AI-Optimized Vanadium Oxide Multilayers Deliver More Than 20-Fold Improvement in Infrared Sensor Performance
Date: 11 Mar 2026 |
Source: Published in Advanced Science
Researchers at UNIST have demonstrated a major advance in infrared sensing technology using artificial intelligence to optimize vanadium oxide materials, achieving more than a 20-fold improvement in bolometric performance. The breakthrough was reported in the journal Advanced Science.
Infrared bolometers are widely used in thermal imaging cameras, night-vision systems, environmental monitoring, and industrial inspection. Many of these devices rely on vanadium oxide (VOx) due to its strong temperature sensitivity. However, achieving both high sensitivity and stable, linear electrical response has long been a challenge for conventional materials.
Vanadium dioxide (VO₂), a well-known phase-transition material, exhibits a very high temperature coefficient of resistance (TCR), making it highly sensitive to temperature changes. However, its non-linear and hysteretic electrical behavior has limited its application in high-performance infrared bolometers.
To overcome these challenges, researchers developed an AI-optimized multilayer architecture based on tungsten-doped vanadium oxide (WxV₁-xOᵧ) thin films. By stacking layers with different doping ratios and using machine-learning algorithms to optimize the structure, the team engineered materials that combine high temperature sensitivity, reduced hysteresis, and improved linear response.
Figure 1. Schematic illustration of the AI-optimized multilayer vanadium oxide structure composed of tungsten-doped WxV₁-xOᵧ thin films with varying compositions, enabling tailored electrical and thermal properties for infrared bolometer applications.
Source: Choi et al., Advanced Science (2026).
The optimized multilayer system simultaneously achieves high TCR and low electrical noise under CMOS-compatible fabrication conditions, making it suitable for integration into modern semiconductor manufacturing processes.
As a result, the researchers demonstrated bolometric performance up to 23.6 times greater than conventional designs, representing a significant step forward for infrared sensing technology.
Figure 3. Performance comparison showing the bolometric performance of the AI-optimized vanadium oxide multilayer device (“This work”) relative to previous materials and device architectures.
Source: Choi et al., Advanced Science (2026).
The study highlights the continuing importance of vanadium-based functional materials in advanced electronics and sensing technologies. By combining machine learning with vanadium oxide materials engineering, the work provides a promising pathway toward next-generation infrared detectors with improved sensitivity, stability, and manufacturability.
Such advances could benefit a wide range of applications, including autonomous systems, security imaging, industrial monitoring, environmental sensing, and scientific instrumentation.
Publication
Choi et al., AI-Optimized Vanadium Oxide Multilayers for More Than 20-fold Enhancement in Bolometric Performance, published in Advanced Science (2026).