The availability of ligands with which the REE can complex, and the temperature, pH and redox properties of the fluid, can be ascertained by two different methods:
1. Direct measurements of fluids from, for example:
(a) Fluid inclusions of REE-bearing fluids (b) Solutions from deep-sea hydrothermal vents (c) Solutions from geothermal fields
(d) Encrustations from fumaroles
2. Indirect inferences of fluid chemistry from minerals, for example:
(a) Mineralogy of REE-bearing deposits
(b) Derivation of fluid chemistry from daughter minerals in fluid inclusions from REE deposits.
Direct measurements of REE-ore forming fluids
The most direct method of understanding how HREE ore deposits form is to study the chemistry of fluids which have led to REE deposits. This is, however, challenging because of the lack of modern analogues. The closest available proxy is through analysis of fluid inclusions where a REE-bearing fluid has been trapped prior to forming an ore deposit. Analysis of fluid inclusions from an ore deposit where the REE have already precipitated will only provide information on the conditions following REE transport. Furthermore, the fluid inclusions must be in minerals which do not readily exchange many elements with a fluid and, for crush leach analysis, must be predominantly primary inclusions of one assemblage.
There are only two published comprehensive studies of the REE chemistry of fluid inclusions from a magmatic source: a granite and a carbonatite, respectively.
Granite-derived orthomagmatic fluid inclusions from the Capitain pluton, New Mexico, were analysed by crush-leach analysis by Banks et al. (1994) and Campbell et al. (1995). Fluid inclusion assemblages were hosted in quartz in two samples each of granophyre and aplite.
Granophyre-hosted inclusions represent a fluid which has not precipitated REE minerals, while aplite-hosted inclusions may have formed after crystallization of allanite and/or titanite. Fluids from both samples were shown to be rich in the REE, and LREE enriched relative to chondrite (Fig. 2.10 A). Concentrations of Cl, SO4 and F, were 40 wt. %, 20 wt. % and 500–5,000 ppm, respectively. Other anions analyzed included B, Br and I (Table 2.6). Homogenization
34
temperatures for all the inclusions were approximately 500 ˚C (Campbell et al., 1995). Thus, the results of this study indicate that appreciable REE concentrations, and LREE-enrichment, can be caused by a hot, Cl-, S-rich acidic brine.
Carbonatite-derived fluid inclusions, trapped in country rock quartz, were analyzed from the Kalkfeld carbonatite complex, Namibia (Bühn and Rankin, 1999). These were analyzed by two techniques: in-situ synchotron XRF and a bulk analysis of crushed and cleaned quartz by ICP-MS. Both techniques gave similar results, with very high REE concentrations in the fluid, up to several wt. % (Fig. 2.10 A). As with the fluids from the Capitain Pluton, the Kalkfeld fluids are LREE-rich, and have a slight Eu anomaly. In contrast to the Capitain Pluton, however, the anion content of the fluids is Cl-poor and CO2-rich (Table 2.6). Fluorine concentrations are slightly higher than the Capitain Pluton, between 5000 and 12,000 ppm. Sulphate concentration was not analyzed. Homogenization temperatures of these fluids were greater than 250 ˚C, at which point the inclusions decrepitated. The results of this study suggest that a CO2-, Cl-, and F-bearing carbonatite-derived fluid is capable of transporting high REE concentrations.
Some authors have alluded to, but not fully discussed, analyses of REE in fluid inclusions from other localities. For example, Dostal et al. (2011) mention Y and Ce concentrations of over 10 ppm and up to 10 ppm, respectively, from the Bokan Mountain complex, USA. Interestingly, this difference in concentration probably equates to a HREE-enriched fluid, with salinity (5–12.4 NaCl eq. wt. %) significantly lower than the inclusions from the Capitain pluton.
The dearth of studies on the chemistry of REE fluids, especially the lack of analyses of inclusions from carbonatites and alkaline igneous rocks, means that the analyses from the Capitan Pluton are often used as an example of a REE ore-forming fluid where analyses of fluids from other, not yet analyzed, rocks may be better suited for modelling.
Chapter2. Carbonatites: An overview
35
Table 2.6: Anion concentrations of carbonatite- and granite-derived REE-bearing fluids from the Kalkfeld carbonatite and Capitain Pluton, (Bühn and Rankin, 1999; Campbell et al., 1995).
Concentrations in ppm except where otherwise labelled. CO2 concentration is estimated.
REE-bearing solutions from deep-sea hydrothermal fluids
The concentrations of the REE and other elements, including important anions, have been studied in deep-sea hydrothermal vents (Michard et al., 1983, 1984; Michard and Albarède, 1986;
Michard, 1989; Klinkhammer et al., 1994; Douville et al., 1999, 2002; Allen and Seyfried Jr, 2005; Cole et al., 2014). Clearly, these systems are not perfect analogues for REE ore-forming fluids, but they do provide an insight into REE mobility at higher temperatures.
Hydrothermal fluids in deep sea hydrothermal vents can range in temperature between approximately 150 °C to supercritical fluids over 400 °C. Values for pH are typically low. The predominant anions in the fluid, in terms of concentration, are Cl- and SO42-, of which Cl- is significantly more abundant. The REE concentration in hydrothermal vent fluids is low relative to that of orthomagmatic fluids, but is approximately two orders of magnitude higher than seawater (Fig. 2.10B). Chondrite-normalized REE distributions are typically LREE-enriched, and often have a prominent Eu anomaly. The most likely cause of this distribution is through REE exchange between the fluids during plagioclase recrystallization (Klinkhammer et al., 1994; Bau and Dulski, 1999; Douville et al., 1999). However, preferential mobility by Cl- has also been suggested as causing fractionation due to preferential mobility of the LREE and Eu2+ in Cl--rich brines (Sverjensky, 1984; Bau, 1991; Wood, 2003; Craddock et al., 2010).
36
Figure 2.10: REE concentrations in natural fluids normalised to chondrite. Fluids are: (A) Igneous-derived fluids from the Capitain Pluton (Banks et al., 1994), and from carbonatite-derived inclusions in quartzite (Bühn and Rankin, 1999). (B) compiled Mid-ocean-ridge-fluid, (Klinkhammer et al., 1994; Douville et al., 1999; Bau and Dulski, 1999; Douville et al., 2002;
Schmidt et al., 2010; Craddock et al., 2010); Seawater: (Cole et al., 2014); Fumarole and F -enriched fluid: (Craddock et al., 2010) (C) Acid-sulphate fluids from Waiotapu, near-neutral fluids from Waimangu, New-Zealand (Wood, 2003).
Chapter2. Carbonatites: An overview
37
The concept of anion-controlled REE distribution is further reinforced by the finding of similar LREE-rich, chondrite normalized patterns from fluids sourced from ultramafic rocks with no plagioclase contribution (Douville et al., 2002), from experimental studies of interaction of Cl- -bearing fluid with peridotite (Allen and Seyfried Jr, 2005), and from vents where fluid may have mixed with magmatic volatiles (Cole et al., 2014). In the cases where mixing with magmatic volatiles are suggested as a control on the REE distribution, the chondrite-normalized REE plots deviate from the typical LREE enriched, Eu-depleted pattern (Craddock et al., 2010; Cole et al., 2014). HREE enrichment in these fluids correlates with higher F- concentration in more acidic fluids, which cannot currently be reconciled with the high-temperature thermodynamic data from Williams-Jones et al. (2012). Craddock et al. (2010) also sampled fluids considered to be submarine equivalents of fumaroles. These fluids had relatively flat REE patterns attributed to low pH and high sulphate concentrations, but the stability of the REE did not seem to be greatly influenced by temperature.
REE-bearing solutions from geothermal fields
REE distributions from continental geothermal fluids are much more varied than submarine hydrothermal fluids (Fig. 2.10 C). Geothermal fluids include both crustal and volcanic fluids, studied from a variety of locations, including: Bulgaria, Tibet, (Michard et al., 1987; Michard, 1989), Yellowstone (Lewis et al., 1997, 1998), Western USA (Van Middlesworth and Wood, 1998; Wood and Shannon, 2003; Wood, 2003), New Zealand (Wood, 2003, 2006), Italy, Turkey, Germany and Jordan (Möller et al., 2003, 2004, 2005, 2006). REE distributions of the fluids are variable. Möller (2002) and Möller et al. (2003, 2004, 2005, 2006) argued that the distribution is dependent on the REE distribution of the minerals which dissolved into the fluid at depth. Little evidence of preferential fractionation of the REE into solution is observable when comparing chondrite-normalized patterns of host-rocks and fluids. Broadly speaking, however, Wood (2003) suggested that geothermal fluids can be divided into acid-sulphate fluids (Lewis et al., 1997, 1998; Wood, 2003, 2006) and near-neutral carbonate fluids (Michard and Albarède, 1986;
Johannesson et al., 1996; Van Middlesworth and Wood, 1998; Wood, 2003; Wood and Shannon, 2003). An example of each is shown in Figure 2.10C. Acid-sulphate fluids have much higher REE concentrations than the near-neutral carbonate fluids, suggesting the pH has a strong control over REE mobility. It is possible that ligands have a subtle control over the REE distribution of the fluid, but any effect is masked by the REE distribution of the source minerals. Carbonate-rich fluids from New Zealand might be expected show HREE enrichment based on the high stability of HREE-carbonate complexes, but this is not observed in the data (Wood, 2003, 2006).
38 REE-bearing vapors from Oldoinyo Lengai
Low-density, vapor-like fluids from volcanic vents can be an important ore-fluid carrier (Williams-Jones and Heinrich, 2005). Analyses of these vapors from fumarolic encrustations at Oldoinyo Lengai provide the only example of carbonic fluids from an active carbonatite (Gilbert and Williams-Jones, 2008; Teague et al., 2011). These studies show that the vapor phase is LREE-enriched, and it is likely that higher concentrations of the REE are deposited at hotter fumaroles (Fig. 2.11). Concentrations of the REE can reach similar levels to those seen in orthomagmatic fluids from granites (Banks et al., 1994). Analysis of a gas sample from one of these fumarolic vents shows that it is CO2-dominated, with order of magnitude lower concentrations of HCl and S, and two orders of magnitude lower HF concentration (Fig. 2.11).
These data suggest carbonate complexation may be able to transport the REE in the vapor phase (Gilbert and Williams-Jones, 2008).
Figure 2.11: Chondrite-normalized REE concentrations from fumarolic encrustations from Oldoinyo Lengai and a single bulk-gas analysis from an active fumarole (inset box). Data from Gilbert and Williams-Jones (2008) and Teague et al. (2011).
Chapter2. Carbonatites: An overview
39
Inferences from the mineralogy of hydrothermal REE deposits
The mineralogy of a rock crystallized from a hydrothermal fluid can be used to infer the composition of the fluid. This is especially true for major elements, but the composition must be interpreted with caution. For example, the absence of a mineral does not necessarily mean that certain elements were not present in the system: water-ice is not required to infer the presence of water, and soluble minerals, such as NaCl, may not remain stable. In addition, minerals can crystallise through reaction with a host-rock (e.g. limestone, or equivalent, in the case of a skarn), and thus, in this example, the presence of Ca-bearing minerals does not mean that Ca was present, or indeed absent, in the fluid. Furthermore, the behavior of trace elements in a fluid will be dictated by, not only the availability and stability of transporting ligands, but also the partition coefficients of the mineral into which they substitute. Recent advances have expanded these parameters to multiple elements under particular experimental conditions (van Hinsberg et al., 2010), but experimental work for many fluids and compositions is still lacking.
Despite the difficulties in inferring fluid composition from mineralogy, some of the best evidence of REE speciation in hydrothermal solutions is derived this way. For example, the first discoveries of REE minerals were from skarn environments at the Bastnӓsite-type deposits, Sweden. Here, clear textural evidence indicates that the REE were mobilized into the skarn, and deposited during reaction with the carbonate minerals, although the source of the fluid is still not understood (Holtstam and Andersson, 2007; Holtstam et al., 2014).
The mineralogy of hydrothermal REE ore deposits is quite similar, suggesting a common link between some of the elements in the gangue mineralogy and the REE. REE-bearing and gangue minerals from different hydrothermal deposit types are listed in Table 2.7. Common gangue minerals from hydrothermal REE mineralization include fluorite, barite and calcite, as well as other F-bearing phases such as apatite, F-bearing amphibole and phlogopite. This association indicates that fluoride and sulphate anions are commonly associated with REE deposits while phosphate and carbonate are occasionally associated. The common association of F-bearing minerals with REE deposits means that F is often interpreted to be a major contributor to REE mobility in hydrothermal REE ore deposits.
Careful interpretation of the paragenesis of a REE mineral deposit can also provide information on the mobility of the REE in solution. Paragenetic sequences can be interpreted as evidence for preferential mobilization of the LREE or the HREE. For example, at the Thor Lake intrusion, Canada, Sheard et al. (2012) interpreted that primary eudialyte had been altered, in-situ, to a mixture of zirconium and REE-bearing minerals which were preferentially enriched in the HREE, while the LREE are transported away from the area of alteration. This is interpreted as evidence for greater mobility of the LREE in hydrothermal fluids, consistent with experimental work from the same research group.
40
Table 2.7: Summary of the mineralogy and fluid-inclusion data from hydrothermal REE deposits.
Chapter2. Carbonatites: An overview
41
Common gangue minerals are italicized for emphasis. Abbreviations: L- liquid, V- vapor, S- solid, M- melt; LL denotes an inclusion assemblage with two liquid phases, usually an aqueous and carbonic phase.
42
In contrast, multiple stages of mineralisation at Khibina, Russia, indicate that later-stage carbonatites have higher HREE concentrations than earlier hydrothermal stages, and are HREE mineral-bearing which suggests that, in some carbonatites, hydrothermal conditions can preferentially mobilize the HREE (Zaitsev et al., 1998).
Mineralogy can also be used to infer temperature and pressure ranges, based on theoretical and experimental mineral stabilities. The P-T ranges of some common REE-minerals found in carbonatites has been reviewed by Wall et al. (2004) (Fig. 2.12). This review compiles the work of Williams-Jones and Wood (1992) who first built a petrogenetic grid based on observational and thermodynamic data in the REE (CO3) F–CaCO 3–F2 (CO3)-1–H2O system, adding additional constrains from subsequent experimental work on bastnӓsite and observational constraints for cerianite, florencite, pyrochlore, monazite, allanite, burbankite and ancylite.
Inferences from fluid inclusions
The temperature, salinity and limited compositional data of REE-bearing hydrothermal fluids can be inferred from fluid inclusions. To date, few studies have been carried out on fluid inclusions from hydrothermal REE deposits, and the data are typically restricted to inclusions in fluorite, quartz, barite, and apatite, with limited additional data from REE-phases, such as bastnӓstite.
Analysis of fluid inclusions has been carried out on samples from carbonatites (Samson et al., 1995a, b; Bühn and Rankin, 1999; Costanzo et al., 2006), fluorite deposits related to carbonatites (Palmer and Williams-Jones, 1996; Bühn et al., 2002; Xie et al., 2009), at the carbonatite-related Bayan Obo REE deposit (Smith and Henderson, 2000; Fan et al., 2004), alkaline-rock related REE deposits (Taylor and Pollard, 1996; Salvi and Williams-Jones, 1996, 1997), hydrothermal REE deposts (Williams-Jones et al., 2000; Samson et al., 2004), and at the Bastnӓs-type skarn deposits (Holtstam et al., 2014). A summary of the homogenization temperature, composition and salinity of inclusions from these locations is included in Table 2.7.
Direct evidence for REE mobility is found where REE-bearing daughter minerals are present in fluid inclusions. For example, burbankite has been found as a daughter mineral in fluid inclusions from carbonatites (Bühn et al., 1999), and fluorcarbonate minerals have been interpreted as daughter minerals at Bayan Obo, China; Tamazeght, Morroco; Strange Lake and Thor Lake, Canada (Fan et al., 2004; Salvi et al., 2000; Salvi and Williams-Jones, 1990; Taylor and Pollard, 1996). In these cases the salinity of the fluid ranges from 6 to 29 wt. % NaCl eq. and the homogenization temperatures are 300–360 °C.
The temperature of REE-bearing fluids can be inferred from the homogenization temperature of fluid inclusions. In this case the homogenization temperature provides a constraint on the minimum temperature of the fluid, due to the effects of pressure.
Chapter2. Carbonatites: An overview
43
Figure 2.12: Temperature and pressure ranges of common REE-bearing minerals found in carbonatites. From Wall et al. (2004).
44
The homogenization temperature of inclusions in REE minerals, or in phases paragenetically related to the formation of REE minerals, generally fall between 200–400 °C (Table 2.7). The salinity of inclusions is very variable, with a median of approximately 10 wt. % NaCl eq. but inclusions can range in salinity from near 0 wt.% NaCl eq. to halite-bearing inclusions with up to 60 wt.% NaCl eq. Similarly, there is a wide range of fluid inclusion types, from simple liquid-vapor inclusions to multi-phase and CO2-rich inclusions.
Where inclusion data are available from both REE-bearing minerals and gangue minerals, the gangue minerals typically have fluid inclusions with a lower homogenization temperature than those from the REE-bearing phase. For example, fluorite often contains inclusions, typically simple LV (liquid-vapor) inclusions, with homogenization temperatures between 100–200 °C (e.g. Bayan Obo, Kizilçaören and Bastnӓs; Smith and Henderson, 2000; Gültekin et al., 2003;
Holtstam et al., 2014). A similar observation is noted from carbonatite-derived fluorite deposits such as Okorusu, Namibia, and Amba Dongar, India, where the homogenization temperature of fluorite is typically less than 160 °C (Palmer and Williams-Jones, 1996; Bühn et al., 2002). These observations potentially indicate that F can remain in solution to relatively low temperatures, and after formation of REE minerals.