Lithium aluminum alloy anodes in Li-ion rechargeable batteries: past developments, recent progress, and future prospects

Aluminum (Al) metal has long been known to function as an anode in lithium-ion batteries (LIBs) owing to its high capacity, low potential, and effective suppression of dendrite growth. However, seemingly intrinsic degradation during cycling has made it less attractive throughout the years compared to graphitic carbon, silicon-blends, and more recently lithium metal itself. Nevertheless, with the recent unprecedented growth of the LIB industry, this review aims to revisit Al as an anode material, particularly in light of important advancements in understanding the electrochemical Li-Al system, as well as the growth of activity in solid-state batteries where cell designs may conveniently mitigate problems found in traditional liquid cells. Furthermore, this review culminates by highlighting several non-trivial points including: (1) prelithiatied Al anodes, with β-LiAl serving as an intercalation host, can be effectively immortal, depending on formation and cycling conditions; (2) the common knowledge of Al having a capacity of 993 mAh g−1 is inaccurate and attributed to kinetic limitations, thus silicon and lithium should not stand alone as the only ‘high-capacity’ candidates in the roadmap for future lithium-ion cells; (3) replacement of Cu current collectors with Al-based foil anodes may simplify LIB manufacturing and has important safety implications due to the galvanic stability of Al at high potentials vs. Li/Li+. Irrespective of the type of Li-ion device of interest, this review may be useful for those in the broader community to enhance their understanding of general alloy anode behavior, as the methodologies reported here can be extended to non-Al anodes and consequently, even to Na-ion and K-ion devices.


Introduction
Whether driven by the desire for a shift to renewable energy sources, or by geopolitical instabilities, improved energy storage technologies are unambiguously poised to play a critical role in a decarbonized world. The coincidental timing of a global energy crisis (1970s) with the emergence of consumer portable electronics (1980s) would spur the development of an interest in electrochemical energy storage with performance characteristics that were previously unheard of. Owing to its atomic arrangement, lithium-based cells are inevitably at the forefront of the push towards batteries with the highest power densities and best specific capacities. Nevertheless, despite the past ∼50 years of continued advancements for lithium-ion batteries (LIBs), a commitment to developing next-generation Li-based cells with largely enhanced energy/power densities is warranted as these devices are only now seeing the widespread adoption in electric vehicles (EVs) and energy storage systems that are poised to lead to societal transformation [1]. Interestingly, even though excitement abounds for the 'next-generation' of devices, a dichotomy has evolved between industry and academia, with the former taking a ruthlessly conservative approach to innovating a commodity, whilst the latter pushes to the bleeding edge of innovation with constant technological advancements, but is also The β-LiAl can store/release a certain amount of Li without altering its crystal structure, and thus giving a small volume change (e.g. ∼1.4%) [9]. Reproduced from [8].
[© 2007 ECS -The Electrochemical Society]. All rights reserved. (b) A schematic illustration of LIB cells when conventional anodes are replaced by Al foil anodes. Reproduced from [10]. CC BY 4.0.
(admittedly a controversial, but defensible statement). At the industrial scale, compromise has been made to mix high-capacity Si or SiO x alloy powders with conventional graphite as additives (e.g. usually ⩽ 20%), such that the volume change is acceptable while the energy density can be increased to some extent [26][27][28][29][30][31][32].
Recently, the concept of using metallic foils has regained interest as the negative electrode for LIB cells amongst some in the academic community [5,[33][34][35]. The concept of using foil anodes is to omit the usage of binders, Cu foils, etc., and to skip the mixing/baking steps. Not only can the material cost be reduced by occupying fewer chemicals/components, but also the multi-step nature of the anode fabrication can potentially be simplified [10]. When Cu foil no longer exists in LIB cells, its risk of being oxidized or even dissolved, which could lead to internal short-circuit at over-discharging states (e.g. 0 V for a full cell), is also eliminated. Heligman and Manthiram have tested 16 different metallic foils and concluded that In, Sn, and Al foils deliver promising capacities as well as formation efficiencies [36]. Some successful demonstrations of In foil [37,38], Sn foil [39][40][41][42][43], and Al foil anodes [43][44][45][46][47][48] can also support their conclusions. Novel alloy strategies using interdigitated eutectic alloys may further be a way to utilize one or more active metals, whilst maintaining some structural supporting matrix [39,49]. Impressively, some of those foil anodes achieved performances that can compete with or even surpass the composite ones, validating the idea that the future LIB anodes may be entirely solid-state in the form of a simple metal or metal alloy, regardless of the state of electrolyte (i.e. solid or liquid). A simple illustration is given in figure 1(b) to describe the internal structure of a LIB cell which consists of the Al foil anode instead of the conventional one. Although thick foils may sacrifice a certain degree of flexibility, the design can readily be used in prismatic cells. Importantly, the illustrated electrode stack may be extremely suitable for the bipolar design [50], which minimizes the weight/volume occupied by cell packaging, thereby improving the gravimetric/volumetric energy densities. Lastly, this simple foil architechture is suggested to possess great potential in making the future LIBs completely solid state if the cathodes made from solid foil ceramic−metal composites are envisioned.
From a practical perspective, it is also necessary to consider how an anode performs under more extreme testing conditions, particularly with respect to environmental considerations such as temperature and pressure. With electric vehicle adoption being widespread across the globe, cells will be exposed to −10 • C to 50 • C ambient variations which are likely beyond the stable operating range for many electrochemical devices, not just batteries. Therefore, the introduction of thermal stability control is essential. For instance, when the battery pack of Tesla EVs heats up to 55 • C, the built-in cooling system is activated. It is believed that slightly higher temperatures above ambient are preferred, which facilitate a faster ion diffusion (i.e. faster charging), and help reduce the ohmic loss that could lead to thermodynamically favorable conditions for lithium plating, but what is this ideal temperature for ensuring rapid charging and optimal safety at the same time? In the case of carbon anodes and carbonate electrolytes, 50 • C may indeed be ideal, and operation at such a moderately elevated temperature warrants further consideration for any new materials.
High-pressure situations are also effectively intrinsic for both liquid cells and solid-state batteries. Commonly, liquid cells will be stacked and assembled into modules under significant compressive force. With swelling and internal expansion with fixed volumes, compressive stress is inevitable. In the solid-state case, ion migration will be accompanied by continuous structural rearrangement, but to what extent stress helps to reduce the likelihood of lithium dendrite formation is still unclear [51]. Furthermore, compressive stress comes at a significant cost in LIB anode materials. Sethuraman et al highlighted how the effective lithiation potential of silicon is shifted according to the extent of applied stress [52]. While 60 mV GPa −1 may seem insignificant, 50 MPa of external stack pressure on a solid-state battery could readily translate into a stress level that is several orders of magnitude higher between particles in powder-based electrodes and electrolytes. When coupled with internal stresses during lithiation, it can be expected to significantly change the scenario for the effective capacity of the selected anode material. Furthermore, in a solid-state cell, Li-ion transport is restricted to solids and interfaces which have been notoriously complex to understand, let alone optimize. For an anode material, volume expansion and phase transformations are both necessary to have a low potential and optimized specific energy density. An optimization between thermodynamic and kinetic performance should be expected and the classical interrelations between structure-process-property and performance will be as interlinked as ever.

Aluminum (Al) as an anode material
Al, as a well-known engineering material, can be found everywhere in our daily lives from aviation applications to food/beverage packaging and already functions as the cathode current collector and pouch cell case in LIBs. Historically, the Al alloy anode was introduced in the early 1970s to replace Li metal since it was found by Exxon that a LiAl alloy could effectively suppress the growth of Li dendrites on its surface [53]. This cell design was the predecessor of LIB where both the anode and cathode serve as Li hosts with no metallic Li present. Attempts had been made to commercialize Al anodes with some success but the discovery of graphite as a negative electrode by Yazami and Touzain in 1983 would effectively mute such pursuits [54]. Later Yoshino's team put together the graphite anode and the LiCoO 2 cathode in 1985 and officially named it a LIB [55]. Since then, the research focus on anodes had been shifted from LiAl alloy to carbon-based one (graphite) and eventually triggered the successful commercialization and popularization of LIBs.
Despite the growth of carbon as the choice for Sony's LIB anodes, some interests remained in Al or Al-alloys. In the 1990s, Li-Al alloys with just 6% depth of discharge were reported to last for 250 cycles with failure attributed to nonuniform reactions between Li and Al and fine particle formation (i.e. pulverization) [56]. Additives of Mn or Cr were introduced to extend the lifetime to 500 cycles, but challenges with complete capacity utilization remained. Pushing towards full cycling performance with Al metal anodes, it is common that electrode lifetimes of only a dozen of cycles are reported [57][58][59][60][61][62]. The underlying reasons behind the drastic capacity fading at room temperature vary a bit and include (but are not limited to): (1) the ∼95% expansion in lattice volume per Al unit cell (face-centered cubic; fcc) upon lithiation [63]; (2) the mechanical brittleness of the lithiated phases [64]; (3) heterogeneous nucleation with unequally distributed interfacial strain [65,66]; (4) formation of nanopores/voids after β-to-α phase transformation [67,68]; (5) lithium trapping issues [59]. Nevertheless, a few studies can still demonstrate the reasonable cycling of Al anodes [43][44][45]. The dichotomy between excellent and poor cycling performance is well captured by examining the Maxell ML series lithium-ion cells. The LiMnO 2 -Al cell utilizes an Al foil with a surface pattern for the anode, and according to the product datasheet in figure 2, can cycle over 10 3 times if the depth of discharge is kept below ∼10%. However, if the depth of discharge is pushed to >80%, barely a handful of cycles can be expected prior to failure.
In the public domain, the origin of the anomalously good performance of Li-ion cells with Al anodes is somewhat of a mystery, and even though several tens of review articles of Si-based anodes can be easily accessed, a comprehensive picture of the evolution and lithium pathways in Al anodes can hardly be found [70]. Perhaps partly due to this situation, Al is still not considered a promising anode candidate for LIB cells. With this report, the authors aim at achieving a unified understanding of why Al anodes often degrade very quickly but sometimes seem to be reliable, by summarizing the available literature and supplementing with our own previously unpublished work (where appropriate). Not only is the recent progress on Al-based anodes for LIB applications highlighted, but some excellent studies published several decades ago are also acknowledged to provide a comprehensive picture of the topic. We have attempted to write this article in a systematic fashion starting with ( § 2) fundamentals in the binary Li-Al system, explaining the origin of the Al anode capacity from a metallurgic point of view. Then ( § 3) the electrochemical lithiation of Al and ( § 4) delithiation of Li-Al alloys. This is subsequently expanded ( § 5) to consider long-term cycling and ( § 6) important scenarios for operation and usage including anode effective-immortality, moderately elevated temperature operation, and implementation in solid-state cells. Although we have tried to adequately cover the range of contributors to this field, we humbly apologize in advance for any references overlooked or omitted, as this is purely unintentional considering our focus on the Li-Al electrochemical system as it pertains to LIB anodes, as opposed to providing a comprehensive overview of the numerous works in this area.  [69]. The degree of utilization refers to the percentage of Al foil being utilized for storing Li.

Fundamentals of the Li-Al system
Firstly, the physio-chemical properties of Li-Al alloys should be discussed from the perspectives of metallurgy and material sciences, followed by an introduction to their electrochemical nature.

Li-Al binary phase diagram
Al can form various alloys with Li, as shown in the phase diagrams published in the 1980s [71,72]. Later a new phase (Li 2−x Al; orthorhombic) was identified and this was included in the range of compounds that can be formed by the two elements not long ago [73]. By summarizing the literature, a modified phase diagram is given in figure 3, with a specific emphasize on the temperatures from ambient to 350 • C, which should cover the common operation conditions for LIBs. Based on their atomic ratios, these lithiated phases may be expected to give very attractive capacities of ∼993 mAh g −1 , ∼1490 mAh g −1 , ∼1986 mAh g −1 , and ∼2234 mAh g −1 for the formations of β-LiAl, Li 3 Al 2 , Li 2−x Al, and Li 9 Al 4 , respectively [74][75][76][77]. One should be aware, among these four Li-Al phases, two of which exhibit some solubility ranges (dashed lines in figure 3) and these phases can accept/release a certain amount of lithium without altering their crystal structures, i.e. β-LiAl and Li 2−x Al. For example, a noticeable Li solubility range of β-LiAl is characterized to range from 48.5 at% to 55 at% by melting and mixing Li and Al ingots at 500 • C [78], while the latter yields a smaller solubility range between Li 1.92 Al 1.08 (64.0 at%) and Li 2 Al (66.7 at%) obtained from the sample that is annealed at 450 • C [73]. The significance of these solubility ranges is rarely discussed in the context of LIBs, which usually operate below 100 • C, perhaps this is because the information of the phase diagram in low temperatures relies on extrapolations of the data obtained at high temperatures [9]. This review will later elaborate ( § 3.3 and § 4.2) on the solubility range of β-LiAl at room temperature and highlight the long-term cycling capabilities of β-LiAl anode as an intercalation material.

Elevated temperature on the formation of Li-Al alloys in LIBs
Among the Li-Al alloys, β-LiAl is the rightful point to focus on with relevance to the LIB field, since it is the first to form electrochemically at room temperature, while other phases may require elevated temperatures. In a comprehensive study done by Ghavidel et al, it is proposed that the electrochemically driven formation of the higher order phases, such as Li 3 Al 2 and Li 2−x Al phases necessitate working temperatures of at least 35 • C for low rates and 60 • C for moderate rates, and the Li 9 Al 4 can only be expected beyond 100 • C [79]. Figure 4 shows some typical galvanostatic profiles from cycling at elevated temperatures with four distinct potential plateaus at 100 • C [79], referring to the four different phases mentioned above, whilst only the first  [71,72], except the Li 2−x Al phase that is identified by Puhakainen et al in 2010 [73]. The dashed lines indicate the lower boundaries of the solubility ranges [9]. Reprinted from [73], Copyright (2010), with permission from Elsevier. plateau, i.e. formation of the β-LiAl, can be observed at room temperature. Furthermore, Ghavidel et al concluded that by operating at temperatures higher than 35 • C, the specific capacity of Al can nearly double, compared to that of β-LiAl. This two-fold capacity is attributed to Li 2−x Al formation and can be achieved by the common operation temperature range of today's commercial cells. As a result, we have drawn a new bar chart that summarizes the capacities of all alloy anode candidates with extra emphasis on Al-based ones to highlight their roles (figure 5). This review will revisit these higher-order phases later in ( § 3.4 and § 4.1), in the context of phase transformation reversibility and cycling characteristics.

Properties of Li-Al intermetallic compounds
The β-LiAl phase with a cubic structure belongs to the Fd3m space group, and is prototypical to NaTl which consists of two distinct diamond sublattices [78]. The lattice parameter of β-LiAl is ∼6.37 Å, which is roughly 95% bigger than that of the α-Al (fcc; 4.05 Å) [63]. This volume difference can naturally be expected to cause issues in Al anodes where expansion and capacity are linked. Mechanically, both Li and Al are known to be soft or ductile metals. Yet a ductile-to-brittle transition is observed by transforming Al into β-LiAl [64]. This combination of brittleness and stress gradients likely leads to electrode damage during cycling, as evidenced by cracking, delamination, and pulverization [80,81]. Kinetically, Li transport in the β-LiAl is several orders of magnitude higher than that in the α-Al [82,83]. One study even called the β-LiAl 'superionic conductor' , underlining the fast diffusion of Li in its lattice [84]. Recently, a new work suggests that this fast diffusion may be a critical factor in determining the ultimate lifetime of β-LiAl phase anodes ( § 6.5) [85]. The ideal stoichiometric composition of β-LiAl is 50:50 in Li:Al ratio and the figure of 993 mAh g −1 is calculated directly from this ratio. However, from the old phase diagram, there clearly exists a Li solubility range in β-LiAl (as already mentioned), which is maximized at 520 • C (∼11 at%, from 46 at% to 57 at%) [74]. Some studies suggest that this range for β-LiAl would largely be maintained when the temperature is decreased to ambient [86], while others believe it should become significantly narrower, being less than one percent below 100 • C [71]. Since early experiments were done at elevated temperatures, no solid conclusion can be drawn regarding the room temperature solubility (e.g. under electrochemical conditions). A recent study has clarified the electrochemically driven solubility range of the β-LiAl at room temperature and suggested that the range should be roughly 6 at% [9], which is also added to the modified phase diagram in figure 3. Therefore, the fully lithiated β-LiAl (i.e. including the solubility), in fact, holds more Li than the ideal stoichiometry (Li 1.000 Al; 50 at%), yielding a real specific capacity of ∼1152 mAh g −1 (Li 1.160 Al; 53.6 at%) [9]. This capacity should also be considered as the theoretical capacity of Al anodes at room temperature instead of the well-adopted 993 mAh g −1 in future investigations since the rapid Li transport characteristics of β-LiAl effectively ensure saturation of this phase during lithiation.
Unlike β-LiAl, which started drawing people's attention in the 1970s, the higher-order phases have largely been neglected in the domain of electrochemical energy storage since these phases seem to be only approachable at elevated temperatures (⩾35 • C for Li 3 Al 2 /Li 2−x Al and ⩾100 • C for Li 9 Al 4 ) as noted previously, or must be synthesized directly [87]. Most literature regarding the phases with a Li content greater than β-LiAl remains in the fields of metallurgy [73] and theoretical calculations [88], e.g. the minima hopping method [89]. In the end, the authors try to summarize available information regarding Li-Al intermetallic compounds in table 1.

Electrochemical lithiation of Al
Given the dynamic electrochemical environment of a LIB anode, we next consider specifically how Li gets incorporated into Al from an electrolyte-electrode interface, following a temporal sequence of initial lithiation, nucleation of β-LiAl, α-to-β phase transformation, β phase saturation, and formation of higher order phases (i.e. true full lithiation).

Initial lithiation and nucleation of β-LiAl
Al is known to be a highly reactive metal, and its superior stability under ambient benefits from the passivation Al 2 O 3 layer at the scale of several tens of nanometers thick, which is formed as soon as Al is exposed to air. Hence the lithiation of the passivation layer together with the breakdown of electrolyte (i.e. SEI formation) should occur with solid solution lithiation of fcc Al taking place nearly simultaneously. Without a phase transformation, the α-Al can withhold up to approximately 2 at% Li into its crystal structure at room temperature [66]. For practical reasons, overlithiation is likely to occur in many scenarios but is limited due to the extra slow Li diffusion, such that the surface layer of α-Al may reach an upper threshold of ∼2.6 at% [94]. Nevertheless, the initial lithiation of Al is usually a rapid process, indicated by . Gravimetric capacities of various alloy anode candidates at the state of full lithiation, with three Li-Al intermetallic compounds specified. The bar of β-LiAl includes its Li solubility range [9]. All data were collected from [36] except the Al ones that are obtained from [79]. Reprinted with permission from [36]. Copyright (2021) American Chemical Society. the sudden potential drop at the beginning of the galvanostatic profile [95], or the steep decay in current as soon as an overpotential is applied [90]. Nucleation of β-LiAl will proceed after lithium saturation of α-Al (or merely its surface layer), at which time the nuclei are distributed in accordance with numerous factors, including the integrity of interfacial layers, local mechanical compliance, and so forth. From two-dimensional observations, as visualized by  ex situ SEM [80], nuclei have a quasi-circular shape, suggesting some isotropy at the macroscale, although this may not hold true at the atomistic scale, as will be discussed later ( § 3.2.1). Figure 6 shows that the distribution of the β-LiAl patches is indicated by the darker color because of the lower mean atomic number of β-LiAl that scatters fewer electrons as compared to Al under a detector for back-scattered electrons (BSE). The difference in the β-LiAl patch size suggests that some nuclei form before others, which is further affirmed by a series of videos provided by an operando light microscopy study (videos S1-S4) [66]. With a sufficient number of growing nuclei, the phase transformation leads to the α-to-β conversion, expansion, and coalescence of the existing β-LiAl patches [96].
Based on classical theory, an energy barrier (e.g. surface free energy) should be overcome to initiate the nucleation of the new phase, as with lithiated compounds in metallic alloys. The first several nuclei should preferentially form at the positions where the energy barrier is minimum, such as grain boundaries and dislocations [97][98][99]. This extra energy required for the nucleation is electrochemically distinguished by the lower equilibrium potential of the lithiation peak of the 1st cycle in the cyclic voltammogram (figure 7(a)) and the potential dip that is below the plateau at the beginning of lithiation in the typical galvanostatic profile of Al anodes (figure 7(b)). Tahmasebi et al suggest that this overpotential might be a result of the mechanical work for deforming the surrounding α-Al [80] (and consistent with later findings [43]), while the grain size and the disorder of Al are also suggested to be responsible [95]. More specifically, Wang et al managed to reduce this nucleation energy by alloying Cu with Al since the addition of Cu would decrease the grain size and increase the disordered fraction of the Al (Cu) solid solution. 15 at% Cu is sufficient to reduce the nucleation potential to nearly zero while 20 at% can diminish the extra energy required for the nucleation of β-LiAl [95]. Additionally, it has been found in our preliminary experiments that a carbon coating on an Al foil (e.g. TOYAL-CARBO ® ) can also effectively reduce this nucleation energy, although further research is needed to confirm why this is the case.
Although the β-LiAl nuclei distribution is seemly random and partly determined by the free energy on the surface of the electrode, the number of nuclei is found to be correlated with the applied potentials. Geronov et al proposed two modes of nucleation: instantaneous and progressive [96]. When an Al electrode is held at a certain potential below equilibrium vs. Li/Li + , a fixed number of β-LiAl nuclei should form instantaneously, which grow and merge until the end of the lithiation. However, if the applied potential jumps to lower levels (i.e. higher overpotentials), additional nuclei are expected, referring to progressive nucleation. Later both modes were visually revealed under a light microscope [66]. In short, a low potential vs. Li/Li + leads to more β-LiAl nuclei in either mode. For most planar geometries (e.g. films and foils), the nuclei will coalesce within the first Z nm of electrode thickness/depth from the electrolyte interface.

α-to-β phase transformation (plateau; lithiation)
After initial lithiation and nucleation, the Al electrode enters a stable stage of phase transformation. The coexistence of the two phases yields a sharp lithiation peak in the cyclic voltammogram in figure 7(a) and a flat lithiation plateau in the galvanostatic profile in figure 7(b), respectively, indicating the phase transition from α-Al to β-LiAl. The growth mechanisms of the β-LiAl may vary with different electrode designs, so thick foils, thin films, and composites are discussed here.

Lithiation kinetics and mechanisms
For thick Al foils, the β-LiAl nuclei initially form at the interface between the electrode and the electrolyte. These nuclei grow 3-dimensionally (3D), and merge with each other, quickly covering the whole surface of the Al foil. Afterward, 1D growth along the Z-axis must take place until the end of the phase transformation. This 1D growth mechanism is often described by the empirical Deal-Grove model (i.e. surface oxidation of silicon), which are suitable for modeling the lithiation process of various materials when the movement of the interface is reaction-or diffusion-controlled [100,101]. Conveniently, oxygen diffusion in the oxide layer of silicon can be considered equivalent to the Li diffusion in the β-LiAl layer [102]. Therefore, the Li diffusivity together with the overall thickness of the Al foil that is being lithiated (i.e. diffusion distance) determines whether the lithiation process is reaction-limited or diffusion-limited. Thanks to the fast Li diffusion in β-LiAl, lithiation of Al in most cases is limited by the chemical reactions at the phase front between the α and the β phase [82,90]. Table 1 also highlights that the Li diffusivity in β-LiAl is roughly three orders of magnitude higher than that in α-Al.
In the case of Al thin films being deposited on a substrate, the initial growth of nuclei will still be 3D. However, the finite thickness can largely limit the in-depth growth as soon as the β-LiAl hits the substrate. Therefore, a shift from 3D to 2D growth must occur, at which the individual nuclei can impinge with each other. Then each nucleus expands laterally following a quasi-circular shape and starts merging with others until the end of lithiation. The whole process was visualized by a series of videos made by a light microscope (videos S1-S4) [66,67]. It should be mentioned that the Deal-Grove is no longer valid in this case. Instead, Avrami equations are the appropriate tool to model the lithiation of Al films [103][104][105]. A kinetic study of lithiation of Al thin films suggests that this 2D growth of the β-LiAl is also a reaction-limited process [66]. In addition, this 2D expansion is likely affected by the orientations of the surrounding Al grains. The <111> out-of-plane orientation with densely packed Al atoms is found to be more reluctant to lithiation than other textures [65].
Lastly, several studies are working on Al composite electrodes following the slurry-based fabrication approach [106][107][108]. The lithiation of composite Al powder is different, where a circular particle is used as the model system. A comprehensive study has schematically illustrated the whole lithiation process of Al particles (figure 8), in which the reaction-limited process at the phase boundary is also highlighted. Interestingly, this study has also introduced the relevancy of Li-rich (i.e. beyond the β-LiAl) amorphous phases at the end of lithiation [109], which will be discussed later.

Mechanical effects
Owing to the volume expansion caused by lithium incorporation, significant stresses are often associated with almost all alloy anodes, such as silicon and germanium, which will give a nominal compressive stress of >1 GPa upon full lithiation [110,111]. This level of stress can cause certain mechanical issues, thereby significantly affecting the electrode integrity and its electrochemical performances [52].
The mechanical effects of Al are discussed here using Al thin films as the model system. The substrate curvature method is applied in a recent study that has quantified the nominal stress (thickness change is not considered) of an Al thin film electrode deposited on a substrate [67]. It can be seen from figure 9(a) that a compressive stress of ∼500 MPa is generated upon a complete α-to-β transformation. Although the stress level is lower than that of Si/Ge electrodes mentioned above, it is still high enough to cause cracking and delamination of the Al thin film electrode, which are also observed under a light microscope [66,67], as well as an electron microscope [81]. Importantly, one can find that an almost linear buildup of compressive stress is observed in accordance with a constant current and the corresponding conversion at a constant rate. The random nucleation and 2D growth of the β-LiAl for Al thin films, which gives the linear change of the phase fractions of α and β, should largely explain this buildup of the compressive stress [66,67]. It should be  acknowledged that this seemly linear behavior of compressive stress is also different from Si and Ge, which yield a rapid stress increase at the beginning of lithiation (quasi-elastic; interstitial sites) and then followed by a stable stress response (quasi-plastic; yielding point) [52,[111][112][113]. At the later stage of lithiation that is indicated by the blue arrow, a stress kink is observed, suggesting a likely different mechanism of the compressive stress accumulation. This will be elaborated together with the Li solubility range of β-LiAl ( § 3.3 and § 4.2). Additionally, an interesting feature of continuous stress accumulation over cycling is observed in figure 9(b). This observation suggests that the lithiation-delithiation of the Al film electrode is not fully reversible, which will also be discussed in detail later ( § 5.1).
Despite generating insightful results with Al thin films, the substrate curvature method is ill-suited for monolithic Al foils. In fact, the stress behavior of an Al foil is more complicated due to the in-depth lithiation. The initial nucleation and growth of the β-LiAl should replicate that of a thin film. When the surface of an Al foil is fully covered with the β-LiAl, the growth via the Z-axis should take over and dominate the lithiation process. This in-depth phase transition of a partly lithiated Al foil is visualized by SEM images. Again, a BSE detector is used to distinguish the two phases with different mean atomic masses. Figures 10(a) and (b) presents the cross-sectional views with and without a tilting angle at two magnifications, where a rough and corrugated phase interface is observed, which is likely a result of the random nucleation and initial 3D growth [66,96]. Although the overall phase transition should follow a 1D growth mode, inhomogeneities may lead to variations in phase evolution. Yet the overall structural integrity of the Al foil electrode seems to be properly maintained, as can be seen in the photographic images.

Structural changes
On the lithiation plateau, the β-LiAl patches keep growing, causing a continued increase of mechanical stress, which can plastically deform the surrounding α-Al grains. This mechanical deformation can lead to the formation of dislocations and smaller grains [63,65,67]. Overall, the structural changes during lithiation are schematically summarized in figure 11. When an Al thin film is being lithiated, ∼95% volume expansion is inevitable due to the Li inclusion (figures 11(a) and (b)). As mentioned, the internal stress drives the grains to deform and become very small (figure 11(c)). After a complete cycle, an Al thin film electrode with finer grains is generated. The Al grains size is found to decrease from ∼100 nm to ∼5-10 nm after repeated lithiation-delithiation cycling (figure 11(d)) [63]. Despite the graphene coating on the electrode surface, the illustrated microstructure evolution is believed to be representative of Al anodes, which can also be supported by other studies [67,68].

Completion of β phase formation (non-plateau; lithiation)
One can notice that the lithiation plateau in the galvanostatic charge and discharge curves no longer exists at the end of lithiation, regardless of the electrode architecture, e.g. Al films (figure 7) and Al composites (figure 8), or even Al foils in [36]. This regime of the last ∼20% of the lithiation has been rarely discussed in the literature, and with some inconsistencies when it is mentioned. Various explanations can be found in the existing literature, such as Li solubility of β-LiAl [9], amorphous Li 1+x Al phases [109], and Li 3 Al 2 /Li 2−x Al/Li 9 Al 4 phases [60,63].

Saturation of β-LiAl
Although there is almost no question that β-LiAl exhibits a solubility range at elevated temperatures, the inconsistency comes from the lower temperature regime. As evidenced by multiple versions of the Li-Al phase diagrams, different extrapolations to room temperature have been suggested, ranging from less than one percent (ca. 49.2-50.1 at% Li) [74] to a solubility range of 9.2 at% Li [114]. Recently, a study is conducted specifically on characterizing the solubility range of β-LiAl at room temperature, determining a ∼6 at% range of Li solubility within β-LiAl (figure 3). As shown in figure 12, this solubility range can yield ∼242 mAh g −1 specific capacity, normalized to Al, which is suggested to be largely correlated with the non-plateau regime [9]. When the electrode contains a single phase, i.e. β-LiAl, its potential would decrease with additional Li inclusion, rather than stay at the plateau of α/β coexistence.

Mechanical effects
Apart from electrochemistry, this non-plateau regime also gives rise to a different path of mechanical evolution. As mentioned above, in figure 9, a kink in the stress signal appears when the potential plateau approaches its end, after which the growth of compressive stress seems to show a higher slope against time. The origin of this faster stress growth at the end of lithiation is still unclear. One explanation may be that it is a result of saturating the solubility range within the β-LiAl as the overall volume change may strain the substrate more significantly than the two-phase coexistence where the stress is restricted at the phase interface [9,67]. This is complicated by the facts that the hardness ( figure 13(a)) and yield strength ( figure 13(b)) of β-LiAl are known to vary markedly across its composition range, but also that significant defect generation and work hardening occur in the residual α-Al matrix [43,67]. More information and discussion will be provided in § 5.1.

Formation of amorphous Li 1+x Al phases
It is still questionable whether the formation of Li-rich phases (i.e. Li content higher than β-LiAl) is occurring at room temperatures. Although these phases are only supposed to exist at elevated temperatures, they are mentioned from time to time in the literature. For instance, both Li 3 Al 2 and Li 9 Al 4 peaks seem to appear in the x-ray diffractograms obtained ex situ [60]. Another study claims that Li 3 Al 2 is detected from a lithiated Al film, in addition to β-LiAl [63]. Yet some other studies claim that Li-rich phases can hardly form at room temperatures [79,80]. Regarding these inconsistencies, Li et al have commented that these Li-rich phases may be formed, but are only limited to a very thin layer on the electrode surface and the overall equilibrium is still α/β [43].
Apart from mentioned above, a study of Al composite electrode (figure 8) proposes a very different mechanism for lithiation [109], especially for the end of lithiation (non-plateau). An in situ x-ray diffraction (XRD) test was done neatly in the same study, providing a series of diffractograms covering the whole lithiation-delithiation processes ( figure 14(a)). At the first glance, the XRD peaks of crystalline Al show reasonable trends with high intensities at the beginning of lithiation and the end of delithiation. The minimum is observed at the end of lithiation where all Al is consumed. Differently, the <111> LiAl peak exhibits a noteworthy behavior: Its intensity in the diffractograms reaches the highest at the end of the lithiation plateau (i.e. ∼80% of the lithiation), and then sharply drops to nearly zero when lithiation finishes. The <220> LiAl peak gives a similar trend, but its intensity decreases more gently and is non-zero at the end of lithiation. It becomes obvious when the peak intensities are extracted and plotted in figure 14(b). Based on these observations, it is interpreted that over-lithiated phases are forming at the expense of the β-LiAl during this regime. Since the β-LiAl and Al peaks are the only ones that are detectable under x-ray, these over-lithiated phases are generalized to be Li 1+x Al and suggested to be amorphous [109], but likely such trends are restricted to particle-based Al anode materials.

Formation of Li-rich phases (extra plateaus; lithiation)
Different from the amorphous Li 1+x Al phases proposed by Qin et al that may occur at room temperature [109], this section focuses on the Li-rich phases (i.e. Li content higher than β-LiAl) that are present in the modified Li-Al binary phase diagram (figure 3) that are expected to be approachable at elevated temperatures. As already mentioned, unlike the β-LiAl, only a few studies have investigated these higher-order phases electrochemically.    β-LiAl capacity) lithiation rate, when the temperature is above 35 • C and 60 • C, respectively. The formation of Li 9 Al 4 , on the other hand, requires a temperature above 100 • C. It is also suggested that the formation of Li 3 Al 2 and Li 2−x Al appears to be linked while Li 9 Al 4 seems to be governed by a different formation mechanism [79].

Opportunity for moderate temperatures
In our daily lives, almost everyone has experienced their smartphones or other portable electronics to heat up during active use and charging. Indeed, even with ambient temperatures of 26 ± 0.5 • C, smartphones can easily reach beyond 43 • C from app usage alone [116]. Considering larger-scale applications of LIBs, such as EVs and stationary energy storage, overheating of the battery pack could be more problematic, leading to drastic incidents like fire catching or even explosion. The situation explains why the sophisticated cooling system is the key component for commercial EVs. Of course, high temperatures, e.g. >100 • C, are beyond the stability of LIB electrolytes, but operating LIBs at slightly elevated temperatures is acceptable or even preferred due to a faster ion diffusion. A commonly acceptable temperature range, as mentioned in the data sheets of many commercial LIB products, is naturally limited to moderate levels (⩽60 • C) to ensure safe and reliable operations [117].
Recently, one study only focuses on the moderate temperature regime and compares the Al thin films that are lithiated at different rates and temperatures. With a small lithiation current of 24 µA cm −2 ( figure 16(a)), the three plateaus referring to the formation of β-LiAl, Li 3 Al 2 , and Li 2−x Al can be clearly observed at 60 • C and 50 • C with the y-axis in a logarithmic scale. However, when the temperature is further decreased to 40 • C, the second (Li 3 Al 2 ) and the third plateau (Li 2−x Al) seem to merge with each other, yielding a long plateau until reaching a specific capacity level that roughly matches the Li 2−x Al one. At 40 • C, when the Al film has been transformed to the β-LiAl (i.e. >47.8 at% Li or >910 mAh g −1 Al), the lithiation potential drops to a level as low as ∼4 mV vs. Li/Li + and slowly recovers to ∼7 mV vs Li/Li + . This potential level is sufficiently low (i.e. <15 mV vs. Li/Li + ) to facilitate the simultaneous formation of Li 3 Al 2 and the Li 2−x Al. This experiment is also repeated at a higher lithiation current of 48 µA cm −2 and the results are shown in figure 16(b). Although the galvanostatic profile at 60 • C almost maintains the same features, the potential plateaus of the Li 3 Al 2 and the Li 2−x Al start merging at 50 • C under this higher rate rather than at 40 • C under the lower rate. Remarkably, the phases that are beyond β-LiAl can hardly form at 40 • C and the lithiation effectively stops with the β-LiAl composition [118].
In the end, what enables Al to reach its full capacity (figure 1) at room temperature is unclear. Certainly, diffusion limitations in the highly ordered Li-Al phases must be considered, but as evidenced by the initial α → β phase transformation, nucleation barriers/potentials can also be expected to play a role. Therefore, the purity of Al (or amount and type of impurities, rather) may have a very big impact on the ability to realize the phases with the highest capacities. This should be anticipated as a very exciting area for research, not just because of the theoretical merits, but also because of the practical implications for the world's most produced base metal [119,120].

Electrochemical delithiation of Li-Al alloys
In this chapter, the reverse phase transformations, i.e. Li extracted from Li-Al alloys electrochemically, are elaborated. The delithiation is presented in different ways depending on whether the fully lithiated phase is Li 9 Al 4 , Li 2−x Al/Li 3 Al 2 , or β-LiAl, with discussions following the trend of the most lithiated, to the least.

Delithiation and cycling of Li-rich phases (extra plateaus; delithiation)
Owing to their difficulty in the formation of Li 3 Al 2 , Li 2−x Al, and Li 9 Al 4 , knowledge about the delithiation pathway from higher-order Li-Al phases to lower-order ones is rather limited. Nevertheless, the extremely low standard potential of Li 9 Al 4 , relative to Li/Li + , should make reversible cycling a challenge. Indeed, as noted by Kim et al [87], 'It is apparent that Li tends to deposit and grow around Li 9 Al 4 ' . So, while from a gravimetric/volumetric capacity perspective, such a phase is probably interesting for primary battery applications, the reversibility of the phase transformations poses a long-term challenge for secondary lithium-ion cells. The Li 3 Al 2 and Li 2−x Al phases, on the other hand, offer excellent capacities but should allow for delithiation and reversibility, at least at elevated temperatures as the delithiation of Li 2−x Al and Li 3 Al 2 corresponds to 0.04-0.05 V vs. Li/Li + for the former and 0.1-0.2 V vs. Li/Li + for the latter. As noted in § 3.4, the extent of thermal energy required to form these phases may only be slightly above room temperature, but this leads to the natural question: Can these phases be cycled reversibly?
Recently, a new cycling approach is proposed to cycle the Al electrode at moderate temperatures, which does not trigger the formation of Li 9 Al 4 while the α ↔ β phase transformations are also intentionally prohibited. As a result, as shown in figure 17(a), the electrode is only cycled between β-LiAl and Li 2−x Al by controlling the temperature and restricting the potential window. Since the phase transformations involved in this new cycling approach are different from the well-studied α ↔ β ones, the cycling behavior is expected to be also different due to different material properties. The outcome turned out to indeed be true, that the cyclability of β-LiAl ↔ Li 2−x Al is significantly better than that of α ↔ β ( figure 17(b)) [118]. Not only does this finding open up opportunities in real applications as the heat generated by ohmic resistance must be dissipated during operation, but also provides new topics to researchers who are interested in Al-based anodes that have greatly improved stability (β-LiAl ↔ Li 2−x Al cycling) and capacity (1986 mAh g −1 Al for Li 2−x Al).

Delithiation of β-LiAl (non-plateau; delithiation)
As β-LiAl is often considered the fully lithiated phase in the LIB field, we return to this section by focusing on the delithiation of β-LiAl at room temperature. Two scenarios of delithiation are discussed: (1) the desaturation of β-LiAl (non-plateau) in this section and (2) the β-to-α phase transformation (plateau) in the next.
Assuming the β-LiAl is the only relevant phase present, the initial delithiation begins with a depletion of the Li while maintaining the crystal structure of β-LiAl [9]. Electrochemically, this desaturation is indicated by the potential increase at the beginning of delithiation, occupying roughly 20% of the β-LiAl capacity prior to the delithiation plateau ( figure 7(b)). This number is generally in agreement with the one characterized by the charging counting method in figure 12, i.e. 242 mAh g −1 -Al. It is still uncertain whether the Li concentration is homogeneous over the whole electrode. Therefore, the desaturation process is possibly location-dependent, e.g. some parts of the electrode may be desaturated prior to others. However, given the good diffusivity of Li in β-LiAl, homogeneity may be expected. When the Li concentration within the β-LiAl approaches its minimum (e.g. 47.8 at%) [9], the nucleation of, and reversion to, α-Al must occur.
The solubility range shifts the lattice parameter of the β-LiAl unit cell, as Li atoms are being inserted/extracted while maintaining the same crystal structure. The shift in lattice parameter as a function of the Li content within β-LiAl was studied and summarized in table 2 [78]. As can be seen, the lattice parameter of the cubic β-LiAl unit cell ranges from ∼6.3902 Å to ∼6.3612 Å when the Li-rich β-LiAl is desaturating to form the Li-poor β-LiAl. Also, insertion/extraction of Li from β-LiAl varies the density, giving >0.1 g cm −3 difference between the Li-rich and the Li-poor β-LiAl. A corresponding asymmetrical mechanical evolution of the electrode is observed during the desaturation of β-LiAl ( figure 9(a)). One can notice that a sudden increase of the stress towards the tensile direction is observed as soon as the current direction is reverted. This is an indication of the overall volume contraction of ca. 1.4% within a single phase, as noted above, that strains the substrate drastically, causing this substantial stress change. After the desaturation, however, the stress change goes to nearly zero, but with some residual deformation preventing it from fully relaxing to the initial state.
Hudak and Huber have suggested that the delithiation processes of Al anodes should be largely responsible for the drastic capacity fading. They observe that the lithiation capacity always matches the delithiation capacity of the previous cycle while the delithiation capacity is usually a fraction of the previous lithiation capacity [61]. Later, an operando light microscopy study makes it clear that the delithiation process causes cracks formation within the lithiated regions of Al thin films, likely due to the mechanical brittleness of the β-LiAl and volume shrinkage. This electrode damage may induce further embrittlement and loss of material. From SEM images and videos S5-S8, cracks can be clearly observed after delithiation, but are notably restricted within the β-LiAl region (figures 18(a) and (c)). The images for the same regions taken by a BSE detector highlight the distribution of the β-LiAl patch (darker color; figures 18(b) and (d)) [67]. Overall, this process includes elongation of the face-centered cubic Al gains during lithiation (figures 18(e)  and (f)) [80], and a grain rearrangement during delithiation ( figure 18(g)), leading to pores left within the delithiated Al matrix.
Although it is suggested that these issues might be less pronounced in bulk Al foils without a substrate-electrode interface [67], the formation of pores after delithiation is also observed. While in general, porosity should be expected to be detrimental to the host, as noted in [121], perhaps the process of pore formation can be controlled at the atomic or microscale to avoid degradation. As illustrated in figure 19, if Li is extracted from the lithiated layer in the first delithiation process, the delithiated Al matrix could be self-organized to form a columnar structure with cleavages and micropores.
As discussed before, Al grains become quite small due to the plastic deformation induced by the larger lattice volume of β-LiAl. Therefore, the Al grains should become smaller and smaller after each successive delithiation (figures 11(c) and (d)). This grain refinement process is also evidenced by the in situ XRD results. In figure 14(b), it is obvious that the peak intensity of the <111> plane of Al becomes weaker after the first cycle, indicating a weaker crystallinity. By taking a closer look at the full set of diffractograms ( figure 14(a)), not only does the intensity of <200> Al exhibit a similar trend, but the peak widths in XRD of both planes seem to be also increased, giving much lower peak-to-width ratios or higher full width at half maximum, which are indicative of smaller crystallite sizes (Scherrer equation). In fact, the delithiated Al is theoretically expected to have a small grain size due to the existence of pores. The typical bicontinuous structure of the porous matrix consists of ligaments or columns, of which the diameter is usually smaller than 1000 nm [121]. Hence, in the case of delithiated Al, the grain size cannot be expected to be larger than this value.
The mechanical stress evolution is surprisingly flat on this delithiation plateau ( figure 9(a)), which means that the mechanical stress remains virtually unchanged. Assuming the β-to-α phase transition is ongoing, the stress change should be only limited to the phase interface that is moving through the volume, which should result in a similar mechanical profile as seen with the lithiation plateau. The observed porous Al formation can well explain this constant stress profile during delithiation. The newly formed pores leave a significant number of voids inside the delithiated electrode, which substantially restrict the volume contraction, and thus do not strain the substrate [67]. Just like other dealloying processes, the selective dissolution of an element from an alloy is considered a common approach to producing porous metals [122]. For example, Chen and Sieradzki have provided SEM images of porous Sn, Pb, Cd, and Bi that are fabricated via electrochemical delithiation [121].

Subsequent electrochemical cycles (α ↔ β)
Although the detailed analyses for the initial cycle help gain fundamental insights into Al anodes, the enablement of Al as an anode candidate for LIBs requires some cyclic stability. For brevity, instead of going through the processes step-by-step, issues and differences stemming from subsequent cycles, in addition to the initial one, are highlighted.

Secondary SEI formation in liquid cells
The porous structure of the delithiated Al leads to a significant increase in the specific surface area. If liquid electrolytes can penetrate these pores and voids, reaching fresh Al surfaces, then it can be expected that it will facilitate more secondary surface reactions. This secondary SEI formation can partly explain the coulombic inefficiency of Al anodes reported in the literature [59,61,106,123]. It should be noted that these reactions are not one-off processes and would occur as soon as fresh Al is exposed to the electrolyte. When an Al anode is going through repetitive cycles, the surface morphology of the electrode is completely changed (figures 20(a) and (b)), from a flat/dense to a porous/loose structure. At the cross-section (figures 20(c) and (d)), the film thickness expands from several hundreds of nanometers to ca. 20 micrometers (a factor of more than 40), explaining the compressive stress built up over cycling (film thickening) observed in figure 9(b). Taking a closer look at figures 20(e) and (f), the nanostructured Al ligaments and columns will become thinner and thinner due to the irreversible formation of the pores, such that there should be always fresh Al surfaces forming during each delithiation [67]. As a result, an optimized coulombic efficiency is difficult to achieve as long as the formation of porous Al is taking place during each delithiation.

Ductile-to-brittle transition
Although Al is known to be a soft metal with decent ductility in its original state, a ductile-to-brittle transition is expected in the presence of pores and/or under tension, just like nanoporous gold [124,125]. This can be evidenced by an interesting SEM image of a partly cycled Al foil. In figure 21, a piece of porous Al (delithiated after ×10 cycles) is falling off from the bulk Al foil, indicating its brittle nature that is more susceptible to deleterious defects. The darker color of the upper (i.e. cycled) layer under a BSE detector indicates a very high level of porosity, where the mean atom density is diluted, thus scattering the electrons less significantly. This observation also allows us to revisit a recent study that relies on backscattered images [35]. The color contrast might not fully indicate the distribution of the β-LiAl but could also be the delithiated Al. Moreover, the brittleness may cause Li trapping issues, which is closely related to the poor cyclability of Al anodes when α ↔ β transitions are engaged. A model developed from the electrochemical data explaining the origin of Li trapping: The delithiation of β-LiAl and the formation of α-Al initiates at the electrode-electrolyte interface while the in-depth β-LiAl may be encapsulated [59]. However, microscopy evidence does not seem to support this model, especially since the β-LiAl is considered a superionic conductor and facilitates fast Li diffusion [84]. Therefore, it is suggested that the loss of active material may be largely responsible for the Li trapping, referring to 'dead Li' rather than the Li encapsulation described in the model.

Safety concerns
With lithiated Al at room temperature, i.e. β-LiAl, it may be expected to be very reactive to oxygen and moisture due to the high Li content (ca. 50 at%) and perceived to be at risk of safety issues. However, recent studies have highlighted the air compatibility of β-LiAl upon some surface modifications, such as an artificial SEI [126] and a shot-peening treatment [127]. The β-LiAl electrodes with modified surfaces can be quite stable under ambient for several hours. Even when obvious oxidation is observed after days of air exposure, the β-LiAl electrodes are generally stable without any drastic reactions, e.g. smoke or fire. We believe that, reasonably, safety concerns should come from the delithiated Al since there is almost no question that dense Al foils/films in their pristine state would evolve over cycling to form a highly porous Al matrix. Overall, the unsolved problems of Al anodes are all closely related to the formation of pores and seemly intrinsically bonded to the α ↔ β phase transformations. Despite the porosity bringing the problems such as brittleness and volume expansion mentioned above, the huge surface area of the cycled electrode can lead to fire-catching upon air exposure (observed experimentally).
In another safety dimension, the omission of copper current collectors in a full cell with an Al anode raises the prospects for zero-volt stability in a lithium-ion cell. Previously considered a key advantage of Na-ion batteries over lithium, the ability to deeply discharge a cathode, such that the anode potential is raised to that of the cathode will almost certainly lead to the dissolution of copper, evident with LiFePO 4 (LFP) cathode cells and made worse with LiNi x Mn y Co 1-x-y O2 (NMC) varieties [128]. As presented in figure 22(a), Cu foil is severely corroded when the potential is above ∼3.6 V vs. Li/Li + , as indicated by the sharp CV oxidation peak (∼17 mA) and the photographs of the disassembled cell. Not only does the Cu foil pulverize to lose its initial shape, but the corroded Cu powder falls off and sticks to the separator. Whereas Al foil exhibits a much weaker oxidation peak in its cyclic voltammogram ( figure 22(b)), even when the α-Al becomes porous, (i.e. delithiated). Furthermore, the appearance of Al foil remains unchanged via eye observation when it is cycled between 1 and 4 V vs. Li/Li + (no lithiation of Al) and the CV current is ∼1000 times smaller than that of Cu foil (inset of figure 22(b)), indicating its superior stability at the high potentials vs. Li/Li + (e.g. an over-discharged full cell).

Strategic utilization of Al anodes
By reviewing most of the works that show superior cycling stabilities of Al anodes and comparing them to the ones with poor cycling performances, it can be noted that stable performance of Al anodes is often observed in the full cell configuration [9,[43][44][45][46][47]129]. When Al anodes are characterized in half cells (i.e. Li metal as the counter electrode), the cycling life is usually restricted to a dozen of cycles [57, 60-62, 67, 80, 95, 106, 107, 123, 130, 131]. With this in mind, it is important to rationalize such behavior in light of the 'full picture' of the Li-Al electrochemical system.

Origin of anomalous performances
To explain this inconsistency between full and half cells, here we introduce the concept of 'DAU' , which is short for 'Degree of Al Utilization' . Ryu et al have partly clarified the importance of DAU, and how the DAU affects the cycling performance of Al anodes [46]. Said briefly, the shallower the lithiation depth (or the lesser the DAU), the longer the cycling life of an Al foil is, similar to what has been discussed together with the Maxell LiMnO 2 -Al cells in figure 2. Several studies have suggested that Al anodes can benefit from partial lithiation because of the heterogeneous nucleation and the ductility of Al [66,67,80]. Returning to the inconsistency, the primary difference between a full cell and a half cell is the Li supply. When a Li foil is used as the counter electrode, the Li supply within the half-cell system is considered infinite. As a result, the Al electrodes will be fully lithiated and transformed into the β phase completely, regardless of whether they are films, foils, or composites. Based on the (de-)lithiation mechanisms presented before, it seems quite reasonable that the Al anodes fail prematurely in these half cells. Yet in a full cell, the Li amount is limited by the cathode. If one carefully revisits the above-mentioned full cell studies that give several hundred stable cycles, the DAUs are usually very low, e.g. 5.5% and 3.5% [44]. This means that most of the Al anode is still fresh Al, which is mechanically more stable and capable of storing significantly more Li, and additionally acts as a consumable Al source, allowing for a minor, but effectively undetectable degradation with each cycle. It is therefore not surprising that a stable cycling performance is achieved. All these DAUs are extracted and summarized in table 3 together with the cell parameters to allow easy comparisons. It then seems logical that a lesser DAU should give a better cycling life to Al anodes, or to any other electrode materials, which helps explain why long cycling is only observed from the full cells with Al anodes, but not from the half cells.
Considering the situation, Chen et al try to benchmark the degradation behavior of Al foil anodes for LIBs, by testing a total of 60 half cells. In figure 23(a), a correlation is established between the number of cycles before failure and the areal capacity per cycle of Al anodes, which, in general, agrees with a better cycling life at a lesser DAU or vice versa. More specifically, figure 23(b) presents the same data, but with both axes in the logarithmic scale. Regardless of whether Al foils are pure or alloys, the correlation seems linear with error bars obtained from multiple cells. Upon understanding why and how DAU affects the performance of Al anodes, the vital role of prelithiation can also be highlighted in full cell applications. If an Al anode is not prelithiated, a certain amount of Li atoms from the cathode may be trapped (e.g. cyclic SEI) during the delithiation of β-LiAl, and subsequently resulting in cell degradation. Pre-storing Li atoms into  the Al anode (i.e. prelithiation) can effectively mitigate this issue since Li stock should no longer be a limiting factor. More details regarding prelithiation techniques can be found in § 6.3. Overall, a low DAU might not be acceptable for some anode candidates since it directly dilutes the specific capacity of a full cell. Al foil anodes, however, should be critically assessed due to the unique situation that they can potentially function as both active material and current collector. More importantly, the usage of Cu element can be completely omitted in the cell, leading to a roughly one-fifth reduction in material cost [132], while electrode manufacturing is expected to be simplified [10]. Cu is known to be a relatively dense metal with a density of 8.96 g cm −3 , which is about 3.3 times higher than that of Al. These facts may make the Al foil anodes with relatively low DAUs acceptable in real applications, especially since the elimination of electrochemically inactive components (e.g. Cu foil, binder, etc) provides extra room for allocating more Al in the LIB cell design.

Mechanical strain engineering
On the other hand, some studies in table 3 show great promises in real applications, regardless of the DAUs. Hong Li has done a practical evaluation of LIBs and stated clearly that the areal capacity of commercial batteries is 3-4 mAh cm −2 , likely by considering two sides of the electrode [12]. Therefore, the LiCoO 2 -Al cell made by Li et al [43] can be quite competitive by delivering ∼2 mAh cm −2 one-side capacities without capacity fading after 120 cycles. By taking a closer look, their work draws some explanatory illustrations that help the readers easily understand the lithiation mechanisms of different Al foils (figure 24). Two modes are proposed here: (1) when uniform lithiation is achieved, the mechanical strain generated by volume expansion is mostly elastic and restricted at the phase interface, such that the overall electrode integrity is properly maintained (left column); (2) when a stochastic lithiation occurs at the surface, huge strains are generated locally causing dislocations, and eventually pulverization of the electrode (right column). The core finding here is that the Al foil electrode with the uniform lithiation pathway may circumvent catastrophic, localized strains. Although this was illustrated with strain engineering, selective doping or alloying, may offer a chemical pathway. For instance, additions of Mn or Cr are known to improve the cycle life of Al anodes [56], but further studies are certainly warranted. Nevertheless, it is clear that, under the right circumstances, the α ↔ β phase transformations can be quite stable during cycling.
Extending some of these concepts to lithium-ion solid-state batteries, the underlying phase transformations should be equivalent to liquid cells. However, the intense applied compressive stress combined with solid electrolytes could change the strategy for the utilization of Al as an anode material. Particularly considering a recent work by Sakka et al the significant void volume fraction in powder-based solid-state cells is inversely related to externally applied pressure, as shown in figure 25(a) [133]. Taking a closer look, this stack pressure will translate to the pressures among circular particles that are several orders of magnitude higher due to extremely low contact areas ( figure 25(b)).
Following this, it has already been noted with in situ x-ray tomography with CuS (also prone to expansion during lithiation) that 'crack formation could be effectively suppressed by increasing the stack pressure from 26 to 40 MPa' [134]. Further complicating the strain scenario is the contribution from internal expansion. For micro Si, this could reach ∼62 MPa and for LiIn, this was measured and estimated to be ∼6 MPa of axial stress (after accounting the for the mismatch between the calculated and experimental values (figure 26(d)) [135], similar to what was measured for CuS [134]. This strongly supports the concern that the delicate mechanical aspects of moving through the relevant phases of the Li-Al system with its range of structures and properties. Adding to this is the interesting case of β-LiAl, which is known to increase its hardness (figure 13(a)) [115] and yielding strength (figure 13(b)) [64] as lithium content is increased within its solubility range. Therefore, the prospects for Al anode utilization in a solid-state battery certainly warrant further research and exploration.

Prelithiation
Prelithiation is known as one of the common ways to improve the performance of Li-based energy storage systems [126,129]. Ryu et al presented a facile but effective method to prepare the prelithiated LiAl anodes for LIB applications. As illustrated in figure 27(a), Li and Al foils are mechanically pressed together to form the LiAl alloy using a liquid electrolyte as the intermediate. The LiAl grains are characterized to be nanocrystalline (figures 27(b) and (c)). At the cross-section (figures 27(d)-(f)), a very homogeneous layer of the β-LiAl can be observed on top of the Al foil. Such a mechanically prelithiated Al anode delivers 200 stable cycles when pairing with an LFP cathode. It should be mentioned that the DAU of the Al anode is ∼26%, which is the highest among all the Al anodes with cycle lives longer than 100 [46]. Not only does the mechanically prelithiated Al anode give impressive cycling performance at a relatively high DAU level, but more importantly, the facile preparation method can easily be scaled up to fulfill industrial requirements. In the absence of mixing-pasting-baking processes, the fabrication of this Al-based anode becomes more competitive both financially and technically. Additionally, labor costs and energy consumption can be potentially reduced without sophisticated heating processes [10].
Of course, prelithiation is not without trade-offs, particularly when most of the lithium is used in the anode for what is effective for structural purposes (stabilization of β-LiAl, described below in § 6.4). For It is important to note that an axis transverse to the fiber is an axis perpendicular to the main symmetry axis (c∞) and therefore the axis 'axial' to the cell is a 'transverse' axis to the fiber. To simplify nomenclature, every time we herein mention 'longitudinal' or 'transversal' will be respected to the fiber and 'axial' will only be respected to the cell. (b) Time-resolved voltage (top), external cycling pressure (middle), and an internal optical signal (bottom) for the aforementioned ASSB cycled at C/30 (5.83 mA g −1 ) and 25 • C in an operando mode. (c) 2D stack view of the reflected spectra, with the corresponding galvanostatic charge/discharge cycle. The charge and discharge processes are plotted in red and blue, respectively. (d) Comparison between operando stress evolution obtained: 1-internally by the FBG sensor and using the mathematical model (green curve) and 2-internally by the FBG sensor and the sensitivity coefficient obtained with the experimental calibration of the sensor (blue curve). The respective galvanostatic charge/discharge is presented (top). The points at the beginning/middle/ end of the charge/discharge are indicated by colored dots, also marked in the corresponding FBG spectra in (d). The external cycling pressure was fixed at 2 MPa prior to performing the battery cycling. Reproduced from [135]. CC BY 4.0.  example, an anode that is limited to β-LiAl can have ∼6% at use for reversible cycling capacity, but ca. 48 at% Li must be used just to form the Zintl structure first. This factor of ×8 for inactive Li is certainly a detractor from the other advantages of this electrode material.

Utilization of β-LiAl solubility
A scientific aspect regarding the Li solubility range of the β-LiAl should be considered, especially for prelithiated Al anodes. As already discussed, the β-LiAl layer of a prelithiated Al foil should exhibit a solubility range that can potentially give a capacity of ∼242 mAh g −1 -Al without a phase transformation. When the β-LiAl crystal structure is maintained (illustrated in figure 28(a)), the cycling life should be theoretically infinite since the issues caused by the α ↔ β phase transformations, such as localized strain, pores formation, and secondary SEI formation, are circumvented or mitigated (∼1.4% volume expansion can be expected between the ends of the β-LiAl solubility range). Figure 28(b) shows the experimental evidence for the argument above, where 500 stable cycles are achieved at various rates for a partly lithiated Al foil. Through reviewing the details from the study, the DAU of the system is translated to be only ∼2%, but it offers a unified understanding of why Al anodes sometimes last hundreds of cycles in full cells. Also, the study suggests that the DAU can be further increased (an optimum value should exist) but necessitates systematic investigations. For instance, a control experiment was done in the same study using a 0.18 mm Al foil to replace the original 0.25 mm one without changing the prelithiation depth (i.e. 0.1 mm is prelithiated) and other cell components, yet both exhibit similar cycling performance [9].
Apart from the applications in various devices, the role of solubility range can also help shed light on the cycling performances summarized in table 3. For the full cell with a DAU lesser than 20%, the solubility range of the β-LiAl should be the primary capacity source, thus giving the Al anode a long cycling life. This Figure 29. Mapping of cycle life for Li4Ti5O12//β-LiAl(Al) cells prior to ⩾80% capacity retention at different current densities. The horizontal lines indicate the common charging current densities for commercial EDLCs and LIBs, considering cell areal capacity of ∼1.5-2 mAh cm −2 (single-side) and the common charging C-rate of 1.5 C. [85]. Reprinted from [85], Copyright (2023), with permission from Elsevier. CC BY 4.0.
hypothesis may be partly supported by the evident difference between the theoretical and the reversible capacity in one of the studies [44]. For instance, the LiCoO 2 -Al, NMC-Al, and LFP-Al should have theoretical capacities of ∼270 mAh g −1 , ∼284 mAh g −1 , and ∼170 mAh g −1 , respectively, normalized to cathode mass. However, the reversible capacities for the three cells are ∼160 mAh g −1 , 170 mAh g −1 , and 126 mAh g −1 [44], suggesting that not all the lithium in the cell is being utilized and the β-LiAl solubility likely plays important roles in these cells.
Regarding the volumetric capacity of β-LiAl, the headline figures are likely not as attractive as it is an insertion/intercalation material. By its nature, this phase is acting as a scaffold for the reversible insertion of Li and thus with only ca. 6 at% of the crystal used for cycling, the number of charges stored per unit volume of the host will not compete with conventional anodes, e.g. graphite. However, the limited volumetric capacity comes with minimal volumetric expansion and relatively fast Li transport. Hence the value proposition of a β-LiAl anode is unlikely to compete directly with alloy anodes in the same application space and/or cell design. Perhaps a better comparison in this regard is with respect to lithium titanate (Li 4 Ti 5 O 12 ), which is also an insertion material with minimal volume expansion and despite no longer being used in EVs, is still seeing healthy market growth by Toshiba for ferry electrification and stationary power supplies.

Targeting effective-immortality of β-LiAl
Regardless of whether the Li solubility range of β-LiAl plays a primary role, a 'user guide' for prelithiated Al anodes is provided in figure 29. This demonstration is built upon a full cell design using Li 4 Ti 5 O 12 (high rate capability, high cycle stability) and β-LiAl(Al) as the cathode and the anode, respectively. The cell capacity is roughly 1.25 mAh cm −2 , determined by the theoretical capacity of the commercial Li 4 Ti 5 O 12 electrode (NEI Corporation). Through cycling 25 cells with various charging currents and anode prelithiation depths, it is suggested that, considering ⩾80% capacity retention after 400 cycles as a key indicator, Al anodes with a nominal prelithiation depth lower than ∼50 µm (i.e. 50 µm Al is lithiated) only survive under low-rate charging conditions (below ∼2 mA cm −2 ). When the prelithiation depth is increased to higher than ∼50 µm, however, the charging rates seem less relevant, giving ⩾300 cycles at a rate as high as 10 mA cm −2 . The variations in performance are suggested to directly stem from Li diffusivity in β-LiAl. Overall, figure 29 highlights different ways of using Al anodes aiming at different application purposes. For instance, a low prelithiation depth may be sufficient for batteries that are always charged at low rates. Yet, when targeting the effective-immortality of an Al anode, higher prelithiation depths and low charging rates are suggested to be always beneficial, as the β-LiAl solubility is the main contributor for Li (de-)intercalation (i.e. no phase transformations). The raw cycling data can be found in the original work [85].

Asymmetric thermal behavior
Whether the investigation on Al anodes in LIBs should remain focused on the β-LiAl phase or shift to the Li-rich phases with much higher capacities will depend on the application requirements. A thermally asymmetric feature, nevertheless, should indeed be clearly stated: only the lithiation process necessitates slightly elevated temperatures to form Li-rich phases. As shown, once the Li 2−x Al phase is formed by lithiating an Al film anode at 60 • C, it can be delithiated to form the α phase and deliver similar levels of delithiation capacity at both 60 • C (figure 30(a)) and room temperature ( figure 30(b)).
Importantly, this asymmetric thermal behavior may be ideal for some applications, considering that the lithiation of the anode is equivalent to the charging process of a full cell. For EVs, the heat accumulation during charging is not necessarily a negative thing, as it can warm up the battery pack in winter and facilitate a faster ion diffusion that positively affects the charging speed and reduces lithium plating risks. With an Al anode, the low-grade waste heat generated during the charge due to internal resistance may already be sufficient to facilitate the formation of Li 3 Al 2 and Li 2−x Al phases. Another example could be the solar farm, which captures energy from sunlight. Solar energy can be more effectively utilized if the excess energy during the daytime or sunny days gets stored in energy storage devices like LIBs and is reused whenever necessary, e.g. during nights or cloudy days. The temperature is often higher when the sun is shining intensively during the daytime, which may create more ideal conditions for Al anodes to form Li 3 Al 2 , Li 2−x Al phases, or even Li 9 Al 4 during the charging process. In both cases, although discharge mostly occurs at relatively low temperatures, it is not an issue due to the electrochemical asymmetry presented here.

Conclusion, perspectives, and prospects
To summarize, despite the long history of Al anodes in LIBs, there is still much to learn about the prospects of Li x Al electrodes as the scale of energy storage devices pushes to unprecedented levels. Scientifically, a conclusive plot (figure 31) is shown at the end of this report, to offer a full picture of how Al behaves as an anode material in LIB systems. For lithiation, in addition to the well-studied α-to-β phase transformation (first plateau), the role of the β-LiAl solubility range has been elaborated to give a specific capacity of 1152 mAh g −1 upon the formation of Li-rich β-LiAl. More importantly, Li 3 Al 2 and Li 2−x Al are reported to be formed at temperatures slightly above ambient, doubling the well-adapted specific capacity of ∼993 mAh g −1 . The formation of Li 9 Al 4 , on the other hand, seems to be governed by a different mechanism and exhibits a higher nucleation barrier, partly explaining why it is not approachable at the temperatures that facilitate the formation of Li 3 Al 2 /Li 2−x Al. Moving to delithiation, Li 9 Al 4 -to-Li 2−x Al and Li 2−x Al-to-Li 3 Al 2 phase transformations seem to occur simultaneously, yielding one plateau until the nucleation of Li 3 Al 2 . Similarly, the delithiation of Li 3 Al 2 and desaturation of β ′′ are also ongoing at the same time. Together with the asymmetric thermal behavior reported in this review, lithiation/delithiation kinetics of the higher-order phases are suggested to be of primary interest and warrant further investigations. Regarding the reversibility of these Li-Al phases, cycling between β-LiAl and Li 2−x Al is significantly more stable than cycling the commonly adapted α ↔ β, which induces the formation of porous Al and must be controlled.
From a macroeconomic perspective, the levels of production forecasted for the coming years would imply looming differentiation in the sector, circumventing the need for a one-size fits all vision for future LIBs. That being said, the first mover advantage of Si as the future anode material should not be understated. With billions of dollars and reputations banked on a future of LIBs containing composite, slurry-coated electrodes, the introduction of a disruptive foil-based platform faces a significant activation energy. Alas, at this stage many investment decisions are no longer made by those who are technology-driven, but rather by those with alternative ambitions. Nevertheless, methods like prelithiation are certainly an emerging necessity in the world of Si-rich anodes, but as is also needed for forming and maintaining the β-LiAl phase in Al foils, thus synergy between these elemental platforms certainly exists. In the end, it may be up to yet-another battery startup company to champion Al before the gigafactories of the world reorient themselves.
Further still, as the value cycle of LIBs comes under closer scrutiny in the West, the prospects for cells with low criticality should be expected. However, this cannot be expected to happen in a vacuum, and thus the rising prominence of post-Li technologies, such as Al-ion or Na-ion, is growing immensely in the battery community due to its prospects for good performance and better sustainability (vis-a-vis material and production costs, in theory). We must then ask the question, how can Al be utilized more in LIBs to such that an Al-rich foil anode be coupled with a cathode that is equally simplified, and in what format should this be packaged? Such are the open questions that remain, but certainly, the answers will only be found by thinking outside of the baux.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).
the IAM-MMI of Karlsruhe Institute of Technology (KIT), Germany, for numerous helpful technical discussions.