Hierarchical stabilization of emulsions with multi- scale interconnected droplets and ultra-low nanoparticle loadings


 Pickering stabilization by colloidal particles is a common strategy to disperse droplets of one fluid into another fluid in food, cosmetics and chemical industries1-3. For over a century, this kind of stabilization has been governed by constant surface coverage concepts in which particles irreversibly attach to the fluid–fluid interface. The need to cover sufficient interfacial area to prevent coalescence typically results in large loadings of particles, uniform droplet size, creation of rigid interface and closed-cell structure with small total area4-7. Here we report a stabilization mechanism that yields hierarchically structured oil-in-brine emulsions with high interfacial area, deformability, connectivity and long-term stability at unprecedentedly low nanoparticle loadings. The hierarchy in structure is achieved via dynamic cation-particle-droplet interactions in cascaded emulsification, which consists of i) formation of submicron oil droplets (~250 nm) lightly covered by hydrophilic polymer-coated iron oxide nanoparticles and polyvalent metal ions; ii) spontaneous formation of small droplets of nonpolar oil (~1 μm) stabilized by the nanodroplets and cations and iii) attachment of nanodroplet/small droplet clusters to bridge large unarmoured oil droplets (5-50 μm) in macroemulsions. This new mode of stabilization enables much more efficient use of nanoparticles, stabilizing a given size macroemulsion droplet at an order of magnitude smaller particle loading. Moreover, particle loading decreases with the 5/3 power of droplet size, rather than the first power typical of Pickering emulsions. Finally, cations play a novel and essential role in this mechanism, which cannot be accommodated in the conventional Pickering model. Our approach provides a new pathway for templating materials with better control over the structure, and for exploiting applications that are currently inaccessible for Pickering and surfactant stabilized emulsions.

Biological organisms have evolved to produce a variety of multi-scale hierarchically structured composites (i.e. wood, bone, shell, or grass). Inspired by such natural processes, biomimetic fabrication can lead to synthesis of unique materials with the possibility of multi-scale functionalization for tissue engineering, separation, catalysis, sensing, energy storage and convection [8][9][10] . Emulsions with controlled strength, droplet sizes and connectivity are candidate soft templates for synthesis of hierarchically porous materials 9,11 . Pickering emulsions stabilized by solid particles have advantages over emulsions stabilized by surfactants because solid particles have low toxicity and larger desorption energies than surfactants and provide a robust mechanical interfacial barrier to droplet coalescence 4,12 . However, current emulsion template requires a high volume fraction (>0.7) of the dispersed phase and high concentration of stabilizers to form crowded droplets trapped in three-dimensional space to induce the interconnectivity 6,9,11,[13][14][15] . For a given particle-fluid-fluid system, the surface coverage of droplets is often invariant, resulting in an inverse relationship between the diameter of emulsion droplet and the loading of Pickering particles 16 : where Re and Rp are the radius of emulsion droplets and particles respectively and φd and φp are the volume fraction of dispersed phase and particles respectively. Thus a given particle loading φp gives a characteristic droplet size, and decreasing loadings lead to larger droplets, eventually too 3 large to remain stable 17,18 . It has long been a major question whether the emulsions can be stabilized with much less particles without sacrificing mechanical properties. Although as little as 5% surface coverage has been reported in one case if particles form monolayer bridges between droplets 19 , it is well-established that particle should cover 23-91% of droplet surface area for steric stabilization 20 . The large loadings of particles at the rigid interface not only reduces deformability of emulsion droplets, but also affects forming pore throats during templated synthesis of hierarchically porous materials 5,6,9,11,[13][14][15]21 .
We report the fabrication of hierarchically structured oil-in-water emulsions at nanoparticle loadings significantly lower than previously reported thresholds for Pickering emulsions. The mechanism of stabilization differs fundamentally from the classical Pickering concept, resulting in scaling relation different from Eq (1) and 2.6 to 12 times less droplet surface coverage than minimum required by the bridging or steric hindrance mechanisms. The stabilization was achieved through the cascaded emulsification ( Fig. 1). At the first level, poly (4-styrenesulfonic acid-comaleic acid)-coated iron oxide nanoparticles (PIONPs) in brine (CaCl2) stabilize submicron droplets, ca. 250 nm in diameter. In the second level, these submicron droplets in combination with polyvalent ions (e.g., Ca 2+ , Co 2+ , Ni 2+ , and Fe 3+ ) spontaneously form macroemulsion small droplets, ca. 1.3 µm in diameter. Finally, these small droplets, in turn, bridge large droplets  µm), forming interconnected hierarchically structured oil-in-brine macroemulsion with droplet size from submicron to tens of microns. The hierarchy greatly increase the surface-to-volume ratio, roughness, strength, viscoelasticity and deformability of emulsion droplets, providing an ideal template for the synthesis of hierarchically structured materials. Neither the large droplets nor the small droplets in macroemulsion require any coverage by individual nanoparticles to remain stable.
Because remarkably few submicron droplets are sufficient, the overall requirement for nanoparticles is much lower than in conventional Pickering emulsions with the same distribution of large droplets.

Metal ion induced stabilization of submicron oil droplets by PIONPs
The hydrophilic PIONPs in brine were chosen for the synthesis of submicron-sized droplets to exploit two advantages. First, transmission electron microscopy (TEM) and scanning electron microscopy (SEM) analyses show that PIONPs are quite monodispersed with a median Fe3O4 core 4 size of 11.0 ± 2.2 nm and a core-shell diameter of 25.1± 2.2 nm (Extended Data Fig. 1a-e). They are negatively charged (-71 mV) and notably stable in high ionic strength solution of CaCl2 (Extended Data Table 1, Extended Data Fig. 1f). Second, these PIONPs can pin the oil-water interface after binding calcium ions through carboxylate group of surface polymers. Conversely, free sulfonic groups on PIONPs preserve sufficient negative charge after binding calcium ions to maintain repulsion between particles and dispersion stability (Extended Data Fig. 1f-i).
We made stable submicron-sized droplets by sonicating 15% (v/v) 1-octadecene oil and PIONPs of 5.1 wt % in 0.4 M CaCl2, based on a series of preliminary tests showing that only at the relatively high concentration (≥4.7 wt%) can PIONPs stabilize homogeneous submicron droplets (Extended Data Fig. 2). The oil-in-brine droplets (Fig. 2a) Fig. 3), and can be stored in brine for at least 35 d. Cryo-SEM analysis coupled with energy-dispersive X-ray spectroscopy spectra show that single oil droplet is surrounded by PIONPs at the oil-water interface and has a relatively higher concentration of Fe, S, Ca, C, and O compared with water background, which indicates that PIONPs and calcium ions serve as stabilizers for these tiny oil droplets (Fig. 2b, Extended Data Fig. 4).
As shown in the high-resolution atomic force microscopy (AFM) image acquired in a liquid cell (Fig. 2c), many PIONPs with diameters of ~25 nm are distributed on the surface of submicron droplet and cover ~26% of the droplet surface, consistent with the value of 33% calculated from mass ratios and projected particle area (Extended Data Table S2). PIONPs are not uniformly distributed on the oil-water interface but form disordered networks due to the capillary forces that are particularly pronounced for rough nanoparticles, as confirmed by the liquid-cell TEM image (Fig. 2d). The oxygen-edge scanning transmission X-ray microscopy (STXM) image (Fig. 2e) distinguishes oxygen-rich water from oxygen-free oil while the iron-edge STXM confirms high concentrations and non-uniform distribution of iron atoms at the oil-water interface (Fig. 2f).
These observations confirm that the first level of the emulsification is different from the principle of wettability-driven particle adhesion, but these submicron droplets at a coverage typical of Pickering emulsions 20 .
Hierarchical stabilization of oil-in-brine macroemulsions by submicron droplets 5 The submicron droplets are less hydrophilic than PIONPs (Extended Data Fig. 5) and have higher affinity for newly created oil/brine interface in the second emulsification step. As shown in the liquid cell SEM images (Fig. 3a), the apparent size of submicron droplets (bright spots) increased with time to form small micron-sized droplets (~1 µm) finally around big oil droplets. Liquid cell STXM and LSCM images also clearly show that the submicron droplets tend to assemble at the oil-brine interface to form a small micron-sized droplet structure, which is comprised of several submicron droplets arranged in a ring surrounding a small volume of oil ( Fig. 3b-d, Extended Data   Fig. 6). These observations provide consistent evidence that the energy barrier creating small droplets is quite low so that the submicron droplets preferentially stabilize such kind of structure due to the strong affinity of the nanoparticles for droplet surfaces at high ionic strength.
We then prepared a macroemulsion through high-shear homogenization of equal parts of 1-  (Fig. 3k, l), further prove that PIONPs serve as bridges between submicron droplets, small droplets, and large droplets. Despite sparse interfacial coverage, our results demonstrate that hierarchy is achieved by a bridging mechanism where submicron droplets on the surface of small droplets in the contact region form a dense, bridging layer as a result of strong capillary attraction caused by the menisci around them.

Effect of metal ions and oil polarity in the hierarchical stabilization mode
The presence of calcium ions on the surface of submicron droplets (Extended Data Fig. 1i, Extended Data Fig. 4) and macroemulsion small and large droplets (Fig. 3m), suggests that these 6 ions contribute to emulsion stability. To test this hypothesis, we used a porous microflow model connected with LSCM for the real-time observation of brine/oil/water interaction, and Ca 2+specific and Na + -specific fluorescence indicators of Fluo-3 and SBFI were used for detecting Ca 2+ and Na + at the oil-water interface, respectively. As shown in LSCM images, a Ca 2+ -specific fluorescence signal is strongly correlated with the location of the nonpolar oil (1-octadecene)water interface (Fig. 4a), but the signal could not be detected at the interface between water and a more polar oil, here ethyl acetate (Fig. 4b). This is consistent with the observation that the submicron droplets (0.1 wt% PIONPs) in 0.4 M CaCl2 do not stabilize ethyl acetate macroemulsions, but stabilize nonpolar oil (e.g., toluene, decane, dodecane, or 1-octadecene) macroemulsions (Extended Data Fig. 9a, b). Moreover, the submicron droplets with very low (0.  Fig. 10a, j). Clearly, the interaction among polyvalent metal ions, oil/water interfaces, and PIONPs contributes to network formation and capillary forces between droplets and droplet aggregates, which arrest the Ostwald ripening from submicron to micron droplets.

Effect of particle loading in the hierarchical stabilization mode
As shown in photographs acquired 38 days after the preparation (Extended Data Fig. 11), most oilin-brine macroemulsions prepared by low (0.01-0.1 wt%) concentration of PIONPs in the presence of 0.4 M CaCl2 showed significant coalescence, whereas oil-in-brine macroemulsions emulsified by submicron droplets with the same concentration of PIONPs showed stability. No coalescence and significant change in size distribution and Sauter diameter of oil droplets were apparent in macroemulsions prepared by submicron droplets even at very low (0.02-0.04 wt%) concentrations of PIONPs among 2, 11, and 38 days (Extended Data Fig. 12). This observation highlights that 7 submicron droplets in brine are more effective, in terms of such low concentration of nanoparticles, at stabilizing macroemulsions than PIONPs dispersed in brine.

Comparison of particle stabilizing efficiency in hierarchical and classic Pickering stabilization mode
As explained above, the surface coverage of the submicron droplets ( Fig. 4i, black symbol) reported here is comparable to many Pickering stabilizers (orange symbol), but the equivalent surface coverage by nanoparticles in hierarchically structured emulsions (green symbols) is an order of magnitude smaller. Moreover, the equivalent surface coverage decreases as the particle loading decreases. At 0.02 wt% particle loadings, the submicron droplet-stabilized droplets are 16 times smaller than the nominal Pickering-stabilized droplet. The multi-step emulsification and bridging effect produce a different scaling law between particle loading and droplet size. The black line in Fig. 4j is the result of using Eq (1) to predict Pickering behavior from the PIONP loading on the submicron droplets. The power-law exponent of -1 is observed for typical Pickering emulsions, orange symbols. In the hierarchical emulsions studied here nanoparticle loading (green symbols) and submicron droplet loading (magenta symbols) scales with the -5/3 power of droplet size, whereas small droplet loading (blue symbols) scales with the -1 power of droplet size. This further proves that the small droplets serve as final Pickering stabilizers for large oil droplets in macroemulsion. 8 Our quantification of droplet surface coverage (Extended Data Tables 2-4), indicates that the ability of submicron droplets to bridge multiple small and large droplets underlies the greater efficiency. At particle loadings of 0.02-0.15 wt%, the combined surface area of small and large droplets is 7 to 16 times greater than that of the Pickering stabilized submicron droplets, so that in effect the PIONPs are 8 to 17 times more efficient to stabilize an order of magnitude more droplet surface area at an effective coverage of 1.9% to 4.0%. The lowest surface coverage at particle loadings of 0.02 wt% is nearly 3 times below the calculated minimal surface coverage of 5.5% required by the bridging mechanism of Pickering stabilization 19 . It is also 26 to 40 times less than that of the other reporting silica 23 and polymer-coated silica 24 nanoparticles for stabilizing macroemulsions by steric hindrance (Extended Data Table 5). The much lower surface coverage, whether calculated by submicron droplets or PIONPs, is thus a consequence of a novel mode of stabilization.

Comparison of emulsion rheology in hierarchical, Pickering, and surfactant stabilization mode
Amplitude sweep measurements demonstrate the strong dependency of macroemulsion viscoelastic signature on solution ionic strength and type and effective PIONP loadings (Extended Data Figs. 14-16). The maximum zero shear storage modulus (G0') and cross over frequency (indicating highest yield stress) were achieved at particle loadings of 0.04 wt% PIONPs in 0.4 M NiCl2. Its G0 normalized by the Laplace pressure is 1.7 and 4.1 times larger than Pickering oil-inwater (1.5 wt% SiO2, ~20 nm) and previously oil-in-brine emulsions (1 wt% SiO2) 25 , respectively.
The creep/recovery signature of macroemulsions confirm these results (Extended Data Fig. 18,   Fig. 4l). During the creep stage, the 0.04 wt % particle loaded hierarchical emulsion in 0.4 M NiCl2 showed very strong deformability as the compliance of hierarchically structured emulsion only reached 0.005 Pa -1 at a shear (3 Pa) of 300 s, whereas SiO2 nanoparticle and SDS stabilized oil-inwater emulsions reached large compliance of 865 and 119 Pa -1 , respectively. After the stress released in 300 to 900s, the final recovery of the hierarchical emulsion is 62%. In contrast, the surfactant stabilized emulsion only can instantaneously recover 15% and the Pickering emulsion 9 nearly cannot be recovered due to its rigid interface and stiffness. These results mean that the viscoelastic properties of hierarchical emulsions can be tuned with polyvalent metal ions and particle concentrations to achieve higher strength and deformability, although the particle loadings are much lower than Pickering and surfactant stabilized emulsions.

Conclusion
Our results conclusively reveal a new hierarchical stabilization mechanism for emulsions. Unlike a Pickering emulsion generated from dispersed nanoparticles and stabilized exclusively via droplet armouring or bridging 20,27 , nanoparticles enter this system already attached to submicron droplets, where they remain during emulsion generation. The integral role of cations in these hierarchical structures also differs qualitatively from their usual influence on stabilization. Their contribution to particle/droplet attachment is independent of classical wettability-driven particle adhesion. This combined with the subsequent spontaneous formation of small droplets greatly enhances the particle's intrinsic interfacial activity, covered surface area, and roughness. This enables more stable macroemulsions formed at ultra-low particle loadings, for which stable Pickering emulsions would be impossible, thereby reducing the amount, cost, and environmental impact of emulsifiers in industrial applications. Moreover, the novel hierarchically structured emulsion integrates the advantages of Pickering and surfactant stabilized emulsions. Due to its ultra-low nanoparticle loadings, this kind of emulsion realizes better control of microstructure and simultaneously achieves high connectivity, deformability and strength. This in turn enables templated synthesis of highly permeable macroporous materials for drug delivery, enhanced in situ or in vivo imaging, and catalysis.

Synthesis of polymer-coated iron oxide nanoparticles
The monodispersed PIONPs were synthesized by sonicating polymer of poly (4styrenesulfonic acid-co-maleic acid) sodium salt (20 kD, Sigma-Aldrich) and iron oxide (Fe3O4) nanoparticles (US Research Nanomaterials, Inc.). A total of 300 mL of MiliQ water was poured into a 400 mL clean beaker, and 12 g polymer was dissolved in water. The pH of the polymer solution was adjusted to 5 by adding concentrated HCl. Then, 3 g of iron oxide nanoparticles were added in solution followed by ultrasonication for 60 min at 50% amplitude in an ice bath. To remove non-coated nanoparticles, after sonication, the solution was centrifuged for 20 min at 4,000 g. Finally, the supernatant was processed by five steps of concentrating and washing with DI water using 100 kD centrifugal filter units (Amicon®Ultra-12, 100 kDa) to remove free polymers. After several centrifuging steps, the PIONPs yield ranged from 20% to 30%.

Structural characterization of PIONPs
The iron oxide content of PIONPs were determined using an inductively coupled optical emission spectrometer (ICP-OES, Avio 200 ICP Optical Emission Spectrometer, PerkinElmer) to measure Fe content in the core. For the ICP-OES measurement, the Fe standard sample was diluted to 0, 1, 2, 4, 8, and 10 ppm, and the PIONPs were dissolved by 5 v/v % hydrochloric acid and diluted to the concentration lower than 10 ppm before measurement. The weight of polymer shell was determined by the whole dry weight of sample minus the weight of Fe3O4 core. The weight ratio of iron oxide to polymer is approximately 1:4.
The core size and shape of PIONPs were imaged using a transmission electron microscopy (TEM, JEOL JEM-2100, Japan) instrument with a 200 keV acceleration voltage and the whole particle size and shape were imaged by a scanning electron microscopy (Zeiss Auriga Compact FIB-SEM) with a 5 KeV acceleration voltage. A dilute aqueous solution of PIONPs was deposited onto an ultrathin carbon-coated copper grid (200 mesh). The size distribution was determined by the ImageJ analysis of more than 800 particles.
The colloidal stability was tested by dynamic light scattering (DLS) and zeta potential measurements at room temperature (25°C) using a DynaPro NanoStar instrument (Wyatt 20 Technology Corporation. USA) and a Malvern Zetasizer Nano ZSP instrument. To evaluate the colloidal stability of PIONPs, the stock solutions of 5 M CaCl2 and PIONPs were diluted to achieve a concentration of 0.1 wt% PIONPs in 0-1 M CaCl2, and the particle sizes in water and brine were determined in the intensity-weighted mode.
The FTIR spectra of polymers at CaCl2 were obtained with a Bruker Vertex 70v spectrometer. 4 wt% polymers were placed in water or brine (0.4 M CaCl2) for 24 hours. Then these samples were frozen-dried and added to a pressed KBr disk to form a thin film. The disk was quickly placed into the sample chamber for FTIR analysis with a 2 cm −1 resolution in the spectral range of 400-4000 cm −1 . The FTIR spectra were collected using an average of 300 scans. The system was rapidly evacuated, and a background spectrum of the KBr disc without a sample was collected under the same instrumental conditions. All spectra were corrected for the background.

Synthesis and colloidal stability of submicron droplets
Oil-in-brine submicron droplets were synthesized using a 3:17 oil:brine volume ratio. The

Hydrophilicity characterization of PIONPs, calcium ion bound PIONPs and submicron droplets
The hydrophilicity of glass cover slips covered with PIONPs, calcium ion bound PIONPs and submicron droplets were performed by a data physics optical contact angle measuring system (OCA 15EC). The Ca 2+ bound PIONPs were prepared by incubation PIONPs in 0.4 M CaCl2 for 24 h and then dialyzed in 1000 times volume of deionized water to remove calcium ions. The submicron droplets were also dialyzed in deionized water to remove calcium ions after preparation.
Then the concentrated PIONPs, Ca 2+ bound PIONPs, and submicron droplets were added to form 21 flat films on the cover slips. A 5 µL drop of water was added on each cover slip, and the contact angle was recorded quickly in 5 miniutes by the contact angle measuring system.

Stability of macroemulsions
The stability of macroemulsions was evaluated by visual and microscope observation and rheology measurements. The visual observation of macroemulsions was recorded during 38 d settling at room temperature, and the destabilization of the emulsion was evaluated by change of color. Once nanoparticles were detached from the oil-water interface and droplets of the oil phase were broken, the upper layer of oil could be observed and the color would become much more yellow-brown. An Olympus upright microscope (Olympus BX-51) was used to capture optical micrographs of macroemulsion oil large droplets. To ensure the structural integrity of the droplets and ensure accurate size analysis, each macroemulsion was loaded into a bacterium counting plate composed of a liquid cell separated by a slide and a cover glass. About 300-500 droplets of each macroemulsion were analyzed in ImageJ software for size analysis. The Sauter average droplet diameter, D32, can be acquired as the sum of all the oil droplets in the emulsion: where Di is the measured droplet diameter and Ni is the number of droplets.
An oscillatory rheometer (Anton Paar MCR-302) was used to determine the rheological characteristics of the macroemulsion at ambient temperature using sandblasted parallel plate Where is the compliance of the system reached maximum deformation, ∞ is the compliance values at 600 s after the stress released.

Cryo-SEM imaging
The oil-water interfacial activities of PIONPs and submicron droplets were characterized by an environmental field emission SEM (Quanta TM 250 FEG) under the cryo-state. A small volume of submicron droplets after dialysis or freshly prepared macroemulsions was put into a rivet and quickly frozen in a nitrogen slush. When the samples were loaded into the SEM chamber, a thin layer of sputtered gold was deposited, and the images were acquired either by secondary electron or backscattered electron detectors at a 5 kV accelerating voltage to detect nanoparticles or submicron droplets at the oil-water interface. The energy-dispersive X-ray spectroscopy analysis was conducted after the overall workflow and optimization of the accelerating voltage.

Liquid cell AFM imaging
AFM (Keysight 5500 scanning probe microscope system) conducted in contact mode was used to scan the 3-D topography of submicron droplets and macroemulsion oil droplets. The specimen was prepared by adding 100 µL submicron droplets or macroemulsion droplets on a 23 cover glass and placed in a moist atmosphere for 10 min. Then, the cover glass was rinsed off with DI water to remove these droplets that did not adhere to the glass, quickly put in a liquid plastic cell, and secured with adhesive tape. A scanner carrying a long rectangular cantilever (PPP-CONT-50, NanoAndMore USA Corp.) with a spring constant of 0.2 N/m was placed above the specimen.
The scanning was conducted in DI water with a scanning rate of 0.5 Hz with a resolution of 512 × 512 pixels. The acquired AFM images were processed by supplied software to create 3D topographic images and calculate surface roughness.

Liquid cell TEM imaging
A TEM (JEOL JEM-2100) was used to characterize the microstructure of submicron droplet in liquid using a microfluidic sample chamber (In situ Technology), which is constituted by two ultrathin (~20 nm) electron translucent SiNx membranes separated by ~100 nm with spacers. A total of 1 µL of submicron droplet solution was added to one end of the chip and pumped through a syringe to allow droplets to enter the chip chamber. Finally, the chip was installed on the conventional electron microscope rod, and imaging was conducted with a low voltage of 120 kV to prevent submicron droplets from electron damage.

Liquid cell STXM imaging
The STXM measurements of submicron droplets, submicron droplet at the oil-water interface, and macroemulsions were conducted at the SM (10ID-1) beamline of the Canadian Light Source using soft X-rays generated from the 2.9 GeV synchrotron storage ring. A fully hydrated sample of submicron droplets or macroemulsions was prepared as a thin layer in a wet specimen cell that comprised two ultrathin (~100 nm) silicon nitride windows. The wet cell was sealed with epoxy glue, which prevented the sample from drying out. The oxygen map was directly obtained by STXM at the photon energy of 540 eV. The iron-edge STXM imaging was acquired from strong and weak absorption of Fe at the photon energies of 710 and 700 eV, respectively. Then, the X- with DI water, a calcium ion-specific fluorescence indicator of Fluo-3 was used to detect whether there were calcium ions at the oil-water interface. To probe Na + /oil/water interaction, a fluorescent 25 indicator of SBFI was used to specifically detect sodium ions at the oil-water interface in the microfluidic channels.

In situ liquid cell SEM imaging
An environmental field emission SEM (Quanta TM 250 FEG) operated at 20 kV was used for the in situ imaging of PND/brine/oil interactions. The in situ technology was achieved through a microfluidic liquid cell (FlowView Aquarius) constituted by an ultrathin (~20 nm) electron translucent SiNx membrane and a solid microchannel substrate separated by ~500 nm with spacers.
A total of 0.5 µL of 1-octadecene oil was loaded into the solid microchannel substrate, and the oil was left for 2 min to allow the oil to slowly diffuse into droplets and adhere to the substrate. Then, 0.5 µL of submicron droplets (0.10 wt%) in brine solution was added, and the cell was quickly closed with a translucent SiNx membrane. The SEM images were obtained at the rate of one frame per 30 µs immediately after the sample was inserted into the SEM chamber and focused on by the Everhart-Thornley detector.

Quantification of PIONP surface coverage on submicron droplets
The quantification of surface coverage on submicron droplets carried according to Dr.
Tiantian Zhang's master thesis 24 . It was assumed that all the droplets formed are spherical, and the quantity of each component in the mixture before and after sonication/homogenization should be conserved, neglecting the loss of experimental observation. The total interfacial area of the submicron droplets was calculated using the data from emulsion droplet size analysis through dynamic light scattering. As the median radius of submicron droplets (Rsmd) was calculated through intensity-weighted DHs, then the total interfacial area of submicron droplets is where Nsmd is the number of submicron droplets in the emulsions and ������� = 4 < 2 > is the average interfacial area. After obtaining the total interfacial area, we calculated the fraction of the total submicron droplet interfacial area covered by PIONPs. The initial mass (Mi) and average radius (Rp) of nanoparticles in the mixture were known. The density (ρ) of PIONPs could be 26 calculated as 2206 Kg/m 3 through ICP-OES analysis of Fe content in the core and polymer to iron oxide weight ratio, then the number of nanoparticles is where ���� is the average volume of nanoparticles. Then the fraction of submicron droplet surface covered with nanoparticles is where ����� = < 2 > is the average covered area of PIONPs. Then the nanoparticle stabilizing efficiency for submicron droplets is

Quantification of PIONP surface coverage on the whole interface of macroemulsions
Three were three different emulsion droplets of submicron droplets, small droplets, and large droplets in the macroemulsion system. As the added oil volume ( ) and the number of submicron droplets ( ) are known. Thus the whole oil volume is Where is the oil volume of submicron droplets added into the macroemulsion system, is the oil volume added with water to make macroemulsions, is the fraction of oil volume forming small droplets, is the oil volume fraction of becoming large droplets. The average radius of small droplets (Rsd) could be estimated by LSCM imaging of the interface between large droplets (Fig. 3g) and the average radius of large droplets (Rld) could be acquired by calculating the Sauter diameter of droplets prepared by different concentration of submicron droplets ( Fig. 4c-g, Extended Data Fig. 13). Then the average interfacial area of submicron droplet is Then the average interfacial area of macroemulsion small droplets is The average interfacial area of large droplets is Because there is nearly the same surface small droplet coverage ( ) between macroemulsion large droplets prepared by submicron droplets of 0.04 w/v% and 0.1 wt% PIONPs (( Fig. 3h, j, Fig. 4h), we assumed that all the macroemulsion large droplets prepared by submicron droplets with low (0.02-0.15 wt%) concentration of PIONPs have the same of 21%. Then the number ratio (Nsd/Nld) is The ratio of Vosd/Vold is The oil volume of small droplets is The oil volume of large droplets is The total number of small droplets is The total number of large droplets is The total number of submicron droplets is The total interfacial area of small droplets is The total interfacial area of large droplets is The total interfacial area of submicron droplets is Where is the number of submicron droplets added for preparing macroemulsions. Then the fraction of small droplet surface covered with nanoparticles is The fraction of macroemulsion large droplet surface covered with nanoparticles is The total interfacial area of macroemulsion system is Then the fraction of macroemulsion droplet surfaces covered with nanoparticles is The nanoparticle efficiency in the whole macroemulsion system is Quantification of submicron droplet surface coverage on the whole interface of macroemulsion ICP-OES measurements of aqueous phase of macroemulsion proved that all the PNDs jumped to oil-brine interface of small droplets and large droplets, because there were no residual PIONPs in the aqueous phase after preparation of macroemulsion with particle loadings (0.02-0.15 wt%). It was assumed that the submicron droplets like nanoparticles to form monolayers with hexagonal close-packed pattern. Then the average covered area of submicron droplets is The fraction of small droplet surface covered with submicron droplets is The fraction of large droplet surface covered with submicron droplet is The total interfacial area of macroemulsion system is Then the fraction of macroemulsion droplet surfaces covered with submicron droplets is The submicron droplet stabilizing efficiency in the whole macroemulsion system is