Introduction

Highly silicic magmatism was long thought to be exclusive to Earth. The extensive silicic magmatic rocks on Earth are compositionally diverse and their presence is a mark of crustal maturation1,2. In contrast, almost all volcanism on the Moon is basaltic, namely the flood basalts of the maria mostly on the lunar nearside3,4. The different magmatic features between the Moon and Earth have usually been ascribed to the absence of water and plate tectonics on the Moon5,6. Therefore, it has long been deemed that highly silicic crust or continent-like massifs do not exist on the stagnant-lid Moon.

However, with advances in remote sensing, several non-mare volcanic deposits have been observed on the Moon (Fig. 1), known as spectral “red spots” first recognized in 19727. The spots are characterized by elevated topography with steep slopes (domes, cones, etc.), high albedo, and strong absorption in the ultraviolet relative to the visible region8,9,10, with high thorium and low iron contents11,12. These features suggest derivation from viscous, evolved magmas with high silica contents, similar to terrestrial rhyolite domes12. Their high silica contents are further confirmed by the Diviner Lunar Radiometer Experiment, which shows that these volcanic areas mainly comprise quartz, silica-rich glass, and alkali feldspar13,14. High abundances of alkali feldspar and silica with few mafic materials also correspond to the high reflectance of these areas14,15. Moreover, a few small granite/felsite clasts (or fragments) were found in returned lunar samples16,17. These observations, taken together, provide robust evidence for the presence of highly silicic magmatism on the Moon, thus challenging conventional views of granite petrogenesis on Earth and offering new clues to the Moon’s thermal history and crustal evolution.

Fig. 1: Silicic volcanic constructs on the Moon.
figure 1

a Locations of silicic volcanic constructs on the Moon (red open circles), with the Procellarum KREEP Terrane outlined (yellow dashed line). The base image is from https://moon.nasa.gov/resources/127/lunar-near-side, published on October 20, 2017. b–e Up-close views of the topography of Mairan Domes, Gruithuisen Domes, Hansteen Alpha, and Lassell Massif. Images from the Lunar Reconnaissance Orbiter wide-angle camera and narrow-angle camera (credit: NASA/GSFC/Arizona State University/Lunar QuickMap).

Full size image

There are two leading hypotheses for the origin of lunar silicic rocks. The first one, silicate liquid immiscibility (SLI), involves near-complete (90–98%) magma differentiation and then segregation of the residual melt into two immiscible liquids, one being Fe-rich and Si-poor (ferrobasalt), and the other being rich in silica and alkalis (rhyolite)18,19. Evidence for SLI on the Moon includes distinct silicate melt inclusions in minerals18,20 and late-stage mesostasis in mare basalts21. This mechanism has been proposed to be responsible for the origin of the small granite clasts/fragments in returned lunar samples19,22,23. However, it is unlikely that the microscopic granitic segregations produced by SLI could have formed the massive silicic volcanic constructs (~5–40 km in diameter) such as the Gruithuisen Domes and Hansteen Alpha12,13,24 (Fig. 1). The silicic volcanic constructs (i.e., large-scale silicic magmatism) have crater-counting ages of ca. 4.0–3.6 billion years ago (Ga)25,26,27 that postdate the radiometric ages of returned granite clasts/fragments (4.38–3.88 Ga)28,29,30, further suggesting different origins between them. Moreover, high-field-strength elements such as thorium are preferentially partitioned into the Fe-rich, Si-poor liquid instead of the Si-rich liquid (the former having Th content ~2.5 times greater than the latter)31,32, in apparent contradiction to the high Th contents of lunar red spots12. Alternatively, a second hypothesis is crustal remelting12,13. This mechanism is widely applied to terrestrial granite petrogenesis33 and can explain the large volumes and high Th contents of lunar silicic volcanic constructs34. However, most lunar samples contain olivine3, the melting of which only produces limited silica-rich melts. Thus, what kind of lunar lithologies could generate granitic magmas and how remains enigmatic.

To interrogate the feasibility of the crustal remelting mechanism and potentially find out the source rock(s) of the silicic volcanic constructs on the Moon, we conducted phase equilibrium modeling to calculate the partial melting of lunar igneous rocks, including both highland rocks and mare basalts. Thermodynamic models35 used here are exclusively developed to model the partial melting of basaltic rocks, and thus, mineral assemblages and melt compositions as a function of melting conditions and source rocks can be well constrained36. Our work shows that large-scale, highly silicic lunar magmatism can be produced by remelting of KREEP basaltic crust driven by internal radiogenic heating or external impact bombardment.

Results

Phase equilibrium modeling of lunar igneous rocks

Lunar igneous rocks modeled here using THERMOCALC v3.47 (see “Methods”) include ferroan-anorthositic suite (FANs), Mg-suite, alkali-suite (excluding granite/felsite), KREEP basalts, and low-Ti [TiO2 ≤ 6 weight % (wt%)] and high-Ti (TiO2 > 6 wt%) mare basalts. Average bulk-rock compositions3,37 were used for phase equilibrium modeling (Supplementary Table 1). A series of pressure–temperature phase diagrams were calculated in the range of 800–1500 °C and 0–5 kbar (Fig. 2). Among these rocks, KREEP basalts (together with high-Ti basalts) and Mg-suite rocks have the lowest (840–960 °C) and highest (1130–1180 °C) solidus temperatures, respectively. At subsolidus conditions, lunar igneous rocks all have plagioclase, high-Ca and low-Ca pyroxenes, and ilmenite. K-feldspar is absent in FANs and Mg-suite rocks because of their low bulk K2O contents. Olivine occurs in most rocks but not in KREEP and high-Ti basalts; instead, they have silica-bearing subsolidus assemblages, in agreement with petrographic observations (see Supplementary Information for details), indicating that KREEP and high-Ti basalts are more silica-fertile than the other lithologies.

Fig. 2: Phase equilibrium modeling of lunar igneous rocks.
figure 2

a FANs rocks. b Mg-suite rocks. c Alkali-suite rocks (excluding granite/felsite). d KREEP basalts. e Low-Ti basalts. f High-Ti basalts. Black dashed curves represent the solidi, and brown dashed lines indicate the mole percent (mol%) of partial melts. g–l Diagrams contoured for melt SiO2 contents. Average bulk compositions of low-Ti and high-Ti basalts are from ref. 37, and the others are from ref. 3 (Supplementary Table 1). Calculated results are shown in Supplementary Data 1–6. Pl plagioclase. Aug augite (high-Ca pyroxene), Opx orthopyroxene (low-Ca pyroxene), Ol olivine, Qtz quartz, Ilm ilmenite, Ksp K-feldspar.

Full size image

Such fertility can be better understood in the olivine–anorthite–quartz pseudoternary diagram projected from diopside38 (Fig. 3). Because all stoichiometric plagioclase and pyroxene project along the anorthite–enstatite join, this join defines a thermal divide when plagioclase and pyroxene dominate the residue at low pressures. Bulk compositions on either side of the join have distinct mineral assemblages ranging from quartz-normative to olivine-normative types. Hence, this dividing line can separate lunar rocks into two groups: silica-excess on the right and silica-deficient on the left. KREEP and high-Ti basalts plot on the silica-excess side with the latter much closer to the join, while the other lithologies have average compositions on the silica-deficient side, consistent with our phase equilibrium modeling. Because partial melting paths (or mineral assemblages from subsolidus to liquidus) diverge greatly on opposite sides of the thermal divide38, silica-excess and silica-deficient lunar rocks are expected to produce different partial melts in terms of silica saturation.

Fig. 3: Lunar igneous rocks in the olivine–anorthite–quartz pseudoternary system projected from diopside.
figure 3

Pentagrams represent average bulk compositions of lunar igneous rocks (FANs rocks, purple; Mg-suite rocks, dark blue; Alkali-suite rocks, green; KREEP basalts, red; Low-Ti basalts, bronze; High-Ti basalts, orange) used in phase equilibrium modeling (Fig. 2). The frequency distributions of high-Ti and low-Ti basalts are shown based on kernel density estimation in the top plot and as probability curves in the bottom plot. The anorthite–enstatite join (dashed line) represents the thermal divide between silica-deficient and silica-excess rocks. Wo wollastonite, En enstatite, Di diopside, An anorthite, Ol olivine, Sill sillimanite.

Full size image

We quantitatively calculated melt compositions for these lunar rocks via phase equilibrium modeling (Supplementary Data 1–6). The contours of melt SiO2 contents are illustrated in Fig. 2g–l. Partial melts extracted from the silica-deficient FANs, Mg-suite, alkali-suite rocks, and low-Ti basalts are overall basaltic in composition, although a small number of intermediate melts (55–60 wt% SiO2) can be generated by very low-degree melting of alkali-suite rocks and low-Ti basalts. In contrast, partially melting the silica-excess KREEP and high-Ti basalts produces granitic melts (SiO2 > 65 wt%). Silica contents of the granitic melt along the solidus increase as pressure decreases (Fig. 2j, l). With the temperature increasing, melts from KREEP and high-Ti basalts retain high SiO2 contents until quartz is exhausted, after which melt compositions become decreasingly silicic over a narrow temperature interval and finally fall into the basaltic range.

Discussion

Highly silicic magmatism from remelting KREEP basalts

Our simulations show that, among all lunar igneous rock types, only KREEP and high-Ti basalts can produce highly silicic melts. However, of the two, high-Ti basalts plot much closer to the thermal divide, and thus, a much narrower parameter space of very low-degree (<5%) melting could yield granitic melts with a high-Ti basalt source (Fig. 4a). Because granitic melts are too viscous to segregate from low melt-fraction (<7%) parental rocks39, it is unlikely that the large silicic volcanic constructs on the Moon formed from crustal remelting of high-Ti basalts. Also, the volcanic constructs occur mostly far away from high-Ti mare deposits (Supplementary Fig. 1), further discounting the possibility of a high-Ti basalt origin. Moreover, despite the silica-excess classification of average high-Ti basalts, nearly half of them (~40%) plot on the silica-deficient side of the thermal divide (Fig. 3). Likewise, low-Ti basalts also straddle the thermal divide, of which nearly one-third plot on the silica-excess side but very close to the divide. Hence, neither high-Ti nor low-Ti mare basalts can be considered ideal sources for large-scale, highly silicic magmatism on the Moon.

Fig. 4: Modeled melt compositions of lunar igneous rocks.
figure 4

a Melt SiO2 content versus melting degree. Granitic melts (≥65 wt% SiO2; gray dashed lines) can be produced either by very low-degree (<5%) melting of high-Ti basalts or by moderate-degree (<20%) melting of KREEP basalts. b Total alkali-silica diagram showing non-modal batch melt compositions. KREEP basalt-derived melts (red line) plot in the rhyolite–dacite field when melting degrees are <20%. In contrast, partial melts of other lunar rocks (FANs rocks, purple line; Mg-suite rocks, dark blue line; Alkali-suite rocks, green line; Low-Ti basalts, bronze dashed line; High-Ti basalts, orange dashed line) mainly plot in the basalt field. Calculations were conducted at 1 kbar.

Full size image

KREEP basalts, by contrast, have an average composition plotting relatively far from the thermal divide. Moreover, almost all reported KREEP basalts fall to the silica-excess side (Supplementary Fig. 2), in accordance with petrographic observations showing the absence of magnesian olivine and the presence of silica in most KREEP basalts3. The non-modal melt compositions calculated along the 1 kbar isobaric melting profile are shown in Fig. 4. For KREEP basalts, granitic melts can be formed by <20% partial melting. With increasing melting degrees, KREEP basalt-derived melts migrate through the andesite field and finally enter the basalt field, while melts from other lunar rocks mainly plot in the basalt field. Furthermore, we also calculated the partial melting of the pristine KREEP (i.e., urKREEP), the last dreg of the magma ocean to solidify, which is proposed to be the parent for KREEP basalts40,41. The results show that no granitic melts are present during urKREEP melting beneath the lunar crust (Supplementary Fig. 3). Therefore, KREEP basalts are the best candidate for producing highly silicic magmas on the Moon.

A striking feature of lunar silicic volcanic constructs is high enrichment in Th (up to ~80 μg g−1)12,42, which can also be achieved by KREEP basalt melting. We combined phase equilibrium and trace-element modeling to compute the Th contents of the silicic melts (see “Methods”). Our modeling shows that KREEP basalt-derived granitic melts have Th contents 2 to 7 times higher than KREEP basalts (Supplementary Table 2). KREEP basalts have average Th contents of 9.3 μg g−1 (ranging from 4.5–14.9 μg g−1)3. Thus, the calculated silicic melts have 28–74 μg g−1 Th, matching estimates from remote sensing (~10–80 μg g−1)12,42. In contrast, high-Ti and low-Ti mare basalts mostly have Th contents less than 1.5 μg g−1 (ref. 3), resulting in partial melts only with <15 μg g−1 Th. For major elements, although no lunar missions visited a silicic target, the returned granite clasts/fragments and granitic glasses may provide preliminary constraints on the silicic compositions of the Moon (Supplementary Table 3), with the caveat that these granitic samples are not equivalent to lunar silicic volcanic constructs. Our calculated melt compositions for KREEP basalts are broadly consistent with the present granitic data (Supplementary Fig. 4). Moreover, all but one of the silicic volcanic constructs are confined to the Procellarum KREEP Terrane (PKT), where most KREEP-rich materials were found (Fig. 1). Furthermore, the exception (i.e., Compton–Belkovich volcanic complex) arguably proves the rule in this case as the single occurrence outside the PKT is associated with a rare enrichment of Th on the lunar farside14. These arguments collectively support a genetic link between lunar silicic magmatism and KREEP basalts.

Potential formation scenarios

As the silicic volcanic constructs are spatially associated with mare basalts in the PKT, previous work suggested basaltic underplating as a heating mechanism to trigger crustal remelting and produce large volumes of silicic lavas12,13. However, mare magmatism in the PKT was active throughout the Nectarian into the Copernican period, mostly from 3.8–3.1 Ga43,44,45, enduring much longer than the more ancient and temporally restricted ca. 4.0–3.6 Ga silicic volcanic constructs25,26,27. Moreover, most basaltic plains surrounding the silicic volcanic constructs are relatively younger26,42. These disparate ages complicate the underplating model and must be considered for evaluating if this model is plausible for lunar silicic magma generation.

Alternative mechanisms for KREEP basalt melting include radiogenic heating and meteoritic impact. KREEP basalts have high concentrations of radioactive elements such as K, Th, and U, with an average heat production value of 4.4 ± 1.8 (1σ) µW m−3 at 4.0 Ga (calculated after ref. 46). We used the one-dimensional heat flow equation with terms for radiogenic heat production and conduction to model the thermal evolution (see “Methods”). As a sensitivity test, the models assumed a buried 5-km-thick pile of KREEP basalts within a 40-km-thick crust at variable depths (20–40 km). The initial crustal thermal gradient was fixed to 15 or 20 °C km−1 (ref. 47). The results are illustrated as a series of temperature-time curves with variable levels of heat production (2–7 µW m−3). KREEP basalts near the base of the crust could have been radioactively heated above the solidus for ~100–200 million years after their formation (Fig. 5; Supplementary Fig. 5). This timing is consistent with the relative timing, temporal overlap, and short time interval between KREEP basalt formation (ca. 4.1–3.8 Ga)43,44,45 and silicic volcanic construct development (ca. 4.0–3.6 Ga)25,26,27. As burial depths decrease, the role of radiogenic heating would be offset by enhanced thermal diffusion (Fig. 5b, c). Therefore, partial melting of KREEP basalts could have been achieved by internal radiogenic heating preferentially in the deep crust.

Fig. 5: Radiogenic heating modeling.
figure 5

We assumed burial of a 5-km-thick pile of KREEP basalts within a 40-km-thick crust at depths of 40 km (a), 30 km (b), or 20 km (c), with a thermal gradient (dT/dz) of 20 °C km−1 (see Supplementary Fig. 5 for 15 °C km−1 results and other conditions). KREEP basalts have 0.54 ± 0.13 (1σ) wt% K2O, 9.3 ± 3.5 μg g−1 Th, and 2.6 ± 1.1 μg g−1 U (ref. 3). Their heat production value was 4.4 ± 1.8 µW m−3 at 4.0 Ga. The red curves represent the mean values, and the reddish areas cover the full range. The solidus temperatures of KREEP basalts at 40, 30, and 20 km are 911, 896, and 883 °C, respectively.

Full size image

As lunar silicic magmatism formed broadly contemporaneously with the ca. 3.9 Ga Late Heavy Bombardment (LHB)48 and silicic volcanic constructs occur mostly near the edges of impact basins49, impact melting may be another mechanism for lunar silicic magma formation. To test this scenario, we used the iSALE-2D hydrocode50,51 to perform two-dimensional impact modeling (see Methods). The models assume a vertical impact of a dunite projectile with a diameter of 20, 40, or 60 km into a 40- or 52-km-thick lunar crust52 with a thermal gradient of 15 or 20 °C km−1, and impact velocity is fixed at 15 km s−1 (ref. 53). The results show that large impacts can produce regional thermal anomalies and form a hot melt pool near the impact site (Fig. 6). At greater distances, the crustal temperature drops gradually, but at shallow depths, there would be a hot layer at temperatures of 850–1000 °C above the solidus of KREEP basalts extending to the edges of impact basins. This model is similar to recent simulations of the early terrestrial bombardment54 and suggests that impacts could provide the heat needed for KREEP basalt melting in the shallow crust. The silicic volcanism erupting near the Apollo 15 site in the southeast edge of the Mare Imbrium basin, where KREEP basalts are a common component of the region4, can provide a good example in support of the impact model. Interestingly, the radiogenic and impact melting models make different predictions for melting occurring in the deep or shallow crust, possibly providing a way to discriminate between their relative contributions. Nonetheless, we note that the above models are not mutually exclusive and might be complementary to one another to generate large volumes of silicic magma on the Moon.

Fig. 6: Hydrocode impact modeling.
figure 6

Post-impact thermal structure for a vertical impact of a dunite impactor with a diameter (Dimpactor) of 20 km (a), 40 km (b, c), or 60 km (d) and an impact velocity of 15 km s−1. The pre-impact crustal thickness (Zcrust) was fixed to either 40 km (a, c) or 52 km (b, d), with a thermal gradient (dT/dz) of 15 °C km−1 (b) or 20 °C km−1 (a, c, d). The turquoise areas on the right side show regions of temperatures of 850–1000 °C where partial melting of KREEP basalts could have occurred.

Full size image

Implications for silicic magmatism in the solar system

Our modeling demonstrates that remelting KREEP basalts can produce evolved, silica-rich magmas without requiring water and plate tectonics. It was long thought that the siliceous crust is exclusive to Earth. However, evolved, SiO2-rich crustal materials are being increasingly discovered on the Moon13,14, Venus55, Mars56, and some asteroids57,58,59, indicating that silicic rocks may constitute a significant part of ancient crusts on rocky bodies in the solar system. Our work indicates that highly silicic magmas can be produced by the partial melting of “silica-excess” (quartz-normative) basaltic rocks under anhydrous conditions in planetary crusts throughout the solar system. The LHB affected the entire inner solar system48 and could have caused melting for early silicic magmatism formation60. Such models imply that the early evolved crust on rocky planets might be generated from an older mafic-dominated reservoir. This inference is consistent with our current knowledge of early Earth54,61,62. Although Earth’s modern felsic crust can be efficiently created in subduction zones6, the Moon, Mars, Venus, and Mercury seem to lack plate tectonics, and possibly even the early Earth63. The silicic magmatism on the stagnant-lid Moon produced by crustal remelting does not require plate tectonics or water and thus provides additional insights into planetary crust formation in the early solar system.

Methods

Phase equilibrium modeling

The modeling was performed using THERMOCALC version 3.4764 with the internally consistent thermodynamic dataset ds62 (updated February 2012)65 in the NCKFMAST (Na2O–CaO–K2O–FeO–MgO–Al2O3–SiO2–TiO2) system. All iron is ferrous, corresponding to the highly reduced conditions of the Moon66. The activity–composition solution models used here were as follows: silicate melt, high-Ca pyroxene (augite)35, low-Ca pyroxene (orthopyroxene), ilmenite67, plagioclase, K-felspar68, and olivine65. Quartz was treated as a pure phase. These solution models have successfully reproduced the phase relations and partial melting of basaltic rocks35,36,54. Thus, they permit quantitative calculations of melt compositions of lunar basaltic igneous rocks (FANs, Mg-suite, alkali-suite rocks, KREEP basalts, and low-Ti and high-Ti mare basalts). Ilmenite saturation is over-predicted under very high-degree melting conditions (other minerals are exhausted) because of the lack of a Ti-bearing endmember in the melt model35; nonetheless, it has little influence on our models in terms of low-degree partial melting. Average bulk compositions were employed for phase equilibrium modeling. Bulk compositions of low-Ti and high-Ti mare basalts are from ref. 37, and the others are from ref. 3 (Supplementary Table 1). All modeling was run under anhydrous conditions due to very low water contents (μg g−1 level) in returned lunar samples69,70. Although Bhattacharya et al.71 conducted spectral and spatial analyses of the silicic Compton–Belkovich volcanic complex and estimated its water content varying from 0.1 to 0.5 wt%, our additional modeling demonstrates that adding water to the dry system of lunar igneous rocks has minimal effect on their melt compositions, especially for SiO2 contents (Supplementary Fig. 6). Modeled results match well with petrographic observations, verifying the reliability of our simulations. Mineral abundances and melt compositions (non-modal batch melting) calculated in the range of 800–1500 °C and 0–5 kbar are presented in Supplementary Data 1–6.

Trace-element (thorium) modeling

Thorium contents of granitic melts extracted from KREEP basalt melting were calculated using the batch melting equation:

$${C}_{{{{{{\rm{melt}}}}}}}/{C}_{{{{{{\rm{source}}}}}}}=1/[D+F\times (1-D)]$$
(1)

where Cmelt and Csource are Th contents of the melt and source rock, respectively, F is the degree of partial melting, and D is the bulk partition coefficient calculated using residual mineral modes and mineral/melt partition coefficients. Mineral and melt modes were obtained by phase equilibrium modeling at 1 kbar. The resultant phase proportions are output as mole percent, very close to volume percent. Thus, the weight percent was further calculated using mineral densities from ref. 72. Most mineral/melt partition coefficients used here were taken from ref. 73, wherein the values are widely used to simulate anatexis of metabasites. Partition coefficients for K-felspar and quartz are from refs. 74,75. The calculated procedures and results are presented in Supplementary Table 2.

Radiogenic heating modeling

The contribution of radiogenic heating to the melting of KREEP basalt was investigated using the one-dimensional heat flow equation76. The standard heat flow equation includes conductive heat flow between rocks of different temperatures, vertical heat transport due to burial or erosion, and heat production by radioactive decay, mechanical deformation, and chemical reaction. Because the vertical heat transport and mechanical and chemical heating can be negligible for lunar crustal rocks, the heat flow equation used here can be simplified as:

$$\frac{\partial T}{\partial t}=\frac{\partial }{\partial z}\left[\kappa \left(T\right)\frac{\partial T}{\partial z}\right]+\frac{{A}_{{rad}}}{\rho \cdot {C}_{p}}$$
(2)

where T is temperature, t is time, κ is thermal diffusivity, z is depth, ρ is density, Cp is specific heat capacity, and Arad is the rate of radiogenic heat production. The models were set with a surface temperature of 250 K, crustal thickness of 40 km, an initial crustal thermal gradient of 15 or 20 °C km−1, a basal temperature of 1300 K at 150 km depth, and a density of KREEP basalt of 3.1 g cm−3 (refs. 47,77). The crust was fixed to comprise a 5-km-thick KREEP basalt layer and a 35-km-thick non-radioactive layer, where the model runs assume variable burial depths of the KREEP basalt layer from 20–40 km. Scenarios shallower than 20 km are not shown because they cannot produce enough heat for KREEP basalt melting. KREEP basalts have 0.54 ± 0.13 (1σ) wt% K2O, 9.3 ± 3.5 μg g−1 Th, and 2.6 ± 1.1 μg g−1 U (ref. 3). Their heat production value was 4.4 ± 1.8 µW m−3 at 4.0 Ga, calculated after ref. 46. As the Cp of different samples is essentially an identical function of temperature78, the widely used Cp equations provided by ref. 79 were selected here. For the κ, we used the temperature-dependent equation determined on Apollo basalt 10020, 44, because it has relatively low porosity among all measured lunar samples80. Also, modeling with variable KREEP basalt thickness (2–7 km) was performed to evaluate its effect on the radiogenic heating mechanism (Supplementary Fig. 5). The results show that the thickness of KREEP basalts should be greater than ~4 km if radiogenic heating alone could raise the temperature above the solidus.

Hydrocode impact modeling

To examine the impact melting process possibly related to the lunar silicic magma formation, we carried out numerical simulations with the shock physics code iSALE-2D50,51. The code is based on the SALE (Simplified Arbitrary Lagrangian Eulerian) hydrocode solution algorithm81 to solve the conservation equations of mass, momentum, and energy. In addition, to describe the thermal and mechanical responses during the hypervelocity impacts, various equations of state and elasto-plastic constitutive models for multiple materials are included82,83. More recent improvements include a modified strength model50, a porosity compaction model51,84, and a dilatancy model50. The iSALE-2D code has been validated against observations85 and widely used in simulating lunar basin formation52,86,87. In this study, all simulations are axisymmetric, and the projectile vertically impacts a spherical Moon-like target along a central axis. The target has a central gravity field, and it is assumed to have a crust, mantle, and core that are represented by granite, dunite, and iron, respectively. The material models and corresponding parameters derive from ref. 52. The projectile is represented by dunite because it is a reasonable approximation for typical ordinary chondrite asteroidal material88,89,90, and granite is often used as a close analog to the lunar anorthositic crust89,90,91. The pre-impact crustal thickness was fixed to either 40 or 52 km, the surface temperature was 250 K, and the crustal thermal gradient was 15 or 20 °C km−1. To avoid reducing mantle material strength to zero, we followed refs. 47,52 by imposing an adiabat of 0.04 °C km−1 at temperatures above 1,300 K. The mesh resolution was set as a constant with a cell size of 250 × 250 m. The impact velocity was fixed at 15 km s−1 (ref. 53). Because the silicic volcanic constructs occur mostly near the edges of impact basins49, a dunite impactor with a diameter of 20, 40, or 60 km was employed in our simulations, forming an impact basin ~250, 500, or 800 km in diameter, respectively. Numerical results indicate the above parameter domain is enough to analyze the lunar silicic magma formation. The representative results are presented in Fig. 6, and the corresponding animations are provided in Supplementary Movies 1–4.