Areas of Expertise
Neuromorphic Computer
Nature Electronics, Published online (2022) | Featured as cover article
Nature Nanotechnology 15, 574-579 (2020)
Nature Materials 17 (4), 335 (2018)
Artificial intelligence (AI) can exceed the capabilities of humans in certain classes of problems including image, speech recognition, and gaming. However, the practical applications of AI, which require the processing of tremendous amounts of data and thus cause significant demands on computing speed and power efficiency, are limited by current computing hardware.
I developed alloyed metal-based artificial synapses, which substantially reduced the non-uniformity and improved analog switching behaviors. The core operation of deep neural networks has been demonstrated reliably and effectively on large-scale artificial synapses crossbars. The chip with tens of thousands of artificial synaptic devices will be a foundational work for future analog AI hardware that can replace the current graphic processing unit (GPU) led by Nvidia.
Furthermore, I am extending the functionality of neuromorphic computing by emulating the structure and working principles of the biological neurons in the human brain. This offers a promising approach to the development of more energy-efficient and highly-parallel AI hardware.
Quantum Nano-Devices
Nature Nanotechnology 18 (5), 464-470 (2023)
Nature Nanotechnology 17 (10), 1054-1059 (2022)
Nature 544 (7650), 340-343 (2017)
Science 362 (6415), 665-670 (2018)
Nature 558 (7710), 410-414 (2018)
Nature Photonics 12 (1), 22-28 (2018)
Nature 2D Materials and Applications 2 (1), 30 (2018)
The transformation of digital computers from bulky machines to portable systems has been enabled by advanced processing technologies. However, as this conventional scaling pathway has approached atomic-scale dimensions, the conventional transistor technology is limited by quantum physics.
Nano-thick two-dimensional (2D) layered materials have emerged as an alternative to conventional technology, promising new paradigms in computation, communication and sensing. The convergence between quantum materials properties and prototype quantum devices is especially apparent in the field of 2D materials.
My findings include light-matter interactions at 0.3 nanometer thick 2D materials (graphene) and interlayer carrier recombination within 1 nanometer thick 2D materials (transition metal dichalcogenide), which can be used for a ultra-high-precision optical clock and optical devices. These various 2D materials offer a broad range of materials properties, high flexibility in fabrication pathways, and the ability to form multi-functioning nano-device systems.
Heterogeneous Systems
Science 377 (6608), 859-864, (2022)
Science Advances Vol. 7, Issue 27 (2021)
Nature 578 (7793), 75-81 (2020)
Nature Nanotechnology 15 (4), 272-276 (2020)
Nature Materials 18 (6), 550 (2019)
ACS Sensors 5, 6, 1582-1588 (2020)
Heterogeneous electronic and optoelectronic systems have been desirable to establish a versatile platform for sensing and processing signals. However, the stiffness of thick-materials and lack of understanding of material science hinder the development of heterogeneously integrated systems.
By introducing a novel material growth technique as known as ‘remote epitaxy’ and a unique layer transfer technique (2DLT), I opened a myriad of possibilities for multifunctional electronic systems. In addition, new coupling phenomena and physics were observed between 3D bulk materials and 2D atomic layers.
In combination with large-scale artificial synaptic devices and quantum nano-devices, my heterogeneous system will become a core to connect human and computer at neuronal interfaces and support human activities as form of medical devices such as artificial eyes and prosthetic skin.
Previous Projects
Many-body Physics in Nano-materials
Novel physics of two-dimensional quantum materials has not been well understood although there is huge potential for quantum computing and flexible electronics due to their strong spin-orbit coupling and nano-meter thickness. The project aimed to unveil carrier dynamics in two-dimensional (2D) nano-materials. The project was funded by National Science Foundation (NSF, 1611598).
Energy-efficient Artificial Intelligence with Analog Epitaxial Synaptic Arrays
State-of-the-art deep learning algorithms tend to present very large network models, which poses significant challenges for hardware, especially for memory. Emerging resistive devices was proposed for synaptic devices and their parallel neural computing. The energy-efficient intelligent hardware for power-/area-constrained local mobile/wearable devices has been developed, funded by National Science Foundation (NSF, 1740184).
Brain-cell Recording for Real-time Processing at Human-computer Interface
To comprehend how the human brain works, there have been many attempts to sense biological signals directly from neurons. However, due to tremendous data bandwidth and different data topology, there is a large need for neuron-like electronic device components. For the purpose of reducing a gap between biological synapses and artificial synapses, ionic and electronic behaviors in memristive devices have been investigated.