On November 14, 2023, Beijing time, Kai Liu’s research group from the School of Materials Science and Technology of Tsinghua University published a research paper entitled “Programmable graded doping for reconfigurable molybdenum ditelluride devices” in Nature Electronics.
Reconfigurable transistors can expand the functionality of integrated circuits and reduce the effective linewidth of the device, thereby improving the integration of the system. However, silicon-based transistors can only be reconfigurable at the chip level through control circuitry and additional memory cells, resulting in increased system complexity and manufacturing costs. Two-dimensional semiconductor materials have great advantages in realizing the reconfigurable functions of transistor devices, but how to realize rich reconfigurable functions on devices with low structural complexity is still a very challenging problem.
In this paper, the authors propose a two-dimensional reconfigurable device construction method based on the channel gradient doping mechanism by using bipolar molybdenum ditelluride (MoTe2) channel material. In this method, the most abundant reconfigurable functions reported so far are achieved in the simple single-gate transistor device, including the bipolar tunable diode, memory, logic memory, and the steady-state function of heterosynaptic plasticity and heterosynaptic plasticity of the three-terminal artificial neuraptic device, which solves the contradiction between the low structural complexity and the rich reconfigurable functions of the reconfigurable device.
Since the invention of integrated circuits, the continuous shrinking of transistor size has driven the rapid development of the information age. However, as the process node moves to sub-10nm, it is becoming increasingly difficult for traditional silicon-based FETs to shrink further due to limitations such as short-channel effects. In the post-Moore era, the concept of reconfigurable transistors was introduced in order to continue to gain the performance benefits of scaling without actually reducing the size of the device. Reconfigurable transistors can realize a variety of freely switching functions on a single transistor device, so they can expand the functionality of integrated circuits and reduce the effective linewidth of the device, thereby improving the integration of the system. However, silicon-based transistors are difficult to achieve reconfigurability at the single device level, and while reconfigurability at the chip level can be achieved through control circuitry and additional memory cells, these additional circuitry and memory cells lead to increased system complexity, reduced integration, and increased manufacturing costs. Therefore, the development of new reconfigurable devices beyond silicon has become an urgent need for the development of integrated circuits.
Two-dimensional semiconductor materials represented by transition metal chalcogenides (TMDCs) have great advantages in realizing the reconfigurable function of transistor devices due to their atomic-level thickness, layered structure without hanging bonds, and electrical properties that are easily controlled by external fields. In order to construct 2D reconfigurable devices, it is necessary to reversibly dop the 2D semiconductor channel materials, and in order to further realize more reconfigurable functions, it is necessary to finely control the reversible doping sites. This makes a natural contradiction between the structural complexity of the device and the richness of reconfigurable functions. For example, devices with low structural complexity, such as single-gate reconfigurable devices, can only achieve two or three reconfigurable functions, and in order to achieve more than three reconfigurable functions, people have developed multi-gate devices and ionic gate devices, but the introduction of multi-gate or heterogeneous materials makes two-dimensional reconfigurable devices like the dilemma faced by silicon-based reconfigurable devices, which inevitably increases system complexity and manufacturing costs. As a result, it remains challenging to achieve rich reconfigurable functionality with devices with low structural complexity.
In order to solve the above key problems, the authors proposed a two-dimensional reconfigurable device construction method based on the channel gradient doping mechanism by using bipolar MoTe2 channel materials. This method achieves the richest reconfigurable function reported to date among the simplest single-gate transistor devices. The basic principle is to apply a large source-leakage voltage and a large gate voltage on both sides of the 2D MoTe2 device at the same time, so that an effective gate voltage with gradient distribution along the channel direction is introduced. The effective gate pressure can accurately control the gas adsorption and desorption on the surface of the channel, so as to realize the gradient doping and polarity control of the bipolar MoTe2 channel. By controlling the source-drain voltage and gate voltage applied during the reconfiguration process, different gradient doping distributions can be introduced in the channel, enabling a variety of reconfigurable functions of the device, including a diode with adjustable polarity, memory, logic memory, and a three-terminal artificial synapse (Figure 1).
Figure 1. Doping mechanism and basic working principle of a single-gate reconfigurable MoTe2 device
Kelvin probe force microscopy images show that the reconfigurable MoTe2 device can be reconstructed into four states: np junction and pn junction, p-doped and n-doped, respectively, through precisely controlled gradient doping, so as to realize the polar tunable diode function and memory function. In the diode state, the device has a rectification ratio of 104 and photoelectric detection capability, and as a memory, the device has a memory window of about 25 V, a memory ratio of more than 103, and a hold time of more than 36 hours (Figure 2).
Figure 2. Gradient doping based on effective gate voltage regulation enables a polar tunable MoTe2 diode and memory
Bioplasticity is the foundation of biological learning and memory. The reconfigurable MoTe2 device reported in this study can be used as an artificial nerve synapse to simulate a variety of plasticity functions of biological synapses (including biological homosynaptic plasticity, heterosynaptic plasticity, and heterosynaptic plasticity), which is very beneficial for the construction of a new artificial neural network with high stability and strong specificity. The device also has a modulation power consumption as low as 7.3 fW when simulating bioheterosynaptic plasticity, which is better than previously reported 2D artificial neural synapses with metaplasticity (Figure 3).
Figure 3. Gradient doping based on effective grating pressure regulation simulates the homeostatic function and heterosynaptic plasticity function of biological synaptic plasticity, heterosynaptic plasticity, and heterosynaptic plasticity
The authors further demonstrate the continuous reconfiguration capability of reconfigurable MoTe2 devices. The device can be continuously switched between multiple reconfigurable functions and remains stable after 100 reconfigurations (Figure 4).
Figure 4. Continuous reconfigurable operation of reconfigurable MoTe2 devices
In conclusion, this study deepens people’s understanding of the concepts of effective gate voltage and gradient doping in 2D semiconductor materials and devices, solves the contradiction between the low structural complexity and rich reconfigurable functions of reconfigurable devices, and provides a new idea for the construction of multifunctional and high-performance 2D reconfigurable devices.
Associate Professor Liu Kai of the School of Materials Science and Technology of Tsinghua University is the corresponding author of this paper, and Peng Ruixuan, a 2019 doctoral student of the School of Materials Science and Technology of Tsinghua University, is the first author of this paper. Other important collaborators include Academician Fan Shoushan of the Department of Physics of Tsinghua University, Professor Zhou Peng of the School of Microelectronics of Fudan University, Professor Song Cheng and Associate Professor Wang Chen of the School of Materials Science and Technology of Tsinghua University, Professor Fan Zhiyong of the Hong Kong University of Science and Technology, and Wu Yonghuang, a 2019 doctoral student of the School of Materials Science and Technology of Tsinghua University. (Source: Web of Science)
Related Paper Information:https://doi.org/10.1038/s41928-023-01056-1