Constructing nitrided interfaces for stabilizing Li metal electrodes in liquid electrolytes

Traditional Li ion batteries based on intercalation-type anodes have been approaching their theoretical limitations in energy density. Replacing the traditional anode with metallic Li has been regarded as the ultimate strategy to develop next-generation high-energy-density Li batteries. Unfortunately, the practical application of Li metal batteries has been hindered by Li dendrite growth, unstable Li/electrolyte interfaces, and Li pulverization during battery cycling. Interfacial modification can effectively solve these challenges and nitrided interfaces stand out among other functional layers because of their impressive effects on regulating Li+ flux distribution, facilitating Li+ diffusion through the solid-electrolyte interphase, and passivating the active surface of Li metal electrodes. Although various designs for nitrided interfaces have been put forward in the last few years, there is no paper that specialized in reviewing these advances and discussing prospects. In consideration of this, we make a systematic summary and give our comments based on our understanding. In addition, a comprehensive perspective on the future development of nitrided interfaces and rational Li/electrolyte interface design for Li metal electrodes is included.


Introduction
Rechargeable lithium (Li) ion batteries (LIBs) have been shaping many aspects of our modern life. [1][2][3][4] Nevertheless, the traditional graphite-based LIBs have nearly reached their theoretical limit in energy density ($250 W h kg À1 ), which hinders the development of portable electrical devices and electric vehicles. [5][6][7][8] Li metal has the lowest electrochemical potential (À3.04 V vs. the standard hydrogen electrode (SHE)) among the alkali metals and a much higher theoretical specic capacity of 3860 mA h g À1 (which is 10 times that of graphite) (Fig. 1a). 4,[9][10][11][12][13][14][15] When paired with high-voltage cathode materials, Li metal batteries (LMBs) are able to provide a 5 V-class output voltage and a 500 W h kg À1 -class energy density (Fig. 1a). [16][17][18][19] Therefore, reviving LMBs is an effective strategy to break the performance limitation of LIBs. [20][21][22][23] The main challenge is that all liquid electrolytes are thermodynamically unstable at 0 V vs. Li/Li + , because the lowest unoccupied molecular orbital (LUMO) of the electrolyte is lower than the Fermi level of Li metal (Fig. 1b). 3,24 Thus, the electrolyte accepts electrons from Li metal and reductively decomposes on the surface of the Li electrode to form a solidelectrolyte interphase (SEI). 16,[25][26][27][28] The inner layer of the SEI (close to Li metal) consists of inorganic components such as lithium oxide (Li 2 O), lithium uoride (LiF), and lithium carbonate (Li 2 CO 3 ), while the outer layer of the SEI (close to the electrolyte) mainly consists of organic components such as polyolens and semicarbonates (Fig. 1c). 25,29,30 The SEI layer is electrically non-conductive but ionically conductive, so that it can block the electron transport at the Li/ electrolyte interface and stop the further decomposition of the electrolyte while Li + diffuses through the layer. 16,26 Unlike graphite which stores Li + in its lattice with acceptable volumetric changes ($12%), the Li metal anode accommodates Li + at the Li/electrolyte interface, leading to unlimited volumetric changes during Li plating/stripping processes. 11,12,31,32 Unfortunately, the native SEI formed on Li is brittle, so it fails to tolerate the stress caused by the volumetric changes of Li metal electrodes. 25,[33][34][35] In addition, Li + is preferentially deposited on the protuberant tips with stronger electrical elds on the substrate, leading to the formation and growth of dendritic Li. 26,36 An Li dendrite has a high Young's modulus of $5 GPa, 37 so it can easily pierce the SEI. Once the SEI is damaged, the newly exposed Li would immediately react with the electrolyte to form a new SEI. 16 Meanwhile, the cracked SEI layer may also expose defects and in turn accelerate the deposition of Li on the defects and form new Li dendrites. Furthermore, Li stripping from the roots of the dendrite would break the electrical contact and produce porous "dead" Li. 38 With battery cycling and continuous SEI build-up, the above problems lead to electrolyte depletion, loss of electrochemically active Li, Li electrode pulverization, and battery performance decay (Fig. 1d). 38,39 Even more troubling, the Li dendrites could lead to internal short-circuits or even safety hazards in working batteries (Fig. 1d). 26,38,[40][41][42] Interfacial engineering is critical to stabilize Li metal electrodes. 16,20,43 Constructing nitrided interfaces on the surface of Li electrodes or current collectors has been proved to be effective to suppress Li dendrite formation and growth, as well as protecting Li from electrolyte erosion. The nitrided interfaces can regulate Li + ux distribution near the Li electrodes or current collectors, facilitate Li + diffusion through the SEI, and passivate the reductive surface of Li, thus improving the electrochemical performance of LMBs. In this paper, we review our current fundamental understanding and recent advances in developing nitrided interphases for stabilizing Li metal electrodes. The strategies for constructing nitrided interphases, including building articial SEI layers, electrolyte engineering, substrate modication, and separator functionalization, have been comprehensively summarized and discussed. In addition, our perspective on the future development of nitrided interfaces and rational Li/electrolyte interface design for Li metal electrode is included. The protuberant tips on the substrate (Li or current collector) have stronger electrical elds. Li + is preferentially deposited on these protuberant tips, leading to the formation and growth of dendritic Li. 12,36 Regulating the uniform deposition of Li + is an important step to eliminate the safety risks and performance decay caused by Li dendrites. Nitrogen (N) has lone-pair electrons and can act as a Lewis base site to adsorb positively charged Li + (Lewis acidic site), thus creating a lithiophilic surface on the Li metal electrode and decreasing the overpotential for Li plating. In addition, N has a high electronegativity (c) of 3.04, and when bonded with atoms with lower electronegativity, such as boron (B) (c ¼ 2.04) and carbon (C) (c ¼ 2.55), the electron cloud in N-C or N-B polar covalent bonds will migrate to the N side. The increased charge density around N can further improve the interaction between N and Li + . Therefore, the nitrided interface is able to regulate the Li + ux distribution near the Li electrode surface or current collectors and thus guide Li + uniform deposition.

Advantages of nitrided interfaces
2.1.2. Facilitating Li + diffusion through the SEI. In general, Li + is solvated with four to six solvent molecules in the electrolyte. 44,45 Before plating onto the substrate (the Li anode or current collectors), the solvated Li + is rstly de-solvated near the SEI, and then the naked Li + ions migrate across the SEI. [45][46][47] The migration speed is the rate-determining step in the Li deposition process. 48 The high Li + ionic conductivity of the SEI helps to improve the kinetics of the Li plating process and thus helps to enhance the electrochemical performance of Li metal electrodes. The diffusion mechanism of Li + through the SEI is complicated and controversial. It was proved that LiF, Li 2 O, and Li 2 CO 3 in the native SEI diffuse Li + via grain boundaries, as their intrinsic ionic conductivity is relatively low (up to $10 À9 S cm À1 ). 25,[49][50][51] Nitrides such as lithium nitride (Li 3 N) and LiN x O y have much higher ionic conductivity (up to $10 À3 S cm À1 ), 52-54 and they can provide faster Li + migration channels in the SEI. Therefore, the nitrided SEI can facilitate Li + diffusion and improve the kinetics of the Li plating process.
2.1.3. Passivating the active surface of Li metal electrodes. As mentioned, the formation of the SEI blocks the electron tunneling at the Li/electrolyte interface and thus stops the decomposition of the electrolyte. The thickness of the SEI is related to its electrical conductivity. Nitrides such as Li 3 N, LiN x O y , carbon nitrides (C 3 N 4 ), and nitrided polymers all have an ultralow electrical conductivity. 54 When used to modied the surface of the Li metal electrode, they can physically and electrically isolated Li from the electrolyte and thus passivate the reductive surface of the Li metal electrode, which helps to reduce the thickness of the SEI.

Comparison of nitride interfaces with other strategies
Besides nitride interfaces, metal oxides (such as MgO), 55 phosphates (such as Li 3 PO 4 ), 56,57 some lithium halides (such as LiCl and LiI), 56,58,59 and lithium chalcogenides (such as Li 2 S and Li 2 Se) 60,61 have also been introduced as modication layers on Li metal. Generally, their precursors are hardly soluble in non-aqueous electrolytes, and as a consequence, these layers are usually fabricated by ex situ methods, serving as articial SEI layers. These modication layers do play a positive role in protecting the Li metal, but they may be damaged by the interfacial stress during Li plating/stripping processes and therefore lose their functions. In contrast, in the case of nitride interfaces, some of their precursors (such as nitrates, amides, N-containing ionic liquids, and nitrocellulose) are soluble in certain non-aqueous solvents, which enables continuous repair of nitride interfaces during battery cycling when these precursors are introduced into the electrolyte as solvents or additives.
Fluorinated interfaces, which feature LiF-rich SEI layers, are widely acclaimed for their outstanding effects on Li metal protection, which is based on their high Young's modulus and the high interfacial energy of LiF. 49,62 Although uorinated interfaces are excellent for inhibiting side reactions between Li metal and the electrolyte, nitride interfaces still show advantages over them in some aspects, especially in bulk ionic conductivity. The transport of Li ions in LiF is much more difficult than in Li 3 N or LiN x O y , obviously limiting the grain growth of deposited Li during the plating process. According to the morphologies, deposited Li with a nitrided interface has a larger grain size but smaller microstructural tortuosity compared with Li with a uorinated interface, contributing to higher reversibility of the active Li during battery cycling.

Methods to construct a nitrided SEI on Li metal electrodes
Since the formation of the SEI is a key factor in controlling the surface properties of Li, one of the effective approaches to stabilize the Li metal electrode is to construct functional arti-cial SEI layers on its surfaces. 28,35,43,63,64 According to the preparation mechanism, the strategies to develop a nitrided articial SEI can be divided into physical methods and chemical methods.
3.1.1. Physical methods. Physical pre-coating methods, such as doctor blading, physical pressing, drop coating, atomic layer deposition (ALD), etc., are simple approaches to easily prepare nitrided interfaces on Li metal electrodes. For instance, a polyurea thin layer was coated on Li metal via the ALD method. 65 The abundant N-containing polar groups in the polyurea were believed to be able to redistribute the Li + ux and lead to a uniform plating/stripping process (Fig. 2a). A poly(butylmethacrylate-acrylonitrile-styrene) (P(BMA-AN-St)) cladding was drop coated on the Li surface. 66 Beneting from the affinity of the polar groups (C^N and C]O) in the polymer chains with both Li + and Li metal, the P(BMA-AN-St) cladding provided channels for regulating the Li + (Fig. 2b), 66 so that a dendrite-free surface and improved electrochemical performance of Li metal electrodes were realized, even with deep cycling. Paik et al. modied copper nitride nanowires (Cu 3 N NWs) on Li foil through one-step roll pressing. The Cu 3 N NWs could be conformally printed onto the Li metal and form a Li 3 N@Cu NW layer on the Li electrode (Fig. 2c). 67 Yu et al. synthesized a polar polymer network (PPN) layer and coated it on a Li metal electrode during battery assembly. 68 The C^N groups of polyacrylonitrile polymer chains in the PPN could reduce the high reactivity of the C]O groups of carbonate solvents and promote the decomposition of salt anions (PF 6 À and bis(triuoromethane)sulfonimide (TFSI À )), forming a stable SEI (Fig. 2d). In addition to these articial SEI layers, other different inorganic articial SEIs were also prepared via physical methods on Li metal electrodes. For instance, a Li 3 N layer can be coated on the Li metal electrode via pressing and rubbing Li 3 N powder to suppress Li dendrite formation. 69 Yang et al. coated a layer of acid-treated graphitic (g)-C 3 N 4 on Li, and its N-containing groups were able to rearrange the concentration of Li + and enhance the transfer of Li + . 70 It should be pointed out that the thickness of the nitrided articial SEI developed via physical methods is normally more than a few micrometres (as summarized in Table 1), which would certainly impose a sacrice on the overall volumetric energy density of Li metal electrodes. In addition, the physical methods could not well control the homogeneity of the articial SEI on Li metal electrodes, and the adhesion between the SEI and the Li metal would not be strong enough, which may lead to the exfoliation of the articial SEI during battery cycling. Besides, the organic articial SEI layers modied by physical methods have poor ionic conductivity, so they normally lead to high electrochemical polarization for Li metal electrodes.
3.1.2. Chemical methods. By using chemical reactions between Li and N-containing precursors, more dense and homogeneous articial SEI layers can be prepared. The most common nitrided articial SEI developed by a chemical method is Li 3 N. The rst reported chemical method to develop a Li 3 N layer was using a N 2 gas ow to treat Li in a desiccator. 71 It was proved that an electrochemically stable Li 3 N protective layer had been coated on Li metal by this method. Furthermore, Tu et al. heated Li chips in a tube furnace under a N 2 ow (Fig. 3a), and the formation of Li 3 N on Li metal was conrmed from the X-ray diffraction patterns (Fig. 3b). 72 They revealed that the Li 3 N layer could efficiently prevent contact between Li and the electrolyte and reduce the side reactions. Similarly, Zhou et al. grew a highly [001] oriented, ower-like Li 3 N lm on Li metal by an N 2 plasma activation method. 73 Because of its high Young's modulus and high ionic conductivity, the Li 3 N lm can physically block direct contact between the reactive Li metal and the liquid organic electrolyte. Despite these achievements, the effect of Li 3 N is limited to some extent by its small grain size (<160 nm), which leads to weak interconnections between the Li 3 N particles. To solve this challenge, Cui et al. heated Li metal in a N 2 atmosphere at a high temperature to develop a pinhole-free Li 3 N layer on the Li metal surface (Fig. 3c). 74 The dense, large, and strongly interconnected grains of Li 3 N in the lm reduced the defects in the articial SEI and effectively improved the stability of the Li metal electrode during battery cycling.
Apart from forming Li 3 N, Xie et al. also dropped AgNO 3 / tetrahydrofuran (THF) solution on the Li surface, and the AgNO 3 particles would further react with Li to form LiNO 3 , which is useful for regulating Li + plating behaviour and suppressing Li dendrite growth. A lithium phosphorus oxynitride layer on a Li metal anode with high ionic conductivity and chemical stability was developed via a nitrogen plasma-assisted deposition method to suppress the corrosion from the electrolyte and promote uniform Li plating/stripping. 75 As summarized in Table 1, the nitrided articial SEI layers prepared via chemical methods are generally inorganic, and most of them are thinner as well as having higher ionic conductivity, so they could reduce the polarization of Li metal electrodes. These inorganic articial SEI layers are normally brittle, however, so the integrity of the SEI would be damaged by the interfacial stress changes caused by Li plating/stripping processes, which would shorten the lifespan of Li metal electrodes.

Building nitrided organic-inorganic composite articial SEIs
Building nitrided organic-inorganic composite interfaces is a good idea that takes advantage of the merits of both individual components and overcome their disadvantages. In this regard, Cui et al. developed a reactive interface constructed from Cu 3 N nanoparticles joined together by styrene butadiene rubber (SBR) as an articial SEI for Li metal electrodes. 76 The inorganic Cu 3 N has high ionic conductivity, and the organic SBR has high mechanical strength and high exibility (Fig. 4a). The Cu 3 N further reacted with Li and a composite articial SEI composed of Li 3 N/SBR/Cu was formed on the surface of the Li metal electrode. The Li 3 N particles provided ionically conductive paths, while the SBR conned the Li 3 N particles and buffered the volume changes of the Li anode. Zheng et al. coated a covalent triazine framework (CTF)-LiI hybrid articial SEI on Li by the doctor blade method (Fig. 4b). 77 The N in CTF could bind with Li + from the electrolyte to form Li-N bonds and thus facilitate uniform Li deposition. The uniformly distributed LiI particles could help to improve the mechanical stress to suppress Li dendrite growth. Yu et al. reported a composite articial SEI consisting of an N-containing organic phase (Norganic) and an inorganic Li 3 N phase by utilizing the hyperthermal reduction of Li and g-C 3 N 4 (Fig. 4c). 78 The obtained Norganic phase could link with the Li 3 N phase and form a conformal and compact coating on Li. The authors believed that the C-N]C and N-(C) 3 groups realized the homogeneous distribution of Li + and provided nucleation sites for Li deposition, while the Li 3 N reduced the resistance to Li + transfer across the Li/electrolyte interfaces. Besides, a dual-layer articial SEI was constructed via in situ polymerization of ethyl a-cyanoacrylate (ECA) monomers on the Li metal surface, in which LiNO 3 was introduced with the ECA monomers as an additive (Fig. 4d). 79 The CN À groups in ECA and the LiNO 3 additive reacted with Li to form a nitrided inorganic interface on Li during battery cycling. Poly(ethyl a-cyanoacrylate) (PECA) was used to cover the outer surface to accommodate the volume changes and buffer the interfacial stress during the Li plating/ stripping processes. Xiong et al. modied a self-healing supramolecular copolymer, which consisted of pendant poly(ethylene oxide) (PEO) segments and ureido-pyrimidinone (UPy) quadruple-hydrogen-bonding moieties, on a Li metal electrode via a drop coating method. 80 During the following drying process, the amide and heterocyclic amine groups in PEO-UPy polymer reacted with Li metal and formed a stable articial SEI (LiPEO-UPy) layer on the Li metal electrode. The developed LiPEO-UPy layer could protect the electrolyte from side reactions and homogenize the fast Li + ux to the surface of the Li metal. Lee et al. developed hybrid polyion complex micelles and coated them on Li foil, in which ionized LiNO 3 combined with block copolymer micelles, polystyrene-block-poly(2-vinyl pyridine) (S2VP), via electrostatic interaction (Fig. 4e). 81 It was believed that the S2VP polymer could isolate the active Li from carbonates so as to reduce the side reaction between them, and meanwhile, the introduced LiNO 3 could further dissolve into the electrolyte during battery cycling. As a result, a composite Nrich SEI with a multilayered structure could be formed on the Li electrode (Fig. 4f). With the designed protective layer, Li metal full cells with a high voltage cathode (LiNi 0.8 Co 0.1 Mn 0.1 O 2 ) delivered superior performance, even under harsh test conditions (thin Li anode, high areal-capacity of 4.0 mA h cm À2 , and high current density of 4.0 mA cm À2 ).
Despite the advantages of organic/inorganic composite arti-cial SEIs, it is challenging to control the homogeneous distribution of organic and inorganic phases. To overcome this, our group synthesized a multi-functional [LiNBH] n layer as an articial SEI for Li metal anodes by utilizing a two-step dehydrogenation reaction between Li and ammonia borane (Fig. 5ac), which features the properties of both organic and inorganic SEIs. 82 The obtained ASEI is composed of [LiNBH] n chains, which are cross-linked and self-reinforced by their intermolecular Li-N ionic bonds, and thus give rise to a exible nature (Fig. 5d). Because of the higher charge density of N in the polar [LiNBH] n chain, Li + from the electrolyte will be absorbed by the N to form additional Li-N ionic bonds, which helps to regulate the homogeneous distribution of the Li + ux on Li electrodes (Fig. 5e). In addition, the [LiNBH] n layer is electrically isolated but has high ionic conductivity, thus facilitating Li + diffusion and deposition beneath the articial SEI layer. Therefore, with the protection of the [LiNBH] n layer, Li dendrite growth has been successfully suppressed and a denser and atter surface was achieved aer Li plating/stripping cycles ( Fig. 5f and g).
In a short summary, inorganic nitrided articial SEI layers have high ionic conductivity and relatively low thickness, but they suffer from low integrity and mechanical exibility. Organic nitrided articial SEI layers (normally prepared via physical methods) can regulate the Li + ux distribution and buffer the volume changes of Li metal electrodes, while their poor ionic conductivity and high thickness usually lead to high electrochemical polarization. Building nitrided inorganicorganic composite articial SEI layers is useful to take advantage of the individual components. Most of the composite articial SEI layers delivered enhanced lifespan compared with pure organic or pure inorganic articial SEIs. Unfortunately, it is still difficult to control the homogeneous distribution of organic and inorganic components in the SEI. In addition, the thickness of articial SEI layers varies from a few nanometres to tens of micrometres. To avoid the sacrice of the volumetric energy density, the thickness of the articial SEI should be lower than one micrometre, especially considering that the anodes in practical LMBs are thin foils with a thickness of <50 mm. Another big challenge to articial SEI layers is their stability during battery cycling. With continuous Li plating/ stripping cycles, the structure and the integrity of articial SEI layers would be destroyed by the interfacial mechanical strength, particularly during long-term cycling. Therefore, improving the stability, reducing the thickness, and increasing the exibility are critical to boost the practical applications of nitrided articial SEI layers in LMBs.

Electrolyte engineering
The deposition behaviour of Li + is strongly related to the physical and chemical properties of the electrolyte. 16,45,48 Electrolyte engineering, including introducing functional additives, using new solvents, regulating the Li + solvation structure, etc., is useful to stabilize Li metal electrodes. The most efficient method to evaluate the effects of these electrolytes is to test Li plating/stripping reversibility in Li-Cu cells.

LiNO 3 additive
LiNO 3 is a practical and economical additive to improve the electrochemical performance of LMBs. It was rst used in an ether-based electrolyte to suppress the shuttle effect of polysulphides in Li-S batteries by Aurbach et al. 87 They studied the surface of the Li anode cycled in the LiNO 3 -containing ether electrolyte and proposed that the LiNO 3 additive was decomposed on Li to form LiN x O y and LiOR. Wen et al. further proved that LiNO 3 is able to improve the coulombic efficiency (CE) of the Li plating/stripping processes in Li-Cu cells (Fig. 6a), as well as suppressing Li dendrite growth. 84 A smoother surface of the Li metal anode was obtained aer adding 0.4 M LiNO 3 into the ether-based electrolyte (Fig. 6b and c). By using the more accurate and sensitive X-ray photoelectron spectroscopy (XPS) depth prole method, Xiong et al. revealed that LiNO 3 in the ether-based electrolyte is reduced on the Li metal surface and forms a complex product consisting of Li 3 N, LiN x O y , and RCH 2 NO 2 (Fig. 6d). 85 Brezesinski et al. also demonstrated that LiNO 3 can form a protective layer on the Li metal anode and suppress gas evolution in the Li-S battery in conjunction with a diglyme-based electrolyte. In particular, the amount of ammable CH 4 and H 2 is dramatically decreased, and either very little or no H 2 is generated during discharge (Fig. 6e). 86 Even though LiNO 3 has achieved a big success in ether-based electrolytes and it also has good solubility in ether-based electrolytes (up to $5 wt%), its application in high-voltage carbonate-based electrolytes is limited due to its ultralow solubility in carbonate solvents (lower than 10 À5 g mL À1 ). 88,89 To boost the application of the LiNO 3 additive in carbonate-based electrolytes, various solubilizers have been utilized to promote its dissolution in carbonate solvents. It was initially reported that 2% vinylene carbonate (VC) can promote the dissolution of 0.1 M LiNO 3 in an ethylene carbonate/dimethyl carbonate (EC/ DMC)-based electrolyte and effectively improve the reversibility of Li plating/stripping processes in Li-Cu cells (Fig. 7a). 90 By analysing the surface of the cycled Li metal electrode with XPS, the existence of Li 3 N in the SEI was conrmed (Fig. 7b). Huang et al. further used a trace amount of CuF 2 to promote the dissolution of 1 wt% LiNO 3 in an EC/diethyl carbonate (DEC)based electrolyte, and proved that LiNO 3 was reduced on Li and formed a nitrided SEI (Fig. 7c). 91 Increasing the concentration of LiNO 3 in carbonate-based electrolytes could improve the electrochemical performance of LMBs. In this regard, Lu et al. used 0.5 wt% Sn(OTf) 2 , where OTf is triuoromethanesulfonate, as a solubilizer to increase the solubility of LiNO 3 in carbonate electrolytes to as high as 5 wt%. 35 Tin(II), which is a Lewis acid, can effectively coordinate NO 3 À and promote complete dissociation between ion pairs without decomposing the solvent molecules (Fig. 7d). By using high-resolution transmission electron microscopy (TEM, Fig. 7e) and XPS depth proling (Fig. 7f)  undergo a site-selective reaction at the inner Helmholtz plane and form an N and O-rich inorganic wavy SEI (Fig. 7g and h), which was experimentally proved by cryo-TEM results (Fig. 7i). The use of extra solubilizers has improved the solubility of LiNO 3 in a carbonate-based electrolyte, but they also increase the cost of the electrolyte. In addition, the solubilizers could be reduced on Li, so that they may destabilize the SEI. Sulfones (such as dimethyl sulfoxide (DMSO), sulfolane, etc.) have high solvability towards LiNO 3 , so they can replace the extra solubilizers and be used as solvents in the electrolyte to dissolve LiNO 3 . In this aspect, Wang et al. used DMSO solvent to dissolve LiNO 3 and prepared a 4 M LiNO 3 /DMSO solution as an additive. 93 They added 5 wt% of this additive into a carbonate-based electrolyte and achieved an ultrahigh CE of 99.55% in Li-Cu cells. It was indicated that distinct NO 3 À anions were involved in the Li + solvation sheath, and a small number of DMSO molecules were also found in the Li + solvation sheath ( Fig. 8a  and b). The NO 3 À in the solvation sheath could be reduced on the Li surface and formed a nitrided inorganic-rich SEI, which was more stable than the SEI formed in the LiNO 3 -free electrolyte ( Fig. 8c and d), while the DMSO molecules could not be decomposed. Therefore, denser and more compact plated Li was obtained on the Cu substrate ( Fig. 8e and f). Wang et al. also used pure sulfolane as the solvent in their electrolyte for LMBs, which contained 3.25 M lithium bis(triuoromethanesulfonyl) imide (LiTFSI) as a salt and 0.1 M LiNO 3 as an additive. 94 By using molecular dynamics (MD) simulations, they pointed out that the NO 3 À anions in the Li + solvation sheath could promote the coordination of TFSI À anions with Li + (Fig. 8e). During battery cycling, these anions in the Li + solvation sheath would be reduced and formed an inorganic SEI. It should be emphasized that LiNO 3 is strongly oxidizing, so it will increase the safety risk of the battery aer being added into the electrolyte, although most of the reported work failed to mention this safety issue. To address this problem, Guo et al. used triethyl phosphate as a solvent to dissolve 1 M LiNO 3 into the electrolyte as well as an extinguishant to eliminate re risk. 95 The developed electrolyte not only generated a nitrided SEI that could suppress Li dendrite growth (Fig. 8f), but also improved the safety of the resultant LMBs. The CE for Li plating/stripping processes only reached $97%, however, which was not high enough for practical LMBs.
In short, the use of LiNO 3 as an additive has effectively optimized the SEI and improved the Li plating/stripping reversibility. The application of LiNO 3 in high-voltage and more practical carbonate-based electrolytes for LMBs is limited, however, due to its low solubility. Different solubilizers were used to increase the solubility of LiNO 3 in carbonate-based electrolytes, although these solubilizers increase the cost of the electrolyte and their decomposition on Li would destabilize the SEI. LiNO 3 has high solubility in organic phosphate esters, sulfones, and amides, and they can be used as solvents or liquid solubilizers to dissolve LiNO 3 in the electrolyte. The thermodynamic stability of these solvents is poorer than that of carbonate solvents, however, which increases the undesirable side reactions between the electrolyte and the Li metal electrode. Furthermore, for the development of safe and practical LMBs, the re risk caused by the oxidizing properties of LiNO 3 should be carefully considered.

Other N-containing additives
Apart from LiNO 3 , some other N-containing additives can also be used to build a nitrided SEI on Li metal electrodes. In this regard, Sun et al. reported that Mg(NO 3 ) 2 can be dissolved in a carbonate-based electrolyte as an additive. 96 They suggested that Mg(NO 3 ) 2 can be dissolved directly as Mg 2+ and NO 3 À ions in the electrolyte even at a concentration of 0.1 M, which was quite different from the situation for LiNO 3 (Fig. 9a). The NO 3 À in the electrolyte could also form a LiN x O y -based SEI and improve the performance of both Li-Cu cells and Li-metal full cells. Wu et al. used metal-organic frameworks (MOF-808) as nanocapsules to load LiNO 3 , and used the MOF-808/LiNO 3 composite as an additive for LMBs (Fig. 9b). 97 The MOF-808 has an internal diameter of 18.4Å and a pore window of 14Å, which can efficiently encapsulate and diffuse LiNO 3 . During battery cycling, the LiNO 3 will be released to react with Li and form a nitrided-rich SEI. Xie et al. introduced nitrofullerene (nitro-C 60 ) as a bifunctional electrolyte additive to smooth the Li surface. 98 The nitro-C 60 in the electrolyte was designed to gather on the protuberances of the Li metal electrode and decompose to NO 2 À and insoluble C 60 . Aer that, NO 2 À further reacted with Li metal and formed a compact and stable Li 3 N/LiN x O y protective layer. The C 60 was anchored on the uneven grooves of the Li surface and resulted in a homogeneous distribution of Li + (Fig. 9c). Similarly, a paradigmatic N-rich polyether, nitrocellulose (NC), was used as an electrolyte additive to stabilize the Li metal electrode (Fig. 9d). 99 The NC additive has low LUMO energy so that it reacts with Li to form an endogenous symbiotic Li 3 N/cellulose double SEI. However, the Li plating/stripping CE only reached $92%, even though the base electrolyte used ethers as the solvents. The use of these N-containing additives also introduces extra cations and organic components into the electrolyte. Their inuence on the SEI composition and the performance of LMBs has not been clearly revealed, however. In addition, as shown in Table 2, the CE for Li plating/stripping in most of these electrolytes is lower than 98%, suggesting that they are not promising for practical applications at the current stage. Also, the stability of these N-containing additives has not been studied.

N-Containing ionic liquids
The decomposition of normal organic solvents on Li metal electrodes leads to the formation of organic components such as ROCOOLi or the inorganic component Li 2 CO 3 in the SEI, both of which have ultralow Li + ionic conductivity and limit the kinetics of Li plating. N-Containing ionic liquids can be used to optimize the SEI composition and generate more effective species for conducting Li + . Due to the high viscosity of ionic liquids, however, they are normally used in mixed solvents. For example, Guo et al. developed an electrolyte with a mixed solvent consisting of N-propyl-N-methylpyrrolidinium bis(tri-uoromethanesulfonyl)amide (Py 13 TFSI) and normal ether solvents 1,3-dioxolane/1,2-dimethoxyethane (DOL/DME) for LMBs. 100 The Py 13 TFSI was reduced to form N + (Py 13 ) and N À (TFSI) species in the SEI, which was able to passivate the active surface of the Li electrode (Fig. 10a). In addition, more ionically conductive Li 3 N was generated on the Li surface during battery cycling. Peng et al. developed an electrolyte that used a mixture of N-propyl-N-methylpiperidinium bis(-uorosulfonyl)imide (PI 13 FSI) and DOL as the solvent and Li [(CF 3 SO 2 )(n-C 4 F 9 SO 2 )N] (LiTNFSI) as the salt. 101 The ionic liquid and the salt decomposed on the surface of the Li metal anode to form an Li 3 N-containing SEI that was highly ionically conductive and exible (Fig. 10b), and a CE of 98.2% was achieved in Li-Cu cells, even at a high current density of 10 mA cm À2 . Choi et al. reported the use of 1-dodecyl-1-methylpyrrolidinium (Pyr1(12) + ) bis(uorosulfonyl)imide (FSI À ) in ordinary electrolyte solutions. 102 The Pyr1(12) + cation with a long aliphatic chain mitigated dendrite growth via the synergistic effects of electrostatic shielding and lithiophobicity, and the FSI À anion induced the generation of a rigid nitrided SEI (Fig. 10c). Even the ether solvents reduce the viscosity of N-containing ionic liquid-based electrolytes, and they also limit the voltage window of the electrolyte. In addition, the high volatility and ammability of ether solvents diminish the safety advantages of ionic liquids in the electrolyte. The high price of ionic liquids is another problem that should be addressed before large-scale applications.

Constructing F-rich and N-rich composite interfaces
LiF has an ultra-high Young's modulus, so it can suppress Li dendrite growth. 49 Although its bulk ionic conductivity is poor, it could form a compact structure and conduct Li + via grain boundaries because of the high grain boundary energy. Inorganic nitrides such as Li 3 N have much higher bulk ionic conductivity, but their particle size is larger than that of LiF, and the connections between the nitride grains are not as compact as that between LiF grains. The F-rich (LiF) and N-rich (Li 3 N and LiN x O y ) composite interface is more effective for stabilizing the Li metal anode. The most common approach to introduce the LiF species on the surface of Li is using F-containing solvents. In this regard, Zhang et al. designed an electrolyte with a mixed carbonate ester containing uoroethylene carbonate (FEC) as the solvent and LiNO 3 as an additive. 103 The FEC and LiNO 3 in the electrolyte altered the solvation sheath of Li + (Fig. 11a), and formed a uniform SEI with an abundance of LiF and LiN x O y on the Li metal anode. Wang et al. pointed out that the simultaneous use of LiNO 3 and FEC in a carbonate-based electrolyte reduced the reactivity of the electrolyte and formed a more compact SEI, thus shortening the diffusion paths of Li + through the SEI and improving the CE of the resultant LMBs (Fig. 11b). 104 Lu et al. dissolved 3 wt% LiNO 3 with the aid of 1 wt% tris(pentauorophenyl)borane as a solubilizer in FECbased carbonate solvents and produced an nitrided and uorinated composite SEI. 105 The co-existence of LiF and Li 3 N in the SEI on Li was veried with a cryo-TEM (Fig. 11c). Zhang et al. proved that under the protection of this F-and N-rich SEI, LMBs delivered much superior electrochemical performance and a prolonged lifespan (Fig. 11d). 106 Li et al. designed an electrolyte with a mixed solvent consisting of uoro-amide (2,2,2-tri-uoro-N,N-dimethylacetamide (FDMA)) and FEC. 107 The FDMA would react with Li via a three-step decomposition mechanism and nally form Li 3 N (Fig. 11e), while FEC would react with Li to generate LiF on the surface of the Li metal electrode. Beneting from the composite SEI, the plated Li was much denser and the stripping of Li was much more homogeneous. Zeng et al. also formulated an electrolyte with a mixture of uorine-rich carbonate and cyclophosphonitrile as the solvent, and formed a uoride-nitride ion-conducting interphase to suppress Li dendrite growth. 108 Optimizing the solvation structure of the electrolyte is another approach to generate a composite SEI. For instance, Zhang et al. proposed that the presence of NO 3 À in the electrolyte could alter the structure of the Li + solvation sheath and promote the decomposition of the FSI À anion, so that an SEI containing LiF and LiN x O y was formed on the Li surface (Fig. 11f). 109 As summarized in Table 2, the reported ether-based electrolytes normally deliver higher CE than ester-based electrolytes, because ether solvents are more stable against Li metal than ester solvents, although the low anodic decomposition voltage (<4 V) of ether-based electrolytes limits their application potential in high-voltage LMBs. In addition, although these reported electrolytes have improved the reversibility of the Li plating/stripping process, most of their CEs were still lower than 99.5%, which is not high enough for long-term and highvolumetric-energy-density LMBs. Furthermore, in most of the published results, the CE of the electrolytes was evaluated under a current density and an areal capacity lower than required for practical applications (higher than 3 mA cm À2 and 3 mA h cm À2 , respectively).
To maintain reasonable viscosity and stability of the electrolyte, the amount of LiNO 3 or other additives has been limited in the reported results. Nevertheless, these additives are continuously consumed/decomposed during battery cycling, and once the additives are depleted, the electrochemical performance of LMBs would decay rapidly. In contrast, the amounts of amides, uoroamides, and N-containing ionic liquids are higher than that of additives when they are used as solvents in the electrolyte, and they can quickly repair the damaged SEI via reacting with newly exposed Li. Therefore, from the point of view of prolonging the lifespan of LMBs, the use of these N-containing solvents is more promising than the use of N-containing additives.

Substrate modification
The Li plating behaviours on the surfaces of metallic Li anodes and current collectors are quite different. The lithiophilicity of the substrate determines the over-potential for Li plating and the size of the nuclei at the initial stage of Li deposition. 33,[116][117][118][119] Nevertheless, most substrates, such as Cu, are lithiophobic. [120][121][122] In addition, a practical current collector substrate is uneven with some protuberant tips on the surface, which induce a non-uniform distribution of the electric eld and inhomogeneous charge distribution near the substrate, which eventually leads to the formation and growth of Li dendrites. Constructing nitrided interfaces on current collector substrates is important to improve their lithiophilicity and adjust the local electric eld, thus regulating the uniform deposition of Li + .

Cu substrate
Cu is the most popular substrate/current collector for negative electrodes in LMBs or anode-free batteries. Nitrided interfaces are able to guide Li + homogeneous plating and improve the Li plating/stripping reversibility on a Cu substrate. Cui et al. synthesized an adaptive polymer with abundant N-H hydrogen bonding sites and applied it for Cu foil modication, where they achieved a much more uniform morphology of plated Li (Fig. 12a). 123 Li et al. used a reactive sputtering method to develop a Cu 3 N layer on Cu foil (Fig. 12b). 124 The Cu 3 N layer was believed to promote uniform surface electronic conductivity of the Cu foil, and it could further react with the deposited Li to form Li 3 N aer the rst plating process. The modied layer improved the CE in Li-Cu cells and the cycling stability of anode-free LiFePO 4 (LFP)kCu cells.
In addition to its function in protecting the Li metal electrode, as discussed above, g-C 3 N 4 can also regulate Li + deposition behaviour on Cu foil. Song et al. reported that the pyridinic nitrogen of g-C 3 N 4 can serve as a Li + affinity centre and help to improve the lithiophobicity of Cu foil (Fig. 12c). 125 The g-C 3 N 4 layer can also facilitate Li + conduction at the SEI through a siteto-site hopping mechanism. In addition, defect engineering of a C-N polymer was proposed to construct an N-decient ultrathin layer (27 nm) on Cu foil via reactive thermal evaporation (Fig. 12d). 126 The lithiophilicity of the defective C-N layer triggered a space charge effect in the SEI and enhanced its chargetransfer capability, leading to a lower nucleation over-potential. In addition, a three-dimensional (3D) porous poly-melamine- formaldehyde (PMF) framework was developed to modify Cu foil and prepare a PMF/Li composite anode. The amine and triazine groups in the PMF can homogenize Li + concentration near the Cu surface and regulate the uniform deposition of Li (Fig. 12e). 127 Decorating LiNO 3 on Cu foil is also effective for forming nitrided interfaces, but the main problem is that LiNO 3 consists of inorganic particles so that it cannot closely adhere to the Cu foil. To solve this issue, with the aid of a polymeric matrix of poly(vinylidene uoride-co-hexauoropropylene) (PVDF-HFP), Cui et al. coated a thin layer of LiNO 3 on the surface of rough Cu. 128 In this design, NO 3 À can be continuously released from the layer into the carbonate-based electrolyte during the Li plating process to maintain an appreciable local NO 3 À concentration at the anode surface (Fig. 12f). In addition, Xie et al. immersed commercially available Cu foam into LiNO 3 aqueous solution to load LiNO 3 particles into the pores and inner surface of the Cu foam. 129 When operating in a carbonatebased electrolyte, the LiNO 3 was reduced and formed an N-rich SEI on the outer and inner surfaces of the Cu foam. The authors believe that this facile method can be applied in large-scale production.
Apart from these advances, polyacrylonitrile (PAN) or PANbased materials were used as interfacial functional layers to guide the uniform deposition of Li + ions. 130,131 AlN interlayers, which simultaneously possessed high Li affinity and an insulating nature, were also used as a surface stabilizer for Li metal anodes. 132 Metal-organic framework (MOF) or MOF-derived materials could increase the affinity of Li + to the Cu substrate, so they were also reported to suppress Li dendrite growth. 121

Other substrates
3D nickel (Ni) foam can also be used as a substrate to accommodate Li. Similar to Cu, the lithiophobic and uneven surface of Ni leads to uniform deposition of Li. Compared with twodimensional (2D) planar Cu, the 3D porous structure of Ni foam offers space to alleviate the volume changes of Li during plating/stripping processes, so it could accommodate high capacity Li plating. To improve the surface lithiophobicity of Ni, Nan et al. decorated cobalt nitride (Co 3 N) nanobrushes on Ni foam. 133 The Co 3 N enabled a low over-potential for nucleation, leading to homogeneous plating of dendrite-free Li. Yang et al. used experimental results and theoretical simulations to prove that a micro-electric eld can be formed by the tri-s-triazine units of g-C 3 N 4 that were used to modify Ni foam, which induced numerous Li nuclei during the initial plating and guided the uniform growth of Li on the substrate ( Fig. 13a and  b). 134 Sun et al. decorated Ni x N on the surface of Ni foam, which improved the specic surface area to reduce the local current density and promoted the uniform plating of Li by the formation of Li 3 N. 135 3D carbon is also a good scaffold to host Li. The modication achieved by nitrided materials can help to increase the affinity of Li + to the carbon matrix and thus decrease the over-potential for Li plating. For example, Li et al. prepared TiN-decorated 3D carbon bres as a scaffold for Li metal anodes (Fig. 13c). 136 The TiN sheath on the carbon bre could absorb Li + and reduce the diffusion energy barrier, thus providing uniform nucleation sites for Li and suppressing the formation of Li dendrites ( Fig. 13d and e). Gong et al. decorated g-C 3 N 4 on 3D graphene to develop a g-C 3 N 4 /graphene/g-C 3 N 4 architecture, which can be used as an electrode to accommodate the Li metal anode (Fig. 13f). 137 The sandwiched structure can guide uniform Li plating/stripping in the van der Waals gap between the graphene and the g-C 3 N 4 . The g-C 3 N 4 can be regarded as an arti-cial SEI to prevent Li deposition on its surface and prevent the direct contact of the electrolyte with the Li metal because of its isolating nature. Other polar nitrides, such as AlN and Mg 3 N 2 , were also reported to modify 3D carbon hosts to regulate the Li plating behaviour and suppress Li dendrite growth. 115,138,139 In a brief summary, nitrided interface modication has improved the lithiophilicity and adjusted the local charge distribution at the surface of substrates/current collectors. As summarized in Table 3, with the functionalization of nitrided interfaces, the Li plating/stripping efficiency on Cu substrates or other substrates has been effectively enhanced to 96-99%. Nevertheless, for real LMBs, especially for anode-free LMBs (without excess Li), the CE should reach a level of 99.9%. This means that the effects of nitrided interfaces need to be further improved. Furthermore, most of the reported results were obtained in ether-based electrolytes, which can undoubtedly improve the CE. As discussed above, evaluating the electrochemical performance with a high-voltage ester-based electrolyte is more practically signicant. Besides, most of the reported results did not mention anode-free battery testing, while one of the most important aims of modifying substrates/current collectors is to build anode-free batteries.
It is worth noting that 3D Cu, Ni foam and 3D carbon hosts are all electronically conductive, so Li + from the electrolyte may accept electrons and be plated on their upper surfaces (separator side). Once this occurs, the effects of the 3D structure towards accommodating Li will be greatly weakened, while the volumetric energy density of the Li anode will be sacriced. As modied nitrides normally have poor electronic conductivity, they could decrease the surface electronic conductivity of the host and thus force Li + diffusion into the pores and lead to Li deposition on the inner surface of the 3D scaffolds.

Separator functionalization
Separators play a key role in all batteries. In LIBs and LMBs, the separator is a porous membrane placed between the positive electrode and negative electrode, which is permeable to ionic ow but prevents electric contact between the electrodes. 144 Previous results proved that coating a polypropylene (PP)-based separator with BN nanosheets was useful for suppressing Li dendrite growth and prolonging the lifespan of LMBs. 145 The separator can also be used to support N-containing materials to induce nitriding of the interface on the Li metal anode. Wang et al. creatively immersed a glass bre separator in a LiNO 3 solution to impregnate the separator with sub-micrometer-scale particles of LiNO 3 . 88 During battery cycling, the LiNO 3 was released into the carbonate-based electrolyte and reacted with Li to form a nitrided protective layer, which could suppress the formation of Li dendrites and "dead" Li. Li et al. proposed a similar concept for sustainably releasing NO 3 À in a carbonate electrolyte by intercalating superuous LiNO 3 particles between bi-layer polypropylene membranes (PP/LiNO 3 /PP). 146 Wu et al. developed a composite separator coated with polyacrylamide-graed graphene oxide molecular brushes (GO-g-PAM) ( Fig. 14a and b). 141 The polyacrylamide chains contained abundant N-H and C]O groups and thus enabled a molecularlevel homogeneous and fast Li + ux on the surface of Li. Besides, a layer of g-C 3 N 4 on commercially available PP separators was prepared ( Fig. 14d and e), and the g-C 3 N 4 on the PP lm was graed to the Li metal surface aer cell assembly. 142 It was proposed that the g-C 3 N 4 can form transient Li-N bonds at the electrode/electrolyte interface to effectively stabilize the Li + ux and thus enable smooth Li deposition at high current densities and capacities ( Fig. 14f and g). Besides, Huang et al. modied a hybrid layer of silk broin and polyvinyl alcohol (SF-PVA) on a PP separator via a freeze drying method. The SF-PVA layer will auto-transferred from the PP separator to the Li surface (Fig. 14h). 143 The N-H and C]O groups in SF are able to regulate the Li + ux distribution, and the SF-PVA layer can form a Li 3 N rich SEI. Therefore, uniform Li nuclei deposition was achieved.
The separator functionalization strategies can remarkably improve the lifespan and the electrochemical performance of Li metal electrodes, even under high current density and high capacity conditions. These strategies are also convenient for largescale production. Unfortunately, the introduction of N-containing materials increases the overall thickness of the separator, which will certainly sacrice the volumetric energy density of LMBs.

Prospects and outlook
Nitrided interfaces have effectively stabilized Li metal electrodes and improved their electrochemical performance as well as lifespan of LMBs. Despite this success, many critical issues and challenges remain to be carefully considered in the future. They are summarized below.

Investigating how nitrided interfaces affect the stripping process
The majority of research has only focused on the plating process, while the stripping process, deciding the utilization of deposited Li, was rarely noticed. There is no doubt that a uniform stripping step will lead to less pulverization and depletion of active Li during each cycle, which is a signicant factor in promoting the lifespan and electrochemical performance of LMBs. The course of Li dissolution with and without nitrided interfaces during the stripping step deserves to be carefully studied.

Understanding how nitrided interfaces affect the components and microstructure of the nal SEI layer
The introduction of nitrogenous additives will denitely inuence the solvent structure of the electrolyte, changing the nal decomposition products, while for nitrided interfaces fabricated ex situ, their existence also alters the decomposition of the electrolyte. In short, the nal SEI layer is composed of nitrogenous compounds and other components, and these products and their distributions have a direct correlation with the Li ion transport. If we only target nitrogenous compounds, we cannot achieve a comprehensive outlook with respect to the interfaces. A better knowledge of the interaction between nitrogenous compounds and other components will help to understand the differences in performance among the various nitrided interfaces.

Developing high-voltage N-containing electrolytes
The application of LiNO 3 and other N-containing additives is mainly conned to low-voltage electrolytes due to their low solubility in most high-voltage non-aqueous solvents. Increasing their solubility in high-voltage electrolytes is of great importance.

Controlling the thickness of nitrided interfaces
The thickness of reported nitrided or nitride-based composite interfaces varies from a few nanometres to more than 20 mm. To avoid the sacrice of the volumetric-energy-density of the Li metal battery, the thickness of the modication layers should be much thinner than that of the Li foil, especially considering that the Li foil used in the practical batteries is only 50 mm in thickness.

Improving the lifespan of the nitrided articial SEI
The reported nitrided articial SEI layers do exhibit positive effects towards suppressing Li dendrite growth and passivating the active Li surface, but their stability is far from satisfactory. They may be destroyed by the interfacial stress resulting from the huge volume changes of the Li metal electrode, so the real effects of nitrided articial SEI layers would be sacriced. Developing exible and robust nitrided articial SEI layers with longer lifespans is critical to promoting their practical effects.

Evaluating the effects of nitrided interfaces under practical conditions
According to the reported designs, most of the electrochemical performance was tested under mild conditions different from the real application situation. To make it more objective, the electrolyte and Li metal should be well quantied, while the cathode capacity should reach the commercial scale. In consideration of the mass energy density and the cost, the amount of electrolyte should be limited to less than 10 mL mA h À1 (lean electrolyte). Meanwhile, the areal capacity of the cathode should be higher than 3 mA h cm À2 , and the thickness of the Li metal anode should be less than 50 mm ($10 mA h cm À2 ), with the capacity ratio of the negative electrode (Li) to the positive electrode (n/p ratio) lower than 3. In addition, the current density and areal capacity in coulombic efficiency and symmetric cell tests should be increased to higher than 3 mA cm À2 and 3 mA h cm À2 , respectively.

Data availability
There is no origional experimental or computational data associated with this article, as it is a Perspective.

Author contributions
Z. Wang and Y. Wang wrote the manuscript. Z. Guo supervised this project. All the authors discussed and polished the manuscript.

Conflicts of interest
There are no conicts to declare.