Benchmarking and Critical Design Considerations of Zero‐Excess Li‐Metal Batteries

Zero‐excess Li metal batteries (ZELMBs), in which a Li‐anode is formed in situ during charging, have received much attention in recent years. ZELMBs bear great potential to increase energy density and facilitate battery production, thereby reducing cost as well as material and energy consumption. Practical application of ZELMBs has so far been limited by challenges related to the non‐uniform deposition behavior of Li, leading to inadequate performance and safety concerns. To address these issues, promising approaches have been developed in recent years, including modifications of the current collector, electrolyte, and cycling protocols. While these approaches improve the long‐term stability of ZELMB, they also reduce the energy density by introducing inactive materials into the cell. Herein, critical design criteria for the various optimization approaches in ZELMB research are established. Nominal volumetric and gravimetric energy densities are determined based on the degree of modification. Thresholds are determined for each of the strategies at which the energy density gain of ZELMB vanishes compared to other cell configurations. These findings are compared to literature results to provide guidance for the further development of ZELMB.

The ORCID identification number(s) for the author(s) of this article can be found under https://doi.org/10.1002/adfm.202214891.
challenge. Li offers a low redox potential (−3.04 V vs SHE) and a high theoretical capacity (3861 mAh g −1 ), resulting in high energy density. So called zero-excess Li-metal batteries (ZELMBs), also called anode-free or anode-less batteries, represent a special type of LMB, in which the Li-metal anode is formed in situ during charging of the cell. A decent number of review articles has been published recently, summarizing the current state of the art of ZELMB with liquid electrolytes [1][2][3][4][5] and first studies employing solid-state electrolytes. [6,7] Besides major advantages regarding the ease of manufacturing as well as material and cost savings, ZELMBs offer the highest gravimetric (GED) and volumetric (VED) energy density among all cell concepts.
The obvious advantages of ZELMB are opposed by several obstacles. ZELMBs have the same problems as Li-metal batteries, with the additional challenge that the lack of a Li reservoir cannot compensate for losses during cycling. At low electrode potentials, a solid electrolyte interphase (SEI) composed of Li and electrolyte decomposition products is formed. [8] The more Li is lost during SEI formation, the lower the initial coulombic efficiency (CE), which directly translates into a reduction in cell capacity. The electron-blocking SEI should protect against further electrolyte decomposition. However, the SEI can break down during Li-plating and stripping. The associated healing of the SEI leads to further Li consumption, lowering CE. [9,10] These processes are also linked to the formation of high surface area Li (HSAL), e.g. dendrites, which bears the risk of short circuits and further Li loss because HSAL is prone to detaching during dissolution, ending as so called "dead Li". [11][12][13][14] This results in further decrease of the CE, a comparably short cycle life and pronounced safety concerns.
To tackle the challenges of ZELMB, several promising approaches emerged in recent years, including modification of the current collector (CC), the electrolyte, cycling protocols, and external measures. [3,4,15] While these approaches improve the long-term stability of a ZELMB, they can also act detrimental in terms of energy density by introducing inactive materials into the cell. At a certain "degree of modification", the corresponding reduction in energy density can become so high, that the theoretical advantages of ZELMB over other battery configurations start to vanish. Consequently, the design parameters for

Introduction
Li-ion batteries (LIBs) currently dominate the market of traction batteries for electric vehicles. In this context, especially high-energy battery cells, offering energy densities beyond 350 Wh kg −1 or 1000 Wh L −1 , are in great demand to extend the driving range, which is often reported to be the main reason for hesitant user acceptance. Li-metal batteries (LMBs) are currently one of the most promising approaches to tackle this the different types of modifications need to be carefully chosen to maintain the advantages of ZELMB over established cell concepts. The aim of the present study is to identify critical dimensions for the different optimization approaches in ZELMB research. Based on theoretical considerations, the dependence of VED and GED on the degree of modification is modelled. This includes the volume and weight of the CC or host structures, excess electrolyte volumes, and inactive Li reservoirs. For each of the strategies, thresholds are identified at which the energy density gain of the ZELMB configuration vanishes with respect to conventional LIB. These findings are compared to literature results to provide guidelines for further development strategies that ensure the preservation of increased energy density of ZELMB. Figure 1 shows schematic drawings of different cell configurations with liquid and solid electrolytes and their nominal VED and GED. An NCM 811-based cathode with an areal capacity of 4 mAh cm −2 was selected in each case for the sake of comparability. The commercialization of LIBs with blended Si/Gr anode materials is progressing rapidly. Therefore, we consider an anode (N/P = 1.2) consisting of 80 wt.% graphite and 20 wt.% Si to be representative for the next generation of LIBs. For the Li-metal batteries with liquid (LMB-l) and solid (LMB-s) electrolyte, an excess amount of Li equal to 120% of the cathode capacity is considered, which corresponds to a Li-layer thickness of 23.5 µm. For ZELMB with liquid (ZELMB-l) and solid (ZELMB-s) electrolytes, a bare Cu-CC is considered at the anode side. For ZELMB-s, a sulfide-based Li-ion conductor is considered as the solid electrolyte separator (SES) and in the composite cathode. Sulfide-based solid electrolytes are currently believed to be the most promising class of materials for all solid-state batteries, and several examples of pouch-level cells with promising practical energy densities exist. [16] The entirety of the design and material parameters used for the calculation of nominal energy densities is provided in Supporting Information. To account for the volume expansion of the silicon (LIB) and the volume of the deposited lithium (LMB, ZELMB) during charging, the VED is determined for the charged state (SOC = 100%). The LIB considered here exhibits a VED (GED) of 1058 Wh L −1 (375 Wh kg −1 ). When replacing the composite anode with a 23.4 µm thick lithium metal foil (LMB-l), GED increases to 462 Wh kg −1 . The corresponding increase in VED (1058 → 1100 Wh L −1 ) is less pronounced. As Li-metal has a very low density of 0.534 g cm −3 , the Li excess stored and the amount deposited at the anode add a lot of volume to the cell in the charged state. When using sulfide solid electrolyte (LMB-s), the GED is reduced compared to LMB-l, due to the higher density of the solid electrolyte, but is still significantly higher than for the LIB. As expected, in ZELMB configuration, the nominal energy density is highest among all cell configurations. The GED and VED are determined to 480 Wh kg −1 and 1324 Wh L −1 for ZELMB-l and 465 Wh kg −1 and 1236 Wh L −1 in the case of ZELMB-s.

Initial Considerations
The above considerations indicate that ZELMB have higher GED and VED, when compared to LIB and LMB. However, as mentioned in the introduction, modifications of the CC, electrolyte and cycling protocols are necessary to achieve sufficient cycling stability. The GED and VED depicted in Figure 1 do not include any modifications to circumvent HSAL growth or reduce the generation of dead Li. Therefore, the corresponding Figure 1. Schematic drawings and nominal gravimetric (GED) and volumetric (VED) energy density of a Li-ion battery (LIB) assuming 4 mAh cm −2 areal capacity and comparison to Li-metal batteries with liquid (LMB-l) and solid (LMB-s) electrolyte as well as zero-excess Li-metal batteries with liquid (ZELMB-l) and solid (ZELMB-s) electrolyte. The color coded bars indicate the mass and volume shares of the cells' components. Please note that for the sake of clarity, the schematic drawings are not to scale. gains in GED and VED over LIB and LMB represent a bestcase scenario. The introduction of inactive materials, e.g. an additional protective or functional layer on the CC, will lower GED and VED to a certain extent, depending on the type and design of modification. To evaluate the impact of different optimization approaches on GED (VED), the LIB configuration (cf. Figure 1) serves as comparison threshold in the following. The corresponding gain in GED (VED) of ZELMB without any modifications over the reference LIB is 28 % (25 %) for using liquid electrolyte and 24 % (17 %) for using solid electrolyte. This gain in GED (VED) can be considered the design window for reasonable modifications. ZELMB modifications that lower GED and VED below the comparison threshold (GED or VED gain < 0%) cannot be considered as suitable improvements. The threshold not only serves as a demonstration which type of modification is suitable for implementation in ZELMB but also points out to which degree (e.g., thickness) an optimization approach can be implemented, before GED or VED advantages of start to vanish.
In the following, the nominal GED and VED of ZELMB with different degrees of modifications are determined. The results are presented in terms of percentage increase in GED and VED with respect to the reference LIB. Corresponding numerical values in terms of Wh kg −1 and Wh L −1 are provided in Supporting Information. Different cell types (pouch, cylindrical prismatic, bipolar configuration, …) have different cell-utilization-factors that furthermore depend on the number of layers (stacks), design of tabs, etc. To keep our statements as general as possible, we consider the single cell element level, [17] including current collectors, electrode layers, electrolyte, and separator. We would like to stress that our analysis only considers the nominal energy density and neglects any effects of the modifications on other important properties, such as safety, rate capability, and cycling stability. For the sake of simplicity and comparability, we also neglect any capacity reduction caused by formation losses or side reactions during continuous cycling. While these limitations do not allow a holistic evaluation of the different concepts and optimization approaches, comparability in terms of nominal energy density is significantly increased.
Furthermore, we would like to emphasize that the following analysis of the literature data is in no way to be understood as negative criticism. Often, the studies do not focus on the optimization of the energy density, but, for example, on fundamental aspects of Li deposition. Nevertheless, we would like to use these data to illustrate how effective the various optimization strategies are using practical examples.

Host Structures
The design of the CC plays a crucial role in affecting the Lideposition behavior and thus ZELMB cycling performance. Microporous structures are designed to confine Li deposits inside the voids (cf. Figure 2). Such host structures with a high surface area can help to decrease the local current density resulting in a smooth Li morphology with high reversibility of plating/stripping. Host structures therefore not only have the potential to accommodate volume changes at the cell level during cycling but also help to increase the cycle life of ZELMB.
When designing host structures, attention must be paid to the geometric dimensions with respect to the amount of Li to be plated into the void volume. The volume of Li to be accommodated by the host can be calculated using Faradays law and the capacity of the cathode. Figure 3a shows design thresholds for host structures capable to accommodate all Li when considering a cathode with a capacity of 4 mAh cm −2 . As expected, the smaller the thickness of the host structure, the larger its porosity needs to be. For example, when using a host structure with a thickness of 20 µm the minimum porosity needs to be ≈98 %, whereas for 50 µm thickness a porosity of 39 % is sufficient to accommodate 4 mAh cm −2 of metallic Li. Falling short of these design requirements inevitably leads to "over-plating", that is, the deposition of Li on top of and not into the host structure, as schematically illustrated in the inset in Figure 3a. It should be noted that the provision of sufficient void volume is a necessary prerequisite, but not a guarantee, for the avoidance of over-plating if the nucleation and growth of Li on the top of  the host is preferred. [18] Figure 3b displays host porosity ranges as a function of the thickness and intended areal capacity. As expected, the higher the areal capacity, the thicker and more porous the hosts have to be to avoid over-plating. For example, when using a host structure with a thickness of 15 µm and a porosity of 65% the areal capacity of the cathode is limited to 2 mAh cm −2 to ensure that all Li plated during charging can be accommodated by the host.
To achieve high energy density at the full-cell level, cathodes with areal capacity > 3 mAh cm −2 are required. From Figure 3b it is obvious that such cathodes need to be paired with relatively thick host structures of high porosity. To investigate how this affects the energy density at the full-cell level, theoretical considerations concerning GED and VED as a function of thickness and porosity of the host structure were carried out to determine reasonable design parameters. Figure 4 shows the gain in VED (Figure 4a (Table S9 and Figure S11, Supporting Information). The shadowed areas indicate host designs not capable to accommodate all Li plated during charging (over-plating). A significant VED gain over LIB can only be achieved in a narrow design window (thickness and porosity). If over-plating is to be avoided, the design restrictions become even more severe. In the case of ZELMB-l, the porosity and thickness of the host need to be > 67% and < 29 µm, respectively, to achieve a gain in VED over the reference LIB while avoiding over-plating. For ZELMB-s, a gain in VED over the reference LIB is obtained for host structures with porosity > 49 % and thickness < 40 µm. While the fabrication of host structures with such parameters seems feasible, the design requirements to maintain higher energy density become more challenging. For example, to achieve 5 % gain in VED over LIB with liquid (solid) electrolytes, the host must have a thickness of 22 (33) µm and a porosity of 89 (60) %. The behavior in the over-plating region differs between ZELMB-l and ZELMB-s. In the case of ZELMB-l, the deposited Li displaces an equivalent volume of liquid electrolyte. This leads to an increase in cell volume during Li-plating into the host as well as during overplating, decreasing the VED in a similar manner. In the case of ZELMB-s, the host structure is initially considered empty. [19] Accordingly, during Li deposition into the host, the cell volume and the VED remain constant until over-plating occurs, where the deposited Li displaces an equivalent volume of solid electrolyte. Accordingly, during over-plating, the cell volume increases and the VED decreases.
Considering GED for ZELMB-l (Figure 4c), the porosity of the host needs to be > 73 % and the thickness between 20 and 66 µm depending on porosity, respectively, to achieve a gain over the reference LIB while avoiding over-plating. For ZELMBs, a gain in GED over the reference LIB is obtained for LLZO host structures with porosity > 56 %. Similar to the VED considerations, the fabrication of host structures that lead to similar GED as the reference LIB seems to be feasible. However, the design requirements to obtain a significant GED improvement might become challenging. For example, to achieve 15% gain in GED over LIB with liquid (solid) electrolytes, the host must have a thickness of 21 (25) µm and a porosity of 94 (80)%. The difference between the design freedom dependencies of ZELMB with liquid and solid electrolytes (i.e., iso-GED lines in Figure 4c,d) is again related to the fact that the host structure is initially considered empty in the case of ZELMB-s. Accordingly, increasing the thickness of a highly porous host hardly affects the mass of the cell, leading to little impact on the GED. In contrast, in the case of ZELMB-l, increasing the thickness of the host structure increases the amount of electrolyte stored in the pores in discharged state. Accordingly, the mass of the cell increases with thickness, leading to significant impact on GED in the liquid electrolyte system.
The impact of the host material's density on the design freedom is illustrated in Figure 4e,f for ZELMB-l and ZELMB-s. We only consider GED here, as VED is independent of the mass density. As expected, the higher the density of the host material, the more restricted the design freedom that allows significant improvement in GED with respect to the reference LIB. While this effect is less pronounced in the case of liquid electrolytes  Li, e.g., NaSICON-type LATP [Li 1+x Al x Ti 2−x (PO 4 ) 3 ]. Sulfide solid electrolytes have the lowest density (1.64 g cm −3 ) with high ionic conductivity, and significant improvements in GED could be achieved with easy-to-implement host structures (e.g., 29 % porosity, 67 µm thickness). However, sulfides are also not stable against metallic lithium and form a passivation-layer, making the direct application as host material challenging.
The results in Figure 4 show that the design freedom of host structures that allow preserving the VED and GED advantages of ZELMB is quite limited. Accordingly, the choice of materials and the architecture of host structures must be considered very carefully.
Previous attempts to introduce host structures for ZELMB are summarized in various review articles. [1,4,5,[20][21][22][23] For example, Umh et al. developed a hierarchical host structure with macropores and a high surface area by electrodeposition of Cu and found significantly improved cycling stability when compared to planar Cu-CC. [24] Based on the SEM images provided, we estimate that the thickness and porosity are ≈ > 100 µm and < 65%, which leads to a GED and VED reduction of −43% and −33% with respect to the reference LIB. As shown in Figure 4e, the design freedom increases when using low-density host materials. Kwon et al. showed excellent improvement in cycling stability by using a modified carbon paper (≈100 µm thickness, ≈83 % porosity) as a host for Li-deposition. [25] When considering these design parameters and the low density of carbon (2.26 g cm −3 ), our calculations suggest a GED and VED significant reduction of −1 % and −33% with respect to the reference LIB. Ikhe et al. used an even thicker (356 µm) but less porous (≈34 %) carbon-based host material and propagated "anode-free operation" for over 3000 cycles, which might be misleading, considering the large Li-storage capacity of the carbon-based host material itself. Anyway, the GED and VED are dramatically reduced by -60% and -70%, respectively, compared to the reference LIB. These examples and our theoretical considerations in Figure 4 shows that even for low-density materials, designing host structures for ZELMB-l is extremely challenging in terms of achieving high energy densities.
In the case of solid electrolytes, the application of host structures that accommodate Li-deposits is still in its infancy. Host structures are described for garnet-based ceramic electrolytes. LLZO is stable against reduction up to 50 mV (vs Li/Li + ) and regarded as fully inert towards Li. [26][27][28][29] Still LLZO separators are prone to dendrite formation and are thus limited by a low critical current density. Using a host structure, the active surface can be increased by a factor 40 compared to planar LLZO surfaces and symmetrical cells Li | LLZO | Li cells can be cycled up to 10 mA cm −2 . [30] Using sacrificial fillers that are removed during sintering [30][31][32] or freeze casting [33] structures with porosity up to 70 % can be prepared, which meets the design requirements to achieve a VED and GED gain over the reference LIB (cf. Figure 4b,d). LLZO based host architectures are described as tri-layer (porous-dense-porous) [30] or bilayer (porous-dense) [31][32][33][34][35] structures where either the anode, the cathode, or both electrodes are hosted. Kravchyk et al. estimated the GED and VED of solid-state batteries based on porous LLZO host structures for multi-layer pouch cells. [19] They showed that trilayer-structures reach VED of 975 Wh L −1 but GED below 275 Wh kg −1 , while bi-layer structures achieve slightly higher GED of up to 285 Wh kg −1 . Bilayer structures with 20 µm dense plus 50 µm porous (70 %) layers have been shown to achieve 329 Wh kg −1 and 972 Wh L −1 . [34,35] We want to emphasize that the above theoretical considerations do not address the initial CE and the resulting loss in capacity. The increase in specific surface area due to the use of host structures also results in greater Li losses, e.g. due to SEI formation, which has a direct impact on the initial CE and thus the capacity loss during formation. In this respect, in addition to the careful design of thickness and porosity, host structures may need to be provided with, e.g., artificial SEIs [36] to prevent a drastic reduction of the nominal energy density.

Lithiophilic and Protective Coatings
Applying lithiophilic and protective coatings on CC has been shown to be a promising approach to increase both, initial CE and cycling stability of ZELMB. The application of lithiophilic or protection layers will add to the volume and mass of the cell. Accordingly, the structure and composition of such layers must be carefully selected to avoid significant reduction of VED and GED, which would negate the advantages of ZELMB concepts.

Lithiophilic Coatings
Due to the relatively lithiophobic behavior of bare Cu foil, typically used as the anode CC, HSAL growth is accelerated leading to a low cycling stability. Therefore, a common approach is the application of lithiophilic layers on the CC to improve the wettability with Li, reduce the nucleation barrier and to improve its ability to anchor Li deposits. Typically applied lithiophilic layers are made of materials that undergo alloying or conversion reactions with Li. As displayed schematically in Figure 5, these layers undergo volume expansion during lithiation. After the lithiophilic layers are saturated with Li, the actual Li deposition begins. Obviously, the type of material and the design parameters of the lithiophilic layer will affect the nominal VED and GED of the ZELMB. Figure 6 shows the gain in VED (Figure 6a,b) and GED (Figure 6c,d) of ZELMB over LIB (for values in terms of Wh kg −1 and Wh L −1 see Figure S2, Supporting Information) for typically used lithiophilic materials in liquid and solid electrolyte systems depending on the lithiophilic layer thickness in the initial state (SOC = 0). The maximum thickness given corresponds to a pure conversion/alloying anode, i.e., all Li is stored by the alloying/conversion reaction and no Li-plating occurs. Considering Zn, Ag, and Sn as lithiophilic materials, the VED gain over LIB (Figure 6a,b) slightly decreases with increasing film thickness. The intersection at thickness = 0 µm equals the energy density of a ZELMB without modification. Interestingly, the VED gain for CuO and SnO conversion type lithiophilic layers rises with the layer thickness. This is because the fully lithiated CuO and SnO layers are thinner compared to a Li layer of similar areal capacity (see Figure S3, Supporting Information). This effect occurs above a certain ratio between specific capacity and density of the lithiophilic material in the fully lithiated state compared to Li metal (see Supporting Information for details). The GED gain over LIB decreases for each type of lithiophilic material when increasing the layer thickness. Furthermore, the reduction in GED is more severe than in the case of VED, as the difference in gravimetric capacity between Limetal and alloying/conversion materials is larger than for the volumetric capacity. When compared to the design freedom of host materials, the VED and GED threshold of the LIB is not met in any case when using lithiophilic layers. However, care must be taken, when considering thick lithiophilic layers, since the volume expansion of alloying/conversion materials is known to lead to fast cell degradation. Accordingly, the application of thin lithiophilic layers appears most suitable, considering VED and GED gain, as well as cycling stability.
Among several chemistries, Ag nanoparticle-based lithiophilic layers are a very popular approach and have been demonstrated to improve nucleation and adhesion of Li. [37] Shin et. al. have proposed an electroless deposition method of Ag nanoparticles on Cu substrate leading to an improved cycling stability in ZELMB full cells compared to bare Cu CC. [38] Considering the extremely low Ag loading of 10 µg cm −2 , our model predicts an almost unaffected gain in VED (GED) of +25 % (+28 %). Zhang et. al. proposed an approach for structured CC with lithiophilic CuO nanosheets, showing prolonged cycling stability in half cells compared to bare Cu-CC. [39] This modification (< 1 µm) could lead to a gain in VED (GED) by +26 % (+27 %). A modification of Cu-CC by Sn was shown by Zhang et al., leading to improved adhesion of Li and increased cycling stability. [40] By applying such thin Sn layers (< 100 nm) the VED (GED) superiority of ZELMB (+25 % (+28 %)) over LIB is almost completely maintained. Our theoretical considerations and the examples www.afm-journal.de www.advancedsciencenews.com from the literature show that the incorporation of lithiophilic materials is generally associated with negligible losses in energy density. On the one hand, the materials themselves have a relatively high volumetric and gravimetric capacity. On the other hand, they can be applied quite easily as nanoparticles or very thin layers, so that the modification accounts for only a very small proportion of the mass and volume of the cell.

Protective Coatings
Protective coatings, also called artificial SEI, have shown their ability to prevent Li loss from continuous SEI formation by electrolyte decomposition while supporting reversible Li plating and stripping. [41] During charging, Li-ions pass the protective layer and are being deposited as metallic Li at the interface between the protective layer and the CC (Figure 7). Protective coatings need to obtain certain properties to ensure sufficient cycling stability of ZELMB. To suppress electrolyte reduction an ideal protective coating is electronically insulating whilst providing high Li-ion conductivity. High mechanical stability and elasticity is needed to suppress HSAL growth and inhibit breaking due to volume changes during cycling. As a protective coating does not contribute to the capacity of the anode, its mass and volume decrease the GED and VED depending on the materials properties and design parameters.   Figure S4, Supporting Information). As expected, the VED of ZELMB is independent from the material density of the protection layer but sensitive to its thickness. In the case of ZELMB-l (Figure 8a), the VED passes the LIB threshold for protection layers thicker than ≈29 µm. To achieve a significant VED gain (> 20%) compared to LIB, the protection layer needs to be < 5.0 µm. In the case of ZELMBs (Figure 8b), the freedom of design to maintain VED superiority is more restricted. The VED passes the LIB threshold for protection layers thicker than ≈20 µm. To achieve a significant VED gain (> 15%) compared to LIB, the protection layer needs to be < 1.5 µm. The design freedom to maintain GED superiority of ZELMB naturally depends on the density of the material used (Figure 8c,d). The higher the protection layer's density, the larger the impact of its thickness on the reduction in GED. For example, in the case of ZELMB-l (Figure 8c), the GED passes the LIB threshold for protection layers thicker than ≈45 and ≈18 µm when considering material densities of 2.0 and 5.0 g cm −3 , respectively. The behavior in the case of ZELMB-s is comparable, but the design range is again somewhat more limited due to the a priori lower VED superiority over LIB.
Our theoretical considerations indicate that the GED and VED of ZELMB are hardly affected by modified protective coatings if the design parameters are chosen appropriately. The application of polymeric protection layers is a common approach to improve cycling stability. Assegie et al. achieved an average CE of 98.88 % over 400 cycles by applying polyethylene oxide (PEO) layers (0.2 µm) on a Cu-CC by spin coating. [41] With such thin layers a gain in VED (GED) of +25 % (+28 %) could be achieved compared to the reference LIB, which is very close to the unmodified ZELMB-l. Thicker PEO layers (10 µm) were used by Geng et al., [42] reducing the gain in VED (GED) to still acceptable 15 % (24 %). Wondimkun et al. introduced graphene oxide layers on Cu-CC achieving smooth Li layers during plating. [43] The layer thickness applied was ≈1 µm resulting in a possible gain in VED (GED) of 24% (27%). By using atomic layer deposition, very thin ceramic layers can be deposited on CCs, as shown by Tu et. al. for Al 2 O 3 and Tan et. al. for TiO 2 . [44,45] Since such layers typically do not exceed thicknesses in the range of 10 -20 nm, the gain in VED (GED) would amount up to 25% (28%) respectively and is therefore quasi-identical to the unmodified ZELMB-l.   Protective layers are not reported for ZELMB-s directly, but in the context of LMB-s, as most solid electrolytes are not stable against Li and thus need interfacial protection. Inorganic interlayers like aluminates, [46][47][48] halogenides, [49,50] LiPON, [51] or organic interlayers like PEO-composites [52] or alucone [53] are discussed. For example, West et al. [51] showed that a 1 µm LiPON coating could prevent the substrate from reaction with Li metal. Similar to ZELMB-l modifications, ALD is applied to form homogenous, dense inorganic coatings in the nm-range. For example, LATP [46] and LLZO [48] were coated by Al 2 O 3 with thicknesses up to 25 nm, showing significant improvements in cycling stability. Wang et al. [53] used molecular layer deposition (MLD) to prepare an inorganic-organic hybrid (alucone) interlayer (< 100 nm) and achieved stable cycling of symmetrical Li-cells up to 0.1 mA cm −2 . These examples show that design requirements of protection layers are fairly met by the existing approaches. Accordingly, corresponding ZELMB-s would maintain their GED and VED advantages over LIB. Overall, our theoretical consideration (Figure 8) and the results from literature indicate that the application of protection layers hardly impedes the gain in VED and GED. They are typically very thin to achieve necessary conductivity, hardly affecting the mass and volume of the cell.

Electrolyte and Separator
Typical tweaks for the success of ZELMB in terms of electrolytes lie not only in the obvious exploration and concentration-tuning of new solvents, salts, and additives, which hardly influences the nominal VED and GED, but also in more hidden and perhaps sometimes unintended details. For example, a number of academic studies reported impressive improvements using thick separators, soaked with excess electrolyte. Such cells are not only invulnerable to capacity losses by electrolyte decomposition, e.g. by repeated SEI formation or dry out, but also benefit from an inflated stability against Li dendrites. This dendrite-blocking nature of thick separators also applies to thick solid-state electrolytes. To estimate which degree of separator/ electrolyte excess in ZELMB is acceptable with regard to maintaining energy density advantages compared to conventional battery concepts, we have carried out corresponding parameter variations and calculated the influence on the GED and VED. Figure 9 shows the gain in VED (Figure 9a) and GED (Figure 9b) of ZELMB-l of different separator thickness and electrolyte excess, respectively, with respect to the reference LIB (for values in terms of Wh kg −1 and Wh L −1 see Figure S5, Supporting Information). As expected, the thicker the separator, the lower the VED and GED. For typical LIB electrolytes (≈1 g cm −3 ), a separator thickness of ≈115 µm (45 µm) would nullify the GED (VED) advantage of ZELMB-l over the reference LIB.

Liquid Electrolytes
Many academic ZELMB-l studies use multiple thick and highly porous separators. While this improves cycling stability by preventing dry-out and reducing the tendency to short-circuit,  the nominal energy density is drastically reduced. Assuming the most often applied two layers of borosilicate-microfiber filter (CAT No. 5401-090 E, Whatman) each with a thickness of 260 µm and 83% porosity, the VED and GED would fall below the reference LIB by -77% and -61%, respectively. Kwon et al. showed extreme differences in cycling stability when comparing cells with excess (24 µl mAh −1 ) and lean (4 µl mAh −1 ) electrolyte conditions. Assuming a 5-fold electrolyte excess in our calculations reduces the GED gain to -14% and the VED gain to -39% with respect to the reference LIB. Accordingly, the improvement in cycling stability by using excess electrolyte comes at the cost of a severe reduction in energy density.
Considering the literature data and the modelling results in Figure 9, applying thick separators and electrolyte excess does not seem to be a suited approach to develop competitive ZELMB-l. While this approach can increase cycling stability by extending the number of cycles before dry-out and short-circuit, the associated reduction in VED and GED is too severe to offset the benefits gained. To maintain a GED (VED) advantage of > 20% over the reference LIB, the separator thickness in ZELMBl needs to be < 34 µm (< 20 µm), when using typical electrolytes with densities of 1.2-1.3 g cm −3 .  Figure S6, Supporting Information). To maintain a VED (GED) advantage of >20% over the reference LIB, the SES thickness in ZELMB-s needs to be < 17 µm (28 µm) when using low density sulfide solid electrolyte and < 17 µm (9 µm) when using high density LLZO solid electrolyte. Due to the larger density of solid electrolytes compared to the infiltrated separators of conventional LIB, the GED is much more sensitive to separator thickness (compare Figure 9b and 10b).

Solid Electrolytes
However, the manufacturing of thin SES currently still considered a major challenge. SES are either manufactured as selfsupporting sheet, on a carrier film or supported by an electrode. Sulfide-based SES of 40 -70 µm thickness can be prepared by supporting the solid electrolyte layer though a nonwoven scaffold. [54,55] Lee et al. [56] reported a ZELMB-s with a 25 µm sulfide SES prepared by doctor blade casting on a PET carrier and laminated to the cathode afterwards. By direct casting of a sulfidebased separator onto a composite cathode a separator thickness as low as 9.5 µm has been realized. [57] Thus, sulfide SES that are already well below necessary design limit to reach significant improvement in GED. The fabrication of large LLZO SES is much more demanding. LLZO SES can be prepared by tapecasting and sintering. However, thin, flat SES larger than 1 cm 2 are not available yet. Sintered discs of 25-250 µm thickness are reported, [58,59] which corresponds to VED and GED gains/ Adv. Funct. Mater. 2023, 33, 2214891  www.afm-journal.de www.advancedsciencenews.com reductions of 12% and -4% down to -59% and -75%, respectively. Bilayer structures with a dense SES and a porous scaffold are simpler to process and to handle, as discussed above in the context of host materials for ZELMB-s. In this case, dense SES of 20 µm thickness can be realized, [34,35] fulfilling the design requirements in Figure 10.
Considering the literature data and the modelling results in Figure 10, careful SES thickness tailoring is key to achieve both, high energy density, and sufficient cycling stability in ZELMB-s. While using thick SES can increase cycling stability by extending the number of cycles before short-circuiting, the associated reduction in VED and GED is too severe to offset the benefits gained. Thus, future developments must focus on scalable processing of thin SES or combinations with host structures considering the design requirements in Figure 10.

Li Reservoir
A specific advantage of ZELMB is that the Li anode is formed in situ. A popular approach in academic ZELMB research is the formation of a Li reservoir. Typically, Li is extracted from the cathode and plated on the anode CC within the first cycle, but subsequent cycling proceeds only with a fraction of this Li deposit. While this results in improved cycling stability, due to the produced Li reservoir, the practically available VED and GED is reduced. Figure 11a displays this idea using the example of a ZELMB consisting of an NMC811 cathode and a bare Cu-CC. Initially, the cell is charged up to SOC = 100%. Neglecting side reactions, an amount of Li corresponding to the cathode capacity is deposited on the anode CC. By stopping the subsequent discharge process at SOC = 20%, a fraction of the Li plated during the previous charging process remains at the anode CC. This remaining amount of Li represents a Li reservoir that can be drawn on during the subsequent cycling in the restricted SOC range of 20 -100%. Accordingly, the cycling stability is increased, as the formed Li-reservoir can compensate for Lilosses, e.g. originating from the formation of dead Li. The larger the Li reservoir formed and thus the restriction of the SOC range, the more the capacity retention is increased. How-ever, the restriction of the SOC range leads to a reduction in the usable capacity of the cathode. At the same time, the formed Li reservoir is not active as long as there are no serious Li losses. Both circumstances lead to the usable GED and VED being reduced compared to the nominal values. Figure 11b shows the percentage gain in GED and VED of ZELMBs (for values in terms of Wh kg −1 and Wh L −1 see Figure S7, Supporting Information) compared to the reference case (cf. LIB in Figure 1) as a function of the SOC range used (or Li-reservoir formed). As expected, the smaller the SOC range used or the larger the Li reservoir formed, the smaller the GED and VED gain compared to the reference LIB. The GED (VED) superiority of ZELMB disappears for Li reservoirs corresponding to >24% (>22%) and >21% (>16%) of the cell capacity in the case of liquid and solid electrolyte, respectively.
Many academic ZELMB studies use the approach of building a Li reservoir (e.g., 50%) to extend the cycling stability of the cell, often combined with thick separators and electrolyte excess, as discussed above. While this indeed improves cycling stability, the nominal energy density is drastically reduced, and the conclusions drawn might not be relevant for practical applications. Louli et al. [13] approached this issue critically and conducted a systematic study regarding the influence of forming a Li reservoir. They showed stable cycling over 1000 cycles in ZELMB, when limiting SOC range to 23%, that is, forming a Li-reservoir equal to 77% of the nominal cathode capacity. However, as shown in Figure 11b, the formation of such a Li reservoir reduces the VED (GED) gain to −63% (−63%). When extending the SOC range to 80%, the capacity retention was 60% after 138 cycles, which is still remarkable, but the gain in GED (VED) over LIB would amount to only 3% (4%).
Considering the literature data and the modelling results in Figure 11, the in situ formation of a Li reservoir does not seem to be a suited approach to develop competitive ZELMBs. While this approach can increase the cycling stability, the associated reduction in VED and GED is too severe to offset the benefits gained. To maintain a GED (VED) advantage of 20 % over the reference LIB, the Li reservoir formed should be < 7% (< 6%) of the cathode capacity.
A promising approach to build a Li reservoir without reducing the cathode capacity is the use of sacrificial Figure 11. a) Illustration of the idea of building a Li-reservoir by cycling in a restricted SOC-range of the cathode after first charging. b) GED and VED gain of ZELMB over reference LIB depending on the amount of Li-reservoir formed during the first charging and restriction of the SOC-range used for cycling, respectively. Please note that for the sake of clarity, the schematic drawings are not to scale. materials/reactions. [60,61] In this case, an additional (side) reaction takes place at the cathode during the first charge besides Lideintercalation. The corresponding electrical charge converted by this sacrificial reaction at the cathode can be used to build up a Li reservoir at the anode. These approaches appear very promising but are still in their infancy. The sacrificial materials introduced and the side reactions (reaction products) taking place must be carefully selected and designed so as not to have any significant impact on the GED (VED) and cycle stability.

Combined Optimization Approaches
For practical applications, combining several approaches might be required to achieve sufficient cycling stability of ZELMB. For example, porous host structures with a lithiophilic gradient properties are considered a promising approach and were subject of numerous studies, e.g. Refs. [62][63][64] As shown above, each of the individual optimization approaches has distinct design restrictions to guarantee improvement over common LIBs. Combining the different approaches obviously will alter these thresholds as more inactive material is added to the cell's volume or mass. To illustrate the extent of the design constraints that result from combining several optimization approaches, we have examined selected examples. Figure 12 shows design constraints for ZELMB-l (Figure 12a-d) and ZELMB-s (Figure 12e-h), defined in such a way that the VED is identical to the LIB reference case (Figure 1). This means that if the parameters are chosen less favorably compared to Figure 12, the VED will fall below that of the reference LIB.
Corresponding numerical values are provided in Table S16  (Supporting Information).
As a starting point, Figure 12a show edge cases with only one optimization approach for ZELMB-l. For example, when using an additional protective layer thicker than 30 µm, the VED will fall below that of the reference LIB. Combining protective with lithiophilic coatings (Figure 12b) reduces the acceptable thickness approximately by the thickness of the lithiophilic coating. Implementing a Li-reservoir by restricting the nominal SOC range to 10% and adding an electrolyte excess with a 10 µm thicker separator reduces the acceptable thickness to 4.1 µm, further restricting the design freedom. Figure 12c shows that the already very limited design window of host structures appears even more difficult to realize practically when combined with additional optimization approaches. Due to the additional inactive masses, the thickness has to be reduced and the porosity increased in order to achieve the VED of the reference LIB at all. As concluded from the results above, building a Li-reservoir is associated with a massive reduction in VED. Accordingly, combining this approach with other modifications largely limits the design freedom. Combinations with lithiophilic and protective layers are an exception, since their small proportions hardly affect the volume of the cell, e.g. when compared to host structures. Figure 12. Thickness or specific volume thresholds for a,e) single approaches as well as b,f) protective layer, c,g) porous host structure and d,h) Lireservoir in combination with other optimization approaches for a-d) ZELMB-l and e-h) ZELMB-s. The labels in the red bars represent the required porosity of the host. Note that in Figure a,c,e, and g) the minimum porosity of the host to avoid "over-plating" is given.
In the case of ZELMB-s, these relationships are quite similar (Figure 12e-h), considering that the a priori VED gain versus LIB is smaller and thus the design thresholds of each optimization approach are somewhat more constrained. The main difference is the design freedom of the host (e.g., cf. Figure 12c,g). As described in detail above, the host structures in ZELMB-s are initially empty, while in ZELMB-l they are filled with liquid electrolyte, which must be displaced from the pores during Li-plating. Figure 12 proves that combining different modifications to ZELMB further restricts the individual design windows in view of achieving high energy density. Accordingly, even more care must be taken when selecting materials and choosing implementation strategies. Guan et al. [65] proposed a strategy to modify a Cu-foam by a lithiophilic gradient to induce uniform lithium deposition in ZELMB-l. With a thickness of 100 µm and an estimated porosity of ≈80 %, the design parameters are far outside the suitable design window (cf. Figure 12c). While excellent cycling stability is achieved, our calculations predict a severe reduction of GED and VED of −33 and −33% with respect to the reference LIB. A similar approach has been used by Cheng et al., [66] employing an Ag (100 nm) coated Cu foam (≈400 µm). Estimating a porosity of ≈85 % from SEM images, our model predicts that such a modification of the CC would result in a reduction of GED and VED of -68% and −73% with respect to the reference LIB.
These examples from the literature show, in agreement with our theoretical considerations, that the use of host structures in combined approaches poses a major challenge with respect to achieving high energy densities. According to Figure 12, the combination of lithiophilic and protective layers appears to be much more appropriate for improving cycling stability while preserving the energy density advantages of ZELMB. Wondimkun et al. [67] used lithiophilic silver nanoparticles with polydopamine coated on a Cu-CC additionally covered by graphene oxide as a protective layer and found significant improvement in cycling stability when compared to the bare Cu-CC. Assuming a total thickness of ≈10 µm, our calculations predict a VED and GED gain of 13% and 23% with respect to the reference LIB. Lee et al. [56] combined a carbon based protective layer with a thickness of ≈10 µm with lithiophilic Ag nanoparticles (≈0.57 µm [cm 3 m −2 ]) and achieved outstanding cycling stability in ZELMB-s. In this case, our calculations predict a significant VED and GED gain of 9 % and 16% with respect to the reference LIB.

Discussion
Based on our theoretical considerations and recent literature reports, it is concluded that combining lithiophilic and protective layers is the most effective and most straightforward combinational approach to increase cycling stability of ZELMB while preserving the energy density advantages over LIB. Implementing host structures, using thicker separators/ SES or excess electrolyte, and building a lithium reservoir can increase cycling stability, but at immense cost to energy density. This is reinforced when different optimization approaches are combined, so that the combination of lithiophilic and protec-tive layers can be clearly identified as the most target-oriented approach. For the other approaches or combinations with/of them, the design criteria derived in this work should be taken into account in order to advance the developments in an application-oriented manner. Of course, basic research work can also be carried out on model systems. However, in this case, transferability to practical applications is not directly given and performance claims should be made with the caveat of the model character and not generalized at an early stage of development.
The VED and GED, which were mainly considered in this study, are of course not the only relevant criteria to evaluate the suitability of a novel battery technology. Costs, safety, environmental compatibility, the availability of raw materials and recyclability also play a crucial role in a fair comparison of emerging and established battery technologies. Unfortunately, most of these aspects are difficult to quantify at lower technical readiness level. Accordingly, a holistic assessment is not fully possible at this stage of ZELMB development.
In principle, ZELMBs should be cheaper than LIBs due to the absence of an active anode material and the corresponding simplified production. In this respect, in the case of identical GED (VED), ZELMB could be preferred over LIB if other relevant properties are equal. However, this consideration does not take into account the costs of the necessary modifications to increase the cycling stability of ZELMB to the level of LIB. It is conceivable that the costs for elaborate CC modifications exceed those of the established LIB anode production. In this case, the cost advantages of ZELMB compared to LIB would be significantly reduced and the superiority of ZELMB starts to vanish.
The same argumentation holds for other evaluation criteria, such as the availability of raw materials and recyclability. ZELMB could be preferred over LIB even with similar GED (VED) if they offer significant advantages in these aspects. In principle, ZELMB have obvious advantages in terms of material consumption and recyclability, since no active anode material has to be used. However, if, for example, rarely occurring materials have to be used for the CC modification or as an electrolyte additive in order to achieve sufficient cycling stability, the actual advantages of ZELMB compared to LIB could disappear. Similarly, it is conceivable that the necessary optimization steps to achieve sufficient cycle stability complicate the recycling of ZELMB. Accordingly, the advantages compared to the LIB could start to vanish. In this respect, in addition to increasing energy density, such aspects must also be urgently considered in the development and optimization of modification strategies for ZELMB.

Conclusion
The current state of knowledge on ZELMB suggests that sufficient cycling stability for practical application can only be achieved by various, possibly combined, modifications of the original cell concept. However, the various optimization approaches introduce inactive materials into the cell, which may negate some of the original advantages of ZELMB. We show that at a certain level of modification, the energy density advantages of ZELMB over conventional LIB start to disappear and derive quantitative design guidelines for future developments.
Theoretical considerations and literature examples show that even for low-density materials, designing host structures for ZELMB with liquid electrolyte is extremely challenging in terms of achieving high energy density. In the case of ZELMB with solid electrolyte, the design freedom of host structures is larger and first results from literature suggest successful implementation of host structures while maintaining high energy density. Lithiophilic and protective coatings have relatively easy to implement design requirements. Literature shows that applying coatings with a thickness < 1 µm using thin film technologies, can improve the cycling stability. Our calculations prove that the addition of such layers hardly affects the mass and volume of the cell, preserving the energy density advantages of ZELMB over conventional LIB. Considering the literature data and the modelling results, applying thick separators/SES, electrolyte excess or building a Lireservoir do not seem to be a suited approaches to develop competitive ZELMB. While these measures increase cycling stability by extending the number of cycles before dry-out, short-circuit, or CE break-down, the associated reduction in VED and GED is too severe to offset the benefits gained. To maintain significant advantage over LIB, the separator thickness in ZELMB-l needs to be < 20 µm. To form a Li-reservoir without losing capacity, the application of sacrificial materials appears a viable solution but is still in its infancy. For practical applications, combining several approaches might be required to achieve sufficient cycling stability of ZELMB. This obviously further restricts the design freedom of the individual optimization approaches. Overall, it is evident that the combination with already severely limited approaches significantly reduces the overall design freedom. Ultimately, the combination of lithiophilic and protective layers appears to be the most target-oriented for future developments. Due to the large design freedom, the approaches can be easily implemented to increase cycling stability while preserving energy density. We hope that the findings of this study and the derived quantitative guidelines will help to target the development of ZELMB with liquid and solid electrolytes.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.