DO MANTLE PLUMES PURIFIED WATER LETS ARSENIC MIDCONTINENT RIFT EXIST? SNEAK INTO AQUIFERS IS A HYBRID BEAST EARTH

EXCAVATING

SNOWMASTODON This PDF is licensed for individual users at University of Wyoming Lib

January 2016 www.earthmagazine.org Feature THE QUESTION OF MANTLE PLUMES Sara E. Pratt

n 1963, J. Tuzo Wilson first pro- Since 1971, the plume hypothesis, the Galápagos Islands, the Azores, the posed that volcanic chains like although never universally accepted, has Canaries and the Afar region of Africa. the Hawaiian Islands form when become the most widely held explana- Over the years, the hypothesis has a tectonic plate drifts over a “hot tion for so-called anomalous volcanism continued to be debated and studied, Ispot” in the mantle. Eight years later, — the type that occurs far from plate resulting in thousands of journal articles. Princeton geophysicist W. Jason Mor- boundaries, like in Hawaii and Yellow- Many of these have described significant gan suggested that such hot spots — he stone, or in excessive amounts along variations on the original hypothesis, yet initially proposed about 20 around the mid-ocean ridges, as in Iceland. In still refer to “plumes” as the underlying world — were fueled by narrow plumes addition to Hawaii, Yellowstone and phenomenon, which has introduced an of hot mantle rock rising from the Iceland, other notable examples include element of semantics to the debate (see core-mantle boundary. Pitcairn Island, Macdonald Seamount, sidebar, page 40). Other researchers

Iceland NORTH AMERICAN EURASIAN PLATE PLATE Eifel Cobb Seamount Azores Yellowstone National Park ARABIAN PLATE PACIFIC PLATE Canary Islands Hawaii COCOS PLATE Cape Verde Afar CARIBBEAN AFRICAN Galápagos PLATE Samoa PLATE Reunion Pitcairn Island Islands NAZCA St. Helena PLATE INDO-AUSTRALIAN SOUTH PLATE Easter Island AMERICAN Macdonald PLATE Prince Edward Seamount Islands Tristan da Cunha Amsterdam Island Bouvet Island This PDF is licensed for individual users at University of Wyoming Lib ANTARCTIC PLATE Balleny Islands

Plate boundary Oceanic plateau Proposed hot spot

Subduction zone Volcanic ridge Over the years, many different locations have been proposed as hot spots. Some of the most well-known pro- posed sites are labeled with red dots in this diagram, which also depicts the borders of Earth’s tectonic plates. Credit: K. Cantner, AGI

page 36 January 2016 EARTH www.earthmagazine.org Feature

suggest plumes are not required at all A Textbook Case? beneath Hawaii thus became something to explain anomalous volcanism. They Morgan proposed his hypothesis in the of a holy grail for mantle researchers. instead propose a “plate hypothesis” wake of the plate tectonics revolution, when in which lithospheric stretching allows geologists were still struggling to explain already melted rock to escape from the anomalous volcanism. Most of Earth’s Can We See the Mantle? mantle to the surface. volcanism occurs on plate boundaries: In the late 1970s and early 1980s, some Despite the debate, Morgan’s hypoth- on mid-ocean ridges or above subduction measurements of ratios of helium-3 to esis, in nearly its original form, has zones, like those that surround the Pacific helium-4 in Hawaiian basalts and else- become entrenched in undergraduate Ocean in the so-called Ring of Fire. But where were discovered to be much higher and high school curricula. But although the cause of volcanism far from a plate than those in mid-ocean ridge basalts. the elegant explanation — often demon- boundary was a conundrum. Because most helium-3 was formed at strated with common household items, In Morgan’s conception, plumes were the same time as Earth about 4.5 billion like a pencil or a candle being used to chimneys of warm, buoyant rock about 100 years ago, it is called a “primordial” iso- punch or burn through a piece of paper to 200 kilometers wide that were rooted tope. It is depleted in surface material — has fascinated generations of stu- at the core-mantle boundary. These nar- because helium escapes into space. Thus, dents, the difficulties of confirming such row conduits of deep-mantle material rise the high ratios in Hawaiian basalts were a hypothesis and the intricacies of the through the solid mantle before spreading interpreted as evidence that plumes are ongoing debate are rarely included in out laterally, like a thunderhead, in the fed by primordial material from deep in the lesson. upper asthenosphere — the If the vast body of mantle plume ductile zone of the upper IF THE VAST BODY OF MANTLE PLUME research has done nothing else, it has mantle that lies below the RESEARCH HAS DONE NOTHING ELSE, revealed the difficulties inherent in trying brittle lithosphere. From IT HAS REVEALED THE DIFFICULTIES to plumb the depths of Earth’s interior. there, they can cause the INHERENT IN TRYING TO PLUMB THE Reaching to a depth of 2,900 kilometers, lithosphere to swell and the mantle cannot be sampled by field- shear, disgorge massive DEPTHS OF EARTH’S INTERIOR. work; it must be remotely sensed and flood basalts, and form modeled. What little we know about the age-progressive volcano chains. the mantle, while mid-ocean ridge sys- mantle’s composition and structure has In addition to having deep roots tems tap recycled upper mantle material been gleaned from geochemical analyses and high temperatures relative to sur- depleted in helium-3. of deep-sea lavas or the rare chunk of rounding mantle rock, Morgan’s other Beginning in the 1980s, seismologists, exhumed mantle rock, and from inter- fundamental criteria were that plumes led by Adam Dziewonski at Harvard, pretation of seismic waves that have transport primordial mantle material began developing new techniques that traveled through the deep Earth. from below the zone of active convection; took advantage of new computing capacity The invention in the late 1970s and are fixed relative to one another; produce and technologies like digitization to build early 1980s of seismic tomography — time-progressive volcanic chains; break upon earlier efforts to use earthquake seis- the use of earthquake seismic waves up continents; and drive plate tectonics. mic waves to image the three-dimensional to image the three-dimensional struc- Morgan formulated his hypothesis structure of Earth. Dubbed “teleseismic ture of the mantle, much like X-rays are around the formation of the Hawaiian tomography” by Dziewonski and the late used to produce medical CT scans of the Island-Emperor Seamount chain in the Caltech seismologist Don L. Anderson in human body — offered a promising path middle of the Pacific Plate. The islands and 1984, the new technique was soon applied toward understanding the core-mantle seamounts exhibit age progression, with in mantle plume studies. This PDF is licensed for individual users at University of Wyoming Lib boundary in more detail. But the method the youngest near present-day Hawaii and Seismic tomography works by measur- has so far proven to be of limited use the oldest near the Aleutian Trench, which ing the travel times of earthquake waves when it comes to visualizing small-scale Morgan suggested was indicative of a plate as they arrive at various stations around features, as plumes are assumed to be. moving over a stationary hot spot. The the world. Waves travel faster through cool However, new models of the mantle that bend in the chain, he suggested, indicated rock than through warm rock. Thus, faster rely on the immense number-crunching that the Pacific Plate changed direction travel times are assumed to indicate zones capacity of supercomputers may offer roughly 47 million years ago. of relatively cool, high-density rock that is the clearest picture yet and move the The island chain became the textbook sinking in the mantle, whereas low-veloc- mantle-plume debate forward (see side- example of a mantle plume hot spot. ity zones are interpreted to indicate hot, bar, page 40). And, confirming the existence of a plume low-density rock that is rising, like mantle

page 37 January 2016 EARTH www.earthmagazine.org Feature

A schematic cross-section of Earth showing the plume hypothesis ver- sus the plate hypothesis. Credit: Anderson, GSA Special Papers 388, 2005

narrow, thus making their detection challenging. Additionally, if seismome- ters are closely spaced in an array, the aperture of their “view” down into the earth is narrowed. This can be problem- atic on ocean islands like Hawaii and Iceland, which have limited land area on which to deploy seismometers. In the late 1990s and early 2000s, several projects — among them SWELL (Seismic Wave Exploration of the Lower Lithosphere) and PLUME (Plume-Lith- osphere Undersea Mantle Experiment) — deployed ocean-bottom seismome- ters in an attempt to answer questions about the structure of the lithosphere and mantle beneath Hawaii. But the plume studies came to various, often conflicting, conclusions. Some critics of the plume hypothesis question if Hawaii was formed by a plume at all, suggesting instead that it could have resulted from purely lithospheric processes.

Do Plumes Exist? One of those critics was Anderson, plumes would be. Researchers translate the southern tip of Africa toward the who in the early 2000s, along with doz- these into maps of zones of upwelling and Afar region of northeast Africa; for- ens of like-minded colleagues including downwelling, which could help improve mally called a large low-shear-velocity Foulger, laid out the alternative plate our view of Earth’s interior and answer province, it is better known as a “super- hypothesis, which they proposed was longstanding questions about the con- plume.” There is another similar zone more consistent with the bulk of obser- vection cells that drive plate movement under the Pacific. But whether such vations collected to date. — including whether the whole mantle regions slow seismic waves because they Although Morgan initially presented convects or just the upper mantle. are hot, or just compositionally different mantle plumes as an assumption, over But there is a major drawback in cor- than surrounding mantle rock, cannot be time, the fact that they were an assump- relating seismic wave speeds with rock determined by tomography. tion — not an observation — has been This PDF is licensed for individual users at University of Wyoming Lib temperature: Mineralogical and chem- “These methods reveal the seismic forgotten, Foulger says. No plume has ical differences in a rock’s composition structure of the mantle,” says Gillian yet been found to satisfy all the criteria can also affect seismic velocities, as can Foulger, a geophysicist at Durham currently attributed to plumes, she says, the presence of partially melted rock. University in England, and a longtime adding that the hypothesis has become Determining whether low-velocity zones collaborator of Anderson’s. They “do not too flexible, with ad hoc variations represent thermal, physical or compo- reveal the geological structure. The two tacked on to accommodate any finding. sitional differences in the mantle has are not the same.” For example, in the 1990s, petrological become a debate of its own. Seismic tomography has other lim- and geochemical analyses revealed that One of the largest low-velocity itations as well. Seismic wavelengths are basalts of suspected plume origin dis- zones rises diagonally from beneath long and plumes are thought to be quite played a much wider range of geochemical

page 38 January 2016 EARTH www.earthmagazine.org Feature

signatures than initially thought, includ- ing helium isotope ratios in the range of depleted shallow mantle or lithospheric material. This prompted plume-hypoth- esis supporters to suggest that the source The age-progressive Hawaiian of the depleted plume material could be Island-Emperor Seamount Chain subducting slabs of upper mantle sinking is the textbook example of a into the deep mantle and becoming incor- mantle plume. porated into preexisting plumes. Credit: NGDC When paleomagnetic analysis revealed that volcanism at the Hawai- ian hot spot had migrated and changed chains and near-fixity — the science had the precise opposite of assumptions” in direction over the course of its exis- progressed to the point where “plumes the mantle plume model. “Hot spots such tence, and is therefore not fixed, some don’t have to meet any particular test.” as Hawaii, Samoa, Iceland and Yellow- researchers proposed that plumes are “Hence the hypothesis now always stone are due to a thermal bump in the distorted by convection in the mantle works with appropriate site-specific shallow mantle, a consequence of the and thus would not be where expected. modifications, but increasingly doesn’t cooling of the Earth.” Such “mantle wind” was also invoked to tell us anything or predict anything, The two theories thus differ on the explain the bend in the Hawaiian chain especially about structures formed in the source of the magma at sites of anoma- and the lack of age progression in other past,” he wrote. “It does, however, make it lous volcanism. Instead of deep-mantle island chains. harder to offer nonplume explanations.” material being drawn to the surface, the “Plumes have been proposed to come In the mid-2000s, several researchers plate hypothesis holds that the chemistry from almost any depth, to rise vertically attempted to start fresh, laying out a and volume of lavas are dependent upon or tilt, to flow for long distances laterally, new definition of a plume as a thermal the local mantle rock type and its ability to have narrow or broad conduits, to instability with a large, bulbous head to melt. “The asthenosphere is mostly have no plume head, one head, or multi- that is heated by the core, arises from near its solidus temperature, hence it is ple heads, to produce steady or variable the bottom of the mantle and is followed widely capable of melting wherever it is flow, to be long- or short-lived, to speed by a narrow tail. But many of the same depressurized by extension,” says Warren up or slow down, to have a source that issues remain. Hamilton, a geophysicist at the Colorado is either depleted, enriched, or both, School of Mines in Golden, Colo. and to have either high or low [ratios of Hawaii is a perfect example, Ham- helium-3 to helium-4],” Foulger wrote A Plate Explanation ilton says. Rather than the result of a in Geoscientist in May 2003. for Hawaii? mantle plume, it is a propagating exten- She wasn’t alone in her thinking that One of the main points raised by critics sional crack in the middle of the Pacific the hypothesis was becoming too broad. of the plume hypothesis is that it requires Plate, which is subducting under Japan In a letter to the editor of Geoscientist that two independent types of thermal in the west and diving under the Aleu- that year, Northwestern University seis- convection be operating in the mantle — tian Islands in the north faster than it is mologist Seth Stein wrote: “I wouldn’t one associated with plate tectonics and spreading at the East Pacific Rise. “The blame anyone for the state of thinking the other causing hot spot volcanism. It Hawaiian-Emperor southeastward-ad- … about hot spots and mantle plumes is much more likely, they say, that only vancing crackspot is a product of the except ourselves. As in Pogo’s dictum, one type — that associated with plate extensional stresses near the midline of This PDF is licensed for individual users at University of Wyoming Lib ‘We have met the enemy and they are us.’” tectonics — is at work, and it wouldn’t the North Pacific between subduction Stein added that in the “absence of produce narrow upwellings. under Asian systems on one side and any other clear model,” the geoscience In 2014, Anderson and James Natland, American ones on the other,” he says. community had accepted “very vague a geochemist at the University of Miami, It’s no coincidence, he adds, that ideas about plumes” and allowed them to wrote in Proceedings of the National the famous dogleg bend in the Hawai- become the “null hypothesis” for anom- Academy of Sciences that Archimedes’ ian-Emperor chain — formerly attributed alous volcanism. He noted that although principle of buoyancy and the laws of to a sudden 60-degree change 47 million the hypothesis initially entailed rigorous thermodynamics dictate that, in a cooling years ago in the direction of the plate as criteria — for example, the presence of a planet, “convection is composed of nar- it drifted over a hot spot — crosses the low-velocity zone, age-progressive island row downwellings and broad upwellings, Mendocino Transform Fault.

page 39 January 2016 EARTH www.earthmagazine.org Feature

“This fault marks a change of 30 mil- melting at ridges, mantle convection operating at different locales. “I suspect lion years in the age of the lithosphere, pulling the plates apart, and Hawaii as how people address the topic in class which represented a huge step in thick- a classic hot spot. I do mention that the varies depending on teaching philosophy ness and properties … 47 million years concept of hot spots being fixed points is — some of us try to discuss unresolved ago,” he says. “This requires that the an assumption that is not always agreed topics to motivate young people into the change [that created the bend in the upon or necessarily reliable. Personally, science,” Stein says. chain] was controlled from the top, not I also am not surprised that things are a “After all, if everything’s known, why from Earth’s core.” little more complicated in how the melt go into it?” In an essay published on mantle- finally makes its way to the surface than plumes.org, the website maintained the simple ‘cartoon’ models. But I’m not Pratt is the senior editor of EARTH. by Foulger to track materials and pub- sure that the complexity of the real world lications related to the plume debate, negates the basic model.” Hamilton and Anderson called the plume Kevin Stewart, a structural geologist hypothesis “zombie science” — a hypoth- and associate professor at the University What Lies Below? esis that, despite contradictory evidence of North Carolina at Chapel Hill, teaches echnological advances con- and the lack of supporting evidence, will that aspects of the plume hypothesis are T tinue to improve the resolution not die. being debated and raises some of the of our view of Earth’s interior, Mantle plumes have become dogma possible alternatives. “I do talk about but disagreement remains over that few researchers are willing to the idea that hot spots may represent what we’re viewing. In a recent contradict in a publish-or-perish envi- a fixed reference system and one way Nature paper titled, “Broad ronment, Hamilton says, especially in people explain this is with plumes orig- plumes rooted at the base of the times of tightening budgets. Those who inating from the core-mantle boundary,” Earth’s mantle beneath major hot should be questioning it are the “young he says. “But then I present some seis- spots,” Scott French, a computa- Turks,” he says, but first they, and the mic tomography data that show that tional scientist at the Department general public, have to learn that there the cartoon images of a thin, rising of Energy’s National Energy are alternatives to what they may have column of hot material traversing the Research Scientific Computing learned in school. entire mantle don’t really match a lot of Center at Lawrence Berkeley the tomography.” National Laboratory, and Bar- In discussing Hawaii, Stewart says, “I bara Romanowicz, a seismologist How Are Plumes Taught? talk about the possibility that if the melt at the University of California at At many colleges and universities, at Hawaii is being generated at shallower Berkeley, reported the develop- especially in entry-level geology courses, levels in the mantle due to plate tectonic ment of the most detailed model the classic model has long been pre- processes, like the shearing of the asthe- yet of the structure of the mantle. dominantly taught. This trend remains nosphere, then there are other ways to The three-dimensional tomo- largely intact, although at least some explain the bend. But I don’t go into it graphic model, similar to a medical say they’re presenting it increasingly as too deeply. The whole story behind a plate CT scan, revealed large swells of unsettled science. origin for hot spot volcanism is pretty what is likely hot mantle material “We touch upon [the plates versus complicated and the full explanation is with diameters roughly 1,000 kilome- plumes controversy] a bit in my upper far beyond the scope of my 101 class.” ters wide. The structures, dubbed division classes, but it is never debated At Northwestern, Stein teaches his plumes by the researchers, rise from as such. So, I present hot spots as plumes introductory geology students that the areas at the core-mantle bound- This PDF is licensed for individual users at University of Wyoming Lib from the deep mantle,” says igneous plume hypothesis for the age-progres- ary with strongly reduced seismic petrologist Erik Klemetti, an assistant sive Hawaiian Islands is an “attractive” velocities, which are embedded in professor at Denison University in Gran- explanation for mid-plate volcanism but larger zones with anomalously low ville, Ohio. “As for how deep, I say that’s it remains “controversial” since the data velocities that are up to 5,000 kilo- still being debated.” don’t always match some predictions. meters wide. Maya Tolstoy, a marine geophysicist He says he tells students it’s still unclear “Previous tomographic mod- and associate professor at Columbia whether hot spots are caused by plumes els have hinted at the existence University, says that she also doesn’t located as deep as the core-mantle of plumes, but it was hard to say touch on the debate in her introductory boundary or by localized upper-mantle unequivocally that they were classes. “I talk about decompression volcanism, or by either of those processes

page 40 January 2016 EARTH www.earthmagazine.org Feature

Scientists using a new tomographic model of Earth’s Hawaii mantle have identified 5,000-kilometer-wide swells of potentially hot material in the lower mantle that the researchers say confirm the existence of plumes. Samoa Credit: French and Romanowicz, Nature, 2015 there,” French says. ”With our model seismic wave- you can say ‘Yes, that definitely is form record.” a plume.’” The research- In the model, plumes were iden- ers used 3 million tified under Hawaii and Iceland but hours of super- not beneath Yellowstone. A few of computer time the structures seen in the model rise (with many Marquesas into the upper mantle — beneath processors run- Pitcairn Island, for example — but ning in parallel) Tahiti Pitcairn most do not get above depths of to hone their 1,000 kilometers. model of the Macdonald “They start to thin out in the upper fast and slow part of the mantle, and they mean- regions, which der and deflect,” Romanowicz said are interpreted in a statement released with the as warm and cool areas, of the few years after W. Jason Morgan new study. “So while the tops of the mantle until the model was able first suggested mantle plumes, one plumes are associated with hot spot to accurately simulate seismogram researcher sought clarification on volcanoes, [the larger plume bases] outputs that matched the real-world how plumes differed substantially are not always vertically under [the results from 273 strong earthquakes from the normal convective models hot spots].” that occurred over the past 20 years. already suggested by plate tecton- French says the remnants of the Previous tomographic models ics, writing: “It would be most helpful plumes in the upper mantle “would faced computing limitations that if someone would explain in terms be too narrow for us to see, even required simplification of the phys- that are meaningful to geophysicists with the techniques we use, which ics to the point that information was in what respects the conventional are limited in local resolution due lost, French says. “Supercomputing geological pictures of rising magma to the fact that we are imaging the allows us to treat the physics more differ from ‘a thermal plume.’” whole mantle.” exactly so we can get around these While what French and Roma- traditional impediments.” nowicz have found may not sound But whether the plumes actually like a traditional plume, French says Applying Supercomputers represent temperature differences the structures they identify may be to Seismograms and show convection remains related to large low-shear-velocity This new view of the structure of unknown. The standard seismological provinces, also called superplumes. Earth’s mantle arose from a model view is that low shear-wave veloci- “Those [low-velocity zones] are in that French and Romanowicz devel- ties indicate hotter material, French close juxtaposition to where we oped using wave-form inversion says, but they can also be caused see these plumes,” he says, “but — which, like typical seismic tomog- by compositional differences, which the plumes are anchored near the raphy, is based on a technique would also explain the plumes’ vast edges of those zones.” This PDF is licensed for individual users at University of Wyoming Lib similar to the statistical regression widths and long-term stability. New models like French and used to best fit data points to a line Romanowicz’s might help research- or curve. But their model differs in ers begin to reconcile what we see how it makes use of seismic data. A Plume by Any on Earth’s surface with what we “The difference between what we Other Name? can know of its interior. And, in the do and what most other researchers The new study emphasizes the role future, perhaps further technological do,” French says, “is that instead of of language in the plume debate, advances will help change the lan- using the arrival times of specific seismic which is no stranger to semantic guage of the debate itself. waves as our data, we use the entire arguments. As early as 1973, just a SEP

page 41 January 2016 EARTH www.earthmagazine.org