Simulating permeabilities based on 3D image data of a layered nano-porous membrane

Reconstruction of highly porous structures from FIB‐SEM using a deep neural network trained on synthetic images

“…Left: stack of backscattered electron images. Centre: reference reconstruction -secondary electron and backscattered electron signals combined using reconstruction by dilation as described by Prill et al (2020); Ott et al (2019). Right: reconstruction result of the trained convolutional neural network.…”

“…Right: reconstruction result of the trained convolutional neural network. The accuracy of the segmentation is dice coefficient 0.88, false negative rate 0.06 and false positive rate 0.12. refer to Prill et al (2020). The finest layer of the nanoporous membrane analysed there was treated exactly the same way as the ZrO 2 sample at hand.…”

Three-dimensional structural comparison of tantalum glancing angle deposition thin films by FIB-SEM

“…For our structural comparison, this is not crucial as both samples are affected. However, this source of error has to be taken into account or corrected for, e.g., when simulating macroscopic material properties (Prill et al, 2019).…”

“…For structures finer than 100 nm, FIB-SEM is the imaging method of choice. For recent reviews of this 3-D imaging technique as well as microstructure characterization and simulation for porous ceramics we refer to Holzer et al (2016); Penner and Holzer (2018); Prill et al (2019). However, in particular for highly porous materials, FIB-SEM causes several typical imaging artifacts, see Bassim et al (2012), Joos et al (2011), Tomus and Pang (2013), Prill et al (2013) and Fitschen et al (2017).…”

“…Thus, the segmentation of highly porous structures is not straightforward. Several segmentation algorithms (Prill et al, 2013;Prill et al, 2019;Salzer et al, 2015) are dedicated to overcoming these artifacts. Here, we apply a new approach to deal with them.…”