WANG Ya-nan, ZHAN Ying-xin, ZHANG Xue-qiang, HUANG Jia-qi

（1. Advanced Research Institute of Multidisciplinary Science, Beijing Institute of Technology, Beijing 100081, China;2. School of Materials Science and Engineering, Beijing Institute of Technology, Beijing 100081, China）

Abstract： To address the issues of non-uniform Li plating/stripping and the large volume fluctuations during their repeated cycling, Li metal anodes, a composite Li anode with a three-dimensional (3D) host has been proposed as a promising strategy to improve the uniformity of Li plating/stripping and relieve volume fluctuations. One key strategy in this area is to develop 3D carbon hosts with gradients of lithiophilicity and conductivity to guide Li deposition from bottom to top, thus maximizing the positive effect of this composite Li anode. Such anodes have recently received significant attention due to the flexibility, adjustability,(electro)chemical stability and light weight of the carbon hosts. This review summarizes recent advances in these anodes, categorizing them into those that have hosts with a lithiophilicity gradient, hosts with a conductivity gradient, and those that have both. Latest research findings are discussed and these different anode categories are reviewed. Prospects for the design of such anodes to promote the use of Li metal anodes in high-energy-density rechargeable batteries are presented.

Key words: Lithium metal battery；Composite lithium anode；Gradient lithiophilicity；Gradient conductivity；Dual gradient

1 Introduction

Rapid development of portable electronic devices and electric vehicles has prompted the exploration of high-energy-density batteries[1-3]. Lithium (Li)-ion batteries have profoundly shaped our daily life by developing a non-fossil and wireless world[4-7]. The energy density of Li-ion batteries increases from about 90 to 270 Wh kg-1after the development of more than 30 years since 1991[8-12]. Nevertheless, the energy density of Li-ion batteries is approaching the ceiling(＜ 350 Wh kg-1) limited by the intercalation mechanism of the anode and cathode[13-15]. In order to further improve the energy density of batteries, Li metal anode is reviving for its high theoretical specific capacity (3860 mAh g-1) and low electrode potential(-3.04 V vs. standard hydrogen electrode)[16-20]. The research of Li metal anodes in batteries began in the 1960s[21-22]. In the 1980s, secondary Li metal batteries were commercialized by Moli Energy Company even though they failed[23-24]. The main obstacle on the way to practical Li metal anodes is non-uniform Li plating/stripping accompanied by large volume fluctuations, which results in low Coulombic efficiency(CE) and short lifespan of Li metal batteries[25-28]. Furthermore, the growth of Li deposition in the form of dendrites could pierce the separator and induce battery safety concerns[29-32]. Despite the above challenges, the high energy density (＞ 400 Wh kg-1) of Li metal batteries is still attractive and competitive.Therefore, improving the uniformity of Li deposition and mitigating volume fluctuations during cycles are under intensive research to stabilize Li metal anodes.

Tremendous efforts have been devoted to improving uniformity of Li deposition and mitigating volume fluctuations, such as liquid electrolyte design[33-38], artificial coating[39-44], lithiophilic sites[45-50], solid-state electrolytes[51-55]and composite Li anodes[56-61]. Among these strategies, a composite Li anode which employs a stable three-dimensional(3D) host emerges as a promising strategy for stabilizing Li metal anodes[62]. The large specific surface area of a 3D host could remarkably reduce local current density, thus mitigating the formation of Li dendrites based on Sand’s time model[63-65]. Surface modification of a host, such as decorating with lithiophilic sites, could further improve the uniformity of Li deposition in addition to the notable decrease in local current density[66-68]. Furthermore, the porous structure of a 3D host could provide space to effectively accommodate Li deposition and the formation of inactive Li, which contributes to relieving the volume fluctuations of composite Li anodes[69-71].

Generally, composite Li anodes with high electrical conductivity outperform those with low electrical conductivity and there is nearly no difference in electrical conductivity between the top and bottom of a host[72]. Once a Li ion contacts the host, it could obtain an electron rapidly. However, the diffusion of Li ions in bulk electrolytes is relatively slower than that of electrons in a host[73]. When the Li ions inside a host or near the surface of a host are consumed rapidly or even depleted, there is a concentration difference of Li ions between bulk electrolyte and electrode surface[74-76]. Then, the diffusion of Li ions from bulk to the surface is driven. When supplemental Li ions reach the top surface of a host, Li is inclined to deposit on the top surface which impedes subsequent Li deposition inside the host[77-78]. The porous structure of a host could not be utilized fully. Thus, inactive Li also inclines to form and accumulate on the top surface of a host, resulting in the rapid increase of polarization and deteriorating stability of composite Li anodes[79-80]. When it comes to practical Li metal batteries, a high cycling capacity (＞ 4.0 mAh cm-2), limited Li (＜ 10.0 mAh cm-2), and lean electrolytes(＜ 3.0 g Ah-1) are indispensable for achieving high energy density at a cell level, which aggravates the selective Li deposition and accumulation of inactive Li on the top surface of composite Li anodes[81].

Gradient composite Li anodes are consequently proposed in order to guide the bottom-up deposition pattern of Li in a host. The difference in lithiophilicity and electrical conductivity between the top and bottom area of a host is designed to guide bottom-up Li deposition[82]. Remarkable progress has been made designing gradient composite Li anodes recently,which improves the utilization efficiency of a host.The design of gradient composite Li anodes emerges as a promising solution to improve the stability of composite Li anodes[83]. To further guide the rational design of gradient composite Li anodes, the systematic analysis of gradient composite Li anodes design is necessary.

In this review, advances in carbon-based gradient composite anodes for stable Li metal batteries were summarized (Fig. 1). In retrospect, various composite Li anodes have been developed, of which carbon-based composite Li anodes are favorable due to the flexible adjustability, (electro)chemical stability and light weight of carbon hosts[84-87]. Numerous carbon materials, such as carbon fiber, carbon nanotube and graphene, have been employed as the 3D hosts for composite Li anode preparations[88-92]. The gradient hosts could be categorized into gradient lithiophilicity,gradient conductivity and dual gradient, which are introduced in the following sections, respectively[93]. In addition, the relationship between gradient host structure and Li plating/stripping behavior is discussed.The merits and limitations of different gradient host structures are pointed out. Perspectives of rational designs on gradient composite Li anodes are further proposed to promote the practical applications of Li metal batteries.

Fig. 1 Advantages and schematic diagram of the gradient composite anodes with lithiophilicity gradient, conductivity gradient and dual gradient

2 Gradient composite Li anodes

2.1 Gradient lithiophilicity

Lithiophilic sites have a strong affinity with Li atoms or ions, which could facilitate uniform Li deposition[94-95]. The lithiophilicity mentioned here can be categorized into 2 types according to the action object. One type is the ion-dipole interaction between Li ions and the polar functional groups on the surface of hosts, which guides lithiophilic sites to adsorb Li ions,such as carbide and doped carbon materials. The other type is that Li atoms will react with the lithiophilic sites on the surface of a host to reduce the Li nucleation barriers, such as metal and metal oxides. The rationale of gradient lithiophilic composite Li anodes is to decorate lithiophilic sites at the bottom of a host and decorate lithiophobic sites at the top of a host, so that the affinity for Li atoms or ions from bottom to top of the host is constantly reduced, forming gradient lithiophilicity inside the host. Gradient lithiophilic host can guide Li ions to preferentially nucleate and deposit at the bottom, thus avoiding deposition of Li at the top of the host.

Zhang et al. developed a host by sequentially dripping carbon nanotubes (CNTs) suspensions with gradient decreased concentrations of zinc oxide (ZnO)layer-by-layer on Li foil (GZCNT) (Fig. 2a-b)[96]. The gradient loading of ZnO from bottom to top serves as gradient lithiophilic sites inside GZCNT host. The bottom layer of lithiophilic ZnO/CNT is tightly anchored to Li foil, which facilitates the formation of stable solid electrolyte interphase (SEI) and inhibits Li corrosion. The top layer of lithiophobic CNT facilitates Li ion diffusion and provides mechanical strength to avoid Li dendrites. In a symmetric Li//Li pouch cell with a surface area of 10 cm2, the GZCNT-coated Li anode maintained stable for over 200 cycles without any dendrites, while the voltage of bare Li anode fluctuated after 55 cycles (Fig. 2c). This strategy could enhance cycling stability of copper (Cu) current collector, 10 cm2Li foil pouch cell, and Li-sulfur (Li-S)battery significantly.

Fig. 2 (a) Schematic diagram of the GZCNT host-coated Li foil[96]. (b) Fabrication of interfacial layer of GZCNT. (c) Cycling performance of symmetric pouch cells. Inset: digital photograph of Li foil, and GZCNT-coated Li foils in pouch cells after 100 and 200 cycles, respectively. (Reprinted with permission)

Yan et al. obtained a gradient lithiophilic 3D conductive host (GSCP)viamagnetron sputtering of silicon (Si) on the bottom of a porous carbon paper(CP)[97]. The content of lithiophilic Si decreases from bottom to top due to unidirectional sputtering, forming a gradient lithiophilic host. Therefore, Li preferentially nucleates and grows in a bottom-up pattern and the space utilization of the host is elevated. GSCP anode maintained an average CE of 99.0% over 400 cycles in a Li//GSCP cell at 1.0 mA cm-2and 1.0 mAh cm-2, displaying excellent electrochemical reversibility. When the current density was increased to 5.0 mA cm-2, the average CE of GSCP was 97.0%over 150 cycles. Before assembling the full cell, 5.0 mAh Li was deposited on a GSCP anode to obtain GSCP@Li anode. Matched with Li4Ti5O12(LTO, with a loading of ～ 1.2 mg cm-2) cathode,GSCP@Li//LTO full cell delivered a specific capacity of 114.8 mAh g-1at 10 C with capacity retention of 84.5% after 5 000 cycles.

The combination of carbon-based gradient lithiophilic host and modified lithiophilic substrate can effectively improve the efficiency of Li deposition. Yun et al. proposed a carbon host derived from metal-organic framework (MOF), which consisted of MOFderived carbon (MOF-C) and Cu substrate modified with silver (Ag) layer (Cu@Ag)[98]. The electrodes with different thicknesses (t, μm) and substrates (X)were denoted as MOF-C(t)/X. To explore Li plating behavior, the voltage profiles of MOF-C(t)/X//Li pouch cells illustrated that MOF-C(30)/Cu@Ag electrode showed the lowest nucleation overpotential(6 mV), while the nucleation overpotential of MOFC(30)/Cu was the highest (21 mV), illustrating Cu@Ag substrate could reduce barriers for Li nucleation. In the initial stage of deposition, Li ions reacted with Ag layer at the bottom of MOF-C(30)/Cu@Ag,which could regulate the subsequent Li deposition process. Experimental and simulation results indicated that local current density was the highest at the bottom of MOF-C(30)/Cu@Ag electrode and decreased upward, which suggested Li preferentially deposited at the bottom (Fig. 3a-b). However, when the thickness of MOF-C increased to 90 μm, the local current density was higher at the top than that at the bottom of MOF-C(90)/Cu@Ag electrode, because the ion transport resistance increased with the thickness increase of electrode. Thus, lithiophilic Cu@Ag substrate with high thickness above to 90 μm could lose its advantage of inhibiting top deposition (Fig. 3c). In MOF-C(t)/X//Li pouch cells, 2.0 mAh cm-2of Li was prestored before the cycling tests, and the MOFC(30)/Cu@Ag electrode exhibited stable cycling for over 250 cycles (500 h) at 0.4 mA cm-2and 0.4 mAh cm-2(Fig. 3d).

Fig. 3 Schematic diagram of Li plating behaviors of (a) MOF-C(30)/Cu,(b) MOF-C(30)/Cu@Ag, and (c) MOF-C(90)/Cu@Ag[98]. (d) Cycling performance of MOF-C(t)/X//Li pouch cells at 0.4 mAh cm-2 and 0.4 mA cm-2. (Reprinted with permission)

Graphene oxide (GO) is widely considered as a carbon host due to its excellent lithiophilicity. Some surface groups of reduced GO (rGO) have high binding energy to Li, indicating high surface lithiophilicity[99]. Nevertheless, tortuous and long Li ion transport pathways induce large transport impedance, and Li tends to be deposited at the upper area of the host.As a consequence, the anode with an aligned structure could provide low tortuosity Li ion transport pathways, leading to Li bottom-up deposition[100-101].Hence, Ni et al. combined GO with aligned structure by 3D printing, and then infused molten Li to form holey GO/Li composite anodes[102]. GO/Li composite anode was flexible and had a controllable thickness.The aligned channels in GO host uniformly distributed Li-ion flux and provided short ion transport pathways, which contributed to enhancing rate performance and increased the energy density of the battery(Fig. 4a, b). In a symmetric cell, the holey GO/Li composite anode exhibited a lower overpotential over 400 h than unpunched GO/Li and bare Li anode at 1.3 mA cm-2and 1.3 mAh cm-2. When matched with LiFePO4(LFP, with a loading of 2.0-2.4 mg cm-2)cathode, the holey GO/Li anode has a capacity of 93.0 mAh g-1at 20 C and stably cycled for more than 600 cycles at 1 C. Likewise, when matched with LiNi0.8Co0.1Mn0.1O2(NCM811, with a loading of 3.4-4.7 mg cm-2) cathode, the holey GO/Li anode displayed capacity of 117.9 mAh g-1at 10 C and maintained a capacity of more than 150.0 mAh g-1for over 150 cycles at 0.5 C. More notably, in a pouch cell matched with a sulfur/polyacrylonitrile (SPAN) cathode with S loading of 9.8 mg, it displayed an initial capacity of 1 015.0 mAh g-1and capacity retention of 80.0% over 100 cycles.

Fig. 4 Schematic diagram of Li plating/stripping of (a) pure GO and (b) holey GO[102]. Schematic diagram of Li plating behaviors of (c) WDC-GDAg and(d) WDC[105]. (e) CE of WDC-GDAg and WDC half cells at 10.0 mA cm-2 and 5.0 mAh cm-2. (f) Cycling performance of WDC-GDAg//LFP, WDC//LFP, and Li//LFP at 10 C. (Reprinted with permission)

Low tortuosity of a host has been proven effective for fast Li ion transport[102-104]. Unfortunately, this structure also results in inhomogeneous local current density, increasing the risk of Li dendrites growth.The gradient distribution of lithiophilic sites in this structure is favorable for achieving synergistic effects.Herein, Zhu et al. introduced the gradient distribution of Ag, ZnO, and Au lithiophilic sites into 3D woodderived carbon (WDC) host with low tortuosityviaa capillary-induced gradient deposition[105]. The WDC host decorated with gradient-distributed Ag nanoparticles (WDC-GDAg) could form a high concentration of Li-ion flux at the bottom and a homogeneous electric field distribution at the top, which eventually achieved the bottom-up deposition (Fig. 4c, d). The WDC-GDAg anode maintained the average CE of 98.3% over 70 cycles at 10.0 mA cm-2and 5.0 mAh cm-2(Fig. 4e). When matched with the LFP(with a loading of 4.0 mg cm-2) cathode, the WDCGDAg//LFP delivered a capacity of 130.0 mAh g-1and a capacity retention of 78.9% over 2 000 cycles at a rate of 10 C (Fig. 4f).

Recently, Cao et al. constructed g-C3N4coated carbon cloth (CC@g-C3N4)viaa thermal polycondensation method, and then infused molten Li into the CC@g-C3N4to obtain CC/Li/Li3N anode[106]. The g-C3N4reacted with Li to form Li3N when molten Li infused, and then during the resting process at high temperature, Li3N gradually diffused and migrated upward to form a Li3N gradient, which created a gradient lithiophilicity structure in the anode. Owing to the high ionic conductivity and electrochemical stability of Li3N, the Li3N-rich artificial SEI layer which was formed by the Li3N gathered at the top could accelerate the diffusion of Li ions at the interface and prevent continuous side reactions between anode and electrolytes. Density function theory (DFT) results prove that Li3N has a low energy barrier of Li ion diffusion. In addition, the rest of Li3N which was not gathered at the top uniformly disperses in the host,regulating uniform Li deposition (Fig. 5a, b). In symmetric cells, the overpotential of CC/Li/Li3N electrode was only 80 mV after cycling 1 000 h at 2.0 mA cm-2and 2.0 mAh cm-2(Fig. 5c). When matched with the LFP (with a loading of 2.0 mg cm-2) cathode,CC/Li/Li3N//LFP cell had an initial discharge capacity of 146.6 mAh g-1at 0.5 C and the capacity retention rate was 93.3% after 150 cycles. Besides,CC/Li/Li3N//LFP full cell also exhibited improved rate performance than Li//LFP (Fig. 5d). At room temperature, the solid-state battery matching the LFP(with a loading of 2.0 mg cm-2) cathode and garnet(LLZO) electrolytes maintained a capacity of 140.0 mAh g-1after 100 cycles at 0.1 C (Fig. 5e).

Gradient lithiophilic composite anodes could preferentially guide Li deposition inside the bottom of hosts, which plays a positive role in alleviating the volume change of Li and improving the overall space utilization of the host. However, there are still many challenges on the way to practical applications of gradient lithiophilic composite anodes. The preparation of gradient lithiophilic composite anodes requires a complex process and large-scale preparation of gradient lithiophilic host is still a challenge. In addition, it remains unclear whether lithiophilic sites would fall off or the structure of lithiophilic sites change during the repeated Li plating/stripping.

2.2 Gradient conductivity

Similar to gradient lithiophilicity, gradient conductivity is constructed by using the material with high conductivity at the bottom of the host and the material with low conductivity or electrical insulation at the top of the host, so that the conductivity decreases from bottom to top continuously. Gradient conductive composite Li anodes could control the distribution of electric field and induce Li to preferentially deposit on the host material with high conductivity.

Initial studies of gradient conductivity hosts focus on combining metal with polymer materials. Hong et al. constructed a conductivity gradient host (CG)composed of conductive copper nanowires (CuNWs)and insulating cellulose nanofibers (CNFs)viaa vacuum-assisted infiltration process[107]. The bottom layer was CuNWs with high conductivity, the middle layer was a mixture of CuNWs and CNFs with low conductivity, and the top insulating layer was a mixture of CNFs and SiO2nanoparticles. SiO2nanoparticles in the top area of CG could exhibited superior wettability with electrolytes, increasing Li-ion flux on polar groups of the CNFs (Fig. 6a-c). Experiments and COMSOL multiphysics simulation revealed that when the thickness of the top insulating layer was 1 μm and the difference in conductivity between the middle and bottom layer was 6 kS cm-1, Li preferentially deposited at the bottom of the CG host. In symmetric cells,CG electrode maintained low voltage hysteresis for about 250 cycles at 1.0 mA cm-2and 1.0 mAh cm-2.Besides, CG was stable for 100 cycles when the current density increased to 5.0 mA cm-2. When matched with a NCM811 (with a loading of ～1.1 mAh cm-2)cathode, the capacity retention of the full cell (N/P ratio of 3.6) was more than 90.0% after 100 cycles at 1 C.

Fig. 6 Schematic diagram of Li plating behaviors of (a) Cu foil, (b) Cu mesh, and (c) CG host[107]. Schematic diagram of Li plating behaviors of (d)CDG-sponge: bottom deposition without dendrite formation. (e) D-sponge:bottom deposition with dendrite formation. (f) C-sponge: top deposition with dendrite formation near the separator[108]. (Reprinted with permission)

Li et al. prepared nickel (Ni) metal on one side of melamine sponge to form a conductive-dielectric gradient hos1 (CDG-sponge) by magnetron sputtering[108]. For comparison, a fully conductive host(C-sponge) and a dielectric host (D-sponge) were also fabricated. The thickness of the conductive Ni coating layer of CDG-sponge attenuated upward, and the top area was bare melamine, forming a conductivity gradient to achieve bottom-up deposition and topdown stripping pattern of Li (Fig. 6d-f). Melamine with polar functional groups (e.g., amine group) could adsorb Li ions and homogenize Li-ion flux. Before assembling symmetric cells, 3.0 mAh cm-2of Li was first deposited on th1 CDG-spong1 (Li@CDGsponge). At 1.0 mA cm-2and 1.0 mAh cm-2,Li//Li@CDG-sponge cell showed a low polarization voltage (＜ 20 mV) and a long lifespan (780 h).Matched with LFP (with a loading of 4.0 mg cm-2)cathode, Li@CDG-sponge//LFP cell maintained a capacity of 127.3 mAh g-1after 400 cycles at 1 C with a capacity retention of 87.6%.

Inspired by these studies above, research on carbon-based hosts with gradient conductivity was initiated. Zou et al. stacked polyacrylonitrile (PAN) membranes and carbon nanofiber (CNF) membranes in sequence to form a periodic PAN/CNF host[109]. PAN membranes acted as dielectric layers to periodically separate the conductive layers (i.e., CNF membranes),blocking the electronic transport pathways from the bottom conductive layer to the upper conductive layer.Only when the plated Li metal connected the 2 conductive layers could electronic pathways be formed.More importantly, if Li dendrites were formed and grown to contact with the upper conductive layer, the electric field could be re-homogenized, and the local current density of the dendrites would decrease significantly, presenting a self-correction effect (Fig. 7a).Consequently, Li could be deposited smoothly at high current density and high cycling capacity, following a bottom-up pattern (Fig. 7b). 10.0 mAh cm-2of Li metal plating in periodic PAN/CNF host was denoted as Li@PAN/CNF. At 5.0 mA cm-2and 5.0 mAh cm-2,Li@PAN/CNF//Li@PAN/CNF symmetric cell delivered a stable voltage hysteresis for over 800 cycles.When matched with the LiNi0.6Co0.2Mn0.2O2(NCM622,with a loading of ～12.0 mg cm-2) cathode, Li@PAN/CNF//NCM622 cell displayed an initial capacity of 153.0 mAh g-1at 1 C and the capacity retention of 87.0% after 100 cycles. Likewise, matched with the NCM811 (with a loading of ～20.0 mg cm-2) cathode,Li@PAN/CNF//NCM811 full cell delivered an initial capacity of 3.5 mAh cm-1at 1 C and maintained 70.0% capacity after 100 cycles.

Fig. 7 (a) Schematic diagram of Li metal evolution in periodic-host based Li metal anode showing a “self-correction” behavior, and (b) schematic diagram of Li plating/stripping of the periodic PAN/CNF host[109]. (c) Schematic diagram of Li plating behaviors of CC and SiC/CC host, and (d) cycling performance of the Li@Cu//LFP and Li@SiC/CC//LFP full cell at 0.5 C[110]. (Reprinted with permission)

Sun et al. synthesized a gradient topological host consisting of SiC whiskers and carbon cloth (SiC/CC)viaa gas-solid reaction of gaseous SiO with porous CC[110]. During the reaction, the gas concentration of SiO increased with the height of the CC host and concentrated at the top surface, forming a CC host with gradient-decorated SiC whiskers. A gradient conductivity structure was derived from the conductivity difference between semiconductive SiC and conductive carbon fibers. The topological structure in SiC/CC host could not only ensure contacts with the deposited Li reducting inactive Li but also effectively regulate the distribution of electric field and Li-ion flux,which was conducive to a uniform deposition of Li(Fig. 7c). In addition, Li would react with C or Si of the SiC/CC host to from Li-Si and Li-C alloys[111],which could reduce Li nucleation barriers. In half cells, SiC/CC electrode remained at a high CE(＞ 95.0% of the initial CE) after 100 cycles at 1.0 mA cm-2and 1.0 mAh cm-2and kept stable after 50 cycles at 5.0 mA cm-2and 3.0 mAh cm-2. Li@SiC/CC (with an average mass loading of 5.0 mAh cm-2)symmetric cell maintained a low overpotential (＜ 20 mV) for over 1 000 h at 1.0 mA cm-2and 1.0 mAh cm-2, and showed a lifespan of over 320 h at 5.0 mA cm-2and 5.0 mAh cm-2. When Li@SiC/CC (3.0 mAh cm-2) anode matched with the LFP (with a loading of 4.5 mg cm-2) cathode, Li@SiC/CC//LFP cell(N/P ratio of 4.4) showed a capacity of 120.0 mAh g-1after 120 cycles with the capacity retention of 80.0%at 0.5 C (Fig. 7d).

Modification of the top area of the carbon-based gradient host is conducive to the formation of stable SEI and maintaining the gradient conductivity structure. Zhou et al. combined a dielectric top layer with the decorated LiNO3particles and CNF layers with gradient conductivity to form a LiNO3-modified conductivity gradient host (LNO-CGH)[112]. LNO-CGH consisted of 4 layers. Namely, the lower 3 layers were CNF layers with different concentrations of CNTs.Among them, the bottom layer corresponded to the spinning solution with 10.0% mass ratio of CNTs and PAN (10CCNF), and the upper 2 layers corresponded to the spinning solution with 5.0% and 0% mass ratio of CNTs and PAN (5CCNF and CNF), forming a conductivity gradient. The top layer was a dielectric PAN film with LiNO3particles (LNO-PAN), which had the lowest conductivity and could continuously release a limited amount of LiNO3into the electrolytes during cycles, promoting the formation of a robust SEI and improving the thermodynamic stability of anode(Fig. 8a-c). In half cells, the average CE of LNO-CGH was 97.4% over 200 cycles at 1.0 mA cm-2and 2.0 mAh cm-2. When the current density and cycling capacity increased to 2.0 mA cm-2and 5.0 mAh cm-2,the LNO-CGH electrode maintained an average CE of 97.3% over 60 cycles. Before assembling the full cell,5.0 mAh cm-2of Li was deposited on the LNO-CGH host (LNO-CGH@Li) as an anode. When matched with the LiNi1/3Co1/3Mn1/3O2(NCM111, with a loading of 9.5 mg cm-2) cathode, LNO-CGH@Li//NCM111 cell (N/P ratio of 4.0) displayed the capacity retention of 74.9% after 120 cycles at 1 C. Similarly, matched with the NCM811 (with a loading of 11 mg cm-2) cathode, LNO-CGH@Li//NCM811 cell(N/P ratio of 2.3) maintained 72.9% capacity retention after 60 cycles at 1 C (Fig. 8d). A flexible quasisolid Li metal battery composed of LNO-CGH@Li(5.0 mAh cm-2) anode, flexible NCM811@PCNF1000(with a loading of 5.5 mg cm-2) cathode, and P(VDFHFP)-based gel polymer separator showed a capacity of 191.4 mAh g-1at 1 C and capacity retention of 78.2% after 100 cycles (Fig. 8e).

Fig. 8 Schematic diagram of Li plating behaviors of (a) Cu foil, (b) 10CCNF and (c) LNO-CGH[112]. (d) Cycling performance of Cu@Li//NCM811 and LNOCGH@Li//NCM811 full cell at 1 C. (e) Schematic of flexible quasi-solid Li metal battery (LNO-CGH@Li//Gel electrolytes//NCM811@PCNF1000).(Reprinted with permission)

Gradient conductive composite Li anodes could guide Li deposition following a bottom-up pattern by reducing the local current density at the top of a host.Nevertheless, due to the high conductivity of Li metal,the electric field would change as Li deposits continuously. Thus, gradient conductive composite Li anodes may not be able to exhibit advantages in the repeated Li stripping/plating process.

2.3 Dual gradient

In order to combine the advantages of gradient lithiophilicity and conductivity, composite Li anodes with dual gradient are proposed, in which lithiophilic sites on the conductive materials are located at the bottom of the host and the lithiophobic materials with low conductivity are placed at the top of the host. The dual gradient could regulate ion and electron conducting pathways simultaneously and synergistically guide homogeneous Li deposition from bottom to top.

Pu et al. constructed a deposition-regulating scaffold (DRS) with a dual gradient[113]. They first prepared a porous bare nickel scaffold (BNS)viatemplated electrodeposition and selective etching. Then,the top of BNS was coated with alumina (Al2O3) for electrical passivation and the bottom area was activated by an Aurum (Au) layer to reduce Li nucleation barrier. The Al2O3coating at the top area of DRS reduced local conductivity, creating a conductivity gradient with conductive metal Ni/Au at the bottom of the host. Likewise, lithiophilic Au at the bottom of DRS showed almost zero nucleation barrier, while high-barrier Al2O3coating was at the top area, together forming a lithiophilicity gradient. Under the synergistic regulation of gradient conductivity and lithiophilicity, Li deposited following a bottom-up pattern in the DRS (Fig. 9a-b). Under different test conditions (i.e., current density and cycling capacity), DRS electrode presented higher CE and a longer lifespan than BNS and Cu foil. It is noteworthy that Li was preferentially deposited at the bottom of DRS even at low-temperature conditions (5 and -15 °C), and the DRS cell could keep 50 cycles with no voltage fluctuation at -15 °C.

Cai et al. designed a sandwich-like gradient host with an artificial SEI (LGH-AS), composed of rGO,SiO2, rGO and poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) in a bottom-up sequence (Fig. 9c, f)[114]. Simultaneously, LGH and AS were fabricated. LGH consisted of two layers of rGO and SiO2interlayer (Fig. 9d), while AS was composed of bottom rGO and PVDF-HFP (Fig. 9e). In LGH-AS, the bottom rGO layer was tightly anchored to Cu foil and conducted electrons from the foil. SiO2interlayer provided lithiophilic sites and simultaneously prevented the bottom and upper rGO layers from contacting and forming electronic pathways. The upper rGO layer prevented SiO2from falling off and endowed mechanical strength to inhibit Li dendrites.The PVDF-HFP layer kept the whole electrode surface insulated and guided Li plating in SiO2interlayer.SiO2and Li would react to form an irreversible Li2O phase on the particle surface, which could uniform Liion flux[115]. The average CE of LGH-AS electrode was 98.1% throughout 275 cycles at 0.5 mA cm-2and 0.5 mAh cm-2. When cycling capacity increased to 5.0 mAh cm-2, LGH-AS exhibited an average CE of 99.1% for 60 cycles. 4.0 mAh cm-2of Li was pre-deposited on LGH-AS (LGH-AS@Li) as an anode,matching with the LFP (with a loading of 7 mg cm-2)cathode to assemble a full cell, which exhibited capacity retention of 90.2% after 300 cycles at 0.5 C and a better rate performance than the other 3 cells from 0.2 C to 2 C (Fig. 9g-h).

Zeolite imidazolate framework (ZIF) is a class of MOFs with excellent chemical and thermal stability[116]. ZIF-8 is a porous material consisting of zinc ions (Zn2+) and imidazolium ligands. Shi et al.prepare1 ZIF-8-derived gradient interfacial layer(ZGIL)viaa two-step lamination method, with ZIF-8 and polyvinylidene fluoride (PVDF) at the top layer and carbonized ZIF-8 (C-ZIF-8) and CNT dispersion at the bottom layer (Fig. 10a)[117]. C-ZIF-8 at the bottom layer of ZGIL provided high Li-ion and electron conductivity, and the Zn clusters or single atoms in it exhibited excellent lithiophilicity. ZIF-8 as a surface electric passivation layer could suppress the top deposition of Li. Moreover, the high Li-ion conductivity of ZIF-8 could not only enhance the rapid diffusion of Li ions and homogenize Li-ion flux, but also release a large amount of Li ions by anchoring anions in electrolytes, such as bis trifluoromethyl sulfonic acid amide ions and hexafluorophosphate ions (TFSI-and PF6-). The average CE of ZGIL-Cu electrode was 97.6% over 300 cycles at 1.0 mA cm-2and 1.0 mAh cm-2. The ZGIL-Li modified symmetric cell cycled for over 700 h at 1.0 mA cm-2and 1.0 mAh cm-2, and it exhibited better rate performance than bare Li and C-ZIF-8-Li. In full cells, the ZGILmodified Cu foil with Li (3.0 mAh cm-2) was used as the anode. While matching with the LiCoO2(with a loading of 4.6 mg cm-2) cathode, LiCoO2//ZGIL cell(N/P ratio of 6.0) remained 56.0% capacity retention after 160 cycles at 2 C (Fig. 10b).

Fig. 10 (a) Schematic diagram of ZGIL coated Cu foil. (b) Cycling performance of LiCoO2//ZGIL and LiCoO2//Cu full cell at 2 C[117].(Reprinted with permission)

The combination of 3D porous hosts and gradient structures could indeed uniform Li deposition, following a bottom-up deposition pattern. Nevertheless,this synergistic effect is generally achieved by adopting a thick host at the expense of volume energy density[118-119]. Yu et al. designed a thin (6-8 μm) composite host, namely a gradient Ag decorated graphene/holey graphene film (G-HGA), which was composed of Ag decorated GO (AgGO), Ag decorated GO mixed with holey GO (AgGO-hGO), and hGO hydrogels in a bottom-up sequence[120]. The concentration of Ag nanoparticles (AgNPs) decreased from bottom to up, constituting lithiophilicity and conductivity gradient simultaneously. hGO provided sufficient mechanical strength, and the hole defects of hGO shortened Li ion transport pathways and facilitated the fast transport of Li ions. The average CE of Li//G-HGA cell was 98.8% for 350 cycles at 1.0 mA cm-2and 1.0 mAh cm-2. When the cycling capacity was increased to 2.0 mAh cm-2, G-HGA electrode maintained high stability over 400 cycles. At 1.0 mA cm-2and 1.0 mAh cm-2, G-HGA@Li//G-HGA@L1 (wit1 4.0 mAh cm-2deposited Li) symmetric cell delivered a cycling lifespan for over 600 h. When G-HGA@Li(3.0 mAh cm-2) anode was matched with LFP (with a loading of 2.7 mAh cm-2) cathode, G-HGA@Li//LFP cell (N/P ratio of 1.1) showed a capacity of 159.0 mAh g-1at 0.2 C, and capacity retention of 99.6%after 175 cycles at 0.5 C.

Dual gradient composite Li anodes could steer Li ions to accumulate at the bottom of the host, and regulate homogeneous stripping/plating behavior of Li,achieving the synergistic effect of guiding Li bottomup deposition pattern. However, most designs of dual gradient composite Li anodes consist of many layers with different properties, which form thick hosts at the expense of the energy density of Li metal batteries.

3 Outlooks

Gradient composite Li anodes could guide the bottom-up deposition pattern of Li and improve the space utilization efficiency of hosts. Since Li stripping/plating process is regulated by the electronic and ionic pathways, the gradient hosts could be divided into gradient lithiophilicity, gradient conductivity and dual gradient. Various strategies have been proposed to prepare carbon-based gradient composite Li anodes (Table 1). However, the veritably working mechanism and effectiveness under practical conditions, especially in pouch cells, of gradient composite Li anodes are still controversial. In order to further promote the development of gradient composite Li anodes, the following aspects are suggested:

Table 1 Comparison of full cell and pouch cell performance of carbon-based gradient composite anodes

(1) Effectiveness under practical conditions.Practical conditions are necessary for achieving high specific energy of Li metal batteries. However, most research employs coin cells under mild conditions(i.e., thick Li and flooded electrolytes). There is a large gap between the performance under mild and practical conditions for Li metal batteries[121-122]. To explore the practical applications of gradient composite Li anodes, practical conditions such as high-loading cathode (＞ 4.0 mAh cm-2), limited Li (＜ 10.0 mAh cm-2), and lean electrolytes (＜ 3.0 g Ah-1) and pouch cell measurements should be considered.

(2) Optimization of host structure parameters.The increase of pore size from bottom to top could accelerate the transport of Li ions[123-124]. The pore gradient could be combined with lithiophilicity gradient and conductivity gradient to synergistically regulate Li stripping/plating process. Owing to the layers with different properties, the gradient hosts are usually thick, which reduces the energy density of batteries.To obtain a high specific energy battery, the thickness of gradient composite Li anodes should be controlled.

(3) Failure analysis of gradient design. For the lithiophilic composite Li anodes, inactive Li could cover lithiophilic sites and block Li ion transport pathways towards lithiophilic sites during cycles, inducing hosts to lose their gradient lithiophilicity[125].Similarly, electric field could vary due to Li deposition with high conductivity. Therefore, it is necessary to investigate the failure mechanism of gradient design and corresponding strategies which could prolong the lifespan of gradient composite Li anodes and enhance the cycling stability of Li metal batteries.

(4) Exploring Li plating/stripping mechanism in gradient composite Li anodes. Understanding the Li plating/stripping mechanism, especially during repeated cycles, in gradient composite Li anodes by advanced characterization tools or theoretical simulations could provide rational guidance for further design of gradient composite Li anodes. Ingenious experimental design is also required to decouple the ef-fect of host properties on the behavior of Li plating/stripping because there are intersectant effects between different properties.

(5) Large-scale preparations of gradient composite Li anodes. Nowadays, the preparations of gradient composite Li anodes are generally complex. Thus, the consistency during large-scale preparations should be carefully evaluated, and simple technologies are pursued. In addition, when it comes to practical applications, the cost should be considered, which should be taken into consideration during the design of gradient composite Li anodes.

In summary, gradient composite Li anodes have the potential to regulate the bottom-up deposition of Li and achieve a uniform Li deposition morphology,which is essential for the development of Li metal batteries. However, there are still practical challenges that need to be addressed in order to implement gradient composite Li anodes in practical applications despite their potential benefits. Therefore, it is important to gain a deep understanding of the regulation and failure mechanisms of these anodes, and to design a host material that considers various aspects to find the optimal solution. By doing so, we could pave the way for the practical applications of high-energy-density Li metal batteries based on composite Li anodes.

Acknowledgements

This work was supported by National Key Research and Development Program(2021YFB2400300), the Beijing Natural Science Foundation (JQ20004), the National Natural Science Foundation of China (22209010 and 22109007), the China Postdoctoral Science Foundation(2021M700404), the Beijing Institute of Technology Research Fund Program for Young Scholars.