arXiv subjects

Shuze Zhu

Publications and source records attributed to Shuze Zhu.

14 recordsLinked to original sources

arXiv2026

Strain-Programmable Interior Topological Channels

Topological channels are typically pinned to physical edges or heterogeneous interfaces. Physical edges expose the channels to localized disorder, whereas interface-based designs generally involve material or structural discontinuities. These constraints limit geometric tunability and can compromise microscopic transport quality. Here we show that smooth strain can program topological channels within a single-component homogeneous lattice through a mechanically constrained inverse-design framework. A strain-dependent Dirac mass provides a direct geometric control principle: its zero contour defines the channel path, the contour-crossing topological mismatch fixes the net chirality, and the normal mass gradient sets the confinement width following an inverse-square-root scaling law. Unlike conventional pinned channels, these strain-engineered interior channels permit continuous control over channel geometry and confinement. We demonstrate the approach in a strained Haldane model, realizing straight channels at prescribed orientations, designed curved paths, and multichannel networks associated with higher-Chern-number phases. The resulting channels exhibit quantized transmission and spatial decoupling from boundary-localized disorder, establishing strain as a mechanical design field for programmable interior topological transport.

cond-mat.mes-hall↗

arXiv2026

Variational Learning of Physical Intuition from a Few Observations: Charting Manifolds of Variational Physics

Humans often generalize physical outcomes from few observations, a desirable capacity known as physical intuition. We show that it can be computationally approached through charting the manifolds of variational physics. We conceive a variational learning framework, where small neural networks trained from merely two or three examples delineate an observation-spanned solution manifold, enabling generalization in unseen scenarios. As demonstrated across classical and quantum regimes including strongly correlated molecules, generalization emerges far beyond the training conditions. This generalization is explained by a unified theory: the Euler-Lagrange operator must be stationary with respect to the observation parameters along the solution manifold, whose complexity predicts a critical network capacity below which generalization fails. Our framework establishes a principled route to solving families of variational problems from a few instances, and shows how physical intuition can be approached through harnessing variational physics.

physics.comp-ph↗

arXiv2026

Real-Symmetric Hamiltonian Enables Near-Linear Scaling for Fast Million-Atom Electronic Structure Computations

The exploration of quantum phenomena in mesoscale materials, such as moire superlattices, is limited by the cubic scaling cost of conventional electronic structure methods. Here, we introduce a scalable tight binding framework that achieves near linear scaling, enabling mesoscopic quantum simulations. By transforming the complex Hermitian Bloch Hamiltonian into an equivalent real symmetric form, the method avoids dense diagonalization by combining sparse LDL decomposition with Sylvester's law of inertia for spectral slicing and global rank calibration. This formulation enables efficient band structure calculations for large scale systems, solving magic angle twisted bilayer graphene in minutes on a standard laptop and extending to 1.5 million atoms within days on a single workstation. Applying this framework to ultra low twist angle structures with atomistic strain relaxation, we find robust isolated low-energy band clusters over several finite ultra low angle windows down to 0.09 degree. Our framework provides an efficient computational platform for studying quantum materials at experimentally relevant length scales and supports data driven discovery in large scale moire systems.

physics.comp-ph↗

arXiv2026

Acquiring Human-Like Data-Efficient Mechanics Prediction from Deep Reinforcement Learning

Humans can infer mechanical outcomes by learning from a few observations. This capacity for mechanics intuition is acquired in a data-efficient manner. Here, we propose a reinforcement learning framework to mimic this process, in which an agent encodes continuous physical observation parameters into its state and is trained via episodic switching across closely related observations. With merely two or three similar observations, the agent acquires robust mechanics intuition that generalizes over wide parameter ranges beyond the training data. Our method is demonstrated on the brachistochrone, a large-deformation elastic plate, and the quantum harmonic oscillator. We explain this generalization through a unified theoretical view: it is associated with cross-parameter Bellman consistency encouraged by episodic switching across neighboring task parameters, promoting approximate stationarity of the Bellman residual with respect to physical variations. This is consistent with a smooth policy that tracks a low-dimensional solution manifold underlying the continuum of tasks. Our work identifies episodic switching as a practically effective and theoretically motivated route to artificial mechanics intuition and suggests a computational analogy to data-efficient generalization in biological learners.

physics.comp-ph↗

arXiv2025

Developing Artificial Mechanics Intuitions from Extremely Small Data

Humans can possess good mechanics intuitions by learning from a few examples, which leads to the question of how to develop artificial mechanics intuitions that can be learned from small data, as we are eagerly entering the era of artificial intelligence. We propose in this Letter the sample-switchable training method, which successfully develops highly-accurate artificial mechanics intuitions that can master brachistochrone problem, catenary problem, and large nonlinear deformation problem of elastic plate by learning from no more than three samples. The model's intuitive prediction ability increases nonlinearly with respect to the number of training samples, suggesting that superb mechanics intuitions can be in-principle achieved based on a finite number of samples, reflecting how human brains form good mechanics intuitions just by learning a few cases. Our current work presents an alternative perspective for educating artificial intelligence capable of intuitively understand and predict how materials deform and move, a scenario that has been frequently seen in Science-Fiction movies.

cs.CE↗

arXiv2025

Programming frictionless interfaces for moiré layers

Structural superlubricity in van der Waals layered systems holds immense promise for diverse nanoscale contacts devices and energy-efficient applications. While all-direction structural superlubricity has been widely investigated, the understanding towards the more fundamental directional structural superlubricity requires further attentions. In this study, we reveal the physical origins of directional structural superlubricity, which reduces to all-direction superlubricity under certain conditions. By investigating the evolution of incomplete moiré tiles at crystalline interfaces, our general scaling approaches establish the mapping from geometry to tunable directional superlubricity, agreeing with large scale molecular dynamics simulations at both homogeneous or heterogeneous interfaces. Furthermore, diverse programmable frictionless motions of nanoflakes traveling inside double-surface nanoconfinement systems can be achieved. Our work delivers new insights into the design of ultra-low frictional interfaces for future nanoscale tribology and nanoconfinement transport.

cond-mat.mes-hall↗

arXiv2024

MGNN: Moment Graph Neural Network for Universal Molecular Potentials

The quest for efficient and robust deep learning models for molecular systems representation is increasingly critical in scientific exploration. The advent of message passing neural networks has marked a transformative era in graph-based learning, particularly in the realm of predicting chemical properties and expediting molecular dynamics studies. We present the Moment Graph Neural Network (MGNN), a rotation-invariant message passing neural network architecture that capitalizes on the moment representation learning of 3D molecular graphs, is adept at capturing the nuanced spatial relationships inherent in three-dimensional molecular structures. MGNN demonstrates new state-of-the-art performance over contemporary methods on benchmark datasets such as QM9 and the revised MD17. The prowess of MGNN also extends to dynamic simulations, accurately predicting the structural and kinetic properties of complex systems such as amorphous electrolytes, with results that closely align with those from ab-initio simulations. The application of MGNN to the simulation of molecular spectra exemplifies its potential to significantly enhance the computational workflow, offering a promising alternative to traditional electronic structure methods

physics.comp-ph↗

arXiv2020

Selection rules of twistronic angles in 2D material flakes via dislocation theory

Interlayer rotation angle couples strongly to the electronic states of twisted van der Waals layers. However, not every angle is energetically favorable. Recent experiments on rotation-tunable electronics reveal the existence of a discrete set of angles at which the rotation-tunable electronics assume the most stable configurations. Nevertheless, a quantitative map for locating these intrinsically preferred twist angles in twisted bilayer system has not been available, posing challenges for the on-demand design of twisted electronics that are intrinsically stable at desired twist angles. Here we reveal a simple mapping between intrinsically preferred twist angles and geometry of the twisted bilayer system, in the form of geometric scaling laws for a wide range of intrinsically preferred twist angles as a function of only geometric parameters of the rotating flake on a supporting layer. We reveal these scaling laws for triangular and hexagonal flakes since they frequently appear in chemical vapor deposition growth. We also present a general method for handling arbitrary flake geometry. Such dimensionless scaling laws possess universality for all kinds of two-dimensional material bilayer systems, providing abundant opportunities for the on-demand design of intrinsic "twistronics". For example, the set of increasing magic-sizes that intrinsically prefers zero-approaching sequence of multiple magic-angles in bilayer graphene system can be revealed.

cond-mat.mes-hall↗

arXiv2016

Strain-induced programmable half-metal and spin-gapless semiconductor in an edge-doped boron nitride nanoribbon

The search for half-metals and spin-gapless semiconductors has attracted extensive attention in material design for spintronics. Existing progress in such a search often requires peculiar atomistic lattice configuration and also lacks active control of the resulting electronic properties. Here we reveal that a boron-nitride nanoribbon with a carbon-doped edge can be made a half-metal or a spin-gapless semiconductor in a programmable fashion. The mechanical strain serves as the on/off switches for functions of half-metal and spin-gapless semiconductor to occur. Our findings shed light on how the edge doping combined with strain engineering can affect electronic properties of two-dimensional materials

cond-mat.mes-hall↗

arXiv2015

Programmable Extreme Pseudomagnetic Fields in Graphene by a Uniaxial Stretch

Many of the properties of graphene are tied to its lattice structure, allowing for tuning of charge carrier dynamics through mechanical strain. The graphene electro-mechanical coupling yields very large pseudomagnetic fields for small strain fields, up to hundreds of Tesla, which offer new scientific opportunities unattainable with ordinary laboratory magnets. Significant challenges exist in investigation of pseudomagnetic fields, limited by the non-planar graphene geometries in existing demonstrations and the lack of a viable approach to controlling the distribution and intensity of the pseudomagnetic field. Here we reveal a facile and effective mechanism to achieve programmable extreme pseudomagnetic fields with uniform distributions in a planar graphene sheet over a large area by a simple uniaxial stretch. We achieve this by patterning the planar graphene geometry and graphene-based hetero-structures with a shape function to engineer a desired strain gradient. Our method is geometrical, opening up new fertile opportunities of strain engineering of electronic properties of 2D materials in general.

cond-mat.mes-hall↗

arXiv2015

Mechanical Control of Graphene on Engineered Pyramidal Strain Arrays

Strain can tune desirable electronic behavior in graphene, but there has been limited progress in controlling strain in graphene devices. In this paper, we study the mechanical response of graphene on substrates patterned with arrays of mesoscale pyramids. Using atomic force microscopy, we demonstrate that the morphology of graphene can be controlled from conformal to suspended depending on the arrangement of pyramids and the aspect ratio of the array. Non-uniform strains in graphene suspended across pyramids are revealed by Raman spectroscopy and supported by atomistic modeling, which also indicates strong pseudomagnetic fields in the graphene. Our results suggest that incorporating mesoscale pyramids in graphene devices is a viable route to achieving strain-engineering of graphene.

cond-mat.mes-hall↗

arXiv2015

Pseudomagnetic Fields in a Locally Strained Graphene Drumhead

Recent experiments reveal that a scanning tunneling microscopy (STM) probe tip can generate a highly localized strain field in a graphene drumhead, which in turn leads to pseudomagnetic fields in the graphene that can spatially confine graphene charge carriers in a way similar to a lithographically defined quantum dot (QD). While these experimental findings are intriguing, their further implementation in nanoelectronic devices hinges upon the knowledge of key underpinning parameters, which still remain elusive. In this paper, we first summarize the experimental measurements of the deformation of graphene membranes due to interactions with the STM probe tip and a back gate electrode. We then carry out systematic coarse grained, (CG), simulations to offer a mechanistic interpretation of STM tip-induced straining of the graphene drumhead. Our findings reveal the effect of (i) the position of the STM probe tip relative to the graphene drumhead center, (ii) the sizes of both the STM probe tip and graphene drumhead, as well as (iii) the applied back-gate voltage, on the induced strain field and corresponding pseudomagnetic field. These results can offer quantitative guidance for future design and implementation of reversible and on-demand formation of graphene QDs in nanoelectronics.

cond-mat.mes-hall↗

arXiv2014

Reversible mechanical and electrical properties of ripped graphene

We examine the mechanical properties of graphene devices stretched on flexible elastomer substrates. Using atomic force microscopy, transport measurements, and mechanics simulations, we show that micro-rips form in the graphene during the initial application of tensile strain; however subsequent applications of the same tensile strain elastically open and close the existing rips. Correspondingly, while the initial tensile strain degrades the devices' transport properties, subsequent strain-relaxation cycles affect transport only moderately, and in a largely reversible fashion, yielding robust electrical transport even after partial mechanical failure.

cond-mat.mes-hall↗

arXiv2014

Effects of surface compliance and relaxation on the frictional properties of lamellar materials

We describe the results of atomic-level stick-slip friction measurements performed on chemically-modified graphite, using atomic force microscopy (AFM). Through detailed molecular dynamics simulations, coarse-grained simulations, and theoretical arguments, we report on complex indentation profiles during AFM scans involving local reversible exfoliation of the top layer of graphene from the underlying graphite sample and its effect on the measured friction force during retraction of the scanning tip. In particular, we report nearly constant lateral stick-slip magnitudes at decreasing loads, which cannot be explained within the standard framework based on continuum mechanics models for the contact area. We explain this anomalous behavior by introducing the effect of local compliance of the topmost graphene layer, which varies when interaction with the AFM tip is enhanced. Such behavior is not observed for non-lamellar materials. We extend our discussion toward the more general understanding of the effects of the top layer relaxation on the friction force under pushing and pulling loads. Our results may provide a more comprehensive understanding of the effectively negative coefficient of friction recently observed on chemically-modified graphite.

cond-mat.mtrl-sci↗