Duration of partial melting in the lower crust of the Limpopo collisional belt

In convergent settings, the duration of partial melting in the lower continental crust dictates how lateral shortening is accommodated by the colliding plates. Here we use the example of the ca. 2700 Ma Southern Marginal Zone of the Limpopo Belt, South Africa, to accurately time the different steps of the granulite facies metamorphic event. We date garnet crystallisation in K‐poor leucosomes using Sm–Nd garnet isotopic data to have occurred at 2734 ± 9 Ma. Following deposition of the protolith and rapid burial, the crust remained molten for 17 ± 14 Ma. This estimate is within error similar to the 24 ± 12 Ma proposed for rocks that have evolved along a Barrovian P–T path. Our results suggest that the Southern Marginal Zone is the result of a collision between a large island arc or a continent with the Kaapvaal Craton, at a time global of geodynamic changes.


| INTRODUC TI ON
The timing of partial melting and how long the crust remains molten are crucial questions to understand the evolution of the rheology of the crust in convergent settings. Time spent above the solidus directly influences crustal differentiation and crustal shortening mechanisms (e.g. Clemens et al., 2020;Maierová et al., 2017;Nicoli, 2020;Sawyer et al., 2011;Vanderhaeghe, 2009). In migmatitic terrains, the presence of K-poor, Ca-rich leucosomes associated with metasedimentary lithologies has been used to investigate the early process of partial melting, that is open system and disequilibrium (Fancello et al., 2018;Fornelli et al., 2002;Nicoli et al., 2017;Zeng et al., 2012;Zuluaga et al., 2017). In this study, we focus on the example of the Neoarchean Southern Marginal Zone (SMZ) of the Limpopo Belt, South Africa (Du Toit et al., 1983) (Figure 1a) to better understand the rate and duration of partial melting processes.
Although the chronology of the high-grade metamorphic event in the SMZ is reasonably well constrained, a debate remains on the timing and source of K-poor, garnet-bearing trondhjemitic bodies, for example internally derived from the partial melting of the surrounding metasediments (e.g. Nicoli et al., 2017;Stevens & van Reenen, 1992;Taylor et al., 2014) or post-peak metamorphism intrusions from an external source (e.g. Belyanin et al., 2014;Sofonov et al., 2014;Sofonov, Reutsky, et al., 2018;Sofonov, Yapaskurt, et al., 2018). This discrepancy is ultimately linked to the nature of the geodynamic mechanisms responsible for the tectonometamorphic evolution of the SMZ-that is subduction and continental collision  and references therein) vs. mantle-driven diapirism (van Reenen et al., 2019 and references therein). To solve these issues, we have measured whole rock and garnet samarium-neodymium (Sm-Nd) isotopic ratios in these leucocratic bodies to retrieve their crystallisation age.

| PRE VIOUS WORK
The SMZ of the Limpopo Belt is located north of the Kaapvaal Craton and consists of a suite of Archean gneissic granitoids-the Baviaanskloof Gneiss, and metasedimentary and metamafic bodies-the Bandelierkop formation (Du Toit et al., 1983) (Figure 1a).
As a result of accretion to the northern edge of the Kaapvaal Craton at ca. 2700 Ma, the SMZ has experienced a single granulite facies metamorphic event at ~850°C and ~11 kbar  and references therein) (Figure 1b). Prograde and peak metamorphism produced a first episode of muscovite and biotite consuming, deformation-assisted, fluid-absent partial melting reactions that was responsible for the formation of garnet-bearing, deformed leucosomes (L1) (Nicoli et al., 2017;Stevens & van Reenen, 1992). L1 leucosomes crystallised near peak pressure conditions due to kinetically induced chemical disequilibrium during partial melting and melt loss (Madlakana & Stevens, 2018;Nicoli et al., 2017). As a result, there is a direct correlation between the width of the leucosomes and their composition, from tonalitic to granitic (Figure 2a). A second generation of orthopyroxene-bearing, nebulitic leucosomes (L2) formed during decompression under granulite facies conditions and crosscut the foliation marked by the L1 leucosomes. The granulite facies metamorphic event was followed by a phase of active post-collisional magmatism (Laurent et al., 2014) and a possible succession of discrete lower-grade metamorphic events focussed on the extended network of shear zones in the SMZ (e.g. Belyanin et al., 2014;Brandt et al., 2018), which culminated in a fluid-assisted retrograde amphibolite facies metamorphic event at ca. 2100 Ma (Madlakana et al., 2020).
The protolith of the Bandelierkop formation metasediments has been identified as psammite/greywacke that was deposited in the foreland of an active collisional setting (Nicoli et al., 2016) at >2 733 Ma (Kreissig et al., 2001;Nicoli et al., 2015). U-Pb ages on zircon have been used to date one trondhjemitic body cropping out in the Pettronela Shear Zone (PSZ) . The resultant 2 726 ± 13 Ma date has been interpreted as an inherited age. However, this age is, within error, similar to the inferred onset of crustal thickening recorded in the footwall of the Hout River Shear Zone (HRSZ) at 2729 ± 19 Ma (Kreissig et al., 2001;Passeraub et al., 1999). The age of peak metamorphism, 2715 ± 5 Ma, has been constrained in several locations (e.g. Kreissig et al., 2001;Rajesh et al., 2014;Taylor et al., 2014;Nicoli et al., 2015;Safonov et al., 2020). U-Pb and Pb-Pb analyses on zircons and monazites from migmatites give ages of ca. 2 691 Ma, which have been interpreted as the end of the granulite metamorphic event (Kreissig et al., 2001;Safonov et al., 2020). Contemporary to the cooling of the granulitic terrane, melting of both the asthenospheric mantle and an enriched, sub-continental lithospheric mantle, metasomatised by subducted supracrustal material, generated a series of post-tectonic magmatic intrusions (Laurent et al., 2013(Laurent et al., , 2014. A series of younger

Significance Statement
The duration of partial melting in the lower continental crust is key to understanding both the efficiency of crustal differentiation and how crustal shortening is accommodated in convergent settings. Combined with field and petrological observation as well as chemical modelling, the use of different geochonometers (U-Pb in zircon, Sm-Nd in garnet) provides a direct way to quantify the different steps followed by a given rock during its metamorphic and anatectic evolution.  . Main melting reactions: 1: Bt + Pl + Qtz + Ms = Al 2 SiO 5 + Liq; 2: Bt + Pl + Qtz + Al 2 SiO 5 = Grt + Liq; 3: Bt + Pl + Qtz = Opx + Liq.  (Dubinina et al., 2015;Nicoli et al., 2017;Safonov, Reutsky, et al., 2018;Stevens, 1997;Taylor et al., 2014). The arrow indicates the increase in the width of the garnet-bearing L1 leucosomes. The numbers in italic next to the arrow indicate the average width of the leucosome in metres.  Safonov et al. (2020). The amphibolite facies event did not occur before 2100 Ma (Madlakana et al., 2020). We grouped the ages in seven events according to the different interpretations found in the literature. The full compilation of ages is available in supplementary materials. Major (Mn, Ca) and trace (Y, P) elements profile through a large garnet in DT06A. (c) Major (Mn, Ca) and trace (Y, P) elements profile through a small garnet in DT06A. Large garnet composition-core: X alm 53, X pyp 39, X spss <1, X grs 5-7; rim: X alm 0.53, X pyp 0.39, X spss 2, X grs 3-4; small garnet composition-core: X alm 52-53, X pyp 41-44, X spss 1-2, X grs 3; rim: X alm 58-62, X pyp 0.32-0.37, X spss 2-4, X grs 3 (Nicoli, 2015). Pl + Qtz Ox U-Pb and Pb-Pb ages, 2 634-2 672 Ma, obtained on monazite and kyanite in the PSZ and the footwall of the HRSZ could be related to this late magmatic activity and the percolation of fluids along pre-existing structures at the end of the granulite facies event (Passeraub et al., 1999;Kreissig et al., 2001;Belyanin et al., 2014;Safonov et al., 2020  facies, probably linked to another metamorphic event, did not occur before ca. 2100 Ma (Madlakana et al., 2020). A full compilation of ages associated with the high-grade metamorphic event in the SMZ is presented in Figure 2b and Table S1.

| S M-ND DATIN G
Sm-Nd dating was performed on three garnet-bearing L1 leucosomes (DT01A, DT02A, DT06A) (Figure 3a) belonging to the Bandelierkop formation in the Brakspruit quarry . The partial melting reactions and kinetic processes responsible for the formation of these leucocratic bodies have been documented and modelled in several prior studies (e.g. Madlakana Nicoli et al., 2017;Taylor et al., 2014). The garnet in the leucosomes varies in size from 5 mm to >1 cm in diameter.
The low solubility of FeO and MgO, and the presence of peritectic minerals, in some cases euhedral in shape, as inclusion in the cores of these garnet crystals is consistent with their formation by partial melting with little to no pre-anatectic history (Madlakana & Stevens, 2018). Melt re-integration modelling  showed that the volume of sub-solidus metamorphic garnet is negligible, <2 vol.%. The interiors of the large garnets are largely  (Nicoli et al., 2017) as the result of internal re-equilibration within the segregated magma (Dorais & Tubrett, 2012;Nicoli et al., 2017). This would have occurred approximately during the late stage of the prograde path (T > 650°C, P > 6 kbar) and at peak metamorphic conditions (850°C, 11 kbar).
No accessory phases (e.g. zircon, monazite, apatite, epidote) have been found as inclusions in both types of garnets from this specific sample which could affect the accuracy of the Sm-Nd results and their geologic significance. Therefore, age estimates obtained using Sm-Nd isotopes should represent episodes of garnet growth in a melt-rich environment (Pollington & Baxter, 2011;Scherer et al., 2000;Smit et al., 2013).
We analysed whole rock fractions and their garnet separates. The rock samples were crushed and pure garnet grains of 500-1 000 μm size were handpicked. To remove phosphate and oxide inclusions, the mineral separates were subjected to aggressive leaching, by sequential sulphuric acid (Anczkiewicz & Thirlwall, 2003) and aqua regia (Ravikant, 2006). The isotopic composition was then determined from whole rock and leached garnet fraction following the protocol given by Bast et al. (2015) (see the detailed method in the supplementary material).

| DISCUSS ION
The new Sm-Nd ages obtained on trondhjemitic leucosome bodies in the SMZ of the Limpopo belt are key to quantifying the F I G U R E 5 Duration of the granulite metamorphic event and associated partial melting in the SMZ. L1 and L2 represent the two different generation of leucosomes, deformed and nebulitic respectively. PPM: prograde partial melting; DPM: decompression partial melting. D: deformation. Number along the metamorphic path indicates the volume fractions of melt that were present . Arrows symbolise melt loss episodes. Partial melts lost form the lower crust ultimately accumulated in the upper and middle crust to form S-type granite (now eroded). Cooling of the granulitic terrane was contemporaneous with the magmatic activity associated with post-collisional extension (Laurent et al., 2013(Laurent et al., , 2014. The whole SMZ was then affected by a network of interconnected shear zones from ca. 2670 Ma (Kreissig et al., 2001;Safonov et al., 2014), which might have acted as the conduits for fluid infiltration during retrogression of the granulites at ca. ~ 2100 Ma (Madlakana et al., 2020).  Kreissig et al., 2001). It is important to note that this study has determined the age of peritectic garnet in leucosomes and that the cores of such peritectic garnets from the SMZ do contain inclusions of other peritectic minerals such as orthopyroxene and rutile (Madlakana & Stevens, 2018). Thus, these garnet crystals are interpreted to be solely peritectic in origin, i.e. that they do not contain any significant quantity of pre-anatectic garnet in their cores. This interpretation leads to the longest inferred range of crystallisation ages. U-Pb zircon and monazite ages in biotite-bearing gneisses, 2 697 ± 36 Ma, 2731 ± 39 Ma and 2 734 ± 37 Ma, have also been interpreted as crystallisation ages linked to the granulitic event (Vezinet et al., 2018). Our new estimate is older than the U-Pb date usually interpreted as the peak metamorphic age in the SMZ, ca. 2 715 Ma. As both L1 and L2 leucosomes give identical U-Pb zircon ages Taylor et al., 2014), it is likely that these ages instead correspond to the final stages of crystallisation of the remaining liquid in the leucosomes (e.g. Hallett & Spear, 2015;Zeiger et al., 2015., Melo et al., 2017Rocha et al., 2018). The near identical ages of garnet formation in the leucosome and of minimum sedimentation suggest a prograde path faster than previously suggested, >2 km Ma −1 . During the burial, prograde partial melting of the metasedimentary lithologies produces up to 26 vol.% of anatectic melt  ( Figure 5).
This is significantly higher than the 7 vol.% melt connectivity threshold, that is a melt film is present in >80% of grain boundaries (Rosenberg & Handy, 2005). As the L1 leucosomes formed in an open system as the result of melt segregation and loss near peak conditions (e.g. Nicoli et al., 2017;Taylor et al., 2014), it is possible that several batches of melt were generated along the prograde path (Taylor et al., 2014). Rheological (Nicoli et al., 2017) and trace element modelling (Villaros et al., 2009) have shown that a single batch of magma is likely to remain in its source rock for <500 years. The great diversity in width, composition and garnet content of leucosomes exposed in the Brakspruit quarry supports this statement (Nicoli et al., 2017).
The range of εNd(t) values also indicates some degree of mixing between crustal and mantle components in the metasediments that were the source of the L1 leucosomes ( Figure 4d) (Zeh et al., 2007), which could, in part, explain the chemical variation in the leucosome population. Near isothermal decompression partial melting produced a maximum of 10 vol.% of anatectic liquid which feeds into the second generation of L2 leucosomes . These leucosomes also lost melt efficiently after accumulation of peritectic garnet, orthopyroxene and plagioclase (Madlakana & Stevens, 2018;Nicoli et al., 2017) and partial crystallisation of quartz and additional plagioclase, as attested by the excellent preservation of the ferromagnesian peritectic minerals in these structures. Therefore, injection of late, externally derived magma in the Bandelierkop formation postpeak metamorphism (e.g. Belyanin et al., 2014;Sofonov et al., 2014;Sofonov, Reutsky, et al., 2018;Sofonov, Yapaskurt, et al., 2018)   Frenquency path and at peak metamorphic conditions, ~ 35 km and 850°C (Nicoli et al., 2017). Using this new information, we reassessed the duration of partial melting in the SMZ ( Figure 5). Our new data suggest the crust remained partially melted for 17 ± 14 Ma.
Rapid decompression to 6 kbar occurred at a rate of 1.2 ± 0.7 km.
We compiled the time spent above solidus for several upper-amphibolite facies to ultra-high temperature (UHT) facies rocks of siliciclastic and felsic origin ( Figure 6). Our compilation differs from that of Harley (2016) as his deals with the duration of UHT metamorphism, that is time spent above 900°C (Δt 900 ). In the SMZ, the time spent above the solidus value is 17 ± 14 Ma, within error similar to the average Barrovian facies value, 24 ± 12 Ma ( Figure 6). The mixed nature of the L1 leucosome indicates that magma underplating is likely to be the heat source that triggered the formation of crustal melts (Zeh et al., 2007).
This protracted partial melting might result in a strong viscosity contrast between the lower crust and the middle crust which can affect the rheological behaviour of the entire orogen. Differences in gravitational potential energy and vertical shear zone can drive exhumation and subsequent lateral crustal flow and exhumation of migmatitic terrains (e.g. Vanderhaeghe & Teyssier, 2001;Weinberg & Mark, 2008). This would be accommodated in the upper crust by the propagation of thrust fronts in the foreland (Cagnard et al., 2006).
In the SMZ, the HRSZ, active between 2600 and 2700 Ma (Kreissig et al., 2001) would be a viable candidate to accommodate this horizontal spread (Smit et al., 2014).
Considering P-T conditions, timing, rates, melt production and isotopic signature, we suggest that the Limpopo orogeny should be regarded as a proto-collisional orogen which resulted from the subduction of a juvenile Neoarchean oceanic crust and the accretion of large island arcs.

ACK N O WLE D G E M ENTS
We thank the editor and the anonymous reviewers for their comments and suggestions. This research was supported through the Alexander von Humboldt Foundation to GN. We thank Prof. R.
Vadlamani of the Radiogenic Isotope Laboratory of IIT Kharagpur for supporting SM for the Sm-Nd isotopic analyses. Open Access funding enabled and organized by Projekt DEAL.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that supports the findings of this study are available in the supplementary material of this article.