arXiv subjects

Niv Drucker

Publications and source records attributed to Niv Drucker.

5 recordsLinked to original sources

arXiv2026

Deep Reinforcement Learning for Individual Atomic Control and Cooling

Real-time feedback control of quantum systems is often limited by partial observations, nonlinear dynamics and measurement noise, which make accurate model-based controllers difficult to design. Here we show that deep reinforcement learning can cool the motion of a single neutral atom coupled to a high-finesse optical cavity using only the continuously monitored cavity transmission. We first train the controller in simulation and then transfer it to the experiment, where online fine-tuning adapts it to unmodeled experimental dynamics. The learned policy damps the atom's motion in real time and achieves a cooling time constant of 388 +/- 14 microseconds, corresponding to only two motional periods in the trap. It also outperforms a standard linear differentiator controller in cooling speed while maintaining comparable atom retention over a broad range of operating conditions. These results establish reinforcement learning as a practical strategy for feedback control in quantum-limited experiments where compact analytical models are incomplete.

quant-ph↗

arXiv2024

Cavity-enabled real-time observation of individual atomic collisions

Using the strong dispersive coupling to a high-cooperativity cavity, we demonstrate fast and non-destructive number-resolved detection of atoms in optical tweezers. We observe individual atom-atom collisions, quantum state jumps, and atom loss events with a time resolution of $100\ \mu$s through continuous measurement of cavity transmission. Using adaptive feedback control in combination with the non-destructive measurements, we further prepare a single atom with $92(2)\%$ probability.

quant-ph↗

arXiv2024

Two-level system hyperpolarization using a quantum Szilard engine

The innate complexity of solid state physics exposes superconducting quantum circuits to interactions with uncontrolled degrees of freedom degrading their coherence. By using a simple stabilization sequence we show that a superconducting fluxonium qubit is coupled to a two-level system (TLS) environment of unknown origin, with a relatively long energy relaxation time exceeding $50\,\text{ms}$. Implementing a quantum Szilard engine with an active feedback control loop allows us to decide whether the qubit heats or cools its TLS environment. The TLSs can be cooled down resulting in a four times lower qubit population, or they can be heated to manifest themselves as a negative temperature environment corresponding to a qubit population of $\sim 80\,\%$. We show that the TLSs and the qubit are each other's dominant loss mechanism and that the qubit relaxation is independent of the TLS populations. Understanding and mitigating TLS environments is therefore not only crucial to improve qubit lifetimes but also to avoid non-Markovian qubit dynamics.

quant-ph↗

arXiv2023

Real-time frequency estimation of a qubit without single-shot-readout

Quantum sensors can potentially achieve the Heisenberg limit of sensitivity over a large dynamic range using quantum algorithms. The adaptive phase estimation algorithm (PEA) is one example that was proven to achieve such high sensitivities with single-shot readout (SSR) sensors. However, using the adaptive PEA on a non-SSR sensor is not trivial due to the low contrast nature of the measurement. The standard approach to account for the averaged nature of the measurement in this PEA algorithm is to use a method based on `majority voting'. Although it is easy to implement, this method is more prone to mistakes due to noise in the measurement. To reduce these mistakes, a binomial distribution technique from a batch selection was recently shown theoretically to be superior, as all ranges of outcomes from an averaged measurement are considered. Here we apply, for the first time, real-time non-adaptive PEA on a non-SSR sensor with the binomial distribution approach. We compare the mean square error of the binomial distribution method to the majority-voting approach using the nitrogen-vacancy center in diamond at ambient conditions as a non-SSR sensor. Our results suggest that the binomial distribution approach achieves better accuracy with the same sensing times. To further shorten the sensing time, we propose an adaptive algorithm that controls the readout phase and, therefore, the measurement basis set. We show by numerical simulation that adding the adaptive protocol can further improve the accuracy in a future real-time experiment.

quant-ph↗

arXiv2017

Demonstration of a passive photon-atom swap gate

Deterministic quantum interactions between single photons and single quantum emitters are a vital building block towards the distribution of quantum information between remote systems. Deterministic photon-atom state transfer has been demonstrated by using protocols that include active feedback or synchronized control pulses. Here we demonstrate a completely passive swap gate between the states of a single photon and a single atom. The underlying mechanism is single-photon Raman interaction (SPRINT) - an interference-based effect in which a photonic qubit deterministically controls the state of a material qubit encoded in the two ground states of a Λ system, and vice versa. Using a nanofiber-coupled microsphere resonator coupled to single Rb atoms we swap a photonic qubit into the atom and back, demonstrating nonclassical fidelities in both directions. Requiring no control fields or feedback protocol, the gate takes place automatically at the timescale of the atom's cavity- enhanced spontaneous emission time. Applicable to any waveguide-coupled Λ system, this scheme provides a versatile building block for the modular scaling up of quantum information processing systems.

quant-ph↗