Quantum emitters (QEs) that emit one photon at a time are key building blocks for numerous quantum technology protocols such as quantum communications, quantum information processing, and quantum key distribution.13-15 In particular, single photon sources operated in the telecom bands (1.25–1.55 mm) are highly desired because they allow implementation of quantum technologies through existing fiber-based optical communication networks. An ideal QE should be on-demand, deterministic, and should deliver exactly one photon in a well-defined polarization and spatiotemporal mode.16 Currently, telecom-compatible single photon sources have been demonstrated in various III-V semiconductor quantum dots11,12 and recently in functionalized carbon nanotubes.10,17 Although these single-photon sources have achieved telecom emission down to 1.55 mm, several challenges such as accurate site positioning and efficient polarization control still remain. Over the past decade, two-dimensional (2D) semiconductors have emerged as a novel optoelectronic platform for both fundamental research and advanced technological applications. Leveraged by the unique membrane-like planar geometry, 2D semiconductors are promising for QEs as they offer high photon extraction efficiency, easy coupling to external fields, and convenient integration with photonic circuits. In addition, as atomically-thin 2D flakes are stretchable and flexible, strain engineering can be readily applied to accurately position the emission sites.3,8,9 Most importantly, atomically-thin 2D transition metal dichalcogenides (TMDCs) have a valley degree of freedom that can be manipulated and accessed through circularly polarized excitonic optical transitions and efficiently tuned via a magnetic field,2,18 bringing new quantum functionalities to embedded QEs. Recently, 2D QEs have been demonstrated in WSe24-7 and hexagonal boron nitride,19,20 covering a broad emission spectrum range from ~500–800 nm. However, SPE emission in the most desirable spectral range – the telecom bands – has never been explored in 2D systems. While most 2D semiconductors have inherent electronic band structures that limit the operating wavelength to the visible range, alpha-phase 2H-MoTe2 has a layer-dependent bandgap in the NIR regime, holding promise for telecom-compatible single photon emission.
We demonstrate the first observation of telecom single-photon emission in MoTe2 mono- and few-layers. We transferred mechanically exfoliated MoTe2 thin flakes onto nanopillar arrays to introduce point-like bi-axial strains, which locally trap excitons into the strain-defined potential landscape, leading to isolated emitters (see Methods and Supplementary Section S1). We observed quantum emitters in both monolayer MoTe2 flakes and relatively thicker MoTe2 samples up to >10 layers. Figure 1a is an optical image of a MoTe2 monolayer on nanopillar arrays. The corresponding wide-field photoluminescence (PL) image, Figure 1b, shows the strained regions give significantly brighter PL emission than the flat areas. Unstrained monolayer MoTe2 features two dominant emission peaks at cryogenic temperature: exciton (X0) emission at around 1050 nm and trion (X±) emission at around 1070 nm (Figure 1c).21,22 With localized strain, a series of narrow PL peaks emerges from the lower energy side of the spectrum and covers a broad spectral range (Supplementary Section S2). The linewidths of such narrow PL emission lines range from a few meV to sub-meV at 10 K temperature, which are significantly narrower than the linewidths of the delocalized MoTe2 exciton PL peaks, and are comparable to those of carbon nanotube (CNT)-based quantum emitters. The narrow near-band-edge emission lines are also frequently observed in 2–4 layered samples but not commonly seen in much thicker flakes. Figure 1d shows a typical narrow PL peak from a localized emitter that displays a linewidth (full width at half maximum, FWHM) of 920 meV. The peak is accompanied by a weak shoulder peak at ~ 2.5 meV higher energy side. Both the width and shape of this PL peak remain essentially unchanged over nearly 3 order of magnitude change in pump power. The PL intensity also vary linearly with pump power over 2 order of magnitude change in pump power and only shows a weak saturation at powers >300 nW (figure 1e), indicating the robustness of the QE emission.
To verify the single photon nature of the narrow emission peaks originated from localized strains, we performed Hanbury Brown-Twiss (HBT) experiments to measure photon antibunching. The second-order correlation at zero time delay, or g2(0), was extracted to evaluate the probability of detecting two photons simultaneously. Figure 2a displays the PL spectrum of a strain-induced localized emitter, of which the PL dynamics and the auto-correlation measurement results are presented in Figure 2b-d. The time-resolved photoluminescence (TRPL) decay curve (Figure 2b) shows a near-perfect single-exponential decay with a lifetime t = 22.2±0.1 ns that is four orders of magnitude longer than that in pristine MoTe2 (~2 ps), which is attributed to the non-radiative decay dominated process.23 This long lifetime provides a clear indication that localization of the exciton in a strain induced potential trap prevents the exciton from recombining via non-radiative defects that dominate the decay of 2D band-edge excitons.
Quantum-dot-like solid-state QEs typically have instability issues such as photon bleaching, blinking, and spectral diffusion, which hinders applications. For instance, the blinking and spectral diffusion behavior of PbS quantum dots (one material capable of quantum light emission at NIR range) is wellknown.24 We monitored the time-dependent emission using both a single-photon detector (Figure 2b, inset) and an InGaAs spectrometer (Supplementary Section S3), finding no detectable photon bleaching, blinking, or spectral diffusion over the timescales presented. Some of our MoTe2 quantum dots were measured continuously over 24 hours and no signs of degradation were detected. Figure 2c and 2d present the photon correlation under pulsed [g2(0) = 0.098±0.003] and CW [g2(0) = 0.181±0.030] excitation, respectively. Both values are well below the photon antibunching threshold of g2(0) = 0.5, which unequivocally reveals the strain-induced MoTe2 localized emitter is a QE.
Obtaining telecom-compatible QEs that emit at around 1.3 mm (O-band) and 1.55 mm (C-band) are required for fiber-based quantum communications as the transmission loss in optical fibers are minimized in these bands. When increasing the MoTe2 layer numbers from monolayer to bulk, the bandgap of MoTe2 decreases from 1.18 eV (1050 nm) to 0.95 eV (1300 nm) monotonically.22 The PL intensity also decreases orders of magnitude due to a direct-to-indirect bandgap transition commonly observed in 2D TMDCs. In our experiment, we observe bright telecom wavelength emissions created from strained few-layer MoTe2 although we have occasionally found such emissions in mono- and bilayer samples (Supplementary Section S4). Figure 3a and Figure S5a,b present PL spectra of telecom-wavelength emitters. We typically observe such highly red-shifted bright emissions spanning 1.25–1.55 mm, covering the full telecom window. In contrast to QEs emitting at wavelengths below 1200 nm, these are characterized by relatively broad linewidths (FWHM 7-~30 meV). Figure 3b is the PL spectrum of a telecom emitter, of which PL dynamics and photon correlation results are presented in Figure 3c-3f, respectively. We observed an initial PL decay with a lifetime of 163 ± 3 ns, followed by an ultra-long decay lifetime of 1.13 ± 0.01 ms. This ultra-long lifetime is attributed to the 1540 nm emission, confirmed by comparing the integrated PL counts from the TRPL curve with the PL counts in spectrum (details given in Supplementary Section S5). The measured lifetime is 6 orders of magnitude longer than that of the MoTe2 band-edge emission and is of 2–3 orders larger than that of the near-band-edge QEs mentioned before. Based on this long lifetime and the fact that these telecom-QEs are observed more on multilayer thick MoTe2, we tentatively attribute the telecom-QEs to indirect excitonic transitions, which are activated by strain induced quantum confinement potentials. Figure 3d show that this telecom-QE is also free of blinking and photon bleaching over 5000 s experiment time. The pulsed excitation photon correlation measurement for the spectral window shown in Figure 3b yield a g2(0) = 0.48±0.03 (Figure 3e). Since this measurement includes contribution from a high-energy shoulder that exhibits shorter PL decay, we employed a time gated g(2) experiment25 (Supplementary Section S6), in which only the photons arriving after the decay of the higher energy shoulder (i.e. after 200 ns delay) were analyzed for the g(2) trace. The time gated g(2) in Figure 3f shows g2(0) = 0.155±0.009 clearly proving the antibunching nature of the 1540 nm telecom emitter. Our result is the first ever demonstration of a 2D material-based C-band (~1.55 mm) telecom QE.
Semiconducting 2D TMDCs are well-known for their valley degree of freedom, which gives rise to interesting phenomena such as valley Zeeman splitting and valley polarization. To investigate the valley physics of the strain-induced QEs in MoTe2, we conducted polarization-resolved magneto-PL spectroscopy with a magnetic field normal to the sample surface (Faraday geometry). Figure 4a presents a helicity-resolved PL study of a MoTe2 QE. The spectra was taken with s + excitation and analyzed for both s + and s – helicities. The valley Zeeman splitting was not observed in the absence of a magnetic field but rose with increasing field, indicating a lifting of valley degree degeneracy. Using the relation between the energy splitting ΔE and the magnetic field B, ΔE = −gμBB, where μB is the Bohr magneton, we extracted a Landé g-factor of -3.61±0.02 for the quantum dot (Supplementary Section S7), which is comparable with g-factors reported in other work.26 At zero field, the degree of circular polarization, PC = (Iσ+−Iσ−)/(Iσ++Iσ−), was 13% as a result of the valley selective-pumping effect. Significant valley polarization was observed with an applied magnetic field, reaching 52% at 8 T, which is more pronounced compared to the 30% polarization reported in magnetic-field induced valley polarization of intrinsic MoTe2 excitons.26
In addition to the Zeeman splitting of valley degenerated trapped excitons, we also observed emission pairs in some of the QEs, which were found to be cross-linearly polarized doublets with sizeable zero-field energy splitting (1-3.7 meV) (Figure 4b and Supplementary Section S8). The two peaks from the doublet have a zero-field splitting of 1.09 meV and reach their maximal PL intensities in opposite linear polarization directions (horizontal/vertical), indicating the presence of fine structure. The observed cross-polarization and zero-field energy splitting have also been reported in III-V quantum dots27-29 and recently in WSe2 QEs.4,6,30 Following prior studies, we attribute this fine structure splitting to hybridization of K and K' valley polarized excitons by an asymmetric potential landscape defined by the localized strain as illustrated in Figure 4d. In our system the zero-field splitting is larger than those (10s – 100s of µeV) observed in III-V quantum dots, indicating a strong electron-hole exchange interaction in such a strained 2D system. When a magnetic field is applied, the Zeeman energy splitting (ΔEZS) and the anisotropic-potential-induced zero-field splitting (ΔE0) compete with each other. Once the field is strong enough to overcome the anisotropic Coulomb potential, that is, ΔEZS > ΔE0, the linearly polarized states vanish and circularly polarized states are recovered. This restoration of valley symmetry is achieved in our QE under an 8 T magnetic field (Figure 4c). The helicities of each peak can be reversed by flipping the direction of the magnetic field, proving that the valley Zeeman effect is now dominating over anisotropic exchange.
In summary, we have deterministically created single-photon emitters in monolayer and multilayer MoTe2 using nano-pillar-based strain engineering. The emission wavelength ranges from 1080–1550 nm, covering all telecom bands. The quantum nature of the localized emitters was verified by photon correlation measurements. We observed cross-polarized doublets with ~1.1 meV zero-field energy splitting in some of the localized emitters, suggesting strongly anisotropic confinement. The polarization of the QEs was found to be tunable by external magnetic field. Our findings extend the operating wavelength of 2D QEs into the NIR regime, bringing in new solutions for creating site-controlled stable telecom-compatible quantum emitters. We envisage various future directions inspired by our early-stage demonstration, including electrically driven telecom quantum emitters and cavity-enhanced tunable NIR QEs. We are also encouraged by the possibility of realizing room-temperature MoTe2 telecom QEs as the energy redshift between the telecom emission and the MoTe2 exciton emission is well above the thermal energy at room temperature. Finally, an in-depth understanding of the excitonic physics of MoTe2 QEs may give rise to new perspectives on manipulating the spin-valley matrix and designing valleytronic devices.