Spatially resolved TiOx phases in switched RRAM devices using soft X-ray spectromicroscopy

Reduction in metal-oxide thin films has been suggested as the key mechanism responsible for forming conductive phases within solid-state memory devices, enabling their resistive switching capacity. The quantitative spatial identification of such conductive regions is a daunting task, particularly for metal-oxides capable of exhibiting multiple phases as in the case of TiOx. Here, we spatially resolve and chemically characterize distinct TiOx phases in localized regions of a TiOx–based memristive device by combining full-field transmission X-ray microscopy with soft X-ray spectroscopic analysis that is performed on lamella samples. We particularly show that electrically pre-switched devices in low-resistive states comprise reduced disordered phases with O/Ti ratios around 1.37 that aggregate in a ~100 nm highly localized region electrically conducting the top and bottom electrodes of the devices. We have also identified crystalline rutile and orthorhombic-like TiO2 phases in the region adjacent to the main reduced area, suggesting that the temperature increases locally up to 1000 K, validating the role of Joule heating in resistive switching. Contrary to previous studies, our approach enables to simultaneously investigate morphological and chemical changes in a quantitative manner without incurring difficulties imposed by interpretation of electron diffraction patterns acquired via conventional electron microscopy techniques.


X-ray photoelectron spectroscopy (XPS)
XPS was used to characterise the TiOx thin film. All spectra were recorded on a Thermo Scientific Theta Probe Angle-Resolved X-ray Photoelectron Spectrometer (ARXPS) system with a monochromated Al Kα X-ray source (hν = 1486.6 eV). The X-ray source was operated at 6.7 mA emission current and 15 kV anode bias and pass energies of 200 eV and 50 eV were used for survey and core level spectra, respectively. Spectra were corrected for any charge shifts by aligning them to the C 1s core level at 285.0 eV and all data were analysed using the Avantage software package. 2 The XPS survey spectrum of the TiOx thin film along with the O 1s and Ti 2p core levels are shown in Fig. S1a, S1b and S1c, respectively. The Ti 2p core level shows charge transfer satellites S3/2 and S1/2 at higher binding energies. 1 Furthermore, a small population of Ti 3+ of the order of 4 % of the total Ti is observed (Fig. S1d). The O 1s core level shows a surface oxygen component, often referred to as non-lattice oxygen, on the higher binding energy side of the main core line. core level. (c) Ti 2p core level including higher binding energy satellites S3/2 and S1/2. (d) Fit of the Ti 2p3/2 peak.

TXM-NEXAFS
The TXM set-up used in this work presents significant advantages compared with previously used geometries. In particular, it does not require replacement of the silicon wafer support with a fragile standing Si3N4 window 3,4 (which could affect the device electrical behaviour due to strain effects and difficulty in sinking the Joule heating 5,6 ) and does not require removal of the top electrode prior to analysis (which is usually performed by a scotch tape method and could lead to unwanted removal of the thin TiO2 layer underneath critical areas 5,7 ), both necessary steps to enable the X-ray transmission if irradiating the device from the top electrode. Most importantly, our geometry allows direct visualization and chemical investigation of the cross-section of the device, as shown in Fig.   3a (main manuscript). Ti 2p and O 1s spectra extracted from the TiOx film in PRI device case are shown in Fig. 3c and 3d (main manuscript), respectively. The first doublet (2p3/2) (457-462 eV) of Ti 2p spectra (Fig. 3c) originates from transitions to (2p3/2, 3d-t2g) and (2p3/2,3d-eg) states while the second doublet (2p1/2) (462-468 eV) originates from transitions to the corresponding 2p1/2 states. The 2p3/2 -2p1/2 splitting is due to spin-orbit coupling while the t2g-eg separation is the crystal-field splitting due to the surrounding O atoms. 8,9 It has to be noted that in all spectra, the (2p3/2 , eg) peak is broader than the (2p3/2, t2g) due to the large degree of hybridization of eg orbitals with O ligand orbitals. 10 Satellite peaks at 470.5 and 476.0 eV due to polaronic transitions are also often observed. 11,12 The O 1s spectra (Fig. 3d,

Chemical mapping
X-ray microscopy image sequences can be analysed to provide quantitative maps of chemical components 15,16 . In this study maps of three spectrally distinct components -TiOx amorphous, TiOx reduced and a third component similar to TiOx reduced but slightly different at the O 1swere generated by fitting to linear combinations of reference spectra extracted from specific regions of the area measured. If additional TiOx components were added, the component maps from the fit contained large regions with unphysical negative coefficients, indicative of an over-determined fit.
If any one of the three reference spectra were removed from the fit, the residual of the fit increased significantly in the regions of the missing component. Figure S7 presents the reference spectra and color coded composite maps for the 3-component fit using stack fit analysis (which also includes a constant to fit the non-Ti, non-O components such as the Pt electrodes) to the separate (Ti 2p or O 1s) and the combined (Ti 2p appended O 1s, aligned) image sequences.
In order to derive quantitative thickness scales for component maps the normal procedure would be to scale the as-extracted reference spectra to the elemental response for a material of the expected elemental stoichiometry and density 15, 16 evaluated from tabulated X-ray absorption data 17 , in which case the grey scales of the derived component maps would give nm thicknesses.
However, because we do not know the composition or density, in this case the edge jumps in background subtracted Ti 2p and O 1s spectra were used to derived amounts of Ti and O based on the scaling to the corresponding edge jumps in pure elemental reference spectra of Ti (Δ480-450eV = 0.0023) and O (Δ 555-428eV =0.0019) at a density of 1.0 g.cm -3 . 17 While this approach has quantitative uncertainties with respect to density, it does have the merit of giving a reliable measure of the O/Ti ratio since the density in any specific compositional region will be constant. This method was used to derive the O/Ti ratios reported in Table S1. All data manipulations were performed using aXis2000. 18 The edge jump values measured from the extracted X-ray absorption spectra have a precision of ~20 %, based on variability of the choice of 'reasonable' background subtraction.