Prolonged photostability in hexagonal boron nitride quantum emitters

The study addresses why room-temperature single-photon emitters in hexagonal boron nitride (hBN) often show poor photostability—manifesting as spectral diffusion, blinking, and irreversible photobleaching—thereby hindering their deployment in quantum photonic technologies. While 2D hosts like hBN promise efficient light extraction and on-chip integration, emitters in CVD-grown hBN frequently bleach under continuous excitation. The authors focus on defining and quantifying the photobleaching mechanisms of a ubiquitous emission species (P3 at 1.98 ± 0.05 eV) across common hBN sample types, with the goal of identifying environmental and intrinsic contributions to bleaching and strategies to improve emitter photostability. The work is significant for engineering emitters with practical photostability for quantum applications.

Prior work established single-photon emission in 3D hosts (diamond, SiC, ZnO) but with extraction inefficiencies due to bulky dielectrics. 2D hosts (TMDs, hBN) mitigate this, with hBN uniquely offering bright room-temperature emission. hBN emitters have been realized via bulk crystal exfoliation (mechanical, liquid), and thin films grown by CVD and MOVPE. Wafer-scale single-crystal monolayer hBN via CVD has advanced, yet CVD emitters often exhibit spectral diffusion, blinking, and photobleaching; mechanisms remain elusive and limited stabilization strategies exist. Liquid exfoliated hBN can show sharp, stable emission after annealing, but yields nanoflakes with limited lateral size for applications. Emission distributions in visible hBN are discretized into species P1–P6; P3 (1.98 eV) and P2 (2.15 eV) are common. Theoretical and experimental studies implicate native point defects (including VB, VN), hydrogen termination, and carbon/oxygen incorporation in defect structure and emission, but correlations to photostability and bleaching processes, particularly in CVD vs liquid-exfoliated samples, have been unclear.

- Samples: Commercial monolayer CVD hBN on Cu (Grolltex), multilayer CVD hBN on Cu (Graphene Supermarket), and multilayer liquid-exfoliated hBN nanoflakes in ethanol/water (Graphene Supermarket). Transfers to SiO2/Si substrates used PMMA-assisted Cu etching and cleaning. A four-layer stack was made by successive monolayer transfers. - Optical measurements: Room-temperature confocal PL with CW 532 nm excitation, typical power density ~900 µW/µm² unless varied. A custom environmental chamber with quartz window enabled controlled atmospheres (N2, O2, saturated water vapor, air). Time-dependent PL spectra and intensity time-traces at fixed emission energies were recorded. For statistics, multiple emitter time-traces were summed to form overall decay curves. - Bleaching lifetime extraction: Fits used single-exponential I=I0 exp(-t/τ) and bi-exponential I=I0[r exp(-t/τ1)+(1−r) exp(-t/τ2)] models with 95% confidence intervals. Alternative dwell-time analyses cross-validated τ. Laser power dependence of bleaching rate 1/τ was measured across decades of power density. - Mechanistic modeling: Interpreted linear 1/τ vs power via a three-level scheme (ground singlet, excited singlet, triplet) allowing photobleaching from excited states; microscopic rate constants for excitation, radiative decay, internal conversion, intersystem crossing, triplet depopulation, and bleaching from S and T considered. - Stacking strategy: Constructed a four-layer hBN stack to test O2 shielding, analyzing the ratio r attributable to quickly bleaching top-layer vs protected layers and potential trapped O2 at interface under N2. - ADF-STEM: Monolayer CVD hBN was transferred to Quantifoil grids with holey carbon. Wide-field PL located P3 emitters; after photobleaching, annular dark-field STEM at 80 kV (JEOL ARM200CF with probe corrector) imaged defects. A low-to-high magnification workflow located vacancies; high-contrast imaging used long camera lengths to enhance low-angle scattering and Z-contrast. Image simulations and line profiles verified species-specific contrast trends (~Z^1.85). Atom-by-atom analysis cataloged monovacancy (VB, VN) configurations and heteroatom substitutions. - XPS: Multilayer CVD hBN (on Cu) and liquid-exfoliated hBN (drop-cast on SiO2) characterized using PHI VersaProbe II (Al Kα). Prior Ar sputtering (500 eV, 500 nA, 1 min) removed adventitious hydrocarbons. Peak deconvolution identified chemical states (B 1s, N 1s, C 1s). Carbon content and bonding (sp2 C–C, C–N variants) assessed; detection limits considered (0.1–1 at%). - Thermal annealing: Ar anneal at 850 °C (1123 K), 1 Torr, 50 sccm, 30 min. PL was measured at identical locations pre/post anneal for liquid-exfoliated samples; overall intensity decay analyses quantified lifetime changes for CVD and liquid-exfoliated samples. - Data analysis: Emission species categorized (P1–P6). Time-traces often showed discrete bleaching steps indicating multiple emitters within the confocal spot; summed decay curves used to extract population lifetimes. Confidence intervals compiled; summary in Table 1 of the article.

- Universal two-timescale photobleaching for the P3 (1.98 ± 0.05 eV) emission across hBN sources and thicknesses: a fast τ1 ≈ 5–10 s and a slower τ2 ≈ 150–220 s. - Monolayer CVD hBN in air: single-exponential overall decay τ = 6.6 ± 0.6 s; dwell-time analysis ~4.7 s. Discrete bleaching steps indicate multiple emitters in focus. - Multilayer CVD hBN (air): bi-exponential τ1 = 11 ± 4.1 s (33%), τ2 = 170 ± 20 s (67%). - Liquid-exfoliated hBN, type-I (air): bi-exponential τ1 = 10 ± 1.6 s (53%), τ2 = 220 ± 30 s (47%). - Liquid-exfoliated hBN, type-II (air): photostable with lifetime >2000 s; sharper ZPLs than type-I; antibunching observed for P3 only in type-II. - Monolayer CVD hBN under N2: bi-exponential τ1 = 5.0 ± 1.1 s (46%), τ2 = 150 ± 10 s (54%); single-exponential fit inadequate, indicating residual fast process likely due to trapped O2 at the hBN/substrate interface. - Four-layer stack (air): bi-exponential τ1 = 11 ± 3.9 s (36%), τ2 = 150 ± 20 s (64%). Ratio r consistent with top-layer exposure and minor contribution of interface-trapped O2 to bottom layer; stacking substantially increases average lifetime vs monolayer. - Laser power dependence: bleaching rate 1/τ scales linearly with excitation power density, consistent with photochemical bleaching from excited states in a three-level system; excludes pure thermal mechanism (ΔT < 100 °C estimated). - Mechanism assignment: Fast τ1 attributed to O2-driven photochemical reactions with emitting defects; mitigated by N2 environment and by hBN stacking. The slower τ2 is consistent across environments and samples, indicating an intrinsic, environment-independent mechanism. - ADF-STEM after bleaching (monolayer CVD): 22 vacancies identified; 15 monovacancies with majority VB over VN (only one VN). Among 12 VB with atomic-scale resolution, 10 show heteroatom substitution at innermost N sites (C and/or O). Frequent C substitution along vacancy edges and in lattice; O substitutes N, C substitutes both B and N. A VB with two O substitutions at innermost N remained stable across scans, indicating relative stability of O-rich configurations. Carbon-related hybrid B–C–N motifs and C–C dimers observed, consistent with simulations and literature. - XPS: After Ar sputter cleaning, CVD hBN shows significant C (≈8.3 at%) mainly as sp2 C–C (graphene-like), and some B–O; liquid-exfoliated hBN shows no detectable C 1s and small B–O, though low-level C cannot be excluded. CVD growth can inadvertently introduce graphene domains (confirmed by SEM/Raman), likely from precursor impurities; phase-separated hBN/graphene favored when B, C, N present. - Annealing effects: In liquid-exfoliated hBN, type-I emitters largely disappear after Ar anneal at 850 °C, whereas all type-II emitters persist with narrow ZPLs. In multilayer CVD, annealing increases lifetimes but does not yield the most stable emitters seen in liquid-exfoliated samples. Proposed explanation: vacancy migration (barriers ~2.3–3.1 eV) at anneal temperature; C at/near defects lowers migration barriers and promotes defect restructuring under excitation. - Overall: Shielding O2 (N2 atmosphere, stacking) enhances photostability by suppressing τ1. The τ2 process likely arises from carbon-assisted defect migration under illumination. Optimizing carbon concentration and preventing graphene-domain formation are critical for scalable fabrication of photostable, narrowband hBN emitters.

The work disentangles two distinct contributions to photobleaching in hBN emitters. The fast component (≈5–10 s) is extrinsic and oxygen-driven, as evidenced by its mitigation under N2 and through hBN stacking that blocks O2 access. Its linear power dependence supports a photochemical mechanism stemming from excited-state reactions. This explains the poorer stability of monolayers in air and suggests straightforward protections (encapsulation, inert atmospheres) for improved device-level stability. The slower component (≈150–220 s) persists irrespective of external gaseous environment, including in multilayer systems and monolayers under N2, indicating an intrinsic mechanism. ADF-STEM reveals extensive C and O substitution at and around VB sites; XPS confirms significant sp2 carbon in CVD films and negligible carbon in liquid-exfoliated samples. Together with annealing trends and theoretical insights, the data support carbon-assisted vacancy migration under laser excitation as the intrinsic bleaching pathway, where C-rich environments facilitate defect restructuring/coalescence into non-emissive configurations. This framework explains why liquid-exfoliated, low-carbon samples host highly photostable, narrowband (type-II) emitters, while CVD samples with substantial carbon content exhibit faster intrinsic bleaching. The findings guide emitters’ engineering: reduce/exclude carbon incorporation during synthesis, avoid graphene-domain formation, and incorporate protective encapsulation to exclude O2.

Photostability of hBN quantum emitters can be substantially improved by addressing two bleaching pathways. The fast, oxygen-mediated photochemical bleaching is suppressed by O2 shielding (inert atmospheres and hBN stacking), increasing average lifetimes by over 20×. A second, slower bleaching process is intrinsic and consistent with carbon-assisted vacancy migration under illumination. Atomic-resolution ADF-STEM and XPS establish prevalent C and O substitution at VB sites in CVD-grown hBN, correlating carbon with reduced photostability. Thermal annealing of liquid-exfoliated hBN eliminates unstable (type-I) emitters while preserving photostable (type-II) ones, consistent with vacancy migration. These insights suggest practical routes: optimize synthesis to minimize carbon incorporation/graphene domains, employ hBN encapsulation, and potentially fabricate tri-layer stacks with a photostable emitter layer sandwiched between defect-free hBN monolayers to achieve both efficient light extraction and robust photostability. Future work should unambiguously link specific atomic defect configurations to emission species and stability, and develop scalable growth protocols that control carbon content and defect chemistry.

- Co-localization between PL and STEM, while improved, remains indirect; definitive assignment of a specific atomic defect to a given emitter was not achieved. - Hydrogen termination at vacancies, potentially stabilizing certain configurations, is difficult to detect via ADF-STEM and remains unresolved. - XPS detection limit (0.1–1 at%) means low-level carbon in liquid-exfoliated samples cannot be fully excluded; sputtering alters near-surface chemistry and removes layers. - Photostability statistics focus on the P3 species; other emission species (e.g., P2) can behave differently, and full generalization requires broader sampling. - The intrinsic bleaching lifetime (~200 s) remains insufficient for many applications; proposed tri-layer encapsulation and optimized growth need experimental validation. - Power-dependent bleaching modeled with a simplified three-level scheme; microscopic reaction pathways and excited-state chemistry with O2 were not directly identified.