Research Article Cite This: ACS Appl. Mater. Interfaces 2018, 10, 17471−17479
www.acsami.org
Crystal Orientation-Dependent Reactivity of Oxide Surfaces in Contact with Lithium Metal Justin G. Connell,*,†,‡ Yisi Zhu,† Peter Zapol,† Sanja Tepavcevic,† Asma Sharafi,∥ Jeff Sakamoto,∥ Larry A. Curtiss,† Dillon D. Fong,† John W. Freeland,§ and Nenad M. Markovic† †
Materials Science Division, ‡Joint Center for Energy Storage Research, and §X-ray Sciences Division, Argonne National Laboratory, Argonne, Illinois 60439, United States ∥ Department of Mechanical Engineering, University of Michigan, Ann Arbor, Michigan 48109, United States S Supporting Information *
ABSTRACT: Understanding ionic transport across interfaces between dissimilar materials and the intrinsic chemical stability of such interfaces is a fundamental challenge spanning many disciplines and is of particular importance for designing conductive and stable solid electrolytes for solid-state Li-ion batteries. In this work, we establish a surface science-based approach for assessing the intrinsic stability of oxide materials in contact with Li metal. Through a combination of experimental and computational insights, using Nb-doped SrTiO3 (Nb/STO) single crystals as a model system, we were able to understand the impact of crystallographic orientation and surface morphology on the extent of the chemical reactions that take place between surface Nb, Ti, and Sr upon reaction with Li. By expanding our approach to investigate the intrinsic stability of the technologically relevant, polycrystalline Nb-doped lithium lanthanum zirconium oxide (Li6.5La3Zr1.5Nb0.5O12) system, we found that this material reacts with Li metal through the reduction of Nb, similar to that observed for Nb/STO. These results clearly demonstrate the feasibility of our approach to assess the intrinsic (in)stability of oxide materials for solid-state batteries and point to new strategies for understanding the performance of such systems. KEYWORDS: solid electrolytes, solid-state batteries, Li metal, buried interface, reactivity, model system, surface science
■
INTRODUCTION Understanding ionic transport across interfaces that are formed between dissimilar materials, as well as the intrinsic chemical stability of such interfaces, is a fundamental challenge that spans disciplines from solid-state physics to materials chemistry and electrochemistry. Interfacial stability impacts a broad scope of applications, including oxygen-ion transport in memristors,1 efficiency of solid oxide fuel cells,2 and the development of new electrochemical energy storage systems that move beyond current Li-ion technologies.3,4 Solid-state batteries are a particularly promising technology due to their inherently safe mode of operation,5 improved energy density due to the higher voltage window enabled by solid oxide/sulfide electrolyte materials relative to conventional organic liquid electrolytes,6,7 and the potential of developing new system architectures that are simply not possible with conventional Li-ion systems.8 Nevertheless, significant barriers exist to implementing solidstate batteries, including modest power densities due to the low ionic conductivities of many solid-state electrolytes9 and, more importantly, significant issues with developing robust interfaces that are stable to contact with Li metal10 and capable of withstanding significant stress/strain during extensive cycling.11,12 © 2018 American Chemical Society
Gaining insight into the interfacial stability of solid-state battery systems is a highly challenging task. Charge/discharge testing and impedance spectroscopy provide useful information about system-level performance but are ultimately insufficient for understanding the atomic-level processes that govern stability in solid-state batteries. Furthermore, significant ambiguities exist over the extent of surface film formation (e.g., Li2CO3) that takes place during cell assembly and sample preparation, complicating the interpretation of such analyses.13 In addition to ambiguities associated with interfacial chemistry, relatively little information exists about the influence of crystallographic orientation/texture on the stability and overall performance of solid electrolyte materials. Although recent efforts have made progress in understanding the influence of grain boundary structure/chemistry on preferential Li migration and eventual cell failure due to dendrite formation along grain boundaries,11,12 the fundamental stability of the individual grains remains largely unexplored. As anisotropies in ionic conductivity along various crystallographic orientations and Received: February 21, 2018 Accepted: April 30, 2018 Published: April 30, 2018 17471
DOI: 10.1021/acsami.8b03078 ACS Appl. Mater. Interfaces 2018, 10, 17471−17479
Research Article
ACS Applied Materials & Interfaces
Figure 1. (a) LEED and (b) AFM images of an etched and annealed STO(001) surface. (c) AFM image of the same STO(001) surface after Li deposition. Scale bars are 1 μm, and height scale units are in nanometers. (d−f) XPS core-level spectra before (red) and after (blue) Li deposition show (d) the clear presence of both Li metal and Li−O species, (e) the reduction of Ti4+ to Ti3+ and Ti2+ and (f) the reduction of Nb5+ to Nb4+ after Li deposition. Increased noise in the Ti 2p and Nb 3d core-level spectra after Li deposition is due to signal attenuation by the Li overlayer.
vacuum (UHV)-based approaches is the ability to verify the absence of interfacial films on both the solid electrolyte and Li metal prior to contact, helping to ensure that the observed interfacial chemistry is representative of the specific electrolyte material under investigation. Although such studies have significantly advanced the understanding of the intrinsic stability of solid electrolyte materials, existing methodologies still do not precisely recreate the conditions “real” materials will experience during cell fabrication and testing. For example, ion beam sputtering processes carried out near the sample surface produce highly energetic (i.e., mobile) Li species, causing transformations in solid electrolyte materials that may not otherwise occur at room temperature. Furthermore, it is unclear what role surface defects, such as those generated during Gaion beam thinning for TEM imaging, may play in (de)stabilizing the interface relative to conventional surface preparation techniques such as mechanical polishing. In this work, we establish a surface science-based approach for assessing the intrinsic stability of oxide materials in contact with Li metal to understand the impact of crystallographic orientation and surface morphology on the extent of the chemical reactions that take place. Through the use of magnetron sputtering to deposit thin films of Li metal on single-crystalline, Nb-doped SrTiO3 (Nb/STO) substrates at room temperature, a more representative interface is created for investigating the intrinsic stability of the Li−STO system.
differences in the reactivity of the various possible crystalline surface terminations will ultimately determine the intrinsic, long-term stability of any solid electrolyte material, developing deeper understanding of these properties may make it possible to design more stable and reversible solid-state electrochemical interfaces. By first working with well-defined, single-crystal model electrodes, it will be possible to develop the atomic-/ molecular-level insights necessary to understand the impact of orientation-dependent interfacial stability and transport properties on performance, as has been established in the past for solid−aqueous interfaces.14−16 Acquiring such insight is complicated by the fact that many existing in situ and operando analysis techniques are limited in providing meaningful information about the chemistry of interfacial processes in solid-state batteries, as the relevant interfaces are, by the nature of the cell architecture, inaccessible during cycling. In light of these difficulties, new methodologies are needed to understand the electrochemical transformations that take place at buried interfaces in solid-state devices. Significant progress has been made in recent years toward developing spectroscopic means of directly investigating interfacial stability in vacuo, using both X-ray spectroscopic17−20 and transmission electron microscopy (TEM)based21,22 techniques to analyze the presence or absence of interphase formation in a variety of solid electrolyte systems in contact with Li metal. The biggest advantage of such ultrahigh 17472
DOI: 10.1021/acsami.8b03078 ACS Appl. Mater. Interfaces 2018, 10, 17471−17479
Research Article
ACS Applied Materials & Interfaces
instrument. XPS analysis of the sample surface before air exposure shows significant Li intensity, with signal from the underlying STO substrate still clearly visible as well (Figure S2). After air exposure, however, XPS indicates that the surface is entirely covered by Li2CO3 (Figure S4), with signal from the underlying STO substrate no longer observable, consistent with volumetric expansion of the islands during conversion to Li2CO3 (Figure 2c). By assuming the as-deposited Li metal is entirely converted to Li2CO3 upon air exposure and using the expected volume expansion upon conversion, it is possible to
Although not a solid electrolyte material itself, STO serves as an ideal model system, as its surface chemistry has been extensively studied and atomically smooth, single-crystal surfaces can be readily prepared with known surface termination.23−26 Furthermore, Ti and Nb are well known to be unstable to reduction by Li metal,27,28 enabling facile validation of our in vacuo methodology. We clearly demonstrate the ability of our approach to resolve the reduction of surface species after Li deposition, with nominally the same amount of Ti and Nb reduction on STO(001), (111), and (110) surfaces. Surprisingly, despite the similar extent of chemical reduction that takes place, significant differences exist between the morphological stability of the three surfaces. The STO(110) surface exhibits significant changes after the deposition and subsequent removal of Li metal, whereas the STO(001) and (111) surfaces remain relatively unperturbed. Density functional theory (DFT) calculations reveal that the observed morphological instability of STO(110) can be explained by destabilization of the surface TiOx layer due to incorporation of Li underneath this layer. Finally, we utilize the above approach to investigate the intrinsic stability of the technologically relevant, polycrystalline Nb-doped lithium lanthanum zirconium oxide (Li6.5La3Zr1.5Nb0.5O12, LLZO) system, which is also found to react with Li metal through the reduction of Nb and a weak, but discernible, interaction with La. These results clearly demonstrate the feasibility of our approach to assess the intrinsic (in)stability of oxide materials for solid-state battery systems.
■
RESULTS Deposition of Li on STO(001). To understand the intrinsic chemistry of Li metal in contact with a model oxide surface, Li was deposited directly onto STO(hkl) surfaces and then transferred directly between the sputter and analysis chambers under ultrahigh vacuum (UHV) conditions (≤5 × 10−9 mbar throughout transfer; see Experimental Methods for more details). Central to these investigations was the preparation of well-defined, single-crystal surfaces to identify differences in stability and reactivity based on crystallographic orientation. We begin by summarizing results on STO(001) surfaces to illustrate our general approach, and Figure 1 illustrates structural and chemical characterization of a representative STO(001) surface both before and after Li deposition. From low-energy electron diffraction (LEED, Figure 1a), a 1 × 1 surface reconstruction is clearly resolved on the as-prepared STO(001) surface and from atomic force microscopy (AFM, Figure 1b), the well-defined crystal terraces of hydrofluoric acid (HF)-etched and annealed23 STO(001) surfaces can be easily seen. X-ray photoelectron spectroscopy (XPS) analysis of the Ti 2p, Nb 3d, and Sr 3p core levels indicates that the pristine STO surface is fully oxidized, with only Ti4+, Sr2+, and Nb5+ oxidation states and minimal surface contamination by adventitious carbon (Figure 1d,e, Supporting Information (SI) Figures S1a and S2). AFM imaging of the sample after Li deposition reveals an island morphology (Figure 1c), which is unsurprising, given the low temperature used to deposit Li metal (25 °C). The average island size depends on the length of the Li deposition (Figure S3), with typical samples analyzed in this work having islands ∼500 nm in size. Although the film morphology shown in Figure 1c should be qualitatively similar to that of the pristine Li deposit, the island size itself is not truly representative as the sample was exposed to air during transfer to the AFM
Figure 2. X-ray absorption spectroscopy (XAS) measurements of STO substrates in (a) total electron yield (TEY) and (b) total fluorescence yield (TFY) modes before (red) and after (blue) Li deposition reveal surface-limited reduction by Li. (c) Schematic of surface structure after Li deposition. XPS signal from reduced Ti species comes from the edges of large Li islands, with the remaining Ti4+ signal coming from unreacted areas between Li islands. After air exposure, the islands coalesce due to volume expansion during transformation to Li2CO3, obscuring signal from the underlying substrate. 17473
DOI: 10.1021/acsami.8b03078 ACS Appl. Mater. Interfaces 2018, 10, 17471−17479
Research Article
ACS Applied Materials & Interfaces
Figure 3. AFM images of STO(001), (111), and (110) surfaces (a−c) before Li deposition and (d−f) after Li deposition/removal, with line profiles given at the bottom to highlight the differing extents of surface morphology change due to reaction with Li metal. Line profiles before (red) and after (blue) Li deposition/removal are taken along the black lines drawn on each image. All images were taken in identical locations before and after Li deposition. Height scales are given in angstrom and are the same for top and bottom images for each orientation. (001) and (110) images are 1 × 1 μm2, and (111) images are 2 × 2 μm2.
indicate charge transfer to Ti 3d states, consistent with the observed reduction of Ti and Nb in XPS. The persistence of Ti4+ and Nb5+ species after Li deposition indicates incomplete reduction of the STO substrate, which can be explained by two possible factors: (i) the reaction between Li and STO is self-limiting (i.e., only the first few monolayers of STO react), and/or (ii) the spaces between Li islands on the surface are essentially Li-free, resulting in regions of unreacted STO. As noted above, the deposition of thicker and thinner Li layers does not result in significant differences in the observed extent of reaction (Figure S3), suggesting that there is not a large fraction of uncoated surface regions between Li islands and that the reaction is instead surface-limited. Angle-resolved XPS (ARXPS) of the film yields slight changes in the relative intensities of the various oxidation states of Ti with increasing sample tilt angle (Figure S6). In the most surface-sensitive case (60° sample tilt angle), 55% of the total Ti signal corresponds to Ti3+ and Ti2+, as compared with 42 and 40% with 30° tilt and no tilt, respectively. This further indicates that the reactions are surface-limited, as tilting to higher angles results in an increase in the amounts of Ti3+ and Ti2+ relative to Ti4+. As the island morphology of the Li films makes it difficult to conclusively assign a surface-limited reaction via ARXPS analysis, X-ray absorption spectroscopy (XAS) measured at grazing incidence was also performed to measure the relative amounts of both surface and bulk reduction of STO(001). This analysis also suggests that the reaction between STO and Li is surface-
put a lower bound on the average diameter of the pristine Limetal islands of ∼90 nm (see the SI for details). Regardless of the true size of the as-deposited islands, they are significantly larger than the escape depth of Al Kα-generated photoelectrons through Li metal (≤16.5 nm).29 This implies that signal from the substrate observed by XPS must come from the regions very near the edges of the Li islands or from any thinly coated or Li-free regions between islands. Nevertheless, Figure S3 clearly demonstrates that the observed extent of Ti reduction does not change significantly with island size, indicating that the underlying chemistry measured with XPS is representative of the intrinsic reactivity of Li with STO. Analysis of the Li 1s core level on the as-deposited Li−STO sample surface (Figure 1d) indicates that the sputtered Li film is primarily present in the metallic state, with some of the Li also present as oxide. This suggests that a fraction of the Li metal reacts with the underlying STO substrate. Indeed, Figure 1e,f demonstrates that after Li deposition, ∼38% of the total Ti signal corresponds to a mixture of 3+ and 2+ oxidation states and 33% of the total Nb signal corresponds to the 4+ oxidation state. The oxidation state of Sr was unaffected by Li deposition (Figure S1a). These results are consistent with thermodynamic data (Figure S5) that indicate the reaction of Li with STO proceeds much easier via Ti reduction than Sr reduction. DFT calculations of the electronic density of states (DOS) for Li adsorbed on the STO(100) surface, discussed in detail below, 17474
DOI: 10.1021/acsami.8b03078 ACS Appl. Mater. Interfaces 2018, 10, 17471−17479
Research Article
ACS Applied Materials & Interfaces
Figure 4. DFT-calculated (a) structures, (b) energy changes, and (c) electronic DOS for Li atoms on (100) and (110) surfaces of STO with different surface terminations. The DOS for a TiO2(001) surface without Li is shown at the top of (c) as a reference.
soluble in water, samples were cleaned ultrasonically in deionized water to remove the Li film from the surface prior to AFM imaging of the reacted STO surface. XPS measurements of STO(hkl) samples afterward reveal that all residual Li is indeed removed by sonication and the surface is fully reoxidized to Ti4+ and Nb5+ (Figure S7). Surprisingly, although AFM imaging of the STO(001) and (111) surfaces before Li deposition and after Li removal reveals little to no changes in the surface morphology (Figure 3a/d,b/e, respectively), the STO(110) surface morphology appears to be substantially modified after deposition and removal of the Li layer (Figure 3c,f), with the morphological changes strongly correlated with the initial morphology. Line profiles taken from the same locations on each surface before and after Li deposition/ removal show the difference in scale of these features (Figure 3, bottom), with randomly distributed, ∼1 Å features present on STO(001) and no differences observed on STO(111). This stands in strong contrast to the ∼2−3 Å features present at the STO(110) that appear to be correlated with the (110) step edges. Note that because of the arbitrary reference of the relative heights of line profiles taken before and after Li deposition/removal, the terrace heights are aligned to compare the profiles before and after Li deposition/removal. Given the lack of any differences in the apparent extent of surface reduction revealed by XPS analysis, coupled with the observation that all three STO(hkl) surfaces are fully reoxidized after sonication (i.e., there is no residual chemical reaction layer present on the surface), this difference in morphological stability on STO(110) relative to (001) and (111) is quite puzzling. To gain insight into the possible structural origins of this difference in morphological stability, DFT simulations were performed to understand the underlying thermodynamic factors that result in the lower stability of the STO(110) surface. The change in energy upon Li adsorption relative to the bulk Li was defined as Eads = −1/n[E(nLiSTO) − E(STO) − nE(Li)], where E(nLiSTO) is the total energy of an STO surface with n adsorbed Li atoms and E(STO) and E(Li) are total energies of the STO slab without Li and body-centered cubic bulk Li metal (per atom), respectively. Figure 4a shows
limited, as measurements of the Ti L-edge in total electron yield (TEY) and total fluorescence yield (TFY) modes before and after Li deposition reveal differing amounts of reduction (Figure 2a,b). The TEY spectra, which are sensitive to the top few monolayers (
