arXiv subjects

Lingyue Liu

Publications and source records attributed to Lingyue Liu.

4 recordsLinked to original sources

arXiv2025

Capillary force-driven particle orientation in rod networks

Hypothesis: Anisotropic rod particles in capillary suspensions form complex network structures with distinctive orientation patterns and rheological properties that differ significantly from spherical particle systems. By identifying the orientation of individual particles, we are able to acquire invaluable experimental insight into the bulk particle orientation measurements. Experiments: Glass microrods were dispersed in capillary suspensions with varied secondary liquid volume fractions. The resulting microstructural characteristics were analyzed using confocal microscopy. Meanwhile, their rheological properties were measured through rheometry and rheoconfocal techniques. Particle networks were quantified in terms of coordination number, clustering coefficient, and orientation distribution. Findings: As the secondary liquid volume fraction increased, rod networks transitioned from point-to-point contact configurations to side-to-side aligned clusters. Unlike spherical systems, the average clustering coefficient decreased with increasing coordination number, indicating the formation of complex particle cluster configurations beyond simple side-to-side alignment. The rod networks demonstrated higher sensitivity to deformation, and samples with higher side-to-side contact probability exhibit higher viscoplastic fragility. These results provide a foundation for designing advanced materials with precisely tunable mechanical properties through controlled anisotropic particle interactions in capillary suspensions.

cond-mat.soft↗

arXiv2025

Dewetting Fingering Instability in Capillary Suspensions: Role of Particles and Liquid Bridges

This study investigates the fingering instability that forms during stretching of capillary suspensions with and without added nanoparticles. The dewetting process is observed using a transparent lifted Hele-Shaw cell. The liquid bridge is stretched under constant acceleration, and the resulting instability patterns are recorded using two high-speed cameras. Finger-like structures, characteristic of the Saffman-Taylor instability are observed. The total length of the dendrites and the intersecting number of branches are quantified. We reveal the roles of microparticles, nanoparticles, and the secondary liquid during the fingering instability. The addition of microparticles to pure liquid enhanced finger length due to increased particle interactions and nucleation sites for bubbles. The addition of secondary fluid reduces fingering length by forming a strong interparticle network. Incorporation of nanoparticles induces an early onset of cavitation and enhances fingering instability. However, nanoparticles make the capillary suspensions' overall microstructure more homogeneous, reduce the sample variation in fingering patterns, and promote the even distribution of gel on both slides during splitting. These findings highlight the complex interactions governing dewetting in capillary (nano)suspensions. This knowledge has potential applications in microfluidics, 3D printing, and thin-film coatings, where controlling dewetting is crucial.

cond-mat.soft↗

arXiv2024

Enhanced contact flexibility from nanoparticles in capillary suspensions

Hypothesis: Sample-spanning particle networks are used to induce structure and a yield stress, necessary for 3D printing of porous ceramics and paints. In capillary suspensions, a small quantity of immiscible secondary fluid is incorporated into a suspension. By further adding nanoparticles with a range of hydrophobicities, the structure of the bridges and microparticle-microparticle contacts should be modified, resulting in a tunable yield stress and shear moduli. Moreover, the compressibility of these samples, important in many processing and application steps, should be sensitive to these changes. Experiment: The nanoparticle hydrophobicity was altered and their position relative to the microparticles and the bridges was examined using confocal microscopy where the correlation between bridge size and network structure was observed. A step-wise uniaxial compression test on the confocal was conducted to monitor the microparticle movement and structural changes between capillary suspension networks with and without nanoparticles. Findings: Our observation suggests that nanoparticles induce the formation of thin liquid films on the surface of the microparticles, mitigating contact line pinning and promoting internal liquid exchange. Additionally, nanoparticles at microparticle contact regions further diminish Hertzian contact, enhancing the capacity for rearrangement. These effects enhance microparticle movement, narrowing the bridge size distribution.

cond-mat.soft↗

arXiv2022

Learning Large-scale Universal User Representation with Sparse Mixture of Experts

Learning user sequence behaviour embedding is very sophisticated and challenging due to the complicated feature interactions over time and high dimensions of user features. Recent emerging foundation models, e.g., BERT and its variants, encourage a large body of researchers to investigate in this field. However, unlike natural language processing (NLP) tasks, the parameters of user behaviour model come mostly from user embedding layer, which makes most existing works fail in training a universal user embedding of large scale. Furthermore, user representations are learned from multiple downstream tasks, and the past research work do not address the seesaw phenomenon. In this paper, we propose SUPERMOE, a generic framework to obtain high quality user representation from multiple tasks. Specifically, the user behaviour sequences are encoded by MoE transformer, and we can thus increase the model capacity to billions of parameters, or even to trillions of parameters. In order to deal with seesaw phenomenon when learning across multiple tasks, we design a new loss function with task indicators. We perform extensive offline experiments on public datasets and online experiments on private real-world business scenarios. Our approach achieves the best performance over state-of-the-art models, and the results demonstrate the effectiveness of our framework.

cs.LG↗