Research

Neuromorphic hardware is a sort of artificial intelligence (AI) hardware that can implement AI functions with ultra-low power by mimicking the human brain that consumes only 20 W. Implementing neuromorphic hardware requires artificial neurons and synapses that can mimic the functions of biological neural networks. We are conducting research on neuron devices and synaptic devices based on various structures (MOSFET, biristor, memristor) and materials (ferroelectric, charge trap layer, electrolyte, RRAM), and neuromorphic systems (deep neural network, spiking neural network) using them.

World smallest sub-5nm all-around gate FinFET with 3nm silicon-fin width were fabricated and reported (VLSI'06). These results show feasibility to continue the Moore's law beyond the sub-5nm using the silicon-based material. World smallest sub-10nm nanowire SONOS Non-volatile memory were were demonstrated (VLSI'07). Furthermore, by combining a floating body and oxide/nitride/oxide stacked gate dielectric, the high speed DRAM operation and the non-volatile memory operation is realized in the single transistor (URAM, unified-RAM). Especially, the DRAM operation is feasible without a cell capacitor by using the floating body capacitor. According to an end-users' demand, one can allocate each operation to utilize each advantageous function. The electrical properties and thermal behaviors of CMOS devices are analyzed through simulation and modeling.

Novel memory technology is invented to overcome limitation of the conventional memory technology. The conventional 1T/1C DRAM cell is facing process challenges due to the high aspect ratio of the capacitor. A novel DRAM (Dynamic random access memory) cell named ‘biristor’ (‘bi-stable resistor’) is the BJT with floating base, and it has the capacitorless and gateless structure for 4F2 high speed DRAM applications. Another novel memory cell named ‘unified-RAM’ realizes the high speed DRAM operation and the non-volatile memory operation in the single transistor without capacitor.

Energy harvesting technology is related to scavenge electrical energy from ambient environment such as radiational, thermal and mechanical energy. Since an energy harvester can provide green and sustainable electrical energy, it becomes a strong candidate as an alternative energy source for sensors in Internet of Things (IoT). Especially, triboelectric nanogenerator, which is a kind of mechanical energy harvester based on triboelectric effect and electrostatic induction, has rapidly developed based on a wide range of material and structure selection since it was invented in 2012.

Security technology makes your information be your own information.

Our lab focus to devise the new security primitive in which all people are marvelous.
- Random Number Generator(RNG): Nature allows human the law of nature. However, Nature also gives us human unpredictable disorder. We call the disorder random. Random number generator is just device to generate digital data from Nature’s disorder. The extracted random bits are applied to enhance the security technology.

Our lab have published RNG technology from Nature’s chaotic properties such as wind.
- Physically Unclonable Function(PUF): Authentication is a process to prove who I am. Biometric data and password, OTP(One-time pad) are methods to prove who I am. Then, how about to attaining authentication of device? Physically unclonable function is devised to make authentication for device. We call PUF to be device finger print.

Our lab have published PUF technology from existing semiconductor device such as flash memory