Existing Condition and Migration Property of Ions in Lithium

Sep 15, 2007 - Yuria Saito,*,† Tatsuya Umecky,† Junichi Niwa,† Tetsuo Sakai,† and Seiji Maeda‡. National Institute of AdVanced Industrial Sc...
3 downloads 0 Views 186KB Size
Download PDF
11794

J. Phys. Chem. B 2007, 111, 11794-11802

Existing Condition and Migration Property of Ions in Lithium Electrolytes with Ionic Liquid Solvent Yuria Saito,*,† Tatsuya Umecky,† Junichi Niwa,† Tetsuo Sakai,† and Seiji Maeda‡ National Institute of AdVanced Industrial Science and Technology, 1-8-31, Midorigaoka, Ikeda, Osaka 563-8577, Japan, and The Nippon Synthetic Chemical Industry Company Limited, 2-13-1, Muroyama, Ibaraki, Osaka 567-0052 Japan ReceiVed: April 18, 2007; In Final Form: July 17, 2007

Ionization conditions of each ionic species in lithium ionic liquid electrolytes, LiTFSI/BMI-TFSI and LiTFSI/ BDMI-TFSI, were confirmed based on the diffusion coefficients of the species measured by the pulsed gradient spin-echo (PGSE) NMR technique. We found that the diffusion coefficient ratios of the cation and anion obs obs obs species Dobs Li /DF of the lithium salt and DH /DF of the ionic liquid solvent were effective guides to evaluate the ionization condition responsible for their mobility. Lithium ions were found to be stabilized, forming the solvated species as Li(TFSI)32-. TFSI- anion coordination could be relaxed by the dispersion of silica to form a gel electrolyte, LiTFSI/BDMI-TFSI/silica. It is expected that the oxygen sites on the silica directly attract Li+, releasing the TFSI- coordination. The lithium species, loosing TFSI- anions, kept a random walk feature in the gel without the diffusion restriction attributed from the strong chemical and morphological effect as that in the gel with the polymer. We can conclude that the silica dispersion is a significant approach to provide the appropriate lithium ion condition as a charge-transporting species in the ionic liquid electrolytes.

Introduction Ionic liquids have been used for several fields of chemistry, such as reaction and extraction solvents for organic and polymer syntheses, catalysis, and extraction for purification.1 They are pure salts of the liquid phase at room temperature and are characterized by the low vapor pressure, inflammability, and thermal and chemical stability. Taking advantage of these attractive features, applications of the ionic liquids to electrochemical devices such as electrolyte solvents and capacitors have been activated.2 In addition, developments of new ionic liquids for improving the device performance were emphasized recently.3-5 In the field of lithium secondary batteries, it is expected that the ionic liquids can be potential substitutions for combustible organic solvents such as propylene carbonate (PC) and ethylene carbonate (EC) from the aspect of increasing safety. In the process of practical applications of ionic liquids to electrolyte materials, we are confronted with some problems, high viscosity that causes low ionic mobility and chemical instability in oxidation and reduction processes at the interface with electrodes. For the latter, structural improvements by substitution and the change of the cation and anion combination have been performed. It is reported that the alkylation of protons on the imidazolium salt effectively lowers the reduction potential and represses decomposition at the limiting potential.6,7 High viscosity of ionic liquids compared with that of the conventional organic solvents are mainly attributed to the Coulombic interaction among the ionic species in the solvent. Further, size and steric factors of the cation and anion affect the viscosity. We also investigated that the degree of ion orientation in the liquid * To whom correspondence should be addressed. E-mail: [email protected]. Tel.: +81-72-751-4527. Fax: +81-72-751-9623. † National Institute of Advanced Industrial Science and Technology. ‡ The Nippon Synthetic Chemical Industry Company Limited.

phase is a dominant factor of the viscosity of alkylimidazolium fluorohydrogenate-type ionic liquids.8 Combination of imidazolium cations and smaller (HF)nF- anions provides highly ordered structure. In the salt, a domain particle composed of several oriented cations and anions could be a unit of the hydrodynamic species responsible for charge transport. The larger the domain particle size due to high orientation of the ions, the lower the charge density of the particle, which leads to the lower Coulombic effect among the particles and macroscopically lower viscosity of the salt. Therefore, one of the ways to control the viscosity of the salt is to search for an appropriate combination of the cation and anion which can provide ordered configuration. However, a more serious problem that should be considered for lithium electrolytes is the existing condition of the lithium species in the ionic solvent. In the analogy of solvated lithium ions in the organic solvent as, for example Li(PC)n+, lithium salt can be dissolved by solvation of the ionic species in the ionic liquid solvent. If it is, we can no longer apply the concept of “dissociation” for the lithium salt in the ionic liquid solvent because the solvation would increase the anion coordination number on a Li+ cation as, for example, Li(TFSI)n+1n- (TFSI-: N(CF3SO2)2-, bis(trifluoromethanesulfonyl)imide), which could be recognized as an associated state compared with the initial state of LiTFSI. Quantitative investigation of the real situation of lithium ions in the ionic solvent is a significant point for designing the electrolyte with high ionic mobility and the cell device with an efficient charge-transport system. Another significant feature of ionic liquids is that they are not ionized completely in the equilibrium state.8-11 The ionization rate depends on the types of salts and the temperature. Addition of a lithium salt to the ionic liquid again changes the ionization rate. We have to systematically consider the effect of the ionization rate of the mixed salt on the individual ionic condition.

10.1021/jp072998r CCC: $37.00 © 2007 American Chemical Society Published on Web 09/15/2007

Ions in Lithium Electrolytes with Ionic Liquid Solvent

Figure 1. Temperature dependence of diffusion coefficients of the ionic species of LiPF6/BMI-TFSI. (a) BMI+ (O) and TFSI- (b) of 0 wt % LiPF6, BMI+(4) and TFSI- (2) of 3 wt % LiPF6, and BMI+(0) and TFSI- (9) of 10 wt % LiPF6. (b) Li+ (O) and PF6- (b) of 3 wt % LiPF6 and Li+ (4) and PF6- (2) of 10 wt % LiPF6.

We have already developed an effective and original NMR technique to measure the dynamic feature of ions in the electrolyte materials accompanied with a new theoretical model for the estimation of static features such as the dissociation degree of the salt and the interaction force between the ions and medium.12,13 In this approach, a constant electric field is applied to detect the ions of the probed nucleus selectively. Different from the conventional electrolytes with the neutral solvent, it is principally hard to measure the lithium ion mobility in the ionic liquid solvent using this technique because not only the target lithium ions but also the solvent ions drift in response to the applied electric field. This changes the equilibrium state of the electrolyte solution, which is a meaningless situation for the identification of mobility. Therefore, we confirm that the diffusion coefficient measurement, which needs no perturbation affecting the ionic condition, is the effective way for detecting the ion migration feature in the realistic condition of the electrolyte. On the basis of the systematic elucidation of the ionic state of lithium ionic liquid electrolytes through the measurements of dynamic properties, we propose the necessity of a second medium dispersion for the electrolyte in order to provide a suitable situation of lithium ions as charge-transport species. Experimental Section Two types of ionic liquids, 1-butyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (BMI-TFSI) and 1-butyl-2,3-

J. Phys. Chem. B, Vol. 111, No. 40, 2007 11795 dimethylimidazolium bis(trifluoromethylsulfonyl)imide (BDMITFSI), were prepared as follows. The 1-butyl-3-methylimidazolium bromide and 1-butyl-2,3-dimethylimidazolium bromide were prepared by the reaction of the respective 1-methylimidazole and 1,2-dimethylimidazole with 1-bromobutane in acetonitrile solvent. Disappearance of imidazole was confirmed using high-performance liquid chromatography (HPLC). Each bromide was mixed with an equimolar amount of K-TFSI in water at 50 °C, and BMI-TFSI and BDMI-TFSI were extracted from the water layer. The products were washed with distilled water several times. They were identified by 1H and 13C NMR and the capillary electrophoretic method, confirming the 98.4% of the cation species. The alkylation of the C-2 position of BMITFSI to form BDMI-TFSI lowered the reduction potential, and the lithium decomposition was reduced at 0 V versus Li/Li+.6,7 LiPF6 or LiTFSI was dissolved in BMI-TFSI and BDMI-TFSI with changing concentration from 3 to 10 wt % for LiPF6 and 3 to 24 wt % for LiTFSI; 10 wt % of LiPF6 and 24 wt % of LiTFSI were the maximum dissolved concentrations for the ionic liquid solvents. This shows that the salts with the same anion are more soluble with each other. In order to prepare gel samples, urethane acrylate (UA) oligomer or silica was mechanically mixed with the lithium electrolyte solutions until the gel was transparent. UA oligomer was synthesized from 2-hydroxyethyl acrylate, isophorone diisocyanate (IPDI), and P(EO/PO) (molecular weight, M ∼ 3000). Polymerization was performed for the gel with the UA using UV radiation by mixing methoxypolyethylene glycol monoacrylate (M ∼ 600) as a polymerizable viscosity reducer and 1-hydroxy phenyl ketone as a photoinitiator. The detailed procedure of the synthesis and the polymer network structure are reported in ref 14. Silica powders of hydrophilic and hydrophobic nanosize SiO2 (AEROSIL300 NIPPON AEROSIL, primary particle size: ∼7 nm, specific surface area: 300 ( 30 m2/g) were used for preparation of another type of gel. For the comparison of ion migration feature in the gels with the UA and silica, the mixing weight ratio of LiTFSI/EMI-TFSI (or EDMI-TFSI)/dispersant was fixed at 0.71:0.09:0.20. Both gels were finally transparent, indicating the homogeneous form. The storage modulus (E′) of the gel electrolyte membranes was measured by using a DVA-225 (ITK Co. Ltd) at 25 °C in the frequency range of 1.6-66 Hz. The shear modulus (G′) was measured by using an AR-1000 (TA instruments) at 25 °C in the frequency range of 0.01-100 Hz, applying a shear stress of 1000Pa. A relation of E′ ) 3G′ was applied to compared the elasticity of the gels with silica and the polymer. Diffusion coefficients of the ionic species were measured using the pulsed gradient spin-echo NMR (PGSE-NMR) technique with a JNM-ECP300W spectrometer and wide-bore probe units adjusting the probed nuclear species, 1H (300.5 MHz), 7Li (116.8 MHz), and 19F (282.7 MHz). In this research, the stimulated echo sequence was used for the measurement.15 The half-sine-shaped gradient pulse was applied twice in sequence after the first and third 90° pulses to detect attenuation of the echo intensity according to the diffusive migration of the probed species. The values of the field gradient pulse parameters were g ) 3-9.5 T/m for pulse strength, δ ) 0-8 ms for the pulse width, and ∆ ) 80 ms for the interval between the two gradient pulses. The diffusion coefficient was measured in the temperature range of 25-70 °C. The NMR spectrum of 1H comprises several peaks attributed to the proton configuration on the imidazolium cation species.8 We estimated the diffusion coefficient from the attenuation of each peak and finally took the average of them to determine a diffusion coefficient value

11796 J. Phys. Chem. B, Vol. 111, No. 40, 2007

Saito et al.

of the cation species. The NMR spectrum of 7Li in the electrolyte solutions was a single peak. After gelation of the electrolyte with the polymer or silica dispersion, a few peaks appeared. This reveals several lithium ion species of different surrounding situations and/or ionization conditions in the gel electrolyte. Therefore, we first separated the peaks by the curve fitting and then evaluated the echo intensity change of the individual peak to elucidate the diffusion manner or, in case of the random walk feature, to estimate the diffusion coefficient, Dobs. Results and Discussion Description of the Ionization Condition using the Ratio of Observed Diffusion Coefficients. In order to evaluate the individual ionic situation in the lithium electrolyte with the ionic liquid solvent, we first selected the combination of the solute, LiPF6, and solvent, BMI-TFSI (LiPF6/BMI-TFSI system), which gives different cation and anion species between the salts. Figure 1 represents the temperature dependence of the diffusion obs + 7 coefficients Dobs Li of the Li species probed by Li, the DPF6 and obs DTFSI of the respective PF6- and TFSI- species probed by 19F, and the DBMIobs of the BMI+ species probed by 1H of LiPF6/ BMI-TFSI (0, 3, and 10 wt % of lithium salt). Three 19F peaks with the chemical shifts of -80, -82, and -89 ppm were observed. By comparison with the 19F peaks of BMI-PF6 (-78 and -82.5 ppm) and BMI-TFSI (-89 ppm) measured independently, the peaks at -80 and -82 ppm were assigned to PF6-, and the peak at -89 ppm was assigned to TFSI-. With increasing LiPF6 concentration, the diffusion values of all species decreased due to the increase in the solution viscosity. In addition, the observed diffusion values reflect the ionization condition of the salts. In practice, Dobs has a contribution from not only the naked ion but also the ion pair and solvated ions coexisting in the electrolyte. The ionization rate of BMI-TFSI is not necessarily unity, and after dissolving LiPF6, the rate would change associated with the ionization condition of LiPF6.8 When we simply assume the equilibrium state of the individual salt dissociation

LiPF6 a Li+ + PF6BMI-TFSI a BMI+ + TFSI-

(1)

observed diffusion coefficients can be represented as

Dobs Li ) x1DLi+ + (1 - x1)DLiPF6 Dobs F1 ) x1DPF6- + (1 - x1)DLiPF6 Dobs H ) x2DBMI+ + (1 - x2)DBMITFSI Dobs F2 ) x2DTFSI- + (1 - x2)DBMITFSI

(2)

where subscripts Li, H, F1, and F2 represent the probed nuclei of 7Li, 1H, and 19F of the PF6- signal and 19F of TFSI- signal, respectively, DLi+, DPF6-, DLiPF6, ... without superscripts represent the inherent diffusion coefficient of Li+, PF6-, LiPF6, ..., respectively, and x1 and x2 are the dissociation degrees of LiPF6 and BMI-TFSI, respectively in eq 1. In the LiPF6/BMI-TFSI system, x1 and x2 are correlated with each other depending on their mixing concentration and temperature. Because Dobs does not exactly correspond to the inherent dynamic feature of the individual species as shown in eq 2, we cannot speculate the ionization condition from the behavior of Figure 1. We here

obs Figure 2. Observed diffusion coefficient ratios Dobs H /DF1 (open obs symbols) and Dobs /D (filled symbols) of the LiPF /BMI-TFSI 6 Li F2 system with several concentrations of LiPF6.

obs propose the evaluation of relative diffusion values, Dobs Li /DF1 obs obs and DH /DF2 . From eq 2, we are led to the relation

obs Dobs Li /DF1 )

obs Dobs H /DF2 )

{

{

DLi+/DPF6- x1 ≈ 1 x1 ≈ 0 1

DBMI+/DTFSI- x2 ≈ 1 x2 ≈ 0 1

(3)

When the dissociation of the salt is promoted (x1 ∼ 1, x2 ∼ 1), the ratio is close to that of the diffusion values of the cation and anion, DLi+/DPF6- or DBMI+/DTFSI-. On the other hand, when the dissociation degree of the salt is low (x1 ∼ 0, x2 ∼ 0), the ratio is close to unity as the associated species are dominant. That is, it is expected that the ionization condition of the salts can be evaluated by the degree of deviation from unity of DLi+/ DPF6- or DBMI+/DTFSI-. Figure 2 represents the plot of the diffusion value ratios DLi+/DPF6- and DBMI+/DTFSI- of the LiPF6/ BMI-TFSI system. The ratio of the solute LiPF6 is below unity, and the ratios of the solvent BMI-TFSI are above unity. We have already confirmed that DLi+ < DPF6- for LiPF6 in several organic solvents and the DBMI+ > DTFSI- for BMI-TFSI in the single salt system.16-19 Therefore, it is acceptable that increasing obs obs obs the separation of Dobs Li /DF1 and DH /DF2 from unity indicates the promotion of salt dissociation. It is characteristic that the obs Dobs H /DF2 of the ionic solvent increased with the increase in the LiPF6 concentration. This change apparently attributes to the dissolved LiPF6. It is presumed that the dissolved lithium attracted TFSI- anions and that the BMI-TFSI dissociation was obs promoted. On the other hand, Dobs Li /DF1 increased with the increasing LiPF6 concentration. This change may correspond to the enhanced association of LiPF6, assuming the simple equilibrium state of the salt dissociation of eq 1. However, a lithium salt dissociation proceeds by the solvation effect of the solvent. On the basis of this idea and the promoted dissociation result of BMI-TFSI by LiPF6 dissolution, it is reasonable to consider the solvation effect of the TFSI- anions from BMIobs TFSI on the lithium species for the increase of Dobs Li /DF1 with LiPF6 concentration. In order to recognize the situation of the ionic condition in more detail, we then selected the electrolyte systems with the salts of the common anion, LiTFSI/BMI-TFSI and LiTFSI/ BDMI-TFSI. This is because the salts of the common anion TFSI- are soluble with each other in a wider concentration range, which is effective to systematically evaluate the concentration dependence of the ionic condition of the salts. In

Ions in Lithium Electrolytes with Ionic Liquid Solvent

J. Phys. Chem. B, Vol. 111, No. 40, 2007 11797 obs Dobs H /DF

TABLE 1: Diffusion Coefficient Ratio of the Cation and obs Anion Species Dobs Li /DF of Several Lithium Electrolytes BMI-TFSI Li+/TFSI(mole ratio)

LiTFSI conc./wt % 3 10 15 17 24

‚[H1-a]: x2 ∼ 1, x1 ∼ 1

BDMI-TFSI Li+/TFSI(mole ratio)

1:23 1:7.1 1:4.9 1:4.3 1:3.2

Dobs H Dobs F

1:22 1:6.9 1:4.7 1:4.2 1:3.1

Dobs F1

Dobs H

Dobs H Dobs F

Dobs F2

) x2DTFSI- + (1 - x2)DBMITFSI

)

x2DBMI+ (2x2/3)DTFSI- + (1/3)DLiTFSI

)

DBMITFSI (x1/3) DTFSI- + (2/3)DBMITFSI

‚[H2-b]: x2 ∼ 0, x1 ∼ 0 Dobs H Dobs F

)

DBMITFSI (1/3)DLiTFSI + (2/3)DBMITFSI

(6)

We found [H1-a] < [H1-b] and [H2-a] < [H2-b] as DBMI+ > DTFSI- > DBMITFSI, which was confirmed experimentally as obs explained above. That means Dobs H /DF increases with decreasing x1. Similarly, [H1-a] > [H2-a] and [H1-b] > [H2-b] are obs accepted. That means Dobs H /DF decrease with decreasing x2 obs Dobs Li /DF

‚[Li1-a]: x2 ∼ 1, x1 ∼ 1

) x1DTFSI- + (1 - x1)DLiTFSI

) x2DBMI+ + (1 - x2)DBMITFSI

(x1/3)DTFSI- + (2x2/3)DTFSI-

‚[H2-a]: x2 ∼ 0, x1 ∼ 1

Dobs F

)

x1DLi+ (x1/3)DTFSI- + (2x2/3)DTFSI-

‚[Li1-b]: x2 ∼ 1, x1 ∼ 0 Dobs Li (4)

As in the case of LiPF6/BMI-TFSI, x1 and x2 are correlated with each other depending on the mixing ratio of LiTFSI and BMITFSI and the temperature. As we explained above, we could obs not distinguish Dobs F1 and DF2 . Therefore, the practically obobs served diffusion value DF can be explicitly represented as

Dobs F )

Dobs F

Dobs Li

) x1DLi+ + (1 - x1)DLiTFSI

Dobs H

x2DBMI+

‚[H1-b]: x2 ∼ 1, x1 ∼ 0

addition, we do not need to consider the possibility of reassociation forming LiTFSI and BMI-PF6 in the LiPF6/BMI-TFSI system, which makes it complicated to identify the ionic situation. In the LiTFSI/BMI-TFSI or LiTFSI/BDMI-TFSI system, we could not distinguish the TFSI- anion echo signals from LiTFSI and BMI-TFSI or BDMI-TFSI due to the fast exchange of them in the equilibrium state. Furthermore, estimation of Dobs from the single 19F peak gave a single F component of decay. These results indicate that all TFSI- anions are equivalent in the electrolyte and have no index of the original salts. We expect that these systems are simple for modeling and evaluating the solvation effect of the anion species on the lithium. By the analogy with LiPF6/BMI-TFSI system, we can obs obs obs speculate that the ratios Dobs Li /DF and DH /DF represent the ionization condition of the solute and solvent salts, respectively. Dobs F is contributed from some species such as TFSI , LiTFSI, BMI-TFSI, and Li+ solvated with some TFSI- in the equilibrium state. In order to prove the speculation, we first started the theoretical consideration on the LiTFSI/BMI-TFSI system. In the equilibrium state of the system, observed diffusion coefficients could be generally represented as follows

Dobs Li

)

1 × [(x1 + Rx2)DTFSI- + (1 - x1)DLiTFISI + 1+R R(1 - x2)DBMITFSI] (5)

where R is the molar ratio between LiTFSI and BMI-TFSI, which is induced from Table 1. In case of a dilute solution of LiTFSI, contribution from DLiTFSI to Dobs F is negligible from eq 5, and DFobs Z Dobs F2 . Therefore, same as the idea of the LiPF6/ obs obs obs BMI-TFSI system, Dobs Li /DF and DH /DF deviate from unity with the promoting of the dissociation of the salts. In the case of a high concentration of LiTFSI, we have to consider the from the three-component species. We contribution on Dobs F obs obs then derived Dobs /D and Dobs depending on the disLi F H /DF sociation condition of the salts as follows. We here put R ) 3, which corresponds to the highest concentration of LiTFSI in this research, to give shape to the equations

Dobs F

)

DLiTFSI (2x1/3)DTFSI- + (1/3)DLiTFSI

‚[Li2-a]: x2 ∼ 0, x1 ∼ 1 Dobs Li Dobs F

)

x1DLi+ (x1/3)DTFSI- + (2/3)DBMITFSI

‚[Li2-b]: x2 ∼ 0, x1 ∼ 0 Dobs Li Dobs Li

)

x1DLi+ (1/3)DLiTFSI + (2/3)DBMITFSI

(7)

We found [Li1-a] < [Li1-b] and [Li2-a] < [Li2-b] as DTFSI- > DLiTFSI > DLi+, explained above, and the consideration of the obs solvated condition of Li+. That means Dobs Li /DF increases with decreasing x1. Similarly, [Li1-a] > [Li2-a] and [Li1-b] > [Li2obs b], indicating that Dobs Li /DF increases with decreasing x2. That is, these evaluations are consistent with the speculation that the obs obs obs ratios Dobs H /DF and DLi /DF represent the ionization condition of the salts even for the mixed salt condition. The ratios deviate from unity with enhancing the salt dissociation. Furthermore, obs obs obs the fact that Dobs Li /DF and DH /DF change oppositely with the obs changes of x1 and x2, that is, for example, Dobs Li /DF increases obs with the decrease in x1 and increase in x2 and Dobs H /DF

11798 J. Phys. Chem. B, Vol. 111, No. 40, 2007

obs Figure 3. Observed diffusion coefficient ratios Dobs (open H /DF obs symbols) and Dobs /D (filled symbols) of the LiTFSI/BMI-TFSI Li F system with several concentrations of LiTFSI.

increases with the increase in x1 and decrease in x2, makes us imagine that the ratios reflect which cation (Li+ or BMI+) attracts the TFSI- anion closer by. obs obs obs Figure 3 represents the Dobs Li /DF and DH /DF with temperature of LiTFSI/BMI-TFSI with several LiTFSI concentrations. obs The plots above unity are Dobs H /DF , reflecting the extent of ionization of the BMI-TFSI solvent, and the plots under unity obs obs represent the Dobs of the solute LiTFSI. Dobs inLi /DF H /DF creased with decreasing temperature. This feature was observed more clearly in the sample with lower LiTFSI concentration. We can suggest two possibilities for this temperature dependence. One is simply due to the difference in the activation obs energies of Dobs H and DF with temperature, associated with the different migration mechanism between the BMI+ and TFSIspecies, and the other is due to the change of the ionization rate of BMI-TFSI with temperature, reflecting the change of obs the contributing species to Dobs H and DF with temperature. For the latter, we could explain as follows. Thermal motion of the species promotes equalizing of the interval space between the cation and anion. The highly oriented species are approximate to the covalent character. As a result, the association degree of obs BMI-TFSI is enhanced, and Dobs goes to unity with H /DF obs obs increasing temperature. DH /DF also increased with LiTFSI concentration. This result indicates that the dissolved lithium species is a trigger of BMI-TFSI ionization. That is, the TFSIanions are attracted to Li+ cations separated from BMI+ cations. Due to the attractive effect, it is reasonable to imagine a solvated species as Li(TFSI)n+1n-. After the solvation, residual BMITFSI, in the case of the low LiTFSI concentration, would be in the equilibrium state of dissociation under the presence of Li(TFSI)n+1n-. This speculation is supported from the change obs - is of Dobs Li /DF . For the lower LiTFSI concentration, TFSI present in several forms, Li(TFSI)n+1n-, BMI-TFSI, and TFSI-. obs Then, Dobs Li /DF ∼ DLi(TFSI)n/D[BMITFSI] < 1, in which D[BMITFSI] is contributed from BMI-TFSI and TFSI- in the equilibrium state. For the higher LiTFSI concentration, most of the obs TFSI- anions contribute to Li(TFSI)n+1n-. Then, Dobs Li /DF ∼ DLi(TFSI)n/DLi(TFSI)n ∼ 1. The temperature dependence of Dobs Li / for the lower LiTFSI concentration may reflect the Dobs F temperature dependence of the ionization rate of BMI-TFSI that is suggested as one of the possibilities above. Dissociated TFSIfrom BMI+ in the lower temperature region is relatively close to Li+, forming a highly solvated Li(TFSI)N+1N- (n < N). In this situation, Dobs F could be approximated to DLi(TFSI)m even for the lower LiTFSI concentration.

Saito et al.

obs Figure 4. Observed diffusion coefficient ratios Dobs (open H /DF obs obs symbols) and DLi /DF (filled symbols) of the LiTFSI/BDMI-TFSI system with several concentrations of LiTFSI.

TABLE 2: Molar Ratio of Li+ and TFSI- in LiTFSI/ BMI-TFSI and LiTFSI/BDMI-TFSI with Several Concentrations of the Lithium Salt obs Dobs Li /DF

1M LiTFSI - EC/DEC 0.5 M 1 M LiCF3SO3 - PC 0.5 M 0.35 M 1 M LiBF4 - EC/EMC 1 M LiPF6 - EC/EMC LiTFSI - BDMI-TFSI

0.80a 0.72 0.85b 0.74 0.70 0.86c 0.66c 0.3

a Estimation using the data of ref 20. b Estimation using the data of ref 21. c Estimation using the data of ref 16.

obs obs and Dobs of the Figure 4 represents the Dobs H /DF Li /DF LiTFSI/BDMI-TFSI system. It is characteristic that the temobs perature dependence of Dobs H /DF is small compared with those of the LiTFSI/BMI-TFSI system. It is possible that the ionization condition is associated with the reduction property of the ionic obs liquid. With increasing the LiTFSI concentration, Dobs H /DF increased, which is similar to the behavior of the LiTFSI/BMITFSI system. We can expect the same mechanism of TFSIattraction of Li species as that explained for the BMI-TFSI obs system. It is characteristic that Dobs Li /DF of BDMI-TFSI was almost constant, ∼0.3, independent of temperature and LiTFSI concentration, except for the solution with 24% of LiTFSI. The gradual decrease of the ratio of 24 wt % LiTFSI with the lowering temperature is due to the deviation of the diffusivity from the random walk behavior. Therefore, ∼0.8 of the higher temperature region is reliable for the 24 wt % LiTFSI. Dobs Li / Dobs F ∼ 0.3 for the lower LiTFSI concentrations is fairly small compared with those of the conventional lithium electrolytes as listed in Table 2. We can see that the ratios were around 0.7-0.8 with organic solvents for a 1 M lithium salt solution, apparently different from 0.3 of the ionic liquid electrolyte. The obs smaller value of Dobs Li /DF confirms the formation of the large solvated species, Li(TFSI)n+1n- and attribute to the size difference of Li(TFSI)n+1n- and TFSI- after solvation. It is expected that Li(TFSI)n+1n- in the ionic liquid solvent is fairly larger in size compared with, for example, Li(PC)n+ in the conventional PC solvent. In order to estimate the solvation number of Li+ in the ionic liquid solvent, we estimated the molar ratio of Li+ to obs TFSI- in Table 1. The abrupt increase of Dobs Li /DF from 0.3 to 0.8 in the LiTFSI/BDMI-TFSI system with increasing Li concentration corresponds to the change from four to three anions for a Li+. This means that at least three TFSI- anions

Ions in Lithium Electrolytes with Ionic Liquid Solvent contribute to lithium solvation in the system. In the solution with 24 wt % LiTFSI, most of the TFSI- anions from the BDMI-TFSI were used to form Li(TFSI)32-, and free TFSIobs anions were exhausted. This results in the situation, Dobs Li /DF = DLi(TFSI)3/DLi(TFSI)3 ∼ 1 because NMR signals probed by 7Li and 19F dominantly detect the same Li(TFSI)32- species. With decreasing the LiTFSI concentration, free TFSI- anions, distinguished from the TFSI- anions contributing to the lithium solvation and equilibrated with BDMI-TFSI, would increase. obs As a result, the contributing species to Dobs are Li and DF nLi(TFSI)n+1 and TFSI (including the contributions from obs BDMI-TFSI), respectively, and consequently, Dobs Li /DF ∼ 0.3. + Coordination conditions of Li and TFSI in this type of mixed salt have been investigated by the measurements of Raman spectra, and it was confirmed that three TFSI- anions per Li+ cation are coordinated, in which two of them are coordinated to more than one Li+ cation, and another TFSI- is coordinated to a single Li+ cation to be an associated structure.22 It is also reported that the most stable coordination of the TFSI- anion on the Li+ cation occurs via the oxygen atoms of the SO2 groups to form a chelate ring.23 The chelate form would be a cause of the larger solvated Li+ species compared with, for example, Li(PC)4+ composed of singly coordinated solvent species. We will practically estimate the size of the solvated lithium in the next paper. Through these considerations, we can revise the ionization condition of the LiTFSI/BDMI-TFSI system from the initial form of eq 1 as follows

LiTFSI + mBDMI-TFSI f Li(TFSI)n+1n- + mBDMI+ + (m - n)TFSImBDMI+ + (m - n)TFSI- h (m - n)BDMI-TFSI + nTFSI- (8) For the 3 wt % LiTFSI electrolyte, m = 21, and plenty of free BDMI+ and TFSI-, which are in equilibrium with the obs can be apassociated BDMI-TFSI, are present. Dobs Li /DF obs can be approximated to DLi(TFSI)n+1/DTFSI-, and DH /Dobs F proximated to DBDMI+/DTFSI-. For the 24 wt % LiTFSI electrolyte, m = 2, and free TFSI- species are not present as all TFSIanions from the BDMI-TFSI coordinate to Li+. As a result, obs obs ≈ DLi(TFSI)n+1/ DLi(TFSI)n+1 ∼ 1 and Dobs ≈ Dobs Li /DF H /DF DBDMI+/DLi(TFSI)n+1, which is larger than DBDMI+/DTFSI- as DLi(TFSI)n+1 < DTFSI-. This speculation can be applied to the LiTFSI/BMI-TFSI system of the higher temperature region. The solvated Li+ species in the ionic liquid solvent has a remarkable feature distinguished from that in the neutral solvent. That is, the net charge of the solvated lithium, Li(TFSI)n+1n- is apparently negative, as far as several anions contribute to solvation, opposite to the positive charge of the original Li+. This anionic species of lithium does not move toward the negative electrode only by the effect of the potential gradient in the charge-transport system. In practice, carrier drifting is dominated by the concentration gradient as well as the potential gradient in the field. Therefore, the negative Li(TFSI)n+1nspecies could move to the negative electrode when the concentration gradient effect is superior to the charge resistance.24 However, it is also the fact that the situation where the negatively charged Li(TFSI)n+1n- plays the part of charge transport is not essentially suitable from the aspect of the potential gradient field for charge-discharge cycles of the lithium battery system. We have to say that the ionic liquid solvent is not an appropriate medium in this situation. In order

J. Phys. Chem. B, Vol. 111, No. 40, 2007 11799 to increase the lithium mobility and prepare a more suitable carrier form in the ionic liquid medium, we need to take an approach that weakens the solvation effect of the ionic solvent on the Li+ cation. One of the ways we propose here is to disperse a second medium of polymer or silica. Composite Electrolytes Dispersed with Polymer or Silica. The purpose of a second solvent addition to the lithium electrolyte with the ionic liquid solvent here is to release the solvating TFSI- anions from the Li+, taking advantage of the attracting effect of the polar sites of the dispersant on Li+ or TFSI-. We selected two types of mediums for gelation, a network polymer with polar groups of the ether oxygen and silica powder. It is expected that the oxygen sites on the polymer or silica particles attract Li+ to weaken the coordination of solvating TFSI- anions. Different from the conventional electrolyte solutions or simple ionic liquids, which give typical random walk migration, carrier diffusivity in the gels showed some characteristic features attributed from the chemical and/or morphological properties of the gel network. Figure 5 represents the NMR echo signal attenuation of the composite LiTFSI/BDMI-TFSI/polymer probed by 1H, 19F, and 7Li attributed from the species including BDMI+, TFSI-, and Li+, respectively. Linear decay of the log plots of the 1H and 19F signal intensity against δ2∆ (δ: pulse width, ∆: duration of the pulses) indicates that the respective BDMI+ and TFSI- species follow random walk migration without any restricting barrier in the medium.25 On the other hand, the echo attenuation of the 7Li signal showed oscillating behavior, which represents the restricted manner in diffusion of the Li+ species.26,27 The NMR signal peak of 7Li in the gel was split into two to three components, changing from the single peak before gelation. Therefore, we first separated the signal into each component analytically using the ECP analytical program of JEOL Ltd. and plotted the peak intensity (area) decay for each separated component (Figure 5c and d). For the gel dispersed with a polymer, the two dominant components, which were meaningful for evaluating the attenuation, showed restricted diffusivity with different oscillation numbers as shown in Figure 5c. This indicates that (a) the polymer selectively affects lithium species geometrically and/or chemically different from the effects on the BDMI+ and TFSI- species and (b) at least two types of lithium under the different sites or a different environment are present in the gel electrolyte with the UA polymer. In the case of the gels prepared with the silica powder, we observed no significant difference in diffusivity between the gels with hydrophilic and hydrophobic silica. BDMI+ and TFSI- species showed a random walk feature the same as those of the gel with the polymer. However, an anomalous feature even different from that of the gel with the polymer was observed in the Li+ diffusion. The 7Li signal split into several components, and the two dominant components, which had enough intensity for decay analyses, showed an independent manner in diffusivity. One peak of the lower-intensity component showed an oscillation change that attributes to the restricted migration, and the other peak of the higher-intensity component attenuated linearly was responsible for random walk migration (Figure 5d). This result means that there are two types of lithium of different migration mechanism in the gel network of silica. In practice, we could not exactly reproduce the ideal restricted diffusion model in ref 28 using the restricted echo results observed here due to the difficulty of fitting the attenuation and oscillation behaviors simultaneously. When we dared to fit the data under the preferential consideration of the oscillation

11800 J. Phys. Chem. B, Vol. 111, No. 40, 2007

Saito et al.

Figure 5. NMR echo signal intensity attenuation of the (a) BDMI+ species probed by the 1H nucleus, (b) TFSI- probed by 19F, (c) Li+ probed by 7Li of LiTFSI/BDMI-TFSI/UA polymer at 25 °C, and (d) Li+ probed by 7Li of LiTFSI/BDMI-TFSI/silica at 25 °C as a function of γ2g2δ2(4∆ δ)π-2 according to the typical random walk relation, M ) M0exp(-γ2g2Dδ2(4∆ - δ)π-2). (a) and (b) decay linearly, indicating the random walk diffusion of the BDMI+ and TFSI- species. Both components of (c) and a single component of (d) showed an oscillating feature, indicating the restricted diffusion attributed from the site-interactive effect and physical barrier of the domain structure. Broken lines in the oscillating echo change are support for the realization of the attenuating behavior and, in the linear echo change, are least-square results for Dobs estimation.

frequency, the restriction size was estimated to be 4-5 µm, which corresponds to the aggregation size of the gel. Restricted diffusion is attributed to the carriers’ conflict with morphological barriers and/or interactions with chemical sites of the medium. A repeat of the increase and decrease in the echo signal intensity shows that the probed species goes back and forth between the barriers. We have already observed the phenomena in dry polymer electrolytes composed of a PEOtype polymer and a glucitol-type polymer.27,28 In the polymer network, oxygen sites on the polymer chains attract lithium through the Coulombic effect, which is the start of salt dissociation. Then, the solvated Li+ by the polymer sites migrates from site to site by partially disconnecting the coordination with the polymer solvent, taking advantage of the segmental motion of the polymer chains.29 This migration mechanism is the dominant process for Li+ migration in the dry polymer electrolyte. We have already said that the domain structure in the polymer electrolyte is responsible for restricted diffusion of lithium species.28 The domain is an aggregation of an entangled polymer chain due to the nonuniform polymer network. We have already observed that this aggregation of micron order was inevitably formed in the preparation process of the polymer and polymer gel electrolytes independent of the polymer type.27,28 As Li+ migrates on the sites along the polymer chain, Li+ diffusivity is influenced by the morphological feature of the chain entanglement and the consequential domain structure. We suppose this situation appeared even in the gels

observed here due to the selectively interactive effect of the oxygen site on Li+. obs Absolute values of the Dobs H and DF of the gels with the polymer and silica were the same order of magnitude, as shown in Figure 6. However, modulus values obtained from the elasticity measurement of the gel dispersed with silica were more than 1 order of magnitude higher than that of the gel with the polymer represented in Figure 7. As BDMI+ and TFSI- species are free from the interactive effect of the solid dispersant due to the linear echo decay, a large difference of G′ and the similarity of Dobs between the gels with the polymer and silica reveal that the network morphology which provides the carrier migration pathway is essentially different between the gels. This would be the cause of the anomalous Li+ diffusivity of two different manners in the gel with the silica. We finally elucidated the ionization condition of LiTFSI and obs BDMI-TFSI by gelation according to the plots of Dobs H /DF and obs obs obs obs DLi /DF in Figure 8, in which the DLi /DF of the gel with the polymer was unable to estimate because of the 7Li echo being attenuated restrictedly as in Figure 5c. We can find from this obs figure that the changing direction of Dobs H /DF by gelation was opposite between the gels with the polymer and silica. That is, obs polymer addition increased Dobs H /DF contrary to the diminishing of the ratio by silica dispersion. On the basis of the discussion for the LiTFSI/BDMI-TFSI system, an increase in obs Dobs H /DF means that the ionization of BDMI-TFSI is promoted

Ions in Lithium Electrolytes with Ionic Liquid Solvent

Figure 6. Temperature dependence of observed diffusion coefficients obs obs obs Dobs Li (b) and DF (9) of LiTFSI/BDMI-TFSI, DLi (O), DF (0), and obs obs DH (4) of LiTFSI/BDMI-TFSI/silica, and DF (shaded square) and Dobs H (shaded triangle) of LiTFSI/BDMI-TFSI/UA polymer.

Figure 7. Modulus values estimated from the elasticity of LiTFSI/ BDMI-TFSI/UA polymer (O) and LiTFSI/BDMI-TFSI/silica (0).

J. Phys. Chem. B, Vol. 111, No. 40, 2007 11801 anions. In practice, Dobs Li apparently increased by gelation with silica, although the macroscopic viscosity increased as in Figure 6. This means that SiO2 released solvated TFSI- anions partially from Li+ but not directly coordinated to Li+ to reduce the lithium mobility. The DLi increase due to the size reduction leads obs to the increased Dobs Li /DF . In the case of the gel with the polymer, on the other hand, we can expect that the solvated lithium species, Li(TFSI)n+1n- is, as it is, attracted to the ether oxygen site. This situation corresponds to indirect attraction of TFSI- anions and leads to the reduction of Dobs F . This may be obs the reason that Dobs /D increased after polymer dispersion. H F These results suggest that the silica is effective in weakening and breaking the TFSI- coordination on the Li+ cation to release the negative charge and affect the size of the Li(TFSI)n+1n-, causing low mobility. In conclusion, our investigation suggests that the lithium species in lithium ionic liquid electrolytes have a solvated form, Li(TFSI)n+1n-. Evaluation of Dcation/Danion revealed that the solvation number (n + 1) was approximated to 3-4 in the LiTFSI/BDMI-TFSI system. As a result, solvated lithium such as Li(TFSI)32- is apparently negative in charge, which resists going to the anode for lithium deposition, although it practically moves when the concentration gradient dominates over the potential gradient. This anionic form of lithium would not be profitable as a charge-transporting species in lithium batteries. If we dare to use the ionic liquid as a solvent of the lithium electrolytes, making use of the good merits of chemical stability and inflammability, we need to release the solvation bonds of TFSI- on Li+. This time, we took the way of dispersing a second medium in the electrolyte. Gelation of the lithium ionic liquid electrolyte by addition of polymer showed a restricted diffusive feature of the lithium species. This means that the oxygen sites on the polymer chains attract the lithium species. Boundaries of the domain particles, which are the aggregations of the entangled polymer chains, in the gel would be barriers for free migration, leading to the restricted diffusive feature. The lithium ion situation in the gel electrolyte by silica dispersion, on the other hand, was completely different from that in the polymer gel electrolytes. Dominant lithium species kept random walk obs behavior even after gelation. Evaluation of the Dobs H /DF and obs obs DLi /DF changes showed that release of the TFSI coordination on Li+ was promoted by silica. It is expected that the silica would attack the lithium directly and substitute for the TFSIanions. Random walk diffusivity of lithium in the gel is attributed to the fact that the silica prepares a homogeneous gel network with the electrolyte solution without local aggregation appearing in the gel with the polymer. References and Notes

obs Figure 8. Observed diffusion coefficient ratios Dobs (open H /DF obs obs symbols) and DLi /DF (filled symbols) of LiTFSI/BDMI-TFSI, LiTFSI/BDMI-TFSI/silica, and LiTFSI/BDMI-TFSI/UA polymer.

due to the attractive effect of the Li+ cation on the TFSI- anions. obs Therefore, a decrease in Dobs by silica addition reveals H /DF that the solvation coordination of TFSI- on Li+ was relaxed by silica, and consequently, the association of BDMI+ and TFSI- proceeded. This situation is supported by the change of obs obs Dobs Li /DF by the silica dispersion in which DLi is estimated from the major peak attenuation responsible for random walk obs migration in Figure 5d. That is, Dobs Li /DF increased from 0.3 to 0.7 by gelation with silica. We can speculate from this result that the silica species attack Li+ and release solvating TFSI-

(1) Welton, T. Chem. ReV. 1999, 99, 2071. (2) Ohno, H. Electrochemical Aspect of Ionic Liquid; Ohno, H., Ed.; John Wiley & Sons, Inc: Hoboken, New Jersey, 2005; p 1. (3) Wilkes, J. S.; Zaworotko, M. J. J Chem. Soc., Chem. Commun. 1992, 965. (4) Hagiwara, R.; Matsumoto, K. Electrochemical Aspect of Ionic Liquid; Ohno, H., Ed.; John Wiley & Sons, Inc.: Hoboken, New Jersey, 2005; p 227. (5) Xu, W.; Angell, C. A. Electrochem. Solid-State Lett. 2001, 4, E1. (6) Hayashi, K.; Nemoto, Y.; Akuto, K.; Sakurai, Y. J. Power Sources 2005, 146, 689. (7) McEwen, A. B.; Ngo, H. L.; LeCompte, K.; Goldman, J. L. J. Electrochem. Soc. 1999, 146, 1687. (8) Saito, Y.; Hirai, K.; Matsumoto, K.; Hagiwara, R.; Minamizaki, Y. J. Phys. Chem. B 2005, 109, 2942. (9) Tokuda, H.; Hayamizu, K.; Ishii, K.; Susan, M. A. B. H.; Watanabe, M. J. Phys. Chem. B 2004, 108, 16593. (10) Tokuda, H.; Hayamizu, K.; Ishii, K.; Susan, M. A. B. H.; Watanabe, M. J. Phys. Chem. B 2005, 109, 6103.

11802 J. Phys. Chem. B, Vol. 111, No. 40, 2007 (11) Tokuda, H.; Ishii, K.; Susan, M. A. B. H.; Tsuzuki, S.; Hayamizu, K.; Watanabe, M. J. Phys. Chem. B 2006, 110, 2833. (12) Kataoka, H.; Saito, Y. J. Phys. Chem. B 2002, 106, 13064. (13) Saito, Y.; Kataoka, H.; Murata, S.; Uetani, Y.; Kii, K.; Minamizaki, Y. J. Phys. Chem. B 2003, 107, 8805. (14) Jiang, G.; Maeda, S.; Saito, Y.; Tanase, S.; Sakai, T. J. Electrochem. Soc. 2005, 152, A767. (15) Kataoka, H.; Saito, Y.; Sakai, T.; Deki, S.; Ikeda, T. J. Phys. Chem. B 2001, 105, 2546. (16) Saito, Y.; Yamamoto, H.; Kageyama, H.; Nakamura, O.; Miyoshi, T.; Matsuoka, M. J. Mater. Sci. 2000, 35, 809. (17) Saito, Y.; Capiglia, C.; Yamamoto, H.; Mustarelli, P. J. Electrochem. Soc. 2000, 147, 1645. (18) Kataoka, H.; Saito, Y.; Uetani, Y.; Murata, S.; Kii, K. J. Phys. Chem. B. 2002, 106, 12084. (19) Noda, A.; Hayamizu, K.; Watanabe, M. J Phys. Chem. B 2001, 105, 4603. (20) Saito, Y.; Kataoka, H.; Capiglia, C.; Yamamoto, H. J. Phys. Chem. B 2000, 104, 2189.

Saito et al. (21) Saito, Y.; Yamamoto, H.; Makamura, O.; Kageyama, H.; Ishikawa, H.; Miyoshi, T.; Matsuoka, M. J. Power Sources 1999, 81-82, 772. (22) Castriota, M.; Caruso, T.; Agostino, R. G.; Cazzanelli, E.; Henderson, W. A.; Passerini, S. J. Phys. Chem. A 2005, 109, 92. (23) Gejji, S. P.; Suresh, C. H.; Babu, K.; Gadre, S. R. J. Phys. Chem. A 1999, 103, 7474. (24) Bockris, J. O’M.; Reddy, A. K. N. Modern Electrochemistry: Ionics, 2nd ed.; Plenum Press: New York, 1998; p 702. (25) Price, W. S. Annu. Rep. NMR Spectrosc. 1996, 32, 51. (26) Callaghan, P. T.; Coy, A.; Haplin, T. P.; MacGowan, D.; Packer, K. J.; Zelaya, F. O. J. Chem. Phys. 1992, 97, 651. (27) Kataoka, H.; Saito, Y.; Tabuchi, M.; Wada, Y.; Sakai, T. Macromolecules 2002, 35, 6239. (28) Saito, Y.; Hirai, K.; Katayama, H.; Abe, T.; Yokoe, M.; Aoi, K.; Okada, M. Macromolecules 2005, 38, 6485. (29) Watanabe, M.; Ogata, N. Br. Polym. J. 1988, 20, 181.