How do you start subduction?


"There are two principal views on the physical mechanism leading to the initiation of subduction.  The first and most common is that as the oceanic lithosphere ages and cools, its density increases so that an instability arises and the plate sinks spontaneously in the mantle under its own weight."3  However, numerical modelling has shown "that it is highly unlikely that the entire lithosphere at a fracture zone will spontaneously founder."4  "A self-sustaining subduction zone does not form from a homogeneous plate."3  Researchers found that "no combination of fault rheology or geometry produced self-sustaining subduction without applied convergence."4  "The Wilson cycle predicts that the Atlantic will close again to form a second Pangaea through widespread initiation of subduction in the Atlantic basin.  Yet, with the exceptions of the narrow Caribbean and Scotia arcs, there is no evidence for subduction initiation (either intra-oceanic or at passive margins) within this ocean basin, despite 100- to 200-million-year-old passive margin sequences...  Thus, the passage of time does not appear to increase the probability of subduction initiation in the Atlantic basin."7

"According to the other view, externally applied compressive stresses and moderate convergence are necessary to form a new subduction zone."3  "The most likely mechanism would be through a transfer of stress induced by a collision, leading to 'forced' subduction initiation elsewhere.  Yet the response to recent collisions suggests otherwise.  The formation of the Alpine-Himalayan chain represents the collision of India and Africa with Eurasia at about 35 to 50 million years ago in the closure of the Tethys Ocean.  If large-scale collisional stress transfer occurred, we would expect subduction to have initiated elsewhere within the Indian and African plates.  However, no new subduction zones have initiated south of either India or Africa...  More than 50 million years have elapsed without the initiation of subduction."7

"In fact, the only subduction initiation that has occurred in the past 80 million years (with the notable exception of the 600-km-long Scotia Arc) has been intra-oceanic and entirely within the Pacific basin."  "Although it is commonly assumed that subduction has operated continously on Earth without interruption, subduction zones are routinely terminated by ocean closure and supercontinent assembly."  Researchers "hypothesize that dramatic reductions or temporary cessations of subduction have occurred in Earth's history."7

That would require subduction on Earth to begin repeatedly virtually from scratch!  The truth is, "subduction initiation [is] probably the least well-understood aspect of plate tectonic theory."7  "While events such as the opening and closure of ocean basins suggest that subduction initiation is common, theoretical models suggest is should be quite difficult."4

Plate tectonics is only observed on the Earth while all other known planets are covered with a stagnant lithosphere.

"There would be no plate tectonics if there were no subduction zones.  Yet how a subduction zone begins remains poorly understood."6  "Most theoretical studies have concluded that it is difficult to initiate a new subduction zone.  Although several studies have examined the initiation of subduction, the dynamics of this process remain obscure.  There remains substantial disagreement and uncertainty about the significance of different processes influencing subduction initiation, the material properties of tectonic plates, and even whether it is possible to initiate a totally new subduction zone in isolation from an existing one."3

"Subduction zones are really only crudely predicted by simple convection theory, and there is much to them that is highly atypical of convection.  First, if one were to only consider the strength of cold super-viscous lithosphere, one would not expect to see subduction zones at all.  Convection with purely temperature-dependent viscosity typical of the Earth's tends to form a cold, hard, and immobile layer on the top, and all convective motion occurs beneath it, as if it were a rigid lid."  "Thus, subduction initiation presents a formidable problem to convection models.  Second, in most forms of thermal convection, both surface boundary layers converging on a sheet-like downwelling will descend, while this occurs nowhere at any terrestrial subduction zone; i.e. all subduction is one-sided, with only one plate descending into the mantle.  This asymmetric downwelling is yet another major enigma not yet well explained in convection theory."1

Pluto is a good illustration.  An old impact crater basin named Sputnik Planum is filled with nitrogen ice.  It is a weak solid with a very low melting point, and should flow slowly even at Pluto's low temperatures.  Internal radioactivity is assumed to heat the ice from below, causing convection.  This image shows the interlocked convection cell "plates" that are from 10 to 40 km across.  From the surface to the bottom of the basin is from .5 to 1 km.  Convection in the ice proceeds at a pace of centimeters per year, comparable to the motion of Earth's crust.2

In this close-up image by the shore of the basin you can see that the edges of each cell descend (two-sided subduction); one does not override the other (one-sided subduction) as on Earth.  If Plate Tectonics theory was right, this is what subduction zones in the lithosphere on Earth would look like.  They don't.  Another mechanism is required.

"Two forces must initially be overcome to make a subduction zone: fault friction (or growth of a lithosphere-cutting shear zone) and plate bending.  Plate bending becomes the principal source of resistance."3  "Materials on both sides of a ridge or transform [fault] should be broadly similar as they are produced by similar processes in similar environments.  Hence, it is unlikely that compositional buoyancy contrast across these weak zones would develop throughout their evolutionary histories."  "Therefore, it is physically implausible why one side of an old ridge, or transform [fault] or fracture zone prefers to sink while the other side chooses to rise under any deviatoric stresses."6

For plate tectonics theory, "the factor which most strongly dictates where subduction initiation will occur is the initial tectonic state of the system.  [That] include[s] former spreading centers, fracture zones, transform faults, passive continental margins, and subduction zones undergoing polarity reversal."3  So how did the very first subduction zones form, before there were plates and spreading ridges to provide compression?  An effort has been underway for decades to model the physics of not just this question, but the whole of plate tectonics theory and how it began.

The familiar image of spreading ridges and slabs diving into the Earth at trenches is simple and elegant, and quickly led to the general acceptance of plate tectonics theory.  Such images are models of 'boundary conditions' (edges of plates).

"Models utilizing dynamically determined boundary conditions to achieve plate-like surface motion have relatively little difficulty with emulating terrestrial convective vigor or simulations of billions of years.  Instead, their weakness is more fundamental; they can only provide insight into the reciprocating dynamics of the mantle and plates once the existence of the plates is assumed and they cannot model any aspects of the dynamics responsible for the origin of the plates."5

These are excerpts from a paper5 published in 2011 assessing the last 40 years of progress in modeling how plate tectonics could have started on Earth, and the work that remains.

"A consensus is beginning to form regarding the necessary requirements for obtaining the primary elements of plate-like surface motion.  However, despite significant progress, the generation of plates over long periods has not yet been modeled with Earth-like convective vigor."

The plates are rheologically and thermally distinguishable from the underlying mantle by their rigidity and temperature.  Their origin is the cool upper thermal boundary layer of the convecting mantle and their stiffness is explained by the mantle's temperature-dependent rheology.  However, it has been known for nearly 40 years that unless some other control on viscosity is present, a temperature-dependent rheology alone cannot produce plate-like behavior at the surface of a convecting fluid."  "Convecting fluids with a temperature-dependent viscosity that varies over the wide range of values appropriate for modeling the rheology of the Earth's mantle exhibit a mode of convection widely referred to as 'stagnant-' or 'rigid-lid convection' where a cool immobile conducting boundary layer forms above a decoupled vigorously convecting underlying fluid."

While some models have been able to overcome the rigid-lid problem, "one cause for concern that applies to many of the studies described in this section is that (except where noted) the findings were obtained in Cartesian geometries.  Cartesian systems, most notably those featuring internal heating, are unrealistically hot in comparison to spherical shell convection models featuring the Earth's core to surface radii ratio.  Given that calculations featuring temperature-dependent rheologies are especially sensitive to the heating mode, future calculations need to focus on the issue of realistic mantle geotherms as well as the generation of plates."

"The challenge ahead for self-consistent rheological plate-modeling methods is to elevate the parameters of the studies to the ranges appropriate for the Earth.  The Earth's Benard-Rayleigh number is approximately 10,000 times the critical value.  The effective Rayleigh numbers of current studies are still more than an order of magnitude less than the desired values. Given the transitions in plate motion behavior discovered in some of the high Rayleigh number studies utilizing force-balance plate formulations, there is the possibility that the rheology based modeling methods that have led to the successes of the past half dozen years may still require significant refinement to obtain plate tectonics with Earth-like convective vigor."

"In addition to modeling convection with Earth-like vigor, plate-like surface motion will be required for periods in excess of a billion years of model time.  Moreover, plate boundaries with terrestrial characteristics have to replace the distinctly un-Earth-like boundaries that persist in present models.  Principal among the outstanding problems is the unresolved issue of one-sided subduction.  Can it be obtained and, of equal importance, can it be sustained in a self-consistent model?"  "The inclusion of realistic tectonic forces and associated continental as well as oceanic plate boundary forces will be required in future studies."

"Of further importance is the ultimate goal of reproducing not just the primary characteristics of plate tectonics: plate-like motion; strike-slip plate boundaries (in addition to convergent and divergent boundaries); a strong toroidal flow component; plate growth and consumption and one-sided subduction, but also some of the secondary features.  For example, Earth-like spreading boundaries consist of ridge sections offset by transform faults, rather than a mixed divergent/strike-slip boundary.  Moreover, stable triple junctions follow well understood trajectories determined by the characteristics of the boundaries from which they are comprised (e.g., convergent boundary).  A realistic mantle convection model featuring self-consistently generated plates will need to reproduce Earth-like triple junctions and emulate their motion on a sphere. Similarly, the models will need to produce spatial heat flow and dynamic topography in agreement with terrestrial data."

"Finally, a fully self-consistent mantle convection model will need to exhibit all of the Earth's surface dynamics on all time scales.  Primary among these requirements will be the manifestation of continental drift and supercontinent assembly and breakup, thus requiring the modeling of distinct oceanic and continental lithosphere.  Indeed, obtaining self-consistently generated oceanic and continental plates from a mantle model is an issue that must be addressed in order to answer some of the most fundamental and rewarding questions: How did the continents form?  How different from mantle convection with just oceanic plates is mantle convection with distinct continental as well as oceanic plates?  For example, do continents really insulate the mantle?  Is cyclic supercontinent assembly inevitable and, if so, why does it happen?"5

What this review article shows is that when plate tectonics theory moves from the cartoon image in the mind to the real world of physics, how it could possibly have gotten started on Earth is neither simple nor elegant.  One may wonder if it is even possible.  The lack of progress in modeling is stunning.



1.  Bercovici, David. 2003. The generation of plate tectonics from mantle convection. Earth and Planetary Science Letters, Vol. 205, pp. 107-121.

2.  Dombard, Andrew J., Sean O'Hara. 2 June 2016. Pluto's polygons explained. Nature, Vol. 534, pp. 40-41.

3.   Gurnis, Michael, Chad Hall, Luc Lavier. 10 July 2004. Evolving force balance during incipient subduction. Geochemistry Geophysics Geosystems, Volume 5, Number 7, pp. 1-31.

4.  Hall, Chad E., Michael Gurnis, Maria Sdrolias, Luc L. Lavier, R. Dietmar Muller. 2003. Catastrophic initiation of subduction following forced convergence across fracture zones. Earth and Planetary Science Letters, Vol. 212, pp. 15-30.

5.  Lowman, Julian P. 2011. Mantle convection models featuring plate tectonic behavior: An overview of methods and progress. Tectonophysics, Vol. 510, pp. 1-16.

6.  Niu, Yaoling, Michael J. O'Hara, Julian A. Pearce. May 2003. Initiation of Subduction Zones as a Consequence of Lateral Compositional Buoyancy Contrast within the Lithosphere: a Petrological Perspective. Journal of Petrology, Volume 44, Number 5, pp. 851-866.

7.  Silver, Paul G., Mark D. Behn. 4 January 2008. Intermittent Plate Tectonics? Science, Vol. 319, pp. 85-88.


 John Michael Fischer