It was recently discovered that covalently bound impurity atoms in crystal lattices can be manipulated with focused electron irradiation, unveiling new perspectives for top-down atomic engineering. This has been made possible by advances in electron optics and instrument stability, but also in the incorporation of impurity elements into the materials using techniques such as ion implantation. We have demonstrated precise control over the movement of silicon impurities in graphene [1], with a manipulation rate already nearly on par with any atomically precise atom manipulation technique. Phosphorus can also be moved with the beam, though this appears to be significantly more challenging [2]. Such manipulation is also possible in large-diameter single-walled carbon nanotubes [3], and we are using ab initio modeling to describe corresponding mechanisms in bulk silicon.

Underlying our understanding of such dynamics are computationally demanding ab initio molecular dynamics simulations of displacement threshold energies [4] combined with a quantum mechanical description of lattice vibrations that allows us to quantify the stochastic ejection of single atoms. In the case of pristine graphene, this has revealed that its electron-beam damage is fully described by knock-on effects and that there appears to be no dose-rate effect at typical TEM or STEM current densities [5]. Precise quantification of the damage cross section of graphene consisting of either of the two stable carbon isotopes further allowed us to map the isotope concentration in selected nanoscale areas of a mixed sample in a technique that might be called a `mass spectrometer in a microscope' [5]. However, accurate quantitative predictions do not appear yet to be possible for graphene impurities [6], and other 2D materials including hBN and MoS2 clearly cannot be described without including electronic excitations [7].

We have also developed a highly efficient ab initio method based on projector-augmented waves described on a real-space grid for the modeling of electron scattering factors using full electrostatic potentials including valence electron bonding. In our benchmarking, the method is as accurate as reference Wien2k calculations, but at a fraction of the computational cost [8]. We are currently applying this to a twisted bilayer graphene model with 1320 atoms relaxed with explicit van der Waals interactions. Using the resulting charge density, we can calculate convergent beam electron diffraction patterns to help interpret 4D-STEM data collected by our collaborators. These simulations are an order of magnitude larger than what has been possible before at this level of theory, and can plausibly be increased by a further order of magnitude within the current approach.

- [1] M. Tripathi et al., Nano Letters 18 (2018) 5319
- [2] C. Su et al., Science Advances 5 (2019) eaav2252
- [3] K. Mustonen et al., Advanced Functional Materials (2019) 1901327
- [4] T. Susi et al., Physical Review Letters 113 (2014) 115501
- [5] T. Susi et al., Nature Communications 7 (2016) 13040
- [6] T. Susi et al., 2D Materials 4 (2017) 042004
- [7] T. Susi et al., Nature Reviews Physics 1 (2019) 397
- [8] T. Susi et al. Ultramicroscopy 197 (2019) 16