View Online PCCP Dynamic Article Links Cite this: Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 PAPER www.rsc.org/pccp Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A Structural modifications of ionic liquid surfactants for improving the water dispersibility of carbon nanotubes: an experimental and theoretical studyw Antonello Di Crescenzo,a Massimiliano Aschi,*b Elisa Del Canto,c Silvia Giordani,c Davide Demurtasd and Antonella Fontana*a Received 16th January 2011, Accepted 20th April 2011 DOI: 10.1039/c1cp20140a The 1-hexadecyl-3-vinylimidazolium bromide (hvimBr), a water-soluble long-chain imidazolium ionic liquid (IL) with surfactant properties, showed the ability to produce stable homogeneous aqueous dispersions of pristine Single-Walled Carbon Nanotubes (SWNTs). The purpose of this study is the improvement of SWNT dispersing ability by assessing the effect of different groups in position 3 of the imidazole ring. In this regard structural analogues were synthesized and, after characterization, their capability to dissolve SWNTs in water was investigated. Molecular Dynamics (MD) simulations have been performed to provide a semi-quantitative indication of the affinity of each dispersing agent toward SWNT and to attempt an explanation of the experimental results. Introduction The main limitation related to the use of Single-Walled Carbon Nanotubes (SWNTs) is their low solubility in organic and aqueous solvents. In the past decade different strategies have been worked out to overcome the dissolution problem.1 Fukushima et al. have demonstrated that ionic liquids (ILs) based on the imidazolium cation can disentangle SWNT bundles.2 1-Hexadecyl-3-vinylimidazolium bromide (hvimBr), a water-soluble long-chain imidazolium ionic liquid with surfactant properties, showed3 the ability to produce stable homogeneous aqueous dispersions of pristine SWNTs. a Dipartimento di Scienze del Farmaco, Università ‘‘G. d’Annunzio’’, Via dei Vestini, I-66100 Chieti, Italy. E-mail: fontana@unich.it; Fax: +39 0871 3554791; Tel: +39 0871 3554790 b Dipartimento di Chimica, Ingegneria Chimica e Materiali, Università degli Studi di L’Aquila, Via Vetoio (Coppito 1), I-67010 L’Aquila, Italy. E-mail: m.aschi@caspur.it; Tel: +39 338 6546526 c School of Chemistry/Centre for Research on Adaptive Nanostructures and Nanodevices (CRANN), Trinity College Dublin, Dublin 2, Ireland d Interdisciplinary Center for Electron Microscopy (CIME), École Polytechnique Fe´de´rale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland w Electronic supplementary information (ESI) available: Lifetime distribution of hpymim residence in the corresponding free energy minimum; conductometric, fluorimetric and turbidimetric CMC determinations; NIR-PL spectra and efficiency data; Raman spectra; AFM topographic images and height profiles; absorbance spectra; stability of dispersions over the time; NMR spectra of hphimBr and hpymimBr; computational estimation of the solubility ratio; point charges and atom types; thermodynamic integration on hphim and on hpymim in vacuum and in water. See DOI: 10.1039/c1cp20140a This journal is c the Owner Societies 2011 The purpose of the present study is to assess the effect of different moieties in position 3 of the imidazole ring on the dispersing ability of ionic liquid surfactants toward SWNTs, in order to develop new derivatives with ideal features. Many long-chain imidazolium ionic liquids demonstrated to have amphiphilic nature and to be able to form micelle in aqueous solution,2,4–6 thus representing a new class of cationic surfactants with peculiar properties. 1-Hexadecyl-3-phenylimidazolium bromide (hphimBr) and 1-hexadecyl-3-(pyren-1-ylmethyl)imidazolium bromide (hpymimBr) (Chart 1) have been synthesized with the aim of evaluating the effect of the substitution of the vinyl group of hvimBr with a phenyl or a pyrenylmethyl group on the affinity of the molecule for the nanotube surface and, as a consequence, on its dispersing ability. Like common ionic surfactants and polymers, the hydrophobic long alkyl chains of the IL cations should adsorb on the nanotube sidewalls via van der Waals and CH–p interactions, thus minimizing the interfacial energy of the nanotube–water interface,7,8 likely behaving as linear alkyl chains at the groove sites on the outside of the nanotube bundles.9 The hydrophilic cationic imidazolium groups should instead flank toward the aqueous phase, providing water solubility. Chart 1 Investigated ionic liquid surfactants. Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 11373 Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A View Online Aromatic moieties are known to strongly interact with carbon nanotubes. In particular pyrene derivatives have been shown to favor the solubilization of SWNTs10,11 and MWNTs12,13 and to allow the supramolecular attachment onto the nanotube surface of a variety of molecules.14 Indeed p–p stacking interactions between the aromatic core of the adsorbing dispersant and the nanotube surface have been demonstrated15 to be the prevailing interactions between the carbon nanotubes and the adsorbate. Nevertheless charge transfer interactions16 have been observed as well for systems containing donor moieties, such as amino groups. A realistic computational approach of the overall investigated system is still prohibitive because of the huge dimensions which should be taken into account. Therefore the choice of the computational–theoretical setup is strictly related to the information which must be obtained. On one hand, it might be possible to keep the overall configurational complexity by using force fields beyond the atomistic detail, e.g. coarsegrained Molecular Dynamics simulations17,18 which are able, in principle, to provide a full description at a semi-quantitative level. On the other hand, an alternative possibility relies on the use of atomistic force fields which, although restricted to a sub-portion of the system, might be able to furnish more quantitative and general information. In both cases, however, a dynamical approach is necessary. In the present study we have focussed our attention on the relationship between the chemical nature of the investigated ILs and their affinity toward SWNTs in aqueous solution. The initial aim of MD simulations was to provide a simplified although physically coherent picture describing the main determinants influencing the affinity of ILs toward SWNTs. Subsequently, we also estimated the solubility and surfactant activity of the investigated ILs and related such observables to the dispersing ability of the ionic liquids. This procedure is obviously far from being exhaustive for understanding the physico-chemical implications of the phenomena underlying dispersing processes, however we feel that it might provide a useful indication for complementary investigations. Results and discussion MD simulations The main aim of the present simulations is to provide an approximate indication of the affinity of a given solute toward SWNT. In this study we evidently disregard any cooperative effect plausibly induced by the ensemble of the adsorbing solutes. In order to evaluate such an affinity we calculated the free energy variation with respect to the minimum distance of the solute from the SWNT surface (rmin). In this first stage of the study we do not check how the solute interacts but only if it interacts. Due to the simulation ensemble, assuming the system as equilibrated (see below, Fig. 1) the overall Helmholtz free energy change for any transition from a reference minimum distance ref to a generic distance i can be calculated from the probabilities p (obtained by the MD simulation) of finding the system in both states i and ref DAi ¼ RT ln 11374 pi pref Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 ð1Þ Fig. 1 Helmholtz free energy variation with respect to the solute– nanotube minimum distance for the investigated system: hvim (red), hphim (green), hpymim (blue). where R is the ideal gas constant and T is the absolute temperature. Note that the ref distance is the one showing the largest probability and therefore might vary from one system to the other. In Fig. 1 we report the free energy differences for the three systems with respect to the minimum SWNT–solute distance of 0.4 nm systematically taken as zero. The error bars represent the standard error calculated by considering three sub-portions of each trajectory. The choice of the reference distance was driven by the fact that at larger distances the relative free energy, although not undergoing significant changes, was affected by very large errors indicative of lack of convergence. Note that the curves reported in the figure must be considered as only indicative of the relative free energy variations when the solutes approach a SWNT surface. In other words, the actual free energy variation with respect to the two partners at (virtually) mutual infinite distance would require the evaluation of the related semiclassical roto-translational partition functions and is far beyond the aim and the possibilities of the present calculations. Results indicate that: (i) hpymim shows the largest free energy variation, about 22.8  0.8 kJ mol1, with respect to hphim (18.6  0.8 kJ mol1) and hvim (5.4  0.8 kJ mol1). (ii) hpymim shows a minimum at a distance of 0.32 nm sharply larger than hphim (0.27 nm). (iii) hvim shows two minima at 0.29 nm and 0.35 nm with the latter being more stable of about 1 kJ mol1. It might be extremely interesting to provide some sort of explanation of the above results. For this reason we have performed an analysis of the actual conformations adopted by the three solutes when interacting with SWNT, i.e. when they reside in the free energy minima basins (see Fig. 1 and computational details for the definition). At this purpose, for each trajectory, we have extracted all the frames corresponding to conformations falling within the free energy basins reported in Fig. 1. Therefore we obtained four sub-trajectories, one for each of the systems showing only one minimum, i.e. hpymim and hphim, and two for hvim showing two free energy minima. This journal is c the Owner Societies 2011 Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A View Online On each of the above sub-trajectories we carried out Essential Dynamics analysis (ED, see Experimental). Note that for this analysis, beyond the usual solute internal coordinates, we have considered its rotational degrees of freedom with respect to the SWNT. ED analysis produced, for all the systems, a typical behaviour characterized by the presence of two eigenvectors showing highest eigenvalues and representing the directions along which about 80% of the overall fluctuation occur. Therefore we could carry out the conformational analysis using the conformational plane defined by the above eigenvectors in which the trajectory has been projected. The most probable nanotube–solute conformations, adopted within the minimum free energy basins, correspond to the zones of highest probability found onto this conformational plane and indicated in violet. The results for the four minima are reported in Fig. 2. Fig. 2 points to a clear difference between the above systems. In particular hpymim minimum, Fig. 2A, and the second minimum of hvim, Fig. 2D, show rather localized violet spots indicating that they are confined in a very limited region of their conformational space. Conversely, the first minimum of hvim, Fig. 2C, spans a very large portion of the space. This means that adsorbed hpymim, as well as adsorbed hvim at its farther distance to SWNT, are conformationally very hindered. On the other hand, hvim residing in its first minimum maintains a high flexibility. Hphim represents a situation somewhat intermediate between the above cases. In order to characterize the structures of the above minima we show explicitly in Fig. 3 the conformations, indicated as a–g in Fig. 2, extracted by the centre of the violet basins (highest probability conformations). We wish to underline that such structures must be considered as just indicative. As far as hpymim is concerned, we observe two similar conformations, Fig. 3a and b, which essentially differ in the relative rotation between pyrene–imidazolium rings. Both the conformations, however, show hpymim evenly lying onto the SWNT surface with the pyrene ring always parallel to the surface. An analogous behaviour might be observed in the case of hphim, Fig. 3c and d. Also in this case we found the solute always lying on the surface with the two minima differing in Fig. 2 Projections of the sub-trajectory corresponding to the ligand–nanotube free energy basins (Fig. 1) onto the corresponding essential plane (nm) of hpymim (A), hphim (B) and hvim (first free energy minimum, C, and second free energy minimum, D). The high probability conformations a–g correspond to violet spots; red spots indicate zero probability conformations. This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 11375 View Online Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A at larger distance, depicted in Fig. 3g, we observe a tendency of the polar head to prevailingly interact with bulk water molecules. All these data, however, do not help in understanding the actual nature of the different affinity found in Fig. 1. As a matter of fact, apparently there is no correlation between the free energy decrease and the solute–SWNT distance and/or the solute conformations. Probably, a more influencing factor, affecting the stability of the system, exists. In this respect, the previous analysis roughly indicates the larger covering provided by hpymim which, it must be noted, also turns out to be the most strongly bound. Therefore we analyzed the number of water molecules which are swept upon solute adsorption. This analysis was performed by counting the number of contacts between water molecules and SWNT surface not exceeding 0.3 nm. The result, reported in Fig. 4, indicates that hpymim removes from the hydrophobic SWNT surface a significantly larger number of water molecules which, consequently, are free to bind to solvent molecules probably reducing the potential energy of the system and increasing its entropy. The final step of this first section concerns the equilibration of the system. In fact, as already remarked, all the considerations so far illustrated are valid only in the presence of exhaustively equilibrated trajectory. The check of the convergence might be rather complicated in the present case. Therefore we decided to carry out the following test. We measured, only for hpymim, the average lifetime of residence of the solute in one of its minima along the trajectory. The value was calculated by counting the number of times hpymim is found in one of the two deep spots in Fig. 2A. The result, shown in Fig. S1 (ESIw), produces a mono-exponential decay with a mean lifetime of 3.12 ps for the first minimum. An analogous value of 2.08 ps was found for the second minimum (not shown). This result indicates that the solute–SWNT dynamical interactions evolve within timescales much shorter than the simulated period, suggesting that the above observations, at least in the regions close to minimum free energy, rely on a Fig. 3 Representative conformations adopted by the solutes when they interact with SWNT. Each a–g conformation corresponds to the highest probability conformation a–g evidenced as violet spots in Fig. 2. the inter-ring rotation. But, differently from hpymim, in the case of hphim the phenyl ring is not always parallel to the nanotube surface. A situation rather similar to the above structures is found for the first two minima of hvim depicted in Fig. 3e and f (the minima closer to the SWNT surface) as extracted from the first essential plane of hvim. On the other hand, for the minimum 11376 Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 Fig. 4 Number of water molecules adsorbed onto the SWNT surface when interacting with the solutes at different distances: hvim (red), hphim (green), hpymim (blue). Circles indicate the location of the free energy minima for the different ILs (see Fig. 1). This journal is c the Owner Societies 2011 Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A View Online sufficiently relaxed system. We could reach the same conclusion also by considering the relatively low values of the error bars reported in Fig. 1. The above considerations lead to the conclusion that hpymim shows a larger affinity toward SWNT mainly because of its larger hydrophobic contact and its capacity to set free a larger number of water molecules. However all these data imply that the three solutes are thermodynamically stable in water with respect to their own aggregation, i.e. their solubility is high enough. Quantitative evaluation of solubility using MD simulations is a tremendously complicated task which might be addressed only indirectly. At this purpose, we adopted the computational strategy described in the Computational details and ESIw (Appendix S1). Results demonstrate that, within the approximations adopted, hpymim shows a hydration free energy of 27 kJ mol1 larger than that of hphim, thus indicating that the former IL is about 25 000 times less soluble in water than the latter. Analogously, they might indicate a higher tendency of hpymim to self-associate into micelles (see infra). A partial explanation of this result might be obtained from Fig. 5 in which we report the hydrophobic area of the two solutes. The result shows that hpymim is characterized by a much larger hydrophobic surface which may prevent its dissolution in aqueous solution and/or favour its self-assembly. In conclusion MD simulations indicate that: -the presence of a highly hydrophobic group may increase the affinity of the ionic liquid toward nanotube both for enthalpic (hydrophobic–hydrophobic interactions) as well as entropic (solvent molecules desolvation from the SWNT surface) effects. -the effect of an aromatic group, such as the phenyl, and an aromatic polar head, such as the imidazolium ion, on the dispersing ability of carbon nanotubes may be comparable. -a too hydrophobic domain may suppress the solubility of the molecule in water thus preventing its ability to disperse the nanotube in water. This evidence is in perfect agreement with the necessity of good dispersing agents of carbon nanotubes (CNTs) to have a well defined balance between the solvophilic Fig. 5 Normalized distribution of hydrophobic surfaces calculated along the trajectories for hvim (red), hphim (green) and hpymim (blue). This journal is c the Owner Societies 2011 and the solvophobic domains of the molecule, i.e. block copolymer19 and pyrene functionalized styrenic copolymers10 with too solvophobic domains are not very good dispersing agents of CNTs notwithstanding their high affinity for the nanotube sidewalls. In contrast, for a too hydrophobic molecule self-assembly into micelles may be favoured over the adsorption onto the nanotube surface. Therefore, for SWNT dispersing purpose, the best ionic liquid should show both good affinity for water and good affinity for the nanotube surface. Characterization of the ionic liquid surfactants The investigated ionic liquid surfactants have been demonstrated to be relatively soluble in water. A solubility of 148 mM at 39 1C, 138 mM at 41 1C and 7 mM at 65 1C has been determined for hvimBr, hphimBr and hpymimBr, respectively. The temperature chosen for ILs chemico-physical characterization is the lowest that ensured a complete solubility of the investigated ILs and that we considered therefore the most informative. It is worth noticing that, once ILs have been solubilized, the obtained micellar solutions were stable at room temperature. As expected, it turns out that the substitution of the vinyl group with groups with increasing weight and lipophilicity decreases, even if not dramatically, the solubility of the surfactants. These measurements are in agreement with the 27 kJ mol1 higher hydration free energy calculated for hpymim with respect to hphim (see above). All of the three ILs form micelles (see as an example, Fig. 6) in aqueous solution. The critical micelle concentration (CMC) values, determined by following different techniques, are 364  25 mM and 232  67 mM for hphimBr and hpymimBr, respectively. In particular CMC values obtained by using the conductometric method (see Fig. S2 and S3 in the ESIw) were slightly higher than those obtained by using spectrofluorimetric (see Fig. S4 and S5 in the ESIw) or turbidimetric measurements (see Fig. S6 in the ESIw), but today it ‘‘is generally recognized that this discrepancy originates from the fact that micellization is not an abrupt phase transition. The association process sets in over a finite concentration range, and to characterize this range by a single number, such as CMC, reflects a somewhat arbitrary decision.’’20 For example, the fluorimetric determination with Fig. 6 Cryo-TEM micrograph of micelles (see black arrows) of hpymimBr. Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 11377 Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A View Online Nile Red may be favoured in the presence of the external fluorescent probe, while pre-aggregation phenomena may interfere with turbidimetric determinations. As far as a comparison among the CMC data of the three different surfactants is concerned, it is difficult to draw a conclusion as the determinations were performed at different temperatures due to solubility problems and an increase of CMC is expected upon increasing the temperature.21 Nevertheless, an appreciable decrease of the CMC has been detected on passing from a phenyl to a pyrenylmethyl substituent in agreement with the very well known22 dependence of CMC on the hydrophobicity of the surfactants and the above determined higher hydration free energy for hpymim. The similar CMC values of hphimBr and hvimBr [330  60 mM]3 could be ascribed to a compensation effect between the CMC decrease expected for a more hydrophobic surfactant and the CMC increase expected for measurements performed at a higher temperature. SWNT dispersing ability of ionic liquid surfactants All of the three investigated ionic liquids are good dispersants for SWNTs at room temperature as confirmed by the structured emission in the NIR region of their dispersions, depicted in Fig. 7 and Fig. S7 in the ESI.w By using the experimental Kataura plot23 to assign the individual optical frequency bands to specific nanotubes, hvimBr and hphimBr seem to favor with respect to sodium dodecyl benzensulfonate (SDBS) the dispersion of small diameter semiconducting SWNTs, corresponding to bands in the wavelength range 900–1000 nm. From average NIR emission efficiency hpymimBr appears to be the worst dispersing agent of semiconducting SWNTs while SDBS appears to be the best among the dispersants considered. Because both the bundling of SWNTs7 and the surrounding medium24 have demonstrated to cause spectral shifts and decrease of PL yield, Raman spectra have also been recorded. Fig. 8 reports the radial breathing mode bands (RBM) of the recorded Raman spectra (full range traces in Fig. S8 in the ESIw) and shows decreased relative intensities of the bands associated with the larger diameter tubes resonant in the RBM region o220 cm1. This strongly indicates that both hvimBr and overall hphimBr bring in solution prevailingly small diameter SWNTs, thus confirming the data obtained by NIR-PL. The similarity between the ID/IG ratios of raw and IL dispersed SWNTs (Fig. 8 in the ESIw) is indicative of the fact that sonication did not strongly affect the nanotube integrity. Atomic force microscopy (AFM) topographic images (Fig. S9 in the ESIw) confirm that all of the investigated IL dispersed samples are composed of debundled nanotubes. The optical properties of the investigated dispersions were also monitored by vis/NIR absorbance measurements (see Fig. 9 and Fig. S10 in the ESIw). All of the spectra showed wellresolved van Hove transitions indicating well debundled SWNTs. The normalized absorption to the first-order semiconducting excitations (E11 semiconductor transitions) at 1000 nm of the obtained debundled solutions of SWNTs, reported in Fig. 9, gives insights into interactions between SWNTs and the investigated ILs. As a matter of fact the bathochromic shifts (i.e., B2 nm) of SWNTs dispersed with hphimBr and 11378 Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 Fig. 7 NIR-PL spectra of SWNT solubilised with the investigated ILs and, for comparison, with SDBS ([SWNTs]i = 0.2 mg mL1, [dispersant] = 1 mM, sonication time 5 h) (lexc = 683 nm). The histogram below reports the average NIR-PL efficiency of SWNTs dispersed with the investigated dispersants. The data are expressed as average (n = 3)  standard deviation. hpymimBr suggest11 mutually interacting p-systems. These shifts are in agreement with that observed, particularly for hpymimBr, in the NIR-PL spectrum. All the above mentioned evidences point to the fact that the van der Waals interactions among alkyl chains and the nanotube surface reinforce synergistically the prospective selectivity that had been proven27 for geometrically constrained polyaromatic molecules thus favouring the capacity of the phenyl ring to operate a selection of nanotubes. Only hpymimBr did not evidence any selection, probably due to its low water solubility (see previous section) and therefore reduced ability to solubilize nanotubes as evidenced also by absorbance data (see Fig. S10 in the ESIw). From absorption spectra performed by keeping constant the initial amount of SWNTs and the concentration of the ionic liquid (Fig. S10 in the ESIw and Fig. 10) it appears that hphimBr is the best dispersant among the three investigated surfactants being capable to solubilize almost twice as much SWNTs as hvimBr at the concentration of 1 mM. This evidence is in conflict with NIR-PL efficiency data (see Fig. 7), nevertheless, as it has been reminded above, NIR-PL efficiency data may be altered by PL yield changes due to the surrounding medium.24 In particular, a common and almost linear increase of dispersing ability has been evidenced on increasing the This journal is c the Owner Societies 2011 View Online Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A Fig. 9 Normalized absorption spectra of SWNT solutions obtained with the three investigated ILs. The concentrations of ILs have been adjusted to allow the solubilisation of ca. the same amount of SWNT. Spectra are shifted vertically of 0.025 a.u. in order to highlight comparisons and differences. S11 and S22 refer to E11 and E22 semiconductor transitions.7,25,26 Fig. 10 Percentage of SWNTs dispersed with the investigated ILs (real absorbance/theoretical absorbance determined at l 377 nm by taking advantage of the calibration curve constructed by Di Crescenzo et al. with e377 nm = 106.0 mL mg1 cm1). Fig. 8 RBM bands in the Raman spectra for raw HiPCO SWNTs (A) and for SWNTs dispersed in 1 mM hvimBr (B), hphimBr (C) or hpymimBr (D). Assignment is based on experimental Kataura plots. Raman spectra of the dispersants are reported in blue. concentration of IL from 0.2 to 1.0 mM surfactant (see Fig. 10), therefore above the CMC values of the corresponding ILs. This increase of solubilized SWNTs appears thus correlated with an increase of adsorbed IL, thus pointing to the fact that: (i) the formation of micelles in this IL concentration range, when SWNTs are present in solution, is highly unlikely, and This journal is c the Owner Societies 2011 (ii) both p–p and van der Waals interactions (i.e. of the aromatic rings or the double bond and of the alkyl chains of the investigated ILs, respectively) are involved in the adsorption. In this same IL concentration interval, hpymimBr appears almost as effective as hphimBr, in perfect agreement with the above considerations that lead to the conclusion that the larger affinity of hpymim toward SWNT with respect to the other ILs is prevailingly due to its larger hydrophobic contact (see Fig. 5) and its ability to set free a larger number of water molecules (see Fig. 4). On the other hand, above the concentration of 1.0 mM the three ILs behave differently. HvimBr demonstrated to increase its dispersing ability in a linear relation to its concentration, passing from a percentage of CNT dispersion of ca. 45% to 98% on a five-fold increase of its concentration.3 This can be seen as a further adsorption of hvimBr onto a not completely saturated surface of the nanotubes in agreement with the lower hydrophobic surface of hvim with respect to the other investigated ILs (see Fig. 5). The levelling off of the degree of SWNT dispersion in the presence of hphimBr appears instead to be associated with a saturation of the nanotube surface in agreement with a higher dimension of the surface area characterizing this IL compared to hvimBr (see Fig. 5). Vice versa, the ability of hpymimBr to solubilize CNTs decreases sharply on increasing its concentration above 1.2 mM. This dramatic decrease is Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 11379 View Online Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A 1-Hexadecyl-3-phenylimidazolium bromide (hphimBr) Fig. 11 Fluorescence intensity of an aqueous solution of hpymimBr 0.75 mM (obtained by a thousand-fold dilution of a solution of 0.75 mM), and of two aqueous dispersions of SWNTs prepared by using hpymimBr 2 mM (dashed line) and 0.75 mM (dotted line). probably due to the much lower solubility of hpymimBr with respect to the other surfactants or due to the preference of hpymimBr to form micelles rather than adsorbed onto the nanotube surface at concentration of surfactants well above its CMC. Actually, it had been previously demonstrated that the dispersibility of MWNTs solubilized by pyrene-labeled hydroxypropyl cellulose (HPC-Py) in an aqueous medium depends13 on the solubility of the corresponding HPC-Py in water and that micelles are28 the main cause of nanotube aggregation at sodium dodecyl sulfate (SDS) concentration well above its CMC. As a matter of fact, simulations have clearly pointed out the higher hydrophobic area (see Fig. 5) and the higher hydration energy of hpymim with respect to hphim and hvim. The interaction of the pyrene group of hpymimBr with the nanotube surface has been confirmed by the strong fluorescence quenching seen in all the SWNT/hpymimBr hybrids. The quenching of the IM is attributed to intramolecular transduction of singlet excited energy from the high-energy pyrene singlet excited state to the low-energy state of SWNT. The quenching is very strong in the presence of 0.75 mM hpymimBr, as shown in Fig. 11, thus highlighting the almost complete adsorption of hpymimBr to the SWNT. The higher fluorescence observed at 2 mM hpymimBr is indicative of the presence of a high amount of free not adsorbed hpymimBr in the solution (see Fig. 11). The obtained dispersion of SWNTs demonstrated to be stable for months (see Fig. S11 in the ESIw) as had been already proven3 for SWNTs dispersed with hvimBr. Experimental Materials Pristine HiPCO SWNTs were provided by Carbon Nanotechnologies, Inc. Houston, USA. Bromohexadecane, phenyl iodide, imidazole, K3PO4, CuI, L-proline, 1-pyrenemethanol, phosphorus(III) bromide, NaH, magnesium sulfate, NaHCO3, silica gel, Nile Red (9-diethylamino-5H-benzo[a]phenoxazin5-one), pyrene and all organic solvents were purchased from Aldrich and were used without further purification. Ultra pure Milli-Q water (Millipore Corp. model Direct-Q 3) with a resistivity of >18.2 MO cm1 was used to prepare all solutions. 11380 Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 1-Phenylimidazole was synthesized by using Ma’s N-arylation method29 and reaction conditions developed by Deng et al.30 One molar equivalent of 1-phenylimidazole and of 1-bromohexadecane were reacted. The mixture was heated to 65 1C and stirred for 4 h. The bromide salt was then dissolved in water and heated to 50 1C for hot extraction. Ten extractions with ethyl acetate were performed to remove impurities. Eventually, the IL was allowed to crystallize in aqueous solution in order to purify the product further. The IL was then filtered and dried under vacuum. 1H-NMR, d ppm (300 MHz; CDCl3): 0.84 (t, 3H, J = 6.8 Hz), 1.20 to 1.32 (m, 26H), 1.94 (quintet, 2H, J = 7.30 Hz), 4.55 (t, 2H, J = 7.30 Hz), 7.49 to 7.81 (m, 7H), 11.02 (br s, 1H); 13C-NMR, d ppm (75 MHz, CDCl3): 14.41, 22.96, 26.56, 29.36, 29.63, 29.68, 29.81, 29.89, 29.92, 29.96, 30.71, 32.18, 50.80, 120.92, 122.03, 123.24, 130.49, 130.86, 134.69, 136.25 (see Fig. S12 and S13 in the ESIw). 1-Hexadecyl-3-(pyren-1-ylmethyl)imidazolium bromide (hpymimBr) 1-Pyrenemethanol was converted into 1-bromomethylpyrene by reacting 2.5 molar equivalent of 1-pyrenemethanol with 1 molar equivalent of PBr3 (CH2Cl2, 0 1C, 4 h).31 The obtained bromide (0.8 molar equivalent) was reacted with 1 molar equivalent of imidazole in the presence of 1.1 molar equivalent of NaH (THF anhydrous, 0 1C) to obtain 1-(pyren-1-ylmethyl)imidazole.32 The desired hpymimBr was synthesized by reacting 1 molar equivalent of 1-(pyren-1-ylmethyl)imidazole and 1-bromohexadecane. The mixture was heated to 140 1C and stirred for 1 h. The final product was extracted with ethyl acetate. 1H-NMR, d ppm (300 MHz; CDCl3): 0.86 (t, 3H, J = 6.5 Hz), 1.12 to 1.28 (m, 26H), 1.79 (quintet, 2H), 4.18 (t, 2H, J = 7.40 Hz), 6.30 (s, 2H), 7.12 (d, 2H, J = 1.4 Hz), 7.88 to 8.32 (m, 9H), 10.74 (br s, 1H). 13C-NMR, d ppm (75 MHz, CDCl3): 14.44, 22.99, 26.50, 29.20, 29.62, 29.66, 29.73, 29.84, 29.91, 29.96, 30.19, 30.25, 30.42, 32.22, 50.52, 51.55, 121.74, 122.06, 122.09, 124.45, 125.03, 125.40, 126.20, 126.30, 126.67, 127.33, 128.67, 128.75, 129.50, 129.85, 130.60, 131.20, 132.57, 137.30 (see Fig. S14 and S15 in the ESIw). Conductometric determination of the CMC of ionic liquids Conductivity measurements were carried out by using a CDM210 Conductivity Meter (Radiometer, Copenhagen), with a cell constant of 0.82 cm1 calibrated with a KCl solution of known concentration. The temperature was kept constant at 41 1C and 65 1C for hphimBr and hpymimBr, respectively. The investigated concentration ranges were 2.5  102 to 0.8 mM for hphimBr and 2.5  102 to 0.4 mM for hpymimBr, respectively. The solutions were stirred in order to favour equilibration before each measurement. The CMC value was calculated by the Williams method from a plot of specific conductivity versus molar concentration. Fluorimetric determination of the CMC of ionic liquids Steady-state emission spectra were recorded at 41 1C and 65 1C for the hphimBr and hpymimBr, respectively, on a Jasco FP-6200 spectrofluorimeter. The investigated concentration ranges were 4.0  102 to 1 mM for hphimBr and 2.5  102 This journal is c the Owner Societies 2011 Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A View Online to 0.5 mM for hpymimBr, respectively. The temperatures were chosen in order to avoid IL precipitation. The Nile Red was used as the fluorescent probe in the case of hphimBr and its concentration was kept constant (0.5 mM) in all of the measurements. Due to the presence of a pyrene group in the molecule, the CMC of hpymimBr was determined by exploiting the fluorescence of that group. The variation of IE/IM (where IE is the fluorescence emission at 480 nm and IM is the fluorescence emission at 377 nm) as a function of the concentration of hpymimBr in water is reported in Fig. S4 in the ESI.w As expected,33 IE/IM is almost constant at low concentration while it increases when the hpymimBr concentration rises up to 190 mM due to the formation of intermolecular excimers. The value of 190 mM is therefore indicative of the onset of aggregation of hpymimBr molecules. The linearity of the plot above 190 mM highlights the formation of micelles with a proper aggregation number and well defined hydrophobic core in which pyrene pendants can easily encounter. Turbidimetric determination of the CMC of hpymimBr The turbidity was recorded at 65 1C, at l 400 nm by using a Varian Cary 100 Bio spectrophotometer. The wavelength was chosen in order to avoid absorption interference by the substrate. The investigated concentration range of hpymimBr was 2.5  102 to 0.4 mM. Data are reported in Fig. S6 in the ESI.w Preparation of SWNT aqueous suspensions Aqueous dispersions of SWNTs were prepared at 25 1C by adding 5 mL of IL aqueous solutions of different concentrations to 1 mg of pristine SWNTs placed into a glass centrifuge tube. The sample was then sonicated with an ultrasonic bath sonicator (Transsonic 310 Elma, 35 kHz) for several hours. The obtained suspension was centrifuged for 10 min at 4000 RPM by using a Universal 32 (Hettich Zentrifugen) centrifuge in order to separate the supernatant aqueous solution from the precipitate. The former consists of mostly exfoliated or fine bundled SWNTs coated with ILs, while the latter contains essentially non-dispersed SWNTs and other ‘‘impurities’’ characterizing pristine nanotubes such as graphite, amorphous carbon and metal catalysts. The results are normalised on the emission efficiency of SWNTs dispersed with SDBS. Raman determinations Micro-Raman scattering measurements were carried out at room temperature in the backscattering geometry using a Renishaw 1000 micro-Raman system equipped with a CCD camera and a Leica microscope. A 1800 lines mm1 grating was used for all measurements, providing a spectral resolution of B1 cm1. As an excitation source the He–Ne laser with 633 nm excitation with variable power was used. Measurements were taken with 20 s of exposure time and 2 accumulations. The laser spot was focused on the sample surface using a 50 objective with short-focus working distance. Raman spectra were collected on numerous spots of the sample and recorded with a Peltier cooled CCD camera. Only one spectrum was collected per spot. The data were collected and analyzed with Renishaw Wire and GRAMS software. Samples were prepared as follows: an aliquot of each of the three ILs dispersed SWNT solutions analysed by NIR-PL was dropped on a glass cover slip and left overnight to dry in an oven at 60 1C. Raman spectra of raw HiPCO SWNTs and IL surfactants were recorded on the as provided powders. AFM measurements AFM topographic images were collected in semi-contact mode with an NT-MDT inverted configuration system. Silicon tips with reflectance gold coated on the back, tip apex radius of 10 nm, force constant of 2 N m1 and frequency of 170 kHz were used. The data were collected and analyzed with NT-MDT Nova software. Samples were prepared as follows: supernatants of SWNTs dispersed in IL surfactants were centrifuged at 9000 RPM for 90 min in order to separate the material from the solution. Fresh water was added following removal of the supernatant for three times. This procedure was repeated three additional times with fresh DMF. The nanotube material was dispersed in high purity DMF by sonication, spray coated onto freshly cleaved mica substrates and dried overnight in an oven at 60 1C. UV/vis/NIR spectroscopic characterization of SWNT dispersion NIR-PL measurements NIR-PL studies were carried out in triplicate on supernatants in a L.O.T. ORIEL NS1 NanoSpectralyzers (diode lasers; lexc = 638 nm, 683 nm and 785 nm) with an integration time of 1 s and 5 accumulations. NIR-PL efficiency data represent the spectrally integrated emission values adjusted to the fraction of excitation light absorbed by the sample. The emission efficiency was determined for the three lexc as follows: The absorption spectra of the suspended SWNTs were recorded on a Varian Cary 100 Bio UV-visible spectrophotometer by using 1 mm path length quartz cuvettes. To quantify the amount of dispersed SWNTs we used the previously determined3 extinction coefficient of 106.0  1.3 mL mg1 cm at 377 nm. The behaviour of the suspensions in the near-infrared region was examined by means of a Jasco V-570 UV-vis-NIR spectrophotometer. Emission efficiency = total emission power/absorption at lexc (2) Cryo-TEM measurements where the total emission power value corresponds to the integer of the total NIR emission spectrum. NIR-PL spectra were acquired three times for each sample to increase the significance of the estimation. The emission intensity was then calculated as average value (nexp = 3)  standard deviation. This journal is c the Owner Societies 2011 An EM grid with a holey carbon film was held in tweezers and 4 to 5 mL of sample solution, pre-heated to 80 1C, was applied on the grid. The tweezers are mounted on a plunge freezing apparatus (guillotine), and after blotting, the grid was immediately immersed in a small metal container with liquid ethane that is cooled from the outside by liquid nitrogen. Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 11381 View Online The speed of cooling is such that ice crystals do not have time to form. Observation was made at 170 1C in a Philips CM12 EM (Eindhoven, The Netherlands) operating at 100 kV and equipped with a cryo-specimen holder Gatan 626 (Warrendale, PA). Digital images were recorded with a Gatan MultiScan charge-coupled device (CCD) camera 1024  1024. The image processing software was a Gatan Digital Micrograph. Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A Computational details The structure of SWNT was generated using TUBE-VBS code.34 In the present study we used a SWNT with the following features: n = 10, m = 3, carbon–carbon distance of 0.1421 nm. No deformations were included. Obviously the dimensions of the SWNT may severely affect the results from MD simulations. For this reason, we wish to underline that the present results are far from being generalized and our aim is only to suggest some of the main determinants influencing the dispersing capability of investigated ionic liquids. Three MD simulations were carried out: (i) 1 molecule of 1-hexadecyl-3-vinylimidazolium (hvim), 1 SWNT, 3857 water molecules and one counterion (chloride), hereafter termed as SYS1; (ii) 1 molecule of 1-hexadecyl-3-phenylimidazolium (hphim), 1 SWNT, 3857 water molecules and one counterion (chloride), hereafter termed as SYS2; (iii) 1 molecule of 1-hexadecyl-3-(pyren-1-ylmethyl)imidazolium (hpymim), 1 SWNT, 3857 water molecules and one counterion (chloride), hereafter termed as SYS3. All the simulations were performed in the NVT ensemble. For all the systems the SWNT was kept frozen at the centre of the box. All the simulations started with the solute (hvim, hphim and hpymim) at a maximum distance of 2.0 nm from SWNT. This distance was selected as it practically implies the absence of appreciable solute–SWNT interaction, i.e. zero potential energy, still maintaining a reasonably reduced dimension of the overall system. It follows that for all the simulations, for which Periodic Boundary Conditions were used, we adopted a rectangular box as large as 125 nm3. After an initial overall energy minimization and a subsequent dynamical solvent equilibration, we carried out a slow heating of the system (using short trajectories of 50 ps length from 50 K to 312 K). After a few ns of equilibration we produced the trajectory at 312 K to be used for the analysis. The lengths of the productive trajectories were: 123 ns for SYS1, 100 ns for SYS2 and 71 ns for SYS3. A time step of 1 fs was adopted, the temperature was kept fixed with the Berendsen temperature coupling35 with the time constant equal to the integration step. The long-range electrostatics was treated by means of the Particle Mesh Ewald method.36 The Single-Point-Charge (SPC) model37 of water was used at the typical liquid density (55.32 mol l1). The simulations were performed adopting a modified version of the Gromacs 3.0 software package38 with the Gromos96 force field (ffG43a1).39 The point charges of the three solutes were calculated by using standard fitting procedures,40 on the optimized structures in vacuo using Density Functional Theory and the B3LYP functional41 with 6-31G* basis set. The Gaussian03 program42 was used for this purpose. Details of utilized atom types and charges are reported in the ESIw (Tables S1–S3). The analysis of the trajectories was performed using either standard tools of Gromacs or home-made routines. 11382 Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 In order to characterize the relative configurations of the solute with respect to the fixed nanotube, we carried out Essential Dynamics analysis (ED).43 This analysis was performed by considering the solute bound to the nanotube, i.e. by taking only the configurations corresponding to the nanotube–solute minimum free energy basins. This was accomplished by extracting from the MD simulation the frames at which the solute was found at the distance of minimum interaction free energy within a range of 2Dr, with Dr being the value at which we observed a free energy variation equal to room temperature energy (RT). This sequence of configurations was used for constructing the covariance matrix of the ionic liquid atom positional fluctuations. Diagonalization of this matrix provides a set of eigenvectors and eigenvalues. The eigenvectors represent the directions in configurational space and the eigenvalues indicate the mean square fluctuations along these axes. Sorting the eigenvectors by the size of the corresponding eigenvalues, the configurational space can be divided into a low dimensional (essential) subspace in which most of the positional fluctuations are confined and a high dimensional subspace in which merely small vibrations occur. Note that the solute rotations with respect to SWNT, but not the relative translations along SWNT, have been included as internal coordinate. In order to evaluate excess chemical potential of the solutes in water, i.e. their hydration free energy with respect to the gas phase, Thermodynamic Integration (TI) was performed on the solutes in a box of pure water adopting the algorithm as implemented in the Gromacs package by using 100 000 steps in water and in the gas-phase at each value of lambda varying from 0 to 0.10 with steps of 0.01 and from 0.10 to 1 with steps of 0.05. Standard values were used such as soft core potential with a 1.51, s 0.3 nm.44 Detailed results of the TI procedure are reported in the ESIw (Tables S4–S7). The value of hydration free energy was then used, according to the thermodynamic cycle reported in the ESIw (Appendix S1) and within the related approximations, to calculate the solubility ratio between hpymim and hphim. The hydrophilic surface has been evaluated according to the g_sas routine as implemented in the Gromacs package, version 4.0.7. Conclusions The introduction in the polar head of the ionic liquid derived surfactant of aromatic groups enhances the affinity for SWNTs, as confirmed by the simulations, but at the same time renders the molecule less water soluble and more prone to self-assembly. NIR-PL, Raman, AFM and vis-NIR absorbance results show that hphimBr has the best features to be used as SWNT water dispersant, as it favors the dispersion of the nanotubes very well, possesses a good water solubility and an optimal hydrophobic/hydrophilic balance which favors adsorption onto the nanotube sidewalls over self-assembly. Acknowledgements We thank Prof. Maurizio Prato for supplying the nanotubes and for useful discussion and Dr Andrea Renzetti for help and support during the synthesis of the investigated ILs. This journal is c the Owner Societies 2011 View Online CASPUR (Rome) is also acknowledged for the use of Gaussian03 package. This work has been supported by MIUR (PRIN 2008, prot. 20085M27SS). The authors wish to thank the support of Science Foundation Ireland (PIYRA 07/ YI2/I1052) and IRCSET and Intel (Postgraduate Research Scholarships to EDC). Downloaded by Rice University on 18 August 2011 Published on 16 May 2011 on http://pubs.rsc.org | doi:10.1039/C1CP20140A Notes and references 1 B. I. Kharisov, O. V. Kharissova, H. Leija Gutierrez and U. Ortiz Méndez, Ind. Eng. Chem. Res., 2008, 48, 572–590. 2 T. Fukushima, A. Kosaka, Y. Ishimura, T. Yamamoto, T. Takigawa, N. Ishii and T. Aida, Science, 2003, 300, 2072–2074. 3 A. Di Crescenzo, D. Demurtas, A. Renzetti, G. Siani, P. De Maria, M. Meneghetti, M. Prato and A. Fontana, Soft Matter, 2009, 5, 62–66. 4 Z. Miskolczy, K. Sebok-Nagy, L. Biczók and S. Göktürk, Chem. Phys. Lett., 2004, 400, 296–300. 5 J. Sirieix-Plénet, L. Gaillon and P. Letellier, Talanta, 2004, 63, 979–986. 6 R. Vanyúr, L. Biczók and Z. Miskolczy, Colloids Surf., A, 2007, 299, 256–261. 7 M. J. O’Connell, S. M. Bachilo, C. B. Huffman, V. C. Moore, M. S. Strano, E. H. Haroz, K. L. Rialon, P. J. Boul, W. H. Noon, C. Kittrell, J. Ma, R. H. Hauge, R. B. Weisman and R. E. Smalley, Science, 2002, 297, 593–596. 8 Y. Kang and T. A. Taton, J. Am. Chem. Soc., 2003, 125, 5650–5651. 9 P. Kondratyuk, Y. Wang, J. K. Johnson and J. T. Yates, J. Phys. Chem. B, 2005, 109, 20999–21005. 10 G. J. Bahun, C. Wang and A. Adronov, J. Polym. Sci., Part A: Polym. Chem., 2006, 44, 1941–1951. 11 D. M. Guldi, G. M. A. Rahman, N. Jux, D. Balbinot, U. Hartnagel, N. Tagmatarchis and M. Prato, J. Am. Chem. Soc., 2005, 127, 9830–9838. 12 S. Meuer, L. Braun and R. Zentel, Chem. Commun., 2008, 3166–3168. 13 Q. Yang, L. Shuai, J. Zhou, F. Lu and X. Pan, J. Phys. Chem. B, 2008, 112, 12934–12939. 14 N. Nakashima, Y. Tomonari and H. Murakami, Chem. Lett., 2002, 31, 638–639. 15 C. Ehli, G. M. A. Rahman, N. Jux, D. Balbinot, D. M. Guldi, F. Paolucci, M. Marcaccio, D. Paolucci, M. Melle-Franco, F. Zerbetto, S. Campidelli and M. Prato, J. Am. Chem. Soc., 2006, 128, 11222–11231. 16 H. Paloniemi, T. Ääritalo, T. Laiho, H. Liuke, N. Kocharova, K. Haapakka, F. Terzi, R. Seeber and J. Lukkari, J. Phys. Chem. B, 2005, 109, 8634–8642. 17 G. Srinivas, J. C. Shelley, S. O. Nielsen, D. E. Discher and M. L. Klein, J. Phys. Chem. B, 2004, 108, 8153–8160. 18 M. Calvaresi, M. Dallavalle and F. Zerbetto, Small, 2009, 5, 2191–2198. 19 E. M. Di Meo, A. Di Crescenzo, D. Velluto, C. P. O’Neil, D. Demurtas, J. A. Hubbell and A. Fontana, Macromolecules, 2010, 43, 3429–3437. 20 D. F. Evans and H. Wennersrom, The colloidal domain, Wiley-VCH, New York, 1999. 21 W. Tong, Q. Zheng, S. Shao, Q. Lei and W. Fang, J. Chem. Eng. Data, 2010, 55, 3766–3771. 22 I. Lin, Tenside Deterg., 1980, 17, 119–123. 23 R. B. Weisman and S. M. Bachilo, Nano Lett., 2003, 3, 1235. This journal is c the Owner Societies 2011 24 Y. Ohno, S. Iwasaki, Y. Murakami, S. Kishimoto, S. Maruyama and T. Mizutani, Phys. Rev. B: Condens. Matter, 2006, 73, 235427. 25 S. M. Bachilo, M. S. Strano, C. Kittrell, R. H. Hauge, R. E. Smalley and R. B. Weisman, Science, 2002, 298, 2361–2366. 26 M. S. Strano, S. K. Doorn, E. H. Haroz, C. Kittrell, R. H. Hauge and R. E. Smalley, Nano Lett., 2003, 3, 1091–1096. 27 R. Marquis, C. Greco, I. Sadokierska, S. Lebedkin, M. M. Kappes, T. Michel, L. Alvarez, J.-L. Sauvajol, S. Meunier and C. Mioskowski, Nano Lett., 2008, 8, 1830–1835. 28 L. Jiang, L. Gao and J. Sun, J. Colloid Interface Sci., 2003, 260, 89–94. 29 D. Ma and Q. Cai, Synlett, 2004, 128–130. 30 W. Deng, Y. F. Wang, C. Zhang, L. Liu and Q. X. Guo, Chin. Chem. Lett., 2006, 17, 313–316. 31 I. Alves, S. Cowell, Y. S. Lee, X. Tang, P. Davis, F. Porreca and V. J. Hruby, Biochem. Biophys. Res. Commun., 2004, 318, 335–340. 32 S. K. Kim, B.-S. Moon, J. H. Park, Y. I. Seo, H. S. Koh, Y. J. Yoon, K. D. Lee and J. Yoon, Tetrahedron Lett., 2005, 46, 6617–6620. 33 C. Wang, Z. Tong, F. Zeng, B. Ren and X. Liu, Colloid Polym. Sci., 2002, 280, 814–821. 34 C. T. White, D. H. Robertson and J. W. Mintmire, Phys. Rev. B: Condens. Matter, 1993, 47, 5485–5488. 35 H. J. C. Berendsen, J. P. M. Postma, W. F. van Gunsteren, A. Di Nola and J. R. Haak, J. Chem. Phys., 1984, 81, 3684–3690. 36 T. Darden, D. York and L. Pedersen, J. Chem. Phys., 1993, 98, 10089–10092. 37 H. J. C. Berendsen, J. P. M. Postma, W. F. van Gunsteren and J. Hermans, Interaction models for water in relation to protein hydration, in Intermolecular forces, ed. B. Pullman, Reidel Publishing Company, Dordrecht, The Netherlands, 1981. 38 E. Lindahl, B. Hess and D. van der Spoel, J. Mol. Model, 2001, 7, 306–317. 39 W. F. van Gunsteren, S. R. Billeter, A. A. Eising, P. H. Hünenberger, P. Krüger, A. E. Mark, W. R. P. Scott and I. G. Tironi, Biomolecular Simulation, the GROMOS96 Manual and User Guide, Verlag der Fachvereine Hochschulverlag AG an der ETH, Zurich, 1996. 40 B. H. Besler, K. M. Merz and P. A. Kollman, J. Comput. Chem., 1990, 11, 431–439. 41 A. D. Becke, Phys. Rev. A, 1988, 38, 3098. 42 M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, Jr, J. A. Montgomery, T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli,J. W. Ochterski, P. Y. Ayala, K. Morokuma,G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz,Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, Revision C.02, Gaussian, Inc., Wallingford CT, 2003. 43 A. Amadei, A. B. M. Linssen and H. J. C. Berendsen, Proteins, 1993, 17, 412–425. 44 A. Villa and A. E. Mark, J. Comput. Chem., 2002, 23, 548–553. Phys. Chem. Chem. Phys., 2011, 13, 11373–11383 11383