A tool to distinguish magmatic from secondarily recrystallized carbonatites—Calcite/apatite rare earth element partitioning

Carbonatites are thought to be part of the deep carbon cycle, and crustal isotopic and trace-element signatures in carbonatites are often interpreted to stem from subduction-related recycling. However, the vast majority of carbonatites are exposed in upper-crustal or subvolcanic complexes of the continental crust, providing ample opportunities for crustal contamination upon magma resurfacing. To better constrain the genesis of carbonatites and its bearing on the deep carbon cycle, a new geochemical approach is needed to distinguish chemical signatures obtained during magma formation and ascent in the mantle from those acquired during crustal emplacement. This endeavor is challenging due to the fact that carbonatites do not represent melt compositions but rather are magmatic cumulates or secondarily recrystallized rocks (Kamenetsky et al., 2021). Experimental studies (Lee and Wyllie, 1998; Brooker and Kjarsgaard, 2011; Weidendorfer et al., 2016) and natural melt inclusions in apatite and magnetite (Guzmics et al., 2011) indicate that primary carbonatite melts yield 5–20 wt% bulk Na₂O + K₂O, while carbonatite rocks are strongly depleted in alkalis. Moreover, eutectic or peritectic magmatic processes do not allow for the characteristic (almost) monomineralic intrusive carbonatites representing melt compositions. Overall, directly deciphering the origin of carbonatites based on bulk-rock majorelement, trace-element, and isotope chemistry is often misleading and, therefore, requires an unbiased geochemical approach such as the one presented in this paper.

Experimental studies that have reconstructed the primary carbonatite melt composition through mineral-melt equilibria have been limited to date to phosphate (Klemme and Dalpé, 2003; Hammouda et al., 2010; Chakhmouradian et al., 2017) and silicate mineral-melt element partitioning. Even though calcite and dolomite are rock-forming minerals in carbonatites, it is surprising how few data (Chebotarev et al., 2019) exist on carbonates equilibrated with carbonatite melt.

Our study provides a tool that allows researchers to distinguish the magmatic or secondary nature of calcite. The tool is based on experimentally determined mineral/mineral rare earth element (REE) partition coefficients (D), thus circumventing the problem of an unknown melt composition. A comparison of the experimental values to natural calcite/apatite (Cc/Ap) pairs allows us to determine whether calcites in a rock sample are primary magmatic minerals or secondarily replaced by fluid-mediated processes.

REE enrichments of economic carbonatites may originate from magmatic differentiation (Nabyl et al., 2020), from segregation of brine-like fluids of magmatic origin (Anenburg et al., 2021), and/or from hydrothermal fluid-rock interaction (Perry and Gysi, 2018; Walter et al., 2021). Secondary processes (including deuteric fluids) can alter the primary mineralogy and change the concentrations of Sr, C, and light REEs (LREEs) by redistributing such fluid-mobile elements and possibly alter the isotopic composition. Most carbonatites form from highly differentiated magmas within the shallow continental crust, where syn-, para-, and postmagmatic fluids inevitably cause alteration (Yaxley et al., 2022). At the same time, many carbonatite plugs, dikes, or veinlets crosscutting igneous or metamorphic rocks appear to be massive and texturally mature, even when strongly hydrothermally altered. This renders textural discrimination between magmatic and fluid-mediated processes difficult.

Apatite helps to disclose the primary nature of calcite since (1) both minerals begin to form in the early stages of crystallization of the carbonatite melt, (2) the high REE contents and its presence in almost every fossil carbonatite render apatite suitable for trace-element partitioning studies, and (3) possible deuteric or hydrothermal alterations in carbonatites mostly result in monazite overgrowth or replacement by other phosphates (Chakhmouradian and Mitchell, 1998), allowing for recognition of primary apatites. Furthermore, a hydrothermal origin for these apatite crystals is inconsistent with the observed Mn and Sr concentrations (Chakhmouradian et al., 2017); see the Supplemental Material¹ for concentrations.

Calcite and apatite were experimentally crystallized from Na-carbonatite melt and equilibrated for 48 h at 1 or 2 kbar and temperatures of 750–1000 °C. The synthetic Na-carbonatite starting material was based on the lava erupted from Oldoinyo Lengai, Tanzania, in 2007 (Weidendorfer et al., 2017) and was mixed from reagent-grade oxide, hydroxide, and carbonate powders. This batch was spiked with a mix of REE oxides and mixed with synthetic calcite and apatite powders in suitable proportions to obtain mineral saturation. Approximately 40 mg aliquots of the homogenized starting materials were filled into 6-mm-long Au₉₀Pd₁₀ capsules (3 mm diameter) and loaded into externally heated rapid-quench cold-seal Hf-C-Mo pressure vessels.

In total, 31 experiments were carried out, 9 of which gave satisfactory results in terms of mineral abundance and size. Among these, 8 samples contained apatites, and 3 samples contained calcites large enough for laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) analysis. We also analyzed calcite trace-element concentrations from three experiments conducted by Weidendorfer et al. (2017), who equilibrated calcite and apatite with natural alkali-carbonatite melt. Major-element analyses were obtained using a JEOL JSM-6390 LA scanning electron microscope and a JEOL JXA-8200 electron probe microanalyzer. Trace-element concentrations were measured by a 193 nm Resonetics ArF excimer laser combined with a Thermo Scientific Element XR LA-ICP-MS. Further details and data are given in the Supplemental Material.

Apatite crystals were idiomorphic and generally 10–50 mm in size (Figs. 1A and 1B); calcite crystals were up to 200 μm and often partly rounded (Fig. 1B).

The experimental partition coefficients were fitted to the lattice strain equation of Blundy and Wood (1994). Figures 2A and 2B show fits to individual experiments, while Figures 2C and 2D show an overall fit of the entire data set based on the temperature (T) dependency of D₀, E, and r₀, where D₀ is the hypothetical strain-free partitioning value, r₀ is the corresponding ideal ionic radius, and E describes the elasticity of the crystal lattice (for details, see the Supplemental Material). Partition coefficients between calcite and apatite were calculated by dividing the modeled values by values and, hence, eliminating the carbonatite melt term:

where N_A is Avogadro's number, R is the gas constant, and r_i is the ionic radius of an element i.

The result was verified by experiment GS4–18, which yielded large apatites and calcites as well as equilibrium carbonatite melt areas free of crystals, allowing for determination of mineralmineral and mineral-melt partition coefficients (striped, gray band in Fig. 3A).

The partition coefficients displayed a flat, slightly U-shaped pattern for La through Lu, and absolute values for (MREE—middle rare earth element) were ~0.15 at 650 °C, increasing to ~0.35 at 1000 °C. Our experimental temperatures ranged from ~100 °C above the Oldoinyo Lengai Na-carbonatite eruption temperature (Keller and Krafft, 1990) to temperatures near the melt inclusion formation temperature estimated by Guzmics et al. (2012) to be between 1050 °C and 1100 °C, consistent with experiments on the intersection of a natural alkaline silicate liquid line of descent with the carbonate melt miscibility gap (Weidendorfer and Asimow, 2022).

The fact that fossil carbonatites do not represent true melt compositions does not imply that they could not be constituted by magmatic minerals. If the carbonatite rocks are magmatic cumulates, the origin and composition of the primary carbonatitic melt can be reconstructed through mineral-melt partition coefficients, an endeavor that may be complicated by a possible melt-composition dependency of said partition coefficients. Our study applies to any carbonatite that crystallizes calcite (>80% of natural carbonatites; Woolley and Kempe, 1989) and apatite. The calcite/apatite partition coefficients neither depend on melt composition nor on additionally crystallizing minerals. However, if more magnesian/ferrous melt compositions lead to high Mg-/Fecalcites or to dolomite/ankerite, the size of the average cation site in the carbonate decreases, which should influence the carbonate/melt partitioning values. Yet, REEs are still expected to substitute for Ca, such that REE-calcite/apatite partitioning patterns (but not absolute values) should shift only slightly.

To test our experimental results, we selected 12 samples from eight alkaline carbonatite complexes (see the Supplemental Material for details). Ten of the samples contain crystals of calcite, and two (MC19-1 and MC19-3) of dolomite. Five samples showed macroscopic and microscopic features indicating a mag-matic origin of the carbonate minerals: comb layers attributed to rapid magmatic crystallization (KS2001), idiomorphic and unaltered nepheline and biotite testifying the absence of pervasive fluid infiltration (X1513, SO199), or interstitial calcite and calcite inclusions in primary and unaltered silicate minerals (B24, B59). Four samples showed clear signs of calcite alteration and/or recrystallization: fluidal (MC19-3) or granuloblastic textures (PLBC1, SI156) and secondary minerals indicating carbonate alteration (e.g., analcime in B58). The remaining three samples (MC19-1, MW2, OKABZ2) lacked any of the above features or other clear signs of alteration; nevertheless, apatite Mn and Sr concentrations were in the igneous field (Chakhmouradian et al., 2017).

A comparison of values of natural samples with the experimental values enabled discrimination between unaltered primary carbonates and those subjected to fluid-mediated secondary recrystallization (Fig. 3B). In samples with magmatic textures, patterns mirrored the experimental patterns. Nevertheless, two samples with Si-rich apatite (KS2001, OKABZ2) showed the same slightly concave-upward pattern, but shifted downward, consistent with a REE enrichment in apatite proportional to Si concentrations. The increase in Si^Ap thus shifted to lower values (Fig. 4B; see also the discussion below). On the other hand, visually altered samples (e.g., MC19-3 and B58) showed a steep REE pattern and a low value (Fig. 4A). Hydrous fluids at submagmatic temperatures (Perry and Gysi, 2018) yield a high solubility of LREEs, leading to steep patterns with low values. For samples where there was no clear textural evidence for recrystallization, the tool indicated primary (MW2 and OKABZ2) and secondary carbonates (MC19-1).

The magmatic origin of carbonatites has long been debated as calcite melting temperatures are excessively high (>1300 °C at 1 kbar; Smyth and Adams, 1923). However, H₂O and alkalis drastically reduce melting temperatures; e.g., in CaO-CO₂-H₂O down to 750 °C (1 kbar; Wyllie and Tuttle, 1960). The Na₂CO₃-CaCO₃ eutectic lies at 700 °C (1 kbar; Cooper et al., 1975), and the chemically complex Oldoinyo Lengai natrocarbonatite melts, dominated by Na₂CO₃-K₂CO₃-CaCO₃-F-Cl-S, exist down to 580 °C (1 kbar; Weidendorfer et al., 2017). For carbonatites formed in alkaline complexes through liquid immiscibility, the miscibility gap crest is near 1150–1200 °C (Martin et al., 2013), while the low-temperature end corresponds to the phonolite minimum near 750 °C (1 kbar), where silicate melt crystallizes (Morse, 1968; Schmidt and Weidendorfer, 2018).

An estimation of crystallization temperatures from our data relative to natural calcite-apatite pairs is only possible for Si con-centrations in apatite comparable to those of experimental apatites (2500–4000 ppm), which cover the range of most natural carbonatites. The partitioning of REEs into apatite is strongly enhanced by high silica activity through the coupled exchange Ca²⁺ + P⁵⁺ = REE³⁺ + Si⁴⁺ (britholite substitution), requiring further experimental calibration of this effect in order to derive crystallization temperatures (e.g., for KS2001 and OKABZ2). Nevertheless, at subvolcanic pressures, Brava nephelinites (B24 and B59) and the carbonatites from Brava (X1513) and Sokli (SO199) indicated crystallization temperatures close to the phonolite minimum. Further complexity arises when apatite crystallizes after calcite, as the evolving melt becomes enriched in REEs (see also the Supplemental Material). However, this does not apply to carbonatites that are immiscible with Brava nephelinites, where apatite and calcite start crystallizing simultaneously between 1050 °C and 1025 °C (Weidendorfer and Asimow, 2022).

The formation of economic REE-rich deposits with >1 wt% REEs may involve multiple processes: an enriched mantle source (e.g., Moore et al., 2015), passive REE enrichment through extreme differentiation of REE-poor magnetite, olivine, clinopyroxene, or calcite (Anenburg et al., 2021), preferential partitioning of REEs into carbonated fluids during immiscibility (Veksler et al., 2012; Martin et al., 2013; Nabyl et al., 2020), and a continuum of carbonate melt to highly alkaline brine-like CO₂-rich fluids (Anenburg et al., 2021).

Nabyl et al. (2020) showed that strong REE enrichment in carbonatite melts at the expense of silicate melts occurs only at temperatures <900 °C, and Martin et al. (2013) indicated that H₂O has a similar effect. Maximum enrichment conditions are hence reached for Ca-poor, highly differentiated phonolitic melts near their minimum temperature of ~750 °C. Our experimental results showed that partition coefficients are below unity only at temperatures <850 °C, with an extrapolated value of ~0.5 at 750 °C. Similar values were obtained by Chebotarev et al. (2019), who reported values of 0.01–0.1 at 650–710 °C. These values are insufficient for extreme REE enrichment; e.g., 95% calcite fractionation at = 0.01 leads only to a 17-fold increase. This increase would also apply to the incompatible H₂O and to any anionic component such as Cl or sulfate. Consequently, with a few weight percent of H₂O in the immiscible carbonatite, this melt would saturate with or transition to a deuteric (brine-like) fluid. The comparatively high fluid solubility of the LREEs (Perry and Gysi, 2018) would then lead to magmatic fluids prone to deposit their REE load in alteration zones within or around intrusive carbonatites.

The experimentally calibrated tool allows researchers to assess processes related to the transition from magmatic crystallization of carbonatites to their fluid saturation and subsequent fluid/rock interaction. Nevertheless, even if carbonatites recrystallize secondarily, this does not necessarily signify that their major constituents, e.g., Ca and C, are completely replaced or even that their isotopic composition has equilibrated with the (crustal) host rock. The next step is then to correlate the magmatic versus secondary nature of the carbonatite minerals with results from Sr, Ca, and C isotopes and elucidate the implications of isotopic crustal values in carbonatites in relation to deep mantle versus secondary processes, where the latter are often related to magma emplacement and solidification processes.

We thank J.M. Allaz, A.C. Calderon, O. Laurent, and P. Tollan for technical support, and A. Giuliani for providing some of the samples. Thanks go to M. Anenburg and three anonymous reviewers for their constructive reviews. This study was financed by Swiss National Science Foundation grant 200020–178948/1.