Effect of Water on a Hydrophobic Deep Eutectic Solvent

Deep eutectic solvents (DESs) formed by hydrogen bond donors and acceptors are a promising new class of solvents. Both hydrophilic and hydrophobic binary DESs readily absorb water, making them ternary mixtures, and a small water content is always inevitable under ambient conditions. We present a thorough study of a typical hydrophobic DES formed by a 1:2 mole ratio of tetrabutyl ammonium chloride and decanoic acid, focusing on the effects of a low water content caused by absorbed water vapor, using multinuclear NMR techniques, molecular modeling, and several other physicochemical techniques. Already very low water contents cause dynamic nanoscale phase segregation, reduce solvent viscosity and fragility, increase self-diffusion coefficients and conductivity, and enhance local dynamics. Water interferes with the hydrogen-bonding network between the chloride ions and carboxylic acid groups by solvating them, which enhances carboxylic acid self-correlation and ion pair formation between tetrabutyl ammonium and chloride. Simulations show that the component molar ratio can be varied, with an effect on the internal structure. The water-induced changes in the physical properties are beneficial for most prospective applications but water creates an acidic aqueous nanophase with a high halide ion concentration, which may have chemically adverse effects.


DES preparation and water saturation
TBAC, the HBA, was weighted in an Erlenmeyer flask; then, DecA, the HBD, was added. Then, the flask was heated in a water bath at 65°C for three hours, periodically mixing the content, until the mixture turned to a homogeneous, clear and colorless fluid. The walls of the flask were rinsed with the liquid in order to ensure that all the weighted solid was in the liquid phase. Figure S1 shows a typical experimental setup, with the water bath used. Figure S1. Experimental setup: water bath used to synthesize the DES.
As previously reported, water is a primary component in these solvents. To highlight differences related to the water content, DESs have been synthesized with the procedure described above, adopting however different conditions. Three sets of samples were prepared:  DESs prepared in ambient conditions, in equilibrium with the ambient atmosphere of known temperature  DESs prepared inside the glovebox, with water and oxygen only present in traces (< 4 ppm).  DESs prepared in ambient conditions, then placed in a water-vapor saturated atmosphere at different temperatures (from 30 °C to 60 °C) for several hours in order to equilibrate the solvent with water vapor. These different synthetic procedures developed allowed to change the water content of the DES.

1.A. Calculation of the water content
For a bicomponent DES with a composition f AB we have BA nf n  (Note: f need not be an integer). If the water-% of the DES (weight-%) is % w the, the water mole fraction is then given by  Figure S2. The stimulated echo 1 H NMR pulse sequence used in this work for measuring the selfdiffusion coefficients. See text for explanations.
Here  is the magnetogyric ratio of proton ( 8 11 2.675 10 s T   ),  the total duration of the bipolar pulse pair, and  the diffusion time. The parameter δ is related to the gradient pulse duration, 23 0 p   (it is the total duration of the bipolar gradient pair G6 and -G6 appearing twice in the pulse program). The diffusion time ∆ is the time given to the molecules to diffuse during the pulse program. It is roughly the time between the dephasing gradient pair at the start and the rephasing gradient pair at the end of the program, 20 16 2 dd   . The time parameters 16,20,30 ddp are set by the user during the optimization of the diffusion experiment, and are thus known in the above equation. In this work, the delays were d20 = 250 ms, d16 = 0.5 ms, and the duration of the gradient pulse G6 (p30) was 6 ms for the wet and 11 ms for the dry DES mixture. In practice, the diffusion experiment is repeated multiple times at increasing values of G (in this work, 32 times). According to the equation above, a Gaussian-type decay ). The diffusion datasets were processed by using Bruker's TOPSPIN 3.5 software. The diffusion measurement yields a "pseudo-2D" data matrix with 32 rows: each row corresponds to a 1 H NMR spectrum at different G and consists here of 64k intensity points. TOPSPIN's "T1/T2 module" may be used to integrate the user-defined 1 H peaks i in the diffusion pseudo-2D dataset as a function of G and to perform a Gaussian fit on the resulting vs.
i IG data for the calculation of the diffusion coefficient i D for each user-defined 1 H peak.     IR spectra were acquired using a Bruker Vertex 70 FT-IR spectrometer, equipped with BR4 Diamond attenuated total reflection (ATR) accessory (Harrick). A room temperature deuterated L-Alanine doped triglycine sulfate (RT-DLaTGS) detector was used. The spectra were measured using both liquid and solid samples without any pretreatment. Adherence to the diamond crystal was controlled with a camera.
The C=O vibration (ca. 1700 cm -1 ) is shifted higher in DES with respect to DecA by ca. 30 cm -1 , in opposite direction than reported earlier. 3 No peaks related to the stretching or the bending of the O-H bond, at ca. 1050 and 650 cm -1 , respectively, or peaks typical of the carboxylic acid dimers (at ca. 1400, 1300 and 900 cm -1 , and its overtones over 2400 cm -1 ). 4 Water has negligible effect on the spectra, except for the appearance of a wide band around 3500 cm -1 (OH vibrations) at high water content. Raman spectra on solid and liquid samples were measured with a Qontor inVia confocal Raman Microscope (Renishaw). The excitation laser wavelength was 532 nm and its power calibration was performed using a Si (111) standard. The maximum power was 50 mW but the laser was set up at 1 % of maximum. Samples were placed on a plastic cap. Water has a negligible effect on the Raman spectra. However, the Raman spectrum of the DES is not exactly the weighted sum of the components, as seen in Fig. S9b. The contribution of TBAC is more pronounced in the DES spectrum with reference to DecA.   The 1 H and 13 C NMR chemical shifts of the mixture components in a dry and wet DES preparation are given in Table S3, together with their assignment. The 14 N NMR spectrum of each preparation consisted of a single broad peak (cf. Fig. S50) due to the TBA nitrogen at −314.5 ppm (dry DES) and −314.7 ppm (wet DES). the two equations only refer to different standard states. Water concentrations must be corrected to the saturation temperature T by using the volume expansion factors given below.  The saturation process is energetically favorable (exothermic), as can be expected because hydrogen bond formation is usually an exothermic process. On the other hand, it is entropically unfavorable. This is a common case when gases dissolve in liquids, and it causes the gas solubility to decrease with increasing temperature. This factor actually limits the amount of water in DES even though temperature would be raised. Because we assume that at every temperature used the DES is saturated with water vapor the solvents so formed and cooled down to room temperature should, in fact, be oversaturated with water.

Differential scanning calorimetry (DSC)
The following sample treatment was applied: 1) equilibration at -55˚C for 10 minutes; 2) heating from -55˚C to 120˚C at 1˚C/min; c) annealing at 120˚C for 10 minutes; d) cooling from 120˚C to -55˚C at 5˚C/min.  H  , "+" sign exotermic, "-" sign endothermic for enthalpy changes; b sharp peak; c level change; d reaction with the substrate Van Osch et al., who studied the same DES with the water mole fraction of 0.086, reported a exothermic peak at around -60˚C (that we are not able to see, because of the instrument used), an exothermic peak at around -40˚C (observed in this work, too) and an endothermic peak (attributed to melting) at -11.95˚C. 3 Taking this value into account the temperature of the major endothermic processes depends nearly linearly on the water mole fraction in this range.

6.A.General
The modelling was applied to TBAC-DecA mixtures (at 1:2 mole ratio) with five different water contents (from dry to close to the observed maximum water content) at different temperatures.
All simulations were performed with GROMACS 2016.1 5-11 using the Bussi-Donadio-Parrinello thermostat (V-rescale) 12 . Simulation pressures were controlled with the Berendsen barostat 13 , and with the Parrinello-Rahman barostat 14 during production simulations. Constraints were solved using LINCS 15 with a LINCS order of 4. All production simulations were run with a 1 fs timestep.
Liquid structures were studied by evaluating the radial distribution functions of key (atom) pairs: Water -chloride, central nitrogen of TBA -chloride, and COOH of DecA -chloride. Selfcorrelations were computed for all components listed before, and for alkyl chains of TBA. Bin volume and density were normalized, and the evaluation was performed on the whole trajectory.
GROMACS compatible molecular topologies and starting structures were obtained from the Automated Topology Builder (ATB) 16 . Corresponding structures are available with the following information: ATB molid 19774 for decanoic acid and 303364 for the tetrabutylammonium-cation. We have provided the parameters for both compounds in the Appendix at the end of the Supporting Information.
OPLS-AA compatible molecular topologies and starting structures were obtained from the LigParGen server. [17][18][19] Partial atomic charges for the OPLS-AA models were further refined using a Hirshfeld population analysis using the Gaussian09 software and the B3LYP/6-311G basis set. 20 GROMOS54a7 topologies had a compatible Hessian-based analysis and partial charge assignment done within the ATB suite, and further optimization was not deemed necessary.
Simulations, viscosity, density and MSD calculations, RDFs and clustering calculations were performed using GROMACS 5.15 and later migrating to GROMACS 2016.1. No meaningful differences were observed during the version migration. Intramolecular interactions were calculated with harmonic potentials, as demonstrated in the appendix of the SI. Electrostatic interactions were computed with the smooth particle mesh Ewald (PME) method. All systems were first minimized using steepest descent scheme, and subsequently equilibrated in a NVT ensemble at 103 K and 303 K, both simulations lasting 500 ps. Final step of the equilibration was performed with a NPT ensemble, lasting 10 ns.
The classical models employed in the computational section of this study are optimized for simulation of organic compounds in the liquid phase, and have been successfully employed in a multitude of studies including similar systems as ours. 19,21 Initial test simulations were performed with the size of the simulation cell set at 55.884 nm 3 , containing 100 decanoic acid and 50 tetrabutylammonium chloride molecules, and either 0, 10, 18, 25, or 33 water molecules, corresponding to water mole fractions in the range 0.0 -0.18. These small simulations were performed in order to verify our approach before applying for HPC resources. Later simulations were extended by a tenfold, containing 500 molecules of tetrabutylammonium chloride, 1000 molecules of decanoic acid, and 0-330 molecules of water. These simulations were performed for 30 ns, with the first half excluded as an equilibration period. Our approach included model validation utilizing two force fields, OPLS-AA and GROMOS54A7, both of which predicted the same behaviour. 22 Furthermore, we observed no change in the measured properties when the system size was increased by a tenfold, suggesting that the simulations have captured a stable state of the system, emerging from the complex interplay of the simplistic potentials used.
Z-density profiles yield qualitative information about the 'microphases' present in our system, by binning the volume of the simulation cell along the z-axis. Each bin has a volume of 0.05 × × Å 3 , where 0.05 Å is the length of the bin along the z-axis, is the length of the x-axis, and is the length of the y-axis. Dimensions of the cell were (30.9 Å)² × 28 Å = 26.735 nm 3 . The resulting graphs clearly indicate how the addition of water drives the system in to a state of 'microphases', where partial density profiles start to locally converge towards zero. The Z densities of two atoms A and B are denoted by A Z and B Z , respectively. The "overlap" of atoms i and j from the Z-density calculations is calculated as where the limits are the z-values of the ends of the simulation cell. This is only a semiquantitative approach because phase segregation can take place in the x-y directions, too. However, the direction of the axis is arbitrary, and a semiquantitative measure of the phase separation can be obtained by comparing the overlap values of the dry DES to those with water added.
The existence of clustered groups was investigated by utilizing a built-in GROMACS analysis tool, which computes distributions of different sized clusters in the simulation cell. The tool is given a set of atoms and a cut-off radius, which is used to define the clusters. In this study the analysis shows (qualitatively) how the ratio of differently sized clusters change as the mole fraction of water is increased. Because the total number of atoms in the clusters equals the number of those atoms in the sample the numbers of clusters observed were normalised by the total number of atoms in the set to obtain the probability distributions of clusters.  The small correction due to the finite-size effects under periodic boundary conditions was taken into account in the simulations. 23 We emphasize that the simulation results should be considered only indicative and semiquantitative if there is no direct experimental evidence. However, the experimental data and simulation results in this work are compatible and support each other.

6.B. Radial distribution functions
The radial distribution function (RDF) of an atom B around atom A is defined as wr is the Helmholtz free energy, () grthe radial distribution function, R the general gas constant, and T the system temperature.

Conclusions
The OW-OW RDF showed the first coordination sphere peak at 2.7 Å, followed by other peaks with rather complicated behavior. The 2.7 Å peak corresponds to hydrogen-bonded water molecule OHO   . 24 The coordination number (amount of water O in the first solvation sphere) is always lowest at lowest water fraction (xw = 0.0063) but there seems to be an "optimum" at water fraction 0.1 -0.15. Temperature broadens the coordination sphere but the behaviour of the coordination number seems rather complicated. In water, the coordination number of OW around OW is approximately 5 but it is much lower (0.2 -1.3) in DES with different water content and temperature. 24 A low self-coordination number of water (1.5) has been predicted for malicine (1:1 mixture of choline chloride and malic acid) even at water fraction 0.5, and has been interpreted to imply a small degree of water self-clustering. 25 The OW-HW RDF has a sharp peak at ca. 1.75 Å, which corresponds to the closest hydrogen atom in the hydrogen-bonded water molecule. Another peak at ca. 3 Å is much wider, and corresponds to the other H atom in that water molecule (because the geometry is not locked the distance can vary and produce a wide peak). This peak has a shoulder, which is clearly separated at 280 K in the driest sample (xw = 0.0063). This suggests two favourable geometries for the closest water molecule. The peak maxima are very close to the values derived from experimental data (1.8 and 3.3 Å), 24 which supports the validity of the method.  The first coordination sphere of Cl around water oxygen (OW) at ca. 3.2 Å contains 1.2 -1.5 chloride atoms, the position independent of temperature and water content. This is very close to the Cl-O distance reported for aqueous chloride solutions. 26,27 Much smaller broad peaks at ca. 4.5 -4.9 Å and 5.5 -5.9 Å. Effect of temperature on coordination number is not clear, the values (below 4 Å) go through a minimum or maximum at different water fractions, being in the range 1.0 -1.5. When the water fraction in DES is increased from zero to xw = 0.0063 the coordination number of Cl around OW is 1.2 -1.5; the same addition of water drops the Cl coordination around OH from 0.9 -1.0 to 0.45 -0.6. The coordination of Cl around OW does not significantly change in DESs with higher water content. Water effectively solvates Cl and "frees" it from OH. As the water concentration is always much smaller than that of OH the hydrogen bonding of Cl to water OH (HW) is favourable to bonding to the carboxylate OH. Because there are more Cl atoms than OW atoms this implies that all water molecules are surrounded by chloride.    Conclusions (OH = oxygen in DecA OH group) One sharp peak at 3.2 Å, very small and broad secondary peak above 5 Å in DES with water (not in dry DES). Coordination number 0.9-1.0 in dry DES, drops to 0.3 -0.56 in DES with water (the higher the higher the temperature). No significant change with water fraction after the initial introduction of water (when xw 0.0  0.0063), except a small change at 280 K. Temperature may increase the mobility of Cl and allow it to approach OH better at higher temperature. Water competes with OH for the coordination of Cl, thereby decreasing the coordination number of Cl around OH. The Cl-HO hydrogen bonding is considered the basis of DES formation and, therefore, water effectively disrupts the DES structure.

Conclusions
In all cases there is a single peak at 4.5 Å, and no structure beyond that. This is close to the observed shoulder in the N-Cl RDF for the choline -Cl pair (ca. 5 Å), which has been attributed to chloride ion next to the positively charged quaternary nitrogen, and to the N-Br distance in concentrated aqueous solutions of TBABr (5 Å). 25,29 This suggests that the halide anion penetrates inside the hydrocarbon arms of TBA + in DES even more than in the reported cases. The coordination number (Cl around N) is ca. 2.7, except in the dry sample (xw = 0.0), where it is 2.2-2.4. Therefore, addition of a small amount of water increases the average number of Cl atoms (chloride anions) around the N atom (N + ). The temperature has only a negligible effect. In concentrated aqueous TBABr solutions the coordination number of Br around N is only ca. 1/3. 29 The peak at 4.5 Å is rather broad, which implies variable geometry around the nitrogen atom. It is close to the RDF peak 4.7 Å for TEA-BF4 in acetonitrile. 30 Because of the large size of the TBA + ion the N-Cl correlation in DES must represent a contact ion pair with no intervening species, evident also because of the halide penetration inside the hydrocarbon arms of TBA. Addition of water frees Cl from OH, to which it is hydrogen-bonded in pristine (dry) DES.   Carboxylic acids form hydrogen-bonded dimers but their structures are considerably less welldefined in solutions than in the gas phase. 31 For acetic acid, in a head-head dimer (COOH-HOOC) the C-C distance is 3.8 Å in the gas phase, and larger (4-4.7 Å) for other dimer conformations. However, in pure or highly concentrated HOAc the maximum in the C-C RDF is at ca. 4.5 Å, which is attributed to a chain of acetic acid molecules (hydrogen bonds C=O ---CH3, C-OH ---O=C). In case of long-chain carboxylic acids, interactions between the chains can modify the dimer geometry in aqueous solutions. 32 The peak at 1.7 Å is the CO(OH) -HOH (CO -HW) distance. The RDF above resembles the one for acetic acid, except that the peak is at 4.1 Å here. This suggests the presence of other kind of dimers. Less than one DecA around each DecA, except approximately one molecule at high water fraction.    The thermal expansion coefficient is given by

Thermal expansion, excess quantities, and surface tension
The density and viscosity of a pure dry DES have been estimated from the intercept of the linear fits of the corresponding data in Figs. 5a,b, and the effective DES molar mass is taken to be (the molar mass of a DES is not an unambiguous concept, and some authors have used mole fraction weighed value or the geometrical mean; however, here we regard dry DES as a compound and the wet DES as a mixture TBAC(DecA)2-water ). 33 The scatter in the data for density, viscosity, and conductivity vs. water mole fraction is attributed mostly to the uncertainty of the Karl Fischer titration of small amounts of water.
The surface tension of the DES formed by the 1:2 ratio of TBAC and DecA is (0.0308 ± 0.0009) Nm -1 at room temperature, and it is independent of the water content (Fig. S34). This impliesapproximately zero surface excess but, because of the measurement technique, we cannot rule out the possibility of rapid water adsorption on the surface of a dry sessile DES drop.   Fig. 3). 34,35 The Vogel-Fulcher-Tammann (VFT) fits of the temperature dependence of viscosity can be used to obtain fragility parameters for the DES samples. The VFT equation for viscosity is

Viscosity and fragility
 where D is the Angell fragility parameter. 36 We can also write the VFT equation as Fragility (also called the steepness index) m is defined using  (or  ) as ( g T = the glass transition temperature) 34 which is the equation (4b) in ref. 35 . Introducing the numerical values we have which has been used in this work to estimate the fragility.
In glass-forming liquids, the following expression has been put forward to estimate the glass transition temperature from fragility parameters, 34 and this estimate is shown in the last column of the Table S5.     to the conductivity data (conductivity corrected for thermal expansion).

11.Local dynamics from NMR relaxation times T1 relaxation times
The major relaxation mechanism of the 13 C nuclei is dipole-dipole relaxations. According to the Bloembergen-Purcell-Pound (BPP) theory the relaxation rate is given by (assuming only dipoledipole relaxation) 37,38 can be used to calculate A0 for that particular nucleus, and the correlation times at all temperatures (at every value of T1).
In this case, the T1 relaxation times of the TBA C1' and C2' carbons in wet DES display minima, which allow calculate the factor A0 for these nuclei. However, because of the strong distance dependence ( 6 r   ) A0 can be assumed to be unaffected by the water content, and the same value can also be used for these nuclei in the dry DES sample.

T2* relaxation times
The T2* relaxation times affect the NMR line widths and include contribution from both the spinspin relaxation and magnetic field inhomogeneity (eq. S20), and they can be calculated from the full width at half height (FWHH) of the peak by    Figure S54. Effect of temperature on the 14 N peak of TBA in dry DES. The spectra are at 5 K intervals from 298 K (light pink) to 333 K (blue). The peak position remains constant within 0.1 ppm.
The peak width is caused by the natural line width and the inhomogeneity broadening, and the true T2 relaxation time and the observed T2* are related by inhomog * 22 11 B TT    (S20) The field inhomogeneity effects all nuclei but the line widths of the 13 C peaks are of the order of 1 Hz, while the line width of the 14 N peaks varies between 13 -81 Hz. Therefore, we can make the approximation in case of 14 N peaks * 22 TT  The T2 relaxation time is a decreasing function of the rotational correlation time , 37      Appendix: GROMACS compatible molecular topologies and starting structures Figure SIApp1: The all-atom structure of tetrabutylammonium-cation. Bonded and non-bonded terms of each butyl-chain are identical, and parameters for the transparent parts are omitted.   Figure SIApp2: The all-atom structure of decanoic acid.