Microscopy Conference 2023

Prof. Dr.

Session chairs

Aberration correctors, electron energy-loss spectroscopy (EELS) and energy-dispersive X-ray spectroscopy (EDX) are nowadays indispensable tools for obtaining atomic-scale information of chemical composition and bonding in materials. Not only can material phases be identified, but important effects such as interdiffusion between phases, or charge ordering, to name but a few, can be studied. In the past decade, in EELS low energy losses were intensively used to identify surface plasmons in nanomaterials, even in a tomographic approach. This was made possible by the implementation of electron monochromators into the electron gun. Recent technological advances have opened the energy-loss window into the sub-10 meV regime where a range of new phenomena can now be studied. These include phonons, low-energy plasmons, and possibly many more low-energy excitations in materials. The wealth of information generates new grand challenges in understanding and modelling the underlying fundamentals of signal generation and intelligent processing of large data sets. Contributions are invited on hardware and technique developments, theory and simulation, data processing as well as applications covering both physical and biological sciences.

Phonon EELS data obtained from first principles

Invited speaker

Invited talk

Modelling spatially-resolved electron energy-loss spectra in the low-loss region

Abstract text (incl. figure legends and references) Mapping low-energy excitations such as plasmons, excitons, or phonons with high spatial and spectral detail is key for understanding and engineering nanostructured materials' optical, electrical, or thermal properties. One of the experimental techniques allowing for analyzing the low-energy excitations and their interaction down to the atomic scale is electron energy-loss spectroscopy (EELS) in a scanning transmission electron microscope (STEM), which enables the acquisition of spectra with a few-meV/Å spectral/spatial resolution [1,2]. Most of the recent state-of-the-art STEM-EELS experiments have been complemented or even guided by theoretical modelling and predictions. This talk will review two main approaches for simulating low-loss EEL spectra, using either classical electrodynamics or ab-initio methods. I will show that the macroscopic electrodynamics calculations are suitable for describing polaritonic excitations, such as plasmon or phonon polaritons that typically emerge in metals or ionic crystals, respectively. I will discuss examples of spatially-resolved EELS of electron-beam excitation of phonon polaritons in hexagonal boron nitride [3] and a system sustaining plasmon and phonon polaritons, which are engineered to strongly couple [4]. On the other hand, STEM-EELS of molecular systems or momentum-resolved EELS requires modelling at a truly microscopic level, which I will demonstrate with an example of a hBN-like molecule. The theoretically predicted results obtained with ab-initio calculations demonstrate that a single isotope impurity in this molecule would significantly affect the EELS probability [5]. Finally, we will discuss possible future directions in STEM-EELS, considering both theoretical and experimental efforts. [1] FS Hage et al., Science 367 (2020), p. 1124-1127. doi:10.1126/science.aba1136 [2] X Yan et al., Nature 589 (2021), p. 65-69. doi:10.1038/s41586-020-03049-y [3] A Konečná et al., Small 17 (2021), p. 2170201. doi:10.1002/smll.202170201 [4] P Gallina et al., arXiv:2112.12832 (2021), https://arxiv.org/pdf/2112.12832.pdf [5] A Konečná, F Iyikanat and FJ García de Abajo, ACS Nano 15 (2021), p. 9890-9899. doi: 10.1021/acsnano.1c01071

IM 2: Spectroscopy

MS 3: Low-dimensional and quantum materials

Spectroscopic coincidence experiments at nanoseconds time resolution in TEM

Abstract text (incl. figure legends and references) Electron energy loss spectroscopy (EELS) and energy dispersive spectroscopy (EDX) are two frequently used analytical techniques in transmission electron microscopy. Both techniques rely on the inelastic scattering of the incoming electron with the core level states of an atom to determine the chemical composition of the material. The two techniques differ with respect to which particle is detected. For EELS the energy of the inelastically scattered electron is measured and in EDX the energy of an emitted X-ray, which is created during the decay of the excited atom, is detected. Hence, for every X-ray emitted there was at least one electron that gave part of its energy to excite the atom which could give the impression that they convey very similar information. However, the energy resolution for EDX is typically 100 times worse compared to EELS due to the detection mechanism of X-rays (100 eV). On the other hand, EELS has a low signal-to-background ratio, (SNR) making it hard to detect low concentration elements. Since these two techniques arise from the same physical process, it should be possible to find a correlation in time between such an inelastic electron and x-ray. By only selecting an electron when an X-ray is detected nearby in time Kruit el al. [1] were able to detect this time correlation between the electrons and X-rays. Since individual electron and X-ray events need to be detected at nanosecond time resolution, a dedicated setup needed to be developed. In the present work, the current setup will be described where we were able to get a temporal resolution of ~250 ns without filtering events on the hardware level. Hence the conventional EELS and EDX signals are retained. Furthermore, we will show how by applying such a time correlation constraint it is possible to remove the background signal in EELS (see Fig.1) and obtain EDX spectra without having any prior knowledge on the shape of this background [2]. This is especially valuable for EELS where conventionally an empirical power-law function is used to fit the background where a pre-edge is needed to extrapolate a fitted background signal. Additionally, it will be shown how one can determine the collection efficiency [JV1] of both EELS and EDX without having the need for a reference sample and knowledge on beam current. Finally, it is discussed how using the time correlation between EELS and EDX can improve the signal-to-noise ratio for trace elements where the signal-to-background is too low for EELS and the energy resolution for EDX is insufficient to discriminate between overlapping x-ray peaks. [1] P. Kruit et al. Detection of X-rays and electron energy loss events in time coincidence. Ultramicroscopy, 13(3):205-213,1984. [2] D. Jannis, K. Müller-Caspary et al., Spectroscopic coincidence experiments in transmission electron microscope, Appl. Phys. Lett. 114, 143101 (2019) [3] D. Jannis et al., Coincidence Detection of EELS and EDX Spectral Events in the Electron Microscope, Appl. Sci. 2021, 11, 9058 [4] D.J., A.B. and J.V. acknowledge funding from the Flemish Research Fund FWO under project no. G093417N and G042920N, J.V. acknowledges funding from No 823717 ESTEEM3[JV1] , K.M.-C. acknowledges funding from the FWO under contract G042920N

IM 1: Progress in instrumentation and ultrafast EM

Abstract talk

Prof.

Mapping local phonon modes in silicon grain boundaries at atomic resolution

Abstract text (incl. figure legends and references) INTRODUCTION In the last few years, mapping of phonon modes by scanning transmission electron microscopy electron energy loss spectroscopy (STEM-EELS) has become possible [1-2]. Phonons, the quantized collective vibrations of atoms, play a fundamental role in many physical properties of materials, especially thermal and electrical conductivity. Changes in the atomic structure at grain boundaries (GBs) affect the local density of states (LDOS) of phonons and thus these physical properties. OBJECTIVES Here, we measure the LDOS across different GBs in silicon at atomic resolution and compare them to results of theoretical calculations. Such measurements have the potential to contribute to our understanding of thermal conductivity across internal interfaces, to the optimization of polycrystalline Si solar cells, and to the design of phononic devices. MATERIALS &amp; METHODS The Si sample in [110] zone axis exhibiting different GBs was prepared from an ingot by FIB and further thinned and cleaned using Ar ions. The experiments were performed in a Nion HERMES whose IRIS spectrometer was fitted with a Dectris ELA direct electron detector. For the atomic-resolution spectral maps (exposure time: 1 ms, dispersion: 1 meV per channel), a probe convergence semi-angle of 30 mrad and a spectrometer collection angle of 45 mrad were used. The energy resolution (FWHM of the zero-loss peak in vacuum) was 8.9 meV. To correct for slow sample drift, series of spectrum images were acquired and non-rigidly registered and integrated [3]. In addition, high-angle annular dark-field (HAADF) images were acquired using 36 mrad convergence angle with greater magnification in the monochromator to increase spatial coherence. Theoretical phonon DOS and LDOS were calculated from velocity-velocity correlation functions of molecular dynamics trajectories using the empirical Tersoff potential [4]. RESULTS Fig. 1 (a) shows an overview image of the sample exhibiting different GBs. In (b) the triple point of two meeting symmetric Σ3 and one symmetric Σ9 can be seen. A raw bulk spectrum, background fit and resulting phonon DOS are depicted in (c), marking the optic phonon peak around 62 meV that is analyzed in the following. In Fig. 2 for three different boundaries (symmetric Σ3, symmetric Σ9 and asymmetric Σ9), the atomic structure (via high-resolution HAADF), a HAADF image acquired simultaneously with the EELS signal, a map of the EELS intensity around 62meV, a line profile across it, and calculated LDOS are given. This observation allows us to correlate the changes in atomic structure with changes in the phonon LDOS from virtually no effect at all for the symmetric Σ3 to drastic modification (compared to bulk) for asymmetric Σ9. CONCLUSION We demonstrated mapping of localized phonon modes at GBs in Si at atomic resolution. The more structurally distorted the boundary configuration is, the stronger the suppression of the phonon LDOS - in agreement with theory. We expect that this research contributes to controlling phonon propagation and to minimizing decoherence in Si-based quantum computing using phonon entanglement. [1] F.S. Hage et al., Science 367, 1124 (2020). [2] X. Yan et al., Nature 589, 65 (2021). [3] L. Jones et al., Microscopy 67, 98 (2018). [4] P. Rez et al., Microsc. Microanal. 28, 672 (2022). Fig. 1: (a) Overview of Si sample, (b) high-resolution of GBs, (c) EELS in bulk with fit. Fig. 2: Structure, EELS and simulation for different GBs.

Beyond conventional chemical mapping: prospects and challenges of orbital mapping

Abstract text (incl. figure legends and references) It is well-known that core-loss EELS data contains not only information about the chemical elements present in the sample, but also about the scattering atoms" chemical environment, which manifests itself in the near-edge fine structure (ELNES). This enables some more advanced characterization techniques such as the fingerprinting of chemical compounds and the mapping of oxidation states. In this work, we explore the possibilities of fine structure mapping at "sub-atomic" spatial resolution, i.e., the real-space mapping of transitions to individual orbitals such as pz or dxy inside the sample. Arguably, an electron is more likely to trigger a transition from, say, a 1s core-electron to an unoccupied px state if it passes close to the region where the 1s-px-overlap is large than it is when passing "far away". Therefore, it is conceivable that by mapping the inelastic scattering probability (i.e., an EEL spectrum) with very high spatial resolution allows to map the orientation and shape of the px orbital. Indeed, this is corroborated by detailed theoretical calculations [1] as well as proof-of-principle experiments [2,3]. Fig. 1 shows an example from the mapping of Ti-eg-like states in Rutile oriented along the [0 0 1] zone axis at 80 kV. Despite the noise, the 90° rotation between adjacent Ti atoms is clearly visible. In this work, we discuss the fundamental requirements for successfully performing orbital mapping, include conditions on the sample (such as geometry, symmetry and stability), the experimental setup (such as spatial resolution, energy resolution and high-tension), and the detection process (such as the signal to noise levels). In particular, it is shown that a low symmetry in the chosen orientation is required to clearly separate the ELNES peaks of different orbitals, that narrow energy ranges are generally necessary, and that high incident doses are required, even with direct electron detectors. Despite the technical challenges, the prospect of mapping the electronic transitions between individual orbitals is exciting not only from a fundamental quantum mechanical point of view, but also for its possibilities for the design and characterization of novel materials in many fields, including electronics, energy storage, and catalysis. [4] [1] Löffler et al., Ultramicroscopy 131 (2013) 39 [2] Löffler et al., Ultramicroscopy 177 (2017) 26 [3] Bugnet et al., PRL 128 (2022) 116401 [4] The authors acknowledge financial support by the Austrian Science Fund (FWF) under grant nr. I4309-N36. Fig. 1: Example of mapping transitions to Ti-eg-like orbitals in Rutile. a) EELS spectrum with selected energy range (yellow). b) charge density calculated with WIEN2k. c) Experimental imaging after averaging over several unit cells. d) Same as c after Gaussian filtering. e) Simulated image at infinite dose. f) Simulated image at realistic dose. g) Same as f after Gaussian filtering. All scalebars indicate 5 Å. Adapted from [2].

IM 5: Quantitative image and diffraction data analysis

Simulations of electron energy loss spectra at ultra-low energy losses

Abstract text (incl. figure legends and references) Recent hardware developments in monochromation of the electron beam [1] in scanning transmission electron microscopy (STEM) have opened new domains of physics to the spectroscopic studies at down to atomic resolution. Most notably, excitation of phonons, happening in the tens to few hundreds of meV range, has been observed at unprecedented spatial resolutions [2,3,4]. Compared to other spectroscopic methods, high spatial resolution is the primary advantage of STEM. It gives access to local changes in the electron energy loss spectra (EELS) near interfaces or defects, also within this ultra-low energy-loss range [3,4]. The primary objects of interest for STEM-EELS are thus local changes of the low energy excitations. To predict and properly interpret the experimental data, reliable simulation methods are necessary. Such methods also need to have a favorable scaling with respect to system sizes (number of atoms, Nat). For instance, a structure model with an interface or a localized defect needs to be sufficiently large to avoid self-interaction of the defect with its periodic images. As an example, an fcc silicon structure model of 1000 atoms would be a cube of dimensions 2.73 nm3. Realistic sample thicknesses are often an order of manitude larger and lateral dimensions may often need to be larger to contain the spreading electron beam wavefunction in multislice calculations. Thus system sizes of tens of thousands of atoms are often needed. Transition potential multislice-based approaches for calculation of inelastic scattering [5,6] provide a transparent and accurate way of simulating STEM-EELS of phonons, and similarly with Bloch-waves based approaches [7] they allow to both include electron challeling effects as well as spectroscopic information. Nevertheless, these approaches require explicit knowledge of phonon modes, which requires diagonalization of dynamical matrix scaling with the third power of Nat. In the domain of tens of thousands of atoms such approaches quickly become computationally too expensive. We have recently presented a method named frequency resolved frozen phonon multislice method (FRFPMS; [8]), which has computing costs scaling linearly with Nat. We will discuss its capabilities and recent developments. This method is also a starting point for a description of another type of excitations in solid state, in particular, the excitations of a magnetic subsystem called magnons. Magnons occupy approximately the same energy range as phonons, however compared to phonons their interaction with electron beam is significantly weaker. Diffuse scattering due to magnons [9] and a semi-classical theory of scattering of electron beams on magnons [10] have been presented recently. We will discuss our recent developments in the theory of magnon scattering in STEM-EELS and routes to potential detection of this type of quasiparticle excitations by monochromated STEM. References: [1] O. Krivanek et al., Nature 514, 209 (2014). [2] F. Hage et al., Phys. Rev. Lett. 122, 016103 (2019). [3] F. Hage et al., Science 367, 1124 (2020). [4] X. Yan et al., Nature 589, 65 (2021). [5] B. D. Forbes and L. Allen, Phys. Rev. B 94, 014110 (2016). [6] C. Dwyer, Phys. Rev. B 96, 224102 (2017). [7] P. Rez and A. Singh, Ultramic. 220, 113162 (2021). [8] P. M. Zeiger and J. Rusz, Phys. Rev. Lett. 124, 025501 (2020). [9] K. Lyon et al., Phys. Rev. B 104, 214418 (2021). [10] B. Mendis, Ultramic. 230, 113390 (2021).

IM 2
Spectroscopy

Appointment

Session chairs

Description

Programme

IM 2Spectroscopy

Appointment

Session chairs

Description

Programme

IM 2
Spectroscopy