Skip to 0 minutes and 3 secondsTell me about some of the architectures that you and your group have actually been working on. You’ve worked on, for example, a design of a photonic quantum computer, right? Yes. So we used photons because photons are very flexible in terms of space. That means you know if we have a solid state implementation, we cannot move qubits around easily. But photons naturally, well probably it is little bit difficult to keep it, but as I said we don’t need really memory concept. So if we can run error correction actively, it doesn’t matter if photon can be stationary or not. We created computation resource. We are continuously running error correction, plus computation, then we can get a result at the end.

Skip to 0 minutes and 53 secondsSo with a photonic architecture, you have some part of the system that’s generating individual photons, is that right? Yes. Okay. And then those photons are put into some sort of device. And inside of that device what happens? Inside the device, we like to two photons interact to make entanglement which is a necessary resource for quantum computation, but problem is that, you know, photon and photon – photons are very weak to interact. So photon do not want to interact with another photon. So what we need to do is we have some, some system to assist a photon-photon interaction. So that is a device you are talking about.

Skip to 1 minute and 37 secondsSo we send photons, massive array of photons which are just individual photons and into the device, then when that comes out, it is all entangled, as the computation resource. So that is the structure. And we need to know which timing we need to send a photon, in what order, and then we design what kind of entangled state will come out. So to do this, we have to sort of, you know, switch the photon which way to go and which photon to interact with another and at the end we have a lattice like entangled state. It is a massive state entanglement, the quantum correlation is throughout that entire sort of photon network.

Skip to 2 minutes and 26 secondsAnd then we measure them to proceed quantum computation and also error correction as well. Okay. So one of the other systems you have been involved in is the design of a large scale system using NV diamond, right? Yes, that is similar, in the sense we use photons. We use photons because we like to consider the quantum system as a network, rather than just a monolthic system. The reason is that you know we will have a problem if the system is large enough, qubit and qubit in a, data qubit and data qubit might be completely apart in the system and we have to spend a lot of resource to interact with them to proceed adequately, for example.

Skip to 3 minutes and 15 secondsSo we like to have flexible systems, then the photons are very good at it. So that is the reason we use the photons to connect NV diamonds. Now NV diamond is not moving around. NV diamond is stationary. Then we can connect NV diamond-NV diamond via photon. So with the NV diamond system you have a couple of small pieces of diamond and inside of that there is a place where you can store a single qubit, right? So then you are taking two of these and you are connecting them together, you are creating the entanglement between two different ones using light that comes out of those diamond chips. Is that correct?

Skip to 3 minutes and 55 secondsWell, there are several ways to entangle them together, but we have a device and we want to have, it’s called NV center which is a defect in the crystal. And put the defect into the cavity, optical cavity and that is a device, and we can sort of couple them via light. You can use the emission, but we tend to use reflections, conditional reflections create photon and electron spin inside and then we can sort of transfer that entanglement, consuming photons. Then at the end we have electron spin-electron spin entanglement. But don’t get excited it is diamond, but it is not the diamond you imagine. It is sort of very small and very cheap one.

Skip to 4 minutes and 49 secondsSo the big question, how big are these systems going to be? What does it take to actually build one of these large-scale systems? Well, well, okay. So large scale quantum computer has to be fault tolerant. There is no question about it at the moment. In this sense, of course, it is dependent on device fidelity. So how well we can create device. Say if our device and control are 100% accurate, of course we don’t need error correction, then we can have – we would don’t need to have many, right. But we cannot have such a system.

Skip to 5 minutes and 36 secondsSo we have to have some kind of error correction on top of it and once you, then not only applying error correction code, we have to make it fault tolerant. To do that, we need lots of resources and the overhead we estimated is 1,000 to 10,000. So, 1,000 to 10,000 times as many physical resources? Yes, physical resources. As we would like to have logical qubits inside our system. Is that? So, yeah, roughly speaking, if you want to control one qubit device, then we have to have 1,000 to 10,000 qubit devices. I see. So, for building a large-scale system, then how many devices does that wind up being?

Skip to 6 minutes and 30 secondsAre we are going to be building systems with 1,000 devices, 10,000 devices, a million, a billon devices? Over a million to one billion. Wow, okay. So, if each one of those components costs us a dollar or $10, is going to be an expensive machine to build, huh? That is an interesting question. You know, good thing about fault tolerance is you just need to reach. Of course, if it is very close to threshold, it is a little bit of a problem, but you need to reach the threshold error rate. And from there, well we estimate about one order of smaller error rate is enough.

Skip to 7 minutes and 12 secondsThat is only 99.9% what I said or really a classical computer scientist probably would say that it is really difficult to achieve. But you know with your computer on your lap, it is doing much better, right? In a classical computation, 99.9% device is rubbish. In that sense each one of them could be cheap compared with creating maybe 99.9999% error rate device. Then maybe we can make it smaller, but it could be much more expensive. Okay, I see. So, talking about the systems, we just talked about the scale of the system we might want to build. Does that actually depend on what we are going to use the computer for?

Skip to 8 minutes and 5 secondsYes, of course and also dependent on what kind of accuracy do you require. Okay. So, it depends on both the application and the error rate we want to have after we have applied all the error correction. Yes, Kae, thanks for being with us here today. This has been our interview with Professor Kae Nemoto of The National Institute of Informatics. Good luck with all of your work on designing quantum computers and good luck on building a large-scale system. Thank you very much.

Large Scale Quantum Computer Systems

Professor Kae Nemoto returns to describe two of the large-scale quantum computer systems she has helped design. One uses photonic chips to continuously create entangled states consisting of many photons. The other uses NV (nitrogen vacancy) diamond and similarly creates a rolling set of entangled qubits. Both of these architectures are suitable for use with one form of quantum error correction, known as a surface code, which we will introduce at the end of the next Activity.

Materials introduced in this video provided by courtesy of Prof. Kae Nemoto and National Institute of Informatic TOKYO, JAPAN.
-Photonic Architecture for Scalable Quantum Information Processing in Diamond
-The Photonic Chip Topological Cluster State Computer

Share this video:

This video is from the free online course:

Understanding Quantum Computers

Keio University