The major goal of this work was to analyze the grain size and orientation of electrodeposited alkali metal films in RFCs as well as to image pore formation after electrodissolution – in representative cross-sections – trying to advance the understanding of microstructure-property relations. However, before the results could be interpreted reliably, a protocol is required that did not alter the metal grain size, including artifacts created by heating during preparation and subsequent annealing. Furthermore, it was difficult to determine if and to what extent lithium and sodium anneal/recrystallize at room temperature, which also needed to be resolved before any conclusions could be drawn. The following section therefore answers these questions, while the subsequent section showcases the first results obtained from electrodeposited films within RFCs.
i. Establishing a Protocol for Alkali Metal Microstructure Analysis by EBSD
As outlined above, the first question is whether lithium or sodium foils exhibit grain growth during room temperature storage. While Singh and Fuchs et al. were able to prepare lithium foils with different average grain sizes,21,22 it is unclear whether further microstructural changes take place during extended storage or instead occur quickly after the initial foil preparation and then stop at a larger grain size, as is the case for electrodeposited silver and copper.24,38
Therefore, four different metal foils were freshly prepared for grain size analysis. Reference lithium and sodium foils (in the following: R-Li and R-Na) were prepared by mechanical removal of all visible passivation layers from a metal ingot, followed by a cut through the ingot using a microtome blade. To alter the grain size of the lithium and sodium metal, each metal was melted by heating to 250°C (Na) / 400°C (Li) on a hotplate in argon atmosphere, followed by rapid quenching in liquid nitrogen according to the procedure used by Singh and Fuchs et al.22 Foils prepared from these quenched ingots are labeled Q-Li and Q-Na. Top-view SEM images of the four foils obtained directly after preparation are shown in Fig. 1a-d. Figure 1e and f show top-view images of Q-Li and Q-Na after several days of room temperature storage, to reveal information about possible grain growth at room temperature. To avoid a potential passivation layer “freezing” the microstructure in place, a new foil was prepared from the same quenched ingot after storage.
First, lines and many triple junctions are found on the freshly prepared metal foils. These are assumed to be grain boundaries made visible by preferential degradation during transfer or morphological changes induced by grain orientation-dependent mechanical resistance to deformation during pressing.21,22 Interestingly, all sodium foils show less pronounced lines, possibly due to differences in mechanical properties, impurity level, and surface chemistry.
With apparent grain sizes of ~ 100 µm – 300 µm for R-Li and 10 µm – 50 µm for Q-Li we confirm that thermal processing strongly influences the lithium grain size as reported previously.21,22 Furthermore, the grain size did not significantly change during room temperature storage of Q-Li over four days, indicating that no visible grain growth occurred despite the high homologous temperature of lithium at room temperature. With less pronounced lines on the surface, sodium metal shows grain sizes close to the millimeter range for R-Na and about 200 µm – 600 µm for Q-Na, larger than that of lithium. Note at this point that the grains were sometimes larger than the SEM field of view, which makes statistical grain size analysis difficult due to the small number of grains in an image. However, when comparing the images, it is also striking that the grain size of sodium could be altered by thermal processing. Furthermore, Q-Na stored for two days did not show grain growth caused by room temperature annealing, which is confirmed below by EBSD imaging.
The apparent lack of grain growth despite the high homologous temperature may be explained considering two stages. Small grains are expected to grow rapidly driven by the reduction of interfacial grain boundary energies, as for example in the case of electrodeposited silver.25 However, with increasing grain size, the driving force to mitigate interfacial grain boundary energies decreases, slowing down further growth of grains. Although the self-diffusion of sodium at room temperature is relatively high (5.81 · 10–9 cm2 s–1) with an average diffusion length of 80 µm over 20 min, the grain growth is kinetically possible but seems to be extremely slow due to the small remaining driving force.39 According to the observed grain size, further grain growth during storage is unlikely within the examined period.
To further validate these initial SEM findings, we used EBSD to analyze freshly prepared surfaces of R-Li, Q-Li, R-Na and Q-Na foils, respectively, which is depicted in Fig. 2 a-d. Exemplary EBSPs for both metals with and without index overlay are depicted in Figures S1 and S2, proving that a crystalline, sufficiently passivation-free surface of lithium and sodium were obtained with our preparation protocol. The inverse pole figure (IPF) maps clearly confirm that the grains identified in Fig. 1 are bcc metal grains with different orientations. Due to the large grain size of R- and Q-Na, several spots were characterized by EBSD to improve the statistics, which are shown in Figures S3 and S4. Similar grain sizes are observed for R-Li and R-Na with EBSD characterization compared to the initial SEM analysis as well as for their quenched counterparts. This controlled grain size change fits well to what was observed previously for lithium21,22 and has not been previously shown for sodium. As the time between quenching the metal and cooling it down for analysis was minimized (i.e., 20 minutes for Q-Na), significant grain growth is not expected to have occurred. Moreover, no significant grain growth due to annealing of Q-Na at room temperature was observed after approximately two weeks, as demonstrated by the IPF in Figure S4. Therefore, grain growth of lithium at room temperature is also unlikely due to its lower homologous temperature compared to sodium.
However, in contrast to sodium, the lithium grains do not follow a typical Voronoi shape in either case, an example being the large beige grain in Fig. 2c. Usually, a more regular grain shape is expected, as seen for sodium (Fig. 2b and Fig. 2d). We assume that these shape distortions were induced by the high ductility of the metal during preparation, where a blade was run across the sample surface to remove any native passivation.
These results for thermally prepared lithium and sodium metal are fully consistent and demonstrate the efficacy of our protocol. However, electrodeposited metal films in RFCs are buried between a CC and an SE separator, meaning that the surface EBSD analysis protocol is not suitable for such films. In addition, no information on the microstructure within the bulk of the metal film is solely obtained by surface imaging, i.e. by top views. Here, cross-section preparation by cryogenic-FIB is required to enable the analysis of electrodeposited metal layers perpendicular to the interface. Therefore, a second step was necessary to validate that cryogenic-FIB preparation does not alter the grain size of prepared samples by local annealing. Cross-sections of samples shown in Fig. 2 a-d were prepared with cryogenic-FIB and analyzed using EBSD (Fig. 2 e-h). A second IPF map in the x-direction of the sample in Fig. 2g is presented in Figure S5, to showcase that the larger green grain on the right actually consists of multiple grains, coincidentally oriented similarly parallel to the y-direction. IPF maps of aged sodium cross-sections of Fig. 2g are depicted in Figure S6.
While the analysis area of a cross-section is much smaller than that of a surface view, strong differences between the thermally processed (quenched) metal foils and reference foils are observed both for lithium and sodium and are in line with the previous obtained results and trends. Furthermore, all cross-sectional maps predominantly show vertical grain boundaries. Likely, this occurs due to the high aspect ratio of the analyzed foil and the grain size being larger than the foil thickness or potentially during texturing when pressing ingots to a foil. This also explains why more curved grain boundaries are observed for thicker sodium foils in Fig. 2f and Fig. 2h with a lower aspect ratio.
We conclude that local grain growth during cryogenic-FIB preparation does not occur and the alkali metal microstructure is unaltered. These findings are supported by a lack of change in grain size and orientation after the second milling step of a cross-section, as demonstrated for sodium (Figure S7). Overall, this confirms that our cryogenic-FIB protocol is suitable to analyze the microstructure of alkali metal cross-sections, including electrodeposited films at CC|SE interfaces, without altering their microstructure.
The established analysis protocol could then also be used to clarify whether the observed white lines on the surface of alkali metal foils in Fig. 1 indeed indicate grain boundaries. Figure S8 shows cross-sections perpendicular to the course of the lines on a sodium foil. The respective EBSD maps confirm that a grain boundary is led into the bulk of the foil and is visible with SEM as the white surface line, supporting the interpretation given in Fig. 1. However, these surface lines do not unequivocally match grain boundaries in every case as highlighted in Figure S8, and thus, for proper grain size determination, microstructural analysis with EBSD is essential.
ii. Analyzing the Microstructure of Electrodeposited Alkali Metal Films by EBSD
After confirming that our protocol for preparation and characterization of the microstructure of alkali metals is reliable and no significant annealing occurs at room temperature, the following analysis is focused on the microstructure of alkali metal films electrodeposited within RFCs. Different RFCs were prepared, namely SS|LPSCl|Li, Cu|LLZO|Li and Al|NZSP|Na cells, representing today´s most investigated solid electrolytes paired with alkali metal anodes. The alkali metal was deposited within each cell on the respective CC. Figure 3a shows the protocol needed to obtain cross-sectional EBSD images thereof, with the three respective IPF maps presented in Fig. 3b. The corresponding voltage profiles during plating are shown in Fig. 3e.
Both lithium films were deposited with 100 µA cm–2 current density at 15 MPa at the SS|LPSCl interface and 5 MPa at the Cu|LLZO interface. Sodium was electrodeposited at the carbon-coated Al|NZSP interface with 200 µA cm–2 at 2 MPa. The current density, pressure and respective capacity deposited were specific to the material system and chosen based on experience to yield a homogeneously deposited film.11,40 Impedance spectra before and after electrodeposition of each cell are depicted in Figure S9, showing a characteristic change from a blocking impedance of the working electrode to the signature of a reversible alkali metal electrode, confirming nucleation and subsequent growth of a metal layer.11,40,41 All three voltage profiles display a characteristic nucleation overpotential in line with previous results, followed by a relatively stable plateau, during which layer growth occurs.11,40 The magnitude of the overvoltage is also similar to what was previously reported with 10 mV – 20 mV for lithium deposition and around 80 mV for sodium deposition.40,41 In the case of lithium plated at the SS|LPSCl interface, a sudden drop in voltage is observed, which indicates a short-circuit induced by dendrite formation. This is confirmed by the impedance data showing a significantly lowered resistance in Figure S9. However, the resulting film could still be analyzed.
After alkali metal deposition, each sample was subsequently cleaved and mounted on a tilted sample holder. The exposed cross-section was then polished using cryogenic-FIB and transferred into the SEM with EBSD using cryogenic transfer under vacuum. In the case of LPSCl, the deposited lithium was easily detachable, which is why only residual SE is attached to the lithium layer.
It is striking that the average grain size of each electrodeposited metal film is quite large, especially compared to other electrodeposited metals,25,28,38,42 limiting the total number of grains visible in each cross-section. The grain width for deposited lithium at the SS|LPSCl interface is around 20 µm – 100 µm and 10 µm – 100 µm for the deposition at the CC|LLZO interface, respectively. Sodium deposited at the carbon-coated Al|NZSP interface shows a grain width around 10 µm – 150 µm. A second cross-section of the sodium film is shown in Figure S10, confirming the given grain size. However, compared to the lithium and sodium metal foils analyzed in Fig. 1 and Fig. 2, the grain size is smaller, further indicating the absence of significant room temperature storage grain growth in electrodeposited films, although the impurity content is expected to be lower for electrodeposited films.
Another observation is that all grain boundaries are perpendicular to the CC|SE interface. This is especially important, as it marks a major difference between the Li/Na|SE interface of an electrodeposited metal compared to an as-built Li/Na|SE interface using alkali metal foil. In the latter case, the grain size is larger and grain boundaries are randomly oriented. This has strong implications for the subsequent discharge performance of the alkali metal films, as the grain size and orientation will likely affect the pore formation.21,22 Additionally, for the lithium film plated at the Cu|LLZO interface, two small grains are observed that do not span the whole thickness of the film. A larger magnification of these areas is visible in Figure S11. This is likely explained by different growth rates for different grains, with the neighboring grains growing faster and thus limiting the growth of the middle grain. We note that this orientation dependence of metal deposition is one of the next targets of more extended and detailed studies via EBSD.
Interestingly, similar predominant columnar grain growth has also been observed for electrodeposited nickel films using cross-sectional EBSD analysis, although the grains in that case did not fully span the whole thickness of the film. While the grain size is nearly constant for nickel films, the fraction of columnar grains increases with thicker deposited layers. A similar phenomenon is not observed here, as the columnar grains appear to grow along the whole layer thickness. On the other hand, similar electrodeposited silver films do not show this columnar grain growth.24 A reason for this could be that silver shows significant room temperature grain annealing within only a few hours, possibly changing the initial grain growth present during electrodeposition. This reasoning also strengthens our conclusion that plated alkali metal does not show grain growth during room temperature storage, as we would be able to observe randomly oriented grain boundaries herein, as in the case of silver and copper. However, it is possible that microstructural changes occur during the deposition process, as discussed later.
It is not clear what sample property or deposition parameter primarily governs the deposited metal microstructure, such as grain size and microstructure of SE and CC. To check for a relationship between the substrate microstructure with the deposited microstructure, IPF maps of lithium plated on a copper CC with both layers being indexed are shown in Figure S12a. Although electrodeposited lithium preferentially deposits on the Cu(111) crystal facet in liquid electrolyte, we found no clear relation between the grain size and orientation of the copper and lithium at the Cu|SE interface.43 This uncorrelated growth may be due to surface passivation or solid electrolyte interphase (SEI) formation masking the underlying grain structure. A partial masking of preferred deposition sites due to passivation was also reported for lithium growth on copper substrates when different liquid electrolyte with different reactivity were used.44 In general, it is unlikely for lithium to undergo epitaxial growth on copper due to a mismatch-induced lattice strain (aCu = 3.6 Å vs. aLi = 3.5 Å).45 The minor effect of the CC microstructure on the resulting grain structure becomes even clearer when lithium is electrodeposited on a Q-Li reservoir foil, as shown by the IPF map in Figure S12b. In this case, neither the grain size nor the grain orientation of the electrodeposited grains matches that of the Q-Li reservoir. Furthermore, the size of the solid electrolyte grains (see Figure S13) does not appear to have a direct correlation with the microstructure of either electrodeposited alkali metal, as the size difference is quite large. It is therefore likely that the deposit microstructure is dominated by the nucleation process and grain growth itself.
As the process of grain growth and evolution during electrodeposition is yet not fully understood, in situ EBSD analysis was performed. Here, cross-sections prepared via FIB of Cu|Li|LLZO|Li and Cu|Na|NZSP|Na were charged and discharged inside the SEM, respectively and stopped for intermittent EBSD analysis of the microstructure. In the case of Cu|Li|LLZO, around 2 µm of lithium were deposited prior to cross-section preparation to fix the CC on the SE. Further, microelectrodes were prepared via FIB at the cross-section to ensure that changes occur in the field of view. The described setup is depicted in Fig. 4a. An overview SEM image is shown in Figure S14a prior to deposition. Front-view images of the pristine cross-section are further provided in Figure S14b and Figure S14c. The left section of Fig. 4b shows the voltage profile during deposition of lithium at ~ 500 µA cm–2. The initial higher voltage results from a high starting current density (1000 µA cm–2) which was then lowered to ensure homogeneous layer-growth. The observed voltage profile is mostly flat, hinting at dendrite free lithium deposition of 10 mAh cm–2, corresponding to a metal layer of ~ 50 µm thickness. The right side of Fig. 4b shows the voltage profile obtained for stripping of the sodium electrode with a typical increase of voltage indicative of pore formation. Both measurements were paused for subsequent EBSD analysis as visible in the potential profiles, which are shown in Fig. 4c and d, respectively.
Figure 4c-(1) shows an SEM image of the pristine cross-section of Cu|Li|LLZO. As the lithium is not freshly deposited, no sufficient EBSPs could be acquired to generate an IPF map. However, upon depositing around 10 µm – 15 µm of new lithium, the IPF map depicted in Fig. 4c-(2) was obtained. Herein, several grains around 10 µm – 30 µm in width can be observed. Surprisingly, after another ~ 5 µm of lithium deposition, the next IPF map shown in Fig. 4c-(3) shows fewer grains with a larger width. The two small blue grains close to a < 111 > orientation from the previous map are apparently fused together with larger neighboring grains close to a < 101 > orientation (green). After another deposition step, IPF map Fig. 4c-(4) shows even wider grains, with another blue grain from the previous map being fused to a neighboring green grain.
After another long deposition step resulting in around 20 µm of new lithium, multiple grains have fused together forming a large grain with > 100 µm width, again close to a < 101 > orientation as visible in Fig. 4c-(5). This post deposition map was acquired after two weeks of room temperature storage and additional FIB polishing, with top- and front-view images depicted in Figure S14g and Figure S14h. Additionally, the map in Figure S15a acquired directly after stopping the deposition process shows the same microstructure. In total, roughly 40 µm – 50 µm metal were plated, which fits to the transferred charge. Small deviations were visible in the map directly acquired after plating, which can be seen in Figure S15a. This is due to the lithium being partly squeezed outside of the prepared electrode area (Figure S14d). The uneven film morphology at the interface also explains the low indexing rate of the shown IPF maps, as the sample is not perfectly tilted at 70° at every spot. Interestingly, our analysis also confirms the absence of grain growth during storage. Additionally, Figure S15b, Figure S15c depict the maps of Fig. 4c-(4) and Fig. 4c-(5) with the IPF in the x-direction, showing that the large green grains actually consist of more grains, coincidentally oriented in the y-direction. These grain boundaries are therefore indicated with black dashed lines in every map in the y-direction. Furthermore, Figure S15d depicts the cross-section shown in Fig. 4c-(5) after storage and additional FIB polishing, further confirming the lack of grain growth during storage. We consider these results gained from the in situ analysis as highly important, i.e., microstructural changes are now shown to occur during the deposition process but to stop once the deposition is finished. This is explored for sodium metal deposition in the following.
An overview of the electrodeposition of sodium on a pre-existing sodium reservoir is depicted in Figure S16. In contrast to lithium, the experiment was conducted using a Q-Na|NZSP|Na cell with an initial electrode thickness of ~ 50 µm. A smooth and flat voltage plateau is observed during deposition of roughly 1 mAh cm− 2. Assuming uniform deposition across the electrode, the deposited capacity corresponds to a sodium layer thickness of approximately 9 µm, which fits to the observed thickness increase at the cross-section shown in Figure S16b-d. Prior to deposition, three main grains are observed at the cross-section, while a few small grains are located directly at the interface (Figure S16e). Similar to the observation for lithium in Fig. 4c, the large grain (green) with an orientation close to < 101 > grows vertically during deposition and consumes the neighboring grain with an orientation close to < 111>. Moreover, the thickness of the smaller grains on the left side of the cross-section directly at the interfaces slightly increases during deposition. The identification of newly formed grains is challenging, as the sodium partly grows out of the image plane leading to shadowing of the interface. However, a new grain appears on the right side of the cross-section as shown by a red arrow in Figure S16g. After finishing the deposition experiment, the cross-section was polished via FIB uncovering a second formed grain indicated by red arrows in Figures S16f and S16h. It should be noted that after polishing, the new cross-section is approximately shifted 10 µm compared to the previous plane, making it difficult to establish an unambiguous correlation. However, upon checking the EBSPs for these grains in Figures S16i and Figure S16j, it can be concluded that this is the same grain. Similar to the deposition of lithium, the grain width of sodium changes during deposition and grains in the < 101 > orientation show preferential growth. Although the microstructure of the discussed cross-section has not been characterized after a defined storage time, it is unlikely that the grain width has changed during storage, as shown by the example in Figure S10. Thus, the microstructural evolution of sodium likely follows that of lithium.
To further study the dependence of pore formation on the metal anode microstructure, a sodium anode was also stripped. A typical voltage profile is achieved with an initial small increase evolving into a step increase, as shown in Fig. 4b. This signature voltage profile clearly indicates pore formation.6,37,46,47 The corresponding IPF maps and Forward-Scatter Electron (FSE) images are shown in Fig. 4d. Starting from an optimal interfacial contact between sodium and NZSP (Fig. 4d-(A)), a dark region close to the interface emerges (Fig. 4d-(B)) after the first stripping interval. We attribute this to pores at the interface. The scattering of primary electrons decreases significantly at a pore, resulting in a loss of contrast in the FSE images. Additionally, the generation of secondary electrons decreases, leading to a fading of brightness in the images (Figure S17a and Figure S17b).
Interestingly, the pores formed mainly within the green grain while the vertical grain boundaries on both sides remain intact (Fig. 4d-(C)). As stripping progresses, another pore nucleates within the large blue grain starting from a grain boundary, supporting the previous description. Moreover, the pores within the green grain further grow into the bulk of it (Fig. 4d-(D)). It should be noted that the electron images and EBSD maps were recorded at an angle of 70° to the cross-section. Although the measuring geometry was corrected in all data, the size and shape of the pores may not have been fully captured. The pores appear to be much larger when compared to the angle-corrected imaged geometry shown in Figure S17c. Finally, the cross-section was polished again to unambiguously confirm that pores are formed by anodic dissolution of sodium (Figure S17d).
Clearly, this observation of preferential pore formation in grains – not in grain boundaries – is at first glance counter-intuitive, but is a direct proof for the fast diffusion of vacancies along grain boundaries.48 Pore nuclei at in grain boundaries can be closed faster by diffusion within the grain boundary than within the bulk. The microstructure, especially the grain size and grain boundaries, strongly influences the pore formation and thus the physical contact at the interface, which drastically affects the performance of the metal anode.
From our perspective, these observations and analyses mark an important milestone on the way to a better understanding of alkali metal electrodes on solid electrolytes. The microstructure and its evolution of electrodeposited alkali metal films was completely unknown due to its elusive nature and difficulty to prepare suitable cross-sections.
iii. Grain Evolution During Cycling
In the following, a mechanism behind the microstructural evolution during stripping and plating of alkali metal is proposed and schematically depicted in Fig. 5.
Figure 5a shows a schematic of a voltage profile during electrochemical deposition of alkali metals on an inert current collector with subsequent storage and anodic dissolution. The metal nucleation at t0 evolves into the early stages of grain formation and growth (Fig. 5b), which is followed by mostly vertical grain growth at t1 with different rates (Fig. 5c and Fig. 5d). If the growth rates of neighboring grains are dissimilar, it also seems to be possible that the growth of slower growing grains is limited compared to the faster growing neighboring grains as shown in Fig. 5d and Fig. 5e or Figure S11, leading to truncated V-shaped grains.
During deposition between t2 and t3, a process similar to grain ripening occurs with larger grains merging with adjacent grains (Fig. 5e), thus widening the average lateral grain dimensions. The exact reason and magnitude of this ripening is unclear as of now and may be related to abnormal grain growth or “secondary recrystallization” which is dependent on the interface and grain boundary energies.49 Additionally, this may also be a reason for the occurrence of truncated V-shaped grains. To elucidate this, further studies will focus on more detailed investigations of the current density and thickness dependence of the film microstructure.
After a prolonged room temperature storage at ts, the microstructure remains unchanged (Fig. 5f). If this microstructure is then subjected to anodic dissolution until pore formation commences, the pores predominantly form within grains at the interface and not at grain boundaries (Fig. 5g). Although grain boundaries are 2D defects and typically have a higher free energy compared to the bulk crystal, fast transport of metal along grain boundaries might be the reason for the observed phenomenon. A higher vacancy flux from the interface into the electrode at grain boundaries suppresses the nucleation and growth of pores where the grain boundaries within the metal meet the interface.
As the pores formed during anodic dissolution seem to form preferentially within grains and not at grain boundaries, we conclude that it is possible to tune the electrochemical properties, such as available stripping capacity,37,50 of the deposited metal layer by controlling the nucleation and growth during deposition and therefore the microstructure. Following this reasoning, achieving a microstructure with small grains would be desired.48 The nucleation density mainly depends on applied current density and temperature as well as the surface properties of the SE.12,13,41 Additionally, our in situ EBSD results further show that not only the grain nucleation is important to control the resulting microstructure, but also the growth process. Potentially, the applied current density will also influence the grain ripening during electrodeposition. Similarly, applying stack pressure may be a suitable tool to guide the lateral expansion of the growing grains.11
In addition, based on our evidence, controlling the microstructure of the CC and SE may not be a successful path to influence the deposited metal microstructure. Lithium plated on a lithium foil does neither match the microstructure of the solid electrolyte nor the foil but plates with a similar, columnar shape as in the case of plating on an inert CC in a reservoir-free setup (Fig. 5h). The microstructure of the deposited metal is still significantly different and has lower overall grain sizes compared to those of commonly used metal foils.
The outcomes and results of this work highly motivate further research on the influence of deposition parameters on the resulting metal film microstructure and thereby subsequent electrochemical performance of the electrode. Both ex situ and in situ experiments will help to develop more advanced plating and stripping strategies.