Barium content of Archaean continental crust reveals the onset of subduction was not global

Tonalite–trondhjemite–granodiorite (TTG) suites dominate Archaean continental crust and are generally interpreted as partial melts of hydrated basaltic protoliths. However, their diverse compositional characteristics (e.g., Sr/Y, La/Yb, Nb/Ta) have led to competing models for their petrogenesis and for the geodynamic regime of early Earth, including melting of subducted slabs, melting of thickened mafic lower crust, and vertical tectonic processes such as drips and sagduction. Conventional indicators like Sr/Y and La/Yb, often used to infer melting depth (LP/MP/HP TTG), can be significantly modified by processes such as fluid-present melting, crustal assimilation, fractional crystallization, and plagioclase accumulation, casting doubt on their reliability as tectonic proxies. This study asks whether barium (Ba), which is difficult to increase after melt generation in the source, can serve as a robust proxy to diagnose the geodynamic setting of TTG formation, thereby constraining when subduction and global plate tectonics began on Earth.

Prior work shows TTG suites are products of partial melting of amphibolite/eclogite and resemble fluid-absent melts of enriched hydrated basaltic sources. Yet experiments and modeling have struggled to reproduce high-MgO TTG solely by basalt melting, suggesting assimilation of MgO-rich materials or other hybridization processes. Numerous studies document that Sr/Y and La/Yb can be elevated by processes unrelated to high-pressure melting (e.g., plagioclase accumulation, hornblende fractionation, fluid-present melting consuming plagioclase at low P–T), and TTG reservoirs can behave as mushy mid-crustal systems. Geodynamically, Archean crustal evolution may have involved both vertical (stagnant-lid style, drips, domes/keels) and later horizontal (mobile-lid, subduction) tectonics. Ba has been proposed as less susceptible to modification by crystal accumulation and by assimilation of MgO-rich rocks; sediment assimilation would raise LILEs including K and Rb alongside Ba, yet many high-Ba TTG have low K2O/Na2O and low Rb, decoupling Ba from other LILE and arguing against sediment assimilation as a general cause of Ba enrichment. Enriched Archean tholeiitic basalts (Th/Nb > 0.1) have been identified as plausible TTG sources and are moderately LILE-enriched, but their primary Ba variability is insufficient to explain the wide Ba range in TTG, motivating P–T–X modeling focused on Ba behavior.

The authors performed petrological and trace element modeling to predict melt compositions, especially Ba, from an average enriched Archean tholeiitic basalt source over geotherms representative of Archean settings: hot subduction (low geothermal gradient ~450 °C GPa⁻¹), drip/vertical tectonics (high gradient ~900 °C GPa⁻¹), and cold subduction (~250 °C GPa⁻¹). They modeled both fluid-absent and fluid-present melting; for fluid-present conditions, an H2O content of 7 wt.% was used to ensure melting occurred with excess fluid. Phase equilibria were calculated using THERMOCALC v3.45 with the ds62 dataset in the NCKFMASHTO–Fe2O3 system, employing appropriate a–x models for amphibole, augite, metabasite melt, orthopyroxene, garnet, biotite, plagioclase, K-feldspar, ilmenite, and magnetite, with quartz and rutile as pure phases. Accessory mineral retention (zircon, apatite) was handled via solubility models using published expressions and stoichiometries to determine Zr and P partitioning in residues as a function of melt fraction and P–T. Trace elements were modeled with the batch melting equation Cmelt/Csource = 1/[D + F(1−D)], where the bulk partition coefficient D is the sum of mineral–melt partition coefficients (Kd) weighted by modal mineral proportions from phase equilibria, and F is the melt fraction. Mineral–melt Kd values were taken from the literature. The source composition for enriched Archean basalts included an average Ba of 107 ppm (median 62 ppm), with ~90% of data <160 ppm, constraining initial Ba availability. To test for secular changes indicative of tectonic transitions, the authors compiled global TTG Ba data (log-transformed due to lognormal distribution) and applied a Bayesian change-point detection algorithm (conjugate partitioned recursion) to identify statistically significant step changes globally and within individual cratons. They also compared the timing of Ba step changes by craton to the ages of the oldest rocks preserved on those cratons using degree-2 polynomial regression.

- Only low geothermal gradients characteristic of hot subduction settings generate Ba-rich TTG melts at low degrees of partial melting (~7–9 mol.%); higher gradients (e.g., drip settings) do not yield similarly Ba-enriched early melts. Fluid-absent vs fluid-present conditions affect Sr/Y magnitudes but not the basic result that Ba-rich TTG are linked to low geotherms. - Fluid-present conditions greatly increase melt productivity along low geotherms (~50 vol.% melt possible if sufficiently hydrated) compared to water-limited conditions (~20 vol.%). In contrast, cold subduction geotherms (~250 °C GPa⁻¹) produce ≤4 vol.% melt, insufficient for extraction, effectively ruling out cold subduction as the TTG source. - Modeled melts match experimental trends: fluid-absent low-gradient melts yield the highest Sr/Y (up to ~2× fluid-present and ~10× higher than high-gradient melts), while all melts have low Ni (<70 ppm) that decreases with increasing Ba, consistent with natural TTG. Plagioclase accumulation and fractional crystallization cannot raise Ba (plagioclase–melt DBa ≈ 1), and biotite accumulation would increase both Ba and Ni, which is not observed in high-Ba TTG (Ni commonly <30 ppm). - The primary Ba of enriched Archean tholeiitic basalts (avg 107 ppm; median 62 ppm; ~90% <160 ppm) is too low and insufficiently variable to explain the observed high Ba in many TTG without invoking specific P–T melting conditions. - Global TTG Ba data show a lognormal distribution and reveal three statistically significant positive step changes in log(Ba) at approximately 3.7, 3.1, and 2.8 Ga, indicating protracted, irreversible shifts in global geodynamics toward conditions favoring Ba-rich TTG (i.e., subduction-like low geotherms). - Ba step-change timings are highly diachronous by craton: Slave and Kaapvaal show high Ba as early as ~4.0 and 3.5 Ga (with a pronounced ~3.5 Ga Ba anomaly in Slave, consistent with a seismically imaged slab at ~3.5 Ga), many cratons shift at ~3.2–3.0 Ga (e.g., Baltica, Amazonia, Pilbara), and some as late as ~2.7 Ga (Superior). This supports regional onset of subduction propagating toward global operation. - There is a strong correlation between the age of the Ba step change and the age of the oldest rocks preserved on each craton (r² = 0.9), suggesting a link between rock preservation and subsequent subduction initiation. - The study constrains a lower bound for global subduction near ~2.0 Ga from independent lines and provides an upper bound for global completeness sometime after ~2.7 Ga, consistent with independent estimates of mantle potential temperature and thermal history models.

The modeling demonstrates that Ba enrichment in TTG primary melts is a sensitive indicator of low geothermal gradient conditions associated with hot subduction, whereas vertical, high-gradient Archean tectonics (e.g., drips, sagduction) do not produce similarly Ba-rich melts at low melt fractions. Because Ba is minimally affected by fractional crystallization, crystal accumulation, or assimilation of MgO-rich rocks—and because sediment assimilation would increase K and Rb alongside Ba, which is generally not observed—Ba serves as a relatively robust proxy for the tectonic setting of TTG formation. Applying this proxy to the global TTG record reveals diachronous increases in Ba through the Archean, implying that subduction initiated regionally and propagated over time rather than beginning synchronously worldwide. The observed step changes at ~3.7, 3.1, and 2.8 Ga, and their diachronous expression across cratons, support a subduction propagation model transitioning from a stagnant-lid regime to plate tectonics. The strong correlation (r² = 0.9) between Ba step-change ages and the oldest rock ages by craton suggests that conditions enabling rock preservation are linked to, or precede, subduction initiation. These inferences are consistent with independent constraints on Earth’s mantle potential temperature and thermal evolution, which allow a broad time window for the emergence of plate tectonics and require geochemical proxies like Ba to refine the timeline. The authors emphasize the importance of evaluating Ba within a broader multi-proxy framework given ongoing debates about the timing and nature of early plate tectonics.

This study introduces and validates Ba content in TTG suites as a proxy for diagnosing the geodynamic setting of Archean crust formation. Petrological modeling shows that only low geothermal gradients associated with hot subduction generate Ba-rich TTG at low melt fractions, while other settings fail to do so. Applying a Bayesian change-point analysis to global TTG Ba data reveals diachronous, stepwise increases in Ba beginning as early as 4.0–3.5 Ga in some cratons and continuing through ~3.2–3.0 Ga to as late as ~2.7 Ga in others, implying regional subduction initiation that propagated to global plate tectonics sometime after 2.7 Ga (with ~2.0 Ga as a lower bound from independent considerations). The results support a subduction propagation model and align with Earth’s thermal history estimates. The Ba proxy provides a framework for a future multi-proxy approach to constrain the onset and evolution of global subduction and plate tectonics. Potential future directions include expanding high-pressure, fluid-present experimental datasets to better calibrate Ba behavior, integrating Ba with other geodynamic proxies across more cratons (including younger or less-studied ones), and refining statistical approaches to account for preservation biases in the Archean rock record.

- Experimental constraints for fluid-present melting at high pressure are relatively sparse due to kinetic limitations, necessitating reliance on thermodynamic modeling for some conditions. - The modeling assumes an average enriched Archean tholeiitic basalt source; while supported by compilations, unrecognized regional source heterogeneities could influence predicted melt Ba. - Sediment assimilation is argued against by decoupling of Ba from K and Rb in many datasets, but localized exceptions cannot be entirely excluded. - The TTG rock record is unevenly preserved across cratons, and diachronous Ba step changes may be affected by preservation and sampling biases. - Broader geodynamic inferences depend on integrating Ba with other proxies; uncertainties in Earth’s thermal history models and observational constraints limit precise global synchronization of the onset of plate tectonics.