In the beginning of this week, we had a fantastic workshop in Würzburg to kick off our new DFG Research Unit FOR5919 “Machine Learning for Complex Quantum States”. Together with the participants and three invited speakers we discussed various topics ranging from machine learning-enhanced quantum gas experiments to agents learning to play quantum games. We’re looking forward to exciting research projects and lively meetings over the course of the next years.
Learn about our group’s work at the DPG Spring Meeting 2026:
We welcome Sven, who joins our team as a new Postdoc. Sven will investigate the inner workings of neural quantum states and what this means for time evolution.
The DFG has decided to fund our Research Unit FOR5919 “Machine Learning for Complex Quantum States”. This collaborative research initiative aims at developing machine-learning-enhanced techniques for simulation, data analysis, and control to enable new insights into complex quantum states and dynamics. It involves participants from eleven institutions across Germany and Switzerland. We’re very excited to kick it off!
Our group continues growing: Massimo joined us as a new PhD student. He will work on investigating the dynamics of correlated two-dimensional materials with neural quantum states.
Our article on explicitly time-dependent NQS has been published in Machine Learning: Science and Technology. Instead of conventional forward-integration of Schrödinger’s equation, this technique optimizes an artificial neural network to solve the dynamics simultaneoulsy across a whole time interval. We demonstrated the efficiency of the approach by simulating quench dynamics and a time-dependent control protocol for a large two-dimensional quantum magnet.
Thanks to great efforts by Wladi, our work on roughening dynamics now appeared in Physical Review Letters! In this paper, we describe how the smooth-interface regime below the roughening transition of the two-dimensional quantum Ising model leads to prethermal plateaux in the non-equilibrium dynamics of domain wall initial conditions.
Two new Master students joined our group: Stefan will work on aspects of many-body quantum chaos and Valentin will continue to develop our explicitly time-dependent NQS approach.
We are looking for a motivated postdoc to join our team and work on new approaches to tackle the quantum many-body problem with machine learning techniques. See the opening for details.
Update: Review of applications has started.
With the beginning of October, Noé joined our group as a new PhD student. He is going to work on our DFG project ``Machine Learning to Tailor Correlated States of Matter’'.
In November we’ll be hosting a workshop on simulating the non-equilibrium dynamics with neural quantum states here in Regensburg. See the workshop website for more information!
Our Postdoc Jonas has secured one of the five places in FZ Jülich’s Young Excellent Scientist Program (YESP). The program offers individual support modules, among which the funding to hold their own symposium on a topic of their choice. Congratulations, Jonas!
We are looking for a new PhD student for a DFG-funded project on “machine learning to tailor correlated states of matter”. The project in cooperation with colleagues at MPI PKS will evolve around neural quantum states for non-equilibrium quantum many-body physics.
Update: The position has been filled.
The advent of quantum simulators and computers has made projective measurement a reality at the many-particle level: Experiments routinely collect high-quality pictures of wave functions made of hundreds of components. But the wealth of information obtained challenges traditional theoretical modeling, which often focuses on a few observables. In our work, that is now published in Phys. Rev. X, we introduce a framework to describe wave-function measurements using network theory, enabling the discovery of a very deep inner structure in quantum wave functions: scale-freeness similar to other completely disconnected ensembles, such as those found in communication, social networks and the internet.
We are lookging for a dedicated student assistant to support our research team. The main tasks will evolve around programming and conducting explorative studies in the area of machine learning for quantum science.
Update: The position has been filled.
Periodic driving is a prime way to engineer quantum matter with interesting properties. For this purpose, it is essential to bring the system into metastable steady states, the properties of which governed by a corresponding effective Hamiltonian. We have developed a deep-learning-assisted variational algorithm to reconstruct such effective Hamiltonians from observations. Thereby, we could recover local Hamiltonians in pre-thermal regimes and observe the growing complexity of instantaneous effective Hamiltonians in subsequent heating regimes. Our findings are now published in Phys. Rev. Research.
Jarzynski’s equality describes a relation in statistical physics between the fluctuations during a work process and the thermal free energy difference corresponding to the initial and final ensemble. While Jarzynski’s equality has been checked in many scenarios, previous experiments focused on systems with only a handful of degrees of freedom in the quantum regime. In our collaboration with colleagues at MPI PKS, the University of Bonn, and UC Berkeley/LBNL, we probed for the first time Jarzynski’s equality in the quantum many-body regime. We devise a dynamical protocol to measure work on quantum computers using a suitably prepared thermal ensemble and tested Jarzynski’s equality on different quantum processors with up to 16 qubits. The article has now been published in Phys. Rev. X.
Together with colleagues from Hamburg, Waterloo, and Augsburg, we are organizing a school and conference “Frontiers at the intersection of quantum simulation and machine learning” to be held from Apr 8 to 19 2024 at the International Centre for Theoretical Physics (ICTP) in Trieste. This will be the first event at ICTP co-sponsored by the WE Heraeus Foundation. More detailed information will follow soon on the event website.
Our group keeps growing and today is Jonas Rigo’s first day as a Postdoc. Welcome, Jonas!
We developed a new approach to simulate spectral functions using neural quantum states, which for example accurately captures the gap closing at the phase transition of a two-dimensional quantum magnet despite the associated diverging time scale. The paper appeared today in Phys. Rev. Lett.!
Last week, Mo Abedi successfully defended his Master’s thesis and he will stay with us to pursue his PhD. In his research, he will continue to develop Reinforcement Learning approaches for optimal quantum control. Congratulations, Mo!
Today, Francesca de Franco joined our group as a new PhD student. She will explore how one can use near-term quantum devices to study many-body physics. Welcome Francesca!
For ML+Quantum Physics enthusiasts and all those who would like to learn about it, we are organizing a Tutorial Session “Physics Meets Machine Learning” and a Focus Session “Machine Learning for Complex Quantum Systems” at the DPG Spring Meeting in Dresden. There will also be another Focus Session on “Understanding Machine Learning as Complex Interacting Systems”. All these sessions promise to be very interesting with great invited speakers – so come by!
Our group is expanding and we are currently looking to fill several Postdoc/PhD positions. One job ad is already posted on the FZJ job portal. To learn about further opportunities, just contact Markus.
Update: All positions were filled in the meantime.
Last week, Dominik Hahn (MPI PKS) presented the outcomes of our collaboration on the quantum many-body Jarzynski equality in the Qiskit Seminar Series. The recording is available on youtube.
During the next two days Markus will mentor a team of talented students and young researchers, who will tackle the quantum control problem with reinforcement learning approaches at the scientific hackathon of the Quantum Ideas Factory 2022 in Heidelberg. Thanks to Mo for the help to set up the challenge. That’s going to be fun!
Our work on phase transition dynamics in the two-dimensional quantum Ising model appeared today in Science Advances. Following up on it, an article for non-experts appeared on pro-physik.de.
Today, Wladislaw Krinitsin joined our group as a new PhD student. He will work on neural quantum states for non-equilibrium dynamics. Welcome Wladi!
Today, we are launching our Helmholtz Young Investigator Group “Machine Learning for Quantum Technology” at PGI-8 (FZ Jülich). Within this project we are going to develop machine learning techniques for the advancement of quantum technologies with a focus on two promising applications: On the one hand, we will employ neural quantum states for simulations of non-equilibrium quantum matter, and, on the other hand, we will develop a versatile framework for optimal quantum control based on reinforcement learning.