Exploring Classes of Co-Solvents for Fast-Charging Lithium-Ion Cells

Fast-charging lithium-ion cells require electrolyte solutions that balance high ionic conductivity and chemical stability. The intro- duction of an organic ester co-solvent is one route that can improve the rate capability of a cell. Several new co-solvent candidates were identiﬁed based on viscosity, permittivity (dielectric constant), and DFT-calculated electrochemical stability windows. Several formate,nitrile,ketone,andamideco-solventsareshowntoincreasetheionicconductivityoflithiumhexaﬂuorophosphateinconven-tionalorganic-carbonate-basedsolutions.BasedongasproductionduringtheﬁrstformationcycleinLi[Ni 1-x-y Co x Al y ]O 2 /graphite- SiO pouch cells, ﬁve candidates were identiﬁed: methyl formate (MF), ethyl formate (EF), propionitrile (PN), isobutyronitrile (iBN), and dimethyl formamide (DMF). High temperature storage (60 ◦ C), long-term cycling, and ultrahigh-precision coulometry results indicate that MF offers the greatest balance between conductivity increase and cell lifetime. Future work is encouraged to develop more stable solution chemistries that incorporate MF. PN may prove useful for low temperature ( < 40 ◦ C) applications. © The Author(s) access Creative Commons Attribution

To further progress the adoption of electric vehicles, it is desirable to develop fast-charging lithium-ion cells that are competitive with the convenience of fuel-powered vehicles, which can be refueled in just a few minutes. Yet even for modern designs, rapidly charging lithiumion cells can cause a decrease in cycle and calendar lifetimes. This is largely attributable to the unwanted electrodeposition of lithium metal (i.e., 'plating') on the graphite electrode surface when large overpotentials are applied to a cell. [1][2][3] Plating leads to a loss in cell lifetime and performance via multiple routes. First, some of the plated lithium is removed from the active lithium inventory of the cell, resulting in an irreversible capacity loss. 4 Second, a reactive interfacial region is created between the plated lithium and the electrolyte solution. Because the solution components are not chemically stable near the lithium electrode potential (i.e., Li/Li + ), they can quickly reduce to form gas by-products and new solid-electrolyte interphase (SEI) material. 5 This represents an additional loss of lithium inventory and can also increase the cell impedance. Finally, plated lithium can lead to dendrite formation, which may cause internal micro-shorts. 6,7 One approach to decrease the overpotential at the negative electrode is to improve the electrode design through improved electrical conductivity and smaller graphite particles to give a shorter characteristic solid-state diffusion time. [8][9][10][11][12][13] However, when negative electrodes become thicker, as they do in high energy density cells, lithium-ion transport in the electrolyte dominates the overpotential. Therefore this manuscript focuses on electrolytes with improved Li-ion transport. To accomplish this it is important to increase the ionic conductivity and lower the viscosity of the electrolyte solution. Depletion of Li + in the electrolyte solution near the negative electrode surface increases the overpotential at the negative electrode and can also contribute to lithium plating. [14][15][16] The introduction of an alkyl ester cosolvent has been shown to increase the ionic conductivity of lithium salts in organic-carbonate-based electrolyte solutions. 17,18 It has been demonstrated that ester-containing electrolyte solutions can improve the capacity retention following cycling at high C-rates in various lithium-ion cells, including Li[Ni 1-x-y Mn x Co y ]O 2 (NMC)/graphite and Li[Ni 1-x-y Co x Al y ]) 2 (NCA)/graphite. [18][19][20][21][22][23] Ester-containing ternary or quaternary solvent blends have also been explored for low temperature applications, where conductivity and transport properties are especially limited. [24][25][26][27][28][29] The simple addition of a new co-solvent is a practical route to improving the performance of cells because it requires little change to existing supply chain and manufacturing infrastructure. This approach is therefore very appealing to industrial lithium-ion cell producers and forms the focus of the present work. Finally, several other factors can affect lithium plating behaviors. 10,30 For example, biphenyl is an electrolyte additive used for overcharge protection and as a fire-retardant. 31, 32 Yet it has been observed that biphenyl polymerization can lead to 'dry-out' regions on the graphite electrode surface and that lithium deposition is enhanced at the boundaries of these regions. 33 It is therefore emphasized that this is a complex problem beyond the relatively simple electrolyte conductivity and stability discussion presented herein.
This work began with the question of how organic esters were chosen and whether new co-solvents could be identified that also increase conductivity without compromising the solution stability. A methodic approach is presented, beginning with a literature review of the physical properties of the various co-solvent candidates and the estimation of their electrochemical stability using density functional theory (DFT) calculations. Ultimately, a selection of new co-solvent candidates were tested and their potential utility for high rate capability cells is discussed.

Methods
Density functional theory calculations.-Density functional theory (DFT) calculations were performed using Gaussian (G09.d01). 34 Geometry optimizations and subsequent normal mode analyses were performed using the M062X hybrid functional, 35 the GD3 dispersion model, 36 the IEFPCM-UFF implicit solvation model (ε = 20), 37-39 and the 6-311++G(2df,2pd) basis set. The M062X double hybrid functional was chosen because it typically offers excellent accuracy for molecules with charge separation. 35,[40][41][42][43] The GD3 dispersion model was included because it has previously been shown to improve the accuracy of M062X calculations with minimal impact on calculation times. 36,[44][45][46][47][48][49] The implicit solvation model was selected to balance calculation speed and accuracy. 37 This is a very fast and easy-to-implement approach to handling the effects of the solvent. However, it is noted that this choice comes with an inherent loss of calculation accuracy. This is because implicit models cannot account for the influence that factors such as solvation sheath structure and preferential solvation in binary solvents can have on reaction barriers and transition state energies. [50][51][52][53][54][55][56] Moreover, the dielectric constant used in this work is an approximate value that does not consider the effects of the dissolved salt. 37,57,58 The method used here to determine standard electrode potentials (i.e., reduction and oxidation) has been described in detail    Table I. Values are shown for 25 • C or as noted.
previously 59 and follows the general formulae for the oxidation 1 and reduction 2 of a solvent, S: This simplistic approach is meant to give a general ranking of oxidative and reductive stability, rather than accurate values. The methods chosen for the present work were selected to balance calculation speed, computational requirements (i.e., CPU time), and accuracy. More rigorous (and more computationally demanding) approaches have been described previously and may be found elsewhere. [60][61][62][63] Optimized energies and molecular geometries are provided as supplemental material.
Electrolyte solutions.-All solutions were prepared in an argonatmosphere glove box and were mixed no more than 24 h before cell filling. Solutions were prepared using LiPF 6 (BASF, 99.94%, <14 ppm H 2 O) and a 25:5:70 solvent blend (by volume) of ethylene carbonate (EC), ethyl methyl carbonate (EMC) and dimethyl carbonate (DMC), received premixed from BASF (<20 ppm H 2 O). Co-solvents in this work, including suppliers, purities and abbrevia-tions, are summarized in Table I. First, the solvents were blended by mixing each co-solvent with the EC:EMC:DMC solution to reach the desired mass percentage (either 5, 20, 40, or 60% co-solvent). LiPF 6 salt was then added to prepare 1.2 mol L −1 electrolyte concentration. Finally, 2% by mass of vinylene carbonate (VC, BASF, 99.97%, <100 ppm H 2 O) was added to each electrolyte.
Conductivity measurements.-Ionic conductivity was measured using four conductivity probes performing independent measurements and connected to a commercial conductivity meter (Hach 3455). Each probe is equipped with an integrated PT1000 resistance temperature detector (RTD) to monitor the temperature of the electrolyte. Probes were calibrated in air and to a known standard (12.99 mS cm −1 , Hanna Instruments HI70030C). Approximately 14.5 mL of electrolyte was added to a custom-made stainless-steel holder under a fume hood. Each probe was sealed to its holder by a Viton O-ring and a custom-made stainless steel clamp to limit electrolyte-air contact. Sealed sensors were placed in a temperature-controlled bath (VWR Scientific model 1151) filled with ethylene glycol. The bath temperature was varied in increments of 5 • C over the range 0 • C-40 • C, evaluated using an external thermocouple thermometer (Omega HH802U) to be accurate within ± 0.5 • C. The electrolyte temperature was allowed to equilibrate with the bath temperature for ≥ 40 min. The accuracy of the ionic conductivity measurements are evaluated to be within ± 2%. Viscosity measurements.-Solution viscosity measurements were performed using an Ostwald viscometer (Sibata Scientific Technology, Japan). Two different capillary diameters were used, 0.50 mm and 0.75 mm, to measure solutions with higher and lower viscosities, respectively. The temperature was controlled at values within the range 10 • C-40 • C, by pumping a water/ethylene glycol mixture from a circulating temperature bath (Thermo Scientific) through a triple-walled Dewar that contained the viscometer. A resistance temperature detector (RTD) was attached to the surface of the viscometer to measure the temperature of the electrolyte. The measurements were completed using computer vision (CV) software developed by Beaulieu et al. The full details of this method, including the experimental setup and the data acquisition software have been described previously. 64 The raw data were interpolated to temperatures in 5 • C increments using a linear interpolation method in MATLAB (Mathworks, Inc.).

Lithium-ion cells.-Vacuum-sealed
LiNi x Co y Al 1-x-y O 2 /graphite-SiO (NCA) pouch cells with a capacity, C, of ∼280 mAh were received dry (no electrolyte). The cells were transferred to an argon-filled glove box, cut below the heat seal, dried under vacuum at 100 • C for 14 h to remove any residual water, and then returned to the glove box. Each cell was filled with 0.85 mL electrolyte (∼1.0 g) and then sealed at −90 kPa and 160 • C using a compact vacuum sealer (MSK-115A, MTI Corp.). Cells were quickly removed from the box and immediately held at 1.5 V at room temperature (21-25 • C) for ∼24 h to allow for cell wetting.
Electrochemical testing.-Immediately following the wetting period, cells were moved to temperature-controlled boxes (40.0 ± 0.1 • C) and connected to a Maccor 4000 Series automated test system (Maccor Inc.). The first charge-discharge cycle, herein referred to as the 'formation' cycle, was performed by applying a constant current of C/10 until the cells reached the upper voltage limit, 4.2 V. Cells were held at 4.2 V for 1 h and then discharged at C/10 until the cells reached the lower voltage limit, 2.8 V. Cells were then charged up to 3.8 V and held for an additional hour. Immediately following this formation procedure, cells were moved into a temperature controlled box at 10.0 ± 0.1 • C and connected to Biologic VMP3 multistat equipped with three frequency response analysis boards for electrochemical impedance spectroscopy (EIS) analysis (100 kHz-30 mHz, 10 mV amplitude). The sum of the charge-transfer impedances in the cell was evaluated by measuring the width of the depressed semi-circle, as reported previously, 65 and normalized by the geometric area of the positive electrode (55.3 cm 2 ). Long-term cycling tests were performed

Results and Discussion
Co-solvent selection.-The questions that prompted this work were: i) why were organic esters chosen as co-solvents for fastcharging lithium-ion cells and ii) are there any other co-solvents that may work just as well. With regards to the first part of the question, the esters methyl propionate (MP) and ethyl acetate (EA) are suitable co-solvents because of their a) low viscosity, b) high polarity, and c) wide electrochemical stability windows. 18 A literature review was therefore conducted to find potential co-solvent classes that meet these three criteria. The polarity of a co-solvent candidate was evaluated by the dielectric constant of the corresponding pure liquid at room temperature. The electrochemical stability windows were estimated by calculating the standard electrode potential of oxidation (Reaction 1) and reduction (Reaction 2) at room temperature. The findings are summarized in Table II and shown graphically in Figure 1. The flash points of the co-solvent candidates are included as a practical consideration that is relevant for industrial electrolyte component manufacturers and distributors.
Based on the literature review and the calculated stability windows, it was identified that organic formates, nitriles, ketones, and amides could meet the criteria for effective co-solvents described above. Formates and nitriles have been used previously for low temperature applications on account of their low viscosities and low freezing points. [107][108][109] This work measured the ionic conductivity of solutions containing MF, EF, nPF, iBF, PN, and iBN ( Figure 2). With regards to ketones and amides, the calculated stability windows are not as good as the esters, formates, or the nitriles. However, these are desirable because of their low costs and wide availability. Therefore one candidate was chosen from each category, MEK and DMF respectively. Several esters were also included for comparison: iBA, MP, and MB. Figure 2 shows that that PN gives the largest improvement to ionic conductivity of all the co-solvents tested. MEK and DMF give a similar magnitude of ionic conductivity improvement. Whereas MF, EF, and iBN provide only small conductivity increase near 40 • C, they offer more substantial improvements near 0 • C. nBF, iBF, MB, and iBA were observed to decrease the solution conductivity over the temperature range 0-40 • C. Based on these results, six new co-solvents were chosen for testing in NCA/graphite pouch cells: MF, EF, PN, iBN, MEK, and DMF. EA and iBF were also included for comparative purposes.
High-temperature storage.-NCA/graphite pouch cells were prepared with electrolyte solutions that contained 5-60% (by mass) of the nine co-solvents identified above. Solutions also contained 2% by weight VC, which is known to form a stable SEI in many cell chemistries. 73,75,110 It is important to note that electrolyte additives can play a significant role in the lifetime, gas evolution, impedance growth, etc. of cells prepared with various solvent mixtures. 24,29 The volume of gas produced during the formation cycle was used to screen co-solvent suitability for the purposes of this comparative study ( Figure 3). Because gas was measured following the full formation cycle, it is unknown whether this gas was produced at lower voltages (i.e., at the negative electrode), higher voltages (i.e., at the positive), or some combination thereof. 111 There is no simple linear relationship between co-solvent content and the volume of gas produced. MEK was determined to be unusable on account of substantial gas production and because cells would not undergo a successful formation cycle at very high co-solvent content. Therefore, no further testing was performed with MEK-containing solutions. It was also decided that ≥40% PN, ≥ 40% iBN, and 60% DMF would be removed from the study because of excessive gas production. iBF was included for comparative reasons, but because the iBF-containing solution had lower conductivity than the control electrolyte (see Figure 2), it is unsuitable for use in fast-charging applications. Therefore, the remainder of this work focuses on just six co-solvents: EA, MF, EF, PN, iBN, and DMF.
The remaining six co-solvent candidates were then evaluated by 500 h open circuit storage at high temperature (60 • C) and at high (4.2 V) or low (2.5 V) cell voltage. Broadly speaking, voltage drop and gas evolution during storage at low cell voltages represents the stability of the electrolyte solution at the negative electrode. Conversely, storage behavior at higher voltages is determined by the electrolyte stability at the charged positive electrode. [112][113][114] The formates MF and EF perform better than EA at high initial cell voltage in that the voltage drop increases only slightly as more co-solvent is added (Figure 4). However, the addition of formate co-solvents significantly increase gas production during high voltage storage ( Figure 5), reflecting gas evolution at the positive electrode, and increases the voltage drop at low voltage (Figure 4), likely indicating instability at the negative electrode. In contrast, 5-20% PN has good storage behavior at low cell voltage but displays a rapid voltage drop when stored at 4.2 V (Figure 4d). This suggests that PN is stable at the negative electrode but oxidizes, either electrochemically or indirectly, at high potentials at the positive electrode. However, unlike the formates, PN does not lead to substantial gas production ( Figure 5). Both iBN and DMF display extremely fast self-discharge behavior at high temperature and voltage, and were therefore excluded from further testing (Figures 4e-4f). Figure 6 shows that the addition of 5-20% MF, EF, PN, or iBN do not significantly affect the cell R ct , relative to control, following storage at either initial voltage tested in this work. This is in contrast to EA and DMF, both of which increase the cell impedance, even when only 5% co-solvent is used. For both formates, 40-60% co-solvent led to significant cell impedance growth following storage. These results support that 5-20% MF, EF, or PN may be viable co-solvent candidates, whereas DMF is an unsuitable co-solvent choice.
Cycling performance.-At this point, the pool of potential co-solvent candidates had been narrowed to three (other than EA): MF, EF, and PN. Although none of these perform as well as the control solution (i.e., cells prepared without any co-solvents), the conductivity boosts were considered beneficial and it was decided that these three co-solvents merited further testing by long-term ( Figure 7) and UHPC (Figure 8) cycling. EA was not further tested for practical reasons (i.e., availability of cells and testing channels) and because long-term and UHPC cycling data and long-term testing data has been reported previously for various ester-containing solutions and may be found elsewhere. [18][19][20][21][22][23] The introduction of 5% MF decreases cycling performance only slightly compared to the control cells. However, when added at higher concentrations, MF leads to a significant in-   crease in capacity fade ( Figure 7a) and V ( V is the difference between average charge and average discharge voltages, Figure 7d) indicating impedance growth in the cells. In long-term cycling, EF performed slightly worse than the corresponding MF-containing cells (Figures 7b, 7e). Curiously, UHPC results indicate a very similar CE for 5% MF and 5% EF solutions. However, UHPC results corroborate the observation that the addition of formate co-solvents decreases cell lifetime relative to the control cells. Finally, PN has the largest capacity fade rate, the steepest V slope, and the lowest CE of the three co-solvents. PN may therefore be excluded for future work at this temperature range. However, it may yet prove useful for low temperature cells, where ionic conductivity and salt diffusion are important factors. 115

Conclusions
A pool of co-solvent candidates for fast-charging lithium-ion cell electrolyte chemistries was identified based on the polarity (dielectric constant), viscosity, and electrochemical stability of the pure liquids. Several were found to indeed improve the ionic conductivity of lithium salt in organic-carbonate-based solutions. New co-solvent classes included formates, nitriles, ketones, and amides. High temperature storage was used to screen the five most promising candidates: MF, EF, PN, iBN, and DMF. The storage performance of the latter two was sufficiently poor that these were excluded from further evaluation. The three remaining co-solvents were tested by UHPC and long-term cycling: MF, EF, and PN. Of these three, MF offers the best balance of ionic conductivity improvements and the least decrease to storage performance and cycling lifetime. Nonetheless, future studies are encouraged to consider whether electrolyte additives may be refined to improve these metrics to more practical levels. Given the substantial conductivity boost associated with the introduction of PN, future work is also encouraged to develop this as a low temperature co-solvent. However, the present results suggest it is unsuitable for use in cells that will regularly see temperatures ≥ 40 • C.