High dielectric 3D scaffold to suppress Li-dendrites and increase the reversibility of anode-less Li-metal anodes

The lithium metal anode is intensively investigated because it considerably increases Li-battery energy density. However, the formation of dendritic/mossy Li-metal microstructures amplifies electrolyte decomposition and Li deactivation. Here we investigate the impact of a high-dielectric porous scaffold, aiming to eliminate the fundamental driver for dendritic/mossy Li-metal growth, the large electrical field gradients at inhomogeneities at the anode surface. In an anode-less (Li-metal free) high-dielectric porous scaffold, this promotes dense plating as observed by operando solid-state NMR. Even in a simple carbonate electrolyte, 1M LiPF 6 in EC/DMC, the high-dielectric scaffold improves the plating/stripping efficiency up to 99.82%, extending the cycle life, indicating that electrolyte decomposition is minimized by the induced compact Li-metal plating. The large porosity of the scaffolds, non-optimized and easy to prepare, enables a specific capacity beyond 2000 mA h g -1 , presenting a facile approach to promote compact Li-metal plating to improve Li-metal anode efficiency and safety. suggesting a high 3D porous scaffolds are through a facile casting approach using both a high dielectric scaffold material, BaTiO 3 (BTO) and a low dielectric scaffold material, Al 2 O 3 (AO), to distinguish between the impact of a porous scaffold and the high dielectric constant on the electrochemical Li-metal plating. These electronically insulating scaffolds show minimal electrochemical activity, and the added weight leads to specific capacities exceeding 2000 mA h g -1 . 7 Li solid-state NMR of an anode-less BTO scaffold-LiCoO 2 full cell demonstrates that the high dielectric scaffold induces compact plating and efficient stripping. Half-cells with a BTO scaffold cycled against Li-metal exhibits 99.82% CE after the first two initial cycles, low overpotentials and an extended cycle life in when the worst-case scenario 1 M LiPF 6 EC/DMC electrolyte is used, under different current and capacity conditions. With the same basic electrolyte, full cells also demonstrate improved performance with an average CE of 99.37%. These results demonstrate the use of high dielectric scaffolds that take away the driving force for inhomogeneous Li-metal deposition as a strategy to improve the reversibility and safety of Li-metal anodes in an anode-less configuration. The next avenues to explore are combinations with more stable SEI forming electrolytes and additives, optimization of the high dielectric scaffold, to minimize the capacity loss during the first cycles and to further extend the cycle life.

The development of high energy density batteries beyond current Li-ion battery technology is necessary to meet the increasing demand for mobile applications such as electric vehicles. 1-5 Lithium metal is considered the ultimate anode material because it possesses the highest theoretical specific capacity, 3860 mA h g -1 , approximately ten times that of current graphite-based anodes, and low redox potential that endows it with a high battery output voltage (-3.040 V vs standard hydrogen electrode). [6][7][8][9][10] However, upon repeated battery charging and discharging, the plating and stripping of Li-metal induces irreversible processes that lead to fast capacity decay which drastically limits the Li-metal battery cycle life. [6][7][8][9][10] The formation of high surface area dendritic and mossy Li-metal morphologies, in combination with the vicious reactivity of Li-metal towards common electrolytes, leads to an irreversible loss of active Li towards solid electrolyte interface products as well as the formation of inactive Li-metal. The latter is often referred to as "dead" Li, which has lost contact with the current collector. [11][12][13][14] Additionally, dendritic structures may penetrate the separator/electrolyte and reach the cathode, causing an internal short-circuit that may induce rapid spontaneous discharge and consequential safety hazards. [6][7][8][9][10] Loss of active Li is in research cells often masked by the excess of Li-metal present, but in practical cells, the amount of excess Li should be minimized to maximize the energy density. 15 Ultimately, all active Li is stored in the cathode in the discharged state. This so-called "anode-less" or "anode free" design, has the additional advantage of not necessitating Li-metal handling.
Strategies that aim to suppress, prevent and block dendrite formation are intensively investigated, [6][7][8][9][10]16,17 typically guided by our current understanding of dendrite nucleation and growth. The space charge model of Chazalviel predicts that when the Li-ion concentration on the surface of the anode drops to zero, after the characteristic Sand's time, plating becomes inhomogeneous and self-amplified growth of dendrites is induced. 18 This motivates the use of strategies that enhance ion mobility, increase the transference number and introduce a large anode surface area to promote a homogenous Li-ion flux and to prevent ion depletion at the anode surface. [6][7][8][9][10]16,17 Modelling of the early stages of nucleation and growth under heterogeneous electrodeposition indicates that the anode surface inhomogeneities, particle size and wettability play a critical role in facilitating dendrite formation. 19 This implies that dendrite growth can be steered by controlling these parameters on the anode, as experimentally demonstrated. 20 The comparison with Mg metal, that does not favour dendrite formation, indicates that both the high surface diffusion barriers for Li and the low surface energy density promote dendrite formation. This is because Limetal diffusion away from the tip where it is initially deposited is hindered and large surface area morphologies are allowed. 21 To regulate these issues, strategies are employed which aim to control the Li-metal-electrolyte interface via the SEI composition. 22 Residual stress within the Li-metal appears to be another driving force for Li-metal growth. 23 This motivates the design of substrates that can release this stress. 10,23 Finally, investigations of how mechanical forces affect dendrite nucleation and growth, have shown that electrolytes with a shear modules, at least 2 times larger than that of Li-metal, can prevent dendrite growth. 24 This has given rise to research where mechanically strong separators, solid electrolytes and protective films are investigated. [6][7][8][9][10]16,17 3D scaffolds are an interesting approach, as they provide the possibility to control the interface, the local interface environment and to a large extent the charge transport. They can also provide a route to mitigate delamination due to their ability to accommodate large volumetric changes upon Li-metal plating and stripping. In a 3D scaffold that is electronically conducting, the electric field is roughly uniform and the local current density is reduced, suppressing dendrite growth. [25][26][27][28][29][30] However, the 3D porous conductive matrix is equipotential, due to which Li-metal can also be deposited on top of the 3D matrix, thus negating its targeted function. 31 Here we explore an alternative approach through the introduction of a material with a high dielectric constant (which we will refer to as a high dielectric material) as a 3D scaffold. Due to the polarizing power of high dielectric material, an effective immobile surface (space) charge density ℎ is established, opposing the applied field in the battery (ferroelectric effect), that scales with the dielectric constant (∇ • = ℎ where = 0 , 0 and the vacuum and relative permittivity and the electrical field). As a consequence, the electrical field lines are drawn towards the high dielectric material (dictated by Gauss Law) which leads to lowering of the divergence of the electrical field in the vicinity of the Li-metal deposition. At the tip of a dendrite near a high dielectric material, this effectively leads to a decrease in the electrical field divergence, and thus to a lower local electrical field gradient. This is shown in Figure   1, where the calculated electrical field gradient near a dendrite with, and without high The high relative permittivity is shown to suppress dendrite growth, promote homogeneous plating, thus damping the self-amplifying cycle of SEI formation and dendritic growth. Nevertheless, several challenges remain, one of which is the reactivity of the scaffold. Though very limited, it contributes to the initial capacity loss, and the cycling of these anodeless scaffolds, still appears to terminate via a short circuit. This prompts the investigation of combinatorial approaches, especially utilizing optimized electrolytes and binders to improve performance. The proposes scaffolds are attractive from a practical standpoint, as they are prepared by means of regular electrode casting. Thereby, the proposed strategy presents new insights into the role of the permittivity on Li-metal deposition and a promising approach for the design of (anode-less) scaffolds to achieve efficient Li-metal anodes.

Results and Discussion
Electrical field calculations. To understand the impact of the dielectric constant of a dielectric block on the electrical field gradient at a nearby Li-metal dendrite, the electrical field is simulated in two dimensions. This is performed for both a bare Cu   using an N-methyl-2-pyrrolidone (NMP) solvent to obtain a slurry that was cast on a regular Cu current collector. The NH4HCO3 acts as a template to achieve a high porosity 33 which is required to achieve high specific anode capacities, taking into account the weight of the scaffold. After that, the electrodes were dried under vacuum at 80 o C to remove the NH4HCO3 template. The resulting porosity of the AO and BTO scaffolds is 62% and 64%, respectively and thus comparable. The thickness of these two electrodes is around 45 μm, including 10 μm copper foil. When this porous volume is completely filled by Li-metal, the specific capacity of the anode is approximately 2240 mAh g -1 . Aiming to expose the impact of a high dielectric scaffold on the cycle life of anode-less configuration of these electrodes, a blank 1M LiPF6 in EC/DMC electrolyte was used. As mentioned earlier, this represents a worst-case scenario because it is well known that this electrolyte formulation leads to uncontrolled dendrite and SEI formation. 9,34 Operando solid-state NMR. To test the hypothesis that the Li-metal deposition is less dendritic and more homogeneous in the pores of the high dielectric scaffold, and that this leads to better reversibility, operando 7 Li solid-state NMR is performed, the results of which are shown in Figure 2. Operando 7 Li solid-state NMR is a direct probe of Li in realistic cell conditions and allows to distinguish differences in the Li-metal microstructure. 12,35,36 As shown in Supplementary Figure S7, the signals from Lispecies in the SEI and electrolyte are in the expected region of diamagnetic materials (-10 to 10 ppm), Li in the LiCoO2 (cathode) in the wide region of -50 to 50 ppm, which both can be readily separated from the Li-metal chemical shift which is dominated by its knight shift. 35 The NMR radiofrequency (RF) has a limited penetration depth into the Li-metal, referred to as skin depth, which at the presently employed B0 field of 11.7 T amounts to approximately 11 m. 13 As a consequence, for deposited Limicrostructures such as dendrites and mossy structures, typically smaller than a few micrometres, the RF penetration can thus be expected to be complete. A Li-metal strip 13,35 gives rise to a resonance at ~245 ppm when placed perpendicular to the fixed magnetic field B0 and at ~270 ppm when the strip is parallel to B0, a consequence of the bulk magnetic susceptibility effect. 13 An anode-less design (absence of Li-metal at the anode in the initial state) is investigated by operando solid-state NMR, utilizing LiCoO2 as the counter electrode (and Li source), such that the observed Li-metal signal arises only due to the deposition at the anodes of interest, similar to a recently reported study 37 . Interestingly, the Li-metal chemical shift in the BTO scaffold is similar to that observed for dense Li-metal plated in the separator when plated under pressure. 12 The density of the BTO scaffold (64 %) is similar to that of the separators, thus the operando 7 Li solid-state NMR indicates that Li-metal plating fills the BTO scaffold pores, preventing the growth of mossy and dendritic microstructures. Discharge (Li-metal stripping) of the bare Cu current collector results in high overpotentials, indicating a strongly increased internal resistance, which is associated with severe SEI formation that hinders Li-ion transport as seen in Figure 2a. The intensity of the 7 Li metal NMR resonance hardly decreases during discharge (Figure 2b and 2c), reflecting the difficulty to strip the Li-metal deposits from the Cu, which upon cycling results in rapid accumulation of inactive "dead" Li-metal (see Figure 2b) comprehensively studied recently using operando NMR. 37 In contrast, the voltage during discharge (Li-metal stripping) for the BTO scaffold is nearly the same as during charge, indicating that the internal resistance is practically unaffected by the SEI formation. Concurrently, the 7 Li metal resonance completely disappears upon the charge, as confirmed in Figure 2f, demonstrating that all Li-metal can be completely stripped from the BTO scaffold, and thus that a high Li-metal stripping/plating efficiency is achieved. The large increase in Li-metal signal after the first cycle can be explained by a low first cycle efficiency, where a significant fraction of the capacity is lost in the formation of SEI products by introducing Li-metal at the anode. In summary, operando 7 Li solid-state NMR demonstrates that the presence of the high dielectric anode-less BTO scaffold results in more compact and less dendritic/mossy Li-metal, in line with the suggested impact of a high dielectric scaffold on the electrical field gradients shown in Figure 1, which reduces the amount of observed "dead" Li-metal and SEI formation, even in a conventional 1 M LiPF6 EC/DMC electrolyte.  The CE is an essential parameter to evaluate the reversibility of the lithium metal anode, reflecting the irreversible loss of Li to both "dead" Li (Li-metal electrically isolated from the current collector) and to the formation of the SEI species during each cycle. Cycling is performed to a fixed Li-metal capacity of 1 mA h cm -2 at rates of 2, 4 and 8 mA cm -2 as shown in Figure 3a-c, and to a lithium metal capacity of 4 mA h cm -2 at a rate of 4 mA cm -2 , as shown in Figure 3d. Li-metal stripping is always limited by a 1.0 V vs Li/Li + cut-off voltage. The two initial cycles are always performed at a current density of 0.5 mA cm -2 to form the SEI, leading to a relatively low CE in the order of ~60% and ~85%. As shown in the cyclic voltammetry curves in Figure S8, the BTO scaffold shows a small peak at 1.2 V (and no activity is observed for AO).

Electrochemical evaluation half cells. Galvanostatic cycling in a half
However, there is no obvious structural change of the BTO scaffold after the 1 st and 100 th cycles as verified by XRD ( Figure S9).  (Figure 3a--c). In sharp contrast, the cells with bare Cu and the AO scaffold show rapid decay in the cycling efficiency, leading to early cell death before reaching 20 cycles. Nevertheless, the BTO scaffold can maintain this for a limited number of cycles as indicated in Figure   3, which is remarkable given the non-optimized LiPF6 EC/DMC electrolyte used. Also cycling towards a larger (practical) capacity of 4 mA h cm -2 at 4 mA cm -2 , which including the BTO scaffold equals a specific capacity of 2000 mA h g -1 , results in a CE of 99.68% that can be maintained for more than 160 cycles (Figure 3d). This is equivalent to a specific capacity above 2000 mA h g -1 , which in combination with a sulphur-carbon cathode would yield a cell level energy density above 500 Wh kg -1 . The evolution of the voltage during cycling shown in Figure 4 and Figure S10 provides more insight into the evolution of the CE seen for the different anodes ( Figure   3). Under all cycling conditions tested, the anode-less Cu and AO scaffold result in a rapid increase in the plating and stripping potentials as shown in Figure 4a- approximately 68, 220 and 375 mV at current densities of 2, 4 and 8 mA cm -2 , respectively. Nevertheless, also for the BTO scaffold, the overpotentials increase, indicating that upon extended cycling, these cells also fail due to SEI growth. These conclusions are further supported by electrochemical impedance spectroscopy (EIS) measurements shown in Figure 4e-g. The total resistance of the cells with the Cu and AO scaffold anode shows a strong increase after 100 cycles at 2 mA cm -2 shown in Figure 4g, associated with the SEI formation driven by mossy/dendritic Li-metal growth. At the same conditions, the BTO scaffold has a much lower interfacial resistance, indicating less SEI formation. In addition to half-cells, full cells were assembled using the BTO scaffold as a lithium free anode (anode-less), or a Li-metal anode, both paired with an NCM cathode, again in combination with a conventional 1 M LiPF6 EC/DMC electrolyte (as shown in Figure S11 and S12). The BTO scaffold outperforms the Li-metal anode in cycling rate and stability demonstrating a high average CE of 99.37%.

Figure 4 Half-cell voltage evolution and impedance of the anode-less Cu and AO and BTO scaffolds vs a Li
Li-metal morphology. An SEM analysis is performed before and after cycling the cells at a current density at 2 mA cm -2 to a specific capacity of 1 mA h cm -2 , comparing bare Cu, the low dielectric constant AO scaffold and the high dielectric constant BTO scaffold, shown in Figure 5. On comparison, the BTO scaffold appears to lead to a more uniform lithium deposition, as no dendritic/mossy structures are observed (Figure 5b,e). The cross-sectional SEM image of the BTO scaffold electrode( Figure   S13a) also taken after plating to 1 mA h cm -2 at a current density of 2 mA cm -2 , suggests that the deposited Li-metal is dense and well-confined in the 3D scaffold of the BTO layer consistent with the operando 7 Li solid-state NMR measurements depicted in Figure 2. In contrast, Li deposition on the bare Cu foil exhibits islands of accumulated dendritic structures (Figure 5a,d) and at the surface of the AO scaffold, dendrites are observed (Figure 5c,f).  (Figure 4), eventually leading to cell death. A comparison of the high dielectric BTO scaffold ( ≈4000) and the low dielectric AO scaffold ( ≈8), having a similar morphology (in terms of loading, particle size distribution and porosity), makes it possible to distinguish between the impact of a non-conducting scaffold morphology and the impact of the dielectric constant of the scaffold on the Li-metal plating. As schematically shown in Figure 6b, the 3D highdielectric BTO scaffold takes away the electrical field gradients at the tip of uneven Limetal deposits, as supported by electrical field calculations (Figure 1). This leads to a more homogeneous Li-ion flux, and thus to less mossy/dendritic and more homogeneous Li-metal deposits as recently shown by Guo 32 and as evidenced by operando solid-state NMR (Figure 2) and SEM (Figure 5). This prevents the accumulation of deactivated Li-metal which is responsible for the highly improved cycling efficiency (Figure 3) and lowers the exposed Li-metal-electrolyte area, minimizing SEI formation and thus preventing the uncontrolled and self-amplified growth of the SEI and dendritic Li-metal deposits, rationalizing the low overpotentials during extended cycling observed in  In conclusion, the challenge faced by Li-metal anodes is to prevent dendritic and Galvanostatic cycling was conducted on a Maccor battery testing system. The impedance measurements of the coin cells were carried out on Autolab between 100 kHz and 0.01 Hz. Cyclic voltammograms (CVs) were recorded using the same electrochemical workstation at a scan rate of 1 mV s -1 in the range of -0.5 V -3 V.
Characterization of the Materials and the Electrodes. SEM images were obtained of the 3D scaffolds after a discharge capacity of 1 mA h cm -2 . Before SEM imaging, the electrodes were rinsed with dimethyl carbonate in a glove box to remove lithium salts and dried several times in a vacuum chamber. Cross-section SEM samples were prepared by cutting with scissors in the glove box. Subsequently, samples were transferred into an SEM (JEOL JSM-6010LA) machine under dry Argon conditions, and images were taken using an accelerating voltage of 2-10 kV (secondary electron).
Nitrogen adsorption-desorption isotherms were recorded using an automatic surface area and porosity analyzer (Micromeritics) at 77 K. The particle size distribution of BTO and AO after ball milling was measured using Microtrac S3500.