Experimental and numerical investigation of the formation of Oil Particle Aggregates (OPA)

The impacts of oil spills have been well recognized and studied for decades. Response measures vary from mechanical removal of the oil to bioremediation, to chemical dispersants (NRC, 2005). Oceans and open water bodies contain particles of different chemical composition (organic, inorganic) that are expected to interact with the oil forming aggregates (Guyomarch et al., 2002; Lee, 2002; Owens and Lee, 2003). The formation of oil-particle aggregates (OPA) changes the buoyancy of the oil droplet, of which the formed OPA becomes heavier than water or near-neutrally buoyant; thus, can be transported easily and eventually settled in the form of OPA (Bragg and Owens, 1995). Formation of OPAs also enhances oil biodegradation and dissolution of hydrocarbons due to the fact that oil-water interfacial area is maintained by the layer of particles on the droplet surface to prevent oil re-coalescence. Understanding and quantifying the formation and the fate of OPA is essential for developing oil spill countermeasures, especially for oil in the nearshore zone (Lee, 2002; Owens and Lee, 2003).

OPAs can be grouped into two major types based on the oil-particle interactions and the structure of oil phase and particles in the aggregates (Zhao et al., 2016): Type I Oil droplets coated by small particles resulting in the so-called pickering emulsion (Frelichowska et al., 2010), which has also been referred to as droplet aggregate (Stoffyn-Egli and Lee, 2002) and Type II, oil covers or bridges large particles (Stoffyn-Egli and Lee, 2002; Waterman and Garcia, 2015). Type I is the most abundant and results in the larger volume of trapped oil, and it is the most common type found in laboratory and field studies (Le Floch et al., 2002; Lee et al., 2003). Particles on the oil droplet surface stabilize the droplet and prevent re-coalescence; thus it is more stable in the marine environment. Very few studies exist on Type II OPA. The structure of type II is complex and the stability of type II OPAs could be less than type I due to the fact that oil exposes to the surroundings rather than armored by particles.

Three phases are involved in the formation of OPAs: oil, particles and a third medium (normally water) and the mechanism of OPA formation are not well understood. Experimental studies have tested different environmental parameters on the effects of the kinetics of OPA formation. The primary factors influence the formation include particle properties (e.g. size and shape, concentration, surface chemistry) (Ajijolaiya et al., 2006; Zhang et al., 2010); oil characteristics (e.g. viscosity, composition, size) (Khelifa et al., 2002; Omotoso et al., 2002); mixing energy in the system (Omotoso et al., 2002; Sun et al., 2010); salinity and temperature (Cloutier et al., 2002; Khelifa et al., 2002; Le Floch et al., 2002; Lee et al., 2002); and the presence of chemical dispersant (Guyomarch et al., 2002).

Predictive models are in need of development for oil-particle interactions. The population balance equation is the most widely used model for the simulation of oil-particle aggregation. Sterling et al. (2004) used a generic population model that includes a simplified aggregation module to predict the oil-particle aggregation process with consideration of density differences of the formed aggregates. They assumed that the oil-particle coagulation efficiency is constant, depending only on the initial physical properties of oil and particles. Bandara et al. (2011) used a similar approach to Sterling et al. (2004) to account for oil-particle interactions in their model. Hill et al. (2002) predicted the equilibrium time for OPA formation using a formulation based on a simplified population balance equation. Khelifa et al. (2005) proposed a model to predict oil-particle aggregation using Monte Carlo method. These models are expedient, but they are not sufficiently mechanistic to be predictive. Zhao et al. (2016) recently developed a predictive model, A-DROP, for the OPA formation. A new conceptual formulation of oil–particle coagulation efficiency is introduced to account for the effects of oil stabilization by particles, particle hydrophobicity, and oil–particle size ratio on OPA formation. The model has the ability to simulate the oil-particle interactions in the nearshore environment.

The objective of this study is to investigate the oil-particle interactions experimentally and numerically. Laboratory experiments were conducted using EPA baffled flask. Trapped oil amount in different layers were obtained and 3D images of the formed OPAs were taken from confocal microscope to study the OPA structures. Results from the experiments were used for the validation of A-DROP model. Finally, the OPA formation in the nearshore environment was simulated.

The oil used in the experiments was Alaska North Slope crude oil (ANS) with a density of 866 kg/m³, viscosity of 11.5 cp, and oil-seawater interfacial tension of 20 mN/m. The clay mineral, kaolinite (Fisher Scientific) was used in this study. The particles were in irregular shapes and many of them appear to be platy (Figure 1), with a density of 2630 kg/m³ and moderately hydrophilic (Delvigne et al., 1987; Shang et al., 2008). The median droplet size of the particles used in the current study is about 9 μm.

A 200 mL baffled trypsinizing flask was used during the experiments. A volume of 120 mL synthetic seawater was poured into the baffled flask (BF). Then 60±7 mg of oil was carefully dispensed onto the center of the water surface, giving an oil concentration of 500 mg/L. Following the addition of oil, the BF was placed on the orbital shaker (Thermo Scientific) and shaked for 1 hour with a speed of 200 rpm. Then, while it was shaking, 180±15 mg particles (1500 mg/L) were added into the BF, and the mixture was continued shaking at designed time (e.g. 10 min, 20 min, 3 hr, 24 hrs). Immediately after each mixing time, the mixture was transferred into the separatory funnel, which was allowed to settle overnight, and three layers were formed for each sample (Figure 2): surface layer with buoyant OPAs and free oil, middle layer with visually clear water, and bottom layer with negatively buoyant OPAs. All three layers were separated, along with the empty BF, for oil concentration measurement. The oil concentration was measured through the absorbance of oil solution in dichloromethane (DCM) using an Ultraviolet (UV)-Vis spectrophotometer (Cole Parmer) (procedures in Pan et al., 2016).

Separate experiments were conducted for microscopy observations with the same parameters and procedure before oil extraction. Samples were extracted from the bottom layer and placed on the microscopy slide for observation using a confocal laser scanning microscopy (Leica DM6000). Simultaneous excitation wavelengths of 488 and 638 nm were used in this study. The signal emitted in the range of 519–605 nm was recorded in the green channel to represent the fluorescent oil, whereas wavelengths of 607–672 nm in the red channel to present particles. The most intense signal in the red channel is the reflectance from the particles. Over 10 scans were acquired digitally in focal planes 0.8 μm apart, constructing a 3D image of the OPA scanned.

The A-DROP model was explained in detail in Zhao et al. (2016). For this reason, only a brief summary of the model is presented herein. Formation of OPAs is simulated based on population balance equation in A-DROP. Three different entities are used in the model: particles, oil droplets, and OPA. For a discrete particle size system, the coagulation of OPAs can be expressed as:

where “m” is the index of an OPA of equivalent diameter D_OPA,m; D_o and D_p represent the equivalent diameter of the oil droplet and particle, respectively; N is the number concentration (number/m³) of OPA of size D_OPA,m; “n” is the total number of size bins of particles; “k” is the total number of size bins of oil droplets; β is the collision frequency; α is the coagulation efficiency; and t is time (s). On the right-side of Eq. 1, the first term represents the birth of OPAs of size D_OPA,m from the coagulation of oil droplets and particles; the second term represents the birth of D_OPA,m from the coagulation of smaller OPAs with particles; and the third term represents the death of D_OPA,m which is the loss of OPAs of size D_OPA,m by coagulation with particles to generate OPA of size larger than D_OPA,m.

The collision frequency β is based on the mechanisms of turbulent shear, differential settling, and Brownian motion between the collided entities. Collision itself will not suffice for aggregation; thus, not every collision will result in successful aggregation. The ratio of the number of successful coagulations to the total number of collision events is described as the coagulation efficiency α. When α = 1, the formation is referred to as “fast” coagulation, where there is essentially no barrier to particle collision and every collision will result in coagulation. However, in reality, α is normally less than 1. The coagulation kernel for oil and particles is still unclear, which may be closely related to the surface chemistry of oil and particles. In A-DROP model, a conceptual formulation for α was established, which was based on the total surface area of oil droplets and the total projection area of particles during the formation of OPAs:

where α_sta is the stability ratio which was estimated based on free energy analysis, A_o is the surface area of a droplet, A_p-proj is the projection area of particles on the droplet surface when an OPA is formed, and F_SP is a factor to account for particle shape and packing effects on the coagulation process.

The mixing speed of 200 rpm was considered in the current study, which corresponds to an average energy dissipation rate of 0.67 W/kg (Figure 3). Using particle image velocimetry (PIV), Zhao et al. (2015) reported that there exists a region with −(5/3) slope in the energy spectra in the BF, which demonstrate the existence of the inertial subrange and the presence of turbulent flow. Delvigne and Sweeney (1988) reported that the energy dissipation rate varied between 10⁻³ and 10⁻² for surface layer and between 1 and 10 W/kg for breaking waves. The energy dissipation rates in natural surf zone varied sparsely in a range of 10⁻⁵ - 5 × 10⁻² W/kg based on different field observations (Bryan et al., 2003; Feddersen, 2012; George et al., 1994). The mixing speed we used may present the mixing energy in the near shore location close to the swash zone.

Figure 4 shows the 3D structure of an OPA seen under the confocal laser scanning microscopy. Most of the OPAs scanned have similar structures. The droplet diameter of the OPA is about 30 μm. Images in the left panel were seen from the transmitted light illumination, showing particles covered the entire surface of the oil droplets. To view the interior structure of the OPA, the OPA was sliced into 6 disks as shown in the middle and right panel. The particles are basically arranged on the perimeter of the oil surface, while the ones inside the oil shown in disk 1 appear to be the particles on the top layer (they did not penetrate into disk 2). Particles can be seen in different depths, and the deepest penetration is about 8 μm as shown in disk 2.

The variation of particle penetration is contradictory with the theory that the position of particles solely depends on the interfacial forces and particle surface chemistry (e.g. hydrophobicity) (Aveyard et al., 2003; Menon et al., 1987) . Based on such theory, for the same oil and particles with the same properties, position of particles relative to the oil surface should be the same. We believe that the hydrodynamic forces in the turbulent flow will build up pressure when the flying droplet and particles contact each other. Such pressure is large enough to make the platy particle stab into the droplet rather than lay flattened on the droplets as many other researchers argued (Ajijolaiya et al., 2006; Zhang et al., 2010). The forces will depend on the size and shape of droplet and particles, the moving speed, and the resistance forces from the entities, thus, resulting in different penetration depths of the surrounding particles into the oil droplet . Further investigation of such new theory is still in our on-going studies.

Figure 5 shows the percentage of oil trapped in negatively buoyant OPAs and on the water surface as buoyant OPAs and free oil. During the experiments, about 20–40% of oil was stuck on the BF wall. This portion of oil was not participated in the oil-particle interactions; thus they were subtracted from the calculation of the total oil. In all the experiments, the middle layer (clear water as shown in Figure 2) contained very smaller amount of oil (less than 4% of the total oil amount), while most of the oil was on the bottom and surface layers. The amount of negatively buoyant OPAs in the bottom (Figure 5a) increased sharply with shaking time (an indication of oil particle interaction time). Then they reached a plateau within about 20 min, which indicates an equilibrium condition has been reached. About 60–70% of oil became negatively buoyant due to the aggregation with heavier particles at steady state. Similarly, the oil on the surface (buoyant OPAs and free oil) decreased quickly with shaking time, and it reached the equilibrium at about the same shaking time as the negatively buoyant OPAs. The percentage of oil (free oil and oil in OPA forms) on the surface is about 30–40% of the total oil at steady state. Free oil can be seen at short shaking time period (e.g. 5 min, 10 min), however, free oil is barely seen by visual observations after 20 min and the oil on the surface was mostly buoyant OPAs (see Figure 2b).

Modeling results from A-DROP are also shown in Figure 5. The model configuration was based on the experiment setup and parameters used were closely resembled the experimental conditions. As shown in Figure 5, simulations match well with the experimental data. The sharp increase in the percentage of oil trapped in negatively buoyant OPA was captured by A-DROP, and the equilibrium has been reached around the same shaking time as the measurements. A logarithmic scale figure was also plotted in Figure 5a, showing that the model underestimate the oil trapping efficiency at early time period of oil-particle interactions. The early interactions may be highly rely on the stabbing mechanisms as discovered in the current study, because there are more surface area available for the interact between oil and particles at the beginning. Such mechanism has not been considered in the A-DROP model. With the establishment of the new theory based on the experimental observations, further improvement of A-DROP model will be conducted in the near future.

The correlation of the shape and packing factor F_SP is plotted in Figure 6 along with the data from Sun et al. (2010) and Khelifa et al. (2008) as reported by Zhao et al. (2016). The relationship is expressed as a function of particle concentration to oil concentration (C_p/C_o):

Though the current experimental conditions were not the same as the one from Sun et al. and Khelifa et al., the F_SP follows the relationship reported by Zhao et al. (2014), which confirms the reliability of the correlation. More experiments are still needed to test such correlation. Meanwhile, by correlating F_SP with different parameters through fitting to experimental data, it may provide more insights of the particle packing and shape effects on the OPA formation; thus, along with the experimental observations, we will further build physical properties of the coagulation efficiency and to improve the model.

In offshore and ocean environment, the sediment concentration close to the surface layer (e.g. < 1 m depth) could be low because of the higher water depth; thus, the interactions of particles and oil spilled on the surface might be minimum. When the spilled oil approaching to the nearshore region, where intense interactions between fluid and sediments cause sediment suspension, the oil-particle interactions will increase dramatically under different mixing intensity and sediment concentration, specifically under the breaking waves of the surf and in the swash zone. To address such process, simulations of OPA formation under a typical nearshore environment are presented. A based case configuration of a typical nearshore environment is shown in Table 1, which were obtained based on the field observations in different beaches in literatures (Zhao et al., 2016). The oil properties were assumed close to South Louisiana 2001 crude oil with a density of 820 kg/m³ and oil-seawater interfacial tension of 21 mN/m, while the particles with a density of 2600 kg/m³ and average size of 5 μm.

The result of OPA formation as a function of residence time is shown in Fig. 7. For the base case of C_p = 5,000 mg/L in swash zone and C_p = 200 mg/L in surf zone, all the free oil droplets become suspended OPAs ~ 4 min after the oil first approaches the Breaker Zone. Then, after another ~ 10 min, some of the suspended OPAs start to become negatively buoyant because of the continuous accumulation of particles on the oil droplets to increase the OPA density. This process is continued and the oil trapped in negatively buoyant OPAs increases with time. The increase slows down after 80 min and at ~110 min residence time, the OPA formation basically reached the equilibrium condition, where about 26% of oil is presented in the negatively buoyant OPAs. When doubles the sediment concentration in swash zone (Cp=10,000 mg/L) while keeping other parameters the same, the time for OPAs becoming negatively buoyant decreases to ~ 9 min (total ~ 14 min for the base case). The time to reach equilibrium condition is also shortened extensively from ~110 min to ~60 min. However, the oil trapping rate in the negatively buoyant OPAs is only 28% at the equilibrium condition (26% for the base case), which did not correspond with the same proportion increase as the sediment concentration. The case of increasing the sediment concentration in surf zone to 5 times of the value for the base case (1,000 mg/L) is also presented in Fig. 7. The OPA formation profile is basically the same as the base case. The oil trapping rate is only increased 0.1%. Results in Fig. 16 implies that addition of sediment particles in the swash zone would largely shorten the equilibrium time, speeding up the OPA formation process; however, it may not affect greatly on the overall amount of OPAs in the suspended and negatively buoyant phases. Addition of sediments in surf zone would not have too much effect on the OPA formation, even increase the sediment concentration to a high rate.

The oil-particle interactions were investigated experimentally and numerically in this study. The experiments were conducted using a laboratory baffled flask. Samples of the microsized OPAs were observed using confocal laser scanning microscopy to reveal the 3D structure of arrangement of particles in the droplet. The images indicate that the platy particles most likely stabbed into the droplet due to the hydrodynamic forces. This is the first evidence observed for the theory of particle stabbing. Further studies are still needed for a full methodical investigation. Oil concentrations in different layers were also reported in the current study, and results were used for the validation of the OPA formation model, A-DROP. The simulation results matched well with the measurements, but slightly underestimate the OPA formation at early contacts between oil and particles. This may suggest a need to include the stabbing mechanisms in the model development. Finally, simulations of OPA formation in a typical nearshore environment were conducted. Results indicate that the increase of particle concentration in swash zone would largely increase the oil-particle interaction process resulting in a shorter equilibrium time, but the oil amount trapped in the negatively buoyant OPAs did not correspond to the increase of particle concentration and only small increase in the oil trapping rate may be expected. The increase of particle concentration in surf zone basically did not have effects on the overall OPA formation even with a large increase rate of particle concentration. The developed A-DROP model could become an important tool in understanding the natural removal of oil and developing oil spill countermeasures by means of oil–particle aggregation.