Effects of Electrolyte Additives and Solvents on Unwanted Lithium Plating in Lithium-Ion Cells

Unwanted lithium plating on the graphite anode of lithium ion batteries can reduce the cycle life and safety of lithium ion batteries. Increased charging rates, lower temperatures, thicker electrodes, lower Li-ion diffusion constant and larger graphite particles all increase the propensity for unwanted lithium plating. In this work, a variety of electrolyte additives and electrolytes, which extend lifetime during low rate cycling, were used in Li[Ni 1/3 Mn 1/3 Co 1/3 ]O 2 /graphite (NMC111/graphite) pouch cells subjected to high rate chargingat20 ◦ C.Itwasfoundthatadditivesandelectrolyteswhichincreasedthenegativeelectrodearea-speciﬁc resistance,R negative , decreasedtheonsetcurrent,I u ,forunwantedlithiumplating.Here,theprocessesofiondesolvation,electronandiontransportthrough the solid electrolyte interphase and contact resistance are lumped into the R negative . Under conditions where R negative is the dominant factor determining when unwanted Li plating occurs, the onset current for lithium plating could be well predicted by the expression: I u = 0 . 080VxS / R negative , where S is the geometric electrode surface area. R negative is easily determined using negative electrode coin-type symmetric cells. This simple rule-of-thumb relation will help guide researchers

by VC on the graphite negative electrode. SEI film modifiers with less resistive SEI-films such as FEC 22 and allyl sulfide (AS) 23 have been reported to reduce the propensity for unwanted lithium plating. Therefore, the effects of different additives on unwanted lithium plating are very important to explore when developing electrolyte additives for commercial applications.
Co-solvents can improve the ionic conductivity of electrolytes at low temperatures. Ternary or quaternary carbonate-based electrolytes with low EC-content and co-solvents having low melting points and low viscosities have exhibited improved low temperature performance. Among these co-solvents, optimized proportions of the ester-based components such as methyl propanoate (MP), ethyl propionate (EP) and methyl acetate (MA) have been reported to provide good low temperature performance including better charge rate capability. 24,25 In cases where electrodes are thick or highly compressed, the addition of such co-solvents can also reduce the propensity for unwanted lithium plating at high charge rates even at higher temperatures.

Pouch cells.-The pouch cells used in this study were
Li[Ni 1/3 Mn 1/3 Co 1/3 ]O 2 (NMC111)/graphite cells with a capacity of 220 mAh balanced for 4.4 V operation. Cells were manufactured by Li-Fun Technology (Xinma Industry Zone, Golden Dragon Road, Tianyuan District, Zhuzhou City, Hunan Province, PRC, 412000) and vacuum sealed without electrolyte before shipping to our laboratory in Canada. The pouch cells are 40 mm long x 20 mm wide x 3.5 mm thick. The separators used in the cells were polyethylene microporous films 16 micrometers thick coated with 2 micrometers of ceramic particles on each side. The porosity of the separators was 41% and the Gurley air flow time was 181 s/100 mL of air using the ISO5636-5:2003 test standard (which has since been replaced by ISO 5636-5:2013). The pouch cells incorporated a rolled electrode design with 10 turns in the jelly roll. Table I summarizes detailed information about the positive and negative electrode materials in these NMC111/graphite pouch cells. The negative-to-positive capacity ratio at 4.1 V is about 1.17. All cells were cut open and vacuum dried at 100 • C for 14 h before electrolyte filling. Then cells were transferred immediately to an argon-filled glove box for electrolyte filling and vacuum sealing. The pouch cells were filled with 0.76 mL (about 0.90 g) of electrolyte. After filling, cells were vacuum-sealed with a compact vacuum sealer (MSK-115A, MTI Corp.).
The formation process was as follows: The pouch cells were first held at 1.5 V for 24 hours at room temperature to ensure full wetting of the electrode stack. Then cells were charged to 3.5 V with a current of C/20 and held at 3.5 V for 1 h (all at 40 • C). Then cells were degassed and vacuum sealed again in an argon-filled glove box. After this step, cells were charged to 4.1 V and discharged to 3.8 V at C/20 and 40 • C, held at 3.8 V for 1 hour and were then ready for electrochemical impedance spectroscopy (EIS) measurements.
Electrochemical impedance spectroscopy (EIS).-EIS measurements were conducted on NMC111/graphite pouch cells before and after cell cycling. Prior to EIS measurement, cells were charged or discharged to 3.8 V and held for 1 h to stabilize cell voltage. The potential of 3.8 V was chosen for the EIS measurements because both graphite and NMC electrodes are at about 50% state of charge at this voltage. EIS spectra were measured at temperatures of 10. ± 0.1 • C or 20. ± 0.1 • C using a Biologic VMP3 electrochemical test station. Data were collected with ten points per decade from 100 kHz to 10 mHz with a signal amplitude of 10 mV. 20 • C cycling.-After EIS measurements, the cells were charged and discharged using different constant currents (C-rates) of 75 mA (C/3), 110 mA (C/2), 220 mA (1 C) and 330 mA (1.5 C) between 2.8 and 4.1 V using a Maccor charger system at 20. ± 0.1 • C. There was no constant voltage segment at the top of charge. Pair cells were tested for every charge rate to make sure the data was reproducible. In order to determine the active lithium loss during cycling, cells were cycled at C/20 (constant current charge and discharge) one time between each of the high charge rate segments. The upper cutoff voltage was set to 4.1 V in order to avoid electrolyte oxidation at the positive electrode and to ensure that the cells were far from having a fully lithiated negative electrode which would occur at 4.4 V for these cells. All pouch cells were cycled with external clamps to maintain pressure on the electrode stack (about 0.5 atm. pressure) and push small amounts of gas that may be produced during cycling to the edge of the pouch. Cells were stopped after about 350 hours cycling or after the capacity loss reached 20%.
Symmetric cells.-Symmetric cells were used to separate the contributions of the impedance of the negative and positive electrodes to the impedance of the full cells. 12,26 Symmetric cells were constructed from electrodes after dissembling some of these pouch cells. The detailed procedures for symmetric cell construction have been described by Petibon et al. 26 The pouch cells were charged or discharged to 3.8 V and held at 3.8 V for one hour (approx. 50% state of charge) before they were cut open in an argon-filled glove box. Eight coin-cell size (1.54 cm 2 ) positive electrodes and negative electrodes were cut respectively from the pouch cell electrodes with a precision punch. Three negative/negative symmetric coin cells, three positive/positive symmetric coin cells and two positive/negative coin cells were reassembled using one polypropylene blown microfiber separator (BMFavailable from 3 M Co., 0.275 mm thickness, 3.2 mg/cm 2 ). The electrolyte used for the symmetric cells was the same as that used in the parent pouch cell.
Electrolyte conductivity measurements.-The ionic conductivity of electrolytes with different proportions of methyl propanoate (MP) was measured using a Mettler Toledo FG3 conductivity meter. Before electrolyte conductivity measurements, the conductivity probe was calibrated using a conductivity standard of 12.88 mS/cm at 25 • C. The conductivity of the electrolyte was measured every 10 to 20 • C between −20 to +55 • C. Figure 1a shows the chemical structure of the additives used. Figures 1b, 1c and 1d show the differential capacity (dQ/dV) vs. cell voltage of NMC111/graphite pouch cells filled with control electrolyte and with control electrolyte in combination with 1% -3%TAP, xPES211 (x = 0.25, 0.5, 0.75 and 1), 2% VC, 2% VC + 1% TMSMS or 2% VC + 1% ES during formation (first charge to 3.5 V). The dQ/dV plot indicates the potential(s) at which the solvents or additives reduce on the graphite anode. 0.85 V ± 0.1 V vs. Li/Li + ) in addition to the reduction peak of EC at 2.90 V, indicating that the peak at 2.70 V corresponds to the reduction of TAP. With increasing content of TAP, the capacity associated with the reduction of TAP increased and the reduction of TAP partially suppressed the reduction of EC. Figure 1c shows that with the addition of xPES211, a new peak appeared at around 2.40 V (graphite potential around 1.15 V ± 0.1 V vs. Li/Li + ) corresponding to the preferential reduction of PES. 27,28 Similar to TAP, as the amount of xPES211 increased, the capacity associated with the reduction of PES211 increased. When x ≥ 0.5 in xPES211, xPES211 passivated the graphite effectively and the reduction of EC (2.90 V) was almost fully suppressed. Figure 1d shows that the reduction of 2% VC occurred at a cell voltage around 2.7 V which fully suppressed the reduction of EC. Cells with 2% VC + 1% TMSMS show a reduction peak at around 2.8 V, while adding TMSMS alone to control electrolyte showed no extra reduction peaks apart from the reduction of EC on the graphite electrode. Therefore, the reduction peak of 2% VC + 1% TMSMS at 2.8 V is probably related to the reduction of VC. Figure 1d shows that reduction peak in cells with 2% VC + 1% ES at about 2.5 V is different from the reduction of VC alone. This peak is caused by the reduction of ES or by the combined reduction of ES and VC. 16,29 Figure 2 shows the Nyquist plots for NMC111/graphite pouch cells with different electrolyte additives measured at 3.80 V and 10 • C after cell formation. Figures 2a and 2b show that the addition of TAP and xPES211 (0 ≤ x ≤ 1) increased cell impedance (diameter of "semicircle" increases) and this impedance increased with increasing content of TAP or xPES211 (0 ≤ x ≤ 1). This result correlates well with the dQ/dV vs. V results shown in Figures 1b and 1c. Figures  1b and 1c show that when more TAP or xPES211 were added, more reduction reactions of TAP or PES occurred, and Figure 2 shows that more resistive SEI films were formed. Figure 2c shows that the addition of 2% VC alone increases the impedance while the addition of 2% VC in combination with 1% ES or 1% TMSMS lowers cell impedance. Figure 2d shows a summary of the resistances, R ct , of NMC111/graphite pouch cells calculated from Figures 2a, 2b and 2c. R ct was taken to be the diameter of the semi-circle in the Nyquist plot (shown in Figure 2a) which includes the resistances associated with ion desolvation, and transport of electrons and ions through the SEIs on both the positive and negative electrodes (collectively called the charge transfer resistance, R ct, here). Figure 2d clearly shows that R ct is strongly influenced by the type and content of additives used. Figures 3a, 3b and 3c show the discharge capacity versus cycle number for NMC111/graphite cells containing different concentrations of TAP tested with different C-rates: C/3 (3a); C/2 (3b) and 1C (3c) during cycling up to 4.1 V at 20 • C. As mentioned above, electrolyte oxidation and transition metal dissolution at the cathode side are negligible when cells were cycled only up to 4.1 V at only 20 • C. 30,31 The capacity loss due to the growth and repair of the SEI on the negative electrode is primarily time dependent and should be similar at different charge rates after the same testing time. Therefore, the rapid capacity loss at high charge rate is due to factors other than parasitic reactions at the positive electrode and SEI growth at the negative electrode. In addition, earlier work by Burns et al. 32 has shown the charge rate dependence of unwanted Li plating. Therefore, rapid capacity loss, if any, during 20 • C cycling was caused by unwanted lithium plating, 32,33 which will be supported by post-mortem pictures of the negative electrodes later. Figures 3a, 3b and 3c show that cells with larger amounts of TAP or higher charge rate have lower capacities and worse capacity retention. Figure 3a shows that rapid capacity loss occurs during cycling at C/3 for cells with control + 3% TAP, indicating that unwanted lithium plating occurs at C/3. Figures  3b and 3c show that unwanted lithium plating begins at C/2 and 1 C for cells with control + 2%TAP and control + 1%TAP, respectively. This result clearly shows that cells with increasing amounts of TAP are more prone to unwanted lithium plating and this correlates with the increasing R ct shown in Figures 2a and 2d. during cycling decreases the cell capacity. This increased polarization is mainly caused by the increased R ct of the negative electrode which will be confirmed by symmetric cell experiments below, thus the increased polarization forces the potential of the negative electrode below 0 V vs. Li/Li + to induce unwanted lithium plating. For cells with control + 3% TAP, V increases faster during cycling at C/3 suggesting that unwanted lithium plating causes impedance growth for control + 3% TAP, presumably by blocking the surface of the negative electrode. Figures 3d and 3e show that V of cells with control electrolyte increases with cycle number much faster than cells with control + 1% TAP which suggests that other parasitic reactions still occur in cells with control electrolyte during long-term cycling even though cells with the control electrolyte show no unwanted lithium plating. The slow increase of V with cycle number for cells with TAP at low charge rates where no unwanted lithium plating occurs may be due continual reduction of TAP at an appreciable rate after the formation cycle. Figures 4a, 4b and 4c show the discharge capacity versus cycle number for NMC111/graphite pouch cells containing different concentrations of xPES211 (x = 0.25, 0.5, 0.75, 1) tested with different C-rates: C/3 (4a); C/2 (4b) and 1 C (4c) during cycling to 4.1 V at 20 • C. One C/20 cycle was also made before and after the testing at higher C-rates. Figures 4a, 4b and 4c show that with increased concentrations of xPES211, the discharge capacity is reduced especially during high rate cycling. Rapid capacity fade and unwanted lithium plating occur at C/2 for cells with xPES211(x = 1) while unwanted lithium plating occurs at 1 C for cells with xPES211(x = 0.5 or 0.75). Cells with xPES211 (x = 0.25) have about the same capacity retention as cells incorporating control electrolyte when cycling at C/3 and C/2 but their capacity retention is slightly better at 1 C (see next paragraph).

Results and Discussion
Figures 4d, 4e and 4f show V versus cycle number for NMC111/graphite pouch cells corresponding to Figures 4a, 4b and  4c, respectively. Figure 4d shows that the V is stable for cells with xPES211 (0 ≤ x ≤ 1) during cycling at C/3. Figure 4e shows that the V does not increase much during cycling even though rapid capacity fade and unwanted lithium plating occur for xPES211 (x = 1) during cycling at C/2. Figures 4d, 4e and 4f show that V of cells with control electrolyte increases slighty faster during C/3, C/2 and 1 C cycling compared with cells containing xPES211 (x = 0.25) at the same charge rate. This result suggests that cells with control elec-trolyte without additives have more parasitic reactions and a small amount of xPES211 (x = 0.25) reduces impedance growth during long-term cycling even at 1 C. Figure 4f shows that during cycling at 1 C, V increases significantly for cells with xPES211 (x ≥ 0.5) which is probably associated with unwanted lithium plating.
Figures 5a, 5b and 5c show the discharge capacity versus cycle number for NMC111/graphite pouch cells containing control electrolyte with 2% VC, 2% VC + 1% TMSMS or 2% VC + 1% ES tested with different C-rates: C/2 (5a); 1 C (5b) and 1.5 C (5c) during cycling to 4.1 V at 20 • C. Figure 5a shows that there is no much difference between cells with the different additives during cycling at C/2 and this is supported by similar V versus cycle number graphs in Figure 5d. Figure 5c shows that during cycling at 1.5 C, cells with control + 2% VC have the worst performance while the combination of 2% VC with 1% ES improves cycling performance at 1.5 C. However the performance is not as good as cells with control electrolyte.
Figures 5d, 5e and 5f show V vs. cycle number for the NMC111/graphite pouch cells described in Figures 5a, 5b and 5c, respectively. Figure 5e shows that during cycling at 1 C, cells with control + 2% VC + 1% ES have the most stable V vs. cycle number which is consistent with the good cycling performance in Figure  5b. This means that the combination of VC and ES can reduce cell impedance and capacity loss during high rate cycling at 20 • C compared to cells containing VC alone. Figure 5f shows that cells with 2% VC and 2% VC + 1% TMSMS have higher polarization which is consistent with the bad performance during 1.5 C cycling shown in Figure 5c.  Figure  6a shows that MP does not influence graphite passivation and the reduction peak at 2.4 V corresponds to the reduction of PES. Figure 6b shows the electrochemical impedance spectra for cells with different contents of MP measured at 3.8 V and 10.0 • C after formation. Adding MP as a co-solvent decreases the diameter of the semi-circle but the changes are relatively small after the MP content increases above 10%. Figure 6c shows the conductivity of electrolytes with different contents of MP as co-solvent. As the amount of MP increases, the conductivity of the electrolyte increases since MP has a lower viscosity and lower melting point than the other solvents. 34   Figures 7b and 7c show that compared to cells containing control + PES211, cells with MP as co-solvent and PES211 as additives have slightly better capacity retention during cycling at C/2 and 1 C at 20 • C. This is most likely due to the higher ion conductivity of the MP-containing electrolytes. However, the improvement is quite limited compared to cells containing control electrolyte without PES211 as additives. Figures 7d, 7e and 7f show V versus cycle number of cells corresponding to Figures 7a, 7b and 7c. Again, the behavior of discharge capacity versus cycle number is correlated with V versus cycle number. Cells with MP-containing electrolyte and PES211 have lower V, thus higher capacity retention compared to cells containing control + PES211. Compared to control electrolyte, cells with MP-containing electrolyte and PES211 still have higher V especially during 1 C cycling (Figure 7f). This result shows that although adding MP increases the electrolyte ionic conductivity, the high cell resistance caused by the PES211 electrolyte additives is still primarily responsible for the poor high rate cycling caused by unwanted lithium plating. Figure 8 shows the overall capacity loss measured by the C/20 cycles before and after the high rate cycling at 20 • C. Two cells were measured for each charge rate and the error bars represent the range of the data from the pair cells cycling at the same current and the same electrolyte.
The blue shaded area in Figure 8 indicates a capacity loss less than 5% where no unwanted Li plating occurs. Capacity loss less than 5% occurs at small charge rates (e.g. C/3) and is probably caused by the growth and repair of the negative electrode SEI during cycling. The larger capacity loss above the shaded area in Figure 8 indicates where unwanted lithium plating occurs as lithium plating accelerates capacity loss. Figures 8a and 8b show that higher contents of TAP or xPES211 cause rapid capacity loss and unwanted lithium plating at lower currents. This result can be explained by the cell impedance shown in Figure 2 and the values of V shown in Figures 3 and 4. The propensity for unwanted lithium plating increases with increased cell resistance when xPES211 or TAP is added to the electrolyte. Figure 8c shows that cells containing VC and ES have nearly the same capacity loss as cells with control electrolyte even at 1.5 C, consistent with Figure 2c which shows that ES lowers the resistance of cells containing VC. Figure 8d shows that the MP co-solvent helps to lower capacity fade and reduce the likelihood of unwanted Li plating at C/2 while rapid capacity loss and unwanted Li plating still occur at 1 C for cells with MP co-solvent and PES211. Figure 9 shows the pictures of negative electrodes extracted from pouch cells after cycling at different C-rates for cells with 0% to 3% TAP. These photos are consistent with the capacity loss versus C-rate shown in Figure 8. Figure 8 shows that the cells with control electrolyte begin to lose large amounts of capacity during cycling at 1.5 C and Figure 9b shows that unwanted lithium plating was observed on the graphite surface after testing at 1.5 C while no unwanted lithium plating was observed after testing at 1 C as shown in Figure 9a. (Note: The photographs were taken through a glove box window and unwanted lithium plating may not be clearly shown in the pictures, however, lithium plating on the graphite surface was clear to the naked eye.) Figures 9c and 9d show that for cells with 1% TAP, no unwanted lithium plating occurred on the negative electrode after cycling at C/2 while unwanted lithium plating was observed after cycling at 1 C, consistent with Figure 8a. Figures 9e and 9f show that after cycling at C/3, unwanted lithium plating occurred for cells with 2% TAP and 3% TAP. Figure 9f shows that more white areas (i.e. lithium plating) were observed on the graphite electrode for cells with 3% TAP than for cells with 2% TAP (Figure 9e). This is consistent with the more severe capacity loss at C/3 for cells with 3% TAP than for cells with 2% TAP as shown in Figure 8a. The post-mortem analysis further confirms that capacity loss during high rate cycling at ambient temperature and modest upper cutoff potential (i.e. 4.1 V) can be used as a detection method for unwanted lithium plating since other parasitic reactions are negligible by comparison. Figures 10a to 10d show the resistance, R ct , (diameter of the semicircle in the Nyquist plot) after formation and after 20 • C cycling at C/3, C/2, 1 C and 1.  Figures S1, S2, S3 and S4 in the supporting information. Figure 10a shows that R ct decreases after cell cycling except for cells with 2% TAP after 1 C cycling. The impedance growth for cells c) e) f) a) b) d) Figure 9. Pictures of the graphite electrodes extracted from pouch cells containing: (a) control electrolyte after 1 C cycling; (b) control electrolyte after 1.  with 2%TAP after 1 C cycling might be caused by the large amount of unwanted lithium plating. Figure 10b shows that R ct decreases for cells with xPES211 (0 ≤ x ≤ 1) after cycling at different charge rates even though unwanted lithium plating occurs at high charge rates. Figure 10c shows that R ct increases for cells with 2% VC and 2% VC + 1% TMSMS after cycling at 1.5 C while the R ct of cells with control and control + 2% VC + 1% ES changed little even after cycling at 1.5 C. Figure 10d shows that R ct increases after 1 C cycling for cells with MP co-solvent and the addition of PES211. This is probably caused by the increased impedance of the negative electrode after unwanted lithium plating. Generally, R ct of full cells decreased during cycling at 20 • C except for cells with a large amount of unwanted lithium plating. However, cell impedance does not always increase when unwanted lithium plating occurs as is the case for xPES211. Figure 11 shows the area-specific Nyquist plots of the reassembled full coin cells, positive electrode symmetric cells and negative electrode symmetric cells constructed from parent NMC111/graphite pouch cells containing different electrolyte additives. In each panel, the results for electrodes from cells with control electrolyte have been included for comparison. The parent pouch cells were dissembled after cycling at 20 • C and were the same pouch cells that were used to generate the data in Figures 3, 4 and 5. Symmetric cells were also reconstructed from cells after formation and EIS spectra for these cells are presented in Figures S5 and S6 in the supporting information. Figures 11a, 11b and 11c show the EIS spectra from full cells and symmetric cells reconstructed from NMC111/graphite pouch cells initially filled with control electrolyte plus 0%, 1%, 2% and 3% TAP. These full and symmetric coin cells were made from the pouch cells after the C/3 cycles at 20 • C (Figure 3a). Figures 11a, 11b and 11c show that most of the impedance increase of the full cells containing 1%, 2% or 3% TAP comes from the negative electrode. Figure  11b also shows that TAP dramatically increases the charge transfer resistance of the negative electrode. This is probably caused by the more resistive SEI on the negative electrode that TAP forms. However, the impact of 1%, 2% or 3% TAP on the positive electrode is small and little changes can be seen in Figure 11c. Figures 11d to 11f show the EIS spectra of full and symmetric cells reconstructed from NMC111/graphite pouch cells initially filled with electrolyte containing xPES211(x = 0, 0.5, 1.0). These full and symmetric coin cells were made from the pouch cells after the C/3 cycles at 20 • C (Figure 4a). Figures 11d, 11e and 11f clearly show that the impedance increase that xPES211 creates in the full cell is due primarily to its effect on the negative electrode. Figures 11e  and 11f show that xPES211 dramatically increases the impedance of the negative electrode as x increases while it decreases the positive electrode impedance. Figures 11g, 11h and 11i show that EIS spectra of symmetric cells and full coin cells containing electrolytes with 2% VC, 2% VC + 1% TMSMS and 2% VC + 1% ES. These full and symmetric coin cells were made from the pouch cells after C/2 cycling at 20 • C (Figure 5a) using the same electrolyte as the dissembled pouch cell. Figures 11g,  11h and 11i show that with the addition of 2% VC, the increased impedance of the negative electrode exceeds the reduced impedance of the positive electrode, thus the full cell impedance is higher than for cells with control electrolyte. Cells with 2% VC + 1% TMSMS and cells with 2% VC + 1% ES have nearly the same negative electrode impedance as cells with control electrolyte. Figure 12a shows a strong correlation between the negative electrode resistance, R negative , and the onset current for unwanted lithium plating. R negative is read from the Nyquist plots shown in Figures 11b,  11e and 11h. Here, the processes of ion desolvation, electron and ion transport through the solid electrolyte interphase (collectively called charge transfer resistance here), current collector/active particle contact resistance and the resistance of Li + ion diffusion in the pores of the negative electrode are lumped into R negative . The effects of the contact resistance and resistance of Li + ion diffusion in the pores of the negative electrode are seen as the small high frequency shoulder in these spectra. [35][36][37] The charge transfer resistance (the diameter of mid-frequency semi-circle) dominates R negative which means changes in R negative are mainly determined by the charge transfer resistance of the negative electrode. R negative in Figure 12a was obtained from symmetric negative electrodes of dissembled pouch cells directly after formation and R negative in Figure 12b was obtained from symmetric negative electrodes of dissembled pouch cells after 20 • C cycling (shown in Figure 11). Detailed Nyquist plots of symmetric cells with different electrolytes are shown in Figures S5 and S6 and the cycling performance at 20 • C of NMC111/graphite cells using pyridine boron trifluoride (PBF) as an additive is shown in Figure S7 in the supporting information. Figure 12 shows that there is a general negative correlation between the magnitude of the negative electrode resistance and the onset current for unwanted Li plating for all cells. Cells with control electrolyte have lower R negative , thus are less likely to have unwanted lithium plating. Electrolyte additives that lead to low negative electrode impedance such as the combination of 2% VC and 1% ES promote high charge currents before unwanted lithium plating while additives with high negative electrode impedance such as high concentrations of TAP or PBF are more likely to lead to unwanted lithium plating. Therefore, cells with additives of that yield lower R negative are expected to have higher onset currents for unwanted lithium plating. Figures 12a and 12b show that the correlation between high R negative and small onset current for lithium plating is observed if R negative is measured directly after formation or after cycling. This is very convenient in the search for additives that simultaneously promote long lifetime and high charge rates -one can evaluate the negative electrode charge transfer impedance in symmetric cells made right after formation. Figures 12a and 12b show that the negative electrode resistance is generally slightly lower when measured after cycling compared to a) The onset C-rate for unwanted lithium plating at 20 • C plotted versus the negative electrode area specific resistance, R negative , measured at 10 • C, obtained from negative electrode symmetric cells prepared from electrodes harvested from NMC111/graphite pouch cells after formation. Corresponding pouch cells and symmetric used the same electrolytes as indicated in the Figure legend. The dashed line is described in the text; b) same as a), except the negative electrode specific resistance (10 • C) was obtained from electrodes extracted from cycled cells. c) The measured onset current for unwanted lithium plating for cells with different electrolytes during cycling at 20 • C and 10 • C as well as the prediction of Equation 1 both plotted versus the negative electrode area specific resistance, R negative from symmetric cells. For the data points in panel c), both the cycle testing and the EIS spectra from symmetric cells were collected at the same temperature. d) The calculated onset current for unwanted lithium plating during cycling at 10 • C (from Equation 1) plotted versus R negative . The measured onset currents at 20 • C are also shown. The dashed line through the 20 • C data is proportional to 1/R negative to capture the trend predicted by Equation 1. the same electrodes measured just after formation. This suggests that the SEI in these electrolytes matures somewhat over time during the cycling of the cells.
A simple model can be used to explain the results in Figures  12a and 12b. Figure S8a shows that the potential of the stage 1-2 plateau of Li x C 6 measured at C/20 and 40 • C is about 0.080 V higher than Li/Li + . Lithium plating will occur thermodynamically when the negative electrode potential is below 0 V vs. Li/Li + . Therefore, it is assumed that unwanted lithium plating can begin when the negative electrode overpotential is about 0.080 V. Figure S8b shows a section of the graphite negative electrode having area, S, and thickness, L. The negative electrode resistance, R, is obtained by dividing the negative electrode area specific resistance, R negative , (obtained from the Nyquist plots in Figures 11b, 11e and 11h) by the active surface area of the negative electrode. In the case where the negative electrode resistance, R, dominates the contributions to the overpotential, η, then where I is the cell current. If η ≥ 0.080 V, one expects unwanted lithium plating to occur. Therefore the onset current for unwanted lithium plating, I u , is given by I u ≥ 0.080 V/R = 0.080 V x S/R negative .
[1] Figure 12c shows the calculated onset current for Li plating, I u , given by Equation 1 for cells tested both at 10 • C and 20 • C. For example, R negative of a negative electrode symmetric cell with control electrolyte is 34.8 × cm 2 measured at 10 • C as shown in Figure S6 in the supporting information. The active negative electrode surface area in the pouch cell is 90.74 cm 2 . This yields I u = 0.080V/(34.8 × cm 2 ) × 90.74 cm 2 = 208.6 mA. Therefore, the calculated onset current for unwanted Li plating is predicted to be about 208.6 mA (0.95 C) for cells with control electrolyte at 10 • C. Figure 12c compares the onset current for unwanted lithium plating from high rate cycling experiments and post mortem analysis to the predictions of Equation 1. The EIS measurements on negative/negative symmetric cells (shown in Figure S9) used for the data points in Figure 12c were made at the same temperature as the determination of the onset current for unwanted Li plating and it appears that data collected at both 10 and 20 • C fall onto the same universal curve.
It is unfortunate that the data shown in Figures 12a and 12b are for measurements of R negative at 10 • C and for the onset current for unwanted Li plating at 20 • C. When these experiments began EIS spectra were taken always at 10 • C in order to amplify differences between cells with different electrolyte additives and it was only realized later, based on the trends in Figures 12a and 12b, that a simple relation like Equation 1 might describe the data. The temperature dependence of R negative for the negative electrodes with SEI layers formed in different electrolytes is unknown and should be characterized. This is most likely why the points in Figures 12a and 12b do not fall closer to a single trend line. The dashed line in Figure 12a is simply a graph of the function: B/R negative with B = 64.34 × cm 2 to show that data in Figure 12a approximately follows the trend of Equation 1. Figure  12d shows the predicted values for the onset currents for unwanted lithium plating at 10 • C calculated using the values of R negative measured at 10 • C using Equation 1. Figure 12d also shows the measured values of the onset currents for unwanted lithium plating at 20 • C plotted versus R negative measured at 10 • C. Figure 12d suggests the onset currents for unwanted lithium plating at 10 • C will be about 30% to 50% of the values at 20 • C.
It is important to stress that the overpotentials due to lithium diffusion in the graphite and lithium-ion diffusion in the electrolyte have not been considered in the simple model that yields Equation 1. The goal of this work was to explore how electrolyte additives, which affect the SEI layers, affect the onset current for lithium plating. Obviously, when very large graphite particles or very thick electrodes are used, those factors must be taken into account.

Conclusions
The impacts of negative electrode impedance and electrolyte ionic conductivity on unwanted lithium plating were investigated with 20 • C high rate cycling, EIS measurements and symmetric cells. Unwanted lithium plating was the main mechanism for rapid capacity loss with cycle number at high charge rates in this study. The negative electrode SEI and its impedance were changed by adding different types and different amounts of electrolyte additives including TAP, xPES211 (0 ≤ x ≤ 1), VC, VC + TMSMS and VC + ES to 1 M LiPF 6 EC:EMC(3:7) electrolyte. The additives used and the amount added all strongly affected cell impedance and unwanted lithium plating. Symmetric cells showed that higher concentrations of TAP or xPES211 increased the resistance of the negative electrode, which decreased the onset current for unwanted lithium plating. Cells with both the additives VC and ES had lower negative electrode impedance and higher onset currents for unwanted lithium plating. Therefore, additives that promote low negative electrode charge transfer resistance are good for low-temperatures and high rate charging. Such additives must be carefully selected so that the lifetime of Li-ion cells is not compromised.
A simple model was proposed for estimating the onset current, I u , for unwanted lithium plating. Based on this model, I u = 0.080 V x S/R negative , where S is the electrode surface area and R negative is the area specific negative electrode resistance measured as the diameter of the "semicircle" in the Nyquist plot of a negative electrode symmetric cell. The predictions of the model agreed very well with the trends observed in experiment.
Researchers hoping to find electrolytes and electrolyte additives that simultaneously promote long lifetime and high charge rates can simply measure the impedance spectrum of a negative electrode symmetric cell made from negative electrodes harvested after formation and then use the simple model to estimate I u . Obviously, other measurements, such as high precision coulometry, are required to estimate the impact of electrolytes and electrolyte additives on cell lifetime.