Understanding the Formation and Evolution of Lithium-Ion Battery Solid Electrolyte Interphase

Lithium ion batteries are widely used as energy storage devices in a variety of products such as smartphones, tablets, laptops and other portable electronics. Thanks to their high energy density and cyclability, they are currently being used by and developed for electric vehicles. There is a growing need for cost reduction; increase in energy density; wider operating temperature range; and improved safety characteristics of the batteries. Organic carbonates are the primary solvents used in lithium-ion battery electrolytes along with electrolyte additives. The reversibility of current lithium-ion batteries is dependent upon the electrolyte used in the batteries. During the initial charging cycles of the cell, a solid electrolyte interface (SEI) is formed by reduction of organic carbonates, electrolyte salts and/or electrolyte additives on the surface of the graphitic anode in lithium-ion batteries. The generation of a stable anode SEI prevents continuous electrolyte reduction on the surface of the anode. The SEI functions as a Li ion conductor but an electrical insulator. The reduction reactions of the electrolytes on the graphitic anode surface have been investigated for many years and it been proposed to contain a complicated mixture of products including lithium oxalate, lithium alkoxides, and lithium oxide from the carbonate solvents and LiF and lithium fluorophosphates from the reduction of LiPF6. Similar ambiguity exists about the components of SEI formed from electrolyte additives and other electrolyte salts. Despite the extensive investigations, the structure, formation mechanisms and evolution of the SEI are poorly understood. Understanding the mechanisms of the reduction reactions of organic carbonates, electrolyte salts and electrolyte additives along with the products of the reactions which result in the generation of the SEI is essential for the development of safer lithium-ion batteries with wider operating temperature range. Lithium naphthalenide has been investigated as a one electron reducing agent for organic carbonates solvents, some of the most robust additives and salts used in lithium ion battery electrolytes. The reaction precipitates have been analyzed by IRATR, XPS and solution NMR spectroscopy. The evolved gases and the volatile components have been analyzed by GC-MS. The reduction products of ethylene carbonate and propylene carbonate are lithium ethylene dicarbonate (LEDC) and ethylene and lithium propylene dicarbonate (LPDC) and propylene, respectively. The reduction products of diethyl and dimethyl carbonate are lithium ethyl carbonate (LEC) and ethane and lithium methyl carbonate(LMC) and methane, respectively. Electrolyte additives, FEC and VC reductively decompose to HCO2Li, Li2C2O4, Li2CO3, and polymerized VC. All the fluorine containing salts generate LiF upon reduction. In addition to LiF, LiBF4 generates LixByFz species; LiBOB and LiDFOB generate lithium oxalate and boron-oxalatoesters; LiPF6 yields LiPF2 species and LiTFSI produces lithium bis[N-(trifluoromethylsulfonylimino)] trifluoromethanesulfonate. The poor thermal stability of the SEI layer has been attributed to exothermic reactions between lithium alkyl carbonates and LiPF6. While the relationship between capacity fade and SEI instability is clear, and there have been some investigations of SEI component evolution, the mechanism of SEI component decomposition is complicated by the presence of many different components. The thermal stability of Li2CO3, LMC, and LEDC in the presence of LiPF6 in dimethyl carbonate (DMC), a common salt and solvent, respectively, in lithium ion battery electrolytes, has been investigated to afford a better understanding of the evolution of the SEI. The residual solids from the reaction mixtures have been characterized by a combination of X-ray photoelectron spectroscopy (XPS) and infrared spectroscopy with attenuated total reflectance (IR-ATR), while the solution and evolved gases have been investigated by nuclear magnetic resonance (NMR) spectroscopy and gas chromatography with mass selective detection (GC-MS). The thermal decomposition of Li2CO3 and LiPF6 in DMC yields CO2, LiF, and F2PO2Li. The thermal decomposition of LMC and LEDC with LiPF6 in DMC results in the generation of a complicated mixture including CO2, LiF, ethers, phosphates, and fluorophosphates.


STATEMENT OF THE PROBLEM AND PROPOSED SOLUTIONS
Organic carbonates are the primary solvents used in lithium-ion battery electrolytes along with electrolyte additives. The reversibility of current lithium-ion batteries is dependent upon the electrolyte used in the batteries. 4 During the initial charging cycles of the cell a solid electrolyte interface (SEI) is formed by reduction of organic carbonates, electrolyte salts and/or electrolyte additives on the surface of the graphitic anode in lithium-ion batteries. The generation of a stable anode SEI prevents continuous electrolyte reduction on the surface of the anode. The SEI functions as a Li ion conductor but an electrical insulator. 3 The reduction reactions of carbonates on the graphitic anode surface have been investigated for many years. Initially a single two electron reduction mechanism of propylene carbonate to generate Li2CO3 and propylene was proposed, 5 later Aurbach and co-workers proposed two sequential one electron reduction reactions of cyclic carbonates to generate lithium alkyl carbonates and alkenes. 6 Numerous other researchers have investigated the composition of the SEI on graphitic anodes in lithium ion batteries. [7][8][9][10][11][12][13][14][15][16][17] In addition to lithium alkyl carbonates and lithium carbonate, the SEI has been proposed to contain a complicated mixture of products including lithium oxalate, lithium alkoxides, and lithium oxide from the carbonate solvents and LiF and lithium fluorophosphates from the reduction of LiPF6. [7][8][9][10][11][12][13][14][15][16][17] Similar ambiguity exists about the components of SEI formed from electrolyte additives and other electrolyte salts. Capacity fade at elevated temperature is connected to the exothermic reactions between lithium alkyl carbonates and LiPF6. 16,18 While the relationship between capacity fade and SEI instability is clear, 16 The results provide significant insights into the formation and decomposition mechanism of the anode SEI.

INTRODUCTION
Organic carbonates are the primary solvents used in lithium-ion battery electrolytes. The reversibility of current lithium-ion batteries is dependent upon the electrolyte used in the batteries. 1 During the initial charging cycles of the cell a solid electrolyte interface (SEI) is formed by reduction of organic carbonates on the surface of the graphitic anode in lithium-ion batteries. The generation of a stable anode SEI prevents continuous electrolyte reduction on the surface of the anode. The SEI functions as a Li ion conductor but an electrical insulator. 2 Understanding the mechanisms of the reduction reactions of organic carbonates along with the products of the reactions which result in the generation of the SEI is essential for the development of better lithium-ion batteries.
The reduction reactions of carbonates on the graphitic anode surface have been investigated for many years. Initially a single two electron reduction mechanism of propylene carbonate to generate Li2CO3 and propylene was proposed, 3 later Aurbach and co-workers proposed two sequential one electron reduction reactions of cyclic carbonates to generate lithium alkyl carbonates and alkenes. 4 Numerous other researchers have investigated the composition of the SEI on graphitic anodes in lithium ion batteries. [5][6][7][8][9][10][11][12][13][14][15] In addition to lithium alkyl carbonates and lithium carbonate, the SEI has been proposed to contain a complicated mixture of products including lithium oxalate, lithium alkoxides, and lithium oxide from the carbonate solvents and LiF and lithium fluorophosphates from the reduction of LiPF6. [5][6][7][8][9][10][11][12][13][14][15] Despite the extensive investigations, the structure and formation mechanisms of the SEI are poorly understood.
A detailed analysis of binder free graphitic anodes cycled in simplified electrolytes composed of a single carbonate solvent and LiPF6 has been reported. 16,17 These investigations suggest that the initial reduction reaction of the carbonates generate lithium alkyl carbonates and LiF as the predominant components of the anode SEI. As an expansion of these investigations, lithium naphthalenide, a well-known one electron

NMR Spectra Of Precipitates
The molecular structures of the precipitates formed in the reaction between lithium naphthalenide and various carbonates are analyzed via a combination of 1 H, 13 C NMR and IR-ATR spectroscopy. As previously reported, the 1 H NMR spectrum of the reduction product of EC contains a singlet at 3.6 ppm characteristic of (-OCH2CH2O-) and peaks at 62.5 and 161.1 ppm in the 13 C NMR spectrum, characteristic of (-CH2O-) and a C=O, respectively (Figure 2.1). 16 The resonances match those previously reported for lithium ethylene dicarbonate (LEDC). 7 The reaction produces LEDC in high yield ~ 95 %, and no other products are observed. The 1 H and 13 C NMR spectra of the precipitate formed by the reaction of lithium naphthalenide with PC is provided in  In addition to analysis of the reduction products of cyclic carbonates, the reduction of dialkyl carbonates has been investigated. The 1 H NMR spectra of the reaction product of the Lithium naphthalenide reduction of DMC is provided in Figure   2  to the peaks characteristic of lithium alkyl carbonates, a weak absorption is observed at ~1450 cm -1 characteristic of Li2CO3. While there appears to be some Li2CO3 present in all spectra, lithium alkyl carbonates decompose to form Li2CO3, as discussed above.
Thus, the thermal, Lewis acid, or Lewis Base catalyzed decomposition of lithium alkyl carbonates is the most likely source of Li2CO3 since Li2CO3 is not observed in the 13 C NMR spectra of fresh reduction products. In addition, all of the samples have increases in the intensity of the absorption of Li2CO3 upon exposure to air.

GC-MS ANALYSIS OF GASES
Upon reaction of lithium naphthalenide with carbonates. significant gas evolution is observed. The gases evolved during reaction were analyzed by GC-MS.
Analysis of the evolved gas during the reduction of EC confirms that the gas produced is ethylene. Reaction of PC results in similar gas evolution and the gaseous product is propylene. The two cyclic carbonates have very similar reaction mechanisms which are consistent with those originally proposed by Aurbach and co-workers. 4 The reduction follows two sequential single electron transfer reactions (Scheme 2.1). The first electron generates a radical anion, the second electron generates lithium alkylene dicarbonates and an alkene. Aurbach and co-workers. 4 There is no evidence for the generation of either Li2CO3 or CO2 during the initial reduction reactions. It is well known that the anode SEI changes upon additional cycling and thermal abuse. 14

02881, United States
The following is published in the journal, Chemistry of Materials, and is presented here in manuscript format.

ABSTRACT
We have synthesized the products of fluoroethylene carbonate (FEC) and vinylene carbonate (VC) via lithium naphthalenide reduction. By analyzing the resulting solid precipitates and gas evolution, our results confirm that both FEC and VC decomposition products include HCO2Li, Li2C2O4, Li2CO3, and polymerized VC. For FEC, our experimental data supports a reduction mechanism where FEC reduces to form VC and LiF, followed by subsequent VC reduction. In the FEC reduction product, HCO2Li, Li2C2O4, and Li2CO3 were found in smaller quantities than in the VC reduction product, with no additional fluorine environments being detected by solid-state nuclear magnetic resonance or X-ray photoelectron spectroscopy analysis. With these additives being practically used in higher (FEC) and lower (VC) concentrations in the base electrolytes of lithium-ion batteries, our results suggest that the different relative ratios of the inorganic and organic reduction products formed by their decomposition may be relevant to the chemical composition and morphology of the solid electrolyte interphase formed in their presence.

INTRODUCTION
Additives are widely used to improve performance of Li-ion batteries, offering an economically viable method of performance enhancement compatible with existing manufacturing infrastructure. 1 Generally, the function of additives is sacrificial: they are reduced at different voltage potentials compared to the base electrolytes to which they are added, forming decomposition products that are incorporated into a protective layer on electrodes. 1−5 This protective layer is called the solid electrolyte interphase (SEI). 6,7 The formation of a stable SEI is essential for all Li-ion batteries, preventing further electrolyte decomposition, thereby underlying capacity retention. 1,5−8 The SEI also represents an electronically insulating barrier between the electrodes and electrolyte, with its composition, thickness, and structure influencing the lithium transport across the interphase. 9 Practically, the SEI layer is extremely air-sensitive. 30,31 Moreover, with a thickness of less than 100 nm, it is very difficult to study experimentally. Here, our strategy is to synthesize the reduction products of FEC and VC in order to experimentally confirm their reduction products. Lithium naphthalenide (Li-Nap), a well-known reducing agent, is known to react with solvents in a similar manner to those which may occur on lithiated anodes. 32 It is used here to reduce FEC and VC, modeling a reduction process in a similar manner to that which may occur in a lithium-ion battery.
Solid products are analyzed with X-ray photoelectron spectroscopy (XPS), solid-state NMR (ssNMR), and Fourier transform infrared spectroscopy (FTIR). Gas evolution is monitored using gas chromatography mass spectrometry (GC-MS). The many techniques provide chemical signatures for future work. Viable reactions to form the detected decomposition products are proposed. For FEC, we propose a reduction scheme where FEC reduces to form LiF and VC, followed by further reduction of VC to polymerized VC (poly(VC)). The poly(VC) contains repeating EC units joined by cross-linking sites; our analysis shows no evidence for F−C bonds in the polymer.
HCO2Li, Li2C2O4, and Li2CO3 are also found in small quantities. For VC, we detect lithium environments of HCO2Li, Li2C2O4, and Li2CO3, in addition to poly(VC).

Synthesis.
All reagents were used as obtained, without further purification. Batterygrade VC and FEC were obtained from BASF. Naphthalene (99 + %, Scintillation grade) and THF (Anhydrous, 99.9%) were purchased from Acros organics. Lithium discs were XPS. XPS spectra of the dried precipitates were acquired using a thermo scientific Kalpha XPS instrument. Samples were made into circular pellets with a press and transferred from the glovebox to the XPS chamber using a vacuum transfer module without exposure to air. C 1s, O 1s, and Li 1s spectra were obtained from the VC precipitate, whereas C 1s, O 1s, Li 1s, and F 1s spectra were acquired from the FEC precipitate. An Argon flood gun was used to avoid surface charge accumulation during sample analysis. The binding energy was corrected on the basis of the C 1s of hydrocarbon at 284.8 eV. The data was processed and analyzed using the Thermo Avantage, XPS Peak 4.1 and the Origin software.
ssNMR. Multinuclear ssNMR spectra were obtained on 16.4 T Bruker Avance III 700 MHz and 11.7 T Bruker Avance III 500 MHz spectrometers. Samples were packed in an Ar glovebox (typically O2 and H2O < 0.1 ppm), avoiding any exposure to ambient air, into rotors of 1.3, 3.2, and 4 mm outer diameters. Magic-angle spinning (MAS) frequencies ranged from 10 to 60 kHz, spinning under N2. 1 H and 13 C chemical shifts were externally referenced to adamantane ( 1 H 1.9 ppm, 13 C 38.5 ppm, CH2) and 7 Li and 19 F to LiF ( 7 Li −1 ppm, 19 F−204 ppm). 33,34 The data were processed using the Bruker TOPSPIN software and analyzed using the dmfit software. 35  Each of the FEC and VC precipitates were investigated using 1 H, 7 Li, and 19 F ssNMR, using 1.3 mm rotors and 60 kHz spinning frequency. 13 C ssNMR experiments were performed using larger 3.2 and 4 mm rotors and spinning frequencies ranging from 10 to 12 kHz. The larger samples provided greater sensitivity. The 13 C spectra were acquired using sweptfrequency two-pulse phase modulation (swfTPPM) 36 1 H decoupling at 80−100 kHz. Direct excitation 13 C experiments provided quantitative information. 1 H− 13 C, 7 Li− 13 C, and 19 F− 13 C correlation experiments were used to probe spatial proximity of these nuclei by transferring magnetization from 1 H, 7 Li, and 19 F nuclei by cross-polarization to C nuclei. Dipolar dephasing (interrupted decoupling) 1 H− 13 C crosspolarization experiments allowed differentiation between protonated and nonprotonated environments. 34,37 Further ssNMR experimental details are given in the Supporting Information.
FTIR. FTIR analysis was performed on each of the precipitates prepared with nondeuterated naphthalene. FTIR spectra of the dried precipitates were acquired on a Bruker Tensor 27 spectrometer, equipped with germanium crystal, in attenuated total reflectance (IR-ATR) mode. Samples were transferred using airtight vials, and the spectrometer was operated inside a nitrogen filled glovebox to avoid sample exposure to ambient air. Each spectrum was acquired with 128 scans from 700 to 4000 cm −1 at the spectral resolution of 4 cm −1 . The data was processed and analyzed using the OPUS and Origin software. Helium was used as carrier gas at a flow rate of 24 mL/min. The initial column temperature was 40 °C, and the temperature was ramped at 10 °C/min to 200 °C and held at that temperature for 2 min with the total run time of 18 min. The mass spectra obtained were compared to the NIST library to determine their molecular structures.

GC-MS
Computational Methods. Chemical shifts were calculated using density functional theory (DFT) using Gaussian 09 38 and estimated using ChemNMR implemented in ChemBioDraw 13.0; see Table S1. ChemNMR approximates 13  Li2CO3. 21,44 The peak at 288.5 eV is assigned to CO2 environments contained in HCO2Li and/or Li2C2O4, while the peaks at lower binding energies of 286.8 and 284.8 eV indicate C−O and C−C bonds, respectively. Li2CO3 has a larger contribution in the VC precipitate compared with the FEC precipitate, also seen by 7 Li and 13 C ssNMR later.
In addition, the relative signal intensity of the HCO2Li/Li2C2O4 peak is larger in the VC precipitate, these carboxylate environments being confirmed by 13 50 We note that residual THF may contribute to the signals of 1.5 and 3.6 ppm. In the 7   to this C environment. A 7 Li− 13 C experiment was also attempted, but no signal was detected: the null result is in agreement with the 7 Li ssNMR assignment (Figure 3. 2) showing that very little Li2CO3 (or similar environment resonating near 0 ppm in the 7 Li spectrum) is present in the FEC precipitate. 1 H− 13 C cross-polarization experiments, Figure 3.3a-iii, were performed, further confirming the 13 C assignments. The signal intensity in these experiments depend on the dynamics of the functional groups and the molecules in the SEI, and spatial proximity of 1 H and 13 C nuclei. Following a similar strategy used in our previous paper to assign different carbon local environments, 52 a delay time is introduced following the cross-polarization step in the experiment to perform a dipolar dephasing (interrupted decoupling) experiment, the experimental details being described in the Supporting Information. When the delay times are varied in the experiment, different C functional groups can be identified on the basis of their attenuation rates. Protonated C is attenuated more rapidly than non-protonated C. Also, the signal intensities for rigid CH/CH2 environments attenuate more rapidly than signal from mobile species such as rotating CH3 methyl groups (due to a reduced dipolar coupling The minor peak I (13 ppm) is characteristic of CH3R environments; the resonance likely has some contribution from residual diethyl ether (CH3CH2−O−CH2CH3) used to rinse the precipitates during synthesis. However, reactions forming these environments in minor quantities may also contribute to the signal.
The 7 Li− 13 C cross-polarization experiment ( Figure 3.3b-ii) indicates the Li + coordination environments by the carboxyl and carbonate groups. The broad resonance of A is consistent with the HCO2Li assignment. The majority of the signal contributing to the asymmetric peak at B is assigned to Li2CO3, the small shoulder being assigned to Li2C2O4. The 7 Li ssNMR spectra are consistent with these assignments (Figure 3.2).
In the dipolar dephasing experiment (Figure 3.3b- (Figure 3.3b-iii), which is not consistent with its assignment solely to HCO2Li. This spectrum was collected at a higher magnetic field strength than the spectra shown in Figure 3.3b-i,-ii, and there is now a severe overlap with the now much more intense D-Napth. spinning sideband (labeled with an asterisk), this signal not being attenuated in the dephasing experiment. Similarly, no attenuation is expected for an acetate resonance.
FTIR. FTIR spectra of the precipitates obtained on reduction of FEC and VC are displayed in Figure 3.4, confirming chemically bonded groups assigned in our XPS and ssNMR spectra. Our assignment here is based on comparison of the spectra to related studies. 18,32,44,47 The FEC and VC reduction products have similar FTIR signatures, with some relative intensity differences at approximately 1300, 1400−1500, and 1750 cm −1 .
In addition to the VC/FEC reduction products, some residual naphthalene is seen (788, 3064 cm −1 ). In each of the samples, the previously assigned Li2CO3 is again observed (878, 1449, and 1488 cm −1 ). As also seen by 13 C ssNMR (Figure 3.3b-ii) and the C 1s XPS spectra (Figure 3.1), the Li2CO3 is more prevalent in the VC sample.
In the FEC reduction product, peaks for carbonate C=O (1795 cm −1 ) and C−O (1080, 1171 cm −1 ) bonds are seen. These peaks are assigned to bonds contained in ROCO2R environments, resembling those assigned to a poly(VC) product in our previous study. 18  To gain further insight into the decomposition mechanisms, an additional experiment was performed with a half molar equivalent of Li-Nap, providing FTIR spectra comparable to that of Figure 3.4 (see Supporting Information). The spectra show much weaker intensities for the peaks assigned to Li2CO3, Li2C2O4, and HCO2Li, relative to the peaks assigned to ROCO2R environments assigned to poly(VC), revealing a Li concentration dependence in the formation of these inorganics.
GC-MS Analysis of Gases. GC-MS analysis was performed, providing more insight for the reduction mechanisms of FEC and VC, resulting in the solid precipitates. For FEC, the reduction with Li-Nap yields a mixture of CO and CO2. The ratio of CO to CO2 peak areas is 1:4.4. For VC, the reduction with Li-Nap yielded carbon monoxide as the only gaseous product (i.e., no CO2 was detected), CO2 having been detected previously. 45,47 The absence of CO2 detection was attributed to its consumption in further reactions, the experiment being performed in a closed system, with an abundance of Li.
Proposed Reduction Products. On the basis of the above analysis, we propose that the reduction product poly(VC) is present in the precipitates of both FEC and VC, as well as Li2C2O4, Li2CO3, and HCO2Li, Figure 3.5. The relative ratio of these products differs for VC and FEC.
The ROCO2R environment observed in the XPS, ssNMR ( group and contribute to the 13 C ssNMR signal of resonance H (Table 3.2). 13 C ssNMR resonance D, assigned to protonated C environments adjacent to two OR groups, indicates the possibility of a cross-linking site for poly(VC) (seen in Figure 3.5). The cross-linking site may also contain C environments adjacent to one O, contributing to resonance E. The signals from the distribution of carbons not adjacent to O (Table 3. Definitive assignments for C-groups and fluorinated-species contained in the reduction products were aided by the large chemical shift dispersion of the 13 Figure 3.5). The carboxylate and carbonate assignments in the 13  With the reduction products detected being nearly identical to FEC, the difference being the LiF product, similar reduction mechanisms are expected. One possible mechanism for the reaction of FEC is nearly identical to the reaction of VC, except that the first step of the reaction involves the reduction of FEC to generate VC, LiF, and 1/2 H2, Scheme 3.1-ii. We note that LiF was generated nearly quantitatively via the Li-Nap reduction of FEC in our experiments. While we were unable to observe H2 generation, as the mass of H2 is below the detection limit of our GC-MS, the detection of H2 during the reduction of FEC was previously reported as part of the 4 electron reduction mechanism of FEC by Jung et al. 54 The rapid polymerization of the VC generated from We note that Shkrob et al. 53  and lithium divinylene dicarbonate (LDVD), these SEI decomposition products being suggested by prior theoretical investigations (see Supporting Information for our estimated NMR shifts of these predicted products). 20 They observed similar 13 C NMR peaks at ∼154 and ∼70−80 ppm in 13 C liquid NMR spectra of the SEI formed on the graphite electrode, dissolved in DMSO-d6, which they assigned to an oligomer of VC and/or poly(VC), the poly(VC) assignment in agreement with the ssNMR results here (see Table 3.2, peaks C and E); they also observed the distinctive 13 C ssNMR resonance at ∼100 ppm, seen in this study (see Table 3.2, peak D) which they assigned to an oligomer of VC. Here, we have assigned the 100 ppm resonance to a cross-linking site of poly(VC) (see Scheme 3.1-i, Figure 3.5). We have also observed broad peaks at 36 and 40 ppm, indicating a distribution of RCH2R′ environments (see Table 3.2, peaks F and G), assigned to the cross-linking site. Finally, in our previous study of the SEI composition on Si anodes formed in the presence of FEC and VC additives, 18 we have observed an FTIR adsorption peak at ∼1800 cm −1 increasing with additive concentration, the adsorption peak being assigned to poly(VC), as in this study.
The absence of the production of CO2 during the reduction of VC, seen by GC-MS, is in contrast to the literature. For example, the study by Ota et al., 47 which used pure VC as an electrolyte solvent, observed CO2 as the major gaseous product and a small amount of CO. Similarly, CO2 has been reported as the major gaseous product, when VC is used as an electrolyte additive. 45,49 The discrepancy is likely due to the reduction of CO2 by excess Li napthalenide to generate CO, Li2CO3, and Li2C2O4 (see Scheme 3.1iii). 55           inorganic components, and the organic phases were utilized for the analyses. Helium was used as carrier gas at a flow rate of 24 mL/min. The initial column temperature was 40°C and the temperature was ramped at 10°C/min to 200°C and held at that temperature for 2 minutes with the total run time of 18 minutes. The mass spectra obtained were compared to the NIST library to determine their molecular structures.
THF, Et2O (solvents) and naphthalene (starting material) and were the only volatile components present in the reaction mixtures. The gas analyses were performed by sampling the head spaces of the reaction mixtures in RB flasks with a 10 µL GC syringe.
Helium was used as the carrier gas at a flow rate of 1.5 mL/min. The initial column temperature was set to 40 °C, and the temperature was ramped at 1°C/min to 43°C and held at that temperature for 2 min with the total run time of 5 min. The mass spectra obtained were compared to the NIST library to determine their molecular structures.
IR-ATR spectra of the dried solid residues were acquired on a Bruker Tensor 27 spectrometer equipped with a germanium crystal in attenuated total reflectance (IR-ATR) mode. Samples were transferred using air-tight vials and the spectrometer was operated inside a nitrogen filled glovebox to avoid air exposure. Each spectrum was acquired with 128 scans from 700 cm -1 to 4000 cm -1 at the spectral resolution of 4 cm -1 . The data were processed and analyzed using the OPUS and Originlab software.
NMR spectra of the samples were collected with a Bruker Avance III 300 MHz NMR spectrometer at room temperature. The solids were dissolved in D2O in the nitrogen filled glovebox and 19 F, 31 P, 11 B, & 13 C NMR spectra of the solutions were acquired. The spectra were processed and analyzed using MestReNova 10.0.2.
XPS spectra of the dried precipitates were acquired using a Thermo Scientific K-alpha XPS. Samples were made into circular pellets with a press or stuck on a conductive carbon tape as a thin layer and transferred from the glovebox to the XPS chamber using a vacuum transfer module without exposure to air. An argon flood gun was used to avoid surface charge accumulation during sample analysis. The binding energy was corrected based on the C 1s of hydrocarbon at 284.8 eV. The data were processed and analyzed using the Thermo Avantage, XPS Peak 4.1 and the Originlab software.

Reduction of electrolyte salts
The

NMR analysis of the solids
The residual organic solvent insoluble solids have been analyzed via a combination of solution NMR spectroscopy in D2O, Infrared spectroscopy with attenuated total reflectance (IR-ATR), and X-ray photo electron spectroscopy. The residual solids have been dissolved in D2O for NMR analysis. While most of the residual solids dissolve in D2O, some of the solid does not readily dissolve. Some of the reduction products may react with water to generate subsequent hydrolysis products.
The dissolved solids were analyzed via a combination of 11 B, 13 C, 19 F, and 31 P NMR spectroscopy. Representative NMR spectra of the solids are provided in

FTIR analysis of the solids
In an effort to further understand the composition of the solids obtained from reduction, the reduction products of salts have been analyzed with IR-ATR. The IR-ATR spectra of the solids generated from the reduction of LiBOB and LiDFOB are provided in Figure 4.3. IR-ATR spectra of the residual solids for the other salts were also acquired, but the spectra were dominated by residual solvent and naphthalene since the decomposition products do not contain any functional groups which strongly absorb IR radiation, consistent with the observation of LiF, in NMR analysis.

INTRODUCTION
Lithium-ion batteries (LIB) are widely used as energy storage devices in portable electronics 1 and increasingly in electric vehicles due to their high energy density. However, LIB exhibit poor capacity retention at moderately elevated temperatures, 2 which is undesirable for many of the intended applications. Impedance growth and loss of cyclable lithium are reported to be the main contributors to capacity fade. [3][4][5] Lithium-ion batteries typically contain a graphite negative electrode, a lithiated transition metal oxide positive electrode, and an electrolyte composed of LiPF6 dissolved in a mixture of organic carbonate solvents. 6 The SEI (solid electrolyte interphase) is formed on the surface of the anode from the electrochemical reduction of the electrolyte and plays a crucial role in the long-term cyclability of LIB. 7 While the SEI has been reported to be a complex mixture of compounds, the initially formed components of the SEI are dominated by LiF, Li2CO3, lithium ethylene dicarbonate ((CH2OCO2Li)2, LEDC) and lithium alkyl carbonates (ROCO2Li). [7][8][9][10][11] The poor thermal stability of the SEI layer has been attributed to exothermal reactions between lithium alkyl carbonates and LiPF6. 12,13 While the relationship between capacity fade and SEI instability is clear, [12][13][14] and there have been some investigations of SEI component evolution 15,16 the mechanism of SEI component decomposition is complicated by the presence of many different components. The limited understanding of the evolution of the SEI components over time has significantly limited efforts to understand the mechanism of ion transport through the SEI via computational modeling. [17][18][19][20][21][22][23][24][25] A more comprehensive understanding of the decomposition reactions will aid computational scientists to develop a better physical understanding of the evolution of ion conducting mechanisms in the SEI and will help to improve the calendar life of lithium-ion batteries.
The SEI components lithium ethylene dicarbonate (LEDC), and lithium methyl carbonate (LMC) were independently synthesized by reduction of EC and DMC with lithium naphthalenide. 10 The decomposition reactions of Li2CO3, LEDC and LMC in the presence of LiPF6 have been investigated. The decomposition products were analyzed via a combination of nuclear resonance spectroscopy (NMR), infrared spectroscopy with attenuated total reflectance (IR-ATR), X-ray photoelectron spectroscopy (XPS) and gas chromatography with mass selective detection (GC-MS).

Experimental
Battery-grade DMC and LiPF6 were obtained from BASF. Li2CO3 was purchased from Sigma-Aldrich. LMC and LEDC were synthesized and purified as previously described. 26 All the reagents were stored in nitrogen filled glove box at room temperature and used without further purification.
The concentrations of LiPF6 and lithium carbonates were fixed at 0.65 mmol/mL in DMC. Samples were prepared inside the nitrogen filled glovebox. The samples were added to dry NMR tubes with DMSO-d6 capillaries, sealed with rubber septa, transferred out of the glovebox, flame sealed without air exposure, and analyzed by NMR spectroscopy. The samples were then stored at 55°C for 48 hours in an oil bath followed by analysis with NMR and GC-MS. Comparable samples were prepared on larger scale in glass ampules for the analysis of the solid resides. The solid residues were washed with DMC three times, dried overnight at room temperature, and analyzed with IR-ATR and XPS. Comparable samples were prepared in a Schlenk tube for overhead gas analyses by GC-MS. Comparable samples were prepared in stainless steel coin cells with comparable results to confirm that the glass containers play no role in the observed reactions.
NMR spectra of the samples were collected with a Bruker Avance III 300 MHz NMR spectrometer at room temperature before and after the high-temperature storage with and without proton decoupling. 19  GC-MS analyses were conducted with Agilent 6890-5973N GC equipped with an Agilent G973N mass selective detector. Liquid samples were diluted with dichloromethane, quenched with water to remove inorganic components, and the organic phase was utilized for the analysis. Helium was used as carrier gas at a flow rate of 24 mL/min. The initial column temperature was 40°C and the temperature was ramped at 10°C/min to 200°C and held at that temperature for 2 minutes with the total run time of 18 minutes. The gas analyses were performed by sampling the head spaces of a Schlenk tubes with a 10 µL GC syringe. Helium was used as carrier gas at a flow rate of 1.5 mL/min. The initial column temperature was set to 40 °C, and the temperature was ramped at 1 °C/min to 43 °C and held at that temperature for 2 min with the total run time of 5 min. The mass spectra obtained were compared to the NIST library to determine their molecular structures.
FTIR spectra of the dried solid residues were acquired on a Bruker Tensor 27 spectrometer equipped with a germanium crystal in attenuated total reflectance (IR-ATR) mode. Samples were transferred using air-tight vials and the spectrometer was operated inside a nitrogen filled glovebox to avoid air exposure. Each spectrum was acquired with 128 scans from 700 cm -1 to 4000 cm -1 at the spectral resolution of 4 cm -1 . The data were processed and analyzed using the OPUS and Originlab software.
XPS spectra of the dried precipitates were acquired using a Thermo Scientific K-alpha XPS. Samples were made into circular pellets with a press or stuck on a conductive carbon tape as a thin layer and transferred from the glovebox to the XPS chamber using a vacuum transfer module without exposure to air. An argon flood gun was used to avoid surface charge accumulation during sample analysis. The binding energy was corrected based on the C 1s of hydrocarbon at 284.8 eV. The data were processed and analyzed using the Thermo Avantage, XPS Peak 4.1 and the Originlab software.

Reactivity of lithium carbonates with LiPF6 in DMC.
In an effort to better understand the stability and decomposition products of components of the anode SEI with the electrolyte, the reactions of three different lithium carbonates, Li2CO3, LMC, and LEDC with LiPF6 have been investigated in DMC.
LEDC and LMC have been independently prepared via the chemical reduction by lithium naphthalenide while Li2CO3 is commercially available. [9][10][11] The stability of Li2CO3 has been investigated in DMC with and without added

NMR Spectroscopy of the solutions
All of the samples were analyzed by 19 F and 31 P NMR spectroscopy before and after storage at elevated temperature. The NMR spectra after storage at 55 o C are provided in Figure 5.1 and the spectral data listed in

FTIR Spectroscopy of the solid residues
The IR-ATR spectra of Li2CO3, LEDC, and LMC before and the residue after reaction of the different lithium carbonates with LiPF6 in DMC are provided in Figure   5.2. The IR spectrum of Li2CO3 contains two strong peaks centered at 1490 and 1450 cm -1 and a weak peak at 858 cm -1 . After storage of Li2CO3 in the presence of DMC for 2 days at 55°C, the insoluble residue was isolated. The IR-ATR spectrum exhibits essentially the same absorbance patterns as of pure Li2CO3 starting material, indicating little reactivity between lithium carbonate and DMC under the storage conditions. However, after storage of Li2CO3 in the presence of 0.65 M LiPF6 in DMC under similar conditions, the residue does not exhibit any peaks associated with Li2CO3 and instead contain several weak absorbances at 1300, 1162, and 758 cm -1 . The structure of the compound associated with these IR absorbances is unclear, but the absence of Li2CO3 in the residual solid is very clear.
Similarly, the characteristic peak of lithium alkyl carbonates, corresponding to C=O bonds, is observed at 1650 cm -1 . The residues obtained from samples containing LMC or LEDC in DMC exhibit IR absorptions similar to the starting material, whereas IR-ATR spectra of the residues obtained from samples containing LMC or LEDC and LiPF6 in DMC contain no peaks associated with LMC or LEDC, respectively. Absorptions are observed in the 720-740 cm -1 region which remain unidentified, but the absence of the lithium alkyl carbonates is clear. The absence of LMC or LEDC in the precipitate suggests that the majority of the lithium carbonates react with LiPF6 in DMC during storage at 55 o C.

X-ray photoelectron spectroscopy of the residues
The XPS spectra of the residues after reaction of the different lithium carbonates with LiPF6 in DMC are provided in Figure 5.3. The XPS spectrum of the residue from the reaction of Li2CO3 with LiPF6 has a very high concentration of Li and F, 38 and 42 %, respectively, and very low concentrations of C, O, and P 8, 6, and 6 % respectively.
The F 1s spectrum is dominated by a peak at 685 eV and the Li 1s spectrum is dominated by a peak at 56.4 eV coupled with the ~ 1:1 ratio of F to Li suggest that the residue is predominantly LiF. The F1s spectrum also contains a small shoulder at 687.5 eV along with related O 1s and P 2p peaks at 533 and 136 eV, respectively, consistent with the presence of a low concentration of LixPFyOz. The C 1s spectrum is dominated by a peak at 285 eV which likely results primarily from residual naphthalene or hydrocarbon contamination. There is no evidence for any residual Li2CO3 at ~290 eV in the C1s XPS spectra, consistent with the IR-ATR spectra.
The The reaction of Li2CO3 with LiPF6 results in the quantitative decomposition of the Li2CO3. A single gas, CO2, is observed by GC-MS. The residual solid from the reaction is predominantly LiF, as supported by XPS and IR-ATR. The solution phase contains a single decomposition product, F2PO2Li consistent with previous reports. 30 The low concentration of P in the residual solid is consistent with the generation of a soluble P containing species, F2PO2Li. The generation of soluble P containing species is the likely reason that the composition of the SEI typically has a much higher ratio of F:P than the 6:1 expected for LiPF6. When a 1:1 stoichiometry of Li2CO3 to LiPF6 is used, ~50 % of the LiPF6 is converted to F2PO2Li and LiF. This reaction is consistent with the equation 1.