Effects of Atmospheric Gases on Li Metal Cyclability and Solid-Electrolyte Interphase Formation.

For Li-air batteries, dissolved gas can cross over from the air electrode to the Li metal anode and affect the solid-electrolyte interphase (SEI) formation, a phenomenon that has not been fully characterized. In this work, the impact of atmospheric gases on the SEI properties is studied using electrochemical methods and ex situ characterization techniques, including X-ray photoelectron spectroscopy, X-ray diffraction, Fourier transform infrared spectroscopy, and scanning electron microscopy. The presence of O2 significantly improved the lithium cyclability; less lithium is consumed to form the SEI or is lost because of electrical disconnects. However, the SEI resistivity and plating overpotentials increased. Lithium cycled in an "air-like" mixed O2/N2 environment also demonstrated improved cycling efficiency, suggesting that dissolved O2 participates in electrolyte reduction, forming a homogeneous SEI, even at low concentrations. The impact of gas environments on Li metal plating and SEI formation represents an additional parameter in designing future Li-metal batteries.

As demand for high-performance energy storage devices continues to grow, “beyond lithium-ion” battery technologies, such as conversion-type chemistries and lithium metal anodes, are necessary to meet these needs.[1] Li–air batteries, for instance, promise extremely high energy capacities, with estimated theoretical values up to 3500 Wh/kg and practical capacities in the range of 500–1000 Wh/kg.[2−6] Since the first report of a Li–air battery, various electrodes, catalysts, electrolytes, and additives such as redox mediators and H2O have been employed to lower charge overpotentials, suppress side reactions, and improve the cyclability.[7−14] Although significant progress has been made at the positive electrode, the generally irreversible reactions that occur at the Li metal anode have not been fully characterized.[2,15]

The reduction potential of lithium is very low, such that it will spontaneously decompose electrolyte, resulting in the growth of a solid-electrolyte interphase (SEI).[16] An ideal SEI is electronically insulating to prevent continuous electrolyte decomposition, yet ionically conductive to lithium ions.[16,17] In a Li–air battery, SEI formation is affected by dissolved gas diffusing from the positive electrode to the Li anode, analogous to undesirable polysulfide shuttling in Li–S batteries and gas crossover in fuel cells (schematic );[18−23] the saturated O2 concentration and O2 diffusion constant for 1 M LiTFSI in tetraglyme were estimated to be around ∼0.6 mM and 10–7cm2/s in 1 atm dry air at 25 °C, respectively.[24−26] Furthermore, while the solubilities of other atmospheric gases are less studied, they also affect SEI formation because of the reactivity of Li metal.[27−30] Several studies have shown that atmospheric gases, including N2, can affect lithium cycling.[31−33]

Scheme 1

Schematic of the Li Metal Electrode during Charge in a Li–Air Battery Where Gas Crossover from the Positive Electrode Can Participate in SEI Formation

The stability of the lithium SEI is crucial to battery performance and lithium cyclability.[16,34−36] Discrepancies in previous literature on the positive or negative impact of gas crossover on SEI formation and lithium stability have yet to be explained.[19,20,31,32,37−39] Many reports state the necessity of using a protected Li metal anode or SEI-forming additives to enable cycling in the presence of O2.[9,20,40−44] Furthermore, the effect of O2 partial pressures has not been explored, yet this is relevant to Li metal batteries prepared in dry room environments. These gaps in understanding necessitate a systematic investigation of SEI and lithium plating in various atmospheric gas environments.

In this study, lithium metal was plated/stripped onto a copper substrate while exposed to O2, Ar, and N2 environments. Ex situ techniques were used to characterize the SEI, including scanning electron microscopy (SEM), Fourier-transform infrared spectroscopy (FT-IR), X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS). The presence of O2 gas during Li plating, while increasing the interfacial resistance, resulted in uniform nucleation and dendrite suppression and enabled good Coulombic efficiencies. A homogeneous SEI consisting of large LiOH domains is likely formed as opposed to a heterogeneous layer formed of many microphases. In mixed O2/N2 environments, Li plating morphologies and cyclability were also improved, but without a notable increase in resistance. These findings indicate that gas crossover in Li–air batteries and other Li metal batteries plays a significant role in SEI homogeneity and overall performance.

Electrochemical Cycling of Li Cu Cells. Lithium metal was plated/stripped in various atmospheric gases with the Swagelok design shown in Figure S1. An initial 2.5 mAh/cm2 of lithium was plated onto the Cu substrate, then subjected to 20 stripping–plating cycles at 10% of the initial capacity, all under galvanostatic conditions (Figure a). At the end, all available lithium was stripped from the Cu. Coulombic efficiencies (abbreviated as CE) were calculated using the Aurbach method, which provides an average efficiency over the whole experiment (Figure S2).[38] Overpotential was averaged from the voltage plateaus during plating. The electrolyte was 1 M LiTFSI in tetraglyme.

Figure 1

Electrochemical cycling of Li|Cu cells. (a) Potential versus time plot for galvanostatic cycling tests performed in O2, Ar, and N2. (b) Coulombic efficiency and (c) plating overpotentials. Four cells were repeated in each gas environment, and error bars show standard deviations. (d) Galvanostatic plating–stripping for one cycle. Current densities of 0.25mA/cm2 were used.

The Li plating and stripping tests performed in O2, Ar, and N2 atmospheres showed stark differences in overpotential and capacity loss. Samples in Ar, while initially cycling with a low overpotential, showed a large jump in polarization after the first 9 cycles (Figure a), indicating that no more lithium can be removed from the Cu electrode.[34,35] Almost 40% irreversible loss in capacity occurred in a single plating/stripping cycle (Figure d). This poor performance is in agreement with previous studies using ether electrolytes.[45]

By contrast, O2 crossover greatly improved the CE (Figure b); however, there was a large increase in overpotential (Figure c). Interestingly, O2 improved the performance in carbonate-based electrolytes without increasing the overpotential (Figure S3). This suggests that improved plating/stripping efficiencies with O2 are not solely due to high overpotentials, the high overpotentials increasing Li nucleation density according to classical nucleation and growth theory.[17] An N2 gas environment slightly improved the CE as well, without a drastic increase in plating overpotential (Figure b).

Impedance spectroscopy was performed following 20 plating–stripping cycles, and the data were fit with an equivalent circuit model (Figure S4). Larger impedances were observed for the SEI formed in O2, consistent with the trend in overpotentials. The shape of the impedance was also different in the O2 sample, indicating differences in SEI structure and ion transport mechanism.

Ex Situ Morphology and Composition Studies After Cycling. SEM images were taken to investigate the effect of gas crossover on the cycled morphology (Figure a). After the cycling and stripping of all the available lithium, electrically isolated Li dendrites were visible in the Ar sample, which explains the large capacity losses (Figure b). This dead Li was observed in the XRD spectra for Ar and N2 samples (Figure S5). In an O2 environment, the cycled morphology was more uniform and less porous (Figures a and S6), which correlates with the weaker Li signal seen by XRD (Figure S5) and the higher CE observed (Figure b).

Figure 2

SEI morphology and composition studies. (a) SEM images and (b) XPS spectra for the SEI formed in various gas environments after cycling.

Ex situ XPS was used to characterize the SEI components: for all samples, salt decomposition products, including LiF, Li2O, and Li2S, were identified (Figure b).[36,46] Peak positions are well aligned with previous studies.[47,48] These peaks increased with depth profiling while the C–C peaks decreased (Figure S7 and Table S4), consistent with widely accepted SEI models: more inorganic components closer to the SEI/Li metal interface.[16] From the F 1s and S 2p spectra, relative peak intensities of LiF (purple) and Li2S (cyan) were much larger for the SEIs formed in Ar and N2 than in O2 (Figure b and Tables S5–S9), and total atomic concentrations showed more F, N, and S content as well (Figures S7 and S8). Online electrochemical mass spectrometry results also indicate increased fluoride species in Ar and N2 samples compared to O2 (Figure S9).

There were fewer salt reduction products in the SEI formed with O2, and LiOH was observed from the XRD and FTIR spectra (Figures S3 and S10). These peaks were absent in the Ar and N2 samples. In addition, Li2CO3 (Figure b, yellow) was detected with larger relative intensities in the O2 sample; a C 1s peak at 283 eV, attributed to Li–C species, was also detected in the O2 sample.[48,49] Although LiF-rich SEIs reportedly improve Li metal anode stability, the poor Coulombic efficiencies in Ar samples, despite high LiF content, suggest that other species—likely LiOH—and mechanisms such as SEI uniformity may play a large role in controlling lithium plating and stripping.[50−53]

Studies of the Initial Li Deposition. The initial Li plating of 2.5 mAh/cm2, prior to cycling, resulted in round, particulate Li morphologies for O2 samples (Figure a). In an Ar environment, whisker-like dendrites were formed (Figure b). Both uniform nuclei and thick dendrites were observed for N2 samples (Figure c). We hypothesize that O2 affects the Cu surface, which subsequently affects lithium nucleation.[17,54] In galvanostatic plating, the low reduction potentials of lithium dictate that electrolyte is first reduced; this continues until an electrically insulating layer is formed across the Cu surface, preventing further electrolyte reduction.[16,36] In the presence of O2, less charge was irreversibly consumed to form this initial surface layer (Figures d and S11); a similar effect is seen with SEI-forming additives such as fluorethylene carbonates.[51,54] The peak beginning around 2.5 V in the voltammetry sweep corresponds to O2 reduction on the Cu electrode prior to Li nucleation (Figure e).

Figure 3

Effect of O2 on Li nucleation. (a–c) SEM images of the first Li plating. Scale bar = 10 μm. (d) Voltage versus capacity during galvanostatic plating. (e) Linear sweep voltammetry to 0.3 V vs Li/Li+ and (f) subsequent FTIR measurements of the Cu electrode.

In order to identify surface species formed prior to Li nucleation, FTIR was measured on Cu electrodes biased to 0.3 V vs Li/Li+ under O2, Ar, and N2 (Figure f). LiOH was observed only in O2 samples. No significant amount of LiOH was observed when biasing to 0.3 V in an N2 environment and subsequently exposing to O2 at open-circuit potential (Figure e, green). This indicates that LiOH was formed electrochemically. These LiOH morphologies are visible in SEM images of the Cu surface when plating in O2 environments (Figure S12a), whereas under Ar, the Cu surface remains rougher (Figure S12b).

To test the hypothesis that the LiOH-rich SEI formed in O2 promotes more uniform Li nucleation, we explored a pretreatment method whereby the cell was first cycled in O2 for 5 cycles before purging and continuing to cycle in Ar. Cells cycled in Ar have low initial CE of 60% which drops rapidly to below 5% after 5 cycles (Figure S13), whereas similar cells cycled in O2 maintained a CE of >90% for more than 60 cycles (where cycling was stopped). Cycling in O2 and then swapping to an Ar atmosphere extended the cycle life to 30 cycles while maintaining high efficiencies of >90% (Figure S13). However, the cell still fails, indicating that O2 continues to play a role in SEI formation during cycling.

Effect of Mixed Gas on SEI Formation. Lithium was next cycled in mixed O2/N2 environments to determine the effect of O2 at different O2 partial pressures. Impressively, CE values in O2/N2 environments were comparable to those of samples cycled in O2 (Figure a), dropping to 80% for a 25:75 O2/N2 ratio. The initial Li nuclei in a mixed O2/N2 environment were also round and nondendritic (Figures c and S14a for the 25:75 and 50:50 ratio, respectively). The plating overpotentials drop steadily with decreasing O2/N2 ratio (Figure b), and the overall SEI composition for the 25:75 sample is closer to that cycled in N2 (Figure d). LiOH is still detected by FTIR and SEM (Figure S14).

Figure 4

Effect of mixed gas on SEI formation. (a) Coulombic efficiencies and (b) plating overpotentials as a function of O2/N2 ratio. (c) SEM image after Li plating in O2/N2 (25:75). Scale bar = 10 μm. (d) XPS atomic percentage for the SEI formed in O2, N2, and 25:75 O2/N2.

Discussion. The XPS, XRD, and FTIR results all suggest that the SEI formed in Ar and N2 contains multiple inorganic products from LiTFSI reduction. This heterogeneous SEI, following the mosaic model,[16] leads to nonuniform ion diffusion through the SEI, uneven Li plating, and dendritic growth,[45,50] consistent with the observed dendritic morphologies (Figure b). In constrast, the O2 SEI contains large LiOH crystallites that form before the onset of plating; this SEI results in much smoother and rounder Li deposits, also coated with LiOH. The positive effects of O2 persist even when either swapping to an Ar atmosphere after the initial cycling (for up to 25 cycles) or when using a lower O2 partial pressure.

From the impedance measurements and equivalent circuit modeling,[55] bulk SEI resistances were similar across all samples; however, the SEI formed in O2 had a larger resistance modeled via a second resistor/capacitor component, likely relating to grain boundaries and the formation of the LiOH layer. Lower grain boundary resistivities were observed in Ar and N2 samples by more than 60 Ω, indicating higher total ionic conductivity and is reflected in the lower plating overpotentials (Figure c). The higher grain boundary resistivity observed in the O2 sample is consistent with a homogeneous and full covering layer of LiOH (SEM, Figure S12) that ions must travel through as discussed in more detail in the Supporting Information (Figure S4 and Table S1).

Two mechanisms have been proposed for SEI formation in the presence of O2 (Figure S15). The first involves the reduction of O2 to form superoxides and then Li2O2, which can then attack the ether solvent, forming Li-carbonates, Li alkoxy species, and LiOH.[38,56] Li2O2 may directly deposit on the Cu surface as well; however, this was not detected with the characterization methods used. Although glymes have improved stability against reduced oxygen species compared to carbonate solvents, the highly reductive potentials during lithium plating may also affect the reactivity of glymes, resulting in radical formation, the radicals then react with O2.[21,57−59] The increased Li2CO3 content for SEI formed in O2 is consistent with increased glyme decomposition (Figure b). Oxygen and nitrogen can also chemically react with lithium metal; this appears to be a minor effect compared to electrochemical SEI formation as shown in EDX measurments of Li metal exposed to gas environments under OCV conditions (Table S2).[27,28,39,60]

Water contamination is another potential source for LiOH,[19,20] and efforts were made in order to minimize this in our study (see Experimental Methods in the Supporting Information). Interestingly, adding 1% H2O to the electrolyte while plating/stripping lithium in an Ar environment showed no improvement to the CE nor suppressed dendritic growth (Figure S17), suggesting that LiOH formed from H2O is not responsible for the large improvements to Coulombic efficiency seen when cycling in an O2 environment. This result has significant implications for understanding lithium–air batteries in the presence of water: multiple studies have shown that these cells cycle even when the Li metal electrode is not protected.[11,38,61] The results presented here indicate that O2 and not simply water is key in protecting the Li metal (Figure S18).

In conclusion, our work demonstrated improved Coulombic efficiencies when cycling lithium in the presence of O2. Even in a mixed O2/N2 gas environment—with an O2/N2 ratio close to that in air (and in a dry room)—O2 helped form a more homogeneous LiOH-containing SEI layer on the Cu substrate surface, enabling uniform Li nucleation and improving subsequent plating–stripping efficiencies. While LiOH is the major component of the SEI, it is the SEI formed in the presence of O2 rather than water that is key to the improved performance. In the absence of O2 or other additives, a more heterogeneous SEI is formed containing a more diverse range of species, which results in nonuniform plating and dendrite growth. Gas crossover could therefore potentially be used as a pretreatment method for promoting homogeneous SEI layers in Li metal anodes. The effects of gas crossover on SEI formation and overall cell performance posit design questions for electrode fabrication in future Li metal batteries, including practical Li–air batteries.