Thus, the mechanism by which short circuit events occur in all‐solid‐state Li‐ion batteries is unclear. Here, a novel technique is employed to monitor lithium metal penetration into solid‐state electrolytes (SSEs) during electrodeposition, using four different SSEs: (1) amorphous 70/30 mol% Li 2 S‐P 2 S 5 , (2) polycrystalline β‐Li 3 PS 4 , (3) polycrystalline Li 6 La 3 ZrTaO 12 garnet, and (4) single‐crystalline Li 6 La 3 ZrTaO 12 garnet. The surface defect population of these SSEs was intentionally varied, through surface treatments as well as internal microstructure. These experiments made it possible to identify both similarities and differences between lithium penetration phenomena in the different SSEs. A chemo‐mechanical model is also developed to evaluate the stresses generated in a sharp flaw during electrodeposition of Li metal. The experimental results and analytical model together allow an interpretation of the mechanism of Li metal penetration into inorganic SSEs.

Nonflammable inorganic solid electrolytes, paired with Li metal anodes, could result in high energy density yet safe rechargeable lithium batteries. 11 - 15 As reviewed by Takada, 12 inorganic solid electrolytes have now been widely studied, 16 - 20 but are not yet commercialized. Monroe and Newman have suggested that dendrite growth during the plating process may be suppressed if the liquid electrolyte is replaced with a Li‐ion conducting solid electrolyte of a sufficiently high shear modulus. 21 , 22 According to this criterion, numerous inorganic solid electrolytes should be able to suppress dendrite formation. However, multiple research groups have recently reported cases where ceramic solid electrolytes paired with a Li metal anode experience a short circuit event. 23 - 27 Several attempts to suppress the shorting via solid electrolyte composition alterations and electrode/solid electrolyte interfacial modifications have been made, 24 , 28 - 31 but with limited success. Only in thin film batteries have Li metal negative electrodes paired with inorganic solid electrolytes (e.g., LiPON) been able to avoid dendrite penetration over extended cycling. 32 The reasons for this difference in behavior between bulk and thin film batteries are not understood. As pointed out in two recent reports, 33 , 34 similar short circuit events were investigated in Na‐ion conducting solid electrolytes in liquid Na batteries in the 1970–1980s. 35 - 42 In these cases, the short circuit event was attributed to stresses created by molten sodium metal plating into cracks within the solid electrolyte bulk, resulting in crack propagation and eventual failure.

The ever‐increasing demand for mobile energy storage has led to extensive development of lithium‐ion batteries due to their exemplary performance, energy density, and cycle life. 1 - 3 However, safety concerns have prompted the need for alternative materials while preserving benefits. 4 Furthermore, vast improvements in energy density may be achieved with lithium metal anodes owing to their high gravimetric capacity (3869 mA h g −1 ) and low density (0.534 g cm −3 ). However, adoption of rechargeable lithium metal batteries has been unsuccessful thus far due to safety concerns associated with short circuits that occur when Li dendrites grow through the liquid electrolyte during the charging process. 4 - 6 Although several approaches have reduced dendrite formation, 7 - 10 to date the phenomenon has not been avoided under all relevant conditions.

2 Results and Discussion

The experimental apparatus used in this work is shown schematically in Figure 1. A solid electrolyte pellet is placed in contact with Li metal foil at the bottom surface and with a metal electrode at the top surface. The top electrode is a spring‐loaded contact to a brass or gold electrode. Galvanostatic experiments are conducted with an applied potential that transports lithium from the bottom electrode through the SSE, causing Li metal plating at the interface between the SSE and the top electrode. We assume that this interface has defects of a distribution of sizes, and that electrodeposition at this interface first fills any defects (cracks and voids) at the surface of the SSE. This is reasonable because once lithium metal has begun to deposit, very little overpotential is required to fill surface defects. It has been observed that electrodeposited lithium is able to fill even nanoscale porosity.43 It is important to note that the phenomenon of interest in the present work is not the filling of initial surface defects or porosity. This only requires small overpotentials if lithium metal does not wet the SSE and a capillary pressure must be overcome, and zero (or negative) overpotential if wetting is present and capillary pressure assists the filling of surface defects by lithium metal. It is instead the propagation of lithium metal into a dense SSE, which requires opening of cracks against mechanical resistance, that is of interest.

Figure 1 Open in figure viewerPowerPoint Schematic of apparatus for Li plating on a metal electrode in contact with a solid electrolyte.

Galvanostatic experiments were conducted, in part to test the proposal by Sharafi et al.34 that there is a critical current density above which lithium dendrites initiate and propagate through the solid electrolyte and result in a short‐circuit event. Galvanostatic current densities were thus chosen to lie in the neighborhood of the critical current density, determined by a step‐wise current density ramp experiment until a short‐circuit event is observed. An optical microscope is used to observe the lithium plating behavior, and the entire apparatus was housed in an argon‐filled glove box with oxygen and water content below 1 and 5 ppm, respectively. Scanning electron microscopy (SEM) and atomic force microcopy (AFM) were used for surface analysis. Since exposure of the samples to air results in nearly immediate surface reactions, a method for anaerobic transfer of the samples from the glove box to the vacuum chamber of the SEM is required. A vacuum‐operated transfer box was designed and constructed for this purpose.44

Experiments performed on the 70/30 mol% amorphous Li 2 S‐P 2 S 5 sample (hereafter referred to as glassy LPS) are discussed first. The ionic conductivity of this material has been measured in other work to be 0.3 mS cm−1 using electrochemical impedance spectroscopy (EIS).45 X‐ray diffraction (XRD) was also used to confirm that the sample is amorphous.45 The surface of this sample was freshly cleaved prior to the experiment and showed a mirror‐like finish. This sample is therefore free of defects due to cutting or polishing, or grain boundaries, multigrain junctions, or pores present in polycrystalline samples. It has the most defect‐free surface of the samples studied herein.

Optical microscopy images recorded during galvanostatic lithium plating at two different locations on the surface, using the brass electrode, are shown in Figure 2a–d and Figure 2e–h, respectively. In the first instance (Figure 2a–d), the electrode tip was placed in contact with a surface region free of any observable pores or cracks. We believe this surface to be the most defect‐free surface evaluated throughout this work. In the second instance (Figure 2e–h), the electrode tip was placed in contact with an area that was intentionally scratched with a diamond‐tipped tool. Due to the low fracture toughness of LPS, measured to be only 0.23 MPa m1/2 in recent work,45 atomically sharp surface flaws are undoubtedly produced. The surface flaws induced in the glassy LPS were measured to be ≈25 µm deep using laser scanning microscopy. In both experiments, a constant current of 0.8 µA was applied, which corresponds to a current density of ≈10 mA cm−2 based on the tip area. This current density in turn corresponds to a Li metal deposition rate of ≈100 µm3 s−1. Both experiments were conducted over ≈20 h, resulting in 16 µA h of charge passed and ≈8 × 106 µm3 of deposited lithium metal.

Figure 2 Open in figure viewerPowerPoint Optical microscopy images showing lithium plating onto a brass tip electrode (≈10 mA cm−2 initial current density) in contact with the fracture surface of glass LPS. The total charge passed, and corresponding volume of lithium metal deposited, are labeled at top. a–d) Lithium deposition at a pristine as‐fractured surface. Surface deposition of lithium metal without crack formation or propagation is observed. e,f) Lithium deposition at a region precracked using a diamond‐tipped tool. No surface deposition of lithium is observed, whereas growth of a crack from the initially damaged region is observed. At top left of images (e)–(h) is seen the lithium deposit from the preceding experiment.

In the first instance where the electrode tip is in contact with the as‐fractured LPS surface, the sequence of images in Figure 2a–d show that lithium metal deposits on the surface of the LPS and gradually propagates laterally from the electrode contact point. No cracking or surface degradation of the LPS was evident, and the cell did not short‐circuit during the experiment.

However, when the electrode was placed in contact near the precracked area, there is no evidence of lithium accumulation on the surface of the LPS under the same galvanostatic conditions (0.8 µA). Instead, as seen in Figure 2e–h, cracks formed and extended into the sample as the experiment progressed. In Figure 2f, the extension and opening of the metal‐filled crack is indicated by the outlined box. We interpret that the progressive crack opening is driven by the electroplating of lithium metal into one or more sharp flaws (Griffith flaws) at the surface of the SSE, despite the low yield strength of lithium metal (0.5 MPa),46 which might otherwise suggest that it should be able to flow and deposit on the surface.

Similar experiments were then conducted on single‐crystal Li 6 La 3 ZrTaO 12 garnet (LLZTO) samples. The ionic conductivity of the sample was measured via EIS to be 0.2 mS cm−1 (Figure S1, Supporting Information). The crystal structure and phase purity have been confirmed via XRD experiments.47 Having seen the impact of surface defects on lithium metal penetration in glassy LPS, the surface of the LLZTO crystal was subjected to two different polishing schemes in order to vary the severity of surface flaws. In one instance, the single crystal was ground using 400 grit SiC abrasive paper, producing a microscopically rough surface with widespread fracture features, as seen in Figure 3a. To obtain a surface with much finer defect size, a highly polished single‐crystal surface was also prepared, as shown in Figure 3b. This surface was prepared by polishing LLZTO single‐crystal samples with successively finer abrasive papers, culminating in a final polish with 50 nm alumina paste. Atomic force microscope (AFM) scans of the two surfaces are shown in Figure 3a,b, respectively. The root‐mean‐square (rms) roughness of the rough surface is 200 nm, and that of the polished surface is 4 nm (each obtained for a scan size of 10 µm × 10 µm). However, of greater importance are the largest flaws on the respective surfaces. It has been shown that during surface grinding of brittle materials, subsurface damage is produced that extends to a maximum depth determined by several factors, including the size distribution of particles (the largest particles bearing most of the load) and the normal force while grinding.48 The maximum damage depth for standard abrasives is observed to range from one to several times the mean particle size.48 Since the ANSI 400 grit specification corresponds to a D50 particle size of 22 µm, it is reasonable to assume that LLZTO surfaces ground with this grit size have surface cracks extending 30–50 µm into the SSE. SEM images of the surface support this assumption (Figure 3c,e). The largest flaws on the highly polished surface were found, using SEM, to be polishing scratches of ≈0.4 µm width, shown in Figure 3d,f, but possibly deeper. Thus, for these two surfaces, we assume the largest surfaces flaws to be in the range of 30–50 µm and 0.4–1 µm, respectively.

Figure 3 Open in figure viewerPowerPoint 48 AFM scans of single‐crystal LLZTO surfaces a) ground with 400 grit SiC abrasive paper and b) polished with successively finer media to a final particle size of 50 nm. The rms roughness is 200 and 4 nm, respectively, measured with a scan size of 10 µm × 10 µm and a tip radius of 7 nm. However, SEM images reveal much larger defects than the mean. c,e) The 400 grit finished surface shows features of ≈10 µm scale, although literaturesuggests that subsurface damage may extend as deep as 50 µm. d,f) The 50 nm polished surface had maximum‐sized defects due to residual polishing scratches that are ≈0.4 µm in width.

Gold electrodes were then sputtered onto the samples, under which lithium metal is electrodeposited during galvanostatic testing. Because gold readily alloys with lithium metal, it has a negligible nucleation barrier (overpotential) for lithium plating at the inception of the experiment.43 This allows our experiment to test the subsequent crack growth and lithium penetration process. Constant current densities of 0.01, 0.1, 0.5, 1, and 5 mA cm−2 (the area being that of the sputtered gold electrode) were applied in a stepwise sequence, with each step lasting for 15 min. The cells reached a short‐circuit state in all cases during the 1 or 5 mA cm−2 steps, with no apparent correlation in short‐circuit conditions with surface roughness. (Four experiments were conducted on the rough surface and nine experiments on the smooth surface.) The absence of any clear differentiation in the current density at which crack propagation occurs for the rough and smooth suggests that both surfaces contain defects of sufficient severity to be “supercritical” under present electrodeposition conditions.

An example of the sequence of crack formation and propagation at 5 mA cm−2 current density (based on the gold electrode area) on LLZTO polished to 50 nm surface finish, is shown using in situ optical microscopy in Figure 4a–d. The galvanostatic current was applied for 15 min, passing a total charge of 2 µA h, which corresponds to a Li‐plated volume of ≈1 × 106 µm3. Figure 4c captures a moment about halfway through the experiment when a crack (white silhouette) has grown from under the gold electrode into the field of view. The crack then extends further as Li‐plating continues (Figure 4d). A second example, conducted at lower current density of 1 mA cm−2, is shown in Figure 5. During real‐time observation, cracks could be observed propagating into the SSE away from the gold electrode. The dark contrast reflections in the image correspond to cracks filled with electrodeposited lithium metal. In the inset figure in Figure 5b, one such lithium‐filled crack is shown after the single crystal was polished to half its initial thickness of 3 mm. Since multiple cracks were created under the sputtered gold electrode, the specific crack responsible for the short circuit was not identified. These results conducted on the polished LLZTO show at practical current densities (1 and 5 mA cm−2), lithium metal penetration can occur from an electrode–SSE interface with starting defect size as small as ≈1 µm.

Figure 4 Open in figure viewerPowerPoint a–d) Optical microscopy images of a polished single crystal of LLZTO (4 nm rms roughness) with sputtered gold electrode during galvanostatic deposition of lithium metal beneath the gold electrode at 5 mA cm−2 current density. Uniform plating behavior appears to take place until a crack is observed to propagate (white silhouette).

Figure 5 Open in figure viewerPowerPoint a) Optical microscopy image of a single‐crystal LLZTO polished with 50 nm particle size polishing media sample, after deposition of the gold electrode. Scratches on the gold film are due to positioning of the contact probe. b) After lithium deposition at 1 mA cm−2 to short circuit, cracks are clearly visible. In the inset, a crack is still clearly visible when the sample has been polished to half its original thickness. The dark background in (b) is the reflection of lithium metal filling the cracks.

Polycrystalline SSE samples were next investigated. Galvanostatic lithium deposition was conducted, using the configuration of Figure 1, onto a densely pressed polycrystalline β‐Li 3 PS 4 sample. The ionic conductivity of the sample was measured via EIS to be 0.2 mS cm−1 (Figure S1, Supporting Information). XRD confirmed that the sample is single‐phase β‐Li 3 PS 4 (Figure S2a, Supporting Information). As shown in Figure 6a–c, the β‐Li 3 PS 4 powder was cold‐pressed into a pellet at 700 MPa uniaxial pressure, after which it was translucent in appearance. The surface of the pellet appeared smooth in optical microscopy (Figure 6a), but SEM imaging revealed fine sub‐micrometer‐scale pores or cracks at the surface (Figure 6e). A current of 4 µA was passed through the cell, corresponding to a high current density of ≈50 mA cm−2 based on the tip electrode area (lithium metal deposition rate of ≈500 µm3 s−1). The duration of the experiment for which data are shown in Figure 6 is 2 min, resulting in ≈0.13 µA h of charge passed and equivalent to ≈6 × 104 µm3 of plated lithium metal. Upon inspection, it was clear that the majority of lithium deposition occurred within the bulk of the β‐Li 3 PS 4 pellet and not on the surface, as can be seen in Video S1 (Supporting Information). Figure 6d shows, in transmission optical microscopy, that the lithium metal deposition occurs in a characteristic branching pattern.

Figure 6 Open in figure viewerPowerPoint a–c) Lithium metal deposition at high current density (≈50 mA cm−2) propagates into densely pressed polycrystalline β‐Li 3 PS 4 from a brass tip electrode. d) Viewed in transmission optical microscopy, the lithium metal network shows a branching pattern. e) The starting as‐pressed surface of the β‐Li 3 PS 4 polycrystal exhibits sub‐micrometer (≈200 nm length scale) cracks or pores.

After the galvanostatic experiment, the β‐Li 3 PS 4 pellet was fractured to expose the lithium metal network shown in Figure 6d. Scanning electron microcopy was used to record images of the fracture surface using secondary electrons (Figure 7a,b,d) and backscattered electrons (Figure 7c), respectively. The sample in Figure 7a–c experienced air exposure for <1 min, while the sample in Figure 7d was transported in a vacuum‐operated transfer box.44 This transfer box is designed to be sealed while in the glove box under high purity argon at 1 atm pressure, but to spontaneously open in the SEM chamber when the vacuum reaches a critical value. In Figure 7a–c, the LPS has a more rounded appearance, which we attribute to volume‐expansive surface reactions that occurred upon exposure to air. This feature is absent in Figure 7d, suggesting that the transfer box has served the intended function. The lithium metal in these images is readily identified in two ways. First, each of the secondary electron images show the fracture features characteristic of a ductile phase, namely, elongated filaments that are necked down at the tips where they have separated from the opposing surface. This ductile phase must be lithium metal rather than LPS, given the low measured fracture toughness of LPS (0.23 MPa m1/2).46 Second, a comparison between the secondary electron images and the backscattered electron image in Figure 7c shows that the phase corresponding to the ductile fracture features has very low backscattered electron contrast, as expected from lithium metal. Between the branches of lithium metal are seen polycrystalline aggregates of β‐Li 3 PS 4 particles. The electrodeposited lithium metal has propagated in a cellular manner, along pore channels, grain boundaries, or both. A recently published study shows similar cellular lithium metal penetration through Al‐doped LLZO,26 although in that instance the penetration appears to follow most grain boundaries rather than isolating polycrystalline aggregates as in the present instance.

Figure 7 Open in figure viewerPowerPoint 3 PS 4 sample into which Li metal has electrodeposited via the experiment shown in Figure 3 PS 4 particle agglomerates. The lithium metal network appears to have penetrated along pore channels and/or grain boundaries of the polycrystalline β‐Li 3 PS 4 , resulting in isolation of packed polycrystalline aggregates. Scanning electron microscope images of the fracture surface of a polycrystalline β‐LiPSsample into which Li metal has electrodeposited via the experiment shown in Figure 6 . a,b,d) Secondary electron images of the fracture surface. c) Backscattered electron images of the fracture surface. Samples in (a)–(c) were exposed to air for <1 min while the sample in (d) was transported in a vacuum operated transfer box. The secondary electron images show ductile fracture of a cellar network of lithium metal within the cells of which are β‐LiPSparticle agglomerates. The lithium metal network appears to have penetrated along pore channels and/or grain boundaries of the polycrystalline β‐LiPS, resulting in isolation of packed polycrystalline aggregates.

In the current experimental design, the cell impedance is dominated by the solid electrolyte layer. An advantage of this design, from the viewpoint of experimental diagnostics, is that the ohmic resistance of the cell can be used to monitor in situ the growth and penetration of a lithium metal network through the solid electrolyte. Before a short‐circuit occurs due to complete lithium metal penetration, there is typically a decrease in cell resistance as the lithium metal network advances through the solid electrolyte. We took advantage of this behavior and conducted a series of experiments to electrically monitor lithium metal penetration under varying electrodeposition conditions.

In one such experiment, the cell voltage under galvanostatic conditions was monitored as a function of time, using polycrystalline β‐Li 3 PS 4 and LLZTO as solid electrolytes (Figure 8a). The ionic conductivity of the polycrystalline LLZTO sample was measured via EIS to be 0.16 mS cm−1 (Figure S1, Supporting Information). The sample was confirmed to have a single garnet phase using XRD (Figure S2b, Supporting Information). Under a constant current of 2 µA, which is ≈25 mA cm−2 initial current density based on tip area, the lithium deposition rate is ≈250 µm3 s−1. For both LPS and LLZTO, there is an initial rapid drop of the cell voltage over a few minutes. Concurrent with this voltage decrease, rapid propagation of the lithium network was observed in the β‐Li 3 PS 4 pellet via optical microscopy. After ≈5 min (0.16 µA h charge passed, 8 × 104 µm3 of plated Li), the cell voltage stabilized, and at the same time, growth of the lithium network appeared to have stopped. Continuing the experiment to 15 min (0.48 µA h charge passed, 24 × 104 µm3 of plated Li) did not result in a short‐circuit of the cell. Even after 24 h of the experiment (data not shown, 46 µA h charge passed, 1.15 × 107 µm3 of plated Li), a cell short‐circuit was not observed. Similar behavior was observed with the polycrystalline LLZTO (Figure 8a). A plausible interpretation of the cell voltage stabilization is that the lithium metal network branching during growth decreases the local current density at the filament tips to a value below a critical current density necessary for propagation. The sharp stepwise decrease in the cell voltage at about 4 min is attributed to preferential Li plating inside a newly formed crack aligned with the thickness direction of the LLZTO crystal.

Figure 8 Open in figure viewerPowerPoint 3 PS 4 and LLZTO solid electrolyte pellet. At a constant lithium plating rate of 250 µm3 s−1 (2 µA), the cell potential initially decreased while lithium network growth was simultaneously observed in β‐Li 3 PS 4 . After about 8 × 104 µm3 of plated lithium (5 min), the voltage reaches approximately a steady‐state value, suggesting that the lithium network is no longer propagating forward through the sample. b) A galvanostatic current of 2 µA is applied for 2 h, following which the current is linearly ramped from 2 to 17 µA at a rate of 0.025 µA s−1, and the cell voltage is monitored. The dashed lines show the expected rate of the cell voltage if the ohmic resistance of the cell is unchanged throughout. The solid curves show the measured voltage versus time. In each case, a negative deviation from linearity indicates that the lithium metal network is advancing. This occurs at ≈9 and ≈15 µA for the polycrystalline LLZTO and β‐Li 3 PS 4 , respectively. c) Shallow galvanostatic cycling of a cell using polycrystalline LLZTO, cycled at ±2 µA. Time spent at each segment before the direction of current is changed is labeled. A decrease in cell voltage indicates advancement of the lithium network. Compared to uninterrupted galvanostatic deposition, cyclic plating and deplating of lithium metal causes much faster growth of the lithium metal network, with a short circuit occurring in the seventh cycle after net lithium electrodeposition of 5 µA h corresponding to 2.5 × 106 µm3 of deposited lithium. The voltage across a cell of the configuration in Figure 1 can be used to monitor the advancement of the lithium metal network through the solid electrolyte. a) Voltage during galvanostatic plating of lithium onto a brass tip through a β‐LiPSand LLZTO solid electrolyte pellet. At a constant lithium plating rate of 250 µm(2 µA), the cell potential initially decreased while lithium network growth was simultaneously observed in β‐LiPS. After about 8 × 10µmof plated lithium (5 min), the voltage reaches approximately a steady‐state value, suggesting that the lithium network is no longer propagating forward through the sample. b) A galvanostatic current of 2 µA is applied for 2 h, following which the current is linearly ramped from 2 to 17 µA at a rate of 0.025 µA s, and the cell voltage is monitored. The dashed lines show the expected rate of the cell voltage if the ohmic resistance of the cell is unchanged throughout. The solid curves show the measured voltage versus time. In each case, a negative deviation from linearity indicates that the lithium metal network is advancing. This occurs at ≈9 and ≈15 µA for the polycrystalline LLZTO and β‐LiPS, respectively. c) Shallow galvanostatic cycling of a cell using polycrystalline LLZTO, cycled at ±2 µA. Time spent at each segment before the direction of current is changed is labeled. A decrease in cell voltage indicates advancement of the lithium network. Compared to uninterrupted galvanostatic deposition, cyclic plating and deplating of lithium metal causes much faster growth of the lithium metal network, with a short circuit occurring in the seventh cycle after net lithium electrodeposition of 5 µA h corresponding to 2.5 × 10µmof deposited lithium.

Thus, in a second experiment, we first allowed the cell voltage to stabilize via galvanostatic discharge, then ramped the current density in order to observe whether there is a critical current density at which the lithium metal network again propagates (as indicated by cell voltage). Figure 8b shows these results. Two cells of the same kind, again using β‐Li 3 PS 4 and polycrystalline LLZTO as solid electrolytes, were first subjected to galvanostatic conditions of 2 µA for 2 h. Based on the experiments in Figure 8a, this is well beyond the duration needed to establish steady state. Then, the current was increased at a constant rate of 0.025 µA s−1, while measuring the cell voltage. Based on the current and resistance at the start of the ramp, the ohmic resistance in the absence of any change in the sample should follow the straight dashed lines in Figure 8b. The data for the two cells (solid curves), however, show in each case a negative deviation from linearity at a critical value of current, which is ≈9 µA for the polycrystalline LPS sample and ≈15 µA for the polycrystalline LLZTO. We attribute this behavior to renewed penetration of the solid electrolyte occurring as the current density at the leading asperities of the lithium metal network reaches a critical value, although the specific local value of current density is not easily established given the extensive crack branching.

Finally, experiments were conducted to compare the effects of unidirectional deposition with the cyclic deposition that would be experienced in a rechargeable battery. Using the polycrystalline LLZTO, a constant current of 2 µA (≈25 mA cm−2) was again applied, but the direction of the current was frequently reversed, as shown in Figure 8c. The cell experienced a short‐circuit after only seven cycles (14 h), while under unidirectional deposition (Figure 8a), no short‐circuit occurs even after 24 h. The net deposition of lithium metal in the cycled case, based on the difference in time under positive and negative current, is 5 µAh, corresponding to 2.5 × 106 µm3 of Li metal, which is about an order of magnitude smaller compared to the unidirectional deposition case. Perhaps more critically, a comparison of the results in Figure 8b,c indicates that galvanostatic cycling has lowered the critical current density at which metal deposition advances through the solid electrolyte, since short‐circuiting occurs in Figure 8c under only 2 µA. Factors that could be responsible for this behavior include localized thermodynamic and kinetic effects,33, 34 stress distribution in the solid electrolyte, and subcritical mechanical damage ahead of the filament tip.