Introduction

Titanium is a moderately incompatible element in magmatic systems and becomes concentrated in crustal rocks relative to the mantle. It has five stable isotopes (46Ti, 47Ti, 48Ti, 49Ti and 50Ti) and its isotopic variations have been increasingly used to probe numerous fundamental geochemical and cosmochemical processes such as planetary differentiation, crust production and crust-mantle interaction (e.g., Millet et al. 2016; Greber et al. 2017a, 2017b, 2021; Deng et al. 2018a, 2018b, 2019, 2023; Mandl 2019; Johnson et al. 2019, 2023; Aarons et al. 2020; Kommescher et al. 2020; Zhao et al. 2020; Hoare et al. 2020, 2022; Rzehak et al. 2021, 2022; Williams et al. 2021; Anguelova et al. 2022; Storck et al. 2023). Early studies demonstrated that Fe-Ti oxide unsaturated terrestrial mafic and ultramafic rocks such as komatiites and global oceanic basalts display limited Ti isotopic variation (δ49/47Ti between − 0.05 and 0.05 ‰), although a small, but resolvable Ti isotopic variation has been observed in komatiites and mid-ocean ridge basalts (Millet et al. 2016; Greber et al. 2017a; Deng et al. 2018b; Zhao et al. 2020). By contrast, differentiated igneous rocks display systematically heavier Ti isotopic compositions that correlate with indicators of magmatic differentiation such as SiO2 or MgO contents, which is assumed to be primarily controlled by saturation and fractionation of isotopically light Fe-Ti oxides during magmatic differentiation (Millet et al. 2016; Greber et al. 2017a). These findings make Ti isotopes a promising tracer of diverse geological and cosmochemical processes. However, subsequent work has demonstrated that the Ti isotope behavior becomes more complicated during magmatic differentiation, and different magmatic systems (such as alkaline, tholeiitic and calc-alkaline) exhibit different Ti isotopic evolution patterns (Millet et al. 2016; Deng et al. 2019; Johnson et al. 2019; 2023; Zhao et al. 2020; Hoare et al. 2020, 2022; Aarons et al. 2021; Greber et al. 2021; Storck et al. 2023). For example, alkaline magmatic series have significantly higher δ49/47Ti (exceeding 2‰) relative to subduction-related systems such as calc-alkaline magmatism and arc tholeiites (up to 0.7‰) at the same SiO2 content, likely related to the much higher initial melt TiO2 contents enabling early saturation and fractional crystallization of Fe-Ti oxide with higher TiO2 contents in the alkaline magmatic series than subduction zone lavas (Deng et al. 2019; Johnson et al. 2019; 2023; Aarons et al. 2020, 2021; Hoare et al. 2020, 2022). Such distinct Ti isotope evolution patterns in igneous systems have led to controversial conclusions on the geodynamic origin of felsic continental crust (Greber et al. 2017b; Deng et al. 2019; Aarons et al. 2020, 2021). Further studies are required to investigate the cause of these complex Ti isotope variations in different magmatic systems.

St. Helena Island (5º 40′W, 16º 00′S), in the South Atlantic, provides us with a natural laboratory in which to investigate the impact of alkaline magma differentiation on Ti isotopes. These lavas experienced fractional crystallization and mineral accumulation processes from a cogenetic mantle source, as deduced from their major and trace element chemistry, and radiogenic isotope characteristics (Baker 1969; Kawabata et al. 2011; Hanyu et al. 2014). This study investigates the Ti isotopic systematics of a well-characterized St. Helena alkaline sample suite (Kawabata et al. 2011; Hanyu et al. 2014; Wang et al. 2021; Zhao et al. 2022). Combined with published Ti isotopic ratios for other igneous rocks, we aim to further explore the origin of Ti isotope variation in different magmatic systems.

Samples and analytical methods

The studied alkaline lavas were collected from St. Helena Island in the middle of the South Atlantic (Chaffey et al. 1989; Kawabata et al. 2011). Twenty-four samples, covering basanite-alkali basalt, trachyandesite and trachyte, along with three hand-picked titanomagnetite separates from St. Helena Island were selected for Ti isotopes analyses. They have previously been studied for petrology, major and trace element composition, and radiogenic/stable isotopic composition, and their magmatic evolution has been well characterized (Kawabata et al. 2011; Hanyu et al. 2014; Wang et al. 2021; Zhao et al. 2022; Zhang et al. 2022). These samples span a range of SiO2 contents from 43.5 to 61.4 wt.%, MgO contents from 0.05 to 15.7 wt.%, total-alkali (Na2O + K2O) contents from 1.7 to 12.6 wt.%, and TiO2 contents from 0.11 to 3.86 wt.% (Fig. 1; Kawabata et al. 2011; Table S1 in Supporting Information S1). They can be classified into three groups according to their petrological and geochemical features (Kawabata et al. 2011). Group 1 samples (MgO content > 13 wt.%) contain a substantial amount of cumulate olivine and clinopyroxene (Fig. 1). Group 2 samples (MgO content; 5 wt.% to 12 wt.%) experienced olivine and clinopyroxene crystallization (Fig. 1; Kawabata et al. 2011). Group 3 samples (MgO content < 5 wt.%) experienced massive fractionation of plagioclase and Fe-Ti oxide, with minor amounts of apatite, olivine and clinopyroxene (Fig. 1; Kawabata et al. 2011). In Group 3 samples, Fe-Ti oxide (titanomagnetite) crystallized and fractionated as evidenced by the appearance of this phase (Wang et al. 2021; Zhao et al. 2022), and the clear decrease in TiO2 and FeOT contents at the 5 wt.% MgO inflection on whole rock diagrams (Fig. 1; Kawabata et al. 2011). Based on detailed scanning electron microscope (SEM) backscattered electron (BSE) images, energy-dispersive spectroscopy (EDS) spectra, and electro-probe microanalyzer data, Wang et al. (2021) and Zhao et al. (2022) have shown that all the Fe-Ti oxide phenocrysts from St. Helena lavas are ulvöspinel-rich titanomagnetite. SEM images also show that titanomagnetite phenocrysts in Group 3 samples are generally in contact with groundmass and devoid of exsolution lamellae of other Fe-Ti oxides, indicating they are equilibrium with the melt. The ulvöspinel-rich titanomagnetite separates analyzed from the three Group 3 samples (SH-62, SH-17, SH-89) were initially separated using Frantz magnetic selection, and then were hand-picked under a binocular microscope. The ulvöspinel-rich titanomagnetite separates in every sample show narrow chemical variations with TiO2 contents ranging from 18.7–22.7 wt.%, FeOT contents ranging from 65.5‒69.9 wt.%, and low contents of MgO (2.45‒4.68 wt.%) and SiO2 (0.04–0.12 wt.%) (Table 1; Wang et al. 2021; Zhao et al. 2022). More information about these volcanic rocks can be found in the literature (Baker 1968; Kawabata et al. 2011; Hanyu et al. 2014; Wang et al. 2021; Zhao et al. 2022).

Fig. 1
figure 1

Magmatic differentiation in the St. Helena lavas illustrated by a SiO2, b TiO2, c FeOT (total iron as FeO) and d Na2O + K2O contents as a function of MgO contents. The major element data and the composition of primary melt are from Kawabata et al. (2011) and are also available in Table S1 of Supporting Information S1. Orange lines represent the MELTS-calculated fractional crystallization paths of St. Helena lavas with initial oxygen fugacity of QFM + 1 and a constant pressure of 3 kbar (see Sect. 5.1 of Wang et al. (2021) for more information). All published data (small gray circles) for St. Helena rock samples are also shown for comparison, which are collected from GEOROC (http://georoc.mpch-mainz.gwdg.de/georoc/) database, and can be found in the Table S2 of Supporting Information S1 with detailed references. Ol  olivine, Cpx clinopyroxene, Pl plagioclase, Tmag  titanomagnetite, Ap apatite

Full size image
Table 1 Ti isotopic composition and selected major element data of alkaline volcanic rock samples from St. Helena Island
Full size table

Chemical separation and Ti isotope measurements were carried out at the Laboratory of Isotopic Geology, Institute of Geology, Chinese Academy of Geological Sciences, Beijing, China, following the methods outlined in Zhao et al. (2020) and Li et al. (2022). Details of the Ti isotope analytical methods are given in Supporting Information S2. Titanium isotope data are presented as δ values in per mil relative to the OL-Ti standard: δ49/47Ti (‰) = [(49Ti/47Ti)sample/(49Ti/47Ti)OL-Ti − 1] × 1000. The long-term average δ49/47Ti of NIST3162a and IGPG-Ti are 1.06 ± 0.04‰ (n = 173, 2SD) and 0.14 ± 0.03‰ (N = 42, 2SD), respectively (Table 1), consistent with published data (Greber et al. 2017a; Deng et al. 2018a; Zhao et al. 2020; Li et al. 2022). In addition, the δ49/47Ti values of BHVO-2, AGV-1 and GSP-2 obtained during this study agree well with previous works (Table 1; Millet and Dauphas 2014; Millet et al. 2016; Greber et al. 2017a, 2017b, 2021; Deng et al. 2018a, 2018b, 2019; Johnson et al. 2019; Mandl 2019; Zhao et al. 2020; He et al. 2020; Hoare et al. 2020, 2022; Li et al. 2022). Replicates of four samples (SH-30, and titanomagnetite phenocrysts of SH-17, SH-89 and SH-62), which were digested from different aliquots of sample powder, display identical results within uncertainty (Table 1).

Samples SH-30 and SH-42 were also analyzed for Fe isotopes to complement the existing dataset. These measurements were carried out on a Thermo Scientific Neptune Plus multi-collector inductively coupled plasma mass-spectrometer (MC-ICPMS) at the State Key Laboratory of Continental Dynamics of Northwest University, China, using the sample-standard bracketing method under “wet” plasma conditions (Chen et al. 2022). Details of the Fe isotope analytical methods are provided in Supporting Information S2. Each sample was measured a minimum of three times. The Fe isotope data are reported using δ notation against the IRMM-014 standard: δX/54Fe (‰) = [(XFe/54Fe)sample/(XFe/54Fe) IRMM-014 − 1] × 1000, where X refers to 57 or 56. The long-term reproducibility for δ57/54Fe data is better than 0.07‰ (2SD; Table S3 in Supporting Information S1). Reference materials BCR-2 and BHVO-2 were analyzed with samples, giving δ57/54Fe values similar to published values (Table S3 in Supporting Information S1; Craddock and Dauphas 2011; Sossi et al. 2012; Zhao et al. 2012; He et al. 2015; Konter et al. 2016; An et al. 2017; Zhao et al. 2022) when the equivalent 2 SD errors are considered.

Results

Titanium isotopic data for the St. Helena lavas are shown in Table 1, along with other geochemical data. The St. Helena whole-rock samples display large Ti isotopic variations with δ49/47Ti ranging from − 0.02 to 1.96‰ (Fig. 2 and Table 1). Specifically, samples from Group 1 and Group 2 (MgO > 5 wt.%) show insignificant δ49/47Ti variation (− 0.02 –0.05‰), regardless of their chemical composition (e.g., SiO2, TiO2, MgO or Na2O + K2O composition; Fig. 2 and Table 1), which agrees well with the limited range in δ49/47Ti reported for global oceanic basalts and komatiites (Millet et al. 2016; Greber et al. 2017a; Deng et al. 2018a, 2019, 2023; Zhao et al. 2020). Compared with the other two groups, Group 3 samples show a remarkable δ49/47Ti variation (from − 0.02 to 1.96‰ for δ49/47Ti; Fig. 2 and Table 1), comparable to other alkaline differentiated lavas from the Afar Rift in Ethiopia, Acension and Heard Islands (0.01 –2.32‰; Fig. 2; Deng et al. 2019; Johnson et al. 2019; Zhao et al. 2020; Hoare et al. 2020). Our new Ti isotope data for the St. Helena samples confirm previous observations that alkaline differentiated lavas show the largest range in Ti isotope composition followed by tholeiitic and calc-alkaline magmas (Fig. 2; Millet et al. 2016; Greber et al. 2017a, 2021; Deng et al. 2019; Johnson et al. 2019, 2023; Zhao et al. 2020; Hoare et al. 2020). In addition, titanomagnetite separates from Group 3 samples have δ49/47Ti values (− 0.54 to 0.01‰) remarkably lower than the corresponding whole rock samples (from − 0.02 to 0.21‰; Table 1) with Δ49/47TiTmag-whole rock (= δ49/47TiTmag − δ49/47Tiwhole rock) values ranging from − 0.68 to -0.22‰. This is consistent with previous investigations (Mandl 2019; Johnson et al. 2019, 2023; Greber et al. 2021; Rzehak et al. 2021, 2022; Hoare et al. 2022; Storck et al. 2023) and confirms the major control that Ti-oxide crystallization exerts on the Ti isotopic characteristics of igneous rocks.

Fig. 2
figure 2

Plot of δ49/47Ti vs SiO2 (a), TiO2 (b), MgO (c) and Na2O + K2O (d) contents for the studied St. Helena lavas. The horizontal line represents the average δ49/47Ti value of the bulk silicate Earth (BSE) (0.05 ± 0.02‰, 2SD; Deng et al. 2023). Also shown are literature data for alkaline (Ascension Island, Afar rift, Heard island; Hoare et al. 2020, 2022), subalkaline (Afar rift, Hekla in Iceland, Kilauea Iki; Deng et al. 2019; Johnson et al. 2019; Zhao et al. 2020), calc-alkaline (Santorini, Agung, Kos, Rindjani; Millet et al. 2016; Hoare et al. 2020, 2022; Greber et al. 2021; Johnson et al. 2023) and tholeiitic (Monowai seamount, Alarcon Rise, Koolau, Hawaii; Hoare et al. 2020, 2022; Zhao et al. 2020) rocks. Error bar denotes two standard deviation (2SD), with the exception of error bars from Millet et al. (2016), Johnson et al. (2019, 2023), Hoare et al. (2020, 2022) and Greber et al. (2021) which are 95% c.i.. The δ49/47Ti data for St. Helena samples are from Table 1

Full size image

The δ57/54Fe values of sample SH-30 and sample SH-42 are summarized in Table S3 in Supporting Information S1. The δ57/54Fe values range from 0.34 to 0.35‰, relatively higher than those previously measured in other Group 3 samples (0.17 to 0.32‰; Zhao et al. 2022).

Discussion

Our new results show significant Ti isotopic variability in St. Helena samples. Titanium is a moderately refractory element during igneous processes and a fluid immobile element (Pearce and Norry 1979; Kessel et al. 2005; Rapp et al. 2010), and thus Ti isotopic signatures are likely unaffected by processes such as low-temperature alteration. The Th/U ratios are commonly used as tracers for low-temperature alteration due to high mobility of U in oxidized geological fluids (Mathieu et al. 2001), while Th behaves as an immobile high field-strength element. U and Th are therefore often fractionated during low-temperature alteration. The absence of correlation between δ49/47Ti and Th/U for St. Helena samples (Fig. 3a), further suggests that low-temperature alteration has an insignificant effect on δ49/47Ti variation. Radiogenic isotope ratios, such as initial 87Sr/86Sr and 143Nd/144Nd, are valuable tracers of source heterogeneity, however, no discernable correlation exists between δ49/47Ti and initial 87Sr/86Sri (Fig. 3b) or 143Nd/144Nd (Fig. 3c) in St. Helena samples. This suggests source heterogeneity did not produce the measured Ti isotopic variability. Therefore, the large variation in δ49/47Ti observed in St. Helena samples likely reflects isotopic fractionation during high temperature igneous processes. This will be more fully explored below, in light of the established fractional crystallization sequence (Wang et al. 2021; Zhao et al. 2022).

Fig. 3
figure 3

δ49/47Ti versus Th/U (a), 87Sr/86Sri (b), and 143Nd/144Nd (c) for St. Helena samples analyzed in this study. The horizontal dashed line and shaded areas represent the average δ49/47Ti values of the BSE (0.05 ± 0.02‰, 2SD; Deng et al. 2023). Error bars are 2SD. δ49/47Ti data for St. Helena samples are from Table 1. The trace element and radiogenic isotope data are from Kawabata et al. (2011) and Hanyu et al. (2014), respectively, and can be found in Table S1 of Supporting Information S1

Full size image

Ti isotopic fractionation during fractional crystallization

St. Helena Group 1 samples (MgO contents > 13 wt.%) have a limited range of Ti isotopic compositions (δ49/47Ti = 0.03– 0.04‰; Table 1 and Fig. 2), suggesting that olivine and clinopyroxene accumulation had a negligible influence on the Ti isotopic composition of the St. Helena lava. Group 2 samples also display a narrow δ49/47Ti range (− 0.02–0.05‰; Table 1), and show no discernable relationship between δ49/47Ti and geochemical indicators of magma differentiation (e.g., SiO2, MgO, TiO2 and total-alkali contents; Fig. 2). This implies that fractional crystallization of olivine and clinopyroxene has little impact on the Ti isotopic composition of St. Helena lava. Titanium isotope fractionation during igneous processes is theoretically controlled by differences in bond stiffness and coordination number (CN) between mineral phases and the melt. A small CN and short bond favor incorporation of heavy isotopes at equilibrium (Urey 1947; Schauble 2004). Titanium is fourfold coordinated in olivine, and four to sixfold coordinated in silicate melt, depending on melt composition (average ∼4.5 to 5.4, Farges and Brown 1997), olivine is therefore predicted to be slightly isotopically heavy relative to silicate melt (Wang et al. 2020). This prediction was confirmed by Greber et al. (2021) who found that olivine had a heavier Ti stable isotope composition than the bulk rock. However, fractional crystallization of olivine seems to have a negligible effect on the Ti isotope composition of St. Helena lavas due to its negligible TiO2 contents. Unlike olivine, pyroxene can contain both six and fourfold coordinated Ti (Farges and Brown 1997). Previous studies (Leitzke et al. 2018; Wang et al. 2020; Aarons et al. 2021) have indicated that if Ti is solely hosted in sixfold coordination in pyroxene, no discernable Ti isotopic fractionation would be expected between the pyroxene and the silicate melt. On the other hand, when Ti is exclusively hosted in fourfold coordination in pyroxene, pyroxene are isotopically heavier than the coexisting melt. In additon, Rzehak et al. (2021, 2022) demonstrated that pyroxene can produce resolvable Ti isotope fractionation under more reducing lunar conditions. Ti primarily occupies a sixfold coordination in pyroxene at upper mantle P–T conditions (Ackerson et al. 2017), and clinopyroxene is the main Ti-bearing phase in St. Helena Group 1 and Group 2 samples. The lack of resolvable Ti isotope variation in the St. Helena Group 1 and Group 2 samples supports the negligible impact of clinopyroxene-induced Ti isotopic fractionation.

In contrast, the St. Helena Group 3 samples (MgO content < 5.0 wt%) display a wide range in δ49/47Ti (− 0.02 to 1.96‰; Table 1 and Fig. 2) and clear correlation between δ49/47Ti and magma differentiation indicators such as SiO2, MgO, TiO2, and total-alkali contents (Fig. 2). As noted above, Group 3 samples underwent massive fractionation of plagioclase and Fe-Ti oxide, with minor amounts of apatite, olivine, and clinopyroxene (Kawabata et al. 2011; Wang et al. 2021). Fractional crystallization of olivine and pyroxene had little effect on the Ti isotope variation in St. Helena Group1 and Group 2 samples according to our discussion above. Minerals like plagioclase and accessory apatite contain almost no TiO2, so the effect of their crystallization is also limited. However, ulvöspinel-rich titanomagnetite is characterized by a high TiO2 content (18.7–22.7 wt.%) and a remarkably lighter Ti isotopic composition (δ49/47Ti = -0.54 to 0.01‰) relative to the corresponding whole rock (δ49/47Ti = − 0.02 to 0.21‰; Table 1). When sufficient titanomagnetite crystallizes, it drives the evolved melt toward a heavier Ti isotopic composition and depletes the melt in TiO2 and FeOT, consistent with our observations (Fig. 1 and Fig. 2). This is because Fe-Ti oxides (such as titanomagnetite, rutile and ilmenite) exclusively host Ti in sixfold coordination (Leitzke et al. 2018; Wang et al. 2020; Aarons et al. 2021) and thus prefer light Ti isotopes relative to other silicate minerals and melts (Leitzke et al. 2018; Wang et al. 2020; Zhao et al. 2020; Aarons et al. 2021; Hoare et al. 2022). Consequently, the removal of Fe-Ti oxides would leave behind a residual magma isotopically heavy in Ti. This inference is supported by our newly measured and published Ti isotopic compositions of natural Fe-Ti oxide samples (Johnson et al. 2019, 2023; Mandl 2019; Aarons et al. 2021; Greber et al. 2021; Nie et al. 2021; Greber et al. 2021; Rzehak et al. 2021; Hoare et al. 2022), as well as experimental results (Rzehak et al. 2021, 2022; Hoare et al. 2022). Recent work by Wang et al. (2021) and Zhao et al. (2022) on the same samples from St. Helena have shown that Mg and Fe isotopic compositions of Group 3 samples vary and correlate with MgO contents, implying that these variations are largely controlled by ulvöspinel-rich titanomagnetite fractionation during magmatic differentiation. Most notably, the δ49/47Ti values of Group 3 samples negatively correlate with δ26/24Mg (Fig. 4a), and positively correlate with δ57/54Fe (Fig. 4b). We therefore conclude that ulvöspinel-rich titanomagnetite crystallisation was the dominant control on progressively heavier Ti isotope compositions in evolving St. Helena Group 3 samples.

Fig. 4
figure 4

δ49/47Ti versus δ26/24Mg (a) and δ57/54Fe (b) for St. Helena lavas. The horizontal dashed line denotes the average δ49/47Ti value of the BSE (0.05 ± 0.02‰, 2SD; Deng et al. 2023). The vertical solid line in panel (a) represents the average δ26/24Mg of the BSE (− 0.25 ± 0.04‰, Teng et al. 2017), and the vertical solid line in panel (b) represents the upper mantle Fe isotopic composition (δ57/54Fe = 0.04 ± 0.04‰, Weyer and Ionov 2007; Craddock et al. 2013). δ49/47Ti data for St. Helena samples are from Table 1. Error bars are 2SD. δ57/54Fe data for samples SH-30 and SH-42 are from Table S3 of Supporting Information S1. Literature Mg and Fe isotope data are taken from Wang et al. (2021) and Zhao et al. (2022), respectively, and can be found in Table S1 of Supporting Information S1

Full size image

Quantitative modeling

We modelled Ti isotopic fractionation during magma differention using the Rayleigh fractionation model described below:

$${{\delta }^{49/47}{Ti}_{melt}=(\delta }^{49/47}{Ti}_{initial melt}+1000)\times {{f}_{Ti}}^{(a-1)}-1000$$

where a ((49Ti/47Ti)mineral /(49Ti/47Ti)melt)) is the Ti isotope fractionation factor between mineral and melt, and fTi is the mass fraction of Ti in the remaining melt. fTi is estimated by comparing the Ti/Th ratios of the samples, assuming a partition coefficient of zero between mineral and melt for Th. In our simulation, sample SH-58 (δ49/47Ti = 0.05‰), with the highest MgO (4.97 wt%) and TiO2 (3.51 wt%) was chosen as the initial melt composition, with MgO ≈ 5 wt.% marking the onset of Fe-Ti oxide saturation (Kawabata et al. 2011; Wang et al. 2021; Zhao et al. 2022). Figure 5a shows that the fTi vs. δ49/47Ti trend of St. Helena Group 3 samples can be reconciled by ∆49/47Timineral-melt (= 1000lna49/47Timineral-melt) of − 0.40 to − 0.30‰ (see Table S4 in Supporting Information S1 for details of quantitative modelling).

Fig. 5
figure 5

Comparison of Ti isotope composition of samples from the St. Helena with a Rayleigh fractionation model using different bulk isotope fraction factors (1000 lnαmineral-melt) between all crystallizing minerals and the silicate melt (see Table S4 of Supporting Information S1 for details) (a). Plot of δ49/47Ti vs MgO contents for the studied St. Helena lavas (see Table S5 of Supporting Information S1 for details of quantitative modelling) (b). The light orange shaded area represents the uncertainty of the calculated δ49/47Ti variation that derives from the uncertainty of the titanomagnetite-melt isotopic fractionation factor (Δ49/47TiTmag-melt = [(− 0.058 ± 0.015)*TiO2 + (0.15 ± 0.24)] × 106/T2) from Johnson et al. (2023). The major element data of primary melt is from Kawabata et al. (2011) and the δ49/47Ti value of primary melt is assumed as 0.00‰. The horizontal dashed line and shaded areas represents the average δ49/47Ti values of BSE (0.05 ± 0.02‰, 2SD; Deng et al. 2023). δ49/47Ti data for St. Helena samples are from Table 1. Error bars are 2SD. Major element data for the studied samples are from Kawabata et al. (2011)

Full size image

To further assess the impact of titanomagnetite crystallization on Ti isotopic composition of St. Helena lavas, we modelled Ti isotopic fractionation following the modelling method of Wang et al. (2021) and Zhao et al. (2022), who quantitatively evaluated the effect of each separated mineral on the Mg and Fe isotopic compositions of St. Helena lavas, respectively. An isotope mass balance model is used to calculate the Ti isotopic composition of residual melts after each crystallization step, and it can be expressed as the following equations:

$$\delta^{49/47} Ti_{melt,1} = \,\,\frac{{\delta^{49/47} Ti_{melt,2} \, \times \,f_{melt,2} \, \times \,TiO_{2 melt,2} \, + \,\sum\nolimits_{i = 1}^{n} {\delta^{49/47} Ti_{i} } \, \times \,f_{i, 2} \, \times \,TiO_{2i,2} }}{{\left( {f_{melt,2} \, \times \,TiO_{2 melt,2} \, + \,\sum\nolimits_{i = 1}^{n} {f_{i, 2} } \, \times TiO_{2i,2} } \right)}}$$
(1)
$$\delta^{49/47} Ti_{i} = \Delta^{49/47} Ti_{{{\text{i}} - melt}} + \delta^{49/47} Ti_{melt, 2}$$
(2)

where δ49/47Timelt, 1 and δ49/47Timelt, 2 are the Ti isotopic compositions of the melts before and after a crystallization step, respectively. δ49/47Tii is the Ti isotopic composition of mineral i, and Δ49/47Tii-melt is the Ti isotopic fractionation between mineral i and melt. \({{TiO}_{2}}_{melt,2}\) and \({{TiO}_{2}}_{{\text{i}}, 2}\) denote the TiO2 contents of residual melts and separated mineral i after a crystallization step, respectively. The proportions of segregated mineral i and residual melts after each crystallization step are marked as fi, 2 and fmelt, 2, respectively. Olivine, clinopyroxene, and titanomagnetite are used in the modeling calculations. Apatite and plagioclase are ignored in our modeling calculations due to their very low TiO2 contents. The chemical compositions and proportions of segregated minerals and residual melts during the evolution history of St. Helena samples are from published MELTS calculation results (Table S5 in Supporting Information S1; Wang et al. 2021). The melt-mineral isotope fractionation between melts and silicate minerals such as olivine and clinopyroxene is assumed to be zero, and the titanomagnetite-melt isotopic fractionation factor is from Johnson et al. (2023). The modeling results are reported in Table S5 in Supporting Information S1 and illustrated in Fig. 5b and Fig. 6. As shown in Fig. 5b, the measured trend of δ49/47Ti versus MgO for St. Helena samples can be well modeled using the above method. Furthermore, the observed trends of δ49/47Ti versus SiO2, TiO2 and total-alkali contents can be modeled for most St. Helena samples except for the two most evolved sample (e.g., SH-30, SH-42) with extremely low MgO (0.05–0.06 wt.%) and TiO2 (0.11–0.23 wt.%) contents (Table 1 and Fig. 6). This discrepancy is probably due to the fact that the MELTS calculations failed to perfectly reproduce the true modes of Fe-Ti oxide minerals and especially the accurate MgO and TiO2 contents of the residual melt and Fe-Ti oxides near the end of magma differentiation at St. Helena (Wang et al. 2021). Nonetheless, our modelling results reaffirm that titanomagnetite crystallization is the major control on the progressive elevation of δ49/47Ti observed in St. Helena Group 3 samples.

Fig. 6
figure 6

Plot of δ49/47Ti vs SiO2 (a), TiO2 (b), and Na2O + K2O (c) contents for the studied St. Helena lavas. The orange lines show the modeled variation of δ49/47Ti with changing SiO2, TiO2, and Na2O + K2O contents (see Table S5 of Supporting Information S1 for details of quantitative modelling). The light orange shaded area represents the uncertainty of the calculated δ49/47Ti variation that derives from the uncertainty of the titanomagnetite-melt isotopic fractionation factor (Δ49/47TiTmag-melt = [(− 0.058 ± 0.015)*TiO2 + (0.15 ± 0.24)] × 106/T2; Johnson et al. 2023). The major element data of primary melt is from Kawabata et al. (2011) and the δ49/47Ti value of primary melt is assumed as 0.00‰. The horizontal dashed line and shaded areas represents the average δ49/47Ti values of BSE (0.05 ± 0.02‰, 2SD; Deng et al. 2023). δ49/47Ti data for St. Helena samples are from Table 1. Error bars are 2SD. Major element data for the studied samples are from Kawabata et al. (2011)

Full size image

Comparison with previous studies

Several studies have shown that differentiated lavas (alkaline, tholeiitic and calc-alkaline series) experience different Ti isotopic evolution pathways during magmatic evolution (Deng et al. 2019; Johnson et al. 2019, 2023; Hoare et al. 2020, 2022; Zhao et al. 2020). New Ti isotopic analyses for the St. Helena samples confirm this conclusion (Fig. 2). For example, at a given SiO2 content, the St. Helena Group 3 samples and other alkaline magmas show a greater increase in δ49/47Ti (− 0.04 to 2.32‰; Deng et al. 2019; Johnson et al. 2019; Hoare et al. 2020; Zhao et al. 2020) than tholeiitic (0.01 to 1.06‰; Zhao et al. 2020; Hoare et al. 2020, 2022) and calc-alkaline (0.04 to 0.65‰; Millet et al. 2016; Greber et al. 2021; Hoare et al. 2020; 2022; Johnson et al. 2023) magma series (Fig. 2a). These differences likely result from caused by crystallization of a range of Fe-Ti oxides (e.g., magnetite, ilmenite, and pseudobrookite) with contrasting Ti isotopic compositions throughout magma evolution (Deng et al. 2019; Johnson et al. 2019, 2023; Hoare et al. 2020, 2022; Zhao et al. 2020). Both theoretical models and measurement of rock samples have demostrated that titanomagnetite is highly enriched in isotopically light Ti relative to other Fe-Ti oxides (ilmenite, rutile, and pseudobrookite) at equilibrium (Johnson et al. 2019, 2023; Wang et al. 2020; Zhao et al. 2020; Hoare et al. 2020, 2022; Greber et al. 2021). Therefore, at magmatic temperatures, progressive crystallization of titanomagnetite-dominanted Fe-Ti oxide raises the melt δ49/47Ti values more than crystallization of other Fe-Ti oxides (ilmenite, rutile, and pseudobrookite). Both Ti4+ and Fe2+ occupy the six-fold site in titanomagnetite. Given that the Fe2+ ion is larger than Ti4+, titanomagnetite has a relatively longer Ti–O bond length (2.05–2.06 Å) than other Fe-Ti oxides such as ilmenite, rutile, and pseudobrookite (1.94–1.99 Å) (Smyth and Bish 1988; Howard et al. 1991; Farges et al. 1996a, b; Bosi et al. 2009). This longer Ti–O bond length means that light Ti is preferentially accommodated in the titanomagnetite structure at equilibrium (Bigeleisen and Mayer 1947; Urey 1947; Schauble 2004). Our results show that the St. Helena Group 3 samples display higher δ49/47Ti than other alkaline differentiated lavas at a given SiO2 from the Afar Rift in Ethiopia, Acension and Heard Islands (Fig. 2; Deng et al. 2019; Johnson et al. 2019; Zhao et al. 2020; Hoare et al. 2020). As discussed above, magmas that crystallize various Fe-Ti oxides (such as titanomagnetite, rutile and ilmenite) with divergent Ti isotopic compositions will preserve distinct Ti isotope compositions. At St. Helena, ulvöspinel-rich titanomagnetite is the only crystallization Fe-Ti oxide phase during the compositional range of our studied Group 3 samples (Fe-Ti oxide saturated), whereas both titanomagnetite and ilmenite are Ti-bearing crystallization phases in other alkaline differentiated lavas. Ulvöspinel-rich titanomagnetite is expected to have much lower δ49/47Ti than other Fe-Ti oxides (e.g., Hoare et al. 2020, 2023; Johnson et al. 2023). The difference between St. Helena Group 3 samples and other alkaline differentiated lavas is probably due to an earlier saturation and higher extent of ulvöspinel-rich titanomagnetite crystallization from St. Helena Group 3 samples compared to other alkaline differentiated lavas.

Another factor controlling Ti isotopic fractionation is the composition of crystallising titanomagnetite. For example, the TiO2 content (15.9–28 wt.%) of titanomagnetite crystallized from alkaline/sub-alkaline lavas in intraplate settings (Afar, Ascension, Kilauea, Hekla, Heard; Deng et al. 2019; Johnson et al. 2019; Zhao et al. 2020; Hoare et al. 2020, 2022) is significantly higher than that of titanomagnetite crystallized from arc-related calc-alkaline lavas (TiO2 = 7–13.0 wt.%; Hailar, Kos Agung, Santorini, Rindjani; Millet et al. 2016; Hoare et al. 2020, 2022; Zhao et al. 2020; Greber et al. 2021; Johnson et al. 2023). This compositional contrast likely explains the difference in the Ti isotopic evolution path of rocks in these two settings (Fig. 2). This speculation is supported by Hoare et al. (2022) who observed that Ti–rich titanomagnetite (~ 22 wt.%, TiO2) was isotopically lighter than Ti-poor magnetite (~ 14 wt% TiO2). This is also evidenced by a subsequent study by Johnson et al. (2023) who observed that Rindjani Ti-poor calc-alkaline lavas crystallizing Ti-poor magnetite (7–10 wt%, TiO2) experienced less Ti isotopic fractionation compared to other magma series. Further corroboration comes from the results produced in this work where the dominant Fe-Ti oxide in the St. Helena Group 3 samples is TiO2 enriched ulvöspinel-rich titanomagnetite (18.7–22.7 wt.% TiO2; Wang et al. 2021; Zhao et al. 2022) with a remarkably light Ti isotopic composition (δ49/47Ti = − 0.54–0.01‰; Table 1). The diverse Ti isotope fractionation trends observed in various magmatic systems is indirectly governed by the parental melt chemistry and redox state, which could control the onset of Fe-Ti oxide crystallizaion, as well as its modal abudance and compostition (Deng et al. 2019; Zhao et al. 2020; Hoare et al. 2020, 2022). For example, alkaline magmas from intraplate settings (reduced and H2O-poor) have higher initial melt TiO2 contents than those of tholeiitic and calc-alkaline lavas (oxidized and H2O-rich; Prytulak and Elliot 2007), which enables early saturation of Ti–rich titanomagnetite or ilmenite with high modal abundance (Hoare et al. 2020, 2022). This will also drive larger Ti fractionation in alkaline magmas relative to tholeiitic and calc-alkaline arc lavas with lower TiO2 contents and less Fe-Ti oxide (Deng et al. 2019; Zhao et al. 2020; Hoare et al. 2020, 2022). Additionally, recent studies by Storck et al. (2023) and Johnson et al. (2023) have demonstrated that the relative Ti fractions in crystallized minerals (silicate minerals such as amphibole or pyroxene vs. Fe-Ti oxides) can also influence Ti stable isotope fractionation in magmatic systems. As outlined in Johnson et al. (2023), this effect might be an indirect result of the elevated Fe3+/Fe2+ ratio in calc-alkaline systems.

Conclusions

This study explores the Ti isotope fractionation behavior during the differentiation of St Helena alkaline magma. Important conclusions include:

  1. (1)

    The δ49/47Ti range of St. Helena Group 1 and Group 2 samples (MgO > 5 wt.%) is quite narrow (− 0.02 to 0.05‰). The absence of systematic correlations between δ49/47Ti and the chemical indicators of magmatic differentiation (e.g., SiO2, MgO, TiO2 and Na2O + K2O contents) indicate that accumulation or crystallization of olivine and clinopyroxene have negligible influence of the δ49/47Ti variation of St. Helena Group 1 and Group 2 samples.

  2. (2)

    Significant Ti isotopic variations (δ49/47Ti =  − 0.02 to 1.96‰) are observed for St. Helena Group 3 samples (MgO < 5 wt.%), and there are clear correlations between δ49/47Ti and SiO2, MgO, TiO2 and total-alkali contents, which are the result of fractional crystallization of isotopically light ulvöspinel-rich titanomagnetite. This is reinforced by the measured lower δ49/47Ti values for ulvöspinel-rich titanomagnetite relative to whole rock (Δ49/47TiTmag-whole rock = − 0.68 to − 0.22‰) of the St. Helena Group 3 samples. Quantitative geochemical modeling also support crystallization and removal of isotopically light ulvöspinel-rich titanomagnetite resulting in isotopically heavy St. Helena Group 3 samples.

  3. (3)

    Our study further confirms that the Ti isotopic fractionation trends observed in different magmatic systems (alkaline, tholeiitic and calc-alkaline series) could be the result of crystallization of different Fe-Ti oxides (Titanomagnetite, ilmenite, magnetite) with distinct light Ti isotopic compositions depending on the different melt TiO2 composition, oxygen fugacity and water activity.