Uncovering the Dynamics of Li–Au Alloying and Li Nucleation at the Au/LiPON Interface with In Situ Z Contrast and Surface Roughness Contrast Imaging via Scanning Electron Microscopy Authors: Munekazu Motoyama, Koji Shimizu, Toshio Kimura, Takayuki Yamamoto, Satoshi Watanabe, Yasutoshi Iriyama Published: 2024 ABSTRACT Controlling the nucleation, growth, and dissolution of Li is crucial for the high cycling stability in rechargeable Li metal batteries. The overpotential for Li nucleation (ηnc) on Li alloys such as Li–Au is generally lower than that on metal current collectors (CCs) with very limited Li solubility like Cu. However, the alloying process of CC and its impact on the Li nucleation kinetics remain unclear. Herein, we experimentally investigate the Li nucleation mechanism on Li–Au alloys using scanning electron microscopy with multiple detectors to distinguish the atomic number and topography contrast. Nonequilibrium processes near the Au CC film/lithium phosphorus oxynitride (LiPON) glass electrolyte interface were observed. Local dissolution of Li occurs concurrently with the Li plating process. This phenomenon indicates that Li nucleates as a metastable phase in Li–Au alloy CC films during Li plating due to the substantially lower interfacial energy of Li with the alloy phase. This low interfacial energy also leads to a low ηnc. Moreover, the local Li redissolution process results in a nonrandom spatial distribution of Li growth sites. Consequently, the alloy layer acts as a buffer, facilitating a more uniform distribution of the reaction sites. Introduction Secondary batteries based on Li metal anodes have high theoretical energy densities.1 However, practical applications of these high theoretical energy densities hinge on maintaining a Coulombic efficiency of nearly 100% while preventing cell short circuits.2 To date, mitigating cell short circuits remains a substantial challenge for both liquid and solid electrolyte battery systems. Growth of Li dendrites2 or filaments,3,4 which extend toward the cathode, causes electrolyte shorts. To address this issue, it is crucial to suppress nonuniform Li plating, which initiates Li dendrite or filament growth at the electrolyte interface. This highlights the necessity of controlling the initial stages of Li plating, particularly the processes of Li nucleation and growth.5,6 The Li nucleation overpotential (ηnc) is typically reduced when the current collector (CC) metal forms Li alloys.5,7−9 Yan et al. measured the ηnc values in an alkyl carbonate electrolyte for various CC metals (Pt, Al, Mg, Zn, Ag, Au, Ni, Cu, etc.).9 They observed that ηnc significantly decreases when CC metals form Li alloys throughout the composition compared to using CC metals insoluble in Li. According to their model, the dissolution of the CC metal in Li reduces the Li nucleation barrier. Okita et al. reported that the presence of a Pt interlayer (<10 nm in thickness) between a Cu CC film and a lithium phosphorus oxynitride (LiPON) glass electrolyte significantly decreases ηnc.10 Moreover, inserting a Pt interlayer improves the Coulombic efficiency of Li plating/stripping. Pt forms various Li alloy phases such as LiPt2 and LiPt.11 Additionally, the Li–Pt alloy/LiPON interface leads to a higher Li nucleation number density compared with the Cu/LiPON interface. Hence, Li alloying affects the Li nucleation process. However, how these alloying processes dynamically interact with the Li nucleation process is not fully understood. The presence of a Li alloy interlayer not only reduces ηnc but also enhances the cycling stability of Li plating/stripping.12−15 Thus, gaining insight into the kinetics of Li nucleation through Li alloying is indispensable for the rational design of rechargeable Li metal batteries. Inorganic solid-state electrolytes (SSEs) have a nonvolatile nature. This nonvolatile property provides a distinct advantage for in situ electron microscopy, which requires vacuum conditions. Previous studies have employed in situ scanning electron microscopy (SEM) to observe the initiation of Li growth and alloying during Li plating at the Cu, Pt, and Al/LiPON interfaces.6,16−19 SEM can visualize nanometer-scale spatial differences in atomic number (Z) on the sample surface using backscattered electrons (BSEs). Z of Li is 3, which is significantly smaller than that of heavy metals typically found in CC films. Due to this difference in Z, BSE imaging is a valuable tool for in situ observations of the nonequilibrium alloying process between Li and CC. The substantial difference in Z between Li and Au (Z = 79) provides a clear contrast, making Au a suitable CC material for studying Li nucleation using a BSE imaging technique. Moreover, the low ηnc of Li on Au makes it an optimal choice for examining the impact of Li alloying on ηnc reduction.7,9,14 Therefore, this study aims to explore Li nucleation by employing a Au CC film and utilizing in situ SEM techniques to differentiate spatial variations in Z and surface roughness. Our findings reveal that local dissolution of Li occurs concurrently with the Li plating process. This phenomenon facilitates a redistribution of Li growth sites, which consequently affects their survival. The redistribution of these growth sites provides insight into how to achieve more uniform Li plating in the presence of interfacial alloy layers as opposed to interfaces without such layers. The rest of the article is organized as follows. Section 3.1 analyzes the in situ observed BSE and lower secondary electron (SE) images to provide insights into the initiation and progression of Li–Au alloying within Au CC films. Section 3.2 examines the changes in BSE image contrast during Li–Au dealloying to confirm the interpretation of BSE contrast changes discussed in Section 3.1. Finally, Section 3.3 explores Li nucleation from a Li–Au alloy CC film, uncovering the redissolution process of metastable Li nuclei during the nucleation process. Methods The solid-state electrochemical cell used in this study is detailed elsewhere16,17 (Figure 1). A 150 μm-thick mirror-polished Li–Al–Si–P–Ti-Ge-based oxide glass ceramic (LATP) plate (Ohara) served as the Li+-conducting substrate. Radio frequency magnetron sputtering coated both surfaces of a 13 × 13 mm LATP plate with approximately 2.5 μm-thick LiPON films. On one side of the LiPON films, a 30 nm-thick Au CC film was deposited within a 5.0 mm-diameter area by pulsed laser deposition (PLD). The duration of PLD required to form a 30 nm-thick Au film was determined using a quartz microbalance (XTC/3, INFICON). The Au film thickness was validated by using a profilometer with a 0.1 nm nominal height resolution (ET200A, Kosaka Laboratory). A YAG laser with a wavelength of 266 nm and a power of 400 mW was used to deposit the CC film. An approximately 3 μm-thick Li film with a 9.0 mm diameter was deposited on the other side of the LiPON film as the counter electrode using a vacuum evaporation method. Schematic illustration of an all-solid-state-thin-film cell with the Au/LiPON/LATP/LiPON/Li structure for in situ SEM observations. The above-mentioned solid-state electrochemical cell was mounted on a custom-made SEM stage with the Au CC film facing up. A Cu plate with a 5.0 mm-diameter hole (viewport) was secured on the top surface of the cell, which corresponded to the Au CC side. The Cu viewport plate was electrically isolated from the SEM stage. Both Au working and Li counter electrodes were individually connected through a feed-through to cables from a potentiostat (SP-300, BioLogic) outside the SEM (SU8030, Hitachi High-Technologies). The edge of the viewport made contact with part of the Au CC film. This contact point is where electrons were supplied to and from the Au CC film. After the cell was secured onto the SEM stage, it was transferred to the SEM chamber without exposure to atmospheric air. The acceleration voltage of the electron beam was set at 3 kV (3–5 μA emission current). The estimated penetration depth of the primary electrons accelerated at 3 kV was approximately 26 nm for the Au film using the Kanaya–Okayama equation (Supporting Information).20 Thus, the observed Z contrast should reflect information across nearly the entire thickness range of the Au film. Normal SEM observations were performed in SE mode. Z contrasts and surface roughness contrasts were independently visualized in the high angle backscattered electron (HA-BSE) mode and the lower-SE mode, respectively. The lower-SE mode detector was positioned below than that of the SE mode detector (Figure S1). A constant current density of 100 μA cm–2 was applied for Li–Au alloying/dealloying and Li plating/stripping. Li plating was terminated at 1800 s after applying the current. Alternating current impedance spectroscopy was employed to measure the cell resistances in the frequency range from 200 kHz down to 100 mHz. Molecular Dynamics Simulation Molecular dynamics (MD) calculations were performed to gain atomic-level insight into Li–Au alloying. The neural network potential (NNP) of Li–Au alloy systems was used as the force field due to its proven accuracy in predicting various Li–Au alloy phases.21 This accuracy was validated in earlier work,22 where the predictions were comparable to those obtained through first-principles MD simulations. We used the LAMMPS software with our homemade NNP interface for all MD calculations.23 Typically, very long simulation times are needed to achieve alloying for relatively large model sizes at low temperatures (e.g., room temperature). Such simulations are particularly challenging for first-principles MD calculations. To address this challenge, we used NNP-based MD simulations and set the temperature to 500 K. This approach accelerates Li–Au alloying in relatively large model sizes while maintaining a temperature close to room temperature. As the initial structures of MD simulations, superlattices combining Au(111) and Li(100) were used. The models consisted of 18 atomic layers of Au (5184 atoms) and 2 (400 atoms), 6 (1200 atoms), 10 (2000 atoms), 14 (2800 atoms), or 18 (3600 atoms) atomic layers of Li. To view the alloying processes, these models were longer in one direction along the interface (x-axis) and perpendicular to the interface (z-axis). MD simulations were performed for 600 ps (time step of 1 fs) at a constant pressure (NPT ensemble) of 0 GPa for the respective models. The Li/Au composition ratio was estimated by counting the neighboring atoms of the constituent Au atoms within a 7 Å sphere (the ratio remained unchanged for longer cutoff distances). Results and Discussion Li–Au Alloying In organic liquid electrolytes, it has been reported that underpotential deposition (UPD) of Li on Au occurs within a potential range of +0.3 to +0.9 V vs Li+/Li.24−26 When the potential of the Au electrode is further reduced to the range of 0 to +0.3 V, Li–Au alloying is observed. This indicates that electrochemical reduction of Li+ at the Au electrode surface results in dissolution of Li atoms into a solid Au electrode before Li plating begins, even at room temperature.9,27 Generally, the activity of an element in an alloy phase is less than 1, which, according to equilibrium thermodynamics, leads to the expectation that the equilibrium potential for the reaction Li+ + e– = Li in LiξAu (ξ > 0) is more positive than the Li+/Li equilibrium potential. Figure 2A shows the voltage transient over time (t) during Li–Au alloying and subsequent Li metal nucleation with a 30 nm-thick Au CC film at 100 μA cm–2. In this experiment, the open circuit voltage is more positive than +1.0 V. Shortly after the measurement was begun, the voltage decreases below +0.10 V but then quickly increases to +0.15 V, indicating a sharp negative voltage peak (c in Figure 2A). The voltage plateaus at +0.15 V for approximately 200 s. After t = 300 s, the voltage abruptly decreases to approximately +0.05 V and gradually decreases until t = 500 s with a small voltage step occurring at t = 400 s. Finally, at t = 520 s, the voltage drops below 0 V. (A) Voltage transient (red curve) and theoretical ξ variation in LiξAu during Li–Au alloying and subsequent Li plating at 100 μA cm–2. Shaded area indicates the positive voltage region. Areal fraction variation of the dark spots (blue plots) in images (b–j) in (B). (B) In situ SEM images observed in SE mode during the voltage transient shown in (A), where images (a–p) are recorded at the times indicated by the corresponding alphabet in (A). Previous studies reported similar voltage profiles with two plateaus during electrochemical Li–Au alloying in liquid electrolytes, as shown in Figure 2A.28,29 Bach et al. provided evidence that the high-energy X-ray diffraction patterns of Au begin to deviate from a pure Au pattern below +0.3 V.28 In the range of +0.3 to +0.9 V, UPD occurs in the Li–Au system in organic liquid electrolytes, as previously described.24−26 However, the atomic structures at the Au/SSE interface remain immobile unless a process like the reduction decomposition of SSE is involved.30 Thus, UPD is not expected to occur at solid/solid interfaces in the same manner as that at solid/liquid interfaces. However, assuming that UPD does occur at solid/solid interfaces, the typical charge passed ranges from 100 to 300 μC cm–2, as UPD involves the formation of a metal monolayer on the electrode surface.24−26 Consequently, this process should be completed within three s at 100 μA cm–2, before reaching point b in Figure 2A. Hence, the voltage plateaus observed until t = 520 s are supposed to be due to Li–Au alloying. Theoretically, the charge passed at t = 520 s results in the formation of Li1.8Au, which falls within the δ-phase range as indicated by the Li–Au binary phase diagram.21 According to the diagram, the δ-phase spans from Li62.5Au37.5 to Li65.5Au34.5. Recent theoretical studies have also suggested that a stable or metastable Li2Au phase22,31 with a composition close to Li1.8Au forms. Bach et al. reported that Li–Au alloying continues until the formation of Li3Au, but all other alloy phases they observed were metastable.28 They proposed that a metastable Li5Au3 phase forms during the first plateau of Li–Au alloying, while Li3Au begins to form in the second plateau. On the other hand, Krauskopf et al. reported that Li nucleation occurs when a Au CC film is lithiated up to Li2Au on garnet-type SSE (Li6.25Al0.25La3Zr2O12).14 Thus, phase formation during electrochemical Li–Au alloying is rather complex. Therefore, operando high-energy X-ray diffraction should be employed in future studies to determine which Li–Au alloy phases form on LiPON during the reaction. Figure 2B shows the in situ SEM images observed in SE mode during the voltage transient in Figure 2A. Visible Li is not observed at a voltage above 0 V (a–k). Thermodynamic theory can explain the absence of observable Li because Li–Au alloying occurs in a potential range that is more positive than the equilibrium potential of Li+/Li. However, Li–Au alloying induces changes in the image contrast of the CC surface (b–i). Specifically, many small bright rings with diameters of roughly 3 μm appear on the CC surface (b–d). These bright rings expand laterally and coalesce with adjacent rings. Subsequently, dark spots begin to expand from the centers of these rings and eventually cover the entire CC surface in j. Figure 2A plots the time evolution of the areal fractions of these darker areas. Compared to the voltage transient, a plateau at +0.15 V is observed, which occurs concurrently with a gradual increase in the areal fraction of the darker areas (b–j). Then, the voltage abruptly decreased to +0.05 V at t = 300 s. This decrease coincides with the complete coverage of the CC surface by the darker areas. HA-BSE and Lower-SE Observations To understand the transitions in image contrast shown in Figure 2B(b–i), we acquired in situ scanning BSE observations on the Au CC film surface (Figure 3A). This method can visualize the local Z contrast on the sample surface. Although the local variation in crystal orientation on the sample surface also slightly affects the BSE image contrast, this effect is negligible in this study because the crystallite sizes in the 30 nm-thick polycrystalline films are too small to be observed at the scale employed. The CC surface begins to show small dark spots at t = 40 s. As the Li–Au alloying progresses, these dark spots are enlarged, until they cover the entire CC surface. By the end of the initial voltage plateau at t = 240 s (Figure 3B), the dark spots completely cover the CC surface, reducing its overall brightness. This darkening indicates that the spatially averaged Z of the CC film decreases. Consequently, the entire CC surface transitions to lower Z states than the initial one (i.e., pure Au). (A) In situ HA-BSE images acquired after initiating Li-Au alloying at 100 μA cm–2. The time, displayed in seconds in the upper left corner, corresponds to the elapsed time t since the reaction started. Squares at t = 450 s indicate the positions where Li growth is identified. (B) Voltage and ξ transients during the observation in (A). Figure 4A shows an in situ observation of the Au CC film surface during Li–Au alloying using the lower-SE mode. The contrast variations reveal roughness differences in the sample surface. However, local variations in Z do not affect the contrast in a lower-SE image. Numerous small bumps appear upon the initiation of Li–Au alloying. These bumps grow unidirectionally into ridge-like shapes (t = 50–150 s), and their widths expand upon coalescence (t = 160–370 s). At t = 410 s, which coincides with the end of the voltage plateau at +0.15 V, the entire CC surface is completely covered by bumps (Figure 4B). (A) In situ lower-SE images acquired after initiating Li-Au alloying at 100 μA cm–2. The time, displayed in seconds in the upper left corner, corresponds to the elapsed time t since the reaction started. Squares at t = 810 s indicate the positions where Li growth is identified. (B) Voltage and ξ transients during the observation in (A). The small surface bumps in Figure 4A correspond to the dark spots observed during the initial voltage plateau in Figure 3A. This correlation suggests that the initially observed dark spots consist of elements with atomic masses lower than that of Au, which leads to a simultaneous volume expansion. The volume expansion induces the formation of bumps, potentially responsible for the bright rings observed in Figure 2B due to edge effects.32 Additional evidence of Au volume expansion, driven by Li–Au alloying, will be discussed later in the paper. Consequently, the small dark spots in Figure 2B, where the Z value decreases while the volume simultaneously expands, likely indicate the initiation sites for Li–Au alloying. Computational Simulation of Atomistic Li–Au Alloying Processes The following equation estimates the backscattering coefficient, η (0 < η < 1), in a BSE image321where ϕ is the electron incidence angle in rad. In this study, ϕ is π/9. The BSE image brightness (0 ≤ bBSE ≤ 255) is linearly proportional to the η. We approximated the average bBSE of the initial Au CC surface at t = 0 s as Z = 79. Additionally, we assigned the smallest bBSE in the center of the grown Li particles, as will be shown later, to Z = 3. bBSE and Z were then linearly correlated between these two end points. Hence, the relationship between bBSE and Z values can be expressed as2where A is 468, which is the proportional coefficient between η and bBSE. Figure 5A shows the time evolution of the Z distribution on the Au CC surface during the Li–Au alloying process and subsequent Li plating. Initially, the CC surface had an average Z of 79. At t = 40 s, small regions with Z values of 55–60 (yellow) randomly appear. These yellow regions then transition to green and gradually spread until they cover the entire CC surface. Throughout the process, the Z values of the regions that changed color earlier remain within the range of 55–60. By 260 s, the entire CC surface exhibits an almost uniform Z with a green color. Subsequently, Z gradually decreases to approximately 30 (light blue). At 460 s, the appearance of a blue color indicates that Li nucleation occurs (Z = 3). These observations suggest that the Li–Au alloying process is initiated from localized spots and spreads laterally. Moreover, further Li enrichment is possible only after the entire CC surface uniformly reaches a Z value of approximately 50 (green). (A) Time evolution of the Z distribution calculated by eq 2 based on the HA-BSE images in Figure 3A. (B) MD-simulated Au slabs consisting of 5184 orange-colored Au atoms, containing (a,b) 400, (c,d) 1200, and (e,f) 3600 green-colored Li atoms. (g) Li/Au atomic ratios as a function of z for 2, 6, 10, 14, and 18 Li layers. Structures are visualized using the VESTA package.33 As Li–Au alloying progresses, the spatial localization of Li-enriched regions gradually relaxes until t = 260 s (Figure 5A). To gain atomistic insight into this relaxation process, we conducted MD simulations. The MD-simulated atomic structures after 600 ps involve alloy systems that consist of slabs containing 5184 Au atoms with the addition of 400, 1200, and 3600 Li atoms [Figure 5B(a–f)]. The Li/Au ratio varies along the z-axis of the Au slab with added Li atoms [Figure 5B(g)]. Introducing additional Li atoms into the Au slabs increases the Li concentration, thereby intensifying the uneven distribution of the Li atoms. Thus, adding more Li atoms does not help to homogenize the nonuniform Li distribution. As a result, the localization of Li within a Au slab remains unrelaxed. However, when the local Li-enriched region reaches a nearly Li2Au composition, it begins to extend, leading to the homogenization of the nonuniform Li distribution while saturating the high Li concentration region to become almost Li2Au. This phenomenon is attributed to the significant energy cost associated with maintaining boundaries between Au and a phase richer in Li than Li2Au. The MD simulations qualitatively explain the experimental results. The Li–Au alloying process is initiated at random spots (Figure 5A). These Li–Au alloying initiation sites progress Li enrichment until they reach Z = 45–50 (green), while other regions remain at Z = 75–85 (red). Once the alloying initiation sites turn green, the local reduction in Z temporarily stops and the green areas expand laterally. Comparing this color transition in Figure 5A with the time evolution of ξ in Figures 2A, 3B, and 4B, the green color roughly corresponds to ξ = 0.9–1.5. The MD simulation results reproduce the homogenization of Li composition, which begins after the alloy composition reaches a certain threshold (Li0.9-Li1.5Au in the experiments and Li2Au in the MD simulations). Dealloying of Li–Au Dealloying Process of Li–Au After the Li plating shown in Figure 3A, we reversed the polarization to proceed with Li stripping and dealloying. We observed the process via in situ BSE imaging (Figure 6A). Figure 6B displays the voltage transient during the Li stripping and dealloying at 100 μA cm–2. Li–Au dealloying is evident in the in situ BSE images after t = 550 s (after the point labeled l in Figure 6A) due to the emergence of numerous holes in the CC film. These holes are attributed to the volume shrinkage that occurs during Li–Au dealloying, indicating that the Au film undergoes a volume expansion during the Li–Au alloying process, as previously mentioned. A rapid increase in the cutoff voltage to +2.0 V is observed at t = 770 s. Consequently, the overall behavior closely follows thermodynamic order: Li stripping occurs first, followed by Li–Au dealloying, with volume shrinkage occurring simultaneously with Li–Au dealloying. (A) (Upper row) in situ HA-BSE images and (lower row) Z-converted images. (B) Variations in the voltage and the average Z value for each square region, numbered 1–5, during Li stripping and subsequent Li–Au dealloying at 100 μA cm–2. Numbers 1–5 correspond to those in (A). The Coulombic efficiency observed in Figure 6B is 43%. During the Li stripping process, the Li particles that had grown outward from the CC film become hollow, indicating that the interior Li is stripped and dissolves back into LiPON (Figure S2). Previous studies have reported similar phenomena.16,34−36 The undissolved surface layers of the Li particles, which are composed of lithium oxides and carbonates, contribute to the low Coulombic efficiency. Moreover, the Z value of the remaining CC film region did not revert to its initial value of 79. Bach et al. reported that during Li–Au dealloying, the alloy phases did not revert along the same pathway as observed during electrochemical Li–Au alloying.28 They found that Au alloy phases, including Li-poor amorphous, Li3Au5, and LiAu2, remained after dealloying. To further identify the remaining Li–Au phases, additional studies employing operando high-energy X-ray diffraction are required, as previously discussed. Figure 6A also shows the Z-converted images. Moreover, Figure 6B shows the time evolutions of the average Z values within the small square regions labeled 1–5 displayed in Figure 6A. Each region measures 0.7 μm × 0.7 μm. The average Z in a square region initially increases until point d, even though the only process occurring is Li stripping (Li–Au dealloying of the CC film has yet to begin). The likely reason for this increase in Z is stripping of plated Li, which grows under the Au CC film and in its grain boundary regions, causing the apparent increase in Z. As previously mentioned, the electron penetration depths are approximately 26 nm for Au and 596 nm for Li in this study. These depths are sufficiently long to reach the bottom of a Li–Au CC film. The average Z then remains between 21 and 25 until point f (t = 250 s). From points g–k (t = 300–500 s), the average Z slowly rises again, which coincides with the detection of a voltage increase followed by a plateau at +0.25 V. The timing of the Z increase slightly varies depending on the location marking the initiation of the dealloying process in the Li–Au CC film. Subsequently, small holes appear in the CC film at point l (t = 550 s). The average Z begins to rise rapidly as the cutoff voltage is approached. The observation that the average Z of the CC film did not fully revert to the original value of Au (= 79) is consistent with the result of low Coulombic efficiency. The observed increase in Z through the BSE images aligns with the direction of the electrochemical dealloying reaction. This examination of the contrast changes in HA-BSE images supports the previous discussion on the progression of Li-rich regions observed during Li–Au alloying, as shown in Figure 5. Lithium Nucleation This section discusses Li nucleation on a Li–Au alloy CC film. Figure 7A shows the time evolution of the BSE images for a Li–Au alloy CC film surface before and after Li nucleation. These images were captured in sequence, following the observation in Figure 3A. The voltage transient curve in Figure 7B corresponds to an enlarged view of the data presented in Figure 3B, centered at t = 500 s. The blank circle at t = 445 s represents the negative voltage peak. The overpotential at the negative voltage peak (ηnc) in Figure 7B is 9 mV by subtracting the ohmic loss and the overpotential of the Li counter electrode (ηCE) from the absolute voltage value (Figures S3 and S4). However, a previous study on a Cu CC film reported ηnc of approximately 78 mV35 at the same current density (100 μA cm–2). These findings are consistent with previous reports, which indicate that ηnc is smaller when CC metals form alloys with Li.7,9,19 (A) (a–l) In situ HA-BSE images during Li nucleation from a Li–Au alloy CC film at 100 μA cm–2. (i) BSE image overlaid with the corresponding Z contrast image. (B) Voltage transient during the observation in (A). Images (a–l) in (A) correspond to (a–l) in (B). I, II: Time evolutions of binarized Li particles from t = 420–590 s. Regions with bBSE below 70 are rendered black, and those above are rendered white. Time intervals of the images are 10 s. I and II correspond to the Li particles denoted by I and II in A, respectively. a to l correspond to the same alphabet in A. Yan et al. propose that this low ηnc may be due to the solubility of Au in Li.9 In our study, nucleation of the Li metal phase is hypothesized to occur at the Li–Au alloy/LiPON interfaces, with more than half of these interfaces occupied by Li atoms. This is because the experimental composition of the Li–Au alloy is roughly Li2Au. At this stage, it is suggested that a monatomic layer of Li, resulting from UPD, is no longer stable on the Li-rich alloy surface, even if UPD occurs at the Au/LiPON interface. The low ηnc of 9 mV is attributed to the high proportion of Li atoms on the substrate surface, implying that the Li–Au alloy surface does not behave as a completely foreign substrate for the formation of three-dimensional Li clusters. At t = 450 s, eight Li nuclei are identified, as indicated by the yellow dashed circles in (b) (Figure 7A). However, only two remain near the right edge of the field of view and continue to grow in size. The other six redissolve back into the Li–Au alloy film. Moreover, the surrounding regions of the dissolved Li nuclei appear darker compared to the other areas. This darker appearance is likely due to Li redissolving back into the Li–Au film. The superimposed Z-contrast image reveals a narrow channel with a smaller Z (blue color) directed toward the surviving Li, as indicated by the arrows. This observation suggests that some nucleated Li redissolves back into the Li–Au alloy film, and these Li atoms are subsequently supplied to the surviving Li. Figure 7I,II shows binarized images of two Li particles, which are labeled as I and II in Figure 7A. Particle I exhibits gradual growth, but the growth rate accelerates after point f (Figure S5). In contrast, Particle II expands until reaching point f, after which it gradually disappears. Throughout the redissolution process of Particle II, the plated Li permeates through the CC film, likely via grain boundaries, retracting reversibly without rupturing the CC film. Unlike nonalloyed CC films such as Cu, a distinct mechanical failure mode like cracking17,35,37 was not observed. A similar noncracking behavior has been reported for Pt CC films, which form various intermediate alloy phases with Li.19 Classical nucleation theory cannot fully explain the nucleation process, in which some nuclei redissolve after nucleation. Classical nucleation theory posits that nuclei continue to grow in size after surpassing a critical nucleation size.38 Instead, the phenomenon observed in this study seems to indicate Ostwald’s rule of stages,39 which explains the nucleation of a metastable phase before the thermodynamically stable phase. This phenomenon is empirically observed under high supersaturation conditions and when the interfacial energy between the metastable phase and the parent phase is lower than that between the stable phase and the parent phase. In our study, the parent phase is assumed to be Li2Au. The critical composition that triggers Li nucleation ranges from 1.6 to 2.9, as shown in Figures 2A, 3B, and 4B. This variability may primarily be attributed to thickness errors of the sample. For convenience, here, we denote the average composition as Li2Au. The metastable phase is Li, and the stable phase is LiξAu (new alloy phase). Although the value of ξ cannot be precisely determined, it should exceed 2. Both Bach et al. and Leenheer et al. identified the formation of Li3Au, a thermodynamically stable phase, during the electrochemical Li–Au alloying in ionic and organic liquid electrolytes, respectively.28,29 As the new alloy phase grows, a significant interfacial energy cost likely arises due to the lattice mismatch associated with volume expansion (Figure 8). The MD simulations suggest that the creation of new interfaces of a Li-richer phase becomes energetically unfavorable once the Li composition reaches a specific threshold. Hence, nucleating the new alloy phase becomes difficult and explains why this phase emerges only through metastable phase nucleation. Schematic illustration of Li nucleation on a Li–Au alloy CC film. Li emerges on the film surface through grain boundaries of the Au CC film, which is kinetically supersaturated with Li. In contrast, Li is not subjected to stress when it grows out of the CC film surface, which is free of stress. Although the interfacial energy between Li and the new alloy phase is unknown, it is likely low because a solid solution-like interface should form between Li and the new alloy phase. Consequently, a lack of stress combined with a low interfacial energy facilitates Li nucleation as a metastable phase before new alloy phase nucleation. Three distinct Li fluxes can be defined at different interfaces: j1 represents the Li flux through the LiPON/Li interface, j2 denotes the Li flux through the Li/LiξAu (ξ > 2) interface, and j3 indicates the Li flux through the new alloy phase/Li2Au interface. When j1 > j2, j3, Li nucleates and grows continuously. Li nucleation sites that meet this condition persist, allowing Li nuclei at these sites to continue growing. At these Li growth sites, the local j1 increases as they grow, drawing j1 fluxes from the surrounding regions. Consequently, some sites near the surviving nucleation site undergo Li redissolution. Statistical Analysis of the Li Nucleation Sites Previous discussions have focused on the role of the Li redissolution process in determining the stable growth sites of Li. We used the equation proposed by Milchev et al.40 to analyze and compare the spatial distribution of Li growth sites in the data obtained from Au, Pt,19 and Cu17 CC films with that of randomly distributed particles, which is written as3where is the probability for a given particle to have its nth nearest neighbor within a distance of rn and N0 is the particle number density. N0 can be obtained from SEM results (e.g., Figure 9a–d). For instance, when the Li growth sites are arranged in a honeycomb pattern, with equal rn for all growth sites, (cumulative probability) exhibits a discrete increase from 0 to 1.0 at a certain value of rn (Figure 9e). SEM surface images comparing Li growth sites at 100 μA cm–2 on the (a) Au CC film with those on (b) Pt and (c) Cu CC films (adapted with permission from refs (17 and 19) Copyright 2015 and 2018 The Electrochemical Society). Scale bars: 5 μm. (d) Lower-SE image of the initiation sites for Li–Au alloying (taken from Figure 4A). Cumulative probabilities as a function of distance for (e) dots arranged in a honeycomb pattern with N0 = 0.63 × 107 cm–2 and (f–h) Li nucleation sites with Au, Pt, and Cu CC films (N0 = 0.63 × 107, 1.36 × 107, and 0.36 × 107 cm–2, respectively) and (i) Li–Au alloying initiation sites (N0 = 1.38 × 107 cm–2). Red, blue, and green indicate first, second, and third nearest neighbors, respectively. The solid lines in Figure 9f–i show the experimental results of obtained from SEM images as a function of rn. The dotted lines represent theoretical curves obtained by eq 3 when N0 is the same as those for the solid curves. The experimental and theoretical curves gradually increase from 0 to 1.0 with increasing rn. However, the experimental curves exhibit a more rapid increase from 0 to 1.0 compared with the theoretical one. This difference is more pronounced in the cases of Au and Pt (f and g). These findings suggest that nearby Li growth sites are less likely to survive because is smaller at shorter distances than in a perfectly random distribution (Poisson distribution). Conversely, the initiation sites for Li–Au alloying exhibit a random distribution because the experimental curve closely aligns with the experimental one (i). Thus, Li–Au alloying is initiated from randomly distributed sites, whereas nucleation of Li from a Li–Au alloy CC film follows a distinct, nonrandom pattern. The coefficient of determination, R2, can be used to quantify the deviation from the Poisson distribution in Li growth sites (Supporting Information). Calculating R2 involves comparing an experimental result (solid line) to its corresponding theoretical curve (dashed line). An R2 value of 1.0 indicates a perfect fit with the Poisson distribution, whereas a negative R2 value suggests that the model fit is less accurate than simply using the average of the experimental values to predict all data points. Table 1 presents the R2 values for the first, second, and third nearest neighbors for the Au, Pt, and Cu CC films. Compared to the value for Cu, the R2 values are clearly smaller for Au and Pt, suggesting that the Li growth sites have a higher degree of nonrandomness in Au and Pt CC films. In contrast, the R2 values for the initiation sites of Li–Au alloying are close to 1.0, indicating that local Li–Au alloying initiation occurs randomly in space. This indicates that the initiation sites for Li–Au alloying do not coincide with stable Li growth sites. That is, the reaction sites for Li–Au alloying, which are initially distributed randomly, subsequently redistribute in a nonrandom pattern during Li nucleation (repulsion of neighbors). This phenomenon occurs because some of the Li nuclei redissolve back into the Li–Au alloy, even during Li plating (Figure 10). Table 1R2 between the Solid (Li Growth Sites with Au, Pt, and Cu CC Films and Li–Au Alloying Initiation Sites) and Dashed Lines in Figure 9f–iCCAuPtCuLi–Au alloying1st–0.11–0.030.750.922nd0.340.350.740.933rd0.410.530.820.94 R2 between the Solid (Li Growth Sites with Au, Pt, and Cu CC Films and Li–Au Alloying Initiation Sites) and Dashed Lines in Figure 9f–i Schematic illustration of the redissolution of Li nuclei resulting in the distribution of growth sites in a nonrandom manner. Li at nonpreferred growth sites dissolves back into Li–Au alloy, providing Li sources to the preferred growth sites. Li redissolution occurs during the transition from a metastable state (more Li nuclei appear) to a stable state (less Li nuclei continue to grow), resulting in a nonrandom distribution of Li growth sites. This phenomenon deviates from classical nucleation theory, which forbids the reverse process (redissolution) after surpassing a critical nucleation size. Previous studies have demonstrated that the presence of interfacial alloy layers leads to the homogenization of current density distribution at the Li/SSE or CC/SSE interfaces.7,8,10,12,13,15 This study microscopically reveals the function of the interfacial alloy film in homogenizing the Li reaction site distribution. A local redissolution process occurs concurrently with Li plating, creating kinetically favorable pathways. This finding was achieved by using in situ HA-BSE and lower-SE observation techniques. Although a comparison with operando X-ray photoelectron spectroscopy measurements is necessary to rigorously ascertain the average surface alloy composition, this study demonstrates that these in situ observation techniques are powerful analytical tools. They provide valuable insights into the mechanisms governing the formation of Li alloy phases in various systems such as Li–Si and Li–Ag. Conclusions We utilized in situ SEM observations to investigate the Li–Au alloying and subsequent Li nucleation processes using HA-BSE and lower-SE imaging. Li–Au alloying was initiated at randomly distributed sites. However, once the composition of the Li–Au film approaches approximately LiAu, further local Li enrichment stops. Then, the relaxation of the nonuniform Li–Au composition distribution begins, and the alloying process initiates a uniform progression. These findings were qualitatively supported by MD simulations. Additionally, we observed Li nucleation on a Li–Au alloy CC film. In-situ HA-BSE observations indicated that some Li nuclei redissolved into the Li–Au alloy film during Li plating, which aligns with Ostwald’s rule of stages. Li, as a metastable phase, likely nucleated earlier than a more thermodynamically stable, Li-richer alloy phase due to its low interfacial energy with the parent alloy phase and its subsequent release from stress upon appearance on the film surface. The redissolution of Li facilitated the organization of Li growth sites, which were distributed nonrandomly. These results suggest that the alloy layer functions as a buffer, homogenizing the nonuniformity of the reaction sites.