Investigating the loss spectrum of individual carbon nanotubes at high energy resolution in real and momentum space

Abstract

Over the last quarter of a century Electron Energy Loss Spectroscopy (EELS) has proved itself as an excellent tool for investigating the plasmonic [1‐5] and electronic [1, 3, 4] response of carbon nanotubes: either by attaching a spectrometer to a Scanning Transmission Electron Microscope STEM [4, 5] or using a purpose built stand‐alone spectrometer [1‐3]. While STEM‐EELS provides superior spatial resolution to that of a stand‐alone spectrometer, it was not until recent improvements in electron source monochromation that the energy resolution of the two techniques became comparable. This was demonstrated by Sato and Terauchi [4] who used TEM –EELS to identify peaks in the loss spectrum of individual CNTs attributable to inter‐band transitions between van Hove singularities (vHs) in the CNTs' electronic density of states. As an extension of the work of Sato and Terauchi [4], the present study is focused on the spatial and momentum dependence of the loss spectrum of individual CNTs. Fig. 1A shows loss spectra acquired from the single walled CNT in Fig. 1B using both a penetrating (1) and an aloof beam geometry (2), (3). The spectra in Fig. 1A are primarily characterised by peaks attributed to vHs (black arrows) [1, 3, 4], and, the π and π+σ plasmon resonances [1‐5]. The apparent red‐shift of the two plasmon peaks with increasing distance to the tube centre is consistent with literature [5]. Note the lack of a comparable red‐shift for the “vHs peaks” as well as a significant change in fine structure of the high energy shoulder of the π plasmon peak. Resolving the CNT loss spectrum in momentum space allows for access to information not readily available in real space. Specifically, the momentum resolved spectra in Fig. 1C reveal that the vHs peaks are non‐dispersive (no peak shift with increasing momentum transfer), consistent with a reported measurement of a “bulk” sample of purified single wall CNTs [1]. Furthermore, it is shown that the “π peak” comprises two distinct plasmon modes: “π 1 “ and “π 2 ”. The π 1 peak is non‐dispersive, thus indicating confinement perpendicular to the CNT axis, while the π 2 peak is dispersive, which in turn indicates significant plasmon propagation along the CNT axis [2, 3].