Looking at a globe, one can easily imagine the continents and oceans as eternal, unchanging aspects of Earth's surface. Geophysicists now know that the appearance of permanence is an illusion caused by the brevity of the human life span. Over millions of years, blocks of Earth's rigid outer layer, the lithosphere, move about, diverging at mid-ocean ridges, sliding about along faults and colliding at the margins of some of the oceans. Those motions cause continental drift and determine the global distribution of earthquakes, volcanoes and mountain ranges.

Although the theory of plate tectonics is well established, the engine that drives the motion of the lithospheric plates continues to defy easy analysis because it is so utterly hidden from view. To confront that difficulty, I and other investigators have focused our research on the mid-ocean ridges. The ridges are major, striking locations where the ocean floor is ripping apart. Examination of the composition, topography and seismic structure of the regions along the mid-ocean ridges is yielding results that often run contrary to conventional expectations. More complicated and fascinating than anyone had anticipated, the chemical and thermal processes in the mantle below mid-ocean ridges dictate how new oceanic crust forms. Mantle activity may also cause islands to emerge in the middle of oceans and deep trenches to form at their edges. In fact, these processes may be so potent that they may even subtly affect the rotation of the planet.

The idea that Earth incorporates a dynamic interior may actually have its roots in the 17th century. In his 1644 treatise Principles of Philosophy, the great French philosopher Ren Descartes wrote that Earth had a central nucleus made of a hot primordial, sunlike fluid surrounded by a solid, opaque layer. Succeeding concentric layers of rock, metal, water and air made up the rest of the planet.

Geophysicists still subscribe to the notion of a layered Earth. In the current view, Earth possesses a solid inner core and a molten outer core. Both consist of iron-rich alloys and have temperatures reaching over 5,000 degrees Celsius and pressures well over a million times the pressure at the surface. Earths composition changes abruptly about 2,900 kilometers below the surface, where the core gives way to a mantle much less dense than the core and made of solid magnesium-iron silicate minerals. Another significant discontinuity, located 670 kilometers below the surface, marks the boundary between the upper and lower mantle (the lattice structure of the mantle minerals changes across that boundary because of the different pressure). An additional major transition known as the Mohorovicic discontinuity, or Moho, separates the dense mantle from the lighter crust above it. The Moho lies 30 to 50 kilometers below the surface of the continents and less than 10 kilometers below the seafloor in the ocean basins. The lithosphere, which includes the crust and the upper part of the mantle, behaves like a mosaic of rigid plates lying above a hotter, more pliable lower part of the mantle called the asthenosphere.

Making ridges from mantle
THIS ORDERED, layered structure might seem to imply that Earth's interior is static. On the contrary, the deep Earth is quite dynamic. Thermal energy left over from the time of Earth's formation, augmented by energy released through the radioactive decay of elements such as potassium 40, uranium and thorium, churns the material within Earth. The heat travels across Earth's inner boundaries and sets into motion huge convection currents that carry hot regions upward and cold ones downward. These processes ultimately cause many of the broad geologic phenomena on the surface, including mountain building, volcanism, earthquakes and the motions of continents.

Among the regions offering the best access to Earth's interior are mid-ocean ridges. These ridges dissect all the major oceans, winding around the globe like the seams of a tennis ball, for a total of more than 60,000 kilometers. The Mid-Atlantic Ridge is a part of that global ridge system. A huge north-south scar in the ocean floor, it forms as the eastern and western parts of the Atlantic move apart at a speed of one to two centimeters per year. In addition to the frequent earthquakes that take place there, the summit of the Mid-Atlantic Ridge spews out hot magma during frequent volcanic eruptions. The magma cools and solidifies, thus forming new oceanic crust. The ridge is higher than the rest of the Atlantic basin floor. At progressively farther distances from the ridge, the seafloor deepens with respect to sea level, presumably because the lithospheric plates that move away from the ridge contract as they gradually cool with age.

The magma that rises at the Mid-Atlantic Ridge obviously originates in the upper mantle. Its composition differs considerably from that of the mantle, however. Magma that cools at ocean ridges forms a common kind of rock known as basalt. But researchers have found that seismic waves travel through the upper mantle at a rate of more than eight kilometers per second, far faster than they would pass through basalt.

One material that could possibly allow such a high velocity of sound is a type of dense, dark-green rock called peridotite. Peridotite consists mostly of three silicon-based minerals: olivine, a dense silicate containing magnesium and iron; orthopyroxene, a similar but less dense mineral; and clinopyroxene, which incorporates some aluminum and is more than 20 percent calcium. Peridotites also have small quantities of spinel, an oxide of chromium, aluminum, magnesium and iron.

How can basaltic magma be produced from a mantle made of peridotite? More than 30 years ago experimental petrologists such as Alfred E. Ringwood and David H. Green of the Australian National University exposed samples of peridotite to elevated temperatures (1,200 to 1,300 degrees C) and high pressures (more than 10,000 atmospheres). These values duplicate the temperature and pressure that exist in the suboceanic upper mantle roughly 100 kilometers below the seafloor. This research showed that gradual decompression of peridotite at those high temperatures melts up to 25 percent of the rock. The melt had a basaltic composition similar to that of mid-ocean ridge basalts.

These experiments support the view that hot, peridotitic mantle material rises under mid-ocean ridges from depths exceeding 100 kilometers below the seafloor. As the material moves upward, the mantle peridotite decompresses and partially melts. The melted part takes on the composition of a basaltic magma, rising rapidly toward the surface and separating from the peridotite that did not melt. Part of the melt erupts on the seafloor along the crest of the mid-ocean ridge, where it cools and solidifies and adds to the ridge crest. The remainder cools and solidifies slowly below the surface, giving rise to new oceanic crust. The thickness of the oceanic crust depends on the amount of melt that is extracted from the mantle.

The ridge crest's depth below sea level marks an equilibrium level determined by the temperature and initial composition of the upper mantle upwelling below the ridge. If the temperature and composition of the mantle were constant all along the ridge, the summit of the ridge would be at the same depth below sea level all along its length.

In the real world such consistency is unlikely. Small variations in mantle temperature or composition along the ridge would cause the summit to settle at varying elevations. Regions of suboceanic mantle where temperatures are higher have lower densities. In addition, a hotter mantle would melt more and produce a thicker basaltic crust. As a result, the ridge summits there will be higher.

The summit of the Mid-Atlantic Ridge shows just such variations in depth below sea level. For instance, along the ridge between about 35 and 45 degrees north latitude lies an area of abnormally high topography. Earth-orbiting satellites have detected in the same region an upward swell in the level of the geoid (the equilibrium level of Earths surface, roughly equivalent to the average sea level).

Researchers generally attribute this swell to the influence of a so-called hot spot centered on the Azores island group. Hot spots are zones that have high topography and excess volcanism. They are generally interpreted as the surface expression of a "mantle plume"--that is, of a rising column of unusually hot mantle material. Most oceanic islands, including the Hawaiian Islands and Iceland, are thought to be the surface expressions of mantle plumes. The source of the heat is thought to lie in the boundary zones deep inside Earth, even as deep as the core-mantle boundary [see "The Core-Mantle Boundary," on page 36].

Minerals offer evidence
MY COLLEAGUES and i set out to test these ideas by exploring how the topography along the Mid-Atlantic Ridge relates to the temperature, structure and composition of the underlying mantle. One way to collect such information is to examine the velocities of seismic waves passing through the mantle under the ridge. Another approach involves searching for local variations in the chemistry of basalts that erupted along the axis of the ridge. Those variations can be used to infer the extent of melting and the physical nature of the mantle source from which they derived.

We followed a third approach by attempting to collect rock samples of mantle peridotite. Peridotite is left as a solid residue after the basaltic magma component melts out of the upper mantle rocks. Mantle rocks usually lie buried under several kilometers of ocean crust, but in some cases blocks of upper mantle peridotite are accessible. They are typically exposed where the axis of the mid-ocean ridge is faulted or where it is offset laterally by transform faults; these rocks can be sampled by drilling or dredging or retrieved directly through the use of a submersible.

To analyze the mantle minerals in the Atlantic peridotite samples, we used an electron microprobe. This instrument focuses a beam of electrons only a few microns in diameter onto a slice of rock. In response, the mineral emits x-rays of characteristic wavelengths. An analysis of the wavelengths and intensities of these x-rays allows a determination of the chemical composition of the mineral. Collaborating with Nobumichi Shimizu of the Woods Hole Oceanographic Institution and Luisa Ottolini of the Italian Research Council in Pavia, we also used a different instrument--an ion microprobe--to determine the concentration of trace elements such as titanium, zirconium and rare-earth elements. The ion probe focuses a beam of ions onto a sample, which dislodges other ions in the sample for measurement. The method enabled us to determine the concentrations of trace elements down to a few parts per million.

Such analyses reveal much about the conditions in the mantle where the sample rocks formed, because the temperatures and pressures there produce distinct compositions in the peridotites. Petrologists, including Green and A. Lynton Jaques of Geoscience Australia, have shown that partial melting modifies the relative abundances of the original minerals in the peridotite. Some minerals, such as clinopyroxene, melt more easily than do others and hence decrease in abundance during the melting. Moreover, the partial melting process changes the composition of the original minerals: certain elements in them, such as aluminum and iron, tend to follow the melt. Their concentration in the minerals decreases as melting proceeds. Other elements, such as magnesium and chromium, tend to stay behind, so that the solid residue becomes enriched with them. Thus, as a result of partial melting, olivine becomes more magnesium-rich and iron-poor; the ratio of chromium to aluminum in spinel increases; and so on.

The composition of these minerals, calibrated by laboratory experiments, allows us to estimate the degree of melting that mantle peridotites undergo during their ascent below the ridge. Our data showed that substantial regional variations exist in the composition of the mantle. For instance, the chromium-to-aluminum ratio of orthopyroxene and spinel is highest in peridotites sampled from a broad area between about 35 degrees and 45 degrees north latitude. The ratio suggests that the average degree of melting of the upper mantle lying below this region may reach as high as 15 percent. In most parts, about 10 to 12 percent of the mantle melts during the trip upward. This area of above-average melting corresponds to the Azores hot-spot region, lending credibility to the theory that hot spots result from unusually hot mantle plumes upwelling deep within Earth. Other findings support that idea, including work by Emily M. Klein, along with Charles H. Langmuir of Columbia University's Lamont-Doherty Earth Observatory, who independently examined the chemistry of basalts along the Mid-Atlantic Ridge.

A hot spot would seem to be the cause of so much melting. In fact, assuming that temperature alone causes the melting in the Azores hot-spot region, we calculated that the hot-spot mantle would need to be more than 100 degrees C hotter than the mantle from elsewhere below the ridge.

Is there a way of testing the validity of this temperature estimate and its underlying assumptions? A number of geothermometers have been proposed. They are based on the observation that certain mineral pairs that coexist in equilibrium in the mantle undergo temperature-dependent chemical reactions. For instance, the orthopyroxene and clinopyroxene in a mantle peridotite react with each other until they reach an equilibrium composition that depends on temperature. Laboratory experiments have calibrated that relation. Thus, determining the composition of the coexisting mineral pair can indicate the temperature at which the members of the pair reached equilibrium.

I applied two geothermometers, one devised by Donald H. Lindsley of Stony Brook University and the other by Peter R. A. Wells of the University of Oxford, to the Mid-Atlantic Ridge peridotites. The results were surprising. They did not show higher temperatures in the hot-spot region; if anything, the region gives temperatures that are slightly lower.

A mantle spiked with water
WHY DID WE NOT FIND higher mantle temperatures for a region that displays high melting? One possibility is that the upper mantle there has a composition that causes it to melt more easily. Water could be the main factor. Experiments by Peter J. Wyllie of the California Institute of Technology and Ikuo Kushiro of the University of Tokyo and the Carnegie Institution of Washington, among others, demonstrated that trace amounts of water and other volatile elements in peridotite drastically decrease its melting temperature. Therefore, if such a "wet" mantle upwelled under a stretch of mid-ocean ridge, it would start melting more deeply in Earth than normal, "dry" mantle would. By the time the peridotite reached the surface, it would have undergone a degree of melting significantly greater than that of dry mantle under similar temperatures [see box on opposite page].

Is there any evidence that the upper mantle below the Azores hot-spot area is wetter than the mantle elsewhere below the Mid-Atlantic Ridge? Indeed there is. Jean-Guy E. Schilling and his co-workers at the University of Rhode Island reported that basalts from the segment of the hot spot situated between 35 and 45 degrees north latitude contain three to four times more water than do normal mid-ocean ridge basalts, as well as higher concentrations of several chemical elements (mostly light rare-earth elements). The anomalously high concentration of those elements means that the parent mantle in the hot-spot area harbors an enriched supply of these elements.

It seems, therefore, that the mantle below the Azores hot spot differs from the normal sub-Mid-Atlantic Ridge mantle not so much by being hotter as by having incorporated at some stage water and other fluids that changed its chemical composition and melting behavior. This chemical transformation of mantle peridotite by fluids is called metasomatism. It would explain why wet mantle near the surface would have experienced more melting than normal mantle would. It may also explain why the equilibrium temperatures estimated from peridotites at the Azores hot spot do not appear higher than average. Melting reactions consume heat, so that partial melting of upwelling mantle may actually have cooled the surrounding mantle. The higher the degree of melting, the greater the heat loss.

So the Azores hot spot may not be linked to a thermal plume originating from the deep mantle or the core-mantle boundary. Instead it may be a melting anomaly of relatively superficial origin in the mantle. These hot spots may not be truly hot and perhaps are best classified as "wet spots" because of the key role that fluids may play in their formation.

Where does the water that produces mantle metasomatism come from? One possible source of this water is the sinking slabs of old oceanic lithosphere in subduction zones at the margin of the oceans. This process recycles water into the mantle. Water could also be released in the upper mantle during degassing processes of the deep mantle. In addition, water molecules can be stored in the actual structure of mantle minerals.

Consider the mineral perovskite, a silicate of magnesium and iron that constitutes the main component of the lower mantle and is therefore the most abundant mineral on Earth. Perovskite can contain water in concentrations up to 1 percent. A lower-pressure form of perovskite, called wadsleyite, prevails in the zone of the mantle at a depth of between 660 and 450 kilometers and can contain water up to concentrations of about 1.5 percent. Totaling up all these water molecules, we can speculate that the total amount of water in Earth's mantle could be equivalent to that of several oceans. Much of this water is probably primordial, captured in Earth's mantle at the time of its formation over four billion years ago. The presence of water molecules dispersed in mantle minerals has important consequences. For example, it significantly lowers mantle viscosity, facilitating convective motions that cause the movement of lithospheric plates and the drifting of continents.

Uneven mantle tilts earths axis
OUR STUDIES OF MANTLE PERIDOTITES from the Mid-Atlantic Ridge suggest that some areas with cooler mantle temperatures may represent the return strokes of the convection cycle in the mantle--that is, the downwelling regions. To understand this notion, we must look south of the Azores region, to the equatorial zone where the Mid-Atlantic Ridge lies deeper than the ridge at higher latitudes. The mineral composition of peridotites recovered from the equatorial Atlantic indicates that they underwent little or no melting, which implies that the mantle temperature was exceptionally low. Schilling and Nadia Sushevskaya of the Vernadsky Institute of Geochemistry of the Russian Academy of Sciences reached similar conclusions after studying basalts from the equatorial Atlantic. In addition, Yu-Shen Zhang and Toshiro Tanimoto of Caltech found that the velocity of the seismic waves is faster in the upper mantle below the equatorial Mid-Atlantic Ridge than at higher latitudes. These observations imply a denser, colder upper mantle below the equatorial region of the Atlantic. The temperature of the upper mantle there may be up to 100 degrees C lower than the mantle temperatures elsewhere below the ridge.

A plausible explanation for the relatively cool and dense equatorial upper mantle is that it results from downwelling mantle currents. Hot mantle plumes upwelling in the northern and southern Atlantic mantle may flow toward the equator, giving up their heat to their cooler surroundings and then sink.

The equatorial position of the "cold" Atlantic mantle belt may not be arbitrary. It is possible that Earth's rotation and convection in the mantle are intimately connected phenomena. In the late 1800s George Darwin (the second son of Charles) pointed out that the distribution of large masses on the surface (such as continents) affects the position of Earth's axis of rotation. Several scientists since then have investigated how density inhomogeneities within the mantle cause true polar wander (that is, the shifting of Earth's axis of rotation relative to the mantle). The wander results from the natural tendency of a spinning object to minimize the energy spent for its rotation.

The redistribution of mass inside Earth may be recorded in the mantle. The late H. William Menard and LeRoy M. Dorman of the Scripps Institution of Oceanography suggested that the depth of mid-ocean ridges generally depends on latitude: ridges become deeper toward the equator and shallower toward the poles. Moreover, gravity measurements revealed that an excess of mass sits below the equatorial areas, at least in the Atlantic. These data suggest that abnormally cold and dense masses exist in the equatorial upper mantle.

The sinking of cold, dense slabs into the mantle may influence true polar wander. Dense masses that find their way to the mantle, such as those that occur in subduction zones at the edge of some oceans, will affect the position of the rotation axis. The equator would tend to shift toward the dense masses. If high-density masses tend to concentrate near the equator, downwelling and cooler mantle spots are most likely to prevail in the equatorial upper mantle, explaining at least qualitatively the cold upper mantle belt and resulting lack of normal melting in the equatorial zone of the Atlantic.

Diving for deep-sea data
A COLDER-THAN-NORMAL equatorial mantle when the Atlantic first opened would imply a colder and thicker continental lithosphere along the equatorial belt. (The equator 100 million years ago crossed the future Atlantic coastlines of Africa and South America roughly along the same position as it does today.) The cold and thick equatorial lithosphere must have resisted the rift propagating from both the south and the north. The equatorial region may have behaved as a "locked zone" (in the sense used by French geologist Vincent E. Courtillot). As a result, the equatorial Atlantic opened sluggishly. This slow, difficult opening may have created the large equatorial fracture zones, visible today as east-west breaks that offset short segments of the mid-ocean ridge.

Now that we know that today's mantle upwelling below mid-ocean ridges is heterogeneous in terms of temperature and composition, the next question is: How do the properties of the mantle upwelling below a given segment of the ridge change over time? This information would enlighten us on an important issue, namely, how ocean basins evolve. But the research to obtain the necessary data would require sampling older oceanic lithosphere at various distances from the axis of the mid-ocean ridge. And unfortunately, the older lithosphere is normally buried deep below sediments.

We felt we might have a chance to reach old lithospheric material in the central Atlantic in the vicinity of the Vema Transform Fault. This fault offsets the crest of the Mid-Atlantic Ridge by 320 kilometers, cutting a deep valley through the oceanic crust. A long sliver of seafloor appears to have been uplifted on the southern side of the transform, and we hoped that this uplifted seafloor would expose a pristine section of lithosphere.

To test this hypothesis, in 1989 we organized an expedition in conjunction with Jean-Marie Auzende of the French oceanographic institution Ifremer. We planned to descend to the seafloor--more than five kilometers down--in the research submersible Nautile. Most of our colleagues viewed our task with skepticism: prevalent opinion held that the normal sequence of upper mantle and crust is completely disrupted near a transform fault.

Nevertheless, we pressed on. We began a series of dives that started at the base of the section and moved up the slope. Each dive lasted about 12 hours, about half of which was spent descending to the seafloor and returning to the surface. The cramped quarters of the Nautile, a sphere of titanium one meter and 80 centimeters in diameter, can accommodate two pilots and one scientist, who lies face down for the duration of the trip.

On our first dive we verified that the base of the section consists of mantle peridotite for a thickness of about one kilometer. On the second day we discovered a layer of gabbros--rocks that form below the seafloor when basaltic melts cool slowly--resting above the peridotite. According to widely accepted geophysical models, gabbros are the main component of the lower part of the oceanic crust. So in going upslope from mantle peridotites to crustal gabbros, we had crossed the Moho discontinuity.

The next day I took the Nautile on a dive that started from the level reached by the submersible the previous day. As I progressed along the slope, skimming the seafloor, a spectacular rock formation called a dike complex gradually revealed itself. Theory holds that dike complexes form where hot molten material, generated by partial melting of the mantle, squirts upward toward the seafloor through many narrow fissures in the crust. Never before had a dike complex been observed on the seafloor.

The dike complex, about one kilometer thick, was topped by a layer of pillow basalt, the form taken by basaltic magma when it cools and solidifies rapidly on eruption to the seafloor. During the next several days, we explored a different section and confirmed our previous findings. We were quite excited because no one had ever before observed a complete and relatively undisturbed section of oceanic upper mantle and crust. We immediately documented our discovery in a short paper that we mailed to Nature as soon as we docked a few weeks later.

Encouraged by the results of the Nautile dives, we conducted two other expeditions and established that the Vema lithospheric section is exposed on the seafloor for more than 300 kilometers. After mapping the magnetic anomalies produced by the seafloor, we could estimate the velocity with which the lithosphere moves away from the ridge axis. We thus established that the Vema section exposes lithosphere created gradually at the axis of the Mid-Atlantic Ridge during a time interval of more than 20 million years--a unique opportunity to study how the creation of lithosphere varies through time!

During the dives, we had used the Nautile's mechanical arm to grab a number of samples of mantle peridotite. We later sampled by dredging mantle peridotites at close intervals along the base of the section in lithosphere of increasing age. From the mineral composition of these rocks we estimated the variations in the degree of melting they had undergone over time during their ascent below the Mid-Atlantic Ridge. At the same time, we could estimate how crustal thickness varied through time, thanks to gravimetric data obtained from both ship and satellite measurements of the gravity field produced by rocks below the seafloor. Crustal thickness depends on the quantity of melt generated by mantle ascending below the ridge.

The results were quite unexpected. The degree of melting of the mantle and the crustal thickness both appear to have increased steadily from 20 million years ago to today. Small oscillations are superimposed on this general trend. The simplest interpretation of these results: the Mid-Atlantic Ridge is becoming steadily "hotter" over time.

Surprisingly, the increase of the temperature of the upwelling mantle is accompanied by a decrease in the spreading rate of the lithospheric plate generated at the ridge axis. This result contrasts with the concept of "passive" upwelling of the mantle in response to the diverging motion of the lithospheric plates--a concept that would require proportionality between spreading rate and degree of melting of the ascending mantle.

We were also able to estimate the velocity of the solid mantle that rises below the ridge, crucial information for refining our models on the formation of the oceanic crust. The speed of the rising mantle depends on its temperature and composition (both affect density and viscosity) and on the diameter of the rising column and is related to the velocity of the lithospheric spreading that diverges from the ridge axis.

How can we estimate the speed of the rising solid mantle? The rising mantle generates melt within a depth interval that can be estimated from experiments and theoretical considerations. The melt fraction rises rapidly, cooling and solidifying as basalt in the crust, while its parent mantle continues to ascend slowly.

When the "parent" mantle peridotite reaches the lithosphere and starts moving horizontally with the plate away from the ridge, the basalt it generated has moved farther away from the ridge. The horizontal distance between the parcel of basaltic crust and its parent mantle, translated as time, would allow us to estimate the velocity of the rising solid mantle. After correlating the temporal variations of the degree of mantle melting with the variations of crustal thickness along the Vema lithospheric section, we estimated the solid mantle rose at an average velocity of about 25 millimeters per year.

To refine this estimate, we need to go back and take additional samples of peridotite from the exposed lithospheric section so that we can achieve a higher resolution in the curve describing temporal variations of degree of melting of the mantle.

Why is the Mid-Atlantic Ridge north of the equator becoming gradually hotter? We can only speculate. Perhaps a wave of plume-derived hot mantle has been flowing southward toward the equator since a few tens of million years ago. We have hints that major oscillations in the intensity of mid-ocean ridge activity occurred in the distant past.

For example, studies by Roger Larson of the University of Rhode Island suggest that a mantle "superplume" roughly 100 million years ago caused swelling of mid-ocean ridges, faster seafloor spreading, rising sea levels, and warming of the climate as a result of larger quantities of carbon dioxide, methane and other greenhouse gases released from the mantle [see "The Mid-Cretaceous Superplume Episode," on page 22].

Much remains to be done before geologists develop a complete picture of mantle dynamics and its influence on surface geology. Debate persists as to the origins of mantle convection and whether it extends into the lower mantle. Indeed, symposia that include theoreticians, geophysicists, geochemists and petrologists invariably yield heated discussions and much dissent. On one point there is unanimity: Earths mantle is very much alive and is an exciting region to study.

ENRICO BONATTI holds degrees in geology from the University of Pisa and the Scuola Normale Superiore in Pisa, Italy. After coming to the U.S. in 1959, he spent several years as a research scientist in marine geology at the University of Californias Scripps Institution of Oceanography and as a professor at the University of Miami's Rosenstiel School of Marine Sciences. Since 1975 he has been with Columbia University's Lamont-Doherty Earth Observatory. Recently he has been teaching and researching in his native country. He has led or participated in expeditions in all the major oceans and in some remote but geologically intriguing lands, from the polar Ural region of Russia to the desert island of Zabargad in the Red Sea. Bonatti wishes to thank Daniele Brunelli, Anna Cipriani and Marco Ligi, who have collaborated with him in his research during the past decade.