Electronic State Analysis by Monochromated STEM-EELS

Introduction

Many efforts to improve the energy resolution of electron energy-loss spectroscopy (EELS) incorporated in a transmission electron microscope (TEM) have been made so far. Recently, the development of a new generation monochromator has made it possible to characterize materials with high energy and spatial resolution by using it together with a spherical aberration corrected scanning transmission electron microscope (STEM). In this report, after briefly introducing the performances of a (scanning) transmission electron microscope equipped with a monochromator installed at Institute for Chemistry Research of Kyoto University, the author will show two examples of high energy resolution EELS experiments. One is the study of localized surface plasmon (LSP) excited in silver nanoparticles (NPs) supported on MgO crystal. It will be shown that the excitation probability of LSP depends on the trajectory of an incident electron probe. The other is the application of energy-loss near-edge structure (ELNES) appearing in carbon K-edge excitation spectra to organic crystal. Owing to the small lifetime broadening of the initial and final states, the fine structures inherent to the molecule are observe in carbon K-edge ELNES, leading to apply to molecular analysis.

Monochromated STEM-EELS (JEM-ARM200F)

Figure 1 shows an appearance of a monochromated (scanning) transmission electron microscope (JEOL; JEM-ARM200F). The monochromator consisted of double Wien filters and deflection coils is incorporated between the Schottky type electron gun and the accelerating tube [1]. An energy dispersed focused beam is formed in the lower part of the first Wien filter, and monochromatization of electrons is performed by inserting an energy selection slit at the dispersion plane. The second filter plays a role of focusing the monochromated beam into an achromatic beam. For this reason, the electrons emitted from the electron source and the electrons focused at the exit plane of monochromator have a 1: 1 relationship. The energy dispersion of the Wien filter is 12.3 μm/eV, and the energy width of electrons is able to be selected by changing the width of the slit. There are seven types of slit widths between 0.1 μm and 4 μm in this device. When inserting a slit, the energy resolution estimated from the full width at half maximum of zero-loss peak can be chosen from 30 meV to 250 meV. The spherical aberration correctors (CEOS; CESCOR and CETCOR) for illumination and imaging lens systems are installed in a column, which makes it possible to perform the high spatial resolution STEM and TEM observations. As analytical apparatuses, an imaging filter (Gatan; Quantum ESR) and an energy dispersed X-ray spectrometer (JEOL; JED-2300T SDD100GV) are equipped with this microscope. The accelerating voltage of 200 kV or 60 kV can be selected, so measurement with low accelerating is also possible. Figure 2 shows the zero-loss peak when an energy selection slit of 0.1 μm is inserted at an accelerating voltage of 200 kV. For comparison, the spectrum measured by a cold field emission gun (Cold-FEG) is also shown. The full width at half maximum of the peak is 33 meV, which is one order of magnitude narrower than that of Cold-FEG, and the tail intensity of the zero-loss peak in the near infrared region of 1 eV or less is greatly reduced. In this way, since the measurable region of the spectrum has extended to low energy side, it has become possible not only to demonstrate its power in the study of surface plasmons as described below but also to detect vibrational excitations [2, 3].

Fig. 1 Appearance of JEM-ARM200F equipped with a monochromator

Fig. 2 Zero-loss spectra acquired with a monochromated gun (red) and cold-FEG (blue)

An energy selection slit of 0.1 mm is inserted. An accelerating voltage is 200 kV.

Dielectric substrate effects on localized surface plasmons

When light or electron is irradiated onto metal NPs, surface modes called localized surface plasmons (LSPs) are excited. This is due to collective oscillation of valence electrons on the surface, and the nano-particle surface is accompanied by strong near-field light. Since the resonance condition of LSP is sensitive to the size of the particle and the surrounding environment, its application to biosensors or photocatalysts has been extensively studied. In order to investigate the physical properties of LSP in more detail, it is necessary to analyze single NPs with high spatial resolution. High energy resolution EELS combined with STEM is a powerful tool for research on LSP, because EEL spectra in the near infrared region can be measured efficiently with sub nm spatial resolution. Many studies on LSP have been reported using this method [4]. In the following, the author will introduce the study of dielectric substrate effect on LSP excited in a silver NP.
Figure 3 shows the results of high energy resolution STEM-EELS measured from a silver NP supported on MgO substrate [5]. The feature of this measurement is that the electron probe is incident on the interface between the substrate and the NP in parallel, allowing us to directly investigate substrate effects on LSP excitations as a function of the distance from the substrate. Spectrum-image (SI) data were obtained with an energy resolution of 70 meV, a collection semi-angle of 29.2 mrad and a spatial sampling of 0.4 nm per pixel. The high-angle annular dark-field (HAADF) image in Fig. 3(a) demonstrates that the silver NP has an almost spherical shape with a diameter of 14 nm. Figure 3(b) shows the spectra extracted from top (indicated by A), side (B) and gap (C) regions around the NP, at a distance of 1 nm from the surface of the particle, as shown in Fig. 3(a). The LSP resonance energy is slightly different depending on the position of the incident probe, and the resonance energy at the top position (A) of the particle is shifted by 80 meV lower energy than the side position (B). The EELS map using the intensity of 3.40 ± 0.20 eV near the resonance energy is shown in Fig. 3(c). Due to the presence of the MgO substrate, the LSP excitation distribution excited in the spherical silver NP is asymmetric; the highest intensity is observed at the top position apart from the interface, while the intensity at the gap position is very weak. Such peak shift and asymmetric intensity distribution are considered to be the effects of the dielectric substrate on the LSP. In order to understand this, we performed simulations by discrete dipole approximation (DDA) for a silver NP on an MgO substrate.

Fig. 3 Localized surface plasmon excited in a silver NP supported on MgO crystal

(a) HAADF image. (b) EEL spectra extracted from three different positions. (c) EELS map using the intensity of LSP resonance peak.

Figure 4(a) presents the EEL spectra calculated for three different electron trajectories, as shown in the model (inset), consisting of a silver nanosphere with a diameter of 14 nm and an MgO substrate with a semi-infinite size. These results are compared with the spectrum calculated for an isolated silver NP in a vacuum, indicated by the black line. The calculation was performed using a DDEELS code [6] and the dielectric function of silver reported by Palik [7]. The dielectric function of MgO was assumed to be a constant value of 3.13 [8]. The LSP resonance energy and the peak intensity depending on the electron trajectories well reproduce the experimental results of Fig. 3(b). The LSP peaks of the NP on the substrate appear at slightly lower energy values than that of the isolated NP. This red shift of LSP peaks represents the substrate effect. On the other hand, the difference in resonance energy at the electron trajectory A and B is related to the polarization direction of the dipole mode of the LSP excited in the silver NP. The LSP dipole modes for which the polarization is perpendicular or parallel to the substrate are excited by electrons having trajectories A and B, respectively. The resonance energy shift depending on the direction of polarization of the LSP with respect to the substrate has also been observed in experiments using linearly polarized light [9]. It should be emphasized that the results equivalent to the experiment of polarized light can be obtained with high spatial resolution by selecting the electron trajectory in the STEM-EELS experiment. The intensity of the LSP peak measured at trajectory A is strong compared to that for an isolated NP, while that for trajectory C is considerably weaker. This characteristic intensity distribution is also noticeable in the EELS map shown in Fig. 4(b), which is calculated from the intensity at 3.40 eV. The LSP excitation probability for the NP is enhanced at the top surface, far from the dielectric substrate, and is suppressed in the gap region. The simulated map reproduces well the experimental EELS map of Fig. 3(c).

Fig. 4 Simulated results by DDEELS

(a) Calculated spectra at three different electron positions shown in the inset and the spectrum for an isolated silver NP (black).
(b) Calculated EELS map using the intensity of LSP resonance peak.

Next, the reason why the spatial distribution of the LSP excitation probability becomes asymmetric is considered. In the DDEELS code, the metallic NP is considered as an aggregate of discrete dipoles. The excitation probability of EELS is calculated by summing the product of the dipole moment P_j at position r_j and the electric field E_j^app applied by an incident electron. The dipole moment is proportional to the local electric field, which is the superposition of the applied field and the fields resulting from other dipoles. To a first approximation, therefore, the energy-loss probability is expected to be strongly affected by the distribution of the applied electric field depending on the location of the incident electron. In the case of NPs supported on a substrate, not only the structural symmetry is reduced, but also the intensity distribution of the applied electric field is modified by the polarization of the dielectric substrate compared to the case of an isolated NP. The effect of the substrate on the applied electric field can be assessed based on the image charge model. The polarization field of the substrate due to an incident electron at (x_e, y_e, z_e) can be described by an image charge, q, located at (−x_e, y_e, z_e) in the substrate when the substrate surface is at x=0. The total electric field applied to the dipole is equal to the sum of the fields generated by the electron and its image charge q=(1−ε_MgO)e(1+ε_MgO), where ε_MgO is the dielectric function of MgO and e is the electron charge. Since ε_MgO is greater than 1 in the visible light region, q is positive, such that the polarization field of the substrate acts to enhance the applied field of the incident electron in the region between the electron and its image charge. Figure 5 shows the calculated applied field, including the substrate effect, for a silver NP with a diameter of 14 nm supported on an MgO crystal. When the incident electron is located at the top surface of the NP, the applied electric field covers the majority of the NP as in Fig. 5(a), meaning that many dipoles are excited in the NP, leading to a high energy-loss probability. In contrast, when the electron is incident in the vicinity of the interface between the NP and the substrate (Fig. 5(c)), the applied field in the NP is weak and its distribution is limited to the region near the interface. This is attributed to the strong cancellation of field in the NP region by the polarization of the substrate, because the NP is located in the opposite direction to the image charge with respect to the position of the incident electron. Therefore, the energy-loss probability becomes low in the vicinity of the interface. In the case of an electron travelling near the side of the NP (Fig. 5(b)), the applied field distribution in the NP is similar to that in an isolated NP (Fig. 5(d)). Therefore, the substrate effect is weak and limited to the region between the incident electron and the interface, which is essentially a vacuum, leading to a similar LSP peak intensity (Fig. 4(a)). The above modifications of the applied field due to the substrate cause the asymmetric distribution observed in the EELS map shown in Fig. 3(c).
The presence of a dielectric substrate under the NP can affect the spatial distribution of the LSP excitation probability as well as its resonance energy. The dipole mode of the LSP excited in a spherical NP splits into two modes corresponding to oscillations parallel or perpendicular to the substrate surface. These modes could be probed separately using different electron trajectories, providing results identical to those obtained from polarized light analyses.

Fig. 5 Distributions of applied electric field calculated by DDELS

(a) to (c) MgO substrate is included. (d) Isolated silver NP.

High resolution EELS of organic thin films

In the measurement of energy-loss near-edge structure (ELNES) appearing in inner-shell electron excitation spectrum, the energy resolution is limited by the lifetimes of initial and final states and the density of states in conduction band as well as the energy spread of the primary beam and the resolution of spectrometer. When an inner-shell electron excited into unoccupied electronic band, a hole is left in the core level. The lifetime of the core hole determines the energy width of the initial state, which is related to the generation of characteristic X-ray and Auger electrons due to decay of electrons existing at shallow levels than the core hole. The energy widths of core levels calculated theoretically show a tendency that it is broader at the core level having larger binding energy [10], indicating that ELNES of shallow core level benefits from improved energy resolution. On the other hand, the energy broadening due to the lifetime of final state depends on the kinetic energy of excited electron. It has been shown that the final state broadening calculated by using the inelastic mean free path of excited electron has very narrow near the threshold region, and becomes broad with increasing the kinetic energy [11]. Considering the lifetime broadening of initial and final states, therefore, it can be said that the effect of high energy resolution appears in the spectral structure near the absorption edge excited from a relatively shallow core level. Furthermore, ELNES reflects the partial density of states of unoccupied band, the band dispersion also affects the broadening of spectrum. Actually, it has been reported that the oxygen K-edge ELNES measured from some transition metal oxides does not improve even with monochromated EELS, which is mainly due to the effect of band dispersion (solid-sate effects) [12]. In the case of organic thin film crystals described below, however, the interaction between molecules is weak and the band dispersion is small, so it is expected that monochromated EELS will be of benefit for ELNES appearing just above the threshold of carbon K-edge with a relatively small binding energy. The spectral features specific to conjugated molecules will appear as sharp π*-resonance peaks.
Figure 6(a) shows the carbon K-edge ELNES measured from copper-phthalocyanine (CuPc) and chlorinated copper-phthalocyanine (CuPcCl₁₆) thin films. CuPc is a four-fold symmetric planar molecule in which copper atom is coordinated at the center of the porphyrin ring as shown in Fig. 6(b). CuPcCl₁₆ is produced by substituting the peripheral hydrogen atoms with chlorine atoms. CuPcCl₁₆ is known as one of the most intense molecules against electron beam irradiation among organic molecules, and its critical electron dose is about 30 C/cm², while since the critical electron dose of CuPc is about 1 C/cm², one should pay attention to the electron irradiation damage of the sample. Spectra shown in Fig. 6(a) were measured with a probe current of 0.05 pA and 1 pA for CuPc and CuPcCl₁₆, respectively. Spectrum image data were acquired below the critical electron dose of each molecule, and then the spatial information was averaged to improve the signal-to-noise ratio of spectra. The fine structures in both spectra show clear differences within 4 eV from the threshold; peaks (A) and (B) appear in both spectra, but in the ELNES of CuPcCl₁₆, an extra peak (C) is observed around 287 eV. These peaks are attributed to the 1s → π* transitions, and the final state may be the lowest unoccupied molecular orbital (LUMO). There are three independent carbon atoms with different bonding in each molecule. As shown in Fig. 3(b), it can be classified into the C1 and C2 atoms bonded to peripheral atoms (hydrogen or chlorine) and a carbon atom, the C3 atoms bonded to three carbon atoms, and C4 atom bonded to two nitrogen atoms and a carbon atom. It has been observed from XPS measurement of CuPc molecule that the 1s level of these carbon atoms has a slight different binding energy [14]. Such chemical shifts of 1s level lead to the different π* peak energy in the ELNES. In the case of CuPc molecule, the binding energies of 1s level at C1, C2 and C3 sites are almost same, while that at C4 site is large. This is because the electronegativity of the nitrogen bonded to the C4 site is large, and the valence electron density on the C4 site becomes lower than that at the other carbon sites, so the Coulomb repulsion energy between the valence electron and the 1s electron decreases. Therefore, the peak (A) in the ELNES of CuPc is due to the excitation of C1, C2 and C3 sites, while the peak (B) corresponds to that of the C4 site [15]. In the case of CuPcCl₁₆, since the electronegativity of the chlorine bonded to the C1 and C2 sites is larger than that of nitrogen, the 1s level of C1 and C2 sites is stabilized more than that of C4 site. Therefore, the extra peak (C) in the ELNES of CuPcCl₁₆ can be attributed to the π*-resonance excited at C1 and C2 sites. In order to make a quantitative interpretation of these ELNESs including the relative intensity of each peak, it is necessary to calculate spectrum taking into account the effect of the core hole on the independent carbon sites. As shown in the above example, the improvement of energy resolution appears effectively in the fine structures just above the threshold. Although the separation between the peaks (B) and (C) of CuPcCl₁₆ is narrow, 0.7 eV, it has been clearly observed. This suggests that the chemical shifts of inner-shell level can be detected in ELNES. It is expected that the analysis of functional groups bonded to organic molecules becomes possible by taking advantage of the features of such high resolution carbon K-edge ELNES.

Fig. 6 Carbon K-edge ELNES of copper-phthalocyanine and its chlorinated thin films (a) and molecular structure model (b)

Finally, the result of vibrational spectrum is shown briefly. Excitations of various vibrational modes are observed in the infrared absorption spectra of organic molecules, but most of them appear below 200 meV. In the case of CuPc molecule, the C-H stretch vibration mode of benzene ring is excited at around 380 meV. As shown in Fig. 7, the broad peak assigned to the C-H vibration is observed, but its intensity is considerably weak compared to the optical phonon peak measured from h-BN. Although this spectrum was measured with an irradiation dose of 0.5 C/cm2 which was smaller than the critical dose of this molecule, the effect of electron irradiation may not be neglected. Actually, when the spectrum was measured at 1.25 C/cm2 slightly above the critical dose, the C-H vibrational peak disappeared, which suggests that the dissociation of hydrogen atoms contributes greatly in the early stage of irradiation damage. This is also confirmed from the fact that the intensity of peak (A) in the ELNES of CuPc decreased with the increasing irradiation dose, indicating the change of bonding state of C1 and C2 sites by the dissociation of hydrogen atoms.

Fig. 7 Vibrational EEL spectrum measured from copper-phthalocyanine thin film

Summary

The monochromated STEM-EELS is a powerful tool to investigate the properties of surface plasmons and vibrational excitation appearing in visible to near infrared region with high spatial resolution. It was also demonstrated that when the low dose measurement is applied to organic crystals, the carbon K-edge ELNES with high energy resolution provides useful information for molecular analysis.

Acknowledgments

This work was partly supported by Grants-in-Aid for Scientific Research (No.16K13625).

References

• M. Mukai, E. Okunishi, M. Ashino, K. Omoto, T. Fukuda, A. Ikeda, K. Somehara, T. Kaneyama, T. Saitoh, T. Hirayama, Y. Ikuhara, Microscopy , 64, 151 (2015).
• T. Miyata, M. Fukuyama, A. Hibata, E. Okunishi, M. Mukai, T. Mizoguchi, Microscopy 63, 377 (2014).
• O. L. Krivanek, T. C. Lovejoy, N. Dellby, T. Aoki, R. W. Carpenter, P. Rez, E. Soignard, J. Zhu, P. E. Batson, M. J. Lagos, R. F. Egerton, P. A. Crozier, Nature, 514, 209 (2014).
• C. Colliex, M. Kociak, O. Stéphan, Ultramicroscopy 162, A1 (2016).
• Y. Fujiyoshi, T. Nemoto, H. Kurata, Ultramicroscopy 175, 116 (2017).
• N. Geuquet, L. Henrard, Ultramicroscopy 110, 1075 (2010).
• E. D. Palik, Handbook of Optical Constants of Solids 1, Academic Press, New York, (1985).
• E. D. Palik, Handbook of Optical Constants of Solids 2, Academic Press, New York, (1991).
• M. W. Knight, Y. Wu, J.B. Lassiter, P. Nordlander, N. J. Halas, Nano Lett . 9, 2188 (2009).
• M. O. Krause, J. H. Oliver, J. Phys. Chem. Ref. Data 8, 329 (1979).
• R. F. Egerton, Ultramicroscopy 107, 575 (2007).
• C. Mitterbauera, G. Kothleitnera, W. Groggera, H. Zandbergenb, B. Freitagc, P. Tiemeijerc, F. Hofer, Ultramicroscopy 96, 469 (2003).
• H. Kurata, Y. Fujiyoshi, Y. Tomisaki, T. Nemoto, M. Haruta, The 16th EMC Proc., Lyon, France, 863 (2016).
• F. Evangelista, V. Carravetta, G. Stefani, B. Jansik, M. Alagia, S. Stranges, A. Ruocco, J. Chem. Phys. 126, 124709 (2007).
• R. De Francesco, M. Stener, G. Fronzoni, J. Phys. Chem. A116, 2885 (2012).