arXiv subjects

Donghoon Kim

Publications and source records attributed to Donghoon Kim.

53 records · Page 3Linked to original sources

arXiv2024

Universal Spin Screening Clouds in Local Moment Phases

When a local impurity spin interacts with conduction electrons whose density of states (DOS) has a (pseudo)gap or diverges at the Fermi energy, a local moment (LM) phase can be favored over a Kondo phase. Theoretically studying quantum entanglement between the impurity and conduction electrons, we demonstrate that conduction electrons form an ''LM spin cloud'' in general LM phases, which corresponds to, but has fundamental difference from, the Kondo cloud screening the impurity spin in the Kondo phase. The LM cloud algebraically decays over the distance from the impurity when the DOS has a pseudogap or divergence, and exponentially when it has a hard gap. We find an ''LM cloud length'', a single length scale characterizing a universal form of the LM cloud. The findings are supported by both of analytic theories and numerical computations.

cond-mat.mes-hall↗

arXiv2024

Preserving Pre-trained Representation Space: On Effectiveness of Prefix-tuning for Large Multi-modal Models

Recently, we have observed that Large Multi-modal Models (LMMs) are revolutionizing the way machines interact with the world, unlocking new possibilities across various multi-modal applications. To adapt LMMs for downstream tasks, parameter-efficient fine-tuning (PEFT) which only trains additional prefix tokens or modules, has gained popularity. Nevertheless, there has been little analysis of how PEFT works in LMMs. In this paper, we delve into the strengths and weaknesses of each tuning strategy, shifting the focus from the efficiency typically associated with these approaches. We first discover that model parameter tuning methods such as LoRA and Adapters distort the feature representation space learned during pre-training and limit the full utilization of pre-trained knowledge. We also demonstrate that prefix-tuning excels at preserving the representation space, despite its lower performance on downstream tasks. These findings suggest a simple two-step PEFT strategy called Prefix-Tuned PEFT (PT-PEFT), which successively performs prefix-tuning and then PEFT (i.e., Adapter, LoRA), combines the benefits of both. Experimental results show that PT-PEFT not only improves performance in image captioning and visual question answering compared to vanilla PEFT methods but also helps preserve the representation space of the four pre-trained models.

cs.CL↗

arXiv2024

Offline Policy Learning via Skill-step Abstraction for Long-horizon Goal-Conditioned Tasks

Goal-conditioned (GC) policy learning often faces a challenge arising from the sparsity of rewards, when confronting long-horizon goals. To address the challenge, we explore skill-based GC policy learning in offline settings, where skills are acquired from existing data and long-horizon goals are decomposed into sequences of near-term goals that align with these skills. Specifically, we present an `offline GC policy learning via skill-step abstraction' framework (GLvSA) tailored for tackling long-horizon GC tasks affected by goal distribution shifts. In the framework, a GC policy is progressively learned offline in conjunction with the incremental modeling of skill-step abstractions on the data. We also devise a GC policy hierarchy that not only accelerates GC policy learning within the framework but also allows for parameter-efficient fine-tuning of the policy. Through experiments with the maze and Franka kitchen environments, we demonstrate the superiority and efficiency of our GLvSA framework in adapting GC policies to a wide range of long-horizon goals. The framework achieves competitive zero-shot and few-shot adaptation performance, outperforming existing GC policy learning and skill-based methods.

cs.LG↗

arXiv2024

A Survey on Exploratory Spatiotemporal Visual Analytics Approaches for Climate Science

Climate science produces a wealth of complex, high-dimensional, multivariate data from observations and numerical models. These data are critical for understanding climate changes and their socioeconomic impacts. Climate scientists are continuously evaluating output from numerical models against observations. This model evaluation process provides useful guidance to improve the numerical models and subsequent climate projections. Exploratory visual analytics systems possess the potential to significantly reduce the burden on scientists for traditional spatiotemporal analyses. In addition, technology and infrastructure advancements are further facilitating broader access to climate data. Climate scientists today can access climate data in distributed analytic environments and render exploratory visualizations for analyses. Efforts are ongoing to optimize the computational efficiency of spatiotemporal analyses to enable efficient exploration of massive data. These advances present further opportunities for the visualization community to innovate over the full landscape of challenges and requirements raised by scientists. In this report, we provide a comprehensive review of the challenges, requirements, and current approaches for exploratory spatiotemporal visual analytics solutions for climate data. We categorize the visual analytic techniques, systems, and tools presented in the relevant literature based on task requirements, data sources, statistical techniques, interaction methods, visualization techniques, performance evaluation methods, and application domains. Moreover, our analytic review identifies trends, limitations, and key challenges in visual analysis. This report will advance future research activities in climate visualizations and enables the end-users of climate data to identify effective climate change mitigation strategies.

cs.HC↗

arXiv2024

Role of Sensing and Computer Vision in 6G Wireless Communications

Recently, we are witnessing the remarkable progress and widespread adoption of sensing technologies in autonomous driving, robotics, and metaverse. Considering the rapid advancement of computer vision (CV) technology to analyze the sensing information, we anticipate a proliferation of wireless applications exploiting the sensing and CV technologies in 6G. In this article, we provide a holistic overview of the sensing and CV-aided wireless communications (SVWC) framework for 6G. By analyzing the high-resolution sensing information through the powerful CV techniques, SVWC can quickly and accurately understand the wireless environments and then perform the wireless tasks. To demonstrate the efficacy of SVWC, we design the whole process of SVWC including the sensing dataset collection, DL model training, and execution of realistic wireless tasks. From the numerical evaluations on 6G communication scenarios, we show that SVWC achieves considerable performance gains over the conventional 5G systems in terms of positioning accuracy, data rate, and access latency.

cs.CV↗

arXiv2023

Exploring the Effects of VR Activities on Stress Relief: A Comparison of Sitting-in-Silence, VR Meditation, and VR Smash Room

In our lives, we encounter various stressors that may cause negative mental and bodily reactions to make us feel frustrated, angry, or irritated. Effective methods to manage or reduce stress and anxiety are essential for a healthy life, and several stress-management approaches are found to be useful for stress relief, such as meditation, taking a rest, walking around nature, or even breaking things in a smash room. Previous research has revealed that certain experiences in virtual reality (VR) are effective for reducing stress as traditional real-world methods. However, it is still unclear how the stress relief effects are associated with other factors like individual user profile in terms of different treatment activities. In this paper, we report our findings from a formal user study that investigates the effects of two virtual activities: (1) VR Meditation and (2) VR Smash Room experience, compared with a traditional Sitting-in-Silence method. Our results show that VR Meditation has a better stress relief effect compared to VR Smash Room and Sitting-in-Silence, and the effects of the treatments are correlated with the participants' personalities. We discuss the findings and implications addressing potential benefits/impacts of different stress-relief activities in VR.

cs.HC↗

arXiv2023

Investigating Psychological Ownership in a Shared AR Space: Effects of Human and Object Reality and Object Controllability

Augmented reality (AR) provides users with a unique social space where virtual objects are natural parts of the real world. The users can interact with 3D virtual objects and virtual humans projected onto the physical environment. This work examines perceived ownership based on the reality of objects and partners, as well as object controllability in a shared AR setting. Our formal user study with 28 participants shows a sense of possession, control, separation, and partner presence affect their perceived ownership of a shared object. Finally, we discuss the findings and present a conclusion.

cs.HC↗

arXiv2023

Hierarchical entanglement shells of multichannel Kondo clouds

Impurities or boundaries often impose nontrivial boundary conditions on a gapless bulk, resulting in distinct boundary universality classes for a given bulk, phase transitions, and non-Fermi liquids in diverse systems. The underlying boundary states however remain largely unexplored. This is related with a fundamental issue how a Kondo cloud spatially forms to screen a magnetic impurity in a metal. Here we predict the quantum-coherent spatial and energy structure of multichannel Kondo clouds, representative boundary states involving competing non-Fermi liquids, by studying quantum entanglement between the impurity and the channels. Entanglement shells of distinct non-Fermi liquids coexist in the structure, depending on the channels. As temperature increases, the shells become suppressed one by one from the outside, and the remaining outermost shell determines the thermal phase of each channel. Detection of the entanglement shells is experimentally feasible. Our findings suggest a guide to studying other boundary states and boundary-bulk entanglement.

cond-mat.str-el↗

arXiv2022

Strain sensitive flexible magnetoelectric ceramic nanocomposites

Advanced flexible electronics and soft robotics require the development and implementation of flexible functional materials. Magnetoelectric (ME) oxide materials can convert magnetic input into electric output and vice versa, making them excellent candidates for advanced sensing, actuating, data storage, and communication. However, their application has been limited to rigid devices due to their brittle nature. Here, we report flexible ME oxide composite (BaTiO3/CoFe2O4) thin film nanostructures that can be transferred onto a stretchable substrate such as polydimethylsiloxane (PDMS). In contrast to rigid bulk counterparts, these ceramic nanostructures display a flexible behavior and exhibit reversibly tunable ME coupling via mechanical stretching. We believe our study can open up new avenues for integrating ceramic ME composites into flexible electronics and soft robotic devices.

physics.app-ph↗

arXiv2022

PORTAL: Portal Widget for Remote Target Acquisition and Control in Immersive Virtual Environments

This paper introduces PORTAL (POrtal widget for Remote Target Acquisition and controL) that allows the user to interact with out-of-reach objects in a virtual environment. We describe the PORTAL interaction technique for placing a portal widget and interacting with target objects through the portal. We conduct two formal user studies to evaluate PORTAL for selection and manipulation functionalities. The results show PORTAL supports participants to interact with remote objects successfully and precisely. Following that, we discuss its potential and limitations, and future works.

cs.HC↗

arXiv2022

Electrostatic Catalysis of a Click Reaction in a Microfluidic Cell

Electric fields have been highlighted as a smart reagent in nature's enzymatic machinery, as they can directly trigger or accelerate redox and/or non-redox chemical processes with stereo- and regio-specificity. In natural catalysis, controlled mass transport of chemical species in confined spaces is also key in facilitating the transport of reactants into the active reaction site. Despite the opportunities the above offers in developing strategies for a new, clean electrostatic catalysis exploiting oriented electric fields, research in this area has been mostly limited to theoretical and experimental studies at the level of single molecules or small molecular ensembles, where both the control over mass transport and scalability cannot be tested. Here, we quantify the electrostatic catalysis of a prototypical Huisgen cycloaddition in a large-area electrode surface and directly compare its performance to the traditional Cu(I)-catalyzed method of the same reaction. Mass diffusion control is achieved in a custom-built microfluidic cell, which enhances reagent transport towards the electrified reactive interface while avoiding both turbulent flow conditions and poor control of the convective mass transport. This unprecedented electrostatic continuous-flow microfluidic reactor is an example of an electric-field driven platform where clean large-scale electrostatic catalytic processes can be efficiently implemented and regulated.

physics.chem-ph↗

arXiv2022

Giant shape-memory effect in twisted ferroic nanocomposites

The shape recovery ability of shape-memory alloys vanishes below a critical size (~50nm), which prevents their practical applications at the nanoscale. In contrast, ferroic materials, even when scaled down to dimensions of a few nanometers, exhibit actuation strain through domain switching, though the generated strain is modest (~1%). Here, we develop free-standing twisted architectures of nanoscale ferroic oxides showing shape-memory effect with a giant recoverable strain (>10%). The twisted geometrical design amplifies the strain generated during ferroelectric domain switching, which cannot be achieved in bulk ceramics or substrate-bonded thin films. The twisted ferroic nanocomposites allow us to overcome the size limitations in traditional shape-memory alloys and opens new avenues in engineering large-stroke shape-memory materials for small-scale actuating devices such as nanorobots and artificial muscle fibrils.

cond-mat.mtrl-sci↗

arXiv2022

Magnetoelectric effect in hydrogen harvesting: magnetic field as a trigger of catalytic reactions

Magnetic fields have been regarded as an additional stimulus for electro- and photocatalytic reactions, but not as a direct trigger for catalytic processes. Multiferroic/magnetoelectric materials, whose electrical polarization and surface charges can be magnetically altered, are especially suitable for triggering and control of catalytic reactions solely with magnetic fields. Here, we demonstrate that magnetic fields can be employed as an independent input energy source for hydrogen harvesting by means of the magnetoelectric effect. Composite multiferroic CoFe2O4-BiFeO3 core-shell nanoparticles act as catalysts for the hydrogen evolution reaction (HER) that is triggered when an alternating magnetic field is applied to an aqueous dispersion of the magnetoelectric nanocatalysts. Based on density functional calculations, we propose that the hydrogen evolution is driven by changes in the ferroelectric polarization direction of BiFeO3 caused by the magnetoelectric coupling. We believe our findings will open new avenues towards magnetically induced renewable energy harvesting.

cond-mat.mtrl-sci↗

arXiv2021

Universal Thermal Entanglement of Multichannel Kondo Effects

Quantum entanglement between an impurity and its environment is expected to be central in quantum impurity problems. We develop a method to compute the entanglement in spin-1/2 impurity problems, based on the entanglement negativity and the boundary conformal field theory (BCFT). Using the method, we study the thermal decay of the entanglement in the multichannel Kondo effects. At zero temperature, the entanglement has the maximal value independent of the number of the screening channels. At low temperature, the entanglement exhibits a power-law thermal decay. The power-law exponent equals two times of the scaling dimension of the BCFT boundary operator describing the impurity spin, and it is attributed to the energy-dependent scaling behavior of the entanglement in energy eigenstates. These agree with numerical renormalization group results, unveiling quantum coherence inside the Kondo screening length.

cond-mat.str-el↗

arXiv2018

Graphene quantum dots prevent alpha-synucleinopathy in Parkinson's disease

While the emerging evidence indicates that the pathogenesis of Parkinson's disease (PD) is strongly correlated to the accumulation of alpha-synuclein (α-syn) aggregates, there has been no clinical success in anti-aggregation agents for the disease to date. Here we show that graphene quantum dots (GQDs) exhibit anti-amyloid activity via direct interaction with α-syn. Employing biophysical, biochemical, and cell-based assays as well as molecular dynamics (MD) simulation, we find that GQDs have notable potency in not only inhibiting fibrillization of α-syn but also disaggregating mature fibrils in a time-dependent manner. Remarkably, GQDs rescue neuronal death and synaptic loss, reduce Lewy body (LB)/Lewy neurite (LN) formation, ameliorate mitochondrial dysfunctions, and prevent neuron-to-neuron transmission of α-syn pathology induced by α-syn preformed fibrils (PFFs) in neurons. In addition, in vivo administration of GQDs protects against α-syn PFFs-induced loss of dopamine neurons, LB/LN pathology, and behavioural deficits through the penetration of the blood-brain barrier (BBB). The finding that GQDs function as an anti-aggregation agent provides a promising novel therapeutic target for the treatment of PD and related α-synucleinopathies.

physics.med-ph↗

arXiv2017

Electron-Hole Separation in Ferroelectric Oxides for Efficient Photovoltaic Responses

Despite their potential to exceed the theoretical Shockley-Queisser limit, ferroelectric photovoltaics (FPVs) have performed inefficiently due to their extremely low photocurrents. Incorporating Bi2FeCrO6 (BFCO) as the light absorber in FPVs has recently led to impressively high and record photocurrents [Nechache et al. Nature Photon. 2015, 9, 61], reviving the FPV field. However, our understanding of this remarkable phenomenon is far from satisfactory. Here, we use first-principles calculations to determine that such excellent performance mainly lies in the efficient separation of electron-hole (e-h) pairs. We show that photoexcited electrons and holes in BFCO are spatially separated on the Fe and Cr sites, respectively. This separation is much more pronounced in disordered BFCO phases, which show exceptional PV responses. We further set out to design a strategy for next-generation FPVs, not limited to BFCO, by exploring 44 additional Bi-based double-perovskite oxides. We suggest 9 novel active-layer materials that can offer strong e-h separations and a desired band gap energy for application in FPVs. Our work indicates that charge separation is the most important issue to be addressed for FPVs to compete with conventional devices.

cond-mat.mtrl-sci↗

arXiv2017

Switchable Ferroelectric Photovoltaic Effects in Epitaxial Thin Films of h-RFeO3 having Narrow Optical Band Gaps

Ferroelectric photovoltaics (FPVs) have drawn much attention owing to their high stability, environmental safety, anomalously high photovoltages, coupled with reversibly switchable photovoltaic responses. However, FPVs suffer from extremely low photocurrents, which is primarily due to their wide band gaps. Here, we present a new class of FPVs by demonstrating switchable ferroelectric photovoltaic effects using hexagonal ferrite (h-RFeO3) thin films having narrow band gaps of ~1.2 eV, where R denotes rare-earth ions. FPVs with narrow band gaps suggests their potential applicability as photovoltaic and optoelectronic devices. The h-RFeO3 films further exhibit reasonably large ferroelectric polarizations, which possibly reduces a rapid recombination rate of the photo-generated electron-hole pairs. The power conversion efficiency (PCE) of h-RFeO3 thin-film devices is sensitive on the magnitude of polarization. In the case of h-TmFeO3 (h-TFO) thin film, the measured PCE is twice as large as that of the BiFeO3 thin film, a prototypic FPV. We have further shown that the switchable photovoltaic effect dominates over the unswitchable internal field effect arising from the net built-in potential. This work thus demonstrates a new class of FPVs towards high-efficiency solar cell and optoelectronic applications.

cond-mat.mtrl-sci↗