Improvement of 3D atom probe tomography reconstruction integrating transmission electron microscopy information

Abstract

We propose to improve atom probe tomography (APT) reconstruction methods to increase the reliability and the accuracy of the resulting 3D volumes. Currently, 3D reconstructions are done iteratively atom by atom, to convert the ion sequence ( #N ) and the detected position ( X D , Y D ) into atomic coordinates ( x,y,z ). To obtain a more realistic reconstruction, a better estimation of the main reconstruction parameters (e.g. image compression factor, detection efficiency, evolution of the tip radius) is required [1]. Some techniques have already been developed in order to estimate the reconstruction parameters. Those calibrations are based on voltage curve, tip image, field ion microscopy or desorption images. Most of these techniques have been shown to give accurate results on homogenous or metallic materials (in this case, the crystallographic pole or atomic planes can easily be observed) [2]. Unfortunately, these techniques are poorly suited for semiconductor materials, since the projection laws and the physics of field evaporation affect the reconstruction parameters, especially in the case of multiphase materials where evaporation fields are different from one phase to another. In this study, we aim to correlate the information obtained by transmission electron microscopy (TEM) with atom probe tomography to improve the current reconstruction model. We focused on high‐K multilayer materials where the evaporation field is very different between each layer. As we can see on figure 1, the original sample contains flat and chemically well‐defined interfaces, whereas those interfaces are strongly distorted (figure2.a) when a classic reconstruction algorithm (with cone angle consideration) is used. This distortion comes from the assumption that two atoms arriving one after the other on the detector were close to one another in the tip, which is not the case in wide field of view APT. In our model, the reconstruction is performed by dealing separately with different (and previously defined) areas of the detector surface. The ions are labeled independently in every sub‐detector, allowing different reconstruction parameters for each subdetector and therefore a better match of the theoretical thickness of every layer. The new reconstruction (figure 2.b) shows homogenous atom density, flat interfaces and more accurate layer thicknesses, making the chemical quantification more reliable. This study could be extended to 3D objects, where the microstructural features observed in TEM could be used to constrain the resulting APT volume. This work has been funded by the French ANR Recherche Technologique de Base (RTB) programme. The experiments were performed on the Nanocharacterisation platform (PFNC) at MINATEC.