In the near-absence of a physically-preserved rock record prior to 4 Ga, insights into the extent and timing of earlier crust-mantle differentiation processes can either be obtained from scarce remnants of Hadean rocks (e.g. detrital zircons and their inclusions)1,2,3,4 or using the geochemical and isotope composition of Archaean rocks. This chemical differentiation history ultimately reflects mantle dynamics and lithosphere tectonics during that time, which, due to a general lack of direct observations, is strongly reliant on numerical models5,6. To date, a wide range of tectonic modes have been defined that differ with respect to the cooling rate of a planet and the ability to cycle elements between its interior and exterior through volcanic and tectonic processes7. In active lid modes, such as modern-day plate tectonics, parts of the lithosphere sink back into the planet’s interior, actively participating in mantle convection, and thus driving mantle cooling as well as continuous crustal growth and recycling8,9. The extent to which plate tectonics was operating in the early Earth remains widely debated, as higher mantle temperatures are thought to impede long-lived subduction and favour instead a stagnant-, sluggish-, or episodic-lid regime7,10,11,12,13. In the stagnant-lid mode, the lithosphere constitutes a single plate that does not take part in convective overturn, so that crustal rejuvenation occurs by other mechanisms such as heat-pipe volcanism and plume tectonics14,15. This mode predicts inefficient crustal recycling and shallow upper mantle convection, consistent with the long-term preservation of primordial mantle heterogeneities16. In a sluggish lid mode, the velocity of the mantle below the lithosphere exceeds surface velocities, resulting in slow plate motions and crustal recycling13,17. In the episodic-lid mode, stagnant phases alternate with short phases of extensive magmatic activity, leading to periodic mantle overturns and large-scale crustal rejuvenation10,18,19.

Geodynamic models of the early Earth show varied outcomes for the longevity of the Hadean lithosphere, the occurrence and frequency of mantle overturns, and the timescales associated with convective re-homogenisation of differentiated mantle reservoirs16,20. Reconstruction of the differentiation history of the Hadean crust-mantle system can provide key temporal constraints on the tectonic evolution of the early Earth that would otherwise remain elusive due to the limited geological record. Short-lived radio-isotope systems, such as the 146Sm–142Nd chronometer, are particularly suited for this purpose. Due to its short half-life (t1/2 = 103 Ma), the 146Sm-142Nd system selectively records crust-mantle differentiation events that took place >4 Gyr ago. Decay of 146Sm in differentiated silicate reservoirs translates into heterogeneities in the abundance of its daughter isotope (142Nd), which in turn can be transferred to the newly formed crust by re-melting of these ancient reservoirs. Using this approach, 142Nd signatures reflecting the formation of both chemically enriched (crustal) and depleted (mantle) Hadean reservoirs have been identified in mantle-derived rocks from the oldest Archaean cratons21,22,23,24. Subsequent investigations of diverse Eo- to Neoarchean rocks25,26,27,28,29,30,31,32 showed that the earliest stages of Earth’s history were marked by extensive crust-mantle differentiation at 4.40 ± 0.03 Ga32. The apparent uniqueness of this event has been suggested to reflect a relatively quiescent evolution of the post-magma ocean Earth16,26, or the extraction of large volumes of crust ca. 4.4–4.5 Gyr ago followed by rapid crustal recycling and rejuvenation through the Hadean33. Models advocating continuous crustal growth and recycling which require the 142Nd composition of the depleted mantle to have gradually declined with time, are only weakly supported by the Archaean 142Nd record32. However, the majority of 146Sm–142Nd studies have so far focused on Eoarchean rocks, and comparatively little is known about the evolution of the depleted mantle through the Archaean. Here we use combined 147,146Sm–143,142Nd systematics to show that the Paleoarchean mantle preserved large-scale heterogeneities related to a period of increased magmatic differentiation several hundred million years after crystallisation of the terrestrial magma ocean. Considered together with previous constraints, the Archaean 142Nd record points towards at least two distinct episodes of Hadean crust-mantle differentiation. These events are also reflected in the Hf-isotope evolution of the oldest zircon grains. This concomitance of mantle depletion and crust extraction suggests that differentiation in the Hadean proceeded via distinct large-scale tectonomagmatic events that led to the creation of long-lived chemical heterogeneities in Earth’s early mantle.

Results and discussion

For this study, igneous rocks from the Paleoarchean to Mesoarchean Bastar and Singhbhum cratons in India were investigated. Tonalite–trondhjemite–granodiorite (TTG) suites were sampled on a regional scale in the central to the western part of the Bastar Craton. Sample descriptions, locations, and zircon U-Pb ages ranging from about 3.45 to 3.55 Ga, are given in Maltese et al.34. Samples from the Singhbhum Craton were collected from the Older Metamorphic Tonalite Gneiss and all units of the Singhbhum Granite (Phase I to III). Sample locations, major and trace element compositions, and 147Sm-143Nd isotope data are published in Pandey et al.35. Zircon U-Pb ages are reported in Upadhyay et al.36 and range from 3.47 to 3.29 Ga. Lutetium-Hf analyses on the same zircon grains by laser ablation inductively coupled plasma mass spectrometry yielded positive ε176Hfi values up to +4, indicating the existence of a depleted mantle domain in the source region of the Singhbhum Craton granitoids37. In addition to the granitoids, samples of komatiitic, basaltic, and andesitic rocks from the Gorumahisani-Badampahar Iron Ore Group (IOG)38, a supracrustal/greenstone belt near the eastern margin of the craton were also analysed. Investigated samples are those of Pandey et al.35, who reported a whole-rock 147Sm–143Nd regression corresponding to 3.75 ± 0.34 Ga, in broad agreement with the emplacement age of 3510 ± 3 Ma inferred from U-Pb zircon dating39.

Results of 146,147Sm-142,143Nd isotope analyses and age constraints are summarised in Supplementary Data 1 and displayed in Figs. 1 and 2. Following usual conventions, 142Nd/144Nd and initial 143Nd/144Nd are reported as relative deviations from the modern silicate Earth in parts per million (μ142Nd) and from the chondritic uniform reservoir40 in parts per ten thousand (ε143Ndi), respectively. The isotope composition of the Bastar granitoids range from −0.9 ± 2.8 ppm to 3.9 ± 3.2 ppm for μ142Nd and from −1.3 to +1.3 for ε143Ndi, but have an average composition close to that of the primitive mantle (Supplementary Data 1). In contrast, the isotope composition of the Singhbhum granitoids displays a spread in μ142Nd values ranging from −2.0 ± 3.6 ppm to +6.3 ± 4.6 ppm which are positively correlated with ε143Ndi143Ndi = −0.5 to +5.1). No clear relationship is observed between the 142,143Nd composition of the granitoids and their U-Pb ages or bulk-rock chemistry. Specifically, five samples with μ142Nd values indistinguishable from bulk silicate Earth also have ε143Ndi within ±1 ε-unit of the chondritic value, while four samples that exhibit a resolved positive μ142Nd anomaly (i.e., +3.9 to +6.3) have positive ε143Ndi values (i.e., +2.5 to + 5.1). Similarly-radiogenic initial Nd isotope compositions are obtained for the mafic and ultramafic rocks from the 3.51 Ga eastern IOG, with μ142Nd between +1.2 and +6.3, and ε143Ndi between +0.9 and +3.8 (Supplementary Data 1). The 146Sm–142Nd and 147Sm–143Nd data, therefore, provide concordant results for all investigated units of the Singhbhum and Bastar cratons—the granitoids of the Bastar Craton bear no record of early differentiation, whereas part of the Singhbhum granitoids and associated supracrustal rocks of the IOG formed from isotopically heterogeneous material originally derived from mantle domains depleted in the Hadean (Fig. 2).

Fig. 1: Model ages of mantle depletion.
figure 1

Isochron diagram showing the 142Nd isotope composition of Bastar and Singhbhum samples against the (147Sm/144Nd)src ratio of their respective mantle sources. (147Sm/144Nd)src is obtained from two-stage isotope evolution modelling of coupled 146,147Sm-142,143Nd systematics (Supplementary Methods). Additional data from Archaean cratons (open circles) mark the onset of global crust-mantle differentiation at 4.4 ± 0.03 Ga (references in the text). Grey-shaded envelopes mark model age uncertainties. Error bars are 2 SE for single analyses and 2 SD for replicated analyses (Supplementary Data 1). Reference lines for bulk silicate Earth (BSE) are shown for context.

Fig. 2: 147Sm–143Nd and 146Sm–142Nd isotope evolution of the Singhbhum mantle source and petrogenesis.
figure 2

The tiles illustrate two possible petrogenetic scenarios for the formation of mafic rocks from the IOG greenstone belt and TTGs from the Singhbhum Craton. Panels a and b consider only the isotope composition of the source melts such that the vertical spread defined by TTGs in Nd isotope vs. time space reflects mixing of melts derived from both depleted and primitive mantle (PM) domains. Panels c and d also consider the petrogenesis of TTGs which were likely produced by re-melting of a mafic protolith, here represented by the 3.51 Ga supracrustal rocks from the eastern IOG belt. In this scenario, the most radiogenic Nd isotope compositions require a so-far elusive mafic precursor extracted from the depleted mantle shortly before formation of the granitoids. The distinct evolution of the Isua depleted mantle reservoir is shown for comparison.

Evidence for late Hadean differentiation

The differentiation history of the Paleoarchean Singhbhum TTGs and supracrustal rocks from the IOG can be further constrained by coupled 146,147Sm–142,143Nd chronometry, assuming a two-stage model as a first-order approximation of the evolution of their mantle source22,32. The first stage corresponds to the evolution of a chondritic reservoir from ~4.57 Ga (T0) to an instantaneous differentiation event at TD that produced a depleted mantle reservoir (DM). The second stage describes the closed-system evolution of this source reservoir from TD until formation of the Singhbhum Craton at TS as a function of its Sm/Nd ratio (147Sm/144Ndsrc). By combining the two chronometric equations of the 146,147Sm–142,143Nd system, both the time of differentiation of the depleted mantle as well as its time-integrated 147Sm/144Ndsrc can be estimated (see ‘Methods’ section, Supplementary Fig. S-2). This allows eliminating uncertainties of conventional model ages related to the composition of the parent Hadean mantle reservoir. The two-stage model indicates differentiation of the Singhbhum TTG source reservoir at \({4.19}_{-0.12}^{+0.06}\) Ga, with 147Sm/144Ndsrc ranging from near-chondritic to 0.25 (Fig. 1). Samples from the IOG belt yield a similar model age and time-integrated 147Sm/144Ndsrc, albeit with larger uncertainties, likely caused by imperfect preservation of their 147Sm–143Nd systematics or improper age assignment (see Supplementary Discussion). Despite these limitations, the model age estimated for the IOG supracrustal rocks is broadly similar to that inferred from the Singhbhum granitoids (Fig. 1 and Supplementary Fig. S-1). These chronological constraints are distinct from those obtained by coupled 146,147Sm–142,143Nd systematics from Eoarchean terranes of southwest Greenland and northern Labrador which yielded an older model age of 4.40 ± 0.03 Ga together with a lower time-integrated 147Sm/144Ndsrc of 0.2132. Although more complex evolution scenarios cannot be excluded, the concordant 142,143Nd model ages obtained from various Archaean cratons suggest that despite these simplified assumptions, a two-stage differentiation history constitutes a generally valid framework that can approximate the evolution of Hadean silicate reservoirs. The simplest interpretation, therefore, is that the isotope signatures of the Singhbhum rocks reflect a late Hadean mantle depletion event while the Bastar Craton formed from more primitive mantle domains that bear no record of Hadean differentiation (Fig. 2). Collectively, these results demonstrate that the Archaean mantle comprised both large-scale primitive and chemically depleted domains. Some of these depleted domains were produced during or shortly after the end of terrestrial accretion, possibly as a result of magma ocean crystallisation, while others, like the Singhbhum source region, may reflect crust extraction in the late Hadean.

A coeval record of crustal extraction and mantle depletion

Geochemical evidence for Hadean crust formation can be found in the Hf-isotope record of Hadean and Eoarchean zircon1,2,3,4,41. Previous studies showed that the majority of Hadean/Eoarchean zircon grains define an array in ε176Hfi vs. time space (with values becoming more negative with decreasing age) which is interpreted to reflect reworking of Earth’s primordial crust that differentiated 4.4–4.5 Gyr ago3,4,41 (Fig. 3). This event is also recorded in rare zircon xenocrysts in Paleoarchean TTGs and detrital zircon from younger siliciclastic successions from the Singhbhum Craton41,42. As shown in Fig. 3, this event is well reflected in the crust-mantle differentiation ages of ~4.4 Ga inferred from the 146,147Sm–142,143Nd systematics of Eoarchean rocks. In addition, several studies pointed out the presence of a population of Hadean grains, at 3.9–4.1 Ga, with ε176Hfi above plausible evolution lines for a 4.4–4.5 Gyr old crustal reservoir3,4 (Fig. 3). This bulge in the ε176Hfi vs. time array requires differentiation of crustal protoliths from juvenile magmas produced in the late Hadean and subsequent evolution of this crustal domain with a time-integrated 176Lu/177Hf ratio between that of mafic and felsic rocks (Fig. 3). A similar late Hadean (~4.2 Ga) crustal extraction age was also inferred for the precursor of the ca. 3.9–4.0 Ga Acasta gneisses43,44,45,46 (Fig. 3), in agreement with the identification of a 4.203 ± 0.058 Ga zircon xenocryst in a granitic rock of the complex47. As is the case for the differentiation event at ~4.4 Ga, crust generation at ~4.2 Ga is also reflected both in the extinct nuclide record of mantle depletion and in the surviving relics of the Hadean rock record. This observation seemingly conflicts with the notion that Hadean crustal extraction ages reflect sporadic preservation of small crustal domains33, as this would not necessarily involve the preservation of their complementary depleted reservoir in the convective mantle. Rather, the record of coeval crust extraction and mantle depletion points to the occurrence of at least two major crust forming events in the Hadean. These events produced depleted mantle domains large enough to escape complete re-homogenisation on a billion-year timescale.

Fig. 3: Comparison of Nd and Hf model ages of mantle depletion and crustal extraction.
figure 3

a Summary of published model ages of crust-mantle differentiation inferred from coupled 146,147Sm-142,143Nd chronometry32,56. b Compilation of zircon Hf-isotope compositions and associated 2 SD external reproducibility; detrital zircon data from the Jack Hills metaconglomerate3,4 (dark red/blue), igneous zircon from the Acasta Gneiss Complex43,45 (green), and detrital and xenocrystic zircon from the Singhbhum Craton41,42 (orange). The DMM (depleted MORB mantle) approximates the isotope evolution of Earth’s upper mantle from its origin (ε176Hfi = 0 at 4.4–4.5 Ga) to the present-day (ε176Hfi ≈ +15). The coloured arrays (blue and orange) show the expected isotope evolution of juvenile crust extracted from a primitive mantle reservoir (bulk silicate Earth) at 4.42 Ga and 4.19 Ga. The arrays are bound by the isotope trajectories of a mafic reservoir with 176Lu/177Hf of 0.022, a tonalite–trondhjemite–granodiorite (TTG) reservoir with 176Lu/177Hf of 0.005; a crustal reservoir with 176Lu/177Hf of 0.008, corresponding to the average upper continental crust (UCC), is shown for comparison3,4. Dark red data points show zircon grains with initial ε176Hf consistent with differentiation of their protoliths during an early crust formation event at ~4.4 Ga. Dark blue circles represent zircon grains with initial ε176Hf requiring a younger crustal protolith, consistent with zircon data on Acasta and Singhbhum.

A temporal frame for geodynamic models

Geodynamic simulations suggest that Hadean-Archaean tectonics may have been dominated by vertical exchanges between the mantle and crust18,48,49. Consistent with this interpretation are studies on Archaean supracrustal belts which show that terranes with geochemical characteristics indicative of subduction settings were uncommon and rather dominated by volcanic rocks, reflecting embryonic stages of arc formation50,51. In the possible absence of self-sustained subduction, crust generation and recycling in the early Earth may have proceeded sporadically11,12,49, in response to disruptive events such as large plumes15 and impacts12, or due to dripping of gravitationally unstable parts of the lithosphere10,18. Consistent with our observations, such catastrophic events would produce large depleted mantle domains during short pulses (~10–20 Myr) of extreme magmatic activity interspersed by long periods of relative quiescence10,18,19. Models advocating an episodic-lid behaviour further impart a strong periodicity in the tectonic evolution of the early Earth19, which may be reflected in the ~200 Myr time period separating the two major mantle depletion events recorded by the 146Sm–142Nd system. In light of the currently available data, it thus appears unlikely that an active tectonic regime, akin of plate tectonics, was operating on the Hadean Earth. Going beyond these different scenarios, our results provide the first evidence that Hadean crust generation is consistently reflected both in the history of mantle depletion and in the geological relics of that epoch. Further investigations of these fragmentary records thus have the potential to assess the frequency of crust-mantle differentiation events during the Hadean as well as the composition of these differentiated reservoirs. Ultimately, chronological constraints on Hadean crust-mantle differentiation, as obtained from the 146,147Sm–142,143Nd isotope systems, may complement geophysical approaches and help to evaluate and refine models of ancient Earth’s geodynamics.


Data acquisition

Whole-rock powders were obtained using standard rock crushing techniques. Details for combined 147Sm-143Nd isotope dilution and 143Nd/144Nd isotope composition determination, conducted at the University of Bern, are described in Pandey et al.35 with the corresponding data displayed in Supplementary Data 1. Data on the Bastar Craton were obtained during this study following the same protocols, but using inductively coupled plasma mass spectrometry. An average 143Nd/144Nd ratio of 0.512078 ± 5 (2 SD, n = 14) was obtained for repeated measurements of the JNdi-1 standard solution during the course of the study (Supplementary Data 2). The external reproducibility of 143Nd/144Nd for samples was estimated using the relationship between the external reproducibility of JNdi-1 measurements over the course of the study (% 2 SD, n = 14) and the average standard error of individual analyses (% 1 SE, 60 × 8 s integrations). All samples were normalised to the recommended JNdi-1 143Nd/144Nd value of 0.512115 for international comparison52. Element purification and isotope analyses for 146Sm–142Nd investigation of all samples were performed at CRPG, Nancy using protocols detailed in Morino et al.32. Neodymium isotope measurements were performed by thermal ionisation mass spectrometry, using a two-line multidynamic scheme. Correction for mass fractionation drift during acquisition53 was performed following Morino et al.32. Measurements were performed in three analytical sessions during which the external precision, expressed as 2 SD, was estimated by repeated measurement of the JNdi-1 standard at 3.2 ppm, 1.3 ppm, and 2.6 ppm, respectively (Supplementary Data 2). 143Nd/144Nd ratios are reported using the conventional ε-notation, as relative deviation in parts per 10,000 from the chondritic value40. 142Nd/144Nd ratios are reported in μ-notation, as deviation in ppm from the JNdi-1 reference material. Internal run statistics (2 standard error, SE) are reported for samples that were analysed once. Replicate analyses are reported using the average value and an uncertainty of 2.5 ppm corresponding to the average 2 SD reproducibility of standards from all analytical sessions. For model calculations, an initial 146Sm/144Sm of 0.0082 was used in conjunction with a half-life of 103 Myr for 146Sm54,55.

Coupled 146,147Sm–142,143Nd model age calculations

The chronological implications of the investigated rock suites are examined using a two-stage isotope evolution model. The first stage corresponds to the evolution of a primitive reservoir (bulk silicate Earth, BSE) between the origin of the solar system (T0 = 4.567 Ga) and an instantaneous differentiation event at T = TD. The second stage represents the closed-system evolution of differentiated silicate reservoirs between TD and the crystallisation age of the samples at T = TS. The 142,143Nd/144Nd composition of the depleted mantle source reservoir (src) at TS is given by:

$${\Bigg(\frac{{\,}^{143}{{{{{\mathrm{Nd}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{src}}}}}}}^{{T}_{{{{{{\mathrm{S}}}}}}}}= \, {\Bigg(\frac{{\,}^{143}{{{{{\mathrm{Nd}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{BSE}}}}}}}^{{T}_{{{{{{\mathrm{P}}}}}}}}+\,{\Bigg(\frac{{\,}^{147}{{{{{\mathrm{Sm}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{BSE}}}}}}}^{{T}_{{{{{{\mathrm{P}}}}}}}}\,\times \left(1-{e}^{{\lambda }_{147}{T}_{{{{{{\mathrm{D}}}}}}}}\right)\\ +\,{\Bigg(\frac{{\,}^{147}{{{{{\mathrm{Sm}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{src}}}}}}}^{{T}_{{{{{{\mathrm{P}}}}}}}}\,\times \,\left({e}^{{\lambda }_{147}{T}_{{{{{{\mathrm{D}}}}}}}}-\,{e}^{{\lambda }_{147}{T}_{{{{{{\mathrm{S}}}}}}}}\right)$$
$${\Bigg(\frac{{\,}^{142}{{{{{\mathrm{Nd}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{src}}}}}}}^{{T}_{{{{{{\mathrm{S}}}}}}}}= \, {\Bigg(\frac{{\,}^{142}{{{{{\mathrm{Nd}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{BSE}}}}}}}^{{T}_{{{{{{\mathrm{P}}}}}}}}+\frac{{({\,}^{146}{{{{{\mathrm{Sm}}}}}}/{\,}^{144}{{{{{\mathrm{Sm}}}}}})}_{{T}_{0}}}{{({\,}^{147}{{{{{\mathrm{Sm}}}}}}/{\,}^{144}{{{{{\mathrm{Sm}}}}}})}_{{T}_{{{{{{\mathrm{P}}}}}}}}}\times \Bigg[{\Bigg(\frac{{\,}^{147}{{{{{\mathrm{Sm}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{BSE}}}}}}}^{{T}_{{{{{{\mathrm{P}}}}}}}}\,\times \,\left({e}^{-{\lambda }_{146}{T}_{0}}-{e}^{-{\lambda }_{146}({T}_{0}-{T}_{{{{{{\mathrm{D}}}}}}})}\right)\\ +\,{\Bigg(\frac{{\,}^{147}{{{{{\mathrm{Sm}}}}}}}{{\,}^{144}{{{{{\mathrm{Nd}}}}}}}\Bigg)}_{{{{{{\mathrm{src}}}}}}}^{{T}_{{{{{{\mathrm{P}}}}}}}}\,\times \,\left({e}^{-{\lambda }_{146}({T}_{0}-{T}_{{{{{{\mathrm{D}}}}}}})}-\,{e}^{-{\lambda }_{146}({T}_{0}-{T}_{{{{{{\mathrm{S}}}}}}})}\right)\Bigg]$$

where λ146 = 6.74 × 10−9 a−1 and λ147 = 6.54 × 10−12 a−1 are the decay constants for 146Sm and 147Sm, respectively, and TP refers to the present-day. The 147Sm/144Nd and 143Nd/144Nd ratios of the bulk silicate Earth are considered chondritic39. The 142Nd/144Nd composition of the BSE is estimated by repeated measurements of the JNdi-1 international standard.

For each sample carrying a positive 142Nd anomaly, the corresponding (147Sm/144Nd)src and TD can be calculated by solving Eqs. (1) and (2) iteratively. The first iteration assumes an arbitrary value of TD = 4.5 Ga and the corresponding (147Sm/144Nd)src is then calculated using Eq. (1). This value is then inserted into Eq. (2) to derive a new value of TD. This calculation is repeated until constant values of TD and (147Sm/144Nd)src are obtained. Alternatively, a two-stage model age can be calculated by combining all samples within a 146,147Sm-142,143Nd isochron (Supplementary Fig. S-2). This approach is similar to that presented above but presents the advantage of integrating all the isotope information into the age calculation. In this case, it is assumed that all samples are derived from cogenetic (mantle) domains with different parent-daughter ratios, or alternatively, that the chemical and isotopic heterogeneities that define the slope of the μ142Nd–Sm/Nd array result from mixing between a Hadean depleted reservoir and a primitive mantle component. The model age is calculated by regression of the 142Nd/144Nd vs. (144Sm/144Nd)src array (Supplementary Fig. S-2). For samples showing no significant 142,143Nd excess, (147Sm/144Nd)src is approximated using Eq. (2), assuming a value of TD = 4.2 Ga. As those samples are characterised by ε143Ndi ≈ 0, the calculation yields near-chondritic (147Sm/144Nd)src and this result is thus only marginally dependent on the assumed value of TD. The slope of the regression line (Fig. S-2) is equal to the value of the 146Sm/144Sm ratio at TD, and the differentiation age is then obtained using the decay equation:

$${T}_{{{{{\mathrm{D}}}}}}=-(1/\lambda_{146})\times {{{{{\mathrm{ln}}}}}}\left[\frac{({\,\!}^{146}{{{{{\mathrm{Sm}}}}}}/{\,\!}^{144}{{{{{\mathrm{Sm}}}}}})_{T_{{{{{\mathrm{d}}}}}}}}{({\,\!}^{146}{{{{{\mathrm{Sm}}}}}}/{\,\!}^{144}{{{{{\mathrm{Sm}}}}}})_{{T_{0}}}}\right]$$

where \(({\,\!}^{146}{{{{{\mathrm{Sm}}}}}}/{\,\!}^{144}{{{{{\mathrm{Sm}}}}}})_{T_{0}}=0.0082\) is the initial isotope composition of the solar system and \({\lambda }_{146}=6.73\) 10−3 Ma−1 is the decay constant of 146Sm54,55.

The Singhbhum results, represented in Fig. S-1, show that the four granitoid samples with the most positive ε143Ndi values (i.e., +2.5 to +5.1) and positive μ142Nd values (+3.9 to +6.3 ppm) yield similar two-stage model ages ranging from 4.18 Ga to 4.26 Ga (Supplementary Fig. S-1a), with an average TD = 4.22 ± 0.07 Ga (2 SD) similar to the model age derived from the 146,147Sm–142,143Nd isochron (\({4.19}_{-0.12}^{+0.06}\) Ga, Fig. 1). Despite the possibility of minor open system behaviour in the investigated IOG rocks, a model age can tentatively be estimated by back-calculating the initial ε143Nd compositions to the accepted emplacement age of 3.51 Ga39, or alternatively by using the ε143Ndi of 2.4 ± 1 obtained from the 3.75 Ga isochron regression (see Supplementary Discussion, Pandey et al.35). Considered together with the average μ142Nd of 4.1 ± 3.2 ppm, these calculations yield imprecise but broadly similar late Hadean model ages of 4.27 ± 0.22 Ga and \({4.15}_{-20}^{+11}\) Ga, respectively (Supplementary Fig. S-1b, c).