Ionic Liquid Vapors in Vacuum: Possibility to Derive Anodic Stabilities from DFT and UPS

Ultraviolet photoelectron spectroscopy (UPS) investigations of several gas-phase ionic liquid (IL) ion pairs have been conducted. [EMIM][OTF], [PYR14][OTF], [EMIM][DCA], [PYR14][DCA], [PYR14][TCM], [PYR14][FSI], [PYR14][PF6], [S222][TFSI], [P4441][TFSI], and [EMMIM][TFSI] vapor UPS spectra are presented for the first time. The experimental low-binding-energy cutoff value (highest occupied molecular orbital, HOMO energy) of the ionic liquid ion pairs, which is of great interest, has been measured. Many studies use calculated gas-phase electronic properties to estimate the liquid-phase electrochemical stability. Hybrid density functional theory (DFT) calculations have been used to interpret the experimental data. The gas-phase photoelectron spectra in conjunction with the theoretical calculations are able to verify most HOMO energies and assign them to the cation or anion. The hybrid M06 functional is shown to offer a very good description of the ionic liquid electronic structure. In some cases, the excellent agreement between the UPS spectra and the M06 calculation validates the conformer found and constitutes as a first indirect experimental determination of ionic liquid ion-pair structure. Comparisons with recent theoretical studies are made, and implications for electrochemical applications are discussed. The new data provide a much-needed reference for future ab initio calculations and support the argument that modeling of IL cations and anions separately is incorrect.


INTRODUCTION
Ionic liquids (ILs) are generally defined as molten organic salts with a melting point below 100°C. ILs have attracted interest because of their uncommon physicochemical properties such as low melting temperatures, excellent solvation ability, relatively high thermal stability, low vapor pressure, nonflammability, high electrochemical stability, etc.
One important application for ionic liquids is as an electrolyte in electrochemical double-layer capacitors (EDLCs) or supercapacitors. In that application, the ionic conductivity and the width of the electrochemical stability window (EW) are the most important properties. Although generally the viscosity of ILs is higher and the ion conductivity is lower than in the conventional electrolytes, ILs are nonetheless considered as the ideal working electrolytes for EDLCs because of their large electrochemical windows, excellent thermal stability, and negligible volatility. 1 In supercapacitors, ILs are close to commercial viability. 2 ILs are also very promising for use in Li batteries, as many ILs are intrinsically stable against the solid Li anode and some ILs form stable solid electrolyte interface layers, thus inhibiting the dendrite growth problem that plagues many Li-battery designs. The wide EW of ILs is of critical importance 3 and allows the use of high cathode voltages and enables the design of high-voltage batteries. ILs may also open up alternative battery chemistries in the standard Li-ion batteries, e.g., Limetal or metal-air. 2 It is possible to synthesize a vast number of different ILs as there are a large number of different cations (and anions) available. Each class of cations for ILs has advantages and disadvantages. Imidazolium-and pyrrolidinium-based ILs, in particular, are very promising for EDLCs and Li batteries. The latter have a more charge-localized aliphatic structure and also a higher EW than the delocalized imidazolium-type aromatic cations. 4 Ammonium-and pyrrolidinium-based ILs have outstanding electrochemical stability because these saturated heterocyclic cations have superior resistance toward reduction. 5 Sulfonium-and phosphonium-based cations should also be of interest as they have a high ionization potential and should therefore also be very stable electrochemically.
The electronic structure of a diverse set of ILs was investigated in this study. We focused on the question of how to determine the intrinsic EWs of the ILs and their oxidation stabilities. The intrinsic (oxidation) stability is the electrolyte stability without interaction with the electrode surface or specific interaction, such as hydrogen bonding with other electrolyte components. 6 Ion transfer is neglected, and the electrode is considered to be chemically inert. In this approach, the electrode in contact with the IL is considered as "electron reservoir" and only electron transfer between electrode and the IL is considered. Only when there are no ion dissociation or chemical reactions occurring, this oneelectron redox mechanism becomes the limiting factor of the anodic and cathodic stabilities. 6,7 There are numerous (recent) studies on this topic, and there is a wide range of EW values and different orders of anion stabilities given in the literature. 3,4,6,8−15 A short overview of some common EW calculation methods is presented before the new data are analyzed.
In the simplest approximation, the electrochemical stability window (EW) of ILs is determined by the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energies of the ion pairs they consist of This so-called frontier molecular orbital method was already proposed by Ong et al. 12 and has been used actively ever since. 8 Ilawe et al. also claimed that the HOMO−LUMO gap of the ion pairs is an indicator of IL stability. 10 Asha et al. suggested that the critical factor that decides the EW of pyrrolidinium-based ILs is the HOMO energy of pairing anions. 4 Some studies go even further and suggest that the electrochemical window of ionic liquids can be estimated by the oxidation and reduction potentials of the constituent ions (in vacuum or in some solvation models) 3 The next approximation to the EW was to take the common electronic gap of the cations and anions (i.e., the overlap Here, the energies of the individual cations and anions (in vacuum) are used. 1,12 Using eq 3, Asha et al. found an agreement with the experimental EW values with a maximum deviation of 5.52% and a minimum deviation of 1.19%. 4 Lian et al. even claimed that this calculation procedure has already been validated. 9 The thermodynamic method has also been used for predicting the liquid-phase EWs. However, in practice, it is also based on the calculation of oxidation and reduction energies of free ions and their solvation energies. Liquid environments can be approximated by calculating the ions in effective mediums (polarizable continuum model, PCM approximation or other solvation models) or by generating molecular dynamics (MD) equilibrated snapshots as input for density functional theory (DFT) calculations.
Computational modeling of ILs requires great care. ILs have been described as "room temperature plasma" due to their charged internal structure with highly correlated motion. For example, most of the currently widely used DFT functionals do not describe anions adequately due to the incorrect behavior of the long-range interactions, while classical force fields used for molecular dynamics (MD) simulations do not readily deal with the partial charge transfer between anions and cations of an IL. 2 All of these HOMO and LUMO energy-based EW estimations (eqs 1−3) rely on the so-called ion-pair approximation. The ion-pair approximation assumes that the electronic structure of the ionic liquid is similar to the ion pairs it consists of. The validity of the ion-pair approximation for ILs  In the case of [EMIM] [BF 4 ], the ion-pair approximation seems to hold overall, as the ion-pair calculation is able to describe the overall shape of the liquid-phase ultraviolet photoelectron spectroscopy (UPS) spectrum very well. However, the description of the top of the valence band of the liquid (or HOMO level of the ion pair) is not accurate using DFT and the ion-pair approximation. 21,22 The situation is even worse in [EMIM][B(CN) 4 ], where the ion-pair approximation is quantitatively unable to describe the liquidphase electronic structure. 23 Importantly however, when performing proper bulk calculations of these ILs, DFT is again able to qualitatively reproduce the correct electronic structure. 22,23 These bulk calculations still lack quantitative accuracy and are computationally too expensive to perform using hybrid functionals. This explains the numerous studies on the liquid IL properties that have been performed using the ion-pair approximation.
There are still very limited experimental data about the gasphase properties of IL vapors. The knowledge of the electronic energies and composition of the topmost valence states is vital for understanding of processes that involve the removal of electrons from the IL. 24 Due to the high practical importance of the EW and the very large number of ab initio calculation studies of ionic liquid ion pairs in vacuum, further experimental data on the HOMO (and LUMO) states of ionic liquid ion pairs are needed.
Therefore, we have investigated the electronic structure of this diverse set of ILs with the emphasis on their intrinsic electrochemical stability.
The ILs under investigation in this work and their abbreviations are shown in Tables 1 and S1. We use the simplest notation for the ILs, which is based on the cations and anions: [ Our assumption is that the intrinsic (maximum) anodic limit determined from the ion-pair approximation may give some insights when comparisons between different ILs are made.

EXPERIMENTAL SECTION
The ionic liquids under study were purchased from different manufacturers. Their stated purities are shown in Table 1. Most ILs were transparent, but some were yellowish The UPS measurements of the ILs were carried out at the new FinEstBeAMS beamline of the new MAX-IV 1.5 GeV storage ring (Lund, Sweden). The beamline is equipped with a collimating grazing incidence plane grating monochromator and toroidal focusing mirrors and covers the excitation photon energy range from 4.5 eV to about 1300 eV. 25 The ionic liquids were evaporated from a quartz crucible in an effusion cell (MBE Komponeten). After inserting the IL into the effusion cell, heating of the IL at around 80°C for several hours was performed to remove residual water from the chemical.
UPS measurements were carried out in the 10 −7 mbar pressure range with a liquid-nitrogen-cooled cold trap.
The spectra were obtained using an electron energy analyzer (Scienta R-4000) in the fixed analyzer transmission mode. The measurements were performed with an excitation energy 40− 50 eV and by using a spectrometer pass energy of 20 or 50 eV. Binding energies were calibrated to the H 2 O 1b 1 (12.62 eV) photoelectron line. Due to the high hydrophilicity of some of the ILs and their high water content, water was still evaporating even at elevated temperatures, and traces of water vapor are visible in most spectra. Water contributes the peaks at 12.62 and 13.0 eV. If the water vapor signal was too strong, a molecular H 2 O spectrum was subtracted from the spectra shown in Figure 1  ). Due to the relatively low vapor density of the ionic liquid, some background gases also appear in the spectra, most notably nitrogen, which contributes peaks at 15.6, 16.7, 16.9, 17.15, and 18.75 eV. If the nitrogen signal was too strong, a molecular N 2 spectrum was subtracted from the spectra shown in Figure  1  . All of the above-mentioned gases are due to the thermal decomposition of the IL. The raw spectra are shown in the Supporting Information ( Figures S1 and S2).
The origin of the spurious peaks at binding energies of 6.8, 11.0, and 22.55 eV in the UPS spectrum of [S 222 ][TFSI] is unknown at this time.
In the case of the [PYR 14 ][OTF] spectrum, the tail that extends from the lowest-binding-energy peak in the UPS spectrum (at about 9.0 eV) is ignored (see Figure 1) DFT calculations were performed using Spartan 14 software. The hybrid functionals M06 (includes 27% exact Hartree− Fock exchange) and ωB97X-D (100% Hartree−Fock exchange for long-range electron−electron interactions) were used for DFT calculations. The Gaussian basis set 6-311++G** (basis set with d,p polarization and diffuse functions) was used throughout.
Many different ion-pair conformers were manually constructed. All geometries (conformers) were optimized (relaxed) for the lowest energy. For example, over 40 different conformers were studied for [S 222 ][TFSI]. Similarly to Fogarty et al., the emphasis was to survey a wide range of cation−anion placements. 29 The density of states (DOS)-type spectra shown in Figure 1 were obtained by convoluting the calculated discrete states with a Gaussian function (0.5−0.6 eV full width at half-maximum) under the assumption that the electron emission intensities from each orbital are equal. Zero point energy and vibronic effects are not taken into account in the calculations.
The structures shown in Figure 1 represent the conformers whose calculated electronic structure agrees best with the experimental UPS spectra. The simulated DOS's of the ACS Omega http://pubs.acs.org/journal/acsodf Article predicted conformers are shown next to the experimental UPS spectra.
It is important to point out that neither the experimental nor the theoretical spectra may represent the lowest-energy conformer, since the evaporation temperature was about 500 K. Furthermore, it has been shown recently that the prediction of IL ion-pair structure is complicated and depends on the calculation method. 30 In some cases, a significant dependence of the calculated DOS on the underlying structure of the ion pair (conformer) was found, unlike in the work of Reinmoller et al. 31  The calculated DOS was shifted to align the peaks and make a comparison between the experimental spectra and the simulated DOS easier. A shift of 2.2 eV was used for the DOS's calculated with the M06 functional, which are shown in Figure  1. The ωB97X-D functional needed just a 0.5 eV shift (not shown). Importantly, these shifts are constant from one ion pair to the next and are probably due to some systematic deficiencies in the functionals.

RESULTS
The experimental UPS spectra along with the calculated DOStype spectra are shown in Figure 1. The figure also shows the ion-pair conformer used for the calculation, i.e., the predicted ion-pair structure. The low-energy cutoff values (HOMO energies) are shown in Table 1. The energy scale of our measurements is very similar to that given by Strasser et al. 1820 However, most other studies have energy scales that are shifted to much lower energies. 16 10 They found that the ωB97X-D functional is superior to M06-2X and B3LYP. Lian et al. also recommended this hybrid functional as it is quantitatively accurate for predicting the electronic properties of individual ions in vacuum. 9 In our previous UPS study of TFSI anionbased ILs, the ωB97X-D functional was shown to perform very well, with only a 0.5 eV energy scale shift required. This is not surprising since the ωB97X-D functional is very close to the best functionals in the benchmark studies. 38 Fu et al. showed that B3LYP was accurate in the prediction of the adiabatic ionization potentials of 160 structurally unrelated organic molecules in gas phase. 39 Tian et al. claimed that PCM calculations are superior to their gas-phase calculations performed using the (similar) B2PLYP-D functional. However, we have shown previously that B3LYP is a poor functional for the description of the electronic structure of ILs. 19,23 The hybrid DFT functionals are better than standard GGA DFT functionals for the description of the [ 19 Therefore, both hybrid DFT and MP2 methods should be considered and investigated.
The energies (HOMO, LUMO, orbital, total, etc.) from the DFT calculations are dependent on the functional and basis set. Unfortunately, this makes ab initio calculation of the absolute binding energies difficult, and in many cases, empirical corrections are needed. We have used shifting of the DOS by 2.2 eV and corrections to the calculated M06 HOMO energies by 1 eV. Unfortunately, the Koopmans theorem is not valid within the DFT formalism, so the Δ-SCF method can be used for an improved estimate of the ionization energy. This calculation shows that the Δ-SCF 40 M06 ionization energy is larger than the HOMO energy by about 1.6 eV, thus leaving only an about 0.8 eV shift. The calculation was performed for conformers at 0 K, and the neglect of zero point and thermal (broadening) energy is another reason why the calculated binding energies are not directly comparable to the experimental data.
Next, some ILs will be discussed separately in more detail. 4 ] has already been studied. 21 It was found that the HOMO energy is about 7.4 eV, which seems too low in the context of the present study. For example, [EMIM][DCA] has a HOMO energy of 7.5 eV (see Table 1 41 Later, they calculated a gap of 6.72 eV, again using the HSE06 functional, which is known to provide relatively accurate band gaps. 42 41 We interpret this as a failure of DFT to correctly describe the HOMO of imidazolium-based (aromatic) ILs. The correct cation π-states emerge as the top of the valence band in bulk calculations and in some hybrid ion-pair calculations. This has already been discussed previously. 21 6 ], the HOMO is dominated by the cation states. 12 These observations challenge the prevailing assumption that it is always the anion, which determines the oxidative stability. 12 3.3. TFSI Anion-Based ILs. , etc. This is because their spectra are mostly dominated by the anion. However, the low-energy region around 9−12 eV is still somewhat different in all of the TFSI anion-based ILs. As mentioned above, the low-binding-energy cutoff value of the TFSI anion-based IL vapors is about 8.6−8.7 eV.

[EMIM][BF 4 ]. The vapor-phase photoelectron spectrum of [EMIM][BF
The TFSI anion-based ILs are also the most thermally stable (see Table 1) under our experimental setup, i.e., they can be evaporated in high vacuum with minor thermal degradation. Yoshimura 20 The M06 calculation is able to reproduce these spectra quite well. However, it tends to overestimate the intensity of the peak at a binding energy of 19.5 eV. It mostly yields HOMO and LUMO binding energies of 7.7 and 2.0 eV, respectively. Therefore, most TFSI anion-based IL ion pairs have 5.6−6.3 eV calculated gaps. Using UPS and IPS, Kanai 45 The [P 4441 ][TFSI] ion pair is structurally the largest of the ILs studied in this work. The calculation of this ion pair is also the most challenging due to the large number of conformers it has. The DOS depends significantly on the underlying ion-pair structure. Over 60 different [P 4441 ][TFSI] conformers were evaluated at the M06/6-311++G** level of theory, which is very time-consuming.
It is important to point out a mistake in our recent paper about the TFSI anion-based ILs. Clearly, the DFT calculation of [PYR 14 ][TFSI] has been performed on the nonsaturated pyrrole-based cation not on the saturated pyrrolidinium, which was measured experimentally. 19 The PYR 14  This observation challenges the prevailing assumption that it is always the anion that determines the oxidative stability. 12 3.4. DCA and TCM Anions. The DFT calculation predicts that the HOMO of these DCA and TCM anion-based ILs is due to the π-orbitals of the anion. Furthermore, the top six molecular orbitals of [ 24 Furthermore, using resonant Auger spectroscopy, they were able to show that nitrogen is significantly contributing to the lowest-bindingenergy feature.
The [EMIM][DCA] UPS spectrum is somewhat similar to the spectra of simple aromatic compounds such as benzylazide and methyl benzyl azides. 28 The [EMIM][DCA] photoelectron spectrum also resembles the pure pyridine and methyl pyridine spectra, 27 as expected. For example, the relatively sharp rise in intensity around 12 eV is very similar to pyridine. The double peak around 10 eV is also similar to pyridines, and it seems to be due to the aromatic nature of the compounds similar features exist in the unsaturated cyclopentene, cyclohexene, and cyanobenzene, but not in the saturated cyclopentane and cyclohexane. Therefore, these features can be assigned to the π-orbitals of the cation.
The However, the lowest-binding-energy feature at 8 eV of [EMIM or PYR 14 ][DCA] is missing in all of these analogues. This again validates the claim that it is due to the DCA anion.
The DFT calculation is able to describe the electronic structure of [PYR 14 ][TCM] very well (see Figure 1).
The [EMIM][DCA] double peak (HOMO) around 8.5 eV is also captured quite well, but the next double peak around 10.5 eV is shifted to higher energies by the M06 functional. It can also be seen that the hybrid DFT has some other inaccuracies in the description of the [ 21 However, this problem will be pursued in a future study.
It should also be kept in mind that the experimental [EMIM][DCA] spectrum is the lowest-quality UPS spectra presented in this study. Thermal degradation products could also be present. This is due to the rather high volatility of this IL, which makes vapor-phase studies rather difficult.

DISCUSSION
Asha et al. claimed that "the HOMO is always located in anions and the LUMO is mainly contributed by cations irrespective of the anions". 8 Very recently, they further claimed that the EW is solely decided by the HOMO energy of the pairing anions. 4 This implies that eq 2 is always valid, i.e., the ion pair can be approximated by its constituent ions.
While this seems to hold for most of ILs, in the case of the strong anions like BF 4 − and PF 6 − , the HOMO of the cation is actually at a higher energy; thus, the HOMO of the ion pair may also be determined by the cation. Based on eq 3, Roohi et al. pointed out that if the HOMO energy of the anion is larger than that of the cation, then the anodic stability is determined by the anion. 46 In the case of the very negative HOMOs of fluorine-containing anions like BF 4 , PF 6 , and TFSI, the anodic stability of the IL could effectively be controlled by the cation. 46 Furthermore, Roohi et al. also made the logical conclusion that the EWs of these fluorine-containing ILs can be equal, if the cation is the same. Fogarty et al. pointed out that the fact that the cation HOMO could also be the ion-pair HOMO might be surprising to some researchers. 24 Tian et al. also pointed out that the anodic limit might be determined by oxidation of cations. 15 Therefore, some further discussion about the localization of the HOMO and energy gaps is warranted.
The HOMO of [EMIM][OTF] is mostly localized on the OTF anion, but small contributions from the cation also exist. The HOMO of [PYR 14 ][OTF] is due to the anion, more specifically, the oxygen atoms (possibly its lone pair electrons). The same conclusion was also reached by Fogarty et al. in their recent liquid-phase work. 24 Using X-ray emission spectroscopy, Kanai et al. were also able to show experimentally that the oxygens are heavily contributing to the HOMO. 45 47 Thus, they predict a very small difference between the EWs of these ILs. This seems to be in contrast with experimental EWs and our gas-phase data and could be due to the use of a nonhybrid PBE functional in their calculations.
Using the thermodynamic method, Kazemiabnavi  anions against oxidation is increasing in the following order: TFSI < OTF < BF 4 < PF 6 . 33 However, Asha et al. put the TFSI anion as more stable than the OTF anion and ordered the anions this way: DCA < TFA < OTF < TFSI < BF 4 < PF 6 . 4,8 This in agreement with our data as generally the HOMO energies are increasing with increasing strength of the anion: DCA, OTF, TFSI, FSI, BF 4 , PF 6 . Therefore, we predict that the TFSI anion should be more stable than the OTF anion. Fogarty et al. also found that the TFSI anion-based ILs had larger HOMO binding energies (by about 0.1−0.7 eV) than the ILs with the OTF anion. 24 By calculating isolated anions, Kazemiabnavi et al. claimed that the HOMO energy of TFSI anion is higher than the OTF anion, therefore making the TFSI anion less stable against oxidation than the OTF anion. This is an example of the failure of the isolated ion HOMO/LUMO methods (eqs 2 and 3), and it illustrates that the calculation of the cation and anion separately is not a good approximation to the ion pair. The TFSI anion is relatively large with a high degree of structural mutation and should not be modeled alone.
Another  24 It is also difficult to understand how these rather small EW values have been derived, as ωB97X-D tends to overestimate the HOMO energies by about 1 eV (see Table  1).
Next, a comparison with the experimentally measured EW values should be performed. Unfortunately, it is very difficult to make comparisons with the experimental EWs, as there is a wide range of reported EWs in the scientific literature. The variance is due to the large number of factors affecting the final value, such as sweep rate, current cutoff value, and electrode (Pt, GC, Au, W, Ta) used. 13 The very commonly used glassy carbon (GC) electrodes also limit the EWs.
The  52 Mousavi found that ammonium-based cations were all limited at −3.4 V (vs Ag + /Ag). 53 The EWs of ILs composed of these stable cations/anions can also be high. For example, [PYR 14 ]-   4 ] has the corresponding value of 8.43 V. 54 The anodic limit on GC electrodes seems to be limited to about +2.5 (vs Ag + /Ag). 5,55 [PYR 14 ][TFSI] on gold was shown to have anodic and cathodic stabilities of more than ± 3.5 V (vs Ag/Ag + ). 56 [DEME][BF 4 ] on platinum also showed limits larger than ± 3.5 V (vs Ag/Ag + ). 57 There are conflicting data about the stability of ILs based on the BF 4 anion. In some studies, the TFSI anion seems to have a larger EW than the same cation paired with the BF 4 anion. 55 However, there are also studies where the BF 4 anion shows higher anodic potential than the TFSI anion.
[Butylpyridinium][BF 4 ] on GC has a reported anodic potential limit of +3.65 (vs Ag + /Ag). 49 4 ] has about 0.5 V higher anodic limit than [DEME][TFSI]. 57 They made the correct conclusion that BF 4 is more difficult to oxidize than TFSI. Thus, BF 4 would seem to be at least 0.5 V more stable than TFSI. Indeed, our HOMO level differences imply a 0.65 V difference.
ILs based on the DCA (and TCM) anion have smaller EWs than most other ILs. 58 60 This anodic limit is smaller than [PYR 14 ][TFSI] by about 1.7 V, which is in good agreement with our gas-phase HOMO-level differences.
There is a correlation between the experimental EW, anodic limit, and the gas-phase HOMO energy. In other words, a more negative HOMO energy implies greater anodic stability and a larger EW. However, there is a wide distribution of EW values in the literature and more study on this is needed.

CONCLUSIONS
Gas-phase ion pairs of several ILs were investigated using valence band photoemission. The DFT calculation using the M06 functional was able to reproduce most of the spectral features, and it performs surprisingly well for most IL vapors (see Figure 1). The ωB97X-D functional offered a very similar level of performance at a somewhat higher computational cost. Systematic shifts were needed to bring the calculated DOS into agreement with the experimental UPS spectra. These shifts can probably be corrected with better functionals or better approaches (GW 43  [TCM]), the excellent agreement between the experimental UPS spectrum and the calculated DOS validates the conformer found for the IL ion pair and provides indirect experimental evidence for the structure of ionic liquid vapors. This is due to the sensitivity of the UPS spectrum on the underlying ion-pair structure.
The prevalent assumption that the cation sets the cathodic limit and the anion sets the anodic limit may not be valid for some ILs. 12 When a "weak" cation is paired with a "strong" anion ([EMIM][BF 4 ], for example), the cation can determine the oxidative stability. 50 However, the approximation still holds in most cases.
It can be shown that many recent theoretical estimations of IL EWs are not accurate. This is not only due to the untested validity of the ion-pair approximation used in the calculation. It seems that a calculation method that incorporates true Hartree−Fock exchange (or the GW approach) is a must for a correct ab initio description of IL ion pairs. The new UPS data will help to validate and further develop the numerous works on the electronic structure and electrochemical stability limits of ILs.
The new data also provide strong support to the argument that the modeling of IL cations and anions separately is incorrect and could lead to wrong conclusions.
It is very difficult to make comparisons with the experimental EWs, as there is a wide range of reported EWs in the scientific literature. However, in some cases, a good correlation between the new gas-phase data and the experimental EW is found.
There is no clear direct connection between the electronic gap and the electrochemical gap. The ion-pair approximation always overestimates the true EW, 3 but is generally considered to provide an upper bound for the true stability potential window. 7 It is also computationally less expensive than other methods. Peljo and Girault claim that these kinds of ion-pair approximations to the EW should be discarded completely. 61 However, we will not go so far and believe that further study of the vapor-and liquid-phase electronic structure and also the excited (LUMO) states of the ion pairs of these ILs is necessary to make final conclusions.
At this point, still a limited number (10+) of experimental photoelectron spectra of IL vapors are published. The liquid IL energy gaps determined by the IPS/UPS method are in excellent agreement with our gas-phase HOMO data and the M06 (+1 eV) calculations. Indeed, the UPS/IPS method is a direct probe of the valence and conduction bands. Therefore, this method is recommended for future IL studies.
Raw "as-measured" UPS spectra and details about the ionic liquid samples (PDF) ■ ACKNOWLEDGMENTS