Oxide semiconductor transistors could be on the verge of a major technological breakthrough after researchers introduced a digital twin platform designed to solve one of the most persistent problems in electronic materials: improving performance without sacrificing long-term reliability. The platform, described by T.H. Pantha, M. Lee, K. Bhojak and colleagues in Communications Engineering, uses computer-based replicas of transistor devices to accelerate the search for manufacturing conditions and material designs that would otherwise require years of laboratory testing.
Oxide semiconductors have become increasingly important because they can be fabricated at relatively low temperatures, deposited over large areas and integrated with glass, plastic or other unconventional substrates. These properties make them attractive for display backplanes, sensors, flexible electronics, transparent circuits and emerging computing systems. Yet the same materials that offer these advantages often face a difficult trade-off. Increasing the operating current or switching speed can improve transistor performance, but it may also increase electrical instability, accelerate degradation and shorten the useful lifetime of the device.
The new platform addresses this challenge by creating a “digital twin”—a computational model that mirrors the behavior of a physical transistor. Unlike a conventional simulation based only on idealized equations, a digital twin can be linked to experimental measurements and updated as new data become available. In principle, the model can learn how processing conditions, device architecture, electrical stress and environmental factors influence transistor behavior. Researchers can then use it to predict which combinations are most likely to deliver high performance while maintaining reliability.
At the heart of an oxide semiconductor transistor is a thin active layer that controls the movement of charge between two electrodes. A gate electrode regulates this current through an electric field, allowing the device to switch between conducting and non-conducting states. In oxide materials, however, defects and imperfections can strongly influence how charges move. Oxygen vacancies, trapped charges, chemical impurities and structural disorder may create localized electronic states that alter the transistor’s threshold voltage, mobility and resistance. These effects can become more pronounced when a device is operated continuously or exposed to heat and moisture.
The performance-reliability trade-off emerges because many strategies that make a transistor faster or more conductive can also make it more vulnerable to instability. Increasing carrier concentration, modifying the semiconductor composition or adjusting the dielectric interface may improve current flow, but these changes can introduce additional defect pathways. Under prolonged electrical stress, charge may become trapped at interfaces or within the semiconductor, causing the transistor’s operating characteristics to drift. For circuits containing millions of devices, even small changes can produce visible image artifacts, inaccurate sensor readings or computational errors.
Instead of treating optimization as a sequence of isolated experiments, the digital twin approach is intended to connect manufacturing, physics and reliability analysis within a single framework. Experimental devices provide the platform with data, while physical models help explain why a transistor behaves in a particular way. Machine-learning methods can then identify patterns across large datasets and estimate the likely outcome of untested designs. This combination could allow researchers to prioritize the most promising experiments rather than exploring every possible material composition, layer thickness, annealing condition or operating voltage.
Such a system could be particularly valuable for oxide semiconductors because their behavior depends on many interacting variables. The composition of the active layer, the choice of electrode materials, the properties of the insulating layer and the conditions used during deposition can all affect the final device. Temperature, oxygen availability and post-processing treatments may change the concentration of defects or modify interfaces between layers. A digital twin can potentially represent these relationships in real time, allowing researchers to distinguish between a transistor that is initially powerful and one that can maintain its performance after thousands of hours of operation.
The researchers’ approach also points toward a more efficient model of materials development. Traditional optimization often depends on fabricating a large number of samples, measuring them individually and repeating the process after each adjustment. This cycle can consume substantial amounts of time and material, particularly when reliability testing requires prolonged stress experiments. A predictive platform could reduce this burden by combining existing measurements with accelerated testing and statistical modeling. It may also reveal hidden relationships that are difficult to identify through manual comparison, such as a processing condition that slightly reduces peak performance but dramatically improves stability.
The implications extend beyond displays, where oxide thin-film transistors are already widely used. More reliable and better-optimized devices could support flexible medical sensors, wearable electronics, large-area imaging systems, environmental monitors and transparent circuitry. Oxide transistors may also play a role in electronic systems that must be manufactured over extensive surfaces or integrated with materials that cannot withstand the high temperatures used for conventional silicon processing. If digital twins can shorten development cycles, they could help researchers move these technologies from laboratory demonstrations toward scalable manufacturing.
The platform does not eliminate the need for physical experiments; instead, it is designed to make those experiments more informative. Predictions must still be tested against real devices, and the quality of the digital twin depends on the quality, diversity and accuracy of the data used to build it. Complex failure mechanisms may also emerge only after long-term operation or under unusual environmental conditions. Nevertheless, the work represents a broader shift in electronics research: from optimizing devices one variable at a time to using connected digital models that learn from fabrication, operation and failure. By bringing these elements together, the researchers aim to overcome a limitation that has constrained oxide semiconductor development for years—and potentially make faster, more durable electronics possible without forcing engineers to choose one advantage at the expense of the other.
Subject of Research: Digital twin-based optimization of oxide semiconductor transistors for improved performance and reliability
Article Title: Digital twin platform for accelerating optimization of oxide semiconductor transistors to overcome fundamental performance-reliability trade-off
Article References: Pantha, T.H., Lee, M., Bhojak, K. et al. “Digital twin platform for accelerating optimization of oxide semiconductor transistors to overcome fundamental performance-reliability trade-off.” Communications Engineering (2026). https://doi.org/10.1038/s44172-026-00760-y
Image Credits: AI Generated
DOI: 10.1038/s44172-026-00760-y
Keywords: oxide semiconductors, thin-film transistors, digital twins, machine learning, transistor reliability, electronics optimization, flexible electronics, materials science
