High-pressure and high-temperature experiments were conducted to simulate melting of a hybrid cumulate lunar mantle. The experimental results show that intermediate to high-Ti lunar pyroclastic glasses (>6 wt% TiO 2 ) can be produced by partial melting of lunar cumulates. High-Ti basalts are generated when the ilmenite/clinopyroxene ratios in the lunar mantle cumulates are between 1/1 and 4/1, depending on the degree of melting. The presence of an urKREEP component in the mantle cumulate strongly influences Al 2 O 3 /CaO of the melts. The experiments provide strong evidence for the model that the compositional diversity of lunar basalts is a consequence of a gravitational overturn of the lunar interior after the lunar magma ocean had solidified. Ilmenite/clinopyroxene in the cumulate mantle, which generates high-Ti melts at partial melting, do not comprise the ratios in ilmenite-bearing cumulates (IBC), which crystallized after ∼90% solidification of the lunar magma ocean and indicate local accumulation of ilmenite in the overturned lunar mantle. However, to fully match the natural composition of the most primitive lunar samples, secondary processes such as assimilation are still required.

The basalt samples, which were returned from the Earth's Moon by astronauts during sample return missions, are chemically different from terrestrial basalts. Whilst terrestrial basalts usually contain less than 1 wt% of TiO 2 , the lunar rocks can contain more than >6 wt% of TiO 2 . The origin of these extraordinarily high TiO 2 contents has long been debated, and hence we conducted experiments at high pressures and temperatures to simulate melting of a heterogeneous lunar mantle. Our experimental results imply that the lunar high-Ti basalts can be produced by melting of a heterogeneous lunar mantle that contains high amounts of Ti-rich minerals, such as ilmenite or similar Fe and Ti-rich oxides, and silicate minerals, such as olivine, orthopyroxene, and clinopyroxenes. Our results support the hypothesis that the stratification of the lunar mantle after it had solidified was not preserved, suggesting a high dgree of heterogeneity within the mantle. Besides Ti, other major elements of our melts do not completely match the natural samples. This can be accounted for by interaction with the host rocks during the ascent of the magmas.