Research Topic

  • Home
  • Research
  • Research Topic

Neuromorphic and Future Logic Device

Neuromorphic device

Despite its high performance, DNN (Deep Neural Network) suffers from its high power consumption to process the immense amount of data and neuromorphic computing inspired by the low-power operation of the human brain was newly suggested. Ferroelectric devices (MFS/MFIS/MFMIS FeFET, FTJs, etc…) are one of the main candidates that can be used as synaptic devices and neuron circuits because of their manifold functionality, CMOS compatibility. Leaky-FeFET shows accumulation and firing operations with leaky characteristics, so it is free from using bulk and complex components such as capacitors, comparators, and amplifiers. Therefore, high integration density, low-cost fabrication process, and low-power operation can be achieved from the SNN system.

Future CMOS device, 3D Monolithic IC

To extend further success of the semiconductor industry, 3D IC, allowing vertical stacking of transistor layers, have emerged as a promising solution. Especially, monolithic 3D (M3D) integration provides several advantages over traditional 3D ICs due to the extremely small contact size to connect two stacked layers. Small size of contacts enables ultra-high integration density, which considerably reduces area and fabricating cost, and reduction of total interconnection length, improving power per performance. However, for the formation of M3D IC, preventing the performance deterioration of already fabricated bottom-tier devices during the sequential processing of top-tier devices is essential. Therefore, to avoid the thermal damage to bottom-tier devices, it is essential to engage in research focused on selective annealing processes and materials that can be processed at lower temperatures.

a