Thanks to the non-linear nature of laser-matter interaction, the use of femtosecond lasers offers a versatile method for encoding information and modifying transparent materials in their volumes and this, with sub-micron resolution. The underlying physical process is a succession of intricate and complex nonlinear phenomena that are sensitive to multiple and multidimensional parameters, such as beam intensity distribution, exposure dose homogeneity and pulse-overlapping sequences as well as propagating wavefront angular orientations and temporal distortions. As a consequence to this inherent and often overwhelming complexity, obtaining a repeatable and accurate result relies strongly on time-consuming machine-specific calibration and experiment-specific fine-tuning attempts until the desired result is reached. Here, we present a digital twin of the processed specimen that not only accurately predicts the exposure outcome in terms of introduced retardance, but also offers a pathway for designing feedforward schemes that compensate for known inaccuracies. We demonstrate the merit of this approach through illustrative examples of arbitrary phase patterns, forming waveplates and images, based on refractive index modulation induced during laser exposure.

由于激光与物质相互作用的非线性特性，飞秒激光器的使用提供了一种通用的方法来编码信息和修改其体积中的透明材料，并具有亚微米分辨率。潜在的物理过程是一系列对多维和多维参数敏感的复杂非线性现象，例如光束强度分布、曝光剂量均匀性和脉冲重叠序列以及传播波前角方向和时间失真。由于这种固有的且通常是压倒性的复杂性，获得可重复和准确的结果在很大程度上依赖于耗时的特定于机器的校准和特定于实验的微调尝试，直到达到所需的结果。这里，我们展示了一个处理过的样本的数字孪生模型，它不仅可以根据引入的延迟准确预测曝光结果，而且还提供了一种设计前馈方案以补偿已知误差的途径。我们通过任意相位模式的说明性示例展示了这种方法的优点，基于激光曝光期间诱导的折射率调制，形成波片和图像。