If you follow quantum computing news for long enough, it can start to seem like any quantum system that can alternate between two well-separated energy states can be used as a qubit. Atoms, ions, photons, electrons, and manufactured devices all have their backers. One of the key things that attracts backers is the tech’s ability to scale. We’ll need a lot of high-quality qubits to start doing any complex computations, and the ability of any technology to get us there is the subject of debate.
So even as some technologies can now support thousands of qubits, some competitors are still working on a small handful of qubits—the companies behind them are convinced that they have the potential to scale more effectively.
One of those technologies involves quantum dots that hold a single electron. Their advantage is that we can manufacture them using the same tech we use to build traditional processors, an approach that has proven to be scalable. This week saw two new papers describing different ways of using quantum dots, one of which was appealing enough that IBM bought the company that developed it. Separately, another company has released a processor showing 100 individual electrons being held in diamond defects, technology that wasn’t obvious could scale.
Flipping spins
The first news comes from a publication in Nature from a company called HRL Laboratories, the descendant of the research program launched by Howard Hughes. HRL is using manufactured quantum dots, a distinct technology from the dots used in displays, although it relies on some of the same physics principles. Both approaches use a structure that traps electrons in a region smaller than the electron’s wavelength. In displays, this allows control of the wavelength that the material emits.
For qubits, the traps are far smaller and serve to hold a single electron in place. Critically, they can be manufactured; with the right wiring, electromagnetic interactions can trap a single electron in a small patch of silicon. Once trapped, the electron’s spin, which can be up, down, or a superposition of the two, can be used as a qubit. While this technology can scale easily—we’re very good at putting wiring into silicon at scale—electron spins are hard to keep stable and are typically controlled via microwaves, requiring a separate control system.
But “typically” doesn’t mean “always,” and HRL is describing a different tech. It requires three separate quantum dots, each holding an electron, with the surrounding electronics controlling how much the spins of these three electrons can interact. That’s critical because, under certain conditions, no two electrons can have the same spin. This explains why atomic orbitals fill up the way they do, with each energy level holding just a pair, one spin-up, the other spin-down. Enabling them to interact can alter their spins.
To do operations on this kind of qubit, you simply need to control which electrons are interacting and to what extent. That is controlled electronically, allowing us to eliminate microwaves entirely. Everything is handled via wiring, eliminating the need for lots of microwave-carrying cabling into the refrigeration system that keeps the hardware near absolute zero.
HRL spends much of the paper describing its control system, which sits at an intermediate level of refrigeration and consists of a traditional processor optimized for low-temperature and low-power operations, consuming less than 3.5 watts despite being manufactured on a 130 nm process. Instructions for operating the qubits are compiled elsewhere, then loaded into the controller, after which it operates autonomously. Communications with the chip that holds the qubits are handled by a superconducting ribbon cable.
The system HRL describes had 18 qubits, and the company ran a simple error-correction code on it, demonstrating a logical error rate of less than 1 percent. That’s well below what has been achieved with other technology, but it’s an important validation that the problems HRL is facing are likely to be in the realm of engineering rather than physics.
Hybrid machines
Although the paper was published by HRL, the progress it has made has since become the property of IBM. That’s notable because IBM has bet big on competing technology in which qubits are held in a manufactured device called a transmon and controlled with microwave pulses—precisely the control system that HRL was trying to avoid. It’s not an obvious fit at first, so we asked IBM Director of Research Jay Gambetta about the decision.
“At a high level, I’m a strong believer in silicon technology—superconducting qubits are built on silicon, spins are built on silicon,” Gambetta told Ars. “It gives us flexibility to keep both parts going in parallel, build on the same foundation.” Right now, that foundation allows rapid iteration and testing of designs. IBM has typically released a new processor architecture every year, refining it over two or three generations before the next one arrives.
But Gambetta suggested that in the longer term, this raises the possibility that HRL tech will appear in an IBM system, operating alongside IBM’s existing tech. “As we look at the error correction codes going forward, some codes are better at being memories, some are better at being magic state creation,” Gambetta said. “I want to have the flexibility to build my hardware that maximizes both of them.”
Other spin tech
The same issue of Nature includes a separate paper on quantum-dot qubits from a group at Delft University of Technology. We covered their technology in May, but in short, it involves qubits based on single electrons in quantum dots and focuses on moving electrons and their spins around.
One problem with quantum dots and other manufactured qubits is that their connections are dictated by the wiring of the chip they’re on. Since error correction codes are based on the geometry of the connections among qubits, this means committing to one or a small number of error correction codes when the chip is manufactured. If better code is developed or if some operations are easier to perform with a different code, you’re out of luck until the next generation of chips.
Qubits based on trapped ions or neutral atoms don’t have this limitation. The qubits can be moved around so that any one can hypothetically be connected to any other. This provides a great deal of algorithmic flexibility and efficiency; their backers expect it’s enough to overcome the relatively slow operations compared to silicon-based qubits.
In May, however, the Delft team showed that an individual spin could be moved from one quantum dot to a neighboring, unoccupied one and that this approach could scale—the spin could be moved through several intervening dots and end up on the opposite side of the chip from where it started. In theory, this means any two spins could be moved around arbitrarily, brought into proximity, and entangled. The approach combines the advantage of manufacturing with the flexibility of atom-based systems.
Wednesday’s paper expands on that, creating a bus that electrons can be moved along, with a number of bus stops where they can be stored when not needed. We need a lot more hardware to get this to work, but that hardware is very compact, and the flexibility can be powerful.
The team describes a 1.2 micrometer bus route and says that the fidelity of sending a spin from one end to another and back is nearly 98 percent. Using a handful of spins, the researchers implement a small error-correction code. They also characterize the sources of error in their system, finding that most of it stems from two sources: noise that arises during the spin exchanges needed for entanglement and the loss of coherence while the qubits are being moved or idling, a common problem with spin qubits.
Diamonds are a qubit’s best friend?
A separate technology went pretty quiet after we looked at it a decade ago: nitrogen vacancies in diamonds. A diamond is a regular array of carbon atoms, all sharing bonds with their neighbors. But impurities can be scattered throughout the array, including the incorporation of nitrogen. Nitrogen can form one less bond than carbon, so it leaves a neighboring carbon with a single unbonded electron. The spin of that electron can be manipulated just like the spin of one held in a quantum dot.
The problem has been that it’s very difficult to control where nitrogen vacancies end up within the diamond. While it’s relatively easy to run wires to them after they’re identified, every chip made to work with nitrogen vacancies is bespoke; if you use a standardized chip layout, there’s no guarantee that there will be one—and only one—nitrogen vacancy where the wiring expects it to be. Saxon Q, a spinout of the Universität Leipzig, has apparently developed a technique that can put nitrogen vacancies within a 10-nanometer radius.
The company’s hardware is striking in several ways, most notably because it runs at room temperature. The machines it is building fit in a standard rack mount and plug right into a normal power supply. The other notable feature is its modular design; the company builds a “core” with eight qubits and can integrate multiple cores in a single rack. It announced last week that it is selling units with over 100 qubits total, which it expects to begin shipping within a few months. Larger systems are expected next year.
Both will be too small to do interesting error-corrected computations. But Saxon Q says its individual hardware qubits have a gate fidelity of over 99.9 percent, making the technology competitive with others on the market. So once again, we return to where we started: The key question remains whether the technology can scale.
