Enhanced Cycle Stability of Crumpled Graphene-Encapsulated Silicon Anodes via Polydopamine Sealing.

Despite silicon being a promising candidate for next-generation lithium-ion battery anodes, self-pulverization and the formation of an unstable solid electrolyte interface, caused by the large volume expansion during lithiation/delithiation, have slowed its commercialization. In this work, we expand on a controllable approach to wrap silicon nanoparticles in a crumpled graphene shell by sealing this shell with a polydopamine-based coating. This provides improved structural stability to buffer the volume change of Si, as demonstrated by a remarkable cycle life, with anodes exhibiting a capacity of 1038 mA h/g after 200 cycles at 1 A/g. The resulting composite displays a high capacity of 1672 mA h/g at 0.1 A/g and can still retain 58% when the current density increases to 4 A/g. A systematic investigation of the impact of spray-drying parameters on the crumpled graphene morphology and its impact on battery performance is also provided.

Introduction

Over the past decade, silicon (Si) has attracted significant attention as a promising anode material for lithium-ion batteries to meet the increasing demand for longer-lasting consumer electronics, longer-range electric vehicles, and large-scale grid storage, which require a higher energy density and lower cost-to-performance ratio. Silicon anodes have the potential to achieve a capacity of 4200 mA h/g based on their fully lithiated state (i.e., Li22Si5),[1,2] which is more than 10-fold higher than that achieved by commercial graphite anodes.[3] Furthermore, more than 90% of the Earth’s crust is composed of silicate minerals,[4] making silicon the second-most abundant element in the Earth’s crust (∼28 wt %) following oxygen.

While silicon-based anodes are touted as the next advance in Li-ion battery technology,[5,6] successful commercialization has not yet been possible due to various unresolved challenges related to poor cycle stability. Chon et al.[7] demonstrated that a large <span class="Disease">stress of >0.5 GPa is generated across silicon particles during the first lithiation, leading to cracking. Repeated expansion and contraction of the particle volume by ∼300% with each cycle leads to further pulverization and capacity fading. Although Liu et al.[8] discovered that the self-pulverization of silicon particles during initial cycles can be effectively slowed down by limiting the size of silicon particles below a critical value (∼150 nm), the volume fluctuation still results in the eventual loss of electrical continuity caused by irreversible deformation. Furthermore, as the potential of the anode decreases below around 1 V (vs Li/Li+), the electrolyte decomposes on the electrode surface forming an electrically insulating but ionically conducting solid electrolyte interface (SEI) layer.[9,10] On materials such as graphite, which exhibit only a small dimensional change, the stable SEI layer can effectively prevent further decomposition of the electrolyte and other undesirable side reactions. In contrast, the SEI layer on Si is continuously destroyed by the large volume change during cycling, which exposes the fresh silicon surface to the electrolyte with each cycle causing low Coulombic efficiency and the continuous consumption of the electrolyte.[11] Additionally, the thickening of the SEI hinders both electronic and ionic charge transport, thereby reducing the rate performance and shortening the cycle life.[12,13] While this SEI instability can be partly ameliorated by the addition of additives which create more elastic SEI layers [e.g., fluoroethylene carbonate (FEC)], these additives only slow down the degradation.

Interfacing Si particles with conductive carbonaceous additives is one of the most common methods used to buffer the volume fluctuation of the silicon and limit the direct formation of the SEI on the silicon surface.[14,15] This has been demonstrated via several methods including mechanical mixing,[16] in situ chemical polymerization,[17] and chemical vapor deposition.[18] However, a conformal but relatively inelastic carbon coating cannot sustain a stable electrode structure from the large volume change of Si during long-term cycling. To prevent the rupture of the carbon protective layer, methods that introduce void space between silicon and carbon have been developed to accommodate the expansion/contraction of the Si. For example, Liu et al.[19] proposed a novel method to prepare carbon-coated Si nanoparticles with a yolk–shell structure. A void space for buffering the volume change was created by removing an artificially created silicon oxide (SiO2) layer between a carbonized polydopamine (pDOPA) shell and silicon core via a hydrofluoric acid (HF) etching. Using this structure, the electrode material maintained a reversible capacity of ∼1110 mA h/g (retention of 74%) after 1000 cycles. Based on a similar method, Liu et al.[20] synthesized a “pomegranate” structure of a microsize Si/C particle using a bottom-up microemulsion method. This large pomegranate-like structure effectively increased the density of the electrode material, which further increased its volume-specific capacity by enabling a higher tap density due to the larger particle size. Despite the improvement in performance of these cells, a mutual drawback of these strategies has been the low density of the electrode, which sacrifices volumetric capacity. Based on the design of the yolk–shell structure, Zhang et al.[21] demonstrated a codeposition method to obtain CaCO3–Si microspheres, which were successively coated by SiO2 and carbon. Finally, the void space was introduced by removing the CaCO3 by immersing the structure in diluted hydrochloric acid (HCl) solution. This method avoided the use of HF and also improved the material density upon the introduction of ferric oxide (Fe2O3) nanoparticles and carbon nanotubes (CNTs). However, these methods involve more complex or even hazardous synthesis techniques (e.g., HF-etching) that also add to the costs associated with fabricating the anode. Hence, a simpler and more efficient strategy for carbon coating is desired to promote the role of silicon-based anodes in the market of commercial Li-ion batteries.

Spray-drying is a scalable unit operation used commercially to generate a significant fraction of commercial powders in the chemical and food industry.[22,23] In the recent decade, this technique has been widely used to controllably prepare electrode materials with desired structures in the field of energy storage and used to prepare materials for lithium-ion batteries,[24−27] lithium–sulfur batteries,[28−30] sodium-ion batteries,[31−33] and electrochemical capacitors.[34−36] Crumpling graphene-based materials around silicon particles via spray-drying has proven to be an effective strategy to create a core–shell structure to buffer the volume change of silicon. Luo et al.[37] first demonstrated that spray-drying of graphene oxide (GO) alone resulted in the formation of 3D crumpled graphene structures formed by the capillary compression of graphene within evaporatively dried, micron-sized aerosols. The resulting crumpled structures significantly enhanced their aggregation and compression resistance compared to powders compressed from more typical 2D morphologies. This is due to the strong graphitic stacking at the folded creases, which hold the structure together through π–π stacking. By adding Si nanoparticles to the dispersion to be aerosolized, the same group[38] found that the GO sheets wrapped around the Si creating a crumpled graphene ball with Si nanoparticles in its core. The graphene was obtained in their process by passing the aerosol droplets through a preheated tube furnace (600 °C) to thermally reduce the GO. The resulting powder was formulated into an electrode and demonstrated a high initial capacity (∼1132 mA h/g), which only dropped by 17% after 250 cycles. After this seminal work, researchers have attempted to improve upon this method by employing spray-dryers to achieve large-scale, automated production of silicon anodes based on crumpled graphene or other carbon shells. However, these works have mainly focused on the introduction of more components into the silicon core to enhance the inside conductivity or robustness of the secondary structure, such as CNTs,[39,40] SiO2,[41] glucose,[42] gelatin,[43] sucrose,[44] polyvinylpyrrolidone,[45,46] polyvinyl alcohol,[47,48] polyacrylic acid-carboxymethyl cellulose,[49] and styrene-acrylonitrile,[50] but the corresponding improvement is still limited compared to Liu’s work.[38] To further optimize this method, a fundamental study about the effect of spray-drying parameters on the physical morphology and dimensions of GO/Si composites and the resulting battery performance is essential, although presently, few papers discuss this in detail. For example, by adjusting process parameters, spray-drying is known to be capable of producing powders with various morphologies including dense or hollow spherical structures and even a donut-like structure.[51,52] Furthermore, the limited inlet temperature of commercial spray-dryers (<220 °C) insufficiently reduces the GO, resulting in a poor electrical conductivity. A following reduction step is required to improve the conductivity of GO, which may impact the structural stability of the crumpled structure.

In this work, a controllable and efficient synthesis route to create crumpled reduced GO (crGO) encapsulated Si nanoparticles was achieved using a mini spray-drying system (Figure ). The effects of spray-drying parameters on the morphology and battery performance of the resulting powders were studied by synthesizing crumpled GO/Si (cGO–Si) and crGO/Si (crGO–Si) composites with varying N2 flow rates, feed concentrations, and GO/Si ratios. Finally, pDOPA was deposited on the surface of the cGO via self-oxidization polymerization, with the goal of further enhancing the structural stability of the cGO and limiting the direct contact between the electrolyte and silicon nanoparticles. This step was followed by heat treatment to reduce the GO, where the pDOPA layer was also carbonized to create nitrogen-doped carbon (cpDOPA). This not only enhances the structural stability of the material but also forms an electrically conductive matrix with efficient lithium-ion transport. This significantly retards the capacity fading from 0.20 to 0.12%/cycle compared to the uncoated sample with an improved capacity of 1379 mA h/g at 1 A/g.

Figure 1

Schematic illustration of the synthesis process of cpDOPA-crGO–Si and diagram of the BUCHI-290 mini spray-dryer used. (① N2 flow in ② spray nozzle ③ drying chamber ④ cyclone ⑤ product vessel ⑥ filter.)

Results and Discussion

In order to achieve the desired highly crumpled GO morphology, we optimized our spray-drying conditions to achieve the best operating conditions for our system. This was first carried out with GO only (no silicon) with a concentration (Cf) of 0.5 mg/mL at varying N2 flow rates (vf) of 246, 473, and 742 L/h through the two-fluid atomizer. The resulting powders were then investigated by scanning electron microscopy (SEM) (Figure S1). The size distribution for each sample is shown in Figure a. With increasing N2 flow rate, the average <span class="Chemical">cGO diameter drops from approximately 3100 to 730 nm. The distribution also narrows with increasing flow, as observed by the standard deviation, decreasing from approximately 630 to 130 nm. The SEM images (Figure S1) reveal the significant morphology changes for samples produced at three different flow rates. At high flow rates, the cGO structures are more spherical in shape with sharp undulations, suggesting a highly crumpled morphology. At lower flow rates the particles transition to a smoother and larger-sized, oblong morphology. The large size difference is attributed to the increasing droplet size expected at lower N2 flow rates (same solution feed rate in each case). As shown schematically in Figure , high-velocity N2 flows through the two-fluid atomizer of the spray-dryer and creates droplet sizes which are inversely proportional to the N2 flow rate. Droplets in the size range of 5–30 μm are achievable according to the data provided by the manufacturer.[53] Thus, low N2 flow rates result in an increase in the number of GO sheets in a single droplet for the same concentration. Since GO is known to accumulate at the interface during droplet evaporation, low flow rates lead to a thicker solid shell. This thicker shell likely prevents further deformation and shrinkage of microdroplets. On the other hand, the smaller number of GO sheets per droplet at high flow rates leads to a thinner, more deformable shell, which is more easily crumpled by the capillary forces generated during evaporative drying. Since Luo et al.[37] demonstrated that the formation of wrinkles enhances the structural stability of crumpled GO, the maximum N2 flow rate (742 L/h) is used to generate highly crumpled structures for all the samples discussed below.

Figure 2

Size distribution of cGO particles prepared based on varying (a) N2 flow rates and (b) GO concentrations; (c) size distribution of crGO–Si particles prepared based on varying GO/Si ratios and GO concentrations; (d) SEM image of crGO–Si; (e) relationship between processing parameters and resulting morphology of crGO–Si.

Figure b shows the size distribution determined by SEM of cGO obtained by spray-drying at varying concentrations of GO (corresponding SEM images are shown in Figure S2). While all particles exhibit a similar crumpled morphology, cGO balls are smaller on average with a narrower particle size distribution when decreasing the dispersion concentration from 1 mg/mL (1165 ± 210 nm) to 0.5 mg/mL (726 ± 128 nm), which is attributed to the decreasing number of GO sheets in each microdroplet. However, there is an insignificant change when the concentration is reduced from 0.5 to 0.2 mg/mL. This may be a limit of the spray-drying system as the cyclone collector can only efficiently capture particles down to ∼1 μm. Since the average size of the resultant particles is close to the threshold, a significant amount of powder may pass to the filter just prior to the system’s exhaust instead of being collected by the cyclone (Figure ). This hypothesis is supported by the drop in the yield observed from >70% at 0.5 mg/mL to <40% at 0.2 mg/mL. Hence, we chose to move forward with cGO prepared at the N2 flow rate of 742 L/h and the concentration of 0.5 mg/mL as the combination exhibiting a smaller particle size, a narrower size distribution, and a relatively good yield with highly crumpled morphology, which we expect would form a more efficient conductive network to electrically contact Si and minimize the effect of volume change in the electrodes.

To incorporate silicon into the <span class="Chemical">cGO, we added Si nanoparticles to the 0.5 mg/mL dispersions and tested two different GO-to-Si mass ratios (1:0.6 and 1:0.3). To investigate the impact of an increased particle size on the performance, we also prepared samples at a dispersion concentration of 1 mg/mL and a GO to Si mass ratio of 1:0.6. The size distributions of the Si-modified powders obtained are plotted in Figure c. Due to the increased solid loading in each droplet (i.e., additional Si), a slight shift to larger diameters compared to that of the unfilled cGO can be observed. All samples display a similar crumpled morphology, as shown in Figure d (more SEM images are shown in Figure S3). There are no silicon nanoparticles evident in the images even under high magnification, suggesting that most, if not all, of the Si nanoparticles are trapped inside the crGO shells. The formation of this core–shell structure is attributed to the diffusivity difference between GO and Si nanoparticles, which is caused by their different hydrodynamic radii (Figure S4). Convection within the droplet during drying (droplet shrinkage) generates the flow of dispersed components to the droplet surface. The resulting concentration gradient causes diffusion in an attempt to redistribute the particles about the droplet volume. The competition between the characteristic timescales associated with these two transport processes is typically quantified by the Peclet number (Pe)[54]where Ri is the initial radius of microdroplets, D is the diffusivity of the dispersed component, and tdry is the drying time. Since the Si nanoparticles are much smaller in hydrodynamic radius than GO sheets (∼41 nm vs ∼ 660 nm on average), according to the Stokes–Einstein relations, they have a much higher diffusivity than that of GO sheets. Thus, Si nanoparticles quickly diffuse to the core to maintain a uniform concentration during droplet shrinkage, while the GO sheets accumulate at the air/water interface.[55,56] The drying time of the droplet in this system is unknown. However, the droplet’s residence time has been estimated to be ∼1.5 s in a previous work that used the same spray-drying system.[57] A drying time of ∼0.1–10 ms has also been estimated,[58] suggesting that drying occurs nearer the inlet. The effect of N2 flow rates and feed concentrations on the resulting morphology of crGO–Si is summarized in Figure e, and samples prepared at high vf and low Cf display a thin shell and small-diameter core.

To evaluate and compare the performance of these core–shell structures as anodes, crGO–Si electrodes with the three different ratios were cycled at varying current densities from 0.1 to 4 A/g, as shown in Figure a. In this work, all gravimetric capacities are reported based on the total mass of Si and <span class="Chemical">carbon shell (including rGO and cpDOPA). For the same GO/Si ratio (1:0.6), the sample prepared with a GO concentration of 0.5 mg/mL retains a higher capacity (38% of 1849 mA h/g at 0.1 A/g) at 4 A/g compared to that of the sample prepared at 1 mg/mL (25% of 1793 mA h/g at 0.1 A/g) at the same current density. This result aligns with our expectation (Figure e) that crGO–Si prepared at low concentration results in smaller structures with less silicon in its core and leads to a more efficient conductive network. For the same GO concentration (0.5 mg/mL), crGO–Si prepared at a GO/Si ratio of 1:0.3 displays a higher capacity at 4 A/g (49% of 1631 mA h/g at 0.1 A/g), exceeding the capacity of crGO (1:0.6) (0.5 mg/mL) at 2 A/g even with a lower content of Si. Since rGO is more conductive compared to Si, this result could also be attributed to the higher overall conductivity due to the increased proportion of rGO in the electrode. The improvement in rate capability due to increased conductivity is also evident in the electrochemical impedance spectroscopy (EIS) results (Figure b). The semicircle in the high-frequency region is attributed to charge transfer resistance (Rct), while the straight line in the low-frequency region represents the Warburg diffusion process (Zw).[59,60] Furthermore, the intersection between the semicircle and x-axis shows the internal resistance (Rs). All Nyquist plots were well-fitted (all fitting curves are displayed in Figure S5) using the equivalent circuit model shown in the inset of Figure b. The order of the electrode conductivity is determined to be crGO–Si (1:0.3) (0.5 mg/mL) > crGO–Si (1:0.6) (0.5 mg/mL) > crGO–Si (1:0.6) (1 mg/mL). As shown in Figure c–d, the voltage profiles show the typical shape for lithiation and delithiation of silicon with plateaus at ∼ 0.3–0.01 and ∼0.2–0.6 V, respectively.[61]

Figure 3

Electrochemical performance of various crGO–Si composites: (a) rate performance; (b) Nyquist plots after the first cycle at 1.2 V vs Li. The equivalent circuit used is shown in the inset; (c) charge/discharge curves at 4 A/g; (d) charge/discharge curves of crGO–Si (1:0.3) (0.5 mg/mL) at varying current densities; (e) long-term cycling performance. (The error bars represent the corresponding standard deviation based on the results from three samples, and all results were measured with an active material (both silicon and carbon) loading of 1 mg/cm2.)

Electrochemical performance of various crGO–Si composites: (a) rate performance; (b) Nyquist plots after the first cycle at 1.2 V vs Li. The equivalent circuit used is shown in the inset; (c) charge/discharge curves at 4 A/g; (d) charge/discharge curves of <span class="Chemical">crGO–Si (1:0.3) (0.5 mg/mL) at varying current densities; (e) long-term cycling performance. (The error bars represent the corresponding standard deviation based on the results from three samples, and all results were measured with an active material (both silicon and carbon) loading of 1 mg/cm2.)

The long-term cycling performance of each sample is shown in Figure e. The low conductivity and larger number of Si nanoparticles in one crGO sack causes a dramatic drop in the capacity of crGO–Si (1:0.6) (1 mg/mL) from 1793 mA h/g at 0.1 A/g to ∼1178 mA h/g at 1 A/g. Only 67% (795 mA h/g) of the initial capacity was retained after 100 cycles. The curves for the other two samples, prepared at low concentration, show similar tendencies and retain a reasonable capacity at high current. However, the crGO–Si (1:0.3) (0.5 mg/mL) displayed a much higher capacity of 1016 mA h/g (78% of the initial capacity), while crGO–Si (1:0.6) (0.5 mg/mL) only retained 68% (925 mA h/g) of the initial capacity after 100 cycles. As shown in Figure S6, crGO–Si (1:0.3) (0.5 mg/mL) displays much better cyclic stability compared to that of rGO–Si (1/0.3), which was prepared without spray-drying the mixture. Moreover, we also examined the cyclic stability of crGO–Si (1:0.3) (0.5 mg/mL) produced at a low N2 flow rate (246 L/h), whose capacity significantly dropped by 52.5% after 100 cycles. Hence, crGO–Si (1:0.3) (0.5 mg/mL) produced at high N2 flow rate (742 L/h) exhibits the best cyclic stability, given its higher rGO/Si ratio, which buffers the volume fluctuation of Si most efficiently among these three samples while maintaining an effective conductive network. From here on, all samples prepared in the following works are based on a GO concentration of 0.5 mg/mL.

Huang et al.[38] directly spray-dried and heat-treated crGO/Si at 600 °C via a modified spray-drying system, which displays better cyclic stability (250 cycles vs 100 cycles for ∼80% capacity retention) and a higher initial Coulombic efficiency (73%) compared to that of crGO (1:0.3) (0.5 mg/mL) (68.6%). We suspect that the reduction step in our procedure might affect the structural stability of cGO/Si due to the possible exfoliation of the GO exposing more Si nanoparticles directly to the electrolyte. Hence, we further reinforced the cGO shell by coating it with an external layer of pDOPA. Inspired by the adhesive nature of mussels for attachment to wet surfaces, pDOPA coating is a simple and versatile method for surface functionalization through spontaneous self-polymerization of DOPA monomers in basic aqueous solutions.[62] The following heat treatment converts the pDOPA into a nitrogen-doped carbon layer on the surface of crGO, which can improve both cyclic stability and charge transfer. To confirm the successful coating of pDOPA, Fourier transform infrared (FTIR) spectroscopy was carried out on pure cGO and pDOPA-cGO, as shown in Figure a. Two new peaks emerge in the spectra of pDOPA-cGO related to the N–H bond and C–N bond at 1504 and 1385 cm–1, respectively, suggesting the existence of nitrogen-containing groups that are introduced by the pDOPA layer. Also, GO is partially reduced during the polymerization of dopamine, as indicated by the decreased peak intensity at 1735 cm–1 (C=O) in the spectra of cGO.[63] Furthermore, pure cGO and pDOPA-cGO were stirred in DI water for 24 h and then freeze-dried for SEM analysis, as shown in Figure b,c, respectively. All pure cGO sacks are found to have a completely different morphology upon redispersion in water. They no longer appeared as crumpled balls but as thick lamellae, suggesting that rGO sheets unfold and aggregate in water, resulting in the collapse of the crumpled structure. On the other hand, the pDOPA-cGO sacks are able to maintain their original crumpled morphology. This suggests that the pDOPA effectively “glues” the structure together and prevents it from restructuring in water. The polymerization reaction involving pure dopamine only takes a few minutes, as indicated by the rapid change in color from transparent to light brown, implying that this process is faster than the restructuring of cGO in water. Furthermore, pDOPA is also found to enhance the structural stability of cGO during heat treatment, as shown in Figure S7. Without the pDOPA coating, the cGO exfoliated in the tube furnace due to the rapid expansion of gases generated upon thermal reduction of GO to rGO. The pDOPA coating rigidifies the structure, enabling the gases to escape without undergoing damaging expansion.

Figure 4

Characterization of pDOPA-coated cGO and crGO: (a) FTIR spectra of cGO and pDOPA-cGO; SEM images of cGO stirred in DI water for 24 h without pDOPA (b) and with pDOPA (c); (d) rate performance and (e) cyclic stability of cpDOPA-crGO scaffold at 2 A/g. (The error bars represent the corresponding standard deviation based on the results from three samples, and all results were measured with an active material (carbon only) loading of 1 mg/cm2.)

Cell performance of pDOPA-crGO without Si was first investigated to study the electrochemical stability of the carbonaceous material formed. Cells assembled using this material were cycled at varying current densities ranging from 0.1 to 4 A/g (Figure d), exhibiting a relatively small drop in capacity from 323 to 216 mA h/g. Additionally, as shown in Figure e, cpDOPA-crGO displays excellent cycling stability with an average Coulombic efficiency of >99.9% and only ∼10% loss in capacity observed over 1000 cycles. According to these results, cpDOPA-crGO is a stable carbon scaffold which will be used to protect Si nanoparticles, as discussed in the sections to follow.

After the addition of Si nanoparticles into cpDOPA-crGO based on a GO/Si ratio of 1:0.6, we first investigated the amount of <span class="Chemical">cpDOPA coated on the surface of crGO–Si using TGA, where samples were pyrolyzed under air flow to 650 °C. As shown in Figure a, there is a significant mass loss at around 500 °C, which should be attributed to the decomposition of the carbonaceous materials. Some mass gain is observed at higher temperatures due to the oxidation of Si nanoparticles. Since the addition of cpDOPA would not affect the mass ratio between crGO and Si, according to the silicon content detected in cpDOPA-crGO–Si (1:0.6) and crGO–Si (1:0.6), the content of cpDOPA is calculated to be around 17.5 wt %, while the total carbon content (rGO + cpDOPA) is around 42.1 wt %. Here, it should be noted that the cpDOPA is coated on crGO–Si (1:0.6) rather than crGO–Si (1:0.3), which exhibits better electrochemical performance, in order to maintain its theoretical capacity by avoiding the use of too much carbon. By comparing cpDOPA-crGO–Si (1:0.6) and crGO–Si (1:0.3) containing a similar carbon content, a fairer conclusion can be drawn about whether or not the extra cpDOPA sealing can help to improve the electrochemical performance of our materials. A representative transmission electron microscopy (TEM) image of a cpDOPA-crGO–Si particle is shown in Figure b. Some parts near the outside of the crumpled structure are nearly transparent to the electron beam and are likely just folds of pDOPA-coated rGO. The inner structure is much darker in contrast likely due to the denser Si-containing core. According to the EDX mapping for carbon, nitrogen, and silicon shown in Figure c–e, silicon is concentrated in the core and the texture of individual silicon nanoparticles approximately 50 nm in diameter can be seen. A strong carbon signal throughout the sample and a lower-intensity nitrogen signal are detected and distributed uniformly across the crGO surface. Moreover, HRTEM was carried out on the crumpled edge of a cpDOPA-crGO–Si (1:0.6) sample at varying magnifications, as shown in Figure S9. The resulting images clearly suggest that silicon nanoparticles are covered by a layer of rGO combined with amorphous carbon (cpDOPA) with a thickness of ∼5 nm. The lattice fringes (d = 0.31 nm) reflect the (111) crystal face of silicon, indicating that the Si nanoparticles inside are still highly crystalline. After spray-drying, Si NPs were not significantly oxidized, as indicated by a negligible increase in oxygen content observed by EDX and no additional peaks appearing in the XRD profiles, as shown in Figure S10.

Figure 5

Characterization of crGO–Si composites with and without cpDOPA coating: (a) TGA curves of varying samples; (b) TEM image of a single cpDOPA-crGO–Si ball and elemental mapping images showing the distribution of carbon (c), nitrogen (d), and silicon (e).

XPS was carried out to characterize the surface elemental composition of bare silicon, crGO–Si (1:0.3), and cpDOPA-crGO–Si (1:0.6). As depicted in Figure a, the presence of a N 1s peak in the spectrum of cpDOPA-crGO–Si (1:0.6) suggests the successful introduction of cpDOPA. Approximately 1.21% N is doped into the carbon shell. Moreover, according to the surface elemental analysis, the silicon content decreased from 47.5 to 7.68%, while the carbon content increased from 27.14 to 73.49% after spray-drying and heat treatment, together with GO sheets. This reduction in the silicon signal, due to the short penetration depth of the XPS system, demonstrates that rGO effectively covers the surface of the Si nanoparticles. After the coating of cpDOPA, the Si content dropped further to 2.13%, while the carbon content reached 82.38%. To obtain chemical bonding information, we carried out a more detailed analysis for the C 1s peak and N 1s of cpDOPA-crGO–Si (1:0.6). For the C 1s spectrum (Figure b), the main peak was well fitted and deconvoluted into three peaks, which are attributed to O–C=O (288.9 eV), C–N (285.7 eV), and C–C (284.8 eV).[64,65] The small area of O–C=O illustrates that most oxygen-containing groups were removed during heat treatment, and the presence of C–N suggests the doping of nitrogen due to the coating of cpDOPA. To determine the type of doped nitrogen in this material, we also did peak fitting for the N 1s spectrum. As shown in Figure c, the three peaks are assigned to graphitic nitrogen (403.6 eV), pyrrolic nitrogen (400.4 eV), and pyridinic nitrogen (397.9 eV).[64,65] XRD profiles are plotted in Figure d. The peaks at 28.4, 47.3, 56.1, and 69.1° are assigned to (111), (220), (311), and (400) lattice planes of crystalline silicon, respectively. A broad and nearly undiscernible peak around ∼27° in the XRD pattern of crGO–Si indicates that the rGO shell indicates little to no restacking of the rGO. However, this peak effectively disappears after the introduction of cpDOPA, which, as expected, forms an amorphous carbon layer.

Figure 6

Characterization of crGO–Si, cpDOPA-crGO–Si, and bare Si: (a) high-resolution XPS survey spectra of varying samples; high-resolution XPS narrow scan spectra of cpDOPA-crGO–Si for C 1s (b) and N 1s (c); (d) XRD patterns of varying samples.

Galvanostatic charge/discharge testing and EIS were carried out on cpDOPA-crGO–Si (1:0.6) to demonstrate the improvement in cell performance by adding the <span class="Chemical">cpDOPA coating as compared to samples without the coating. As shown in Figure a, cpDOPA-crGO–Si (1:0.6) delivers 1672 mA h/g at 0.1 A/g, which drops to 975 mA h/g at 4 A/g, retaining 58% of its capacity at the high rate of lithiation. In contrast, crGO–Si (1:0.3) retained only 49% of its initial capacity at 0.1 A/g after the current density increases to 4 A/g. The improved rate capability suggests that the conductive network is further improved in the cpDOPA-coated sample, which is consistent with the results of EIS, where cpDOPA-crGO–Si (1:0.6) exhibits the smallest Rct (72 Ω), as indicated by the smallest semicircle in the high-frequency region of the Nyquist plot (Figure b). Although cpDOPA-crGO–Si (1:0.6) has a carbon content similar to that of crGO–Si (1:0.3), the nitrogen-doped carbon coating provides a more efficient pathway for lithium-ion transfer. The first charge/discharge curves of cpDOPA-crGO–Si (1:0.6) and crGO–Si (1:0.3) are plotted in Figure c, and the sharp slope at ∼ 1.5–0.1 V in the first discharge curve indicates the formation of the SEI layer, while the long platform at ∼ 0.1–0.01 V is attributed to the transformation from crystalline Si to an amorphous LiSi. Also, crGO–Si (1:0.3) exhibits more irreversible capacity to form an SEI layer, leading to a lower initial Coulombic efficiency (68.6%) compared to 76.3% for the cpDOPA-crGO–Si (1:0.6). The improvement suggests that the cpDOPA-crGO shell may better seal the Si from direct contact with the electrolyte than just the crGO alone. For reference, pure Si exhibits an even poorer initial Coulombic efficiency of 57.2%. The improvement in initial Coulombic efficiency may also be explained by the significant drop in the surface area, estimated by gas adsorption of 48.4 to 9.2 m2/g before and after the cpDOPA coating procedure, respectively. After the first discharge, the silicon anode only converts between amorphous LixSi and amorphous silicon corresponding to the typical potential plateaus at ∼ 0.3–0.01 V during discharging and ∼0.2–0.6 V during charging, as shown in Figures c and 6d.[61] With increasing current density, the polarization of Si leads to an apparent increase in the delithiation potential and reduction of the lithiation potential.

Figure 7

Electrochemical performance of crGO–Si composites with and without cpDOPA coating: (a) rate performance, (b) Nyquist plots after the first cycle and equivalent circuit (inset); (c) first charge/discharge curves at 0.1 A/g; (d) charge/discharge curves of cpDOPA-crGO–Si (1:0.6) at varying current densities. (The error bars represent the corresponding standard deviation based on the results from three samples, and all results were measured with an active material (both silicon and carbon) loading of 1 mg/cm2.)

Electrochemical performance of crGO–Si composites with and without <span class="Chemical">cpDOPA coating: (a) rate performance, (b) Nyquist plots after the first cycle and equivalent circuit (inset); (c) first charge/discharge curves at 0.1 A/g; (d) charge/discharge curves of cpDOPA-crGO–Si (1:0.6) at varying current densities. (The error bars represent the corresponding standard deviation based on the results from three samples, and all results were measured with an active material (both silicon and carbon) loading of 1 mg/cm2.)

The long-term cyclic stability of cpDOPA-crGO–Si (1:0.6) was investigated by galvanostatic charge/discharge and compared to uncoated <span class="Chemical">crGO–Si (1:0.3), as shown in Figure a. cpDOPA-crGO–Si (1:0.6) displays an outstanding cyclic stability, which only dropped from 1379 to 1206 mA h/g after 100 cycles and 1038 mA h/g after 200 cycles with corresponding capacity retentions of 87.5 and 75%, respectively. Uncoated crGO–Si (1:0.3) only achieved a capacity of 783 mA h/g after 200 cycles, exhibiting a capacity decay rate of 0.20%/cycle, which is significantly worse than the cpDOPA-coated sample, where a decay rate of 0.12%/cycle is observed. The SEM images of cpDOPA-crGO–Si (1:0.6) after 200 cycles are shown as in Figure S12, which indicates no obvious cracking or delamination and that the original crumpled structure of the crGO appears largely maintained. The Nyquist plot of cpDOPA-crGO–Si (1:0.6) after 200 cycles was also measured, as shown in Figure S13, which still exhibits a similar shape, comprising a semicircle at the high-frequency region and a straight line at low frequencies compared to the plot obtained after the first cycle. The small increase in the semicircle diameter indicates that this cpDOPA-coated crGO–Si can maintain a relatively intact electrochemical structure even after long-term cycling. Based on the capacity of the pure cpDOPA-crGO shell (323 mA h/g at 0.1 A/g and 216 mA h/g at 4 A/g, as shown in Figure d) and TGA analysis (containing ∼42.1 wt % of carbon, as shown in Figure a), we calculated the contribution of carbon in cpDOPA-crGO–Si (1:0.6), which suggests that only 8.1 and 9.3% of capacity should be attributed to the carbon shell at 0.1 and 4 A/g, respectively. Based on the above calculations, the silicon alone displays capacities of ∼2654 mA h/g at 0.1 A/g and ∼1527 mA h/g at 4 A/g, which is around 73.9 and 42.5%, respectively, of its theoretical capacity (3590 mA h/g) at room temperature.[66] To meet the high energy densities required for practical application of this battery technology, we also checked the cycling stability of cpDOPA-crGO–Si (1:0.6) with increased active material loading (2.5 mg/cm2 including both silicon and carbon). Even with a 2.5× increase in mass loading, this material still shows a promising cycle life, as shown in Figure b. The areal capacity of cpDOPA-crGO–Si (1:0.6) only dropped from 2.46 to 2.05 mA h/cm2 after 100 cycles, and this capacity is very close to the value of commercial graphite anodes.[67] Hence, we can conclude that the coating of cpDOPA significantly improves the structural stability of the crGO–Si composite. Compared to other Si/C anodes prepared based on the same spray-drying technique (Table S2), cpDOPA-crGO–Si displays competitive capacity and improved cycling stability with a simple and efficient production process. Although Yan et al.[68] very recently conducted a similar pDOPA coating method, our material exhibits better cycle life and uses much less electrolyte additives [5% FEC vs 10% FEC and 2% vinylene carbonate (VC)], which is attributed to the optimization of spray-drying parameters.

Figure 8

Cyclic stability of cpDOPA-crGO–Si with varying loadings: (a) Comparison of cycle life between crGO–Si and cpDOPA-crGO–Si; (b) cycling stability of cpDOPA-crGO–Si with an increased mass loading. (The error bars represent the corresponding standard deviation based on the results from three samples; all results in Figure a were measured with an active material (both silicon and carbon) loading of 1 mg/cm2.)

Conclusions

In summary, the effect of the feed concentration, nitrogen flow rate, and GO/Si ratio on the resulting physical morphology, size distribution, and battery performance of crGO–Si was evaluated. It is confirmed that a highly crumpled structure with a small and uniform size distribution is produced at a high N2 flow rate of approximately 742 L/h and a diluted GO dispersion concentration of 0.5 mg/mL with a low GO/Si ratio (1:0.3), which results in an electrode with the highest structural stability and most efficient conductive network to buffer the volume change of Si. When cycled at a current density of 1 A/g in the range of 0.01–1.5 V, crGO–Si (1:0.3) (0.5 mg/mL) delivers an improved cycle life with a capacity loss rate of 0.20% per cycle over 200 cycles. A layer of nitrogen-doped carbon (cpDOPA) was then coated on the surface of crGO–Si (1:0.6) (0.5 mg/mL) to further enhance the sealing and stability of the Si–C core–shell structure. The HRTEM images clearly reveal that crystalline Si nanoparticles are encapsulated by a layer of rGO combined with amorphous carbon (cpDOPA). Based on a silicon content (57.9 wt %) which is similar to that of crGO–Si (1:0.3), cpDOPA-crGO–Si (1:0.6) displays much better rate capability and cyclic stability. It can still maintain a high capacity of 1007 mA h/g after 200 cycles at 1 A/g, the reinforced material displays significantly enhanced cycle stability with a lower decay rate of 0.12% per cycle. According to the SEM images after cycling, the crumpled structure is still largely maintained without severe accumulation of the SEI. Meanwhile, the optimized initial Coulombic efficiency (76.3%) also demonstrates that the addition of cpDOPA coating effectively decreases direct contact of Si by the electrolyte.

Experimental Methods

Synthesis of GO

GO was synthesized from natural flake graphite (Alfa Aesar) by Tour’s modified Hummer’s method;[69] 3 g of <span class="Chemical">graphite and 18 g of KMnO4 were added into a continuously stirred acid mixture composed of 360 mL of sulfuric acid (H2SO4) and 40 mL of phosphoric acid (H3PO4). The oxidation reaction was conducted at 50 °C for 16 h. Around 6 mL of hydrogen peroxide (H2O2) was added to the resulting mixture after cooling down to room temperature where the color changed from purple to golden yellow. The resultant suspension was then centrifuged for 30 min (3000 rpm, rotor diameter 15 cm). The supernatant was decanted and discarded, and the precipitant GO was dispersed in 30% HCl and centrifuged again using the same settings. The above washing step was repeated twice using 10% HCl and four times using wash ethanol. In order to transfer GO from ethanol to deionized (DI) water, a dialysis bag with 12–14 kDa molecular weight cutoff (MWCO) was used to exchange the GO dispersion’s ethanol solvent with DI water. The DI water was frequently replaced with a fresh amount for 3 days, and the resulting GO aqueous dispersion was stored for subsequent steps.

Preparation of Encapsulated Silicon

Both neat cGO and cGO–Si were produced via a BUCHI-290 mini spray-dryer, as shown in Figure . An inlet temperature of 200 °C (with an outlet temperature of ∼80 °C) and a feed pump rate of 20% of the machine’s maximum (which converts to ∼6 mL/min of water) were used. As a control, neat crumpled GO (i.e., not with Si) was produced at varying N2 flow rates (246, 473, and 742 L/h) and GO concentrations (0.2, 0.5, and 1 mg/mL) to determine the ideal structure of carbon scaffold to encapsulate Si. Si nanoparticles (∼50 nm, Strem Chemical Inc.) were then added into a GO aqueous dispersion based on two GO/Si mass ratios (1:0.3 and 1:0.6) and spray-dried at different GO concentrations (0.5 and 1 mg/mL) at a fixed N2 flow rate of 742 L/h. For subsequent electrochemical testing, the obtained cGO–Si was thermally reduced to crGO–Si to increase its electrical conductivity in a tube furnace (Carbolite). To maintain its crumpled structure during thermal treatment, cGO–Si was reduced at a low initial heating rate (1 °C/min) under a flowing mixture of argon (95%) and hydrogen (5%) gases from room temperature to 400 °C. It was held at 400 °C isothermally for 2 h. It was then heated at a faster rate (5 °C/min) up to 800 °C and held isothermally for 3 h before allowing to cool down to room temperature.

Preparation of Carbonized Poly(dopamine)-Coated crGO–Si (cpDOPA-crGO–Si)

For a typical synthesis of cpDOPA-crGO–Si, 150 mg of <span class="Chemical">cGO–Si was mixed with 150 mg of dopamine hydrochloride (DOPA·HCl) in 75 mL of Tris–HCl buffer (10 mM, pH = 8.5) and stirred for 24 h. This suspension was then divided into two 50 mL centrifuge tubes and centrifuged for 15 min (3000 rpm, rotor diameter 15 cm). The supernatant was discarded, and the precipitated pDOPA-cGO–Si was dispersed in DI water and centrifuged thrice again. The dried pDOPA-cGO–Si powder was collected using a LABCONCO freeze-dryer. In order to carbonize pDOPA and reduce cGO, pDOPA-cGO–Si powder was heat-treated in the same way as cGO–Si in the last section.

Materials Characterization

SEM images were taken on a field-emission scanning electron microscope (Zeiss LEO1550) with an acceleration voltage of 10 kV. The size distributions of crGO and crGO–Si prepared at varying conditions were obtained by counting at least 100 composite particles in SEM images for each sample. TGA (Q500, TA Instruments) was performed by heating the sample under air flow from room temperature to 650 °C at a rate of 5 °C/min. TEM images were taken on an energy-filtered transmission electron microscope (Zeiss Libra 200MC) with an acceleration voltage of 200 kV. A Bruker energy-dispersive X-ray (EDX) spectrometer system integrated into the TEM system was used for elemental mapping, and FTIR spectra were acquired using an FTIR spectrometer (PerkinElmerSpectrum TwoTM). Hydrodynamic radii were determined by dynamic light scattering (DLS, Zetasizer Nano-ZS90, Malvern). X-ray photoelectron spectroscopy (Thermal Scientific KAlpha XPS spectrometer, 150 eV) was carried out to analyze the surface elemental composition and chemical bonding. XRD was carried out using an XRG 3000 X-ray diffractometer (Cu Kα radiation). An FEI Titan 80–300 LB was used to obtain high-resolution TEM (HRTEM) images of cpDOPA-crGO–Si. The specific surface area of various samples was measured by a Micromeritics Gemini VII surface-area analyzer based on Brunauer–Emmett–Teller (BET) theory.

Electrochemical Testing

In order to investigate and compare the electrochemical performance of crGO–Si and c<span class="Chemical">pDOPA-crGO–Si prepared at varying conditions, the working electrode was prepared by mixing the active material, carbon black (carbon super P, MTI), and sodium alginate (Sigma-Aldrich) in DI water with a mass ratio of 65:20:15 using a rotor/stator homogenizer. The resulting slurry was cast onto a copper foil by a typical film-casting doctor-blade method, followed by drying at 80 °C under vacuum overnight. All rate capability and cyclic stability results were obtained by assembling the fabricated working electrode [∼1 mg/cm2 of active material (both silicon and carbon)] for all studies except for Figure b, where ∼2.5 mg/cm2 of the active material (both silicon and carbon) was applied with a lithium metal foil (Sigma-Aldrich, 99.9% trace metal basis) in a coin-type half-cell. A Whatman glass microfiber (Grade GF/A) was used as a separator, and 1 M LiPF6 in a 1:1 v/v mixture of ethylene carbonate (EC) and dimethyl carbonate (DMC) containing 5 vol % FEC purchased from Canrd China was used as the electrolyte. All cells were assembled in an Ar-filled glovebox (<1 ppm O2 and water) and cycled between 0.01 and 1.5 V versus Li/Li+ using a LANHE multichannel battery tester (Wuhan LAND Electronics Co.). EIS was carried out on a SP-300 potentiostat (BioLogic) in the range of 1 MHz to 100 mHz with an AC amplitude of 10 mV. Nyquist plots were recorded after the first full cycle (after one discharge and one charge at 0.1 A/g) or after 200 cycles at 1 A/g. The electrodes were charged (delithiated) to 1.5 V, disconnected from the battery tester, and connected to the potentiostat, where the OCV was around 1.2 V for each sample. EIS was then carried out at this DC voltage vs the lithium metal.