In this colorful world, the construction of a complete physical theory to reveal the nature of the world needs to be conducted in high dimensionality, for example, in the M-theory established by Edward Witten, the basic particles and interaction forces between them are described in eleven-dimensionality. On the other hand, in material science and engineering, reducing the dimensionality of materials will greatly mitigate the complexity of the atomic and molecular systems and benefit the study on their properties. Moreover, the quantum confinement effects induced by the reduction of the material dimensionality will result in many novel electronic/optoelectronic properties which are desired in designing advanced electronics/optoelectronics. In this dissertation, the fabrication and performance characterization of electronics/optoelectronics based on the low-dimensional materials represented by two-dimensional (2D) nanoflakes are presented, the comprehensive device performance characterization, working mechanism uncovering, and subsequent computing systems simulated based on these advanced devices will contribute to the material science community and pave the way for future smart signal processing systems at the edge.

First, we develop a growth strategy for high-quality tellurium (Te) nanobelts, where the hexagonal boron nitride (h-BN) with the atomically flat surface is introduced into the chemical vapor deposition (CVD) system as the growth substrate. Based on the obtained Te samples, we fabricate high-performance p-type global-gate field-effect transistors (FETs) and the field-effect hole mobilities at room temperature is up to 1370 cm²V^-1s^-1, which is among the highest values reported in literature. The ultrahigh field-effect hole mobilities can be explained by the significantly reduced photon scattering induced by the ultra-flat surface without dangling bonds of h-BN substrate. Moreover, to further demonstrate the compatibility of our growth strategy with semiconductor processing technology, we successfully fabricate p-type Te FETs in local-bottom gate architecture and this device also exhibit considerable performance.

Besides the basic Te FETs, we also design and fabricate two memory devices based on van der Waals (vdW) heterostructures enabled by Te nanoflakes. The first one is the electronic/optoelectronic memory unit based on CuInP₂S₆/graphene/h-BN/Te (CIPS/Gr/h-BN/Te) vdW heterostructure, which presents both long-term potentiation/depression memory states triggered by electrical pulses and short-term memory behaviors induced by 1550-nm laser pulse (a typical wavelength in the conventional fiber optical communication band). Both the long-term electronic and short-term optical memory behaviors can be ascribed to the ferroelectric effects of CIPS nanoflake. Leveraging these rich dynamics, a fully memristive in-sensor reservoir computing (RC) system is demonstrated to simultaneously sense, decode and learn messages transmitted by optical fibre. After feeding the RC system with 60,000 handwritten digits for training, the obtained accuracy in 10,000 testing digits reaches ~0.8 that is comparable to the identification achieved by the software.

The second device is a multifunctional floating gate (FG) memory device enabled by Te/h-BN/MoS₂ vdW heterostructure. Different from the previous one where Te nanoflakes serve as the channel material, Te nanoflakes are used as the charge trapping layer in this device. Under the intense input stimuli, this device exhibits both superior long-term electronic behaviors including ~10⁸ extinction ratio, ~100-ns switching speed, >4000 cycles, >4000-s retention stability, which are comparable to the most advanced FG memory devices based on 2D nanoflakes. Besides, this device also presents admiring nonvolatile multi-bit optoelectronic programmable characteristics including >60 optoelectronic memory states and >1000-s retention stability. However, after weakening the stimuli intensity, the long-term memory behaviors will degrade into short-term ones. Based on the short-term electronic and optoelectronic memory behaviors, we simulate a multi-modal RC system with a high recognition accuracy of 90.77% for event-type multimodal handwritten digit-recognition. The successful fabrication of the devices based on CIPS/Gr/h-BN/Te and Te/h-BN/MoS₂ vdW heterostructures with the simulation of the RC system working at the optical communication band will offer new possibilities for advanced electronics/optoelectronics in low-dimensional material family and provides insights for future smart signal processing systems at the edge.

In the end, this thesis shares personal insights on the existing challenges and outlook on the research field about advanced electronics/optoelectronic based on low-dimensional materials.