Physics-based reconstruction algorithm for 3D atom probe microscopy

原子探针断层扫描代表了一种具有出色分辨率的3D材料分析方法，该方法遵循一种不寻常的功能原理：单个原子作为离子从样品中解吸，在强电场中加速，由空间分辨检测器装置检测，并通过其飞行时间进行识别。在测量之后，通过数值重建生成原子空间排列的3D图。到目前为止，已建立的重建算法是基于一个简单的几何投影。这导致不可接受的伪影，并将空间定位的精度限制在约1纳米。该项目的主要目标是设计和探索一种新的重建方法，该方法基于对离子轨迹的物理精确计算。在该项目的第一阶段，对精确计算特定样品几何形状的电场和离子轨迹，然后将其与检测器事件的位置进行匹配的原始想法进行了测试。在这些研究的过程中，推导出了一个确定原子位置的附加原理。 现在，这一原则应测试不同的样品几何形状。应优化重建的剩余参数，以便重建正确的样品形状和微观结构。应提高程序的准确性。甚至原子晶格结构的测量是否成为可能，也应该加以研究。算法的数值效率应得到充分的提高，以便重建可以在小型多内核CPU卡上处理。最后，应开发一个软件，使实验数据集的实际处理。

Atom probe tomography represents a method of 3D material analysis in outstanding resolution that follows an unusual functional principle: Single atoms are desorbed from the sample as ions, accelerated in strong electric fields, detected by a spatially resolving detector setup, and identified by their time of flight. Subsequent to the measurement, a 3D map of the spatial arrangement of the atoms is generated by a numeric reconstruction. Up to now, the established reconstruction algorithms are based on a simple geometric projection. This leads to unacceptable artifacts and limits the accuracy of spatial positioning to about 1 nanometre. The primary goal of the project is the design and exploration of a new reconstruction method that is based on the physically exact calculation of the ion trajectories. In a first phase of the project, the original idea of exactly calculating the electrical fields and the ion trajectories for a particular sample geometry and then matching them with the positions of the detector events was tested. In the course of these investigations, an additional principle was derived to determine the positions of the atoms unambigously. Now this principle shall be tested with different sample geometries. Remaining parameters of the reconstruction shall be optimized so that the correct sample shape and microstructure is reconstructed. The accuracy of the procedure should be improved. It should be investigated whether even the measurement of the atomic lattice structure becomes possible. The numerical efficiency of the algorithms should be sufficiently improved so that the reconstruction can be processed on small multi-kernel CPU cards. Finally a software should be developed that allows the practical processing of experimental data sets.