arXiv subjects

Mingyu Yu

Publications and source records attributed to Mingyu Yu.

11 recordsLinked to original sources

arXiv2026

Multi-modal machine learning analysis of GaSe molecular beam epitaxy growth conditions

Autonomous synthesis platforms integrating machine learning with in situ diagnostics have the potential to revolutionize thin-film growth by enabling real-time process optimization and reducing the need for manual tuning. However, their application to molecular beam epitaxy (MBE) remains underdeveloped. Here, we present a machine learning-guided framework for MBE growth of GaSe films, leveraging reflection high-energy electron diffraction (RHEED) as an in situ diagnostic alongside ex situ characterization via X-ray diffraction and atomic force microscopy. Unsupervised learning on RHEED patterns reveals a well-defined boundary between high- and low-quality samples, capturing physically meaningful features. Mutual information analysis shows a strong correlation between RHEED embeddings and rocking curve full-width at half-maximum (fwhm), while the correlation with AFM root-mean-square (RMS) roughness is weak. Among key growth conditions, growth rate most strongly influences fwhm, whereas the Se/Ga flux ratio primarily affects RMS roughness and the RHEED embeddings. Supervised learning models trained to predict fwhm and RMS roughness demonstrate moderate accuracy, with significant improvement achieved by incorporating RHEED embeddings. Furthermore, anomaly detection via residual analysis in supervised learning aligns well with unsupervised classification from RHEED, reinforcing the reliability of the predictive models. This study establishes a data-driven framework for machine learning-assisted MBE, paving the way for real-time process control and accelerated optimization of thin-film synthesis.

cond-mat.mtrl-sci↗

arXiv2026

Beyond Visual Safety: Jailbreaking Multimodal Large Language Models for Harmful Image Generation via Semantic-Agnostic Inputs

The rapid advancement of Multimodal Large Language Models (MLLMs) has introduced complex security challenges, particularly at the intersection of textual and visual safety. While existing schemes have explored the security vulnerabilities of MLLMs, the investigation into their visual safety boundaries remains insufficient. In this paper, we propose Beyond Visual Safety (BVS), a novel image-text pair jailbreaking framework specifically designed to probe the visual safety boundaries of MLLMs. BVS employs a "reconstruction-then-generation" strategy, leveraging neutralized visual splicing and inductive recomposition to decouple malicious intent from raw inputs, thereby leading MLLMs to be induced into generating harmful images. Experimental results demonstrate that BVS achieves a remarkable jailbreak success rate of 98.21\% against GPT-5 (12 January 2026 release). Our findings expose critical vulnerabilities in the visual safety alignment of current MLLMs.

cs.CV↗

arXiv2026

Machine Learning Guided Polymorph Selection in Molecular Beam Epitaxy of In2Se3

Indium selenide (In2Se3), a layered chalcogenide with multiple polymorphs, is a promising material for optoelectronic and ferroelectric applications. However, achieving polymorph-pure thin films remains a major challenge due to the complex growth space. In this work, Bayesian optimization (BO) is successfully leveraged to guide the molecular beam epitaxy growth of In2Se3 on Al2O3 substrates. By training a predictive Gaussian process regressor with sequential learning, we efficiently explored substrate temperature, indium flux, selenium flux, and cracker temperature, reducing experimental trials required for successful synthesis. A {\gamma}-In2Se3 film with 91% phase purity was achieved in fewer than 10 BO run samples. Attempts to isolate {\alpha}-In2Se3 were limited by amorphous film formation at low temperatures, indicating that single-step codeposition is unsuitable for crystalline {\alpha}-In2Se3 on Al2O3. Overall, this study validates BO as a powerful approach for phase-selective growth in complex material systems.

cond-mat.mtrl-sci↗

arXiv2025

Examining the Spin Structure of Altermagnetic Candidate MnTe Grown with Near Ideal Stoichiometry

Altermagnets are a recently-discovered class of materials with magnetic ordering that have a zero net magnetization and a momentum-dependent spin splitting in their band structure, arising from a collinear spin arrangement with alternating polarizations in the crystal lattice. The nickeline-structured manganese telluride ({\alpha}-MnTe) is an attractive altermagnet candidate due to its predicted large spin splitting energy and a transition temperature near 300K. In this work, we present a thorough investigation of the spin structure of {\alpha}-MnTe thin films grown by molecular beam epitaxy with very high crystal quality and low residual magnetization. The epitaxial {\alpha}-MnTe films have a full-width-at-half-maximum of 0.1{\deg} as measured by x-ray-diffraction rocking curves and a root-mean-square roughness below 1 nm. Neutron diffraction measurements confirm the antiferromagnetic order in the {\alpha}-MnTe film and show a N\'eel temperature of 307 K. Polarized neutron reflectometry detects a vanishingly small net magnetization which may be confined to the MnTe/InP interface, highlighting the near-ideal stoichiometry in the sample. In vacuo angle resolved photoemission spectroscopy reveals that the bulk band spectrum of the MnTe films is consistent with the weak altermagnetic order as theoretically predicted and observed for the high symmetry nodal plane in the center of the Brillouin zone. This study establishes optimized growth conditions for the synthesis of stoichiometric {\alpha}-MnTe thin films which exhibit exceptional structural and magnetic ordering, thereby providing a robust platform for the precise characterization of their altermagnetic properties.

cond-mat.mtrl-sci↗

arXiv2025

AJF: Adaptive Jailbreak Framework Based on the Comprehension Ability of Black-Box Large Language Models

Recent advancements in adversarial jailbreak attacks have exposed critical vulnerabilities in Large Language Models (LLMs), enabling the circumvention of alignment safeguards through increasingly sophisticated prompt manipulations. Our experiments find that the effectiveness of jailbreak strategies is influenced by the comprehension ability of the target LLM. Building on this insight, we propose an Adaptive Jailbreak Framework (AJF) based on the comprehension ability of black-box large language models. Specifically, AJF first categorizes the comprehension ability of the LLM and then applies different strategies accordingly: For models with limited comprehension ability (Type-I LLMs), AJF integrates layered semantic mutations with an encryption technique (MuEn strategy), to more effectively evade the LLM's defenses during the input and inference stages. For models with strong comprehension ability (Type-II LLMs), AJF employs a more complex strategy that builds upon the MuEn strategy by adding an additional layer: inducing the LLM to generate an encrypted response. This forms a dual-end encryption scheme (MuDeEn strategy), further bypassing the LLM's defenses during the output stage. Experimental results demonstrate the effectiveness of our approach, achieving attack success rates of \textbf{98.9\%} on GPT-4o (29 May 2025 release) and \textbf{99.8\%} on GPT-4.1 (8 July 2025 release). Our work contributes to a deeper understanding of the vulnerabilities in current LLMs alignment mechanisms.

cs.CL↗

arXiv2024

DDIM-Driven Coverless Steganography Scheme with Real Key

With the advancement of information hiding techniques, generation-based coverless steganography has emerged as an alternative to traditional methods, leveraging generative models to transform secret information into stego-objects rather than embedding it within the redundancy of the cover. However, existing generation-based approaches require pseudo-keys that must be shared between communication parties, leading to high overhead of frequent key exchanges and security risks due to their tight coupling with the secret information. This paper proposes a DDIM-driven coverless steganography scheme that utilizes a real-key mechanism, improving the key management. By integrating reversible data hiding (RDH) and chaotic encryption into generation model, the proposed method eliminates excessive key exchanges and reduces the correlation between the key and the secret information. Furthermore, it requires only a single key negotiation for multiple communication, which lowers overhead. Experimental results demonstrate that the proposed scheme resists substitution attacks, enhancing the security of covert communication.

cs.CR↗

arXiv2024

Quasi-van der Waals Epitaxial Growth of {\gamma}'-GaSe Thin Films on GaAs(111)B Substrates

GaSe is an important member of the post-transition metal chalcogenide family and is an emerging two-dimensional (2D) semiconductor material. Because it is a van der Waals material, it can be fabricated into atomic-scale ultrathin films, making it suitable for the preparation of compact, heterostructure devices. In addition, GaSe possesses unusual optical and electronic properties, such as a shift from an indirect-bandgap single-layer film to a direct-bandgap bulk material, rare intrinsic p-type conduction, and nonlinear optical behaviors. These properties make GaSe an appealing candidate for the fabrication of field-effect transistors, photodetectors, and photovoltaics. However, the wafer-scale production of pure GaSe single crystal thin films remains challenging. This study develops an approach for the direct growth of nanometer-thick GaSe films on GaAs substrates using molecular beam epitaxy. It yields smooth thin GaSe films with the rare {\gamma}'-polymorph. We analyze the formation mechanism of {\gamma}'-GaSe using density functional theory and speculate that it is stabilized by Ga vacancies since the formation enthalpy of {\gamma}'-GaSe tends to become lower than that of other polymorphs when the Ga vacancy concentration increases. Finally, we investigate the growth conditions of GaSe, providing valuable insights for exploring 2D/3D quasi-van der Waals epitaxial growth.

cond-mat.mtrl-sci↗

arXiv2024

Review of Nanolayered Post-transition Metal Monochalcogenides: Synthesis, Properties, and Applications

Nanolayered post-transition metal monochalcogenides (PTMMCs) stand out as promising advanced two-dimensional (2D) materials. Beyond inheriting the general advantages associated with traditional 2D materials, they exhibit unique properties, including a wide bandgap range covering the ultraviolet to the mid-infrared spectral ranges, thickness-dependent bandgap behaviors, good nonlinear optical performance, high thermoelectric coefficients, and ferroelectricity. Consequently, these materials hold significant potential in diverse applications such as photodetectors, field effect transistors, thermoelectrics, ferroelectrics, photovoltaics, and electrochemical devices, especially in the manufacturing of nanoscale devices. However, there is still a lack of systematic understanding of the PTMMC family. This study provides a broad overview of the crystal structures, bandgap structures, synthesis methods, physical properties, and state-of-the-art applications of PTMMC materials with a motif of X-M-M-X (M=Ga, In, Ge, Sn; X=S, Se, Te). An outlook for the development trends is emphasized at the end, underscoring the critical importance of this work to the future exploration of nanolayered PTMMCs.

cond-mat.mtrl-sci↗

arXiv2024

Treatment and Aging Studies of GaAs(111)B Substrates for van der Waals Chalcogenide Film Growth

GaAs(111)B is a semiconductor substrate widely used in research and commercial fields due to its low cost, mature synthesis technology, and excellent properties for manufacturing electronic devices. It is not only used to grow three-dimensional (3D) strongly-bonded materials, but has also been used as a substrate for layered, van der Waals (vdW)-bonded chalcogenide film growth. However, GaAs(111)B wafers cannot be directly used for growing epitaxial vdW chalcogenide films for two reasons: (1) the GaAs surface has a substantial number of dangling bonds that need to be passivated for vdW layers growth; (2) the substrate surface is covered with a thin epi-ready oxide layer which must be removed before film growth. In this paper, we optimize the method for deoxidizing GaAs(111)B substrates under a Se overpressure and successfully create a smooth, deoxidized, and passivated substrate for subsequent growth of vdW chalcogenide materials. We demonstrate the benefits of this method for the growth of vdW chalcogenide thin films using GaSe as a representative of vdW chalcogenides. In addition, we find that severely aged substrates have difficulty maintaining a smooth surface during the deoxidation and passivation process and cause GaSe crystals to nucleate in random shapes and orientations. We describe a method using water droplet testing to determine the age of the substrate. Finally, X-ray photoelectron spectroscopy (XPS) characterization reveals that the natural aging of GaAs(111)B in the air results in an increase in surface oxides, Ga2O3 and As2O3, while exposure to ultraviolet (UV)-ozone not only enhances the contents of these two oxides but also generates a new oxide, As2O5. Our research contributes to expanding the compatibility of GaAs(111)B with diverse growth materials and the production of high-quality heterostructure devices.

cond-mat.mtrl-sci↗

arXiv2022

Epitaxial growth of atomically thin Ga2Se2 films on c-plane sapphire substrates

Broadening the variety of two-dimensional (2D) materials and improving the synthesis of ultrathin films are crucial to the development of the semiconductor industry. As a state-of-the-art 2D material, Ga2Se2 has attractive optoelectronic properties when it reaches the atomically-thin regime. However, its van der Waals epitaxial growth, especially for the atomically-thin films, has seldom been studied. In this paper, we used molecular beam epitaxy to synthesize Ga2Se2 single-crystal films with a surface roughness down to 1.82 nm on c-plane sapphire substrates by optimizing substrate temperature, Se:Ga flux ratio, and growth rate. Then we used a 3-step mode to grow Ga2Se2 films with a thickness as low as 3 tetralayers and a surface roughness as low as 0.61 nm, far exceeding the performance of direct growth. Finally, we found that the surface morphology strongly depends on the Se:Ga flux ratio, and higher growth rates widened the suitable flux ratio window for growing Ga2Se2. Overall, this work advances the understanding of the vdW epitaxy growth mechanism for post-transition metal monochalcogenides on sapphire substrates.

cond-mat.mtrl-sci↗

arXiv2022

Topological insulator-based Dirac hyperbolic metamaterial with large mode indices

Hyperbolic metamaterials (HMMs) are engineered materials with a hyperbolic isofrequency surface, enabling a range of novel phenomena and applications including negative refraction, enhanced sensing, and subdiffraction imaging, focusing, and waveguiding. Existing HMMs primarily work in the visible and infrared spectral range due to the inherent properties of their constituent materials. Here we demonstrate a THz-range Dirac HMM using topological insulators (TIs) as the building blocks. We find that the structure houses up to three high-wavevector volume plasmon polariton (VPP) modes, consistent with transfer matrix modeling. The VPPs have mode indices ranging from 126 to 531, 10-100x larger than observed for VPP modes in traditional media while maintaining comparable quality factors. We attribute these properties to the two-dimensional Dirac nature of the electrons occupying the topological insulator surface states. Because these are van der Waals materials, these structures can be grown at a wafer-scale on a variety of substrates, allowing them to be integrated with existing THz structures and enabling next-generation THz optical devices.

physics.optics↗