Real-Time Control of Sputtering Parameters to Achieve Stable Vanadium Oxide Thin Films with Consistent Electrical and Optical Properties

Vanadium oxide (VOx) possesses various oxidation states, each exhibiting unique electrical and optical properties that have made it a subject of research for decades, with potential applications in smart windows, batteries, catalysts, and various sensors. Particularly, VO2 exhibits metal-insulator transition (MIT) characteristics at 60-70 degrees Celsius, promising diverse applications. However, within the range of x = 1.5 to 2.5, there are more than 15 oxidation states, which can easily lose their properties due to minor process variations. Thus, maintaining a consistent oxidation state and crystal structure presents a significant challenge. Even if the entire thin film maintains a consistent oxidation state and crystal structure, limiting the second phase that occurs during MIT is another major challenge. Due to these characteristics, it is essential to monitor the plasma state in real time and adjust the process variables to create a stable growth environment during the growth process of vanadium oxide thin films.<br/> <br/>The primary equipment used to observe the sputtering process is Optical Emission Spectroscopy (OES). OES operates during plasma process to measure the intensity of light in the 200-1000 nm wavelength range. Process variables include working pressure, sputtering power, and the distance between the sample and target, while the preparation process before and after sputtering is consistently maintained. Various combinations of sputtering process variables are used to analyze changes in the OES results, defining the relationship between each variable and the overall spectrum. Based on this, a controller is developed to ensure the spectrum remains consistent. The VO₂ thin film is subjected to resistance measurements in response to temperature changes to determine the total resistance change, maximum resistance change rate, and transition temperature according to MIT. A multilayer neural network is constructed to examine the relationships between the complete spectrum data and these results. The goal is to adjust the process variables based on the real-time observed OES spectrum to achieve a thin film with consistent performance.