On the origin of oldest Iron-Oxide-Copper-Gold (IOCG) deposits at the transition from Archean drip to plate tectonics

Geological evidence supports a signicant change in Earth’s behaviour in the mid- to late Archaean, between 3.2 and 2.5 Ga, reecting stabilization of the lithosphere and replacement of vertical tectonics by linear imbricated belts. At the heart of this change, the oldest (c. 2.75 Ga) Iron-Oxide-Copper-Gold deposits (IOCG) were formed in the Carajás Mineral Province (CMP) of the Amazon craton. U-Pb ages, Lu-Hf isotopes and trace element composition of detrital zircons from modern drainages record the crystallization ages of the exposed rocks of the CMP. Combined with the geochemistry of Archean granitoids in the CMP, we recognize four different age and compositional groups: 3.01-2.92 Ga TTG, 2.87-2.83 Ga transitional TTG + sanukitoid + K-rich granitoids, 2.78-2.72 Ga A-type crustal granites accompanying IOCGs, and 2.59-2.53 Ga alkaline high-K intrusions accompanied by renewed IOCG mineralization. The rst two groups have a dominantly juvenile isotopic signature whereas the last two have evolved Hf-isotope signatures, accompanied by increase in K 2 O/Na 2 O, reecting addition of old crustal components in the melting sources over time. The older juvenile granitoids are associated with dome-and-keel structures typical of granite-greenstone terranes, whereas the younger granitoids were emplaced along a linear shear belt associated with new mac-ultramac intrusions and remelting of older TTG. Based on the tectono-magmatic evolution, we argue that metasomatism and fertilization of the underlying lithospheric mantle by incompatible elements, necessary for the development of IOCG deposits, were related to vertical drip-tectonics during development of the TTG proto-continent. This proto-continent made the lithosphere rigid enough to allow linear translithospheric deformation to localize at c. 2.85 Ga, allowing decompression melting of the metasomatized lithospheric mantle in a restricted extensional setting to form abundant mac and A-type

accompanying IOCGs, and 2.59-2.53 Ga alkaline high-K intrusions accompanied by renewed IOCG mineralization. The rst two groups have a dominantly juvenile isotopic signature whereas the last two have evolved Hf-isotope signatures, accompanied by increase in K 2 O/Na 2 O, re ecting addition of old crustal components in the melting sources over time. The older juvenile granitoids are associated with dome-and-keel structures typical of granite-greenstone terranes, whereas the younger granitoids were emplaced along a linear shear belt associated with new ma c-ultrama c intrusions and remelting of older TTG. Based on the tectono-magmatic evolution, we argue that metasomatism and fertilization of the underlying lithospheric mantle by incompatible elements, necessary for the development of IOCG deposits, were related to vertical drip-tectonics during development of the TTG proto-continent. This proto-continent made the lithosphere rigid enough to allow linear translithospheric deformation to localize at c. 2.85 Ga, allowing decompression melting of the metasomatized lithospheric mantle in a restricted extensional setting to form abundant ma c and A-type granitoids at c. 2.75 Ga, and the rst IOCG deposits on Earth.

Background
The onset of plate tectonics is a heavily debated topic [1,2], but geological evidence from paleomagnetism, development of passive margins, metamorphic rocks and granitoid suggests that it was progressively established from 3.2 to 2.5 Ga [2]. Archean granitoid archives from many cratons show a change in rock chemistry between 3.2 and 2.5 Ga from intrusions of the sodic tonalite-trondhjemitegranodiorite (TTG) series to granitic intrusions with increasing potassic content [2,3,4].
The elevated Na/K, Sr/Y and high La/Yb in many TTGs is ascribed to partial melting of ma c rocks at depth, leaving a stable garnet residue [5,6]. Through time there was a decrease in TTG Na+Ca and Sr contents, which led to the distinction between high-Sr TTG (deeper, high-P melts with residual garnet) and low-Sr transitional TTG (lower-P melts with residual plagioclase) sub-types [3,4,5,6]. Associated depletion in heavy rare earth elements (HREE) and residual garnet requires melting of source ma c rocks in a range of 1 to 3 GPa [7,8], but depending on the source composition stability of garnet can be achieved at pressures bellow 1 GPa [9].
Contemporaneously with the transitional TTG, a speci c and restricted type of granitoid, known as sanukitoid [10] began to appear in nearly all cratons, where they are late additions accompanied in some cases by K-rich granitoids [3]. These sanukitoids are mostly diorites to granodiorites that have a clear mantle signature, especially in their high MgO contents, but are also enriched in incompatible elements, most notably Ba and Sr, and sometimes high-eld strength elements [3,4,10]. Melting experiments and petrogenetic modelling show that they may have formed either by peridotitic melting of mantle previously metasomatised by felsic melts of TTG composition, or by reaction between TTG melts and mantle [11].
Although there is no reported genetic relationship between sanukitoids and Iron Ore-Copper-Gold (IOCG) deposits, they share the same geochemical signature with enrichment in trace and minor elements typical of both basic (Co, Ni) and felsic (LREE, LILE, U, F) sources [11,12]. The source of Cu, Au, incompatible elements, and volatiles in IOCG deposits is commonly attributed to metasomatism of the underlying subcontinental lithospheric mantle (SCLM) during previous subduction events on continental margins [13,14]. Precambrian IOCG deposits occur inboard of lithospheric boundaries in necked transitions between thick Archean and thinner Proterozoic underlying mantle that focused and guided melting of this previously subduction-metasomatized mantle. In contrast, Phanerozoic IOCG deposits are preferentially located in extensional or transtensional zones of arcs [13,14]. The existence of "modernstyle" subduction in the Archean is contentious [1,2,4] and the stagnant lid tectonic hypothesis is at least equally plausible to explain Archean geodynamics [15,16] and the evolving nature of Archean magmas [4].
Thus, understanding the origin and timing of Archean IOCGs and their temporal and chemical connection with the evolving Archean magmatism could help constrain the tectonic setting at the time.
The Carajás Mineral Province (CMP) of the Amazon craton in northern Brazil includes the only Archean IOCG deposit discovered so far. In this region, a long history of Archean granitoid magmatism (3.01 to 2.5 Ga) records the same compelling change in rock chemistry reported in other cratons [3]. Here we report on ages, Hf isotopic composition and trace elements of zircons from modern drainages within the CMP, which re ect the nature of the exposed Archean rocks. We combine these results with available zircon crystallization ages and bulk-rock chemistry of granitoids to de ne a temporal evolution linked to the tectonic evolution of the region. We argue that emplacement of the 2.87-2.83 Ga transitional TTG and sanukitoid magmas records early fertilization and melting of the lithospheric mantle, which was re-melted at c. 2.75 Ga to generate the IOCG mineral systems in the transition from drip to mobile tectonics in the Amazon Craton. This differs from the models of mantle enrichment via subduction proposed for the younger IOCGs [12].

GEOLOGICAL SETTING OF CARAJÁS MINERAL PROVINCE (CMP) AND ITS IOCG DEPOSITS
The CMP region encompasses three main rock associations: i) Archean basement rocks (3.00-2.83 Ga) represented by TTG-greenstone belt associations, with minor sanukitoids and K-rich granitoids [17,18,19,20]; ii) the Carajás basin composed of a meta-volcanosedimentary sequence, including thick Superior-type BIFs coeval with the intrusion of A-type granitoids and layered ma cultrama c complexes (2.78-2.72 Ga), alkaline high-K intrusions (c. 2.56 Ga) and younger low-grade siliciclastic cover rocks [21,22]; and iii) widespread Paleoproterozoic (c. 1.88 Ga) A-type intrusions [23] (Fig.   1). To the north, these lithologies are bounded by high-grade rocks (gneisses and migmatites) of Paleoproterozoic age (2.50-1.96 Ga) within older Archean terranes. Interpretations based on geological, bulk-rock geochemistry and isotopic data have led to two contrasting interpretations for the tectonic setting of the Carajás Basin: an intracontinental rift forming the Carajás basin [21,22] or a compressional arc-related setting [24].
In the CMP, the region to the south of the Carajás basin, the environment related to the emplacement of the 3.0 to 2.83 Ga TTGs is preserved, with the dome-and-keel structures typical of many granite-greenstone terranes [25] such as the Mogno dome [26,27] in Fig. 1. The dome-and-keel terrane is bound to the north by a high-to low grade 200-km wide anastomosing set of E-W to ESE-WNW strike-slip shear zones and faults (Itacaiúnas shear belt) developed around 2.85 Ga. This shear belt was later reactivated by sinistral motion at the time of development of Carajás basin at 2.78-2.72 Ga, accompanied by the intrusion of syn-kinematic A-type granites [20,28,29].
The IOCG deposits are located along or near these reactivated shear zones, in proximity to intrusions of different compositions, and are characterized by Fe-oxides and enrichment in Cu-Au-REE-(U-Y-Ni-Co-Pd-Sn-Bi-Pb-Ag-Te) [27]. They have been affected by intense proximal K-Fe and distal Na-Ca hydrothermal alteration with late-stage chlorite and carbonates, associated with hypersaline uids (250°-570°C) and late sulphide deposition with different sul dation states [30]. Geochronology points to two main IOCG mineralization events at 2.75-2.68 Ga and 2.62-2.50 Ga [31], the rst associated with reactivated regional shear zones and syn-kinematic intrusions of A-type granitoids [20] and the later to a second reactivation stage accompanied by minor alkaline high-K granitoids [32].

Results
Detrital zircons from modern drainages in the CMP have either oscillatory zoning or are homogeneous in CL images, indicating that they crystallized from magmas ( Fig. 2A). However, some zircons are rounded to ovoid and show core-mantle relationships. Their 207Pb/206Pb ages de ne four distinct groups (G1 to G4) that correspond to the zircon U-Pb and Pb-Pb crystallization ages of the exposed rocks, and hence they record peaks in magma addition into the CMP crust ( Fig. 2B and supplementary table S1). Peaks in magmatism, from G1 to G4, occurred at 3.05-2.92, 2.87-2.83, 2.78-2.72, and 2.59-2.53 Ga. In general, and based on the known regional geology, they are associated with an evolution from Na-rich high-HREE TTGs (G1) grading into low-HREE TTGs and MgO-rich sanukitoids (G2), and then to high-K granitoids (G3), revealing an enrichment in K over Na throughout time ( Fig. 2C and D). Trace-element discriminants (Y, Hf, U, Th, Yb and Lu) used to infer the zircon host rock during crystallization [33] points to a dominance of granitoid sources with important ma c addition in G3, during the development of Carajás basin. Three zircon grains suggest crystallization from a carbonatites or other highly alkaline rocks. The evolving granitoid chemistry is accompanied by a trend in zircon Hf isotope ratios from the early radiogenic juvenile signature of TTG rocks of G1 group (eHf (t) = +1.13 to +6.56) to progressively evolved and less radiogenic values where G2 has eHf (t) = -5.31 to +6.76, G3 has eHf (t) = -6.83 to +1.19 and G4 has eHf (t) = -4.41 to +8.27 (Fig. 2B). Hf model ages for the younger G2, G3 and G4 magmatism varies from 3.28 to 3.01 Ga which overlaps with the U-Pb and Hf model ages of G1.

Discussion
As mentioned, in general, the enrichment in K over Na is marked by the transition from TTG to K-rich granitoids, re ecting the increased importance of crustal components in the melt sources, indicated by the trend in zircon Hf isotope ratios (Figs. 2 and 3A). The TTGs of G1 and G2 show a wide range of Sr/Y and Eu*, suggesting that both garnet-rich and plagioclase-rich residues were involved, while G3 granitoids have lower Sr/Y (Figs. 3B and C) However, an increase in Sr/Y and (La/Yb)N during G1 (3.05-2.92 Ga) TTG magmatism ( 3B) indicates a deepening of the melt source [5] towards G2 associated transitional TTGs. This shift is also captured by the overall development of a positive Eu anomaly over the same interval, re ecting the increasing presence deeper plagioclase-free residues from G1 to G3 (Fig. 3D). The T DM ages of all the igneous zircons coincide between 3.25and 2.90 Ga, indicating that they could be generated by reworking G1 juvenile TTG or rocks with similar isotopic signature.
The dome-and-keel geometry preserved in the south CMP region preserved in the Mogno dome ( Fig. 1) shows that vertical tectonics played an important role during granitoid emplacement from 3.01 to 2.83 Ga. In the absence of Archean "modern-style" subduction [1,2,4], crustal tectonics could be driven by forces related to mantle convection under a cold lithospheric lid [34]. Numerical models indicate that this process creates lithospheric drips where mantle convection cells downwell and is an e cient way to drag crustal rocks deep into the mantle, where they can melt at high pressures [34]. Models also have shown that drips can evolve from symmetric (vertical) to asymmetric (tilted), mimicking "modern-style" subduction zones with important implications for the melts arising from the drips [4]. We use these drip models to interpret the evolution of magmatism in the CMP and the origin of its IOCGs.
We propose that during the 3.00-2.92 Ga magmatism in the CMP region is associated with the development of a symmetric drip dragging down crustal ma c rocks from the lid causing melting, below the garnet-in isograd, to form the juvenile G1 TTG melts (Fig. 4A). Lateral displacement of the lid due convection led the drip to become asymmetric starting the G2 magmatic stage. At this stage, previouslyformed TTGs were dragged down by the drip and re-melted in conjunction with the melting of ma c rocks [4], forming transitional TTGs and accounting for the increased K content of the G2 granitoids between 2.87-2.83 Ga. Recycling of older crustal rocks is re ected in their Hf isotopic values. The increase in Sr/Y values along with the positive Eu anomaly from G1 to G2 re ects the deepening of melt source in a transition from symmetric to asymmetric dripping. Because of this drip asymmetry new TTG magmas rose through and metasomatized the lithospheric mantle wedge (Fig. 4B) enriching it in LREE, LILE and some HFSE [4]. The contemporaneous dehydration and release of uids from the tilted ma c crustal drip could also trigger partial melting of this metasomatized mantle wedge, generating the high-Mg sanukitoid rocks observed in the 2.87-2.83 Ga interval (G2). The ascent of these hot G2 sanukitoid magmas would then favour melting of older TTG crust, generating the coeval high-K granitoids (Fig.4B) with ancient T DM model ages.
Enrichment of the mantle in LILE, REE and Cl has been attributed to fertilization by aqueous uids from dehydrating oceanic plates [35]. The same 'subduction' effect would be expected from the dehydration of the crustal ma c lid in the asymmetric drip phase, arguably responsible for the metasomatism of the lithospheric mantle associated to the sanukitoid magmatism in the 2.87-2.83 Ga interval. On the other hand, metasomatism of the mantle wedge above the drip would also be achieved through uid-free processes by the interaction of felsic TTG melts with mantle peridotites [11].The role of mantle uids in IOCG deposits has been demonstrated by the positive δ37Cl values (+0.2 to +2.1‰) derived from uid inclusions from selected deposits, including the Sossego deposit of the CMP [36].
The development of linear belts in the CMP region started during late G2 magmatism, with the nucleation of the Itacaiúnas shear belt at 2.85 Ga [28] along with metamorphism [37] and emplacement of G2 2.85-2.83 Ga sanukitoids and high-K granitoids along the shear belt [19]. This stage indicates that the protocontinent formed by the older TTGs of G1 and G2 was rigid enough to allow strain localization and linear deformation, arguably associated to far-eld stresses resulting from the development of an Archean orogens elsewhere.
Continuous far-eld stresses reactivated earlier structures during G3 magmatism at 2.78-2.72 Ga, developing the localized transtensional Carajás rift basin [29]. These structures could have reached the base of the lithosphere [38], and extension would have promoted heating and decompressional partial melting of the drip-metasomatized mantle (Fig. 4C). Comparison of empirical melting relations with modelled P-T conditions in the lithosphere during extension indicates that large volumes of magma may be produced by decompressional partial melting of melt-metasomatized mantle during the earliest stages of extension in localized rift basins [39,40]. These melts would pond at the base of the Itacaiúnas shear belt and ascend to generate the ma c-ultrama c rocks of the CMP and the Carajás basin, while the heat transferred by such underplating would favour melting at the base of the previous TTG proto-continent, yielding coeval syn-kinematic [20] A-type granites with ancient Hf-isotope signatures. Transport of the Cu and REE to into IOCG mineralization is commonly attained by chloride complexes [41,42]. We argue that the Itacaiúnas shear belt and its successive reactivated structures acted as conduits connecting the mantle to the upper crustal levels of the CMP. This event along the translithospheric shear zones of the Itacaiúnas shear belt would be volatile-rich [43] and able to transfer the Cu, REE and LILE from the dripmetasomatised mantle to shallow crustal levels where mixing with the basinal brines [30] allow the breakdown of the chloride complexes and deposition of the rst pulse of Cu endowment in the CMP at c. 2.76 Ga.
The last magmatic event at 2.59-2.53 Ga (G4) is marked by granitoids with exclusively evolved zircon Hf signatures, which suggests yet another reactivation event. This reactivation is corroborated by the partial or total resetting of the U-Pb and Rb-Sr systems and by Ar-Ar cooling ages together with the third pulse of IOCG mineralization at 2.60-2.45 Ga [31,44]. This magmatic event includes also minor high-K alkaline intrusions related to the sinistral reactivation of the northern shear zones of the CMP [28,32,37]. During this reactivation minor partial melting of the drip-metasomatized mantle tapped by the lithospheric shear zones, accompanied by volatile-rich batches, could produce both the alkaline incompatible-element-rich magmatism and renewed IOCG mineralization.
We present here an evolutionary history of the granitoid archive of the CMP where we integrate the origin of the IOCGs, where both magmas and mineralization inform on the tectonic setting and mantle alteration processes. In this evolution, we use published numerical models for Archean geodynamics to argue that that no subduction is necessary to explain the genesis of magmatic rocks from 3.01 to 2.50 Ga and that enrichment of the mantle, a necessary ingredient for IOCG mineralization, could be attained through drip tectonics [4]. The Hf-isotope record demonstrates that after a period of crustal growth ending at c. 2.9 Ga, the crust evolved through progressive reworking of the oldest rocks to produce younger generations of granitoids, a process that would be favoured by an essentially in-situ mechanism such as drip tectonics. The early TTGs were part of a broader process that formed this increasingly rigid protocontinent, and allowed strain localization into linear shear belts that were later reactivated to form narrow extensional rift zones connected by shear zones to the underlying lithospheric mantle. Decompression melting by mantle upwelling during rifting from 2.78-2.72 Ga was responsible for concentrated ma cfelsic activity in the rift zone and rst Earth`s IOCG deposits tapping the uids in the metasomatized lithospheric mantle. This event was followed by reactivation of the process at 2.59-2.53 Ga aided by volatiles from the mantle, promoting renewed IOCG mineralization in the CMP.

Methods
Sampling and data compilation. Four samples from modern drainages of the Carajás Mineral Province were collected on the Rio Plaque (sample BA236432), Rancho Refugio Drainage (sample BA236438), Rio Novo Drainage (sample BA236443) and Rio Sereno drainage (sample BA236447). For sample locations see Figure 1. Data are reported in the supplementary table S1. In order to strength interpretation on detrital zircon data we have extensively compiled whole-rock geochemistry and zircon U-Pb and Pb-Pb ages from Archean rocks of Carajás Mineral Province region. The compiled data is reported in the supplementary table S2.
Major and Trace Element Determinations. The major constituents of zircon, Zr and Si, and selected minor elements such as Hf, Y, U and Th were analysed using GEMOC's CAMEBAX SX50 electron microprobe. More precise Uranium and Thorium concentrations were calculated from LAM-ICPMS U-Th-Pb analyses by comparison of U and Th raw counts with those obtained for the GEMOC GJ1 standard. Yb/Hf and Lu/Hf ratios collected as part of the multi-collector ICPMS Hf-isotope analysis were converted to Yb and Lu concentrations using the Hf contents determined by electron microprobe. All grains were imaged using back-scattered electrons in a Zeiss Scanning Electron Microprobe to provide maps for the selection of appropriate spots for U-Pb and Lu-Hf analyses.
Zircon U-Pb dating. The zircon grains were mounted in epoxy discs and polished. Grain mounts containing the samples and zircon standards were cleaned in 2N nitric acid for ca. 1 hour prior to analysis. Analyses were performed using a custom built UV LAM [46] coupled to an Agilent 4500, series 300, ICPMS at the GEMOC Key Centre, Macquarie University. Samples and standard were ablated in a custom-built sample chamber using He to carry ablated material out of the sample cell and into the ICPMS. Relative to ablation in Ar, ablation in He minimises deposition of ablation products around ablation sites, improves sample transport e ciency, results in more stable signals and gives more reproducible Pb/U fractionation. To minimise dynamic U/Pb fractionation as the laser beam penetrates into the sample [47] all analyses were performed with the laser focused above the sample (typically 2 00 µm). Identical laser operating conditions (laser energy and degree of defocusing) were rigorously maintained throughout each run of 20 analyses to ensure constant U/Pb fractionation. It should be noted that the effect of any uncorrected U-Pb fractionation is to move data points along a chord corresponding to zero-age Pb loss or gain in the concordia plot, without affecting the 207Pb/206Pb ages. Ablation pit diameter was generally 40-50 µm. Samples were analysed in separate "runs" of 20 analyses comprising 12 analyses of unknowns bracketed, before and after, by 4 analyses of the GEMOC GJ-1 zircon standard, a gem quality zircon believed to be from an African pegmatite. Each analysis took ~180 seconds, with gas background measurements being taken over the rst ~60 seconds, prior to initiation of ablation. Data were acquired on ve isotopes using the instrument's time-resolved analysis data acquisition software with short dwell times to provide quasi-simultaneous measurement of the ve mases and optimum precision. 204Pb was not determined due to low count rates and the isobaric interference from Hg, which is a signi cant contaminant in the Ar supply gas. The time-resolved analysis software determines signal intensity data for each mass sweep performed by the mass spectrometer. Time-resolved signals (i.e., signals as a function of time, which is a proxy for ablation depth) generally allow isotopic heterogeneity within the ablation volume to be clearly identi ed (e.g., zones of Pb loss or common Pb related to fractures or areas of radiation damage; also inclusions, inherited cores, etc.). Raw data were processed using GLITTER, an in-house on-line data reduction program. 207Pb/206Pb, 208Pb/232U, 206Pb/238U and 207Pb/235U (235U = 238U/137.88) ratios were calculated for each mass sweep and the timeresolved ratios for each analysis were then carefully examined. Optimal signal intervals for the background and ablation data were selected for each sample and automatically matched with identical time intervals for the standard zircon analyses, thus correcting for the effects of ablation/transportrelated U/Pb fractionation and mass bias of the mass spectrometer. Net background-corrected count rates for each isotope were used for calculation of sample ages. The data in the result Tables show internal precision (1s) based on counting statistics on standards and sample. Concordia ages were determined using Isoplot 2.32 [48]. To monitor the quality of the data obtained with this technique, analyses of well-characterised zircons such as the Mud Tank zircon 734±32 Ma [49] and 91500 1064 Ma [50], have been run within each set separate "run" of 20 analyses. The U-Pb age data obtained for Mud Tank and 91500 zircon during this project are given by comparison with TIMS data. Common-Lead Correction: Traditionally, common-lead is measured by analysis of 204Pb, and the isotopic composition of lead corrected accordingly. However, LAM-ICPMS U-Pb analyses do not report 204Pb, because of the low peak/background ratio of the 204Pb peak due to the presence of Hg in the argon nebuliser gas. An alternative approach to common-lead correction of U-Pb data proposed by [51] neither requires knowledge of the amount of 204Pb present, nor assumes that corrected compositions plot on the concordia. Assuming that the observed discordance due to a combination of (1) lead-loss at a given time t L and (2) contamination by common-lead of known composition, the common-lead component is removed by stepwise subtraction of common 206 Pb c , 207 Pb c and 208 Pb c from the observed composition, until the residual composition matches the isotopic pattern of a lead which has only suffered lead-loss. The procedure allows the simultaneous determination of the amount of common-lead present and the corrected 206 Pb r / 238 U, 207 Pb r / 235 U and 208 Pb r / 232 Th ratios. This method is thus applicable to U-Pb analyses which do not report 204Pb, or do so only with inferior precision, and to grains which have suffered lead-loss in addition to contamination by common-lead. A more detailed description of the correction methodology is given by ref. [51].
Hf Isotope Determination. Hf isotope data for the same set of zircons were collected as described by ref [52]. Hf isotope analyses were carried out in situ with a Merchantek EO LUV 266nm laser-ablation microprobe, attached to a Nu Plasma multi-collector ICPMS at GEMOC, Macquarie University, Sydney. Most analyses were carried out with a beam diameter of 60-80 µm, a 10 Hz repetition rate, and energies of 0.6-1.3 mJ/pulse. Typical ablation times were 80-120 s, resulting in pits 40-60 µm deep. The methodology and analyses of standard solutions and standard zircons are described by ref. 7. Masses 172,175,176,177,178,179 and 180 were analysed simultaneously in Faraday cups; all analyses were carried out in static-collection mode. Data were normalised to 179Hf/177Hf = 0.7325, using an exponential correction for mass bias. Interference of 176Lu on 176Hf is small and was corrected using the recommended 176Lu/175Lu = 0.02669. Interference of 176Yb on 176Hf was corrected using 176Yb/172Yb = 0.5865; this value was determined by successive spiking of the JMC475 Hf standard (1 ppm solution) with Yb, and iteratively nding the value of 176Yb/172Yb required to yield the value of 176Hf/177Hf obtained on the pure Hf solution. Excellent precision and accuracy are obtainable even with large corrections on 176Hf [7]. All laser-ablation analyses were carried out using time-resolved data acquisition mode, in which the signal for each mass and each ratio is displayed as a function of time during the analysis. This allows the more stable portions of the ablation to be selected for integration, before the data are processed to give the nal results. Background was collected on peak for 45 seconds before ablation was initiated Declarations Page 10/21 ACKNOWLEDGMENTS CEG acknowledges the Center of Technological Development of the Geological Survey of Brazil for the support in this research. Part of this discussion was initially presented by CEG in December 2015 during an internal CMP workshop of the Geological Survey of Brazil, held in the city of Belem, and bene ted from estimated recent advances in geodynamics numerical modelling by several authors.

AUTHOR CONTRIBUTIONS
CEG, WLG and RW conceived the idea and wrote the paper. WLG and EB acquired the zircon U-Pb, Hf and trace element data. CEG and LBT assisted in the zircon U-Pb/Pb-Pb data compilation and organization. CEG, LLL, CML and LDC assisted in the whole-rock geochemical data compilation and organization.

ADDITIONAL INFORMATION
Competing interests: The authors declare no competing interests. Geological context of CMP region granitoids, shear zones and IOCG deposits. Simpli ed geological map modi ed from refs. [18,26]. Location of IOCG deposits after ref. [30].  Transition from drip to plate tectonics and fertilization of the non-convective lithospheric mantle in the CMP. A. Foundering and melting of crustal ma c drips due to convection in the underlying asthenospheric mantle [4]. The drip includes juvenile basalt that melts at variable depths producing the juvenile G1 TTGs.
Older shallower TTGs can also be produced at the base of the thick ma c crust [45]. B. With time and as a function of mantle convection and lid strength contrasts [31], the drip becomes asymmetric allowing for increased variety in magma compositions in the G3 suites (2.87-2.83 Ga). Transitional TTGs form by mixing basalt-derived melts with anatexis of older TTGs in the drip. Drip asymmetry allows TTG melts to rise into and metasomatize the non-convective lithospheric mantle in LILE, REE and HFSE. Fluids released from the dehydrating crustal part of the drip promote partial melting of the previously metasomatized lithospheric mantle to form hot, high-Mg sanukitoids. These hot magmas rise to trigger minor melting of the TTG crust producing the high-K granitoids typically associated with sanukitoids. These magma suites are emplaced during the initial deformation of the Itacaiúnas shear belt. C. Lithospheric transtension at 2.78-2.72 Ga reactivates the Itacaiúnas shear belt to form the localized Carajás rift basin, promoting local mantle upwelling and decompression melting, generating ma c-ultrama c layered intrusions and basalts within the Carajás basin. The under-and intraplating of these ma c hot melts trigger the remelting of the TTG proto-crust to generate the A-type, high-K granitoids of G3. At this stage, the same drip-metasomatized non-convecting lithospheric mantle that previously produced sanukitoids releases uids that use the translithospheric Itacaiúnas shear belt as conduits allowing them to reach shallower levels within the Carajás rift basin where they interact with basinal uids [30] generating the rst IOCG mineralization on Earth.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.