Neural systems form networks embedded in physical space. Local self-regulation, spanning fast synaptic dynamics to slower homeostatic plasticity, refines this scaffold into collective dynamics that are both stable and flexible. This stability, however, is not guaranteed, as recurrent excitation can push such networks toward runaway activity, reminiscent of epileptic seizures. Yet, biological networks reliably settle into stable regimes between silence and saturation, and they can shift their working regime on demand. It remains unclear whether those stable yet flexible dynamics arise from the same mechanisms, and how much of the network's dynamical repertoire is fixed by its topology alone. In this talk, I combine statistical physics of nonequilibrium phase transitions with spiking models, experimental data, and neuromorphic hardware. Together, these approaches show that topology and local self-regulation jointly define a shared repertoire of collective states - from stationary activity to traveling waves and synchronized bursts - and that external stimulation can steer a network between them. Our results provide insight into how topology and local self-regulation shape the stability-flexibility trade-off in biological neural networks. Extending this framework toward associative learning mechanisms could guide new generations of scalable, bio-inspired AI systems.
On the Sun, active regions tend to emerge close to earlier ones, forming nests that persist over several rotations. On Sun-like stars, the scaling of flux emergence with rotation sets the observed level of magnetic activity. I present recent work on both questions, along with new problems that emerge. First, I describe a density-based clustering method that identifies nests in latitude-sliced longitude-time data. Applied to 151 years of sunspot-group records, it shows that nesting increases with activity level in the time-latitude windows considered. I then show a first application to SDO/HMI magnetograms. Solar Orbiter observations indicate that a few nests can produce most flares near cycle maximum. I also discuss how nesting statistics can be carried over to stellar light curves, which lack latitude information. Second, I turn to the flux emergence and transport model (FEAT) and its extensions. A single scaling of emerging flux with rotation reproduces the disc-integrated mean magnetic field of Sun-like stars and the photometric variability of solar-type *Kepler* stars. Yet it underestimates the Ca II H&K emission of fast rotators. I discuss what this may imply for chromospheric heating, and how coronal-field models of active stars would offer an independent test and a basis for studying their flares. Both lines of work prepare us to interpret PLATO data.
In this study, we present our progress in developing a novel quantum simulator based on neutral ytterbium atoms. By combining itinerant fermionic dynamics in optical lattices with local, state-dependent control via optical tweezers, our aim is to realize a robust, scalable and programmable architecture. The ground- and metastable-state manifold of ytterbium makes it an attractive platform, and its fermionic isotopes provide nuclear-spin degrees of freedom. State-selective control of local Hamiltonian terms is enabled through magic and tune-out wavelengths, offering strong potential for various quantum simulations.
A key recent achievement was the precise determination of the excited-state tune-out wavelength for ytterbium atoms, which was previously hindered by severe losses resulting from inelastic collisions and Raman-induced photon scattering. This was made possible by isolating the atoms in an optical lattice at a magic wavelength, combined with resolved-sideband cooling on the ultra-narrow clock transition. This allows measurements to be taken at shallow lattice depths. The resulting reduction in loss rates significantly enhances the lifetime of the excited clock state. Precision measurements are further enabled by fluorescence imaging in a triple-magic configuration, where molasses cooling on the 3P1 transition is realized using a newly identified magic angle, which is reported here.
In a key step towards the quantum simulation, we realize a high-NA microscopy of fermionic ¹⁷¹Yb in a clock-magic-wavelength lattice. This provides the single-site resolution required to observe the targeted dynamics. Since the atoms are initially loaded into a three-dimensional lattice, a single two-dimensional layer must be isolated for imaging. To this end, we have developed a matter-wave refocusing scheme. Atoms that are initially occupying multiple layers of the vertical lattice are released into a tightly focused light sheet. After evolving for a quarter of a period, the atoms refocus at the center of the harmonic confinement provided by the light sheet. This allows them to be recaptured efficiently into a single lattice layer. We discuss the relevant experimental steps, including clock-state sideband cooling and fluorescence imaging with simultaneous Raman sideband cooling. Together, these results establish the imaging and control toolbox needed to perform hybrid tweezer–lattice experiments with fermionic ¹⁷¹Yb at the single-atom, single-site level. This paves the
