Understanding the Degradation of a Model Si Anode in a Li-Ion Battery at the Atomic Scale.

To advance the understanding of the degradation of the liquid electrolyte and Si electrode, and their interface, we exploit the latest developments in cryo-atom probe tomography. We evidence Si anode corrosion from the decomposition of the Li salt before charge-discharge cycles even begin. Volume shrinkage during delithiation leads to the development of nanograins from recrystallization in regions left amorphous by the lithiation. The newly created grain boundaries facilitate pulverization of nanoscale Si fragments, and one is found floating in the electrolyte. P is segregated to these grain boundaries, which confirms the decomposition of the electrolyte. As structural defects are bound to assist the nucleation of Li-rich phases in subsequent lithiations and accelerate the electrolyte's decomposition, these insights into the developed nanoscale microstructure interacting with the electrolyte contribute to understanding the self-catalyzed/accelerated degradation Si anodes and can inform new battery designs unaffected by these life-limiting factors.

To meet the rapidly increasing demand for Li-ion batteries for electric vehicles,[1,2] tremendous efforts have been devoted to discovering cheap and abundant anode materials that can replace graphite that is in short supply.[3] A crystalline Si anode, which can offer nearly 10 times the capacity of a commercial graphite anode (QSi = 4200 mAh g–1 vs Qgraphite = 372 mAh g–1)[4] has emerged as an attractive anode material for next-generation Li-ion batteries since the first development of the Li–Si anode by Lai in 1976.[5] Compared to graphite, in which each of the six in-plane C atoms can only bond with one Li ion, each Si atom can bond with up to 4.4 Li ions.[6] Thus, finding a path to exploiting Si as an anode material can be a revolutionary approach for reaching batteries with ultrahigh energy density. Tesla, Inc., has revealed its plans to gradually increase the use of Si anodes in its future batteries,[7] and Amprius Technologies, Inc., recently announced the shipment of its first commercially available Si anode based Li-ion batteries with an energy density of 450 mWh g–1.[8]

An efficient Si anode remains some sort of holy grail for rechargeable Li-ion batteries, and their widespread use is hindered by rapid capacity fading.[9,10] The enormous volume changes occurring during lithiation/delithiation cycles (e.g., +300% volume increase from Si to Li22Si5) result in irreversible damage:[4] deformation and residual stresses accumulate and create an ensemble of structural defect features, including interfaces, dislocations, grain boundaries, and nanocracks forming within the Si anode. An array of approaches has been explored to overcome this critical issue. For example, the use of nanocomposite/structured Si,[11] including nanowires,[12,13] core–shell[14,15] and hollow[16,17] nanoparticles, and porous Si,[18,19] has been reported to be effective for the enhanced suppression of the initiation of mechanical fracture from the large volume changes.

Most studies use techniques providing a bulk average or two-dimensional information,[20−22] which, even in combination, cannot fully analyze the nanoscale compositional distribution and microstructural evolution of electrodes and electrolytes. In situ[11,23,24] and cryogenic transmission electron microscopy (TEM)[25,26] have already revealed an undesirable removal or destruction of the passivating solid-electrolyte interphase (SEI) and severe pulverization of Si nano/microparticles from bulk Si during the expansion and shrinkage cycles.[26] Despite impressive empirical advances, numerous fundamental aspects of the microstructural degradation remain elusive, making it impossible to devise targeted strategies to circumvent these specific issues and enable a breakthrough in Si-based anodes. Elucidating the mechanisms that led to mechanical failure has emerged as a crucial topic to achieve a high-capacity Si anode.

The combination of high-resolution microscopy of the fine scale of the microstructure that develops and the precise microanalysis of the electrode’s evolving composition can be achieved by using the latest development in cryogenic atom probe tomography (cryo-APT), Figure a. APT provides direct and three-dimensional, near-atomically resolved analytical imaging of materials and has the ability to collect all elements irrespective of their mass. APT is underpinned by an intense electric field that provides controlled removal of individual ions from a sharp, needle-shaped specimen. However, this field can cause outward electromigration of Li[27] in battery materials, making the detailed analysis of its distribution often impossible but also affecting the overall data quality,[28] which explains why battery materials have rarely been analyzed by APT.[29−32] However, we demonstrated recent approaches enabling analysis of lithiated anode and cathode materials,[28] and delithiated samples still bear traces of crucial processes taking place during battery operation. Compared with cryo-TEM, the outstanding advantage of cryo-APT is the ability to resolve sub-nm-scale structure and chemistry at the same time and more importantly in three dimensions. Cryo-TEM usually provides a 2D projection of a sometimes complex 3D nanostructure. Tomography can be challenging to interpret from the reconstruction of images acquired through a tilt-series. Chemical information within the sample usually requires the additional incorporation of EELS or EDX with cryo-TEM, which is not always available, and with a sensitivity that is in the range of 1 or more at. % and quantification is particuarly challenging for light elements (e.g., Li). Cryo-APT is however not without challenges, from the difficulties in specimen preparation and transfer to, e.g., the mass ranging with numerous molecular ions and fragmentation paths that can affect their quantification, and this is particularly the case for organic compounds for which the literature is limited.

Figure 1

(a) Unique infrastructure for the cryo-atom probe enabling the study. (b) SEM images of the LN2-quenched anode containing the frozen-electrolyte surface. (c) The Si electrode where (d) the cryo-milled pillar was made to prepare (e) the final APT specimen. The scale bars are 50 μm in parts b and c, 20 μm in part d, and 1 μm in part e.

Here, we leverage cryo-APT for the first time to obtain compositional mapping of Li-ion battery materials, the abutting electrolyte, and the solid–liquid interface between the two at an increasing number of charge–discharge cycles. Custom cells were disassembled inside a N2 glovebox (H2O and O2 < 10 ppm);[33] see the Methods section in the Supporting Information. The collected Si anode with the electrolyte was immediately plunge-frozen in liquid N2 (LN2) and then transferred by using the cryogenically cooled, ultrahigh-vacuum suitcases into a scanning-electron microscope/Xe-plasma focused ion beam (SEM/PFIB) for imaging and cryogenic specimen preparation, Figure b,c. APT specimens of the electrolyte and electrode were prepared at cryogenic temperature using the method we introduced in ref (34), Figure d,e.

The locations of the cryo-APT analyses of the uncycled electrode and electrolyte, Figure c, are indicatively marked in Figure b. Within the electrolyte, individual, isolated Si ions are already detected. We conducted a cryo-APT analysis of the frozen raw electrolyte on a different metallic substrate (NP-Au) that shows no Si ion (see Figures S1–S4). These additional analyses confirm that dissolved Si ions originated from the corrosion of the Si anode. Veith et al. observed non-electrochemically driven Si–O and Si–F bonds on a Si anode soaked in a similar electrolyte.[35] Si–O groups can react with HF generated by hydrolyzed or thermally decomposed LiPF6 electrolyte,[36] resulting in the dissolution of Si ions and two additional H2O molecules, which trigger further HF generation and a self-sustaining corrosive cycle.[37−39] The hydrolysis could be initiated by residual atmospheric moisture during cell assembly.[40] Degradation of the anode and the electrolyte hence starts even before cycling, with any oxygen-containing Si species that generate more water and accelerate the failure of the Si battery cell.

Figure 2

(a) Schematic of the Si electrode and cycling process. (b) Voltage vs current curves of the Si(111) anode in a Li–Si cell. (c) Cryo-APT analysis of the electrolyte and anode before cycling; scale bars are 20 nm. (d) Cryo-APT reconstructed atom map of the one-cycle electrolyte; the blue isosurface delineates regions containing at least 25 at. % Si; the scale bar is 20 nm. Movie #1, the corresponding mass spectra, and additional analyses can be found in the Supporting Information: (i) a close-up showing dissolved Si ions (scale bar = 2 nm) and (ii) a delaminated Si debris in the electrolyte (scale bar = 5 nm). Green, blue, and yellow dots represent reconstructed carbonate species, Si, and SiO compounds, respectively. (e) Transmission-electron micrograph of the 10-cycled Si anode along the [110] zone axis of the single crystal, along with fast Fourier transformation (FFT) patterns from different regions highlighted by colored boxes. The white scale bar is 20 nm. (f) 3D reconstructed atom map of the Si electrode after 25 cycles (scale bar = 20 nm). Blue, yellow, and pink dots represent reconstructed Si, SiO, and Li, respectively. Movie #2 and mass spectra of the corresponding data set are presented in the Supporting Information. The inset shows the extracted Si grain boundary with the 2D contour density map of P atoms (scale bar = 5 nm).

After one cycle, Figure d, cryo-APT reveals also dissolved isolated Si ions, accompanied by an ∼10 nm pulverized Si fragment covered with an oxide shell (see Figures S5 and S6). Such a fragment could potentially block the pores of the separator for Li-ion diffusion, raising cell impedance and deteriorating the rate performance of the battery. The presence of the SiO species at the surface supports the hypothesis that the dissolution was associated with the formation of HF.

After 10 cycles, TEM was performed on the dried electrode after removal of the electrolyte and thorough cleaning, Figure e, complemented by additional APT experiments (Figures S7–S13). We evidence that the originally single crystalline Si has transformed into a nanocrystalline microstructure, containing numerous nanoscale grains and grain boundaries with different crystallographic orientations that have been formed during the lithiation/delithiation process, confirming previous reports.[41] The volume change associated with the formation of Li-rich phases imposes strong compressive loading on the silicon matrix.[42] Indentation of Si single crystals has demonstrated that the breaking of covalent Si–Si bonds injects a high number of vacancies in the crystal and results in amorphization,[43,44] also reported experimentally during battery cycling.[45] Upon relaxation during delithiation, depending on the rate, new crystals nucleate with no orientation relationship with the surrounding crystal matrix.[43,44] The discharging rate must influence this process.

After 25 cycles, Figure f, cryo-APT analysis of the very surface of the anode contains two Si grains, as confirmed by atom probe crystallography[46] (Figure S14), and a faceted grain boundary. No chemicals expected from the SEI layer (LiCO2, LiOH, LiF) are observed at the interface (see Movie #3), which can be attributed to the high reversibility of the SEI layer on Si anodes.[25,47,48] On the surface, we find several isolated nanoscale islands rich in SiO species. Such oxides promote the formation of HF which corrodes the Si, passivate the Si anode, and act as a mechanical clamping layer that restricts swelling.[48,49] SiO can store Li ions (QSiO = 1543 mAh g–1)[50] with lower volume expansion (approximately 120%) when irreversibly lithiated,[51] that can cause stress build-up at the interface and facilitate crack initiation,[52] decohesion, and pulverization, explaining the presence of SiO on the Si fragment’s surface in Figure d.

After full delithiation, 20–30 nm below the surface, Li (8 ± 1 appm) is still detected within the Si matrix, as readily visible from the corresponding mass spectrum, Figure f. Density-functional theory predicts an attraction between vacancies and Li in Si,[53] which can combine with a strong Coulomb attraction between an electron-rich vacancy and the electropositive Li. The image Li atoms are hence likely trapped by remaining vacancies injected under plastic loading.

At the grain boundary, Li does not appear segregated, conversely to P, that has even seen partitions to specific facets and to the facet junction indicated by a black arrow (see Figure S15). Such a distribution was previously suggested to be associated with local strain.[54] P can diffuse along grain boundaries in Si,[55] and its segregation can be energetically favorable due to the passivation of dangling bonds,[56] which modifies the conductivity.[57] In addition, atomistic simulations have indicated that the combined effect of the presence of P and a stress concentrator (i.e., a grain boundary) decreases the fracture strength of Si nanowires.[58] Lastly, the presence of P (209 ± 11 appm), originally from the LiPF6 salt, also suggests the liberation of F and the facile formation of HF that is a known embrittler of polycrystalline Si through void formation along grain boundaries.[59] These effects collectively make these newly created grain boundaries particularly brittle and critical to the lifetime of the Si anode.

To summarize, cryo-APT allowed us to track the evolution of the three-dimensional, nanoscale elemental distributions of species in the electrolyte, a model Si anode, and their interface over increasing charge–discharge cycles. We provide measured data that advance the understanding of the degradation mechanism—or actually degradation mechanisms—and emphasize the often-overlooked role of microstructural defects created and evolving throughout the battery operation lifetime. In addition, the nucleation of the LiSi (metastable) phases[42] during the first cycle can be assumed to be homogeneous, occurring randomly across the surface of the anode. However, the combined presence of crystalline defects and remaining Li impurities in the anode will undoubtedly assist heterogeneous nucleation of these phases during subsequent lithiation, potentially enhanced by accelerated diffusion of Li through structural defects.[60,61] Segregants can energetically destabilize grain boundaries, already weakened by HF,[59] or form space charges, that can favor decohesion. Nucleation in the parts of the microstructure with a high density of defects localizes the volume expansion to mechanically weaker regions, thus facilitating the pulverization of fragments from the anode. This combination of (electro)chemical reactions, phase transformation, and mechanical failure, assisted by the localized decomposition of the electrolyte, accelerates the delamination/mass loss and localized lithiation causing fast loss of capacity[51] (see Figure S16). Strategies for the development of robust and durable Si-based anodes for next-generation Li-ion batteries can draw from our findings on the degradation of the Si electrode—the role of the newly formed grain boundaries that may be exploited through segregation, but also the details of the electrolyte degradation that can guide the selection of P-free and F-free salts and avoiding exposure to moisture during fabrication, which can be difficult to achieve in large-scale production.