A recent review about memristors and their applications as memories and synapses can be found in ^[1][7]. Furthermore, the majority of memristors are oxide memristors, in which the reversible transformation from HRS to LRS is due to conduction filaments formed inside the oxide; for this reason, they are also termed as filament memristors.

In the Set process, pairs of oxygen vacancies (anions) are created inside the oxide due to the impact ionization process: O_L → V_O²⁺ + O_i²⁻, where O_L are the oxygen atoms in the oxide lattice, V_O²⁺ are the oxygen vacancies and O_i²⁻ are supplementary oxygen ions produced in the lattice. The O_i²⁻ ions are migrating to the top electrode, while the oxygen vacancies are migrating in the opposite direction to the bottom electrode, and trap the electrons. Thus, at the bottom electrode the population of oxygen vacancies is formed and is propagating to the top electrode, which causes the formation of tiny current filaments, hence providing conduction in the oxide and, as a consequence, the transition from an insulating to a conduction state (i.e., the transition from HRS to LRS). In the Reset process, the large densities of current filaments are cancelling oxygen vacancies via Joule heating: V_O²⁺ + O_i²⁻ → O_L. Therefore, we have the inverse transition from LRS to HRS. These processes are depicted schematically in Figure 2.

Figure 2. Schematic representation of the filament memristor and its states.

Soon after the discovery of the memristor, it was used as a high-frequency switch ^[8]. This was a logical step, since memristors have low R_ON (in the range of 10–100 Ω), while R_OFF is in order of tens or hundreds of MΩ. The oxide memristor is embedded in a planar waveguide termed coplanar waveguide (CPW), consisting in three metallic electrodes separated by two gaps, with the central conductor being the signal line and the outer electrodes being the ground planes; all the three electrodes are deposited on an insulator substrate. The CPW line has both input and output impedances of 50 Ω (standard reference impedance for microwave circuits). In the case of the microwave memristor switch described in ^[8], the substrate is a 525 μm-thick high-resistivity (HR, with resistivity of 1000 Ω·cm) silicon (Si) wafer having a 200 nm-thick silicon dioxide (SiO₂) layer grown over it, such that the memristor is isolated from the substrate containing surface charges. The memristor is a MIM-like device, having two dissimilar metallic electrodes: the bottom electrode is Ti (5 nm)/Pt (20 nm), while the top electrode is Ta (30 nm)/Pt (200 nm). The oxide is amorphous Ta₂O₅ having the thickness of 7 nm, with a high density of oxygen vacancies. This memristor is integrated in the central electrode of the CPW as it is shown in Figure 3. The Set pulse has an amplitude of +2 V and a duration of 105 ps, while the Reset pulse has an amplitude of +3.3 V and a duration of 120 ps. The ON- and OFF-state are evidenced by the sub-ns time response, meaning that the memristor works as a switch up to 20 GHz.

Figure 3. The microwave filament memristor in coplanar waveguide (CPW) technology: (a) cross-section; (b) top-view.

Further developed memristors based on monolayers, such as MoS₂ or h-BN, rely upon the concepts described above, i.e., the monolayer is embedded in the central conductor, the ON resistance is kept low (at 10–20 Ω), and the results show that the monolayer memristors are very good microwave switches, having low insertion loss (<2 dB) and high isolation (between −25 and −30 dB); moreover, they are able of high-power handling ^[9][10]. The first experimental results show that these memristors are filament memristors, even if they are monolayers.