Room-Temperature, Highly Pure Single-Photon Sources from All-Inorganic Lead Halide Perovskite Quantum Dots.

Attaining pure single-photon emission is key for many quantum technologies, from optical quantum computing to quantum key distribution and quantum imaging. The past 20 years have seen the development of several solid-state quantum emitters, but most of them require highly sophisticated techniques (e.g., ultrahigh vacuum growth methods and cryostats for low-temperature operation). The system complexity may be significantly reduced by employing quantum emitters capable of working at room temperature. Here, we present a systematic study across ∼170 photostable single CsPbX3 (X: Br and I) colloidal quantum dots (QDs) of different sizes and compositions, unveiling that increasing quantum confinement is an effective strategy for maximizing single-photon purity due to the suppressed biexciton quantum yield. Leveraging the latter, we achieve 98% single-photon purity (g(2)(0) as low as 2%) from a cavity-free, nonresonantly excited single 6.6 nm CsPbI3 QDs, showcasing the great potential of CsPbX3 QDs as room-temperature highly pure single-photon sources for quantum technologies.

Introduction

Single-photon sources (SPSs) could play an important role in several emerging quantum technologies.[1] They could be employed in quantum key distribution (QKD) protocols (e.g., BB84), which exploit specific properties of quantum states (e.g., the noncloning theorem) to secure the distribution of symmetric encryption keys.[2] SPSs are also exploited for quantum-imaging applications, where the peculiar photon statistics of single QDs is utilized to enhance sensitivity as well as spatial resolution.[3,4] Furthermore, indistinguishable single photons are the workhorses of optical quantum computing (boson sampling), potentially offering a solution to computing tasks that are hard to implement with classical computers.[5,6] An ideal SPS emits single indistinguishable photons on demand with a vanishing probability of multiphoton emission. The single-photon purity is characterized by the second-order intensity correlation functionas a function of the delay time τ between two-photon detection events, measured in a Hanbury Brown and Twiss (HBT) experiment. The quality of an SPS is evidenced by a dip at zero time delay, also termed photon antibunching, where g(2)(0) = 0 represents the ideal case. As a practical criterion, an experimental source is considered to be a SPS if g(2)(0) < 0.5. Specifically, QKD and quantum imaging applications require high single-photon purity (e.g., g(2)(0) much lower than 0.1).[1]

For boson-sampling applications, it is crucial that the single photons are quantum mechanically indistinguishable, so photon–photon gates could be implemented with linear optics, produced and collected with sufficient efficiency, to ensure scalability.[6,7] Practical deployment into devices might benefit from an “ideal” SPS that is produced at low cost, operates at room temperature (RT) and at a high (GHz) rate (requiring sub-ns emission lifetime), and can be electrically driven for further device miniaturization.[8−10] A tunable emission wavelength, potentially reaching the near-infrared region (800 nm, 1.3–1.5 μm) would facilitate interfacing of SPS with optical fibers for long-distance quantum communications. Promising materials systems developed in the past 20 years include trapped atoms/ions,[11−16] molecules,[17−19] color centers in diamond,[20−26] monolayer transition metal dichalcogenides,[27−30] epitaxial[9,31−36] and colloidal[37−40] semiconductor quantum dots (QDs), operated at RT and cryogenic temperature. However, since each of the currently explored platforms presents its own challenges and advantages,[1] the quest for a quantum emitter satisfying all the above requirements simultaneously is still ongoing.

To date, III–V semiconductor QDs are the best-performing SPS platform in terms of efficiency and indistinguishability. For example, resonantly excited InGaAs QDs embedded in microcavities provide over 97% indistinguishability for thousands of emitted photons, g(2)(0) ∼ 2–3%, and a high efficiency of around 60% at a repetition rate of 1 GHz.[35] However, such outstanding performance comes at the cost of high complexity in both material fabrication (typically requiring growth via molecular-beam epitaxy) as well as operation (typically requiring cryogenic temperatures).[41]

In contrast, colloidal QDs recently emerged as promising alternative SPS enabling RT operation,[40] with the compelling opportunity of electrically driven single-photon devices.[42] The newest member in this material class are lead halide perovskite QDs, of particular interest due to their low-cost synthesis, solution processability, tunability of the emission wavelength via size and composition, narrow-band emission, short radiative lifetime (approximately nanoseconds at RT), as well as high photoluminescence (PL) quantum yield (QY).[43,44] While this compelling set of attributes has already driven the development of a new generation of optoelectronic devices including lasers,[45−48] LEDs,[49−52] and photodetectors,[53−56] lead halide perovskite QDs have recently also indicated their promise at the single-QD level,[57−67] demonstrating single-photon emission with a high rate at RT in single-QD PL measurements.[68−74] Furthermore, single CsPbBr3 QDs have showcased significant potential as coherent photon sources,[75−77] demonstrating the beneficial combination of long coherence times and short radiative lifetimes[61] at cryogenic temperatures, without the need for embedding them in optical microcavities. However, the mechanisms underlying their single-photon purity are still not fully understood, as systematic studies (e.g., aimed at exploring several QD sizes and different material formulations) were hindered by their poor optical stability, particularly strong for small QDs with edge length shorter than 7–8 nm. Typical signatures of photodegradation are a dynamic blueshift of the PL emission as a result of light-triggered QD structural transformations. Such insufficient optical stability under ambient conditions can be attributed to the soft ionic lattice and the highly dynamic binding between capping ligands and QD surface.[78] In response to these challenges, recent efforts have shown that the detrimental photodegradation in ambient conditions could be suppressed by embedding the QDs in a protective matrix to prevent exposure to moisture.[79−84]

Here we report single QD spectroscopy studies combined with photon-correlation measurements on single perovskite QDs of different composition (CsPbBr3 and CsPbI3) and size (down to 6.6 nm). By systematically studying spectrally stable QDs, obtained by proper encapsulation procedures, we demonstrate that improved single-photon purity can be obtained from strongly quantum-confined QDs, with g(2)(0) as low as ∼ 2% for 6.6 nm CsPbI3 QDs; these results are in line with the enhanced Auger-mediated nonradiative biexciton recombination in strong quantum confinement.[85−87]

Results and Discussion

For our systematic study of the effect of quantum confinement on the single-photon purity in perovskite QDs, three QD samples of different sizes and composition were studied: CsPbBr3 QDs with a mean edge length of 10 nm, as well as CsPbI3 QDs with mean edge lengths of 6.6 and 10 nm, respectively. Figure a displays the PL emission spectra of all samples, with an inset showing a TEM image of the 6.6 nm CsPbI3 QDs. Size histograms of these three samples are shown in Figure S1. Further details on the QD synthesis according to previously published procedures,[88,89] as well as sample preparation, can be found in the Supporting Information. In the following, we will first assess the degree of quantum confinement in our samples and then discuss how such size- and composition-dependent confinement is expected to affect Auger rates, biexciton emission, and single-photon purity. Finally, experiments are presented to verify our hypothesis.

Figure 1

Optical properties of CsPbX3 QDs. (a) PL spectra at RT from an ensemble of 10 nm CsPbBr3 QDs (dispersed in toluene, centered at 2.419 eV), 6.6 nm CsPbI3 (dispersed in hexane, centered at 1.895 eV), and 10 nm CsPbI3 (dispersed in toluene, centered at 1.816 eV). Inset: TEM image of the 6.6 nm CsPbI3 QDs. (b) Illustration of the dependence of the Auger rate on QD size and composition. Gray dashed circles represent the respective Bohr diameter (BD) in CsPbBr3 and CsPbI3. (c) Schematic of a typical g(2)(τ) trace under pulsed laser excitation. g(2)(0) is defined as the ratio between the area A1 of the peak at zero time delay to the average area A2 of the side peaks, which is approximately the ratio between biexciton (BX) and exciton (X) PL quantum yield (QY).

On the basis of the Bohr diameters (BD) of CsPbBr3 (7 nm) and CsPbI3 (12 nm),[43] our studied QD samples cover the range from weak/intermediate to strong confinement, with relative sizes of 1.43 BD (10 nm CsPbBr3), 0.83 BD (10 nm CsPbI3), and 0.55 BD (6.6 nm CsPbI3), respectively. Figure b schematically shows how quantum confinement and, hence, Auger rates[90,91] scale with composition and size of the QD. For each composition, larger Auger rates are expected in smaller QDs;[85−87] for QDs of the same size, the composition with a larger BD is expected to exhibit larger Auger rates. To investigate how quantum confinement enhances the Auger rates and the single-photon purity, we will carry out photon-correlation measurements in a HBT setup under pulsed laser excitation (405 nm, 10 MHz repetition rate). Figure c schematically displays the expected g(2)(τ) trace from a single emitter, exhibiting a series of discrete peaks temporally separated by the inverse of the laser repetition rate. The peak at zero time delay (i.e., g(2)(τ = 0)) reflects the probability that the two detected photons in the HBT setup originated from the same laser pulse, while the adjacent peaks report coincidence events originating from consecutive laser pulses.[92] Any residual of g(2)(0) > 0 may be attributed to the probability of biexciton creation and subsequent two-photon emission,[92,93] eventually imposing a limit on the maximum attainable single-photon purity.

In this work, the g(2)(0) value refers to the ratio of the integrated area of the zero-delay peak to the average area of the first three side peaks. The areas are determined by fitting single-exponential decay to the experimental data (following eqs S3 and S4). It is worth mentioning that PL blinking can, in principle, lead to a bunching effect (g(2)(τ) > 1) at short time delays (i.e., in the first few nanoseconds to milliseconds), depending on the blinking time scales.[70,94−96] However, based on a detailed discussion of this effect in the Supporting Information, PL blinking in our QDs contributes only marginally to the estimation of the single-photon purity.

We started the exploration by measuring photon statistics of 10 nm CsPbBr3 QDs. To directly relate the acquired g(2)(τ) to the emission energy of the studied QD and to exclude photodegradation effects in our analysis, we acquired PL spectra before and after the acquisition of g(2)(τ); see Figure a for a representative time series of PL spectra. The shaded area denoted as “g(2) measurements” corresponds to the time during which the emission is rerouted for g(2)(τ) measurements, and only dark counts are observed in the PL spectra. Importantly, the time series attests bright and photostable emission over the entire acquisition time (i.e., both before and after the g(2)(τ) measurements). Such photostability is also evidenced via single-Lorentzian fitting (eq S2) on each of the acquired spectra: Both the fitted PL peak center of 2.423 ± 0.003 eV and the line width (full width at half-maximum, FWHM) of 72 ± 2 meV exhibit only a small standard deviation. The corresponding g(2)(τ) is plotted in Figure b, where the obtained g(2)(0) = 27% < 50% indicates that we are studying a single QD. As discussed above, the finite value of 27% suggests contribution from radiative biexciton emission.

Figure 2

CsPbBr3 QDs (10 nm) at the single particle level. (a) Time-series of PL spectra from a single 10 nm CsPbBr3 QD in inert atmosphere with 1 s of binning time. The time window denoted as g(2) measurements indicates the time during which the g(2)(τ) is acquired. (b) Corresponding g(2)(τ) trace normalized to the average amplitude of the peaks at τ = −0.1, 0.1, and 0.2 μs (see the Supporting Information for a detailed discussion). g(2)(0) = 0.27 is obtained from exponential fits (see Supporting Information for details). (c) g(2)(0) as a function of the central energy of the PL, acquired from over 65 QDs. Error bars represent the spectral dynamics within the acquisition time, typically < 5 meV, indicating spectrally stable QDs. (d) Line width (FWHM) as a function of the PL peak center, obtained from single Lorentzian fitting. Error bars indicate the spectral dynamics of peak center and line width. (e) Time series of PL spectra from a single 10 nm CsPbBr3 QD under ambient conditions, featuring rapid blue-shifting of the PL. (f) Corresponding normalized g(2)(τ) trace exhibits g(2)(0) = 0.025.

Benefiting from the photostable PL emission, we explore g(2)(0) as a function of emission energy, hereby addressing the effect of quantum confinement on single-photon purity. To this end, we acquired PL spectral time series and g(2)(τ) for over 65 CsPbBr3 QDs from the same batch. Figure c,d correlate the thereby obtained g(2)(0) and PL line width with the QD emission energy. All studied QDs have less than 7 meV (most < 5 meV) standard deviation in emission energy, confirming the spectrally stable emission during g(2)(τ) acquisition. As shown in Figure c, we observe that g(2)(0) decreases from 0.8 to 0.1 for emission energies increasing from 2.415 to 2.455 eV, in agreement with the trend reported in ref (70). Since within a single QD batch the PL emission energy can be directly linked to the QD core size, the observed trend suggests a gain in single-photon purity upon increasing the degree of quantum confinement. This clear trend is also consistent with our hypothesis that the stronger quantum confinement suppresses biexciton emission due to accelerated Auger decay.[85−87,93] QD-to-QD variations in the g(2)(0) values with the same emission energy could arise from the subtle variations in quality and/or smoothness of the QD-core/ligand interfaces affecting the sensitive Auger mechanism without a measurable change in emission energy.[92] Narrow line widths of 70–80 meV are obtained with emission energies ranging from 2.415 to 2.455 eV. The only weakly size-dependent line width suggests similar exciton surface-phonon coupling strengths for those QDs in weak/intermediate quantum confinement.[97] QDs with a high g(2)(0) (> 50%), as observed in our data, reflect pronounced biexciton emission. While achieving a high biexciton QY is desired for the realization of efficient lasers and entangled photon sources for optoelectronics and quantum technologies, such QDs cannot be implemented as SPSs.

Carrying out the measurement at ambient conditions (no encapsulation), continuous PL blue-shifting with photobleaching is observed as shown in Figure e. This observation has frequently been reported[68,69,79,98,99] and attributed to the core-size reduction resulting from an irreversible light-triggered degradation in the presence of oxygen and moisture. The corresponding g(2)(τ) plotted in Figure f shows a low g(2)(0) ≈ 2.5%, in line with early reports of room-temperature single-photon emission from perovskite QDs.[68,70,100,101] In our case, although the pristine QD originally featured a size of ∼10 nm, its PL emission blue-shifted rapidly to ∼ 2.56 eV. Hence, even before the start of the g(2)(τ) acquisition, the transformed QD has already assumed an equivalent core size below 5 nm based on the PL emission characteristic. This low g(2)(0) of 2.5% is comparable to the value reported in ref (70), with similar size and comparable emission energy. Despite their good single-photon purity, such QDs undergoing rapid photodegradation cannot be implemented in practical applications of SPSs. Nevertheless, the low g(2)(0) for blue-shifting QDs highlights the beneficial role of quantum confinement for suppressing biexciton emission.

In the search for an even higher single-photon purity, we now proceed to studying CsPbI3 QDs, motivated by their stronger quantum confinement based on the larger CsPbI3 Bohr diameter (see Figure b). To date, CsPbI3 QDs have received less attention in single-QD studies at room temperature, in part due to their lower stability compared to CsPbBr3 QDs. To achieve equivalent spectral stability as for the previously reported CsPbBr3 QDs, we add an epoxy-encapsulation step to the polystyrene-embedding of the QDs to further protect the QDs against oxygen and moisture (see the Supporting Information for more details). Figure a displays the time series of the PL spectra for CsPbI3 QDs with an average edge length of 10 nm, demonstrating stable PL emission with average emission energy at 1.851 ± 0.002 eV. Figure b displays the corresponding g(2)(τ) with g(2)(0) ≈ 8.5%. Compared to 10 nm CsPbBr3 QDs whose mean and lowest g(2)(0) are around 35% and 10%, respectively (see Figure c), the 10 nm CsPbI3 QDs show improved single-photon purity with a mean and lowest g(2)(0) values of 18% and 5.6%, respectively (see Figure d). Over half of the studied 10 nm CsPbI3 dots exhibit g(2)(0) below 15%. This significant improvement in single-photon purity achieved by changing the material formulation and exciton Bohr diameter supports our hypothesis of confinement-controlled multiphoton emission.

Figure 3

Single-photon purity vs QD composition. (a) Time series from a single 10 nm CsPbI3 QD in inert atmosphere. (b) Corresponding normalized g(2)(τ) trace, with g(2)(0) = 0.085 obtained from exponential fits. (c) and (d) Histograms of g(2)(0) values of 10 nm CsPbBr3 and CsPbI3 QDs, respectively.

To further increase quantum confinement and improve single-photon purity, we explored CsPbI3 QDs with an average edge length of 6.6 nm. To the best of our knowledge, single-QD spectroscopy studies of these QDs have not been reported yet due to their fast photodegradation and short shelf life. However, in this work, epoxy-encapsulation enables us to obtain spectrally stable PL, free of photodegradation. The time series of PL spectra from such QDs shown in Figure a indeed attests to stable emission (at 1.900 ± 0.005 eV) over a very long acquisition period. The encapsulation also enabled a first systematic characterization of the size-dependent emission broadening in CsPbI3 QDs at the single-particle level, with a linear increase of the PL line width for smaller QDs (see Figure b). A similar trend has previously also been observed for colloidal CdSe[102,103] and perovskite CsPbX3 QDs,[97,104] attesting to a kind of universal trend in emission line broadening versus nanoparticle sizes, possibly triggered by an enhanced coupling to surface-phonon modes. Figure c reports the intensity–time trace and corresponding intensity histogram acquired by avalanche photodiodes (APDs) through the time-tagged time-resolved (TTTR) method.[105] The intensity–time trace and histogram reflect that the QD remains in its bright emissive state, with a count rate around 700 counts per 10 ms, for most of the time (the ON fraction is higher than 92% with a threshold of 200 counts per 10 ms). Additional PL blinking traces can be found in Figures S3 and S5. The bright emissive state stems from the exciton recombination, while the low-intensity state originates from trions or biexcitons with strongly quenched PL QY, as reported in CdSe QDs[106,107] and perovskite QDs.[66,68] In addition, the occurrence of events with intermediate brightness (i.e., a “gray state”) points to either trion/biexciton emission with relatively high PL QY or to fast switching between the emissive excitonic state and trion/biexciton states, with switching times shorter than the employed binning time (10 ms). The g(2)(τ) (see Figure d) attests very high single-photon purity with g(2)(0) ≈ 2%, comparable to the best reported values for colloidal QDs.[40,42,69]Figure e shows a histogram of the g(2)(0) values across 77 of such 6.6 nm CsPbI3 QDs. We report a mean g(2)(0) around 9.6%, with 30 QDs showing g(2)(0) ≤ 5% and over 50 QDs showing g(2)(0) ≤ 10%. The dependence of g(2)(0) on the emission energy is given in Figure S2, featuring a systematically lower g(2)(0) values for larger emission energy. The large number of QDs with ultralow g(2)(0) values firmly establishes quantum confinement as a solid design principle toward ultrapure single-photon emission, while the narrow distribution of g(2)(0) and the newly achieved photostability highlight the importance of a high degree of material control for achieving reproducible SPS operation. It is worth mentioning that alternative strategies to achieve high single-photon purity have been explored which utilized a temporal filtering scheme to reject multiphonons emitted by biexcitons, as reported in refs (108−111), with typical g(2)(0) values of < 0.01; in addition, embedding QDs into microcavities and exploring resonant excitation schemes, as reported in ref (34), proved to be effective in obtaining g(2)(0) ≈ 0.0075% at cryogenic temepratures.

Figure 4

Ultrahigh single-photon purity in strongly confined 6.6 nm CsPbI3 QDs. (a) Time series from a single 6.6 nm CsPbI3 QD in inert atmosphere, featuring a PL center at 1.900 ± 0.005 eV across the entire acquisition period. (b) Line width as a function of the PL peak center, assembled from 77 of such 6.6 nm CsPbI3 QDs; the line width increases from 70 to 140 meV for PL peak energies increasing from 1.80 to 2.00 eV. (c) Intensity time trace with a binning time of 10 ms acquired by APDs through the time-tagged time-resolved (TTTR) method. Emissive bright states exhibit count rates of around 700 counts per 10 ms. The intensity histogram is plotted on the right. The red line in both graphs represents the threshold for ON-states. (d) One of the best g(2)(τ) traces, with g(2)(0) = 0.02, obtained from exponential fitting. (e) Histogram of g(2)(0) values for 6.6 nm CsPbI3 QDs with a mean value of around 9.6%. Around 39% of the QDs feature g(2)(0) ≤ 5%, and about 65% of QDs feature g(2)(0) ≤ 10%.

Conclusions

In conclusion, by systematically studying single CsPbX3 QDs of different sizes and compositions, we established quantum confinement as an effective control knob for ultrahigh single-photon purity. Stronger quantum confinement improves single-photon purity via suppressed biexciton emission, likely due to enhanced Auger rates in smaller QDs. These observations were enabled by an improved sample preparation protocol providing enhanced QD photostability. For bright and photostable CsPbI3 QDs with edge lengths of approximately 6.6 nm, we report g(2)(0) values as low as ∼ 2%, equal to the current record for colloidal QDs. Our results help to elucidate the photophysics of single-photon emission in single perovskite QDs and suggest strong quantum confinement as an effective material design rule for optimized single-photon sources.