Garnet Pyroxenite Cumulates from Cretaceous Alkaline Intraplate Magmas Underplate the Zealandia Mantle Lithosphere

Abstract

The elemental and isotopic properties of garnet pyroxenites can yield information on lithospheric mantle composition, thermal state, and evolution. The 34Ma Kakanui Mineral Breccia in New Zealand contains spectacular but little-studied mantle peridotite and pyroxenite xenoliths that yield new insights into the evolution of a portion of the underlying mantle lithosphere of a former Gondwana margin. The moderately depleted and metasomatized spinel peridotites, as judged from spinel and olivine compositions and bulk rock major and platinum group element abundances, give mineral equilibration temperatures <1020°C and are derived from the middle to shallow (~35 to 50 km) lithospheric mantle when projected onto a 70 mW∙m⁻² geotherm. These residues have low Re/Os and Re-depletion ¹⁸⁷Os/¹⁸⁸Os model ages that range from Eocene (0.05 Ga) to Paleoproterozoic (1.9 Ga), consistent with extraction from a lithospheric mantle comprising fragments with complex depletion histories. Although the peridotites have restricted δ¹⁸O (olivine +5.2 to 6.2), evidence for an isotopically heterogeneous mantle column in addition to the ¹⁸⁷Os/¹⁸⁸Os is seen in clinopyroxene ⁸⁷Sr/⁸⁶Sr (0.70244 to 0.70292), εNd (+4.1 to 18.8), ²⁰⁶Pb/²⁰⁴Pb (17.8 to 20.3), and εHf (+10 to +101). Higher metamorphic equilibrium temperatures of the garnet pyroxenites (Fe–Mg exchange of >1150°C) compared to the peridotites indicate their Eocene extraction was from towards the base of this isotopically heterogeneous mantle lithosphere. Pyroxenite bulk compositions point to cumulate origins, and the mineral isotope ratios of ⁸⁷Sr/⁸⁶Sr (0.70282 to 0.70294), εNd (+5.5 to 8.0) and ²⁰⁶Pb/²⁰⁴Pb (18.1 to 19.3) match many of the Zealandia metasomatized mantle peridotite xenoliths as well as the primitive intraplate basalts but not the garnet pyroxenite host magmas. In contrast to many global pyroxenite studies, the garnet pyroxenite ⁸⁷Sr/⁸⁶Sr and δ¹⁸O (+5.2 to 5.8) data provide no evidence for subducted crustal material in the primary magma source region, and Sm–Nd and Lu–Hf isotope data yield mid-Cenozoic ages that are probably related to isotope closure during eruption. An exception is one sample that yields a Lu–Hf isochron age of 111.9 ± 9.1 Ma, which corresponds to the convergence of the Lu–Hf isotope evolution curves of three other samples. Liquids calculated to have been in equilibrium with these cumulates have trace element compositions comparable to primitive alkaline intraplate basalts like those found at the surface of Zealandia. The new data, therefore, indicate that a pulse of intraplate magmatism occurred during or directly after the cessation of long-lived subduction on the former Zealandia Early Cretaceous forearc Gondwana margin, despite any volcanic surface exposure having been long eroded away. The lower lithospheric mantle emplacement of the garnet pyroxenites suggests that the source of the alkaline parent magmas was probably the convecting mantle, which supports conclusions that intraplate magmas in Zealandia have asthenospheric and lithospheric mantle sources.

Kakanui Mineral Breccia, Zealandia, garnet pyroxenite, mantle peridotite, isotopes

INTRODUCTION

The wealth of thermal, chemical, and isotopic data that garnet pyroxenites can provide makes them important constituents for assessing deep lithosphere evolution. These rock types have been interpreted to be deeply trapped mafic melts (e.g. Mason, 1968; White et al., 1972; Sen, 1988), fragments of oceanic crust (Allègre & Turcotte, 1986; Yu et al., 2010), zones of silicate magma-peridotite reaction (e.g. Garrido & Bodinier, 1999; Liu et al., 2005), or cumulates (e.g. Irving, 1974; Pearson et al., 1993; Becker, 1996; Lassiter et al., 2000; Bizimis et al., 2005; Lee et al., 2006; Downes, 2007). Garnet pyroxenite xenoliths and veins in orogenic peridotites have provided evidence for former slab-derived melts stored in the mantle (e.g. Pearson et al., 1993; Becker, 1996; Xu, 2002; Lu et al., 2020) and evidence for ancient components in the source of some ocean island basalts (e.g. Lassiter et al., 2000; Bizimis et al., 2005; Sobolev et al., 2005). Garnet pyroxenite xenoliths have also provided critical evidence for the lithospheric thickening and magmatic processes affecting the roots of arcs (e.g. Lee et al., 2006; Ducea et al., 2021). Garnet pyroxenites, therefore, provide good links to subduction processes. However, their formation cannot be ascribed to a single geological process because any magma traversing the mantle lithosphere could leave behind a cumulate, depending upon the melt-rock ratio.

We investigate a little-studied garnet pyroxenite suite from the intraplate Kakanui Mineral Breccia in the South Island of New Zealand and characterise it's formation. With a wealth of recent information on the composition and evolution of the Zealandia peridotitic mantle lithosphere, there is now a good reference frame with which to place these garnet pyroxenites. We do this through petrography, mineral trace element, isotope (Hf–Sr–Nd–Pb–O) and whole-rock geochemistry of the xenoliths, coupled with calculation of the trace element compositions of liquids that were theoretically in equilibrium. We also present new data and synthesize existing data on the co-erupted peridotite xenoliths to understand the mantle lithosphere within which the garnet pyroxenites occur.

GEOLOGICAL SETTING

The islands of New Zealand represent the emergent ~5% of an undersea 5 Mkm² continent known as Zealandia (Fig. 1a). This continent comprises a variety of Late Cambrian-Mesozoic terranes that represent the Cretaceous back-arc, arc, and fore-arc to the Gondwana subduction margin (Mortimer, 2004) assembled on the outboard continental margin of the East Antarctic Craton (Scott et al., 2021). Long-lived subduction along the Zealandia margin of Gondwana was terminated by Cretaceous collision of the oceanic Hikurangi Plateau, which now resides in the former trench position under the Chatham Rise (Fig. 1a) (Davy et al., 2008; Hoernle et al., 2010; Jacob et al., 2017). Following plateau collision, much of central-southern Zealandia rapidly transitioned to an extensional tectonic regime that extensively thinned the continental lithosphere before the formation of oceanic lithosphere in the Tasman Sea and from Antarctica via oceanic lithosphere formation in the Southern Ocean at 84 Ma (Gaina et al., 1998) (Fig. 1a).

Fig. 1

A. The continent of Zealandia comprises the rifted fore-arc, arc, and back arc of the paleo-Pacific Gondwana margin. Stars indicate location of ultramafic rocks. Figure modified from Scott et al. (2019). B. Distribution of the Waiareka-Deborah Volcanic Field and three mantle xenolith locations (Kakanui, Alma and Round Hill). Figure modified from Scott et al. (2020b).

Throughout the last 100 Ma, intraplate magmas have erupted through the remnants of the Gondwana margin Cretaceous fore-arc and back arc (Mortimer & Scott, 2020). Regional scale examination of mantle rock occurrences has revealed that the central-southern Zealandia lithospheric mantle contains regional-scale fertile and depleted domains (Reay & Sipiera, 1987; McCoy-West et al., 2013; Scott et al., 2014a, 2014b; McCoy-West et al., 2015; Czertowicz et al., 2016; Scott et al., 2016a, 2016b, 2019; Scott, 2020; Delpech et al., 2023). Some peridotites have Re-depletion Os model ages as old as Archean (Liu et al., 2015) but Proterozoic and younger ages are vastly more common (McCoy-West et al., 2013; Liu et al., 2015; Scott et al., 2019; Delpech et al., 2023). While there is no doubt that there is ≥2.0 Ga peridotitic material beneath the Zealandia crust, it remains debated whether this material represents large areas of lithospheric mantle that have been stable for Ga (McCoy-West et al., 2013) or small mantle fragments accumulated from the heterogeneity of the convecting mantle during young continent stabilisation (Liu et al., 2015; Scott et al., 2019; Delpech et al., 2023).

Regardless of when assembly of the Zealandia lithospheric mantle occurred, peridotite xenoliths across the continent show that this mantle has been metasomatically altered (Klemme, 2004; Scott et al., 2014a, 2014b; McCoy-West et al., 2015; Czertowicz et al., 2016; McCoy-West et al., 2016; Scott et al., 2016a, 2016b; Dalton et al., 2017; Li et al., 2018; Bonnington et al., 2023; Delpech et al., 2023; Cooper et al., 2024). This likely occurred in the Mesozoic and mainly imprinted a FOZO or HIMU-like isotopic signature (Scott et al., 2014a, 2014b; McCoy-West et al., 2016; Wang et al., 2016; Scott et al., 2016a; Dalton et al., 2017; van der Meer et al., 2017; Serre et al., 2020).

Kakanui mineral breccia

Located on a shoreline platform at the small coastal Kakanui township (Lat: −45.194, Long: 170.901) in the South Island of New Zealand, the Kakanui Mineral Breccia represents a small intraplate eruption within the Eocene 34 Ma Waiareka-Deborah Volcanic Formation (Fig. 1b) (Thomson, 1907; Dickey, 1968a, 1968b; Coombs et al., 1986; Reay & Sipiera, 1987; Zack et al., 1997; Hoernle et al., 2006; Corcoran & Moore, 2008; Moorhouse et al., 2015; Scott et al., 2020b). It contains an exceptional cargo of mantle-derived ultramafic xenoliths (Fig. 2a, b) (Mason, 1966; Dickey, 1968a; Philpotts et al., 1972; White et al., 1972; Reay & Sipiera, 1987; Scott et al., 2020b), as well as xenocrysts of kaersutite, apatite, phlogopite, ilmenite, augite, anorthoclase, and pyrope (Mason, 1966, 1968; Mason & Allen, 1973; Reay & Wood, 1974; Reay et al., 1989, 1993; Fournelle & Scott, 2017; Burgin et al., 2023) and crustal granulites (Jacob et al., 2017). The ultramafic xenoliths—which form the topic of the present study—comprise spinel-bearing peridotite and pyroxenite, some of the latter of which are garnet-bearing. The garnet pyroxenites were referred to as ‘eclogite’ in early studies; however, the clinopyroxene is generally not omphacitic and garnet pyroxenite is typically a more appropriate classification.

Fig. 2

A. Outcrop of the central tuff breccia facies forming the Kakanui Mineral Breccia. B. Peridotite xenoliths are enclosed in a thin anulus of nephelinite in a tuffaceous volcanic matrix that has a zeolite cement.

METHODS

Mineral major element analyses at the University of Otago were obtained using a Zeiss Sigma field emission gun scanning electron microscope equipped with an energy dispersive spectrometer operated using an accelerating voltage of 15 kV and a 2.7 nA beam current, with a live time of approximately 30 s. Mineral analyses at the University of Cape Town were gathered using a JEOL Superprobe JXA-8100 electron probe micro-analyzer equipped with four wavelength dispersive spectrometers. The electron beam was set to an accelerating voltage of 15 kV, a beam current of 20 nA, and operated with a spot diameter of 1 to 3 μm. Measurements were performed with counting times of 10 seconds on peak and 5 seconds on background positions. Mineral analyses measured at Shimane University, Japan, were obtained on a JEOL 8530F field emission microprobe using a 15 kV accelerating voltage, a 20 nA current and a 1 μm beam diameter with counting times of 20 seconds on element peaks and 10 seconds at background positions. In each case, Smithsonian Microbeam reference materials were used to standardize the results and confirm the analytical routines.

Mineral trace element concentrations were obtained at the Centre for Trace Element Analysis at the University of Otago using a RESOlution M-50193 nm ArF excimer laser ablation system coupled with an Agilent 7900 quadrupole inductively coupled plasma mass spectrometer. Operating conditions were 2.5 J cm⁻² fluence, 5–10 Hz, and with a beam diameter of 75 micrometers. The mass peaks were collected in time-resolved mode with one point per peak and an integration time of 10 ms per element. Raw mass peak count rates were background subtracted, corrected for mass bias drift, and converted to concentrations using an offline spreadsheet. NIST 610 and 612 glasses (Pearce et al., 1997) were analyzed before and after each set of ~10–20 unknown analyses. The trace element concentrations were then obtained by normalizing to count rates of Ca using NIST 610 glass and CaO values obtained by the scanning electron microscope or microprobe, with 612 as a monitor for precision and accuracy.

Samples selected for whole-rock analysis had their altered surfaces removed with a diamond-coated rock saw. Trimmed samples were then sanded and washed, and then crushed to a fine powder using an agate rock mill. Whole rock major and trace element geochemistry was obtained by ALS (geochemistry laboratories) in Brisbane, Australia. Here, a lithium borate flux was added to the powders, followed by fusion at 1000°C and then digestion in a mixture of dilute HNO₃ and HCl. Measurements of major elements were by inductively coupled plasma optical emission spectrometry. For trace elements, the sample powders were fused at 1025°C, then digested in a HF-HNO₃-HCl mixture, and measured by inductively coupled plasma mass spectrometry using a Perkin–Elmer Elan 9000 system. In-house standards were reproduced within error. Loss on ignition was calculated after heating a powdered fraction to 1000°C.

Strontium, neodymium and lead isotopes were measured by multi-collector inductively coupled plasma mass spectrometry using a Nu Instruments NuPlasma HR at the University of Cape Town, with Nd and Pb analyses also using a DSN-1000 desolvating nebulizer. Strontium, Nd and Pb elemental separation was conducted after the method of Pin et al. (2014). Strontium isotope data were corrected for Rb interference using the measured signal of ⁸⁵Rb and the natural ⁸⁵Rb/⁸⁷Rb ratio. Strontium isotope data were corrected for instrumental mass fractionation using the exponential law and ⁸⁶Sr/⁸⁸Sr value of 0.1194. Data were normalized to an ⁸⁷Sr/⁸⁶Sr value of 0.710255 for bracketing analyses of NIST SRM987. For Nd isotope analyses, JNdi-1 was measured as reference standard and all data are reported normalized to a ¹⁴³Nd/¹⁴⁴Nd value of 0.512115 for this standard (Tanaka et al., 2000). All Nd data were corrected for Sm and Ce interferences using the measured ¹⁴⁷Sm and ¹⁴⁰Ce signals and natural Sm and Ce isotope abundances. Data were corrected for instrumental mass fractionation using the exponential law and ¹⁴⁶Nd/¹⁴⁴Nd value of 0.7219. For Pb isotope measurements, a NIST SRM997 Tl solution was added to all standards and samples to give a ≈ 10:1 Pb/Tl ratio. All Pb isotope data were corrected for Hg interference using on-peak background measurements, as well as for instrumental mass fractionation using the exponential law and a ²⁰⁵Tl/²⁰³Tl value of 2.3889. Data were normalized to NIST SRM981 values for ²⁰⁸Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb, ²⁰⁶Pb/²⁰⁴Pb of 36.7219, 15.4963, 16.9405, respectively (Galer & Abouchami, 1998). BHVO-2 was run simultaneously with samples and yielded average values (n = 3) for ⁸⁶Sr/⁸⁸Sr = 0.703471 ± 0.000011, ¹⁴³Nd/¹⁴⁴Nd = 0.512980 ± 0.000025, and ²⁰⁸Pb/²⁰⁴Pb = 38.1893 ±0.0032, ²⁰⁷Pb/²⁰⁴Pb = 15.5296 ± 0.0011, ²⁰⁶Pb/²⁰⁴Pb = 18.6059 ± 0.0011, which agree with published values (Weis et al., 2006).

Lutetium-hafnium isotopes were measured at the University of British Columbia (UBC; sample OU 25262), the Universität zu Köln (OU 25261, OU 49320, OU 49324, and OU 49325), and the University of Cape Town (OU 20225). Garnet and clinopyroxene at the University of British Columbia and the Universität zu Köln were first washed with de-ionized water and then rinsed in 1 N HCl at room temperature for 1 h. Each aliquot then had a mixed ¹⁷⁶Lu/¹⁸⁰Hf spike added and were then progressively dissolved using sequential addition of HF:HNO₃, HNO_3, and HCl (Köln) or HF:HNO₃–HClO₄ (UBC) interspersed with full sample dry-down steps. Lutetium and Hf fractions were isolated using an Eichrom Ln Spec ion cation exchange chemistry following Münker et al. (2001). Isotope measurements at UBC were measured using Nu Plasma multi-collector inductively coupled plasma mass spectrometer and at Köln with a Neptune multi-collector inductively coupled plasma mass spectrometer. ¹⁷⁶Lu/¹⁷⁵Lu was determined using an ¹⁷⁶Yb interference correction based on a linear correlation between ln(¹⁷⁶Yb/¹⁷¹Yb)/ln(¹⁷⁴Yb/¹⁷¹Yb) determined by replicate Yb standard measurements. Mass bias correction for Hf assumed the exponential law and a ¹⁷⁹Hf/¹⁷⁷Hf = 0.7325. Hafnium isotopes are reported relative to JMC-475 (¹⁷⁶Hf/¹⁷⁷Hf = 0.282160; Blichert-Toft et al., 1997), which has a long-term external reproducibility of 0.4 εHf units. External reproducibility at 2 standard deviations of the mean was estimated from the external reproducibility for JMC-475 measured at concentrations that bracketed those of the unknowns (10–40 ppb; Bizzarro et al., 2003). The isochron and age calculation was calculated using Isoplot version 3.27 and applying a value of 1.867 × 10¹¹ yr⁻¹ for λ¹⁷⁶Lu (Scherer et al., 2001; Söderlund et al., 2004). At the University of Cape Town, Hf was purified using the anion–cation separation procedure in a HCl–HF medium adapted from Blichert-Toft et al. (1997). Lutetium was separated from the other rare earth elements using Eichrom Ln-spec resin in an HCl medium. The isotopes were measured using static multi-collection on a Nu Plasma MC-ICP-MS instrument at the University of Cape Town. JMC475 was measured prior to all unknowns and after every second unknown. All uncertainties are reported at the two standard deviation levels.

Whole rock Os isotope ratios and platinum group elements + Re concentrations were obtained at the University of Alberta following the approach of Pearson & Woodland (2000), modified for digestion in a high pressure asher. Approximately 1 g of whole-rock powder and a ¹⁸⁵Re, ¹⁹⁰Os, ¹⁹¹Ir, ¹⁹⁴Pt, and ¹⁰⁶Pd mixed spike was dissolved by inverse aqua regia in a high-pressure (130 bar) asher (HPA-S, Anton Paar) for ∼16 hours at ∼260°C. Osmium was extracted from the solution by solvent extraction into CHCl₃, back-extracted into HBr, and then purified via micro-distillation using a H₂SO₄-dichromate solution. Osmium isotopes were measured as OsO⁻₃ using multiple Faraday collectors and amplifiers equipped with 10¹² W resistors using a Thermo Scientific Triton Plus thermal ionization mass spectrometer. Typical intensities of mass 240 (mainly ¹⁹²Os¹⁶O⁻₃) were 0.1 to 0.4 V with 50 ratios measured using 8 s integrations per ratio. Raw ratios were first corrected for gain and baseline, followed by oxygen correction using ¹⁷O/¹⁶O = 0.0003749 and ¹⁸O/¹⁶O = 0.0020439 and spike correction, and then instrumental mass fractionation assuming ¹⁹²Os/¹⁸⁸Os = 3.083 via the exponential law. ¹⁸⁷Os/¹⁸⁸Os internal precision was typically better than 0.05% (2σ). Repeated measurements of Os reference material DROsS yielded ¹⁸⁷Os/¹⁸⁸Os = 0.160927 ± 0.000173 (2σ, n = 14) and University of Maryland Johnson Matthey standard yielded ¹⁸⁷Os/¹⁸⁸Os = 0.113796 ± 0.000022 (2σ, n = 10). Highly siderophile elements (Ir, Pd, Pt, Os, Re, Ru) were separated from the acid solution using cation exchange column chromatography and BPHA solvent extraction. The solution was then evaporated to dryness and the residues dissolved in 5 ml of 6 M HCl. One ml of 30% H₂O₂ was added to reduce Cr^VI to Cr^III, and the solution dried down and then dissolved in 10 ml of 0.1 M HCl and loaded onto the cation exchange columns. The dried down residues were dissolved in 2 ml of 0.5 M HCl and 300 μl of fresh 0.025 M BPHA CHCl₃ was added. After three solvent extraction runs, 2 ml of 0.5 M HCl was pipetted. All analyses were measured in low-resolution mode on a Nu Instruments Attom. For Pd, a zirconium (⁹⁰Zr) standard solution was measured to determine the oxide production rate (ZrO+/Zr+) and this was found to be <1% during the analytical measurements. Isobaric interference correction of ⁹⁰ZrO+ was <0.1%. Correction for instrumental mass fractionation was achieved by measurements of in-house elemental reference solutions and assumption of natural isotopic abundances.

Minerals were analyzed for O isotopes at the University of Cape Town, South Africa, and GNS Science, New Zealand. Those undertaken at the University of Cape Town were obtained by laser fluorination following the method described by Harris & Vogeli (2010), except that ClF₃ was used as the fluorinating agent. Monastery garnet (MONGT, +5.38; Harris et al., 2000), as recalibrated by Harris & Vogeli (2010), was run in duplicate as an in-house standard. The variation in MONGT duplicates during this work was 0.14, which gives a standard deviation of 0.09‰ (n = 19 pairs). Yields for the MONGT standard were calculated as 13.92 μM O₂ per mg. Isotope ratio measurements of O₂ were made off-line using a ThermoFinnigan Delta XP mass spectrometer run in dual-inlet mode housed at the University of Cape Town Department of Archaeology. The analyses conducted in New Zealand were undertaken at the GNS Science National Isotope Centre, where oxygen was extracted from minerals by heating with a CO₂ laser in an atmosphere of BrF₅. Oxygen was converted to CO₂ in a graphite furnace, yield measured, and isotopic ratios determined on a GVI IsoPrime mass spectrometer. Measured ¹⁸O/¹⁶O ratios of samples were normalized to the accepted value (5.8‰, Valley et al., 1995) of the UWG-2 garnet. The 1σ external precision, based on 10 replicate analyses of UWG-2, was 0.08‰. The oxygen isotope results are reported in the standard delta notation (δ, where δ = (R_sample/R_standard − 1) × 1000) in per mil (‰) compared to Vienna Standard Mean Ocean Water.

RESULTS

Garnet pyroxenite mineralogy and mineral and bulk rock chemistry

The garnet pyroxenites, which reach 20 cm in diameter but are more commonly between 1 and 5 cm, are commonly dark green due to the abundance of clinopyroxene (Fig. 3a). The garnet grains are red, fine-grained, and similar in grainsize to matrix clinopyroxene (Fig. 3b, c). An exception is OU 25652 in which grains can be over 1 cm in diameter and are significantly coarser than matrix clinopyroxene and amphibole (Fig. 3d). The garnet grains are typically anhedral and sometimes have inclusions or embayments of clinopyroxene and/or amphibole. Counts of >1500 points per sample indicate that garnet ranges from 2 to 50 modal % (Table 1). Clinopyroxene in these samples constitutes between 41% and 80%, comparable to the 18% to 95% range determined by Mason (1968). Orthopyroxene occurs as rare (<0.5%) grains in OU 49320 and OU 49325, or as exsolution lamellae in some clinopyroxene grains. The main Ti-bearing phase is ilmenite, although it is typically a minor component and is absent in OU 49320 and OU 49324, which instead have hercynitic spinel. Amphibole, from 0% to 18%, is distinctive because of its strong dark brown to light brown pleochroism. Its paragenesis is complicated; in some samples, it appears to partially replace clinopyroxene, whereas in others, it occurs as large optically continuous grains (Fig. 3d). Clay (?) alteration in surrounds some garnet grains, and patches of calcite and/or zeolite occur in some samples. Although Mason (1968) reported some composite xenoliths comprising peridotite and garnet pyroxenite, we have not found any such specimens in our investigation.

Fig. 3

A. Typical appearance of a garnet pyroxenite xenolith in outcrop. B–D. Thin sections viewed in plane polarised light. E. Garnet Mg# versus garnet Ca content is a means of distinguishing the different pyroxenites.

Table 1

Summary of Kakanui garnet pyroxenite xenolith data

	Modal %							Bulk rock data																		Geothermometry
	cpx	grt	ilm	spl	opx	amp	SiO2	TiO2	Al2O3	FeO	MnO	MgO	CaO	Na2O	K2O	LOI		Mg#	Cr	Os	Ir	Ru	Pt	Pd	Re	ToC	ToC
							wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %		μg/g	ng/g	ng/g	ng/g	ng/g	ng/g	ng/g	Sun & Liang	Sudholz et al.
KGP1	62.1	18.0	19.5			0.5	No bulk rock data measured										101.99	50								-	1209
recalculated							37.6	11.1	10.0	19.9	0.1	9.9	11.7	1.6	0.0	0.01
OU 20225	40.6	50.1	0.0			9.3	42.07	0.86	17.21	13.86	0.23	12.79	8.49	1.25	0.52	1.00	98.29	65	228							871	1168
recalculated							44.48	1.07	16.61	12.74	0.23	14.09	9.91	1.37	0.12	0.18	100.80	69
OU 25651	43.0	39.4	0.1			17.4	43.55	2.04	14.49	13.32	0.16	11.00	11.10	1.75	0.84	1.36	99.61	62	40	0.0003	0.004	0.014	0.41	0.139	0.30	892	1273
recalculated							44.17	1.81	15.54	13.96	0.24	12.60	11.16	1.38	0.30	0.29	101.43	64
OU 25652	Too coarse to point count						No bulk rock data measured																			881	1160
OU 49320	79.8	12.8	0.0	7.3	<0.1		42.79	1.18	15.26	10.34	0.13	14.45	13.40	1.44	0.06	0.86	99.91	73	40	<LOD	<LOD	<LOD	0.24	0.017	2.82	956	1303
recalculated							44.76	1.29	15.37	10.12	0.13	14.57	14.07	1.52	0.00	0.00	101.84	74
OU 49324	80.2	19.6		0.1			44.23	1.06	15.61	10.84	0.17	14.33	11.84	1.15	0.05		99.28	72	40	<LOD	<LOD	<LOD	0.19	<LOD	3.92	-	1309
recalculated							46.61	1.59	12.95	9.55	0.08	14.16	14.04	1.68	0.05	0.10	100.81	75
OU 49325	71.1	2.2	0.8		0.2	25.8	47.23	1.56	9.41	13.66	0.20	11.95	12.65	2.52	0.36	0.86	100.40	63	495	0.0002	<LOD	<LOD	0.25	<LOD	0.07	1025	1230
recalculated							47.51	2.42	9.64	12.52	0.17	12.15	13.22	2.73	0.28	0.56	101.20	66
OU 53629							42.58	4.36	10.77	15.20	0.18	11.77	12.31	1.85	0.31		99.33	61	103
OU 53631							44.50	1.94	13.71	12.93	0.18	12.79	10.94	1.84	0.49		99.32	66	368
OU 53635							42.18	1.66	14.24	14.10	0.20	12.10	11.09	1.72	0.31		97.60	63	64
OU 53644							46.35	1.24	9.49	12.00	0.20	12.14	13.43	2.43	0.35		97.63	67	390
EM1							44.08	1.85	13.37	12.36	0.19	12.49	12.16	1.53	0.29		98.32	67	80
K14	18	18	2			62	41.20	3.20	16.35	15.54	0.20	10.93	9.61	2.33	0.86	0.58	100.80	58
K15	68	25	1			6	46.11	1.97	12.56	12.81	0.17	12.83	12.16	0.94	0.16	0.59	100.30	66

^*Bold italic font are bulk compositions calculated from mineral chemistries, densities and modes.

^**Data for OU 53629, 53631, 53635, 53644 and EM-1 are from Maidens (1992), and data for K14 and K15 are from Mason (1968).

Mineral major element compositions within samples show little internal variation (Supplementary 1). All garnets are pyrope-rich, with Mg# (=100*Mg/(Mg + Fe)) varying from 49 to 69 (Fig. 3e). TiO₂ reaches 0.47 wt % but Cr₂O₃ is below detection limit. Clinopyroxene is augite and is unzoned apart from some rims that were in contact with either the host magma or with hydrothermal veins formed after the rocks were deposited. Orthopyroxene is enstatite. Following the classification scheme of Hawthorne et al. (2012), the amphiboles are kaersutite. As with the major element mineral chemistry, the garnet, clinopyroxene and amphibole grains have similar trace element compositions between samples and little internal variation (Supplementary 1; Fig. 4a–g), except for elements present in low concentrations such as large ion lithophile and high field strength elements. Garnet has the highest HREE content, and clinopyroxene and amphibole have very similar REE concentrations. The REE partitioning between garnet/clinopyroxene, or clinopyroxene/amphibole is consistent between samples (Fig. 4h).

Fig. 4

A–G. Normalised trace element patterns of phases in garnet pyroxenites, and measured and calculated normalised bulk compositions. Sample OU 25652 was too coarse-grained to meaningfully point count. H. The samples also show large similarities in the partitioning of REE between garnet/clinopyroxene and amphibole/clinopyroxene. Primitive mantle is from Sun & McDonough (1989).

Since some of the garnet pyroxenites contain carbonate and zeolite, bulk rock compositions were calculated from measured modal proportions, mineral densities and mineral compositions and then compared with measured values. Although the measured versus calculated bulk rock major element compositions are generally in broad agreement (Fig. 5a), there is commonly a small difference between elements such as Rb, Sr, K, Nb, U, and Pb. Therefore, the recalculated values are used here and in figures instead of the measured values. Garnet pyroxenite whole rock Mg# ranges from 50 to 73 (Table 1). There is a large variation in SiO₂ (37.6 to 47.5 wt %) and MgO (9.9 to 14.6 wt %) (Fig. 5c), as well as in TiO₂ (0.9 to 4.4 wt %, but one outlier at 11.1 wt %), Al₂O₃ (9.5 to 16.6 wt %), Fe₂O₃ (9.6 to 19.9 wt %), and CaO (9.6 to 14.1 wt %). The large major element dispersion of the suite is, however, mainly due to KGP1, which is ilmenite-rich and so has high Fe and Ti. Bulk rock highly siderophile element (Os, Ir, and Ru) concentrations in three of the four samples were below detection limit, with the concentrations in the one sample that does contain measurable levels (OU 25651) being extremely low (Table 1). Platinum is <0.41 ng/g, Pd was detected in only two samples, and Re varies between 0.07 and 3.92 ng/g.

Fig. 5

A. A comparison of measured versus calculated major element compositions in the Kakanui garnet pyroxenites. B. MgO versus SiO₂ of the garnet pyroxenites compared to local mafic magmas and global garnet pyroxenites. C. Total alkalis versus silica diagram (Le Bas, 1986) showing the range of compositions. D. Although the Cr concentrations of garnet pyroxenite suites vary widely, the Kakanui samples all tend to have low concentrations. Data in B, C, and D from Pearson et al. (1993) for Beni Bousera, Lee et al. (2006) for the Sierra Nevada xenoliths, Griffin et al. (1988) for SE Australia xenoliths, Hoernle et al. (2010) for the Hikurangi Plateau, Mckenzie & O’Nions et al. (1995) for N-MORB and E-MORB, this study for peridotite xenoliths, Scott et al. (2020a) for Dunedin Volcanic Group and Scott et al. (2020b) for the Waiareka-Deborah Volcanic Field.

Peridotite xenolith mineralogy, mineral and bulk rock geochemistry

The peridotite xenoliths in the Kakanui Mineral Breccia are typically between about 2 and 10 cm in diameter but, like the pyroxenites, can reach ~20 cm (Fig. 2a, b). They are significantly more common in outcrop than pyroxenite but commonly have calcium carbonate, talc, and zeolite replacing mantle minerals. This makes the original compositions of some of the samples difficult to determine reliably using the relative proportions of mineral phases; we therefore refer to the suite as ‘peridotite’ but recognize that some samples classify as websterite (e.g. McCoy-West et al., 2013). However, there is no obvious textural distinction between the peridotite and websterite lithologies, and all contain high mantle-like bulk Cr and Ni contents (≥ 1774 μg/g; Table 2).

Table 2

Summary of new Kakanui peridotite xenolith data

	ol mg#	spl Cr#	SiO2	TiO2	Al2O3	Fe2O3	MnO	MgO	CaO	Na2O	K2O	P2O5	LOI	Total	Cr	Ni	Os	Ir	Ru	Pt	Pd	Re	187Re/188Os	187Os/188Os	TRD	TMA
			wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	wt %	μg/g	μg/g	ng/g	ng/g	ng/g	ng/g	ng/g	ng/g			Ma	Ma
KAK-1	89.4	-	44.43	0.12	3.17	8.56	0.14	37.37	4.06	0.37	0.02	0.05	1.38	99.67	-	2064	1.25	3.34	6.01	3.65	1.13	0.02	0.060	0.12762	96	112
KAK-2	90.9	19.5	46.09	0.06	2.21	7.67	0.12	40.77	1.53	0.20	0.01	0.01	0.71	99.38	2238	2849	1.10	3.36	7.12	2.94	1.17	0.01	0.053	0.11451	1929	2200
KAK-3	90.1	41	42.98	0.01	0.76	9.08	0.12	45.03	0.83	0.09	0.01	0.00	0.63	99.53	2716	2182	1.54	2.52	5.34	1.51	0.29	0.01	0.021	0.11669	1629	1713
KAK-4	90.6	39	43.73	0.02	0.90	8.39	0.12	44.50	1.27	0.13	0.00	0.03	0.45	99.56	2531	1933	3.62	4.83	8.10	9.70	1.43	0.03	0.043	0.11962	1222	1360
KAK-5	89.8	16.5	42.89	0.28	3.16	9.48	0.14	32.88	6.34	0.53	0.09	0.07	3.80	99.67	2174	2788	1.46	3.01	5.83	3.24	2.14	0.01	0.043	0.12798	46	51
KAK-6	90.4	13.9	43.93	0.03	2.62	8.58	0.12	38.49	3.82	0.22	0.02	0.01	1.81	99.65	2331	2880	1.47	3.39	7.75	4.51	3.55	0.03	0.090	0.12234	842	1068
KAK-7	89.7	10	44.86	0.10	3.71	8.52	0.13	36.91	3.72	0.34	0.01	0.01	0.99	99.30	2111	2770	1.85	3.66	7.68	7.08	5.96	0.04	0.092	0.12577	359	459
KAK-8	89.5	10	44.93	0.11	3.67	8.88	0.13	37.33	3.55	0.32	0.01	0.01	0.69	99.63	2149	2904	1.92	3.16	6.74	2.81	1.87	0.01	0.037	0.12783	67	74
KAK-9	-	26.1	28.67	0.13	3.29	7.15	0.27	14.11	26.63	0.78	0.15	0.14	22.09	103.41	1774	3634	3.00	4.17	10.56	5.42	1.66	0.01	0.023	0.12464	519	549
KAK-10	89.9	11.4	43.78	0.04	3.57	8.25	0.24	26.00	8.92	0.39	0.09	0.03	9.04	100.34	2775	2448	1.93	3.62	8.07	4.39	2.61	0.02	0.046	0.12518	442	496

Non-carbonated peridotites comprise olivine-clinopyroxene-orthopyroxene-spinel (Fig. 6a). Olivine Mg# values are 87.9 to 90.6 and spinel Cr# (=100*Cr/(Cr + Al)) varies between 9 and 41 (Table 2; Supplementary 2). On an olivine Mg#-spinel Cr# plot (Fig. 6b), most samples plot on or close to the post-Archean melting array. The clinopyroxene grains are Cr-diopside and the orthopyroxene grains are Cr-enstatite. Several samples are either crosscut by 0.5 to 1 cm wide veins of brown amphibole or have amphibole surrounding spinel. Following the Hawthorne et al. (2012) classification scheme, this amphibole is pargasite in KAK-1 and ferri-kaersutite in KAK-5. Klemme (2004) documented the occurrence of apatite, as well as silicate and fluoride glasses along grain boundaries, in two Kakanui wehrlite xenoliths.

Fig. 6

A. Plane polarised light image of KAK-4, one of the rare fresh spinel peridotites from the Kakanui Mineral Breccia. B. The spinel-olivine data plot on or close to the post-Archean melting curve. Melting curves from the compilation of Scott et al. (2019). Additional Zealandia peridotite data are from McCoy-West et al. (2013), Scott et al. (2014b), Dalton et al. (2017), Scott et al. (2019), Shao et al. (2021) and Scott et al. (2023).

The 10 peridotite xenoliths reported here have a wide range of loss on ignition values (0.45 to 22.1 wt %) (Table 2). When recalculated as anhydrous, the talc-, carbonate- and zeolite-bearing KAK-5 (LOI = 3.80 wt %), KAK-10 (LOI = 9.04 wt %) and KAK-9 (LOI = 22.1 wt %) give anomalous major element compositions for a typical peridotite, especially MgO (<34.3%). Since these samples have high LOI and high CaO (Table 2), the occurrence of CaCO₃ will have had an effect on the normalization. During the LOI process, CO₂ will be lost Ca will remain, which means that the normalization process overestimates the mantle Ca abundance and consequently underestimates the other elements. Thus, more weight should be given to the results of samples with low LOI (e.g. KAK-2, 3, 4, 7, and 8). For these aforementioned samples, the suite has a moderate to low (volatile-free) Al₂O₃ range from 0.76 to 3.78 wt % (Table 2,  Supplementary 2).The concentrations of platinum group elements in the 10 peridotites (Table 2; Fig. 7), and one sample (MS179C) analyzed by McCoy-West et al. (2013, 2015), are mostly above 1 ng/g. Rhenium concentrations, however, are very low (<0.04 ng/g), except for MS179C (0.265 ng/g; McCoy-West et al., 2015). Peridotites with the highest olivine Mg# tend to have lower (Pd/Ir)_N values (Fig. 7).

Fig. 7

Platinum group elements of Kakanui peridotites normalised to primitive upper mantle (PUM) values of Becker et al. (2006). The data for the ‘fertile Zealandia’ peridotites are from East Otago (McCoy-West et al., 2015) and the ‘depleted Zealandia’ field is from data in Liu et al. (2015) and Scott et al. (2019). These fertile versus depleted groups are differentiated by geographical position (see Scott et al., 2019).

Peridotite and pyroxenite thermobarometry

Two-pyroxene equilibration temperatures for the peridotites, using the Taylor (1998) geothermometer, range from 795°C to 1020°C when a pressure of 1.5 GPa is assumed (Table 2). Although a single pressure (depth) cannot account for such a temperature range, varying the P by 0.5 GPa (to cover the fertile spinel facies peridotite range of ~1–2 GPa) makes only ~ ± 15°C difference.

Garnet pyroxenite temperatures were determined in two ways. First, they are calculated using the recalibrated Mg–Fe exchange thermometer of Sudholz et al. (2022), assuming a pressure of 1.5GPa. As per the peridotites, there is no firm constraint on their precise depth of extraction but varying the pressure by ±0.5 GPa makes only ± ~ 30°C difference. The results (Table 1) are OU 20225 = 1160°C, OU 25652 = 1170°C, KGP1 = 1209°C, OU 49325 =1230°C and OU25651 = 1273°C for kaersutite-bearing samples, and 1309°C for both the kaersutite-free OU 49320 and OU 49324. Second, garnet-clinopyroxene REE thermometry on five of the seven samples that have appropriate Mg#, following the method and criteria Sun & Liang (2015), gives lower temperatures than the Fe–Mg exchange method: OU 20225 is 871°C, OU 25651 is 892°C, OU 25652 is 881°C, OU 49320 is 956°C, and OU 49325 is 1025°C.

Isotopic composition of the Kakanui mantle

O isotopes

Garnet pyroxenite δ¹⁸O values for KGP1, OU 20225, OU 25651, OU 49324, and OU 49325 have a very small spread, with garnet typically having +5.4 to 5.7‰ and clinopyroxene having +5.3 to 5.8‰ (Fig. 8a; Table 3). These data overlap with the published garnet (+5.3‰) and clinopyroxene (+5.2‰) data from a Kakanui ‘eclogite’ (Gonzaga et al., 2010), as well as a pyrope megacryst (+5.7‰; Urosevic et al., 2018). The pyroxenite data are like olivine data from the Kakanui peridotites, although these tend to have a slightly broader range (δ¹⁸O of +5.0 to 5.9). Olivine data from four depleted (+5.3‰ to +6.2‰) and four metasomatized (+5.3‰ to +6.0‰) peridotites from the nearby Dunedin Volcanic Group yield comparable results to those of the Kakanui peridotite silicate minerals (Fig. 8a; Table 3).

Fig. 8

O–Sr–Nd–Pb–Hf isotope data for Kakanui garnet pyroxenites and peridotites. A. Oxygen isotope ratios of the Otago mantle, with duplicates included. B. Age-corrected ⁸⁷Sr/⁸⁶Sr versus εNd. As there is no ⁸⁷Sr/Sr for garnet, it is shown as a field. C and D. Pb isotopic data. Note that these Pb data are not age corrected because of imprecise U-Th-Pb concentration data. The mantle array is from Hart (1984). E. Age corrected εHf versus εNd. The εHf array is from Vervoort & Blichert-Toft (2000). F. Evolution lines of garnet pyroxenite calculated bulk rock εHf values. The 111 Ma age comes from the Lu–Hf age data of OU 25652. Data sources for the diagrams are from this study, with additional pyroxenite oxygen isotope data from Gonzaga et al. (2010), Otago zircon megacrysts from van der Meer et al. (2019), a Kakanui garnet megacryst from Urosevic et al. (2018), and Zealandia intraplate basalts from Timm et al. (2009) and McCoy-West et al. (2010). Radiogenic isotope data for peridotites are Scott et al. (2014b) and McCoy-West et al. (2016); the field for the Dunedin Volcanic Group is from Hoernle et al. (2006), Sprung et al. (2007), Scott et al. (2020a), Pontesilli et al. (2021) and Wilson et al. (2004); the Waiareka-Deborah volcanic field is from Hoernle et al. (2006) and Scott et al. (2020b). Otago crust isotopic values are from Scanlan et al. (2018, 2020) and Scott et al. (2020a). The Hf isotopic data for the pyrope and amphibole megacrysts and the Kakanui nephelinite are from Fulmer et al. (2010).

Table 3

Oxygen isotope data for the Kakanui garnet pyroxenite and spinel peridotite xenoliths from Kakanui and the Dunedin Volcanic Group.

		δ18O
Garnet pyroxenite
Clinopyroxene	KGP1	5.5	5.5
	OU 20225	5.3
	OU 25651	5.8	5.7
	OU 49324	5.3	5.2	5.3
	OU 49325	5.5	5.5
Garnet	KPG1	5.5	5.4
	OU 20225	5.6
	OU 25651	5.4	5.3
	OU 49324	5.7	5.6	5.5	5.5
	OU 49325	5.3	5.7
Peridotite
Olivine	KAK-1	5.0	5.1
	KAK-2	5.7
	KAK-3	5.9
	KAK-4	5.1
	KAK-6	5.1	5.0
	KAK-7	5.8	5.7
	BAL-8	5.5
	BER-1	5.8
	BER-4	5.3
	FOU-9	6.0
	MUR-2	5.8
	SUM-4	6.2
	TRL-5	5.3
	TRL-10	6.0
Orthopyroxene	KAK-6	5.9
	KAK-7	5.7
Clinopyroxene	KAK-1	6.0
	KAK-2	6.2
	KAK-3	6.4
	KAK-4	6.4
Spinel	KAK-1	4.4
	KAK-4	4.6
	KAK-6	4.5
	KAK-7	5.9

Rb–Sr isotopes

Measured clinopyroxene ⁸⁷Sr/⁸⁶Sr values for 6 garnet pyroxenites range from 0.70282 to 0.70305, with the low ⁸⁷Rb/⁸⁶Sr < 0.01 making age correction negligible (Fig. 8b; Table 4). Amphibole ⁸⁷Sr/⁸⁶Sr for OU 20225 (0.70290–0.70292) and data reported by Basu (1978; ⁸⁷Sr/⁸⁶Sr = 0.70305) overlap clinopyroxene, with the low Rb contents (<0.03 μg/g) also resulting in minimal age correction (0.70287 to 0.70303). No new peridotite ⁸⁷Sr/⁸⁶Sr data are reported here; however, the published clinopyroxene data from 5 Kakanui peridotite xenoliths (Scott et al., 2014b; McCoy-West et al., 2016), as well as two xenoliths from Alma and Round Hill within the same volcanic field (0.70243 to 0.70315; Scott et al., 2014b) overlap the pyroxenites (Fig. 8b).

Table 4

Strontium–Nd–Pb–Hf isotopic data for Kakanui garnet pyroxenite and spinel peridotite xenoliths

	Rb	Sr	87Rb/ 87Sr	87Sr/  86Sr	±2 s internal	Sm	Nd	147Sm/  144Nd	143Nd/  144Ndm	±2 s  internal	εNd	208Pb/  204Pb	±2 s internal	207Pb/  204Pb	±2 s internal	206Pb/  204Pb	±2 s internal	Lu	Hf	176Lu/  177Hf	176Hf/  177Hf	±2 s internal	εHf
Sample	μg/g	μg/g				μg/g	μg/g											μg/g	μg/g
Garnet pyroxenites
Clinopyroxene
OU 20225	0.00	151	0.00	0.70283	0.00001	4.48	14.48	0.1870	0.512940	0.00008	5.9	39.076	0.007	15.638	0.003	19.266	0.003	0.08	2.29	0.0052	0.282878	0.00002	3.8
OU 25651	0.20	101	0.01	0.70291	0.00001	5.68	16.58	0.2069	0.512938	0.00002	5.9	38.650	0.004	15.621	0.001	19.013	0.002	0.05	2.27	0.0032	0.282956	0.00003	6.5
OU 25652	0.15	89	0.00	0.70295	0.00001	4.00	11.69	0.2070	0.512945	0.00001	6.0	38.456	0.003	15.616	0.002	19.163	0.001	0.05	3.24	0.0021	0.282914	0.00003	5.0
duplicate																		0.05	3.19	0.0021	0.282922	0.00002	5.3
OU 49320	0.02	101	0.00	0.70293	0.00001	2.44	7.07	0.2087	0.512933	0.00001	5.8	38.498	0.004	15.599	0.002	18.371	0.002	0.02	1.61	0.0021	0.282945	0.00001	6.1
OU 49324	0.01	94	0.00	0.70282	0.00001	2.43	6.90	0.2125	0.512949	0.00001	6.1	38.037	0.007	15.596	0.002	18.504	0.003	0.02	1.59	0.0020	0.282946	0.00001	6.1
OU 49325	0.05	120	0.00	0.70294	0.00001	3.96	13.39	0.1787	0.512922	0.00009	5.5	37.788	0.004	15.561	0.002	18.180	0.002	0.09	3.04	0.0043	0.282949	0.00001	6.3
Garnet
KGP1						2.06	1.28	0.9718	0.513017	0.00001	7.4
OU 20225						1.43	1.08	0.8003	0.513046	0.00004	8.0							1.07	1.13	0.1354	0.282996	0.00001	7.9
OU 25651						1.62	1.43	0.6864	0.513038	0.00002	7.8							1.22	0.52	0.3348	0.283172	0.00002	14.1
OU 25652						1.41	0.91	0.9405	0.513040	0.00001	7.8							0.76	0.77	0.1402	0.282675	0.00004	−3.4
duplicate																		0.80	0.82	0.1373	0.282698	0.00004	−2.6
duplicate																		0.79	0.80	0.1414	0.282702	0.00003	−2.5
duplicate																		0.78	0.81	0.1361	0.282684	0.00004	−3.1
OU 49320						0.92	0.74	0.7477	0.513070	0.00001	8.4							0.22	0.72	0.0438	0.283188	0.00001	14.7
duplicate						0.92	0.75	0.7378	0.513070	0.00001	8.4							0.23	0.73	0.0447	0.283211	0.00001	15.5
OU 49324						0.91	0.71	0.7799	0.513099	0.00001	9.0							0.36	0.90	0.0573	0.283215	0.00001	15.7
duplicate						0.95	0.89	0.6496	0.513055	0.00002	8.1							0.29	0.77	0.0539	0.283197	0.00001	15.0
Amphibole
OU 20225	28	723	0.11	0.70292	11	6.34	22.62	0.1693	0.512942	0.00002	5.9	38.812	0.003	15.613	0.001	19.420	0.001	2.00	0.56	0.5044	0.283294	0.00008	18.5
duplicate	28	723	0.11	0.70290	10							38.866	0.007	15.631	0.003	18.997	0.003
OU 25651						3.20	9.74	0.1983	0.512947	0.00001	6.0							0.05	1.56	0.0045	0.282961	0.00001	6.7
OU 49325						4.77	17.46	0.1653	0.512929	0.00001	5.7							0.08	2.41	0.0048	0.282943	0.00001	6.0
Bulk rock (calculated)
OU 20225						3.12	8.52	0.4925	0.512994		6.9							0.63	1.65	0.0545	0.282943		6.1
OU 25651						3.64	9.41	0.3948	0.512979		6.7							0.51	1.45	0.0502	0.283042		9.6
OU 49320						2.23	6.19	0.2834	0.512952		6.1							0.05	1.49	0.0049	0.282949		6.3
OU 49324						2.13	5.69	0.3241	0.512979		6.6							0.09	1.45	0.0088	0.282955		6.5
Peridotites (data from  Scott et al. (2014b))
Clinopyroxene
KAK-1	0.07	87	0.02	0.70244	0.00001	1.80	5.25	0.2067	0.513207	0.00001	11.1									0.0406	0.283350	0.00001	20.4
KAK-2	0.15	111	0.00	0.70254	0.00001	1.81	5.82	0.1878	0.513085	0.00001	8.7	37.581	0.003	15.407	0.002	17.898	0.003
duplicate												37.468	0.011	15.388	0.004	17.766	0.005
KAK-3	1.66	505	0.01	0.70281	0.00001	10.18	46.10	0.1333	0.512875	0.00001	4.6
KAK-4	0.03	139	0.00	0.70283	0.00001	2.33	8.08	0.1737	0.512847	0.00001	4.1	39.991	0.007	15.639	0.005	20.300	0.007
ALM-1	0.02	219	0.00	0.70292	0.00001	3.57	15.39	0.1400	0.512910	0.00001	5.3	39.088	0.014	15.647	0.004	19.508	0.007			0.1276	0.28563	0.000013	101.0
ALM-4	0.02	165	0.00	0.70260	0.00001	2.79	10.48	0.1604	0.512995	0.00001	7.0	38.579	0.011	15.599	0.004	19.030	0.005
ROU-1	0.06	30	0.01	0.70288	0.00001	1.59	3.42	0.2814	0.513003	0.00001	7.1	39.431	0.010	15.662	0.004	19.916	0.006
ROU-2	0.14	165	0.00	0.70292	0.00002	1.98	7.71	0.1550	0.512919	0.00000	5.5	39.241	0.007	15.632	0.003	19.702	0.003			0.0444	0.28318	0.000005	14.5

Sm–Nd isotopes

Garnet pyroxenite clinopyroxene ¹⁴³Nd/¹⁴⁴Nd values have a tight range of measured εNd of +5.5 to +6.1, with eruption age-corrected εNd₃₄ = +5.6 to +6.0 (Fig. 8b; Table 4). Amphibole Nd isotope values (εNd of +5.7 to +6.0; εNd₃₄ = +5.8 to +6.1) largely overlap clinopyroxene. The garnet separates have measured εNd of +7.4 to +9.0 (Fig. 8b) that age-correct to εNd₃₄ + 4.0 to +6.5. As per the ⁸⁷Sr/⁸⁶Sr peridotite data, no new peridotite ¹⁴³Nd/¹⁴⁴Nd are reported here; however, published peridotite xenolith clinopyroxene values have a wider array than the pyroxenites (εNd of +4.1 to +18.8; Scott et al., 2014b; McCoy-West et al., 2016) (Fig. 8b; Table 4).

Pb isotopes

The pyroxenite clinopyroxene separates have ²⁰⁶Pb/²⁰⁴Pb of 18.180 to 19.266, ²⁰⁷Pb/²⁰⁴Pb = 15.561 to 15.638, and ²⁰⁸Pb/²⁰⁴Pb of 37.788 to 39.076 (Fig. 8c, d; Table 4). Duplicate amphibole measurements from OU 20225 have an internal variation of ²⁰⁶Pb/²⁰⁴Pb = 18.997 to 19.420, ²⁰⁷Pb/²⁰⁴Pb = 15.613 to 15.631, and ²⁰⁸Pb/²⁰⁴Pb = 38.812 to 38.866 show it to be internally heterogeneous but within the clinopyroxene range. The three Kakanui peridotite samples are either quite radiogenic (²⁰⁶Pb/²⁰⁴Pb of 20.150 to 20.300; MS179C and KAK-4) or distinctly less radiogenic (17.767, 17.898; KAK-2) (Scott et al., 2014b; McCoy-West et al., 2016) than the pyroxenites (Fig. 8c, d; Table 4). The peridotite clinopyroxene Pb data from the nearby Alma and Round Hill locations (Scott et al., 2014b) overlap the pyroxenite values (Fig. 8c, d).

Lu–Hf isotopes

The Lu–Hf garnet and clinopyroxene isotope data for the garnet pyroxenites show significant variability (Table 3). Clinopyroxene separates in OU 25651, OU 25652, OU 49320, OU 49325, and OU 20225 have ¹⁷⁶Lu/¹⁷⁷Hf values of <0.005 and εHf of +3.8 to 6.5 values that correct at 34 Ma to +4.4 to +7.2 (Fig. 8e). Amphibole Lu–Hf results for OU 20225, OU 25651 and OU 49325 give εHf of +6.7 and + 18.5, with OU 20225 having an unusually high ¹⁷⁶Lu/¹⁷⁷Hf (0.50), but all correcting at 34 Ma to a tight range of + 6.1 to +8.4 (Fig. 8e). Measured garnet values for OU 20225, OU 25651, OU 25652, OU 49320, and OU 49325 range from εHf of +7.1 to +15.7 and have ¹⁷⁶Lu/¹⁷⁷Hf = 0.04 to 0.33. Garnet in OU 20225, OU 25651, OU 49320, and OU 49324 correct to + 6.9 to +8.4 at 34 Ma, wwhereas all four fractions of the coarse-grained OU 25652 give +12.3 to +12.7 (Fig. 8e). Calculated bulk rock εHf values, for the four samples for which point counting could be reliably undertaken to determine representative modes (OU 20225, OU 25651, OU 49320, and OU 49324), yield +6.1 to +9.6 for the present day, with depleted mantle model ages of ~ −1.0 Ga for OU 20225 and OU 25651 but ~ 0.5 Ga for OU 49320 and OU 49324 (Table 4). The εHf of three of these four samples converge at +8.3 at ~111 Ma (Fig. 8f). Scott et al. (2014b) found the peridotite εHf to be extremely variable within the volcanic field, with one Kakanui (+20.4) and two Alma samples (+14.5, 101) (Table 4), correcting at 34 Ma to +14.2, +20.1, and + 98.9. This wide range was also found in peridotite xenoliths from the adjacent Dunedin Volcanic Group.

Re–Os isotopes

Attempts to gather ¹⁸⁷Os/¹⁸⁸Os on the garnet pyroxenites were unsuccessful due to their extremely low Os contents; however, 11 Kakanui peridotites, including the websterite analyzed by McCoy-West et al. (2013), have bulk rock ¹⁸⁷Os/¹⁸⁸Os of 0.11451 to 0.12783 (Table 2). The low ¹⁸⁷Re/¹⁸⁸Os (<0.09, except for 0.271 in MS179C of McCoy-West et al. (2013)) means that the mantle and Re-depletion Os (T_RD) model ages are very similar (Table 2). The Re-depletion Os model ages (T_RD) range from 1.9 to 0.05 Ga and do not form a statistically robust aluminochron (Fig. 9a).

Fig. 9

Summary of T_RD in spinel peridotites from Kakanui compared with the data from the rest of Southern-Central Zealandia from McCoy-West et al. (2013), Liu et al. (2015), Scott et al. (2019) and Delpech et al. (2023). The hatched samples are those from depleted rocks with low Re/Os. The global abyssal peridotite population is from the sulphide compilation of Pearson et al. (2007) and the Kaapvaal Craton population is from Pearson & Wittig (2008).

DISCUSSION

Kakanui spinel peridotites: Zealandia mantle lithosphere melt residues

The main component of the lithospheric mantle is peridotite, and the Kakanui spinel peridotites therefore give insight into the mantle column within which the garnet pyroxenites formed. When the peridotite temperatures of equilibration are projected onto the estimated 70 mW∙m⁻² Cenozoic geotherm for Otago determined by Scott et al. (2014a, 2014b) from spinel peridotites and heat flow measurements, these give depths from 35 to 50 km and equate to the middle to upper portion of the Zealandia lithospheric mantle (Fig. 10). The moderately depleted olivine Mg# (<90.6), moderate to low spinel Cr# (< 41) and the low bulk rock Al₂O₃ (<4.0 wt %) indicate that these rocks are moderately depleted melt residues. This is supported by platinum group element patterns, which are more like fertile Zealandia xenoliths than extremely fractionated patterns of highly depleted ones (Liu et al., 2015; McCoy-West et al., 2016; Scott et al., 2019; Delpech et al., 2023) (Fig. 7). The δ¹⁸O values for peridotite olivine grains are similar to typical mantle values of ~ + 5.5 (Fig. 8a).

Fig. 10

The Fe–Mg geothermometric results projected onto the Otago Cenozoic geotherm of Scott et al. (2014a, 2014b), which was estimated from spinel peridotite geothermometry and heat-flow measurements.

The Kakanui peridotite Os isotope data show no clear correlation with depletion indices such as Al₂O₃, and have minimum Re-depletion Os model ages ranging from Eocene to Paleoproterozoic (Fig. 9a). Samples with low LOI also tend to have low ¹⁸⁷Os/¹⁸⁸Os, which suggests that post-depositional alteration has not affected these isotope ratios. Clinopyroxene trace element data from five of the peridotites, previously presented and discussed in Scott et al. (2014b) and McCoy-West et al. (2015), have revealed that the peridotites are slightly enriched via metasomatism. While metasomatism could potentially have caused the dispersion in Os isotopes from, for example, isotopic data that would otherwise yield a mid-late Proterozoic isochron, even heavily metasomatized Archean peridotite suites commonly retain strong evidence for their ancient melt depletion (e.g. Liu et al., 2020; Pearson & Wittig, 2008) (Fig. 9b). The low ¹⁸⁷Re/¹⁸⁸Os ratios (mostly <0.09) indicate that the Kakanui Re-depletion Os model ages could approximate melting ages if only a single stage of melting has taken place, which may be unlikely in the case of many Zealandia mantle rocks (e.g. Liu et al., 2015; Scott et al., 2019). Even so, the occurrence of several unradiogenic Os isotope ratios indicate that this lithospheric mantle column must contain ancient fragments (Fig. 9a) (McCoy-West et al., 2013; Scott et al., 2014b; Liu et al., 2015; Scott et al., 2019; Delpech et al., 2023), with the lack of correspondence between the Os isotopes and any other element or isotope ratio supporting the interpretation that this mantle lithosphere is composed of fragments with different depletion ages (Liu et al., 2015).

Kakanui garnet pyroxenites: Cumulates of alkaline mantle melts

Since the Zealandia margin of Gondwana has a long and complicated history of accretion, subduction, and extension, the genetic processes that led to the formation of Kakanui garnet pyroxenites are difficult to unravel. On the basis that REE diffusion is much slower Fe and Mg diffusion (Sun & Liang, 2015), the Fe–Mg geothermometry of the garnet pyroxenite suite, rather than the REE thermometry results, should give the ambient mantle temperatures from which the rocks were extracted. When projected onto the ~70 mW m⁻² NZ geotherm, the calculated temperatures indicate depths of 60 to 80 km. These depths are slightly greater than the peridotite xenoliths and close to the base of the lithosphere (Fig. 10). The garnet pyroxenites, therefore, form an underplate in the Zealandia lithospheric mantle deep beneath Kakanui.

Although derived from the middle to lower lithospheric mantle, the whole-rock trace element and major element concentrations are inappropriate for the garnet pyroxenites to be deeply trapped melts. Their low SiO₂ also discounts them being mildly evolved basalt, and the whole-rock trace element patterns are too depleted in incompatible LILE and LREEs compared to intraplate OIB-like melts that form intraplate magmas through New Zealand (Fig. 4). Their moderate MgO and low Cr concentrations (Fig. 5b, d) mean that they are unlike pyroxenites formed by strong wall-rock reaction between a percolating silicate melt and the surrounding mantle (e.g. Liu et al., 2005), although such evidence is strongly dependent upon the composition and volume of the infiltrating melt. The most plausible explanation for these garnet pyroxenites is that they are cumulates. The extremely low PGE abundances contrast with those of typical mantle-derived basalts (Day, 2013), and are likely a result of forming from melts before those magmas achieved low pressure sulfide saturation.

While a cumulate origin is a widely accepted mechanism for garnet pyroxenite generation (e.g. Downes, 2007), the tectonic affinities of the parental melts are important for establishing the rocks' origin. The geochemical data show that the Kakanui pyroxenites are distinct from some well-characterized garnet pyroxenite suites (Fig. 5b, c, d). For example, while they are similar in MgO and SiO₂ to some of the Sierra Nevada pyroxenites (Lee et al., 2006), they are mostly less Mg-rich than the majority of garnet pyroxenite xenoliths from SE Australia (Griffin et al., 1988) or those in the orogenic peridotite at Beni Bousera (Pearson et al., 1993) (Fig. 5b). Furthermore, the Beni Bousera (Pearson et al., 1993) and orogenic pyroxenites in Austria (Becker, 1996), as well as intraplate xenoliths in China (Xu, 2002) and SE Australia (Lu et al., 2020), have δ¹⁸O values and radiogenic ⁸⁷Sr/⁸⁶Sr that indicate that the parental melts were at least in part derived from subducted crust; these properties contrast with the mantle-like δ¹⁸O values (Fig. 8a), the unradiogenic ⁸⁷Sr/⁸⁶Sr, and radiogenic ¹⁴³Nd/¹⁴⁴Nd (Fig. 8b) and ¹⁷⁶Hf/¹⁷⁷Hf (Fig. 8e) of the Kakanui pyroxenites. Moreover, the amphibole Nb concentrations are too high for rutile as a residual phase and the rocks to be related to arc magmatism (e.g. Coltorti et al., 2007).

Any magma traversing the lithospheric mantle could form a cumulate and, therefore, knowing the composition of the melts that were in equilibrium with the garnet pyroxenites would help elucidate their origin (e.g. Becker, 1996; Lu et al., 2020). If subsequent re-equilibration has not drastically changed the mineral trace element abundances, the compositions of the equilibrium melts from which garnet, clinopyroxene, and amphibole crystallized can be estimated utilizing garnet/liquid, clinopyroxene/liquid, and amphibole/liquid partition coefficients. Most of the mineral/melt partition coefficients used here (Rb, Ba, Th, U, Nb, La, Ce, Pb, Sr, P, Nd, Zr, Hf, Sm, Ti, Tb, Ho, Tm, Yb, and Lu) are from Adam & Green (2006), with K partitioning between clinopyroxene/melt and garnet/melt from Gaetani et al. (2003) and amphibole/melt from Dalpe & Baker (1994). Partition coefficients for Pr, Eu, Gd, Dy, and Er are from Zack et al. (1997) (Supplementary 2).

The modelled equilibrium liquids show general agreement between different minerals within samples, particularly in the middle and heavy REE (Fig. 11a-c). The scatter increases in large ion lithophile elements (LILE) and high field strength elements (HFSE), likely because of lower analytical precision caused by low elemental abundances and/or less well-constrained partition coefficients (e.g. Fulmer et al., 2010). Although ilmenite is rarely observed, it does not appear to have significantly influenced the melt composition by depleting HFSE prior to crystallization of the main phases because clinopyroxene lacks negative Ti anomalies (Fig. 4) and texturally, ilmenite may represent a late stage intercumulus phase (Fig. 3). An exception is sample KGP1, where the liquids modelled from amphibole compositions have flatter HREE patterns as well as depletion in Rb, Ba, and K and enrichment in Th and U, compared to those modelled from garnet and clinopyroxene in the same sample (Fig. 11c); these minerals in this sample may not be in equilibrium. Additionally, equilibrium melts calculated from sample OU 49324 have very steep middle- to heavy REE patterns. These may result from early crystallization of garnet from the primary melt to deplete the heavy REE prior to forming the sampled pyroxenite, as garnet can be the liquidus phase at high pressure in strongly alkaline liquids (e.g. Foley, 1990). The summary of averaged equilibrium liquids shows the theoretical melts to resemble OIB but also having greater enrichment in LILE, HFSE, and light REE (Fig. 11d). The predicted melt compositions lack negative Nb anomalies, but have positive Th, U, and Nb and negative K anomalies. In this regard, they are like Zealandia intraplate alkaline rocks, such as the host 34 Ma Kakanui nephelinite (Scott et al., 2020b) or ~ 25 Ma Alpine Dike Swarm lamprophyres (Cooper, 2020). This can be seen also in the δ¹⁸O values, which overlap the published range for Zealandia intraplate volcanic rocks (Timm et al., 2009: McCoy-West et al., 2010) (Fig. 8a).

Fig. 11

A–C. Representative modelled liquid compositions from the different mineral phases in the garnet pyroxenites. D. equilibrium melts averaged from the individual mineral equilibrium melt models in each sample apart for KGP1 where amphibole is excluded owing to its disequilibrium with garnet and clinopyroxene. Abundances are normalized to primitive mantle. Primitive mantle, OIB and MORB compositions are from Sun & McDonough (1989). KAK is a Kakanui host nephelinite from Scott et al. (2020b), DVG is the Dunedin Volcanic Group from Scott et al. (2020a) and ADS is an ~24 Ma Alpine Dike swarm lamprophyre from Cooper (2020).

Fig. 12

A. Sm–Nd isotopic data for Kakanui garnet pyroxenites. B. Lu–Hf isotopic data for the garnet pyroxenites, and the host rock nephelinite and amphibole and garnet xenocrysts from Fulmer et al. (2010).

Age and significance of the Kakanui garnet pyroxenites

Intraplate alkaline magmatism has been occurring in Zealandia since shortly after the cessation of subduction and associated calc-alkaline magmatism along the Gondwana margin at ~110 to 100 Ma (Baker et al., 1994; Muir et al., 1997: Timm et al., 2010; van der Meer et al., 2016). Combinations of garnet plus clinopyroxene ± amphibole in five garnet pyroxenite samples for which there is sufficient spread of ¹⁴⁷Sm/¹⁴⁴Nd yielded isochron ages of 18.8 ± 4.6 to 40.2 ± 3.7 Ma (Fig. 12a) that span the known eruption age of 34 Ma, as determined from kaersutite ⁴⁰Ar/³⁹Ar data (Hoernle et al., 2006). Since the amphibole and clinopyroxene separates have very little isotopic variation, the differing isochron ages are caused by variation in garnet ¹⁴⁷Sm/¹⁴⁴Nd, which probably is at least in part due to garnet separates containing small amounts of included clinopyroxene or amphibole coupled with small amounts of depositional-stage alteration of the rocks. The Sr, Nd, and Pb isotopic data show that the garnet pyroxenites cannot be cumulates of the host Waiareka-Deborah Volcanic field magmas (Fig. 8b, c, d); thus, given the high calculated equilibration temperatures of the samples, the Sm–Nd isochrons therefore probably date cooling at the time of eruption (Mezger et al., 1992; Scherer et al., 2000; Smit et al., 2013).

Like the Sm–Nd data, the garnet pyroxenite Lu–Hf isochron ages for OU 20225 (43.9 ± 11.2 Ma), OU 25651 (33.9 ± 3.8 Ma), OU 49320 (31.9 ± 13.2 Ma) and OU 49324 (54.4 ± 13.6 Ma), as well as the host nephelinite and garnet xenocrysts (Fulmer et al., 2010) (Fig. 12b), overlap the 34 Ma eruption age (Hoernle et al., 2006). However, there is one exception: garnet pyroxenite OU 25652, which has unusually coarse garnet and one of the lowest Fe–Mg exchange (1168°C) and lowest garnet-clinopyroxene REE (881°C) equilibration temperatures of the pyroxenite suite (Table 1). This sample yields a six-point isochron age of 111.9 ± 9.1 Ma (Fig. 11b). Although it is effectively only a two-point isochron due to the homogeneity of the fractions, this age is the same as that given by the intersection of bulk rock εHf evolution lines of OU 25261, OU 49320, and OU 49324 (Fig. 8f). Depleted mantle extraction ages of ~111 Ma for the parental melts, therefore, indicate that initiation of alkaline intraplate mafic magmatism in Zealandia occurred almost immediately after the cessation of long-lived subduction along this portion of the Gondwana margin, which occurred between ~110 and 100 Ma (Tulloch et al., 2009; van der Meer et al., 2017). The oldest to-date known (~102–100 Ma) post-subduction intraplate Zealandia magmatism occurs in Westland on the western side of the South Island of New Zealand within the former back arc (Fig. 1a) (Muir et al., 1997; Tulloch et al., 2009). These rocks have been dextrally displaced 480 km northwards along the Alpine Fault in the last 25 Ma. Kakanui, by contrast, resides in the position of the former forearc and significantly closer to the former trench (Fig. 1a). Two felsic tuffs from the East Otago region have been tentatively attributed to hot melting of the lower continental lithosphere in an extending crust at ~112 Ma (Tulloch et al., 2009). These could conceivably be associated with the magmatic event that led to the Kakanui garnet pyroxenites.

With a 200°C range in the results of Fe–Mg exchange geothermometry, projection of the garnet pyroxenites onto the Cenozoic Otago geotherm indicates that they have vertical distribution of at least 10 km above the base of the lithosphere (Fig. 10, 13). Since the garnet pyroxenites originate from near the base of the lithosphere and have isotopic compositions at least in part distinct from peridotitic components, the primary magmas to the cumulates were therefore probably derived from the asthenosphere. This also means that the host nephelinite that picked up the xenoliths was also derived from either the base of the lithosphere or the asthenosphere. In both cases, debate about the occurrence of pargasitic mantle amphibole in the source of Zealandia intraplate alkaline magmas (e.g. Panter et al., 2006; Scott et al., 2016a, 2020a; Cooper et al., 2024) is not relevant here because this mineral is not thermally stable at temperatures near the base of the lithosphere (1350^oC). In contrast, the occurrence of amphibole in Kakanui mantle peridotites is consistent with the observation that they are derived from cooler temperatures than the pyroxenites and consequently from shallower in the mantle column (Fig. 10). The peridotite metasomatic clinopyroxene grains have similar Sr–Nd and Pb isotopic compositions to amphibole from the garnet pyroxenites (Fig. 8b, c, d) and hence may represent the products of shallower mantle enrichment caused by parental magmas associated with the garnet pyroxenites. Although further work is required to examine this hypothesis, amphibole veins in peridotites from elsewhere in New Zealand have recently been shown to have the appropriate ⁸⁷Sr/⁸⁶Sr values (~0.7028–0.7029) of primitive intraplate magmas (Cooper et al., 2024).

Fig. 13

Summaryschematic of the lithosphere column beneath Kakanui. The garnet pyroxenite is present in the lower lithosphere, with the examined spinel peridotites representing fragments from the mid-mantle lithosphere. The granulite facies middle and lower crust is from Jacob et al. (2017).

CONCLUSIONS

The mantle cargo in the 34 Ma Kakanui Mineral Breccia in New Zealand reveals a heterogeneous mantle lithosphere beneath a former Cretaceous forearc. Spinel peridotite xenoliths indicate that shallower portions of this mantle lithosphere are moderately depleted and enriched, and have a wide range of Re–Os minimum model ages (and low Re/Os) from Eocene to Paleoproterozoic. These rocks also have highly variable Sr, Nd, Pb, and Hf isotope ratios, but mantle-like O isotopes. The garnet pyroxenite xenoliths, on the other hand, were extracted from the middle to lower New Zealand mantle lithosphere. Their mantle-like δ¹⁸O and unradiogenic ⁸⁷Sr/⁸⁶Sr signatures are distinct from many other global garnet pyroxenite occurrences, which are commonly interpreted to be derived from subducted oceanic lithosphere. Geochemical characteristics indicate the Kakanui garnet peridotites to be cumulates that were in equilibrium with alkaline melts of similar trace element composition to known Zealandia intraplate alkaline magmas. Although most of the garnet peridotites yield Sm–Nd (18.8 ± 4.6 to 40.2 ± 3.7 Ma) and Lu–Hf isochron (54.4 ± 13.6 Ma to 31.9 ± 13.2 Ma) ages close to the age of eruption (34 Ma), one coarse grained garnet-bearing sample yielded an isochron age of 111 Ma that is consistent with the intersection of Hf isotope evolution lines for three other samples and likely documents the parental melt formation event. The primary magmas to these pyroxenites, therefore, appear to have been extracted from the mantle in the Early Cretaceous and hence cryptically record one of the first phases of alkaline magmatism in Zealandia after the cessation of long-lived subduction along this portion of the Gondwana margin.

Acknowledgements

This work was supported by an NSERC Discovery grant (RGPIN-2020-04692) to Pearson, an Accelerator Grant (RGPAS-2020-00069) and the Canadian Foundation for Innovation and British Columbia Knowledge Development Fund (Joint Project 229814) to Smit, and an Alexander von Humboldt grant (NZL1222206) to Scott. We thank K. Panter, M. Bizimis and an anonymous reviewer for their comments, and T. Waight for handling the manuscript.

Data Availability Statement

Data used in this article are available in the text and supplementary files.