Since beginning his undergraduate studies in 1999, physics and materials have always been two inseparable "tags" for Liu Kai. Since joining Tsinghua University as an associate professor in 2015, he has been committed to researching low-dimensional materials.

Recently, he and his team have proposed a new method for constructing two-dimensional reconfigurable devices based on the channel gradient doping mechanism, using the bipolar molybdenum ditelluride (MoTe2) two-dimensional semiconductor channel material.

This method allows for the control of gas adsorption and desorption on the channel material, thereby enabling a wide range of reconfigurable functions on the simplest single-gate reconfigurable devices.

Through this, the contradiction between the low structural complexity and the rich reconfigurable functions of two-dimensional reconfigurable devices can be resolved.

Advertisement

At the same time, without changing the structural complexity, the reconfigurable functions of the device can be greatly expanded, which is expected to have a significant impact in the field of electronics.At the same time, this will bring reference significance to other reconfigurable devices and demonstrate the potential for developing multifunctional chips by integrating with silicon-based chips.

Currently, the challenges of miniaturization of integrated circuits are increasing. The reconfigurable device proposed this time, based on gradient doping, allows each transistor to have reconfigurable characteristics.

In the future, if this reconfigurable device is used to make integrated circuits, it is expected to achieve a new type of chip with high flexibility, high performance, and multifunctionality.

Such chips can not only undertake logical computing tasks and store computing results in real time, but also perceive the outside world through light, and also simulate the learning and forgetting functions of the human brain.

Furthermore, through advanced packaging technology, these chips can be combined with silicon-based chips to fully exert their respective strengths and provide a solid foundation for the next generation of computing architecture.By that time, such chips are expected to play a key role in fields such as artificial intelligence, the Internet of Things, and wearable devices.

The Dilemma of Silicon-based Reconfigurable Devices

Since the invention of integrated circuits, the development of information technology has mainly relied on the continuous reduction in the size of transistors on integrated circuits. However, now the size of transistors has approached the physical limit, and further miniaturization has become increasingly difficult.

As a result, people have started to study a special class of semiconductor devices - reconfigurable transistors. The characteristic of reconfigurable transistors is that multiple electrical functions can be implemented on a single transistor, and they can be switched between these functions.If every transistor on an integrated circuit possesses such reconfigurable capabilities, then the functionality of the entire circuit can be greatly expanded, thereby further enhancing the effective integration of the circuit without the need to continue to scale down the size of the transistors.

However, the mainstream integrated circuits are currently based on silicon, and a single silicon-based transistor cannot achieve the aforementioned reconfigurable capabilities.

Although some silicon-based chips can achieve reconfigurability at the chip level through additional circuits and added storage units, these additional circuits and units will lead to an increase in system complexity and production costs.

Therefore, people are very much looking forward to developing devices that can surpass silicon-based ones to create new reconfigurable devices.

Over the years, the Liu Kai research group has carried out a number of studies on two-dimensional semiconductor materials represented by transition metal dichalcogenides (TMDCs).This is a type of material with a natural ultra-thin layered structure, which has excellent electrical properties and can be effectively controlled.

However, to construct two-dimensional reconfigurable devices, reversible doping of two-dimensional semiconductor channel materials is required. To further achieve more reconfigurable functions, it is necessary to localize and refine the control of the reversible doping position.

Previously, other research teams have reported two-dimensional reconfigurable devices. In these achievements, scholars have used different reversible doping methods. Among them, gate control is the most common means used to regulate doping.

It is understood that a single-gate reconfigurable device is the simplest type of reconfigurable device. This type of device controls the doping of the entire or part of the channel through a single gate. Although the structure is relatively simple, it can only achieve two or three reconfigurable functions.

To achieve more functions, it is necessary to construct more gates to allow each gate to control a part of the channel separately. Although multi-gate reconfigurable devices can achieve more than three reconfigurable functions, multiple gates will increase the structural complexity of the device.Another approach to enhancing reconfigurability is to introduce other heterogeneous materials, but this would similarly lead to increased complexity in the device.

In general, the dilemma faced by silicon-based reconfigurable devices reemerges in two-dimensional reconfigurable devices: to achieve more functions, it is inevitable to increase system complexity and manufacturing costs.

Under these circumstances, Liu Kai's research group carried out this study.

After 100 reconstructions, performance stability can still be maintained.

Laser direct writing is a method that uses the photothermal effect produced by laser irradiation on a material, utilizing localized heating channels to achieve channel property control. Initially, they planned to use laser direct writing technology to fabricate devices on molybdenum ditelluride channels.Previously, the team had accumulated a significant amount of research results in the field of laser direct writing of two-dimensional materials and devices. However, during the process of this project, they encountered some technical challenges.

So they changed their approach and tried to use Joule heating instead of laser thermal effects to control the channel. The reason for doing this is that Joule heating is expected to be more stable and faster.

Regarding this, Liu Kai said: "Our initial idea was relatively simple, without considering too many semiconductor physics mechanisms. The introduction of Joule heating is achieved by inputting a larger source-drain voltage. According to the most intuitive Ohm's law, the source-drain voltage will introduce current, and this current grows linearly with the voltage."

But in the actual process, they found that the experimental phenomena deviated significantly from the original expectations: that is, the growth of the current not only did not conform to the linear law, but also decayed rapidly over time.

In response to this abnormal phenomenon, through long-term experiments and analysis, they discovered two overlooked factors: First, a large source-drain gate voltage will introduce a large effective gate voltage; second, the effective gate voltage will cause molybdenum ditelluride to adsorb water and oxygen molecules from the air.After a period of further exploration, the research team discovered that these two overlooked negative factors could be "turned into friends." That is, by using effective gate voltage and gas adsorption, they could directly construct np junctions and pn junctions, which are two types of diodes.

Subsequently, they began to shift their experimental goals: first, they verified the possibility of constructing diodes and achieved high-performance diodes.

After discovering the impact of gas adsorption and desorption, they managed to verify the concept on the device within just one or two months and successfully obtained high-quality diode data. However, in the dry winter of Beijing, the device was quickly "damaged" by static electricity.

In the following four or five months, the research team prepared dozens of devices to replicate the previous results. Liu Kai said, "Repetitive experiments are often very painful, we all hope to replicate the previous results, and even obtain higher quality data, but it is often not satisfactory."

After a long period of repetitive experiments, on a sample that was not originally expected to be promising, the previous phenomenon was finally replicated.In subsequent experiments, they gradually mastered the experimental patterns and, by summarizing the construction ideas of this device, refined the gradient doping method.

Based on new data and experience, the gradient doping mechanism was further perfected and could be extended to more functions. With new experimental experience, subsequent repetitive experiments also became easier.

Through this, they used gradient doping to build memory devices, neuromorphic devices, and logic memory devices on the same structural device, creating this function-rich reconfigurable device.

The reason why this device can continuously switch between various reconfigurable functions and still maintain stable performance after 100 reconstructions is that the reconfigurable molybdenum telluride device is built on gas adsorption and desorption doping. Gas adsorption and desorption itself has a high degree of reversibility, and the gradient doping method has high flexibility.

By introducing effective gate voltages with different gradients, different gas adsorption and desorption states can be triggered on the same device, thereby introducing different channel gradient doping distributions, allowing the device to switch to various reconfigurable functions. At this time, by continuously applying different reconstruction voltages, it is possible to continuously switch between different functions.The key to achieving the stability of restructuring lies in the process of gas adsorption and desorption, which is repeatedly carried out, and the channels can always remain stable.

This stability can be attributed to the following reasons:

Firstly, the two-dimensional molybdenum ditelluride material itself is very stable in the air.

Secondly, because the selected material is a few layers of molybdenum ditelluride with a thickness of 2-11nm, instead of using a single-layer material, the stability of the few-layer material is better than that of the single layer.

Thirdly, the gradient doping method uses a relatively low voltage, which will not cause significant thermal effects in the channel material, thereby avoiding irreversible thermal oxidation reactions in the channel material.Liu Kai said: "The deepest impression this work has given us is that when the initial idea of a task encounters difficulties, it is essential to delve into the underlying principles and promptly adjust the research direction and content. Once the target of effort is identified, it must be meticulously refined."

Ultimately, the related paper was published in Nature Electronics (IF 34.3) with the title "Programmable graded doping for reconfigurable molybdenum ditelluride devices". Tsinghua University Ph.D. student Peng Ruixuan is the first author, and Liu Kai served as the corresponding author.

At the same time, a commentary article [3] was also published in the same issue of Nature Electronics, highlighting the high performance and high reconfigurability of the device.

Next, Liu Kai and his team will continue the concept of graded doping, exploring other structures and mechanisms of reconfigurable devices, striving to enhance the performance of reconfigurable devices and expand their other reconfigurable functions. They will also explore the array application of reconfigurable devices, further demonstrating the potential and prospects of the graded doping method.