Linking mantle plumes, large igneous provinces and
Stephan V. Sobolev
*, Alexander V. Sobolev
*, Dmitry V. Kuzmin
, Nadezhda A. Krivolutskaya
, Alexey G. Petrunin
Nicholas T. Arndt
, Viktor A. Radko
& Yuri R. Vasiliev
Large igneous provinces (LIPs) are known for their rapid produc-
tion of enormous volumes of magma (up to several million cubic
kilometres in less than a million years)
, for marked thinning of the
, often ending with a continental break-up, and for
their links to global environmental catastrophes
. Despite the
importance of LIPs, controversy surrounds even the basic idea that
they form through melting in the heads of thermal mantle
. The Permo-Triassic Siberian Traps
example and the largest continental LIP
—is located on thick
and was synchronous with the largest
known mass-extinction event
. However, there is no evidence of
pre-magmatic uplift or of a large lithospheric stretching
, as pre-
dicted above a plume head
. Moreover, estimates of magmatic
degassing from the Siberian Traps are considered insufficient
to trigger climatic crises
, leading to the hypothesis that the
release of thermogenic gases from the sediment pile caused the
. Here we present petrological evidence for a
large amount (15 wt%) of dense recycled oceanic crust in the head
of the plume and develop a thermomechanical model that predicts
no pre-magmatic uplift and requires no lithospheric extension.
The model implies extensive plume melting and heterogeneous
erosion of the thick cratonic lithosphere over the course of a few
hundred thousand years. The model suggests that massive degass-
ing of CO
and HCl, mostly from the recycled crust in the plume
head, could alone trigger a mass extinction and predicts it happen-
ing before the main volcanic phase, in agreement with stratigraphic
and geochronological data for the Siberian Traps and other LIPs
Petrological studies of Siberian Traps and associated alkaline rocks
reveal high temperatures (1,600–1,650 uC)
in their mantle sources.
Olivine compositions in samples from lower units of the Norilsk lava
section provide evidence that the mantle source of the Siberian Traps
wasunusually richin ancientrecycled oceanic crust
, in agreement with
. For the main volcanic phase, however, such data
were unavailable. Here we report 2,500 new olivine analyses and host-
rock compositions for 45 basalts covering the main stages of tholeiitic
magmatism in threekey localities:theNorilsk area,thePutorana plateau
and the Maymecha–Kotuy province (Fig. 1a). Almost all olivine com-
positions possess significantly higher NiO and FeO/MnO than expected
for olivine in peridotite-derived magmas (Fig. 1b, c and Supplementary
Fig. 1), suggesting a contribution of melts from pyroxenitic sources
Alternative explanations for these observations seem less plausible (see
Methods for discussion). Our interpretation of the olivine compositions
implies that the source of the Siberian Traps contained 10–20 wt%
recycled oceanic crust (Methods). More specifically, all lavas erupted
during the first stage of magmatic activity (Gudchikhinskaya and earlier
suites of the Norilsk area) are depleted in heavy rare-earth elements
indicating residual garnet and derivation within or below the base
of thick lithosphere (more than 130 km depth)
. The source of
Gudchikhinskaya lavas was probably almost entirely pyroxenitic
(Fig. 1b–d). Younger magmas are not depleted in heavy rare-earth
elements, indicating their formation at shallow depths and marked
thinning of the lithosphere. Our calculation suggests that these
magmas had a near-constant proportion of pyroxenite-derived melt
of about 50% (Fig. 1d and Supplementary Table 1) and were strongly
contaminated by the continental crust
. Because the main Norilsk
section spans less than 1 Myr (ref. 1), it is likely that the lithosphere
was thinned in only a few hundred thousand years.
High mantle temperatures over a vast area (Fig. 1a) are consistent
with the head of a hot mantle plume
. On the basis of the petrological
constraints we develop a thermomechanical model of the interaction of
the plume and lithosphere (see Methods). We assume that the plume
arrived below the lithosphere at about 253 Myr ago (model time 0),
perhaps near the northern border of the Siberian Shield, where the
hottest melts (meimechites) erupted
. We further assume that the
plume head was hot (T
5 1,600 uC; 250 uC excess temperature) and
containeda high content (15 wt%) of recycled oceanic crust. In our two-
dimensional model we approximate the plume head by a half-circle of
radius 400 km located below cratonic lithosphere of variable thickness
corresponding to the margin of the Archean craton (130–250 km of
depleted lithosphere and 160–250 km of thermal lithosphere; Fig. 2b
and Supplementary Fig. 2).
The arrival of a large and hot mantle plume head at the base of the
lithosphere has been predicted
to cause about 0.8–1 km of broad
surface uplift per 100 uC of plume excess temperature. For a purely
thermal plume with an excess temperature of 250 uC, we do indeed
obtain about 2.0 km of surface uplift (Fig. 2a, red curve). However, if a
large fraction (15 wt%) of dense recycled material is present within the
plume, its buoyancy is strongly decreased, resulting in little regional
uplift (250 m) (Fig. 2a, black curve). Other processes leading to surface
subsidence, such as the plume-induced rise of the 670-km phase
or the crystallization and evacuation of melts, may easily
counteract such a small uplift.
The plume head erodes the lowest part of the thermal lithosphere
and rapidly spreads below the more refractory depleted lithosphere
(Fig. 2b). Its ascent leads to progressive melting of recycled eclogitic
material in the plume and to the formation of reaction pyroxenite,
which melts at depths of 130–180 km, well before the peridotite
(Fig. 2e). The early, purely pyroxenite-derived, melts yielded the lavas
of the Gudchikhinskaya and earlier suites that display the ‘garnet
signature’ (Fig. 1d).
We propose that massive intrusion of the Gudchikhinskaya suite by
dykes imposed compressive stress in the upper brittle part of the litho-
sphere, ‘locking’it to the magmatransport (Fig. 2b, e shows the moment
of ‘locking’). After that, the melt could intrude only into the lower
Deutsches GeoForschungsZentrum GFZ, Telegrafenberg, 14473, Potsdam, Germany.
O.Yu. Schmidt Institute of the Physics of the Earth, Russian Academy of Sciences, 10 ul. B. Gruzinskaya, Moscow,
ISTerre, CNRS, University Joseph Fourier, Maison des Ge
osciences, 1381 rue de la Piscine, BP 53, 38041 Grenoble Cedex 9, France.
Max Planck Institute for Chemistry, 27 J.-J.-Becher-
Weg, Mainz, 55128, Germany.
V. I. Vernadsky Institute of Geochemistry and Analytical Chemistry, Russian Academy of Sciences, 19 ul. Kosygina, Moscow, 119991, Russia.
V. S. Sobolev Institute of
Geology and Mineralogy, Siberian Branch of Russian Academy of Sciences, 3 prosp. Akad. Koptyuga, Novosibirsk, 630090, Russia.
Limited Liability Company ‘Norilskgeologiya’ Norilsk, PO Box 889,
*These authors contributed equally to this work.
312 | NATURE | VOL 477 | 15 SEPTEMBER 2011
Macmillan Publishers Limited. All rights reserved
55 60 65 70
75 80 85
5000 1,000 km
N deep Fo>60
N shallow Fo<60
N shallow Fo>60
P shallow Fo>60
M shallow Fo>60
P deep Fo>60
P deep Fo<60
0.2 1 32
Petrological constraints. a, Geological map of the Siberian Traps
Dark green areas are lavas, light green areas are tuffs. The dashed black line
marks the border of the province. Red lines outline areas with different
magmatic activities: solid indicates maximal, dashed is moderate, and dotted is
minimal. The three studied regions are Norilsk (N), Putorana plateau (P) and
Maymecha–Kotuy province (M). White numbers stand for the potential
mantle temperature estimated for lavas of the corresponding areas
. b, FeO/
MnO ratios of olivine phenocrysts over normalized Gd/Yb ratios of host lavas.
The blue line marks the pressure that divides ‘deep’ lavas depleted in heavy rare-
earth elements from ‘shallow’ lavas. The green oval is the reference for the
almost pure shallow peridotitic mantle source and indicates the compositions
of olivine and lavas from the mid-ocean ridge (Knipovich Ridge, North
Atlantic) with minimum amounts of recycled ocean crust in their sources
olivines are the averages of the three highest Fo percentages of each sample. GA,
garnet in the mantle source. c, The proportions of pyroxenite-derived melt in
the mixture of pyroxenite-derived and peridotite-derived melts calculated
independently of Mn deficiency (X
Mn) and Ni excess (X
d, Integrated lava section for Siberian Traps based on the Norilsk section
(Supplementary Information). X
is the proportion of pyroxenite-derived
melt, calculated as the average of X
Mn and X
Ni for high-forsterite olivines
and as X
Mn for low-forsterite olivines, because X
Ni for the latter yields
systematic overestimation (Fig. 1c). Small black dots show lavas of the Norilsk
. For abbreviations indicating the lava suites of the Norilsk area and
normalization for Dy/Yb ratio, see Supplementary Information. Per, peridotite-
derived melt component.
0.29 Myr 0.32 Myr 0.39 Myr 0.42 Myr 0.52 Myr
Depth (km) Depth (km) H (km)
Model. a, Maximum pre-magmatic surface uplift (H) atop a
spreading mantle plume with an excess temperature of 250 uC. The red curve
corresponds to the purely thermal plume, and the black curve corresponds to a
thermo-chemical plume containing 15 wt% of recycled crust. b, c, Temperature
distributions (uC) in the model cross-section at model times of 0.15 Myr (b) and
0.5 Myr (c). The solid line marks the boundary of the depleted lithosphere, and
the dashed half-circle denotes the initial shape of the starting plume.
d, Snapshots of the plume breaking through the lithosphere in the domain
shown by the white rectangle in f. Colours show concentrations of the
pyroxenitic component in the plume or in the crystallized melt.
e, f, Distribution of the pyroxenite component in the plume (C
crystallized melt in the model cross-section at model times of 0.15 Myr (e) and
0.5 Myr (f). The solid line marks the boundary of the depleted lithosphere.
15 SEPTEMBER 2011 | VOL 477 | NATURE | 313
Macmillan Publishers Limited. All rights reserved
lithosphere (see Methods). The intruding melt cools and crystallizes to
dense eclogite. It also strongly heats and weakens the lithosphere, pro-
moting Raleigh–Taylor instabilities
. The lower part founders, and the
base of the lithosphere is mechanically eroded (Fig. 2c). Enriched in
eclogite, the lithospheric material in the boundary layer above the
plume escapes to the sides of the plume and then downwards, allowing
the plume to ascend (Fig. 2d, f). The plume breaks through the litho-
sphere in several zones, and in only 100–200 kyr reaches its minimum
depth of about 50 km (Fig. 2d and Supplementary Fig. 2). At this level,
mafic melts crystallize to a garnet-free assemblage and have a density
lower than that of the ambient mantle, thus preventing the formation of
Raleigh–Taylor instabilities. This mode of rapid lithosphere destruction
does not require regional stretching and matches observations for the
The extent of lithospheric destruction depends, among other factors
(Supplementary Information), on the initial density of mantle litho-
sphere, which is controlled by its composition (Fig. 3a and Supplemen-
tary Information). In the case of re-fertilized or moderately depleted
mantle lithosphere, the volume of the melt that intrudes into the crust
(melt crossing 50 km depth) reaches few per cent of the plume volume
(Fig. 3a), which leads to substantial melting of the crust and contam-
ination of basalts. Using the proportion of the magma-to-plume
volumes from a two-dimensional model for a three-dimensional
plume head with a radius of 400 km, we estimate the volume of the
magma intruded into the crust to be (6–8) 3 10
, which is realistic
for the Siberian Traps
In agreement with geochemical data, the model predicts that most
magma contains about 50% of pyroxenite-derived melt (Fig. 3b, blue
curves and symbols) and lack the ‘garnet signature’ because they are
generated at depths of less than 60 km. For the melt generated deeper
than 100 km, the model predicts a much higher proportion of pyrox-
enite melt (75–100%; Fig. 3b, red curves), again in agreement with
observations (Fig. 3b, red symbols).
Our model allows us to estimate the volume of CO
and HCl gases
released from the plume. For these calculations we consider separately
the recycled crust and peridotitic components by using data from melt
inclusions in olivine in Gudchikhinskaya picrites and mantle peridotite
as well as published estimates (Methods). For the composition of recycled
crust this yields HCl 5 137 p.p.m., S 5 135 p.p.m., H
O 5 800 p.p.m.
. 900 p.p.m. The model predicts that most of the CO
HCl in the recycled-crust component of the plume is extracted during
its interacti on with the lithosphere (Fig. 4a), and a major part is extracted
before the main phase of magmatism. For a three-dimensional plume
with a radius of 400 km, the mass of extracted CO
, which comes mostly
from the recycled component of the plume, is more than 170 3 10
exceeds the maximum estimate of the CO
released from the magmatic
heating of the coals from the Tunguska basin
Our prediction of the mass of CO
extracted from the plume is
consistent with the amount of CO
released during the Permo-
Triassic mass extinction estimated from Ca isotope data
Moreover, if we use a d
Cvalueof212% for pyroxenite-derived melt
as measured in Koolau (Hawaii) basaltic melt for the source dominated
by the recycled crust component
, we can alsoexplainthe
associated with the main mass-extinction event
from the plume alone may have triggered the main extinction
event. We speculate that low-density and low-viscosity volatiles were
the first to penetrate the compressed and mechanically locked crust,
triggering the extinction (Fig. 4a, upper axis). Alternatively, a sufficient
quantity of gases may have been released before lithospheric locking,
together with the deep-sited magmas of pre-Gudchikhinskaya suites,
which could also produce metamorphic gases by magmatic heating of
the coals and carbonates. This alternative is supported by the recent
discovery of coal fly ash in Permian rocks from the Canadian High
Arctic immediately before the mass extinction, interpreted as a result
of combustion of Siberian coal and organic-rich sediments by flood
. In either case, according to our model, the major mass
extinction happened before the main phase of flood basalt extrusion.
In contrast, most of the CO
and other gases released by contact meta-
morphism of carbon-rich and sulphur-rich sediments, which have been
suggested as a trigger for the mass extinction
, would be released
during the main phase of magmatism. Precise U–Pb dating of
Siberian magmatic units and the Permo-Triassic boundary is required
to distinguish between the two hypotheses. Nonetheless, existing geo-
and the presence of abundant pyroclastic rocks
underlying lavas of the main magmatic phase
support the idea that
the major mass extinction predated the main phase of magmatism (see
Fig. 4a, upper axis). Additional large amounts of gases released from
may have been the cause of
C excursions during
the later phases of the biotic crisis
According to our data and model, the plume also generates a sur-
prisingly large amount of HCl (about 18 3 10
tonnes; Fig. 4a), mostly
0.4 0.8 1.2 1.6 2.0
Re-fertilized (Δρ = 25 kg m
0.4 0.6 1.0
Depleted (Δρ = 50 kg m
Fraction of volume
Fraction of pyroxenite melt
Model predictions. a, Evolution in time of a melt volume crossing
the 50-km depth and normalized to the volume of the plume. The solid and
dashed curves correspond to the models with re-fertilized lithosphere and
moderately depleted lithosphere, respectively. b, Plot of the fraction of
pyroxenitic component in basalts of the Norilsk cross-section against the
fraction of the volume of extruded magmas. The blue colour corresponds to the
‘shallow’ melts that do not retain a garnet signature; the red colour corresponds
to the deep melts that retain a garnet signature. Symbols denote data from
olivine compositions; see Fig. 1b for details. Error bars correspond to 1 standard
deviation of the mean of pyroxenite-derived melt proportions estimated
independently from Ni excess and Mn deficiency of olivine (Methods and
Supplementary Table 1). The solid and dashed curves show the modelled
average melt compositions with re-fertilized and moderately depleted
lithosphere, respectively. The grey rectangle shows the range of variation of the
melt compositions predicted by the model.
314 | NATURE | VOL 477 | 15 SEPTEMBER 2011
Macmillan Publishers Limited. All rights reserved
also derived from the recycled component. This quantity of toxic HCl
must have been extremely damaging for the terrestrial species and was
also sufficient to trigger deadly instability of the stratospheric ozone
By accepting our viewpoint that degassing of the plume, rather than
thermogenic gases from sediments, triggered the biotic crises, we lose
an elegant explanation of why the Siberian LIP was so much more
damaging to biota than other LIPs of comparable size (Karoo, Parana
and North Atlantic) that extruded through other types of sediment or
. An alternative explanation is based on the correlation
of the intensity of mass extinctions with the age of Phanerozoic LIPs
(Fig. 4b), a relationship that can be explained by the temporally dif-
ferent response of the ocean to acidification by the large amounts of
(refs 23, 30). In contrast to the pre-Mid-Mesozoic
‘Neritan’ ocean, the more recent ‘Cretan’ ocean was buffered against
acidification by deep-sea unlithified carbonate sediments and was thus
much more resistant to acidification
. Therefore, CO
degassing of a
pre-Mid-Mesozoic LIP caused much more severe ocean acidification
and mass extinction than later LIPs (Fig. 4b). The only exception is the
Deccan LIP and the contemporaneous mass extinction at 65.5 Myr
ago; however, in this case the Chicxulub impact was an additional
Numerical tests (Supplementary Figs 4–6) suggest that rapid litho-
spheric destruction associat ed with melting in the heads of thermoche-
mical plumes is valid for the large range of plume parameters and
lithospheric thicknesses, and therefore may apply not only to the
Siberian Traps but also to other LIPs. An absence of prominent pre-
magmatic uplift does not argue against a plume origin of LIPs, but may
instead point to a high content of recycled crust within the plume. In such
cases, other parameters being equal, the model predicts that eclogite-rich
plumes caused the most extensive delamination and thinning of the
lithosphere, thus best preparing it for a possible break-up. They also
produced the strongest volcanism and led to the most marked climatic
Another suggestion of our model—that major mass extinctions are
triggered by degassing of plume magmas that predate the main magmatic
phase—also seems to be consistent with the observations for many LIPs
implying that gas output from plume heads may be much larger than
We report new data on 45 representative olivine-bearing samples of Siberian flood
basalts from the Norilsk, Putorana and Maimecha–Kotui regions (Supplementary
Tables 1 and 3). The bulk rocks were crushed, melted and analysed for major and
trace elements with an electron probe microanalyser (EPMA) and by laser ablation
inductively coupled plasma mass spectrometry (LA-ICP-MS) at the Max Planck
Institute for Chemistry in Mainz, Germany. Olivine phenocrysts (about 2,500
analyses) were analysed by EPMA with a special high-precision protocol
Max Planck Institute for Chemistry. Using this new information and published
approaches we estimated the amount of recycled oceanic crust in the sources of
basalts and their potential temperatures, and discuss possible alternative models of
the source compositions. We further estimate the amounts of H
O, Cl, S and CO
in the recycled oceanic crust and develop a model of its degassing during plume–
We model the thermomechanical interaction of the plume and lithosphere by
numerically solving a coupled system of momentum, mass and energy conser-
vation equations in two dimensions. We employ nonlinear temperature and
stress-dependent elasto-visco-plastic rheology, consider pressure-dependent and
temperature-dependent melting of a heterogeneous mantle and employ simple
models of melt transfer and extraction of volatiles.
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 17 February; accepted 26 July 2011.
1. Reichow, M. K. et al. The timing and extent of the eruption of the Siberian Traps
large igneous province: implications for the end-Permian environmental crisis.
Earth Planet. Sci. Lett. 277, 9–20 (2009).
2. White, R. & McKenzie, D. Magmatism at rift zones—the generation of volcanic
continental margins and flood basalts. J. Geophys. Res. Solid Earth Planets 94,
3. Garfunkel, Z. Formation of continental flood volcanism—the perspective of setting
of melting. Lithos 100, 49–65 (2008).
4. Courtillot, V. E. & Renne, P. R. On the ages of flood basalt events. C. R. Geosci. 335,
5. Wignall, P. B. Large igneous provinces and mass extinctions. Earth Sci. Rev. 53,
6. Campbell, I. H. & Griffiths, R. W. Implications of mantle plume structure for the
evolution of flood basalts. Earth Planet. Sci. Lett. 99, 79–93 (1990).
250 200 150 100
Magma volume crossing
50 km depth (Fig. 3a)
Ref. 15 Ref. 26
and metamorphic gas
Deep gas and
Generic extinctions (%)
‘Cretan’ ocean mode
Production of volatilesand its consequences for mass extinctions.
(left axis) and HCl (right axis) amounts extracted from the
plume against model time (lower axis). Solid curves show the minimum estimate
and dashed curves the maximum estimate of CO
and HCl extracted from the
plume (Methods). The grey rectangle shows the estimated range of the released
during the Permo-Triassic mass extinction
. The green area shows tim e
dependence of the nor malized volume of the magma crossing the 50-km depth,
calculated for the re-fertilized lithosph ere (Fig. 3a). On the top axis we show
geological time and a possible model for triggering the Permo-Triassic mass
extinction. GBT, gases break through. Also shown is U–Pb dating of the extinction
and U–Pb dating of main-phase Siberian basalts
of mass extinction intensity (light blue field) with major LIPs (circles) against
geological time (modified from ref. 33), together with the timing of different ocean
. Circle colours denote the timing of LIPs relative to ocean modes: blue,
‘Cretan’ mode; red ‘Neritan’ mode; blue and red together, transition mode. The
scale of circle sizes is in millions of cubic kilometres. CAMP, Central Atlantic
Magmatic Province; NAMP, Northern Atlantic Magmatic Provinces, OJP,
Ontong Java; CP, Caribbean Plateaux; CR, Columbian River basalts.
15 SEPTEMBER 2011 | VOL 477 | NATURE | 315
Macmillan Publishers Limited. All rights reserved
7. Czamanske, G. K., Gurevitch, A. B., Fedorenko, V. & Simonov, O. Demise of the
Siberian plume: paleogeographic and paleotectonic reconstruction from the
prevolcanic and volcanic record, north-central Siberia. Int. Geol. Rev. 40, 95–115
8. Elkins-Tanton, L. T. & Hager, B. H. Melt intrusion as a trigger for lithospheric
foundering and the eruption of the Siberian flood basalts. Geophys. Res. Lett. 27,
9. Richards, M. A., Duncan, R. A. & Courtillot, V. E. Flood basalts and hot-spot tracks—
plume heads and tails. Science 246, 103–107 (1989).
10. Cordery, M. J., Davies, G. F. & Campbell, I. H. Genesis of flood basalts from eclogite-
bearing mantle plumes. J. Geophys. Res. Solid Earth 102, 20179–20197 (1997).
11. Sobolev, V. S. Petrology of Siberian Traps (Transactions of All-Union Arctic Institute,
vol. 43) (Glavnogo Upravleniaya Sevmorputi, 1936).
12. Dobretsov, N.L., Kirdyashkin, A. A., Kirdyashkin, A. G., Vernikovsky,V. A. & Gladkov, I.
N. Modelling of thermochemical plumes and implications for the origin of the
Siberian traps. Lithos 100, 66–92 (2008).
13. Self, S., Widdowson, M., Thordarson,T. & Jay, A. E. Volatile fluxes during flood basalt
eruptions and potential effects on the global environment: a Deccan perspective.
Earth Planet. Sci. Lett. 248, 518–532 (2006).
14. Sobolev, A. V., Krivolutskaya, N. A. & Kuzmin, D. V. Petrology of the parental melts
and mantle sources of Siberian trap magmatism. Petrology 17, 253–286 (2009).
15. Svensen, H. et al. Siberian gas venting and the end-Permian environmental crisis.
Earth Planet. Sci. Lett. 277, 490–500 (2009).
16. Ganino, C. & Arndt, N. T. Climate changes caused by degassing of sediments
during the emplacementof large igneous provinces. Geology 37, 323–326 (2009).
17. Sobolev, A. V., Sobolev, S. V., Kuzmin, D. V., Malitch, K. N. & Petrunin, A. G. Siberian
meimechites: origin and relation to flood basalts and kimberlites. Russ. Geol.
Geophys. 50, 999–1033 (2009).
18. Sobolev, A. V. et al. The amount of recycled crust in sources of mantle-derived
melts. Science 316, 412–417 (2007).
19. Fedorenko, V. A. et al. Petrogenesis of the flood-basalt sequence at Noril’sk, North
Central Siberia. Int. Geol. Rev. 38, 99–135 (1996).
20. Wooden, J. L. et al. Isotopic and trace-element constraints on mantle and crustal
contributions to Siberian continental flood basalts, Norilsk Area, Siberia. Geochim.
Cosmochim. Acta 57, 3677–3704 (1993).
21. Farnetani, C. G. & Richards, M. A. Numerical investigations of the mantle plume
initiation model for flood-basalt events. J. Geophys. Res. Solid Earth 99,
22. Leng, W. & Zhong, S. J. Surface subsidence caused by mantle plumes and volcanic
loading in large igneous provinces. Earth Planet. Sci. Lett. 291, 207–214 (2010).
23. Payne, J. L. et al. Calcium isotope constraints on the end-Permian mass extinction.
Proc. Natl Acad. Sci. USA 107, 8543–8548 (2010).
24. Hauri, E. SIMS analysis of volatiles in silicate glasses. 2. Isotopes and abundances
in Hawaiian melt inclusions.
Chem. Geol. 183, 115–141 (2002).
25. Grasby, S. E., Sanei, H. & Beauchamp, B. Catastrophic dispersion of coal fly ash into
oceans during the latest Permian extinction. Nature Geosci. 4, 104–107 (2011).
26. Kamo, S. L. et al. Rapid eruption of Siberian flood-volcanic rocks and evidence for
coincidence with the Permian–Triassic boundary and mass extinction at 251 Ma.
Earth Planet. Sci. Lett. 214, 75–91 (2003).
27. Mundil, R., Palfy, J., Renne, P. R. & Black, P. The Triassic timescale: new constraints
and a review of geochronological data. Geol. Soc. Lond. Spec. Pub. 334, 41–60
28. Payne, J. L. & Kump, L. R. Evidence for recurrent Early Triassic massive volcanism
from quantitative interpretation of carbon isotope fluctuations. Earth Planet. Sci.
Lett. 256, 264–277 (2007).
29. Beerling, D. J., Harfoot, M., Lomax, B. & Pyle, J. A. The stability of the stratospheric
ozone layer during the end-Permian eruption of the Siberian Traps. Phil. Trans. R.
Soc. Lond. A 365, 1843–1866 (2007).
30. Ridgwell, A. A mid Mesozoic revolution in the regulation of ocean chemistry. Mar.
Geol. 217, 339–357 (2005).
31. Schulte, P. et al. The Chicxulub asteroid impact and mass extinction at the
Cretaceous–Paleogene Boundary. Science 327, 1214–1218 (2010).
32. Masaitis, V. L. Permian and Triassic volcanism of Siberia: problems of dynamic
reconstructions [in Russian]. Zapiski Vsesouznogo Mineralogicheskogo Obshestva
112, 412–425 (1983).
33. White, R. V. & Saunders, A. D. Volcanism, impact and mass extinctions: incredible
or credible coincidences? Lithos 79, 299–316 (2005).
Supplementary Information is linked to the online version of the paper at
Acknowledgements S.V.S. and A.V.S. are especially grateful to Vladimir Stepanovich
Sobolev, who excited their interest in the origin of the Siberian Traps. We thank
G. A. Fedorenko for providing data on the Norilsk lavas and for discussions;
N. Groschopf for help in managing the electron probe microanalyser; O. Kuzmina,
N. Svirskaya and T. Shlichkova for sample preparation; P. Cardin, N. Dobretsov,
E. Galimov, C. Herzberg, A. Hofmann, L. Kogarko, H.-C. Nataf, J. Payne, Y. Podladchikov,
I. Ryabchikov, A. Turchyn and G. Wo
rner for discussions; and P. Kelemenfor comments.
S.V.S. thanks the Deutsche Forschungsgemeinschaft (DFG) SPP 1375 SAMPLE (SO
425/4) for support. The study by A.V.S. was funded by the Agence Nationale de la
Recherche, France (Chair of Excellence Grant ANR-09-CEXC-003-01) and partly
supported by a Gauss Professorship in Go
ttingen University, Germany, the Russian
Foundation for Basic Research (09-05-01193a), a Russian President grant for leading
Russian scientific schools (
-3919.2010.5) and an Earth Sciences Department of
Russian Academy Grants.
Author Contributions S.V.S. and A.V.S. provided major contributions to
thermomechanical (S.V.S.) and petrological (A.V.S.) modelling, to the interpretation of
data and to the writing of the paper. N.A.K. provided geological background and
contributed to interpretation. A.G.P. contributed to the thermomechanical modelling at
an initial stage. N.T.A. contributed to interpretation and writing of the paper. D.V.K.
processed samples and performed the measurements. N.A.K., V.A.R. and Y.R.V.
provided carefully selected samples. All authors contributed intellectually to the paper.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of this article at
www.nature.com/nature. Correspondence and requests for materials should be
addressed to S.V.S. (email@example.com) or A.V.S.
316 | NATURE | VOL 477 | 15 SEPTEMBER 2011
Macmillan Publishers Limited. All rights reserved
Samples. The samples studied and their localities are described in Supplementary
Analytical methods. Olivine grains were manually separated from crushed lavas,
then mounted and polished in epoxy. The compositions of olivine were analysed
with an EPMA on a Jeol JXA 8200 SuperProbe at the Max Planck Institute for
Chemistry (Mainz, Germany) at an accelerating voltage of 20 kV and a beam
current of 300 nA, following a special procedure which allows 20–30 p.p.m. (2s
error) precision and accuracy for Ni, Ca, Mn, Al, Ti, Cr and Co, and 0.02 mol%
accuracy for the forsterite component in olivine
Bulk rocks were crushed, melted to glass
and then mounted and polished in
epoxy. Major and trace elements were also determined by EPMA at the Max
Planck Institute for Chemistry. Major-element abundances in glasses were mea-
sured at an accelerating voltage of 15 kV and a beam current of 12 nA with a
reference sample of natural basaltic glass USNM111240/52 (VG2)
, with a relative
error of 1–2%. LA-ICP MS was used to determine trace elements in glasses of melt
inclusions and in olivines, on an ELEMENT-2, Thermo Scientific mass spectro-
meter with a UP-213 New Wave Research solid-phase laser at the Max Planck
Institute for Chemistry, with reference to the KL-2G and NIST 612 standard
samples of basaltic glass
Proportions of pyroxenite-derived melt and recycled crust. We interpret excess-
ive Ni and deficient Mn concentrations in Siberian olivine phenocrysts relative to
olivine in peridotite-derived melt as a result of the contribution of olivine free
pyroxenite lithology in their source
. The alternative explanations of this phe-
nomenon are discussed briefly in the next section. Relative proportions of pyroxenite
and peridotite derived melts were estimated from MnO/FeO (X
NiO 3 FeO/MgO (X
for each sample, using the average composition
of most magnesian olivines (defined by olivines with Fo within 3 mol% from a
The amount of recycled crust in the plume is linked to the proportion of
pyroxenite-derived melt by the degree of melting of the eclogite component, the
amount of eclogite-derived melt needed to producehybridpyroxenitefromperidotite,
and the degrees of melting of peridotite and pyroxenite
. We calculated the
amount of recycled crust in the Siberian plume using the approach described
and their equation S3, and the following assumptions: a maximum
degree of melting of eclogite and pyroxenite of 60%; an average proportion of
pyroxenite-derived melt in shallow magmas of 46% (Supplementary Table 1);
and melting of peridotite at 50 km depth. The amounts of recycled crust are
10% and 20% for 10% and 25% melting of peridotite, respectively.
Alternative explanations of the unusual olivine composition. Alternative
explanations of high Ni/Mg and FeO/MnO ratios in olivine include: (1) effect of
clinopyroxene crystallization, (2) an underestimated temperature effect on olivine-
melt partition of Ni, and (3) contribution of core material to the mantle source.
None of these alternatives require a significant role of olivine-free pyroxenite in the
mantle source. In addition there are different models for the pyroxenite origin,
which may affect our estimation of proportions of pyroxenite in the Siberian plume.
These include (4) solid-state reaction of peridotite and recycled crust in the lower
mantle and (5) partial reaction between eclogite-derived melt and peridotite. Below
we briefly discuss these alternatives.
1. Crystallization of clinopyroxene together with olivine may increase both Ni/Mg
and Fe/Mn ratios in olivine. This effect could be particularly important for the low
magnesian evolved olivine.Howeverclinopyroxene crystallizationcanbe recognized
situation for most studied olivines, which do not show a significant decrease in Ca
concentrations. In particular, early clinopyroxene crystallization cannot explain the
composition of the olivines from picrites of the Gudchikhinskaya formation and the
Ayan river, which are extremely rich in Ni and deficient in Mn. In addition, melt
inclusions in the former
and olivines in the latter (Supplementary Table 3) do not
indicate early clinopyroxene crystallization. We conclude that the fractionation of
clinopyroxene could not have produced the observed anomalies in the olivine
2. An underestimated temperature effect on olivine-melt partition coefficient, if
present, may increase Ni concentration in the shallow olivine compared with
olivine in the deep source as a result of temperature difference between the sites
of generation and crystallization. This issue has been discussed previously
where it was shown that any temperature effect additional to the compositional
one considered in the model of Ni partitioning between melt and olivine used in
is too small to explain the extent of Ni excess observed in Siberian
3. The contribution of core material to increase Ni/Mg (ref. 41) and Fe/Mn
(ref. 42) ratios has been discussed previously
, where it was shown that this
explanation is highly unlikely because of a lack of correlation between Ni excess
and high Co concentrations in the olivines.
4. The solid-state reaction between recycled crust and per idotite in the lower
mantle may produce pyroxenite with a composition much closer to that of peridotite,
which in the upper mantle will be transformed to olivine-bearing pyroxenite
. If this
lithology exists, it could potentially be the source of parental melts of typical Siberian
basalts. However, olivine-bearing lithologies could not be the source of deep-sourced
lavas such as the Gudchikhinskaya formation and Ayan river picrites, whose olivine
compositions demonstrate derivation from a dominating olivine-free source. In
addition, solid-state reactions of the type envisaged in ref. 37 will be limited by slow
volume diffusion in the mantle
, whereas the production of reaction olivine-free
pyroxenite will be restricted only by melt percolation velocity
5. Incomplete reaction between eclogite-derived melt and peridotite may pro-
duce olivine-bearing lithologies
, which could be potential sources of the parental
melts of typical Siberian traps. However, these olivine-bearing lithologies cannot
produce Siberian deep-sourced magmas (see above).
We conclude that although we cannot fully exclude some of proposed alter-
natives, our explanation of the olivine compositions is based on solid grounds and
seems the most plausible.
Potential temperatures. The published potential temperature of 1,540 uCforthe
source of Gudchikhinskaya magmas
has been corrected for the effect of 40% melting
of pyroxenitic source
, the amount of pyroxenite in the plume (15%) and the latent
heat of melting
. The value obtained is about 1,600 uC. Potential temperatures for the
and the Putorana plateau
(see Fig. 1a) were obtained
formagmasstronglyenrichedinhighlyincompatible elements. These magmas origi-
nated at low degrees of melting and thus were not corrected for the melting effect.
Amount of volatile elements. The concentrations of volatile elements in the
recycled oceanic crust in the Siberian plume were constrained using the composi-
tions of inclusions of uncontaminated melt in early olivine phenocrysts from
Gudchikhinskaya magmas. For these magmas it was shown from both olivine
and melt compositions that they probably represent the melting of a pure
. Inclusions in olivine from these magmas have been shown
to represent primary melts
and thus their concentrations of Cl, S and H
used to estimate the contents of these volatiles in the mantle source. For the com-
position of source eclogite, this yields the following values
:Cl5 137 p.p.m.,
S 5 135 p.p.m. and H
O 5 800 p.p.m., after normalization of the values to K. In
the deep mantle, these amounts of Cl, S and H
O could reside in chloride
sulphide, and garnet or pyroxene respectively. In contrast with Cl, S and H
the amount of CO
in relatively shallow melts does not represent the primary
concentration, because of almost complete degassing at high pressures. Thus for
an assessment of CO
in the recycled oceanic crust we use global estimations of
3,000 p.p.m. CO
for the bulk 7-km-thick oceanic crust and its maximum
outgassing rate through arc volcanism of 70% (ref. 48). This gives a conservative
minimum estimate of 900 p.p.m. CO
in the deeply recycled oceanic crust. The
maximum estimate would be about 1,800 p.p.m. using the same initial bulk con-
centrations of CO
and minimal outgassing of 40% (ref. 48). In the deep mantle this
amount of CO
could reside in carbonates or diamond
. For our model we use the
minimum conservative estimate of CO
5 900 p.p.m.
Thermo-mechanical numerical technique. We use a fully coupled thermo-
mechanical formulation for the system of momentum, mass and energy con-
servation equations in two dimensions with nonlinear temperature and
stress-dependent elasto-visco-plastic rheology, described in detail in ref. 49 (for para-
meters and procedure for calculating density, see Supplementary Information).
Equations are solved numerically using the explicit Lagrangean FEM technique
LAPEX2D (ref. 50) based on a FLAC algorithm (prototype described in ref. 51)
combined with a particle-in-cell approach. All time-dependent fields including
full stress tensor are stored at particles.
Melting models. We use a simplified model for batch melting of four components:
peridotite, pyroxenite and two eclogites formed through the crystallization of
peridotitic and pyroxenitic melts, respectively. Melting temperatures are defined
as follows: for peridotite we use the dry batch melting model
; for pyroxenite we
use the following relation from experiments
for dry batch melting of pyroxenite:
where P is pressure in kbars, T is potential temperature in uC and X is degree of
melting, 0 , X , 0.55.
For eclogite of both types (peridotite-derived or pyroxenite-derived) we use the
following relations approximating experiments
for dry batch melting of eclogite
at 50% melting:
at P , 55 kbar, and
at 55 ( P , 80 kbar.
Macmillan Publishers Limited. All rights reserved
The melting for each finite element is organized sequentially, beginning with the
component with the lowest melting temperature (usually eclogite), then usually
pyroxenite and finally peridotite. If the current temperature (T) in a finite element
exceeds the melting temperature (T
) of the component that exists in this element,
than a certain amount of melt (dC
) is generated that lowers the current temper-
ature to the melting temperature, dC
5 (T 2 T
T), where C
capacity and DS
is entropy of melting, set to 1,200 J kg
and 400 J kg
respectively, for all components.
Model for melt transfer. We assume that melt transfer within the melting domain
occurs much faster than the Raleigh–Taylor instability develops in the lithosphere.
This assumption holds if the velocity of melt transfer (V
) is much higher than
H/t, where H is the typical distance of melt transfer and t is typical time of
Raleigh–Taylor instability. With values of H , 50 km and t < 50,000 years (see
Supplementary Fig. 2), this assumption is valid if V
is much higher than 1 m yr
In reality, V
is at least tenfold higher in the upper mantle regions where intensive
If present, the entire melt is assumed to move rapidly upwards within the
domain where the local temperature is higher than the melting temperature. As
usual for melt porous flow, we also assume that it is thermally equilibrating at each
element. In practice, at every nth calculation time-step, for every element we check
the melting condition and move the entire volume of melt (if present) one element
upwards, recalculating the temperature of that element according to the local
energy conservation law.
For the melt in the uppermost elements of the melting domain, we consider two
transfer modes. First we consider a mode that mimics transfer through fractures:
in this case, a fraction or the entire melt is assumed to move to the surface; in
practice it is just taken out from the model. Second, we consider a mode of
mechanically locked lithosphere: in this case the entire melt from each uppermost
element of the melting zone is moved to, and evenly distributed between, K ele-
ments in the column just above it (usually we take K 5 4) and is assumed to
crystallize. Simultaneously, the rheology of melt-accepting lithospheric elements
is switched from ‘dry’ to ‘wet’ olivine rheology if the crystallized melt content
exceeds some critical value (we take this value as 1%). The temperature and
composition of the accepting elements are recalculated according to the local
energy and mass conservation laws.
Extraction of volatiles. We consider two endmember models for the extraction of
volatiles. In the first model we assume that CO
and HCl are fully extracted from
the plume if the temperature approaches that of the carbonatite solidus
model gives an upper bound for melt mobility,assumingthat melts produced by an
infinitely low degree of melting can move out of the plume. In the second model we
assume that CO
and HCl are fully extracted from both peridotitic and pyroxenitic
components only if 1% melting is achieved. This model gives lower boundary for
melt mobility, assuming that only 1% carbonate–silicate melts can move out of the
plume. For both models we assume a concentration of HCl in recycled crust of
137 p.p.m. derived from melt inclusions in olivine, no HCl in the peridotitic
component and minimum conservative estimates of CO
content in both recycled
crust (900 p.p.m.) and plume peridotite (70 p.p.m.) (see above).
Expected mode of motion of volatiles in the lithosphere. Melt from the plume is
trapped and crystallizes in the lithosphere; it then returns to the mantle as the
lithosphere founders. The volatiles extracted by melting of the plume are released
as the melt crystallizes because host phases are not stable at the high temperatures
at the base of the lithosphere. They migrate upwards, then react and are fixed in
carbonates or chlorides in the cooler upper part of the lithosphere. Continuous
upward migration of high-temperature isotherms then decomposes these phases,
promoting further displacement of the volatile front ahead of the basalt-melting
front. According to our model, more than 70% of the mafic magmas generated in
the plume crystallized to eclogites and subsided back into the mantle as the
densified lithosphere foundered; less than 30% were intruded into the crust.
However, almost all carbonatite melts traversed the lithosphere and crust, because
the temperature of the detached blocks was significantly higher than the melting
temperature of carbonatite. It is therefore likely that a significant part of the
volatiles released from the plume finally reached the surface, promoting explosive
eruptions, which are very common in the early stage of Siberian Traps
other LIPs, that is, Emeishan flood basalts
. Note that the volatiles that were
initially stored in the minerals of the destructed portions of the lithosphere should
be also melted out and could finally reach the surface as well.
Why output of volatiles from LIPs could be drastically underestimated. The
amount of volatiles released was previously estimated using only the volume of
extruded magmas or magmas intruded into the shallow crust. In these studies were
disregarded the magmas that crystallized in the deep crust and, more importantly,
the much larger volumes of magmas that we propose were involved in the destruc-
tion of the lithosphere and never reached the crust, although most of the volatiles
extracted from them probably did (see above). Additionally, the recycled crust
component of the plume, which contains much more volatiles than the peridotitic
component, was not previously considered in the balance calculation.
C excursion. The
C-isotope change due to the released CO
(considered to be instantaneous) can be estimated from a simple mass-balance
. According to our model, about 172 3 10
tonnes of CO
from the plume—about 70% from recycled crust and the remaining 30% from
peridotite. Assuming that d
C 5212% for crust-derived CO
(ref. 24) and
C 525% for peridotite-derived CO
, we obtain an average isotopic composi-
tion of the plume-released carbon of d
C 529.9%. Assuming d
C 5 3.6% for
the initial carbon isotope composition and 300 3 10
tonnes of CO
(or 82 3 10
tonnes of C)
for the Late Permian CO
reservoir, we estimate that the magnitude
of the carbon isotope excursion was 4.9% if all plume-released gases migrated to
the surface, and 3.5% if only half of them arrived. Both numbers are well within
the range of reported values for the Permian–Triassic excursion
34. Stoll, B. et al. An automated iridium-strip heater for LA-ICP-MS bulk analysis of
geological samples. Geostand. Geoanal. Res. 32, 5–26 (2008).
35. Jarosevich, E. J., Nelen, J. A. & Norberg, J. A. Reference sample fro electron
microprobe analysis. Geostand. Newsl. 4, 43–47 (1980).
36. Jochum, K. P. et al. The preparation and preliminary characterisation of eight
geological MPI-DING reference glasses for in-site microanalysis. Geostand. Newsl.
24, 87–133 (2000).
37. Herzberg, C. Identification of source lithology in the Hawaiian and Canary Islands:
implications for origins. J. Petrol. 52, 113–146 (2011).
38. Sobolev, A. V., Hofmann, A. W., Brugmann, G., Batanova, V. G. & Kuzmin, D. V. A
quantitative link between recycling and osmium isotopes. Science 321, 536
39. Sobolev, A. V., Hofmann, A. W., Sobolev, S. V. & Nikogosian, I. K. An olivine-free
mantle source of Hawaiian shield basalts. Nature 434, 590–597 (2005).
40. Beattie, P., Ford, C. & Russell, D. Partition coefficients for olivine-melt and
orthopyroxene-melt systems. Contrib. Mineral. Petrol. 109, 212–224 (1991).
41. Ryabchikov, I. D. High NiO content in mantle-derived magmas as evidence for
material transfer from the Earth’s core. Dokl. Earth Sci. 389, 437–439 (2003).
42. Humayun, M., Qin, L. P. & Norman, M. D. Geochemical evidence for excess iron in
the mantle beneath Hawaii. Science 306, 91–94 (2004).
43. Holzapfel, C., Chakraborty, S., Rubie, D. D. & Frost, D. J. Effect of pressure on Fe-Mg,
Ni and Mn diffusion in (Fe
olivine. Phys. Earth Planet. Inter. 162,
44. Kelemen, P. B., Hart, S. R. & Bernstein, S. Silica enrichment in the continental upper
mantle via melt/rock reaction. Earth Planet. Sci. Lett. 164, 387–406 (1998).
45. Manglik, A. & Christensen, U. R. Effect of lithospheric root on decompression
melting in plume–lithosphere interaction models. Geophys. J. Int. 164, 259–270
46. Ryabchikov, I. D., Solovova, I. P., Ntaflos, T., Buchl, A. & Tikhonenkov, P. I.
Subalkaline picrobasalts and plateau basalts from the Putorana plateau (Siberian
continental flood basalt province). II. Melt inclusion chemistry, composition of
‘primary’ magmas and P–T regime at the base of the superplume. Geochem. Int.
39, 432–446 (2001).
47. Safonov, O. G., Kamenetsky, V. S. & Perchuk, L. L. Links between carbonatite and
kimberlite melts in chloride–carbonate–silicate systems: experiments and
application to natural assemblages. J. Petrol. 52, 1307–1331 (2011).
48. Dasgupta, R. & Hirschmann, M. M. The deep carbon cycle and melting in Earth’s
interior. Earth Planet. Sci. Lett. 298, 1–13 (2010).
49. Sobolev, S. V. & Babeyko, A. Y. What drives orogeny in the Andes? Geology 33,
50. Babeyko, A. Y., Sobolev, S. V., Trumbull, R. B., Oncken, O. & Lavier, L. L. Numerical
models of crustal scale convection and partial melting beneath the Altiplano-Puna
plateau. Earth Planet. Sci. Lett. 199, 373–388 (2002).
51. Poliakov, A. N., Cundall, P. A., Podladchikov, Y. Y. & Lyakhovsky, V. A. in Flow and
Creep in the Solar System: Observations, Modelling and Theory (eds Stone, D. B. &
Runcorn, S. K.) 175–195 (Kluwer Academic, 1993).
52. Katz, R. F., Spiegelman, M. & Langmuir, C. H. A new parameterization of hydrous
mantle melting. Geochem. Geophys. Geosyst. 4, doi:10.1029/2002gc000433
53. Spandler, C., Yaxley, G., Green, D. H. & Rosenthal, A. Phase relations and melting of
anhydrous K-bearing eclogite from 1200 to 1600 degrees C and 3 to 5 GPa.
J. Petrol. 49, 771–795 (2008).
54. Kelemen, P. B., Hirth, G., Shimizu,N., Spiegelman, M. & Dick, H. J. B. A review of melt
migration processes in the adiabatically upwelling mantle beneath oceanic
spreading ridges. Phil. Trans. R. Soc. Lond. A 355, 283–318 (1997).
55. Dasgupta, R. & Hirschmann, M. M. Effect of variable carbonate concentration on
the solidus of mantle peridotite. Am. Mineral. 92, 370–379 (2007).
56. Wignall, P. B. et al. Volcanism, mass extinction and carbon isotope fluctuations in
the Middle Permian of China. Science 324, 1179–1182 (2009).
57. Retallack, G. J. & Jahren, A. H. Methane release from igneous intrusion of coal
during late Permian extinction events. J. Geol. 116, 1–20 (2008).
Macmillan Publishers Limited. All rights reserved
WWW.NATURE.COM/NATURE | 1
Studied samples are olivine phyric, olivine-plagioclase phyric or
aphyric basalts and rarely picrites from outcrops or drill core of lava
flows (most samples), and outcrops of dykes (few samples) see
Supplementary Table S1. They contain visually fresh olivine
phenocrysts, microphenocrysts or olivine-pyroxene intergrowths and
were carefully selected from lavas from Norilsk area, Putorana
plateau and Maymecha-Kotuy province.
We studied samples from most lava units of the Norilsk
section, covering most parts of the region
. Samples were
carefully located relative to the lower boundaries of the units and
were placed within the integrated lava section of Siberian traps using
this information and absolute positions of unit boundaries after
Fedorenko et al, 1996 ref(19). Abbreviations at Fig. 1D indicate the
following lava suites of the Norilsk area placed in the order of
decreasing age: Iv-Ivakinskaya, Gd-Gudchikhinskaya, Tk-
Tuklonskaya, Nd-Nadezhdinskaya, Mr-Morangovskaya, Mk-
Mukulaevskaya, Hr-Haerlakhskaya, Km-Kumginskaya, Sm-
indicates the element ratio normalized to the
primitive mantle composition
. We also studied olivine phyric
basaltic dykes (samples 22-12 and 22-6), cutting the upper
Mukulaevskaya suite and thus representing any of upper units from
Hr to Sm.
Studied samples came from collection of V.
and Y. Vasiliev (picrites of Ayan river)
estimated approximate positions of studied samples in the integrated
lava section (Fig 1D) using data reported in ref(19, 61).
Maymecha-Kotuy province. Most lavas of this province are highly
and are not discussed in this paper. Here we report olivine
compositions from only one sample, which represents the major type
of voluminous tholeiitic lavas from the Maymecha-Kotuy province
It is similar in composition to the predominant type of Norilsk and
Putorana basalts and is placed in the integrated lava section (Fig 1D)
using the data
Magmas originated from mantle peridotite or reaction pyroxenite
should crystallize magnesian olivine at low pressures
minimum forsterite component is expected for olivine in equilibrium
with melt originated from the pure pyroxenitic source. Sobolev et al
(2009) ref(14) estimated the composition of such an olivine to be in
the range of Fo84-86 similar to the most magnesian composition of
actual olivine from lavas of Gudchikhinkaya formation (up to Fo
83.5, ref(14)). This suggests minimal fractionation of these lavas. The
subalkaline picrites from Ayan River on the Putorana plateau have
also high-Mg olivine (up to Fo 90, see Fig S1) and thus a low degree
of fractionation. The shallow-derived Siberian lavas (no garnet in the
source) underwent more crystal fractionation and highly magnesian
olivine is exceptionally rare in these lavas. Here we report olivine
with up to Fo79 in upper, most voluminous volcanic formations of
Norilsk area and corresponding lava suites at Maymecha-Kotuy and
Putorana regions (see Fig S1). This value is more magnesian than
previously reported in these suites. Figure S1 compares the
compositions of the most magnesian olivines reported in this paper
with compositions of olivine expected to crystallize from peridotite-
. For the same Fo content, all Siberian olivines
are significantly higher in Ni and lower in Mn (seen also in higher
FeO/MnO ratio) than olivine from peridotite derived magmas.
Crystallization of large amounts of early clinopyroxene could
potentially create a similar affect in coexisting olivine. This is
demonstrated by compositional trend of olivine from sample CY-315
from Tuklonskaya formation (Fig S1). This olivine has lowest Ni of
the all Siberian samples and defines the flattest trend in Fo-NiO
space. In this olivine CaO decreases rapidly and FeO/MnO increases
with decreasing Fo. Thus for this lava we cannot exclude the
possibility that the primary high-Mg melt was derived from a
peridotite- dominated source. However, the CaO contents of olivine
from other samples do not support this interpretation (see Fig S1C).
Instead, we suggest that high Ni and low Mn in Siberian olivines
imply a large fraction of non-peridotitic lithology in their source
We note that FeO/MnO ratio of olivine does not change significantly
during shallow fractional crystallization as indicated by the trends on
Figure S1D. This supports the use of this ratio for estimating amount
of pyroxenitic component for highly evolved low magnesian olivine.
The most magnesian olivine reported here (Fo79) comes from the
most voluminous traps lavas but it is still much less magnesian than
expected for their primary melts. This may suggest that an early stage
of their fractionation took place at deep magma chambers, perhaps at
40-50 km depth close to the level of their origin predicted by model.
Calculation of density
Apart from temperature, pressure and initial composition, the
densities of all materials are assumed to be affected by variable
content of melt, crystallized melt or melt extraction. All these effects
are linearized and are considered in Boussinesq approximation.
Density of material i is
is density of pure non molten i-th material, X
concentration of extracted peridotitic melt, X
is concentration of the
pyroxenitic component, X
is concentration of crystallized melt
derived from the pyroxenitic component, X
is concentration of
crystallized melt derived from the peridotitic component, X
concentration of melt derived from the pyroxenitic component, X
concentration of melt derived from the peridotitic component. All
concentrations are volume concentrations. Multipliers of
concentrations (“a” parameters) are constants depending on material
and pressure. The “a” parameters related to melt were assumed to be -
0.1 for all materials. Values for the “a” parameters related to
crystallized melt and pyroxenitic component in mantle materials
(mantle lithosphere, asthenosphere and plume) were estimated
calculating rock densities for different pressure and temperature from
bulk chemical compositions of pyroxenite-derived
using a thermodynamic modeling approach
density effect of melt extraction from peridotite (depletion) we use
. All these parameters were found to be close to 0.05 (0.045-
0.055) in the eclogite stability field. For simplicity, we assume them
to be equal to 0.05 in the eclogite stability field (here at depth > 55
km). Change of density due to gabbro-eclogite transformation
considered by linearly decreasing “a” parameters from 0.05 at depth
of 55 km to -0.05 at depth of 40 km.
With the accepted values of parameters, the buoyancy of the
thermo-chemical plume with an excess temperature of 250 °C
(Tp=1600°C) and an eclogite content of 15 % is only 10% of the
buoyancy of the purely thermal plume with the same temperature.
2 | WWW.NATURE.COM/NATURE
doi:10.1038/nature10385RESEARCHR RESEARCH SUPPLEMENTARY INFORMATION
Fig.S1. Compositions of representative Siberian trap olivine, compared with olivine from peridotite derived melts. A. Numbered lines and field are
compositions of olivine crystallized from peridotite derived melt with initial MgO content indicated by corresponding number
. Blue field outline compositions of
low magnesian olivines (Fo<60) from Siberian traps (Supplementary Table S3). Knipovich MORB K22-23-stands for olivine compositions from Knipovich ridge
MORB sample with almost zero content of pyroxenitic component in the melt
. C. Solid black lines indicate linear regression lines for the olivine of different
samples. Other symbols are as on Fig S1A. D. Dashed red and green lines are suggested boundaries for FeO/MnO ratios of end-member olivines crystallized from
pyroxenite and peridotite derived melts respectively. Slope of these lines constrained to be parallel to the dominant slope of trend lines. Other symbols are as on
Fig S1A,B. Error bars outlined by the blue circle are 2-standard deviations estimated from repeated analyses of olivine standard.
The setup and boundary conditions for our
preferred model are presented in Fig. S2. We
assume that the plume head arrives at the base
of the cratonic root at model time 0. The left
model boundary is open for inwards- and
outwards flow of material, while the right
boundary and bottom are closed, to avoid
instabilities. We use a non-uniform grid to
better resolve the central part of the model.
Finite element size is 5 X 5 km in the best-
resolved part of the model. Tests show no
significant influence on model results of mesh
coarsening out of that region. Plume potential
temperature for the preferred model is
Tp=1600°C and the content of recycled oceanic
crust (eclogite) is 15 wt%. Evolution in time of
temperature and pyroxenite content are shown
in Fig. S3.
Fig. S2. Initial and boundary conditions for the preferred model. Background colors show initial
distribution of rock-types. Dark blue-upper crust, blue- lower crust, green- depleted mantle lithosphere,
yellow- asthenosphere, red- plume. White dashed line shows bottom of the thermal boundary layer,
where potential temperature, reaches asthenospheric temperature of 1350 °C. Potential temperature in
the plume head is 1600°C. Also shown is the finite element grid which is finest in the central-upper
part of the model.
WWW.NATURE.COM/NATURE | 3
SUPPLEMENTARY INFORMATION RESEARCH
doi:10.1038/nature10385RESEARCHR RESEARCH SUPPLEMENTARY INFORMATION
Fig. S3. Evolution of the potential temperature distribution Tp (left column) and of distribution of the pyroxenite component in the plume or in the
crystallized melt C_Px (right column))
Model sensitivity and preferred model
The intensity of lithospheric destruction and melt production
depend on (1) the initial thickness and degree of depletion of the
lithosphere (Fig. S4), (2) plume temperature, and volume (Fig. S5),
and (3) content of the recycled crust in the plume (Fig. S6).
Our preferred model is constrained by the petrological data,
i.e. a plume head potential temperature of 1600°C, 15 wt% of
recycled crust, and an initial thickness of depleted lithosphere of
130 km. A plume head radius of 400 km was chosen arbitrarily.
With the fixed initial thickness of the lithosphere, a lower plume
potential temperature (1500-1550°C instead of 1600°C in preferred
model), or a smaller plume (200-300 km radius instead of 400 km)
would result in much less extensive lithospheric delamination and
lower melt production (Fig. S5). A content of recycled crust
significantly higher than 15 wt% makes the plume negatively
buoyant (at plume potential temperature of 1600°C), while a lower
content results in large pre-magmatic uplift (Fig. S6).
4 | WWW.NATURE.COM/NATURE
doi:10.1038/nature10385RESEARCHR RESEARCH SUPPLEMENTARY INFORMATION
We prefer a model of re-fertilized mantle lithosphere because
the lithosphere of Siberian Craton experienced multiple
interactions with mantle plumes before the emplacement of
Siberian Traps, as evidenced by multiple phases of kimberlite
magmatism and synchronous flood basalts
. Choosing depleted
lithosphere with two-times larger density deficit does not
significantly change the results (Fig. 3a main text, dashed curve
and Fig. S4). However, further rise of the average degree of
depletion and density deficit, to 60 and 70 kg/m
impedes lithospheric destruction (Fig. S4).
Fig. S4. Effect of thickness of the depleted lithosphere Hl (a) and degree of depletion (b) on the intensity of the lithospheric destruction. Shown are
potential temperature distributions at model time 1.0 Myr.
Fig. S5. Effect of a plume potential temperature, Tp (a) and a plume radius R (b) on the intensity of the lithospheric destruction. Shown are potential
temperature distributions at model time 1.0 Myr.
WWW.NATURE.COM/NATURE | 5
SUPPLEMENTARY INFORMATION RESEARCH
Fig. S6. Effect of a plume composition on pre-magmatic surface topography (H) and on intensity of the lithospheric destruction. Shown is the maximum
pre-magmatic surface topography (upper panel) and temperature distributions at model times 0.5 and 1.0 Myr. Content of the recycled crust component in
plume (Xe) is 15 Wt% (a) and 7.5 Wt% (b).
Krivolutskaya, N. & Rudakova, A. Structure and Geochemical
Characteristics of Trap Rocks from the Noril’sk Trough, Northwestern
Siberian Craton. Geochemistry International 47, 675-698,
Krivolutskaya, N. A., Sobolev, A. V., Mikhailiov, V. N. & Svirskaya, N.
M. New data concerning the high-Mg rocks of the Siberian trap
formation in the Noril'sk region. Geochimica Et Cosmochimica Acta 71,
Hofmann, A. W. Chemical differentiation of the Earth: the relationship
between mantle, continental crust, and oceanic crust. Earth Planet. Sci.
Lett. 90, 297-314 (1988).
Nesterenko, G. V., Tikhonenkov, P. I. & Kolesov, G. M. Rare-earth
elements in plateau-basalt of the Siberian platform. Geokhimiya, 823-
Nesterenko, G. V., Tikhonenkov, P. I. & Romashova, T. V. Basalts of
Putorana plateau. Geokhimiya, 1419-1425 (1991).
Sharma, M., Basu, A. R. & Nesterenko, G. V. Nd-Sr isotopes,
petrochemistry, and origin of the Siberian flood basalts, USSR.
Geochimica Et Cosmochimica Acta 55, 1183-1192 (1991).
Sharma, M., Basu, A. R. & Nesterenko, G. V. Temporal Sr-isotopic, Nd-
isotopic and Pb-isotopic variations in the Siberian flood basalts -
implications for the plume-source characteristics. Earth and Planetary
Science Letters 113, 365-381 (1992).
Vasilev, Y. R. Plagioclase bearing picrites of Ayan river. Russian
Geology and Geophysics, 68-75 (1988).
Vasiliev, Y. R. & Zolotukhin, V. V. The Maimecha-Kotui alkaline-
ultramafic province of the northern Siberian Platform, Russia. Episodes
18, 155-164 (1995).
Arndt, N., Chauvel, C., Czamanske, G. & Fedorenko, V. Two mantle
sources, two plumbing systems: tholeiitic and alkaline magmatism of the
Maymecha River basin, Siberian flood volcanic province. Contributions
to Mineralogy and Petrology 133, 297-313 (1998).
Vasilev, Y. R. Low potassium basalts of Maimecha-Kotuy province and
their likely geodynamic setting. Doklady Akademii Nauk 366, 507-510
Hirth, G. & Kohlstedt, D. L. in In Inside the Subduction Factory Vol.
138 (ed J. Eiler) 83-105 (American Geophysical Union 2003).
Walter, M. J. Melting of garnet peridotite and the origin of komatiite and
depleted lithosphere. Journal of Petrology 39, 29-60 (1998).
Sobolev, S. V. & Babeyko, A. Y. Modeling of mineralogical
composition, density and elastic-wave velocities in anhydrous magmatic
rocks. Surveys in Geophysics 15, 515-544 (1994).
Schutt, D. L. & Lesher, C. E. Effects of melt depletion on the density
and seismic velocity of garnet and spinel lherzolite. Journal of
Geophysical Research-Solid Earth 111, B05401, doi:
Griffin, W. L. et al. The Siberian lithosphere traverse: mantle terranes
and the assembly of the Siberian Craton. Tectonophysics 310, 1-35
Kuzmin, M. I., Yarmolyuk, V. V. & Kravchinsky, V. A. Phanerozoic hot
spot traces and paleogeographic reconstructions of the Siberian continent
based on interaction with the African large low shear velocity province.
Earth-Science Reviews 102, 29-59, doi:10.1016/j.earscirev.2010.06.004
6 | WWW.NATURE.COM/NATURE
Table S1 notes. Group: deep (high Gd/Ybn>1.6 =garnet signature), Sh shallow (low Gd/Ybn< 1.6 = no garnet in the
MK MaymechaKotuy province. Unit: (Norilsk region GdGudchikhinskaya, TkTuklonskaya, NdNadezhdinskaya, Mr
Morangovskaya, MkMukulaevskaya, HrHaerlakhskaya, KmKumginskaya, SmSamoedskaya; Putorana plateau Ay
Ayanskaya, HoHonumanskaya; MaymechaKotuy province On Onkuchakskaya. Ref: references. Type: lavas or dykes.
N2:amountof olivinegrainshavingcompositionwithin maximum 3mol%Fo.Avg Fo (3)averageof mostmagnesian
olivinegrainshavingcompositionwithinmaximum 3 mol% Fo. Xpx Ni and Xpx Mnproportionofpyroxenitederived
melt calculated from Ni excess and Mn deficiency correspondingly in average most magnesian olivine. Xpxaccepted
proportionofpyroxenitederived melt (see text).S.e.standarderror of mean. FeO/MnO correspondingrationforthe
Sample Group Region !"#$ Ref Type Thick m N1 Max Fo N2 Avg Fo (3) Xpx Ni S.e Xpx Mn S.e Xpx S.e %&'()"' Gd/Yb n
SU-50 Deep Nor Gd S3 lava 345 128
75 79.71 1.32 0.01 0.99 0.00 1.15 0.17 !%#! 2.60
4270/13 Deep Nor Gd S3 lava 375 47
33 79.18 1.34 0.01 0.98 0.00 1.16 0.18 !%#( 2.34
Deep Nor Gd new lava 380 52
7 79.12 1.38 0.03 1.02 0.01 1.20 0.18 !$#! 2.24
XS-51/130 Deep Nor Gd S3 lava 408 120
85 81.51 1.16 0.00 0.89 0.00 1.03 0.14 (+#) 2.27
991a Deep Put Ay new lava 1540 98
44 87.85 0.71 0.01 0.90 0.01 0.80 0.10 (+#+ 4.25
991b Deep Put Ay new lava 1560 103
42 87.26 0.73 0.01 0.88 0.01 0.80 0.08 (+#$ 4.41
81-133 Deep Put Ay new lava 1580 100
73 86.59 0.75 0.00 0.84 0.00 0.80 0.05 (!#$ 4.50
Sh (Fo>60) Nor Tk new lava 580 9
9 70.87 0.41 0.01 0.59 0.01 0.50 0.09 ("#, 1.37
SU31 Sh (Fo>60) Nor Tk S3 lava 586 75 76.66 14 74.92 0.32 0.01 0.49 0.01 0.40 0.08 )!#+ 1.28
SU33 Sh (Fo>60) Nor Tk S3 lava 587 8 73.57 8 73.02 0.34 0.01 0.52 0.01 0.43 0.09 )+#! 1.30
SU36 Sh (Fo>60) Nor Tk new lava 590 16 76.66 7 74.22 0.35 0.01 0.44 0.01 0.40 0.04 )(#+ 1.32
Sh (Fo>60) Nor Tk new lava 592 62
54 75.59 0.27 0.00 0.42 0.01 0.35 0.07 )(#' 1.36
530/12 Sh (Fo>60) Nor Nd S3 lava 840 17 79.41 1 79.41 0.49 0.59 0.54 0.05 ("#, 1.46
Sh (Fo>60) Nor Mr new lava 1200 34
2 67.74 0.61 0.04 0.55 0.08 0.58 0.03 (&#' 1.33
Sh (Fo>60) Nor
Mk new lava
1 62.78 0.64 0.48 0.56 0.08 )!#! 1.30
Sh (Fo>60) Nor
Mk new lava
1 72.78 0.47 0.36 0.41 0.06 ))#" 1.22
Sh (Fo>60) Nor
Hr new lava
23 66.99 0.56 0.00 0.37 0.00 0.46 0.10 ))#$ 1.21
Sh (Fo>60) Nor
Sm new lava
2 72.38 0.54 0.02 0.37 0.02 0.45 0.09 ))#% 1.19
85-63 Sh (Fo>60) Put Ay new lava 1500 57
3 70.29 0.56 0.03 0.51 0.00 0.54 0.02 )+#, 1.14
83-50 Sh (Fo>60) Put Ho new lava 1800 39
2 68.83 0.57 0.03 0.44 0.01 0.50 0.06 )(#+ 1.32
1934-2 Sh (Fo>60) M-K
On new lava
5 75.61 0.57 0.03 0.45 0.02 0.51 0.06 )!#" 1.20
141-7 Sh (Fo<60) Nor Nd new lava 1175 48
28 46.54 0.79 0.01 0.61 0.01 0.61 ("#! 1.22
Sh (Fo<60) Nor Mr new lava 1232 67
50.77 0.80 0.02 0.58 0.03 0.58 ("#" 1.28
Sh (Fo<60) Nor Mr new lava 1258 48
50.70 1.05 0.02 0.70 0.01 0.70 ('#% 1.26
Sh (Fo<60) Nor Mr new lava 1321 36
49.06 0.91 0.01 0.56 0.01 0.56 (&#) 1.21
Sh (Fo<60) Nor Mk new lava 1659 59 56.25
56.25 0.62 0.44 0.44 )(#+ 1.20
Sh (Fo<60) Nor Mk new lava 1736 69
56.88 0.67 0.02 0.44 0.03 0.44 )(#+ 1.23
Sh (Fo<60) Nor Mk new lava 1818 52
58.74 0.78 0.02 0.55 0.01 0.55 (&#' 1.28
Sh (Fo<60) Nor Mk new lava 1906 73
52.17 0.93 0.02 0.55 0.01 0.55 (&#, 1.25
Sh (Fo<60) Nor Mk new lava 2080 45
48.33 1.08 0.01 0.68 0.01 0.68 ($#! 1.21
Sh (Fo<60) Nor Mk new lava 2130 12
46.70 1.01 0.04 0.61 0.02 0.61 (%#& 1.22
Sh (Fo<60) Nor Mk new lava 2235 51
47.30 0.98 0.01 0.52 0.01 0.52 )+#! 1.15
Sh (Fo<60) Nor Hr new lava 2310 60
46.16 1.14 0.01 0.63 0.00 0.63 (%#' 1.23
Sh (Fo<60) Nor Hr new lava 2430 27
50.02 1.08 0.02 0.56 0.01 0.56 (&#) 1.06
Sh (Fo<60) Nor Hr new lava 2460 12
45.85 0.92 0.02 0.48 0.01 0.48 )!#! 1.18
115-1 Sh (Fo<60) Nor Hr new lava 2650 111 59.01
59.01 0.71 0.34 0.34 ),#( 1.16
115-9 Sh (Fo<60) Nor Hr new lava 2800 75
51.21 0.70 0.01 0.26 0.02 0.26 )'#& 1.29
Sh (Fo<60) Nor Km new lava 2862
46.53 0.69 0.26 0.26 )'#& 1.30
Sh (Fo<60) Nor Km new lava 2890 37
45.78 0.67 0.01 0.20 0.01 0.20 )%#! 1.22
119-1 Sh (Fo<60) Nor Km new lava 2900 48
,)#$" " ,)#$"
0.58 0.29 0.29 )'#) 1.21
Sh (Fo<60) Nor Km new lava 2930 18
52.55 0.67 0.02 0.28 0.08 0.28 )'#' 1.22
125-4 Sh (Fo<60) Nor Sm new lava 3140 74
57.24 0.79 0.01 0.37 0.01 0.37 ))#$ 1.19
Sh (Fo<60) Nor Sm new lava 3420 83
49.82 0.85 0.02 0.36 0.01 0.36 ))#" 1.16
Sh (Fo<60) Nor Sm new lava 3480 48 55.36
55.36 0.68 0.47 0.47 )!#, 1.18
81-134 Sh (Fo<60) Put Ay new lava 1300 26 59.80
59.80 0.69 0.51 0.51 )+#) 1.15
Sh (Fo<60) Nor Hr-Sm new dyke - 71
15 73.67 0.58 0.01 0.49 0.01 0.53 0.05 )!#+ 1.19
Sh (Fo<60) Nor
dyke - 51
28 76.64 0.37 0.00 0.21 0.01 0.29 0.08 )$#% 1.13
12-1 Sh (Fo<60) Nor
dyke - 56 48.62 12 46.72 0.69 0.02 0.35 0.01 0.35 ))#& 1.26
39-3 Sh (Fo<60) Nor
dyke - 98
87 52.46 0.84 0.00 0.36 0.00 0.36 ))#% 1.22
125-5 Sh (Fo<60) Nor
dyke - 50
24 45.80 0.78 0.01 0.51 0.01 0.51 )+#) 1.18
120 Sh (Fo<60) Nor
dyke - 75
34 59.49 0.73 0.00 0.28 0.00 0.28 )'#, 1.16
WWW.NATURE.COM/NATURE | 7
SUPPLEMENTARY INFORMATION RESEARCH