Diffusion-zoned pyroxenes in an isotopically heterogeneous mantle lithosphere beneath the Dunedin Volcanic Group, New Zealand, and their implications for intraplate alkaline magma sources

Intraplate alkaline volcanic rocks have been argued to originate from melting of oceanic crust mixed with sedimentary or peridotitic components (e.g., Hofmann and White, 1982; Chauvel et al., 1992) or from metasomatized lithospheric mantle peridotite (e.g., Class and Goldstein, 1997; Pilet et al., 2008). In the latter case, amphibole may be an important component because it can provide the high concentrations of incompatible elements to a melt that are commonly lacking in depleted peridotite and occurs in a variety of tectonic settings (e.g., Class and Goldstein, 1997; Panter et al., 2006; Sprung et al., 2007; Rooney et al., 2014; Pilet, 2015; Scott et al., 2016b).

Small volume and sporadically erupted late Oligocene–Miocene intraplate alkaline basalts within the ∼10,000 km² Dunedin Volcanic Group in New Zealand (Fig. 1) have distinctive isotopic source signatures that trend toward and into the high time-integrated U/Pb–like (HIMU) isotopic field (Reay et al., 1991; Price et al., 2003; Hoernle et al., 2006). Note that “HIMU-like” is used to emphasize that while the Sr, Nd, and ²⁰⁶Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb trend toward and into the HIMU field compiled by Stracke et al. (2005), ²⁰⁷Pb/²⁰⁴Pb is typically less radiogenic. In keeping with the worldwide debate on intraplate alkaline volcano source regions, the Dunedin Volcanic Group source reservoir has been inferred to be oceanic crust located within the asthenosphere (Reay et al., 1991; Hoernle et al., 2006; Coombs et al., 2008; Timm et al., 2010) or metasomatized peridotite within the lithosphere (Price et al., 2003). The common New Zealand lithospheric source model involves melting of amphibole-veined peridotitic mantle (Finn et al., 2005; Panter et al., 2006; Sprung et al., 2007).

Rather than using the derivatives of melting to characterize possible mantle sources, we examine the peridotitic xenolith cargo for signs that the lithospheric mantle may have played a role in Dunedin Volcanic Group alkaline basalt genesis. This paper summarizes the results of a petrological study of 10 peridotite xenoliths from 3 geographically close locations on the central-western flank of the province and integrates them with existing data. Mineral major and trace element and radiogenic (Sr, Nd, Pb) isotopic compositions permit a comprehensive characterization of the upper portions of the lithospheric mantle column. A ubiquitous feature of the xenoliths is the zoned pyroxenes that, when coupled with diffusion calculations, permit insight into the time scales of equilibration and thermal changes that have occurred within this lithospheric mantle. The results highlight the utility of peridotite xenoliths in aiding interpretation of alkaline intraplate magma sources.

The exposed New Zealand crustal rocks, which are part of the largely undersea Zealandia continent, comprise numerous tectonostratigraphic terranes that were accreted to the paleo-Pacific Gondwana subduction margin throughout the Phanerozoic (Mortimer, 2004, and references therein). Between ca. 110 and 105 Ma this margin began to undergo extension and crustal thinning (e.g., Gibson et al., 1988; Gray and Foster, 2004; Scott and Cooper, 2006), which led to the 84 Ma formation of oceanic seafloor between Zealandia, Australia, and Antarctica (Gaina et al., 1998). Seafloor spreading ceased between Australia and Zealandia at 55 Ma but is ongoing in the Southern Ocean between Antarctica and Zealandia.

Intraplate alkaline basalts have erupted through Zealandia over the past ∼102 m.y. (e.g., Timm et al., 2010; van der Meer et al., 2016); the ∼10,000 km² late Oligocene–Miocene (24–9 Ma) Dunedin Volcanic Group in East Otago is one of the larger spatially distributed intraplate magmatic provinces. The dispersed flows, dikes, and plugs are dominantly composed of alkaline basanites that have ocean island basalt–like chemistries (Price and Taylor, 1973; Hoernle et al., 2006; Coombs et al., 2008) and HIMU-like isotopic source signatures (Reay et al., 1991; Price et al., 2003; Hoernle et al., 2006; Coombs et al., 2008; Timm et al., 2010). The volcanic rocks were erupted through a crust dominated by Jurassic metasedimentary rocks (Mortimer, 2004; Adams et al., 2007). Godfrey et al. (2001) seismically detected the Moho to be at a depth of 25–30 km beneath Dunedin. This depth, coupled with high heat flow measurements of ∼90 mWm² reported in Godfrey et al. (2001) and high equilibration temperatures in some fertile spinel facies peridotite xenoliths, led Scott et al. (2014b) to infer that the late Oligocene–Miocene mantle lithosphere was 30–40 km thick.

The mantle under Zealandia, at least for the Cretaceous to Miocene portions that have been sampled by intraplate basalts or exhumed in orogenic massifs, is heterogeneous and composed of large adjacent fertile and refractory domains (Scott et al., 2014b, 2016a, 2016b; Czertowicz et al., 2016; McCoy-West et al., 2015). Analysis of >100 xenoliths from a combined 15 locations shows the mantle under the Dunedin Volcanic Group to be moderately fertile (Scott et al., 2014a, 2014b; McCoy-West et al., 2015, 2016). Peridotite xenoliths from across Otago show the area to have clinopyroxene Hf (Scott et al., 2014b), Nd and Sr (McCoy-West et al., 2016) and whole-rock Os isotopic compositions (McCoy-West et al., 2013; Liu et al., 2015) that indicate melt extraction from the Archean to the Mesozoic. There is no evidence to date that points to there being crust older than Cambrian in Zealandia and therefore some portions of the mantle are >2 b.y. older than the oldest known crust. The ancient mantle fragments have been suggested to represent a long-lived stable continental platform since the Proterozoic (McCoy-West et al., 2013), or of an amalgamation of fragments that were recycled through the asthenosphere and incorporated into the lithosphere throughout Mesozoic continent construction (Liu et al., 2015). The mantle lithosphere was subsequently metasomatized by carbonatitic or CO₂-bearing silicic melts that imparted a HIMU-like isotopic imprint (Scott et al., 2014a, 2014b, 2016b; McCoy-West et al., 2016). Isochrons fit to Nd and Hf isotopic data suggest that this mantle metasomatism most likely occurred in the Mesozoic, but the fit of data is poor and the age interpretations are complicated by multicomponent mixing and multistage depletion and enrichment histories (Scott et al. 2014b; McCoy-West et al., 2016).

Peridotite mineral major element analyses were obtained in situ on 30 µm thick polished microscope sections by wavelength dispersal spectrometry (WDS) on a JEOL JXA-8600 electron microprobe analyzer at the University of Otago. The microprobe, now retired, had 2 spectrometers and was operated with an accelerating voltage of 15 kV, a current of 20 nA, and a 20 μm beam diameter. Counting times were 30 s on peak positions and 10 s on background positions. Smithsonian minerals were used as standards and data were reduced using an in-house ZAF (atomic number, absorption, and fluorescence excitation effects) correction program. Quantitative line scans and subsequent mineral analysis were subsequently obtained using a Zeiss Sigma variable pressure scanning electron microscope with an energy dispersal spectrometer (EDS) attachment at the University of Otago. This instrument was operated with a 60 or 120 µm aperture at a working distance of 8.5 mm with a 15 kV accelerating voltage and a current of 20 nA. The beam has a resolution of several micrometers. Data were standardized using Smithsonian microbeam standards, with the analytical routine tested against known composition minerals and Oxford Instruments Aztec program (https://www.oxford-instruments.com). The EDS data are comparable in quality and reproducibility to the WDS data, and there are no systematic differences in samples that have had both methods applied.

Pyroxene trace element concentrations were measured in situ on polished thin sections and on grain mounts at the University of Otago, and on grain mounts at the University of Alberta. Both instruments are Resonetics RESOlution M-50-LR laser ablation systems. For the polished sections and grain mounts, background data were acquired for 20 s followed by a 40 s analysis at a 5 Hz repetition rate, a fluence of ∼4 J/cm², and a spot diameter of 75–50 μm. Each set of 10–14 analyses was standardized against NIST610 (Jochum et al., 2011). Orthopyroxene grains were reanalyzed at Alberta using NIST612 and 614 as standards. Analyses were conducted at 5 Hz, 4 J/cm² with a beam diameter of 75–90 μm. Data were reduced using the Iolite software package (Paton et al., 2011) and the average pyroxene SiO₂, measured by WDS from the cores of grains, was used for internal normalization. A comparison of the Alberta versus Otago laboratory results revealed no systematic differences (Scott et al., 2016a).

Clinopyroxene was separated from 6 samples, leached in cold 2 N HCl for 1 h, and then dissolved in HF, HCl, and HNO₃. Strontium isotopes were determined by thermal ionization mass spectrometry at the University of Copenhagen and Nd and Pb isotopes were analyzed using multicollector–inductively coupled plasma–mass spectrometry at the University of Cape Town, South Africa. The isotopic data were collected at the same time and using the exact same standard column chemistry, standards, blank and data corrections as results presented in Scott et al. (2016b) and Czertowicz et al. (2016); see these articles for the detailed methodology.

Peridotite xenoliths were extracted from basaltic flows on the central-western side of the ca. 24–9 Ma Dunedin Volcanic Group at Summer Hills and Bald Hill and from a dike at Foulden Maar (Fig. 1). Basalt at the base of Foulden Maar is ca. 23 Ma (Timm et al., 2010; Fox et al., 2015) and the associated xenolith-bearing dike is assumed to be about the same age. The precise ages of the Summer Hills and Bald Hill flows have not been determined, although Bald Hill is within 1 km of Foulden Maar and Summer Hills is <2 km northeast of a flow dated by the K-Ar method as ca. 16 Ma (Coombs et al., 2008). Xenoliths in each location are ovoid, mostly fresh, and as much as 5 cm in diameter. We collected ∼100 xenoliths; 15 randomly selected largest xenoliths (∼5–3 cm diameter) from each locality were petrographically inspected and mineral major elements collected. This paper reports these data, along with trace elements and Sr-Nd-Pb isotopes, on a subset of 10 best-characterized samples. The xenoliths suite is dominantly lherzolitic and has a spectrum of textures; most are the porphyroclastic variety of Harte (1977) (Table 1). Additional specimen images, representative microprobe, EDS, and laser ablation data are presented in the GSA Data Repository Item¹.

Olivine core Mg# [100 × Mg/(Mg + Fe)] span a very small range in each sample, with average values of 5–10 analyses per sample yielding values from 89.4 to 91.4 across the sample suite (Table 1). Many grains show minor serpentinization at grain margins or in cracks, but no zoning was noted except for where the olivine is adjacent to the host basalt (these data are not included in the average values). Average spinel Cr# [100 × Cr/(Cr + Al)] varies from 9.3 to 35.0 (Table 1) with darker colors in plane polarized light typically equating to a more Cr-rich composition. Like olivine, the spinel grains are not discernibly zoned except where partially converted to magnetite at xenolith margins, and these are not included in the average values.

The maximum measured width of an orthopyroxene grain in the sample suite is 8.5 mm and the maximum width of a clinopyroxene grain is 3.0 mm, with coexisting orthopyroxene generally being slightly to significantly coarser than clinopyroxene (Table 1). Orthopyroxene cores in samples SUM-4, SUM-6, and SUM-15 contain <5 μm wide rods of exsolved clinopyroxene (Fig. 2A). EDS line scans across these zones of exsolution lamellae and reintegration of chemistries modify the result of an pyroxene single spot analysis insignificantly (<0.5 wt% in all but SiO₂), although this very small change may also partially reflect the WDS or EDS spot analyses having incorporated exsolution within the excitation area beneath the crystal surface. Orthopyroxene WDS and EDS point analyses of rims and cores demonstrate that all grains, including those in samples where the pyroxenes are devoid of exsolution lamellae, are systematically zoned (Figs. 3A, 3B). In all samples but sample FOU-9, Al₂O₃ and Cr₂O₃ decrease and MgO increases toward the rim of the grains (Figs. 3A, 3B). CaO in many samples has a slightly lower concentration at the rim (Fig. 3B). The chemical zoning commonly extends to ∼150 μm from orthopyroxene rims (Fig. 4A). FeO shows no systematic variation. The one exception to the general trends is FOU-9. In this sample the orthopyroxene rims have higher concentrations of Cr₂O₃, Al₂O₃, and CaO but lower MgO than the cores (Figs. 3A, 3B), which is the reverse of the other samples. Line scans show development of heterogeneous Al₂O₃ and Cr₂O₃ concentrations over ∼25–50 μm from the rims, with the CaO and MgO concentrations beginning to increase within ∼60–100 μm of the rims (Fig. 5A).

Clinopyroxene is typically the smallest and least abundant of the three silicate phases (Table 1). Very narrow exsolution lamellae of orthopyroxene are present in only several samples. Like orthopyroxene, all clinopyroxenes show core-rim chemical variations; rims typically have lower concentrations of Al₂O₃ and Cr₂O₃ and higher MgO than the cores (Figs. 3A, 3B). These trends generally mimic those of coexisting orthopyroxene grains except for CaO, which tends to be slightly higher at clinopyroxene rims than cores (Figs. 3A, 3B, and 4B). The zoning penetrates as much as 150 μm into the grains. FOU-9 clinopyroxene has sieve-textured borders (Fig. 2B) and opposing Al₂O₃ and MgO core-rim trends compared to other samples (Fig. 3A). Line scans of FOU-9 clinopyroxenes show the compositional heterogeneity in Cr₂O₃ and Al₂O₃ occurring <50 μm from the rims, with the CaO, FeO, and MgO profiles extending further into the grain (Fig. 5B).

Trace elements were measured in 10 samples. Four samples contain clinopyroxene trace element patterns with flat heavy rare earth element (HREE) concentrations at ∼10× chondrite and a downward inflection in light (L) REEs. These patterns mimic theoretically modeled clinopyroxene REE compositions formed following extraction of melt (Fig. 6A). Exceptions are BAL-4, FOU-3, FOU-9, FOU-15, SUM-6, and SUM-12, which have LREE ± middle (M) REE signatures that depart from predicted melting trajectories (Fig. 6A). In the case of FOU-3, the departure from the modeled REE trends is very subtle and is restricted to the LREEs. Orthopyroxene grains mostly have positive trending REE patterns when plotted against decreasing cation radius (Fig. 6B). Exceptions, relative to modeled values, are FOU-3 and FOU-15, which have distinct upward inflections from Nd to La; FOU-9, which has high LREEs and MREEs; and SUM-6, which has distinctly low HREEs. Primitive mantle-normalized patterns show the clinopyroxenes typically have negative Pb, Zr, and Ti anomalies (Fig. 6C), whereas orthopyroxenes have positive Pb, Zr, Hf, and Ti anomalies (Fig. 6D). FOU-9 is notable for the absence of a positive orthopyroxene Ti anomaly to offset the large negative clinopyroxene Ti anomaly. Laser ablation line scans from the core to rim of pyroxenes were unable to detect any core to rim zonation in REE concentrations (see the Data Repository Item). On the basis of the clinopyroxene trace element distributions, the peridotites are subdivided into unmetasomatized varieties named type 1 (BAL-8, SUM-4, SUM-11, SUM-15), and the remainder are classified as type 2 (Table 1; Fig. 6).

Pyroxene REE equilibration in 8 samples is tested using the lattice strain theory for cation ionic radii (Agranier and Lee, 2007). If orthopyroxene and clinopyroxene equilibrium is not reestablished after some perturbation, then the concentration ratios should deviate dramatically from those theoretically predicted at the respective isotherms. HREEs in all the inspected peridotites generally have linear distributions subparallel to the isotherms (Fig. 6E). MREEs show some dispersion in FOU-15 and SUM-12, and LREEs (especially La and Ce) commonly show a small to large dispersion from the isotherms in the most samples.

Sr, Nd, and Pb isotopes have been measured in clinopyroxene extracted from two type 1 (LREE depleted) xenoliths (BAL-8 and SUM-11) and four type 2 (variably LREE enriched) xenoliths (BAL-4, FOU-3, FOU-9, and SUM-6) (Table 2). Sr and Nd data are corrected to 20 Ma to account for ingrowth since the late Oligocene–Miocene eruption, although this correction makes virtually no difference because Rb and Sm concentrations are so low. No Pb age corrections were undertaken because the Pb, Th, and U concentrations are close to laser detection limits. The new data are plotted with all other published peridotite clinopyroxene isotope data from East Otago (Scott et al., 2014b; McCoy-West et al., 2016), which are here also subdivided into type 1 or type 2 on the basis of their chondrite-normalized trace element patterns.

The type 1 xenoliths tend to have radiogenic Nd [εNd_{(20 Ma)} ≥ + 15.5] that is outside the range of modern mid-oceanic ridge basalt (MORB) and unradiogenic ⁸⁷Sr/⁸⁶Sr (<0.7028) (Table 2; Fig. 7A). Type 2 xenoliths display a wide array of Sr and Nd values but cluster around εNd_{(20 Ma)} = + 7.7 to + 5.1 and ⁸⁷Sr/⁸⁶Sr_{(20 Ma)} = 0.7028 (Table 2), where they overlap with the restricted fields of the Dunedin Volcanic Group basalts (Timm et al., 2010) and the HIMU mantle reservoir (Stracke et al., 2005) (Fig. 7A). FOU-3, which was classified as a type 2 xenolith because the clinopyroxene LREE pattern departs very slightly from predicted melting trends (Fig. 6A), also has radiogenic Nd [εNd_{(20 Ma)} = + 22.5] but is distinctly more radiogenic in ⁸⁷Sr/⁸⁶Sr_{(20 Ma)} than the type 1 xenoliths. Measured Pb isotopes in all three samples have radiogenic ²⁰⁶Pb/²⁰⁴Pb (19.8–20.2), ²⁰⁷Pb/²⁰⁴Pb (15.5–15.6), and ²⁰⁸Pb/²⁰⁴Pb (39.6–40.0) compared to MORB mantle (Stracke et al., 2005) and do not clearly separate type 1 from type 2 peridotites.

Clinopyroxene-orthopyroxene major element exchange temperatures calculated using the Brey and Köhler (1990) calibration at a nominal pressure of 15 kbar and using the cores of the multiple adjacent coarse pyroxene pairs yield equilibrium temperatures of ∼805–1128 °C (Table 1). There is no discernable temperature difference between type 1 or type 2 xenoliths, and varying the pressure by ±5 kbar to reflect the approximate upper and lower limits of the spinel facies in fertile peridotites has only a <20 °C effect on estimates. Temperatures calculated using the pyroxene REE thermometer of Liang et al. (2013) overlap with the two-pyroxene major element geothermometry results for BAL-4, FOU-9, SUM-4, and SUM-12; however, there is some separation in estimates for FOU-15 and SUM-6, with the major element results plus errors being lower than the REE temperatures. Because CaO and MgO concentrations vary in the pyroxenes near rims (Figs. 4 and 5) and the Brey and Köhler (1990) calibration is based on the exchange of Ca, Mg, and Fe, the difference between methods temperatures may be due to the true core of a grain pyroxene not being analyzed as a result of the microscope slide not dissecting the center of the grain. FOU-3 is anomalous in that the two-pyroxene thermometer yields a much higher temperature than the REE thermometer. Because of the evidence for major element zoning, the REE geothermometer results are preferred (814–1086 °C). These results overlap with peridotite equilibration estimates made on other Dunedin Volcanic Group xenoliths documented by Scott et al. (2014a, 2014b) and McCoy-West et al. (2015).

The major, trace element, and isotope data presented here enable a thorough petrological history to be established for upper lithospheric mantle under the Dunedin Volcanic Group. The interpreted mantle composition is summarized in Figure 8 and discussed in the following sections. The data and interpretations are then used to assess the role that the mantle lithosphere may have played in the source of the Dunedin Volcanic Group alkaline magmas.

Olivine Mg#, spinel Cr# (Table 1), and clinopyroxene trace element data (Fig. 6A) show that the type 1 peridotites have undergone melt extraction. The degrees of partial melting were low; most apparently underwent <∼8% melt extraction based upon comparisons of measured the clinopyroxene HREE concentrations and melt extraction models (Fig. 6A). The melt models make no correction for subsolidus trace element exchange, and these are therefore minimum estimates (e.g., McCoy-West et al., 2016). The radiogenic Nd isotopes of the BAL-8 and SUM-11 clinopyroxene separates (Table 2) are indicative of moderately long isolation from convecting mantle and are comparable with several other xenoliths from the Dunedin Volcanic Group classified as type 1 (Fig. 7A).

Single-stage depleted mantle Nd model ages in BAL-8 and SUM-11 are ca. 0.3 Ga (Table 1) and the data plot close to the 0.3 Ga isochron (Fig. 7B). Several published type 1 xenoliths have radiogenic Nd and/or a high ¹⁴⁷Sm/¹⁴⁴Nd and plot between the 0.1 and 0.3 Ga isochrons. However, the likelihood of multiple melt extraction events and/or metasomatism complicates the significance of these simplistic Nd model ages. For example, SUM-11 has a model age of 0.33 Ga but the unradiogenic ⁸⁷Sr/⁸⁶Sr = 0.7020 suggests long-term isolation on a billion-year time scale. Os and Hf isotope data have shown that there are residues of Proterozoic to Archean melting present within the mantle lithosphere under the Dunedin Volcanic Group (McCoy-West et al., 2013; Scott et al., 2014b), although debate remains on whether these data mean that the entire underlying mantle lithosphere, or just small portions of it, is Archean–Proterozoic in age (McCoy-West et al., 2013; Liu et al., 2015). FOU-3 is classified as type 2 on the basis of a slight degree of LREE enrichment, but has a clinopyroxene trace element pattern very similar to type 1 xenoliths (Fig. 6A) and an even more radiogenic Nd isotopic composition [εNd_{(20 Ma)} = + 22.6] than BAL-8 or SUM-11. However, this sample also plots close to a mixing array from a very radiogenic composition and the main cluster of type 2 xenoliths and has a future model age (Table 2). Such combined REE and Nd isotope systematics require multiple stages of modification (depletion and enrichment) of the sample’s Sm/Nd isotope system. Therefore, placing absolute age constraints is difficult and the model age and its position close to an isochron age of 1 Ga are unlikely to be significant.

An important observation in the context of this study is that the type 1 xenoliths are completely distinct from the isotopically homogeneous Dunedin Volcanic Group basalts (Figs. 7A, 7B). This means that mantle lithosphere represented by the type 1 xenoliths, which are moderately fertile and occur in multiple locations through the volcanic province, has not contributed a detectably significant volume to any of the intraplate alkaline magmatism.

The similarities in equilibrium temperatures of the type 1 and type 2 xenoliths (Table 1) indicate that the type 1 peridotites must be embedded within a variably chemically modified mantle (Fig. 8). The type 2 xenoliths have REE systematics consistent with being type 1 protoliths that were subsequently enriched to varying degrees by LREE-bearing fluids (Fig. 6). In the present data set, FOU-9 is an example of a strongly enriched xenolith, whereas FOU-3 is an example of a weakly modified xenolith (Fig. 6A). Furthermore, the most strongly modified xenoliths have similar isotopic Sr and Nd compositions (⁸⁷Sr/⁸⁶Sr_{(20 Ma)} = 0.7028–0.7029; εNd_{(20 Ma)} = + 5.1 to + 5.2; Fig. 7A), whereas the less strongly metasomatized xenoliths plot in arrays that mimic mixing between type 1 peridotite and a metasomatic agent with an isotopic composition similar to that of the host basalts (Fig. 7B). Thus, the trends subparallel to mid-Mesozoic isochrons (Fig. 7B) likely represent mixing between heterogeneous mantle and a homogeneous metasomatic agent (Fig. 7C) and have little or no geochronological significance.

It is not easy to determine the nature of the metasomatic agent that has affected the Bald Hill, Foulden Maar, and Summer Hills peridotites because the evidence for enrichment is commonly restricted to only the LREEs and fluid mobile elements (Figs. 6A–6D). One exception is FOU-9; clinopyroxene HREE concentrations indicate that it was a moderately refractory rock prior to LREE enrichment (Fig. 6A) and thus susceptible to inheriting the trace element and isotopic character of the metasomatic agent. In this sample, the LREE-rich clinopyroxene has Th/U > 4 (both elements are above detection limit) and very low Ti/Eu (493). The clinopyroxene negative Ti anomaly is not offset by a positive orthopyroxene Ti anomaly (Figs. 6C, 6D). These trace element properties are consistent with enrichment by a Ti-free or Ti-poor LREE-rich melt, such as carbonatite or CO₂-bearing silicate melt (e.g., Coltorti et al., 1999; Wittig et al., 2009). Thus, at least for FOU-9, the enrichment agent was chemically and isotopically similar to that inferred for other Otago peridotites (Scott et al., 2014a, 2014b, 2016b; McCoy-West et al., 2015, 2016). However, it is also striking that many of the type 2 xenoliths are isotopically similar to the host intraplate alkaline basalts (Figs. 7A, 7B), and therefore these peridotites could have been metasomatized by an earlier precursor melt or the host melt (Figs. 7B, 7C); the former option is nearly impossible to test because of the moderately fertile nature of the xenoliths prior to enrichment; the latter possibility is explored in the following.

The largest clinopyroxene diameter for a type 2 peridotite is 2400 μm (SUM-15; Fig. 2A) and REE thermometry indicates that HREEs in this sample equilibrated at ∼800 °C (Table 1). At this temperature, the Yb diffusion data of Van Orman et al. (2001) indicate that clinopyroxene homogenization from the core to the rim (1200 μm) would require ∼2 m.y. (Fig. 9A; Table 3). If equilibrium had been attained at 900 °C, then the time required would be 0.096 m.y. Eu has a larger ionic radius and is a slower diffuser than Yb, and appears to have equilibrated between pyroxenes in 6 of the 8 analyzed samples (Fig. 6E). Using an orthopyroxene radius of 3250 μm from SUM-6, which is a type 2 xenolith that equilibrated at 1035 °C, and Eu diffusion data from Cherniak and Liang (2007), the required time to achieve a homogeneous concentration at 1000 and 1100 °C would be ∼5–69 m.y. (Fig. 9A; Table 3); lower REE equilibration temperatures increase the time scale of diffusion dramatically. Because peridotite xenoliths typically remain in magmas for a duration of hours to days (e.g., Spera, 1984; Demouchy et al., 2006), the results of these calculations indicate that the attainment of HREE and MREE equilibrium between pyroxenes in type 2 samples FOU-3, FOU-9, SUM-6, and SUM-12 (Fig. 6E) must have predated xenolith entrainment by a least several million years and probably longer. Because LREEs have larger ionic radii than HREEs, the fact that Ce has equilibrated in pyroxenes in SUM-6 and FOU-9 (Fig. 6E) indicates that an even longer time scale is possible. Metasomatism therefore cannot be due to reaction with the host basalt.

In four phase spinel facies peridotites, diffusive loss of Al, Cr, and Mg in orthopyroxene and clinopyroxene requires exchange with spinel, since olivine only incorporates trace amounts of Al and Cr (e.g., De Hoog et al., 2010). The exchange of Al and Cr between orthopyroxene and spinel is temperature sensitive, with orthopyroxene housing less Al and Cr and spinel housing more as temperature decreases (Witt-Eickschen and Seck, 1991). Data shown in Figures 3A and 4 are therefore consistent with subsolidus exchange in Al and Cr during cooling for all samples except FOU-9. Ca exchange between pyroxenes is also temperature sensitive (Bertrand and Mercier, 1985; Brey and Köhler, 1990). Furthermore, because the pyroxenes are the only major reservoirs for Ca in the peridotite xenoliths studied here, the generally elevated clinopyroxene rim Ca and generally depleted orthopyroxene rim Ca concentrations (Fig. 3B) also likely represent cation exchange during cooling.

Element diffusion data enable calculation of an approximate time scale of pyroxene cooling in the type 1 and type 2 xenoliths. Mg is a fast-diffusing species in orthopyroxene (D₀ = 1.1 × 10⁻⁴ to 4.34 × 10⁻⁹ m²/s; Cherniak and Dimanov, 2010). The diffusion data of Schwandt et al. (1998) indicate that a heterogeneous profile penetrating as much as 150 μm into orthopyroxene, such as shown in SUM-11 in Figure 4, would develop at 800 °C in 0.33–0.82 m.y., or 0.018–0.032 m.y. at 900 °C with the variation at a single temperature dependent on the crystallographic axis along which diffusion occurred (Fig. 9B; Table 3). Note that diffusion rates at 800 °C and 900 °C are modeled because the major elements in many xenoliths equilibrated at ≤950 °C (Table 1). Cr in orthopyroxene is a slower diffusive species than Mg (Cherniak and Dimanov, 2010) and the orthopyroxene Cr diffusion data from Ganguly et al. (2007) indicate that development of a 150 μm profile at 800 °C would require between 1.9–7.8 m.y., or 0.16–0.67 m.y. at 900 °C (Fig. 9B; Table 3). Al diffusion rates in orthopyroxene are not known, but coupled diffusion of Cr and Al is indicated by the similar extents of zoning (Fig. 4A). Ca is a relatively slow diffusive species in clinopyroxene at mantle temperatures (D₀ = 3.2 × 10⁻⁷ to 2.4 × 10⁻¹⁰ m²/s; Zhang et al., 2010). A 150 μm Ca zoning will develop in clinopyroxene in 2–7.1 m.y. at 800 °C and in 0.095–0.47 m.y. at 900 °C, with the variation in time scale once again depending upon the crystallographic axis along which diffusion occurred (Fig. 9C; Table 3). Al diffusion in clinopyroxene is less well experimentally constrained than Ca but thought to be slow (Sautter et al., 1988); a 150 μm profile would develop in ∼16.7 m.y. at 800 °C or 1.2 m.y. at 900 °C (Fig. 9C; Table 3). The calculated time scales, although subject to large uncertainties in diffusion parameters, lead to the conclusion that the mantle cooling recorded in the pyroxene rims occurred on the order of millions of years prior to xenolith entrainment.

As pyroxene zoning is ubiquitous in peridotites exhumed across East Otago in an otherwise heterogeneously depleted and metasomatized mantle (Scott et al., 2014a, 2014b), the process that caused the cooling must have been region-wide. Similar core-to-rim Al gradients in peridotitic pyroxenes have been attributed to lithospheric mantle thermal perturbations caused by continental-scale tectonism; for example, in the case of cooling profiles, mechanisms that have been appealed to are mantle diapirism due to thinning of the lithosphere (e.g., Preβ et al., 1986; Werling and Altherr, 1997; Xu et al., 1998) or slab impingement and upper mantle insulation (Chin et al., 2015). The insulation effects of a slab are not possible for the mantle under the Dunedin Volcanic Group because subduction had ceased ∼80 m.y. before intraplate magmatism. However, since ca. 110–105 Ma, the Zealandia lithosphere has had a dynamic history in which it was thinned as a precursor to opening of the Tasman Sea and Southern Ocean and Zealandia-Gondwana separation (Gibson et al., 1988; Gaina et al., 1998; Gray and Foster, 2004; Scott and Cooper, 2006). Through the Cenozoic, Zealandia gradually became submerged until maximum transgression in the late Oligocene, at which time most, if not all, of the South Island was under the sea (Landis et al., 2008). Just prior to eruption of the late Oligocene–Miocene Dunedin Volcanic Group, large portions of Zealandia emerged from beneath sea level as a result of deformation distributed from the Australia-Pacific plate boundary that was simultaneously forming through the center of the continent. The small-scale uplift, perhaps caused by tectonic-induced lithosphere thickening, may have promoted cooling within the mantle in the Oligocene–Miocene.

Although type 2 peridotites have isotopic compositions the same as or very similar to those of the Dunedin Volcanic Group basalts (Figs. 7A, 7B), the diffusion modeling results are interpreted to show that the host basalt did not impose this isotopic composition. Melting of amphibole in the lithospheric mantle has been argued as a potentially important source component to alkaline basaltic magmas (Class and Goldstein, 1997; Panter et al., 2006; Pilet et al., 2008; Rooney et al., 2014; Pilet, 2015) and has been inferred from basalt data to be in the HIMU source for the New Zealand Cenozoic intraplate magmatism (Finn et al., 2005; Panter et al., 2006; Sprung et al., 2007). Low-degree melting of amphibole plus an enriched but formerly refractory peridotite has also been suggested to be capable of generating a melt with a trace element composition very similar to those of the New Zealand intraplate basalts (Scott et al., 2016b). However, as discussed in Scott et al. (2016b), one of the problems is that typical mantle amphibole (Ti-pargasite) stability is restricted to ≤1100 °C (e.g., Foley, 1991; Green et al., 2014; Mandler and Grove, 2016), significantly cooler than the lithosphere thermal base (∼1350 °C; McKenzie and Bickle, 1988). The main mechanism for expanding the thermal range of amphibole is the replacement of OH by F (Foley, 1991), but most published data on mantle amphiboles for which F has been measured show them to have very low F concentrations (e.g., Bonadiman et al., 2014).

To perturb F-free (or low- F) pargasite that had been stable in the upper lithospheric mantle would require a rise in the mantle geotherm immediately prior to melting or significant decompression. The lack of evidence for extensive lithosphere thinning rules out the latter explanation, and the pyroxene core-rim major element data (Figs. 3 and 4) and the results of the diffusion calculations (Fig. 8) attest to moderately slow cooling having occurred in all samples but FOU-9. However, this sample has the highest pyroxene core equilibration temperatures in the suite (∼1086 °C) and therefore probably comes from the deepest level. Using the diffusion parameters discussed here for a host magma temperature of 1200 °C, the ∼60 μm Ca and Mg profiles in clinopyroxene (Fig. 5) would require between 19 and 889 yr to form, and the ∼60 μm Mg and ∼20 μm Cr profiles in orthopyroxene would require between 1.5 and 56 yr (Table 3), all of which are outside the time scales for residence in the host magma during ascent. Heating of pyroxene rims in FOU-9 could therefore be indicative of a rise in the geotherm, or it could also be due to a localized effect of heating from a stalled batch of magma that crystallized nearby, or it could reflect the imprecision of the diffusion calculations. In either case, FOU-9 must come from above the basalt melt source, which leaves only a very small range for amphibole, if it were present, to have been available to melt.

A feature that indirectly links the upper lithospheric mantle to the intraplate magmatism is highlighted in similarities in the type 2 Sr and Nd isotopic data and the alkaline basalts (Fig. 7A). These data, along with the similarities in Pb isotopes, indicate that the HIMU-like source of the metasomatic fluids that affected portions of the upper mantle was likely related to the source of the HIMU-like basaltic magmas. Mg isotope data from a small suite of Zealandia alkaline basalts and a suite of strongly metasomatized xenoliths that both have HIMU-like Sr-Nd-Pb isotopic compositions reveal distinctly light δ²⁶Mg isotopic values that are indicative of subducted carbonate in their sources (Wang et al., 2016). Therefore, deep melting of oceanic-related lithosphere may have been an important component in the Dunedin Volcanic Group alkaline intraplate basalt as well as the fluids that have enriched the lithospheric mantle column. More Mg isotope work is required to test this theory.

Major element, trace element, and isotopic analyses of peridotite xenolith minerals show the lithospheric mantle under the Dunedin Volcanic Group to be composed of LREE-depleted peridotitic domains (type 1) embedded within variably enriched mantle (type 2).

The type 1 xenoliths tend to have more radiogenic Nd than moderately to strongly metasomatized type 2 xenoliths. No Dunedin Volcanic Group basalt has a comparably radiogenic Nd value, and thus these old yet fertile mantle domains have not contributed significantly, if at all, to the intraplate magmatism. The strongly metasomatized type 2 peridotites xenoliths display HIMU-like Sr, Nd, and Pb isotopic signatures that are similar to, or overlap with, the host basalts. Diffusion calculation results for the time required for pyroxene REEs and MREEs equilibration in type 2 peridotites indicate that metasomatism must have occurred millions of years before xenolith entrainment. Trace element concentrations are interpreted to show that one metasomatic agent was a LREE-rich CO₂-bearing fluid.

All pyroxenes have distinctive Al, Cr, and Mg ± Ca zonation patterns that commonly extend as much as 150 μm from rims. The pyroxene element trends and diffusion rates on all but the inferred deepest xenolith are interpreted to record slow cooling, and therefore that there was evidently little or no change to the upper mantle geotherm under the Dunedin Volcanic Group during intraplate magmatism. The hottest pyroxene core equilibration temperatures in the xenolith suite (∼1086 °C) indicate that there was only a very small depth range over which typical mantle amphibole (thermal stability of ∼1100 °C) could have potentially been a component in the Dunedin Volcanic Group source.

Dalton was supported by an Otago Museum Postgraduate Geology Scholarship and an Australasian Institute of Mining and Mineralogy Education Endowment Trust. Pyroxene trace element and isotope analyses were funded by a Foundation for Research Science and Technology Fellowship (contract UOOX1004) and University of Otago Research Grants to Scott, and a Canada Excellence Research Chairs Government of Canada program and a Canada Foundation for Innovation instrumentation grant to Pearson. We thank Y. Luo for providing assistance with orthopyroxene analyses at the University of Alberta. C. Stirling permitted analysis at the Centre for Trace Element Analysis at Otago. We thank S. Pilet and S. Kidder for comments on a draft, and K. Panter and two anonymous reviewers for thorough and constructive reviews.