Skip to content

Mantle Convection and Plumes

CSIR NET Geophysics Part C: Rayleigh number, boundary layers, whole vs layered convection, 410 and 660 phase changes, plumes, hotspots, plate driving forces.

  • 20 min read
  • Updated
On this page · 14 sections
  1. Prerequisites
  2. The Hook: A Solid That Flows
  3. Rayleigh–Bénard Convection
  4. Viscosity and the Structure of Mantle Flow
  5. Whole-Mantle or Layered Convection?
  6. Mantle Plumes
  7. What Drives the Plates
  8. Memory Tricks
  9. Don’t Confuse
  10. Common Mistakes
  11. Practice
  12. Quick Revision Table
  13. Key Takeaways
  14. What’s Next

Prerequisites

The Hook: A Solid That Flows

Rayleigh–Bénard Convection

The Rayleigh number

A fluid layer of thickness \( d \), heated from below with a temperature difference \( \Delta T \) across it, convects if buoyancy beats the two things that resist it: viscosity (which slows motion) and diffusion (which removes the temperature difference of a rising blob before it can rise). The ratio is the Rayleigh number:

$$\boxed{Ra = \frac{\rho g\alpha\Delta T\,d^3}{\kappa\eta}}$$

Convection begins when \( Ra \) exceeds a critical value \( Ra_c \) of order \( 10^3 \), which depends on the boundary conditions:

Boundaries\( Ra_c \)
both free-slipabout 657.5
one rigid, one free-slipabout 1,100.65
both rigidabout 1,707.8

For a layer heated internally (by radioactivity) rather than from below, an analogous Rayleigh number is written with heat production in place of \( \Delta T \); the conclusion for the mantle is the same.

Other dimensionless numbers

NumberDefinitionMantle valueMeaning
Prandtl \( Pr \)\( \eta/(\rho\kappa) \)about \( 10^{23} \)effectively infinite: momentum diffuses instantly, inertia is negligible
Reynolds \( Re \)\( \rho vL/\eta \)about \( 10^{-20} \)flow is perfectly laminar (creeping, Stokes flow)
Peclet \( Pe \)\( vL/\kappa \)a few thousand at plate speeds over mantle depthheat is carried mainly by advection
Nusselt \( Nu \)total heat flux / conductive heat fluxtenshow much convection boosts heat transport

Because inertia is negligible, mantle flow at any instant is set by the balance of buoyancy and viscous forces alone: the Stokes equations. Change the buoyancy and the whole flow field responds at once.

Boundary-layer theory

At high \( Ra \) the interior of a convecting layer is well mixed and nearly adiabatic; almost all the temperature change happens in thin thermal boundary layers at the top and bottom. A boundary layer grows by conduction until it is thick enough to be locally unstable (its own Rayleigh number reaches critical), then it detaches. This gives the scaling laws

$$\boxed{\frac{\delta}{d} \sim \left(\frac{Ra_c}{Ra}\right)^{1/3}} \qquad \boxed{Nu \sim \left(\frac{Ra}{Ra_c}\right)^{1/3}}$$

With \( Ra = 10^7 \), \( Ra_c = 10^3 \) and \( d = 2900 \) km, \( \delta \approx 2900\times(10^{-4})^{1/3} \approx 135 \) km: the thickness of oceanic lithosphere. That match is not a coincidence. The lithosphere is the top thermal boundary layer of mantle convection; the base of the mantle, D″, is the bottom one.

graph TD
    A[Top thermal boundary layer: plates] -->|cools, thickens, becomes dense| B[Subducting slabs: cold downwellings]
    B --> C[Sink through the transition zone into the lower mantle, or stall]
    C --> D[Bottom thermal boundary layer: D-double-prime]
    D -->|heated from the core, becomes buoyant| E[Plumes: hot, narrow upwellings]
    E --> F[Hotspots and large igneous provinces]
    G[Passive upwelling at ridges] --> A

Heating mode and the Urey ratio

The mantle is heated from within (decay of U, Th and K), from below (heat crossing the core–mantle boundary) and by its own secular cooling. Internally heated convection is dominated by strong cold downwellings with broad, diffuse return flow; bottom heating produces hot narrow upwellings. The mantle has both: slabs are the dominant downwellings, plumes carry the core’s heat. The Urey ratio (radiogenic heat production divided by total heat loss) is less than one: the Earth is losing heat faster than it makes it, and is cooling.

Viscosity and the Structure of Mantle Flow

ConstraintWhat it gives
Post-glacial rebound (Fennoscandia, Hudson Bay)upper-mantle viscosity about \( 10^{21} \) Pa s
Long-wavelength geoid and rebound data togetherlower mantle stiffer than upper mantle, by one to two orders of magnitude in most models
Laboratory creep lawsviscosity falls steeply with temperature (Arrhenius dependence) and with water content; dislocation creep is non-Newtonian

Strongly temperature-dependent viscosity makes the cold top boundary layer very stiff. On a planet where nothing breaks it, it becomes a stagnant lid (one-plate planets). On Earth the lithosphere breaks along weak plate boundaries, so the lid moves and sinks: plate-mode or mobile-lid convection.

Whole-Mantle or Layered Convection?

Line of evidencePoints towards
Seismic tomography: fast (cold) slabs beneath the Americas and the former Tethys continue into the lower mantlewhole-mantle flow
Tomography: slabs beneath parts of the western Pacific (Japan, Izu–Bonin) flatten and stall in the transition zoneat least temporary impedance at 660 km
Geochemistry: MORB source depleted in incompatible elements; ocean island basalts (OIB) more enriched, some with high \( ^3\text{He}/^4\text{He} \)distinct, long-lived reservoirs
Heat budget: depleted upper mantle cannot produce the heat lostsome deeper, more enriched material
Large low-shear-velocity provinces (LLSVPs) under Africa and the Pacificdense, possibly chemically distinct piles at the base

The prevailing view is whole-mantle convection with complications: slabs can stall at 660 km before descending, and chemically distinct material can survive in the deep mantle (LLSVPs, ultralow-velocity zones) and feed plumes. Strict two-layer convection, with no mass exchange at 660 km, is hard to reconcile with tomography.

Phase transitions as valves

The olivine transitions that make the 410 and 660 km discontinuities respond differently to temperature. For a transition with Clapeyron slope \( \gamma = dP/dT \):

$$\boxed{\Delta P = \gamma\,\Delta T} \qquad \boxed{\Delta z = \frac{\Delta P}{\rho g} = \frac{\gamma\,\Delta T}{\rho g}}$$
DiscontinuityReaction\( \gamma \)In a cold slabEffect on sinking
410 kmolivine → wadsleyitepositive (exothermic)occurs shallower (uplifted)dense phase appears earlier: aids sinking
660 kmringwoodite → bridgmanite + ferropericlasenegative (endothermic)occurs deeper (depressed)buoyant low-pressure phase persists: resists sinking

The negative slope at 660 km is why slabs can stall there and why the transition zone is thicker inside cold slabs and thinner in hot regions such as above plumes. Receiver-function measurements of transition-zone thickness are therefore used as a thermometer.

The base of the mantle

D″ is the lowermost few hundred kilometres of the mantle, a thermal (and probably chemical) boundary layer with strong lateral variation. The phase change from bridgmanite to post-perovskite has been proposed to explain a seismic discontinuity within it. Two continent-sized LLSVPs sit beneath Africa and the Pacific, and thin ultralow-velocity zones (possibly partially molten) lie at the core–mantle boundary in places. It has been proposed that plumes rise preferentially from the margins of the LLSVPs.

Mantle Plumes

What a plume is

A plume is a hot, narrow upwelling that rises from a thermal boundary layer, most likely D″. Laboratory and numerical models give a characteristic shape: a large head (a starting plume) followed by a thin tail (conduit) that can feed volcanism for tens of millions of years.

StageSurface expression
Head reaches the lithosphere and flattensdomal uplift, then a large igneous province (flood basalts) erupted in a few Myr or less; sometimes continental breakup
Tail continues to feedan age-progressive volcanic chain on the moving plate (hotspot track)
Throughoutbroad swell and geoid high; excess temperature of about 100–300 °C; OIB geochemistry; high \( ^3\text{He}/^4\text{He} \) in some

The classic Indian example is the Réunion plume: its head produced the Deccan Traps around the Cretaceous–Paleogene boundary, and its tail left the Chagos–Laccadive Ridge and Mascarene Plateau track as India moved north. The Kerguelen plume is linked to the Rajmahal Traps and the Ninetyeast Ridge. (Ages and the full list of LIPs are on the mantle plumes section of the Structural Geology chapter.)

How fast does a plume head rise?

A buoyant sphere of radius \( a \) and density deficit \( \Delta\rho \) rising through fluid of viscosity \( \eta \) moves at the Stokes velocity

$$\boxed{v_s = \frac{2\,\Delta\rho\,g\,a^2}{9\,\eta}} \qquad \Delta\rho = \rho\,\alpha\,\Delta T$$

A sphere much less viscous than its surroundings rises up to 1.5 times faster (the Hadamard–Rybczynski correction).

Buoyancy flux and heat flux from a swell

A plate moving at speed \( v \) over a plume is lifted into a swell of cross-sectional area \( S \) (perpendicular to the plate motion). The mass deficit supplied per second, the buoyancy flux, is

$$\boxed{B = (\rho_m - \rho_w)\,S\,v}$$

(\( \rho_w \) because water fills the space the swell displaces). The heat carried by the plume follows from \( B = \rho\alpha\Delta T\,Q_V \) and heat flux \( H = \rho c_p\Delta T\,Q_V \) (\( Q_V \) is volume flux):

$$\boxed{H = \frac{c_p}{\alpha}\,B}$$

Summed over all hotspots, plume heat flux estimated this way is only a small fraction of the global heat loss, much less than the heat lost through the plates. Since plumes are thought to carry heat out of the core, this has been used to argue that core heat flow is a modest share of the total, although estimates of core heat flow vary widely and some models give much larger values.

The plume debate

Not every hotspot is generally accepted as plume-fed. Points in dispute include the depth of origin (some low-velocity conduits are imaged into the lower mantle, others not), whether hotspots are fixed (paleomagnetism of the Emperor seamounts shows the Hawaiian hotspot moved south during their formation), and whether every LIP had the uplift a plume head predicts. The alternative “plate” view explains some intraplate volcanism by shallow processes (lithospheric extension and cracking, edge-driven convection, fertile mantle). For the exam: know the plume model, its predicted signatures, and that it is debated.

What Drives the Plates

Since plates are the top boundary layer of convection, “what drives plates” is the same question as “where is the buoyancy”. Order-of-magnitude estimates per unit length of boundary:

ForceEstimateComment
Slab pull\( \Delta\rho\,g\,h\,L \): with mean excess density about 65 kg m\(^{-3}\) (\( \rho_m\alpha\Delta T/2 \)), slab thickness 100 km, length 600 km, about \( 4\times 10^{13} \) N m\(^{-1}\)largest driving force; much is balanced by resistance as the slab sinks
Ridge push\( g\rho_m\alpha\Delta T\kappa t \approx 4\times 10^{12} \) N m\(^{-1}\) at 100 Myran order of magnitude smaller; a body force over the whole cooling plate
Basal tractiondepends on the relative motion of plate and underlying mantlecan drive or resist
Resisting forcesslab resistance, collisional resistance, transform frictionoppose motion

The Forsyth–Uyeda observation that plates attached to long subducting slabs move fastest, and the fact that plates without slabs move slowly, both point to slab pull as the leading term. Plates are not passive rafts dragged by convection cells beneath them; they are the convection’s own cold, sinking limb.

Dynamic topography. Mantle flow also pushes the surface up over upwellings and pulls it down over downwellings, by hundreds of metres to a kilometre or so. It contributes to swells, to long-wavelength geoid anomalies and to the vertical movements of continents.

Memory Tricks

Don’t Confuse

ThisvsThat
Rayleigh number (convective vigour)Reynolds number (inertia vs viscosity; tiny in the mantle)
Hotspot (the surface volcanic expression)Plume (the proposed deep upwelling beneath it)
Passive upwelling at ridges (plates pull apart)Active upwelling in plumes (buoyant from below)
Positive Clapeyron slope (410: shallower in cold slabs)Negative Clapeyron slope (660: deeper in cold slabs)
LLSVP (continent-sized, low shear velocity, base of mantle)ULVZ (thin, very low velocity, at the core–mantle boundary)
Whole-mantle convection (mass exchange across 660 km)Layered convection (separate upper and lower mantle cells)
Stagnant lid (one-plate planet)Mobile lid (plate tectonics)

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. What does the Rayleigh number compare?
  2. What is the critical Rayleigh number for a layer with two free-slip boundaries?
  3. Which mantle discontinuity has a negative Clapeyron slope?
Answers
  1. Buoyancy (driving convection) against viscous resistance and thermal diffusion (opposing it).
  2. About 657.5 (\( 27\pi^4/4 \)).
  3. The 660 km discontinuity (ringwoodite to bridgmanite plus ferropericlase).

Level 2: Part B Style

Practice Q1. With \( \rho = 4000 \) kg m\(^{-3}\), \( g = 10 \) m s\(^{-2}\), \( \alpha = 2\times 10^{-5} \) K\(^{-1}\), \( \Delta T = 2500 \) K, \( d = 2900 \) km, \( \kappa = 10^{-6} \) m\(^2\) s\(^{-1}\) and \( \eta = 10^{22} \) Pa s, the Rayleigh number of the mantle is closest to:

(a) \( 5\times 10^{3} \) (b) \( 5\times 10^{6} \) (c) \( 5\times 10^{9} \) (d) \( 5\times 10^{12} \)

Answer: (b) 5 × 10^6

\( \rho g\alpha\Delta T = 4000\times 10\times 2\times 10^{-5}\times 2500 = 2000 \). \( d^3 = 2.44\times 10^{19} \). Numerator \( 4.9\times 10^{22} \). Denominator \( 10^{-6}\times 10^{22} = 10^{16} \). \( Ra \approx 4.9\times 10^{6} \), thousands of times critical.

Practice Q2. For an upper-mantle viscosity of \( 10^{21} \) Pa s and rigidity \( 7\times 10^{10} \) Pa, the Maxwell relaxation time is about:

(a) 4.5 years (b) 450 years (c) 45 kyr (d) 4.5 Myr

Answer: (b) 450 years

\( \tau = \eta/\mu = 10^{21}/(7\times 10^{10}) = 1.43\times 10^{10} \) s \( \approx 450 \) years.

Practice Q3. Inside a cold subducting slab, compared with the surrounding mantle:

(a) the 410 km transition is deeper and the 660 km transition shallower (b) the 410 km transition is shallower and the 660 km transition deeper (c) both transitions are shallower (d) both transitions are deeper

Answer: (b)

At 410 km \( \gamma > 0 \): lower \( T \) means lower transition pressure, so it is shallower. At 660 km \( \gamma < 0 \): lower \( T \) means higher pressure, so deeper. The transition zone is thicker in the slab.

Practice Q4. If the Rayleigh number of a convecting layer increases by a factor of 1,000, the thickness of its thermal boundary layers changes by a factor of about:

(a) 1/1000 (b) 1/31.6 (c) 1/10 (d) 10

Answer: (c) 1/10

\( \delta\propto Ra^{-1/3} \): \( 1000^{-1/3} = 1/10 \). (b) would be the \( -1/2 \) power.

Level 3: Part C Style

Practice Q5. A spherical plume head of radius 200 km and excess temperature 300 K rises through mantle of viscosity \( 10^{21} \) Pa s (\( \rho = 3300 \) kg m\(^{-3}\), \( \alpha = 3\times 10^{-5} \) K\(^{-1}\), \( g = 10 \) m s\(^{-2}\)). Treat it as a rigid Stokes sphere. Which statements are correct?

  1. Its density deficit is about 30 kg m\(^{-3}\).
  2. It rises at about 8 cm yr\(^{-1}\).
  3. It would take about 35 Myr to rise through 2,900 km.
  4. A head of twice the radius would rise about twice as fast.

(a) 1 and 2 only (b) 1, 2 and 3 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4

Answer: (b) 1, 2 and 3 only
  • 1: \( 3300\times 3\times 10^{-5}\times 300 = 29.7 \) kg m\(^{-3}\). True.
  • 2: \( v = \frac{2\times 29.7\times 10\times 4\times 10^{10}}{9\times 10^{21}} = 2.64\times 10^{-9} \) m s\(^{-1}\) = 8.3 cm yr\(^{-1}\). True.
  • 3: \( 2900/83 \approx 35 \) Myr. True.
  • 4: False. \( v\propto a^2 \): four times as fast.

Practice Q6. A hotspot swell has a cross-sectional area of \( 10^9 \) m\(^2\) perpendicular to the motion of a plate moving at 8 cm yr\(^{-1}\). Take \( \rho_m - \rho_w = 2300 \) kg m\(^{-3}\), \( c_p = 1250 \) J kg\(^{-1}\) K\(^{-1}\), \( \alpha = 3\times 10^{-5} \) K\(^{-1}\). Which statements are correct?

  1. The buoyancy flux is about \( 5.8\times 10^{3} \) kg s\(^{-1}\).
  2. The plume heat flux is about 0.24 TW.
  3. This plume carries more than 5% of Earth’s total heat loss.
  4. For the same plume beneath a plate moving half as fast, the swell cross-section would be halved.

(a) 1 and 2 only (b) 1, 2 and 4 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (a) 1 and 2 only
  • 1: \( v = 2.5\times 10^{-9} \) m s\(^{-1}\); \( B = 2300\times 10^9\times 2.5\times 10^{-9} = 5.8\times 10^3 \) kg s\(^{-1}\). True.
  • 2: \( H = (1250/3\times 10^{-5})\times 5.8\times 10^3 = 2.4\times 10^{11} \) W. True.
  • 3: False. \( 0.24/46 \approx 0.5\% \).
  • 4: False. \( B = (\rho_m - \rho_w)Sv \) is fixed by the plume; halving \( v \) doubles \( S \).

Practice Q7. A slab is 600 K colder than the surrounding mantle at the transition zone. Take Clapeyron slopes of \( +3 \) MPa K\(^{-1}\) at 410 km and \( -2 \) MPa K\(^{-1}\) at 660 km, and \( \rho g = 4\times 10^4 \) Pa m\(^{-1}\) (illustrative values). Which statements are correct?

  1. Inside the slab the 410 km transition is raised by about 45 km.
  2. Inside the slab the 660 km transition is depressed by about 30 km.
  3. The deflected 660 km transition makes the slab denser than its surroundings at that depth and so aids sinking.
  4. The transition zone inside the slab is about 75 km thicker than outside.

(a) 1, 2 and 4 only (b) 1 and 3 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 4 only
  • 1: \( \Delta P = 3\times 10^6\times(-600) = -1.8 \) GPa; \( \Delta z = 1.8\times 10^9/4\times 10^4 = 45 \) km shallower. True.
  • 2: \( \Delta P = (-2\times 10^6)\times(-600) = +1.2 \) GPa; 30 km deeper. True.
  • 3: False. Between the normal and the depressed 660, the slab still has low-density ringwoodite while the mantle around it has transformed to denser bridgmanite plus ferropericlase: the slab is relatively buoyant there, and the transition resists sinking.
  • 4: \( 45 + 30 = 75 \) km. True.

Quick Revision Table

IdeaFact or formula
Maxwell time\( \eta/\mu \approx 450 \) yr for \( 10^{21} \) Pa s
Rayleigh number\( Ra = \rho g\alpha\Delta T d^3/\kappa\eta \); mantle about \( 10^6 \)–\( 10^8 \)
Critical \( Ra \)657.5 (free–free), 1,100.65 (rigid–free), 1,707.8 (rigid–rigid)
Prandtl, Reynolds\( Pr\sim 10^{23} \) (infinite); \( Re\sim 10^{-20} \) (creeping flow)
Boundary layers\( \delta/d\sim(Ra_c/Ra)^{1/3} \); \( Nu\sim(Ra/Ra_c)^{1/3} \)
Lithospherethe top thermal boundary layer; D″ the bottom one
Heatinginternal (U, Th, K), basal (core), secular cooling; Urey ratio below 1
Viscosityupper mantle about \( 10^{21} \) Pa s; lower mantle stiffer
Convection stylewhole-mantle with slab stalling and deep chemical piles
Clapeyron410 positive (aids sinking), 660 negative (resists); \( \Delta z = \gamma\Delta T/\rho g \)
Deep mantleD″, post-perovskite, LLSVPs (Africa, Pacific), ULVZs
Stokes\( v = 2\Delta\rho ga^2/9\eta \), \( \Delta\rho = \rho\alpha\Delta T \)
Buoyancy flux\( B = (\rho_m - \rho_w)Sv \); heat flux \( H = c_pB/\alpha \)
Plume stageshead → uplift and LIP; tail → age-progressive track
Indian plumesRéunion (Deccan, Chagos–Laccadive), Kerguelen (Rajmahal, Ninetyeast)
Forcesslab pull about \( 10^{13} \) N m\(^{-1}\) > ridge push about \( 10^{12} \) N m\(^{-1}\)

Key Takeaways

What’s Next

Advertisement
Advertisement