Lithium Salt Dissociation in Diblock Copolymer Electrolyte Using Fourier Transform Infrared Spectroscopy

Polymer electrolytes are important materials in the manufacture of all-solid-state batteries due to their ionic conductivity, achieved by doping the polymer with salt, and mechanical strength, achieved by use of a block copolymer with a rigid block. High salt concentration is advantageous to achieve high ionic conductivity, but it makes estimation of battery performance difficult due to the breakdown of dilute-solution theory, which assumes complete ion dissociation. Therefore, practical battery design would benefit from an empirical understanding of the relationship between ion dissociation and salt concentration in block copolymer electrolyte. In this study, the dissociation of lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) in polystyrene (PS)—poly(ethylene oxide) (PEO) diblock copolymer electrolyte was investigated using Fourier transform infrared (FTIR) spectroscopy. Quantitative analysis was performed to reveal the appearance of ion pairs and interactions between the salt and the ethylene oxide moieties with increasing salt concentration. FTIR peaks associated with polymer functional groups were found to be more useful than those of the TFSI anion for understanding the chemical state of the block copolymer electrolyte. In particular, PS peaks were used to quantify polymer dilution upon salt addition and verify that the Beer–Lambert law was valid at all concentrations investigated. PEO peaks revealed conformational changes of the polymer upon coordination with lithium ions. A previously unidentified FTIR peak was discovered that relates to polymer–salt interaction. It was used to determine the extent of salt dissociation, which compares well with a Raman study of a homopolymer electrolyte. This work definitively shows that LiTFSI dissolves into the PEO phase of the block copolymer, essentially unaffected by PS presence. It also establishes FTIR as a useful technique for quantifying dissociation state of concentrated polymer and composite electrolytes for lithium batteries.


INTRODUCTION
The development of future energy storage systems should include the improvement of battery materials to have higher power and higher energy density with low cost and safety (Goodenough et al., 2007;Li et al., 2018). All-solid-state batteries are a promising battery type in terms of safety, stability and energy density Judez et al., 2018). Polymer electrolytes are important materials in manufacture of allsolid-state batteries due to their adhesivity, processability, and chemical resistance (Ruck et al., 2019). For these reasons, they will likely play a role in composite/hybrid electrolytes, considered necessary to enable future energy storage systems.
After it had been found that the ether groups in the amorphous phase of poly(ethylene oxide) (PEO) coordinate with cations and conduct ions by segmental motion (Wright, 1975;Borodin and Smith, 2006), there have been a great number of studies on the electrochemical properties and the transport mechanisms of polymer electrolyte systems. This is, in part, because practical batteries require higher charge/ discharge rate than is currently achievable with PEO-based electrolytes . There are two complementary approaches that could address transport limitations: (1) the use of composites, and (2) the use of concentrated polymer electrolyte. Ion transport experiments and theory are significantly more complicated in composite and concentrated systems (Pesko et al., 2017;Villaluenga et al., 2018). Furthermore, salt concentration impacts not only polymer morphology and dynamics (Chintapalli et al., 2016;Oparaji et al., 2018) but also transport (Berliner et al., 2019;Kim and Hallinan, 2020). Increasing lithium salt concentration provides an increasing number of charge carriers in the form of dissociated ions. However, strong polymer-cation interactions can negatively impact ionic mobility (Suo et al., 2016;Ford et al., 2020). In addition, salt dissociation appears to decrease with increasing salt concentration in polymer electrolyte, resulting in the formation of triplets or higher ion clusters (Suo et al., 2016). These large agglomerates and neutral ion pairs negatively impact the diffusion and migration of the ions, making estimation or analysis of transport behavior complex or even impossible without direct knowledge of thermodynamics (via dissociation and interaction of ions with the chemical environment). This motivates an in-depth study of lithium salt dissociation and ion interaction with the chemical environment in concentrated composite electrolytes, such as block copolymer/salt mixtures.
The dissociation of N(SO 2 CF 3 ) 2 (bis(trifluoromethanesulfonyl) imide, TFSI) salts with group 2 elements (Mg, Ca, Sr and Ba) in PEO electrolyte was studied using Fourier transform infrared spectroscopy (FTIR) by Bakker et al. (Bakker et al., 1995). Two different geometric structures of ion pairs were found for different metal ions and ion pairs were found at high salt concentration. Dissanayake et al. studied the effect of temperature on the dissociation and phase transition of PEO/LiSO 3 CF 3 (lithium trifluoromethanesulfonate, LiTf) electrolyte using FTIR (Dissanayake and Frech, 1995). The degree of dissociation was measured by the band intensities of SO 3 and CF 3 groups in associated or free ions. FTIR and Raman spectroscopy of PEO/ LiN(SO 2 CF 3 ) 2 (lithium bis(trifluoromethanesulfonyl)imide, LiTFSI) were performed by Rey et al. (1998b). Attempts to reveal the conformational changes of ether groups or TFSI anions, and ionpairing effects of Li + and TFSI − were carried out with limited quantitative analysis. FTIR studies of ion pairs and polymercation interactions in block copolymer electrolyte systems with lithium salts have been conducted for PEO-poly(propylene oxide)-PEO triblock copolymer with imide salts (Bakker et al., 1996), poly(vinylpyrrolidone-co-methyl methacrylate) (PVP-PMMA)/LiClO 4 (Chiu et al., 2007), and PEO-containing polyester copolymer with sodium salts (Lu et al., 2009). In this study, LiTFSI dissociation is investigated in a high molecular weight polystyrene-poly(ethylene oxide) (SEO) block copolymer which is highly promising for lithium batteries due to its superior dendrite suppression and mechanical strength derived from the presence of the glassy PS block (Stone et al., 2012;Oparaji et al., 2018).
A recent molecular dynamics (MD) study on ion solvation found a decrease in the diffusion coefficient with increasing ion concentration in a diblock copolymer (Seo et al., 2019). In our previous study, the mutual diffusion coefficient of LiTFSI in SEO block copolymer was studied using time-resolved FTIR (Kim and Hallinan, 2020). The diffusion coefficient in the low salt concentration regime was found to decrease followed by an increase of the diffusion coefficient in the high salt concentration regime, where salt dissociation is presumed to be limited. Changes of the FTIR spectra were observed at various salt concentrations that appeared to indicate the presence of associated ion pairs in the concentrated polymer electrolyte samples. This current study takes a systematic, detailed look at the equilibrium FTIR spectra of SEO/LiTFSI electrolytes in an attempt to provide better understanding of the dissociation state of the salt and the coordination between the ions and the polymer backbone. All the regions of the mid-infrared region of the spectrum have been analyzed, including peak deconvolution where necessary, to track changes in peak intensity and position. A discussion, in light of other studies in the literature, connects these observations to the chemical interactions underlying the spectral changes.

EXPERIMENTAL Materials
A diblock copolymer of SEO was synthesized via anionic polymerization as described elsewhere (Oparaji et al., 2016). The molecular weight of PS and PEO blocks were 121 and 165 kg/mol, respectively. The PDI was 1.11, and the volume fraction of the PEO was 0.56 at 120°C. The synthesized polymer was freeze-dried under vacuum and stored at −20°C. LiTFSI was purchased from BASF and dried under vacuum for 48 h at 120°C in an antechamber connected to an argon-filled glove box. The dried LiTFSI was transferred to the glovebox without exposure to air, and is denoted "as-received".

Polymer Electrolyte Preparation
The SEO and LiTFSI were mixed in n-methyl-2-pyrrolidone (NMP), and the mixtures were stirred overnight at 60°C. The mixtures were cast on a hot plate covered with nickel foil, and the solvent was allowed to evaporate at 60°C. The cast films were removed from nickel foil after 24 h and dried under vacuum at 90°C for 24 h in the antechamber mentioned above. The thicknesses were measured by a micrometer. The average thickness of films was 100 ± 10 μm. The salt molarity in the resulting films was from 0 to 1.54 mol LiTFSI /L total (r 0-0.17 mol LiTFSI /mol EO ). Polymer electrolyte membrane casting was performed in an argon-filled glove box with <0.2 ppm of oxygen and water.
The molarity of species i, c i , for each electrolyte membrane was calculated with assumption of volume additivity of the mixture, where x i is the moles of species i, r is the molar ratio of salt and ethylene oxide (EO) moieties, M is the molecular weight, ρ is the density, and subscripts i, EO, S, PEO, PS and LiTFSI are species, ethylene oxide monomer, styrene monomer, PEO, PS and LiTFSI, respectively. x i of the LiTFSI, EO and styrene monomers were r, 1, and MEOMPS MPEOMs , respectively. Density values of 1.056 g/cm 3 for PEO (500 kg/mol) and 1.009 g/cm 3 for PS (110 kg/mol) at 120°C from reference data were used in this study (Zoller and Walsh, 1995). The density of LiTFSI at 20°C was taken because reference data at 120°C could not be found, and the temperature is much lower than the melting point of LiTFSI (236°C) (Aravindan et al., 2011).
The values of r, c and the volume fractions of each component are shown in Table 1. Assuming volume additivity amounts to at most 5% error (Teran and Balsara, 2014).

Fourier-Transform Infrared-Attenuated Total Reflectance (FTIR-ATR) Spectroscopy
The polymer electrolyte films were placed on a single reflection diamond ATR crystal (Golden Gate, Specac) in an argon-filled glove box and sealed by an o-ring. All spectra were the result of four scans, collected at a 4 cm −1 resolution in the range of 4,000-650 cm −1 on a PerkinElmer Frontier with a liquidnitrogen-cooled MCT detector. The background of the diamond crystal was taken immediately following each measurement after removing the sample. The absolute absorbance values reported in this work were calculated directly from the detected infrared energy transmitted through the sample at each wavenumber (E s ) and the energy of the background (E b ) according to (2)

Density Functional Theory Simulation
The interactions between the lithium cation and polymer or TFSI anion were simulated using density functional theory (DFT). All optimization and frequency calculations were conducted using the Kohn-Sham formulation of DFT with the B3LYP function and the 6-31++G(d,p) basis set. All geometry optimizations and vibration analyses were calculated using Gaussian 09 software (Frisch et al., 2009). All structures were optimized and frequency calculations were performed in order to ensure that a local minimum energy structure had been found. The data for the IR spectra was extracted from the output files using GaussView 5.0.9 (GaussView, 2009) and plotted using a scaling factor of 0.964 (Johnson III, 2019). Natural bond orbital (NBO) calculations were conducted on the optimized structures using NBO Version 3.1 (Glendening et al., 1998). Three-dimensional structures were produced with CYLView 1.0.1 (Legault, 2009). The chemical structures and the coordination of the materials and the simulation results are given in the Supplementary Material.

RESULTS
The FTIR spectra of pure LiTFSI and SEO/LiTFSI polymer electrolytes of different concentrations at 120°C are shown in Figure 1. The assignment of peaks of pure materials is shown in Table 2. The characteristic peaks of LiTFSI and PEO in the SEO electrolyte shifted with respect to the pure components, indicating that the salt dissolved into the PEO phase of SEO. The PS peaks do not exhibit any changes in peak position because the PS component is not involved in solvation of the salt. Therefore, all changes of the IR bands of LiTFSI arise from the interaction only between the PEO and LiTFSI. The shift of the peaks with addition of salt is presented in Table 2, where peak locations are reported for all salt concentrations investigated. The increase of lithium salt concentration resulted in the increase of the absorbance of the peaks associated with LiTFSI.  The most obvious increases can be found for the SO 2 asymmetric vibration (] a SO 2 , 1,349 and 1,330 cm −1 ), CF 3 asymmetric vibration (] a CF 3 , 1,180 cm −1 ), and SNS asymmetric vibration (] a SNS, 1,053 cm −1 ). Conversely, the absorbance of peaks associated with PEO and PS decreased with increasing salt concentration. Decreasing CH or CH 2 stretching bands were observed with increasing salt concentration at high wavenumber (νCH of benzene rings, 3,100-3,000 cm −1 , and ]CH and ]CH 2 of PS and PEO backbone, 2,925 and 2,860 cm −1 ). The COC vibration of PEO (]COC, 1,093 cm −1 ) was found to have decreased and shifted by addition of lithium salt. The various infrared bands associated with the phenyl groups of PS showed decreasing absorbance with increasing salt concentration, but did not show any change in the peak position. Since some peaks shifted as a function of salt concentration, to avoid confusion the peaks of the salt components are referred to by their location in the polymer/salt system with the highest salt concentration in our experiments (r 0.17 mol LiTFSI /mol EO ), and the peaks of the polymer components are referred to by their location in pure SEO (r 0 mol LiTFSI /mol EO ). Refer to Table 2 for detailed information regarding peak shifting according to the FTIR-ATR spectra before deconvolution.
The left-hand side of Figure 1 is the FTIR-ATR spectra of the polymer electrolytes between 3,150 and 2,500 cm −1 . In this wavenumber range, there are no LiTFSI peaks, and IR bands in this range are due to carbon-hydrogen vibrations of PS or PEO. The small peaks above 3,000 cm −1 are CH stretching of the phenyl groups (Jabbari and Peppas, 1993). These bands have small changes in intensity with increasing salt concentration, but the peak positions are constant. The broad set of convoluted bands between 3,000 and 2,500 cm −1 are due to CH 2 stretching of PS and PEO, CH stretching of the main chain of PS, and aromatic ring vibrations (Jabbari and Peppas, 1993). Although the highest peak at 2,859 cm −1 has components of PS and PEO, it seems to be dominated by the signal from CH 2 of PEO since the peak shifts with salt addition. The peaks in this region from the backbone of the PS block seem to have only a slight decrease with increasing salt concentration and no change in position and shape, because the PS component is inert to the interaction with the lithium salt (Oparaji, 2017). The changes in the absorbance of PS peaks are thought to be due to dilution by addition of LiTFSI, which increases the volume fraction of the PEO/LiTFSI phase. The depth of penetration of the FTIR-ATR evanescent wave is on the order of 1 μm and the infrared spot size is on the order of 1 mm 2 . Thus, this technique samples a volume on the order of 1 nL (100 µm) 3 , which is a sufficiently large control volume that the continuum concentration of the polymer electrolyte should be considered, unaffected by the block copolymer nanostructure, which is on the order of 100 nm. Therefore, addition of LiTFSI into the PEO phase of the block copolymer not only decreases the concentration of PEO, but also decreases the effective, macroscopic concentration of PS. Intensity changes of peaks associated with SEO are examined quantitatively and in more detail below using more intense peaks that occur in the fingerprint region. The right-hand side of Figure 1 is the fingerprint region from 1,750 to 650 cm −1 . The IR spectra of the polymer electrolytes in this region will be discussed subsequently using Figures 2-9, which present enlarged sections with more detail.

DISCUSSION
The results were examined thoroughly in terms of salt dissociation and ion interaction because the degree of dissociation and ion interaction largely impact the polymer electrolytes' physical and electrochemical properties such as the structure, mechanical strength, or ionic conductivity. The mechanical properties of block copolymer electrolytes are known to be dependent on the volume fraction of each component (Niitani et al., 2005;Devaux et al., 2015), the microphase separation (Rosedale and Bates, 1990;Ruzette et al., 2001), and the sequence of blocks (Matsuo et al., 1968). Since microphase separation of SEO becomes stronger with increasing salt content due to strong ionic interactions of lithium salt in PEO phase (Metwalli et al., 2015;Chintapalli et al., 2016;Oparaji et al., 2018), the mechanical strength of PS-PEO-based block copolymer electrolytes that are microphase separated in the neat state merely depends on the molecular weight and the volume fraction of glassy phase (Devaux et al., 2015;Oparaji et al., 2018).
The polymer-salt interaction of PEO-based polymer electrolytes was related to the thermodynamic and electrochemical properties using Lewis-base approach by Wieczorek et al. (Wieczorek et al., 1995). Different types of fillers (polymeric, organic, or inorganic) acted as Lewis acid or base centers and caused non-uniform distribution of metal cations in the vicinity of the fillers forming different types of complexes. This caused unique features of the polymer composites with each filler in terms of the crystallinity, the glass transition temperature, and the dependence of ionic conductivity on temperature or filler concentration. Pesko et al. (2016) studied the electrochemical, thermodynamic, and structural properties as a function of solvation-site connectivity which is defined as the ratio of number of oxygen atoms to the total number of atoms in a repeat unit. The solvation-site connectivity was controlled by varying the chain lengths of vinyl or EO groups in the repeat unit. Higher connectivity makes more solvation-sites and narrows the distance between solvation-sites, which facilitates ion hopping between the solvation-sites. Ionic conductivity showed a linear increase to increasing number of solvation-sites for 0.06 ≤ r ≤ 0.14 and T g + 55°C ≤ T ≤ T g + 95°C. A recent study of PEO-based polymer electrolytes with sodium triflate (NaTf) showed a rapid decrease of ionic conductivity at high salt concentration (EO : Na + ≤ 8) where appearance of aggregates was observed by FTIR spectroscopy (Lehmann et al., 2019). In block copolymer electrolytes, strong interaction of lithium salts with conductive polymer induces strong microphase separation and uniform distribution of salt in the conductive phase leading to improved ion transport (Seo et al., 2019). The interaction of both cations and anions with the non-conductive phase reduces the ion transport (Jo et al., 2013;Seo et al., 2019).
Since ions play an important part in battery electrolytes, there are quite a few studies of the dissociation state of alkali salts in various solvents. For example, the solvation and ionic state of LiTFSI can be investigated using the spectroscopic behavior of the block copolymer electrolytes. Vibrational spectroscopies are powerful methods to study the state of ions or molecules influenced by intermolecular or intramolecular interactions. Although its specialty is for identification of compounds, detection of functional groups, and qualitative analysis of chemical interactions, quantitative analysis is also possible. This requires precise and careful control of samples and experimental conditions along with appropriate analysis and proper assumptions (Kakihana et al., 1990;Fieldson and Barbari, 1993;Ferry et al., 1995;Sammon et al., 1998;Elabd et al., 2003;Philippe et al., 2004;Hallinan and Elabd, 2007;Oparaji et al., 2016;Yang et al., 2018;Kim and Hallinan, 2020). In this section, the dissociation of LiTFSI in SEO is investigated via detailed analysis of the fingerprint region.
Absorbance of FTIR is proportional to concentration according to the Beer-Lambert law; appropriate concentration units are molarity (M or mol/L). The molarity (and therefore absorbance) of the polymer is expected to decrease with increasing salt content. Pure SEO peaks (δCH 2 + phenyl group, 1,491 and 1,451 cm −1 ) as well as pure PS peak (ρCH 2 , 696 cm −1 ) were chosen to study the dilution of SEO by addition of LiTFSI. These are strong polymer peaks without any overlap by LiTFSI peaks. They are shown in Figures 2A,B. The peak positions do not show any significant change as shown in Figure 2C. However, the absorbance of these peaks increased when salt was first introduced (r 0.02) then steadily decreased with increasing salt concentration as shown in Figure 2D. The initial increase of the IR absorbance at low salt concentration (r 0.02) can be explained by (1) the density change of polymer/salt mixture due to non-ideal mixing (Teran and Balsara, 2014) or (2) the change of chemical environment of the system by addition of salt influencing the molar extinction coefficient which affects the IR intensity (Lappi et al., 2004). The further decrease of absorbance with increasing LiTFSI concentration is due to dilution as mentioned above. Figure 3A shows the FTIR-ATR spectra of SEO/LiTFSI between 1,400 and 1,150 cm −1 . The ] a SO 2 appears as a doublet at 1,348 and 1,331 cm −1 . The intensity of the doublet increases with increasing salt concentration. This doublet was found at lower frequencies for pure LiTFSI, 1,332 and 1,309 cm −1 ( Table 2). The position of the lower wavenumber peak appears to have a slight blueshift as salt concentration is increased from r 0 to r 0.05 and then a slightly more significant redshift from r 0.085 to r 0.17. This qualitatively tracks the trend of ionic conductivity of these electrolytes which increases to a maximum in the vicinity of r 0.085, followed by decreasing conductivity at yet higher salt concentration (Panday et al., 2009;Majeed, 2019). Unfortunately, the change in peak location is too small (less than the resolution of the measurement) for quantitative analysis. This doublet is interesting because of the unique behavior compared to the other  LiTFSI peaks. It is worth noting that the blueshift from pure LiTFSI to the SEO/LiTFSI polymer electrolyte (i.e., the peak shift to higher frequency by dissolving LiTFSI in polymer) only occurs for this doublet. The other LiTFSI peaks in the polymer electrolyte are redshifted with respect to pure LiTFSI (in the infrared range in this study). The ratio of the two peaks in the doublet has been used as an indication of the presence of conformational isomers of TFSI anions (Rey et al., 1998b). However, the peaks at 1,348 and 1,331 cm −1 overlap with the symmetric and asymmetric CH 2 wagging (ω s CH 2 , 1,348 cm −1 , ω a CH 2 , 1,324 cm −1 ) making quantitative analysis ambiguous and suspicious. The most intense peak in this region is the asymmetric CF 3 stretching of the TFSI anion (] a CF 3 ). The symmetric vibration of this functional group (] s CF 3 ) appears as a shoulder on the high wavenumber side. The symmetric CF 3 vibration is hard to notice at lower salt concentrations because it overlaps with the CH 2 twisting band (τCH 2 , 1,249 cm −1 ) and it is much smaller than ] a CF 3 . Like the SO 2 stretching peaks, the intensity of the CF 3 peaks increase with increasing salt concentration. Unlike the SO 2 vibrations, the CF 3 stretching peaks are located at lower frequency than those of pure LiTFSI. Regarding the weak SEO peaks, above a salt concentration of r 0.085, the τCH 2 at 1,249 cm −1 appears as a shoulder on the ] s CF 3 peak. The same behavior is apparent for the other twisting CH 2 peak at 1294 cm −1 near the ] a SO 2 (1,330 cm −1 ) peak. Figure 3B demonstrates that a deconvolution into six peaks fits this region of the spectrum quite well.
The deconvoluted IR bands from Figure 3B in the region from 1,400 to 1,200 cm −1 are presented in Figure 4A. With increasing salt concentration, the deconvoluted ] a SO 2 peaks at 1,348 and 1,331 cm −1 do not significantly change position but they do increase with increasing salt concentration ( Figure 4B) as expected. This is shown quantitatively in Figure 4C, where the normalized integrated absorbance of the deconvoluted ] a SO 2 peaks increase monotonically with r. The peak at 1,348 cm −1 is known to be affected more by C1 conformer (cis) than C2 (trans), and vice versa for the other (Herstedt et al., 2005). The ratio of these peaks therefore have been shown to be different with different content of geometric isomers. However, for PEObased polymer electrolytes, as we mentioned above, the ] a SO 2 doublet overlaps with ωCH 2 , such that decoupling and quantification of these peaks are very difficult. With this in mind, the ratio of the ] a SO 2 peak at 1,331 cm −1 to that at 1,348 cm −1 is shown in Figure 4D, and no significant physical implication was found in this study.
In Figure 4B, the increase of absorbance of the ] a SO 2 doublet seems to deviate from the expected linear relationship with the concentration. The peak absorbances were normalized for comparison where A n is the normalized absorbance, A is the absorbance at each salt concentration, and the superscripts denote the sample numbers in Table 1. In order to remove the effect of volume changes, the peaks were calibrated by the absorbance of δCH 2 + phenyl group (1,500-1,420 cm −1 ).
where A c is calibrated absorbance, A is the absorbance of the peak to be calibrated, and A(δCH 2 + phenyl group) is the absorbance of δCH 2 + phenyl group (1,500-1,420 cm −1 ). The calibrated absorbances were then normalized where superscripts denote the sample numbers and the subscripts c and n mean calibration and normalization, respectively. As shown in Figure 4C, the normalized calibrated absorbances were linear as a function of the molarity ratio of salt to SEO. The linearity of the plot demonstrates the validity of the Beer-Lambert law across the entire concentration range since A salt /A SEO ∝ c salt /c SEO . Note that c SEO c EO + c S . The calibration and normalization process continues to be applied in the following analysis.
The τCH 2 (1,292 and 1,249 cm −1 ) show interesting behavior in Figure 4A. These peaks initially have similar intensities and shape when there is no salt in the polymer. With addition of salt, the τCH 2 band at 1,292 cm −1 gradually shifts to higher wavenumber and the peak area increases ( Figure 4B) while the other τCH 2 becomes smaller and broader. The ratio of the two CH 2 bands from this study [CH 2 (1, 249 cm −1 )/τCH 2 (1, 292 cm −1 )] is shown in Figure 4D. Matsuura and coworkers studied the conformational change of molten PEO and reported the gauche (1,296 cm −1 ) and gauche + trans (1,246 cm −1 ) isomerism of CH 2 deformations (Matsuura and Fukuhara, 1986). Shieh and Liu identified the IR bands at 1,280 and 1,242 cm −1 as gauche and trans conformations of τCH 2 , respectively (Shieh and Liu, 2004). Therefore, in this paper we denote the band at 1,292 cm −1 for the pure SEO (r 0) as gauche and 1,249 cm −1 as trans conformer. The transition from trans to gauche conformations with increasing LiTFSI content implies that there is chain reorientation due to the attractive force of the oxygen atoms in the ether groups coordinating lithium cations in a crownether-like structure. MD simulation of the gauche-trans conformations of EO chains according to coordination with hydrogen ions forming hydrogen bonds has been reported by Begum and Matsuura, and gauche conformation has been found to be most favorable (Begum and Matsuura, 1997).
The deconvoluted IR bands from Figure 3B at 1,227 cm −1 (] s CF 3 ) and 1,180 cm −1 (] a CF 3 ) are presented in Figure 5A. The ] a CF 3 band arises at 1,189 cm −1 for r 0.02 and monotonically increases until r 0.17, shifting to 1,180 cm −1 . It is unclear what causes the significant shift in this peak. Despite the significant number of spectroscopic studies of TFSI, there is surprisingly little discussion of this intense IR peak. This may be partially due to the large number of vibrational modes of this functional group (Herstedt et al., 2005). The significant shift of ] a CF 3 to lower wavenumber from pure LiTFSI to that in the block copolymer electrolyte and the further redshift with increasing LiTFSI concentration could be related to (1) the strong electronegativity of fluorine and the delocalized negative charge upon dissociation, (2) conformational changes of TFSI, (3) hydrophobic interactions between different CF 3 groups, and/ or (4) Li + -F coordination (Arnaud et al., 1996). Due to the  uncertainty and ambiguity in assigning physical significance to the spectral changes of ] a CF 3 , attention is turned to the weaker ] s CF 3 shoulder. The absorbance of ] s CF 3 increases with increasing salt concentration. This can be seen quantitatively in Figure 5B, where the normalized integrated absorbance of both CF 3 stretching peaks is shown. The weaker normalized absorbance increase of ] s CF 3 as compared to ] a CF 3 is most likely due to the overlap of ] s CF 3 with τCH 2 , as mentioned above. With increasing salt concentration, the symmetric CF 3 vibration appears to blueshift slightly, but again physical significance is difficult to ascribe. Based on the linearity of A c,n (] a CF 3 ) in Figure 5B, it is clear that ] a CF 3 accurately represents salt concentration, verifying it use for diffusion studies in prior work (Kim and Hallinan, 2020). Figure 6 shows the FTIR-ATR spectra of SEO (r 0), LiTFSI, and the salt-doped polymer electrolytes between 1,160 and 713 cm −1 . As shown in Figure 6A, the symmetric stretching vibration of SO 2 (] s SO 2 , 1,131 cm −1 ) of the polymer electrolytes is found at lower wavenumber than that of the dry solid LiTFSI (1,139 cm −1 ). The other IR peak from the salt is the SNS vibration mode (] a SNS) at 1,053 cm −1 . This is also found at lower position than that of pure LiTFSI (1,060 cm −1 ). The COC stretching (]COC, 1,093 cm −1 ) decreases and redshifts with increasing salt concentration, in agreement with other reports of PEO/salt electrolytes (Bakker et al., 1995;Bakker et al., 1996;Zhang and Wang, 2009). The ]COC is known to be related to the crystalline structure of PEO which appears strongly below the melting temperature (T m ) of PEO then reduces above T m due to crystal melting. The spectra were collected at 120°C where there is no crystallinity in PEO. Thus, the decrease of the COC peak with increasing salt concentration is due to (1) dilution of ether groups with the addition of salt, and (2) change of COC conformation, forming crown-ether-like coordination with the lithium cations. The dilution effect is examined by calibrating and normalizing the deconvoluted COC absorbance. This is shown as a function of EO molarity in Figure 6C. At high EO molarity (i.e., low salt concentration) a linear trend is observed, but a plateau is found at lower EO molarity (i.e., higher salt concentration) indicating that the effect of interaction with lithium ions is stronger in the concentrated electrolyte. The peak position, also shown in Figure 6C, mirrors the absorbance trend. Bakker et al. (1995) and Bakker et al. (1996) claimed that stronger interaction between ether groups and cations causes a larger shift. In the present work, the shift was stronger at low salt concentration then became smaller at high salt concentration implying that the interaction between EO segments and lithium ion saturates at high salt concentration. This apparent contradiction as well as the mirroring of absorbance and peak position could be an artifact of deconvolution due to the strong overlap of the COC peak and LiTFSI peaks. This demonstrates that the COC absorbance is not appropriate for quantitative FTIR analysis.
The peaks between 1,100 and 800 cm −1 shown in Figure 6A originate from SEO, with the exception of ] a SNS at 1,053 cm −1 . The peak that is observed at 1,027 cm −1 in this study is more significantly contributed by the vibration of the phenyl groups of PS (1,027 cm −1 ) (Liang and Krimm, 1958;van Asselen et al., 2004). The small plateau region between 1,046 and 1,037 cm −1 for the pure SEO (yellow solid line in Figure 6A) is thought to be contributed by a C-C-O-C-C deformation. The C-C-O-C-C conformational change around 1,045 cm −1 has been reported for pure PEO as having very weak intensity (Matsuura and Miyazawa, 1969;Yu and Wu, 2007). Despite its weak intensity, the peak at 1,045 cm −1 along with the PS peak at 1,027 cm −1 makes the analysis of ] a SNS difficult due to strong overlap. The CH 2 rocking at 941 cm −1 (ρCH 2 ) showed an absorbance increase with increasing salt concentration while the other rocking bands showed decreases (990, 860-845, 696 cm −1 ). It is worth noting that the tail of the strong ] a SNS ( Figure 6A) extends into this region and could impact the apparent absorbance changes of the PEO and PS bands, such as the CH 2 increase.
The deconvoluted IR bands from Figure 6B between 805 and 713 cm −1 are shown in Figure 7A. The assignments of these peaks are not clear in many literature reports (Pennarun and Jannasch, 2005;Kam et al., 2014;Velez et al., 2016). These three peaks are thought to be shifted from 808, 773, and 745 cm −1 for the solid LiTFSI to 787, 759, and 739 cm −1 for the SEO/LiTFSI electrolyte at r 0.17. The assignments of these peaks are shown in Table 2. For pure LiTFSI, the IR peaks at 808 and 773 cm −1 seem to be complex combination peaks including the vibrations of CS, SN, and CF 3 components (Rey et al., 1998a). However, there is agreement that the peak at 745 cm −1 in pure LiTFSI is the symmetric SNS vibration (] s SNS). Upon incorporation of LiTFSI into SEO, all three of these peaks redshift. The peak at the highest frequency shifted from 808 cm −1 in pure LiTFSI to 787 cm −1 at r 0.17 mol LiTFSI /mol EO ; the middle peak shifted from 773 to 755 cm −1 ; and the lowest peak shifted from 745 to 740 cm −1 . As a result, the peak at 787 cm −1 showed monotonic increase in absorbance without changing the peak position with increasing salt concentration with only minor overlap with the CH 2 rocking band (ρCH 2 ). The absorbance increase is presented quantitatively in Figure 7B. Like ] a CF 3 and ] a SO 2 , the normalized, calibrated absorbance of the peak at 787 cm −1 ( Figure 7B) is more-or-less linear with salt concentration. The 787 cm −1 peak position also exhibits minor shifts mirroring those of ] a SO 2 that are too small for quantitative analysis. The other LiTFSI peaks seemingly split from 748 cm −1 , but in fact, it is due to the growth of two different peaks (νCS, ]sN, νCF 3 at 759 cm −1 and ] s SNS at 740 cm −1 ) around the peak of benzene ring vibration (748 cm −1 ). Due to this overlap, no quantitative analysis was attempted. Figure 8 presents the FTIR spectra of SEO, SEO/LiTFSI, and LiTFSI from 1,715 to 1,615 cm −1 . No absorption was found in this region for the pure SEO nor for solid LiTFSI, but a peak clearly appears in the electrolyte at 1,675 cm −1 . The peak location was constant for all samples. To the best of our knowledge, this peak has not been reported in literature for any PEO-based polymer electrolytes nor any LiTFSI-doped electrolytes. We propose that this peak is attributed to the ether-lithium interaction. An approximate argument for this assignment can be made using liquid water stretching as a reference (] 3, 400 cm −1 ). A simple rule of thumb that can be used to examine how vibrational Frontiers in Energy Research | www.frontiersin.org September 2020 | Volume 8 | Article 569442 frequency will change upon change in mass of an atom in a functional group follows where c is speed of light, κ is the force constant (related to the stiffness of the bond), and reduced mass is (Silverstein et al., 2005) μ m 1 m 2 m 1 + m 2 .
As a zeroth order approximation assuming that κ is constant, the shift on going from H 2 O stretching (μ H 2 O 1.79 g/mol) to Li 2 O stretching (μ Li 2 O 7.43 g/mol) is This is remarkable agreement with observation (1,675 cm −1 ), but clearly involves somewhat arbitrary assumptions. This peak assignment should be verified with simulations. A peak has been observed in liquid electrolyte at 1,650 cm −1 (Pyun, 1999). Along with several other peaks, it was assigned to a degradation product of LiPF 6 and ethylene carbonate that contains Li-O functionality and can be catalyzed by water presence. However, unlike the other peaks in that study, the absorbance at 1,650 cm −1 did not increase monotonically with water content. This could indicate that it actually corresponds to the coordinated Li-O intermediate in the reaction scheme, which would support our peak assignment. It is worth mentioning that the water bending absorbance occurs in this region (1,640 cm −1 ). It is unlikely that the peak at 1,675 cm −1 is due to water contamination because our experimental setup has been design and tested to rigorously exclude exposure to air. Returning to experimental observations, the peak did not show any further absorbance increase on going from r 0.125 to r 0.17. This might be due to the number of ether groups per lithium ion being saturated at r 0.125 (EO:Li + 8), resulting in incomplete dissociation of LiTFSI with addition of more salt. If additional Li cations are not generated with the addition of salt, then additional Li + -O coordination cannot form and the peak intensity will not increase. The normalized absorbance of this peak is shown in Figure 8B (blue dotted line with × marks), where it appears to plateau. However, if the effect of dilution is taken into  account using the normalized, calibrated absorbance (red solid line with ○ marks in Figure 8B), it can be seen that LiTFSI dissociation is not completely shut down at the highest salt concentration (r 0.17, EO:Li + 6), but it is incomplete. The data for A c,n (1, 675 cm −1 ) in Figure 8B can be used to estimate the degree of dissociation. A line is fitted to the experimental data of r 0.02-0.085 that are assumed to be fully dissociated. The data at r 0 was excluded from the fitting due to the significantly different chemical environment indicated by the trend shown in Figure 2D. As shown in Figure 9, A c,n (1, 675 cm −1 ) falls below the line at r 0.125 and 0.17. The degree of dissociation at the two highest salt concentrations was calculated by the deviation of the experimental data from the line. The SEO/LiTFSI electrolyte is found to be 90% dissociated at r 0.125 and 70% dissociated at r 0.17. This finding is in relatively good agreement with a study by Edman, who used a Raman mode associated with Li + -PEO coordination to conclude that dissociation in PEO-LiTFSI electrolyte above EO:Li + 8 is incomplete (Edman, 2000). In particular, Edman found that LiTFSI nearly completely dissociates in PEO at r 0.125, but that LiTFSI is only 76% dissociated at r 0.17. Agreement was also found from a recent study of PEO-based polymer electrolyte and NaTf system (Lehmann et al., 2019). A strong decrease of ionic conductivity was observed above EO:Na + 12 (20 wt%) for dry linear PEO-NaTf. The FTIR spectra of PEO-NaTf showed that aggregates appeared for the salt concentrations above EO: Na + 8:1. The agreement among the three studies using two different techniques strengthens the conclusions and indicates that the presence of PS in the block copolymer does not have a significant impact on dissociation state of the salt in the PEO phase.
Several computations were conducted following the approach of Han et al. (2017), in which DFT calculation has been used to support experimental results of FTIR and Raman spectroscopy. In this study, computations were conducted in order to identify a potential origin for the peak that appears at 1,675 cm −1 as the LiTFSI is added to the system. We have initially tested if the peak originates from the coordination of the Li cation and the PEO backbone. In order to look at Li-PEO coordination, a fragment of the PEO backbone was modeled as coordinated with a Li-cation. Backbone fragments consisting of a single chain containing either two or three oxygens (1,2dimethoxyethane and 1-methoxy-2-(2-methoxyethoxy)ethane respectively), and two individual chains of 1,2-dimethoxyethane were explored as potential Li-PEO interaction sites that could be occurring experimentally. Vibrations corresponding to the C-H bonds before and after Li-coordination were in agreement with the experimentally determined spectra suggesting that our model system is a good approximation to the experimental system. The experimental peak at approximately 1,500 cm −1 corresponds to C-H stretching on the PEO-backbone and therefore was used as a reference peak for the calculated IR spectra. All three potential Li-PEO coordination calculations had the expected peak around 1,500 cm −1 but did not have a peak around the region of interest [both before and after scaling (by 0.964)], suggesting that the peak at 1,675 cm −1 is not from Li-PEO coordination. Next, the potential for the TFSI counterion interacting with the Li-PEO system was explored. These simulated IR spectra also had the internal standard peak at ∼1,500 cm −1 corresponding to C-H stretching of the PEO backbone. The other PEO backbone peaks overlap with the peaks experimentally shown to correspond to the LiTFSI salt making it difficult to differentiate most of the other IR signals. Despite this, the expanded system also did not result in a peak around 1,675 cm −1 .
The potential presence of water was explored next where the Li-PEO system formed a Li + /OH 2 complex with a water molecule. Based off the computations of the Li-PEO systems with and without water, there are expected to be additionally three peaks resulting from water. The first peak will appear around 1,660-1,600 cm −1 and the other two peaks are expected to occur around ∼3,750 and ∼3,650 cm −1 ( Figure 10A). When calculated all three peaks had similar intensities meaning if the first peak at 1,675 cm −1 truly corresponds to water then the other peaks should be observable unless PS is IR-active in these regions. This system displayed a peak at 1,660 cm −1 before scaling (∼1,600 cm −1 after scaling by 0.964) that corresponded to O-H wagging from the water molecule. This result indicates that the presence of water might cause the IR band around 1,675 cm −1 in the experimental system. However, there is no indication of OH stretching (3,130-3,740 cm −1 ) absorbance in the experimental spectra ( Figure 10B), and the OH stretching vibration is a stronger absorber than the HOH bending vibration ( Figure 10B).
Other potential interactions between Li and oxygen-based impurities were explored in order to rule out their involvement. A hydroxy-anion coordinated to the Li cation in the Li-PEO system did not produce a peak around the 1,675 cm −1 region. Finally, the IR spectra for an oxygen-bridge between the two Li-PEO systems was explored too. Again, formation of a Li-O-Li bridge did not result in a peak around the region of interest.
Interestingly, stereoelectronic effects were found in the calculated IR spectra for the C-H bond vibration region (∼3,000 cm −1 ) when converting from the uncoordinated trans conformer to the Li-coordinated gauche configuration. These changes originate from the attenuation of anomeric effect (Juaristi and Cuevas, 1992;Alabugin, 2000;Kirby, 2012;Gomes et al., 2015) When uncoordinated, the lone pairs of oxygen are able to donate electron density to the antibonding C-H orbitals, effectively weakening the C-H bonds. However, in order for the PEO backbone to coordinate with the lithium cation it must adopt a gauche conformation and the oxygen uses one of its lone pairs to coordinate with lithium. As a consequence, oxygen has only one lone pair that can participate in hyperconjugation with the C-H bond and the orbital overlap between the donor and acceptor has been compromised due to the now gauche geometry. Both of these factors decrease the average donor ability of oxygen from 6 kcal/mol in the uncoordinated configuration to 4.2 kcal/mol in the Li-coordinated complex according to NBO analysis (Glendening et al., 2012). Due to the weaker donor ability of oxygen in the gauche conformer the IR peaks corresponding to C-H bond vibrations are blue-shifted in comparison to the trans-conformer spectra, as shown in Figure 11A. Although somewhat less pronounced, the same blue shift is seen in the experimental spectra of Figure 11B, where the C-H bond vibration in question monotonically shifts from 2,860 cm −1 in the neat polymer to 2,882 cm −1 at r 0.17 mol Li /mol EO (Table 2). In other words, the simulated and experimental spectra both indicate that the amount of gauche PEO conformations increase with addition of LiTFSI salt to the polymer, and the computational modeling provides molecular insight into the origin of these changes.
The interaction of cation and NMP was also studied experimentally since the C O double bond has a peak around 1,675 cm −1 . The spectrum of a solution of NMP and LiTFSI, shown in Figure 12A, exhibits a large C O peak. After collecting the spectrum of the solution, the LiTFSI/NMP mixture on the ATR crystal was brought into the argon-filled glovebox, unsealed, and dried overnight at 60°C. The sample was further dried for one more overnight under vacuum at 60°C. Then another FTIR spectrum was collected of the LiTFSI dried from solution. This spectrum is shown in Figure 12A along with a spectrum of as-received LiTFSI that had never been exposed to NMP. There are distinct differences between the LiTFSI salt dried from solution and the as-received LiTFSI. For the most part, the LiTFSI peaks in the wavenumber range from 1,550 to 650 cm −1 agree between the dried and the as-received salt. On the other hand, a weak peak remains at 1,675 cm −1 (and 3,050-2,800 cm −1 , not shown) in the LiTFSI salt dried from solution, whereas there are no indications of these peaks in the as-received salt. This seems to indicate that there is strong solvation between LiTFSI and NMP, and that NMP-LiTFSI complex remains in polymer electrolytes processed with this solvent. It is known that large dielectric constant of PEO reduces ion-ion interaction and leads to higher ion mobility (Wheatle et al., 2017). However, when there is solvent with larger dielectric constant than polymer, lithium cation is likely coordinated to solvent rather than polymer (Ford et al., 2020). We propose that the formation of Li-NMP complex depends on the amount of lithium salt. To prove this, salt diffusion test was set up as described in our previous study (Kim and Hallinan, 2020) with the membrane of higher salt concentration placed below the membrane with lower salt concentration. Figure 12B represents the time-resolved FTIR spectra where the LiTFSI diffuses from the bottom to the top membrane. The initial spectrum is represented as orange line and the final spectrum is blue. The peak at 1,675 cm −1 decreased with decreasing LiTFSI concentration supporting the assertion that this peak is from a complex containing LiTFSI.

CONCLUSION
In this study, the dissociation of LiTFSI in the amorphous PEO conducting phase of an SEO diblock copolymer electrolyte was investigated using FTIR-ATR spectroscopy. Quantitative analysis showed increasing interactions between the salt molecule and ethylene oxide chains with increasing salt concentration and a lack of complete dissociation at the highest salt concentration studied. The inert polystyrene block worked as a reliable indicator of the consistency of the Beer-Lambert law at high salt concentration. In other words, the absorbance of polystyrene peaks decreased with increasing LiTFSI content (due to dilution), but lacked any detectable interaction with the salt. The PEO phase, however, was found to interact strongly with Li + . Conformational changes of the PEO backbone upon Li + complexation were detected via relative changes of CH 2 vibrations (τCH 2 ) and a blue shift of CH stretching due to the anomeric effect that was identified via simulations. Note that TFSI peaks, for the most part, exhibited only weak changes connected to anion conformations with changing salt concentration. On the other hand, large spectral changes indicative of conformational changes of the ether backbone of PEO were evident with changing salt concentration. The appearance of a previously unidentified peak at 1,675 cm −1 was attributed to NMP-LiTFSI complexation, indicating that polymer electrolytes processed with NMP contain residual solvent strongly solvating the LiTFSI salt. Analysis of this peak at 1,675 cm −1 , led to the same conclusions as those reached in a separate Raman study of PEO-LiTFSI electrolyte, namely that incomplete salt dissociation begins above r 0.125 (EO:Li + 8). Dissociation and chemical environment of the lithium salt was therefore surprisingly determined primarily from FTIR-ATR spectral changes associated with the polymer rather than those of the TFSI anions. The agreement between FTIR and Raman spectroscopy strengthens the claims of both studies. The comparison of composite SEO electrolyte with PEO electrolyte further indicates that PS does not significantly affect salt dissociation nor ion-PEO interaction. The findings from this work gives many interesting aspects to be explored and will work as a useful resource in study of polymer electrolyte systems.

DATA AVAILABILITY STATEMENT
All datasets presented in this study are included in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
The manuscript was written through contributions of all authors. KK conducted the FTIR experiment and analysis. LK conducted the DFT calculations. All authors have given approval to the final version of the manuscript.

FUNDING
This work was supported by the National Science Foundation award number 1804871.