A MoS2 Hafnium Oxide Based Ferroelectric Encoder for Temporal‐Efficient Spiking Neural Network

• Published in: Advanced materials (Weinheim) 2023-01, Vol.35 (2), p.e2204949-n/a
• Main Authors: Chien, Yu‐Chieh, Xiang, Heng, Shi, Yufei, Duong, Ngoc Thanh, Li, Sifan, Ang, Kah‐Wee
• Format: Article
• Language: English
• Subjects: 2D materials; Accuracy; Artificial intelligence; Biomimetic materials; Charge efficiency; Coders; Data processing; ferroelectric encoder; Ferroelectric materials; Ferroelectricity; Hafnium oxide; hafnium zirconium oxide; Machine learning; Materials science; Molybdenum disulfide; Neural networks; Noise levels; Noise prediction; Random noise; Spiking; spiking neural networks; Switching; time‐to‐first‐spike encoding scheme
• ISSN: 0935-9648; 1521-4095
• DOI: 10.1002/adma.202204949

Spiking neural network (SNN), where the information is evaluated recurrently through spikes, has manifested significant promises to minimize the energy expenditure in data‐intensive machine learning and artificial intelligence. Among these applications, the artificial neural encoders are essential to convert the external stimuli to a spiking format that can be subsequently fed to the neural network. Here, a molybdenum disulfide (MoS2) hafnium oxide‐based ferroelectric encoder is demonstrated for temporal‐efficient information processing in SNN. The fast domain switching attribute associated with the polycrystalline nature of hafnium oxide‐based ferroelectric material is exploited for spike encoding, rendering it suitable for realizing biomimetic encoders. Accordingly, a high‐performance ferroelectric encoder is achieved, featuring a superior switching efficiency, negligible charge trapping effect, and robust ferroelectric response, which successfully enable a broad dynamic range. Furthermore, an SNN is simulated to verify the precision of the encoded information, in which an average inference accuracy of 95.14% can be achieved, using the Modified National Insitute of Standards and Technology (MNIST) dataset for digit classification. Moreover, this ferroelectric encoder manifests prominent resilience against noise injection with an overall prediction accuracy of 94.73% under various Gaussian noise levels, showing practical promises to reduce the computational load for the neural network. A temporal‐efficient noise‐resilient ferroelectric encoder, exploiting the time‐to‐first‐spike encoding scheme, is proposed for information preprocessing in spiking neural networks. Our device manifests prominent ferroelectric dynamics, showing an excellent classification accuracy of 95.14%, in addition to resilience against noise attacks. This work manifests the potential to enable spike‐driven computations for energy‐efficient machine learning and artificial intelligence.