Constructing hierarchical porous structure in microsized silicon/carbon nanotubes composite anode with LiF-rich solid-electrolyte interfaces for highly stable lithium-ion batteries

Lithium-ion batteries (LIBs) with silicon microparticle anodes provide a high capacity, low cost, low environmental impact, and ease of production. However, the rapid capacity degradation and low Coulombic efficiency (CE) are impediments to their further development and commercialization, which are mainly caused by large volume variation and unstable solid–electrolyte interface (SEI) of silicon. To break this bottleneck, here, we demonstrate that designing silicon microparticles with nanoporous structure (PSi) and confining the PSi in the carbon nanotube (CNT) segregated network can effectively suppress the volume expansion of silicon, enabling the fabrication of high-performance electrodes. The rate capability and cycling performance of the electrode are further improved by creating a hierarchical open porous structure for the PSi-CNT composite anodes via freeze drying. In addition, the mixTHF electrolyte was employed to get a thin and uniform SEI, which can reduce the breakage of SEI during cycling and improve the CE and stability of the LIBs. As a result, the PSi-CNT composite anode delivers a high specific capacity of 3210.1 mAh g−1 at 1/15 °C rate and an initial Coulombic efficiency of 87.3%. After 100 cycles, the capacity could be maintained at over 2000 mAh g−1 with 99.5% CE. In addition, hierarchical porous structured PSi-CNT composites exhibit excellent rate performance, the specific capacity could reach 2264.5 mAh g−1 at 5 °C rate. The work suggests several effective solutions that could be used to facilitate the future commercialization of silicon anodes.


Introduction
In the next two or three decades, an energy crisis is likely to occur due to the continuous increase in global energy consumption, as well as the shortages and environmental issues associated with conventional energy sources. Today, there is a global transition from fossil fuels to sustainable and renewable energy sources (e.g. solar, wind, and hydropower) to alleviate the energy crisis and environmental problems [1][2][3][4]. Apparently, all further applications of these new energy sources rely on the development of advanced batteries for energy storage [1][2][3][4]. Lithium-ion battery (LIB) is the standard battery technology, which has been widely used in consumer electronics such as cell phones, laptops, and electric vehicles and even shows great potential for large-scale applications in electrical grids [3][4][5][6][7]. To apply to upgraded products and several other next-generation applications, current LIBs must improve their performance to provide higher energy density and higher power density [8][9][10][11]. In general, the performance of LIBs is largely determined by the electrode materials [11]. Therefore, novel designs of electrode materials are urgently required for boosting the performance of LIB. Graphite anode materials have long been commercially used due to their high stability and low cost, but their advanced applications are limited by the low theoretical capacity of graphite (372 mAh g −1 ) [12]. Recently, silicon-based anode materials are increasingly attracting attention because of the significant advantages of silicon, such as unprecedented theoretical capacity at room temperature (3579 mAh g −1 ), low discharge voltage (an average delithiation voltage of Si is 0.4 V), crustal abundance, and environmental friendliness [6,[11][12][13][14][15]. Unfortunately, approaching the theoretical performance of silicon remains challenging because silicon materials suffer from poor electrical conductivity (<10 −3 S cm −1 at 25 • C) and large volume change during the charge/discharge process (320%), which inevitably leads in cracking of the silicon particles and disassembly of the anode and results in poor electrical contact between current collectors and active materials, severely affecting the capacity [14][15][16][17][18][19][20][21]. In addition, the cracking of silicon particles exposes more fresh Si surface, which would react with electrolyte during cycling to form a passive layer, also known as solid electrolyte interphase (SEI), resulting in low Coulombic efficiency (CE) and degraded electrochemical performances [17][18][19][20][21].
A number of solutions have been proposed to address these issues, including nanosized silicon. The issue of silicon pulverization and fragmentation can be effectively mitigated by reducing the silicon to the nanoscale [15]. However, it will necessitate a complicated preparation process and high production costs, which are not conducive to commercialization. Therefore, compared to nanosized silicon, microsized silicon offers cost advantages and is easier to realize subsequent commercializing [16]. Compositing microsized silicon materials with functional additives has been proven effective to improve their performance stability during operation. For example, Park et al demonstrated that forming a segregated structure composed of silicon microparticles (µSi) and carbon nanotubes (CNTs) network can successfully toughen the composite architecture to suppress mechanical instabilities while simultaneously improving the electrical conductivity, enabling the fabrication of high-performance electrodes with fast charge-delivery and near-theoretical specific capacity [16]. However, the cycling stability of the microsized silicon-based composite electrode is still far from satisfactory because of the inferior intrinsic stability of microsized silicon particles.
The main objective of this work is to improve the cycling performance of µSi and CNT composite (µSi-CNT) on the basis of achieving high capacity, further complying with industrial standards. Several approaches to mitigate the capacity fading of µSi-CNT are proposed, including optimization of the silicon structure, µSi-CNT composite structure, and modification of the electrolyte. First, a nanoporous structure was introduced to the µSi to improve the poor cycling stability of pristine µSi anode material by minimizing the effect of large volume changes during cycling. Then, freeze-drying method was used to prevent the oxidation of Si particles during the processing process and create a hierarchical open porous structure in the nanoporous silicon microparticles (PSi) and CNTs composite (PSi-CNT), in which the PSi particles are well confined by the segregated CNTs network, simultaneously enabling fast electron and ion transport and the enhanced robustness of the electrode. Finally, the mixTHF electrolyte was used to further improve the CE and cycling stability by forming the thin and uniform lithium fluoride-based SEI layers which have low adhesion to lithiated alloy surface [19][20][21]. It can effectively avoid the frequent breaking and reforming of SEI layers, which results in low initial Coulombic efficiency (ICE) and poor stability. Combining all of these, we could obtain the PSi-CNT composite anode with high specific capacity and cycling stability.

Electrodes fabrication and characterization
The CNT aqueous dispersion (0.2 wt% SWCNT in water, ∼0.2 wt% PVP as a surfactant stabilizer,) is mixed with the nanoporous PSi (∼10 µm) to fabricate the composite electrodes without any additional additives such as polymeric binder and carbon black. A slurry-casting method was employed to form the composite electrodes onto a copper substrate. The CNT dispersion was first mixed with PSi with different porosities by wet milling to obtain a uniform slurry and a small amount of isopropanol is added into the mixture during milling to facilitate the dispersion of Si particles. The CNT mass fractions (0 ∼ 9 wt%) are controlled by changing the mass ratio between the PSi and CNT dispersion. As a typical fabrication example, 3.76 ml of CNT dispersion was mixed with 100 mg of PSi to obtain an electrode with 7 wt% CNTs. The slurry was then cast onto Cu foil using a doctor blade, then slowly dried at 40 • C for 2 h, followed by vacuum drying at 100 • C for 12 h to remove residual water. To remove the PVP surfactant from the CNT dispersion, the dried electrodes were annealed in Ar gas at 700 • C for 2 h. To obtain a PSi and CNT composite (PSi-CNT) anode with a hierarchical porous structure, after the slurry was cast onto the Cu foil by a doctor blade, the wet films were immediately frozen in a freezer at −20 • C and the frozen films were freeze-dried (Labconco FreeZone 2.5 l−50 • C Benchtop Freeze Dryer) at 0.04 mbar and −50 • C to remove the solvents. The average thickness of the electrodes was 20-30 µm (Sfigure 2(b)) and the mass loading of the active material was 1-1.5 mg cm −1 . For comparison, we used nonporous µSi to fabricate the corresponding anodes (µSi-CNT) using the same fabrication protocol as the PSi-CNT composite anode.
The morphology and microstructure of the samples were examined by field emission SEM (Zeiss Ultra Plus, Zeiss) in high-vacuum mode with an acceleration voltage of 5 keV. The mass (M) and thickness (t) of the electrodes were determined using a microbalance (MSA6, Sartorious) and a digital micrometer after subtracting the mass or thickness of the Cu foils. Energy-dispersive x-ray spectroscopy (EDX) was performed in the same microscope and analyzed using the INCA program.

Cell assembly and electrochemical characterization
The CR-2032 coin cells were assembled to evaluate the Li-ion storage behavior of PSi-CNT anodes. Typically, a layer of polypropylene membrane (K2045 coated PP, Celgard LLC, Charlotte, NC) was used as the separator. The PSi-CNT electrode was used as the working electrode, while Li metal discs with a diameter of 16 mm were used as the counter and reference electrodes. The PSi-CNT films were cut into 12 mm diameter discs, resulting in a geometric electrode area of 1.13 cm 2 . The thickness of the anode was found by measuring the disc using a digital micrometer and subtracting the thickness of the Cu foil. Similarly, the mass of the anode was found using a microbalance and subtracting the mass of the Cu foil. The separator was cut into 18 mm diameter discs. The cell was assembled in an argon atmosphere glove box. The electrolytes used for cell assembly were: (1) 1. To evaluate the electrochemical responses of the material, all half cells were measured using cyclic voltammetry (CV) and galvanostatic charge-discharge (GCD) with a voltage range of 0.005-1.2 V on a potentiostat (VMP3, Biologic) and LAND-CT2001A (Wuhan LAND electronics Co., China). To activate the SEI layer, the cells were cycled five times at 1/30 C rate. The cyclabilities of the cells were then evaluated at 1/15 C rate for 200 cycles. The discharge rate capabilities of the electrodes were investigated using asymmetric charge/discharge conditions; the cells were charged at a fixed 1/30 C rate and then discharged at varied rates. For the post-mortem analysis, the cycled cells were carefully disassembled inside a glove box under an inert atmosphere. The cycled electrodes were then rinsed with dimethyl carbonate several times and dried inside the glove box at room temperature. In the CV measurements, the cells were cycled 0.2 mV s −1 for ten times.

Porous silicon with CNT segregated network composites
The development of the active material focused on modifying the structure of the silicon particles to accommodate the large volume expansion associated with the lithiation and delithiation processes. As depicted in figure 1(a), nanoporous silicon microparticles (PSi) were used to compound with CNTs to obtain a porous silicon and CNTs composite electrode (PSi-CNT). Compared with the conventional nonporous Si microparticles (µSi), PSi with the unique nanoporous structure can effectively reduce the overall volume expansion during operation and, consequently, prevent pulverization and loss of active materials. PSi microparticles with high purity (>99.99%) and an average particle size of ∼10 µm were obtained by electrochemical etching with hydrofluoric acid (HF). As shown in figures 1(b) and (c), uniform open nanosized pores can be clearly observed for those PSi particles. In previous work, Park et al demonstrated that µSi particles can be formed into robust electrodes by mixing with small amounts of CNTs without any additional polymeric binder or conductive additive [16]. Intriguingly, these composites exhibit an unusual segregated network structure [22], as opposed to the typical network structure observed in CNT with nanoparticle composites [23][24][25]. Such segregated networks occur only when the particle size is larger than the filler length (CNT, ∼1 µm). The excluded volume associated with particles is the driving force behind the creation of the segregated network [16]. The PSi-CNT composite electrodes were fabricated using the same method. In a typical process, the CNT dispersion was first mixed with PSi by grinding to obtain a homogeneous slurry. Then, the slurry was cast onto the Cu foil using a doctor blade followed by natural drying. Finally, the obtained electrodes were annealed in Ar gas at 700 • C for 2 h to remove the polymer additives (i.e. PVP) introduced by the CNT dispersion. As shown in figures 1(d) and S1(a)-(c), PSi microparticles can be completely covered by the CNT network when the CNT mass fraction exceeds 7 wt%. Therefore, these CNTs form networked two-dimensional (2D) membranes that wrap and interconnect the PSi particles (figures 1(e), (f) and S2(a)). The membranes are comprised of entangled, van der Waals bonded networks of CNT bundles that envelop the µSi. The network improves both the mechanical and electrical properties of the electrode allowing for the fabrication of improved electrochemical performance [16]. The homogeneous distribution of both PSi particles and CNTs is confirmed by elemental mapping of Si and C, according to EDX ( figure 1(g)). However, in addition to Si and C, O was visible in the EDX mapping ( figure 1(g)). This is due to the fact that the etching process of PSi results in a high concentration of OH on the surface (PSi particles used in this work are produced by electrochemical etching with HF according to Poroussilicon, Ltd), which in turn induces the oxidation of silicon during the annealing process. The oxidation of Si could influence its electrochemical performance.
The electrochemical performance of µSi-CNT and PSi-CNT composite electrodes was investigated by performing GCD and CV tests. Figure 2(a) shows that, compared with conventional µSi, PSi successfully enabled better cycling stability. As depicted in figure 2(a), even though a very high specific capacity (∼3462 mAh g −1 at the 1st cycle) was achieved for the µSi-CNT composite electrode, it dropped rapidly after 50 cycles due to the large volume changes and the cracking of Si during the charging and discharging cycles. In contrast, the specific capacity of the PSi-CNT composite electrode is more stable over cycling. After 200 cycles, the specific capacity of the PSi-CNT composite remained higher than 1000 mAh g −1 , while the specific capacity of the µSi-CNT composite degraded to 284.2 mAh g −1 . Impressively, the capacity retention of PSi is considerably higher than those of previously reported Si nanoparticles or nanosized Si/carbon composites [26][27][28][29][30][31]. We attribute the excellent stability of the PSi-CNT composite electrode to the nanoporous structure of the PSi, which provides additional space for internal expansion, effectively preventing the large volume change of the Si particles [27][28][29][30][31]. Thus, the structural integrity can be well retained during the charging-discharging processes and capacity degradation is successfully prevented. Moreover, increasing the porosity (from 40% to 80%) provides more space for the volumetric expansion, thereby further improving the cycling stability of the PSi-CNT electrode (Sfigures 3(a)-(c)). In addition, the porous structure can increase the active surface area of the Si and facilitate faster ion transport, leading to improved ion intercalation kinetics and enhanced rate performances (i.e. the specific capacity of PSi-NT dropped slowly than Si-CNT with the increase of the current density) (figure 2(b) and Sfigure 3(d)) [32,33]. However, the PSi-CNT electrode shows inferior initial CE and specific capacity because the large surface area results in excessive SEI formation (figure 2(a) and Sfigure 3(c)) [27]. Besides, the inevitable oxidation of the PSi during the electrode preparation process (i.e. heat treatment at 700 • C for 2 h) further reduces its capacity (figure 1(g)). Figures 2(c) and (d) display the CV peaks of the activated µSi-CNT and PSi-CNT electrodes, respectively. According to them, both µSi-CNT and PSi-CNT electrodes exhibit two delithiation peaks at 0.36-0.37 V and 0.51-0.52 V, respectively, and a lithiation peak at 0.15-0.19 V [34][35][36][37][38][39]. The cathodic peak at 0.09 V was observed, indicating the electrochemical contribution of CNTs [39]. However, the shapes of the oxidation and reduction peaks of µSi-CNT and PSi-CNT electrodes are slightly different due to the different silicon structures (i.e. porous and nonporous). In addition, during the first cathodic scan, the peak at approximately 0.01 V corresponds to the lithiation of Si particles, and there are no other reduction peaks because Si particles undergo conversion from crystalline to Li-Si alloy (Sfigure 3(a)) [40]. The subsequent cathodic scan process reveals a new peak at around 0.19 V, which is attributed to the transformation from amorphous Si to Li-Si alloy (figure 2(d)) [41,42]. A broad peak near 1.0 V indicates the formation of a SEI layer (Sfigure 3(e)) [43].

Freeze-dried PSi-CNT electrode
PSi used in this work is obtained by electrochemical etching with HF. Because of this process, there is a significant amount of OH groups on the surface of the PSi, which contribute to the oxidation of the PSi during the heat treatment. This oxidation was responsible for the lower specific capacity of the PSi-CNT composite electrode compared to the nonporous µSi-CNT composite electrode ( figure 2(a)). Consequently, the use of PSi as the active material drives the need for an alternative method to fabricate the electrode that does not involve heat treatment. The CNT dispersion used in this work contains a large amount of water (99.94 wt%) and a small amount of PVP binder (0.2 wt%), therefore, the freeze-drying method could be utilized to create the open porous structure for electrodes ( figure 3(a)). As shown in figures 3(b) and (c), compared to naturally dried the PSi-CNT composite, after freeze-drying, a hierarchical open porous structure can be formed for PSi-CNT composite and the PSi particles are well confined in the CNT/PVP segregated network. The hierarchical porous structure of PSi-CNT composite consists of two levels of porous structures. First is the nanoporous structure of Si microparticle which is obtained by the electrochemical etching, as shown in figures 1(b) and (c). The second porous structure is obtained by the freeze-drying approach ( figure 3(b)). The average size of pores created by the freeze-drying method is 6.7 µm (SFigure 4). This unique structure ensures unblocked ion transport. Meanwhile, the oxidation of PSi can be effectively prevented during the freeze-drying process. Note that the CNTs mass fraction would significantly affect the quality of the PSi-CNT composite electrode (Sfigure 5). Larger CNT mass fraction can enhance the conductivity and improve the integrity and robustness of the electrode by forming highly continuous CNT networks but the concomitant increase of insulative polymer additive loading results in a decrease in capacity (Sfigure 6). Therefore, in this work, an optimized CNT loading of 7 wt% was selected for electrode preparation. And this optimized CNT mass fraction was obtained by comparing series of PSi-CNT composite with different CNT mass fractions (Sfigure 5 and Sfigure 6). Figure 3(d) depicts EDX mapping of the freeze-dried PSi-CNT composite electrode, which confirms that the oxidation of the PSi is successfully prevented. Note that part of the O signal is attributed to the PVP which comes from the CNT dispersion. Although PVP cannot be removed without heat treatment and would degrade the conductive properties of the electrode, the formation of the open porous structure and the high quality of the unoxidized PSi can compensate for the negative effects of PVP, ensuring the improved ultimate performance of the composite electrode. As shown in figures 3(e) and (f), the freeze-dried PSi-CNT composite electrode demonstrates a higher specific capacity than the PSi-CNT composite prepared by thermal annealing treatment. More importantly, due to the open porous structure, the rate performance of the freeze-dried PSi-CNT composite electrode was significantly improved (figure 3(e)), while the long-cycling stability was maintained (figure 3(f)). It is because the large specific surface area of hierarchical porous structured PSi-CNT composite is beneficial to the transport of Li + , thus improving its rate performance.

The mixTHF electrolyte
Compared with Si nanoparticles, Si microparticles offer a low-cost alternative, which is key to commercial applications. However, Si microparticles are more prone to mechanical failure and in principle require more effective methods to inhibit the SEI formation, ensuring higher initial CE, especially the porous Si microparticles because of their higher specific areas [20,21]. Electrolyte plays a significant role in passivating the active material surface and electrolyte during the initial charging and discharging process. To passivate the interphase and suppress the continuous side reactions between silicon particles and electrolyte during the repeated charging and discharging (lithiation and delithiation) processes, adding functional additives into the electrolyte and creating cosolvent systems have been widely used, which can effectively stabilize the SEI film. Various electrolytes and additives were also developed to further improve the CE [44]. Among all electrolytes, carbonate electrolytes containing fluoroethylene carbonate (FEC) as an additive enable the best cycling performance in Si-based anodes, which is also used in this work [45]. Nevertheless, the thick and inhomogeneous SEI (figure 4(a)) on Si is still not robust enough to tolerate the large volume change of porous Si microparticles, resulting in continuous consumption of the Li and electrolyte and the loss of active material [17]. Thus, when we use the conventional EC/DEC/FEC electrolyte non-uniform organic−inorganic SEI forms and it induces high stress and strain at places where expansion is highly inhomogeneous, which easily breaks the weak, mixed organic−inorganic SEI. Consequently, repeated breaking and/or reforming of the SEI leads to a low ICE and poor stability. Recently, novel electrolyte designs have been proposed to stabilize alloy-type anode materials such as microsized Si [17,20,21]. Among them, LiF-based SEIs could be a good choice to accommodate structural change during cycling, because of the low adhesion between LiF and the Si-alloy [20,21]. Additionally, LiF is highly electrically insulating, which prevents electrolyte decomposition and consequently restricts excessive SEI formation [20,21]. It is also mechanically robust enough to suppress alloy pulverization. In this work, we used the mixTHF electrolyte (2.0 M LiPF 6 in 1:1 v/v mixture of THF and MTHF) for PSi-CNT composite anodes to promote the formation thin, uniform, and elastic LiF-based SEI with low adhesion to lithiated alloy surfaces ( figure 4(a)). Therefore, it can reduce the breaking/reforming of the SEI layers during the charging and discharging process, and further improve the ICE and cycling stability of PSi-CNT composite electrodes. The electrochemical performance of the PSi-CNT anode using the mixTHF as the electrolyte was compared the that of the anode using carbonate electrolyte (i.e. 1.2 M LiPF 6 in 3:6:1 (v/v) mixture of EC/DEC/FEC). For the electrochemical test, CV was conducted in a voltage range of 0.01−1.2 V (versus Li + /Li) at a sweep rate of 0.2 mV s −1 to reveal the redox reaction in the cycling process. As shown in figure 4(b), two anodic peaks at 0.37 V and 0.55 V can be attributed to the low-voltage delithiation and high-voltage delithiation of amorphous Li-Si alloys, respectively, whereas one cathodic peak at 0.15 V indicates the lithiation reaction of Si, which is associated with the formation of Li-Si alloy. [34][35][36][37][38][39][40][41][42] However, the anodic peaks merged into one peak at 0.60 V for the cell containing the mixTHF electrolyte and the peak shape becomes sharp and narrow, indicating an improved electrochemical activity ( figure 4(b)) [46,47]. As presented in figures 4(c) and (d), rate and cycling capabilities were both improved obviously when using the mixTHF electrolyte. This is because the optimized electrolyte addressed the issue of an unstable SEI relating to the large volume change during the lithiation and delithiation of silicon. The significantly increased ICE in the mixTHF electrolyte can also verify the electrolyte design principle that the formation of LiF-rich SEI on Si can improve Li plating/stripping CE (the ICE is 57.6% in carbonate electrolyte and 87.3% in the mixTHF electrolyte at 1/15 C rate). The CE remains stable for over 100 cycles in the mixTHF electrolyte. Using the mixTHF electrolyte to improve the SEI allows the specific capacity of the PSi-CNT composite electrode to maintain over 2000 mAh g −1 after 100 cycles with ∼99.5% of CE. The performance enhancement mechanism is a result of the thin and uniform LiF-based SEI with low adhesion to lithiated alloy surfaces, which accommodates the expansion of silicon and reduces breakage of the SEI during cycling. Furthermore, the wide bandgap and insulating nature of LiF reduce the thickness of the SEI [20,21]. Figure 4(e) shows the digital and SEM images of the PSi-CNT composite electrode after the cycling (cycled 100 times at a 1/15 C rate). After charging and discharging cycles, the PSi-CNT electrode maintained its original open porous structure.

Conclusion
In summary, first, we exploit the nanoporous structured Si microparticles to reduce the stress from the large volume changes and accelerate ion transport kinetics to improve the rate and cycling performance of the µSi-CNT composite anode of LIBs. Nanoporous structure can reduce the volume expansion and, consequently, pulverization of silicon microparticles. The electrochemical tests revealed that the nanoporous PSi-CNT composite electrode is more stable over cycling compared to the nonporous µSi-CNT composite electrode. The specific capacity of the PSi-CNT composite remained above 1000 mAh g −1 after 200 cycles, while the µSi-CNT composite dropped to 284.2 mAh g −1 . The nanoporous structure in PSi allows for internal expansion, effectively reducing the volume change of the particles. Based on the nanoporous structure, we further created the micro open porous structure for PSi-CNT composite electrodes via the freeze-drying method. Therefore, the hierarchical porous structure of PSi-CNT composite consisting of two levels of porous structures was created. The freeze-drying technique not only allows us to obtain a highly porous structure for PSi-CNT composites but also prevents the oxidation of PSi and reduces the effect of PVP polymer binder which is electrochemically inert. As a result, the specific capacity and rate performance of the PSi-CNT composite electrode were further improved, while the long-cycling stability was maintained. This is because the highly porous structure provides a larger specific surface area, which is beneficial to the transport of Li + , allowing good performance even at a high C-rate. However increase of the specific area leads to low ICE and stability. To address this issue, we utilized the mixTHF electrolyte to produce a thin and uniform SEI with low adhesion to the lithiated Si surfaces, reducing repeated breaking and reforming of the SEI layers during the charging and discharging cycles to further improve the ICE and cycling stability of the PSi-CNT composite electrode. Combining optimization of the silicon particle structure, composite structure, and electrolyte, the PSi-CNT composite anode can reach a specific capacity of 3210.1 mAh g −1 at 1/15 C rate and with 87.3% initial CE. After 100 cycles, the specific capacity was maintained above 2000 mAh g −1 with ∼99.5% CE. The results of this work offer a potential solution for the commercialization of silicon anodes in the near future.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.