Si is the most widely studied anode material, and entire reviews have been published on specific types of Si anodes. This is partly due to the extremely large theoretical capacity of elemental Si. The practical limit of Si anode capacity is 3579 mAh g−1 in the pure phase. Along with such impressive theoretical values, Si processing technology is also advanced due to the extensive Si use in a semiconductor (including photovoltaic) industry. Si anode research has thus developed at a rapid speed. The performance advantages and limitations of Si anodes as well as their operational and failure mechanisms have been studied extensively and the knowledge gained may serve as a basis to understand other types of alloying electrodes.
Upon lithiation, both amorphous Si (a-Si) and crystalline Si (c-Si) anodes first become amorphous and may later form a metastable Li15Si4 crystalline phase below ≈0.05 V versus Li. If initially crystalline, Si will expand anistropically, primarily in the <110> direction as seen in Figure 2. This initial lithiation is thought to be limited by the reaction rate at the interface between the crystalline Si and amorphous lithiated phases. However, as the lithiated phase becomes thicker, the large volume change can cause GPa levels of stress to build up, especially at high charge rates. Nanostructures are therefore necessary to relieve the stress at the surfaces and provide necessary void space for expansion.[9f] Without such measures, stress will cause crack formation and limit capacity by inducing a large polarization. Ex situ experiments have shown that cracks become less frequent for c-Si pillars of diameters 240–360 nm diameter,[9c] and in situ experiments showed no cracks for particles below ≈150 nm in spite of the very high (minutes or less) lithiation rates.
Figure 2. Lithiation of Si pillars, showing their anisotropic growth in the <110> directions. Reproduced with permission.[9a] Copyright 2011, American Chemical Society.
Download figure to PowerPoint
In large dimensions or continuous films, a-Si initial material may be preferable over c-Si due to its isotropic lithiation and volume expansion. Experimental evidence agrees for thin-film experiments, and for nanowires, cycle life was improved by lithiating a CVD deposited a-Si layer over the as-grown c-Si nanowires. a-Si is rate limited by Li transport through the alloyed phase, assuming that the surface reaction is not hindered by a thick SEI. SEI growth is facilitated by the formation of solvent-permeable defects and fractures in the SEI layer caused in turn by large-volume changes upon Li insertion and extraction.[10b]
Volume change in a-Si has been measured to proceed almost linearly with degree of lithiation, after about 0.8 Li per Si have been inserted. Based on measurements of curvature in a thin-film anode, Si anodes have been observed to undergo an initially elastic buildup of stress before plastic flow can occur. Independent measurements of the start of plastic flow and the start of linear volume expansion roughly align as shown in Figure 3, supporting this theory.[17b,] The comparison shows volume expansion occurring after the start of plastic strain because volume measurements were done with a thicker film (210 nm versus 50 nm for stress measurement), causing greater stress buildup. Also, a linear volume change agrees with the densities of crystalline Li alloys with Group IV elements. It should be noted, however, that mechanical properties are geometry dependent, and the onset of plastic flow also depends on the charge rate.
Figure 3. Stresses and volume changes within Si thin films upon Li insertion: a) voltage profile and stress for 50 nm Si thin-film electrode, showing the transition from elastic to plastic strain; b) voltage profile and relative volume for 210 nm Si thin-film electrode. The start of volume change is abrupt, indicating the transition from elastic to plastic strain[17, 18]
Download figure to PowerPoint
The chemical diffusion coefficient of lithiated Si is in the range of 10−13–10−11cm2 s−1.[9e,] Such sluggish transport is associated with the breaking and forming of chemical bonds upon electrochemical alloying of Li with Si, and may cause a buildup of high Li concentration near the surface of Si. The high Li concentration, in turn, causes large stress and volume change, resulting in surface fracture similar to c-Si.[12, 20] Nanosized dimensions therefore prevent localized stress by reducing the diffusion length. Geometry and charge rate also plays a role, as discussed earlier. Experimentally, even a 150-nm thick amorphous thin-film Si deposited on a rigid metal substrate and cycled at a C/2 (1800 mA g−1) rate has been observed to crack, although it did maintain contact with the substrate.
In turn, Li mobility also changes with composition and stress. Therefore, Li mobility changes with time and space, depending on geometry and charge rate. The strain induced by high Li composition should aid diffusion, while an overall PITT measurement using a polymer electrolyte to correct for crack formation in thin films suggest a slightly rising chemical diffusion coefficient with increase in Li content.[21b] The measurement of diffusion coefficients is further complicated by the limitations of various electrochemical methods used. Understanding this diffusion process is critical to understanding the operation of Si anodes. However, a detailed study is beyond the scope of this review.
Outside of the Si-active material, Li transport can also be limited by the SEI layer build-up. Structurally stable nanostructured electrodes may fail due to high impedance in a thick SEI layer that plug voids between individual particles, particularly if the electrode thickness and mass loading are reasonably large. Thin electrodes with low mass and capacity loadings, in contrast, may show excellent stability in half cells because Li transport within the SEI-plugged anode may still be sufficiently fast to achieve high capacity utilization.
Although the SEI composition is largely controlled by the components in the electrolyte, the native oxide on Si also contributes a layer of the SEI closest to the active material. The oxide layer is irreversibly lithiated in the first cycle, and a slight decrease in CE is observed. XPS (X-ray photoelectron spectroscopy) analysis shows that SiO2 decomposes to form SiOxFy if left to age in contact with standard LiPF6 and ethylene carbonate (EC)-based electrolyte. However, if a formation step is done immediately, LixSiOy is formed instead, with Li2O at the immediate interface between the active material and the SEI. An SEI layer composed of organic and fluorinated compounds, similar to what is found on graphite, then covers these inner layers. As with standard graphite anodes, commercially available and widely used additives such as lithium bis(oxatlato)borate (LiBOB), fluoroethylene carbonate (FEC), and vinylene carbonate (VC) reduce the SEI layer thickness, improve its stability, and greatly enhance overall performance of high-capacity anodes.
Finally, the bulk of the high-profile research conducted in the field of Si anodes has been in the design of fabrication methods. Various researchers have created nanostructures that are well spaced and sized to allow good transport and prevent sintering and cracking.[9c] Because of the large theoretical capacity of Si, this has been much more successful for Si than for any other material. In 1999, Si nanoparticles were used to assemble a high-capacity anode, although this degraded very quickly. Later studies showed significantly improved stability of electrodes with Si nanoparticles, some with over a thousand stable cycles. High-rate capability has been demonstrated with Si anodes at low mass loadings. For example, a porous nanowire structure even achieved a 18 A g−1 rate at >1000 mAh g−1 for 250 cycles, at low loading of 0.3 mg cm−2. Many other electrodes have also achieved cycle life on the order of 1000+ cycles.[33, 35]
Various carbon-silicon (C-Si) composite particles have been used to enhance Si anode performance. For example, 2D layers of graphite/Si hundreds of nm in thickness have been used to achieve 840 mAh g−1, 1.1 mg cm−2, C/3.4, and 100 cycles, while C-coated Si nanoparticles  and porous Si-C composite spheres achieved similar performance at higher rates. More complex electrodes have achieved high capacities and excellent stability at ultra-high mass and capacity loadings without major compromises in other properties. Such results show that well-designed composite Si anodes with proper dimensions can, in principle, achieve most of the necessary characteristics of a well performing high-capacity anode. The remaining major challenge of Si and other high-capacity anodes is the low CE, which is due to their high specific surface area and insufficient SEI stability. How composite architecture and geometry may overcome this challenging problem will be discussed in a separate section at the end of this review.
Pre-lithiation of anodes before their assembling and/or cycling may reduce the impact of volume expansion, while also reducing first-cycle irreversible capacity loss. The concept has been utilized for other systems where irreversible capacity is an issue. Ma et al. used this method to improve the capacity and cycle stability of Si. However, about 1/3 of the capacity was lost after 20 cycles, possibly due to large particle size (up to 1 μm in diameter) and insufficient electrode fabrication control combined with still significant volume changes within individual particles and the resulting SEI growth. In addition, due to their extremely reductive potential, Li-Si alloys are very reactive with air, particularly if produced as nanoparticles, and need to be handled with extreme care in an Ar-filled glovebox environment. The use of Ar increases the cost of battery fabrication significantly, which creates a major obstacle for the widespread adoption of lithiated anodes by industry.
Binders keep particles in electrical contact, and are critical to assembling high-quality electrodes, regardless of the particle size. Polyvinylidene fluoride (PVDF) is the standard binder used in Li-ion batteries, but has been shown to be inferior to carboxymethylcellulose (CMC), polyacrylic acid (PAA), and alginate, which have become the three binders of choice for Si anodes. All three of these binders (and their derivatives) have higher strength and elastic moduli than PVDF, and have carboxyl groups to form strong chemical bonds to the Si surface.[33, 41] In contrast to PAA, CMC, and alginate, PVDF also swells in electrolyte and experiences a dramatic reduction to its already inferior elastic modulus.[41b] While PVDF withstands greater strain before failure than CMC, it becomes very soft and incapable of holding expanded Si particles in electrical contact upon delithiation and the resulting particle contraction. In addition, permeation of the solvent through PVDF leads to the formation of the SEI between the Si and the binder, which further weakens the binder–Si interface.
Alternatively, the use of electrically conductive binders may improve the electrode's ability to maintain electrically conductive network between individual particles. Such polymer binders have been also shown to achieve high capacity and long cycle life in half cells, but tend to reduce CE in the initial cycles.[35c,d,] Conversely, binder-less anodes have also been fabricated using 1D geometries, and those utilizing carbon nanotubes have achieved perhaps the most promising results. Carbon nanotubes can also be used to replace not only the binder and conductive additive but also the current collector, which comprises a much larger portion of the battery's volume and mass. This results in even greater energy density and is possible due to the high conductivity and strength of carbon nanotubes. The various benefits of such a system are discussed in further detail later in this review.
In spite of the significant progress in the field of Si anodes, no commercial anodes are currently based on Si or Si-containing composites, except in a few very recent designs utilizing a small quantity of small carbon-coated silicon oxide (SiOx)-containing particles added to graphite anodes to slightly increase their gravimetric capacity or pulse-power performance. The authors are optimistic, though, and believe that Si-based anode technology will likely be commercialized within the next 2–4 years. This will require further technology development and overcoming the main challenge of maintaining a constant particle volume and size, and doing so with particles synthesized on a large commercial scale.
Despite its very high cost (prohibitively high in most applications), Ge has recently received increasing interest due to its superior conductivity and Li diffusivity compared with Si. Interestingly, Ge lithiation can result in intermediate lithiated crystalline phases until Li15Ge4 is ultimately formed, while its delithiation results in the formation of a porous amorphous phase. Unlike Si, the entire process is isotropic and does not lead to crack formation, even at a high rate and particle size of ≈620 nm.[50c,] An in situ study by Liang et al. exemplifies this difference by comparing the lithiation process of similarly sized particles (Figure 4). Furthermore, thin-film experiments have shown reversible lithiation at rates as high as 1000 C.[48a] A recent paper analyzing Si–Ge alloys aptly demonstrates Ge's superior rate capabilities (Figure 5), showing the gradually increasing rate capability as the transition is made from Si to Ge.
Figure 4. In situ lithiation of Si and Ge: a–c) Initial lithiation and crack formation of a crystalline Si particle, d–f) Initial lithiation without crack formation of a crystalline Ge particle. Both particles are ≈160 nm in diameter. Reproduced with permission. Copyright 2013, American Chemical Society.
Download figure to PowerPoint
Figure 5. Impact of Ge content on the rate performance of Si–Ge alloy Li-ion battery anode. Reproduced with permission. Copyright 2013, American Chemical Society.
Download figure to PowerPoint
The properties of Ge allow for using particle sizes larger than that of Si. This is a major advantage over Si, as nanoscale particles and the resulting high specific surface area result in the formation of excessive SEI, resulting in the build-up of irreversible capacity and impedance. Indeed, even 500 nm Ge particles have been reported to exhibit stable cycle life, albeit at 150 mA g−1 rate and 800 mAh g−1 capacity. 200 nm sheets of Ge on graphene have been shown to exhibit <20% loss in capacity over 400 cycles, at nearly theoretical capacity although with only 45 wt% Ge. However, to date, most studies have focused on nano-Ge with dimensions often <100 nm, and even as small as 3 nm. This suggests that the size of Ge particles has yet to be fully optimized. Nevertheless, truly impressive rate capability has been achieved, and even 40 C rate capability has been reported after 400 cycles at 1000 mAh g−1 and 15 mg cm−2 mass loading with a LiFePO4 cathode as a counter electrode.[50b]
Sn anode research is different from Si and Ge anodes in that much focus has been placed on oxides, which showed significantly improved cycle stability. In the fully lithiated state, Sn forms Li22Sn5, which places its theoretical limit at 990 mAh gSn−1. While Sn specific capacity is significantly inferior to Si, its volumetric capacity is comparable (Figure 1). However, the theoretical capacity of pure Sn has not been achieved with stable cycling, possibly due to the brittleness of the fully lithiated phase. A recent ex situ study showed that 10 nm particles decrease in size and increase in size distribution after lithiation and delithiation, indicating that fractures occur even at such small particle sizes. Indeed, embedded Sn anodes have shown slightly higher stable capacities, and tin oxide has been suggested to be superior very early on due to the oxide phase “gluing” or holding the Sn grains together.
One would assume that the change in properties would be continuous from Si to Ge to Sn. However, the Sn lithiation process follows the formation of crystallographic phases (also see Figure 6):
Figure 6. Formation of crystalline phases upon electrochemical Li insertion to and extraction from a Sn anode. Reproduced with permission.[61a] Copyright 1998, American Physical Society.
Download figure to PowerPoint
Sn is different from Ge and Si in that it has an extremely low melting point, so much that room temperature is already 60% of the melting point. Sn atoms are therefore much more mobile than in Ge and Si, and crystallize more easily. Crystallization helps to explain the easy fracturing of lithiated Sn, as well as the 5–10 orders of magnitude difference in the measured chemical diffusion coefficient of Li between electrochemically and chemically lithiated Li-Sn alloys. In the former case, the brittle Li17Sn4 phase easily forms on exterior regions of a particle. In the latter case, chemically lithiated Sn most likely has large grain size and a uniform degree of lithiation. Taking such properties into consideration, Sn anodes appear to require composite structures to help maintain electrical contact and Li diffusivity.
Tin oxide composites received heightened attention in the late 1990s due to its commercial development by Fujifilm and a high-profile publication by Idota et al. Based on their achieved capacities, many researchers have traditionally viewed tin oxide composites to form an electrochemically inert irreversible Li2O, which hold together Sn phase regions.[60, 65] Such claims were bolstered by the observation that Sn-phase regions could be observed to lithiate/delithiate by in situ XRD.[60, 66] However, more recent advanced nanostructures have improved uniform electron transport and surpassed the theoretical capacity of SnO2 by reversing the Li2O formation reaction, although at a higher Li extraction potentials. This reaction is believed to lithiate at ≈0.8 V and delithiate at ≈1.3 V,[65a,67b,67e] with extremely broad peaks. Unfortunately, as the overall cell voltage is reduced, this extra capacity is generally of lower value. Tin oxide anodes are also believed to degrade due to the coarsening of Sn particles,[65a,] and oxidizing the Li2O phase may accelerate this process. Even so, Zhou et al.[67g] have reported high capacity of 1000–1400 mAh g−1, good cycle life of 500 cycles (at a high mass loading of ≈10 mg cm−2), and good rate capability, although at the expense of low CE due to the use of high-surface area graphene oxide.[2a,]
Lastly, a review of Sn anodes would not be complete without visiting the Sn-Co-C anodes used in the Nexelion battery initially commercialized and advertised by Sony. Although the battery had a relatively short life, it boasted a 20% improvement in overall volumetric capacity. Lithiated Sn–Co alloys may at first form a Li-Sn-Co phase, but eventually separates into two phases to form a new Li-Sn phase. Then, if sufficient Co is present (20% is sufficient), an amorphous Co phase can be formed as well. When the anode is delithiated, the Sn then re-alloys with the Co instead of forming its own Sn phase. This process is believed to be the cause of the superior cycle life of Sn–Co alloys over pure Sn, perhaps preventing fracture. Other Sn-transition metal alloys also appear to have similar properties.