Article pubs.acs.


Making Sense of Enthalpy of Vaporization Trends for Ionic Liquids: New Experimental and Simulation Data Show a Simple Linear Relationship and Help Reconcile Previous Data
Sergey P. Verevkin,*,† Dzmitry H. Zaitsau, Vladimir N. Emel’yanenko, and Andrei V. Yermalayeu
Department of Physical Chemistry, University of Rostock, Dr.-Lorenz-Weg 1, 18059 Rostock, Germany

Christoph Schick†
Department of Physics, University of Rostock, Rostock, Wissmarsche Strasse 43-45, 18057 Rostock Germany

Hongjun Liu and Edward J. Maginn
Department of Chemical and Biomolecular Engineering, University of Notre Dame, Notre Dame, Indiana 46556, United States

Safak Bulut and Ingo Krossing
Institut für Anorganische und Analytische Chemie, Universitat ̈ Freiburg, Albertstrasse 21, 79104 Freiburg, Germany

Roland Kalb
Proionic GmbH, Parkring 18, Trakt H/1, A-8074 Grambach bei Graz, Austria
S Supporting Information *

ABSTRACT: Vaporization enthalpy of an ionic liquid (IL) is a key physical property for applications of ILs as thermofluids and also is useful in developing liquid state theories and validating intermolecular potential functions used in molecular modeling of these liquids. Compilation of the data for a homologous series of 1-alkyl-3-methylimidazolium bis(trifluoromethane-sulfonyl)imide ([Cnmim][NTf2]) ILs has revealed an embarrassing disarray of literature results. New experimental data, based on the concurring results from quartz crystal microbalance, thermogravimetric analyses, and molecular dynamics simulation have revealed a clear linear dependence of IL vaporization enthalpies on the chain length of the alkyl group on the cation. Ambiguity of the procedure for extrapolation of vaporization enthalpies to the reference temperature 298 K was found to be a major source of the discrepancies among previous data sets. Two simple methods for temperature adjustment of vaporization enthalpies have been suggested. Resulting vaporization enthalpies obey group additivity, although the values of the additivity parameters for ILs are different from those for molecular compounds.

Research on ionic liquids (ILs) and their industrial applications has expanded rapidly over the past decade. One of the most attractive features of ILs is their negligible vapor pressure at ambient temperatures. At elevated temperatures relevant for many applications of ILs, however, the vapor pressure is no longer negligible, even though it remains small (approximately at the level of a few Pa). Hence, for chemical processing with ILs, knowledge of the vapor pressure and vaporization o enthalpies, Δg l Hm, is indispensable. Moreover, vapor pressures and enthalpies of vaporization also play a crucial role in the development of liquid state theories, and so there is additional
© 2013 American Chemical Society


motivation to understand these properties for ILs. Experimental measurements of the vaporization enthalpies are extremely challenging because of two main problems. The vapor pressures of ILs at ambient temperature are so low as to be practically immeasurable, whereas at high temperatures where vapor pressures can be measured, possible thermal decomposition processes can distort the results. As a matter of fact, with the exception of the Knudsen method,1,2 traditional experimental
Received: November 19, 2012 Revised: March 27, 2013 Published: May 1, 2013
6473 | J. Phys. Chem. B 2013, 117, 6473−6486

However. Figure 1.3−10 A homologous series of the 1-alkyl-3-methylimidazolium bis(trifluoromethane-sulfonyl)imides or [Cnmim][NTf2] is beyond any doubt the most frequently investigated class of IL for the study of vaporization enthalpies. Chem.6 kJ mol−1). We have collected the available literature data on vaporization enthalpies of [Cnmim][NTf2] in Table S1 (Supporting Information (SI)) and present these data graphically in Figure 1. 117. it is unclear whether ILs follow the same monotonic pattern or not. the obvious questions are: What is the right trend within the “Milky Way” spread of the experimental results? Do vaporization enthalpies exhibit a typical linear increase in vaporization enthalpy with increasing chain length. It has been shown that the vaporization enthalpies for the reference compounds dx. and octyl. (□) LOSMS [3]. not without bias.3. Given all these conflicting results. combined with a QCM. in alkanes.9 were used for the experimental investigation. o Δg l Hm (298 K) chain length (n) dependence for [Cnmim][NTf2].1021/jp311429r | J.5−7 high-temperature UV spectroscopic technique (UV). The second method used a commercially available TGA for the vaporization enthalpy measurements.19 The TGA procedure has been carefully developed for very low volatility compounds. thereby lowering the vaporization enthalpy for the longer alkyl chains.4 thermogravimetry (TGA). B 2013. the Langmuir vaporization is significantly more sensitive since the total open surface is exposed to the QCM under vacuum conditions. the vaporization studies of ILs have been conducted at temperatures between 360 and 600 K.doi.8 and quartz crystal microbalance (QCM).95 ± 0. Phys. do they show a “bent up” increase or an aggregate-driven decrease beyond a certain chain length? We attempt to resolve this contradiction using a combination of precise experimental measurements and atomistic simulations to examine the vaporization enthalpies for the [Cnmim][NTf2] family with odd and even chain length of the alkyl-imidazolium cation. especially for the vaporization enthalpy. that the available vaporization enthalpies of [Cnmim][NTf2] are in total disarray. butyl-.17 but this was the first time such a phenomenon was used to explain10 an anomaly in an enthalpy of vaporization trend.9.16 and experimental studies. What is curious about the calorimetry data.10 Two distinct slopes in the vaporization enthalpy versus alkyl chain length curve were clearly observed for the [Cnmim][NTf2] ILs (+) calorimetry [13].18 The combination of high-vacuum conditions with the extremely sensitive QCM allows measurements of mass loss rates for ILs at temperatures down to 363 K.9. However. There are at least two reasons for this extended interest.9 ± 0.12 The second aspect is that the increasing length of the cation alkyl chain should lead to predictable monotonic structure−property relations such as those exhibited by molecular liquids such as alkanes. The vaporization enthalpy chain length dependence in Figure 1 resembles the “Milky Way” rather than any reasonable relationship.10 Such an aggregation phenomenon has been reported in several simulation15. hexyl-. For example.3 By contrast. (○) QCM-Knudsen [10]. This behavior was explained in terms of a structural percolation phenomenon in which the longer alkyl chains form aggregates in the liquid phase. It is apparent Article Figure 2. (◊) Knudsen effusion method [2]. The first method was a homemade set up for the Langmuir vaporization from the open liquid surface coupled with the QCM. the incremental increase 6474 in vaporization enthalpy per CH2 group is about half this (4. a very unusual (for a homologous series) and noticeably o nonlinear and even “bent up” dependence of Δg l Hm on alkyl chain length (in the imidazolium cation) was observed (see Figure 2) for ethyl-. it is well-known that these ILs have remarkable thermal stability such that they can be distillated without decomposition at elevated temperatures of 473 to 573 K. This has stimulated the development of new direct experimental methods such as temperature programmed desorption and line of sight mass spectrometry (LOSMS).14 Another and more sophisticated dependence of vaporization enthalpies was reported recently using the Knudsen effusion apparatus. Two recently developed methods7. Some possible trends in the vaporization enthalpy chainlength dependence.The Journal of Physical Chemistry B techniques for vapor pressure measurement have not been developed for extremely low volatility liquids such as ILs. however. some o definite trends in the Δg l Hm values of [Cnmim][NTf2] (see Figure 2) can be suggested and even justified.13 For comparison. is the unusually high contribution each additional CH2 group makes to the vaporization enthalpy (8.11.10 As a rule. 6473−6486 . albeit at a greater slope than alkanes? Or.derivatives determined by the Knudsen method2 and by LOSMS. In contrast to the conventional Knudsen method. the direct calorimetrically measured values of vaporization enthalpies13 for the same set of ILs demonstrated o (see Figure 2) the expected linear dependence of the Δg l Hm on the chain length.10 kJ mol−1). First. with the change in slope occurring at [C6mim][NTf2] (see Figure 2). Available literature data on the enthalpy of vaporization.

As to objective (b). Samples of [Cnmim][NTf2] with even alkyl chain numbers were commercially available from IoLiTec. In our study. the residual IL in the cavity and the IL-deposit on the QCM were analyzed by attenuated total reflectance-infrared (ATR-IR) spectroscopy. it depends on the vacuum conditions and possible deposits on the internal parts of the vacuum chamber. and NMR spectroscopy (1H and 19F nuclei). The conditioning was repeated until a reproducible mass loss within two consecutive runs was recorded. In order to reduce the impact of the background noise on the QCM. dx. Previous work from our group20 and others21 has shown that vaporization enthalpies can be reliably computed with MD. In order to detect a possible decomposition of IL under the experimental conditions. 117. The objectives of the present work were twofold: (a) to investigate structure−property relations within the [Cnmim][NTf2] family using experimental and computational methods. Chem. The QCM is placed directly over the measuring cavity containing the IL. No changes in the spectra before and after the experiment were detected for the ILs under study. to ensure the absence of lithium halides. The ILs were dried at 60 °C/10−3 mbar until water contents of <20 ppm were achieved (Karl Fischertitration). 6473−6486 . or 298 K. and the effect of background noise was checked. The value Δg l Cpm = Cpm(g) − Cpm(1) is the difference of the molar heat capacities of the gaseous o o Cpm (g) and the liquid phase Cpm (1). Isothermal mass loss rate dm/dt was measured in a few consecutive series with increasing and decreasing temperature steps. B 2013. As a rule.18 6475 ■ The frequency change rate df/dt was measured in a few consecutive series with increasing and decreasing temperature steps. After each run. TGA: Measurements of Vaporization Enthalpy. Since the aggregation phenomena used to explain the change in slope of the vaporization enthalpy occurring at [C6mim][NTf2] is observed in MD simulations. No changes in the spectra were detected with the ILs under study in this work. and Merck. The o temperature-dependent vaporization enthalpy ΔlgHm (T) is given by g o g o o Δg l Hm(T ) = Δl Hm(T0) + Δl Cpm(T − T0) (3) EXPERIMENTAL SECTION General Procedures. is 9 obtained by g o o Δg ⎛ df ⎞ 1⎞ l Hm(T0) − Δl CpmT0 ⎛ 1 ln⎜ T ⎟ = A′ − ⎜ − ⎟ ⎝ dt ⎠ R T0 ⎠ ⎝T + o Δg l Cpm R ⎛T ⎞ ln⎜ ⎟ ⎝ T0 ⎠ (2) with a constant A′. Isothermal mass loss rate dm/dt was monitored in the temperature range 480−620 K at a nitrogen flow rate of 140 mL·min−1.5−1% of the frequency change rate at the lowest temperature of determination. The value of Δf was measured as a function of time at different temperatures. The sample was stepwise heated and a mass loss of 0. 542 K. Raman. required for the proper comparison of the experimental data. SC is the area of the crystal.1−0. respectively. The change in the vibrational frequency of the crystal Δf is a measure of the amount of IL deposited on the cold QCM. In order to confirm the absence of decomposition of IL in the experimental conditions. The experimental procedure has been reported In short. Enthalpies of vaporization of the even and odd ILs have been measured using the QCM technique. Two simple and robust procedures will be suggested for the proper adjustment of the vaporization enthalpies to any required temperature. Preliminary experiments have revealed that the background noise was less than 0.8 mg from the crucible was recorded at each isothermal step. it was kept at a constant temperature of 30 K higher than the temperature of walls of the vacuum system.doi.9. Δg l Hm (T0). QCM: Measurements of Vaporization Enthalpy. Δf = −C·f 2 ·Δm ·SC−1 Article (1) where f is the fundamental frequency of the crystal (5 MHz in this case) with Δf ≪ f. Purity was additionally always controlled by 7Li NMR spectra. the sample of IL was cooled down.1021/jp311429r | J. The change of the vibrational frequency Δf is directly related to the mass deposition Δm on the crystal surface according to equation9. T0 appearing in eq 2 is an arbitrarily chosen reference temperature. the ILs were subjected to vacuum evaporation at 333 K for more than 24 h. the residual IL in the crucible was analyzed by ATR-IR spectroscopy. T0 was set equal to o o o 380 K.The Journal of Physical Chemistry B could be obtained with an accuracy better than ±2−3 kJ·mol−1 in comparison to the reliable literature data. The QCM method provides very reproducible temperature dependences of the frequency change rate df/dt. Sigma-Aldrich. The samples with odd numbers were prepared and purified at the University of Freiburg (the procedure is given in the SI). All ILs were fully characterized by IR. which is essentially unknown including all empirical parameters which are specific for the apparatus and the substance under study. and (b) to develop an approach for the temperature adjustments of IL vaporization enthalpies. a sample of IL placed in an open cavity of a thermostatted metal block is exposed to a vacuum system (at 10−5 Pa). Phys. it will be of interest to see if this causes a change in slope for the simulation results. As to objective (a). Prior to the experiments. About 50−70 mg of the IL sample was placed in a plain platinum crucible inside of the measuring head of the TGA. We used a carefully calibrated Perkin-Elmer Pyris 6 TGA in this work. it will be shown that most of the literature experimental data on vaporization enthalpies for the [Cnmim][NTf2] series are of a good quality and the scattered “Milky Way” behavior depicted in Figure 1 is simply the consequence of oversimplified adjustment of the results from the elevated experimental temperatures to the reference temperature 298 K.7 Atomistic molecular dynamics (MD) simulations were also carried out using force fields validated to capture the liquid phase properties of this class of ILs. and C is a constant. The highly pure sample of [C1mim][NTf2] was prepared and purified by Proionic GmbH. we will show that the chain length dependence of the vaporization enthalpy of this family of ILs is (within the boundaries of experimental and simulation uncertainties) represented by a straight line. Additional conditioning of an IL-sample was performed inside of the experimental set up in order to remove possible traces of solvents and moisture. The experimental uncertainties assessed for the vaporization enthalpy from the df/dt temperature dependences were always better than ±1 kJ·mol−1 (calculated as the twice standard deviation). The background noise can impact the QCM signal.18 Using the frequency change rate df/dt measured o by the QCM the molar enthalpy of vaporization.

B 2013. o Δg l Hm(Tav). 3. Δg l Hm(Tav). The structure of the ions was optimized and partial atomic charges were determined using the restrained electrostatic potential (RESP) method. Chem. but within the experimental uncertainties. enthalpies of vaporization Δg l Hm(Tav) derived from both methods (see column 3. For comparison as well as in order to keep consistency with our previous work. with the residual heat 6476 RESULTS AND DISCUSSION o Vaporization Enthalpies. Table 1) were not possible to compare because they referred to different Tav values. Coulombic interactions were treated using the particle−particle−particle mesh approach. Each calculation constrains a dihedral angle of interest to a specific value and lets all other degrees of freedom relax. For comparison with our own vaporization enthalpies measured by TGA and QCM. enthalpies of vaporization o Δg l Hm(Tav) (column 3. A force field for each cation and the [NTf2] anion was developed using procedures described in detail elsewhere. n. Force Field and MD Methodology. with thermostat and barostat time constants of 0. we have added to dx.0 kJ·mol−1 (calculated as the twice standard deviation). the linear correlation (omitting n = 1) is also apparent for the data from TGA.12 nm and tail corrections were used. Lennard-Jones parameters. For the QCM studies. and the mass loss rate dm/dt by TGA for the odd members of the homologous series [Cnmim][NTf2] with the alkyl chain lengths n = 1. Comparison of the vaporization enthalpies at the selected temperatures 380 K for QCM and at 572 K for TGA should be least affected by an ambiguity of the o Δg l Cpm value because of the deliberately short extrapolations from Tav of the individual measurements. while for the TGA results we chose 572 K for comparison (see Figure 3). 6473−6486 ■ .33−35 The enthalpy of vaporization was computed from the difference between the molar energies of the vapor and liquid phases o g l Δg l Hm(T ) = ⟨U ⟩(T ) − ⟨U ⟩(T ) + RT g l Article capacity obtained from the simulations via the following expression: res CP = 1 [⟨U NBĤ ⟩ − ⟨U NB⟩⟨Ĥ ⟩ 2 kBT + P(⟨VĤ ⟩ − ⟨V ⟩⟨Ĥ ⟩)] − R (5) where U is the intermolecular “non-bond” energy. although since differences Δg l Cpm are required and the ideal contribution is unaffected by the condensed phase.36 The heat capacity can be split into ideal and residual components. angle bending. as well as on the method used: long-chained species were studied at significantly higher temperatures in comparison to shorter one. which ideally is close to the average temperatures Tav of the individual measurements. The ideal contribution to the heat capacity can be obtained readily from electronic o structure calculations. the following expression was used: NB res res Δg l C P = C P (g) − C P (l) (6) to estimate the heat capacity difference by using MD and to compare it with the experimental value.27−30 The GAFF database does not do a particularly good job with alkyl chain dihedral angles.1 and 1. followed by single point energy calculations at the MP2/6-31g(d) level. Electronic structure calculations were carried out on an isolated cation and anion using the Gaussian 09 package25 at the B3LYP/6-31g(d) level. when measured by using the QCM.19 It is apparent from Table 1 that Tav values for ILs under study are significantly different depending on the chain length n. o As a matter of fact. Intermolecular pairwise Lennard-Jones and Coulombic terms were also used. (4) where ⟨U ⟩ and ⟨U ⟩ are the ensemble average internal energies of an ion pair in the gas and liquid phases. dihedral rotation and improper terms.31 The liquid phase consisted of 200 ion pairs simulated at constant pressure and temperature. and 7 are given in Table 1 (the primary experimental data are listed in Table S1 in the SI). The difference in heat capacity at constant pressure between the o liquid and gas phases Δg l Cpm was computed following a procedure given by Lagache et al.8.26 The ion charges were constrained to be ±1. and intramolecular terms were taken directly from the generalized Amber force field (GAFF).0 ps. It is apparent from Figure 3 that the vaporization enthalpies for the [Cnmim][NTf2] family (omitting n = 1) show an impeccable linear dependence on the number of C-atoms in the alkyl chain of the imidazolium cation. and Ĥ = U + PV is the configurational enthalpy.24 Lennard-Jones terms for cross interactions were derived using Lorentz−Berthelot combining rules. a standard Class I force field was used. we f irst used in the o current work the commonly acknowledged value37 Δg l Cpm = −1 −1 −100 J·K ·mol in eq 3. containing intramolecular bond stretching. Standard procedures were used to ensure the system was well equilibrated. as well as the TGA studies were performed at temperatures of 30−150 K higher in comparison to the QCM. Previous simulations and experiments have shown that the ion pair is the dominant vapor phase species at low pressure.1021/jp311429r | J. and then the resulting partial charges were uniformly scaled to achieve a net ion charge of ±0.22 MD simulations of the ILs [Cnmim][NTf2] with n = 1−12 were carried out using the LAMMPS package. Phys. In order to get the full picture. Combined experimental uncertainties of the determination of vaporization enthalpy with the TGA were assessed7 to be at the level of ±3. For the sake of comparison. of [Cnmim][NTf2]: Experiment and Simulation. respectively. MD simulations are then carried out with similar constraints. This reduction of net charge has been shown to be an effective way to mimic charge transfer without using a polarizable force field.The Journal of Physical Chemistry B The relationship between the mass loss dm/dt and the vaporization enthalpy was derived according to eq 1 but by using the mass loss rate dm/dt measured by the TGA (instead of the frequency change d f/d t by QCM). Table 1) have to be adjusted to any arbitrary but reasonable common temperature.23 Atom types. The algorithm of Shinoda et al. and the dihedral angle parameters were fit to match the ab initio energies. 5. and is expected to give much more accurate enthalpies of vaporization. respectively. Vaporization enthalpies. The Lennard-Jones potential was truncated at 0. the suitable temperature was 380 K. referred to average temperatures of the experimental temperature ranges obtained from temperaturedependent measurements of the frequency change df/dt by the QCM.32 as implemented in LAMMPS was used. The liquid phase was simulated for 4 ns and an independent 40 ns simulation of a single ion pair in the vapor phase was also carried out. we also collected in Table 1 our previous results for the even members of this family measured just recently by QCM18 as well as by TGA. so the dihedral angle potential of the alkyl chains was parametrized by matching energy scans in which B3LYP/6-31g(d) geometry optimizations were performed. The scatter of the results obtained by the TGA is somewhat larger.22 Briefly.doi.

0 104.2 ± 1. Δg l Hm(Tav) for 1-alkyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imides on the length of the cation alkyl chain (n).4 Enthalpies of vaporization for ILs were adjusted to 298 K with Δg l Cp = −100 J·K−1·mol−1. and in this −1 −1 o case we again used the value Δg for l Cpm = −100 J·K ·mol interpolation of vaporization enthalpies from Tav to the selected T = 475 K.6 ± 1.0 110.2 ± 1.0 136.0 ± 1.5 ± 1.5 ± 1. In our opinion.4 533.0 125.0 152.0 160.0 [C10mim][NTf2] 142. measured by the combined Knudsen-QCM method.0 147.0 118.0 166.0 149. of [Cnmim][NTf2] at Tav and at 298 K Article T-range/K 345−400 543−623 362−395 480−570 362−395 480−570 362−395 513−572 362−404 523−603 362−405 513−571 364−406 503−603 372−402 513−571 380−410 493−552 392−425 503−562 400−432 513−571 405−445 513−571 410−452 546−621 a Tav/K 381. where the authors consider their Knudsen-QCM results as an unusual evidence for “structural percolation phenomenon. there is also no break in slope.1 ± 1.1021/jp311429r | J.0 114.8 394.0 137. there is no change in slope.4 ± 1.doi. As can be observed.0 ± 1.5 ± 1.2 ± 1. but then also confirm both observations with simulations (see below).8 ± 1.10 (right panel) where there is a change in slope claimed at about n = 6.0 ± 1.7 ± 1.4 ± 1. Experimental Vaporization Enthalpies.6 ± 1.0 113.5 129. if one chooses to take the n = 3 point as an outlier (for example).8 381. o The enthalpies of vaporization Δg l Hm(Tav) computed in the present work using MD simulations are shown in Figure 4 along with the experimental results from Rocha et al.8 ± 1.4 128.0 ± 1.8 [C7mim][NTf2] 133. [C1mim][NTf2] 119.10 The solid lines are linear fits to the n = 2−12 data.5 ± 1.0 126.10 6477 one must throw out the n = 2 data point.0 378 520. This is clearly seen in Figure 3 of the present work. uncertainties in the heat capacity differences were not taken into account.7 387 541.3 ± 1.10 Rather. we do not observe any discontinuities in the linear trend on enthalpy of vaporization with chain length (except for n = 1) within the error bars for all three data sets.0 139.’s data and consider the associated uncertainties.5 o Δg l Hm(Tav) kJ·mol−1 a o Δg l Hm (298 K) kJ·mol−1 method QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA QCM TGA o Figure 3.3 ± 1.5 ± 1.8 ± 1.8 541.5 ± 1.0 110.13 are in reasonable agreement with the present set of results or most of the results found in the literature.7 560.0 108.6 ± 1.8 383 541.3 546.3 553. Δg l Hm. Experimental uncertainties of our measurements as well as of the literature data are presented in Figure 3 as the error bars.9 ± 1.6 [C2mim][NTf2] 118.3 ± 1.8 142.0 ± 1.0 122. In addition to pointing this out.0 142.’s data.9 ± 1. which is thought to occur in ILs.0 169.5 ± 1. the MD results show a monotonic increase in the enthalpy of vaporization with no discontinuity.8 ± 1.6 126.0 147.6 ± 1. the experimental vaporization enthalpies from the combined Knudsen-QCM method10 also fit very well (within the error bars) the linear dependence on the chain length n.5 416.2 541.3 582. In our opinion.6 132. it seems that this is an artifact that was overinterpreted by the authors within the boundaries of their experimental uncertainties.0 133. despite the fact that Rocha et al.10 does not result in two distinct trends in the enthalpy of vaporization with chain length.8 ± 1.8 430. which shows why one should see an outlier at n = 1 (due to Coulombic and van der Waals trends).5 [C5mim][NTf2] 127. Chem. in agreement with the experimental results of the present QCM and TGA study.7 ± 1. It could be argued that there is a slight change in slope. Figure 3 recent literature data from Rocha et al. B 2013.8 385.5 136. This conclusion disagrees with the interpretation given in ref 10.3 ± 1.4 175.6 ± 1.0 132.4 ± 1. state in their paper that the enthalpy of vaporization “results for [C2mim][NTf2] and [C4mim][NTf2] from Santos et al.0 157.The Journal of Physical Chemistry B o Table 1.2 380. We also tried to fit two different slopes through the simulation results for n = 2−6 and n = 7−12.1 164. with increment dx. in order for one to observe a change in slope of the enthalpy of vaporization in Rocha et al.” It is also clear from Figure 3 that. Or. and these fits are shown as dashed lines in the upper panel of Figure 4.0 ± 1. 117.0 [C16mim][NTf2] 156.5 ± 1.2 378 541.4 [C18mim][NTf2] 162.10 We selected for comparison with their data the temperature 475 K. the “structural percolation phenomenon” or nanoscale aggregation of alkyl chains.6 [C4mim][NTf2] 124.9 145.5 [C3mim][NTf2] 121.6 522 408.6 ± 2. Simulation results described below give further support for this conclusion.0 132. but in contrast to the experimental data of Rocha et al.2 ± 1.9 [C8mim][NTf2] 136.0 133.8 424. we provide our own experimental data for even and odd [Cnmim][NTf2] with n = 1 to n = 18 (see Table 1) and show that from n = 2 to n = 18 there is no change in slope. We show that if one uses all of Rocha et al.0 138. We do see an “outlier” at n = 1.3 ± 1.4 ± 1. Dependence of the enthalpy of vaporization.4 ± 1.” Thus we are unsure what the justification is to exclude the n = 2 data from the fitting.5 ± 1.8 ± 2. Phys.6 590.9 ± 1. as it has been claimed recently.9 140.7 ± 1.8 ± 1.5 ± 1.0 [C12mim][NTf2] 147.0 136. From the data presented in Figure 3. similar to our own finding.1 ± 1. 6473−6486 .1 ± 1.1 [C14mim][NTf2] 152.9 [C6mim][NTf2] 131.0

we also find that the increase in enthalpy of vaporization for [Cnmim][NTf2] with increasing chain length arises completely from van der Waals interactions. 6478 and is consistent with the physical argument presented above. of vaporization enthalpy per CH2 actually increasing slightly for the longer alkyl chains. Consistent with the findings of Köddermann et al. and thus within the statistical uncertainties of the simulations. and shows a slightly larger drop per CH2 in going from [C1mim][NTf2] to [C2mim][NTf2]. this change is miniscule compared to the uncertainties in the calculations. This latter feature is similar to the calculations by Köddermann et al.10 did to see whether our MD simulations exhibited a similar aggregation behavior.39 where the Coulomb energy drops by 4. The differences for all three contributions or the driving forces for the vaporization enthalpy obtained from the MD simulations are shown in Figure 5 for the ILs [Cnmim][NTf2]. The results are shown in Figure 6. To quantify this behavior.10 Color-coded simulation snapshots were used to provide qualitative evidence that the alkyl tails of the cations tend to aggregate as n > 6. we carried out the same analysis that Rocha et al. They observed a growth in the height of the primary CT−CT peak for n = 2−6. Chem. Similar to what Rocha et al. The Coulomb energy monotonically decreases from [C2mim][NTf2] to [C12mim][NTf2] by 4 kJ·mol−1 for 380 K (7 kJ·mol−1 at 572 K) for all ILs. and intramolecular terms.1 nm in going from n = 4 to n = 8.doi. identified as the CT−CT RDF. Effect of alkyl chain length on the total enthalpy of vaporization and the various components at 380 K and 572 K. In contrast to the present study.39 the internal energies for the liquid and the vapor phases were divided into the contributions from Coulomb.85 nm as the chain length increases above n = 6. B 2013.. our MD simulations show clear evidence of nanoscale ordering of the alkyl tails.39 suggest that the Coulombic energy contribution remains unchanged as the length of the alkyl chains increase beyond n = 2. However.10 (right) where a discontinuity in the trend is observed at around n = 6. Figure 5. Following previous procedures. Unfortunately. which are the dominant driving force of the vaporization process. diamonds (intramolecular contribution). and then a plateau of the peak height for n > 6. It can be concluded that the MD simulations show that the enthalpy of vaporization exhibits a simple linear dependence with chain length for n = 2−12. left) compared with the experimental data of Rocha et al. Rocha et al. which was then used to explain the change in slope of the experimental vaporization enthalpy curve. The exact values of the peak heights and positions obtained in the present work differ from those of Rocha et al. They also observed that the distance of the second peak shifts from about 0. To examine this further. The percolation phenomena was first observed by MD simulations15. 117. Phys. the authors did not report the enthalpies of vaporization from the MD simulations where this aggregation behavior was observed. The lower left panel of Figure 6 shows that the primary peak height plateaus at around n = 6− kJ·mol−1 (at 298 K) from [C1mim][NTf2] to [C2mim][NTf2]. squares (Coulombic contribution). however. the results of Köddermann et al. van der Waals.10 dx.10 attributed the change in slope of the vaporization enthalpy with chain length to a structural percolation phenomenon associated with the aggregation of cation alkyl tails for n > 6.38. Vaporization enthalpy of [Cnmim][NTf2] at 380 and 572 K from MD simulations (this work. circles (van der Waals contribution)..16 and further quantified by Rocha et al.39 however.The Journal of Physical Chemistry B Article Figure 4. for n = 2−10. These features provide clear evidence for nanoscale ordering in the MD simulations. Rocha et al.95 to 0. 6473−6486 . The symbols have the following meaning: triangles (total enthalpy). The lower right panel of Figure 6 shows that the location of the second peak changes significantly for n < 6 and decreases by about 0.1021/jp311429r | J.10 computed the height of the first and second peaks on the terminal alkyl carbon−carbon radial distribution function (RDF).10 observed.

the MD results are in better agreement with the experimental TGA data.’s force field results in a more structured liquid than that obtained in our simulations.8 −1. What is clear is that the MD simulations and two different sets of experiments have essentially the same slope.6 −12.4 112.8 where used in the present study.6 119.0 138. clusters could be formed.0 reveal reliable structure−property relationships for the o [Cnmim][NTf2] family. it is very common to adjust the vaporization enthalpies also to the reference temperature o 298 K. (Top): Terminal alkyl carbon RDF (CT−CT) for [Cnmim][NTf2]. of [Cnmim][NTf2] o Δg l Hm (QCM) o Δg l Hm (MD) Δ (QCMMD) o Δg l Hm (TGA) o Δg l Hm (MD) Δ (TGAMD) n in [Cnmim] [NTf2] 1 2 3 4 5 6 7 8 9 10 11 12 119. the experimental values Δg l Hm(Tav) (column 3.2 −6.5 151.5 −5.8 −11.1 132. (Lower right): Height of the second peak of the CT−CT RDF as a function of position of the second peak.8 −10. However. which were reasonably close to the Tav of the individual experiments.3 132.5 −1.1 114. and.0 −12.The Journal of Physical Chemistry B Article Figure 6. Both sets of simulations show clear evidence of nanoscale ordering.40 we have observed that at high temperature and pressure. the present MD simulations do not show deviations f rom the monotonic change in vaporization enthalpy with chain length.5 572 K 111.8 111. B 2013. n = 1−12. that we do not expect to observe any significant amount of clustering in the vapor 131.8 118.4 −1.6 107.0 121. the presence of nanoscale ordering does not necessarily result in the kind of slope change in the vaporization enthalpy curve reported by Rocha et al.7 −5. there is a consistent offset with the MD simulations predicting vaporization enthalpies about 10% higher than the experimental QCM data. in fact.4 124.1 137.7 106. At 380 K. while scaled charges of ±0.10 o A comparison of the absolute Δg l Hm(T) values from the MD simulations and experiments is given in Table 2 and Figure 7.6 −13.2 165.8 162. which is likely because they used integer charges on their cations and anions. o (T) Values to the How Does One Adjust Δ lg Hm Reference Temperature 298 K Properly? In order to 6479 Table 2.5 121.2 113.4 115. Experimental and MD Calculated Enthalpies of o Vaporization. the similarity between the two simulations is striking.1 149.6 127.0 −10.7 116.5 111.9 134. Such a procedure has allowed a proper interpretation of the chain-length trends at the selected temperatures. (lower left): Height of the first peak of the CT−CT RDF as a function of chain length.4 −14.doi.5 109. This suggests that. Nevertheless. In particular.3 −1. Phys.0 157.8 153. because a different force field was used.8 380 K 130.0 130.7 −11. however. in recent work.6 136. Table 1) were adjusted using eq 3 to 380 K (for QCM) and 572 K (for TGA).9 142. At 572 K.1021/jp311429r | J. However.8 −4.4 −13.4 −10. 117.1 109. That is why the values of vaporization enthalpies were hardly affected by uncertainty in o the Δg l Cpm value used for extrapolation.4 146. and no discontinuities are observed. The present experiments are at sufficiently low pressure. because the Δg l Hm(298 K) data are required for validation of the high-level first principle calculations41 as well as for development of the reliable force fields for MD dx.4 119. The similarity of slopes for short and long chains supports the assumption made in the MD simulations that the vapor phase consists of single ion pairs.4 144. Rocha et al. at least to the extent that the MD simulations are capturing the energetics of these ILs.0 105.2 129.2 −11. Δg l Hm. although the relatively large uncertainties of the TGA experiment make it harder to be certain of this.5 −8. van der Waals interactions go down with increasing chain length.2 119. Chem.9 118.3 128.4 124. 6473−6486 . but despite this.

Once again. this Δg l Cpm estimate contradicts recent experimental results. where we used the experimental Co pm(1). The common use of Δg l Cpm = −100 J·K−1·mol−1 as a constant for all ILs regardless on the structure o is obviously an oversimplification since Δg l Cpm will differ for ILs with different cations and anions. it should be mentioned that the estimation procedure by Chickos and Acree43 was parametrized only for molecular liquids and not for o ILs. because the assessed resulting Δg l Cpm values are independent of the experimental method.1021/jp311429r | J. Hence. Being simultaneously surprised and confused with the o general level of the heat capacity difference Δg l Cpm for ILs. Small o differences in ΔlgCpm will result in large differences in g o o Δl Hm(298 K). which are then used to estimate Co pm(g) using well-known statistical thermodynamics procedures. the g o Δl Cpm values derived according to eq 7 using all available data collected for [Cnmim][NTf2] in Table S1 (SI) are indistinguishable within the boundaries of their uncertainties (see Table S1. B 2013. it became apparent that at least one additional approach is required to assess this difference independently from the dx. In this work we provide a simple and elegant procedure for o properly assessing the correct Δg l Cpm values to use for the [Cnmim][NTf2] series. we estimated the value Δg l Cpm(298 K) = 44 −137. Table 1) using QCM and TGA at two significantly different average temperatures Tav. They are also quite different from the “acknowledged” constant o −1 −1 value Δg used by most of the l Cpm = −100 J·K ·mol 37 community. o accurate Δg l Hm(298 K) data will not be available until reliable 19 g o values of Δl Cpm or Co pm(g) for ILs have been obtained. Δg l Cpm is the difference of molar heat capacities of o o o the liquid and the gaseous samples: Δg l Cpm = Cpm(g) − Cpm(1).The Journal of Physical Chemistry B Article Figure 7. This procedure is based on the experimental measurements of vaporization enthalpies o Δg l Hm(Tav) with two different methods at two different Tav. in the recent o −1 −1 literature. which showed that enthalpy of vaporization data for [C2mim][NTf2] obtained using different experimental methods agree only for the heat capacity difference not higher than −50 J·K−1·mol−1 (see details in ref 9).g. however. o and Cpm(g) was calculated in this work by using density o functional theory (DFT) calculations. For the calculation according to eq 7. It is quite o surprising that the level of absolute Δg l Cpm values for the initial representatives of the [Cnmim][NTf2] family were significantly o −1 −1 lower than the “expected” value of Δg l Cpm = −100 J·K ·mol .41 A similar value of Δg l Cpm= −105. the o value of Δg l Cpm used in the extrapolation is crucial.40 Extrapolation to T = 298 K is usually done by using eq 3. as well as it should also vary as a function of alkyl chain length n. The latter fact serves as good o evidence that the simple Δg l Cpm procedure designed in the o current study is correct. column 4 and 5). there are no other data or even ideas of how to o obtain Δg l Cpm values for ILs. but it is essential to realize that over large differences (over 100 K) between Tav and the reference temperature. it has turned out that use of this procedure in the case of the [Cnmim][NTf2] family seems to be incorrect. the values of Cpm(1) can be reliably measured by calorimetry or even be assessed by simple additivity rules. ambiguity of the Δg l Cpm values required for the extrapolation of experimental vaporization enthalpies to the reference temperature 298 K has been discussed recently. the value Δg (regardless of l Cpm = −100 J·K ·mol the structure of the IL) has been systematically used for the o temperature adjustments according to eq 3. we need only to rewrite eq 3 as follows: g o g o o Δg av )QCM − Δl Hm(T av )TGA ) l Cpm = (Δl Hm(T /[(Tav )QCM − (Tav )TGA ] (7) in order to obtain the experimental differences between heat o capacities Δg l Cpm indirectly. However. o As a rule. Therefore the heat capacity Co pm(g) is usually derived using the frequencies calculated by quantum chemistry. 117.9. we were very surprised o that the absolute values of Δg l Cpm as well as the general trends with increasing chain length are very close for ILs and the polar molecular compounds like linear alkanols (see Table S3 column 4).5−19 This fact also has thwarted further development o of these methods. o for [C2mim][NTf2]. we have also calculated Δl Cpm for the similarly shaped linear alkanols CnH2n+1OH with n = 1−14 (see Table S3 column 4). the experimental determination of the heat capacity of a gaseous IL is not possible. for ILs.doi. we deliberately measured for each IL o under study the enthalpies of vaporization Δg l Hm(Tav) (column 3. These values for the [Cnmim][NTf2] series calculated with eq 7 are given in Table 3.4 J·K ·mol was also derived. However. Thus. The experimental and calculated enthalpies of vaporization for [Cnmim][NTf2] at 380 and 572 K. g o For the sake of comparison. 6480 How to resolve this contradiction? As a matter of fact.44 In contrast. 6473−6486 . the Δg l Cpm(298 K) data are generously overestimated.43 However.45 However. according to a procedure developed by Chickos and Acree42 using the experimental isobaric molar heat capacity of [C4mim][N(CN)2]. o It is apparent from Table 3 (column 3) that Δg l Cpm values estimated from eq 7 are definitely chain length dependent. simulations. by using the g o o value of Δl Cpm = −100 J·K−1·mol−1. Chem.7 J·K−1·mol−1. and this fact has heavily aggravated comparison of vaporization enthalpies measured by different methods. e. As a consequence. In the current work. A significant amount of the discrepancies among the literature vaporization enthalpies for the [Cnmim][NTf2] series at 298 K as measured by different methods is most likely due to the long temperature range (80− o 270 K) over which extrapolation of the measured Δg l Hm at Tav to the reference temperature is performed. in the literature.19 As a matter of we have o deliberately used only our own Δg l Hm(Tav) data. the current ΔlgCpm value broadly applied for temperature adjustments for ILs is a rough estimate used regardless of the structure of an IL. This Δg l Cpm value is based on the calorimetric liquid heat capacity measurements for a single IL [C4mim][PF6] and statistical thermodynamic o −1 −1 calculations. Phys. o Formally.7. As a consequence.

1 b o Δg l Cpm (Tav) o Δg l Cpm (298 K)c Co pm (g) (298 K)d 5 348. as well as the similarity for mixtures of 6481 where ρ is the density of an IL (kg·m ) and M is the molar mass (kg·mol−1).9 506.6 885.9 942.7 −147.999) and the o (l.p).2 −155.0 552.doi. From the oscillation theory. Calculated in ref 45 using DFT at the B3LYP/6-311+G(d. Importantly. Estimated from (Δg l m av av d o Co pm − Cvm difference (see text). 6R for the sum of rotational and translational contributions.2 817. SI). 8. 298 K) were estimated by interpolation and missing Cpm extrapolation. o Assuming that for the ideal gas the relation (Co pm − Cvm)g = R is valid. g) o o o + Cvm (conf. B 2013.8 −201. l) + (Cpm − Cvm )l o o o o Cpm (g) = Cvm (transl. The contribution of vibrations at low frequencies into the heat capacity is equal to R for each degree of freedom: all together.2 470. both procedures suggest that o Δg l Cpm is dependent on the chain length of the cation.9 631.8 −169. Assuming the equality of the vibrational contributions into the heat capacity of the liquid and the gaseous phase. Values of Co pm(g. a sum of contributions for the free rotation and the free translational motion of a molecule into the ideal gas heat capacity can be assigned to be equal to 3 R.1 −181. Heat Capacity Differences between Liquid and Gas Phases for [Cnmim][NTf2] and Heat Capacities for Liquid [Cnmim][NTf2] in J·K−1·mol−1 [Cnmim] [NTf2] 1 [C1mim] [NTf2] [C2mim] [NTf2] [C3mim] [NTf2] [C4mim] [NTf2] [C5mim] [NTf2] [C6mim] [NTf2] [C7mim] [NTf2] [C8mim] [NTf2] [C10mim] [NTf2] [C12mim] [NTf2] [C14mim] [NTf2] [C16mim] [NTf2] [C18mim] [NTf2] a Article conformers at equilibrium in these phases.6 389. They were dx. 6.1 o Δg l Cpm (298 K)e 3 −73 ± 9 −56 ± 13 −76 ± 17 −67 ± 11 −96 ± 12 −86 ± 13 −114 ± 13 −92 ± 9 −136 ± 11 −167 ± 12 −161 ± 11 −170 ± 15 −157 ± 11 4 6 −125. the heat capacity difference was expressed as follows: o o o o o Δg l Cpm = Cvm(transl.8 (with r2 = 0. it is useful to o analyze separately the heat capacities Co pm(g) and Cpm(l) for the [Cnmim][NTf2] series. l) − Cvm (rot. l) o o o + Cvm (conf.P) measurements as follows: κT = 2 ⎞ TαP M 1⎛ 1 ⎜ ⎟ + ⎜ o 2 ρ⎝W Cpm ⎟ ⎠ −3 (13) experimental procedure designed above.9 (10) −85 −73 −88 −89 −95 −113 −124 368.6 429. Phys. we decided to use the basics of statistical thermodynamics where the heat capacities of liquid and gaseous phases can be assessed as a sum of translational. Pa−1. 6473−6486 . rotational. 298 K) have been critically evaluated just recently.6 755.6 491. column 4.6 532. g) + (Cpm − Cvm )g o Cvm (conf) (8) (9) where the contribution is responsible for the equilibrium mixture of conformers.4 598. We collected the volumetric properties and the speed of sound data for the [Cnmim][NTf2] series from the ILthermo database48 (see Table S5. e Difference (column 5 − column 2). Chem.7 n + 440. K . 298 K) = 20. and vibrational contributions. 4.5 (with r = 0.8 593. The resulting heat o capacity differences ΔlgCpm derived from the volumetric properties are given in Table 3.3 536. g) + Cvm (rot.8 −137.4 −243.The Journal of Physical Chemistry B Table 3.4 661.6 −333. In order to gain more insight in this problem.8 −411. For this purpose.5 −369. The experimental values for Co pm(l.2 511. Thus two independent procedures have ascertained the o level of the Δg l Cpm differences for the ILs under study.1021/jp311429r | J. Comparison of the o Δg l Cpm values derived from the QCM and TGA experiments (Table 3 column 3) with those from the volumetric properties (Table 3 column 4) have shown that these results are hardly distinguishable for the homologous series of [Cnmim][NTf2] ILs.4 565. It turns out that the evaluated experimental data44 (for n = 2.44 We report the recommended values in Table 3. l) + Cvm (rot.6 −285. Compressibilities are more often derived from the speed of sound W(T. column 2.4 n + 327. g) + Cvm(rot. 298 K) for n = 10−18 were estimated by interpolation and extrapolation. κT is the isothermal compressibility. The molar volume Vm as well as the thermal expansion coefficient αP are usually determined from the temperature dependence of the density of liquid. where αP is the thermal expansion coefficient.6 1005. g) + Cvm (vib. 117. bCalculated using eq 7 from QCM and TGA data measured in this work (uncertainty is estimated according to to the o (TGA))2 + f o l l o w i n g e q u a t i o n Δ Δ lg C pom = [ ( Δ lg H m c o 2 1/2 H (QCM)) ] /[( T (TGA) − ( T (QCM))].46 the free rotation or linear motion of a molecule in the condensed state is converted into the low frequency vibrations. and 14) fit very well to the linear chain length dependence.9999) and the missing Co pm(g. It has (12) The evaluated experimental data44 were approximated with the linear 2 regression: Co pm(l.2 − level. and these values were approximated in this work with the linear regression: Co pm(g.4 692. 298 K) were estimated in ref 45 with B3LYP/6-311+G(d. The isobaric and the isochoric heat capacities of both phases are usually expressed with the following equations: o o o o Cpm (l) = Cvm (transl. l) + Cvm (vib. l) − (Cpm − Cvm )l Co pm(1) (298 K)a 2 473.2 409.2 From common rules of statistical thermodynamics. The compressibilities κT can be calculated from the pressure dependence of density in the isothermal conditions.4 572. 298 K) = 31. g) + (Cpm − Cvm)g o o o o − Cvm (transl. and both procedures could be recommended for the practical estimations required for temperature adjustment of experimental vaporization enthalpies. we can simplify eq 10 to o o o Δg l Cpm = − 2R − (Cpm − Cvm)l (11) It is obvious from the main part of turned out that this contribution can be easily estimated from the volumetric properties according to following equation:47 o o (Cpm )l = − Cvm 2 αP VmT κT −1 o eq 11 that the contribution (Co pm − Cvm)l is g o the heat capacity difference Δl Cpm.8 450.

Therefore we have calculated these values for the [Cnmim][NTf2] family using DFT at the B3LYP/6-311+G(d. o −1 −1 can be obtained. column 5). The differences between the gaseous heat capacities (from DFT) and the liquid heat capacities (from experiment) are given in Table 3 (column 6). the MD absolute values for Δg l Cpm are about twice as large as the experimental values. 298 K) data. column 6).7 ± 37. use of eq 14 for temperature adjustment of vapor pressure measurements opens a new way for comparison and validation of experimental results measured by different dx. 298 K) for the [Cnmim][NTf2] are quite reliable. and they 6482 allow one to make recommendations on how to adjust vaporization enthalpies to the reference temperature properly.0) with r 2 = 0. However. and 18) and the literature Co pm( eq 14 seems to be of a crucial o importance because it helps to avoid ambiguity of the Δg l Cpm values commonly used nowadays in the literature. The results are given in Figure 8 and in Table S4.g. the Co pm(l. It turns out that a simple linear correlation: g o o Δ1 Cpm = Cpm (l. This simple empirical correlation was shown to be reliable within 3% (or ±16 J·K−1·mol−1).1021/jp311429r | J. o Clearly.39 also reported the overall linear decrease of Δg l Hm(Tav) for −1 −1 [C2mim][NTf2] from 132 kJ·mol to 111 kJ·mol over the temperature range of 273−473 K. which are commonly found in modern IL laboratories. For the DFT calculations. We are reticent now to give any reasonable explanation for the significant difference between experimental and theoretical o Δg l Cpm values. From their MD simulations. column 3). it seems to be reasonable to develop an empirical correlation between the g o experimental Co pm(l. Taking into account that the experimental data on Co pm(l. available experimental methods are unable to measure Co pm(g. 117. as all the procedures show that there is a strong dependence on chain length. pyrrolidinium. but we would recommend applying this simple correlation for other ILs instead of using −1 −1 o the conventional constant Δg l Cpm = −100 J·K ·mol . the disagreement could be attributed to the simple rigid rotor-harmonic oscillator approximation used in the first principles calculations of the Co pm(g. which is a value of Δg l Cpm = −108 J·K ·mol in agreement with our MD values (see Figure 8). to o our surprise. it should be mentioned that there is no rigid rotor approximation in MD. but. pyridinium.49 The molar volume Vm values are usually obtained from the densities of ILs. which are given in Table 3 (column 2). 298 K) for the [Cnmim][NTf2] were estimated by interpolation and extrapolation (see Table 3. footnote a). column 2).950 (14) can be derived using our own QCM and TGA data (except for n = 1. o In spite of this fact. These differences follow the already expected increasing trend with the increasing chain length.p) level (see Table 3. and the missing Co pm(l. 298 K)( −0. the size of the ILs under study in the current work is too large to make calculations without this approximation. Residual heat capacities of the liquid and gas phase obtained from MD simulations (left) and computed Δg l Cpm (right) as a function of cation alkyl chain length. 298 K) for ILs.05) + (68. A search of the literature has revealed that Ködermann et o al.The Journal of Physical Chemistry B Article o Figure 8. taking into account at least the chain-length dependence. 298 K) values required for eq 14 could be also predicted with a reasonable accuracy49 as a function Co pm(l. 298 K).doi. which are routinely determined as a part of a physical-chemical attestation of new ILs.. but we expect that the fluctuations of these coefficients should be not too large. Moreover. or ammonium based ILs). column 3). Thus. approximated with a linear regression (see Table 3. but disagree with the experimental values o (Table 3. It is interesting that the results of the MD simulations are in a good agreement with those obtained from the DFT calculations (Table 3. 298 K) = f(Vm) of the molar volume Vm. 298 K) and the experimental Δl Cpm. the use of a constant value for ΔlgCpm is not recommended. Unfortunately. 12. At the current state of our knowledge. We are conscious that the coefficients in eq 14 could vary depending on the particular class of IL under study (e.26 ± 0. the general trends regarding Δg l Cpm from the experiment and MD simulations are very similar. B 2013. 298 K) of ILs are easily measured using the commercially available DSC (differential scanning calorimeters). The correlation according to eq 14 is adjusted now for the [Cnmim][NTf2] series. Equation 14 is easy to apply because the Co pm(l. but the absolute values are about twice as large as those obtained from our experimental findings from TGA and QCM methods (Table 3. The Δg l Cpm values from experiment and MD are deliberately referred to the similar temperatures. In order to better understand this discrepancy. Thus the currently observed disagreement of the absolute experimental and o theoretical Δg l Cpm values merits further extended investigation. Chem. we also calculated o values for the ILs [Cnmim][NTf2] using MD the ΔlgCpm simulation. As has been already pointed out. which will be very useful for o quick appraisal of Δg l Cpm for the [Cnmim][NTf2] series with any chain length. 6473−6486 . Phys.

the data were shifted by 60 kJ·mol−1). We calculated and collected the appropriate numbers in Table 4 for comparison. (□) n-alkyl benzenes (for comparison with ILs.997) (MD) (16) Figure 9. some scatter of the Δg l Hm(298 K) data still remains.995) o −1 Δg = 124. However. (Δ) n-alcohols (for comparison with ILs.. A most simple manifestation of additive rules is the correlation of any property. column 3. 6473−6486 .65n l Hm(298K)/kJ · mol (15) from MD (with r 2 = 0. with the number of C-atoms. In this context. the dependence of vaporization enthalpy on the number of C-atoms. provided that a general transfer of the group contributions established for the molecular compounds to the ILs is valid. Using the QCM and TGA results obtained in our work. but in our opinion this scatter is rather the evidence of the challenging task to measure the vapor pressure and the vaporization enthalpy of extremely heavy. In contrast to the Article prediction of thermodynamic properties of molecular liquids are well established. a larger range of about 30−50 K is necessary to provide a reliable o slope and ΔlgHm (Tav). alkyl benzenes. and we have even scaled them in order to fit them on the same plot. (◊) [Cnmim][NTf2] (joined treatment of the QCM and TGA results from this work).51 In our recent studies we have tried to establish whether the thermodynamic properties such as enthalpy of vaporization and enthalpies of formation of ILs obey the group additivity rules. a careful analysis of the primary experimental data of this work has revealed that both ILs were measured in a very narrow (about 15 K. low volatility 5 o ILs. Phys.7 It is very disheartening that the data for C10 and C12 measured10 with the very good QCM-Knudsen method are significantly out of the linear correlation apparent on Figure 9.50. The resulting plot is given in Figure 9. see Table S1) temperature range. apparent from this figure that the intercepts for the different families are totally different.50 Is there any Difference Between ILs and Molecular Liquids? CH2−Increment.1021/jp311429r | J. and alcohols at 298 K. dx. or sublimation enthalpy but not for vaporization enthalpy. (○) nalkyl nitriles (for comparison with ILs. the data were shifted by 45 kJ·mol−1). due to a systematic error discussed elsewhere. o Δg l Hm(298 K) adjusted with o Δg l Cpm from “Milky Way” discussed at the beginning of this paper and presented in Figure 1. The outlying values of Δg l Hm(298 K) obtained from TGA 8 and a high-temperature spectroscopic technique could be an indicator that these methods still require further development. It is Δg l Hm(298 Figure 10. in the alkyl chain of the imidazolium cation follows the equation o −1 Δg = 115. n.7 + 3.51 This knowledge could simplify reliable predictions for ILs. fusion. and it is obvious that Δg l Hm(298 K) of ILs follows the same pattern as the molecular liquids. o For comparison. and they generally represent the contribution of the CH2-group to the vaporo ization enthalpy Δg l Hm(298 K). This correlation in the series of homologues is additionally a valuable test to check the internal consistency of the experimental results. where the odd and even effect was observed for such thermodynamic properties as melting point. Certainly. all the slopes presented in Figure 10 seem to be similar. we have adjusted experimental results for the [Cnmim][NTf2] series measured by all available methods from Tav to the reference temperature 298 K. Group-additivity procedures for 6483 from which the enthalpy of vaporization K) of other representatives in this series with different n can be calculated. the proper adjustment of vaporization enthalpies to 298 K using simple empirical rules developed in the current study has converted the mess of experimental points available in the literature (see Figure 1) into the logical structure−property dependence presented in Figure 9. it is now quite obvious that there is a simple linear dependence of the vaporization enthalpy on the o chain length. the data were shifted by 60 kJ·mol−1. 117.doi.89n l Hm(298K)/kJ · mol from experiment (with r 2 = 0. Thus. the enthalpy of vaporization. The enthalpies of vaporization for 1-alkyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imides. The experimental Table 3. In order to demonstrate the advantage of using eq 14 and o ΔlgCpm values derived in this work. From our experiences. it is also important to point out that enthalpies of vaporization for the [Cnmim][NTf2] family with odd and even chain length of the alkyl-imidazolium cation fit o the same straight line.g. alkyl nitriles. we present the Δg l Hm(298 K) chain length dependence for some molecular homologous series together with the results for the ILs [Cnmim][NTf2] in Figure 10. We also deliberately omitted from Figure 9 the calorimetric data measured by drop microcalorimetry13 for [Cnmim][NTf2].The Journal of Physical Chemistry B techniques and could significantly contribute to the development and refinement of experimental methods dealing with ILs. However.8 + 3. B 2013. Chem.

44 ± 0.7 kJ mol−1 for the MD simulations).5 kJ·mol−1). in contrast with the available literature data. This finding is also in agreement with the MD-simulation studies of the [Cnmim][NTf2] series by Köddermann et al.39 and also in this work. Experimental Vapor Pressures of 1-Alkyl-3-methylimidazolium Bis(trifluoromethylsulfonyl)imides and a Correlation Scheme for Estimation of Vaporization Enthalpies of Ionic Liquids. University of Rostock. The results of MD simulations generally agree with the experiments and. ■ According to the data given in Table 4. the Coulomb energy for [C1mim][NTf2] is slightly higher than that for other ILs in the series (by about 4.5 kJ·mol−1 is also observed for n-alkylbenzenes and for n-alkyl-nitriles. it is also interesting to compare how different MD procedures are able to obtain this value. U. but the additive contributions become unique in comparison to the well-established group-contributions developed for the molecular liquids. pyridinium. J. (2) Zaitsau. 7303−7306.. Present Address Faculty of Interdisciplinary Research. B 2013. A. A. Finally. J. The Kinetic Theory of Gases.65 4. Chem.39 using a different force field. Y. Fax: +49 381 498 6524. S. Kabo. o Having established the CH2 contribution to Δg l Hm(298 K) in ILs from experiment.12 4.7 kJ·mol−1 per CH2 group in the [Cnmim][NTf2]. A.71 ± 0.7 kJ·mol−1. Tschersich. 6473−6486 . which exactly coincides with the increase in the enthalpy of vaporization for n-alkanols but is in disagreement with our MD results.20 3. However. and this compound could be already considered as the “regular” representative of the family with the constant CH2-contribution.The Journal of Physical Chemistry B Table 4. a somewhat lower contribution of 4. ammonium. REFERENCES (1) Knudsen. developed in their group having integer partial charges on the ions calculated at 298 K an increase of about 4.. 3rd ed. Paulechka. which impacts the enthalpy of vaporization for [C1mim][NTf2]. H. pyrrolidinium. Notes The authors declare no competing financial interest. Probably.97 4. This outlier could be understood by the competition of the corresponding Coulomb and van der Waals interactions in the comparably small [C1mim][NTf2]. A portion of this material is based upon work supported by the Air Force Office of Scientific Research under AFOSR Award Number FA9550-10-1-0244. Dz... Köddermann et al. 1950. This effect can be also attributed to the symmetry of the cation and a specific structuring of the bulk liquid due to this symmetry as well as due to the delicate 6484 CONCLUSIONS Our new experimental results based on the concurring results from QCM and TGA methods have revealed. Tel: +49 381 498 6508. ■ ■ † ASSOCIATED CONTENT S Supporting Information * Experimental details and data evaluation procedure. in spite of the obviously small difference of about 1 kJ·mol−1 between CH2 contributions in n-alkanes and in AUTHOR INFORMATION Corresponding Author *E-mail: sergey.doi.0 kJ·mol−1. although evidence is found in the simulations to support the formation of nonpolar domains of aggregation for cations having longer alkyl chain lengths. M. P. The specific value for the CH2 contribution in ILs is the clear evidence that the additivity rules are generally valid for ILs. and this value will significantly contribute to the reliable predictions especially for the long chained imidazolium.acs. the values of the additivity parameters for ILs are different from those for molecular compounds. The Values of the CH2 Group Increment into the Enthalpy of Vaporization for the Different Classes of Molecular and Ionic Compounds compound [Cnmim][NTf2] CnH2n+1CN CnH2n+1OH CnH2n+1C6H5 CnH2n+2 CH2CH-CnH2n+1 HS−CnH2n+1 Cl−CnH2n+1 Br−CnH2n+1 CnH2n+1CO2−CH3 CH2 increment. This seems to be a consequence of high dipole−dipole interactions in these molecular liquids. kJ·mol−1 3.9 kJ·mol−1 for the ILs under study (3. it is interesting to note that the enthalpy of vaporization of the fi rst representative of the series [C1mim][NTf2] deviates slightly from the linear correlation. ■ ■ ACKNOWLEDGMENTS This work has been supported by the German Science Foundation (DFG) in the frame of the priority program SPP 1191 “Ionic Liquids”.48 ± 0. most of the homologous series exhibit CH2 group contribution to the vaporization enthalpy very close around 5. and it is this contribution that results in the larger enthalpy of vaporization than would be expected. Chem. We have shown that enthalpies of vaporization generally obey group additivity.2 kJ·mol−1).1021/jp311429r | J. Phys. Methuen: London. however.76 4. this aggregation does not appear to change the slope of the enthalpy of vaporization versus chain length.03 refs this work (exp) this work (MD) 52 53 54 14 14 14 14 14 14 Article balance of Coulomb and van der Waals interactions.. It is important to note for further development of the additivity rules51 for ILs that the aforementioned competition of the Coulomb and van der Waals interactions in [C1mim][NTf2] seems to be completed for [C2mim][NTf2]. Strechan. Light and Matter”. the Δg l Hm correlation holds for all other values of n (see Figure 10). The MD simulations of the present study find that the increment is 3. A. Department “Life. the definite linear dependence of the vaporization enthalpies on the chain length. starting with [C2mim][NTf2].. and a simple method based on the experimental liquid heat capacities has been suggested. According to the MD simulations..89 ± 0. A 2006. dx.08 4.85 4. the intensive Coulomb forces specific for ILs also have a consequence in decreasing the CH2 group contribution to 3. Heintz. 110. this CH2 contribution was determined very precisely (±0. o However. This material is available free of charge via the Internet at http:// pubs.04 4.95 4. 117. or phosphoniumbased ILs. which is in quite good agreement with our current G. Phys.verevkin@uni-rostock. It is important to note that.80 5. Ambiguity of o the Δg l Cpm values required for temperature adjustments of vaporization enthalpies was resolved..

C. J.. I. Chem. Magee.. Phys. Enthalpies of Vaporization of Some 1-Substituted n-Alkanes. P.. Phys.. J. E. Chem. A. Verevkin. J.. C. 831−834. Phys. Chem. P. Prediction of Thermodynamic Derivative Properties of Fluids by Monte Carlo Simulation. M. J. The Distillation and Volatility of Ionic Liquids... Fast Parallel Algorithms for Short-Range Molecular Dynamics.. Krekeler. Vapor−Liquid Coexistence and Critical Behavior of Ionic Liquids via Molecular Simulations. Heintz. Verevkin. 519−879. Direct UV-spectroscopic Measurement of Selected Ionic-Liquid Vapors. H. P. J. J. 7246−7250. 3191−3195.. Schröder. Inc. A. Vapor Pressure Measurements. Gilea. Licence. 2010.. A. A Study of the Alkyl Chain Length Dependence. M. Acta 2012. A. Rev. 2006. Magee.. Ionic Liquids: First Direct Determination of their Cohesive Energy. Rebelo. A. J. A. L. A.. J. and ab Initio Calculations. (27) Morrow. Bleif. 11734−11742. W. P. P.. J.. S. Y. Phys. A.. Phys. B 2008. F. Baker. 154. (20) Kelkar. Critical Behaviour and Vapour−Liquid Coexistence of 1-Alkyl-3-methylimidazolium Bis(trifluoromethylsulfonyl)amide Ionic Liquids via Monte Carlo Simulations. Ionic Liquids: Dissecting the Enthalpies of Vaporization. C. V. J.. A. G. L.. G. Phys.. S. J. (13) Santos.: Pittsburgh. 6150−6155. (29) Youngs. Data 2003. Y. J. Santos. S. M. (25) Frisch. J. C.. S.. Deyko. (37) Esperanca.. Coutinho. Phys. Int. 1995. Zaitsau. Chem. Marrucho. 9424−9427. (4) Emel’yanenko. Am. A Molecular Dynamics Investigation of the Structural and Dynamic Properties of the Ionic Liquid 1-n-Butyl-3methylimidazolium Bis(trifluoromethanesulfonyl)imide. Kollman. (33) Strasser.. A. Heintz. Comput.. M... Maginn. Jess. C. J. Pyrrolidinium Based Ionic Liquids. Balasubramanian. 1977. R. J.. Paschek. J. P. 2011. Blokhin. Corfield. N. A. (36) Lagache. J.. A. G. Phys. C. M. A. M. R. 111. Chem. Chem. 9. S.. Kabo.. 2007. Berger. F. 3609−3613.. Emel’yanenk. 124507−16. Vaporisation Enthalpies of Imidazolium Based Ionic Liquids. Phys. (7) Verevkin. Tildesley. G. Y. 9. B. F. 129. Soc. Zaitsau. 110.. 1439−1443. 134103−134111. (41) Paulechka. Thermodyn. C. J. 4641−4644.The Journal of Physical Chemistry B (3) Armstrong. Frenkel. W. (40) Rai. Schick. Ludwig. Relative Volatilities of Ionic Liquids by Vacuum Distillation of Mixtures. Chem. B 2007. 2001. Eng. 6485 Article (22) Liu. E. J. (35) Ludwig. Phys. 106. J... Magee.. Mikami. Delle Site. (5) Luo. Stridh. N. J.. S. Phys. Marrucho. M. G. Vydrov. Schick. Cieplak. B 2010. Chem. P. J. 1−19. R. R... PA. P. Green Chem. L. R. W. 1453−1466. K. J. I. Phys. Widegren. Emelyanenko. S.. J. J.. N. J. 439. J. J. P. Case. Chem. P. Holm. Chem 2008. Zhao. 55−62. 2007. 1157−1174.. (28) Bhargava. Chem. Hurst. S. H. Jr. Phys. 117. Nanoscale Segregation in Room Temperature Ionic Liquids. T.. B. 48. Maginn. Analysis of Evaporation and Thermal Decomposition of Ionic Liquids by Thermogravimetrical Analysis at Ambient Pressure and High Vacuum. S. K. Rebelo. N. Ed.3-Dimethylimidazolium Chloride. Gomes. L. H. Wang. 3930−3937. (24) Wang. Mass Spectrometry and ab Initio Calculations. J. Gaussian. Comput. N... 127. L. J.. Y. 112. N. S. (42) Chickos. J. ̀ (18) Zaitsau. Ralys.. 8. 13. 2007. G. A. Wolf.-A. Dai.. 10077−10081. M. H. 549−555.. Magee. 457−462. J. W. (26) Bayly.. 55. A. A. N. V. Express Thermo-gravimetric Method for the Vaporization Enthalpies Appraisal for Very Low Volatile Molecular and Ionic Compounds. F. S. P. W. Lima. J. Yermalayeu. A Thermogravimetric Study of the Alkyl Chain Length Dependence.. S. Sunner. J. Tail Aggregation and Domain Diffusion in Ionic Liquids. Voth. N. Zaitsau. L. P. J. N. Computer Simulation of Liquids. Molecular Dynamics Study of the Ionic Liquid 1-n-Butyl-3-methylimidazolium Hexafluorophosphate.. O. Rebelo. Emelyanenko. 91−97. Nature 2006. Phys. Tariq.. O.. J. N.1021/jp311429r | J.. Phys. Chem. D. Phys. Vaporisation Enthalpies of Imidazolium Based Ionic Liquids. P. (8) Wang. M... L. C. I. P. R. (34) Rai. B 2006. et al. H. Licence. Faraday Discus. B. B 2013. P. (30) Schmidt. 2011.. W. J. Ralys. W.. N. Santos. M. B 2007. A. 97. J. V. (12) Widegren. B. Acree. N. H. M. (6) Heym. J.. (31) Plimpton. Enthalpies of Vaporization of Organic and Organometallic Compounds. P. Lovelock... Chem. Phys. Li.. Kragl. M. Dai. E. Chem. J. M. C. J. Wu. P. Vapourisation of Ionic Liquids. Heintz. J. Nanostructural Organization in Ionic Liquids. T. J.. Maginn. Phys. J. Chem. 53−69. V.. Russina.. 18601−18608... 3. H. D. J. Phys. Thermochim. dx. C. (11) Earle... M. 111. Phys. (21) Liu.. H. I. A Novel United-Atom Force Field for Imidazolium-Based Ionic Liquids. C. H. Am. Chem. 2012. A.. V. Caldwell. 982−990. Leone.. Kelkar.. 1548−1558. Maginn. Chem. Refined Potential Model for Atomistic Simulations of Ionic Liquid [bmim][PF6]. 2011. M. Do We Understand the Volatility of Ionic Liquids? Angew. Esperanca. P. (10) Rocha. 433− 437. A New Method for the Determination of Vaporization Enthalpies of Ionic Liquids at Low Temperatures. Boutin. J.. Application of Static Charge Transfer within an Ionic-Liquid Force Field and Its Effect on Structure and Dynamics. 9. Chem. Isothermogravimetric Determination of the Enthalpies of Vaporization of 1-Alkyl-3methylimidazolium Ionic Liquids... Thermodyn. Chem. C. Phys. D. Lett. A. J. 129... A. Chem.. Chem.. A. J. Goulay. J. T. J. J. 111. (43) Emel’yanenko. V. N. R. 6473−6486 . J. S. 135. M. Esperanca. Dommert. Chem. Chem. 9424−9427. The Gaseous Enthalpy of Formation of the Ionic Liquid 1-Butyl-3-methylimidazolium Dicyanamide from Combustion Calorimetry.. H.. 117.. L.. 3−12. 2007. Dz. Luo. G. Chem. Dz. C.. A. K. J. F. C. N. Villar-Garcia. E. R.. (32) Shinoda.. Kollman. Emel’yanenko. (17) Triolo. Padua. Wang. 1-Butyl-1-Methyl Pyrrolidinium Dicyanoamide: Thermochemical Measurement. E. W. Canongia Lopes.. N. Lopes. Eng.. S.. Esperanca. J. Gaussian 09. M. 10269−10280. Chem. Seddon. 8959−8964. 238. X. Etzold. A. Dz. 2011. J. Jones. 111. B 2006. J.. J... 284−285. V. Rapid Estimation of Elastic Constants by Molecular Dynamics Simulation under Constant Stress. Phys. J. 2004. B 2008... 2. Lovelock. E. N. Chem. B. Phys. Maginn. 4333−4339. Phys. Chem. Henderson. M. 12889−12895. Data 2010. J. Jones. Data 2003. Chem. 12. N. J. R.. E... S. Ionic Charge Reduction and Atomic Partial Charges from First-Principles Calculations of 1. A. 2009. 114. Verevkin. M. U... 3330−3335. J.. B 2002.. S. M. Kern. 12807−12813. J. S. HighAccuracy Vapor Pressure Data of the Extended [CnC1im][Ntf2] Ionic Liquid Series: Trend Changes and Structural Shifts. Cornell.. B 2007. L. R. W. Claredon Press: Oxford. 1096−1104.. R. 12. Calculating the Enthalpy of Vaporization of Ionic Liquid Clusters. 46. 25.. I. Maginn. M. P. R. Satterley. J. Phys. Fuchs. Lopes. Di Cola.. Heintz. M. Ref. H. J. Rebelo. Phys. Calculating the Enthalpy of Vaporization for Ionic Liquid Clusters. 111. P. Dz. Sellers. A.. K. (16) Wang. A 2007. V. Phys... S. Phys. 2006. Chem. Chem. E. Maginn. A. E.. 10919−10926. 32. 110. A. Soc. J. P. Chem. A. Chem. 69. 7. Chem. F. C. J. (39) Köddermann. 112. L. Hardacre. Gomes..doi. B 2007. 2012. P. B 2011. (38) Kelkar. A. B 2011. B. Development and Testing of a General Amber Force Field. Chem.. 6582−6584. C. N. Ungerer. 115. Shimizu. P. S. Thermodynamic Properties of 1-Butyl-3methylimidazolium Hexafluorophosphate in the Ideal Gas State. ̀ (9) Verevkin. A WellBehaved Electrostatic Potential Based Method Using Charge Restraints for Deriving Atomic Charges: The RESP Model. 2007. R. V. J. (15) Lopes. Phys. R. Chem. Chem. Coutinho. D. Phys. V. R. Phys. (14) Mansson. B 2004. S. P. Photoelectron Spectrum of Isolated Ion-Pairs in Ionic Liquid Vapor. Chem. C. F... N. S. S. N.. Chem. 1880−2002. U... Z. Phys. 1993. 114510−114516... Chem. 54. Volatility of Aprotic Ionic Liquids  A Review. Chem. 115.. (23) Allen. Canongia Lopes. (19) Verevkin. A. 1987.

Mol. Data 2010. I. 2005. Liq. Y. Thermodyn. (51) Verevkin. A. 15708−15717. Chem. Dz. Fluid Phase Equilib. P.. Chem. Enthalpies of Vaporization of a Series of Aliphatic Alcohols: Experimental Results and Values Predicted by the ERAS-Model. Ref. 1954. J.. E. 73−81. G. A... 187−207. (50) Roux. 2006. 2005.. D. Phys.. S. G.1021/jp311429r | J. Pergamon Press: New York/London/Paris. Kabo. B 2008. J. (45) Paulechka. Zaitsau. Koutek. nist. Doubsky. 2004.. U. Conformations. 14994−15000. J. Kabo. N. S. Heat Capacity of Room-Temperature Ionic Liquids: A Critical Review. H.. J. Phys. (46) Moelwyn-Hughes. A. P. 39. 37.. 2001. Chickos. Thermochim. Dz. Thermochemistry of ImidazoliumBased Ionic Liquids: Experiment and First-Principles Calculations. Thermodyn. Blokhin. (48) Ionic Liquids Database − ILThermo: http://ilthermo. Chem. Emel’yanenko. 115. (49) Strechan. V. E. (47) Paulechka. Paulechka. Thermodyn. N. B 2013. S. Verevkin. J. J. Vapour Pressures and Enthalpies of Vapourization of a Series of the Linear Aliphatic Nitriles. (54) Verevkin. V.. J. S. Structure. Shaplov. J. J. (53) Kulikov. 112. Thermochemical Properties of 1-Butyl-3-methylimidazolium Nitrate. Heintz. Y. 6473−6486 . 941−953.. S. 12. Fusion and Sublimation Enthalpies of the Dicarboxylic Acids from C4 to C14 and C16. C.. G. P. 37.doi. 105−111. B. U. D. Chem. Frenkel. Chem. A. Kabo. Vapour Pressures and Enthalpies of Vaporization of a Series of the Linear n-Alkyl-benzenes. Zaitsau.. Phys. U. (52) Emel’yanenko. Phys.The Journal of Physical Chemistry B (44) Paulechka. Kabo. Vibrations. A. 117.. Chem.. M.. J. G. Temprado.. M.. Vaporization. Lozinskay. Acta 2008.. Muzny. Phys. Y. U. J. 2010.boulder. V. A. A. 38. Verevkin. Article 6486 On the Difference between Isobaric and Isochoric Heat Capacities of Liquid Cyclohexyl Esters. Heintz.. Y. N. M. 474. Physical Chemistry.. V. H. Emel’yanenko.. and Ideal-Gas Properties of 1-Alkyl-3methylimidazolium Bis(trifluoromethylsulfonyl)imide Ionic Pairs and Constituent Ions. V. P. S. Chem. Chem. 192. 1111−1123. A. 25−31. 033108− 1−033107−23.

Sign up to vote on this title
UsefulNot useful