We were the first to publish in 2004 a review of mean concentrations of major components and volatile and trace elements in magmatic melts in the Earth’s dominant geodynamic environments based on our database, which had been composed starting from 1994 in the system Paradox for Windows. Data in the database are analyses of glass in naturally quenched and experimentally homogenized melt inclusions in minerals and quench glasses in volcanic rocks (Naumov et al., 2004). The total number of analyses in the database was then almost 14 000, and the database comprised 190 500 determinations of 60 elements. Our next review included 33 000 analyses for 73 elements and was published in 2010 (Naumov et al., 2010). The number of measurements in the database amounted to 480 000. The further intensification of such studies is well illustrated by data in Table 1.

Table 1. Number of publications on mineral-hosted melt inclusions and quench glasses in rocks included in our database during various periods of time
Full size table

The total number of the publications was 1936, and the overall number of the objects was 145 000, which were analyzed for 75 elements. Nowadays the database comprises more than 2 600 000 measurements: 1 312 000 analyses for major components, 225 000 analyses for volatiles (H2O, Cl, F, S, and CO2), 640 000 analyses for ore and trace elements, 415 500 analyses for REE, and the database additionally contains 24 500 measurements of homogenization temperatures and >7700 measurements of pressures in natural magmatic melts.

Acute interest of many researchers in mineral-hosted inclusions is explained by that these inclusions provide likely the most straightforward and reliable means of estimating the composition and physicochemical parameters of natural magmatic melts. These studies were even more activated after quantitative analytical methods of high spatial resolution were invented: electron, ion, and proton probes; Raman spectroscopy; laser ablation–inductively coupled plasma–mass spectrometry, local IR spectroscopy, etc. Some lately published papers present data even on the isotopic composition of some elements in individual melt inclusions (e.g., Eiler et al., 2007; Bouvier et al., 2008; Le Voyer et al., 2008; Harlou et al., 2009; Layne et al., 2009; Wittenbrink et al., 2009; Anderson et al., 2021; Hartley et al., 2021; Li et al., 2021; Kawaguchi et al., 2022).

Our work was centered on analysis of similarities and differences between the behaviors of elements in natural magmatic systems and factors that control the most general features of the geochemistry of rocks in the Earth’s dominant geodynamic environments. This analysis was carried out based on evaluation of the mean concentrations of elements in corresponding melts of mafic, intermediate, and felsic composition.

CHARACTERISTICS OF THE EARTH’S DOMINANT GEODYNAMIC ENVIRONMENTS

Herein we distinguish the following geodynamic environments that differ from one another in the parameters of generation and evolution of magmatic melts: (I) spreading environments of oceanic plates (mid-oceanic ridges), (II) environments in which effects of mantle plumes are discernible (on oceanic islands and lava plateaus), (III and IV) environments related to subduction processes (III is environments of arc magmatism, and IV is zones of magmatism on active continental margins in which continental crust is involved in the magma-generating processes), (V) environments of continental rifts and continental hotspots, and (VI) environments of backarc spreading.

Figure 1 presents a histogram of the distribution of SiO2 concentrations in mineral-hosted homogeneous melt inclusions and in quenched glasses in volcanic rocks from all of the aforementioned geodynamic environments.

Fig. 1.
figure 1

Histogram showing the distribution of SiO2 concentrations in magmatic melts from all of the Earth’s principal geodynamic environments: data on mineral-hosted inclusions and quench glasses (n is the number of determinations).

Full size image

The distribution is obviously bimodal, with one of the maxima occurring at SiO2 = 50–52 wt % and the other at 72–74 wt %. The falls on um SiO2 concentrations of 62–64 wt %. We have previously mentioned the bimodal character of naturally occurring magmatic melts, with this conclusion based on 3465 analyses (Naumov et al., 2004), and later confirmed this based on 33 000 analyses (Naumov et al., 2010). It seems to be reasonable to conclude that the further amassing of the analytical data will not any significantly modify this conclusion.

The histograms shown in Fig. 2 demonstrate how SiO2 concentrations are distributed between magmatic melts in the various geodynamic environments (I–VI).

Fig. 2.
figure 2

Histograms showing the distribution of SiO2 concentrations in natural magmatic melts from all of the Earth’s geodynamic environments (I−IV): data on mineral-hosted homogeneous inclusions and quench glasses (n is the number of determinations). Environments: (I) mid-oceanic ridges, (II) oceanic islands, (III) island arcs, (IV) active continental margins, (V) continental rifts and hotspot areas, (VI) backarc basins.

Full size image

Melts in environment I (mid-oceanic ridges) are mostly of mafic composition. Environment II (oceanic islands) is also dominated by mafic melts, and ultramafic ones are quantitative strongly subordinate. Similar distribution types are characteristic of environments III and IV (island arcs and active continental margins). Intraplate continental environments (V) are characterized by a trimodal SiO2 distribution. The magmatic melts show broad variations in SiO2 concentration from 40 to 80 wt %. Data on backarc basins (environment VI) are still relatively sparse (2150 determinations), but they still show the dominance of mafic melts. In general and with regard to this distribution of the SiO2 concentrations (Figs. 1 and 2), the mean concentrations of major components and volatile, trace, and rare-earth elements were calculated for three types of magmatic melts: mafic and ultramafic ones (SiO2 = 40–54 wt %), intermediate and low-SiO2 felsic melts (SiO2 = 54–66 wt %), and felsic ones (SiO2 > 66 wt %).

For each of the melt types, mean concentrations were calculated separately for each of the geodynamic environments (see above). Our earlier publications (Naumov et al., 2004, 2010, 2016, 2022) demonstrated that it is more reasonable to calculate geometric (but not arithmetic) mean concentrations, because the distributions of many elements are closer to lognormal ones. Lognormal distributions of trace elements and, hence, more adequate usage of geometric means have also been stressed by other researchers (Gale et al., 2013). Geometric mean concentrations were calculated at 95% confidence level. Determinations that did not meet this criterion were rejected, and the mean values were recalculated. The calculated mean concentrations of major oxides and volatile and trace elements are summarized in Tables 2–5.

Table 2. Mean concentrations of major components and volatile, ore, and trace elements in ultramafic and mafic magmatic melts (SiO2 = 40–54 wt %) in the dominant geodynamic environments (I−VI): data on mineral-hosted melt inclusions and quench glasses in volcanic rocks
Full size table
Table 3.   Mean concentrations of major components and volatile, ore, and trace elements in intermediate magmatic melts (SiO2 = 54–66 wt %) in the dominant geodynamic environments (II−VI): data on mineral-hosted melt inclusions and quench glasses in volcanic rocks
Full size table
Table 4.   Mean concentrations of major components and volatile, ore, and trace elements in felsic magmatic melts (SiO2 > 66 wt %) in the dominant geodynamic environments (II−VI): data on mineral-hosted melt inclusions and quench glasses in volcanic rocks
Full size table
Table 5. Ratios of elements in mafic magmatic melts (SiO2 = 40–54 wt %) in the dominant geodynamic environments (I−VI): data on mineral-hosted melt inclusions and quench glasses in volcanic rocks
Full size table

How much the mean concentrations of elements have changed when new data were added and the number of the determinations has been significantly increased compared to the earlier ones? To estimate these changes, we compared data on mafic melts from all of the geodynamic environments published in (Naumov et al., 2010) with data on melts from the same environments in Table 2. The difference for major elements was 6.8 rel. %, that for volatile elements (H2O, Cl, F, S, and CO2) was 19.2 rel. %, that for nine trace elements (Li, V, Rb, Sr, Y, Zr, Ba, Th, and U) was 10.7 rel. % and that for REE was 9.4 rel. %.

Figures 3 and 4 show primitive mantle-normalized multielemental patterns for the mean concentrations of elements (Tables 2−4) in magmatic melts from geodynamic environments I−VI.

Fig. 3.
figure 3

Primitive mantle-normalized (Sun and McDonough, 1989) trace-element and REE patterns of mafic, intermediate, and felsic melts in (I) mid-oceanic ridges, (II) oceanic islands, (III) island arcs, (IV) active continental margins, (V) continental rifts and continental hotspot areas, and (VI) backarc spreading areas.

Full size image
Fig. 4.
figure 4

Primitive mantle-normalized (Sun and McDonough, 1989) REE patterns of mafic, intermediate, and felsic melts in (I) mid-oceanic ridges, (II) island arcs, (III) oceanic islands, (IV) active continental margins, (V) continental rifts and continental hotspot areas, and (VI) backarc spreading areas.

Full size image

No such analysis of the distribution of trace elements in melts of different composition from different geodynamic environments extend beyond the scope of this paper. Many related issues have been discussed in papers on the analysis of data on some of these environments (e.g., Kovalenko et al., 2006, 2007, 2009). Herein we would only like to stress some issues commonly emerging when the whole data set on mineral-hosted inclusions and glasses in rocks is analyzed.

(1) The mean compositions of mineral-hosted inclusions and glasses in rocks confirm the presence of characteristic geochemical fingerprints of mafic, intermediate, and felsic melts in each of the environments. First of all, this is the fan-shaped configuration of the normalized patterns of trace elements in mafic and intermediate melts. The greatest differences were found between the melts of environments I and V.

(2) The normalized trace-element patterns of intermediate and mafic melts are generally similar and differ from those of felsic rocks. This may indicate that the sources of intermediate and mafic melts were similar in each environment. It is also reasonable to hypothesize that the leading mechanism generating intermediate melts is the differentiation of mafic (mantle) magmas, whereas the sources and generation mechanisms of felsic magmas were different (derivation from crustal rocks).

(3) A remarkable feature of the patterns, first of all, those of mafic melts, is the very narrow range of the concentrations of the least incompatible elements in the right-hand parts of the patterns (from Tb through Lu). This uniformity indicates that the mantle sources and generation parameters of the magmas were generally similar, because the concentrations and ratios of HREE and other mildly incompatible elements most strongly depend on the generation parameters of the melts, in contrast to those of the most incompatible elements, whose concentrations in mantle sources significantly vary when even small melt and fluid portions are transferred (mantle metasomatism). It should be kept in mind that the relative concentrations of the least incompatible elements are least significantly affected by processes of crustal contamination, changes in the source compositions under the effect of fluid- and melt-assisted material transfer, etc. The small differences in these parts of the patterns are the most illustrative in this context. These components are most significantly enriched in the melts of mid-oceanic ridges, a fact that may indicate that the degrees of melting were the lowest, and the magma generation depths were the shallowest, in this environment.

(4) The increase in the normalized concentrations of elements with an increase in their incompatibility generally replicates the distribution character of elements in the mean composition of the continental crust. At the same time, a simple model of the direct contamination of mantle melts (like the depletion of melts in environment I) with crustal material is inacceptable, because the mean concentrations of some incompatible elements in mafic melts are higher than in the continental crust. For example, the Ta and La concentrations in the continental crust are 0.7 and 20 ppm, respectively (Rudnick and Gao, 2003), whereas those in the melts of environment V are 1.47 and 30.2 ppm, respectively. More promising models seem to be those involving the derivation and migration of small portions of highly enriched melts and/or fluids that modify the compositions of the mantle sources and/or mantle magmas.

(5) Melts related to the environments of continental margins (III and IV) are noted for significant variations in the left-hand parts of the patterns, a phenomenon commonly described with reference to geochemical anomalies. A shining example of such anomalies is the widely known negative Ta−Nb anomaly. The nature of this anomaly is still uncertain. A strong decrease in Nb and Ta concentrations in melts could not result from the mixing of magmas from different sources, because the concentration levels of these elements are very low (particularly in melts of environment III, in which Ta and Nb concentrations are lower than in the mafic magmas of mid-oceanic ridges). Both Nb and Ta were likely retained in the residue in the course of melting. These elements are selectively concentrated in, for example, rutile, but mafic melts in equilibrium with rutile would have been notably enriched in Ti, which is not the case.

(6) Another interesting feature of the mafic melts of environments III and IV is their positive Pb anomaly (high Ce/Pb = 7.41 and 6.29, respectively). The origin of this anomaly is also disputable. Both elements are strongly incompatible when silicate minerals crystallize from melts. The fractionation of Ce and Pb can be driven by sulfides, because Pb is a typical chalcophyle element (Hart and Gaetani, 2006). However, sulfides are quite commonly found in mafic magmas in all environments, whereas a Pb anomaly is obvious in the melts only of two of them. Another possible enrichment mechanism of magmas (or their sources) in Pb relative Ce is material transfer with aqueous fluids derived at the degassing of a subducted slab (Ayers, 1998). Note that a Pb anomaly is also typical of intermediate and felsic melts, and the Pb anomaly of the latter is pronounced even better than their Nb anomaly.

Table 5 lists some average ratios of components in mafic magmas from various environments. Obviously, the number of analyses for the elements has significantly increased over the past 12 years. This information quantitatively appends the conclusions that can be derived from the analysis of the multielemental patterns. Many of the variation ranges significantly overlap and very little change from one environment to another, for example, those of the Th/U, P2O5/F, and TiO2/Dy ratios. These ratios are insusceptible to geodynamic environments, and the insignificant variations in these ratios were likely caused by local processes. The rest of the ratios are distributed more contrastingly, but none of them can be used to reasonably reliably determine the affiliation of a composition with any of the six environments. Moreover, some parameters make it possible to combine some environments into larger groups within which differences are insignificant. For example, environments of platform boundaries (III and IV) notably differ from environments related to mantle plumes in Ce/Pb, Nb/U, Zr/Nb, and Th/Ta ratio. Some parameters of environment I are closely similar to those of “plume” environments (for example, the Th/Ta ratio), whereas other parameters make this environment similar to environments III and IV (for example, the Zr/Nb ratio). Environment I principally differs from all other environments in the La/Yb and Th/Yb ratios (these ratios can be employed as indicators of the depleted mantle). It is worth mentioning that this depletion is not associated with any significant changes in the ratios of incompatible elements: for example, the H2O/Ce and Ce/Pb ratios in environments I and II are similar but principally differ from those in III + IV. Quantitative interpretations of these differences is a challenging task, because they may be related to the redistribution of components with the involvement of melts and fluids that had been derived under different thermodynamic parameters. Some of the differences may have likely been inherited during various evolutionary stages of geospheres. In this context, it would be interesting to trace the changes in the ratios of elements in similar melts with time, but such data on ancient complexes are still very sparse.

CONCLUSIONS

(1) Our database has been remarkably extended and now includes more than 2 600 000 determinations of 75 elements in mineral-hosted melt inclusions and in quench glasses in volcanic rocks. These data were generalized, and mean concentrations of major components and volatile, ore, and trace elements in magmatic melts were calculated for all of the dominant Earth’s geodynamic environments.

(2) Our newly acquired data confirm that, from the viewpoint of geochemical specifics, all melts data on which are available from the current version of the database definitely belong to any of the following six types, which correspond to the previously distinguished geodynamic environments: (I) environments of the spreading of oceanic plates (mid-oceanic ridges), (II) environments with mantle plumes at oceanic plates (oceanic islands and lava plateaus), (III and IV) environments related to subduction processes (III is arc magmatic zone on the oceanic crust, and IV is magmatic zones in active continental margins in which magma-generating processes involve the continental crust), (V) continental rifts and areas with continental hotspots, and (VI) environments of backarc spreading (Naumov et al., 2010).

(3) The distribution of SiO2 concentrations in natural magmatic melts is bimodal in all of the geodynamic environments: one of the maxima occurs at SiO2 = 50–52 wt %, and the other at 72–76 wt %. The smallest number of analyses corresponds to SiO2 concentrations of 62−66%.

(4) The primitive mantle-normalized multielemental patterns for the mean compositions of mafic, intermediate, and felsic rocks in environments I−VI show inherent features of melt compositions in each of the environments.

(5) Generalized and averaged data on the composition of mineral-hosted melt inclusions and glasses in rocks were used to calculate the average ratios of incompatible trace and volatile components (e.g., H2O/Ce, K2O/Cl, Nb/U, Ba/Rb, and Ce/Pb) in the magmatic melts of each of the environments. The variations in these ratios were calculated, and it is demonstrated that ratios of incompatible elements are mostly statistically significantly different in the different environments. The differences are particularly significant in the different environments, and these differences are the largest between the ratios of elements of different incompatibility (for example, Nb/Yb) and some volatiles (for example, K2O/H2O).