The effort made in recent years in the development of new memories has led to a significant advance in emerging technologies, which have proven their usefulness not only in the field of memories, but also to obtain devices that perform an artificial synapse and thus emulate biological neurons. Among the novel concepts, resistive-switching random access memory (RRAM), or memristive device, has attracted a great deal of interest for its ability to store multiple states. In fact, its operation is based on the so-called resistive switching, which consists of the formation of conductive filaments through a dielectric that separates two metal electrodes, allowing current to flow between them. The conductive filaments can be partially dissolved and re-formed, which explains the variation in conductivity and thus the existence of several different states, opening up possibilities for analog applications and brain-inspired computing systems. In the field of memories, two clearly differentiated states are used, called low resistance (LRS, when the filament is formed) and high resistance (HRS, when the filament is partially dissolved and does not completely join the two electrodes). The set process takes the structure to the LRS state, while the reset takes it to the HRS. Since the resistance of the memristor is not constant but depends on the previous states it has gone through, this device is non-volatile, i.e., it remembers its history [1].At the present time, when artificial intelligence-based applications are experiencing considerable momentum, there is a call for intensified research into memristive devices-based neuromorphic computing systems [2]. To achieve efficient, high-performance neuromorphic circuits, it is not enough to be able to switch the device resistance between two distinct values. In addition, the specific value of the conductivity at each time instant must be properly controlled so that the synaptic weights can be encoded. For this purpose, it is necessary to focus on the analysis of the dynamic evolution of the internal variables of the memristive devices, which necessarily implies a deeper understanding of the underlying physical processes that cause the variation of their resistance [3].In this work, a thermoelectric analysis of memristive devices is carried out, with special emphasis on time response and multilevel control. The studied structures are TiN/Ti/HfO2/W and TiN/Ti/HfO2/Pt/Cr metal-insulator-metal (MIM) capacitors, with several HfO2 -thickness values varying between 8 and 20 nm. Dielectric layers were atomic layer deposited (ALD). It is known that in the case of hafnium oxide layers sandwiched between these specific electrode materials, the resistive switching phenomenon is related to the distribution of oxygen vacancies inside the dielectric film, which occurs when oxygen atoms are drawn to the electrodes [4].Fig.1 shows the control of intermediate states in multilevel behavior, both in terms of I-V and small signals (conductance and capacitance measured at 500 kHz). The current evolution in the reset process after applying -0.8 V height-voltage pulse trains can be seen in Fig.2, which also shows the schematic of the experimental setup. On the other hand, Fig.3 depicts some results of the transient analysis; the amplitude of the current transients in both set and reset states have been superimposed on the external I-V loop. An example of the study of set and reset dynamics performed by applying variable speed voltage ramps is shown in Fig. 4, together with the experimental setup, this analysis has been carried out at different temperatures, between 77 and 350 K. All results shown in Figs. 1-4 correspond to a TiN/Ti/10 nm-HfO2/W structures, being TiN the top electrode and W the bottom one. Bias voltage is applied between top and bottom electrodes.The influence of the dielectric thickness on the resistive switching modes that the structures can exhibit is also addressed. Finally, some results on the physical and geometrical morphology of the conductive filaments, obtained from the analysis of the conduction mechanisms, are discussed.REFERENCES “Memristive devices for brain-inspired computing”. Edited by S. Spiga et al. Woodhead Publishing (2020). https://www.sciencedirect.com/science/article/pii/B9780081027820000253?via%3DihubLanza et al. Science 376, 6597 (2022). https://doi.org/10.1126/science.abj9979 García et al. J. of Phys. D: Appl. Phys. 56, 365108 (2023) https://iopscience.iop.org/article/10.1088/1361-6463/acdae0/meta C. Jasmin “Filamentary model in resisitive switching materials”. AIP Conf. Proc. 1901, 060004 (2017). https://doi.org/10.1063/1.5010507 Figure 1