Searcharxiv⌕ Search

arXiv subjects

Jinwoo Hwang

Publications and source records attributed to Jinwoo Hwang.

At least 19 recordsLinked to original sources

SIRIS: Can LLMs Tell When Simulation Errors Matter in AI-Driven Systems Research

LLM agents increasingly drive systems research and rely on simulators to evaluate candidate designs at scale. Simulators are imperfect by design: their modeling limitations can distort predictions yet leave the recommended design correct, or change which design should be selected. We introduce Siris to test whether LLMs can tell these cases apart, judging from their systems understanding when a limitation invalidates the simulator's recommendation. By perturbing a simulator, Siris constructs workload pairs with the same candidates, limitation, and simulator recommendation. The perturbation changes predictions on both workloads, yet a reference simulator supports the recommendation on only one. Evaluating thirteen frontier LLMs on LLM serving design problems, we find that identifying a limitation does not ensure a correct choice, and even the strongest models sometimes abandon valid recommendations. Self-correction through a review turn helps these models restore valid choices but introduces new errors in several others. These findings motivate evaluating systems understanding through design decisions under imperfect simulators.

cs.AI↗

A Unified Kinematic Representation Enables Reusable Biological Joint Moment Estimation

Objective: Data-driven models that estimate physiological states, particularly biological joint moments, are widely used in exoskeleton control. However, these estimators are often coupled to device-specific sensor configurations, limiting controller transfer and the use of open-source biomechanics datasets. Methods: We proposed joint kinematics as an intermediate representation that decouples hardware-specific sensing from downstream biological joint moment estimation. A joint-moment estimator using joint angles and angular velocities was trained exclusively on open-source biomechanics data and evaluated using kinematics from a hip exoskeleton, knee exoskeleton, and inertial measurement unit (IMU) sensor suite during level-ground, ramp-ascent, and ramp-descent walking. Results: The estimator achieved an root mean square error (RMSE) of 0.17 Nm/kg and coefficient of determination (R2) of 0.79 using hip exoskeleton kinematics, 0.19 Nm/kg and 0.62 using knee exoskeleton kinematics, and 0.15 Nm/kg and 0.85 using kinematics derived from IMUs across bilateral hip, knee, and ankle joints. Conclusion: Joint kinematics enabled an estimator trained only on open-source data to operate across distinct wearable platforms. Significance: This framework may reduce target-device data collection and support transferable biological joint moment estimation for exoskeleton control and wearable biomechanics.

cs.RO↗

Linking Topological Medium Range Order and Density-Wave Coherence to Metallic Glass Ductility

Metallic glasses exhibit pronounced composition-dependent mechanical properties, yet identifying the structural origins of these changes remains difficult. An outstanding issue is how to correlate different forms of medium range order (MRO), defined at different length scales and accessed through different experimental observables, and to understand their connection to properties. We combine synchrotron reduced density function, G(r), with machine-learning-assisted four dimensional scanning transmission electron microscopy (4D-STEM) to examine geometric and topological MRO in five Zr-based metallic glasses. Density wave analysis of G(r) provides a geometric MRO descriptor through the coherence length of atomic density correlations, while 4D-STEM reveals nanoscale topological and chemical MRO motifs, including crystal-like local ordering with distinct rotational symmetries. We show that the G(r)-derived coherence length captures an important baseline tendency for cooperative ductile relaxation, while composition-dependent topological MRO provides an additional structural contribution that helps explain the observed mechanical responses. In particular, increasing Zr content suppresses Cu-rich FCC-like MRO, increases the cooperative coherence volume, and promotes fine Zr-rich HCP-like local order associated with enhanced ductility. These results identify the combined structural influence of geometric and topological MRO as a key basis for composition-dependent mechanical response in metallic glasses.

cond-mat.mtrl-sci↗

Assessing the Transferability of General-Purpose MachineLearning Interatomic Potentials for Heterogeneous Catalysis with HetCat26

Foundation machine learning interatomic potentials (MLIPs) promise near-density functional theory (DFT) accuracy across broad areas of chemistry and materials science. However, their performance in describing systems and processes relevant to heterogeneous catalysis remains underexplored. Here, we introduce HetCat26, a collection of benchmark tests designed to assess pre-trained MLIPs across key aspects of catalytic modeling, including surface energetics, metal-metal oxide interactions, adsorption, and catalytic reaction networks. Evaluating fifteen foundation models, we find that performance on existing general materials benchmarks is only weakly predictive of performance on HetCat26; transferability to heterogeneous catalysis cannot be inferred from these general benchmarks. Across the benchmark tests, current models describe surface energetics and, perhaps surprisingly, reaction barriers well, whereas larger errors are observed for adsorption, and DFT site preferences are often not reproduced. Two models, eSEN-30M-OAM and MACE-MH-1-OMAT, nevertheless achieve high accuracy across the properties evaluated. Beyond model benchmarking, HetCat26 highlights the importance of training data consistency: PBE and PBE+U calculations should not be mixed within a training set. Overall, we identify challenges limiting the transferability of current foundation MLIPs to heterogeneous catalysis and, more broadly, to chemical reactions at interfaces, providing guidance for the development of the next generation of foundation models.

cond-mat.mtrl-sci↗

Nanoscale Sr$_2$IrO$_4$ Freestanding Thin-Films for Flexible Electronics

We report the structural and optical properties of nanoscale Sr2IrO4 freestanding thin-films fabricated using a water-soluble Sr3Al2O6 layer. The coherent lattice structure, phonon modes, two-magnon Raman scattering, and optical absorption spectra of the Sr2IrO4 nanomembrane are analogous to those of the layered iridate epitaxial thin-films and single crystals. Remarkably, the formation of 3-unit-cell-thick SrIrO3 and interfacial composite layers alleviates antiphase boundaries at the Sr2IrO4/Sr3Al2O6 interface, resulting in structurally-robust nanomembranes. Our experimental results show that this freestanding thin-film approach of layered oxides can provide techniques for tuning or realizing unprecedented states beyond conventional thin-film methods, suggesting a pathway in achieving flexible layered-oxide electronics.

cond-mat.mtrl-sci↗

Laser-Ablated Au Electrodes with Preferred Orientation and Flat Interface on Oxides

Gold (Au) thin-film electrodes deposited on oxides by pulsed laser deposition (PLD) exhibit distinctive structural properties compared with thermally evaporated Au films. While thermally evaporated Au forms polycrystalline films with randomly oriented grains, PLD-grown Au films show a strong preferred [111] orientation perpendicular to the substrate surface, likely due to the energetic nature of the laser-ablated Au plume. High-resolution transmission electron microscopy reveals atomically flat interfaces between the PLD-grown Au electrodes and epitaxial oxide thin films, separated by nanometer-scale gaps comparable to interlayer spacings in van der Waals materials. The highly ordered microstructure and flat interfacial morphology result in enhanced adhesion of the Au films to oxide surfaces. These results demonstrate that PLD can produce both high-quality Au electrodes and oxide thin films within a single deposition platform, offering improved interfacial control and potential benefits for a wide range of oxide electronic and functional devices.

cond-mat.mtrl-sci↗

Coordination-Sensitive Nanoscale Analysis of Defect-Driven Phase Transformation in Si-Doped (AlXGa1-X)2O3

Defect-driven phase instability critically influences the structural reliability of ultrawide bandgap oxides, yet direct nanoscale metrics linking local chemistry to structural transformation remain limited. Here, we introduce a coordination-sensitive atom probe tomography framework that quantitatively resolves reductions in local cation coordination and links them directly to defect-driven phase transformation. Using Si-doped beta-(AlxGa1-x)2O3 heterostructures with controlled Al composition (6-17%) and doping levels (10^17-10^20 cm^-3), we show that gamma-phase inclusions emerge exclusively under the combined conditions of elevated Al content and heavy Si doping. Two-dimensional compositional mapping reveals pronounced lateral Al/Ga inhomogeneity in these regions, while nearest-neighbor and radial distribution analyses quantitatively resolve a significant reduction in first-shell Ga coordination, consistent with local cation deficiency. Correlative scanning transmission electron microscopy confirms that these coordination-depleted regions coincide spatially with gamma-phase inclusions. Density functional theory further supports this mechanism, demonstrating that Al incorporation reduces monoclinic lattice stability and, in conjunction with donor-induced vacancy formation, facilitates vacancy-mediated cation rearrangement and coordination collapse. Together, these results establish coordination loss as a measurable nanoscale signature directly linked to defect-driven phase instability. This framework provides a generalizable approach for probing defect-driven phase instability in doped and alloyed ultrawide bandgap semiconductors.

cond-mat.mtrl-sci↗

From pore collapse to crystal growth: ultrafast laser-induced stishovite formation in nanoporous silica

The crystallization of amorphous solids under ultrafast laser irradiation represents a paradigm of non-equilibrium phase transitions, where the interplay between electromagnetic energy localization and atomic-scale dynamics remains largely uncharted. By using a multiscale framework that couples finite-difference time-domain simulations of nonlinear femtosecond laser pulse propagation with molecular dynamics of the atomic response, we demonstrate that field enhancement around the pores of nanoporous amorphous silica confines laser energy and drives rapid pore collapse. In nanoporous silica, the enhanced local electromagnetic field leads to stronger energy absorption compared with smaller-pore and homogeneous systems. This heterogeneous energy localization provides preferential nucleation sites within the dense glass network, leading to ultrafast formation of stishovite on a sub-nanosecond timescale, faster than in homogeneous silica. This accelerated crystallization can outpace pressure relaxation making the transition to a high-pressure phase possible. These results are confirmed by experimental observations of femtosecond-laser-induced crystallization in confined geometries, and show that electromagnetic hotspots in nanoporous glass structures can be tailored to control solid-state transformations.

cond-mat.mtrl-sci↗

Microwave Performance of all MOCVD-grown AlScN/GaN MIS-HEMTs on Semi-Insulating GaN Substrates

We report on the design, fabrication, and characterization of all MOCVD-grown long-gate AlScN/GaN metal-insulator-semiconductor high electron mobility transistors (MIS-HEMTs) on semi-insulating GaN substrates. Devices with a gate length of $1~μ$m and gate-drain spacing of $0.9~μ$ m exhibit a maximum drain current density of 1 A/mm, an on/off current ratio of $2\times 10^5$, and a three-terminal breakdown voltage of 63 V. The device has near-ideal subthreshold characteristics with a subthreshold swing of 63 mV/dec and a current dispersion as low as 7.8$\%$ at 10 V due to the excellent interfacial quality with a trap density ($D_\mathrm{it}$) of $2.11\times{10}^{11}~\mathrm{cm}^{-2}eV^{-1}$ and the semi-insulating GaN substrate with a low threading dislocation density. Small-signal RF measurements reveal an $f_\mathrm{T}/f_\mathrm{max}$ of 25.8/51.1 GHz, while large-signal load-pull characterization at 10 GHz demonstrates an output power density of 4.04 W/mm with a power-added efficiency of 22.7$\%$. In addition, a minimum noise figure below 2.5 dB was measured over a wide drain current range from 100 mA/mm to 700 mA/mm below 6 GHz. These results extend previous demonstrations of short-gate MOCVD-grown AlScN/GaN HEMTs to the long-gate, high-voltage regime, confirming the robustness of this material system for both high-frequency and high-power device applications with favorable microwave noise performance.

physics.app-ph↗

Stabilizing Magnetic Bubble Domains in Epitaxial 2D Magnet/Topological Insulator Heterostructures through Interfacial Interactions

Epitaxial heterostructures of two-dimensional van der Waals magnets and topological insulators offer a powerful platform for probing interfacial spin interactions that govern magnetic textures in low-dimensional quantum systems, while simultaneously enabling highly efficient, atomically thin spin-orbit-torque memory and computing architectures. Despite this promise, the fundamental role of these interfacial interactions in determining magnetic domain-phase stability remain largely uncharted. Here, we perform scanning transmission X-ray microscopy to image nanoscale magnetic textures in epitaxial Fe3GeTe2 Bi2Te3 heterostructures, enabled by a thermal-release-tape dry transfer process onto X-ray transparent silicon-rich nitride membranes. Under zero-field-cooled conditions, we observe robust bubble domain phases from 75 to 165 K, and across different number of folds of the multilayer Fe3GeTe2 Bi2Te3 heterostructures. This is in stark contrast with exfoliated single-crystal Fe3GeTe2 flakes, where ZFC stripe domains are observed for flakes thicker than 20 nm and no domains have been reported for thin flakes less than 15 nm. First-principles calculations and micromagnetic simulations reveal that interfacial coupling to Bi2Te3 modifies the magnetic anisotropy and introduces interfacial Dzyaloshinskii-Moriya interaction, shifting the magnetic phase space towards bubble-domain stabilization without field-cooling. Together, our results offer a new strategy for phase-selective control of magnetic domains through interfacial engineering.

cond-mat.mtrl-sci↗

Mimicking the earth core conditions with ultrafast laser materials interaction

Ultrafast lasers create extreme, non-equilibrium thermodynamic conditions that can transiently reach pressures and temperatures comparable to interior core of the earth. Here we show that femtosecond excitation of amorphous silica-hafnia multilayer dielectrics drives the formation of high-pressure crystalline phases of silica including stishovite, seifertite, and the pyrite-type high density structure, within confined subsurface regions.Using TEM, SAED, and 4D-STEM, we directly map nanoscale phase evolution and identify crystalline motifs embedded inside laser generated blisters.Complementary molecular dynamics simualtions reveal the thermodynamic pathway underlying these transformations, where rapid electronic pressure initiates densification and octahedral coordination, followed by temperature driven crystallization and displacive transitions during ultrafast quenching. The resulting polymorphs reflects a dual-stage pathway inaccessible under equilibrium processing. Our results establish femtosecond laser excitation as a viable route to synthesize and stabilize ultrahigh-density high pressure silica phases under ambient conditions, without a diamond anvil cell, with implications for laser-damage mechanisms, high-energy-density materials, and planetary physics.

cond-mat.mtrl-sci↗

Understanding the Performance Behaviors of End-to-End Protein Design Pipelines on GPUs

Recent computational advances enable protein design pipelines to run end-to-end on GPUs, yet their heterogeneous computational behaviors remain undercharacterized at the system level. We implement and profile a representative pipeline at both component and full-pipeline granularities across varying inputs and hyperparameters. Our characterization identifies generally low GPU utilization and high sensitivity to sequence length and sampling strategies. We outline future research directions based on these insights and release an open-source pipeline and profiling scripts to facilitate further studies.

cs.ET↗

Neo: Real-Time On-Device 3D Gaussian Splatting with Reuse-and-Update Sorting Acceleration

3D Gaussian Splatting (3DGS) rendering in real-time on resource-constrained devices is essential for delivering immersive augmented and virtual reality (AR/VR) experiences. However, existing solutions struggle to achieve high frame rates, especially for high-resolution rendering. Our analysis identifies the sorting stage in the 3DGS rendering pipeline as the major bottleneck due to its high memory bandwidth demand. This paper presents Neo, which introduces a reuse-and-update sorting algorithm that exploits temporal redundancy in Gaussian ordering across consecutive frames, and devises a hardware accelerator optimized for this algorithm. By efficiently tracking and updating Gaussian depth ordering instead of re-sorting from scratch, Neo significantly reduces redundant computations and memory bandwidth pressure. Experimental results show that Neo achieves up to 10.0x and 5.6x higher throughput than state-of-the-art edge GPU and ASIC solution, respectively, while reducing DRAM traffic by 94.5% and 81.3%. These improvements make high-quality and low-latency on-device 3D rendering more practical.

cs.AR↗

Band Alignment Tuning from Charge Transfer in Epitaxial SrIrO$_3$/SrCoO$_3$ Superlattices

Understanding charge transfer at oxide interfaces is crucial for designing materials with emergent electronic and magnetic properties, especially in systems where strong electron correlations and spin-orbit coupling coexist. SrIrO$_3$/SrCoO$_3$ (SIO/SCO) superlattices offer a unique platform to explore these effects due to their contrasting electronic structures and magnetic behaviors. Building on past theory based on continuity of O 2p band alignment, we employ density functional theory (DFT) to model electron transfer from Ir to Co across the SIO/SCO interface. To characterize these effects, we synthesized epitaxial SIO/SCO superlattices via molecular beam epitaxy. Structural and transport measurements confirmed high crystallinity, metallic behavior, and suppression of Kondo scattering that has been reported in uniform SIO films. Further characterization via X-ray absorption spectroscopy (XAS) revealed orbital anisotropy and valence changes consistent with interfacial charge transfer. Co K- and L$_{2,3}$-edge and Ir L$_2$-edge spectra verified electron donation from Ir to Co, stabilizing the perovskite SCO phase and tuning the electronic structure of SIO via hole-doping. O K-edge XAS showed band alignment shifts in the SIO layer consistent with DFT predictions. Our work here provides a pathway for engineering oxide heterostructures with tailored magnetic and electronic properties.

cond-mat.mtrl-sci↗

Broad-Range Tuning of Ferroelectric Switching of LaxBi1-xFeO3 Epitaxial Films via Digital Doping using Off-Axis Co-Sputtering

To investigate the scope of ferroelectric behavior in La-substituted BiFeO3 films, LaxBi1-xFeO3 epitaxial films were synthesized using off-axis co-sputtering on SrTiO3(001) and DyScO3(110) substrates with a SrRuO3 bottom electrode layer. A digital-doping deposition method was used to enable precise control and continuous tuning of La concentration in high-quality LaxBi1-xFeO3 films across a wide range of x = 0.05-0.60, which was systematically investigated using piezoresponse force microscopy. Robust and reversible out-of-plane ferroelectric switching has been observed up to x = 0.35, while films with x $\geq$ 0.37 exhibit no measurable ferroelectric behavior, indicating a sharp ferroelectric-to-paraelectric phase transition between x = 0.35 and 0.37. This represents the highest reported La concentration in LaxBi1-xFeO3 films that retains ferroelectric ordering, highlighting opportunities to engineer ferroelectric and multiferroic properties in complex oxide heterostructures.

cond-mat.mtrl-sci↗

Pimba: A Processing-in-Memory Acceleration for Post-Transformer Large Language Model Serving

Transformers are the driving force behind today's Large Language Models (LLMs), serving as the foundation for their performance and versatility. Yet, their compute and memory costs grow with sequence length, posing scalability challenges for long-context inferencing. In response, the algorithm community is exploring alternative architectures, such as state space models (SSMs), linear attention, and recurrent neural networks (RNNs), which we refer to as post-transformers. This shift presents a key challenge: building a serving system that efficiently supports both transformer and post-transformer LLMs within a unified framework. To address this challenge, we analyze the performance characteristics of transformer and post-transformer LLMs. Despite their algorithmic differences, both are fundamentally limited by memory bandwidth under batched inference due to attention in transformers and state updates in post-transformers. Further analyses suggest two additional insights: (1) state update operations, unlike attention, incur high hardware cost, making per-bank PIM acceleration inefficient, and (2) different low-precision arithmetic methods offer varying accuracy-area tradeoffs, while we identify Microsoft's MX as the Pareto-optimal choice. Building on these insights, we design Pimba as an array of State-update Processing Units (SPUs), each shared between two banks to enable interleaved access to PIM. Each SPU includes a State-update Processing Engine (SPE) that comprises element-wise multipliers and adders using MX-based quantized arithmetic, enabling efficient execution of state update and attention operations. Our evaluation shows that, compared to LLM-optimized GPU and GPU+PIM systems, Pimba achieves up to 4.1x and 2.1x higher token generation throughput, respectively.

cs.AR↗

Déjà Vu: Efficient Video-Language Query Engine with Learning-based Inter-Frame Computation Reuse

Recently, Video-Language Models (VideoLMs) have demonstrated remarkable capabilities, offering significant potential for flexible and powerful video query systems. These models typically rely on Vision Transformers (ViTs), which process video frames individually to extract visual embeddings. However, generating embeddings for large-scale videos requires ViT inferencing across numerous frames, posing a major hurdle to real-world deployment and necessitating solutions for integration into scalable video data management systems. This paper introduces Déjà Vu, a video-language query engine that accelerates ViT-based VideoLMs by reusing computations across consecutive frames. At its core is ReuseViT, a modified ViT model specifically designed for VideoLM tasks, which learns to detect inter-frame reuse opportunities, striking an effective balance between accuracy and reuse. Although ReuseViT significantly reduces computation, these savings do not directly translate into performance gains on GPUs. To overcome this, Déjà Vu integrates memory-compute joint compaction techniques that convert the FLOP savings into tangible performance gains. Evaluations on three VideoLM tasks show that Déjà Vu accelerates embedding generation by up to a 2.64x within a 2% error bound, dramatically enhancing the practicality of VideoLMs for large-scale video analytics.

cs.DC↗

Thermal Annealing and Radiation Effects on Structural and Electrical Properties of NbN/GaN Superconductor/Semiconductor Junction

In the rapidly evolving field of quantum computing, niobium nitride (NbN) superconductors have emerged as integral components due to their unique structural properties, including a high superconducting transition temperature (Tc), exceptional electrical conductivity, and compatibility with advanced device architectures. This study investigates the impact of high-temperature annealing and high-dose gamma irradiation on the structural and superconducting properties of NbN films grown on GaN via reactive DC magnetron sputtering. The as-deposited cubic δ-NbN (111) films exhibited a high-intensity XRD peak, high Tc of 12.82K, and an atomically flat surface. Annealing at 500 and 950 °C for varying durations revealed notable structural and surface changes. High-resolution STEM indicated improved local ordering, while AFM showed reduced surface roughness after annealing. XPS revealed a gradual increase in the Nb/N ratio with higher annealing temperatures and durations. High-resolution XRD and STEM analyses showed lattice constant modifications in δ-NbN films, attributed to residual stress changes following annealing. Additionally, XRD phi-scans revealed sixfold symmetry in NbN films due to rotational domains relative to GaN. While Tc remained stable after annealing at 500 °C, increasing the annealing temperature to 950 °C degraded Tc to ~8K and reduced the residual resistivity ratio from 0.85 in as-deposited films to 0.29 after 30 minutes. The effects of gamma radiation (5 Mrad (Si)) were also studied, demonstrating minimal changes to crystallinity and superconducting performance, indicating excellent radiation resilience. These findings highlight the potential of NbN superconductors for integration into advanced quantum devices and their suitability for applications in radiation-intensive environments such as space, satellites, and nuclear power plants.

cond-mat.supr-con↗