"A silicon-based quantum-computing platform has met key standards for reducing error — setting the stage for quantum devices that could benefit from established semiconductor microchip technologies.
Quantum bits (qubits) that use the quantum properties of electrons in silicon devices offer enormous potential for developing compact and robust quantum computers that take advantage of the existing silicon-microchip industry. But quantum operations are subject to error, and getting error rates low enough to make quantum silicon devices feasible remains a challenge. Three papers in this issue, by Xue et al.1, Noiri et al.2 and Mądzik et al.3, report demonstrations of qubit operations in silicon devices with fidelities above the threshold of one of the most popular quantum error-correcting codes. The results suggest that these devices could be a competitive platform for scalable quantum-information processing.
The basic idea behind any quantum computer is that the quantum nature of qubits enables them to be in a state that is not simply ‘1’ or ‘0’, but some combination known as a superposition. This means that two qubits can be in a superposition of ‘01’, ‘10’, ‘11’ and ‘00’, which leads to even more possibilities. This ability can be used to speed up certain computations that are too complicated for a classical computer to perform in a reasonable amount of time. These include Shor’s algorithm, a strategy for factorizing large numbers, which could compromise existing encryption schemes for internet security, and other algorithms that could be used in materials science and drug design.
The effort to build a large, operational quantum computer is an extremely ambitious technological undertaking. Currently, the most widely used qubits are made of superconducting circuits that are used by technology companies such as IBM and Google in their quantum processors. Important demonstrations have been achieved with these systems, including proof-of-principle chemistry simulations4 by IBM. The quantum-computing researchers at Google reported that one of their superconducting-qubit devices took around 200 seconds to perform a computation for which they claimed a classical supercomputer would require 10,000 years5. However, superconducting qubits are relatively large, and this makes it difficult to fit a lot of them on one chip housed in a single cooling system, and to scale to larger devices.
Another problem, common to all qubits, is that they can remain in a given superposition of states for only a limited amount of time, known as the coherence time. This limitation introduces errors into computations performed with qubits and has motivated the development of error-correction protocols aimed at mitigating these errors, by implementing quantum algorithms using more qubits than the computations require. This has the effect of grouping many physical qubits into fewer logical qubits, leading to redundancy, which can be used for quantum error correction.
A compact alternative to the superconducting qubit also has long coherence times: the electron spin qubit6 is based on the quantum nature of the magnetic moment (or spin) of the electron. These qubits must be formed through a mechanism that isolates individual electrons from their environment while ensuring that they are accessible and can still be controlled with applied electromagnetic fields. The most common way to do this is with quantum dots7, which are tiny traps that form at the interface of two semiconductor materials. Multiple quantum dots can be engineered by using metallic leads to create separate traps, each of which can host a single electron.
Silicon is an attractive material with which to create spin qubits. This is mainly because it can be isotopically purified, which means that the vast majority of its atoms will not have a net spin associated with their nuclei. If the atoms had non-zero nuclear spin, this spin would interact with the qubits and lead to the loss of quantum information. Furthermore, silicon is the material used for the circuits in everyday computers, so silicon-based quantum computers stand to benefit greatly from the existing nanoelectronics infrastructure, which is very sophisticated. This explains why the quantum-computing research programme of Intel, one of the leading chip-manufacturing companies, has focused on this platform8.
To implement quantum algorithms, single-qubit control and two-qubit interactions are required to realize both single- and two-qubit logic gates. These gates are the building blocks of any quantum algorithm — similar to the logic gates used in classical computing. They can be produced using magnetic fields, with which spins interact naturally. However, these magnetic interactions are weak. By contrast, electric fields offer faster control by coupling the spin to the motion of the electron, and can be used for both types of gate.
For two-qubit gates, two electrons are brought close together — close enough that they have overlapping quantum-mechanical wavefunctions, which can be thought of as the spatial extent of the electrons. This overlap gives rise to an effective spin–spin interaction that, when carefully controlled, results in the qubits becoming ‘entangled’, meaning that they share a common state and can no longer be described independently. Changes to the state of one qubit depend on that of the other. The accuracy with which these conditional operations are carried out is measured by a quantity called fidelity, which needs to meet a minimum threshold for quantum error-correction strategies to be viable. Xue et al. and Noiri et al. met this milestone in experiments using electron spin qubits in isotopically enriched silicon quantum-dot arrays (Fig. 1a).
Mądzik et al. made similar progress on a different silicon spin qubit made from the nuclear spin of a phosphorus atom that had been substituted in place of a silicon atom in a silicon lattice (Fig. 1b). In this case, the nuclear spin of a single atom can be used as a resource. Nuclear spins have extremely long coherence times, which makes them attractive spin-qubit candidates, but methods to induce interactions between nuclear spins to enable accurate operations have been scarce. Mądzik and colleagues used an electron to mediate the interactions between two nuclear spins, and generated high-fidelity two-qubit entangling gates, bringing the accuracy of nuclear spin-qubit operations to the quantum error-correction threshold.
The results of all three groups move silicon-based quantum-information processing a step closer to offering a viable quantum-computing platform — a status so far held by only a few other systems, including superconducting qubits9 and trapped ions10. However, there are still challenges to be overcome if the groups’ devices are to become scalable. A key issue is that a lot of the qubits’ calibration, benchmarking and achieved fidelities will be negatively affected when the system size is increased — even by a single qubit. The next experimental milestone for this system would therefore be to build a larger array of quantum dots hosting two-qubit gates with fidelities as high as those demonstrated by Xue et al. and Noiri et al., despite the presence of more qubits. A further breakthrough for such a system would be the demonstration of quantum error correction.
At some point, adding qubits to the array will do more harm than good. This is because it will become too difficult to calibrate and control a large system with multi-qubit interactions. A modular architecture could instead be developed based on resonators, which are microstructures that use light to tune spins to certain ‘resonant’ frequencies. Resonators have already been shown to be capable of coupling two quantum-dot spins11, and could potentially be used to build arrays of quantum dots that are connected in a network. The details of how large these modules need to be and how they would be connected is an open problem from both a theoretical and an experimental point of view." [1]
1. Nature 601, 320-322 (2022)
Komentarų nėra:
Rašyti komentarą