arXiv subjects

Hongji Zhang

Publications and source records attributed to Hongji Zhang.

8 recordsLinked to original sources

arXiv2026

Dynamics of amorphous membranes in the two-dimensional limit

Atomically thin mechanical resonators have been realized predominantly in crystalline two-dimensional (2D) materials, such as graphene, where long-range crystalline order sets their elastic properties and defines their nonlinear resonant behavior. Extending these concepts to the amorphous 2D limit has remained largely unexplored. Here, we demonstrate that monolayer amorphous carbon (MAC) forms suspended membranes that support optothermal actuation and sensitive interferometric readout across both linear and nonlinear regimes of its resonant motion. We resolve thermomechanical motion, driven resonances, and multimode spectra in MAC nanodrums. The frequencies of fundamental vibration modes correspond to unusually low pretensions, placing monolayer MAC nanodrums in a regime where geometric nonlinearities, stress heterogeneity, and mode coupling emerge at comparatively low drive powers. Consistently, we observe pronounced nonlinear dynamics, including hardening, softening, and mixed Duffing responses, nonlinear damping, parametrically excited modes, and signatures of intermodal coupling. These results establish MAC as a robust nanoelectromechanical platform and open an experimental route to disorder-governed nanomechanics in the 2D amorphous limit.

cond-mat.mes-hall↗

arXiv2026

HINT: Toward an Executable Hardware-Intent Representation Layer for LLM-Driven RTL Generation

Generating implementation-quality RTL with large language models (LLMs) remains difficult because direct generation must resolve microarchitecture while simultaneously producing and debugging low-level code. We present HINT, an executable hardware-intent intermediate representation layer between behavioral specifications or executable oracles and RTL. HINT makes RTL-relevant microarchitecture explicit, supports pre-RTL checking, and supplies explicit RTL-lowering obligations. We evaluate HINT using both a minimal single-agent flow and a full staged workflow. Across seven operator cases, the HINT-mediated route, with no post-synthesis QoR refinement, produces contract-compliant synthesizable RTL on 7/7 cases; Direct C2RTL and C2HLSC apply to five cases and succeed on 5/5 and 1/5, respectively. Under matched Design Compiler synthesis, HINT reduces area by 5.0\%--26.2\% relative to five manual RTL implementations and by 8.9\%--86.1\% relative to five accepted Direct C2RTL results. RealBench AES and SDC, together with a Vortex VPU synthesizing to 561.67k~$\mu\mathrm{m}^2$, further demonstrate specification-driven, protocol-rich, memory-rich, and hierarchical designs. In the controlled operator study, the HINT-mediated route shows better observed convergence and avoids the severe implementation-quality degradation seen in several direct-generation results.

cs.AR↗

arXiv2026

Atomically Thin Amorphous Carbon with an Ultralow Dielectric Constant

Two-dimensional (2D) materials exhibit excellent properties at monolayer thickness and are viable replacements for various microelectronic components as scaling gradually approaches the atomic limit. Despite significant advancements in the ongoing 2D revolution of integrated circuits, one crucial building block, namely a 2D ultralow-k (ULK) dielectric, remains unreported. The challenge lies in achieving a dielectric constant less than 3, as traditional low-k dielectrics are inherently unstable at the 2D limit due to their amorphous or porous nature. The realisation of ultrathin dielectrics with low-k is also needed to address current bottlenecks in integrated circuits scaling. Specifically, low-k materials are necessary to minimise parasitic capacitances as the distance between conductive elements shrinks below 10 nm. Moreover, advanced architectures like gate-all-around field effect transistors (GAA FET) require even lower dielectric constants (k<2) at sub-3nm thickness. Here, we show that layer-by-layer grown multilayer amorphous carbon (ML-AC), as thin as 0.8 nm, is a mechanically robust 2D ULK dielectric with k of 1.35 and dielectric strength of 28-31 MV cm-1. The lack of any long-range order, its intrinsic 2D nature, sp2 carbon character and low density are all essential for minimising dielectric permittivity. Moreover, ML-AC overcomes the vulnerability of existing dielectrics to ion diffusion degradation with a record metal ion diffusion time to failure (TTF) of 10^10 s for even a single layer. Therefore, otherwise necessary additional layers occupying up to 3 nm can be eliminated, which is especially significant as metal line widths approach 10 nm. Combined with its low-temperature, direct and conformal growth even on a dielectric, these critical features enable substantial improvements in silicon-based semiconductor electronics and ensure compatibility with future 2D electronics.

cond-mat.mtrl-sci↗

arXiv2026

UniRTL: Unifying Code and Graph for Robust RTL Representation Learning

Developing effective representations for register transfer level (RTL) designs is crucial for accelerating the hardware design workflow. Existing approaches, however, typically rely on a single data modality, either the RTL code or its associated graph-based representation, limiting the expressiveness and generalization ability of the learned representations. For RTL, the control data flow graph (CDFG) offers a comprehensive structural representation that preserves complete information, while the code modality explicitly encodes semantic and functional information. We argue that integrating these complementary modalities is essential for a thorough understanding of RTL designs. To this end, we propose UniRTL, a multimodal pretraining framework that learns unified RTL representations by jointly leveraging code and CDFG. UniRTL achieves fine-grained alignment between code and graph through mutual masked modeling and employs a hierarchical training strategy that incorporates a pretrained graph-aware tokenizer and staged alignment of text (i.e., functional summary) and code prior to graph integration. We evaluate UniRTL on two downstream tasks, performance prediction and code retrieval, under multiple settings. Experimental results show that UniRTL consistently outperforms prior methods, establishing it as a more robust and powerful foundation for advancing hardware design automation.

cs.LG↗

arXiv2026

Quantum Criticality in Monolayer Amorphous Carbon

Amorphous solids represent the extreme limit of broken translational symmetry, in which the absence of long-range order removes well-defined crystal momenta and invalidates the Bloch description of electronic states. Monolayer amorphous carbon (MAC) has emerged as a unique realization of a strictly two-dimensional (2D) amorphous lattice defined by a structurally contiguous but topologically disordered $sp^2$-bonded random network devoid of any defined long-range crystal symmetry. From atomic-resolution measurements of multifractal wavefunctions, we show that disorder in MAC effectively localizes the low-energy part of the electronic spectrum but retains an extended critical-like state near the band centre ($E\sim 0$). We conjecture that this state is protected from topological disorder by remnant chiral symmetry surviving within the continuous random network, described by a Wess-Zumino-Witten (WZW) topological term. Near criticality, we verify the multifractal scaling relation $\eta = -\Delta_2$, providing quantitative agreement between independently measured spatial correlation decay and multifractal scaling exponents. Our results are confirmed by atomistic tight-binding calculations that closely mirror the multifractal scaling near $E\sim 0$. Our results establish MAC as the first strictly 2D amorphous electronic system to exhibit Anderson criticality driven purely by topological disorder

cond-mat.dis-nn↗

arXiv2025

Beyond Tokens: Enhancing RTL Quality Estimation via Structural Graph Learning

Estimating the quality of register transfer level (RTL) designs is crucial in the electronic design automation (EDA) workflow, as it enables instant feedback on key performance metrics like area and delay without the need for time-consuming logic synthesis. While recent approaches have leveraged large language models (LLMs) to derive embeddings from RTL code and achieved promising results, they overlook the structural semantics essential for accurate quality estimation. In contrast, the control data flow graph (CDFG) view exposes the design's structural characteristics more explicitly, offering richer cues for representation learning. In this work, we introduce StructRTL, a novel structure-aware graph self-supervised learning framework for improved RTL design quality estimation. By learning structure-informed representations from CDFGs, StructRTL significantly outperforms prior art on various quality estimation tasks. To further boost performance, we incorporate a knowledge distillation strategy that transfers low-level insights from post-mapping netlists into the CDFG-based predictor. Experimental results demonstrate that StructRTL establishes new state-of-the-art results, highlighting the effectiveness of combining structural learning with cross-stage supervision.

cs.LG↗

arXiv2025

DeepRTL2: A Versatile Model for RTL-Related Tasks

The integration of large language models (LLMs) into electronic design automation (EDA) has significantly advanced the field, offering transformative benefits, particularly in register transfer level (RTL) code generation and understanding. While previous studies have demonstrated the efficacy of fine-tuning LLMs for these generation-based tasks, embedding-based tasks, which are equally critical to EDA workflows, have been largely overlooked. These tasks, including natural language code search, RTL code functionality equivalence checking, and performance prediction, are essential for accelerating and optimizing the hardware design process. To address this gap, we present DeepRTL2, a family of versatile LLMs that unifies both generation- and embedding-based tasks related to RTL. By simultaneously tackling a broad range of tasks, DeepRTL2 represents the first model to provide a comprehensive solution to the diverse challenges in EDA. Through extensive experiments, we show that DeepRTL2 achieves state-of-the-art performance across all evaluated tasks.

cs.AR↗

arXiv2021

Synthesis and properties of free-standing monolayer amorphous carbon

Bulk amorphous materials have been studied extensively and are widely used, yet their atomic arrangement remains an open issue. Although they are generally believed to be Zachariasen continuous random networks, recent experimental evidence favours the competing crystallite model in the case of amorphous silicon. In two-dimensional materials, however, the corresponding questions remain unanswered. Here we report the synthesis, by laser-assisted chemical vapour deposition, of centimetre-scale, free-standing, continuous and stable monolayer amorphous carbon, topologically distinct from disordered graphene. Unlike in bulk materials, the structure of monolayer amorphous carbon can be determined by atomic-resolution imaging. Extensive characterization by Raman and X-ray spectroscopy and transmission electron microscopy reveals the complete absence of long-range periodicity and a threefold-coordinated structure with a wide distribution of bond lengths, bond angles, and five-, six-, seven- and eight-member rings. The ring distribution is not a Zachariasen continuous random network, but resembles the competing (nano)crystallite model. We construct a corresponding model that enables density-functional-theory calculations of the properties of monolayer amorphous carbon, in accordance with observations. Direct measurements confirm that it is insulating, with resistivity values similar to those of boron nitride grown by chemical vapour deposition. Free-standing monolayer amorphous carbon is surprisingly stable and deforms to a high breaking strength, without crack propagation from the point of fracture. The excellent physical properties of this stable, free-standing monolayer amorphous carbon could prove useful for permeation and diffusion barriers in applications such as magnetic recording devices and flexible electronics.

cond-mat.mtrl-sci↗