In recent years, nanoionics-based memristive devices with resistive switching behaviors have garnered significant attention as promising candidates for next-generation nonvolatile memory known as resistance random access memory (RRAM) due to their potential high scalability and low-power consumption. In addition, memristive devices are also promising for logic and neuromorphic computations. The memristive devices typically have a simple stacked structure of metal-insulator-metal (MIM), where the resistance can be reversed changed through local ion migration and electrochemical reaction after an initial electroforming process. Nanoionics–based memristive devices can be classified as cation migration cells and anion migration cells. When an electrochemically active electrode such as one made of Ag or Cu is used in a MIM device, resistive switching occurs through formation and annihilation of a metallic atom bridge resulting from the migration of highly mobile cations such as Ag+ and Cu+ ions. This type of nanoionics-based device is referred to as an atomic switch or electrochemical metallization systems.[1] Resistive switching behavior can also be observed for a MIM device without an electrochemically active electrode. In this case, the migration of anions, usually oxygen ions plays a crucial role. Local oxygen ion migration, which is better described as the migration of oxygen vacancy (V O ‥) and a resultant change in the electronic barrier at the metal electrode/TMO interface, is associated with the resistive switching behavior. In general, the cation migration cell and anion migration cell both need electroforming process to obtain resistive switching behaviors. For the cation migration cells, the formation of Ag or Cu bridges in the forming process has been directly observed by TEM or in-situ TEM.[2]In contrast, the mechanism of the electroforming process in the anion migration cell is still uncertain due to the lack of direct characterization. In this study, the electroforming process and resistive switching behaviors of the anion migration cell of Pt/WO3-x/Pt have been investigated as a function of temperature. As shown in Figure 1a, electroforming process with soft-breakdown was triggered as the sweep voltage was raised to about 6.0 V at 296 K and 396 K. But, at 110 K, no soft-breakdown was observed even by increasing the sweep voltage to 20 V, which indicates that the devices cannot be electroformed at ultralow temperature. In contrast, the device shows resistive switching behaviors at ultralow temperature after being electroformed at room temperature, see Figure 1b. In order to real-time observe the structure changes during electroforming process, we prepared a Pt/WO3-x/Pt devices that operated inside a TEM, see the inset of Figure 1g. As reported in Figure 1c to 1f, the initial amorphous WO3-x layer gradually crystallized by applying voltage weeping. And, crystallized nanoparticles of Magnéli phase were observed, as shown in Figure 1g. Actually, the current level of in situ measurements is much smaller than that in the conventional electroforming processes. This may be the reason why no conductive filament composed of the Magnéli phases was observed. The crystallization and formation of the conductive filament in WO3-x layer have been observed by cutting the electroformed device in our previous work.[3] Based on above results, it is proposed that the Joule heating effect plays a key role in the electroforming process. Under ultralow temperature, the current level of the device is too low to generate enough heating effect to introduce the filament formation. Moreover, the heat dissipates much quickly at ultralow temperature. Therefore electroforming process could not be triggered at ultralow temperature. In contrast, the resistive switching behavior obtained after electroforming process mainly occurs in the nanogap region between the electrode and the conductive filaments, where the V O ‥ migration under external electric field determines the resistive switching process. So, the resistive switching behavior can be observed at low temperature. Based on these results and discussions, it is reasonable to conclude that Joule heating effect domains the electroforming process, instead of the resistive switching processes. Terabe, K.; Hasegawa, T.; Nakayama, T.; Aono, M., Nature 2005, 433, 47-50. Yang, Y.; Gao, P.; Gaba, S.; Chang, T.; Pan, X.; Lu, W., Nat.Commun. 2012, 3, 732. Tan, Z. H.; Yang, R.; Terabe, K.; Yin, X. B.; Zhang, X. D.; Guo, X., Adv. Mater. 2016, 28, 377-384. Figure 1