4.1 Sapphire formation in the lithospheric mantle
The corundum gem deposits (a brief introduction see Supplementary information) consist of ruby and sapphire deposits, and this study focuses on six Cenozoic magmatic sapphire deposits in eastern China: Muling, Kuandian, Changle, Luhe, Mingxi, and Penglai (details see Supplementary Dataset 1). In these deposits, sapphires are found in mantle xenoliths or as xenocrysts, not in the basaltic matrix. Previous works showed that sapphires did not crystallize in the basaltic magma but were trapped during magma upwelling26. They initially formed in the lithospheric P-T conditions of 790 − 1200°C and 0.85 − 2.5 GPa.
In these sapphire deposits, the sapphire gems have been widely discovered in mantle xenoliths of alkaline basalts26, which typically originate from the lithosphere-asthenosphere boundary (LAB)27, and are rarely found in the Mesozoic calc-alkaline basalts (lithospheric origin)8. This suggests that the sapphire primarily formed in the deep lithosphere. Additionally, the MORB-like 3He/4He and high 40Ar/36Ar isotopic ratios28 and the presence of multiple fluid inclusions26 in these sapphires indicate the involvement of asthenosphere-derived melts or fluids in their formation. Previous models did not explain why sapphires form at such deep lithosphere and how water influences their production28,29.
Our experiments show that sapphire forms when phlogopite interacts with water beyond 1.5 GPa, which equals > 50 km depth in the lithospheric mantle (Fig. 3). This reaction answers two key questions about the sapphire deposits: how do sapphires form in the lithospheric mantle, and why do these deposits only form in the Cenozoic?
The depth of sapphire formation, as indicated by this study, approximates the melting depth of Mesozoic calc-alkaline basalts8 and can only be brought to the surface by Cenozoic alkaline basaltic magma from LAB27. These sapphires are initially produced in the lithosphere, trapped by deep-origin magmas, and then transported to the surface. From a traditional viewpoint, an Al-saturated environment rarely occurs in the mantle. This study provides a relatively universal mechanism for sapphire formation in the lithospheric mantle (> 50 km depth) (Fig. 4).
The timescale of sapphire formation is controlled by water migration in the mantle. Before the Cenozoic era, LAB beneath eastern China was generally more than 90 km8, and the upward-moving water was first captured by pargasite30, which hindered the direct hydration of the lithospheric mantle. However, during the Cenozoic era, with the destruction of the cratonic lithosphere, LAB become less than 90 km, and thus pargasite breaks down. Abundant water can directly react with phlogopite in the lithospheric mantle to form sapphire. Phlogopite is frequently found in Paleozoic mantle xenoliths, even as phlogopitites (~ 100% phlogopite), while Cenozoic xenoliths lack phlogopite-bearing rocks31,32, possibly due to mantle metasomatism consumption.
Our hydrothermal diamond-anvil cell experiments of the “phlogopite + H2O” system at 500 − 950°C and 1.0 − 1.8 GPa simulate a typical mantle metasomatism process, essential to understanding sapphire formation. On the other hand, this finding highlights the role of P-T conditions in phlogopite dissolution and sapphire formation (Fig. 3). Thus, the presence of sapphire deposits provides unique information about hydration events.
4.2 Spatial and temporal constraints on hydration
The distribution and age of sapphire deposits provide spatial and temporal constraints on hydration events. The six Cenozoic sapphire deposits are all distributed along the eastern coast of China (Fig. 1), roughly parallel to the Pacific subduction zone. The horizontal distance between these deposits and the subduction trench ranges from 1200 to 1600 kilometers, suggesting that plate subduction has triggered the hydration of the entire mantle wedge. This big mantle wedge extends to a maximum depth of 410 km, with a latitudinal and longitudinal distance of 1600 km and 2000 km, respectively. Notably, Cenozoic basalts in Kuandian and Chifeng have similar petrologic features, approximate ages, and close latitude33,34. However, sapphires-bearing xenoliths are only reported in Kuandian and rarely in Chifeng. This is probably due to the deeper LAB beneath Chifeng33. Deeper LAB means less destruction of the lithospheric mantle, less pargasite breakdown, and thus less water release.
There are two timelines for these sapphire deposits: the ages of sapphire formation (Ts) represent when sapphires initially appeared at the lithospheric mantle, while the ages of sapphire-bearing host rocks (Th) record when those sapphires were transported to the surface28. Thus, sapphire formation cannot occur later than the eruption of associated alkaline basaltic magma (Th≤Ts). These basaltic magmas in eastern China erupted from ~ 19.2 Ma to Pleistocene (Th), which represents the deadlines of hydration events. Recent studies reported the accurate dating of zircon inclusions inside the sapphire crystals or associated zircon megacrysts26,35,36. However, whether these zircons’ ages can represent when sapphire formed remains controversial. The relatively high temperature of alkaline basaltic magmas may lead to Pb diffusion in zircons and reset the U-Pb ages37. However, the preservation of ancient zircons (e.g., ~ 400 Ma zircons inclusions) in Cenozoic sapphire deposits suggests the lack of age resetting in zircons26. A recent study36 suggests that the zircon inclusions trapped in Changle sapphires can be used as the age of sapphire formation (~ 15.5 Ma), which is consistent with the eruption age of the host basalts. This indicates that sapphires formed contemporary to the eruption of basaltic magma. This model can also be applied to Cenozoic magmatic sapphire deposits in Southeastern Asia and eastern Australia.
4.3 A globally applicable model?
Eastern China and Australia are both typical examples of the destruction of ancient craton lithosphere due to the subduction of the western Pacific Plate (Fig. 1; Extended Data Fig. S4, S5). Extensive magmatic sapphire deposits in Southeast Asia are also distributed parallel to the subduction zone. In these regions, sapphire deposits are all related to alkaline basalts and have the same metallogenic features26. All the sapphire deposits can be explained by the model presented in this study. These sapphire deposits provide constraints on the hydration of the lithospheric mantle beneath these areas, respectively.
Sapphire deposits in these three regions are all distributed along the coastlines. The deposits in Australia are located farther from the subduction trench (~ 3400 km on average) compared to those in China (~ 1400 km on average), while the sapphire deposits in Southeast Asia are closest to the subduction zone trenches (~ 650 km on average). Some of the early sapphires in eastern Australia may have been influenced by fluids from the subducting slab before 80 Ma. However, with the Tasman Sea spreading38, the impact of the subducting slab on eastern Australia has diminished. Seismic profiles reveal that stagnant slabs beneath the lithosphere are only present in northeastern Australia39. This suggests that the mantle transition zone or mantle plumes, rather than the slab itself, contribute to the water, needed for forming eastern and southeastern sapphire deposits. The shallow LAB beneath the three regions40 indicates the destruction of the thick pargasite layer, releasing fluids that interacted with phlogopite in the ancient craton lithosphere to form mantle sapphire.
The formation age of sapphire deposits (Th) in eastern Australia varies from 74.5 Ma to 2 Ma, while those in eastern China and Southeast Asia are concentrated in the Late Cenozoic. Moreover, Asian sapphires are limited to early-stage basalts in the locality, whereas Australian sapphires are found in basalts from all stages, implying that the lithospheric mantle in eastern China was hydrated during a short period, while that beneath eastern Australia experienced hydration in multiple stages throughout the Cenozoic.