If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
5 Present Address: Materials Science Department, University of Milano-Bicocca, R. Cozzi 55, Milano, MI 20126, Italy
5 Present Address: Materials Science Department, University of Milano-Bicocca, R. Cozzi 55, Milano, MI 20126, Italy
Research Laboratory of Electronics, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USADepartment of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USA
Research Laboratory of Electronics, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USADepartment of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USADepartment of Materials Science and Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USA
Li2O formation in molten-salt Li–O2 batteries is enabled by the redox of NO3− ions
NO3− is reduced to NO2− and Li2O, following which O2 oxidizes NO2− back to NO3−
Ni electrodes have higher discharge voltage due to optimum binding of NO3− and NO2−
Context & scale
Rechargeable Li–O2 batteries using molten-salt electrolytes exhibit better cycling stability, higher round-trip efficiency, and form Li2O instead of Li2O2, leading to higher gravimetric energy density. However, there is limited understanding of the reaction mechanisms during discharge and charge, which hinders their development. In this work, we find that the redox activity of NO3− ions in the molten salt is what enables the formation of Li2O, where the reaction occurs through the reduction of NO3− to NO2− and Li2O, which is catalyzed by transition metal oxides on the electrode surface. In an O2 environment, NO2− can be chemically oxidized by O2 to reform NO3−, leading to an apparent four-electron oxygen reduction. Based on this understanding of the reaction mechanism, we demonstrate that the oxygen 2p-band center can serve as a descriptor for the transition metal oxide catalyst, where the highest discharge voltages are observed on NiO, which has the optimum binding of NO3− and NO2−.
Li–O2 batteries can provide greater gravimetric energy than Li-ion batteries but suffer from poor efficiency and cycle life due to the instability of aprotic electrolytes. In this study, we show that the apparent four-electron oxygen reduction to form Li2O in Li–O2 batteries with molten nitrate is facilitated by the electrochemical reduction of nitrate to nitrite, and subsequent chemical oxidation of nitrite to nitrate by molecular oxygen, instead of a four-electron oxygen reduction aided by disproportionation of Li2O2 generated from two-electron reduction of molecular oxygen. By examining a series of transition metal catalysts using experiments and computation, optimizing the surface binding of nitrate to enhance the kinetics of the electrochemical reduction of nitrate to nitrite, as well as increasing the kinetics of nitrite oxidation by O2 was shown to increase the discharge voltage and render the observed high-rate capability for NiO-based surfaces in Li–O2 batteries.
Considering that oxygen reduction products such as Li2O2 and Li2O are largely insoluble in organic electrolytes, computational studies have shown that thermodynamic barriers for oxygen reduction and evolution on the surface of Li2O2 are small,
suggesting fast oxygen reduction and evolution kinetics. Unfortunately, Li–O2 batteries with aprotic electrolytes suffer from large voltage hysteresis between discharge and charge, low round-trip energy efficiency (<80%),
have shown that Li–O2 batteries can be cycled in molten nitrates (LiNO3/KNO3) with a reversible two-electron oxygen reduction, having a small voltage hysteresis of ∼0.1 V. Subsequently, reversible four-electron oxygen reduction has been reported with Ni
electrodes. These Li–O2 batteries with LiNO3/KNO3 can deliver gravimetric energy of ∼220 Wh/kgelectrode (∼11 mAh/cm2geo, where “geo” refers to the geometric surface area of the electrode and only the weight of Ni is considered) and a cycle life of 150 cycles (∼10 Wh/kgelectrode at 0.5 mAh/cm2geo) but at moderate current densities of up to 0.2 mA/cm2geo.
Many open questions regarding the mechanism of oxygen reduction and evolution in such Li–O2 cells with molten nitrate electrolyte remain, and further studies can provide insights into increasing cell energy and power. Discharge with carbon-based oxygen electrodes produces Li2O2 by a two-electron reduction,
where electrochemical oxygen reduction is proposed to give rise to peroxide and peroxide disproportionates to oxide (Figure S1). Recently, molten nitrate electrolytes were shown to be redox active in Li–Ar cells,
where nitrate can be reduced reversibly to produce nitrite and Li2O (LiNO3 + 2e− + 2Li+ => LiNO2 + Li2O, Eo = ∼2.6 VLi estimated in the supplemental information). As this nitrate/nitrite redox reaction can take place at similar discharge voltages to those of the Li–O2 batteries,
the nitrate/nitrite redox may participate in the oxygen reduction and evolution in Li–O2 batteries, which is the subject of this study.
In this study, we show that four-electron oxygen reduction to form Li2O in Li–O2 batteries with LiNO3/KNO3 is facilitated by the electrochemical reduction of nitrate to nitrite and subsequent chemical oxidation of nitrite to nitrate by molecular O2. Isotopic labeling experiments revealed that the oxygen atoms in the O2 that evolved during charging of the molten-salt Li–O2 batteries came from the nitrate electrolyte instead of molecular oxygen. Raman spectroscopy measurements of the oxygen electrode following discharge in 36O2 (18O18O) with 16O-based LiNO3/KNO3 reveal that Li2O does not contain 18O but LiNO3/KNO3 does contain 18O. Further support comes from differential electrochemical mass spectrometry (DEMS) that detected only 32O2 upon charging of such electrodes discharged in 36O2 (18O18O). The discharge voltage of such Li–O2 cells was sensitive to the surface chemistry of the oxygen electrode. The kinetics of the apparent four-electron oxygen reduction to Li2O (4Li+ + 4e− + O2 => 2Li2O) can be limited by the slowness in the electrochemical reduction of nitrate to nitrite (2LiNO3 + 4e− + 4Li+ => 2LiNO2 + 2Li2O) due to weak NO3− adsorption or the desorption of NO2−, which can be impeded by the strong surface binding strength of NO2− and the poor kinetics for surface oxidation of NO2− by O2. Oxygen electrodes with NiO-like surfaces were found to have fast kinetics for electrochemical NO3− reduction to NO2− and surface oxidation of NO2− by O2, which is responsible for the observed high voltages and rate capability for discharge in O2 and Ar with LiNO3/KNO3.
Results and discussion
Four-electron oxygen reduction is facilitated by nitrate redox
We first show that the redox of nitrate anions in molten-salt electrolytes is essential to enable the four-electron reduction of molecular oxygen. Discharging Li–O2 (32O2) cells in LiNO3/KNO3 at 423 K (Figure 1A) prepared similarly to previous work
at 791 cm−1) in Figure S2, X-ray diffraction (XRD) and scanning electron microscopy (SEM) in Figure S3, and titration using titanium (IV) oxysulfate (Figure S4). In addition, the amount of Li2O determined from acid-base titration (0.170 mmol, Figure 1C) agreed with that estimated from the discharge capacity (10 mAh = 0.187 mmol Li2O). Unfortunately, these observations are not sufficient to confirm that oxygen molecules are reduced directly to Li2O (4Li+ + 4e− + O2 => 2Li2O, Eoeq. = 2.91 VLi
), as discharging in LiNO3/KNO3 without oxygen (in Ar, Figure 1B) can also make Li2O (Figure 1C). The discharge capacity found in Ar can be attributed to the electrochemical reduction of nitrate to nitrite (LiNO3 + 2e− + 2Li+ => LiNO2 + Li2O, Eo = 2.57 VLi estimated in the supplemental information), producing equimolar nitrate and Li2O (Figure 1C), in agreement with recent work.
Using 18O-isotope labeling experiments, we show that Li2O formed on discharge derived its oxygen from reducing nitrate anions in the molten-salt electrolyte. Discharging Li–O2 cells with LiN16O3/KN16O3 in 36O2 (18O18O) at 423 K showed similar discharge and charge voltages (Figures 1D, S5, and S6), comparable capacities (Figure S6), and amounts of Li2O (Figures 1E and S3A) to those in 32O2 (16O16O). Oxygen-pressure-tracking measurements of 32O2 (16O16O) and 36O2 (18O18O) consumption during discharge showed a 4e−/O2 reduction process (Figure S7). Remarkably, DEMS measurements of oxygen evolution upon charging electrodes discharged in 36O2 did not yield 36O2 (18O18O) but instead 32O2, where signals from 36O2 as well as 34O2 (16O18O) were negligible within experimental uncertainty (Figure 1D). This observation indicates that molecular oxygen was not reduced directly to form Li2O (2Li+ + 1/2O2 + 4e− => Li2O).
Instead, the detection of 32O2 upon charge indicates the formation of Li216O by oxygen transfer from nitrate upon discharge of Li–36O2 cells. This hypothesis was confirmed by comparable Raman Li–O stretching of Li2O discharged in 36O2 to that of the reference Li216O and that discharged in 32O2 (Figure 1E).
The formation of Li216O came from the presence of N16O3− anions in the electrolyte, the only significant 16O source in the Li–36O2 cell. The participation of NO3− in the formation of Li2O was substantiated by Raman spectroscopy (Figure 1E). The two major peaks at 1,045 and 1,065 cm−1 found for LiN16O3/KN16O3 before discharge can be attributed to nitrate anion symmetric stretching (n1(Ag)),
dominated largely by K+ and Li+ coordination, respectively, which is derived from comparisons with reference to LiN16O3 and KN16O3 spectra in Figure S8. Of significance to note is that discharge in 36O2 led to the appearance of new peaks (1,033, 1,028, 1,023, 1,014, and 1,005 cm−1) in the region of nitrate anion symmetric stretching, which were not observed after discharge in 32O2 (Figure 1E). These new peaks could be attributed to red shifts in the symmetric stretching of nitrate anions induced by different degrees of 18O substitution relative to N16O3−. This assignment is supported by the simulated wavenumber of symmetric stretching as a function of 18O-substituted-N16O3− (Figure 1E) obtained from density functional theory (DFT) calculations of the isolated anion in vacuum at the B3LYP/6-31G(d,p) level, where greater red shifts in the symmetric stretching were correlated with more 18O substitution in N16O3−. The higher relative peak intensity at ∼1,045 cm−1 after discharge in 36O2 than that after discharge in 16O2 could be explained by additional contributions from red-shifted stretching modes from 18O-substituted-LiNO3. This hypothesis is in agreement with the observation that these red-shifted peaks resulting from 18O-substituted N16O3− remained, but those from Li216O disappeared after charging (Figure S9). Further support comes from the fact that replacing LiNO3/KNO3 with lithium bis(trifluoromethanesulfonyl)imide (LiTFSI)/potassium bis(trifluoromethanesulfonyl)imide (KTFSI) resulted in much reduced discharge capacities of Li–O2 batteries (Figure 1B), with no Li2O shown from Raman and titration measurements (Figure S10). Therefore, these observations indicate that the reduction of nitrate to nitrite plays a critical role in the formation of Li2O from a four-electron reduction of molecular oxygen to provide high capacities in Li–O2 batteries.
We propose that the formation of Li2O during the discharge of Li–O2 cells with LiNO3/KNO3 is accompanied by the reduction of NO3− to NO2− (2LiNO3 + 4e− + 4Li+ => 2LiNO2 + 2Li2O), where NO2− can be oxidized subsequently to NO3− by O2 (2NO2− + O2 => 2NO3−), giving rise to apparent 4e−/O2 reduction (4Li+ + 4e− + O2 => 2Li2O). The oxidation of NO2− by O2 is in agreement with the observation that NO2− was not detected after discharge in O2 using the Griess method
(Figure 1C) or Raman (Figure S2). In contrast, equimolar Li2O and NO2− (Figure 1C) were found after discharge in Ar, in agreement with NO3− reduction to NO2− (2LiNO3 + 4e− + 4Li+ => 2LiNO2 + 2Li2O), which was accompanied by Raman peaks
showing that the reaction is energetically favorable (LiNO2 + 1/2O2 = LiNO3, ΔGo = −0.68 eV, details in the supplemental information). Moreover, the kinetics of NO2− oxidation by O2 was fast enough to replenish NO3− consumed in the discharge of Li–O2 cells. The rate of NO2− oxidation by O2 was estimated from pressure tracking measurements of LiNO3/KNO2 in the presence of O2 (Figure S13), where the NO2− oxidation rate was 8.86 mmol/h, and the O2 consumption rate was 4.43 mmol/h, corresponding to 0.47 mA, which is sufficient to sustain the current (0.127 mA) applied to Li–O2 cells (Figure 1). Furthermore, this mechanism is supported by DEMS measurements of discharged electrodes in Ar (Figure S14), which showed no O2 evolution upon charging to 3.5 VLi, having charge attributable to the oxidation of NO2− to NO3− (Li2O + LiNO2 + 2e− <=> LiNO3 + 2Li+), in agreement with Giordani et al.
Enhancing the power and gravimetric energy of Li–O2 batteries
The power of Li–O2 cells was increased by adding nano-Ni (80–150 nm, 4.84 m2/g) particles to the Ni325 particles (1–5 μm, 0.68 m2/g) in the electrode. The nano-Ni/Ni325 (1:10 weight ratio) electrodes (0.11 gNi/cm2geo) were shown to maintain similar discharge voltages up to current densities as high as 2.0 mA/cm2geo (Figure 2A), providing a power density up to ∼5.5 mW/cm2geo, while a significant reduction in the discharge voltage at current densities greater than 0.5 mA/cm2geo was found for those without nano-Ni (Figure S15). The reduction in the discharge capacity with increasing current density (Figures 2A and S18) can be attributed to the slow diffusion of molecular oxygen in the electrode, considering the limiting current density of ∼3.7 mA/cm2geo (estimated from oxygen solubility and diffusivity in the electrolyte, detailed in the supplemental information). Li–O2 cells with nano-Ni/Ni325 electrodes were shown to provide stable cycling at 1.0 mA/cm2geo (Figure 2B), having the Coulombic efficiency stabilized at ∼100% after the first 25 cycles (Figure S17).
The gravimetric capacities of such Li–O2 cells can be increased greatly by replacing nano-Ni/Ni325 electrodes with Ni-plated Vulcan carbon (XC72) supported on carbon paper (Ni/VC/CP having 787 cm2VC/cm2geoFigure S19A). By examining the different current densities of Ni plating (Figures S19B–S19D), dense and uniform Ni was plated onto the surface of VC/CP at low plating currents such as 0.25 mA/cm2geo, which was critical to prevent the direct O2 reduction from forming LiO2 and Li2O2 on the exposed surface of VC, which could lead to parasitic reactions with the carbon surface and poor cycling performance
(Figures S20–S22). Analyses of discharged electrodes via acid-base and titanium (IV) oxysulfate titrations as well as the Griess method, revealed that the amount of Li2O (23 μmol corresponding to 1.6 mAh/cm2geo) was in reasonable agreement with the discharge capacity (Figure S22), whereas there was no detectable Li2O2. After a few cycles of activation (Figure S23A), stable cycling of Ni/VC/CP electrodes at 0.2 mA/cm2geo was obtained over 80 cycles with reversible capacities of ∼0.35 mAh/cm2geo (Figures 2C and S23B). If the total mass of the Ni/VC/CP electrode were considered, a gravimetric discharge capacity of 25.5 mAh/gNi/VC/CP was obtained (Table S5). Here, we show that CP is inactive based on the following observations. First, negligible discharge capacities were found for discharge with neat CP (Figure S20A). Second, SEM images of the electrodes after Ni deposition showed that Ni was mainly deposited on VC (Figure S19), and SEM images after discharge revealed that Li2O was only observed on Ni/VC but not on CP (Figure S24). Third, energy dispersive spectroscopy (EDS) analysis after discharge showed negligible oxygen (from either Li2O or the molten salt) on the CP fibers, suggesting they remained inactive during discharge, whereas signals from O, K (from the molten salt), and Ni were mainly found in the Ni/VC layer (Figure S25). Therefore, we report the high gravimetric energy of ∼2,900 Wh/gNi/VC (based on 1,100 mAh/gNi/VC, though this number becomes ∼680 mAh/gNi/VC/Li2O when the weight of deposited Li2O is included) without considering the weight of CP, and such findings provide insights into strategies to develop practical Li–O2 batteries with the power and energy rivaling that of Li-ion batteries. In the following section, we focus on the fundamental understanding of the kinetics of NO3− reduction to NO2−, the oxidation of NO2− to NO3− by O2 (2NO2− + O2 => 2NO3−), and 4e−/O2 reduction to Li2O, and examine whether there are other electrodes than Ni that can provide comparable or greater rate capability.
The discharge voltage of Li–O2 cells with LiNO3/KNO3 was found to be surface dependent. Replacing Ni with Ti in Ti/LiNO3/KNO3/SS electrodes significantly reduced the discharge voltage plateau to ∼2.1 VLi (Figure 3A). The adding of Ni(NO3)2 to Ti/LiNO3/KNO3/SS electrodes increased the major voltage plateau to 2.8 VLi (Figure 3A), similar to Ni/LiNO3/KNO3/SS (Figure 1B). In addition, a minor voltage plateau at ∼2.9 VLi appeared with Ni(NO3)2 addition, where the capacity increased with the increasing concentration of Ni(NO3)2. This minor plateau for Ti/LiNO3/KNO3/SS electrodes with added Ni(NO3)2 is similar to that observed for Ni/LiNO3/KNO3/SS electrodes where discharge was terminated following the minor plateau (Figure 3B), suggesting a common origin. This hypothesis is supported by the following facts. First, Ni(NO3)2 was detected after stirring NiO powder in LiNO3/KNO3 at 423 K in Ar, as evidenced by the appearance of a Raman peak at 1,090 cm−1 (indicative of the coordination of NO3− with Ni2+ in Figure S27). Second, inductively coupled plasma (ICP) analysis revealed water-soluble Ni species from Ni/LiNO3/KNO3/SS electrodes (Table S1; Figure S26). Third, X-ray photoelectron spectroscopy (XPS) analysis revealed similar Ni 2p features for Ti/LiNO3/KNO3/SS with added Ni(NO3)2 (Figure S28) and Ni/LiNO3/KNO3/SS (Figure S29). Discharging Ti/LiNO3/KNO3/SS with added Ni(NO3)2 in Figure 3A, as well as Ni/LiNO3/KNO3/SS electrodes to the end of the minor plateau only (Figure 3B), resulted in the emergence of a broad feature at ∼500 cm−1 via Raman spectroscopy, which was absent before discharge and exhibited greater peak intensities discharged in O2 than Ar (Figure 3D). This peak feature could be attributed to the Ni–O stretching mode
of NiO-like species (Figure 3D). The absence of the Raman signal for Li2O following discharge to the end of the minor plateau is in agreement with negligible amounts of Li2O (Figure 3C) detected from the Griess and acid-base methods (with an Li2O detection limit of 0.2 μmol), which were much lower than those expected from the discharge capacities (17.6 μmol of Li2O expected from 0.74 mAh/cm2geo in O2 in Figure 3B). We therefore propose that the discharge process for the minor plateau at ∼3.0VLi stems from the formation of NiO-like species and not Li2O via 2Li+ + ½O2 + Ni(NO3)2 + 2e− = NiO + 2LiNO3 (Eo = 3.68 VLi), or 3Li+ + O2 + Ni(NO3)2 + 3e−= LiNiO2 + 2LiNO3 (Eo = 3.47 VLi) in an O2 environment, and 2Li+ + Ni(NO3)2 + 2e− = NiO + LiNO2 + LiNO3 (Eo = 3.41 VLi) or 3Li+ + Ni(NO3)2 + 3e− = LiNiO2 + 2LiNO2 (Eo = 3.12 VLi) in an Ar environment (where calculation details can be found in the supplemental information). Therefore, the formation of NiO-like species on the surface of the oxygen electrode in lieu of Li2O upon initial discharge is critical to enable Li–O2 discharge with small overpotentials (at ∼2.8 VLi in Figures 1B and 3A) and fast kinetics of four-electron oxygen reduction to generate Li2O, which contrasts with large overpotentials (∼2.1 VLi in Figure 3A) for TiO2-like surfaces of Ti/LiNO3/KNO3/SS.
Adding transition metal nitrates other than Ni(NO3)2 to Ti/LiNO3/KNO3/SS electrodes could also increase the voltages of the major plateau during discharge in O2 (Figure 4A) as well as in Ar (Figure 4B). Interestingly, the voltages for the major discharge plateau in O2 (Figure 4A) were found to systematically increase in the order from Mn(NO3)2 < Cu(NO3)2 = Fe(NO3)3 < Co(NO3)2 = Ni(NO3)2. Li2O was detected in all these electrodes discharged in O2 from Raman (Figure 4C) and XRD (Figure S31), where amounts determined from titration (Figure 4D) are in good agreement with those expected from the discharge capacities in the major plateau. In addition, NO2− (Figure 4D) and Li2O2 (Figure S32) were not detected in these discharged electrodes in agreement with Ni/LiNO3/KNO3/SS electrodes (Figure 1C). Similar to adding Ni(NO3) to Ti/LiNO3/KNO3/SS (Figure 3A), Mn3O4, Cu2O, Fe3O4, and Co3O4-like species were formed on the surface of Ti/LiNO3/KNO3/SS electrodes from corresponding transition metal nitrates
during the initial minor plateau upon discharge in O2, as revealed from XPS (Figure S28; Table S2) and SEM (Figure S30). As the metal oxides deposited on the Ti/LiNO3/KNO3/SS electrodes were expected to have similar total surface areas due to adding the same amount of metal nitrate and their similar morphologies (Figure S30), the increasing discharge voltage in Figure 4A is indicative of greater kinetics for four-electron oxygen reduction to Li2O from Mn3O4, Cu2O, Fe3O4, and Co3O4 to NiO.
In order to understand the origin of this activity trend, we further examine how these surfaces affect the kinetics of NO3− to NO2− (2LiNO3 + 4e− + 4Li+ => 2LiNO2 + 2Li2O) and NO2− oxidation to NO3− by O2 (2NO2− + O2 => 2NO3−), as our proposed four-electron oxygen reduction to Li2O (4Li+ + 4e− + O2 => 2Li2O) involves these two processes. The kinetics of electrochemical NO3− reduction to NO2− (2LiNO3 + 4e− + 4Li+ => 2LiNO2 + 2Li2O) and chemical NO2− oxidation to NO3− by O2 (2NO2− + O2 => 2NO3−) were found to be oxide-dependent. The discharge voltage of the major plateau in Ar was found to increase in the order from Fe3O4, Mn3O4, Cu2O, and Co3O4 to NiO, generated from corresponding M(NO3)x added to Ti/LiNO3/KNO3/SS electrodes, indicative of increasing kinetics for NO3− reduction to NO2−. Discharge in Ar generated comparable amounts of Li2O (Raman in Figure 4C and XRD in Figure S33) and NO2− (Figure 4D), as expected from the process of NO3− + 2e− + 2Li+ => NO2− + Li2O. This trend is in reasonable agreement with recent work that reports increasing kinetics of NO3− reduction to NO2− from Cu = Fe, < Co < Ni-based electrodes.
We next examine the molecular origin for the kinetics trend of NO3− reduction to NO2− on these oxide surfaces using DFT. The reduction of LiNO3 to LiNO2 was computed in three steps (Figure 5A). First, NO3− and Li+ adsorb onto the oxide surface to form adsorbed –NO3−–Li+, where –NO3− adsorbs in a bidentate configuration on surface metal sites and Li+ adsorbs on a surface oxygen site (step 1). Second, –NO2−–Li+ is formed by the reaction with 2Li+ and 2e−, yielding Li2ONO3 (step 2), where –NO2−–Li+ adsorbs in a bidentate configuration similar to the one of –NO3−–Li+. We here consider the overall two e− reduction of –NO3−–Li+ to –NO2−–Li+ and Li2ONO3, where this reaction could be a single 2e− step or could contain two 1e− steps. Third, –NO2–Li+ desorbs and dissolves into the electrolyte (step 3). Although the Gibbs free energy barrier of –Li+–NO3− adsorption (step 1) was found to increase from Mn3O4, NiO to Cu2O, that for –Li+–NO2− desorption (step 3) decreased greatly. Specifically, the kinetics of LiNO3 reduction to LiNO2 are limited by –Li+–NO2− desorption from the metal sites of Mn3O4, while those on Cu2O are limited by –Li+–NO3− adsorption. It should be noted that the electrochemical reduction of NO3− to NO2− catalyzed on the metal sites of oxides in this work occurs at more reducing conditions than those of chemical NO oxidation by molecular oxygen catalyzed on the oxygen sites of oxides.
The computed trend for the kinetics of NO3− reduction to NO2− (Figure 5A) is in good agreement with the discharge voltage of the major plateaus in Ar (Figure 4B), where NiO showed the lowest Gibbs free energy barrier for all three steps among the oxides examined in Figure 5A and the highest discharge voltage for Ti/LiNO3/KNO3/SS with Ni(NO3)2 addition (Figure 4B). The discharge voltage of the major plateau (Figure 4B) was found to exhibit a volcano trend, with the computed bulk (Figure 5B) and surface (Figure S34) O 2p band center of Mn3O4, Fe3O4, Co3O4, NiO, and Cu2O
It is proposed that the kinetics on the left branch is limited by –Li+–NO3− adsorption due to weak binding and that on the right branch is limited by –Li+–NO2− desorption due to strong binding (Figure 5B). These results are in agreement with previous findings of electroreduction of nitric acid on metals, which have identified nitrate binding as the critical descriptor for reaction kinetics.
The formation of large Li2O particles upon discharge (Figure S3B) can be rationalized based on the solubility of Li2O in the LiNO3/KNO3 molten salt, which was measured to be 0.3 wLi2O/w% or 0.22 molLi2O/L (Figure S35).
We further investigate the oxidation kinetics of NO2− by O2, which were also found to be surface dependent. A mixture of LiNO3/KNO2 eutectic salts and metal oxides (Mn3O4, Fe3O4, Co3O4, NiO, and Cu2O) was pressed as pellets and then transferred into an O2-filled cell, after which the residual NO2− was quantified using the Griess method (supplemental information) to deduce the NO2− oxidation rate by O2. NiO was the most reactive among the examined transition metal oxides and the kinetics increased with the order from Mn3O4 < Cu2O < Fe3O4 < Co3O4 < NiO. The lowest reaction rate found for Mn3O4 was ∼0.2 mol/L·h, which is comparable with the rate measured in bulk electrolyte solution without adding transition metal oxides (∼0.2 mol/L·h; Figure S13). Such oxidation rates of NO2− by O2 would project limiting currents of 0.7 mA/cm2geo (Mn3O4) and 3.1 mA/cm2geo (NiO), which is in agreement with no significant voltage reduction for the major plateau of Li–O2 cells Ni/LiNO3/KNO3/SS at current densities up to 2 mA/cm2geo. The oxidation of NO2− by O2 during discharge in O2 can result in a lower buildup of –NO2−–Li+ than that in Ar for metal oxides on the right branch in Figure 5B. This argument is in agreement with a larger voltage reduction in Ar for Ti/LiNO3/KNO3/SS with added Co(NO3)2, Fe(NO3)3, or Mn(NO3)2, relative to that with Ni(NO3)2, than that in O2. Further support came from the greater reduction observed in the discharge voltage, with increasing current densities from 0.1 (∼2.5 VLi) to 2.0 mA/cm2geo (∼2.2 VLi) in Ar (Figure S16) than in O2 for nano-Ni/Ni325 electrodes. Therefore, we propose that the apparent four-electron oxygen reduction to Li2O (4Li+ + 4e− + O2 => 2Li2O) in LiNO3/KNO3 involves electrochemical NO3− reduction to NO2− (2LiNO3 + 4e− + 4Li+ => 2LiNO2 + 2Li2O) and NO2− oxidation to NO3− by O2 (2NO2− + O2 => 2NO3−) in Figure 5B. The kinetics are limited by NO3− adsorption or the desorption of NO2−, which can be influenced by surface binding strength of NO2− and the kinetics for surface oxidation of NO2− by O2. The higher discharge potentials in an O2 environment compared with an Ar environment can be rationalized by the higher overall reaction potential in O2 (2.91 VLi) versus Ar (2.57 VLi) and the negligible amount of NO2− found in the electrode by titration experiments (Figure 1C). Although Cu-based electrodes were found to exhibit similar discharge voltages to Ni and cycling stability at low rates (Figure S36), NiO-like surfaces with fast kinetics for electrochemical NO3− reduction to NO2− and surface oxidation of NO2− by O2 are expected to have the highest rate capability for the discharge of Li–O2 cells with LiNO3/KNO3. Such mechanistic insights can provide insights into the design of the oxygen electrode for high-energy and high-power Li–O2 batteries.
We here disclosed a new mechanism of 4e−/O2 oxygen reduction reaction in molten-salt Li–O2 batteries, where nitrate plays a critical role in Li2O formation. Raman spectroscopy and DEMS were conducted on electrodes discharged in 36O2, indicating that the Li2O was mainly from the reduction of NO3− and not O2. By changing the metal oxide surface, the discharge voltage corresponding to the reduction of NO3− to NO2− exhibited a volcano-shaped dependence on the O 2p-band center of metal oxides (Mn3O4, Fe3O4, Co3O4, NiO, and Cu2O). Through understanding the NO3− redox reaction and its reduction steps on metal oxides via experiments and theoretical calculations, this work provides design principles for new catalysts in the application of Li molten salt batteries and molten-salt Li–O2 batteries with a 4e−/O2 process. Based on these design principles, molten-salt Li–O2 batteries with high areal rate capability up to 2.0 mA/cm2geo, good cycling stability up to 250 cycles, and gravimetric capacity up to 1,100 mAh/gNi/VC have been demonstrated.
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Graham Leverick ( firstname.lastname@example.org ).
This study did not generate new unique reagents.
Nickel metal powder (325 mesh, 99.8%, Fisher Scientific), nano nickel metal powder (80–150 nm, 99.8%, Alfa Aesar), and Ti metal powder (325 mesh, 99.5%, Fisher Scientific) were used for electrode preparation. LiNO3 (99.98%, Fisher Scientific.), KNO3 (99.99%, Fisher Scientific), and KNO2 (97%, Alfa Aesar) were used to prepare the LiNO3/KNO3 (with a weight ratio of 1/2) and the LiNO3/KNO2 (with a weight ratio of 35/65) eutectic molten salts. LiTFSI (99.99% extra dry grade provided by Solvay) and KTFSI (97%, Sigma Aldrich) were used to prepare the LiTFSI/KTFSI (with a weight ratio of 2/3) eutectic molten salt. GF (GF/A, Whatman) was used as a separator in molten-salt cells. Homemade LAGP was used as a ceramic membrane to prevent soluble Li2O and NO2− cross-over between the positive electrode and negative electrode, which was synthesized using the same method as the literature.
Nickel standard ICP solution was used for titration of solubility of Ni(NO3)2 in LiNO3/KNO3 molten salts. Mn(NO3)2·4H2O (>97%, Sigma Aldrich), Fe(NO3)3·9H2O (>98%, Sigma Aldrich), Co(NO3)2·6H2O (>98%, Sigma Aldrich), Ni(NO3)2·6H2O (99.99%, Alfa Aesar), and Cu(NO3)2·2.5H2O (99.99%, Sigma Aldrich) were used as additives in Ti-based electrodes. Li2O (97%, Sigma-Aldrich), Li2O2 (90%, Sigma-Aldrich), NiO (99.995%, Sigma-Aldrich) and LiNiO2 (>98%, Sigma-Aldrich) were used as standard references. TiO2 (99.98%, Sigma Aldrich), Mn3O4 (97%, Sigma Aldrich), Fe2O3 (>99%, Sigma Aldrich), Co3O4 (linear formula, Sigma Aldrich), NiO (99.995%, Sigma Aldrich), and Cu2O (97%, Sigma Aldrich) were used for the measurement of reaction rates of NO2−/O2 on metal oxides. VC XC72 (VC, Premetek) and Polytetrafluoroethylene (50 wt % PTFE solution, Sigma Aldrich), and CP (Freudenberg H23C2, FuelCellStore) with a gas diffusion layer (GDL) were used to prepare VC/CP electrodes. Here, we used the “Watt’s bath” method
for Ni electrodeposition with NiSO4·6H2O (Sigma Aldrich), NiCl2 (98%, Sigma Aldrich), and H3BO3 (99.5%, Sigma Aldrich) aqueous solution as electrolytes.
The preparation of Ni metal electrode (Ni325): 2 g of Ni powders (325 mesh, 99.8%, Fisher Scientific) were added into 4 mL of LiNO3/KNO3 (with a weight ratio of 1/2) solution (0.25 g/mL in deionized water [DIW]). Then the above suspensions were sonicated for 10 min and transferred into an oven for drying at 453 K for 2 h. Then, the composite powder was grounded for a half an hour and pressed as electrodes (0.2 g and 12.7 mm of diameter) on stainless steel mesh (120 × 120) with 2 tons of pressure for 1 min in an Ar-filled glovebox. These electrodes were transferred into a vacuum Buchi glass oven at 473 K for 2 days and then stored in an Ar-filled glove box. After preparation, there is a small amount of NO2− in the Ni/LiNO3/KNO3 (with a weight ratio of 1/2)/SS electrodes, which could be attributed to the reactions:
Ni + LiNO3 = NiO + LiNO2 (ΔG0 = −132.0 kJ/mol from thermodynamic data in Table S6), Ni + KNO3 = NiO + KNO2 (ΔG0 = −124.4 kJ/mol from thermodynamic data in Table S6) or 3Ni + 6LiNO3 = Ni(NO3)2 + 2LiNiO2 + 4LiNO2 (ΔG0 = −174.1 kJ/mol from thermodynamic data in Table S6).
The preparation of modified Ni metal electrode (nano-Ni/Ni325): 0.2 g of nano-Ni (80–150 nm, 99.8%, Alfa Aesar) powers and 2.0 g of micro-sized Ni powders (325 mesh) were added into the 4.4 mL of LiNO3/KNO3 (with a weight ratio of 1/2) solution (0.25 g/mL in DIW). The following process is the same as the preparation of the above Ni325 electrode.
The preparation of Ti-based electrodes with transition metal cations: 2 g of Ti powder (325 mesh, 99.5%, Fisher Scientific) were added into 4 mL of LiNO3/KNO3 (with a weight ratio of 1/2) solution (0.25 g/mL in DIW) with 10–50 mM of metal nitrates (Mn(NO3)2·4H2O (>97%, Sigma Aldrich), Fe(NO3)3·9H2O (>98%, Sigma Aldrich), Co(NO3)2·6H2O (>98%, Sigma Aldrich), Ni(NO3)2·6H2O (99.99%, Alfa Aesar), and Cu(NO3)2·2.5H2O (99.99%, Sigma Aldrich). Then the above suspensions were sonicated for 10 min and transferred into an oven for drying at 423 K for 2 h. Then, the composites were grounded for a half an hour and pressed as electrodes (0.2 g/pellet) with 12.7 mm of diameter on stainless steel mesh (120 × 120) with 2 tons of pressure for 1 min in an Ar-filled glove box. These electrodes were transferred into a vacuum glass oven at 423 K for 2 days and then stored in an Ar-filled glove box.
The preparation of Ni metal electrodes (Ni/LiTFSI/KTFSI [with a weight ratio of 2/3]/SS: 2 g of Ni powders [325 mesh, 99.8%, Fisher Scientific] were added into 4 mL of LiTFSI/KTFSI [with a weight ratio of 2/3] solution [0.25 g/mL in DIW]). Then the above suspensions were sonicated for 10 min and transferred into an oven for drying at 453 K for 2 h. Then, the composite powder was grounded for a half an hour and compressed as a 12.7 mm electrode (0.2 g) on stainless steel mesh (120 × 120) with 2 tons of pressure for 1 min in an Ar-filled glove box. These electrodes were transferred into a vacuum Buchi glass oven at 473 K for 2 days and then stored in an Ar-filled glove box.
Preparation of Ni/VC/CP electrodes: 90 mg of VC XC72 (VC) and 10 mg PTFE (from 50 w% PTFE solution, Sigma Aldrich) were mixed in 1 mL of isopropanol (IPA), and ball milled for 30 min. Then, the suspension was doctor blended on a piece of CP with GDL (Freudenberg H23C2, FuelCellStore). Then, the VC-coated electrode was dried at 333 K for 1 h. The dried electrode VC/CP is used for Ni electrochemical plating. Here, we used the “Watt’s bath” to produce Ni metal on VC/CP electrode. 1 M of NiSO4·6H2O (Sigma Aldrich), 0.2 M of NiCl2 (98%, Sigma Aldrich), and 0.5 M of H3BO3 (99.5%, Sigma Aldrich) solutions were prepared as the electrolyte for the Ni electrodeposition. The Ni foam was used as the negative electrode, and the VC/CP electrode was used as the positive electrode. The two above electrodes were immersed into the electrolyte and then applied a constant current (0.25–1.0 mA/cm2geo) to produce Ni layers on VC/CP electrodes (Ni/VC/CP). Then, a layer of LiNO3/KNO3(with a weight ratio of 1/2) was cast onto the dried Ni/VC/CP electrodes using the drop-casting method. The Ni/VC/CP electrodes with salts were transferred into the Buchi glass oven at 423 K for 48 h and then placed into the glove box for battery testing.
Preparation of glass fiber separators with molten salts
The preparation of GF separators: the GF separators (18 mm, GF/A, Whatman) were soaked into 0.25 g/mL of LiNO3/KNO3 (with a weight ratio of 1/2) solution and dried on a hotplate at the temperature of 393 K. The loading of salts on a piece of GF was ∼120 mg. Then, these separators with molten salts were dried in a vacuum Buchi glass oven at 423 K for 2 days and then stored in an argon-filled glove box.
Assembly of molten-salt cells
All parts of molten-salt cells were dried in a vacuum oven at 353 K for 12 h before use. The Li negative electrode (12.7 mm in diameter) and the oxygen metal electrode (12.7 mm of diameter) were separated by a piece of GF with salts and a piece of LAGP ceramic membrane. The schematic structure of the Li molten-salt cell is shown in Figure 1A. After assembling, the cells were charged with O2 or Ar. The charged O2 or Ar pressure ranged from 15 to 40 psi at room temperature.
Molten-salt cells were measured in an oven with a temperature ranging from 423 to 443 K. The operation voltage window was set from 2.0 to 3.6 V based on different types of batteries. The applied current densities are from 0.1 to 2 mA/cm2geo. The battery measurement was conducted on a Biologic VMP3 electrochemical workstation.
The DEMS measurements were conducted on a custom-made DEMS setup.
32O2 (16O16O), 34O2 (16O18O), and 36O2 (18O18O) were detected during the charge with 15 min of accumulation time for each point. O2 pressure was measured during the discharge process to quantify the O2 consumption. Helium (Ultra High Purity 5.0 Grade, Airgas) was used as the carrier gas in DEMS measurements. The effective area of electrodes was 1 cm2 for DEMS measurement. The operation temperature was 423 K. In the DEMS measurement, molten-salt cells were kept at the open-circuit voltage (OCV) for 5 h at 423 K to ensure complete melting of LiNO3/KNO3 eutectic. The applied discharge and charge current densities were 0.1–0.2 mA/cm2geo.
Characterizations of electrodes
Ni electrodes were characterized through XRD (PANalytical X’Pert Pro), SEM (Zeiss Merlin), and Raman spectroscopy (HORIBA Scientific LabRAM HR800). XRD and Raman spectra measurements were conducted in air-tight cells. The applied voltage and current in XRD measurements are 45 kV and 40 mA, respectively, using Cu-Kα radiation (λ = 1.54178 Å). A red laser (λ = 632.8 nm) was used with a 50-fold magnification in the Raman spectra measurements. An exposure time of 15 s with 600 gratings was used, and each spectrum was accumulated 5 times.
All Ni and Ti-based electrodes were washed by DIW 3 times and dried at 323 K in a vacuum glass oven overnight before XPS measurements. All the XPS spectra were collected using a PHI 5000 VersaProbe II (ULVAC-PHI) using a monochromatized Al Kα source and a charge neutralizer. Pass energy of 23.5 eV was used, and adventitious carbon at 285 eV (C1s spectra) was used to calibrate all XPS spectra. After subtraction of a Shirley-type background, photoemission lines were fitted using combined Gaussian-Lorentzian functions.
Quantification of discharge products
The quantification of NO2− using the Griess method:
the NO2− titration was conducted on a UV-vis spectrophotometer (Genesys 180, Thermo Fisher Scientific). The standard NO2− solutions (0, 0.01, 0.02, 0.05, 0.1, 0.5, 1 mM) were prepared for the calibration curve. 50 μL of standard NO2− solution and 50 μL of sulfanilic acid (10 mg/mL solution in 5% phosphoric acid, Promega) were added into the 1.2 mL plate deep well and then was kept in the dark environment for 3–5 min. Then, 50 μL of N-(1-naphthyl)ethylenediamine dihydrochloride (1 mg/mL) solution (Promega) were added to the above solution and was kept in the dark environment for 3–5 min. After that, we transferred 100 μL of the above solution into a quartz cuvette (10 mm of path length, VWR), and then 1.9 mL of DIW was added as well. The solution in the cuvette was tested immediately using Genesys 180 (Thermo Fisher Scientific) with a scanning rate of 1 nm/s from 450 to 700 nm. The curve of absorbance versus NO2− concentration was linear fitted, as shown in Figure S11B. The samples were dispersed in 20 mL of DIW, and then the clear solution will be collected by centrifugation. The clear solution is diluted to ranging from 1/20 to 1/100. The diluted solution will be titrated using the above procedure.
The quantification of Li2O2 using titanium oxysulfate solution: discharged electrodes were extracted from discharged Li molten-salt cells in the Ar-filled glove box. Then, the electrode was dispersed in 20 mL of DIW (stored in a refrigerator at 298 K) and then stirred for 4 min. During this time, the following reaction occurred: Li2O2 + 2H2O = 2LiOH + H2O2. There is a side reaction Li2O2 + H2O → 2LiOH + 0.5O2, which can be negligible based on literature.
Next, 1 mL of the filtered solution using 0.2 μm filter was taken from the vial and added to 1 mL cooled DIW and titrated with 0.5 mL standardized titanium (IV) oxysulfate solution (Aldrich, ∼15 wt % in dilute sulfuric acid, 99.99% trace metals basis). This step allows the fast reaction between H2O2 and Ti4+ oxysulfate to form pertitanic acid, which exhibits a stable, yellow color. Ti4+ + H2O2 + 2H2O → H2TiO4 + 4H+. The concentration of the yellow pertitanic acid was determined using UV-vis spectroscopy Genesys 180 (Thermo Fisher Scientific) with a scanning rate of 1 nm/s from 350 to 650 nm. The UV-vis spectra of the titration of standardized H2O2 solutions (Certified ACS 31.7%, Fisher Chemical) at various concentrations and the corresponding calibration curve are shown in Figures S4A and S4B. The amount of Li2O2 produced by molten-salt Li–O2 batteries can be quantified using the above protocol.
The quantification of Li2O using acid-base titration: the acid-based titrations for Li2O quantification were done on a pH meter (PH 700 meter, VWR) using 0.01 N of HCl standard solution (VWR). All discharged electrodes were firstly dispersed in 20 mL of DIW, and then the clear solutions were collected via centrifugation and filtration using a 0.2 μm filter for the acid-based titration. The reaction of electrode in DIW is Li2O + H2O = 2LiOH.
When pH value approaches 7 (6.5–7.5), we assume all LiOH is consumed by HCl. In this titration, one mole of Li2O can generate two moles of LiOH, which can be quantified using HCl titration. Thus, we can get the consumed HCl and then the mole of Li2O. If there is Li2O2 according to Li2O2 quantification, we can deduct the contribution from Li2O2 and then get the mole of Li2O.
The quantification of Ni2+ in Ni/salts/SS electrode: the ICP measurements for solubility of Ni(NO3)2 in LiNO3/KNO3 eutectic molten salts were conducted on an ICP-optical emission spectrometer (ICP-OES, Agilent 5100 DVD). The solid sample was firstly dissolved in 10 mL of DIW, and then the clear solutions were collected via filtration. After adding HNO3 to make the above solution neutral and then 3 w% HNO3 was added to the solution for the ICP measurement. NiO has negligible solubility in water (Table S1), but Ni(NO3)2 has high solubility ∼8 M,
so the Ni species detected by ICP are attributed to Ni(NO3)2.
Characterization of the reactions between NO2− and O2
The LiNO3/KNO2 (with a weight ratio of 35/65) eutectic molten salts (100–130 mg) were loaded on a piece of GF (1.27 cm2 of diameter). Then, the GFs with LiNO3/KNO2 were dried in a vacuum Buchi glass oven at 423 K for 2 days and then stored in an Ar-filled glove box. A GF separator with LiNO3/KNO2 was placed in a Li–O2 cell with O2 pressure of 40 psi at room temperature. Then, the cell was placed in an oven at the temperature of 423 K for 48 h. After that, we disassembled the cell in an argon-filled glove box and characterized the reacted LiNO3/KNO2 using Raman spectroscopy in an air-tight Raman cell with Ar protection. A piece of GF (1.27 cm2 of diameter) with LiNO3/KNO2 was placed in a cell with an Ar pressure of 30 psi at room temperature. Then, the cell was stabilized in an oven with a temperature of 423 K for 2 h. After that, the cell is evacuated and purged O2 with a pressure of 30 psi. Then, the O2 pressure was recorded with time to show the reaction rate of NO2− and O2. As a controlled experiment, a piece of GF with LiNO3/KNO3 (1/2 in weight ratio) is tested using the above procedure.
The LiNO3 and KNO2 were dried in vacuum at 423 K for 48 h, and commercial metal oxides (TiO2, Mn3O4, Fe2O3, Co3O4, NiO, Cu2O) were vacuum dried at 383 K overnight. Then, these dry powder samples were transferred into an Ar-filled glove box for use. The Brunauer-Emmett-Teller (BET) measurements were conducted to obtain the surface area of these metal oxides. The surface area of these metal oxides is listed in Table S4. Then, the dried LiNO3/KNO2 salts (with a weight of 35/65) were mixed with these dried metal oxides by a normalized ratio of 0.8 g salts/1 m2 metal oxide. The mixture was manually grounded for 20 min for homogeneity in the Ar-filled glove box. Then, we pelletized a small amount of the above grounded powder samples and put them in the air-tight cell. After being charged 40 psi of O2 (at room temperature), these cells were placed in the oven at 423 K for 12 h. Here, we assume the reaction rate of NO2−/O2 is linear with time. After the reaction, the pellets were dispersed into DIW, and the clear solutions were collected via filtration using a 0.2 mm filter. Then, the residual NO2− in the clear solutions was titrated using the Griess method described in the previous section. The experiment for each metal oxide sample was repeated 3–5 times to check the reproducibility.
Data and code availability
This study did not generate any datasets.
This work is in part supported by Shell and a Department of Navy award N00014-20-1-2221 issued by the Office of Naval Research. This work made use of the MRSEC Shared Experimental Facilities at MIT, supported by the National Science Foundation under award number DMR-1419807. G.L. gratefully acknowledges partial support from a Natural Sciences and Engineering Research Council of Canada (NSERC) PGS-D and Siebel Scholarship (Class of 2020). S.F. gratefully acknowledges the Link Foundation for an Energy Fellowship .
Y.G.Z., G.L., and Y.S.-H. conceived and designed the study. Y.G.Z., G.L., L.G., S.F., R.T., and J.R.L. conducted experiments, computation, and analysis. Y.Z. and Y.Y. conducted the XPS measurements. Y.G.Z., G.L., and Y.S.-H. wrote the manuscript, and all authors edited the manuscript.