2026-07-10 | | Total: 10
Experimental investigations of forced stably stratified turbulence (SST) have shown that the step-like density profile, made of well-mixed density layers and sharp interfaces alternating along the gravity direction, undergo a slow coarsening dynamics with either decay or merging of interfaces. In this Letter, we focus on the coarsening dynamics phenomenon, by means of Direct Numerical Simulations of forced SST at moderate resolutions, and very long temporal integration. We show that the vertical drift and merging of interfaces is associated to the emergence of spatially-uniform, vertically-sheared helical structures that break the mirror-symmetry of the system. When these are absent, interfaces decay is observed instead. %how the kinetic energy excursions observed at $Fr=0.076$, occurring in parallel to vertical drift of interfaces, are due to the emergence of spatially-uniform, vertically-sheared helical structures that break the mirror-symmetry of the system. This is absent at larger $Fr=0.22$, where interface decay is observed instead. A dynamical correspondence between helicity dissipation rate by buoyancy effects and the vertical buoyancy flux allows to establish a (causal) connection between the chiral structures and the vertical movement of interfaces leading to merging.
Inverse energy transfer to large-scale coherent structures in idealized models of geophysical flows has been of interest for over four decades. Extensive knowledge exists regarding inverse transfer in rotating and stratified dry dynamics, characterized by the Rossby number and a single dry Froude number. The current study includes effects of water and phase changes, with dynamics characterized by the Rossby number and two Froude numbers for unsaturated and saturated environments. Using numerical computations with random forcing, inverse energy transfer is examined for a model with a Boussinesq dynamical core, incorporating water vapor and liquid water in the limit of asymptotically-fast cloud microphysics. Besides kinetic energy, total energy includes buoyant potential energies from each phase, and latent moist energy responsible for potential energy transfer at phase boundaries. The rotation and stratification terms are large and comparable, such that the dry version of the evolution equations is dominated by inverse transfer of pseudo potential vorticity(PV). For fixed Rossby and dry (unsaturated) Froude numbers, compared to dry dynamics, there is a reduction in energy transfer rate, associated with the larger Froude number of saturated regions. The upscale transfer to moist PV is influenced by nonlinear waves at lowest order resulting from nonlinear buoyancy near phase interfaces. These nonlinear waves lead to coherent updrafts and downdrafts roughly aligned with fuzzy, large-scale phase boundaries identified by the time average of a cloud indicator function. Statistical relationships between phase boundaries, updrafts/downdrafts and moist PV are explored in flow regions dominated by moist PV-vortices.
Chaotic systems often evolve on a low-dimensional attractor whose geometry varies from one region to another. We propose a non-intrusive reduced-order model that reads this local geometry by clustering and uses it to shape a radial basis library whose kernels adapt to each region. Fitting the reduced velocity onto this library by one global regularised least-squares solve gives an explicit, differentiable vector field that reproduces the long-term statistics, that is, the invariant measure, without any use of the governing equations. Since a radial basis field decays away from the data and cannot by itself return an escaped state, the integration is stabilised by a kinematic corrector whose magnitude is reported as a measure of how far each result rests on the learned field rather than on the corrector. On Lorenz-63 the model recovers the attractor, its marginal densities, and the positive and neutral Lyapunov exponents, while under-recovering the strong transverse contraction. On Lorenz-96 its valid prediction time is competitive with tuned neural-network and reservoir-computing forecasters, and the invariant measure is reproduced on both the full state and a reduced observable. On the Kuramoto--Sivashinsky equation and the quasiperiodic Kolmogorov flow the model matches the energy distribution and spectrum of an intrusive quantised-local Galerkin model, and improves on a global Galerkin projection of the same dimension, without ever projecting the governing equations.
This paper investigates the behaviour of a fluid characterized by a viscosity simultaneously depending on pressure and shear rate within a Hele-Shaw cell featuring a sharp corner geometry. The study extends previous analyses conducted on purely pressure-dependent (piezo-viscous) and yield-stress fluids, providing a new perspective on confined complex flows. Motivated by practical applications related to designing biomedical devices and flows of relevance to biomedicine area, thin film technologies, injection molding -- to name only a few -- the flow configuration considered here can highlight essential features of complex fluid behavior in narrow-gap geometries around a sharp edge. Starting from the governing equations for an incompressible generalized Newtonian fluid and employing an appropriate rheological model, we derive the modified flow equations adapted to the Hele-Shaw flow. A particular solution is obtained near the corner region. Numerical simulations complement the theoretical results, illustrating the influence of the rheological parameters on the flow behavior.
Inspired by biological systems, extensive research has explored how fluid-structure interactions in compliant channels and confined geometries control fluid transport. While local nonlinearities can be induced by individual components, arranging these elements into larger architectures gives rise to increasingly complex, collective responses. Predicting these collective behaviors, however, remains largely restricted to steady-state characterization, as the dynamic coupling between time-varying flows and multiple interacting structures is difficult to model. In this paper, we investigate numerically the collective interaction of multiple asymmetric leaflets within a channel at low-Reynolds number. By utilizing symmetrically oscillating plates rather than a pressure-driven flow to isolate the system from background asymmetries, we characterize how these interacting structures generate a net fluid transport. We develop an analytical framework to evaluate transport in the steady limit, which we subsequently extend to account for time-dependent channel oscillations, providing a complete dynamic description of the coupled fluid-structure system. Our results demonstrate that high leaflet densities maximize collective interactions and net transport. Furthermore, we define an elastoviscous number comparing viscous hydrodynamic forces to the restorative elastic forces of the leaflets, and uncover an optimal value that maximizes the net flow. This framework establishes a foundation for analyzing how collective slender structures interact dynamically within viscous environments, laying the groundwork for future studies on flow control in biological fluid transport and microfluidic design.
Determining absolute/convective instability boundaries conventionally requires repeated saddle searches in the complex-wavenumber plane and a subsequent scan of the physical parameter space to locate zero absolute growth. Such nested calculations become costly and sensitive to modal branch association for large non-normal eigenvalue problems. This work develops a direct continuation method for neutral stationary-saddle boundaries of frequency-affine generalised eigenvalue problems. The zero-group-velocity condition is expressed as an adjoint solvability residual and solved together with the direct and adjoint eigenproblems, complex gauge constraints and the neutral-growth condition. The resulting one-dimensional solution manifold in the combined state--parameter space is tracked by scaled pseudo-arclength continuation, allowing parameter folds to be crossed without switching the physical continuation variable. The formulation recovers the analytical Ginzburg--Landau boundary and, for a Gaussian-wake Orr--Sommerfeld problem, agrees with separately formulated finite-difference saddle corrections to approximately $10^{-8}$ in relative critical Reynolds number. Compared with nested complex-wavenumber and parameter-plane saddle scanning, the scanning calculations require $14.0$--$30.6$ times the wall time of the direct adjoint continuation, with the cost increasing as the reconstructed boundary is refined. Application to a coupled Oldroyd--B free-surface film reveals genuine folds of the neutral-saddle manifold and a re-entrant CI--AI--CI boundary geometry for the selected saddle family. The results show that adjoint-augmented pseudo-arclength continuation can replace nested saddle searches and parameter-plane reconstruction by direct and computationally efficient tracking of the neutral boundary itself.
The transport of inertial particles in unsteady flows is often governed by the competition between multiple migration mechanisms. We investigate the interplay between inertia-induced drift and thermophoretic migration in a time-periodic vortex containing a localized temperature field. Starting from the small-Stokes-number limit of the particle equations of motion, we derive a cycle-averaged radial migration model describing the slow evolution of suspended particles over timescales much longer than the forcing period. The competition between outward inertia-induced drift and inward thermophoretic migration gives rise to stable and unstable fixed points of the reduced radial dynamics, corresponding respectively to particle trapping states and separatrices bounding trapped trajectories. The existence and location of these states are shown to be governed by the dimensionless control parameter $Π$, which measures the relative strength of inertia-induced transport to the thermophoretic transport. As $Π$ is decreased below a critical value, the stable and unstable fixed points coalesce and disappear, resulting in the loss of particle trapping. Phase portraits, bifurcation diagrams, and local asymptotic analysis demonstrate that trapping is destroyed through a saddle-node bifurcation. The transition is further characterized by the vanishing of the dominant eigenvalue and the associated divergence of the relaxation time, indicative of critical slowing down. Additional calculations employing various other velocity and temperature profiles demonstrate that the trapping-loss mechanism is robust and not specific to a particular profile.
Physics-informed neural networks (PINNs) have shown remarkable prospects in solving forward and inverse problems involving partial differential equations (PDEs). However, PINNs still face challenges in solving fluid mechanics problems involving shocks, especially in steady inviscid flows around an airfoil, where they may even fail to capture shocks. In this study, we first point out that the reason PINNs fail to capture shocks is that the steady Euler equations used to construct the loss function impose weak constraints, which are difficult to correct the continuous function approximation preference of neural networks, causing gradient descent converges to a smooth local optimum. Based on this insight, we propose to strengthen the physical constraints by reconstructing steady shock capturing as temporal evolution that gradually converges to the steady state solution. The unsteady Euler equations constructed by introducing time derivative terms into the steady equations are used to constrain PINNs. The output of PINNs is no longer required to directly approximate a flow field with shocks by minimizing the residuals of the steady Euler equations. Instead, shocks gradually form under the guidance of the temporal evolution law of the flow field. This additional temporal penalty alleviates the tendency of PINNs to converge to a smooth local optimum. Since obtaining the steady state solution requires solving the unsteady Euler equations over a long time in the time dimension, while the capability of PINNs to solve such problems is poor, we introduce a PDE loss function that embeds the concept of pseudo time-stepping to avoid this issue. In addition, to further improve the shock capturing accuracy, we develop a simplified formulation of the Euler equations. By solving four forward problems involving different flow conditions and geometries, we validate the effectiveness of the proposed method.
While the direct simulation Monte Carlo (DSMC) is a mainstream stochastic particle method for simulating rarefied gas flows, it incurs excessively high computational costs in the near continuum regime. As a hybrid acceleration approach coupling DSMC with macroscopic synthetic equations, the direct intermittent general synthetic iterative scheme (DIG) delivers fast convergence and asymptotic-preserving characteristics, which effectively alleviate the kinetic scale limitations inherent to standard DSMC. In this study, we develop a parallel DIG augmented DSMC solver for three dimensional rarefied gas flow simulations on unstructured meshes. On top of the standard DSMC algorithms for particle transport and collisions, a reliable intermittent coupling framework is constructed to exchange macroscopic flow data between the stochastic DSMC module and deterministic macroscopic synthetic equations. For parallel execution on unstructured grids, we employ a hybrid MPI architecture equipped with ghost cells to enable local particle tracking and batch inter-rank particle migration. A graph partitioning based dynamic load balancing strategy is also integrated to mitigate uneven particle distribution over computational domains. Numerical results demonstrate that the proposed solver achieves satisfactory agreement with the SPARTA DSMC. Leveraging the fast convergence and asymptotic-preserving properties of the DIG method, the required number of spatial cells and statistical sampling steps are drastically decreased, leading to substantial reductions in computational memory and runtime. This work presents an efficient high-performance numerical tool for high-fidelity simulations of rarefied flows over complex geometries. The code is available in the developer repository at the github link.
Turbulence in the solar interior and atmosphere plays a crucial role in energy transport, yet modeling its subgrid-scale effects remains a major challenge. This study leverages machine learning (ML) models to predict components of the Reynolds stress tensor using high-resolution StellarBox simulations of the quiet Sun. Previously, we have compared a Multi-Layer Perceptron (MLP) and a 3D Convolutional Neural Network (CNN) against physics-based baselines to achieve a lower Mean Squared Error (MSE) and better generalization across various heights and depths in the solar atmosphere. To enhance learning, in this work, we investigate cluster-weighted training using K-Means and Hierarchical Agglomerative Clustering (HAC). By weighing the loss function based on cluster-specific prediction errors, we direct the model's attention to high-error regions. It significantly improves CNN performance, achieving 34% lower MSE and a significantly higher R2 score indicating that integrating deterministic clustering with ML is a promising technique for modeling subgrid turbulence, in particular, and regression in diverse environments, in general.