Research Article

Lithium Isotope History of Cenozoic Seawater: Changes in Silicate Weathering and Reverse Weathering

See allHide authors and affiliations

Science  17 Feb 2012:
Vol. 335, Issue 6070, pp. 818-823
DOI: 10.1126/science.1214697


Weathering of uplifted continental rocks consumes carbon dioxide and transports cations to the oceans, thereby playing a critical role in controlling both seawater chemistry and climate. However, there are few archives of seawater chemical change that reveal shifts in global tectonic forces connecting Earth ocean-climate processes. We present a 68-million-year record of lithium isotopes in seawater (δ7LiSW) reconstructed from planktonic foraminifera. From the Paleocene (60 million years ago) to the present, δ7LiSW rose by 9 per mil (‰), requiring large changes in continental weathering and seafloor reverse weathering that are consistent with increased tectonic uplift, more rapid continental denudation, increasingly incongruent continental weathering (lower chemical weathering intensity), and more rapid CO2 drawdown. A 5‰ drop in δ7LiSW across the Cretaceous-Paleogene boundary cannot be produced by an impactor or by Deccan trap volcanism, suggesting large-scale continental denudation.

Lithium, the lightest of the alkali elements, is a conservative cation in seawater. The residence time of Li in seawater (~1.2 million years) is much shorter than for the other major alkalis (Na+ and K+) but is much longer than the oceanic mixing time (~1000 years), so that Li in seawater is well mixed and homogeneous vertically and laterally in both concentration (ratio to salt) and in isotopic composition ([Li]SW ≈ 26 μM; δ7LiSW = 31‰) (1) (fig. S1). Lithium is a trace component in continental and seafloor rocks, and unlike other recorders of oceanic chemistry changes (namely Sr and Os), it is hosted almost exclusively in silicate minerals. Because of the large relative mass difference between its two stable isotopes (6Li and 7Li), low-temperature Li isotope fractionation exhibits a very large range, making Li a powerful tracer of low-temperature geochemical processes. Among continental granites, mid-ocean ridge basalt (MORB), marine authigenic aluminosilicate clays (MAACs), dissolved Li sources and sinks to and from the sea, and seawater itself, the spread in δ7Li values is more than 31‰—an enormous isotopic range.

The Li isotopic composition of seawater reflects a balance between input and removal fluxes and their isotopic compositions. The two dominant sources of dissolved Li to seawater are rivers (low-temperature chemical weathering of continental silicate rocks) and hydrothermal (HT) fluxes from mid-ocean ridge spreading centers (high-temperature weathering of oceanic silicate rocks) (26). The removal of Li from seawater is entirely by incorporation into marine sediments and low-temperature altered oceanic crust (AOC) via formation of Li-, Mg-, and Fe-bearing MAACs (“reverse weathering”) (713) (Fig. 1 and Table 1). The 7Li enrichment of seawater (much heavier than all sources) requires the existence of marine reverse weathering that produces secondary clays bearing Li that is isotopically much lighter than the seawater from which their Li is derived. Secular variations in δ7LiSW must thus reflect imbalances between the sources and sinks of Li to and from the ocean, driven by perturbations in the global silicate weathering and reverse-weathering cycles (14, 15).

Fig. 1

(A) Box model of Li cycle in modern ocean (Table 1). At steady state, the input and output fluxes and isotopic compositions balance each other. The total input flux of Li to the ocean (FInput) is the sum of riverine (FRiv), hydrothermal (FHT), and subduction reflux (FReflx). The outgoing flux (FSink) is the sum of Li removal via MAAC and AOC formation during reverse weathering. The mass balance is numerically expressed as Embedded Image. The isotopic composition of the input flux (δ7LiInput) is a flux-weighted average of the compositions of rivers (δ7LiRiv), hydrothermal fluids (δ7LiHT), and refluxed Li (δ7LiReflx). The composition of the output flux (δ7LiSink) is dependent on the seawater δ7LiSW value and has a constant offset from seawater (∆Seawater–Sink = δ7LiSW – δ7LiSink = 16‰). Steady-state isotope balance is expressed as Embedded Image. (B) One possible steady-state isotope mass balance for the Paleocene-Eocene (P-E) Ocean (60 Ma) when δ7LiSW was 22‰, 9‰ lighter than today. FRiv is set at modern values and δ7LiRiv is then constrained to lie near δ7LiUCC, reflecting congruent weathering of the peneplained transport-limited continents (29, 30) plus perhaps dissolution of the basaltic Deccan Traps and North Atlantic Igneous Provinces. This P-E scenario is an end-member paradigm. Variants of this scenario in which river and hydrothermal fluxes are allowed to vary are explored in text S4 and fig. S10. These alternative scenarios demonstrate that the river condition is constrained to remain between 2‰ and 6‰ regardless of changes in river Li fluxes, hydrothermal fluxes, or seawater Li concentrations.

Table 1

Best estimates of dissolved Li input and output fluxes and their compositions from published results (text S1).

View this table:

Seawater lithium budget. The fluxes of river Li and HT Li to the sea are comparable in magnitude but isotopically very distinct. High-temperature HT vent fluids (>350°C) are highly enriched in Li above seawater ([Li]HT ≈ 840 μM) but are only slightly fractionated from their source rocks (δ7LiHT ≈ 8.3‰; δ7LiMORB ≈ 3.7‰). The ~4‰ enrichment of HT Li above MORB is probably because of 6Li sequestration in HT Mg-rich greenstone alteration minerals (asbestos) during various stages of hydrothermal recirculation at mid-ocean ridges. In contrast, riverine Li (δ7LiRiv ≈ 23‰) is ~21‰ heavier than continental source rocks (upper continental crust; δ7LiUCC = 1.7‰) (3, 4, 16, 17) (figs. S2 to S5 and text S1).

Modern-day riverine dissolved Li displays a large spread in concentration and isotopes. Only about one-fifth of continentally weathered Li is carried in the dissolved load. The remainder is carried as Li in secondary clays (2, 6) (fig. S3). The large isotopic offset (21‰) observed between δ7LiRiv and δ7LiUCC is a function of chemical weathering intensity (18, 19). The preferential uptake of 6Li into secondary aluminosilicate clay minerals and oxides formed during weathering and transport drives δ7LiRiv much heavier than the average continental crust. The partitioning of riverine Li into dissolved and secondary phases as a function of weathering intensity and weathering regimes determines both dissolved Li flux (FRiv-Li) and δ7LiRiv. Peneplained terrains, especially those in the tropics with transport-limited regimes, exhibit congruent weathering, little or no secondary dioctahedral clay formation to entrap and transport cations, and hence dissolved Li isotope ratios that reflect their source silicate bedrocks. High-relief terrains (uplifted mountains and downcutting high plateaus) with weathering-limited regimes exhibit high physical and chemical weathering and denudation rates with extreme incongruent weathering, large rates of secondary clay formation that carry most of the weathered Li down rivers, and hence dissolved Li isotopes ratios that are 7Li-enriched relative to their source rocks (5, 20, 21) (text S1).

At steady state, marine Li removal processes must balance the fluvial and hydrothermal inputs. We group the Li removal processes together and term them “reverse weathering” (13). Preferential removal of 6Li into marine clays, similar to that on continents, leads to a large removal-induced fractionation. This reverse weathering–driven fractionation, representing “light” fractionation from the seawater Li source of about 16‰ (ΔSW–Sink ≈ 16‰), drives seawater isotopically heavy (11, 12) (text S1). Without this removal-induced Li isotope fractionation, seawater would simply reflect the flux-weighted isotope ratio of its sources (δ7LiInput ≈ 15‰). For a steady-state ocean, this removal-driven fractionation puts a boundary condition on composition of the input fluxes because δ7LiSW must remain about 16‰ heavier than δ7LiInputs. This characteristic of Li isotopes in seawater with its large isotopic separation is unique.

The removal of Li by low-temperature reverse-weathering reactions includes MAAC formation in sediments (22) and low-temperature seafloor basalt alteration (AOC formation) (7, 8, 16), both of which involve formation of Mg-rich smectites (911), and Fe-rich aluminosilicates in muddy shallow-water sediments (23). Lithium is prone to substitute for octahedral Mg in both high-temperature and low-temperature “clays.” Low-temperature weathering (<250°C) of ridge flank basalts away from spreading centers acts as a net sink for Li. In many respects, Li is a crustally recycled cation (4, 8, 9), being sourced to the ocean from HT circulation at the ridge axis and then reconsumed into basalts during low-temperature alteration off axis. In today’s ocean, MAAC is responsible for ~70% and AOC for ~30% of the total marine lithium removal by reverse weathering (Table 1).

Seawater lithium isotope record. Our seawater Li isotope record was constructed by analyzing chemically cleaned planktonic foraminifera that incorporate trace quantities of seawater dissolved Li into their calcium carbonate tests during growth in the surface ocean ([Li+]Foram ≈ 1 to 2 ppm). This intrinsic lattice-bound Li isotope ratio is independent of temperature and Li concentration (15). The analytical challenges associated with precise δ7Li determinations from mass-limited foraminifera samples (<1 mg) without contamination and laboratory fractionation artifacts were overcome by developing new methods (24, 25). We have validated the application of foraminifera as faithful recorders of δ7LiSW (fig. S6) without significant diagenetic overprints (text S3). We selected eight Deep Sea Drilling Project (DSDP)–Ocean Drilling Program (ODP) sites with existing high-resolution Sr isotope records, minimal diagenesis, and good chronologies (Fig. 2). We analyzed age- and species-overlapping foraminifera samples (n = 301) at a resolution of 500,000 to 1 million years, which included both individual species and bulk foraminifera (Fig. 3).

Fig. 2

DSDP/ODP drill sites 588, 757, 758, 926, 1262, 1263, 1265, and 1267, with preexisting high-resolution seawater strontium isotope records, from which the foraminifera samples for this work were collected (25) (text S2 and table S1).

Fig. 3

Late Cretaceous to Holocene Li isotope record (covering 68 million years) and published values of seawater records for 87Sr/86Sr and 187Os/186Os. Lithium isotope values (δ7Li), expressed as per mil (‰) variation from NIST L-SVEC standard (SRM 8545) (25), are plotted on the top left y axis. The error bars represent 2σ uncertainty associated with each quintuplicate measurement. The gray line represents 5-point running mean of δ7LiForam record; the two parallel black lines are the corresponding ±2σ uncertainty based on the average SD of all δ7LiForam measurements (σ = ±0.55‰, n = 301). The individual foraminifera species symbols are listed in fig. S8. Foraminiferal Li and Sr data are color-coded according to drill sites and are plotted on the same age chronology (33, 5053). The Cretaceous-Tertiary (K-Pg) boundary is set at 65.68 Ma (54). The Cenozoic marine Os isotope record (187Os/186Os) is plotted on the bottom left y axis (34, 35) (text S2). Because of osmium’s short residence time in the ocean and its isotopic sensitivity to impacts and mantle sources (LIPs), the 187Os/186Os record reflects large abrupt shifts that are not discernable in either the 87Sr/86Sr or 7Li/6Li records.

We make six important assumptions for mass balance calculations that drive the observed changes in δ7LiSW: (i) In contrast to previous δ7LiSW mass balance studies (7, 15), we include a Li “subduction reflux” term that recognizes solutions resulting from dewatering and breakdown of MAAC in the downgoing slab during subduction and expulsed via the decollement back to the ocean (26, 27) (Table 1 and text S1). This subduction reflux of Li from the convergent margins is held constant over time at ~6 × 109 moles of Li per year (the present-day value), and its isotopic composition is set at a constant offset from contemporaneous seawater (δ7LiReflx = δ7LiSW – δ7LiSink). This estimate of subduction reflux terms is based on instantaneous steady-state balances, whereas on the real Earth there is a time delay of 50 to 100 million years between sediment deposition and crustal subduction. However, this flux of Li from MAAC and AOC breakdown has a minor influence on seawater Li mass and isotope budgets. (ii) Initially, for simplicity, we take the fluvial dissolved FRiv-Li as constant over the Cenozoic (this constraint is later dropped). Nevertheless, from the early Paleogene to the present, the river-borne total weathered Li flux (suspended + dissolved) has increased by ~300% as a result of increased orogeny. More important, increasingly incongruent chemical weathering of the continents has changed the dominant Li-bearing phase in rivers from dissolved Li in the Paleocene to suspended Li (in clays) today. This weathering-driven redistribution of riverine Li flux has changed the ratio of dissolved to suspended partitioning from 4:1 in Paleocene to 1:4 today. (iii) δ7LiHT is kept at the modern-day value of 8.3‰. (iv) FHT-Li is kept constant over the Cenozoic (28). (v) Fractionation of Li upon removal from seawater during reverse weathering (∆SW-SED = δ7LiSW − δ7LiSink ≈ 16‰) is kept constant because the processes of Li removal, whether sediment-hosted or by alteration of oceanic crust, likely has not changed over the Cenozoic. (vi) The removal flux of Li out of the ocean (FSink) has remained near steady state with the inputs and exhibits first-order removal kinetics with respect to the Li concentration of seawater (LiSW). Our interpretation of changes in δ7LiSW is based on the weathering and reverse-weathering processes that have the largest δ7Li fractionation factors and thus the highest likelihood to be the main drivers not only of ocean δ7LiSW change but also other changes in oceanic and atmospheric chemistry and climate driven by silicate weathering (29, 30) (text S4).

Our seawater Li isotope record is consistent with previous Neogene reconstructions from chemically cleaned planktonic foraminifera (15, 31) (fig. S7). Also, our two paleo-δ7LiSW records from individual species and from bulk foraminifera are indistinguishable from one another (fig. S8). This similarity implies an absence of biogenic fractionation (vital effects) in δ7LiSW recorded by different species of foraminifera (both extant and extinct), despite strong discrimination against Li by foraminifera [lithium distribution coefficient KDLi(Calcite–Seawater) = (Li/Ca)Calcite/(Li/Ca)Seawater ≈ 4.2 × 10−3 mol/mol] and large variations in observed (Li/Ca)Foram (fig. S6). Also, fossil foraminifera from different drill sites at different paleolatitudes and paleodepths and under markedly different calcite preservation states display no offset or memory effects or diagenetic resetting (text S3 and fig. S9). The absence of any systematic differences in foraminifera-based δ7LiSW values between samples for both individual species and bulk foraminifera samples of the same age from different drill sites demonstrates that within the resolution of the present record, dissolved Li in seawater was isotopically homogeneous, and cleaned bulk foraminifera can be used to reconstruct the long-term evolution of δ7LiSW. Our modern foraminifera δ7LiSW appear to average ~1‰ lighter than seawater (fig. S6).

Results and interpretations. Our paleo-δ7LiSW record exhibits a 9‰ rise during the past 60 million years, implying a shift in chemical weathering of continental rocks consistent with the seawater 87Sr/86Sr record (Fig. 3). The isotope record shows no change over the past 6 million years; thus, the Li content and δ7LiSW of modern seawater are presumed to be near steady state (Fig. 1). Although our Li isotope record is similar to the seawater Sr isotope history, the Cenozoic δ7LiSW does not exhibit the monotonically increasing trend of seawater 87Sr/86Sr. The rise in δ7LiSW during the Cenozoic is nonlinear, punctuated by transient flat steady states and quasi-linear rises that may coincide with major climatic and tectonic events (32). The history of δ7LiSW over the past 60 million years can be divided into periods of stepped rises (Fig. 4). From 52 to 47 million years ago (Ma), 35 to 31 Ma, and 14 to 6 Ma, the average rate of increase in δ7LiSW (∆δ7LiSW / ∆t) is ~ 0.4‰ ± 0.1‰ per million years (2σ). The net result is a 9‰ rise in δ7LiSW during the past 60 million years. We argue that this increase in δ7LiSW is caused primarily by increases in δ7LiRiv (text S4 and fig. S10) that drive increases in the isotopic value entering the sea. Within several residence times (several million years), the marine reverse-weathering removal sink must adjust to obtain a new balance. The mechanism for this clay formation feedback is likely linked to the aluminum source from the continents (clays) and changes in the lithium and magnesium concentrations of seawater (LiSW/MgSW).

Fig. 4

(A) The 5-point running mean of all δ7LiForam values from all foraminiferal Li isotope ratio analyses (red line) plotted according to their average age with 2σ uncertainty (the two parallel gray lines) (Fig. 3). The vertical green downward arrow marks the rapid drop in δ7Li across the K-Pg boundary. The horizontal and vertical purple arrows reflect the overall 9‰ rise in δ7LiSW during the Cenozoic. (B) The ~9‰ rise in δ7LiForam over the last 60 million years has been divided into four distinct periods of steady state (black horizontal lines, ∆δ7Li/∆t ≈ 0.0‰ per million years) and three periods of rapid increase in δ7LiSW (red lines, ∆δ7Li/∆t ≈ 0.4‰ per million years).

Seawater 87Sr/86Sr and 187Os/188Os are dependent on differences in isotopic compositions (no isotope fractionation) in their continental (radiogenic) and mantle (nonradiogenic) sources to the ocean (Fig. 3) (3235). However, weathering of continental carbonates and redissolution of marine carbonates (for Sr) and weathering of organic-rich black shales and the rain of cosmic dust (for Os) complicate the uplift and continental runoff connections because these elements are not hosted solely in silicates, in contrast to Li (2, 36, 37). Increased silicate weathering during the Cenozoic has also been suggested from the records of seawater δ44/40Ca (38) and Sr/Ca (39), but both suffer from nonsilicate sources and sinks. The sulfur isotope history of seawater (40), an ocean redox record of S burial in marine sediments, suggests an oceanic anoxic event near the Paleocene-Eocene Thermal Maximum best explained as an increase in sediment sulfide burial, but is mute on changes in tectonics and weathering. Lithium is unique in tagging processes involving silicates, which is the key to carbon dioxide consumption during weathering and thus the connection between uplift tectonics and climate (15, 29, 41, 42).

The Paleocene-Eocene δ7LiSW minimum. Isotope and mass balance estimates for the mid-Paleocene δ7LiSW minimum (~22‰), based on our understanding of the modern-day oceanic Li cycle (Fig. 1B), predict an extremely light δ7LiRiv (~2 to 6‰) reflecting near-congruent weathering of the UCC regardless of secular changes in the Li river flux (text S4 and fig. S10). The early Cenozoic hothouse climate—with high sea levels, swamp continents, and cation-depleted peneplained continents dominated by neotropical congruent weathering and high weathering intensity under hot rainy conditions—is consistent with this interpretation. Low-latitude rivers during this period probably carried sparse suspended material or secondary clays. Thus, the cations delivered to the ocean were mostly in the dissolved load. The implication is that over the Cenozoic the proportion of dissolved Li flux to the ocean decreased (while clay Li increased) and its isotopic value (δ7LiRiv) increased as the chemical weathering regime of the continents became more weathering limited because of tectonic uplift in the rain belts (e.g., the Himalayas). The interpretation of the Mg/Ca change in seawater, which implies a slowdown in HT fluxes with time, is not inconsistent with the direction of change of our Li isotope record, but the magnitude of HT changes required seems unreasonable (text S4).

The δ7LiSW minimum in the early Cenozoic may have also been influenced by the northward migration of India carrying the recently erupted Large Igneous Province (LIP) Deccan Traps across the equatorial rain belt (43, 44). LIPs probably have a Li concentration and δ7Li composition similar to MORB (6 ppm; 3.7‰). Because basalts are aluminum-deficient relative to granites, they weather very rapidly (45) and highly congruently (few secondary cation-bearing clays to fractionate Li isotopes) (12). Thus rapid chemical dissolution of the Deccan basalts during the 10-million-year equatorial passage in a geological period of much more intense equatorial rainfall (46, 47) may have played a part in depressing δ7LiSW during the Oligocene. The subaqueous extent of the Deccan eruptions is unknown; thus, its impact on depressing δ7LiSW cannot be quantified. Similarly, eruption of the LIP North Atlantic Igneous Provinces (NAIP) may also have contributed to subaqueous dissolution of basalts (45) and had a similar impact on δ7LiSW during the Oligocene. It is also likely that quiescence of active mountain building in the rain belts and absence of uplift into orographic rain-catching mountains such as the Andes, Himalayas, and Rockies minimized the early Cenozoic occurrences of transport-limited weathering regimes that are today responsible for formation and transport of secondary clays to the sea and fractionation of the fluvial dissolved river Li flux toward heavy values.

During the Cenozoic climate cooling caused by increased CO2 drawdown, higher tectonic activity in the rain belts shifted the dominant continental weathering regime from transport-limited (congruent) to weathering-limited (incongruent), which led to heavier Li dissolved flux to the oceans and more weathered Li carried in clays (14). Thus, the similarity of our δ7LiSW record with the global ocean bottom water δ18O (temperature and ice volume) record is perhaps not accidental (fig. S11). This interpretation suggests that the rises in δ7LiSW punctuated with stable plateaus every 10 million years or so since the late Eocene might reflect periods of active uplift and denudation followed by periods of tectonic inactivity, at least in the low-latitude rain belts where most continental chemical weathering is expected to occur (Fig. 3). This scenario of Cenozoic continental silicate weathering is different from a monotonically increasing continental chemical weathering regime after initiation of the Himalayan uplift as suggested by the seawater Sr isotope history (32, 33). Coupled Sr and Li isotope models may lead to a better understanding of secular changes in chemical weathering of continental silicates and carbonates and the importance of cation-bearing clay fluxes to the sea in the CO2-consumption term of climate models (41).

The Cretaceous-Paleogene boundary (KTB) δ7LiSW event. An abrupt 5‰ drop in δ7LiSW occurs in less than 500,000 years across the Cretaceous-Paleogene (Tertiary) boundary (K-Pg or KTB), simultaneous with the seawater iridium and osmium isotope spikes (Fig. 3) (44). This rapid drop in δ7LiSW must be due to a large instantaneous delivery of isotopically light Li to the sea comparable to the Li content of the entire Cretaceous ocean. It is probably not due to fast addition of Li to seawater from congruent weathering of freshly erupted Deccan Traps (δ7LiBasalt ≈ 3.7‰) (43). Given the inventory of Li in seawater today (~3.4 × 1016 moles) and the possibility that this may have been higher in the Cretaceous, the Li mass of the Chicxulub bolide (~10 km diameter; ~9 × 1011 moles of Li) carrying chondritic δ7Li (~2‰) or even the 200-km impact crater itself (~4.5 × 1014 moles of Li) are insufficient Li (48, 49), even if the chondrite or the granite/gneiss basement of the impact crater were presumed to be instantly pulverized and dissolved congruently into the ocean. The cause of this large fast δ7LiSW drop across the K-Pg boundary remains enigmatic. We are working on a hypothesis that suggests that this drop in δ7LiSW exactly at the K-Pg boundary might be due to massive continental denudation and acid rain weathering of continental soils that were partially incinerated and deforested by the impact aftermath and washed into the sea.

Conclusions. With increased mountain building and changes in continental silicate chemical weathering regimes through the Cenozoic, the isotopic signature of riverine dissolved δ7Li increased from about 3‰ to 23‰. The partitioning of river-borne Li shifted from dissolved Li to clay Li as a result of increasingly incongruent weathering of silicate rocks and secondary clay formation. Overall, this scenario requires that the uplift- and weathering-driven cooling of the climate shifted the global weathering pattern from transport-limited to weathering-limited, increasing secondary clay mineral formation during weathering. As a result of preferential retention of 6Li by secondary clays, the δ7Li of river-borne Li became progressively heavier, driving seawater to its present heavy value. An increase in total Li weathered and delivered to the sea (clay Li plus dissolved Li), if extended to the weathering intensity of the major igneous cations stored in the sea (Na, K, and Mg), suggests faster CO2 drawdown due to more rapid weathering rates. Direct quantification of the influences of forward chemical weathering of continental silicate rocks and reverse weathering of marine silicates on δ7LiSW might provide alternative estimates of atmospheric CO2 consumption by the silicate weathering and reverse-weathering cycles (42).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S11

Table S1

References (55157)

References and Notes

  1. See supplementary material on Science Online.
  2. Acknowledgments: We thank the U.S. National Science Foundation (MG&G), American Chemical Society (Petroleum Research Fund), and The Francis Eppes Society of Florida State University for providing financial support. We sincerely thank S. Clemens, D. Hodell, E. Hathorne, E. Martin, E. Thomas, W. Landing, and H. Spero for providing samples. The manuscript greatly benefited from the constructive reviews of M. Bender, V. Salters, M. Humayun, A. Paytan, P. Pogge von Strandmann, and an anonymous reviewer. We are very thankful to B. Peucker-Ehrenrbrink and G. Ravizza for providing the osmium isotope data from GTS 2012 (in press). We also thank H. Elderfield, I.N. McCave, and A. Galy for helpful discussions. Tabulated data are provided on Science Online.
View Abstract

Navigate This Article