Authors: Ke Zhang, Dehuan Meng, Feiming Bai, Junyi Zhai, Zhong Lin Wang
Published: 2020-07-02
Source: Full article
AbstractMemristor‐based architectures have shown great potential for developing future computing systems beyond the era of von Neumann and Moore's law. However, the monotonous electrical input for dynamic resistance regulation limits the developments of memristors. Here, a concept of a photon‐memristive system, which realizes memristance depending on number of photons (optical inputs), is proposed. A detailed theoretical derivation is performed and the memristive characteristics, as stimulated by the optical inputs based on a hybrid system, consisting of a low‐dimension photoelectric semiconductor and a ferroelectric substrate are determined. The photon‐memristive system is also suitable for nonvolatile photonic memory since it possesses three or more‐bit data storage, desirable resistance‐change space, and an ON/OFF ratio of nearly 107. The integrated circuit based on several photon‐memristive systems also realizes available photon‐triggered in‐memory computing. The photon‐memristive system expands the definition of memristors and emerges as a new data storage cell for future photonic neuromorphic computational architectures.