1. The dynamics of subduction throughout the Earth's history
Jeroen van Hunen
Durham University, UK
Thanks to: Jon Davidson (Durham)
Jean-Francois Moyen (St. Etienne)
Arie van den Berg (Utrecht)
Taras Gerya (ETH)

2. In this talk Subduction and Earth evolution
Since when did subduction operate?
Theories
Observables
Did subduction style change over time?
What I plan to discuss:

3. How was Earth different in the past?
Produced 3x as
much radiogenic heat
So what was different in Archaean?
Earth is a very strong heater: it produces a lot of heat insight from decay of radiogenic elements. This decay is going on for 4.5 Ga, suggests that’s why is was higher in past: about 3x in the Archean. This is fairly well constrained, apart from some uncertainty about the original composition of the Earth’s
Also Earth has been cooling: as I said today K/Grs, and geological record of MORBs gives indication of mantle T in past: depending on a few assumptions hotter in Archean.
was K hotter
(Herzberg et al., 2010)

4. Consequences of more radiogenic heat
Today’s surface heat flux of 80 mW/m2:
50% = from radiogenic heat production
50% = Earth cooling
To have Earth cooling
in Archaean, we need
a cooling mechanism
more efficient than
plate tectonics (PT)
This additional heat production can be a serious problem for PT:
Today, about 80 mW/m2 of heat flows from Earth, so football field produces enough heat energy to run a microwave oven on. This is mostly coming out of oceans and is the result of PT: without PT there would be much less heat escaping.
½ is produced by radioactive heating, other half is primordial heat: heat that is still there from Earth formation, and is slowly released (rough numbers, if Earth is chondritic, radiogenic component is larger than 50%).
In past, more radiogenic heat. So if we want to avoid heat accumulation, and we assume PT in past, then PT must have been capable to remove this extra heat as well. So PT must have been faster/more efficient. If not, the Earth was not cooling as fast, or perhaps not cooling at all, but heating up.
Same arguments for any other kind of tectonics.
(Sleep, 2000; Turcotte and Schubert, 2002)

5. Archaean mantle was 100-300 K hotter
Significantly hotter Archaean
mantle (Nisbet et al., 1993; Abbott et al., 1994)
Wet, slightly hotter
Archean mantle
(Grove and Parman, 2004)
Some debate about how much hotter. Main evidence comes from komatiites, which suggest large degrees of mantle melting.
Depending on how this is interpreted, this suggests a mantle of K hotter. Lower bound if mantle was also wetter, upper bound if extra melting comes mainly from the higher mantle temperatures
Peak temperature
in Archaean?
(Herzberg et al., 2010)

6. Consequences of a hotter mantle
More melting at mid-ocean ridges
thicker oceanic crust
thicker harzburgitic melt residue
layer
Weaker plate and mantle material:
h = exp (T)
~1 order of magnitude for every 100 K
Effect of dehydration strengthening?
early Earth?
today
Hotter Earth also had consequences:
Most materials (including rocks) become weaker when heated up: rule of thumb: 1 oom per 100 degrees.
Mantle melts close to surface, 7 km crust.
In hotter Earth more melting, thicker crust, perhaps 20 km instead of 7.
Plot shows expected crusal thickness. Btw, Tpot is mantle T extrapolated to Earth’s surface.
(van Thienen & al., 2004)

7. Consequences of more melting
thick crust/harzburgite
low average density r
no slab pull?
no subduction? (Ontong Java)
no plate tectonics?
very low r
low r
normal r
lithosphere
today
Archaean
crust
harzburgite
peridotite
This has significant dynamical consequences: crust (basalt) is buoyant: its density is lower than mantle density. So it floats. Today this is not hugely important, and doesn’t prevent plate from subducting. But if crust is 3x as thick, it IS a problem: no way this material is going to go down.
Today we see this at Ontong Java, has anomalously thick crust, and is approaching the subduction zone. This socalled ‘oceanic plateau’ is not going to subduct.
(Davies, 1992)

8. Effect of basalt-eclogite transition?
No subduction
Subduction
density basaltic crust
density eclogitic crust
Costa Rica subduction zone
Meta-stable basalt
transition gradual
Not full story, and there might be a way out.
B->E
Eclogite is denser, even a little denser than mantle material, so more of it would only help subduction.
There is one problem: transition only once material at 40 km depth. So how to get it there? Chicken&egg problem.
So subduction might benefit from this transition, but the fact remains that crust has to be brought to 40 km depth first, and this requires force.
(Cloos, 1993; Hacker, et al., 2003)

9. Stronger Archaean plates?
harzburgite = dry = strong
plate bending more difficult?
slower Archaean plate motion?
fits with supercontinent ages
But:
plate strength in cold top part
plates bending induces faulting + rehydration
(Faccenda et al., 2008)
(Korenaga, 2006)

10. Weaker Archaean plates?
colors =
viscosity
black
basalt
white
eclogite
time
Adding this transition and other complexities to the problem, it quickly becomes too large for a back-of-the-envelope calculation.
I did computer modelling to look at subduction dynamics in hotter Earth, when crust is thicker.
Explain plots: colours, axes, columns, rows.
viscosity
DTmantle = 0oC oC oC oC
(van Hunen & van den Berg, 2008)

11. Weaker Archaean plates?
colors =
viscosity
black
basalt
white
eclogite
time
Today and 100 K hotter are similar. Quantitative differences, but qualitatively same: PT works in a similar way.
viscosity
For low Tmantle subduction looks like today’s
(van Hunen & van den Berg, 2008)

12. Weaker Archaean plates?
colors =
viscosity
black
basalt
white
eclogite
time
Increasing T more gives problems: thick crust starts to produce large buoyancy stresses, and at same time slab becomes weak, because it is hotter. This leads to more and more slab breakoff. Consequence: broken-off slab offers no slab pull anymore: subduction slows down …
viscosity
For higher Tmantle frequent slab break-off occurs …
(van Hunen & van den Berg, 2008)

13. Weaker Archaean plates?
colors =
viscosity
black
basalt
white
eclogite
time
And may even completely stop.
So for too hot mantle, PT is not viable anymore.
viscosity
… or subduction completely stops.
(van Hunen & van den Berg, 2008)

14. Summary of many model calculations
Summary of all results in terms of average subduction speed vs. Tpot.
Almost all models show speedup for DT=0100K
Most models have peak vsubd at DT=100 or 200 K and start to decrease afterwards again.
So what kind of cooling history would those models give?
Are these subduction velocities enough to cool early Earth?

15. Possible parameterizations of vsubd1 3 2
To answer this, I cleaned up this spaghetti plot, and parameterized the lines into simple functions, looking at 3 different endmember scenarios.
Simply assume vsubd didn’t change
vsubd keeps on increasing with incresaing mantle temperature.
Vsubd increased at first (weaker mantle), but then decreased (resistance from buoyant crust and breakoff)
Are these subduction velocities enough to cool early Earth?

16. Thermal evolution the Earth
Heat production
Surface heat flow 1 3 2 ?
We want to calculate the temperature of the Earth T.
Translating T to heat energy requires heat capacity C (in J/K): total heat of the Earth = C*T
What we know is the sources and sinks of the heat, so we know what causes T to change over time, see eqn in slide.
H=source, Q=sink.
(Reymer and Schubert, 1984; Sleep, 2000; van Thienen et al., 2005)
(Herzberg et al., 2010)

17. Model 1: subduction for all Tm
Flat vsubd rate:
cooling since early Archean
Cooling curve similar to Korenaga,’06
and Labrosse & Jaupart,’07.
Main plot here is on right: T and v versus time.
Top-left shows again vsubd as function of Tpot, and bottom-left shows resulting heat flow.
Flat vsubd  almost constant qs, but since early Earth produced more heat, it couldn’t get rid of it all, and heated up in beginning. But we have to be careful of any plots like these for Hadean: highly speculative
Results look a lot like those suggested by Korenaga,’06 and Labrosse and Jaupart (’06)

18. Model 2: Rapid plate tectonics  efficient cooling  ‘thermal catastrophe’
Increasing vsubd with Tpot:
‘Thermal catastrophe’
2nd scenario: hotter weaker Earth gives faster plate tectonics.
This scenario almost always leads to a ‘thermal catastrophe’: an incredibly hot Archaean Earth:
Positive feedback: hotter Earth  faster tectonics  faster cooling  Earth started off even hotter, etc.

19. Model 3: Inefficient subduction  hotter Archaean mantle
Peak in vsubd:
Recent rapid cooling since Proterozoic
Model 3 gives more reasonable results: Until Arch-Prot. Earth was in downgoing branch of curve in top-left: hotter Earth gives slower vsubd, and therefore less surface heat release, and so even hotter Earth. However, heat production diminishes more rapidly than surface heatflow, so heating up of Earth reduces and even stops at 2.5 Ga, after which Earth rapidly cools down. This curve suggests that PT was most effective around 1 Ga.

20. Observations Slave Isua Barberton Pilbara Jack Hills
Only very few old rocks are preserved.
Slave
Isua
Barberton
Pilbara
Jack Hills
Clearly, theories and models have their uncertainties. Can we find evidence for PT in the rock record?
The problem there is: there is not so much left to take the evidence from. Continents are the only old pieces on Earth, so we have to look for evidence on continents.
Very few parts are old enough. And those available are often very much reworked
Those rocks are often very much reworked: metamorphosed, altered, and deformed

21. Linear features? Abitibi, Superior Province Pilbara, Australia
Seismic observations: dipping reflectors suggest remnants of fossil subduction zones?
(Calvert et al., 1995, JF Moyen, pers.comm.)

22. Oldest ophiolites Oldest ophiolite 3.7 Gyrs old?
Oldest generally accepted
ophiolites are ~2 Gyrs old
(Jormua, Finland; Purtuniq,
Canada)
Ophiolites become wide-spread
after 1.0 Gyrs ago
Oldest in Greenland, Isua. Not everyone agrees that this is a proper one: parts too distributed.\
Oldest accepted ones: 2 Gyrs, and these provide pretty solid evidence for oceanic lithosphere 2 Ga, and therefore PT, since it is hard to come up with how oceanic lithosphere would form without PT.
However, for ages < 1 Ga, the number of ophiolites is increasing significantly. Presenrvation problem, or different PT mechanism?
(Stern, 2005; Furnes et al., 2007)

23. Seismic observations Horizontal vs. vertical motion
Sub-horizontal dipping
reflectors suggest
fossil subduction?
Seismic observations: dipping reflectors suggest remnants of fossil subduction zones?
(Calvert et al., 1995)

24. Plate tectonics in Archaean?
Paleo-magnetism
Paleo-latitudes of old continents
varied over time
Only during supercontinent
(formation/breakup)
Data sparse!
Episodic early
plate tectonics?
Looking at magnetic record of rocks of different ages in same continent might tell us how this continent was travelling N-S. This travelling would be evidence for PT if different continents travel with different speeds.
Here are results for the few leftover old cratons.
They suggest rapid motion around when supercontinents existed.Only then PT? Episodic?
(O’Neill et al., 2007; Silver and Behn, 2008)

25. Subduction as site for crust formation
Bulk continental crust:
Today: andesites
Formed in subduction zone
Mantle wedge hydration and -melting
Archaean: tonalite-trondhjemite-granodiorite (TTGs)
(slab?) melting of mafic crust (Similarities with adakites?)
Interaction with a mantle wedge?
Suggested
formation
scenarios:
Continental crust might give us another clue. Today, andesites are thought to form primarily in subduction settings, so how far back in time can we trace this? Archaean continental crust (TTG), however, TTGs are not the same geochemically: this complicates things. BUT (abundant) TTG looks a lot like (rare) modern adakites: slab melts. Makes sense for hotter Earth. There are some problems with this model however: no MW influence (unlike for adakites)
Alternative model ‘base-of-lithosphere’ has its own problems: TTGs are thought to be ‘wet’ melting products, and the base of the lithosphere is most likely not wet at all. Also low T ( degC, high p (>12-15 kbar) conditions are more in line with a cold subduction geotherm than a normal continental geotherm.
(Defant and Drummond, 1993; Foley et al., 2002, 2003; van Thienen et al., 2004; Bédard, 2006)

26. Arc signature in Archaean TTGs?
rare-Earth
element (REE)
pattern
fluid
immobile
elements
(JF Moyen, pers. comm.)

27. “Arc” signature in Icelandic dacites !?!
How to explain this then? Well, it is thought that remelting from mafic lower crust and fractional crystallisation to upper felsic crust can produce this signature as well. This doesn’t mean that a slab melt scenario is not likely anymore (it still is the favourite model), but that subduction is not required to create this geochemical fingerprint, and that melting at the base of the crust is still a viable model.
(Willbord et al., 2009; JF Moyen, pers. comm.)

28. Key characteristics of plate tectonics
A compilation of these ‘hallmarks’ of PT is given by Bob Stern. Many typical PT features (passive margins, UHPM, ophiolites) are absent in early Prot. and Archaean. He takes this as evidence that PT didn’t start before late Proterozoic.
I’ll come back to this interpretation in a few slides.
No subduction in Precambrian?
(Stern, 2008)

29. Absence of UHPM by slab break-off?
Phanerozoic
Archaean
UHPM
no subduction
of continental
crust: absence
of UHPM
To explain: UHPM associated with cont.coll: continent dragged down  breakoff  rebound = UHPM
Perhaps Archaean continents not dragged down, and therefore no UHPM?
subduction of
continental crust
gives UHPM
(USGS website; Wortel and Spakman, 2000; van Hunen and Allen, 2010, subm.)

30. Long-term episodicity in subduction?
CC provides another interesting piece of information.
When CC is plotted in a histogram against age, it is clear that most of the crust is formed episodically, in peak periods, at 3.3, 2.7, 1.9, and 1.1 Gyrs.
To explain this, we probably have to look into the dynamics of the mantle over time, and this is the next topic I would like to go through briefly.
(McCulloch and Bennett, 1994; Davies, 1995; O’Neill et al., 2007)

31. Short-term episodicity in subduction?
Abitibi, Superior Province
Clastic
Calc-alkaline
Tholeitic
Time (Ma) 
(van Hunen & van den Berg, 2008; Moyen and van Hunen, in prep.)

32. Did subduction style change over time?
Evolution flat  steep subduction (Abbott et al., 1994)?
 No, because:
If too buoyant, slabs won’t subduct at all
A hot, weak mantle is unable to support flat subduction
A suggested change in PT style is flatsteep subduction. Initially based on buoyancy arguments: buoyant slabs don’t want to sink and will ‘float in mantle’.
However, if slab too buoyant to subduct, subduction and PT will stop altogether. Later used to explain geochemical characteristics in CC (next slide)
Furthermore, flat subduction (as observed today) requires some component of overriding continent motion (S.Am moves westward rapidly  flat subduction where buoyant plateaus subduct). In that case a flat slab occurs only when the slab is overridden faster than it is sinking in the mantle, and sinking rates are larger in a weak mantle. But an Archaean mantle was probably to weak to support those flat slabs.
(Abbott et al., 1994; van Hunen et al., 2004)

33. ‘Archaean water world’
Today: regassing > degassing
Early Earth:
More volcanism  more degassing
Hotter mantle  faster slab dehydration  less regassing?
Weaker rocks  no high mountain ranges?
Buoyant oceanic lithosphere  shallow ocean basins
Water is one of the volatiles: it enters ocean/atmosphere during MOR/OI volcanism, but can also re-enter mantle at subduction zones (most of water that enters subduction zone comes out again with arc volcanism, but some makes it to deep earth).
De/regassing rates quite unconstrained, but regassing seems more important than regassing today, so ocean shrinks? Observations suggest a very early (Hadean) ocean present.
In hotter Earth this was probably different, as slabs would dehydrate fasater during subduction, and less water made it to deep mantle.
(Wallmann, 2001; Rüpke et al., 2004; Rey & Coltice, 2009; Flament et al., 2008)

34. Concluding remarks Subduction evolution:
Changing dynamics due to changing mantle temperature:
Different crustal thickness, plate strength
Episodic subduction (long / short time scale)?
Observational evidence:
Geology/Geophysics: Ophiolites, dipping reflectors, palaeomagnetism
Geochemistry: Continental crust, geochemical fingerprints
Petrology: Ultra-high pressure metamorphism, blueschists

36. Computer model simulations
High Tmantle  thick crust
 no subduction? But …
Thick crust is only
sustainable if mantle
doesn’t ‘deplete’, i.e. if
crust mixes back in
after subduction
Perhaps heavy eclogite
settled at bottom of
mantle?
Other people did computer calculations that focus on other aspects.
Today, plates subduct and mix with mantle again. After long time of mixing, material homogeneous again, and can melt again.
But in past, mantle hotter, so weaker, and perhaps mixing of heavy eclogite not so easy. Maybe all eclogite sank to bottom of mantle, and stayed there?
Then mantle material becomes more depleted over time. This means it cannot melt as much anymore. Material can only melt once, and won’t melt again unless you enrich it again with the melt products (mixing). Then crust might have been thinner than today, and subduction wasn’t a problem in past.
(Davies, 2006)

37. Heat flow, cooling, and vsubd through time:
parameterized cooling model assumptions
today: qoc,0 =32 TW
qcont,0 =14 TW (Jaupart et al., 2007)
Present-day cooling rate (118 K/Gyrs, Jaupart et al., 2007)
extrapolated. Too high? Steady state reasonable?
only plate-tectonic cooling, no alternative mechanisms
(e.g. magma ocean, flood basalts)
Urey ratio (ratio of internal heating to surface heat flow, ~0.33)
uncertain
46 TW
From these results: peak T in early Archean
NB1: Present-day cooling rate (118 deg/Ga, Jaupart et al., 2007) extrapolated. Too high?
NB2: only PT cooling, no alternative mechanisms (magma ocean, flood basalts)
NB3: Urey ratio (ratio of internal heating to surface heat flow, ~0.33) uncertain

38. Subduction in Archaean?
Dipping seismic reflectors
Ophiolites (?)
These models can show insight, but are poorly constrained. What they do show is that for some models, PT is viable back to early Archaean, for others it isn’t.
Seismic dipping reflectors occur in many old cratons, and are mostly interpreted as fossil subduction zones.
Ophiolites are one of the main characteristics of today’s PT, and one of the few remnants of oceanic lithosphere older than 180 Ma. Phanerozoic ophiolites occur widespread, Proterozoic ones are often debated, and I think only one Archaean one is reported in Greenland, although the essential components of the ophiolite are not found as one continuous sequence, but rather a scattered set of components.
(Calvert et al., 1995; Furnes et al., 2007)

39. Ultra-high pressure metamorphism (UHPM)
Oceanic subduction
Continental collision, UHPM rocks form (gneisses, eclogites).
Slab break-off, and ‘rebound’
of continent, UHPM to surface
UHPM is typical for modern PT,
but is absent in rock record older
than 600 Myrs. Why?
No subduction.
Different slab temperature, so different UHPM rocks formed.
No continent subduction, so no UHPM rocks formed.
UHPM
Another indicator is UHPM: eclogites and gneisses that have been buried to large depth (>>100km), and somehow made it back to surface.
We think we understand how this works today:
- subduction brings eclogites down to large depth
- continental collision stops this process, because CC is too buoyant.
- slabs might break off, which causes the bit of CC that was dragged down to bounce back to surface (gneisses), pulling some eclogites with it.
Now UHPM doesn’t occur prior to 600 Ma. Why?
- no PT, no subduction
- different pT conditions, rocks might have looked different, and we can’t recognize them
- my favourite: weaker slabs couldn’t drag CC down as much as they can today, so CC never makes it to depths where UHPM is formed.

40. Different cooling scenarios
cooling of a
fluid
constant cooling
(after Sleep, 2000;
Turcotte & Schubert, 2002;
Korenaga, 2005)
surface heat flow
by plate tectonics
as today
Tm(t) strongly depends
on surface tectonics
What determines the cooling rate of the Earth:
Heatr production
Relase of heat through surface, i.e. vigour of PT.
Some scenarios:
Green: If Earth simply cooled the way it does now  green line. No physical basis for constant cooling, but it is a simple assumption.
Red=thermal catastrophe: has a fluid-dynamical basis: (explain: hotter mantle is weaker, convects faster, and therefore cools faster. That means a faster drop of Tm through time, so a steeper line when going back in time. This eventually leads to ‘thermal catastrophe’).
Blue=constant surface heatflow (explain: assume pl.tect. Always operated as it does today, but radiogenic hp was larger in past): has a physical basis.
So 3 totally different mantle cooling curves. We are looking for a curve like the green one (the one without a physical basis), but two first attempts show something totally different. Apparently, the physical processes responsible for the cooling are quite important. And apparently these models are too rough to determine Earth’s cooling history.
So either there’s something seriously wrong with at least one of these theories, or plate tectonics only operated recently, and we cannot use the ‘older’ parts of the blue/red curves.
So this is where my work will hopefully fit in: I modelled pl.tect. in hotter Earth.

41. Melting by episodic mantle avalanches?
Sudden warming of upper mantle may give:
Wide-spread (re-)melting  new continental crust
Unfavourable PT conditions: intermittent PT?
Illustrated how UM and LM T’s can evolve.
This would also have sign. Consequences for PT, as the larger amount of melting would give a thicker crust that suddenly doesn’t want to subduct anymore?
(Davies, 1995)

42. Cooling ability of flood basalts
Plumes transport heat from CMB and gives melting at surface
Diapir tectonics results in large-scale melting
Latent heat can be
important cooling
agent: enough to
cool Archaean
Earth if 100x more
vigorous than in
Phanerozoic
(van Thienen et al., 2005)

43. Did style of PT change over time?
Evolving style of PT?
TODAY
Unsubductable crust to form stacks (?)
Mechanism to form greenstone belts (?)
IN HOTTER EARTH
Also possible if crust was too buoyant to subduct, it might stack up at surface. This could explain presence of greenstone belts, but a problem is that if this stacking up occurred on large scale, we’d expect a lot more Archaean oceanic crust, which we don’t find.
(after Davies, 1992)

44. Alternative tectonics: diapir/delamination tectonics
Mechanism:
Crust built by eruptions
Deepest crust transforms to dense eclogites: delaminates
Downwellings  melting  TTG formation
Abundant melting releases latent
heat
One of the more popular models is that of diapir tectonics (or delamination tectonics).
(Zegers and van Keken, 2001; van Thienen et al., 2004, 2005)

45. Alternative tectonics: diapir/delamination tectonics
Why this model?
Explains ovoid extrusions (e.g. Pilbara)
No need for PT before late Archaean or Proterozoic
Efficient cooling mechanism
(Zegers and van Keken, 2001; van Thienen et al., 2004, 2005)

46. Archaean sea level and emerged continents
Constant continental freeboard (±200 m)
Very early ocean present
Continental growth model (?)
DT < K
unless orogenies
were weaker
2-3% late
Archaean
continent
emergence
‘Archaean water world’ 42 25
% continental area
Last slide:
Related to ocean volume is a useful observation that Archaean cratons have been around for Gyrs without significant erosion/sedimentation, suggesting roughly constant elevation of continents above sealevel (+-200 m), or constant continental freeboard.
Given changed amount of continental area, buoyancy of oceanic lithosphere, and all other potential changes that we saw before, this is quite remarkable.
E.g., hotter mantle gives thicker oceanic crust, that would be buoyant, and would sink less deep than today’s oceanic crust, and would lead to shallower oceans. If ocean volume was roughly the same, this means that part of that ocean water was flooding the continents.
Constant freeboard argues that that was not happening, and this limits early mantle T to at most K hotter than today, which is fairly low.
In plot (continental area vs Tm), gray area is constant freeboard area, purple area is likely Archaean situation.
The authors of this paper suggested that continents were quite weak, and therefore high mountains didn’t form, which relaxes freeboard constraint somewhat.
They argue that perhaps during Archaean only 2-3% of Earth’s surface was above sea level as continental area.
T(oC)
(Harrison et al., 2005; Flament et al., 2008)