About

Jianhui Li is associated with Northwestern University's Department of Mathematics, which hosts the RTG Dynamics: Classical, Modern, and Quantum. The research group supports activities aimed at exposing young researchers to the broad field of dynamical systems, including summer schools, research workshops, and undergraduate research experiences. The group's focus is on the study of dynamics, a vibrant area of mathematics that describes the time-evolution of mechanical systems and their abstract models, originating with the work of Poincare in celestial mechanics. The group emphasizes training a broad and strong workforce in dynamical systems, integrating mentoring and public outreach to enhance research and dissemination of the field. The principal investigators involved in this effort include faculty members Bryna Kra, Nir Avni, Aaron Brown, and Jared Wunsch, whose specialties cover a wide range of dynamical fields and intersect with various other areas of mathematics such as number theory, combinatorics, partial differential equations, quantum mechanics, group theory, and representation theory.

Research topics

• Materials science
• Chemical engineering
• Chemistry
• Composite material
• Organic chemistry
• Meteorology
• Optoelectronics
• Physics
• Thermodynamics
• Optics

Selected publications

• Hybrid Network Structure of Hexagonal Boron Nitride-Silicon Carbide Whisker to Improve the Performance of the Polybenzoxazine with KH560-Boron Nitride

  Polymers · 2026-03-29

  articleOpen access

  In this study, NH2-MgO was employed as a crosslinking agent to covalently link boron nitride (BN) and silicon carbide whiskers (SiCw) via an amidation reaction, yielding the BN-MgO-SiCw hybrid filler. The BN-MgO-SiCw/PBz composites were fabricated using a ball-milling-assisted solution mixing method combined with hot-press molding, and their comprehensive properties were systematically evaluated. The results demonstrate that the BN-MgO-SiCw/PBz composite exhibits excellent thermal conductivity, favorable dielectric properties, superior thermal stability, and outstanding mechanical performance. At a filler loading of 50 wt%, the composite achieved a thermal conductivity of 1.41 W/mK, which is substantially higher than that of the KH560-BN/PBz composite (0.91 W/mK) and approximately 5.2 times that of the neat PBz matrix. The dielectric constant (ε) and dielectric loss (tan δ) of the BN-MgO-SiCw/PBz composite were 6.81 and 0.013, respectively, remaining at relatively low levels. The thermal degradation temperature at 30% weight loss (T30) and the heat resistance index temperature (THRI) reached 572 °C and 244 °C, respectively, both higher than those of the KH560-BN/PBz composite at the same filler loading (511 °C and 224 °C). The tensile strength and flexural strength of the BN-MgO-SiCw/PBz composite were 50.0 MPa and 72.3 MPa, respectively, exceeding those of the KH560-BN/PBz composite (39.4 MPa and 56.2 MPa) while remaining slightly below those of the neat PBz matrix. Collectively, these findings indicate that the BN-MgO-SiCw/PBz composite holds great promise as a novel material with well-balanced comprehensive properties, making it a strong candidate for applications in fields such as electronic packaging.

  PublisherOA PDFDOI

• Machine learning-driven alignment architecture of heterogeneous data with transient varying semantics

  Nature Communications · 2026-04-23

  articleOpen access

  Via cross-correlation algorithms or synchronized acquisition of signals, the alignment of heterogeneous data with unknown semantic time shifts and intermittent semantic variations cannot be solved. The shift is caused by different data acquisition principles of sensors, different response discrimination principles using heterogeneous data, etc. Here, we report an unsupervised alignment architecture with a supervised learning model as the kernel to overcome the limitations of brain cognition, perception, and storage in aligning complex heterogeneous data. A set of data with a time shift is input into the kernel model of the architecture to predict the semantic labels, features or continuous values corresponding to another set of data. The time shift corresponding to the maximum testing accuracy or the minimum mean squared error is the alignment parameter for the two heterogeneous datasets. This architecture is expected to serve as a preprocessing step for semantic mining of signals and for information fusion.

  PublisherOA PDFDOI

• Zinc Diffusion Technique for Fabrication of Narrow Band Infrared Detectors

  2025-08-13

  articleSenior author

  Diffusion technique was used to fabricate planar type II superlattice infrared photodetectors. In this work, we present mid-wavelength infrared photodetectors based on III-V InAs/InAs<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1-x</inf>Sb<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</inf> superlattice. Metal Organic Chemical Vapor Deposition growth technique was used to grow the material and perform the diffusion. Zinc was used as a p-type dopant for junction generation. Fabrication technique, diffusion process along with the electrical and optical performance of the planar devices are presented.

  PublisherDOI

• Cavitation characteristics of water guided laser nozzle on water jet fiber stability

  Physics of Fluids · 2025-03-01 · 3 citations

  articleSenior author

  The generation of a stable and reliable water jet fiber is essential for optimizing the performance of water-guided laser systems, as cavitation significantly impacts its stability. This study utilizes computational fluid dynamics to simulate various nozzle structures and analyze the effect of cavitation on water jet fiber stability. Numerical simulations were conducted to examine the behavior of downward conical nozzles both with and without cavitation effects. The research aims to explicate the mechanisms governing cavitation formation and its impact on jet stability. Additionally, this study investigates how nozzle structural parameters, including the length-to-diameter ratio, divergence angle, and orifice diameter, affect jet stability under cavitation conditions. A multi-objective genetic algorithm is subsequently employed to globally optimize the Kriging surrogate model, thereby facilitating the identification of Pareto-optimal solutions for enhancing the stability characteristics of the water jet fiber. High-speed camera visualization was utilized to systematically investigate the stability and fragmentation mechanisms of the water jet. Experiments conducted using a 532 nm green laser source confirmed that the optimized downward conical nozzle can produce a stable water jet fiber. Specifically, an optimized nozzle with a 0.08 mm aperture can generate a stable water jet fiber extending up to 84 mm in length under an inlet pressure of 5.0 MPa, thus meeting the requirements for efficient water-laser coupling. This study provides valuable insights and guidance for enhancing water-guided laser processing technology and its practical engineering applications.

  PublisherDOI

• Tuning the contact time of an impacting droplet by superhydrophobic particles

  Physical review. E · 2025-04-08 · 1 citations

  article

  The interaction between a droplet and superhydrophobic particles has a duration that significantly deviates from the inertia-capillary timescale. Both the rapid and delayed rebounds of the droplet are observed on the superhydrophobic particles with various particle sizes and impact velocities. Typically, the former occurs with little attachment of the particles to the rebounding droplet, while the latter emerges with the signature of a liquid marble formation. We propose a simplified model to predict the contact time of the droplet on the superhydrophobic particles by considering the additional surface deformation underneath the droplet and the possible formation of a liquid marble induced by the particles. This proposed model not only supplies the upper and lower bounds of the contact time that covers all the experimentally measured data, but also provides a way to estimate the effective viscosity in the liquid-particle mixture, which collapses the maximum spreading factor of all tested impact cases to one curve.

  PublisherDOI

• DSAS: A Universal Plug-and-Play Framework for Attention Optimization in Multi-Document Question Answering

  ArXiv.org · 2025-10-14

  preprintOpen access1st authorCorresponding

  While large language models (LLMs) show considerable promise across various fields, they have notable limitations in handling multi-document question answering (Multi-doc QA) tasks. The first challenge is long-range dependency modeling, where LLMs struggle to focus on key information in long texts, which weakens important semantic connections. Second, most LLMs suffer from the ''lost-in-the-middle'' issue, where they have difficulty processing information in the middle of long inputs. Current solutions either truncate global dependencies or demand costly finetuning, ultimately lacking a universal and simple solution for these challenges. To resolve these limitations, we propose Dual-Stage Adaptive Sharpening (DSAS) containing two modules. (i) The Contextual Gate Weighting (CGW) module alleviates ''lost-in-the-middle'' by assessing paragraph relevance through layer-wise attention tracking and position-aware weighting. (ii) The Reciprocal Attention Suppression (RAS) module enhances focus on critical paragraphs by suppressing information exchange between key and irrelevant texts, thus mitigating the limitations in long-range dependency modeling. Notably, DSAS functions as a plug-and-play solution requiring no architectural modifications or extra training parameters. Extensive experiments on four benchmarks demonstrate DSAS's efficacy across mainstream LLMs (Llama, Qwen, Mistral, and Deepseek), with an average F1-score improvement of 4.2% in Multi-doc QA tasks on Llama-3.1-8B-Instruct and Qwen2.5-14B-Instruct. Ablation studies confirm the essential contributions of both the CGW and RAS modules. In addition, detailed discussions in the Appendix further validate the robustness and scalability of DSAS.

  PublisherOA PDFDOI

• 3D-printed flower-inspired evaporator for simultaneous solar seawater desalination and hydrogen production

  Chemical Engineering Journal · 2025-05-16 · 13 citations

  article1st author
  PublisherDOI

• AI-driven advances in metal–organic frameworks: from data to design and applications

  Chemical Communications · 2025-01-01 · 29 citations

  review

  Metal-organic frameworks (MOFs) are a versatile class of porous materials with unprecedented structural tunability, surface area, and application potential in areas such as gas storage, carbon capture, and biomedicine. However, their immense chemical design space poses significant challenges for conventional discovery and optimization methods. Recent advances in artificial intelligence (AI) and machine learning (ML) have introduced transformative capabilities to this field, enabling accurate property prediction, automated structure generation, and synthesis planning at scale. This review provides a comprehensive overview of AI-driven strategies for accelerating MOF research. It discusses key databases, deep learning architectures, generative models, and hybrid AI-simulation frameworks that have reshaped the design and screening of high-performance MOFs. Techniques such as graph neural networks and AL have enabled breakthroughs in structure-property prediction, while integration with robotics is advancing autonomous laboratories. Despite these advancements, challenges remain in data quality, model interpretability, and experimental validation. Future directions include physics-informed ML models, standardized data protocols, and deeper integration of AI with chemical robotics. By highlighting both opportunities and current limitations, this review aims to provide a roadmap for the next generation of AI-accelerated MOF innovation.

  PublisherDOI

• Modeling of InAs/InAs _1-x Sb _x Strained-Layer Superlattice Using a Modified Empirical Tight Binding Method

  IEEE Journal of Selected Topics in Quantum Electronics · 2025-07-21

  articleSenior author

  We report on quantum mechanical based modelling on electronic band structure modeling of ternary-based InAs/InAs<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1-x</sub>Sb<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</sub> strained-layer superlattice using a modified sp3s* empirical tight binding method. By using a virtual crystal approximation with a bowing of the s-on-site tight-binding energy, ternary superlattices layers were modeled. Additionally, a theoretical explanation of atomic segregation in superlattices is suggested and implemented in calculations. Our simulations show good agreement with experimentally measured band gap of InAs/InAs<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1-x</sub>Sb<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</sub> superlattices and their variants.

  PublisherDOI

• Study on the Time-Varying Stiffness Characteristics of Four-Point Contact Ball Bearings

  Lubricants · 2025-03-10 · 5 citations

  articleOpen access

  This paper takes a four-point contact ball bearing of a wind turbine as the research object, analyzes the force and deformation relationship under the combined action of axial load and radial load, obtains the load distribution of rolling elements, and establishes a time-varying stiffness model of four-point contact ball bearings without clearance. The stiffness variation law of the case bearing in one rolling period is analyzed, and the time-varying characteristics of stiffness are characterized by the average stiffness and stiffness amplitude variation rate. The influence laws of the number of rolling elements, initial contact angle, axial load, and radial load on the time-varying characteristics of bearing stiffness are analyzed. The results show that within one rolling period, the average value of axial stiffness is about 2.21 times that of radial stiffness, and the amplitude variation rates of radial stiffness and axial stiffness are 0.0047% and 0.002%, respectively. The time-varying characteristics of both are not obvious. The influence of the number of rolling elements on the two stiffnesses is almost linear, while the influence of axial load on stiffness is small; the initial contact angle is positively correlated with axial stiffness and negatively correlated with radial stiffness. With the increase in radial load, the two stiffnesses also increase. Finally, the stiffness test of four-point contact ball bearings was carried out, and the error between the test value and the theoretical value was less than 15%, which preliminarily verified the correctness of the stiffness model.

  PublisherOA PDFDOI

Frequent coauthors

• Manijeh Razeghi

  64 shared

• Arash Dehzangi

  60 shared

• Donghai Wu

  34 shared

• Ryan McClintock

  Northwestern University

  17 shared

• Luquan Ren

  Jilin University

  13 shared

• Zhichao Ma

  Jilin University

  12 shared

• Hongwei Zhao

  12 shared

• Patrick Willis

  University of Washington

  10 shared

Labs

• RTG Dynamics: Classical, Modern and QuantumPI

Education

• PhD, Electrical Engineering

  Northwestern University