XPS Investigation of the Electrolyte Induced Stabilization of LiCoO 2 and “AlPO 4 ”-Coated LiCoO 2 Composite Electrodes

The “AlPO 4 ” coating has been shown to improve the electrochemical performance of LiCoO 2 batteries. We previously showed that the “AlPO 4 ” coating promotes the formation of metal ﬂuorides, which could act as a stable surface ﬁlm and protect LiCoO 2 from continuous degradation upon cycling. In this work, we removed the ﬂuorine source in the LiPF 6 salt by using the LiClO 4 salt and investigated the effectiveness of the “AlPO 4 ” coating. Interestingly, the “AlPO 4 ” coating was found to improve the voltage efﬁciency and capacity retention when cycling in the LiPF 6 electrolyte, but was detrimental when cycling in the LiClO 4 electrolyte. XPS revealed that the “AlPO 4 ” coating promotes the formation of metal ﬂuoride in both electrolytes, with the surface ﬁlm formed in LiClO 4 being more electrically resistive compared to that formed in LiPF 6 . The source of ﬂuorine in the coated electrode cycled in LiPF 6 is largely attributed to the LiPF 6 salt whereas the source of ﬂuorine in the coated electrode cycled in LiClO 4 is the binder PVDF. We believe that the coating could react with HF impurity in the LiPF 6 electrolyte or from the binder PVDF and form stable metal ﬂuoride ﬁlms on the surface. vs. Li + if cycled to than 4.2 in capacity loss. 3,4 Structural instability 3,5 and reactivity of the cathode with the electrolyte 3 proposed as modiﬁcation

have shown improvement when cycled to high voltages compared to bare LiCoO 2 positive electrodes.However, the origin responsible for the increased performance is not well understood.Understanding the mechanism responsible for the enhancement in cycling stability provided by the coatings/surface modification is essential to further stabilize positive electrode materials for applications that demand high cycle life such as electrical vehicles and stationary storage.
Enhanced performance in cycling associated with surface coating has been attributed to phase transitions, 3,4,[12][13][14][15] coatings, 3,5,15,16 suppression of electrolyte decomposition, reduced oxygen activity associated with cation mixing, 3,[17][18][19] physical barriers and HF scavenging, 6,[19][20][21] suppression of side reactions, [7][8][9][10]22,23 electrolyte acidity, 16,24 and the PVDF binder as a source of HF. 25,26 The recent investigation by Dahèron et al., 18 showed that the substitution of Al for Co not only increases the ionic nature of the Co-O bond through orbital mixing, but also that the substitution reduces the basicity of the LiCoO 2 surface, thus making the surface less receptive or vulnerable to acidic attack in the electrolyte. Howevr, this benefit can be temporary as recent work indicates that the Li-Al-Co-O surface region is consumed during cycling. 27 Th "AlPO 4 "-coating on LiCoO 2 can induce the formation of Al-rich regions (LiAl y Co 1-y O 2 ) and P-rich regions (Li + conducting Li 3 PO 4 ), 28 and that Co-and Alcontaining oxyfluoride species and species like PF x (OH) y developed on the surface during cycling can lower the rate of Co dissolution and deposition of Co-containing species on the negative electrode and further degradation of LiPF 6 .27 For a more detailed discussion of the metal fluoride formation mechanism, the reader is directed to Figure 12 and the corresponding discussion in the report by Lu et al. 27 In this paper, we examine the influence of the fluorine source on the formation of metal fluorides and the efficacy of the surface coating by relating cycling performance to the surface chemistries of bare and "AlPO 4 "-coated LiCoO 2 electrodes cycled in 1 M LiPF 6 or LiClO 4 in EC:DMC (1:1) to 4.6 V.For a more detailed discussion of the metal fluoride formation mechanism, the reader is directed to Figure 12 and the corresponding discussion on page 4421 (second column) of Lu et al. 27 The LiClO 4 salt was used to help determine if the formation of the metal fluorides was a product of the interaction between the active material with the LiPF 6 electrolyte and/or the PVDF binder. The cntribution of the F-containing LiPF 6 salt on the efficacy of the surface coating in contrast to the non F-containing LiClO 4 salt will be discussed.

Experimental
Bare LiCoO 2 and "AlPO 4 "-coated LiCoO 2 powder samples were prepared as previously described. 28The reversible capacities and cycling stability of bare and "AlPO 4 "-coated LiCoO 2 composite electrodes were measured by using a two-electrode lithium cell (Tomcell type TJ-AC).Lithium cells were constructed inside the glove box (H 2 O, O 2 < 0.1 ppm, Mbraun, USA) using a lithium metal foil as the negative electrode and the composite positive electrode separated by two polypropylene microporous separators (Celgard 2500).The electrolytes used were 1 M LiPF 6 or 1 M LiClO 4 (battery grade, dry, 99.99% trace metals basis, Aldrich) in a 3:7 volume ratio EC/DMC solvent (Kishida Chemical Corp, Japan).The assembled lithium cells were set aside for 6 h prior to electrochemical testing on a Solartron 1470 battery testing unit in order for the electrolyte to wet the electrodes.The cells with bare and coated LiCoO 2 electrodes were cycled at the C/5 rate between voltage limits of 2.0-4.6 V with 4 h hold at 4.6 V for 20 cycles.The C-rate was defined based on the theoretical capacity of LiCoO 2 (274 mA h g −1 ).The discharged cells were disassembled in the glove box, after which the Li x CoO 2 electrodes were extracted and stored in hermetically sealed containers in the glove box.
The XPS spectra of bare and "AlPO 4 "-coated LiCoO 2 electrodes before and after cycling were measured using a Physical Electronics model 5400 X-ray photoelectron spectrometer.Each electrode was removed from the hermetically sealed container in an argon-filled glove box (< 3 ppm O 2 , < 3.5 ppm H 2 O) and mounted onto the sample puck with the aid of copper clips placed on the edge of the electrode.The electrode was then transferred via a sealed chamber to the introduction chamber of the XPS spectrometer without exposure to ambient conditions.The introduction chamber was evacuated for 10-15 min before the sealed chamber was opened, then the sealed chamber was opened and evacuated for another 10-15 min before transferring the sample to the analysis chamber.The data were collected at room temperature using non-monochromatic Al Kα (1486.6 eV) X-ray source operating at 400 W (15 kV and 27 mA).Data collection proceeded when the pressure in the analysis chamber reached ≈2×10 −8 Torr.Survey spectra (0 -900 eV at an electron takeoff angle of 45 • ) were collected at analyzer pass energy of 89.45 eV with energy resolution of 1.34 eV, 0.5 eV/step and an integration interval of 50 ms/step.The survey spectra consist of the average of 20 cycles with total acquisition time of 30.42 min.Multiplex spectra were collected at analyzer pass energy of 35.75 eV with energy resolution of 0.54 eV, 0.2 eV/step and an integration interval of 50 ms/step.The binding energies for the pristine (bare and coated) electrodes are reported as measured since there was no indication of electrode charging as indicated by the binding energy of the C 1s line for carbon black at 284.4 eV, which is in close agreement with a number of previously reported results. 29,30The cycled electrodes displayed a small degree of charging in the range of 0.2-0.4eV, but their binding energies are reported with respect to carbon black at 284.4 eV.The binding energies of reference compounds in powder form are reported with respect to hydrocarbon at 285.0 eV.These procedures yielded the same binding energies for the C 1s and F 1s lines for powder PVDF and an electrode, which consisted of a mixture of PVDF and carbon black.
Depth profile analyses were made using 4 keV Ar ions with a raster size of 4×4 mm 2 .The sputtering was made in 0.5 or 1 minute intervals for a total sputtering time of 6 minutes.The C 1s, O 1s, F 1s, Li 1s, Co 2p, P 2p and Cl 2p photoemission lines were collected after each interval of sputtering at analyzer pass energy of 35.75 eV and an energy increment of 0.2 eV.The sputtering rate was calibrated using a 1000 Å SiO 2 film on a Si substrate and was found to be in the range of 25-30 Å of SiO 2 per minute.
Another set of XPS spectra were collected on beamline X24A of the National Synchrotron Light Source utilizing monochromatic Xrays with energy of 2555 eV compared to 1486.6 eV for Al K α X-rays in order to increase the depth of the analyzed region.The experimental setup consisted of a 2-stage high vacuum chamber with base pressure below 10 −7 Torr and a Scienta R4000 detector running in transmission mode at pass energy of 200 eV with 0.3 mm analyzer aperture.The pass energy and aperture combination gives a resolution of about 0.29 eV at beam energy of 2139 eV as measured by the width of the Ag Fermi foot; though it will change with energy.The spectra were collected at an electron takeoff angle (between surface and detector) of 85 • .The angle between the beam axis and the detector is 90 • .
The sampling depth, defined as three times the electron mean free path, of C 1s, O1s and F 1s photoelectrons generated by Al X-rays (1486.6 eV) at an electron takeoff angle of 90 • relative to the plane of the sample (SD 90 ) are 8.7, 7.3 and 6.3 nm, respectively. 31Note that (1) the sampling depth at an electron takeoff angle of θ (SD θ ) is equal to SD 90 * sin(θ) and ( 2) the sampling depth is proportional to (E) m where E is the kinetic energy of the photoelectrons and the exponent m for many elements and compounds is very close to 0.77. 32,33Hence, the sampling depth of Li 1s, C 1s, O 1s, F 1s, P 2p and Co 2p photoelectrons generated with Al X-rays at an electron takeoff angle of 45 • are estimated to be 7.0, 6.1, 5.2, 4.5, 6.7 and 4.1 nm, respectively.On the other hand, the sampling depth of Li 1s, C1s, O 1s, F 1s, P 2p and Co 2p photoelectrons for the synchrotron X-rays with excitation energy of 2555 eV at an electron takeoff angle of 85 • are 15.2, 14.1 12.9, 12.2, 14.9, and 11.7 nm, respectively.

Results and Discussion
Electrochemical performance.-The"AlPO 4 " coating was found to improve the voltage efficiency and capacity retention when using LiPF 6 electrolyte, but was detrimental when cycled in the LiClO 4 electrolyte.As shown previously, the galvanostatic voltage profiles of lithium cells having bare and "AlPO 4 "-coated LiCoO 2 electrodes revealed that the coated electrode exhibits higher capacity retention in comparison to the bare electrode upon cycling in 1 M of LiPF 6 in EC:DMC between 2.5-4.7 V vs. Li + /Li with 4 hours hold at 4.7 V. 27 The galvanostatic voltage profiles of lithium cells having bare and "AlPO 4 "-coated LiCoO 2 electrodes cycled in LiPF 6 or LiClO 4 between voltage limits of 2.0-4.6 V vs. Li + /Li with 4 h hold at 4.6 V are shown in Figure 1.Clearly, the coating as well as the type of salt used during cycling has a significant effect on the degree of capacity retention and the degree of polarization.Variations  These binding energy values were obtained from the referenced literature or determined via standards where noted.It should be noted that the y-axis is the same for all spectra but were shifted vertically for clarity.
in discharge capacity (mAh/g), discharge energy (Wh/kg) and energy efficiency (percent of discharge energy to charge energy) with cycle number are shown in Figure S1.The coated electrode cycled in LiPF 6 has the highest capacity retention.It is interesting to note that the bare electrode cycled in LiClO 4 has higher capacity retention than the bare electrode cycled in LiPF 6 or the coated electrode cycled in LiClO 4 .Hence, the coating doesn't provide an advantage when cycling was performed in LiClO 4 electrolyte.The advantages of the "AlPO 4 " coating are more dramatic when one examines the discharge energy and energy efficiency (Figure S1).The discharge energy and energy efficiency for the 20 th cycle decreased in the following order: coated electrode cycled in LiPF 6 (≈770 Wh/kg, ≈90%), bare electrode cycled in LiClO 4 (≈500 Wh/kg, ≈63%), coated electrode cycled in LiClO 4 (≈300 Wh/kg, 56%), bare electrode cycled in LiPF 6 (≈300 Wh/kg, ≈54%).
It is known that structural instability can occur when LiCoO 2 is cycled to high cutoff voltages (>4.2 V) and can cause degradation in battery performance and capacity retention.Previously, based on synchrotron X-ray diffraction data, we were able to show that the "AlPO 4 "-coating prevented any structural instability, even when cycling in LiPF 6 electrolyte to 4.7 V. 27 Therefore, we are attributing the observed reduction in capacity retention to surface chemical changes and not to lattice instability when comparing the coated electrode cycled in LiPF 6 versus the coated electrode cycled in LiClO 4 .
Scanning electron or transmission electron microscopy micrographs of the cycled active material to look for cracks or degradation of the primary particles of LiCoO 2 were not taken.Though we believe a mechanical degradation study to be of interest, our extensive literature search suggested that LiClO 4 is commonly used as a model electrolyte salt because it does not mechanically damage the active particles.A study by Ozawa, Yazami, and Fultz 34i ndicated that LiClO 4 caused some measure of lithium re-intercalation, but no mechanical damage, when the cathode material was thermally aged for several weeks.Aurbach et al. 35 found that Fe dissolution in LiClO 4 was negligible even at elevated temperatures due to the lack of acidic contaminants and that Co dissolution occurred in LiPF 6 based electrolytes.Hereafter, we describe the surface chemical changes evolved after cycling in LiClO 4 and LiPF 6 as determined by XPS.
XPS/Surface chemistry.-Allcycled electrodes were examined in the discharged state.A comparison of survey spectra of various electrodes is shown in Figure S2.The C 1s, O 1s, Co 2p and F 1s spectra are shown in Figures 2-5, respectively.The Li 1s, P 2p, Cl 2p and Al 2p spectra are shown in Figures S3-S6, respectively.The concentrations based on these spectra are provided in Table I.The compositions of the near surface region for the pristine bare and coated electrodes are dominated by carbon, fluorine and then oxygen.The binding energy, full width-at-half-maximum (FWHM) and concentrations (in atomic percent) of various elements for each proposed chemical state are summarized in Table SI.The depth profile spectra of the O 1s, Co 2p, C 1s and F 1s regions are shown in Figures S7-S10, respectively, with concentrations based on these spectra are shown in Figure S11.Hereafter, we discuss the results for each element and correlate surface chemistry with electrochemical performance characteristics.
C 1s region.-Thesurface film developed in coated electrodes were found to be thicker with more oxygenated nature compared to that developed in bare electrode in both electrolytes based on the changes in C 1s intensity and binding energy.The C 1s photoemission spectra of pristine bare (Figure 2a) and pristine coated (Figure 2d) electrodes are dominated by the contributions from carbon black and PVDF.Various carbon chemistries are highlighted by As previously described, reference spectra of Super P carbon black, PVDF, and a composite electrode of 50 wt% PVDF and 50 wt% Super P carbon black were used as a guide to deconvolute the spectra of pristine composite electrodes. 27This procedure resulted in satisfactory fits for all of the cycled electrodes, except for the coated electrode cycled in LiClO 4 .In this case, the binding energies for all components were allowed to vary except those for carbon black, where the binding energy separation and FWHM for the main peak and its satellite were constrained to the reference electrode values.The additional small components, which we assign to -CHF and -CF 3 , were also observed for a powder sample of PVDF and we believe are at least, in part, due to sample damage by the X-rays.
As can be seen from the data in Figure 2 and Table SI, the concentration of carbon black (≈284.4eV) for the bare and coated electrodes cycled in LiPF 6 and LiClO 4 decreased significantly relative to concentrations for the pristine electrodes, which can be related to the buildup of surface films on the electrode surfaces.Both bare electrodes (cycled in LiPF 6 and LiClO 4 ) have a larger contribution of the carbon black signal compared to their coated coun-terparts.The sub-peaks associated with the CH 2 -CF 2 chemistries (≈286.1 and 290.6 eV) for the cycled electrodes have concentrations close to those observed for the pristine electrodes except in the case of the bare electrode cycled in LiPF 6 , which has significantly lower contributions.Furthermore, the broader nature of these sub-peaks for the cycled electrodes relative to the pristine electrodes may indicate the presence of multiple chemistries and therefore these sub-peaks may not only be associated with -CH 2 -CF 2from PVDF, but also with C-O and Li 2 CO 3 type of species, respectively.For instance, Dedryvère et al. 36 assigned a peak at 286.0 eV as oligomer species of poly(ethylene oxide) type, (CH 2 CH 2 O) n , resulting from the decomposition of ethylene carbonate in the electrolyte.Another investigation by the same group 37 38 The third sub-peak for the coated electrode cycled in LiClO 4 is again at a higher binding energy (≈288.8eV), which can be related to a carboxylic chemistry.The final sub-peak at ≈293.1 eV is assigned to CF 3 chemistries.This sub-peak contribution is relatively weak for all electrodes, though the bare and coated electrodes cycled in A relative estimate for the thickness of the surface films can be made based on the reduction of the C 1s intensity for carbon black in the cycled electrodes relative to that in the pristine electrodes.It is well known that a thin over layer with a uniform thickness will attenuate the intensity of the photoelectrons from the underlayer following the relationship where d is the thickness of the over layer, λ c is the mean free path of the photoelectrons, and θ is the electron takeoff angle relative to the plane of the sample which is 45 • in this experiment with Al Xrays. 39,40For C 1s photoelectrons using Al X-rays, λ c is ≈2.9 nm 31 and, hence, the overlayer film thicknesses are estimated to be ≈ 4.2 and 1.6 nm for the bare electrode cycled in LiPF 6 and LiClO 4 , respectively, and ≈ 5.3 and 4.4 nm for the coated electrode cycled in LiPF 6 and LiClO 4 , respectively.It should be noted, and stressed, that a uniform thickness or composition may not be the case for the cycled electrodes.Therefore, these estimates are a crude measure for film thickness and may be compared on a relative basis.It should also be noted that XPS measurements offer a non-destructive means of measuring the film thickness without interfering with the chemical integrity of the film.
O 1s region.-TheO 1s photoemission spectra for the pristine bare (Figure 3a) and coated (Figure 3d) electrodes display prominent contributions from LiCoO 2 lattice oxygen (O 2− ), surface oxygen (O − ), and Li 3 PO 4 at 531.5 eV based on our reference (in the case of the coated electrode) in good agreement with previously reported results. 27,28The lattice oxygen at ≈529.5 eV, the surface oxygen species O-C = O * and Li 2 CO 3 collectively at ≈532.1 eV, the O * P(OR) 3 and O * -C = O species at ≈533.5 eV and Li x PF y O z at ≈534.6 eV are identified by the dashed lines (Figure 3).
The sub-peak at 529.8 eV for the bare electrode cycled in LiPF 6 is close to that observed for our Co(OH) 2 reference at 530.1 eV indicating that surface Co is present in the form of Co(OH) 2 in agreement with the assignment based on the Co 2p region.On the other hand, the sub-peak at 529.9 eV for the bare electrode cycled in LiClO 4 is similar to that observed for our CoO reference at 529.9 eV indicating that surface Co is present in the form of CoO in agreement with the assignment based on the Co 2p region.The differentiation between Co(OH) 2 and CoO was based mainly on the Co 2p region as will be discussed later in the text.The sub-peak at 530.2 eV for the coated electrode cycled in LiPF 6 , cannot be attributed to LiCoO 2 , CoO or Co(OH) 2 as the Co 2p region confirmed the absence of these phases.Recent work by the Edström group 41,42 45 Clearly the literature indicates that this sub-peak is associated with oxygen as a polyanion.It is to be noted that such a low binding energy component is absent in the case of the coated electrode cycled in LiClO 4 .
The O1s component located between 531.6 -532.2 eV can be attributed to the surface oxygen of LiCoO 2 and oxygen in Li 2 CO 3 .Contributions at ≈531.6 eV (pristine bare electrode) and ≈531.9 eV (pristine coated electrode) are significantly reduced upon cycling in LiPF 6 .This contribution, however, is significantly increased upon cycling the bare electrode in LiClO 4 , but is absent in the case of the coated electrode cycled in LiClO 4 .The binding energy of this component increased upon cycling from 531.6 to 531.9 eV (bare electrode in LiPF 6 ), 531.6 to 532.1 eV (bare electrode in LiClO 4 ), and 531.9 to 532.2 eV (coated electrode in LiPF 6 ) indicating that this subpeak is dominated by contributions from carbonate chemistry and/or oxygen doubly bonded to carbon (C = O) upon cycling.The sub-peak at ≈533.7 eV, which can be associated with oxygen singly bound to carbon (RO * -CO-O * R) in carbonate, emerged at a moderate level upon cycling bare and coated electrodes in LiPF 6 and bare electrode in LiClO 4 , which has the highest contribution of this component.
The high binding energy sub-peak at ≈535.5 eV, seen in the pristine bare electrode, the bare electrode cycled in LiClO 4 , and the bare electrode cycled in LiPF 6 , was not observed in any of our reference compounds.This sub-peak has been reported to be the result of carbonate impurities on the surface of the LiCoO 2 particles. 46This contribution emerged at significant levels upon cycling bare electrodes in LiPF 6 and LiClO 4 .A weak contribution (≈1 atomic %) with a slightly lower binding energy (535.1 eV) was observed for the coated electrode cycled in LiPF 6 .Also, a sub-peak is observed at ≈536.4 eV for the bare electrode cycled in LiClO 4 and a sub-peak is observed at ≈535.8 eV for the coated electrode cycled in LiClO 4 , which were not present in any of the reference materials we used for calibration purposes.For the coated electrode cycled in LiClO 4 , a unique sub-peak at ≈532.9 eV emerged, which can be related to oxygen doubly bound to carbon.Another unique sub-peak at 534.5 eV also emerged, which can be related to oxygen singly bound to carbon.Due to the observed high binding energies, these components are thought to be the result of differential charging as a result of low electrical conductivity of the electrode sample after cycling.This further complicates not only the O 1s assignments and differentiation of the peaks associated with Li x PF y O z and alkyl carbonates, but the C 1s assignments as well and can only be speculated currently.
The lack of a contribution from the LiCoO 2 lattice oxygen subpeak for the bare electrode cycled in LiPF 6 in this study, which was not the case in our previous study, 27 indicates that the surface film being formed is thicker in the current investigation.The formation of a thicker film could be the result of cycling between 2.0-4.6 V in the current investigation versus 2.5-4.7 V in the previous investigation, as the electrolyte is more likely to break down while discharging to 2.0 V.As previously reported for the coated electrode cycled in LiPF 6 , the oxygen species were considered to be associated with the formation of Co-Al-O-F and PF x (OH) y species.On the other hand, the oxygen species for the bare electrode cycled in LiPF 6 included significant contributions from lattice oxygen, Li 2 CO 3 , O-C=O, and Li x PF y O z .As will be discussed later, based on the binding energy of the Cl 2p line and its concentration, none of the oxygen signal for the electrodes cycled in LiClO 4 can be attributed to the salt.The electrodes cycled in LiClO 4 have contributions from higher binding energy sub-peaks.This would suggest that the salt is decomposing and reacting with the solvent material, forming insoluble reaction products that are on the surface of the electrode or become part of the surface film.The bare electrode cycled in LiPF 6 has a larger contribution from the sub-peak at ≈535.4 eV, which is considered to play a role in the performance profile.The coated electrode cycled in LiClO 4 has the highest concentration of high binding energy oxygen chemistries, indicating a thicker resistive surface film that is forming on the surface and accounting for the poor electrochemical performance.
8][49][50] These species are resistive as illustrated by the differential charging observed in the hard X-ray data, which will be discussed later in the text.[50] Co 2p region.-Cobaltfluoride was found to dominate the cobalt chemistry in coated electrodes cycled in both electrolytes but not in bare electrodes.The Co 2p spectra for the pristine bare (Figure 4a) and coated (Figure 4d) electrodes, with the Co 2p 3/2 peak at ≈780 eV and the satellite at approximately +10 eV, are characteristic of LiCoO 2 and LiAl y Co (1-y) O 2 , respectively.The cobalt signal for the bare electrode cycled in LiPF 6 at 779.8 eV is very weak and, therefore, resulting in a poorly developed satellite structure at ≈+6.1 eV.
These values compare well with our values for Co(OH) 2 (Co 2p 3/2 at 779.8 eV) with a satellite structure at 6.1 eV higher indicating that the majority of surface Co is present in the form of Co(OH) 2 .The Co(OH) 2 chemistry is also supported by results based on the O 1s region.For the bare electrode cycled in LiClO 4 , the Co 2p 3/2 line has a binding energy of 780.3 eV with a satellite structure at approximately +7 eV.These values compare well to our values for the CoO reference (Co 2p 3/2 at 780.0 eV and satellite contribution at +7 eV) indicating that a large portion of surface Co is present in the form of CoO.This assignment is also supported by results based on the O 1s region.However, the presence of a modest amount of CoF 2 and/or oxyfluoride cannot be ruled out as the peak is slightly shifted to higher binding energy relative to that for CoO.The presence of the CoF 2 and/or oxyfluoride is confirmed based on synchrotron XPS data as will be demonstrated later in the text.The spectrum for the coated electrode cycled in LiPF 6 with the 2p 3/2 peak at ≈783.7 eV and a satellite separation of ≈+5 eV is consistent with the reference spectrum of CoF 2 indicating the formation of cobalt fluoride in agreement with previous results. 27The spectrum for the coated electrode cycled in LiClO 4 with a Co 2p 3/2 sub-peak at ≈783.9 eV also indicates the formation of CoF 2 .
The Co concentrations for the electrodes cycled in LiClO 4 are higher than those for the electrodes cycled in LiPF 6 (Table S1).It is noted that the sampling depth for the Co 2p electrons is 4.1 nm, which is smaller than or comparable to the overlayer film thickness as determined based on the C 1s region except in the case of the bare electrode cycled in LiClO 4 where it is greater than the overlayer thickness.Hence, the Co signal is from Co residing in the overlayer film except in the case of the bare electrode cycled in LiClO 4 where the Co signal can have an additional contribution from Co residing below the overlayer film.Therefore, Co dissolution may occur more frequently in LiClO 4 cycled electrodes, which then results in Co products being deposited in the overlayer film. 3s reported previously, 27 the formation of CoF 2 in the case of the LiPF 6 electrolyte can be due to a reaction between LiCo (1-y) Al y O 2 and/or LiCoO 2 with HF present as an impurity in the electrolyte.However, in the case of the LiClO 4 electrolyte, the only source of F is the PVDF binder, which can produce HF as a result of a dehydrofluorination reaction as reported previously. 25,26The reaction mechanism, therefore, could be similar for both the LiPF 6 and LiClO 4 electrolytes but the source of HF is different.In one case, it is the HF impurity in LiPF 6 and, in the other case, dehydrofluorination of the PVDF binder.F 1s region.-TheF 1s spectra from the pristine electrodes (Figures 5a and 5d) display mainly a single peak at ≈687.8 eV, which corresponds to fluorine chemistry from PVDF.A weak component with binding energy of ≈685 eV is also observed and can be related to the presence of LiF.The origin of LiF was attributed previously to a dehydrofluorination reaction of PVDF generating HF, which then reacts with LiCoO 2 or Li 2 CO 3 present on the surfaces of the LiCoO 2 particles to form LiF. 25 Upon cycling the bare electrode in LiPF 6 , the majority of the F signal arises from species at 687.9 eV, associated with PVDF, LiPF 6 and its degradation products.However, a high binding energy peak (≈690.9eV) is also present.This high binding energy was reported in recent Li-air work and was assigned to highly ionic species formed with the addition of organic groups promoted by organo-lithium. 47The high binding energy species, however, could also be the result of highly resistive species that is experiencing differential charging.For the coated electrode cycled in LiPF 6 , a single broad peak is observed at ≈688.0 eV, which is assigned to PVDF, LiPF 6 and its degradation products.
For the LiClO 4 -cycled electrodes, the fluorine contribution was limited to PVDF and the relative intensity of the F 1s signal reflects this (the signal in Figure 5 is multiplied by 5 for clarity).For both the bare and coated electrodes cycled in LiClO 4 , the F 1s region consists of a major contribution with binding energy of ≈688.0 eV, which can be associated with PVDF and AlF 3 (with F 1s binding energy of 687.4 eV from our AlF 3 reference).Al comes from the current collector in the case of the bare electrode, but also from the coating in the case of the coated electrode.For the bare electrode cycled in LiClO 4 , a component with binding energy at 685.3 eV, which can be related to LiF, and a high binding energy component at 689.6 eV, which could be associated with a highly ionic species within organic material as mentioned previously or differential charging, were also present.
Based on our CoF 2 reference, a subpeak which corresponds to the CoF 2 chemistry is expected at ≈686.0 eV in the F 1s region.However, due to the relatively large concentration of F from PDVF as well as LiPF 6 and its degradation products in the case of coated electrode cycled in LiPF 6 , the subpeak associated with CoF 2 chemistry is not clearly resolvable.However, the assignment for the CoF 2 chemistry was clearly confirmed based on the Co 2p photoemission region as discussed earlier.
Al 2p region.-TheAl 2p spectrum for the pristine coated electrode (Figure S6) displays a peak at ≈73.7 eV that is associated with LiAl y Co (1-y) O 2 , while that for the coated electrode cycled in LiPF 6 displays a peak at ≈76.4 eV that is associated with Al-F-O and/or AlF 3 species.As expected, no Al contribution is observed for the pristine bare electrode.The Al 2p spectrum for the bare electrode cycled in LiPF 6 consists of a single peak at 76.9 eV, which is associated with Al-F-O and/or AlF 3 from the interaction of the current collector with the LiPF 6 electrolyte and/or PVDF.The Al spectrum for the bare electrode cycled in LiClO 4 consists of two sub-peaks at 76.1 eV and 78.2 eV, which can be associated with Al-F-O and/or AlF 3 .The Al 2p spectrum for the coated electrode cycled in LiClO 4 , however, consists of a relatively large intensity sub-peak at ≈77.4 eV that is considered to be associated with AlF 3 and/or an Al-F-O oxyfluoride chemistries.The higher binding energies observed for the bare electrode cycled in LiPF 6 and LiClO 4 relative to that observed for our AlF 3 reference (76.7 eV) suggest a resistive film that is differentially charging.It should be noted that the atomic concentrations for aluminum are significantly higher for the LiClO 4 cycled electrodes than for the LiPF 6 cycled electrodes and the pristine electrodes.2][53] This could be due to the detachment of the active material at the current collector interface and the subsequent Al dissolution into the electrolyte and its re-deposition on the electrode surface.Similar corrosion products of the Al current collector were reported to occur after cycling in LiPF 6 in PC:DEC and EC:DMC electrolytes. 51nchrotron based XPS.-In an effort to examine the depth distribution of the respective surface films and the thickness of the metal fluoride phases, the electrodes were also inspected using monochromatic synchrotron X-rays with energy of 2555 eV.The O 1s photoemission spectra (Figure S12) are similar to spectra obtained using our conventional in-house spectrometer (Figure 3) with one very noticeable exception.Only electrodes cycled in LiClO 4 exhibit intense high binding energy peaks beyond 536 eV.The F 1s spectra (Figure S13) are also similar to the in-house spectra (Figure 5), except for the high binding energy peaks (∼693 eV), which appeared in electrodes cycled in LiClO 4 .High binding energy peaks were also observed in the Al 2p and Al 2s spectra (not shown) and were all shifted by similar amount as those of the F and O high binding energy peaks.We believe that these high binding energy peaks associated with O, F and Al are the result of differential charging.Thus, the highly resistive species appear to contain Al, O, and F. Clearly, the films formed on the surfaces of the electrodes cycled in LiClO 4 are more resistive in nature than the films formed on the surfaces of the electrodes cycled in LiPF 6 , regardless of the coating.Therefore, we speculate that these species could be due to dissolution of the Al current collector and its redeposition onto/into the electrode surface.Differential charging occurs less frequently with spectrometers employing conventional nonmonochromatic Al or Mg X-rays relative to spectrometers employing monochromatic synchrotron radiation.In the case of conventional spectrometers, the secondary electrons, generated by the X-rays passing through the window of the source, tend to neutralize the sample surface.
Synchrotron based XPS helps to reveal the thickness of the CoF 2 formed in the coated electrode cycled in LiPF 6 .The Co 2p spectrum (Figure S14) for the bare LiCoO 2 electrode cycled in LiPF 6 is consistent with conventional data showing no contribution from CoF 2 .The Co 2p spectrum for the bare electrode cycled in LiClO 4, displays contributions from lattice LiCoO 2 as well as CoF 2 .However, the Co 2p spectrum for the coated LiCoO 2 electrode cycled in LiPF 6 displays a contribution from LiCoO 2 in addition to the CoF 2 contribution.Since CoF 2 is the only chemistry observed in the case of conventional data, it suggests the depth of the region that contains CoF 2 for the coated LiCoO 2 electrode cycled in LiPF 6 to be greater than 4 but less than 11 nm.The formation of CoF 2 also occurs in the case of the electrodes cycled in LiClO 4 as the surface doesn't contain enough chlorine to account for all of the highly oxidized Co.The extent of the formation of CoF 2 is greater in the case of the coated LiCoO 2 electrode cycled in LiClO 4 relative to the bare electrode cycled in LiClO 4 .The CoF 2 peak was also observed in conventional data for both the bare and coated LiCoO 2 electrodes cycled in LiClO 4 and cannot be attributed to differential charging.In this case, the formation of CoF 2 suggests an interaction with the PVDF binder since it is the only source of F. Such an interaction could lead to degradation of PVDF and the loss of electrical connectivity between particles and be responsible for the poor performance.

Conclusions
Our previous work based on high resolution synchrotron X-ray diffraction data, 27 showed that the addition of the "AlPO 4 "-coating prevented lattice instability when cycling in LiPF 6 electrolyte even at higher voltages (4.7 V).Based on surface analysis by XPS and the XRD results, we proposed that the coating offers protection against corrosion by HF, in addition to lattice stability of the coated LiCoO 2 in the cathode.Based on voltage-capacity profiles in Figure 1, the observed trends in discharge energy and energy efficiency after 20 cycles are as follows: "AlPO 4 "-coated electrode in LiPF 6 (≈770 Wh/kg, ≈90%), bare electrode cycled in LiClO 4 (≈500 Wh/kg, ≈63%), "AlPO 4 "-coated electrode cycled in LiClO 4 (≈300 Wh/kg, 56%), bare electrode cycled in LiPF 6 (≈300 Wh/kg, ≈54%).Chemical analysis via XPS in the current investigation revealed several interesting results.The electrodes cycled in LiClO 4 showed CoF 2 formation in the Co 2p photoemission region (Figure 4).The only source of fluorine for these electrodes is the PVDF binder, suggesting dehydrofluorination or leaching of the F from the material.Also, Al was detected in the surface films for the bare electrodes.The Al concentrations are significantly higher for the electrodes cycled in the LiClO 4 electrolyte.This suggests that there is corrosion of the current collector and then redeposition from the electrolyte onto/into the surface film.The corrosion of the current collector has been shown previously, [51][52][53] but the degree of corrosion and exactly what variable has caused the corrosion is currently unclear.An estimation of the film thickness was made based on attenuation of the carbon black intensity for the cycled electrodes relative to that of the corresponding pristine electrodes.Sputter depth profiling with energetic Ar-ions (4 keV) provided some information (Figures S7-S10), but information from the C 1s region is of limited use as the bombardment of carbon films is known to cause some chemical changes to the carbon morphologies. 54,55Additionally, differential charging observed in the synchrotron-based XPS studies support the electrically resistive nature of the Al oxyfluoride (Al-O-F) formed during cycling in the LiClO 4 electrolyte.
The "AlPO 4 " coating was found to improve the voltage efficiency and capacity retention when using the LiPF 6 electrolyte, but was detrimental when using the LiClO 4 electrolyte.This indicates that the enhanced cycling performance with surface coating is only realized when cycling in LiPF 6 , suggesting that while the coating material may prevent HF corrosion of the active material, the electrolyte composition also plays a key role in cell performance.XPS reveals that the "AlPO 4 " coating promotes the formation of metal fluoride (e.g.CoF 2 and Al-O-F) in both electrolytes.The source of fluorine in the coated LiCoO 2 electrode cycled in LiPF 6 is largely attributed to the LiPF 6 salt whereas the source of fluorine in the coated LiCoO 2 electrode cycled in LiClO 4 is the binder PVDF.Furthermore, the data appear to indicate that the surface film formed on the bare electrodes is thinner than the surface film formed on the coated LiCoO 2 electrodes, with the film formed on the bare electrode cycled in LiClO 4 being the thinnest (≈ 1.6 nm) and the film formed on the coated LiCoO 2 electrode cycled in LiPF 6 being the thickest (≈ 5.3 nm).We stress that these are crude estimates based solely on the assumptions of a consistent photoelectron mean free path across chemistries and that these estimates are to aid the reader in addressing the questions of film thickness on a relative basis.Our study suggests that the working mechanism of "AlPO 4 " coating is the formation of stable metal fluoride and/or Co/Al oxyfluoride surface film via reactions between the surface coating and the F source from the electrolyte (e.g., HF impurity in the LiPF 6 electrolyte).When the F source from the electrolyte is removed such as in the case of LiClO 4 , the coating no longer works positively, but appears to mainly extract the fluorine from the PVDF binder causing degradation of the electrode.The fluorine source from PVDF is likely to be also HF, as a result of dehydrofluorination reaction as discussed earlier in the text. 25,26

FigureFigure 4 .Figure 5 .
Figure XPS spectra of the O 1s photoemission line for (a) pristine, bare LiCoO 2 , (b) bare LiCoO 2 cycled in LiClO 4 , (c) bare LiCoO 2 cycled in LiPF 6 , (d) pristine, "AlPO 4 "-coated LiCoO 2 , (e) coated LiCoO 2 cycled in LiClO 4 , and (f) coated LiCoO 2 cycled in LiPF 6 .The dashed vertical lines represent binding energy values for lattice oxygen (LiCoO 2 lattice oxygen), oxygen present in lithium carbonate (Li 2 CO 3 ), oxygen doubly bound to carbon in a carboxylic acid arrangement (O-C=O * ), oxygen bound to phosphorus (O * P(OR) 3 ), oxygen singly bound to carbon in a carboxylic acid arrangement (O * -C=O), and oxygen in a fluorine based compound from LiPF 6 degradation (Li x PF y O z ).It should be noted that the y-axis is the same for all spectra but were shifted vertically for clarity.
assigned peaks at 286.4 and 286.7 eV to the carbon singly bound to oxygen in lithium ethyl carbonate (CH 3 C * H 2 -O-CO 2 Li) and lithium methyl carbonate (C * H 3 -O-CO 2 Li), respectively.The third sub-peak at ≈288.3 eV can be assigned to a carbon doubly bound to oxygen or singly bound to two oxygen atoms, or to fluorinated alkyl carbonate (R-CF 2 -C * H 2 -O-CO 2 -R).

Table I . Summary of elemental concentrations (in atomic percent) as determined from XPS multiplex spectra.
Downloaded on 2017-03-27 to IP LiClO 4 have larger contributions, suggesting a reaction involving the PVDF binder.The above assignments are supported with assignments from the other XPS regions in the forthcoming discussions.
) unless CC License in place (see abstract).ecsdl.org/site/terms_useaddress.Redistribution subject to ECS terms of use (see 18.51.1.63 has indicated a binding energy of 530.3 eV is indicative of a Li 4 SiO 4 and work by Ramana et al. assigns a binding energy of 530.1 eV to LiMn 2 O 4 43 and 530.9 eV to LiFe 2 P 3 O 10 . 44Early work by Morgan et al. suggests a variety of pyrophosphates result in a binding energy of approximately 530.2 eV.