate atoms on the other side. So all kinds of technical
- skills and techniques had to be put together so that we could actually manipulate the atoms
- whilst running the crystal growth system. We're using that technique. What we've done now
- is it's a lithography process is we've patterned large regions where we can then put the phosphorus
- down and we can create metallic regions where the phosphorus is in insulating regions around.
- We can create very fine metal wires, and we can go all the way down to the atomic scale.
- So the very first choices we made were something called a single electron transistor. So this is
- the lithographic pattern. We basically have a quantum dot in the middle with a drain
- and a source that we can pass current through it. Then we have a plunger gate here that controls the
- energy levels of the dot, and then either side, here we also have gates that control
- the tunnel barrier. So there's little gaps here so that we create this separate island.
- So when we look at this device we basically have essentially a kind of two dimensional
- channel either side to a dot in the middle. This is represented by this energy level diagram.
- Because the dot is small we basically confine the energy states. So we get these red lines here of
- the different energy states in the dot. Then we have our kind of reservoirs over the side.
- So what we can do is we can apply the source drain bias here. Then by changing this gate bias
- we can move these up and down until they align with the Fermi energy. An electron will hop on
- and off that central island. What you see now is you've measured the conductance through the device
- as a function of this gate bias. You see these peaks in the current as electrons hop on and
- off that central conducting island. Now this device is sufficiently small, it's nanometres,
- 100 nanometres or so in size that we actually only have 4000 atoms on that central island.
- What we can also do is we change the source drain bias so that now we can see lots of different
- energy levels within the window and we get these beautiful what we call Coulomb diamond patterns.
- With this device that has about 4000 atoms on it, every time we remove an atom from
- this we get this little Coulomb diamond where basically they're all the same height. From
- measuring the heights of those diamonds, we can tell what the energy level separation is.
- So this is a way now that we've lithographically patterned a very
- small device. It's encapsulated with silicon. We can now start to see that quantum behaviour.
- Well the next thing Feynman asked was well hang on a minute why can't we make machines out of very,
- very small wires either ten or 100 angstroms in diameter? So that's exactly what we started to do.
- So here we've got a device where this grey region is the STM patterned region. There's our etched
- registration markers. The blue is the encapsulated silicon. Then these yellow regions on top of the
- gold that connects the device that's buried beneath the surface. So in such a way we're
- able to pattern from 1.7 nanometres just a few atoms wide and then take it through to the STM
- with our markers up to the optical microscope and then to a little chip that you can see.
- So we finally now connected all those different length scales using this lithographic patterning
- process. If we look at those wires we can basically zoom into the lithography here.
- So here's our rows of silicon atoms. When we phosphine dose it we get lots of phosphorus
- sitting close together. This is quite amazing because they're so close. It means that basically
- you create a conducting state along the wire, the electron wave function from all of the individual
- phosphorus atoms that incorporate into the wire overlap to create a metallic state.
- So what was quite surprising about this work is even if we make that wire very, very thin,
- the resistivity is maintained. It's got such a high doping density. It's a beautiful
- encapsulated thin wire, only a few atoms wide. Whereas if you look at a lot of the semiconductor
- wires in the literature, normally when they make those wires very thin on a freestanding wire,
- they become very resistive. So this was quite remarkable. It basically meant that Ohm's law
- was maintained down to the atomic scale with a resistivity that was comparable to copper. So just
- by putting phosphorus inside the silicon crystal, you're creating these very fine metallic wires.
- Also the other thing that was fascinating about this is these wires are what we call epitaxial
- wires. They're crystalline wires. So you can take the wire with these gate voltages on either side,
- and you can make the wire different thicknesses and measure the conductance through the wire.
- If we do that what we see here is the conductance is a function of these in-plane gate biases.
- For a thick wire of about five nanometres and a thin wire of about 1.7 nanometres.
- We can see that when we're above the quantum conductance the E-squared 1H, that basically
- the resistivity or the conductivity, the wire is very high. So even if we try and deplete the
- electrons from the wire, there's so many electrons in there that it stays a very nice metallic wire.
- If we make it very thin and we apply a voltage, we start to get rid of the electrons on the
- wire and we become aware of the electrons hopping between the dopants within the wire.
- If we now sit on these different points in the conductivity and we measure the change in the
- conductance as a function of time, we can see when we're above E-squared 1H. When it's nice
- and metallic we have a very low noise wire. When we start to deplete the electrons from the wire,
- we see the electrons hopping on and off the different dopants within the eye. We get
- these very strong RTS signals. From this we can measure what we call the power spectral density.
- To try and figure out how we can compare one device to another. If we normalise that with
- respect to the carrier density in the wire, we can get something called the phenomenological
- huge parameter. That allows us to tell which of the wires has low noise. The amazing thing
- about this is if we look at that huge parameter as a function of all these different materials,
- we find that these heavily doped wires have very, very low noise. So we've got these crystalline
- wires. Phosphorus is substituted for silicon very, very fast. We can get gigahertz frequencies down
- there. They conduct like a copper wire all the way down to where the atoms are in your crystal.
- Then noise is low. Now this is very important in the long-term if we want to make devices where
- we're controlling the quantum behaviour, we don't want noise to cause a loss of the quantum state.
- One of the other things that Richard Feynman talked about was what happens when we get to
- the small, well say circuits of seven atoms. They behave like nothing on a large scale
- and they satisfy the laws of quantum mechanics. So now we make devices here. So we basically
- have taken that central island of our single electron transistor. We've shrunk it right down.
- We can actually go in and look at the size of this and look at the silicon atoms that are exposed to
- these green regions of the silicon atoms are exposed. There's no hydrogen terminating them,
- and the red ones are just single dangling bonds. We can start to see that what happens is the
- chemistry happens in there. We phosphine dose it. The phosphine when it lands on
- the surface is pH 3 immediately likes to lose the hydrogen. So you get pH 2 + 2.
- These are the silicon dimer rows represented by bars. Then as it sits on the surface it
- likes to get rid of more and more hydrogen. It doesn't isn't likely to sit there with hydrogen
- on this. The next thing you see is it goes to pH and two hydrogens, and then eventually I'll
- lose even more hydrogen and it will go to what we call an Enbridge phosphorus atom.
- So when you dose with phosphine this natural chemical reaction occurs on the surface.
- That ultimately dictates the number of atoms that you get in there. So by changing the size
- of that lithography we can control the number of atoms. On this particular device we have seven
- different charge transitions. So you can see there are seven different peaks across here.
- Now what's amazing about this, it's very like the last one I showed you. You can see the size of
- these diamonds is changing because the device is so small. As you remove electrons from the system,
- you're actually changing the energy separation quite dramatically. In the last one I had 4000
- atoms. I changed one or two and the actual diamonds didn't change a lot. Here now
- these quantum effects are starting to become very strong as we remove the electrons in that
- energy level spacings going further and further. If we look at this, we also see these very sharp
- lines on the edges. This is now because on these devices, the actual wires, the source and drain
- leads are no longer two dimensional. They're one dimensional wires. So we get this density
- of states on the side that moves up and down the energy levels of the dopants. So this is really
- now nature at the level of seven atoms. Every time we make a device with a different number of atoms,
- we see a different signature, a different fingerprint of what the device looks like.
- What's fascinating for us is we've now made hundreds and hundreds of these devices,
- and we're building up a map of exactly how the world behaves as you move atoms around
- and you change the number and you change the number of electrons on those atoms.
- So then Richard, Richard Feynman said, well, what about you know, we can use not just circuits, but
- some system involving the quantised energy levels, or even the interactions of the quantised spins.
- So now he's starting to say not just the charge states, but let's look at the spins.
- So now over the last 20 years or so, the last ten years since we've developed the technology
- to make devices, we've literally made a whole series of devices. We've figured out how to
- reproducibly make them so that you can go in and make the same device twice, which is quite
- critical in the quantum world and non-trivial. Then we started to bring all these different
- components together, whether it's a single atom device we've started to see we can measure a
- single electron spin on those devices. We can get spin states to move between atoms. Then we've
- started to look at how you can actually image what the atoms look like beneath the surface.
- Then over the last five years or so, we've really tried to figure out can we integrate all of these
- to make a machine where we're inputting information, transmitting information in
- the machine, reading the information out to see if we can actually control the quantum
- world? There's lots of patents coming out of this. It's a very exciting field to be in.
- So what I'm going to do now is just go through four devices that I think epitomise some of the
- exciting stuff that we're doing. The first one is the single atom transistor. So it's going to be
- a little movie now that shows you how you get just a single phosphorus atom in the surface.
- So I think I showed you before that you need to start with the etch registration markers.
- This is what the chip looks like before it goes in the STM in ultrahigh vacuum heat up to high
- temperature reconstruct the surface. You bring down the hydrogen onto the surface that forms
- the atomic scale mask. You come along with your STM chip and you deliberately dissolve just six
- hydrogen atoms from the surface. You dose with the phosphine gas that brings a phosphorus atom
- in. That's going to be the active component and it only sticks with that silicon is exposed.
- We then do a controlled chemical reaction. We heat it up. What we find is a pH 2 likes to
- grab a hydrogen and come off allowing space for another pH 2 to dissociate to pH hit it again,
- another pH 2 comes off. Now if we heat it just a slight bit more, we find that the phosphorus atom
- goes into the top layer and it kicks out the silicon and the silicon stays above it so you
- know where it is in the surface. We now take this device and we encapsulate with silicon so that we
- basically form this nice protective environment where that phosphorus atom is sitting beneath.
- That's the full three dimensional encapsulation. Interestingly, not just do we put that atom in
- the middle, but we also open up these leads at the same time. These leads have very heavily
- phosphorus doped silicon. That allows us to bring those wires the current through to the actual atom
- itself. Once it's encapsulated, you take it out of the UHV and you make contacts to the buried
- device. So this is what it looks like. This is an STM image of the source and drain leads.
- This is the lithography stage. You can see the atomic step sites of the silicon surface.
- If you zoom in now you can see where the ejected silicon is sitting in the middle.
- I guess one of the amazing things about this is because it's an atom. You can actually look at
- the actual Coulomb potential that's created just by the presence of the atom in the surface. So
- just by putting the phosphorus there, that extra positive charge from the phosphorus creates a
- Coulomb potential that captures the electron. We can actually sit with that phosphorus atom
- with, we can ionise it, get rid of the electron, we can put one electron there or we can even start
- to load two electrons. So now if we look at the conductance through this device you can
- see here's the current going through as a function of these in-plane gate biases. We now have one,
- two, three peaks where we have zero, one or two electrons on the system. If again we put it as a
- function of the source drain bias, we can actually measure what we call the charging energy. From the
- energy we know that we get a distinct energy that tells you that it's a phosphorus atom.
- If you put a different atom down there, it would have a completely different energy.
- So that's really exciting. That really tells us that we've put down phosphorus atom. We know that
- that's the process. We've seen it. We can measure it electrically. Then one of the interesting
- things is we can also now start to image what the wave function looks like. So if I ask people
- what is the wave function of a phosphorus, extra electron spin look like a lot of people think it's
- a cloud, a fuzzy ball. In reality when you image it with the STM, you get this beautiful picture
- of exactly what the wave function is and it interacts with the atoms in the surface directly.
- How do we do that? We use the scanning tunnelling microscope in a slightly different mode.
- So normally we use it in a topographic mode where I've described it. Constant current
- measuring over the surface. We also can have a conducting backplane now so that we can get
- electrons to hop from the back plane through to the atom and then through to the surface.
- So now if we look it's on the side. If we now look at the energy level diagram of this. This is the
- barrier, the vacuum barrier which sits here. We've got a voltage on our tip. Then we've basically
- got this conducting back plane here so we can get electrons to hop onto the atom and then off again.
- By changing this bias now we can actually probe the energy levels of the atoms.
- This particular form of spectroscopy that we do inside the STM. So we change the bias here.
- When we reach the density of states we get a peak in the current.
- This is where the phosphorus atom sits beneath the surface. If we now park
- the STM tip at a slightly different bias just above that here we can go back and we can now,
- at the fixed height actually image what the wave function looks like beneath the surface.
- Now why is that interesting for us? That's fascinating because the device is so small
- now that we can both image it directly. And then from the images we can actually take
- classical computers that put all the atoms in and do atomistic modelling that tells us what that
- atom should look like in a particular position in the crystal structure. Here is the measurement,
- and here's the theory. Basically you can see these green and blue regions correspond to the actual
- states of the silicon atoms and the surface. So we can actually tell exactly where that phosphorus
- atom is. If you move the atom just one or two lattice spacings you get a completely different
- image. So using this technique we can actually pinpoint exactly where the atom is in the surface.
- The reason why that's exciting is you can mathematically model it. Then if you bring a
- second atom in you can actually see how the two wave functions overlap. So you can start to see
- how the interactions between the electrons of the atoms in the surface change. If you want to
- build a machine where you're controlling where the electrons go in the system, this kind of precision
- allows you to measure it directly. Then predictively model how you'd make the next device.
- What's the next one I want to talk about is now a gate. I want to put information in
- and read information off. So I'm going to use the same kind of structures. I have my single
- electron transistor here, which I'm going to use as a sensor. It's now going to sense what's
- happening to the electrons in the device. I've got two different quantum dots here with single
- electrons on those quantum dots. By changing the number of phosphorus atoms in those quantum dots,
- I can change the number of electrons I have in the system. So I've got this whole world to play with.
- What I want to do now is measure the current through here, align the energy level of my island
- with that of the quantum dot, and see if I can measure the spin states of the electrons on that
- quantum dot. What we typically do is we measure the current through the device here as a function
- of these two gate biases. We get these lines at 45 degrees, and those are the single electron
- transistor peaks that I showed you before. There are points where the energy level of
- the dot aligns that with the island. You get a break in the conductance here. It's around
- those points that we start to tell if an electron moves from this dot over to here or over to here,
- we get a change in conductance. It's at that point we can measure the spin state.
- Here's the energy level diagrams now of the island of the SET. The energy for our left and right dot.
- If I now put it in a magnetic field I spin split the states. So the spin up is at a different
- energy level to the spin down. By changing these biases either side, I can move the spin split
- states with respect to this island. If I'm on a spin up state, an electron will move across
- and another electron will move down and I will get this kind of dip in the signal through the SET.
- Now the thing I'm not talking about here, because there's so many things to tell is we're attaching
- here an RF tank circuit and that allows us to apply an AC signal so that we can measure this
- incredibly quickly. So here's my left dot here. If I've got a spin down state I'll have no signal. If
- I have a spin up state I'll get a signal. So what we can do with the same transistor island here I
- can measure both the spin states and either these dots, even though there are only 13 nanometres
- apart, they're incredibly close. So using the same sensor I can come along and I get no signal
- at all, I know that both spins are down, if I get a signal on the right that's down that basically
- shows that my left one is down and my right one is up. In this situation, I've got a signal here.
- No signal here. So it's up down, and here I've got basically two signals. So it's an up up.
- So in this way now I'm starting to be able to measure and control individual spins by changing
- voltages on gates in a magnetic field to be able to both sense and initialise the spins on both
- quantum dots or qubits. Once we get to this state, we now want to do a kind of operation between the
- two. So all computers are a basis of having what we call single gate operations and two qubit gate
- operations. So in this particular device, what I want to do is I want to load a spin up on one of
- my atoms. I want to load a spin down. I then use my gate biases either side to bring them together,
- and when they're close enough together they entangle. They start to oscillate
- and if I change the time, I can then pull them apart and read them out again. So this is quite
- amazing. Load electron on one, load electron on the other, bring them together, they oscillate.
- They're now entangled in the quantum world, and then I can bring them apart and measure them.
- If I do that, I can see that when they're in this spin up down state, you can see it
- oscillates out of phase with the spin down up state. After about literally 0.8 nanoseconds,
- I can swap the electron up to down on the two different atoms, quite literally 0.8 nanoseconds.
- So this is a way now to operate a very fast gate inside this system. It's hugely exciting.
- So with single and two qubit gates you can do any kind of computation in the quantum world.
- So now I want to see what happens if I put lots of these dots together. This is what we call an
- analogue quantum simulator. One of the things that Richard Feynman said was, 'If you really want to
- understand how nature works, you better build it at the same length scale that nature is made of.'
- So one of the questions we've always had is can you actually build a machine where you
- encode the Hamiltonian of the system and try and solve a problem that you can't do classically?
- So with phosphorus atoms in silicon is a very busy slide. One of the great things
- about having these atoms is that you can bring them very, very close together.
- Because the natural energy well that forms you have what we call a very large on site energy.
- So it's this ratio basically of the tunnel coupling to the temperature that allows you
- to form these very strongly coherent states. It comes because you can bring these atoms
- very close together compared to other systems. So you can see the distance here
- gets bigger each time. In superconducting qubits is quite big. It's about five to ten microns.
- So the energy scales are very small. By having this value up to about 1000 it means
- you form this very strongly coherent regime and it allows you to start to look at how electrons
- interact in matter. So this device is a very recent result. We basically made a ten dot system.
- By changing the relative distances between the dots we can create either this system
- which is like a series of diamonds. It's like a polyacetylene molecule. Whereas on the left
- hand side we can actually move the atom at the beginning and end of the chain further away.
- Now this basically is trying to mimic how a polyacetylene molecule works.
- We've got source and drain leads either side. So in this case if I put an electron onto the system
- it will come through and hop off the other side, and what I'll see because I've got ten different
- dots is I'll see ten different current peaks as I get electrons to travel through that chain.
- However, on this side, if I move these atoms further apart on either side, what I find is that
- I shouldn't expect to get ten peaks as I expect up here, I should expect only to get a couple.
- So what's happening now is as electrons load onto the system, the first four electrons
- will be localised in the central region here. As I load the fifth electron it will exist on both
- sides of the chain at the same time. If this is in the coherent regime. As I load the sixth electron,
- they will localise either side and as I load the seventh electron again they will switch
- either sides. So I'll get another current peak. So by changing just the distances in the device
- I can actually now get what we call a topological state. This is the first time that we've seen this
- kind of interacting predictive state that shows you've got coherence across this whole chain.
- is predicted for a long time. Never been able to see because you have to have three things
- you need to have very high on site energies. You've got to be able to precision engineers.
- You can see what we're changing here is distances between about seven to ten nanometres down to nine
- to 7.8 nanometres. So sub-nanometre precision in the two different devices.
- The other thing is you need to have a minimal number of gates that you can
- control the energy level of each of the dots, but also all of the dots together.
- So just to give you a sense of that, if you were to just use individual atoms,
- they'd have to be absolutely identical through the chain. If you had one atom moving out of position,
- it would be very difficult to create that chain. If you make the dots very big then basically you
- create a wire so all the dots start to overlap. So you've got to find the right size of dots
- where you can systematically make it so the energy levels are identical.
- You need to have that sub-nanometre precision so you can bring them slightly
- closer together, as you can move those atoms at either end of the chain further away. Then also
- you want to have a very low gate density. So typically you want to tune every dot
- individually, but then move them up and down together. If you have too many gates, it's very
- hard to do that without back action. So this is a very unique regime. The exciting thing about this
- quantum simulator is we can now start to simulate much more complicated materials.
- One of the things we're very excited about is looking at high temperature superconductivity.
- So in the bottom right hand side now I'm looking to the future. What do we do with
- this technology. We can do lots of little devices, but fundamentally we want to build a
- quantum processor. It's well known in the quantum world. If you want to create a quantum computer,
- you've got to be able to control the individual spin states of the atoms, but you've also got to
- error correct. That means you need ancilla atoms around that you when you can't actually measure
- your qubit directly because you'll collapse the state. You can probe the ancilla states to tell
- what's happening to the errors on qubits. This is something called the 2D surface code. So in
- the long-term you need to make a two-dimensional grid of atoms in order to build an error corrected
- processor. That's hugely challenged now to bring all your leads into that device.
- However, with this technique you can go to 3D so you can pattern on one plane so you can make
- wires in one direction in one plane, grow some silicon. You can make some islands and atoms on
- the second plane where you can individually read out the individual electron spins. You can grow
- more silicon and then you can pattern wires on the second, and the third plane that are
- perpendicular to those in the first plane. In that central plane, obviously you want to have lots of
- atoms. You can get bigger and bigger processes. So what we did here was we actually patterned
- 1024 atoms in the surface to show that with this technique, this all fits in to a very, very small
- device. So it will still fit in a dilution fridge. So it's very exciting that we can do this.
- Then we started to see can we actually measure the spin state in a fully three-dimensional
- sub-nanometre precision device. So here we have a single electron transistor with a
- single atom in there. We've grown silicon above it and we've got a wire that sits above that. So
- we can actually control the spin states with the wire. The key thing we have to do here is each
- layer that we grow on, we want to make sure it's nice and smooth so that when we pattern on here
- with our lithography, you can actually see the wire there where the hydrogen is being removed
- or not. So it's a huge technical challenge to make those surfaces flat and be able to create
- wires in multiple planes. If we do that, we actually find that we can use these wires,
- whether they're the in-plane gates, the red ones or the surface gates to actually both measure
- the spin state and to measure the conductance through the device. It's fascinating for me
- that these devices are actually a lot less noisy than the other devices. So by putting a metal,
- essentially a metal plane above where the dopants are, we get rid of the noise that comes from
- surface states even though they are far away. Also at the moment what we're doing is basically
- building up these integrated circuits. So the first concept in 1998, putting single atoms down,
- wires interconnects, making two qubit gates and now this ten dot quantum simulator.
- The key thing that I've learned through that whole process is in the quantum world, it really makes
- a difference if you can see what you're making. If you can fabricate it with atomic precision.
- Then if you can actually control the signals that come down to control the spin states on the atoms.
- So that ability to both see, build, control at the atomic scale is crucial to build these new
- machines. What we've learned from these systems is that they are very long coherence times, very high
- control fidelity, is incredibly fast. We know that they're low noise because they're crystalline.
- We have very few gates in there. The actual presence of the phosphorus atoms creates the gates
- and there's only two atoms. That's something that I'll come back to in a second. That's truly
- amazing that in these devices there are only two atoms in the active device. We're able to
- make the whole device with atomic precision. Even in 3D. What that means is by having these very low
- noise devices where there's only crystalline qubits around them, it means that we'll need
- a lot fewer of them to do error correction. So I just want to go back to that. We're literally
- in the whole active region of the device. We literally take our silicon and we replace it with
- phosphorus and whether phosphorus is the active element or the control wires from the device.
- They sit next to each other in the periodic table, so there's very little strain in the device.
- By just putting the atoms there we create these little potentials where the electrons sit.
- So it's a very strongly quantised system. It's perfect each time we do it. It's relatively simple
- to model because it's a small system and it has a very low gate density. By encapsulating in silicon
- we've got this ability to manufacture devices in all three dimensions with no materials interfaces.
- So where are we going with this technology? We can go back and we can look at the computing
- industry. We know the first transistor in 1947, 1958 the integrated circuit, the first kind of
- commercial use they had was 1964, and the first PC in 1981. What this does for us is it gives a
- little bit of a roadmap. So unlike when they first made computers where they didn't know where it was
- going, we can see that the first single atom transistor. 2012 an integrated circuit 2023,
- first commercial product I reckon would be about 2028, and then an error corrected
- system could be as close as 2033. So it's very exciting to bring all of those things together.
- If you now start to look at the future, what can it do? A lot of people don't
- know where this kind of ability to control the quantum world is going to take them,
- but we do know that simulating materials is one of the first things that we're already doing.
- So if we can understand 2D materials, how do we make high temperature superconductors,
- how we can look at artificial photosynthesis. These are the kind of areas that are interesting.
- In the chemical world for me, people are talking about looking at designing drugs.
- Obviously drug design is a key area where if you can control quantum states and you use quantum
- computing to actually solve those problems, it would be quite dramatic. One of the interesting
- things for me looking at that field is that obviously it takes a long time to solve a drug
- design problem at the moment. It takes about ten years and it's hundreds of millions of dollars.
- One of the things I think is amazing about that is we will do that with classical computers.
- That limits that we have. Actually to build a quantum computer is going to be
- less than that I think, in the long term. So it's quite interesting to compare them.
- In the financial world. A lot of people are interested in using quantum states to solve
- optimisation problems. There's huge numbers of problems that they have in finance and likewise
- in transport. So going back to how do we optimise the grid, how do we get different
- buses and trains and everything else to work together to minimise fuel costs?
- Who knows where the quantum world is going to take us? There are certainly a lot
- of problems out there that the classical computers simply can't solve as we stand.
- So with the end of the lecture, what I'd like to leave you with is just a few thoughts.
- If we go back to 460 BC, Democritus basically said, imagine if you took the world,
- take a slab of any material and you chop it up and you chop it up and you chop it up,
- and you keep going. Eventually you'll get down to some small, indivisible component which he
- thought at the time were like atoms. He was right all that time ago he envisaged a world like that.
- With Henry Baker, he basically said, 'If you could take microscopes and look at the world,
- you'd understand how nature forms.' He knew that microscopes had that power. I don't think he
- would have known, though, that you could take an imaging tool and turn it into a fabrication tool,
- which is essentially what we've done. But he was essentially right. Understanding the
- world is the ability to see it at smaller and smaller length scales.
- Richard Feynman, likewise, Richard Feynman's whole lecture was about if you make things
- smaller, you'll basically be able to create new machines. What's fascinating for me now is 2014
- we can actually make devices with a functional element is an individual atom. We're starting
- now to make machines that never existed before that are made of individual atoms. Perhaps
- what's even more fascinating about this is if you look at the number of atoms on average in a human
- is seven octillion, and where we've gone is from seven octillion
- down to one. I think that gives an indication of the power of the imagination. So thank you.
- M1:
- Thank you very much for that. We have a few minutes for questions. So let's first
- take questions from this audience and then ask if there are any remote questions. So,
- Jeremy, I think a microphone will come to you. Yes, good.
- M2: Of course. Incredibly beautiful. You talk about two atoms and you don't in your
- imagination here, tell us about actually you want to do many other sorts of atoms and put them in.
- So where do you see that on the roadmap. Do you actually see the key thing is actually
- to keep at two because that's where you have control or that actually starting to try and
- develop other ways to put other individual atoms in here is going to be the key.
- Michelle Simmons: So just to give me an understanding so different
- atoms you mean or more than two different atoms.
- M2: Different atoms.
- Michelle Simmons: Absolutely. So I think when we started out as a technique, there were probably
- about four or five different atom types that were compatible with this hydrogen resist. So whether
- it's aluminium. People have recently done boron. So you can start to make atomic size PN junction.
- So there are lots of atoms that are compatible with this technology and indeed lots of
- substrates. You can do it in Germanium. Some people are hoping to do it in graphene. So for me
- the interesting thing is you can do that if the chemistry works. The chemical world is
- quite fragile and you have to understand it quite deeply. So to build the devices that we've got,
- it's an ultrahigh vacuum system. So whatever chemistry you have has to work in ultrahigh
- vacuum with the tools that you have. So you have temperature basically that you can change
- things and then you're trying to force a chemical reaction under those conditions. So that's really
- for me I can see it's going to keep opening up. There will be lots of materials you can do it in
- and there'll be some that you can't, but you can make insulators, metals, different atom
- dopant types, different substrates. So it's yes, I think it's going to change quite dramatically
- over the next ten years or so. Huge numbers of groups jumping into this field at the moment.
- M1: Thank you.
- M3: Thank you, Michelle, a lovely talk. Can I just ask a
- hard-nosed engineering question? Do you have a concept of yield?
- Michelle Simmons: Yes, I do.
- M3: Can you say something about that.
- Michelle Simmons: So I'll say something because I know you're very keen on this, Mike. So I think
- one of the one of the things I found frustrating actually Chris, is in the audience in Cambridge
- was trying to make the same device twice. It drove me mad. So in the gallium arsenide system, we had
- a lot of control of all the different parameters, but we found that we couldn't make the same device
- twice no matter how we tried. One of the reasons that I got frustrated with that was because I felt
- you couldn't build something unless you could do that, with this deterministic doping. When you put
- the atoms in place, they will stay there. So then the challenge becomes how well can you control
- the process itself. If you have control of the process which we do your own clearing
- where nobody else is using the tools, then the yield is incredibly high. So that's for me very
- rewarding because we can make lots of devices and something like 90 per cent, 95 per cent of them
- will work. So contact problems are not an issue. One of the things we're learning though is that
- you can move atoms around and things change. So we joke every atom counts. You really have to
- know where the atom is and where you're putting it as to how the device is going to behave.
- By systematically making lots of devices as a function of time, we now understand
- how that works. So it's quite, the yield is high for the process that we've developed.
- Every atom counts. So you have to have that precision all the way down to the atomistic level.
- So exactly where the atoms are going. I think once you get that then these devices are very stable.
- So it's quite fascinating. I haven't seen anything quite like this
- in terms of yield and obviously the semiconductor industry at a higher
- length scale. At this quantum length scale I haven't seen anything like it.
- M1: Yes. Thank you. Just a moment. You need a microphone I think. Yes.
- F1: Thank you for an exhilarating lecture. As a
- scientist, but not in this field. Following on from that question,
- could you give a sense of how long it takes to make one of these precious things?
- Michelle Simmons: So that's you know, this is another it's one of the things that was
- fascinating to me when we started this whole field. People would ask us single writing
- with a tip. So it takes time to do that. You can't stamp out millions so you know who's
- ever going to be interested in it. My argument to that is if you want to control the world at
- this link scale and control quantum states for a computer, then you have to have that precision.
- So we've looked at how can you make the devices as fast as possible. So in our labs you will
- literally design a device at the beginning of the week you will put it into the UHV system. You have
- to put the markers in, it takes half a day, into the UHV system, maybe a day or two to get the
- process working in there. Back to the clean room and by the end of the week you've got a device,
- and so it's quite fast. Again, if you have to put that aside how it takes other people.
- How long does it take. If you send a device through a fabrication line
- it takes an awful long time. It can take months to get that back.
- So for me again it's quite exciting because in this field where you're trying to build processes,
- the faster your cycle time to understand how the devices work, the more you understand the
- way they're working by making lots and lots of them and iterating that process,
- I think the faster you'll make something useful. Yes, it's quite it's a very different way. I've
- spent a lot of time working with different groups as to how they make devices and depending on how
- you make it, you come with a very different mindset but single right precision fast cycle
- time. Then you've got to build that understanding of where all the atoms are and how they work.
- M1: There was wrong at the back there.
- M4: Yes. Thanks Michelle for a fantastic talk. I have a follow up to Mike's question. I mean, is
- the variability that you naturally get from the kind of doping getting incorporated in
- different sites and maybe the environment that there are different environments that
- that doping experience is, is that something you can utilise or is this something you really have
- to get down to an absolute control in order to make a large, large array of quantum bits?
- Michelle Simmons: Yes, it's a great question because I think we've always had the
- very hard core route that we have to control everything. So we've got lots of techniques
- now where we can do deterministic doping placement. Along the line, along the way,
- we've obviously had devices where atoms haven't been where we wanted and then
- we've understood actually how you can utilise them. So by you know, I think in the original…
- Bruce Cain had an architecture to build a quantum computer. He had literally individual atoms that
- he was encoding information on. The challenge there is if you try and address one atom,
- they have the same energy as the next atom. So you end up hitting multiple atoms at the same time
- and you create errors in that way. So if you now have atom qubits with different number of atoms,
- you can actually address them at completely different frequencies. If you have different
- number of atoms you get different tunability. So there are actually huge advantages from
- engineering every aspect of the qubit and the control and everything else.
- So it's actually there's a lot to play with and actually having imperfections if you actually
- can create the imperfections, you understand how to get them. They can actually help you along.
- You have to I think fundamentally you need that precision control.
- M5:
- Lovely talk Michelle. Really excited to see the 1000 qubit vision. My question to you is
- whether the coherence time, the T2 time starts degrading as you increase the number of atoms.
- Do you think at the 1000 atom scale, will it remain realistically long?
- Michelle Simmons: Yes. Look, it's a great question and the honest answer is until we do it,
- I won't know. It's well known that in any system, as you add more qubits, the T2 time drops down. I
- think with our system I'm hoping that because we start with higher quality that the drop will be
- not so bad. I guess the fundamental question is if you understand what limits your coherence times,
- you can actually work around it. I think again with the system that we've got, we've actually
- looked at all the different parameters of how you engineer the atoms with respect to electromagnetic
- fields, and you can stop a lot of the decoherence mechanisms with the atomistic control.
- So you know where the electrons, where the spin is losing its coherence and
- where it's losing its spin state because of the actual environment of the qubit itself.
- If you know how to engineer that environment, you know how to mitigate a lot of that. So
- that's again something if I didn't know what I was doing and I just was constantly tuning, I would be
- very nervous. So having the ability to see it, we've got the ability to engineer around it.
- M1: Okay. Any questions here? Just a minute. Here it comes.
- M6: It's a wonderful story, Michelle, that you've told us a wonderful narrative.
- As you went through it, you identified what were the next problems and how you systematically
- tackled them. In your ITRS as it were, for getting to 1000 qubits. There are a lot of things that you
- already know how to do, and you just have to do them better and harder. What are the things that
- that are the real challenges that are going to be really hard? New problems getting to 1000 qubits.
- Michelle Simmons: Yes. So I think for us because our devices are so small, having the classical
- control chip connecting to the quantum chip is going to be challenging. So we at the moment have
- chips that are down at the same temperature that are closely bonded to our chips. But obviously
- in the long-term, to get that even better, you would want to bring all that control
- onto the chip directly. So I think when we first started, a lot of people thought that
- as you make things smaller and smaller, you're never going to be able to get connections in.
- Because of the STM technique you can make those connections very small. Then you can fan them out.
- By having a three-dimensional control, you can make that fan out easier.
- Fundamentally, whichever qubit you've got, you've got to have some kind of classical control
- that works well. I think that's still in development globally for semiconductors.
- So I'm very, very keen to keep an eye on that as to how we connect to our chip.
- So the quantum chip I'm not so worried about it's going to take time for us to figure out all the
- atoms exactly where they go. The classical trial would be the thing that I'm most worried about.
- M1: Thank you. Oh, there's a question remotely. Okay.
- M7: So and Ashley Naismith, an incredible research and inspiring lecture. Thank you.
- I'm interested in the future applications. Will the devices be durable enough to
- withstand the rigours of use. Do we have the technology to repair quantum devices.
- Michelle Simmons: Yes. So I pretty much have this ever growing belief having the experience that
- I've got that you need this precision to make a fully functional device. Those devices in the
- first instance we use them in dilution fridges which are in controlled environment. I think
- that's going to be the way for a long time. So there's no way that everyone's going to have one.
- They will be controlled environment with some kind of internet access to be able to use them.
- I do think it's coming. It really feels to me that the more we make devices,
- the more we're learning and the more power we're seeing coming
- out of them. So I think you will get them, but they'll be remote access.
- M1: Good. We have time for one more question. Brief question. Here we are.
- M8: Yes.
- Hi, Michelle. Great talk. I was interested in your simulator actually, because
- that seems to be a sort of parallel way of solving quantum problems.
- I was wondering how you map your simulator onto the quantum problem that you're interested on?
- Michelle Simmons: In the long-term you mean?
- M8: Well even the short-term?
- Michelle Simmons: Yes. So at the moment we literally will design the system to
- match the system that we're working on. So this is the polyacetylene molecule. You're
- literally creating these dimer structures in the surface. Is that what you mean Charles or not.
- M8: Yes, yes. So you try to create the same wave function that you're
- trying to solve that into silicon and say yes.
- M1: Maybe you should repeat the repeat the question. Use a microphone.
- M8: Are you scaling the quantum problem or are you taking trying to map it exactly?
- Michelle Simmons: So at the moment we're trying to map it, exactly.
- But, I guess we've already looked at going to much larger systems and I think that's what's
- exciting is that that was a really, like a demo model for us, a very simple system.
- You can actually go to much larger systems and start to look at a lot of 2D materials.
- M8: Very interesting.
- M1: Well thank you everybody. I have one more duty, which is to congratulate you on
- this Bakerian lecture and present you with the Bakerian medal.
- Michelle Simmons: Thank you very much. M1: Thank you very much.
Join Professor Michelle Simmons to find out how scientists are delivering Richard Feynman’s dream of designing materials at the atomic limit for quantum machines.
Sixty years ago, the great American physicist Richard Feynman delivered a famous lecture in which he urged experimentalists to push for the creation of new materials with features designed at the atomic limit. He called this the "final question": whether ultimately "we can arrange the atoms the way we want: the very atoms all the way down!"
Professor Simmons will explain how to manufacture materials and devices whose properties are determined by the placement of individual atoms, and will highlight the creative explosion in new devices that has followed and the many new insights into the quantum world that this revolution has made possible.
About the Royal Society
91TV is a Fellowship of many of the world's most eminent scientists and is the oldest scientific academy in continuous existence.
/
Subscribe to our YouTube channel for exciting science videos and live events.
Find us on:
Bluesky:
Facebook:
Instagram:
LinkedIn:
TikTok: