Neuromorphic computing, inspired by the energy efficiency of biological neural networks, aims to overcome limitations of von Neumann architectures. Memristive switching devices, with their high switching speeds and scalability, are promising candidates for next-generation digital memory. Two-dimensional (2D) materials offer advantages over conventional metal oxides for memristive devices due to their power efficiency, tunability, and heterointegration capabilities. Existing approaches utilizing 2D materials for memory functions rely on mechanisms like grain boundary vacancies, phase transitions, and Schottky barriers, but these methods often suffer from uncontrollable material properties. Substitutional doping presents a more controllable alternative, allowing precise tuning of electronic properties. Previous research has demonstrated the potential of doped 2D materials in memristive devices, but the exact role of dopant atoms in the memristor mechanism remains unclear. This study investigates a p-type doping strategy in WS₂ using Nb atoms, aiming to improve the conductance ratio and linearity of weight updates in synaptic transistors. The researchers used liquid-phase precursor chemical vapor deposition (CVD) to achieve uniform Nb incorporation into WS₂ crystals.

The introduction section reviews existing literature on neuromorphic computing, memristive devices, and the use of 2D materials in memory applications. It highlights the advantages of 2D TMDCs and discusses various mechanisms previously employed to achieve memory functions in 2D material-based devices, such as grain boundary-induced vacancies, phase transitions, and Schottky barriers. The authors also mention limitations of these methods, emphasizing the need for more controllable techniques like substitutional doping. Examples from previous studies using doped 2D materials for resistance-switching devices are provided, including work on MoTe₂, MoS₂Ox, and V-doped MoS₂, demonstrating the potential of this approach but also pointing to the need for further clarification on the role of dopant atoms in the memristor mechanism.

The researchers utilized atmospheric pressure CVD to grow Nb-doped WS₂ layers. A precursor solution containing niobium oxalate and sodium tungstate was prepared, with the ratio of W and Nb atoms precisely controlled to achieve different doping concentrations (0%, 2%, and 5%). The solution was spin-coated onto a SiO₂/Si substrate, which was then heated to 875 °C in a quartz tube under Ar flow. Monolayer WS₂ crystallites with triangular morphologies were obtained. The grown materials were characterized using various techniques. Raman spectroscopy, photoluminescence (PL), and X-ray photoelectron spectroscopy (XPS) were employed to assess the quality of the WS₂ flakes and confirm doping-induced alterations in the atomic structure. Scanning transmission electron microscopy (STEM) provided high-resolution imaging of the crystal structure, enabling the visualization of Nb atom distribution. Atomic force microscopy (AFM) was used to measure the thickness of the grown WS₂. Finally, synaptic transistors were fabricated using the monolayer Nb-doped WS₂ as the channel material, and their electrical properties were characterized. The performance of the synaptic transistors was evaluated by testing their ability to recognize handwritten digits from the MNIST dataset.

The CVD growth process yielded uniformly distributed monolayer WS₂ flakes. Nb doping reduced the average size of the WS₂ flakes. Raman and PL spectroscopy confirmed the successful incorporation of Nb into the WS₂ lattice, and the observed redshift in emission and binding energies indicated p-type doping. High-resolution XPS spectra confirmed the presence of Nb in the doped samples and showed shifts in the binding energies of W and S atoms, further supporting the p-type doping. HAADF-STEM imaging provided direct visualization of Nb atoms substitutionally incorporated into the WS₂ lattice. The fabricated synaptic transistors based on Nb-doped WS₂ exhibited a significantly enhanced switching ratio (10³), 100 times larger than undoped WS₂ devices. This improvement is attributed to the Nb atoms' role in trapping and detrapping electrons. The gate-tunable modulation of channel conductivity effectively simulated synaptic potentiation, depression, and repetitive learning processes. When applied to the MNIST handwritten digit dataset, the Nb-WS₂ synaptic transistor achieved a recognition accuracy of 92.30% after 125 training iterations.

The significantly improved switch ratio and recognition accuracy of the Nb-doped WS₂ synaptic transistors demonstrate the effectiveness of the p-type doping strategy. The results confirm the crucial role of Nb atoms in modulating the channel conductivity through electron trapping and detrapping, mimicking the behavior of biological synapses. The high recognition accuracy on the MNIST dataset highlights the potential of Nb-doped WS₂ for practical applications in neuromorphic computing. The study provides valuable insights into the design and optimization of high-performance synaptic transistors based on 2D TMDCs, offering a pragmatic and accessible atomic doping methodology.

This study successfully demonstrated the fabrication of high-performance synaptic transistors using Nb-doped WS₂. The atomic doping strategy led to a significant enhancement in the switch ratio and recognition accuracy, demonstrating the potential of this approach for neuromorphic computing applications. Future research could explore other dopant atoms and doping concentrations to further optimize device performance. Investigating different 2D TMDCs and exploring the integration of these synaptic transistors into more complex neuromorphic circuits would also be valuable.

The study primarily focused on a specific doping concentration range. Further investigation is needed to determine the optimal doping level for maximizing device performance. The sample size for evaluating the impact of doping on the size of the WS2 flakes could be larger to increase statistical significance. The long-term stability and reliability of the Nb-doped WS₂ synaptic transistors under continuous operation need further investigation.