2026 Volume 35 Issue 8
Article Contents

Ying-Wen Zhang, Yang Wang, Chen-Long Xu, Yi-Bo Wang, Peng Xu. Coherent Rydberg excitation of single atoms using a pulsed fiber amplifier[J]. Chinese Physics B, 2026, 35(8): 083204. doi: 10.1088/1674-1056/ae5db3
Citation: Ying-Wen Zhang, Yang Wang, Chen-Long Xu, Yi-Bo Wang, Peng Xu. Coherent Rydberg excitation of single atoms using a pulsed fiber amplifier[J]. Chinese Physics B, 2026, 35(8): 083204. doi: 10.1088/1674-1056/ae5db3

Coherent Rydberg excitation of single atoms using a pulsed fiber amplifier

  • Received Date: 09/02/2026
    Accepted Date: 30/03/2026
    Available Online: 01/08/2026
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(6)

Article Metrics

Article views(95) PDF downloads(3) Cited by(0)

Access History

Coherent Rydberg excitation of single atoms using a pulsed fiber amplifier

Abstract: In recent years, the growing scale of programmable neutral-atom arrays has led to an increasing demand for higher-power Rydberg excitation light. Although pulsed amplifiers deliver higher peak power than continuous-wave lasers, their application to efficient coherent Rydberg excitation of single atoms in arrays has been limited by challenges such as pulse distortion, synchronization with excitation sequences, and spectral linewidth broadening. Here, we address these issues using a fiber-based master-oscillator power-amplifier system. We demonstrate efficient coherent Rydberg excitation of single atoms in a rubidium atom array, achieving performance comparable to that of continuous-wave methods. This study provides a new technical pathway toward future large-scale quantum simulation and computation with Rydberg atom arrays.

1.   Introduction
  • Neutral-atom arrays have rapidly developed into a leading platform for realizing programmable quantum systems. In this approach, individual cold atoms are trapped in arrays of optical tweezers, and controllable interactions are introduced by exciting the atoms to Rydberg states. The system exhibits excellent qubit coherence,[1,2] scalability,[3,4] flexible qubit connectivity,[5] high-fidelity Rydberg excitation control,[6] and two-qubit logic gates.[7–9] It has been successfully used to prepare complex quantum many-body states[10–12] and provides an experimental testbed for constructing scalable architectures toward universal fault-tolerant quantum computing.[13,14]

    To advance quantum simulation and computation, experimental systems are being progressively scaled up,[15,16] placing higher demands on experimental control — particularly the need for higher-power Rydberg excitation light to address larger atom arrays and achieve faster Rabi frequencies.[8] For example, in rubidium atoms, the 420 nm + 1013 nm two-photon excitation scheme is widely adopted, largely because readily available near-infrared continuous-wave (CW) lasers — such as Ti:sapphire and fiber lasers delivering tens to hundreds of watts — can efficiently drive the upper transition from the intermediate state to the Rydberg state, despite its small transition dipole moment. In neutral-atom-array experiments, the excitation sequence typically operates in a pulsed mode: the Rydberg excitation light is switched on only during short windows (typically on the order of microseconds) via components such as acousto-optic modulators (AOMs) and remains off otherwise. In such a pulsed configuration, the attainable peak power remains limited by the available CW laser power.

    In contrast, pulsed lasers concentrate energy within extremely short time windows, delivering peak powers significantly higher (> 1 kW) than those of CW sources.[17–19] This characteristic makes them a compelling alternative to CW lasers for supplying higher-power Rydberg excitation. Previous studies have utilized picosecond pulses for ultrafast Rydberg excitation[20] and nanosecond pulses for electromagnetically induced transparency (EIT) spectroscopy.[21] However, efficient coherent excitation of single atoms with microsecond pulsed lasers remains elusive due to challenges such as severe pulse distortion,[22,23] synchronization with excitation sequences, and spectral linewidth broadening.

    In this paper, we implement a fiber-based master-oscillator power-amplifier (MOPA) scheme to achieve efficient coherent Rydberg excitation of single atoms in a rubidium atom array. We address waveform distortion and pulse-sequence synchronization. The demonstrated performance is comparable to that of continuous-wave methods, indicating that the MOPA architecture largely preserves the narrow linewidth of the seed laser during power amplification. This work thus provides a potentially new technical route toward scalable, high-fidelity Rydberg-based quantum control. The rest of this paper is organized as follows. Section 2 describes the pre-shaping of the 1013 nm seed-pulse waveform to suppress amplified pulse distortion. In Section 3, we synchronize the selected amplified 1013 nm pulse with a 420 nm pulse generated from a CW laser to drive coherent Rydberg excitation of single atoms in a rubidium atom array and compare the results with those obtained using a conventional CW scheme. Finally, Section 4 summarizes the main findings and discusses potential applications of the pulsed fiber amplifier scheme.

2.   Pulse pre-shaping compensates for waveform distortion during amplification
  • In the experiment, a pulsed fiber amplifier module (Connet MFAP-Yb-1013-M-SF) with a peak power of 50 W is employed to amplify the input 1013 nm seed laser. This amplifier utilizes an ytterbium-doped fiber as the gain medium, with pump light establishing the initial population inversion between the upper and lower energy levels of the Yb3+ ions. The seed laser, first modulated by an AOM to generate optical pulses with tunable pulse width and repetition rate, is then introduced into the gain medium. Within the fiber, photons from the seed pulse stimulate the excited Yb3+ ions to return to the ground state, emitting additional photons through stimulated emission and thereby amplifying the pulse.

    During the amplification process, the leading edge of the pulse first enters the gain medium, where the population inversion has not yet been depleted, resulting in high gain and strong amplification. As the pulse propagates through the medium, the population inversion is progressively depleted. By the time the trailing edge arrives, the gain decreases, leading to weaker amplification.[22,23] This variation causes distortion in the output pulse waveform. Figure 1 illustrates this phenomenon using a square input pulse as an example. The input waveform is shown in Fig. 1(a), while the black curve in Fig. 1(b) represents the experimental output waveform. It can be seen that the output waveform exhibits significant distortion due to gain saturation.

    The dynamic evolution of the photon density of the pulse and the population inversion during amplification can be quantitatively described by the Frantz–Nodvik equation[23]

    where n denotes the photon number density, σ is the stimulated emission cross-section, c represents the speed of light in the medium, and Δ0 denotes the initial population inversion. Furthermore, we can derive the expression for the output pulse Is,z (T) after an input pulse Is,0 (T) passes through a fiber of length L, given by

    where I is the optical intensity, T = t – z / c, G0(z)=exp(σ∫0zΔ0(z′)dz′) is the small-signal gain, Jout(T)=∫−∞TIs,0(t′)dt′ represents the output pulse fluence, and Jsat is the saturation fluence, given by hν / (2σ). At the onset of amplification, Jout = 0, the denominator equals 1 / G0, and the gain is G0. As amplification proceeds, Jout increases, causing the denominator to increase and the gain to gradually decrease. When Jout ≫ Jsat, the denominator approaches unity, at which point the gain drops to 1, indicating that the gain is fully saturated and no further power amplification occurs.

    Based on this model, the correspondence between the input and output waveforms can be readily established. Taking the measured waveform in Fig. 1(a) as the input, we adjust the two undetermined parameters in Eq. (2), namely the initial population inversion and the stimulated emission cross-section, so that the theoretically calculated output waveform progressively approaches the measured output waveform. This process determines the values of these parameters. As indicated by the red curve in Fig. 1(b), the theoretical waveform shows good overall agreement with the measured one. The main discrepancy is that the theoretical waveform is slightly broader than the measured one. This difference arises because the actual amplification process is accompanied by temporal pulse compression,[24] an effect not accounted for in the theoretical calculation.

    To correct the waveform distortion, we employ an input waveform pre-shaping scheme. Based on the established relationship between the input and output waveforms, the Frantz–Nodvik equation is inversely solved to obtain[25]

    Substituting the target output waveform into Eq. (3) yields the theoretical pre-shaped waveform. The basic workflow of waveform pre-shaping and the corresponding experimental setup are illustrated in Figs. 2(b) and 2(c). In practice, the theoretical pre-shaped waveform is loaded into a function generator to pre-modulate the seed light entering the pulsed fiber amplifier. The amplified optical pulse is detected with a high-speed photodetector; the resulting waveform is converted into an electrical signal and recorded by an oscilloscope. The amplified output waveform obtained directly from the theoretical pre-shaping still deviates from the target waveform because the experimentally applied pre-shaped waveform is the radio-frequency (RF) waveform that drives the internal AOM, rather than the optical pulse itself. Since the actual modulated optical pulse cannot be measured directly, the pre-shaped input waveform is fine-tuned iteratively based on the measured output.[26]

    The effectiveness of the pre-shaping scheme in suppressing output waveform distortion is illustrated here for a target 1 μs square pulse, as shown in Fig. 3. In Fig. 3(a), the red curve shows the theoretically pre-shaped waveform, while the black curve corresponds to the experimentally implemented waveform after fine-tuning. The measured output (black curve) agrees well with the theoretical output (red curve), with a root-mean-square (RMS) deviation of 1% over the interval between 0.6 μs and 1.3 μs, as shown in Fig. 3(b). Compared to the unshaped input, the pre-shaped waveform effectively suppresses the overall pulse distortion.

    Although the output waveform distortion has been significantly reduced, it still deviates to some extent from an ideal square pulse. We attribute this discrepancy mainly to two factors. First, the waveform pre-shaping scheme we employed adjusts only a few key parameters to control the overall shape of the pre-modulated waveform, offering limited precision for capturing finer details.[26,27] Second, the pre-shaping approach is based on the Frantz–Nodvik equation, which inherently incorporates certain idealizations, such as neglecting nonlinear effects like stimulated Raman scattering (SRS)[28] and stimulated Brillouin scattering (SBS),[29] which are unavoidable in practical systems.

    In addition, we tested square pulses under different pulse-width and pump- power conditions. The results show that when varying the pulse width at a fixed pump power, the output waveform distortion characteristics remain highly consistent because they follow the same gain profile. Under a fixed pulse width, increasing the pump power raises the initial gain, extends the time required to enter the gain-saturation regime, and consequently intensifies the waveform distortion. In both scenarios, applying the waveform pre-shaping method effectively suppresses the distortion and yields square-pulse outputs that closely approximate the ideal shape.

3.   Coherent Rydberg excitation of single atoms
  • We then apply the pre-shaped pulse to drive Rydberg excitations in an atom-array platform. First, a cloud of cold 87Rb atoms is prepared in a magneto-optical trap (MOT). Subsequently, an 830 nm beam is phase-modulated by a spatial light modulator (SLM) and tightly focused through a high-numerical-aperture lens to form an optical tweezer array. The tweezer array overlaps spatially with the MOT and is capable of trapping up to 10 × 20 atoms. Individually loaded atoms are further cooled via polarization gradient cooling and adiabatic cooling, reaching a final temperature of approximately 7 μK. Atoms are then optically pumped into the | F = 1, mF = 0〉 state and excited to a Rydberg state via a two-photon transition. Finally, fluorescence detection is used to measure the probability of atoms remaining in the ground state.

    The relevant energy-level structure of the 87Rb atom and the two-photon Rydberg excitation scheme are shown in Fig. 4(a). Atoms are excited from the state | 0〉 = |F = 1, mF = 0〉 to the 60S1/2 Rydberg state via a two-photon process. Both the 420 nm and 1013 nm excitation beams are π-polarized, with an intermediate-state detuning of 1.69 GHz. The π-polarization of the Rydberg excitation light is adopted to facilitate the subsequent implementation of a magic-intensity optical trap scheme, ensuring that the required magnetic field direction is perpendicular to the excitation beam.[1,2] The 1013 nm light originates from a homemade cat-eye laser; most of the output serves as seed light for the pulsed fiber amplifier, while a small portion is coupled into an ultra-stable cavity for frequency stabilization, resulting in a linewidth of a few hundred hertz (Fig. 2(a)). The time sequence for Rydberg excitation and the method used to select a single 1013 nm pulse are shown in Fig. 4(b). To maintain stable population inversion in the gain medium and prevent runaway accumulation of nonlinear effects, the pulsed fiber amplifier is operated in a burst mode.[17] For Rydberg excitation, a single pulse is extracted from the burst using a TTL-controlled switch. When the TTL signal is high, the amplifier outputs a single pulse (pulse 2) for excitation; when low, it continues to deliver the original pulse burst (pulse 1). The 420 nm laser is generated by frequency-doubling an 840 nm source, and its frequency is stabilized to the same ultra-stable cavity as the 1013 nm laser. The Rydberg excitation geometry is shown in Fig. 4(c). Two counter-propagating beams are aligned to a selected row of four atoms spaced about 20 μm apart. At this spacing, interatomic interactions are negligible, allowing each atom to be excited independently. This configuration makes use of the uniformity of the excitation light while improving data acquisition efficiency.

    To enable a direct comparison between the CW and pulsed schemes, the output of the pulsed amplifier is coupled into the pre-aligned 1013 nm delivery fiber via a free-space optical path. This configuration allows rapid switching between the pulsed and CW excitation schemes as needed. Accounting for coupling losses in the free-space interface, approximately 50% of the pulsed amplifier’s power is available for excitation.

    The 1013 nm Rydberg excitation pulse is generated using the pre-shaping scheme under conditions of 10 μs pulse duration, 3.5 W pump power, and a 2 kHz repetition rate, corresponding to a peak power of approximately 5 W from the pulsed fiber amplifier. Rabi oscillation is measured by scanning the duration of the 420 nm pulse. To minimize decoherence due to power fluctuations, the 420 nm pulse is positioned at the center of the flat region of the 1013 nm pulse, as shown in Fig. 4(b).[30] The power fluctuation between amplified 1013 nm Rydberg excitation pulses is approximately 2%, while that for the 420 nm excitation pulses is about 1%.

    We apply an external magnetic field to lift the degeneracy of the 60S state, splitting the mj sublevels by approximately 40 MHz. By scanning the frequency of the 1013 nm excitation laser, we locate the absorption peak corresponding to the transition from the ground state to the Rydberg state |r〉 = |60S, mj = −1/2〉. With the 1013 nm laser tuned to resonance, we then scan the duration of the 420 nm pulse to drive coherent evolution between the ground and Rydberg states.

    The Rabi oscillation between |0〉 and |r〉 obtained using the conventional CW scheme is shown in Fig. 5(a). A fit to the experimental data (orange solid points) yields a two-photon Rabi frequency of 2π × (1.98 ± 0.01) MHz, a 1/e lifetime of 11.14 ± 3.62 μs, and a contrast of ∼ 0.94. The corresponding oscillation measured with the pulsed fiber amplifier scheme is shown in Fig. 5(b). Fitting the data (red solid points) yields a Rabi frequency of 2π × (1.35 ± 0.01) MHz, a 1/e lifetime of 15.10 ± 7.54 μs, and a contrast of ∼ 0.98. Both schemes complete approximately 20 oscillation cycles within the 1/e lifetime, confirming their comparable coherence performance.

    The pulsed fiber amplifier scheme enables the Rabi frequency to be increased directly by raising the pump power and shortening the pulse duration. To explore the performance limits of the amplifier module, we compare the achieved Rabi frequencies under different pump powers for two extreme pulse widths: 1 μs and 10 μs, and the results are shown in Fig. 5(c). Under the condition of a 1 μs pulse width, a 2 kHz repetition rate, and a pump power of 4 W, the pulsed fiber amplifier reaches its maximum output (∼ 50 W peak power), yielding a two-photon Rabi frequency of 2π × (4.25 ± 0.06) MHz.

    To evaluate potential linewidth broadening introduced by the pulsed amplifier, we perform Ramsey interferometry between |0〉 and |r〉 using a π/2-gap-π/2 pulse sequence. The results for the conventional CW scheme and the pulsed fiber amplifier scheme are compared in Fig. 6. For the pulsed fiber amplifier scheme, the fitted dephasing time extracted from the Ramsey fringes is T2∗=5.25±0.78 μs (Fig. 6(b)). At a temperature of T = 7 μK, the random shot-to-shot variation in atomic velocity results in a Gaussian-distributed Doppler detuning δD / (2π), corresponding to an expected dephasing time of 6.24 μs. This indicates that the dephasing is dominated by thermal Doppler effects. The measured dephasing time for the pulsed fiber amplifier scheme is comparable to the value of 6.42 ± 0.73 μs obtained with the CW scheme. Currently, the ∼ 2% shot-to-shot power fluctuations of the 1013 nm pulsed output are about twice those of the CW source; improving this stability could further extend the dephasing time.[31] These results indicate that the MOPA architecture largely preserves the narrow linewidth of the seed laser during power amplification, thus enabling efficient coherent control of Rydberg atoms.

4.   Conclusion and outlook
  • This study demonstrates efficient coherent Rydberg excitation of single atoms using a fiber-based MOPA system, achieving performance comparable to CW excitation. The MOPA architecture provides a promising route to deliver higher peak power while preserving the spectral quality of the seed laser. The Rydberg Rabi frequency can be flexibly tuned through pump power and pulse width adjustments, potentially exceeding the coupling strength achievable with conventional CW lasers. This approach provides a scalable route to large-scale quantum simulation and computation based on Rydberg atom arrays.

    In the future, closed-loop feedback[32,33] could further improve the amplified output waveform quality and enable arbitrary waveform generation. Such capabilities would be suitable for single off-resonant modulated pulse schemes[34–36] aimed at high-fidelity two-qubit gates. For these implementations, controlling the phase noise introduced during amplification — which has not been systematically examined here — remains an important task[37,38] and offers a compelling direction for further study. Furthermore, the high-repetition-rate, short-pulse burst generated by the amplifier module is naturally suited for periodically driving quantum systems, enabling the study of novel nonequilibrium phenomena such as discrete time crystals.[39,40]

Figure (6)  Reference (40)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return