Nanosegregation and Structuring in the Bulk and at the Surface of Ionic-Liquid Mixtures

Ionic-liquid (IL) mixtures hold great promise, as they allow liquids with a wide range of properties to be formed by mixing two common components, rather than by synthesizing a large array of pure ILs with different chemical structures. In addition, these mixtures can exhibit a range of properties and structural organization that depend on their composition, which opens up new possibilities for the composition-dependent control of IL properties for particular applications. However, the fundamental properties, structure and dynamics of IL mixtures are currently poorly understood, which limits their more widespread application. This paper presents the first comprehensive investigation into the bulk and surface properties of IL mixtures formed from two commonly encountered ILs: 1-ethyl-3methylimidazolium and 1-dodecyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide Page 1 of 45 ACS Paragon Plus Environment The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 2 ([C2mim][Tf2N] and [C12mim][Tf2N]). Physical property measurements (viscosity, conductivity and density) find that these IL mixtures are not well described by simple mixing laws, suggesting that their structure and dynamics are strongly composition-dependent. Small-angle X-ray and neutron scattering (SAXS and SANS) measurements, alongside molecular dynamics (MD) simulations, show that at low mole fractions of [C12mim][Tf2N], the bulk of the IL is composed of small aggregates of [C12mim] + ions in a [C2mim][Tf2N] matrix, which is driven by nano-segregation of the long alkyl chains and the polar parts of the IL. As the proportion of [C12mim][Tf2N] in the mixtures increases, the size and number of aggregates increases until the C12 alkyl chains percolate through the system and a bicontinuous network of polar and non-polar domains is formed. Reactive atom scattering-laserinduced fluorescence (RAS-LIF) experiments, also supported by MD simulations, have been used to probe the surface structure of these mixtures. It is found that the vacuum-IL interface is enriched significantly in C12 alkyl chains, even in mixtures low in the long-chain component. These data show, contrary to previous suggestions, that the [C12mim] + ion is surface active in this binary IL mixture. However, the surface does not become saturated in C12 chains as its proportion in the mixtures increases and remains unsaturated in pure [C12mim][Tf2N].


Introduction
Ionic liquids (ILs) are molten salts that are liquid at relatively low temperatures. They are often defined as salts with melting temperatures below 100 °C, but a great many are liquid at room temperature and below. There has been much interest in ILs in recent years and their unusual combination of properties, such as very low vapor pressures, wide electrochemical windows, good thermal stabilities, and ability to dissolve a wide range of solutes, has led to their use in a range of 3 downside of this flexibility is that there are a huge number of potential combinations of ions that will form ILs and it would be impossible to synthesize and test all of these for every application. Recent studies have shown that it is possible to generate a library of liquids with a range of properties by simply mixing two or more ILs. [33][34] As such, IL mixtures offer the attractive prospect of being able to access a range of properties from a relatively small number of common ILs by mixing, rather than having to synthesize and store large numbers of different ILs. However, while significant progress has been made in recent years in our understanding of pure ILs, [35][36] comparatively little is known about the fundamental properties of IL mixtures and how these relate to their composition.
The structure and dynamics of ILs, both in the bulk and at the gas-liquid interface, are critical to many of their applications. In the bulk, the amphiphilic nature of many IL ions results in nanosegregation of the liquid into polar and non-polar domains, 37 leading to the development of nanostructure that has been studied using a range of experimental and computational approaches. 35 Considerably less is known about the effects of mixing ILs on their nanostructure, or indeed their fundamental physical properties compared to pure ILs. Some reports of physicochemical investigations of IL mixtures have found that they exhibit properties that are quite well described by simple mixing laws, [38][39][40][41][42][43][44][45][46][47][48][49][50][51][52][53] although this is somewhat surprising given that the chemical structures of some of the ions involved are significantly different. However, non-ideal behavior has also been observed in other studies (e.g. of conductivity, [40][41][54][55][56] density, 46, 50, 57-59 viscosity 48,52,[60][61] and phase behavior 62 ) and, interestingly, some IL mixtures that display ideal mixing behavior in some of their physical properties exhibit significantly non-ideal behavior in others. This has led to the suggestion that IL mixtures are best considered as 'double salt ionic liquids' (DSILs) 34 whose properties are defined by their ultimate ionic composition, rather than by the properties of the pure ILs from which they are derived.
In most cases the origins of non-ideal behavior in IL mixtures are not well understood, but factors such as size differences between the cations and/or the different hydrogen-bonding abilities of ions in a mixture have been proposed as important. 34,48 As is so often the case with ILs, general explanations for these phenomena that apply to all of the diverse range of ILs that are known are unlikely to be found. Instead, detailed experimental and theoretical investigation of a range of 4 different types of IL mixtures is important to understand their behavior. 42,[63][64][65][66][67][68][69] One area that has not received significant attention, but is likely to play a key role in dictating the properties of IL mixtures, is the nanostructure that results from aggregation of one IL in another. A small number of studies have investigated the self-organization of ionic amphiphiles (e.g. long-chain imidazolium, pyridinium and piperidinium halide or [BF 4 [77][78][79] ILs. It is also well known that ILs can be used as solvents for the self-organization of non-ionic surfactants. [80][81] What is currently unclear is the extent to which selforganization plays a role in the observed properties of IL mixtures containing ions that are more similar in their chemical structure. To the best of our knowledge there have been no reports of selforganization (i.e. aggregation/micelle formation) in binary IL mixtures (mixtures containing a common anion and two different cations or vice versa).
The structures of IL surfaces have also been studied in some detail and are of particular interest in applications where the gas-liquid interface is important e.g. CCS, multi-phase catalysis and gas separation. A range of approaches have been used to study IL surfaces and some general features and design rules are beginning to emerge, although there is still much to learn in this area. 36,[82][83][84][85][86] surface structure is influenced by variations of both the anion and cation and, within a series of ILs, increasing the alkyl chain length on the cation leads to a surface that is less polar and more saturated with alkyl chains. 87 Moving from large anions (e.g. [Tf 2 N] -) to smaller anions (e.g. [BF 4 ] -) for a given chain length increases the density of alkyl chains at the surface, due to more efficient packing of the ions. 83 However, there have been relatively few investigations into the surface structure of IL mixtures. 67,84,[88][89] Information of the surface structure can be obtained from indirect classical measurements such as surface tension measurements, which we do not consider further here. More direct experimental techniques used include angle-resolved X-ray photoelectron spectroscopy (ARXPS), 90 Rutherford backscattering spectroscopy (RBS), 86, 91 low-energy ion scattering (LEIS) 82 and time-of-flight secondary ion mass spectrometry (ToF-SIMs), 86,92 and the technique which we have pioneered and use here, reactive-atom scattering with laser-induced fluorescence detection (RAS-LIF). 83,[93][94] They have also been investigated by MD simulations. [95][96] There are large discrepancies in the degree of surface enrichment reported, ranging from stoichiometric surface 5 compositions (detected by ARXPS) 90 to the complete absence of some constituting ions at the surface (reported in an LEIS study). 82 It should be noted that these studies were concerned with different IL combinations, which might have distinct mixing behaviors. However, the same IL mixture investigated using two techniques (RBS and ToF-SIMs) showed different surface compositions. 86  known to form a nano-segregated structure in the bulk and display significant surface ordering, interesting bulk and surface phenomena are observed. 87,[93][94] This is the first time that many of these features have been reported, and that a unified discussion of both bulk and surface properties has been presented for binary IL mixtures.

Results
The bulk properties of [C 2 mim] 1-x [C 12 mim] x [Tf 2 N] mixtures (x = 0 to 1) were investigated by measuring physical properties (viscosity, conductivity and density), as well as with SANS and SAXS experiments, and MD simulations. The structure of the vacuum-liquid interface was investigated using reactive-atom scattering probed via laser-induced-fluorescence (RAS-LIF) and MD simulations.
As such, this is a most extensive characterization of bulk and surface properties of these mixtures.

Physical property measurements
The viscosity and conductivity of ILs are known to be quite sensitive to the nature of both the anion and cation. However, in previous work on IL mixtures, composition-dependent viscosities and 7 conductivities have often been found to be quite well described by simple mixing laws. In particular, the Arrhenius or Grunberg-Nissan mixing law: [98][99][100] (where x is the mole fraction of one component of a binary mixture and η, η 1 and η 2 are the viscosity of the mixture, component 1 and component 2, respectively) or the Katti and Chaudhri mixing law: 101 1 and V m,2 are the molar volumes of the mixture and the two components, respectively, and ∆G E is the excess molar Gibbs energy of activation for flow).
as expected for ILs (see ESI for details). However, large deviations from ideal mixing behavior are also seen when using the Grunberg-Nissan mixing law (figure 3). Attempted fitting of the data to the Katti and Chaudhri mixing law showed similar behavior, as the molar volumes of mixtures in this system show only small deviations from ideal mixing behavior with changing composition (vide infra). Conductivity measurements on the [C 2 mim] 1-x [C 12 mim] x [Tf 2 N] mixtures give a similar picture (figure 3, red circles), although the conductivity data deviate less significantly from the Grunberg-Nissan mixing law compared to the viscosity data. Again, there is a significant (within experimental error) difference between the observed and predicted conductivity from 24 mol% [C 12 mim][Tf 2 N] onwards, being greatest at around 50 mol%. In the region where there is a significant difference between the observed and predicted conductivities, the mixtures have lower conductivities than expected from Grunberg-Nissan mixing behavior. This is consistent with the viscosity data, as conductivities are usually found to be approximately inversely proportional to viscosities in ILs. 1,[102][103][104] The density (figure 4, black squares) and molar volumes of the [C 2 mim] 1-x [C 12 mim] x [Tf 2 N] mixture system exhibit smaller deviations from ideal mixing behavior (using a linear mixing law as shown in equation 3) than the viscosity and conductivity. The observed densities are lower than expected based on a linear mixing law and the maximum deviation occurs at around x = 0.5. Excess  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  molar volumes calculated from the density data are somewhat scattered (figure 4, red circles), probably because they represent very small differences and any small errors in the measured densities are magnified, and do not show a particularly clear trend across the composition range. However, there is a tendency for small, positive excess molar volumes in these mixtures.

Small-angle X-ray scattering measurements
SAXS measurements have been shown in a range of studies to give good insight into the nanostructure present in pure ILs. [105][106][107][108][109][110][111][112][113][114][115][116][117][118][119] In order to further probe the bulk structure of [C 2 mim] 1-   Three peaks are often seen in SAXS data for ILs. These can be assigned to i) the characteristic distances between ions of different charge at close-contact (contact peak, CP) at high q (~1.3 Å -1 ); ii) characteristic distances related to the alternation of ions within the polar network of the IL (charge-ordering peak, COP) (~0.8-0.9 Å -1 ); and iii) characteristic separations caused by the segregation of polar and non-polar domains within the IL (polar/non-polar peak, PNPP) at low q (~0.2-0.25 Å -1 ). [120][121] In pure [C 2 mim][Tf 2 N] only the COP is clearly visible, although there is a suggestion of a broad CP, which has been identified by others for this IL. 113,120  N] increases, although its position appears to vary little. These changes will be discussed further below in the context of our complementary neutron scattering results and MD simulations, and compared to data obtained for a series of pure [C n mim][Tf 2 N] ILs for 2 ≤ n ≤

Small-angle neutron scattering measurements
In order to gain additional insight into the structure of the bulk phase, SANS measurements were   The SANS data suggest a picture that is qualitatively similar to that obtained from the SAXS data, in that the PNPP appears in the data between 24 and 32 mol% [C 12 mim][Tf 2 N]. However, a more detailed picture can be obtained from the SANS data by fitting to scattering models; two different models have been employed. An ellipsoidal model describes a system composed of ellipsoidal scattering aggregates with a particular SLD (in this case the approximate SLD of non-  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 13 deuterated alkyl chains, -3.7 × 10 -7 Å -2 ) dispersed within a solvent with a particular SLD (in this case the SLD of the [C 2 mim-d 11 ][Tf 2 N] and cationic head groups of the [C 12 mim] + ions, which depends on the composition of the mixture). The ellipsoids are characterized by axial and equatorial radii, and the volume fraction of the system that they occupy. At low loadings of [C 12 mim][Tf 2 N] (4 and 24 mol%) the ellipsoidal model fits the data quite well (see ESI for fits), although the fit is poorer at 24 mol%. This is potentially because an incipient PNPP at this composition changes the form of the scattering data without being obvious as a peak, due to the low q scattering that is present. The ellipsoidal model does not fit the data at x ≥ 0.32.
As the SANS data for the [C 2 mim-

Molecular dynamics simulations of the bulk
In order to better understand the bulk structure of this system and to aid interpretation of the SANS/SAXS data, molecular dynamics (MD) simulations were performed on [C 2 mim] 1-  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 density configurations, with ions placed randomly in periodic cubic boxes; (ii) typical equilibrations were implemented for more than 1 ns; (iii) multiple re-equilibrations through the use of temperature annealing and switching off and on of the Coulomb interactions were performed; (iv) further simulation runs were used to produce equilibrated systems at the studied temperatures. The equilibrated systems were used in production runs of at least 4 ns at 500 K (see ESI for details).The simulations were performed at 500 K to improve the dynamics and equilibration of the simulated systems. Nonetheless, comparisons between the main structural features of systems simulated at 500K and 300K are possible, as proven by the green and grey lines in figures 6(a) and 6(b), corresponding to pure [C 2 mim][Tf 2 N] and [C 12 mim][Tf 2 N] at the two temperatures (vide infra). The simulations allow structure factor functions, S(q), to be calculated for each IL (shown in figure 7). The structural analysis has focused on the low-q region of the S(q) functions (0.2 < q/Å -1 < 1.8). In that range most ionic liquids feature three peaks, the CP, COP and PNPP, as described above. These can be compared to the SAXS data presented above, with which they are in excellent agreement, giving confidence in the robustness of the computational methodology. In addition, the MD trajectories can be analyzed to provide a more detailed understanding of the aggregation and structural characteristics of the ILs under study (discussed in detail below).  119,126 All S(q) functions were shifted, in proportion to the volume fraction of the liquid occupied by the non-polar alkyl chains, in the y-axis direction in order to improve legibility. Shifting the data in this way also allows the IL mixtures to be compared to the pure ILs (see text for details and also note the position of the colored markers on the right hand side of the middle graph). In both sets of data, the PNPP begins to appear as an incipient peak at x = 0.32, which grows in intensity and shifts to higher q with increasing mole fraction of [C 12  and 500 K (green and grey lines in figures 6b and 6c) shows that the S(q) differences between the two simulation temperatures are small (COPs and CPs become less pronounced at higher temperatures).
Thus, it is possible to make a comparison between the IL mixtures and pure ILs simulated at the two temperatures.
The first composition (4 mol% [C 12 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57 58 59 60 Figure 9: Aggregate volume-to-box volume ratio, R V (n a ), as a function of maximum aggregate lengthto-box-diagonal ratio, R d (n a ) for all chains aggregates in the mixture series. The black guidelines represent the volume-to-maximum-length ratios for spheres, oblate ellipsoids and prolate ellipsoids in order to guide the eye. Figure 9 shows the relationship between the length and volume of the tail aggregates -a measure of their shapes -as their sizes become progressively larger (from the lower left to the upper right corners of the graphs). It shows that the shape of the aggregate is more or less independent of the type of chains it contains, but strongly dependent on its size: small aggregates have a prolate shape that evolves gradually as their size increases to more globular structures, perhaps best viewed as fragments of a non-polar network, which eventually coalesce and percolate throughout the system.

RAS-LIF experiments
As in our previous work on IL surface structure, 83 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 0.08, 0.16, 0.24, 0.52, 0.74, 0.87 and 1 (see ESI for experimental details). RAS-LIF is a very surfacesensitive technique that aims to detect selectively the abstractable hydrogen atoms exposed at the surface of an IL by reaction with gas-phase O( 3 P) atoms. 83,137 This produces OH radicals that scatter back into the gas phase, where they are detected by laser-induced fluorescence (LIF). The translational energy distribution of the incident O( 3 P) atoms is such that a significant fraction of the O( 3 P) atoms are capable of abstracting an H atom from weaker, 2° C-H bonds, but not from stronger C-H bonds. 83,136 Consequently, the detection of CH 2 groups in the alkyl chains is strongly favored over other types of H, including those in the CH 3 groups or attached to the imidazolium ring. Thus, interfacial reactivity observed in a RAS-LIF experiment is a quantitative measure of the accessibility of these CH 2 groups at the surface.
The rotational distributions of the OH products were characterized by recording LIF excitation spectra at a fixed delay between the photolysis and probe laser pulses in order to confirm that the distributions did not vary materially across the range of mixtures. This was indeed the case as described in detail in the ESI. The LIF intensity of a single transition could, therefore, be taken to be representative of the total number density of OH scattered from the surface.
Appearance profiles (the LIF intensity as a function of delay between photolysis and probe lasers) were measured on the most intense line in the spectrum corresponding to the Q 1 (1) transition.
Details of the procedure and the correction for minor background signals are given in the ESI. The resulting profiles for the pure ILs, selected mixtures and the squalane reference are presented in figure   11. Consistent with our previous results, 83   Each profile is an average of three independent sets of ten profiles each as described in the text. Error bars are 1σ standard error of the mean.

Discussion
A detailed molecular insight into the bulk-phase structure of the [C 2 mim] 1-x [C 12 mim] x [Tf 2 N] system has been gained using a range of complementary techniques. This understanding allows a proposal to be made as to why the physical properties of this system change as they do with changing composition, and importantly to speculate on why some properties in this system deviate more strongly from ideal behavior than others.
Four key observations on the bulk properties in particular require explanation and they are considered in the discussion that follows: 1) The onset of significant deviation from ideal bulk physical property behavior occurs at around 24 mol% [C 12 mim][Tf 2 N].
The fact that viscosity is the property that deviates most from ideal behavior in this system could be linked to the reorganization of the longer-range ordering of the non-polar sub-phase that is required in response to a shearing force. This may provide an additional resistance to flow alongside local interionic interactions in the polar network. Recent work has suggested that network structures, albeit based on rather different intermolecular interactions, formed in molten-salt mixtures also contribute to higher than expected viscosities. 143 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 35 The corresponding changes taking place at the surface of the mixtures as the composition is varied are now considered. Contrary to previous reports based on ARXPS measurements, 90  The picture that has emerged from the RAS-LIF measurements, supported by the MD simulations, can be summarized in figure 16, which shows approximately two unit cells from the MD simulations of each mixture slab, viewed from the side (camera perpendicular to surface normal). N]anion and the proximity of cation and anion near the surface means that 'true' saturation will not be achievedi.e. the situation as visualized in Figure 13 is as saturated as this system can become, because the cations cannot pack more closely together than this. In conventional surfactant chemistry, addition of surfactant to water leads to a rapid reduction in the surface tension as the surfactants assemble at the surface, with a sharp change in slope when the critical micelle concentration (cmc) is reached. However, careful analysis and consideration of these IL mixtures, for which the C12 component is much more modestly surface active than a typical surfactant in aqueous solution, shows that the surface only becomes saturated relatively slowly and continues to change with [C 12 mim][Tf 2 N] mole fraction all the way to x = 1.
A cmc is not evident in this system and it is clear that surface saturation per se does not drive aggregation. Given that formation of aggregates in the [C 2 mim] 1-x [C 12 mim] x [Tf 2 N] mixtures is observed by x = 0.24, at which composition the surface is clearly not saturated, it is clear that there is little real analogy between this IL mixture system and conventional aqueous solutions of surfactants.
If the system were to be considered as a [C 12 mim][Tf 2 N] surfactant dissolved in [C 2 mim][Tf 2 N], the data suggest that its surface activity is relatively weak and so it is clear that bulk aggregation in this system is not driven by the absence of available surface sites.
Finally, it is observed ( figure 12) that there is an initial, near stoichiometric, reduction in C12 This is consistent with a primarily entropic, colligative effect, where the C2 chains distribute statistically amongst highly aggregated C12 chains at both the surface and in the bulk.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 larger aggregates of C12 alkyl chains are present in the liquid, which coalesce with increasing x to form a continuous non-polar sub-phase by x = 0.52 and it is proposed that increased chain-chain interactions in these larger aggregates are linked to the observed deviations in viscosities and conductivities. The length scales measured by scattering studies demonstrate that these IL mixtures are structurally distinct from pure ILs with which they share similar non-polar volume fractions. In the systems studied, the length scale of local bilayer structure in the liquids is clearly related to the length of the longest alkyl chain in the IL. However, the appearance of the PNPP in SAXS and SANS studies and the progression from small non-polar alkyl-chain aggregates to a continuous non-polar sub-phase in MD simulations do show some correlation with the non-polar volume fraction of both the pure ILs and mixtures. To sum up, there are some similarities and some differences between the IL mixtures investigated here and related pure ILs, but it is clear that they are a distinct system. The observations reported here are exciting, as they show that a range of structural and physical properties, both in the bulk and at the surface, can be controlled simply by the composition of an IL mixture, rather than by synthesizing a large number of pure ILs. The detailed understanding that this work brings promises to allow the rational selection of specific surface or bulk structure/properties in [C 2 mim] 1-x [C 12 mim] x [Tf 2 N] mixtures when they are being considered for particular applications.

Supporting Information.
Full experimental and computational details are given in the supplementary information.