Mobilization of lithospheric mantle carbon during the Palaeocene-Eocene thermal maximum

The early Cenozoic exhibited profound environmental change inﬂuenced by plume magmatism, continental breakup, and opening of the North Atlantic Ocean. Global warming culminated in the transient (170 thousand year, kyr) hyperthermal event, the Palaeocene-Eocene thermal maximum (PETM) 56 million years ago (Ma). Although sedimentary methane release has been proposed as a trigger, recent studies have implicated carbon dioxide (CO 2 ) emissions from the coeval North Atlantic igneous province (NAIP). However, we calculate that volcanic outgassing from mid-ocean ridges and large igneous provinces associated with the NAIP yields only one-ﬁfth of the carbon required to trigger the PETM. Rather, we show that volcanic sequences spanning the rift-to-drift phase of the NAIP exhibit a sudden and ∼ 220-kyr-long intensiﬁcation of volcanism coincident with the PETM, and driven by substantial melting of the sub-continental lithospheric mantle (SCLM). Critically, the SCLM is enriched in metasomatic carbonates and is a major carbon reservoir. We propose that the coincidence of the Iceland plume and emerging asthenospheric upwelling disrupted the SCLM and caused massive mobilization of this deep carbon. Our melting models and coupled tectonic–geochemical simulations indicate the release of > 10 4 gigatons of carbon, which is su ﬃ cient to drive PETM warming. Our


30˚W
15˚W 08  the North Atlantic ridge peaked after, not during, early Eocene 49 hyperthermals (Fig. 1d). Therefore, some other major, but tran- 50 sient, source of volcanic carbon appears to be required if the 51 volcanic outgassing hypothesis 14,16 is correct. 52 We have investigated several volcanic sequences spanning 53 the Palaeocene-Eocene boundary (Fig. 1c). The Deep Sea 54 Drilling Project Leg 81 Site 555 lies on the Rockall Plateau 55 (Fig. 1a), near the proto-North Atlantic ridge (Fig. 1b). Here, 56 Phase 1 volcanism 21 (Fig. 2a) is coeval with the Milne Land 57 basalts in East Greenland and the Middle to Upper Series lavas 58 in the Faroe Islands 1 (Fig. 2b-c). In the Rockall sequence, we 59 found a sharp increase in the frequency of volcanic tuffs just be- 60 low the Palaeocene-Eocene boundary (Fig. 2a) (Methods). The 61 PETM is defined by δ 13 C, however the volcanostratigraphy in 62 our study area is not conducive to developing a high-resolution 63 carbon isotope stratigraphy (Methods). Therefore, we rely on a 64 combination of radiometric, magnetostratigraphic and paleon- 65 tological age constraints, in addition to well-defined sediment 66 accumulation rate estimates (Methods). Mudstones interbedded 67 with the uppermost tuffs contain the dinoflagellate cyst, Apec- 68 todinium augustum 27 , which is biostratigraphically diagnostic 69 of the PETM as it signifies a sudden prevalence of tropical sea- 70 surface temperatures 28 . Based on sedimentation rates (50 cm 71 ka −1 ) 4 , this volcanic flare-up lasted for 171-213 kyr, similar to 72 the duration of the coeval PETM 11,15 , and was followed by a 73 sharp decline in volcanism 21 . The tuffs exhibit wide composi- 74 tional diversity from basanites to dacites (Supplementary Figs. 75 1, 2a-b), an increased range of magnesium number (Mg#; to a 76 maximum observed value of 63), and a marked shift to highly 77 negative ǫNd signatures at ca. 56.03 Ma (Fig. 2a). This activity 78 signals a step change in magmatic processes and volcanic un- 79 rest along the ridge, as recorded across a major area of the NAIP 80 (>130,000 km 2 ) 1,29 . Sampled PETM-age tuffs are enriched in 81 Rb and Ba and depleted in Nb and Sr ( Supplementary Fig. 2c), 82 and have similar compositions to some of the lowermost ('neg- 83 ative ash series') tuffs of the Danish Basin, which likely derive 84 from nearby volcanoes along the continental shelf 29 . The geo- 85 chemical similarities between tuffs from Rockall and the Dan- 86 ish Basin is consistent with paleogeography ( Fig. 1b), palyno-87 logical constraints 27 and the stratigraphic position of these tuffs 88 towards the end of the PETM elsewhere in the NAIP 6 .  Nansen Fjord Formation were emplaced subaerially near the ridge axis (Fig. 1). The Greenland are included (Hold with Hope) 34 (Fig. 1b), which 155 have an upper age of ca. 57 Ma (C25n-C24r) 10 . These lavas ex- Ridge + low LIP prod. (S1) S1 + 4% melting S1 + 5% melting

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s12345-111-2222-3.

366
Recoveries of JA-2 and BRR-1 are shown in Supplementary 367 Table 5.  straints at this site (Fig. 1a). The PETM as geochemically de- where C 0 is the initial concentration of some element in the where ol is olivine, opx is orthopyroxene, sp is spinel, cpx is 499 clinopyroxene and gt is garnet.  we assumed a PETM duration of 170 kyr (ref. 15 ), which is sup-518 ported by recent astrochronological solutions 11 .

519
We use Beta distributions to represent uncertainty in the pa-520 rameters (Supplementary Table 8 where µ = the mean, σ = standard deviation, and σ 2 = the vari-526 ance. To estimate Beta distribution parameters, we use best esti-527 mates (from published data and observations) of the minimum, 528 mean and maximum values for each variable (Supplementary   529   Table 8, and discussed below), and apply these to equations 6 530 and 7 above (using re-scaled values for µ and σ). For simplic-531 ity we assume that the standard deviation for a given variable is where W COcean = total weight of C released from the ocean crust 56-55 Ma (Fig. 1d), assuming a plausible range of thicknesses 594 and melting widths for this layer (Supplementary Table 8).

595
The precise thickness of the c-SCLM of the North Atlantic

Data availability
All data generated or analysed during this study are provided in the online version of this article and in Supplementary Tables  1-8.

Code availability
The numerical modelling codes associated with this paper are available from the corresponding author (Thomas.Gernon@noc.soton.ac.uk) upon reasonable request. The map in Fig. 1a was plotted with open software GMT (under a GNU Lesser General Public License), and the map in Fig. 1b was plotted with open software GPlates (licensed for distribution under a GNU General Public License).
M.P. provided support with geochemical analysis and interpretation. G.F. carried out the melt modelling. A.M. calculated the seafloor production rates and provided support with GPlates and pyGPlates. T.G. wrote the manuscript with input from all co-authors. Figure 1 Early

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryInformationNGS20210300519.pdf