Insights into the reactivity and lithium plating mechanisms of ultra-thin metal oxide coatings for anode-free solid-state lithium metal batteries

Solid-state batteries (SSBs) in an “anode-free” cell format using lithium metal anodes are the best candidates for high energy density battery applications. However, low lithium metal Coulombic efficiency and charge loss due to solid electrolyte interphase (SEI) formation severely limit the cycle life of anode-free SSBs. Here, we explore ultra-thin (5–20 nm) Al 2 O 3 and ZnO coatings deposited by atomic layer deposition (ALD) on copper electrodes for anode-free cells with a solid polymer electrolyte. Voltammetry shows that lithium inventory loss from SEI formation is reduced over 50% with Al 2 O 3 @Cu electrodes, but these electrodes experience orders of magnitude higher interface resistances than bare Cu and ZnO@Cu electrodes due to low ionic and electronic conductivities. The electrochemical differences are reflected in XPS, where Al 2 O 3 undergoes a self-limiting lithiation reaction with Li 0, while ZnO reacts completely with Li 0 to form LiZn and Li 2 O. These chemical differences result in higher and lower lithium plating nucleation overpotentials for Al 2 O 3 (up to 220 mV) and ZnO (down to 15 mV) coatings, respectively, relative to uncoated Cu electrodes (35 mV). ToF-SIMS reveals lithium plating underneath a Li y AlO x coating and through emergent defects and pinholes with Al 2 O 3 @Cu electrodes, while it plates exclusively on top of converted ZnO@Cu electrodes. SEM corroborates these mechanisms, showing sparse coverage of isolated Li clusters plated with Al 2 O 3 @Cu electrodes, while Cu and ZnO@Cu grow more dense and interconnected deposits. Despite both coatings improving different aspects of anode-free battery design, unmodified Cu electrodes show higher Coulombic efficiencies (∼77%) than Al 2 O 3 @Cu (up to 70%) and ZnO@Cu (up to 75%) electrodes. Increasing Al 2 O 3 thickness decreases the practical current density compared to unmodified Cu (30 µA/cm 2), but increasing ZnO thicknesses can double or triple this value. These (electro)chemical and morphological observations suggest two mechanisms: less-reactive metal oxides develop lithium ion conductivity through their structure to plate lithium underneath, while more-reactive metal oxides undergo full reduction and conversion with lithium plating above the coating. This fundamental research opens future work to leverage these mechanisms and explore other materials for high-efficiency anode-free SSBs.

The further development of rechargeable batteries with high energy density, long cycle life, and low cost are necessary for future portable electronics, electric vehicles, and energy storage at large (Albertus et al., 2021). Solid-state batteries (SSBs) with solid electrolytes and lithium metal anodes can have a higher energy density than typical lithium-ion batteries (Bonnick and Muldoon, 2022), especially in “anode-free” cell configurations that use no lithium excess on the anode side (Heubner et al., 2021; Huang et al., 2022). The success of an anode-free battery heavily relies on the even and efficient plating and stripping of Li metal on the current collector over time, a challenging goal (Salvatierra et al., 2021; Fuchs et al., 2023). During the plating steps with most electrolytes, there is the formation of a solid electrolyte interphase (SEI) layer, which passivates the lithium anode but also creates an irreversible loss of lithium in the battery, reducing the usable energy in the cell (An et al., 2016; Peled and Menkin, 2017; Mao et al., 2018; Hobold et al., 2021; Wu et al., 2023). Additionally, the formation of lithium dendrites that lead to more capacity loss and cell shorting is even more of an issue in anode-free cells (Weber et al., 2019; Lewis et al., 2023). One of the strategies to address these issues is developing an artificial SEI that can offer fast Li-ion transport, promote uniform lithium plating, suppress Li dendrite growth, and prevent side reactions between lithium metal and electrolyte (Yu et al., 2020; Tamwattana et al., 2021; Tong et al., 2021; Sun et al., 2023).

Atomic layer deposition (ALD) is an advanced deposition technique for developing an ultrathin and uniform artificial SEI, offering a variety of metal oxide coatings to improving battery performance. However, the fundamental understanding of how these coatings help has not been well understood. Thin conformal layers of metal oxides like Al 2 O 3 or ZnO have been used extensively on lithium-ion battery cathode particles to prevent degradation on the positive electrode (Li et al., 2014; Park et al., 2014; Cabana et al., 2018; Neudeck et al., 2019; Cheng et al., 2020; Kaur and Gates, 2022). On the anode side, however, these coatings are relatively understudied, especially with lithium metal anodes. For lithium-ion batteries, ALD Al 2 O 3 and other coatings have been shown to improve capacity retention and cycling rates with graphite (Jung et al., 2010), graphene-like (Lahiri et al., 2011; Yu et al., 2014), TiO 2 (Sopha et al., 2017), and silicon anodes (Shin and Cho, 2018). Direct deposition of Al 2 O 3 on lithium metal anodes to generate a lithiated lithium aluminate (Li y AlO x) coating has been shown to enhance cycling in liquid (Kazyak et al., 2015; Kozen et al., 2015; Chen et al., 2017) and solid polymer electrolytes (Wang et al., 2018). This coating can help prevent side reactions in other ways, such as capacity loss from polysulfide shuttling in Li-S batteries (Kozen et al., 2015). These studies did not reveal any particular performance dependence on the ALD layer thickness. Much less work has been done on these coatings for true anode-free applications, especially in solid-state systems. Mohammadi et al. showed that increasing ALD ZnO thickness on Ni foam electrodes greatly improved lithium Coulombic efficiency, achieving 99.0% in ether-based liquid electrolyte (Mohammadi et al., 2023). A similar trend was seen with 10 nm SnO x coatings on Cu electrodes, also with liquid carbonate electrolytes (Kim et al., 2023a). In SSBs, Li 2 Te conversion-type interlayers have been shown to greatly improve lithium plating and efficiency with sulfide electrolytes by improving wetting at the current collector interface (Wang et al., 2023).

In this work, we investigated the reactivity of two metal oxide thin films prepared by ALD on copper current collectors: Al 2 O 3 and ZnO. Half cells using Cu, Al 2 O 3 @Cu, and ZnO@Cu electrodes were tested using a solid polymer electrolyte and showed markedly different interphase formation, interface impedance, and anode performance depending on the chemistry and coating thickness. The electronically insulating Al 2 O 3 @Cu limits capacity loss during SEI formation but at the expense of higher interface resistance, while ZnO@Cu consumes more Li than Cu but has lower resistance, reflected in both electrochemical impedance spectroscopy and overpotential analysis. These differences are ascribed to fundamental differences in reactivity, as shown by X-ray photoelectron spectroscopy (XPS), where Al 2 O 3 lithiates to form a Li y AlO x compound while ZnO converts to a LiZn alloy and Li 2 O upon Li deposition. Secondary ion mass spectrometry (ToF-SIMS) 3D imaging and scanning electron microscopy (SEM) imaging show this results in lithium metal plating both underneath the ALD layer and through emergent defects and pinholes in Al 2 O 3 @Cu coatings but on top of ZnO@Cu coatings. All of these phenomena are connected to cell performance quantified by Coulombic efficiency and shorting current density measurements. This work provides a first step towards developing ultra-thin metal oxide coatings as artificial SEIs for anode-free SSBs.

2.1 Materials

Trimethyl aluminum (TMA, 98%) and diethyl zinc (DEZ, 95%) were purchased from Strem (United States). Poly (ethylene oxide) (PEO, Mw = 600 kDa), lithium bis(trifluoromethylsunfonyl)imide (LiTFSI, 99.95% metals basis), acetonitrile (ACN, anhydrous, 99.8%), dimethoxyethane (DME, anhydrous, 99.5%), and hydrochloric acid (HCl, ACS reagent) were purchased from Sigma-Aldrich. PEO was dried under vacuum at 60 °C for 2 days, and LiTFSI was dried under vacuum at 150 °C for 5 days, before use. Lithium bis(fluorosulfonyl)imide (LiFSI, 99.9%, <20 ppm H 2 O) was purchased from Solvionic, and lithium nitrate (LiNO 3, anhydrous, 99.999% metals basis) was purchased from Fisher Scientific. Certified ACS acetone and isopropanol were purchased from Fisher Scientific. Copper foil (6 µm thickness, 99.8%) was purchased from MSE Supplies, aluminum foil (50 µm thickness, 99.0%) was purchased from Goodfellow, and zinc foil (250 µm thickness, 99.95% metals basis) was purchased from Fisher Scientific. Lithium chips (250 µm thickness) and stainless steel 316 coin cell components were purchased from MTI Corporation.

Working electrodes were punched into 1.0 cm 2 discs. Cu, Al, and Zn electrodes with native oxide on the surfaces were rinsed with acetone and isopropanol; electrodes referred to simply as “Cu” below have the native oxide in this way. Cu-HCl electrodes were first exposed to dilute hydrochloric acid for 30 s to remove the native oxide from as-received foil, then rinsed four times with excess acetone. ALD-coated Cu electrodes were punched and dried with no rinsing steps after deposition. All electrodes were dried under vacuum before transferring to the glovebox. Celgard 2,500 separators (25 µm thickness) were punched to 16 mm diameter and wetted with 26 µL electrolyte for liquid electrolyte cells.

PEO-LiTFSI solid polymer electrolytes were prepared by a solvent casting method. PEO and LiTFSI (EO:Li 15:1 mol:mol ratio) were dissolved in acetonitrile for a 6.7 wt% total dissolved solids solution in the glovebox. This mixture was stirred to homogeneity overnight, then coated onto a PTFE film (McMaster Carr) with a doctor blade. This was allowed to dry ambiently for several hours before vacuum drying at 60 °C overnight to remove all residual solvent. The film was then punched to 16 mm diameter and carefully peeled from the PTFE film to yield a freestanding membrane with thickness typically around 50 µm.

2.2 Atomic layer deposition

Copper foil was cut into 4 x 6 cm pieces, covered with laboratory-grade aluminum foil on the bottom side, and gently cleaned with acetone and isopropanol to remove surface contaminants. The native oxide was left untouched to provide a better surface for the deposition process. This was then transferred to the atomic layer deposition (ALD) chamber for coating. The backing foil was used to prevent deposition on the back side of the copper that would produce an electronically insulating layer and was removed prior to cell assembly.

The Al 2 O 3 and ZnO ALD were performed using a custom hot-walled, viscous flow reactor similar to the one discussed in a previous report (Elam et al., 2002). In a typical coating run, the protected copper substrate and two pieces of Si(100) wafer were loaded onto a sample tray as shown in Supplementary Figure S1 and installed in the ALD system. The Si(100) coupons were used to verify the ALD coating thickness and uniformity using spectroscopic ellipsometry measurements (alpha-SE, J. A. Woollam Co.). All the ALD coatings were carried out at 150 °C using ultrahigh purity (99.999%) nitrogen as carrier gas at a flow rate of 225 sccm, and the base pressure of the reactor was ∼0.68 Torr. Al 2 O 3 and ZnO ALD used alternating exposures of TMA or DEZ, respectively, and deionized H 2 O, where all three precursors were maintained in room-temperature containers. The typical Al 2 O 3 and ZnO ALD timing sequence can be described as t 1 –t 2 –t 3 –t 4, where t 1 = 1 s and t 3 = 1 s are the exposure times for TMA or DEZ and H 2 O, respectively, and t 2 = 25 s and t 4 = 25 s are the corresponding purge times used between the precursor doses. The Al 2 O 3 and ZnO ALD growth per cycle were found to be 1.18 Å/cycle and 1.31 Å/cycle, respectively. The Al 2 O 3 ALD used 45 and 85 ALD cycles to produce coating thicknesses of ∼5 nm and ∼10 nm, and the ZnO ALD used 40, 80, and 160 ALD cycles to produce coating thicknesses of ∼5 nm, ∼10 nm, and 20 nm. Copper electrodes coated with Al 2 O 3 and ZnO are denoted as Al 2 O 3 @Cu and ZnO@Cu, respectively.

2.3 Electrochemistry

All electrochemical tests were performed in 2032 coin cells using bare or ALD-coated copper as the working electrode (1.0 cm 2 area) and freshly polished lithium foil (0.9 cm 2 area) as the counter electrode. Freestanding PEO-LiTFSI (2.0 cm 2 area) was used as the electrolyte. Two 0.5 mm steel spacers and wave spring added to the cell stack to maintain pressure. Cells were then slowly heated from 20 °C to the operating temperature of 60 °C in a temperature-controlled chamber (Espec). All measurements were performed with a VSP-300 potentiostat (Biologic). Potentiostatic electrochemical impedance spectroscopy (EIS) was performed with a frequency range of 7 MHz to 1 Hz, DC bias of 0.0 V vs open circuit or 0.1 V vs Li 0 when noted, AC amplitude of 10 mV, and sampling of 20 points per decade. Prior to any cycling, cyclic voltammetry (CV) was performed with a scan rate of 0.2 mV/s, scanning negative from the cell open circuit potential to 0.1 V then to 1.5 V and back to 0.1 V. The first step is shown in Figure 1 and Supplementary Figure S2 here as a linear sweep voltammogram (LSV) for clarity. Galvanostatic cycling was then performed in 10 µA/cm 2 increments (as described in the Results section) with a stripping potential limit of 0.5 V; a constant voltage was then held at 0.5 V until the current dropped to 1 µA/cm 2 to ensure all lithium was stripped from the electrode. The cells were rested at open circuit for 1 minute before performing EIS steps.

2.4 X-ray photoelectron spectroscopy and Li deposition

A combination of Li deposition and subsequent surface analysis was performed in a home-built ultra-high vacuum (UHV) system (10 −10 mbar of base pressure) where a glove box, a surface analysis module, and a physical vapor deposition (PVD) module are interconnected through a UHV linear transfer system to avoid any air exposure. Samples were initially mounted in a glove box and transferred to the surface analysis module. X-ray photoelectron spectroscopy (XPS) was performed using a Specs PHOIBOS 150 hemispherical energy analyzer with a monochromatic Al K α X-ray source (1486.7 eV). The survey spectra were measured using a pass energy of 40 eV at a resolution of 0.2 eV/step and a total integration time of 0.1 s/point. The core level spectra were measured using a pass energy of 20 eV at a resolution of 0.05 eV/step and a total integration time of 0.5 s/point. After surface analysis in the pristine state, the samples were transferred to the PVD module. Li deposition was performed via e-beam evaporation for 15 min for all samples as described previously (Connell et al., 2020), and the coated samples were subsequently transferred to the surface analysis module for following XPS measurements. The survey and core level spectra of the Li deposited surfaces were measured in the same manner. Peak analysis was performed using CasaXPS software with a Shirley-type background and 70-30 Gaussian-Lorentzian peak shapes. The sp 3 signal in C 1s (284.8 eV) was used for charge referencing.

2.5 ToF-SIMS

Time-of-flight secondary ion mass spectrometry (ToF-SIMS) imaging and depth profiling analysis were carried out using a commercial ToF-SIMS instrument (M6, IONTOF, Münster, Germany). A focused 30 keV Bi 3 + primary (probing) ion beam was generated by a pulsed liquid metal ion gun (LMIG) equipped with a Wien filter (∼0.26 pA current internal to the LMIG). Depth profiling was performed by sputtering a 1 keV beam of O 2 from a dual source ion column (DSC-S) equipped with a Wien filter with total sample exposure of 2.8 × 10 17 ions/cm 2 (∼250 nA current internal to the DSC-S). The primary beam raster size (pixels) was 128 μm × 128 µm with a field of view of 50 μm × 50 µm and a sputter beam crater size of 300 μm × 300 µm. The flood gun was kept on during the measurements in order to manage surface potential by reducing sample charging. Cycled, rinsed, and dried samples with plated lithium sealed in coin cells were opened in an argon filled glove box, mounted with copper tape on a top-mount sample plate, and placed in an air-tight sample transfer box. Thus, samples were not exposed to air at any point in time. Five data sets were collected at different regions on each sample, with continuous collection of ToF-SIMS data during continuous DSC-S sputtering for 160 s to obtain the depth profiles. All the 3D images and depth profiles were generated using SurfaceLab 7.3.134065 with ion signals plotted as recorded, without normalization.

2.6 SEM-FIB

The cycled sample morphology was characterized using scanning electron microscopy (SEM) via secondary electron imaging in a Thermo Fisher 5CX Helios SEM-FIB. Prior to imaging, samples were transferred from an Ar glovebox into the SEM using an air-free transfer workflow. The sample was prepared and loaded into a transfer container in the Ar-filled glovebox. The transfer container was then cooled in a liquid nitrogen bath to liquid nitrogen temperature with the sample inside. The cooled container was opened to N 2 atmosphere, exposing the sample to nitrogen gas only at liquid nitrogen temperatures. The sample was then quickly transferred into a vacuum cryo transfer system (Leica EM VCT500), maintaining an atmosphere of cold N 2 and a temperature below −140°C, before the cold N 2 gas was pumped out. The cooled transfer shuttle was then connected to the chamber of the SEM and the sample transferred in vacuum to the SEM stage, which was also cooled to liquid nitrogen temperatures.

During characterization, imaging was performed using a beam voltage of 5 kV and a current of 0.69 nA. Energy dispersive spectroscopy (EDS) was conducted using an Oxford detector to confirm the elemental distribution of the Cu substrate, coatings, and plated products. EDS measurements were done using different voltages, including 5 kV and 20 kV, under a high beam current such as 0.69 nA. The depth and density of the plated products were also investigated with focused ion beam (FIB) milling, where Ga ions were used at a voltage of 30 kV and a current range of 1–47 nA. Larger currents were used to do rough milling of a larger volume, while lower currents were used to do fine polishing and cleaning of the cut cross-sectional area. After milling and polishing, the exposed cross-sections of the grown products were imaged and analyzed with EDS.

3.1 Voltammetry of Oxide Layers

To begin interrogating the effect of oxide coating layers in SSBs, we tested a simple PEO-LiTFSI solid polymer electrolyte with Cu, Al 2 O 3 @Cu, and ZnO@Cu electrodes. Copper, a metal that shows very little alloying with lithium metal under these conditions (Rupp et al., 2019; Huang et al., 2023), was tested with the native oxide untouched to ensure better comparison to the artificial metal oxides deposited with ALD. PEO-LiTFSI was used as a solid electrolyte for two main reasons. First, it is a well-studied, versatile solid electrolyte system that is simple to fabricate (Xue et al., 2015). Second, unlike other solid electrolytes that suffer from electrode contact loss during cycling due to their hardness and inflexible nature, PEO-LiTFSI should provide a soft electrode-electrolyte interface that maintains contact throughout the cell lifetime (Janek and Zeier, 2016; Homann et al., 2020). This enables us to exclude delamination between the electrode and electrolyte as a failure mode here.

Figures 1A–C shows representative LSVs for Cu with native oxide, 5 and 10 nm Al 2 O 3 @Cu, and 5, 10, and 20 nm ZnO@Cu. The reductive charge passed, as integrated from voltammetry prior to 0.0 V vs Li 0 (the onset of lithium plating), is shown in Figure 1D and is analogous to the charge lost during the formation step in a practical battery (An et al., 2016; Mao et al., 2018). On the surface of Cu with native oxide, we observed two distinctive groups of reduction signals: reduction of native oxide above 2.0 V vs Li 0 and SEI formation from LiTFSI and PEO at around 1.2–1.4 V and 0.8–0.9 V vs Li 0 (Ushakova et al., 2022). With Al 2 O 3 coatings, much lower current densities (<2 µA/cm 2) and reductive charges than Cu electrodes are seen, with current density and charge decreasing with increasing coating thickness (Figures 1B,D). We observed reduction signals only below 2.0 V vs. Li 0, with the lack of CuO x reduction indication that any pinholes in the film that would expose bare Cu are very small in area. The lower current densities agree with the expected, extremely low electronic conductivity of Al 2 O 3 (10 −16 S/cm) (Kusunose et al., 2019) that hinders electron transfer to the electrolyte.

In contrast, ZnO@Cu electrodes show similar magnitude of current densities to the Cu electrode (2–7 µA/cm 2) with peaks related to PEO-LiTFSI reduction and ZnO reduction below 1.5 V. The presence of reduction peaks above 2.0 V might be related to exposed CuO x caused by non-conformal coatings, especially with thinner ZnO. Regardless, the larger current densities from 5 to 20 nm ZnO compared to Al 2 O 3 @Cu indicate that ZnO layers of any thickness have functionally high electronic conductivity (likely in the range of 100 S/cm) (Elam et al., 2003), since electrons can move from the Cu current collector to the electrolyte interface and replicate the same current density as unmodified Cu. Additionally, 20 nm ZnO@Cu shows higher current densities and charges related to oxide reduction and alloying (Figure 1C inset), as discussed later. It was also observed with voltammetry that the native oxides on Al and Zn foil electrodes behave similar to ALD coatings—aluminum oxides are electronically insulating and lead to low current densities, while zinc oxides are electronically conductive and show similar SEI formation current densities as copper electrodes (Supplementary Figure S2).

Figure 1D shows that coating Cu with Al 2 O 3 lowers the lithium inventory loss, while coating with ZnO tends to increase this loss. From a practical perspective, the Al 2 O 3 @Cu electrodes would be desirable in anode-free SSBs since the lithium inventory loss during SEI formation is reduced by at least 50% compared to Cu alone, thus increasing the usable capacity after the formation step. Although the 2–22 µAh/cm 2 lithium losses reported here are small compared to the >3 mAh/cm 2 required for practical applications (Albertus et al., 2018), this is related to total electrode surface area, not just geometric area. For 3D structured anodes (Park et al., 2020; Pathak et al., 2021), this lithium loss number quickly rises compared to the planar 2D electrodes used here, significantly eating into the usable capacity. Since ALD is a useful process for the conformal coating of 3D structures, developing ALD coatings as artificial SEIs may be a valuable strategy for limiting lithium loss with 3D structured electrodes (Kim et al., 2017).

The origin of this lithium loss may be different as well as the coating chemistry changes. Cu electrodes lose charge solely to SEI formation and oxide reduction if the native oxide is present (∼0.3 µAh/cm 2 oxide reduction is seen here). Al 2 O 3 @Cu could lose charge to SEI formation and reduction of the Al 2 O 3 coating. Assuming the maximum of a three-electron transfer to every Al 3+ ion in the coating, the estimated charge is 1.0 µAh/cm 2 for 10 nm Al 2 O 3 @Cu (approximately 0.1 µAh/cm 2 per 1 nm Al 2 O 3). Given the difficulty of reducing Al 2 O 3, the lack of clear peaks in voltammetry, and the observed charge being at least 2 µAh/cm 2 for all coatings, some or all charge must go towards SEI formation. This is observed with the small peaks in Figure 1B being in the same potential range as PEO-LiTFSI reductions peaks seen with Cu electrodes. For ZnO@Cu, charge is also lost through SEI formation and the reduction of ZnO. Assuming a two-electron transfer to every Zn 2+ ion in the coating, the estimated charge for 20 nm ZnO@Cu is 5-8 µAh/cm 2 (approximately 0.4 µAh/cm 2 per 1 nm ZnO). This helps explain why charge loss is greater with ZnO over Cu, as the ZnO is much more redox active than Al 2 O 3, and the electronic conductivity of ZnO@Cu coatings explains a similar charge (∼9 µAh/cm 2) being lost to SEI as with bare Cu (Figure 1D). The reduction charge of 20 nm ZnO@Cu electrodes (20 µAh/cm 2) may be larger than the straight addition of SEI and ZnO reduction charges (estimated 14–17 µAh/cm 2) due to the increased surface area after reduction causing more SEI formation than expected based on the geometric area; see the Discussion section below.

3.2 Interface resistance

As demonstrated above, the electronic conductivity of the ALD layers can control capacity fade by limiting or increasing the amount of charge lost before lithium plating. In addition to electronic differences that contribute to battery lifetime, these interface layers have different ionic conductivities that control the effective cell resistance, which affects the energy efficiency and power density of the battery during operation. This interface resistance is expected to change as the coatings undergo (electro)chemical changes. Figure 2A shows the modulus of impedance, |Z|, calculated at 1 Hz from impedance data before any electrochemistry was performed in the cell. This represents a rough estimate of the cell impedance at an un-pinned open circuit potential, which can be increased if the ALD coating is resistive even in the absence of a redox process. With the resistive Al 2 O 3 coatings, the impedance modulus is at least triple that of Cu alone (56 kΩ cm 2). Meanwhile ZnO coatings have little effect (within error) on the impedance, again indicating their conductive nature.

Figures 2B,C show impedance data and interface resistance (R Int) values following electrochemical reduction to form the SEI and/or react the ALD coatings following the voltammetry in Figure 1. R Int is represented by the medium-frequency semicircle apparent in all Nyquist plots in Figure 2B and is taken to be the sum of the formed SEI’s and the coating layer’s resistances. Following electrochemical reduction before the onset of Li plating, R Int values on the Cu electrodes are around 450 Ω cm 2 and reflect the resistance of the SEI layer. ZnO@Cu electrodes show overall lower impedances (300–450 Ω cm 2 for 5–10 nm ZnO@Cu), especially with 20 nm ZnO@Cu (around 126 Ω cm 2). R Int values for 10 nm ZnO coatings are slightly higher than 5 nm ZnO@Cu coatings, which may be due to the thicker Li 2 O layer formed from the ZnO conversion (vide infra), effectively increasing the SEI thickness. The decrease from 5 to 10 nm ZnO@Cu to 20 nm ZnO@Cu may be related electroactive surface area increases (thus lowering the geometric area resistance R Int) alluded to above. On the other hand, Al 2 O 3 @Cu electrodes have R Int values exceeding 10 kΩ cm 2 after going to reducing potentials, indicating that these layers have very slow ion transport (slower than the SEI alone), as R Int reflects the conductivity of ions (i.e., Li +) through the electrolyte-surface layers-electrode interfaces.

After electrochemically reducing the interface to form the SEI and reduce the ALD coatings as much as possible with applied potential, lithium metal was plated with galvanostatic cycling (Supplementary Figure S3). Once Li present at the interface, there is the possibility of further chemical reduction of the coating. Impedance data for the same half cells with Li metal plated are shown in Figure 2D & 2E. For Cu electrodes, the impedance decreases slightly (by ∼130 Ω cm 2), something we often observe once Li is on the surface and the potential of the working electrode is pinned to that of a stable redox couple (Li +/0). The greatest changes are seen for Al 2 O 3 @Cu after Li plating, where the R Int for 5 nm Al 2 O 3 @Cu is 37-fold decreased to around 295 Ω cm 2, similar to that of Cu and ZnO@Cu electrodes. This could be attributed to reaction of the coating to increase Li + transport through Al 2 O 3 or to large surface areas of now-exposed lithium electrode, mechanisms that will be discussed more below. For 10 nm Al 2 O 3 @Cu, the interface resistance after Li plating is still several kΩ cm 2 due to slow Li + conductivity and low electronic conductivity through the layer, and many cells short due to dendrite formation before enough lithium can be plated to get a reliable EIS measurement. As with electrochemical reduction (Figure 2B), ZnO@Cu electrodes show lower R Int values than Cu alone, and R Int decreases with increasing thickness.

3.3 Reactivity of Al 2 O 3 vs ZnO with lithium

To further understand the chemical reactivity of the ALD coatings upon Li contact, we compared the surface chemistries before and after Li deposition via e-beam evaporation. A combined XPS and Li deposition protocol in the ultra-high vacuum system (described in the Materials and Methods section) used previously in numerous solid electrolyte systems (Zhu et al., 2019; Connell et al., 2020; Liu et al., 2022) was applied to rule out air exposure or undesired surface contamination. Surfaces of as-prepared Al 2 O 3 @Cu and ZnO@Cu samples were initially examined via XPS prior to the Li deposition, and the Li-deposited surfaces were subsequently analyzed via XPS again.

Core level spectra and optical images of the Al 2 O 3 @Cu and ZnO@Cu surfaces before and after Li deposition are shown in Figure 3. For Al 2 O 3 @Cu, a clear Al 2 O 3 signal at 74.2 eV is observed in Al 2p core level spectra on the pristine surface (Figure 3A, top), with superpositioned Cu 3p core levels derived from the Cu substrate (Cu 0 and Cu 2 O at 74.8 eV and CuO 76.7 eV) (Khalakhan et al., 2021). Upon contact with Li 0, we observed a decrease in Al 2 O 3 peak intensity and a new peak emerging at 73.5 eV, which we tentatively assign as Li y AlO x (Figure 3A, bottom). This shift is commensurate with Al 2 O 3 coatings grown directly on lithium metal (73.2 eV) (Chen et al., 2017) and coatings on argyrodite solid electrolytes in contact with lithium (73.4 eV) (Hood et al., 2023). Meanwhile, ZnO shows a more distinct and clear reduction reaction with Li 0 (Figure 3B). The surface of the as-prepared ZnO@Cu shows a ZnO signal at 1022.1 eV in Zn 2p 3/2 core level spectra (Figure 3B, top), while the Li-deposited surface mainly exhibits alloying between Zn and Li to form a reduced peak at 1020.5 eV along with some residual ZnO (Figure 3B, bottom).

The O 1s and Li 1s core level spectra corroborate these mechanisms of the Li-derived reaction. In Figure 3C (top) and Supplementary Figure S4c, the O 1s core level spectra of Al 2 O 3 @Cu show a typical Al 2 O 3 feature with surface hydroxyls and some residual carbon-oxygen species at higher binding energies. This oxygen signal overall shifts to lower binding energies upon lithium deposition, indicating partial reaction of the ALD Al 2 O 3 coating leading to a more reduced environment (in agreement with the Al 2p spectra). Due to the complexity of O 1s core level spectra, especially for metal oxides, in-depth deconvolution of this narrow binding energy range is challenging (Supplementary Figure S4c). Some lithium oxide species may form at the Li-Al 2 O 3 @Cu interface as well, as apparent at lower binding energies (Figure 3C, top). We assume that the lithium oxide species mainly consist of Li-OH and Li-O bonding, which we carefully speculate originated from the residual hydroxyl species on the ALD Al 2 O 3 surface that reacted upon Li deposition and not necessarily from the Al 2 O 3 layer, although this is worth investigating in future.

A similar trend is seen in the O 1s core level spectra on the ZnO@Cu surfaces, as shown in Figure 3C (bottom) and Supplementary Figure S4d. The as-deposited coating contains Zn-OH and sub-stoichiometric ZnO 1-x in addition to ZnO. Oxygen vacancies in the ZnO coating might explain the high electronic conductivity of the coated electrode, leading to the trends in Figures 1C,D and Figure 2. Upon Li deposition, we observed the same overall shift of the O 1s spectra to lower binding energy like with Al 2 O 3 @Cu, this time with a clear Li 2 O signal emerging at 528.5 eV, as shown in Figure 3C (bottom). This indicates a near-complete conversion reaction of ZnO@Cu in contact with Li.

In the Li 1s core level spectra (Figure 3D), metallic Li is observed at 52.5 eV on the surface of the Li-deposited Al 2 O 3 @Cu, while the ZnO@Cu surface does not show any evidence of this after Li deposition. Given the asymmetric shape of the peak at 54.7 eV on the Li-deposited Al 2 O 3 @Cu surface, we speculate a co-existence of Li y AlO x and Li oxide species, which is consistent with the Al 2p and O 1s core level spectra. The Li-deposited ZnO@Cu surface exhibits a major signal at 54.5 eV with a small shoulder at 53.7 eV, which is probably derived from a combination of Li 2 O formation and Li-Zn alloying, respectively. These results suggest that while both chemistries show reactivity upon Li deposition, Al 2 O 3 coatings react in a self-limiting way to form a Li y AlO x phase that eventually becomes stable to Li 0, while ZnO undergoes a full conversion reaction to LiZn alloy and Li 2 O, something seen visually in Figure 3E. These differences in reactivity help to explain both the trends in pre-plating charge in Figure 1D and differences in initial interfacial impedance in Figure 2.

3.4 Overpotentials during lithium plating

Given the differences in (electro)chemical reactivity between Al 2 O 3 and ZnO, these coatings should change how lithium metal plates during battery charging. This can initially be quantified by looking at the nucleation and growth overpotentials in the first plating step (Figure 4A). Here, we quantify η Nucleation as the magnitude of the peak voltage of the first plating step; the absence of lithium prior to this step is crucial both for the definition of nucleation overpotential (as discussed in-depth by Yu et al.) (Yu et al., 2023) and to show how electrochemical reduction of the oxide coatings affects Li nucleation.

The nucleation overpotentials quantified in Figure 4B follow the impedance data in Figure 2—there is little difference between Cu and thin ZnO@Cu (η Nucleation ∼35 ± 4 mV for 0–10 nm thickness) but a relatively large decrease when increasing the ZnO thickness to 20 nm (15 ± 1 mV), while Al 2 O 3 @Cu increases the overpotential drastically with increasing thickness (up to 223 ± 60 mV at 10 nm thickness). This reflects the relative ionic conductivities of the interface layers following electrochemical reaction (Figure 2C), which will determine the cell voltage at the early Li plating stage through V cell = I applied × R cell (R cell will be mainly controlled by R Int as apparent by the larger interface vs small bulk semicircles in the Nyquist plots, Figures 2B,D). Surprisingly, the growth overpotentials show virtually no difference (η Growth ∼ 6 mV) between all coating conditions (Figure 4C), with only a slight drop in overpotential with increasing ZnO thickness (down to 5.1 mV) in line with the impedance in Figure 2E. This indicates that the initial plating dictated by either the reacted Al 2 O 3 or the converted ZnO surfaces is probably not a governing factor of the Li growth reaction at longer times, as another mechanism takes over under these conditions.

3.5 Visualizing coating reactivity and plated lithium

To visualize the differences in coating reactivity and lithium plating behavior between Al 2 O 3 @Cu and ZnO@Cu, we employed ToF-SIMS and SEM-FIB imaging. ToF-SIMS provides spatially resolved chemical analysis of the anode interface, while SEM-FIB directly characterizes the plated Li 3D structure. To prepare better samples for post-mortem analysis, the PEO-LiTFSI solid electrolyte in Cu||Li half cells was replaced with a liquid electrolyte (1 M LiFSI, 1 M LiNO 3 in DME) and cycled at room temperature so the anode could be rinsed well and expose the plated lithium with minimal disruption or corrosion. For each electrode, 1 mAh/cm 2 was plated at 10 µA/cm 2 after initial formation plating and stripping steps at 0.1 mAh/cm 2, 10 µA/cm 2. The cell was then disassembled, rinsed with pure DME solvent, dried, and stored in closed containers in the glovebox before analysis.

Three-dimensional ToF-SIMS images in Figures 5A,B highlight the ALD layer while contrasting the differences in lithium plating for Al 2 O 3 vs ZnO. With Al 2 O 3@ Cu, the ALD layer has Li integrated into the Al 2 O 3, detected as LiAlO + and other Li-containing ions (Figure 5A and Supplementary Figure S5 and Supplementary Figure S6). Within and mainly under this Li y AlO x layer, small amounts of lithium metal exist as evident from the Li 3 + signal (Counihan et al., 2023), especially in the depth profile (Figure 5B). Importantly, there is no Li on top of the Li y AlO x layer (Region A1), indicating no appreciable electronic conductivity in this layer relative to other regions on the electrode. Instead, in regions where lithium is seen plated in large amounts, it appears in a distinct Li/Li x AlO y /Cu triple interface (Region A2). This may result from the lithium metal plating onto the copper then breaking through the ALD coating or from lithium plating at pinholes in the coating, as discussed later. More than one region with lithium metal deposits exhibited this behavior, indicating this could be a prominent mechanism for lithium plating with Al 2 O 3 coatings (Supplementary Figure S7). Finally, the bottom of Region A1 shows that the DSC-C has sputtered through completely to metallic copper, as evident from the Cu 3 + signal, while Region A2 only shows metallic copper in the part of the bottom region where metallic lithium plated onto the copper and broken through the ALD coating; presumably the lithium metal is thick enough in that region that the copper substrate was not exposed during the short sputtering time.

With ZnO@Cu, no plating below the coating or triple interfaces were observed (Figure 5B, Region Z1). Rather, all Li was detected above a Zn-rich layer that separated the deposited lithium and copper metal substrate. This was detected as LiZnO + and Zn + separately. Note that sputtering with the DSC-S will form oxide species and lead to the appearance of LiZnO + during Bi 3 + sputtering, or it may form directly from an existing LiZn-Li 2 O mixture. However, XPS does not detect any “lithiated zinc oxide” structure present here. This also applies for the Li-Al 2 O 3 mixtures in Figure 5A, but since XPS detects a different chemical species (Figure 3A), we assume for now that LiAlO + results from a specific species, Li y AlO x. Importantly, the Zn + is concentrated exclusively between the Li 3 + and Cu 3 + interfaces, i.e., virtually no zinc dissolves into the lithium phase as it stays in the converted LiZn layer. This is important from a mechanistic standpoint, since other anode-free designs with alloying metal interlayers can involve metal dissolution into the plated Li and exsolution during Li stripping (Lee et al., 2020; Sandoval et al., 2023). Finally, the bottom of Region Z1 shows that the DSC-S has sputtered through completely to metallic copper, albeit with small amounts of alloyed Li(Huang et al.; Rupp et al., 2019) (which are also present with copper in Region 1 and part of Region 2).

To visualize the structure of lithium deposits on a larger level, we used SEM-FIB. Figure 6 shows the morphology of the plated lithium on the same Cu, 5 nm Al 2 O 3 @Cu, and 5 nm ZnO@Cu electrodes as Figure 5. Panel i is a low-magnification image of the sample surface, panel ii is a higher-magnification image of the Li morphology, and iii is an image of the cross-section after milling and polishing for each electrode. In general, the plated Li forms circular, flower-like patches on all three electrodes, appearing in the SEM images with a darker contrast than the Cu or coated Cu substrates due to its low Z value (Goldstein et al., 2018).

The image scales of Figures 6A–C,I are the same, and the differences between the sizes and morphology of the plated Li can be directly observed. The Li plated on Cu appears in clusters interconnected with each other, covering around 73% of the surface area measured (Figures 6A,I). Meanwhile, Li plated on 5 nm Al 2 O 3 @Cu surfaces forms individual, separated clusters (Figures 6B,I) and covers only around 36% of the electrode area on average. For Li on 5 nm ZnO@Cu, the average coverage is around 61%, and the clusters grow to be both separated and connected (Figures 6C,I).

The images in Figures 6A–C, 6ii are again all shown in the same magnification. The Li growth morphologies are similar to each other, forming circular flower-like clusters that are joined together by several “petals”. The size differences can be observed in these images as well, where the clusters grow largest on bare Cu (Figures 6A, 6ii) and smallest on 5 nm ZnO@Cu (Figures 6C, 6ii). Supplementary Figure S9 shows histograms of the Li clusters on the surface of 5 nm Al 2 O 3 @Cu and 5 nm ZnO@Cu, which quantitatively shows the distribution of the deposits. ZnO@Cu electrodes especially show very small Li cluster sizes (areas <10 −3 mm 2).

Figures 6A–C, 6iii show the cross-sectional view of a randomly selected, yet representative, Li cluster on each electrode. An averaged thickness marker is included in each image, where the measured Li growth thickness is around 5.36 µm on Cu, 9.03 µm on 5 nm Al 2 O 3 @Cu, and 5.20 µm on 5 nm ZnO@Cu. The dark contrast is the Li layer, whereas the layer underneath with grains is the Cu substrate in each case. EDS signal confirms the position of Cu, and the lack of signal in the dark layer confirms it is a low Z element, specifically Li, as shown in Supplementary Figure S10. The nearly double thickness of Li deposits on 5 nm Al 2 O 3 @Cu compared to Cu and 5 nm ZnO@Cu corroborate the reason for the lower coverage area observed in the lower magnification images. Since 1 mAh/cm 2 Li (∼5 µm theoretical thickness) is plated in all cases, Li must be accumulated on the sparsely formed Li clusters on 5 nm Al 2 O 3 @Cu, as opposed to growing freely on the underlying conductive surfaces of Cu and 5 nm ZnO@Cu.

3.6 Cycling Performance

In totality, all factors like interface resistance, overpotentials, and chemical differences that influence lithium plating will impact the performance of the battery. Two key metrics for battery performance are lifetime and charging rate, which are often tested at the lab scale by quantifying Coulombic efficiency and critical current density, respectively. Along these lines, we performed a series of plating and stripping measurements in Cu||Li half cells with PEO-LiTFSI electrolytes for all coatings. The testing protocol, known as the reservoir or Adams method (Adams et al., 2018; Mohammadi et al., 2023), follows that of Figure 7A: 50 µAh/cm 2 of lithium is plated and stripped, the same charge is re-plated, then 20% of this charge is stripped and plated for several cycles before a final stripping step. A small charge was used to avoid shorting the cells before useful information could be extracted, as will become evident. Each cell was run using this protocol at 10 µA/cm 2, then the protocol was repeated at increasing 10 µA/cm 2 increments until the cell shorted.

The current density during which the cell shorted will be referred to as the “shorting current density”, which we differentiate slightly from the common “critical current density” determined under different current or voltage ramp testing conditions (Sharafi et al., 2016; Flatscher et al., 2020; Gopal et al., 2023). The Coulombic efficiency (CE) of the initial plate-strip step and the reservoir step were quantified separately, following: where Q initial strip and Q initial plate are the respective charges from the first two steps, Q plate is the reservoir charge, n is the number of 20% strip-plate steps, and Q final strip is the charge from the last reservoir stripping step. We chose to quantify both CE values since CE Initial will reflect charge loss from lithium plated from the base ALD coating layer, while CE Reservoir will reflect charge loss from lithium plated on existing lithium.

Figures 7B,C compare CE Initial and CE Reservoir as a function of ALD coating thickness at 10 and 20 µA/cm 2 current densities. Although these are still well below the >99% values needed for practical batteries, this low Coulombic efficiency of lithium metal is known for solid polymer and composite electrolytes (Counihan et al., 2023) and is an active area of exploration. CEs of the 5 nm Al 2 O 3 @Cu show a 7%–13% decrease as compared to the CEs of Cu, and 10 nm Al 2 O 3 @Cu cells failed before plating even 50 µAh/cm 2 at 10 µA/cm 2. All ZnO@Cu electrodes performed more similarly to unmodified Cu, though still with 4%–11% lower CE values. At each current density tested, CE Reservoir was almost always equal to or greater than CE initial (Supplementary Figure S11). Although this is somewhat expected given the protocol differences (Mohammadi et al., 2023), this indicates that under these conditions, there is at least no particular charge efficiency benefit to plating onto the base ALD coating compared to a lithium reservoir and that keeping a lithium reservoir may even boost CE slightly (Fang et al., 2021). Somewhat surprisingly, in all cases, unmodified Cu electrodes show the highest CE values. Although there is a noticeable decrease in CE when the thickness of ZnO@Cu is increased from 10 to 20 nm, 20 nm ZnO@Cu still shows the same or higher CE than 5 nm Al 2 O 3 @Cu. This is opposite of the impedance and overpotential trends above and in contrast to the trend observed by Mohammadi et al. (Mohammadi et al., 2023), where increasing the ZnO thickness on Ni current collectors increased the CE in liquid electrolytes. Such a finding points to the differences between liquid and solid electrolytes and the many unique challenges in the solid electrolyte field. Charge loss due to Li trapping in the LiZn phase (Rehnlund et al., 2017; Crowley et al., 2022; Liu et al., 2023) is unlikely to contribute much to this, since the alloy layer is so thin and the cycling protocol includes potential holds to strip and de-alloy as much as possible.

Finally, the shorting current densities observed across all tested cells are shown in Figure 7D. With Al 2 O 3 coatings, the accessible current density drops off drastically with increasing thickness and is no better than the baseline Cu electrodes (30 µA/cm 2). In contrast, ZnO coatings offer a general increase in achievable current density, with 20 nm ZnO@Cu (60–100 µA/cm 2) doubling or tripling that of Cu alone and Al 2 O 3 @Cu. This may be related to a higher surface area and/or better nucleation sites on LiZn clusters compared to Cu, leading to less lithium filament and dendrite formation. The low Coulombic efficiencies with all electrode types point to the limitations of this simple PEO-LiTFSI electrolyte system, where other problems relating to lithium metal stability need to be addressed.

Given the above results, we postulate two distinct mechanisms for Li plating on Al 2 O 3 @Cu and ZnO@Cu (Figure 8) and discuss their effects on battery performance. We can describe these two mechanisms generally. In Mechanism 1, a metal or metalloid forms strong bonds to oxygen and is relatively unreactive to Li metal. The ALD coating remains as a metal oxide, though with varying amounts of lithium, after (electro)chemical reaction. In Mechanism 2, a metal or metalloid is readily reduced or reactive to lithium metal in its oxide form, converting completely to the metal or lithium-metal alloy and lithium oxide. Since the redox potentials of most metal ions are positive of Li 0, we believe the metal-oxygen bond strength will be a primary factor in determining the stability of the metal oxide, at least kinetically, and exhibiting Mechanism 1 vs 2 for a given material.

When reducing these metal oxides electrochemically (potentials greater than 0.0 V vs Li +/0) or chemically with Li 0, the reaction pathway will be determined primarily by this bond strength difference. In Mechanism 1, if the metal ion reduces and Li + enters the structure, the M-O bond will not break, leading to a stable lithiated structure:

It has also been hypothesized that Li + can enter the structure without any permanent redox reaction taking place due to space charge or other effects (Cheng et al., 2020):

For discussion purposes, we will refer to both of these pathways as “lithiation”, one with and one without redox processes occurring.

In Mechanism 2, a complete conversion reaction will be seen: M is reduced to its metallic state, and Li + ions replace its bonds with oxygen, leading to:

If the metal is able to alloy with lithium, it will form the alloy(s) appropriate for the metal-lithium mixture, possibly in a concerted mechanism during the conversion step:

These last two pathways are known in the field of conversion anodes (Bresser et al., 2016; Puthusseri et al., 2018; Marques et al., 2023). This “lithiation” vs “conversion”, along with electronic conductivity of the coatings, is at the heart of the differences between Al 2 O 3 @Cu and ZnO@Cu.

When MO x lithiates to Li y MO x, the ionic conductivity of Li + through the layer increases, as shown in Figure 2B. However, the maximum level of lithiation might only be achieved by chemical reduction with Li 0, as shown in Figure 2D & 3. This is represented in Figure 8 (top) with the electrochemically reduced Li y– MO x lithiating further to Li y+ MO x in the presence of Li 0. The exact Li content likely varies depending on the metal center, and it may be possible to undergo further lithiation and phase decomposition to form lithium oxides (i.e., driving Mechanism 1 partially or completely to Mechanism 2), although this needs more study to confirm.

Al 2 O 3 is a popular coating choice for lithium batteries, mainly on the cathode side due to its oxidative stability (Park et al., 2014; Cheng et al., 2020). Its reductive stability is less well-known and under debate. It is difficult to rationalize an Al 3+ cation reducing to Al 2+ or Al 1+, and Al 0 would not be stable within the oxide structure and is not observed experimentally in Figure 3. XPS clearly shows a shift to lower binding energies for both Al, O, and Li species (Figure 3A; 3C; 3D), indicating a reduced environment for all of those elements. Whatever reaction does occur between Li 0 and Al 2 O 3 appears to be self-limiting, as apparent here and in the work of Fan et al., who showed experimentally that reacting Li 0 and Al 2 O 3 particles yields a thin (∼2 nm) Li y AlO x layer on the particle surface (Fan et al., 2018). Given the 3 nm thickness but incomplete reaction of our ALD Al 2 O 3 @Cu in Figure 3, it is reasonable to assume this reaction also self-limits at ∼2 nm with ALD Al 2 O 3. In support of the direct reduction of Al 2 O 3 by Li 0 is work by Jung and Han, predicting that as Al 2 O 3 incorporates Li atoms (electrons and ions), Li replaces Al as the metal coordinating to O (Jung and Han, 2013). This causes increased Al-Al interactions resembling multi-atom clusters, a process thermodynamically favorable up to Li 3.4 Al 2 O 3 and nominal Al valencies of +1. Others have put more reserved Li:Al molar ratios of 0.5:2 (Fan et al., 2018) and 1.1:2 (Cheng et al., 2020) in this redox-driven lithiation. We also observe a clear change in the interface impedance when electrochemically reacting these coatings at reducing potentials before lithium plating (Figures 2B,C), although these changes could be coupled to decomposition of the electrolyte that enables reactivity by supplying other atoms to the reaction in Eq. 3.

Against the direct reduction of Al 2 O 3 is the obvious lack of evidence for Al 2+ and Al 1+ outside of exotic organometallic compounds (Dohmeier et al., 1996). Several Li-Al-O compounds are known, but all exhibit Al 3+ ions with different amounts of oxygen bond sharing with Li + ions (Byker et al., 1979). If Al 3+ was inherently redox active in Al 2 O 3, it is expected to appear in voltammetry (Figure 1B), although this could also be a kinetically limited process that never appears before lithium plates at low enough potentials. The self-limiting surface reaction mechanism also implies that reactivity can only occur under certain conditions. This could include oxygen vacancies or other non-stoichiometries, something certainly possible in ALD coatings (Robertson, 2005; Li et al., 2017), or (sub-)nanometer changes in Al 2 O 3 structure leading to undercoordination at the coating’s interfaces (Paranamana et al., 2022). Whether these non-stoichiometries can enable (electro)chemical reduction, if the chemical and electrochemical differences we observe in Al 2 O 3 @Cu coatings after e-beam Li deposition and electrochemical treatment involve direct reduction of the coating, and what the exact Li content in the coating is, remain open questions and an area for research. In spite of this, during cell operation, the coatings do somehow chemically change to incorporate Li + ions (apparent in ToF-SIMS data, Figure 5) to increase conductivity (Figures 2B,C), so Al 2 O 3 @Cu is still used as an example for Mechanism 1 here.

Once ionic conductivity is established in the Li y MO x layer, lithium metal can, in theory, be plated underneath the coating since Cu is the only electronically conductive component in the system (Figure 8, top). This is evidenced in Figures 5A,B (Region A1), as Li was detected only underneath the coating, pointing to a continued lack of electronic conductivity even in the lithiated state. However, most of the plated lithium seems to grow through or around the coating. This could be due to two primary reasons. First, the coating is not mechanically resilient enough to handle the stress of lithium plating underneath, so it delaminates from the surrounding Cu substrate or breaks, letting the Li cluster grow around or through the coating. These “emergent defects” arising during cell operation then lead to lithium plating above the plane of the coating into the bulk electrolyte. Second, inherent pinholes or thickness deviations in the layer from the ALD process provide “hot spots” for current to focus through the coating, since the resistance is lower in these spots than most of the coating. Both alternative mechanisms are shown in Supplementary Figure S12 and would result in images like Figure 5A, Region A2. In either case, the increased current density focusing through the Al 2 O 3 layer would lead to lithium filament or dendrite formation that shorts cells faster and makes higher surface-area deposits that cause lower Coulombic efficiency with Al 2 O 3 @Cu compared to unmodified Cu (Figure 7). This is corroborated by the lower areal coverage but thicker Li clusters seen in SEM-FIB (Figure 6) with 5 nm Al 2 O 3 @Cu electrodes. The breakthrough of lithium metal through the coating leading to higher electroactive surface area also explains the drastic decrease in interface impedance when Li is plated (Figure 2); differentiating this mechanism from increased Li + transport through a more lithiated Li y AlO x layer requires further investigation.

All experiments under the conditions tested here point to a competition between the ideal case of lithium plating directly below the Li y AlO x layer, lithium plating leading to emergent defects, and pinholes contributing to either direct plating of lithium through the film or enabling the emergent defect mechanism by providing weak points in the coating. If pinholes were the prevailing mechanism, lithium metal would never been seen plated under the much more resistive coating (Figure 5ba & 5b). Regardless, pinholes can be eliminated by careful processing conditions in ALD (Zhang et al., 2009). The prevalence of emergent defects at the expense of ideal plating indicates that Al 2 O 3 as a coating on Cu is a larger issue, so many chemistries should be explored for higher Li + conductivity and better mechanical properties to enable Mechanism 1 (Figure 8).

The lithium plating mechanism with ZnO@Cu is markedly different and more straightforward (Figure 8, bottom). ZnO more readily converts to LiZn and Li 2 O before lithium starts plating (Figure 1C; 3). Thus, lithium can nucleate preferentially on LiZn regions in contact with the electrolyte instead of just copper, as evidenced by the much lower nucleation overpotential for 20 nm ZnO@Cu (Figure 4B). However, it is most probable that LiZn clusters are isolated in space or form a loosely connected aggregate network instead of forming a uniform flat surface. These higher surface area electrodes would show lower geometric area interface resistance (as is indeed seen in Figure 2), larger lithium loss due to SEI formation over a larger area (seen in Figure 1D), and grow smaller Li clusters than the Cu case, which is confirmed in SEM images (Figure 6C). This would lead to lithium nuclei forming on top of LiZn separately at the nanometer scale. It is reasonable, though, that these nuclei grow and merge into one another (Figure 8, bottom), leading to lithium plating mostly resembling that of unmodified Cu (Figure 6). Apart from the initial nucleation, there is virtually no chemical or structural difference between Li plating on Cu or ZnO@Cu (except for a difference in the Li-covered area as mentioned earlier) since the surface of the Li will be controlled by SEI formed from the electrolyte, in contrast to the artificial SEI of Li y AlO x with Al 2 O 3 @Cu electrodes before lithium breaks through the coating. This homogeneous Li deposit on top of a thin LiZn layer is seen in ToF-SIMS (Figure 5B), and the plating morphology at larger scales is shown to be similar to Cu in Figure 6C. Even in the presence of defects, the lithium plating mechanism should not change much, since both the converted LiZn and underlying Cu are electronically conductive, in contrast to the Al 2 O 3 @Cu case (Figure 8 and Supplementary Figure S12). The higher shorting current density (Figure 7D) with increasing ZnO thickness using PEO-LiTFSI solid polymer electrolytes may reflect better Li nucleation, favorable morphology, and less dendrite formation at the current densities and lower charges not captured in our imaging experiments here.

This lithiophilic ZnO plating mechanism has been employed in some studies by using ZnO nanoparticles (Shin et al., 2023; Song et al., 2023). We believe ZnO ALD coatings would be advantageous over ZnO nanoparticles in two ways. First, particles distributed over the electrode surface would inherently produce more heterogeneity than the ALD coating given the same amount of ZnO. Exposed regions of the bare current collector could plate lithium directly and not take advantage of the lithiophilic LiZn alloy during the nucleation step. Second, for the same areal coverage of ZnO on a surface (ideally 100%), there should be much less lithium loss with the thin ALD coating since total coverage with nanoparticles would likely involve packing multiple layers of particles together, leading to more ZnO irreversibly converted to Li 2 O. One possible benefit of particles over ALD coatings would be less mechanical stress on the substrate electrode as the ZnO reacts to convert. As the coating takes in lithium, it should expand both normal to and laterally on the substrate; particles in this case have room to expand by pushing the flexible polymer electrolyte out laterally, leading to less stress on the substrate. This can be especially important over large areas typical of practical cell formats (Rodrigues et al., 2023). Electrochemo-mechanical stress during lithiation/conversion is something to explore in future for in situ reacted coatings.

Despite the clearly better electrochemical performance of ZnO@Cu compared to Al 2 O 3 @Cu with these PEO-LiTFSI solid polymer electrolytes, it is difficult to judge one type of metal oxide mechanism as better than the other in all cases. The chosen chemistry will probably depend on the type of electrolyte used for both mechanical and chemical compatibility reasons, making it important to study a range of materials on a fundamental level. Recent work in liquid electrolytes has shown that conversion-type (Mechanism 2) coatings like ALD ZnO on Ni(Mohammadi et al., 2023) and ALD SnO x on Cu(Kim et al., 2023a) have better Coulombic efficiency and charge retention than the bare metals or ALD Al 2 O 3. However, the electronic conductivity and conversion mechanism of these coatings still lead to irreversible charge in the first battery cycle, something to avoid for high energy density anode-free battery applications. Related to the lithiation-type (Mechanism 1) coatings, there exist other materials designed to guide lithium plating between the interlayer and current collector, rather than the interlayer and the solid electrolyte (Lee et al., 2020; Kim et al., 2023a). Although these coatings use carbon and sometimes alloying metal particles, the theory is the same as the top mechanism in Figure 8 and provides the better path forward for limiting electrolyte degradation by Li metal.

Beyond the two chemistries tested here, we believe other materials can be utilized to take advantage of and provide additional proof for the two mechanisms in Figure 8. Most metal oxides are expected to proceed through Mechanism 2. The oxides of common metals like zinc, iron, and tin are already well-studied conversion anode materials (Xu et al., 2023), with several metals alloying with lithium as well. Other non-transition metals or metalloids can form (lithium) oxides that are stable to lithium metal and utilize Mechanism 1. Like aluminum, the other group 13 elements of boron and gallium show Li-stable Li 3 BO 3 (Nagao et al., 2019) and LiGaO 2 (Tsai et al., 2022) compounds. Oxides with early transition metals like zirconium may also be able to form stable reduced structures; Zr 4+ in lithium garnets has been shown to form stable Zr 2+ species at the lithium metal-oxide interface (Zhu et al., 2019). SiO x, a promising anode material itself, is known to undergo a mixture of Mechanisms 1 and 2, where residual pockets of Si 0 alloy with Li while the remaining SiO x species reversibly reduce and form lithium silicates, Li y SiO x (Kim et al., 2023b). There is wide room to explore a range of single-metal and multiple-metal or layered oxide coatings to achieve efficient lithium metal plating.

In conclusion, we investigated how the fundamental chemistries of Al 2 O 3 and ZnO coatings influence lithium metal plating in anode-free SSB configurations with PEO-LiTFSI solid polymer electrolytes. ALD was used to create ultra-thin metal oxide layers on Cu current collectors with 5–20 nm thickness. Due to their electronic conductivity and reactivity differences, Al 2 O 3 @Cu electrode lose less lithium inventory in the first formation cycle relative to Cu while ZnO@Cu electrodes lose more with increasing coating thickness. Al 2 O 3 coatings show orders of magnitude higher interfacial impedance compared to Cu and ZnO@Cu until lithium is plated. XPS on Li-deposited surfaces reveals that Al 2 O 3 coatings react to form Li y AlO x, while ZnO coatings easily reduce and convert to a LiZn alloy and Li 2 O. These reactions help explain the lower nucleation overpotentials for ZnO@Cu electrodes and the higher overpotentials with thicker Al 2 O 3 @Cu coatings relative to unmodified Cu. ToF-SIMS imaging reveals that lithium plates underneath the lithiated Al 2 O 3 @Cu coating initially but mostly plates through emergent defects and pinholes; in contrast, lithium plates exclusively on top of the converted ZnO@Cu coating. The different plating mechanisms help explain the lower Coulombic efficiency and usable current densities with Al 2 O 3 coatings, as well as the different plated lithium coverages and thicknesses observed with SEM-FIB. Two general mechanisms resulting in the lithiation of less-reactive metal oxides and the conversion of more-reactive metal oxides during the initial lithium plating step are proposed based on these results. Future work will focus on understanding the redox processes in these coatings as well as the evolution and control of defects during lithium plating. This research opens new avenues to optimize coating thicknesses, chemistries, layering, and other parameters along with electrolyte chemistries to achieve high-Coulombic efficiency, anode-free lithium metal solid-state batteries.