Electrical storage technologies are of vital importance to enable effective utilization of intermittent renewable energy sources and the creation of sustainable electric transportation. Conventional Li-ion batteries cannot meet the long-term energy storage requirements for electric vehicles, owing to their inherent gravimetric energy limitation associated with Li intercalation1,2,3. To increase gravimetric energy, one promising approach is to employ four-electron redox of oxygen, where Li-O2 batteries have recently shown the potential to provide gravimetric energy ~4 times that of conventional Li-ion batteries4,5,6,7,8. Unfortunately, little is known about the oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) mechanisms in the presence of Li+ ions at room temperature, which has led to a number of challenges3,5 at the oxygen electrode, limiting the practical use of this technology, such as poor round-trip efficiency4,9, the reactivity of aprotic electrolytes with oxygen redox reaction intermediates10 and cycle life3. Improving the mechanistic understanding of oxygen reduction and evolution is critical to develop strategies to overcome these challenges.

Here we study the redox of oxygen on the surface of a mixed electronic and Li+ ionic conductor, LixV2O5, using a specially designed, all solid-state Li-ion battery11, which eliminates parasitic reactions between oxygen reduction/evolution reaction intermediates and aprotic electrolytes used in conventional Li-O2 batteries reported to date. Fourier transform infrared spectroscopy10,12,13,14 and Raman spectroscopy13,15 have shown that carbonate solvents commonly used in Li-ion batteries are not suitable for the oxygen electrode as they react with the ORR intermediate species such as the superoxide radical ion (O2)10,13 and yields parasitic reaction products such as Li2CO3 and lithium alkyl carbonate species. On the other hand, ether-based solvents such as dimethoxyethane (DME) are reasonably stable against O2 15, where oxygen reduction leads to the formation of Li2O2 in first few cycles7,16. Unfortunately, cycling in ether-based electrolytes gradually converts Li2O2 to lithium carbonate-based species16 and Li salts used in ether-based electrolytes can react with oxygen reduction products17. Moreover, researchers have very recently reported that carbon in the oxygen electrode can react with Li2O2 to form carbonates in DME18, which increases the complexity in unraveling the Li-O2 reaction mechanisms in aprotic electrolytes. Utilizing the special all solid-state cell design and ambient pressure X-ray photoelectron spectroscopy (APXPS)19,20,21, we directly visualize the formation and disappearance of Li-O2 reaction products (namely Li2O2) on an LixV2O5 surface in situ as a function of applied battery potential.


In situ electrochemical APXPS measurements were conducted using a solid-state Li-O2 battery11 to probe reaction products during discharge and charge to avoid the evaporation of liquid electrolytes with low vapor pressure in the high-vacuum XPS chamber (Fig. 1). This solid-state cell consisted of lithiated lithium titanate (LLTO) as the negative electrode (having a chemical formula of Li4+xTi5O12)22, lithium phosphorous oxynitride (LiPON) as the Li+-conducting solid electrolyte (~1,000 nm thick) and vanadium oxide (V2O5)23 as the positive electrode (~23 nm in thickness). The LLTO/LiPON/V2O5 cell was placed on a holder outfitted with electrical contacts to enable in situ electrochemical measurements in the XPS chamber (Supplementary Fig. S1)24. The Li 1s, O 1s, C 1s and V 2p spectra were collected from the top cell surface, which consisted of LixV2O5 and LiPON (Fig. 1), while the battery was discharged and charged potentiostatically under ultra high vacuum (UHV, pressure < 10−11 atm) and followed by similar testing conditions in an oxygen environment at a partial pressure of p(O2) = 5 × 10−4 atm (Methods and Supplementary Fig. S2). Substantial differences in the Li 1s, O 1s and V 2p spectra were observed between UHV and oxygen conditions upon charge and discharge, which will be discussed in detail below.

Figure 1
figure 1

Solid-state cell (LLTO/LiPON/LixV2O5) used for in situ APXPS measurements.

Lithiated Li4+xTi5O12 (LLTO) of 750 nm supported on a Pt coated alumina disk, LixV2O5 of 23 nm and Li-ion conducting LiPON electrolyte of 1,000 nm were used as the negative electrode, positive electrodes and electrolyte, respectively. LLTO was encapsulated by LiPON. XPS data were collected from the top surface of the cell, which consisted of both LixV2O5 and LiPON, as a function of voltage applied externally.

In situ XPS evidence of reversible intercalation/de-intercalation in LixV2O5 under UHV

Figure 2a shows the Li 1s, O 1s and V 2p spectra as a function of potential applied across the LLTO/LiPON/LixV2O5 cell (Vcell) during discharge in UHV, where detailed information of fitted components is shown in Supplementary Table S1 and Fig. S3. We note that all the cell potentials (Vcell) used in this study refer to the measured open-circuit-voltage (OCV) or the applied potentials across the solid-state cell (LLTO/LiPON/LixV2O5). Upon discharge, the Li 1s peak position remained nearly unchanged while the peak intensity was increased slightly (up to 20% in Supplementary Fig. S3), which is due to the increasing lithium content in LixV2O5 as expected from Li ion migration from the LLTO to LixV2O5 with decreasing applied voltage from 1.8 to 0 Vcell. The Li 1s peak centered at 55.5 eV could be assigned to surface carbonate species such as Li2CO3 (55.5 eV)25 that was formed upon air exposure on LixV2O5 and LiPON, LixV2O5 (55.7–55.9 eV)26 and partially to LiPON (56.0 eV)27. The increasing Li 1s intensity was accompanied with the broadening and gradual shift of the V 2p peak to lower binding energy, indicating lowered valence state of vanadium ions upon discharge. The V 2p peak could be fitted to three peaks at 514.5 eV, 516.0 eV and 517.4 eV, which are attributed to V3+, V4+ and V5+ in LixV2O5, respectively27,28. The relative fractions of V3+ and V4+ ions were found to increase upon discharge while that of V5+ decreased (Supplementary Fig. S4a), which was used to determine the vanadium valence state of LixV2O5 from 4.6+ (at 1.85 Vcell) to 3.6+ (at 0 Vcell), as shown in Fig. 3a. The changes noted in the O 1s region are in agreement with lithium intercalation into LixV2O5. The O 1s region includes contributions from LixV2O5, LiPON and surface lithium carbonate species (Li2CO3) formed upon air exposure of LiPON and V2O5 (Fig. 2a). Three components centered at 530.2, 532.0 and 534.0 eV were used to describe lattice oxygen (O2−) in LixV2O527,28, both oxygen doubly bound to phosphorus (P = O)27 and oxygen in Li2CO325 and oxygen singly bound to two phosphorus (P-O-P)27 in LiPON, respectively. Upon discharge, the lattice oxygen (O2−) in LixV2O527,28, became increasingly pronounced and led to the gradual growth of the O 1s peak shoulder at the low binding energy side. In addition, there was a systematic shift in the binding energy of the O2− lattice component in LixV2O5 to lower binding energy by 0.4 eV from 1.85 to 0 Vcell, which is in agreement with decreased covalency of V-O bonds with decreasing vanadium valence state28.

Figure 2
figure 2

In situ XPS data of Li 1s, O 1s and V 2p collected under UHV.

(a) Discharge from OCV (1.85 Vcell) to 0 Vcell. (b) Charge from OCV (0.1 Vcell) to 3.0 Vcell.

Figure 3
figure 3

LixV2O5 surface vanadium valence state as a function of applied voltage during cycling in UHV and p(O2) = 5 × 10−4 atm.

(a) The vanadium valence state of the surface of the LixV2O5 electrode as a function of applied voltage during discharge (left) and charge (right) under UHV. (b) The vanadium valence state of the surface of the LixV2O5 electrode as a function of applied voltage during discharge (left) and charge (right) under p(O2) = 5 × 10−4 atm. The changes of the vanadium valence state as a function of applied potentials observed in this study (open circle) is compared with that estimated from previously reported lithium content of LixV2O5 from Li/LixV2O5 thin film batteries (solid circle)29.

Upon charging, the changes in the Li 1s, V 2p and O 1s upon discharge were reversed, which is indicative of reversible lithium de-intercalation in LixV2O5, as shown in Fig. 2b. The intensity of the Li 1s region was decreased by 40% at voltages equal to and greater than 2.0 Vcell. (Supplementary Fig. S3). This is accompanied with increased vanadium valence state upon charging from 3.6+ (at 0.1 Vcell) to 4.6+ (at 3.0 Vcell) (Fig. 3a). It is interesting to note that the relative fractions of V3+, V4+ and V5+ ions and the covalency in the V-O bonds (having the binding energy of the O2− component shifted to higher binding energy) at the end of charge (3.0 Vcell) were found comparable to those at the onset of discharge (1.85 Vcell, Supplementary Fig. S4a–b). It should be mentioned that the relative intensity of the O 1s component at 532.0 eV (assigned to Li2CO3 and P = O in LiPON, Fig. 2b) was considerably reduced upon charge to 3.0 Vcell, which suggests the decomposition of Li2CO3 upon charging. This hypothesis is further supported by the decreases in the C 1s component at 290.0 eV (assigned to Li2CO325 in Supplementary Fig. S5).

The vanadium valence state estimated from the XPS V 2p data as a function of applied voltage was further compared with that obtained from Li/LixV2O5 thin film batteries reported previously29, as shown in Fig. 3a. Upon discharge, the vanadium valence state was changed from V4.6+ to V3.6+ while the valence state of vanadium was changed reversibly from V3.6+ to V4.6+ upon recharge. As the charge capacity (0.21 mC) was found to be larger than the discharge capacity (0.16 mC), the additional charge observed during charging can be attributed to the electrochemical oxidation of Li2CO3 as mentioned above. Comparable voltage profiles were noted in Fig. 3a, indicating that these in situ electrochemical XPS measurements in UHV showed reversible lithium intercalation in LixV2O5 comparable to that reported previously from Li/LixV2O5 thin film batteries29.

In situ APXPS evidence of reversible formation and removal of Li2O2 on LixV2O5 under p(O2) = 5 × 10−4 atm

During discharge, the intensity of the Li 1s region was found to increase while that of V 2p region decreased significantly (Fig. 4a and Supplementary Fig. S6). Although the LLTO/LiPON/LixV2O5 cell in the presence of oxygen had a comparable discharge capacity (0.18 mC) to that in UHV (0.16 mC) (Supplementary Fig. S2), the vanadium valence state was found to decrease only slightly from 4.5+ (at 2.1 Vcell) to 4.25+ (at 0 Vcell) in Fig. 3b. The charge can be attributed largely to the formation of reaction products associated with the reduction of molecular oxygen by lithium ions on the surface of LixV2O5, which gradually covered the LixV2O5 surface upon discharge. This observation is further substantiated by the appearance and growth of a new peak centered at 531.3 eV in the O 1s region, which became dominant at the end of discharge (at 0 Vcell) (Fig. 4a). Considering the Li 1s binding energy shifts to 54.8 eV, this new species can be assigned to Li2O2 based on the binding energies of both O 1s and Li 1s peaks of a Li2O2 reference sample (O 1s: 531.3 eV and Li 1s: 54.7 eV)30. More interestingly, a minor peak centered at 528.5 eV was found to appear and grow at voltages lower than 0.2 Vcell, indicative of Li2O formation (Li2O: 528.5 eV)31, where further studies are needed.

Figure 4
figure 4

In situ APXPS data of Li 1s, O 1s and V 2p collected under p(O2) = 5 × 10−4 atm.

(a) Discharge from OCV (2.1 Vcell) to 0 Vcell. (b) Charge from OCV (0.1 Vcell) to 3.0 Vcell.

Upon charging, the intensity of Li 1s peak was decreased while that of V 2p was increased greatly in contrast to discharge (Fig. 4b and Supplementary Fig. S6). The charge capacity of the LLTO/LiPON/LixV2O5 cell in the presence of oxygen (Supplementary Fig. S2) can be attributed to not only lithium de-intercalation as evidenced by the increased vanadium valence state from 4.2+ (0.1 Vcell) to 4.8+ (3.0 Vcell) in Fig. 3b, but also the oxidation and removal of Li2O2 on the surface of LixV2O5. The latter is supported by the reduction of O 1s peak for Li2O2 centered at 531.3 eV and the shift of the Li 1s binding energy from 54.8 eV (Li2O2)30 back to 55.5 eV (Li2CO325/LixV2O526/LiPON27). The onset of Li2O2 oxidation began at low applied potentials such as 1.6 Vcell, corresponding to voltages no greater than ~3.2 V versus Li+/Li (VLi) and Li2O2 was removed completely at the end of charge (3.0 Vcell), which re-exposed the LixV2O5 surface (Fig. 4b). Considering the thermodynamic reversible potential of bulk Li2O2 at 2.96 VLi32, the oxidation of Li2O2 can occur at overpotentials as low as ~0.2 Vcell ( = 3.2 VLi), which is in contrast to large charging overpotentials noted typically for Li-O2 cells with aprotic electrolytes (~1.0 V cell = 4.0 VLi)33.


Further studies are clearly needed to elucidate the physical origin to low overpotentials of Li2O2 electrochemical oxidation observed in this study. We here discuss three possibilities to highlight the differences between Li2O2 formed in this study and those formed in Li-O2 batteries with aprotic electrolytes. Li2O2 particles formed on the surface of LixV2O5 are extremely thin. Considering the volumetric capacity Li2O2 (9716 C/cm3) and the measured capacity associated with Li2O2 formed on discharge (0.71 mC/cm2), the average thickness of maximum Li2O2 coverage on the LixV2O5 surface was estimated to be ~0.7 nm, corresponding to ~1–2 unit cells of Li2O2 (Supplementary Information). This is in contrast to Li2O2 particles having sizes in the range of 100 to 1000 nm formed in Li-O2 batteries with aprotic electrolytes7,33,34. Second, the surfaces of Li2O2 are free of carbonate species formed on the surface of the LLTO/LiPON/ LixV2O5 cell (no growth of the C 1s peak intensity during discharge in presence of O2 in Supplementary Fig. S7), which is in contrast to the coverage of Li2CO3-like species on Li2O2 particles even with ether-based electrolytes16. Third, very thin Li2O2 particles may have stoichiometry and electronic properties considerably different from Li2O2 particles. This is supported by a recent density functional theory study reporting that the reduced coordination of oxygen atoms at the surface of Li2O2 yields the formation of a thin metallic region localized at the surface35, which may facilitate the electro-oxidation of the Li2O2.

This study has, for the first time, revealed reversible oxygen reduction and evolution on an oxide surface using a solid-state lithium cell using in situ APXPS, as summarized in Fig. 5. Through the direct visualization of the formation and disappearance of Li-O2 reaction products, this study connects the electrical potential to oxygen redox chemistry in presence of lithium ions at room temperature and leads to numerous opportunities for exploiting in situ APXPS to gain mechanistic insights into air-based electrochemical reactions for efficient energy storage.

Figure 5
figure 5

A schematic summarizing the reaction mechanism.

During discharge and charge in UHV, lithium ions reversibly intercalate/de-intercalate into/from the LixV2O5 electrode. During discharge in p(O2) = 5 × 10−4 atm, lithium ions meet with reduced oxygen on the surface of the LixV2O5 electrode forming Li2O2, which is decomposed upon recharge in p(O2) = 5 × 10−4 atm.


Negative electrode fabrication

Lithiated Li4Ti5O12 (LLTO) thin film was used as the negative electrode in this study. Li4Ti5O12 powder was synthesized by ball milling the stoichiometric amounts of Li2CO3 (Mallinckrodt – 99.5+%) and TiO2 (Aldrich –Anatase 99.9+%), which was pressed into a 2″ diameter disk and then fired in air at 950°C for 10 hr. To ensure O2 recovery upon cooling, the disk was cooled slowly at 0.5°C/min to 700°C, which was held at 700°C for 20 hr and finally cooled down at 10°C/min to room temperature. The dense pellet was subsequently bound to a Cu plate and used as a target for magnetron sputtering. LTO films were grown by means of r.f. magnetron sputtering in an Ar atmosphere (Air Liquide – Research Grade), at an applied RF power of 80 W at a pressure of 5 mtorr, where deposition was performed once the base pressure in the chamber reached 10−6 torr. The films were grown on 1 cm Al2O3 disks (Valley Design) coated with 0.5 μm of Pt (Refining Systems, Las Vegas Nevada, USA −99.99%) on both sides, which acted as the negative electrode current collector. Sputtering of the Pt was performed in direct current mode using 25 W and 15 mtorr Ar pressure. The as-deposited LTO films with a typical thickness of 750 nm were annealed in air at 700°C for 1 hr to develop the spinel structure.

The LTO films were lithiated electrochemically in standard Swagelok cells to form Li7Ti5O1222. The cells were prepared in an Ar filled glove box using an Li metal anode (0.75 mm Alfa Aesar −99.9%), two pieces of Celgard 2500 separator and a few drops of 1.2 M LiPF6 in ethylene carbonate/ethyl methyl carbonate electrolyte (Ferro), which was found to have water content < 2 ppm by Karl-Fisher Titration (Metler-Toledo). The cells were discharged at 5 μA current to 1.2 V using a Maccor Battery Tester. A representative discharge profile for the lithiation of the LTO films used for solid-state cells is shown in Supplementary Fig. S8, which reveals a large plateau at ~1.53 VLi associated with lithium intercalation into the LTO spinel structure. The Swagelok cells were disassembled in the glove box and the lithiated LTO anodes were washed for 5 min by soaking in dimethyl carbonate (Aldrich −99% Anhydrous) to remove residual electrolyte and left to dry on filter paper in the box.

Solid electrolyte fabrication

The lithiated LTO electrodes were loaded into a vacuum chamber and pumped down until a base pressure below 3 × 10−6 torr was attained. A 1 μm thick lithium phosphorus oxynitride (LiPON) solid electrolyte layer was prepared by r.f. magnetron sputtering of a Li3PO4 target in a N2 atmosphere (Air Liquide – Research Grade) at an applied power of 100 W, 20 standard cubic centimeters per minute (sccm) N2flow and a pressure of 20 mtorr11.

Positive electrode fabrication

Vanadium oxide (V2O5)23 thin films were prepared by dc magnetron sputtering using a vanadium target (Kurt J. Lesker −99.9%). The films were deposited in a mixture of argon (20 sccm-Air Liquide – Research Grade) and O2 (4 sccm-Air Liquide – Research Grade) at an applied power of 19 W and a deposition pressure of 11 mtorr. A homemade mask, cut from a stainless steel mesh, was used to add some small lines to the cathode surface. The thickness of the V2O5 electrode with a diameter of 0.5 cm was approximated to be 23 nm (assuming an expected deposition rate of 2.5 Å/sec for 90 seconds and on a flat surface) (corresponding to an electrode volume of 4.52·10−7 cm3). The solid-state cells were stored in an Ar filled glove box and shipped in sealed argon filled stainless steel vessels to the Lawrence Berkeley National Laboratories (LBNL) Advanced Light Source (ALS) for in situ experiments.

Experimental details of in situ APXPS

XPS data of Li 1s, C 1s, O 1s and V 2p were collected from the top surface of solid-state LLTO/LiPON/LixV2O5 cells in UHV and in presence of oxygen as a function of applied cell potential at beamline 9.3.2 at LBNL-ALS19. The cell was placed onto an insulating ceramic holder, where the Pt-coated cell bottom surface was in contact with a piece of Pt foil served as the negative electrode current collector. Spring-loaded Au-Pd coated tungsten tips were used to hold the cell in place and to connect the negative and positive electrodes to an external potentiostat24. The XPS sampling size was ~0.8 mm in diameter. No other pre-cleaning (e.g. sputtering of the surface) was carried out. XPS data were collected under the following conditions (details in Supplementary Information, Fig. S9 and S10): a) UHV (5 × 10−12 atm): and open-circuit potential (OCV) (~0.5 Vcell); b) p(O2) = 5 × 10−4 atm in the cell potentials from OCV (~0.5 Vcell) to 0 Vcell and then from OCV (~0.1 Vcell) to 3.0 V; c) UHV (5 × 10−12 atm) at potentials from OCV (~2.2 Vcell) to 0 Vcell and then OCV (~0.1 Vcell) to 3.0 Vcell; d) p(O2) = 5 × 10−4 atm at potentials from OCV (~2.2 Vcell) to 0 Vcell upon discharge and then from OCV (~0.1 Vcell) to 3.0 Vcell. Potentials were applied using a Bio-Logic SP-300 potenstiostat and the cell was held constant for 15 mins after each applied potential before XPS data collection at each experimental condition. XPS spectra were collected in the following sequence: low-resolution in the binding energy range from −10 to 595 eV and then high-resolution Li 1s, O 1s, C 1s and V 2p at a photon energy of 670 eV. Data collected from steps (c) and (d) are discussed in detail in this work.

XPS data analysis

The binding energy of all the spectra was calibrated to the C 1s photoemission peak of adventitious hydrocarbons at 285.0 eV. Shirley-type background subtraction was applied to the photoemission lines, which were fitted using a combined Gaussian-Lorentzian line shape (CasaXPS). The V 2p spectra were fitted to three components (V3+, V4+ and V5+) based on the binding energy values reported previously27,28, where all three components were constrained to have the same full-width-half-maximum (FWHM). The integrated intensity of each element and each component was extracted from the fitting, which was then normalized by its respective photo-ionization cross section. The binding energy, FWHM and integrated peak area CasaXPS) divided by their respective photo-ionization cross section36 (0.008985 Mbarn for Li 1s, 0.3134 Mbarn for O 1s, 1.084 Mbran for V 2p and 0.1257 Mbarn for C 1s at 670 eV) for each component are shown in Supplementary Table S1. The cross-section normalized intensity values of Li 1s, C 1s, O 1s and V 2p as a function of applied voltage in UHV and in presence of oxygen are shown in Supplementary Fig. S11.