Vanadium dioxide (${\mathrm{VO}}_2$) exhibits an insulator-to-metal transition (IMT) above room temperature, which makes it an ideal candidate for resistive-switching- (RS) based artificial neural networks (ANNs). To enhance the viability of RS-based ANNs, it is beneficial to operate ${\mathrm{VO}}_2$ devices close to the transition temperature, minimizing the switching power required to induce the IMT. Therefore, to avoid spontaneous switching and enable energy-efficient operation, it is important to develop a thorough understanding of thermal effects and heat propagation through these systems. In this work, we investigate the RS behavior of electronically and thermally coupled ${\mathrm{VO}}_2$ multibridge devices grown on substrates with distinct thermal and crystalline properties. Multiple switching states are observed independently of the substrate material. IR-emissivity mapping is used to study switching events and locally probe the temperature. Our results show that the substrate with lower thermal conductivity enhances thermal crosstalk, which decreases the power needed to induce the IMT in thermally coupled neighboring devices. The present study shows that the substrate choice can play an important role in the energy efficiency of IMT-based devices for ANN applications.

Close

PrevNext

View figure in article

• Heat transfer
• Metal-insulator transition
• Resistive switching
• Thermoelectric effects
• Artificial neural networks
• Transition metal oxides
• Electrical techniques
• Infrared spectroscopy