Liquid-activated quantum emission from pristine hexagonal boron nitride for nanofluidic sensing

Liquids confined down to the atomic scale can show radically new properties. However, only indirect and ensemble measurements operate in such extreme confinement, calling for novel optical approaches that enable direct imaging at the molecular level. Here we harness fluorescence originating from single-photon emitters at the surface of hexagonal boron nitride for molecular imaging and sensing in nanometrically confined liquids. The emission originates from the chemisorption of organic solvent molecules onto native surface defects, revealing single-molecule dynamics at the interface through the spatially correlated activation of neighbouring defects. Emitter spectra further offer a direct readout of the local dielectric properties, unveiling increasing dielectric order under nanometre-scale confinement. Liquid-activated native hexagonal boron nitride defects bridge the gap between solid-state nanophotonics and nanofluidics, opening new avenues for nanoscale sensing and optofluidics.

its 6 eV band gap have been identified as room-temperature quantum emitters [13][14][15] .While emitters in hBN have been induced artificially using techniques such as irradiation 16,17 or carbon doping 18,19 , the potential of liquid treatments remains largely unexplored.Recent studies have combined liquid and irradiation treatments to activate plasma-induced surface defects in hBN using water 20 and binary mixtures of water with organic solvents 21 .Post-treatment of ion beam-exposed hBN with liquids has also been shown to modify defect emission properties 22 .Furthermore, plasma-induced surface defects have been utilized to study single interfacial charge dynamics, such as proton hopping 20,21 .However, even mild plasma treatment induces mechanical and chemical changes that result in hBN crystals no longer having atomically smooth surfaces 23 , thus preventing the integration of defects in ultra-flat van der Waals heterostructures, which are crucial for advancements in angstrom-scale fluidics 4,8 .
In this study, we demonstrate that organic solvents can activate visible-range quantum emission from pristine high-quality hBN crystals.We attribute this phenomenon to the interaction between organic molecules and native surface defects [24][25][26][27] .
By employing spectral super-resolution microscopy 28 , we observe defect-mediated molecular random walks and couplings between defect dipoles and the liquid medium, leading to tunable emission wavelengths through the dielectric properties of the liquid.Leveraging intrinsic properties of widely used 2D materials and common solvents, the fluorescence activation mechanism reported here is utilized to image nanofluidic structures, with emitters serving as nanoscale probes of the order and dynamics of liquid media confined to the nanoscale.

Liquid-activated fluorescence from pristine hBN
The surface of untreated hBN crystals exhibits visible-range fluorescence when in contact with common organic solvents like ethanol.To demonstrate this effect, we exfoliated high-quality hBN crystals 24 onto a glass coverslip, which was placed in a liquid-filled chamber on an inverted microscope (Fig. 1a).Photoluminescence (PL) of the crystals under 561 nm wide-field laser illumination (0.35-3.5 kW/cm 2 ) was collected using a high-numerical aperture objective and projected onto a camera chip.
Pristine hBN in air or water did not exhibit fluorescence under these illumination conditions.However, we observed intense fluorescence from as-exfoliated crystals in contact with ethanol (Fig. 1b).The fluorescence intensity gradually decreased over continuous illumination and stabilized after several seconds, revealing emission from sub-diffraction spots (Fig. 1c and Supplementary Video 1) which can be localized with ≈10nm precision using single-molecule localization microscopy 29 (details in Methods).We attribute this emission to the activation of defects present in the as-exfoliated crystal through contact with the liquid, resulting in randomly distributed transient emitters on the surface (Supplementary Fig. 1).Under constant 3.5 kW/cm 2 illumination, the number of emitters decreased to a stable value of approximately 0.4 per square micron (Fig. 1d).This process does not deteriorate the crystal (Supplementary Fig. 2 & 3), and the decrease observed in Fig. 1c is reversible: when left in the dark, the crystal fluorescence recovered within tens of minutes (Supplementary Fig. 4) without inducing additional emitters in the steady state (Supplementary Fig. 5).
Remarkably, the fluorescent activation of the surface occurred with most common organic solvents, including n-alkanes (pentane to hexadecane) and primary alcohols (methanol to 1-pentanol).However, we did not observe any emission in pure water, heavy water, and hydrogen peroxide.To quantitatively compare the steady-state fluorescence in different liquid media, we imaged freshly cleaved hBN crystals in several liquids under 3.5 kW/cm 2 illumination.The steady-state fluorescence can be quantified using the crystal brightness I crystal , defined as the sum of localized emitter intensities per surface unit and time unit (Fig. 1e).From these observations, we can classify solvents into three types based on the extent of hBN fluorescent activation.
Most organic solvents, such as primary alcohols, n-alkanes, and chloroalkanes, exhibited intense fluorescence (I crystal > 500 photons/µm 2 /s), representing type I activation.Glycerol and other high-boiling point liquids (≥ 200 • C) exhibited a limited but measurable level of fluorescence, classified as type II.On the other hand, pure water showed no activation, falling into type III (I crystal < 10 photons/µm 2 /s).Furthermore, the addition of 10v/v% of water in ethanol reduced the number of emitters, resulting in a 5-fold reduction in the fluorescence signal (Supplementary Fig. 6).The distinction between activation types could not be solely explained by physical parameters of the solvents (Supplementary Fig. 7), indicating a chemical specificity.
The steady-state emitter density was found to depend on the liquid medium (Supplementary Fig. 7 & 8), the illumination power (Supplementary Fig. 9), and macroscopic flow over the crystal (Supplementary Fig. 10).Considering the dynamics of these fluorescent emitters, a striking observation is the presence of fluorescent trajectories on the crystal surface (Supplementary Movie 2).These trajectories indicate the correlated activation of neighboring defects, corresponding to molecular random walks.By linking the super-resolved localizations of emitters (see Methods), we can extract the associated trajectories, as illustrated in Fig. 2b.Previously observed trajectories on plasma-exposed hBN in water and binary mixtures of water and alcohols were attributed to proton hopping 20,21 .A similar phenomenology is thus expected here, with (i) defect activation due to reversible charge transfer from and to the solvent and (ii) correlated activation of neighboring defects, mediated by the lateral motion of charge-bearing solvent molecules that remain physisorbed on the crystal surface.However, the aprotic nature of some solvents used here, and the spectral differences point to a distinct emitter type and reactivity.Since neither pristine hBN nor the liquids used possess visible-range electronic transitions, the activation of emitters at the interface must arise from an electronic structure rearrangement that generates this new optically addressable electronic transition.This could be explained by the chemisorption of organic molecules onto hBN defects 30 .The chemical selectivity observed in Figure 1e suggests that a necessary and sufficient condition for a pure liquid to activate native hBN defects is the presence of a carbon atom in its molecular structure.This finding aligns with recent research indicating the crucial role of carbon in activating visible-range emitters in hBN 18 .While a direct observation of the exact chemical structure of the emitters is challenging, the physicochemical interactions between the defects and the liquid, along with the photophysical properties of the emitters will guide structural assignment.

Analysis of emitter dynamics
To understand the trajectories quantitatively, we acquired 50k, 6ms-long frames of a 13×13µm crystal area in the steady state, yielding 700k localizations leading to 100k trajectories.Figure 2a shows a subset of these trajectories overlaid with the super-resolved image obtained from 5000 frames.This visualization reveals that while some trajectories exhibit free hopping 3/30 behavior, others remain trapped for extended periods, resulting in bright spots in the super-resolved image.Most emitters are active for tens to hundreds of milliseconds, but a fraction displays stability for several seconds (Fig. 2c), with some activations lasting over a minute (Supplementary Fig. 11).As measurements were conducted at a low emitter density where trajectories do not split or merge with statistical significance, we assign trajectories to single molecules binding to single defects.We analyzed the observed molecular random walks through their trajectory residence times on the crystal surface T T res , which comprise complex information on both the chemisorption energy at defect sites and the physisorption energy on pristine hBN in between defects, as well as residence times of molecules at single defect sites T D res corresponding to chemisorption only.Both residence times were found to follow a double exponential decay (Fig. 2d), with slow exponential decay components τ D res =35±1 ms and τ T res =82±3 ms, respectively.Assuming that single-defect residence times follow the Arrhenius equation τ D res = ν −1 e ∆G/kT where ν ≈ 10 12 − 10 13 s −1 is a molecular attempt rate 21 , we obtain a desorption energy barrier ∆G ≈ 24-27 kT≈ 0.6-0.7 eV which is larger than typical physisorption energies (tens of meV) and smaller than covalent bonding energies (several eV) 31 .This can be rationalized in terms of a lowered energy barrier under illumination 20 consistent with the observed light-induced reduction in number of emitters in Fig. 1d as well as the illumination power dependency of the density of emitters (Supplementary Fig. 9).Turning our attention to the emitter motion, we find that their one-dimensional displacement probability density function 32 PDF(x,τ) follows a two-component Gaussian distribution (Fig. 2e).The central part of the distribution remains of constant size (≈20 nm) corresponding to the localization uncertainty when a trajectory is trapped.The tails of the distribution however enlarge with increasing lag time, characterizing hopping events.As shown in Figure 2f, the hopping tail size scales as √ 2Dτ with lag time τ, which corresponds to Brownian diffusion with a diffusion coefficient D= 9.1×10 −14 m 2 /s, which is over four orders of magnitude slower than bulk liquid molecular diffusion coefficients.While this observation makes it impossible to resolve the molecular travel time between defects, this slowdown enables the detection of bright emission from localized spots, which we now propose as a spectral sensing tool.

Spectral properties and solvatochromic sensing
We examined the spectral response of the emitters to their liquid environment using spectral single-molecule localization microscopy (sSMLM) 28 which enables simultaneous localization and spectral characterization (Fig. 3a, details in Methods).
Single-emitter spectra were found to be homogeneously distributed, with the appearance of a single population of emitters when exposed to the same liquid environment.Ensemble averaged spectra were consistently characterized by two peaks, classically attributed to the zero-phonon line (ZPL) and the phonon side band (PSB) for emitters embedded in a matrix (Fig. 3b).
Interestingly, these ensemble spectra appeared to depend strongly on the activating liquid, and more precisely on its static dielectric constant ε liq .We present in Figure 3b spectra of emitters obtained in the following liquids of increasing polarity: pentane, tert-butyl alcohol and DMSO.A notable polarity-induced solvatochromic redshift of the emission was gradually observed from nonpolar pentane (615 nm) to more polar tert-butanol (626 nm) to highly polar DMSO (641 nm).Beyond the ZPL shift, we observed changes in the PSB, which is less clearly defined for polar solvents.We report in Figure 3c the center wavelengths of both peaks as obtained from fitting to a two-Lorentzian model for several solvents ordered by increasing ε liq .
We tested 1-pentanol, isopropanol and methanol on top of previously introduced liquids to interpolate dielectric constant values and found that both the ZPL and the PSB are redshifted by over 25 nm (≈80 meV) in highly polar liquids compared with nonpolar alkanes.In the range ε liq < 25, a linear dependency was observed between the ZPL wavelength and the dielectric constant, indicated by the dashed line (Fig. 3c) with a slope of approximately 1 nm per unit.The Jablonski diagram presented in Figure 3d illustrates the process giving rise to the observed spectra.By absorbing a photon (excitation, green arrow), the emitter is excited to a dipolar state which interacts with the solvent and relaxes before radiating (dipole relaxation, curved arrow).Direct evidence of the dipolar nature of our liquid-activated emitters is presented in Figure 3e where either the linearly polarized light used for excitation (green) or the PL emission (orange) was rotated while monitoring the signal through an analyzer (details in Supplementary Fig. 11).As sketched in Figure 3f, the solvatochromic redshift can be described by the presence of liquid molecular dipoles stabilizing the excited state dipole, thus lowering its energy and redshifting the ZPL.After this step, as shown in Figure 3d, the transition back to the ground state can occur in two ways: direct emission of a photon (ZPL, orange arrow), and phonon-assisted emission (PSB, red arrow), which is redshifted compared to the ZPL as a fraction of the energy leads to lattice vibrations (phonon, zigzag arrow).Our analysis of the phonon side bands revealed increasing phonon broadening and decreasing phonon side band content with increasing solvent polarity (details in Supplementary Discussion and Supplementary Fig. 12).In the case of a mixture of polar ethanol and apolar heptane, the spectrum was very similar to that of the polar liquid, demonstrating a strong affinity between the emitters and polar molecules (Supplementary Fig. 13 & 14).

Time-resolved measurements reveal quantum emission
To prove that the measured fluorescence originates from single-photon emitters, we performed time-correlated photon counting.For this, a 0.7 mW continuous-wave 561 nm laser beam was focused to a ≈ 1 µm 2 spot onto hBN crystals in liquid, and fluorescence signal was collected by two single photon detectors (SPDs) in a Hanbury Brown and Twiss interferometer configuration (inset of Fig. 4a).The typical time trace of a stable single emitter in acetonitrile is shown in Figure 4a.The analysis of photon arrival times from a 10s window shows a clear photon antibunching dip g (2) (0) = 0.25 ± 0.02 at zero delay time, demonstrating single-photon emission (Fig. 4b).This result implies that the bright spots are single emitters and not clusters, thereby their optical readout truly reports on nanoscale properties of the liquid.This activation of quantum emission through the strong chemisorption interaction between a single activating molecule and a single defect was observed in carbon nanotubes 33 , but the mechanism at play here exhibits the particularities of being transient and observed in liquid.Photon statistics under pulsed excitation show the suppression of the correlated pulse peak at zero delay time, confirming single-photon emission (Fig. 4c).This feature was also found in hexadecane with a measured g (2) (0) = 0.45 ± 0.04.We thus demonstrated liquid-tunable single-photon emission with a ZPL shift of 21 nm (inset in Fig. 4b).This shift is comparable to those achieved by hBN defects in response to strain 34 or electric fields 35 , which shows the potential of liquid-activated emitters as dielectric environment sensors.

Integration in single-digit nanofluidic systems
Building upon the characterization of the emission in bulk liquids, we probed hBN-liquid interfaces in molecular confinement in two-dimensional nanoslits.As sketched in Figure 5a, the nanoslits are obtained by van der Waals assembly of heterostructures comprising 3 crystals: bottom, spacer and top.The top crystal was chosen to be muscovite mica for its transparency and lack of fluorescent properties, and the bottom crystal was pristine hBN to be activated by the liquid.The middle crystal, composed of few-layer graphene patterned by electron beam lithography acted as a spacer, defining a slit-shaped channel between the hBN and mica crystals, whose height was set by the number of 3.4 Å-thick graphene layers 4 (details in Methods and in Supplementary Fig. 15).
An optical micrograph of a device with h = 2.4 nm is provided in Figure 5c.The bottom blue region corresponds to the full heterostructure with nanoslits, and the top purple region corresponds to the open hBN crystal masked by graphene spacers but without encapsulation by the mica.We verified that covering the pristine hBN crystal with a patterned few-layer graphene crystal masks liquid-activated emitters, as was observed for other types of hBN emitters 36,37 .On bare hBN, emitters are randomly distributed (Supplementary Fig. 16), but the graphene mask allows for their precise positioning on the basal plane of hBN in liquid.An overlay of the graphene spacer atomic force microscopy (AFM) topography and the super-resolved image is shown in Fig. 5b, demonstrating the correspondence between the lithographically defined graphene pattern and the optically measured fluorescence from masked hBN.We further verified that capping masked hBN with mica does not quench its fluorescence, allowing direct imaging of emitters in confined liquid (Fig. 5d).We show in Figure 5e that the localization intensity distributions with and without the confining mica top are similar, but the number of emitters in confinement is reduced by two thirds.Single-defect residence times inside nanoslits were found to be slightly longer in nanoslits than in masked hBN (Supplementary Fig. 17).Therefore, the observed three-fold decrease in emitter number comes not from faster photobleaching of the emitters but rather from confinement-induced slowdown of their activation kinetics.
Integrating emitters into nanofluidic structures allows probing the effect of confinement on liquid structure and dynamics.
We first confirm the observation of trajectories in 1.4 nm confinement in Figure 5g, where a set of emitters is shown in the the 150 nm wide slit.We then focus on the spectral properties of confined emitters, which can be robustly extracted through sSMLM with relatively low numbers of localizations (<1000).We present sSMLM spectra obtained in the bare, masked, and confined geometries for ethanol (Fig. 5h) and acetonitrile (Fig. 5i).For both solvents, bringing the confinement size from 2.4 nm down to 1.4 nm leads to a clear blueshift pinpointed by the dashes indicating the ZPL position.For ethanol, the ZPL blueshifts from 637 nm to 624±2 nm, and for acetonitrile, from 636 nm to 621±1 nm under 1.4 nm confinement, bringing the spectral signature of a strongly polar solvent close to that of nonpolar alkanes.This substantial confinement-induced blueshift suggests that emitters experience a reduced dielectric constant of ε conf ≈11 in ethanol and 7 in acetonitrile when considering Figure 3c as a calibration curve.
Bringing the top mica wall close to the emitters can impact their emission in two ways.Firstly, confinement by the opposite wall of lowered dielectric constant ε wall ≈ 8 could reduce the effective number of solvent molecules interacting with the emitter and confine the electric field lines within the slit 38 , potentially destabilizing the excited state.Secondly, the emission can be affected by reducing the out-of-plane component of the liquid dielectric tensor 8 .As depicted in Figure 5f, the interaction range between a solvent molecule with dipole µ S and the defect with dipole µ D is given by where a ≈ 0.25 nm is the in-plane lattice parameter of hBN and e is the elementary charge, we estimate µ D ≈ 12 D. Using the dipole moments of ethanol (1.7 D) and acetonitrile (3.4 D), we find ℓ dip ≈ 1 nm and 1.3 nm, respectively, which are smaller than the height of the nanoslit.Hence, the observed effect under 1.4 nm confinement might not arise solely from geometrical effects due to the proximity of the top wall, and could be explained by the confinement-induced reduction of the solvent dielectric constant, as observed for water 8 and predicted for other liquids 39 .These results consolidate the picture of dipolar environment-tuned emitters, whose properties are affected by changes in the sensing hemisphere with volume ∼ 2/3πℓ 3 dip , which encloses fewer than 100 molecules in the case of acetonitrile.Beyond passive diffusion and dielectric sensing of confined liquids, tracking emitter dynamics in confinement may be used to directly image nanoscale flow and study its interplay with defects 40 .18). f, Illustration of the effect of confinement: the liquid dielectric constant can be changed by confinement (ε conf ) and the defect dipole can interact with solvent molecules (yellow ellipses) within a range ℓ dip , comparable to the confinement size h.g, Representative trajectories in 2.4 nm-high nanoslits filled with ethanol, overlaid with the super-resolved image.h,i, sSMLM spectra of liquid-activated defects in nanoslits filled with ethanol and acetonitrile, respectively.The confinement size is tuned from the open geometry to 2.4 nm down to 1.4 nm.Solid lines correspond to two-component Lorentzian fits, and black dashes indicate the extracted ZPL position.

Outlook
hBN crystals, already known for exceptional optical properties, exhibit a peculiar interaction with liquids.When in contact with organic solvents, native point defects on the atomically smooth surface of the crystal become emissive.This unique system, where the encounter of a single defect with a single organic molecule yields a single-photon emitter, combines solid-state emitters and organic fluorophores, providing a new tool for studying solid-liquid interfaces.We demonstrated two sensing approaches using liquid-activated hBN: the activation dynamics provide insights into interfacial charge transfer between defects and single molecules, while the emission spectra of the emitters offer information about the nanoscale dielectric environment.These phenomena were found to hold in confinement as small as a few nanometers, where only ensemble-averaged measurement techniques have been successful so far.As it relies on common samples and widely available single-molecule microscopy techniques, this approach could be readily applied for optical imaging and sensing in nanofluidic systems operando.

Sample preparation
Pristine hBN flakes from high-quality crystals 24 were exfoliated onto borosilicate glass coverslips (no.1.5 Micro Coverglass, Electron Microscopy Sciences, 25 mm in diameter, 170 µm thick), using low-adhesion blue tape.The glass coverslips were cleaned either by (i) sonication in acetone followed by rinsing in isopropanol (ii) sonication in 2% Hellmanex III glassware cleaning agent solution, following by sonication in deionized water.No differences were observed between (i) and (ii).The coverslips were rinsed three times in the last solution, after which they were dried with a nitrogen gun covered with a 20 nm-pore filter.Adhesion to the substrate upon exfoliation was promoted by oxygen plasma cleaning of the coverslip (2 min, 100W).
High-quality crystals purchased from HQ Graphene and hBN crystals obtained by crystallization of hBN from molten iron in nitrogen-hydrogen atmosphere were also tested, and exhibited no notable difference with samples grown by the authors.The crystals were immersed immediately following exfoliation, and the PEEK chamber was thoroughly rinsed three times at each solvent exchange with fresh solvent.The chamber was covered by an additional glass coverslip to prevent solvent evaporation and contamination.The fabrication details for the microfluidic flow cell are given in the corresponding Supplementary Methods.
Solvents used are detailed in Supplementary Table 1.
Nanoslits are made of a van der Waals (vdW) heterostructure of a spacer layer sandwiched between a top layer and a bottom layer following the same protocol as previously reported 4 .Here, the vdW stack is composed of top mica-graphene spacer-bottom hBN.In brief, thin (few atomic layers) graphene was first patterned via EBL into parallel strips with a width of 1 µm and a separation of 150 nm.A mica crystal (≈ 200 nm thick) was then transferred on top of the graphene spacer, via PMMA based transfer method.Then this mica-graphene spacer stack was lifted and transferred onto a freshly exfoliated hBN layer.This whole stack was then transferred onto a glass coverslip for the imaging.The slit dimensions of the final device is shown in Figure 5a where it has a width of approximately 150 nm, a length of 20 µm and a height equivalent to the thickness of the graphene spacer.Fabrication flow chart and materials are described in detail in Supplementary Figure 15.

Chemicals used
All chemicals were purchased with maximum purity grade available, and some were purchased anhydrous.The full list is provided in Supplementary Table 1.No effect of solvent purity or residual water traces on hBN fluorescence activation was observed.
groups of 100 single-emitter sSMLM spectra.This ensured sufficient signal-to-noise ratio for fitting while reflecting variations within the ensemble spectra.

Trajectory analysis
Wide-field fluorescence frames acquired with the EMCCD camera were first localized using ThunderSTORM 41 .Trajectories were obtained by applying the Crocker-Grier linking algorithm 43 to the localization microscopy tables.We used the implementation of the algorithm provided by the Python package trackpy 44 .Briefly, for each localization event at a given frame, the algorithm links another event if it is found at the next frame within a specified search range.This search range was set to 120 nm in Fig. 2, as the probability of having a 1D displacement exceeding this value is less than 1% according to the analysis in Fig. 2e.
In the statistical analysis of trajectories, the emitter position is described as a random variable of time X(t).The trajectory residence time on the crystal surface T T res corresponds to the duration of this single trajectory.The one-dimensional displacement probability distribution function is given by PDF(x, τ) = P X(t + τ) − X(t) = x .The residence times of molecules at single defect sites T D res were obtained using the same linking algorithm with a linking range of 35 nm (about twice the typical localization uncertainty) instead of 120 nm.

Figure 1 .
Figure 1.Liquid-induced fluorescence from pristine hBN crystals.a, Sketch of the experimental setup.b, Wide-field fluorescence images of a hBN crystal under 3.5 kW/cm 2 561 nm laser light illumination with 1 second exposure time.No fluorescence was observed in water, but in ethanol the entire crystal surface became fluorescent.The images underwent linear contrast enhancement.c, A zoomed-in view of the dashed yellow box in b reveals dense clusters of emission when turning the laser is turned ON, with 6 ms exposure time.After 10 s of wide-field illumination, the crystal surface reached a stable number of diffraction-limited isolated emitters.d, Localization microscopy-based counting of the emitters as a function of the illumination time: after 5 seconds, a steady state was reached.The dashed line is a fit to an offset exponential relaxation.e, Liquid dependency of the crystal fluorescence, showing strongly activating liquids (type I), mildly activating liquids (type II) and no activation in water (type III).

Figure 2 .
Figure 2. The surface of pristine hBN reveals interfacial molecular dynamics.a, Overlay of a super-resolved image from 5000 frames showing hopping emitters in isopropanol as the linked trajectories, as well as trapped spots.b, Artist's view of the correlated activation of neighboring defects leading to trajectories.c, Representative intensity traces from the same images, taken from 7x7 pixel bins around emitters as delimited by the dashed yellow box in Fig. 1c, with 6 ms exposure time.The top right trace corresponds to a long defect activation.The top left trace corresponds to a short activation of the same defect, magnified on the bottom panel.d, Distribution of residence times on single defects and for the entire trajectories.Dotted lines are fits to a two-component exponential decay.e, Displacement probability density functions (PDF) of trajectories after different lag times τ=6,24,66,142 ms.Dashed lines are fits to two-component Gaussians.f, Visualizing the evolution of the two modes of the Gaussian fit in Fig. 2e with increasing lag time.The central region, corresponding to the trapped state, remains of constant width, whereas the tails, corresponding to hopping, enlarge with time.The solid line is a fit to a standard diffusion curve.

Figure 3 .
Figure 3. Spectral properties of surface dipole emitters coupled to both solid and liquid environments.a, Spectral single-molecule localization microscopy (sSMLM) splits the fluorescence signal from an emitter into a localization component (left, SMLM) and a spectral component (right) on the same camera chip.b, Ensemble spectra of liquid-activated emitters in different type I solvents exhibiting a clear zero-phonon line (ZPL) and phonon side band (PSB).c, Visualizing the wavelength shifts of both ZPL (circles) and PSB (squares), which correlate with the dielectric constant of the liquid.Peak positions were obtained by fitting to a sum of Lorentzians.The dashed line indicates the linear solvatochromic range, with a slope of 1 nm per unit.Error bars correspond to standard deviations of fitting parameters from groups of 100 single-molecule spectra.d, Jablonski diagram of processes at play: 561 nm laser excitation induces a dipolar excited state, which can emit directly (orange arrow, ZPL) or with emission of a phonon (red arrow, PSB).e, Normalized intensity as a function of input or output light polarization angle α relative to the emitter axis.The solid lines correspond to fits to ideal electric dipole emission cos 2 α + constant.More details are provided in Supplementary Figure 11.f, Sketch of an excited emitter that can interact with both the crystal through phonons and with surrounding molecules (yellow ellipses).

Figure 4 .
Figure 4.Quantum emission from liquid-activated emitters.a, Representative PL trace from an isolated emitter in acetonitrile under 0.7 mW confocal excitation.The shaded region corresponds to the 10s-long trace used for photon statistics.b, Normalized coincidences g(2)  measured from time correlated single photon counting in a Hanbury Brown and Twiss geometry, in hexadecane and acetonitrile.In both liquids, a pronounced antibunching is observed with g (2) (0) < 0.5 without background correction, proving the single-photon emission.The fluorescent lifetimes, corresponding to the width of the antibunching dip, were found to be 2.73±0.09ns and 2.20 ± 0.20 ns for acetonitrile and hexadecane, respectively.Spectra are shown in the inset, demonstrating liquid-tunable single-photon emission from 615 nm to 636 nm.c, Single-photon statistics under pulsed laser excitation showing a suppression of the central peak due to antibunching.

Figure 5 .
Figure 5. Nanoslit-embedded liquid-activated emitters.a, Sketch of the heterostructure nanoslit device.The red glow indicates an emitter inside the nanoslit.b, Overlay of a super-resolved image of masked ethanol-activated hBN and the AFM mapping of the graphene spacers.c, Optical micrograph of the heterostructure.On the purple colored part of the image, only graphene spacers on the hBN bottom crystal are present, which leads to masked hBN.The bottom blue region corresponds to the full heterostructure with slits.d, Super-resolved image of acetonitrile-activated emitters embedded in 2.4 nm-high nanoslits, from 30k frames with 20 ms exposure time and 1.4 kW/cm 2 illumination.e, Comparison of localization intensity distributions for masked hBN and 2.4 nm nanoslits in acetonitrile, showing no loss of photons but an overall reduction in number of localizations (details in Supplementary Fig.18).f, Illustration of the effect of confinement: the liquid dielectric constant can be changed by confinement (ε conf ) and the defect dipole can interact with solvent molecules (yellow ellipses) within a range ℓ dip , comparable to the confinement size h.g, Representative trajectories in 2.4 nm-high nanoslits filled with ethanol, overlaid with the super-resolved image.h,i, sSMLM spectra of liquid-activated defects in nanoslits filled with ethanol and acetonitrile, respectively.The confinement size is tuned from the open geometry to 2.4 nm down to 1.4 nm.Solid lines correspond to two-component Lorentzian fits, and black dashes indicate the extracted ZPL position.
τ) = P X(t + τ) − X(t) = x 1D displacementprobability density function constant ε wall Mica top wall static dielectric constant µ D Defect dipole moment µ S Solvent molecule dipole moment ℓ dip Range of dipolar interactions