Streamlining a Cold-Atom Experimentby Marius GÄCHTER Cold-atom based experiments for quantum simulation are very sophisticated machines with many parameters, which require a lot of configuration and calibration before every measurement.Their complexity is furthermore steadily growing with the advancements of the field.This development even has the power to render certain experiments practically infeasible as the preparatory steps required keep prolonging and delaying the actual measurements.This thesis addresses this major problem by exploring different strategies to streamline and automatize the preparatory configuration and calibration steps in the Cavity Experiment at ETH Zurich.These strategies are mainly based on technological advancements in machine learning and computing power in general, which have been made in the time since the inception of the field of cold gases more than twenty years ago.In Part I, we streamline standard experimental techniques for optical lattice calibration, spin preparation and imaging.In Part II, we implement a versatile optimization framework and connect it to our experimental setup.This offers manifold possibilities, which we showcase by letting the optimization algorithm find optical evaporation ramps to produce a Bose-Einstein condensate, without giving it prior knowledge about the system.The techniques developed in the framework of this thesis can be readily extended and applied to other state of the art cold-atom experiments.
Streamlining a Cold-Atom Experimentby Marius GÄCHTER Cold-atom based experiments for quantum simulation are very sophisticated machines with many parameters, which require a lot of configuration and calibration before every measurement.Their complexity is furthermore steadily growing with the advancements of the field.This development even has the power to render certain experiments practically infeasible as the preparatory steps required keep prolonging and delaying the actual measurements.This thesis addresses this major problem by exploring different strategies to streamline and automatize the preparatory configuration and calibration steps in the Cavity Experiment at ETH Zurich.These strategies are mainly based on technological advancements in machine learning and computing power in general, which have been made in the time since the inception of the field of cold gases more than twenty years ago.In Part I, we streamline standard experimental techniques for optical lattice calibration, spin preparation and imaging.In Part II, we implement a versatile optimization framework and connect it to our experimental setup.This offers manifold possibilities, which we showcase by letting the optimization algorithm find optical evaporation ramps to produce a Bose-Einstein condensate, without giving it prior knowledge about the system.The techniques developed in the framework of this thesis can be readily extended and applied to other state of the art cold-atom experiments.
As ultracold atom experiments become highly controlled and scalable quantum simulators, they require sophisticated control over high-dimensional parameter spaces and generate increasingly complex measurement data that need to be analyzed and interpreted efficiently. Machine learning (ML) techniques have been established as versatile tools for addressing these challenges, offering strategies for data interpretation, experimental control, and theoretical modeling. In this review, we provide a perspective on how machine learning is being applied across various aspects of quantum simulation, with a focus on cold atomic systems. Emphasis is placed on practical use cases -- from classifying many-body phases to optimizing experimental protocols and representing quantum states -- highlighting the specific contexts in which different ML approaches prove effective. Rather than presenting algorithmic details, we focus on the physical insights enabled by ML and the kinds of problems in quantum simulation where these methods offer tangible benefits.
We apply an online optimization process based on machine learning to the production of Bose-Einstein condensates (BEC). BEC is typically created with an exponential evaporation ramp that is optimal for ergodic dynamics with two-body s-wave interactions and no other loss rates, but likely sub-optimal for real experiments. Through repeated machine-controlled scientific experimentation and observations our 'learner' discovers an optimal evaporation ramp for BEC production. In contrast to previous work, our learner uses a Gaussian process to develop a statistical model of the relationship between the parameters it controls and the quality of the BEC produced. We demonstrate that the Gaussian process machine learner is able to discover a ramp that produces high quality BECs in 10 times fewer iterations than a previously used online optimization technique. Furthermore, we show the internal model developed can be used to determine which parameters are essential in BEC creation and which are unimportant, providing insight into the optimization process of the system.
Abstract The generation of cold atom clouds is a complex process which involves the optimization of noisy data in high dimensional parameter spaces. Optimization can be challenging both in and especially outside of the lab due to lack of time, expertise, or access for lengthy manual optimization. In recent years, it was demonstrated that machine learning offers a solution since it can optimize high dimensional problems quickly, without knowledge of the experiment itself. In this paper we present results showing the benchmarking of nine different optimization techniques and implementations, alongside their ability to optimize a rubidium (Rb) cold atom experiment. The investigations are performed on a 3D 87Rb molasses with 10 and 18 adjustable parameters, respectively, where the atom number obtained by absorption imaging was chosen as the test problem. We further compare the best performing optimizers under different effective noise conditions by reducing the signal-to-noise ratio of the images via adapting the atomic vapor pressure in the 2D+ magneto-optical trap and the detection laser frequency stability.
This thesis presents work toward a quantum simulator with cold atoms. Characteristic and defining subsystems of the experimental setup are presented. Among them there are: the self-made viewports for ultra-high vacuum using indium seals, work on a modular laser system, for laser cooling, with miniaturised optics for the optimization of double path acusto-optic modulator. Work on a computer control to implement past slot is also presented. \nA versatile detection system for absorption imaging was designed and built. This detection system is stable, easy to align and exchange of magnification is simple. As well, a high resolution imaging system was designed, built and tested. This imaging system uses commercial microscope objectives and has been designed to work with four different microscope objective. Magnification of 5x, 10x 20x and 50x are possible. With the maximum magnification of 50x a resolution of 1.74μm for 780nm was achieved. Also a camera control for in-situ absorption imaging was programmed. Calculation of number of atoms and temperature were implemented in the program, as well as a defringing algorithm. \nA new method for temperature calculation for atoms held in a magnetic quadrupole trap was developed and tested, giving similar calculation of temperature as the commonly known time of flight (TOF) method. It is shown that this new method gives better results than the common TOF method for short time of flight. \nCharacterization of atom losses and heating in the production of a Bose-Einstein condensate were performed.
Virtually all aspects of many-body atomic physics are challenging: experiments are technically demanding, datasets have become enormous, and the memory and CPU requirements for classical simulation of generic quantum systems often scale exponentially with system size. Machine learning (ML) methods are already assisting in each of these areas and are poised to become transformative. Here, we focus on two specific applications of ML to cold-atom-based quantum simulators. These devices generally generate data in the form of images; we first showcase denoising of raw images and then identify solitonic waves in Bose-Einstein condensates. In both of these examples, we comment on the interplay between performance, model complexity, and interpretability.
The entry into the new millennium was marked in the quantum community by the first experimental realizations of the Bose-Einstein condensation of atoms in a vapor, achieved in different research groups. Twenty-five years later, such systems are 'routinely' produced and are the core of many experiments. At the frontier between condensed matter their y, quantum optics, and atomic physics, ultra-cold atoms are subjects of high scientific interest.Ultra-cold gas can be trapped in optical lattices, giving a realistic way of experimentally realizing a new range of lattice Hamiltonians, that are used to simulate condensed matter systems. In these simulations, ultra-cold atoms are used to mimic the behavior of electrons in a solid-state material and could be used to simulate a wide range of complex condensed matter phenomena, such as the fractional quantum Hall effect, topological insulators, and high-temperature superconductivity.In order to study such systems, diffraction-limited quantum gas microscopes have emerged as a way to manipulate and detect atoms trapped in lattice potentials. Using far-field interference, experiments have mainly been restricted to a lattice spacing of 500 nm. This limits the speed of dynamics of all physical processes.Higher energy regimes could be accessed by reducing the lattice spacing. For this purpose, one of the long-term goals of the team is to trap atoms in the near-field of a surface and to make use of the Casimir-Polder forces to shape sub-wavelength potentials. Thanks to nano-fabrication capabilities, the spacing can be reduced to a few tens of nanometers. However, this requires to adapt the imaging techniques to observe cold-atoms with sub-wavelength spatial resolutions.In this PhD thesis, I report first on the advances toward implementing an atom chip in our experiment, which will serve as an integrated platform to manipulate the quantum gas. I present simulations and study an experimental solution allowing the cloud to be transported to the near-field. The method is based on the interference of a laser radiation in the near field of a rotating surface. It has been fully simulated, and the limitations have been numerically and experimentally assessed. I conclude this part by presenting the first model of chip that has been fabricated.In a second part, I report on the development of a sub-wavelength imaging method of cold Rubidium 87 atoms which is based on excited state dressing. I detail the concept of this nano-metric spatial resolution method and apply it to super-resolve a single longitudinal wave function trapped in a tight lattice that exhibits a width below 30 nm.