Roadmap | The review process | Home
   Definition of a plume

A cactoplume, by Erik Lundin

Definition of a Plume

Don L. Anderson

Seismological Laboratory, California Institute of Technology, Pasadena, CA 91125

…you cannot prove a vague theory wrong. – Richard P. Feynman

It is impossible to dissociate language from science or science from language, because every natural science requires words in which the concepts are expressed. To call forth a concept, a word is needed; to portray a phenomenon, a concept is needed… – Antoine Lavoisier


The fluid dynamic definition of a plume in an homogeneous fluid may not be appropriate for features in the mantle. Thus, the petrological and geochemical usages of the term have become vague and flexible and it is difficult to obtain agreement among workers. As Lavoisier famously said “…it is impossible to dissociate science from language…”; in order to make progress we need clear concepts and a precise language.


Recent studies do not support the existence of Core-Mantle Boundary (CMB) type temperatures (ΔT~1500°C) or thermally driven mantle plumes originating from the CMB as the cause of ocean-island magmatism. Most Ocean Island Basalts (OIB) have inferred temperatures that overlap the range inferred at midocean ridges (see Thermal pages). In some studies, the temperatures inferred from magma at Hawaii fall outside the range of MORB but this is the only such location, for modern basalts, and the maximum anomaly is ΔT < 150°C. Even these modest temperature anomalies are contested [see Primary magmas at mid-ocean ridges]. Hawaiian xenoliths do not support such high temperatures and other studies of magmas have a smaller excess temperature. Plausible plume temperatures also cannot provide large degrees of melting of peridotite at midplate locations. The rarity of high-MgO melts from the mantle is difficult to understand. The uplifts and heatflows expected to be associated with plumes are also not observed at places thought to be affected by plumes. It is important therefore to reconsider the definition of ‘plume’ and the criteria by which true plumes can be distinguished, if they exist, from normal mantle variations. Recently, the definitions of plumes and plume criteria have multiplied and the concept has become rather fuzzy. It has even been suggested that plumes cannot be defined, and that all plumes are different [see The Great Plume Debate].

What are the other options for creating the volcanic features that have been attributed to plumes or to high temperatures, or for localizing the volcanism? Options include lithospheric (stress, architecture, fabric), petrologic (mantle fertility, lithology, melting point, volatile content), and tectonic (subduction of seamount chains, edge effects, dikes, focusing and delamination). These are all basically athermal features or mechanisms and involve parameters other than absolute temperature. They are all natural manifestations of plate tectonics. It is possible that the full range of the inferred temperature excess at the CMB, ~1500°C, is not represented in surface magmas, and that the range in MORB is not the full range of mantle potential temperature below the surface TBL.

The conditions for the formation of a volcano include high homologous temperature, the eruptability of the magma and the stress state of the lithosphere; high absolute temperature is not a requirement. The absence of a volcano does not imply the absence of melt. The whole upper mantle, or asthenosphere, could in theory be above the solidus and volcanoes would still be restricted to those points or lines where the local stress state or plate fabric allows extrusion. The same mantle can provide underplating, ponding and sill intrusion but eruptions do not occur unless the stress conditions allow extrusion and dikes. Alternatively, the whole plate could be under extension but volcanism can be confined to those places where the melting temperature is particularly low. The thermal uplift, lithospheric erosion, high heatflow and high mantle potential temperatures that were once the hallmark of the mantle plume hypothesis have been abandoned by many in favor of non-thermal criteria, including helium isotopic ratios, buoyancy flux, linearity and age progressions of island chains, although there is some circular reasoning involved in these choices.

There is now confusion amongst plume advocates about just what a plume is and what are the criteria that can be used to identify plumes [see also What is a plume? for numerous definitions of mantle plumes]. ‘Low’ seismic velocities and ‘anomalous’ geochemical signatures are now considered the main diagnostics although high seismic velocities have also been considered as evidence for plumes. Data are invariably filtered and selected 'to remove plume influence' before the judgments are made, so standard statistical methods cannot be applied. What constitutes ‘ambient’ mantle, and what constitutes an ‘anomaly’ are central to the question of whether plumes even exist. Currently, the mantle is assumed to be well homogenized by chaotic convection and any deviation in temperature or composition from what is assumed to be ambient is attributed to a plume, viz. 'plumes all have in common the intrusion of anomalous mantle into ambient mantle'. If ambient mantle is chemically heterogeneous and variable in temperature then even this broad definition of plume fails. Delamination and subduction also involve the insertion of one kind of material into another.

Definition of a Plume

Stress- and dike-based models recognize that the difference between a surface volcano (‘hotspot’) and sills, intrusion, magma chambers and underplating is simply one of the orientation of the stress field, not the absolute temperature of the mantle, or even the magnitude of the stress. Low seismic velocities, large igneous provinces and large buoyancy fluxes are treated by some as unambiguous evidence for deep mantle thermal plumes, but this is not a universal view. One is tempted to pick and choose among criteria and features to identify those that are real plumes. Alternatively, if the definition of a plume is too broad, or flexible, or circular, then the term ’plume' carries no information. The words ‘anomaly’ or ‘anomalous’ are parts of most modern definitions of ‘plume’ and therefore there is a statistical element; how anomalous, how hot, and how large (in the case of LIPs) become issues, as does the definition of ‘ambient’ and ‘average’.

The general usage of the term ‘plume’ in the geophysics and geochemistry literature implies the following definition; A geophysical plume is the result of a thermal instability created in a homogeneous fluid heated from below that rises as a bulbous head followed by a cylindrical tail. [but for a myriad of other definitions, see What is a plume?]. Recently, the word has been applied to a variety of features and properties, and plumes have been invoked to explain or rationalize a large number of different observations.

A plume is basically a fluid dynamic and thermal concept. The concept as applied in geophysics and geochemistry was not based on any direct measurements of morphology, depth or temperature; magma volumes and seismic velocities were used as proxies for temperature. These properties plus heatflow, composition and age progressions were used to support the contention that plumes exist in Earth’s mantle, but they can also be satisfied by non-thermal processes and features such as fertile or low melting blobs, and internal instabilities created by heating of material dumped into the mantle from above. If the word ‘plume’ is applied to any ‘anomalous feature’ then it is not a useful term if one is interested in investigating the cause of the ‘anomaly’. Indeed, it can proactively do harm, since by implying that the cause of the surface observations is known, researchers are discouraged from further enquiry.

The cactoplume – the ultimate fix for any surface phenomenon (by Erik Lundin). Click here or on figure for enlargement.

Definition of ‘Anomalous’

Considering the range of basalt chemistries observed at ocean islands, island arcs and on continents, midocean ridge basalts (MORB) are remarkably uniform. But even these have been subdivided into DMORB, NMORB, TMORB, EMORB, PMORB and OIB. MORB found along ridges at the onset of continental separation, in narrow ocean basins and at slow spreading ridges differs from MORB found at mature and fast spreading ridges. Normal MORB (NMORB) is fairly homogeneous geochemically and thermally and is considered, in many studies, to be a perfect starting point for looking at anomalies. Fast-spreading ridges sample and average a large volume of the mantle and, as predicted by the Central Limit Theorem (CLT; see also Mantle reservoirs and Statistics pages), yield a uniform product with small variance, even if the mantle is heterogeneous. A similar rationale is applied to Standard Mean Ocean Water (SMOW), which is a well-mixed reference material used for studying oxygen isotopes. Basalts from volcanoes that sample restricted regions will display larger variance and will contain extremes of composition that are averaged out in MORB. If the CLT is the only factor that controls the composition of basalts then the variance will increase as the sampled volume decreases, e.g., from East Pacific Rise MORB, to slow-spreading ridges, to OIB, to seamounts, to xenoliths, and finally to fluid inclusions. The challenge to petrology is to distinguish sampling variations form true lithologic variations, called ‘reservoirs’. Large heterogeneities such as delaminated crust and subducted seamount chains may survive in the mantle for long times and will not be averaged out when sampled by a single volcano.

Currently, the mantle is often assumed to be isothermal and homogeneous, mainly because the most homogeneous product of the mantle, NMORB, is relatively homogeneous and isothermal. The CLT tells us that a homogeneous product does not imply a homogeneous source. Normal ridges are expected to have the same elevation and productivity and constant geochemical properties; departures from these conditions are then viewed as ‘anomalous’. A convecting fluid cannot be isothermal, and a plate tectonic mantle, involving recycling, cannot be homogeneous or have a constant melting temperature. The ‘need’ for plumes can only be established if temperatures or compositions are outside the range expected from plate tectonic processes, which involves heating from within, cooling from above, and recycling. Only things that the Earth ‘needs’ exist. Since MORB involves large-scale sampling and averaging, and blending of magmas, it is not sufficient to find components that fall outside the MORB range.

Normal mantle convection involves thermal features that are thousands of km in extent, and chemical features that can range from millimeters to thousands of km in scale. Most are tens to hundreds of km across, however. Blobs that are tens of km in dimension are interesting since they can decouple from mantle flow and are in the range inferred from seismic scattering experiments. Delaminated crust can become buoyant as it warms up to ambient mantle temperature. Dikes are buoyant upwellings that are of the order of meters to tens of meters across. Downwelling dense eclogite slabs, particularly those containing CO2, can have very low seismic velocities. There are several low-shear-velocity features in the deep mantle that have dimensions of thousands of km, but they also have high bulk modulus and density. It is not helpful to call all of these features ‘plumes’ or ‘superplumes’. Likewise, some magmas and restites can have very high – 3He/4He ratios - and very low 3He contents – and by some conventions these are plumes. Although it is polite to be inclusive, a wide open ‘definition’ of ‘plume’ does not advance understanding. Science requires precise, testable and useful definitions. The vague and flexible definitions of ‘aether’ and ‘phlogiston’ held up progress in physics and chemistry for centuries.

Mantle Temperatures and the Plume Hypothesis

Mantle temperatures inferred from petrology depend on assumed source lithology and volatile content and most current petrological studies either find no difference in potential temperature between ridge and ‘hotspot’ magmas [see also Falloon et al., 2007], or exhibit differences that are in the range of normal plate tectonic models, with no lower thermal boundary layer (~200°C). In fact, for a given heatflow, the interior temperatures of the mantle are higher, and show more variability, for purely internal heating and cooling from above than for plume models (lower boundary heating from below). When there are two or more TBL the temperature rise across the mantle is distributed among them. If the intrinsic density increases with depth (e.g., crust, harzburgite, perisphere, fertile peridotite, eclogite) then the temperature profile will be superadiabatic – a conduction geotherm – and the temperatures of basalts removed form various levels will increase by about 10°C/km. This is one way to explain the higher temperature magmas that emerge from beneath thick lithosphere. In a mantle that is heated by radioactivity and cooled by sinking or bottomed-out slabs, the temperature gradient can be subadiabatic. Therefore, in contrast to the interior of an ideal convecting fluid heated entirely from below, there can be a substantial dependence of magma temperatures with depth of extraction. Potential temperatures inferred from magmas do not automatically yield the potential temperature of the underlying mantle nor do they imply that the mantle geotherm is an adiabat.

Potential temperatures at ‘normal’ oceanic ridges, or for ‘ambient’ mantle, are given by various recent authors as 1243-1351°C, 1280-1400°C, 1300-1570°C and 1400-1600°C. ‘Normal’ is defined as ridges ‘away from the influence of hotspots or mantle plumes’. Usually, ‘normal’ also means ‘a mature ridge’, i.e. not a new ridge and not near a continent, or in a narrow ocean basin. Particularly deep or shallow, or particularly slowly spreading ridges are also not considered normal. The fuzziness in the definition of ‘normal’ is partly responsible for the above disagreements but there is still controversy on how to infer source temperatures from magma compositions, and how to allow for source composition and volatile content. ‘Ambient’ or normal mantle is also considered to be the same as ridge mantle although some basalts may come from depths that are shallower than the so-called ‘fully convective layer’ or ‘adiabatic mantle’. The TBL may extend deeper than the ‘lithosphere’ and need not be entirely subsolidus.

The above values may be representative of the shallow mantle under deep, mature and fast spreading ridges, at least if that mantle is similar to dry pyrolite. How representative are they of ‘normal’ mantle, considering that most mantle is not represented? First, they are probably biased to the low side. Mantle newly uncovered by continental drift, under thick lithosphere, shallow ridges, and the interiors of large long-lived plates is likely to be hotter. Magmas from these regions are usually considered to be due to plumes rather than due to normal variations of mantle temperature away from mature ridges. If the mantle is not adiabatic then shallow magmas are likely to be colder than deeper magmas. On the other hand, ridges that are deep or near subduction zones may be colder, but these are not excluded as systematically as are shallow ridges and newly formed ridges. The actual range, or variance, of ‘normal’ mantle temperatures is likely to be much greater than given in the above complilations.

A conservative estimate of the range of mantle potential temperatures, given the above considerations, might be 250°C. If ridge temperature represents the low end of mantle temperatures then 1500°C seems to be a plausible upper bound, at least for long wavelength variations. The CLT also tells us that variance and extreme values depend on wavelength. If ridge chemistry, temperature and elevation average over, say, 400 km, then variations can be greater over individual island and seamount scales. The most rapid spatial variations are, however, vertical – across TBL – except where dikes and delamination allow asthenospheric temperatures to be brought into contact with lithospheric and crustal temperatures. A temperature of 1350°C is achieved at a depth of 135 km for a conduction gradient of 10°C/km. If the TBL extends to a depth of 155 km under Hawaii or through a buoyant perisphere or continental lithosphere then the mantle temperature can be 200°C hotter, or 1550°C. Such temperatures are not inconsistent with geophysical data; they are a little high, if extended adiabatically to 420 and 650 km, if these mantle discontinuities are due to equilibrium phase changes in olivine. However, subadiabatic gradients are likely if slabs and delaminates bottom out near the transition zone, and if the mantle is heated from within.

The above estimates of temperature may be representative of the shallow mantle under deep, mature and fast spreading ridges; other locations are often excluded from analysis. An assumption in the petrological estimates is that the mantle is similar to dry pyrolite. How representative are these values of ‘normal’ mantle, considering that most mantle is not represented? First, they are probably biased to the low side. Mantle newly uncovered by continental drift, under thick lithosphere, shallow ridges, and the interiors of large long-lived plates is likely to be hotter. Magmas from these regions are usually considered to be due to plumes rather than due to normal variations of mantle temperature away from mature ridges. If the mantle is not adiabatic then shallow magmas are likely to be colder than deeper magmas. On the other hand, ridges that are deep bathymetrically or near subduction zones may be from colder mantle; these are not excluded as systematically as are shallow ridges and newly formed ridges. The actual range, or variance, of ‘normal’ mantle temperatures is likely to be much greater than given in the above complilations.

According to early influential estimates the potential temperature below ridges is 1280°C and beneath Hawaii is ~1550°C. Hawaii is often considered to be a plume because there is no evidence that it is on a preexisting tectonic trend or is part of a broad region of hot ambient mantle [see also Hawaii pages]. Magmatism may, however, be localized by lithospheric stress or architecture, or by low melting point or fertile mantle. Early studies predicted that the lithosphere under Hawaii would be thinned by plume heating [see also Why is heat flow not high at hotspots?]. The lithosphere is actually thick, but it is not yet clear whether this condition preceded the magmatism, or was a result of it. Some authors conclude that OIB can be '50-150°C hotter than maximum ridge potential temperatures', away from the influence of hotspots, assuming identical compositions and derivation from below the surface TBL. Some ocean island basalts (Hawaii, Iceland, Reunion) imply potential mantle temperatures of 1286-1372°C, when volatile contents are taken into account, although this range is contested. Part of the Emperor chain may lie on preexisting tectonic features [see also Norton, 2007]. The subduction of an island chain or the delamination of an island arc is perhaps the easiest way to visualize a fertile streak in the mantle that is unlikely to be homogenized by chaotic convection. Global stress maps show extensional stresses in the vicinity of Hawaii [see also Stuart et al., 2007], and most other hotspots and ridges. The orientation of the stress is consistent with the orientation of the islands and the direction of new volcanism.


The original definitions of plumes involved high temperature, great depth, rapid upwelling and a mechanism that included bottom heating of an isothermal homogeneous fluid. The mantle is far from these conditions. Magmatism on the Earth involves heterogeneity of the mantle, stress state of the lithosphere, recycling, plate tectonics and a range in melting behaviors. Large volumes of melt can be generated from particularly fertile mantle as well as from particularly hot mantle.

last updated 3rd December, 2006