Recently, single photons have been observed emanating from point defects in two-dimensional (2D) materials including WSe2, WS2, hexagonal-BN, and GaSe, with their energy residing in the direct electronic bandgap. Here, we report single photon emission from a nominal weakly emitting indirect bandgap 2D material through deterministic strain induced localization. A method is demonstrated to create highly spatially localized and spectrally well-separated defect emission sites in the 750–800 nm regime in a continuous epitaxial film of few-layer WSe2 synthesized by a multistep diffusion-mediated gas source chemical vapor deposition technique. To separate the effects of mechanical strain from the substrate or dielectric-environment induced changes in the electronic structure, we created arrays of large isotropically etched ultrasharp silicon dioxide tips with spatial dimensions on the order of 10 μm. We use bending based on the small radius of these tips—on the order of 4 nm—to impart electronic localization effects through morphology alone, as the WSe2 film experiences a uniform SiO2 dielectric environment in the device geometry chosen for this investigation. When the continuous WSe2 film was transferred onto an array of SiO2 tips, an ∼87% yield of localized emission sites on the tips was observed. The outcomes of this report provide fundamental guidelines for the integration of beyond-lab-scale quantum materials into photonic device architectures for all-optical quantum information applications.
Single photon generation is a requirement for quantum key distribution and all-optical quantum computing, and is crucial for the advancement of quantum information technology.1–6 Beginning with WSe2 in 2015,7 several recent studies have observed single photons originating from defect structures in two-dimensional (2D) materials such as mechanically exfoliated WSe2,8–18 WS2,14,15 hexagonal-BN,19 and GaSe,20,21 and in chemical vapor deposition (CVD) synthesized WSe27 and hexagonal-BN22 where second order photon correlation parameters have reached as low as 0.07–0.39 (Table S1, supplementary material). Optical emission energy in these systems resides within the electronic bandgap, and excitation had been provided by optical pumping8–11,14–16,18,20–22 and electrical charge injection.14 Previous studies have postulated that nonuniform strain fields govern quantum emission in these materials,15,16,23 which may benefit secure communication technology as the use of single photon sources requires both spatial control of the emission site and no more than one emitter per site.24–26 However, the mechanism responsible for recent spatial control based on dielectrically nonuniform tent-pole-style pillars15,16 remains unclear, as does the generality of conclusions based on these studies. Additionally, scalability also currently limits progress: lab-scale mechanical exfoliation and powder vapor transport growth of small crystallites constitute the only cases where this effect has been observed in semiconducting 2D materials.7–16,20,21
Here, we demonstrate a method to create highly spatially localized and well-separated defect emission sites in a continuous film of few-layer epitaxial WSe2 synthesized by a multistep diffusion-mediated gas source chemical vapor deposition (CVD) technique.27 The synthesis method employs W(CO)6 and H2Se vapor-phase sources enabling control over WSe2 nucleation density and lateral domain growth which are necessary to achieve uniform epitaxial films on sapphire (0001). When the coalesced epitaxial WSe2 film was transferred onto an array of ultrasharp SiO2 tips, we observed one order of magnitude longer bound exciton lifetimes from the tip apex compared to defect emission intrinsic to the film, where both defect types emitted in the ∼725–810 nm wavelength regime. Narrow linewidth emission was seen arising from 13 out of 15 tips, and we present a detailed analysis of power, temperature, and quantum emission characterization of one tip. Single photon generation at the tip apex was confirmed, where single photon purity reached ∼70% at 3.8 K.
Uniaxial and biaxial tensile (compressive) strain reduces (increases) the electronic bandgap in 2D transition metal dichalcogenide (TMD) materials such as MoS228–31 and WSe2.31–33 This may lead to a quantum dot-type energy landscape if strain is applied locally and has been hypothesized to be the mechanism responsible for quantum emission from WSe2 and WS2 transferred onto tent-pole-style engineered substrates12,15,16 or patterned surfaces.13 To separate the effects of mechanical strain from substrate- and/or dielectric-environment induced changes in the local electronic structure, we have created arrays of silicon dioxide tips with well-defined tip radii. Large 20 μm × 20 μm sputtered chromium pads were used as the hard mask, while a buffered oxide etchant was used to isotropically etch the tips from a ∼10 μm-thick CVD SiO2 film deposited onto substoichiometric silicon nitride which acted as a dielectric etch stop. On the top (bottom) surfaces of the WSe2 film, the tensile (compressive) strain can be estimated as
where h is the thickness of the WSe2 film and r is the tip radius. Figure 1 illustrates that by applying Eq. (1) to an n = 1–5 layer WSe2 film with layer thicknesses of 6.491 Å34 placed onto a 10-nm radius tip, a maximum strain of 3.2%–16.2% will arise on the film surface with a strain gradient on the order of 1010% m−1 perpendicular to the film. This symmetric strain profile thus should have an observable effect on the excitonic emission33 and lifetime,35 although the true strain experienced will depend on a number of other issues such as compliance and adhesion energy. We note that a recent study36 found an order of magnitude increase in the thermal expansion coefficient (α) of several monolayer TMD materials compared to the bulk material. The relationship between α and elastic modulus (E) can be expressed as37
where γ is the Grüneisen parameter, ρ is the mass density, and cv is the specific heat. Since α is inversely proportional to E, it is likely that mono- and few-layer TMD materials are significantly more compliant than their bulk counterparts and thus are more able to be strained by the sharp tips in this work.
Ultralarge strain and strain gradients are possible for atomically thin crystals transferred onto ultrasharp tips. (a) Calculated maximum strain and (b) perpendicular strain gradient that will arise on the top (tensile) and bottom (compressive) surfaces of an n-layer WSe2 film as a function of tip radius according to Eq. (1). (c) and (d) Scanning electron micrographs of a representative sharp tip. (e) Details of cross-sectional transmission electron microscopy analysis illustrating a 5–10-layer thickness for epitaxial WSe2 in this work.
Ultralarge strain and strain gradients are possible for atomically thin crystals transferred onto ultrasharp tips. (a) Calculated maximum strain and (b) perpendicular strain gradient that will arise on the top (tensile) and bottom (compressive) surfaces of an n-layer WSe2 film as a function of tip radius according to Eq. (1). (c) and (d) Scanning electron micrographs of a representative sharp tip. (e) Details of cross-sectional transmission electron microscopy analysis illustrating a 5–10-layer thickness for epitaxial WSe2 in this work.
We make use of a recently developed vertical cold-wall CVD reaction scheme27 to deposit WSe2 over a roughly 1 cm2 sapphire c-plane (0001) substrate, where prior challenges in wafer-scale synthesis have been solved by using H2Se as the selenium precursor with a significant amount of excess chalcogen to obtain epitaxial films (Figs. S1 and S2, supplementary material). Using an aberration-corrected transmission electron microscope (TEM), we estimate the film thickness was on the order of 5–15 layers. The epitaxial WSe2 was then transferred onto the tips using a wet transfer technique (see the supplementary material for additional details).
In order to determine the underlying carrier relaxation dynamics and single photon generation characteristics of WSe2, ultrafast optical characterization was performed using a Ti:Sapphire femtosecond laser for time resolved photoluminescence (TRPL), and Hanbury Brown-Twiss (HBT) interferometry was performed using a HeNe continuous wave laser with spectral windows defined by inserting band-pass and tunable long- and short-pass filters into the beam path. A diagram of the beam paths for both experiments is given in Fig. S3, supplementary material. Figures 2 and S4, supplementary material, give the spectroscopic characterization results of optical emission arising from WSe2 on the apex of an ultrasharp SiO2 tip. An additional comparison of the emission spectra from the unstrained material and from the strained material at the tip apex, along with representative photoluminescence (PL) from nonquenching tip emitters is given in Figs. 3 and S5, supplementary material, where the free exciton emission at the tip apex exhibits a blue shift of 22.1 ± 3.4 nm (47.3 ± 8.3 meV) compared to the unstrained free exciton emission peaks. This is 1.2–8.5 times larger than that observed in tent-pole-based studies,15,16 although we note that since the elastic deformation is tensile at the upper WSe2 surface and compressive at the lower surface, quantitative deconvolution of the strain-dependent emission spectra as per Ref. 33 is not trivial. Our gas source CVD WSe2 films contain grains smaller than the 0.7–1 μm excitation beam diameter, thus the contribution of grain boundary defects to quantum emission is an additional remaining unknown in this field. As the TRPL intensity I(Δt) at any defect emission wavelength contains contributions from both free and localized (bound) excitons, we implement a bi-exponential rise and decay model to gain insights into the lifetimes of bound and free excitons as
where Ibkgd is the background count rate, the subscript j indicates the fast and slow processes which we attribute to free and localized excitons, respectively, the subscript i is the peak index, I0,j is the exciton emission intensity, Δt is the time delay, T is the period, and τrise,j and τdecay,j are the characteristic lifetimes of excitonic rise and decay, respectively. To minimize the number of adjustable parameters used in our model, all TRPL data were fit with a fully unconstrained model to obtain T = 12.4415 ± 0.0008 ns and τrise,fast = τrise,slow = 7.95 ± 0.71 ps. As the obtained period uncertainty and rise time are well below the detection limit of the silicon avalanche photodiode used in the time resolved measurements, we do not ascribe physical significance to these quantities only to state that T and τrise are sufficiently consistent between measurements, so as to hold these parameters fixed at the listed values during data analysis. Subsequently, the model was constrained to allow only I0,j and τdecay,j as adjustable parameters. Ibkgd was dark count limited at ∼80–100 counts s−1 which was more than 50 times below the TRPL count rate. Subsequently, the model was constrained to allow only fast and slow components of I0,j and τdecay,j as adjustable parameters. Power dependence of the emission I0,j can be analyzed using a saturation model
where P is the excitation power, I0,j(P = ∞) is the saturation intensity, Phalf-sat is the half-saturation power, and j = fast and slow relaxation processes. Figure 2(b) demonstrates that excitation power dependences of the deconvolved emission intensities exhibit (i) a nearly linear trend for the delocalized transition (free exciton) with I∞,fast = 16.4 ± 20.0 events s−1 and Phalf-sat,fast = 32.1± 44.9 μW (large uncertainty is indicative of linear dependence on the excitation power) and (ii) emitters located on the top of the ultrasharp tips (bound exciton) exhibit saturation behavior with I∞,slow = 5.89 ± 0.71 events s−1 and Phalf-sat,slow = 5.36 ± 1.16 μW. Although the emission intensity for the localized emitter is not fully saturated at the maximum power used in this study, we find that at higher powers defect emission from the sharp tips can spontaneously quench, and hence we limit excitation to the low power regime in this study.
Time resolved photoluminescence (TRPL) and photon field intensity correlation demonstrate spatially localized quantum emission at the apex of an ultrasharp SiO2 tip. (a) TRPL spectra of localized emission from WSe2 on an ultrasharp SiO2 tip obtained using femtosecond excitation at 540 nm and 5.5 μW. Equation (3) is used to deconvolve bound (slow decay, dashed line) and free (fast decay, dotted line) exciton contributions to the measured event count. (b) Bound (slow decay, circles) and free (fast decay, triangles) exciton intensities (I0, black) and decay times (τdecay, red) obtained from TRPL modeled by Eq. (3), where error bars are defined by the 99% confidence intervals. Intensity was characterized using a saturation model given by Eq. (4) (slow decay, dashed line and fast decay, dotted line). (c) Measured second order photon correlation g(2)(Δt) as a function of time delay Δt and modeled according to Eq. (5).
Time resolved photoluminescence (TRPL) and photon field intensity correlation demonstrate spatially localized quantum emission at the apex of an ultrasharp SiO2 tip. (a) TRPL spectra of localized emission from WSe2 on an ultrasharp SiO2 tip obtained using femtosecond excitation at 540 nm and 5.5 μW. Equation (3) is used to deconvolve bound (slow decay, dashed line) and free (fast decay, dotted line) exciton contributions to the measured event count. (b) Bound (slow decay, circles) and free (fast decay, triangles) exciton intensities (I0, black) and decay times (τdecay, red) obtained from TRPL modeled by Eq. (3), where error bars are defined by the 99% confidence intervals. Intensity was characterized using a saturation model given by Eq. (4) (slow decay, dashed line and fast decay, dotted line). (c) Measured second order photon correlation g(2)(Δt) as a function of time delay Δt and modeled according to Eq. (5).
Photoluminescence (PL) reveals the inherent differences between “intrinsic” and “engineered” emission sites. (a) High resolution pulsed excitation PL of defect emission in WSe2 occurring intrinsically in the material (tan) and at the apex of ultrasharp SiO2 tips (blue, light blue). (b) TRPL of intrinsic [tan, corresponding to the left panel in (a)] and strain-engineered emitters [blue, corresponding to the center panel in (a), light blue corresponding to right panel in (a)]. Control over the spatial positioning of localized emission sites is demonstrated through the (c) optical micrograph and (d) scanning continuous wave excitation PL of epitaxial WSe2 transferred onto an ultrasharp SiO2 tip, where the integrated area of the localized exciton peak is shown vs x-y coordinate.
Photoluminescence (PL) reveals the inherent differences between “intrinsic” and “engineered” emission sites. (a) High resolution pulsed excitation PL of defect emission in WSe2 occurring intrinsically in the material (tan) and at the apex of ultrasharp SiO2 tips (blue, light blue). (b) TRPL of intrinsic [tan, corresponding to the left panel in (a)] and strain-engineered emitters [blue, corresponding to the center panel in (a), light blue corresponding to right panel in (a)]. Control over the spatial positioning of localized emission sites is demonstrated through the (c) optical micrograph and (d) scanning continuous wave excitation PL of epitaxial WSe2 transferred onto an ultrasharp SiO2 tip, where the integrated area of the localized exciton peak is shown vs x-y coordinate.
Figure 2(c) demonstrates the quantum emission of a defect emission site at the apex of an ultrasharp SiO2 tip at ∼3.8 K through the time-dependent photon field intensity correlation, g(2)(Δt). The data can be modeled using a single exponential decay (two level) photon antibunching model22
where Δt is the time delay, τdecay is the lifetime, g(2)(Δt = 0) is the second order photon correlation parameter for single photon emission, and [1 − g(2)(Δt = 0)] is defined as the single photon purity. A high degree of photon antibunching was obtained, g(2)(Δt = 0) < 0.3 over a collection time of 45 min, and exhibited stable emission up to 8 h. The g(2) spectra are normalized at far from zero (240.32 ns ≤ Δt≤ 2097.44 ns). Stability in emission intensity allowed us to obtain the intrinsic g(2)(Δt = 0) = 0.284 ± 0.062 and an intrinsic τdecay = 9.01± 1.56 ns from collections over three different times from 45 min to 8 h. We note that although the obtained τdecay is in agreement with that obtained by TRPL, HBT measurements of lifetime are heavily influenced by the excitation power used in the measurement18 making TRPL the appropriate technique for lifetime determination.
By comparing the TRPL of WSe2 defect emission on the tips with a WSe2 defect on the substrate, we were able to decouple the intrinsic defects in the material from defects arising through engineered morphology which may elucidate the characteristics of strain induced emission [Figs. 3(a), 3(b), and S5, supplementary material). In comparison with engineered emitters where I0,slow > I0,fast, the intrinsic localized emission site we located on the substrate exhibited a slow-component emission intensity 2.9 times lower than that of the fast component, indicating the dominance of the PL by the free exciton for this intrinsic defect. TRPL also revealed that the decay time for localized excitons at the apex of three representative sharp tips was 12.75, 19.47, and 68 times longer than that of the free exciton, with τdecay,slow = 11.203 ± 0.660, 15.58 ± 0.50, and 56.40 ± 7.71 ns and τdecay,fast = 0.800 ± 0.040, 0.828 ± 0.059, and 0.879 ± 0.047 ns for the emitters shown in Figs. 2(a) and 3(b). This is in comparison to the case for localized defect emission from WSe2 on the substrate for which τdecay,slow (2.36 ± 0.24 ns) was only 4.8 times longer than τdecay,fast (0.516 ± 0.029 ns). The slower relaxation times for both the free and localized excitons on the tips may also be due to phonon-mediated dark state recombination—as recently modeled for SeW antisite defects18—or changes in the density of states and optical phonon energies—where phonon softening may proportionately increase the lifetime38—although understanding the exact mechanism requires rigorous experimental and theoretical treatment in future works. The time window used in our experiment was on the order of 12.5 ns for TRPL, which was limited by the repetition rate of the laser. This explains why the longest decay time possessed a large error, and is thus more qualitative in nature. We note that previous reports also indicate a wide range of decay times, roughly 0.5 to 225 ns (Table S1, supplementary material), and thus a major remaining challenge in this field is understanding the mechanisms responsible for this especially from the theoretical side. In an attempt to more accurately quantify defects intrinsically present in the epitaxial WSe2 of this work, we have conducted Rutherford backscattering spectrometry (RBS). RBS analysis (Fig. S5, supplementary material) allowed us to obtain the elemental atomic ratios of the material in this study. Analyzing the scattering yield ratios between W and Se, we obtained a W:Se ratio of 1:1.91 ± 0.04 corresponding to a selenium deficiency of ∼9%. This is considerably higher than the ∼2.2% selenium deficiency obtained for crystallites synthesized through powder vaporization,39 which may lead to the variation in relaxation times we observe in this work. Confocal photoluminescence mapping around an ultrasharp tip is shown in Figs. 3(c) and 3(d) and indicates defect emission that originates at the tip center.
Figure 4 gives the temperature dependent emission spectra of a localized defect emission site. Over the temperature range of 3.8–25 K, the emission energy remains relatively constant with a mean and standard deviation of 772.953 nm (1.604 eV) and 0.830 nm (0.00172 eV), respectively. This is in agreement with the power dependent peak emission wavelength we observed, 772.88 ± 0.27 nm (1.604 ± 0.001 eV) as shown in Fig. S7, supplementary material, and the spectral wandering demonstrated in previous reports.14,15 Similar to previous studies,7–11,14–16,20 localized emission is not observable at temperatures above 25 K. Figure 4(b) shows that the temperature dependent emission intensity diminishes exponentially and can be modeled as I(T) = I1·exp[−T/Tc], with a characteristic temperature of Tc = 5.98 ± 0.24 K (corresponding to a kBT energy of 515 μeV, where kB is the Boltzmann constant). This is in agreement with the value obtained for quantum emission from monolayer WSe2 (∼300 μeV).11 Additionally, we observe that the full width at half maximum (FWHM) increases exponentially from 1.42 nm (2.94 meV) at 3.8 K to 2.48 nm (5.15 meV) at 25 K, and can be modeled as Γ(T) = Γ0 + Γ1·exp[−T/Tc], with a characteristic temperature of Tc = 5.4 ± 0.3 K and Γ0 = 1.41 ± 0.01 nm (2.91 ± 0.03 meV). We note that increasing the defect trapping energy, so that room temperature operation can be demonstrated, may require the use of chemical-functionalization or solitary dopants as has recently been demonstrated for carbon nanotubes,25,40 along with novel optical engineering approaches.17
Temperature dependent characteristics of localized emission. (a) Pulsed excitation PL spectra of localized emission from WSe2 on an ultrasharp SiO2 tip as a function of temperature. (a, inset) Peak emission wavelength over the measured temperature range exhibits a mean of 772.95 nm and a standard deviation of 0.83 nm. (b) Peak emission intensity, I, and full width at half maximum (FWHM), Γ, vs temperature of the localized emitter. I and Γ can be modeled by exponential decay and growth models, respectively.
Temperature dependent characteristics of localized emission. (a) Pulsed excitation PL spectra of localized emission from WSe2 on an ultrasharp SiO2 tip as a function of temperature. (a, inset) Peak emission wavelength over the measured temperature range exhibits a mean of 772.95 nm and a standard deviation of 0.83 nm. (b) Peak emission intensity, I, and full width at half maximum (FWHM), Γ, vs temperature of the localized emitter. I and Γ can be modeled by exponential decay and growth models, respectively.
In conclusion, we have demonstrated a route to deterministically create spatially localized defect emission sites in the 750–800 nm regime using few-layer epitaxial WSe2 and ultrasharp SiO2 tips, where a detailed analysis of a characteristic tip proved single photon generation with g(2)(Δt = 0) = 0.284 ± 0.062. The exciton lifetime increases from ∼1 to 2 ns for a defect lying in the planar WSe2 to ∼10 ns when the defect is located on a tip apex; although this is still too small to be relevant for quantum memory applications, it is an advance in the right direction. Engineering of quantum emission in 2D materials is still in a very early stage, and further efforts are required to increase both the yield of quantum emitters and the thermal detrapping energy for higher temperatures and brighter operation. Future large-scale statistical investigations are also required to better understand defect emission including the dark state, fine structure, and polarization characteristics in the 2D material/ultrasharp tip array system. As the atomic structure of defects responsible for localized emission is not observable in this or prior works, sustained in-depth investigations of quantum emission in 2D materials are required both from the experimental and theoretical communities.
See the supplementary material for additional comparison with existing literature, materials, methods, defect spectroscopy, and RBS results.
Fabrication, confocal spectroscopy, and modeling by M.T.P., W.W., and R.D.M. were supported by the U. S. National Science Foundation Award Nos. CAREER-1553987 (M.T.P., W.W.) and REU-1560098 (M.T.P., R.D.M.), and a FEI Company Graduate Fellowship (W.W.); ion beam analysis by Y.W. and transmission electron microscopy by M.T.P. were performed under user Grant No. 2018AU0058 at the Center for Integrated Nanotechnologies, an Office of Science User Facility operated for the U. S. Department of Energy (DOE) Office of Science, and the Laboratory Directed Research and Development Program of Los Alamos National Laboratory under Project No. 20190516ECR (M.T.P.). Los Alamos National Laboratory, an affirmative action equal opportunity employer, is managed by Triad National Security, LLC for the U. S. Department of Energy's NNSA, under Contract No. 89233218CNA000001; epitaxial WSe2 was synthesized by X.Z., T.H.C., and J.M.R. at The Pennsylvania State University Two-Dimensional Crystal Consortium – Materials Innovation Platform (No. 2DCC-MIP) which is supported by NSF Cooperative Agreement No. DMR-1539916; time resolved photoluminescence and photon correlation spectroscopy performed by C.K.D., J.R.H., M.T.P., and R.D.M. at the U. S. AFRL was supported by the Air Force Office of Scientific Research (Program Manager Dr. Gernot Pomrenke) under Contract No. FA9550-15RYCOR159 (J.R.H., C.K.D.). The project was conceived and led by M.T.P., and written by M.T.P. with the assistance of W.W., C.K.D., J.R.H., and contributions from all co-authors.