Roadmap | The review process | Home
   Ridge-crossing hotspots

"Ridge-crossing" hotspots

E. K. Beutel1 & D.L. Anderson2

1Dept. of Geology, College of Charleston, 66 George St. Charleston, SC 29424

2Seismological Laboratory, California Institute of Technology, Pasadena, Ca. 91125

The authors of this page request feedback to help them mature the ideas described herein.


Ridge-crossing volcanic island chains have not been satisfactorily explained before using non-plume hypotheses and as such, ridge-crossing chains are cited as evidence for relatively fixed deep-mantle plumes. In this paper we describe two melt sources whose interaction with ridges may produce age-progressive, ridge-crossing volcanic seamount chains without invoking a deep mantle source. We use both conceptual and finite-element models to illustrate the viability of these mechanisms.


Hypotheses for ridge-crossing seamount chains have focused on the interaction of mantle plumes with ridge segments (DePaolo & Manga, 2003; Sleep 2002). A stationary plume of hot mantle material rises to the base of the lithosphere and creates a chain of age-progressive seamounts. As a migrating ridge approaches the plume, material rises along the lithospheric gradient to the ridge and chains of seamounts are created on both sides of the ridge. Eventually the ridge moves away from the plume and seamount production on a single plate resumes (e.g., DePaolo & Manga, 2003; Sleep 2002; Ribe 1996; Kincaid et al., 1996). This results in the formation of a lopsided V-shaped array of seamounts (Figure 1).

Figure 1: Cartoon of a ridge-crossing seamount chain geometry in the Atlantic (after Sleep, 2002).

Because current non-thermal models of seamount formation are unable to explain this geometry, (e.g. propagating cracks are presumed to be unable to cross ridges), ridge-crossing seamount chains have been cited as evidence for relatively stationary plumes. However, we propose that when ridges are viewed as active rather than passive features, non-thermal ridge-crossing seamount chains are not only viable, but also likely. In this paper we propose an alternative model for ridge-crossing seamount chains that does not invoke plumes but is still able to explain the four primary elements:

  • a source of melt,
  • a means of bringing melt to the surface,
  • a ridge-crossing seamount chain, and
  • an explanation for the age progressive geometry of the chain.


The source of melt

We focus on two alternative sources for the melt:

  • a fertile region in the mantle, and
  • mantle in a nascent melting condition.

In the former case, the mantle is heterogeneous and contains fertile streaks, some of which are large and produce major outpourings of magma when tapped, i.e. when the stress condition in the overlying plate permits dikes and volcanoes to form. In the latter case, the mantle is fairly homogeneous but is at or near its melting point and melt is created at any time pressure drops or when the lithosphere extends and passive upwelling occurs as a result. Elements of these scenarios have been discussed previously by various authors, e.g., Meibom & Anderson (2003), Sleep (2002), Natland & Dick (2001), Niu et al. (2001), McNutt & Bonneville (1999), Anderson (1998) and Sleep (1997).

Transport of melt to the surface

Melt is generated from fertile mantle regions and mantle in a nascent melting condition when it is decompressed. Decompression melting may occur at and ahead of ridge-transform intersections (RTIs) when slip along the transform is impeded (Beutel, 2005) This model for creating lithospheric extension is suggested by the observation that many seamount chains intersect or originate near RTIs, e.g., the Foundations and the St. Helena and Easter chains.

Finite-element models show that large areas of extension develop when transform slip is impeded (Figure 2a,b) (Beutel, 2005, in which a discussion of the possible causes of impeded transform slip may be found; see also Ridge-transform intersections page). If the mantle is already partially molten, or if magma is ponded beneath the plate, then the condition for dikng is simply that the least compressive axis is horizontal. If the mantle is entirely subsolidus, then decompression is required to generate excess melt. This is expected to occur in response to exension, such as occurs in the neighborhood of RTIs. Seamount chains rather than linear ridges are postulated to form as a result of periodic strengthening and weakening of the transform fault due to changes in the local stress state, which occur partially as a result of the injection of magma into the crust. We also postulate that the age difference between a seamount and the underlying oceanic crust can vary greatly depending on the offset of the transform and the distance ahead of the RTI that the seamount is emplaced.

Figure 2: Results of finite-element modeling of the nascent-melting case. Panels show a series of 2D models of a reorganizing ridge. Background colors indicate maximum stress intensity, warmer colors indicate larger stresses, white and black bars indicate maximum and minimum stress vectors (white: compressional, black: extensional). Panels A, C and E show the results for weak transforms (slip not impeded), and panels B, D and F show strong transforms (slip impeded). Note the increased extensional stress at and ahead of the ridge-transform intersection when slip is impeded. Adjacent to each panel is a cartoon showing the seamounts that would be produced. Seamounts are color-coded according to age, as shown in the key at the bottom. These seamounts form time-progressive chains. Panel F shows the switch from a dual seamount chain on either side of the ridge to a single chain on the western side of the ridge. Note the gap between the chain on the east and the chain on the west. The change in trend of the seamount chains on the western plate reflects a relative, not absolute plate motion change. Click here for enlargement.

Ridge-crossing seamount chains

Based on our understanding of non-thermal melt sources and the transport of melt to the surface we have developed two models by which non-plume ridge-crossing seamount chains may be generated. These models are based on the following assumptions:

  1. most transforms experience periods of decreased slip, due to welding or tectonic forces,
  2. when transform slip is impeded, extension is concentrated at and ahead of the RTI (Figure 2a,b),
  3. if the mantle is fertile or near its solidus, extension at an RTI will result in the formation of a seamount, and
  4. ridges migrate and reorganize.

Model 1 – A reorganising ridge: Model 1 is illustrated in Figure 2. We examine a possible ridge-crossing geometry for seamount chains formed during a ridge reorganization over a near-solidus mantle. We assume that any large, concentrated extension in the oceanic lithosphere will result in the formation of a seamount. Finite-element models of an evolving ridge were constructed to determine the location of these extensional stresses and the resultant seamount pattern that would emerge, during a ridge reorganization. More than 12 initial time frames were constructed, and the 6 most relevant are shown. Figure 2 presents the finite-element model results as maximum stress type by color (red is extensional and blue is compressional) and maximum and minimum stress vectors, along with conceptual models of seamount emplacement adjacent to the stress maps. The exact model parameters are given in the appendix. Stresses applied to the model consist chiefly of ridge-perpendicular gravity forces applied to the whole of the plate.

  1. Time 1: Three north-south-trending ridge segments are connected by weak transforms. The oceanic plate is under relatively little east-west stress.
  2. Time 2: A change in plate motion direction results in NE-SW ridge-push forces and the strengthening of both transform faults. Large areas of extensional stress are concentrated ahead of and at the RTIs. Seamounts are emplaced in older crust ahead of the RTIs.
  3. Time 3: In response to changing plate motions a new NW-SE-oriented ridge propagates between the southern two ridge segments, eradicating the transform. The northern transform is modeled as weak. Extension is concentrated where the new NW-SE trending ridge intersects the middle NS-trending ridge segment. Seamounts are emplaced on the western plate only.
  4. Time 4: Re-strengthening of the northern transform results in extensional stresses at and ahead of the RTIs there. New seamounts are emplaced.
  5. Time 5: A new NW-SE-trending ridge segment propagates over the northern transform. Large extensional stresses are concentrated where the NS ridge segments intersect the NW-SE trending segments, more seamounts are emplaced on the western plate.
  6. Time 6: A new, weak transform forms between the NW-SE trending ridge segments. Extension is still concentrated at the intersection of the NS- and NW-SE-trending ridge segments. Seamounts are emplaced on the western plate at the intersection.

Seamount Pattern: The resulting seamount pattern is a pair of chains that age progressively away from the ridge and appear to switch from emplacement on both sides of the ridge to emplacement only the west side. This gives the appearance of a ridge that overran a hotspot.

While no one specific seamount chain is modeled in Figure 2, the ridge reorganization modeled is similar to ridge reorganizations seen throughout the Pacific plate. The high density of seamounts in the Pacific compared with the Atlantic suggests that the mantle beneath the Pacific is closer to its solidus than Atlantic mantle.

Model 2 – A stationary fertile region: In this conceptual model we apply the finite-element results of extension at or ahead of RTIs to a ridge migrating over a stationary fertile mantle region. The assumption is that seamounts will only be formed when there is strong extensional stress above the fertile mantle region, the presumed source of the extensional stress being impeded slip on a transform fault.

Figure 3 is a schematic diagram illustrating the modeling results separated into discrete time periods. During the first four time periods (T1 - T4) a fertile mantle region is intersected by a series of RTIs as the ridge moves to the northwest. This results in a single chain of seamounts on the western plate, the spacing of which is determined by the production rate at, and spacing of, the RTIs. The seamounts move to the northwest relative to one-another but to the west relative to the ridge. In time period T5 the ridge overrides the fertile mantle region and chains of seamounts on both sides of the ridge are created. Finally, the ridge moves off the fertile mantle region and the chain on the western plate is terminated. In time period T8 the fertile mantle region is intersected by an RTI on the eastern side of the ridge and a new series of seamounts is created, giving the impression that the seamount chain has crossed the ridge.

Figure 3: Time series for a conceptual model of a ridge and a fertile mantle region migrating with respect to a fixed point (crescent) on the eastern (African) plate. In time periods T1-T4 impeded slip along the transform increases extension over the fertile mantle region, resulting in extension and volcanism, and seamounts are emplaced on the western plate. During time periods T5-T7 the fertile mantle region lies beneath the ridge and seamounts are emplaced on both plates. In time period T8 the ridge moves away from the fertile mantle region, resulting in the formation of seamounts on the eastern plate. Because the region of fertility migrates south relative to the fixed point on the eastern plate, a pair of time-progressive NW- and NE-trending seamount chains forms.

Other Locations

The model above can account for the geometries of some Atlantic seamount chains. However, the principles of RTI-generated seamount chains may be applied to many other observed seamount geometries. Short seamount chains may result from small regions of mantle fertility and/or the migration of a ridge away from the fertile region. Large areas of melt and aseismic ridges may represent large regions of fertility or mantle in a nascent melting state. The more numerous seamounts and faster spreading rates in the South Pacific indicate that different processes may affect seamount formation there including strong thermal contraction of the lithosphere, a plate close to the tensile state, and/or numerous fertile patches or proximity to the nascent melting state.

Large igneous provinces (LIPs)

Plume-based models for ridge-crossing-hotspots involve the separation of postulated plume heads (LIPs) from the ends of their tails (volcanic chains) by actively spreading ridges. Most volcanic chains, however, do not start at a LIP and most LIPs are not associated with a volcanic chain. Perhaps the best documented case is the separation of the Kerguelen Plateau from the Ninety-East Ridge and the Rajmahal basalts (Weis et al., 2002; Kent et al., 2002; Coffin et al., 2002). In order to link the basaltic outpourings on mainland India with the Kerguelen Plateau and Broken Ridge a series of ridge jumps are postulated to have occurred as the Kerguelan hotspot drifted slowly to the south (Kent et al., 2002; Antretter et al., 2002). Other models involve multiple plumes (e.g., Coffin et al., 2002). In this paper we highlight, as others have before us, the links between LIPs, ridges and transform faults, whereby many of the proposed splits between "plume heads" and "plume tails" may be ridge-reorganizations and transform fault interactions rather than ridges crossing fixed thermal anomalies. The prevalence of "coincidental" relationships between hotspot features and tectonic features is thought-provoking. These include:

  • the Ninetyeast ridge lies is along an extensive offset of seafloor magnetic anomalies – a fossil transform fault,
  • Réunion is located on the intersection of an abandoned ridge and a fracture zone,
  • Réunion and Mauritius developed on Paleocene fossil spreading centers and were transported away from each other by a fracture zone that lies in between (Hirn, 1993), and
  • the Ontong Java plateau is thought to have been created at a triple junction (e.g., Korenaga, 2005).

Seamount chains that apparently cross ridges do not require a plume-based model, but can be explained by ridge dynamics and mantle heterogeneities. Our models demonstrate that volcanism can migrate from one side of a ridge to the other when extensional stress at and ahead of RTIs is considered. We further suggest that the nature of the melt sources may also affect seamount-chain geometry. Regions of fertility in the mantle may have the ability to produce larger volumes of magma than thermal anomalies of the magnitudes suggested by observations (see also Thermal pages). The recognition that hotspots move relative to one another and relative to the geomagnetic reference frame, led to the concept of drifting plumes, the predicted consequences of which are not different from those of a passive fertile heterogeneity. A stress-controlled mechanism of magma release is, however, more able to cause the rapid high-rate volcanism that builds LIPs, and the rapid switching on and off of magmatism along volcanic chains, than a thermal plume explanation. Furthermore, the size of a fertile heterogeneity is not important since it is stress, and the extension of the lithosphere, that controls and localizes the volcanism.

The melting of fertile patches of mantle requires no additional heat input to explain even the largest volumes of melt produced (Korenaga, 2005). The largest LIP on Earth is the Ontong-Java Plateau (see also Ontong Java pages). If the 20-km-thick crust there resulted from draining an area three times larger than the plateau itself (as a result of focusing at the apex of a triple junction) and if 20% melting was involved, then only a 30-km-thick section of the mantle may have been involved. The normal mantle geotherm is usually close to or above the solidus in the depth range ~30 to ~50 km. Clift (2004) has shown that the subsidence patterns of many oceanic plateaus attributed to plumes are consistent with normal mantle temperatures and not the elevated temperatures expected for plumes (see also Mantle temperature under LIPs page). This is consistent with the model we propose here, which attributes ridge-crossing seamount chains and LIPs to the extraction of melt from regions of fertility and does not invoke greatly elevated temperature.


Many so-called hotspot tracks lie along pre-existing fracture zones and transform faults, or emerge from RTIs. This, along with the inability of propagating cracks to cross active spreading ridges, led us to explore mechanisms for the migration of stress conditions as an explanation for the volcanism. A fertile region in the mantle can also have similar effects to a hotspot (plume). Finite-element models demonstrate the viability of off-ridge extensional stress migration associated with ridge-transform intersections (RTIs). Combined with fertile mantle and/or mantle in the nascent melting condition, such areas of extensional stress may create ridge-crossing seamount chains without the need to invoke a thermal anomaly for which there is no independent evidence.

Many of the geochemical arguments for “plume" compositional components could apply equally well to passive compositional hetereogeneities. Our models do not rule out a thermal plume origin for ridge-crossing seamount chains, but they demonstrate that such a model is non-unique. A passive mantle heterogeneity is not expected to be relatively fixed. The bulk of the upper mantle is probably moving, albeit much more slowly than plates and ridges. However, a source sufficiently deep, i.e. beneath the plate, is all that is required, as in the original model of Wilson (1963).


Appendix: Finite Element Model Parameters and Geometry

Model Type: Two-dimensional plane strain elastic finite element model, program by Gobat & Atkinson (1996).

Model Parameters:

Applied Forces: Applied forces are ridge-perpendicular and applied to the whole of the plate based on the modeled age of the crust. As new ridges are propagated to the southwest from the N-S trending ridges, new forces perpendicular to the now NW-SE trending ridges are applied. The basic applied forces are as follows:

Age (Myr)
Force (N/m)

Strength: The model consists of material of two end-member strengths, weak and strong. The strong material is 3 orders of magnitude stronger than the weak material. This is based on the strong decoupling expected between ridge material (weak) and oceanic crustal material (strong). Others (Richardson et al., 1979) have demonstrated that the exact ratio is not important as long as it is > 1. Because this model was testing end-member conditions the transform was alternately modeled as both weak and strong.

Boundary Conditions: The model is fixed in space along its eastern edge. Numerous configurations were tested to determine the configuration with the least edge effects. By fixing the model on its eastern edge we allow it to move as a whole while still providing a baseline to move against. This is similar to modeling the apparent movement of the entire South Atlantic basin away from a pinned African plate.

last updated 21st November, 2005