Angle-Dependent Structural Color-Based CuO/ZnO Nanopatterns with Deep Learning Authentication

To prevent counterfeiting and piracy, various anti-counterfeiting technologies are used. In the past, labels such as holograms and watermarks have been used, but due to their high predictability, small electronic tags (RFID, QR code, etc.) or high-precision nanopatterns have been widely used in recent years. For these methods, complex processes such as MEMS, lithography, and high-precision writing are required to produce sophisticated patterns, which increases production costs and limits mass production. In this study, we propose a fabrication method for unpredictable nanopatterns through electrospinning, and independently depositing CuO and ZnO through a two-step solution process to provide patterns with high information density. We also demonstrate that the pattern has an excellent pattern discrimination accuracy of 97% through simple deep learning verification.<br/><br/>The fabrication process of nanopatterns is as follows: 1) fabrication of the base pattern by electrospinning, 2) deposition of CuO & ZnO materials. First, a polymer solution containing palladium ions is collected vertically between parallel aluminum electrodes and transferred to the target substrate. The aligned transferred nanofibers are periodic diffraction grating structures of one-dimensional (1D) order, which can exhibit optical effects by showing separated colors that vary significantly with the angle with respect to the input light. Here, palladium is used as an active catalyst for copper deposition. The aligned nanofibers transferred to the substrate undergo solvent decomposition and fiber stabilization steps by heat treatment at 500 °C. Subsequently, Cu nanofibers are obtained by Cu electroless plating. A polymer solution containing nitrate ions is then transferred to the target substrate using similarly parallel electrodes. A heat treatment of 500 °C results in the formation of a seed layer for ZnO growth. During this process, thermal oxidation of the already prepared Cu NFs takes place under an open atmosphere. The raman spectrum shows bands of CuO centered at 302, 338, and 624 cm-1, copper(II) oxide has been formed. ZnO hydrothermal synthesis is then carried out at 80 °C to obtain the final CuO/ZnO composite nanopatterns. By adjusting the ZnO hydrothermal synthesis time, the ZnO particle size and the diameter of the NFs can be controlled, which enables the fabrication of different structural colors. The fabricated patterns are applied to a deep learning-based discrimination model, which has an excellent pattern discrimination accuracy of about 97%. It was compared with the binary pattern and ZnO pattern, and the CuO/ZnO composite pattern showed a high angle dependence based on the structural color, making it suitable as a strong anti-counterfeiting pattern.<br/><br/>In this study, we proposed a fabrication method for unclonable anti-counterfeiting patterns that can be discriminated without specialized optical devices. The nanopatterns with artificially aligned grating structures, which are non-replicable through a simple setting during electrospinning, become the base pattern for structural color induction. Then, CuO/ZnO composite nanopatterns with two materials deposited by solution process are obtained. At this time, the structural color can be controlled by adjusting the ZnO synthesis time, and the mixing of CuO with high absorption in the visible light region can improve the color saturation and angle dependence of the system by reducing the incoherent scattering of the grating structure, thus ensuring high information density. The discrimination algorithm based on deep neural network supervised learning achieves a high pattern identification accuracy of about 97% for an area of at least 80 (mm²). It also proves that the pattern can be estimated with a probability of about 73.5% for untrained patterns, demonstrating that it is a practical anti-counterfeiting technology.