Probing the Reaction Kinetics of the Charge Reactions of

---

Letter pubs.acs.org/JPCL

Probing the Reaction Kinetics of the Charge Reactions of Nonaqueous Li−O2 Batteries Yi-Chun Lu*,†,‡,§ and Yang Shao-Horn*,†,‡,§ †

Department of Mechanical Engineering, ‡Department of Materials Science and Engineering, and §Electrochemical Energy Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States S Supporting Information *

ABSTRACT: Understanding the reaction mechanism of nonaqueous oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) is key to increase the low roundtrip efficiency and power capability of rechargeable Li-air batteries. Here we show that the ORR kinetics are much faster than OER kinetics and OER occurs in two distinct stages upon Li-air battery charging. The first OER stage occurs at low overpotentials (4.0 VLi) and that the application of catalysts can effectively decrease the Li−O2 recharge potential compared with the uncatalyzed carbon.10,12 Providing mechanistic insights into the OER kinetics and addressing these apparent discrepancies are key to develop strategies for improving the round-trip efficiency of Li−O2 batteries. In this study, we examine the influence of the discharge/ charge rates and discharge capacity prior to charging on the reaction kinetics of the Li−O2 battery charging using galvanostatic and potentiostatic measurements. We report three distinct stages upon charging including an initial sloping voltage region from 2.9 VLi to 3.4 VLi, a subsequent flat voltage plateau that was highly sensitive to charge/discharge rates and © XXXX American Chemical Society

Received: November 11, 2012 Accepted: December 17, 2012

93

dx.doi.org/10.1021/jz3018368 | J. Phys. Chem. Lett. 2013, 4, 93−99

The Journal of Physical Chemistry Letters

Letter

Figure 1. (a) Discharge profiles of Li−O2 cells with VC electrodes in O2-saturated 1.0 M LiClO4 in tetraethylene glycol dimethyl ether (TEGDME) at 10, 100, and 1000 mA/gc to 200 mAh/gc. (b) Blue square: discharge potentials of the Li−O2 cells shown in panel a at the 10% of discharge capacity (i.e., 20 mAh/gc) as a function of the true-carbon-surface-area normalized discharge current density. Orange square: capacitive- and IRcorrected specific ORR polarization curve of a GCE in O2-saturated 0.1 M LiClO4 in dimethoxyethane (DME) at 100 rpm and 20 mV/s reported in ref 13.

Figure 2. (a) Charge profiles of Li−O2 cells (all predischarged to 200 mAh/gc at 100 mA/gc) at 2, 5, 10, 20, and 100 mA/gc in O2-saturated 1.0 M LiClO4 in TEGDME. (b) Charge potentials (at the first 0.05 μAh/cm2true = 0.18 nm (∼0.2 nm) of Li2O2 in thickness) of select studies8,9,17 and this work as a function of total TSAC (μAh/cm2true) formed at the end of discharge. The gray dashed lines denote the estimated Li2O2 thickness estimated from the total TSAC assuming a uniform distribution of Li2O2 on the true electrode surface. The volumetric capacity of Li2O2 is 2699 mAh/cc (molecular weight of Li2O2 = 45.88 g/mol, density of Li2O2 = 2.31 g/cm3)18 The number in the parentheses denotes the TSAR during charging in μA/cm2true. (c) Green and red symbols: charge potentials (at the first 30% of charge capacity) of select studies8,9,17 and this work. The number in the parentheses denotes the TSAC in μAh/cm2true. For blue symbols, the specific surface area of Li2O2 is estimated to be 7.53 m2/gLi2O2 assuming spherical particles with ∼345 nm in diameter19 and the Li 2O2 density of 2.31 g/cm318 using a formula 6(1000/ (densityLi2O2*diameterLi2O2).20 (d) Charge potentials of three Li−O2 cells that were predischarged to 200 mAh/gc in O2-saturated 1.0 M LiClO4 in TEGDME at 10, 100, and 1000 mA/gc.

At low coverage (e.g., 20 mAh/gc, translating ∼0.1 nm Li2O2 in thickness assuming uniform Li2O2 formation on the VC surfaces), the discharge potential, which largely reflects the ORR kinetics on VC, was decreased from 2.75 VLi at 10 mA/gc, to 2.70 VLi at 100 mA/gc, and to 2.54 VLi at 1000 mA/gc, giving

rise to a Tafel slope of ∼150 mV/dec. Such a Tafel slope is similar to that found for the Li+-ORR kinetics obtained on a glassy carbon electrode (GCE) via rotating disk electrode technique,13 as shown in Figure 1b. Interestingly, the discharge potentials of Li−O2 cell are ∼75 mV lower than those of GCE 94

dx.doi.org/10.1021/jz3018368 | J. Phys. Chem. Lett. 2013, 4, 93−99

The Journal of Physical Chemistry Letters

Letter

Figure 3. (a) Potential steps (blue) and corresponding current response (red) during PITT measurements (10 mV steps, 1.3 mA/gc current cutoff or 50 h durations) on charging of an electrode that was discharged to 200 mAh/gc at 100 mA/gc in O2-saturated 1.0 M LiClO4 in TEGDME. (b) Zoom-in of panel a in stage I. (c) Zoom-in of panel a in stage II. (d) Comparison between the reported19 Li2O2-oxidation current density on VC electrodes at 4.1 VLi and the Li−O2 charge current of stage II (shown in panel c) as a function of the potential holding time.

at comparable rates, which might be attributed to the Li+-ORR activity difference between GCE and VC. At higher coverage (e.g., 200 mAh/gc, translating ∼0.74 nm Li2O2 in thickness), the discharge potential, which reflects the ORR kinetics on Li2O2, gave rise to a higher Tafel slope (∼180−200 mV/dec) than that of low coverage. Interestingly, the overpotential of the Li+-ORR kinetics observed on the VC surfaces at 10 mA/gc at 200 mAh/gc (i.e., 200 mV) is much smaller than the computed overpotential for Li+-ORR on the surface of Li2O2 (11̅00) (450 mV)16 but closer to the overpotentials predicted along the kink (190 mV) and step (280 mV) sites of Li2O2 (110̅ 0).17 The reaction kinetics upon Li−O2 battery charging are more sluggish and sensitive to rates compared with that of the discharge ORR reaction. Figure 2a shows the charge profiles of five VC electrodes that were predischarged at 100 mA/gc to 200 mAh/gc at charging rates of 2 to 100 mA/gc. The charge profiles qualitatively consist of three reaction stages. Stage I is a sloping region between 3.0 VLi to 3.4 VLi with negligible ratedependence. Stage II is a relatively flat plateau region (especially at slow rates) between 3.4 VLi (2 mA/gc) and 4.1 VLi (100 mA/gc) with strong rate dependence. Stage III is a second plateau following stage II between 3.6 VLi (2 mA/gc) and 4.3 VLi (100 mA/gc). The percentage of the charge capacity associated with stages I, II, and III was about 25, 50, and 25%, respectively. We applied a potentiostatic intermittent titration technique (PITT)21−24 and a galvanostatic intermittent titration technique (GITT) to study the Li−O2 battery charge kinetics, which are quasi-equilibrium techniques to probe the kinetic of each redox process. Figure 3a shows the PITT response of a VC

electrode that was predischarged to 200 mAh/gc at 100 mA/gc. The characteristic three-stage response is consistent with the charge profiles at low rates (Figure 2a, 2−10 mA/gc). Using low rate to probe reactions can reveal electrode processes at quasiequilibrium. At high rates, different potential-dependence of these processes on current can blur the distinction of these reaction stages or fundamentally different processes. Therefore, we discuss Li2O2 oxidation reaction mechanisms and processes using low-rate electrochemical and PITT/GITT results, where the electrodes reach quasi-equilibrium. The zoom-in of the PITT response in stage I (Figure 3b) shows a maximum current at the start of each potential increment, followed by monotonic decay. Such characteristic monotonic decay is indicative of diffusion-dominated processes (e.g., Li+ diffusion in Li2O2) comparable to the kinetics of lithium intercalation/ deintercalation24−26 and the absence of a significant barrier to the oxidation kinetics in stage I.26 Interestingly, the PITT response of stage II is completely different from that of stage I (Figure 3c), where the current first decays, slowly rises to reach a maximum, and then gradually decays. This kinetic response suggests a barrier for reaction kinetics that is gradually overcome at constant potential as the reaction proceeds.23,26 The current response can be described qualitatively by a nucleation and growth process.23 First, the initial current decay suggests the formation and coalescence of subcritical-size nuclei to yield supercritical nuclei23,27,28 (e.g., lithium and/or oxygen vacancies). Second, the subsequent increase in current can be attributed to the growth of the supercritical nuclei in size and number, which is accompanied by an increase in the electrochemically active area for the reaction kinetics.23,29,30 95

dx.doi.org/10.1021/jz3018368 | J. Phys. Chem. Lett. 2013, 4, 93−99

The Journal of Physical Chemistry Letters

Letter

deintercalation process via a solid-solution route from the outer part of Li2O2 to form LiO2-like species on the surface (Li2O2 → LiO2 + Li+ + e−), where LiO2-like species disproportionate to evolve O2 (LiO2 + LiO2 → Li2O2 + O2), yielding an overall 2e−/O2 OER process (Li2O2 → 2Li+ + O2 + 2e−), as shown in Figure 4. This hypothesis is consistent with a recent study by

Third, the decrease in current after the maximum can be attributed largely to the decrease in the available Li2O2. The nucleation and growth-type processes associated with stage II is further supported by GITT tests, as shown in Figure S3 of the Supporting Information. During the GITT charging in stage II (Figure S3c of the Supporting Information), the potential first increases to support the initial nucleation events (e.g., forming supercritical nuclei from subcritical-size nuclei),23,27,28 then decreases as a result of increased total surface area of supercritical nuclei (i.e., lower overpotential needed for sustaining the same current). This PITT current response of stage II is comparable to the oxidation current profile of electrodes preloaded with Li2O2 reported by Harding et al.,19 as compared in Figure 3d. The time needed to reach the maximum oxidation current for electrodes preloaded with Li2O2 is longer than that in the Li− O2 cell, which can be attributed to the fact that Li2O2 particles in the preloaded electrodes (∼345 nm)19 are much larger than those in the discharged electrodes examined here (coating-like morphology, Figure S1 of the Supporting Information). More interesting to note is that the nucleation and growth kinetics of Li2O2 oxidation can be promoted greatly by catalysts (Figure S4 of the Supporting Information),19 where the initiation time is much shorter with Pt/C or Ru/C than VC, even at a significantly lower charging potential (3.6 VLi) compared with VC (4.1 VLi)19 (Figure S4 of the Supporting Information). The separation between stages I and II can be attributed to the differences in stoichiometry and/or oxygen local environments between the outer part and the bulk part of the Li2O2.9 To support this, we show that the oxygen local environment of the outer part of the Li2O2 is different from that of the bulk part of the Li2O2 using X-ray absorption near-edge structure (XANES) technique with surface sensitive total electron yield (TEY) mode and bulk fluorescence yield (FY) mode5,9 (Figure S5 of the Supporting Information). We further evaluate the OER kinetics of stage I (i.e., the sloping region) by examining the charge potentials at a small capacity. This allows one to investigate the OER kinetics of oxidizing the first subnanometer layer or the outer part of Li2O2 formed on discharge. Figure 2b shows the charge potentials, which were estimated at the charging capacity of 0.05 μAh/ cm2true of Li−O2 cells (∼0.2 nm of Li2O2 oxidized assuming a uniform distribution of Li2O2 on the true electrode surface) with pure carbon electrodes as a function of the total discharge capacity.8,9,17 The charge potentials appear to be independent of the overall Li2O2 thickness formed in the Li−O2 cells up to 18 nm at comparable rates (5 × 10−3 to 0.1 μA/cm2true). This suggests that the OER process is not limited by transport processes that are dependent on the Li2O2 thickness, for example, the hole tunneling process described by Viswanathan et al.16 For instance, it is reported16 that the highest current density that can be drawn from a 5 nm thick Li2O2 is ∼10−9 μA/nm2Li2O2 (∼105 μA/cm2Li2O2) at an overpotential of 0.6 V and ∼10−14 μA/nm2Li2O2 (∼1.0 μA/cm2Li2O2) at an overpotential of 0.1 V. Clearly, these “limiting currents” predicted for tunneling (1.0−105 μA/cm2Li2O2) are at least 10 times higher than the rates used in the Li−O2 cells included in Figure 2b (5 × 10−3 to 0.1 μA/cm2true), which suggests that the OER process in stage I is not limited by the tunneling process. Here we discuss possible reaction processes responsible for stage I. Considering the sloping characteristics of stage I with relatively low overpotentials (4.0 VLi at comparable rates).9,11,12 To examine further the influence of the intrinsic charge current density on the charge plateau potential (stage II), we compare the charge potentials as a function of TSAR upon charge in Figure 2c (the green and red symbols), which 96

dx.doi.org/10.1021/jz3018368 | J. Phys. Chem. Lett. 2013, 4, 93−99

The Journal of Physical Chemistry Letters

Letter

were estimated at ∼30% of the charge capacity of Li−O2 cells with carbon electrodes (having TSACs in the range of 0.2 to 1.0 μAh/cm2true). The charge potentials at ∼30% charge capacities were selected to have a large enough depth of charge to capture the dominant process responsible for this charge plateau and to minimize the extent of parasitic reactions such as electrolyte decomposition at voltages greater than ∼4.0 VLi. As shown in Figure 2c, the charge potential increases sharply with TSAR, forming a Tafel slope of ∼300 mV/dec. Interestingly, the OER potentials (blue symbols) of electrodes that were prefilled with chemically synthesized Li2O2 via potentiostatic measurements were reported by Harding et al.19 as a function of rate following a comparable Tafel slope. Note that the rates were normalized to the true surface area of Li2O2 loaded (blue symbols), instead of the electrode surface area, due to the fact that the Li2O2 used in these experiments were large discrete particles (∼345 nm in diameter).19 This further supports the fact that the Tafel slope of OER kinetics in stage II (∼300 mV/ dec, Figure 2c) is associated with the oxidation of the bulk of Li2O2 particles. One cannot exclude the formation of LiO2 via a two-phase route and its subsequent disproportionation reaction. Further studies are needed to identify the reaction intermediate species involved in stage II. We note that the reactions occurring in stages I (3.0 to 3.4 VLi) and II (∼3.4 VLi) in this study are significantly lower than the electrolyte decomposition potential reported for the ether-based electrolytes (∼3.8 VLi).31 Nevertheless, considering recent suggestions about the formation of carbonate species17,31,34 and singlet oxygen21,34 during the initial charging, it is important to further investigate their influences on the reaction kinetics and mechanisms of the Li−O2 recharge. The voltage plateau in stage II was found to be influenced by a number of additional factors besides the charge rate. First, stage II voltage plateau was shown to be lowered significantly by catalysts in both Li−O2 cells10,12 and Li2O2-filled electrodes (blue symbols in Figure 2c).19 Harding and coworkers.19 have shown that the electrochemical oxidation kinetics of Li2O2 can be significantly promoted by Pt and Ru catalysts in electrodes preloaded with Li2O2. SEM and XRD characterization have confirmed the complete removal of Li2O2 after potentiostatic charging.19 However, the physical origin associated with the enhanced OER kinetics by catalysts is not understood, and further studies are needed to reveal the working mechanism of catalyzed Li2O2-oxidation. Second, increasing the discharge capacity led to increased charging voltages in stage II, as shown in Figure S6a of the Supporting Information. This is likely due to the decrease in the electrical conductance with increasing Li2O2 thicknesses.16,17 Assuming that the voltage gap between the charge profiles with different discharge capacities (or Li2O2 thickness) is only related to the differences in the electrical conductance of the Li2O2 deposit, the electronic conductivity of Li2O2 can be estimated to be ∼2 × 10−13 S/cm (Supporting Information), which is in reasonable agreement with that estimated from a recent study16 (∼1 × 10−12 S/cm, Supporting Information), reporting an increase of 0.1 V of overpotential associated with an increase in Li2O2 thickness of 1 nm.16 The low Li2O2 conductivity estimated here (10−12 to 10−13 S/cm) suggests that the electrical transport losses will be significant in Li−O2 cells with increasing Li2O2 thickness or at high rates. Third, increasing the discharge rate of Li2O2 prior to charging from 10 to 1000 mA/gc was found to increase stage II charging voltage (at the same charge rate of 2 mA/gc), as shown in Figure 2d. Remarkably, the charge potential plateau of the

discharged electrode at 10 mA/gc was as low as 3.3 VLi. The physical origin responsible for the decreased charge overpotential with decreasing discharge rate is currently not understood. Further studies are needed to resolve the differences in the chemical and structural nature of the Li2O2 and byproducts formed at different discharge rates to provide insights into the enhanced mechanism. We believe that the experimentally obtained minimum overpotential at the charge plateau (stage II) (η = 340 mV) is an important basis for future studies on the “theoretical overpotential” of bulk Li2O2oxidation, from which can potentially shed light on the electro-oxidation mechanism of Li2O2 in the rechargeable Li− O2 batteries. Reaction processes responsible for stage III at the end of charge are likely related to the decompositions of both carbonate-type byproducts and electrolyte to evolve CO2 (Figure 4), which is in agreement with the detection of largely CO2 at the end of charge by McCloskey et al.17,31,32 This is supported by the PITT current response of stage III (Figure S7 of the Supporting Information), where at a potential of ∼3.8 VLi the current increases sharply and reaches its steady-state at ∼4.6 mA/gc, and the steady-stage current at 3.8 VLi in Figure S7 of the Supporting Information continued far beyond the total capacity of 200 mAh/gc (Figure S7 of the Supporting Information). In summary, we examine the OER kinetics and reveal three major reaction stages upon the Li−O2 battery recharge, as shown in Figure 4. During the oxidation of the first subnm layer of Li2O2, the reaction exhibits a sloping voltage profile, requires relatively low overpotentials (