741 papers
This paper introduces FDTO, a memory-efficient and accurate optimization-based solver that combines finite-difference time-stepping with body-fitted curvilinear grids to overcome the conditioning and efficiency limitations of existing methods like PINNs and ODIL for solving incompressible flow problems.
This paper presents a novel theoretical and numerical framework to demonstrate that non-rotating quasi-2D viscous flows over topographies settle into steady states where large-scale vortices occupy topographic valleys, a behavior distinct from rotating systems and significant for understanding slow planetary flows.
By applying geometric branch growth and renormalization to real mouse suprachiasmatic nucleus networks, this study reveals that circadian rhythms remain robust across different network scales, demonstrating that average connectivity degree, rather than network size or clustering, is the primary structural driver of rhythmic stability.
This paper presents a robust semi-analytical pseudo-spectral method utilizing mapped associated Legendre polynomials and an advanced exponential time differencing scheme to efficiently and accurately simulate rotating, stratified flows in unbounded cylindrical domains by overcoming the numerical stiffness typically caused by strong shear and fast wave forces.
This study demonstrates that a scalable, dispatcher-led contrail avoidance workflow integrated into standard airline operations significantly reduces contrail formation rates without increasing fuel consumption, as validated by a randomized control trial using satellite imagery.
This paper presents a "Unified Blockage Model" that analytically predicts the performance of rotors in confined flows across arbitrary thrust coefficients and misalignment angles, successfully bridging the gap between existing simplified theories and complex fluid dynamics validated by simulations and experimental data.
This paper establishes that the earthquake b-value arises from the power-law scaling of fault rupture area and slip magnitude within three-dimensional fault networks, linking earthquake statistics directly to fault mechanics and fracture energy dissipation.
This paper introduces Extracted Mode Tracking (EMT), an unsupervised machine learning framework that resolves gravity-capillary wave evolution in vessels with unknown boundary conditions by directly extracting wave modes from spatio-temporal data, thereby enabling quantitative analysis of nonlinear dynamics without requiring prior theoretical modeling.
This paper demonstrates that machine learning models trained to predict the two-electron reduced density matrix (2-RDM) can accurately surrogate correlated wavefunction methods, enabling coupled-cluster-quality electronic structure calculations for large solvated systems at a fraction of the conventional computational cost.
This paper demonstrates that Oscillator Ising Machines outperform Bistable Latch Ising Machines in solving combinatorial optimization problems because the configuration-dependent stability of oscillators allows for the selective destabilization of high-energy states, whereas the uniform stability of latches limits their computational efficiency.
This paper reports the first optical observation of negative ion drift at surface pressure in a He:CF:SF mixture using an optically read-out TPC, demonstrating multi-species ion transport and validating the feasibility of large-scale, low-diffusion optical TPCs for rare event searches.
Experiments at the Scarlet Facility using a $10^{21}^2$ laser on thick tantalum targets revealed that while bare targets produced the highest MeV electron and X-ray yields, thicker front-surface coatings like foam and nanowires enhanced heavy ion acceleration, highlighting the critical role of coating density and thickness in optimizing particle generation and suggesting post-damage crater analysis as a viable method for benchmarking laser absorption.
This paper demonstrates that the Boltzmann-Curtiss formulation, which incorporates both rotational and translational degrees of freedom, significantly improves the accuracy of modeling shock structures and nonequilibrium flow profiles for gases like argon and nitrogen compared to traditional Navier-Stokes equations and Maxwell-Boltzmann-based theories.
This paper proposes a physics-informed Bayesian latent diffusion model that integrates probabilistic encoding, diffusion dynamics, and physics constraints to achieve stable and reliable anomaly detection for new physics searches in collider data.
This paper proposes a logarithmic dark energy model (CDM) constrained by recent cosmological datasets, which yields a higher Hubble constant () that partially alleviates the Hubble tension while remaining statistically competitive with the standard CDM model.
This paper applies thermodynamic and economic principles to reframe AI tokens as physical quantities with measurable costs, establishing a finite global "question budget" to argue that the critical challenge for humanity is not the volume of computable answers but the agency required to determine which questions are worth asking.
This paper utilizes stochastic modeling and Boston University data to predict that the proposed 40% federal research funding cuts under the 2026 Trump Administration would drastically increase the proportion of R1 universities where over half of their faculty face subcritical annual expenditures, thereby threatening the viability of quality research and doctoral programs across the United States.
This paper introduces a maximum-entropy framework for continuous-time temporal networks that factorizes interactions into global time processes and static edge probabilities, enabling closed-form analytical solutions and improved generative modeling via non-homogeneous Poisson processes.
This paper presents the first successful *ab initio* simulation of non-equilibrium recombination in evolving ultracold plasmas by employing a co-moving reference frame and trajectory-based identification criteria, achieving a 20% recombination efficiency that aligns with laboratory measurements.
NMIRacle is a novel two-stage generative framework that integrates IR and NMR spectra with count-aware fragment representations to accurately elucidate molecular structures, outperforming existing baselines across varying levels of complexity.