Toward an Understanding of SEI Formation and Lithium Plating on Copper in Anode-Free Batteries

“Anode-free” batteries present a significant advantage due to their substantially higher energy density and ease of assembly in a dry air atmosphere. However, issues involving lithium dendrite growth and low cycling Coulombic efficiencies during operation remain to be solved. Solid electrolyte interphase (SEI) formation on Cu and its effect on Li plating are studied here to understand the interplay between the Cu current collector surface chemistry and plated Li morphology. A native interphase layer (N-SEI) on the Cu current collector was observed with solid-state nuclear magnetic resonance spectroscopy (ssNMR) and electrochemical impedance spectroscopy (EIS). Cyclic voltammetry (CV) and time-of-flight secondary ion mass spectrometry (ToF-SIMS) studies showed that the nature of the N-SEI is affected by the copper interface composition. An X-ray photoelectron spectroscopy (XPS) study identified a relationship between the applied voltage and SEI composition. In addition to the typical SEI components, the SEI contains copper oxides (CuxO) and their reduction reaction products. Parasitic electrochemical reactions were observed via in situ NMR measurements of Li plating efficiency. Scanning electron microscopy (SEM) studies revealed a correlation between the morphology of the plated Li and the SEI homogeneity, current density, and rest time in the electrolyte before plating. Via ToF-SIMS, we found that the preferential plating of Li on Cu is governed by the distribution of ionically conducting rather than electronic conducting compounds. The results together suggest strategies for mitigating dendrite formation by current collector pretreatment and controlled SEI formation during the first battery charge.

relatively low capacitance of the semi-circle (around 10 -9 F) also could be due to the significant roughness of the Cu substrate. 4 After 5 days' rest time, the fit of the impedance spectrum Nyquist plot for c-AcH-Cu required an additional RC component, assigned to the SEI. The diameter of the high-frequency semicircle decreases with time, while the total impedance has not changed significantly. The slope of the Nyquist plot of d-HCl-Cu decrease with time. This indicates a relative increase in the time constants of the diffusional as compared to the faradaic processes. 5 We attribute these trends to an increase in the heterogeneity of the SEI after prolong rest time. 6 Table S1. The resulting fitting parameters for the c-AcH-Cu symmetric cells shown in Figure  S1 and Figure S2.  Figure S3. 19  The effect of soaking time was studied by comparing the 19 F NMR spectra of samples of d-HCl-Cu soaked for 1 hour and 96 hours and of c-AcH-Cu soaked for 1 hour and 18 hours ( Figure S4). Both sets of spectra consist of resonances corresponding to LiPF6 and its decomposition products and a broad LiF resonance. 7 In addition, sharp resonances around -225 ppm were observed in all spectra except c-AcH-Cu soaked for 1 hour.
In all the 19 F NMR spectra (Figures 5, S3-5, S10-11), a group of poorly resolved resonances in the range of -130 ppm to -180 ppm appears in varying intensities (detailed in Figure S5). The -156 ppm and -153 ppm resonances are attributed to HF in several publications. 8 However, it is not fully understood how HF can co-exist with lithium metal. The resonance around -180 ppm could be attributed to fluoride ions coordinated to hydroxide 9 or oxide species ( Figure   S5). 10 In addition, several of the spinning side bands for the LiPF6 and LiF appear in this range, adding to the complexity of the spectrum. Figure S4. 19  hour which gives rise to broader 7 Li resonance compared to that of the N-SEI formed for 96 hours ( Figure S6).         Figure S17). 20 The broad lithium peak around 56 eV indicates the presence of LiPF6 along with its decomposition products (LixPFy, LixPOyFz) and LiF ( Figure S16d) which is also in corroboration with the presence of F1s peaks at 685.3 eV (typical for LiF) and 687.2 eV (typical for LiPF6 decomposition products) ( Figure S16c). 21 These findings suggest that the N-SEI on Cu is composed of Cu oxides and the LiPF6 decomposition products also seen in the 19 F NMR.
The XPS C 1s, O 1s, F1s and Li1s spectra of e-SEI2V are very similar to the spectra recorded for the N-SEI ( Figure S16). However, the C1s spectrum shows an increase in the intensity of C-O and C=O components in comparison to C-C/C-H components ( Figure S16). In addition, the increase in the intensity of the P 2p peak ( Figure S17     The composition of the SEI on lithium plated Cu substrates was studied with XPS both on the plated lithium and on the surface of the exposed Cu SEI areas ( Figures S20-21). Similarly to the e-SEI1.4V ( Figure S15), the Cu 2p signals were not observed on the surface, presumably since the SEI on the lithium plated sample is thicker and the Cu compounds are buried deeper than 20 nm ( Figure S20). After sputtering 20 nm of the exposed Cu area, the Cu 2p spectrum was observed with more significant negative shifts (929.7 eV and 949.0 eV), this could be due to both or either Cu oxide lithiation and Cu metal nano-cluster formation. [18][19][20]23 Sputtering of 20 nm of the lithium-plated area did not give rise to any Cu 2p peaks ( Figure S20).

In situ NMR
The skin depth of Li metal in this study is = √ 0 = 12.1 μm where ρ is the resistivity of the metal (94.7 n Ω for Li metal at 298 K), μ0 is the permeability of the vacuum (4π 10 -7 m kg/ s 2 A 2 ), μr is the relative permeability of the medium (μr = 1.4 for Li metal) and ν is the frequency of the applied rf field (116.7 MHz). 24,25 As shown in our previous work using in situ NMR on Cu-LFP full cells, 26 skin depth issues are not expected to be an issue. Similar to the previous study, only 1 mAh cm -2 of Li metal is plated in each cycle and by performing a "nutation experiment" of plated Li, we showed that the electrodeposits nutated like a sample experiencing no skin depth effects. Furthermore, the changes in Li metal intensity (in Figure 7) are approximately linear with charge, indicating that all plated Li contributes to the NMR signal. Figure S22. a) The 7 Li in situ NMR spectra at the start of cycling where only the diamagnetic peak originating from the electrolyte is observed. b) The 7 Li in situ NMR spectra at the end of charge (1 mAh/cm 2 ) showing both the Li metal peak centered around approximately 260 ppm and the electrolyte peak.
The coulombic efficiency (CE) of the galvanostatic cycling performed in in situ NMR cells and the corresponding normalized intensity of the Li metal peak were quantified with in situ NMR at the end of charge (plating) and discharge (stripping) for d-HCl-Cu ( Figure S23) and c-AcH-Cu ( Figure S24).These are additional examples of the trends described in the manuscript. The CE and the amount of dead Li quantified with NMR (the intensity of Li metal at the end of discharge) is similar for both surface treatments whereas the intensity of Li metal at the end of charge is different for the two cells.

TOF-SIMS
The measurements were conducted in the burst alignment mode (BAM) for better lateral resolution of images with a Bi + primary beam (25 keV) and a Cs + sputtering beam (500 eV) over an area of 250 µm × 250 µm (sputtering area 500 µm × 500 µm). During the measurements, the current for the sputtering beam was steady and constant (35 nA), thus over the same sputtering span, the fluence dose density were expected to be the same for all the samples. Moreover, the high current bunch mode (HCBM) with a mass resolution up to 10000 was also applied to ensure that within the limitation of the instrument, the species analysed in BAM mode were solely contributing. For these air sensitive samples, they were mounted inside an Ar-filled glovebox and transferred into the instrument within the vacuum transport suitcase which was opened until the pressure of the loadlock chamber lower than 10 -5 mbar.
The intensity of different species in SIMS are affected by a variety of different factors and the concentration of species normally cannot be obtained directly from the intensity data. However, for the same species, assuming the chemical environment was similar within the same sample and among samples with similar treatments, the relative intensities correlated with related concentrations are comparable.
Native and electrochemical SEI on Cu were studied by ToF-SIMS. CuxO, LiF, CuFx and OH -related species were observed in N-SEI and e-SEI at all tested conditions. Since the penetration depth of the SIMS is on the order of the thickness of N-SEI and constant-voltage e-SEI, the acquired depth profiles were analysed qualitatively.
In Figure Figure S27b.
The intensity of 65 CuOis significantly higher for d-HCl-Cu samples. This could be an indication of a thicker or more homogeneous SEI on c-AcH-Cu.
Copper fluorides (detected as 65 CuF2 -) were observed in N-SEI and SEI on copper by TOF-SIMS ( Figure S27d). The 65 CuF2depth profile of N-SEI and SEI reveals similar behaviour to copper oxides. The intensity of 65 CuF2is higher for N-SEI and eSEI2.8 compared to eSEIs formed at lower voltages. This suggests that CuFx are found mainly in inner SEI layers.
Moreover, CuF2 was observed in the XPS spectra. F1s gives rise to a broad peak for CuF2 around 685 eV, however this peak overlaps with LiF. Copper fluorides were not found by ssNMR, this is probably because CuF2 contains paramagnetic Cu(II) that is not visible to ssNMR. The presented ToFSIMS depth profiles are the results of signal collection from the analysis area (250 µm × 250 µm). However, further data analysis of the 65 CuO signal from different regions of interest, i.e., the exposed copper and plated lithium, confirmed the presence of copper oxides in the SEI on both areas. The calculated thickness of the plated lithium is approximately 1.5 µm assuming 100% efficiency and coverage of 50% of the copper area.
Thus, it is assumed that for all the SIMS analysis, the copper oxides remaining on the underlying copper substrates cannot contribute to the copper oxide signals recorded from the lithium surface.
The 3D reconstruction of 65 CuOsignal (XZ plane) of eSEI0.1V on lithium platted d-HCl-Cu (right) and c-AcH-Cu (left) in depicted in Figure S28. For each sample the 65 CuOsignals for SEI on lithium and the exposed Cu are compared. The intensity of 65 CuOin the SEI on the exposed d-HCl-Cu is higher compared to the exposed c-AcH-Cu and that the increase of 65 CuOsignal with sputtering is more gradual on d-HCl-Cu. The intensity of the 65 CuOsignal from the SEI on the plated lithium is similar to the signal from the top layers of he SEI on the exposed copper. Figure S28. 3D reconstruction of 65 CuO signal (XZ plane), comparing the intensity of 65 CuO on exposed and lithium plated areas on a copper disk after the application of lithium plating procedure (d-HCl-Cu-right, c-AcH-Cu-left).

Lithium plating on Cu
The XY maps of OHand LiF (detected as LiF2 -) in the eSEI0.1v on c-AcH-Cu ( Figure S29), LiF and OH -XY and XZ maps of eSEI2V on c-AcH-Cu ( Figure S30) and d-HCl-Cu ( Figure  S31) reveal a similar trend to the SEI composition trend on d-HCl-Cu ( Figure 10). In Figures  S29-S33 the areas with the higher intensity are SEI on Cu areas.
The OHintensity in N-SEI on d-HCl-Cu is higher. The intensity of the OHsignal for the N-SEI on d-HCl-Cu is significantly higher and more heterogeneous compared to that of N-SEI on c-AcH-Cu. These findings strengthen the reliability of the OHmapping and support the assumption of preferential Li plating on OHpoor areas. Figure S34. OHdepth profile in SEI formed at N-SEI, normalised to total counts