Main

Substantial accumulation of free oxygen in the atmosphere occurred between ~2.45 and 2.20 billion years ago (Ga) (refs. 1,2,3), with permanent atmospheric oxygenation between 2.3 and 2.2 Ga (refs. 4,5). This period is known as the Great Oxidation Event (GOE) and marks the most dramatic change in Earth’s surface chemistry and habitability6. However, it remains unclear if these major atmospheric changes affected the amount of free or chemically available oxygen in the mantle and, consequently, the redox state of mantle-derived magmas. In the modern Earth, considerable amounts of surface-oxidized components infiltrate the mantle via slab fluids and subducted sediments7, ultimately influencing the oxidation state of the mantle wedge and arc magmas8,9. A better understanding of how magmas changed oxidation through time would clarify ocean–atmosphere influence on mantle redox potential and whether deep-ocean oxygenation is a feature restricted to the Phanerozoic10,11.

Mantle oxygen fugacity (\(f_{\mathrm{O}_2}\)) probably changed in the early Earth as a result of metallic Fe retention during core formation and further homogenization12,13, but subsequent variations through time are debatable. The mantle \(f_{\mathrm{O}_2}\) is either described as largely unchanged14,15,16 or overall having a near-constant rate of increase through time17,18,19,20.

Common explanations for the absence of identifiable \(f_{\mathrm{O}_2}\) change in the mantle are based on the ‘infinite reservoir’ argument, given its relatively larger size compared to surface reservoirs (that is, atmosphere, hydrosphere and crust6,15). Nonetheless, mid-ocean ridge basalts show an increase in the Archaean potentially linked to early ‘whiffs’ of atmospheric oxygen18, suggesting that the upper mantle can be affected by changes in the atmosphere. However, despite evidence for recycling of continental materials during at least the past 3.0 Gyr (for example, ref. 21), it is unclear how much this has affected the redox state of the mantle, even across the GOE. Here we test the hypothesis of a change in the redox state of magmas in the sub-arc mantle region via recycling processes akin to subduction during the Palaeoproterozoic.

Oxygen fugacity and recycling in the geological record

One of the recognizable changes in the magma record that can provide insights into surface–mantle interactions is the transition from tonalite–trondhjemite–granodiorite (TTG) to sanukitoid magmas recorded in most cratons worldwide22,23. This transition started in the Palaeoarchaean and progressed with reduced occurrence to the Palaeoproterozoic22,24,25. TTGs are formed by partial melting at depth of (often) altered basaltic rocks, whereas sanukitoids require interaction of TTG such as melts and/or sediments within a metasomatized mantle wedge22,26. Sanukitoid petrogenesis thus constitutes prima facie evidence for crustal recycling into the upper mantle via ‘subduction’27. The tectonic settings for this secular magmatic transition are still debated23,28,29, but all models involve a mechanism that includes a return to the mantle of material that was previously at the Earth’s surface (for example, ref. 30).

Magmas can contain sulfur in variable valence states (S2−, S0, S4+, S6+), which are chiefly controlled by melting conditions and assimilation of surface material at source31,32. The preferential assimilation and fugacity variation of sulfur from mantle source to crustal emplacement is controlled by crystallization and/or degassing of the magma33. As a result, whole-rock analyses of sulfur speciation may involve a large uncertainty due to the volatility and instability of sulfur during and, especially, after crystallization. This problem is greatly reduced when melt or mineral phase inclusions are analysed as they are protected by the host crystals and therefore more likely to preserve pristine speciation through subsequent tectono-metamorphic-hydrothermal events34,35.

The accessory mineral apatite has great potential to record sulfur speciation as it can seamlessly incorporate both reduced and oxidized sulfur36,37 and so retain a proxy for the oxygen fugacity of the crystallizing magma38,39. Unfortunately, apatite grains easily recrystallize when the host rock is affected by hydrothermal fluids and/or deformation. Hence our approach includes analysing the sulfur speciation of apatite inclusions shielded in host zircon grains and of apatite grains present in the rock’s matrix. This avoids secondary alteration processes and offers a robust way to monitor the redox state of the primary magmas (for example, refs. 40,41,42). Importantly, U–Pb–Hf isotopes of the host zircon can be used to provide information on the age and the source (that is, mantle or crustal origin) of the magma from which they crystallized43, and stable oxygen isotopes can identify a supracrustal component and aqueous alteration at varying temperatures44,45. Zircon trace element analyses can also provide insights into the composition and redox state of host magmatic rocks46. Thus, combining a multi-isotopic and elemental study of zircon host minerals with extra information on the sulfur speciation of its apatite inclusions is a robust approach to study redox changes in magmas, applied here to those which are primary components of the continental crust.

A Palaeoproterozoic magmatic transition

We investigate samples from two well-characterized suites of intermediate granitoids from the juvenile Mineiro Belt, Brazil. These magmatic suites were classified as a 2.35 Ga TTG and 2.13 Ga sanukitoid based on field evidence, geochemistry and isotope studies, both in whole-rock and accessory phases24,47,48,49. They are particularly suitable for this study because they record the youngest transition from TTG to sanukitoid magmatism on Earth, taking place in the Palaeoproterozoic rather than in the Archaean as in many other cratons23,24. The transition started shortly before the global ‘tectono-magmatic lull’50 and evolved across the time period of irreversible oxygenation of Earth’s atmosphere5.

Isotopically, both whole-rock and individual zircon analyses show that the TTGs plot above the chondritic reservoir value (CHUR) and that the sanukitoids have approximately chondritic values24,49. This indicates that both magmatic suites have a ‘juvenile’ signature43. Moreover, a projected crustal evolution line typical of granitic crust (176Lu/177Hf = 0.01) demonstrates that the sanukitoid suite can be traced back 200 Myr to the TTG composition, and in turn, those two can be traced back to a depleted mantle (DM) source at around 2.5 Ga (Fig. 1a). Thus, shortly (150 Myr) after mantle extraction, the mafic parental source first underwent partial melting and generated the 2.35-Gyr-old TTGs48. A similar mafic source was remelted about 200 Myr later47, but this time the magmas record increased depth of melting, as a result of steepening of the subducted slab and opening of a mantle wedge24. This second stage generated 2.13 Ga sanukitoid magmas that contain abundant mafic enclaves, with geochemical and oxygen isotope analyses suggesting the involvement of sediments24 (Fig. 1b). The absence of any zircon inheritance and chondritic zircon ɛHf(t) values (176Hf/177Hf ratios relative to chondrite at the time of crystallization) alongside substantial range in zircon δ18O (+4.7‰ to +7.5‰ deviation from the 18O/16O ratio of the Vienna Standard Mean Ocean Water) suggests that crustal assimilation is unlikely and the sediments originated from the 2.13 Ga magmatic arc itself—a geological phenomenon described as a ‘self-feeding arc’, where arc sediments form, are deposited in the adjacent ocean and then are subducted24 (Fig. 1c). The high concentration of elements that are strongly mobile in aqueous fluids (for example, Ba and Sr > 1,000 ppm) alongside enrichment in mantle-compatible elements (Mg number: ~55, Ni: ~80 ppm and Cr: ~200 ppm; ref. 47) strongly support sediment admixture in the sub-arc mantle wedge, where sanukitoid magmas were sourced.

Fig. 1: Palaeoproterozoic TTG–sanukitoid magmatic transition.
figure 1

a, Zircon U–Pb ages versus 176Hf/177Hf(t) ratios (expressed as ɛHf(t) values relative to chondrite at the time of crystallization t). Zircons from TTG magmas (n = 31) have significantly positive ɛHf(t), whereas zircons from the sanukitoid magmas (n = 33) are near the CHUR. A crustal evolution line links both suites of rocks to a DM melting event at ~2.5 Ga. b, Zircon 18O/16O ratios (expressed as δ18O relative to Vienna Standard Mean Ocean Water) show that the basaltic crust was hydrothermally altered at high temperature (~4.5‰) before generating TTG magmas at 2.35 Ga and before remelting in the metasomatized mantle wedge at 2.13 Ga. The latter event generated sanukitoids that have zircons with heavier oxygen (~6.5‰). Individual error bars in a and b are shown at 2 standard errors. c, Tectonic model for the generation of magmas in the Palaeoproterozic pre- and post-GOE peak. SCLM stands for subcontinental lithospheric mantle.

Full size image

The magmas described above evolved within a geologically short duration (2.5–2.35–2.13 Ga) in an environment akin to intra-oceanic arcs and away from continental sources. Their evolution was nonetheless long enough to witness variations in atmospheric oxygenation consequent upon the GOE. In this context, the intrinsic relationship of the suites as derived from a similar basaltic source remelted at different geological intervals47,48,49 suggests that comparative changes of fugacity are feasible.

Sulfur speciation in matrix apatite and inclusions

We have determined the valence(s) of sulfur in apatite using the X-ray absorption near-edge structure (μ-XANES) technique at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France (Methods and Supplementary Data). We present 40 analyses in 26 matrix apatite grains and 136 analyses in 20 apatite inclusions in zircons. For all studied samples, we targeted crack-free apatite inclusions that show no crystallographic misorientation and display a ‘hotspot’ zone in sulfur concentration maps (Fig. 2). The patterns show higher sulfur concentration in the centre of the inclusions and gradual decrease towards the host zircon crystal. This core-to-rim zonation is taken to represent sulfur incorporation and retention during early crystallization and thus records primary signatures of the crystallizing magma. It is unlikely that they represent sulfur exchange with the host, as zircon has negligible sulfur content and extremely slow diffusion rates51.

Fig. 2: Backscattered electron images of inclusions in zircon and sulfur speciation spectra.
figure 2

a–b, Twenty-nine μ-XANES analyses in six apatite inclusions from the TTG sample and 108 analyses in 14 inclusions from the sanukitoid sample show that the average spectra differ greatly from a reduced (high S2−) speciation in TTG host zircons (a) to an oxidized speciation (high S6+) in sanukitoid host zircons (b). Inset diagrams show the crystallographic orientation of the inclusions and their relative sulfur concentration. Sample labels are indicated in the top left corner of each image.

Full size image

Although many apatite inclusions from the 2.35-Gyr-old TTG suite have proven to be sulfur barren, six inclusions and a few matrix apatites have sufficiently high sulfur to yield measurable μ-XANES spectra (Fig. 2); these show remarkably lower S6+/STOTAL (that is, they have dominant S2− peak) compared to the apatite sulfur speciation in the 2.13 Ga sanukitoid (Fig. 3). Matrix apatites and three out of six inclusions have much larger uncertainties and S6+/STOTAL between 0.4 and 0.7, while analyses of the other inclusions show S6+/STOTAL < 0.1 (Fig. 3). As matrix apatites are more prone to recrystallization for not being shielded in zircon, we suggest that these three inclusions with much higher S6+/STOTAL were affected by post-crystallization interaction with fluids in relative higher fugacity. Moreover, these inclusions with higher S6+/STOTAL have no detectable ‘hotspot’ zoning, indicating sulfur redistribution or concentrations straddling the detection limit. One of the analysed inclusions with very low S6+/STOTAL is hosted by a zircon grain that contains an ilmenite inclusion, reinforcing the interpretation of magmatic crystallization under possible reduced conditions (Fig. 2). Ilmenite and magnetite occur in the matrix, but magnetite is thought to have been produced during recrystallization of the host rock during deformation/metamorphism48, a process that apparently did not affect the armoured inclusions.

Fig. 3: Oxygen fugacity of the samples studied with respect to the fayalite–magnetite–quartz buffer.
figure 3

a, ΔFMQ calculated via S6+/ΣSTOTAL oxybarometer proxy. Large dark diamonds are averages of μ-XANES analyses within a single inclusion (sample size as indicated in Fig. 2 caption). Averages of 15 and 25 analyses, respectively, in six matrix apatite grains from the TTG sample (red) and 20 from the sanukitoid samples (blue) are plotted as smaller lighter diamonds. Error bars are 2 standard deviations. b, ΔFMQ calculated via zircon trace element proxy versus U/Yb ratios of individual zircon grains from sanukitoid (n = 20) and TTG (n = 23) samples. Error bars are 2 standard errors.

Full size image

Our results show that matrix apatite grains from the 2.13 Gyr sanukitoid depict a conspicuous oxidized pattern with dominant S6+ and high S6+/STOTAL peak area ratio that is in agreement with apatite inclusions from the same sample (Fig. 2). The high S6+/STOTAL for both inclusions and apatite grains in the rock matrix suggests that if any recrystallization event has occurred, it was similarly oxidizing if compared to the primary fugacity of the inclusions (Fig. 3). Moreover, magnetite is a common opaque mineral phase in these rocks, formed as isometric homogeneous crystals included in hornblende47, which overall points towards primary oxidizing crystallization conditions.

Quantifying changes in oxygen fugacity

To determine \(f_{\mathrm{O}_2}\) and ascertain the magnitude of change, we used a sulfur-in-apatite experimental calibration oxybarometer that combines the fayalite–magnetite–quartz buffer (FMQ) and S6+/STOTAL (ref. 39) (Fig. 3a). Additionally, we used trace element analyses of the host zircons in a proxy that correlates zircon composition to oxygen fugacity relative to FMQ46. The \(f_{\mathrm{O}_2}\) distance from the FMQ buffer (ΔFMQ) is then plotted against the zircon U/Yb to relate to sediment input at the source and/or crustal contamination of the magmas52 (Fig. 3b).

Our results demonstrate that the isotopically and petrologically linked TTG and sanukitoid suites in the Mineiro Belt of Brazil crystallized under different redox conditions (Fig. 3). The 2.35-Gyr-old TTG formed when atmospheric oxygen levels fluctuated across a threshold of 10−5 of present atmospheric levels53 and primary apatite inclusions are reduced, with the three most precise measurements ranging from ΔFMQ –0.75 ± 0.30 to –0.28 ± 0.38 (2σ standard deviation—average of several analyses within each inclusion; Fig. 3a). The younger 2.13-Gyr-old sanukitoid post-dates the GOE when atmospheric oxygen levels stabilized permanently between 0.1 and 40% present atmospheric levels54, and primary inclusions show oxidized sulfur speciation ranging from ΔFMQ + 0.80 ± 0.16 to +1.24 ± 0.36 (Fig. 3a). Despite some intergrain variation, trace element analyses in zircons confirm that the 2.13-Gyr-old sanukitoid suite is more oxidized than the 2.35-Gyr-old TTG and corroborates the sulfur speciation data in apatite inclusions (Fig. 3b). The higher U/Yb in zircons from the sanukitoid suite is in agreement with elevated δ18O and is attributed to surface-derived oxidized sediments that metasomatized the mantle.

An atmospheric window to the mantle

Modern marine sediments can be highly oxidized compared to upper mantle-derived basalts8. Their subduction will carry this surficial redox signature and affect the oxidation state of modern arc magmas7,9. The generation of a deep subduction-like setting in the Palaeoproterozoic would have facilitated sediment input into the sub-arc mantle source (>10% by volume)24. These sediments may have carried sulfate and ferric iron into the mantle wedge in a similar way to that observed in Phanerozoic arcs10,11. Oxidized sanukitoid magmas were then formed at 2.13 Ga during melting of a 2.5-Gyr-old altered oceanic slab and its sedimentary carapace (Fig. 1). We suggest that this change in tectonic style allowed the atmosphere to dramatically influence sub-arc mantle chemistry through transfer of oxidized sediments (for example, refs. 7,9). The fortuitous occurrence of a TTG–sanukitoid transition in the Mineiro Belt across the GOE has allowed this change in oxidation state to be observed.

The global TTG–sanukitoid transitional record characterizes vigorous and protracted crustal recycling and mantle metasomatism in the early Earth (for example, refs. 22,26). Although localized and with less magnitude than the GOE, Archaean ‘whiffs’ of oxygen may have similarly altered the nature of subducted sediments55 and started to modify the oxygen fugacity of mantle sources in the Archaean (compare with ref. 42). It then suggests that similar mantle–surface interactions could have had a major influence on magmatism and metallogenic endowment since the Archaean56 and that the oxidation of the mantle was diachronous and related to the onset of global plate tectonics. The subsequent irreversible step-rise of oxygen in the Palaeoproterozoic may have led to the dominance of characteristically oxidized calc-alkaline magmas in the geological record.

Methods

Sample preparation

Zircons were hand-picked and mounted on tape on a 5 × 5 mm area. A 25-mm-diameter round mount was then centred on top of the grains where epoxy resin was poured and dried in a vacuum chamber. The resin mounts were then polished to reveal the zircon interiors. Cathodoluminesce and backscatter electron images for U–Pb, oxygen and Lu–Hf analyses targeting were acquired at the University of Portsmouth. They show oscillatory zoning typical of magmatic grains24,49. Host grains and inclusions were identified using elemental mapping by scanning electron microscopy-energy dispersive X-ray spectrometry. The spectral composition and ‘spot’ identification of the different phases were determined using a silicon drift (SDD-EDS) Oxford X-max 80 mm2 detector attached to a SEM Zeiss EVO MA 10 LaB6 at the University of Portsmouth.

Synchrotron data acquisition and data processing

Sulfur speciation analyses were carried out using the ID21 X-ray microscopy beamline at the ESRF in Grenoble, France57. During the session, the sulfur speciation (S−2, S4+ and S6+ and their proportions) was measured in both apatite grains and apatite inclusions in zircon. X-ray beam was produced using a 42-period undulator, and the energy was selected with a double-crystal monochromator. XANES spectra were acquired in the range of 2.46 to 2.53 keV, in continuous mode, with steps of 0.2 eV and 0.1 s per point, taking approximately 1 minute per point. The calibration of the double-crystal monochromator was done using a gypsum reference powder (maximum of white line of the XANES spectrum at 2.48184 keV). The fine tune and focusing was achieved using a set of two Kirkpatrick–Baez mirrors with a Ni coating. The beam was reduced to 0.3 μm V × 0.4 μm H. Sample mounts were mounted vertically in the microscope, and all measurements were carried out under vacuum (10−5 mbar). The beam was then used for construction of 2D μXRF elemental maps and μ-XANES spectra for chemical speciation. The detector used was an SGX 100 mm² SiriusSD silicon drift diode and a multichannel analyser. Dead time of the XRF detector was controlled, and beam intensity was attenuated so that dead time was always in the range of 17–22% throughout the session. XANES spectra were produced after correction by beam intensity (constantly measured with a photodiode upstream the sample) and the XRF dead time. Zircon host grains were analysed a few times to assess possible interferences. To separate the XRF K lines from S from the Zr L2 and L3 emissions lines, XRF spectra were batch fitted using the PyMca software58. Inclusions were located thanks to μXRF map and appeared as ‘hotspot’ domains with high sulfur contents. To increase the signal to noise of XANES spectra, several points of interest were selected over each inclusion, and an average spectrum was obtained. This approach enables focusing the targets/regions rich in sulfur and optimizes the time spent for each sample.

The ESRF S K-edge XANES spectra inorganic database was used to identify the different peak energy positions for natural minerals with distinct sulfur speciation. To identify S6+(~2,482 eV), S4+ (~2,478 eV) and S2−(~2,470 eV), we used gypsum, pyrrhotite and pyrite. The high-focused beam yields high sulfur X-ray counts (in the order of ~6,000 counts per s for S6+ when measuring gypsum and ~4,500 counts per s for S2− when measuring pyrite). The same Durango crystal analysed in ref. 38 was also analysed during the beginning and at the end of the session and the spectra was reduced as per the unknown apatites. As a similar procedure developed in ref. 38, we analysed across the grain perpendicularly to its c axis. Those comprise ten points measured twice each, testing homogeneity and beam damage. A single point at the core of the crystal was further analysed ten times and matched individual measurements, reassuring the minor existence of the beam damage for S6+ and S2− peaks59. Additionally, we analysed the same grain after a rotation of 90 degrees to assess the possible effect of the crystallographic orientation of the apatite. No differences were noticed in the spectra. The same experiment was performed with grain AMT_03_Zr12b that contains inclusions 1 and 2. XANES data filtering and reduction was done using Python scripts and PyMca 5.5.5 software for visualization and selection of the reduction parameters. First, XANES spectra were filtered to exclude noisy spectra. Only spectra with a white line intensity that were at least two times higher than the signal measured at the pre-edge background were considered. XANES normalization values were 2.47006 keV for the edge and −0.008840, −0.004420 keV for pre-edge and 0.004550, 0.045496 keV for the post-edge. Spectra were fitted as a linear combination of Gaussian peaks, using a Python code and the advanced fitting from PyMca for Gaussians curves. Finally, the script extracted the peak area ratios (S6+/STOTAL). The average integrated S6+/STOTAL peak area ratio obtained for Durango is 0.956 ± 0.004 (2 standard error), in agreement with the peak area ratio obtained in ref. 38 (0.956 ± 0.002; 2 standard error).

Electron backscatter diffraction

Electron backscatter diffraction (EBSD) analyses were performed in zircon, apatite and apatite inclusions to evaluate the microstructural relationship of these phases and the possible relationship between sulfur speciation and crystallographic orientation. Analyses were performed at Géosciences Montpellier, France, using a CamScan X500-FEG Crystal Probe. This SEM was developed from its conception for high-resolution crystallographic mapping by backscattered electron diffraction. It thus has the particularity of having its electronic column tilted at 70°, which makes it possible to keep the sample horizontal during analyses, and allows backscattered electron diffraction detection of surfaces with uneven polishing. The EBSD Symmetry detector of CMOS technology was used for crystallographic analysis and a detector EDS UltimMax 100 was used for chemical analysis. Reduction of the data was performed using the AZtec software of Oxford Instruments. The crystallographic data were processed using Tango for orientation map display. Noise reduction was applied to the raw crystallographic data and the final EBSD maps contain the band contrast for quality of the diffraction pattern and lattice orientation of each phase analysed. Within each phase analysed, no misorientation >3° was observed, suggesting high crystallinity of individual phases and that processes of recrystallization are unlikely to have occurred. The small misorientation observed is possibly caused by polishing and hardness of the apatite inclusions against the zircon host. Nonetheless, these features do not compromise the refraction of part of the inclusions and are sufficient enough quality to assess crystallographic orientation. We analysed apatite inclusions with variable crystallographic orientations, which show no correlation in their sulfur μ-XANES spectra (Fig. 2).

Zircon trace element analysis

Trace element composition was obtained by laser ablation inductively coupled plasma mass spectrometry on distinct zircon domains. Analyses were conducted using an ASI RESOlution Excimer 193 nm laser system, coupled to an Analytik Jena PlasmaQuant Elite ICP-MS at the University of Portsmouth, United Kingdom. Laser ablation settings for the first session were 30 μm, 3 Hz and 3.5 J cm−2. Analyses were located on top of previous electron microprobe or U–Pb analyses. Oxide formations in the plasma were estimated by monitoring the ThO+/Th+ and UO+/U+ on the National Institute of Standards and Technology (NIST) glass standard reference material 612, and tuning was done to reach values below 0.3% and 0.2% for the ThO+/Th+ and UO+/U+ ratios, respectively. Elemental fractionation was monitored through repeated measurements of the 232Th/238U ratio in NIST 612, with variations between 0.9 and 1 observed during analytical sessions. He carrier gas flow in the ablation cell was set to 0.310 l min−1 and combined with 0.003 l min−1 of N2 gas before introduction to the ICP. Plasma generation was set at 1,300 W with 10 l min−1 of Ar plasma flow, auxiliary gas flow of 1.65 l min−1 and nebulizer flow between 0.89 and 0.9 l min−1. The glass NIST 612 was used as a primary standard and zircon reference materials 9150060, GJ-161,62, Plešovice63 and Temora-264 were used as secondary standards. The glass NIST 612 was analysed four times at the beginning and at the end of the session and twice each group of unknowns. Each group of unknowns consisted of a combination of four investigated zircons and two secondary standards. The following isotopes were measured and had an integration time of 10 ms: 25Mg,27Al, 29Si, 31P, 39K, 43Ca, 45Sc, 49Ti, 51V, 53Cr, 55Mn, 85Rb, 87Sr, 89Y, 90Zr, 93Nb, 95Mo, 118Sn, 121Sb, 133Cs, 137Ba, 139La, 140Ce, 141Pr, 146Nd, 147Sm, 153Eu, 157Gd, 159Tb, 163Dy, 165Ho, 166Er, 169Tm, 172Yb, 175Lu, 178Hf, 181Ta, 182W, 208Pb, 209Bi, 232Th, 238U. The raw data were reduced and corrected for instrumental drift using the Iolite 3.4 software65. Zirconium content was used for internal calibration and measured using a Cameca SX100 electron microprobe at the University of Bristol, United Kingdom (below), avoiding using stoichiometry when normalized. NIST 610 and secondary zircon reference materials yielded results within 10–15% in accuracy relative to the recommended published values.

Pressure and temperature dependence

Experimental studies in silicate melts demonstrate that an increase in 100 °C temperature imparts a ΔFMQ –0.5 deviation in the magma \(f_{\mathrm{O}_2}\); whereas an increase in 0.3 GPa would impart a ΔFMQ + 0.2 deviation66,67. Direct measurement of temperature crystallization via Ti-in-zircon thermometry68 indicates that zircon grains from the sanukitoid samples crystallized at 800 ± 13 °C (2 standard error), and TTG samples have zircons with average crystallization temperature of 700 ± 10 °C (2 standard error), considering SiO2 and TiO2 activity = 1. A lower TiO2 activity such as 0.7 yields higher temperatures of crystallization, 839 °C and 726 °C, respectively, for sanukitoid and TTG zircons. Al-in-hornblende oxybarometer indicates crystallization pressures at around 0.6–0.4 GPa for the sanukitoid samples47, whereas higher pressure conditions were suggested for the TTG samples (~0.7 GPa) based on the presence of igneous almandine garnet and overall trace element modelling48. The ~0.3 GPa difference in pressure between these two magmatic suites is not enough to explain a > 1.0 ΔFMQ difference between the calculated \(f_{\mathrm{O}_2}\) for sanukitoid and TTG magmas. Moreover, the ~100 °C difference in temperature, if considered, would amplify the contrast in fugacity obtained via sulfur oxybarometry.

Electron probe microanalysis

Zircon grains were analysed by electron probe microanalysis at the University of Bristol, United Kingdom, using a Cameca SX100. An electron beam of 5 µm was used with an acceleration voltage of 17 kV and 100 nA beam current. Element concentrations (Si, P, Y, Zr, Yb, Lu, Th and Hf) were determined using thallium acid phthalate, large pentaerythritol, large lithium fluoride and pentaerythritol crystals. The 91500 zircon standard60 was analysed during the microprobe sessions to monitor data quality. Only analyses with concentrations more than the sum of the individual detection limit and the absolute 2σ error were considered. The Zr content was used for internal calibration of trace element analyses presented above and vary from 47 to 48%.