Time series learning in a many-body Rydberg system with emergent collective nonlinearity
Zongkai Liu 1,2,⋆,§, Qiming Ren 1,2,§, Chris Nill 3,4,†,§, Albert Cabot 3, Wei Xia 5, Yanjie Tong 1,2, Huizhen Wang 1,2, Wenguang Yang 1,2, Junyao Xie 1,2, Mingyong Jing 1,2, Hao Zhang 1,2, Liantuan Xiao 1,2, Suotang Jia 1,2, Igor Lesanovsky 3,6, Linjie Zhang 1,2,‡ · 2025
Read Time series learning in a many-body Rydberg system with emergent collective nonlinearity by Zongkai Liu 1,2,⋆,§, Qiming Ren 1,2,§, Chris Nill 3,4,†,§, Albert Cabot 3, Wei Xia 5, Yanjie Tong 1,2, Huizhen Wang 1,2, Wenguang Yang 1,2, Junyao Xie 1,2, Mingyong Jing 1,2, Hao Zhang 1,2, Liantuan Xiao 1,2, Suotang Jia 1,2, Igor Lesanovsky 3,6, Linjie Zhang 1,2,‡ (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-15047.txt.
Introduction
Zongkai Liu, Qiming Ren, and thirteen co-authors show that a cloud of interacting Rydberg atoms can learn to predict time series, especially near a phase change. Rydberg atoms, boosted to giant excited states, respond to each other strongly and can amplify weak outside signals when the system sits close to a transition. The team feeds the series in by wiggling the laser that drives the vapor.
Close to the transition, the vapor's ability to learn the input grows, reflected in how faithfully its response can be read out as a forecast. The platform is attractive because its collective behavior can be tuned and read with simple tools. The result hints at atomic vapors as small, physics-built forecasting devices.
Catalog record
| CREATOR | Zongkai Liu 1,2,⋆,§, Qiming Ren 1,2,§, Chris Nill 3,4,†,§, Albert Cabot 3, Wei Xia 5, Yanjie Tong 1,2, Huizhen Wang 1,2, Wenguang Yang 1,2, Junyao Xie 1,2, Mingyong Jing 1,2, Hao Zhang 1,2, Liantuan Xiao 1,2, Suotang Jia 1,2, Igor Lesanovsky 3,6, Linjie Zhang 1,2,‡ |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-15047
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Time series learning in a many-body Rydberg system with emergent collective nonlinearity Zongkai Liu 1,2,⋆,§, Qiming Ren 1,2,§, Chris Nill 3,4,†,§, Albert Cabot 3, Wei Xia 5, Yanjie Tong 1,2, Huizhen Wang 1,2, Wenguang Yang 1,2, Junyao Xie 1,2, Mingyong Jing 1,2, Hao Zhang 1,2, Liantuan Xiao 1,2, Suotang Jia 1,2, Igor Lesanovsky 3,6, Linjie Zhang 1,2,‡ Abstract Interacting Rydberg atoms constitute a versatile platform for the realization of non-equilibrium states of matter. Close to phase transitions, they respond collectively to external perturbations, which can be harnessed for technological applications in the domain of quantum metrology and sensing. Owing to the controllable complexity and straightforward interpretability of Rydberg atoms, we can observe and tune the emergent collective nonlinearity. Here, we investigate the application of an interacting Rydberg vapour for the purpose of time series prediction. The vapour is driven by a laser field whose Rabi frequency is modulated in order to input the time series. We find that close to a non-equilibrium phase transition, where collective effects are amplified, the capability of the system to learn the input becomes enhanced. This is reflected in an increase of the accuracy with which future values of the time series can be predicted. Using the Lorenz time series and temperature data as examples, our work demonstrates how emergent phenomena enhance the capability of noisy many-body quantum systems for data processing and forecasting. 1 1 footnotetext: [email protected] 2 2 footnotetext: [email protected] 3 3 footnotetext: [email protected] 4 4 footnotetext: Z.L, Q.R, and C.N contribute equally to this work. I Introduction Figure 1: Time series prediction with a Rydberg vapour. The time-dependent input signal x(t) is sampled at discrete time steps t_{n} , yielding a step-wise constant function with values x_{n}=x(t_{n}) . The signal is converted into the Rabi frequency \Omega_{c}(t_{n}) of the coupling laser, which couples the intermediate state 6P_{3/2} to the Rydberg state 48D_{5/2} detuned by \Delta_{c} . This Rabi frequency ranges within the interval [\Omega_{\mathrm{min}},\Omega_{\mathrm{max}}] . The caesium (Cs) atoms are first excited by a probe laser with a constant Rabi frequency \Omega_{p} , which drives the transition from the ground state 6S_{1/2} to the intermediate 6P_{3/2} . Subsequently, the atoms are excited by the modulated coupling laser. The optical response of the Cs vapour is measured by monitoring the transmission T(t) of this probe laser. The measured transmission T(t) is post-processed using a Savitzky-Golay filter and down-sampled by a factor of 20 , yielding the transmission signals T^{(i)}=T(t_{i+20n}) . Here, i\in[1,20] labels the index of the downsampled time series and s_{n}=t_{i+20n} denotes the corresponding time. The transmission signal within the time interval [s_{n-M},s_{n}] is used to predict the signal x(s_{n+1}) (denoted as \hat{y}(s_{n+1}) ). This prediction is generated by a linear regression layer, which was previously trained (see text for details). Computation, information processing, and dynamics in many-body systems are inherently intertwined. One paradigmatic example is the Hopfield network Krotov 2023 , whose architecture, energy function, and memory retrieval are directly derived from the Ising model. Other examples include swarm intelligence algorithms, such as particle swarm optimization Kennedy and Eberhart 1995 ; Parsopoulos and Vrahatis 2002 ; Wang et al. 2018 and ant colony optimization Dorigo et al. 2006 ; Dorigo and Blum 2005 . These systems exploit collective interactions across scales to exhibit emergent computational capabilities, where nonlinear coupling between microscopic degrees of freedom generates information-processing capabilities. This interplay underscores both the potential of physical systems as computational platforms and the need to decode how collective behaviour enables specific computational tasks. A particularly relevant and challenging task is time-series prediction of complex systems, for example chaotic dynamics such as the Lorenz attractor Ghys 2013 ; Shen 2023 or weather forecasting Mammedov et al. 2022 . A computational framework that has proven effective for such problems is reservoir computing Jaeger 2001 ; Tanaka et al. 2019 ; Nakajima 2020 ; Wringe et al. 2025 . In reservoir computing, only a linear readout layer is trained, while the high-dimensional, nonlinear reservoir dynamics remain fixed. This significantly simplifies training, as it avoids tuning a large number of internal degrees of freedom and instead leverages the computational power of the intrinsic dynamics of the system, provided they are sufficiently rich Dale et al. 2019 ; Carroll 2022 ; Gauthier et al. 2021 . Reservoir computing has been implemented in a variety of settings, such as in echo state networks Ozturk et al. 2007 ; Sun et al. 2024 ; Jaeger and Haas 2004 and liquid state machines Maass 2011 as well as in physical systems such as delay-based electronic circuit reservoirs Soriano et al. 2015 , and quantum reservoir computing platforms Fujii and Nakajima 2017 ; Mujal et al. 2021 ; Sannia et al. 2024 . Despite this diversity, understanding the fundamental mechanisms that endow complex physical reservoirs with learning and prediction capabilities remains subject to current research Yan et al. 2024 . Thermal Rydberg vapours are a promising platform in this context. These many-body platforms are known to exhibit a rich phase diagram with complex internal dynamics, as extensively studied theoretically Ates et al. 2006 ; Marcuzzi et al. 2014 ; Lesanovsky and Garrahan 2014 ; Bai et al. 2020 ; Klocke et al. 2021 and demonstrated experimentally Carr et al. 2013 ; Ding et al. 2020 ; Wadenpfuhl and Adams 2023 ; Wu et al. 2024 ; Liu et al. 2024 . These features make Rydberg ensembles attractive candidates as platforms for reservoir computing, and, at the same time, ideal systems for probing the microscopic origin of learning capabilities in complex many-body dynamics at the atomic level. In this work, we exploit the response of a Rydberg vapour as a physical reservoir for time-series prediction. The experimental control over the collective behaviour of the system enables us to systematically analyse how its internal dynamics give rise to learning capabilities. We find that the learning performance is improved in the vicinity of a phase transition, where nonlinear and collective responses are strongly amplified in the Rydberg vapour. This behaviour is in good agreement with a corresponding theoretical minimal model, providing insight into how critical many-body dynamics augment the computational power of physical reservoirs. II Experiment Figure 2: Time series prediction. (a) Amplitude modulated signal \Omega_{c}/2\pi in MHz which is input into the Rydberg system. (b-d) Output transmission signal T(t) in arbitrary units recorded for different detunings \Delta_{c} . The detuning is chosen such that the system is out of (panel b) and within the bistable region (panels c-d). (e, f) Time series prediction for the Lorenz series x_{\mathrm{Lor}} . A Rydberg vapour operating in the bistable region (e) results in a better prediction (red curve), compared to outside (f). The ground truth is shown in blue. (g-h) Prediction results on daily temperature of Beijing for detuning \Delta_{c} , near (g) and out (h) of the bistable region. (i) The lower yellow and lower green curves correspond to the minimal Rabi frequency \Omega_{\mathrm{min}} of the coupling laser but with different scan directions of the detuning \Delta_{c} (as indicated by arrows and colours). The upper yellow and upper green curves correspond to \Omega_{\mathrm{max}} . A closed hysteresis loop formed by the two scan directions characterizes the bistable phenomenon. The detuning interval where the bistability occurs is