Overview#
Computer simulations have become an indispensable part of modern research. To reliably interpret the data from experiments, a careful comparison between theoretical models and experiments is often required. The use of dynamical scattering simulations based on the multislice algorithm has, for several decades, been an invaluable aid for transmission electron microscopy (TEM). As a result, there are many excellent simulation codes, and abTEM is another one; however, our approach differs in many key aspects.
How abTEM differs#
No GUI; Jupyter notebooks#
abTEM does not implement a graphical user interface, instead focusing on scripts and especially
Jupyter notebooks. A Jupyter Notebook is a mix of code, explanatory text, and visualizations, which
immediately provides documentation of the simulation, allowing others to understand, share and reproduce it. abTEM
aids a visual workflow by implementing methods for creating quick visualizations of each subtask, for example, a heatmap
of the projected potential or the profile of the electron probe. The use of open formats improves reproducibility
and transparency, and we always encourage users to share their code with publications.
Flexible, no fixed simulation modes#
abTEM does not directly implement any of the common TEM experiments, instead inviting the user to mix and match objects to construct the desired simulation. This provides the user with enormous flexibility to design the simulation while abTEM takes care of the computational details. See below for examples.
“Pure Python”, but fast#
Python has become the most popular programming language in physics, thanks in particular to its extensive base of open-source libraries. In the electron microscopy community, this includes packages such as HyperSpy, Libertem and Py4DSTEM. Effective use of fast open-source numerical libraries such as NumPy, CuPy, Numba and PyFFTW, allows abTEM to be extremely competitive in raw performance. The combination of easy-to-understand Python code, excellent interoperability combined with uncompromising simulation speed make for an attractive combination.
Deploys anywhere, powered by Dask#
abTEM is created from the ground up to work with Dask. Dask allows scaling from a single laptop to hundreds of nodes at high-performance computing (HPC) facilities with minimal changes to the code. This makes the code very well suited for diverse settings, from teaching to cutting-edge research.
Open-source and open to contributions#
The abTEM developers are not alone in seeing the necessity of open-source software in electron microscopy. We are striving to make this an inviting project for new contributors and collaborators. Everybody is invited to contribute to use and develop the code; see our guide for contributors.
The abTEM object model#
abTEM
follows domain-driven and object-oriented
design principles. In short, that means that our implementation of the multislice algorithm is based on objects
representing physical concepts; for example, a plane wave function is represented by the PlaneWave object, and a
focused STEM probe wave function is represented by a Probe object. This makes the code easy to understand, not only for
specialists but anyone familiar with transmission electron microscopy.
To simulate a (basic) high-resolution TEM (HRTEM) image, we need a PlaneWave at a given energy (by default in units of
\(\mathrm{eV}\)), a Potential with a given real-space sampling (in units of \(\mathrm{Å}\)), and a CTF (contrast
transfer function) representing the objective lens with a given aperture (in \(\mathrm{mrad}\)) and spherical aberration
coefficient (also in \(\mathrm{Å}\)). We show how this could be implemented in the code below; please change the tabs for
other simulation modes.
An image is measured by getting the intensity of the exit wave, calculated by propagating the plane wave through
the potential using the multislice algorithm and applying the CTF.
wave = PlaneWave(energy=80e3)
potential = Potential(atoms, sampling=.05)
ctf = CTF(semiangle_cutoff=30, Cs=1e-3 * 1e10, defocus="scherzer")
exit_wave = wave.multislice(potential)
measurement = exit_wave.apply_ctf(ctf).intensity()
We can turn the HRTEM simulation into a selected area electron diffraction (SAED) simulation by getting rid of the CTF
and calculating diffraction_patterns from the exit wave instead of the intensity.
wave = PlaneWave(energy=80e3)
potential = Potential(atoms, sampling=.05)
exit_wave = wave.multislice(potential)
measurement = exit_wave.diffraction_patterns()
The SAED simulation becomes a converged-beam electron diffraction (CBED) simulation by only changing the
PlaneWave for a Probe; the semiangle cutoff keyword now describes the condenser aperture of the probe.
wave = Probe(energy=80e3, semiangle_cutoff=10)
potential = Potential(atoms, sampling=.05)
exit_wave = wave.multislice(potential)
measurement = exit_wave.diffraction_patterns()
The CBED simulation can become a precession ED simulation by setting the tilt property of the Probe, with
a specified array of angles.
wave = Probe(energy=80e3, semiangle_cutoff=30, Cs=1e-3 * 1e10, defocus="scherzer")
potential = Potential(atoms, sampling=.05)
wave.tilt = precession_tilts(precession_angle=20, num_samples=60)
exit_wave = probe.multislice(potential)
measurement = exit_wave.diffraction_patterns()
A STEM-ADF simulation requires defining the scan area using the GridScan (defaulting to the entire area of the
cell) and an AnnularDetector defining the inner and outer detector angles (again in \(\mathrm{mrad}\)).
wave = Probe(energy=80e3, semiangle_cutoff=30, Cs=1e-3 * 1e10, defocus="scherzer")
potential = Potential(atoms, sampling=.05)
detector = AnnularDetector(inner=60, outer=120)
scan = GridScan()
measurement = probe.scan(potential, detectors=detector, scan=scan)
The STEM-ADF simulation is easily modified into a 4D-STEM simulation by swapping the
AnnularDetector for a PixelatedDetector.
wave = Probe(energy=80e3, semiangle_cutoff=30, Cs=1e-3 * 1e10, defocus="scherzer")
potential = Potential(atoms, sampling=.05)
detector = PixelatedDetector()
scan = GridScan()
measurement = probe.scan(potential, detectors=detector, scan=scan)
These examples demonstrate simplified abTEM code for simulating several common electron microscopy experiments.
For clarity, we have omitted import statements, and the atomic models are assumed to be given as atoms.
Equivalent complete executable code examples can be found further in the documentation
.
We recommend that new users work through our walkthrough to familiarize themselves with ab TEM and its features.