01 the electrochemistry of electrodes in organic solvents

225 J. Electroanal. Chem., 297 (1991) 225-244 Elsevier Sequoia S.A., Lausanne The electrochemistry of noble metal electrodes in aprotic organic solvents containing lithium salts D. Aurbach *, M. Daroux, P. Faguy and E. Yeager ** Case Center for Electrochemical Sciences, and The Department of Chemistry, Case Western Reserve University, Clevelanrl, OH 44IO6 (USA) (Received 9 August 1989; in revised form 23 March 1990) Abstract The electrochemical behavior of various non-aqueous organic electrolyte systems has been investi- gated using inert metal electrodes. The systems studied included propylene carbonate, dimethoxyethane and tetrahydrofuran solutions of LiClO,, L&F,, LiSO,CF, and Bu,NClO,. The electrode metals included polycrystalline gold and silver. Various techniques including cyclic voltammetry, FTIR and XPS were used to characterize the main electrochemical reactions that occur in these systems. Several separate film forming processes have been identified, including reduction of solvent, the salt, and traces of oxygen and water. The surface films formed in these processes lead to the apparent stability of these systems at low potentials. Li UPD was also examined and was found to be controlled by the nature of the surface films through which lithium is deposited. INTRODUCTION There is increasing interest in carrying out electrochemical reactions in non-aque- ous systems [l], both because of the wide range of potentials that can be reached before solvent decomposition occurs in these systems when compared with aqueous systems, and because of the solubility and/or stability requirements of many reagents. The electrochemical behavior of noble metal electrodes in non-aqueous systems is important in defining the potential windows available and the reactivity of the solvents, salts and contaminants such as water and oxygen that may be present. The objective here is to survey the behavior of noble metal electrodes in lithium ion solutions in a number of solvent systems of commercial importance to the battery industry. The emphasis is on identification of the general features and * Present address: Department of Chemistry, Bar-Ban University, Ramat Gan 52100, Israel. ** To whom correspondence should be addressed. 0022-0728/91/$03.50 0 1991 - Elsevier Sequoia S.A. 226 processes common to all these systems, rather than on the individual details of specific reactions. Many studies of electrochemical stability of non-aqueous systems have been carried out by organic electrochemists searching for inert solvents in which to study the reaction of more electroactive materials [2]. The electrochemical stability of a variety of dipolar aprotic systems including propylene carbonate (PC), di- methoxyethane (DME), tetrahydrofuran (THF) and acetonitrile solutions of salts such as LiClO,, NaCIO, and R,NClO, has been investigated by Mann [3] using platinum, silver and carbon working electrodes. Electrochemical windows of more than 5 V could be obtained with most of the systems listed above when a platinum working electrode was used. When silver was used, silver dissolution limited the electrochemical window to a width of ca. 3.0 V. Similar work has been carried out by others [4,5]. The limiting reaction of Li salt solutions in alkyl carbonates and ethers at noble metal electrodes at potentials approaching that of the reversible Li/Li+ electrode is usually lithium deposition rather than solvent reduction [6]. In many instances however the anodic or cathodic reactions limiting this voltage window are still unknown. The reason for the stability of many non-aqueous solvents at such negative potentials is not clear. Another important aspect regarding the electrochemistry of nonaqueous systems is the role of impurities such as water and oxygen which are always present in dipolar aprotic systems. None of the purification methods for non-aqueous solvents is capable of eliminating water totally from the system. De-aeration is difficult in many cases because of the volatility of many non-aqueous solvents. The reduction of water on platinum in some non-aqueous solvents containing lithium salts has already been studied [7,8]. With other noble metal electrodes such as gold and silver, less work has been done. Oxygen reduction in a few nonaqueous systems has also received attention [see for example refs. 9-13 and references therein]. The present research has been concerned with the electrochemical behavior of gold and silver electrodes in propylene carbonate (PC), dimethoxyethane (DME) and tetrahydrofuran (THF), with a variety of electrolyte salts. FTIR and X-ray photoelectron spectroscopy have been used to characterize the electrode surfaces and to identify surface species. EXPERIMENTAL General All of the electrochemical measurements and sample preparations for the spec- troscopic studies were carried out at room temperature under an Ar atmosphere in a Vacuum Atmosphere Corp. (VAC) dry box as described elsewhere [14,15]. Propylene carbonate (Burdick and Jackson, distilled in glass), dimethoxyethane (Grant Chemicals, triply distilled in glass), tetrahydrofuran (Aldrich), LiClO, (G. Fredrick Smith), L&F, (USS Agrichemical), LiSO,CF, (Grant Chemicals), Bu,NClO, (Fisher) and Li foils l-2 mm thick (Foote mineral) were used. Their purification and drying methods have been described elsewhere [14-161. 221 Electrochemical measurements For the electrochemical measurements, a PARC potentiostat (Model 176) and programmer (Model 179, together with a Honeywell 522 XY recorder were used. Two types of working electrodes were used: (1) Gold or silver disks (6 mm o.d.) mounted in polypropylene holders were used both as stationary and rotating electrodes. (2) Gold and silver were deposited on glass or mica substrates (average thickness 200 nm, 1 to 6 cm*) in a high vacuum system (10-6-10-7 Torr) either by evaporation or sputtering. Since gold does not adhere well to glass, a thin film of niobium (5-10 nm) was deposited on the glass by sputtering prior to gold deposi- tion. All the voltammograms presented in this paper were obtained with such noble metal on glass electrodes. Two types of cells were used. For lithium salt solutions, a single compartment cell with both lithium counter and reference electrodes was satisfactory. For Bu,NClO, solutions, a three-compartment cell was used. The counter electrode was an electronically conducting polypyrrole film deposited on gold foil. This film was formed by oxidative electrodeposition of the polymer under galvanostatic conditions in a pyrrole solution in DME/Bu,NClO,. Use of this counter electrode avoided contamination arising from counter electrode reactions since the redox processes in polypyrrole are localized in the polymer film with anion migration providing charge compensation. The reference electrode was Li foil in a 0.1 M solution of LiClO, in a 0.1 M Bu,NClO, solution of the solvent under study. The reference compartment was separated from the rest of the cell by an additional compartment filled with Bu,NClO, solution. Ionic contact between the compartments was maintained through a finepore glass frit. Sample preparation for FTIR measurements Highly reflective (mirror-like), large area gold or silver electrodes were prepared either by evaporation or sputtering gold and silver on glass substrates or by polishing silver plates (45 mm X 20 mm X 0.5 mm) with 0.05 pm y-alumina. These plates were used as working electrodes in a one-compartment electrochemical cell with the appropriate solution. Li foils were used as counter and reference electrodes. The potential was stepped or swept from the open circuit value (OCV) (2.5-3.0 V vs. Li/Li+) to the predetermined potential. Then the samples were taken out, washed with pure solvent, dried, and transferred under argon to the spectrometer. In a number of the experiments, the samples were protected by KBr plates as described in an earlier paper [14]. The silver sample with the salt plate on it was wrapped with 0.5 mm thick lithium foil (25 mm X 50 mm). The edges of the Li foil were then pressed toward the sides of the KBr plate to form seals [14]. For each set of experiments a reference sample was prepared. This consisted of a silver (or gold) plate, treated in the same way, washed with solvent, dried and then sealed with a KBr plate and lithium foil as described above. The effectiveness of this transfer technique has been verified in earlier work [14], and the FTIR measurements are described in detail elsewhere [14,15]. 228 XPS measurements XPS measurements were carried out using a Varian ESCA (VIEE-15) spectrome- ter. The pressure in the spectrometer chamber was lo-’ Torr. Silver strips (20 mm x 3 mm x 0.5 mm) polished with 0.05 pm y-alumina were used as working electrodes in PC, DME and THF solutions of 0.2 M L&F,, with Li foils as counter and reference electrodes. The electrochemical treatment was similar to that of the samples for FTIR. The specimens were transferred to the spectrometer and intro- duced inside under argon, protected from atmospheric contamination by a polyeth- ylene glove bag. RESULTS AND DISCUSSION The gold and silver electrodes were studied in propylene carbonate (PC), di- methoxyethane (DME) and tetrahydrofuran (THF). The gold electrode has the wider electrochemical window and the initial discussion is focused on results obtained with this metal (Figs. l-7), although similar behavior is seen with silver electrodes. When the electrode was immersed initially in the solution, the open-circuit potential fell in the range 2.5 to 3.0 V (vs. Li/Li+) in all cases, regardless of the solvent, the salt or the electrode metal. As the potential was scanned in the negative direction in Li+-containing solutions, evidence is seen for at least five separate (although not independent) reduction reactions prior to the deposition of bulk lithium. The following reactions are believed to occur (in approximate order of increasingly negative potential): reduction of oxygen, reduction of water, reduction of solvent, reduction of the anion, and a UPD-like reduction of Li+ prior to bulk deposition. In systems where the tetrabutylammonium cation was used in place of Lif, the negative limit to the sweep was set by the large reduction currents for the cation. This normally occurred at 0.2 V in the case of ethereal solvents (DME, THF) [16]. In the case of tetrabutylammonium perchlorate (TBAP) + PC solutions, substantial solvent reduction occurs at potentials around 0.7-0.8 V (vs. Li/Li+) [16]. With gold electrodes the positive limit is set at approximately 4.5-5.0 V by large anodic currents (of the order of hundreds of pA/cm’) which are attributed to solvent oxidation [17]. With silver electrodes, anodic dissolution of silver sets the limit at 3.6-3.7 V. Oxygen reduction The initial cyclic voltammogram obtained in 0.1 M LiAsF, with a fresh gold electrode is shown in Fig. 1. A pronounced cathodic peak is seen at approximately 1.8 V on the first cycle, but this peak is completely absent on the second and subsequent cycles. There is no corresponding anodic feature. A second, somewhat more pronounced cathodic peak is seen at about 1.1 V. This peak is greatly reduced in the second and subsequent sweeps, although some activity remains in this region at steady state. The disappearance or marked diminution of these two peaks on the 229 Au /DMEtO.IM Li AsF6 --60 U - f wst cycle ----- second cycle _._.._......_ &,,,,y state --so - 20 mV/s 10 E/V (vs. LilLi 2.0 30 I I Fig. 1. Cyclic voltammetry, gold in DME+O.l M L&F,. Sweep rate 20 mV/s. (p ) First cycle, no additives (- - -). Second cycle. (. . . . .) Steady state, no difference if solution is saturated with 0,, once steady state is achieved. second sweep is seen even if the electrode is held at open circuit voltages for an hour or more between the first and second sweeps. Essentially the same behavior is seen with either L&F, or LiClO, solutions in DME, THF or PC. Oxygen and water have pronounced effects on the first cycle behavior obtained with fresh electrodes, but the addition of either once steady state has been reached has no effect on the voltammograms. Figure 2 shows the first cycle voltammetry of a gold electrode in DME + 0.1 M L&F, which has been saturated with.0,. The peak at 1.8 V has shifted to - 2.0 V, and is greatly enhanced in comparison with Fig. 1, while the peak at 1.1 V is greatly reduced. Again, however, the 1.8 V peak is absent on the second cycle. On the basis of this result the peak at 1.8 V is attributed to oxygen reduction. Figure 3 shows the voltammetry in the region of the - 2.0 V peak obtained when the Li+ cation is replaced by the tetrabutylammonium ion for both DME and PC Au/DMEtO.IM LiAsFct02 sat. --60 20 mV/s E/V (vs. LilLi? LO 20 30 8 I Fig. 2. Effect of 0, on the cyclic voltammetry of Au electrode in DME + 0.1 M LiAsF, saturated with Oz. 20 mV/s. (- ) First cycle with a new electrode. (- - -) Second cycle, same electrode. 230 j//LA c~n-~ DMEtO.IM TBAPtOz sot &LA cm-2 PCtO.IM TBAP to2 sat. ~~mv/s~~~dit,~~m~~~~~~~~,~4 -“““.” steady state after Of 0.005M LEIO4 LiCl04 addition --600 E/V (vs. Li/Li+) --2’o E/V (vs. LijLi+) I.0 20 3.0 0 10 2p J.0 co I I 1 I I I I I I I I A B Fig. 3. Effect of Li+ on the oxygen reduction process on Au in DME+O.l M TBAP and PC+O.l M TBAP. 20 mV/s. ( -) Steady state behavior. (- - -) Next cycle after the addition of 0.005 M LiClO,. solutions. The solutions were saturated with 0, before the electrode was introduced. In contrast to the presence of lithium ions, in TBAP solutions there is no significant difference between the first cycle and steady state behavior. In both DME and PC, at steady state, there are pronounced cathodic peaks in the region of 2.0 V, which are considerably larger than the first cycle counterparts observed in Li+-containing solutions. In DME there is a corresponding anodic peak, while in PC the voltammo- gram has the form of a totally irreversible reduction process. The addition of 0.005 M Li+ causes a marked change in the oxygen reduction behavior. On the cycle immediately following the addition (dashed line in Fig. 3) a peak at 2.0 V similar to that seen in Fig. 2 appears. On subsequent cycles the electrode is completely passivated with respect to oxygen reduction (dotted line). There have been a number of previous investigations of oxygen reduction in organic solvents [9-111. The behavior found in the present work for O,-saturated DME + TBAP solution is very similar to that seen previously for O,-saturated acetonitrile (ACN) + TBAP solution [9-111. The cathodic peak at 1.6-1.7 V and the corresponding anodic peak around 2.2-2.3 V (vs. Li/Li+) (see Fig. 3a) are probably equivalent to the peaks at - 1.1 V (SCE) and - 0.65 V (SCE) found by Sawyer et al. [9] for O,-saturated ACN + TBAP solutions. These authors attribute this pair of peaks to the formation and oxidation of superoxide ion according to the equation O2 + e- e o;- (1) As Fig. 3 shows, the behavior of PC + TBAP solution is very different from that in DME and ACN. In PC the reverse reaction is not seen, and the probable explanation is that an EC process occurs in which the superoxide ion reacts with the solvent in an irreversible process. This behavior may reflect the difference in electrophilicity between PC and the other two solvents. The greater ease of reduc- 231 tion of PC has been noted in earlier work [16]. PC should be a good substrate for nucleophilic attack since the leaving group is -0CO; in which the negative charge is stabilized by resonance involving two equivalent oxygen atoms (or the stable carbonate anion). In DME, nucleophilic attack on the solvent is less likely, since alkoxide is not an effective leaving group [18]. Therefore, in DME in the absence of strong electrophiles, the superoxide ion is relatively stable and survives long enough to be reoxidized. When a lithium salt was introduced into O,-saturated TBAP solutions of PC or DME, the 0, reduction process described above became irreversible and the electrode passivated, as is also shown in Fig. 3. The most likely explanation for this result is that when Li+ ions are present a precipitate of lithium superoxide precipitates and blocks the electrode surface. Similar behavior was observed by Sawyer et al. [9] when metallic cations were introduced into O,-saturated ACN + TBAP solutions. Although the film thus formed prevents further oxygen reduction, it does not passivate the electrode totally for other processes which occur at more negative potentials (Fig. 2). The charge under the 0, reduction peak obtained with a solution containing only the usual background levels of oxygen (Fig. 1) is on the order of 400-600~ +/cm2. This indicates that a few monolayers of superoxide ( - 200 PC/cm2 per monolayer, assuming one electron transfer) is enough to passivate the electrode for further 0, reduction. In some cases, the first cycle oxygen reduction peak is split into a broad peak centered around 2.1 V and a shoulder at 1.8 V (Fig. 7). This suggests that two film forming processes may be occurring. One is probably lithium superoxide formation as discussed above, while the other could be the formation of lithium peroxide according to schemes such as the following: 2 o;- e 0, + o;- (2) I 2 Lit Li 202 Li02 Li 202 Both LiO, and Li,O, should be highly insoluble in ethereal solvents or PC. WATER REDUCTION (3) Figure 4 shows a typical first cycle voltammogram for Au in PC + 0.2 M LiClO, solution with 0.01 M H,O added. In this case the peak at 1.2 V is greatly increased. This peak is therefore attributed to the’reduction of water. On the second cycle, this 232 1 j/pA cm-2 Au/PCtO.ZM LiAsFgtO.OlM Hz0 --100 - Li B: --EQ - first c~& - --- second cyck 20 mV/s --a0 E/V (vs. LilLi Fig. 4. Effect of H,O on cyclic voltammetry at Au in PC+O.2 M LiAsF, +O.Ol M H20. Sweep rate 20 mV/ s. (- ) First cycle with a fresh electrode. (- - -) Second cycle. feature is very much reduced, but a residual cathodic wave is seen on this and subsequent cycles, indicating that the water reduction process is not totally self-pas- sivating. If, after the electrode has been swept through the water reduction peak, the potential is then swept positive again, a pronounced anodic peak appears at about 3.5 V. On the following negative sweep a new reduction peak also appears at 2.8 V. From the initial positive sweep in Fig. 4, it is evident that this pair of peaks is never apparent until after the electrode has been cycled to potentials negative of the water reduction peak (1.0 V or less). Figure 5 shows the effect of the addition of O.OlM H,O on the steady state voltammogram of a gold electrode in DME + 0.1 M L&F,. The reduction wave around 1.2 V is enhanced somewhat, although to nowhere near the extent of the first cycle, as is the pair of peaks around 3.0 V. In addition, there is an enhancement of the cathodic processes that occur below about 0.6 V. Again, the effect of water addition on both the first cycle and steady state behavior was substantially the same for all the salts (L&F,, LiClO,, LiSO,CF, of LiBF,) and all the solvents (PC, DME or THF) studied. The most probable reaction for the process yielding the water reduction peak around 1.2 V is: H,O + Li++ e-= LiOH + f H, (4) LiOH should be insoluble in these systems and can be expected to precipitate on the electrode surface. FTIR spectra obtained from the surface of gold and silver electrodes treated in H,O contaminated solutions at potentials below 1.2 V show bands typical of LiOH, particularly the sharp band at about 3680 cm-