NEWS FEATURE

The perplexing physics of oil dispersantsMassive amounts of oil, gas, and dispersant streamed into the Gulf of Mexico during the

Deepwater Horizon disaster. Understanding the chemistry and physics of this mix as it

churned through the salt water turns out to be an exceedingly complex problem with plenty

of unknowns.

M. Mitchell Waldrop, Science Writer

On April 30, 2010, 10 days after a blowout destroyedthe offshore drilling platform Deepwater Horizon offthe coast of Louisiana and triggered what was fastbecoming the worst oil spill in US history, the well’sowner, British Petroleum, sent a remotely piloted sub-marine 1,500 meters down to the floor of the Gulf ofMexico. Once the vehicle arrived at the broken

wellhead, which was still spewing more than 6,000 litersof oil per minute, it stuck the end of a kilometers-longhose into the erupting plume and started pumping indispersants: detergent-like chemicals designed to frag-ment the hydrocarbons into tiny droplets. It was the startof a campaign that would ultimately inject the plumewith almost 3 million liters of the chemicals.

No one knew the ecosystem impact of using huge amounts of dispersants in deep water to break up the massive oil slickcaused by the 2010 British Petroleum disaster, seen here via satellite one month after the blowout. Image credit:Science Source/NASA.

Published under the PNAS license.

www.pnas.org/cgi/doi/10.1073/pnas.1907155116 PNAS | May 28, 2019 | vol. 116 | no. 22 | 10603–10607

NEW

SFEATURE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

Using dispersant at that depth was a roll of thedice; the chemicals had been used before on surfaceoil slicks with varying degrees of success but neverin such cold, deep waters. No one could be surewhat effect they would ultimately have on the oceanecosystem, on coastal fisheries, or even on the oilitself. The responders could only hope that the in-jection would work as intended and that the resultingoil droplets would be consumed by the Gulf’s manypetroleum-eating bacteria without ever making it tothe surface.

So did it work? That depends on whom you ask.The oil companies certainly think it did, says Tamay

Özgökmen, a mechanical engineer at the University ofMiami in Coral Gables, FL, who has spent much ofthe past eight years studying the Deepwater Hori-zon incident and its aftermath. The companiespoint to plummeting concentrations of toxic vaporsover the oil slick—cleanup crews could finally workwithout respirators—and to aerial photographs sug-gesting that less oil was reaching the surface. Sofrom the companies’ perspective, says Özgökmen,deep-sea dispersants have gone from being a des-peration move to being standard operating proce-dure. “They’re preparing for it in future oil spills,”he says.

The National Academy of Sciences’Ocean ScienceBoard tends to agree: In a draft consensus report re-leased on April 5, the Board’s panel on oil-spill-dispersant use concluded that yes, the deep-sea in-jection had generally been effective at dispersingthe oil, making the hydrocarbons easier for bacteriato digest, preventing surface oil from fouling nearbyshores, and enhancing worker safety by mitigatingexposure to hazardous oil-related chemicals (1).

But the report—and the numerous researchersstudying dispersants’ effects—also emphasized themany remaining uncertainties. The benefits of mas-sive deep-sea dispersants are still a matter of intense

debate among scientists. And the risks are evenmurkier. Dispersants by themselves don’t pose muchof a near-term risk. They produce little more thanburning eyes and coughing in humans, and exceptin the immediate vicinity of the Deepwater Horizonplume, the National Academy of Sciences reportconcluded, they never got close to acute toxicitythresholds for sea life living at the water’s surface.But biologists are still trying to figure out the long-term threat to human and ecosystem health posedby millions of liters of the stuff combined with un-known quantities of crude oil laced with its ownbrew of toxins and carcinogens.

Underlying it all is a mystery: where did the oilactually go, and how did the dispersants affect itsmovements? Before thewellheadwas finally capped onJuly 15, 2010, it had released an estimated 760 millionliters of oil—and as much as 25% of it remainsunaccounted for.

Although definitive answers are hard to come by,major clues have emerged in the years since theaccident as researchers have studied the real-worldphysics of oil, water, and dispersants. They haveanalyzed and reanalyzed the data recorded duringthe disaster, studied oil-droplet formation in thelaboratory (with and without dispersants), trackedcurrents in the Gulf with fleets of high-tech buoys,and constructed innumerable computer simula-tions. Researchers know vastly more than they oncedid about what happened to the oil in the deep seaplume as it rose from the wellhead; how the oilinteracted with sunlight, wind, and waves as itspread across the surface; and exactly what rolethe dispersants played.

And in June 2018, researchers embarked on thelargest experimental simulation of the Deepwater Ho-rizon spill to date at a huge saltwater tank in NewJersey. In the two-phase experiment, which will con-clude with a second series of experiments in July2019, the scientists will gather a trove of data in hopesof pinning down some of the last remaining uncer-tainties stemming from a disaster whose scale andspeed took everyone by surprise.

On the SurfaceThe real-world chemistry and physics of the air-seainterface are about as complicated as it gets. As soonas oil from any spill hits the surface, for example, itstarts baking in the sun, boiling off volatile compoundsand losing almost half its volume as it turns into a tarrygunk that resists dispersant action (2). The fumes werebad news for the Deepwater Horizon cleanup crews;not only were the gases a fire hazard but also theyincluded some 40 times the allowed exposure levelsfor benzene, a known carcinogen. As much as 25% ofthe oil in that incident seems to have evaporated inthis way.

In addition, explains Eric D’Asaro, an oceanogra-pher at the University of Washington in Seattle, thesurface of the ocean isn’t like a flat puddle of rain-water. It moves, surges, and heaves. Breaking wavesand ocean currents are constantly shattering the oil

An airplane releases oil dispersant over oil from the Deepwater Horizondisaster off the shores of Louisiana in May 2010. All told, about 3 million litersof dispersant was used on the spill. Image credit: Science Source/UnitedStates Coast Guard.

10604 | www.pnas.org/cgi/doi/10.1073/pnas.1907155116 Waldrop

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

slicks back into droplets and dragging them underagain, he says, “until there’s an equilibrium betweenthings that are carried up and carried down.” Thefinest droplets go deepest, says D’Asaro, who’s amember of the Consortium for Advanced Researchon Transport of Hydrocarbon in the Environment(CARTHE). This means that the so-called oil slick is ac-tually a thick layer of oil droplets extending down asmuch as 10 meters.

Dispersants add another level of complexity (seeFig. 1), says Joseph Katz, a mechanical engineer atJohns Hopkins University in Baltimore, MD, whostudies the effects of these chemicals with fundingfrom a consortium funded by the Gulf of Mexico Re-search Initiative, which separately funds CARTHE.He works with a laboratory wave tank that allows himto introduce oil slicks and then watch through asystem of lasers and microscopes as the breakerssmash the slicks into an underwater cloud of oildroplets.

“Without dispersants,” says Katz, “I found the sizedistribution to be understandable.” That is, the dropletsshowed a range of sizes down to about 100 microme-ters, or about as small as a turbulent eddy can get be-fore it’s dissipated by fluid viscosity. “But withdispersants, I couldn’t predict the distribution,” he says.Instead of a cutoff at 100 micrometers, he saw dropletsas small as 1 micrometer (3).

A closer look showed what was happening, saysKatz: in the presence of dispersants, which lower thesurface tension between oil and water, the dropletswere developing all sorts of threads and tails. “Theylook like sperm cells,” he says. In fact, the dispersantswere concentrating in the tails, which would growlonger and longer until they broke up to producethe microdroplets.

Above the surface, Katz found that dispersantscause a 100-fold increase in the concentration of ultra-fine oil droplets floating in the air (4). It’s less clear howthese floating droplets form—the popping of bub-bles, maybe?—but their presence raises new healthconcerns: what happens when people breathe ininfinitesimal droplets that are filled with toxinsand carcinogens from the oil and are so small thatthey can penetrate deep into the lungs? “Go tothe literature, and you find we don’t know much,”says Katz.

Adding still more complexity are the currents thatstir the Gulf on every size scale, from local riptides atthe beach to the giant Loop Current: a powerful flowthat rises between the Yucatan Peninsula and Cuba,wanders around the Gulf in an erratic and hard-to-predict path, and finally exits between Cuba andFlorida to become the Gulf Stream. One nightmarescenario during the Deepwater Horizon incident wasthat the Loop Current would capture the spreadingoil slick and end up fouling a good chunk of Floridaor conceivably even the East Coast. That this didn’thappen was purely a stroke of luck: the Loop Currentwas flowing south of the Deepwater Horizon site atthe time of the accident and was in the process of

spinning off a giant eddy that kept the oil relativelyclose to shore.

However, that simply meant that the fate of theDeepwater Horizon oil was subject to a host of poorlyunderstood, smaller-scale currents. In August 2012,CARTHE members sought to map those flows inunprecedented detail with the Grand LAgrang-ian Deployment (GLAD)—an experiment that seededthe blowout region with 317 custom-made floatsdesigned to drift with the currents the way oil would, andthen tracked them via GPS for 10 days (5). GLAD was thelargest experiment of its kind ever conducted untilearly 2016, when the consortium followed up withmore than 1,000 drifters in the Lagrangian sub-mesoscale experiment (LASER) (6).

In both cases, says D’Asaro, the drifter paths showedthe currents very clearly. But strikingly, he says, “wefound that sometimes there were places where thedrifters gathered together”—typically at a junctionbetween waters of different density.

One recurring example is at the mouth of theMississippi River, he says. “There is a fan of fresh watercoming out, making a rather sharp boundary with the

Fig. 1. Dispersants consist of surfactant molecules composed of a hydrophilichead group and a lipophilic tail (A). In seawater and oil, the hydrophiliccomponent turns toward the seawater and the lipophilic side toward the oilphase, spurring the formation of small oil droplets (B). Dispersants break up oilslicks, sending dispersant-stabilized oil droplets into the water column (C ).Reprinted by permission of ref. 10, Springer Nature: Nature ReviewsMicrobiology, copyright 2015.

Waldrop PNAS | May 28, 2019 | vol. 116 | no. 22 | 10605

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

saltwater in the ocean.” Salt water is heavier, so it di-ves underneath and creates a “front” that can collectfloating things.

During an oil spill, says D’Asaro, that can be goodnews or bad news: “If there is oil on the salty side, itwill be prevented from going on shore. But if the frontintersects the shore, it will become a conduit for theoil.” Either way, he adds, modelers need to learn howto predict these fronts so that clean-up crews in futureoil spills will know the best places to pick the stuff up.

In the PlumeMeanwhile, another group of CARTHE investigatorswas finding a whole new set of complexities in theplume of oil and gas rising from the Gulf seafloor.

“Think of it like a volcanic eruption, where the heatand force of explosion [send] the rock and hot gaseshigh into the atmosphere,” says Scott Socolofsky, acivil engineer at Texas A&M University in CollegeStation. The heat and force were there in plenty, saysSocolofsky. From a combination of observation andexperiment, as well as a detailed computer model of

the plume (7) that incorporated factors such as fluiddynamics, the buoyancy of oil and gas, and their sol-ubility in seawater, Socolofsky and other researchersknow that what came roaring up the broken drill pipewas a 100 °C, high-pressure mix of oil and natural gasthat abruptly decompressed as it slammed into thefrigid, 4 °C bottom waters of the Gulf. The gas reactedlike the fizz from a shaken soda can, flashing into amass of bubbles that helped break the oil into a cloudof fine droplets. And from there, says Socolofsky, “asthe gas bubbles and oil droplets started to rise, theirlightness created a plume that entrained the ambientseawater and carried it along with them.”

But bubbles and droplets have only a limited ca-pacity to lift the dense bottom water, says Socolofsky.So at a certain point, he says, “they started getting offthis upward rising train: ‘This is as high as I can go.’” InDeepwater Horizon, this happened at a depth ofroughly 1,100 meters, or about 400 meters above theseafloor. The smaller drops and dissolved compoundsspread out into an intrusion—a kind of underwatermushroom cloud that was very dilute and hard to seedirectly but that was detected from chemical traces.

Meanwhile, says Socolofsky, the larger dropletsand bubbles kept rising. But they didn’t make it to thesurface, either, because the gas inside them steadilydissolved into the surrounding seawater as they rose.So did everything soluble in the oil droplets: his plumemodel estimates that 27% of the original mass of theoil disappeared in this way.

The model also suggests that the dispersants in-jected at the wellhead enhanced this effect byshrinking the droplet and bubble sizes by about athird, which increased the surface-to-volume ratio andmade it easier for volatiles such as benzene to dissolveon the way up. That didn’t appreciably cut downthe total amount of oil reaching the surface, saysSocolofsky, but it definitely improved the air quality forthe cleanup crews. “The workers’ respirator alarms quitgoing off,” he says.

In short, says Socolofsky, researchers now have agood general understanding of what the plumelooked like. Unfortunately, he adds, “that doesn’tanswer the question of where the oil went.” For that,he says, you’d need to calibrate the models with theactual size distribution of the oil droplets coming outof the wellhead, with and without dispersants. “That’sa challenging measurement,” he says, “and it was notdone on Deepwater Horizon.”

Frustratingly, it’s also a measurement that’s almostimpossible to make in the laboratory. Droplet forma-tion depends on the surface tension between the oiland water, which doesn’t scale. So to reproduce thefull range of droplets coming out of the 50-centimeterDeepwater Horizon pipe, an experimenter wouldneed a model pipe at least twice the size of the largeststable oil droplets, which are about 12 millimetersacross. (Anything bigger will quickly break up fromunstable oscillations.) That works out to a minimumpipe size of roughly 25 millimeters. But a nozzle thatbig would fill up any lab-sized tank in minutes, turningit an impenetrable black. Most laboratory experimentsuse nozzles with diameters of 1 or 2 millimeters.

This uncertainty in the droplet size leaves plenty ofroom for interpretation. For example, University ofMiami oceanographer Claire Paris and her collabora-tors have created their own model of the plume (8). Itincorporates much of the same chemistry and physicsas the model developed by Socolofsky and his co-workers, including factors such as solubility andbuoyancy. But it uses different experimental data,suggesting that the violence of the eruption from thewellhead smashed the oil into droplets so small thatthe dispersants couldn’t have made themmuch smaller.And if that was the case, says Paris, “the injection ofdispersants did not significantly change the amount ofoil that reaches the surface. Maybe 3%.”

Complicating things still further is the presence ofall that gas in the outflow, mainly methane, ethane,and propane. CARTHE member Michel Boufadel, anenvironmental engineer at the New Jersey Institute ofTechnology in Newark, recently worked withÖzgökmenand several other colleagues to reanalyze the Deep-water Horizon data (9) and concluded that therehad been a lot more gas in the jet than people hadassumed. “It was not just big blobs of gas, but a veryviolent tumbling and churning,” says Boufadel, whowas a member of the panel that prepared the NationalAcademy of Sciences’ recent consensus report ondispersants. So who knows what really happenedwhen dispersants were injected into this maelstrom.“Dispersants like to stay at the interfaces,” he says. So

“Think of it like a volcanic eruption, where the heat andforce of explosion [send] the rock and hot gases high intothe atmosphere.”

—Scott Socolofsky

10606 | www.pnas.org/cgi/doi/10.1073/pnas.1907155116 Waldrop

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

maybe they were reacting with the gas all along andnot the oil. “There are not many experiments, or evenmodels for this kind of churn flow,” he says.

To sort all this out, says Boufadel, “we need a full-scale experiment.”

What’s Left in the TankThat’s where the huge tank in New Jersey comes in.By US law, says Özgökmen, you can’t put oil in theocean even for an experiment. So the researchershave turned to the Ohmsett facility, an above-groundsaltwater tank operated by the US Interior Departmentin Leonardo, NJ. Roughly the size of four Olympic-length swimming pools placed end to end, the facil-ity is designed for testing oil cleanup methods. But aCARTHE team led by Boufadel took over the tankJune 18–29, 2018 in the first phase of their effort torecreate the Deepwater Horizon disaster at somethingapproaching full scale.

The focus in this initial phase was to nail down thedynamics of droplet formation without dispersants.For each run, Boufadel and his colleagues injectedsome 10 tons of oil through a pipe that was beingtowed along the bottom of the tank to simulate cur-rent flow. The pipe was 25 millimeters across, bigenough to generate the full range of droplet sizes,which the researchers measured with a camera thatwas developed for the task at the University of Miami

and that could image even very small droplets over awide range of distances.

The results are still being prepared for publication,says Boufadel. But the data taken so far cover the fullgamut of conditions, including churn flow with 50%/50% oil and gas, bubbly flow with 5–10% gas, andsmooth flow with no gas. In the experiment’s secondrun in July 2019, the group will measure how dropletformation is affected under each condition by differ-ent levels of dispersants.

Hopefully, the results will clear up some of theuncertainties about where the oil went after Deep-water Horizon. But it will definitely be a culmination ofCARTHE’s work on the accident, says Boufadel. Likeall the other Gulf of Mexico Research Initiative con-sortia, the group is now moving into a data consoli-dation phase that’s geared toward integrating theimmense amount of science that’s been done on theGulf of Mexico—and improving the computer modelsthat response teams will use in the next oil spill.

That’s when, not if, says Boufadel. Future blowoutsmay or may not be as inaccessible as Deepwater Ho-rizon, he says—although with oil companies drilling indeeper and deeper waters around the world, that’salways a possibility. “But there are a lot of pipes un-derwater,” he says. “And if you have an oil release, itdoesn’t have to be a mile below the surface.”

Since Deepwater Horizon, says Boufadel, “it’sgood to be ready.”

1 National Academies of Sciences, Engineering, and Medicine, The Use of Dispersants in Marine Oil Spill Response (NationalAcademies Press, Washington, DC, 2019).

2 C. P.Ward, C. J. Armstrong, R. N. Conmy, D. P. French-McCay, C. M. Reddy, Photochemical oxidation of oil reduced the effectivenessof aerial dispersants applied in response to the Deepwater Horizon spill. Environ. Sci. Technol. Lett. 5, 226–231 (2018).

3 C. Li, J. Miller, J. Wang, S. S. Koley, J. Katz, Size distribution and dispersion of droplets generated by impingement of breaking waveson oil slicks: Oil droplets generated by breaking waves. J. Geophys. Res. Oceans 122, 7938–7957 (2017).

4 N. Afshar-Mohajer, C. Li, A. M. Rule, J. Katz, K. Koehler, A laboratory study of particulate and gaseous emissions from crude oil andcrude oil-dispersant contaminated seawater due to breaking waves. Atmos. Environ. 179, 177–186 (2018).

5 A. C. Poje et al., Submesoscale dispersion in the vicinity of the Deepwater Horizon spill. Proc. Natl. Acad. Sci. U.S.A. 111, 12693–12698 (2014).

6 E. A. D’Asaro et al., Ocean convergence and the dispersion of flotsam. Proc. Natl. Acad. Sci. U.S.A. 115, 1162–1167 (2018).7 J. Gros et al., Petroleum dynamics in the sea and influence of subsea dispersant injection duringDeepwater Horizon. Proc. Natl. Acad.Sci. U.S.A. 114, 10065–10070 (2017).

8 C. B. Paris et al., Evolution of the Macondo well blowout: Simulating the effects of the circulation and synthetic dispersants on thesubsea oil transport. Environ. Sci. Technol. 46, 13293–13302 (2012).

9 M. C. Boufadel et al., Was the Deepwater Horizon well discharge churn flow? Implications on the estimation of the oil discharge anddroplet size distribution. Geophys. Res. Lett. 45, 2396–2403 (2018).

10 S. Kleindienst, J.H. Paul, S. B. Joye, Using dispersants after oil spills: Impacts on the composition and activity of microbialcommunities. Nat. Rev. Microbiol. 13, 388–396 (2015).

Waldrop PNAS | May 28, 2019 | vol. 116 | no. 22 | 10607

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1