In the mid-20th century, new technologies like sonar and magnetometers provided the ability to map the seafloor and its magnetic anomalies, the basis of observations which led to the theory of plate tectonics. According to this theory, Earth’s uppermost layer (lithosphere) is divided into a mosaic of plates which move with respect to each other and the underlying mantle. Oceanic subduction zones play a key role in plate tectonics because about 90% of plate tectonic driving forces arise from slab pull (e.g., Lithgow-Bertelloni and Richards, 1998). Deformation, igneous activity, and natural disasters occur mainly at plate boundaries; however, igneous and tectonic activity also occurs within the plates. Examples are hotspot volcanoes above rising mantle plumes like Hawai’i and Yellowstone. The idea of a mantle plume was first proposed to describe the migration of the Pacific Plate over a hotspot in the mantle to explain the Hawai’i-Emperor volcano-island-seamount (Wilson, 1963). Following Wilson’s proposal, many studies have been carried out to refine this idea. An important recent hypothesis is that the impingement of a new mantle plume head on the base of the lithosphere can sometimes cause a new subduction zone to form (plume-induced subduction initiation, PISI, Whattam and Stern, 2015 and references therein). Numerical and observational evidence have shown that sinking of lithosphere into mantle induced by plume head is irrelevant to the nature of the overlying plate, it can be oceanic or continental lithosphere (e.g., Ueda et al., 2008; Burov and Cloetingh, 2009; Burov and Cloetingh, 2010; Gerya et al., 2015). Plume-induced sinking of intracontinental lithospheric mantle has been reviewed by Cloetingh et al. (2021). Here, we focus on plume-induced oceanic subduction initiation and briefly revisit plume-induced sinking of intracontinental lithospheric mantle. Formation of a new subduction zone as a result of plume-lithosphere interplay can also address the earliest subduction and the beginning of plate tectonics on Earth (Gerya et al., 2015).

Different mechanisms of subduction initiation have been proposed including onset of a new subduction zone along lithospheric weaknesses like transform faults zones (e.g., McKenzie, 1977; Mueller and Phillips, 1991; Hall et al., 2003; Lebrun et al., 2003; Gurnis et al., 2004; Stern 2004; Stern and Gerya, 2018), conversion of passive margins into active plate boundaries (e.g., Wilson 1966; Regenauer-Lieb et al., 2001; Nikolaeva et al., 2010; Duarte et al., 2013; Zhou et al., 2020), subduction initiation induced by pre-existing plate forces (e.g., Kemp and Stevenson, 1996; Baes et al., 2011; Levy and Jaupart, 2012; Zhong and Li, 2019) and subduction initiation induced by mantle suction flow (e.g., Lu et al., 2015; Baes and Sobolev, 2017; Baes et al., 2018). These mechanisms all depend on pre-existing plate tectonic forces and/or pre-existing lithospheric weak zones which are products of plate tectonics. PISI does not require such forces or weaknesses. For this reason, Whattam and Stern (2015) and Gerya et al. (2015) suggested that plume-lithosphere interaction may have initiated Earth’s first subduction zone, triggering the modern regime of plate tectonics.

Modeling studies (Ueda et al., 2008; Gerya et al., 2015; Baes et al., 2016, 2020a and, 2020b; Davaille et al., 2017; Lu et al., 2015; Crameri and Tackley, 2016) and geological and geochemical observations (Whattam and Stern, 2015; Davaille et al., 2017; Gao et al., 2019; Stern and Dumitru, 2019; Gülcher et al., 2020; Li et al., 2020; Yang et al., 2020; Rodriguez et al., 2021; van Hinsbergen et al., 2021) have advanced our understanding of plume-lithosphere interplay and its role in forming new subduction zones. Here, we revisit models and natural examples of oceanic PISI. We first explain mantle plumes and their role in plate tectonics. Then, we briefly explore previous studies of plume-induced sinking of intracontinental lithospheric mantle. This is followed by a review of numerical modeling studies of oceanic PISI in greater detail. Then, we revisit proposed sites of PISI on Earth and finally, we end with discussion and conclusions. A short video overview of PISI can be watched at https://www.youtube.com/watch?v=Hb47L8S7fMU&t=5s.

Mantle plumes are zones of upwelling hot mantle whose surface manifestations include linear volcanic chains, plateaus, large igneous provinces and continental flood basalts. Thinking about mantle plumes goes back to the early 1960’s. Wilson (1963) proposed the hotspot theory based on observations of the Hawaiian Islands which formed “within-plate,” far from any plate boundary. He argued that motion of the Pacific Plate over a “hotspot,” which represented a long-lasting exceptionally hot region in the mantle, generated the chain of Hawaiian Islands. The main evidence for this proposal was the progressive aging of the islands towards the NW. Wilson proposed that if lithosphere moves over a relatively fixed magma source—possibly the source is within the relatively stagnant center of a mantle convection cell—it would produce a chain of volcanoes like the Hawaiian Islands (Figure 1A). He noted that the rising mantle plume is not necessarily stationary but may move slower than the overlying plate. Morgan (1971) refined Wilson’s idea, stating that hotspots are stationary hot mantle plumes which originate from the core-mantle boundary. He proposed that these hotspots are expressions of lower mantle convection. Wilson (1973) indicated that surface expressions of mantle plumes in the oceans are: 1) uplift, 2) active volcanoes (producing alkaline basalts, rhyolite and tholeiite basalts), 3) high gravity anomalies, 4) extension, and 5) high heat flow. He documented more than thirty plumes, most of which lie close to mid-oceanic ridges and a few which are within the plate, producing isolated chains of seamounts and islands. The Wilson-Morgan hypothesis has advanced in the last 5 decades. Duncan and Richards (1991) noted that fixed hotspots on Earth can serve as a convenient reference frame for plate motion, through the orientations and age distributions of volcanic trails. They argued that plate motions in two reference frames of plume-supported hotspot and paleomagnetic reference frames reveal small but systematic differences. As hotspots are the upward flow of a long-lived, stable pattern of whole mantle convection (Figure 1B) these differences indicate whole-mantle motion relative to the Earth’s spin axis.

There are different views about the origin of mantle plumes. One is that these are deep-rooted, uprising hot mantle rocks which come from the core-mantle boundary (primary plumes of Courtillot et al. (2003) and shown as Type 1 plumes in Figure 1C). Burke and Torsvik (2004) indicated that these plumes mainly form at the margins of LLSVPs (Large Low Shear wave Velocity Provinces which are also known as Sub-African and Sub-Pacific regions (Figure 1C) or TUZO and JASON (Burke, 2011)), which are approximately stable through time. It has been proposed that these two LLSVPs in the deep mantle beneath Africa and the South-Central Pacific which cover ∼20% of the core–mantle boundary and extend up to ∼1,000 km above it (e.g., Dziewonski, 1984; Davies et al., 2015), have higher density than the surrounding mantle, indicating that they are chemically different (e.g., Becker and Boschi, 2002; Cottaar and Lekic, 2016). The second type of plumes are those that come from the transition zone at the top of the superplumes of the two LLSVPs (called secondary plumes by Courtillot et al. (2003) and shown as Type 2 plumes in Figure 1C). Anderson (2000) challenged the role of the deep mantle in forming plumes (Type 3 plumes in Figure 1C). Based on geophysical data such as bathymetry, subsidence rates, heat flow, global plate motion modeling, depths to mantle phase changes and seismic velocities, he suggested that lateral variations of temperature in the asthenosphere is ∼200°C, which is in the range of estimated plume-excess temperatures (e.g., Herzberg and Gazel, 2009). He argued that upwelling of asthenosphere through the melting zones are responsible for excess magmatism, not deep-rooted hot mantle plumes. Anderson (2005) proposed that plumes are asthenospheric features which form due to stress- and crack-controlled magmatism, mechanisms that are associated with plate tectonics.

Plumes rise towards the surface due to their buoyancy, which depends on plume temperature, composition and size. The plume excess temperature with respect to ambient mantle has been estimated to be 200–350 K (White and McKenzie, 1995; Herzberg and Gazel, 2009). Montelli et al. (2004) presented tomographic evidence for the existence of deep-mantle plumes having diameters of several hundred kilometers. They mentioned that because of the limitations in estimating plume diameters in the lower mantle where tomographic data has low resolution, the actual plume size may be much smaller. Some studies suggest that plumes might include a denser component with different composition than ambient mantle (e.g., Deschamps et al., 2012). According to these studies, recycling subducted materials (e.g., continental crust (Willbold and Stracke, 2006) or subducted oceanic crust (Hofmann and White, 1982)) within the mantle is required to explain the geochemical variability in hotspot lavas. In contrast, some scientists believe that plumes are purely thermal features (e.g., Schuberth et al., 2009).

The role of mantle plumes in plate tectonics and the causes of plate reorganization has been previously studied (e.g., Ratcliff et al., 1998; Gerya et al., 2015; Van Hinsbergen et al., 2021). Using 2D numerical models, Ratcliff et al. (1998) showed that a rising mantle plume head can decouple a passive plate (a plate whose speed and direction of motion are driven by its adjacent active plate) from an active plate (Figure 1D). Their modeling results showed that following plume-lithosphere interaction, the passive plate changed its motion from eastward (Figure 1D2) to westward (Figure 1D3). They argued that this scenario is applicable for the tectonic evolution of the Pacific Plate and its neighboring plates during ∼143–122 Ma (Figure 1E). The Pacific, Izanagi and Farallon plates were likely all pulled to the southwest but with different velocities, due to the southwest-ward motion of the Phoenix plate prior to 122 Ma (Figure 1E1). There was no subduction around the Pacific Plate and the only mature subduction zone was the Phoenix Plate beneath what is now New Zealand and New Caledonia (i.e., Zealandia, Mortimer et al., 2017). Flood basalt eruption around 122 Ma to form the Ontong-Java plateau decoupled the Pacific, Izanagi and Farallon plates from the Phoenix’s mantle traction and led to plate reorganization (Figure 1E2). Recently, van Hinsbergen et al. (2021) proposed that a rising mantle plume below Madagascar-India around 105 Ma formed a >12,000 km long plate boundary between the Indian and African plates with subduction segments in the eastern Mediterranean and a ridge between India and Madagascar (Figure 1F). According to this study, plume-lithosphere interaction near a mid-oceanic ridge can cause plate rotation which may form a new subduction zone and/or rift. Subduction initiation induced by impingement of a mantle plume on lithosphere is another scenario which illustrates the importance of plumes in plate tectonics (e.g., Ueda et al., 2008; Gerya et al., 2015; Whattam and Stern, 2015). Plume-induced subduction initiation (PISI), because it doesn’t require any pre-existing lithospheric weaknesses and/or plate forces, can explain the development of initial subduction on Earth, which could have triggered the onset of plate tectonics (Gerya et al., 2015; Whattam and Stern, 2015).

Impingement of a mantle plume on continental lithosphere differs in terms of rheological and mechanical properties from plume-oceanic lithosphere interaction. Using numerical models, (Burov and Cloetingh, 2009; Burov and Cloetingh, 2010) investigated plume-induced sinking of intracontinental lithospheric mantle. They showed that arrival of a mantle plume below continental lithosphere eroded mantle lithosphere and allowed the plume to intrude into the crust. Brittle failure of the upper crust and detachment of mantle lithosphere forms a subduction-like motion (Figure 2A). Gravitationally unstable mantle lithosphere and the stratified structure of continental lithosphere plays a key role in causing a subduction-like process in this situation (Burov and Cloething, 2009; Burov and Cloething, 2010). One of the critical issues in these studies is the assumption of extremely high plume temperatures of 2,300°C. As indicated in the previous section, observations show that plume excess temperature with respect to ambient mantle can be 200–350°C. Recently, Cloetingh et al. (2021) revisited plume-induced sinking of intracontinental lithospheric mantle. They used the same model setup as Burov and Cloetingh. (2009) and Burov and Cloetingh. (2010) but with various continental lithospheric ages and different plume temperatures. Their 2D numerical models revealed that horizontal intrusion of a plume may occur at different levels of the continental lithosphere depending on lithosphere age and internal structure. Plume intrusion into buoyant continental crust and attendant sinking of continental mantle lithosphere occurs only if continental lithosphere is young (60–150 Myr) and plume is extremely hot (Figure 2B). In the case of older continental lithosphere, lithospheric strength prohibits penetration of plume material through the lithosphere and as a result plume-lithosphere interaction causes downthrusting of only the lowermost lithospheric mantle (Figure 2C). Cloetingh et al. (2021) carried out numerical experiments with more realistic plume excess temperature of ∼350 K. Regardless of lithosphere age, plume-lithosphere interaction in these models resulted only in sinking of the lowermost lithosphere into the asthenosphere (Figure 2D). It should be noted, however, that no melt-induced weakening of the lithosphere (e.g., Sobolev et al., 2011; Gerya et al., 2015; Gulcher et al., 2020) has been applied in these models.

Evidence of subduction-like sinking of continental lithosphere associated with upwelling of hot mantle is documented in Anatolia, the Caucasus, Central Asia, NE China, Iberia, Central Europe and the Colorado Plateau (Memiş et al., 2020; Cloetingth et al., 2021 and references therein). Seismic data commonly show a negative velocity anomaly—reflecting upwelling mantle material- and one or two areas with positive seismic anomalies—indicating sinking of continental lithosphere.

PISI has been the focus of several studies. In this section, we first review modeling studies that explore plume-induced subduction initiation. Then, we revisit works which document geologic evidence of subduction initiation triggered by a hot mantle plume head.

Recent studies based on modeling and geological-geochemical observations increasingly support the importance of plume-lithosphere interactions in plate tectonics. When the first modeling study of plume-induced subduction initiation was carried out by Ueda et al. (2008) there was no support from terrestrial geological evidence. At that time, the only observational evidence was those of Venusian coronae. However, after recognition of the first strong geologic evidence for plume-induced subduction initiation in the Caribbean region by Whattam and Stern (2015), the number of studies identifying examples of plume-induced subduction initiation has been growing. The main geological observation of PISI is the magmatic record. Overlapping isotopic compositions of post plume-lithosphere interaction and arc rocks can serve as key evidence for PISI. We note that the uncertainties in observations increase in the case of ancient sites of PISI (e.g., Eastern North China craton) due to overprinting of geological records.

The PISI hypothesis highlights the ability of rising plume heads to: 1) create new subduction zones, which can evolve into new plate boundaries on modern Earth (e.g., Ueda et al., 2008; Baes et al., 2016) and 2) may have triggered the onset of plate tectonics on Earth (Gerya et al., 2015). It is generally accepted that a plume head arriving at the base of the lithosphere partially melts to produce enormous volumes of magma. These flood basalt events can break the lithosphere to start a new subduction zone. It should be noted that only the plume head rather than its thinner tail is capable of triggering the onset of new subduction zones. Even then, the plume head must be unusually powerful and other conditions must be satisfied for a new subduction zone to form. This explains the rare nature of plume-induced subduction initiation events in the geologic record (Baes et al., 2016; Rodriguez, et al., 2021).

Numerical studies provide valuable information on facilitating and hindering factors involved in plume-induced subduction initiation (Figure 15). As the first step in PISI is penetration of plume into the lithosphere, the main difficulty in the earliest stage is to reduce and overcome lithospheric strength and break the plate across a region that is broad enough to allow adjacent portions to flex down. At this stage, flexural bending, strength (age) of the lithosphere, and thickness and strength of the crust are the hindering factors. Aging of the lithosphere has a dual effect; older lithosphere is stronger, hindering subduction initiation, but it is also a facilitating factor because it is denser and thus more unstable. Plume buoyancy is a critical driving force in the PISI process. Hotter and bigger plumes ease lithosphere breakage and onset of subduction. Gravitational forces due to overriding of the plume rocks is another factor helping to bend and sink the lithosphere into the mantle. Other facilitating parameters are far-field extensional or compressional forces from neighboring plates. Pre-existing lithospheric weaknesses and particularly magmatic weakening are also critically required for plume penetration. As lithosphere breaks and sinks into the mantle, water plays a key role in lubricating the plate interface to decouple subducting and overriding plates.

To investigate interaction of mantle plumes with lithosphere, two types of models have been employed: regional (e.g., Gerya et al., 2015) and global (e.g., Crameri and Tackley, 2016). Both regional and global PISI models show that subduction can be initiated in an extensional regime, similar to the result of Kemp and Stevenson (1996). It should be noted that in PISI regional models, there might be some artificial compressive stresses due to boundary conditions but they have secondary effects as the main driving force to start subduction in a PISI scenario is extension induced by plume head uprising. This can also be seen in experiments with extensional regimes which show even faster subduction initiation (e.g., Baes et al., 2020a).

Regional and global models have different strengths and weaknesses. The advantages of regional models are their higher resolution, more complex geometries and more appropriate material properties and physics (rheological stratification) of the lithosphere. The main weakness of these models are the assumed initial conditions. As these models simulate a piece of the whole Earth, caution is needed to appropriately define their initial conditions. In PISI regional models, which are the pioneering models, plume head size, shape and temperature are specified as initial conditions. These parameters should be defined based on observations or outcomes of global convection models. The other closely related weakness of these models is artificial compressive stresses due to closed box which have secondary-order effects on the results. Global 3D spherical models of convection have advantages of self-organized properties and structures of tectonic features like plumes, subduction zones and mid-oceanic ridges. As they simulate the whole Earth, they enable us to include interaction of different tectonic features in one model. This is in contrast to regional models which are able to simulate specific tectonic (not the whole) features of the Earth. The disadvantages of 3D spherical models are their lower resolution, refinement of the lithospheric rheology and lack of rheological stratification and minimal physics of the lithosphere compared to regional models. The other disadvantage of these models is that, due to self-organised tectonic structures and collective behavior, it is not easy to distinguish the effect of an individual feature on the resultant deformation regime. Despite the aforementioned differences, both regional and 3D spherical models to investigate PISI showed common results illustrating formation of a new subduction due to penetration of a plume head into the lithosphere and sinking of negatively buoyant lithosphere into the mantle.

It should be pointed out that although previous modeling studies have advanced our understanding about the central role of mantle plumes on formation of new converging plate boundaries, there is still room for further investigations. For instance, based on the model of Whattam and Stern (2015), in the Caribbean region a subduction had been ongoing along the northeastern margin of the Caribbean plateau for about 20 Myr when the Galapagos plume began to interact with the lithosphere at about 100 Ma. In modeling studies the effect of pre-existing subduction was only implicitly investigated by considering extensional regimes above the plume head associated with back-arc extension of this pre-existing subduction. Similarly, in the scenario of Stern and Dumitru (2019) the interaction of a plume head with lithosphere led to cessation of pre-existing subduction. The interaction of a plume head with a subducting slab was not studied in modeling studies. Therefore, more sophisticated modeling studies are required to address these and other unexplored cases.

There are several lines of evidence indicating that plume-lithosphere interaction was more common in the Precambrian than today (e.g., Gerya et al., 2015; Gao et al., 2019). One of the characteristics of the Precambrian is the likely existence of hotter plumes and thinner lithosphere. Geochemical observations of Precambrian greenstone belts suggest close relations between plume activity and subduction initiation (Puchtel et al., 1999; Moyen and Van Hunen, 2012; Gerya et al., 2015). Numerical models reveal that, depending on lithospheric strength and plume buoyancy, plume-lithosphere interaction in the Precambrian could have led to either plume-induced self-sustained subduction or plume-induced episodic drips (Gerya et al., 2015; Baes et al., 2020a). In either case, these were likely short-lived events (Gerya et al., 2015). These short-lived subduction zones or lithospheric drips may have facilitated lithospheric damage to start plate tectonics in accordance with Bercovici and Ricard (2014). Gerya et al. (2015) suggested that the transition from short-lived Archaean plume–induced self-sustained subduction or episodic drips to modern-style plate tectonics, which may have required secular cooling of the upper mantle, began with thickening and strengthening of the lithosphere. They also proposed that PISI could have started the first self-sustained subduction zone, triggering the onset of plate tectonics. Sobolev and Brown (2019) suggested that multiple PISI events might have dominated the tectonic regime before establishing global plate tectonics, causing episodic regional resurfacing at the Earth and production of the proto-continental crust. An alternative scenario for beginning plate tectonics was proposed by Rey et al. (2014), who suggested that gravitational spreading of continents at passive margins kick-started global plate tectonics. One of the uncertainties of this scenario is whether buoyant continental crust could have formed without plate tectonics.

The initiation of subduction alone is not necessarily indicative of the beginning of plate tectonics as modern plate tectonics is defined by the interaction of a mosaic of plates, which diverge from each other at mid-oceanic ridges and move towards each other at convergent plate margins. Numerical models show that trench retreat following PISI can produce a mosaic of new oceanic lithosphere in the back-arc region, indicating that a mantle plume could have started plate tectonics (Gerya et al., 2015). As plate tectonics began, an increasing number of plate tectonic features such as subduction zones, mid-oceanic ridges, transform faults, large-igneous provinces and plateaus created more complexity. Despite a long debate about whether lithospheric forces or mantle convection drives plate tectonics, there is common agreement among solid Earth scientists that slab pull and slab suction provide the most important driving force in the plate tectonics. This speaks to the importance of continued investigations of modeling and other studies of subduction processes from beginning to end.