The race to build practical quantum computers may be entering a new phase. Some of the prior art are currently facing size constraints, while others are rapidly falling behind.
Over the years, two main approaches have allowed physicists to make progress by packing devices with qubits in amounts equivalent to computer memory bits. One of these methods encodes qubits as currents flowing in superconducting loops. The other uses excited states of individual ions confined in vacuum by an electromagnetic field.
But in the past two years, qubits composed of single neutral atoms (rather than ions) and held in “tweezers” made of laser light have suddenly come into competition. And other technologies in even earlier stages of development may still catch up.
Theoretical physicist Barbara Terhal of QuTech, a quantum research institute at Delft University of Technology in the Netherlands, said: “But this is no guarantee that these platforms will continue to lead.”
Exploring qubits
Quantum computers have the potential to solve problems that conventional computers cannot solve by using phenomena such as quantum superposition. In quantum superposition, an object can exist in two simultaneous states (e.g. rotating both clockwise and counterclockwise). Physicists call such states qubits to distinguish them from ordinary bits that are only ‘0’ or ‘1’.
Quantum states are notoriously fragile. In quantum computers, the information they carry, which can span multiple qubits and form ‘entangled’ states, tends to degrade or be lost as the computation progresses. To keep the state as long as possible, the qubits should be isolated from the environment. However, they cannot be too isolated from each other because they must interact to perform computations.
This, among other factors, makes building useful quantum computers difficult. But the field is more advanced than QuTech’s head of research Lieven Vandersypen anticipated a decade ago. “The progress is really impressive.”
In 2019, Google made headlines when it claimed that a machine made up of 54 superconducting qubits had performed the first quantum computation that would have taken an incredibly long time on a conventional computer. rice field. Technology company IBM, which has invested heavily in superconducting qubits, expects to reach a milestone in the coming months when it unveils a quantum chip named Condor, the first to break the 1,000-qubit barrier. increase.
Last November, the company announced its previous chip, the 433-qubit Osprey. This follows Eagle’s record-setting 127 cubits in 2021. says Jerry Chow, who leads the quantum computer program at IBM’s Thomas J. Watson Research Center in Yorktown Heights, New York.
quality and quantity
According to Chow, IBM’s aim is not only to scale up the number of qubits, but also to improve their quality. Some of the company’s superconducting elements can hold a quantum state for over 300 microseconds, he says. This is a record of this technique. Now he is error-free for 99.9% of his operations involving two qubits, one key metric.
With the number of superconducting qubits on a chip well above 1,000, scaling up becomes impractical. This is because each qubit must be individually wired to external circuitry for control and readout. Therefore, IBM takes a modular approach. After 2024, further steps on the roadmap aim not to increase the number of qubits on a chip, but to link multiple chips to his one machine. Entangle the qubits on separate chips. The chip is the heart of a large machine encased in a cryogenic system that brings the chip to near 0 Kelvin.
Trapped ion computers can be even more size constrained than superconducting computers because they require individual laser devices to control each ion. Typically, this means limiting the trap to columns of about 32 ions per tip. But IonQ, a startup spun off from the University of Maryland in College Park, says its approach will allow him to cram multiple rows of ions into a single chip, possibly reaching 1,024 qubits. . To go beyond that, IonQ also plans to move to a modular approach that connects multiple chips. In laboratory experiments, the fidelity of trapped ions he reached 99.99 percent, according to a company spokesperson.
tweezers technique
Another technique that received little attention until a few years ago could soon break the 1,000-qubit barrier. Focused laser beams, called optical tweezers, are used to trap neutral atoms and encode qubits into the electronic states of atoms or the spins of nuclei. The approach has evolved over her decade-plus, but is now “booming,” says Giulia Semeghini, a physicist at Harvard University in Cambridge, Massachusetts.
To assemble multiple qubits, physicists split a single laser beam into many, for example, by passing them through a screen made of liquid crystals. This allows you to create arrays of hundreds of tweezers, each trapping its own atom. Atoms are typically several micrometers apart from their neighbors and can remain in a quantum state for seconds or more. To make the atoms interact, the physicist points another laser at one of them to excite it into an excited state. In this excited state, the outer electrons orbit farther from the nucleus than they normally would. This promotes electrostatic interactions with neighboring atoms.
Using tweezers, researchers are building arrays of more than 200 neutral atoms and rapidly combining new and existing techniques to turn them into fully functional quantum computers.
One of the main advantages of this technique is that it allows physicists to combine multiple types of tweezers. Some tweezers move around quickly and can be used with the atoms they carry. Harvard physicist Dolev Bluvstein said: This makes the technology more flexible than other platforms such as superconductors, where each qubit can only interact with its direct neighbors on the chip. A team including Semeghini and Bluvstein demonstrated this flexibility in their April 2022 paper..
Tweezers-based qubits should soon be 99% error-free, but further improvements will require significant work, Semeghini said.
The pace of improvement of neutral atoms has taken the quantum computing community by surprise. Physicist Chao-Yang Lu of the University of Science and Technology of China (USTC) in Hefei said:
spin control
Other qubit technologies are still in their early stages, but are making steady progress. One method encodes information into the spins of individual electrons trapped by electric fields inside conventional semiconductors such as silicon. Last year, Vandersypen and his collaborators demonstrated a fully functional six-qubit his machine of this kind.2As in optical tweezers, the electron spins can be shuttled around the device and moved next to other devices as desired. But, as with other types of qubits, the main difficulty is keeping the spins from affecting each other when they aren’t supposed to, what physicists call crosstalk.
The advantage of semiconductor-based qubits is that they can be made in the same type of factories that make today’s computer chips, but a team led by physicist Michelle Simmons of the University of New South Wales in Sydney, Australia, has Assembling the device. atom by atom using the tip of an automated scanning tunneling microscope. “Everything is patterned with subnanometer precision,” she says.
Yet another approach is still in the concept stage, but has received significant investment, especially from Microsoft. The technique aims to make use of ‘topological states’ to make qubits robust against degradation, like a knotted string that can be twisted and pulled but not untied. In 2020, researchers observed the underlying physical mechanism of his one type of topological protection and are now working on demonstrating the first topological qubit.
“Every platform being pursued today has some promise, but developing it can sometimes require very novel ideas that are unpredictable,” says Vandersypen. Pan Jian-Wei, a physicist working on multiple quantum computing approaches at USTC, agrees. As for the race to develop quantum computers, “it’s still too early to tell which candidate will win.”
This article is reproduced with permission and was first published on February 6, 2023.
