Research Projects

Neuromorphic computing

‐ 1R-based ANN arrays for online training/inference

– Artificial synapses based on single-crystalline ReRAM

Remote epitaxy, Graphene-based layer transfer


– Freestanding InGaAs-based IR Photodectors

– Freestanding Multifunctional complex oxides for magnetoelectric coupling

– SiC/III-N power electronics

Heterointegration, Flexible electronics

– Skin strain sensor arrays

– Flexible/transparent microLEDs

– Self-powered IoT system

Two-dimensional materials

– Monolayer-by-monolayer splitting of wafer‐scale 2D materials

– Wafer-scale single-crystalline 2D materials

– Wafer-scale 2D heterostructures

Renewable energy, Energy storage

– Wafer recycling technique for GaAs solar cells based on remote epitaxy

– High efficiency III-V multi-junction solar cells based on remote epitaxy

– Single-crystalline all solid-state battery

2D Material-Based Layer Transfer

A 2D Material-based layer transfer (2DLT) technology offers infinitive growths & transfers of high quality single-crystalline semiconductor films on 2D materials. We develop a method to perform remote epitaxy of defect-free single-crystalline films on epitaxial graphene in general material system and we study mechanics for repeatable & precise exfoliation of epilayers on graphene. Our focus is to fabricate high performance electronic/photonic/photovoltaics devices with low manufacturing cost based on this 2DLT technique.

Brain-Inspired Neuromorphic Computing

Shortcomings in supercomputer originate in their von Neumann architecture, which involves a CPU, an input/output unit, and a storage (memory) unit. Because communication between these CPU and memory requires data buses, this system inherently cannot accommodate the speed necessary for real-time AI computation. In addition, implementing AI with von Neumann computing would involve training hundreds of thousands of CPUs, GPUs, and memory devices. Brain-inspired neuromorphic computing has recently emerged as a plausible alternative computing method for AI because it enables ultrafast real-time data processing in small footprint. The parallelism property of the ReRAM’s dot-product calculation algorithm enables its fast computing speed. More importantly, ReRAMs’ ability to represent multiple bits in a single cell can enable real-time data processing at low power in a small footprint. Therefore, employing ReRAM is the one of the very promising pathways to realize real-time processing in AI. In ReRAM, imperfection of the switching medium promotes filament formation, but it also leads to uncontrolled filament formation during each switching cycle. Therefore, the device exhibits non-uniform performance. Our group is working on a total new type of ReRAM devices to overcome the material limitations that have been holding back use of ReRAM arrays as an AI computing platform. Our current device uniformity is > 95% with the highest on-off ratio ever reported without utilizing selecting devices.

Skin Electronics

The operating space of electronic devices have extended from rigid and flat office desks to soft and pliable human body, improving the quality of life. Development of wearable (or epidermal) electronics drives not only in-situ monitoring health of the internal organs, but also instantaneous treatment for diseases. Furthermore, human beings have logged into the Internet of Things (IoTs) through wearable electronics.

Surprisingly, amid this sweeping trend, human skin has been treated as merely a flexible, stretchable, and soft space for mounting of wearable electronic devices. In fact, the skin is the outmost and the largest organ covering the external body surface and plays a vital role to maintain human life. Wearable electronics should not affect the physiological behavior of the skin (e.g., perspiration). Furthermore, wearable electronics should exhibit high durability on the skin for continuous operation for IoT applications.

Based on this criterion, our aim is to demonstrate non-invasive and long-lived (even symbiotic) electronics on the skin, called ‘Skin Electronics’. We have developed a skin-like patch which exhibit outstanding breathability, not blocking physiological behavior of the skin as well as providing conformal contact with the skin. Furthermore, high-performance inorganic devices can be fabricated on the skin-like patch without additional brittle polymer substrates (~100% yield), resulting in maximization of electromechanical performance of the devices such as strain sensors.

Single-Crystalline Monolayer 2D Materials

We fabricate wafer-scale single-crystalline 2D materials and study their heterostructures. We also study unique nanoscale mechanics in two-dimensional materials like graphene for single-atom-thickness precision control.