Ouzo Effect Examined at the Nanoscale via Direct Observation of Droplet Nucleation and Morphology

Herein, we present the direct observation via liquid-phase transmission electron microscopy (LPTEM) of the nucleation and growth pathways of structures formed by the so-called “ouzo effect”, which is a classic example of surfactant-free, spontaneous emulsification. Such liquid–liquid phase separation occurs in ternary systems with an appropriate cosolvent such that the addition of the third component extracts the cosolvent and makes the other component insoluble. Such droplets are homogeneously sized, stable, and require minimal energy to disperse compared to conventional emulsification methods. Thus, ouzo precipitation processes are an attractive, straightforward, and energy-efficient technique for preparing dispersions, especially those made on an industrial scale. While this process and the resulting emulsions have been studied by numerous indirect techniques (e.g., X-ray and light scattering), direct observation of such structures and their formation at the nanoscale has remained elusive. Here, we employed the nascent technique of LPTEM to simultaneously evaluate droplet growth and nanostructure. Observation of such emulsification and its rate dependence is a promising indication that similar LPTEM methodologies may be used to investigate emulsion formation and kinetics.


II. Sample Preparation
Solutions of trans-anethole in ethanol were prepared and stored for no more than 24 hours before use so as to ensure sample integrity. All formulations are given in volume percentages. To preform the droplets, an aliquot of this solution was chilled, and DI water was added dropwise until the solution was homogeneously cloudy. The same protocol was followed for the N, Ndimethylaniline solutions.

III. Liquid Cell Assembly
LPTEM experiments were performed using the Hummingbird Scientific Dual Flow Mixing holder as previously described. 1 Briefly, a non-glow discharged SiN x chip was seated in the holder tip,

S4
and an 0.8μL droplet of the sample was dropcast onto it. The top SiN x chip was placed such that the rectangular windows were oriented orthogonally (Figure S1), and the top plate and clamp were used to seal the cell. The liquid cell's integrity was verified using an external pumping station fitted with an optical microscope, such that the windows could be visually inspected for fracture and to ensure they could withstand the vacuum of the microscope (8.6 x 10 -6 mbar). The microfluidic lines were left unfilled but sealed during this test.

IV. Fluorescence and Optical Microscopy
Brightfield optical micrographs were taken on a Nikon Ti-U inverted microscope using an Imaging Source 23UX249 color camera. Nikon Plan Fluor 100x/1.30 Oil objective was used to image the samples. Differential interference contrast was used. Fluorescence images were taken by a Zeiss Axio Observer inverted microscope with Zen pro software and an Axiocam 503 mono camera. A Zeiss Plan-APOCHROMAT 63x/1.4 oil objective was used to observe the droplets and Nile Red, the fluorescent dye used for dying trans-anethole droplets, was excited by Colibri 7 LED light. A 91 HE CFP/YFP/mCherry filter was used to give off excitation from 494 -528 nm and receive emission from 546 -564 nm. The apparatus for containing the solution is as shown below ( Figure   S2). Adding 3 drops (~150 μL) of DI water to the 1 mL trans-anethole in ethanol solution would induce formation of trans-anethole droplets. To observe the droplets on an inverted microscope, 1 mL of trans-anethol / ethanol solution was transferred to a coverslip-bottom dish to which 3 drops of DI water is subsequently added.  Figure S1), there is some exterior volume that must be filled before mixing can occur.
Preliminary emulsification experiments were carried out with surfactant-loaded oil phases which were imaged prior to flowing in water. In such experiments, anisotropic appearance of the dispersed phase from the port of origin confirms these calculations and demonstrates the diffusivedriven nature of the in situ transport ( Figure S3).

Figure S3
Time series of representative micrographs of in situ experiment flowing water into sample of AOT dissolved in isooctane.
(A) Initial image of cell prior to beginning flow of water (t=0 min), demonstrating homogeneous and structure-less morphology.
Water flowed at 5 µl/min resulted in appearance of structures in one corner (B), but not another (C) after 11 minutes of dilution.

VI. Microscope and Imaging Conditions
A JEM-ARM300F (JEOL Ltd., Tokyo, Japan) transmission electron microscope operating at a voltage of 300keV and a current of 15μA was utilized to perform the in situ experiments. Images were acquired with a Gatan 2k x 2k OneView-IS CMOS camera (Gatan Inc., Pleasanton, CA,

S8
USA) at an exposure of 0.5 s via Gatan Digital Micrograph imaging software (Roper Technologies, Sarasota, FL, USA). Electron fluxes were calculated using the measured beam current, which was calibrated previously with a Faraday Holder using the relevant apertures.

VII. Image Analysis
Image processing and analysis was done in ImageJ. Acquired images were rebinned to a dimension of 1024x1024 pixels and background adjusted by subtracting a Gaussian blurred duplicate. Videos were created by aligning and stacking images of the same field of view. Droplets were measured manually, as the contrast was insufficient to yield reliable automated thresholding.

VIII. μFTIR Analysis
In order to verify the molecular integrity of the trans-anethole after imaging, post-mortem μFTIR was performed. A Bruker MicroFTIR was used to analyze the windows of the SiN x chips after the conclusion of the in situ experiment. The chips were removed from the holder, pried apart, and allowed to dry prior to analysis. Given the thickness of the SiN x chips and windows, it was necessary to run these experiments in reflectance mode, which also required 500 scans in order to generate signal with sufficient intensity. 2cm wavelength resolution was used. As a control, the same trans-anethole solution was dropcast onto a SiNx chip and allowed to dry, so as to serve as a representative spectrum of the material without imaging. High flux imaging conditions yielded visible damage in addition to spectra inconsistent with unimaged regions and controls, corroborating the low flux conditions used as safe for the material ( Figure S4. (A) Initial image of cell prior to beginning flow of water (t=0min), demonstrating homogeneous and structure-less morphology.
Water flowed at 5 μl/min resulted in appearance of structures in one corner (B), but not another (C) after 11 minutes of dilution

IX. Droplet Growth Data and Discussion
Using the image analysis techniques outlined above, we analyzed droplet growth data from emulsification experiments across a range of concentrations (5, 10, and 20 v.% trans-anethole) and flow rates (1 or 3 μL/min) ( Figure S5) in order to evaluate differences in nucleation and growth rates.
Growth rates were determined via logistic fit in Prism software, and k values (or logistic growth rates) are compared below (Table S1).
It was anticipated that slower flow rates would result in slower growth rates and a greater, more homogeneous droplet population than a higher flow rate, resulting from the increased equilibration time. A weak concentration dependence was observed, but the high standard deviations for these measurements rendered them statistically insignificant. However, one way ANOVA indicated significant differences in k values between flow rates ( Figure S6, S7). The primary variation observed was droplet number as a function of trans-anethole concentration ( Figure S66). Both linear and logarithmic growth fits were considered for both droplet diameter and the cube of the droplet radius (Error! Reference source not found.). Here, we see that the logistic fit has both a higher R 2 and lower standard error, indicating it is a better fit than the linear model. Additionally, we see that the r 3 plot is not well fit by a linear curve, which suggests that these droplet growth kinetics are not consistent with traditional models of Ostwald ripening   One way ANOVA indicated that there is significant difference in growth constants as a function of flow rate (p<0.001) ( Figure S7).