Identifying Carbon as the Source of Visible Single Photon Emission from Hexagonal Boron Nitride

Single photon emitters (SPEs) in hexagonal boron nitride (hBN) have garnered significant attention over the last few years due to their superior optical properties. However, despite the vast range of experimental results and theoretical calculations, the defect structure responsible for the observed emission has remained elusive. Here, by controlling the incorporation of impurities into hBN and by comparing various bottom up synthesis methods, we provide direct evidence that the visible SPEs are carbon related. We also perform ion implantation experiments and confirm that only carbon implantation creates SPEs in the visible spectral range. In addition, we report room-temperature optically detected magnetic resonance from an ensemble of carbon-based defects, confirming their potential deployment as spin-qubit systems. We discuss the most probable carbon-based defect configurations and suggest that the CBVN is the most suitable candidate. Our results resolve a long-standing debate about the origin of single emitters at the visible range in hBN and will be key to deterministic engineering of these defects for quantum photonic devices.

Part of the challenge stems from standard hBN bulk crystal synthesis via high pressure high temperature not being amenable to the deterministic control of impurity incorporation. This is aggravated by the induced impurities often segregating and forming regions of inhomogeneous defect concentration. 24 In addition, the two-dimensional, layered nature of hBN makes ion implantation difficult to control. 25 These limitations have precluded identifying the exact origin of the single photon emission in the material. Here, we address this problem.
We carry out a detailed study surveying various hBN samples grown in different laboratories by metal-organic vapor phase epitaxy (MOVPE) and molecular beam epitaxy (MBE). We find compelling evidence that to observe photoluminescence from SPEs the inclusion of carbon atoms in hBN is required.
By systematically growing samples with different carbon concentrations, we show that the carbon content determines whether the photoluminescence signal originates from ensemble of emitters (high carbon concentration) or isolated defects (low carbon concentration).
We carry out multi-species ion-implantation experiments on MOVPE films and rule out the possibility that the emitters are associated to native vacancy complexes; quantum emitters in the visible spectral range (the most commonly studied) appear only when carbon is involved strongly indicating they are carbon related. Finally, we demonstrate that these carbon-based emitters display room-temperature optically detected magnetic resonance (ODMR) confirming their potential use as addressable spin-qubits in emerging quantum technologies. Table 1 summarizes the materials we analyzed. They are epitaxial hBN samples grown by different methods and under various conditions. The rationale was to understand whether the single defects are intrinsic to hBN (e.g. substitutional or interstitial nitrogen or boron complexes) or they involve foreign atoms (e.g. carbon). We investigated hBN samples grown by four methods. (1) Metal organic vapor phase epitaxy (MOVPE) with varying flow rates of the precursor triethyl boron (TEB)-a parameter known to systematically alter the levels of incorporated carbon. (2) Molecular beam epitaxy (MBE) on sapphire with and without a source of carbon. (3) High-temperature MBE on SiC with a varying orientation of the Si face to explore the possibility of carbon incorporation occurring from the substrate. (4) Growth by the conversion of highly oriented pyrolytic graphite (HOPG) into hBN. Note, that in the current work we focus on bottom up growth of hBN as it offers an opportunity for large (centimeter) scale films of desired thickness (down to ~1 nm), as well as better control over the inclusions of impurities. 26 The table also includes additional growth details, sample characteristics, shorthand names used throughout the rest of the paper, as well as general photoluminescence properties. Expanded growth details of each sample are included in the experimental section. Notably, the fourth column of the table indicates a direct correlation between the appearance of the visible emission and the presence of carbon in the hBN films. This is discussed in detail below. We first explore the photoluminescence (PL) from a series of hBN samples grown by MOVPE as the triethyl boron (TEB) flow rate is increased and the ammonia flow is kept constant (Fig. 1a). 27 The aim of this measurement is to engineer an ensemble of hBN emitters, and to compare their line shape and linewidth with isolated SPEs grown using the same growth technique. A region of the TEB 10 (µmol/min) sample with the lowest percentage of carbon shows negligible fluorescence. Increasing the flow rate to TEB 20 is accompanied by the appearance of a bright fluorescence signal with two clear peaks appearing at ~585 nm and ~635 nm. Further increasing the flow rate to TEB 30 and 60 provides a similarly structured PL signature, with higher fluorescence intensity. Moderate fluctuations in the peak positions, and intensity ratio of the 585 nm and 635 nm peaks at different sample locations are consistent with emission from dense ensembles of hBN emitters. This also confirms previous findings showing that hBN emitters possess zero-phonon line (ZPL) wavelengths clustered at ~585 nm when the sample is grown epitaxially. 28,29 The separation in energy between the ZPL of the ensemble and phonon sideband (PSB) peak is ~176 meV on average. 30,31 In specific examples, we can resolve the energy detuning between the ZPL and each of the two-known longitudinal-optical (LO) phonon modes, LO1 ~165 meV and LO2 ~195 meV (Fig. S1), matching values reported in the literature. 30,31 The confocal scans for the samples MOVPE TEB 20/30/60 ( Fig. S2) display consistent fluorescence intensity across the sample surface, indicative of homogeneous incorporation of ensembles of SPEs.
Note that higher PL intensity correlates directly with higher TEB flux. Previous studies on these samples have confirmed that with sufficient slow growth kinetics (i.e. TEB 10), the excess carbon atoms are largely removed in the ammonia atmosphere leaving a very low percentage of carbon incorporation during growth. 27 When the precursor flux increases, the layer-by-layer growth proceeds faster than the process of etching away the excess carbon-resulting in a relatively high incorporation of carbon in the hBN matrix. This is confirmed by analyzing the relative intensity of the peaks associated to C-B bonds in the C1s region of the X-ray photoelectron spectroscopy (XPS), Figure S3. A nearly linear correlation between C-B bonding and TEB flux is observed (Fig.  1b) from <0.5% for TEB 10 to ~2.5% for TEB 60, reproduced from ref [23]. We also analyze the relative concentration of C-N bonding from the samples, Figure 1c, showing a similar near linear trend in C-N bonding with increased TEB flux during growth. The clear preference for the formation of C-B bonds as opposed to C-N bonds follows logically from noting the B species are introduced with three pre-existing bonds to C. Figure S4 shows the average intensity of ensemble emission vs. the percentage of carbon bonding to both B and N via XPS, confirming a near perfectly linear trend for TEB 10, 20, 30. We attribute the slight deviation from this trend (i.e. a decrease in PL signal) observed for the TEB 60 sample to increased disorder in the material-which has been reported to quench photoemission from SPEs due to non-radiative decay pathways. 32 It should be noted that previous structural analysis of hBN samples synthesized with increasing TEB flow rates concluded that the layered nature of the material is preserved, which confirms the material is C-doped hBN rather than a hybrid BNC material. 27 Based on these results, we advance that the SPE emission at ~580 nm in hBN is likely to originate from a carbonrelated defect complex.
We next employ a lab-built confocal PL setup with a 532 nm excitation source, to study in detail the TEB 10 sample. The level of carbon doping is low enough that we can isolate single quantum emitters; a representative spectrum for one such emitter is shown in Figure 1d. The quantum nature of the emission was confirmed by measuring the second order auto-correlation function; the value of g (2) (τ = 0) < 0.5 ( Fig. 1e) is conventionally attributed to a single photon source with sub-Poissonian emission statistics. We measured the zero-phonon line (ZPL) wavelength of 77 SPEs in the MOVPE hBN (TEB 10) sample, Figure S5, finding that ~78% of the emitters are located at (585 ± 10) nm, and 95% at wavelengths < 600 nm. This matches the distribution of ZPL wavelengths reported previously for CVD hBN grown on copper foils (reproduced in Figure S5). 28 The typical line shape of these emitters at room temperature is also consistent with previous studies, including the ZPL and a PSB centered at ~177 meV from the ZPL energy. 30 This suggests that when the carbon concentration is sufficiently low, individual quantum emitters can be isolated. Their optical properties and spectral distribution are consistent with those observed in samples with higher carbon doping, with the difference merely being due to the density of emitters. This reinforces the hypothesis of a common-carbon-related-structural origin of these defects (see discussion below).  To further confirm that carbon-based defects are responsible for SPE emission from hBN we analyze a series of hBN samples grown by a different method, MBE. 33 Figure 2a displays the PL spectrum observed from undoped MBE hBN grown on sapphire substrate using an e-beam evaporator to supply boron from a boron nitride crucible. The resulting PL signal was relatively low; no SPEs could be found despite the material being of good quality as shown by a clear hBN E2g Raman line. Interestingly however, when the elemental boron precursor was placed inside a carbon crucible-with otherwise identical growth conditions-we observed the appearance of sharp spectral lines, shown in figure 2b. The carbon crucible used for e-beam thermalization of the boron source shows clear signs of sidewall etching, which suggests that carbon was present in the gas phase during growth.
In contrast to the undoped MBE hBN on sapphire showing only the hBN Raman (~573.7 nm) and the Cr in sapphire line (~694 nm), we find clear emission lines, corresponding to hBN SPEs. The SPEs are incorporated at a high density, up to 5 per confocal spot (~400 nm). Due to the high density of emitters incorporated, we could not unambiguously confirm their quantum nature via second-order autocorrelation measurements. We instead probed the polarization dependence of the emission by placing a polarizer in the collection path. Figure 2c shows one such collection, where emission from a presumed ZPL at ~577 nm is linearly polarized, with the PL intensity dropping when the polarizer is perpendicular to the polarization direction of the probed emitter.
We recorded the ZPL locations of 65 different emitters from the carbon doped MBE hBN on sapphire, Figure S5. The ZPL wavelength spectral distribution of the samples grown by MBE is approximately ~31% <600 nm, ~55% in the 600-700 nm range, and ~14% >700 nm. This is in contrast to the ZPLs observed in the CVD and MOVPE samples discussed above and for which ~95% of the ZPL wavelengths is <600 nm. A possible explanation is that in the MOVPE sample only one type of carbon-based defect is formed while in the MBE sample different defects do form. The observation of the extended ZPL range in the MBE sample (appearing only in the presence of carbon) suggests that the entire range of visible SPEs is due to carbon-related defects: either multiple defects (e.g. carbon-related complexes) or a single structural defect with variations in the ZPL energy due to local strain and Stark effects. 9, 10 Further discussion on the possible structure for different carbon defects is included below.
We next explored MBE growth of hBN on silicon carbide (SiC), investigating different crystal orientations: specifically, with the top Si face-on (0˚) and slightly off (8˚). All hBN films grown on SiC utilized a Knudsen high temperature source for boron evaporation, at a growth temperature of 1390 ˚C, and identical conditions other than the substrates. Representative spectra from both sample types (Si at 0˚ and at 8˚) are displayed in Figure 2d. When growth was performed with the Si face at 0˚, only a single SPE peak was located across a 40 µm 2 scan. In contrast, when the Si face is oriented at 8˚ we found a number of SPE peaks across the sample, often with a number of different ZPL wavelengths appearing within the same confocal spot-similarly to the carbon doped MBE hBN on sapphire, yet at a lower density.
These results further support the role of carbon in the origin of hBN SPEs in the visible spectral range. We attribute the incorporation of these SPEs during hBN growth on SiC to carbon diffusion from the substrate. At the growth temperature of 1390 ˚C, some sublimation of Si from the surface of the SiC substrates is expected, with the subsequent formation of an extra carbon layer on the surface of SiC. 34 While these temperatures are sufficient to sublime Si, they are not sufficient to evaporate C from the SiC surface. 34 Note that these temperatures are close to those used for graphene formation by SiC annealing, 34 suggesting that diffusion of carbon provides an available source for incorporation into the growing hBN layer. Interestingly, C incorporation into hBN appears significantly enhanced when the Si face is oriented 8˚ out of plane. Figure S5 shows the distribution of 26 ZPL wavelengths from the MBE hBN on SiC (Si 8˚) sample. Similarly, to the carbon doped MBE on sapphire we found that the ZPL positions were evenly distributed across the visible range: 35% <600 nm, 46% in the range 600-700 nm, and 19% >700 nm.
Finally, we analyze a third technique for hBN growth, the conversion of HOPG to hBN. 35 The process yields high quality hBN layers at the surface of the HOPG crystal, as confirmed by Raman spectroscopy (Figure 2e), with the hBN E2g mode located at 1366.9 cm -1 with a FWHM of 21.8 cm -1 . 36 The conversion from graphite, proceeding via atomic substitutions, provides a high availability of carbon defects in the resulting hBN conversion layer. A typical PL spectrum observed from the HOPG to hBN conversion sample is shown in figure 2f. We observe a structured emission with ZPL and PSB peaks reminiscent of that from emitters ensembles in the high carbon MOVPE samples. In certain confocal spots, we observed extremely bright SPEs; they undergo intermittent blinking events. Figure 2f displays such an example, while at other spots a broader ZPL peak appeared. As with the high carbon doped MOVPE samples, we found the energy detuning between the ZPL and first and second order PSBs to be consistent with hBN SPEs. 30,31 We now turn our attention to using ion implantation for defect incorporation, in an attempt to confirm the role of carbon. We performed a series of implantation experiments (dose: 10 13 ions/cm -2 , energy 10 keV) with carbon as well as silicon and oxygen used as controls to rule out the possibility for the photoemission to be due to native vacancy defects. The implantation experiments were performed on the MOVPE hBN (TEB 10) films (film thickness ~40 nm, Figure  S6) so to compare the relevant results to those for the samples synthesized while increasing carbon content during growth. Figure 3a shows the confocal scan of an MOVPE hBN (TEB 10) film after carbon implantation, but prior to annealing, where a TEM grid with 50 µm 2 square apertures was used as a mask. The implanted region is labelled I, while the masked region is labelled II. Figure 3b displays spectra collected from emitters within the implanted region (I), and a representative g (2) (τ = 0) < 0.5, confirming the quantum nature of the emission from these centers. Figure 3c displays a representative emitter from the masked region (II), showing the typical line shape of the ZPL and the PSB peaks found in MOVPE hBN (TEB 10) films, with the corresponding g (2) (τ = 0) shown to the right.  (2) ( ) is shown to the right. c. Spectra and g (2) ( ) from the masked area (II) display borderer ZPLs, and prominent PSBs consistent with that observed in Figures 1f and 1g. d. Confocal scan of the carbon implanted sample, post annealing, where areas marked (I) and (II) were implanted, while area (III) was masked. e. A representative spectra and g (2) ( ) from area (I), showing an ensemble of hBN emitters, and a corresponding g (2) ( ) measurement showing no dip as expected for ensemble emission. f. A representative spectra and g (2) ( ) from area (II) showing single SPE, but with significant spectral contributions from nearby SPEs resulting in a g (2) ( ) value of ~ 0.75 consistent. g. A representative spectra and g (2) ( ) from the masked area (III) post annealing showing a well-resolved SPE and PSB and a g (2) ( ) confirming a single emission center. h. Typical spectra observed in the oxygen-implanted region pre-annealing, showing only background emission the VBpeak ~ 800 nm are observed. i. Characteristic spectrum from oxygen implanted region post annealing, with the only spectral signature observed is a broad peak at ~ 630 nm. j. Typical spectra observed in the siliconimplanted region pre-annealing, showing only background emission the VBpeak ~ 800 nm are observed. k. Characteristic spectrum from silicon implanted region post annealing, with the only spectral signature observed is a broad peak at ~ 630 nm.
Inside the C-implanted region, most emitters (~ 80%) display narrow ZPL peaks (~5 nm FWHM) and extremely weak PSBs, compared to the typical ZPL/PSB found in these MOVPE hBN (TEB 10) films (Figs. 3c and 2a). As shown in Figure S7, no SPEs were found in the non-implanted MOVPE hBN (TEB 10) films with linewidths below ~8 nm, demonstrating the unique nature of these emission lines. The remaining ~20% of the emitters within the implanted region display similar line shapes (typically ~20 nm FWHM) and phonon coupling to those for the emitter in figure 3c, consistent with previous studies and demonstrating that SPEs with energies ~2.2 eV tend to possess broader lines and increased phonon coupling. 19 Our results indicate that the sharp emission lines belong to SPEs created via implantation of C ions. The reasons for the observed narrow line shape and the minimal phonon coupling are not clear and require further study. Nevertheless, our results suggest definitively that carbon implantation creates new color centers in hBN and strongly supports our hypothesis.
After characterizing the implanted hBN films, the samples were then annealed in high vacuum (1000 ˚C, <10 -6 Torr, 2 hours), and the same set of measurements was performed. As shown in figure 3d, the implanted regions are still visible, they however show variations in PL intensity. This effect is likely due to ion scattering around the mask edges and vacancy diffusion-which have been observed for implantation in diamond. 37 The PL spectra from three different areas are shown in figure 3e-f, and correspond to (I) the implanted region of high PL intensity, (II) the implanted region of lower PL intensity, and (III) the masked region of the film. Figure 3e displays a representative spectrum from inside region I, where we found emission characterized by broad ZPL and PSB profiles similar to those observed in the high TEB flux growths. This emission is confirmed to be due to an ensemble of SPEs as the corresponding g (2) ( ) measurement shows no anti-bunching despite the associated ZPL/PSB structure. A similar spectral signature is observed consistently throughout region (I), again implying the creation of an ensemble of C-based SPEs. Figure 3f displays a representative spectrum from the implanted region II, where we again observe luminescence with a similar line shape. The overall ensemble signal remains homogeneous in this region, although appears less dense and bright, and a g (2) ( ) function shows a value of ~0.75, confirming the presence of fewer emitters within a confocal spot. Note that in both implanted areas (I and II) we no longer find the narrow emission lines with low phonon coupling found prior to annealing. Finally, figure 3g displays a representative spectrum from region III (masked area), showing a typical ZPL and PSB profile and a g (2) ( ) < 0.5, which confirms the quantum nature of the emission. As expected, this is similar to the pattern observed for the unimplanted samples.
We next analyze the results of identical implantation experiments performed with oxygen and silicon. As with carbon implantation, for both Si and O the implanted regions display an increased intensity prior to annealing. Figure 3h displays the typical spectrum from the implanted region of the oxygen-irradiated samples, prior to annealing. Only two notable spectral features are observed, the silicon Raman line from the underlying substrate, and a broad peak spanning over the spectral range ~720-820 nm. Note that the peak artifacts above 800 nm are due to the optics used in our confocal setup which are optimized for the visible region. Recent work has identified this broad emission at ~800 nm as the negatively charged boron vacancy (VB -). 16 Importantly, the observation of the VBemission spectra which is relatively strong and homogeneously distributed throughout the implanted region, confirms a high degree of vacancy creation upon implantation, consistent with our SRIM simulations, as shown in Figure S8. No evidence of sharp emission lines similar to that from carbon implantation are found, and almost no SPEs were located within the 50-µm 2 area, suggesting that potentially pre-existing SPEs were largely destroyed by oxygen implantation. Figure 3i displays the typical spectrum observed from the oxygen implanted area after high temperature annealing. The implanted region displays a broad signal centered at ~630nm and shows no evidence of the ensemble emission at ~585 nm observed with high carbon doping during growths or carbon implantation. Additionally, the lack of a PSB or other features consistent with an ensemble of hBN SPEs allows us to exclude that this emission may originate from an ensemble of the carbon-based SPEs centered at ~630 nm.
The silicon implanted MOVPE hBN (TEB 10) samples display nearly identical fluorescence properties within the implanted regions as the oxygen implanted samples. Figures 3j and 3k display the results for before and after annealing, with similar formation of (VB -) before annealing and broad emission in the visible spectral region after. The lack of formation of the visible SPEs (or ensemble of SPEs) upon Si and O implantation experiments leads to ruling out native vacancyrelated defects as the source of the sharp SPEs. Additionally, we can rule out the possibility that carbon incorporation plays a secondary role in the increased emission intensity from MOVPE hBN by means of stabilizing or modifying the charge state of other defects in the material-as previously suggested for the trend of increasing intensity of electron paramagnetic resonance signals from hBN. 38 The compiled evidence indicates in a compelling manner that the structural nature of hBN SPEs emitting in the visible is directly associated with carbon-based defect(s).
Considering our results on the creation of carbon-based defects by multiple methods, we deem important commenting on previous reports in the literature aimed at deliberately creating emitters in hBN. A number of methods have been employed: high temperature annealing, 19, 39 plasma treatment, 40 strain activation, 41 electron beam irradiation, 19,42 ion implantation (although never involving carbon), 25 and focused ion beam (FIB) irradiation. 43 The results from these experiments have been largely inconclusive and, at times, conflicting. Considered in their entirety, previous results largely suggest that preexisting emitters are activated rather than created in hBN. 39 This is plausible as even the highest quality hBN material, grown by HPHT precipitation, is known to incorporate carbon. 24 Inspired by the recent work showing that hBN defects emitting in the visible region can exhibit optically detected magnetic resonance (ODMR), 17 we performed such a measurement on ensembles of hBN emitters on our MOVPE hBN (TEB 60). Figure 4a shows the ODMR spectra displayed as relative contrast, spin-dependent variation in photoluminescence (∆PL/PL) for different values of the static applied magnetic field. Note that the ODMR signal shows an increase (~0.5%) in emission intensity when the microwave excitation is resonant with the spin transition. Our results are in line with some preliminary reports suggesting that this ODMR signal is potentially associated with a carbon-related defect. 17 Importantly however, in previous experiments, ODMR was only observed at low temperature (8.5 K), while our measurements show that the ODMR contrast is resolvable at room temperature. By varying the static magnetic field B, we measure resonances at ~523, ~668.5 and ~815.4 MHz for B = 19, 24 and 29 mT, respectively. The near perfect linear trend suggests an isotropic g-factor for the transition (∆g = 0). Figure 4b displays the observed ODMR resonance frequencies as a function of the applied magnetic field, resulting in a slope ge of ~2.09. Unfortunately, these results do not allow us to be conclusive about the multiplicity of the spin sublevels we are probing, as our observation may be consistent with either a S=½ or S≥½ system with a small ground state splitting.
When comparing our room-temperature ODMR results from ensembles emitting at ~580 nm to that recently found at low temperature for a single or few peaks with a primary ZPL at ~727 nm, we find remarkable similarities. 17 Both display a single resonant peak suggesting either a spin-1/2 system or a spin-1 system with a small zero-field splitting. The peak width is narrow, ~30-35 MHz and we do not observe hyperfine splitting.

Figure 4-ODMR from MOVPE hBN (TEB 60) and Structural
Candidates. a. ODMR contrast observed from ~585 nm ensemble emission of MOPVE hBN (TEB 60) at applied fields of 19, 24, and 29 mT respectively. b. ODMR resonance frequencies as a function of applied magnetic field, fit with equation S1, with an extracted g value of ~ 2.09. c. Carbon based defect candidates for visible spectrum single photon emission from hBN. CBVN contains a carbon substituted for a boron and an associated nitrogen vacancy. CNVB contains a carbon substituted for nitrogen and a corresponding boron vacancy. CN is a carbon substituted for nitrogen, with no associated vacancy. Figure 4c displays three possible carbon-based defects we considered to explain the observed SPE emission. First, we consider the substitutional CN defect, which has been suggested in previous ODMR measurements. 17 The agreement between theory and experiment for this case requires a modification of the simulated hyperfine coupling parameter in order to obtain such narrow linewidths. 17 Additionally, recent calculations have not been able to produce results matching the spin-dependent transitions with the CN defect. 44 Furthermore, density functional theory (DFT) calculations have predicted a D3h symmetry for the defect, 21 which importantly possesses inversion symmetry. This is inconsistent with the growing body of literature demonstrating Stark tuning of emission lines, which is only possible for defects without inversion symmetry. 9,45,46,47 Finally, observations of large strain shifts in hBN, 10 and a linearly polarized optical dipole, 48 are inconsistent with a high symmetry defect like CN. As a result, the CN defect is an unlikely candidate for explaining our observations. We thus turn our attention to carbon-based defects which lack inversion symmetry, i.e. CBVN and CNVB.
In an attempt to differentiate between the two possibilities, we note that only the boron-vacancy (VB) is mobile under the annealing conditions (1000˚C) we used in our implantation studies, while the nitrogen-vacancy (VN) is not. 49 While this is only indirect evidence, it indicates that the CBVN defect is a more favorable candidate defect to explain our implantation results. Theoretical calculations on the CBVN defect in hBN have predicted multiple electronic levels within the band gap as well as ZPL transition energies compatible with our measurements. We note though, for completeness, that only the neutral form of the defect has been simulated. 21,50,51,52 To conclude, we have presented rigorous results to confirm the role carbon plays in the formation of hBN quantum emitters in the visible spectral range. We compared samples grown by three different methods: MOVPE, MBE, and HOPG conversion. They all exhibited a direct correlation between the introduction of carbon as a precursor/substance and the formation of SPEs in the visible spectral range. The bottom-up growth using MOVPE technique further enabled us to control the carbon incorporation and vary the density of the quantum emitters. We have successfully reproduced equivalent results using direct ion implantation of carbon. This, besides strengthening the hypothesis of carbon being directly involved in the formation of hBN color centers, also offer a promising avenue for direct engineering of emitters in hBN, on demand. Finally, we performed ODMR measurements on the ensembles of carbon related emitters, demonstrating room-temperature ODMR from this family of defect, for the first time. We put forward the CBVN defect as the most likely candidate for the hBN defect emitting in the visible spectral range studied in this work. Our results provide important insight into the ongoing effort towards understanding the origin of hBN color centers, while advancing potential strategies for the controlled engineering of these quantum emitters for nanophotonic devices. boron (TEB) and ammonia were used as the boron and nitrogen precursors, respectively, while hydrogen was the carrier gas. The precursors were introduced into the reactor as short alternating pulses, in order to minimize parasitic reactions between TEB and ammonia. hBN growth was carried out at a reduced pressure of 85 mBar and the growth temperature was set to 1350˚C. In the present study, the TEB flux was varied from 10 µmol/min to 60 µmol/min to study the effect on carbon incorporation on sub-bandgap luminescence from the hBN films. For ion implantation, PL and SPE measurements, cm-sized hBN films were transferred from sapphire on to SiO2/Si substrates, using water-assisted self-delamination. 27 Thickness of the hBN films was also measured using atomic force microscopy, as shown in the supplementary information. X-ray photoelectron spectroscopy was used for determining the impurity levels in the as-grown MOVPE-hBN films, as shown in the supplementary information. A gentle etching using Ar beam was performed in-situ to remove adventitious carbon and impurities from the surface; all spectra were collected from the bulk of hBN films. Molecular beam epitaxy. BN epilayers were grown using a custom-designed Veeco GENxplor MBE system capable of achieving growth temperatures as high as 1850 °C under ultra-high vacuum conditions, on rotating substrates with diameters of up to 3 inches. Details of the MBE growth have been previously published. 53 In all our studies, we relied on thermocouple readings to measure the growth temperature of the substrate. For all samples discussed in the current paper the growth temperature was in the range 1250-1390 o C. We used a high-temperature Knudsen effusion cell (Veeco) or electron beam evaporator (Dr. Eberl MBE-Komponenten GmbH) for evaporation of boron. High-purity (5 N) elemental boron contains the natural mixture of 11 B and 10 B isotopes. To have boron in the e-beam evaporator we used boron nitride and vitreous carbon crucibles. We used a standard Veeco RF plasma source to provide the active nitrogen flux. The hBN epilayers were grown using a fixed RF power of 550 W and a nitrogen (N2) flow rate of 2 sccm. We used 10 × 10 mm 2 (0001) sapphire and on-and 8 o -off oriented Si-face SiC substrates.
HOPG to hBN Conversion. The conversion takes place in a graphite crucible. A HOPG crystal is placed in the center of the crucible on a separate graphite holder. Small holes in the stage holding the HOPG allow vapors from the boron-oxide powder, placed at the bottom of the crucible, to flow to the HOPG crystal. A radio frequency induction furnace is then heated to 2000 ˚C, and N2 gas is introduced as the nitrogen precursor. A central tube mixing the nitrogen gas with the boron-oxide vapor pre-mixes the precursors prior to arriving at the HOPG crystal. Further details can be found here. 35 Ion Implantation. Ion implantation was carried out on 40 nm-thick MOVPE-hBN films, grown using a TEB flux of 10 µmol/min. For this, the hBN films were first transferred on to SiO2/Si substrates. A copper grid with a square mesh (GCu300, ProSciTech) was used as the implantation mask. Carbon, silicon and oxygen were separately implanted into the hBN films. During implantation, the ion energy and fluence were 10 keV and 10 13 ion/cm 2 , respectively.
Confocal Microscopy. PL studies were carried out using a lab-built scanning confocal microscope with continuous wave (CW) 532-nm laser (Gem 532, Laser Quantum Ltd.) as the excitation source. The laser was directed through a 532 nm line filter and a half-waveplate and focused onto the sample using a high numerical aperture (100×, NA = 0.9, Nikon) objective lens. Scanning was performed using an X−Y piezo fast steering mirror (FSM-300). The collected light was filtered using a 532-nm dichroic mirror (532 nm laser BrightLine, Semrock) and an additional 568-nm long pass filter (Semrock). The signal was then coupled into a graded-index multimode fiber (fiber aperture of 62.5 μm). A flipping mirror was used to direct the emission to a spectrometer (Acton Spectra Pro, Princeton Instrument Inc.) or to two avalanche photodiodes (Excelitas Technologies) in a Hanbury Brown-Twiss configuration, for spectroscopy and photon counting measurements, respectively. Correlation measurements were carried out using a time-correlated single photon counting module (PicoHarp 300, PicoQuant). All of the second-order autocorrelation g (2) (τ) measurements were analyzed and fitted without background correction unless specified otherwise.
ODMR. The ODMR spectra were measured with a confocal microscope setup. A 100× objective (Olympus MPLN100X) was used to focus a 532-nm laser (LaserQuantum opus 532) onto the sample and collect the PL signal. The PL signal is collected back through a 650-nm short-pass dichroic mirror for separation from scattered laser light. Additionally, a 532nm and 550nm longpass filter were used before the PL was detected by a silicon avalanche photodiode (Thorlabs APD440A) to filter out the laser light. The microwave field was applied through a signal generatorplus-amplifier system (Stanford Research Systems SG384 + VectaWave VBA1000-18 Amplifier); the sample was placed on a 0.5-mm-wide copper stripline. In order to detect the ODMR signal (i.e. the relative ∆PL/PL contrast) by lock-in technique (Signal Recovery 7230), the microwaves were driven with an on-off modulation. The resonant condition was changed with the external magnetic field by mounting a permanent magnet below the sample.