Petrology and petrogenesis of an intraplate alkaline lamprophyre-phonolite-carbonatite association in the Alpine Dyke Swarm, New Zealand

ABSTRACT

The Alpine Dyke Swarm (ADS), intruding Haast Schist in the Southern Alps, New Zealand, comprises dykes, sills and diatremes of alkaline and ultramafic lamprophyres, phonolites and carbonatites. Intrusion peaked at ∼25 Ma during inception of dextral transtensional displacement on the Alpine Fault plate boundary. In a chamber beneath Haast River, magmas evolved by fractional crystallisation from primitive lamprophyres to phonolites, and then by liquid immiscibility to carbonatite magmas. Carbonatitic magmas coexisted with a highly sodic fluid that metasomatised adjacent quartzofeldspathic schist to aegirine-albite fenites. Carbonatites fractionated from Ca- to Fe-rich and, under late-stage, hydrothermal conditions, to Ba-Sr-REE–rich varieties. Some lamprophyres rose directly from a highly refractory spinel- or possibly garnet-spinel peridotite mantle that had been extensively metasomatised prior to the low-degree partial melting event and Cr-diopside series nodule entrainment. Al-augite series nodules give ages similar to host lamprophyres and are interpreted as deep-seated cognate cumulates. Compared to the broadly coeval basaltic-basanitic magmas of the Dunedin Volcanic Group (DVG) of East Otago, ADS magmas are enriched in volatiles, LILE, and HFSE (including REE). DVG magmas were derived from a less metasomatised mantle source, and, although undergoing extensive fractionation, failed to achieve the extreme alkali enrichment necessary for silicate melt-carbonatite immiscibility.

KEYWORDS:

• New Zealand
• Alpine Dyke Swarm
• alkaline/ultramafic lamprophyres
• phonolites
• carbonatites
• fenites
• mineralogy
• geochemistry
• petrogenesis
• Dunedin Volcanic Group

Introduction

Lamprophyres are an important component of continental intraplate volcanism in the South Island, New Zealand. Cretaceous lamprophyres occur as dyke swarms in the Tapuaenuku Igneous Complex (Baker et al. 1994) and as ring dykes in the Blue Mountain Complex (Grapes 1975) of Marlborough in the Eastern Province. They also occur as a component of the Cretaceous Westland dyke swarms in the Western Province (Wellman and Cooper 1971; Waight et al. 1998; van der Meer et al. 2017). In the Eastern Province, intrusion of alkaline lamprophyres of late Oligocene-early Miocene age occurred in West Otago and South Westland, forming the Alpine Dyke Swarm (ADS, Cooper 1986). This magmatism, described here, initiated at the same time as extensive alkaline volcanism of the Dunedin Volcanic Group in East Otago, summarised by Scott et al. (2020). Although sharing many petrological features, the two suites differ significantly in the H₂O and CO₂ contents of their magmas, reflecting contrasting mantle source compositions.

Field relationships

Dykes, sills and diatremes of predominantly lamprophyre, comprising the ADS, intrude country rock Haast Schist in the Southern Alps between Lake Wanaka and Paringa River (Figure 1). An isolated limburgite dyke (Hutton 1943), occurs a further 48 km to the south at Nevis Bluff, Kawarau River. The terrain is mountainous, so many areas of the swarm have been explored only at reconnaissance level. However, the zone of intrusion extends north-northeast for approximately 110 km from Lake Wanaka to the Paringa River, and is ∼25 km wide (Figure 1). Widespread and more volumetric basaltic activity occurs in East Otago, where the Dunedin Volcanic Group comprises the central Dunedin volcano and outlying vents with associated dykes, flows and pyroclastic deposits (Coombs et al. 2008).

Figure 1. Locality map showing extent of Alpine Dyke Swarm and locations of those intrusions that are logged in the GNS QMAP database. Dykes and sills in black, diatremes in red. Boundaries of the proposed magma chamber underlying the Haast River area are based on the distribution of evolved magmas (phonolites and carbonatites). Red asterisks mark localities of Cr-diopside series, mantle xenolith-bearing lamprophyres, with data from Wallace (1975), Cook (1984), Cox (1984), Brodie (1985), Briggs (2011), and Cooper and Scott (pers. observation) (colour online).

In the ADS, lamprophyre intrusions dominate in the extreme north and south, but in the area between the Blue River and immediately north of the Haast River, dyke lithologies are more varied. Here lamprophyres (both alkaline – AL, and ultramafic – UML, using the classification of Rock 1991) are associated with phonolites and carbonatites (Turner 1932; Mason 1961; Cooper 1971, 1986, 1996; Wellman and Cooper 1971; Blattner and Cooper 1974; Waters 1983; Brodie 1985; Barreiro and Cooper 1987; Brodie and Cooper 1989; Cooper et al. 1987, 1995, 2015, 2016; Paterson 1993; Norrie 2000; Cooper and Paterson 2008; Cooper and Beck 2009; Briggs 2011, 2017; Briggs et al. 2016). Lamprophyres are generally thin, with an average thickness of ∼ 1 m (although with a maximum of 85 m). Dykes of phonolite (originally described as tinguaites by Turner 1932) can be up to 15 m thick, and carbonatites range from a maximum thickness of ∼1.25 m down to thin, vein-like intrusions. In the Haast River area, dykes have a preferred east-west strike (095–100°) and northerly dip (Cooper 1974), and in areas where the country rock foliation is predominantly planar, intrusion has also occurred parallel to the schistosity, producing north-easterly striking (∼030°) sills (Cooper 1974, his Figure 19).

Along the Haast River road, a geophysical survey by Garrick and Hatherton (1973) identified a seismic reflecting surface at a depth of ∼3 km that was restricted to the area of known lamprophyre dyke intrusion. They interpreted this surface to be the top of a lamprophyre magma chamber. This hypothesis is compatible with the wide range of magma types represented by dyke intrusions in the Haast River area (suggesting fractionation of ponded magma in a crustal magma chamber), and the presence of coarse-grained xenoliths of ultramafic rocks, gabbro and syenite in lamprophyre and phonolite dykes derived from this plutonic complex.

To the south, in the Matukituki River and Lake Wanaka area, the dykes strike preferentially east-southeast to southeast (120° and 155°) (Cooper et al. 1987). All dyke orientations will have been modified by rotation in the Southern Alps as a consequence of ongoing dextral shear related to Pacific–Australian plate boundary deformation.

Dykes/diatremes are strongly discordant to country rock foliation with sharp intrusive contacts and chilled, typically microcrystalline margins (Figure 2A). In both the southern part of the swarm (Cooper 1979, 1986; Cook 1984; Cox 1984; Gamble 1984; Turnbull 2000; Maloney 2016; and in unpublished Otago University BSc theses), and in an isolated occurrence in the Moeraki River (Wallace 1975), intrusions can take the form of steep-sided diatremes. Here, rock types include breccias of both country rock schist and lamprophyre, commonly cemented or impregnated by an ankeritic carbonate (Figure 2B). Diatremic breccias are intruded by late-stage coherent lamprophyre. Diatremes are commonly rich sources of mantle-derived Cr-diopside series nodules (Wallace 1975; Brodie 1985; Scott, Hodgkinson, et al. 2014; Scott et al. 2016; Figure 2A). In the southern diatremes, open space fillings within breccias, coupled with intersertal textures in dyke chilled margins, suggest a high level of emplacement.

Figure 2. A, Margin of the Moeraki River diatreme. Dark coloured kaersutite-biotite-olivine UML containing ocelli of carbonates-albite-apatite-pyrite is chilled against isoclinally folded quartzofeldspathic Haast Schist. UML contains coarse-grained Cr-diopside series harzburgite nodules with pronounced talc-magnesite (T-M) alteration rims adjacent to the lamprophyre groundmass. Analysis of UML, OU86381, is reported in Supplementary Table 2. B, Unsorted lamprophyre breccia with a carbonate matrix. Open cavities with euhedral quartz infillings testify to the shallow depth of formation of the exposed Mt Alta diatreme. C, Thin lamprophyre dyke intruding quartzofeldspathic schist, Fish River. Note selvedge of carbonate at dyke contacts (colour online).

In the northern part of the swarm, wall-rocks are commonly carbonated, in places with an ankeritic selvedge (Figure 2C). Intrusion both postdates and predates emplacement of cross-cutting arrays of carbonate veins. The evolved carbonatites have contacts with quartzofeldspathic schist that are bleached, with metasomatic replacement by albite, aegirine and riebeckite–arfvedsonite forming sodic fenites (Cooper 1971, his Figures 4–9; Paterson 1993; Cooper et al. 2016). Although fenitisation is extensive adjacent to carbonatite, it also occurs to a progressively more restricted extent adjacent to phonolites and lamprophyres. Xenoliths of quartzofeldspathic schist in ultramafic lamprophyres are also metasomatically converted to aegirine-riebeckite schists, indicating the widespread presence of an alkaline, Na-rich, fluid phase coexisting with these magmas.

Dykes and sills contain internal bands of variable grain sizes, evidence of multiple intrusion (Figure 3A). AL, and more rarely UML, are characterised by ocellar textures (Figure 2A and 3B), where individual ocelli are defined by spherical or ellipsoidal bodies dominated by the minerals albite, K-feldspar and calcite (Cooper 1979, his Figures 3, 7). Ocelli are commonly concentrated in planar zones parallel to the dyke contact (Figures 2A and 3B) and segregations of felsic material in places form anastomosing networks (Figure 3C). Pyroxene, kaersutite and biotite are minor phases of ocelli, and in places are consistently concentrated at one end of the ocellus, with calcite defining the central or opposite end. This texture has been interpreted as due to gravity settling of mineral phases from residual magma that infilled vesicles (Cooper 1979). In phonolites, ocelli are infilled with sodalite, K-feldspar, ferroan carbonates (ankerite and siderite), cancrinite, Nb-rutile and zircon (Cooper and Beck 2009, their Figure 2).

Figure 3. A, Syenite (S) and hornblendite (H) nodules in lamprophyre cut by thin dykelet (D) of chilled lamprophyre containing vertically aligned infilled vesicles, north bank of Haast River. B, Parallel bands of ocelli and chilled margin(s) in Fish River lamprophyre dyke. C, Lamprophyre with anastomosing array of felsic segregation net-veins, north bank of Haast River (colour online).

Age of the dyke swarm

Early attempts to determine the age of intrusion of the ADS at Haast River by K-Ar techniques encountered problems with both unrealistically young and old dates, and wide discordance (Wellman and Cooper 1971; Adams 1980). This was attributed to argon loss and re-incorporation into K-poor minerals during shearing and uplift associated with the nearby plate boundary. A subsequent K-Ar study of ADS rocks near Wanaka, distant from the plate boundary, indicated a more coherent distribution of ages, with 10 whole rock ages ranging from 25.2 to 31.9 Ma and kaersutites separated from four samples yielding a tighter group at 22.9–27.8 Ma (Adams and Cooper 1996). A previously unpublished age on a kaersutite megacryst from the Lake Wanaka lamprophyre diatreme has given a highly precise ⁴⁰Ar/³⁹Ar plateau age of 23.8 ± 0.1 Ma (J. M. Scott, pers. comm. 2019, Supplementary Table 1).

Rb-Sr isochron ages from a sodalite microsyenite dyke and biotite-kaersutite-rich nodules in ultramafic lamprophyres of 24.1, 20.0, and 22.5 Ma, and U-Pb ages of zircon from phonolite of 24.6 and 25.1 Ma in the Haast River area were reported by Cooper et al. (1987). The structural relationships between the dykes and host schist structures are compatible with intrusion along Riedel shear zones associated with dextral transtensional movement on the Alpine Fault plate boundary in the late Paleogene-early Neogene (Cooper et al. 1987). A very similar genetic relationship between lamprophyre intrusion and initiation of transtensional tectonism was described by Scarrow et al. (2011).

U-Th/Pb analysis of zircon and monazite from syenite nodules, phonolite and carbonatite from the Haast-Burke River area by laser ablation at UCSB has yielded ages between 25 and 23.8 Ma for these evolved ADS differentiates (Briggs 2017). A LA-ICPMS date of 19.7 on thorite from a Haast River carbonatite vein figured in Cooper (1971), and a single analysis on an altered thorite domain from the same sample of ∼17 Ma (Cottle 2014), likely record late-stage carbothermal and alteration processes. The limburgite dyke at Nevis Bluff, Kawarau River has been dated by ⁴⁰Ar/³⁹Ar methods to be 20.7 Ma (Hoernle et al. 2006). The true age of ADS intrusion is clearly Late Oligocene-Early Miocene.

Exotic xenoliths

Many dykes contain nodules up to 40 cm diameter of gabbro, and syenites (Figure 3A), the latter contain zircons whose age is identical to the age of intrusion determined from dykes (Briggs 2017). Peridotite nodules include lherzolite, harzburgite, and dunite, of the Cr-diopside series suite (e.g. Wilshire and Shervais 1975), olivine-clinopyroxene-amphibole ± biotite assemblages are correlated with Al-augite series nodules (Wilshire and Shervais 1975), and apatite-sodian ferrosalite (with up to 20 mol % acmite component)-titanomagnetite-amphibole-titanite assemblages with an adcumulate/poikilitic texture, are correlated with amphibole-apatite series nodules (Wass 1979).

Peridotites are predominantly harzburgitic (Figure 2A) and are interpreted as residual mantle assemblages resulting from depletion by partial melting events. Olivine (Mg# (100*Mg/(Mg + Fe) up to 93)) is the dominant phase with accessory Cr-rich spinel (Cr# (100*Cr/(Cr + Al) up to 80)) (Wallace 1975; Brodie and Cooper 1989; Norrie 2000; Scott, Waight, et al. 2014; Scott et al. 2016). Such compositions are characteristic of highly refractory mantle (e.g. Carswell 1980; Dick and Bullen 1984). In some nodules, chromite forms symplectitic intergrowths with orthopyroxene and subordinate clinopyroxene (Wallace 1975; Brodie 1985; Norrie 2000; Crase 2014; Scott et al. 2016). These symplectites have been interpreted as pseudomorphs of original garnet. Recent isotope studies of peridotites from the Lake Wanaka diatreme have yielded Re depletion Os model ages that range from 0.5 to 2.7 Ga (Liu et al. 2015). This spectrum of ages implies that the underlying lithospheric mantle is grossly heterogeneous in age (with some fragments as old as Archean), that have been assembled beneath Zealandia (Liu et al. 2015; Scott et al. 2019).

Some peridotite nodules have been modally metasomatised with titanian chromian pargasitic hornblende or colourless chromian phlogopite occurring along grain boundaries, or in veins of pargasite ± diopside ± titanian magnetite ± calcite ± apatite that predate entrainment in the lamprophyre (Brodie and Cooper 1989; Norrie 2000; Scott, Waight, et al. 2014). In extreme cases, Cr-diopside and Cr-spinel are highly corroded and surrounded by patches of phlogopite and dolomite (Brodie and Cooper 1989, their Figure 5.3; Norrie 2000). Recent research on peridotite nodules in ADS lamprophyres has shown that metasomatic clinopyroxene is enriched in (La/Yb)_N, and Th/U and depleted in Ti/Eu (Scott, Waight, et al. 2014, 2016). These trends were attributed to mantle metasomatism by both carbonatite and silicate/CO₂-bearing volatile-rich melts. A similar style of cryptic metasomatism has also affected the mantle beneath DVG rocks in East Otago (Scott, Hodgkinson, et al. 2014; Scott, Waight, et al. 2014; McCoy-West et al. 2016; Dalton et al. 2017).

The Al-augite suite of xenoliths are the largest nodules observed, and in the Fish River UML these reach sizes up to 40 × 25 × 20 cm. Nodules are in places banded, so the mineralogy is variable on a mm/cm scale. Olivine, amphibole, clinopyroxene and biotite dominate, with titanomagnetite, ilmenite, pyrite and chalcopyrite locally significant. Textures are commonly poikilitic suggesting a cumulate origin, with early olivine enclosed within oikocrysts of amphibole. The nodule mineralogy is similar to, and individual phases have similar compositions to the essential phases of the host UML. For example, in UML OU 28051, a variety of nodules gave composition ranges of olivines Fo_75–80, amphiboles TiO₂ 2.34–3.80%, Mg# 72.9–76.8, clinopyroxenes Mg# 78.0–80.2, and biotites TiO₂ 3.47–4.34%, Mg# 78.3–83.0, compared to host lamprophyre compositions of olivine Fo₇₉, amphiboles TiO₂ 3.39–4.98%, Mg# 72.9–74.5, clinopyroxenes Mg# 75.7–81.0, and biotites TiO₂ 4.06–5.15%, Mg# 81.3–81.9. This similarity in mineral chemistry to UML and the similar Rb-Sr age to ADS dykes (Cooper et al. 1987) would suggest that the Al-augite suite nodules are cognate cumulates, derived from a slightly more primitive, hydrous, lamprophyric precursor.

Amphibole-apatite nodules have a crystallisation sequence of apatite, titanomagnetite, amphibole, and titanite, resulting in adcumulate/poikilitic textures. They too are interpreted as cumulates or the products of vein crystallisation from metasomatic fluids, or more carbonatitic magmas, at deep crustal or mantle depths (Wass 1979; Brodie and Cooper 1989).

Many dykes also contain megacrysts, in order of decreasing abundance these include kaersutite, clinopyroxene, olivine, biotite and titanomagnetite.

Texture and mineralogy

Ultramafic lamprophyres

UMLs have a porphyritic or seriate texture dominated by combinations of euhedral olivine, clinopyroxene, brown amphibole, biotite and titanomagnetite (TiO₂ up to 17.0 wt %, OU71140; Norrie 2000) with a modally variable, interstitial component that in some dykes is predominantly carbonate, but can be feldspathoids (nepheline, sodalite or analcime) with trace albite (Ab_97–99) (Norrie 2000). Apatite is ubiquitous and perovskite and/or titanite occur in some rocks. Pyrite is generally an accessory phase and alkali feldspar may occur in trace amounts. Olivine (and rarely clinopyroxene), are porphyritic with compositions typically Fo_75–79, ranging to Fo₈₅ (Brodie and Cooper 1989). Extreme compositions of Fo₉₂ probably represent disaggregated phases from peridotite nodules, an origin also attributed to Cr-rich spinel grains overgrown by magnetite. Olivine is commonly enclosed in amphibole or rarely biotite, and is either marginally altered or completely pseudomorphed by combinations of serpentine, talc and carbonate. Pyroxenes are varieties of ferroan diopside (Morimoto 1988) (Mg# phenocryst 0.74–0.81, groundmass 0.79–0.61, OU71128; Norrie 2000). Amphiboles are Ti-rich (3.17–6.91 wt. %) and range from Mg# 0.57–0.77 and are classified predominantly as potassian kaersutites (Leake 1978). In some dykes, grains are overgrown or altered by a blue-green, richteritic to arfvedsonitic amphibole, typically developed adjacent to groundmass-rich carbonate. Micas are titanian phlogopitic biotite to phlogopite with TiO₂ ranging from 2.73 to 6.93 wt. %, and with Mg# 0.62–0.82. Ocellar textures are common, defined by circular or ellipsoidal sections in thin sections infilled by combinations of carbonate, alkali feldspar, analcime, apatite, colourless micas, ilmenite and pyrite. Biotite occurs in some ocelli, generally as isolated flakes, but in places (e.g. OU86379) as pseudohexagonal plates in optical continuity with biotite in the immediate groundmass.

On the basis of their mineralogy, these dykes and diatremes have been previously classified as ouachitites (Cooper 1986; Brodie and Cooper 1989) based on the criteria of containing groundmass carbonate and feldspathoid (Rock 1986). However, a more recent reclassification of the UML (Tappe et al. 2005) recommended that the name ouachitite be discarded, and proposed that aillikite (containing primary groundmass carbonate) or damtjernite (containing nepheline and/or alkali feldspar) are more appropriate. Since ADS UMLs can contain accessory interstitial amounts of both carbonate (Cooper 1971), and sodalite and albite (Brodie 1985), they share characteristics of both rock types and their classification remains ambiguous.

Alkaline lamprophyres

ALs differ from UMLs in containing an essential feldspathic groundmass. They typically have a panidiomorphic or seriate texture defined by euhedral phenocrysts of olivine, Ti-amphibole, clinopyroxene, and Ti-mica enclosed in a groundmass that lacks olivine, but is composed of second generations of the other phases, including Ti-magnetite. As is typical of lamprophyres (Rock 1977), the primary mineralogy has invariably been intensely altered deuterically.

Interstitial plagioclase feldspar, when fresh, ranges from compositions as calcic as An_43.5Ab₅₅Or_1.5 (Wallace 1973), but is commonly albite, An₁Ab₉₈Or₁. Feldspar can evolve in groundmass and ocelli to sanidine, An₁Ab₄₁Or₅₈ (Cooper 1979). Calcite, cancrinite, and sodalite occur in varying proportions. Olivine varies in composition from Fo₈₅ to Fo₇₂, and both normal and reverse zoning has been observed. Ti-rich amphibole, with TiO₂ contents up to 5.99 wt. % generally increasing from core to rim of grains, is typically kaersutite. It commonly shows optical colour zoning with sharp, resorbed zone boundaries. Mg# ranges from 40 to 67, exhibiting both normal and reverse zoning patterns. A blue-green deuterically altered rim is typically sodic (Na₂O 4.39 wt. %) and has a subcalcic Mg-hastingsite composition. In some grains, clinopyroxenes have a corroded, xenocrystic, green core rich in Fe and Na (2.59 wt. % Na₂O) with a typical composition of En₃₁Fs₂₂Wo₄₇. As with grains from the southern part of the swarm described by Cooper (1979), they have complicated sector and concentric zoning patterns. Sector zoning is defined by prism sectors enriched in Al₂O₃ and TiO₂ (up to 8.90 and 4.41 wt. %, respectively) relative to basal sectors (up to 4.24 and 1.16 wt. %, respectively). An Fe-enrichment trend from core to rim in each sector is typical (e.g. Fs₁₃En₃₈Wo₄₉ to Fs₃₆En₂₀Wo₄₄), although reverse zoning does occur across internal concentric zone boundaries. A greenish colour developed in some grain rims suggests that the increase in Fe is accommodated in an aegirine component. In two dykes (OU86376, 28095), orthopyroxene occurs as isolated crystals. In each case, the grains are zoned from Fe-rich cores (e.g. En₄₀Fs₆₀) to more magnesian rims (En₄₅Fs₅₄ to En₆₀Fs₄₁).

Both biotite and spinel are titaniferous, with TiO₂ concentrations up to 6.49 and 19.40 wt. %, respectively. Titano-magnetite contains up to 5.8 wt. % Al₂O₃ and 2.2 wt. % MnO. Some dykes contain titanite.

Interiors of thick intrusions are coarse grained, with gabbro described by Mason (1961). In this gabbro, olivine has a composition of Fo_73.6, plagioclase ranges from An₅₇Ab₄₁Or₂ to An₁₈Ab₈₀Or₂, augite approximates En_44.7Fs_8.5Wo_46.8, biotite has TiO₂ up to 3.95 wt % and Mg# 60, and kaersutite has an Mg# ranging from 64 to 68.

The fine-grained lamprophyres, in which the groundmass is dominated by plagioclase are classical camptonites, while those magmas that have cooled more rapidly to produce a glassy groundmass are monchiquites (Rock 1977).

Phonolites

Phonolites were first described by Turner (1932) as boulders in the Fish River. In situ occurrences in the headwaters of this river were documented by Norrie (2000) and the most southerly occurrences crop out in the headwaters of the Blue River, a few km further south. The most northern occurrences are a few km north of Haast River (Briggs 2011). This distribution is interpreted as outlining the geographic limits of the underlying magma chamber where fractionation of ADS magmas has taken place.

As described by Turner (1932), Mason (1961), Cooper (1971, 1986), and Cooper and Beck (2009), phonolites and coarser microsyenite variants contain phenocrysts of perthitic alkali feldspar (sanidine rims to phenocrysts are An₀Ab₂Or₉₈, groundmass and ocellar compositions range to An₀Ab_99.5Or_0.5), nepheline (Ne_81.5Ks_18.5, commonly partially pseudomorphed by cancrinite, which is an Na-Ca rich variety, containing essential SO₄), clinopyroxene, and green biotite set in a groundmass of sodalite, cancrinite, muscovite, titanite, apatite and zircon (e.g. OU 28076, Figure 4A). Kaersutite (Mg# 56–62, TiO₂ 5.16–5.78 wt. %) is uncommon, and olivine (Fo_77–85) occurs rarely and is surrounded by a reaction rim of biotite suggesting it is xenocrystic. Clinopyroxene in most of the phonolites is typically fine-grained acicular aegirine. In the microsyenites, clinopyroxene is intensely zoned (Figure 4A) and ranges from core compositions of En₃₀Fs₂₄Wo₄₇ with Na₂O 2.56, Al₂O₃ 2.86, and TiO₂ 1.02 wt. % in low Al-Ti sector zones to titan-acmite rims of En₃Fs_84.5Wo_12.5 with Na₂O 12.7, Al₂O₃ 2.35, and TiO₂ 4.11 wt. %.

Figure 4. A, Photomicrograph of zoned clinopyroxene with pale green core and dark green aegirine rim with sodalite (SOD), green biotite (BIOT), alkali feldspar (K-F) and cancrinite (CAN) in microsyenite, OU28076, Haast River. B, Contact between phonolite (PHON, OU86382) and calcite-dolomite carbonatite (CARB, OU86396) intruding carbonated and fenitised Alpine schist (SCH), Burke River. Analyses of these dyke compositions are included in Supplementary Table 2 and their REE compositions are illustrated in Figure 7 (colour online).

In a pyroxene-free phonolite described by Cooper and Beck (2009), the perthitic orthoclase/sanidine-cancrinite assemblage contains elliptical micropegmatitic, ocellus-like patches in which euhedral crystals of an albite-rich feldspar are intergrown with ferroan carbonates comprising manganoan-magnesian siderite and ankerite, zircon, a niobian rutile (containing up to 12.2 wt. % Nb₂O₅) and pyrite on the margin of sodalite segregations (Cooper and Beck 2009, their Figure 2). These micropegmatitic patches are interpreted as the crystalline products of alkaline and halogen-rich residual magma. In phonolites from Haast River, these pegmatitic segregations can reach diameters of 20 cm (e.g. OU85416; Briggs et al. 2016), with individual crystals of cancrinite and sodalite reaching 2 cm diameter, with important albite, carbonate, phyllosilicate, Nb rutile, ilmenite and zircon.

Carbonatites

Carbonatites by definition comprise > 50% carbonate minerals with variable, generally small amounts of silicates (albite and a titanian acmite), oxides (mainly magnetite and rutile, with hematite in some samples) and sulphides (pyrite, galena and sphalerite with rare pyrrhotite and chalcopyrite). These carbonate-rich rocks occur as a minor intrusive component (∼1 volume per cent, Cooper 1986, his Figure 2) in the northern part of the dyke swarm. They occur as selvedges adjacent to lamprophyres (Figure 2C), as intrusions spatially associated with phonolites (e.g. Figure 4B; Cooper and Paterson 2008, their Figure 2B) and as separate concordant sill-like bodies or discordant veins and dykes. In the Burke River example illustrated in Figure 4B (a locality studied in detail by Thomas 2018), the intimate association of phonolite and carbonatite occurs over a distance of at least 60 m, with the carbonatite switching along strike from one side of the phonolite to the other. The carbonatites have produced marked fenitisation against their host schists. Quartzofeldspathic schists are converted distally to albite-aegirine-riebeckite fenites and proximally to albite-aegirine-rutile fenites by significant additions of Na, C, Mn, Fe³⁺, Nb, Ba, Sr and possibly total Fe, Zn and Pb and the removal of Mg, H, Fe²⁺, Cu, K and Rb (Cooper 1986; Paterson 1993; Cooper et al. 2016). In carbonatites from the Burke River area there has been addition of Ca and Li, the latter incorporated into the Li-mica, taeniolite (Paterson 1993). Metabasic hosts (greenschists and amphibolites), are converted to carbonate-albite-muscovite-hematite fenites.

Carbonatites textures are typically allotriomorphic granular, with banding defined by variations in grain size or segregations of carbonate, silicate and sulphide phases. Some carbonatites are inequigranular, with coarse rounded grains either primary phenocrysts or porphyroclasts produced during sub-solidus deformation-induced recrystallisation.

ADS carbonatites are mineralogically diverse, some with four or more coexisting carbonate minerals (Paterson 1993; Cooper and Paterson 2008). Carbonatites range from varieties dominated by calcite, dolomite, ankerite or siderite to those containing essential norsethite (BaMg(CO₃)₂) or strontianite. Others contain the carbonates daqingshanite–(Ce) (3(Ba, Sr)CO₃.Ce(PO₄), (Cooper 1986)) where Ce is the dominant Rare Earth Element (REE), carbocernaite ((Ca, Na)(Sr, Ce, Ba)(CO₃) ₂) and barytocalcite or paralstonite (Cooper and Paterson 2008), ancylite (Sr(Ce,La)(CO₃)₂(OH)·H₂O; Cooper et al. 2015), synchysite (CaCe(CO₃)₂F; Thomas 2018) and other minerals rich in REE that remain unidentified. The most common silicate minerals include albite and a titanian acmite, while baotite (a complex Ba-Ti-Nb chloro-silicate, Cooper 1996) and taeniolite (a lithium-magnesium fluor-mica in fenites, Paterson 1993; Cooper et al. 1995) indicate the involvement of Cl, F and Li respectively. Phosphates include ubiquitous apatite and monazite.

Summary of mineralogy

The sequence UML-AL-phonolite-carbonatite is interpreted as a fractionated series with UML and AL representing primary, and in the case of those bearing Cr-diopside series nodules, primitive magmas. This nodule suite indicates that primitive magmas are derived by melting of spinel-, or conceivably garnet-spinel-bearing depleted harzburgitic peridotite that has undergone metasomatic alteration prior to entrainment of fragments by lamprophyre magma.

Olivine, kaersutite and clinopyroxenes of UML and AL magmas are Mg-rich, and these phases become more Fe-rich and less modally abundant in the phonolites. However, many of the mafic phases are concentrically zoned and exhibit corroded grain cores with many examples of reverse zoning. These grains are most likely the product of mixing and recycling of minerals from more evolved magmas at depth. Certainly the presence of xenocrystic Fe-rich orthopyroxene in several AL magmas suggests that the products of earlier magmatism, probably unrelated to the ADS, are being sampled from the deep crust. Thermobarometric calculations using equations of Putirka (2016) suggest the ADS kaersutitic amphiboles for UML crystallised at conditions of 11.8–8.9 kb at an average temperature (from eight geothermometers) of 1034°. Crystallisation conditions for AL varied from a pressure of 7.5–4.6 kb at ∼1000°C. A single mineral hygrometer of Ridolfi and Renzulli (2012) suggests ADS amphiboles crystallised from magmas containing 4.20–5.98 wt. % H₂O. These conditions are compatible with experimental synthesis by Pilet et al. (2010) who demonstrated kaersutite crystallisation can start at 1130° C and 1.5 GPa in basanitic melts containing 5–6 wt % H₂O.

Phonolites and carbonatites have many mineralogical compositional similarities, including the albitic nature of their feldspars, acmitic clinopyroxenes, the presence of Fe-bearing carbonates and Nb-rich rutile. The two magmas are also commonly closely associated in the field, evidence that is compatible with them representing conjugate immiscible magmas (Cooper 1986; Cooper and Paterson 2008). Whether all carbonatites are derived by this mechanism is unclear. Several of the UML are carbonate-rich, and it has been suggested that carbonatite may also be derived directly from these parent magmas by liquid immiscibility (Rock 1986; Tappe et al. 2006, 2009). In the ADS, many lamprophyres have selvedges (Figure 2C), veins and offshoots composed largely of carbonate. Similar features have been described for lamprophyres elsewhere and interpreted by Currie and Ferguson (1970) as due to boiling off, from the ascending lamprophyre magma, of a low viscosity, volatile-rich precursor fluid that effectively opens up the fracture system facilitating intrusion of the associated magma. This process implies immiscibility between lamprophyre magma and a carbonate-rich fluid.

Geochemistry-major and trace elements

A full compilation of 161 analysed ADS dyke rocks is presented in Supplementary Table 2. Analytical techniques are described in Supplementary File 1.

On a total alkalis versus silica (TAS) diagram (Figure 5), the most primitive magma compositions, petrographically identified as UML, and plotting in the foidite field, are compositionally variable. Some of these intrusions contain mantle peridotite nodules and the magmas are interpreted to be near-primitive compositions. Their variability may reflect, in part, melting of inhomogeneous mantle domains (see section on the role of mantle metasomatism). The fractionation path, or paths, of the ADS magmas that can be inferred from Figure 5 are somewhat diffuse, probably due to both the mixing of magma batches from different sources, and, as inferred petrographically, with mixing of magmas that have undergone different degrees of fractionation.

Figure 5. Total alkalis versus silica diagram (TAS) showing geochemical variation in the ADS. Fields for volatile-poor rocks are taken from Le Bas et al. (1986). Average compositions of AL and UML (taken from Rock 1991) and ADS rock types are represented by larger symbols (colour online).

Ultramafic lamprophyres (ouachitite, aillikite)

ADS UML tend to be lower in SiO₂ (25.3–41.5 wt %), and Al₂O₃ but enriched in MgO, CaO and FeO compared to AL, although there is considerable overlap. The average analysis of ADS UML is similar to the UML of Rock (1991) (Figure 5). A high proportion (56%) of UML have K₂O/Na₂O > 1, which, taken with the greater Ba contents, is reflected in the presence of biotite as an essential mineral. UMLs average 303 ppm Cr and 293 ppm Ni, and many of these dykes carry Cr-diopside series nodules, which is compatible with them representing near primitive magmas derived from the source mantle.

Alkaline lamprophyres (camptonite, monchiquite)

Alkaline lamprophyres of the ADS are typically silica-undersaturated (normative ne up to 20 mol %, rarely lc normative, normative plagioclase An₃₀-An₇₀), high alkali rocks (with Na > K), of hydrated and carbonated basanite and nephelinite compositions (Figure 5). The average AL from the ADS (n = 89) is very similar to the AL composition (n = 854) reported in Rock (1991) (Figure 5). ADS AL are REE-rich (up to 600 ppm) and have a LREE-enriched, HREE-depleted composition, with a range of La_N from ∼540 to ∼100 and a range of HREE from Lu_N 18 to <7 (Figure 6A). In terms of the comparison between AL and UML, the UML tend to have greater LREE enrichment relative to chondrite, but are more depleted in HREE (Figure 6A). With the almost total lack of feldspar in UML and the absence as a phenocryst phase in AL, there are no Eu anomalies (Figure 6A). The relative depletion in HREE of primitive UML may reflect derivation from a mantle source region that is richer in residual garnet than the peridotitic source of some of the AL. The high contents of Ba, Sr, Nb, Zr, and REE in AL and UML indicate that these magmas were generated by small degrees of melting of this mantle source.

Figure 6. A, Chondrite normalised REE distributions of AL and UML. Yellow outline shows limits of AL REE distributions, individual UML ratios are shown as black curves. B, Normalised REE distributions of phonolites. All analyses in Figures 6, 7, 9 and 12 are normalised against chondrite values of Sun and McDonough (1989) (colour online).

There are sufficient analyses from different parts of the dyke swarm to make tentative conclusions about regional geochemical variation in AL compositions. The swarm was arbitrarily subdivided into a southern section (S – Lake Wanaka and the Matukituki River), a middle section (M – Wilkin and Young Valleys), and a northern section closest to the Alpine Fault (N – Fish, Burke, and Haast Rivers, with a smaller population from Moeraki and Paringa Rivers). The average SiO₂ contents of these three sections is very similar, ranging from 41.09 (S) to 40.98 (M) and 42.83 (N) wt. %, but despite this there are significant enrichments in S lamprophyres in TiO₂, Ba, Sr, Zr, Nb, and REE, and more marginal enrichment in K₂O, P₂O₅, and Th. In comparison, the N suite is enriched, on average, in MgO, Cr and Ni.

Phonolite

Phonolites are predominantly intermediate rocks with a range from ∼49 to 60 wt % SiO₂, high alumina (17–21.7 wt % Al₂O₃) and very high combined alkalis (up to 16.9 wt. % Na₂O + K₂O), almost exclusively with Na₂O>>K₂O. Total REE contents range from 99 to 439 ppm, with most analyses showing a steep negative slope between La-Sm, ranging in La_N values from 530 to 85. HREE concentrations are relatively depleted with Lu_N 27–9, with La/Lu ranging from 58 to 299 (La_N/Lu_N 6–32). In many analyses, the MREE segment of the phonolites pattern, extending from Eu to Er, defines a shallow trough, with a small increase towards the HREE. ADS phonolites have either a very small positive or negative anomaly in Eu (normalised Eu/Eu*, defined as Eu_N /((Sm_N + Gd_N)/2), ranging from 0.92–1.15, with an average of 15 analyses of 1.03). This reflects very minor accumulation or fractionation of feldspar (Figure 6B).

One ADS felsic dyke (OU28083) is anomalous, showing increase from La_N to Sm_N, with a decrease through to Lu_N (Figure 7). This felsic dyke is also geochemically anomalous in major elements compared to other felsic intrusions, exhibiting a marginally quartz normative composition. The dyke occurs adjacent to a ferrocarbonatite (OU 28122) that shows an identical REE trend, but at higher total concentrations (Figure 7). This feature is discussed more fully in the carbonatite section.

Figure 7. Chondrite normalised REE compositions of phonolitic/trachytic-carbonatite pairs, dykes having an intimate field association (see Figure 4B) (colour online).

Relative to AL and UML, phonolites are enriched in SiO₂, Al₂O₃, K₂O, Na₂O, Zr, Hf, Rb, Nb, Ta, Zn, Ga, Pb, U, and Th, and depleted in CaO, TiO₂, P₂O₅, total Fe, REE, Cr, Ni, V, Sr, and Cu (Supplementary Table 2, and Figures 8 and 10).

Figure 8. Harker variation diagrams of UML to phonolite compositions plotted against SiO₂: A, MgO, B, Al₂O₃, C, CaO, D, TiO₂, E, P₂O₅, F, Zr (ppm). The variation SiO₂ v Na₂O + K₂O is shown in Figure 5. Symbols as in Figure 5 (colour online).

Figure 9. Chondrite normalised REE distributions of selected carbonatites. Several analyses have been omitted to improve the clarity of the figure. Each individual carbonatite curve has been annotated with the nature of the dominant carbonate species (colour online).

Figure 10. Multi-element variation diagram for average compositions of ADS AL, UML, and phonolite. Rock’s (1991) average AL and UML compositions are plotted for comparison together with an average composition of DVG basalt-basanite-hawaiite-mugearite compiled from data in Price and Chappell (1975), Reay et al. (1991), Price et al. (2003), Coombs et al. (2008), Timm et al. (2010) and McLeod and White (2018). Compositions have been normalised to Primitive Mantle values of Sun and McDonough (1989) (colour online).

Carbonatites

The small modal amount of silicate minerals (predominantly albite and aegirine), contribute a variable SiO₂ content, ranging from 0.6 to 23.8 wt. % to carbonatite compositions. Cross-cutting relationships in the ADS carbonatites have not been observed, so the relative stages of carbonatite evolution can only be inferred. However, in other carbonatite complexes around the world there is a common evolutionary trend from early Ca-dominated sövites to progressively more Mg- and Fe- rich compositions (beforsites and ferrocarbonatites), with Ba, Sr and REE rich-varieties thought to represent the extreme fractionates (e.g Chilwa Island, Garson 1966). Carbonatite compositions range from CaO contents of 42.8 wt. % for the calcite + dolomite/ankerite carbonatite (OU 86396) associated with phonolite (Figures 4B and 7), to 3.71 wt. % in the norsethite carbonatite (OU86398), and 0.54 wt. % in a ferrocarbonatite dominated by siderite (OU58753). The norsethite carbonatite contains extreme concentrations of BaO (25.69 wt. %), and high SrO (2.39%), REE₂O₃ (2.23%), PbO (1.21%) and ZnO (5.95%). Carbonatites show the greatest variability in ΣREE concentrations from 19070 ppm (OU86398; Cooper 1986 and new analysis in Table S2) to 10 ppm (ΟU58753), and in their geochemical trends within the REE spectrum (Figure 9). There is a general pattern of La and total REE enrichment reaching high values in calcite or ferroan calcite-rich varieties (e.g. OU86396 La 1825ppm, ΣREE 6150 ppm; OU 58679 La 1205 ppm, ΣREE 4588 ppm) and low contents in siderite-rich variants (e.g. OU 50436 La 6.1 ppm, ΣREE 127 ppm; OU 58753 La 0.8 ppm, ΣREE 10.0 ppm). Despite the general tendency for REE concentrations to decrease towards the ferrocarbonatites there are many exceptions reflecting the variation of REE–bearing minerals (monazite, ancylite, synchysite, daqinshanite, carbocernaite, a burbankite-like Ba-REE-Na carbonate and fergusonite) and the ease with which these high density phases could fractionate from low viscosity carbonatite magmas.

Several carbonatites (e.g. OU66211, 86396, 86397, 86398, 50439, 58679, 58688) show chondrite-normalised steep negative slopes, with pronounced enrichment in LREE and depletion in HREE with La/Lu ratios up to 6709. These patterns have long been considered as ‘typical’ of carbonatites. Many other ADS carbonatites, however, show a La and Ce depletion with maximum enrichment relative to chondrite occurring for Nd or Sm (e.g. siderite-ankerite carbonatite OU 28122), resulting in a hump-shaped overall pattern (Figure 9). Such LREE depletion with maximum concentrations in the MREE have been described for dolomite–ankerite and siderite ferrocarbonatites at Haast River by Cooper and Paterson (2008) and Cooper et al. (2015) and ascribed to crystallisation during the carbothermal, fluid-saturated, stages of carbonatite magma evolution. The selected REE patterns shown in Figure 9 show that some ferrocarbonatites share this REE characteristic, but that other ankerite carbonatites (e.g. OU66211 Figure 9) exhibit a ‘normal’ LREE enrichment and steep negative slope. These LREE-depleted patterns in both carbonate minerals and in carbonatite whole rocks are being increasingly reported (e.g. Hornig-Kjarsgaard 1998; Midende et al. 2014; Cooper et al. 2015; Moore et al. 2015; Weidendorfer et al. 2016; Broom-Fendley et al. 2017) and are variously interpreted to result from crystallisation and/or fractionation of LREE-rich minerals such as apatite or monazite prior to crystallisation of the LREE-depleted carbonate phase, to precipitation of hydrothermal carbonatites influenced by the preferential mobility of LREE in chloride or fluoride complexes (e.g. Migdisov et al. 2009; Migdisov and Williams-Jones 2014), to hydrothermal leaching of LREE from previously crystallised carbonatite minerals (e.g. Cheng et al. 2018), or to fenitic alteration during the later stages of crystallisation (Weidendorfer et al. 2016).

The carbonatite-silicate pair (OU28122-OU28083, Figure 7) which shows LREE depletion is, based on the otherwise ‘normal’ behaviour of conjugate phonolite REE spectra (Figure 6B), interpreted as an example of subsequent leaching of both rock types by circulation of pervasive late- to post-magmatic hydrothermal fluids.

Incompatible elements

A multi-element variation diagram (Figure 10) illustrates that lamprophyres show relative enrichment in LILE (Cs to U) and HFSE (Nb, Ta, REE with LREE > HREE, and Zr) compared to adjacent elements in the plot, but are relatively depleted in K, Pb, and Hf when normalised to primitive mantle. Compositions are very similar to the AL and UML averages of Rock (1991) and to ocean island basalts (not plotted). However, in comparison, phonolites are depleted in Ba, K, REE, P and Ti, with large positive anomalies in Th-U, Ta-Nb, Pb, and Hf-Zr.

Multi-element plots for carbonatites (not shown) are complicated by the variation in behaviour of the REE, but have major enrichments in Ba-Th, Nb, REE and Zr accompanied by depletions in Rb, K, Hf, and Ti. Pb has a variable behaviour, enriched in some rocks, but not in others.

Stable isotopes

Oxygen and carbon isotope analyses of calcite from the ADS were reported in Cooper and Paterson (2008), including data from Blattner and Cooper (1974), Paterson (1993), White (1998) and new analyses of calcite from UML. Most carbonates from lamprophyres and carbonatites have values of δ¹⁸O ranging from +6.7 to +12.0‰ and δ¹³C −4.4 to −7.1‰ and plot on the low δ¹³C edge of the worldwide carbonatite concentration of Deines and Gold (1973), Deines (1989) (Supplementary Figure 1). There is also a spread towards higher δ¹³C (−2.1‰) and δ¹⁸O (20.8‰) in a trend common to many carbonatite complexs and ascribed to coupled silicate-carbonate fractionation, derivation from sedimentary carbonate, or sedimentary contamination (Deines 1989; Demeny et al. 1998). Given the carbonate-poor nature of the host Haast Schist, processes involving carbonate contamination in the ADS are considered unlikely.

Wellnitz (2017) analysed carbonate from lamprophyre ocelli in the Niger Peak diatreme that yielded ratios of δ¹³C −5.64‰ and −6.06‰, and δ¹⁸O 8.35‰ and 7.03‰. These values lie within the range of other ADS carbonates and suggest that the carbonate in ocelli is of primary carbonatitic origin, rather than resulting from secondary hydrothermal infilling.

Radiogenic isotopes

Sr, Nd and Pb isotope analyses of ADS intrusive rocks have been determined by Barreiro and Cooper (1987) with Hoernle et al. (2006) contributing two analyses, including the Nevis Bluff limburgite and Timm et al. (2010) listing four analyses from the southern part of the swarm. The ¹⁴³Nd/¹⁴⁴Nd ratio is remarkably similar for all UML, AL, phonolites and carbonatites (average 0.5128670 ± 0.000025). Although Sr isotope ratios for all rock types are non-radiogenic (⁸⁷Sr/⁸⁶Sr ranges from 0.70279 to 0.703679, average 0.703148, n = 20), two additional AL analyses, CB56 of Barreiro and Cooper (1987) and the Lake Hawea sample of Timm et al. (2010), show elevated ⁸⁷Sr/⁸⁶Sr ratios (0.70461 and 0.70405, respectively). These are interpreted to reflect some interaction with the more radiogenic host schist. Carbonate-rich ocelli in a Niger Peak diatreme lamprophyre (Wellnitz 2017) also have a slightly higher Sr ratio (average 0.70417) than the ADS average (0.70315), but the Nd isotope ratio (0.512852) is indistinguishable from the overall ADS population. A similar relationship has been observed for Cretaceous lamprophyres on the West Coast of New Zealand (van der Meer et al. 2017).

All ADS magmas have radiogenic Pb, with isotope ratios higher than many OIB. All samples have ²⁰⁶Pb/²⁰⁴Pb >19, with some samples having ratios >20.5. These characteristics have been attributed to a HIMU-like component in the source, a characteristic confirmed with isotope analyses of peridotite nodules from the ADS (Scott, Waight, et al. 2014, 2016). The Nd–Sr–Pb isotope ratios of the ADS magmas were interpreted by Barreiro and Cooper (1987) as being generated by melting of a metasomatic mantle component. Subsequent analyses by Hoernle et al. (2006), Timm et al. (2010), Cooper et al. (2015), Scott, Waight, et al. (2014), Scott et al. (2016), and Thomas (2018) are compatible with this interpretation. The general consensus is that the alkaline intraplate, low-SiO₂ magmas of the lower South Island have been derived from a depleted (garnet-bearing?), spinel-facies lithosphere that has been metasomatised by carbonatite or CO₂-bearing, LREE-rich melts producing amphibole and/or phlogopite. Metasomatism occurred in the late Cretaceous (McCoy-West et al. 2016; van der Meer et al. 2017), with the HIMU-like isotopic compositions of subsequent melts the result of in situ, radiogenic ingrowth from the metasomatically introduced U and Th. Some workers (Hoernle et al. 2006; Timm et al. 2010) prefer an asthenospheric origin for Zealandia’s intraplate magmas, but this is still the subject of debate. Derivation of the metasomatising agents from the asthenosphere is less contentious.

Petrogenesis

The primitive or near-primitive magma compositions in the ADS, recognised by their cargo of mantle nodules, are ultramafic and alkaline lamprophyres. Lamprophyric enrichment in incompatible trace elements but depletion of HREE relative to LREE, the symplectitic intergrowths of spinel and orthopyroxene in highly refractory mantle xenoliths, coupled with the presence therein of mica-amphibole-clinopyroxene veins, suggest that ADS magmas were formed by low degrees of partial melting of a possibly garnet-bearing, depleted mantle that had been subsequently metasomatised and volatile-enriched. There is now a consensus that the metasomatic agent was dominantly carbonatitic (Scott, Waight, et al. 2014) and that the metasomatism was likely Cretaceous in age (McCoy-West et al. 2016; van der Meer et al. 2017).

Variation in the composition of lamprophyre magmas suggest that the mantle domains undergoing melting may have had a variable composition (Scott, Hodgkinson, et al. 2014; Scott, Waight, et al. 2014; Scott et al. 2016; van der Meer et al. 2017), although chemical zoning in lamprophyre minerals caused by resorption suggest mixing of magmas in the deep plumbing system also contributes to the variability.

Liu et al. (2015) and Scott et al. (2019) determined Re-depletion Os model ages from mantle nodules of the ADS, showing that melt extraction events range from Archean to modern. They suggested this variation was due to accretion to Zealandia’s subduction margin of mantle fragments of different ages, with most ages significantly older than the overlying Phanerozoic crust.

With regard to subsequent magma evolution in the ADS, there is a low abundance of rock compositions between UML/AL and phonolite (Figures 5, 8, 11), a scarcity referred to elsewhere as the Daly Gap. In lamprophyre studies by Philpotts and Hodgson (1968), Philpotts (1971) and Ferguson and Currie (1971), the gap has been interpreted as due to liquid immiscibility, but other studies (e.g. Mackenzie and White 1970) appeal to the separation of late-stage liquids of radically different compositions. Cooper (1979) showed that interstitial glasses in monchiquite lamprophyres were of phonolitic composition and that these late-stage liquids segregated as veins (e.g. Figure 3C), or migrated into vesicles producing the ocellar textures so characteristic of lamprophyres. Clague (1978) showed that only a very small degree (∼15%) of additional crystallisation in the middle stages of fractionation, including the incoming of titano-magnetite as a cumulus phase, was needed to produce a radical change from mafic to felsic magma compositions, effectively explaining their bimodal distribution. The variation in REE spectra between lamprophyres and phonolites (Figure 6) requires the fractionation of MREE-rich kaersutitic amphibole (see Ulrych et al. 2018).

On a ‘Hamilton’ diagram, (Na₂O + K₂O)–(SiO₂ + Al₂O₃ + TiO₂)–(CaO + MgO + FeO), the ADS geochemistry defines a trend from UML, through AL to phonolites, compatible with the fractionation of olivine, titanomagnetite, clinopyroxene and kaersutite, typical compositions of which are plotted in Figure 11.

Figure 11. Immiscible phase relations between silicate and carbonate magmas (taken from experimental data and compositions of natural melt inclusions: Kjarsgaard and Peterson 1991, Kjarsgaard 1998, Brooker and Kjarsgaard 2011, Guzmics et al. 2011, Martin et al. 2013) plotted in the ternary system (Na₂O + K₂O)–(SiO₂ + Al₂O₃ + TiO₂)–(CaO + MgO + FeO). The ADS compositional spectrum from UML to AL and phonolites is shown, with fractionation dominated by olivine (Ol), pyroxene (Cpx), titano-magnetite (Mag) and kaersutite (Krs) driving the residual, highly alkaline magma compositions to intersect the immiscibility solvus. A possible conjugate carbonatite magma, coexisting with ADS phonolite, is shown by the tie-line, the slope of which is taken from Brooker and Hamilton (1990). This immiscible carbonatite loses alkalis to a fenitisation fluid resulting in bulk ADS carbonatites plotting close to the CaO + MgO + FeO apex. Symbols as in Figure 5 (colour online).

The close spatial association of phonolites and carbonatites, their similar radiogenic and stable isotopic compositions (Barreiro and Cooper 1987; Cooper and Paterson 2008), coupled with the similarity in composition of minerals common to the two rock types (e.g aegirine-rich clinopyroxenes, Cooper 1986, his Figure 4) led to the interpretion that the two magmas most probably originated by immiscibility. In Figure 11, the most alkali-rich phonolites have compositions that plot along the silicate limb of a silicate melt-carbonatite miscibility gap defined by both experimental work (e.g. Brooker and Hamilton 1990; Lee and Wyllie 1997, 1998) and the compositions of natural melt inclusions in minerals from carbonatite complex rocks (e.g. Guzmics et al. 2011). The miscibility gap is dependent on P-T conditions, CO₂ and H₂O saturation, and parent-rock geochemistry, but with the solvus shown in Figure 11, the conjugate liquid that would coexist with alkali-rich ADS phonolite is an alkali-rich carbonatite. Compositions of ADS carbonatites, like the vast majority of carbonatites elsewhere in the world, cluster along the (SiO₂+Al₂O₃+TiO₂)–(CaO + MgO + FeO) join and are depleted in alkalis. The development of albite-aegirine ± riebeckite/arfvedsonite fenite from the country rock quartzofeldspathic schist along carbonatite contacts clearly demonstrates how alkalis can be removed from the host carbonatite as a metasomatic fenitising fluid.

In Figure 11 there is a spread of carbonatite compositions away from the CaO–MgO–FeO apex. This spread is in part due to the presence of silicate minerals (albite and aegirine in particular) in some samples resulting in an increase in SiO₂+Al₂O₃+TiO₂, but it is predominantly an artefact of fractionation of carbonatite to progressively more Ba- and Sr-rich compositions, components that are not included in Figure 11.

The validity of the proposal that phonolite-carbonatite pairs in the ADS represent immiscible conjugate liquids could also be tested against geochemical constraints determined from experimental petrology. A major caveat is that ADS carbonatites do not represent magma compositions, having lost significant but unquantifiable components to a fenitising fluid. For the phonolite-carbonatite pairs OU86382-86396 and OU 28983-28122, partition coefficients (D = concentration in carbonatite melt/concentration in silicate melt) for La of 31 and 15; Lu 7 and 3.8; Y 12.8 and 5.8; Sr 22.9 and 39.2; Ba 8.2 and 7.7; and Th 2.7 and 7.84 are all >1. Despite the major differences in the REE distributions of the two rock pairs (Figure 7), these elements are consistently enriched in carbonatite. Conversely Rb, Zr, Ga, Hf, Ta, Nb, Si, Ti, Al are preferentially partitioned into phonolite. A very similar enrichment/depletion pattern has been established for experimental conjugate liquids formed by immiscibility from a hydrous silicate melt with high degree of polymerisation (Martin et al. 2013). Significantly, such enrichment in the REE does not occur experimentally during immiscibility under anhydrous conditions.

Alpine Dyke Swarm and Dunedin Volcanic Group, similarities and differences

The Dunedin Volcanic Group (DVG) rocks of East and North Otago comprise a very similar lithological suite to the ADS. Volcanism in East Otago started at approximately the same time as in the ADS, ∼24 Ma (Hoernle et al. 2006), but eruption at the central Dunedin Volcano finished at ∼10 Ma, similar to volcanism in the outlying province (8.9 Ma, Coombs et al. 2008; see also Scott et al. 2020, this volume).

Although there are seemingly identical fractionation series from basanite/foidite to phonolite in both provinces, there are some significant differences:

1. The DVG contains no equivalent of the low-SiO_2, UML compositions (Supplementary Table 2), that occur in the ADS.

2. Carbonate veins and carbonate in fenitised schist in the Port Chalmers Breccia unit of the DVG have primitive stable isotope ratios (δ¹⁸Ο 7.2–8.4‰, δ¹³C −5.8 to −7.2‰) that have carbonatitic affinities (Price et al. 2003). However, carbonatites (s.s.) have not been recorded, and the fenitisation of the schist basement could result from interaction with fluids from alkaline silicate magmas, as is observed adjacent to ADS lamprophyres. In contrast, ADS magmas are volatile-rich, with hydrous silicates and carbonate minerals occurring as essential primary phases throughout the fractionation series. The CO₂-enrichment leads to the formation of carbonatites in the ADS.

3. Basic magmas in the DVG (basanite-basalt) and ADS (both AL and UML) have average compositions of TiO₂ (2.64, 3.35, and 3.20 wt. %), Ba (497, 652, and 854 ppm), Sr (855, 1048, and 1263 ppm), Zr (306, 362, and 360 ppm), Nb (83, 104, and 115 ppm), and REE (286, 364, and 430 ppm) (see also Figure 10). Despite similar SiO₂ and Mg# contents, these significant compositional differences can probably be explained by variation in the source mantle that has been melted (compare Scott, Hodgkinson, et al. 2014; Scott, Waight, et al. 2014).

4. Apart from two lherzolite-bearing mafic phonolites analysed by Reay et al. (1991), other DVG phonolites and nepheline syenites show pronounced negative Eu anomalies (Figure 12), with Eu/Eu* ranging from 0.21 to 0.93, with an average of 0.52 (n = 13). In contrast, ADS phonolites (Figure 6) have a very small anomaly, with an average Eu/Eu* of 1.03 (n = 15). The most alkali-rich DVG rocks have clearly undergone significant feldspar fractionation during their evolution. As a result, the DVG residual magmas failed to achieve the critical degree of alkali enrichment needed to cross the boundary into the field of silicate melt-carbonatite immiscibility shown in Figure 13. In contrast, the ADS phonolites, enriched in alkalis, underwent very late stage immiscibility with the formation of conjugate magmas of phonolite and carbonatite respectively (see Weidendorfer et al. 2016, 2017; Schmidt and Weidendorfer 2018). The subtle difference in alkali enrichment between the two suites may explain why carbonatites have not been recorded in the DVG.

Figure 12. Comparison between the chondrite normalised REE distribution of phonolites from ADS (yellow field, see also Figure 6B) and DVG (black curves). Data sources as in Figure 10 (colour online).

Figure 13. Total alkalis versus silica diagram (TAS) showing comparison between geochemical variation in the ADS and DVG. The most evolved, alkali-rich phonolite compositions of the ADS fractionation trend cross the field boundary representing the maximum extent of silicate melt-carbonatite immiscibility for CO₂-saturated compositions (taken from Schmidt and Weidendorfer 2018). In contrast, the composition of DVG magmas (dashed outline) fails to reach this critical composition (colour online).

Conclusions

The Alpine Dyke Swarm is composed predominantly of lamprophyres, both alkaline and ultramafic, that were generated by partial melting of refractory, but subsequently metasomatised, volatile-rich garnet-bearing spinel peridotite mantle at approximately 25 Ma. Fractionation of olivine, kaersutite, clinopyroxene and titano-magnetite drove the evolved magma to intersect the silicate melt-carbonatite miscibility gap, resulting in generation of conjugate phonolite and alkaline carbonatite magmas coexisting with a fluid phase. Intrusion from a subvolcanic magma chamber beneath Haast River resulted in diffusion of this fluid phase into the country rock Haast Schist forming metasomatic sodic fenites. Carbonatites have a wide range of compositions, fractionating to late-stage ferro- and REE-Ba-Sr-rich varieties. The ADS is enriched in volatiles and incompatible elements compared to the DVG of East Otago. This enrichment suggests derivation of the ADS by melting of an intensely metasomatised mantle source that contains a complex mixture of depleted components, some of which may be as old as Archean (Liu et al. 2015).

Acknowledgements

I thank research students who have contributed to study of the Alpine Dyke Swarm over the years. I am indebted to James Scott, Quinten van der Meer and Dejan Prelivic for helpful comments on the manuscript and to Stephen Read for help in the preparation of figures. Mount Aspiring National Park is thanked for granting permits to collect rock specimens.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Funding

I thank the Otago University Research Committee, the Benson Fund and the Ministry of Economic Development for their financial support.