Spectroscopic Characterization of Solid Discharge Products in Li–Air

Jun 30, 2011 - Lucas D. Griffith , Alice E.S. Sleightholme , John F. Mansfield , Donald J. Siegel , and Charles W. Monroe. ACS Applied Materials & Int...
0 downloads 0 Views 2MB Size
ARTICLE pubs.acs.org/JPCC

Spectroscopic Characterization of Solid Discharge Products in Li Air Cells with Aprotic Carbonate Electrolytes Gabriel M. Veith,* Nancy J. Dudney, Jane Howe, and Jagjit Nanda Oak Ridge National Laboratory, Materials Science and Technology Division, Oak Ridge, Tennessee 37831, United States

bS Supporting Information ABSTRACT: Raman, infrared, and X-ray photoelectron spectroscopies were used to characterize the thick coating of reaction products on carbon and MnO2-coated carbon cathodes produced during discharge of Li air cells. The results show that neither Li2O2 nor Li2O are major components of the insoluble discharge products; instead, the products are composed primarily of fluorine, lithium, and carbon, with surprisingly little oxygen. The complex reaction chemistry also appears to involve the formation of ethers or alkoxide products at the expense of the carbonate solvent molecules (ethylene carbonate and dimethyl carbonate). The irreversible discharge reaction is likely electrochemically promoted with Li+, anions, and dissolved oxygen. Exactly how the molecular O2 participates in the reaction is unclear and requires further study. The addition of a conformal coating of MnO2 on the carbon lowers the cell’s operating voltage but does not alter the overall discharge chemistry.

’ INTRODUCTION Rechargeable lithium air batteries are a potential replacement for conventional Li-ion intercalation batteries due to their high theoretical specific energy (11 970 Wh/kg).1,2 Li air cells typically consist of an electrically conductive porous air electrode; a nonaqueous electrolyte, such as ethylene carbonate/ dimethyl carbonate; a lithium salt, for example, LiPF6; and a Li foil anode.2 11 The air electrode is porous to store the products that are generated from the reaction of Li+ and O2 during the discharge reaction. The design of the cell is critical to optimize performance. For example, the pore volume of the cathode will affect the extent of discharge6,7 as will the volume12 and the type of liquid electrolyte.8 By applying a large enough potential, or introducing a catalyst, the discharge reaction can be reversed, making the Li air battery rechargeable.13 Bruce and others have demonstrated that the addition of a metal oxide catalyst, for example, MnO2, can lower the charge overpotential and increase the cycleability of the cathode significantly.2,3,14 16 However, at this point, with a few exceptions,17,18 there is no fundamental understanding of the role of the catalyst, the structure of the catalyst (e.g., MnO2 phase), the mechanism of the charging reaction, or the products formed during the discharge reaction. Raman spectroscopy and X-ray diffraction results have been reported indicating the formation of Li2O2 in both polymer electrolyte13 and Pd/ MnO2 electrocatalyst based Li air cells.19 MnO2 electrocatalysts have also been shown to decompose Li2O2 that was purposefully added to a cathode.4 However, results by Mizuno et al.20 and Xu et al.21 report that the discharge products were not Li2O2 but were Li carbonates from the decomposition of the liquid electrolyte. r 2011 American Chemical Society

There are many different electrolyte and lithium-salt combinations being investigated for Li air cells.22 The goal of this work was to experimentally determine the chemistry of the insoluble discharge products from a relatively widely used liquid electrolyte—ethylene carbonate/dimethylcarbonate (EC/ DMC)—8,9,23 26 with LiPF6 as the salt. Understanding the mechanism of the discharge chemistry and identification of the correct discharged products would help predict ways to develop new catalysts and electrolytes to achieve good rechargeability and round trip efficiencies in Li air cells.

’ EXPERIMENTAL SECTION Swagelok cells were constructed in an argon-filled glovebox using 0.75 mm thick Li foil (99.99%, Alfa), a Celgard 2500 separator, and 1.5 mL of 1.2 M LiPF6 in 1:1 wt % ethylene carbonate/dimethyl carbonate (EC/DMC) (Ferro), which is a widely used electrolyte for Li-ion batteries and Li air cells.4,9,14,23 25,27,28 Two carbon cathodes were used in this work. The first is a graphitized carbon foam (0.035 g, 1 mm thick, surface area = 2.2 m2/g as measured by N2 physisorption on a Quantachrome Autosorb 1C) made at ORNL.29,30 The graphite foam was formed from Mitsubishi ARA24 pitch at 1000 °C, followed by graphitization in Ar at 2800 °C to give a highly conductive foam with open cells (0.4 mm diameter). The second cathode is a carbon veil from Hollingsworth and Vose (grade 08000020, 0.5 m2/g, four layers assembled in cell, 1.2 mg/layer). Manganese oxide-coated electrodes were prepared by Received: May 9, 2011 Revised: June 2, 2011 Published: June 30, 2011 14325

dx.doi.org/10.1021/jp2043015 | J. Phys. Chem. C 2011, 115, 14325–14333

The Journal of Physical Chemistry C soaking the carbon cathodes in an excess of a 0.1 M NaMnO4 (Alfa Aesar) + 0.1 M Na2SO4 (Aldrich) solution for 5.5 h.31 The cathodes were washed with 18 MΩ water until the wash solution was clear and then soaked in 100 mL of water for 10 min, followed by vacuum drying for 18 h. The cells were assembled in a vertical geometry with the cathode located at the top of the cell closest to the oxygen/argon supply. Research grade argon or oxygen (Air Liquide) was used. To exchange argon from the prepared cells, they were attached to a manifold where the argon (10 mL initial) in the cell was serially diluted with O2 (100 mL, 20 psi) 15 times. The cells were operated with 20 psi O2 or atmospheric pressure argon with enough electrolyte to saturate the cathode. Cells were discharged to 2.2 V at 5 μA (graphite foam) or 1 μA of current on a Maccor battery tester. Cells were disassembled in an argon-filled glovebox, washed with DMC for 2 min, and dried for 24 h on filter paper under vacuum. A JEOL-840 scanning electron microscope (SEM) or a Hitachi S-3400 VP SEM operated at 15 kV of accelerating voltage was used to image the cathodes. Energy-dispersive X-ray spectra (EDS) were collected using an EDAX Si(Li) detector. The specimen was placed on carbon tape for imaging without any coating. Solid products formed during the discharge reaction were characterized using a PHI 3056 X-ray photoelectron spectrometer (XPS) with an Al anode source operated at 15 kV and an applied power of 350 W. XPS samples were load locked into the XPS prior to analysis without exposing the sample to ambient air. Some XPS samples were sputtered using 3 kV argon ions at a pressure of 1  10 5 Torr for 30 min. Lithium XPS standards were obtained from Sigma Aldrich (LiPF6) and Alfa Aesar (LiF, Li2O2, Li2CO3) or made by physical vapor deposition in a N2 plasma (lithium phosphate oxynitride, Lipon).32 Fourier transform infrared (FTIR) spectroscopy data were collected on the carbon veil sample with a N2-purged FTIR spectrometer (BioRad 575C), equipped with a room-temperature deuterated triglycine sulfate (DTGS) detector and KBr beamsplitter, controlled with WinIR-Pro software. A washed carbon veil (not discharged) was used as the background. Raman spectroscopy of the discharged air cathodes was performed using a WITec alpha-300 confocal microscope using a 532 nm excitation wavelength using a 50 microscope objective. The Raman signal was collected in the confocal geometry and dispersed through a spectrometer (600 lines/mm grating) and detected using a thermoelectric cooled CCD detector. Raman shifts were collected in the range of 100 3000 cm 1.

’ RESULTS The graphite foam (Figure 1, top left) and carbon veil (Figure 1, top right) used in this study were selected due to either the chemical simplicity (graphite) or the open structures (veil), which reduces the system complexity and enabled the application of a wide variety of analytical techniques. This differs from most other cells reported in the literature that use polyvinylidene fluoride (PVDF) as a binder, allowing us to simplify spectroscopic analysis due to overlapping peaks. XPS data collected on unused carbon (raw) current collectors revealed that the graphite foam surface contained primarily C C (284.8 eV) and the C O functionality (532 eV) with less than 2 at. % oxygen on the graphite surface (Figure 2 and Table 1). The XPS surface chemistry of the carbon veil was more complex due to an organic binder (Figure 3). The veil surface contained a C N functionality (9.4% N, BE = 400 eV, not shown) and C O

ARTICLE

(286.4 and 533.5 eV) and O CdO functionalities (289 eV, 532 eV), originating from 20.7% surface O species. Because of the low F concentration in the binder (2.2%, Table 1), the C 1s C F functionality, which would be observed around 292 eV (Figure 3), was not detected; however, the F 1s binding energy was 688.0 eV, consistent with C F bonds. MnO2 was autocatalytically reduced on the carbon surface at neutral pHs, resulting in a homogeneous coating of MnO2 on the carbon surface.31 This approach varies significantly from the typical methods of adding MnO2 to carbon-based Li air cells, which generally rely on the mixing of powdered MnO2 with carbon black and PVDF binder.15 As shown by the XPS results in Figures 2 and 3, the addition of MnO2 to the carbon surface did not change the C 1s spectra significantly for either the graphite foam or the carbon veil. The O 1s spectra all showed the formation of an O peak around 529.5 eV due to oxygen from MnO2. XPS characterization still shows significant concentrations of fluorine and nitrogen (1.4 and 6%, respectively) from the binder used on the veil, indicating that the MnO2 coating is very thin, less than 2 nm thick. Given that the deposition of the MnO2 occurs at neutral pHs, it is unlikely that the coating is inhomogeneous, resulting in the observed signals from the carbon veil.31 XPS analysis of the MnO2 coating show a single Mn 2p3/2 peak at 642.5 eV (Figure S1, Supporting Information), which is consistent with the formation of MnO2.33 Because of the low surface areas of the cathodes, the concentration of MnO2 is low and there is no X-ray diffraction from the MnO2 phase (cf. Figure S2, Supporting Information). However, previous reports show the formation of a birnesitte MnO2 by this methodology.31 There was no change in cathode morphology evidenced in the SEM by the addition of the MnO2. Discharge reactions were performed with bare and coated carbon veil cathodes in both air and O2 atmospheres, as shown with selected curves in Figure 4. The initial open-circuit voltages (OCVs) of cells were about 3.4 V, which is the expected potential for carbon versus lithium. Cells were discharged at relatively low fixed currents because of the small surface areas provided by the veils and foam cathodes. Figure 4 clearly demonstrates that oxygen is required for extended discharge to proceed. With the argon cover gas, the cell potential drops to 2.2 V within 2 7 h, which is consistent with studies of the Li intercalation reaction for these carbon fibers.34 Only with the availability of O2 is the discharge extended at ∼2.7 V for 1200 h. A similar, although less dramatic, reduction in discharge capacity was observed for the MnO2-coated veils discharged in argon versus oxygen (Figure 4, inset). Each of the curves includes the relaxation at open circuit following the 2.2 V cutoff. The potentials approach or exceed 3 V, indicating that only a small amount of the Li has inserted into the carbon during discharge. It should be noted that odd features, such as the increase in voltage at ∼800 h, are reproduced with other cells. Figure 5 shows the discharge plots for the graphite foam-based cathodes operated at 5 μA of current in O2. Again, the initial OCV (∼3.4 V) and the rapid voltage drop upon the initial discharge are consistent with the lithium intercalation reaction for the graphite foam.35 Relaxations at 450 h and after 2.2 V termination are indicative of high potentials associated with the Li carbon reaction, rather than lower voltages expected for Li + O2 reactions. The slopping discharge curves initiating at 2.9 and 2.55 V are due to reactions requiring oxygen. The addition of the MnO2 coating adds to the cell polarization, resulting in a lower cell voltage and increasing the cell resistance due to a barrier layer 14326

dx.doi.org/10.1021/jp2043015 |J. Phys. Chem. C 2011, 115, 14325–14333

The Journal of Physical Chemistry C

ARTICLE

Figure 1. SEM images of graphite foam (top left), the carbon veil (top right), discharge products on graphite foam (lower left), and discharge products on the carbon veil (lower right).

Table 1. XPS Composition of the Raw Cathodes and Discharge Products Insoluble in DMC sample

at. % Li

at. % P or N

graphite

30.3

(P) 3.0

11.5

9.4

graphite, sputtered

32.4

(P) 3.3

16.6

10.8

36.8

MnO2 on graphite

45.8

(P) 0.7

11.8

6.0

35.7

(N) 9.4

67.7

20.7

2.2

(P) 1.3

20.7

6.7

25.0

raw graphite

raw carbon veil carbon veil

Figure 2. C 1s and O 1s XPS data for graphite foam and carbon veil cathodes.

of MnO2. There are large fluctuations in cell voltages, the origins of which are unknown; however, the variations in voltages are somewhat reproducible over multiple samples. These fluctuations could be due to changes in the Li anode, aspects of the cathode reaction, or the formation of an extended solid-electrolyte-interphase (SEI), but these reasons do not seem important as the reaction products measured for short times are the same as those that are measured for longer reaction times. The discharge products after 800 1200 h discharge are shown in the lower images of Figure 1. The coatings appear

46.2

at. % C

at. % O

98.0

2.0

at. % F

45.8

quite thick and dense. To characterize the discharge products, XPS studies were performed and compared to several lithiumcontaining reference compounds (Table 2). Figure 6 shows the C 1s and O 1s XPS data collected for the discharge products on graphite foam- and carbon veil-based cathodes, respectively, after long discharges (750 1200 h). Figure 7 shows the F 1s, P 2p, and Li 1s XPS data collected for the discharge products on these same cathodes. Largely, the reaction products are very similar. To investigate changes in the discharge products as a function of time, sputtering experiments were performed on the same graphite foam sample used in the XPS experiments described above. After sputtering the discharge products for 30 min, there was no change in the Li, O, or P XPS data (Figure 7). There were minor changes in the fractions of C 1s and F 1s species (Figure 6), but there were no changes in the binding energy or type of species present in the samples. XPS results on discharge products formed during shorter reaction times (48 h) were similar to results shown above. For the graphite foam-based cathodes, the C 1s data reveal a new C 1s peak at ∼288.5 eV that accounts for 80 90% of the insoluble carbon species. Analysis of the MnO2/graphite foam sample showed similar oxygen-containing species with nearly identical binding energies. The O 1s data for all the graphite foam-based cathodes show the formation of a single oxygen species centered at 535.9 eV. These oxygenated species cannot 14327

dx.doi.org/10.1021/jp2043015 |J. Phys. Chem. C 2011, 115, 14325–14333

The Journal of Physical Chemistry C

ARTICLE

Figure 5. Discharge data for graphite foam-based cathodes in oxygen. Curves include the relaxation under open-circuit condition at the end of discharge.

Table 2. XPS Binding Energies of Li, F, and P in Reference Materials and Discharge Products Figure 3. C 1s and O 1s XPS data of MnO2-coated graphite foam and carbon veil cathodes.

Figure 4. Discharge data of carbon veil-based cathodes. The inset shows data for the MnO2-coated carbon veil, including the subsequent relaxation at open circuit.

be attributed to typical C O or CdO functionalities observed on the native carbon surface.36 There was no evidence for CdO originating from the carbonate electrolyte in the C 1s XPS data for the graphite-based cathodes, which would have a binding energy of ∼292 eV, or the O 1s XPS data, which would show a binding energy of 534 eV.37 Instead, this functionality is due to the formation of C O C ether species, which have an O 1s binding energy around 536 eV,37 40 and C 1s binding energies centered at 288 eV.37 There was a slight improvement in the C 1s fits when a small concentration (∼15%) of secondary carbon species attributed to carbon bound to an ether (C* C O C, BE = 287.4 eV) and CO3 (290 eV) were added to the refinement (χ2 = 12 and 2, respectively).37,41 The low concentrations of CO3 and C* C O C versus C O C indicate that the majority of the solid discharge products is ether-like. The C 1s and O 1s data collected for the products grown on carbon veil cathodes showed the same C 1s species, at 288.2 eV,

material

Li binding energy

F binding energy

P binding energy

(eV)

(eV)

(eV)

Li2O2

54.58

Li2CO3 Li3 xPOyNz

55.15 55.50

LiF

56.01

684.95

LiPF6

57.33

688.5

137.60

Li air cell

59.7

689.5/691.0

138.1 138.3

134.02

and O 1s species, at 535.8 eV, that were recorded for the graphite foam-based cathodes due to ether compounds (Figure 6). However, there was a slight change in the C,O species in the discharge products, compared to the product on graphite foam-based cathodes. Specifically, there was an increase in C C and C H species in the C 1s spectra (284.8 eV) and CdO in the O 1s data (532.4 eV). These peaks originate from the carbon veil cathodes. SEM data collected for these samples reveal that the discharge products spall off the cathode, presumably during the disassembly of the cell and peeling of the veil layers (Figure 1) (in addition, cf. Figures S1 and S2, Supporting Information). This exposes the underlying CdO functionality, which is detected in the XPS, leading to the observed substrate peaks. Phosphorus and fluorine XPS data collected for the graphite foam and MnO2-coated graphite cathodes show the formation of P and F species with binding energies around 138 eV (P 2p) and 690 691 eV (F 1s) (Figure 7). The P 2p binding energies are similar to the P binding energies observed for pure LiPF6 (138.1 138.3 eV) (Table 2). However, the F 1s binding energies have shifted to slightly higher binding energies versus LiPF6 (F 1s = 688.5 eV) and show the formation of two F species: one with a binding energy of 691 eV and the second with a binding energy of 689.5 eV. This shift cannot be attributed to the C Fx functionality, which would exhibit a C 1s peak centered at 291 eV40 and a F 1s peak around 688 eV.42 On the carbon veil, there is a single F species detected with a binding energy of 689.0 eV, which is slightly lower energy than the F species detected on the graphite foam, but higher in energy than the F species detected for LiPF6. The binding energy of the P was consistent with what was detected on the graphite foam cathodes (P 1s BE = 138.0 eV). 14328

dx.doi.org/10.1021/jp2043015 |J. Phys. Chem. C 2011, 115, 14325–14333

The Journal of Physical Chemistry C

Figure 6. C 1s and O 1s XPS data for discharge products on graphite foams.

The Li 1s spectra recorded for the discharged products on the graphite foam- and carbon veil-based cathodes shows a single Li species in the discharge products with a binding energy of 59.7 eV. This binding energy is significantly higher than what was experimentally measured for Li in Li2O2, Li2CO3, Lipon, LiF, and LiPF6 (Table 2). In addition, the observed Li binding energy is significantly higher than what has been reported for Li alkylcarbonates (Li-OCO2), 55.6 eV.41,43 The Li in Li2O2 has a binding energy of 54.6 eV due to the covalent bonding of this solid. Conversely, the Li in LiPF6 has a binding energy of 57.3 eV due to the highly ionic nature of this material, which allows it to readily dissociate and facilitate Li+ transfer in liquid electrolytes. These results would strongly indicate that the Li is highly dissociated or bound to a strongly electrophilic site. Table 1 presents elemental analysis information obtained from the XPS data of the discharge products insoluble in the DMC wash or the liquid electrolyte during the reaction using standard sensitivity factors.44 In all samples, there is a significant concentration of Li in the discharge products, 30 46%. The relative

ARTICLE

ratio of F to P in the XPS data (Table 1) would strongly indicate that the LiPF6 is decomposing during the reaction, resulting in higher F concentrations and forming a P species that is soluble in the electrolyte or easily washed off during the cathode rinse. For comparison purposes, the fluorine-to-phosphorus ratio for the LiPF6 standard was 5.6:1, close to the theoretical 6:1 concentration predicted by stoichiometry. EDX data collected on the carbon veil discharge products showed similar enrichments in F compared to P (49.1:2.2 at. %), which were in excellent agreement with the F/P ratio determined by the XPS. The constant P concentration in the graphite and sputtered graphite samples would strongly indicate that the P is dissolved in the liquid electrolyte and not buried in the discharge products. The carbon veil discharge products have the highest carbon-to-oxygen ratio of the discharge products. This is consistent with the observed CdO functionality that may originate from the incomplete cleavage/oxidation of the CdO functionality during the reaction or exposed veil surface evidence in the Supporting Infomation, Figures S3 and S4. However, despite oxygen being required for the reaction to proceed, there is relatively little oxygen (2 11%) in the discharge products. This would strongly indicate that the oxygen plays a catalytic role in the reaction or is trapped in a product that is volatile (e.g., CO2) or soluble in the electrolyte, such as a lithium ethyl(methyl) carbonate.45 In conjunction with XPS, we also performed micro-Raman and FTIR studies on the discharged air cathodes. Figure 8 shows the Raman spectra at several locations (indicated in the optical micrograph) of a carbon foam air cathode without the MnO2 coating. Among the Raman signals, most prominent were the strong carbon graphitic G bands at 1587 cm 1 due to the sp2 bonding arrangement of carbon and the disorder D band at 1360 cm 1 mostly from sp3 carbon. Most notable was the absence of a Raman feature corresponding to the Li2O2, which should appear at 790 cm 1 corresponding to the O O stretch of the anhydrous peroxide structure.13 If we had formed hydrated peroxides (because of sample exposure or moisture contamination), the Raman band would further shift by more than 50 cm 1 to higher wavelengths.46 There is no evidence for CdO stretching in the Raman data (1794 cm 1), which is consistent with the cleavage of CdO and the formation of ether species. The observed band at 908 cm 1 is most likely attributed to the shift of the EC ring breathing bands (from 891 cm 1) due to the presence of the electropositive Li+.46,47 Another likely evidence of the reduction of CdO and subsequent interaction with salt (LiPF6), forming a Li O C-type complex, is the appearance of the 733 cm 1 peak, which, in a normal EC molecule, appears at 714 cm 1 due to the ring bending mode.46,47 The feature at 1204 cm 1 is attributed to C H bending due to species in the discharge products. The variations in signal intensity/presence would be consistent with local variations in discharge products. It is worth noting that peak positions reported here originate from the solid discharge/decomposition product and could be different from solution studies reported in the literature due to solvent effects. The Raman data collected for the discharged MnO2-coated air cathode shown in Figure 9 did not show any of the characteristic decomposition salt/solvent peaks. We only observed the prominent carbon Raman signal at 1356 and 1585 cm 1, respectively. The absence of the Raman signal corresponding to the MnO2 phase is most likely due to very low loading of the catalyst due to the low surface area of the graphite foam. Further, the Raman measurement was performed at a low laser incident power (less 14329

dx.doi.org/10.1021/jp2043015 |J. Phys. Chem. C 2011, 115, 14325–14333

The Journal of Physical Chemistry C

ARTICLE

Figure 7. Li 1s, P 1s, and F 1s XPS data of discharge products on graphite foam-based cathodes.

than 0.1 mW) to avoid sample heating with a 50 objective having a typical spot diameter of less than 400 μm, which may not be able to show weaker Raman bands if present. It is not clear at present if the absence of the Raman peaks for the discharge products is because of the role of MnO2 or a change in reaction product chemistry. The Raman spectra of a discharged carbon veil sample is shown in Figure 10 showing prominent D and G bands at 1360 and 1587 cm 1, respectively. Unlike the carbon foam air cathode, the carbon veil sample is only partially graphitic, as given by the broad D band extending from 1000 cm 1. There could be a possibility that the relatively weaker discharge peaks are spectrally buried under this huge carbon signal. Again, there is no evidence of Li2O2 formation. The origins of the peak at 952 cm 1 (for the carbon veil) and the

peak at 1204 cm 1 (for the graphite foam) are currently under investigation to determine weather they are intrinsic to any of the discharge products species or from the cathode. Figure 11 shows FTIR data of the discharge products obtained on the carbon veil. Strong vibrational bands are observed at 1755 cm 1 (CdO), 1075 cm 1 (C O), and 845 cm 1 (C O C)48 and starting at 564 cm 1 due to Li O or P F bands.46,49 Smaller vibration bands, labeled in Figure 11, are consistent with the formation of C O species.46,49 The FTIR data are consistent with lithium alkoxides, that is, C O Li, along with ether C O C, and F P solid products that are observed in the XPS data. Because of the difficulty of separating carbon veil layers after the discharge reaction, it was impossible to get the same thickness as the carbon veil used as an FTIR 14330

dx.doi.org/10.1021/jp2043015 |J. Phys. Chem. C 2011, 115, 14325–14333

The Journal of Physical Chemistry C

Figure 8. Raman spectra collected on four different spots on the graphite foam cathode after discharge.

ARTICLE

Figure 10. Raman spectra collected on two different spots on the carbon veil cathode after discharge.

Figure 11. FTIR data of discharge products on the carbon veil.

Figure 9. Raman spectra collected on two different spots on the MnO2coated graphite foam cathode after discharge.

background. Consequently, the CdO species may be due to CO from the cathode that was evidenced in the XPS. The reduced transmission of the sample and absence of the CdO peak at 292 eV in the XPS data, due to the carbonate electrolyte, would support this argument. Alternatively, discharge products may not form on carbon surfaces that are not electrically connected.

’ DISCUSSION The XPS, Raman, and FTIR results indicate that a complex discharge chemistry is occurring in the ethylene carbonate/ dimethyl carbonate electrolyte. It is important to note that the products we are analyzing are insoluble in DMC and the liquid electrolyte, so we obtain an incomplete picture of the total discharge chemistry. However, understanding the solid products is critical because they will block catalyst and surface sites and

need to be decomposed to obtain maximum efficiency of the Li air cell. At this point, the observed reaction appears to be different than reactions typically observed at low potentials (below 1.5 V) for anodes that typically produce lithium alkoxides and alkyl carbonate oligomers.50 53 Given the lack of Li2O2, a better description of the reaction occurring under our conditions may be the formation of an SEI layer at 2.5 2.9 V on a carbon surface in an oxygen atmosphere. Quantitative analysis of the XPS data shows that the solid discharge products contain high concentrations of lithium (30 46 at. %) and fluorine (25 46 at. %) and low concentrations of phosphorus (