Chemical and Electrochemical Differences in Nonaqueous Li–O2

Angela Speidel and Rouven Scheffler of Volkswagen Group, along with Seok Ju Kang at IBM are also gratefully acknowledged for fruitful conversations...
0 downloads 0 Views 1MB Size
Letter pubs.acs.org/JPCL

Chemical and Electrochemical Differences in Nonaqueous Li−O2 and Na−O2 Batteries Bryan D. McCloskey,*,†,§ Jeannette M. Garcia,† and Alan C. Luntz†,‡ †

IBM Research, Almaden Research Center, San Jose, California 95120, United States SUNCAT, SLAC National Accelerator Laboratory, Menlo Park, California 94025, United States



ABSTRACT: We present a comparative study of nonaqueous Li−O2 and Na−O2 batteries employing an ether-based electrolyte. The most intriguing difference between the two batteries is their respective galvanostatic charging overpotentials: a Na−O2 battery exhibits a low overpotential throughout most of its charge, whereas a Li−O2 battery has a low initial overpotential that continuously increases to very high voltages by the end of charge. However, we find that the inherent kinetic Li and Na−O2 overpotentials, as measured on a flat glassy carbon electrode in a bulk electrolysis cell, are similar. Measurement of each batteries’ desired product yield, YNaO2 and YLi2O2, during discharge and rechargeability by differential electrochemical mass spectrometry (DEMS) indicates that less chemical and electrochemical decomposition occurs in a Na−O2 battery during the first Galvanostatic discharge−charge cycle. We therefore postulate that reactivity differences (Li2O2 being more reactive than NaO2) between the major discharge products lead to the observed charge overpotential difference between each battery. SECTION: Energy Conversion and Storage; Energy and Charge Transport

L

battery employing an electrolyte with an ethereal solvent (diglyme) was convincingly shown to be:22,24

i−air batteries have attracted considerable attention as a potential high energy alternative to existing state-of-the art rechargeable batteries.1−3 In previous work, the active reversible electrochemical reaction occurring within the nonaqueous Li− air battery cathode was shown to be:4−8 2(Li+ + e−) + O2 ↔ Li 2O2 ,

Na + + e− + O2 ↔ NaO2 ,

(2)

a 1 e−/O2 process compared to the 2 e−/O2 process observed in the Li−O2 battery employing a similar electrolyte solvent. The formation of sodium superoxide, and not sodium peroxide, is rather surprising given the analogous Li−O2 electrochemistry leads to Li2O2 formation and that Na2O2 formation (E0 = 2.33 V vs Na/Na+) is thermodynamically favored, although only slightly, to form over NaO2. The origins of these electrochemical differences in Na−O2 and Li−O2 batteries are still poorly understood. Hartmann et al. suggest that the 1e− formation of NaO2 is a kinetically favored process over 2e− Na2O2 formation.22 In contrast, lithium superoxide (LiO2) has been observed as a transient intermediate in a Li−O2 battery,25 but is not stable at room temperature and quickly converts to Li2O2 via an additional Li+-induced charge transfer or disproportionation.8 Kang et al. suggests that NaO2 is more likely to form than Na2O2 in a Na−O2 battery simply because its surface energies are lower and this overwhelms the small difference in bulk formation energies (at least for nanoparticulate deposits).26

E 0 Li2O2 = 2.96 V vs Li/Li+ (1)

where E0Li2O2 is the standard potential of bulk Li2O2 formation calculated using the Nernst equation. However, the production and oxidation of Li2O2 leads to many significant challenges.9−15 Among them, we have suggested that Li2O2-induced decomposition of the electrolyte and cathode limit battery rechargeability and reduce voltaic efficiency.16 The voltaic inefficiency arises primarily from a large observed overpotential during battery charge, even though mounting theoretical evidence predicts O2 evolution from Li2O2 occurs at inherently low overpotentials.17,18 Therefore, the origin of the charge overpotential is still poorly understood, with studies suggesting various possible causes, including charge transport limitations,19,20 sluggish O2 evolution kinetics,21 and, as mentioned previously, parasitic side reactions of the electrolyte and cathode. In a series of interesting recent reports, a similar system, the nonaqueous Na−O2 battery, was explored as a possible alternative to the Li−O2 battery.22−24 These reports highlighted many intriguing differences between the two batteries; for example, the active cathode electrochemistry of the Na−O2 © 2014 American Chemical Society

E 0 NaO2 = 2.27 V vs Na/Na +

Received: March 10, 2014 Accepted: March 17, 2014 Published: March 17, 2014 1230

dx.doi.org/10.1021/jz500494s | J. Phys. Chem. Lett. 2014, 5, 1230−1235

The Journal of Physical Chemistry Letters

Letter

Figure 1. Galvanostatic discharge−charge of a Li−O2 cell (a−c)33 and a Na−O2 cell (d−f). (b,e,) O2 consumption, nO2,d, and Li2O2 or NaO2 formation, ni,d, during discharge; (c,f) O2 evolution, nO2,c, and Li2O2 or NaO2 oxidation, ni,c, during charge (the dashed black line is a 2 e− process for (c) and a 1 e− process for f). 1N LiTFSI and 0.2N NaOTf in DME were used as electrolytes. The Li−O2 cell was discharged at 2 mA and charged at 0.5 mA, whereas the Na−O2 cell was both discharged and charged at 0.5 mA.

charge in a Li−O2 battery. Lithium carbonates, which do not decompose until high (>4 V vs Li/Li+) potentials, continuously form in a Li−O2 battery during charge, whereas very little sodium carbonates, or any other side products, form in the Na−O2 battery. These results bolster the contention that the formation of parasitic side products during battery charge lead to the observed charging overpotential in Li−O2 batteries. Figure 1 compares the galvanostatic discharge−charge of a Li−O2 and Na−O2 battery using DEMS and pressure decay/ rise measurements.5 Of significant interest in this study, the voltage profiles of each battery are distinctly different during charge (Figure 1a,d), with the Na−O2 overpotentials being consistent with those reported by Hartmann et al.22,24 For the Li−O2 battery, the charge overpotential has been discussed extensively in previous publications and is briefly discussed here.8,16,17,29 Initially, the Li−O2 charging potential is quite low and is only slightly higher than the battery’s open circuit potential after discharge, which is 2.85 V. We have previously ascribed this value (2.85 V) to be the equilibrium potential of Li2O2 growth on Li2O2 deposits (Uo,Li2O2).8 The potential then monotonically increases over the initial ∼0.3 mAh portion of charge, leading to a sudden jump in potential around 0.7 mAh, where the potential reaches a high plateau of around 4.6−4.7 V. The time-average overpotential during Li−O2 charge is approximately 1.2 V. Other carbon electrodes exhibit similarly high overpotentials, although the voltage profiles are qualitatively different, e.g., a Vulcan XC72 carbon cathodebased battery also has a low initial charging potential, but the

The differences in the Na and Li−O2 electrochemistry appear to have a sizable effect on their respective performance. An appropriate metric of rechargeability, the ratio of oxygen evolved during charge (OER) to oxygen consumed during discharge (ORR), showed that the Na−O2 battery is more rechargeable than the Li−O2 battery, at least after the first galvanostatic cycle.22 Although the Na−O2 system has a significantly lower theoretical specific energy (1100 Wh kg−1 to 3450 Wh kg−1),22,27 the capacity of the Na−O2 battery was much higher at low current densities (∼100 μA/cm2) than a similarly discharged Li−O2 battery.22 Further, although both Li−O2 and Na−O2 batteries exhibited low overpotentials on discharge, the Na−O2 overpotential remains low during most of battery charge, unlike in the Li−O2 battery, where a low charge overpotential is initially observed, but then continuously increases to an ultimate potential of >4 V vs Li/Li+.22,28 Given the fascinating differences between these two batteries, as highlighted by Hartmann et al., we set out to explore the electrochemistry and chemistry of each in hopes that we could learn more about the phenomenological origins of their different galvanostatic voltage profiles. We report here our initial findings on the chemical differences between these two batteries. We find that less decomposition of the electrolyte/ cathode occurs in the Na−O2 battery compared to the Li−O2 battery, likely resulting from differences in reactivity of the primary O2 reduction products, NaO2 vs Li2O2. Most parasitic decomposition in the Na−O2 battery occurs during discharge, whereas decomposition occurs nearly equally on discharge and 1231

dx.doi.org/10.1021/jz500494s | J. Phys. Chem. Lett. 2014, 5, 1230−1235

The Journal of Physical Chemistry Letters

Letter

potential and an oxidation ‘tail’ that extends above the initial anodic peak that is essential to eliminate cycle-to-cycle passivation. The difference between Uo and the onset potential at which reduction/oxidation currents commence, are the ORR and OER overpotentials, respectively. 8 For the Na−O 2 electrochemistry, these overpotentials are very low to nonexistent, with the onset reduction/oxidation currents commencing right at Uo. In the Li−O2 cell, these onsets are only ∼0.2 V above/below Uo, which still correspond to very low overpotentials (a more exhaustive and quantitative study comparing Li−O2 and Na−O2 kinetics is currently in progress). Reasonably assuming that the fundamental Li or Na−O2 electrochemistry occurring in the bulk electrolysis cell and battery are similar, bulk electrolysis results predict that the Li− O2 overpotentials would be at most only slightly higher (∼0.2 V) than the Na−O2 overpotentials in the battery. However, we see a dramatic difference in the observed battery overpotentials, and this suggests that these overpotentials are not related to the fundamental kinetic overpotential of the electrochemical reactions between Li or Na and O2. Therefore, some other difference in battery chemistry/ electrochemistry must be leading to the differences in charge overpotentials. Our original hypothesis on the ever-increasing Li−O2 overpotential is that it is a result of parasitic chemistry of Li2O2 with the electrolyte.16 A comparison of oxygen consumption and evolution to desired product formation (Li2O2 or NaO2) during galvanostatic discharge−charge allows us to quantify the total parasitic chemistry occurring in each battery (Figure 1, methods used to quantify each species are discussed in the Experimental Methods section).33 Overall, the Na−O2 cathode chemistry is clearly ‘cleaner’ than the Li−O2 battery on the first galvanostatic cycle, with salient points on each discussed in the ensuing paragraphs. Table 1 provides definitions of variables discussed and summarizes important parameters for both Na−O2 and Li−O2 batteries.

potential continuously monotonically increases during the entirety of charge to a final potential of 4.5 V.28 The cause of the rise in charge potential is currently a point of contention in the field. We have postulated that it is parasitic product formation at the Li2O2−electrolyte interface that leads to the high observed overpotentials, not the inherent kinetic Li2O2 overpotential, which we have asserted is quite low.16,30 Hummelshoj et al.17 and Kang et al.18 have also calculated that Li2O2 oxidation should occur at low overpotentials. Others have reported that the mechanism for OER from Li2O2 leads to sluggish kinetics and the resulting high overpotential.21,31,32 However, the monotonic charge potential increase and high average overpotential are not observed in a Na−O2 battery. Under an O2 headspace, the open circuit potential of a discharged cell is 2.24 V vs Na/Na+ (Uo,NaO2), which is equal to the calculated potential for bulk NaO2 formation at 0.2N Na+ and 1.5 atm O2 (the O2 pressure used in this study). From the open circuit potential, the charge potential of the Na−O2 battery initially jumps to 2.6 V vs Na/Na+ and then decreases to 2.5 V, leading to an average overpotential of only 0.25 V, where it stays for ∼80% of the charge. Only at the end of charge (the final 0.15 mAh) does the potential suddenly increase to >4 V. If the significantly lower Na−O2 battery charging potential, relative to that of a Li−O2 battery, is due to differences in the fundamental kinetic overpotentials of each electrochemical reaction, these overpotentials should then be apparent when measuring Na and Li−O2 electrochemistry in a three electrode bulk electrolysis cell. However, only small differences in the inherent ORR and OER kinetic overpotentials are observed for each on a smooth glassy carbon working electrode in a wellstirred cell (Figure 2). Characteristic features of ORR and OER are similar in both cells, with a single, large cathodic and anodic current peak present on either side of each cells’ equilibrium

Table 1. Definition of Variables Used in This Study and Their Values during the First Galvanostatic Discharge− Charge Cycle of a Li−O2 and Na−O2 Battery (e−/O2)d (e−/O2)c YLi2O2,d; YNaO2,d OER/ ORR n′xCO3,d (a) n′xCO3,c (a)

definition

Li−O2

Na−O2

average electrons consumed per O2 on discharge average electrons consumed per O2 on charge ratio of the amount of Li2O2/ NaO2 formed to amount Li2O2/NaO2 expected given the Coulometry ratio of O2 evolved on charge to O2 consumed on discharge average Li2CO3 or Na2CO3 formation rate on discharge average Li2CO3 or Na2CO3 formation rate on charge

2.02 ± 0.02

1.02 ± 0.02

2.59 ± 0.02

1.10 ± 0.02

0.91 ± 0.01

0.945 ± 0.01

0.78 ± 0.03

0.93 ± 0.02

0.04

0.10

>0.3

4 V.30 No such Na2CO3 accumulation occurs at any time during cell charge, and therefore a low overpotential is observed, consistent with the overpotentials measured in the bulk electrolysis measurements. It should be noted that NaO2 oxidation does not contribute exclusively to the initial 0.75 mAh of the Na−O2 charge electrochemistry: only 0.96 NaO2/e− is oxidized rather than the desired 1.00 NaO2/e−. Most of the parasitic chemistry in the Na−O2 battery occurs on discharge, as the total OER/ORR for the Na−O2 cell was measured to be 0.93, in agreement with Hartmann et al.,24 and YNaO2,d was only slightly higher (0.945). It is therefore likely that the additional electrochemical process occurring during the initial stage of charge is related to oxidation of some decomposition product formed on discharge. The chemical and physical properties of the parasitic chemistry occurring are unknown, as we were unable to detect any decomposition species present using NMR, aside from the small amount of NaOAc and NaHCO2 initially observed during discharge. The origin of the sudden charge potential increase near the end of charging is not yet entirely understood. However, the potential increase is consistent with decomposition products accumulating in the solid phase on the surface of the cathode (as is the case in the Li−O2 battery35). As the ratio of NaO2 to decomposition products reaches a critical value, a sudden increase in charge potential may occur as a result of a mixed oxidation potential. It should be noted that this sudden potential rise would also be consistent with the possibility that NaO2 is slightly soluble in the electrolyte, which would lead to a small portion of it losing electronic contact with the cathode. Further studies to elucidate the exact mechanistic origins of the charge potential rise are warranted. Finally, only a small amount of NaO2 (∼0.1 μmol), and perhaps another basic solid product, was observed on the cathode surface via a base and iodometric titration after a full galvanostatic cycle (no products were detected using NMR). No solid cathode products could be observed after an additional 0.05 mAh charge beyond the full 1 mAh charge cycle. In conclusion, we present evidence that the differences in charge overpotential in Li−O2 and Na−O2 batteries arise from the difference in their primary discharge products’ reactivity. OER in both Li−O2 and Na−O2 batteries should have similar overpotentials given a comparison of their respective CVs on a smooth glassy carbon electrode. However, Li2O2 is more reactive than NaO2 and therefore promotes more electrolyte and carbon decomposition than NaO2 as observed by quantitative titrations and DEMS. The decomposition products formed during Li−O2 battery charge likely accumulate at the

discharge−charge cycle by taking more care during the product extraction process).33 Only a very small amount of sodium acetate and an even smaller amount of NaHCO2, neither of which could be quantified given their low concentration, were observed using NMR analysis of D2O-extracted Na−O2 discharge products. No other parasitic side products have been identified, e.g., NaF, in the Na−O2 cells. During the initial ∼0.3 mAh of charge, the Li2O2 oxidation rate is higher than the O2 evolution rate (0.94 Li2O2/2e− vs 0.91 O2/2e−), with much stronger deviations between O2 evolution and Li2O2 oxidation at higher charging potentials. This indicates that a portion of Li2O2 oxidation contributes to parasitic reactions throughout cell charge.33 In contrast, NaO2 oxidation and O2 evolution are statistically equal to each other for the first 0.5 mAh of charge (0.96 NaO2 and O2 per e−), and only slightly deviate away from each other toward the end of charge at higher potentials. In other words, NaO2 oxidation leads exclusively to O2 evolution and not other parasitic electrochemistry during the initial half of the charge. In a previous study, we suggested that the Li2O2 parasitic oxidation/ charge reaction forms lithium carbonates, among other decomposition products. This result was confirmed using a quantitative carbonate analysis method outlined by Thotiyl et al. (Figure 3a,b also summarize these results).33,34 In fact,

Figure 3. (a) Galvanostatic discharge−charge curve, and (b) total Li2CO3 formed as measured by CO2 evolution from discharged and partially charged cathodes immersed in 3 M H2SO4.33 The labels of the black arrows in (a) are the charge-normalized Li2CO3 (in μmol Li2CO3/mAh) over the corresponding regions. Both the plot and table are color coded, with the arrow color in panel a, corresponding to the similarly labeled row in b: green corresponds to a 2 mAh discharge, and red, blue, and magenta corresponds to a 2 mAh discharge, followed by a charge to 4.1 V, 1 mAh, and 1.4 mAh, respectively. 1.4 mAh charge corresponds to the approximate onset of CO2 evolution during charge. No carbonate was present at the end of a full galvanostatic cycle (orange arrow), as indicated by a base titration on a fully charged cathode. P50 was used as the cathode, 1N LiTFSI in DME as the electrolyte, i = 2 mA (discharge), 0.5 mA (charge). (c,d) Similar measurements on a Na−O2 cell.