A central goal in quantum optics and quantum information science is the development of quantum networks to generate entanglement between distributed quantum memories. Experimental progress relies on the quality and efficiency of the light–matter quantum interface connecting the quantum states of photons to internal states of quantum emitters. Quantum emitters in solids, which have properties resembling those of atoms and ions, offer an opportunity for realizing light–matter quantum interfaces in scalable and compact hardware. These quantum emitters require a material platform that enables stable spin and optical properties, as well as a robust manufacturing of quantum photonic circuits. Because no emitter system is yet perfect and different applications may require different properties, several light–matter quantum interfaces are being developed in various platforms. This Review highlights the progress in three leading material platforms: diamond, silicon carbide and atomically thin semiconductors.
The material hosting an embedded quantum emitter strongly impacts its properties, as has been shown over the past few decades for low-dimensional semiconductors and atomic impurities with quantized optical transitions. A wide range of possible platforms for quantum emitters exist, including semiconductor quantum dots (QDs) and atomic colour centres. These ‘artificial atoms’ are now being developed for several applications, from quantum computing and networks to quantum sensing and metrology.
The material platforms for artificial atoms developed to date entail important trade-offs in their optical properties and their quantum memory performance. Self-assembled semiconductor QDs are currently the best-performing photonic system as they offer an attractive combination of photon generation rate, quality of optical coherence, range of spectral tunability and mature methods for their integration into nanostructures. Indium arsenide (InAs) QDs have been used for the efficient generation of spin–photon1,2,3 and spin–spin4,5 entanglement, quantum teleportation6, quantum-state transfer7 and quantum relays8, as well as on-chip integration with polarization-demultiplexing waveguides9,10. QD-based qubits have been realized using excitons, charges and spins. Electron or hole spins provide the longest coherence time, but this time is currently limited to a few microseconds11,12. Another challenge for QDs is forming local qubit clusters by finding additional ancilla qubits nearby, as required by many applications13,14,15. Although magnetic dopants or coupled QDs may provide a solution16,17 to the latter challenge, other quantum systems offer more feasible routes for the production of ancilla qubits. Among these systems, a leading platform is the negatively charged nitrogen-vacancy (NV) centre in diamond18, which is formed by a N atom substituting a C atom and by a lattice vacancy. This atomic defect has an optically addressable spin-triplet ground state that is well decoupled from the almost spinless carbon lattice19. Major recent advances in the field of NV centres include the demonstration of spin coherence times exceeding 1 second20,21 at room temperature, a two-node quantum network realized across a distance of more than 1 km (the longest distance for all matter-based qubits)22, the addressing of several proximal nuclear-spin data qubits23 and quantum error correction with non-destructive measurements and real-time feedback13. However, extending spin–photonic quantum networks beyond two NVs requires advances in spin–photon interfaces or the use of improved entanglement protocols. Practical quantum network nodes (Fig. 1) will require the control of many spin–photon interfaces, likely with the integration of many quantum emitters in compact and efficient photonic chips. This scaling challenge is driving engineering efforts to further develop NV-diamond systems and is also stimulating the exploration of alternative solid-state quantum emitters that may be more suitable for scalable on-chip integration and may offer a better combination of stable optical transitions, multiple data qubits and high-fidelity logic gates. For example, defects in silicon carbide (SiC) or atomically thin layered materials24,25, such as transition metal dichalcogenides (TMDCs), may be integrated into photonic circuits more easily than diamond, which is fuelling research efforts on the realization of artificial atoms in these materials.
In this Review, we start by discussing diamond as a host material for quantum network applications, with a focus on NV centres and the recently investigated silicon-vacancy (SiV) centre. In particular, we review the outstanding challenges for SiVs and other similar centres that have the potential to become the best spin–photon interface. Next, we consider the SiC platform. Recent studies have revealed a wide variety of atom-like colour centres in different SiC polytypes; some of these are well understood, while others require more investigation. We also discuss device integration and scaling opportunities enabled by the industrial maturity of SiC growth and processing methods. Finally, we cover the very recent progress on artificial atoms in single-layer and few-layer 2D materials, such as TMDCs. Although their origin is in many cases still uncertain, these quantum emitters exhibit unusual photostability and emission efficiency despite their surface proximity; these properties bode well for atomic-precision control of quantum defects and their heterogeneous integration with electro-optic circuits. Moreover, the ability to create new types of atomically precise heterostructures by stacking layered materials offers great flexibility in the control of the spectral, charge and spin properties of the resulting system.
Distributed quantum networks
A central goal in quantum information science is the development of quantum networks to distribute entanglement between distant nodes26 (Fig. 1). A promising approach is to connect stationary qubit nodes using photons (Fig. 2a). In this quantum network approach, each node consists of a quantum memory whose register can have several ancillary qubits for efficient and fast single-qubit and two-qubit gates. The presence of several physical qubits in a node enables quantum error correction to produce resilient (ultimately fault-tolerant) logical qubits27. If at least one qubit in a node couples to an optical channel shared by other nodes, multiple nodes can be entangled with high fidelity using heralded optical measurements, even if the optical links are lossy28. The NV centre is a promising platform for this networking architecture because it has a stable electronic spin that mediates the coupling to a quantum register made of several nuclear spins that act as data qubits, while also coupling to photons through its optical transitions.
Several protocols have been proposed for entangling distant spin qubits through light detection29,30,31. The process relies on state-dependent photon scattering from two memories, followed by an operation known as a quantum erasure measurement, which serves to remove any information regarding the origin of the scattered photons creating quantum indistinguishability (Fig. 2a). The photonic qubit can be encoded in a dual-rail basis (such as photon polarization or two waveguide spatial modes) or in the photon number state (photon number states can be 0 or 1). Either way, the quantum erasure step collapses the quantum states of the spin memories into a non-local quantum superposition. Various blueprints for modular quantum computers based on this heralded entanglement approach have been proposed28,32,33,34.
Many experimental platforms are currently being pursued, with registers encoded in trapped ions, neutral atoms and solid-state spins. Atomic systems have long led the way, but atom-like systems in solids have been showing comparable or, for some metrics, even better performance. Indeed, the highest rate of entanglement generation was achieved with semiconductor QD spins at nearly 10 kHz4,5, whereas the current record for measurement-based entanglement generation fidelity is held by NV centres in diamond with a fidelity of up to 92%22. The high operational bandwidth for QDs arises from their strong optical transitions and fast (on the picosecond scale) all-optical spin-state control; however, the quantum storage time in QDs is modest owing to the spin noise of QD nuclei in III–V materials. By contrast, each NV centre has a stable electron spin-triplet state (with coherence time approaching 1 second) that couples coherently to photons at a wavelength of ~637 nm and also to nearby nuclear spins, such as those of the naturally present (intrinsic) N and proximal 13C nuclei (typically within 1 nm). Thus, every NV centre, with its surrounding nuclei, can be considered a small quantum computer coupled to photons, which makes NVs particularly attractive for quantum repeater networks or modular quantum computing. The main experimental obstacle is that only a small fraction of the NV's fluorescence occurs through the zero-phonon line (ZPL), directly coupling electronic states only, whereas the remainder goes into the undesirable phonon side band, which involves simultaneous excitation of the lattice. This fraction, given by the Debye–Waller factor, is approximately 3% and thus sharply limits the efficiency of the spin–photon interface. A promising solution is to incorporate the NV into an optical cavity to reshape its spectrum and improve overall photon collection. Cavity coupling results in a much stronger zero-phonon emission (more than 50% of the overall emission) for a cavity-coupled NV centre than for a similar NV centre in bulk diamond (Fig. 2b). The SiV and germanium-vacancy (GeV) centres in diamond inherently have a higher Debye–Waller factor (Fig. 2c,d), though they still have drawbacks, as discussed below.
Scalability route with diamond spins
The coupling between photons and quantum states of an emitter is efficient in the strong Purcell regime, in which the emitter interacts primarily with one optical mode. This regime can be reached by coupling the emitter to optical cavities with a high ratio of quality factor (Q) to mode volume (Vmode), yielding a Purcell factor (spontaneous emission rate modification) proportional to Q/Vmode. For semiconductor QDs, the strong Purcell regime is routinely achieved35,36,37. Because of the difficulty of fabricating complex nanostructures in diamond, initial work followed a more complex hybrid approach, coupling diamond nanocrystals to cavities fabricated in other materials38,39,40. However, even for a high Q/Vmode of ~103, the Purcell factor was small (<10) because of a weak spatial mode overlap between the NV centre and the cavity. Therefore, it has not been possible to bring the fraction of ZPL emission close to unity using hybrid approaches.
To increase this overlap, various all-diamond cavities have been developed, including microrings41, microdisks42 and photonic crystals (PhCs)43,44. Compared with hybrid approaches employing nanocrystals, the all-diamond cavity has the advantage of resulting in better spin and optical properties of the NV centres. Experiments on all-diamond cavities have increased the emission rate of the NV centre into the ZPL by more than 50 times, producing ZPL emission fractions in excess of 60% in PhC nanocavities45,46. Some representative state-of-the-art approaches to NV–cavity systems are shown in Fig. 3. Such nanocavities can be fabricated in large numbers and closely packed on a chip, favouring applications that require many individually addressable spin qubits, such as modular quantum computing and multiplexed quantum repeater schemes. For applications requiring a small or intermediate number of spin qubits, another promising approach consists of fibre-based microcavities fabricated around NV centres in diamond slabs, in which emitters have particularly promising optical stability47. These microcavities facilitate coupling to a targeted emitter by lateral and vertical translation of their curved mirror for spatial and spectral coupling, respectively.
Despite proof-of-principle demonstrations, a remaining challenge is to improve the fabrication yield by spatial alignment between the emitter and the cavity mode. One approach uses lithographic positioning of cavities around prelocalized emitters47,48,49. An advantage of diamond colour centres is that they can be produced at desired locations by ion implantation and annealing50,51,52. By targeted N implantation, NV centres have been incorporated into cavities through nanopatterning by means of an atomic force microscope tip53 (Fig. 3a) or by masks for self-aligned implantation and diamond dry etching54,55 (Fig. 3b). Diamond membranes containing NV centres produced during diamond growth enable excellent vertical alignment with the cavity mode (Fig. 3c). An alternative approach uses focused ion beam implantation of ions, such as N (ref.56) or Si (Ref.57). In particular, nanometre-scale ion beams of Si atoms have been used to implant SiV centres into 1D and 2D PhC cavities with sub-wavelength resolution58,59.
Another important question is how to connect and control potentially thousands of quantum registers. Photonic integrated circuits60 provide an attractive platform, but a major challenge is fabrication yield: one or more faulty components can cause the whole chip to fail. Whereas standard lithography methods can produce photonic integrated circuits with high yield (remaining fabrication imperfections can be compensated with post-production tuning61,62), fabrication methods for quantum registers are far less mature. Even with fabrication improvements, such as spatial alignment of emitters with cavities, the yield of acceptable emitter–cavity systems will likely remain low in the foreseeable future: too many aspects still lack sufficient control, including the spectral diffusion of the NV centres, the number of NV centres per cavity, the alignment of nuclear-spin physical qubits with the NV electron spin registers and the NV–cavity spectral overlap. One option is to integrate only preselected, functional NV-nanophotonic devices into photonic integrated circuits63. Similar approaches have been adopted for semiconductor QDs64,65,66,67.
In the past decade or so, several new, promising diamond quantum emitter systems beyond NV centres have been investigated. One is the negatively charged SiV centre. Compared with the NV centre, the SiV centre has a favourable ZPL branching ratio68 of 0.7, a shorter radiative lifetime on the scale of nanoseconds and lower spectral diffusion that results in photon emission with high indistinguishability69. The spectral stability derives from the inversion symmetry of the SiV centre, which prevents first-order Stark shifts (the linear shifting of spectral lines in an external electric field). The downside is that spectral tuning by electric fields is not available to shift SiV emitters into alignment, so other schemes, such as the use of Raman transitions, are required for spectral tuning59. In addition, the radiative efficiency of the SiV centre is significantly lower than that of the NV centre (Table 1), but it can be increased through the cavity Purcell effect. The ground state of the SiV centre comprises two orbital branches, which are split by 50 GHz; these states can be controlled coherently by means of laser pulses70,71. Moreover, the negatively charged SiV centre has an optically accessible spin 1/2 ground state72, which can be controlled coherently by microwave fields73.
A key challenge in the optimization of SiV centres has been the need to improve the spin coherence time, which is only ~100 ns at 4 K (Refs74,75) because of electron–phonon scattering between the two ground-state orbitals. Recently, it has been shown that by cooling to lower temperatures (~100 mK), the spin coherence time can be increased by orders of magnitude to ~10 ms (Ref.76). Alternatives to cooling to dilution-refrigerator temperatures are also being investigated. In particular, increasing the energy splitting of the orbitals of the SiV centre by applying strain through voltage-controlled nanoelectromechanical cantilevers resulted in an increase of the spin coherence time to almost half of a microsecond77 and this timescale has already enabled the observation of coupling to nearby 13C spins that can serve as ancilla data qubits. In parallel, other experiments suggested the existence of a neutral SiV centre charge state with longer spin coherence timescales reaching the millisecond regime78,79; however, an understanding of the underlying photophysics of the neutral SiV centre and signatures of optically detectable magnetic resonance are still missing. Recently produced GeV centres possess spin coherence properties similar to those of SiV centres but have higher radiative efficiency80,81.
Colour centres in silicon carbide
The exceptional spin coherence of diamond colour centres derives from the wide bandgap, weak spin-orbit coupling, dilute nuclear-spin bath (which can be suppressed further by isotope engineering) and availability of high-purity material growth protocols82. While SiC shares most of these properties with diamond, it also inherits important features from Si, such as well-established large-scale production methods, high-quality nanofabrication83 and controlled doping without adverse effects, such as the graphitization occurring in diamond owing to ion implantation84. More than 200 polytypes of SiC exist, each with a relatively large bandgap of 2–3 eV (Ref.85). Because of this large bandgap (compared with that of Si), SiC-based electronics are promising for high-temperature applications86. In addition, the large breakdown voltage and high thermal conductivity, which are roughly ten and three times higher than those of Si, respectively, make SiC an interesting platform for high-power electronic devices85,86. SiC is also considered a promising material for applications in microelectromechanical systems87. The good electrical properties and well-developed doping methods available for this material enable various sensor applications, such as optical sensors for UV and X-ray, and gas sensors88. Like diamond, SiC is chemically inert and biocompatible; thus, it is a relevant material for applications in biotechnology89.
The properties of SiC can also be harnessed for quantum applications: numerous colour centres have been reported in SiC substrates and optoelectronic devices (Table 2). Many of these defects can be excited by below-bandgap optical pumping. These colour centres commonly consist of substitutional impurities, such as transition metal impurities and vacancy-related defects. Vacancy-related defects can be created by high-energy electron irradiation or ion implantation90,91,92,93,94,95, often followed by annealing96,97. The first isolated single-colour centre in SiC was the positively charged carbon antisite–vacancy pair with bright (~1 million photons per second detected) photon emission near 700 nm at room temperature98. Although theoretical studies have suggested that the spins in the ground state of the neutral carbon antisite–vacancy pair may be coherently manipulated and optically addressed as the diamond NV centre99, its spin ground state has not yet been reported.
Among optically active deep defects, many electron paramagnetic defect centres have been reported. The neutral divacancy (DV) and negatively charged Si vacancy (VSi) in the 4H–SiC and 6H–SiC polytypes (Fig. 4a,b), in particular, are the most studied100,101,102,103,104,105,106,107,108,109,110. They show strong photoluminescence (PL) peaks at approximately 1.1 eV and 1.4 eV (Refs100,101), respectively. Spin-selective intersystem crossing allows spin polarization of the ground state by optical illumination (90%112 and ~80%103 for DV and VSi, respectively) and thus enables optically detected magnetic resonance of DV and VSi ensembles in hexagonal100,101 and rhombic polytypes110.
Isolated colour centres have been realized in high-purity SiC substrates. In high-purity semi-insulating SiC substrates grown by high-temperature chemical vapour deposition (CVD)113 and physical vapour transport114, the concentration of dominant residual impurities, such as shallow N donors and B acceptors, is on the order of 1015 cm−3. The concentration of paramagnetic intrinsic defects, such as the VSi, C vacancy and DV in high-purity semi-insulating SiC materials, is typically in the range of 1015–1016 cm−3, as estimated from electron paramagnetic resonance measurements114. This concentration is similar to that found in CVD-grown diamond, in which a spin coherence time (T2) of 600 μs was reported for NV electronic spin ensembles115. By optimizing the growth conditions, it is currently possible to achieve a concentration of residual paramagnetic impurities on the order of 1013 cm−3 (Ref.116). In CVD-grown epitaxial layers, the dominant intrinsic defect is the C vacancy, which typically has a concentration of ~1013 cm−3 or slightly lower117. The two most abundant nuclei in SiC (28Si and 12C) are spinless, whereas the second most abundant nuclei (29Si and 13C) have non-zero spins with a natural abundance of 4.7 and 1.1%, respectively. The mutual interactions of these types of nuclei can be suppressed by applying an external magnetic field owing to the mismatch between their Zeeman energies. As a result, the nuclear-spin bath in SiC is expected to permit long coherence times for embedded spin defects118,119. Experiments demonstrating the single-spin coherent control of stable isolated deep defects (DV and VSi) are shown in Fig. 4c (Refs93,96). T2 exceeds 160 µs for a single VSi at room temperature93 and 1 ms for DV ensembles at cryogenic temperatures96,119. VSi in natural 4H–SiC displays a temperature-dependent spin relaxation time (T1) of ~300 µs at room temperature and ~10 s at 17 K, and a T2 of up to ~20 ms, as measured by dynamic decoupling120. The coupling between the electron spins and nuclei provides candidates for ancilla qubits14,121; for example, 29Si (Refs93,96,100,111,122) has a hyperfine interaction of up to ~10 MHz for both VSi (Ref.105) and DV97 (for 13C, see Refs111,123).
The electronic S = 1 spin in the neutral charge state DV has a long electron–nuclear interaction time in the excited state111,124, which was used to obtain near-unity polarization in 29Si nuclear spins. Efficient coupling of DV spins to electric fields (Stark parameter up to ~30 Hz cmV−1) and strain (sensitivity ~10−7 Hz−1/2N−1/2) also enables sensing applications102. Electrical driving of the DV spins109 will be useful for developing integrated small-scale spintronic devices. Experiments on DV ensembles indicate that it is possible to produce electron–nuclear-spin-entangled states with fidelity approaching 0.9 (Refs125,126). The width of the ZPL of single DV defects is nearly lifetime-limited (~80 MHz, Fig. 4d)112, which is a key prerequisite for the efficient optical quantum interference necessary for long distance entanglement22,69.
The negatively charged VSi also has a high spin (S >1/2) in both the ground and excited states, with a zero-field splitting ranging from a few MHz to ~400 MHz (Refs104,106,127) and an optical spin polarization107,128 that persists up to 250 °C in some polytypes110. The ground state was often assigned as S = 1 (Ref.101), but experimental evidence proves that the ground state is a quartet manifold of S = 3/2 (Refs104,105,106,108,110,122,131). This multilevel ground state with a long spin lifetime93,120,122 enables the use of these defects for quantum magnetometry129,130,131, similarly to NV centres in diamond132,133. Whereas spins of other defects, such as the NV centre in diamond134,135,136 and DV in SiC (Ref.92), exist in more than one orientation, VSi spins have only one spin orientation along the crystallographic c-axis101,110. This enables an unambiguous assignment of the observed resonance transitions130,131, without the need to track the crystal orientation135,136,137,138. A recent experimental study showed that one of the SiV defects in 4H–SiC features two strong ZPL emissions (Debye–Waller factor of up to 40%)139. Theoretical work indicates that this centre exhibits polarization-selective transitions to two spin ground states128. The thermal stability of the centre's zero-field splitting over a wide range of temperatures (10–320 K) in some hexagonal polytypes140 holds promise for stability in sensing applications. Optically induced polarization inversion also enables stimulated microwave emission108.
Both the SiC DV and VSi are promising for high-fidelity spin-state readout through spin-selective resonant optical excitation107,112,139 (Table 2). These defects have excited-state lifetimes of ~10 ns (Refs94,102,112,139), similar to that of the diamond NV centre (Table 1). The collected photon flux is improved by embedding these emitters in photonic devices, such as solid immersion lenses93, nanowires141,142 or PhC cavities143,144,145. To produce nanophotonic structures, cubic polytype 3C–SiC thin layers have been heteroepitaxially grown on a Si wafer144,145,146,147,148, and hexagonal-polytype thin films have been produced on oxides by the smart-cut technique143,149. Selective etching of these substrates produces suspended structures, such as PhC cavities and whispering gallery mode resonators, with quality factors up to 50,000 (Refs143,144,145,146,147,148,150,151). Coupling emitters to SiC PhC cavities has increased photon collection by up to ten times145. Optically detected magnetic resonance152 was observed for an ensemble of Ky5 colour centres in 3C-SiC substrates, which is attributed to a neutral DV in this polytype. Dopant-selective photoelectrochemical etching of hexagonal-polytype layers has been used to fabricate a microdisk resonator150, as well as a nanobeam PhC cavity coupled to a VSi centre with a Purcell enhancement of 84 (Refs151,153) (Fig. 4e). Designs for PhC nanobeam cavities with triangular cross sections are a possible alternative154.
Ensemble studies on many other optically active defects in SiC, including the NV centre in 4H–SiC (Refs155,156) and other defects in 4H–SiC and 6H–SiC (Ref.92), have reported spin ground states (Table 2), though the isolation of individual defects has not been achieved. Various transition metal impurities157,158,159,160,161 are interesting as emitters in the telecom-wavelength band, and some of them exhibit high-contrast optically detected magnetic resonance158,160.
Bright colour centres without controllable and detectable spins can still be useful as single-photon sources. Examples include the positive carbon antisite–vacancy pair centres in 4H–SiC (Ref.98), oxidation-related defects at the surface of 4H–SiC and 6H–SiC (Ref.162), annealing-related defects in 4H–SiC with bright single-photon emission rates of up to 2 Mcps (Ref.163) and electrically driven single silicon antisites, known as D1 centres, in 4H–SiC p–n junction diodes164 (Fig. 4f). Because SiC allows for p and n doping, the realization of optoelectronics-based spintronic devices is also feasible. For example, a magnetoconductivity-based magnetic-field sensor for astrophysical research was realized165. With zero-field splitting values ranging from ~100 MHz (for the SiC VSi) to ~10 GHz (for SiC–Ti impurities), defects in SiC offer a wide operating range for magnetometry applications. In addition, polytype control92 and polytype stacking164,166 offer paths towards spin and optical control in SiC devices.
van der Waals materials
Groundbreaking work on graphene167 has stimulated the search for other atomically thin 2D layered materials with appealing electronic and optical properties. Such materials offer opportunities in basic science and are promising for the development of novel optoelectronic devices for energy-efficient information and communications technologies. Of particular interest are monolayer materials that exhibit semiconducting or insulating behaviour and have electronic bandgaps in the visible or near-infrared spectrum, such as TMDCs168,169. Many TMDCs, when reduced to a single layer, exhibit direct-band PL24,168,169. The 2D nature of the materials, the availability of semiconductors, metals and insulators, and the possibility of assembling layer-by-layer designer van der Waals heterostructures170,171,172 have enabled the observation of new physical phenomena, as well as the realization of advanced optoelectronic and nanophotonic devices173,174,175,176. Moreover, the electronic excitations in these materials exhibit a robust valley degree of freedom that coherently interfaces with photon polarization and can provide the physical basis of next-generation information processing applications177,178,179,180,181.
Whereas much research has focused on how defects influence the electronic properties182 of TMDCs, atomic defects are also promising for addressing the usual integration challenges in quantum photonics183. Novel atomically thin materials, such as TMDCs and hexagonal boron nitride (hBN), also support localized excitons184,185,186,187,188,189,190,191. This section reviews a selection of 2D materials with documented quantum defects and summarizes the state of knowledge of their optical properties, including their interactions with external electric, magnetic and strain fields192. An exciting prospect is that the TMDCs valley degree of freedom may be inherited by the confined excitons, providing an additional degree of freedom for quantum-optical applications193,194,195.
Tungsten diselenide (WSe2) is a TMDC that, when thinned to a single layer (consisting of three atomic layers), exhibits a direct bandgap at its K and K´ points. The first experimental evidence of exciton localization in WSe2 has been observed in the past few years by five research groups that concurrently reported on isolated optically active quantum defects184,185,186,187,188. Although research on such defects is currently intensifying, the field is so young that even the exact origin of these defects is an open question; in particular, their atomic configurations and electronic structures still need to be determined. Low-temperature photoluminescence spectra of two WSe2 defects188 are shown in Fig. 5a. Most observed WSe2 defects emit in the band between 1.63 eV (760 nm) and 1.72 eV (721 nm), with spectral linewidths on the order of 120 μeV. The spatially localized nature of the emitters is clearly seen in fluorescence images recorded on WSe2 flakes (Fig. 5a). WSe2 is not the only TMDC to support localized exciton emission. Optical excitation of defects in molybdenum diselenide (MoSe2)196,197 was demonstrated, as well as optical and electrical excitation of quantum defects in tungsten disulfide (WS2)198. The wavelength emission for these two materials is lower than that for WSe2 (650–700 nm). Large-scale deterministic creation of quantum emitters — a key step towards reliable fabrication of quantum devices on chip with high yield — has also recently been independently demonstrated by two research groups199,200: the fabrication method uses the effect of localized strain on the optical properties of layered materials to produce quantum emitters189 with almost unity probability when the layered material is transferred onto a patterned substrate. Similar quantum emitters were recently achieved in hBN layers on micropillar arrays; these function even at room temperature201. An alternative approach for the creation of full quantum confinement relies on patterning the TMDC into laterally confined disks. This was demonstrated for an array of disks with a diameter of ~20 nm patterned into molybdenum disulfide (MoS2)193, in which the confinement gave rise to localized photoluminescence (Fig. 5b). The photoluminescence spectra indicate that in patterned MoS2, the energy of the exciton is shifted compared with the exciton energy in bare MoS2 (Fig. 5b); the shift depends on the size of the disk, which is attributed to excitonic interactions. However, studies on single disks are still needed to elucidate these interactions and realize the promise of lithographically defined 2D disks as quantum emitters.
Similar to NV centres in diamond, the 2D insulator hBN also supports an optically active defect that emits single photons, even at room temperature190. These defects emit over a wide spectral range exceeding 200 meV centred around 2 eV (or 623 nm, Fig. 5c). Density functional theory (DFT) predicts that this defect is an antisite complex, in which a N atom replaces a B site, leaving the N site vacant, but this prediction still needs verification190. The electronic structure of this antisite complex is shown in Fig. 5c, and the resulting spectrum is in good agreement with the measured PL spectrum. Recent advanced DFT models also indicate the presence of hBN defects with spin-triplet ground states accessible by optically detected magnetic resonance202. Extensive optical spectroscopy studies on WSe2 and hBN defects have recently provided important insight into their photophysical properties. For example, fluorescence intensity autocorrelation measurements indicate single-photon emission between atom-like discrete energy levels. Such identification is confirmed by anti-bunched (below one at zero time delay) correlations for both WSe2 (Ref.184) (Fig. 6a) and hBN. Time-dependent spectroscopy reveals that WSe2 and hBN defects have excited-state lifetimes on the order of 1–3 ns, significantly longer than the ~30 ps lifetime of delocalized (bulk) excitons in WSe2 (Ref.181). Photoluminescence excitation spectroscopy measurements have corroborated the localized nature of these emitters and provided detail on the depth of the confinement potential well and its energy-level structure188; the excited-state resonances of a WSe2 defect are shown in Fig. 6b. In this experiment, PL from the n = 0 excitonic ground state is monitored as the energy of a narrow-band tunable laser is scanned across the higher-lying states. These measurements show excited-state absorption from defect resonances (n > 1) and from the free neutral exciton at 1.75 eV. These studies reveal that the depth of the confinement potential varies depending on the emitter in the range 50–200 meV, likely owing to inhomogeneous strain or structural defects.
Realizing a spin–photon interface in such atomic defects in 2D layered materials requires access to spin-selective optical transitions. Polarization-resolved spectroscopy has shown that emission from WSe2 (Ref.188) and hBN (Ref.190) defects is linearly polarized, independent of pump polarization. An external magnetic field can change the selection rules from linear to circular. A natural question that arises is what the role of the environment in influencing the photophysics of the defects is. Initial studies of WSe2 defects revealed a spectral wandering that was intrinsic to the sample and that could be modified by optical charging of the flake187. An important future direction is to understand the interplay between the quality of the substrate and the stability of embedded emitters.
An external magnetic or electric field can modify the optical transition energies, lifetimes, polarization selection and other photophysical properties of quantum emitters. Such interactions open opportunities for sensing, as well as spectral tuning, of quantum emitters. Moreover, controlled doping of the charge environment184 influences the stability of the defects and their transition energies (Fig. 6c). In this type of experiment, a back gate allows doping of the WSe2; this modifies the spectral location of the emission of the defect and changes its lifetime. Electrical excitation has also been recently demonstrated for this class of quantum defects198,203,204. The electroluminescence from a quantum light-emitting diode — in which a quantum emitter in a single layer of p-doped WSe2 is driven by injection of electron minority carriers through an hBN tunnelling barrier — is shown in Fig. 6d. Under these conditions, exciton recombination occurs in the quantum-confined region, leading to a sharp emission spectrum. Alternatively, magnetic fields applied perpendicular to the flake can shift the spectroscopic lines of the defects via the Zeeman response. It has been observed that many WSe2 defects possess zero-field exchange splitting with an average value of ~0.7 meV (although some defects have been observed with no or very small zero-field splitting). These experiments also revealed a remarkably large Landé factor of ~10, nearly an order of magnitude larger than that of QDs in III–V materials205, even though the polarization properties of WSe2 neutral exciton transitions are similar to those of self-assembled QDs in III–V materials. Specifically, at zero magnetic field, the emitted photons are linearly polarized owing to the anisotropic electron–hole exchange interaction. An out-of-plane magnetic field dominates over the electron–hole exchange interaction and recovers circularly polarized optical transitions, but an in-plane magnetic field has no measurable effect185. The reason for this magnetic-field dependence is that the 2D valley degree of freedom couples only to out-of-plane magnetic fields. These studies indicate that the electronic structure of defects in 2D materials inherit the valley properties of the host material. This observation on WSe2 defects suggests that the valley degree of freedom may also be exploited to build a localized quantum memory, much like the spin of an electron. Finally, directional strain on 2D crystals influences the emission characteristics of WSe2 and hBN defects189,201 and opens a route to nanometre-scale band-structure engineering. With continued experimental progress in the understanding and control of the internal properties of such systems, we may expect optically active defects in 2D materials to play an increasingly important role in future quantum photonic technologies.
Conclusions and outlook
In this Review, we focused on quantum emitter systems in diamond, SiC and layered 2D materials, three emerging platforms with distinctly promising properties. In the quest to find one quantum defect that ticks all the boxes for building scalable quantum networks, efforts should continue along two routes: improvements of current systems to push the performance of single- and few-qubit devices and exploration of new material platforms that might have even better properties. Another promising perspective is provided by hybrid systems, in which quantum materials are combined with other platforms (such as waveguide circuits) to make the best use of the properties of each material. How well such hybrid approaches will perform depends on how effectively materials can be combined and quantum states exchanged between them. The progress in the field over the past decade has been remarkable. Many start-up companies have already begun to commercialize artificial-atom technologies, such as InAs QDs and diamond NV centres, for quantum information processing and sensing applications206. Within the next decade, we can expect that today's leading quantum systems will be refined, benchmarked and increasingly put to use in applications, while new artificial-atom technologies with even greater potential will be developed.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors acknowledge financial support from the European Research Council (ERC) Consolidator grant PHOENICS agreement no. 617985, the Engineering and Physical Sciences Research Council (EPSRC) National Quantum Technologies Programme NQIT EP/M013243/1, Korea Institute of Science and Technology (KIST) institutional programs (Grants No. 2E27231, 2E27110), the Army Research Laboratory Center for Distributed Quantum Information (CDQI) and the National Science Foundation program ACQUIRE: “Scalable Quantum Communications with Error-Corrected Semiconductor Qubits”, NSF EFRI EFMA-1542707, NSF CAREER DMR 1553788 and AFOSR FA9550-16-1-0020. S.Y.L. and J.W. thank S. Nguyen and B-S. Song for comments.
About this article
Nature Photonics (2018)