Signatures of Early Earth Differentiation in the Deep Mantle? Richard W. Carlson Department of Terrestrial Magnetism Carnegie Institution of Washington COMPRES, June 15, 2011

Continental Crust Formation has Caused Chemical Differentiation of the Mantle Mass Fraction 0.45% Sample/Bulk-Silicate-Earth 30-70% Partial melting of the shallow (< 200 km) mantle ultimately results in the creation of continental crust that is highly enriched in incompatible trace elements. Though small in volume, the extreme incompatible element enrichment of the continental crust has left the residual mantle depleted in these elements. This depleted mantle now serves as the source of mid-ocean ridge basalts. Continental crust formation started by at least 4.3 Ga and continues today. A critical question developed in the last decade is the composition of the mantle before continental crust was extracted from the mantle. Conventional wisdom models this “primitive” mantle through chondritic analogs, at least for refractory lithophile elements, but some critical differences have been found that suggest that the “primitive” mantle is instead the remnant of early Earth differentiation processes, perhaps occurring at very high pressures and temperatures. 70-30% WHEN DID THIS SEPARATION OCCUR? Sm-Nd model ages for MORB = 200 - 2000 Ma Pb-Pb model age for oceanic basalts ~1800 Ma “Average” continental crust Sm-Nd model age ~2000 Ma

A Remnant of Early Differentiation or Modern Subduction? LLSVPs: A Remnant of Early Differentiation or Modern Subduction? Seismic examination of the base of the mantle is showing a good deal of velocity heterogeneity structure. Two large areas of low shear velocity, termed Large Low Shear Velocity Provinces (LLSVPs) originally were proposed to be starting thermal plumes, but recent work identifying sharp “edges” to the LLSVP’s suggest there is a compositional distinction as well. Are the LLSVP’s just hot segregated components of the plate tectonic recycling of oceanic lithosphere produced in the shallow mantle, or could they instead be long-lived compositionally distinct regions produced by early Earth differentiation events? This talk will explore the evidence in support of the latter possibility. Garnero and McNamara, 2008

Some Meteorites are Compositionally Similar to the Sun Some Meteorites are Compositionally Similar to the Sun. These Serve as a Starting Point for Estimating Bulk Earth Composition, but how well is the Chondrite Model Matched by Real Earth Rocks? N C In? For most elements, CI chondrites provide a good approximation of solar composition Most models for the composition of Earth are based on chondrites as a starting point as at least some groups of these meteorites, especially the CI’s, have compositions similar to the Sun for all but the most volatile (H, He, C, N) elements and for some elements that are consumed by nuclear reactions (Li) in the Sun. We know that the shallow Earth (crust, upper mantle) does not have chondritic abundances of most of these elements, but the presumption has been that extraction of continental crust from the mantle has created two chemically complimentary reservoirs that together would combine to a “primitive” composition close to that predicted from chondrite analogs. Given the amount of continental crust and its degree of incompatible element enrichment, some portion of the mantle, perhaps the whole lower mantle, would retain its primitive composition with only the upper mantle reflecting the incompatible element depletion caused by extracting the continental crust. Li Solar and CI compositions from Palme and O’Neill, Treatise on Geochemistry, 2003

The Bulk Earth is NOT CI Chondritic: Volatile Depletion is a Characteristic of Many Solar System Objects, Including Earth Although “primitive mantle” and “chondritic mantle” often are used interchangeably, many element abundances in the Earth, and in Earth’s mantle, clearly are not chondritic. In this example, Earth shows a depletion in elements plotted according to their condensation temperature in the Solar nebula. The more volatile, the more depleted. Other types of chondrites show similar patterns of volatile depletion, though not as extreme. This feature suggests that Earth’s volatile depletion was inherited from its building blocks – in other words, Earth never had its full compliment of volatile elements. From McDonough TOG, 2003 CI-normalized terrestrial volatile element abundances decrease with decreasing condensation temperature. Same pattern, though less extreme, is seen in “primitive” meteorites. Volatile depletion of Earth may be a “pre-accretion” phenomena

Dating Early Earth Differentiation Actively-used short-lived radioactive isotopes Parent Isotope Atom % Half-life (Ma) Daughter Isotope 26Al 0.005 0.73 26Mg 60Fe 3.7 x 10-7 1.5 60Ni 53Mn 0.00063 3.7 53Cr 107Pd 0.0015 6.5 107Ag 182Hf 0.0037 9 182W 129I 0.011 15.7 129Xe 244Pu 244Pu/238U = 0.0068 80 Fission Xe 146Sm 0.026 103 142Nd The continuing chemical differentiation of Earth has obscured much of the evidence for early Earth differentiation that can be extracted from compositional characteristics or from radiogenic isotope systems based on radioactive isotopes with half-lives similar to, or longer than, the age of the Earth. Significant advances have been made in the investigation of differentiation events occurring within tens to 100 Ma of planet formation through the application of short-lived radionuclides that were present when the Solar system formed. Besides their high temporal resolution, short-lived systems more than double the number of chronometers available and have a wide range of chemical properties that allow the investigation of numerous processes important in planet differentiation. They have the additional advantage that changes in the daughter isotopic composition can only be created by parent-daughter fractionation that occurs while the parent is still extant. This makes them very sensitive to early differentiation, but less sensitive to being overprinted by events occurring later in Earth history. Condensation – Volatile Loss: Al-Mg, Mn-Cr, Pd-Ag, I-Xe Metal – Silicate Separation: Fe-Ni, Pd-Ag, Hf-W Silicate Differentiation: Al-Mg, Fe-Ni, Mn-Cr, Hf-W, Sm-Nd

The Bulk Earth is NOT CI Chondritic: Volatile Depletion is a Characteristic of Many Solar System Objects, Including Earth Although “primitive mantle” and “chondritic mantle” often are used interchangeably, many element abundances in the Earth, and in Earth’s mantle, clearly are not chondritic. In this example, Earth shows a depletion in elements plotted according to their condensation temperature in the Solar nebula. The more volatile, the more depleted. Other types of chondrites show similar patterns of volatile depletion, though not as extreme. This feature suggests that Earth’s volatile depletion was inherited from its building blocks – in other words, Earth never had its full compliment of volatile elements. From McDonough TOG, 2003 CI-normalized terrestrial volatile element abundances decrease with decreasing condensation temperature. Same pattern, though less extreme, is seen in “primitive” meteorites. Volatile depletion of Earth may be a “pre-accretion” phenomena

Earth Formed Volatile Depleted Chondrite Mn/Cr variation correlates with 53Cr/52Cr. Earth has a lower 53Cr/52Cr than almost all chondrites. Mn more volatile than Cr. Earth’s volatile depletion occurred while 53Mn was alive (t1/2 = 3.7 Ma) The timing of Earth’s volatile depletion can be estimated from the 53Mn-53Cr decay system because Mn is moderately volatile while Cr is refractory. With a half life of only 3.7 Ma, any variation in the 53Cr/52Cr ratio is a sign of Mn/Cr fractionation within the first few million years of Solar system evolution. Although there is some scatter, likely indicative of both the details of parent body formation and later metamorphism, whole rock chondrites of different types form a Mn/Cr vs. 53Cr/52Cr (expressed in parts in 10,000 difference from the terrestrial standard) correlation indicative of an age of 4566 ± 2 Ma, overlapping the best estimate of the Solar system formation age of 4568 Ma. The 53Cr/52Cr of Earth is well determined and using an estimate of the Mn/Cr ratio of the bulk Earth puts Earth on the chondrite line indicating that its volatile depletion was present within 2 ± 2 Ma of Solar system formation. Earth From Qin et al., GCA 2010

Earth’s Mantle is Depleted in Siderophile Elements Besides Earth’s depletion in volatile elements compared to CI chondrites, we also know that Earth’s mantle is depleted in siderophile elements. A great deal of experimental effort has been put into determining the pressure, temperature and compositional space that would allow one to deduce from these siderophile element abundances the conditions of core formation on Earth. Substantial advances in our understanding of the P and T dependency of the metal-silicate distribution coefficients for these elements have arisen from this work. One feature of the siderophile element pattern that remains difficult to explain by core-mantle equilibrium is the chondritic relative abundances, and unexpectedly high abundances of the highly siderophile elements (PGE, Au, Re) that have metal-silicate D’s over 100,000 in some cases. Palme and O’Neil, TOG, 2003

Reconciling Mn-Cr, Pd-Ag, and Hf-W Constraints on the Timescale of Earth Volatile-Depletion and Core Formation 26 Myr accretion of volatile-poor material (86% of Earth mass) 4% CI added at 26 Myr Adding Mn-Cr, Pd-Ag, and Hf-W results together allow constraints to be placed both on the timing of core formation and the composition of accreting material. The low 53Cr of Earth requires initial accretion of volatile-poor material. To extend the Pd-Ag core formation age to within the range calculated from Hf-W requires the late (after 53Mn has decayed away) addition of volatile-rich material. One (non-unique) model that simultaneously explains the 3 systems has Earth growing to 86% of its current mass by accretion of volatile-depleted material for 26 Ma followed by addition of another 4% of its mass from volatile-enriched material. A final 9% of Earth’s mass of volatile-rich material is added in the giant impact that forms the Moon, but the timing of this impact is not well constrained because it likely occurred after 107Pd and 182Hf had decayed away. (Adds another 9% of Earth Mass) Schonbachler et al., Science 2010

Refractory Lithophile Elements SHOULD be Present in the BSE in Chondritic Relative Abundances, but Often They are Not Finding rocks with chondritic relative abundances of refractory lithophile elements, however, has proven difficult. Even a group of peridotite xenoliths selected on the basis of major element compositions close to expectations of “primitive” mantle show depletion in the more incompatible of the refractory lithophile elements. This could reflect the extraction of continental crust from the upper mantle because the extreme enrichment of the crust in the most incompatible elements will strongly deplete them in the residual mantle, but the small volume of crust means that its formation will have only minimal effect on the major element composition of the residual mantle. “Fertile” mantle xenoliths (from Palme and O’Neill, TOG, 2004, after Jagoutz et al., 1979)

146,147Sm-142,143Nd Systematics Short-lived chronometer: 146Sm 142Nd (T1/2= 103 Ma) 146Sm exists only in the first 500 Ma of Solar System history Coupled to the long-lived chronometer: 147Sm 143Nd (T1/2 = 106 Ga) 147Sm abundance decreased by only 3% in 4.56 Ga Zircon 4.4 Ga Isua 3.8 Ga Refractory lithophile element differentiation is well tracked with the Sm-Nd radiometric system that includes two embedded decays. The long-lived 147Sm (106 Ga half life) decay to 143Nd records every change in Sm/Nd ratio that has occurred during all of Earth history. The short-lived 146Sm (103 Ma half life) records changes in Sm/Nd ratio only until 146Sm is extinct, some 500-600 Ma after Earth formation. Variation in 142Nd relative abundance thus can only be created by differentiation events occurring in the first few hundred million years of Earth history. Mixing between early formed reservoirs can destroy 142Nd variation later in Earth history.

142Nd Variation in Earth Materials Limited and Restricted Only to Rocks Older than 3.5 Ga 142Nd excesses measured in 3.8 Ga samples from SW Greenland and Anshan, China (up to 0.15e). 142Nd deficiencies in Nuvvuagittuq, Quebec, Canada Evidence for early differentiation, but not all old rocks show this No heterogeneities preserved after 3.5 Ga in the convecting Earth’s mantle The only variation in 142Nd/144Nd seen in Earth rocks occurs in rocks older than 3.5 Ga including rocks from Isua Greenland and Anshan China, both of which have 142Nd/144Nd roughly 15 ppm higher than most other Earth rocks. This variation in 142Nd/144Nd does not correlate with the rocks Sm/Nd ratio and thus probably reflects a characteristic of their mantle source. In contrast, rocks from the Nuvvuagittuq greenstone belt in northern Quebec show low 142Nd/144Nd that does correlate with Sm/Nd providing a 146Sm-142Nd isochron of age near 4.3 Ga making these the oldest crustal rocks on Earth. An important point about this isochron, however, is that the initial 142Nd/144Nd at the time of formation of these rocks is the same as that the mantle source of all modern igneous rocks would have had at 4.3 Ga. External Precision

Is “Terrestrial” 142Nd/144Nd Chondritic? – No! 142Nd/144Nd ratios measured in carbonaceous, ordinary and some enstatite chondrites, and eucrites, are lower than laboratory standard and terrestrial samples Excess 142Nd in Earth rocks indicative of higher than chondritic Sm/Nd ratio while 146Sm was still extant. The first expectation if the bulk silicate Earth (BSE) has chondritic relative REE abundances is that the 142Nd/144Nd of Earth should be the same as chondrites. It isn’t. All Earth rocks so far analyzed, with the exception of the LREE enriched Nuvvuagittuq samples, have 142Nd/144Nd about 20-30 parts per million higher than chondrites. Implies that even when 146Sm was extant, the mantle had a Sm/Nd ratio at least 6.5% higher than chondritic. Open symbols show data from Nyquist et al., 1995; Andreasen and Sharma, 2006; Rankenburg et al., 2006. Closed symbols are data from Boyet and Carlson, 2005; Carlson et al., 2007.

Constraints on the Timing of Earth Differentiation 5 Ma, 147Sm/144Nd=0.209 30 Ma, 147Sm/144Nd=0.212 60 Ma, 147Sm/144Nd=0.216 100 Ma, 147Sm/144Nd=0.222 Mid-ocean ridge basalts The longer one waits to form the high Sm/Nd ratio source, the higher the Sm/Nd has to be in order to create the observed 20 ppm offset in 142Nd/144Nd between chondrites and Earth. This will have consequences for 143Nd/144Nd evolution such that if the high Sm/Nd ratio was established later than 30 Ma after Earth formation, this reservoir would have 143Nd/144Nd higher than observed in MORB, which has the highest 143Nd/144Nd of any significant mantle reservoir. Archean samples chondritic evolution Differentiation event occurred during the first <30 Ma of Earth history

Predicted Parental Mantle Reservoir from 142Nd Overlaps with high 3He/4He Reservoir (Ra) Reservoir parental to terrestrial mantle “primordial” chondrite reservoir The super-chondritic 142Nd/144Nd of all modern mantle and crustal rocks requires a Sm/Nd ratio that would lead to 143Nd/144Nd between 0.5129 to 0.5131 today compared to chondritic of 0.51263. Perhaps not coincidently, the rocks with highest 3He/4He (picrites from the Baffin Island component of the North Atlantic flood basalts) have Nd isotopic compositions within the range predicted from the superchondritic 142Nd/144Nd of the modern mantle.

The Broader Trace Element Characteristics of this Ancient Depleted Source The search for magmas derived from the “most primitive” mantle sources reveals refractory lithophile trace element patterns and isotopic compositions that are subtly different from chondritic models. Jackson et al., Nature 2010

Similar Normalized Incompatible Element Patterns Found for Other Major Flood Basalts, in this case, Ontong-Java Basalts The flat primitive-mantle-normalized patterns defined by alteration-resistant incompatible elements in the Kwaimbaita- and Kroenke-type basalts (see Fitton & Godard, 2004) point to a mantle source not too different from estimated primitive mantle in most of its inter-element ratios. However, the observed isotopic values (e.g., eNd(t) ~ +6) are clearly far-removed from those estimated for primitive mantle (eNd = 0). (Tejada and Mahoney, MantlePlumes.org) 10 1 Several other flood basalt provinces also have units with trace element and isotopic systematics suggestive of derivation from this same ancient incompatible element depleted source. In most continental flood basalts the signature of this source is strongly overprinted by contamination from the crust through which the magmas flowed to eruption. In oceanic flood basalts, as in this example from the Ontong-Java Plateau, the issue of crustal contamination is not as severe, and sure enough, these basalts display minimal compositional variation and their trace element patterns and Nd isotopic compositions are consistent with a source similar to that of the Baffin Island picrites. 10 1 All these samples have e143Nd between +4 and +7

Size and Composition of the Reservoirs. So What? Reservoir Mass(1025g) Th(ppb)U(ppb) K(ppm) TW Cont. Crust 2.26 5600 1300 15000 7.3 Enriched=D” 17 920 230 2650 9.3 Enriched>1600km 111 150 40 440 10.4 Primitive (60%) 242 79 20 240 11.7 These subtle trace element difference may not seem too important, but among these elements are the 3 main heat producing elements in the mantle, K, U, and Th. Chondritic models suggest present day abundances of these elements sufficient to generate nearly 20 TW of heating. Most of this (7.3 TW) has been segregated into the continental crust. Non-chondritic models suggest that Earth may have half the abundance of these elements. Besides the obvious implications for mantle heat production, if the whole Earth is this incompatible element depleted, then most of the mantle must now be as depleted as the MORB source in order to provide all the incompatible elements that are in the continental crust. The alternative is that the source we have been calling “primitive” is in fact the result of an early global differentiation event on Earth. In this case, the early depleted reservoir (EDR) must have a chemically complimentary early enriched reservoir (EER) whose degree of incompatible element enrichment depends on its size. Early Depleted 290-390 43-53 11-13 ~150 9.5-10.3 MORB Mantle 161 7.9 3.2 50 1.1

Two Ways to Create an EDR – EER Pair Shallow Differentiation Magma Ocean Overturn Basal Magma Ocean (Labrosse et al., Nature 2007) Two ways to generate chemically complimentary sources in a magma ocean. Top panel shows a lunar analogy that differentiates the magma ocean from the bottom to the top, concentrating an iron and incompatible element rich residual liquid near the top that overturns to end up at the base of the mantle. The lower panel assumes that liquids in the deep mantle are denser than crystallizing solids in which case the magma ocean crystallizes from the middle and concentrates incompatible element rich liquids both at its top and at its bottom. Given the lack of rocks on Earth with low 142Nd/144Nd, this ancient incompatible element rich reservoir (early enriched reservoir or EER) must have sufficient density to keep it out of mantle circulation, and thus likely would be concentrated at the base of the mantle – in the LLSVPs?

How Did the Non-Chondritic Mantle Form? Melting is the easiest way to fractionate the lithophile elements, but what were the conditions of melting? Corgne et al., 2005 – 25 GPa How to decide between a deep or shallow production of the depleted reservoir. Because the fractionation of refractory lithophile elements is most likely accomplished by partial melting/fractional crystallization, compare the trace element patterns of the EDR with the distribution coefficient patterns of minerals possibly involved in the differentiation. Garnet has too steep a distribution coefficient pattern for the moderately compatible elements to match the EDR pattern. Clinopyroxene is a close match except for the most incompatible elements. Either Mg- or Ca-perovskite alone fractionate the HFSE from REE too severely to explain the EDR pattern, but combining Mg- and Ca-perovskite in the right proportions might provide a reasonable fractionating assemblage to create the EDR pattern. A big problem here is the uncertainty and possible variability of the perovskite D’s at high P and T.

Signatures of Early Earth Differentiation in the Deep Mantle? Earth accreted first, and mostly, from volatile-depleted material Core formation occurred while the accreting material shifted from volatile-poor (reduced?) to volatile-rich (oxidized) First ~85% of Earth’s mass mostly volatile-poor What has been called “primitive” mantle is in fact incompatible element depleted Earth is non-chondritic in refractory lithophile element abundances? Signature of an early differentiation event? Deep fractionation of perovskite or subduction of a shallow “KREEP” crust? Only the depleted reservoir is sampled at Earth’s surface – the complementary enriched reservoir must be buried in the deep mantle – LLSVPs?

When Did Earth’s Core Form? If core formation were simple 33 ± 2 Ma after Solar System formation or 4.534 Ga If Earth grew slowly and involved many “accumulation events”, then the answer depends on the details of Earth accumulation Parts in 10,000 Parts in 10,000 The 9 Ma half life decay of 182Hf into 182W provides a chronometer of core formation because W is moderately siderophilic while Hf is lithophilic. Core metal will thus have a Hf/W = 0 and will not grow in radiogenic 182W whereas the mantle’s Hf/W will be raised from a bulk Earth value of 1 to 10, resulting in a mantle with a 182W/184W higher than chondritic by 2 parts in 10,000 (2-epsilon). Extracting accurate core formation ages from this system, however, require understanding of the details of the core formation process – one event? A small core forming event with each new planetesimal impact? 182Hf  182W (t1/2 = 9 Ma) Chondrite Hf/W = 1 Metal Hf/W = 0 Mantle Hf/W = 10

Pd-Ag Core Formation Timescale Too Fast for Hf-W! Accrete volatile-rich material first, but this violates Mn-Cr 107Pd  107Ag (t1/2 = 6.5 Myr) Pd/Ag CI = 3 Pd/Ag Earth = 13 Pd/Ag Core > 400 Pd/Ag Mantle = 0.5 The 6.5 Ma decay of 107Pd to 107Ag does not suffer the same sensitivity to the mode of core formation as does Hf-W because in this case, the parent (Pd) is quantitatively extracted into the core while the daughter (Ag) remains in reasonable abundance in the mantle. Thus, each new impactor is less able to modify the Ag isotopic composition of the growing Earth’s mantle because it must mix with the Ag already present in Earth’s mantle. Both single- and multiple-stage core formation models result in Pd-Ag core formation ages (< 11 Ma) that are much quicker than suggested by Hf-W. Unlike Hf-W, however, Pd-Ag also is sensitive to volatile loss due to the moderate volatility of Ag, but the refractory nature of Pd. If Earth accreted from material as volatile-rich as CI chondrites, then the Pd/Ag ratio of Earth would be too low to provide any chronological constraints on the timing of core formation. Earth cannot have formed from material with CI-like volatile enrichment, however, because it would then violate the Mn-Cr evidence for early volatile depletion of Earth. Dashed curves are for accumulation of material as volatile-depleted as Earth today (Pd/Ag = 12.9). Solid curves are for accumulation of CV3 chondrites (Pd/Ag = 8.5). Numbers along the curves give the mantle Pd/Ag ratio after core formation. If Earth accumulated from volatile-rich material, then Pd-Ag offers no constraints on the timing of core formation. (From Schonbachler et al., Science 2010)

The Importance of that Last 1% Earth = 6 x 1024 kg Ocean = 1.4 x 1021 kg CI Chondrite = 18 wt% H2O 1% Earth Mass of CI Chondrite contains 1021 kg water The chondritic relative abundances of the highly siderophile elements has long been used to suggest that a small fraction of Earth (~1% of its mass) was added to the mantle after core formation. A similar mass as rich in H2O as CI chondrite would add a mass of water to the Earth that is equivalent to the mass of the current oceans.

Early Earth 142Nd/144Nd and 143Nd/144Nd Evolution The +15 ppm 142Nd/144Nd of the SW Greenland Archean rocks require a 147Sm/144Nd > 0.225. The reduction in 142Nd/144Nd between 3.9 and 3.5 Ga requires mixing between high- and low-Sm/Nd reservoirs formed within tens of Ma of Earth formation. 142Nd/144Nd in Archean Mantle-Derived Rocks Initial 143Nd in Mantle-Derived Rocks

r-, s-process Variability Explains at Least some of the 142Nd/144Nd Range Between C- and O-, E-Chondrites, but not the Earth-Chondrite Offset Because 142Nd is made by a different nucleosynthetic process than other Nd isotopes, variability in the mixing of s-process and r-process produced Nd can produce variations in 142Nd/144Nd that have nothing to do with Sm/Nd ratio and 146Sm decay. S-, r-process mixing is well tracked by variations in 148Nd because this isotope is almost purely r-process while 144Nd is produced by both s- and r-process nucleosynthesis. C-chondrites show a significant range in 148Nd/144Nd that correlates with 142Nd/144Nd with exactly the slope predicted for s-, r-process mixing. This correlation, however, passes below terrestrial 142Nd/144Nd at terrestrial 148Nd/144Nd indicating that the offset in 142Nd/144Nd between Earth and chondrites is more likely due to 146Sm decay than to nucleosynthetic variations.

Example – the 4.3 Ga Nuvvuagittuq Terrane, Quebec, Canada Crust Formation (with its characteristic LREE enrichment) Started Early Example – the 4.3 Ga Nuvvuagittuq Terrane, Quebec, Canada 3.8 Ga 4.0 Ga The low 142Nd/144Nd of the Nuvvuagittuq are only found in samples that have low Sm/Nd. We suggest that this is in fact an isochron and that these rocks were formed at 4.28 Ga from a source that had the same Nd isotopic composition as modern mantle. O’Neil et al., Science, 2008