MANY things that scientists study are invisible to them. Announcements about the discovery of planets around other stars do not come because astronomers have seen the planets, although they would certainly love to. The discovery is more likely based on observations of the star's motion that indicate strong nearby gravitational forces. So it is with discoveries about the Earth's core and mantle. Because the deepest well ever drilled extends down just 12 kilometers, not even pricking the mantle, researchers have to employ indirect methods to study the Earth.
Using meteorites and seismological evidence as clues, scientists have known almost since the beginning of the century that the Earth has a solid, mostly iron, inner core and a molten outer core with a mantle and crust of rocky, silicate material. But for just as long they have been puzzled about how the core and mantle separated. The primordial planet Earth grew out of bits of gas and dust, aggregating over time into a larger, more solid body. Was there then some cataclysmic event billions of years ago that melted much of the planet, prompting the metals and silicates to separate as oil and water do? Or was the separation the result of a more gradual process, a trickling down of the denser molten metals between solid silicate mineral grains to the center of the Earth?
Recent research at Lawrence Livermore National Laboratory by geochemist William Minarik has helped to dispel the second, "trickle down," theory. Using the larger of Livermore's two multi-anvil presses to mimic the pressures and temperatures that exist deep in the Earth, he has shown that metals like those in the Earth's core could not have trickled down (see figure below).
The materials used in the experiments were olivine, a silicate mineral that makes up much of the Earth's upper mantle, and an iron-nickel-sulfur-oxygen combination to represent the core.

The multi-anvil press is a relatively rare research tool. Livermore's two presses have been used for a variety of material property studies, including diffusion and deformation of ceramics and metals, deep-focus earthquakes, and the high-pressure stability of mineral phases. The larger, 1,200-ton hydraulic press can produce pressures of 25 gigapascals (GPa), which is equivalent to 250,000 times the atmospheric pressure at sea level, or the pressure that occurs 700 kilometers deep in the Earth. In addition to pressing on the sample, the experiment passes an electric current through a furnace within the assembly to generate temperatures up to 2,200°C.
These experiments using the multi-anvil press generated a high pressure of 11 GPa, or 110,000 times the atmospheric pressure at sea level. This corresponds to 380 kilometers deep into the Earth, or pressures at the center of moons or asteroids 2,500 kilometers in radius (about the size of Mercury). The sample was also heated to a temperature of 1,500°C. Under those conditions, the metal melts and the olivine remains solid.
The geometry of the press is key to creating these enormous pressures. For the 11-GPa experiment, a ceramic octahedron had a 10.3-millimeter-long hole with a tiny rhenium furnace, a thermocouple to measure temperatures, and a graphite capsule containing the olivine and iron-nickel-sulfur-oxygen sample inserted in it. The octahedron rested in the center of eight 32-millimeter tungsten carbide cubes whose inside corners were truncated to accommodate the sample. (Tungsten carbide is used for the cubes, or anvils, because of its hardness, which is close to that of diamonds but at a much lower cost.) Tiny ceramic gaskets were placed at the edges of the carbide cubes to contain the pressure. This assembly of an inner octahedron and eight carbide anvil cubes was put in the press's split-cone, steel buckets as shown in the figure above. In several stages, the steel buckets pushed on the carbide cubic anvils, which pushed on the octahedral volume inside.
The multi-anvil press is not Livermore's only device for studying the behavior of the Earth's innards, but in many ways it is the best for this type of study. The diamond-anvil cell can produce 100-GPa pressures, comparable to the pressures at the center of the Earth (see Science & Technology Review, March 1996). But it can accommodate only a 20-micrometer sample, too small for much post-experiment evaluation. With the piston-cylinder press, the sample volume is about 500 millimeters3, but it has only a 4-GPa pressure capability, which is comparable to a depth of just 120 kilometers. The multi-anvil press is in the middle, providing pressures useful for studies of this type and accommodating a sample large enough for evaluation after the experiment.
In Minarik's experiments, the press took about 4 hours to bring the samples to full pressure, after which the samples were heated for periods ranging from 4 to 24 hours. During this time, the porosity of the sample collapsed, and the stable microstructure developed. Then the unit was cooled down and allowed to decompress for about 12 hours. During this process, the graphite capsule turned to diamond, which must be ground off before the sample could be sliced and polished for evaluation.
Despite being molten and much denser than the olivine, the metallic melt showed no signs of separating and draining to the bottom of the capsule. For the molten metal to drip down along the silicate grain edges, it has to be able to wet the edges. But in none of the experiments did wetting occur.1 Rather, the iron-nickel mixture beaded up at the corners of the silicate grains like water does on a waxed car, as shown in the figure below.
Livermore's findings agree with similar, lower-pressure studies that have melted meteorites and iron-nickel-sulfur- oxygen mixtures and failed to wet the silicate minerals. Together, these experiments indicate that much higher temperatures were required to separate the Earth's core and mantle--temperatures high enough to melt most of the Earth.
All of these data lend credence to the theory that the young, growing Earth was repeatedly bombarded by large planetoids, with some of these collisions generating temperatures high enough to form a magma ocean from which drops of dense molten metal separated. The largest collision may have been when a large celestial body, about the size of Mars, collided with Earth nearly 4.5 billion years ago, melting most of it and causing the core and mantle to separate. The leading theory today for the Moon's formation postulates that some material from that collision was ejected into orbit and condensed into the Earth's Moon.
Livermore scientists have long studied material properties and the effects of high temperatures and pressures. Their work has resulted in some mighty big bangs but none as large as the ones Minarik has postulated.
"We plan to look next at the geochemical aspects of this project, the partitioning of trace elements between molten metal and silicates at the same high temperatures and pressures," says Minarik. "There are many scientists in this country and elsewhere studying the formation of the Earth, and all of us are in the same boat. With all of the direct evidence of the Earth's formation buried far beneath our feet, these laboratory experiments are our only way to recreate what might have happened."

Key Words: Earth core formation, multi-anvil press.

1. W. G. Minarik, et al., "Textural Entrapment of Core-Forming Melts," Science 272, 530-533 (April 26, 1996).

For further information contact William Minarik (510) 423-4130 ( or Frederick Ryerson (510) 422-6170 (

Back to December 1996