Quantifying intermolecular interactions of ionic liquids using cohesive energy densities.

For ionic liquids (ILs), both the large number of possible cation + anion combinations and their ionic nature provide a unique challenge for understanding intermolecular interactions. Cohesive energy density, ced, is used to quantify the strength of intermolecular interactions for molecular liquids, and is determined using the enthalpy of vaporization. A critical analysis of the experimental challenges and data to obtain ced for ILs is provided. For ILs there are two methods to judge the strength of intermolecular interactions, due to the presence of multiple constituents in the vapour phase of ILs. Firstly, cedIP, where the ionic vapour constituent is neutral ion pairs, the major constituent of the IL vapour. Secondly, cedC+A, where the ionic vapour constituents are isolated ions. A cedIP dataset is presented for 64 ILs. For the first time an experimental cedC+A, a measure of the strength of the total intermolecular interaction for an IL, is presented. cedC+A is significantly larger for ILs than ced for most molecular liquids, reflecting the need to break all of the relatively strong electrostatic interactions present in ILs. However, the van der Waals interactions contribute significantly to IL volatility due to the very strong electrostatic interaction in the neutral ion pair ionic vapour. An excellent linear correlation is found between cedIP and the inverse of the molecular volume. A good linear correlation is found between IL cedIP and IL Gordon parameter (which are dependent primarily on surface tension). ced values obtained through indirect methods gave similar magnitude values to cedIP. These findings show that cedIP is very important for understanding IL intermolecular interactions, in spite of cedIP not being a measure of the total intermolecular interactions of an IL. In the outlook section, remaining challenges for understanding IL intermolecular interactions are outlined.

Why are intermolecular interactions of ionic liquids important?

Ionic n>an class="Chemical">liquids (ILs), liquids composed entirely of mobile cations and anions, are important from both an academic and an industrial standpoint [1,2]. ILs have been proposed as electrolytes, solvents for synthesis and catalysis, gas capture/storage media, lubricants, engineering fluids and in sensors [3-10].

The strength of inten>an class="Chemical">rmolecular interactions have historically been used to understand and predict a range of properties for molecular liquids, including surface tension [11], wettability [12], the ability of a solvent to promote solute self-assembly [13,14], solubility [15] and viscosity [16-19]. All of these properties are very important for potential IL applications. For example, surface tension and wettability are key properties in understanding supercapacitors [20,21], and solubility is crucial for understanding and predicting gas separation/capture/storage [22]. Therefore, establishing links between these and other properties will give significant insight into the properties that underpin a wide range of applications.

A key propn>en>an class="Chemical">rty of ILs is their very low vapour pressure [23]. Hence, it is of interest to quantitatively investigate the intermolecular interactions of ILs to answer the fundamental question: do ILs have strong intermolecular interactions? In many publications it has been assumed that strong electrostatic intermolecular interactions drive the very low vapour pressures of ILs [24-27].

A pure molecular liquid contains only one constituent, the molecule; intermolecular interactions are controlled by molecule–molecule interactions. The situation is more complicated for ILs, as there are two constituents in each liquid: a cation and an anion. Both constituents will strongly influence the intermolecular interactions, making understanding intermolecular interactions of ILs much tougher than for molecular liquids. There are also many possible ILs and, therefore, many possible cation–anion combinations, providing an additional challenge to understanding intermolecular interactions of ILs.

In this artin>an class="Chemical">cle, attempts are made to find correlations for experimental measures of intermolecular interactions that hold across both molecular liquids and ILs; such correlations are particularly valuable, as these would reveal remarkable underlying similarities in intermolecular interactions. However, such correlations are expected to be very rare, given the very different chemical constituents of molecular liquids and ILs. Almost as valuable would be correlations for all ILs; such relationships that hold across ILs with a wide range of structural features are relatively rare. Lastly, family-based correlations can give significant insights into IL properties, although such correlations for relatively small groups of ILs have limited use for predicting IL properties. In this article the aim is to develop correlations for ILs using only experimental data. For certain values the required amount of experimental data does not yet exist. In such cases, quantitative correlations were not sought; however, qualitative observations were made. Data from simulations and calculations can provide an excellent complement to experimental data, and in some cases provide access to information that has not yet been experimentally measured.

In this article I use abbreviations to refer to individual ILs, instead of the IL names. The IL names, along with the relevant abbreviations, are given in the ESI, electronic supplementary material, table S1 (for cations) and electronic supplementary material, table S2 (for anions).

Quantifying the strength of intermolecular interactions of molecular liquids

How are the strengths of intermolecular interactions of molecular liquids quantified?

For molecular liquids the strength of intermolecular interactions is quantified using the cohesive energy density, cedML: [28,29] where ΔvapU is the internal energy of vaporization, i.e. the energy required to vaporize the liquid to its saturated vapour, Vm is the liquid molar volume (Vm = M/ρ, where M = molar mass and ρ = liquid density), ce is the cohesive energy, ΔvapH is the enthalpy of vaporization, R is the gas constant and T is the liquid temperature. cedML is a measure of the total intermolecular cohesion per unit volume in the liquid phase, assuming negligible intermolecular interactions in the vapour [28]. Alternatively, cedML quantifies the work required to produce a hole (often called a cavity) of unit volume in the molecular liquid [30].

Conditions used for vaporization measurements for molecular liquids

Vaporization is a physin>an class="Chemical">cal process; no chemical change to form new products occurs. There are two types of vaporization: evaporation and boiling. Evaporation is a surface process; molecules escape from the liquid surface in the gas phase, where the gas phase is not saturated with the vapour of that molecule. Boiling is a bulk process; bubbles of saturated vapour are formed by molecules in the liquid phase. In this article the focus is on evaporation. ILs have been categorized as superheated liquids, and hence, observing IL boiling is very unlikely [31].

There an>an class="Chemical">re two extremes of vaporization conditions. Firstly, those close to equilibrium conditions, which are preferred for obtaining thermodynamic data. These conditions are generally produced using a Knudsen effusion cell, which is a heated source containing a pinhole to allow vapour to slowly escape under effusive conditions, thus giving near-equilibrium conditions. Secondly, those far from equilibrium conditions. Evaporation occurs from the liquid–gas surface and the vapour is detected directly (with no further interactions and no chance of condensation then re-evaporation of the vaporized sample); these conditions are known as free or Langmuir evaporation. Methods for measuring ΔvapH that involve a Knudsen effusion cell require higher sample temperatures (and, therefore, higher vapour pressures) than methods involving Langmuir vaporization (due to the effective smaller sample surface area available for vaporization). Higher sample temperatures also increase the probability of the chemical process of thermal decomposition (TD) occurring. For most molecular liquids the probability of vaporization, compared to the probability of TD, is sufficiently large (i.e. only sample vaporization occurs) that near-equilibrium methods are used to study vaporization, meaning that cedML can be readily determined for most molecular liquids [32].

For most molecular liquids there are negligible intermolecular interactions in the vapour, as they generally vaporize as individual isolated molecules. Therefore, cedML is a measure of the total intermolecular cohesion per unit volume in the liquid phase.

Results for the cohesive energy density for molecular liquids

cedML,298 values n>an class="Chemical">range from 195 J cm−3 for squalane, through 385 J cm−3 for acetone, 679 J cm−3 for ethanol, 1055 J cm−3 for ethanolamine and up to 2293 J cm−3 for water [32]. Liquid hydrocarbons can be used to explain how ΔvapH298 and ced relate to intermolecular interaction strengths, and how size is vital. ΔvapH298 increases with increasing size, from n-hexane (ΔvapH298 = 31 kJ mol−1) to n-hexadecane (ΔvapH298 = 81 kJ mol−1) to squalane (ΔvapH298 = 105 kJ mol−1) [32,33]. However, clearly only van der Waals (vdW) interactions are added as the hydrocarbons increase in size. This fact is reflected in the cedML,298 values, which are in the range of approximately 190 to approximately 270 J cm−3 for liquid hydrocarbons [32]. These cedML,298 values are all very small, as expected when only vdW interactions are involved. Each liquid-phase squalane molecule forms a large number of relatively weak intermolecular interactions, all of which need to be broken in order to vaporize a squalane molecule. The increase in ΔvapH298 from n-hexane to squalane is not due to stronger intermolecular interactions; there are simply more intermolecular interactions per molecule (but not per unit volume).

Many molecular liquids have significant electrostatic interactions, e.g. water, glycerol, formamide, ethanolamine. For water ΔvapH298 = 44 kJ mol−1 and for formamide ΔvapH298 = 61 kJ mol−1, much lower than ΔvapH298 for n-hexadecane and squalene [32,33]. However, water and formamide are significantly smaller than n-hexadecane and squalene [32,33]. Therefore, water and formamide give much larger cedML values than for hydrocarbons [32], demonstrating that the intermolecular interactions in molecular liquids that have significant electrostatic interactions are much stronger than in hydrocarbons.

How the cohesive energy density relates to other properties for molecular liquids

Given the interest in correlating ced with other properties, it is vital to know how cedML values correlate to other liquid-phase properties.

Surface tension

Surface tension, γ, is a key parameter for colloid science and engineering [14]. Surface tension represents the energy required to separate a bulk liquid to create two new liquid–gas surfaces, i.e. the energy per surface area. Forming such surfaces requires breaking bulk intermolecular interactions [34-38]. In 1896, Stefan proposed that the work necessary to bring a molecule from the interior of the liquid to the surface is half that needed to vaporize it [37]. For samples that will thermally decompose before vaporization occurs, i.e. involatile liquids and polymers, the Gordon parameter, G [11], has been used as a substitute for ced: [14,39] where γ is the surface tension and Vmol is the molecular volume (Vmol = Vm/Na, where Na = Avogadro's number). Owing to the possible influence of the surface structure on γ and therefore G, ced is generally preferred over G for investigating intermolecular interactions. However, for non-volatile samples G is easier to measure than ΔvapH. A very good linear correlation between cedML and GML is observed for a small selection of molecular liquids (figure 1).

Figure 1.

cedML versus GML for select molecular liquids.

Liquids as solvents for self-assembly

Solute self-assembly in solvents is vital in many areas, inn>an class="Chemical">cluding in nature (in particular in water), drug delivery, inorganic materials and many more [13,14]. Until recently only approximately 20 solvents were known to promote self-assembly [40], although more have been found recently [41]. In particular, many protic ILs and a number of aprotic ILs have been found to promote self-assembly [40,42,43]. For ILs there are essentially two categories of cation: aprotic ILs are formed by transfer of an alkyl chain from e.g. an alkylchloride to a base, whereas protic ILs are formed by transfer of a proton from a Brønsted acid to a Brønsted base [44,45]. ced can be considered to represent the tightness or structuredness of liquids as caused by intermolecular interactions [30]. It has been demonstrated that structured liquids support self-assembly of amphiphilic molecules, due to the large ced [13,41]. Structural properties of liquids acting as solvents can be categorized by stiffness, as measured by the work that must be expended to create a hole in the liquid, as expressed by the liquid ced [46]. Evans has proposed that liquids with G (as a substitute for ced) > ∼ 110 J cm−3 tend to promote aggregation of amphiphilic molecules, i.e. G and ced can be used as measures of a solvent's solvophobic strength [13,14,47].

Solubility

Pn>an class="Chemical">redicting the solubilities/miscibilities of gases/solvents in/with ILs is of vital importance [4,22], and a general prediction scheme would be of great use. Hildebrand showed that for non-polar liquids: where δH is the Hildebrand solubility parameter [28]. δH is typically small for non-polar solvents and large for polar solvents [48]. If δH is similar for solvent and solute then it is predicted that solvation will occur, as similar strength interactions will be broken and formed. δH predicts solubility well for non-polar and slightly polar molecules, where vdW interactions are dominant. For systems where electrostatic interactions (e.g. polar liquids) are important, a one-component parameter such as δH does not represent the system sufficiently, and more components are required to explain solubility [15,49]. Hansen solubility parameters comprise: polar, non-polar and hydrogen bonding, and they can be estimated using indirect methods.

Activity n>an class="Chemical">coefficients give a measure of interactions between solute and solvent, and are used to obtain δH,indirect for liquids (δH values for ILs determined from activity coefficients are referred to in this article as δH,indirect, to differentiate between δH values obtained from ced). Activity coefficients at infinite dilution for solutes in ILs have been measured using inverse gas chromatography, and δH,indirect was obtained (ced values for ILs determined from these δH,indirect values are referred to in this article as cedindirect) [50].

Viscosity

Viscosity is a key van>an class="Chemical">riable in process design, e.g. mixing, separation, lubrication. Viscous flow is often modelled by the activated jumping of a molecule from an initial configuration to a second position, separated by an intermediate activated state [18]. Eyring and co-workers postulated that the activated state of the molecule requires a larger volume than the initial state. As this larger volume requires hole creation in the liquid, the activation energy for viscous flow, Ea,vis, should be a fraction of ΔvapU: where C is a constant for ‘normal’ liquids. For non-spherical molecules (e.g. hexane and acetone) C ≈ 4; for small spherical molecules (e.g. CCl4 and cyclohexane) C ≈ 3. For both liquid metals and molten salts the unit of flow was considered to be important; to obtain a satisfactory linear correlation between ΔvapU and Ea,vis, atomic/ionic radii were included in correlations [19,51]. A further possible complication is that the relationship between viscosity and temperature does not follow Arrhenius-like behaviour just above the glass transition temperature for all ILs (i.e. many ILs are fragile liquids), leading to possible errors in measured Ea,vis for ILs [52].

δH, and therefon>an class="Chemical">re ced, can also be determined using measurements of intrinsic viscosity (ced values for ILs are again referred to as cedindirect). The viscosity of a dilute solution of a solute (e.g. a polymer) is a maximum in the ‘best’ solvent, i.e. the solution where cohesion properties of solvent and solute are comparable [53].

Internal pressure

The internal pn>an class="Chemical">ressure of a liquid, Pint, is another property that can be used to quantify intermolecular interactions. cedML depends on breaking all intermolecular interactions. Pint for molecular liquids depends on small isothermal expansion; intermolecular distances increase slightly so not all intermolecular interactions are broken. The ratio: is the standard method to compare Pint to ced. r can be used to categorize solvents. Liquids where r > 1.2 are labelled as loose, e.g. fluorocarbons, triethylamine (repulsive interactions are relatively large) [54]. Liquids with r < 0.8 are collectively labelled as associated/tight/stiff liquids, e.g. water, ethanolamine (attractive interactions are strong) [54]. Liquids with r ∼ 1 are non-polar liquids, e.g. hydrocarbons, diethyl ether, toluene [28,54]. This ratio also highlights that Pint and ced have the same units, generally MPa or J cm−3.

Measuring the cohesive energy density for ionic liquids: overcoming the challenges

There an>an class="Chemical">re two major experimental challenges for obtaining ced for ILs (Vm can be readily measured with good accuracy for ILs [55]). Firstly, measuring the amount of IL ionic vapour at different temperatures to obtain ΔvapH. The inherently low vapour pressure of ILs makes these measurements difficult. For [C2C1Im][NTf2] at T = 298 K the vapour pressure can be estimated as approximately 10−9 Pa [56], a much lower value than the expected detection limits for any apparatus used for studying vaporization of molecular liquids. Secondly, the ionic vapour composition needs to be known under the conditions for which ΔvapH is measured (ionic vapour refers only to the vaporization of intact IL) [57], so that the nature of the intermolecular interactions broken is known.

Ionic liquids can be distilled

For many years, it was assumed that IL TD followed by vaporization of the non-ionic TD products occurred more readily than IL vaporization of intact ionic vapour. Hence, it was assumed that ILs had no detectable vapour pressure [58-60].

Since 2005 it has ben>an class="Chemical">come accepted that ILs can be vaporized, in contradiction of previous assumptions. The major evidence for IL vaporization was their successful distillation, i.e. the physical process of IL vaporization, followed by condensation of the same species that was vaporized. Bulk-scale distillation of intact IL has been achieved for 1,3-dialkylimidazolium-, 1,1-dialkylpyrrolidinium- and tetraalkylammonium-based ILs at a temperature range of 475 to 575 K [61-63]. Most of the focus has been on [NTf2]−-based ILs, which have the required combination of relatively good thermal stability and relatively high volatility [31,64,65]. Aprotic ILs are the focus of this article, as very few vaporization studies of protic ILs have been published, in particular on identification of the vapour phase species [66-69].

The vapour phase composition of ionic liquids: the ionic vapour

In the seminal 2006 Nature an>an class="Chemical">rticle demonstrating distillation of ILs, it was speculated that IL vaporization occurred as isolated ions or as ion aggregates [61,70]. Additionally, a chemical process of IL transfer could not be ruled out.

The ionic vapour: intact ions in the vapour phase

For a wide range of ILs using positive mode mass spectrometry (MS) (with electron ionization, field ionization or photoionization) the parent cation, [C]+ [57,63,66,71-89], has been detected intact in the vapour phase (at 350 K < T < 700 K). In addition, the parent anion, [A]− [66,83,85], has been detected intact in the vapour phase using negative mode MS (with electron ionization or photoionization). The cations investigated include [CC1Im]+, [CC1Pyrr]+, [CPy]+, tetraalkylphosphonium and alkylisouronium; the anions investigated include [NTf2] −, [PF6] − and [TfO] −. If IL TD products were vaporizing instead of intact IL vaporizing, one would not expect to detect intact parent cations and parent anions. For example, electron ionization MS of the products of heating [CC1Im][A] ILs (A− = halide ion) found non-ionic TD products such as 1-methylimidazole and alkyl halides [77,90]. Therefore, intact ionic has definitely been detected for a wide range of ILs. Leading on from this detection of intact parent ions in the mass spectra, there are two key questions to be answered about the ionic vapour.

The ionic vapour is composed of neutral species, not isolated ions

For a wide range of ionic vapours, when the MS ionization source, e.g. electrons, was turned off, no ions were detected; it was concluded that the vapour phase of ILs is composed primarily of neutral species, not isolated ions [57,63,66,71-77]. More recently, using a specialist Knudsen effusion cell to produce the vapour of [C2C1Im][NTf2] and electric fields to extract any overall charged ions in the vapour phase, [C]+, [C2A]+, [A] − and [CA2] − were detected [87,91]. However, the amount of ions in the vapour phase was approximately 108 to 1011 times lower than the amount of CA ion pairs [56], demonstrating the lack of significant amounts of overall charged ions in the vapour phase for [C2C1Im][NTf2]. Lastly, the intact molecular ion, [CA]+• = [C4C1ImC(CN)3]+•, was detected for [C4C1Im][C(CN)3] using both field ionization and photoionization at m/z 229 [78,82], demonstrating that the lack of [CA]+• detected for other ILs should not be taken as an indication of the lack of neutral vapour prior to ionization. Overall, these findings show that the vapours of the ILs studied to date are composed predominantly of neutral species.

The ionic vapour is composed of one cation and one anion, not larger clusters

No higher mass n>an class="Chemical">clusters of the form [CA–1]+ (where m ≥ 2) have been observed after either electron ionization or photoionization of a wide range of ILs [57,63,66,71-77,79-82]. It has been suggested that for electron ionization too much excess energy is deposited into the ionic vapour, which will lead to the break-up of higher mass clusters such as C3A3 [92] (if such clusters are present in the ionic vapour). Vaporization as CA where m ≥ 2 occurs for alkali halides such as LiCl, and after electron ionization cluster ions such as [Li3Cl2]+ were detected [93]. This finding suggests that electron ionization MS may be able to detect higher mass clusters such as C3A3 (if such clusters are present in the vapour phase). For [C4C1Im][C(CN)3] the intact molecular ion, [CA]+•, has been detected using field ionization and photoionization [78,82], but no higher mass clusters were detected up to m/z 600; no [C2A]+ (m/z 368), [C2A2]+• (m/z 458) or [C3A2]+ (m/z 597) were observed [82]. [CA]+• is expected to have a significantly weaker intermolecular cation–radical interaction than the cation–anion intermolecular interactions in [C2A]+. Therefore, observation of [CA]+• but not [C2A]+, [C2A2]+• or [C3A2]+ shows that [C2A]+, C2A2 or [C3A2]+ were not present in significant concentrations in the ionic vapour.

Rean>an class="Chemical">ctions of the form [C]+ + (ionic vapour) → (product ions) and [A] − + (ionic vapour) → (product ions) were carried out for [C6C1Im][NTf2], [P6,6,6,14][TfO] and [C4C1Pyrr][NTf2] [66]. [C2A]+ and [CA2] − were detected for the respective reactions, as would be expected if the ionic vapour was composed of CA neutral ion pairs. However, larger cluster ions, e.g. [C3A2]+, were not detected, demonstrating that the ionic vapour was not composed of C2A2 clusters, or any larger clusters.

Two ionization mechanisms have been proposed for ILs, demonstrated below for photoionization: dissociative ionization [71] (mechanism 1) and CA neutral ion pair dissociation [83] (mechanism 2):

It should also be noted that one dicationic IL has successfully been vaporized. The ionic vapour was shown to consist of CA2 neutral ion triplets [94,95].

The ionic vapour under equilibrium conditions

Most of the studies descn>an class="Chemical">ribed in §§4.2.1 to 4.2.3 were performed under Langmuir vaporization conditions (for ILs with a wide range of different structures), i.e. not under equilibrium conditions [57,63,66,71-85]. Therefore, it is very important to highlight those studies that were performed under near-equilibrium conditions. A combination of a Knudsen effusion cell with a small orifice with MS detection has been used by a number of research groups to study the ionic vapour of [CC1Im][NTf2] [86-89,91]. All evidence is in agreement with that given already in §4, demonstrating that the equilibrium ionic vapour for [CC1Im][NTf2] ILs is composed of neutral ion pairs. At present ILs with other anions have not been successfully vaporized using a Knudsen effusion cell, e.g. [CC1Im][PF6] ILs thermally decomposed in a Knudsen effusion cell [96].

When flash heating an IL/acetonitrile mixtun>an class="Chemical">re, i.e. heating from room temperature to very high temperature very rapidly (nearly instantaneously to incandescence, expected to give T > 1000 K), larger clusters of the type [C2A]+ and [CA2] − were detected using MS [92]. It was suggested that these clusters were formed via gas-phase decomposition of neutral clusters, e.g. C2A2 [92]. However, such conditions are very far from equilibrium conditions and ΔvapH was not measured in this study; the composition of the ionic vapour at such temperatures does not appear to be relevant for ΔvapH measurements.

Investigating the ionic vapour using simulations and calculations

Simulations and calculations in the tempn>en>an class="Chemical">rature range relevant to IL experimental vaporization studies (360 K < T < 630 K, see electronic supplementary material, table S3) have shown mainly isolated, neutral ion pairs in the ionic vapour [97,98]. Simulations at temperatures well above those at which vaporization experiments have been carried out (i.e. T > 800 K) have suggested the presence of higher mass clusters [97,99-102]. In addition, below T approximately 350 K calculations found that the major constituent of the ionic vapour was C4A4 for [C4C1Im][CF3SO3] [97]. Measurements of ΔvapHT have been made at a wide variety of different T, ranging from 370 to 625 K [57]. As will be explained in §4.4.2, extrapolations of ΔvapHT are regularly made to T values for below those at which ΔvapHT was measured, especially to T = 298 K. An important assumption for such extrapolations is that the ionic vapour remains the same at the target T as that T at which ΔvapHT was measured. For example, to be confident in an extrapolation to T = 298 K, one needs to know what the ionic vapour composition is at 298 K. At present, the best effort to experimentally determine the ionic vapour composition near T = 298 K is at T = 340 K for [C2C1Im][NTf2], which suggested that the ionic vapour was composed of neutral ion pairs [57], although checking for C4A4 was not possible on the apparatus used due to m/z 1285 for [C4A3]+ being beyond the detector m/z range. Thus, there is currently no experimental evidence to confirm or deny the presence of clusters larger than ion pairs in the vapour of ILs at T = 298 K.

Isolated ions in the ionic vapour

It has previously been stated that the enthalpn>y of deson>an class="Chemical">rption for an IL ion pair from the bulk IL, ΔdesH(total), cannot be measured directly, which were reasonable statements at the time, given the very challenging nature of these experiments [103,104]. As noted in §4.2.2, for [C2C1Im][NTf2] high temperature-capable thermal ion emission MS has been used to detect ionic vapour species with net overall charge, i.e. very small quantities of isolated [C]+ and [A] − ions have been detected in the ionic vapour [87,91]. From measurements of the amount of [C]+ and [A] − ions in the equilibrium ionic vapour at different temperatures the enthalpy of cation desorption from the bulk IL, ΔdesH([C]+) and the enthalpy of anion desorption from the bulk IL, ΔdesH([A] −) have been determined, respectively [87,91]. Therefore, experimental values for ΔdesH([C]+) and ΔdesH([A] −) (and therefore, ΔsolvH([C]+) and ΔsolvH([A] −)) have now be obtained [87,91].

Summary

All evidence leads to the n>an class="Chemical">conclusion that vaporization occurs primarily as CA neutral ion pairs (i.e. no net overall charge), and vaporization as CA where m ≥ 2 does not occur to any significant level. There are very small amounts of isolated ions in the ionic vapour (i.e. species with net overall charge). There are open questions about the nature of the equilibrium ionic vapour composition for many ILs and the ionic vapour composition at room temperature; these will be outlined in §9.

Two different measures of cohesive energy density for ionic liquids

The moleculan>an class="Chemical">r unit for an IL is made up of one cation and one anion, i.e. the molar mass of an IL is always taken from one cation and one anion, MIP. Much research has been carried out into the occurrence of ion pairs in bulk ILs; the general consensus is that long-lived ion pairs do not exist as individual entities in the bulk of ILs, at least at room temperature [105]. Therefore, the molecular units in the bulk IL (i.e. isolated ions, see §4.2) and in the ionic vapour (i.e. isolated neutral ion pairs) are different. As the bulk IL is made up of individual ions, the cation–anion interaction present in the neutral ion pair ionic vapour is deemed as intermolecular, as has been concluded elsewhere [104]. This judgement leads to two different versions of ced for ILs, here labelled cedIP (ced when the ionic vapour is a neutral ion pair CA) and cedC+A (ced when the ionic vapour is isolated ions, [C]+ and [A] −).

The traditional equation for ced (see equation (2.1)) can be used for ILs: where IP denotes ion pair, as the ionic vapour is a neutral ion pair (scheme 1, step (i)) and Vm is the IL molar volume (Vm = MIP/ρ, where MIP = IL molar mass and ρ = IL density). Therefore, cedIP of an IL can readily be obtained using ΔvapH (scheme 2, step (i)).

Scheme 1.

Extended Born–Fajans–Haber cycle for transformations of an ionic liquid in terms of enthalpies. Adapted from Preiss et al. [103].

Scheme 2.

Transformations of an ionic liquid in terms of ced.

The energy to form isolated cations and isolated anions in the gas phase from the liquid phase can be obtained from ΔdesH(total) (scheme 1, step (ii)). ΔdesH(total) = −ΔsolvH(total), where ΔsolvH(total) represents the sum of the enthalpy of cation solvation from the vapour phase into the bulk IL, ΔsolvH([C]+), and the enthalpy of anion solvation from the vapour phase into the bulk IL, ΔsolvH([A] −). cedC+A can be obtained by measuring ΔdesH(total) (scheme 2, step (ii)).

Isolated gas-phase cations and anions n>an class="Chemical">can also be formed by vaporizing a neutral ion pair (scheme 1, step (i)) and then breaking the cation–anion intermolecular interaction (scheme 1, step (iii)). Therefore, cedC+A can be obtained by measuring ΔvapH and the enthalpy of vapour phase neutral ion pair dissociation, ΔdissH(CA) (scheme 1, step (ii)).

As there are significant intermolecular interactions in the vapour, cedIP does not capture all liquid-phase intermolecular interactions for an IL, i.e. cedIP is not a measure of the total intermolecular cohesion per unit volume in the IL. cedC+A captures all liquid-phase intermolecular interactions for an IL, i.e. cedC+A is a measure of the total intermolecular cohesion per unit volume in the IL.

Problems caused by the low vapour pressure of ionic liquids

Monitoring the amount of ionic liquid vaporized

Accurately detecting the very small amounts of IL vaporization with respect to temperature is very challenging experimentally, especially with apparatus developed for molecular liquids. Hence, methods have been developed to measure IL vaporization at the temperatures required. Improvements in the sensitivity of vaporization detection mean that measurements have been made at increasingly lower temperature; therefore, TD is expected to be less of a complication for measurements made at lower temperatures.

There an>an class="Chemical">re broadly two methods to monitor the amount of IL vaporized with respect to temperature (scheme 3). Method 1: measure the amount of IL lost on vaporization from the liquid phase. Techniques that can be used to monitor the mass lost as the IL vaporizes include Knudsen effusion mass loss [56,88], calorimetry [106,107], thermogravimetric analysis (TGA) [31,64,65,108-116] and transpiration [117]. Method 2: measure the amount of ionic vapour in the gas phase produced by IL vaporization. Within Method 2, there are two approaches to determine the amount of ionic vapour in the gas phase: (a) measure the amount of ionic vapour in the gas phase (temperature-programmed MS (TPMS) [57,63,71-77,86], temperature-programmed photoelectron spectroscopy (TPPES) [80], UV absorption spectroscopy [118]), (b) measure the amount of ionic vapour that has condensed from the gas phase onto a solid surface (quartz crystal microbalance (QCM) [96,110-112,114-116,119-136], magnetic suspension balance (MSB) [31]).

Scheme 3.

The two methods used to monitor the amount of IL vaporized with respect to temperature. Within Method 2, there are two approaches to determine the amount of ionic vapour in the gas phase. TGA, thermogravimetric analysis; TPMS, temperature-programmed mass spectrometry; TPPES, temperature-programmed photoelectron spectroscopy; QCM, quartz crystal microbalance; MSB, magnetic suspension balance.

The two methods used to monitor the amount of IL vapn>on>an class="Chemical">rized with respect to temperature. Within Method 2, there are two approaches to determine the amount of ionic vapour in the gas phase. TGA, thermogravimetric analysis; TPMS, temperature-programmed mass spectrometry; TPPES, temperature-programmed photoelectron spectroscopy; QCM, quartz crystal microbalance; MSB, magnetic suspension balance.

Comparing measured ΔvapHT: adjusting ΔvapHT to a common temperature

Measurements of Δvapn>HT have been made at a wide range of T, from 370 to 625 K [57]. To make valid comparisons of ΔvapHT a common T is needed. To obtain ΔvapHT at a constant T (usually T = 298 K) an extrapolation is required from the measurement T to the constant T used for ΔvapHT comparisons. Such an extrapolation requires knowledge of the heat capacity at constant pressure, ΔglCp, between the IL in the vapour and liquid phases, over the T range of interest. It has been noted that this extrapolation was a major source of discrepancies for published ΔvapH298 values for ILs [111]. For ILs, the liquid phase heat capacities can be readily measured [137]. However, the gas-phase heat capacities have not been measured to date, and have been judged to be impossible to measure [111]. Thus, different methods of obtaining ΔglCp, have been proposed. However, the variety of different methods and ΔglCp values used means that a summation is required.

Quantum mechanin>an class="Chemical">cal and statistical thermodynamic calculations on ionic vapour neutral ion pairs have been used to obtain the gas-phase heat capacity for [CC1Im][PF6] [138]. Based upon this value, a value for ΔglCp in the region of −100 J K−1 mol−1 has been used for a wide variety of ILs [56,57,63,71-73,75,77,108,117,118,122,123,131,134,136]. Quantum mechanical and statistical thermodynamic calculations on ionic vapour neutral ion pairs have been used to obtain gas-phase heat capacities for [CC1Im][NTf2] (n = 2, 4, 6, 8) [139]. Correlations developed from these values have then been used to obtain ΔglCp for [CCIm][NTf2] ILs, ranging from ΔglCp = −112 J K−1 mol−1 for [C2C1Im][NTf2] to ΔglCp = −203 J K−1 mol−1 for [C10C10Im][NTf2] [129,130,132,133,135]. Clearly, there are very large and unsatisfactory discrepancies between these ΔglCp values and the widely used ΔglCp = −100 J K−1 mol−1 value; these differences can lead to significant variation in ΔvapH values, up to 20 kJ mol−1 for ILs with long alkyl chains [111].

In 2013, Verevkin et al. [111] analysed thein>an class="Chemical">r ΔvapH data for [CC1Im][NTf2] (n = 1–18) measured using two different techniques, TGA and QCM. As the T at which the ΔvapHT data were recorded using the two techniques were different by approximately 150 K, values for ΔglCp could be indirectly determined using this experimental ΔvapHT data. ΔglCp was found to range from ΔglCp = −56 J K−1 mol−1 for n = 2 to ΔglCp = −170 J K−1 mol−1 for n = 16. This procedure using variable −T ΔvapHT data has since been applied to more ILs, although to date they have all been for ILs with an [NTf2] − anion [88,106,114-116]; for example, ΔglCp = −67 J K−1 mol−1 for n = 4 has been found, in excellent agreement with the findings of Verevkin and co-workers [88]. Overall, ΔglCp has generally been found to be in the range of −40 to −120 J K−1 mol−1 for ILs with relatively few CH2 groups, whereas for ILs with relatively large amounts of CH2 groups ΔglCp has generally been found to be in the range of −120 to −200 J K−1 mol−1. All groups have found a chain length dependence of ΔglCp. Most recently, ΔglCp = −62 J K−1 mol−1 has been proposed using a combination of experimental ΔvapHT data measured at different T and educated guesses [140]. When compared to ΔglCp derived from both methods explained above (calculations and indirectly from experimental data), ΔglCp = −62 J K−1 mol−1 appears too small for ILs with relatively large numbers of CH2 groups, e.g. [C12C1Im][NTf2] and [C6C6Im][NTf2].

Overall, then>an class="Chemical">re are two possible options for obtaining ΔvapHT at a common T. Firstly, use ΔglCp values defined in the literature where available, and otherwise use a constant, IL independent ΔglCp value. This mix-and-match approach has the advantage of including the expected alkyl chain dependence of ΔglCp for certain ILs [111]; the clear disadvantage is the glaring inconsistency, as ΔglCp has not been determined by any method for many ILs for which ΔvapH values have been measured. The second option is to use a constant, IL independent ΔglCp value for all ILs. This approach has the advantages of simplicity and consistency, but with the considerable disadvantage of missing the alkyl chain length dependence. In this contribution a constant, IL independent ΔglCp = −100 J K−1 mol−1 was used, as this approach is a reasonable compromise, as it represents the approximate mid-point of ΔglCp values published to date.

In terms of selen>an class="Chemical">cting a common T, in almost all literature examples for ILs T = 298 K has been chosen; thus, ΔvapH298 is obtained. Using T = 298 K allows ready comparison to many other properties of ILs, e.g. surface tension. However, T = 450 K has also been used to ensure that the T extrapolation is minimal, as T = 450 K is similar to the T at which ΔvapH is measured [140]. In this contribution T = 298 K is used. As noted above, when using T = 298 K and ΔglCp = −100 J K−1 mol−1, large differences can occur for certain ILs. However, T = 450 K was also tested; using T = 450 K gave the same correlation observed in figure 3b (see electronic supplementary material, figure S1).

Figure 3.

cedIP,298 for ILs (extrapolated to T = 298 K using constant ΔglCp = −100 J K−1 mol−1) versus: (a) , (b) GIL, (c) GIL. The same data are plotted in (b) and (c); in (c) different IL families are plotted as the same colour data points. Details on the families are given in §7.1; [P6,6,6,14][BF4] and the dicationic IL [C3(C1Im)2][NTf2]2 do not fit into any of the four categories used here.

Monitoring ionic liquid vaporization only

IL vaporization only must be measured to ensure reliable ΔvapH data, i.e. the vaporization of liquid-phase non-ionic TD products must not be measured. Methods that can distinguish in situ between IL vaporization and vaporization of liquid-phase TD products are ideal. However, there are relatively few such methods available; for ILs, TPMS and TGA (using two carrier gases) are the main methods used for distinguishing IL vaporization and vaporization of liquid-phase TD products [31,64,65,77].

Methods that involve measuring the amount of vapn>oun>an class="Chemical">r that condenses from the gas phase onto a solid surface (e.g. QCM and MSB), when operated under high vacuum conditions (i.e. apparatus base system pressure less than 10−8 mbar), can be used to detect only IL vaporization. ILs are clearly sufficiently involatile that they will condense onto a solid surface at room temperature, whatever the system pressure. Many TD products of ILs are highly volatile, e.g. for halide-ion ILs [77,90]. Therefore, it is expected that such volatile TD products will not condense under high vacuum conditions at room temperature. Consequently, using QCM and MSB under high vacuum conditions would detect only IL vaporization, as long as the TD products were relatively volatile. However, not all TD products for ILs are highly volatile, e.g. for [CC1Im][BF4] ILs [141].

A number of popn>ulan>an class="Chemical">r methods of obtaining ΔvapH for ILs involve measuring the total amount of vaporization. Almost all methods that involve measuring the mass lost from the liquid-phase measure the total amount of IL vaporization, e.g. TGA using only one carrier gas [108-116], Knudsen effusion mass loss [56] and calorimetry [142]. Also, some methods that involve measuring the amount of ionic vapour rely on there being no significant vaporization of liquid-phase non-ionic TD products, e.g. vapour phase spectroscopy [80,118]. Therefore, these methods rely on no significant vaporization of liquid-phase TD products occurring. In some studies attempts have been made to check if TD occurred by characterizing either the condensate or the remaining IL that had not vaporized. Neither of these approaches is foolproof. For the condensate, as noted above, TD products may be sufficiently volatile that they do not condense under experimental conditions; for the remaining IL sample, observing no TD products in the liquid phase does not mean there was no TD product vaporized.

For [n>an class="Chemical">CC1Im][NTf2] ILs (n = 2 and 4) it has been demonstrated, using two carrier gas TGA, that ΔvapH is far smaller than the activation energy of TD [31,64]. In addition, [CC1Im][NTf2] ILs (along with [N2,2,2,2][NTf2] and [C4C1Pyrr][NTf2]) have been readily distilled without significant TD occurring [61-63]. These results indicate that for [cation][NTf2] ILs, where the cation does not contain relatively reactive functional groups, only IL vaporization is significant under the experimental conditions generally used for vaporization experiments. Therefore, the assumption that many researchers make that only IL vaporization occurs for [cation][NTf2] ILs appears to be valid, making ΔvapH values measured for [cation][NTf2] more trustworthy than for other ILs.

There is the absun>an class="Chemical">rd situation where the same technique, TGA with a single carrier gas, for the same families of ILs has been used to determine either ΔvapH or the activation energy of TD, i.e. in different papers it is assumed that only IL vaporization or only liquid-phase non-ionic TD product vaporization was occurring. This situation has occurred for carboxylate-based ILs [143,144], [NTf2] −-based ILs [110,111,145,146] and Cl−-based ILs [147,148]. Therefore, when using TGA it would appear best to use at least two different carrier gases (to allow ionic vapour and TD products to be distinguished) or only study ILs for which vaporization is known to be dominant, e.g. [cation][NTf2]. Hence, for the correlations discussed in §§7 and 8, ΔvapH values measured using TGA (and the transpiration method) for ILs other than [cation][NTf2] ILs have been excluded. This exclusion of ILs such as [cation]Cl reflects the unknown composition of the vapour phase; such ILs are known to have relatively similar ΔvapH and the activation energies of TD, and are known to thermally decompose, even under vacuum using Langmuir vaporization conditions [77,90]. For the correlations by Kabo et al. [140] of literature ΔvapH values with ρ and γ, all TGA data were excluded due to uncertainty over the data analysis methods. However, in this article TGA data were selected based upon the ILs studied and concerns over competition from TD; the analysis methods used were trusted.

The ΔvapH value from the gn>an class="Chemical">roup of Vaghjiani and co-workers for [C4C1Pyrr][NTf2] is excluded from correlations here, as it appears to be a very large outlier [80] relative to the other three published ΔvapH values for the same IL [73,112]. The microcalorimetry ΔvapH values for [CC1Im][NTf2] are very large outliers [142] and were excluded for reasons explained in [129]. A contributing factor may have been the relatively high temperatures at which these measurements were carried out. The TPUV ΔvapH data from Ogura et al. [149] was excluded due to the very large ΔvapH values and the very large errors given.

The ΔvapH values from Ven>an class="Chemical">revkin and co-workers for three protic ILs are excluded [150]. It is expected that the vapour phase for the three ILs studied will mainly comprise ionic vapour [67] and not non-ionic TD product vapour (most likely the products of proton transfer from the cation to the anion), but no evidence is presented to confirm this supposition. For protic ILs, ideally a ΔvapH technique would be used that can distinguish between IL vaporization and TD product vaporization.

Comparing vaporization conditions: Knudsen effusion versus Langmuir

For ILs, both Knudsen effusion [56,86-89,91,119,129-133,135,151,152] and Langmuin>an class="Chemical">r [31,57,63,64,71-77,80,96,106-128] vaporization methods have been used to obtain ΔvapH. For Langmuir (i.e. non-equilibrium) vaporization, an activation barrier to vaporization due to the IL–gas surface structure could exist. Consequently, measurements under Langmuir vaporization conditions would not probe the vaporization energy required to obtain a reliable and accurate ΔvapH value, i.e. vaporization kinetics would be contributing to any measurement.

Any possible effect of kinetin>an class="Chemical">cs on vaporization energetics can be judged by comparing data for techniques that used near-equilibrium and far-from-equilibrium methods, e.g. Knudsen mass loss versus TPMS, respectively. For [CC1Im][NTf2], ΔvapH values published in 2006 and 2007 from Knudsen cell mass loss [56] and from Langmuir vaporization [71] matched extremely well (note that the ΔglCp values used were very similar and thus no mismatch was caused by ΔglCp). Since then, both near-equilibrium and Langmuir methods have confirmed agreement in ΔvapH values for [CC1Im][NTf2] [111,129]. That the values are the same within the relatively small errors strongly suggests that there is no significant kinetic effect for Langmuir vaporization; therefore, there is no significant activation barrier for vaporization at the IL–gas surface. Good matches of ΔvapH values from near-equilibrium and far-from-equilibrium methods have also been published for [CC1Pyrr][NTf2] and [CPy][NTf2] ILs. For ILs with anions other than [NTf2] − there is currently no data to judge kinetic effects on vaporization, as no Knudsen cell (i.e. near-equilibrium) measurements have been made for ILs with anions other than [NTf2] −. For example, for [C4C1Im][PF6] the primary process for near-equilibrium conditions was vaporization of TD products [153], whereas the primary process for Langmuir evaporation was vaporization of intact IL [96]. This difference is due to the higher temperature required for near-equilibrium conditions (compared to Langmuir evaporation) to detect vaporization.

ΔvapH values for approximately 115 ionic liquids

There is now a signifin>an class="Chemical">cant quantity of ΔvapH data available in the literature (approx. 40 papers); there are ΔvapH298 values published for approximately 115 ILs (see electronic supplementary material, table S3). ΔvapH values published in the literature to date have mainly been produced by three groups, who are led principally by Verevkin, Santos and Jones. The Santos group have published only for [cation][NTf2] ILs, whereas the Verevkin and Jones groups have published for [cation][NTf2] ILs and also for a wider variety of anions, e.g. [PF6] − and [BF4] −. Publications from groups other than these three groups have mainly focused on [cation][NTf2] ILs. This concentration on [cation][NTf2] ILs reflects firstly the wish for researchers new to the area to investigate ILs for which ΔvapH values have already been published, secondly the relative thermal stability of [cation][NTf2] ILs (as noted in §4.4.3, and quantified in [31,64]) and thirdly the large number of [cation][NTf2] ILs that are liquid at room temperature.

Measuring ΔdesH

Measuring cedC+A requires Vm and either ΔdesH([C]+) and ΔdesH([A] −) (equation (4.3)) or ΔvapH and ΔdissH(CA) (equation (4.5)). Therefore, there are two potential experimental approaches to obtain cedC+A.

Experimental values for ΔdesH([C]+) and ΔdesH([A] −) can be obtained [87,91]. These measurements have been carried out in the range of T approximately 490 K. To obtain ΔdesH([C]+) and ΔdesH([A] −) at T = 298 K, an extrapolation is required. Therefore, the heat capacity at constant pressure for the isolated ions in the vapour phase will be required (the heat capacity for the liquid phase can be measured, as noted in §4.4.2). At present, these values are not available, although they could be readily calculated.

ΔdissH(CA) has not been measun>an class="Chemical">red experimentally to date. For a [C2C1Im][NTf2] ionic vapour neutral ion pair the energy to produce electronically excited but bound ([C2C1Im][NTf2])* was found to be 5.5 eV, i.e. 531 kJ mol−1 [85]. This value represents an upper limit for ΔdissH(CA) for [C2C1Im][NTf2].

Obtaining the cohesive energy density from simulations and calculations

ΔvapH can be obtained using simulations and n>an class="Gene">calculations. Enthalpies for the ionic vapour neutral ion pair can be calculated using high-level methods. The most widely used approach to obtain the liquid-phase enthalpy per ion pair is from molecular dynamics simulations [26,154-156]; an ion pair in a continuum solvation model has also been used [157,158]. Vm can also readily be obtained from molecular dynamics simulations; hence, cedIP(calc.) can be obtained from simulations and calculations for ILs.

ΔdesH(total) (i.e. cen>an class="Chemical">C+A + 2RT) can be obtained either from a combination of experimental (ΔvapH, table 1) and calculated (ΔdissH298(CA), table 2) [103] results or solely from calculations [104]. ΔdissH(CA) values have been published for many ILs (e.g. [103,196,197]). A combination of simulations and calculations has been used for ILs to obtain ceC+A for a limited selection of ILs ([CC1Im][BF4] and [CC1Im][PF6], where n = 2, 4, 6) [104]; cedC+A were not provided in the article.

Table 1.

ΔvapH298, Vm (at 298 K), Vmol (at 298 K), γ (at 298 K), G (at 298 K) and cedIP,298 data for ILs (for which all of this data could be found in the literature from experimental measurements). For ΔvapH298 and cedIP,298 a constant ΔglCp = −100 J K−1 mol−1 value was used in all cases. ΔvapH298 values were determined for each IL by taking the average (see electronic supplementary material, table S4) of all reliable individual ΔvapH298 literature values (how these individual ΔvapH298 values were chosen is explained in §4.4.3). All literature references used to obtain ΔvapH298, Vm, Vmol (liquid density, ρ, for both Vm and Vmol) and γ are given. .

ionic liquid	ΔvapH298kJ mol−1	ref. for ΔvapH298	Vmcm−3 mol−1	Vmolnm−3	ref. for ρ	γmN−1 m−1	ref. for γ	G J cm−3	cedIP,298J cm−3
[C1C1Im][NTf2]	132	[111,132]	241	0.400	[55]	36.3	[55]	49.3	539
[C2C1Im][NTf2]	136	[56,57,71,86,106,108,111,117,118,129]	257	0.427	[159]	36.9	[160]	49.1	519
[C3C1Im][NTf2]	131	[111,129]	275	0.456	[161]	34.9	[160]	45.3	469
[C4C1Im][NTf2]	136	[56,71,88,108,111,113,118,129]	292	0.486	[55]	33.6	[160]	42.8	457
[C5C1Im][NTf2]	138	[111,129]	309	0.513	[161]	32.9	[160]	41.1	439
[C6C1Im][NTf2]	142	[56,71,108,111,129]	328	0.545	[55]	32.3	[160]	39.6	425
[C7C1Im][NTf2]	143	[111,129]	341	0.567	[162]	32.0	[160]	38.7	412
[C8C1Im][NTf2]	149	[56,71,108,111,129]	363	0.603	[55]	31.9	[160]	37.8	403
[C10C1Im][NTf2]	152	[108,111,129]	397	0.659	[55]	32.1	[160]	36.9	378
[C12C1Im][NTf2]	156	[111,129]	427	0.710	[55]	32.3	[163]	36.2	358
[C3C1C1Im][NTf2]	142	[86,108,152]	288	0.478	[164]	41.0	[165]	52.4	485
[C4C1C1Im][NTf2]	145	[121,152]	306	0.509	[166]	37.4	[167]	46.9	466
[C3(C1Im)2][NTf2]2	183	[94]	476	0.791	[168]	44.7	[168]	48.3	378
[Me(EG)1C1Im][NTf2]	130	[115]	280	0.465	[169]	35.6	[169]	45.9	455
[Me(EG)2C1Im][NTf2]	136	[115]	320	0.531	[55]	36.5	[55]	45.1	419
[C2C2Im][NTf2]	126	[130]	275	0.457	[170]	35.6	[163]	46.2	450
[C3C3Im][NTf2]	131	[130]	310	0.515	[170]	32.4	[163]	40.4	416
[C4C4Im][NTf2]	134	[130]	344	0.571	[170]	31.1	[163]	37.5	383
[C5C5Im][NTf2]	142	[130]	379	0.629	[170]	30.1	[163]	35.1	368
[C6C6Im][NTf2]	157	[130]	412	0.685	[170]	29.5	[163]	33.5	373
[C7C7Im][NTf2]	150	[133]	445	0.740	[170]	29.3	[163]	32.4	332
[C8C8Im][NTf2]	146	[133]	480	0.798	[170]	29.2	[163]	31.5	298
[C9C9Im][NTf2]	150	[133]	515	0.855	[170]	29.5	[163]	31.1	286
[C10C10Im][NTf2]	146	[133]	547	0.908	[170]	29.6	[163]	30.6	262
[C3C2Im][NTf2]	132	[132]	292	0.485	[171]	33.4	[171]	42.5	445
[C3C1Pyrr][NTf2]	147	[112]	284	0.472	[172]	32.5	[167]	41.8	510
[C4C1Pyrr][NTf2]	150	[73,112]	300	0.498	[172]	34.9	[167]	44.1	490
[C6C1Pyrr][NTf2]	152	[73,112]	341	0.567	[173]	31.7	[173]	38.3	438
[C10C1Pyrr][NTf2]	161	[112]	405	0.673	[173]	31.4	[173]	35.8	391
[C2Py][NTf2]	138	[124,131]	253	0.420	[174]	37.4	[174]	50.0	536
[C4Py][NTf2]	143	[118,124,131]	286	0.476	[174]	33.4	[174]	42.8	491
[C5Py][NTf2]	144	[124]	303	0.503	[174]	32.5	[174]	40.9	469
[C6Py][NTf2]	151	[73,124]	320	0.532	[174]	31.7	[174]	39.1	463
[C2C1Im][NPf2]	136	[108,118]	308	0.512	[175]	33.9	[175]	42.4	432
[C4C1Im][NPf2]	137	104, 114	343	0.570	[175]	31.7	[175]	38.2	392
[C6C1Im][NPf2]	139	[108]	378	0.627	[175]	30.3	[175]	35.4	361
[C8C1Im][NPf2]	145	[108]	412	0.685	[55]	27.7	[55]	31.4	346
[C2C1Im][BF4]	143	[73,123]	155	0.257	[175]	54.4	[175]	85.5	905
[C4C1Im][BF4]	154	[75]	188	0.313	[175]	46.9	[175]	69.1	806
[C8C1Im][BF4]	163	[71]	257	0.427	[55]	30.8	[55]	40.9	626
[P6,6,6,14][BF4]	199	[75]	605	1.005	[176]	28.3	[176]	28.2	325
[C4Py][BF4]	161	[73,127]	184	0.305	[177]	46.6	[177]	69.2	860
[3C11C4Py][BF4]	153	[127]	200	0.333	[178]	44.8	[178]	64.7	752
[4C11C4Py][BF4]	152	[127]	200	0.333	[178]	45.5	[178]	65.6	746
[C2C1Im][C2SO4]	155	[57,123]	191	0.317	[179]	47.0	[179]	68.9	800
[C2C1Im][C8SO4]	172	[123]	292	0.485	[180]	31.0	[180]	39.5	580
[C4C1Im][C8SO4]	182	[75]	328	0.544	[181]	26.7	[181]	32.7	548
[C2C1Im][TfO]	138	[123]	188	0.312	[182]	41.3	[183]	60.9	721
[C8C1Im][TfO]	152	[71]	302	0.501	[55]	28.5	[55]	35.9	495
[C4C1Im][PF6]	150	[96]	207	0.345	[173]	44.0	[184]	62.8	710
[C6C1Im][PF6]	153	[96]	241	0.401	[185]	38.2	[184]	51.8	624
[C8C1Im][PF6]	164	[71,96]	276	0.458	[55]	32.5	[55]	42.2	585
[C2C1Im][SCN]	153	[73,123]	152	0.252	[186]	57.8	[186]	91.5	993
[C8C1Im][N(CN)2]	163	[73]	256	0.426	[187]	36.9	[188]	49.0	627
[C4C1Pyrr][N(CN)2]	162	[72]	198	0.329	[189]	56.4	[189]	81.7	806
[C2C1Im][C(CN)3]	139	[122]	186	0.308	[190]	50.4	[191]	74.6	733
[C4C1Im][C(CN)3]	143	[122]	218	0.363	[190]	49.2	[191]	69.0	643
[C2C1Im][B(CN)4]	136	[123]	217	0.361	[192]	48.7	[191]	68.4	613
[C4C1Im][FeCl4]	171	[73]	247	0.410	[193]	46.0	[193]	62.0	684
[C2C1Im][FAP]	126	[123]	342	0.568	[194]	32.5	[194]	39.2	361
[C6C1Im][FAP]	144	[75]	395	0.656	[194]	31.6	[194]	36.4	359
[C4C1Pyrr][FAP]	153	[73]	370	0.614	[195]	38.0	[165]	44.7	408
[C8C1Im]Cl	167	[77]	229	0.380	[55]	30.9	[55]	42.7	718
[C8C1Im]I	167	[77]	247	0.410	[55]	32.7	[55]	44.0	667

Table 2.

ΔvapH298 (constant ΔglCp) and cedIP,298 (constant ΔglCp) are taken from table 1. ΔdissH490(CA) (exp.) = ΔdesH490(total) − ΔvapH490. ΔdissH298(CA) (calc.) are taken from Preiss et al. [103]. cedC+A(exp. + calc.) have been rounded to three significant figures.

ionic liquid	ΔvapH298(exp.) kJ mol−1	cedIP,298(exp.) J cm−3	ΔdissH298(CA) (calc.)kJ mol−1	ΔdesH298 (total) (calc.) kJ mol−1	cedC+A,298 (exp. + calc.) J cm−3
[C2C1Im][NTf2]	136	519	344	480	1850
[C4C1Im][NTf2]	136	457	351	487	1650
[C2C1Im][BF4]	143	905	372	515	3290
[C4C1Im][BF4]	154	806	372	526	2770
[C4C1Im][PF6]	150	710	353	503	2400
[C2C1Im][C2SO4]	155	800	385	540	2800
[C2C1Im][SCN]	153	993	371	524	3420
[C2C1Im][C(CN)3]	139	733	340	479	2550

ΔvapH298, Vm (at 298 K), Vmol (at 298 K), γ (at 298 K), G (at 298 K) and cedIP,298 data fon>an class="Chemical">r ILs (for which all of this data could be found in the literature from experimental measurements). For ΔvapH298 and cedIP,298 a constant ΔglCp = −100 J K−1 mol−1 value was used in all cases. ΔvapH298 values were determined for each IL by taking the average (see electronic supplementary material, table S4) of all reliable individual ΔvapH298 literature values (how these individual ΔvapH298 values were chosen is explained in §4.4.3). All literature references used to obtain ΔvapH298, Vm, Vmol (liquid density, ρ, for both Vm and Vmol) and γ are given. .

As explained in §3.3, δH (and therefon>an class="Chemical">re cedML) can be separated, using indirect methods, into non-polar, polar and hydrogen bonding contributions [15]. For ILs, at least two different methods have been used to obtain contributions to cedIP (or contributions to ΔvapH, from which contributions to cedIP can be obtained), and these methods are explained in the next two paragraphs. Methods to separate the different contributions to intermolecular interactions, and which contributions get grouped together, can vary widely between different researchers.

Electn>an class="Chemical">rostatic contributions are generally those from Coulomb's Law; vdW interactions are usually considered as all other intermolecular interactions (therefore, induction, dispersion and hydrogen bonding are included) [155,196]. ΔvapH can be separated into electrostatic and vdW contributions [155,156]. To obtain the electrostatic and vdW contributions to cedIP, the electrostatic and vdW contributions to ΔvapH simply need to be divided by Vm. To obtain the electrostatic and vdW contributions to cedC+A, the electrostatic and vdW contributions to ΔdesH(total) (or ΔvapH + ΔdissH(CA)) need to be divided by Vm. Such information is available in the literature [155,196].

A second method to separate cedIP labels contributions as polar (similar to electrostatic) and non-polar (similar to vdW) [26]. These two contributions are determined by the user defining atoms in the IL that are polar and atoms that are non-polar. Interaction energies for these atoms are then summed to obtain the polar and non-polar contributions to ceIP (when there is an interaction between a polar atom and a non-polar atom, the energy is shared equally). The polar and non-polar contributions to cedIP are then obtained by dividing by the partial liquid volumes for the polar and non-polar atoms, respectively.

Results for ΔvapH, ΔdesH([C]+) and ΔdesH([A] −) for ionic liquids

ΔvapH for ILs an>an class="Chemical">re much larger than ΔvapH for MLs, as expected. ΔvapH298 values for the ILs investigated to date range from approximately 130 kJ mol−1 for [CC1Im][NTf2] (n = 1 to 3) to 199 kJ mol−1 for [P6,6,6,14][BF4] (table 1 and electronic supplementary material, table S4). For molecular liquids, ΔvapH298 ranges from 30 kJ mol−1 for hexane to 102 kJ mol−1 for triethanolamine and 105 kJ mol−1 for squalane; for water ΔvapH298 = 44 kJ mol−1 [32,33,198].

ΔvapH from experiments: trends for ionic liquids

For ILs, Δvapn>H values inn>an class="Chemical">crease as n is increased (either on the cation or the anion). This conclusion has been demonstrated for [CC1Im][NTf2] (figure 2b) [56,71,108,111,129], [CC1Im][NPf2] [108,118], [CC1Pyrr][NTf2] [73,112], [CPy][NTf2] [124], [CCPy][NTf2] [134], [CC1Im][PF6] [96], and [C2C1Im][CSO4] [57,123]. A broad conclusion for ΔvapH values is that [NTf2] −-based ILs exhibit lower ΔvapH values than for ILs with anions such as [BF4] − or [N(CN)2] − (when the cation is the same, table 1 and electronic supplementary material, table S4). For ILs with the same anion and different cation cores, there is relatively limited data available. A broad conclusion is that, where n is the same, [CC1Im][NTf2] gave lower ΔvapH298 values than [CC1Pyrr][NTf2] and [CPy][NTf2] (where n = 4 or n = 6) [56,71,73,108,111,112,124,129].

Figure 2.

Variation of properties with respect to n for [CC1Im][NTf2]: (a) molar volume, Vm, (b) enthalpy of vaporization at 298 K, ΔvapH298, (c) surface tension, γ, (d) cohesive energy density with an ion pair as the molecular unit, cedIP and (e) Gordon parameter, G. The data used and references are given in table 1. The ΔvapH298 values (and, therefore, the cedIP values) were obtained using a constant ΔglCp = −100 J K−1 mol−1; it has been discussed elsewhere how using such an approach can lead to non-linearity in ΔvapH298 values for [CC1Im][NTf2] at small n, but when variable ΔglCp values are used the relationship is closer to linear with respect to n [111].

Variation of pn>an class="Chemical">roperties with respect to n for [CC1Im][NTf2]: (a) molar volume, Vm, (b) enthalpy of vaporization at 298 K, ΔvapH298, (c) surface tension, γ, (d) cohesive energy density with an ion pair as the molecular unit, cedIP and (e) Gordon parameter, G. The data used and references are given in table 1. The ΔvapH298 values (and, therefore, the cedIP values) were obtained using a constant ΔglCp = −100 J K−1 mol−1; it has been discussed elsewhere how using such an approach can lead to non-linearity in ΔvapH298 values for [CC1Im][NTf2] at small n, but when variable ΔglCp values are used the relationship is closer to linear with respect to n [111].

ΔVapH from simulations and calculations

There is a vast amount of Δvapn>H(n>an class="Gene">calc.) data available in the literature, far too much to give a complete summary here. Therefore, the focus here will be on key examples. The most widely used force field for ILs is the CL&P force field [199]. The aim of the developers of this force field was to provide a model that could describe a large range of ILs; hence, only liquid-phase densities and structural data were used for parametrization [199]. ΔvapH298(calc.) using the CL&P force field are in the range of 150 to 250 kJ mol−1 [23,156,199]. These ΔvapH(calc.) are larger than ΔvapH(exp.), which are in the range of 130 to 180 kJ mol−1 for these ILs. This finding is not surprising, given the relatively low levels of parametrization used to produce the CL&P force field. However, the trend of ΔvapH values increasing as n increases is found, and [C4C1Im][NTf2] giving a small ΔvapH298(calc.) relative to other [C4C1Im][A] ILs, suggests that correct trends in intermolecular interactions are captured. Köddermann et al. [154,155] parametrized a force field against experimental data for [C2C1Im][NTf2], including structural, dynamic and thermodynamic data (liquid-phase density, self-diffusion coefficients of cations and anions and NMR rotational correlation times for cations). ΔvapH298(calc.) for [CC1Im][NTf2] (n = 2, 4, 6, 8) using this force field matched very well to ΔvapH298(exp.) [56,71,154], demonstrating that force fields can be produced for ILs that do an excellent job of capturing the intermolecular interactions. Schröder and Coutinho used the continuum solvation model COSMO–RS for a structurally diverse set of ILs to obtain ΔvapH298(calc.) [158]. ΔvapH298(calc.) matched very well to ΔvapH298(exp.) (±10 kJ mol−1), demonstrating that the continuum solvation model captured the bulk liquid very well.

ΔdesH([C]+), ΔdesH([A] −) and ΔdissH(CA) from experiments, simulations and calculations

ΔdesH([C]+) and ΔdesH([A] −) have been expn>en>an class="Chemical">rimentally measured for [C2C1Im][NTf2] only. Therefore, cedC+A can only be obtained from experimental values for [C2C1Im][NTf2]. ΔdesH490([C]+) = 206 kJ mol−1 and ΔdesH490([A] −) = 207 kJ mol−1 at T ∼ 490 K. Therefore, ΔdesH490([C]+) +ΔdesH490([A] −) =ΔdesH490(total) = 413 kJ mol−1. Using ΔdesH490(total) (exp.) = 413 kJ mol−1 and ΔvapH490(exp.) = 117 kJ mol−1 (constant ΔglCp) an experimentally derived value of ΔdissH490(CA) (exp.) = 296 kJ mol−1 can be obtained for the first time.

For [C2C1Im][NTf2] ΔdesH490(total) (exp.) = 413 kJ mol−1 (table 3) is approximately 3½ times larger than ΔvapH490(exp.) = 117 kJ mol−1 (constant ΔglCp = −100 J K−1 mol−1). It takes 3½ times more energy to vaporize two isolated ions than to vaporize a neutral ion pair.

Table 3.

ΔvapH490 was extrapolated using constant ΔglCp = −100 J K−1 mol−1. ΔdesH490([C]+) and ΔdesH490([A] −) are taken from Dunaev et al. [91]. ΔdissH490(CA) (exp.) = ΔdesH490(total) − ΔvapH490. cedC+A,490(exp.) has been rounded to three significant figures.

ionic liquid	ΔvapH490 (exp.) kJ mol−1	ΔdesH490([C]+) (exp.) kJ mol−1	ΔdesH490([A] −) (exp.) kJ mol−1	ΔdesH490(total) (exp.) kJ mol−1	ΔdissH490(CA) (exp.) kJ mol−1	cedC+A,490 (exp.) J cm−3
[C2C1Im][NTf2]	117	207	206	413	296	1580

ΔvapH490 was extrapn>olated using constant ΔglCp = −100 J K−1 mol−1. ΔdesH490([C]+) and ΔdesH490([A] −) are taken from Dunaev et al. [91]. ΔdissH490(CA) (exp.) = ΔdesH490(total) − ΔvapH490. cedC+A,490(exp.) has been rounded to three significant figures.

Calculated ΔdissH298(CA) for [CC1Im][A] ILs gives 340 to 430 kJ mol−1 at 298 K (table 2) [103]. ΔdissH(CA) varies very little with increasing n for the high-level calculations in [103,196], whereas for the lower-level calculations in [104] ΔdissH(CA) decreases with increasing n (which is a very surprising finding, given one would expect both the electrostatic and vdW interactions for a neutral ion pair to be relatively unaffected by alkyl chain length). A combination of these ΔdissH298(CA) (calc.) values from [103] and ΔvapH298(exp.) gave 480 kJ mol−1 < ΔdesH298(total) (calc.) < 540 kJ mol−1 (table 2).

Separating ΔvapH, ΔdissH(CA), ΔdesH([C]+) and ΔdesH([A] −) into electrostatic and vdW contributions

The electn>an class="Chemical">rostatic and vdW contributions to ΔvapH obtained from molecular dynamics simulations of the liquid phase vary greatly for different ILs (and often give overestimates of the experimental ΔvapH values, §5.2). The electrostatic contributions range from 60 to 200 kJ mol−1, depending on the IL; the vdW contributions range from 70 to 210 kJ mol−1 [156]. For example, for [C8C1Im][NTf2] the electrostatic and vdW contributions to ΔvapH are 73 and 114 kJ mol−1 respectively; in contrast, for [C4C1Im][BF4] the electrostatic and vdW contributions to ΔvapH are 117 and 64 kJ mol−1, respectively [156].

Using moleculan>an class="Chemical">r dynamics simulations for [CC1Im][NTf2] it has been found that as n increases the electrostatic contribution stays approximately constant while the vdW contribution increases significantly [155,156]; the same has been found for [CC1Im][NO3] and [CC1Im][PF6] [26,156]. Therefore, the increase in ΔvapH with increasing n is due to extra vdW contributions. For the IL [P6,6,6,14][NTf2], which contains 28 CH2 groups and 4 CH3 groups, the electrostatic and vdW contributions to ΔvapH are 67 and 200 kJ mol−1, respectively, demonstrating that vdW contributions can dominate for certain ILs [156].

For [C4C1Im][A] ILs the vdW contributions are all in the range of 60 to 100 kJ mol−1, whereas the electrostatic contributions have a far larger range, from 80 to 210 kJ mol−1. [C4C1Im][NTf2] has a far smaller electrostatic contribution than [C4C1Im][BF4]. These findings demonstrate that [C4C1Im][BF4] has a larger ΔvapH than [C4C1Im][NTf2] due to stronger electrostatic contributions.

Electn>an class="Chemical">rostatic intermolecular interaction contributions dominate ΔdissH(CA), with vdW contributions near zero, for the ILs studied ([CC1Im][A] and [CC1Pyrr][A] with a wide range of anions) [155,196]. For example, for [C2C1Im][NTf2] the electrostatic contribution to ΔdissH(CA) is actually larger than ΔdissH(CA) itself, i.e. the vdW interactions are stronger in the isolated [C]+ and isolated [A] − than in the CA neutral ion pair [196]. Given that ΔdissH(CA) are always significantly larger than ΔvapH, this finding shows that electrostatic contributions dominate ΔdesH.

Cohesive energy densities: results

As noted in §4.4.5, there an>an class="Chemical">re many ΔvapH values available in the literature for ILs. Unsurprisingly, some of these ΔvapH values have been used to obtain cedIP, e.g. in [200-203]. A key difference is that in this article the ΔvapH data used is rigorously selected; only ΔvapH values obtained directly from heating experiments are included, and where doubts exist over the ΔvapH values they are not included (as outlined in §4.4.3).

Trends in cedIP for ionic liquids

cedIP,298 values range from 262 J cm−3 for [C10C10Im][NTf2] to 9932 J cm−3 for [C2C1Im][SCN] (table 1). The range for ΔvapH298 is approximately 130 kJ mol−1 to approximately 200 kJ mol−1. The range for Vm is approximately 150 cm3 mol−1 to approximately 600 cm3 mol−1. The range for cedIP,298 is clearly larger than for ΔvapH298, caused by the larger range for Vm over ΔvapH298 (table 1). A plot of cedIP,298 against (figure 3a) shows a very good linear correlation (R2 = 0.92). Marcus recently found a similar correlation between cedIP (using ΔvapH both from experimental and indirect measurements) and the ionic volume (which is proportional to Vm) [202]. This finding demonstrates conclusively that the ΔvapH298 values are not the key factor in determining cedIP for ILs; the key factor is Vm, i.e. IL size. Large ILs have small cedIP, and small ILs have large cedIP; size matters. cedIP can now be predicted with good reliability from calculations alone with no experimental input, as you just need to know Vm (or Vmol), and Vm can be predicted without IL synthesis [195,204].

cedc+A for ionic liquids

Despite ΔdesH([C]+) and ΔdesH([A] −) being experimentally measured for an IL, cedC+A has not been obtained from the experimental data to date, as far as I am aware. For [C2C1Im][NTf2] experimentally derived cedC+A,490 = 1580 J cm−3 (rounded to three significant figures) at approximately 490 K.

cedC+A,298 obtained from a combination of experimental (ΔvapH298, table 1) and calculated (ΔdissH298(CA)) results gives approximately 1600 J cm−3 < cedC+A,298 < approximately 3400 J cm−3. cedC+A,298 obtained solely from calculations for [CC1Im][BF4] and [CC1Im][PF6] gave approximately 2000 J cm−3 to approximately 3200 J cm−3 [104].

cedC+A,298 obtained fn>an class="Chemical">rom a combination of experimental (ΔvapH298, table 1) and calculated (ΔdissH298(CA)) [103] results gives approximately 1850 J cm−3 for [C2C1Im][NTf2]. Given the potential for errors from both experiments (the signals to measure ΔdesH([C]+) and ΔdesH([A] −) in particular are very small and this measurement has only been made for one IL to date) and calculations (they have not been validated against experimental data) the match between the experimentally derived cedC+A,490 = 1580 J cm−3 and the value above cedC+A,298 = 1850 J cm−3 is good (and the T values are also different). This match gives confidence that the cedC+A,298 values derived from a combination of experimental and calculated data give a reasonable measure of the IL total intermolecular interactions; certainly comparisons between two cedC+A,298 values can be trusted when the difference between the two cedC+A,298 values is large.

Separating ced into electrostatic and vdW contributions

There is currently little data in the literature from MD simulations for electrostatic and vdW contributions to cedIP. Data needed to obtain electrostatic and vdW contributions to cedIP is available [155,156], but the simple calculations have not been carried out; such data are presented in table 4. As explained in §5.2, the ΔvapH298(calc) values in [156], and, therefore, cedIP, are overestimates of the ΔvapH298(exp.) values; it is expected that the vdW contributions are too large [156]. The electrostatic contributions to cedIP,298 are between 90 and 620 J cm−3, whereas the vdW contributions to cedIP,298 are between 160 and 350 J cm−3.

Table 4.

Separating ced into electrostatic and vdW contributions using a combination of simulations and calculations, along with experimental molar volume, Vm. The references given are for the simulation/calculation literature.

ionic liquid	ΔvapU (vdW) kJ mol−1	ΔvapU (elec) kJ mol−1	Vm (exp.) cm3 mol−1	cedIP (exp.) J cm−3	cedIP (vdW) J cm−3	cedIP (elec) J cm−3	δH,IP J cm−3	δH,IP (vdW) J cm−3	δH,IP (elec) J cm−3	ref.
[C2C1Im][NTf2]	42	82	258	481	162	318	21.9	12.7	17.8	[155]
[C4C1Im][NTf2]	50	78	292	438	171	268	20.9	13.1	16.4	[155]
[C6C1Im][NTf2]	60	78	328	421	182	237	20.5	13.5	15.4	[155]
[C8C1Im][NTf2]	71	79	363	413	195	217	20.3	14.0	14.7	[155]
[C2C1Im][NTf2]	89	81	258	659	345	314	25.7	18.6	17.7	[156]
[C4C1Im][NTf2]	99	79	292	610	339	270	24.7	18.4	16.4	[156]
[C6C1Im][NTf2]	108	76	328	561	329	232	23.7	18.1	15.2	[156]
[C8C1Im][NTf2]	114	73	363	515	314	201	22.7	17.7	14.2	[156]
[P6,6,6,14][NTf2]	200	67	728	367	275	92	19.2	16.6	9.6	[156]
[C4C1Im][PF6]	68	116	207	889	329	560	29.8	18.1	23.7	[156]
[C4C1Im][BF4]	64	117	188	963	340	622	31.0	18.4	24.9	[156]

For [CC1Im][NO3] (n = 2, 4, 6, 8) the polar contributions to cedIP,298 are between 1300 and 1400 J cm−3 for all four ILs, whereas the non-polar contribution to cedIP,298 are between 290 and 400 J cm−3 for all four ILs [26]. These values are not comparable to the values given in the previous paragraph, as their origins are very different (see §4.5).

Using data from [155] and [196] fon>an class="Chemical">r [C2C1Im][NTf2], ΔdesH298(elec) = 82 + 358 = 440 kJ mol−1 and ΔdesH298(vdW) = 42 – 16 = 26 kJ mol−1. Therefore, the electrostatic contribution to cedC+A,298 for [C2C1Im][NTf2] is 1710 J cm−3, whereas the vdW contribution to cedC+A,298 is 100 J cm−3. The electrostatic contribution dominates the total intermolecular interaction for [C2C1Im][NTf2]. Such dominance is expected to hold for other ILs; compared to [C2C1Im][NTf2], the vdW contribution is expected to be even less important for [C2C1Im][SCN], but more important for [C8C1Im][NTf2].

Can cohesive energy densities be used to understand other properties?

A key question to answer is: an>an class="Chemical">re either of the two measures of IL intermolecular interactions, cedIP and cedC+A, useful for understanding IL liquid-phase properties? The only way to answer that question is through obtaining data and developing trends using cedIP and cedC+A. At this stage only cedIP can be used to understand other properties; there are insufficient experimental cedC+A values to make meaningful comparisons and develop trends.

Gordon parameter, G

A range of methods have been used to measun>an class="Chemical">re γ for ILs [205]. There is a large variation in γ values for the same IL across different publications [205]. This occurrence is most likely due to impurities present in the sample. Many IL γ values are measured under conditions where water and other volatile impurities can contaminate the sample; in addition, grease-type impurities have been identified at IL–gas surfaces [206]. These contaminants are likely to affect γ.

ILs have γ values in the range of 26 to 60 mN m−1 [181,205,207,208]. Molecular liquids have γ values in the range of 18 to 72 mN m−1 [209-211]. Therefore, ILs and molecular liquids have γ values of the same magnitude (unlike ΔvapH298). For ILs, γ values decrease as n is increased [205]. It is more difficult to draw conclusions on the effect of the anion on IL γ values, due to both anion complexity and variations in γ values across different publications. However, a broad conclusion is that [NTf2] −-based ILs exhibit lower γ values than for ILs with anions such as [BF4] − or [N(CN)2] − (for the same cation) [205].

γ and ΔvapH are not linean>an class="Chemical">rly correlated. This observation is most readily highlighted by considering the effect of increasing the alkyl chain length, n; ΔvapH increased and surface tension decreased (e.g. for [CC1Im][NTf2], figure 2b,c). Most importantly, this observation clearly demonstrates that γ and ΔvapH should not be used for quantfying the strength of intermolecular interactions of ILs; comparisons need to be made for values in the same units.

G values range fn>an class="Chemical">rom 28.2 J cm−3 for [P6,6,6,14][BF4] to 91.5 J cm−3 for [C2C1Im][SCN] (for those ILs for which experimental ΔvapH values are also available, table 1). G for molecular liquids range from 29 J cm−3 for squalane, 48 J cm−3 for ethanol, 128 J cm−3 for glycerol and 231 J cm−3 for water (electronic supplementary material, table S5).

There is a good linean>an class="Chemical">r correlation between cedIP and G for ILs (figure 3b). Marcus recently found a similar correlation between cedIP (using ΔvapH both from experimental and indirect measurements) and G [212]. This observation strongly suggests that cedIP and G both capture the same intermolecular interactions for ILs. G shows a good linear correlation with the inverse of size, , i.e. a small IL has a large G value (electronic supplementary material, figure S2). This is as expected, given the correlations given previously, and reiterates that the strength of intermolecular interactions in ILs can be obtained, to a good degree of accuracy, from their size.

There is a good linear correlation between ced and G when both molecular and ILs are considered (figure 4). This relationship strongly suggests that there is an underlying relationship between ced and G that is not dominated by the ionic nature of ILs.

Figure 4.

ced (including both cedML and cedIP) versus G (including G for ILs and G for a selection of molecular liquids).

Although a major aim of this an>an class="Chemical">rticle is to find correlations that hold across all ILs, it is important to compare cedIP and GIL for select IL families/categories. Categorizing the ILs in figure 3 into different families is potentially arbitrary. ILs with a large anion [75,195] ([NTf2] −, [NPf2] − and [FAP] −) are grouped together here. It must be noted that [C4SO4] − and [C8SO4] − are relatively large, but this is caused by the large alkyl chain, and hence these anions are classified here as small (the reasons for this classification will become clear shortly). ILs with small anions are categorized here as all those that do not contain [NTf2] −, [NPf2] − or [FAP] −.

For ILs with a large anion and any length alkyl chain both cedIP and GIL were relatively small (figure 3c), as all data points are located in the bottom left of figure 3c. In addition, all of these ILs with large anions match relatively well to the line of best fit for all ILs, first presented in figure 3b. Within that family, ILs with long alkyl chains (defined here as having at least one alkyl chain of length C6H13, i.e. [C6C1Im][NTf2], although ILs with two intermediate alkyl chains were also included in this category, i.e. [C4C4Im][NTf2]) are coloured black (figure 3c), and ILs with short alkyl chains are coloured purple (figure 3c). Clearly, ILs with long alkyl chains gave smaller cedIP and GIL than those with long short alkyl chains.

ILs with small anions and short alkyl n>an class="Chemical">chains (plotted as green points on figure 3c) gave the largest cedIP and GIL. Therefore, to obtain an IL with the combination of both a large cedIP and a large GIL a short alkyl chain and a small anion is required. ILs with small anions and long alkyl chains (plotted as red points on figure 3c) gave relatively large cedIP but smaller than expected GIL, based upon their cedIP. While conclusions based on individual data points should be treated with caution, the presence of a cluster of red data points (figure 3c) suggests that the relationship for ILs with small anions and long alkyl chains is different to other ILs reported here. One possible explanation for this observation for ILs with small anions and long alkyl chains is the influence of the IL–gas surface structure on GIL. There is a large amount of literature showing that ILs tend to orient with their longer alkyl chains, on average, at the outer IL–gas surface, with the rest of the IL, on average, located just below the outer IL–gas surface [213,214]. Intermolecular interactions for alkyl chains are weaker than those for the other parts of the ILs (e.g. ion–ion electrostatic intermolecular interactions). Therefore, having more alkyl chains located near the outer IL–gas surface than expected based on the stoichiometry will lead to lower GIL but will have no effect on cedIP. At this stage, there is insufficient evidence to draw this conclusion with high confidence, but there is a growing amount of evidence to support it.

Ionic liquids as solvents for self-assembly

In 2013 a summary was pn>an class="Chemical">resented of aprotic ILs that support amphiphile self-assembly; at least 11 aprotic ILs were found, in addition to at least 37 protic ILs [40]. ILs that promote self-assembly for which cedIP and G have also been measured are listed in table 5. Clearly, the rule [13,14] given by Evans of self-assembly only occurring in solvents with G > 110 J cm−3 does not hold. The lower limit from this data would appear to be GIL > 42.8 J cm−3 and cedIP > 457 J cm−3. This would suggest that any IL with GIL and cedIP greater than these values would be able to support self-assembly. ILs that meet these criteria include [C6C1Im][PF6] and [C8C1Im][N(CN)2]. Certain ILs meet the cedIP rule but not the G rule, e.g. [C8C1Im]Cl and [C4C1Im][C8SO4]. This lack of agreement between cedIP and GIL may well be due to the influence of the IL–gas surface structure, as outlined in §7.1. Therefore, cedIP would appear not to be the best guide for determining self-assembly in ILs; GIL is a much better guide for determining whether an IL can support self-assembly.

Table 5.

Aprotic ILs which are known to support amphiphile self-assembly [40] and for which cedIP and G have also been measured.

ionic liquid	cedIP,298 / J cm−3	GIL/J cm−3
[C4C1Im][NTf2]	457	42.8
[C2C1Im][NTf2]	519	49.1
[C4C1Im][PF6]	710	62.8
[C2C1Im][C2SO4]	800	68.9
[C4C1Im][BF4]	806	69.1
[C4C1Pyrr][N(CN)2]	806	81.7

δH,indiren>an class="Chemical">ct for ILs have been obtained from activity coefficients at infinite dilution, giving 19 J1/2 cm−3/2 < δH,indirect < 30 J1/2 cm−3/2 [50,200,201,215-218] and, therefore, 350 J cm−3 < cedindirect < 900 J cm−3. At present, comparisons of cedIP and cedindirect obtained from measurements of δH for ILs are difficult, as there is insufficient data to draw conclusions across a sufficiently large dataset. Marcus has compared δH,IP for ILs (obtained from cedIP data) with the solubility of liquid organic solutes in ILs [203]. The cedIP data (and therefore, δH,IP for ILs) were obtained from both experimental ΔvapH values and ΔvapH from less reliable indirect/calculation methods. However, as outlined in §6.1, there is an excellent correlation of cedIP and , so less reliable ΔvapH values are unlikely to have a large effect on cedIP values. Therefore, the findings in this paper are expected to hold if only experimental ΔvapH values were used to obtain cedIP. Overall, no satisfactory correlation was found, demonstrating that the theories that underpin the use of δH for predicting solubilities do not hold for ILs, at least for organic solutes.

Hansen solubility parameten>an class="Chemical">rs have been obtained for a range of ILs using indirect methods [219-222]. These values could potentially be compared to δH,IP(vdW) and δH,IP(elec) (obtained from cedIP(vdW) and cedIP(elec) respectively), examples of which are given in table 4, even though the categories are different. However, at this stage there is insufficient data both in terms of Hansen solubility parameters from indirect measurement and cedIP(vdW) and cedIP(elec) to make satisfactory comparisons.

ced values fon>an class="Chemical">r ILs have been estimated from viscosity measurements of ILs and IL-based solutions [166,221,223-227]. These give values in the range 600 J cm−3 < cedindirect < 900 J cm−3. At present, as for data from measurements of solubility, there is an insufficiently large dataset to draw definitive conclusions on any correlations with cedIP. Very recently, a linear correlation has been presented between ce and Ea,vis for three protic ILs [228]. C = 5 was found, for ceIP = C · Ea,vis. The ΔvapH values used to obtain this correlation were not included in correlations in this article, for reasons outlined in §4.4.3. It is assumed that, at temperatures at which the vaporization measurements were made, the unit of viscous flow in the liquid phase was ion pairs. In such a situation, using data based on ΔvapH and not ΔdesH(total) makes sense.

Conclusions of cedIP versus cedindirect

Indiren>an class="Chemical">ct methods of obtaining cedindirect for ILs (i.e. intrinsic viscosity or solubility data) have so far proved unsuccessful in matching cedIP data. There are a number of possible reasons for this finding: insufficient/poor quality data (ILs are known for containing impurities), entropy may be vital (especially for solubility), ILs are large and only need to break select intermolecular interactions for processes such as solvation or new surface formation to occur, and the importance of non-polar interactions in ILs (non-polar interactions are known to cause problems for δH [15,49]). These last reasons would suggest that an approach similar to the Hansen solubility parameter approach might prove useful, if reliable data can be obtained.

Insights from cohesive energy densities of ionic liquids

Ionic liquids have very large total intermolecular interactions

As explained in §2.2 and §4.3, cedML and n>an class="Chemical">cedC+A (but not cedIP) represent the total intermolecular interactions, i.e. the total cohesion in a unit volume. cedML,298 values range from 195 J cm−3 for squalane, through 385 J cm−3 for acetone, 679 J cm−3 for ethanol and finishing with 2293 J cm−3 for water [32]. For [C2C1Im][NTf2] experimentally derived cedC+A = 1576 J cm−3 at approximately 490 K. Clearly, [C2C1Im][NTf2] has stronger intermolecular interactions than most molecular liquids. Furthermore, based upon computationally derived cedC+A values, some ILs have stronger intermolecular interactions than water. However, ILs certainly have significantly weaker intermolecular interactions than classical molten salts; Marcus has estimated cedC+A for molten salts ranging from 6241 for CsI to 64 520 J cm−3 for LiF [229]. Based upon the values for both cedIP and cedC+A presented here, ILs have very strong intermolecular interactions (many larger than water). Calculations and simulations indicated that these very strong intermolecular interactions are because ILs are ionic (§6.3); overcoming the electrostatic interactions to break charge neutrality costs a great deal of enthalpy per ion pair.

Intermolecular interactions for ions in the liquid phase versus ions in the vapour phase

A very simpn>listin>an class="Chemical">c, but potentially insightful, model can be developed for [C2C1Im][NTf2], given the large amount of data available for this IL. Liquid-phase molecular dynamics simulations for [C2C1Im][NTf2] gives the number of neighbouring cations surrounding an anion as approximately seven [230]. Therefore, ΔdesH490(total) = 413 kJ mol−1 can be viewed as the enthalpy to break 14 cation–anion intermolecular interactions (as seven cation–anion intermolecular interactions need to be broken for the cation and seven for the anion). Consequently, each cation–anion intermolecular interaction is approximately 30 kJ mol−1. ΔdissH490(CA) = 296 kJ mol−1 is the enthalpy to break one cation–anion intermolecular interaction in the vapour phase. Very clearly, the vapour phase CA intermolecular interaction is substantially stronger than the average liquid-phase CA intermolecular interaction. This stronger vapour phase intermolecular interaction is most likely because having just one cation and one anion, with no other electrostatic intermolecular interactions that might weaken this one interaction, would lead to a particularly strong single electrostatic CA intermolecular interaction.

Given the strength of the liquid-phase total intermolecular interactions for ILs, ΔvapH values are actually smaller than one might expect. The reason for this is that vapour phase cation–anion intermolecular interactions are very strong.

Ionic liquids are associated liquids

r(n>an class="Chemical">cedIP) and r(cedC+A) are the ratios between Pint and cedIP and cedC+A, respectively. For ILs, 0.60 < r(cedIP) < 0.85 when cedIP values are used, e.g. for [C2C1Im][NTf2] r(cedIP) = 0.81. Therefore, cedIP is considerably larger than Pint for all ILs [54]. These r(cedIP) values are similar to what are collectively termed associated/tight/stiff liquids, e.g. acetone (r = 0.67), n-hexanol (r = 0.65) [54]. For [C2C1Im][NTf2] cedC+A,490 = 1580 J cm−3 and Pint = 392 J cm−3 [54,231]; therefore, r(cedC+A) = 0.25. This r(cedC+A) value still leads to categorization of [C2C1Im][NTf2] as an associated liquid, but puts [C2C1Im][NTf2] in the region of methanol (r = 0.33) and formamide (r = 0.36); r = 0.07 for water, r = 0.24 ± 0.02 for liquid metals [232] and r < 0.10 for molten salts [232]. Overall, ILs would be classified as having stronger attractive intermolecular interactions than fluorocarbons, but weaker attractive intermolecular interactions than liquid metals, molten salts and water.

The importance of charge neutrality

cedC+A values an>an class="Chemical">re much larger than cedIP values (tables 2 and 3). Vm used to calculate both cedIP and cedC+A are obviously the same, so the very large difference is caused by the intermolecular interactions that are broken for vaporization of neutral ion pairs and isolated ions, respectively.

Values obtained from both intn>an class="Chemical">rinsic viscosity and solubility measurements are in the range of 350 J cm−3 < cedindirect <  J cm−3. These values match the magnitude of the experimental values for cedIP (250 J cm−3 < cedIP < 1000 J cm−3) and are very different from both the experimental value for cedC+A and also the cedC+A values obtained via indirect methods (1550 J cm−3 < cedC+A < 3400 J cm−3). In addition, there is the good linear correlation between ced and G for a dataset including both ILs and molecular liquids, where cedIP is used for ILs. All of this evidence indicates that the intermolecular interactions that are broken for ILs when measuring cedIP are similar to those broken for ILs when measuring cedindirect.

A possible rationalization fon>an class="Chemical">r the above observations centres on maintaining charge neutrality. When breaking intermolecular interactions to either form holes in a liquid or to form a new IL–gas surface, charge neutrality will be maintained, i.e. all ions will always have at least two close-contact neighbours [230], and no ion will be left without a counterion. In essence, in the liquid phase, charge neutrality is always maintained. For IL vaporization the ionic vapour is composed almost exclusively of neutral ion pairs; therefore, charge neutrality is also maintained, as every ion has a counterion present. Therefore, the physical processes that underpin cedindirect, G and cedIP all maintain charge neutrality. However, for ion desorption all intermolecular interactions are broken; therefore, the processes that underpin cedC+A break charge neutrality. Therefore, it is possible that cedIP—rather than cedC+A—is the key value for understanding other properties that are underpinned by the strength of IL intermolecular interactions, even though cedIP is not a measure of the total IL intermolecular interactions.

Why are ionic liquids involatile compared to molecular liquids?

A key question is: why are ILs so involatile n>an class="Chemical">compared to molecular liquids? Is it because ILs have very strong electrostatic intermolecular interactions in the liquid phase? cedIP is a measure of the strength of intermolecular interactions that control IL vaporization. There are both liquid-phase and vapour-phase intermolecular interactions that contribute to ΔvapH and therefore, cedIP (see equation (8.1), which is simply equation (4.4) rearranged).

For ILs, elen>an class="Chemical">ctrostatic intermolecular interactions dominate the total intermolecular interactions in the liquid phase, relative to the vdW intermolecular interactions (i.e. electrostatic intermolecular interactions control ΔdesH, see §6.3). However, contributions of electrostatic intermolecular interactions are relatively large for the vapour phase of ILs (i.e. for ΔdissH). Consequently, the contribution of electrostatic intermolecular interactions to ΔvapH for ILs is not dominant; as the liquid and vapour phase electrostatic contributions are similar in magnitude they subtract (see equation (8.1)) to give a relatively small electrostatic contribution to ΔvapH. Certain ILs have electrostatic contributions to ΔvapH of approximately 60 kJ mol−1 (see §5.4), which is significantly less than ΔvapH for certain molecular liquids, e.g. ΔvapH = 102 kJ mol−1 for triethanolamine [32]. Hence, it can be concluded that electrostatic intermolecular interactions are not the dominant factor controlling the large ΔvapH of ILs (relative to molecular liquids); vdW intermolecular interactions are important. These observations contradict the theory that the low volatility of ILs is due to their ionic nature, i.e. their strong electrostatic intermolecular interactions [24-27].

Having established that vdW intermolecular interactions matter for ILs for controlling ΔvapH, it is important to understand what determines the contribution of these vdW intermolecular interactions. For the intermolecular interaction vdW contributions, size matters. The vdW contributions to ΔvapH for many, if not all, ILs are expected to be significantly larger than ΔvapH for water. [C8C1Im][NTf2] is far larger than water, so for [C8C1Im][NTf2] the vdW contribution to ΔvapH is much larger than for water. The relatively large vdW contributions in ILs are not due to stronger vdW intermolecular interactions; there are simply more such vdW intermolecular interactions for ILs, as ILs are significantly larger than most molecular liquids. It is the same argument as to why certain hydrocarbons, e.g. squalene, have relatively large ΔvapH but small cedML (see §2.3) [33].

Clean>an class="Chemical">rly, ILs are less volatile than molecular liquids. However, ILs can actually be considered more volatile than their liquid-phase intermolecular interactions alone would suggest. The large contribution from the vapour phase intermolecular interactions, determined by ΔdissH(CA), effectively cancels out a large amount of the liquid-phase intermolecular interactions contributing to ΔvapH (see equation (8.1)).

Why are cedIP and Vm−1 linearly correlated?

Both ΔvapH and Vm feature in the n>an class="Chemical">cedIP equation (equation (4.1)). However, there is a very good linear correlation between cedIP and (figure 3a, and no correlation between ΔvapH298 and ), showing that Vm is a far more important variable for cedIP than ΔvapH. The range of ΔvapH298 values for ILs is relatively small, 130 to 200 kJ mol−1, approximately 50% increase from the smallest to the largest value. The range of Vm, 150 to 600 cm3 mol−1 (table 1), approximately 400% increase from the smallest to the largest value. Hence, Vm is a far more important variable in cedIP than ΔvapH298. Vm does not dominate cedML, i.e. ΔvapH298 is an important variable for cedML (electronic supplementary material, table S5) [32]. It is important to unpack why this is the case.

The relatively small van>an class="Chemical">riation in ΔvapH298 for ILs with a large variation in structures reflects that all ILs have considerable electrostatic and vdW intermolecular interaction contributions to ΔvapH298 (§5.4), and that large changes in IL structure, and hence large changes in Vm, do not have a significant impact on these contributions. Additionally, increasing IL size can either increase or decrease ΔvapH. When n is increased (i.e. decreased) from n = 2 to n = 8, ΔvapH298 increases (§5.1). When the anion size is increased (i.e. decreased) from [BF4] − to [NTf2] −, ΔvapH298 decreases (§5.1). So by increasing Vm, it is possible either to increase or decrease ΔvapH298. Hence, a large change in Vm can lead to a small change in ΔvapH298. Therefore, it is understandable why ΔvapH298 and are not correlated, and furthermore, why dominates cedIP.

Outlook: the challenges remaining

Knowledge of cohesive energy density, ced, for ILs has advanced a great deal in recent years. It is no longer acceptable to state that ced cannot be directly experimentally measured for ILs. Significant progress has been made on using ced values for ILs to test theories of intermolecular interactions developed for molecular liquids.

There an>an class="Chemical">re many challenges ahead for experimental measurements of ced for ILs; there are also challenges for calculations and simulations related to ced. Many of these challenges are focused around making vaporization measurements for ILs that do not contain the [NTf2] − anion; IL vaporization using Knudsen effusion apparatus has only been successfully achieved for [cation][NTf2] ILs to date. A summary of areas that could be explored are given in table 6.

Table 6.

Summary of the challenges still remaining for intermolecular interactions of ILs.

experiments/calculations needed	ILs	information that could be accessed about intermolecular interactions of ILs
high sensitivity MS to determine ionic vapour composition at T = 298 K	any ILs	Calculations have suggested that neutral ion pairs are not the most favoured ionic vapour composition at T = 298 K. Such data are needed to have complete confidence in extrapolation of ΔvapH data to T = 298 K.
Knudsen effusion MS to determine ionic vapour composition	any ILs apart from [CnC1Im][NTf2]; ILs expected to have low ΔvapH and high thermal stability would be good candidates (perhaps [C2C1Im][FAP])	To demonstrate that the equilibrium ionic vapour composition of ILs (other than [CnC1Im][NTf2]) is neutral ion pairs. Langmuir vaporization studies for a wide range of ILs strongly suggest that the equilibrium ionic vapour composition for all ILs is neutral ion pairs; however, no experimental has been recorded.
Knudsen effusion ΔvapH	any ILs apart from [CnC1Im][NTf2] for which ΔvapH has already been measured using Langmuir vaporization; ILs expected to have low ΔvapH and high thermal stability would be good candidates (perhaps [C2C1Im][FAP])	To demonstrate that measuring ΔvapH using the non-equilibrium Langmuir gives the same ΔvapH values as the Knudsen effusion method, i.e. that there is no significant kinetic effect for Langmuir vaporization for ILs other than [CnC1Im][NTf2].
ΔvapH at very different temperatures	any ILs apart from [CnC1Im][NTf2]	To obtain ΔglCp directly from experiment for ILs other than [CnC1Im][NTf2]. Such data would allow validation of ΔglCp values from other methods.
ΔvapH	ILs with small anions and long alkyl chains, e.g. [C8C1Im][SCN]	More data for such ILs has the potential to give significant insight into the underlying interactions that determine GIL, and therefore, γ.
γ and ρ	any IL for which ΔvapH has already been measured, but either γ or ρ has not been measured	To provide more cedIP and GIL data, which will allow further insight into intermolecular interactions without the need for further challenging measurements of ΔvapH.
ΔdesH	any ILs	Delta;desH(total) provides a measure of the total intermolecular interaction for an IL; hence, cedC+A can be obtained. Currently ΔdesH(total) data have only been published for [C2C1Im][NTf2]. Having stated in §8.4 that cedIP may be more useful than cedC+A for correlating with other IL properties, it might appear that knowledge of cedC+A is not useful; that is very much not the case. A key question is: are cedIP versus cedC+A linearly correlated? Do they essentially measure the same thing? Does size matter for cedC+A as much as it does for cedIP? How much does counterion affect ΔdesH? Is desorption really enthalpically favourable for larger ion clusters?
calculations of ΔdissH(CA)	any ILs	How does ΔdissH(CA) vary with alkyl chain length? How does ΔdissH(CA) vary with the anion? Do electrostatic intermolecular interactions dominate ΔdissH(CA) for all ILs?
calculations of ΔdesH(total) to get cedC+A	any ILs	cedC+A is very hard to measure experimentally, so insight into cedC+A (and, therefore, the total intermolecular interactions for ILs) is needed.
calculations of vdW and electrostatic contributions to ΔdesH(total) and, therefore, to cedC+A	any ILs	Knowledge of the vdW and electrostatic contributions to IL total intermolecular interaction will give significant insight into the underlying reasons for trends. Do electrostatic intermolecular interactions dominate cedC+A for all ILs?

Conclusion

cedML values have histon>an class="Chemical">rically been very useful for understanding properties of molecular liquids, both as liquids and as solvents. For ILs, a general conclusion would be that many attempts at correlations between cedIP and another property have proven unsuccessful to date. The lack of a correlation in itself can provide useful information about the underlying assumptions involved in those theories, and provide insight into IL properties and intermolecular interactions.

In spite of the general lan>an class="Chemical">ck of success in finding correlations of cedIP to other IL properties, several positive linear correlations have been found. The excellent linear correlation between two variables derived solely from experimental data, cedIP and , is an exciting development for ILs. Clearly, size matters when judging intermolecular interaction strengths for ILs. From this correlation, the ability to make a priori predictions of cedIP without the need for synthesizing or characterizing an IL is a huge step forward in understanding IL intermolecular interactions, both in terms of being able to readily predict cedIP in itself, but also giving the ability to test theories for ILs against other physical property data.

The good linear n>an class="Chemical">correlation between two variables derived solely from experimental data, cedIP and G, is a further exciting development for ILs. This correlation gives significant insight into IL intermolecular interactions. However, further insight could be gained if more data were available for cedIP and G, particularly for ILs with long alkyl chains and small anions.

ILs have very stn>an class="Chemical">rong intermolecular interactions relative to most molecular liquids, as evidenced by the very large cedC+A value, presented here for the first time. Electrostatic intermolecular interactions dominate for ILs, as might be expected for liquids composed solely of ions.

ILs are not involatile solely due to their ionic nature, and therefore, due to strong electrostatic interactions. Their vapour composition—neutral ions pairs—and their relatively large size (giving considerable vdW intermolecular interactions) contribute greatly too.