Opt Express. 2021 Oct 25;29(22):35414-35425. doi: 10.1364/OE.434846.
The fabrication of complex integrated photonic devices via direct laser writing is a powerful and rapidly developing technology. However, the approach is still facing several challenges. One of them is the reliable quantitative characterization of refractive index (RI) changes induced upon laser exposure. To this end, we develop a tomographic reconstruction algorithm following a modern optimization approach, relying on accelerated proximal gradient descent, based on intensity images only. Very recently, such algorithms have become the state of the art in the community of bioimaging, but have never been applied to direct laser written structures such as waveguides. We adapt the algorithm to our concern of characterizing these translation-invariant structures and extend it in order to jointly estimate the aberrations introduced by the imaging system. We show that a correct estimation of these aberrations is necessary to make use of data recorded at larger angles and that it can increase the fidelity of the reconstructed RI profiles. Moreover, we present a method allowing to cross-validate the RI reconstructions by comparing en-face widefield images of thin waveguide sections with matching simulations based on the retrieved RI profile.
通过直接激光写入来制造复杂的集成光子器件是一种强大且快速发展的技术。然而，该方法仍然面临着一些挑战。其中之一是可靠地定量表征激光辐照引起的折射率 (RI) 变化。为此，我们开发了一种基于强度图像的基于加速近端梯度下降的现代优化方法的层析重建算法。最近，此类算法已成为生物成像领域的最新技术，但从未应用于直接激光写入结构，如波导。我们将该算法应用于我们对这些平移不变结构进行特征描述的关注，并对其进行扩展，以便联合估计成像系统引入的像差。我们表明，正确估计这些像差对于利用在更大角度记录的数据是必要的，并且可以提高重建 RI 分布的保真度。此外，我们提出了一种通过将薄波导部分的共焦宽场图像与基于所获取 RI 分布的匹配模拟进行比较来交叉验证 RI 重建的方法。
