Compensation of thermal drift effects in atomic force microscopy using probabilistic state estimation

二十多年来，原子力显微镜（AFM）在生命科学、生物学、材料科学、半导体工业、微观和纳米技术等各个学科中都是必不可少的仪器。它通过在微、纳米和原子尺度上提供高分辨率的结构可视化，彻底改变了表面分析。虽然AFM最初仅用于对样品的形貌进行成像，但随着时间的推移，不同的操作模式不断发展，包括电、磁和化学测量。对比传统光刻工艺提供更高分辨率的新型光刻方法的需求，以及能够加工新材料（如生物材料）的需求，推动了基于afm的光刻领域的研究。在利用原子力显微镜尖端操纵单个微/纳米实体方面也进行了深入的研究。这种基于原子力显微镜的组装有潜力的应用是基于碳纳米管、DNA、单石墨烯层、纳米线等纳米材料的新型纳米电子器件和系统的原型制作。影响原子力显微镜几乎所有应用领域的一个问题是热漂移的存在。热漂移源于AFM（或一般的SPM）不同组件的温度变化和热膨胀系数的差异。这将导致探针相对于样品在所有三个维度上的未知时变位移。这种运动通常是缓慢的，但它会导致图像和光刻过程的扭曲，伪造光谱结果，并危及纳米操作的成功。在这个项目中，我们的主要目标是开发一种灵活的漂移补偿系统，适用于成像，光谱学以及纳米操作任务。所开发的方法应该允许在图像采集和实时操作过程中对热漂移进行主动补偿，而无需事先了解样品性质或漂移状态。此外，它应该补偿所有三个维度的漂移，从而不限制应用领域。为了达到这些目标，将开发一种概率算法，结合非光栅扫描方法，描述AFM地形数据的一般有效模型以及描述漂移的概率模型。该算法将基于贝叶斯滤波，以纳入不准确模型引入的不确定性。该系统将在不同的系统上进行实验验证，使用定制的AFM控制架构，在基于AFM的不同纳米物体自动操作的背景下进行。

Since more than two decades, the atomic force microscope (AFM) represents an essential instrument in various disciplines covering life science, biology, material science, semiconductor industries, and micro- and nanotechnology in general. It has revolutionized surface analysis by providing high-resolution visualization of structures at micro-, nano-, and atomic scales. While AFM was first only used for imaging the topography of a sample, different operation modes evolved over time including electrical, magnetic, and chemical measurements. The need for novel lithographic methods providing higher resolution than conventional lithographic processes and enabling to process new materials (e.g. biomaterials) have motivated research in the domain of AFM-based lithography. Intense research has also been carried out on the manipulation of individual micro-/nanoentities with the AFM tip. Promising potential applications of such AFM-based assembly are e.g. prototyping of novel nanoelectronic devices and systems based on nanomaterials such as CNTs, DNA, single-graphene layers, nanowires, etc. A problem that is affecting almost every application area of AFM is the presence of thermal drift induced. Thermal drift originates from small changes in temperature and differences in the coefficient of thermal expansion of the different components of the AFM (or SPM in general). This results in an unknown, time-variant displacement of the probe relative to the sample in all three dimensions. This motion is generally slow, but it is causing distortions in images and lithographic processes, falsifying spectroscopy results, and compromising the success of nanomanipulations. In the proposed project our major objective is the development of a flexible drift compensation system applicable in the context of imaging, spectroscopy as well as nanomanipulation tasks. The developed methods should allow for an active compensation of thermal drift during image acquisition and manipulation in real-time without any prior knowledge of the sample properties or drift state. Moreover, it should compensate drift in all three dimensions thus not limiting the field of application. To reach these objectives a probabilistic algorithm will be developed incorporating non-raster scanning methods, generally valid models describing AFM topography data as well as a probabilistic model describing drift. The algorithm will be based on Bayesian filtering to incorporate the uncertainties introduced by inaccurate models. The system will be experimentally validated on different systems using a custom-made AFM control architecture in the context of AFM-based automated manipulation of different nanoobjects.