Cold atoms on the road to Honda

NaKa Origins

Within SynQS we are using cold atoms for the quantum simulation of complex problems. Over the years these studies spanned diverse topics like superfluidity [1], non-thermal fixed points [2], entanglement [3], cosmology or high-energy physics [4]. Nowadays, we are taking the next steps and started to look into the study of optimization within the project EnerQuant or machine learning problems with partners like the Honda Research Institute Europe. The list of topics has actually become so diverse that the reliability of the machines and their rate of producing data is a bottleneck in new studies and their deployment in industrial environments. A prime example is the NaLi ( = NAtrium + LIthium, which is sodium and lithium in German ) machine which is shown in the picture below.

Exciting studies on polaron physics or lattice gauge theories have been performed on the machine, but there is no question that it takes a grand master of experimental atomic physics to extract those results from the machine. The reasons are numerous, however they can be mostly traced back to the sheer complexity and size of the setup. This motivated us to look for leaner implementations that focus on a robust implementation of the problems under investigation. A first step was taken in the construction of the SoPa( = SOdium + POtassium ) machine, which already took numerous steps in this directions as discussed in Ref. [5]. Given this experience we started to wonder if it might be even possible to shrink down the complexity of the machines towards a system that would be compatible with data-center environments. This is where the work on the NaKa( = NAtrium + KAlium, once again, German for sodium and potassium) machine started. When the nearby Honda Research Institute Europe (HRI-EU) located in Offenbach, Germany, showed interest in realizing such a robust platform for processing quantum information, we were happy to join forces. Being already involved in the matter of using quantum technology for simulation and computing techniques, our collaborators became the “Godfathers” of the NaKa, accompanying investigation of theoretical optimization models potentially being tested on the future hardware as well as observing our experimental progress. With the opportunity to test our experiment-to-be at the HRI-EU facility and with their support we could make the rather bold attempt to transfer the whole quantum experiment to the HRI-EU data-center in Offenbach, including laser, vacuum, and electronics.

NaKa Beginnings

In the spring of 2021, we started the construction of the NaKa machine within the SynQS group of the Kirchhoff Institute of Physics at the University of Heidelberg. First of all, we had to specify the rough framework, the possibilities of the hardware- and software-realizations and to define the physical insights we wanted to achieve with running the experiment. Nowadays, online services like IBM Quantum make quantum technology accessible to people from all over the world ranging from high school students to highly trained experts by simply logging in to their website. It is therefore natural to aim for a completely remote experimental control. Once the experiment is set up, an authorized user should be able to access the experiment (e.g. tweak some parameters and of course receive the results) regardless of his or her physical location and time. The natural location for deploying such an experiment at a company facility would be their data-center. Like in a lab, only a limited number of people have access, electrical power supply is properly taken care of (including failure backups), and the temperature and humidity is at least somewhat controlled. So they seem like an almost perfect choice for running such sensitive experiments as we were planning to set up. This motivated the decision to assemble the whole device into standardized 19-inch racks which are widely used not only in data centers or networking centered facilities but also gained more popularity in different areas of application as well. After these considerations, we could finally start with getting our hands on all the optics, vacuum parts and electronics modules to start our journey!

A Rocky Road

Making a cold atom experiment fit into two $\sim 2$m high but rather narrow rack systems proved to be more challenging than we had anticipated. Stuffing all the electronics devices (e.g. power supplies, frequency controllers and small monitor devices) into our e(lectronics)-rack, was not too complicated. Many providers already adapted to this 19-inch standard being introduced long time ago and design their product to be compatible to this convention. By far more ambitious was the realization of its unlikely twin, our o(ptical)-rack. Starting from the size of the experiments already alive and running at our institute, this meant miniaturizing our vacuum system substantially, since we had a base platform of only $60 \text{cm} \times 40 \text{cm}$ at our disposal. Having to deal with tiny screws and their appropriate small-sized tools tested our patience more than once.

Furthermore, optical set-ups, as used for our experiments, are highly susceptible to temperature changes or vibrations because we have to rely on the sometimes hyperfine alignment of laser beams and interference patterns to being disturbed over a long period of time. The textbook tells you to buy big, suspended tables being several square meters in size, quickly filling up half a laboratory. However, we had to completely disregard this well-established and justified advice. Our o-rack was a home-built construction of simple alloy rods stacked together and completely static without any buffer zone between the hard ground and our fragile lenses, waveplates and beam splitters.

Finally, the vacuum chambers, where the atoms are being cooled to temperatures of a few mK, had to be built up. Next to standard components like Tees (a T-shaped tube), valves and viewports, a unique 2D-MOT (magneto-optical trap) chamber had to be designed to meet all the requirements and constraints. It has to maintain an ultra-high vacuum (UHV), withstand temperatures of over 200°C for “baking” the vacuum and should be not magnetizable since trapping the atoms involves the deployment of permanent magnets. The result was a 3 kg heavy, 10 cm wide vacuum chamber made entirely from stainless steel. As soon as it arrived in our lab and was connected to a strong vacuum pump, we were ready to equip it with optical mounts, windows for the laser beams and the magnets.

You never walk alone…

After having assembled all the necessary parts, we could finally aim for our first experimental break-through: the trapping of ultra-cold potassium-39 atoms in a two-dimensional cloud, by utilizing laser cooling and a magnetically controlled confinement inside the small vacuum chamber (i.e. the 2D MOT - two-dimensional magneto-optical trap). Being already familiar with this so called “signal hunting” and a bunch of luck, we could celebrate this first achievement after basically pushing the “on” button of our experiment. The next step to test was by far more crucial: would we be able to partly disassemble the NaKa in order to make it mobile and transport it to a suitable, remote location, power it up and expect the same results as we got so far? On November 3rd, the mission was just about to start. Firmly packaged and partly taken apart, the NaKa arrived safely in its full size at the HRI-EU facility in Offenbach and was quickly set up to be turned on in a server room. Without having to suffer any losses in hardware, we were able to re-construct our 2D MOT in no time and could leave it to stay at the Institute for a couple of weeks.

Far from Home

Having full access to the experiment from a remote location in mind, every device was integrated in a network based experimental control system. Because we did not want to rely on existing infrastructure at the remote location or adjust our setup to any company specific network setup, remote accesswas realized with the help of an LTE module. This way, the NaKa is reachable even in locations where no physical connection to the internet may be provided as long as a phone signal is available.

For this, a small side project called DjangoControlServer was created. Django is a widely used Python web framework, which focuses on user-friendliness where many web-developing structures are already implemented and the user is encouraged to write reusable apps instead of coding everything up from scratch again. This was very convenient for us as we were able to create an uniform platform for all our smaller lab devices measuring and putting out signals. Normally, every single device comes with its own software. Our work-around not only enables an overview over all processes running on the experiment but also saves a lot of time by not needing to click through a dozen different applications. In addition, it can be easily extended in case of new devices being deployed as long as one can communicate with them through Python (a widely used programming language).

Having set up everything, we could test the robustness of the NaKa to variations in its environment of a longer time. In the plots below we present the results of these tests for one day at the HRI-EU. The leftmost figure (a) shows the environment variables humidity and temperature inside the o-rack as measured with the help of a small sensor. While the overall mean value of both variables does not change significantly, one may observe oscillations in a 30 minute cycle. The server room was equipped with an air conditioning system, which caused these variations by its on-off-regulation. Additionally, people entering and leaving the adjacent rooms, one can observe some irregularities around 8:00 and 17:00. This in turn influences the laser such that the control input is also oscillating with the same frequency and shows some disturbances around the times mentioned above, which can be observed in figure (b). The strength/intensity of the 2D-MOT is similarly affected, see panel (c). Overall, we observed a high robustness of the machine to the variations as the operation did not require any realignment of the optics for weeks and the laser cooling worked continuously throughout the whole time.

Homecoming - What we have learned so far

So far we learnt several encouraging things:

• Building the cold atom hardware in a compact and mobile setup was more feasible than could have been initially expected.
• The systems can be operated within 24h after transport.
• Realignments have not been necessary over the time span of two weeks.
• A secure software access is a bottleneck in the deployment of the machines in data-centers as they usually require high-level certification.

Now, after the NaKa is back home at the Kirchhoff-Institute and thus having survived the trip twice, we are looking forward to further enhance our set-up while keeping it transportable and therefore highly flexible. With planning to add confinement in the third dimension (3D MOT), the respective vacuum parts are already in production. ****While preceding experiments like the NaLi will never be able to leave the lab, the NaKa might fulfill the desire of having a quantum simulator right inside their own doors and realize proposals like [6] in an industrial environment. Other opportunities lie in stress testing single components or assemblies as well as educational purposes. With this, we end our journey for now, however we are eager to get started on our next one and look towards the next river bend.

The Team

Synthetic Quantum Systems @ The Kirchhoff-Institute of Physics

Synthetic Quantum Systems (SYNQS) is a collaboration of four research groups with overlapping research interests. The collaboration exists to strengthen and better integrate theoretical and experimental research. Our work is generally related to leveraging the potential of synthetic quantum systems, with the aim of pushing the understanding of fundamental phenomena in quantum many-body physics, as well as developing novel quantum technologies.

The Honda Research Institute Europe

The Honda Research Institute Europe conducts research in artificial intelligence and intelligent systems. We believe that innovation is created through science and that intelligent systems will shape our future in many ways. Our focus is on fundamental research in domains such as computer science, bioinformatics, computational intelligence, optimization, and robotics. With about 50 scientists and researchers our goal is to develop novel technologies together with our university partners and provide innovative solutions to support Honda’s current and future technology roadmap.

[1] Eckel et al. Hysteresis in a quantized superfluid “atomtronic” circuit. Nature 506, 200 (2014).

[2] Prüfer et al. Observation of universal dynamics in a spinor Bose gas far from equilibrium. Nature 563, 217 (2018).

[3] Kunkel et al. Spatially distributed multipartite entanglement enables EPR steering of atomic clouds. Science 360, 413 (2018).

[4] Mil et al. A scalable realization of local U(1) gauge invariance in cold atomic mixtures. Science 367, 1128 (2020).

[5] Bhatt et al. Stochastic dynamics of a few sodium atoms in a cold potassium cloud. arXiv:2101.01135 (2021).

[6] Kasper et al. . Universal quantum computation and quantum error correction with ultracold atomic mixtures. Quantum Science and Technology 7, 015008 (2021).