The neuromorphic nature of the resistive switching in TiO2 memristors has triggered a series of studies addressing their functional coupling with living biological systems. The common features of the electroconductive behavior of memristive and biological neural networks have been revised in terms of physical, mathematical, and stochastic models (Chua, 2013; Feali and Ahmadi, 2016). The memristive electronics was shown to support important synaptic functions such as spike timing-dependent plasticity (Jo et al., 2010; Pickett et al., 2013). Recently, a memristive simulation of important biological synaptic functions such as non-linear transmission characteristics, short-/long-term plasticity, and paired-pulse facilitation has been reported for hybrid organic–inorganic memristors using Ti-based maleic acid/TiO2 ultrathin films (Liu et al., 2020). In relation to this, functionalized TiO2 memristive systems may be in competition with the new generation of two-dimensional memristive materials such as WSe2 (Zhu et al., 2018), MoS2 (Li et al., 2018), MoS2/graphene (Kalita et al., 2019), and other systems (Zhang et al., 2019a) with ionic coupling, ionic modulation effects, or other synapse-mimicking functionalities. Furthermore, the biomimetic fabrication of TiO2 (Seisenbaeva et al., 2010; Vijayan and Puglia, 2019; Kumar et al., 2020) opens up new horizons for its versatile microstructural patterning and functionalizations.
As they mimic the synapses in biological neurons, memristors became the key component for designing novel types of computing and information systems based on artificial neural networks, the so-called neuromorphic electronics (Zidan, 2018; Wang and Zhuge, 2019; Zhang et al., 2019b). Electronic artificial neurons with synaptic memristors are capable of emulating the associative memory, an important function of the brain (Pershin and Di Ventra, 2010). In addition, the technological simplicity of thin-film memristors based on transition metal oxides such as TiO2 allows their integration into electronic circuits with extremely high packing density. Memristor crossbars are technologically compatible with traditional integrated circuits, whose integration can be implemented within the complementary metal–oxide–semiconductor platform using nanoimprint lithography (Xia et al., 2009). Nowadays, the size of a Pt-TiOx-HfO2-Pt memristor crossbar can be as small as 2 nm (Pi et al., 2019). Thus, the inherent properties of memristors such as non-volatile resistive memory and synaptic plasticity, along with feasibly high integration density, are at the forefront of the new-type hardware performance of cognitive tasks, such as image recognition (Yao et al., 2017). The current state of the art, prospects, and challenges in the new brain-inspired computing concepts with memristive implementation have been comprehensively reviewed in topical papers (Jeong et al., 2016; Xia and Yang, 2019; Zhang et al., 2020). These reviews postulate that the newly emerging computing paradigm is still in its infancy, while the rapid development and current challenges in this field are related to the technological and materials aspects. The major concerns are the lack of understanding of the microscopic picture and the mechanisms of switching, as well as the unproven reliability of memristor materials. The choice of memristive materials as well as the methods of synthesis and fabrication affect the properties of memristive devices, including the amplitude of resistive switching, endurance, stochasticity, and data retention time.