Papers

The primary challenge of conducting PDE-constrained optimization for high-dimensional problems, such as kinetic equations, is the often prohibitive memory cost. Computing gradients using the adjoint state method would require the storage of the entire time history of the forward solution. For such problems, where the memory cost for storing a single instance of the forward solution can already be a limiting factor, this is clearly not feasible. In this paper, we propose a memory-efficient adjoint state method that compresses the forward and adjoint solution with a dynamical low-rank approximation (a model order reduction technique) and bypasses the need to store the entire forward solution by employing a time-reversible low-rank integrator. The dynamical low-rank approach introduces a number of challenges: reversibility can fail in the rank-deficient case and the low-rank trajectories can show chaotic behavior. In particular, the latter has a number of important consequences for the optimization problem. We address those challenges and show that our method can drastically reduce the memory requirement for gradient-based optimization of kinetic equations. In particular, we consider two examples from kinetic plasma physics: optimizing beam profiles to suppress a bump-on-tail instability and shaping a particle beam using external electric fields.

Neural quantum states provide flexible variational representations of quantum many-body wave functions by combining neural-network parametrizations with Monte Carlo sampling. In this perspective, we review recent advances in their application to condensed-matter systems, focusing on frustrated quantum magnets, interacting lattice fermions, and non-equilibrium dynamics. We discuss the architectures, symmetry constraints, optimization methods, and sampling strategies underlying state-of-the-art calculations, and summarize practical guidelines for reliable simulations. We also examine the principal remaining challenges, including learning non-trivial sign and phase structures, controlling variational bias, enforcing physical symmetries, scaling optimization to large networks, and achieving stable real-time evolution. Finally, we outline promising directions in which neural quantum states may extend the reach of classical simulations of strongly correlated quantum matter.

We study the frustrated $J_1$-$J_2$ Ising model on the honeycomb lattice with ferromagnetic nearest-neighbor couplings fixed at $J_1=1$ and strong antiferromagnetic next-nearest-neighbor interactions, i.e., $J_2 \leq -1/4$. Little is known for this range of $J_2$, whereas for less negative values of $J_2$ the system orders ferromagnetically at low temperatures and appears to remain in the Ising universality class. In previous work it was shown that the model has a largely degenerate ground state, and it was conjectured that there is some kind of phase transition. We introduce a complex-valued nematic order parameter, which can differentiate between the high-temperature paramagnetic phase and the observed partially-disordered stripe phase at lower temperatures. Configurations in this phase consist of stripes of spins parallel with respect to one lattice direction, which collectively meander along the remaining two, producing partially disordered ground states. The sharp peaks in the specific heat observed in earlier work only appear when using periodic boundary conditions and are absent for free boundaries. Additionally, we reveal a striking dependence of the behavior on the aspect ratio of the considered samples. Ultimately, even a careful finite-size scaling analysis for $J_2 = -0.5$ and $J_2 = -1$ is unable to clearly discern between a crossover without any singularities and some form of continuous transition, including the possibility of an infinite-order transition of the Berezinskii-Kosterlitz-Thouless (BKT) type. For $J_2=-1/4$ we find that the system remains disordered at all temperatures and that it exhibits a finite ground-state entropy per site, for which our simulations provide the accurate asymptotic estimate $S(T=0)/N = 0.230\,960\,93(14)$ in the thermodynamic limit $N\rightarrow\infty$.

This work advances a fundamentally new approach to space-based construction by using liquid self organization in microgravity as a generative design and fabrication principle. Building on the LiquiFab method, we demonstrate how minimal surface architectures - traditionally dependent on complex additive manufacturing - can instead emerge directly from the physics of fluid interfaces shaped by programmable boundary conditions. We present a simulation-to-fabrication workflow, that includes a boundary-driven minimal-surface solver integrated in Grasshopper/Rhino, and an experimental system that implements in a neutral buoyancy environment simulating microgravity. This enables the generation of customizable Schwarz-P-inspired with tunable geometry and thickness, illustrating a scalable pathway for material-efficient, on-orbit fabrication. To validate performance in true microgravity, a flight experiment on the International Space Station is scheduled for the first quarter of 2027 and we here detail the additional considerations required for this experiment.

Quantum error mitigation (QEM) is at the very heart of near-term quantum simulations and lattice gauge theories are no exceptions, rather their physical symmetries provide natural consistency checks on noise quantum states. In this work, we exploit the parity and fermion-number symmetries of a gauge theory, the (1+1)-dimensional Schwinger model, under depolarising noise and investigate symmetry verification under digital quantum simulation. We investigate two set-ups - symmetry- sector post-selection in adiabatic state preparation followed by real-time measurements of the chiral condensate and symmetry verification within a variational quantum eigensolver (VQE). In the first case, post-selection reduces the bias in the chiral condensate consistently removing up to 60% of the quantum noise induced error in our system. Motivated by the observed regularity of the residual bias (in the low-noise regime), we further introduce a global-noise calibration obtained from classically accessible smaller lattices and implemented on larger lattices recovering noiseless chiral condensate values within statistical uncertainty. However, in VQE, symmetry verification does not seem to generally improve the optimised parameters or the fidelity of the prepared states, although it reduces the bias in the estimated ground-state energy. This demonstates that improving a noisy cost-function estimator in variational algorithms does not necessarily improve the outcome of the algorithm. Our results show the strength of symmetry verification in different approaches while establishing that its usefulness critically depends on where it is applied in the computational workflow, and provide practical guidance for symmetry-assisted quantum error mitigation in quantum simulations of lattice gauge theories.

{Multislice electron ptychography commonly accounts for inelastic scattering by including an absorptive object potential in the reconstruction forward model that attenuates the elastic signal. However, it remains unclear how the reconstruction is quantitatively impacted by thermal scattering. Here, we use the quantum excitation of phonons (QEP) formalism, which explicitly separates elastic and inelastic (thermal diffuse) scattering within a single multislice propagation, to simulate four-dimensional scanning transmission electron microscopy (4D STEM) datasets of PbTiO$_3$ and SrTiO$_3$ and reconstruct their phase and amplitude. We find that reconstructions of the QEP total and elastic-only datasets are nearly indistinguishable across sample thicknesses from 11 to 40 nm and across atomic species, demonstrating that the reconstruction is largely insensitive to incoherent thermal diffuse scattering within the collected angular range. Reconstructions from multislice phase-only simulations further confirm that appreciable amplitude does not arise from multiple elastic scattering, but instead reflects attenuation of the coherent elastic signal due to thermal scattering. Further comparison of the QEP elastic-only reconstruction with absorptive multislice simulations reveals substantial amplitude contrast deviations for heavy Pb columns (up to 17%), arising from the approximations used to evaluate the absorptive potential for the 4D STEM simulation. These results thus indicate that reconstruction phase and amplitude accuracy are not significantly limited by the absorptive potential forward model, even in the presence of thermal diffuse scattering.

Ultrafast photoexcitation offers a controllable route to steer structural dynamics in solids, yet predicting how nonequilibrium electronic excitation drives lattice motion across extended spatial and temporal scales remains a major computational challenge. Here we introduce time-dependent ab-initio propagation with electronic machine learning (TDAP-eML), a framework that explicitly incorporates electronic evolution into scalable simulations of photoexcitation-induced lattice dynamics. By integrating machine-learned electronic structure with atomistic propagation, TDAP-eML describes how photoexcitation reshapes the evolving energy landscapes and forces governing structural motion. Across representative examples including silicon and FeSe, the framework reproduces key photoexcited lattice responses obtained from first-principles time-dependent density functional theory calculations and captures coherent phonon dynamics together with their dependence on excitation conditions. Its computational advantage increases with system size, reaching nearly three orders of magnitude reduction in computational cost for the large systems examined. TDAP-eML thus establishes a scalable framework for coupled electronic and lattice evolution, linking nonequilibrium excitation to photoinduced forces, predictive structural dynamics, and experimentally accessible observables.

Many modern machine learning models can be understood as kernel-based function-space models, including Gaussian processes and neural tangent kernels. In scientific machine learning, differential operators are increasingly used to encode physical structure directly into such models. However, it has remained unclear under which conditions these constructions are valid machine learning models, i.e. preserve positive semi-definiteness, and whether observed instabilities arise from ill-posed modeling or numerical effects. Here, we establish a simple and sufficient condition for positive semi-definiteness: for linear differential operators of order m, the base kernel must be m-times continuously differentiable. Crucially, this guarantee holds for operators with non-constant and even discontinuous coefficients. Examples are ubiquitous in physical systems, including diffusion, material elasticity, wave propagation in inhomogeneous media, and quantum systems. We conclude that remaining instabilities are attributable to numerical issues, providing a unifying validity foundation for operator-informed kernel methods.

We investigate how the realism of synthetic room impulse response (RIR) datasets affects the training of DeepFilterNet3 for single-channel speech enhancement. We compare a DNS4 image-source-method (ISM) RIR dataset with a higher-acoustic-fidelity dataset generated using hybrid wave-based and geometrical acoustics simulation. Rather than isolating individual simulation factors, we compare complete RIR generation pipelines while keeping the enhancement model unchanged. Models are evaluated on unseen measured RIRs using objective speech enhancement metrics and downstream automatic speech recognition (ASR). Training with the higher-fidelity dataset consistently yields modest improvements in objective metrics and substantially lower ASR word error rates than the ISM dataset. Although the experiments do not attribute these gains to individual modelling components, they show that increasing the overall realism of synthetic acoustic training data improves the generalization of DeepFilterNet3 to unseen measured environments.

The numerical accuracy and computational efficiency of free-space field propagation directly determine what physics can be predicted and what devices can be designed, from nanophotonic metasurfaces to optically addressed quantum processors. Existing methods face a fundamental accuracy-speed trade-off, which is further compounded in the vectorial regime, where major formalisms remain fragmented with no unified computational framework. Here we address both challenges: a theory that unifies the vectorial diffraction framework, and an algorithm that resolves the accuracy-speed trade-off, each reinforcing the other. We prove that all major vectorial diffraction formalisms and vector-potential seeding methods arise from a single surface equivalence principle and reduce, without approximation, to a finite sequence of scalar angular-spectrum operations. Any advance in scalar propagation can therefore be exploited across the entire vectorial ecosystem. The error-compensating angular spectrum method (E-ASM) serves as a fast and accurate scalar solver, delivering peak signal-to-noise-ratio gains exceeding 60 dB over the state-of-the-art band-extended angular spectrum method at order-of-magnitude faster speed, with a freely configurable observation window. Their combination, implemented in a GPU-accelerated, fully differentiable form, turns rigorous vectorial field computation, previously prohibitive for higher-order beams over macroscopic distances, into a practical, inverse-design-ready tool. We demonstrate the framework on optically addressed local qubit gates and on vectorial Hermite-Gaussian and Laguerre-Gaussian beam propagation, resolving field structures inaccessible to scalar or approximate vectorial methods.

Hierarchical equations of motion(HEOM) provide exact dynamics of open quantum systems coupled to harmonic baths, but their computational cost becomes prohibitive for systems with many independent local environments. In this work, we develop a symmetry-adapted HEOM formalism to significantly reduce the computational cost for the permutationally invariant Holstein-Tavis-Cummings (HTC) model. The method removes redundant information in two stages. First, all auxiliary density operators (ADOs) related only by relabeling identical molecules and their bath channels are replaced by a single canonical occupation-pattern representative. Second, molecules with the same local hierarchy occupation produce repeated matrix elements within each representative, allowing only the distinct complex variables to be propagated instead of the full (N + 1) $\times$ (N + 1) ADO matrices. The resulting matrix-free equations are evaluated using precomputed connections and molecular multiplicities. At fixed hierarchy depth L and number of bath correlation exponentials m, the number of canonical representatives becomes independent of the ensemble size for N $\geq$ L and the number of unique variables saturates for N $\geq$ L + 2. The formulation easily extends to multiple-exponential bath decompositions, arbitrary initial density operators, static disorders, and cavity loss. Our benchmarks reproduce conventional HEOM dynamics while requiring far fewer propagated variables and substantially less memory.

Closed-loop discovery systems increasingly execute experiments and update decisions autonomously, turning record integrity into part of the experimental apparatus. We show that false-science induction arises when legitimate physical objects and measurements are paired incorrectly, driving neural surrogates to faithfully learn record-induced associations that do not correspond to the true object-outcome relationship while marginal data distributions remain unchanged. Across green fluorescent protein fitness and materials band-gap prediction loops, coherent paired misbinding systematically redirects experimental budgets toward low-performing basins, whereas same-volume random swaps have negligible effects. These observations identify error coherence, rather than raw error frequency, as the primary variable controlling this budget misallocation in the tested loops. The resulting binding identifiability boundary supports monitored-axis quarantines and feedback-conflict triage, which intercept over-concentrated proposals before execution and isolate the corrupted hypothesis axis.

A power series for the Wigner-Kirkwood pair commutation function for quantum statistical mechanics in classical phase space is given with terms automatically generated by recursion. The calculated second virial coefficient agrees with the measured values of helium for temperatures greater than 65 K. Prospects for a general quantum Monte Carlo algorithm are discussed.

The Marketron model of \cite{HalperinItkin2025Mark} and its option pricing extension in \cite{HalperinItkinMarketron2} suffer from structural non-identifiability: an eighteen-parameter space traps solvers in suboptimal local minima and renders economic quantities unmeasurable. By removing exact scaling gauges and sign symmetries, freezing non-financial parameters by explicit criteria, and adiabatically eliminating the fast hidden signal, we derive a robust nine-parameter reduced model. A Gauss-Newton Hessian with empty null space and a manifold-boundary analysis confirm that the reduced core carries no exact symmetry and admits no further reduction. A diffusive correlation between flow and return innovations captures the short-maturity skew. A staged calibration from the physical measure to the risk-neutral measure, illustrated on SPX options, fits the whole surface with a single parameter set. The same reduction turns the wedge between the physical and pricing values of the flow block into a well-defined market price of flow risk rather than a ridge artifact, identifiable here for the first time, though a single surface constrains its level only weakly. Finally, our analysis reveals that in the Marketron model the log-price obeys a generalized Langevin equation with a closed-form, state-modulated memory kernel, and that the memory variable itself is the exact Markovian lift of this kernel. This mapping also yields a testable condition, the equality of the signal and memory relaxation rates, which on the SPX surface come out well separated, though both weakly identified, placing the fitted market tentatively in the driven, non-equilibrium regime and turning the active-matter reading from an analogy into a falsifiable constraint.

The fabrication of highly ordered laser-induced periodic surface structures (LIPSS) on thin metallic films is dictated, predomaninatly, by a synergy of periodic electromagnetic energy deposition and complex fluid dynamics. In this work, we present a combined experimental and theoretical study on the formation of ultra-regular LIPSS on 32-nm-thick Au films using a double-pulse femtosecond laser scheme. We demonstrate that for thin films, the excitation of coupled Surface Plasmon Polaritons (SPPs) at both interfaces dictates the initial energy distribution. On the othe hand, the final morphology is greatly influenced by hydrodynamical processes. Interestingly, due to the low electron-phonon coupling of Au and the high thermal confinement of the thin film, single-pulse irradiation leads to uncontrolled hydrodynamic instabilities a non uniform topographies. Thus, we demonstrate that a double-pulse approach with an optimized interpulse delay ({\Delta}{\tau}=1.2 ns) effectively controls the melt duration and viscosity, suppressing complex fluid motion and promoting the growth of highly ordered arrays. These structures support narrow surface lattice resonances (SLRs) suitable for high-sensitivity plasmonic biosensing.

We develop analytical and particle-based methods for uncertainty propagation in random neural network models, where both the inputs and network parameters are allowed to be random. Building on the piecewise-linear structure of the Leaky ReLU activation function, we derive a local approximation of the neural network output with respect to perturbations in both its inputs and parameters. This approximation is exact for perturbations that preserve the network activation pattern, and it allows us to compute analytical expressions for the probability density function and characteristic function of the network output, together with closed-form approximations for its mean and covariance. We extend this uncertainty propagation framework to autonomous dynamical systems whose one-step evolution map is represented by a random neural network. Repeated application of this map defines an autoregressive model, for which we derive recursive equations to propagate uncertainty in both the state and network parameters over time. These equations explicitly account for the state-parameter cross-covariance that develops under successive iterations of the network. Numerical experiments on the Lorenz-63 system and the Kuramoto-Sivashinsky equation demonstrate accurate uncertainty propagation through the predictability horizon and the applicability of the proposed framework to high-dimensional dynamical systems.

Selecting the optimal neural-operator prediction during deployment is challenging when high-fidelity reference solutions are unavailable. We demonstrate that under a squared Hilbert-space loss, ranking a finite model library depends strictly on the low-dimensional span of candidate differences, allowing us to score all models simultaneously using a single anchor-based linearized response of the governing equation. This shared physical diagnostic accurately recovered over 99.6% of pairwise preferences and 99.0% of optimal checkpoints across diverse Fourier and convolutional operator libraries for fluid, reaction-diffusion, and wave dynamics. Furthermore, the corrected physical proxy frequently outperformed the best individual candidates, and we establish computable sufficient conditions that rigorously certify exact decisions for strongly monotone discretizations. By exploiting the local dynamical response rather than raw defect magnitude, this framework enables the reliable and highly efficient deployment of scientific surrogates without requiring ground-truth data.

Neural PDE operators are increasingly trained on reusable solver archives, yet validation often relies on clean prediction error and parameter-agnostic plausibility checks. We introduce cross-parameter relinking, a data-poisoning primitive that makes a triggered input select a valid solution from the same PDE family under an incorrect physical parameter. We term this a wrong-physics backdoor: the output remains physically plausible but is wrong for the intended parameter. The attack exploits tensor-to-parameter provenance failures in multi-parameter archives by stamping the surrogate input and relinking its supervision to a cached alternate-parameter solution for the same latent sample. Across 476 attack campaigns, we evaluate Burgers, advection-diffusion, two-dimensional Navier-Stokes, and an elliptic Poisson case. Fourier Neural Operators and DeepONet provide the primary evidence, with Transformer, GRU, and LSTM models as support. FNO reaches a backdoor success rate of 1.0000 on both advection-diffusion and two-dimensional Navier-Stokes while retaining low clean relative L2 error. Clean-label, label-only, and shuffled controls show that high attack success alone is insufficient: successful attacks must move predictions toward the intended alternate-physics target while preserving bounded clean error. These results expose a structural validation gap: smoothness or generic solver-like behavior is insufficient unless the provenance of the intended physical parameter is also verified.

We study the translocation of a polymer chain through a nanopore where the chain length fluctuates stochastically due to the polymerization-depolymerization processes at the chain ends. We map this process to an equivalent representation where the pore performs a stochastic random-walk-like process on a line in the presence of two diffusing sinks on either side of it with diffusion constants $D_1$ and $D_3$ respectively. The translocation process terminates when the pore hits either of the two outer diffusing sinks. In the case where the pore motion itself is diffusive with diffusion constant $D_2$, we compute exactly the splitting probability that the pore hits the left (right) sink before hitting the right (left) sink. We show that the splitting probability in the presence of mobile sinks is rather nontrivial compared to the classical case of immobile sinks (the latter corresponds to the case when the chain length is fixed). Furthermore, we also compute exactly the probability distribution of the translocation time and that of the chain length at the completion time of the translocation. We show that both distributions have power law tails with exponents that depend continuously on the diffusion constants $D_1$, $D_2$ and $D_3$. We validate our analytical predictions via numerical simulations. We then present numerical results for the case when the pore performs a fractional Brownian motion with Hurst exponent $0<H<1$, while the sinks are still diffusive.

Few-mode photonic circuits can increase functionality without multiplying waveguide paths, but bends and fabrication errors can mix their transverse modes. We investigate a strategy in which waveguide confinement and perturbation parity are engineered together in vertically elliptical, femtosecond-laser-written glass waveguides. The intended modal basis comprises the even $1S$ mode and the vertically odd $2P_y$ mode. No window-converged $2P_x$ state is resolved for a lower-confinement (LC) design, whereas a higher-confinement (HC) design guides $2P_x$, which must therefore be isolated by symmetry. In scalar beam-propagation calculations, the $1S$-$2P_y$ propagation-constant splitting predicts the optimized periods of a vertically modulated coherent modal splitter to within $1.0\%$. Horizontal S-bends remain parity-mismatched for $1S \leftrightarrow 2P_y$ coupling, while the symmetry-allowed HC $1S \rightarrow 2P_x$ transfer reaches only $0.6\%$ at the largest displacement. At a displacement of $150~\mu\mathrm{m}$, the HC design retains approximately the same $2P_y$ power as the LC design retains at $40~\mu\mathrm{m}$. Thermal and stochastic writing-error calculations reveal the resulting trade-off: stronger confinement improves modal-power retention, but writing jitter that breaks $x$-parity can populate the guided $2P_x$ mode. These results demonstrate how modal-basis engineering can shift part of the crosstalk-control burden from the trajectory to waveguide symmetry, supporting joint path–mode degrees of freedom in quantum photonic applications.

By combining coherence thermofield dynamics with the single-Hessian approximation, we enable simulations of low- to medium-resolution vibronic spectra of weakly anharmonic systems at nonzero temperatures, at negligible additional cost relative to zero-temperature calculations. Single-Hessian coherence thermofield Gaussian wavepacket dynamics is exact in any harmonic potential, provided that the reference Hessian is that of the final surface. When applied to Morse systems of increasing anharmonicity and varying temperature, this method successfully captures excited-state anharmonicity and key temperature-dependent spectral features, including hot bands and broadening. By combining the method with on-the-fly ab initio dynamics, we demonstrate its utility by computing the absorption spectra of naphthalene, aminocoumarin C450, and phenyl radical, and the photoelectron spectrum of SeO$_{2}^{-}$ . Within the ab initio single-Hessian approximation, after the zero-temperature spectrum is obtained at the cost of classical molecular dynamics (on the order of hours), all nonzero-temperature spectra are computed in seconds.

We summarise the outcome of two summer internship projects based at the University of Manchester, focused on the break-even point in terms of environmental sustainability for ML-based data compression algorithms. Using the example of a ML-based lossless compression algorithm, we compare estimates for the carbon-equivalent of the infrastructure needed for ML training and inference with the carbon-equivalent savings from reduced disk storage requirements, and discuss their break-even point.

We develop a resonant-state expansion formulation for acoustic scattering by individual resonators. The scattered pressure and particle-velocity fields are expanded over the resonant states of the system, with excitation amplitudes determined by overlap integrals between the incident field and the resonant states over the resonator volume. Using the acoustic energy flux, we derive expressions for the extinction, scattering, and absorption cross-sections and show that the extinction spectrum can be resolved into contributions from individual resonant states. The formulation is first validated for a homogeneous two-dimensional cylinder, where it reproduces the analytical Mie-theory solution. We then consider a sectorally perturbed cylinder with coupled azimuthal modes and demonstrate agreement with finite-element simulations. Finally, we combine the eigenvalue and scattering formulations for a material-programmed hard-wall annular metaatom and reproduce its scattering spectra and near fields. The developed framework provides a physically transparent modal approach to acoustic scattering by open resonators with reduced symmetry and spatially structured material parameters.

We present a practical, ab initio time-dependent method using Hagedorn wavepackets to simulate resonance Raman (RR) spectra of polyatomic molecules. Hagedorn functions—Gaussians multiplied by specific polynomials—are used to represent RR initial and final states because these functions are exact solutions to the time-dependent Schrödinger equation for at-most-quadratic potentials and can be propagated at zero cost beyond that of propagating the guiding Gaussian. Using efficient recursive formulae to compute overlaps between Hagedorn wavepackets, we can evaluate RR excitation profiles for arbitrary spectral signals, such as fundamental, overtone, combination, and hot bands. We then construct the Stokes and anti-Stokes RR spectra from these profiles. We first validate the method in a two-dimensional displaced, distorted, and Duschinsky-rotated harmonic model against numerically exact split-operator calculations. Then, we apply the method to compute RR spectra of anthracene by performing dynamics on a 66-dimensional harmonic potential energy surface constructed from density functional theory calculations.

Resonant-state expansion (RSE) is a powerful modal framework for the perturbative analysis of open resonant systems, providing direct access to complex eigenfrequencies and eigenmodes. While RSE is well developed in electromagnetism, a comparably systematic formulation for acoustics remains less established. Here, we develop a general Green-function-based formalism for acoustic RSE and illustrate it for a class of two-dimensional acoustic resonators. Using the resonant states of an analytically solvable cylindrical reference system as a basis, we derive explicit perturbation matrix elements for uniform, radial, and sectoral variations of density and compressibility, representing homogeneous tuning, graded profiles, and symmetry-induced modal coupling. The resulting complex eigenfrequencies and eigenmodes are validated against exact analytical solutions and finite-element simulations, showing excellent quantitative agreement. The framework provides a systematic and physically transparent approach for analyzing perturbed open acoustic resonators and establishes a basis for resonant-state methods in acoustic metamaterials and non-Hermitian acoustics.

In this paper, we study the numerical computation of eigenfunctions of the Landau Hamiltonian with a periodic potential. We propose a spectral method using the eigenfunctions of the Landau operator that allows us to numerically compute the band structure for potentials of different strengths, including the well-studied weak and strong potentials, and additionally systems in the intermediate regime. We then apply the method to simulations across the full range of coupling constants from weak to strong coupling and present an analysis of the band structure as a function of the potential strength.

We use physics-informed neural networks (PINNs) to solve the gravitational quasinormal-mode (QNM) eigenvalue problem for Kerr spacetime directly in the two-dimensional hyperboloidal formulation of the Teukolsky equation. This formulation does not require separation of variables and thus retains the coupled radial–angular structure. Such a scheme provides a prototype for calculating the QNMs of beyond-Kerr black holes for which the perturbation equations are non-separable. Sequences with increasing angular momentum are constructed, reaching close to the extremal limit. We focus on the fundamental modes $(\ell,m,n)=(2,0,0)$, $(2,1,0)$, $(2,2,0)$, $(3,3,0)$ and $(4,4,0)$, together with the first overtone $(2,2,1)$. Independent benchmark evaluation shows that every reported real and imaginary frequency component remains below $0.5\%$ error, with a median deviation of $0.1\%$. This accuracy is maintained in the near-extremal regime, where the damping rate becomes small and the modes are longest-lived. The results establish a non-spectral numerical route to multidimensional black-hole perturbation eigenproblems which does not match the substantially higher precision of dedicated Kerr solvers but offers greater flexibility and requires less analytical pre-processing. Non-separable rotating backgrounds and coupled systems, such as gravitational–electromagnetic Kerr–Newman perturbations, are natural extensions of the same construction.

Atomically precise phosphine-stabilized gold nanoclusters are commonly characterized by single-crystal X-ray diffraction, yet the extent to which these static structures represent finite-temperature behavior remains unclear. To explore the free-energy landscapes, equilibrium populations, and isomerization kinetics of these nanoclusters in the gas phase, we establish a general framework that combines molecular dynamics simulations based on a machine-learned interatomic potential with Markov state models (MSMs). Analysis of the MSMs indicates that experimentally reported crystal structures frequently correspond to minor metastable states or transient configurations rather than the dominant finite-temperature structures. Increasing ligand coverage systematically alters both the thermodynamics and kinetics of structural rearrangements, driving the transition from planar to three-dimensional gold cores while accelerating isomerization dynamics. Moreover, catalytically accessible geometries are often only minor members of the equilibrium ensemble, highlighting a trade-off between structural stability and surface accessibility. These results emphasize that ligand-protected nanoclusters need to be viewed as dynamic ensembles and their finite-temperature behavior cannot be fully captured by their corresponding crystallographic structures alone.

We identify different regimes of quantum state transfer in long-range coupled spin-$\frac{1}{2}$ systems, where naturally occurring power-law interactions enable rapid, high-fidelity transfer with minimal engineering. Across a broad range of interaction profiles, from effectively nearest-neighbour coupling to Coulomb interactions, we show how long-range connectivity fundamentally reshapes the mechanisms underlying information propagation within such systems. For effectively short-range interactions, transfer follows familiar ballistic transfer dynamics: an initially localised excitation spreads across many eigenmodes concentrated within the approximately linear region of the spectrum, enabling robust wavepacket motion. In contrast, increasing long-distance interactions via lowering the power-law exponent $\alpha$ ($\alpha=1-2$) drives a striking transformation, where the initial state becomes confined to progressively fewer eigenmodes, ultimately reducing the dynamics to the coherent participation of only a few states corresponding to the highest eigenenergies. This spectral localization gives rise to emergent long-range oscillations between distant sites, revealing a distinct – and faster – transfer mechanism arising from the intrinsic structure of long-range quantum interactions rather than from full-system engineering pathways.

No interatomic potential has offered universality across chemistry, near-first-principles accuracy and the speed of empirical potentials at once. Here we introduce DPA4C, an equivariant potential whose architecture and compressed CUDA operators are co-designed under deployment constraints to pursue accuracy and efficiency together. Five variants spanning a 49-fold parameter range form the high-throughput end of the measured accuracy–throughput frontier. The largest variant approaches the accuracy of the MACE-Omat models at about two orders of magnitude higher measured throughput. The most compact reduces the energy, force and stress errors of the fastest existing universal MLIP by 61.4%, 48.1% and 34.3% at 1.92 times its saturated throughput. All five variants complete multimillion-atom simulations on a single GPU and run molecular dynamics for 2.048 billion atoms on 1,024 16-GB NVIDIA V100 GPUs at 83.3–91.2% weak-scaling efficiency. Compared with the MEAM empirical potential, DPA4C-Nano reaches 1.8 and 2.5 times the saturated throughput in single-GPU scans on the same V100 hardware for diamond carbon and FCC copper, respectively. DPA4C therefore brings quantum-trained universal accuracy into a regime of speed and system size previously associated with empirical potentials.