Solvents' Critical Role in Nonaqueous Lithium–Oxygen Battery

Apr 27, 2011 - Joseph K. Papp , Jason D. Forster , Colin M. Burke , Hyo Won Kim , Alan C. Luntz , Robert M. Shelby , Jeffrey J. Urban , and Bryan D. M...
0 downloads 5 Views 2MB Size
LETTER pubs.acs.org/JPCL

Solvents’ Critical Role in Nonaqueous LithiumOxygen Battery Electrochemistry B. D. McCloskey,* D. S. Bethune,† R. M. Shelby,† G. Girishkumar,† and A. C. Luntz† Almaden Research Center, IBM Research, 650 Harry Road, San Jose, California 95120, United States

bS Supporting Information ABSTRACT: Among the many important challenges facing the development of Liair batteries, understanding the electrolyte’s role in producing the appropriate reversible electrochemistry (i.e., 2Liþ þ O2 þ 2e T Li2O2) is critical. Quantitative differential electrochemical mass spectrometry (DEMS), coupled with isotopic labeling of oxygen gas, was used to study LiO2 electrochemistry in various solvents, including carbonates (typical Li ion battery solvents) and dimethoxyethane (DME). In conjunction with the gas-phase DEMS analysis, electrodeposits formed during discharge on LiO2 cell cathodes were characterized using ex situ analytical techniques, such as X-ray diffraction and Raman spectroscopy. Carbonate-based solvents were found to irreversibly decompose upon cell discharge. DME-based cells, however, produced mainly lithium peroxide on discharge. Upon cell charge, the lithium peroxide both decomposed to evolve oxygen and oxidized DME at high potentials. Our results lead to two conclusions; (1) coulometry has to be coupled with quantitative gas consumption and evolution data to properly characterize the rechargeability of Liair batteries, and (2) chemical and electrochemical electrolyte stability in the presence of lithium peroxide and its intermediates is essential to produce a truly reversible LiO2 electrochemistry. SECTION: Energy Conversion and Storage

W

ith a theoretical specific energy of 11 700 W h per kg of lithium, nonaqueous Liair batteries potentially offer a significant improvement in energy density over that provided by current state-of-the-art rechargeable Li ion batteries.1,2 However, many challenges remain to develop a practical secondary Liair battery. Among these challenges, published Liair and LiO2 embodiments have only achieved a small fraction of their theoretical specific energy with limited rechargeability. At a fundamental level, inefficient cycling and limited specific energy may be a result of the poorly understood LiO2 chemistry occurring at the cathode. The nonaqueous Liair battery was first demonstrated by Jiang and Abraham and reported to have modest rechargeability.2 They suggested, as is now widely accepted, that the electrochemical reaction describing the cathode chemistry is lithium peroxide formation via a two-electron process 2ðLiþ þ e Þ þ O2 T Li2 O2 ðsolidÞ

ð1Þ

with the forward direction describing discharge and the reverse direction describing charge. The standard potential of Li2O2 formation (and decomposition) is U0 = 2.96 V (with respect to Li/Liþ), as given by the Nernst equation: U0 = ΔG/2F (ΔG is the Gibbs free energy of reaction 1 under standard conditions, and F is the Faraday constant). Recently, irreversible reactions given by eqs 2 and 3 have been suggested to contribute to the cathode chemistry during deep r 2011 American Chemical Society

discharges3,4 4ðLiþ þ e Þ þ O2 f 2Li2 O

U0 ¼ 2:91 V

ð2Þ

Li2 O2 þ 2ðLiþ þ e Þ f 2Li2 O

U0 ¼ 2:72 V

ð3Þ

Many studies have reported extended dischargecharge cycles with modest loss in capacity.510 In addition, the importance of nanoparticles catalysis has been suggested to enhance rechargeability and lower discharge and charging overpotentials.3,5,6,11,12 Most of these reports are based on coulometry, that is, by following the loss of discharge capacity with repeated dischargecharge cycles. This approach assumes that chemical reaction 1 describes the fundamental processes occurring in both cell discharge and charge, with a secondary parasitic process describing the loss in capacity. However, quantitative experimental evidence is lacking that definitively proves that reaction 1 is the dominant reversible electrochemical process in many of the reported Liair embodiments. Raman spectroscopy and permanganate titrations have been used to qualitatively identify Li2O2 as a discharge product.2,6 Furthermore, Read quantified the ratio of O2 consumed per electron during discharge of cells employing a variety of different solvents.13 The resulting Received: March 15, 2011 Accepted: April 18, 2011 Published: April 27, 2011 1161

dx.doi.org/10.1021/jz200352v | J. Phys. Chem. Lett. 2011, 2, 1161–1166

The Journal of Physical Chemistry Letters

LETTER

Table 1. Cathode Weight Gain during Dischargea discharge product

cathode weight gain (mg/mAh)

DME discharge

1.0

PC/DME discharge

2.3

EC/DMC discharge

2.0

Li2O2b

0.86

Li2CO3b

1.4

Li2Ob

0.56

The standard deviation of all discharge weight gains was ∼0.15 mg/ mAh based on replicate trails. b Indicates theoretical cathode weight gain if only this product was formed during discharge, assuming that the 2 e/product molecule formed (see eq S1, Supporting Information). a

Figure 1. Dischargecharge curves for (a) pure DME, (b) 1:1 (v:v) EC/DMC, and (c) 1:2 (v:v) PC/DME-based cells. Discharge conditions: 0.09 mA/cm2 under 18O2 to 2 V or 10 h. Charge conditions: 0.09 mA/cm2 under Ar. Regions IIV during charge in (a) represent different stages of gas evolution and are also labeled in Figure 4 for comparison. Each figure also includes the number of electrons consumed per oxygen molecule for various discharge regions.

noninteger e/O2 ratios were ascribed to a mixture of Li2O2 (a 2e/O2 process) and Li2O (a 4e/O2 process) production, although no spectroscopic evidence of either product was presented. Ogasawara et al. used mass spectrometry to characterize gas evolution from a cathode artificially packed with Li2O2 and found that O2 is predominantly evolved.7 To avoid complications related to H2O and CO2 contamination, oxygen is usually used in place of air to study the fundamental Liair electrochemistry. Recently, multiple conference proceedings and publications have hinted that the LiO2 electrochemistry may be substantially more complicated than a simple combination of reactions 13.8,1419 These reports include X-ray diffraction (XRD) and Raman spectroscopy on discharged cathodes and DEMS during cell charging. To date, most nonaqueous LiO2 batteries have used

carbonate solvent-based liquid electrolytes. Carbonates are typically employed as solvents in state-of-the-art rechargeable Li ion batteries as a result of the stable and cyclable solid electrolyte interface that they form at each electrode.2022 However, Laorie et al. have shown, using cyclic voltammetry, that the choice of solvent significantly influences fundamental processes occurring during oxygen reduction and oxygen evolution in the presence of Li ions.23 In this Letter, to better understand the electrochemistry occurring in LiO2 batteries, we combined in situ quantitative gas-phase mass spectrometry (DEMS, schematically described in Supporting Information Figure S1) with ex situ cathode analysis (XRD and Raman) following discharge. We compared results of cells employing various solvents, including carbonates, dimethoxyethane (DME), and a mixture of the two, because they have been used in previous LiO2 studies.6,7,11,12,18 Employing carbonate-based solvents resulted in a complex LiO2 cell discharge chemistry, including predominant decomposition of carbonates to form nonvolatile deposits in the cathode, which decomposed further to CO2 upon cell charging. In contrast, Li2O2 was predominantly formed during discharge of pure DME-based cells. Upon charging, Li2O2 decomposed to form O2, although the amount of O2 evolved was less than the amount consumed during discharge. Decomposition of DME also occurred at high potentials during charging. Our results indicate that the LiO2 electrochemistry is strongly dependent on the choice of solvent used in the cell. A brief summary of the experimental methods is given at the end of the Letter, while method details are given in the accompanying Supporting Information. The following section will describe O2 consumption and product characterization during discharge of LiO2 cells and will be followed by gas evolution characterization during charge of the same cells. Characterizing Processes during Discharge. Figure 1 presents typical galvanostatic dischargecharge curves for LiO2 cells employing various carbonate- and ether-based electrolyte solvents, with lithium trifluoromethane sulfonimide (LiTFSI) as the electrolyte salt in all cells. Cells were discharged at 0.1 mA (0.09 mA/cm2) under 1.5 atm of isotopically labeled O2 (18O2) and charged at 0.1 mA under 1.5 atm of Ar. The cells were discharged to a voltage cutoff of 2 V or a capacity cutoff of 1 mAh, whichever occurred first. No gas evolution, as monitored by the DEMS, was observed during discharge of cells identical to those studied in Figure 1. We first discuss the discharge portion of these curves. During discharge of pure DME-based cells, one broad voltage plateau, at 2.65 V, was observed (Figure 1a). The number of electrons per oxygen consumed during discharge was 2.05 ( 0.05, with the error estimate from many repeated similar measurements. Furthermore, a 1.0 ( 0.15 mg/mAh cathode weight gain 1162

dx.doi.org/10.1021/jz200352v |J. Phys. Chem. Lett. 2011, 2, 1161–1166

The Journal of Physical Chemistry Letters

LETTER

Figure 3. Raman spectra of discharged carbon cathodes from a pure DME-based cell, a 1:1 (v:v) EC/DME-based cell, and a 1:2 (v:v) PC/ DME-based cell. Neat P50 carbon paper is included for comparison. Discharge conditions: 0.09 mA/cm2 under O2 to 2 V.

Figure 2. XRD patterns of discharged carbon cathodes from cells employing (a) pure DME, (b) 1:1 (v:v) EC/DMC, and (c) 1:2 (v:v) PC/DME as electrolyte solvents. (9) and (b) denote reference Li2O2 and graphite peaks, respectively.

was observed (see Table 1). Both of these measurements are quite close to values expected for the exclusive formation of Li2O2 via eq 1 (i.e., Li2O2 formation would consume 2 e per O2 and result in a 0.86 mg/mAh cathode weight increase). The reasonable agreement between observed and theoretical values suggests, but does not conclusively prove, that the dominant electrochemistry observed during discharge of a DME-based cell is the production of Li2O2. However, the observed values are still slightly higher than the theoretical values, indicating that a parasitic electrochemical reaction, such as electrochemical reduction of the solvent, could account for a small fraction of the discharge current. The reductive and oxidative stability of the solvents explored in this study are shown in Figure S2 (Supporting Information) and discussed in more detail in the Supporting Information. To date, most reported nonaqueous LiO2 batteries employed carbonate-based electrolytes as a result of the stable and cyclable solid electrolyte interface carbonates formed at each electrode in Li ion batteries. In contrast to the constant discharge plateau in DME, the pure carbonate-based cell (Figure 1b, 1EC/ 1DMC) exhibited a 2.6 V voltage plateau that only lasted 0.25 mAh and was followed by a gradual voltage drop until the 2 V cutoff was reached at 0.63 mAh; 2.70 e/O2 were consumed during the initial voltage plateau (2.6 V), with 2.35 e/O2 consumed during the voltage drop region. Therefore, electrochemical processes other than the 2 e/O2 process defined in reaction 1 must contribute to the discharge current. The weight gain of the cathode (∼2.0 mg/mAh, Table 1) during discharge clearly indicates that the average molecular weight of products formed on the cathode is higher than that of Li2O2. We show later that, during discharge, both the e/O2 consumed and the cathode weight gain are consistent with solvent decomposition to form solid carbonate deposits on the cathode, especially Li alkyl carbonates.

The 1PC/2DME-based cell (Figure 1c) exhibited two distinct voltage plateaus upon discharge, with 2.60 e/O2 consumed during the initial 2.6 V plateau and 2.05 e/O2 consumed in the short discharge plateau at ∼2.3 V. The initial plateau’s O2 consumption was similar to that observed in the pure carbonate-based cell, while the 2.3 V plateau exhibited oxygen consumption similar to that of the DME-based cell. In addition, the cathode weight gain observed in the full discharge for the 1PC/2DME-based cell was ∼2.3 mg/mAh (Table 1). Similar to the pure carbonate-based cell, both facts suggest that the discharge products are not pure Li2O2, but predominantly include higher molecular weight species (e.g., Li alkyl carbonates). However, the 2.05 e/O2 consumed during the 2.3 V plateau suggests that Li2O2 was formed during this stage and not Li2O (which is produced in a 4 e/O2 process), as has been suggested as a potential discharge product by others.3,4,13 XRDs for pure DME, 1PC/2DME, and 1EC/1DMC-based electrolyte discharges are presented in Figure 2. All cells were discharged to 2 V at a current density of 0.09 mA/cm2. The DME-based cell had a total capacity of 8.2 mAh, compared to only ∼0.7 mAh for the PC/DME and EC/DMC-based cells. The discrepancy in these cells’ capacities is currently under investigation. The broad peaks at 2θ = 27 and 47° arose from small amounts of crystallinity in the cell’s mylar window. Graphitic carbon (from the P50 cathode) and Li2O2 crystalline phases were observed in the DME-based discharged cathode (Figure 2a). No other crystalline phase was observed in the DME-based cathode. Scherrer equation peak-broadening analysis on the Li2O2 peak centered at around 2θ ≈ 35° in Figure 2a indicates that the Li2O2 typical crystallite dimension is approximately 15 nm. This crystallite dimension remains constant regardless of the depth of discharge (i.e., a discharge to 2 mAh yielded the same crystallite dimension as the 8 mAh discharge). Only graphitic carbon was observed in cathodes removed from cells employing EC/DMC-based and PC/ DME-based electrolytes, which indicates that no measurable crystalline product was produced during their discharge. Raman spectra for the same cathodes studied in Figure 2 are shown in Figure 3. The neat P50 carbon paper Raman spectrum is also included here for reference. As in the XRD, Li2O2 (strongest 1163

dx.doi.org/10.1021/jz200352v |J. Phys. Chem. Lett. 2011, 2, 1161–1166

The Journal of Physical Chemistry Letters peaks at 790 and 250 cm1) was observed as the major discharge product in the DME-based cathode. In addition, Raman and XRD of DME-based cathodes discharged down to a potential of 1 V showed only Li2O2 as a discharge product, with no Li2O formation observed. Therefore, no evidence was found to indicate that reactions 2 and 3 contribute significantly to the discharge chemistry. Raman spectroscopy on the 1PC/2DME and 1EC/1DMCbased discharged cathodes shows Li2CO3 and possibly Li alkyl carbonates as the dominant products. The peaks observed at around 3000 cm1 correspond to CH bond stretching and are similar to peaks observed in a Li methyl carbonate Raman spectrum (Figure S3 in the Supporting Information). No Li2O2, Li2O, or LiOH were observed in either of these cases. A Baker test strip analysis (a qualitative test for the presence of peroxides; please see the Methods Section for a description of this analysis) of the discharged cathode does, however, indicate the presence of at least trace amounts of peroxide in both carbonatebased discharged cathodes. Of course, a similar positive Baker strip test was found when using the DME-based electrolyte. DEMS Analysis of Gas Evolution during Charging. Figure 4 presents measurements of gas evolution during charging for the cells studied in Figure 1. As expected, the DME-based cell (Figure 4a) discharged under 18O2 predominantly evolved 18O2 upon charging, indicating that the O2 evolved was derived exclusively from O2 consumed by the cell during discharge. The DME-based cell exhibited four distinct gas evolution processes (labeled IIV in Figures 1a and 4a), all of which correlate with changes in the cell voltage. As the cell potential initially increased almost linearly from 3.1 to 4 V, 18O2 was evolved at its highest rate (stage I). The potential then increased rapidly from 4 to 4.5 V (stage II), during which the 18O2 evolution rate remained low. 18O2 evolution during the high potential plateau (4.5 V, stage III) remained essentially constant until near the end of the charge, where the onset of CO2 evolution was correlated with a slight potential increase (4.54.6 V, stage IV). These gas evolution regions are more clearly visualized in Figure 5, where a discharged DME-based cell is subjected to a linear oxidative potential scan. Four distinct peaks (at 3.2, 3.4, 3.8, and >4.5 V) are observed in this

LETTER

potential scan, with the first three evolving exclusively O2 and the last coincident to O2 and CO2 evolution. Because Li2O2 was the only observed discharge product in a DME-based cell, we speculate that the oxidative potential scan reveals a wide distribution of Li2O2 decomposition overpotentials. This overpotential distribution is the origin of the initial linear increase in cell potential (stage I) and the high charging voltage during stages III and IV of Figure 1a. In other words, O2 is evolved initially at low voltages in Figure 4a, but the cell voltage increases as discharge species corresponding to each peak in the oxidative potential scan are depleted. Determining the source of the overpotential distribution in Figure 5 is an important and ongoing investigation in our laboratory. The total O2 evolved in the DME-based cell (Figure 4a), as determined by integrating the O2 generation curve, corresponded to only 60% of O2 evolution expected from a 1 mAh discharge (11.7 μmol of O2 was produced, and 18.2 μmol of O2 was consumed during the 1 mAh discharge). In other words, an average of 3.2 e

Figure 5. Gas evolution and current versus cell voltage during a 0.075 mV/s linear oxidative potential scan of a discharged DME-based cell. The cell was discharged at 0.1 mA (0.09 mA/cm2) for 10 h under 16 O2 prior to the scan. The scan was performed under Ar.

Figure 4. Isotopically labeled O2 and CO2 gas evolution during charging of (a) DME-based, (b) 1:1 (v:v) EC/DMC-based, and (c) 1:2 (v:v) PC/DMEbased cells studied in Figure 1. 1164

dx.doi.org/10.1021/jz200352v |J. Phys. Chem. Lett. 2011, 2, 1161–1166

The Journal of Physical Chemistry Letters was consumed during charging for every O2 molecule evolved (from reaction 1, the ideal e/O2 ratio for Li2O2 decomposition is 2). The cause of this high e/O2 ratio is currently under investigation. However, Li2O2 fully disappeared during cell charging as no Li2O2 was observed on charged cathodes using Raman spectroscopy, XRD, and Baker test strip analysis. A possible explanation of the high e/O2 ratio could be a slow thermal chemical or electrochemical reaction between Li2O2 and DME that consumes a portion of the Li2O2 prior to or during cell charge. The CO2 isotopes evolved at the end of the charge were in a 45:40:15 molar ratio for C18O2/C16O18O/C16O2. The large molar concentration of 18O present in the CO2 isotopes indicates that Li218O2 formed during discharge participates in DME oxidation at high voltages (4.54.6 V) and that under these conditions, the electrochemical stability of the electrolyte in the presence of Li2O2 is poor (Figure S4 (Supporting Information) shows that CO2 evolution during charge of a nondischarged DME-based cell occurs only at potentials > 4.75 V). The O2/CO2 total molar generation ratio for the entire charge cycle was approximately 7. H2 was also evolved in small quantities for E > 4.5 V (approximately 50% of the total CO2 generation); however, no other gas (from molar mass 180) besides O2, CO2, and H2 was evolved upon charging. During charging, the 1EC/1DMC-based cell (Figures 1b and 4b) exhibited a very brief ∼3.2 V plateau and then rose to high potentials (4.54.6 V), similar to those observed in previous studies where carbonate electrolytes were employed and no cathode catalysts were used.7,11 Although a small amount of 18O2 was initially evolved (coincident with the ∼3.2 V plateau), CO2 was by far the dominant species formed during charging; the total O2/CO2 molar generation ratio during cell charge was 1/27. Total O2 evolution during charging was 0.5 μmol for the 1EC/1DMC-based cell, whereas 9.4 μmol of O2 were consumed during its discharge. The isotopic composition of the CO2 evolved was C18O2/C16O18O/C16O2 = 5:70:25. These results clearly indicate that electrolyte decomposition occurs, as the only available source of 16O is the electrolyte. The low amount of O2 evolution, coupled with the high concentration of the mixed carbon dioxide isotope (C16O18O), also implies that the discharge chemistry involves reduction of carbonates by electrochemical products formed from 18O2 (solvent reduction is discussed more completely in the description of Figure S2 in the Supporting Information). Although electrolyte decomposition was the primary source of CO2 during cell charging, a small amount of cathode carbon oxidation contributed to CO2 evolution (