Lö Mfuti wa Graphene Ungapangitsa Mabatire a EV Kukhala Motalikirapo ndipo Sungani Mphamvu Zambiri

Authors: In Hyuk Son Jong Hwan Park Soonchul Kwon Seongyong Park Mark H. Rümmeli Alicja Bachmatiuk Hyun Jae Song Junhwan Ku Jang Wook Choi Jae-man Choi Seok-Gwang Doo Hyuk Chang Abstract Silicon is receiving discernable attention as an active material for next generation lithium-ion battery anodes because of its unparalleled gravimetric capacity. However, the large volume change of silicon over charge–discharge cycles weakens its competitiveness in the volumetric energy density and cycle life. Here we report direct graphene growth over silicon nanoparticles without silicon carbide formation. The graphene layers anchored onto the silicon surface accommodate the volume expansion of silicon via a sliding process between adjacent graphene layers. When paired with a commercial lithium cobalt oxide cathode, the silicon carbide-free graphene coating allows the full cell to reach volumetric energy densities of 972 and 700 Wh l−1 at first and 200th cycle, respectively, 1.8 and 1.5 times higher than those of current commercial lithium-ion batteries. This observation suggests that two-dimensional layered structure of graphene and its silicon carbide-free integration with silicon can serve as a prototype in advancing silicon anodes to commercially viable technology. Introduction The theoretical gravimetric capacity of silicon (Si) reaches almost 4,000 mAh g−1. This unparalleled value has stimulated the battery community to invest considerable research efforts because the high gravimetric capacity enables one to increase the energy densities of lithium-ion batteries (LIBs) significantly, and thus bring future LIB applications, such as electrical vehicles, to a reality , , , , . In the past decade, diverse advanced electrode structures , , , , , , , , , , and binder designs , , were developed to resolve chronic capacity fading issues originating from the large volume change of Si, leading to substantially improved cycling performance even over thousands of cycles . In spite of the promising gravimetric value and substantial progress in cycle life, most of Si anodes demonstrated to date have focused primarily on the gravimetric capacity but have not offered a similar promise in their volumetric capacity because existing electrode designs rely on pre-defined void space to accommodate the volume expansion of Si. In many LIB applications including portable electronics, however, the volumetric energy density is a critical parameter in determining battery performance. Together with a relatively inferior cycle life, weak volumetric energy density is presently a major bottleneck in implementing Si anodes in commercial cells. To meet this critical demand, Si anode technology needs to be revisited with different electrode designs that offer stable cycling performance while the electrode volume is minimized. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 11 Although a variety of Si morphologies and their composites with other conductive materials are currently available, for an immediate solution to actual manufacturing that requires high-standard quality control, an easy and scalable synthesis of active Si components is crucial. From a practical application standpoint, among a vast number of viable candidates, the use of commercial Si nanoparticles (NPs) with a simple and efficient conductive surface coating would be highly desirable. Various conductive materials including amorphous carbon (AC) have been investigated as coating materials for Si NP anodes , , . However, most of them fail to deliver stable long-term cycling performance because the implemented coating materials are unable to accommodate the volume expansion of Si and consequently fracture over repeated cycles. 8 20 21 In an attempt to address the limitation of previous conductive coatings as well as to achieve good cycling performance with a significantly higher volumetric energy density, in this study, we adopt multilayer graphene directly grown on the Si surface as a coating material. The two-dimensional (2D) layered character of graphene provides a unique and efficient operation of Si anodes since multilayered graphene can accommodate Si volume expansion via a sliding process between adjacent layers without the need to provide void space in the as-made electrode. Also, the graphene-coated Si NP pellet exhibits 12.8 S cm−1 at a marginal graphene content of 1 wt% through a high percolation network. To this end, we overcome a challenge of growing graphene directly on Si surfaces without Si carbide (SiC) formation by developing a chemical vapour deposition (CVD) process that involves a mild oxidant. SiC formation is fatal in Si anode operations because SiC is an electrical insulator with poor defect characteristics. Moreover, SiC is inactive in reacting with Li ions and consequently hinders Li ion diffusion into the Si phase. With the assistance of the graphene interlayer sliding process and enhanced conductivity, the graphene-coated Si NPs reach a volumetric capacity of 2,500 mAh cm−3 (versus 550 mAh cm−3 of commercial graphite), the highest value among those reported to date for any LIB anodes while exhibiting excellent cycling and rate performance. a priori Results SiC-free graphene growth on Si The direct growth of high-quality graphene on Si via CVD process has proven challenging because typical graphene synthesis conditions require a reducing atmosphere that tends to strip the native Si oxide layer off the Si surface, and then drives a reaction between Si and decomposed carbon precursors to form SiC . Our initial approach of using methane (CH4) as a carbon precursor mixed with H2 was indeed unable to achieve a graphene growth on the Si NPs ( ), thereby yielding only β-SiC ( ). To overcome this limitation, we included carbon dioxide (CO2), a mild oxidant, in the CVD process along with CH4 (refs , ). The inclusion of CO2 allows one to avoid the formation of SiC and also lower the growth temperature compared with well-known graphene growth on surface . In our experiment, at 900 °C, the graphene growth was incomplete or inhomogeneous ( ), whereas graphene growth at 1,100 °C produced oxide layers on the Si surface that are too thick for efficient Li ion diffusion ( ). At an intermediate temperature of 1,000 °C, 2–10 layers of graphene were formed as shown in transmission electron microscope (TEM) images ( ) clearly displaying the layered structure as well as the interlayer distance near 3.4 Å ( , inset). Closer inspection indicates that individual layers are anchored directly to the Si particle surface at their ends (red arrows in ) and lie parallel to the Si surface. These well-aligned graphene layers can maintain their layered stacking structure even during lithiation via a sliding process ( ), thus providing an elegant means to accommodate the volume expansion of Si. Graphene growth has been attempted on Si dioxide (SiO2) surfaces using CH4 and H2 (refs , ). However, the graphene growth in those processes was very inefficient or does not avoid SiC formation. The inefficient growth is ascribed to insufficient catalytic sites on SiO2 surface. By contrast, in our process, the CO2 addition generates more catalytic sites in the form of SiO with some defects. 22 23 Supplementary Fig. 1 Supplementary Fig. 2 24 25 23 Supplementary Fig. 3 Supplementary Fig. 4 Fig. 1a,b Fig. 1b Fig. 1c Fig. 1d 26 27 x ( ) A low-magnification TEM image of Gr–Si NP. ( ) A higher-magnification TEM image for the same Gr–Si NP from the white box in . (Insets) The line profiles from the two red boxes indicate that the interlayer spacing between graphene layers is ∼3.4 Å, in good agreement with that of typical graphene layers based on van der Waals interaction. ( ) A high-magnification TEM image visualizing the origins (red arrows) from which individual graphene layers grow. ( ) A schematic illustration showing the sliding process of the graphene coating layers that can buffer the volume expansion of Si. a b a c d The SiC-free graphene growth was verified by both bulk scale and single-particle scale analyses. X-ray photoelectron spectroscopy profiles in Si 2 band ( ) as well as X-ray diffraction (XRD) spectra ( ) show no peaks corresponding to SiC for graphene-coated Si (Gr–Si), AC-coated Si (AC–Si) and pristine Si, in contrast to a control (SiC–Si) sample synthesized through a CO2-free route using CH4 and H2. On the other hand, a scanning TEM imaging using high-angle annular dark field showed that the Si NP have a core–shell structure with brighter core and thin relatively darker shell attributed to a Si core and an oxide coating ( ). Electron energy loss spectroscopy (EELS) spectra ( ) obtained for multiple spots across the NP did not exhibit any signals reflective of SiC formation, confirming SiC-free growth in the current growth process. Also, SiO2 signals at 108 eV were detected at points 1 and 2 more strongly than the other spots in the centre, implying that the SiO2 surface serves as catalytic sites for graphene growth. The persistent presence of the SiO2 surface layers was also verified by XRD ( ) and EELS ( ) characterization. p Fig. 2a Supplementary Fig. 2 Fig. 2b Fig. 2c 28 Supplementary Fig. 2 Fig. 2b,c ( ) X-ray photoelectron spectroscopy spectra in Si 2 band for Gr–Si, SiC–Si, AC–Si and pristine Si. ( ) A scanning TEM image of a Gr–Si NP. ( ) EELS spectra for the spots denoted in . a p b c b TEM analysis In situ We now turn to the use of Gr–Si as an anode material in LIBs. The lithiation of the Gr–Si NPs (1 wt% graphene) was monitored in real time during their volume expansion using TEM analysis , , , . In the actual experiment, some Gr–Si particles were placed onto a gold (Au)-fixed electrode. A second, but movable, electrode with Li/LiO2 at its tip was positioned to just come into contact with the Gr–Si NPs to lithiate them ( ). Once contact with the second electrode was made, the particles began to swell ( ). Greater details on this process can be seen in and . Two types of expanding structures were observed, namely non-defective particles and defective particles as highlighted in . In cases where the graphene fully encapsulates the particle and has no obvious defects ( ), the diameter increased by ∼30% (220% volume expansion). At the end of lithiation, the interlayer distance of the graphene layers increased to 3.8 Å reflective of Li intercalation , and the layered characteristic of the graphitic coating was preserved all round the NP ( ), suggesting a sliding process between layers as illustrated in . In cases where a defective region exists (see green circle in ), upon swelling, the inner particle pulverizes and ruptures through the defect ( ). The observed distinct fracture behaviours are unlikely to be from particle size, as an encapsulated particle bigger than the fractured one in did not fracture (see the indicated particles at 0 and 40 s in ). Also, recent observation indicates that if not encapsulated, even smaller Si particles ( 200 nm (ref. ). in situ 29 30 31 32 Fig. 3a Fig. 3b Supplementary Fig. 5 Supplementary Movies 1–3 Fig. 3b,c Fig. 3d 33 Fig. 3d Fig. 3c Fig. 3a Fig. 3e Fig. 3 Supplementary Fig. 5 34 35 ( ) Gr–Si NPs attached to the surface of Au wire and a second Li/LiO2 electrode. ( ) The same Gr–Si NPs after lithiation. ( ) A schematic summary of lithiated Gr–Si NPs for both non-defective and defective graphene encapsulation. ( , ) Close up TEM images for ( ) non-defective particle (the one circled with the red line in and ) and ( ) defective particle (the one circled with the blue line in and ). The EELS spectra in both cases confirm the lithiation. The line profiles from the red boxes in both cases show increased interlayer distances of ca. 3.8 Å, reflective of lithiation into the graphene interlayer space. a b c d e d a b e a b Local EELS analysis during lithiation verifies the lithiation of Si for both cases through a blurred Li -edge near 55 eV as well as Si -edge at 100 eV ( ; ref. ). The layer shown on the surface of graphene in is attributed to Li/Li2O residue. In essence, the graphene layers preserve the integration with Si through their chemically bonded roots on the Si surface, and at the same time, the sliding process allows the few layer graphene coating to accommodate the volume expansion of Si. In a separate TEM experiment with the Gr–Si LIB electrode, it was observed that the average number of graphene layers measured at different locations decreased from 9.4 to 5.6 layers ( ) after the first lithiation, reconfirming the interlayer sliding process. K L Fig. 3d,e 36 Fig. 3d ex situ Supplementary Fig. 6 The graphene encapsulation and interlayer sliding process exert a clamping force that helps to maintain Si particle integrity during volume expansion. However, this clamping mechanism does not appear to affect Li diffusion, as good specific capacities are achieved in the electrochemical measurements. It is also notable that the volume expansion of particles was, on the whole, uniform in all radial directions even though the contact point with the second electrode is highly localized. This highlights the graphene layers’ propensity for homogeneous Li ion diffusion into the Si core (see red and orange dotted lines in ). The analysis also indicates that Li diffusion through the graphene layers is pretty fast, in agreement with previous studies , . Also, it is noted that the observed clamping effect is mainly from the graphene coating layers rather than surface native oxide layers because the surface oxide layers are too thin to have a significant effect. The clamping effect based on surface oxide layers has been observed only when Si has hollow structures , or the oxide layers are very thick . Fig. 3a,b in situ 30 37 7 11 38 Conductivity and electrochemical measurements Upon graphene growth, the conductivity of the active material assembly becomes markedly enhanced. When pelletized in powder, pristine Si exhibited a low conductivity of <10−7 S cm−1, whereas even 1 wt% of graphene addition increased the conductivity remarkably to 12.8 S cm−1 ( ). This enhancement is ascribed to an excellent percolation behaviour of the graphene coating layers in a way that the 2D graphene and its well-aligned layered stacking promote a well-connected conductive network ( ), and is also in line with previous graphene–polymer composites . The conductivity increased further to 38.3 S cm−1 for a graphene content of 5 wt%. Fig. 4a Supplementary Fig. 7 39 ( ) The pellet conductivities of Gr–Si, AC–Si and Super P–Si at different carbon contents. ( ) The first lithiation–delithiation profiles. The electrode thicknesses of 5 wt%-Gr–Si, 1 wt%-Gr–Si, 2 wt%-AC–Si and pristine Si (Super P 1 wt%) are 4.5, 5.0, 8.3 and 12.3 μm, respectively. ( ) The lithiation (filled circle) and delithiation (open circle) capacity retentions of the same four samples. The potential was swept within a voltage window between 0.01 and 1.5 V at 0.5C. ( ) The top-viewed SEM images: the 5 wt%-Gr–Si electrode (left) before and (middle) after 200 cycles and (right) the 2 wt%-AC–Si electrode after 200 cycles. ( ) A high-magnification TEM image of 5 wt%-Gr–Si after 200 cycles obtained from the white box in the low-magnification TEM image (top inset). Middle inset: selected area electron diffraction showing amorphous nature of Si after cycling. Bottom inset: a line profile from the graphene layers in the red box displaying an interlayer distance of 3.8 Å. ( ) Rate capability of 5 wt%-Gr–Si with different initial areal capacities (3.0, 2.0 and 1.0 mAh cm−2). a b c d e f The electrochemical performance of the Gr–Si NPs was examined by preparing coin-type half-cells in which Li metal serves as the counter/reference electrode. For comparison, in addition to Gr–Si with different graphene contents, AC–Si and pristine Si were also investigated. The graphene content was controlled by the growth time while the growth temperature was fixed to 1,000 °C. All samples exhibited characteristic lithiation and delithiation plateaus at 0.1 and 0.4 V versus Li/Li+, respectively, in the first cycle ( ; 0.05C, 1C=2,000 mA g−1, 0.01–1.5 V versus Li/Li+). 5 wt%-Gr–Si, 1 wt%-Gr–Si, 2 wt%-AC–Si and pristine Si exhibited lithiation/delithiation capacities of 2,820/2,340, 2,341/1,629, 3,143/1,160 and 3,632/741 mAh g−1, respectively, when Si and coated graphene are taken into account, leading to initial Coulombic efficiencies (CEs) of 83%, 69%, 37% and 20%. The low reversible capacity and initial CEs of AC–Si are attributed to the defects of the AC layers that trap Li ions. Fig. 4b The graphene coating makes a drastic difference in the cycling performance ( and ). For a fair comparison among the samples, we first set the areal capacity to 1 mAh cm−2. Under this constraint, the capacity fading of both 2 wt%-AC–Si and pristine Si was significant even for early cycles, indicating that a simple AC coating is not operational at all. Between both Gr–Si samples, 5 wt%-Gr–Si exhibited better performance than that of 1 wt%-Gr–Si (85% versus 70% retention after 200 cycles), suggesting that depending on the areal capacity, the graphene content becomes critical. The lower retention of 1 wt%-Gr–Si might be explained by occasional defect formation as observed in our TEM investigation based on the same graphene content ( ). For the 5 wt% graphene sample, the testing was expanded to even higher areal capacities up to 6 mAh cm−2 ( ). For areal capacities of 1, 3 and 6 mAh cm−2, the capacity retentions after 100 cycles were 96%, 91% and 43%, respectively. These samples also delivered good CEs of 99.6, 99.3, and 99.4% after 100 cycles, respectively. The effect of the conformal graphene coating on the structural integrity of Si NPs was elucidated by post-mortem scanning electron microscope (SEM) and TEM images after 200 cycles. Even after 200 cycles, the SEM image of 5 wt%-Gr–Si visualized individual NPs with clear particle boundaries ( ). By contrast, the SEM image of 2 wt%-AC–Si showed that solid electrolyte interphase (SEI) layers overgrew and buried individual NPs, indicating the importance of the graphene coating on the stability of SEI layers. The TEM analyses ( and ) of 5 wt%-Gr–Si visualized the preserved individual layers of the graphene coating, confirming their robust integration with Si NPs during the repeated large volume change again through the sliding process. In addition, the film thickness of 5 wt%-Gr–Si with 3 mAh cm−2 increased by 51.3% (15.0→22.7 μm) after the first cycle, which is substantially smaller than that (77.4%, 10.2→18.1 μm) of 5 wt%-Super P-pristine Si even with half the areal specific capacity of 1.5 mAh cm−2. Fig. 4c Supplementary Fig. 8 in situ Fig. 3e Supplementary Fig. 8b Fig. 4d Fig. 4e Supplementary Figs 9–11 According to TEM images after the first delithiation ( ) and the 200th delithiation ( ), the interlayer distance of the graphene layers remained at ∼3.8 Å, indicating that the sliding process is not fully reversible. These phenomena are attributed to a combined effect from SEI layer formation and residual Li ions between the graphene layers, together with the smaller volume shrinkage during the first delithiation than the volume expansion of the first lithiation , . Nonetheless, the graphene layers remain rooted on the Si surfaces, continuously fulfilling their original roles as a conducting agent and a buffering medium to accommodate volume expansion of Si. Supplementary Fig. 10 Supplementary Fig. 11 40 41 The graphene coating also has a marked effect on the rate capability utilizing its well-developed percolating network. Even at a high areal capacity at 3 mAh cm−2, a 20 times C-rate increase from 0.5 to 10 C still retained 90% of the initial capacity. It should be emphasized that the direct graphene growth on Si NPs used here is fundamentally different from previous works in which a simple mixture of Si NPs and reduced graphene oxide is prepared via a solution process in that in the case of the solution processes, complete encapsulation of Si NPs is not feasible , , , 42 43 44