Influence of Binder Coverage on Interfacial Chemistry of Thin Film LiNi

In this work, we explore the in ﬂ uence of binder coverage and chemistry on the interfacial properties of the textured Ni-rich cathode LiNi 0.6 Mn 0.2 Co 0.2 O 2 . We ﬁ nd that the formation of the cathode/electrolyte interphase (CEI) composition varies signi ﬁ cantly for cathodes coated with either poly(vinylene ﬂ uoride) (PVDF), carboxymethyl cellulose (CMC), or lithium polyacrylate (LiPAA) after cycling to high upper cutoff voltages (4.5 V vs Li/Li). The PVDF-coated samples had a thinner CEI and twice the relative concentration of LiF and Li 2 CO 3 to Li x PO y F z species in the CEI compared to the uncoated sample. This correlated with signi ﬁ cantly lower interfacial impedance (285 vs ∼ 1700 Ohm-cm 2 ) and improved capacity retention between cycles of the PVDF- coated samples compared to the other binder compositions and the uncoated sample. CMC-coated samples performed worst, with a CEI comprised of greater amounts of Li x PO y F z . In addition, we ﬁ nd the choice of binder results in the selective protection or promotion of electrolyte reactions at the (104) surface of the 622 cathode. This suggests that the choice of binder can impact the surface chemistry and performance of high voltage cathodes and supports an avenue for interest in multifunctional binders for stabilizing the CEI.

Electrolyte decomposition products deposited on the surface of the cathode are referred to as the cathode/electrolyte interphase (CEI). This surface layer is typically comprised of lithium fluorophosphates (Li x PO y F x ) and LiF from salt decomposition 1 and lithium carbonates, polycarbonates (Li 2 CO 3 , ROCO 2 Li), and lithium alkoxides (ROLi) from solvent decomposition. [2][3][4] The CEI is often referred to as a passivation layer against continuous electrolyte decomposition, but continuous transition metal dissolution 5 and gas evolution from the cathode at high voltages 6 suggest that this layer does not fully stabilize the interface. Indeed, little is known about the factors which influence CEI formation when compared to the solid electrolyte interphase (SEI) at the anode.
In promising cathodes, such as Ni-rich LiNi x Mn y Co z O 2 (where x + y + z = 1 and x > 0.5), the cathode can be involved in side reactions. At high upper cutoff voltages (>4.4 V vs Li/Li + ), LiNi 0.6 Mn 0.2 Co 0.2 O 2 (NMC622) undergoes surface structural rearrangement from layered R3m to spinel-like and rock salt phases (Fm3m). 7 This structural rearrangement has been correlated with the onset of oxygen evolution as atomic oxygen or singlet oxygen 8 which induces chemical oxidation of the electrolyte prior to electrochemical oxidation. 9 The reaction mechanism proposed by Gasteiger et al. 10 for this decomposition involves ethylene carbonate (EC) oxidation into vinylene carbonate (VC) and H 2 O 2 at the interface followed by electrooxidation of H 2 O 2 above 3.8 V vs Li/Li + to form H 2 , O 2 , and H 2 O. These species are known to initiate additional electrolyte decomposition reactions such as the formation of HF with fluorinated salts like LiPF 6 . 10 Oxidation of EC may be followed by a ring opening reaction, producing CO 2 and oligomers at the surface which contribute to CEI structure. 2 Composite lithium-ion battery electrodes are comprised of an electrochemically active material, conductive additives, and a binding agent which are primarily responsible for Li + storage, providing continuous electronic conductivity, and mechanical cohesion and adhesion, respectively. Binder and conductive carbon are commonly referred to as inactive components, but likely play a role in the observed chemistry. Indeed, the carbon of the electrode has been shown to decompose upon charging, producing CO 2 . 9,11 Furthermore, recent work on novel binders has provided a promising route toward forming a stable CEI for high voltage operations of Nirich NMC materials. Song et al. reported a fluorinated polyimide binder which covalently bound to the surface of Ni-rich NMC via its carboxylic acid group to form a highly stable chemical network which reduced capacity fade at high voltage operation. 12 Manthiram et al. proposed lithium polyacrylate (LiPAA) could form an artificial CEI on the surface of high voltage spinel LiNi 0.5 Mn 1.5 O 4 (LNMO). 13 Dou et al. demonstrated the improved rate capability of NMC333 electrodes with carboxymethyl cellulose over PVDF when cycled between 2.5-4.6 V vs Li/Li + . 14 Despite this evidence of promising CEI stabilization options, common binders such as poly(vinylidene fluoride) (PVDF), CMC, and LiPAA have not been thoroughly investigated for their influence on CEI formation of Nirich NMC cathode materials. Additionally, the morphology of these binders, i.e. coverage/uniformity, is rarely considered in composite electrodes. This raises the question whether the extent of coverage might influence the formation of the CEI.
In this study, we examined the effects of binder morphology and composition on the formation of the CEI on Ni-rich NMC622. Thin planar films comprised solely of active material were used to model potential binder coverage environments in composite electrodes by spin and spray coating. PVDF is selected as a reference case to typical composite systems, CMC represents a common waterprocessed binder, and LiPAA is chosen as a potential multifunctional binder (i.e. Li source for CEI formation). These systems are compared to the baseline uncoated case to discern trends in surface composition and electrochemical performance.
Following deposition of 500 nm NMC622 on 1 cm diameter Al 2 O 3 substrates coated with Co and Pt, thin films were annealed at 700 o C for 1 h under high purity air flow (∼0.2 lPM, Airgas) with a ramp rate of 5 o C min −1 then stored in an Ar-filled glove box.
Binder solutions were prepared by roller milling (U.S. Stoneware) powders with solvents for 24 h. 0.1 wt% PVDF (5130, Solvay) was dissolved in N-methyl pyrrolidinone (NMP) and 1.0 wt% CMC-Na (Acros Organics, 90 K MW) was dissolved in ultrapure (18 MΩ) deionized water. LiPAA was produced by dissolving polyacrylic acid (PAA, Sigma Aldrich, 450 K MW) in water to ∼10 wt% then titrating the binder solution with LiOH until a neutral pH was reached. 16 The binder solution used here was diluted to 0.2 wt% LiPAA in water.
Binder coatings were spin-coated on a Cee Model 100CB by dropping 20 μl binder solution onto the center of the sample disk which was spun at 3000 RPM for 30 s such that excess binder solution was spun off into the catch cup. Spray coating were done with a Talon Gravity Feed Airbrush Set in a single pass across the sample at a 2'' fixed standoff distance. Spin and spray-coated films were dried in air at 80 o C then placed in a vacuum oven at 100 o C overnight to remove residual water. Samples were weighed on a microbalance before and after deposition, but the mass change was not significantly above the error margin of the balance so alternative methods of characterization were employed.
Characterization.-Surface morphology and composition data were collected with a Scanning Electron Microscope equipped with an Energy-Dispersive X-ray Spectrometer (SEM/EDS, Hitachi TM3030 Plus with Quantax 70). X-ray Photoelectron Spectroscopy (XPS) was conducted on a PHI 3056 XPS spectrometer operated at 350 W and 15 kV with an Mg Kα (1253.6 eV) source. Pristine samples were transferred in air and cycled films were transferred under vacuum for measurement in a cryo-pumped vacuum chamber at 10 −9 Torr or less (10 −11 Torr base pressure). Survey scans were collected at 93.9 eV pass energy with 0.5 eV energy steps while high-resolution scans were acquired at 23.5 eV pass energy and 0.05 eV energy steps with 20−60 repeated scans of all spectra to improve the signal-to-noise ratio. Spectra were shifted relative to the adventitious carbon peak (284.8 eV) to correct for charging. Raw data was fit to component peaks based on pure binder powders and Li salts collected in the same instrument to provide standard references for relative peak area, position, and full width at half maximum (FWHM). Quantification of species by atomic concentration, X A , was calculated in CasaXPS based on the corrected area: Where A is the raw area intensity, RSF is the relative sensitivity factor of the element based on Scofield cross sections, T is the transmission factor for the instrument, and λ is the inelastic mean free path of an electron at the given binding energy.
Electrochemical measurements.-Electrochemical cells were assembled in an Ar-filled glove box in Swagelok cells vs Li metal (1 cm diameter), in duplicate. Two 1.3 cm diameter separators (Dreamweaver Gold 40) were soaked in 300 μl 1.2 M LiPF 6 in ethylene carbonate (EC) and ethyl methyl carbonate (EMC) in a 3:7 wt. ratio (Tomiyama). Cells rested at open circuit voltage for 2 h before cycling for two formation cycles between 3.0-4.5 V vs Li/Li + in a constant current, constant voltage protocol (CCCV). Constant current steps of 10 mA g −1 (∼C/20, where theoretical capacity, C = 200 mAh g −1 ) were held until the upper cutoff voltage (UCV) followed by a constant potential hold at UCV until measured current dropped below 5 mA g −1 (C/40) then cells were discharged at 10 mA g −1 constant current to the lower cutoff voltage. Electrochemical Impedance Spectroscopy tests were conducted after the 2 formation cycles on a BioLogic MPG-2 Battery Tester when cells were at open circuit voltage over a frequency range of 10 mHz to 20 kHz with 6 mV applied signal. Cycled electrodes were extracted in an Ar-filled glove box, rinsed in 1 mL dimethyl carbonate (DMC) for 30 s before drying and transferring to the XPS chamber under vacuum.

Results and Discussion
The active material of a composite electrode is difficult to study with surface sensitive techniques such as XPS due to the presence of conductive carbon and binders. These additional components support good electronic contact and mechanical robustness of the electrode but attenuate the signal of the active material when examined in XPS, as demonstrated for the Ni2p signal collected from the composite NMC622 electrode in Fig. 1a (bottom) with 90 wt% NMC622, 5 wt% PVDF, and 5 wt% carbon black. In contrast, one can see enhanced sensitivity to the active material by preparing planar thin film (0.5 μm) electrodes by magnetron sputtering, resulting in clear Ni2p signal from the cathode ( Fig. 1 a-middle). 15 These films adopt a (104) texturing and provide a controlled environment for experimenting on the cathode material and enable detection of interfacial signals which are attenuated by inactive components in composite electrodes. Consequently, the cathode/electrolyte interface of NMC622 can be studied, with attenuation of the surface signal caused by the formation of a layer of electrolyte degradation products ( Fig. 1a-top). The differences are schematically illustrated in Fig. 1b and provides a platform for studying the effects of artificial surface layers.
The morphology of a polymer composite electrode can vary depending on the mixing method, slurry composition, and quality of slurry coating. 17 Hypothetical morphologies are schematically depicted in Fig. 2 beginning with minimal surface coverage (Fig. 2a), where binder and conductive carbon fills voids between active material particles to form a cohesive network while leaving a majority of the active material surface exposed to the electrolyte, as has been proposed for PVDF-based electrodes. 18 A heterogeneous layer of binder and carbon may form instead (Fig. 2b), where the active material particles are coated sporadically with regions of substantial binder agglomerates alternating with bare or thin regions of binder coverage. A third case is shown in Fig. 2c, where uniform coverage of binder across the surface of the active material effectively encapsulates particles in a thin layer of polymer. These different binder configurations can be modeled by controlling the coating of binders on a well-defined thin film surface. The minimal surface coverage situation (Fig. 2a) is represented with an uncoated thin film of NMC622 and serves as our baseline. Sporadic globular coverage (Fig. 2b) is studied by spray coating binders on the surface of thin films, while uniform thin coverage (Fig. 2c) is accomplished by spin coating binder solutions. Conductive carbon is omitted so that the influence of binder coverage on the cathode/electrolyte interface may be studied in isolation.
The baseline coverage case-an uncoated thin film of NMC622 -was probed with ex situ XPS before and after 2 cycles between 3.0-4.5 V vs Li/Li + in half cells at 10 mA g −1 . Overall, the CEI comprised of semicarbonates, oligomeric species, and inorganic components formed during the first few cycles which may be (1) discontinuous and/or (2) thin due to the observation of lattice oxygen in the O1s core spectra. The C1s spectra for the pristine and cycled samples can be seen in Fig. 3a as green and blue traces, respectively. The F1s and P2p signals of the pristine material had no peaks above the baseline signal so they are excluded from the figure. A substantial increase in the relative contribution of C-O and O-C-O bonds can be seen after cycling (at ∼286.2 eV and ∼287.1 eV, respectively). There is also a modest increase in the CO 3 signal (∼288.5 eV) from 2 at% to 6 at% after cycling, suggesting that more oligomeric and polymeric species form on the surface relative to semicarbonates or Li 2 CO 3 in the first two cycles. A similar trend was observed on binder-free and carbon-free LiCoO 2 cathodes cycled to 4.6 V vs Li/Li + , 19 indicating the growth of a CEI layer during the first few cycles in high voltage environments. This is reflected in the O1s spectra shown in Fig. 3b, where initially the largest component around 529.5 eV, attributed to lattice oxygen bound to transition metals (M-O), diminishes from 52 at% to 18 at% of the O1s signal after cycling. It is important to note that while the intensity of this peak is substantially lower for the cycled film, the fact that it is still visible suggests that either the CEI is discontinuous or is sufficiently thin for photoelectrons emitted from the lattice can escape the surface layer and reach the detector. The thickness of the surface layer, in this case, would be ∼2 nm, based on the inelastic mean free path of an electron through a polymeric surface layer. 20,21 The O1s spectra shows a greater increase of the oligomer/C-O contribution (532.1 eV) than the carbonate peak (531.1 eV), which was similarly seen in the C1s spectra. An additional peak in the cycled O1s spectrum was assigned to lithium fluorophosphates (Li x PO y F z ) and was confirmed in the F1s and P2p spectra seen in Fig. 3c and Fig. 3d at 687.3 eV and 134.8 eV, respectively. A signal contribution by LiF was detected at 685.6 eV, likely from the decomposition of the electrolyte salt, LiPF 6 , with trace amounts of water in the electrolyte. 22 Partially decomposed or residual LiPF 6 not removed by rinsing the electrodes was detected at ∼688.5 eV. The position of Li salts observed here match well with the spectra for pure Li salts (Li 2 CO 3 , LiPF 6 , LiF) collected on this instrument.
NMC films were coated with binder solutions by spin coating or spray coating before drying overnight under vacuum at 100 o C. The morphology of a PVDF spin coating can be seen in Fig. S1a (available online at stacks.iop.org/JES/167/040521/mmedia). The thin coating provides poor contrast relative to the background, so EDS maps were collected. Binder signals for C and F can be seen in Figs. S1b and S1d, respectively. These signals have relatively low intensity compared to the bulk element signals due to the thin surface layer (<5 nm) compared to the penetration depth of EDS (several μm). The Pt-coated Al 2 O 3 substrate signals are visible in the O, Al, and Pt spectra (Figs. S1c, S1e, and S1f) and have some variation in intensity, likely due to the top layer of 500 nm NMC622. The elements of the bulk NMC622 film can be seen in the Mn, Co, and Ni spectra (Figs. S1g and S1i) and are uniformly dispersed. This morphology is a good representation of the full thin coating of binder on NMC particles in composite electrodes, which would likely have some nonuniformities due to variation in particle size and shape. 15 XPS was used to characterize the surface chemistry of the pristine binder coated films. The spectra for these samples can be seen in Fig. 4 which were fit using reference XPS spectra of pure binder powders collected on our instrument (Fig. S2). A schematic of the three binders selected for this study are included in Fig. 4a as a reference for the bonds expected in the XPS data. The C1s spectra of the pristine binder-coated samples are compared to the uncoated sample in Fig. 4b with the regions corresponding to characteristic bonds of each binder marked. The PVDF-coated sample has primary contributions at ∼285.5 eV and ∼290 eV, corresponding to the CF 2 and CH 2 bonds of the binder, respectively. There is a minor amount of adventitious carbon at ∼284.8 eV used for charge correction in all the samples as well as C-O, C=O, and CO 3 -type species which will be discussed later. The C1s profiles of LiPAA-coated samples are seen in green in Fig. 4b, with a strong COO − at ∼288 eV corresponding to the carboxylic acid group of the binder. The CMC-coated samples are plotted in purple in Fig. 4b and have a characteristic peak for C-O-C and C-O-H bonds at ∼286.6 eV. The relative area of the ether bond signal of the spin-coated CMC films is more intense than that of the spray-coated sample, corresponding to a greater amount of CMC in the XPS sampling volume. This is expected for a continuous thin coating vs a sporadic coating of similar overall thickness and the same trend is observed for the C-F bonds of the PVDF-coated films and the carboxylic acid bonds of the LiPAA-coated films.
These trends are similarly observed in the O1s spectra of Fig. 4c, where the peak for lattice oxygen bonded to a transition metal is denoted as M-O and appears at ∼529.5 eV. The spray-coated spectrum has a 22% lower M-O peak area than the spin-coated sample, meaning that more of the NMC622 film is obscured from detection by overall thicker coverage of spray-coated LiPAA than spin-coated LiPAA. The O1s spectra for CMC and PVDF samples follow the same trend as the C1s spectra, with a substantially lower intensity of the M-O signal in the spin-coated samples compared to the spray-coated samples in each case (28% and 51% lower, respectively). Although the M-O signal is significantly attenuated in these spin-coated samples, it is still visible. Based on the inelastic mean free path of a Ni2p photoelectron in a uniform polymeric coating, 20,21 the thickness of the spin-coated films are approximately 1.8 nm. Furthermore, given the variation in M-O coverage this supports the different distributions of binders on the surfaces.
NMC materials are known to form a surface layer of Li 2 CO 3 when processed in water for 15 min or more via the formation of LiOH and Li 2 O by reaction with water which are rapidly converted to Li 2 CO 3 upon exposure to CO 2 . 23, 24 We investigated whether the short-term (∼1 min) exposure of our NMC thin films to water had a pronounced effect on interfacial chemistry. Deionized (18 MΩ)  water was spin-coated onto a pristine NMC622 thin film with the same procedure as binder solutions and studied with XPS after drying overnight under vacuum at 100 o C. These spectra can be seen in Fig. S3, where the relative contribution of carbonate bonds in the C1s and O1s signals are nearly identical for the pristine and water processed samples. There is a 10% increase intensity of the C-C, C-H peak for the water processed sample, which may be attributed to additional carbon contamination of the surface. These minor differences and the clear Ni2p signal demonstrates that any surface layer formed by water content must be less than the ∼10 nm calculated by others using magnetic susceptibility measurements of NMC cathode powders mixed in water for 15 min or more. 24 There is also no change in the shape of the Ni2p spectra which would be expected if the surface Ni were reduced or the formation of NiOOH. 25 Uncoated and binder coated NMC622 thin films were cycled between 3.0-4.5 V vs Li/Li + for two cycles at 10 mA g −1 (∼C/20) to study the influence of binder coverage on CEI formation and initial cathode performance. The first cycle voltage profiles for each sample are included as a function of specific capacity in Fig. 5a, with filled and hollow symbols corresponding to spin-coated and spray-coated samples, respectively. Except for the CMC coated cathodes, the cells show a sloping plateau at 3.7 V followed by a second plateau at 4.4 V. The 4.4 V plateau appears to be related to the (104) orientation of the cathodes, as discussed later, and not related to the electrochemical oxidation of the electrolyte as this typically occurs at lower voltages (4.3 V) and attributes lower capacities (1 mAh g −1 ). 26 The CMC coated cathode differs significantly as this binder appears to promote the structural rearrangement of the cathode as discussed later. A contributing factor to capacity loss is likely due to surface structural rearrangement and oxygen evolution from the cathode which occurs when cycling NMC622 to 4.5 V vs Li/Li + . 6,27 Specific capacity, coulombic efficiency, and capacity retention between cycles are listed in Table I. The first cycle charge capacity of the spin-coated PVDF and LiPAA samples were 208 and 191 mAh g −1 , which were closest to the theoretical capacity of NMC622 in this study (∼200 mAh g −1 at 4.5 V vs Li/Li + ). The baseline uncoated sample charged to 177 mAh g −1 which was close to that of the spray-coated PVDF and LiPAA samples (165 and 164 mAh g −1 , respectively). The CMC spin and spray-coated samples had the worst first cycle charge capacity overall (112 and 100 mAh g −1 , respectively), but all binder coated samples followed the same trend of spin-coated samples higher first cycle charge and discharge capacities. This is somewhat counterintuitive because one would expect insulating polymers to inhibit Li + diffusion into the cathode. This might be explained by the low rates studied here, where cells were cycled at a constant current of 10 mA g −1 (C/20) until reaching the upper cutoff voltage at which point the voltage was held until the measured charge current decreased below 5 mA g −1 (∼C/40).
The capacity retention between cycles is 27% larger for the spincoated PVDF sample than the uncoated material (Table I). This is emphasized in Fig. 5b, where the green traces corresponding to PVDF-coated samples have retained more of their initial capacity than the baseline and other binder coatings. The PVDF spin-coated sample has retained the distinct sloping plateaus characteristic to Ni 2+ /Ni 4+ redox in NMC whereas most other samples have suppressed plateaus, likely due to polarization of the cell or loss of electrochemical capacity from the surface layer. This may be caused by the build-up of a highly resistive surface reconstruction layer and a layer of electrolyte decomposition products, as observed by others for Ni-rich NMC materials cycled to high voltages. 4, 28 We examine whether this is the case by AC impedance spectroscopy in Fig. 5c. These Nyquist plots are fit to an equivalent circuit model corresponding to the processes occurring in series in this cell: bulk electrolyte resistance to Li + conduction in solution, R e , desolvation of Li + at the interface-referred to here as charge transfer resistance, R ct and Q dl , Li diffusion through the CEI, R CEI and Q CEI , and a constant phase element for Li diffusion into the thin film, Q b

29-31
Charge transfer resistance is sometimes attributed to a convolution of effects at the interface (i.e. intercalation, diffusion through SEI, desolvation), but work by Xu et al. demonstrated the serial nature of charge transfer resistance as desolvation of Li + followed by diffusion through the interface. 31,32 Therefore, we employ discrete elements for these processes in our equivalent circuit.
The bulk electrolyte and charge transfer resistances are relatively small and comparable between samples (12-54 Ohm * cm 2 ), but the medium frequency semicircle corresponding to R CEI varies significantly between samples depending on coating morphology and chemistry. The cycled PVDF spin-coated and spray-coated samples had R CEI of 399 Ohm * cm 2 and 285 Ohm * cm 2 , respectively, lower than the 1650 Ohm * cm 2 of the baseline. At 910 Ohm * cm 2 , the LiPAA spin-coated had a slightly lower R CEI than the baseline, whereas the LiPAA spray-coated sample had a higher value of 1762 Ohm * cm 2 . Both CMC samples had higher R CEI than the baseline, at 1864 Ohm * cm 2 and 1745 Ohm * cm 2 for the spin and spraycoated samples, respectively. These values match the general trend of the specific capacity and capacity retention data, with lower R CEI corresponding to a higher specific capacity. This contrasts previous work where LiPAA was used as an artificial CEI in composite electrodes of LNMO, providing lower interfacial resistance compared to PVDF. 13 This is likely due to the difference in surface chemistry of NMC622 (e.g. oxygen evolution from the surface at high upper cutoff voltages), which will be discussed in the next section.
XPS was employed to determine whether the differences in interfacial impedance could be related to the morphology or compositions of the binders and CEI between samples. No F1s or P2p component spectra are shown for pristine uncoated or pristine CMC or LiPAA-coated samples because there were no distinguishable peaks in those regions. Components of all spectra were deconvoluted using reference spectra of the pure binders and common Li salts, collected on our instrument. Due to the presence of similar binding energies between the binders and decomposed species (e.g. ether bonds in CMC and oligomeric C-O from EC decomposition) absolute identification of the quantity of certain species is ambiguous and will be compared qualitatively. Components which do not have significant overlap with signals observed in the pure binder samples will be compared quantitatively (i.e. Li 2 CO 3 , Li x PO y F z , and LiF signals in the C1s, P2p, and F1s spectra, respectively).
The P2p and F1s data for the three binder compositions and two coating techniques are compared to the uncoated material before and after cycling in Fig. 6. The F1s spectra of Fig. 6a show that pristine PVDF-coated samples have a primary peak at ∼687.5 eV corresponding to the CF 2 bonds of the binder which was fit to the FWHM of the pure binder spectrum. There was also a minor peak at ∼685.1 eV for the spin-coated PVDF which may be due to trace LiF. No signals were detected above the background for the F1s and P2p regions of the other pristine samples, so they are excluded from Fig. 6. The F1s spectra of cycled PVDF-coated samples exhibit two component peaks corresponding to salt decomposition products: LiF and Li x PO y F z at ∼687 eV, which aligns with what was observed in the O1s spectra. An additional minor contribution around ∼689.2 eV corresponds to residual LiPF 6 on the sample surface. Samples were rinsed in DMC before analysis, so residual salt may have been loosely bound in the CEI or as a precipitate on the surface. The CF 2 signal of the binder was still visible in the cycled samples with a relative intensity of 63 at% in the F1s spectra of the cycled  spray-coated sample compared to 37 at% for the spin-coated sample. The difference in signal attenuation suggests that the larger binder agglomerates on the spray-coated samples were less obscured by the CEI than the spin-coated PVDF samples. This means that the PVDF either coexists within the CEI layer (for the spray coated sample) or has a thin coating of decomposition products deposited on top of the binder layer (for either the spin or spray-coated samples). The cycled spin-coated PVDF sample had noticeably more LiF when compared to the other cycled samples in Fig. 6. Decomposition of salt species can be confirmed by considering the P2p spectra of cycled samples in Fig. 6b. For both PVDF-coated samples and the uncoated sample, Li x PO y F z is detected at ∼134.8 eV and residual LiPF 6 can be seen at ∼137.4 eV.
The same species are observed of the PVDF sample analysis can be observed in the LiPAA F1s spectra of Fig. 6c except for the CF 2 bond due to the difference in binder chemistry. In this case, the cycled uncoated sample has a higher LiF:Li x PO y F z ratio of 2.0 when compared to the cycled LiPAA-coated samples with 1.6 and 1.5 for spin and spray-coating, respectively. The P2p spectra of Fig. 6d show a ratio of Li x PO y F z to LiPF 6 which was slightly larger between the LiPAA-coated samples (1.4) than the uncoated sample (0.96).
The cycled CMC-coated samples had a different overall F1s peak ratios than the uncoated sample and the other binder samples. The CMC F1s spectra seen in Fig. 6e demonstrates similar concentrations of LiF and Li x PO y F z , whereas the baseline sample had predominantly LiF. Additionally, the CMC-coated samples have the opposite trend of other samples for the P2p data in Fig. 6f, with a Li x PO y F z to LiPF 6 ratio of 0.67 and 0.59 for the spin and spray-coated samples, respectively. The comparable amounts of LiF and Li x PO y F z at the surface of CMC demonstrates a different CEI chemistry compared to the other samples which correlates with the poorer cycling performance discussed earlier. To summarize the F1s and P2p XPS data, salt decomposition products were observed in the spectra of all cycled samples. The PVDF-coated samples had a greater ratio of LiF:Li x PO y F z than all other samples, and the cycled LiPAA and CMC-coated samples all had similar ratios of LiF:Li x PO y F z which were lower than the baseline (cycled uncoated) sample. Figure 7 shows the core level spectra of C1s and O1s for the three binder compositions and two coating techniques compared to the uncoated material before and after cycling. Similar bonds are detected in the C1s spectra of both the spin and spray-coated PVDF samples shown in Fig. 7a. There is a appearance of -CO 3 species (∼288.5 eV) after cycling for the PVDF-coated and uncoated samples indicate the presence of Li 2 CO 3 or semicarbonates formed by oxidation of EC. 33 The characteristic CF 2 bond of PVDF was detected at ∼290 eV in the pristine samples and was depressed in the cycled samples in good agreement with the F1s spectra of Fig. 6a. O-C-O/ C=O and C-O bonds were detected at ∼287.1 eV and ∼286.2 eV, respectively. These likely correspond to oligomers and lithium alkoxide species (ROLi), which can form from EC oxidation followed by a ring-opening polymerization reaction. 33,34 Compounds appear in a narrow range in the O1s spectra seen in Fig. 7b, with the lowest energy peak at ∼529.5 eV attributed to lattice oxygen (M-O), and ROLi species assigned to ∼531.0 eV. CO 3 and O-C=O groups were observed at ∼531.9 eV, and oligomer (C-O) species were assigned to ∼533 eV. The highest energy component appears at ∼534 eV and corresponds to Li x PO y F z species. Cycled PVDF-coated samples had more intense M-O signals of 45 at% and 32 at% for spin and spray-coated PVDF, respectively, compared to the 18.5 at% of the uncoated sample, indicating a thinner CEI or more exposed active material surfaces. The Li x PO y F z contribution in the O1s spectra of the uncoated sample was larger than both the spin and spray-coated PVDF-coated sample, as was the case in the F1s spectra of Fig. 6a. Such fluorinated species necessarily originate from the decomposition of the electrolyte salt, There are similar C1s moieties detected for the uncoated sample and the LiPAA-coated samples in Fig. 7c. There C-C/C-H intensity is greater for the LiPAA-coated samples, which can be attributed to the bonds of the binder. This is reflected in the O1s spectra of Fig. 7d, in which the greater contribution around 532 eV in the LiPAA-coated samples aligns with the primary component of the O1s spectra for the pure binder (Fig. S2).
The CMC-coated samples had larger contributions of C-O and O-C-O/C=O components than the baseline, as seen in Fig. 7e. There was a substantial decrease in the relative concentration of M-O for the CMC-coated samples compared to the uncoated sample's O1s spectra in Fig. 7f. This could be attributed to the formation of a thicker CEI, given that the M-O signal was visible in the coated samples before cycling.
After cycling, a metal oxide (M-O) signal is visible in the XPS O1s spectrum for each sample and can be seen in Fig. 8a. This signifies either a discontinuous CEI or a thin enough surface layer for electrons to escape the NMC surface and reach the detector. The intensity of these signals matches the attenuation of Ni2p spectra in Fig. S4. Such a thin CEI is in line with previous reports of the CEI. 4,35 The M-O signal for the uncoated sample is 18% of the overall O1s signal, which is substantially lower than the spin and spray-coated PVDF samples at 45% and 32%, respectively. This suggests a thinner overall coating on the PVDF-coated samples because the combined signal effects of the PVDF coating and CEI on those samples suppresses the M-O signal less than the CEI alone on the uncoated sample. In contrast, the relative amount of M-O detected from the LiPAA-coated samples is close to that of the baseline. The CMC-coated samples had low relative M-O signals, indicating a thicker or more attenuating surface layer. These trends match the impedance trends observed in Fig. 5c, suggesting that lower interfacial resistance and higher capacity retention between cycles might be due to a thinner overall surface layer. Figure 8b shows graphically the relative atomic percent of different bonds detected at the surface of each cycled cathode thin film by XPS. Quantification of at% is based on the area of each component peak fit to the raw data and weighted according to the relative sensitivity factor of each element based on Scofield cross sections, the transmission factor of the instrument, and the inelastic mean free path of an electron at the given binding energy. The active material and electrolyte salt contributions were excluded for clarity in comparing decomposed species present in the CEI. Bonds with significant overlap with binder contributions were also excluded to reduce ambiguity in discussing common bonds shared between CEI components and the binders (e.g. -COO − of ROLi and LiPAA). This method of comparison necessarily omits potential CEI species which overlap with binder binding energies (e.g. ROLi, oligomeric C-O, carboxylates) and does not account for any species formed below thick regions of binder which attenuate signals of surface species. The remaining components-Li 2 CO 3 /polycarbonates, Li x PO y F z , and LiF -do not have significant overlap with signals observed in the pure binder samples for their respective spectra: C1s, P2p, and F1s. These species are normalized in Fig. 8b to compare their relative proportion in the CEI. The majority of this three species ratio of the CEI on each sample is comprised of LiF, except for the CMC samples which have 48 at% and 45 at% LiF for the spin and spray-coated samples, respectively. There is a larger proportion of Li 2 CO 3 /RCO 3 observed on the binder coated samples compared to the uncoated film, which might be attributed to processing conditions for the aqueous binders (LiPAA and CMC), but the increase for the PVDF samples as well suggests a mechanistic role of the binders. There is slightly more Li x PO y F z observed on the spray-coated samples compared to the spincoated samples in each case, although the PVDF-coated samples show the least relative amount of Li x PO y F z overall. The PVDF samples also exhibit the largest relative LiF signals after the uncoated sample, which might contribute to their higher capacity retention when considered alongside the M-O data in Fig. 8a.
The role of LiF in passivating interfaces is not well understood, 36 although a thin layer of predominantly LiF has been observed as an effective passivation layer for NMC333 21 as well as high voltage operation of LMNO. 37 In this case, the CEI formed on PVDF-coated  samples may differ from the LiPAA and CMC-coated samples due to a difference in water content at the interface. While these samples were dried in a vacuum oven, residual water may have been trapped at the surface by the aqueous binders (LiPAA and CMC). This water could be hydrolyzed by the strong Lewis acid PF 5 according to Eqs. 1 and 2 to form POF 3 . Residual Li 2 O at the surface of the cathodes could then react to form lithium fluorophosphates: 38,39 POF sol.
Li O s LiF s Li PO F s 3 The increased proportion of Li x PO y F z observed on the LiPAA and CMC-coated samples as well as the increased thickness of the CEI can be explained by the additional water content at the surface. This highlights the importance of minimizing water content at the surface for high voltage cathode materials. The improved performance and surface chemistry of the PVDF-coated NMC622 over the LiPAA-coated samples somewhat contradicts the observation of previous work on LiPAA-coated LNMO composite electrodes which had reduced charge transfer resistance and similar capacity at low rates compared to PVDF-coated samples. 13 This difference in observations of performance might be attributed to different mechanisms of CEI formation in LNMO compared to Ni-rich NMC. The crystal structure of LNMO is known to remain stable when cycling to voltages up to ∼5 V vs Li/Li + , at which electrochemical oxidation of the electrolyte becomes the primary source for CEI components. Ni-rich NMC systems, on the other hand, undergo structural rearrangement of the cathode and release oxygen from the lattice, causing chemical oxidation of the electrolyte.
Beyond the changes in interfacial chemistry there were significant changes in the structure of the cathode that varied with respect to the polymer binder. Figure 9 shows XRD data collected for a pristine electrode and electrodes cycled twice from OCV to 4.5 V → 3.0 V. The pristine electrode shows clear evidence of a 104 texturing, as evident by the large (104) peak, along with the standard (003) peak from the NaFeO 2 -type layered structure. Upon charging the (003) and (104) peaks are reduced for the PVDF and LiPAA coated cathodes and eliminated for the CMC coated cathode. This degradation/loss of structure corresponds with the capacity retention observed in Fig. 5. Indeed, the complete loss of capacity of the CMC coated cathode is consistent with the voltage profile of the electrode not demonstrating a plateau at 3.7 V.
Within the literature there is evidence that ethylene carbonate preferentially reacts with the lattice oxygen atoms exposed in the (104) planes of layered cathodes NMC333 and LiCoO 2 resulting in vacancy sites. 40 These vacancies promote structural rearrangements resulting in a disordered rock salt and amorphous structures 41 which have significantly different voltage profiles. This structural rearrangement has been observed as an additional plateau in the voltage profile of NMC systems. An in situ XRD study of NMC622 by Doeff et al. included a minor plateau around 4.35 V, as did a study of NMC442 and NMC532 by Dahn et al. This corresponds to the 4.3-4.4 V plateau observed in our system, although they are not as significant as the regions observed here, likely due to the texturing of these films.
Our data indicates that the (104) orientation does undergo a preferential decomposition mechanism as predicted by theory. Furthermore, the binder plays a role in this process. Specifically, we think the solvent interaction with the binder promotes (CMC, PAA)/hinders (PVDF) EC interaction with the surface. In addition, the preferential formation of LiF from PVDF and LiPAA acts to impede the EC promoted decomposition of the structure while Li x PO y F z is not effective at blocking EC interactions with the surface. This hypothesis is supported by our recent work on solid state batteries which show the solid electrolyte Lipon effectively prevents the 4.4 V reaction and promotes (but not completely) structural retention over 100 cycles to 4.5 V. 42 This work supports theoretical work on optimal binder content in composite electrodes being a thin layer of physisorbed or chemisorbed polymer (1-5 nm) with minimal excess binder agglomerated with conductive carbon. 43 This role of the binder might be extended to other cathode systems which suffer from structural degradation at high voltages, particularly from EC-driven decomposition. One way to reduce degradation is through coating of active material with a thin uniform layer of PVDF which promotes the formation of a LiF-rich CEI layer. Additionally, the structural rearrangement of (104) planes in NMC could be suppressed through preferential modification of those surfaces by polymer coatings or crystal engineering. Bulk cathode processing could be tailored to achieve this morphology and chemistry by adding a dry mixing step of the active material and binder. Finally, the data supports that the PVDF may be participating in CEI formation and could be a critical component to this reaction.

Conclusions
In this study, we examined the effects of binder morphology and composition on the formation of the CEI on Ni-rich NMC622 thin film electrodes and the degradation of the electrode structure. Spin and spray coating of PVDF, LiPAA, and CMC allowed for experimental control of the planar interface such that the surface chemistry of cycled films could be probed with XPS. The PVDF-coated samples had the thinnest CEI which were comprised of greater relative concentrations of LiF and Li 2 CO 3 than Li x PO y F z when compared to the other samples. CMC had the thickest CEI and a more even composition of the three species of interest, whereas LiPAA had a similar CEI thickness to the uncoated sample with similar surface chemistry. These trends aligned with capacity retention between formation cycles and the substantially lower interfacial impedance of PVDF-coated samples. All binder coated samples had more Li 2 CO 3 on the surface than the uncoated sample, which might be attributed to processing conditions. This work demonstrates that the presence of binder on the interface of cathode active materials directly influences the chemistry of CEI formation which impacts the performance of the cell. The presence of residual water trapped at the interface by binders changes the composition and thickness of CEI, so a good binder should provide adequate adhesion between active material particles and conductive carbon while minimizing water introduced to the interface. As such, close attention to processing conditions and binder selection is imperative for the application of high voltage cathode materials such as Ni-rich NMC.
contract DE-AC05-00OR22725, was sponsored by the Office of Energy Efficiency and Renewable Energy (EERE) Vehicle Technologies Office (VTO) (Deputy Director: David Howell) Applied Battery Research subprogram (Program Manager: Peter Faguy) (N.D.P., C.D., and G.M.V.). N.D.P. also thanks Beth Armstrong for constructive feedback on the structure of this report, Robert Sacci for assistance on FTIR, and Katie Browning for assistance on spin coating and insightful discussions.This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (https://energy.gov/downloads/doepublic-access-plan).