Effects of Physical Properties and Dispersion Conditions on the Chemical Dispersion of Crude Oil

Abstract

Effects of oil type and mixing energy on the performance of chemical dispersants were investigated in a baffled flask mixing system using the commercial dispersant Corexit 9500. Effects of viscosity and interfacial tension were investigated by testing three crude oils (Arabian Light, Mars, and Lloyd) at two dispersant-to-oil ratios (DOR; 1:100 and 1:25) and three energy dissipation rates (0.00075, 0.016, and 0.16 W/kg). Dispersion effectiveness (i.e., fraction entrained as droplets in the water column) of all three oils was maximal at a mixing energy of 0.016 W/kg, which is similar to the energy dissipation rate in the surface layer of the open ocean. For Arabian Light and Mars, dispersion effectiveness was affected by DOR only at the lowest mixing energy, but it was proportional to DOR at all mixing energies for the more viscous oil, Lloyd. Droplet-size distributions were multimodal for all oil–dispersant combinations that were tested, indicating the involvement of multiple droplet-formation mechanisms. Diameter of mean volume of the major droplet-size modes was not sensitive to changes in mixing energy or DOR, but the fraction of dispersed oil in each mode was affected by these factors as well as by the oil type. Under all dispersion conditions (DOR and mixing energy), droplets produced by dispersion of more viscous oils were larger than those produced by less viscous oils. In general, higher dispersant concentration, favored the formation of smaller droplets, regardless of oil viscosity.

Introduction

On average, about 3 million gallons of crude oil are spilled into U.S. marine waters annually (NRC, 2005), but this figure has been dwarfed by the recent massive oil spill in the Gulf of Mexico. Although the trends in volume and number of accidental releases have been decreasing for several decades, the history of oil spills indicates that large spills occur intermittently (Schmidt-Etkin, 1999). When they occur, the environmental (e.g., contamination of surface-occurring animals and sensitive shorelines) and economic (e.g., disruption of commercial and recreational uses of near-shore areas) effects can be devastating. So, there is a continuing need to develop and improve spill response technologies. Although environmental damage cannot be completely prevented when accidental releases of petroleum occur, it may be possible to minimize the damage if a variety of complementary response alternatives are available.

Although mechanical response, which removes oil from the environment, is preferred, chemical dispersion can provide an effective alternative in cases where mechanical response would not be effective (Lessard and Demarco, 2000). In this approach, dispersants are applied to the oil to promote the formation of oil droplets that can be dispersed into the water column (Clayton et al., 1993; Lessard and Demarco, 2000). The oil droplets can be transported vertically and horizontally in the water column by turbulent diffusion, ultimately diluting the oil into a large volume of seawater. The active components of chemical dispersants are surfactants, which promote droplet formation by reducing the oil–water interfacial tension (NRC, 2005). Surfactants are amphiphilic compounds that contain hydrophobic and hydrophilic regions in the same molecule. These compounds reduce the interfacial tension by accumulating at the oil–water interface and interacting with both phases simultaneously. Although the basic physical chemistry of chemical dispersion is understood, prediction of dispersant performance for specific crude oils is largely based on empirical testing.

Previous studies have shown that the physical and chemical properties of the oil, composition of the dispersant, mixing energy, mixing time, fluid dynamics, temperature, and salinity are all important factors that affect dispersant performance (Clayton et al., 1993; NRC, 2005; Mukherjee and Wrenn, 2007, 2009a). Weathering also affects dispersant performance. For example, the viscosity of light oils increases when the volatile components evaporate and the water-soluble components dissolve into the water column (Aravamudan et al., 1981). More viscous oils resist breakage into droplets, and therefore less of the floating oil is transported into the water column (Chandrashekar et al., 2005) and larger droplets are formed (Wang and Calabrese, 1986; Delvigne and Sweeney, 1988). Increased viscosity also reduces the penetration of the dispersant into the oil slick and the transport rate of the dispersant through the oil mass (Clayton et al., 1993). The type and the amount of dispersant used affects the oil–water interfacial tension (Fingas et al., 1993; Tcholakova et al., 2004). Lower interfacial tension favors breakage of the oil slick into small oil droplets (Li and Garrett, 1998; Lessard and Demarco, 2000). The presence of surfactants can also significantly reduce the coalescence of suspended droplets (Ivanov et al., 1999). Coalescence is undesirable because it results in formation of larger, more buoyant droplets that may quickly rise back to the water surface and reduce dispersion effectiveness (DE). Mixing energy, provided by wind and waves, also plays a significant role in the initial droplet formation (Clayton et al., 1993; Shaw, 2003; Chandrashekar et al., 2005; Sis et al., 2005) and the kinetics of breakage and coalescence of suspended oil droplets (Coulaloglou and Tavlarides, 1977; Aravamudan et al., 1981; Fingas et al., 1996; Li and Garrett, 1998; NRC, 2005).

This report describes the results of an empirical investigation of the effects of several important factors on the performance of the commercial dispersant Corexit 9500. These factors include the physical properties of the oil, dispersant-to-oil ratio (DOR), and mixing energy. A wide range of physical properties were encompassed by the use of three crude oils with different initial densities and viscosities: Arabian Light (low density and viscosity), Mars (intermediate density and viscosity), and Lloyd (high density and viscosity). Bench-top dispersion experiments were conducted using a baffled-flask mixing system (Sorial et al., 2004a, 2004b). The mixing energy varied over a range that encompassed different marine environments (e.g., estuary and open ocean surface). This research will provide a better understanding of chemical dispersion of crude oil that can aid spill response planning and decision making.

Materials and Methods

Oil–dispersant mixtures

Evaporatively weathered Arabian Light, Mars, and Lloyd crude oils were used in these experiments. The physical and chemical properties of these oils are listed in Environment Canada's ETC database. The oils were evaporated under a stream of air for 24 h at room temperature (20°C–22°C) in a fume hood. The mass lost during this process differed due to differences in the relative concentrations of volatile components in each oil: the mass of Arabian Light was reduced by 23%, Mars was reduced by 21%, and Lloyd was reduced by 17%. Two separate oil–dispersant mixtures were prepared for each weathered oil. One mixture had a DOR of 1:25 (4%, v/v), and the DOR of the second was 1:100 (1%, v/v). To ensure homogeneous distribution of dispersant in the oil, the oil–dispersant mixtures were mixed for 24 h at 200 rpm on a gyratory shaker. The physical properties of the oil–dispersant mixtures are provided in Table 1.

Table 1.

Physical Properties of Mixtures of Evaporatively Weathered Crude Oils with Corexit 9500

Oil type	Weathering extent (%)	DOR	Density (ρ_d) (kg/m³)	Viscosity (μ_d) (g/cm-s)	Interfacial tension (σ) (g/s²)
Arabian Light	23	1:25	895	0.182	0.18
		1:100	880	0.162	0.21
Mars	21	1:25	928	1.53	0.45
		1:100	930	1.63	1.35
Lloyd	17	1:25	960	64.10	7.01
		1:100	960	85.80	14.73

DOR, dispersant-to-oil ratio.

Commercially available Corexit 9500 (Nalco Energy Services) was used as the dispersant. Corexit 9500 contains about 48% nonionic and 35% anionic surfactants in a solvent consisting of a mixture of food-grade aliphatic hydrocarbons (NRC, 2005). The nonionic surfactants include ethoxylated sorbitan mono- and tri-oleates and sorbitan monooleate. Sodium dioctyl sulfosuccinate is the major anionic surfactant in the dispersant (Pollino and Holdway, 2002).

Artificial seawater

Oil was dispersed in artificial seawater that was prepared by dissolving 35 g of artificial sea salts (Sigma-Aldrich) in 1 L of ultrapure deionized water. The seawater was filtered through a 0.2-μm membrane filter (Millipore) before use. Filtration removed suspended particles that could interfere with measurement of dispersed oil droplets.

Mixing

Dispersion experiments were conducted using baffled flasks, which were constructed by modifying 150-mL trypsinizing flasks with four glass baffles (Wheaton Science Products). The flasks were modified by attaching a drain port near the bottom to facilitate sample collection (Sorial et al., 2004a, 2004b). The contents in the baffled flasks were mixed by shaking on a Lab-Line Orbit Environ Shaker (Lab-Line Instruments, Inc.) with an orbital diameter of 1.9 cm. Three rotational speeds, 125, 150, and 200 rpm, were used. The specific energy dissipation rates corresponding to these speeds were 0.00075, 0.016, and 0.16 W/kg water, respectively. These specific energy dissipation rates were determined by Kaku et al. (2006) using a mixing system with the same characteristics as described above. For one rotational speed (125 rpm), the specific energy dissipation rate was estimated by exponential extrapolation of the reported data for 50 and 100 rpm.

Dispersion experiments

One-hundred twenty milliliters (V_aq,tot) of filtered artificial seawater was added to each baffled flask, and 0.1 mL (V_oil,tot) of an oil–dispersant mixture was added to the water surface using a Repeater Plus Pipette (Eppendorf) with a 0.5 mL pipette tip. The contents of the flasks were mixed for 45 min at a predetermined rotational speed. Preliminary results showed that dispersion of all six oil–dispersant combinations used in this research reached steady state within 45 min of mixing. After the mixing period, 1 mL of the dispersed phase was withdrawn through the stopcock, at the bottom of the baffled flask, and discarded and a 40 mL sample was collected and subjected to analysis as described below. Three independent replicates were conducted for each experimental condition (Table 2) for a total of 54 independent tests. The order of the tests was randomized to preclude confounding of systematic (e.g., temporal) effects with treatment effects.

Table 2.

Experimental Design for Evaluating Treatment Effects on Dispersion Effectiveness and Droplet-Size Distribution

Oil type	DOR	Rotational speed (rpm)
23% weathered Arabian Light	1:25	125
		150
		200
	1:100	125
		150
		200
21% weathered Mars	1:25	125
		150
		200
	1:100	125
		150
		200
17% weathered Lloyd	1:25	125
		150
		200
	1:100	125
		150
		200

DOR, dispersant-to-oil ratio.

Analytical methods

The mass concentration of dispersed oil was determined by extracting the oil into dichloromethane methane (DCM) (Fisher Scientific) followed by spectrophotometric measurement of the concentration of extracted oil (Sorial et al., 2004a, 2004b). The oil was extracted from the aqueous phase by transferring 30 mL (V_oil,ext) of the sample to a 125-mL separatory funnel, followed by extraction with three 5-mL portions of DCM. The DCM extracts were combined in a 100-mL beaker. The solvent was dried by passing it through a bed of anhydrous sodium sulfate (Fisher Scientific) and collected in a clean beaker. The sodium sulfate was rinsed with additional DCM to recover any oil trapped in the bed. The volume of the pooled extracts was adjusted to 50 mL (V_ext,dilute) in a volumetric flask, and 3 mL of the extract was transferred to a quartz cuvette. The absorbance was measured relative to a DCM blank at 0.5-nm intervals from 340 to 400 nm using a Perkin-Elmer LAMBDA 2 UV-visible spectrophotometer. The trapezoidal rule was used to integrate the area under the absorbance versus wavelength curve. If a sample exhibited absorbance >2.5 at any wavelength, it was diluted by a factor of two and the integrated absorbance was measured again. The method was calibrated by determining the integrated absorbance for a series of six standards that were prepared by extracting stock dispersant–oil solutions, separately for each of the six oil–dispersant combinations, from synthetic seawater into DCM using a procedure that was identical to that described above.

The size distribution and number concentration of the dispersed oil droplets was measured using an optical particle counter (OPC; Particle Measuring Systems Inc.). The OPC was equipped with an LS-200 sample module and a Liquilaz E20P detector with a built-in 12-mW laser diode (λ = 785 nm). The OPC reported droplet sizes in 15 adjustable channels from 2 to 125 μm. The number concentration of the particles must be <10⁴ mL⁻¹ to prevent coincidence counting. Therefore, samples were diluted with deionized water by a factor of 100–5,000 before measurement. The pipette tips used to transfer these samples were cut to enlarge the opening to at least 2 mm to minimize the effects of sample transfer on the droplet-size distribution (APHA, 1999). Our previous research (Mukherjee and Wrenn, 2009b) showed that the size of the oil droplets reported by the OPC (diluted sample) matched well with that obtained using microscopy (undiluted sample). Moreover, the mass concentrations of the oil dispersed estimated from the OPC generated size distribution were statistically indistinguishable from those estimated using gravimetric analysis and UV-spectroscopy.

The interfacial tension between the oil–dispersant mixtures and the filtered artificial seawater was measured using a digital tensiometer (Model: KSV Sigma 703 D) using the Du Nouy ring method. The ring diameter was 1.909 cm and was made of platinum–iridium wire with diameter of 0.037 cm. The ring was hung from a balance hook and was slowly lowered into the filtered artificial seawater. A thick layer of oil–dispersant mixture was then carefully added on to the water surface. The Du Nouy ring was then slowly raised through the oil–water interface and the force needed to break through the interface was recorded to estimate the oil–water interfacial tension. The lower and upper measuring range for the instrument was 0.001 and 1,000 dyne/cm, respectively, at room temperature (20°C–22°C).

The viscosities of the oil–dispersant mixtures were measured using a constant stress and strain rheometer (TA Instruments AR 2000). The rheometer had a cone and plate geometry with a diameter of 60 mm (cone angle 0° 59 min, 49 s: truncation 27 μm). The torque was between 0.1 and 200 μN-m. The lower limit of the stress that the equipment can accurately measure was 0.0008 Pa. The temperature was maintained at 25°C during measurement using a Peltier plate. The Peltier plate was cleaned with DCM and isopropyl alcohol (Sigma-Aldrich) and dried before and after all measurements.

Dispersion effectiveness and droplet size distribution

Dispersion effectiveness (DE) is defined to be the fraction of added oil that was dispersed in the water column at the end of the mixing period. DE was estimated from the dispersed oil concentration (C_oil,ext; g/mL) as shown below: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6} \begin{document} \begin{align*} { \rm DE } = \frac { M_ { \rm oil, aq } } { M_ { \rm oil , o } } = \frac { \left( \frac { V_ { \rm aq , tot } } { V_ { \rm oil , ext } } \right) [ ( C_ { \rm oil , ext } ) ( V_ { \rm ext , dilute } ) ] } { V_ { \rm oil , tot } \rho_ { \rm oil } } \tag { 1 } \end{align*} \end{document}

where M_oil,o, M_oil,aq, and ρ_oil are the initial mass of oil, the mass of dispersed oil, and the oil density, respectively.

Droplets formed under different treatments were compared based on the volume distributions of the droplets. An important parameter used to characterize each size distribution was the diameter of mean volume (DMV): \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6} \begin{document} \begin{align*} { \rm DMV } = \left( \frac { \sum \limits_ { i = 1 } ^ { 15 } N_i \overline{d} _i^3 } { N_ { \rm tot } } \right) ^ { 1 / 3 } \tag { 2 } \end{align*} \end{document}

where N_i (mL^–1) is the number concentration of droplets in channel i, d_i (μm) is the average droplet diameter in channel i (i.e., the arithmetic mean of the upper and lower size limits), and N_tot (mL^–1) is the total droplet number concentration (i.e., the sum of the number concentrations of droplets in all 15 channels).

Statistical analysis

Analysis of variance (ANOVA) was used to evaluate the statistical significance of the main effects and interactions due to mixing energy, oil type, and DOR on DE and the overall DMV of the dispersed oil. In addition, a general linear model (GLM) was used to evaluate the significant main effects and interactions of the oil physical properties (i.e., viscosity and interfacial tension) and mixing energy using forward stepwise multiple linear regression to estimate the best-fit coefficients. A limited sensitivity analysis was performed to determine whether the statistically significant parameters of the GLM could have been obtained by chance due to the specific set of values that were measured for the response variables (DE and DMV). The sensitivity of the model to the specific data observed in this study was evaluated by generating 10 alternative data sets and conducting the GLM analysis independently for each data set. The alternative data sets were produced by generating 10 populations of normally distributed random errors with mean of zero and variance equal to that measured for the original data set, and then the errors were added to the measured response for each independent experimental run (total of 540 errors; 10 errors/replicate × 3 replicates/treatment × 18 treatments). The coefficients estimated for each significant factor were averaged over the 10 independent implementations of the GLM, and the 95% confidence intervals were determined. When a factor was not identified as being statistically significant in the model for one or more of the data sets, the coefficient was considered to be zero for that realization.

Results

Dispersion effectiveness

The effects of mixing energy and DOR on the DE for weathered Arabian Light, Mars, and Lloyd are shown in Fig. 1. The DE for all three oils was relatively insensitive to changes in mixing energy above an energy dissipation rate of 0.016 W/kg, but mixing energy exerted a strong effect on DE at lower energy dissipation rates for all three oils at the lowest DOR and at both DOR for the most viscous oil (Lloyd). The effect of DOR was most important for viscous oil or low energy dissipation rates. For example, Arabian Light and Mars, which were almost completely dispersed at higher energy dissipation rates regardless of DOR, were also well dispersed (>75%) at the lowest energy dissipation rate when the DOR was 1:25 but were very poorly dispersed (<15%) at this mixing energy when the DOR was reduced to 1:100. For weathered Lloyd, which is 40–50 times more viscous than weathered Mars and several hundred times more viscous than weathered Arabian Light, the maximum DE was limited by DOR. So, although the trends are consistent with a priori expectations, the interactions among DOR, mixing energy, and oil viscosity exert complex effects on DE.

FIG. 1.

Effect of mixing energy on dispersion effectiveness for weathered Arabian Light (white), Mars (gray), and Lloyd (black) crude oils at DOR of 1:25 (no hatching) and 1:100 (hatched). The error bars represent one standard deviation of independent triplicate experiments.

Treatment effects on DE were evaluated using ANOVA (Table 3). Mixing energy, oil type, and DOR exerted significant main effects on DE (p < 0.001). In addition, the two-factor interaction between mixing energy and DOR and the three-factor interaction (i.e., mixing energy, oil type, and DOR) were statistically significant (p < 0.001), indicating that the main effects of the treatment factors were not simply additive. The relative importance of the main effects, as indicated by the F-ratio, was mixing energy >DOR >oil type.

Table 3.

Three-Way Analysis of Variance for the Effect of Treatment Factors on Dispersion Effectiveness and Diameter of Mean Volume

		Dispersion effectiveness			Diameter of the mean volume
Source	df	Mean squares	F-ratio	p^a	Mean squares	F-ratio	p^a
Energy	2	13,952.4	393.7	<0.001	113.2	54.6	<0.001
Oil type	2	6,308.5	178.0	<0.001	378.6	182.7	<0.001
DOR	1	10,661.6	300.8	<0.001	131.3	63.4	<0.001
Energy × oil type	4	33.6	0.9	0.447	19.1	9.2	<0.001
Energy × DOR	2	1,843.1	52.0	<0.001	23.2	11.2	<0.001
Oil type × DOR	2	51.8	1.4	0.245	10.8	5.2	0.010
Energy × oil type × DOR	4	1,836.3	51.8	<0.001	6.6	3.1	0.024
Error	36	35.4			2.0

p is the probability that the factor or interaction has no effect on the response variable.

In this study, DOR and oil type were used to provide a wide range of oil–water interfacial tension and oil viscosity, which are more fundamental physical properties that are known to affect dispersion. General linear modeling was conducted using these physical properties and mixing energy as the independent variables in a forward stepwise regression analysis. In this analysis, only the main effects of mixing energy and interfacial tension were determined to be statistically significant (Table 4). The main effect of viscosity and all possible interactions were not statistically significant. The sensitivity analysis suggested that the main effects of mixing energy and interfacial tension were real because the averages for their coefficients over all 10 alternative realizations of the dispersion data were significantly different from zero (Table 4). In addition, the main effect of mixing energy was identified as being statistically significant in all 10 alternative data sets, and the main effect of interfacial tension was statistically significant in 6 of 10 cases (Table 4). Note, that the GLM was only able to explain 43% of the total variation that was observed in DE when only mixing energy and interfacial tension were used as independent variables (Fig. 2A). Moreover, the residuals of the best-fit model were correlated with the measured DE (Fig. 2B; slope, −0.56 ± 0.13), which indicates that unidentified factors exerted important effects on DE.

FIG. 2.

Fit of general linear model (GLM) for dispersion effectiveness using model coefficients shown in Table 4: (A) comparison of model-predicted and observed dispersion effectiveness and (B) model residuals. Solid line in (A) represents perfect correspondence between GLM-predicted and observed dispersion effectiveness. Solid line in (B) shows relationship between the residuals and observed effectiveness.

Table 4.

Significant Main Effects and Two-Way Interactions of Oil Physical Properties and Mixing Energy on Dispersion Effectiveness

	GLM		Sensitivity analysis
Significant effects	p^a	Coefficient	Coefficient ^b	Frequency (%) ^c
Constant	<0.001	0.687	0.68 ± 0.02	100
Energy	<0.001	1.85	1.73 ± 0.21	100
Interfacial tension	<0.001	−0.034	−0.02 ± 0.01	60

p is the probability that the coefficient is zero.

Average of the coefficients (±95% confidence interval) over all runs of the GLM sensitivity analysis; the coefficients were assumed to be zero for any factor not identified as being statistically significant in a given run.

Percentage of runs in which a factor was found to be statistically significant in the sensitivity analysis.

Normalized volume distribution

The normalized volume distributions of the oil droplets generated during dispersion, as a function of mixing energy and DOR, are shown in Figs. 3, 4, and 5 for weathered Arabian Light, Mars, and Lloyd, respectively. The volume distributions produced by dispersion of weathered Arabian Light and Mars were mostly tri-modal (Figs. 3 and 4), although the largest size mode was less prominent at high DOR and energy dissipation rates. The DMV for the three droplet-size modes were relative insensitive to changes in the mixing energy and DOR. On average, the DMV for the small, medium, and the large-size droplets were ∼3.5, 10.5, and 30 μm, respectively, for both oils. A fourth mode consisting of very small droplets was observed in the volume distribution for Arabian Light at the highest DOR and mixing energy, but the DMV could not be estimated because it included droplets that were smaller than the detection limit for the OPC. For the weathered Lloyd, the droplet volume distributions were tetra-modal at both DOR and all three energy levels (Fig. 5). The DMVs of the small, medium-small, medium-large, and large droplet-size modes were ∼3.3, 7.8, 11.5, and 33.0 μm, respectively (i.e., the medium size mode observed for dispersed Arabian Light and Mars appears to have split into two distinct modes for dispersed Lloyd). Like the weathered Arabian Light and Mars, the DMVs for the four droplet-size modes observed in dispersions of weathered Lloyd were not affected by changes in mixing energy and DOR. Unlike the less viscous oils, for which most of the oil was entrained either as small- or medium-sized droplets, most of the weathered Lloyd was entrained as large droplets.

FIG. 3.

Normalized volume distributions for dispersed weathered Arabian Light crude oil at DOR of (A) 1:25 and (B) 1:100 for mixing energies of 0.00075 W/kg (circles), 0.016 W/kg (triangles), and 0.16 W/kg (squares).

FIG. 4.

Normalized volume distributions for dispersed weathered Mars crude oil at DOR of (A) 1:25 and (B) 1:100 for mixing energies of 0.00075 W/kg (circles), 0.016 W/kg (triangles), and 0.16 W/kg (squares).

FIG. 5.

Normalized volume distributions for dispersed weathered Lloyd crude oil at DOR of (A) 1:25 and (B) 1:100 for mixing energies of 0.00075 W/kg (circles), 0.016 W/kg (triangles), and 0.16 W/kg (squares).

The volume fraction of dispersed oil that was present in each droplet-size mode is shown in Fig. 6 for all experimental conditions tested in this study. The fraction of oil dispersed as small droplets increased with increasing DOR and mixing energy, and the ease of formation of small droplets was inversely proportional to the viscosity of the oil. For example, the amount of weathered Arabian Light and Mars that was dispersed as droplets in the smallest size mode (DMV = 3.5 μm) increased with increasing mixing energy regardless of DOR, but the small size mode dominated the volume distributions only at the high DOR (1:25) and high mixing energy (0.16 W/kg). Although the effect of DOR on DE was difficult to discern for these oils at mixing energies >0.016 W/kg, the effect on the droplet size was clear because medium-sized droplets dominated the size distributions for both oils at the highest mixing energy when the DOR was low (1:100). At the high DOR, the relative amount of Mars crude oil dispersed as medium-sized droplets was maximal at the intermediate energy dissipation rate (0.016 W/kg), and the relative amount dispersed as large droplets was maximal for both lower viscosity oils at the lowest energy dissipation rate. Similar behavior was observed for dispersed Lloyd crude oil, except the large size modes (large and medium-large) contained the majority of the dispersed oil. The largest size mode dominated the volume distribution when Lloyd was dispersed at the low DOR (80%–90%, depending on mixing energy) and the smallest droplets were almost nonexistent (<2%).

FIG. 6.

Ternary plot of the distribution of dispersed oil among small, medium, and large droplet-size modes for weathered Arabian Light (circles), weathered Mars (triangles), and weathered Lloyd (squares) at DOR of 1:25 (shaded) and 1:100 (unshaded) as a function of mixing energy. Each symbol is marked with a letter to indicate whether mixing energy was low (0.00075 W/kg, “l”), medium (0.016 W/kg, “m”), or high (0.16 W/kg, “h”).

Diameter of mean volume

Because the mixing energy, oil type, and DOR affected the amount of the oil entrained as different sized droplets, the overall DMV of the oil droplets was also affected. The variation of DMV with mixing energy is shown in Fig. 7 for all three oils at both DOR. The DMV was inversely proportional to mixing energy (i.e., DMV decreased for all three oils as the energy dissipation rate increased). Also, except for the lowest energy dissipation rate, higher dispersant concentration resulted in smaller DMV for all three oils. As expected, the most viscous oil (Lloyd) formed droplets with the largest DMV.

FIG. 7.

Effect of mixing energy on diameter of mean volume (DMV) of dispersed oil droplets for weathered Arabian Light (white), Mars (gray), and Lloyd (black) crude oils at DOR of 1:25 (no hatching) and 1:100 (hatching). Error bars represent one standard deviation of independent triplicate experiments.

The main effects of mixing energy, oil type, and DOR (p < 0.001), all two-factor interactions (p < 0.01), and the three-factor interaction (p = 0.024) were statistically significant (Table 3). Based on the F ratio, the oil type exerted the strongest effect on DMV. Regression analysis using the physical properties as the treatment factors indicated that the main effects of mixing energy and interfacial tension on DMV and the two-way interactions between mixing energy and viscosity and between viscosity and interfacial tension were statistically significant (p < 0.05) (Table 5). The sensitivity analysis showed that the average coefficients for both main effects and both two-factor interactions were significantly different from zero in the GLM. Also, all of the main effects and interactions were identified as being statistically significant in at least 80% of the alternative realizations of the data. The GLM explained 85% of the total variation that was observed in measured DMV (Fig. 8A), but the residuals were correlated with the measured DMV (slope = −0.16 ± 0.09) (Fig. 8B), indicating that some important factors controlling droplet size have not been identified.

FIG. 8.

Fit of the GLM for DMV using model coefficients shown in Table 5: (A) comparison of model-predicted and observed DMV and (B) model residuals. Solid line in (A) represents perfect correspondence between GLM- predicted and observed dispersion effectiveness. Solid line in (B) shows relationship between the residuals and observed DMV.

Table 5.

Significant Main Effects and Two-Way Interactions of Oil Physical Properties and Mixing Energy on Diameter of Mean Volume

	GLM		Sensitivity analysis
Significant effects	p^a	Coefficient	Coefficient ^b	Frequency (%) ^c
Constant	<0.001	6.62	6.59 ± 0.27	100
Energy	0.001	−16.46	−16.22 ± 1.89	100
Interfacial tension	<0.001	2.51	2.63 ± 0.40	100
Viscosity × interfacial tension	<0.001	−0.019	−0.019 ± 0.003	100
Energy × viscosity	<0.001	−0.428	−0.325 ± 0.13	80

p is the probability that the coefficient is zero.

Percentage of runs in which a factor was found to be statistically significant in the sensitivity analysis.

Discussion

The effectiveness of chemical dispersion is known to depend on the specific combination of oil and dispersant, the dispersant concentration, and the environmental conditions under which dispersion occurs. In this study, the effectiveness with which three oils (Arabian Light, Mars, and Lloyd) were dispersed by the commercially available dispersant Corexit 9500 was investigated at two dispersant concentrations (DOR of 1:100 and 1:25) over mixing energies spanning more than two orders of magnitude (0.00075–0.16 W/kg). The oils were selected to encompass a wide range of viscosities (>15 to <8,600 cP), and addition of dispersant at two concentrations to each oil gave a range of interfacial tensions that spanned almost two orders of magnitude (0.2–15 dyn/cm). The effects of the treatment factors were evaluated based on the amount of oil entrained in the water column as droplets (DE) and the average size of the entrained droplets (DMV). The effects of qualitative treatment factors (oil type and DOR) were evaluated using ANOVA, and fundamental physical properties (viscosity and interfacial tension) were evaluated using general linear modeling.

In general, ANOVA identified more significant main effects and interactions than the GLM, which suggests that these physical properties are not sufficient to predict the performance of specific dispersant–oil combinations. For example, whereas ANOVA showed that all three main effects (mixing energy, oil type, and DOR) and their three-factor interaction were statistically significant for both response variables (Table 3), viscosity was not found to exert a significant main effect on either response (Tables 4 and 5). Similarly, the three-factor interaction among mixing energy, interfacial tension, and viscosity was not statistically significant in the best-fit GLM for either response variable. Because viscosity and interfacial tension were both affected by the oil type and DOR, one might not expect a simple one-to-one correspondence between viscosity and oil type or DOR and interfacial tension, but in general, the differences between these properties due to the interaction between oil type and DOR was smaller than the differences that were attributable by the oils alone. The differences between the conclusions of these two methods of statistical analysis suggest that other properties, such as the chemical composition (Mukherjee et al., in preparation), are also important because, if identical treatment factors and responses are used, both methods produce identical conclusions.

One advantage of general linear modeling over ANOVA is that the coefficients of the GLM give a direct indication of the size and direction of each significant treatment effect or interaction, and the model can be used to predict the response to any given combination of treatment conditions. For example, mixing energy positively affected DE (i.e., DE increased with increasing mixing energy), whereas interfacial tension exerted a negative effect on DE (i.e., DE decreased as the interfacial tension increased; Table 4). Both of these factors are consistent with our a priori expectations because increasing interfacial tension increases the energy required to create new oil–water interfacial area and increasing mixing energy provides additional energy that can be used to create new interfacial area. All other things being equal, the coefficients of the GLM predict that increasing the mixing energy from 0.016 to 0.16 W/kg should increase DE by about 27% (i.e., an additional 0.27 mL/mL of the floating oil was dispersed when the mixing energy was increased from the intermediate to high level), and increasing the oil–water interfacial tension from 7 to 14.7 dyn/cm (Lloyd) should decrease the DE by about 26% (i.e., 0.26 mL/mL). Although the effect of interfacial tension on dispersion of Lloyd crude oil was predicted reasonably well (average decrease in DE was 31% [i.e., 0.31 ± 0.2 mL/mL]), increasing the mixing energy from 0.016 to 0.16 W/kg had essentially no effect on the DE regardless of oil type. Further, the GLM predicted that increasing the mixing energy from 0.00075 to 0.016 W/kg should increase DE by only about 3% (i.e., 0.03 mL/mL), whereas the average observed increase was 47% (i.e., 0.47 mL/mL). The highly nonlinear response to mixing energy is one of the reasons for the relatively poor fit of the GLM for DE (Fig. 2; R² = 0.43).

The best-fit GLM for DMV, on the other hand, provided a reasonably good description of the observed DMV over the range of treatment conditions that were tested (Fig. 8; R² = 0.85). This model includes statistically significant main effects due to mixing energy and oil–water interfacial tension and significant interactions between mixing energy and viscosity and between viscosity and interfacial tension (Table 5). As expected, mixing energy exerted a negative effect on DMV (i.e., the average droplet size decreased as mixing energy increased), and interfacial tension exerted a positive effect (i.e., droplet size increased as the oil–water interfacial tension increased). The interactions of the main effects with viscosity were both negative. The negative interactions with viscosity indicate that the effect of mixing energy was greater for more viscous oils (e.g., Lloyd) than for less viscous oils (e.g., Arabian Light), whereas the effect of interfacial tension decreased with increasing viscosity. The greater effect of mixing energy on more viscous oils may reflect the simple fact that droplet size cannot decrease indefinitely (i.e., it approaches a lower limit of zero); so, oil that can be broken into small droplets at relatively low energy levels has less additional size reduction potential. The smaller negative effect of interfacial tension for more viscous oils, on the other hand, may reflect a transition from a droplet-formation mechanism that is primarily controlled by interfacial tension to one that is primarily controlled by viscosity.

The design of this study precludes significant interpretation with respect to the mechanisms of droplet formation because the samples were collected at steady state where droplet breakage and coalescence occurred simultaneously. Nevertheless, the presence of multiple modes in the normalized volume distributions of the dispersed oil droplets for all treatments suggests that multiple droplet-formation mechanisms may have been involved (Figs. 3, 4, and 5). The observed size distributions were a result of the mechanisms and rates of droplet formation and coalescence, which depend on the fluid dynamics (e.g., mixing energy and flow velocity) and the properties of the dispersed (oil) and the continuous (water) phases (Ceylan et al., 2003; Kolev, 2004). Droplets of different sizes can be produced either during the initial breakup of a floating oil slick (e.g., simultaneous formation of large droplets and small satellite droplets) or by breakage of droplets in the water column (Shaw, 2003). Droplet breakage in the water column may result in formation of two daughter droplets (binary breakage) or many (e.g., erosive breakage and thorough breakage) (Narsimhan et al., 1980). Similarly coalescence results in formation of larger droplets from the collision of two smaller droplets and can have a strong influence on the overall efficiency of dispersion because larger droplets are more buoyant and have faster rise velocities than smaller droplets (Sterling et al., 2004). At lower mixing energy, coalescence might be favored over breakage resulting in the formation of larger droplets. The large droplets can quickly rise back to the surface and thereby reduce DE dramatically (Sterling et al., 2004). On the other hand, droplet formation could reasonably be expected to increase with energy dissipation rate at a different rate than coalescence. So, the net result could be increased dispersion with increased mixing energy, as observed in our experiments. The important role of droplet formation mechanisms in the chemical dispersion of crude oil is shown by the differences in steady-state droplet-size distributions that can be produced when the same combination of oil, dispersant, and mixing energy are treated in mixing systems with different hydrodynamic characteristics (Mukherjee and Wrenn, 2009a).

Conclusion

The effects of oil type, dispersant concentration (i.e., DOR), and energy dissipation rate on the chemical dispersion of crude oil were investigated using a baffled-flask mixing system. Three evaporatively weathered crude oil—Arabian Light, Mars, and Lloyd—were selected based on the differences in their physical properties, and the commercially available product, Corexit 9500, was used as the dispersant. ANOVA showed that oil type, DOR, and mixing energy exerted statistically significant effects on DE and the DMV, which was used to characterize the droplet-size distributions. ANOVA also indicated the significance of the two- and three-factor interactions. When the fundamental physical properties of viscosity and interfacial tension were used in GLMs in place of oil type and DOR, however, fewer statistically significant main effects and interactions were identified, suggesting that other characteristics of these oil–dispersant combinations were also important in controlling dispersion. Although mixing energy and oil–water interfacial tension were identified by the GLM as exerting statistically significant effects on DE and DMV, viscosity only occurred through its interactions with mixing energy and interfacial tension in the best-fit model for DMV.

The normalized volume distributions of the dispersed oil droplets were multimodal for all treatments at steady state, and the central tendencies of these modes were relatively consistent regardless of oil type, DOR, or mixing energy. The only important difference among oils was that medium-size mode (DMV about 10 μm), observed in the case of weathered Arabian Light and Mars, split into two modes with DMV of about 8 and 12 μm when weathered Lloyd was dispersed. The relative amount of oil in each droplet-size mode, however, was a function of the experimental conditions. In general, higher dispersant concentration and higher mixing energies favored the formation of smaller droplets, but the effects were moderated by viscosity (i.e., weathered Mars produced dispersions with substantial amounts of oil present as medium-sized droplets even under the most aggressive dispersion conditions and dispersions of weathered Lloyd were dominated by large droplets under all conditions).

Successful application of chemical dispersion in the mitigation of marine oil spills requires a better understanding of the effects of oil properties, dispersant chemistry, and environmental conditions on the relative amount of oil transferred as droplets to the water column and the size distributions of the dispersed oil droplets. Empirical studies, such as this one, can provide information on the relative performance of specific dispersants with specific oils that can inform spill response planning and decision making. The number of combinations that must be tested under different conditions are, however, staggering. A more fundamental understanding of the physical, chemical, and hydrodynamic factors that control dispersion is necessary to reduce the scope of this task to something that is manageable.

Footnotes

Acknowledgments

The authors would like to thank Larry Heugatter (Conoco Philips Co.) for providing the crude oils used in the research, and Pratim Biswas (Washington University) for providing access to the optical particle counter. The Department of Energy, Environmental, and Chemical Engineering provided financial support to B.M. during this research.

Author Disclosure Statement

No competing financial interests exist.

References

American Public Health Association (APHA). 1999. Standard Methods for the Examination of Water and Wastewater. 20th. Washington, DC: American Public Health Association, American Water Works Association, and Water Environment Federation.

Aravamudan

K.S.

, Raj

P.K.

, Marsh

1981. Simplified models to predict the breakup of oil on rough seas. Proceedings of the International Oil Spill Conference. American Petroleum Institute: Atlanta, GA, 153–159.

Ceylan

, Kelbaliyev

, Ceylan

2003. Estimation of the maximum stable drop sizes, coalescence frequencies and the size distributions in isotropic turbulent dispersions. Colloids Surf. A. Physicochem. Eng. Aspects, 212:285.

Chandrashekar

, Sorial

G.A.

, Weaver

J.W.

2005. Dispersant effectiveness of three oil under various simulated conditions. Environ. Eng. Sci., 22:324.

Clayton

J.R.

, Payne

J.R.

, Farlow

J.S.

1993. Oil Spill Dispersants: Mechanisms of Action and Laboratory Tests. Boca Raton, FL: CRC Press, Inc.

Coulaloglou

C.A.

, Tavlarides

L.L.

1977. Description of interaction processes in agitated liquid-liquid dispersions. Chem. Eng. Sci., 32:1289.

Delvigne

G.A.L.

, Sweeney

C.E.

1988. Natural dispersion of oil. Oil Chem. Pollut., 4:281.

Fingas

M.F.

, Huang

, Fieldhouse

, Wang

, Mullin

J.V.

1996. The effect of energy, settling time and shaking time on the swirling flask dispersant apparatus. Spill Sci. Technol. Bull., 3:193.

Fingas

M.F.

, Kyle

D.A.

, Tennyson

E.J

. 1993. Physical and chemical studies on dispersants: the effect of dispersant amount and energy. Proceedings of the 16th Arctic and Marine Oil Spill Program Technical Seminar. Environment Canada, Ottawa, Canada, 861–876.

10.

Ivanov

I.B.

, Danov

K.D.

, Kralchevsky

P.A.

1999. Flocculation and coalescence of micron-size emulsion droplets. Colloids Surf. A. Physicochem. Eng. Aspects, 15:161.

11.

Kaku

J.V.

, Boufadel

M.C.

, Venosa

A.D.

, Weaver

2006. Flow dynamics in eccentrically rotating flasks used for dispersant effectiveness testing. Environ. Fluid Mech., 6:385.

12.

Kolev

N.I.

2004. Multiphase Flow Dynamics 2: Thermal and Mechanical Interactions. Berlin: Springer.

13.

Lessard

R.R.

, Demarco

2000. The significance of oil spill dispersants. Spill Sci. Technol. Bull., 6:59.

14.

, Garrett

1998. The relationship between oil droplet size and upper ocean turbulence. Marine Pollut. Bull., 36:961.

15.

Mukherjee

, Wrenn

B.A.

2009a. Influence of dynamic mixing energy on dispersant performance: role of mixing systems. Environ. Eng. Sci., 26:1725.

16.

Mukherjee

, Wrenn

B.A.

2009b. Size distribution as a measure of dispersant performance. Can. J. Civ. Eng., 36:540.

17.

Mukherjee

, Wrenn

B.A.

2007. Effects of mixing energy and flow dynamics on chemical dispersion of crude oil. Proceedings of the 30th Arctic Marine Oil Spill, Program Technical Seminar. Environment Canada, Edmonton, Canada, 189–202.

18.

Narsimhan

, Ramkrishna

, Gupta

J.P.

1980. Analysis of drop size distributions in lean liquid-liquid dispersions. AIChE. J., 30:457.

19.

National Research Council (NRC). 2005. Oil Spill Dispersants: Efficacy and Effects. Washington: National Academies Press.

20.

Pollino

C.A.

, Holdway

D.A.

2002. Toxicity testing of crude oil and related compounds using early life stages of the crimson-spotted rainbowfish (Melanotaenia fluviatilis) Ecotoxicol. Environ. Saf., 52:180.

21.

Schmidt-Etkin

1999. Historical overview of oil spills from all sources (1960–1998) Proceedings of the International Oil Spill Conference. American Petroleum Institute, Seattle, WA, 1097–1102.

22.

Shaw

J.M.

2003. A microscopic view of oil slick break-up and emulsion formation in breaking waves. Spill Sci. Technol. Bull., 8:491.

23.

Sis

, Kelbaliyev

, Chander

2005. Kinetics of drop breakage in stirred vessels under turbulent conditions. J. Dispers. Sci. Technol., 26:565.

24.

Sorial

G.A.

, Venosa

A.D.

, Koran

K.M.

, Holder

, King

D.W.

2004a. Oil spill dispersant effectiveness protocol I: impact of operational variables. J. Environ. Eng., 130:1073.

25.

Sorial

G.A.

, Venosa

A.D.

, Koran

K.M.

, Holder

, King

D.W.

2004b. Oil spill dispersant effectiveness protocol II: performance of revised protocol. J. Environ. Eng., 130:1073.

26.

Sterling

M.C.

, Bonner

J.S.

, Ernest

A.N.S.

, Page

C.A.

, Autenrieth

R.L.

2004. Chemical dispersant effectiveness testing: influence of droplet coalescence. Marine Pollut. Bull., 48:969.

27.

Tcholakova

, Denkov

N.D.

, Danner

2004. Role of surfactant type and concentration for the mean drop size during emulsification in turbulent flow. Langmuir, 20:7444.

28.

Wang

C.Y.

, Calabrese

R.V.

1986. Drop breakup in turbulent stirred-tank contactors. II: Relative influence of viscosity and interfacial tension. Am. Inst. Chem. Eng., 32:667.