Predicting Autoxidation Stability of Ether- and Amide-Based Electrolyte

Article pubs.acs.org/JPCA

Predicting Autoxidation Stability of Ether- and Amide-Based Electrolyte Solvents for Li−Air Batteries Vyacheslav S. Bryantsev*,† and Francesco Faglioni†,‡ †

Liox Power, Incorporated, 129 N. Hill Avenue, Suite 103, Pasadena, California 91106, United States Dipartimento di Chimica, Università di Modena e Reggio Emilia, Via Campi 183, 41100 Modena, Italy

‡

S Supporting Information *

ABSTRACT: Finding suitable solvents remains one of the most elusive challenges in rechargeable, nonaqueous Li−air battery technology. Although ether and amides are identified as stable classes of aprotic solvents against nucleophilic attack by superoxide, many of them are prone to autoxidation under oxygen atmosphere. In this work, we use density functional theory calculations coupled with an implicit solvent model to investigate the autoxidative stability of ether- and N,Ndialkylamide-based solvents. The change in the activation free energy for the C−H bond cleavage by O2 is consistent with the extent of peroxide production for each class of solvent. Conversely, the thermodynamic stability alone is not sufficient to account for the observed variation in solvent reactivity toward O2. A detailed understanding of the factors influencing the autoxidative stability provides several strategies for designing molecules with enhanced air/O2 stability, comparable or superior to that of structurally related hydrocarbons. The mechanism of superoxide-mediated oxidation of hydroperoxides derived from ethers and amides is presented. The degradation mechanism accounts for the primary decomposition products (esters and carboxylates) observed in the Li−air battery with etherbased electrolytes. The identification of solvents having resistance to autoxidation is critical for the development of rechargeable Li−air batteries with long cycle life.

1. INTRODUCTION One prerequisite for the development of organic electrolyte rechargeable Li−air batteries1−15 with high specific energy and good capacity retention over long cycle life is the identification of stable electrolyte solvents that are not consumed during the charge−discharge process. The identification of solvents having long-term stability in the operating environment of the Li−air battery remains an elusive challenge.13−31 There is now strong experimental evidence that most commonly used organic carbonate solvents react with the discharge components of the Li−air battery to form lithium carbonates and alkyl carbonates, which are decomposed on charge to give CO2 rather than O2.16−21 The irreversible loss of the solvent in each cycle causes rapid capacity loss and early cell failure. In our previous work, we used theoretical calculations to unravel the initial mechanism of superoxide-induced decomposition of organic carbonates.28,29 We found that the nucleophilic attack of superoxide on the O-alkyl carbon is a common initial mechanism of decomposition of organic carbonates, alkyl carboxylates, and alkyl esters of sulfonic, phosphinic, phosphonic, and phosphoric acids. The powerful nucleophilicity of superoxide in aprotic media sets tight limits as to which classes of solvents are suitable for Li−air batteries. Based on the combined computational and experimental evidence, ether- and N-alkyl-substituted lactams and amides © XXXX American Chemical Society

have been identified as among the most stable classes of aprotic solvents toward nucleophilic substitution by superoxide.29 Indeed, recent studies by McCloskey et al.20 and Freunberger et al.27 have shown that Li2O2 is a primary product formed on the first discharge in the Li−air battery with ether-based electrolytes. However, low current efficiency for O2 release is found during the charging process,20 and a significant amount of electrolyte decomposition is observed on cycling.27 These factors have led to the conclusion that ether-based solvents are unsuitable for long-term cycling of the Li−air battery.20,27 Similarly, evidence for solvent decomposition has been observed in amides, though at a reduced rate compared to ethers.31 It is well-known that ethers slowly decompose under ambient conditions in the presence of oxygen.32−34 The study of autoxidation of organic materials generally and ethers specifically is a well-established field of research.32−37 Indeed, explosive peroxides are known to form spontaneously from diisopropyl ether (iPrOiPr), tetrahydrofuran (THF), and diethyl ether (EtOEt) upon exposure to air at room temperature.32−34 Initiation of autoxidation by molecular Received: February 15, 2012 Revised: May 11, 2012

A

dx.doi.org/10.1021/jp301537w | J. Phys. Chem. A XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry A

Article

Table 1. Comparison of Standard Gibbs Free Energies of Propagation and Degenerate Chain Branching in the Autoxidation of Hydrocarbons, Ethers, and Amides in the Gas Phase (kcal/mol) R· + O2 → ROO· a

R·

B3LYP

CH3Ċ HCH3 CH3CH2OĊ HCH3 CH3CON(CH3)Ċ H2 CH3CON(C2H5)Ċ HCH3

−17.81 −18.70 −10.32 −11.52

CCSD(T)

ROO· + THF → ROOH + THF_H b

B3LYP

−25.66 −27.07 −18.51

10.82 10.25 8.24 9.18

a

CCSD(T) 8.64 8.09 6.21

b

ROOH → RO· + ·OH B3LYPa

CCSD(T)b

25.53 22.62 25.32 25.01

35.48 32.93 35.58

a

B3LYP/6-311++G**. bExtrapolated CCSD(T)/CBS energies (eq 2). ZPE and thermal correction were calculated at the B3LYP/6-311++G** level.

oxygen is an endothermic process requiring a considerable activation energy, which accounts for the relatively slow rate under ambient conditions. 32−34 However, the rate of autoxidation during the operation of the battery could be more significant, because hydroperoxides formed in the autoxidation process are expected to be highly reactive with superoxide formed upon discharge,38,39 which could generate more reactive species40−42 capable of solvent degradation. Likewise, N-methylpyrrolidone (NMP) and related lactams containing an α-hydrogen atom in the ring alkyl group adjacent to amide nitrogen are susceptible to slow oxidation by molecular oxygen under mild conditions, yielding hydroperoxides and cyclic imides as major products.43,44 Understanding reactivity trends and determining the factors influencing autoxidation stability would provide important guidelines for the design of functional ethers and amides that are less likely to produce peroxides and imides. Herein we report the computed free energy barriers (ΔGlact) and reaction free energies (ΔGlr) for hydrogen atom abstraction by molecular oxygen from a series of ethers and N-alkyl-substituted lactams and amides. Good correspondence for each class of solvents is obtained between the calculated activation free energies and the extent of peroxide production through uncatalyzed autoxidation. This serves as a validation of first-principles theoretical calculations as an efficient screening method to identify autoxidation-resistant solvents for Li−air battery applications. Furthermore, we investigate whether the slow process of autoxidation could affect the stability of ether, lactam, and amide solvents in Li−air cells and whether autoxidation may be accelerated under conditions of operation.

Figure 1. Variation in the MP2 reaction energy for hydrogen atom abstraction from tetrahydrofuran by molecular oxygen, RH + O2 → R· + HOO·, with basis set size (aug-cc-pVnZ, n = 2−5). An extrapolated complete basis set limit (CBS) is indicated as a horizontal line.

basis set size on the MP2 reaction energy for hydrogen atom abstraction by molecular oxygen from THF. To estimate the MP2 CBS limit, we used an extrapolation scheme based on a polynomial function of inverse powers of 4 and 555,56 ΔE(n) = ΔECBS + B /(n + 1)4 + C /(n + 1)5

(1)

where n = 3, 4, 5 for n = T, Q, and 5 in aug-cc-pVnZ, respectively, and ΔECBS, B, and C are the fitting parameters. The CCSD(T)/CBS reaction energies and barrier heights were estimated by combining the MP2/CBS energies with CCSD(T)/aug-cc-pVDZ corrections5758

2. COMPUTATIONAL METHODS Electronic structure calculations were carried out using the Jaguar 7.545 and the NWChem 5.146 programs. We used the B3LYP47,48 flavor of density functional theory (DFT) in the 6311++G** basis set throughout the study (Tables 1−4). Control calculations for hydrogen abstraction reactions show that using the B3LYP functional with a more extended aug-ccpVTZ basis set does not lead to improved accuracy when compared to high-level coupled-cluster calculations. As a benchmark for determining the accuracy of the B3LYP method, we employed calculations at the second-order Möller− Plesset perturbation theory (MP2)46,49 in the complete basis set (CBS) limit and coupled-cluster theory46,50−52 with single, double, and perturbative triple excitations CCSD(T) in the aug-cc-pVDZ basis set. MP2 and CCSD(T) calculations used unrestricted reference wave functions, keeping core electrons frozen. For the basis set expansion in our MP2 calculations, we used a family of augmented correlation-consistent basis sets (aug-cc-pVnZ, n = D, T, Q, 5).53,54 The basis set for geometry optimization was aug-cc-pVDZ. Figure 1 illustrates the effect of

ΔE(CCSD(T)/CBS) = ΔE(MP2/CBS) + δCCSD(T) (2a)

δCCSD(T) = ΔE(CCSD(T)/aug‐cc‐pVDZ) − ΔE(MP2/aug‐cc‐pVDZ)

(2b)

Transition-state search was performed using a standard quasi-Newton method in Jaguar,45 starting from the partially optimized geometry along the chosen reaction coordinate and the precalculated Hessian. The nature of all transition states obtained at the B3LYP/6-311++G** level was confirmed by the presence of a single imaginary frequency in the vibrational spectrum, corresponding to the motion along the reaction coordinate. The standard Gibbs free energy of each species in the gas phase (T = 298 K, P = 1 atm) was calculated using the rigid rotor−harmonic oscillator approximation without scaling. Solvation calculations were carried out using the Poisson− Boltzmann continuum model in Jaguar,45 with the default values of solute atomic radii. The dielectric constants59,60 for B



Article

Table 2. Reaction Free Energies (ΔGlr) and Free Energy Barriers (ΔGlact) for Hydrogen Atom Abstraction from Ether-Based Solvents by Molecular Oxygen Calculated in the Implicit Solvent Reaction Field (kcal/mol) ΔGlr ε (solvent)a

solvent 1 THF 2 tBuOEt 3 EtOEt 4 iPrOPr 5 CPME 6 EtOMe 7 THP 8 DME 9 tBuOMe 10 propanee 11 2DME−Li+ 12 PhOMe 13 CF3CH2OMe 14 7-OBCH 15 tBuOtBu 16 HF−THP

7.58 4.34 4.34 3.88 4.76 4.34 5.68 7.20 4.34 1.84 7.20 4.45 4.34 4.34 3.05 5.68

ΔGlactb

B3LYPc

CCSD(T)d

B3LYPc

34.36 34.01 35.32 34.88 35.05 35.78 36.34 34.90 36.85 40.48 37.23 39.83 35.49 40.26 42.73 39.16

35.74

36.53 37.11 37.37 37.43 38.11 38.40 38.40 38.89 39.41 43.60 43.74 43.82 44.15 44.74 47.97 48.79

(EtOEt)

(EtOEt)

(EtOEt) (pentane) (DME) (EtOEt) (EtOEt) (THP)

36.35 37.23 37.28 35.93 37.70 41.88

37.74

CCSD(T)d

43.10

46.99

a

Solvent dielectric constant taken from refs 59 and 60. If dielectric constant is not available, calculations employ parameters for a solvent shown in parentheses. bThe calculated ΔGlact (eq 4) includes a statistical correction term −RT ln(n) to account for the presence of n equivalent H atoms in a molecule. cB3LYP/6-311++G**. dExtrapolated CCSD(T)/CBS energies (eq 2). ZPE, thermal, and solvent correction were calculated at the B3LYP/6-311++G** level. eUsed as hydrocarbon reference.

Here, ΔGo→* = RT ln(24.46) = 1.89 kcal/mol (T = 298.15 K) is a standard-state correction,61,62 associated with moving a solute from a standard-state gas-phase concentration of 1 atm to a standard-state solution-phase concentration of 1 mol/L, and RT ln([RH]) is a free energy change of 1 mol of RH ideal gas from liquid-state concentration to 1 mol/L standard state in solution.62 The latter conversion term needs to be included if the pure solvent RH(l) is adopted as the reference state for the solvent in the lower leg of the thermodynamic cycle.62 Similarly, ΔGlact in the liquid phase can be written as a sum of the gas-phase activation free energy (ΔGoact,g), the difference in the solvation free energies of the transition state (RH···O2) and the reactant, and the correction term (−RT ln([RH]))

tetrahydrofuran (THF), diethyl ether (EtOEt), diisopropyl ether (iPrOiPr), cyclopentyl methyl ether (CPME), tetrahydropyran (THP), dimethoxyethane (DME), anisole (PhOMe), di-tert-butyl ether (tBuOtBu), N-methylpyrrolidone (NMP), N,N-dimethylacetamide (DMA), N,N-diethylacetamide (DEA), and N,N-dimethylformamide (DMF) are given in Tables 2 and 4, and probe radii are defined from their molecular weight and liquid density at room temperature as follows: 2.524 (THF), 2.741 (EtOEt), 3.036 (iPrOiPr), 2.848 (CPME), 2.690 (THP), 2.749 (DME), 2.788 (PhOMe), 3.240 (tBuOtBu), 2.674 (NMP), 2.643 (DMA), 2.934 (DEA), and 2.485 (DMF). Reaction free energies for hydrogen atom abstraction from the solvent by molecular oxygen in the condensed phase are calculated using the thermodynamic cycle shown in Scheme 1.

ΔGl act = ΔGoact,g + ΔG*solv (RH···O2 ) − ΔG*solv (RH)

Scheme 1. Thermodynamic Cycle Used in the Calculation of the Liquid-Phase Reaction Free Energies for Hydrogen Atom Abstraction from the Solvent by Molecular Oxygen

− RT ln([RH])

3. RESULTS AND DISCUSSION In this work, we combine density functional theory (B3LYP) with the Poisson−Boltzmann continuum solvent model to predict the reactivity of a variety of ether- and amide-based solvents with molecular oxygen (Scheme 2). The basic mechanism of the autoxidation of organic solvents in the liquid phase with degenerate chain branching35−37,63,64 is represented in Scheme 3. Degenerate branching occurs when hydroperoxide product formed in reaction III decomposes into free radicals available for initiating new chain reactions. The chain initiation step I is strongly endothermic and requires high activation energy (Tables 2 and 4).34,65,66 In contrast, the propagation step (reactions II and III) is exothermic (Table 1), making the overall reaction thermodynamically favorable. Therefore, the oxidation process is often characterized by an induction period43 after which the reaction proceeds more readily. A relevant example is the oxidation of ethers under the usual storage conditions.32−34

From Scheme 1, ΔGlr in the liquid phase can be expressed as a sum of the gas-phase free energy of reaction (ΔGor,g), the difference in the solvation free energies of the products and reactants (ΔΔG*solv), and the standard state (ΔGo→*) and concentration (−RT ln([RH])) correction terms61,62 ΔGl r = ΔGo r,g + ΔΔG*solv + ΔGo →* − RT ln([RH]) (3a)

ΔΔG*solv = ΔG*solv (R·) + ΔG*solv (HOO·) − ΔG*solv (RH)

(4)

(3b) C



Article

of Gibbs free energies of reactions II, III, and V, which are summarized in Table 1. We find very similar reaction free energies for the addition of O2 to the secondary hydrocarbon radical and secondary ether radical, as well as for the subsequent hydrogen atom abstraction from THF by the corresponding peroxy radicals. This is consistent with the fact that peroxy radicals derived from hydrocarbons show about the same reactivity toward hydrogen atom abstraction as structurally related peroxy radicals derived from ethers.67,68 On the other hand, the RO−OH bond dissociation energy is significantly lower in ethers compared to that in structurally related hydrocarbons and amides, which could account for a higher reactivity of ether hydroperoxides under thermooxidative conditions.67−73 This result is in agreement with previous calculations.34,69 Finally, the thermodynamics of the propagation step is significantly less favorable for amides than for ethers and hydrocarbons. This is consistent with relatively short kinetic chain lengths observed in the autoxidation of amides at elevated temperatures.71−73 Due to a complex nature of the autoxidation process that may involve dozens of intermediates for each solvent under investigation,34,69 we do not attempt to develop a comprehensive quantitative kinetic model of oxidative degradation. In this study we focus on comparing ab initio results for the initial rate-limiting step I of chain initiation by molecular oxygen, which provides a simple theoretical framework for computational screening of solvents with improved autoxidative characteristics. Computational results for ethers and amides, including comparison with available experimental data, are discussed in sections 3.1 and 3.3, respectively, while sections 3.2 and 3.4 examine the effect of autoxidation products formed in ether- and amide-based solvents on the discharge chemistry of Li−air cells. 3.1. Autoxidation of Ethers. Computed reaction free energies (ΔGlr) and free energy barriers (ΔGlact) for the hydrogen atom abstraction reaction of molecular oxygen with 15 ether-based solvents in the liquid phase are summarized in Table 2. Only the lowest value of ΔGlact is listed in Table 2, as determined via a thorough systematic search of the plausible low-lying transition states. Figure 2 exemplifies transition states

Scheme 2. Ether-, Lactam-, and Amide-Based Solvents Investigated in This Worka

a

Each class is arranged in the order of increasing activation free energy for hydrogen atom abstraction by molecular oxygen (ΔGlact, Tables 2 and 4). Additionally, propane is shown as hydrocarbon reference. The most reactive position is marked by asterisk.

Scheme 3. General Mechanism of the Liquid-Phase Autoxidation of Organic Solvents with Degenerate Chain Branching Figure 2. Three distinct transition state structures (bond lengths are in Å) and activation free energies calculated in the implicit solvent reaction field (kcal/mol) for hydrogen atom abstraction from ethyl methyl ether by molecular oxygen.

structures and their relative energies for the reaction of O2 with ethyl methyl ether. The value of 38.40 reported in Table 2 refers to TS1 in Figure 2. There have been few studies investigating the autoxidative stability of ethers in air under normal storage conditions.32,33,74 In one such comparative study,32 several solvents were exposed to air at room temperature for 1 month without a stabilizing agent. A significant amount of peroxide was observed in iPrOiPr and THF (hundreds of parts per million), while the extent of peroxide production in CPME and t-BuOMe was

Further insight into the autoxidation activity of hydrocarbons, ethers, and amides can be gained from the comparison D



Article

in ΔGlr (and ΔGlact) is as follows: MeOEt > EtOEt > tBuOEt. This is likely reflective of the increased destabilizing effect of electron-donating alkyl groups on the oxygen nonbonding orbitals in the parent ether. It is evident from Table 2 that nonrigid alkyl ethers containing an α-hydrogen atom (1−9) are less stable toward O2 than structurally related hydrocarbons (10). This is due to the stabilization of ether radicals by the interaction of the partially filled p-orbital of the carbon atom with the neighboring oxygen p-type lone pair.76 The maximum overlap, and hence the maximum radical stabilization, is obtained in planar systems when the p-orbitals are parallel. The higher reactivity of the five-membered cyclic ether THF compared to the sixmembered cyclic ethers THP can be explained on the basis of more favorable overlap between the p-orbitals, since the CH2−O−Ċ (H)−CH2 dihedral angle in the THF radical is 7.7° (a near planar geometry), whereas this angle in the THP radical is 43.8° (a large distortion from planarity). Theoretical predictions are in line with experimental measurements67,68 indicating that THF is about 10 times more reactive toward peroxy radicals than THP. Similar arguments were used to justify the relatively high autoxidative stability of CPME, despite the presence of a tertiary hydrogen.32 Finally, higher activation energy required for the C−H cleavage in DME than in EtOEt and EtOMe can be attributed to the additional repulsion between the O2 lone pairs and the lone pairs on the second oxygen atom in the diether. The analysis of the factors leading to destabilization of the transition state for the C−H bond cleavage by O2 provides several strategies for designing molecules that offer increased autoxidative stability, possibly comparable or superior to that of structurally related hydrocarbons. Representative examples to illustrate these strategies are given in Table 2 (ethers 11−16). Apparently, with all reactive α-hydrogen atoms substituted by methyl groups, the cleavage of the β−C−H bond in tBuOtBu is difficult. Alternatively, the autoxidative stability of ethers can be improved through fluorination at β-carbon positions. Calculations reveal that the activation energy for the reaction of O2 with CF3CH2OMe and HF-THP increases by 5.8 and 10.4 kcal/mol, respectively, compared to the nonfluorinated analogues. We ascribe this effect to the unfavorable electrostatic interaction between the oxygen and fluorine atoms in the transition state.77 Another strategy for achieving enhanced air/ O2 stability is to conformationally constrain a radical into an unfavorable pyramidal geometry. As we discussed earlier, a strong nonplanarity also prevents favorable overlap between the half-occupied p-orbital of the carbon atom and the oxygen nonbonding p-orbital. As a result, a typically reactive C−H bond at the tertiary α-carbon atom is completely deactivated in the bicyclic 7-OBCH ether (ΔGlr = 49.78 kcal/mol). Remarkably, the β-carbon site in 7-OBCH is more susceptible to autoxidation (ΔGlr = 41.60 kcal/mol). Finally, if an ether oxygen lone pair interacts strongly with a π-system (PhOMe) or is involved in a donor−acceptor interaction with Li+ cation (2DME−Li+), it loses the ability to stabilize the radical center, thereby increasing resistance against autoxidation. Thus, based on model calculations for a complex of Li+ cation with two monoglyme molecules (2DME−Li+), longer glyme−Li salt electrolytes composed of equimolar amounts of a glyme and a Li salt78,79 are expected to have an improved autoxidative stability over glyme−Li salt mixtures containing an excess of glymes.

relatively low ( 34.88 (iPrOiPr) and 36.85 (tBuOMe) > 34.01 (tBuOEt). By varying the functional group that is positioned away from the radical center, the trend E



Article

LiClO4/CH3CN electrolyte is k = 2.9 × 10−3 s−1, which, using the classical reaction kinetic theory, can be converted to a ratelimiting barrier of ΔGact = 16.6 kcal/mol. We suggest that the decomposition of ether-based electrolytes observed in Li−air batteries arises from the initial oxidation by O2 to form ether hydroperoxides at T = 25−30 °C (eqs I−VI).68,69 Solvent oxidation during charge20,27 and subsequent oxygenation during the following discharge may also significantly contribute to hydroperoxide production. These hydroperoxides are not stable in the presence of superoxide and easily decompose into esters, formate, and other oxygenated products during discharge, as outlined in Scheme 4 and discussed in detail below. Furthermore, lithium

3.2. Reactivity of Ethers and Ether Hydroperoxides with Superoxide. Theoretical calculations predicted high stability of ethers (DME) against nucleophilic substitution by the superoxide anion radical (O2•−).29 However, recent experimental studies by Freunberger et al.27 have indicated that ether-based electrolytes in the Li−air battery exhibit significant decomposition on cycling. A mechanistic pathway of solvent decomposition was proposed, suggesting that superoxide is directly engaged in the hydrogen atom abstraction from ethers. To test this hypothesis, we computed reaction free energy profiles for α-hydrogen atom abstraction from EtOMe and DME to O2•− and LiO2 (Table 3). Treatment of solvation Table 3. Reaction Free Energies (ΔGor) and Free Energy Barriers (ΔGoact) for α-Hydrogen Abstraction from EtOMe and DME by Superoxide Calculated Using a Dielectric Continuum Solvent Model and a Mixed Cluster/Continuum Representation of the Solvent by Inclusion of Three Additional Solvent Molecules (kcal/mol)a reactant complex

ΔGor

ΔGoact

1EtOMe−O2− 1DME−O2− 4EtOMe−O2−b 4EtOMe−LiO2

27.79 26.61 29.73 25.83

29.72 30.73 32.74 29.88

Scheme 4. Proposed Mechanism of Superoxide-Mediated Oxidation of 1-Hydroperoxy-1,2-dimethoxyethane

a Obtained at the B3LYP/6-311++G** level. bThe structural model of the reaction site for 4EtOMe−O2− is exemplified in Figure 4.

effects using a dielectric continuum model is augmented by explicit inclusion of three additional solvent molecules to construct a full first solvation shell around the O2•− ion and LiO2. A structural model of the solvated O2•− species is illustrated in Figure 4. The results, independent of the solvation model used (with and without explicit solvent), indicate that direct hydrogen atom abstraction by O2•− and LiO2 is unlikely as it requires an activation energy of at least ∼30.0 kcal/mol. This reaction has too high a barrier to be competitive with LiO2 disproportionation to Li2O2 + O2 and direct reduction to Li2O2 in the Li−air battery. For comparison, the reported26 first-order rate constant for the transformation of LiO2 to Li2O2 in 0.1 M

carboxylates formed upon discharge can only be irreversibly oxidized to CO2 upon charge.27 This reaction scheme could account for the primary decomposition products observed in the Li−air battery with ether-based electrolytes.27 It should be noted that formates and esters can also be produced during the oxidative degradation in air at elevated69,70 (T = 50 °C) and high80,81 (T = 150 °C) temperatures. However, this is unlikely to be a significant pathway for the production of esters and

Figure 4. Reaction free energy profile for α-hydrogen abstraction from EtOMe by superoxide calculated in the presence of three additional solvent molecules and the dielectric field of a solvent continuum (kcal/mol). B3LYP optimized bond lengths (Å) are shown. F



Article

carboxylic acids at 25−30 °C, since autoxidation of ethers at this temperature gives hydroperoxides in almost 100% yield.68 The proposed mechanism of superoxide-mediated oxidation of ether hydroperoxides is depicted in Scheme 4. Superoxide acts as a strong base in aprotic solvents82,83 and will likely react with hydroperoxide via initial proton transfer (eq VII).38,39 A subsequent one-electron reduction of the generated HOO· radical by O2•− to give HOO−, followed by the second proton transfer from ROOH to HOO− (eqs VIII, IX) is a well-known mechanism of superoxide ion disproportionation in weakly acidic media.83 There is important experimental evidence39,40 that H2O2 formed in the disproportionation reaction (eq IX) does not accumulate in aprotic solvents but presumably reacts further with HOO− to generate superoxide, water, and the highly reactive hydroxyl radical (eq X). The ·OH radical could serve as radical initiator for solvent autoxidation, possibly explaining the larger extent of electrolyte degradation during the operation of the Li−air battery. Peroxy anions ROO− containing easily oxidizable α-hydrogen atoms undergo facile superoxide-induced oxidation at room temperature.41,42,84 This process constitutes a major pathway for ester production (eq XI). However, esters themselves are not stable against superoxide as they are susceptible to O−alkyl cleavage and nucleophilic displacement of the carboxylate ion (eq XII).29,85,86 Further oxidative decomposition of the reactive intermediates yields formates and other carboxylates41 (eqs XIII−XIV) that precipitate as insoluble salts in the presence of Li+ ions. It is now generally accepted that tertiary hydroperoxides act solely as a proton source toward O2•− in nonreactive organic solvents. This is supported by several studies38,39 indicating that the reaction between t-BuOOH and O2•− in toluene, pyridine, and benzene produces t-BuOO− as a stable product. In contrast, primary and secondary ROO− anions in the presence of O2/O2•− can be easily oxidized to the corresponding ketones and carboxylic acids (eq XI).41,42,84 Alternatively, the reaction between the primary and secondary hydroperoxides and O2•− may proceed via an electron-transfer pathway. A facile electron transfer from O2•− to MeOOH and EtOOH has been recently observed in the gas phase.87 This experimental observation is corroborated by the present DFT calculations (Figure S1, Supporting Information). It is not known if the assumed electron-transfer process between ether hydroperoxides and O2•− could also occur in solution. Model calculations with explicit inclusion of solvent molecules (Figure S2, Supporting Information) suggest that the calculated activation energy for the rate-limiting step is not sufficiently high (below 20 kcal/ mol) to rule out this possibility. 3.3. Autoxidation of N-Alkyl Lactams and N,N-Dialkyl Amides. Table 4 summarizes the thermodynamic parameters (ΔGlr) and activation energies (ΔGlact) for hydrogen atom abstraction from 14 aprotic lactam and amide solvents by molecular oxygen. In several instances we report results for several possible reaction sites within a given molecule in order to provide a mechanistic understanding of the relevant reaction pathways of solvent autoxidation. Since no experimental data are available for solvent degradation at ambient conditions, our theoretical predictions will be compared with the available experimental data on the uncatalyzed autoxidation of amidebased solvents at 75−131 °C.43,44,71−73 The oxidation of alkylamides and lactams is known to occur primarily at the N-alkyl carbon. 43,71 There is general experimental evidence of the increased reactivity of ring CH

Table 4. Reaction Free Energies (ΔGlr) and Free Energy Barriers (ΔGlact) for Hydrogen Atom Abstraction from Lactams and Amides by Molecular Oxygen Calculated in the Implicit Solvent Reaction Field at the B3LYP/6-311++G** Level (kcal/mol) solvent 17 18 19 20

DMI 5-Me−NMP NMPI NMP

21 5,5-diMe− NMP 22 NAP 23 1-Ac−Pyr 24 2-MeO− DMA 25 DMA 26 DEA 27 DMF 28 DMTFA 29 DIPA 30 NMDA

ε (solvent)b

ΔGlr

ΔGlactc

−CH2−N< −CH(Me)−N< −CH2−N< −CH2−N< CH3−N< −CH2−C(O)− CH3−N

Predicting Autoxidation Stability of Ether- and Amide-Based Electrolyte

Recommend Documents