A Neutron Reflectometry Study of a Metal Oxide/Li-Ion Battery Electrolyte Interface

Abstract

The Li-ion battery solid-electrolyte interface (SEI) is critical to battery performance, but its characterization is challenging. SEI constituents are sensitive to the ambient environment and contain significant quantities of light elements (e.g. H and Li) which are not readily observed with electron and x-ray-based techniques. The nanometer-scale features of the SEI may be difficult to probe in situ, and its structure is potential-dependent, necessitating in situ measurements. Neutron reflectometry (NR) can help address these challenges.[i] In this work, in situ NR was used to characterize the interface between a model oxide, WO3, and a lithium ion battery electrolyte, 1 mol/dm3 LiPF6 in an ethylene carbonate/diethyl carbonate mixture, at a series of applied potentials. The NR data were fit to obtain scattering length density (SLD)-depth profiles of the interface. Because the SLD is nuclide-dependent, a composition depth profile may be inferred. Tungsten oxide was selected as a model surface to build on earlier work which used tungsten as a model metallic electrode.[ii] In that work, tungsten was selected as it does not alloy with lithium, allowing the interface to be considered in the absence of bulk transformations. The model oxide surface was prepared by thermal oxidation of a DC magnetron sputtered tungsten thin film on a Si\\SiO2 substrate. NR and x-ray reflectivity (XRR) of the bare wafer confirmed the electrode had the intended Si\\SiO2\\W\\WOx layer structure, that the oxide stoichiometry was WO3, and that the layers were of low porosity. The W layer underlying the WO3 served as a current collector. NR data measured at open circuit were best fit with a model containing the layers: Si\\SiO2\\W\\WO3\\interface layer\\solution. NR was subsequently measured at a series of potentials: 2.50 V, 1.50 V, 0.75 V, 0.25 V, 0.75 V, 1.50 V, and 2.50 V. As the electrode potential decreased, the layer that was originally WO3 grew thicker and its SLD decreased, indicating lithiation (lithium has a negative scattering length). The reduction may initially proceed by intercalation of Li, but at more negative potentials may proceed by a conversion mechanism. This is supported by the magnitude of the thickness increase (1.3-fold) and by the SLD (1.6×10-4 nm-2) of the layer. The values predicted for reduction of WO3 to homogeneously mixed W and Li2O are 1.7-fold and 1.2×10-4 nm-2, respectively. The SLD-depth profile across the oxide layer remained flat throughout the cycle, showing that reduction was homogeneous, rather than there being a front between reacted and unreacted material. The W layer thickness did not change over the entire series of potentials, indicating that any metallic W produced was well interspersed with other species and not plated onto W current collector. This was also supported by ex situ XRR data measured after removing the electrode from the cell which showed no significant changes in the W layer thickness and interfacial roughness. The interface layer, which could be considered the SEI, also grew thicker during reduction, and its SLD decreased to near 0 nm-2, which indicates the layer was rich in Li. Since deuterated solvents were used 1H was not expected to be one of the constituents of this layer (barring the presence of adventitious 1H-bearing species). The reduction charge passed was larger than expected for complete conversion of WO3 to W and Li2O; the excess charge might be accounted for by SEI formation. When the electrode was returned to 2.5 V the thicknesses of both the layer that was originally WO3 and the interface layer decreased. The SLDs of both layers increased but did not fully return to their original levels, indicating only partial removal of lithium. This is also reflected in the low coulombic efficiency of the cycle. This confirms the dynamic nature of the SEI noted in our earlier work. Comparing SLD-depth profiles at 0.25 V for a thermally-oxidized electrode and for an electrode that was not pre-oxidized showed that in the latter case, the low SLD surface layer was probably an SEI and not simply the reduction product of the oxide; the SLD profile for the thermally oxidized wafer did not look like a thicker version of that observed for the thin wafer. Instead, it showed an additional layer which we attribute to mixed W and Li2O. [i] Owejan, J. E.; Owejan, J. P.; DeCaluwe, S. C.; Dura, J. A., Chem. Mater. 2012, 24 (11), 2133-2140. [ii] Lee, C. H.; Dura, J. A.; LeBar, A.; DeCaluwe, S. C., J. Power Sources 2019, 412, 725-735.