During this period, a group of Cambridge University students in white gloves were poring over library archives and ships’ logs. Brought together by Dave Gubbins, their supervisor, their mission was to retrieve and collate every scrap of declination, inclination and intensity data ever recorded.
The students painstakingly pieced together the history of Earth’s magnetic field as far back as 1590. Using these data they then calculated the field at the boundary of the mantle and the core. The results showed that the magnetic field at the core–mantle boundary was much more complex than the field observed at the Earth’s surface. Amid the general complexity, one feature stood out: much as Halley and Hansteen had observed earlier for the surface field, there appeared to be not two but four locations at which the magnetic field lines were concentrated. Two were in the northern hemisphere, where the field lines entered the core, and two in the southern hemisphere, where the field was outwards. None of these locations lay on Earth’s rotation axis, which indicated there was a complicated pattern of convective motion within the outer core. This was another feature against which geodynamo models had to be tested.
In 1986 Englishman Paul Roberts and American Gary Glatzmaier met and began working together on the geodynamo problem. Roberts had recently joined the University of California at Los Angeles (UCLA) as professor of mathematics and geophysical sciences. He would later describe himself as “a doyen of geomagnetic theory, one of the old fogies of the subject:” his work in geomagnetism went back to the days of Elsasser and Bullard.
At the outset of Roberts” graduate studies at Cambridge University in the early 1950s, his supervisor had proposed he should “either prove that fluid dynamos could not exist, or find a working model.” Bamboozled by the magnitude of the demand, he had promptly changed his thesis topic to the slightly more tractable problem of the origin of the secular variation and found a new supervisor—none other than Keith Runcorn, who was presumably delighted to take on yet another such able student.
Roberts’ enthusiasm for the geodynamo problem as a whole had been reignited in the 1960s, by which time he was working in the geomagnetism group at Newcastle upon Tyne and news was leaking out from behind the Iron Curtain that a Soviet geophysicist, Stanislav Braginsky, had cracked some of the long-standing issues and discovered a set of almost symmetrical solutions. Meantime, new results on the magnetic fields of the sun and other stars were accumulating thick and fast, and it was now known that Jupiter too had a strong dipolar field, and presumably an internal dynamo. Roberts began to work tirelessly to advance the theory of kinematic dynamos, to understand the balance between the various forces acting on the core fluid—iron-rich metal in the case of the Earth, but “metallic” hydrogen in the case of the sun and Jupiter. Once again, he was drawn towards the grail of deriving a fully self-consistent homogeneous magnetohydrodynamic model of Earth’s core—with no prior assumptions regarding fluid flow, and minimal simplifications.
When Roberts arrived at UCLA, Gary Glatzmaier had just finished writing his PhD thesis—on convection in a rotating spherical shell —at the University of Colorado, and had joined the United States National Laboratory at Los Alamos in New Mexico, where he was engaged in modeling magnetohydrodynamic processes in the sun. Computers had come a very long way since Bullard had first used the mechanical differential analyzer to solve his differential equations and since Vine and Matthews had fed their seafloor models on paper tape into one of its early electronic successors, the Edsac 2.
In 1995, after several years of collaboration, Glatzmaier and Roberts seized the opportunity to harness the unprecedented speed and capacity of the new Cray C90 supercomputer at Pittsburgh University. Their goal was to produce a computer simulation of Earth’s magnetic field that went through all the processes and reproduced all the features that had been documented—not just from the four hundred years of direct, first-hand observation, but from the paleomagnetic records that stretched back to the beginning of the planet’s history. Their results would stun the world of geophysics.
To make the computer mimic all the complex processes taking place in Earth’s core, Glatzmaier and Roberts programed it to calculate temperature, pressure and density, the velocity of the fluid, and the magnetic field by simultaneously solving all the equations of the magnetohydrodynamic problem. Using spherical geometry— latitude, longitude and radius—the computational problem was similar in principle to, but enormously larger than, the one that Gauss’s team of “calculators” had taken on in the 1830s.
A single solution, a snapshot for one instant of time, would not be sufficient. Since the geodynamo evolves with time, the equations had to be solved over and over, stepping forward by the equivalent of twenty days each time. And as the solutions corresponding to different times depended on each other, it was necessary to look backwards and forwards at each step of the calculation. It was this more than anything else that took up so much computer time: the exercise would have been all but impossible before the advent of the supercomputer.
Finally, the solutions had to be converted to values of temperature, pressure, density, fluid velocity and magnetic fields at locations on a huge spatial grid: at a total of 100,352 points in the outer core (49 different radii at each of 32 latitudes and 64 longitudes) and 34,816 points in the inner core (17 radii x 32 latitudes x 64 longitudes).
Initially, Glatzmaier and Roberts ran their program for the equivalent of 40,000 years. At twenty-day intervals, this amounted to 730,000 steps, and a total of 100,000 million calculations for each component. The process took over a year, and required 2000 hours of central processor time on the Pittsburgh supercomputer.
The results were incredible. Graphic representations of Earth’s magnetic field, such as the one reproduced on page 234, would become icons of late twentieth-century science and grace the covers of numerous books and journals, including Science and Nature.
After the equivalent of about 10,000 years, the model settled so that the field at the Earth’s surface was predominantly dipolar. So far, so good. However, the image of the field emerging from the core–mantle boundary showed the surface field was just a smoothed version of a much more complicated picture.
At the core–mantle boundary, bundles of magnetic field lines emerge at high latitudes, but just as Gubbins’ students had found in their extrapolation of historical data to this boundary, there are two bundles in each hemisphere, and they are not actually at the poles. In fact, at the poles of the core–mantle boundary the field is relatively weak.
Within the core, the picture is even more complicated. As well as the field lines emerging from the core, which make up what is known as the “poloidal” field, others are wrapped up inside the core like tangled spaghetti. This “toroidal” field is locked within the core, and since its field lines never penetrate the core–mantle boundary, it is impossible to detect from outside. As time progresses, the twisting and shearing caused by convection and the rotation of the Earth convert toroidal field lines into poloidal field lines and vice versa—essential processes in the operation of the geodynamo.
A snapshot of the magnetic field lines inside and near Earth’s surface during a period of stable, normal polarity, as modeled by Glatzmaier and Roberts. The inner circle represents the core–mantle boundary, and the outer circle the surface of Earth. White and grey lines represent inward-directed and outward-directed magnetic field lines respectively. Outside the core the familiar dipole field predominates, with field lines exiting the core and Earth’s surface in the southern hemisphere, and looping round to re-enter the surface and core in the northern hemisphere. Within the core, the additional “toroidal” magnetic field makes the pattern much more complicated.
Gary Glatzmaier (left) and Paul Roberts. The two men began working together in 1986 on the problem of how to create an accurate model of the processes taking place in Earth’s core. In 1995 they seized the opportunity to use a new supercomputer at Pittsburgh University to carry out calculations that would previously hav
e been impossible, and ultimately solved the mystery of what causes Earth’s magnetic field.
At the Earth’s surface, the simulated magnetic field goes through changes that look very much like observed secular variation, with features growing and decaying, while generally drifting westward at a rate close to the 0.2 degrees a year that Bullard measured in the early 1950s. This, however, belies what is going on in the outer core, where the field is quite unstable, and in fact regularly tries to reverse polarity. Most attempts to reverse polarity are unsuccessful only because the inner-core fails to follow suit: it has a longer magnetic decay time, and so takes longer to respond to the driving field from the outer core. More often than not, the outer core “changes its mind” and flips back before the inner-core field can react, and so a polarity reversal does not occur. In this way, the electrically conducting properties of the inner core has a crucial stabilizing effect on the geomagnetic field as a whole.
Glatzmaier and Roberts’ simulation of Earth’s magnetic field during a polarity reversal from reversed to normal. In each case, the inner circle represents the core–mantle boundary and the outer circle the surface of Earth. In (a) the field is still predominantly dipolar, but already tilted significantly from the geographic rotation axis. In (b) the pattern of field lines is more complicated, and the field intensity at Earth’s surface has weakened considerably. In (c) the dipole is beginning to regenerate in the opposite (normal) direction, although a complex field structure persists, particularly in the southern hemisphere.
Nonetheless, it was right at the end of its run that Glatzmaier and Roberts’ computer simulation produced its most startling event. After the equivalent of nearly 40,000 years, the outer-core field destabilized and waited just long enough for the inner-core field to reverse too. Instead of regenerating in the original direction, the whole magnetic field flipped and grew back in the opposite polarity. The computer model had undergone a full geomagnetic polarity reversal.
Glatzmaier and Roberts’ dream had come true. The simulation had maintained a stable dipole field for 40,000 years. It had mimicked secular variation in every required manner. It seemed to have tried to reverse several times but had almost always reverted to its original polarity. And eventually, just once, it had succeeded in executing a full polarity reversal.
The two men had answered Einstein’s ninety-year-old challenge. In landmark papers in the journals Nature and Physics of the Earth and Planetary Interiors, they convinced the world of science that electric currents in the Earth’s molten, iron-rich outer core, brought about by the combined churning effects of convection and the planet’s rotation, were all that was needed to account for everything known about Earth’s magnetic field—that and a very large and powerful computer.
Epilogue
The Earth is not simple … Significant breakthroughs … will come from young people who are inspired to work on what others proclaim is impossible.
—GARY GLATZMAIER, 2002
Amazing as Gary Glatzmaier and Paul Roberts’ achievement was, it would mark not an end but the beginning of a new era of Earth science, one in which the geodynamo was no longer mere theory. Their computer model had demonstrated that a magnetohydrodynamic dynamo that obeyed Maxwell’s equations of electromagnetism and Navier and Stokes’ equations of fluid dynamics could generate an Earthlike magnetic field. Their 1996 papers unleashed a whole new burst of enthusiasm and activity.
First, the world watched closely as Glatzmaier and Roberts’ model chugged through more and more computer time, but after the equivalent of 300,000 years it had still not achieved another polarity reversal. This was not altogether unexpected: in real time it has been 780,000 years since the last documented reversal, and if paleomagnetism has told us anything it is that the past is no indication of the future. Polarity reversals occur randomly, and the intervals between them range from a few thousand to tens of millions of years. Perhaps Glatzmaier and Roberts had been lucky the first time round.
But what about the repeated attempts of their model’s outer-core field to reverse that had been thwarted by the failure of the inner-core to follow suit? Examination of paleomagnetic records has revealed there have been several major “excursions” from the stable magnetic field direction, much bigger than regular secular variation but not always recorded worldwide, and not eventuating in full polarity reversals. Examples are the so-called Laschamp excursion, recorded in 40,000-year-old lava flows in the Chaîne des Puys region of France, and the Mono Lake excursion, thought to have occurred in California about 25,000 years ago. Although replicated and credited by many as genuine features of the geomagnetic field, neither the Laschamp nor Mono Lake excursions had been found in paleomagnetic records from sites some distance away. The significance of these regional excursions had proved elusive, but now here was an explanation: attempts of the outer core to reverse polarity that had been aborted because the inner core had failed to follow suit.
The picture was, however, still far from complete; many questions remained unanswered. Glatzmaier and Roberts’ model had been successful, but the limitations of computing power had forced them to assign unrealistic values to some of the quantities. Other researchers now began to publish their own geodynamo models and gradually close the gap between computation and reality, but as Glatzmaier wrote in the 2002 Annual Review of Earth and Planetary Science:
Recent geodynamo simulations have provided new insights and predictions for convection and magnetic field generation in the Earth’s core … [but] … we still have a long way to go.
The computational costs of producing a fully realistic geodynamo model would, he pointed out, be staggering by today’s standards.
A major question remained about the source of the energy required to drive the dynamo—the mechanism behind convective motion in the outer core—and how it should be modeled. Compared to convection in the Earth’s virtually solid mantle, which is a sluggish affair resulting in seafloor spreading rates of only a few centimeters a year, convection in the outer core, where the fluid is thought to flow as easily as water, is up to a million times faster—millimeters per second, or tens of kilometers per year. What drives this motion?
There is one obvious source: the heat imparted to the core when the Earth first formed. A large enough temperature difference between the inner core and the core–mantle boundary would drive thermal convection. Heat supplied by decay of radioactive elements in the core would aid the process, but it is uncertain whether a sufficient source of radiogenic heat exists within the core.
In addition to this, there is a process known as “compositional convection.” The core is not only much denser than the mantle, it is also made of vastly different material. This, together with the fact that pressure increases enormously towards the center of the planet, has led geophysicists to believe that as the Earth has cooled from its original completely molten state, the inner core has gradually formed by “freezing” from the center outwards. The process is ongoing: the inner core presumably continues to grow as the Earth cools further and outer-core fluid progressively solidifies at the boundary between the outer and inner cores.
The inner core is also known to be denser than the outer core, and as outer-core material solidifies on to the inner core, lighter, albeit minor, components of the core fluid such as sulfur and oxygen are released. Just as a cork held under the surface of water will rise upwards when released, this lighter material, being buoyant, will rise naturally in the outer core—the movement known as compositional convection. Solidification also involves the release of heat (it is the opposite of melting, which requires heat input) so thermal convection will be further enhanced by this process.
Although scientists now agree that both thermal and compositional convection contribute energy to the geodynamo, the balance between these two processes, and how to incorporate them in models, is still intensely debated.
Another unsolved mystery is what actually happens during a polarity reversal. Computer simulations and paleomagneti
c records agree that the field intensity is reduced considerably. Does this mean that Earth’s cosmic ray shield is weakened, to the extent that it can no longer defend life on Earth from space’s bombardment of charged particles and radiation?
Some scientists believe this is the case, and is what has led to genetic mutations and extinctions of whole species. Is there evidence to correlate geomagnetic polarity reversals with such extinctions? Certainly not on a simple one-for-one basis. Mass extinctions— for example, the demise of the dinosaurs 65 million years ago, at the famous Cretaceous−Tertiary boundary—have not occurred close enough to reversals that we can validly claim cause and effect. Given the complex interrelations we know must exist between the various Earth processes, it is probably naïve to expect such a simplistic correlation.
In the computer models, polarity reversals occur at seemingly random intervals of time. This is compatible with statistical analyses of the geomagnetic polarity timescale, and is the outcome of what mathematicians call a chaotic process. Like Gauss, Glatzmaier and Roberts were content to let the computations tell the story. This is not the case with most of us: we need a physical picture on which to pin mathematical models, so we insist on searching for a physical trigger that will initiate the necessary changes in core fluid motions and make the field try to reverse.
One candidate is what Richard Muller, a physicist at the University of California, Berkeley, describes as an avalanche on the core–mantle boundary. Muller has argued that if a massive glob of the slushy material on the boundary were suddenly to slip back into the core it would seriously disrupt the pattern of convective flow, possibly enough to initiate a reversal. Such an avalanche, Muller suggests, could be triggered by the jolt of an astronomical impact—for example, a large asteroid hitting the Earth.
North Pole, South Pole: The Epic Quest to Solve the Great Mystery of Earth's Magnetism Page 21
