The spin state of a single erbium atom in a tiny silicon transistor has been probed using laser light. The finding opens a path towards a hybrid spin–photon quantum-computing architecture. See Letter p.91
The relentless drive to scale down semiconductor devices, which are at the heart of computers, tablets and video game consoles, has long been predicted to hit a brick wall when atomic dimensions are reached. However, with smaller devices also comes increased access to new functionalities, which can be based on the control of single atoms and atomic defects1. Access to properties other than the commonly used electric charge of electrons in a semiconductor, such as the spin of electrons and nuclei, is a possible route to improved device performance, and perhaps even to quantum computing2. Indeed, researchers have demonstrated3 robust control of the spin of an electron bound to a single phosphorus atom in silicon. In this issue, Yin et al.4 (page 91) describe another exciting advance in spin control — the spin-selective 'optical addressing' of a single atom in silicon.
In their study, Yin and colleagues shine light from a precisely tuned laser on a single erbium atom that has been implanted in an electrical switch (a transistor). The laser light is absorbed by the erbium atom and changes the atom's charge state in a way that depends on its spin state (Fig. 1). The transistor is operated at cryogenic temperatures in a mode in which electric transport consists of a trickle of electrons that tunnel through the device one at a time. In this single-electron transistor (SET) mode, the device can reliably detect the ionization of a single erbium atom.
Erbium is a poster child of silicon-based photonics because its dominant optical transition, which occurs at a wavelength of about 1,540 nanometres, lies right in the transmission window of the optical fibres commonly used in telecommunication. But optical transitions of erbium atoms in silicon are relatively slow, a fact that has precluded their use in the readout of spin states of single atoms through the detection of emitted photons. A key advance of the present study is the use of a SET as a sensitive electrical sensor of the charge state of an erbium atom during optical excitation. What's more, when the authors added a magnetic field to this mix, they could also use the SET to measure the spin states of the atom's electrons and nucleus. In this way, Yin et al. were able to observe the eight spectral lines that are associated with the 'hyperfine splitting' of energy levels of one erbium atom with nuclear spin 7/2. Remarkably, the laser light used in the set-up did not cause excessive charge noise in the SET. Although the nuclear spin state was detected using multiple laser shots, single-shot detection and manipulation of single nuclear spins should become possible with further refinements of the technique.
These results are a major step towards access to both single spins and photons in a silicon-based platform — a combination that might be used to attain efficient quantum communication. The findings are also exciting because they expand the list of atoms and atomic defects for which reliable single-spin access can be obtained, beyond phosphorus in silicon3 and nitrogen-vacancy centres in diamond5. And, as Yin et al. point out, other atomic defects can be probed with this method, too. Compared with the popular optical methods for measuring the spins of nitrogen-vacancy centres, which can be performed at room temperature, the authors' hybrid opto-electronic approach circumvents the problem of limited photon-collection efficiency that plagues purely optical methods.
It is also noteworthy that the SET used here is a FinFET6, a three-dimensional type of transistor that is mass-produced for processors in consumer electronics. Devices based on the authors' hybrid approach might therefore serve as a platform for expanding the functionality of semiconductor systems beyond the mere shuffling of electrons to encode 'zeros' and 'ones' — the classical bits of digital information and computing. With regard to quantum computing, which requires integration of quantum memory, logic and communication modules, Yin and colleagues' work offers a promising route towards the integration of spin- and photon-based quantum bits (qubits) in silicon. This is because nuclear spins of dopant atoms such as erbium and phosphorus have extremely long 'coherence times', which are desirable for quantum memories7. In addition, quantum information could be transferred between electron and nuclear spins8 and possibly also encoded in single photons emitted by an erbium atom.
To integrate erbium-based qubits into a photonic network, the issue of efficient photon collection will also have to be addressed, perhaps by placing the qubits in optical cavities (arrangements of highly reflective mirrors that trap light). And although many challenges remain, integrating these elements in silicon has the tantalizing potential to achieve distributed quantum-computing architectures (see, for example, ref. 9) at the exact wavelength that is used for classical telecommunication.