The qubit is the basic unit of quantum computation. So quantum hardware developers obviously like to boast about how many they have. Though some claim to have thousands in their devices, there’s a very real sense in which no one has built even a single one yet.
There are a couple of different things that we use the qubit name for. One is a physical qubit. The physical part refers to the fact that these are actual, real-life objects. The bit part tells us that these objects should have two possible states. And qu is for quantum, since we need to manipulate the states in a quantum mechanical way.
Any qubit that deserves the name should also have extremely low noise. The way we manipulate them and interact them should be almost perfect. As an achievement of experimental physics, they must be at the pinnacle: A marvel of science and engineering.
Even so, they are not good enough. For quantum computers, almost perfect is almost useless.
This is no more than we expect of normal computers. There are millions of pixels on your screen, but you’d notice if just one was doing some random. The same is true for all the millions of bits swimming around in your programs. It only takes a few switching value because they’re bored for everything to turn to nonsense.
When we program, we often forget that the bits in our computer have an actual corporeal form. We think of them as an abstract concept, pure and incorruptible. Software development would be a very different activity otherwise.
Quantum programs are designed with the same degree of perfection in mind. To run them, we need logical qubits: Incarnations of the idea of quantum information itself.
Building logical qubits requires us to tame the wildness of their physical cousins. We need quantum error correction. Many physical qubits are herded together and tricked into becoming greater than the sum of their parts. The more physical qubits we use, the better the effect. Noise decreases exponentially, until we can be sure than not a single error will occur during a computation.
This is not without its cost. We must think nothing of spending a few hundred physical qubits to build a single logical one. But if it means achieving the full promise of quantum computing, it will be worth it.
The most popular design for quantum error correction is the surface code. For the very smallest surface code, 17 physical qubits are needed. These would build one logical qubit, but not with enough complexity to actually do anything with it.
Nothing like this has yet been achieved. To see why, let’s take a look at what would be needed.
This is a surface code. The 17 dots, both black and white, are the physical qubits. The 24 coloured lines represent a certain kind of quantum operation, the controlled-NOT. For each pair of connected qubits, this operation should be possible to do cleanly and directly.
Wiring up all these controlled-NOTS is the main challenge. Just having 17 qubits in our quantum processor is not enough. We also need the instruction set to support this specific web of processes.
Having a bunch of physical qubits on a line is old news, Two lines next to each other is also doable. But the 2D lattice of connections needed for the surface code is much more tricky.
It’s so tricky that it could change the landscape of quantum computing. IBM and Google are arguably in the lead at the moment, with their methods of building qubits from superconducting circuits. But other approaches are not far behind. Qubits based on trapped ions have already been used to do some impressive things based on quantum error correction (see here).
Even so, Google is promising this and much more by the end of the year. They promise a 7x7 lattice of 49 physical qubits.
This would be a huge step forward in comparison to other devices, like IBMs 2x8 lattice of 16 physical qubits.
The IBM device has enough connectivity to make a logical bit from physical qubits. It’ll do a lot more cool stuff in the coming months, as you’d expect from the device at the forefront of its field. But making a logical qubit will not be among its achievements.
The fact that Google’s 49 qubits will be so revolutionary makes it hard to believe we’ll see it before the year is out. More realistic landmarks for this year are a 17 qubit device from IBM, and a 20 qubit one from Google. Both have enough qubits to get started with the surface code. But do they have the right layout? Only time will tell.
We might not have to wait for long. John Martinis, the guy in charge of building Google’s quantum devices, is giving a talk next week. The title…
Measuring logical error scaling with the surface code
Surface codes are definitely on the radar of the tech giants. The world’s first logical qubit is coming. Has Google’s 20 qubit device already managed it?
Update: Betteridge’s law of headlines states that “Any headline that ends in a question mark can be answered by the word no.” The law has not failed us this time. Google’s 22 qubit device will just be two rows of 11: not well connected enough for a surface code. They are certainly aiming towards building a logical qubit, but so are many others. It could be that the superconducting qubits of IBM and Google will be overtaken by trapped ion qubits before this milestone is reached. Attempts to reach future milestones could also bring other architectures, like spin qubits, to the fore. There are many exciting years ahead in the development of quantum computers.