Bulk rock major and trace element analyses of fine powdered samples were carried out at CSIR-National Geophysical Research Institute, Hyderabad. The major elements were analyzed using XRF (Phillips Axios mAX4), followed by pressed pellet sample preparation technique. The details of the analytical methods, such as instrument calibration, data acquisition, accuracy, and detection limits are provided in Krishna et al. (2016). Trace elements were analyzed using AttoM HR-ICP-MS (Nu Instruments, United Kingdom). 50 mg of finely powdered samples were taken in Savillex^® vials and 10 mL of the acid mixture containing HF and HNO₃ (7:3 ratio) was added. Further, the vials were heated at 150°C for 48 h and a clear solution was obtained. 1 mL of HClO₄ was added and the sample solutions were dried to form a solid residue. Twenty ml of 1:1 ratio HF and Millipore water mixture was added to the vials and heated at 80°C for 1 h. The sample solutions were further transferred to 25 mL conical flasks, and 5 mL of 1 ppm Rh was added as an internal standard. To obtain the optimal total dissolved solids (TDS) level, the sample solution was initially diluted to 250 mL, and 5 mL of the diluted solution is further diluted to 50 mL. The standards UB-N, JP-1 and blank solutions were analyzed along with the samples. The details of the sample digestion method, instrumental parameters, data acquisition, and quality are described in Satyanarayanan et al. (2018).

Sample preparation for isotopic measurements was carried out at the Institute of Geological Sciences, University of Bern, Switzerland. Around 90 mg of finely powdered sample was spiked with ⁸⁷Rb-⁸⁴Sr and ¹⁴⁹Sm-¹⁵⁰Nd spikes in Savillex^® beakers by keeping an empiric error magnification of ∼1.5–2 (Stracke et al., 2014). The samples underwent a digestion procedure using concentrated HF, HNO₃ and HCl solutions separately for 2 days each at 120°C. Each step was carried out after drying the samples on a hotplate at low temperatures. After drying the samples, the samples were loaded on DOWEX^® AG 50 W-X8 (200–400 mesh) cation columns to separate Rb and Sr The loading reagent was 2.5 M HCl. Rubidium and Sr cuts were collected using 2.5 M HCl with a distinct eluting step in between. Chemical separation of Sm from Nd was done using the procedure outlined by Ravindran et al. (2021).

The Rb and Sr isotopic measurements were done at the Institute of Geological Sciences, University of Bern, Switzerland. The spiked samples were measured for Rb and Sr isotopes on a Thermo Scientific™, Neptune Plus™ multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS). Dry plasma mode using a Cetac Aridus II desolvation system was used for efficient sample introduction and sensitivity during the measurements on the multi-collector ICPMS. The standards and samples were measured at the same concentration for better accuracy. Background measurements of the washing acid reagent, 0.5 M HNO₃, were subtracted from each measurement of sample and standard solutions. Interferences of ⁸⁴Kr and ⁸⁶Kr on ⁸⁴Sr and 86Sr were monitored and were close to the baseline (<10^–5 V). Total procedural blanks were < 500 pg that were used for the blank correction of the samples.

The Sr isotopic ratios were corrected for mass bias using exponential law and iteration (Stracke et al., 2014) with a normalization ratio of ⁸⁶Sr/⁸⁸Sr = 0.1194. The NIST standard SRM^® 987 yielded ⁸⁷Sr/⁸⁶Sr = 0.710253 ± 0.000012 (2SD; n=8). The Rb isotopic ratios were corrected for mass bias by the standard-sample-standard bracketing method using the exponential law.

Neodymium and Sm isotopes were measured on a Thermo Scientific™, Neptune™ Multi Collector ICPMS at the Institute of Geochemistry and Petrology, ETH Zurich, Switzerland. The same measurement technique used for Rb and Sr isotope measurements was carried out for Nd and Sm isotope measurements. Additionally, apart from using a normalization ratio of ¹⁴⁶Nd/¹⁴⁴Nd = 0.7219, Nd isotope ratios were also internally normalized for instrumental mass bias following Vance and Thirlwall (2002). Standard JNdi yielded an average ¹⁴³Nd/¹⁴⁴Nd = .512084 ± 0.000005 (2σ; n = 14), with an external reproducibility of 14 ppm.

Analysis of clinopyroxene composition was performed at Institute of Earth Sciences, Academia Sinica, Taiwan. Selected samples were first examined with a scanning electron microscope (JEOL SEM JSM-6360LV) for microtextural observation, qualitative mineral identification, and spotting sites of interest for subsequent electron probe microanalysis (EPMA), which was undertaken by an electron probe micro analyzer (JEOL EPMA JXA-8900 R) equipped with four wavelength dispersive spectrometers (WDS). Secondary electron images and backscattered electron images were used to guide the analysis of the target positions of minerals. For quantitative analysis, a 1 µm beam was used at an acceleration voltage of 15 kV with a beam current of 12 nA. Corrections on the measured X-ray intensities were carried out by oxide-ZAF method using the standard calibration of synthetic standard minerals with various diffracting crystals. The standard minerals analyzed are as follows: wollastonite for Si with TAP crystal, rutile for Ti with PET crystal, corundum for Al (TAP), chromium oxide for Cr (PET), hematite for Fe with LiF crystal, Mn-oxide for Mn (PET), periclase for Mg (TAP), nickel oxide for Ni (LiF), wollastonite for Ca (PET), albite for Na (TAP), apatite for P (PET) and adularia for K(PET). Peak counting for each element and both upper and lower baseline X-rays counted for 10 s and 5 s, respectively. Standards run as unknowns yielded relative standard deviations of < 1% for Si, Na and K, and < 0.5% for other elements. Detection limits were less than 600 ppm for all elements. Analyses were made within the cores of mineral grains to minimize the effects of zonation, alteration and re-equilibration. The results are presented in Supplementary Table S1.

Major and trace element data for samples from the Ghatgaon, Keonjhar and Pipilia dyke swarms are given in Supplementary Table S2. Geochemical variations of the samples range from basalt, basaltic andesite to andesites (Figure 3), showing substantial compositional variations. The samples from Keonjhar exhibit compositional variations from basalt to andesites with 3.7–12.8 wt% MgO and 11.3 to 15.8 wt% FeO_t. Most of the Ghatgaon samples have 6.4 to 9.2 wt% MgO, except two samples (KD-12A, KD-12B) collected from Dhenkikote that have ∼21 wt% MgO. Such high MgO is consistent with their cumulate nature. From our data and previous studies (Dasgupta et al., 2019; Adhikari et al., 2021; Pandey et al., 2021) the basalt compositions are filtered out and further plotted into Zr/P₂O₅*10⁴ vs. Nb/Y diagram of Winchester and Floyd (1976) to discriminate between alkaline and tholeiitic compositions (Figure 3, inset). The dyke swarms of Ghatgaon, Keonjhar and Pipilia exhibit tholeiitic basalt compositions. A relative compositional variability is observed mainly in Ghatgaon and Pipilia dyke swarms. No systematic trend exists between TiO₂ versus Mg# (Figure 4). Variations in Al₂O₃/TiO₂ versus Mg# (Figure 4) in relatively less evolved (Mg# >50) lavas indicate moderate degree of partial melting from the mantle source (Nesbitt et al., 1979).

All samples display LREE-enriched patterns in chondrite-normalized variation diagrams (Figure 5). Though the REE patterns of Pipilia swarms are subparallel, the variations in other dyke swarms highlight the differences between the compositional variability of individual dyke swarms. Especially, sample KD-21of Keonjhar dyke swarm exhibit relatively enriched REE patterns (Figure 5C). In the multi-element diagram, compositional variations can be observed in individual dyke swarms, pointing towards a complex petrogenetic relation (Figure 5). Almost all samples exhibit negative Ti anomalies, positive Pb anomalies and negative Nb anomalies (Figure 5). The Ghatgaon samples show negative Y anomalies, whereas the Keonjhar and Pipilia samples exhibit either negative or positive Y anomalies. Negative Sr anomalies relative to Pr and Nd can also be noticed for most of the samples.

The Sr-Nd isotopic data for nine selected samples with basaltic composition are listed in Table 2, which include four samples from Ghatgaon dyke swarm, four samples from Keonjhar dyke swarm and one sample from the Pipilia dyke swarm. They display large variations. For instance, initial ⁸⁷Sr/⁸⁶Sr of samples from Keonjhar dyke swarm ranges from .705543 to .707282, and ε_Ndt values from -12.0 to + 3.9 (Table 2). The lack of negative correlation between them (Figure 6) indicates the Rb-Sr isotopic system might have been disturbed by post-magmatic processes (e.g., alteration); the Sm-Nd isotopic system is relatively insensitive to such processes. Both positive and negative ε_Nd values are evident for the Ghatgaon and Keonjhar samples. In Keonjhar dyke swarm, the sample that exhibits relatively enriched REE patterns (KD-21) also shows an enriched ε_Ndt value of -12. A single sample (KD-19A) from the Pipilia dyke swarm yielded an ε_Nd value of -8.8.

Due to the altered nature of the samples, representative samples containing fresh clinopyroxene are selected from the three dyke swarms for mineral chemical analysis. The selected samples are from the Keonjhar and Ghatgaon dyke swarms; no Pipilia samples were selected. Clinopyroxene in the Keonjhar samples exhibits mostly augitic compositions (En_48-36Wo_44-34Fs_22-13), and are characterized by 51–54 wt% of SiO₂ and 13–17 wt% MgO. Clinopyroxene in the Ghatgaon samples has compositions varying from pigeonite to augite (En_50-37Wo_36-11Fs_44-19) (Figure 7A). All analyzed clinopyroxene have relatively low TiO₂. Further, most clinopyroxene compositions plot on or below the equilibrium curve in the Rhodes diagram (Figure 7B), indicating most clinopyroxene grains formed by equilibrium crystallization. In other words, those grains crystallized from or continually equilibrated with the groundmass that eventually solidified as the framework of the dolerites. Compositional variations in the clinopyroxenes are also exhibit significant overlap with mid-ocean ridge basalts (MORB) and Island arc tholeiites (IAT) (Figure 7C).

The incompatible trace element patterns of the Singhbhum dykes are markedly different from typical patterns of N-MORB and OIB (Sun and McDonough, 1989). Instead, the patterns of the dykes display strong resemblance to continental crustal rocks (Figure 5). Such a trace element signature is commonly explained by assimilation of continental crustal rocks by mantle-derived magmas during their ascent to the surface, magma derivation from a mantle wedge above a subduction zone or a mantle source already modified by the crustal components. Below, we address these mechanisms in detail.

We have assessed the possibilities of crustal contamination from geochemical signatures (Figure 8A). The Nb and Th ratios have negligible effects on fractional crystallization and mantle melting, therefore the Th/Nb ratios can be used to evaluate the effect of crustal contamination or mantle source heterogeneity (Tegner et al., 2019). The continental crust has high Th/Nb (0.875; Rudnick and Gao, 2014). The Paleoarchean and Mesoarchean TTGs and Mesoarchean K-rich granitoids of Singhbhum craton also exhibit high Th/Nb (0.786; Upadhyay et al., 2019). In the present study, samples from Ghatgaon dyke swarm (Th/Nb = 0.13–0.44), Keonjhar dyke swarm (Th/Nb= 0.15–0.32) and Pipilia dyke swarm (Th/Nb = 0.15–0.32) exhibit Th/Nb ratios lower than the continental crust, which points out that the contamination from the upper crustal sources is minimal. Further, we have evaluated the role of crustal contamination or possible involvement of recycled components in the source using the Th/Yb and Nb/Yb (Pearce, 2008). The mantle array in the Th/Yb versus Nb/Yb diagram represents melts formed from the oceanic mantle sources. Two mixing curves generated between MORB sources (N-MORB and E-MORB) and average continental crust (Rudnick and Gao, 2014) indicate the samples have assimilated 5% to more than 40% crustal components, in which 10%–30% of crustal contamination can explain the Th/Yb and Nb/Yb variations observed in the Pipila and Keonjhar swarms (Figure 8A). The Neoarchean Ghatgaon swarm exhibit assimilation of more than 25% of the crustal components. However, it is unlikely that a sub-alkaline primitive magma assimilates more than 20%–30% of crustal components (Heinonen et al., 2022). Roy et al. (2004) also proposed that variations in δ¹⁸O (avg. + 3.97‰ ± .75‰) of newer dolerite swarms are inconsistent with the incorporation of a large volume of continental crust.

There are several differences in the geochemical and isotopic variations of the dyke swarms that cannot be easily attributed to crustal sources or assimilation of crustal components and/or magma mixing or fractional crystallization alone. The Paleoarchean granitoids of the craton are proposed to be derived from the partial melting of proto-mafic crust at ca. 10–15 kbar (Pandey et al., 2019; Upadhyay et al., 2019), formed from a near chondritic mantle reservoir (Pandey et al., 2019) (ε _Ndt = -0.3 to + 2.2 and ε _Hfi = -0.3 to + 2.0). The dyke swarms of Singhbhum craton, however, exhibit more enriched Nd isotope values. For example, part of Paleoarchean older metamorphic trondhjemite and Singhbhum granite III which are basement to the Ghatgaon dyke swarm exhibit depleted mantle values (ε _Ndt = + 2 .1 to + 4.5; Pandey et al., 2019). However, the Nd isotopes of the Ghatgaon samples show wide variations (ε _Ndt=+ 4.6 to -4.8). In the basement of Keonjhar swarm, the Singhbhum granite II exhibits ε _Ndt = + 2.7. The ε_Ndt values of Keonjhar dyke swarm are -11 to + 3.8. Upadhyay et al. (2019) also noticed that the Paleoarchean-Mesoarchean granites and tonalite–trondhjemite–granodiorites (TTGs) of the craton are derived from a depleted mantle source. If we recalculate the isotopic variations of basement granites it can be a potential contaminant, but it is unlikely that the observed isotopic variability results only from the assimilation of basement granitoids, as the source characteristics of the samples are inconsistent with enriched high-temperature regimes. Moreover, the mantle source characteristics appears to be similar despite their discrete emplacement ages, which points out the isotopically distinct basement granitoids cannot be attributed solely to the compositional variations.

The negative anomalies in Nb, Zr and Hf in the primitive mantle normalized diagrams (Figure 5), relatively high initial ⁸⁷Sr/⁸⁶Sr, and negative ε_Ndt values can be correlated with subduction zones (Figure 6) (Zheng, 2019). Also, the clinopyroxene in the Keonjhar and Ghatgaon samples shows significant overlap between the compositional fields of clinopyroxenes from MORB and IAT (Beccaluva et al., 1989) (Figure 7C). However, similar signatures can also be produced from crustal contamination or crustal components in the source. The age, distribution and cross-cutting relationship of dyke swarms are inconsistent with their origin related to a subduction zone (Figure 8) (Dasgupta et al., 2019; Adhikari et al., 2021; Pandey et al., 2021). The dyke swarms are distributed in an intracratonic regime and the stabilization of the Singhbhum craton occurred during Mesoarchean (Dey et al., 2017; Olierook et al., 2019; Srivastava et al., 2019; Hofmann et al., 2022). The emplacement of these dyke swarms occurred at discrete magmatic episodes from Neoarchean to Mesoproterozoic, and their subparallel distribution and areal extent of more than 100 km are consistent with an intracontinental rift system (Pandey et al., 2021)

We propose the observed geochemical signatures are most likely to be the characteristics of the mantle source that has incorporated crustal components (Adhikari et al., 2021; Pandey et al., 2021). Evidence for a pre-Neoarchean crustal recycling and subsequent enrichment of DMM mantle source followed by the metasomatism of SCLM has been noticed in the 2.8 Ga Jagannathpur volcanics and Ghatgaon and Keshgaria dyke swarms (Adhikari et al., 2021). Pandey et al. (2021) proposed that subduction-related crustal recycling modified the refractory SCLM. Further episodic plume impingement triggered the partial melting of the enriched SCLM between 2.8 Ga and 1.76 Ga. Moreover, delineation of thick crust under Proterozoic domains is interpreted to be evidence for Neoarchean convergent tectonics in the craton (Mandal, 2017). Recycling of Eoarchean and Paleoarchean crust (Olierook et al., 2019; Sreenivas et al., 2019; Upadhyay et al., 2019) and Mesoarchean and Neoarchean subduction-related magmatism (Mondal et al., 2006; Manikyamba et al., 2015), have been noticed in the Archean tectonic framework of Singhbhum craton. Even though the plate tectonics-related crustal recycling models during Archean are debatable (Hamilton, 2019; Zhu et al., 2021), it is conceivable that either through subduction or through non-plate tectonic modes, like melting of metasomatized lithosphere delaminated from the base of the continental crust (Cousens et al., 2001; Arndt et al., 2009), the earlier crust of Singhbhum craton have been recycled into the mantle. The resulting heterogeneous mantle source(s) can be attributed to the source of dyke swarms of Singhbhum craton (Upadhyay et al., 2014; Olierook et al., 2019; Pandey et al., 2019; Hofmann et al., 2022).

Earlier discussion arrived at a view that the mantle source giving rise to the studied Singhbhum dyke swarms acquired a crustal trace element signature. Logical questions that follow are in which part of the mantle did that source reside, and how did it partially melted to yield the magmas that eventually solidified as the dykes. This section addresses issues relevant to geochemical and mineralogical characteristics of the mantle source, and temperature-pressure of partial melting.

First, we employ the parameter ΔNb to examine the relative enrichment and depletion of the mantle source. Calculated as 1.74+ log(Nb/Y) - 1.92 log(Zr/Y), ΔNb is used to express deviation from a reference line (i.e., ΔNb = 0) separating enriched oceanic mantle end-members (E-MORB and OIB) and N-MORB (Fitton, 2007). Fitton (2007) proposed that all N-MORB has ΔNb<0, whereas E-MORB and OIB lavas exhibit ΔNb>0. Figure 8B shows the ΔNb variations of the Gatagaon, Keonjhar and Pipilia dyke swarms. The Ghatgaon and Keonjhar dyke swarms exhibit both positive and negative ΔNb values, whereas the Pipilia dyke swarm exhibit mostly negative ΔNb values. One sample from Ghatgaon (KD-09) exhibits a maximum ΔNb of + 0.27, which means Nb is enriched 2.7 times relative to the ΔNb of zero. Other positive ΔNb values of Ghatgaon dyke swarm range from + 0.04 to + 0.18. One sample from the Keonjhar dyke swarm (KD-05) exhibits + 0.06 ΔNb whereas others exhibit negative ΔNb (-0.2 to -0.5). The variations in the ΔNb values also indicate a heterogeneous mantle source, that has both enriched and depleted components and which is distinct from MORB sources (Figure 8B). Along with ΔNb, elevated La/Sm_n values illustrate high degrees of partial melting or a less depleted source (Figure 8B). Most of the ΔNb values overlap in all three dyke swarms, which again illustrates secular compositional variations (e.g., Pandey et al., 2021) in the mantle source may not be applicable, rather it can be attributed to mantle heterogeneity.

Further, we used the variations in MREE to HREE to evaluate the effects of residual garnet in the source, as garnet prefers HREE relative to MREE. The (Dy/Yb)_N ratios (Davidson et al., 2013), along with (La/Sm)_N ratios have been considered as they reflect the degree of mantle melting (Figure 9) (Tegner et al., 2019). The Keonjhar swarm and Ghatgaon swarms exhibit Dy/Yb of 1.8 and 1.9 (median) whereas the Pipilia swarm exhibit Dy/Yb =2.2. The Dy/Yb variations are slightly higher than MORB values (∼1.6) and considerably lower than the OIB values (∼2.5) (Figure 9B). The Himalayan magmatic province and CAMP proposed to be derived from a subduction-modified mantle source (Tegner et al., 2019; Manu Prasanth et al., 2022) also show overlap with the Dy/Yb variations of the Ghatgaon, Keonjhar and Pipilia swarms (Figure 10). The Dy/Dy* vs. Dy/Yb (Davidson et al., 2013), relations exhibit LREE enriched field with garnet control on some of the Keonjhar and Pipilia swarms. Most of the Ghatgaon swarm samples exhibit clinopyroxene fractionation (Figure 9A). Some samples from Dhenkikote area of Ghatgaon dyke swarm (KD-12A, KD-12B) possibly stalled at the base of the crust or coalesced into magma chambers as it is evident form their cumulate nature.

The primary magma calculation schemes depend on the addition or removal of olivine (e.g., Herzberg and Asimov, 2015; Lee et al., 2009) were not able to retrieve realistic thermal and melting conditions as the dyke swarms exhibit fractionation of clinopyroxenes (Figure 4). Therefore, we further explored the mantle melting models using the REEBOX PRO application of Brown and Leshar (2016). This application model the composition of melts produced by lithologically heterogeneous dynamically upwelling mantle. We simulated a range of forward partial melting models of peridotitic and mixed peridotite and pyroxenite source lithologies, using different modes of melting, and variable mantle potential temperatures. The pyroxenites in the model represent recycled crustal domains in the mantle (Figure 9).

For the first two models (Figures 10A, B), we used a pyrolite peridotite source (Sun and McDonough, 1989). In the first model (Figure 10A), we used the columnar melting regime and T _p is constrained between 1350°C to 1550°C, which shows trace element variations can be achieved by 5%–10% partial melting under a T _p of 1400°C to 1,450°C. Two samples from the Keonjhar dyke swarm exhibit melting within the T _p range of 1350°C–1400°C. At 10% of melting (T _p = 1450°C–1350°C) the pressure is 2.4 to 1.3 GPa which indicates melting in the garnet-spinel transition zone and spinel peridotite zone (Su et al., 2010), and also corroborate Dy/Yb variations and both positive and negative anomalies of Y in the primitive mantle normalized diagrams (Figure 5).

In the second model (Figure 10B), we simulated a triangular melting regime with a pyrolite source as the starting composition. Relatively lower T _p (1350°C–1400°C) and 5%–10% of melting can produce compositional variations in some of the Keonjhar dyke swarms. When the T _p increases (1400°C–1450°C), 10%–20% melting is required to produce the compositional variation. Most of the melting took place in the spinel stability field. (Brown and Lesher, 2016).

To address the heterogeneous nature of the mantle source we used combinations of peridotite and pyroxenite under a columnar melting region in the third model. Considering the thick continental lithosphere of the Singhbhum craton and discrete episodes of dyke emplacement, we assumed the columnar melting region is more appropriate (Brown and Lesher, 2016; Tegner et al., 2019). Along with pyrolite peridotite, we also tried modelling with a depleted mantle (Workman and Hart, 2005) with different proportions of pyroxenites. But the melting curves did not follow the observed compositional variations. The results show anhydrous pyrolite peridotite (T _p = 1,350°C and 1,400°C) with 1%–10% pyroxenite (silica-saturated pyroxenite, G2; Pertermann and Hirschmann, 2003) were involved in the petrogenesis.

The SCLM sources combined with convecting asthenospheric regions likely contributed to the wide range of compositions in the dyke swarms of Singhbhum craton (Kamenetsky et al., 2017). The geochemical variations can be compared with LIPs like the Himalayan magmatic province, Central Atlantic Magmatic Province and North Atlantic Magmatic Province (Korenaga, 2004; Coltice et al., 2007; Marzoli et al., 2018; Tegner et al., 2019; Manu Prasanth et al., 2022), as their origins are inconsistent with a deep-seated mantle plume and are proposed to be derived from lithospheric and/or asthenospheric mantle sources with recycled crustal components (Korenaga, 2004; Shellnutt et al., 2021; Manu Prasanth et al., 2022). Narrow zone upwelling of asthenospheric melts associated with the discrete rifting events in the craton might have triggered the melting of the lithospheric mantle domains (Foley, 2008; Tang et al., 2013), which led to the melt generation and emplacement of Ghatgaon, Keonjhar and Pipilia dyke swarms.

Pandey et al. (2021) identified a secular Nd isotopic evolution trend for the dyke swarms of Singhbhum craton. In detail, the 2.8 Ga Keshgaria swarm exhibit near chondritic ε_Ndt values (-1.0 to + 2.1), whereas younger generations of dyke swarms have progressively more negative values. Such a trend is consistent with a lithospheric mantle source in which crustal materials were incubated since ∼2.8–3.0 Ga, and the dyke swarms formed by periodic melt extraction from that source. However, our new data hint complexity beyond a simple, secular Nd isotopic trend. The 2.76 Ga Ghatgaon dyke swarm, which yielded negative ε_Ndt (-2.4 to -2.6) values (Pandey et al., 2021), displays a greater range of ε_Ndt values from negative to positive (ε _Ndt = -4.8 to + 4.6). Also, the Neoarchean-Paleoproterozoic Keonjhar dyke swarm exhibits a similar but more extreme range of ε_Ndt values (ε _Ndt = -11.9 to + 3.8) (Figure 6). Undoubtedly, at least for the Ghatgaon and Keonjhar dyke swarms, a depleted mantle-like component must also be involved at the time of magma genesis. One way to explain the Nd isotopic range was interaction between asthenosphere and mantle lithosphere that had been previously enriched. However, samples having positive ε_Ndt values in this study also show continental crust-like trace element signature as strong as samples having negative ε_Ndt values, inconsistent with variable degrees of asthenosphere-lithosphere interaction. An alternative explanation is that the positive ε_Ndt values reflect recent mantle enrichment relative to the formation age of the Ghatgaon and Keonjhar dyke swarms. If the secular trend argued by Pandey et al. (2021) tracks evolution of an enriched mantle source formed by recycling of ∼2.8–3.0 Ga crustal material, our new data imply that crustal materials younger than that range might also be involved, implying crustal recycling might have been episodic rather than a discrete, single-stage process. In other words, there might be multiple secular trends depending on the age of the recycled material(s). We suggest that this mechanism offers a better explanation for all Nd isotopic data available for the Singhbhum dyke swarms so far.