Colloidal dispersions of oxide nanoparticles in ionic liquids: elucidating the key parameters

The combination of ionic liquid and nanoparticle properties is highly appealing for a number of applications. However, thus far there has been limited systematic exploration of colloidal stabilisation in these solvents, which provides an initial direction towards their employment. Here, we present a new and comprehensive study of the key parameters affecting the colloidal stability in dispersions of oxide nanoparticles in ionic liquids. Twelve diverse and representative ionic liquids are used to disperse iron oxide nanoparticles. The liquid interface of these nanoparticles has been carefully tuned in a molecular solvent before transferring into an ionic liquid, without passing through the powder state. Multiscale-characterisation is applied, on both the micro and the nano scale, incorporating both small angle X-ray scattering and dynamic light scattering. The results show the surface charge of the nanoparticles to be a crucial parameter, controlling the layering of the surrounding ionic liquid, and hence producing repulsion allowing efficient counterbalancing of the attractive interactions. For intermediate charges the strength of the repulsion depends on the specific system causing varying levels of aggregation or even none at all. Several samples consist of sufficiently repulsive systems leading to single dispersed nanoparticles, stable in the long term. Thanks to the magnetic properties of the chosen iron oxide nanoparticles, true ferrofluids are produced, appropriate for applications using magnetic fields. The strength and breadth of the observed trends suggests that the key parameters identified here can be generalised to most ionic liquids.

preparation is detailed in the following sections.

S-1.3.1 Bare nanoparticles with hydroxyl groups
1M NaOH aqueous solution was added to 1 vol% of initial ferrofluid in water (prepared as described above, with nitric acid, pH around 1.5) till the point of zero charge (PZC) of maghemite was reached at pH ≈ 7. The nanoparticles were magnetically precipitated and the supernatant was removed. The particles were washed several times with ultra-pure water until the free NO 3 concentration was reduced below 10 -6 M. At the PZC these uncharged bare nanoparticles have hydroxyl groups on their surface. To test the influence of the nature of the counter-ion, several different aqueous acids were added, including perchloric acid (HClO 4 ), benzenesulfonic acid (C 6 H 5 SO 3 H), bistriflimide (HTFSI), and aqueous 1-(4-sulfobutyl)-3-methyl imidazolium bistriflimide (HSMIM TFSI) solution. For each case the pH was reduced to ≈ 1.5, leading to stable colloids except for the addition of SMIM +/-TFSIsolution which led to flocculated nanoparticles in water.

S-1.3.2 Citrate coated particles
The addition of citric acid to 1 vol% of initial ferrofluid in water led to sedimented nanoparticles. They were washed several times with a 10 -2 M aqueous citric acid solution with the help of magnetic induced sedimentation and removal of the supernatant to obtain a free citric acid concentration of ≈ 10 -2 M. A 1 M base of the type XOH, with X = Li, Na, tetrabutyl ammonium (TBA) or 1-butyl-3-methyl imidazolium (BMIM), was added till a pH ≈ 7 was reached, leading to a stable colloid in all cases. The negatively charged citrate coated nanoparticles balance their charge with the cations X of the added base.

S-1.3.3 Polymer coated particles
68% concentrated nitric acid was added to 100 mL of 1 wt% aqueous Poly (Acrylic Acid Sodium salt) solution till a pH of ≈ 2 was reached. 50 ml of 1 vol% aqueous initial acid ferrofluid with NO 3 counter ions was added. After 1 minute of magnetic stirring the nanoparticles were magnetically precipitated and the supernatant was removed. Ammonia solution was added to disperse the nanoparticles at pH ≈ 7. In a last step the remaining free ions and polymer is removed by dialysis with 10-12 kDa membranes. The counter-ion is then NH 4 . The same steps were applied for Poly (Acrylic co-Maleic Acid) (PAAMA) except the addition of nitric acid in the beginning as the PAAMA solution was already at pH ≈ 2. The counter ions were changed by the following steps. First, nitric acid (1M) was added till a pH ≈ 2 was reached and the precipitated particles were washed several times. Second, BMIMOH (0.02M) or TBAOH (1M) was added to disperse the nanoparticles at pH ≈ 7. In a last step the remaining free ions are removed by dialysis with 10-12 kDa membranes. The solubility of poly (acrylic acid) was tested in the 12 ionic liquids in order to connect the colloidal stability of PAAcoated NPs with the solubility of the polymer in the ionic liquids. Therefore, ammonia solution was added to poly (acrylic acid) with 16wt% PAANa and 47wt% PAAH till the ratio of H + and NH 4 + was 1:1 and the pH was close to 7 to exchange H + with NH 4 + . NH 4 + was one of the counter-ions that lead to colloidal stability in some of the ionic liquids. The solution was freeze-dried and heated at 120 • C while pumping for 3 hours to remove the water. Around 0.01 mol of the polymer powder per litre of ionic liquid was added. After one day of magnetic stirring the polymer was dissolved in EMIM X (X= AcO, EP, and DEP). The other samples were brought into contact with the moisture of the air and after another week, the polymer was dissolved only in HSMIM TFSI. This can be due to the large water uptake of this ionic liquid, here around 13wt% after 1 week.

S-1.3.4 Transfer from water to DMSO
In contrast to water, dimethyl sulfoxide (DMSO) is miscible with all tested ionic liquids. Therefore, all these different systems were transferred from water to DMSO in order to test the influence of the way of transfer from the molecular solvent towards the ionic liquids. A dialysis step was used, with 10-12 kDa membranes. As the nanoparticles were bigger than the holes in the membrane, only the solvent molecules and the ions were exchanged. The outside DMSO solution with the adapted free ion concentration is changed several times until the water concentration is lower than 10 -2 vol%. In a last step the different systems were added to the ionic liquid in adapted proportions to obtain, after removing DMSO by freeze-drying, a final colloid with a volume fraction of ≈ 0.01.
Table S-2. Structure factor at low Q, S(Q→ 0), of ionic liquid-based colloids from SAXS (see text for details). * In the case of collapsed polyelectrolyte. If the polymer is not collapsed, part of the charge is on the surface of the oxide and part in the solvent, which is a different and more complicated situation. See main text for more information. "yes/no" means that the sample destabilised before the SAXS measurements. "no" means that dispersions with aggregates on the micron scale or bigger are obtained and ** means that there are no aggregates on the micron scale but there was not enough sample left to perform the SAXS measurement. OH 2 + : hydroxyl protonated ending groups on the nanoparticles; citrate -: citrate coated nanoparticles; PAAand PAAMA -: polymer coated nanoparticles (poly acrylate and poly(acrylate-co-maleate), respectively); SMIM +/-TFSI -: butylsulfonate methylimidazolium bistriflimide (SMIM +/being a zwitterion); TBA + : tetrabutylammonium; BMIM + : butylmethylimidazolium; NH 4 + : ammonium.

S-1.4.1 Flame Atomic Absorption Spectroscopy (FAAS)
The maghemite-based ferrofluids were dissolved in a concentrated hydrochloric acid solution and their total iron concentration were determined by flame atomic absorption measurements (FAAS) with an Analyst 100 spectrometer from PerkinElmer. The volume fractions φ were determined from the total iron concentration in the sample. Taking the molar weight (159.7 g mol -1 ) and density (5.07 g cm -3 ) of maghemite, 9 the volume fraction of NPs can be calculated as: φ (vol%) = [Fe](mol L -1 )*1.577. Room temperature magnetisation measurements were performed on the initial aqueous ferrofluid and two ionic liquid-based ferrofluids using a home-made Vibrating Sample Magnetometer (VSM) up to H=700kA/m. When magnetised, the vibrating sample generates in the measuring coil a voltage at the same frequency proportional to the magnetisation M of the material.  for the nanoparticles in the initial sample in water and after transfer to the ionic liquid, here as an example the case of EMIM TFSI (same data as (a)), using SANS. The form factor of the nanoparticles is also plotted.

S-1.4.3 Small Angle X-ray and neutron Scattering (SAXS and SANS)
SAXS experiments were carried out with a laboratory XEUSS 2.0 (W)SAXS. The beam energy was fixed at 8 keV and the wavelength at λ =1.54 Å. The sample to detector distance was 3m to yield an accessible Q-range of 0.004 Å -1 -0. However, as explained in the main text of the paper, the samples strongly absorb X-rays due to the ionic liquids and to iron. Therefore, either the transmission was very low when using classical calibrated polyimide capillaries (thickness 1.5mm) or the thickness was not accurate when reducing the thickness with home-made cells (thickness < 0.1 µm). As a consequence, absolute intensity cannot be determined and the high Q region (> 0.2 Å −1 ) is highly noisy (see Fig. S-1a). Therefore SANS were performed on the PAXY spectrometer at the LLB facility (CEA Saclay, France). Three different configurations were used (neutron wavelength λ =6Å, sample to detector distance d=1m; λ =6Å, d=3 m and λ =8.5Å, d=5 m) leading to an accessible Q-range of 0.004 Å -1 to 0.2 Å -1 . Samples were measured in quartz cells with 1 mm inner thickness. Standard correction procedures were performed 10 using the Pasinet software (available free of charge at http://didier.lairez.fr/) in order to determine the absolute intensity (cm -1 ). The SAXS curves rescaled on the SANS measurement in figure S-1a prove that these SAXS measurements are reliable in the small Q range. Figure S-1-b shows SANS measurements of colloidal dispersions of the same NPs in water and in an ionic liquid (namely EMIM TFSI). The high Q region where the scattered intensity can be identified to the form factor of the NPs (Q above 0.06 Å −1 ) is properly measured by SANS (by opposition to SAXS in Fig. S-1-a). We then conclude that the nanoparticles keep their shape and their size distribution after transfer from water to the ionic liquid. This conclusion is valid for all dispersions measured by SANS.

S-1.4.4 Dynamic Light Scattering (DLS)
Light scattering measurements were performed using a Vasco DLS Particle Analyzer from Cordouan Technologies to study translational diffusion properties of water-based and IL-based dispersions for volume fractions φ ranging from 0.1vol% to 1.53vol%. The laser of the device is operating at a wavelength λ =656 nm. An around 200 µm thin film is analysed in backscattering mode at 135 • (i.e. Q=2.3*10 -3 Å -1 in water). These conditions prevent both the effects of light absorption by the colloidal suspension and multiple scattering even in strongly absorbing media. 11 Field autocorrelation curves with a precise baseline at long times were obtained by optimizing parameters such as the incident laser power, the sampling time and the number of channels. The measured intensity correlation function G(t) was transformed into G 1 (t) using the expression: The normalized intensity auto correlation functions G 1 (t) were analysed with a stretched exponential function, e -(t/τ ) β with a distribution of relaxation times described by a decay time τ and a stretching exponent β . The nanoparticle translation time τ is probed. It corresponds to the translational diffusion coefficient D t = (<τ> Q 2 ) -1 , which, in non-interacting conditions, is related to the hydrodynamic radius R H using Stokes-Einstein's equation: where k B is the Boltzmann constant, T the absolute temperature, η the solvent shear-viscosity (0.89*10 -3 Pa s for water; see η in table S-1) for ionic liquids; the values are given for T=25 • C at which all experiments were performed. The values for the determined hydro-dynamical diameter d H of the samples depicted in figure 7b) of the main paper are given in table S-4. The reference hydro-dynamical diameter d H of the initial sample in water is 13.4 nm for 1 vol% NP concentration. The stretching exponent β is close to 1 for all measurements meaning that there exists only one population of particle size. For NPs with hydroxyl protonated groups on their surface and SMIM +/-TFSIas counter-ions the zwitterion concentration is around 2%. This could increase the viscosity of the sample whereas the samples viscosity would be decreased by a higher water content. A plausible order of magnitude of an eventual amount of molecular solvent remaining after freeze-drying is the amount of water taken by the ionic liquid alone after 24 hours at air. Assuming that both contributions influence linearly the viscosity, an average hydro-dynamical diameter d H can be calculated with equation 2 and an error bar is estimated.