Quantitative Delineation of the Low Energy Decomposition Pathway for Lithium Peroxide in Lithium–Oxygen Battery

Abstract

Identification of a low‐potential decomposition pathway for lithium peroxide (Li₂O₂) in nonaqueous lithium–oxygen (Li–O₂) battery is urgently needed to ameliorate its poor energy efficiency. In this study, experimental data and theoretical calculations demonstrate that the recharge overpotential (η _RC) of Li–O₂ battery is fundamentally dependent on the Li₂O₂ crystallization pathway which is intrinsically related to the microscopic structural properties of the growing crystals during discharge. The Li₂O₂ grown by concurrent surface reduction and chemical disproportionation seems to form two discrete phases that have been deconvoluted and the amount of Li₂O₂ deposited by these two routes is quantitatively estimated. Systematic analyses have demonstrated that, regardless of the bulk morphology, solution‐grown Li₂O₂ shows higher η _RC (>1 V) which can be attributed to higher structural order in the crystal compared to the surface‐grown Li₂O₂. Presumably due to a cohesive interaction between the electrode surface and growing crystals, the surface‐grown Li₂O₂ seems to possess microscopic structural disorder that facilitates a delithiation induced partial solution‐phase oxidation at lower η _RC (<0.5 V). This difference in η _RC for differently grown Li₂O₂ provides crucial insights into necessary control over Li₂O₂ crystallization pathways to improve the energy efficiency of a Li–O₂ battery.

Herein, a fundamental correlation between the lithium peroxide crystallization pathway and subsequent recharge mechanism is established by combination of theory and experiment. The lithium peroxide grown by concurrent disproportionation and surface reduction is found to form two coexisting phases. Quantitative deconvolution shows that the lithium peroxide grown on electrode surface decomposes at low overpotential by a partial solution phase oxidation.

1. Introduction

Nonaqueous lithium–oxygen (Li–O₂) battery, with a theoretical specific energy of ≈3.5 kWh kg⁻¹, is considered to be one of the most promising energy storage technologies to cope with the ever‐increasing energy demand of the modern society for both mobile and stationary applications. The overall reversible electrochemical reaction in a nonaqueous Li–O₂ battery involves a two‐electron (2e⁻) redox of oxygen (O₂) in the presence of lithium ions (Li⁺) (2Li⁺ + O₂ + 2e⁻ ↔ Li₂O₂(s); E ⁰ = 2.96 V vs Li/Li⁺) producing solid lithium peroxide (Li₂O₂) as the discharge (DC, oxygen reduction reaction (ORR)) product, which subsequently decomposes in the reverse reaction during recharge (RC, oxygen evolution reaction (OER)).^{[ 1 ]} Despite the promise of Li–O₂ battery as a next generation storage system, low round‐trip energy efficiency (ratio of energy densities during DC and RC respectively) due to high RC overpotential (η, difference between thermodynamic (E ⁰) and measured (E) potential (η = |E − E ⁰|); usually η _RC > 1 V) and as a whole poorly understood RC mechanism pose a significant challenge before its realization.^{[ 2 ]} Although several electrocatalysts and soluble redox‐mediators were used to reduce the η _RC, they could not truly improve the reversibility of oxygen‐electrochemistry, rather the solid catalysts exacerbated parasitic reactions and shuttling of redox‐mediators on the unprotected negative electrode led to passivation of Li metal.^{[ 3 ]} Moreover, use of an electrocatalyst seems redundant due to low kinetic overpotential for OER in Li–O₂ battery as reported by density functional theory (DFT) calculations.^{[ 4 ]} Consistent with these theoretical studies, several experimental reports also found that the onset potential for Li₂O₂ oxidation is indeed very close to the equilibrium potential.^{[ 5 ]} However, despite the low kinetic overpotential for OER, the η _RC gradually increases during the course of RC and a two‐stage RC profile consisting of a low‐potential (<4 V vs Li/Li⁺) sloping region and a high‐potential (>4 V vs Li/Li⁺) flat plateau is usually observed for catalyst/redox‐mediator free Li–O₂ batteries.^{[ 6 ]} The rise in η _RC has been qualitatively attributed to several interrelated factors such as physicochemical properties of the discharged Li₂O₂, stability of the electrode/electrolyte, presence of additives/impurities in the electrolyte etc.^{[ 6} , ^{7 ]} Unfortunately, the effects from these factors have not been decoupled yet and combining results from independent experiments is often difficult due to completely different experimental conditions. As a result, contradictory reports explaining decomposition pathway of Li₂O₂, number of steps involved, reaction intermediates, kinetics of OER and exact reaction interface remain specific to a particular system and provide inconsistent understanding about the overall RC mechanism of Li–O₂ battery.^{[ 4} , ⁶ , ⁷ , ^{8 ]} Therefore, in order for realizing practical applications of Li–O₂ battery by suppressing the increase of η _RC, it is essential to identify a general low energy decomposition pathway for Li₂O₂.

Herein, we combine several in situ and ex situ experimental measurements with theoretical modelling to systematically decouple the crucial factors affecting the RC of Li–O₂ battery and delineate a non‐catalytic low energy decomposition route for Li₂O₂. By detailed analysis we clearly demonstrate that RC potential of a Li–O₂ battery fundamentally depends on the Li₂O₂ crystallization pathway which is intrinsically related to the microscopic structural properties of the growing Li₂O₂ crystals during DC of the cell. It is known that nucleation and growth of Li₂O₂ occurs via either chemical disproportionation or electrochemical reduction of intermediate reaction species LiO₂.^{[ 9 ]} However, here we propose that there is no sharp partition between these two reaction pathways. Rather, depending on the thermodynamic feasibility, Li₂O₂ grows by concurrent electrochemical surface reduction and solution mediated chemical disproportionation reaction over a wide range of DC conditions. The Li₂O₂ grown by these two routes seems to form two coexisting discrete phases with subtle structural differences that result in distinctly different η _RC. With the help of theory and experimental results we have quantitatively deconvoluted and estimated the amount of surface and solution grown Li₂O₂ during DC of the cells at different conditions. Unlike previous reports, here we clearly demonstrate that, no matter what the DC conditions, electrode/electrolyte properties and bulk morphology of Li₂O₂ are, solution‐grown Li₂O₂ always shows higher η _RC (>1 V) compared to that (<0.5 V) of surface‐grown Li₂O₂. The difference in η _RC can be attributed to intrinsic structural differences in Li₂O₂ crystals grown by two different routes. The higher η _RC of solution grown Li₂O₂ can be attributed to relatively higher structural order in Li₂O₂ crystals grown by chemical disproportionation. In contrast, due to a possible cohesive interaction between the electrode surface and growing crystals, the surface‐grown Li₂O₂ seems to possess microscopic structural disorder and improved charge transport properties that facilitate a delithiation induced partial solution‐phase oxidation at relatively lower η _RC (<0.5 V). Our observations suggest that ensuring surface‐growth of Li₂O₂ by strategic modification of electrode and proper choice of electrolyte can simultaneously provide a large DC capacity with high RC efficiency to mitigate the capacity‐rechargeability trade‐off in Li–O₂ battery.

2. Results and Discussions

2.1. Discharge Rate Dependent Recharge Behavior of Li–O₂ Batteries

Coin type CR2032 cells were assembled using Ketjenblack (KB) as the positive electrode prepared by KB‐slurry (in an N‐methyl‐2‐pyrrolidone (NMP) solvent with 20 wt% LITHion (Ion Power) as the binder) pasted on Toray carbon paper (CP, TGP‐H‐30, 1.6 cm diameter), 1 m lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) in tetraethylene glycol dimethyl ether (TEGDME or tetraglyme; <20 ppm H₂O by Karl Fischer titration, unless otherwise mentioned) and lithium foil (Li, 1.6 cm diameter) as the negative electrode. TEGDME has been chosen as the representative solvent due to its acceptable stability in Li–O₂ battery, good compatibility with Li metal and intermediate value of the Gutmann donor number (16.6 kcal mol⁻¹) that is known to be a major solvent property influencing the ORR and OER mechanisms.^{[ 10 ]} The configuration of a typical Li–O₂ coin cell is schematically shown in Figure  1a and the detailed cell‐assembly procedure is provided in the Supporting information Experimental section.

Figure 1.

a) Schematic representation of a Li–O₂ coin cell with Ketjenblack (KB) as the positive electrode. b) Potential (E) versus capacity (Q) plot for KB electrode in LiTFSI/TEGDME electrolyte (<20 ppm H₂O). The cells are discharged at different current densities (j _DC) but recharged at the same current density (j _RC) of 100 mA g⁻¹. c) Anodic linear sweep voltammetric (LSV, 0.05 mV s⁻¹) curves after discharge at different j _DC up to the fixed capacity (Q _DC) of 1 mAh.

Figure 1b shows the first cycle of galvanostatic DC/RC of Li–O₂ cells with KB electrodes where DC of the cells is carried out at four different current densities (j _DC) 25, 50, 100, and 200 mA g⁻¹ up to a fixed capacity (Q) of 1 mAh. On the contrary, all the cells are recharged at 100 mA g⁻¹ with cut‐off conditions of 100% RC or 4.5 V versus Li/Li⁺. DC/RC currents for all the cells are normalized to the mass of KB (≈0.5±0.05 mg cm⁻²) in the electrode. During discharge, all the cells show a nearly constant‐potential plateau in the DC curve with gradual increase in average η _DC from 223 mV at 25 mA g⁻¹ to 316 mV at 200 mA g⁻¹. A magnified view of the DC curves in the range of 0.4–0.6 mAh is shown in Figure S1 in the Supporting Information. This increase in average η _DC with j _DC is characteristic of iR drop due to buildup of insulating solid Li₂O₂ (having ionic and electronic conductivities in the range of 10⁻¹⁹–10⁻²⁰ S cm⁻¹ at 25 °C) during DC of the cells.^{[ 4} , ^{11 ]} In contrast, despite recharged at the same current density (j _RC), all the cells show more pronounced difference in RC behavior. Qualitatively, all the RC curves consist of two stages; an initial sloping region with gradual increase in η _RC usually up to 20–40% state of charge (SOC) followed by a flat high‐voltage plateau above 4 V versus Li/Li⁺. From Figure 1b, it is apparent that the fraction of charge passed in the sloping region or in other words, the amount of Li₂O₂ decomposed at low η _RC (<1 V) depends on the DC current. Nonetheless, the potential in the plateau region remains quite similar. In addition to galvanostatic RC, we have also recharged the cells by anodic linear sweep voltammetry (LSV, sweep rate 0.05 mV s⁻¹) from 2.95 to 4.7 V versus Li/Li⁺ after galvanostatic DC at different j _DC up to a Q _DC of 1 mAh. The LSV curves over the full range of potential are provided in Figure S2 in the Supporting Information, whereas a magnified section from 2.95 to 4 V versus Li/Li⁺ is shown in Figure 1c. Two distinctly different oxidation peaks, one centered at ≈3.35 V versus Li/Li⁺ and the other at >4.2 V versus Li/Li⁺, are observed for all the cells. Consistent with the trend in sloping region of galvanostatic curves, the intensity of the LSV peak at ≈3.35 V versus Li/Li⁺ also increases with increase in j _DC. These observations clearly show that larger fraction of Li₂O₂ deposited at relatively higher j _DC can be decomposed at lower η _RC.

2.2. Morphological and Structural Analysis of Rate Dependent Discharge Products

A nanoscale continuum model for the growth of Li₂O₂ crystals in Li–O₂ battery predicted a transition from large toroids (at lower discharge rates) to thin‐film/nanoparticles (at higher discharge rates) at a critical j _DC implying that large toroids and thin‐film/nanoparticles cannot coexist in the DC product.^{[ 12 ]} Consistent with this hypothesis, the DC current dependent RC profile of Li–O₂ battery was previously attributed to morphological differences observed in Li₂O₂ deposited at different j _DC.^{[ 6b,c ]} These studies demonstrated that thin‐film/nanoparticle‐like Li₂O₂ formed at higher j _DC can be decomposed at lower η _RC sloping region, whereas large toroidal Li₂O₂ formed at lower j _DC gets decomposed at higher η _RC (>1 V) flat plateau.^{[ 6b,c ]} However, it is not understood why a two‐stage RC profile, which according to those reports should be indicative of a mixed morphology in deposited Li₂O₂, is observed over a wide range of j _DC. Moreover, a later study has evinced that a sufficiently dry glyme‐ether based electrolyte is not expected to show any j _DC dependent morphological differences in the discharged Li₂O₂, rather a film‐like Li₂O₂ is expected at all j _DC.^{[ 7c ]} As a result, sharp correlation between j _DC‐dependent Li₂O₂ morphology and subsequent RC profile seems ambiguous. These contradictions in reported data lead to inconsistency in understanding the DC/RC process and ask for careful reconsideration of all the qualitative and quantitative parameters that affect the operation of Li–O₂ battery. Besides, combining results from different independent experiments to provide a complete idea about DC/RC processes is often difficult due to high sensitivity of Li–O₂ electrochemistry towards variation in experimental conditions.^{[ 13 ]} Therefore, here we have carried out several in situ and ex situ analyses under the same experimental set up inside a super‐dry room (maximum dew point −60 °C) in order to maintain similar effects by external factors in all of our experiments. The scanning electron microscopy (SEM) images in Figure  2a–d show the morphology of the electrodes after DC at four different j _DC. Possibly due to low water content (<20 ppm H₂O by Karl Fischer titration) in our tetraglyme based electrolyte, no large particle/toroidal Li₂O₂ is observed at any current density from 25 to 200 mA g⁻¹. In fact, all the electrodes look indistinguishable in the SEM images after DC up to a fixed Q _DC of 1 mAh. A closer look at the DC products by transmission electron microscopy (TEM), shown in Figure 2e–h, indicates that at all j _DC the deposited Li₂O₂ is composed of small nano‐sized conglomerated particles (30–40 nm diameter) that appear to be a film at low magnification. Another important revelation from TEM images is that the deposition is not conformal, rather particles of Li₂O₂ are nonuniformly deposited as islands leaving uncovered carbon surfaces around the deposits (bare carbon is pointed by white arrows in Figure 2e–h). Considering very small measurement area of TEM imaging, we have provided TEM images at three different locations of the same electrodes, marked as (i) to (iii) in Figure 2e–h, to show extensive nonuniformity of Li₂O₂ deposition throughout the electrode. The SEM and TEM images of pristine KB samples are also shown in Figure S3a,b in the Supporting Information, respectively, for comparison with the discharged electrodes. However, despite nonuniform deposition, holistically the average thickness of the deposited Li₂O₂ film decreases with increase in j _DC. From these morphological analyses it is clear that the two‐stage RC profile and its dependence on j _DC in Li–O₂ battery cannot be simply explained by the bulk morphologies of Li₂O₂. Crystallinity of Li₂O₂ is another important factor that has been considered by several groups to explain j _DC dependent RC profile of Li–O₂ battery.^{[ 6b,c ]} Operando X‐ray diffraction (XRD) studies during RC observed a non‐linear decay of peak intensity for Li₂O₂ where in the beginning, the peak intensity decays very slowly followed by a rapid decay at high RC potential.^{[ 14 ]} The slow decay in XRD peak intensity gives an indirect evidence for decomposition of Li₂O₂ with smaller domain size (or nano‐sized/amorphous Li₂O₂) at low RC potential, whereas the sharp decay directly indicates disappearance of larger domains of Li₂O₂ (or toroidal/crystalline Li₂O₂) at flat high‐voltage plateau. This explanation of the two‐stage RC profile seems quite reasonable since higher electronic as well as ionic conductivities of amorphous Li₂O₂ make it feasible to decompose at low η _RC, whereas crystalline Li₂O₂ that behaves like an insulator decomposes at high voltage plateau.^{[ 5} , ^{15 ]} However, exact mechanism of simultaneous deposition of amorphous and crystalline Li₂O₂ at a particular DC condition is not understood. Moreover, j _DC dependent variation in Li₂O₂ domain size is also questionable in a sufficiently dry ether‐based electrolyte.^{[ 7c ]} Figure 2i and Figure S4 (Supporting Information) show the XRD patterns of the electrodes discharged at different current densities from 25 to 200 mA g⁻¹. Superposition of peaks for both (100) and (101) reflections of Li₂O₂ discharged at all current densities is quite evident from the XRD patterns. The average domain sizes of Li₂O₂ discharged at four different j _DC vary in the ranges 13.7–14.2 and 9.7–11.1 nm for (100) and (101) reflections respectively as displayed in Figure 2j. It is necessary to mention that no correlation between j _DC and domain size can be established from these results due to experimental uncertainty over the small differences in full width half maximum (FWHM) values for corresponding peaks of the four samples. However, an apparent difference in domain size (≈3 nm) between (100) and (101) reflections for all the samples indicates anisotropy in the crystal structure of deposited Li₂O₂.^{[ 14 ]} An important corollary of the XRD data shown in Figure 2i,j is that it is difficult to discriminate the crystallinity of Li₂O₂ deposited at different j _DC. Having said that, we do not disregard the effects of Li₂O₂ crystallinity on the RC profile of Li–O₂ battery, we rather emphasize the limitation of ex situ XRD analysis in identifying subtle differences in j _DC dependent structural features of electrochemically formed Li₂O₂.^{[ 14a ]} Instability of carbon electrode and nonaqueous liquid electrolyte towards reduced oxygen species is a serious problem in Li–O₂ battery leading to buildup of insulating solid parasitic products that increase the impedance of the cell.^{[ 16 ]} As the cell‐impedance rises, decomposition of Li₂O₂ should require high η _RC. In order to investigate any effect of parasitic side reactions on the j _DC dependent RC profile of Li–O₂ battery, we have quantified the amount of Li₂O₂ after DC at different j _DC up to a fixed Q _DC of 1 mAh by titrating the electrodes with titanium (IV) oxysulfate (TiOSO₄) solution. The yields of Li₂O₂ after DC have been shown in Figure 2k and are found to be 74 ± 1, 73 ± 1, 72 ± 2, and 71 ± 2% at j _DC 25, 50, 100, and 200 mA g⁻¹ respectively. Lower than theoretical yield of Li₂O₂ is ascribed to parasitic reactions as mentioned above and the Fourier transform infrared spectrum (FTIR) of the KB electrode discharged at 100 mA g⁻¹ in Figure S5 in the Supporting Information shows the existence of lithium carbonate (Li₂CO₃) and lithium carboxylates (RCO₂Li) as the parasitic products. It has to be mentioned here that the parasitic products may have detrimental effect in increasing the η _RC of the cells but quite similar yield of Li₂O₂ (hence similar extent of parasitic reactions) after DC at different j _DC should exhibit comparable effects on η _RC.

Figure 2.

a–d) Scanning electron microscopy (SEM) images of KB electrodes after discharge at different j _DC up to a Q _DC of 1 mAh in LiTFSI/TEGDME electrolyte (<20 ppm H₂O). e–h) Transmission electron microscopy (TEM) images and i) X‐ray diffraction (XRD) patterns of the same electrodes. All the XRD data are normalized to the intensity of KB+CP peak at 2θ = 54.3. j) Plot of domain size versus j _DC for both (100) and (101) reflections of Li₂O₂. Domain size is calculated by using Scherrer equation. k) Plot of Li₂O₂ yield measured by titration with TiOSO₄ solution after DC at different j _DC. TEM images marked as (i–iii) represent three different locations of the same electrode at respective j _DC. The white arrows indicate parts of carbon surface not covered with Li₂O₂. The Yellow dotted lines in (e–h) demarcate the Li₂O₂/carbon interface. XRD peak positions for standard Li₂O₂ sample are shown by black vertical lines at the bottom of (i). The vertical lines in (k) represent the error range of measured data from three sets of experiment.

The structural, morphological and quantitative analyses demonstrated above could only point out a j _DC dependent difference in average thickness of the deposited Li₂O₂. A possible correlation between η _RC and thickness of Li₂O₂ deposit was reported earlier for a flat electrode and it was argued that relatively thinner parts of Li₂O₂ film could be decomposed at low η _RC.^{[ 5b ]} However, interestingly, the positions of anodic LSV peaks (in the ranges of <3.5 and >4 V vs Li/Li⁺) of a Li₂O₂ film of <3 nm thickness grown on a flat electrode resemble the positions of the LSV peaks for the oxidation of several tens of nanometer thick Li₂O₂ deposited on KB electrode as well as on other high surface area carbons reported previously.^{[ 5} , ^{17 ]} Therefore, exact correlation between RC profile and Li₂O₂ thickness seems inconclusive. Besides, the stochasticity of Li₂O₂ layer thickness due to nonuniform deposition also makes it difficult to establish a quantitative relationship between thickness and RC potential. Nonetheless, based on the observation of nonuniform deposition, we have modelled the nucleation‐and‐growth of Li₂O₂ in our system in order to provide a quantitative insight about the relation between j _DC dependent Li₂O₂ deposition and subsequent RC profile.

2.3. Theoretical Modelling of Li₂O₂ Growth During Discharge

The DC of Li–O₂ battery is known to involve multiple steps of ORR in nonaqueous conditions^{[ 8c ]}

$$\begin{array}{c}\hfill {{\mathrm{O}}_2}^{\ast }+\mathrm{L}{\mathrm{i}}^{+}+{\mathrm{e}}^{-}\to \mathrm{Li}{{\mathrm{O}}_2}^{\ast }\end{array}$$ (1)

$$\begin{array}{c}\hfill \mathrm{Li}{{\mathrm{O}}_2}^{\ast }\leftrightarrow \mathrm{Li}{\mathrm{O}}_2\left(\mathrm{sol}.\right)\end{array}$$ (2)

$$\begin{array}{c}\hfill \mathrm{Li}{{\mathrm{O}}_2}^{\ast }+\mathrm{L}{\mathrm{i}}^{+}+{\mathrm{e}}^{-}\to \mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2\left(\mathrm{s}\right)\end{array}$$ (3)

$$\begin{array}{c}\hfill 2\mathrm{Li}{\mathrm{O}}_2\left(\mathrm{sol}.\right)\to \mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2\left(\mathrm{s}\right)+{\mathrm{O}}_2\end{array}$$ (4)

(Here the asterisk (*) represents the species adsorbed on the electrode surface)

Based on these reaction steps we propose a nucleation‐and‐growth mechanism that agrees with the nonuniform deposition of Li₂O₂ as observed in the TEM images. In our model, reaction (1), i.e., formation of surface bound LiO₂* is regarded as the first step of ORR. This ORR step is followed by concurrent nucleation (reaction (3)) and growth (reaction (4)) of Li₂O₂ on the KB surface. Negative free energies for both surface reduction and solution phase disproportionation indicate that simultaneous occurrence of these two reactions is thermodynamically feasible.^{[ 6c ]} Here we specify that nucleation refers to surface‐growth that accounts for the coverage of Li₂O₂ layer on the electrode surface, whereas growth means precipitation of Li₂O₂ from solution phase on top of the nucleation sites. Li₂O₂ layer with thickness >10 nm (maximum electron tunneling distance through Li₂O₂ layer) observed in TEM images also indicates involvement of solution mediated vertical growth of Li₂O₂ in our cells.^{[ 18 ]} The solubility, if any, of intermediate LiO₂ in sufficiently dry LiTFSI/TEGDME electrolyte (<20 ppm H₂O) has been verified by cathodic LSV using rotating ring disk electrode (RRDE, rotation speed of 900 rpm) under 3‐electrode set‐up. From the LSV data shown in Figure S6 in the Supporting Information, a pronounced reduction current is observed at the KB disk below 2.76 V versus Li/Li⁺ and simultaneously an oxidizing current is observed at the GC ring held at 3.5 V versus Li/Li⁺. The reduction current corresponds to ORR at the KB disk while the ring current originates from the oxidation of solvated LiO₂ (sol.) that diffuses from the disk to the ring due to the rotation of the electrode. The emergence of ring current confirms the solubility, at least partially, of LiO₂ in dry LiTFSI/TEGDME electrolyte (<20 ppm H₂O) and the results are consistent with previously reported studies.^{[ 9 ]} Higher binding energy of lithium as well as of oxygen leading to preferential nucleation at randomly distributed defect sites results in nonuniform nucleation of Li₂O₂ on carbon surface. Density functional theory (DFT) calculations showed that the adsorption energies at the defect sites can be higher by ≈1.5 and ≈4.4 eV for lithium and oxygen respectively compared to a perfect defectless site.^{[ 19 ]} KB indeed manifests considerable defects and consists of oxygen functional groups that can act as the adsorption sites for both oxygen and lithium. The Raman spectrum of KB in Figure S7 in the Supporting Information shows D‐band (I _D, representative of disorder) to G‐band (I _G, graphitic structure) intensity ratio (I _D/I _G) of 1.06 that certainly confirms presence of defects in the carbon structure.^{[ 17 ]} X‐ray photoelectron spectroscopic (XPS) analysis of the KB surface estimates oxygen to carbon ratio (O/C) of 0.027 where oxygen functionalities are identified to be C—O (e.g., phenol, ether, epoxy etc. at C1s ≈ 286.7 eV and O1s ≈ 533.5 eV), C=O (e.g., carbonyl, quinone etc. at C1s 287.5 eV and O1s 531.8 eV) and —COO⁻ (e.g., carboxylic acid, ester, lactone, etc. at C1s 289.3 eV) groups that are shown by deconvoluted XPS data in Figure S8 in the Supporting Information and the physicochemical properties of KB are summarized in Table S1 in the Supporting Information.^{[ 20 ]} According to our proposed mechanism, nucleation of Li₂O₂ starts preferentially at the defect sites of KB electrode and concurrent nucleation‐and‐growth occurs progressively until complete coverage of Li₂O₂ layer on the electrode surface. A schematic representation of the proposed DC mechanism is shown in Figure  3 .

Figure 3.

Schematic representation of a nonuniform nucleation‐and‐growth model presented in this work.

In order to model the nucleation‐and‐growth mechanism described above, we have considered the kinetic equations describing the rates of reactions (1) (r ₁, mol m⁻² s⁻¹) and (3) (r ₂, mol m⁻² s⁻¹)

$$\begin{array}{c}\hfill r_1=k_1{c}_{O_2}\exp \left(-\frac{{\alpha }_{1}F}{RT}{\eta }_1\right)\end{array}$$ (5)

$$\begin{array}{c}\hfill r_2=k_2{c}_{Li{O}_2}\exp \left(-\frac{{\alpha }_{2}F}{RT}{\eta }_2\right)\end{array}$$ (6)

where k ₁ (m s⁻¹) and k ₂ (s⁻¹) are the rate constants for reduction of oxygen (O₂) and surface adsorbed lithium superoxide (LiO₂*) respectively, $c_{{\mathrm{O}}_2}$ (mol m⁻³) and $c_{\mathrm{Li}{\mathrm{O}}_2}$ (mol m⁻²) are the concentrations of oxygen in the electrolyte and surface concentration of LiO₂* on the electrode respectively, α ₁ and α ₂ are the charge transfer symmetry factors approximated to be 0.5, η ₁ and η ₂ are the overpotentials for the respective steps assumed to be η ₁ ≈ η ₂ = η _DC. F (96 485 C mol⁻¹), R (8.314 J mol⁻¹ K⁻¹), and T (298 K) have their usual meaning. The solvation kinetics of LiO₂* is represented by

$$\begin{array}{c}\hfill r_{\mathrm{s}}=k_{\mathrm{s}}{c}_{\mathrm{Li}{\mathrm{O}}_2}\end{array}$$ (7)

where r _s (mol m⁻² s⁻¹) is the solvation rate of LiO₂* and k _s (s⁻¹) is the solvation rate constant. These equations show that unlike nucleation rate (r ₂), which depends on the DC overpotential (η _DC), the growth rate depends on the rate of LiO₂* solvation (r _s) and its supersaturation in the electrolyte. For simplicity, here we consider same rates of reactions at the superficial surface and also inside the bulk of the porous electrode. We further assume that since nucleation occurs by surface reduction, the nucleating layer on the electrode surface can have a maximum average thickness (h ₀) limited by electron tunneling length and based on this assumption the areal growth rate of Li₂O₂ can be described as

$$\begin{array}{ccc}\hfill \frac{d{A}_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}}{dt}=A_{\mathrm{electrode}}& \times & \frac{M_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}}{{\rho }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}\times h_0}\times c_{O_2}\hfill \\ & & \times \frac{k_1\times \exp \left(-\frac{2\alpha F}{RT}{\eta }_{DC}\right)}{\exp \left(-\frac{\alpha F}{RT}{\eta }_{DC}\right)+\frac{k_s}{k_2}}\hfill \end{array}$$ (8)

where A _electrode is the electrochemically active surface area (ECSA, 0.028 m²) of the KB electrode (measured by cyclic voltammetry (CV, Figure S9, Supporting Information) of a symmetric KB|KB cell and the calculation is shown in the supporting information), $M_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ and ${\rho }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ are the molar mass (45.88 g mol⁻¹) and bulk density (2.31 × 10⁶ g m⁻³) of Li₂O₂ respectively and as previously mentioned, h ₀ is the maximum average thickness of the nucleating layer. Full derivation of Equation (8) is provided in the supporting information. According to this equation, sudden‐death during DC (or full saturation of electrode surface) occurs at a time (t _s) when $A_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}=A_{\mathrm{electrode}}$. Figure S10 in the Supporting Information shows DC of the cells at four j _DC from 25 to 200 mA g⁻¹ up to the cut‐off potential of 2 V versus Li/Li⁺. A sharp decay in DC potential up to 2 V versus Li/Li⁺ (sudden‐death) is observed at the end of DC for all the cells in Figure S10 in the Supporting Information. This sudden‐death denotes the complete coverage of the electrode surface by insulating Li₂O₂ as the DC product. Had the sudden death been due to charge transport limitation through a conformal layer of Li₂O₂, as observed in flat electrodes, the Q _DC would be inversely proportional to j _DC.^{[ 21 ]} Figure S11 in the Supporting Information shows a plot of Q _DC versus j _DC for the KB electrodes where the inverse relationship, shown by the gray solid line, significantly deviates from the experimental data points.

2.4. Quantitative Correlation between Li₂O₂ Growth Mechanism and Recharge Profile

We have plotted the t _s versus η _DC for the cells discharged at j _DC from 25 to 200 mA g⁻¹ in Figure  4a. Values of η _DC for different j _DC are obtained by subtracting the equilibrium potential (E ⁰) of 2.96 V versus Li/Li⁺ from the DC potentials (E) in Figure 1b at 0.5% state of DC. Although the maximum tunneling length for a compact Li₂O₂ layer is reported to be ≈10 nm, presence of defects and grain boundaries can increase the tunneling length.^{[ 11} , ^{18 ]} Therefore, to fit the experiemntal data in Figure 4a by using Equation (8) and for subsequent calculations we have considered different h ₀ values in the range 10–20 nm. A representative fitting of the experimental data points considering h ₀ value of 15 nm in Equation (8) is shown by the solid gray line in Figure 4a. The numerical values for all the fitting parameters used in Figure 4a are provided in Table  1 . Using the values of the fitting parameters in the t _s versus η _DC plots, next, we have computed the area covered by the Li₂O₂ layer discharged at different j _DC up to a limited Q _DC of 1 mAh. In order to correct the effect of overlapping Li₂O₂ particles, Kolmogorov's phase transformation theory is applied to calculate the fractional area coverage of the electrode by Li₂O₂ layer at a given time frame

$$\begin{array}{c}\hfill {\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}=1-\exp \left(-\frac{A_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}}{A_{\mathrm{electrode}}}\right)\end{array}$$ (9)

Figure 4.

a) Plot of the time (t _s) required for galvanostatic full discharge (up to 2 V vs Li/Li⁺) of a cell versus DC overpotential (η _DC) using KB as the electrode and LiTFSI/TEGDME as the electrolyte (<20 ppm H₂O) and fitting of this experimental data based on the nucleation and growth model presented in this work. b) Plot of the amount of charge (Q _RC) passed during RC below 4 V versus Li/Li⁺ and the calculated charge (Q _DC/SL) stored as Li₂O₂ layer with 10, 12, 15, 18, and 20 nm thickness grown by surface reduction against the respective current densities. For all the cases in (b) a Q _DC of 1 mAh is considered.

Table 1.

Physical constants and fitting parameters

$M_{\mathbf{L}{\mathbf{i}}_2{\mathbf{O}}_2}$ [g mol−1]	${\rho }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ [g m−3]	h 0 [m]	$c_{{\mathrm{O}}_2}$ [mol m−3]	α	F [C mol−1]	R [J K−1 mol−1]	T [K]	k 1 [m s−1]	k s/k 2
45.88	2.31 × 106	15 × 10−9	4.43[ 9 ]	0.5	96 485	8.314	298	10−10	2.7 × 103

${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$is the fraction of KB electrode covered by the Li₂O₂ layer.^{[ 22 ]} A plot of ${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ versus j _DC for Q _DC of 1 mAh and h ₀ value of 15 nm in Figure S12 in the Supporting Information shows gradual increase of area covered by Li₂O₂ film as the DC rate increases. This observation justifies the decrease in average Li₂O₂ film thickness with increase in j _DC as observed by TEM images in Figure 2e–h. However, we assume that, regardless of j _DC, the surface grown Li₂O₂ layer has a fixed thickness determined by the maximum electron tunneling length through this layer. Therefore, the amount of Li₂O₂ deposited by surface reduction at different j _DC can be scaled with ${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}.$The actual area of the electrode covered by Li₂O₂ ($A_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}^{\prime })$is calculated by multiplying ECSA of KB with ${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}.$We have estimated the amount of charge (Q _DC/SL) stored in the surface grown layer in terms of deposited Li₂O₂ from this $A_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}^{\prime }$ and different h ₀ values at each j _DC from 25 to 200 mA g⁻¹ for a capacity of 1 mAh. Expectedly, Q _DC/SL increases with j _DC as shown in Figure 4b. We have also plotted the amount of charge passed (Q _RC < 4 V) at low RC potential (<4 V vs Li/Li⁺) during RC (both galvanostatic RC and anodic LSV) of the cells discharged up to 1 mAh at respective j _DC. It is interesting to see that Q _DC/SL for a layer of 15 nm and Q _RC < 4 V measured by both galvanostatic RC and anodic LSV are quantitatively comparable. That means the 15 nm thick Li₂O₂ layer grown by surface reduction decomposes at potential <4 V versus Li/Li⁺. Conversely, based on this argument, any Li₂O₂ deposited by chemical disproportionation decomposes at higher potential (>4 V vs Li/Li⁺) flat plateau of the RC curve. This observation shows an unprecedented deconvolution of RC potentials of Li–O₂ battery based on mechanism of Li₂O₂ nucleation‐and‐growth and provides a quantitative evidence that electrochemically deposited Li₂O₂ on the surface of the electrode decomposes at low η _RC. It is important to note that this rationale does not specify the effect of any particular morphology of Li₂O₂ in explaining the RC curve of Li–O₂ battery. Rather it just deconvolutes the effects of two concurrent Li₂O₂‐growth mechanisms on η _RC. However, this general explanation can also be extended to justify the apparent effects of Li₂O₂ morphology on RC profile of Li–O₂ battery. The change in morphology of Li₂O₂ from large toroids to conformal thin‐film is actually the shift in reaction mechanism from predominant solution phase disproportionation towards electrochemical reduction on the electrode surface. As the morphology changes from toroids to thin‐film, the fraction of surface‐grown Li₂O₂ increases and consequently more Li₂O₂ is decomposed at relatively lower η _RC. Therefore, the apparent effect of Li₂O₂ morphology on RC potential is actually a corollary of fundamental growth mechanism described above. Although the visible morphology of Li₂O₂ was reported to change sharply at a particular DC condition, here we reiterate that depending on thermodynamic feasibility, the co‐occurrence of the two underlying growth processes is possible over a wide range of DC conditions and the equilibrium shifts quite steadily with small changes in external factors.^{[ 6} , ^{12 ]} As a result, DC condition dependent changes in the fractions of sloping region and the high voltage plateau of a two‐stage RC profile can be simply correlated to the change in equilibrium between the Li₂O₂ growth mechanisms regardless of the bulk morphology. According to the nonuniform nucleation‐and‐growth mechanism explained in Figure 3, increase in depth of discharge (DOD) corresponds to progressive growth of Li₂O₂ in the lateral direction maintaining a fixed ratio of the two concurrent growth processes at a particular DC condition. Therefore, the ratio of the sloping region and flat plateau should not change with the DOD if the DC conditions remain unchanged. To verify this hypothesis, we have discharged and recharged three more cells up to Q _DC of 1.5, 2, and 3 mAh at j _DC of 100 mA g⁻¹. The galvanostatic DC/RC curves shown in Figure S13a in the Supporting Information demonstrate similar two‐stage RC profile for all the cells. Quite interestingly, all the RC curves have overlapped upon normalization of Q _DC in Figure S13b in the Supporting Information. XRD data of the discharged electrodes shown in Figure S14a in the Supporting Information demonstrates gradual increase in Li₂O₂ peak intensity that signifies the addition of more Li₂O₂ with increase in DOD. However, no significant growth of individual grains in Li₂O₂ crystals is observed as shown in Figure S14b in the Supporting Information. This means that the fundamental crystallization mechanism of Li₂O₂ does not change with DOD and maintains same recharge efficiency. Summarizing all these results it is understood that promoting reaction (3) (surface reduction of LiO₂*) over reaction (4) (solution phase disproportionation of LiO₂ (sol.)) during DC can contribute to low η _RC of the cell. The competition between these two reactions represented by k _s/k ₂ ratio depends on η _DC (or j _DC) and properties of the electrode as well as of the electrolyte.^{[ 6} , ¹⁰ , ^{17 ]} We have already explained the effect of η _DC (or j _DC) dependent growth process of Li₂O₂ on η _RC. Next, in order to prove generality of our hypothesis, we have further modified KB electrode as well as the electrolyte to manipulate the k _s/k ₂ ratio in our system. It is known that defects and surface functional groups on carbon electrode promote surface‐growth of Li₂O₂, whereas controlled amount of water (H₂O) as additive in the electrolyte increases the solubility of LiO₂ (sol.) and enhances the rate of disproportionation.^{[ 7} , ^{17 ]} We have oxidized KB, hereafter denoted as KB_ox, in H₂SO₄ + HNO₃ acid mixture to introduce defects and surface oxygen functional groups that are expected to promote surface‐reduction of LiO₂*. On the other hand, to decrease the propensity for surface reduction, KB is reduced, denoted as KB_r, in Ar+H₂ (4% H₂) gas mixture at 1000 °C to remove oxygen functional groups. Raman spectra and XPS results of KB_ox and KB_r are shown in Figures S7 and S8 in the Supporting Information respectively. The I _D/I _G and O/C ratios for both the samples are summarized in Table S1 in the Supporting Information. In short, these results clearly indicate significant enhancement of oxygen functionality (O/C = 0.098) and disorder (I _D/I _G = 1.54) in KB_ox and on the contrary, after reduction at high temperature the structure has become more ordered (I _D/I _G = 0.71) and oxygen content is much decreased (O/C = 0.003) in KB_r. KB, KB_ox, and KB_r are discharged and recharged at 100 mA g⁻¹ up to 1 mAh and the DC/RC curves are shown in Figure  5a. Unsurprisingly, as the oxygen functionality/defects increases in the electrode active material in order of KB_r < KB < KB_ox, the sloping region becomes more extended with lower η _RC. The span of sloping region extends from ≈25% SOC in KB_r to ≈33% SOC in KB to ≈60% SOC in KB_ox. The extension of sloping region can certainly be corroborated to enhanced surface reduction of LiO₂* with increase in surface‐functionality of the electrode. Morphologies of the as prepared and discharged KB_ox and KB_r electrodes are shown in Figure S15a–d in the Supporting Information. Figure S15b in the Supporting Information shows predominant film‐like deposition of Li₂O₂ on the superficial surface of KB_ox electrode. On the contrary, disk‐like Li₂O₂ with roughly 100 nm diameter and 20 nm thickness, as shown in Figure S15d in the Supporting Information, are formed on KB_r electrode. The large sized disk‐like particles give more evidence in favor of a dominant solution phase growth of Li₂O₂ in KB_r electrode. To further enhance the solution phase disproportionation, KB electrodes have been discharged and recharged in LiTFSI/TEGDME electrolyte with different concentrations of H₂O added as solvating agent. Toroidal DC products in the H₂O containing electrolytes, as shown in Figure S15e,f in the Supporting Information, certainly indicate enhanced solution mediated discharge upon addition of H₂O. Figure 5b shows that the low‐potential sloping region becomes shorter with increase in concentration of H₂O. In particular the length of the sloping region decreases from ≈33% SOC in dry electrolyte (<20 ppm H₂O) to ≈27% SOC in 1000 ppm to ≈18% SOC in 5000 ppm H₂O containing electrolytes. To provide a quantitative validation of our hypothesis that solution phase disproportionation decreases the surface coverage of Li₂O₂, we have calculated the dependence of ${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ on η _DC (or j _DC) and k _s/k ₂ ratio for a fixed DC time. In our calculation we have considered commonly observed η _DC values of 0–0.5 V and k _s/k ₂ values 0–10⁵ at a DC time of 10 h. Here k _s/k ₂ = 0 implies exclusive surface reduction of LiO₂*, whereas k _s/k ₂ = 10⁵ means greater contribution from solution phase disproportionation of LiO₂ (sol.). The contour plot in Figure  6a shows that at a particular η _DC, ${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ increases with decrease in k _s/k ₂ value. On the other hand, increase in η _DC that enhances the rate of LiO₂* reduction (reaction (3)) also increases ${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ at a fixed value of k _s/k ₂. This clearly demonstrates that with the enhancement of LiO₂* reduction, more fraction of electrode surface gets covered with Li₂O₂. Although increase in ${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ decreases the η _RC as explained above, at the same time the electrode surface becomes passivated with insulating layer of Li₂O₂ that causes early death of DC resulting in capacity‐rechargeability trade‐off in Li–O₂ battery. In Figure 6b we have plotted the sudden‐death time (t _s) at different η _DC and k _s/k ₂ values. The results unequivocally demonstrate that dominant surface reduction of LiO₂* leads to early death and hence manifests low Q _DC.

Figure 5.

a) Potential (E) versus capacity (Q) plot of pristine and modified KB electrodes in LiTFSI/TEGDME electrolyte (<20 ppm H₂O). b) Same plot of pristine KB electrodes in LiTFSI/TEGDME electrolyte with different water (H₂O) content.

Figure 6.

a) Contour plot showing the variation of the fraction of total electrode area covered by Li₂O₂ (${\theta }_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2})$ grown by surface reduction with η _DC and the ratio of the kinetic rate constants for solution mediated disproportionation (k _s) and surface reduction (k ₂) respectively for a fixed DC time of 10 h. b) Contour plot showing variation of t _s with η _DC and k _s/k ₂ ratio.

2.5. Establishing Mechanism of Low‐Potential Recharge Pathway

In order for developing a more efficient Li–O₂ battery, it is necessary to understand why and how the surface‐grown Li₂O₂ decomposes at relatively lower η _RC. Therefore, we now focus on understanding and establishing the RC mechanisms that correlate to the decompositions of surface and solution‐grown Li₂O₂ at different RC potentials. The apparent sharp transition from a sloping trend to a flat plateau in the RC curve (Figure 1b) and two markedly different oxidation peaks (≈3.35 and >4.2 V vs Li/Li⁺) in anodic LSV (Figure 1c; Figure S2, Supporting Information) certainly indicate a switch in decomposition mechanism for these two types of Li₂O₂. Since RC of Li–O₂ battery involves oxidation of ostensibly insulating solid Li₂O₂, η _RC seems to be sensitive towards subtle structural differences that affect charge transport through the solid Li₂O₂ layer.^{[ 15 ]} As it has been mentioned earlier in this article, due to higher adsorption energies of lithium and oxygen, Li₂O₂ starts nucleating preferentially at the defect sites on carbon surface.^{[ 19} , ^{23 ]} According to prior DFT calculations, the strong binding interactions between nucleating sites on the carbon surface and Li₂O₂ lead to electron transfer from the anti‐bonding O₂ orbitals to the carbon electrode inducing a p‐type conductivity at Li₂O₂/electrode interface.^{[ 24 ]} Furthermore, due to this strong binding, reorganization of bonding electrons often results in elongation of Li–O bond length that diminishes the interaction between lithium atom and oxygen atom in Li₂O₂.^{[ 25 ]} As a result, mobility of Li⁺ in Li₂O₂ crystal becomes easier in the layer that interacts strongly with carbon electrode surface. These defect‐induced improvements in both ionic and electronic conductivities of surface grown Li₂O₂ perhaps make the decomposition possible at low η _RC. In contrast, the Li₂O₂ precipitated from solution phase is believed to be structurally more ordered, stoichiometric and electronically insulating that requires high η _RC for decomposition.^{[ 7} , ^{15 ]} The higher structural order of solution grown Li₂O₂ is evinced by the XRD patterns shown in Figure S16a,b in the Supporting Information. The XRD data in Figure S16a in the Supporting Information show that the intensity of peaks and the average domain size of Li₂O₂ crystals increases as the propensity towards solution growth increases on different carbons in the order KB_ox < KB < KB_r. Consistent with this trend, Figure S16b in the Supporting Information shows, increasing H₂O content in the electrolyte also increases both the XRD peak intensity and the domain size in Li₂O₂ crystals. Figure  7a–d shows galvanostatic RC curve in combination with quantitative estimation of OER to gain clear understanding of the RC process. The online electrochemical mass spectrometry (OEMS) data in Figure 7b show a plot of gas evolution rate (r _g) with Q _RC. A high rate ($r_{\mathrm{g}/{\mathrm{O}}_2}$) of oxygen evolution at the beginning of RC confirms the decomposition of Li₂O₂ below 3.5 V versus Li/Li⁺. However, the $r_{\mathrm{g}/{\mathrm{O}}_2}$ value is lower than that calculated for a 2e⁻/O₂ process and more surprisingly, although the RC potential gradually increases, $r_{\mathrm{g}/{\mathrm{O}}_2}$ shows a decreasing trend up to ≈15% SOC that corresponds to ≈3.5 V versus Li/Li⁺. This means either the rate of Li₂O₂ decomposition ($r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2})$ is lower than expected with a similar decreasing trend or the evolved oxygen somehow remains trapped inside the electrode during the course of RC. In order to quantify $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$, we have titrated the KB electrode at different SOCs by TiOSO₄ solution and the results are shown in Figure 7c. The titration data show that the $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ values are quite high, roughly close to the theoretical value of 2.38 × 10⁻⁸ g s⁻¹, at the beginning of RC and remain significantly higher than $r_{\mathrm{g}/{\mathrm{O}}_2}$ up to ≈30% SOC. These results evince that the decomposition of Li₂O₂ is considerably facile at low η _RC sloping region. The difference between $r_{\mathrm{g}/{\mathrm{O}}_2}$ and $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ can be attributed to generation of highly reactive singlet oxygen (¹O₂) that reacts with electrode as well as with electrolyte and does not evolve to be detected by OEMS despite continuous decomposition of Li₂O₂ during RC.^{[ 26 ]} These oxidations of electrode and electrolyte by ¹O₂ produce parasitic side products in the form of Li₂CO₃, RCO₂Li, etc.^{[ 16a ]} We have also measured the change in mass of the whole cell in situ during DC/RC. Ideally, any change in the mass of a cell should account for the consumption and evolution of oxygen during DC and RC respectively. We have estimated a ≈1.97 e⁻/O₂ ORR from the mass of consumed O₂ during DC of the cell and the trend of O₂ consumption is shown in Figure S17 in the Supporting Information. On the other side Figure 7d shows the rate of mass change (r _w) of the cell during RC. Generally, the value of r _w should account for the decrease in mass due to oxygen evolution from decomposition of Li₂O₂. Interestingly, despite high value of $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ the graph in Figure 7d shows a decreasing trend of r _w up to ≈30% SOC and follows the trend of oxygen evolution. This observation clearly indicates entrapment of oxygen and accumulation of insulating side products that balance the decrease in mass due to Li₂O₂ decomposition. Nevertheless, the non‐catalytic decomposition of Li₂O₂ below 3.5 V versus Li/Li⁺ provides optimism about achieving Li–O₂ batteries with high round‐trip energy efficiency. Considering the high energy barrier for a direct 2e⁻ oxidation, stepwise 1e⁻ process with low delithiation barrier seems to be a preferred oxidation pathway for Li₂O₂.^{[ 8c ]} Advantageously, defect induced structural disordering of surface grown Li₂O₂ lowers both the formation and migration energies of Li⁺ vacancy leading to facile delithiation and formation of non‐stoichiometric Li_{2− x}O₂ (0 < x < 1) phase.^{[ 27 ]} Despite the energetically favorable delithiation, according to first‐principles calculations, oxygen evolution from the solid Li_{2− x}O₂ species remains as the highest energy‐demanding step during Li₂O₂ decomposition.^{[ 28 ]} Alternatively, a concerted delithiation of Li₂O₂ clusters leading to disintegration and dissolution of the non‐stoichiometric Li_{2− x}O₂ at a critical value of x may instead provide a solution‐phase oxidation pathway on the bare electrode surface at relatively low η _RC in the range 0.3–0.4 V.^{[ 27 ]} An anodic LSV has been carried out on a pre‐discharged KB electrode using RRDE set up similar to that described for Figure S6 in the Supporting Information. The current responses of the KB disk swept from 2.9 to 4.5 V versus Li/Li⁺ and the GC ring held at 3.5 V versus Li/Li⁺ are shown in Figure 7e. It is indeed found that the oxidation peak for the surface grown Li₂O₂ at ≈3.4 V versus Li/Li⁺ is associated with the occurrence of ring current at the GC ring electrode. This observed ring current directly provides evidence in favor of a partial solution mediated oxidation of the Li₂O₂ that is deposited by surface reduction. The oxidation in the solution phase is likely to be energetically favorable over the solid phase oxidation due to circumvention of the charge transport limitation of insulating solid Li₂O₂. Interestingly, the rationale for decomposition of surface‐grown Li₂O₂ at low η _RC can be extended to explain the apparent electro‐catalytic effect in Li–O₂ battery. Any catalyst that shows strong binding towards Li⁺ and O₂ is found to promote surface‐growth of Li₂O₂ and subsequently exhibits low η _RC.^{[ 25} , ^{29 ]} Therefore, regardless of the electrode type, low η _RC is fundamentally linked to the Li₂O₂ growth on the surface of the electrode during DC. Now the question is why the potential gradually increases and forms a slope before reaching a flat plateau. One possible reason is the structural nonuniformity in the surface‐grown Li₂O₂ layers that may show a range of RC potential. According to DFT calculations, the interaction between carbon and the Li₂O₂ layers changes as the distance from the carbon surface increases.^{[ 24 ]} As a result, the degrees of disorder in the surface‐grown Li₂O₂ is also expected to change and may exhibit gradual change in RC potential. Another obvious reason for increase in RC potential is the accumulation of parasitic side products that increase the impedance of the cell.^{[ 16 ]} After complete decomposition of surface grown Li₂O₂ when the RC potential reaches the plateau and η _RC becomes high (>1 V), the solution grown structurally more ordered and stoichiometric Li₂O₂ starts decomposing. Based on first‐principles calculations, the insulating bulk Li₂O₂ becomes conductive only when the η _RC is high enough to activate the delithiation process and the charge transport arises from the mixture of Li vacancies and hole polaron's contributions.^{[ 11 ]} Therefore, oxidation of Li₂O₂ in this case also presumably starts with delithiation and progresses via non‐stoichiometric phases. In fact, decomposition of Li₂O₂ by forming Li vacancy in the crystal was indeed observed by an in situ XRD study during RC of a Li–O₂ battery.^{[ 14a ]} However, unlike decomposition of surface grown Li₂O₂, no solution mediated oxidation has been observed for the Li₂O₂ grown by disproportionation. High enough η _RC to permit successive delithiation and oxygen evolution from the Li₂O₂ crystal before it disintegrates and dissolves in the electrolyte may be the reason why no solution mediated oxidation is observed in this case. It is interesting that although the galvanostatic RC curve reaches a flat‐potential (≈4.2 V vs Li/Li⁺) plateau at ≈30% SOC and roughly maintains the potential up to ≈80% SOC, the values of $r_{\mathrm{g}/{\mathrm{O}}_2}$ and $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ are not consistent with this Coulometric data. Both $r_{\mathrm{g}/{\mathrm{O}}_2}$ and $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ are quite low at the beginning of the plateau and steadily increases up to 60–70% SOC where in fact $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ reaches the theoretical value. These low values of $r_{\mathrm{g}/{\mathrm{O}}_2}$ and $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ can be attributed to delithiation or Li vacancy creation process in stoichiometric and structurally more ordered solution grown Li₂O₂ at the initial part of the plateau.^{[ 6b ]} This is consistent with the oxygen evolution trend during RC of solution grown large toroidal Li₂O₂ that also shows very low $r_{\mathrm{g}/{\mathrm{O}}_2}$ at ≈5–20% SOC as demonstrated in Figure S18 in the Supporting Information. After this initial delithiation process, formation of non‐stoichiometric phases and gradual decrease in both particle size and thickness of Li₂O₂ deposits leading to improved charge transport can be one of the reasons behind the high OER rate at the later stage of RC. A possible mechanism for the decomposition of Li₂O₂ is schematically described in Figure 7f. After complete decomposition of Li₂O₂, high amount of carbon dioxide (CO₂) is evolved due to decomposition of parasitic products. From the ratio of consumed oxygen to evolved oxygen measured by in situ mass measurement and OEMS respectively, we have calculated an OER/ORR ratio of 0.47. Considering ≈70% yield of Li₂O₂ after DC, a correction of OER/ORR ratio gives a value of 0.67 which still indicates loss of oxygen during RC due to parasitic reactions. As a result, the overall mass change rate (r _w) of the electrode does not follow the trend of $r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2}$ rather shows excellent consistency with the cumulative rate of gas evolution (r _gshown by the red curve in Figure 7b) that include the information about parasitic reactions. The problem of parasitic reactions also affects the cycling stability of the electrodes. Figure S19a–c in the Supporting Information shows the 10 galvanostatic cycles of KB, KB_r and KB_ox electrodes, respectively. Deterioration of the cycling performance is evident in all the cases. However, the comparison of the 5th and 10th cycles of these electrodes in Figure S20a,b in the Supporting Information, respectively, shows that KB_ox consistently maintains the lowest RC potential among these three electrodes. Nonetheless, considering gradual increase in η _RC over the cycles, it should be noted that in order to achieve an efficient Li–O₂ battery with large capacity, high rechargeability and long cycle life, it is essential to improve the reversibility of ORR and OER by mitigating the problem of parasitic reactions.

Figure 7.

a) E versus Q plot for RC of KB electrode at j _DC of 100 mA g⁻¹ in LiTFSI/TEGDME electrolyte (<20 ppm H₂O). b) Online electrochemical mass spectrometry (OEMS) results showing the rate of gas evolution (r _g), c) Rate of decomposition of Li₂O₂ ($r_{\mathrm{L}{\mathrm{i}}_2{\mathrm{O}}_2})$ measured by titration of the electrodes with TiOSO₄ solution at different SOCs and d) Rate of mass change of the whole cell by in situ weight measurement during RC of the cells. e) Anodic linear sweep voltammetry (LSV) plot using rotating ring disk electrode (RRDE) with KB disk and glassy carbon (GC) ring at a sweep rate of 5 mV s⁻¹ and rotation speed of 900 rpm. f) Schematic representation of the proposed RC mechanism. The broken gray lines in b)‐d) represent the rates corresponding to 2e⁻/O₂ process. The green line in (d) is just a visual guide to the trend of the data.

3. Conclusion

In summary, we have investigated the recharge mechanism of Li–O₂ battery in order to identify a non‐catalytic low energy decomposition route for Li₂O₂. The onset potential for Li₂O₂ decomposition has been found to be very close to the equilibrium potential of 2.96 V versus Li/Li⁺. However, during the course of recharge, the potential gradually increases, goes above 4 V versus Li/Li⁺ and forms a flat‐potential plateau. Our detailed analyses have shown that neither the bulk morphology of Li₂O₂ nor the parasitic side reactions can completely explain this two‐stage recharge profile. Here we have demonstrated that the recharge potential of a Li–O₂ battery can rather be fundamentally correlated to the crystallization pathway of Li₂O₂. Depending on thermodynamic feasibility, Li₂O₂ crystallization occurs by concurrent electrochemical surface reduction and solution mediated chemical disproportionation reactions over a wide range of discharge conditions. A combination of experimental data and theoretical calculations shows, the Li₂O₂ grown by these two routes forms two coexisting discrete phases with subtle structural differences. In an unprecedented attempt, we have deconvoluted and quantitatively estimated the amount of Li₂O₂ deposited by these two different processes. Our quantitative analyses precisely specify that the surface‐grown structurally disordered Li₂O₂ decomposes at relatively low recharge potential (<3.5 V vs Li/Li⁺). Detailed mechanistic study identifies a delithiation induced partial solution phase decomposition of surface grown Li₂O₂ to be a non‐catalytic low energy oxidation pathway. On the contrary, regardless of bulk morphology, solution‐grown Li₂O₂ decomposes at recharge potential >4 V versus Li/Li⁺. Therefore, any discharge condition that promotes predominant surface growth of Li₂O₂ should be beneficial for efficient recharge of the cell. We have further shown that the equilibrium between these two growth processes and the subsequent recharge profile can be controlled by the properties of the electrode as well as of the electrolyte. In short, our results provide important insights into necessary control over fundamental crystallization pathways of Li₂O₂ by proper choice of electrode and electrolyte to improve the energy efficiency of Li–O₂ battery.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Dutta A., Ito K., Nomura A., Kubo Y., Quantitative Delineation of the Low Energy Decomposition Pathway for Lithium Peroxide in Lithium–Oxygen Battery. Adv. Sci. 2020, 7, 2001660 10.1002/advs.202001660