Article

pubs.acs.org/JPCC

Oxygen Reduction Reactions in Ionic Liquids and the Formulation of a General ORR Mechanism for Li−Air Batteries † † † † ‡ Chris J. Allen, Jaehee Hwang, Roger Kautz, Sanjeev Mukerjee, Edward J. Plichta, ‡ † Mary A. Hendrickson, and K. M. Abraham*, † Department of and Chemical Biology, Northeastern University Center for Renewable Energy Technology (NUCRET), Northeastern University, Boston, Massachusetts 02115, United States ‡ Power Division, U.S. Army RDECOM CERDEC CP&I, RDER-CCP, 5100 Magazine Road, Aberdeen Proving Ground, Maryland 21005, United States

ABSTRACT: Oxygen reduction and evolution reactions (ORRs and OERs) have been studied in ionic liquids containing singly charged cations having a range of ionic radii, or charge densities. Specifically, ORR and OER mechanisms were studied using cyclic and rotating disk electrode voltammetry in the neat ionic liquids (ILs), 1-ethyl- 3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMITFSI) and 1-methyl-1-butyl-pyrrolidinium bis- fl (tri ouromethanesulfonyl)imide (PYR14TFSI), and in their solutions containing LiTFSI, NaPF6, KPF6, and tetrabutylam- fl monium hexa uorophosphate (TBAPF6). A strong correlation was found between the ORR products and the ionic charge density, including those of the ionic liquids. The observed trend is explained in terms of the Lewis acidity of the cation present in the electrolyte using an acidity scale created from 13C − 13  NMR chemical shifts and spin lattice relaxation (T1) times of C O in solutions of these charged in propylene carbonate (PC). The ionic liquids lie in a continuum of a cascading Lewis acidity scale with respect to the charge density of alkali metal, IL, and TBA cations with the result that the ORR products in ionic liquids and in organic electrolytes containing any conducting cations can be predicted on the basis of a general theory based on the hard soft (HSAB) concept.

■ INTRODUCTION electrolytes5,6 judicially selected on the basis of their Lewis − fi Lithium- (Li-ion) batteries have an undeniable influence acid base properties as de ned by Guttmann donor and over our daily lives. Vested in this battery technology are acceptor numbers. The Guttmann donor number (DN) mobile electronics, load-leveling infrastructures, and electric measures the electron-donating capacity of the to propulsion vehicles. Despite their ubiquity, limits on energy form complexes with Lewis such as Li+. density and the high cost of commercialized Li-ion batteries Ionic liquids as a class of electrolytes for Li batteries offer have accelerated efforts for alternative rechargeable battery several potential advantages over traditional nonaqueous − fi systems, such as the nonaqueous Li O2 battery, rst realized organic . Besides a negligible vapor pressure, high 15 years ago.1 The 5280 Wh/kg theoretical energy density of ionic conductivity, and nonflammability, they can be designed − − ’ Li O2 is 7 8 times that of today s best Li-ion battery, and it to offer enhanced hydrophobicity and large electrochemical offers a long-term solution to energy independence. stability windows that are highly desirable for their use in the ffi A primary concern facing this power source is the ine cient Li−air battery. In this work, we have studied oxygen reduction rechargeability of insoluble LixOy discharge products that 2 and evolution reactions (ORRs and OERs, respectively) in two accumulate on the O2 electrode. This leads to poor cell ionic liquids (ILs), namely, 1-ethyl-3-methylimidazolium bis- performance, stemming from large cathode impedances and the (trifluoromethylsulfonyl)imide (EMITFSI), and 1-methyl-1- associated voltage gaps between oxygen reduction reactions 3,4 butyl-pyrrolidinium bis(triflouromethanesulfonyl)imide (ORRs) and oxygen evolution reactions (OERs). Our recent work on the mechanism of ORRs in nonaqueous electrolytes (PYR14TFSI). The PYR14 cation is abbreviated as PYR in this has revealed that the properties of the organic solvent play a report. The cations and common anion of these ionic liquids significant role in the nature of the final reduction product are shown in Scheme 1. formed, and the stability of the intermediates through which − the conversion of O2 to Li2O occurs in the discharge of a Li Received: July 6, 2012 O2 cell. We have gained this insight from a detailed study of the Revised: September 2, 2012 ORR intermediates and products in a series of organic Published: September 5, 2012

© 2012 American Chemical Society 20755 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

Scheme 1. Ionic Liquid Cations and Anion Used in This was first measured versus Li/Li+ using a Li foil placed next to Study the reference electrode in the ionic liquid. With this value of the Ag/Ag+ electrode potential versus Li/Li+, the measured electrode potentials from the CV and RDE cells were converted to the Li/Li+ voltage scale reported throughout this article. Propylene carbonate (PC) (anhydrous, 99.7%, Sigma- Aldrich) was used as the solvent for all NMR measurements along with various salts, including lithium bis- (trifluoromethanesulfonyl)imide (LiTFSI) (Purolyte), lithium fl fl hexa uorophosphate (LiPF6) (Purolyte), sodium hexa uoro- phosphate (NaPF6) (98%, Sigma-Aldrich), potassium hexa- fl uorophosphate (KPF6) (98%, Sigma-Aldrich), tetrabutylam- fl monium hexa uorophosphate (TBAPF6) (Fluka), and the two ionic liquids, EMITFSI and PYRTFSI. 13C NMR samples run on a Varian 400 MHz NMR utilized The ORR and OER mechanisms were studied using cyclic external referencing to preserve salt−solvent interactions. The and rotating disk electrode voltammetry (CV and RDE, tube configuration consisted of three components: (1) a respectively) in the neat ionic liquids and in their solutions Wilmad NMR tube (100 mHz, Economy) with 1 M salt containing singly charged cation salts, which included LiTFSI, solutions in propylene carbonate, (2) an inner PYREX capillary NaPF , KPF , and tetrabutylammonium hexafluorophosphate melting point reference tube with d-6 and tetrame- 6 6 fl (TBAPF6). A correlation between the ORR products and the thylsilane (TMS) (1% v/v), and (3) Master ex peroxide-cured Lewis acidity of the cation present in the electrolyte was silicone to hold the capillary melting point tube in place. developed with the help of an acidity scale of the various Electrolyte preparation and tube assembly was done in a cations, including those of the ionic liquids. The acidity scale controlled-atmosphere argon glovebox. Before all NMR was created from 13C NMR chemical shifts and spin−lattice experiments, the instrument was locked with acetone d-6 fi 13  13 relaxation (T1) times, speci cally of the C O moiety, of the placed in the capillary tube. The C chemical shifts of PC and solutions of these salts in propylene carbonate (PC). its solutions were referenced to the 13C peaks of TMS. − 13 Remarkably, we have found that the ionic liquids lie in a Inversion recovery C T1 measurements utilized the same continuum of a falling Lewis acidity scale with respect to the tube configuration and were made on a Varian 500 MHz charge density of alkali metal, ILs, and TBA cations with the instrument. The 90° pulse width was calibrated on each sample result that the ORR products in ionic liquids and in organic before measuring T1, and eight acquisition arrays were used. electrolytes containing any conducting cations can be predicted Two scans were taken for each sample. on the basis of a general theory based on the hard soft acid base (HSAB) concept. These results are presented in detail in this ■ RESULTS AND DISCUSSION paper. The present results complement our recent work that An initial look at the role cations play on ORRs can be gleaned provided initial evidence for the correlation between the charge from the cyclic voltammograms (CVs) obtained with and 5,6 + density of ions and the ORR products. This work provides without 25 mM TBA in O2-saturated EMITFSI and PYRTFSI. experimental evidence for the ability of the HSAB theory to These CVs measured on Au and glassy carbon (GC) working explain the ORR mechanism and reaction products in any electrodes are presented in Figure 1. There is no significant nonaqueous electrolytes and, perhaps, even in aqueous change in the ORR following addition of TBA+. Increasing the electrolytes. TBA+ concentration 20× to 0.5 M made no change to the peak positions. An estimate of formal potentials (Eo′) in both neat ■ EXPERIMENTAL SECTION The ionic liquids 1-ethyl-3-methylimidazolium bis- (trifluoromethylsulfonyl)imide (EMITFSI) and 1-butyl-1- methylpyrrolidinium bis(trifluoromethylsulfonyl)imide (PYRTFSI) were synthesized through an aqueous ion-exchange reaction. The details of the synthesis and purification are described elsewhere.7,8 Structures were confirmed with 1H fi NMR, and a H2O content below 25 ppm was veri ed with Karl Fisher coulometry. Cyclic voltammetry (CV) measurements were performed on an Autolab potentiostat (Eco Chemie B.V.) equipped with a frequency response analyzer for iR correction. Electrochemical O2 half-cell studies were carried out in a controlled-atmosphere argon glovebox outfitted with a high- purity O2 gas source (99.995%). The O2 half-cell consisted of a planar glassy carbon (⌀ = 6 mm) or Au (⌀ = 5 mm) disk working electrode, a Ag/Ag+ reference electrode, and a Pt mesh counter electrode. The reference electrode was composed of 0.1 M silver fl + tri ouromethanesulfonate (AgCF3SO3) (99% , Sigma Aldrich) Figure 1. Cyclic voltammograms on Au and GC electrodes in O2- EMITFSI solution contained in a glass tube with a Ag wire and saturated EMITFSI and PYRTFSI both with (red) and without fi tted with a Vycor frit. In an argon atmosphere, its potential (black) 0.025 M TBAPF6. Scan rate = 100 mV/s.

20756 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

11,12 ILs can be calculated based on peak potentials (Epa + Epc/2). with initial IL ORR reports, including our results presented The 120 mV lower reduction potential of PYR+ relative to in a preliminary communication.7 EMI+ on both surfaces may be explained by the slightly higher The ORR reactions in ILs are those in the equations of + − solvating ability of the EMI cation with O2 , in agreement with Scheme 2, which are changed little with the addition of TBA 9 the observations of Sawyer, who correlated a higher O2 formal cations, and the anodic and cathodic reactions can be reduction potential with improved electrolyte solvating ability represented as depicted in eqs C1 and A1, respectively. − + of O2 . It can be seen from the data in Table 1 that EMI , with Scheme 2. Neat EMITFSI and Their TBA+ Solutions Table 1. Crystallographic Ionic Radii of Cations Used in Both ORR and NMR Studies

cation ionic radius (pm) ref Li+ 60 13 Na+ 96 13 K+ 133 13 EMI+ 239 14 PYR+ 330 15 TBA+ 494 13 Rotating disk electrode (RDE) polarization curves were a smaller ionic radius or charge density (hence, stronger Lewis generated in both ILs to gain further support for the above acidity as explained below), appears to exhibit improved ORR reactions formulated from the CV, and to see if there are of the O reduction product compared to PYR+, 2 differences in the kinetics of the ORRs in the two IL which has a larger ionic radius. electrolytes, which differ in their viscosities and hence mass The reversible processes exhibited in the CVs in Figure 2 transport properties. RDE polarization curves in Figure 3a have been established to be the one-electron reduction of O to 2 reveal some differences between these two ILs during ORRs under forced convective conditions. The ORR onset potential is approximately 200 mV higher in EMITFSI, in agreement with quiescent electrochemistry in Figure 1. The mass transport limiting current region below 1.8 V reveals noticeable differences in ORR transport properties between the two IL cations. EMITFSI is unable to reach a convection controlled steady-state current at either 400 or 2500 rpm at 10 mV/s. The requirements needed to achieve a steady-state RDE current following a potential step is dependent on scan rate and rotation speed along with properties intrinsic to the electro- active species and solvent.16 For example, the Schmidt number ff (Sc), a ratio of solvent kinematic viscosity to O2 di usion coefficient, as seen in Table 2, favors improved mass transport at low values (kinematic viscosity is calculated from the solvents' dynamic viscosity divided by density). However, the Sc of PYRTFSI is an order of magnitude higher than that of Figure 2. Cyclic voltammetry of PYRTFSI on a Au electrode at EMITFSI, which conflicts with limiting current behavior in various scan rates (25−500 mV/s) indicated with arrows. Inset: Figure 3a. An explanation for this may be found from the − Randles Sevcik analysis for n = 1 (red circle), n = 2 (blue square), and higher charge density of EMI+. experimental (triangle). Mass transport limiting currents in the neat EMITFSI were achieved at lower scan rates and rotation speeds (i.e., 5 mV/s, − − O2 from the linear response between the magnitudes of 400 1000 rpm) seen in Figure 3c. This was determined by ν1/2 − cathodic peak currents (Ip) versus scan rate ( ) in the extrapolating to the origin of the Koutechy Levich plot (inset), Randles−Sevcik plots, as shown in Figure 2 (inset) for O in which relates current as a function of electrode rotation speed. − 2 PYRTFSI (negative currents). The oxidation of O2 to O2 is In doing so, we found no contribution of kinetically limiting also shown (positive currents). In all solutions of Figure 1, the current at 1.82 V and lower, indicative of a mass transport ORRs involve diffusion-limited processes on both electrodes. controlled region.20 The ORR Tafel slopes extracted from RDE According to eq 1 for a reversible system data and listed in Table 2 reveal slightly enhanced kinetics in neat EMITFSI relative to PYRTFSI. Using these slopes, I =×2.69 1053/2nADC1/2 ν1/2 p O22O (1) transfer coefficients of 0.49 and 0.59 were calculated for the one-electron ORR, as shown in eq C1 of Scheme 2 in EMITFSI n is number of electrons, A is electrode area (cm2), D − is O2/O2 and PYRTFSI, respectively. − ff ffi × −6 × −6 the O2/O2 di usion coe cient (1.8 10 and 0.86 10 Effect of Alkali Metal Cations on ORR in Ionic Liquids. 2 10 fl + cm /s, respectively), CO2 is concentration of electroactive Figure 4a illustrates the in uence of Li on the ORR species (13.6 × 10−6 mol/cm3),10 and ν is scan rate (V/s). electrochemistry in PYRTFSI. At a 25 mM LiTFSI Comparing experimental values to theoretical plots of n =1,2 concentration, the final ratio of products differed slightly (Figure 2, inset), the ORR reaction in PYRTFSI unequivocally from that found in EMITFSI.7 Those data are presented here − consists of a one-electron O2/O2 redox couple, in agreement for clarity in Figure 4b. The PYRTFSI cathodic peak current

20757 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

Figure 4. Cyclic voltammograms (CVs) at 100 mV/s on a Au electrode. (a) 0.025 M LiTFSI in PYRTFSI. Inset: CVs at various cathodic limits of 1.95 V and higher. (b) 0.025 M LiTFSI in EMITFSI.

− O2 ion pairing. The negative shift in reduction potential for − − − − the PYR O2 relative to EMI O2 along with mass transport differences described in Figure 3 may account for the different behaviors in these Li+-spiked ILs. Following ORR, oxidation of multiple reduction products is evident in the anodic sweep of Figure 4a. Peak A1 results from ORR products at C2, since a cathodic voltage limit of 1.95 V eliminates peak A1 (see Figure 4a inset). In that illustration, the decrease in cathodic limit from 2.23 to 1.95 V results in formation of peak A2, followed by peak A3. Hence, these peaks are assigned to the oxidation of LiO2 Figure 3. RDE polarization curves on a Au electrode. (a) O2-saturated and Li O , respectively, with the reactions summarized in neat PYRTFSI (green) and EMITFSI (black) at 400 (solid) and 2500 2 2 (dashed) rpm at 10 mV/s. (b) EMITFSI with 0, 10, and 15 mM Scheme 3. These assignments are consistent with our previous results and those of others.5,6,22 Consistency between reaction LiTFSI at 400 rpm and 10 mV/s. Inset: Expanded view of ORR onset ff region. (c) ORR in neat EMITFSI at 5 mV/s, 400−1000 rpm, and time scales (at di erent cathodic limits) and respective A2/A3 corresponding Koutecky−Levich plot at 1.82 V (inset). peak area ratios is in agreement with the reaction sequence in Scheme 3. At cathodic limits > 1.95 V (Figure 4a inset), the extent of LiO2 disproportionation is limited due to reaction and potential at C1 in Figure 4a is in agreement with a one- time constraints. Its abundance and hence oxidation exceeds 7 − electron reduction to form LiO2. Nonaqueous Li O2 cell that of Li2O2. Increasing the reaction time for LiO2 electrochemistry also supports an initial one-electron prod- decomposition leads to more Li2O2, as seen in the decreased uct.3,21 A second reduction peak (C2) follows at 1.8 V, a A2/A3 ratio in Figure 4a, following a cathodic limit of 1.5 V. − − +− − consequence of further reduction of O2 to form O2 and PYR The fact that the PYR O2 ion pair can be seen at potentials

a Table 2. Ionic Liquid Properties ff ffi ffi density dynamic viscosity O2 di usion coe cient Schmidt number ORR Tafel slope transfer coe cient ionic liquid (g/cm3) (mPa s) (cm2/s) (Sc) (mV/dec) (α) EMITFSI 1.518 (17) 0.33 (18) 7.3 × 10−6 (19) 2.9 × 104 120 0.49 PYRTFSI 1.39 (15) 0.640 (10) 1.8 × 10−6 (10) 3.6 × 105 101 0.59 aReferences to selected values are indicated in parentheses.

20758 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

Scheme 3. 0.025 M LiTFSI PYRTFSI Scheme 4. 0.025 M NaPF6 EMITFSI

− − Initially, no EMI O2 products form on the GC electrode either in the presence of 25 mM Li7+ or in Na+, as seen in Figure 5a. Insulation of the GC electrode after several cycles, − − likely due to insoluble Na2O2 coverage, does initiate EMI O2 oxidation, as seen in the 20th cycle. Like the Li+-containing IL, below where LiO2 is formed suggests that the Li2O2, whose fi oxidation is seen at A3, is formed from the chemical the NaO2 ORR product with oxidation at A2 evolves rst, decomposition reaction shown in Scheme 3, and not from following high voltage cathodic limits in Figure 5b. The peak A3 intensity grows as the cathodic limit is expanded and the electrochemical reduction of LiO2. + reaction time scales increase, a result of NaO2 chemical Relative to Li /EMITFSI, cyclic voltammograms of O2 in the 25 mM Na+/EMITFSI presented in Figure 5 reveal only subtle decomposition to Na2O2. The high-voltage peaks in Figure 5 may come as a result of oxidation of proton-stabilized oxide + formation (e.g., H2O2) from EMI or residual H2O in solution, as we recently explained.7 In a 25 mM K+/EMITFSI solution, one ORR peak at C1 is followed by unique anodic features between 2.1 and 2.8 V in Figure 6a. Despite the 500 mV peak separation between the major cathodic and anodic processes, consistent rechargeability of ORR and OER products is observed after 20 cycles. The ORR reaction sequence presented in Scheme 5 accounts for the anodic features in Figure 6 and is described further below. The +− −− + [EMI O2 K ] represents a transition state for the +− − −− + conversion of EMI O2 to O2 K . However, direct ion − + pairing of O2 with K cannot be ruled out. Correlation of Cation Radius with its Lewis Acidity. The ORR and OER results presented above show a correlation between the nature of the conducting cation present in the IL electrolytes and the nature of ORR products formed and their oxidation potentials. An obvious difference among the various cations studied is the relative ionic radii (Table 1). Since ORR − 2− ff products, whether it is O2 or O2 , do not di er substantially in their ionic size, but only in their relative Lewis basicity, a most probable property of the cation that influences the stability of the ORR products is the Lewis acidity. To shed light on this view, we devised an experiment to measure the relative Lewis acidity of the various cations, including those of the ionic liquids examined. It is based on the fact that salt cations (M+) in nonaqueous electrolytes are generally solvated by the solvent − + − by forming acid base complexes, M (solvates)n(X ), where X is the anion of the conducting salt. The value of n in Li salt 13 Figure 5. Cyclic voltammograms of 0.025 M NaPF in EMITFSI on a solutions is believed to be 4. We hypothesized that the C 6 − GC electrode at 100 mV/s (a). Selected cycles: 1st (black), 2nd NMR chemical shift and spin lattice relaxation times (T1)of (blue), and 20th (red) scan. (b) Scans extending to various cathodic the CO functional group in propylene carbonate (PC) due limits as listed in the inset. to interaction (or formation of acid−base complex) with each of the cations studied would provide information on the relative acidity of the cations. Although the Li+-to-solvent molar ratio in changes. The single ORR peak, followed by multiple OER a 1 M Li salt solution is much less than 4, only one 13C signal peaks, witnessed here is a signature of an electron transfer, will be seen because of the rapid exchange of the solvated and followed by a chemical decomposition reaction, like that seen in unsolvated solvent molecules with Li+ in the NMR time scale.23 Li+ solution. Consequently, both Li+ and Na+ ORRs likely share A cation with a stronger Lewis acidity would form a stronger a similar reaction sequence in this IL. The probable ORR and complex with CO, as depicted in structure 1, and hence OER products are given in Scheme 4. would exhibit a larger 13CO chemical shift relative to PC.

20759 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

Scheme 5. 0.025 M KPF6 EMITFSI

LiPF6 and LiTFSI, the counteranion is shown here to have little effect on the 13C shift. Since all of the ions studied have the same single positive charge, the radius (and, hence, ionic volume) is roughly inversely proportional to the charge density, which is a measure of the Lewis acidity. Therefore, 13C NMR chemical shifts of the CO provide a measure of the complexing ability of the Lewis base CO (by donating the electron pair on O) with the Lewis acid cations, and hence the relative acidity of the cations. This effect is illustrated in Figure 7b with a cascading chemical shift as a function of ionic radius for these cations. The cations of the ionic liquids are shown to have a similar weak acidity as TBA+ with similar 13C chemical shifts closer to those of neat PC, while the 13C peaks are farther downfield for Li+ and Na+ solutions, indicating their strong Lewis acidity. K+ has an intermediate Lewis acidity among the cations studied, and its electrochemistry clearly supports it. To gain additional information on the relative acidity of the various cations from solvent−salt interactions with PC as a Figure 6. (a) Cyclic voltammograms of O2-saturated 0.025 M KPF6 in 13 EMITFSI at 100 mV/s on a GC electrode: cycle 1 (black) and cycle reference, the C longitudinal relaxation times (T1), also called − 13 20 (red). (b) Incremental cathodic limits ranging from 2.38 to 1.19 V. spin lattice relaxation times, were measured using C NMR. (c) Various scan rates (10−300 mV/s). Inset shows expansion on The T1 measurements were conducted on all carbon atoms of anodic peaks. propylene carbonate in the presence of the various salts. A typical compilation of inversion recovery spectra for 1 M LiPF6 The very large IL and TBA cations with their low charge in PC obtained with various delay times is shown in Figure 8. densities would exhibit the least Lewis acidity and, therefore, The T1 of the four PC carbon atoms are plotted as a function probably the least chemical shift change. The chemical shift of cation radius (Lewis acidity) in Figure 9. Comparing trends data would be complemented by the 13CO spin relaxation of all four carbon environments, the overall relaxation efficiency times, with the shortest T1 in solutions of the salt having the increases (i.e., shorter T1) as the number of protons attached to  − ff strongest Lewis acid cations. carbon increases (C O>CH>CH2). The e ect of 13 13  ff The C NMR spectra highlighting the C O chemical di erent cations on T1 times is most pronounced on the shifts of PC with and without the various salts, including the carbonyl carbon. The random and rapid motion of neat PC ionic liquids, are presented in Figure 7a. This plot shows that, molecules yields the longest relaxation time for the given + + + in LiTFSI and NaPF6 salt solutions, Li and Na ions are carbon transition frequency. The Li , which deshields the strongly solvated to the carbonyl oxygen atoms of PC since carbonyl carbon most effectively, as seen in Figure 7, produces 13 ffi 13  they have the largest negative chemical shifts (less C the most e cient relaxation and shortest C O T1 time. The + +−  − shielding) relative to PC while K and the three large cations, strong Lewis acidity of the solvate that forms Li (C O)n + + + − TBA , EMI , and PYR , are weakly coordinated. Comparing TFSI may inhibit internal motion within PC to shorten the T1.

20760 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

13 Figure 8. C NMR inversion recovery spectra to calculate T1 for 1 M LiPF6 in EMITFSI with partially relaxed spectra.

Correlation between ORR and OER Processes and Relative Acidity of the Cation in the Electrolyte. As we have seen earlier, there is no significant change in ORR following addition of TBA+. We can estimate relative Lewis acidities according to the inverse of their atomic radii. The large single-charged cations with more shielded, polarizable valence electrons make the three cations TBA+, EMI+, and PYR+ soft acids. They confer exceptional stability to the one-electron − +− − ORR product O2 , which is a soft base, as M O2 (M = PYR, EMI). This affords one-electron reversibility, consistent with our previous observations in organic electrolytes.5,6,24 The effect of hard acids on the ORR products is evident in solutions containing Li and Na salts. Both Li+ and Na+ are hard acids in the ionic liquids, in agreement with their behavior previously observed in organic electrolytes. In O2 solutions containing LiTFSI in Figure 4, for example, the cathodic peak currents versus scan rate and potential are in agreement with a one-electron reduction of O to form LiO , which then may Figure 7. (a) 13C NMR chemical shifts of the propylene carbonate 2 2 (PC) carbonyl carbon, concentrated with various 1 M salt solutions: 1 disproportionate to Li2O2. An initial one-electron product was + + + + confirmed with scan-rate-dependent studies in Li-concentrated M LiPF6 (Li ), 1 M LiTFSI (Li ), NaPF6 (Na ), KPF6 (K ), EMITFSI + + + EMITFSI resembling closely the voltammetry behavior (EMI ), PYRTFSI (PYR ), and TBAPF6 (TBA ). The structure of PC 13 is shown in the inset. (b) C(O) C chemical shifts relative to neat observed here. Furthermore, the presence of both LiO2 and/ fi PC as a function of cation ionic radius. or Li2O2 as ORR products has been identi ed from in situ Raman spectroelectrochemistry in organic electrolyte and X-ray diffraction patterns of Li−air cell discharge products.22,24 The − 2− An increase in radius of the coordinating cation yields longer T1 softness of O2 relative to O2 favors chemical decomposition + + − 2− times up until the EMI coordination, where relaxation times of an unstable LiO2 to form the hard(Li ) hard(O2 ) product ff fi decrease again. Molecular motion, di usion, and rotation are all in Li2O2. These data con rm the general trend in the nature of modes that contribute to the relaxation process. Lack of charge the ORR products vested in the hard soft acid base (HSAB) density and the soft acid nature of the largest cations PYR+ and theory of Pearson25 that hard acids want to be associated with TBA+ appear to enhance relaxation, perhaps through a different hard bases while soft acids associate with soft bases. Although means related to the way they interact with the solvents than hard acids, such as Li and Na+, have a strong electrostatic 13 − the smaller cations. C NMR in Figure 7 revealed only slight attraction toward soft bases, such as O2 , they cannot remain as  + chemical shifts in the C O environment for either PYR or stable LiO2 or NaO2 because of the stability founded in the TBA+ medium. A decrease of n in the cation+−(CO) − HSAB theory. The cations investigated in this work listed in − n TFSI complex with these larger cations may enhance Table 1 rank in terms of Lewis acid hardness in the manner, + + + + + + molecular tumbling during relaxation and improve their T1 TBA < PYR < EMI

20761 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

13 + Figure 9. Calculated C NMR T1 values of propylene carbonate with various 1 M salt solutions: propylene carbonate (neat PC), LiPF6 (Li ), + + + + + + LiTFSI (Li ), NaPF6 (Na ), KPF6 (K ), EMITFSI (EMI ), PYRTFSI (PYR ), and TBAPF6 (TBA ).

+ 2− Na2O2 caused by the hard acid Na -to-hard base O2 stabilization interaction. Lewis acidity differences between EMI+ and K+ are apparently not substantial enough for ORR products to be ion paired with either EMI+ or K+ to dominate the reaction. As seen in Figure 6b, increments in the cathodic voltage limit results in an initial broad anodic feature following a cathodic limit of 2.0 V (highlighted in red) in the vicinity of peak A1. The broadness may come as a result of the intermediate +− −− + (EMI O2 K ), which is oxidized at A2. Peak A2 achieves improved resolution as the cathodic limit is decreased toward the 1.2 V limit (blue curve). At these lower voltages, this −− + transitioning species probably transforms to O2 K and is oxidized at A3. Cycling over a range of scan rates illustrated in fi Figure 6c further clari es the time dependence of this reaction. Figure 10. Cyclic voltammogram of neat EMITFSI along with various Peak A3 intensity dominates at slower scan rates (10−65 mV/ salts at 0.025 M concentration on a GC electrode at 100 mV/s. s) on the right side of the hash mark (expanded in Figure 6c inset). This is followed by a gradual intensity change in favor of uncompensated resistance on the electrode surface from peak A2 at faster scan rates. The degree of softness from the − insoluble product accumulation may, in part, also contribute cations in steps 1 3 of Scheme 5 stabilizing the superoxide to this occurrence. directly determines shifts in oxidation potentials and hence ORR Mechanism in Nonaqueous Electrolytes. ff The a ects the degree of reversibility. results presented here provide strong evidence for the role The trend of shifting oxidation potentials extends to the cations play in controlling the reaction mechanism and OER peak positions of all four cations in EMITFSI, as products of O2 reduction reactions in ionic liquids with strong summarized in Figure 10. The OER overpotentials of cation− − correlations between the Lewis acidity of the cations and O2 ion pairs increases as a function of cation hardness. basicity of the ORR products. For conducting cations, the According to HSAB theory, stronger, covalent bond character is Lewis acidity is inversely proportional to the ionic charge more favorable for soft/soft interactions, resulting in improved density, as confirmed by the NMR data. For oxygen reduction 2− 2− − equilibrium of the redox species at the electrode surface and products, basicity decreases in the order O >O2 >O2 . better reversibility. As the cation hardness increases, it imparts With this concept, it is now possible to predict the mechanism − − instability in the cation O2 ion pair, which coupled to and stability of ORR products in ionic liquids. The findings in chemical decomposition to form the peroxide, forces redox this work unify our previous observations in organic electro- concentrations farther from Nernstian-like conditions. The lytes5 and lead to the formulation of a general theory for −− + strongly ion-paired O2 M (where M = Li and Na) ORR predicting the ORR mechanism and products in nonaqueous products are oxidized at the highest anodic potentials. Some electrolytes based on the HSAB concept. The results in the

20762 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article ionic liquid media with a mixture of cations, specifically IL+ and ■ CONCLUSION + + Li or Na , provide unambiguous evidence for the Lewis acidity A strong correlation was found between the ORR products and of the cation to stabilize the reaction products. All soft acid the ionic charge density of the conducting cations in + + cation-containing electrolytes, such as IL and/or TBA , nonaqueous electrolytes. The observed trend is explained in provide a happy medium for the highly reversible one-electron terms of the Lewis acidity of the cation present in the − + + 13 O2/O2 process. When a strong Lewis acid, such as Li or Na , electrolyte using an acidity scale created from C NMR − 13  is present, even at low concentrations, it dominates the reaction chemical shifts and spin lattice relaxation (T1) times of C mechanism, ultimately leading to the formation of the stable O in solutions of the salts of these charged ions in propylene 2− 2− hard base products O2 or O . The CV results in ionic liquids carbonate. The ionic liquids lie in a continuum of a cascading that contained Li+ show clear evidence for the chemical Lewis acidity scale with respect to the charge density of alkali disproportionation of LiO2 to Li2O2 in order to satisfy the metal, IL, and TBA cations with the result that the ORR stability requirements underlying the HSAB concept. products in ionic liquids and in organic electrolytes containing 13 any conducting cations can be predicted using a general theory The C NMR chemical shift and T1 data also provide further support to the view of solvation of strong acid cations, such as based on the hard soft acid base (HSAB) concept. Li+ and Na+, in organic electrolytes for these ions to exist as + ■ AUTHOR INFORMATION solvated ions M (solvent)n. Interestingly, the NMR data show that the large, soft acid cations IL+ and TBA+ are little solvated Corresponding Author through point charge (CO → M+) complexation between *Address: Northeastern University, 360 Huntington Avenue, the solvents and the cations. The interactions between the soft 317 Egan Research Center, Boston, MA 02115. Phone: (617)- acid cations TBA+ or IL+ and organic solvents, such as PC, 373-5630. Fax: (617)-373-8949. E-mail: kmabraham@comcast. , , or dimethoxyethane, leading to net. the formation of their respective electrolyte solutions, probably Notes − − involve dipole dipole or dipole induced dipole involving The authors declare no competing financial interest. dispersion energy. The solvation of Li+ in organic solvents evidenced by the NMR data also provides support to our earlier ■ ACKNOWLEDGMENTS proposal6 for the modulation of the Lewis acidity of the hard Financial support for this work was provided by the U.S. Army acid Li+ by the basicity of the organic solvents in Li+-conducting Cerdec through Subcontract No. GTS-S-10-392. nonaqueous electrolytes. As mentioned earlier, in a nonaqueous electrolyte composed of a solution of a Li salt in an organic + ■ REFERENCES solvent, the solvent complexes with the Li through the donor electrons on its oxygen, nitrogen, or other heteroatom to form (1) Abraham, K. M.; Jiang, Z. J. Electrochem. Soc. 1996, 143,1. solvent-separated ion pairs. Typically, the Li+ is solvated by four (2) Christensen, J.; Albertus, P.; Sanchez-Carrera, R. S.; Lohmann, T.; Kozinsky, B.; Liedtke, R.; Ahmed, J.; Kojic, A. J. Electrochem. Soc. solvent molecules to form ion pairs of the general formula, 2012, 159, R1. + − − 26 + Li (solvent)4 X , where X is the anion of the Li salt. Li , (3) Xiao, J.; Hu, J.; Wang, D.; Hu, D.; Xu, W.; Graff, G. L.; Nie, Z.; because of its small ionic radius and the resulting high positive Liu, J.; Zhang, J.-G. J. Power Sources 2011, 196, 5674. charge density, is a hard Lewis acid. However, the formation of (4) Zhang, S. S.; Foster, D.; Read, J. J. Power Sources 2010, 195, 1235. + − − (5) Laoire, C. O.; Mukerjee, S.; Abraham, K. M.; Plichta, E. J.; the solvates, Li (solvent)4 X , modulates (lowers) the acidity of Li+ in organic electrolytes. The extent to which Li+ acidity is Hendrickson, M. A. J. Phys. Chem. C 2009, 113, 20127. (6) Laoire, C. O.; Mukerjee, S.; Abraham, K. M.; Plichta, E. J.; lowered by solvation depends on the Gutmann donor (DN) Hendrickson, M. A. J. Phys. Chem. C 2010, 114, 9178. number, which is a measure of the basicity or electron-donating (7) Allen, C. J.; Mukerjee, S.; Plichta, E. J.; Hendrickson, M. A.; property of the solvent. A higher donor number solvent lowers Abraham, K. M. J. Phys. Chem. Lett. 2011, 2, 2420. the acidity of Li+ to a greater extent to make it a softer Lewis (8) Appetecchi, G. B.; Scaccia, S.; Tizzani, C.; Alessandrini, F.; acid than by a solvent possessing a lower donor number. We Passerini, S. J. Electrochem. Soc. 2006, 153, A1685. have found a strong correlation between the modulated Lewis (9) Sawyer, D. T. Oxygen Chemistry; Oxford University Press: New York, 1991. acidity of the Li+ solvates formed, judiciously chosen for their ff (10) Katayama, Y.; Sekiguchi, K.; Yamagata, M.; Miura, T. J. DN di erences, and the ORR mechanism as well as the Electrochem. Soc. 2005, 152, E247. 6 27 ultimate products of ORR. Dimethyl sulfoxide with a donor (11) Carter, M. T.; Hussey, C. L.; Strubinger, S. K. D.; Osteryoung, number of 29.8 lowers the Lewis acidity of Li+ to the extent R. A. Inorg. Chem. 1991, 30, 1149. − + − − (12) AlNashef, I. M.; Leonard, M. L.; Kittle, M. C.; Matthews, M. A.; that it provides a longer lifetime for O2 as Li (DMSO)n O2 . The ORR and OER results in ionic liquids presented here Weidner, J. W. Electrochem. Solid-State Lett. 2001, 4, D16. 6 (13) Paul, R. C.; Johar, S. P.; Banait, J. S.; Narula, S. P. J. Phys. Chem. together with our earlier observations in organic electrolytes 1976, 80, 351. lead to an all-encompassing theory based on the HSAB concept (14) Fawcett, W. R.; Ryan, P. J. Collect. Czech. Chem. Commun. 2009, to predict the ORR mechanism and products in nonaqueous 74, 1665. electrolytes. HSAB concepts also provide an explanation for the (15) Appetecchi, G. B.; Montanino, M.; Zane, D.; Carewska, M.; very transient or lack of existence of O − in aqueous electrolytes Alessandrini, F.; Passerini, S. Electrochim. Acta 2009, 54, 1325. 2 ’ where the strong acid H+ wants to stabilize the strong base (16) Boxall, D. L.; O Dea, J. J.; Osteryoung, R. A. J. Electrochem. Soc. O 2− or O2− with the result that the observed ORR products in 2002, 149, E468. 2 (17) Noda, A.; Hayamizu, K.; Watanabe, M. J. Phys. Chem. B 2001, aqueous acid electrolytes are H2O2 and H2O. Indeed, the one- 105, 4603. fi electron reduction of O2 may be the rst ORR step in the (18) Matsumoto, H.; Sakaebe, H.; Tatsumi, K.; Kikuta, M.; Ishiko, E.; aqueous electrolytes as in the nonaqueous electrolytes. Kono, M. J. Power Sources 2006, 160, 1308.

20763 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764 The Journal of Physical Chemistry C Article

(19) Buzzeo, M. C.; Klymenko, O. V.; Wadhawan, J. D.; Hardacre, C.; Seddon, K. R.; Compton, R. G. J. Phys. Chem. A 2003, 107, 8872. (20) Zurilla, R. W.; Sen, R. K.; Yeager, E. J. Electrochem. Soc. 1978, 125, 1103. (21) Lu, Y.-C.; Gasteiger, H. A.; Shao-Horn, Y. Electrochem. Solid- State Lett. 2011, 14, A70. (22) Peng, Z.-Q.; Freunberger, S. A.; Hardwick, L. J.; Chen, Y.-H.; Giordani, V.; Barde, F.; Novak, P.; Graham, D.; Tarascon, J.-M.; Bruce, P. G. Angew. Chem., Int. Ed. 2011, 50, 6351. (23) Abraham, K. M.; Pasquariello, D. M.; Martin, F. J. J. Electrochem. Soc. 1986, 133, 661. (24) Laoire, C. O.; Mukerjee, S.; Plichta, E. J.; Hendrickson, M. A.; Abraham, K. M. J. Electrochem. Soc. 2011, 158, A302. (25) Pearson, R. G. J. Am. Chem. Soc. 1963, 85, 3533. (26) Gnanaraj, J. S.; Thompson, R. W.; DiCarlo, J. F.; Abraham, K. M. J. Electrochem. Soc. 2007, 154, A185. (27) Mamantov, G., Popov, A. I., Eds. Chemistry of Nonaqueous Solutions: Current Progress; VCH Publishers: New York, 1994.

20764 dx.doi.org/10.1021/jp306718v | J. Phys. Chem. C 2012, 116, 20755−20764