A room-temperature single-photon source based on strongly interacting Rydberg atoms

See allHide authors and affiliations

Science  26 Oct 2018:
Vol. 362, Issue 6413, pp. 446-449
DOI: 10.1126/science.aau1949

Single photons from inflated atoms

Single-photon emitters are of interest for many applications, such as quantum sensing and quantum secure communication. Although efficient on-demand solid-state sources based on quantum dots, defect color centers in diamond, and single molecules are available, most of these systems require cryogenic temperatures for optimal performance. Ripka et al. used a cloud of warm excited rubidium atoms confined to a gas cell and exploited the exaggerated interaction of Rydberg states to generate single photons on demand. The results demonstrate the viability of atomic gases for application in the development of quantum technologies.

Science, this issue p. 446


Tailored quantum states of light can be created via a transfer of collective quantum states of matter to light modes. Such collective quantum states emerge in interacting many-body systems if thermal fluctuations are overcome by sufficient interaction strengths. Therefore, ultracold temperatures or strong confinement are typically required. We show that the exaggerated interactions between Rydberg atoms allow for collective quantum states even above room temperature. The emerging Rydberg interactions lead both to suppression of multiple Rydberg state excitations and destructive interference due to polariton dephasing. We experimentally implemented a four-wave mixing scheme to demonstrate an on-demand single-photon source. The combination of glass cell technology, identical atoms, and operation around room temperature promises scalability and integrability. This approach has the potential for various applications in quantum information processing and communication.

Single-photon emitters are of interest for manifold applications in quantum computation, simulation, sensing, and especially in quantum secure communication. In particular, the latter will demand efficient quantum repeaters (1) that require efficient single-photon creation and storage schemes, ideally at the same platform. Room-temperature alkali gases can perfectly serve as such storage media, and second-long storage times have already been demonstrated for weak coherent states (2). However, the generation of single photons with stable wavelength and adapted bandwidth is still a challenge.

The first observation of antibunched photon counting statistics from single-atom fluorescence dates back to 40 years ago (3). Today, very efficient on-demand solid-state sources based on quantum dots (4), color centers (5), and single molecules (6) are available. Although in principle most of these systems are working at room temperature as well (79), they do require cryogenic temperatures for optimal performance. Because they are solid-state embedded, they suffer from interactions with phonons, spin noise, strain, and drifting electric fields. These result in large variations in frequencies, spectral wandering, and additional phase noise causing spectral diffusion, which is a fundamental limitation for scalability.

In contrast, atoms (3) and ions (10) inherently produce spectrally indistinguishable single photons. Here, the highest fidelities have been achieved with ultracold atoms inside high-finesse cavities (11). Small photonic networks have been realized with them (12). With room-temperature gases, so far, heralded single photons with sub-Doppler linewidth can be created (13). There, two photons are generated in a probabilistic process; one of them then gives the information on the existence of the other one. However, as with parametric down-conversion (14), they undergo a spontaneous generation process, and this requires a triggered memory for temporal synchronization of the created photons.

A rather new approach makes use of strongly interacting Rydberg atoms, which suppress multiple excitations within a certain radius (15). Related to this, Rydberg interactions can also lead to dephasing between Rydberg polaritons (16, 17). These interaction mechanisms can be converted into an effective optical nonlinearity, even on the single-photon level (18). On the basis of the Rydberg blockade effect in ultracold samples, researchers have demonstrated photon antibunching (19, 20) and even higher-order correlated light states (21).

However, Rydberg-excited atoms can be coherently excited and optically probed not only at ultralow temperatures but also in thermal vapors (22). Nanosecond-pulsed and intense excitation fields at GHz Rabi coupling strengths allow for coherent dynamics faster than 1 ns per Rabi cycle (23). On this time scale, thermal atoms move less than the optical wavelength and can be assumed to be nearly frozen. As soon as they are excited, the collective excitation can be retrieved within 1.2 ns, before dephasing due to spatial inhomogeneities and the Doppler effect set in (24). For strong Rydberg interaction effects to be observable, the atom-atom interaction strength must be larger than the excitation bandwidth. The equivalent of 3 GHz of van der Waals–type Rydberg interactions can be achieved in a dephasing regime where the excitation volume is larger than the interaction distance (25).

Blockade of Rydberg excitations occurs at distances shorter than the blockade radius rB = (C6/Ωeff)1/6, where C6 is the van der Waals coefficient, is the reduced Planck constant, and Ωeff is the effective two-photon coupling strength (26). If we assume that the addressed Rydberg state is 40S1/2 and that Ωeff = 400 MHz, then (27) C6 = 2π × 806.3 MHz μm6 and the blockade radius results in rB = 1.1 ± 0.1 μm. For multiple excitations in an atomic ensemble to be suppressed by Rydberg blockade, the size of the ensemble must be smaller than that radius. Rubidium atoms can be coherently excited to Rydberg states in vapor cells of such a size, without being limited by atom-wall interactions (28). Therefore, it has been predicted that a four-wave mixing experiment in a microscopic cell will result in a suppression of multiple excitations and hence to an emission of single photons (29). Note that even when multiple Rydberg excitations cannot be prevented, the superposition of the shifted excitations leads to destructive interference in the detection mode (16). It is also important to note that the size of such a small ensemble is still larger than the optical wavelength. This guarantees a directed emission pattern during the four-wave mixing process (30). Besides, the emitted photon reflects the shapes of the excitation pulses and is not created in a spontaneous emission process. Furthermore, the emitted photons do not undergo any jitter in either the center frequency or the bandwidth.

To truncate the Rydberg-excited ensembles to μm3 volumes, our setup focuses the 795-nm excitation beam to a waist with w0 = 1.45 ± 0.05 μm Gaussian 1/e2 diameter (Fig. 1C). The vapor cell is designed in a wedge-shaped geometry and provides a distance range from 0 to several tens of micrometers (Fig. 1, A and B). Thus, we can choose the degree of longitudinal confinement of the atomic ensemble. The cell is filled with rubidium at natural abundance and the experiment is done using the 85Rb isotope. The vapor pressure is set by a temperature of 140°C, which corresponds to a steady-state density of approximately 30 atoms per μm3 in the hyperfine ground state F = 3. Yet this is not sufficient for an efficient single-photon generation. However, because a large number of atoms are attached to the surfaces of the glass windows, we desorb a fraction of them via light-induced atomic desorption (31). A 2.5-ns pulse at 480 nm raises the number of atoms to more than 1000 per μm3 within 5 ns (Fig. 2B). The Rydberg-excited volume then contains about 2000 atoms. In the following few nanoseconds, the optical thickness decreases again as they hit the opposite surface.

Fig. 1 Experimental setup.

(A) Photograph of the wedge-shaped microcell. The outer dimensions are 25 mm × 50 mm. The glass windows have a thickness of 1.21 ± 0.01 mm, matching the high–numeric aperture aspheric lenses (focusing and collimation), so that their diffraction limits are reached. The inner distance of the glass windows varies from touching each other to about 100 μm. The thicknesses are determined via residual-finesse resonances. In the region of 400 nm to few μm, interference effects can be observed (so-called Newton’s rings). (B) Schematic of the setup. (C) The excitation volume inside the vapor cell. The excitation of the rubidium atoms is transversally limited to the diameter of the 795-nm beam; 475- and 480-nm beams are much larger; all optical beams are copropagating (32).

Fig. 2 Experiment scheme.

(A) Excitation level scheme. The excitation path is clockwise starting from 5P1/2 to 5P3/2 via 40S1/2. (B) Pulse sequence. The desorption pulse (gray) leads to a drastic increase of the optical depth within 5 ns (gray dots; error bars denote SD). The optical thickness is determined via statistical measurements at the single-photon level. Magenta and blue, excitation pulses; cyan, deexcitation pulse. Desorption pulse and deexcitation pulse are plotted with an offset relative to the power axis (left). (C) Typical signal shape retrieved from a cell length of 1.14 μm. The bin width is 0.425 ns. Nonclassical emission is observed in an early time interval of the signal, marked in gray (32).

The photons at 780 nm are generated in a pulsed four-wave mixing cycle about 5 ns after the pulsed desorption (Fig. 2, A and B). The durations of the excitation pulses are 2.5 ns each. In the limit of adiabatic elimination of the intermediate 5P1/2 state, the Rabi frequency of the coupling between the ground state and the Rydberg state is Ωeff/2π = 0.41 ± 0.16 GHz. The Rabi frequency of the transition from the Rydberg state to the excited state 5P3/2 is Ω480/2π = 1.15 ± 0.09 GHz. The photons at 780 nm are emitted into the same spatial mode as the 795-nm beam. They are collected by an aspheric high–numeric aperture lens behind the cell. For spatial and spectral filtering, the emission is coupled into a single-mode fiber and frequency-filtered by bandpass filters and a temperature-stabilized etalon with a bandwidth of 1.7 GHz. Afterward, the photons are detected in a Hanbury Brown and Twiss–type setup. The excitation pulses are recorded in order to enable a post-selection of events with respect to intensities and timing jitters (32).

Figure 2C shows a histogram of photoelectric detection events. The signal is mainly composed of two kinds of origins: (i) early photons that are coherently generated in the four-wave mixing process with a decay time of around 1 ns, and (ii) partly incoherently generated photons that take over at larger delays after the excitation pulses and are mainly originating from collisions of excited atoms with the glass windows. Both processes are much faster than the natural lifetime of the excited atoms. These two kinds of photons differ not only in their temporal appearance but also in their spatial emission pattern. The reason is that the four-wave mixing photons emit into the excitation mode, whereas the background photons are assumed to emit in the full solid angle.

In Fig. 3A, a normalized second-order intensity correlation function g(2)(τ) is shown, where the discrete delay is given in multiples of the repetition time (20 ms). The detection events are limited to a time window of 0 to 1.7 ns (gray area in Fig. 2C). All measurements from cell thicknesses 0.94 to 1.14 μm are concatenated. At zero delay, we observe clear antibunching with g(2)(τ = 0) = 0.215+10, where the error is the statistical uncertainty given by the error propagation of ±1 standard deviation of the correlation fluctuations at τ ≠ 0. The four-wave mixing should give g(2) = 1, so this constitutes clear evidence for strong Rydberg interactions that lead to excitation blockade and polariton dephasing. The detection probability in the measurement shown in Fig. 3A is on average 0.60%. Accounting for all attenuating elements between generation (behind the first lens) and detection of the photons, such as the coupling into a single-mode fiber, imperfect spectral filtering, and finite detection efficiencies of the single-photon counters, we obtain a mean generation efficiency (brightness) of ε = 4.0%.

Fig. 3 Normalized second-order photon correlation functions.

(A) Normalized second-order intensity correlation function g(2)(τ) as a function of coincidence delay time in multiples of the cycle repetition time (20 ms); time window, 0 to 1.7 ns; cell thickness L < L0; g(2)(τ = 0) = 0.215+10. (B) Dependence on the cell thickness in two time windows: 0 to 1.7 ns (blue), and 1.7 to 10 ns (red) as a reference. The blue dashed line is a guide to the eye based on a generic model g(2)(0) = (1 – g(2)bg) exp[–(L0/L)α] + g(2)bg (19), which is related to the hard-sphere character of the blockade. The characteristic length is L0 = 1.16 μm. (C) Dependence on the upper limit of the time window, for cell thicknesses L < L0 (blue) and L > L0 (red) as a reference. The blue dashed line is a guide to the eye and is described by a generic model similar to that of Fig. 3B. Error bars in the y direction are obtained by error propagation of ±1 SD due to statistical fluctuations; error bars in the x direction mark the span of x-positions of the cell setting the individual data points.

The effect of the excitation volume can be seen in Fig. 3B. Here, the zero-delay photon-pair correlation is depicted as a function of the cell length L. The blue data points show pair correlations in the time window 0 to 1.7 ns; the red data points show the time span of 1.7 to 10 ns as a reference. When the cell length is below a characteristic length L0 = 1.16 ± 0.01 μm, Rydberg-Rydberg interaction will prevent the emission of more than one photon into the detection mode. This value L0 is related to but not equivalent to the blockade radius. As a comparison to the strong interactions, we also investigated the 32S1/2 Rydberg state with rB = 0.85 ± 0.13 μm. We obtained Poissonian light statistics for all cell lengths around 1 μm and above.

The evolution of the imprint of the Rydberg interactions with time can be seen in Fig. 3C. It shows the photon-pair correlation versus the upper limit of the correlation time window that has a constant duration of 1.7 ns. At cell lengths smaller than L0 (blue), g(2)(τ = 0) reveals antibunching again. The coherent many-body state lasts about 1.9 ns, after which the imprint of the antibunching vanishes. As a comparison with negligible Rydberg interaction, the photon-pair correlation at larger cell lengths L > L0 is shown in red.

We have shown that strong Rydberg interactions in a hot atomic vapor cell lead to the generation of single photons. This has two consequences: (i) Strong atomic interaction effects do not need ultracold temperatures to be observable in the laboratory, and (ii) a single-photon emitter can be realized with promising features for future applications. Further experimental work will be needed to determine the indistinguishability of the generated photons, apply optical integration technologies, and combine this source with photon memories based on atomic vapors.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 and S2

References (3335)

References and Notes

  1. See supplementary materials.
Acknowledgments: We thank Y.-H. Chen for her contributions in the early stage of the experiment, F. Schreiber for crafting the vapor cell, N. Gruhler for the coating of the vapor cell with sapphire, C. Müller and M. Gryzlova for contributions to Fig. 1, P. Michler and M. Schwartz for useful discussions about single-photon sources, and H. Alaeian for reading and commenting the manuscript. Funding: Supported by the ERC under contract 267100 and the BMBF within (project 16KIS0129). Author contributions: The experiment was conceived by all authors; execution of the experiment as well as data analysis were performed by F.R.; F.R. wrote the manuscript with contributions from all authors. Competing interests: The authors declare no competing financial interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.

Stay Connected to Science

Navigate This Article