Lithium metal stabilization for next-generation lithium-based batteries: from fundamental chemistry to advanced characterization and effective protection

Intrinsic properties of Li metal anodes

Li metal possesses the smallest radius of all metal atoms and the lowest density (0.534 g cm^-3) of alkali metals, thereby endowing it with ultrahigh storage capacity. The theoretical specific capacity (Qc) of Li can be calculated based on the following equations, with an ultrahigh Qc of 3860 mAh g^-1 obtained^[8]:

Li - e^- ↔ Li                     (1)

As an electrode, Li has the most negative electrochemical potential (-3.040 V vs. a standard hydrogen electrode) among all anodes, which endows LMBs with high discharge voltage and energy density. Furthermore, Li metal anodes are able to provide an energy-dense Li-ion source, which enables the anode to be compatible with extensive cathode materials, including non-lithiated cathodes, leading to the versatility of battery design. For LMBs, the electrochemical deposition/dissolution of Li metal takes place on the anodic side. Simultaneously, Li⁺ ions are consumed/accumulated at the cathodic side. According to the paired cathode materials (e.g., intercalation cathode materials, O₂ and S), LMBs can be divided into three primary categories, namely, lithium-lithium intercalation compound, Li-O₂ and Li-S batteries. The reaction principle of LMBs during the charge/discharge processes is described as follows^[9-11]:

Lithium-lithium intercalation compound batteries (taking Li_1-xMO₂ as an example):

Anodic reaction: xLi = xLi⁺ + xe^-          (3)

Cathodic reaction: Li_1-xMO₂ + xLi⁺ + xe^- = LiMO₂     (4)

Li-O₂ battery:

Anodic reaction: 2Li = 2Li⁺ + 2e^-          (5)

Cathodic reaction: 1∕2O₂ + H₂O + 2e^- = 2OH^-        (6)

Or 2Li⁺ + 2e^- + O₂ = Li₂O₂     (7)

Li-S battery:

Anodic reaction: 16Li = 16Li⁺ + 16e^-          (8)

Cathodic reaction: S₈ + 16Li⁺ + 16e^- = 8Li₂S        (9)

Challenges facing Li metal anodes

Despite the tremendous potential of next-generation secondary battery applications, the Li anode still faces a multitude of challenges to overcome: (i) electrolyte depletion due to the continuous reaction with Li metal and mechanical instability of the SEI layer; (ii) dendrite formation deriving from initial uneven Li nucleation and the subsequent selective deposition; and (iii) infinite relative volume change. These shortcomings inevitably lead to unsatisfactory Coulombic efficiency (CE), short lifetimes and even safety problems, such as thermal runaway and explosion hazards.

Unstable formation of SEIs on Li surface

The SEI has become a critical research topic since the pioneering work of Aurbach and Peled^[12-15]. Different models, including mosaics (polyhetero microphase), vacancies (Schottky lattice defects), double-layer capacitors, multilayers and monolithic SEIs, can be used to explain the chemical and physical properties of SEIs [Figure 2A-E]^{[12,13,16,17]}. Among them, the mosaic model with various components and a dual-layer architecture is commonly accepted. The inner layer, which is close to the anode, consists of species with low oxidation states, including LiF, Li₃N, Li₂O, Li₂CO₃ and LiOH, and is labeled as an inorganic layer. In contrast, the outer layer of the SEI is composed of species with high oxidation states, such as RCOO₂Li, ROLi and ROCO₂Li, and is labeled as the organic layer.

Figure 2. Schematic illustrations of various SEI models: (A) mosaics (polyhetero microphase); (B) vacancies (Schottky lattice defects); (C) double-layer capacitors; (D) multilayers; (E) monolithic SEI. Reprinted with permission from Ref.^[126]. Copyright (2021) Cell Press.

Due to the significantly negative electrochemical potential of Li⁺/Li, virtually any available electrolyte can be reduced by Li metal to form electrically insulating and ionically conductive species on the electrode surface [Figure 3A]. Goodenough et al. explained the SEI formation mechanism based on the frontier molecular orbital theory^[18]. As shown in Figure 3B, μ_A and μ_C refer to the electrochemical potentials of the anode and cathode, respectively. Accordingly, E_LUMO and E_HOMO are representative of the voltage that corresponds to the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) of the electrolytes, respectively. If μ_A > E_LUMO, electrons tend to transfer from the anode surface to the unoccupied orbital of the electrolyte, triggering the reduction reaction of the electrolyte. The generated SEI serves as a barrier, preventing further decomposition of the electrolyte. Analogously, redox reactions facilitate the formation of the SEI between the electrode and electrolyte in the case of μ_C < E_HOMO.

Figure 3. (A) Schematic illustration of challenges facing Li metal anodes. (B) SEI formation conditions in a liquid electrolyte. Reprinted with permission from Ref.^[8]. Copyright (2017) American Chemical Society. (C) Time-lapse TEM images of Li whisker growth. Reprinted with permission from Ref.^[22]. Copyright (2020) Springer Nature. (D) Schematic representation of dead Li including Li⁺ ions in SEI formation and Li metal wrapped by insulating byproducts. Reprinted with permission from Ref.^[92]. Copyright (2020) American Chemical Society. (E) Fire and (F) explosion phenomena of Li metal anode. Reprinted with permission from Ref.^[38]. Copyright (2020) Springer Nature. (G) Schematic diagram and scanning electron micrographs of high-rate cycled angled sections showing failure within two stacked Celgard 2325 separators. Reprinted with permission from Ref.^[39]. Copyright (2021) American Chemical Society.

The component and structure of the SEI film are strongly dependent on the composition of the electrolyte, mainly organic carbonate- and ether-based electrolytes, which determine the primary properties of the SEI film. Organic carbonates are the most commonly used electrolyte solvents for LMBs. The preliminary composition of the SEI in a carbonate-based electrolyte system is mainly made up of Li alkyl carbonates (ROCOOLi, where R is an organic group related to the solvent), which are generated by the one-electron reduction of organic carbonates^[14]. More stable components of Li₂CO₃, Li₂O and Li halides can be presented in the SEI depending on the used salts^[15]. However, the SEIs formed in organic carbonate-based electrolytes lack flexibility, which makes them susceptible to interfacial fluctuations. With significant dendrite suppression capacity, ethers are considered as strong candidates for electrolyte solvents for LMBs in multiple systems, which can be ascribed to the formation of oligomers with strong affinity to the Li surface and excellent flexibility^[19]. For example, 1,3-dioxolane (DOL), as a widely used ether solvent in LMBs, can form oligomeric SEI films with superior flexibility and strong binding affinity to Li metal. As the Li metal anode is cycled in the DOL-based electrolyte, DOL is reduced on the Li metal surface to form different alkoxy species, leading to its anionic partial polymerization to form LiO-(CH₂CH₂OCH₂O)_nLi type species. A surface film is thus formed by these short-chain oligomers of poly(ethylene oxide) that is elastomeric and can accommodate the morphological change of the anode surface upon long-term cycling.

The SEI film is an essential path for the desolvation and reduction of Li⁺ ions on the electrode, where the diffusion behavior of Li⁺ ions crucially affects the Li deposition morphology. Specifically, once solvated Li⁺ ions are transported to the SEI surface under the electric field, the SEI is capable of desolvating Li⁺ ions and the naked Li⁺ ions are then transported to the current collector where they are reduced. Furthermore, robust and dense SEI can prevent Li exposure, which directly causes the Li corrosion. A desired SEI for efficient and safe LMBs should possess the following features: (1) excellently ionic conductivity; (2) proper thickness, which is sufficient to prevent electron penetration to electrolyte completely and appropriate for the diffusion resistance of Li⁺ ions; (3) extraordinary chemical and electrochemical stability in composition during long-term cycling; and (4) excellent mechanical properties to adapt the volume variation of Li during repeated charge/discharge processes.

Li dendrite and relative issues

During electrodeposition, the growth of metal “dendrites” is a common phenomenon. Many metals, such as Zn, Cu, Ag, Sn and so on, were found to exhibit ramified morphologies under given electrodeposition conditions^[20,21]. Generally, metallic electrodeposition can be divided into two stages: early nucleation and late deposition [Figure 3C]^[22]. During the deposition process, a critical thermodynamic radius needs to be reached to favor dendrite formation energetically and a critical kinetic radius is required to keep the isolated embryo growing. As the growth continues, localized electric fields dominate the morphology of the Li deposits in the late growth regime to form dendrites with prolonged length and sometimes augmented diameter.

To explain the propagation of dendrites during the electrochemical deposition of metals, various models have been proposed: (i) surface tension model; (ii) diffusion-limited model; and (iii) electromigration-limited model^[23]. Barton et al. comprehensively proposed the surface tension model in which surface tension between the electrolyte and metal was considered to be one of the driving forces for Li dendrite growth^[24]. At the initial stage, the electrodeposition of metal ions on an elevated region of the anode surface with spherical diffusion was faster than that on a flat surface with the linear diffusion, leading to preliminary protrusions with enhanced diffusion conditions^[24-26]. Later, the possible Li morphology could be estimated via a fluid dynamics mathematical model and it was confirmed that particle-shaped Li deposition would deform Li dendrites under high surface tension^[27]. In the classical diffusion-limited model, metal ions move randomly and reach active sites of the electrode where the deposition probability of mobile ions can be defined as a balance between the rate of the electrochemical reaction and the bulk diffusion^[28]. The diffusion becomes dominant rather than the reductive deposition while a deposition probability is low, leading to the suppression of dendrite formation. In contrast, the high deposition probability always causes the ramified dendrite structure. The model was applied to simulate the reductive deposition of Li⁺ ions and found that the formation of dendrites was related to the competition between the diffusion of Li⁺ ions in the SEI film and the interfacial deposition reaction^[29]. Moreover, dendrite propagation can be successfully inhibited by pulse charging conditions, which altered the electrodeposition morphology of various metals, thus effectively suppressing dendrite growth^[30].

In the Chazalviel model, dendrite initiation was described to be limited by the electromigration process rather than the diffusion process^[20]. It was demonstrated that the current density can change the ion concentration gradient, i.e., the lower current density maintains a stationary ion distribution without propagation of Li dendrite, while the higher current density introduces the depletion of cations and anions at the electrode interface and the resulting local space charge field leads to the growth of dendritic Li. The critical current density, J*, was defined as the boundary behavior between the high and low current densities as follows:

where e represents the electronic charge, C₀ represents the initial concentration of the electrolyte, D represents the ambipolar diffusion coefficient, t_a represents the transport number of anions and L represents the distance between the electrodes. In particular, the nuclei of Li dendrites were formed when the concentration of ions near the electrode dropped to zero. The time of the concentration of ions dropping to zero was defined as Sand’s time, τ_Sand. Furthermore, the corresponding current density was labeled as the limiting current density when the ion concentration approached zero.

In the 1990s, Chazalviel’s space charge model was verified and supplemented^[31,32]. In particular, to explain the Li dendrite formation and growth even at a low current density below J*, the heterogeneity of the electrode surface was considered in the model based on the fact that the variation in local current density could be caused by the inhomogeneous microstructures of the electrode surface^[33]. Severe Li dendrite growth usually produces a large number of derivative problems, such as dead Li (i.e., Li losing contact with the electrode surface and can no longer participate in the electrode reactions) [Figure 3D]. The formation of dead Li indicates an irreversible loss of battery capacity^[34]. Both uncontrollable dendritic Li growth and dead Li are believed to be the reason for the low CE (less than < 99%) of Li deposition/dissolution in nonaqueous electrolytes^[35]. In addition to the formation of dead Li, the newly exposed Li dendrites on the surface can be rapidly corroded by the nonaqueous electrolyte, leading to the formation of a thick SEI with poor conductivity. Due to the repeated generation/fracture of the unstable SEI during charge/discharge processes, the exposed fresh Li reacts with the electrolyte continuously. Electrolyte depletion and electrochemical corrosion of Li metal eventually lead to low CE and capacity decay^[36,37]. Among the above-discussed issues triggered by dendritic growth, safety concerns, including battery fires [Figure 3E] and even explosion hazards [Figure 3F], which result from short circuit risks [Figure 3G], are also top priorities for developing practical LMBs^[38,39].

Morphological and microstructural characterization of Li surface

In situ imaging techniques, including in situ optical image techniques, electron-based analysis, scanning probe-based tools, X-ray and synchrotron-based analysis, magnetism-based techniques, neutron-based analysis and other in situ techniques, have been widely applied to qualitatively study the composition of Li-containing species^[43-46]. Optical image and microscopy technologies are based on the light wave of the object being reflected and projected by the detector to form the image. Therefore, optical microscopy (OM) in conjunction with an in situ reaction cell and electrochemical workstation is the most widely used image observation equipment. More importantly, it is the most direct, rapid, feasible and low-cost method to analyze the deposition and dissolution behavior of Li metal. In 1998, Osaka et al. employed in situ OM to explore the effect of the electrolyte on the Li deposition morphology and anode cycling efficiency^[47]. Arakawa et al. investigated the relationship between dead Li formation and the Li plating morphology by recording the film attached to an optical microscope to obtain photographs^[34]. Subsequently, Yamaki et al. employed in situ OM to probe the internal mechanism of deposition and dissolution of Li metal in LiClO₄-PC and proposed the famous whisker model based on a mathematical analysis showing that dendritic Li may grow from the current collector during deposition^[27]. Inevitably, stress from the bottom regions can seriously destroy the SEI interphase, enabling Li whiskers to continuously grow in the form of extrusion. Nevertheless, Steiger et al. proposed a different view of Li cycling behavior in a LiPF₆-EC/DMC electrolyte^[48]. They claimed that crystal defects control the Li deposition behavior, i.e., newly fresh Li atoms are electrodeposited to the defective regions. In addition to defects, Li metal can also grow from grain boundaries, kinks and chemical inhomogeneities, such as contaminants. Furthermore, they proved that dendritic growth is based on the defect-controlled deposition mechanism, challenging the conventional mechanism based on ion transport^[49]. Moreover, Shen et al. studied the impacts of current density, magnetic field and electromigration on dendritic Li growth via in situ OM observation^[50-52]. Although in situ OM has shown its effectiveness in observing dendrite growth and battery failure, it still possesses some drawbacks, the most important of which are the low resolution and poor image clarity in complicated electrolyte circumstances^[53].

Due to higher spatial resolution than in situ OM, in situ scanning electron microscopy (SEM) has been widely applied to battery research. Krauskopf et al. investigated the Li growth kinetics on a garnet-type solid-state electrolyte (SSE) at the microscale via in situ SEM^[54]. Motoyama et al. used in situ SEM to observe the plating and stripping behavior of Li metal in a LiPON electrolyte^[55]. As shown in Figure 4A, a Li island grows laterally over time without damaging the wolfram film. During the Li dissolution process, the dome top of the island first begins to shrink and then the lateral bulges reduce gradually. In situ electrochemical SEM (EC-SEM) provides a powerful tool to reveal the reaction mechanism in electrochemical systems. For instance, Rong et al. found that the addition of single LiNO₃ in a LiTFSI-DOL/DME electrolyte could hardly inhibit the Li dendrite formation and growth via EC-SEM^[56].

Figure 4. (A) Sequential SEM images during Li electrodeposition/dissolution processes based on a wolfram film with a thickness of 90 nm and a LiPON electrolyte at 50 μA cm^-2. (B) Schematic illustration of liquid confining cell. Reprinted with permission from Ref.^[59]. Copyright (2017) Elsevier. (C) Preserving and stabilizing Li by cryo-EM. (D) Atomic-resolution TEM images of Li metal and SEI. Reprinted with permission from Ref.^[62]. Copyright (2017) American Association for the Advancement of Science. (E) AFM (two-dimensional, 2D) images of buckled surface topography from three RTIL electrolytes: no-salt P₁₄ TFSI; 1.0 M Li TFSI in P₁₄ TFSI; 1.0 M Li FSI in P₁₄ TFSI. Reprinted with permission from Ref.^[66]. Copyright (2018) American Chemical Society. (F) SECM cell for battery electrodes including photograph, cross-sectional schematic, reactions at the microelectrode (ME) and sample during imaging (ME radius and ME-sample distance are enlarged for clarity). (G) Sequence of SECM feedback images recorded at different charge/discharge processes. Reprinted with permission from Ref.^[67]. Copyright (2020) Wiley-VCH. (H) Sketch of X-ray tomography setup. Reprinted with permission from Ref.^[70]. Copyright (2013) American Association for the Advancement of Science. (I) X-ray tomography slices showing cross sections of symmetric lithium cells cycled to various stages and magnified 3D reconstructed volumes of cells. Reprinted with permission from Ref.^[71]. Copyright (2014) Springer Nature.

As an advanced technology, in situ transmission electron microscopy (TEM) has been developed gradually. With high energy electrons as the excitation source, the operation principle of in situ TEM is similar to that of in situ SEM, but its spatial resolution is higher than in situ SEM. Furthermore, TEM can provide chemical and structural information for electrodes at the nanoscale^[57]. Initially, Huang et al. reported an “open” in situ TEM cell^[58]. However, ‘‘open’’ TEM inevitably has its own defects: (1) the electrode is in point contact with the electrolyte, and thus the mode of Li⁺ ion diffusion is quite distinct from the actual cell; (2) the strong electron beams are detrimental to some polymer electrolytes; and (3) some SSEs can cause additional overpotential. A ‘‘closed’’ cell with a chamber was then designed to seal off the liquid electrolyte [Figure 4B] to avoid the effect of the high vacuum on the system^[59]. Moreover, clear dynamic information regarding Li stripping/plating processes can be obtained using high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM)^[60]. Gong et al. employed operando HAADF-STEM to study Li electrodeposition and dissolution, which explained the F-enriched interface formation on Li deposits^[61]. For the past few years, Li et al. upgraded Li dendrite images at the atomic scale via cryogenic-electron microscopy technology (cryo-EM) [Figure 4C and D]^[62]. Wang et al. observed the phase transition of ordered (crystalline) Li and disordered (glassy) Li via cryo-TEM, revealing the transient structural evolution during Li deposition^[63].

In situ atomic force microscopy (AFM) has also been generally applied to obtaining surface information, such as the analysis of the morphology of deposits and exploration of the Young’s modulus, which cannot be detected by other microscopy techniques. Aurbach et al. studied the impact of the electrolyte component on Li deposition behavior at low and high current densities via in situ AFM observation^[64]. Subsequently, Kitta et al. employed in situ AFM combined with adhesion mapping to confirm that fresh Li protrusion possessed a thinner SEI film and lower adhesion might grow rapidly^[65]. Yoon et al. used in situ AFM to detect the plane strain modulus of an SEI that was formed in an ionic liquid at room temperature through a mechanical analysis method^[66]. As exhibited in Figure 4E, under compressive stress, the SEI layer generates surface buckling. According to the calculated results, the strain modulus of the SEI film for all the samples ranges from 1.2 to 2.0 GPa. Based on an electrochemical reaction, scanning electrochemical microscopy (SECM) is a kind of scanning technology that uses a small driving probe to scan samples and obtains corresponding electrochemical information. The interfaces between measured liquid and solid samples, which can be insulators, semiconductors and conductors, are effectively monitored, enabling SECM to timely obtain the current on the electrode. Krueger et al. investigated the current evolution of the selected SEI region via an in situ SECM device [Figure 4F]^[67]. As shown in Figure 4G, it can be found that the gap between the minimum and maximum currents gradually increases with the enhancement of applied current during cycling. The SEI will be increasingly challenged by the heavy current due to the continually changing surface shape of metallic Li in the subsequent cycles; thus, it can be applied to detect the appearance of lithium protrusions. Furthermore, laser scanning confocal microscopy (LSCM) and scanning tunnelling microscopy (STM) have also been exploited to investigate Li metal anodes^[68].

In situ X-ray imaging technologies, such as transmission X-ray microscopy (TXM) and X-ray tomography (XRT), can establish real-time three-dimensional (3D) images of Li deposits without damaging cells. Considering the high energy of X-rays, TXM can be employed to study thick samples. Using in situ TXM, Cheng et al. first revealed the dynamics of dendritic Li growth in 1 M LiPF₆-ethylene carbonate (EC)/diethyl carbonate (DEC)^[69]. Notably, as a brand-new 3D detection technology, XRT is capable of investigating the inner electrode structure without slicing treatment. It photographs electrodes from different angles by rotating samples and then reconstructs 3D images through remodeling [Figure 4H]^[70]. Harry et al. reported that dendritic Li was mainly distributed inside the electrode at the initial stage, i.e., below the interface of the electrode/polymer electrolyte^[71,72]. Crystal impurities and the subsurface structure are the nucleation seeds of Li dendrites. Subsequently, Li branches pierce the electrolyte and reach the cathode side, eventually leading to short circuit of the battery [Figure 4I]. Yu et al. systematically estimated the impact of various electrolyte systems on the plating and stripping processes of Li metal via synchrotron-based X-ray imaging technology^[45]. They proposed that the excessive overpotential generated a higher nucleation density, leading to smaller nuclei during the charge process, which is consistent with the previously reported result^[73].

In situ nuclear magnetic resonance (NMR) is commonly used to monitor the component variation of the electrolyte, electrode and interface. As a noninvasive imaging technique developed from in situ NMR, in situ nuclear magnetic resonance imaging (MRI) can achieve sufficient temporal and spatial resolution of the ⁷Li chemical shift. Ilott et al. reported the utilization of indirect MRI with 3D imaging to investigate the microscopic structure of Li metal, showing that dendritic growth was distorted rather than unidirectional until the short circuit was formed^[74]. Because this method not depending on measuring Li directly, it is also suitable for several metallic dendrites, such as sodium, zinc and magnesium anodes. In situ electron paramagnetic resonance (EPR), in which the unpaired electron spins rather than nuclear spins are present in EPR when applying a magnetic field, is an emerging technique in Li metal anode research. Sathiya et al. first observed uneven Li deposition with a LiPF₆-EC/PC/DMC electrolyte in a Li₂Ru_0.75Sn_0.25O₃/Li battery by in situ EPR^[75]. Wandt et al. found that the formation of microstructured (mossy/dendritic) Li was greatly reduced after the addition of FEC via in situ EPR and ex situ SEM^[76].

Neutron radiography and tomography (NR/NT), whose imaging principles are based on the neutron absorption gradient determined by the permeability and absorption of sample composition and density, are typical 2D and 3D imaging technologies, respectively. Recently, Song et al. investigated the distribution and growth mechanism of Li dendrites by in situ NR/NT techniques^[77]. Combined with electrochemical tests, it was confirmed that the dendritic Li growth between the anode and LiMn₂O₄ contributed to the short circuit. Sun et al. used in situ NT combined with in situ XRT to demonstrate the relationship between the overall electrochemical performance of a Li-O₂ battery and the evolution of Li morphology, which fully revealed that the irreversible conversion of Li metal from the bulk structure to porous state^[78]. This work has advanced the mechanism of battery failure, which contributes to new insights into future battery design.

Chemical composition and bonding analysis of SEI components

A number of methods have been developed to investigate the evolution of the electrode/electrolyte interface during cycling. Li detection and quantification can be acquired by in situ X-ray diffraction (XRD), Raman spectroscopy, Fourier transform infrared (FTIR) spectroscopy, NMR, MRI and neutron depth profiling (NDP) analysis. The composition of the surface reaction products can be revealed by in situ X-ray photoelectron spectroscopy (XPS) and cryo-TEM analysis. The salt and Li-ion concentration gradients can be investigated by holographic interferometry, in situ stimulated Raman scattering (SRS) microscopy and MRI analysis. In situ XRD [Figure 5A] provides a range of signals, such as some metastable information that cannot be obtained through ex situ XRD^[79]. Shen et al. proved the excellent stability of a LiF-modified Li anode in the 1 M LiPF₆/EC/DEC electrolyte by in situ XRD^[80]. As shown in Figure 5B, at the first cycle, the electrolyte in the battery system based on bare Li anode is decomposed to the formation of the SEI film and Li dendrites. However, the diffraction peak change in the battery system based on the LiF-modified Li electrode is negligible, indicating the stable SEI.

Figure 5. (A) Schematic illustration of in situ XRD cell. Reprinted with permission from Ref.^[79]. Copyright (2019) Wiley-VCH. (B) In situ XRD patterns of (top) bare Li anode and (bottom) LiF-Li anode upon first charge-discharge process. Reprinted with permission from Ref.^[80]. Copyright (2019) Springer Nature. (C) Schematic illustration of in situ Raman cell. Reprinted with permission from Ref.^[81]. Copyright (2015) Wiley-VCH. (D) In situ Raman spectra of ether and MIP-based electrolytes, respectively. Reprinted with permission from Ref.^[83]. Copyright (2019) Springer Nature. (E) Schematic diagram of cell used for in situ MRI. Reprinted with permission from Ref.^[90]. Copyright (2015) American Chemical Society. (F) In situ NMR and susceptibility calculations enable to characterize the types of Li microstructures nondestructively in a functioning battery. Reprinted with permission from Ref.^[94]. Copyright (2015) American Chemical Society. (G) Schematic of ⁷Li in situ NMR technique used to study dead Li formation and (H) corresponding ⁷Li NMR spectra. Reprinted with permission from Ref.^[92]. Copyright (2020) American Chemical Society. (I) Experimental setup for in situ NDP measurements. Reprinted with permission from Ref.^[96]. Copyright (2019) Elsevier. (J) Operando NDP measurements of first plating and stripping cycle including the plating and striping activity at (top) 0.5 mAh cm^-2 and (bottom) 2 mAh cm^-2. Reprinted with permission from Ref.^[97]. Copyright (2018) Springer Nature. (K) 2D evolution of Li distribution as a function of charging time based on normalized neutron radiographs. Reprinted with permission from Ref.^[77]. Copyright (2019) American Chemical Society.

In situ Raman measurements [Figure 5C] have advantages over in situ XRD due to their applicability to analyze poorly crystalline and amorphous compounds^[81]. Furthermore, Raman spectroscopy is feasible for studying the electrolyte, electrode and interface. Qiua et al. found that minute amounts of O₂ in Li-O₂ batteries can react rapidly with metallic Li to generate an SEI enriched with Li₂O, Li₂O₂ and LiOH species, which protect the anode from further reacting with salts and solvents^[82]. The inhomogeneity of the interfacial Li-ion concentration is closely related to the dendritic Li growth; hence, significant studies have been devoted to the concentration variation of Li⁺ ions at the electrode/electrolyte interface. For example, by using in situ Raman spectroscopy measurements, Chen et al. found a Raman band at 741.8 cm^-1, which was attributed to the S-N stretching and coupled CF₃ bending in TFSI^-, proving the uniform distribution of Li⁺ ions in the MIP-based electrolyte [Figure 5D]^[83]. In addition, as a non-linear Raman technique, SRS has been successfully applied to investigate the variation of the interfacial Li⁺ ion concentration, realizing the 3D visualization of ion depletion and the corresponding dendrite growth^[84]. Due to the high measurement accuracy, rapid measurement speed, high resolution and detection sensitivity, in situ FTIR spectroscopy has been used to track electrolyte decomposition and SEI construction. According to in situ FTIR results, Hu et al. discovered that nitrile compounds are attractive electrolyte candidates with wide electrochemical windows, outstanding thermal stability and high safety at high voltages^[85,86]. However, nitrile compounds are highly corrosive to Li metal. XPS, which is usually used to characterize the electronic structure of surface elements and local chemical environments, is an effective means for investigating solid surfaces. Schwöbel et al. first reported the SEI layer between metallic Li and LiPON using in situ XPS^[87]. A more fundamental study of the reaction of the active metallic Li with the atmosphere was carried out by exposing Li to CO₂ by Etxebarria et al.^[88]. Yan et al. investigated the influence of the FEC on the SEI constituents by sputtering XPS, via which it was revealed that the SEI consisted of a dense two-layer structure with the bottom layer composed of inorganic species, such as Li₂CO₃ and LiF, and the bottom layer composed of organic species, such as ROCO₂Li and ROLi^[89].

In situ NMR can be used to realize the quantitative evaluation of various Li morphologies because the radiofrequency can completely penetrate the Li microstructure [Figure 5E]^[90-92]. According to the reported results, the position variation of the Li microstructure could be detected by the changed signal^[93]. The resonance displacement of NMR shifts with various metallic Li structures, such as the characteristic peaks of dendritic Li and mossy Li at 270 and 261 ppm, respectively [Figure 5F]^[94]. In addition, the intensity of NMR varies with the amount of Li deposits. This difference in peak position and intensity enables the deposited Li to be quantitatively evaluated. For instance, Gunnarsdóttir et al. calculated the amount of ‘‘dead Li’’ from the NMR spectra in Figure 5G and H^[92]. In addition, 3D matrix images on the timescale of an electrochemical reaction can be generated by specific chemical shift points, which allows in situ MRI to detect the Li-ion concentration variation during electrochemical processes in real time. Romanenko et al. conducted a quantitative MRI study on solid-state ionic plastic crystal electrolyte-based batteries^[95]. Li-ion transfer into the electrode matrix during the ongoing discharge of the anode results in partial liquefaction of the electrolyte at the metal interface. Displacement of the liquefaction front is accompanied by faster Li-ion transfer through grain boundaries at the electrode. The arrival of Li⁺ ions then leads the electrolyte to be partially liquefied at the electrode interface, leading to an enhancement in the charge transmission.

NDP, which is based on low energy neutrons, can quantify the concentration of Li⁺ ions directly. In situ NDP, as a nondestructive analysis tool, provides a feasible method to determine the electrochemical reaction sites and electromigration and kinetic processes of the metal anode. Li et al. exhibited the Li electrodeposition behavior in a Li-Li_6.4La₃Zr_1.4Ta_0.6O₁₂ (LLZTO)-Ti solid-state battery (with a 3D Ti electrode) by in situ NDP (Figure 5I)^[96]. Lv et al. investigated the spatial heterogeneity of Li metal during deposition and dissolution processes with NDP [Figure 5J]^[97]. They demonstrated the impact of high current density in reducing the Li distribution density for the first time. It was found that Li integration under low current density (0.5 mA cm^-2) and high current density (2 mA cm^-2) was significantly different, indicating the effect of initial electroplating current density on the SEI and subsequent cycles. Recently, Song et al. used in situ NR and NT to study Li distribution dynamics^[77]. As shown in Figure 5K, Li consumption increased with time during the charge process. Further studies confirmed that high current densities result in a high density of Li deposits, while low current densities generate a low density of Li deposits and thick Li deposition, thereby forming significant “dead Li”. Titration gas chromatography (GC) techniques can be used to quantify the fraction of dead Li^[98]. Recently, Hsieh et al. estimated the amount of dead Li by combining GC with in situ⁷Li NMR analytical techniques^[99]. Based on in situ holographic interferometry, in which a variation in the refractive index caused by a concentration variation of chemical species leads to a discrepancy in the optical path length between the reference and objective beams, Ota et al. proposed the ion concentration-dependent basal growth mechanism^[100]. The results showed that the concentration gradient of Li⁺ ions around the tip and arm of Li dendrites was steeper than that at the electrode/electrolyte interface, indicating that the local current density converged here^[101-103].

Anode structural design

Inhomogeneous Li nucleation behavior due to uneven distribution of electrons and Li⁺ ions can trigger fatal Li dendrite growth. On this basis, various host strategies have been reported to homogenize the interface electric field, distribute the local current density and uniformize the Li⁺ flux. Furthermore, a stable host minimizes the volume variation to avoid stress fluctuations. The hosts usually fall into three categories: electron-conductive (EC) frameworks; ionic-conductive (IC) frameworks; mixed ion and electron-conductive (MIEC) frameworks.

3D current collectors (i.e., EC frameworks) with high specific surface areas (SSAs) are conducive to mitigating huge volume changes and reducing the local current density. For instance, 3D Cu current collectors with submicron pores have been reported to solve the dendrite issue directly due to the large pore volumes and high SSAs^[110]. Furthermore, the tip effect of the porous electrode induced the initial deposition of Li metal on the microchannel walls, which constrained dendrite growth. Because the initial Li nucleation sites evolve into bumps distributed on the current collector, whose edges are unable to capture Li ions under the electric field, insufficient Li ions preferentially tend to gather on the tip, which is known as the so-called “tip effect”, leading to dendritic deposition. Yan et al. explored the Li nucleation pattern on a variety of metal substrates and revealed a special growth phenomenon that depends on the substrates [Figure 6A]^[111]. The nucleation overpotentials of Li metal on Au, Ag, Zn and Mg substrates are almost zero, which can be explained by the definite solubility of these metals in Li metal, leading to solid solution buffer layers before the formation of Li metal. Carbon-based host materials can promote the energy density of batteries due to low density. For example, Zuo et al. used a graphitized carbon fiber host for Li anodes, which guarantees high areal capacity and metal plating/stripping efficiency, low voltage hysteresis and long lifespans^[112]. Jin et al. reported a Li restoration method via a biochar capsule to host iodine that can effectively rejuvenate electrochemically inactive Li based on an iodine redox chemistry^[113].

Figure 6. (A) Voltage profiles of Li deposition on Cu and Au substrates at 10 µA cm^-2. Reprinted with permission from Ref.^[111]. Copyright (2016) Springer Nature. (B) Summary of calculated binding energies between heteroatom-doped carbon and a Li atom. Reprinted with permission from Ref.^[114]. Copyright (2019) American Association for the Advancement of Science. (C) Schematic of dual functionality of an electron-deficient carbon surface. Reprinted with permission from Ref.^[115]. Copyright (2021) Springer Nature. (D) Schematic diagram of Li deposition on lithiophilic-lithiophobic gradient layer-coated Li foils. Reprinted with permission from Ref.^[117]. Copyright (2018) Springer Nature. (E) Electrokinetic phenomena in 3D PPS under an electric field. Reprinted with permission from Ref.^[124]. Copyright (2018) Springer Nature. (F) Schematic diagrams of different Li metal anode host structures. Reprinted with permission from Ref.^[35]. Copyright (2019) Cell Press.

To effectively achieve the inhibition of dendrites, it is critical to adjust the charge transfer process and regulate the initial nucleation sites. Heteroatomic doping in a carbon matrix is one of the most effective methods to guide the initial Li nucleation process. Specifically, 20 carbonaceous-based species (pristine and heteroatom doped) were modeled with various dopants^[114]. As shown in Figure 6B, among the various single-doped carbon matrices, aO-carbon matrices exhibit the largest binding energy toward Li atoms, thus significantly decreasing the Li nucleation overpotential. In addition to heteroatom doping, a multivacancy defect-enriched carbon matrix was presented as the anode current collector [Figure 6C]^[115]. Furthermore, the modification of metallic oxides, such as ZnO, Co₃O₄, Cu₂O, TiO₂ and other polar particles, on carbon matrices is also an effective strategy to induce the migration and diffusion of Li⁺ ions to modify the nucleation and deposition of Li metal^[116]. Zhang et al. demonstrated a promising strategy based on a lithiophilic-lithiophobic gradient layer in which the bottom lithiophilic ZnO layer tightly anchors the whole layer onto the Li foil to facilitate a stable SEI formation and the top lithiophobic carbon nanotube sublayer can effectively suppress Li corrosion [Figure 6D]^[117].

It is notable that in a matrix host material with high electronic conductivity, Li deposition could undesirably take place outside the host, especially at high current densities. To resolve the problem, Yang et al. proposed a 3D garnet-type Li₇La_2.75Ca_0.25-Zr_1.75Nb_0.25O₁₂ framework as an IC Li host to achieve bottom-up deposition^[118]. The structure with a planar Cu current collector guarantees the bottom plating of Li deposits, while the dense garnet-type layer in the middle blocks potential dendritic growth, which prevents safety risks caused by short circuit. It is noted that the features of high ionic conductivity (3 × 10^-4 S cm^-1) and single-ion conductive feature of the garnet electrode also favors uniform and stable Li deposition. The unique porous garnet structure also enlarges interfacial contact between the electrode and electrolyte, facilitating the rapid acquisition of sufficient Li⁺ ions and reducing interfacial resistance significantly. Hitz et al. reported that Li symmetric cells based on doped Li₇La₃Zr₂O₁₂ ceramic Li-conductor electrodes are able to cycle at an ultrahigh current density of 10 mA cm^-2 with an ultralow interfacial resistance of 7 Ω at room temperature^[119].

Apart from the reduced local current density and mitigated Li-ion depletion via optimized 3D EC and IC scaffolds, the stability of Li metal electrodes at large current densities is a complicated dynamic issue. Both electron and ion transport need to be strengthened under high rates^[120]. Based on previous knowledge, different electron/ion transport rates are required for Li plating and stripping processes during high current density operation. During the Li plating process, the fast electron transport is beneficial for the efficient distribution of electrons on the electrode surface, which induces a homogenous distribution of the local current density, leading to the inhabitation of the dendrite propagation. In contrast, during the Li stripping process, the rapid transport of Li⁺ ions seems to be more desirable because rapid Li⁺ departure and homogenized Li⁺ flux can accelerate and stabilize its dissolution according to the principle of chemical equilibrium. Therefore, the rational design of 3D MIEC frameworks is considered as a promising solution and has attracted significant attention. Considering that Li anodes with excess Li capacity are needed to offset irreversible Li loss during cycling in full batteries due to continuous parasitic reactions, it is a rational strategy for employing MIEC anodes through molten infusion technique on a large scale. Guo et al. reported that molten Li could be successfully absorbed into the channels of carbon wood modified with ZnO nanoparticles, forming a Li/carbon wood composite anode that enabled a homogeneous Li⁺ flux behavior^[121]. More recently, Luo and co-workers proposed a 3D MXene aerogel MIEC scaffold for a Li metal anode with excellent electrochemical performance under high current density^[122]. Represented by Ti₃C₂, such MXene aerogels have both high electronic conductivity and fast ionic transport capability due to the abundant lithiophilic oxygen- and fluorine-containing functional groups.

In addition, some IC frameworks, such as Li_6.4La₃Zr₂Al_0.2O₁₂ (LLZO), were incorporated into the 3D EC framework of carbon nanofibers via an electrospinning process, followed by heat treatment, fabricating a 3D MIEC composite framework^[123]. The outstanding performance of this framework could be attributed to the improved electrolyte wettability caused by decreased interface energy and homogenized Li⁺ flux to compensate Li⁺ depletion in the vicinity of the electrode. Li et al. developed a 3D crosslinked porous polyethylenimine sponge (PPS), which was capable of providing strong Li-ion affinity and promoting electrokinetic phenomena [Figure 6E]^[124]. The lithiophilicity of the PPS concentrates Li⁺ ions in the sponge, leading to a higher local concentration of Li⁺ ions at the electrode/electrolyte interface compared to that in the bulk electrolyte. Chen et al. reported that a prelithiated Li_xSi anode can effectively suppress the formation of dead Li when the total lithiation/delithiation capacity in each cycle is higher than the Li plating/stripping capacity on Li_xSi because the slight delithiation from Li_xSi shows a higher potential than the stripping of deposited Li from Li_xSi [Figure 6F]^[35].

Fabrication of artificial interfaces

Fabricating artificial SEIs (ASEIs) is one of the most attractive solutions to overcome the shortcomings of spontaneously formed interphases. Due to multiple degrees of freedom, ASEIs with unique properties and compositions created outside an electrochemical cell can be precisely tuned. A stable and homogeneous ASEI with a specific composition on the electrode surface is favorable for the definition and characterization of structure-property relationships. Ideally, as shown in Figure 7A, an ideal ASEI must meet several intrinsic requirements: (1) chemically and electrochemically stable to avoid parasitic reactions with the electrolyte; (2) mechanically flexible and robust to withstand the volume change and retard dendrite growth during repeated cycles; and (3) uniform and abundant ion conductive pathways to facilitate facile single Li-ion diffusion^[125,126].

Figure 7. (A) Schematic illustrations of Li deposition at various interfaces. Reprinted with permission from Ref.^[125]. Copyright (2021) Springer Nature. (B) Polyionic liquid coating on Li metal anode. Reprinted with permission from Ref.^[132]. Copyright (2020) American Chemical Society. (C) Stability comparison of PEO-coated Li and bare Li in water. Reprinted with permission from Ref.^[133]. Copyright (2020) Elsevier. (D) Schematic illustrations of mitigated surface concentration during concentration polarization with an artificial SEI with a high dielectric constant. Reprinted with permission from Ref.^[135]. Copyright (2021) American Chemical Society. (E) Schematic of half-blocked coin cell before charging and after first charging. Reprinted with permission from Ref.^[143]. Copyright (2018) National Academy of Sciences. (F) MOF-based SSI formed on LMA with immobilized ionic channels for fast Li⁺ transport. Reprinted with permission from Ref.^[148]. Copyright (2020) Cell Press. (G) Uniform and stable ASEI on Cu. Reprinted with permission from Ref.^[150]. Copyright (2020) Springer Nature. (H) XPS depth profiles of Li 1s with three stages in regions I (red), II (green) and III (blue), with the detection depth of XPS covering only the LiF layer, LiF+Li and Li, respectively. Reprinted with permission from Ref.^[151]. Copyright (2017) American Chemical Society.

An ideal ASEI must be stable to the Li metal anode, preventing electrolyte component decomposition apart from the reduction of Li⁺ ions^[127]. Polymer networks have shown the ability to stabilize Li metal anodes and block electrolyte penetration to some extent. In addition, ASEIs are attractive due to the polymer networks, such as polyethylene oxide^[128], Nafion^[129] and polyvinylidene fluoride^[130], having a maximum limit in swelling due to the balance between the entropy of mixing and the entropy of the polymer chain configuration^[131]. Huang et al. integrated tether cations onto a polymer backbone to form a polyionic liquid coating that is chemically stable even at ultranegative potentials below Li reduction (< -3.04 V vs. SHE), resulting in an improved Li deposition morphology, as well as superior cycling performance of LMBs [Figure 7B]^[132]. A Li metal electrode featured with an air-stable and waterproof surface using a PEO-based composite ASEI film was also prepared by Yang and co-workers for Li-S battery applications [Figure 7C]^[133].

The mechanical modulus of the interphase can be tuned by modulating the physical structure and chemical compositions of the ASEI. During the electrodeposition process, mechanically unstable SEIs are prone to fracture and allow dendritic penetration when confronted with large amounts of stress and strain, leading to the generation of hotspots in which uneven diffusion of Li⁺ ions aggravates dendritic Li growth^[134]. Even without dendrites, the mechanical degradation of the SEI will expose fresh Li to solution, leading to the consumption of both electrolyte and Li metal. Therefore, the degradation of mechanical stabilities accelerates the failure of batteries. Tamwattana et al. showed that the considerably high tensile strength and Young’s modulus in α-PVDF are attributable to a dense microstructure and less porosity, which can effectively promote a uniform Li deposition by decreasing the overpotential, lowering the local current density and suppressing Li protrusions [Figure 7D]^[135,136]. A flexible Young’s modulus (1-0 GPa) for an ASEI can be achieved with semi-crystalline and crystalline polymers that can increase the strength of the backbone. In contrast, inorganics, such as ceramics, generally provide a high Young’s modulus (10-100 GPa), which enables resistance to the stress imposed by the growing dendrites (4.2 GPa). Furthermore, Li-ion conductive inorganics, like Li₃N^[137], LiF^[138], Li₂S^[139] and Li₃PO₄^[140], as well as passive inorganics that solely impart mechanical stiffness at the interface, like Al₂O₃^[141] and nano-diamond coatings^[142], are considered as high mechanical modulus candidates.

The heterogeneous transport of Li⁺ ions in the native SEI is due to differences in diffusion rates in different regions, which is the main triggering factor of dendrites. Maintaining fast and uniform Li-ion flux is critical to suppressing the formation of dendrites. Shi et al. revealed the stripping mechanism on a Li anode by visualizing the interface between the stripped Li and the SEI [Figure 7E]^[143]. After the cations migrate through the SEI layer, the metal vacancies will be left on the Li electrode below the SEI layer. The aggregation of metal vacancies can lead to the formation of large cavities at the Li/SEI interface, which is fatal for the adherence of the SEI layer. Tewari et al. revealed that ion depletion easily resulted in a scarcity of Li⁺ ions beneath the SEI, thereby making it a diffusion-controlled reaction. Reaction-controlled rather than diffusion-controlled conditions are favorable for compact and stable deposition^[144]. Lithiated NafionTM was reported as a single ion-conducting polymer to fabricate an ASEI^[145,146]. Crosslinked ionic networks offer the multifunctionality of both electrolyte blocking and single ion conduction^[147]. Xu et al. proposed that coating an ion-transport rectifier on the Li metal anode allows the anions (e.g., PF^6-) to complex with the Lewis acid sites and the solvated cations (Li⁺) to effectively translocate within the pore channels with both high σ_Li+ and t_Li+ [Figure 7F]^[148]. Apart from the diffusivity of ions within the interphase, the surface diffusivity also has a significant influence on the Li deposition morphology. High surface diffusivity is generally enabled by halides, such as LiF, LiBr and LiI^[149]. Gao et al. reported a self-assembled monolayer of electrochemically active molecules that regulates the composition and nanostructure of the SEI [Figure 7G]^[150]. A multilayer SEI that contains an amorphous outer layer and a LiF-rich inner phase seals the Li anode effectively. Lin et al. also developed a conformal LiF coating technique on Li surface with commercial Freon R134a as the reagent [Figure 7H]^[151].

Electrolyte modification

In general, the interaction between electrolyte solvent molecules plays a critical role in deciding the bulk properties of an electrolyte, such as the dielectric constant, boiling and melting points and viscosity, which is largely driven by van der Waals forces and tends to be much weaker than the other interactions among the solvent, cations and anions. As shown in Figure 8A, the strategies of electrolyte modification can be divided into three classes based on the regulation of the interactions in the electrolyte system: (1) regulating the interaction between the cations and solvent to modulate the redox stability of solvent molecules; (2) regulating the interaction between cations and anions to optimize the formation of solvent separated ion pairs (SSIPs), contact ion pairs (CIPs) and aggregates (AGGs) and salt solubility; and (3) regulating the interaction between the anions and solvent to adjust the anion solvation behavior^[152].

Figure 8. (A) Schematic of interaction relationship in electrolytes. Reprinted with permission from Ref.^[152]. Copyright (2020) American Chemical Society. (B) Schematic illustrations of interfacial stability enabled by the HOMO-LUMO gap of SEI and stable SEI treated chemically by fluorination. Reprinted with permission from Ref.^[156]. Copyright (2019) American Chemical Society. (C) Schematic illustration of fabrication and working principle of LNO-SRF. Reprinted with permission from Ref.^[158]. Copyright (2018) Springer Nature. (D) Schematic of morphologies of Li deposited on substrate in different electrolytes. Reprinted with permission from Ref.^[160]. Copyright (2015) Springer Nature. (E) Schematic illustrations of passivation films with 1 M and 3 M LiPF₆ in a mixed solvent of EC-EMC-DMC (1:1:1 by vol.). Reprinted with permission from Ref.^[163]. Copyright (2020) Springer Nature. (F) Schematic illustrations of effect of NL cosolvent on solvation structures and thereafter the SEI formation process. Reprinted with permission from Ref.^[164]. Copyright (2020) Wiley-VCH. (G) Optimized geometrical structures of (left) TPFPB-BF^4- and (right) TPFPB-PF^6- complexes. (H) Typical mass spectra via electrospray ionization and (inset) 3D isosurface rendering of F in PF^6- around DMC and EC molecules. Reprinted with permission from Ref.^[165]. Copyright 2015 American Chemical Society.

In electrolytes, especially in non-aqueous systems, Li⁺ ions are generally solvated by the surrounding solvent molecules. Thus, the cation-solvent interactions have an important impact on guiding the cation solvation/desolvation and transport behavior. For example, the Li⁺ solvation structure can be intuitively tuned by adjusting the electrolyte compositions. The components featuring a high Gutmann donor number are more favorable to be recruited into the inner solvation shell. Frontier molecular theory can be employed to uncover the redox stability of the modified solvation shell. The coordinated Li⁺ ions attract electrons from the solvent molecules and correspondingly lower their LUMO^[153]. In the electrolyte, the solvent molecules coordinated with Li⁺ ions can obtain electrons much more easily from the anode than free solvent molecules, leading to electrolyte decomposition. Therefore, according to the different affinities of solvents toward Li⁺ ions, the chemical and electrochemical stabilities of electrolytes can be regulated and designed. For example, fluoroethylene carbonate (FEC) possesses a lower LUMO energy level than DEC and EC and can more easily obtain electrons from the electrode due to the strong electron-withdrawing effect of the F functional group^[154]. Consequently, FEC preferentially reacts with the Li anode to induce a LiF-rich SEI. Zhang et al. created a highly stable organic interphase with a well-tuned LUMO energy to improve the anti-reduction ability of SEI components and enhance the long-term cyclability of LMBs^[155,156]. Introducing trifluoromethyl functional groups (-CF₃) to the molecular structure in the SEI can significantly tune the orbital energies and the HOMO/LUMO gap due to the strong electron-withdrawing property of -CF₃ functional groups [Figure 8B]^[155,156].

In contrast to aqueous electrolytes, where anions and cations are completely separated by the solvent molecules, organic solvents tend to possess much lower dielectric constants than water and the cation-anion interaction cannot be weakened effectively. Consequently, Li salts cannot be dissociated completely, and therefore the Li⁺-anion interaction significantly affects the coordination of SSIP, CIP and AGGs, further regulating the features of the SEI components. In commercial dilute electrolytes, Li⁺ ions are generally solvated by the electrolyte solvent molecules with the SSIP structure, leading to a solvent-derived SEI. To avoid this structure, many anion additives, like LiNO₃, have been widely developed as effective electrolyte additives to induce CIP structures for fabricating stable anion-derived SEI and realizing uniform Li deposition^[157-159]. Liu et al. found that the previous knowledge regarding the preferential reduction of nitrate anions during SEI formation in ether-based electrolytes can be extended to carbonate-based systems despite their extremely limited solubility, which has a positive effect on the ion transport and charge transfer kinetics, dramatically altering the nuclei from dendritic to spherical [Figure 8C]^[158]. Li et al. introduced both Li₂S₈ and LiNO₃ additives to an ether-based electrolyte, which enabled a synergistic effect on the Li surface that can effectively minimize the electrolyte decomposition and prevent dendrites from propagating [Figure 8D]^[160].

More effectively, the concept of high-concentration electrolytes (HCEs) has been widely accepted^[161]. With increasing salt concentration, the number of free solvent molecules decreases, leading to an increase in the proportion of anions in the solvation shells. The involved coordination of anions transforms the Li⁺ solvation structure from SSIP to CIP and AGGs, which induces an inorganic-rich SEI and stabilizes Li metal anodes^[90,162]. Liu et al. found that both a favorable SEI and cathode electrolyte interphase (CEI) can be formed in HCEs [Figure 8E]^[163]. However, an ultrahigh concentration of Li salt commonly sacrifices the ionic conductivity of electrolytes, leading to batteries with high costs. To maintain the favorable solvation structure, non-solvating fluorinated ethers with weak interactions with cations have been innovatively adopted to decorate the electrolyte, which decreases the apparent concentration of the lithium salt and retains the same solvation structure as that in the HCEs. Therefore, local HCEs that function similarly to HCEs but with reduced viscosity, improved ionic conductivity and reduced costs have been reported. Ding et al. reported that a type of non-solvating and low-dielectric (NL) cosolvent can intrinsically facilitate the anion-cation interaction and thus promote the proportion of AGGs via model investigations [Figure 8F]^[164].

Unlike both nucleophilic and electrophilic aqueous solvents, the electrolyte solvents employed in LMB systems are nucleophilic to prevent active protonation reactions from occurring. Furthermore, Li salts possessing anions with large ionic radii are generally adopted to guarantee ideal solubility in organic solvents with small dielectric constants. The delocalized charges and large radii of the anions both weaken the interaction between cations and solvent molecules. As a result, in most cases, the anions in electrolytes are conventionally considered to be unsolvated and the solvation of anions is rarely investigated. Nevertheless, the solvation behavior of anions inevitably influences the stability of anions and solvent molecules, the anion diffusivity and the transference number of cations. Furthermore, solvation might achieve the stability of the radicals in the electrolyte, leading to the inhibition of parasitic interfacial reactions. Solvents with a high acceptor number (AN) are expected to interact strongly with anions in the electrolyte. For instance, tris(pentafluorophenyl)borane (TPFPB) with the electron-deficient B element acting as the high AN solvent could strongly interact with PF^6- and BF^4-, with free energy variations of -12.7 and -24.4 kcal mol^-1, respectively [Figure 8G]^[165]. These free energy variations are significantly more obvious than those between routine ester solvents and anions, which are mostly achieved by H-F interactions [Figure 8H]. In addition, cyclic carbonates rather than linear counterparts are favorable due to the greater polarity with the ring constraint. The effect of anions on the solvent molecule stability is contrary to that of cations, which improves the reductive stability and reduces the oxidative stability. In addition, the anion-solvent complexes lower the mobility of anions, thus increasing the transference number of the cations, which is conducive to the electrochemical rate performance. Furthermore, the anion solvation promotes salt dissociation and increases the solubility of the salt.

Replacing flammable liquid electrolytes with non-flammable SSEs is a feasible method to improve the safety of batteries. Current SSEs can be divided into three categories: polymer SSEs; inorganic ceramic SSEs; polymer/ceramic composite SSEs. Polymer electrolytes exhibit better flexibility and thus better interfacial contact with the Li electrode. However, their relatively low ionic conductivity and narrow electrochemical window limit their application in high-voltage batteries. Ceramic electrolytes usually possess desirably high moduli for dendrite suppression and high ionic conductivity. However, most inorganic ceramics are fragile, which restricts their application in mass production. Therefore, polymer/ceramic composite electrolytes are promising for achieving high conductivity and proper flexibility in practical LMBs. Because of the high ionic conductivity and thermal stability of PEO, it is frequently used to fabricate polymer/ceramic composite electrolytes. PEO-in-ceramic electrolytes (80% ceramic)^[166], polyvinylidenfluoride-hexafluoropropylene/Li_1.3Al_0.3Ti_1.7(PO₄)₃^[167,168], Li₇P₃S₁₁-PEO-LiClO₄^[169], Li_0.33La_0.557TiO₃ nanofiber-enhanced PEO^[170] and Li_6.4La₃Zr_1.4Ta_0.6O₁₂/PEO/succinonitrile^[171] have been successfully used as novel composite electrolytes.

Separator modification

The most direct and simple approach to protect LMBs from short circuit is blocking the growth of Li dendrites. However, conventional PP- and PE-derived separators can be easily pierced by sharp dendritic Li^[172]. Stiff coating layers can effectively enhance the penetrating resistance of the separators. Liang et al. reported that curved surfaces are more favorable for dendrite suppression due to the fact that they can resist puncture from the sharp tips of Li dendrites by redistributing the interfacial stress^[173]. Based on this perspective, they proposed a separator with the coating layer of a nano-shield, like SiO₂ spheres, which can minimize the piercing force of Li dendrites and significantly improve the cycling life of LMBs. The nano-shield protection theory is applicable to polystyrene sphere-coated separators, which can extend the discharging lifespan of Li/Li symmetric cells to 70 h. Based on the same principle, Al₂O₃^[174]- and AlN^[175]-coated separators were further used to suppress dendrite growth by enhancing the mechanical barrier. In addition to inorganic coating materials, integrated separators consisting of polymers with high strength are another strategy for the suppression of dendritic growth. For example, Hao et al. prepared a poly(p-phenylene benzobisoxazole) PBO nanofiber membrane-based separator with high strength via a highly scalable blade casting process^[176]. These membranes possess a strength of 525 MPa, a Young’s modulus of 20 GPa, a thermal stability of up to 600 °C and impressively low ionic resistance, thereby enabling the inhibition of dendrites in electrochemical cells^[176].

The irregular pores and skeletal structures of conventional separators lead to an inhomogeneous flux of Li⁺ ions due to the crowded Li⁺ ions in the pores during transport. However, the Li-ion concentration in the vicinity of the skeletons remains unchanged. Under this condition, while the Li⁺ ions reach the anode surface and obtain electrons to be reduced, the nonuniform Li deposition and formation of dendritic Li occurs. Therefore, the redistribution of Li-ion flux is crucial to inhibit the dendritic formation of LMBs. For example, ceramic and gel electrolytes with rapid Li-ion conductivity are generally employed in functional separators for balancing the Li-ion distribution^[177,178]. Compared with separators with electron-conducting coating layers, Li-ion conductor-modified separators can avoid Li reduction in the coating layer due to the fact that Li-ion conductors are electrically insulated. Furthermore, these separators possess remarkable flexibility in composition and structure for ion redistribution compared to commercial separators. Ma et al. proposed a 3D-ordered microporous separator using regenerated eggshell^[179]. Cations can transfer in 3D, leading to superior electrochemical performances compared to batteries with conventional separators. Similarly, Zhao et al. proposed that LLZTO ceramic modulated PP separators can be used as rectifiers to redistribute Li⁺ ions for dendrite-free LMBs [Figure 9A]^[180]. The separator with a LLZTO coating layer achieved the uniformity of ion concentration [Figure 9B]. Liu et al. decorated conventional Celgard separators with metallic Mg nanoparticles [Figure 9C]^[181]. Due to lithiophilic Mg nanoparticles, Li⁺ ions prefer to homogenously electroplate on the separator surface, which faces the anode side. Hao et al. developed MOFs coated on a functionalized PP separator to regulate ion transport [Figure 9D]^[182]. The well-defined intrinsic nanochannels in MOFs and the negatively charged channels both restrict the free migration of anions, contributing to a high Li⁺ transference number of 0.68. Sheng et al. reported the effective suppression of electrolyte-Li metal reactivity through a nanoporous separator [Figure 9E]^[183]. Calculations assisted by diversified characterization reveal that the separator partially desolvates Li⁺ in space created by its uniform nanopores and deactivates solvents before Li deposition occurs. The surface contact gaps between the Zr-MOCN@PP and Li electrode are significantly mitigated compared to those between the UiO-66@PP separator and Li electrode [Figure 9F]. Li et al. reported a polyacrylamide-grafted graphene oxide molecular brush-modified PP separator [Figure 9G]^[184]. Furthermore, Liu et al. developed an approach to immobilize Li⁺ ions by coating the separator with functionalized nanocarbon (FNC), leading to Li dendrites growing toward each other simultaneously from both the FNC layer on the separator and the Li metal anode, which changes the Li growth direction [Figure 9H]^[185].

Figure 9. (A) Schematic illustration of Li deposition behavior with PP separator and LLZTO-coated separator. (B) Simulation of Li-ion distribution through LLZTO-coated separator. Reprinted with permission from Ref.^[180]. Copyright (2018) American Association for the Advancement of Science. (C) Schematic illustration of magnetron sputtering system for preparation of Mg-coated separator. Reprinted with permission from Ref.^[181]. Copyright (2019) Elsevier. (D) Schematic illustration of MOFs@PP separator regulating the transport of Li⁺ and anions in LMBs. Reprinted with permission from Ref.^[182]. Copyright (2021) Wiley-VCH. (E) Fabrication of Zr-MOCN@PP membrane via photopolymerization and corresponding SEM images. (F) Schematic illustration of Li deposition in cells with different separators: (top) UiO-66@PP; (bottom) Zr-MOCN@PP. Reprinted with permission from Ref.^[183]. Copyright (2022) Springer Nature. (G) Schematic illustration of Li deposition with GO-g-PAM-modified separator. Reprinted with permission from Ref.^[184]. Copyright (2017) Springer Nature. (H) A blank cell with (top) pristine separator and (bottom) FNC cell using FNC-coated separator. Reprinted with permission from Ref.^[185]. Copyright (2014) Springer Nature. (I) Schematic diagrams of Li dendrite detection in (left) routine and (right) dual-layer separator-based cell. Reprinted with permission from Ref.^[146]. Copyright (2017) Wiley-VCH.

Generally, it is difficult to detect Li dendrites before the eventual short circuit and battery failure. However, before the battery short circuit, dendritic Li must first reach the separator, which can act as an alarm to detect Li dendrites. Thus, a functional separator was designed with a sensing terminal to realize dendrite detection [Figure 9I]^[146]. The separator evolved into a triple-layer configuration of polymer-metal-polymer. The voltage gap between the anode and the sandwiched metal can be identified by the electrochemical potential difference. During the charge process, once Li dendrites appeared and expanded until they eventually reached the sensing metal layer, a sharp voltage drop can be effectively recorded to indicate dendritic Li growth. Once this is detected, the working battery is terminated to avoid the final short circuit, battery thermal runaway, possible fire circumstance and even potential explosion. Apart from the detection of dendrites, the sensing layer can even eliminate Li dendrites via incorporation with the Li metal reactive materials. It is reported that a nanosized Si layer sandwiched between two layers of conventional polyolefin separators can ensure the high safety of LMBs^[186].

These strategies to suppress dendrite growth in LMBs, such as anode structural design, artificial SEI layers and electrolyte and separator modification, are summarized in Table 1. Each one of them seems weak for practical applications when facing tough and indomitable Li metal anodes. Mutual combination and integration will be strong to face the challenges associated with Li metal anodes. For example, structured Li metal anodes have more advantages in suppressing dendrite growth in the nucleation stage. If this strategy is matched with optimized electrolytes (or other artificial SEI layer), the composite anode can hopefully reduce the side reactions of Li metal, leading to high Coulombic efficiency with a dendrite-free Li depositing morphology. In addition, the anode matrix can be designed to tightly contact the solid-state electrolyte to reduce the interfacial impedance. Therefore, safe and high rate LMBs can be obtained.