Search

Experimental control involves a lot of manual effort with non-trivial decisions for precise adjustments. Here, we study the automatic experimental alignment for coupling laser light into an optical fiber using reinforcement learning (RL). We face several real-world challenges, such as time-consuming training, partial observability, and noisy actions due to imprecision in the mirror steering motors. We show that we can overcome these challenges: To save time, we use a virtual testbed to tune our environment for dealing with partial observability and use relatively sample-efficient model-free RL algorithms like Soft Actor-Critic (SAC) or Truncated Quantile Critics (TQC). Furthermore, by fully training on the experiment, the agent learns directly to handle the noise present. In our extensive experimentation, we show that we are able to achieve 90% coupling, showcasing the effectiveness of our proposed approaches. We reach this efficiency, which is comparable to that of a human expert, without additional feedback loops despite the motors' inaccuracies. Our result is an example of the readiness of RL for real-world tasks. We consider RL a promising tool for reducing the workload in labs., Comment: 10 pages + 10 pages appendices, 3 + 11 figures

Optical quantum memories are essential for quantum communications and photonic quantum technologies. Ensemble optical memories based on 3-level interactions are a popular basis for implementing these memories. All such memories, however, suffer from loss due to scattering. In off-resonant 3-level interactions, such as the Raman gradient echo memory (GEM), scattering loss can be reduced by a large detuning from the intermediate state. In this work, we show how electromagnetically induced transparency adjacent to the Raman absorption line plays a crucial role in reducing scattering loss, so that maximum efficiency is in fact achieved at a moderate detuning. Furthermore, the effectiveness of the transparency, and therefore the efficiency of GEM, depends on the order in which gradients are applied to store and recall the light. We provide a theoretical analysis and show experimentally how the efficiency depends on gradient order and detuning.

Squeezed states are an interesting class of quantum states that have numerous applications. This work presents the design, characterisation, and operation of a bow-tie optical parametric amplifier (OPA) for squeezed vacuum generation. We report the high duty cycle operation and long-term stability of the system that makes it suitable for post-selection based continuous-variable quantum information protocols, cluster-state quantum computing, quantum metrology, and potentially gravitational wave detectors. Over a 50 hour continuous operation, the measured squeezing levels were greater than 10 dB with a duty cycle of 96.6%. Alternatively, in a different mode of operation, the squeezer can also operate 10 dB below the quantum noise limit over a 12 hour period with no relocks, with an average squeezing of 11.9 dB. We also measured a maximum squeezing level of 12.6 dB at 1550 nm. This represents one of the best reported squeezing results at 1550 nm to date for a bow-tie cavity. We discuss the design aspects of the experiment that contribute to the overall stability, reliability, and longevity of the OPA, along with the automated locking schemes and different modes of operation., Comment: 7 pages, 5 figures

Strong nonlinear interactions between single photons have important applications in optical quantum information processing. Demonstrations of these interactions in cold atomic ensembles have largely been limited to exploiting slow light generated using electromagnetically induced transparency (EIT). However, these EIT implementations have limited achievable phase shifts due to spontaneous emission. Here, we demonstrate and characterize a scheme free from these limitations using gradient echo memory with inferred single photon phase shifts of $0.07\pm0.02$ $\mu\text{rad}$. Excellent agreement with theoretical modelling was observed. Degradation of memory efficiency was observed for large phase shifts but strategies to overcome that are presented., Comment: 7 pages, 6 figures

Efficient quantum memories will be an essential building block of large scale networked quantum systems and provide a link between flying photonic qubits and atomic or quasi-atomic local quantum processors. To provide a path to scalability avoidance of bulky, difficult to maintain systems such as high vacuum and low temperature cryogenics is imperative. Memory efficiencies above 50% are required to be operating above the quantum no-cloning limit. Such high efficiencies have only been achieved in systems with photon sources tailored to the memory bandwidth. In this paper we explore the combination of an ultralow spectral bandwidth source of single photons from cavity-enhanced spontaneous parametric down-conversion with a gas-ensemble atomic memory. Our rubidium vapour gradient echo memory achieves 84$\pm$3% recall efficiency of single photons: a record for an always-ready, hot, and vacuum system free optical memory., Comment: 11 pages, 10 figures

In the recent years, we observe a dramatic boost of research in photonics empowered by the concepts of machine learning and artificial intelligence. The corresponding photonic systems, to which this new methodology is applied, can range from traditional optical waveguides to nanoantennas and metasurfaces, and these novel approaches underpin the fundamental principles of light-matter interaction developed for a smart design of intelligent photonic devices. Concepts and approaches of artificial intelligence and machine learning penetrate rapidly into the fundamental physics of light, and they provide effective tools for the study of the field of metaphotonics driven by optically-induced electric and magnetic resonances. Here, we introduce this new field with its application to metaphotonics and also present a summary of the basic concepts of machine learning with some specific examples developed and demonstrated for metasystems and metasurfaces.

In two-color optical nanofiber-based dipole traps for cold alkali atoms, the trap efficiency depends on the wavelength and intensity of light in the evanescent field, and the initial laser-cooling process. Typically, no more than one atom can be trapped per trapping site. Improving the trapping efficiency can increase the number of filled trapping sites, thereby increasing the optical depth. Here, we report on the implementation of an in-loop stochastic artificial neural network machine learner to trap $^{87}$Rb atoms in an uncompensated two-color evanescent field dipole trap by optimizing the absorption of a near-resonant, nanofiber-guided, probe beam. By giving the neural network control of the laser cooling process, we observe an increase in the number of dipole-trapped atoms by $\sim$ 50%, a small decrease in their average temperature from 150 $\mu$K to 140 $\mu$K, and an increase in peak optical depth by 70%. The machine learner is able to quickly and effectively explore the large parameter space of the laser cooling control to find optimal parameters for loading the dipole traps. The increased number of atoms should facilitate studies of collective atom-light interactions mediated via the evanescent field., Comment: 12 pages, 5 figures

The efficiency of an ensemble-based optical quantum memory depends critically on the strength of the atom-light coupling. An optical cavity is an effective method to enhance atom-light coupling strength, with the drawback that cavities can be difficult to integrate into a memory setup. In this work we show coherent enhancement of atom-light coupling via an interference effect. The light to be absorbed into the atomic ensemble is split and used to drive the atoms from opposite ends of the ensemble. We compare this method theoretically to a cavity enhanced scheme and present experimental results for our coherent enhancement in cold rubidium-87 atoms that show an efficiency of $72\pm5\%$ and a storage lifetime of $110\pm 10$ us.

A number of techniques exist to use an ensemble of atoms as a quantum memory for light. Many of these propose to use backward retrieval as a way to improve the storage and recall efficiency. We report on a demonstration of an off-resonant Raman memory that uses backward retrieval to achieve an efficiency of $65\pm6\%$ at a storage time of one pulse duration. The memory has a characteristic decay time of 60 $\mu$s, corresponding to a delay-bandwidth product of $160$.

Many important physical processes have dynamics that are too complex to completely model analytically. Optimisation of such processes often relies on intuition, trial-and-error, or the construction of empirical models. Machine learning based on artificial neural networks has emerged as an efficient means to develop empirical models of complex systems. We implement a deep artificial neural network to optimise the magneto-optic cooling and trapping of neutral atomic ensembles. Cold atomic ensembles have become commonplace in laboratories around the world, however, many-body interactions give rise to complex dynamics that preclude precise analytic optimisation of the cooling and trapping process. The solution identified by machine learning is radically different to the smoothly varying adiabatic solutions currently used. Despite this, the solutions vastly outperform best known solutions producing higher optical densities. This may provide a pathway to a new understanding of the dynamics of the cooling and trapping processes in cold atomic ensembles.