Two-dimensional materials, such as molybdenum disulfide (MoS₂) and graphene, have emerged as promising candidates for neuromorphic computing due to their unique atomic-scale thickness and tunable electronic properties. These materials exhibit exceptional mechanical flexibility, high carrier mobility, and strong light-matter interactions, making them ideal for replicating the behavior of biological synapses and neurons. Their ultrathin nature allows for precise control over charge transport, enabling the design of energy-efficient and high-density neuromorphic devices.
One of the key advantages of 2D materials in neuromorphic computing is their defect-mediated switching behavior. Defects, such as vacancies or grain boundaries, can create localized states that influence conductivity. In MoS₂, sulfur vacancies act as trapping sites for charge carriers, enabling memristive switching analogous to synaptic plasticity. This property allows for the emulation of long-term potentiation (LTP) and long-term depression (LTD), which are critical for learning and memory in biological systems. Graphene, with its zero-bandgap structure, exhibits tunable resistance switching when interfaced with other materials, further enhancing its applicability in artificial synapses.
Heterostructure engineering plays a pivotal role in optimizing 2D material-based neuromorphic devices. By stacking different 2D layers, such as graphene-hBN-MoS₂, researchers can create tailored electronic band structures that mimic synaptic functions. For example, hBN acts as an insulating barrier that enables precise control over tunneling currents, while MoS₂ provides active switching sites. These heterostructures can replicate spike-timing-dependent plasticity (STDP), a fundamental learning rule in biological neural networks. Additionally, the integration of ferroelectric materials with 2D semiconductors introduces non-volatile memory effects, essential for low-power neuromorphic systems.
Optoelectronic synapses represent another breakthrough in 2D material neuromorphics. Unlike purely electronic devices, optoelectronic synapses leverage light-matter interactions to modulate synaptic weights. MoS₂-based phototransistors, for instance, exhibit persistent photoconductivity, allowing them to retain synaptic states after optical stimulation. This property is particularly useful for neuromorphic systems inspired by optogenetics, where light pulses control neural activity. Graphene’s broadband optical absorption further enhances its utility in such systems, enabling ultrafast photodetection and synaptic response times in the picosecond range.
Scalability and integration remain significant challenges for 2D material-based neuromorphic computing. While individual devices demonstrate impressive performance, achieving uniform large-scale fabrication is difficult due to material variability and defect distribution. Chemical vapor deposition (CVD) growth of 2D materials often results in inhomogeneous films, necessitating advanced transfer techniques to ensure device consistency. Moreover, integrating 2D materials with conventional silicon-based electronics requires novel fabrication approaches to mitigate interface defects and contact resistance.
Recent breakthroughs in optogenetics-inspired systems highlight the potential of 2D materials for brain-like computing. Researchers have demonstrated artificial synapses that respond to both electrical and optical stimuli, closely mimicking biological neurons. For example, graphene-MoS₂ heterostructures have been used to create devices that exhibit paired-pulse facilitation (PPF), a short-term synaptic plasticity mechanism. These systems can process visual information in a manner similar to the human retina, paving the way for energy-efficient neuromorphic vision sensors.
The mechanical flexibility of 2D materials also opens new avenues for wearable and implantable neuromorphic devices. Unlike rigid silicon-based systems, flexible 2D material circuits can conform to biological tissues, enabling seamless integration with the human body. This property is particularly advantageous for developing brain-machine interfaces and prosthetics that require adaptive learning capabilities.
Despite these advancements, several hurdles must be overcome to realize practical 2D material-based neuromorphic systems. Device-to-device variability, limited endurance, and the lack of standardized fabrication protocols remain critical issues. Additionally, the energy consumption of 2D neuromorphic devices, while lower than traditional CMOS systems, must be further reduced to compete with biological neural networks.
In summary, 2D materials offer unparalleled opportunities for neuromorphic computing due to their atomic-scale precision and multifunctional properties. Defect-mediated switching, heterostructure engineering, and optoelectronic synapses provide versatile mechanisms for emulating neural functions. While scalability and integration challenges persist, ongoing research continues to push the boundaries of brain-inspired computing, bringing us closer to realizing efficient and adaptive artificial intelligence systems.