Highlight

Intro Current computing technology is based on von Neumann architecture, which consists of a processing unit (PU) for performing logic operations and a memory for storing information. As the PU and memory are physically separated, data transfer between them requires a substantial energy consumption and time delay. This so-called von Neumann bottleneck results in a serious problem for a state of the art computing technology, in particular for large-capacity information processing such as artificial intelligence (AI), internet of things (IoT), and big-data analytics. In this work, we developed a spin logic device by employing electric-field controlled spin-orbit torque switching, which performs both logic and memory functions simultaneously. Moreover, by tuning the polarity of the electric-field effect, we firstly realized both “n-type” and “p-type” spin logic devices and demonstrated complementary logic operations. This paves the way for the development of non-volatile and reconfigurable logic devices, which can tackle challenges of conventional semiconductor-based devices such as the von Neumann bottleneck or the physical limit of miniaturization. 
Principal Investigator Prof. Byong-Guk Park 
Date 2019-01-31 

* List of 'KAIST Top 10 Research Achievements of 2018'

 

College

Department

Principal Investigator(s)

Project Title

College of Natural Sciences

Dept. of Physics

Jaewook Ahn

 Heung-Sun Sim

Rydberg-atom quantum simulator development

 Dept. of Chemistry

 Mu-Hyun Baik

From C-H to C-C Bonds at Room Temperature

Graduate School of Nanoscience and Technology

Yong Woon Kim

Role of rodlike counterions on the interactions of DNAs

College of Life Science &Bio-Engineering

Dept. of Biological Sciences

 Daesoo Kim

Medial preoptic area induces hunting-like behaviors to target objects and prey

Graduate School of Medical Science and Engineering

 Jeong Ho Lee

Identification of the origin of brain tumors and new therapeutic strategy

College of Engineering

Dept. of Mechanical Engineering

 Bumki Min

Linear Frequency conversion of light at a spatiotemporal boundary

School of Electrical Engineering

 Jun-Bo Yoon

Industrial-grade Flexible Transparent Force Touch Sensor

Dept. of Industrial and Systems Engineering

Heeyoung Kim

Detection and clustering of mixed-type defect patterns in wafer bin maps

Dept. of Materials Science and Engineering

Byong-Guk Park

Development of reconfigurable spin-based logic device

Dept. of Nuclear and Quantum Engineering

Sung Oh Cho

Development of Miniaturized X-ray Tube based on Carbon Nanotube and Electronic Brachytherapy Device

 ※ The order of listed projects is based on the university’s organizational structure, without relation to the ranking.

 
 
1. Research Background: Next-generation semiconductor material technology
 
• Currently used computing technology uses a von Neumann structure in which logic elements and memory elements are spatially separated (Figure 1a). Therefore, power consumption and signal delay due to signal transmission between two devices can not be avoided for information processing, and power consumption is caused by continuous power supply for information maintenance due to volatility characteristics. In addition, since the performance of a computer CPU and a smartphone AP is proportional to the number of memory / logic devices, in order to increase the capacity continuously, improvement of integration of devices and reduction of power consumption must be preceded.
 
• One of the ways to solve this fundamental problem is to introduce logic memory devices with information storage capability in logic devices (Figure 1b). To do so, nonvolatile and programmable logic devices must be developed.
 
1.png
Fig 1. Comparison on memory device structure

 

 Spintronics is a research field that develops electronic devices with new functions that are not possessed by existing materials / devices by using both electric charge and spin, which are inherent properties of electrons. Representative device based on spintronics are MRAM, which has non-volatile and high-speed operation characteristics and are being actively developed as the next-generation memory by domestic and foreign semiconductor companies such as Samsung, Hynix, Intel and Qualcomm. Recently, following the success of magnetic memory, attention has been focused on the development of logic devices based on spintronics, However, spintronics based logic devices have been studied to perform only simple logic functions, and complementary logic operations, a core technology for configuring logic devices, have not yet been developed.
 
2. Research Description: Development of complementary spin-logic device: Smart electronic device that simultaneously operate logically and memorize
 
 In this study, spintronic based logic device was developed by using the technique of controlling the spin orbit torque, which is one of the next generation magnetic memory operation method, by using external electric field (Fig. 2a). The spin-orbit torque is a technique for controlling the magnetization direction by using the interaction between the electron's spin and the orbit. The magnetization inversion speed is about 10 times faster than the conventional spin transfer torque, which is very advantageous for the logic device application.
 
 First, we fabricated a device with Ta / CoFeB / MgO structure which has vertical anisotropy characteristic, and developed a technology to control the spin-orbit torque-based switching current by applying an electric field to this device. By applying an electric field of about 5 MV/cm, the switching threshold current can be controlled by more than 30% (Fig. 2b). Based on these characteristics, we developed a spin-based logic device that implements logic functions such as "AND" and "OR".
 
 The polarity of the field effect could be controlled by adjusting the oxidation state of the CoFeB / MgO interface. We have developed an "n-type" logic device which can be easily switched in the (+) electric field depending on the oxidation state and a "p-type" logic device in which the switching is easy in the (-) electric field application (Fig. 2b). We have developed a logic device that combines these "n-type" and "p-type" devices to perform complementary logic functions (Figure 2c). This is the first device to replace CMOS, the basic structure of existing semiconductor logic devices, as a spin-based technology.
 
 We have demonstrated that spin-based logic devices can increase integration and reduce power consumption compared to conventional semiconductor-based devices by using computer simulation. We have designed AI and IoT chips and evaluated their performance.
 
 In addition to these results, we developed material technology to improve the performance of spin-logic devices. Especially, ferromagnetic / nonmagnetic interfacial induced spin current generation materials can be used to develop low power, high spin torque - based devices by enabling arbitrary control of the spin direction and enabling switching efficiency and non - magnetic field operation.
 
 
2.png

Fig 2. Spintronics based logic device

 
3. Beneficial Effects
 
 A spin-based logic device is a completely new device that can simultaneously operate and memorize, and is a technology that can overcome the fundamental limitation of existing computing technology. For example, it breaks the boundaries between currently existing memory devices (DRAM, FLASH) and logic devices (CMOS), and integrates them into a single nonvolatile logic device. This nonvolatile logic device can be applied not only to lowering the process cost by simplifying the device but also to the neuromorphic computing which can do logical operation and memorize simultaneously.
 
 The spin-orbit torque technology developed in this study is a core technology that can replace high-speed SRAM-based cache memory when applied to magnetic memory. It is nonvolatile and will contribute to the development of related technologies by reducing memory power consumption and applying to electronic devices that require low power operation such as wearable, mobile, biosensor and IoT (Internet of Things).
 
 The domestic semiconductor industry continues to lead the global market in memory semiconductors, but the recent threat from large-scale memory semiconductor investments in China is becoming visible. The results of this study can be used as the next generation semiconductors related technology, which will help to expand the market of non - memory logic device part with strengthening the leadership of memory industry.