Determining the Composition of Carbonate Solvent Systems Used in Lithium-Ion Batteries without Salt Removal

: In this work, two methods were investigated for determining the composition of carbonate solvent systems used in lithium-ion (Li-ion) battery electrolytes. One method was based on comprehensive two-dimensional gas chromatography with electron ionization time-of-ﬂight mass spectrometry (GC × GC/EI TOF MS), which often enables unknown compound identiﬁcation by their electron ionization (EI) mass spectra. The other method was based on comprehensive two-dimensional gas chromatography with ﬂame ionization detection (GC × GC/FID). Both methods were used to determine the concentrations of six different commonly used carbonates in Li-ion battery electrolytes (i.e., ethylene carbonate (EC), propylene carbonate (PC), dimethyl carbonate (DMC), diethyl carbonate (DEC), ethyl methyl carbonate (EMC), and vinylene carbonate (VC) in model compound mixtures (MCMs), single-blind samples (SBS), and a commercially obtained electrolyte solution (COES). Both methods were found to be precise (uncertainty < 5%), accurate (error < 5%), and sensitive (limit of detection <0.12 ppm for FID and <2.7 ppm for MS). Furthermore, unlike the previously reported methods, these methods do not require removing lithium hexaﬂuorophosphate salt (LiPF 6 ) from the sample prior to analysis. Removal of the lithium salt was avoided by diluting the electrolyte solutions prior to analysis (1000-fold dilution) and using minimal sample volumes (0.1 µ L) for analysis.


Introduction
Lithium-ion (Li-ion) batteries are the predominant energy storage and conversion device in various applications, such as portable consumer electronics, electric vehicles (EVs), grid storage, space applications, and military applications [1][2][3].Li-ion batteries offer several advantages over other rechargeable batteries, such as high energy density and power density, low self-discharge rate, and long life [4].However, decomposition reactions occur within Li-ion batteries under certain operating conditions, which negatively impact the performance of the battery and impose safety hazards by generating toxic and flammable compounds.For instance, under abuse conditions, decomposition reactions generate a significant amount of heat and gaseous products, leading the cell into thermal runaway.Thermal runaway is frequently accompanied by the release of flammable and toxic gases during the venting process and afterward.The rates and pathways of these decomposition reactions, the composition and amount of the gaseous products, and heat generation rates are affected by the solvent(s) used within the electrolyte solution, which can vary significantly [5].The most common electrolytes used in Li-ion batteries comprise a conducting salt such as lithium hexafluorophosphate (LiPF 6 ) dissolved in a mixture of organic carbonate solvents.
Energies 2022, 15, 2805 2 of 14 Accurate analytical methods for determining the composition of solvent systems used in Li-ion batteries are needed to better understand decomposition processes occurring within Li-ion batteries and their effects on battery performance and safety.Recovering organic carbonate compounds during recycling may also benefit from accurate analytical methods to verify the solvent composition [6].Several studies have implemented different analytical methods to identify and quantify electrolyte components.For example, in a qualitative study, Horsthemke et al. [7] used gas chromatography combined with mass spectrometry (GC/MS) to identify electrolyte components and aging products for various commercially available Li-ion cells.A headspace solid-phase microextraction (SPME) technique was used to prevent the injection of salts into the GC column [7].On the other hand, Ellis et al. [8] have used Fourier-transform infrared (FTIR) spectroscopy to determine the concentrations of LiPF 6 and common solvents in Li-ion battery electrolytes.An FTIR spectral database of known solutions with known concentrations was first created.Machine learning was then used to quantify unknown electrolyte solutions by interpolating data in the FTIR spectral database [8].Further, Gachot et al. [9] have used the GC/MS technique to identify the compounds within the Li-ion battery electrolytes after cycling and heating.A postcapillary column was used to preserve the ion source from HF and the bleeding of the first column.The same research group has coupled gas chromatography with mass spectrometry and Fourier transform infrared spectroscopy (GC/MS-FTIR) to identify electrolyte compositions.This method was used to characterize volatile compounds released during thermal runaway and gaseous and soluble volatile products in a swollen battery [10].Schultz et al. [11] have employed high-performance liquid chromatography (HPLC) coupled to tandem mass spectrometry (LC/MS/MS) for the identification and quantification of components of Li-ion battery electrolytes.The method was then used to quantify chemical species produced under different aging processes [11].Petibon et al. [12] have developed a semi-quantitative approach based on the GC/MS technique for the determination of the consumption and transesterification of additives and solvents after formation cycles and storage at high potential.Liquid-liquid extraction was used to remove LiPF 6 prior to analysis [12].However, liquid-liquid extraction is time-consuming and may introduce additional uncertainty in the measurement.Terborg et al. [13] have employed GC/MS and GC/FID to identify organic solvents in Li-ion battery electrolytes.Thompson et al. [14] have used GC/MS to analyze the changes in the composition of the carbonates within the battery electrolytes after cycling at different cutoff potentials.The salt was extracted from the samples using liquid-liquid extraction and centrifuge [14].Weber et al. [15] have used a GC-MS-based analytical method to identify phosphorous/organic degradation products in aged cells that are known to be toxic.Quantification of dimethyl phosphorofluoridate (DMPF) and diethyl phosphorofluoridate (DEPF) has been done using calibration curves [15].Finally, several previous studies have used gas chromatography to investigate venting behavior and venting gas composition of lithium-ion cells during thermal runaway [10, [16][17][18][19].
Most of the previously reported techniques required removing LiPF 6 salt from the samples prior to analysis because it is well known that the conductive salt reacts with the silicon coating of the GC column and results in excessive column bleeding [7].However, salt removal is time-consuming, introduces additional costs, and reduces the precision of the method.
Comprehensive two-dimensional gas chromatography (GC×GC) coupled with FID (GC×GC/FID), or mass spectrometry detection (GC×GC/EI TOF MS) is a powerful analytical technique due to its high resolution and high peak capacity.MS can often be used for the identification of unknown compounds based on a comparison of their electron ionization (EI) mass spectra to mass spectral libraries, while FID is often used for compound quantification because of the similar response of FID to different types of compounds.In this paper, two methods are described for the determination of the compositions of solvent systems used in Li-ion battery electrolytes, one based on GC×GC/FID and the other GC×GC/EI TOF MS.GC×GC has been widely used to analyze complex mixtures related to food, petrochemical, environmental, biomedical, hardware, and software industries [20].However, to the best of the authors' knowledge, determination of the compositions of the solvent systems used in Li-ion battery electrolytes has not been performed based on GC×GC/FID or GC×GC/EI TOF MS.The purpose of the present work was to determine the accuracy and precision of these two analytical techniques when analyzing compounds found in electrolyte solutions of Li-ion batteries.To quantify and compare the performance of the two analytical techniques, the limit of detection (LOD), limit of quantification (LOQ), uncertainty, and repeatability were determined.

Carbonate Mixtures and Li-Ion Battery Electrolyte
Various model mixtures of known carbonates were used to determine the accuracy and precision of the GC×GC/FID and GC×GC/EI TOF MS methods.Model compound mixtures (MCMs) with three different compositions were prepared and analyzed (MCM #1, MCM #2, and MCM #3; Table 1).One MCM was prepared and analyzed on three different days (MCM #1A MCM #2A, and MCM #3A; Table 1) to determine the intraday repeatability of the methods.A commercially obtained electrolyte solution (COES) of 1.0 M LiPF 6 solution in 50:50 (v/v) (±5%) EC and DMC was purchased from Sigma-Aldrich and analyzed using both methods (Table 1).Two single-blind samples (SBSs) were also analyzed (using both methods) that were prepared by members of a different laboratory with compositions that were revealed only after they were analyzed (SBS; Table 1).SBS #1 was prepared with pure EC (≥99%), EMC (≥99.9%),VC (≥99.5%), and isopropanol (IPA; 99.8%).SBS #2 was prepared from an electrolyte solution in 50:50 (v/v) (±5%) EC and EMC, an electrolyte solution in 50:50 (v/v) (±5%) EC and DEC, and pure VC (≥99.5%).SBS #1 and SBS#2 were shipped (~3 h) then stored (up to 6 months) prior to preparation and injection.Samples were prepared by diluting 10 µL of each mixture in 10 mL DCM or acetone.Each sample was injected 10 times into both instruments with an injection volume of 0.1 µL and a split ratio of 24.

GC×GC/FID Instrument Operating Conditions and Parameters
All GC×GC/FID measurements were performed using an instrument composed of a 7890B GC oven (Agilent, Santa Clara, CA, USA), a 7683 series autosampler (Hewlett-Packard, Palo Alto, CA, USA), a 7683B series injector (Agilent, Santa Clara, CA, USA), an FID (Agilent, Santa Clara, CA, USA), and a quad-jet dual-stage thermal modulator (LECO Corporation, Saint Joseph, MI, USA) cooled with liquid nitrogen.The capillary column contained within the primary oven (primary column) was a midpolar 30 m DB-17MS column.The capillary column contained within the secondary oven (secondary column) was a nonpolar 0.8 m DB-1MS column.A 0.3 m guard column (Ultimate Plus Deactivated Fused Silica) was used between the secondary column and the FID.All columns had a 0.25 mm inner diameter and a 0.25 µm film thickness (Agilent, Santa Clara, CA, USA).Ultrahigh purity (99.9999%) helium was used as the carrier gas, with a flow rate of 1.5 mL/min.Primary oven temperature was maintained at 40 • C for 1.0 min, increased to 200 • C at a temperature ramp rate of 3 • C/minute, and then held constant for five more minutes.The temperature offsets (relative to the primary oven) for the secondary oven and modulator were +5 • C and +15 • C, respectively.The modulation period was 3.0 s, with a hot pulse duration of 0.9 s.The temperatures of the FID and the inlet of the injection port were 300 • C and 280 • C, respectively.The acquisition rate of the FID was 200 Hz, and the acquisition solvent delay was 150 s.The injection volume was 0.1 µL.The split-ratio for the inlet of the injection port was 24.The instrument was operated using ChromaTOF software (version 4.71.0.0).A S/N threshold of 50 was used for data processing of all chromatograms.

GC×GC/EI TOF MS Instrument Operating Conditions and Parameters
All GC×GC/EI TOF MS measurements were performed using a Pegasus GC-HRT 4D (LECO, Saint Joseph, MI, USA), which was composed of a 7890B GC oven (Agilent, Santa Clara, CA, USA), an Agilent injector (G4513A), a quad-jet dual-stage thermal modulator (LECO, Saint Joseph, MI, USA) cooled with liquid nitrogen, an electron ionization source (LECO, Saint Joseph, MI, USA), and a high-resolution time-of-flight mass spectrometer (LECO, Saint Joseph, MI, USA).The capillary column contained within the primary oven (primary column) was a polar 60.0 m ZB-35HT column.The capillary column contained within the secondary oven (secondary column) was a nonpolar 1.2 m ZB-1HT column.All columns had a 0.25 mm inner diameter and a 0.25 µm film thickness (Phenomenex, Torrance, CA, USA).Ultra-high purity (99.9999%) helium was used as a carrier gas with a flow rate of 1.25 mL/minute.Primary oven temperature was maintained at 40 • C for 1.0 min, increased to 200 • C at a temperature ramp rate of 3 • C/min, and then held constant for five more minutes.The temperature offsets (relative to the primary oven) for the secondary oven and modulator were +5 • C and +15 • C, respectively.The modulation period was 3.0 s, with a hot pulse duration of 0.9 s.The temperatures of the transfer line and the inlet of the injection port were 300 • C and 280 • C, respectively.The temperature of the ion source was maintained at 200 • C. The kinetic energy of the electrons used for electron ionization (EI) was 70 eV.The acquisition solvent delay was 150 s.The injection volume was 0.1 µL.The split-ratio for the inlet of the injection port was 24.The instrument was operated using ChromaTOF software (version 1.90.60.0.43266).A S/N threshold of 50 was used for data processing of all chromatograms.The ChromaTOF software automatically identified compounds by comparing their EI mass spectra to the EI mass spectra in the Wiley (2011) and NIST (2011) databases.The match factor threshold, which is the minimum match factor an EI mass spectrum must exhibit with a database entry for a compound to be identified, was 800.

Calibration Curves and Sample Preparation
Calibration curves were established for each carbonate (DMC, DEC, PC, EMC, VC, and EC) by preparing a stock solution that had the same concentration of each carbonate as the pure carbonate solution prepared for analysis (discussed in Sections 2.3 and 2.4), which corresponded to 100% (v/v), and diluting the stock solution in DCM or acetone to prepare solutions corresponding to 1, 5, 10, 20, 40, 60, and 80% (v/v) used for calibration.The stock solution was prepared by dissolving 200 µL of DMC, DEC, PC, EMC, VC, and 264.2 mg of EC (equivalent to 200 µL considering ρ EC = 1.321 g/cm 3 ) in 200 mL of acetone.Each calibration curve solution was then analyzed with GC×GC/FID and GC×GC/EI TOF MS according to Sections 2.3 and 2.4, respectively.The concentration of each carbonate in each calibration curve solution is shown in Table 2.In addition, each of the calibration curve solutions was analyzed five times with GC×GC/FID and five times with GC×GC/EI TOF MS.SBSs, MCMs, and COES were analyzed 10 times each with GC×GC/FID and ten times each with GC×GC/EI TOF MS.

Identification and Quantitation of the Carbonates
Testing the ability of GC×GC/FID to identify the carbonates was carried out by preparing samples of individual carbonates, i.e., 10 µL of DMC, DEC, PC, EMC, VC, and 13.21 mg of EC in 10 mL acetone, and injecting them separately into the GC×GC system to create a map for individual carbonates based on their first and second-retention times.Testing the GC×GC/EI TOF MS method was based on the comparison of the measured EI mass spectra to the EI mass spectral library.Testing the quantitation of DMC, DEC, PC, EMC, VC, and EC was carried out by exporting the peak areas for each carbonate in the calibration solutions after data processing in GC×GC/FID and GC×GC/EI TOF MS to Microsoft Excel 365.The slope and intercept associated with the calibration curve of each compound were then determined using conventional linear regression analysis.

LOD and LOQ
The statistical LOD and LOQ for each compound was determined based on the standard deviation (σ) of the intercept over the slope (M) in the calibration curve data by using Equations ( 1) and (2), respectively [21].

Analysis of Column Bleeding
Aiming to determine whether the samples containing LiPF 6 salt cause excessive column bleeding, an experiment was carried out by injecting a blank acetone sample (considered as the baseline sample) before and after five injections of COES and SBS #2.All chromatograms were inspected for any visual signs of column bleed, changes in peak shapes, or changes in retention times.

Uncertainty Analysis
The uncertainty analysis was carried out similar to a method proposed by Moffat [22].In the present work, an uncertainty analysis was performed in two steps.First, the uncertainty of the calibration curves was quantified.Second, the uncertainty of sample concentrations was quantified.Figure 1 presents the calibration curve for VC with a slope of m and an intercept of b.The slope and the intercept depend on the concentrations of the compound and the measured GC peak areas.However, both the concentrations and measured peak areas are subject to experimental uncertainties.In other words, the concentrations (x-axis) are subject to fixed errors (B) due to the sample preparation process, and the measured peak areas (y-axis) are subject to random errors (S) due to the repeated measurements.Thus, there is an overall uncertainty associated with the slope and the intercept of each calibration curve.

Analysis of Column Bleeding
Aiming to determine whether the samples containing LiPF6 salt cause excessive column bleeding, an experiment was carried out by injecting a blank acetone sample (considered as the baseline sample) before and after five injections of COES and SBS #2.All chromatograms were inspected for any visual signs of column bleed, changes in peak shapes, or changes in retention times.

Uncertainty Analysis
The uncertainty analysis was carried out similar to a method proposed by Moffat [22].In the present work, an uncertainty analysis was performed in two steps.First, the uncertainty of the calibration curves was quantified.Second, the uncertainty of sample concentrations was quantified.Figure 1 presents the calibration curve for VC with a slope of m and an intercept of b.The slope and the intercept depend on the concentrations of the compound and the measured GC peak areas.However, both the concentrations and measured peak areas are subject to experimental uncertainties.In other words, the concentrations (x-axis) are subject to fixed errors (B) due to the sample preparation process, and the measured peak areas (y-axis) are subject to random errors (S) due to the repeated measurements.Thus, there is an overall uncertainty associated with the slope and the intercept of each calibration curve.
where = {∑( ) } was the overall fixed error arising from preparation of the solution with concentration of i.The subscript n referred to the number of pieces of equipment used in the process of preparing the i solution.̅ = √ was the mean precision index calculated as the ratio of Si, which was the standard deviation of the i solution, and N, which was the number of measurements made.The fixed and random uncertainties of the slope (B Rm and S Rm , respectively) and the intercept (B Rb and S Rb , respectively) of each calibration curve were determined with Equations ( 3) and (4), similar to Moffat et al. [22]: (3) where was the overall fixed error arising from preparation of the solution with concentration of i.The subscript n referred to the number of pieces of equipment used in the process of preparing the i solution.
was the mean precision index calculated as the ratio of S i , which was the standard deviation of the i solution, and N, which was the number of measurements made.
From the calibration curves, the concentration of each compound can be determined as x = (y − b)/m.Note that y is the measured peak area with a precision index (random Energies 2022, 15, 2805 7 of error), b is the intercept with fixed errors, and m is the slope with fixed errors.Therefore, the fixed and random error associated with concentration, x, can be determined as follows: The total uncertainty associated with concentration of a compound, x, can be written as:

Testing the Accuracy of Identification and Quantitation of Carbonates by Using GC×GC/FID
Carbonate compounds were identified using GC×GC/FID, based on their first-and second-dimension retention times (1D-RT and 2D-RT) of each carbonate.Figure 2 shows a three-dimensional chromatogram obtained by using GC×GC/FID for the 100% calibration solution.All the compounds were completely separated by GC×GC, allowing each carbonate to be detected without interference from the others.Table 3 summarizes the retention times as well as the slope and the Y-intercept for the calibration curve of each carbonate.Linear correlation coefficients (R 2 ) were found to be greater than 0.977 for all cases.
From the calibration curves, the concentration of each compound can be determined as x = (y − b)/m.Note that y is the measured peak area with a precision index (random error), b is the intercept with fixed errors, and m is the slope with fixed errors.Therefore, the fixed and random error associated with concentration, x, can be determined as follows: The total uncertainty associated with concentration of a compound, x, can be written as:

Testing the Accuracy of Identification and Quantitation of Carbonates by Using GC×GC/FID
Carbonate compounds were identified using GC×GC/FID, based on their first-and second-dimension retention times (1D-RT and 2D-RT) of each carbonate.Figure 2 shows a three-dimensional chromatogram obtained by using GC×GC/FID for the 100% calibration solution.All the compounds were completely separated by GC×GC, allowing each carbonate to be detected without interference from the others.Table 3 summarizes the retention times as well as the slope and the Y-intercept for the calibration curve of each carbonate.Linear correlation coefficients (R 2 ) were found to be greater than 0.977 for all cases.

LOD and LOQ of the GC×GC/FID Method
As shown in Figure 3, the minimum concentration of all the carbonate solvents, i.e., DMC, EMC, DEC, VC, PC, and EC, visually detected by the GC×GC/FID method was 0.2% at a signal to noise ratio of 50, corresponding to injections of 0.084, 0.080, 0.077, 0.110, 0.098, and 0.106 ppm in acetone for DMC, EMC, DEC, VC, PC, and EC, respectively.

LOD and LOQ of the GC×GC/FID method
As shown in Figure 3, the minimum concentration of all the carbonate solvents, i.e., DMC, EMC, DEC, VC, PC, and EC, visually detected by the GC×GC/FID method was 0.2% at a signal to noise ratio of 50, corresponding to injections of 0.084, 0.080, 0.077, 0.110, 0.098, and 0.106 ppm in acetone for DMC, EMC, DEC, VC, PC, and EC, respectively.The statistical LOD and LOQ were determined using Equations ( 1) and ( 2), respectively.The visual LOD, statistical LOD, and statistical LOQ are summarized in Table 4. EC and EMC have the highest and lowest LOD and LOQ, respectively.These LODs are much lower than the LODs reported for one-dimensional gas chromatography with FID

Testing the Accuracy of the Identification and Quantitation of Carbonates by Using GC×GC/EI TOF MS
Figure 4 shows the electron ionization (EI) mass spectra obtained by using GC×GC/EI TOF MS for DMC, EMC, DEC, VC, PC, and EC in the 100% calibration solution.The similarity factors upon comparison of the measured mass spectra to mass spectral libraries for DMC, EMC, DEC, VC, PC, and EC were 864, 808, 941, 888, 895, and 929, respectively.Table 5 summarizes the retention times as well, as the slopes and the y-intercepts of the calibration plots for each carbonate.Correlation coefficients for the linear regression of GC×GC/EI TOF MS peak area to compound concentration ranged from 0.9533 for DMC to 0.9990 for PC.The statistical LOD and LOQ were determined using Equations ( 1) and ( 2), respectively.The visual LOD, statistical LOD, and statistical LOQ are summarized in Table 4. EC and EMC have the highest and lowest LOD and LOQ, respectively.These LODs are much lower than the LODs reported for one-dimensional gas chromatography with FID [13].The LOQ for EMC, DEC, DMC, VC, PC, and EC were 3.04, 3.60, 4.23, 11.43, 12.13, and 16.41 ppm, respectively.4 shows the electron ionization (EI) mass spectra obtained by using GC×GC/EI TOF MS for DMC, EMC, DEC, VC, PC, and EC in the 100% calibration solution.The similarity factors upon comparison of the measured mass spectra to mass spectral libraries for DMC, EMC, DEC, VC, PC, and EC were 864, 808, 941, 888, 895, and 929, respectively.Table 5 summarizes the retention times as well, as the slopes and the y-intercepts of the calibration plots for each carbonate.Correlation coefficients for the linear regression of GC×GC/EI TOF MS peak area to compound concentration ranged from 0.9533 for DMC to 0.9990 for PC.

LOD and LOQ of the GC×GC/EI TOF MS Method
Carbonates in the solutions with the lowest concentrations, i.e., 0.1%, 0.2%, 0.4%, 0.6%, and 0.8% (v/v) were not detected by GC×GC/EI TOF MS.The minimum concentration (v/v) detected was for the 1% calibration solution (corresponding to injection of 0.42, 0.40, 0.39, and 0.55 ppm of DMC, EMC, DEC, and VC, respectively) or 5% (v/v) (corresponding to injection of 2.44 and 2.66 ppm of PC, and EC, respectively).The LOD and LOQ for the GC×GC/EI TOF MS method are summarized in Table 6.The statistical LOD was the greatest for DMC, EMC, DEC, and VC and lower for PC and EC.Statistical LOD was lower than the visual LOD for PC and EC.No solutions were prepared at concentrations between 1-5% for PC and EC.The LODs and LOQs were generally lower for the GC×GC/FID method than they were for the GC×GC/EI TOF MS method.

LOD and LOQ of the GC×GC/EI TOF MS Method
Carbonates in the solutions with the lowest concentrations, i.e., 0.1%, 0.2%, 0.4%, 0.6%, and 0.8% (v/v) were not detected by GC×GC/EI TOF MS.The minimum concentration (v/v) detected was for the 1% calibration solution (corresponding to injection of 0.42, 0.40, 0.39, and 0.55 ppm of DMC, EMC, DEC, and VC, respectively) or 5% (v/v) (corresponding to injection of 2.44 and 2.66 ppm of PC, and EC, respectively).The LOD and LOQ for the GC×GC/EI TOF MS method are summarized in Table 6.The statistical LOD was the greatest for DMC, EMC, DEC, and VC and lower for PC and EC.Statistical LOD was lower than the visual LOD for PC and EC.No solutions were prepared at concentrations between 1-5% for PC and EC.The LODs and LOQs were generally lower for the GC×GC/FID method than they were for the GC×GC/EI TOF MS method.

Testing the Accuracy of the Quantitation of MCMs
The intraday repeatability of quantitation was determined for both methods by measuring three samples that were prepared in the same way but on different days.The compositions determined for MCM #1A, MCM #1B, and MCM #1C are shown in Table 7.The results for MCM #1A-C were similar.The average volume percentage determined using GC×GC/FID for DMC, DEC, EMC, VC, and PC were 19.6 ± 2.4, 20.5 ± 1.5, 19.3 ± 0.8, 19.3 ± 3.1, and #20.9 ± 1.9% (v/v), respectively.The average volume percentage determined using GC×GC/EI TOF MS for DMC, DEC, EMC, VC, and PC were 17.0 ± 1.5, 20.2 ± 1.3, 21.6 ± 1.6, 20.1 ± 1.4, and 19.8 ± 1.9% (v/v), respectively.Measurements made on the same day were similar to each other (average relative standard deviation = 3.9%).In general, both instruments predicted the volume percentages of the carbonates with good accuracy.The largest error for the GC×GC/FID method was observed for PC, and the largest error observed for the GC×GC/EI TOF MS method was associated with DMC.Additional MCMs were also analyzed using GC×GC/FID.The compositions determined for MCM #2 and MCM #3 are shown in Figure 5. MCM #2 was 20, 20, and 60% EMC, VC, and DMC, respectively.MCM #3 was 60, 20, and 20% EC, DEC, and PC, respectively.The uncertainties associated with the measurement of each compound and the actual concentration are also shown in Figure 5.The percentage errors of the GC×GC/FID method for EMC, VC, and EC in MCM #2, were 1.1, 0.7, 0.4%, respectively (average percentage error = 0.7%).The percentage errors of the GC×GC/FID method for EC, DEC, and PC in MCM #3, were 2.6, 3.1, 2.8%, respectively (average percentage error = 2.8%).Furthermore, the greatest uncertainty associated with the measurements was 4.1% for DMC in MCM #2, and 4.3% for EC in MCM #3.The compositions determined for SBSs and COES are shown in Figure 6.The errors in the expected volume percentages of the electrolytes in SBS #2 (which was prepared with a commercially obtained electrolyte solution as described in the experimental section) and the COES were especially large because of large uncertainties mentioned in the product specifications from the manufacturer.Both methods were very accurate.The percentage

Testing the Accuracy of Quantitation of SBSs and COES
The compositions determined for SBSs and COES are shown in Figure 6.The errors in the expected volume percentages of the electrolytes in SBS #2 (which was prepared with a commercially obtained electrolyte solution as described in the experimental section) and the COES were especially large because of large uncertainties mentioned in the product specifications from the manufacturer.Both methods were very accurate.The percentage errors of the GC×GC/FID method for EC, EMC, and VC in SPS #1 were 0.8, 1.0, and 0.4%, respectively (average percentage error = 0.8%).The percentage errors of the GC×GC/FID method for EC, DEC, EMC, and VC in SPS #2 were 1.2, 2.0, 6.4, and 0.3%, respectively (average percentage error = 2.5%).The percentage errors of the GC×GC/FID method for EC and DMC in COES were 2.5 and 5.7%, respectively (average percentage error = 4.1%).The percentage errors of the GC×GC/EI TOF MS method for EC, EMC, and VC in SPS #1 were 0.2, 2.4, and 1.9%, respectively (average percentage error = 1.5%).The percentage errors of the GC×GC/EI TOF MS method for EC, DEC, EMC, and VC in SPS #2 were 1.8, 2.9, 2.6, and 0.2%, respectively (average percentage error = 1.9%).The percentage errors of the GC×GC/EI TOF MS method for EC and DMC in COES were 11 and 2%, respectively (average percentage error = 6.5%).Furthermore, the maximum uncertainty associated with the measurements using GC×GC/FID was 5.5% for DMC in COES (average uncertainty = 2.43%), and the maximum uncertainty of the measurements using GC×GC/EI TOF MS was 4.4% for EMC in SBS #1 (average uncertainty = 2.13%).

Analysis of Column Bleeding before and after Salt Injection
To determine whether the samples containing LiPF 6 salt cause excessive column bleeding, an experiment was conducted by injecting a blank acetone sample (considered as the baseline sample) before and after five injections of COES and SBS #2. Figure 7 shows the GC×GC-FID/chromatograms of acetone before and after injection of salt-containing samples.No changes were observed in the amount of column bleeding, noise, or retention times before and after samples that contained LiPF 6 were analyzed, which suggests that the effects of the LiPF 6 on the GC columns were negligible.The solvent-bleeding also remained unchanged.As shown in Figure 8, no additional column bleeding or changes in retention times or peak shapes were observed for COES or SBS #2, which further suggests that the effects of the LiPF6 salt were negligible.This is likely because very little salt was introduced into the instrument (samples were diluted 1000-fold prior to analysis, and only 0.1 µL of the diluted samples was analyzed).Additionally, inlet liners with glass wool were used at the inlet of the GC×GC instruments to trap nonvolatile compounds, such as LiPF6, and the liners were replaced when needed.As shown in Figure 8, no additional column bleeding or changes in retention times or peak shapes were observed for COES or SBS #2, which further suggests that the effects of the LiPF 6 salt were negligible.This is likely because very little salt was introduced into the instrument (samples were diluted 1000-fold prior to analysis, and only 0.1 µL of the diluted samples was analyzed).Additionally, inlet liners with glass wool were used at the inlet of the GC×GC instruments to trap nonvolatile compounds, such as LiPF 6 , and the liners were replaced when needed.

Conclusions
Two methods are described to determine the composition of carbonate solvent systems commonly used in Li-ion battery electrolyte solutions, one based on GC×GC/FID and another based on GC×GC/EI TOF MS.The accuracy and precision of the methods were tested using various model compound mixtures (MCMs), a commercially obtained electrolyte solution (COES), and single-blind samples (SBSs).Both methods were found to be accurate, precise, and sensitive.The high sensitivity of the methods allowed the use of high dilution factors and low injection volumes, eliminating the need for salt removal.No adverse effects on column bleeding, retention times, or GC peak shapes were observed while analyzing samples containing LiPF6 salt.The methods employed in this study will be used in future studies to investigate changes in the composition of electrolyte solutions from Li-ion batteries over time, after calendar aging, and after cycle aging.

Conclusions
Two methods are described to determine the composition of carbonate solvent systems commonly used in Li-ion battery electrolyte solutions, one based on GC×GC/FID and another based on GC×GC/EI TOF MS.The accuracy and precision of the methods were tested using various model compound mixtures (MCMs), a commercially obtained electrolyte solution (COES), and single-blind samples (SBSs).Both methods were found to be accurate, precise, and sensitive.The high sensitivity of the methods allowed the use of high dilution factors and low injection volumes, eliminating the need for salt removal.No adverse effects on column bleeding, retention times, or GC peak shapes were observed while analyzing samples containing LiPF 6 salt.The methods employed in this study will be used in future studies to investigate changes in the composition of electrolyte solutions from Li-ion batteries over time, after calendar aging, and after cycle aging.

Figure 2 .
Figure 2. Three-dimensional chromatogram exhibiting effective separation of the carbonates in the 100% calibration solution when using the GC×GC/FID system.

Figure 2 .
Figure 2. Three-dimensional chromatogram exhibiting effective separation of the carbonates in the calibration solution when using the GC×GC/FID system.

Figure 4 .
Figure 4. 70 eV EI mass spectra obtained by using GC×GC/EI TOF MS for DMC, EMC, DEC, VC, PC, and EC.

Figure 4 .
Figure 4. 70 eV EI mass spectra obtained by using GC×GC/EI TOF MS for DMC, EMC, DEC, VC, PC, and EC.

Figure 7 .
Figure 7. GC×GC/FID chromatograms of acetone before (a) and after (b) ten samples containing LiPF6 were analyzed.

Table 1 .
Composition of various mixtures analyzed by GC×GC/FID and GC×GC/EI TOF MS methods.

Table 2 .
Concentrations of carbonates in calibration solutions.

Table 3 .
GC×GC/FID calibration results including the retention times and linear correlation parameters for common organic solvents found in Li-ion batteries.

Table 3 .
GC×GC /FID calibration results including the retention times and linear correlation parameters for common organic solvents found in Li-ion batteries.

Table 4 .
Limit of detection (LOD) and limit of quantification (LOQ) of DMC, EMC, DEC, VC, PC, and EC when using GC×GC/FID.

Table 4 .
Limit of detection (LOD) and limit of quantification (LOQ) of DMC, EMC, DEC, VC, PC, and EC when using GC×GC/FID.

Table 5 .
GC×GC/EI TOF MS calibration results, including the retention times and linear correlation parameters for common organic solvents found in Li-ion batteries.

Table 5 .
GC×GC /EI TOF MS calibration results, including the retention times and linear correlation parameters for common organic solvents found in Li-ion batteries.

Table 6 .
Limit of detection (LOD) and limit of quantification (LOQ) for DMC, EMC, DEC, VC, PC, and EC obtained using the GC×GC/EI TOF MS method.

Table 7 .
Volume fraction of the components in MCM #1A-C measured using GC×GC/FID and GC×GC/EI TOF/MS.The expected concentration of each carbonate was 20 ± 1.5% for MCM#1A-C.