利用混合态分解算法提高重叠关联成像的质量

X射线重叠关联成像技术所采用的常规相位恢复迭代算法均把入射光视为完全相干的单色光(即单态入射光),把样品视为稳定平衡的(即单态样品),把实验过程中成像系统内所有光学元器件的响应视为固定不变的。然而在实际实验中,上述三个条件很难得到同时保证。在非理想的实验条件下,任何单态条件发生变化或者出现多重态混合,其造成的退相干效应均会使衍射图样变得模糊,严重降低利用单态相位恢复迭代算法获得的重建图像质量。为了消除各种来源的混合态效应的负面影响,本文利用多模式拓展重叠关联迭代引擎算法(Mm-e PIE)和子像素上采样拓展重叠关联迭代引擎算法(Us-e PIE)这两种不同的混合态分解算法对实验数据进行处理。重建结果表明,不同来源的混合态效应(退相干效应)的分解结果在重建探针函数和重建样品函数中均得到了体现,两种混合态分解算法均能够显著改善重建图像的质量。其中多模式算法Mm-e PIE在消除入射探针的非相干效应方面更胜一筹。

Improving the reconstructed image quality in ptychography using mixed states decomposition iterative phase retrieval algorithms

Background:In most commonly used iterative phase retrieval algorithms in X-ray ptychography, the incident X-ray is considered to be fully coherent and monochromatic (i.e. in single state), the sample is considered to be stable or unchanged, and the responses of all optical elements in the system are considered to be fixed throughout the data acquisition process. However, in actual experiments, the above three conditions cannot be guaranteed simultaneously. Under non-ideal conditions, decoherence effects from single state changing or states mixture could severely deteriorate the reconstructed image quality.Purpose:This study aims to improve the reconstructed image quality using mixed states decomposition iterative phase retrieval algorithms to eliminate the negative influences of various types of decoherence effects. Methods:Two mixed state decomposition algorithms, the multi-mode extended ptychographical iterative engine (Mm-ePIE) algorithm and the sub-pixel up-sampling extended ptychographical iterative engine (Us-ePIE) algorithm are used to reconstruct the sample function and the probe function(s) simultaneously. Results:The results show that the decomposition effect of state mixture from different sources is embodied in both the reconstructed sample function and probe function(s). Conclusion:These two mixed-state decomposition phase retrieval algorithms can both improve the quality of the reconstructed image significantly. Moreover, the multi-mode algorithm performs better than the up-sampling algorithm in removing the decoherence effects.