Clay hydroxyl isotopes show an enhanced hydrologic cycle during the Paleocene-Eocene Thermal Maximum

The PETM (~55.9 Ma) was a period of rapid global warming linked to a large carbon release, marked by a ~3–4‰ negative carbon isotope excursion (CIE) and widespread carbonate dissolution. Global mean temperatures rose by ~4–5 °C and oceans experienced deoxygenation and acidification. Multiple lines of evidence indicate significant perturbations to the hydrologic cycle, but regional responses varied greatly. Prior palynological and biomarker studies suggest sustained increases in terrestrial runoff in some regions (e.g., Venezuela, Arctic Spitsbergen, New Zealand, North Sea, Arctic Ocean), while other areas show increased aridity or enhanced seasonality and extreme rainfall. Despite these observations, quantitative constraints on the timing and magnitude of hydrologic change relative to the CIE remain limited due to a lack of suitable proxies. This study aims to develop and apply a new clay hydroxyl isotope proxy to resolve hydrologic variability across the PETM in the North Sea Basin and to test whether hydrologic intensification coincided with, preceded, or lagged carbon-cycle perturbations.

Previous work documents regional hydrologic responses during the PETM: increases in terrestrial runoff and low-salinity indicators at high and low latitudes; heightened aridity or seasonality in subtropical to mid-latitude regions (e.g., southern Rocky Mountains, Spanish Pyrenees, Normandy). The origin of increased kaolinite during the PETM has been debated: early interpretations favored contemporaneous formation under warm, humid conditions (enhanced weathering and pedogenesis), whereas later studies argued for inheritance from pre-existing, slowly formed regolith (kaolinization on Myr timescales). However, transformations of 2:1 phyllosilicates to kaolinite in soils can occur on much shorter timescales (100 days at 150 °C in lab; significant kaolinization over 5–10 ka in tropical conditions). Prior bulk clay δ18O studies (e.g., Bass River, New Jersey) suggested that δ18Obulk variability largely reflects changing mineral proportions (kaolinite vs smectite) rather than within-mineral isotopic changes. Literature also notes the need for better hydrologic proxies and fractionation constraints for clay hydroxyl isotopes to reconstruct paleoprecipitation isotopic composition.

• Study site and samples: 22 separated <4 μm fractions from the Sele Formation, well 22/10a-4 (57°44′8.47″N; 1°50′26.59″E; central North Sea; water depth up to ~500 m) spanning the PETM. Bulk organic δ13C data for the same section define CIE onset between 2614.3 and 2613.5 m, with a large drop to −30±1‰ at 2612.0 m.
• Clay mineralogy: Assemblage includes kaolinite, illite, illite-smectite (I-S), and chlorite. Pre-PETM base: illite (13–23 wt%), I-S (44–54 wt%), chlorite (7–10 wt%), low kaolinite (1–4 wt%). Toward and just after CIE onset, kaolinite increases up to 44 wt% while I-S decreases to 19 wt%; then mineral proportions vary through the section.
• Hydroxyl isotope measurements (DTIA): Samples were carbonate-removed with sodium acetate/acetic acid buffer, dried at 40 °C. Differential Thermal Isotope Analysis (DTIA) employed controlled heating (40 °C/min; isothermal steps at 250 and 390 °C) to isolate hydroxyl water. δ2H and δ18O of the hydroxyl peak were measured using a Picarro L2130i analyzer. Three to four replicate DTIA runs per depth (two at 2610.82 m). Analytical uncertainties are ±1σ of replicates.
• Determination of hydroxyl peak and dehydroxylation: A preliminary 5 °C/min ramp to 1030 °C characterized dehydration/dehydroxylation peaks using standards (kaolinite, rectorite, chlorite) and a PETM sample; PETM samples show no high-T chlorite dehydroxylation peak.
• Bulk oxygen isotope measurements: Carbonates removed with 1 M HCl; organics removed with 30% H2O2; samples rinsed, dried, pre-fluorinated with F2 to remove hydration water, then laser fluorinated (CO2 laser) in F2 to liberate O2. O2 purified and measured on Thermo MAT253 IRMS; δ18O normalized to VSMOW-SLAP using VSMOW-2 and SLAP-2 standards. Uncertainty from replicate aliquots. Treatments confirmed negligible impact on kaolinite δ18O.
• Estimation of hydroxyl source water δ2H (δ2HOH-SW): Used mineral-specific hydrogen isotope fractionation factors at 30 °C for kaolinite (1000 ln α = −31.6) and illite-smectite (−54.7). Calculated hydroxyl water contributions from each mineral based on assemblage proportions and hydroxyl contents, assuming I-S composition does not affect α but changes hydroxyl H contribution. Computed δ2HOH-SW via weighted correction; also calculated 100% kaolinite and 100% I-S end-member scenarios to bracket uncertainty. No reliable hydroxyl oxygen fractionation factors exist; thus, δ18OOH-SW not estimated. Quartz presence (~up to 31%) complicates bulk oxygen partitioning.
• Statistics: Correlations (ordinary least squares) between mineral oxygen/hydrogen contributions and δ18Obulk, δ18OOH, δ2HOH were assessed; significant r² values reported where p<0.05. Data and analysis scripts are provided in Supplementary materials.

• Hydroxyl hydrogen isotopes (δ2HOH): Values decrease gradually prior to the PETM and then show an abrupt ~8‰ VSMOW decrease at the CIE onset, coincident with increased kaolinite and low-salinity dinoflagellate cyst abundances. δ2HOH returns to pre-PETM values by 2606.4 m, well before the CIE ends. Overall δ2HOH range: 34.1±0.3‰ (27.6±0.3‰ excluding a one-point minimum).
• Hydroxyl oxygen isotopes (δ18OOH): Average 3.2±0.3‰ VSMOW at base, increase to 10.7±0.2‰ (excluding a one-point minimum at 1.38±0.13‰ VSMOW associated with kaolinite maximum and I-S minimum). At CIE onset, δ18OOH decreases by 10.2±0.2‰ to 0.42±0.04‰ and remains between −0.7±0.2‰ and 3.7±0.2‰ thereafter. δ18OOH shows weak to moderate correlations with mineral oxygen contributions (e.g., kaolinite r²=0.46; p<0.01; n=22).
• Bulk clay δ18O (δ18Obulk): Positively correlated with I-S oxygen content (r²=0.30; p=0.01; n=22) and strongly negatively correlated with kaolinite oxygen content (r=0.83; p<0.01; n=22). Variability magnitude is smaller (3.7±0.8‰) than δ18OOH (11.3±0.3‰). Weak correlation between δ18Obulk and δ18OOH (r²=0.45; p<0.01; n=22). δ18Obulk exhibits complex pre- and post-onset fluctuations, partially recovering after the CIE onset.
• Hydroxyl source water δ2H (δ2HOH-SW): Reconstruction closely tracks δ2HOH (r²=0.90; p<0.01; n=22), with absolute values from −2.5 to −50.9‰ VSMOW and greater variability (48.5‰) than measured δ2HOH due to mineral fractionation corrections.
• Mineralogical controls: δ18Obulk strongly reflects kaolinite proportion (r²=0.83), implying near-constant kaolinite δ18O and compositional control on bulk values. By contrast, δ2HOH and δ18OOH show poor correlations with kaolinite proportion, indicating hydroxyl isotopes are less influenced by mineral mixing and better record hydrologic signals.
• Process inference: The δ2HOH decrease and its covariance with runoff indicators support intensified precipitation/runoff at PETM onset. Clay formation/alteration during the PETM (rapid kaolinization of 2:1 clays in soils) is favored over erosion of pre-existing kaolinite to explain hydroxyl isotope signals. TEX86 suggests ≥~10 °C SST rise at onset, consistent with enhanced convection and possible tropical cyclone activity driving low δ18O/δ2H rainfall.
• Timing: The δ2HOH and δ18OOH decreases begin prior to or at the CIE onset and return to baseline before CIE termination, indicating hydrologic intensification was short-lived relative to carbon-cycle perturbation.

The hydroxyl isotope records demonstrate that the North Sea Basin experienced an intensified hydrologic cycle at the PETM onset, characterized by more negative rainfall isotopic composition consistent with higher precipitation amounts (amount effect) and/or high-intensity events. The strong covariance between δ2HOH and low-salinity-tolerant dinoflagellate cyst abundances corroborates increased freshwater input and runoff. Weak relationships between hydroxyl isotope values and clay mineral proportions, together with the expected direction of temperature effects on fractionation (which would increase, not decrease, δ values), indicate that source water isotope changes dominate the signal rather than temperature or compositional artifacts. The pre-CIE trend toward more negative hydroxyl isotopes suggests hydrologic reorganization began before major carbon release, possibly linked to sea-level lowering or regional climate shifts, while the early return to baseline during the CIE implies a transient hydrologic response. Comparisons with other North Sea and Danish sections show broadly similar short-lived hydrologic excursions, though stratigraphic resolution and local factors modulate proxy responses. The mineralogical and isotopic evidence supports rapid in-soil kaolinization during the PETM, adding hydroxyl groups that record meteoric water isotopic composition, rather than dominant input of pre-formed kaolinite with fixed isotopic signatures.

By applying differential thermal isotope analysis to measure δ2H and δ18O of clay hydroxyl groups, alongside bulk clay δ18O, this study provides new constraints on PETM hydrology in the North Sea Basin. The records reveal a sharp ~8‰ decrease in hydroxyl isotopes at the CIE onset, consistent with intensified precipitation and runoff—potentially including enhanced tropical cyclone activity—followed by an early return to baseline, indicating hydrologic changes were shorter in duration than the carbon-cycle perturbation. The data, together with mineralogical trends, imply increased kaolinization and silicate weathering under warm, humid conditions. The work establishes clay hydroxyl isotopes as a promising proxy for paleohydrology and clay provenance, provided mineralogical effects are carefully evaluated. Future research should: (1) refine hydrogen isotopic fractionation factors for diverse clay minerals under environmentally relevant conditions; (2) improve methods to remove organic matter without altering hydroxyl isotopes; (3) expand measurements to other PETM sections and compare with isotopically enabled climate model simulations to enable quantitative reconstructions of precipitation amount and seasonality.

• Fractionation constraints: Limited, composition-independent hydrogen isotope fractionation factors (αH–OH) are available only for kaolinite and illite-smectite; other hydroxyl-bearing phases may contribute without well-constrained α values, introducing uncertainty in δ2HOH-SW reconstructions.
• Mineralogical assumptions: Source water reconstructions assume hydroxyl contributions from kaolinite and illite-smectite only and typical mineral stoichiometries; variability in interlayer chemistry or I-S illite:smectite ratio could affect results.
• Post-depositional effects: Although significant hydrogen exchange in deep-sea buried clays is considered unlikely outside the <0.1 μm fraction over ~Myr timescales, some post-depositional isotopic exchange or marine bias cannot be entirely excluded.
• Organic contamination: Potential influence of residual organics on DTIA δ2HOH measurements is acknowledged; while considered minor, it adds uncertainty.
• Bulk oxygen inference: Presence of quartz of unknown provenance complicates interpretation of bulk δ18O; quartz was not separated from clays for bulk analyses.
• Spatial limitation: Results are from a single, albeit highly expanded, North Sea section; regional or global generalization requires additional sites.
• CIE stratigraphic complexity: The recorded CIE onset may be influenced by mixing of carbon sources and potential pre-onset excursions, complicating precise temporal alignment between hydrologic and carbon-cycle signals.