arXiv subjects

Lailai Zhu

Publications and source records attributed to Lailai Zhu.

At least 19 recordsLinked to original sources

arXiv2026

Revisiting creeping viscoelastic cross-slot flow: Global linear stability and structural sensitivity analyses

The viscoelastic instability of cross-slot flow was first observed experimentally almost half a century ago and reproduced numerically two decades ago, yet its physical origin remains unresolved. We revisit this problem for two-dimensional creeping flow of Oldroyd-B fluid by combining direct numerical simulations, global stability analysis, structural sensitivity analysis, and energy-budget analysis. Our simulations reproduce the canonical pitchfork bifurcation, and the stability analysis consistently predicts the threshold and perturbation growth rates. The leading eigenmode consists of a chiral velocity--stress perturbation that tilts and rotates the birefringent strand generated by the extensional flow. Structural sensitivity and energy-budget analyses identify narrow high-extension-rate ridges within the extensional flow as both the spatial core and energetic source of the instability. In these ridges, the stress-based wavemaker co-localizes with large positive disturbance polymeric stress power density, indicating localized transfer of stored elastic energy to the disturbance flow. Analyses of cross-slot variants with rounded corners and with a centered cylinder further reveal that neither sharp corners nor a free central stagnation point is the essential destabilizing ingredient; rather, the instability originates from elastic-energy release in extension-dominated regions characteristic of cross-slot flow.

physics.flu-dyn↗

arXiv2026

Effects of surface viscosities on the motion of a droplet enclosing a translating particle

We investigate the influence of interfacial rheology on the motion of a compound particle consisting of a viscous droplet enclosing a translating rigid particle in the Stokes flow regime. The droplet interface is modeled using the Boussinesq-Scriven constitutive law, incorporating both surface shear and dilatational viscosities. An exact analytical solution is derived for the concentric configuration, and the analysis is extended to eccentric geometries using a spectral boundary integral method, enabling a systematic examination of confinement, viscosity contrast, and interfacial properties. For concentric configurations, we show that the induced droplet velocity is independent of surface shear viscosity, while surface dilatational viscosity can either enhance or suppress the droplet motion depending on the interplay between confinement and viscosity ratio. This behavior is rationalized in terms of competing effects between reduced interfacial mobility and increased driving force required to maintain the prescribed particle speed. In contrast, when the particle is eccentrically positioned within the droplet, a dependence on surface shear viscosity emerges, leading to a consistent enhancement of droplet motion that becomes more pronounced with increasing eccentricity. The analytical and numerical results are in excellent agreement and reveal how interfacial rheology, confinement, and symmetry breaking jointly govern the dynamics of compound particle systems. These findings provide mechanistic insight and establish a quantitative benchmark for future studies of active compound particles with complex interfaces.

physics.flu-dyn↗

arXiv2026

FlowDA: Accurate, Low-Latency Weather Data Assimilation via Flow Matching

Data assimilation (DA) is a fundamental component of modern weather prediction, yet it remains a major computational bottleneck in machine learning (ML)-based forecasting pipelines due to reliance on traditional variational methods. Recent generative ML-based DA methods offer a promising alternative but typically require many sampling steps and suffer from error accumulation under long-horizon auto-regressive rollouts with cycling assimilation. We propose FlowDA, a low-latency weather-scale generative DA framework based on flow matching. FlowDA conditions on observations through a SetConv-based embedding and fine-tunes the Aurora foundation model to deliver accurate, efficient, and robust analyses. Experiments across observation rates decreasing from $3.9\%$ to $0.1\%$ demonstrate superior performance of FlowDA over strong baselines with similar tunable-parameter size. FlowDA further shows robustness to observational noise and stable performance in long-horizon auto-regressive cycling DA. Overall, FlowDA points to an efficient and scalable direction for data-driven DA.

cs.LG↗

arXiv2025

Hydrodynamic Instability Induces Spontaneous Motion of Floating Ice Discs

Spinning ice discs in nature have been reported for more than a century, yet laboratory experiments have yielded diverse observations and contradictory explanations, leaving the mechanism behind the disc motion elusive. Here we combine numerical simulations and scaling analysis to investigate a freely moving ice disc in a lab-scale water tank. We observe the disc remaining stationary or experiencing spontaneous motion, depending on the disc-water temperature difference and water depth. The motion is initiated by a buoyancy-driven, downward plume arising from water's density anomaly -- its density peaks near $4^\circ$C. Crucially, the plume breaks rotational and mirror symmetries after descending beyond a critical distance due to a thermoconvective instability, thereby inducing the disc to move autonomously. Our findings quantitatively unify disc behaviors observed across independent experiments and establish a predictive criterion for the onset of disc motion. More broadly, we point to a route for thermally-driven transport: coupling of bulk thermoconvection and moving bodies, relevant to geophysical processes such as continental drift and iceberg capsizing.

physics.flu-dyn↗

arXiv2025

Automated Code Development for PDE Solvers Using Large Language Models

Foundation models -- large language models (LLMs) in particular -- have become ubiquitous, shaping daily life and driving breakthroughs across science, engineering, and technology. Harnessing their broad cross-domain knowledge, text-processing, and reasoning abilities for software development, e.g., numerical libraries for solving partial differential equations (PDEs), is therefore attracting growing interest. Yet existing studies mainly automate case setup and execution for end users. We introduce LLM-PDEveloper, a zero-shot, multi-agent LLM framework that automates code development for PDE libraries, specifically targeting secondary developers. By translating mathematical and algorithmic descriptions directly into source code, LLM-PDEveloper generates new solvers/modules and adapts existing ones. This end-to-end math-to-code approach enables a self-augmenting pipeline that continuously expands the codebase of a library, extends its capacities, and broadens its scope. We demonstrate LLM-PDEveloper on three tasks: 1) build a solver for a new PDE, 2) implement new BCs for a given PDE, and 3) modify an existing solver to incorporate additional terms, achieving moderate success rates. Failures due to syntactic errors made by LLMs are analyzed and we propose effective fixes. We also identify the mechanisms underlying certain semantic errors, guiding future research.

cs.SE↗

arXiv2025

Theoretical framework for designing phase change material systems

Phase change materials (PCMs) hold considerable promise for thermal energy storage applications. However, designing a PCM system to meet specific performance presents a formidable challenge, given the intricate influence of multiple factors on the performance. To address this challenge, we hereby develop a theoretical framework that elucidates the melting process of PCMs. By integrating stability analysis with theoretical modeling, we derive a transition criterion to demarcate different melting regimes, and subsequently formulate the melting curve that uniquely characterizes the performance of an exemplary PCM system. This theoretical melting curve captures the key trends observed in experimental and numerical data across a broad parameter space, establishing a convenient and quantitative relationship between design parameters and system performance. Furthermore, we demonstrate the versatility of the theoretical framework across diverse configurations. Overall, our findings deepen the understanding of thermo-hydrodynamics in melting PCMs, thereby facilitating the evaluation, design, and enhancement of PCM systems.

physics.flu-dyn↗

arXiv2025

Colloidal Magnus effect in polymer solutions

Rotating particles moving in fluids undergo a transverse migration via the inertia-induced Magnus effect. This phenomenon vanishes at colloidal scales because inertia is negligible and the fluid flow is time reversible. Yet, recent experiments discovered an inverse Magnus effect of colloids in polymeric and micellar solutions supposedly because their viscoelasticity breaks the time reversibility. Our study shows that classical viscoelastic features -- normal-stress differences and/or shear-thinning cannot explain this phenomenon. Instead, it originates from local polymer density inhomogeneities due to their stress-gradient-induced transport, a mechanism increasingly important at smaller scales -- indeed relevant to colloidal experiments. Incorporating this mechanism into our model leads to quantitative agreement with the experiments without fitting parameters. Our work provides new insights into colloidal motion in complex fluids with microstructural inhomogeneities, offers a simple mechanistic theory for predicting the resulting migration, and underscores the necessity of assimilating these findings in future designs of micro-machinery including swimmers, actuators, rheometers, and so on.

cond-mat.soft↗

arXiv2024

Using Large Language Models for Parametric Shape Optimization

Recent advanced large language models (LLMs) have showcased their emergent capability of in-context learning, facilitating intelligent decision-making through natural language prompts without retraining. This new machine learning paradigm has shown promise in various fields, including general control and optimization problems. Inspired by these advancements, we explore the potential of LLMs for a specific and essential engineering task: parametric shape optimization (PSO). We develop an optimization framework, LLM-PSO, that leverages an LLM to determine the optimal shape of parameterized engineering designs in the spirit of evolutionary strategies. Utilizing the ``Claude 3.5 Sonnet'' LLM, we evaluate LLM-PSO on two benchmark flow optimization problems, specifically aiming to identify drag-minimizing profiles for 1) a two-dimensional airfoil in laminar flow, and 2) a three-dimensional axisymmetric body in Stokes flow. In both cases, LLM-PSO successfully identifies optimal shapes in agreement with benchmark solutions. Besides, it generally converges faster than other classical optimization algorithms. Our preliminary exploration may inspire further investigations into harnessing LLMs for shape optimization and engineering design more broadly.

cs.CE↗

arXiv2024

Enabling microrobotic chemotaxis via reset-free hierarchical reinforcement learning

Microorganisms have evolved diverse strategies to propel in viscous fluids, navigate complex environments, and exhibit taxis in response to stimuli. This has inspired the development of synthetic microrobots, where machine learning (ML) is playing an increasingly important role. Can ML endow these robots with intelligence resembling that developed by their natural counterparts over evolutionary timelines? Here, we demonstrate chemotactic navigation of a multi-link articulated microrobot using two-level hierarchical reinforcement learning (RL). The lower-level RL allows the robot -- featuring either a chain or ring topology -- to acquire topology-specific swimming gaits: wave propagation characteristic of flagella or body oscillation akin to an ameboid. Such flagellar and ameboid microswimmers, further enabled by the higher-level RL, accomplish chemotactic navigation in prototypical biologically-relevant scenarios that feature conflicting chemoattractants, pursuing a swimming bacterial mimic, steering in vortical flows, and squeezing through tight constrictions. Additionally, we achieve reset-free, partially observable RL, where the robot observes only its joint angles and local scalar quantities. This advancement illuminates solutions for overcoming the persistent challenges of manual resets and partial observability in real-world microrobotic RL.

cond-mat.soft↗

arXiv2024

Self-propulsion of an elliptic phoretic disk emitting solute uniformly

Self-propulsion of chemically active droplet and phoretic disk has been widely studied; however, most research overlooks the influence of disk shape on swimming dynamics. Inspired by the experimentally observed prolate composite droplets and elliptic camphor disks, we employ simulations to investigate the phoretic dynamics of an elliptic disk that uniformly emits solutes in the creeping flow regime. By varying the disk's eccentricity $e$ and the P'eclet number $\Pe$, we distinguish five disk behaviors: stationary, steady, orbiting, periodic, and chaotic. We perform a global linear stability analysis (LSA) to predict the onset of instability and the most unstable eigenmode when a stationary disk spontaneously transitions to steady self-propulsion. In addition to the LSA, we use an alternative approach to determine the perturbation growth rate, offering valuable insights into the competing roles of advection and diffusion. The steady motion features a transition from a puller-type to a neutral-type swimmer as $\Pe$ increases, which occurs as a bimodal concentration profile at the disk surface shifts to a polarized solute distribution, driven by convective solute transport. An elliptic disk achieves an orbiting motion through a chiral symmetry-breaking instability, wherein it repeatedly follows a circular path while simultaneously rotating. The swinging periodic motion, emerging from a steady motion via a supercritical Hopf bifurcation, is characterized by a wave-like trajectory. We uncover a transition from normal diffusion to superdiffusion as eccentricity $e$ increases, corresponding to a random-walking circular disk and a ballistically swimming elliptic counterpart, respectively.

physics.flu-dyn↗

arXiv2024

Self-diffusiophoretic propulsion of a spheroidal particle in a shear-thinning fluid

Shear-thinning viscosity is a non-Newtonian behaviour that active particles often encounter in biological fluids such as blood and mucus. The fundamental question of how this ubiquitous non-Newtonian rheology affects the propulsion of active particles has attracted substantial interest. In particular, spherical Janus particles driven by self-diffusiophresis, a major physico-chemical propulsion mechanism of synthetic active particles, were shown to always swim slower in a shear-thinning fluid than in a Newtonian fluid. In this work, we move beyond the spherical limit to examine the effect of particle eccentricity on self-diffusiophoretic propulsion in a shear-thinning fluid. We use a combination of asymptotic analysis and numerical simulations to show that shear-thinning rheology can enhance self-diffusiophoretic propulsion of a spheroidal particle, in stark contrast to previous findings for the spherical case. A systematic characterization of the dependence of the propulsion speed on the particle's active surface coverage has also uncovered an intriguing feature associated with the propulsion speeds of a pair of complementarily coated particles not previously reported. Symmetry arguments are presented to elucidate how this new feature emerges as a combined effect of anisotropy of the spheroidal geometry and nonlinearity in fluid rheology.

physics.flu-dyn↗

arXiv2024

Shaping active matter from crystalline solids to active turbulence

Active matter drives its constituent agents to move autonomously by harnessing free energy, leading to diverse emergent states with relevance to both biological processes and inanimate functionalities. Achieving maximum reconfigurability of active materials with minimal control remains a desirable yet challenging goal. Here, we employ large-scale, agent-resolved simulations to demonstrate that modulating the activity of a wet phoretic medium alone can govern its solid-liquid-gas phase transitions and, subsequently, laminar-turbulent transitions in fluid phases, thereby shaping its emergent pattern. These two progressively emerging transitions, hitherto unreported, bring us closer to perceiving the parallels between active matter and traditional matter. Our work reproduces and reconciles seemingly conflicting experimental observations on chemically active systems, presenting a unified landscape of phoretic collective dynamics. These findings enhance the understanding of long-range, many-body interactions among phoretic agents, offer new insights into their non-equilibrium collective behaviors, and provide potential guidelines for designing reconfigurable materials.

cond-mat.soft↗

arXiv2024

Training microrobots to swim by a large language model

Machine learning and artificial intelligence have recently represented a popular paradigm for designing and optimizing robotic systems across various scales. Recent studies have showcased the innovative application of large language models (LLMs) in industrial control [1] and in directing legged walking robots [2]. In this study, we utilize an LLM, GPT-4, to train two prototypical microrobots for swimming in viscous fluids. Adopting a few-shot learning approach, we develop a minimal, unified prompt composed of only five sentences. The same concise prompt successfully guides two distinct articulated microrobots -- the three-link swimmer and the three-sphere swimmer -- in mastering their signature strokes. These strokes, initially conceptualized by physicists, are now effectively interpreted and applied by the LLM, enabling the microrobots to circumvent the physical constraints inherent to micro-locomotion. Remarkably, our LLM-based decision-making strategy substantially surpasses a traditional reinforcement learning method in terms of training speed. We discuss the nuanced aspects of prompt design, particularly emphasizing the reduction of monetary expenses of using GPT-4.

cs.RO↗

arXiv2023

Tunable Memory and Activity of Quincke Particles in Micellar Fluid

Memory can remarkably modify the collective behaviors of active particles. We show that in a micellar fluid, Quincke particles driven by a square-wave electric field exhibit a frequency-dependent memory. Upon increasing the frequency, a memory of directions emerges whereas the activity of particles decreases. As the activity is dominated by interaction, Quincke particles aggregate and form dense clusters in which the memory of the direction is further enhanced due to the stronger electric interactions. The density-dependent memory and activity result in dynamic heterogeneity in flocking and offer new opportunity for study of collective motions.

cond-mat.soft↗

arXiv2022

Viscoelastic levitation

The effects of viscoelasticity have been shown to manifest themselves via symmetry breaking. In this investigation, we show a novel phenomenon that arises from this idea. We observe that when a dense sphere is rotated near a wall (the rotation being aligned with the wall-normal direction and gravity), it levitates to a fixed distance away from the wall. Since the shear is larger in the gap (between the sphere and the wall) than in the open side of the sphere, the shear-induced elastic stresses are thus asymmetric, resulting in a net elastic vertical force that balances the weight of the sphere. We conduct experiments, theoretic models, and numerical simulations for rotating spheres of various sizes and densities in a Boger-type fluid. In the small Deborah number range, the results are collapsed into a universal trend by considering a dimensionless group of the ratio of elastic to gravitational forces.

physics.flu-dyn↗

arXiv2022

Optimising low-Reynolds-number predation via optimal control and reinforcement learning

We seek the best stroke sequences of a finite-size swimming predator chasing a non-motile point or finite--size prey at low Reynolds number. We use optimal control to seek the globally-optimal solutions for the former and RL for general situations. The predator is represented by a squirmer model that can translate forward and laterally, rotate and generate a stresslet flow. We identify the predator's best squirming sequences to achieve the time-optimal (TO) and efficiency-optimal (EO) predation. For a point prey, the TO squirmer executing translational motions favours a two-fold L-shaped trajectory that enables it to exploit the disturbance flow for accelerated predation; using a stresslet mode significantly expedites the EO predation, allowing the predator to catch the prey faster yet with lower energy consumption and higher predatory efficiency; the predator can harness its stresslet disturbance flow to suck the prey towards itself; compared to a translating predator, its compeer combining translation and rotation is less time--efficient, and the latter occasionally achieves the TO predation via retreating in order to advance. We also adopt RL to reproduce the globally-optimal predatory strategy of chasing a point prey, qualitatively capturing the crucial two--fold attribute of TO path. Using a numerically emulated RL environment, we explore the dependence of the optimal predatory path on the size of prey. Our results might provide useful information that help design synthetic microswimmers such as \textit{in vivo} medical micro-robots capable of capturing and approaching objects in viscous flows.

physics.flu-dyn↗

arXiv2021

Low-Reynolds-number, bi-flagellated Quincke swimmers with multiple forms of motion

In the limit of zero Reynolds number (Re), swimmers propel themselves exploiting a series of non-reciprocal body motions. For an artificial swimmer, a proper selection of the power source is required to drive its motion, in cooperation with its geometric and mechanical properties. Although various external fields (magnetic, acoustic, optical, etc.) have been introduced, electric fields are rarely utilized to actuate such swimmers experimentally in unbounded space. Here we use uniform and static electric fields to demonstrate locomotion of a bi-flagellated sphere at low Re via Quincke rotation. These Quincke swimmers exhibit three different forms of motion, including a self-oscillatory state due to elasto-electro-hydrodynamic interactions. Each form of motion follows a distinct trajectory in space. Our experiments and numerical results demonstrate a new method to generate, and potentially control, the locomotion of artificial flagellated swimmers.

cond-mat.soft↗

arXiv2020

Particle motion nearby rough surfaces

We study the hydrodynamic coupling between particles and solid, rough boundaries characterized by random surface textures. Using the Lorentz reciprocal theorem, we derive analytical expressions for the grand mobility tensor of a spherical particle and find that roughness-induced velocities vary nonmonotonically with the characteristic wavelength of the surface. In contrast to sedimentation near a planar wall, our theory predicts continuous particle translation transverse and perpendicular to the applied force. Most prominently, this motion manifests itself in a variance of particle displacements that grows quadratically in time along the direction of the force. This increase is rationalized by surface roughness generating particle sedimentation closer to or farther from the surface, which entails a significant variability of settling velocities.

physics.flu-dyn↗