2614
J . Phys. Chem. 1984,88, 2614-2621
homologous series: acetonitrile, propionitrile, and butyronitrile. Acetonitrile has the greatest effect, propionitrile a lesser effect, and butyronitrile the least. That order is the same as the order of dielectric constants for the solvents, and one would expect that the greater the dielectric constant the more effective the solvent would be in solvating the ions. Benzene presents an especially interesting case. Because it is quite nonpolar, it is somewhat surprising that benzene is soluble at all in the totally ionic liquids, although Osteryoung and coworkers have reported that it was soluble in the butylpyridinium chloroaluminates. They found that it modified the electric conductivity and it affected proton and chemical shifts of the cation.'* Benzene is not as soluble in the basic melts as the series of nitriles (which are miscible in all proportions), but it has a greater effect on the chemical shifts of the cation protons in the range where it is soluble. This may be due to a more specific interaction between the imidazolium cation and benzene, such as the complex between the two aromatic species suggested for the butylpyridinium melts.I2 This could result in aromatic solvent-induced shifts, where the ring current of the benzene shields the protons of the imidazolium ~ a t i 0 n s . l ~ The dependence of chemical shifts on N in basic melts implies that proton N M R measurements could be used as an analytical
method for composition determination. Such a method has been reported recently.I6
Conclusions Proton N M R chemical shifts of the hydrogens located in the cations of organic chloride-aluminum chloride room-temperature molten salt mixtures are sensitive to composition. The changes in chemical shifts may be explained in terms of anion-cation interactions, where the imidazolium cation is distributed among five different anionetion complexes. Addition of lithium chloride or one of several organic solvents disrupts the complexes and changes the proton chemical shifts. Acknowledgment. We thank Lloyd H u g for running the NMR spectra reported here and Dr. Charles Hussey for many helpful discussions. L.A.K. gratefully acknowledges the support of Universal Energy Systems, Inc., Dayton, OH, and of the Air Force Wright Aeronautical Laboratories, Wright-Patterson AFB, OH. Registry No. AlC13, 7446-70-0; LiC1, 7447-41-8; ImCl, 1467-16-9; MeEtImCl, 65039-09-0; 1-methyl-3-butylimidazolium chloride, 7991790-1; butyronitrile, 109-74-0;propionitrile, 107-12-0; acetonitrile, 7505-8; benzene, 71-43-2.
Properties of 1,3-Dialkylimldazollum Chloride-Aluminum Chloride Ionic Liquids. 2. Phase Transitions, Densities, Electrical Conductivities, and Viscosities Armand A. Fannin, Jr., Danilo A. Floreani, Lowell A. King, John S. Landers, Bernard J. Piersma, Daniel J. Stech, Robert L. Vaughn, John S. Wilkes,* and John L. Williams The Frank J . Seiler Research Laboratory, US.Air Force Academy, Colorado Springs, Colorado 80840 (Received: February 18, 1983; In Final Form: April 4, 1983)
Experimental values are reported for the densities and electrical conductivities of several 1,3-dialkylimidazoliurnchlorides and for the densities, electrical conductivities, and viscosities of representative compositions of binary mixtures of these salts with aluminum chloride. These data were collected over wide temperature and composition ranges. The transport properties are reported as specific and equivalent conductivities and as kinematic and absolute viscosities. These data and the densities are interpreted in terms of a model of the structure of the binary melts. The solid-liquid phase diagram for the 1methyl-3-ethylimidazolium-aluminumchloride system was determined, including glass transition temperatures.
The work described here is a part of a continuing study designed to develop new low-temperature electrolytes for battery applications. This study of the phase transitions, densities, electrical conductivities, and viscosities of dialkylimidazolium chloridealuminum chloride binary melts is the second in a series of papers on the properties of these melts.' In an earlier paperZ from this laboratory the densities, conductivities, and viscosities were reported for several 1-alkylpyridinium chlorides and their binary mixtures with AlCl,. The dialkylimidazolium chlorides used in this study are shown in Chart I. The 1-methyl-3-ethylimidazolium chloride (MeEtImCl)-AICl, system was chosen as a "base-line" melt, primarily because of its relatively favorable condpctivity and viscosity behavior and its wide liquid range. Only a few compositions were studied for the other binary systems, in order to rule out unexpected anomalies. Studies of the proton nuclear magnetic resonance (NMR) spec(1) Fannin, A. A., Jr.; King, L. A.; Levisky, J. A.; Wilkes, J. s.J . Phys. Chem., preceding article in this issue. (2) Carpio, R. A.; King, L. A.; Lindstrom, R. E.; Nardi, J. C.; Hussey, C. L. J . Electrochem. SOC.1979, 126, 1644-50.
CHART I.
R +C I R,/N+N\R3
compd MeMeImCl MeEtImCl MePrImCl MeBuImCl BuBuImCl
R1 methyl methyl methyl n-butyl n-butyl
R3
methyl ethyl n-propyl n-butyl n-butyl
troscopy of these systems were presented in the first paper in the present series.'
Experimental Section Sample Preparation. The dialkylimidazolium salts and their binary mixtures with AlCl, were prepared as described elsewhere., All sample preparation and handling (except in sealed apparatus) were conducted in an argon-filled glovebox (Vacuum/Atmosphere Co. box and Model MO-40 DRI TRAIN), having a combined (3) Wilkes, J. S.; Levisky, J. A,; Wilson, R. A.; Hussey, C. L. Inorg. Chem. 1982, 21, 1263-4.
This article not subject to U.S. Copyright. Published 1984 by the American Chemical Society
1,3-Dialkylimidazolium Chloride-Aluminum Chloride moisture and oxygen concentration less than 10 ppm. Melting and Freezing Point Measurements. Melting points, and in some cases glass transition points, were determined by two methods: visually and by differential scanning calorimetry (DSC). Since most transitions were at subambient temperatures, a conventional melting point apparatus could not be used for the visual determinations. A temperature-controlled Dewar (Wilmad WG-821 variable-temperature insert, WG-836 transfer Dewar, and WG-838 heater) designed for an electron paramagnetic resonance (EPR) cavity proved to be convenient for measurements to -100 "C. Samples were sealed in quartz EPR tubes, and their temperature was controlled by a stream of dry nitrogen gas cooled by a glass coil immersed in liquid nitrogen. The phase transitions were observed with a magnifying glass through the double quartz walls of the Dewar. Many of the samples supercooled, so melting points, rather than freezing points, were taken as the true transition temperature. Glass transitions were clearly distinguished by an increase in viscosity with cooling, terminated by a sudden fracturing of the melt near the glass transition temperature. The DSC was done on a Perkin-Elmer DSC-2 calorimeter fitted with the subambient accessories. Temperatures below 30 "C were attained by using liquid nitrogen cooling with a helium purge of the DSC head. The instrument was calibrated with acetone and with water before and after each set of experiments. The samples were contained in stainless-steel large-volume pans which were loaded and sealed inside a glovebox. The temperature was scanned at 10 or 20 "C/min. Most compositions were observed both visually and by DSC. For melting point determinations, the two methods usually agreed within 1 "C, with a "worst-case" discrepancy, of 2 O C . The DSC also confirmed the visual determinations of glass transitions; i.e., a sharp peak corresponded to the melting point, while a change in base-line slope indicated a glass transition. Density Measurements. Densities were measured in sealed Pyrex dilatometric tubes whose volumes had been calibrated with mercury or distilled water in the conventional manner. Etched on each dilatometer was a reference mark midway up the stem. Samples which could be handled conveniently as liquids were loaded into dilatometers with bulbs on the bottom of a relatively small-diameter stem. The volume of this type of dilatometer to the reference mark was typically 6.5 cm3 and the cross section of the stem typically 0.085 cm3/cm. The remaining samples were loaded into straight tubes of typical volumes and cross sections of 1.5 cm3 and 0.24 cm3/cm, respectively. Weighings were made inside the glovebox. Loaded dilatometers were stoppered, removed from the glovebox, evacuated, and sealed with a torch. The dilatometers were placed in a B. Braun Thermonix Model 1420 water bath and temperatures monitored with an Air Force standard platinum reference thermometer. Estimated uncertainty in sample temperature was f0.05 "C. At temperatures above 90 "C the dilatometers were placed in a silicone oil constant-temperature bath wherein the estimated uncertainty in sample temperatures was f0.5 "C. The experimental measurements of sample volumes were made by measuring with a cathetometer the distance of the bottom of the meniscus from the reference mark. Cathetometer readings of the index mark and meniscus locations were made to a precision of f0.05 mm. Appropriate corrections were made in calibration and sample measurements for buoyancy, thermal expansion, and meniscus shape effects. All of the compositions studied exhibit negligible vapor pressures at the temperatures reached, and no volatilization corrections were made. Overall precision in density was estimated to be fO.l% and *I% for samples in the large and small dilatometers, respectively. Conductance Measurements. The same conductance cell was used for all samples. It was a Pyrex capillary approximately 0.5 cm long with a nominal i.d. of 0.05 cm. The capillary was sealed to a 0.6-cm4.d. Pyrex tube. Bright platinum wire coils were used. One was placed inside the larger Pyrex tube, immediately above the capillary, and the other coil was located on the outside surface of the capillary. A thermocouple was also inserted into the larger Pyrex tube. The assembly was immersed to approximately the
The Journal of Physical Chemistry, Vol. 88, No. 12, 1984 2615 same depth in small containers of each sample. Before each filling the cell was carefully cleaned by washing with acetonitrile and water and was dried in a 100 "C oven. The conductance cell was calibrated at 25 "C by using 0.1 demal aqueous KCl! The cell constant was corrected for thermal expansion as appropriate for each individual experimental measurement. The sample containers were loaded, and the conductance measurements made in the glovebox. The containers were immersed in a well-stirred mineral oil bath to a depth where the surface of the sample was at least 5 cm below the surface of the oil, in order to minimize temperature gradients within the sample. Temperature stability of at worst f0.05 "C was attained over the entire temperature range, and the actual temperatures were known to within f0.1 "C. A Bayley Model 124 proportional temperature controller was used. Overall precision in specific conductivity was estimated to be f l % . Conductance measurements were made at 1 kHz with a Beckman Model RC-18A conductivity bridge. Measurements at 1 and 3 kHz were identical within experimental error, so no frequency corrections were considered necessary. Viscosity Measurements. A closed, submersible, all-Pyrex viscometer was employed. It could be opened for emptying, cleaning, and refilling and then resealed. The viscometer was calibrated at various temperatures with cyclohexanol, ethylene glycol, and glycerol. Flow times for calibration varied between 19 and 4550 s. The calibration data were fit to a straight line passing through the origin of a time-log (kinematic viscosity) plot. The average deviation of calibration data from the line was i l % . The viscometer was mounted on a vertical platform submerged in a well-stirred silicone oil bath. The platform could be rotated in the vertical plane by remote control to fill the upper chamber of the viscometer. The passage of the liquid meniscus past two arrow marks above the capillary was timed with a stopwatch with a precision of better than i1.5%. At least six runs were made for each sample at each temperature, and the mean efflux times were used in the calculation of kinematic viscosity. The oil bath was equipped with submerged heaters, but the principal temperature control was achieved with submerged coils attached to a NESLAB Endocal Model RTE-9B refrigerated circulating bath. The oil bath could be maintained at temperatures ranging from -15 to 100 "C with a long-term stability of 50.5 "C. The actual sample temperatures were known to within hO.1 OC at near-ambient temperatures and f0.5 "C at the temperature extremes. When the above uncertainties were combined with those in composition and misalignment of the capillary from the vertical, an overall error in kinematic viscosity of i 2 . 5 % was estimated.
Results The concentrations of binary melts are expressed as the "apparent mole fraction", N, of AlCI3. This is the melt composition calculated as though the melt were comprised of monomeric AICI, and the appropriate imidazolium chloride. The molecular weight of the binary melts is given by M = MI,,,
+ 133.34N/(l
- N)
(1)
where MI,,, is the molecular weight of the dialkylimidazolium chloride. Notice that this is not the customary formula for the evaluation of the molecular weight of a binary molten salt mixture; rather, it allows for the fact that the number of ions present in the mixture is governed solely by the number of moles of imidazolium chloride present and is not changed by the addition of AlC13. It was not convenient to determine the densities and transport properties at an apparent mole fraction of AIC13higher than 0.666, for the melting temperature rose extremely rapidly with increasing AICI3 content in this composition region. Pure aluminum chloride is a nonelectrolyte, and its physical properties are not considered here; the density of pure aluminum chloride was reported earlier.5*6 (4) Jones, G.; Bradshaw, B. C. J . Am. Chem. SOC.1933,55, 1780-1800. (5) King, L.A.; Seegmiller, D. W. J. Chem. Eng. Dafa 1971, 16, 23-6.
2616 The Journal of Physical Chemistry, Vol. 88, No. 12, 1984
TABLE I: Melting, Freezing, and Glass Transition Temperatures N t.OC N t , OC N t,OC MeMeImCl and MeMeImC1-AlC1, 125 0.50 75 0.66 15 0.00 MeEtImCl and MeEtImC1-A1C13
0.10 0.15 0.20 0.25 0.30 0.31 0.32 0.32 0.33 0.33 0.34
84 72 62 54 41 39 32 28 21 -59' 19 -64' 17
0.00
60
0.00 0.05
0.36 0.38 0.40 0.41 0.42 0.42 0.44 0.46 0.48
-68' -72" -75' -27 -24
0.58 0.60 0.61 0.61 0.62 0.64 0.66 0.67 0.68 0.68 0.70 0.73
-80a
-19 -10 -3 7 -3 -9 -18
0.50
0.52 0.54 0.56
-23 -3 1 -33 -94' -95" -95" -96' -90' 45 -93' 79 110
Fannin et al. TABLE II: hast-Squares-Fitted Parameters for Eq 2 Densities
-1046, N
a, g/cm3
g/(cm3 "C) MeMeImC1-A1C13
0.0000, 0.33
1.1745 1.2738 1.3053 1.3725
0.50
0.66
0.66
0.2500 0.3333 0.3999 0.4500 0.5000 0.5499 0.6001 0.6666 0.0000 0.33 0.50 0.66
1.0918 1.1734 1.2348 1.3198
0.666were done in sealed tubes and ( 6 ) Seegmiller, D. W.; Fannin, A. A., Jr.; Olson, D. S.; King, L. A. J. Chem. Eng. Datu 1972, 17, 295-8. (7) Hurley, F. H.; Weir, T. P. J . Electrochem. Soe. 1951, 98, 203-6.
1.15
-
I
2
I
I
I
3
4
5
6
7
8
NUMBER OF ALKYL GROUP CARBONS
Figure 2. Dependence of density on imidazolium cation size presumably under significant AlCl, vapor pressure. We made no measurements on compositions with N > 0.73. Densities. The raw experimental data were converted into densities and are presented elsewhere (see paragraph at the end of the text regarding supplemental material). Experimental densities were least-squares fitted to equations of the form p = u b(t - 60) (2)
+
where t is the temperature in O C . The values of the fitted parameters are given in Table 11, together with the valid temperature range for each composition. No unusual features were found in the densities of individual samples. As expected, there was a relatively smooth density decrease as the size of the imidazolium cation increased. This is illustrated in Figure 2 for N = 0.50melts at 60 OC; similar behavior was noted at other compositions and temperatures. The composition dependence of the density of MeEtImC1-A1C13 melts is not a smooth function. When a and b from eq 2 are
The Journal of Physical Chemistry, Vol. 88, No. 12, 1984 2617
1,3-Dialkylimidazolium Chloride-Aluminum Chloride
TABLE III: Least-Squares-Fitted a (g/cm3) and 104b (g/(cm3 "C)) Parameters for Eq 3 Densities imidazolium chloride 02 a5 a11 u14 4 2 -bS -b8 MeMeImCl 1.2308 1.2808 1.3053 1.3518 1.3754 7.9765 7.6488 6.7460 8.0153 MeEtImCl 1.1279 1.2208 1.2662 1.3263 1.3567 6.1096 6.6170 MePrImCl 1.0891 1.1870 1.2937 1.3234 6.3867 6.7864 7.8879 1.2348 MeBulmCl 1.0873 1.1706 1.2113 1.2748 1.3069 2.0088 3.5039 7.6234 1.1391 1.1977 1.2273 BuBuImCl 1.0086 1.0962 6.1846 6.4536 7.1948 O O
-b, 1
-b14
8.2393 8.7004 8.5598 8.3704 7.8094
9.2609 9.1690 9.0195 8.8815 8.2299
g/cm3 0.002 0.0004 0.0005 0.0004 0.0003
u:
== [ ~ ( p c a l c-dp,,,,l)2/(number of points - number of fitted parameter^)]'/^.
1.30 m 8.0
\
\
I
0.3
I
0.4
I
I
I
I
05
I
0.6
MOLE F R A C T I O N A I C l z
Figure 3. Dependence of eq 2 parameters on MeEtImC1-A1C13 melt composition: points are a and b calculated at each experimental sample composition;solid lines calculated from eq 3; dashed lines calculated from a coordination number 1 model.
plotted as functions of N (from 0.33 to 0.66), structural features clearly can be seen, as is evident in Figure 3. In our earlier study of proton N M R spectroscopy of MeEtImC1-A1C13 melts,' we found that composition dependence of chemical shifts could interpreted in terms of a structural model. The observed chemical shifts were attributed to the presence of cation-anion "complexes". The structural model which best reproduced the chemical shift data assumed that each cation was associated with two anions, and each anion with two cations. The only cations, of course, are the imidazolium ions, while the anions are Cl-, AlC14-, and Al2CI7-. The possible species are ...Cl-...Im+...CI-..., ...C1-...Im+...AICl4-..., ...AlCl4-...Im+...A1C1,-.., ...AlC14-...Im+...Al,C1,-..., and Al2C1,-...Im+...Al2C1,-....These will be denoted below by the subscripts 2, 5, 8, 11, and 14, respectively, indicating the total number of chlorides in the anions associated with a given cation. This picture of the melt structure was discussed in detail in our previous paper.' As may be seen in Figure 3, the change in composition dependence of density between the acidic and basic composition regions is not as sharply delineated as for some of the proton chemical shifts. Even so, we assumed the densities of MeEtImC1-A1C13 binary melts could be represented by an analogue of eq 10 in that paper. Since both a and b in eq 2 are functions of composition alone, and not of temperature, the overall MeEtImC1-AlC1, density equation is p = X2az X5as X8a8 X l , a l , X,,a,, (X2b2 + X5b5 X8b8 X11b11 + X14bI4)(t - 60) (3)
...
+
+
+
+
+
where the ai's and b,'s represent the eq 2 parameters for the individual "complexes" with 2, 5, 8, 11, and 14 chlorides. The calculation of the Xi's was described in detail in our previous paper.' A computationally simpler method follows: Letfbe the anion fraction of AIC14-. Then,f= 1 - I(2N- 1)/(1 -MI. For 0 < N < 0.5, X 2 = (1 -A2, X s = 2f(l -j),and X l l and X14 = 0. For 0.5 < N 50.667, X , and X , = 0, X , , = 2f(l -A, and X14= (1 -A2. At all compositions X8 = f'. The experimental densities of the MeEtImC1-A1C13 melts were least-squares fitted to eq 3, and the ten fitted parameters are given in Table 111. Equation 3 with the fitted parameters reproduced the observed densities for MeEtImC1-A1C13 binary melts with a standard deviation, u, of 0.0004 g/cm3. Furthermore, each of the parameters appeared to have a reasonable value for its particular "complex". As was observed with the proton N M R data,
the structural model with coordination number 2 reproduced the data significantly better than did a coordination number 1 model (for which o = 0.003 g/cm3). This also may be seen in Figure 3. The densities of the AlC13 binary mixtures with the remaining four dialkylimidazolium chlorides were not measured at enough different compositions to fit the data sets directly to eq 3. Nevertheless, we wished to be able to calculate densities at intermediate compositions. If these binary liquids behaved as did MeEtImC1-A1C13, intermediate densities obtained by just linearly interpolating the existing data would most probably contain systematic errors. To avoid this, we assumed that the parameters for all complexes were similarly related to each other in all five binary systems. Thus, as was calculated in terms of a2 and as, a,, in terms of as and a14, bs in terms of b2 and b8,and b l l in terms of b8 and b14. For example, in the basic region (0 < N < 0.5), if the deviation, Abs, of b5 from a value linearly interpolated between b2 and b8 is Ab5 = b, - (l/2)(b8
+ b,)
(4)
and if we assume that the density curves have similar shapes and differ only in scale, then this deviation is proportional to the difference in b2 and b8. The plot of b vs. composition in the basic region (see Figure 3) may then be characterized by the ratio
B5 = Ab,/(b8 - b,)
(5)
This ratio is sufficient to describe the deviation from linearity of the density vs. composition plots. Now, to obtain consistent formulations for density equations among the various binary liquid systems, the value b, can be calculated from eq 4 and 5 by using b8and b2 for the mixture and Bs from the MeEtImC1-A1C13 system: bS
= B5(b8 - b2) + (1/2)(b8 + b2)
(6a)
Analogous equations can be obtained for a's in the basic region and for a's and b's in the acidic composition region:
- a21 + (1/2)(a8 + 0 2 ) bll = Bll(b8 - b14) + ( 1 / 2 ) ( b 8 + bl4) as = Adas
(6b) (6c)
= A l l ( a 8 - a141 + ( l / 2 ) ( a 8 + a141 (64 The values of the ratios A,, A , , , and B , , also were obtained from the MeEtImC1-A1C13 data. The experimental densities of each of the four remaining binary systems then were least-squares fitted to eq 3, where as, a l l , bS,and bll were replaced by the functions given as eq 6a-d. For each binary system the six fitted parameters and the four others calculated from them are given in Table 111. The standard deviations in the reproduction of the experimental densities with eq 3 also are given in Table 111. The standard deviations for all but the MeEtImC1-AlC1, system indicate the precision to be expected for the compositions actually examined and not necessarily for intermediate compositions. None of the pure 1,3-dialkylimidazoliumchloride data was used in the least-squares fitting to eq 3, primarily because of the limited temperature span of these data. Even so, it is gratifying to note that extrapolation of eq 3 to N = 0 yields densities for all but MeMeImCl that lie within the estimated uncertainty of the experimental values. Conductivities. The raw experimental data were converted into specific conductivities and are presented elsewhere (see paragraph all
2618 The Journal of Physical Chemistry, Vol. 88, No. 12, 1984
Fannin et al.
TABLE I V Least-Squares-Fitted Parameters for Eq 7 Specific Conductivities
0.00 0.33 0.50 0.66
102Ko, 104~,, 106 mho/cm mho/(cm "C) mho/(cg°C2) MeMeImCl and MeMeImC1-AlCI, -5.0593 15.8865 1.3204 1.4978 4.4066 4.6565 4.6323 8.4086 0.7210 3.1296 5.3115 1.7539
0.00 0.30 0.33 0.34 0.36 0.40 0.44 0.48 0.49 0.50 0.51 0.52 0.58 0.64 0.66
MeEtImCl and MeEtImC1-A1CI3 0.5485 3.0871 6.4687 1.0401 3.6487 3.3282 1.2276 4.0044 3.5180 1.3301 4.1150 3.4135 1S267 4.4399 3.3882 2.0239 5.0962 3.3497 2.7631 5.7540 2.5818 3.8661 6.7385 2.5529 4.1523 6.7567 2.1262 4.4799 7.0092 1.9378 4.3883 6.8528 1.7390 4.2454 6.6604 1.8959 3.6404 5.9530 2.1441 3.0935 4.8956 1.4423 2.9277 4.6334 1.3382
36 21 22 22 16 31 19 17 22 30 18 36 17 22
110 58 104 100 100 85 100 56 100 100 100 106 58 100 104
0.40 0.50 0.60
MePrImCl and MePrImC1-A1C1, 0.2087 0.7552 4.8038 1.2284 3.5281 2.7283 2.8821 5.4980 2.1200 2.5659 4.6650 1.6184
70 21 26 30
152 100 100 100
0.00 0.33 0.50 0.66
MeBuImCl and MeBuImC1-AlCI, 0.1231 0.1456 3.2260 0.5726 2.0957 2.2558 2.4127 4.6523 1.9272 2.0170 3.6113 1.3517
0.00 0.33 0.50 0.66
BuBuImCl and BuBuImC1-AIC13 0.2515 -0.9381 2.0592 0.2268 0.8820 1.3791 1.3407 2.9590 1.6109 1.4003 2.6490 1.0825
N
0.00
tmi",
tmam
153 80 80 20
164 106 110 100
52
"C
75 36 18 18 102 60 21 18
"C
110 100 100 100 110 100 100 100
at the end of the text regarding supplemental material). Experimental specific conductivities, K, were least-squares fitted to equations of the form
= KO
+ ~ , ( -t 60) + K Z ( ~ - 60)'
(7) where t is the temperature in "C. The values of the fitted parameters are given in Table IV, together with the valid temperature range for each composition. The dependence of specific conductivity on cation molecular weight was not as regular as for densities: the conductivity of any given MeEtImCl-AlCl, binary mixture is nearly as high as that of the corresponding dimethyl salt binary and very substantially higher than that of the methyl-propyl analogue. The heavier dialkyl analogues were progressively less conductive. This behavior is shown in Figure 4. Similar conductivity dependence on the alkyl groups also was observed for 1-alkylpyridinium chloride-AQ melts? Figure 4 also clearly shows that the specific conductivity is markedly dependent upon composition of the melt, with equimolar melts being the best conductors. (The dibutyl system appears to be an exception to this at lower temperatures, where the A1C13-rich melt is the better conductor.) In the composition region N C 0.5, the relative decrease in conductivity in going from MeMeImCl to BuBuImCl becomes even greater at lower temperatures. Figure 4 shows only the data at 100 OC, since at that temperature all of the binary systems are liquid over the entire composition range studied (except for pure MeMeImCl), but the same general behavior was noted at all temperatures investigated. Our earlier study2 with 1-alkylpyridinium chlorides also showed a similar dependence upon composition: at N = 0.5 1-ethylpyridinium chloride-AlC1, was a better conductor than K
MOLE FRACTION AICI,
Figure 4. Dependence of specific conductivity on R1R,ImC1-A1C13melt composition at 100 OC. MM, ME, MP, MB, and BB represent MeMeImC1-, MeEtImCI-, MePrImCI-, MeBuImCI-, and BuBuImCIAIC13, binaries, respectively. I
I
1
I
I
I
I
TEMPERATURE
I 0
I
\
i
0.4
0
-
9 0.4
\
5
0.0
i-
250c
-0.4
I
0.3
BO
AB. 0.3
-0.4
0 0 0.4
I
0.5
oc
0.5 0.7 MOLE FRACTION AICI,
0.6
0.7
MOLE FRACTION AICI,
Figure 5. Dependence of equivalent conductivity on R,R,ImCI-AlCI, melt composition and temperature: ( 0 )calculated and (0)extrapolated from eq 10 fits of individual samples; (-) (except for inset), locus of eq 10 simultaneous fit in composition and temperature;- (inset), interpolated values. MM, ME, MP, MB, and BB represent MeMeImCI-, MeEtImCl-, MePrImCI-, MeBuImCl-, and BuBuImCl-AlCI, binaries, respectively. either the N = 0 or N = 0.67 melts. Specific conductivities were converted into equivalent conductivities, A, by (8)
= KM/p
where M is from eq 1 and the densities, p, were calculated above. (The equivalent weights are the molecular weights of these binary systems.) The effects of temperature and composition on the equivalent conductivities of MeEtImC1-AlC1, mixtures are illustrated in Figure 5, as are 60 "C isotherms for the other systems which are liquid at that temperature. The composition dependence is similar to the behavior noted for specific conductivity, but the equivalent conductivity is less sensitive to the substitution of AI,Cl,for AlC14- as N approaches 0.666. As was observed with the l-alkylpyridinium-AlC13 melts,2 the equivalent conductivities did not display Arrhenius behavior, but they did follow the Vogel-Tammann-Fulcher (VTF) equation for conductivitys A = A A T 1 I 2exp[-k,/(T
- To)]
(9)
In this equation, A,, is a scaling factor dependent upon the units chosen to express the conductivity, Tois the "ideal" glass transition temperature, and kAis a constant characteristic of the material, (8)
Fulcher, G. S.J . Am. Ceram. SOC.1925,8, 339-55.
The Journal of Physical Chemistry, Vol. 88, No. 12, 1984 2619
1,3-Dialkylimidazolium Chloride-Aluminum Chloride TABLE V Least-Squares-Fitted Parameters for Eq 10, 11-13, and 16 Equivalent Conductivities and Absolute Viscosities of 1-Methyl-3-ethylimidazolium Chloride-Aluminum Chloride Binary Mixtures eq 10 os 16
eq 11-13 ”complex” species parameter
parameters
2
5
To, K k,,, K In A,,
259.5 493.4 8.0711
189.1 763.0 8.7437
To, K k,, K In A,,
252.3 7323 -4.669=
8
11
14
123.4 790.9 8.9753
126.5 802.1 8.9519
121.9 813.3 8.9253
137.2
132.2
TABLE VI: Least-Squares-Fitted Parameters for Eq 10 Equivalent Conductivities N To, K k,i, K In Ah lo4,“ MeMeImCl and MeMeImC1-AlCb 0.00 0.33 0.50 0.66
28 1.9 131.7 144.5 133.8
0.00
222.4
387 1297 678 775
Eq 10-13”
189.9
142.2
with khR being the formal analogue of the Arrhenius activation energy, where R is the universal gas constant. The general properties9 of glass-forming molten salts, and their glass transitions in particular,1° have been discussed elsewhere. The experimental values of A, obtained from eq 8 were fitted to the logarithmic form of the VTF equation, i.e. In A = - k A / ( T - To) - (1/2) In T + In AA
(10)
Every sample could be fitted very well to eq 10 (except N = 0 BuBuImC1, which was studied over a very small temperature range), and plots of the experimental In A (1 /2) In T vs. 1/( T - To)were straight lines. For the MeEtImC1-AlC13 binaries, the fitted values of the three parameters, -kh, In A,,, and especially To, were quite smooth functions of N . This suggested the possibility of modifying eq 10 to a single equation which would yield equivalent conductivities thoughout the experimental temperature and composition ranges investigated for MeEtImC1-A1C13 binaries. We chose to attempt to represent the equivalent conductivities in terms of the same structural model that reproduced the proton NMR chemical shifts and densities for this binary system, described above. We assumed the equivalent conductivities of MeEtImC1-A1C13 binary melts could be represented by an analogue of eq 3. All three fitted parameters in eq 10 are functions of composition alone and not of temperature. On a mole fraction scale consistent with the model, they can be expressed as weighted averages by
+
To = Xzaz
+ &a5 + &as + Xllal, +Xl4aI4
+
In A = X2c2+ X5c5 X8cs + Xllcll+ X14c14
(11)
(13)
where the a’s, and b’s, and c’s represent the To,k,,, and In A,,, respectively, for each of the individual “complexes” with 2, 5 , 8, 11, and 14 chlorides. For any individual sample, relatively small changes in the values of To correspond to quite large changes in k,, and In A , and vice versa; there was more scatter in the dependence of the latter two parameters upon N than there was for To. Initially, we obtained values of k,, and In A,, by fitting individual sample data to eq 10. Then we calculated the parameters in eq 12 and 13 by fitting these (9) Angell, C. A,; Tucker, J. C. In ’Physical Chemistry of Process Metallurgy-The Richardson Conference”; Jeffs, J. H. E., Tait, R. J., Eds.; Institute of Mining and Metallurgy: London, 1974; pp 207-14. (10) Angell, C. A.; Moynihan, C. T. In “Molten Salts Characterization and Analysis”; Mamantov, G., Ed.; Marcel Dekker: New York, 1969; pp 3 15-75.
716
9.029
11
MePrImCl and MePrImCl-A1C13 0.40 0.50 0.60
“The parameters for equivalent conductivity yield u = 0.02 in In (A/(cm2/(equiv ohm))), where u = [C(ln Aatlcd- In &,tl)2/(number of points - number of fitted parameter^)]'/^. *The parameters for absolute viscosity yield u = 0.03 in In (q/cP), where u = [Cln qcalcd - In q,,ptl)2/(number of points - number of fitted parameter^)]'/^. Averaged values of k, and In A , used for all compositions for absolute viscosity calculations.
6 18 1 8
MeEtImCl
0.00
Eq 11-13 and 16b
8.259 10.254 8.802 9.019
221.6 172.4 146.6 137.4
867 806 752 792
9.340 8.798 8.813 8.882
11 14 7 6
MeBuImCl and MeBuImC1-AlC1, 0.00 0.33 0.50 0.66
215.1 188.8 151.1 136.9
0.33 0.50 0.66
168.6 148.0 145.5
1024 798 725 789
9.565 8.483 8.609 8.762
13 11 19 6
8.957 8.654 8.480
16 6 13
BuBuImCl-A1C13 1129 844 758
“ u is defined in Table V. values of k g and In Ag. Using values of k,, and In A , from eq 12 and 13, we refitted the individual samples to eq 10 and obtained a new set of T i s . These latter values were nearly the same values as were obtained at first, but they even more smoothly varied with composition. The newly obtained T i s were then fitted to eq 11. They are shown in Figure 1, together with the locus of eq 11. Equivalent conductivities of MeEt1mC1-A1Cl3 at arbitrary compositions and temperatures within the ranges experimentally investigated may be calculated with eq 10, with the parameters To,kA,and In A,, calculated from eq 11-13. All of the fitted values of the parameters are given in Table V. This procedure reproduced the equivalent conductivities for the 103 experimental points with a standard deviation of 0.02 in In (A/(cm2/(equiv ohm))) (which may be compared to standard deviations of approximately 0.001 for fitting individual sample data to eq 10). As was observed with the proton N M R data’ and densities, the coordination number 2 structural model reproduced the data significantly better than did a coordination number 1 model (for which u = 0.04). This rather cumbersome method of calculating equivalent conductivities probably is not warranted for the remaining 1,3dialkylimidazolium chloride-A1C13 binary systems. The dependence of In A on N (at least for MeEtImC1-AlC1,) is much more nearly linear than were the chemical shifts and densities, especially at low temperatures. Furthermore, the data exist only at three compositions for each of the other binary systems, and the conductivity data are not as precise as were the NMR shifts and densities. Accordingly, in Table VI are the fitted parameters from eq 10 for the individual samples of the other four binary melt systems and of the pure dialkylimidazolium chlorides. The equivalent conductivities of intermediate compositions can be obtained by linear interpolation (in N) between the appropriate sample compositions. Viscosities. The raw experimental data were converted into kinematic viscosities and are presented elsewhere (see paragraph at the end of the text regarding supplemental material). The viscosities of the pure salts and of the two lower A1C13-content binaries with MeMeImCl were not measured because of the experimental difficulties which were anticipated in working with these relatively high-melting compounds in the present viscometer. As was the case with equivalent conductivities, the experimental kinematic viscosities did not display Arrhenius behavior but did follow the VTF equation for kinematic viscosity: In v = k , / ( T - To)+ (1/2) In T
+ In A ,
(14)
2620
Fannin et al.
The Journal of Physical Chemistry, Vol. 88, No. 12, 1984
TABLE VIII: Least-Squares-Fitted Parameters for Eq 16 Absolute Viscosities
TABLE VII: Least-Squares-Fitted Parameters for Eq 14 Kinematic Viscosities tmim
OC
k., K
To, K
N
-In A, MeMeImC1-A1C13 116.4 921 5.465
0.66
20
tmm
OC
205.1 186.5 111.0 148.O 151.6 128.8
12 11 11 11 11 0
90 90 90 90 90 90
0.33 0.50 0.66
195.9 145.2 164.0
MePrImC1-A1C13 I81 4.169 823 5.159 560 4.428
0 0 0
94 91 94
0.33 0.50 0.66
186.0 188.2 158.3
MeBuImC1-A1C1, 938 5.259 441 3.834 619 4.610
20 20 -1 1
90 95 95
0.33 0.50 0.66
161.4 159.1 154.0
BuBuImC1-A1C13 1213 5.619 150 4.161 102 4.114
20 -10 -10
95 90 90
0.50
0.61 0.66
3.0) I
I
I
I
I
I
I
u
I
5
s
I
I
0.5
1
I
0.6
MePrImC1-A1C13 811 4.139 819 5.151 608 4.351
11 29 9
0.66
MeBuImC1-AlC1, 185.0 958 5.208 184.5 484 3.819 154.8 666 4.531
5 4 I
0.33 0.50 0.66
165.4 156.5 150.1
BuBuImC1-A1C13 1261 5.193 192 4.81 1 150 4.165
12 4 6
I
is defined in Table V.
+
1
I
TEMPERATURE / 'C
I
194.8 141.5 160.0
the Arrhenius equation (over a somewhat smaller experimental temperature range).* The absolute viscosity data were handled in the very much the same way as were the equivalent conductivities. The viscosities of each individual sample were fitted to the VTF equation for absolute viscosity: In 9 = k , / ( T / T o ) (1/2) In T In A , (16) I
0.4
0.33 0.50 0.66
0.50
U
I
9
0.33
nu
0.51 0.3
MeMeImC1-A1C13 109.1 1024 5.444
94
MeEtImC1-A1C13 645 4.469 142 4.911 690 4.651 644 4.516 604 4.581 186 5.118
0.31 0.36 0.42
0.66
I
0.7
MOLE FRACTION AICI,
Figure 6. Dependence of absolute viscosity on RIR3ImC1-AlCl3melt composition and temperature: filled and open symbols calculated from eq 16 fits of individual samples; (-) for ME data, locus of eq 16 simultaneous fit in composition and temperature; for others, interpolated values. ME, MP, MB, and BB represent MeEtImC1-, MePrImC1-, MeBuImC1-, and BuBuImC1-A1C13 binaries, respectively.
where v is kinematic viscosity. This equation differs from eq 10 in the signs of the first two terms. The experimental kinematic viscosities were least-squares fitted to eq 14, and the values of the fitted parameters are given in Table VII, together with the valid temperature range for each composition. The kinematic viscosities were converted to absolute viscosities by 9 = PV (15) where 9 is the absolute viscosity. The dependence of viscosity on the size of the dialkylimidazolium ion is even more anomalous than was observed for conductivities. The MeEtImC1-AlC1, melts have substantially lower viscosities than do any of the other melts studied, including MeMeImC1-AlC1,. The effects of temperature and composition on the absolute viscosities of the MeEtImC1AlCI, mixtures are illustrated in Figure 6, as are 25 OC isotherms for the other systems which are liquid at that temperature. As would be expected, the absolute viscosities were "mirror images" of the equivalent conductivities. The absolute viscosities did not exhibit Arrhenius behavior. This is in contrast to the l-alkylpyridini~m-AlCl~ melts, which did obey
+
Furthermore, the MeEtImC1-AlC13 binary data were fitted to yield a single equation which gives In 9 at arbitrary compositions and temperatures within the ranges experimentally covered. These values may be calculated with eq 16 by using the parameters To, k,, and In A , from eq 11-13. The T i s obtained in this way and the locus of the absolute viscosity version of eq 11 are shown in Figure 1. (There was considerably more scatter in k , and In A , for viscosities than was observed for the conductivities. We found the best fit of eq 16 to the data was obtained at all compositions when the average values for all samples were used for k, and In A,.) This procedure reproduced the absolute viscosities for the 57 experimental points with a standard deviation of 0.03 in In (v/cP) (which may be compared to standard deviations of approximately 0.001 for fitting individual sample data to eq 16). All of the fitted parameters are given in Table V. Viscosity appears to be a less sensitive function of melt structure than were the conductivity, density, and proton N M R shifts. The structural model with coordination number 2 produced a standard deviation only very slightly better than for the coordination number 1 model. The former model is retained here for consistency in calculations. For the same reasons as were given above for equivalent conductivities, the data for the other binary melt systems were not fitted to combined temperature-composition equations. In Table VI11 are given the fitted parameters from eq 16 for the individual samples of the other four binary melt systems. Discussion The experimental glass transition temperatures, Tg,obtained from visual and DSC determinations, lie slightly above the "ideal" glass transition temperatures, Toobtained from fitting the VTF equation to conductivity and viscosity data. The two sets of glass transition temperatures vary in the same way with composition of the MeEtImC1-AlC1, binary system. These characteristics, which may be seen in Figure 1, apparently are typical of molten salts. For example, Easteal and Angell" observed similar behavior in the pyridinium chloride-zinc chloride system. The results of several studies were summarized by Angell and Moynihan'O in a comprehensive discussion of transport properties of low-melting molten salt systems. These authors have interpreted the composition dependence of TBand Toin terms of Coulombic cohesive energy changes in the (11) Easteal, A. J.; Angell, C. A. J . Phys. Chem. 1970, 74, 3987-99.
The Journal of Physical Chemistry, Vol. 88, No. 12, I984 2621
1,3-Dialkylimidazolium Chloride-Aluminum Chloride w
\
i
FROM ABSOLUTE VISCOSITIES \
FROM EQUIVALENT CONDUCTIVITIES
-
-
TEMPERATURE
OC 30
60 YO
-
c> 4
I
’0013
I
I
04
I
1
I
0 5
0 6
I
I 0 7
MOLE FRACTION AICI,
Figure 7. ”Activation energies” for viscous flow and for equivalent conductivity of MeEtImC1-A1C13 binary melts (locus of eq 17).
liquid as charge density varies with ionic size or as longer range particle interactions occur. The present data are consistent with these ideas. The values of Tg and To rise with increasing chloride concentration, which may be a reflection of increasing charge density and/or increasingly strong “complex” formation. In the chloride-deficient region, i.e., N 1 0.5, the anions are large, and Tgand To do not change appreciably with melt composition. The mass transport data presented here do not provide as strong an argument in support of the coordination number 2 structural model as did the N M R chemical shift’ and density data. On the other hand, the data certainly are consistent with such a picture of melt structure. One might expect the presence of a relatively small chloride ion to yield an enhanced mass transport in melts with high chloride ion concentrations. In fact, just the opposite effect was noted for both specific and equivalent conductivities and for kinematic and absolute viscosities, suggesting some sort of bonding together of species by the relatively high charge density of the chloride ions. (We noted a similar lowering of electrical conductivity when small amounts of LiCl were substituted for MeEtImCl.) Conductivities and viscosities were quite insensitive to the relative concentrations of the bulkier AlC14- and AlzC17ions. For any structural model to be believable, the parameters calculated in terms of the model must be physically reasonable. The equivalent conductivity and viscosity parameters presented in Table V for each of the five “complexes” are not only reasonable values individually but also are mutually consistent with one another. As will be shown below, the similarity of activation energies for electrical conductivity and viscous flow also supports the idea of some sort of simultaneous interaction among several particles in the melt and is consistent with the structural model proposed here. Although the mass transport properties investigated here do not obey Arrhenius behavior, an “activation energy”, E,may still be obtained from the expressions EA = RF(d In A/dT) or E, = RF(d In s/aT). Both expressions, when applied to eq 10 and 16,respectively, yield T
where E,k,and To are parameters for the appropriate transport property. These activation energies are of course temperature
dependent. Activation energies calculated from eq 17 for MeEtImC1-A1C13 melts are given in Figure 7 as functions of composition and at three temperatures. A comparison between the activation energies for viscous flow and for equivalent conductivity is of theoretical importance. In the case of several molten alkylammonium picrates, iodides, nitrates, and perchlorates, it has been shown that the activation energies for the two modes of mass transport were very similar, with the viscosity values from 0 to 10% larger.’* These similar activation energies have been attributed to similar configurational changes involved in the transport processes.13 Newman et al.I4 found that E, was only slightly greater than EA for l-methylpyridinium bromide and iodide. In our earlier work with l-alkylpyridinium halide-AlC1, binary melts? we found the two activation energies at 50 O C identical (within the experimental uncertainties) for N = 0.66 methyl-, ethyl-, and n-propylpyridinium chlorides and the values of E, 9% and 15% greater than EA for N = 0.66n-butylpyridinium chloride and ethylpyridinium bromide melts, respectively. In the present work, E,, is 14% and 17% greater than E A for N = 0.5 MeEtImC1-A1C13 at 30 and 90 OC, respectively. The relative difference is less a t lower N, and the two become indistinguishable as N approaches 0.66,as may be seen in Figure 7. The similarity in activation energies for the two mass transport modes for these low-temperature ionic liquids with large cations and/or anions is conspicuously different from what is observed with more familiar molten salts. In the case of several inorganic molten salts and molten salt mixtures, the energy barriers for viscous flow were found to be several times greater than for equivalent conductivity. The present interpretation of our physical property data in terms of ionic “association” or ”complexes” suggests that the barriers to mass transport by the conduction of electrically charged particles or by viscous flow may involve in both cases the disruption of forces involving several particles simultaneously. The temperature dependence of the activation energy for the two modes of mass transport also is consistent with the melt structure model. The increased energy at higher temperatures decreases the degree of association. Thus, as the temperature is increased, there will be fewer associated species, and a deviation from Arrhenius behavior will be noted. As was observed earlier,zs12the present activation energies become somewhat larger as the cation size increased. As was also observed with the 1 -alkylpyridinium binaries: the smallest cation in the present work displayed anomalously high activation energies.
’
Acknowledgment. L.A.K, gratefully acknowledges the support of Universial Energy Systems, Inc., Dayton, OH, and of the Air Force Wright Aeronautical Laboratories, Wright-Patterson AFB,
OH. Registry No. AIC13,7446-70-0;MeMeImC1,79917-88-7;MeEtImCl, 65039-09-0; MePrImC1, 79917-89-8; MeBuImCl, 79917-90-1; BuBuImCI, 83608-75-7.
Supplementary Material Available: Tables listing the densities, specific conductivities, and kinematic viscosities of binary mixtures of 1,3-dialkylimidazolium chloride and aluminum chloride (10 pages). Ordering information is given on any current masthead page. (12) Jam, G . J.; Reeves, R.D.; Ward, A. T. Nature (London) 1964,204, 1188-9. (13) Reinsborough, V. C. Reu. Pure Appl. Chem. 1968, 18, 281-90. (14) Newman, D. S.; Tillack, R. T.; Morgan, D. P.; Wan, W.-C. J. Electrochem. SOC.1977, 124, 856-60. (15) Harrap, B. S.; Heymann, E. Trans. Faraday SOC.1955,51,259-67.
