SearcharxivSearch

arXiv subjects

Yijian Wu

Publications and source records attributed to Yijian Wu.

7 recordsLinked to original sources

Nonclassical condensation pathways revealed by the multivariable theory of nucleation

We extend classical nucleation theory (CNT) by explicitly incorporating the multidimensional nature of nucleation and the coupled roles of kinetics and thermodynamics. Specifically, we treat the cluster density as an independent variable, within both sharp-interface and diffuse-interface descriptions. The kinetics are governed by dynamical density functional theory. Applied to liquid condensation in the Lennard-Jones system, our two-variable (size--density) and three-variable (size--interface width--density) models reveal nonclassical nucleation mechanism. At low supersaturation, both models recover the classical picture, in which clusters nucleate and grow at the equilibrium liquid density. As supersaturation increases, a nonclassical behavior emerges: the critical cluster density decreases, and the nucleation pathway involves concomitant evolution in cluster size, density, and, within the diffuse-interface description, interfacial width. Our model with diffuse interface reveals a rapid increase in interfacial diffuseness at high supersaturation. Near the spinodal limit, both models predict the critical cluster with diverging sizes, densities approaching that of the metastable initial phase, and vanishing work of formation, which provides a smooth connection between nucleation and spinodal decomposition. Comparison with molecular dynamics simulations demonstrates that both models substantially outperform CNT. However, the weak non-monotonic dependence of the critical cluster density observed at very low supersaturation is captured only by diffuse-interface models. Overall, our findings indicate that CNT should be applied only in the low-supersaturation regime, and our work provides a robust foundation for its refinement beyond this limit.

physics.chem-ph

Taming System Complexity: Demystifying Software Engineering Agents in Diagnosing Linux Kernel Faults

The Linux kernel is a critical system, serving as the foundation for numerous systems. Bugs in the Linux kernel can cause serious consequences, affecting billions of users. Fault localization (FL), which aims at identifying the buggy code elements in software, plays an essential role in software quality assurance. While recent LLM agents have achieved promising accuracy in FL on recent benchmarks like SWE-bench, it remains unclear how well these methods perform in the Linux kernel, where FL is much more challenging due to the large-scale code base, limited observability, and diverse impact factors. In this paper, we introduce LinuxFLBench, a FL benchmark constructed from real-world Linux kernel bugs. We conduct an empirical study to assess the performance of state-of-the-art LLM agents on the Linux kernel. Our initial results reveal that existing agents struggle with this task, achieving a best top-1 accuracy of only 41.6% at file level. To address this challenge, we propose LinuxFL$^+$, an enhancement framework designed to improve FL effectiveness of LLM agents for the Linux kernel. LinuxFL$^+$ substantially improves the FL accuracy of all studied agents (e.g., 7.2% - 11.2% accuracy increase) with minimal costs.

cs.AI

The seeding method: A test case for classical nucleation theory in small systems

Molecular dynamics simulations are widely used to investigate nucleation in first-order phase transitions. Brute-force simulations, though popular, are limited to conditions of high metastability, where the critical cluster and the nucleation barrier are small. The seeding method has recently emerged as a powerful alternative for exploring lower supersaturation regimes by initiating simulations with a pre-formed nucleus. In confined systems (NVT ensemble), the seeded simulations are particularly effective for determining stable cluster properties and provide a stringent test case for classical nucleation theory (CNT). In this work, we perform NVT seeded simulations of Lennard-Jones condensation in small systems and compare them with CNT predictions based on several thermodynamic models, including equations of state, perturbation theory, and ideal gas approximation. We find that CNT accurately predicts stable cluster radii across a wide range of conditions. Notably, even the very simple ideal gas approximation proves useful for initializing seeded simulations. Furthermore, seeded simulation results correspond to the critical cluster radii of infinite systems: CNT predictions with good equations of state show very good agreement with simulations, while the perturbation theory and the ideal gas approximation perform well at low temperatures but deviate significantly at high temperatures.

physics.chem-ph

Nonclassical Nucleation Pathways in Liquid Condensation Revealed by Simulation and Theory

Using state-of-the-art rare-event sampling simulations, we precisely characterize the nucleation of liquid droplets from a supersaturated Lennard-Jones gas and uncover a key physical feature: critical clusters nucleate with a density that differs substantially from that of the macroscopic equilibrium liquid. Our atomistic simulations also reveal a nonclassical nucleation pathway showing simultaneous growth and densification in liquid condensation. We then exploit these insights to develop a two-variable nucleation theory, in which the cluster density is allowed to vary. Our accessible model based on the capillary approximation is able to quantitatively retrieve the numerical results in nucleation rate and critical cluster properties over a large range of supersaturation. Remarkably, the two-variable model successfully captures the observed nucleation pathway. The effectiveness of this integrated numerical and theoretical framework demonstrates that the cluster density is a decisive variable in nucleation, highlighting the limitations of the single-variable description while offering a robust foundation for its refinement.

physics.chem-ph

Classical theory of nucleation applied to condensation of a Lennard-Jones fluid

The classical nucleation theory (CNT) and its modified versions provide a convenient framework for describing the nucleation process under the capillary approximation. However, these models often predict nucleation rates that depart significantly from simulation results, even for a simple Lennard-Jones fluid. This large discrepancy is likely due to the inaccurate estimation of the driving force for nucleation, which most traditional models estimate within the ideal solution approximation. In this study, we address this issue by directly calculating the driving force for nucleation using equations of state (EOS) and integrating this approach into the calculation of nucleation rates within the framework of CNT and its modified model. We apply this method to examine the condensation of a Lennard-Jones fluid and compare the resulting nucleation rates with molecular dynamics (MD) simulation data. Our results demonstrate that at relatively low supersaturation, where the capillary approximation is reasonable, our thermodynamic models exhibit excellent agreement with MD results, significantly outperforming traditional models. At moderate and high supersaturation, our approach continues to show a reasonable agreement with MD results. Furthermore, when comparing the results obtained by using different EOS, we find that more precise EOS generally yield better agreement with MD data.

cond-mat.soft

RegMiner: Towards Constructing a Large Regression Dataset from Code Evolution History

Bug datasets consisting of real-world bugs are important artifacts for researchers and programmers, which lay empirical and experimental foundation for various SE/PL research such as fault localization, software testing, and program repair. All known state-of-the-art datasets are constructed manually, which inevitably limits their scalability, representativeness, and the support for the emerging data-driven research. In this work, we propose an approach to automate the process of harvesting replicable regression bugs from the code evolutionary history. We focus on regression bug dataset, as they (1) manifest how a bug is introduced and fixed (as normal bugs), (2) support regression bug analysis, and (3) incorporate a much stronger specification (i.e., the original passing version) for general bug analysis. Technically, we address an information retrieval problem on code evolution history. Given a code repository, we search for regressions where a test can pass a regression-fixing commit, fail a regressioninducing commit, and pass a working commit. In this work, we address the challenges of (1) identifying potential regression-fixing commits from the code evolution history, (2) migrating the test and its code dependencies over the history, and (3) minimizing the compilation overhead during the regression search. We build our tool, RegMiner, which harvested 537 regressions over 66 projects for 3 weeks, created the largest replicable regression dataset within shortest period, to the best of our knowledge. Moreover, our empirical study on our regression dataset shows a gap between the popular regression fault localization techniques (e.g, delta-debugging) and the real fix, revealing new data-driven research opportunities.

cs.SE

An Empirical Study of Usages, Updates and Risks of Third-Party Libraries in Java Projects

Third-party libraries are a central building block to develop software systems. However, outdated third-party libraries are commonly used, and developers are usually less aware of the potential risks. Therefore, a quantitative and holistic study on usages, updates and risks of third-party libraries can provide practical insights to improve the ecosystem sustainably. In this paper, we conduct such a study in the Java ecosystem. Specifically, we conduct a library usage analysis (e.g., usage intensity and outdatedness) and a library update analysis (e.g., update intensity and delay) using 806 open-source projects. The two analyses aim to quantify usage and update practices holistically from the perspective of both open-source projects and third-party libraries. Then, we conduct a library risk analysis (e.g., potential risk and developer response) in terms of bugs with 15 popularly-used third-party libraries. This analysis aims to quantify the potential risk of using outdated libraries and the developer response to the risk. Our findings from the three analyses provide practical insights to developers and researchers on problems and potential solutions in maintaining third-party libraries (e.g., smart alerting and automated updating of outdated libraries). To demonstrate the usefulness of our findings, we propose a bug-driven alerting system for assisting developers to make confident decisions in updating third-party library versions. We have released our dataset to foster valuable applications and improve the ecosystem.

cs.SE