Resumen de Towards a new dimension in analytical TEM: EELS, Tomography and the Spectrum Volume

Lluís Yedra Cardona

• The main goal of this thesis is to combine electron tomography and electron energy loss spectroscopy (EELS) in the TEM, in order to obtain chemical and electronic information in 3D in the nanoscale. Tomography in the TEM has been discussed and the advantages and disadvantages of HAADF STEM tomography have been analyzed. - A Fe3O4 nanocube sample has been considered to illustrate the acquisition, alignment, tilt axis assessment, reconstruction and visualization processes. - HAADF STEM tomography has been used to reconstruct the shape of a series of Cu2O nanoparticles. For one of the samples, the facets of the Cu2O nanoparticles, octahedral in shape, have been found to correspond to {111} planes. The physics and the data analysis methods of core-loss EELS have been briefly summarized. Two relevant contributions have been made: - The effects of beam precession on EELS have been investigated. A signal enhancement due to electron beam precession in the TEM when in zone axis conditions has been discovered for the first time. - A homemade software, Oxide Wizard, for valence state determination of transition metals has been developed and applied to map the Mn oxidation state of a MnOx/FeOy nanoparticle. EELS spectrum imaging and electron tomography have been combined to recover chemical information in 3D for FexCo(3-x)O4@Co3O4 mesoporous nanoparticles. - The samples have been first characterized by HAADF STEM tomography and EELS. - EELS chemical information is, of course, limited to 2D maps, while only structural information is be recovered in 3D form HAADF tomography. . An EELS-SI tilt series has been acquired at 80 kV and low acquisition times, resulting in a very noisy dataset. - Principal component analysis (PCA) has been used to separate the noise from the signal. - Quantifications of the O, Fe and Co signals have been reconstructed, and the shape of the particle recovered. - From the noise clean dataset, components with physical meaning (iron oxide, cobalt oxide and thickness) have been extracted using independent component analysis (ICA), and reconstructed in 3D. - A new kind of signal with chemical and thickness information has been obtained by combining the thickness component with the quantification results. This signal is able to properly reconstruct the chemical structure of the sample in 3D, detecting a higher presence of iron in the surface, and an even distribution of Fe inside of the particle. . Another EELS-SI tilt series has been acquired at 80 kV and low acquisition times with enough spatial resolution to resolve the porous structure. - The noise has been reduced using PCA. - Four signals extracted from the spectrum have been reconstructed: the integrated edge intensities for oxygen, iron and cobalt, and the integrated background prior to the oxygen edge. EELS-SI tomography has been applied to reconstruct CoFe2O4 (CFO) nanocolumns embedded in a BiFeO3 (BFO) matrix grown on LaNiO3 buffered LaAlO3 substrate (BFO-CFO//LNO/LAO). - The nanocomposite sample has been prepared in a nanoneedle shape by means of focused ion beam (FIB). This preparation has the advantage of keeping the thickness of the sample constant throughout the tilting experiment. The nanoneedle contains an island of CFO in a matrix of BFO, as well as the LNO/LAO substrate. - The noise has been discarded using PCA. - The oxygen, iron and lanthanum edges intensities have been extracted and reconstructed. A new kind of data hypercube, the Spectrum Volume, has been proposed. Just as a Spectrum Image contains a spectrum per pixel, a Spectrum Volume is to contain a spectrum per voxel. EELS-SI tomography has been applied to successfully acquire a spectrum volume (SV) of the (BFO-CFO//LNO/LAO) system. - PCA has been used to discard the noise and Bayesian linear unmixing (BLU) has been used to extract independent components (eigenspectra). - Each eigenspectrum has a 2D weighting map for each tilt; weighting maps have been reconstructed into 3D weighting volumes. - Full single spectra for any point in the reconstructed space can now be recovered as a weighed sum of eigenspectra, using the 3D weighting distributions. This approach is used to extract single spectra, spectrum lines and spectrum images from the inside of the sample. In the context of this thesis, tomography and EELS-SI have been combined to successfully recover chemical information in 3D at the nanoscale, and a new kind of data hypercube, the Spectrum Volume, has been proposed and experimentally recovered.