A5039025339- Advanced Memory and Neural Computing
- Transition Metal Oxide Nanomaterials
- Ferroelectric and Negative Capacitance Devices
- Machine Learning and ELM
- ZnO doping and properties
- Gas Sensing Nanomaterials and Sensors
- Photoreceptor and optogenetics research
- Ga2O3 and related materials
- Magnetic and transport properties of perovskites and related materials
- Advancements in Solid Oxide Fuel Cells
- TiO2 Photocatalysis and Solar Cells
- Modular Robots and Swarm Intelligence
- CCD and CMOS Imaging Sensors
- Multiferroics and related materials
- Electronic and Structural Properties of Oxides
- Neural Networks and Reservoir Computing
- Conducting polymers and applications
- Neuroscience and Neural Engineering
- Neural dynamics and brain function
University of California, San Diego
2023-2025
Pohang University of Science and Technology
2018-2021
Government of the Republic of Korea
2018-2021
Abstract CMOS-RRAM integration holds great promise for low energy and high throughput neuromorphic computing. However, most RRAM technologies relying on filamentary switching suffer from variations noise, leading to computational accuracy loss, increased consumption, overhead by expensive program verify schemes. We developed a filament-free, bulk technology address these challenges. systematically engineered trilayer metal-oxide stack investigated the characteristics of with varying...
Transition-metal oxides (TMOs) with brownmillerite (BM) structures possess one-dimensional oxygen vacancy channels (OVCs), which play a key role in realizing high ionic conduction at low temperatures. The controllability of the channel orientation, thus, possesses great potential for practical applications and would provide better visualization diffusion pathways ions TMOs. In this study, orientations OVCs BM-SrFeO2.5 are stabilized along two crystallographic directions epitaxial thin films....
A key aspect of how the brain learns and enables decision-making processes is through synaptic interactions. Electrical transmission communication in a network synapses are modulated by extracellular fields generated ionic chemical gradients. Emulating such spatial interactions synthetic networks can be potential use for neuromorphic learning hardware implementation artificial intelligence. Here, we demonstrate that hydrogen-doped perovskite nickelate devices, electric bias across single...
Filament-free bulk resistive-switching random access memory (RRAM) devices have been proposed to offer multilevel conductance states with less variations and noise forming-free operation for neuromorphic computing applications. Understanding conduction mechanism switching dynamics of filament-free RRAM is crucial optimize device characteristics build large-scale arrays compute in Here, we first analyze by temperature-dependent <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML"...
The Complementary metal oxide semiconductor-Resistive random access memory (RRAM) integration presents significant potential for energy-efficient and high-speed neuromorphic computing. However, conventional filamentary RRAM technologies that rely on switching face challenges such as high variability, noise, reduced computational accuracy, increased energy consumption. To address these limitations, we developed a filament-free, bulk-switching technology. By designing trilayer metal-oxide...
Here, we demonstrate the scalable and continuous production of large coatings with thermochromic VO<sub>2</sub> nanoparticles for “smart” windows.
Abstract Reversible phase transformation of correlated oxides by field‐driven ionic process present opportunity to efficiently transduce between transfer and electrical currents in insertion‐based reconfigurable transistors. However, the switching rate insertion transistors is fundamentally limited slow into lattices oxides. Here, it demonstrated that preformed oxygen vacancies VO 2− δ strongly accelerate proton low gate voltage synaptic As degree deficiency increases transistors, steepness...
Electronic phase modulation based on hydrogen insertion/extraction is kinetically limited by the bulk diffusion or surface exchange reaction, so slow kinetics has been a fundamental challenge to be solved for realizing faster solid-state electrochemical switching devices. Here we accelerate electronic that occurs insertion in VO2 through vertically aligned 2D defects induced symmetry mismatch between epitaxial films and substrates. By using domain-matching growth of monoclinic with lattice...
The control of field-driven ionic redistribution guided by crystal anisotropy increases the retention H<sup>+</sup>s in VO<sub>2</sub> lattices locating H<sup>+</sup> into deep regions from interfaces, and thus strengthens long-term memory artificial synaptic devices.
Abstract Unlike the substitutional dopants, interstitial hydrogen effectively supplies significant amount of carriers in empty narrow d band correlated electronic systems by reversibly adding it into sites. Here, is demonstrated that hydrogenated VO 2 , a heavily insulating phase with 3d configuration, can be thermodynamically stabilized topotactically preserving its lattice framework regardless facet direction epilayers. However, kinetics modulation and response out‐of‐plane expansion are...
Several metastable polymorphs, diverse crystal structures with the same chemical formula, could be accessed through epitaxial interfaces low energy between film and symmetry-matched substrates. Here, we fabricated tungsten oxide (WO3) nanoscaffolds composed of hexagonal WO3 (h-WO3)/monoclinic (m-WO3) modulated proportions these phases by tuning W arrival rate during growth. The nanoscaffold is vertically aligned coherent interphase boundaries h-WO3 m-WO3; this structure leads to persistence...
Abstract Spinodal decomposition, the spontaneous phase separation process of periodic lamellae at nanometer scale, correlated oxide ((Ti, V)O 2 ) systems offers a sophisticated route to achieve new class mesoscale structures in form self-assembled superlattices for possible applications using steep metal–insulator transitions. Here, we tunable self-assembly (Ti, with transitions (Δ T MI < 5 K) by spinodal decomposition accurate control growth parameters without conventional layer-by-layer...
Abstract In-memory computing with emerging non-volatile memory devices (eNVMs) has shown promising results in accelerating matrix-vector multiplications. However, activation function calculations are still being implemented general processors or large and complex neuron peripheral circuits. Here, we present the integration of Ag-based conductive bridge random access (Ag-CBRAM) crossbar arrays Mott rectified linear unit (ReLU) neurons for scalable, energy area-efficient hardware (HW)...
In this work, we present a hardware implementation of spiking neural network (SNN) model using trilayer bulk RRAM crossbar arrays for an autonomous navigation/racing task. We demonstrate multi-level switching in MΩ regime (Al <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> O xmlns:xlink="http://www.w3.org/1999/xlink">3</inf> /TiO xmlns:xlink="http://www.w3.org/1999/xlink">x</inf> ) devices without needing compliance current. also...
Resistive memory-based reconfigurable systems constructed by CMOS-RRAM integration hold great promise for low energy and high throughput neuromorphic computing. However, most RRAM technologies relying on filamentary switching suffer from variations noise leading to computational accuracy loss, increased consumption, overhead expensive program verify schemes. Low ON-state resistance of devices further increases the consumption due high-current read write operations, limits array size parallel...