Ultrafast, nonblinking single-photon sources from perovskite quantum dots in plasmonic nanocavities

Abstract

Achieving strong light-matter interaction to manipulate emission requires integrating colloidal perovskite quantum dots (PQDs) with plasmonic nanocavities, yet this integration is challenged by their vulnerability to polar solvents. We successfully synthesized highly emissive, solvent-resistant CsPbI₃ PQDs and integrated them into nanoparticle-on-mirror structures. This integration enabled a 435-fold reduction in emission lifetime and a 250-fold increase in total emission intensity. Key results include a very short radiative lifetime below 12 picoseconds and a record-high single-photon emission rate exceeding 2.3 × 10⁹ counts per second at room temperature. Notably, we also observed nonblinking single-photon emission with high purity arising from nanocavity-enhanced radiative electron-hole recombination. Finite-difference time-domain simulations confirmed ultrasmall mode volumes of ~3 × 10⁻⁵ (λ/n)³, effectively enhancing spontaneous emission via the Purcell effect. These ultrabright and nonblinking properties highlight the strong potential of this platform for future quantum technology applications.

Ultrabright, nonblinking perovskite quantum dots in plasmonic nanocavities achieve record-fast single-photon emission.

INTRODUCTION

Metal halide perovskites have attracted widespread attention over the past decade. Their composition is represented by the chemical formula ABX₃, where A represents a monovalent cation [e.g., CH₃NH₃⁺ (MA⁺), NH₂CHNH₂⁺ (FA⁺), or Cs⁺], B is a divalent metal cation (e.g., Pb²⁺ or Sn²⁺), and X is a halide anion (e.g., Cl⁻, Br⁻, I⁻, or mixed). The perovskite crystal structure has versatile properties, including high absorption coefficients, strong photoluminescence (PL), a narrow spectral bandwidth, precisely tunable bandgaps, high carrier mobility, and high defect tolerance (1–3). These characteristics make perovskites promising candidates for next-generation optoelectronic devices, such as solar cells, photodetectors, light-emitting diodes, and single-photon sources, also known as quantum emitters (4–6).

Efficient and bright single-photon sources are critical elements for quantum information systems, such as quantum key distribution, quantum state teleportation, and quantum communication (7–9). However, although the intrinsic emission rates from perovskite quantum dots (PQDs) are considered fast among all light-emitting materials, their emission lifetime can still be substantially shortened by harnessing strong light-matter interaction (10). A promising approach is to enhance the local photonic density of states at the emitter’s position, also known as the Purcell effect (11). One effective strategy to achieve high Purcell enhancement is integrating quantum emitters with plasmonic structures, where mode volumes of plasmonic cavities can range from 10⁻⁴ (λ/n)³ for single open cavities down to 10⁻⁷ (λ/n)³ for nanogaps in nanoparticle-on-mirror (NPoM) geometry (12, 13). Despite the fact that the spectral confinement in the plasmonic nanocavity is typically much weaker than in the optical microcavity because of plasmonic losses in the metals (14), plasmonic cavities with an extremely small mode volume exhibit an extraordinarily large enhancement of spontaneous emission rates, nonlinear optical responses, and strong light-matter interaction (15–18).

Several promising studies have demonstrated the interaction between quantum dot (QD) light sources and optical or plasmonic resonant cavities, successfully maintaining the single-photon property. For example, a Purcell factor (F_p) of ≈24 was achieved with CdSe/CdS/Zn QDs near gold nanoparticles (19), and a F_p of ≈43 was achieved with InGaAs QDs in photonic crystals (20). Other single-photon emitters, such as organic molecules (21), diamond pillars (22), and silicon vacancy centers in diamond (23), have also been studied to integrate with plasmonic structures. Various geometries have been proposed for plasmonic nanocavity-single quantum emitter systems, including nanopatch (24–26), nanogap (27–29), simple metal films (30, 31), bullseye antennas (32, 33), and ordered/disordered arrays (34–36). The fabrication of these metallic plasmonic structures often relies on the solution process with polar solvent. As a result, PQDs are difficult to integrate into these nanostructures because of their intrinsic ionic properties and, hence, high solubility in polar solvents. For instance, while silver (Ag) nanocubes are commonly well dispersed in high-polarity solvents like water, ethanol, and isopropanol (37), traditional PQDs with oleic acid (OA) and oleylamine (OAm) ligands suffer from severe phase degradation and ligand desorption in polar solvents (38), which often leads to a total annihilation of PQD emission after integration with a nanocavity. Furthermore, all quantum emitters operating at room temperature are known to suffer from blinking—a PL intermittency at the single-dot level (38–40). This phenomenon refers to individual nanoscale emitters randomly switching between an ON state (emitting light) and an OFF state (being dark) under continuous excitation, with blinking becoming more conspicuous at higher excitation power. Common methods to suppress blinking include surface passivation of PQDs with a core/shell structure (39, 41) or operating experiments at cryogenic temperatures (42, 43). Recently, an important advance was reported by Wang et al. (28), who demonstrated that a blinking CdSe/ZnS QD can exhibit nonblinking PL when deterministically coupled to a plasmonic nanocavity. Unlike approaches relying on engineered core-shell structures, their work reveals that plasmon-enhanced radiative decay can outcompete nonradiative Auger recombination, thereby equalizing the quantum yields (QYs) of neutral and charged excitons within the same dot. This cavity-mediated acceleration of radiative rates not only suppresses the observable blinking but also enables unprecedented single-QD brightness, with photon generation rates exceeding 500 MHz. Despite this breakthrough, achieving robust, ultrabright, and nonblinking single-photon emission—particularly for PQDs integrated with nanophotonic cavities—remains a major challenge for advancing quantum photonics.

In this study, we successfully synthesized CsPbI₃ PQDs that are resistant to polar solvents by partially replacing the traditional ligands (OA and OAm) of PQDs with zwitterions. Our spray-synthesized, zwitterion-capped PQDs displayed near-unity PLQY, high stability, and excellent surface passivation through ligand encapsulation. Cubic-shaped PQDs with an average size of ~15 nm were observed via transmission electron microscopy (TEM), and various material analysis experiments including Fourier-transform infrared (FTIR) spectroscopy and x-ray photoelectron spectroscopy (XPS) were conducted to confirm the presence of zwitterions on the QD surface. Consequently, we coupled these single photon–emitting PQDs to plasmonic nanocavities. The structures comprised 100-nm Ag nanocubes separated from a flat Ag film by a PQD/polystyrene (PS) composite gap layer. To investigate the relationship between gap distance and interaction strength, we varied the thickness of the gap layer from 10 to 30 nm. From the scattering spectrum, the plasmonic structure with a 10-nm gap layer overlapped well with the emission spectrum of the PQDs, resulting in a 435-fold decrease in the emission lifetime and a 250-fold increase in the total emission intensity. This produced a tremendous maximum single-photon emission rate exceeding 2.3 × 10⁹ counts/s obtained before reaching saturated brightness. Furthermore, we unexpectedly observed the nonblinking emission in PQDs with plasmonic nanocavity at room temperature, while good single-photon purity was also maintained. Finite-difference time-domain (FDTD) simulations were implemented to study the Purcell enhancement of such a structure. With an ultrasmall mode volume of ~3 × 10⁻⁵ (λ/n)³ in this study, the spontaneous emission rate was effectively enhanced with a Purcell factor (F_p) of ≈250 with an electrical field concentrated at a corner of the nanocube-mirror gap. By combining the properties of high single-photon purity, record-high brightness, and nonblinking emission at room temperature, these achievements for CsPbI₃ PQDs with the NPoM structure are poised to advance the applications of perovskite-based QDs in quantum technologies, including quantum computing, quantum telecommunication, and quantum sensing.

RESULTS

Material characteristics of CsPbI₃ PQDs

The pristine CsPbI₃ PQDs (P-PQDs) were synthesized through our previously reported spray synthesis method (for details of synthesis, see Materials and Methods) (44). To improve the resistance to polar solvents, the synthesis was modified by partially replacing the traditional ligands (OA and OAm) with zwitterions, 3-(decyldimethylammonio)-propane-sulfonate inner salt (DPIS). The chemical structures of these surface ligands are shown in Fig. 1A.

Fig. 1. Characterization of D-PQDs and P-PQDs.

(A) Chemical structures of three surface ligands: OA, OAm, and DPIS. (B) Absorption and PL emission spectra of the D-PQDs. The red peak represents the D-PQD solution emission. a.u., arbitrary units. (C) PL decay lifetimes of D-PQDs and P-PQDs. (D) TEM image of separate and cubic-shaped D-PQDs. (E) FT-IR spectra of the D-PQDs and P-PQDs. XPS spectra for D-PQDs and P-PQDs: (F) Pb 4f and (G) I 3d. (H) P-PQDs (left) and D-PQDs (right) in hexane under white light and UV illumination. (I) P-PQDs (left) and D-PQDs (right) in IPA under white light and UV illumination.

Figure 1B shows the absorbance and PL spectra of the CsPbI₃ DPIS PQDs (D-PQDs), which exhibit clear excitonic absorption features and a high absolute PLQY of up to 95 ± 3% with an emission peak wavelength of 693 nm and a full width at half maximum (FWHM) of ~30 nm. Figure 1C displays the results of time-resolved PL (TRPL) measurement. The D-PQDs exhibit a considerably prolonged lifetime and slightly higher PLQY, indicating that DPIS effectively passivates surface defects. In addition, the much longer emission lifetime can be attributed to the polar nature of zwitterion bonding, which implies that DPIS was successful anchored on the surface of D-PQDs (45, 46)

The TEM image of the D-PQDs is shown in Fig. 1D, with their statistical size distributions displayed in fig. S1. The D-PQDs are cubic shaped with an average size of 15.5 ± 2.3 nm, and the size distribution appears homogeneous seemingly because of the larger contact area of the reaction species achieved by the spray synthesis method. Energy-dispersive x-ray spectroscopy was also conducted to inspect the chemical composition of CsPbI₃ P-PQDs (fig. S2A) and D-PQDs (fig. S2B). The percentages of Cs, Pb, and I in P-PQDs are 23.4, 19.0, and 57.6%, respectively. The ratio of Cs:Pb:I is almost close to the typical atomic ratio of 1:1:3 for CsPbI₃. On the other side, the percentage of S in D-PQDs is around 0.9%, confirming the existence of DPIS on CsPbI₃ QDs. FTIR analysis has also been performed to confirm the existence of DPIS in D-PQDs. According to the FTIR spectra, the peaks at 1030 and 1039 cm⁻¹ are only found in the D-PQD sample (Fig. 1E), and these two peaks are associated with S─O⁻ (1030 cm⁻¹) and S═O (1039 cm⁻¹) in the SO₃⁻ group of DPIS (47, 48). In addition, in XPS measurements, a noticeable shift of 0.3 to 0.5 eV toward lower binding energy is observed in the Pb 4f and I 3d peaks of D-PQDs compared to P-PQDs (Fig. 1, F and G), implying the presence of chemical interactions between the SO₃⁻ groups of DPIS and Pb²⁺ ions in CsPbI₃ QDs (49, 50). On the basis of the abovementioned material analysis evidence, the coordination of DPIS on the surface of CsPbI₃ D-PQDs is further confirmed. The resistance of D-PQDs to polar solvents was further evaluated by directly dispersing both P-PQDs and D-PQDs in isopropyl alcohol (IPA), as illustrated in Fig. 1I. In contrast to their comparable PL observed in hexane solutions (Fig. 1H), only the D-PQDs were able to retain their luminescent properties when dispersed in IPA. So far, stable metallic nanoparticles are usually coated with a poly(vinylpyrrolidone) (PVP) shell, which makes them available only in highly polar solvents (37, 51–54). We emphasize that only the high polar-solvent resistance of our D-PQDs enables the integration of perovskite nanomaterials with polar solvent–processed metallic nanostructures.

Plasmonic nanocavity

We used polar solvent–resisted D-PQDs as single-photon sources for constructing the plasmonic nanocavity structure. The schematic of the NPoM nanocavity is shown in Fig. 2A, and the nanocavity is designed as a sandwich structure in which a single highly emissive D-PQD is located between a 100-nm Ag nanocube and a flat 200-nm Ag film (for details of fabrication, see Materials and Methods).

Fig. 2. Plasmonic nanogap cavity for PQDs.

(A) Schematic of plasmonic nanogap cavity in which a Ag nanocube surrounded by PVP is situated on a Ag film, separated by a 10- to 30-nm bilayer spacer containing PQD and PS. (B) SEM top-view image of Ag nanocubes with a size of 100 nm. (C) Normalized PL spectrum of the single D-PQD (red area) and normalized scattering spectra of the plasmonic nanocavities with different gap layers: 10 nm (pink), 20 nm (orange), and 30 nm (green). (D) Dark-field image of plasmonic nanocavity with a 10-nm gap layer (inset: schematic of the reflected dark-field microscopy setup, where a ring-shaped illumination suppresses substrate reflection and only scattering from the NPoM cavity is redirected into the collection path and captured by the objective).

In this work, the D-PQD was deposited within a thin layer of PS to avoid PL quenching via direct contact between the QD and the conductor. This emission layer was prepared by spin coating a dilute solution of D-PQDs in PS onto the Ag layer. We designed different layer thicknesses (10 to 30 nm) to elucidate the dependence of cavity strength on gap distance. Next, 100-nm Ag nanocubes with a PVP shell in IPA were drop-casted on top of the PQDs. The scanning electron microscopy (SEM) image of the NPoM nanocavity is shown in Fig. 2B. Furthermore, the TEM image in Fig. 1D shows that the D-PQDs also exhibit a well-dispersed distribution without aggregation. This greatly increased the probability of a single D-PQD corresponding to only one Ag nanocube in the designed structure, which was achieved by carefully tuning each concentration. We also investigated the SEM cross-sectional view in fig. S3 to measure the thickness of each layer, which is consistent with the thickness measured via an ellipsometer.

According to Fermi’s golden rule, the spontaneous emission rate of an emitter is proportional to the local density of optical states (16). For the NPoM structure, the gap modes generated between a metallic cube and a metallic film have been proven to present a strong enhancement of the local density of optical states, leading to high Purcell enhancement (16, 24, 55, 56).

To evaluate the plasmon resonance, the scattering spectra of the plasmonic nanocavity samples with different gap layers were obtained and are shown with the PL spectrum of the single D-PQD in Fig. 2C, and the dark-field microscopic images are also illustrated in Fig. 2D and fig. S4 (A and B). Illustrations of the details of dark-field microscopic images are shown in Materials and Methods. In these dark-field images, the red spots signify the strong coupling mode (red-shifted scattering band) of isolated single nanocavity structures, while the yellow spots are attributed to the aggregated Ag nanocubes. Consistent with a previous study, the blue shift of the scattering spectrum peak with increasing layer thickness is a result of the strong dependence of plasmon resonance on the gap layer thickness, nanoparticle size, and the refractive index of the medium (57). For the plasmonic structure with a 10-nm PS gap layer, a far-field scattering spectrum was measured with a peak wavelength of 670 nm and an FWHW of 140 nm. The quality factors (Q ≡ λ/Δλ) of our NPoM cavities are estimated to be 4.7, 3.6, and 3.4 for gap layers of 10, 20, and 30 nm, respectively. The low quality factors of these NPoMs are attributed to plasmonic loss. Notably, the scattering spectrum of the 10-nm PS gap layer cavity exhibits maximum overlap with the emission spectrum of the single D-PQD, which suggests the optimal structure for Purcell enhancement for the CsPbI₃ PQDs and provides ideal resonance conditions.

FDTD simulation

To evaluate the enhancement of our NPoM nanocavity, we calculated the field enhancement factors and mode volumes of NPoM with different gap distances on the basis of FDTD simulations. As shown in fig. S5A, the coordinate system for the simulation is defined with the x and y axes set to in-plane, while the z axis is the normal direction of the substrate. Following previous studies on NPoM, the nanocube shows higher scattering efficiency, stronger field enhancement, and a smaller mode volume compared with a nanosphere of similar size. These advantages stem mainly from its reduced structural symmetry and sharper edges, which enable the nanocube to concentrate the near field around its surface more effectively (52, 53). Therefore, we calculated the enhancement factor (EF ≡ |E_max|²/|E₀|²) and effective mode volume $\left(V_{\mathrm{m}}\equiv \frac{\int {\mid E\left(\overrightarrow{r}\right)\mid }^{2}d\overrightarrow{r}}{\text{max}\left[{\mid E\left(\overrightarrow{r}\right)\mid }^2\right]}\right)$ to quantify the performance of this structure.

As shown in fig. S5B, the field enhancement values, |E_max|²/|E₀|² (plotted as log scale), decrease with gap distance, especially when the gap distance is below 15 nm, where |E_max|² denotes the maximum field intensity across all positions. The enhancement factors are 229, 152, and 145 for gap distances of 10, 20, and 30 nm, respectively. As for the effective mode volume $V_{\mathrm{m}}^{\text{eff}}$ (plotted as log scale in fig. S5C), it roughly follows the opposite trend of the field enhancement, which decreases as the gap distance decreases. The corresponding values are 3.2 × 10⁻⁵, 1.6 × 10⁻⁴, and 4.6 × 10⁻⁴ (λ/n)³ for gap distances of 10, 20, and 30 nm, respectively.

According to the literature, the Purcell factor (F_p) describes the modification of the spontaneous emission rate of a quantum emitter inside a resonant cavity and can be theoretically derived as $F_{\mathrm{p}}=\frac{3}{4{\mathrm{\pi }}^2}\frac{Q}{V_{\mathrm{m}}}{\left(\frac{\mathrm{\lambda }}{n}\right)}^3=\frac{3}{4{\mathrm{\pi }}^2}\frac{Q}{V_{\mathrm{m}}^{\text{eff}}}~\frac{{\mathrm{\gamma }}_{\text{cav}}}{{\mathrm{\gamma }}_0}$, where γ_cav and γ₀ refer to the spontaneous emission rates inside and out of cavity, respectively (16, 58). Despite the NPoM structures exhibiting a low quality factor or weak spectral confinement, the structure still provides pronounced enhancement with its extremely small mode volume. The expected Purcell factors, based on effective mode volumes, are 2800, 427, and 140 for NPoMs with 10-, 20- and 30-nm gap layers, respectively. Such a strong Purcell effect can facilitate the enhancement of spontaneous emission in these plasmonic nanocavity structures. On the basis of the in-plane field distribution and the field distribution near the edge surface of the nanocube, as shown in fig. S5 (D to F), the Purcell enhancement stemming from the gap-plasmon cavity is localized at the corner of the Ag nanocube and within the NPoM gap. Moreover, the electric field becomes increasingly confined with smaller gap distances, corresponding to the calculated increasing enhancement factor and consistent with other NPoM studies (13, 59).

Purcell-accelerated emission lifetime and Purcell-enhanced PL intensity

After the integration of D-PQDs and NPoM nanocavity, the PL characteristics of these three PQD samples were measured with homebuilt confocal microscopy. As shown in Fig. 3A, the spectra show no obvious spectral shift, which is consistent with the ensemble solution PL results (Fig. 1B), implying that there is no phase degradation and no perceptible modification from the low-Q cavity after the integration of the D-PQDs with the NPoM. Figure 3 (B and C) presents the TRPL decay curves for both coupled and uncoupled systems, alongside their corresponding baseline (background) measurements. Note that separate timescales are used for plotting the coupled and uncoupled decay curves because of the substantial different emission rates between coupled and uncoupled samples. In addition, fig. S6 separately displays the instrument response function (IRF) of the system, which exhibits an FWHM of ~50 ps. The average emission lifetime of D-PQDs on a glass substrate is 50 ns, and the decay is fitted by a biexponential model [$I\left(t\right)\sim A_1{e}^{-t/{\mathrm{\tau }}_1}+A_2{e}^{-t/{\mathrm{\tau }}_2}$], with the fast lifetime τ₁ = 2.82 ± 0.05 ns (contribution of 34%). The overall TRPL lifetime fitting results for both the ensemble D-PQD sample and single D-PQDs are now comprehensively presented in Table 1. All decay curves were successfully fitted using a biexponential decay model, and the lifetime components of the ensemble D-PQD sample are highly consistent with those measured for the single D-PQD, specifically the uncoupled emitters. The fast (τ₁) and slow (τ₂) component values show good agreement between the ensemble and single-emitter data, confirming that the characteristics of the single D-PQDs measured are representative of the material batch. The fluorescence lifetimes of the D-PQDs in NPoM are obviously shorter than that of the D-PQDs on glass. Purcell-enhanced emission from the band-edge exciton surpasses the relaxation pathway to defect states, leading to ultrafast band-edge exciton emission on picosecond timescales (60–62). To measure the lifetime of this fast emission, the fitting is performed by convolving the measured IRF (FWHM ~50 ps) with a biexponential function (24). Figure 3D shows the fastest lifetime of the single D-PQD in the NPoMs with different gap distances, with more than 10 dots in each sample recorded for statistical analysis. For gap distances of 30 and 20 nm, the emission lifetimes are shortened to fast ${\mathrm{\tau }}_1$ = 54 ± 2 and 35 ± 2 ps, respectively. On the basis of the intrinsic radiative lifetime (≈5.2 ns), the resulting Purcell enhancements are determined to be ~97 and 150, respectively. In comparison, the NPoM with a 10-nm gap layer shows an ultrashort lifetime of 12 ± 1 ps, which corresponds to a ≈435-fold acceleration.

Fig. 3. Enhanced emission and suppressed blinking of single PQDs coupled to NPoM.

(A) Emission spectra of single D-PQD on glass and single D-PQD in NPoM with 10- and 30-nm gap layers. (B) Time-resolved PL curves from a single PQD coupled to NPoM and on-glass. The fast component is limited by the IRF of the APD, also shown (light gray). (C) Time-resolved PL curves from a single PQD on glass. (D) Statistics of fast decay components for coupled single QDs and D-PQD on glass. (E) Excitation power dependence of the PL counts from a single PQD on glass and a single PQD coupled to NPoM. Dotted lines show linear fits to the data before saturation. To avoid the nonlinear response of APD, ND filter is used in measurement. (F) Emission time traces, corresponding g⁽²⁾(τ) curves, and the estimated average number of excitons (<N>) for a single D-PQD in a 10-nm gap NPoM cavity under elevated excitation intensities. The gray traces represent data collected under CW excitation (time trace bin time, 10 ms), while the black trace represents the g⁽²⁾(τ) measurement under pulsed excitation (2.5-MHz repetition rate). ND filters were used to optimize the count rates: 1 and 5 W/cm² (no filter), 20 to 500 W/cm² (OD 1), 1000 to 3000 W/cm² (OD 2), and 10,000 W/cm² (OD 3). The bin time for g⁽²⁾(τ) calculations is 1.2 ns.

Table 1. Biexponential lifetime fitting of D-PQD.

Lifetime fitting results of D-PQD by a biexponential model.

D-PQD	A 1	τ1 (ns)	A 2	τ2 (ns)
Ensemble	0.28	4.38	0.72	67.5
On glass	0.34	2.82	0.66	51.5
10-nm gap	0.98	0.012	0.02	0.58
20-nm gap	0.97	0.035	0.03	1.31
30-nm gap	0.95	0.054	0.05	2.67

We further compared the time-averaged emission rate from a single D-PQD on glass relative to a single D-PQD coupled to the nanocavity under 405-nm continuous-wave (CW) excitation (Fig. 3E). The structures with 30- and 20-nm gap layers show 40- and 70-times enhancement, respectively. In comparison, the highest PL intensity was obtained from a plasmonic nanocavity containing a 10-nm gap layer, primarily due to the optimal spectral overlap between the PQD emission and the plasmon resonance, as well as the smallest mode volume. A maximum single-photon emission rate exceeding 2.3 × 10⁹ counts/s was extracted before its saturated brightness, and the emission intensity was 250 times larger than that of a reference sample consisting of D-PQD on a glass substrate. To accurately quantify these high emission rates and ensure the linearity of the measurements, the total photon count rate is calculated by summing the signals detected by the two avalanche photodiodes (APDs) with the Hanbury-Brown and Twiss (HBT) experiment setup. Moreover, neutral density (ND) filters are introduced into the optical path when the measured PL intensity is high to attenuate the signal and maintain it within the linear response region of the APDs. As detailed in Fig. 3E, the overall dataset is composed of four distinct measurement regions, corresponding to measurements taken without an ND filter and with ND filters of optical density 1 (OD 1) 1 to OD 3. We emphasize that this represents the brightest room-temperature single-photon emission ever reported, delivering a brightness at least two orders of magnitude higher than previous records—namely, 9 × 10⁶ counts/s from noncavity CsPbI₃ PQD (63) and 1.5 × 10⁷ counts/s from PbS/CdS QD in a plasmonic nanocavity (64).

To elucidate the physical origin of this giant emission enhancement, we quantitatively decomposed the total enhancement factor into three distinct contributions: collection efficiency (η), excitation rate (${\mathrm{\gamma }}_{\text{exc}}$), and QY (24). First, regarding the collection efficiency, the radiation pattern from the NPoM is highly directional compared to the emission of a dipole on glass. With our high–numerical aperture (NA = 1.45) objective, this directionality theoretically boosts the collection efficiency from ≈37% (on glass) to ≈85% for the NPoM structure (16, 58), yielding a collection enhancement factor of η ≈ 2.3. Second, the excitation rate enhancement was evaluated via FDTD simulations performed specifically at the 405-nm excitation wavelength. The results indicate that the electric field intensity is notably amplified within the 10-nm gap, corresponding to an excitation enhancement (${\mathrm{\gamma }}_{\text{exc}}$) of ~118-fold. Third, the effective QY enhancement was derived from the statistical analysis of emission time traces (duty cycle of ON/OFF states). The analysis reveals that the effective PLQY recovers from ≈28% for PQDs on glass to ≈68% within the cavity, driven by the Purcell-enhanced radiative decay. This restoration contributes a factor of QY ≈ 2.4. Combining these contributions, the ideal emission intensity enhancement is estimated to be ~650-fold. We note that the experimentally observed enhancement (≈250-fold) is lower than this theoretical maximum. This discrepancy is primarily attributed to the spatial misalignment of the D-PQDs. While the theoretical calculation assumes that the emitter is located precisely at the corner of the gap, the PQDs in our experiment are randomly distributed within the gap layer, leading to a spatially average field enhancement that is naturally lower than the peak theoretical value.

Single-photon and nonblinking behavior at room temperature

Moreover, we investigated the single-photon characteristics of an individual CsPbI₃ PQD by using a homebuilt confocal microscope system integrated with an HBT interferometry setup at room temperature. Figure 4 (A, C, and E) shows the HBT results of D-PQD on glass and D-PQD in NPoM with different gap layers (10 and 30 nm). The single D-PQD without an optical cavity exhibits a second-order correlation value at zero time delay [g⁽²⁾(0)] with a value of 0.089, which implies high single-photon purity. In the case of D-PQD in the plasmonic nanocavity, both the 10- and 30-nm samples display antibunching characteristics, with g⁽²⁾(0) < 0.15. Furthermore, to confirm the single-photon emission characteristics under elevated excitation intensities, we measured the emission time trace and the corresponding second-order correlation function, g⁽²⁾(τ), for a single D-PQD placed in a 10-nm gap NPoM cavity using both pulsed and CW excitation sources (Fig. 3F). Consistent with recent findings (65), the emission remained nonblinking and stable across different excitation intensities. The measured g⁽²⁾(0) values consistently stayed between 0.12 and 0.19 under pulsed excitation and 0.17 and 0.23 under CW excitation. As g⁽²⁾(0) remains well below 0.5 in all cases, this confirms the single-photon emission characteristics of the detected signal, even under conditions of increased excitation power density. However, with a more extreme excitation intensity, the emission could contain more light from multiple excitons, resulting in poorer single-photon purity, as shown in fig. S7. Moreover, to ensure the statistical reliability of our findings, Fig. 4I presents the statistical distribution of g⁽²⁾(0) for each sample (on glass and in NPoMs at different gap distances), derived from measurements of more than 10 individual QDs. On the basis of the HBT experiment results, the D-PQD maintains high single-photon purity in the nanocavity structure under pulse operation, along with high brightness and fast emission rate.

Fig. 4. Photon statistics and photodynamic mechanisms of single D-PQDs in NPoM.

Second-order correlation function of single D-PQD (A) on glass and in NPoM with (C) a 30-nm gap layer and (E) 10-nm gap layer, with the values of g⁽²⁾(0) reaching 0.089, 0.115, and 0.105, respectively. The time traces of single-photon emission counts among 300 s of single D-PQD (B) on glass and in NPoM with (D) a 30-nm gap layer and (F) 10-nm gap layer. Additional time traces of single-photon emission counts for the D-PQD (G) on glass and (H) in NPoM with a 10-nm gap layer. NC, nanocavity. (I) Statistic of g⁽²⁾(0) of D-PQDs. Proposed photodynamic mechanisms of (J) bare PQDs on glass and (K) PQDs embedded in NPoM structures.

Although NPoM structures typically enhance biexciton emission, which can reduce single-photon purity via the biexciton-exciton cascade mechanism (66, 67), the D-PQD within our NPoM system consistently exhibits strong antibunching characteristics. We attribute this favorable result to selective plasmonic quenching of the biexciton state. This mechanism is likely facilitated by the PQD’s lack of a protective shell and its close proximity to the metallic surface. Considering the reported near-zero biexciton binding energy in CsPbI₃ at room temperature (68), the Coulomb interaction within the biexciton is presumed negligible. Consequently, the biexciton wave function is expected to be spatially larger than that of the exciton. This extended biexciton wave function may partially penetrate the metal film, thereby introducing a powerful, additional nonradiative quenching pathway. This disparity suggests that nonradiative quenching mechanisms dominate the biexciton dynamics, whereas the single-exciton emission remains primarily governed by the beneficial Purcell enhancement.

In addition to the high single-photon purity of CsPbI₃ D-PQD, we also found intriguing results when measuring the PL time traces of emission intensity (Fig. 4, B, D, and F). The D-PQD shows noticeable flicking during continuous excitation. In contrast, D-PQD in NPoM with a 10-nm gap layer exhibits nonblinking and nonflicking behavior with no appreciable PL decay; the time trace of emission intensity is very stable. While the D-PQD in NPoM with a 30-nm gap layer also displays blinking suppression, the suppression is not at the same level as that of the D-PQD in NPoM with a 10-nm gap layer. It is important to note that the excitation conditions used for the HBT measurements shown in Fig. 4 were intentionally controlled to avoid the nonlinearity of APD. Specifically, the QDs on glass were excited at 2.5 MHz (≈10 μW), whereas the QDs in cavities were excited at 20 MHz but with an extremely low power (≈5 nW). This substantial reduction in power density (four orders of magnitude lower) for the cavity samples was necessary because the high Purcell enhancement enables the QD to achieve the required counting rate at minimal excitation, thus preventing photocharging and preserving high single-photon purity. To mitigate the influence of individual QD variability (dot-to-dot variations) and establish a statistically robust foundation for our findings, we have substantially expanded our analysis. Specifically, in addition to the representative single D-PQD time traces shown in Fig. 4 (B, D, and F), we also present time traces for both uncoupled (on glass) and coupled (in NPoM with a 10-nm gap layer) systems in Fig. 4 (G and H). Furthermore, multiple supplementary time traces for each condition are included in figs. S8 to S10.

The mechanisms behind blinking and flicking have been extensively researched and are associated with the formation of nonemissive trions from environmental fluctuations or photocharging pathways, as illustrated in Fig. 4 (J and K) (69, 70). In light of this, we carefully reexamined the blinking and flickering mechanism in our CsPbI₃ D-PQD sample on glass. Consistent with prior studies, we performed the fluorescence lifetime-intensity distribution (FLID) analysis to figure out the underlying mechanisms (fig. S11). Our analysis revealed that both hot-carrier blinking and Auger-type blinking are evident in the single CsPbI₃ QD. On the basis of the FLID statistics, the exciton lifetime is determined to be ~5 to 6 ns, and the trion lifetime is less than 0.5 ns, which aligns well with reported literature values (71–73). This extremely short trion lifetime is primarily attributed to fast Auger recombination, causing the PLQY of trion to approach zero. Consequently, the observed blinking and flickering arise from the random switching between the neutral (emissive, high PLQY) and charged trion (nonemissive, near-zero PLQY) states of the QD. Through the isolation of the neutral exciton component via FLID analysis, the intrinsic radiative lifetime of the exciton is determined to be ~5 ns (corresponding to a radiative rate of 2 × 10⁸ s⁻¹). This intrinsic value is longer than the ensemble fitted result (~2.8 ns). This discrepancy is attributed to the glass substrate–induced nonradiative recombination channels, such as surface defects and Auger recombination, which reduce the effective lifetime and PLQY of the uncoupled nanocrystals.

We believe that blinking suppression stems from the substantial enhancement of the radiative process of excitonic emission by high Purcell enhancement, which dominates the photodynamics and markedly reduces the probability of trion formation (74, 75). Crucially, the enhanced radiative rate effectively competes with the nonradiative Auger recombination rate of the charged states (trions). In line with recent observations in other plasmon-coupled systems (28), this competition leads to a high QY even in the trion state, resulting in stable, nonintermittent emission intensity. This mechanism allows the nanocavity-coupled QD to exhibit bright, nonblinking PL while the underlying photophysical states (neutral exciton and trion) still switch randomly, potentially manifesting as a “blinking lifetime”—a definitive sign that the suppression of blinking intensity is due to increased quantum efficiency in the charged state, rather than the total elimination of trion formation. Apart from other blinking suppression strategies, integrating with NPoM provides an efficient way to maintain the PL spectrum and single-photon purity with considerably enhanced brightness and emission rate.

DISCUSSION

In conclusion, we successfully synthesized CsPbI₃ PQDs with high resistance to polar solvents by incorporating the DPIS zwitterions as capping ligands. With the excellent polar solvent resistance of these D-PQDs, we were able to integrate them into an NPoM plasmonic nanocavity. For samples with a 10-nm gap layer, we achieved 435-fold acceleration in the radiative emission rate and 250-fold enhancement in the total emission intensity. The shortest radiative lifetime observed was under 12 ps, which is near the instrument limitation, and a maximum single-photon emission rate surpassing 2.3 × 10⁹ counts/s before the saturated brightness at room temperature was also recorded. In contrast, the NPoM with 20- and 30-nm gap layers displayed relatively “moderate” Purcell enhancement resulting from diminished field confinement within the nanogap; the experimental results showed good agreement with FDTD simulations. Furthermore, although blinking has been commonly observed in single PQDs because of elevated photocharging rates at ambient temperatures, ultrastable, nonblinking single-photon emission with high single-photon purity was successfully demonstrated with our D-PQDs in the NPoM with a 10-nm gap layer at room temperature. Combining these unrivaled characteristics, the architecture presented here could provide important guidance for the design of plasmonic nanocavities coupled with perovskite-based quantum emitters and could also be used for future high-speed displays and quantum photonic applications.

MATERIALS AND METHODS

Synthesis of CsPbI₃ P-PQDs and D-PQDs

The CsPbI₃ P-PQDs were prepared by the spray synthesis method adopted from the traditional hot-injection method. Cs-oleate was prepared by dissolving 0.064 g (0.196 mmol) of Cs₂CO₃ (99.995%, Sigma-Aldrich) in 2.4 ml of 1-octadecene (90%, Sigma-Aldrich) with 0.2 ml of OA (90%, Sigma-Aldrich). The solution was heated to 300°C for complete dissolution of Cs₂CO₃. Meanwhile, 520 mg (1.128 mmol) of PbI₂ (99.9985%, Thermo Fisher Scientific) was dissolved in 30 ml of 1-octadecene, 3 ml of OA, and 3 ml of OAm (80 to 90%, Acros) at 120°C for the preparation of Pb-oleate, followed by heating up and holding at 180°C for the reaction of spray synthesis. Then, Cs-oleate was swiftly sprayed into the Pb-oleate solution, accompanied by the formation of CsPbI₃ P-PQDs. The mixture was quenched to room temperature in an ice bath after the reaction time of 10 s. The abovementioned steps were all performed in a N₂-filled glovebox. The products were centrifuged at 5000 rpm for 10 min after quenching. The supernatant was discarded, and the precipitates were redispersed in hexane (99%, Macron Chemicals) for further centrifugation at 5000 rpm for 20 min. The clear supernatant solution was collected and stored in a refrigerator. Half amount of OAm (1.5 ml) was replaced with 60 mg (0.195 mmol) of 3-(decyldimethylammonio) propane sulfonate inner salt (Sigma-Aldrich) for the preparation of CsPbI₃ D-PQDs.

Sample fabrication

The plasmonic nanocavity fabrication began with a deposition of a flat 200-nm Ag film onto a clean glass substrate by sputtering. Consequently, a solution of PS and CsPbI₃ D-PQD was spin coated onto the Ag film. To prepare the solution for spin coating, PS was fully dissolved in toluene for 1 hour, and the D-PQD was diluted to a concentration of 1 mg/ml in toluene. The solutions of PS and D-PQD were mixed and diluted to a final D-PQD concentration of 2 μg/ml and PS concentrations ranging from 1.7 to 5 mg/ml for different thicknesses. The PS/D-PQD films were spin coated onto the Ag film at 1000 rpm for 30 s to produce films with 10- to 30-nm thickness.

Ag nanocubes were purchased from nanoComposix Inc. with an average edge size of 100 nm in a 1 mg/ml ethanol solution. The nanocubes were coated with a ~3-nm PVP stabilizer layer, and we redispersed nanocubes to IPA to reduce the polarity of the solution. Then, a solution of Ag nanocubes (2 μl) was drop-casted onto PS/D-PQD on the Ag film, with the aid of sonication at the same time to ensure the uniform distribution of nanocubes.

Material characterization

Absorption spectra in the ultraviolet (UV)-visible region were acquired by a Shimadzu UV-2600 spectrophotometer. The PL spectra of the ensemble PQDs were acquired with a charge-coupled device camera (PIXIS256BR, Princeton Instruments) by exciting the sample with a 375-nm diode laser. PL lifetimes were collected by a photomultiplier tube (PMC-100-1, Becker & Hickl GmbH) and a time-correlated single photon counting module (SPC-150, Becker & Hickl GmbH), with a picosecond pulsed laser (LDH-P-C-375 M, PicoQuant) as the pumping source. TEM and HRTEM (high-resolution TEM) images and energy-dispersive x-ray spectroscopy elemental spectra were recorded via an HRTEM-200F microscope operating at an accelerating voltage of 200 kV. SEM top-view and cross-sectional-view images were recorded by a JSM-IT800 Scanning Electron Microscope. The chemical structures of the samples were analyzed using FTIR spectroscopy (Bruker, Tensor 27). X-ray photoelectron spectra were acquired by electron spectroscopy for chemical analysis (PHI 5000 VersaProbe II, ULVAC-PHI).

Optical measurements

This study of dark-field microscopy configuration used a reflected dark-field microscopy configuration, using the objective’s annular illumination aperture to generate a hollow conical light beam incident upon the sample at high oblique angles. The system is designed so that the illumination NA is greater than the central collection NA, which effectively blocks specular reflection from the flat substrate and suppresses background noise. Only light scattered or reradiated by nanoscale structures—specifically, NPoM cavities—is redirected into the lower-angle collection channel. This method enables high-contrast dark-field imaging by highlighting individual NPoM cavities with strong signal-to-background discrimination while minimizing contributions from the surrounding substrate.

The measurements of single-photon emission and plasmonic nanocavities were performed using a self-modified confocal microscope (SouthPort Co.). The samples were excited by a single-mode 405-nm pulsed diode laser (LDH-P-C-405B, PicoQuant) or a 405-nm CW laser (Civil Laser) with an oil-immersion objective lens (UplanSApo, Olympus; NA, 1.45). The 405-nm pulsed laser was used as the pumping source for measuring photon antibunching and PL lifetime, whereas the 405-nm CW laser was used for the measurements of PL spectra and power-dependent PL intensity. To facilitate a valid comparison between these two regimes, the pulsed energy density ($N_{\text{Pulse}}$) was calibrated to the CW power density ($J_{\text{CW}}$) using the relationship $N_{\text{Pulse}}={\mathrm{\tau }}_1\times J_{\text{CW}}$, where ${\mathrm{\tau }}_1$ denotes the fast component of the enhanced emission lifetime. This calibration equates the incident energy per pulse to the energy integrated over one emission lifetime under CW excitation. The lasers were focused on the samples with spot sizes of ~0.35 μm (oil-immersion objective lens), corresponding to the diffraction limits based on the Rayleigh criterion. The second-order correlation function g⁽²⁾(t) was measured by using an HBT optical setup, and the emitted photons were collected by the APDs (SPCM-AQRH-16, Excelitas) connected to a time-correlated single photon counter (Hydraharp 400, PicoQuant). The PL spectra of PQDs in plasmonic nanocavities were measured with a spectrometer equipped with an intensified charge-coupled device camera. The confocal images, PL time traces, PL lifetimes, PL spectra, and values of g⁽²⁾(0) of the samples were all realized in this system.

FDTD simulation

The simulations were performed using OptiFDTD. In the simulations, the total-field scattered-field source was used in a simulation box containing the NPoM structure. The plasmonic nanocavities were excited using a plane wave incident at 0° relative to the surface normal to the Ag film and the excitation electromagnetic source set to 680 nm in wavelength to evaluate the response under resonance condition. In the calculation, a mesh with a size smaller than 1 nm was used. The mesh size was below 0.5 nm with small gap distances. The dielectric function of the Ag metal film and the nanocubes was from Johnson and Christy’s data (76).

Supplementary Materials

This PDF file includes:

Figs. S1 to S11