Bayesian optimization of laser processes to maximize structural color gamut

Laser-induced printing is a fast, low-cost, and contactless method for producing high-resolution images on thin films containing metallic nanoparticles. While it enables visual effects and color rendering, its color gamut remains narrower than that of inkjet printing, mainly due to limited saturation and incomplete sRGB hue coverage. To address this, laser parameters, such as scan speed, power, repetition rate, and polarization must be precisely tuned. The color prediction being extremely complex and tedious, the preferred strategy is to build a parameter-to-color database by printing multiple samples under varying conditions and measuring outcomes. The manual tuning of parameters to obtain optimal colors is highly sensitive to sample variability. In this paper, we propose to replace an existing method with genetic algorithm by a novel Bayesian optimization approach to find the optimal laser parameters with the following advantages: simpler as reformulating the problem as multiobjective is not needed, less costly in laser inscription to reach the optimal gamut, and has a better gamut at fixed inscription number.