arXiv subjects

Cong Su

Publications and source records attributed to Cong Su.

15 recordsLinked to original sources

arXiv2026

SonarLLM: A Native Sonar--Optical Multimodal Large Language Model for Underwater Perception

Reliable underwater perception requires complementary sensing under variable visibility. Optical cameras capture appearance and semantics but degrade rapidly with turbidity, whereas imaging sonar preserves geometry while exhibiting distinct range-azimuth structure and acoustic artifacts. Existing MLLMs, built primarily on optical encoders, are therefore ill-suited to model sonar or adaptively exploit sonar-optical complementarity. We propose SonarLLM, a sonar-optical MLLM that treats sonar as a native perceptual modality. It combines a sonar-specific encoder, modality-specific physics-aware feature enhancement, and reliability-aware hierarchical fusion to align acoustic structure with optical semantics and dynamically adjust their contributions as sensing quality changes. We also introduce SonarBench, a paired benchmark that spans four tasks: recognition, counting, visual question answering, and captioning; and, across the benchmark, three input settings: sonar-only, optical-only, and fusion. By fixing the scene and sonar observation while varying optical degradation, SonarBench enables controlled measurement of cross-modal complementarity. SonarLLM achieves 72.0% macro accuracy across sonar-only recognition, counting, and VQA, outperforming the strongest baseline by 34.4 percentage points, and 68.7% under fusion, exceeding the best baseline by 25.1 points. For recognition and counting, the fusion-over-optical gain grows from 6.0 to 36.0 points as turbidity increases, indicating the increasing complementary value of sonar under controlled optical degradation. Together, these results show that robust heterogeneous perception depends not only on adding sonar, but on representing and weighting it according to its sensing characteristics.

cs.AI↗

arXiv2026

AirflowAttack: Thermal-Airflow Adversarial Perturbations against Infrared Remote-Sensing Vision-Language Models

Vision-language models (VLMs) are increasingly deployed on infrared (IR) remote sensing imagery in security-critical settings, yet their adversarial robustness remains unexamined. We present AirflowAttack, to our knowledge the first adversarial attack for IR remote-sensing VLMs and the first to weaponize thermal-airflow turbulence as the perturbation prior. A lightweight generator synthesizes a single input-agnostic perturbation regularized toward physically plausible airflow patterns. Optimized on one surrogate CLIP model, it attains a mean zero-shot scene-classification attack success rate (ASR, the fraction of samples whose top-1 class changes) of 48.5% across five diverse CLIP backbones, far exceeding four IR-specific physical baselines (27.7--37.0%). Applied to six state-of-the-art VLMs, it cuts scene-classification accuracy by up to 38.2% relative, yet paradoxically makes some models more confident in their IR analysis, confabulating the perturbation as genuine thermal evidence such as temperature gradients and convection. Ablations show the airflow prior raises physical plausibility at no measurable cost to attack success. Together with a benchmark spanning eleven models and four tasks, these findings expose critical vulnerabilities in the rapidly expanding IR VLM ecosystem.

cs.CV↗

arXiv2025

Direct Observation of Massless Excitons and Linear Exciton Dispersion

Excitons -- elementary excitations formed by bound electron-hole pairs -- govern the optical properties and excited-state dynamics of materials. In two-dimensions (2D), excitons are theoretically predicted to have a linear energy-momentum relation with a non-analytic discontinuity in the long wavelength limit, mimicking the dispersion of a photon. This results in an exciton that behaves like a massless particle, despite the fact that it is a composite boson composed of massive constituents. However, experimental observation of massless excitons has remained elusive. In this work, we unambiguously experimentally observe the predicted linear exciton dispersion in freestanding monolayer hexagonal boron nitride (hBN) using momentum-resolved electron energy-loss spectroscopy. The experimental result is in excellent agreement with our theoretical prediction based on ab initio many-body perturbation theory. Additionally, we identify the lowest dipole-allowed transition in monolayer hBN to be at 6.6 eV, illuminating a long-standing debate about the band gap of monolayer hBN. These findings provide critical insights into 2D excitonic physics and open new avenues for exciton-mediated superconductivity, Bose-Einstein condensation, and high-efficiency optoelectronic applications.

cond-mat.mtrl-sci↗

arXiv2023

Multi-dimensional Fair Federated Learning

Federated learning (FL) has emerged as a promising collaborative and secure paradigm for training a model from decentralized data without compromising privacy. Group fairness and client fairness are two dimensions of fairness that are important for FL. Standard FL can result in disproportionate disadvantages for certain clients, and it still faces the challenge of treating different groups equitably in a population. The problem of privately training fair FL models without compromising the generalization capability of disadvantaged clients remains open. In this paper, we propose a method, called mFairFL, to address this problem and achieve group fairness and client fairness simultaneously. mFairFL leverages differential multipliers to construct an optimization objective for empirical risk minimization with fairness constraints. Before aggregating locally trained models, it first detects conflicts among their gradients, and then iteratively curates the direction and magnitude of gradients to mitigate these conflicts. Theoretical analysis proves mFairFL facilitates the fairness in model development. The experimental evaluations based on three benchmark datasets show significant advantages of mFairFL compared to seven state-of-the-art baselines.

cs.LG↗

arXiv2023

Three-dimensional imaging of buried heterointerfaces

We report three-dimensional (3D) structure determination of a twisted hexagonal boron nitride (h-BN) heterointerface from a single-view data set using multislice ptychography. We identify the buried heterointerface between two twisted h-BN flakes with a lateral resolution of 0.57 {\AA} and a depth resolution of 2.5 nm. The latter is a significant improvement (~2.7 times) over the aperture-limited depth resolution of incoherent imaging modes such as annular-dark-field scanning transmission electron microscopy. This is attributed to the diffraction signal extending beyond the aperture edge with the depth resolution set by the curvature of the Ewald sphere. Future advances to this approach could improve the depth resolution to the sub-nanometer level and enable the identification of individual dopants, defects and color centers in twisted heterointerfaces and other materials.

cond-mat.mtrl-sci↗

arXiv2022

Imaging the electron charge density in monolayer MoS2 at the {\AA}ngstrom scale

Four-dimensional scanning transmission electron microscopy (4D-STEM) has recently gained widespread attention for its ability to image atomic electric fields with sub-{\AA}ngstrom spatial resolution. These electric field maps represent the integrated effect of the nucleus, core electrons and valence electrons, and separating their contributions is non-trivial. In this paper, we utilized simultaneously acquired 4D-STEM center of mass (CoM) images and annular dark field (ADF) images to determine the electron charge density in monolayer MoS2. We find that both the core electrons and the valence electrons contribute to the derived electron charge density. However, due to blurring by the probe shape, the valence electron contribution forms a nearly featureless background while most of the spatial modulation comes from the core electrons. Our findings highlight the importance of probe shape in interpreting charge densities derived from 4D STEM.

cond-mat.mtrl-sci↗

arXiv2022

Deep Learning Coherent Diffractive Imaging

We report the development of deep learning coherent electron diffractive imaging at sub-angstrom resolution using convolutional neural networks (CNNs) trained with only simulated data. We experimentally demonstrate this method by applying the trained CNNs to directly recover the phase images from electron diffraction patterns of twisted hexagonal boron nitride, monolayer graphene and a Au nanoparticle with comparable quality to those reconstructed by a conventional ptychographic method. Fourier ring correlation between the CNN and ptychographic images indicates the achievement of a spatial resolution in the range of 0.70 and 0.55 angstrom (depending on different resolution criteria). The ability to replace iterative algorithms with CNNs and perform real-time imaging from coherent diffraction patterns is expected to find broad applications in the physical and biological sciences.

cond-mat.mtrl-sci↗

arXiv2021

Tuning color centers at a twisted interface

Color center is a promising platform for quantum technologies, but their application is hindered by the typically random defect distribution and complex mesoscopic environment. Employing cathodoluminescence, we demonstrate that an ultraviolet-emitting single photon emitter can be readily activated and controlled on-demand at the twisted interface of two hexagonal boron nitride flakes. The brightness of the color center can be enhanced by two orders of magnitude by altering the twist angle. Additionally, a brightness modulation of nearly 100% of this color center is achieved by an external voltage. Our ab-initio GW calculations suggest that the emission is correlated to nitrogen vacancies and that a twist-induced moir\'e potential facilitates electron-hole recombination. This mechanism is further exploited to draw nanoscale color center patterns using electron beams.

cond-mat.mtrl-sci↗

arXiv2021

Revealing the Br{\o}nsted-Evans-Polanyi Relation in Halide-Activated Fast MoS2 Growth Towards Millimeter-Sized 2D Crystals

Achieving large-size two-dimensional (2D) crystals is key to fully exploiting their remarkable functionalities and application potentials. Chemical vapor deposition (CVD) growth of 2D semiconductors such as monolayer MoS2 has been reported to be activated by halide salts, yet clear identification of the underlying mechanism remains elusive. Here we provide unambiguous experimental evidence showing that the MoS2 growth dynamics are halogen-dependent through the Br{\o}nsted-Evans-Polanyi relation, based on which we build a growth model by considering MoS2 edge passivation by halogens, and theoretically reproduces the trend of our experimental observations. These mechanistic understandings enable us to further optimize the fast growth of MoS2 and reach record-large domain sizes that should facilitate practical applications.

cond-mat.mtrl-sci↗

arXiv2020

Designing Artificial Two-Dimensional Landscapes via Room-Temperature Atomic-Layer Substitution

Manipulating materials with atomic-scale precision is essential for the development of next-generation material design toolbox. Tremendous efforts have been made to advance the compositional, structural, and spatial accuracy of material deposition and patterning. The family of 2D materials provides an ideal platform to realize atomic-level material architectures. The wide and rich physics of these materials have led to fabrication of heterostructures, superlattices, and twisted structures with breakthrough discoveries and applications. Here, we report a novel atomic-scale material design tool that selectively breaks and forms chemical bonds of 2D materials at room temperature, called atomic-layer substitution (ALS), through which we can substitute the top layer chalcogen atoms within the 3-atom-thick transition-metal dichalcogenides using arbitrary patterns. Flipping the layer via transfer allows us to perform the same procedure on the other side, yielding programmable in-plane multi-heterostructures with different out-of-plane crystal symmetry and electric polarization. First-principle calculations elucidate how the ALS process is overall exothermic in energy and only has a small reaction barrier, facilitating the reaction to occur at room temperature. Optical characterizations confirm the fidelity of this design approach, while TEM shows the direct evidence of Janus structure and suggests the atomic transition at the interface of designed heterostructure. Finally, transport and Kelvin probe measurements on MoXY (X,Y=S,Se; X and Y corresponding to the bottom and top layers) lateral multi-heterostructures reveal the surface potential and dipole orientation of each region, and the barrier height between them. Our approach for designing artificial 2D landscape down to a single layer of atoms can lead to unique electronic, photonic and mechanical properties previously not found in nature.

cond-mat.mtrl-sci↗

arXiv2019

Evidence for a pressure-induced phase transition of few-layer graphene to 2D diamond

We unveil the diamondization mechanism of few-layer graphene compressed in the presence of water, providing robust evidence for the pressure-induced formation of 2D diamond. High-pressure Raman spectroscopy provides evidence of a phase transition occurring in the range of 4-7 GPa for 5-layer graphene and graphite. The pressure-induced phase is partially transparent and indents the silicon substrate. Our combined theoretical and experimental results indicate a gradual top-bottom diamondization mechanism, consistent with the formation of diamondene, a 2D ferromagnetic semiconductor. High-pressure x-ray diffraction on graphene indicates the formation of hexagonal diamond, consistent with the bulk limit of eclipsed-conformed diamondene.

cond-mat.mtrl-sci↗

arXiv2019

Molecular monolayer stabilizer for multilayer 2D materials

2D van der Waals materials have rich and unique functional properties, but many are susceptible to corrosion under ambient conditions. Here we show that linear alkylamines are highly effective in protecting the optoelectronic properties of these materials such as black phosphorous (BP) and transitional metal dichalcogenides. As a representative example, n-hexylamine can be applied in the form of thin molecular monolayers on BP flakes with less-than-2nm thickness and can prolong BP's lifetime from a few hours to several weeks and even months in ambient environments. Characterizations combined with our theoretical analysis show that the thin monolayers selectively sift out water molecules, forming a drying layer to achieve the passivation of the protected 2D materials. The monolayer coating is also stable in air, hydrogen annealing, and organic solvents, but can be removed by certain organic acids.

physics.app-ph↗

arXiv2018

Competing dynamics of single phosphorus dopant in graphene with electron irradiation

Atomic-level structural changes in materials are important but challenging to study. Here, we demonstrate the dynamics and the possibility of manipulating a phosphorus dopant atom in graphene using scanning transmission electron microscopy (STEM). The mechanisms of various processes are explored and compared with those of other dopant species by first-principles calculations. This work paves the way for designing a more precise and optimized protocol for atomic engineering.

cond-mat.mtrl-sci↗

arXiv2013

Tailoring exciton dynamics by elastic strain-gradient in semiconductors

As device miniaturization approaches the atomic limit, it becomes highly desirable to exploit novel paradigms for tailoring electronic structures and carrier dynamics in materials. Elastic strain can in principle be applied to achieve reversible and fast control of such properties, but it remains a great challenge to create and utilize precisely controlled inhomogeneous deformation in semiconductors. Here, we take a combined experimental and theoretical approach to demonstrate that elastic strain-gradient can be created controllably and reversibly in ZnO micro/nanowires. In particular, we show that the inhomogeneous strain distribution creates an effective field that fundamentally alters the dynamics of the neutral excitons. As the basic principles behind these results are quite generic and applicable to most semiconductors, this work points to a novel route to a wide range of applications in electronics, optoelectronics, and photochemistry.

cond-mat.mtrl-sci↗

arXiv2012

Novel Two-dimensional Carbon Allotrope with Strong Electronic Anisotropy

Two novel two-dimensional carbon allotropes comprised of octagons and pentagons are proposed based on the first-principles calculations. The two carbon allotropes, named OPG-L and OPG-Z, are found to have distinct properties. OPG-L is metallic, while OPG-Z is a gapless semimetal. Remarkably, OPG-Z exhibits pronounced electronic anisotropy with highly anisotropic Dirac points at the Fermi level. A tight-binding model is suggested to describe the low-energy quasiparticles, which clarifies the origin of the anisotropic Dirac points. Such an anisotropic electronic characteristic of OPG-Z is expected to have wide implications in nano-electronics.

cond-mat.mtrl-sci↗