Journal of Materials Science

, Volume 48, Issue 14, pp 4823–4833

A silicon nanoparticle/reduced graphene oxide composite anode with excellent nanoparticle dispersion to improve lithium ion battery performance

Authors

  • Rhet C. de Guzman
    • Department of Chemical Engineering and Materials ScienceWayne State University
  • Jinho Yang
    • Department of Electrical and Computer EngineeringWayne State University
  • Mark Ming-Cheng Cheng
    • Department of Electrical and Computer EngineeringWayne State University
  • Steven O. Salley
    • Department of Chemical Engineering and Materials ScienceWayne State University
    • Department of Chemical Engineering and Materials ScienceWayne State University
Article

DOI: 10.1007/s10853-012-7094-7

Cite this article as:
de Guzman, R.C., Yang, J., Cheng, M.M. et al. J Mater Sci (2013) 48: 4823. doi:10.1007/s10853-012-7094-7

Abstract

Composite anodes of Si nanoparticles (SiNPs) and reduced graphene oxide (RGO) sheets with highly dispersed SiNPs were synthesized to investigate the performance-related improvements that particle dispersion can impart. Three composites with varying degrees of particle dispersions were prepared using different ultrasonication, and a combination of ultrasonication and surfactant. With more dispersed SiNPs, the capacity retention and rate performance as evaluated by galvanostatic cycling using increasing current density rates (500–2500 mA/g) also improved compared with anodes that have poor particle dispersion. These results demonstrate that better nanoparticle dispersion (small clusters to mono-dispersed particles) between the stable and the highly conducting RGO layers, allows the carbonaceous matrix material to complement the SiNP-Li+ electrochemistry by becoming highly involved in the charge–discharge reaction mechanisms as indicated by chronopotentiometry and cyclic voltammetry (CV). Particle dispersion improvement was confirmed to be a key component in a composite anode design to maximize Si for high-performance lithium ion battery (LIB) application.

Introduction

The promise of increasing the anode-specific capacity of lithium ion batteries (LIBs) by as much as ten-fold through the substitution of graphite (theoretical capacity of 372 mAh/g [1]) with Si (theoretical capacity of 3,572 mAh/g [2]) has tremendously influenced the direction of recent scientific efforts. Utilization of Si promotes a high-capacity Li alloying reaction which produces a Li-rich phase (Li15Si4) compared with the intercalation reaction with graphite (LiC6). However, the increased accommodation of Li+ ions during charge–discharge cycles induces large volume variations (as much as 370 %) and stress on the bulk anode matrix that ultimately leads to failure. In view of this hindrance, different strategic schemes have been pursued to alleviate the effect of volume expansion including: amorphous thin films [36], nanowires [79], nanotubes [10], and porous morphologies [1113]. Despite these advances, capacity degradation during cycling is still problematic, suggesting electrode fracturing and eventual electrical contact losses. The most promising recent advances have employed Si particles limited to within few nanometers in dimension coupled with carbonaceous material [1422] as a form of support. Here, we focus on composites from SiNP dispersed within graphene sheets to display performance improvements via particle dispersion.

A recent state-of-the art material, graphene, is such a carbonaceous support material. The highly organized sp2-bonded carbon atoms in graphene provide outstanding mechanical and electrical properties compared with other known materials [23, 24]. It has the potential to provide limit/control of Si volume expansion while assisting in electron conductivity within composite anodes. It also has excellent chemical stability that is crucial in minimizing intensive side reactions between the SiNPs and the electrolyte which forms an unstable non-conducting solid–electrolyte interphase (SEI) [14] that degrades electrochemical performance.

Nano-sized Si particles benefit from having high surface area: increasing sites for Li alloying to improve electrode capacity (and rate capability) and increased fracture toughness [25] to promote stability. These properties were demonstrated in recent reports [1517] using 30–50-nm-sized SiNP within a graphene-based matrix producing a SiNP/graphene composite. In these reports the composite anodes were synthesized through two different methods: thermal reduction of graphene oxide (GO)/silicon oxide composite [15]; or physical mixing of exfoliated graphene and Si particles [16, 17]. The composites exhibit high capacities (applied current densities: 100–300 mA/g) with improved retention in spite of significant degradation during the early cycles, a phenomenon that may be attributable to inadequate dispersion of the SiNP. Proper dispersion of the particles is recognized as a critical design parameter of these studies [1517]; very recent works pertaining to voids/spacing acting as buffer region for particle expansion both theoretically [26] and experimentally [18, 27, 28] (for Si and Sn) support this. For this specific simple composite design: SiNP within reduced graphene oxide (RGO) matrix, however, there has been no systematic experimental investigation yet of particle dispersion and the resulting electrochemical performances. Stable dispersion should be the key for the SiNPs to minimize volume expansion stress, display advantageous nanomaterial behavior, and avoid the poor macro-scale performance. Likewise, improvement of dispersion will likely delay the predominant merging/agglomeration of particles during cycling as observed by Li et al. [29] that diminishes the size advantage of nanoparticles as well as introduce more irreversible capacity losses (as a result of hindered ion transport and/or to unstable SEI). In order to contribute further with the advancement of composite anode design, this study investigated the dispersion–electrochemical cycling performance relationship through the use of SiNP/RGO composites prepared using three means of particle dispersion as depicted in Table 1: (S1) sonication, (S2) high-power/energy sonics, and (S3) sonication combined with a surfactant. The effect of preparative methods on the SiNP distribution was evaluated by imaging the size of particles (or aggregates) relative to the lone particle dimension and the resulting electrochemical performances was assessed using variable current density cycles.
Table 1

Effect of preparative condition on SiNP dispersion in RGO matrix

Composite type

Technique

Dispersion Illustration

Particle Distribution

S1

Sonication

https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Figa_HTML.gif

Large agglomerates (>30 × single particle size)

S2

Probe sonication

https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Figb_HTML.gif

Agglomerates (20–10× single particle size)

S3

Surfactant (with probe sonication)

https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Figc_HTML.gif

Small clusters to lone particle distribution

Experimental procedures

Graphite oxide synthesis

The RGO precursor, graphite oxide was prepared using a two-stage Hummers’ method according to Kovtyukhova [30]. Graphite (2 g, Dixon Microfyne, Ashbury, NJ) was pre-oxidized in an oil bath at 80 °C for 4.5 h using 30-mL H2SO4 (95 %) with pre-dissolved K2S2O8 (1 g) and P2O5 (1 g). After cooling, the solution was diluted with 1 L of deionized (DI) water, and then filtered and washed until the filtrate was pH neutral. The pre-oxidized graphite was mixed with 80-mL H2SO4 in an ice bath. While maintaining stirring, KMnO4 (10 g) was added stepwise for the reaction to proceed just below room temperature for 2 h then followed by careful dilution using 150-mL DI water. During this process, the solution temperature was kept below 50 °C. After additional stirring for 2 h, further dilution with 500-mL DI water was then followed by slow addition of H2O2 (30 %, 8.3 mL). The mixture was then allowed to settle overnight and then decanted. The product was purified using repeated rinsing and centrifugation with 5 % HCl and DI water. The resulting graphite oxide suspension (about 10 mg/mL) was then stored in an amber bottle at room temperature.

Nano-Si preparation

A 5–10-nm distribution of SiNP was acquired from Meliorum Nanotechnology (Rochester, NY). The particles were stored in an Ar-filled glovebox with O2 and moisture content both <1 ppm. Prior to use, the particles were exposed to air overnight to develop ample surface hydrophilic oxide layer for proper dispersion.

Composite anode formation

The oxidized SiNPs were weighed and then dispersed in three different ways: (1) sonication (Branson Model 2510, Danbury, CT, 100 W, 50/60 Hz) for 1 h (S1) or (2) sonic probing (Misonix, Ultrasonic Cell Disruptor, Farmingdale, NY, 100 W, 22.5 kHz) for 1 h (S2). Dispersion (3) was prepared by dispersing the SiNP in methanol with a 1 % (v/v) content of n-octyl alcohol (99 %) by sonic probing under the conditions described for S2 (S3). After the dispersion step, the standalone composite anode was produced without the need for binders and conductive diluents following the procedures as outlined in Lee et al.’s [15] work. An appropriate amount of graphite oxide suspension was added to the dispersion to make a 1:1 weight ratio (SiNP:RGO). It was then subjected to the same sonication technique (from prior step) for 2 h. Sonics exfoliate the expanded structure of graphite oxide to form the GO platelets while at the same time dispersing the SiNPs. All sonics-based steps were performed in room temperature conditions and were closely monitored to minimize temperature increase. After this, the sample was vacuum-filtered (setup: Millipore 47 mm all-glass vacuum filter holder: funnel and flask; filter: 0.2 μm pore, Whatman Anodisc) forming a solid composite. During filtration, SiNP cross-over was minimized due to the initial deposition of GO platelets on the filter surface. The resulting anode was then air-dried followed by thermal reduction. The reduction was done using 10 % H2 (balance Ar, 100 mL/min) at 700 °C for 1.25 h. After the reduction, the standalone SiNP/RGO composite material was sampled, weighed, and prepared for testing.

Battery assembly

The cells were assembled in a Swagelok-type cell [31], loaded typically with ~1 mg of the composite (1.5 mg/cm2, 5–10 μm thick) as the working electrode and Li metal (99.9 %, 0.75 mm thick, Alfa-Aesar) acting as both counter and reference electrode (half-cell configuration). A solution of 1.0 M LiPF6 dissolved in ethylene carbonate (EC)/dimethyl carbonate (DMC) 1:1(v/v) solution was used as the electrolyte. The cells were fabricated and cycled in an Ar-filled glove box.

Characterizations

General characterizations

Scanning electron microscopy (SEM) images were taken using a JEOL (Peabody, MA) Model JSM-6510LV-LGS at 25 kV. Chemical composition analysis was done using the equipped Energy Dispersive X-ray Spectrometer (EDS). Thermogravimetric analysis (TGA) was carried out on a Model TA Instruments 2960 (New Castle, DE) at a heating rate of 10 °C/min in air, from which the contents of Si in the composites were determined considering the adjustments for Si oxidation. GO platelet imaging was acquired by atomic force microscopy (AFM) (Digital Instruments Dimension 3100, Plainview, NY) by immobilizing a representative area onto a freshly cleaved mica surface at room temperature under N2 purge. A normal tapping mode silicon cantilever (300 kHz, 40 N/m, T300, nanoScience Instruments/Vista Probes, Phoenix, AZ) was utilized for optimum resolution.

Dispersion characterization

The composite morphology was investigated using a transmission electron microscope (TEM). TEM micrographs were acquired on a JEOL-2010 FasTEM at 200 kV. The TEM samples were prepared by dispersing a small area of dry anode in ethanol with sonication for 1 h. Then a drop of sample solution was casted on 300-mesh copper TEM grids covered with thin amorphous carbon films.

Electrochemical characterizations

Electrochemical cycling of the assembled cells was performed galvanostatically with a cut-off voltage range of 0.02–2.0 V while maintaining a constant current density of 500 mA/g for 40 cycles. This current density was used to determine the highest (optimal) capacity of the composite material. After this step, varying densities of 900, 1,500, and 2,500 mA/g were used after every 10 cycles to assess the rate performance of the cells. The cells were then cycled back to a current density of 500 mA/g to measure the changes in capacity following previous high current density cycles. Cyclic voltammetry (CV) measurement was performed using a single scan rate of 0.04 mV/s over a range of 0.01–2.0 V to gain a better understanding of the reaction mechanisms. The baseline performance was evaluated using around 1 mg of RGO (0.7 mg/cm2, 25 μm thick) without any SiNP loading. In all the electrochemical tests, a Gamry (Warminster, PA) Reference 3000 and series G 300 potentiostat/galvanostat was used.

Results and discussion

RGO characterizations

Figure 1a and b shows that the GO (around 1 μm × 1 μm) from the synthesis are about a nanometer thick platelets. After the filtration/formation process, GO platelets stacked together to form a 4-μm-thick film (Fig. 1c) that appears similar to those of previous studies [16, 32, 33]. Thermal reduction produced a RGO film (Fig. 1d) with an expanded thickness of 25 μm. The RGO film produced is comparable to those that exhibit defects [34], as indicated by the similarity of the Raman profile (Fig. 2). EDS surface chemical analysis shows that the GO has 33 % (w/w) oxygen content (66 % C and <1 % trace: Al, Cl, and S), suggesting the presence of imperfections. Moreover, 10 % oxygen (89 % C and <1 % trace: Al, Cl, and S) is still retained after reduction to RGO, further suggesting that few sp3-hybridized carbon atoms are recovered back into a sp2-hybridized state [16]. This in turn retains distortions in the RGO morphology. Close examination of charge–discharge profiles of high-quality graphene-based electrodes [32] indicate a different insertion mechanism (e.g., edge type and adsorption at both sides of sheets) compared with stable graphite electrodes (charge: 0.1 V, discharge: 0.2 V). Using graphene, charge–discharge occurs at higher potentials, as a result, stable interactions with the electrolyte are possibly compromised and poorer cycling stability (compared with graphite) was observed [32]. Thus, a matrix consisting of imperfect RGO layers may offer better cycling stability (vs. mono to few layers of pristine graphene) and conductivity (vs. graphite).
https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Fig1_HTML.gif
Fig. 1

a AFM image of a representative GO platelet, b sectional analysis indicate that the platelet has 0.92 nm thickness. Cross-sectional SEM images of the c GO film and the d RGO film after reduction

https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Fig2_HTML.gif
Fig. 2

Raman spectra of the RGO film after thermal reduction indicating imperfections in the RGO sheets

The electrochemical performance of the RGO film without any SiNP is summarized in Fig. 3. As a convention, the accommodation of Li+ (lithiation) into the working electrode is identified as the charge reaction while the release of Li+ (delithiation) is called the discharge reaction. The charge–discharge curve (Fig. 3a) demonstrates a gradual changing potential profile compared to the steep profiles of high-quality graphene [32]. The trend indicates that the interaction between Li+ and RGO (vs. high-quality graphene) happens at a lower, more stable potential indicating better cycling stability. Using a lower current density (150 mA/g), the change in charge–discharge voltage during cycling is less pronounced than when larger currents are applied. Figure 3b shows the rate capability of the film. Cycling using 150 mA/g, a first discharge of ~300 mAh/g with 40 % Coulombic efficiency is observed. Such low efficiency is also observed by others [17, 3537] which are attributed to the high surface area of graphene (and RGO) and the formation of SEI as evidenced by their initial charge–discharge curves, CV plots (and differential capacity curves). After 40 cycles using the same current density, the discharge capacity is at 230 mAh/g with >99 % efficiency. Varying the current densities to 300, 750, and 1,500 mA/g displayed average discharge capacities of 200, 160, and 120 mAh/g, respectively. Cycling back to 150 mA/g the capacity returned to 230 mAh/g. It should be noted that the average Coulombic efficiency during the variable current runs was >99 %. This further demonstrates the stability and rate performance of the film.
https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Fig3_HTML.gif
Fig. 3

a Charge–discharge profile of the RGO film using different current densities and b cycling performance of the RGO film

SiNP/RGO composite anode

Kim et al. [14] reported that ~10 nm is the optimal size for SiNPs as a LIB anode material. This particle diameter was found to be the most stable when cycling at 900 mA/g, showing no appreciable changes in particle size observed after 40 cycles. Likewise, modeling by Dimitrijevic et al. [26] predicted that cracking damages during lithiation are negligible for Si (and Sn) particles that are 20 nm and smaller. Thus, the SiNP employed in this study was chosen considering both works: distribution of 5–10 nm.

Dispersion

Figure 4 shows the TEM images of SiNP/RGO composites produced through different dispersion techniques. Large particle agglomerations >300 nm are clearly observed in S1 (Fig. 4a), which suggests that the applied energy is not strong enough to disperse the SiNPs and thus agglomeration occurs. In S2 (Fig. 4b), the agglomerations are smaller than in S1; and smaller particles are now evidently seen dispersed throughout the material. The average size of the smaller particles appears to be around 30 nm. Finally, for procedure S3 (Fig. 4c), a surfactant assisted S2, the agglomerated particles are less evident and small particle aggregates in a range of 10–20 nm are more predominant. This can be attributed to micelle formation which promotes the separation of the SiNPs and mitigates re-agglomeration. Closer examination of the S3 dispersion (Fig. 4d) reveals that small particle clusters (highlighted, some are even single particles) are dispersed uniformly. EDS and TGA results indicate almost identical 50 % (w/w) Si content for all these reduced composite anodes. These tests coupled with the fact that the particles are smallest after procedure S3 further indicate that dispersion of SiNPs is most complete in this composite. Close comparison of degree of Si particle dispersion with the published results of others [1517] illustrates improved single particle distribution with interparticle spacings more than three times the diameter.
https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Fig4_HTML.jpg
Fig. 4

TEM images of the morphology of a S1, b S2, and c S3 composite anodes. d Closer examination of S3 reveals the dispersion of non-agglomerated particles

Electrochemical performance

The electrochemical cycling performance of the three composite anodes is presented in Fig. 5. In order to determine the highest capacity of the composite anodes, a relatively low current density of 500 mA/g was applied. Cycling at this value, the first discharge of all three composites was at ~2,100 mAh/g with the following Coulombic efficiencies: 80 % (S1), 72 % (S2), and 67 % (S3). As the composite composition is 50 % Si, this initial capacity output is expected and approached the theoretical limit. The initial inefficiency can be attributed to the formation of the SEI layer and possibly other side reactions as a result of the high surface area of both SiNP and RGO, with the greatest bearing observed with S3. The efficiencies demonstrated here is a significant improvement compared with previously reported graphene anodes which range from 22 to 58 % [17, 37] and comparable with other reports for SiNP of 58 to 75 % [16, 17, 20]. Figure 5 shows cycling values of pristine Si cycled at 300 mA/g as demonstrated in [16], the result clearly indicates that all three composites exhibit better cycling performance even with higher current densities. Along with high Coulombic efficiency, this strongly suggests that effectively coupling SiNP and RGO can lead to a promising anode material.
https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Fig5_HTML.gif
Fig. 5

a Cycling performance of S1, S2, S3, and Si [16] using various current densities. b Discharge capacity retention for the first 10 cycles

The next 14 cycles of S1 (Fig. 5a) showed a very steep drop in discharge capacity from about 2000 to 1000 mAh/g (~50 % of initial) with an average efficiency of 98 %. At this same point Si alone displayed similar capacity (32 % of initial). However, considering the first discharge, S1 shows improved capacity retention. Continuing until cycle number 40, a much slower rate of capacity decay was observed, with a capacity value of 650 mAh/g at 98 % efficiency. In relation to the results of the pure graphene anode, where only the initial run demonstrated very low Coulombic efficiency (afterward the reversible capacity is stabilized), this observed early cycle decay can be attributed to the SiNP component stability. This predominant observation in the literature has led some to conclude that pristine Si particles have poor cycling stability [17, 38]. Based on the morphology of S1, it seems as the agglomerated SiNPs are acting as macro-scale Si (together with its detriments such as severe volume expansion, structure destruction, etc.) that masks the stabilizing effect of the RGO matrix. With this, it is presumed that even using graphene material that exhibits improved cycling stability such as in [39] would bring minimal improvements unless proper focus to the SiNPs is given.

For composite S2, the capacity retention presents a slight improvement compared to S1. At cycle 15, the reversible capacity is 1,200 mAh/g (98 % efficiency) which translates to a 20 % improvement compared with both S1 and Si. It is also evident that the decay slope is less steep in S2 compared with S1, and at the end of cycle 40 there is still an observed capacity of 850 mAh/g (98 % efficiency). These improvements are likely due to the better dispersion of the SiNPs, as that is the only differentiating trait between the two. As particles are less agglomerated, the SiNP behaves better in cycling the Li+ ions.

Further improvement of the dispersion with the surfactant approach of S3 resulted in improvements in capacity retention as demonstrated in Fig. 5. After 15 cycles, the discharge capacity output is ~1700 mAh/g, a 70 and 42 % improvement from S1 and S2, respectively. The key for this improved performance can be attributed to the capacity retention at early runs. A closer observation of the composites’ relative capacity (vs. the first discharge) after the first 10 cycles as indicated in Fig. 5b distinctly illustrates that significant capacity degradation occurs during the early cycles. At cycle 5, S3 showed substantial capacity retention (96 %) compared with S1 (74 %), S2 (80 %), and Si (60 %). Beyond this point, S3 displays <1 % capacity drop every cycle, strongly correlated with the extent of SiNP dispersion. With this improved capacity retention, S3 has retained a capacity of 1200 mAh/g (57 % of initial) after 40 cycles, 85 and 41 % above from S1 and S2, respectively.

To elucidate the improvement brought by particle dispersion, Fig. 6a shows the charge–discharge profiles (1st, 2nd and 5th runs) while cycling the three composites using 500 mA/g. In general, highly distinct plateau regions were observed beginning at 0.3 V during charge and at 0.5 V during discharge. These characteristic potential values of Si alloying/dealloying reactions are in agreement with others [7, 10, 16, 17, 38]. During the composites’ first charge, SEI formation-related features are the predominant characteristic of the curves. The interphase is known to have diminished ionic conductivity compared with the electrolyte and, as a result, reduced discharges are observed with efficiencies around 70 %. The 2nd and 5th cycles of the composites appear to be very similar except for some key characteristics displayed by the S3 discharges. Two insets of Fig. 6a highlight the discharge curve from (i) 0.4 to 1.7 V and from (ii) 0 to 0.5 V (at 1100 to 2200 mAh/g range). While the S1 and S2 performances reach the dealloying potential at around 0.5 V immediately, S3’s curve involved another possible mechanism that resulted in a linear potential climb with a smaller slope than the other two. After reaching the plateau region, exponential curves continue almost vertically upward to the high potential states. In S3 however it is slightly different; sloping exponential curve with less defined inflection point forming a more circular arch similar to Fig. 3a. This profile appears to have the same characteristics as from S1 and S2 but it may indicate additional reaction mechanism. Further investigation of the distinct features of the S3 cycling curves and their respective current peaks in the CV curves (Fig. 6b) suggest that aside from SiNP alloying/dealloying mechanism, RGO intercalation/deintercalation also occurs. Specifically, S3’s cycling peaks at <0.1 V and those that are broadened in between 0.25 and 0.3 V which are also present in Fig. 6c may be attributed to the RGO’s contribution during lithiation [17] and delithiation, respectively. Through these observations, the following speculations are derived: as improvements in SiNP dispersion happens the occurrence of non-agglomerated particles also improves, through this the network of RGO matrix allows for proper coverage of isolated particles. During charging, Li+ ions get accommodated by both the stable RGO and the high-capacity SiNPs (majority). As the RGO/conduction network covers high surface area nanoparticles, maximized Li accommodation (capacity at highest) with highly efficient electron transport is achieved. At discharge, CV curves indicate a predominant discharge peak at 0.5 V corresponding to a SiNP-based delithiation. The structure likewise allows for efficient transport of Li+ ions out of the composite material with limited irreversibilities as evidenced by the small decrease in intensity of the CV peaks. Also, as the SiNPs are non-agglomerated, cycling stresses that would otherwise cause failure on bulk materials are mitigated via dislocation movement [14, 25]. All these contribute to the improvement of cycling stability of the overall composite anode.
https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Fig6_HTML.gif
Fig. 6

a Charge–discharge profiles of S1, S2, and S3 using 500 mA/g. Insets expanded areas to highlight curve characteristics. CV curves of b S3 and c RGO corresponding with their respective cycling curves

With these desirable traits S3 has displayed improved rate capacities in contrast with the other two composites as seen in the latter part of Fig. 5a. Increasing the current density applied to 900 mA/g, the capacity output of S3 ranged from 1000 to 880 mAh/g. For S1 and S2, smaller discharge capacities were observed, ranging from 600 to 525 mAh/g and 650 to 575 mAh/g, respectively. At the highest current applied (2500 mA/g), S3 still outperformed S1 and S2 by having a stable retained capacity of 440–490 mAh/g. In general, S3 showed improvements in two tested areas: (a) capacity retention (at low current density) and (b) improved rate capability as compared with S1 and S2. However, it is noteworthy that S3 still exhibits a decreasing capacity (<1 % drop every cycle) which can be credited to the structural change brought by the electrolyte penetration forming non-conductive SEI barrier between SiNP and conducting carbon [27]. S3 also shows that at increased rates the drops in discharges are more pronounced than the other two composites. The margins between capacities at each current density level used are becoming smaller: at 900 mA/g 55 % (330 mAh/g), at 1500 mA/g 42 % (200 mAh/g) and at 2500 mA/g 27 % (100 mAh/g). A possible explanation can be seen in the Si morphology after the electrochemical cycles are completed as depicted in Fig. 7a. In this image it is apparent that the originally spherical SiNPs are no longer present; instead a vein-like distribution of Si (highlighted, elemental composition confirmed via point-by-point EDS analysis) is predominant. It appears that, during high rates of electrochemical cycling, phase transformations occur [29], with the electrolyte saturation the SiNPs getting dislodged from the matrix, then agglomerated, and then apparently filling up the volume within the folds of the RGO matrix. A parallel observation was noted by Iwamura et al. [27] that SiNPs (76 nm) held within a void/buffer space of about four times greater volume may drop out of the carbon matrix during lithiation and/or delithiation, leading to a decrease of capacity retention. This phenomenon could likely be minimized provided that the particles are first coated with an overlayer (e.g., carbon coating as in [18, 40]) that will not only improve mechanical anchoring with the RGO matrix but also the interaction with the electrolyte. Aside from the irreversible capacity linked to the transformations, the decrease of the nanoparticle characteristics of Si eventually led to declined performance of S3 (compared with the early cycles). Figure 7b also highlights similar morphological change of Si in S2 but with even larger Si agglomerates than S3. The particle transformation presents the highest surface area decrease in S3 followed by S2 then finally S1. It follows then that the largest observed decrease in capacity occurred in S3. Upon cycling back to the original current loading (500 mA/g), S3 displayed the best recovery obtaining around 80 % of its discharge capacity from cycle 40. However, from cycles 70 to 80 there is evident decay observed in the curves of both S2 and S3 which are attributed to the morphological change. As S3 still has the better surface area and size morphology than S2, a lower capacity decay was consequently observed. This is a direct indication that even as the morphology of SiNPs tend to change at high current rates having a good starting particle dispersion such as in S3 will lead to better overall performance.
https://static-content.springer.com/image/art%3A10.1007%2Fs10853-012-7094-7/MediaObjects/10853_2012_7094_Fig7_HTML.jpg
Fig. 7

Morphological change of Si after high-rate electrochemical cycling of a S3 and b S2

In summary, RGO has been demonstrated as a stable matrix material that supports high-capacity SiNPs. Likewise, improvement of the SiNP dispersion within the composite results in improved capacity retention as well as rate capability. As the anode preparation was mainly through simple sonics-based techniques, future design aspects such as improvements in anchoring/adhesion of SiNPs to the matrix and/or employment of particle protection means to isolate the active particles from unstable SEI can be considered to improve the electrochemical performance.

Conclusion

The combination of high power/energy sonics and surfactants can produce improved particle dispersions containing non-agglomerated SiNPs (10–20 nm) to small particle clusters (<100 nm) within the composite. With this attribute, charge–discharge profile and CV curves of the anode suggest that the mechanism of Li+ accommodation/release highly involves the stable and conducting RGO matrix. Together with the electrochemical benefits of having SiNPs, the composite anode material displays higher capacity retention and better rate performance in contrast with other composites with poor particle dispersion. Particle dispersion improvement is a step forward in composite anode design to maximize Si for high-performance LIB application.

Acknowledgements

Financial support from the Department of Energy (Grant DE-EE0002106) for this research is gratefully acknowledged.

Copyright information

© Springer Science+Business Media New York 2013