Performance of Full Cells Containing Carbonate-Based LiFSI Electrolytes and Silicon-Graphite Negative Electrodes

The energy density of lithium-ion cells can be significantly increased by the use of silicon-containing negative electrodes. However, the long-term performance of these cells is limited by the stability of the silicon electrode-electrolyte interface, which is continually disrupted during electrochemical cycling. Therefore, the development of electrolyte systems that enhance the stability of this interface is a critical need. In this article, we examine the cycling of ∼20 mAh pouch cells with lithium bis(fluorosulfonyl)imide (LiFSI)-containing carbonate-based electrolytes, silicon-graphite negative electrodes, and Li1.03(Ni0.5Co0.2Mn0.3)0.97O2 based positive electrodes. The effect of fluoroethylene carbonate (FEC) and vinylene carbonate (VC) addition on cell performance is also examined and compared to the performance of our baseline LiPF6-containing cells. Our data show that cells containing only LiFSI show rapid loss of capacity, whereas additions of FEC and VC significantly improve cell capacity retention. Furthermore, the performance of LiFSI-FEC and LiPF6-FEC cells are very similar indicating that the electrolyte salts play a much smaller role in performance degradation than the electrolyte solvent. Future efforts to enhance longevity of cells with silicon-graphite negative electrodes will thereby focus on developing alternative solvent systems. © The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0981602jes] All rights reserved.

The ever-increasing demand for high energy density lithium-ion cells has led to a resurgence of interest in silicon-based negative electrodes. [1][2][3][4] Unlike conventional graphite-based negative electrodes which have a theoretical capacity of 372 mAh/g-graphite, siliconbased negative electrodes can deliver capacities up to 3579 mAh/gsilicon, thus making possible the development of thinner, highercapacity electrodes. However, the commercialization of silicon anodes has been limited by factors that include the following: (a) the relatively large amounts of conduction additives required to ensure good interparticle electric conductivity because of silicon's semi-conducting nature; (b) the large volume expansion/contraction resulting from silicon lithiation/delithiation processes leading to excessive cracking and delamination from the current collector; (c) the development of binder-systems that can maintain cohesion between the coating components and adhesion to the current collector during electrochemical cycling; (d) the excessive solid electrolyte interphase (SEI) formation on the electrodes that irreversibly trap lithium leading to rapid capacity fade. 1 In order to improve the performance of silicon-based negative electrodes, researchers have developed various strategies that include the following: (a) nanosizing the silicon particles, whereby the spaces between the particles can accommodate the volume expansion of individual particles; the use of nanoparticles, nanowires, and nanotubes have been shown to significantly improve cyclability; [5][6][7] (b) using non-traditional binders including water soluble polymers, such as polyacrylic acid (PAA), carboxymethyl cellulose (CMC), and alginates to improve mechanical integrity of the electrodes; 8,9 novel polymers that allow electronic conduction while maintaining electrode integrity during cell operation have also been described; 10 (c) improving electronic conductivity through use of carbon coatings applied by techniques such as physical/chemical vapor deposition and graphene wrapping; 11,12 (d) developing composite electrodes containing silicon and carbon constituents that limit electrode expansion during cycling. 13,14 These approaches are among the many strategies being actively pursued at Argonne National Laboratory as part of the U.S. DOE's Applied Battery Research (ABR) for Transportation program. The research is intended to accelerate the commercialization of silicon-based electrodes that are viable for use in batteries for vehicular applications.
An important focus area is the development of electrolyte systems that enhance the stability of the silicon electrode-electrolyte interface. This interface is continually disrupted during silicon expansion and contraction often exposing fresh surfaces for SEI formation that trap additional lithium. Electrode and cell lifetimes are strongly dependent on the electrolyte solvents, salts and additives. For example, cells with electrolytes containing LiPF 6 salt and conventional alkyl carbonate solvents display poor cycle life. However, the addition of fluoroethylene carbonate (FEC) and/or vinylene carbonate (VC) to these conventional systems cause remarkable improvements in cell cyclability. [15][16][17] Specifically, the addition of these compounds facilitates formation of a robust SEI that withstands the stresses and strains induced by expansion and contraction during electrochemical cycling. Several mechanisms have been proposed for the beneficial effects of these compounds. For VC these mechanisms include the creation of an opened VC 2− anion at low to intermediate levels of silicon lithiation, and a radical anion •OC 2 H 2 O 2 − at high levels of lithiation; further reaction and/or polymerization of these radicals results in crack-resistant SEI layers. 2 For FEC these mechanisms include (a) fluoride modification of the oxide over-layer covering the silicon particles; (b) radical polymerization of VC generated by reactions of electrolyte breakdown products with FEC; (c) enhanced deposition of LiF in the SEI layer; and (d) radical polymerization leading to formation of cross-linked elastomers that maintain electrode cohesion during cycling. 18 An electrolyte salt that has generated significant interest in recent years is lithium bis(fluorosulfonyl)imide Li[N(SO 2 F) 2 ], 19 henceforth referred to as LiFSI in this manuscript. This salt is reported to have higher ionic conductivity, better high temperature stability, and improved stability toward hydrolysis than LiPF 6 . 20 In addition, LiCoO 2 //Li and LiCoO 2 //graphite cells containing LiFSI are shown to outperform LiPF 6 -bearing cells. The performance of nanoSi//Li cells is also better with LiFSI than with LiPF 6 as shown recently by Phillipe and coworkers. 21 From depth-resolved Photoelectron Spectroscopy (PES) studies they concluded that, unlike LiPF 6 that forms SiO x F y by reaction with the SiO x surface oxide on silicon, LiFSI does not fluorinate the SiO x , thereby preserving favorable interactions between the binder and the active material surfaces. More recently, Piper et al. showed good reversible cycling of full cells containing nanosiliconbearing negative electrodes and LiFSI-bearing pyrrolidinium-based ionic liquid electrolytes. 3 Other researchers have indicated that the long-term cycling of lithium-metal cells is enabled in electrolytes composed of ether solvents and high LiFSI concentrations, which apparently suppress dendrite growth. 22 This dendrite suppression may be due to the unique properties of the SEI structure in FSI-based electrolytes. Shkrob et al. have indicated that reduction of the FSI − anion results in inorganic SEI phases without the concurrent generation of organic radicals and gaseous products. This mineralized SEI continually cracks and reforms during Li + deposition causing uniform growth on the Li metal surface instead of the metal whiskers that form in the SEI of conventional carbonate electrolytes. 23 The research literature contains numerous half-cell studies of silicon-based electrodes with small mass loadings; the cyclability of electrodes with high mass loadings in a full-cell configuration is rarely described. Therefore in this article, we report the performance of full cells with LiFSI-bearing carbonate-based electrolytes, silicon-graphite negative electrode (also referred to as Si-Gr) and Li 1.03 (Ni 0.5 Co 0.2 Mn 0.3 ) 0.97 O 2 -based positive electrodes (also referred to either as oxide or NCM523). The silicon-graphite electrode was developed at Argonne as a drop-in replacement for conventional graphite-based electrodes; it is physically robust, highly uniform, and very flexible even at capacity loadings exceeding 3 mAh/cm 2 . The NCM523 electrode and the LiFSI-based electrolytes were prepared at Argonne from commercially-available constituents. The performance of LiFSI-containing cells is compared to performance of our baseline LiPF 6 -containing cells. Such investigations are important steps toward developing electrolytes that enable high-performance, highenergy-dense, long-lasting silicon-containing batteries.

Experimental
Materials.-All electrodes in this study are from Argonne's Cell Analysis, Modeling, and Prototyping (CAMP) Facility; details on constituents and composition of these electrodes are shown in Table I. The positive electrode contains Li 1.03 (Ni 0.5 Co 0.2 Mn 0.3 ) 0.97 O 2 as the active constituent; the C45 carbon, added to enhance the electrode's electronic conduction, displays negligible electrochemical activity at voltages < 4.5 V vs. Li/Li + . In contrast, the C45 carbon is electrochemically active at potentials experienced by the negative electrode. 24 However, the negative electrode capacity is mainly determined by the graphite (73 wt%) and nanosilicon (15 wt%) content. The positive electrode binder is PVdF (polyvinylidene fluoride), which provides both coating cohesion and adhesion to the current collector. The negative electrode binder is partially-lithiated polyacrylic acid (Li-PAA), prepared by titrating polyacrylic acid (Sigma Aldrich, average M v ∼450,000) with LiOH (Sigma Aldrich). This Li-PAA binder enables the fabrication of mechanically robust electrodes, even at relatively high active-material loadings, as indicated earlier.
All electrolytes (see Table II) were prepared in-house in a glove box filled with nitrogen (< 1 ppm O 2 , H 2 O). To prepare the formulations, ethylene carbonate (EC) and ethyl methyl carbonate (EMC), both from BASF, were pre-mixed in 30/70 w/w% ratio. Then the salt, either LiFSI (Sarchem Laboratories) or LiPF 6 (Strem), was added to the solvent, the mixture was stirred 1 h at room temperature, filtered through 0.2 μm PTFE filter and the volume of the filtrate was adjusted  Pouch cell assembly and testing setup.-The full cell tests were conducted in single-layer xx3450-type pouch cells (see Figure 1) assembled in a dedicated climate-controlled dry-room with a dew point less than −42 • C (<100 ppm moisture). 24 These stack-type cells consist of two manually-aligned electrode layers (one single-sided positive and one single-sided negative electrode) separated by a single sheet of Celgard 2325 (PP/PE/PP) separator. The electrodes are dried overnight at 120 • C (positive) and 150 • C (negative) prior to cell assembly. The positive electrode has an average coating area of 14.1 cm 2 and the negative electrode has an average coating area of 14.9 cm 2 ; the larger negative electrode area reduces the probability of lithium plating during electrochemical cycling. An aluminum tab is welded to the positive and a nickel tab is welded to the negative electrode for connection to the external circuit; both tabs are 7 mm wide and 100 μm thick. The electrode stack is placed in a pouch container and a side area heat sealer is used to seal 3 sides of the pouch; a tab area heat sealer is used to prevent leaks at the tab locations. The assembly is then dried overnight at 60 • C.
The electrolyte of interest is added by pipette to the single open edge of the pouch in 0.5 mL aliquots. The pouch is then subjected to two 18 second vacuum and air-refill stages in a vacuum sealer to minimize air pockets and enhance electrode wetting by the electrolyte. After these soak and degassing steps, the final side of the pouch is heat-sealed while the cell is under vacuum. At this stage one side of the pouch contains an extra region that can fill with gases that may be generated during formation cycling.
For the electrochemical tests, the cells are fixed to rigid, nonconductive boards with female banana plugs in 2 rows of 2 cells per fixture layer. Standard stainless steel plates, xx3450-sized acrylic sheets,   Table III to obtain capacity retention and impedance rise information. The extra region of the pouch cells is sometimes inflated after formation because of gases generated during cycling. Such inflation was not observed in our cells probably because of the small electrode area-to-cell volume ratio of the single-layer pouch cell design. Therefore, the extra region was left intact and the cells proceeded with performance testing. The electrochemical cycling was conducted in the 3.0-4.1 V range. The upper cutoff voltage (UCV) was limited to 4.1 V to minimize the likelihood of Al current-collector pitting that is known to occur at full cell voltages > 4.2 V. The lower cutoff voltage (LCV) was set at 3.0 V because prior experiments showed that lower voltages shorten cell life.
Coin cell assembly and testing.-The full cell data were complemented by half-cell capacity data obtained in 2032-type coin cells (1.6 cm 2 area electrodes). These cells contained the Si-Gr electrode, Celgard 2325 separator, and a Li-metal counter electrode. Each electrolyte listed in Table II was examined mainly to determine features associated with solid electrolyte interphase (SEI) formation. These cells were cycled in the 1.5-0.0 V vs. Li/Li + range with a 0.085 mA/cm 2 current at room temperature.

Results and Discussion
As indicated previously, multiple cells containing identical chemistries were tested under the conditions shown in Table III. Negligible performance differences were seen for cells with identical chemistries. Therefore, only representative data are shown in this article to highlight aging trends and differences between various cell chemistries. that of the LiFSI cell around 2.4 V. The LiFSI-VC cell profile also shows a small deviation from that of the LiFSI cell around 2.8 V. The inset table shows that charge capacity of the LiFSI cells decreases slightly with VC and FEC addition. However, the discharge capacity and coulombic efficiency (CE) are marginally better for the LiFSI-VC cell. Note that the coulombic efficiencies for all cells are around 80%. Higher efficiencies, around 90%, are obtained when the LCV is lowered to 2.5 V; however, as indicated previously, cell lifetimes decrease when the LCV is reduced to 2.5 V. Figure 3 shows differential capacity profiles for the data shown in Figure 2 shapes and locations differ based on the electrolyte composition. For the LiFSI cell the peak maximum is at 3.05 V, whereas for the LiFSI-VC and LiFSI-FEC cells the maxima are at 2.85 V and at 2.44 V, respectively. The behavior of the LiPF 6 -FEC cell is distinctly different; after an initial increase, the dQ/dV values become negative around 2.64 V, and remains so for a brief period, before rising sharply and displaying a maxima at 2.62 V. This change in sign in the dQ/dV data directly results from the peak seen in the LiPF 6 -FEC initial charge data (inset, Figure 2), where the potential drops briefly before rising again. As such, it is a clear marker for this behavior, which can otherwise appear subtle in the voltage profiles themselves.
The full cell data contains contributions from both positive and negative electrodes. Because the locations, shapes, and electrolyte composition dependence of peaks in Figure 3 appear consistent with electrochemical reduction reactions we obtained half-cell cycling data on the Si-Gr electrodes. Figure 4 show differential capacity (dQ/dV) profiles calculated from the capacity-voltage data collected on these Si-Gr//Li cells. Only data during the first Si-Gr lithiation cycle is shown, which corresponds to the first charge cycle of the full cell. Figure 4 is divided into two panels for clarity; the upper panel shows data for the LiFSI and LiFSI-VC cells, and the lower panel shows data for the LiFSI-FEC and LiPF 6 -FEC cells. The data show that electrolyte reduction is initiated for the cells in the following order: LiFSI-FEC, LiPF 6 -FEC, LiFSI-VC and LiFSI, with peaks at ∼1.2 V vs. Li/Li + , at ∼1.04 V, at ∼0.74 V, and at ∼0.58 V, respectively. This trend is the same as that for the peaks observed in the full cell data (Figure 3). Some noteworthy observations from the data in Figure 4 are as follows: (i) the peak maxima for the cells containing VC and FEC are higher than that for the LiFSI cell without additives, which suggests that these compounds are reduced earlier than the carbonates in the LiFSI cell; (ii) the differences in peak maxima for the LiFSI-FEC and LiPF 6 -FEC suggest that the solvation structure of FEC varies with the electrolyte salt; (iii) the dQ/dV data for the LiPF 6 -FEC cell briefly changes sign, from negative to positive, similar to the sign change observed for the full cell dQ/dV data in the inset of Figure 3. This sign change could result from the breakdown of a resistive film formed during initial reduction reactions at the electrolyte-silicon interface; our experiments show that this sign-change is seen for all LiPF 6 -FEC cells containing negative electrodes with >5 wt% silicon, but is not observed for electrodes based solely on graphite; (iv) additional peaks at 0.85 V and 0.92 V are seen in the LiFSI-FEC and LiPF 6 -FEC data, which suggests further solvent reduction; (v) additional peaks (not shown), with intensities that are two orders of magnitude smaller than the main peaks, are seen in the 1.8-2.8 V range. These initial reduction reactions likely contribute to SEI structure on the electrodes, but are not explored here because SEI formation mechanisms are beyond the scope of this manuscript. Additional details from cycling the Si-Gr//Li cells in the 1.5-0.0 V range are shown in Table IV. It is evident that the irreversible capacity (lithiation minus delithiation) for cycle 1 is greater for the cells containing VC and FEC, which is consistent with greater lithium trapping during SEI formation in the presence of these additives. On the other hand, the irreversible capacity for cycle 2 is similar for all cells; in addition, the values are much smaller than those for cycle 1 indicating that the major SEI formation processes occur during the 1 st cycle, as expected. The coulombic efficiency, which is also a measure of lithium trapping, also increases for cells from ∼91-93% in cycle 1 to ∼99% in cycle 3 (data not shown). i.e., the efficiencies are lowest for LiFSI, but are in the same range for the other cells.
On further C/3 cycling, the capacity of all cells decline (see Figure 5a). The fastest and slowest capacity decline is observed for the LiFSI and LiFSI-VC cells, respectively. The LiFSI-FEC and LiPF 6 -FEC cells show an intermediate capacity decline; the capacities for these cells at the 100 th cycle are similar. Table V shows capacity retention at the 100 th cycle relative to the 1 st cycle. The retention decreases as follows: 61.4% (LiFSI-VC), 58.1% (LiFSI-FEC), 56.1% (LiPF 6 -FEC data), 18.7% (LiFSI), which is similar to the 3 rd cycle CE trend shown above. The capacity retention is consistent with the CE values (at C/3 rate) with cycle number (see Figure 5b). In all cases, the CE is highest initially, shows a steady decrease, then rises again, i.e., it shows a "hammock" trend.  shows an expanded view of the C.E. axis, and highlights differences between the cells. sistent with its rapid capacity decline; the sharp peaks in this dataset, which indicates a higher charge-to-discharge ratio probably results from occasional fracturing and reformation of the negative electrode SEI layer. Finally, for all cells, the CE for the 100 th cycle, which is at ∼C/20 rate, is lower than the CE for the prior C/3 cycles, and is likely related to the longer measurement times for the C/20 cycles compared to that for the C/3 cycles. The expansion/contraction of the Si-Gr electrode during lithiation/delithation results in breakdown/reformation of the SEI layer thereby increasing the amount of lithium immobilized by these SEIforming side-reactions. Electrolytes that yield SEIs with higher inorganic (more mineralized) content are expected to show a lower CE because of continual cracking and reformation of the SEI. On the other hand, electrolytes that form an elastomeric SEI, which is more resistant to cracking, will show higher CE values. This explanation is in agreement with the above CE data, which is lowest for the LiFSI (89.4% at C/20) electrolyte, which is expected to form a more mineralized SEI. The values are higher for the LiFSI-FEC (98.3% at C/20), LiPF 6 -FEC (98.4% at C/20) and LiFSI-VC (99.0% at C/20) cells because the VC and FEC compounds are expected to form cross-linked elastomers in the SEI during electrolyte reduction. 18 In addition to lithium-consuming side reactions in the negative electrode SEI, cell capacity fade can result from factors that include phase changes in the electrode active materials, isolation of active material particles through loss of electronic or ionic conductivity, and depletion of electrode active material through dissolution into the electrolyte. The NCM523 electrode is not expected to be a major contributor to capacity fade because it does not display any major phase changes, or dissolve to any significant extent, in the voltage range examined. In contrast, nanosilicon particle loss and a reduction in electronic connectivity are known contributors to capacity fade in the Si-Gr electrode, especially in cells that lack VC or FEC in the electrolyte. Taking all factors into consideration we surmise that capacity fade in our cells arises at the negative electrode. Capacity retention is enhanced by the electrolyte additives (VC, FEC) that improve coating cohesion and lessen lithium-consuming side reactions at the negative electrode.
Extended cycles -Impedance rise.-For vehicular applications, in addition to excellent capacity retention, battery cells need to sustain high current pulses under rapid discharge and charge conditions without degrading significantly. This ability is often determined by the hybrid pulse power characterization (HPPC) tests, in which cell impedance is determined over its useable charge and voltage range. 25 In a typical HPPC test, the cell is charged to its UCV (4.1 V, here), then subjected to repetitions of a pulse profile that contains constant-current discharge (3C, here) and charge (2.25C, here) pulses, followed by 10% depth of discharge (DOD) constant-current C/1 discharge segments, each followed by a 1h rest period.
The area specific impedance (ASI) data for the various pouch cells, obtained by using the HPPC protocol, is shown in Figure 6. ASI values at 3.5 V and 3.7 V, interpolated from the measured data, are listed in Table V. The data indicate that initial cell ASI's are quite similar, especially in the 3.4-4.0 V range. For example, at 3.7 V, the ASI values range from 29 (LiFSI) to 31 (LiFSI-VC); i.e., the values are the lowest for LiFSI and highest for the LiFSI-VC cell. Cell impedances are higher after the 100 charge-discharge cycles. We were unable to determine ASI values for the LiFSI cell because of its excessive capacity loss. For the other cells, however, the ASI values at 3.7 V are 42 (LiFSI-VC), 47 (LiFSI-FEC), 44 (LiPF 6 -FEC data). Although these data lie within a relatively narrow ASI band, Table V indicates the impedance increase is lowest for LiFSI-VC and highest for the LiFSI-FEC cell.
The causal mechanisms leading to impedance rise in these cells are yet to be explored. However, initial experiments in cells containing similar electrolytes and a Li-Sn reference electrode indicate that cell impedance rise mainly arises at the positive electrode; contribution of the negative electrode to impedance rise is, surprisingly, small. 26   The ASI data shown were obtained with a 3C discharge pulse, where C refers to the initial C/1 capacity of the cells.
that hinder the motion of lithium ions. These factors include the following: (a) microcrack development in the oxide particles; (b) crystal structure changes at the oxide-electrolyte interface; (c) decrease in ionic conductivity of electrode surface films; (d) degradation of electronic pathways between the carbon matrix and the oxide active material; and (e) deterioration of electrolyte transport properties within the pores of the electrode coating. 27 Prior experiments have suggested that the oxidation products of VC create a barrier layer that minimizes degradation of the positive electrode by the electrolyte. 28 On the other hand, FEC is highly resistant to oxidation 29 and is not likely to form a passivation layer at the positive electrode-electrolyte interface. This difference may explain the lower impedance rise for the LiFSI-VC cell. Further electrochemical and physicochemical experiments are needed to identify the appropriate mechanistic scenarios that lead to cell performance degradation. These experiments are currently in progress and their results will be detailed in future articles.

Conclusions
The electrochemical performance of cells containing Li 1.03 (Ni 0.5 Co 0.2 Mn 0.3 ) 0.97 O 2 -based positive electrodes, 15 wt% nanosilicon-bearing negative electrodes, and LiFSI-based electrolytes were examined in the 3.0-4.1 V cycling range. The resulting data were compared to cells containing an LiPF 6 -FEC electrolyte that is commonly used in silicon-bearing lithium-ion cells. The conclusions from our studies include the following: 1. The initial discharge capacity (139 mAh/g) of LiFSI cells is comparable to that of the LiFSI-VC and LiFSI-FEC cells. However, after 100 cycles, the LiFSI cells showed a relatively low capacity (26 mAh/g), when compared to the LiFSI-VC (86 mAh/g) and LiFSI-FEC (79 mAh/g) cells. 2. The LiFSI-VC cells displayed the best performance on cyclinglowest impedance rise and highest capacity retention -of all cells tested. 3. The performance of LiFSI-FEC and LiPF 6 -FEC cells is very similar. This result, in conjunction with data from the other cells, indicates that these salts play a much smaller role in performance degradation than the electrolyte solvent. Therefore, future efforts to enhance cell lifetimes will focus on designing alternative solvent systems.