Superior compatibility of silicon nanowire anodes in ionic liquid electrolytes

Electrochemical measurements

The electrochemical measurements were carried out on lithium metal cells using CR2032 coin cells. The lithium metal disk (500 μ thickness, 10 mm diameter) was purchased by Linyi Gelon LIB Co and used as the counter electrode. The a-Si and/or Sn-Si NW anodes were used as working electrodes. The electrodes were separated using two disks of glass fiber (16 mm diameter) separators (Whatman TM, Maidstone, UK). The cells were kept under vacuum for 30 min to allow complete loading of the IL electrolyte into the electrode and the separator.

A preliminary evaluation of the electrochemical processes taking place in Li/Si cells was run in a-Si anodes through cyclic voltammetric (CV) analysis paired with potentiostatic electrochemical impedance spectroscopy (PEIS) measurements. The following electrolyte formulations were used: 0.2LiTFSI:0.8EMITFSI, 0.2LiTFSI:0.8EMIFSI, and 0.2LiFSI:0.8EMIFSI. The CV tests were performed between 0.01 to 2 V vs. Li⁺/Li° at increasing scan rates, i.e., from 0.05 to 1 mV s^-1 (four consecutive CV cycles were run at each selected scan rate), followed by 500 cycles at constant scan rate of 1 mV s^-1. The PEIS was carried out on the fresh cells and at the end of each increasing scan rate CV family in the 10 kHz-0.1 Hz frequency range with 10 mV amplitude voltage.

Galvanostatic charge-discharge (GC) measurements were carried out on Li/Sn-Si NW cells using 0.2LiTFSI:0.8EMIFSI, 0.2LiFSI:0.8EMIFSI, and 0.2LiFSI:0.8N₁₁₁₄FSI as the electrolytes. The cells were cycled between 0.01 and 2 V vs. Li⁺/Li° at current rates from 0.1C to 10C.

All the electrochemical tests were carried out in a climate chamber at 20 °C using a Biologic (Seyssinet-Pariset, France) multichannel battery cycler.

Interfacial characterization

XPS was carried out to get information on electrochemical passivation layer (SEI) formation on the Sn-Si NW surface. XPS analysis was performed under an ultra-high vacuum (~10^-9 mbar) using a Kratos AXIS ULTRA spectrometer with a monochromatic Al Kα X-ray radiation source (hν = 1,486.6 eV). Pass energies of 160 and 20 eV were used for survey spectra and narrow regions, respectively. The C 1s line at 284.8 eV was used as a charge reference. These spectroscopic studies were performed on the Sn-Si NW surface after electrochemical tests at different states of charge in lithium salt-IL electrolytes: (i) pristine material; (ii) 1 cycle @0.1C; and (iii) 1 cycle @0.1C + 10 cycles @1C. After the cell dismantling, the Si electrodes were rinsed with Tetrahydrofuran (Sigma Aldrich, anhydrous, ≥ 99.9%) at least three times to remove the residual electrolyte, vacuum-dried at room temperature, and then transferred directly to the XPS analysis chamber.

Lithium intercalation process

The reversibility of the Li⁺ storage process was studied on a-Si anodes in the 0.2LiTFSI-0.8EMITFSI, 0.2LiTFSI-0.8EMIFSI, and 0.2LiFSI-0.8EMIFSI electrolytes through CV tests at different scan rates and for prolonged cycles. The results are displayed in Figure 1. The current value is normalized with respect to the Si electrode mass loading (A g^-1).

Figure 1. Cyclic voltammetry of Li/Si cells containing 0.2LiTFSI-0.8EMITFSI (A), 0.2LiTFSI-0.8EMIFSI (B), and 0.2LiFSI-0.8EMIFSI (C) electrolytes run at different scan rates and for prolonged cycles with a scan rate of 1 mV s^-1carried out on 0.2LiTFSI-0.8EMIFSI (D)- and 0.2LiFSI-0.8EMIFSI (E)-based cells. T = 20 °C.

A broad feature located around 1.2 V vs. Li⁺/Li was observed during the first cathodic scans in all three cases [Figure 1A-C], which relates to the passivation layer (SEI) formation onto the Si anodes^[23,33]. This peak (magnified in Supplementary Figure 1) disappears in the subsequent cycles, suggesting stabilization of the electrode/electrolyte interface. The Li/Si cell with 0.2LiTFSI-0.8EMITFSI shows the peaks relative to the lithiation of a-Si in the first cathodic scan, but no evident peaks can be noticed in the following cycles; this suggests non-optimal SEI formation (instead progressive lithiation of the amorphized Si anode)^[23,33], highlighting for large irreversible capacity and, consequently, low initial coulombic efficiency and indicating poor reversibility of the lithium dealloying. The progressive replacement of the TFSI anion with the FSI one in the IL [Figure 1B] and in both IL and lithium salt [Figure 1C] results in very beneficial effect on the cycling behavior. The lithiation of Si initially causes the formation of the amorphous phase Li_xSi, which turns into a crystalline phase below 0.06 V vs. Li⁺/Li^[24], corresponding to two well-defined peaks, located at 0.15 and 0.05 V vs. Li⁺/Li, as can be observed in Figure 1B and C. In the reverse scan, two similar features around 0.3 and 0.5 V vs. Li⁺/Li arise from the two-phase reaction from crystalline Li₁₅Si₄ to amorphous Li_xSi (x ≈ 2), followed by the formation of the delithiated amorphous Si^[24,34]. This indicates good reversibility of the delithiation process during the anodic scan, i.e., leading to much higher stored and delivered capacity, possibly due to an optimal SEI growth onto the Si electrode, given by the much better film-forming ability of the FSI anion with respect to the TFSI one^[23,24]. The complete overlap of the voltammograms, even at high scan rates and for prolonged cycles [Figure 1D and E], suggests good reversibility and high coulombic efficiency of the alloying/dealloying process during cycles in the 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8EMIFSI electrolytes. The increase of the scan rate results in progressive enhancement of the current flow through the cell, leading to more pronounced and defined current vs. voltage features in both verses, with only a moderate shift of the maximum value due to diffusive processes that take place into the electrolyte, according to the Randles-Sevcik equation^[35,36]. At 0.2 mV s^-1, two distinguished profiles are still observed, both in the cathodic (Li⁺ alloying) and anodic (Li⁺ dealloying) verse, whereas at faster scan rates, they are not split. Larger values of specific current associated with the electrochemical processes, resulting in higher capacity involved in the (de)lithiation process, were observed in the 0.2LiFSI-0.8EMIFSI cells with respect to the 0.2LiTFSI-0.8EMIFSI ones. Prolonged CVs were run only on 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8EMIFSI cells after CV tests displayed in Figure 1B and C to investigate the reversibility of the Li⁺ alloying process under hard operative conditions, i.e., 500 cycles at 1 mV s^-1. The results, as illustrated in Figure 1D and E, show two defined features around 0.25 and 0.1 V (cathodic scan) and around 0.6 V (anodic scan). A very good reversibility of the lithium alloying process is observed even after 500 cycles, witnessed by the good correspondence between the cathodic and anodic features. A remarkable reproducibility of the CV traces is recorded for the first 200 cycles, whereas progressive, even modest, decrease of the CV profile is observed up to the 500th cycle. To summarize, the 0.2LiFSI-0.8EMIFSI electrolyte exhibits comparable voltammetric features to those of 0.2LiTFSI-0.8EMIFSI; however, it shows evidence of higher capacity decay (i.e., more depleted CV profile) after prolonged CV tests (> 200 cycles).

The mobility of Li-ions at the electrolyte/silicon interface directly depends on the redox reaction rate according to the Randles-Sevcik equation^[35,36] and can be modulated by the SEI layer formed onto the silicon particles, which can enhance the rate performances and reduce the lithium trapping^[37]. Therefore, the Li-ion diffusion coefficient D_Li+ (cm² s^-1) was calculated for a-Si anode cycled into 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8EMIFSI electrolytes, evaluating the CV tests recorded at different scan rate [Figure 1B and C] described in the previous paragraph. The D_Li+ was calculated based on the Randles-Sevcik equation (Equation 1)^[38,39] for a semi-infinite diffusion of Li-ions:

where the I_p is the peak current, and R, F, and T are the gas constant, Faraday constant, and temperature (K), respectively. A (cm²) is the active surface area of the a-Si anode, n is the number of electrons in the reaction, C (mol cm^-3) is the Li₊-ion concentration in the electrolyte, and ν (V s^-1) is the scan rate. The relationship of the cathodic and anodic current peaks (I_c and I_a) with the square root of the scan rate is reported in Figure 2. The current peaks in anodic A_n and cathodic C_n scan, taken into consideration for the evaluation of D_Li+, are displayed in Supplementary Figure 2. The slope of the linear fit [Supplementary Figure 3] between I_p and ν^0.5 was used to estimate the D_Li+^[40-42]. Both systems show a linear behavior, meaning the diffusion-controlled reaction^[43] and a peak current ratio (I_pa/I_pc) ≈ 1, suggesting reversible lithiation process and no parasitic reaction associated with the electron transfer^[39]. The D_Li+ was found to be equal to (9.6 ± 0.5) × 10^-10 cm² s^-1 for 0.2LiTFSI-0.8EMIFSI and (2.5 ± 0.1) × 10^-10 cm² s^-1 for 0.2LiFSI-0.8EMIFSI electrolytes, comparable with the results reported in the literature for Li_xSi phases in organic electrolytes and obtained through different techniques^[44,45]. The error related to the D values is equal to 5%.

Figure 2. Linear fits for the anodic and cathodic peak currents versus square root of the scanning rates of a-Si electrodes using 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8EMIFSI electrolytes. T = 20 °C.

Impedance measurements

Impedance measurements were conducted after each CV family test run at different scan rates [Figure 1] and reported in Supplementary Figure 4 as Nyquist plots. The impedance responses display a semicircle pattern at medium-high frequencies and a straight line at low frequencies, in agreement with other studies on similar anode materials in IL electrolytes^[43,46]. The high-frequency intercept with the real axis corresponds to the bulk resistance (R_bulk) of the IL^[47-49], and the semicircle diameter gives the overall interfacial resistance R_int^[47-49].Figure 3 plots the dependence of R_bulk (A) and R_int (B), determined from the impedance responses [Supplementary Figure 4] taken [in open circuit voltage (OCV) conditions] after each CV family of Figure 1, as a function of the scan rate. The resistance values summarized in Table 3 were evaluated through the equivalent circuit models depicted in Supplementary Figure 5, where Q_dl is the constant phase element taking into account the double layer capacitance, and C_l is the electrode limit capacitance^[47-49]. R_bulk values equal to 1.7, 1.6, and 1.3 Ω cm were detected (at 20 °C) for the 0.2LiTFSI-0.8EMITFSI, 0.2LiTFSI-0.8EMIFSI, and 0.2LiFSI-0.8EMIFSI cells, respectively, i.e., close to those reported in previous work and summarized in Table 2^[32]. Significantly, no change of R_bulk is observed during the CV tests, confirming the good electrochemical stability of the IL-based electrolytes and the negligible lithium depletion from the electrolyte. However, the replacement of the TFSI anion with FSI results in a sharp decrease of R_int. For instance, more than 200 Ω cm² are recorded for the 0.2LiTFSI-0.8EMITFSI cells [Supplementary Figure 4A], whereas only 40 and 20 Ω cm² are approximately exhibited by those with 0.2LiTFSI-0.8EMIFSI [Supplementary Figure 4B] and 0.2LiFSI-0.8EMIFSI[Supplementary Figure 4C], respectively. These results, in very good agreement with the voltammetry data of Figure 1, indicate large decrease in interfacial resistance accompanying the transformation of the original discrete a-Si into a stable porous network (nano-ligaments)^[23,33]. Once more, it highlighted the beneficial effect of the FSI anion at the interface with the Si electrode, playing a key role in the SEI composition and, therefore, the active material morphology. For instance, a suitable SEI layer leads to a larger surface area offered for the alloying process by the a-Si network and to the accommodation of more facile charge transfer at the electrode/electrolyte interface. Conversely, the TFSI anion is not able to promote the growing-up of suitable SEI, leading to poor Li⁺ alloying and, therefore, low performance. This issue is in good agreement with both the CV results of Figure 1. The impedance response recorded after 500 cycles (shown in Supplementary Figure 4D and E) exhibits similar shape with respect to that of the Nyquist plots of the first family cycles [Supplementary Figure 4B and C]. No practical change of R_bulk is observed, whereas a more slightly depressed semicircle is exhibited by the 0.2LiFSI-0.8EMIFSI cells, nevertheless demonstrating the electrochemical reliability of the 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8EMIFSI electrolytes. An overall interfacial resistance lower than 100 Ω cm² is recorded in both electrolyte formulations. To summarize, the a-Si anodes seem to show comparable behavior in both 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8EMIFSI.

Figure 3. Evolution of the electrolyte R_bulk (A) and overall interfacial R_int (B) resistance during CV tests run at different scan rates for Li/a-Si cells in 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8EMIFSI. T = 20 °C.

Cycling behavior

The CV investigation on a-Si anodes has shown how EMIFSI is a more appropriate IL solvent compared to EMITFSI for Si electrodes. Therefore, this IL was used as the electrolyte component for further investigations on Sn-Si NW anodes. The N₁₁₁₄FSI IL was employed for comparison purposes. The performances in Li-metal cells of Sn-Si NW anodes were estimated by galvanostatic cycling measurements. Preliminary examinations [Supplementary Figure 6] have evidenced good and reproducible cycling behavior for a Si:Sn weight ratio above 2.6:1. As a consequence, only anode samples satisfying this parameter were subjected to cycling tests.

Effect of anion and cation

The voltage-capacity profile of the 1st charge-discharge cycle for Sn-Si NW electrodes in EMIFSI and N₁₁₁₄FSI electrolytes is reported in Figure 4A. A series of features, between 1.2 and 0.2 V, can be noticed during the first discharge curve at 0.1C, likely due to the IL electrolyte degradation leading to SEI layer formation onto the electrode surface^[23,33]. These features are followed by two plateaus located around 0.25 and 0.15 V, ascribable to the lithiation process. The value of the initial coulombic efficiency (η) and the nominal capacities (Q) delivered during the rate capability tests on Li/Sn-Si NW cells are reported in Table 4. The 0.2LiFSI-0.8EMIFSI cell shows the lowest initial irreversible capacity (Q_irr), amounting to 28%, with respect to 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8N₁₁₁₄FSI cells, which show Q_irr equal to 55% and 35%, respectively, but quickly leveling to 99% in all cases, as can be seen in Figure 4A. A stable capacity value is observed in all electrochemical formulations up to 100 cycles, even at very high rates (10C). Higher capacity values were delivered by the 0.2LiFSI-0.8EMIFSI cells at different scan rates, with a gain of ~550 mA h g^-1 up to 1C and of ~800 mA h g^-1 at higher current rates (> 2C) with respect to the other electrolyte systems investigated. The LiFSI-EMIFSI cell can deliver above 50% (1,300 mA h g^-1) of the nominal capacity at 10 C; meanwhile, the 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8N₁₁₁₄FSI cells deliver less than 29% and 16%, respectively, of their nominal capacity at the same current rate. The 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8N₁₁₁₄FSI electrolyte formulations show very similar capacity values which, at high rates (> 1C), slightly decay for the N₁₁₁₄FSI cells, likely due to the faster ion transport properties of the EMIFSI IL with respect to N₁₁₄FSI^[32]. The best performances of the full-FSI cell configuration, in agreement with the CV results, are attributable to the better film-forming ability of the FSI anion with respect to TFSI^[23,24]. Conversely, the N₁₁₁₄FSI IL displays lower compatibility towards Sn-Si NW anodes than EMIFSI. Supplementary Figure 7 displays the voltage vs. capacity profile of Sn-Si NW anodes in different electrolyte formulations at increasing scan rates. The voltage of Si alloying plateaus is distributed around 0.1 V, and the dealloying plateaus are at about 0.4 V, as reported in the literature^[34,50,51]. The presence of Sn in the working electrode is evidenced by the two less-defined plateaus in the potential range of 0.45-0.6V[Supplementary Figure 7], attributable to the two-phase alloying reaction that involved the formation of the dominant phases Li₂Sn₅ and LiSn^[34,51]. The increase of the current rate from 0.1C to 5C does not result in significant qualitative change of the voltage profile shape, but it does lead to an increase of lithiation/de-lithiation plateau slope attributed (and to an increase in ohmic drop) to diffusive phenomena within the IL electrolyte, especially at high rates. The rate capability tests [Figure 4B] were followed by prolonged tests run at 1C [Figure 4C] and have shown a capacity recovery up to the starting value at 1C. Very interesting capacity retention, despite a decay observed after 100 cycles, is generally exhibited by these electrolyte formulations; i.e., 1,005 (0.2LiFSI-0.8EMIFSI cells), 608 (0.2LiTFSI-0.8EMIFSI), and 509 (0.2LiFSI-0.8N₁₁₄FSI) mA h g^-1 are still delivered after 2,000 consecutive charge-discharge cycles (1C) at 100% of DoD, corresponding to 46% (capacity fading equal to 0.027 per cycle), 37% (0.031% per cycle), and 33% (0.034% per cycle) of the initial capacity discharged at 1C. The remarkably enhanced performances of the 0.2LiFSI-0.8EMIFSI electrolyte, comparable to or overcoming those observed in alkyl carbonate and other IL electrolytes^[23,33], are once more likely ascribed to the better film-forming ability of the FSI anion, in combination with the transport properties of the EMI cation. The reversible (charge) capacity vs. current density dependence of the Li/Sn-Si NW Sn cells [Figure 4D] evidences a lower rate region, where the capacity is limited by the diffusion phenomena taking place within the Si electrode^[52,53], distinguished by the higher rate one characterized by a more pronounced slope of the capacity vs. current density curve due to diffusive phenomena occurring in the IL electrolyte^[52,53]. The lower rate region limit is around 0.3 mA cm^-2 for the 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8N114FSI electrolytes: above this current rate, the electrochemical behavior of the Li/Si cells is governed by diffusive phenomena. Interestingly, the lower rate region extends up to 1 mA cm^-2 in 0.2LiFSI-0.8EMIFSI cells, indicating that diffusion processes become predominant in this electrolyte at much higher current density with respect to that observed in 0.2LiTFSI-0.8EMIFSI and 0.2LiFSI-0.8N₁₁₄FSI. To summarize, the 0.2LiFSI-0.8EMIFSI electrolyte exhibits, among the investigated IL formulations, the best cyclic behavior in Sn-Si NW anodes.

Figure 4. 1st discharge-charge curves (A), cycling performance at different rates (B and C), and reversible capacity vs. current density dependence (D) of Li/Si cells in 0.2LiTFSI-0.8EMIFSI, 0.2LiFSI-0.8EMIFSI, and 0.2LiFSI-0.8N₁₁₁₄FSI electrolytes. The cycling behavior of Li/Si cells in organic electrolytes (1M LiPF₆ in EC:DMC 1:1 v:v 3 wt.% FEC) is reported for comparison purposes. T = 20 °C.

Effect of FEC organic additive content

The Sn-Si NW anodes have also been studied in the presence of FEC organic additives to investigate a possible improvement of the electrolyte/electrode interface^[17]. The GC tests, as reported in Figure 5, were carried out in LiTFSI-EMIFSI electrolytes containing different FEC molar fractions (summarized in Table 1). The first discharge profile at 0.1C [Figure 5A] exhibits the features already discussed in the previous paragraph; however, the progressively increasing FEC content is seen to enhance the features located in the 1.7-0.2 V range and ascribable to irreversible electrolyte decomposition. In particular, at FEC content equal to 3 wt.%, a very wide plateau is observed around 0.4 V, leading to much larger irreversible capacity. The reversible nominal capacity, the initial coulombic efficiency (i.e., from 45% to 55%), and the cycling behavior (even at 10C) up to about 100 cycles are seen slightly increasing (i.e., from 2,000 to about 2,100 mA h g^-1) after incorporation of a FEC mole fraction equal to 0.01, which exhibits at 1C and 10C a delivered capacity of 1,800 and 750 mA h g^-1 (vs. 1,650 and 450 mA h g^-1 of the FEC-free cells), respectively. Further increases of the FEC up to 3 wt.% content lead to decreased performance. The nominal capacity and the coulombic efficiency fall down to 1,500 mA h g^-1 and 14%, respectively. Therefore, the presence of a FEC mole fraction > 0.01 in IL electrolytes seems to support larger consumption of electrolytes and/or hinder the formation of an optimal SEI onto the Sn-Si NW surface. As reported in the literature^[54,55], the uncoordinated FEC can passivate the anode surface by forming LiF, which can undergo the formation of HF in the presence of small amount of water, causing the dissolution of the SEI on the Si anode surface. However, more prolonged discharge-charge tests (C) reveal progressive decay in performance. For instance, after 1,000 cycles, the 0.01 FEC cells exhibit lower capacities (950 mA h g-1) than those delivered by the FEC-free ones (1,200 mA h g-1), even approaching the values (900 mA h g-1) shown by the 0.07 FEC and 3 wt.% FEC cells (i.e., no practical difference is seen for FEC mole fractions exceeding 0.07). This means an increase in capacity fading from 0.03% (FEC-free cells) to 0.05% per cycle.

Figure 5. 1st discharge-charge curves (A) and cycling performance at different rates (B and C) of Li/Sn-Si NW cells in LiTFSI-EMIFSI electrolytes with different FEC contents. T = 20 °C.

XPS analysis on Sn-Si NW anode surface

To get insight into the SEI layer composition formed on the Sn-Si NW anodes, ex-situ XPS analysis was carried out on pristine electrodes after the first charge-discharge cycle and several cycles (reported in Supplementary Figure 8). In order to avoid the binding energy (BE) shifts associated with charging phenomena and to gain good accuracy of the peak assignments, the XPS spectra were collected in the delithiated state. The atomic concentrations of the post-mortem Sn-Si NW anodes are summarized in Table 5 (the pristine electrode is reported for comparison). The SEI layer of the Si electrodes cycled in FEC-based and 0.2LiTFSI-0.8EMIFSI electrolytes exhibits high concentration of fluorine and a bit lower oxygen concentration in conjunction with higher silicon concentration onto the anode surface after several cycles, suggesting thinner passive layer onto a Si electrode, as previously reported in literature^[56,57]. Otherwise, the Si anodes in 0.2LiFSI-0.8EMIFSI and 0.2LiFSI-0.8N₁₁₁₄FSI show high sulfur concentrations, which depend on the degradation products of the LiFSI salt^[55].

High-resolution XPS spectra of C 1s, Si 2p, F 1s, Li 1s, S 2p, O 1s, and N 1s, collected for each sample after the first charge-discharge cycles, are displayed in Supplementary Figures 9 and 10, whereas those taken after several cycles are given in Figures 5 and 6. The Si 2p spectra of pristine Sn-Si NW anodes exhibit a narrow feature at 99.8 eV corresponding to the bulk silicon, and the presence of silicon oxides is observed at 100.5 and 102.5 eV^[56,58], also confirmed by the O 1s spectra at 532.4 eV^[56,58].

Figure 6. High-resolution C 1s, Si 2p, F 1s, and Li 1s core level spectra obtained on the Sn-Si NW anodes after consecutive lithiation/delithiation steps in different electrolyte formulations (see legend).

The N 1s XPS region of the Sn-Si NW [Figure 6] electrodes, cycled in IL electrolytes, displayed three well-separated peaks: the first two are related to the cation (-C-N-, N⁺), i.e., EMI⁺ (401.5 eV) and N⁺₁₁₁₄(402.6 eV)^[59,60], and the anion, which appears around 400 eV (C-N, N-SO) for both TFSI^- and FSI^-[55,61]. The presence of the IL signals is probably due to the caught IL in the porous structure of the Si anodes^[62]. The last peak at 398.5 eV is ascribable to Li₃N/LiC_xH_yN or -N=C- double bond in a polymeric network^[63-66]. These compounds are associated with the reduction of the electrolyte, which gave place to graphitic carbon^[62], evidenced in C 1s spectra [Figure 6] with the sp2 graphitic carbon feature at 284.8 eV. The relative atomic concentration increases upon cycles for the all-TFSI-based electrolyte, decreases for LiFSI-N₁₁₁₄FSI and LiFSI-EMIFSI + 3 wt.% FEC, and finally remains unchanged in LiFSI-EMIFSI, suggesting the growing-up of a stable Li₃N/LiC_xH_yN polymeric network during the first cycle.

The high-resolution XPS spectra of C 1s [Figure 6] display four main features centered around 284.8, 286.1, 288.5, and 289.5 eV in all cases, associated with C-C/C=C^[55,67], C-O/C-N^[64,68,69], O-C=O/C-F, and $$\mathrm{CO}_{3}^{2-}$$^[55,67,70] species, respectively. The C-C/C=C, C-O/C-N, and CF groups arise from the decomposition of IL (i.e., EMI⁺, N₁₁₁₄⁺, TFSI^-, and FSI^-). For Si-based electrodes cycled in LiTFSI-EMITFSI, the C 1s spectra show a very small peak at 293 eV associated with the CF₃ group^[59,61] derived by the TFSI anion, also confirmed by the well-pronounced peak at 688.6 eV in the F 1s XPS spectra [Figure 6]. The $$\mathrm{CO}_{3}^{2-}$$ can be assigned to Li-OCOOC₂H₅ and/or Li₂CO₃ arising from FEC degradation^[59] but also from salt decomposition, as was demonstrated in other works^[63,71]. The presence of this compound was also detected in the Li 1s spectra at 54.7 eV. This peak is more prominent in FEC-based electrolytes and in LiTFSI-EMITFSI. The area ratio of C_286.1/C_288.5 does not substantially change for LiFSI-EMIFSI and organic electrolytes during cycling, indicating the growth-up of stable SEI during the first cycle. The C 1s spectra of the Sn-Si NW cycled in organic electrolytes reveal quite similar features to those detected in ILs, as already reported in the previous paper^[62].

The deconvolution of O 1s spectra [Figure 7] highlights the presence of carbonate species, already observed in the C 1s spectra and evidenced around 532 eV^[55,72]. Lithium silicate species (i.e., Li_xSiO_y), related to the lithiation of Si oxide^[56,58,73], can be attributed to the peak at 531 eV, which are visible also in the Si 2p spectra at 100.5 eV^[55,72]. As described in detail elsewhere^[74-76], the broad peak at 533.6 eV, ascribable to O-C=O species, depends on solvent and/or additive polymerization, and it was also observed in the C 1s spectra at 286.1 eV. In all electrolytes, the Si surface is rich in carbonate species, as confirmed by the features in the C 1s and O 1s spectra [Figures 6 and 7].

Figure 7. High-resolution S 2p, O 1s, and N 1s core level spectra obtained on pristine Sn-Si NW anodes and after consecutive lithiation/delithiation steps in different electrolyte formulations (see legend).

A large LiF peak at 684.8 eV^[55,77] can be observed in the F 1s spectra, in agreement with the Li 1s spectra at 55.6 eV^[55,77], due to decomposition of salt (LiTFSI, LiFSI, and LiPF₆) by electrochemical reduction of S-F bonds^[58] and, during cycling, it can be observed that TFSI-based and FEC-based electrolytes exhibit a higher amount of LiF. As described by Piper et al., FSI^- and TFSI^- are subject to a quite different reduction process, where the FSI^- rapidly releases F^-, forming LiF and SO₂, which suggests the formation of SEI rich in small inorganic compounds^[60]. Conversely, TFSI^- can form different products, such as -SO₂CF₃. The -SO₂ species can also be observed at 168.8 eV in the S 2p spectra, where results are more prominent in the FSI-based systems. The fast release of F^- and SO₂ may be correlated to the high cycling performance shown by Si-Si NW anodes in LiFSI-EMIFSI electrolytes. This speculation derived from modeling studies performed on the FEC system^[78], demonstrating its ability to rapidly release F^- (to form LiF) during its decomposition. The SiOF signal at 686.8 eV corresponds to the Si substrate, indicating the thickness of the SEI layer: in organic electrolytes, it might be covered by the LiPF₆ signal, resulting from incomplete removal during the electrode rinsing and/or from intermediate decomposition products with similar BE to LiPF₆^[76,79]. This feature is more pronounced in LiFSI-EMIFSI and organic (1M LiPF₆ in EC:DMC 1:1 v:v + 3 wt.% FEC) electrolytes, indicating thinner SEI on the Si surface. The presence of residual lithium salts (LiTFSI, LiFSI, and LiPF₆) and/or their incomplete decomposition products is detected around 688 eV^[58,77]. In the case of LiTFSI, it is completely overlapped by the CF₃ signal at 688.6 eV, and the SiOF is not visible.

The S 2p deconvolution [Figure 7] results in a pretty complex spectrum for the Si anode cycled in IL-based electrolytes. It can be divided into three main components. The first one was represented by the Sp_3/2 peak, which confirms the presence of residual LiTFSI and/or LiFSI salt at 168.8 and 169.1 eV^[77], respectively. However, the second one encloses several features, in the range between 169.8 and 168 eV, due to degradation products of imide-based salts that might be assigned to S=O bonds^[77,80]. The last weak component, at around 167 eV, is also related to the reduction of Li-imide salt, but this process can also be induced by the X-ray beam^[55,76]. The s 2p spectra display very weak peaks, probably due to residual traces of electrolytes.

The Si 2p spectra show the peak related to the bulk silicon at ~98 and ~99 eV^[55,56], which, in the case of LiFSI-EMIFSI, tends to increase, meaning a reduction of the SEI layer thickness and in agreement with the results from impedance measurements discussed in Paragraph 3.2. A signal at 100.5 eV can be attributed to Li_xSiO_y, as already previously described, and the last peak at 102.8 eV is associated with the SiO₂ species.

The XPS results suggest a bilayer SEI onto the Si surface, i.e., an outer part composed of organic compounds within a polymeric network that can provide flexibility and good mechanical stability against the volume changes during the Si cycling, and an inner layer dominated by inorganic compounds such as Li₃N and LiF, delivering high Li-ion conductivity, and a graphitic network that provides electronic conductivity^[62,63,71]. The different cycling performances can be due to the different densities and porosity of the SEI on the Si surface. The SEI layer onto the Si anodes cycled in LiFSI-EMIFSI and organic electrolytes may be rich in stable LiF and -Si-F compounds which, thanks to their high bonding energy, do not decompose during cycling and promote stable interface with the electrolyte^[81,82]. Although, in the other case, the surface layer can contain less stable compounds, such as metastable and less dense, linear alkyl carbonates [-Si-OCH₂CH₂OCO₂Li, -Si-CH₂CH₂OCO₂Li, R(OCO₂Li)₂] and, due to their low bonding energy, can continue to decompose upon cycling, thus feeding irreversible redox reactions that promote silicon dendritization and pulverization, as has been shown by Bongiorno et al. The low density of the SEI layer allows the transition of the electrolyte during the Li⁺ insertion, which causes the breakup of the -Si-Si- network bond, giving rise to porous structures and cracking^[83].