Fast-chargeable lithium-ion batteries by μ-Si anode-tailored full-cell design

Significance

In the current design of lithium-ion battery full cells, the capacity ratio of the anode to the cathode, the so-called N/P ratio, is kept close to 1 to maximize energy density. However, this cell setting is not conducive for high-capacity electrode materials, such as silicon (Si), which undergo significant volume changes over the full state-of-charge (SOC) range. For micrometer-sized Si (μ-Si), we propose an anode-tailored full-cell design (ATFD). Through fundamental analysis focusing on lithiation kinetics and the degree of volume expansion, we have identified the SOC range of 30 to 70% as most suitable for long-term cycling. The μ-Si ATFD-based full cells demonstrated higher energy density and improved fast-charging capabilities compared to their graphite-based counterparts in the conventional cell settings.

Abstract

Silicon (Si) anodes have long been recognized to significantly improve the energy density and fast-charging capability of lithium-ion batteries (LIBs). However, the implementation of these anodes in commercial LIB cells has progressed incrementally due to the immense volume change of Si across its full state-of-charge (SOC) range. Here, we report an anode-tailored full-cell design (ATFD), which incorporates micrometer-sized silicon (μ-Si) alone, for operation over a limited, prespecified SOC range identified as 30−70%. This range allows homogeneous (de)lithiation throughout the electrode, accompanied by an acceptable level of volume change. The ATFD-based cell exhibits 21.3% higher gravimetric energy density than that of its graphite-based counterpart in a commercial 18650 cylindrical cell and 84.6% capacity retention after 500 cycles even at a fast-charging rate of 3 C. This study indicates that the partial, intermediate SOC operation of the μ-Si anode can markedly increase the energy density and boost the fast-charging capability of a LIB cell, a challenging task in traditional cell engineering.

High-capacity electrode materials play a decisive role in extending the driving range of electric vehicles (EVs) powered by lithium-ion batteries (LIBs) to facilitate the smooth transition from vehicles powered by internal combustion engines (1–3). On the anode side of LIBs, silicon (Si), which stores lithium (Li) ions via an alloying mechanism, is well aligned with this transition because Si can offer much higher specific capacity compared to that of its conventional graphite counterpart. Along with the alloying reaction mechanism, the possibility of decreasing the electrode thickness enables the Si anode to shorten the charging time of a LIB cell, thereby targeting so-called extreme fast-charging (XFC) (4–6). To the benefit of fast-charging, the higher redox potential of Si compared to that of graphite renders the Si anode less prone to Li plating during the charging process, which would degrade the cyclability and heighten the likelihood of safety event (7). The utilization of Si in anodes is also beneficial in ensuring more sustainable supply chains, as the procurement of raw materials for graphite is confined to a limited number of nations worldwide (8).

The battery community has invested considerable effort in developing Si anode materials that can sustain the immense volume expansion during lithiation. Research has identified the critical nature of nanometer dimensions to withstand pulverization (9–11). As a result, a variety of nanostructured Si materials were shown to demonstrate enhanced cyclability. In particular, Si-carbon blends and SiO_x (x ~ 1) have been most widely adopted by the commercial sector (12). Nonetheless, these nanodimensional Si anodes are characterized by relatively low initial Coulombic efficiency (ICE) as a result of their large surface-to-volume ratios, sacrificial volumetric capacity because of the needful inclusion of inactive components, and high cost of their synthesis processes. Industry is particularly concerned about the limited room for lowering the cost of nano-Si-based materials. Although pure Si microparticles (μ-Si) are superior in all these respects, they have not yet been commercially adopted because μ-Si inevitably pulverizes when stress accumulates in a particle during volume expansion. This leads to the undesired, indiscriminate growth of a solid-electrolyte-interphase (SEI) and ultimately to a severe degradation of the cycle life and power capability (13).

Several strategies have been proposed to address these issues with μ-Si, including employing graphene cages (14), carbon nanotubes (CNTs) (15), polymeric binders (16–19) with high elasticity and self-healing properties, and electrolyte designs (20–22) that induce the formation of a lithium fluoride (LiF)–rich SEI. These strategies have largely improved the cycle life, particularly at low charging rates (<0.5 C). Yet, owing to the structural instability at high C-rates, fast-charging, one of the main advantageous capabilities of Si anodes, has not yet been satisfactorily demonstrated.

Restricting the range of the state of charge (SOC) of μ-Si could be a realistic solution (23, 24), but has not yet been intensively explored mainly because of the superficial belief that the partial use of the full capacity of μ-Si may demotivate the exclusive use of Si in the anode. Nonetheless, estimation of the capacity of a 18650 cell as a function of the anode capacity (Fig. 1A) indicates that an increase in the anode capacity does not linearly contribute to an increase in the total cell capacity; in fact, the effect of increasing the anode capacity is steep until ~1,300 mAh g⁻¹, whereas its effect weakens significantly thereafter. This nonlinear relation between the anode capacity and that of the total cell in the early anode capacity regime points to the possibility that the exclusive use of Si as an anode material could make a large impact even when a limited SOC range of Si was to be applied. As noted in Fig. 1A, for example, by taking an SOC range of merely 40% for μ-Si, which corresponds to 1,300 mAh g⁻¹, the total cell capacity increases by 19.1% compared to the anode composed exclusively of graphite. Notably, expanding the SOC range of Si to 100% would only increase the total cell capacity by an additional 4%. This indicates that the aforementioned nonlinear relation dictates the choice of the SOC range of Si for determining the total cell capacity and that this selection is of a highly sensitive nature. The nonlinear relation implies the use of μ-Si to be an effective strategy for increasing the energy density of a cell, on condition that a suitable SOC range is identified such that the volume expansion issues can be largely managed. The detailed calculations on which the plot in Fig. 1A is based are presented in SI Appendix, Table S1.

Fig. 1.

Characterization of μ-Si. (A) Total cell capacity of a 18650 cylindrical Li-ion cell as a function of the anode specific capacity. (B) Particle size distribution of μ-Si. (C) SEM image of μ-Si/CNT electrode. (D) SEM and EDS images of μ-Si particle. (E) Raman spectrum of μ-Si particle. (F) Initial charge-discharge profiles and initial Coulombic efficiencies of μ-Si/CB and μ-Si/CNT electrodes. (G) Half-cell cycling at 0.5 C for both charge and discharge at 25 °C.

Having noticed that an SOC range for μ-Si of merely 40% could increase the overall cell capacity substantially, we endeavored to identify electrochemical protocols that would permit a highly reversible SOC range of μ-Si. Our investigation, which focused on the electrode-level reaction heterogeneity and volume expansion at various SOCs, revealed an SOC range of 30 to 70% for μ-Si to be the most optimal for increasing the total cell capacity while warranting excellent rate capability and long-term cyclability. Based on our understanding of the effect of controlling the SOC range on the energy density and cycle life, we established a so-called anode-tailored full-cell design (ATFD) for LIB full cells containing an anode composed exclusively of μ-Si. The implementation of ATFD based on μ-Si||LiNi_0.9Mn_0.06Co_0.04O₂ (NCM) full cells demonstrated stable cyclability (84.6% retention after 500 cycles) even when imposing a 20-min limit for each charge at 3 C. Our work highlights μ-Si as a viable option for the anodes of upcoming “ultimate LIBs” once its SOC range is predetermined in consideration of the key physicochemical properties of the electrode at different degrees of lithiation.

Results

Preparation and Basic Analysis of the μ-Si Anode.

We used commercial, coating-free μ-Si with an average particle size of 4.6 μm (Fig. 1B). Scanning electron microscopy (SEM) (Fig. 1C) and energy-dispersive X-ray spectroscopy (EDS) mapping (Fig. 1D) confirmed the absence of coating material on μ-Si. Particle Raman analysis (Fig. 1E) showed a crystalline Si peak at 510 cm⁻¹ without any carbon peak. This μ-Si electrode was fabricated by introducing CNTs instead of carbon black (CB) to not only supplement the electrical conductivity of the electrode through a percolated network of CNTs, but also enhance the mechanical integrity of the electrode by interconnecting μ-Si particles (15). Details of the electrode and test conditions are presented in the Materials and Methods. Assisted by the CNTs, the specific capacity and ICE of the μ-Si/CNT electrode were superior to the corresponding properties of its carbon black counterpart according to their half-cell tests (Fig. 1F): 3461.3 vs. 3275.3 mAh g⁻¹ and 95.6% vs. 90.3%, respectively. The μ-Si/CNT electrode also displayed higher capacity retention after 50 cycles compared to that of the μ-Si/CB electrode when cycled at 0.5 C for both charge and discharge (Fig. 1G). In spite of the markedly improved performance by replacing CB with CNTs, stabilizing μ-Si cycling in the full SOC sweep still remained a formidable challenge.

Partial Cycling and Selection of SOC Range.

As mentioned above, cycling μ-Si only within a range of 40% of the entire SOC can be an effective approach to increase the total capacity of a cell while minimizing the drawbacks associated with the huge volume expansion of μ-Si. Confinement of the SOC range to 40% means that an infinite number of options are available depending on the specified minimum level of SOC; in principle, any SOC range of 40% could be selected between the lowest possible 0 to 40% and highest possible 60 to 100% ranges. To cover the entire range from 0% to 100%, in our analysis, we defined seven different ranges: R1 (0 to 40%), R2 (10 to 50%), R3 (20 to 60%), R4 (30 to 70%), R5 (40 to 80%), R6 (50 to 90%), and R7 (60 to 100%). The effect of the selected SOC range on the cyclability of the μ-Si/CNT electrodes was evaluated by establishing an electrochemical protocol (Fig. 2A). Sequentially, this test protocol consists of two formation cycles, 100 cycles in the 40% SOC range of choice (denoted as SOC-swing cycling), and delithiation to 1.5 V to evaluate the remaining amount of Li, followed by 100 full-range cycles in the voltage range of 0.05 to 1.5 V to assess the cyclability after the SOC-swing cycling. The tests with R1 to R7 indicate that Li loss steadily decreases from R1 to R4, but rebounds to increase from R4 to R7 (Fig. 2B and SI Appendix, Fig. S1). These findings point to R4 as the optimal condition under which to obtain the best cyclability with minimal overpotential increase (SI Appendix, Fig. S2 A–G) and voltage hysteresis (SI Appendix, Fig. S2H) during SOC-swing cycling. Moreover, R4 delivered superior cycling performance during full-range cycling in the last step (Fig. 2C), reconfirming the suitability of R4 for sustaining the cycling of the μ-Si/CNT anode. Although some works (23–26) reported enhanced cyclability of Si anodes by shallow charging, to the best of our knowledge, none of them have identified optimal SOC ranges that do not begin at SOC0.

Fig. 2.

Investigation of various 40% SOC ranges for μ-Si anodes. (A) Cycling protocol that was used to determine the optimal SOC range of 40%. (B) Percentage of lithium loss after 100 SOC-swing cycles run over different 40% SOC ranges. Error bars represent the SD for multiple (n = 4) cells. (C) Capacity retention during SOC-swing and full-range cycles. The charge capacity is normalized with respect to the 2nd discharge capacity during formation cycling. Cross-sectional SEM images of R1, R4, R6 electrode after (D) 10 and (E) 30 SOC-swing cycles. (F) Electrode thickness during SOC-swing cycling.

The superior reversibility of R4 was also reflected in the changes of the electrode thickness during SOC-swing cycling (Fig. 2 D and E). Consistent with the Li loss results in Fig. 2B, the most significant electrode swelling occurred for R6, followed by R1 and R4. For example, the electrodes in R1, R4, and R6 swelled by 58.0%, 33.9%, and 62.1% after 30 SOC-swing cycles as compared to their thickness of 25.0 μm after the formation cycles (Fig. 2F). Importantly, these series of data indicate that, for a fixed SOC range of 40%, the cycling stability can vary widely depending on the absolute starting and ending SOC.

This motivated us to unveil the underlying reason for this observation, particularly focusing on the homogeneity of the reaction at the electrode level. The inferior performance of μ-Si in certain SOC ranges appeared to be associated with nonuniform lithiation, which in turn leads to the formation of cracks. Indeed, in R1, cracks were predominantly apparent in particles in certain locations throughout the electrode (Fig. 3A and SI Appendix, Fig. S3), suggesting that the lithiation is largely confined to these cracked particles. R1 exhibited a wider range of particle degradation (red box for a more severely degraded particle and blue box for a less degraded particle), whereas the degradation of particles in R4 was less variant and severe (Fig. 3B). The variation of particle degradation was not very significant in R6. However, R6 suffered from more serious pulverization than R4 as evidenced by crack formation (Fig. 3C). To note, the degradation of a particle can be judged based on how clear the particle edge is. This uneven reaction distribution would accelerate unwanted SEI growth and would explain the more prominent electrode swelling in R1 over R4.

Fig. 3.

Fast kinetics and homogeneous particle reaction in R4. (A−C) SEM images of (Top) electrode cross-sections and (Bottom) individual μ-Si particles in (A) R1, (B) R4, and (C) R6 after 50 cycles. (D) Schematic illustration of the lithiation process of μ-Si along the degree of lithiation. (E) Schematic trends of active particle fraction of μ-Si and electrode swelling as a function of SOC. (F) Schematic illustration of lithiation behaviors and mechanical degradation of R1, R4, and R6.

The reaction heterogeneity over the active particles in the electrode is closely affected by the reaction kinetics. Consistent with previous studies (27), tests with the galvanostatic intermittent titration technique (GITT) (SI Appendix, Fig. S4) indicate that Li diffusivity initially increases with lithiation, but levels off toward the end of charge. These distinct kinetics during the course of lithiation prompted us to divide the Li-Si alloying process into three regimes, each characterized by its own phase transformation (Fig. 3D):

$$\mathrm{Si}\stackrel{K_1}{\to }{\mathrm{Li}}_{\mathrm{x}}\mathrm{Si}\stackrel{K_2}{\to }{\mathrm{Li}}_{\mathrm{x}+\mathrm{y}}\mathrm{Si}\stackrel{K_3}{\to }\mathrm{c}-{\mathrm{Li}}_{3.75}\mathrm{Si}.$$

According to the GITT results, K₂ is kinetically more favorable than K₁, presumably due to an increased number of Li migration pathways (28) and enhanced electronic conductivity (29). This interpretation is also in line with a previous report (30) that Li-Si alloys offer higher Li ion diffusivity than bare Si. Moreover, as supported by our GITT and half-cell electrochemical impedance spectroscopy (EIS) results (SI Appendix, Fig. S5), K₂ is also expected to be kinetically more favorable than K₃ because the latter step involves two-phase transformation from amorphous (Li_x+ySi) to the crystalline (c-Li_3.75Si) phase (30).

The superiority of R4 in terms of the cyclability and electrode swelling can be understood in consideration of the reaction heterogeneity and volume expansion of Si with respect to the SOC. From the viewpoint of volume expansion, a Si electrode is known (31, 32) to swell parabolically with increasing SOC (Fig. 3E), implying that utilizing a lower-level SOC range is beneficial for managing the volume expansion. However, as observed in the cross-sectional SEM image of R1 (Fig. 3A and SI Appendix, Fig. S3), cycling in the lower-level SOC range is largely dominated by the relatively sluggish K₁ process, which boosts the reaction heterogeneity among active particles and accelerates the capacity decay. This trade-off relation justifies the selection of R4 as our target SOC range for μ-Si (Fig. 3E) and explains its superior performance. To re-emphasize, R4 is the optimal SOC range with regard to both the reaction homogeneity and volume expansion (Fig. 3F); contrary to this, the lower-level SOC range induces the heterogeneous participation of active particles and in the higher-level SOC range, the active μ-Si particles undergo a more significant volume change via the two-phase transformation (33, 34) in K₃. In contrast, the intermediate-level SOC range offers a good compromise in that the dominant K₂ step with relatively fast kinetics enhances the reaction homogeneity of μ-Si while the volume expansion of Si is not extremely severe yet.

Anode-Tailored Full-Cell Design Strategy.

In conventional full-cell design (CFD), it is uncommon to use a specific, limited SOC range for the anode while using the full SOC range for the cathode. For this reason, in CFD the anode capacity is usually set to slightly exceed the cathode capacity, thereby maximizing the energy density of the cell while minimizing the risk of Li plating (35) (Fig. 4A). With this cell configuration, when the cathode operates in the full SOC range, the anode invariably operates from 0% to the specified SOC level, depending on the N/P ratio (the anode capacity over the cathode capacity) (Fig. 4C). With the prespecified N/P ratio of CFD, the use of an SOC of only 40% for the anode would necessarily sacrifice the energy density of the cell because the capacity of the cathode would only be partially utilized, which would discourage the use of high-capacity Si anodes.

Fig. 4.

Full-cell design strategy to utilize the target SOC range of the μ-Si anode. Schematic illustration of (A) conventional and (B) anode-tailored full-cell design. Schematic voltage profiles of cathode and anode in (C) conventional and (D) anode-tailored full-cell design. Schematic illustration of (E) graphite under CFD, (F) μ-Si under CFD, (G) μ-Si under ATFD before and after fast-charging cycles. The gray and yellow spheres represent graphite and μ-Si materials, respectively. Green sphere represents precharged μ-Si.

In this regard, the anode-tailored full-cell design (ATFD), the main cell configuration used in this study, would be a more appropriate approach when a partial SOC range is predetermined for the anode, whereas the full SOC range of the cathode is used (Fig. 4 B and D). Execution of the ATFD therefore makes it necessary to first determine the target SOC range of the anode that is aligned with the full SOC range of the cathode; in other words, the desired starting SOC of the anode would need to be attained by prelithiation.

These distinct cell configurations have a significant impact on both cycle life and rate capability. Graphite anode-based full cells employing CFD (Fig. 4E) are susceptible to Li plating during fast-charging, compromising battery safety and cycle life, aside from their inferior energy density. Full cells with μ-Si utilizing CFD (Fig. 4F) could mitigate the risk of Li plating due to the higher redox potential of μ-Si and the possibility of reducing the anode thickness. However, the severe volume changes of μ-Si may not warrant sufficient cycle life. In stark contrast, μ-Si anode-based full cells with ATFD (Fig. 4G) can realize robust cycling under fast-charging conditions by taking advantage of the rapid kinetics of the specified SOC of μ-Si and the reduced anode thickness, while maintaining competitive energy density levels.

Effect of μ-Si-Based ATFD on Energy Density and Fast-Charging.

To verify the feasibility of ATFD, we conducted a series of evaluations for anode-tailored full cells with an N/P ratio of 2.55 (Fig. 5A). This test also included conventional full cells with graphite and μ-Si anodes, both with an N/P ratio of 1.1 as control cells. All of these anodes were paired with NCM cathodes with the same areal capacity of 2 mAh cm⁻² (see the Materials and Methods and SI Appendix, Table S2 for detailed cell specifications).

Fig. 5.

Superior electrochemical performance of μ-Si in ATFD. (A) Initial charge-discharge voltage profiles of full cells at 0.1 C. The specific capacity is calculated by dividing the discharge capacity of the cell by the weight of the active material on the cathode. (B) Specific discharge capacity with respect to the mass of active material on the cathode as a function of cycle number when cycled under 3 C charge with 20 min cut-off and 1 C discharge. (C) Fast-charging capability test with the charging rate varied from 0.5 C to 12 C and the discharging rate fixed at 1 C, followed by recovery cycles at 1 C for both charge and discharge. (D) Retention of discharge capacity during 6 C charge–1 C discharge cycling and (E) corresponding charging voltage profiles of R4 and R6. (F) Long-term capacity retention of ATFD-based R4 cell under 3 C charge with 20 min cut-off and 1 C discharge.

Fig. 5A shows the specific capacity of the full-cell that was calculated by dividing the discharge capacity of the cell by the weight of the cathode active material. At 0.1 C (Fig. 5A), the CFD cells with graphite and μ-Si exhibited reversible capacities of 188.2 and 192.2 mAh g⁻¹ with ICEs of 86.1% and 85.4%, respectively. On the other hand, the anode-tailored R1, R4, and R6 full cells had specific capacities of 188.4, 213.9, and 211.3 mAh g⁻¹, respectively, along with ICEs of 83.5%, 90.1%, and 89.4%. Notably, the initial specific capacities of the R4 and R6 cells are comparable to that of the Li metal-based cell with an unlimited source of Li. All these results are summarized in SI Appendix, Fig. S6.

Although Si anodes have long been considered (36, 37) a key element for fast-charging LIBs, guaranteeing a robust cycle life while imposing fast-charging conditions is nontrivial. With this challenge in mind, we conducted a fast-charging test for various cells under the following cycling conditions: charging proceeded at 3 C with a 20 min cut-off and discharging proceeded at 1 C with 100% depth (Fig. 5B and SI Appendix, Fig. S7). The graphite cell gradually lost its capacity from the beginning and largely maintained it after the 40th cycle, whereas the Li metal half-cell initially delivered high specific capacities but ceased to operate at around the 130th cycle because of sudden failure. The capacity drop in both cells is attributed (38, 39) to well-known inhomogeneous Li plating and subsequent uncontrolled SEI formation during charging at a high current density, which consequently increases the cell resistance or gives rise to short-circuiting. The CFD-based μ-Si cell with the N/P ratio of 1.1 displayed a low initial specific capacity of 143.9 mAh g^–1 and poor cycle life (20.3% retention after 150 cycles) mainly due to the severe volume change of μ-Si over the broad SOC range of the anode. The anode-tailored R1 cell initially exhibited higher specific capacity of 164.6 mAh g^–1 as a result of the limited SOC sweep of the anode with the higher N/P ratio of 2.55, but suffered from inferior capacity retention (53.9% after 200 cycles) because of the aforementioned inhomogeneous reaction of μ-Si particles. In contrast, the anode-tailored R4 cell performed outstandingly with respect to both the specific capacity and cycle life; it delivered 191.4 mAh g^–1 in its first cycle at 1 C discharge and preserved 87.9% of its capacity after 250 cycles. The superior performance of the R4 cell was also reflected in its higher CEs throughout cycling (SI Appendix, Fig. S7B). These results attest to the effectiveness of our ATFD strategy, in which we employ a targeted SOC range for the anode to endow a LIB cell with fast-charging capability without sacrificing its energy density. The R6 cell also demonstrated robust cyclability of 92.6% after the same number of cycles, but delivered a lower capacity of 171.9 mAh g^–1 because of the larger volume change of μ-Si, which induces greater interfacial resistance at the SEI. R6 is additionally disadvantaged by its slower reaction kinetics of step K₃ compared to K₂, which dominates the reaction in R4.

Subsequent investigation of the fast-charging capability of R4 entailed varying the charging rate from 0.5 C to 12 C while using a discharging rate of 1 C (Fig. 5C). The graphite cell steadily maintained its capacity up to 3 C but experienced severe capacity loss at higher charging rates of 6 C and 12 C, leading to poor capacity recovery when returning to the charging rate of 1 C. Among all the cells tested, the R4 cell demonstrated the highest tolerance against high charging rates (see voltage profiles in SI Appendix, Fig. S8), thereby reconfirming the impact of its cell design on maintaining the interfacial stability and electrode integrity under harsh operating conditions. In addition to its fast-charging capability, the R4 cell exhibited excellent cycling stability even at an extreme fast-charging rate of 6 C with a 10-min cut-off (Fig. 5D). Under this charging condition, the discharging capacity of this cell at 1 C started with a decent value of 182.7 mAh g⁻¹ and retained the capacity of 163.2 mAh g⁻¹ after 150 cycles. By contrast, the R6 cell began with 164.6 mAh g⁻¹ and, after the same number of cycles, retained only 107.5 mAh g⁻¹. As shown in the charging profiles (Fig. 5E), severe polarization of R6 with cycling drove it to reach the cut-off potential of 4.2 V earlier to extend the CV period. In the same line, the R4 cell maintained the CC period more persistently for a longer cycling time owing to its smaller polarization toward much shorter CV periods. Once again, these results of R4 were made possible by its operation strategy involving a targeted SOC range of the anode, which enables homogeneous reaction by the μ-Si electrode together with an acceptable level of volume change. The CC-to-CV (equally, galvanostatic-to-potentiostatic) ratio, a direct indicator of the resistance of a cell against fast-charging, is plotted in SI Appendix, Fig. S9 for the R4 and R6 cells at different cycle numbers. The superior cyclability of R4 was observed in long-term cycling tests at 3 C charge and 1 C discharge, delivering a discharge capacity of 190.7 mAh g^–1 in its first cycle and maintaining 84.6% capacity after 500 cycles (Fig. 5F). The better cyclability of R4 was also evident at the mild C-rate of 1 C for both charge and discharge (SI Appendix, Fig. S10) and was reflected in the EIS measurements (SI Appendix, Fig. S11). Consistent with the charging profiles in Fig. 5E, the R4 cell exhibited the smallest and most persistent semicircles over the given SOC range of 40% during both charge and discharge, verifying its superior interfacial stability. It should be noted that the enhanced performance is not solely attributable to prelithiation. As indicated in Fig. 2 B and C, engaging the optimal SOC range also plays a critical role in achieving stable cycling. Simply 30% prelithiated µ-Si under CFD exhibited even more rapid capacity degradation due to Li plating under fast-charging conditions (SI Appendix, Fig. S12). Needle-like Li plating was observed on the µ-Si electrode surface after 30 cycles through SEM analysis (SI Appendix, Fig. S13). In this context, our current investigation deviates from established prelithiation studies, which persistently initiate charging at SOC0.

The effectiveness of our ATFD strategy was additionally validated by evaluating the amounts of Li remaining in the anode and cathode after the above cycling tests at a charging rate of 3 C. To this end, the cycled full cells were disassembled after a different number of cycles and reassembled into half-cells that were subsequently delithiated to determine the amounts of Li remaining in the individual electrodes (Fig. 6A and SI Appendix, Fig. S14). As the R1 cell was not subjected to prelithiation before cycling, the anode contained a negligible amount of Li after cycles that ended in discharge. However, the R1 cell lost the Li content of the cathode with cycling owing to the irreversibility related to the inhomogeneous reaction that consumed Li. In the case of R6, the cathode capacity was maintained to a greater extent with cycling, whereas the Li content in the anode decreased significantly, revealing that the Li source added by prelithiation was continuously consumed to compensate for the irreversibility arising from the excessive volume change of μ-Si. Contrary to this, the R4 cell largely maintained the Li contents of both the cathode and anode for 100 cycles. The total amount of Li remaining in both the anode and cathode as well as the amount of Li lost in both electrodes are plotted in Fig. 6B, which conspicuously assures the enhanced reversibility of R4.

Fig. 6.

Validation of ATFD strategy. (A and B) Quantification of Li amounts of anodes and cathodes in full cells after 3 C fast-charging cycles determined by reassembly experiments. (A) Areal delithiation capacity of the anodes and cathodes. (B) Sum of the Li source and Li loss of anodes and cathodes in full cells. (C and D) XPS analysis of anodes after 100 cycles under 3 C charge and 1 C discharge. (C) XPS F 1s spectra of μ-Si (CFD), R1, R4, and R6. (D) Relative proportions of chemical components according to the F 1s spectra. (E) Schematic illustration of three-electrode pouch cell. (F) Cycling performance and Coulombic efficiency of three-electrode pouch cells at 3 C charge–1 C discharge cycling in the voltage range of 2.7 to 4.2 V. Voltage profiles of the anodes, cathodes, and full cells of (G) CFD-based μ-Si and (H) ATFD-based R4 three-electrode pouch cells as a function of time.

The SEI composition is known to play a crucial role in determining the cycle life of Si anodes. X-ray photoelectron spectroscopy (XPS) was employed to analyze the compositions of the SEI layers on different anodes after 100 cycles under 3 C charge and 1 C discharge (Fig. 6 C and D). As shown in SI Appendix, Fig. S15, although the SEI of µ-Si in CFD consists of similar atomic contents to those of the other electrodes, the LiF content in the SEI of µ-Si in CFD is significantly lower (Fig. 6D). This is attributed to the large volume expansion of µ-Si in CFD, which continually decomposes the electrolyte and thus reduces the relative content of LiF. On the other hand, analogous peak positions and patterns observed in R1, R4, and R6 indicate that the SEI properties are primarily governed by the electrolyte composition rather than the SOC range employed during main cycling. From these XPS results, we concluded that the superior performance of R4 is mainly associated with its unique physicochemical properties in relation to the designated SOC range rather than the SEI composition.

To further validate the effectiveness of the ATFD strategy, we conducted three-electrode pouch cell tests at 3 C charge–1 C discharge for both CFD-based and ATFD-based R4 cells (Fig. 6E). Throughout cycling, R4 demonstrated superior capacity retention and higher Coulombic efficiencies (Fig. 6F). Analysis of the initial voltage profiles (SI Appendix, Fig. S16) revealed that the cathode operating ranges for CFD and R4 were 3.58 to 4.30 V and 3.39 to 4.30 V, respectively. The distinct bottom cut-off voltages of the cathodes originate from the lower operating potential of the prelithiated μ-Si under R4, enabling greater cathode utilization in R4 as compared to CFD in the designated full-cell voltage range. Moreover, as cycling progressed, the anode voltage in CFD shifted below 0 V (Fig. 6G), in reflection of a more rapid overpotential increase in the corresponding anode due to its more server degradation. In the full-cell voltage range of 2.7 to 4.2 V, this anode voltage shift lowered the cathode voltage, eventually dropping below 4.2 V after 20 cycles (SI Appendix, Fig. S17). Our analysis indicates that the unstable behavior of the anode during cycling is the main reason for the rapid capacity decay of the CFD-based full-cell. On the contrary, R4 exhibited far more stable voltage profiles for both its anode and cathode throughout the cycles (Fig. 6H). In a similar context, the end potential of the cathode of R4 during charge remained more stable compared to that of CFD (SI Appendix, Fig. S17).

The current research extends beyond simply improving the cycle life of µ-Si by reducing the SOC range and incorporating additional Li in its pristine state. A fundamental comprehension of μ-Si with a focus on its reaction kinetics and mechanical properties with respect to the degree of lithiation offers insights into unconventional full-cell designs that enhance energy density and fast-charging capability in a balanced manner (Fig. 7 A and B and SI Appendix, Fig. S18). Specifically, the R4 ATFD cell increases the gravimetric energy density compared to its graphite-based CFD cell counterpart by 151% during 0.1 C formation and by 213% after 200 cycles (Fig. 7C). By employing the SOC-confined operation strategy and exclusively using µ-Si on the anode, the ATFD cells in this study achieved superior performance in all of energy density, fast chargeability, and cycle life, compared to any prelithiated Si (40–47) and µ-Si reported (14, 15, 17–22, 24, 36) to date (Fig. 7 D and E). Although the SOC-specified operation requires prelithiation to begin charging at the designated SOC, prelithiation technology has recently made significant progress, particularly through the implementation of a cost-effective roll-to-roll scheme for scalability (41). In line with this progress, we demonstrated the uniform prelithiation of a precise amount for our μ-Si electrodes, covering an area of 3 × 3 cm², through physical contact with Li foil (SI Appendix, Fig. S19A). The electrochemically active amount of prelithiated Li indeed matched well with the thickness (6 μm) of the transferred Li foil (SI Appendix, Fig. S19B). Moreover, prelithiation can further be refined through a wet chemical process to tune the SEI properties. As a demonstration of such opportunity, we employed 4M lithium bis(fluorosulfonyl)imide (LiFSI) in dimethyl ether (DME) during pressurization in the prelithiation process and observed fluorine and oxygen-containing interfacial components (SI Appendix, Fig. S20). The effect of these modified interfacial characteristics will be detailed in a forthcoming paper. All in all, scalable prelithiation technology is anticipated to be applicable to other similar asymmetric cell configurations with superior electrochemical properties.

Fig. 7.

High-energy density and fast-chargeable μ-Si-based LIBs with ATFD. (A) Schematic representations of CFD configurations with μ-Si and graphite anodes, and ATFD configurations with R1 and R4. (B) Radar plot of electrochemical properties of CFD and ATFD systems. (C) Comparison of the practical energy density of different types of cells. Note that the ATFD with R4 takes the Li weight into account in its electrode mass. Detailed information is presented in SI Appendix, Table S3. Performance comparison of (D) prelithiated Si anode full cells and (E) μ-Si full cells with regard to charging time and cycle life. Detailed performance is presented in SI Appendix, Tables S4 and S5.

Conclusions

It remains technologically challenging to develop LIBs for which robust cycling is warranted under fast-charging conditions without sacrificing their energy density (48–50). In this regard, consensus has been gathering momentum on the potential of Si anodes, particularly μ-Si anodes, in consideration of their cost and manufacturing compatibility. Nevertheless, fast-chargeable LIBs with a μ-Si anode are not readily available because various drawbacks originating from the volume change of μ-Si have not yet been overcome to an acceptable level. We demonstrated the homogeneous (de)lithiation of a μ-Si anode with a relatively moderate volume change by engaging a limited, intermediate SOC range of μ-Si. The SOC-swing tests identified an SOC of 30 to 70% as the most optimal range for sustainable cycling, and this was confirmed by full-cell evaluations in which the μ-Si anodes were paired with state-of-the-art high-nickel cathodes. From the consideration of the materials alone (SI Appendix, Table S6), the μ-Si anode with an operating range of 30 to 70% SOC is estimated to offer a higher volumetric energy density than a lithium-metal battery with 40-µm-thick Li metal foil as the anode (SI Appendix, Fig. S21A). In a 18650-cell setting (SI Appendix, Table S1) (12), the given cell with the μ-Si anode and controlled SOC operation is estimated to increase the energy density by 21.3% compared to that of its graphite anode counterpart. In fact, this energy density is even close to that of a Li-metal battery with an N/P ratio of 1.1 (SI Appendix, Fig. S21B). Importantly, the 40% SOC sweep of μ-Si underlies this attractive position between the graphite anode and Li metal anode by taking advantage of the nonlinear relation between the anode capacity and total cell capacity (Fig. 1A). It is also noted that the demonstrated cycle life based on the ATFD can be further improved by employing functional binders (17–19) and engineered electrolytes (20–22). We anticipate that upcoming LIBs facing a growing demand for higher Si content in the anodes could benefit from SOC-controlled operation, even if their active materials are not solely composed of Si. Looking ahead, it is conceivable that tailored operational protocols could be devised for different alloy-based electrodes that suffer from volume expansion and consequential unstable interfacial properties, taking into account their own reaction behaviors along the SOC range.

Materials and Methods

Cell Materials and Fabrication.

The graphite anodes (MCMB, Osaka Gas) were fabricated by first preparing slurry by dispersing the active material, carbon black (Super P, Timcal), sodium carboxymethyl cellulose (Na-CMC, Nippon Paper), and styrene butadiene rubber (SBR, Zeon) in the ratio of 95:1:2:2 by weight. μ-Si (Alfa Aesar)/Super P slurry was prepared by dispersing the active material, CMC, SBR, and Super P in the ratio of 80:5:5:10 by weight. μ-Si/CNT (CNT, OCSiAl) slurry was prepared by the same procedure, but in the weight ratio of 80:7.5:12:0.5. The slurries were then coated on copper foil (18 μm) using the doctor blade technique, and were subsequently dried at 80 °C for 1 h. The dried electrodes were calendered to meet the designated electrode density (SI Appendix, Table S2) and further dried at 120 °C for 7 h in a vacuum oven. The cathodes, LiNi_0.9Mn_0.06Co_0.04O₂ (NCM, homemade) were fabricated by preparing a slurry by dispersing the active material, carbon black, and polyvinylidene fluoride (PVDF, Solvay) in N-methyl-2-pyrrolidone (NMP, JUNSEI) in the weight ratio of 92:4:4. The subsequent procedure for fabricating the cathodes was the same as for the anodes, with the exception that the slurry was cast onto aluminum foil (20 μm) and dried at 120 °C. The loading levels of the NCM cathodes were adjusted to achieve the target N/P ratios by taking the capacities of the paired anodes into account. Detailed electrode information about each full-cell is provided in SI Appendix, Table S2. Electrochemical tests were performed by preparing 2032-type coin cells. Each working electrode was punched into a disc with a diameter of 12 mm (NCM), 13 mm (graphite, μ-Si), or 15 mm (Li metal). 1.3 M LiPF₆ in a mixture of ethylene carbonate, ethyl methyl carbonate, and diethyl carbonate (3/5/2 vol%) with 10 wt% fluoroethylene carbonate, 0.5 wt% vinylene carbonate, 1 wt% propane sultone, 0.2 wt% lithium tetrafluoroborate (Enchem) was used as the electrolyte. The amount of the electrolyte injected was 5 μl mAh⁻¹. A microporous polypropylene film with a thickness of 15 μm (Celgard) was used as the separator. For the anode-tailored R4 and R6 full cells, the μ-Si electrodes were electrochemically prelithiated to the target capacity at 0.1 C in coin half-cells to demonstrate the feasibility of SOC-specified operation. These prelithiated electrodes were then assembled into the corresponding full cells. Prelithiation of μ-Si electrodes with 3 × 3 cm² and a μ-Si loading of 3.5 mg cm⁻² (SI Appendix, Figs. S19 and S20) was accomplished by direct contact with a 6 µm Li foil coated on a PET substrate, followed by pressurizing the assembly at 7.5 MPa and 80 °C for 6 min. After Li transfer, the PET substrate was removed. The wet chemical prelithiation was carried out using a similar approach, with the addition of the electrolyte comprising 4 M LiFSI in DME dropped onto the surface of the μ-Si electrode prior to pressurization. All coin cell assembly procedures were conducted in an argon-filled glove box (O₂ and H₂O < 1.0 ppm). Three-electrode pouch cells were assembled in a dry room with a dew point below –55 °C. As illustrated in Fig. 6E, the reference electrode was located between two separators. The same electrodes, electrolyte, and separators were used as those employed in the coin cell tests. For the anode-tailored R4 three-electrode pouch cells, the μ-Si electrodes were electrochemically prelithiated to the target capacity at 0.1 C in μ-Si||Li metal pouch cells before being assembled into the corresponding three-electrode pouch cells.

Electrochemical Measurements.

Electrochemical tests were conducted by charging in constant current–constant voltage (CC-CV) mode and discharging in CC mode at room temperature using a battery cycler (WBCS3000, WonATech). For the half-cell evaluation, two formation cycles were employed at 0.1 C for both charge and discharge, followed by cycling tests at 0.5 C for both charge and discharge in the voltage range of 0.01 to 1.5 V. The CV mode during charge was held until the current reached 0.05 C. For the full-cell evaluation, the same test procedure was adopted as for the half-cell evaluation except for the voltage range: 2.7 to 4.2 V for graphite and μ-Si full cells and 3.0 to 4.3 V for Li metal full cells. The fast-charging and rate capability tests were carried out in CC-CV mode for charging and the CV period was set to end by time cut-off; the duration of the entire CC-CV mode was set to 60/C-rate minutes. Discharging proceeded in CC mode at 1 C.

In the SOC-swing test (Fig. 2A), each electrode underwent two formation cycles after prelithiation to the designated SOC level. In subsequent SOC-swing cycles, the electrode was cycled at 0.5 C for 100 cycles. In each SOC-swing cycle, charging and discharging were conducted in CC-CV and CC mode, respectively. In the CV mode, the voltage was held until the capacity reached 40% of the capacity in the second formation cycle. Additionally, the CV step was also terminated (And condition) when the current reached 0.05 C. The CC discharging step was terminated when the voltage reached 1.5 V. After the SOC-swing cycles, the cell was discharged to 1.5 V to determine the remaining amount of Li. The Li loss was calculated as

$$\mathit{Li}\mathit{loss}\left(\mathit{mAh}\right)=Q_i-Q_f+{\sum }_1^{100}{N}^{\mathit{th}}\mathit{irreversible}\mathit{capacity},$$ [1]

where Q_i is the initial prelithiated capacity and Q_f is the discharge capacity with 1.5 V cut-off after the SOC-swing cycles. Subsequently, the cell was cycled at 0.1 C for three cycles, followed by cycling at 0.5 C for 100 cycles in the voltage range of 0.01 to 1.5 V. Cycling was carried out in CC mode for both charge and discharge. The GITT measurements were conducted after formation cycles. Each cell was galvanostatically scanned at 0.2 C in the voltage range of 0.01 to 1.5 V, during which each pulse was held for 10 min, followed by 4 h of relaxation. Electrochemical impedance spectroscopy (EIS) results were obtained in the frequency range of 1 MHz to 0.1 Hz with a voltage amplitude of 10 mV (SP-150, Biologic). The acquired spectra were fitted to the equivalent circuit model (SI Appendix, Fig. S11C) using the EC-lab Z-fit software.

Characterization.

The cross-sections of electrodes and the morphologies of particles were visualized using SEM (GeminiSEM-560, ZEISS). Elemental analysis was performed using energy dispersive spectroscopy (EDS, Aztec, Oxford) linked to SEM. For these analyses, the cycled cells were disassembled in a glove box and electrodes were washed with anhydrous dimethyl carbonate (DMC, Sigma-Aldrich). Samples for the cross-sectional analysis were prepared using an ion-milling system (IM400, Hitachi). The particle size distributions were obtained using a laser diffraction particle size analyzer (Microtrac S3500, Microtrac). The characteristics of chemical bonds were investigated by Raman spectroscopy (DXR3xi, Thermo Fisher Scientific). The SEI components were analyzed using XPS (NEXSA, Thermo Fisher Scientific), coupled with etching with a 2 kV argon ion beam for 15 s.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.