Detailed 40Ar/39Ar Geochronology of the Loyalty and Three Kings Ridges Clarifies the Extent and Sequential Development of Eocene to Miocene Southwest Pacific Remnant Volcanic Arcs

The 2015 VESPA voyage (Volcanic Evolution of South Pacific Arcs) was a seismic and rock dredging expedition to the Loyalty and Three Kings Ridges and South Fiji Basin. In this paper we present 33 40Ar/39Ar, 22 micropaleontological, and two U/Pb ages for igneous and sedimentary rocks from 33 dredge sites in this little‐studied part of the southwest Pacific Ocean. Igneous rocks include basalts, dolerites, basaltic andesites, trachyandesites, and a granite. Successful Ar/Ar dating of altered and/or low‐K basalts was achieved through careful sample selection and processing, detailed petrographic and element mapping of groundmass, and incremental heating experiments on both phenocryst and groundmass separates to interpret the complex spectra produced by samples having multiple K reservoirs. The 40Ar/39Ar ages of most of the sampled lavas, irrespective of composition, are latest Oligocene to earliest Miocene (25–22 Ma); two are Eocene (39–36 Ma). The granite has a U/Pb zircon age of 23.6 ± 0.3 Ma. 40Ar/39Ar lava ages are corroborated by microfossil ages from associated sedimentary rocks. The VESPA lavas are part of a >3,000 km long disrupted belt of Eocene to Miocene subduction‐related volcanic rocks. The belt includes arc rocks in Northland New Zealand, Northland Plateau, Three Kings Ridge, and Loyalty Ridge and, speculatively, D’Entrecasteaux Ridge. This belt is the product of superimposed Eocene and Oligocene‐Miocene remnant volcanic arcs that were stranded along and near the edge of Zealandia while still‐active arc belts migrated east with the Pacific trench.

VESPA was a program of closely spaced dredge sampling, supported by seismic reflection profiling (Mortimer & Patriat, 2016). The work plan took advantage of the fact that the Loyalty and Three Kings Ridges are cut by the Cook Fracture Zone (CFZ), a sinistral strike slip fault of ∼250 km lateral displacement (Herzer et al., 2011;Lapouille, 1977;Recy & Dupont, 1982). Importantly, the 1,000-1,500 m relief of the bathymetric walls of the CFZ allowed the expedition to acquire samples at different stratigraphic depths within the Loyalty and Three Kings Ridges. A total of 43 dredges were made on the voyage, nine on the Norfolk Ridge and 34 in the Loyalty Ridge-Three-Kings Ridge-Cook Fracture Zone area (Mortimer & Patriat, 2016;Mortimer et al., 2021; this paper; Figures 2 and 3).   Agranier et al. (2023). Labels show how the regional terms Loyalty and Three Kings Ridges, and Norfolk Basin (bold italic uppercase text) include smaller basins, troughs, terraces, and ridges (italic lowercase text). (b) More detailed bathymetric map showing sample sites in the Loyalty Ridge work area. Symbols indicate rock type and interpreted petrotectonic setting of lavas as explained later in this paper. Background bathymetry is from GEBCO Compilation Group (2019) supplemented by VESPA multibeam bathymetry tracks (shaded; Mortimer & Patriat, 2016). Tectonic information is from Herzer et al. (2011).

Analytical Methods
Dating by Ar/Ar methods at the University of California Santa Barbara, U.S.A (UCSB) followed procedures described in Gans (1997) and Mortimer et al. (2014). One sample was dated by Ar/Ar methods at the U.S. Geological Survey, Menlo Park (cf. Calvert & Lanphere, 2006). All Ar/Ar ages in this paper are reported relative to a Fish Canyon sanidine age of 28.198 Ma. U/Pb dating of zircon was done by LA-ICP-MS (Laser Ablation-Inductively Coupled Plasma-Mass Spectrometry) methods at the University of Otago, Dunedin, New Zealand following methods described in . A Resonetics RESOlution M-50-LR excimer laser ablation system was used with a spot diameter of 33 μm, 5 Hz repetition rate and 40 s acquisition time. Isotopes were measured on an Agilent 7500cs quadrupole ICP-MS. Fractionation of Pb, Th and U was corrected using the TEMORA-2 zircon standard. TEMORA-2, R33, and NIST glass 610 were measured during analytical sessions and used to calculate correction factors. Foraminifera and radiolarian extraction and examination at GNS Science Lower Hutt, New Zealand followed standard methods (Crundwell et al., 2016). Electron microprobe elemental maps and analyses of the groundmasses of selected dated volcanic rocks were conducted at UCSB. A detailed description of the procedures and approaches used to obtain optimal Ar/Ar results is outlined in Section 5 below. Further details of other analytical methods and equipment used are given in Supporting Information S1. Throughout the paper, all radiometric age errors are quoted as ±2σ and/or 95% confidence level and the timescale of Raine et al. (2015) is used.

General Considerations
Dredged mafic submarine lavas present special problems for Ar/Ar dating. These include absence of stratigraphic control and context, and limited choice of datable material (i.e., rock type, phenocryst content, grain size and freshness). Other problems include the very low K/Ca ratios of plagioclase phenocrysts, low overall K content of the whole rock and common replacement of K-rich groundmass crystalline phases and interstitial glass by secondary minerals such as clays, zeolites, oxides. In rocks that lack a suitable phenocryst phase, the only option is to attempt to date the microcrystalline groundmass. Such fine-grained and composite aggregates of primary and secondary phases commonly yield complex 40 Ar/ 39 Ar age spectra that are difficult to interpret because of varying degrees of hydrothermal alteration, low temperature argon loss, low radiogenic yields, low K contents, and reactor-induced recoil of 39 Ar, 37 Ca, and 36 Ar. In subaerial basalts these problems are minimized by sampling the most slowly cooled holocrystalline interior of thick lava flows that show the least amount of alteration (e.g., Gans, 1997). With dredged rocks, dating must be attempted on a very limited selection of less-than-ideal samples such as the quenched and vesicular upper portions of lavas, and/or samples that often show substantial clay alteration.
The principles of interpreting 40 Ar/ 39 Ar step heating results to obtain rock ages have been investigated for more than 50 years (e.g., Dalrymple & Lanphere, 1974;Turner, 1971), and dating principles are explained in textbooks such as McDougall and Harrison (1999) and Dickin (2005). A review paper by Schaen et al. (2021) addressed technical matters of irradiation, standards, monitors, and statistical treatment of data, but paid less attention to sample selection strategies, screening and sample preparation, and interpretation of "disturbed" age spectra. Many studies have noted that incremental heating 40 Ar/ 39 Ar analyses from whole rock or groundmass separates of fine-grained mafic rocks often do not yield straightforward plateaus (e.g., Fleck et al., 2014;Gans, 1997;Iwata & Kaneoka, 2000). While plausible explanations and recommendations for interpreting these "disturbed" spectra have been offered, rarely is a rigorous assessment made as to how potassium is actually distributed between various primary and secondary groundmass phases and how this might influence the resultant age interpretation.
Natural samples rarely behave according to theory and are often difficult to interpret, especially when dealing with many of the complexities outlined above. In this paper, we address problems with dating the very fine-grained to glassy groundmass of relatively low K and variably altered basalts and basaltic andesites.

Analytical Approach
Hand specimens of all VESPA lavas were examined on board ship to identify those possibly suitable for argon geochronology. Post-cruise petrographic inspection eliminated samples that were too altered to consider dating, for example, where matrix was mostly replaced by secondary minerals and where plagioclase was substantially altered. Petrographic examination at high optical magnifications (20X and 40X objectives; Supporting Information S2) allowed visual estimates of what groundmass phases (e.g., plagioclase, pyroxene, glass, oxides, secondary clays, etc.) were present and their approximate proportions.
Rocks that passed these initial screenings underwent phenocryst and groundmass separation. Pure separates of plagioclase and other phenocryst phases were obtained from sieved fractions (125-250 μm) using standard density and magnetic mineral separation techniques described in Mortimer et al. (2014); plagioclase phenocrysts were handpicked and had no adhering groundmass material. Groundmass concentrates from 125 to 250 μm sieve fractions were obtained by light magnetic separations which selectively removed phenocrysts and secondary phases. The groundmass concentrates then underwent intensive ultrasonic cleaning for several hours in repeated changes of deionized water to remove secondary clay and zeolite. No samples were treated with acid, as we have found from past experience that this tends to greatly reduce radiogenic yields by increasing atmospheric contamination due to increased surface area; clays, zeolite and carbonate can be satisfactorily removed by magnetic separation and ultrasonic cleaning alone. Importantly, whenever possible, both plagioclase phenocrysts and groundmass from the same sample were separated for dating to provide a mutual crosscheck of the Ar/Ar results. Our previous experience (e.g., Mortimer et al., 1998Mortimer et al., , 2007Mortimer et al., , 2014Mortimer et al., , 2018Mortimer et al., , 2021 has demonstrated that it is not always possible to predict in advance whether groundmass or phenocrysts would yield the best result. The 40 Ar/ 39 Ar ages were obtained by incremental heating experiments on both phenocryst and groundmass separates. There was a wide range in complexity of step heating spectra, reflecting the multiple issues mentioned in Section 5.1 above. Some plagioclase phenocrysts had very low K/Ca <0.005 and low signal/noise ratios; these required a large correction for the interfering isotopes of Ca-derived 36 Ar and 39 Ar. Normally, such material would be regarded as unsuitable for Ar/Ar dating. However, by taking great care to monitor uncertainties in system blanks, reactor constants (especially for Ca), mass discrimination and tailing corrections, meaningful (albeit imprecise) ages were obtained from these low-K plagioclases. Many incremental heating experiments use a 500-1300°C temperature range. Our experience in dating groundmass concentrates from mafic lavas has consistently shown that the loosely held argon released at the lowest temperatures (350-550°C) and the highly retentive argon released at the highest temperatures (≥1150°C) is both analytically problematic and are almost never used in age interpretations (e.g., Nauert & Gans, 1994). Specifically, the low T gas yields apparent ages that can either be too old (due to recoil) or too young (due to Ar loss from non-retentive phases) and include hydrocarbons that contaminate the mass spectrometer. Gas released at the highest T steps commonly also yields ages that can be either too young (due to recoil) or too old (excess Ar trapped in the most retentive phases). In the latter case, steps are associated with elevated 37 Ar signals and a substantial drop in apparent K/Ca ratios due to the degassing of highly retentive calcic phases (pyroxenes and the cores of plagioclase crystals). All the step heating experiments on groundmass reported here commenced with a 30-min degassing of each sample at ∼500°C, followed by 9-10 incremental heating steps spanning 600-1100°C, and a final high temperature degassing of any residual gas. Thus, our approach was to focus on the middle T fraction of the gas that is the largest volume and analytically most meaningful.
As our VESPA Ar/Ar dating work proceeded, it was clear that different groundmass separates were yielding very different results in terms of the quality of the data as revealed by the shape and complexity of the resultant age spectra. We set out to better understand this variability by defining the different K reservoirs in primary and secondary groundmass phases. To do this, we conducted a detailed electron microprobe investigation of eight samples representing a range of petrographic textures and argon spectrum types (Table 1, Figure 4, Supporting Information S3 and S4). Quantitative X-ray 10 element (K, Na, Ca, Fe, Mg, Ti, Al, Si, Cr, Mn) maps of representative 600 × 600 μm groundmass areas supplemented by quantitative point analyses of the principal K-bearing domains for each of these sample provided detailed information on the proportions of groundmass phases present, and their precise chemical composition.

Groundmass Petrography
The groundmasses of dated samples range from mostly glassy to holocrystalline and display varying degrees of alteration (Supporting Information S5). Primary groundmass crystalline phases nearly always include plagioclase, clinopyroxene, and iron titanium oxides, and may also include minor amounts of olivine, orthopyroxene, and late crystallized sanidine. Interstitial glass may be entirely fresh, but more commonly is partially or completely altered to fibrous aggregates of Fe-and K-rich clays, Fe-oxides, and zeolites ( Figure 4). These primary and secondary phases occur in highly variable proportions for different samples, ranging from holocrystalline coarse-grained (75-200 μm) interlocking fresh crystal aggregates of plagioclase-pyroxene-olivine in a diabasic texture, to highly vesicular and glassy groundmass samples containing less than 25% microlites of flow-aligned acicular plagioclase (25-75 μm long and 4-10 µm wide) and tiny blebs of pyroxene and oxide in a matrix of glass or its alteration products. Phenocryst phases include plagioclase ± clinopyroxene ± olivine and account for anywhere from a few percent to 30% of total rock volume.

Groundmass K Reservoirs From Element Mapping
Any interpretation of 40 Ar/ 39 Ar ages implicitly assumes a knowledge of exactly what phase or phases are being dated. This is self-evident for pure phenocryst separates, but far less so for the various fine-grained polycrystalline aggregates that make up the groundmass of mafic lavas. We used detailed electron microprobe studies to assess exactly which phases contribute to the 40 Ar/ 39 Ar ages we obtain from groundmass samples.
Our element maps and microprobe analyses of eight representative groundmass samples from VESPA lavas were used to calculate how K and Ca are distributed in different groundmass phases (Table 2, Supporting Information S4). The primary K reservoirs in all samples are plagioclase, interstitial glass and late crystallizing sanidine. Secondary K reservoirs include clay and zeolite. The distribution of K varies enormously between samples (Table 2). In well-crystallized groundmass samples, a large percentage (35%-85%) of the total K resides in groundmass plagioclase (and in sanidine in more potassic rocks), whereas in the quenched lavas, 80%-90% of the K resides either in interstitial glass or its clay and zeolite alteration products. Groundmass plagioclase in all samples is inevitably zoned from more calcic (An 50-75 ) cores to more sodic (An 30-50 ) rims, a zonation that is accompanied by a marked increase in wt% K and K/Ca ratios. Groundmass plagioclase cores may contain as little as 0.02 wt% K with K/Ca ratios as low as 0.001, to as much as 0.5 wt% K (with K/Ca ∼0.06). Similarly, K contents of plagioclase rims range from c. 0.1 to 1.0 wt% K, with K/Ca ratios ranging from 0.005 to 0.15. Although the K-rich rims are often quite thin (1-5 μm), they typically account for the majority (50%-80%) of the total K residing in plagioclase. Interstitial glass and the high-K clays that replace them have significantly higher K contents (1.0-2.5 wt%) and K/Ca ratios of 0.5-2.0. Sanidine (and some high K zeolites) have even higher K contents (3-9 wt%) and K/Ca ratios of 3-10. It is important to keep in mind that significant fractions of the loosely held clays and zeolites noted in Table 2 were likely removed during ultrasonic cleaning of the groundmass separates, thus removing a portion of the original K. Therefore, the quantitative K-budgets in the separates that were subsequently irradiated and dated may be different from what was present in the analyzed sample.

Types of Groundmass Ar/Ar Spectra and Their Interpretation
A cursory inspection of all age spectra from groundmass separates of the dredge samples collected in this study (Supporting Information S6), as well as from our previous studies of southwest Pacific submarine lavas (e.g., Mortimer et al., 1998Mortimer et al., , 2007Mortimer et al., , 2021 reveals that they tend to fall into five distinct spectral types ( Figure 4). In the best-cases, incremental heating experiments of groundmass yields reasonably flat age spectra (Type 1 in Figure 4), where most (>70%) of the gas released defines a plateau of successive steps whose age errors are within two sigma of each other, resembling what is commonly obtained from separates of fresh K-bearing phenocryst phases. In Type 2 spectra, apparent ages start relatively high, steadily decrease over the first 50%-80% of the gas released, and then plummet downward to much younger ages at the highest temperature steps. Type 3 spectra have ages that initially step down over the first 10%-30% of the gas released, flatten into a well-defined plateau in the central part of the spectrum, and then step down again at the highest temperatures. Type 4 and Type 5 are both hump-shaped spectra ( Figure 4), with ages that initially climb, reach a maximum in the central part of the age spectrum, and then step down at the highest temperatures, the main difference being that Type 4 are somewhat broader humps with a well-defined flat "mini-plateau" at the top of the hump. Each of these different spectrum types commonly has a characteristic associated K/Ca distribution (Figure 4), and the deviations from a simple flat age spectrum for Types 2-5 can be explained in terms of varying contributions from reactor induced recoil, partial argon loss in the natural environment, and a trapped "excess argon" component. Inverse isochron plots help with age interpretation as "excess Ar" steps lie well off isochron lines.
Recoil is the redistribution of 39 Ar and 37 Ar because of the neutron flux during irradiation and is important only on very small length scales (≤1-2 μm) (e.g., Fleck et al., 2014;McDougall & Harrison, 1999). During irradiation some fraction of the 39 Ar atoms may be dislodged from high K sites (e.g., K-rich rims of plagioclase or K-rich clays and zeolites), losing some to the atmosphere and embedding others in lower K sites (e.g., in the interiors of plagioclase crystals or in adjacent refractory non-K bearing phases). A similar recoil redistribution is expected for Ca-derived 37 Ar. The predictable consequence of recoil is that the earliest released argon (from the smallest grains and outer rims of larger grains) will have elevated 40 Ar/ 39 Ar ratios and thus older apparent ages, whereas the argon released at highest temperatures will yield anomalously younger ages due to the presence of excess 39 Ar. Similarly, in very calcic phases, recoil of 37 Ar may yield slightly lower 37 Ar/ 39 Ar (apparent K/Ca ratios), at low T balanced by slightly elevated K/Ca at high T in the spectrum. On inverse isochron plots, 39 Ar recoil often yields loops instead of lines.
Argon loss may be the consequence of diffusive loss of the 40 Ar daughter product from non-retentive phases (e.g., glass) or the replacement of primary igneous phases by the new growth of distinctly younger secondary K-bearing phases-in both cases yielding younger apparent ages (e.g., McDougall & Harrison, 1999). These younger ages Note. Values in representative 600 × 600 μm groundmass areas are normalized to 100% non-potassic phase and void-free. Based on electron microprobe element mapping (Supporting Information S3).
will be most evident in the gas released early in the incremental heating experiment that is coming from the outer portions of primary phases and the generally non-retentive secondary phases. Excess argon is the ambient argon that is trapped during initial crystal growth having an initial 40 Ar/ 36 Ar ratio that is distinctly higher than normal atmospheric argon. This yields anomalously old ages and is most evident in gas released at the highest temperatures from highly retentive and very low K/Ca sites such as the cores of plagioclase phenocrysts or pyroxenes.
None of the above complexities are mutually exclusive. For example, K-rich rims on plagioclase, tiny patches of interstitial sanidine and glass, and secondary K-rich clays and zeolites would be susceptible to both argon loss and reactor induced recoil, such that the low temperature steps might be either anomalously old (recoil dominant) or young (argon loss dominant). Similarly, ages at the highest temperature steps might either increase (excess argon) or decrease (recoil), depending on which is dominant. Notably, much of the very loosely held argon situated in K-rich clays, zeolites, and/or micron-scale blebs and rims of sanidine and glass was not analyzed in our incremental heating experiments as it was pumped away during the initial degassing. This low temperature argon is of little use, being the most severely affected by recoil and argon loss, and often contaminated by excessive adsorbed atmospheric argon (low radiogenic yield) and hydrocarbons whose masses overlap the target argon isotopes. Indeed, one of the principal advantages to obtaining 40 Ar/ 39 Ar ages from groundmass separates using well-calibrated incremental heating is that it allows us to selectively degas and analyze different argon domains within each sample and focus our analytical efforts on those primary K reservoirs and phases that are least affected by some of the interferences outlined above.

Type 1 Spectrum-Flat With a Simple Plateau
Type 1 groundmass samples like DR13Aii (P84621) (Figure 4a) generally have elevated K/Ca ratios and high analytical precision such that there is little ambiguity in their age assignment. This sample contains a very fresh, mostly crystalline groundmass composed of interlocking 25-75 μm crystals of plagioclase (30% An 47-52 ), clinopyroxene (7%), and Fe-Ti oxide (4%) with abundant interstitial sanidine (45%) in patches up to 40 μm across. Minor phases include pale brown glass (10%) and traces of biotite and calcite (<2%). Of the total potassium present, c. 87% is held in sanidine (Table 2) with K/Ca ∼10, the remainder is in groundmass plagioclase and glass. The groundmass argon spectrum ( Figure 4d) yields slightly older ages in the lowest temperature steps due to the recoil loss of 39 Ar from small non-retentive glass domains, a broad central plateau associated with high K/Ca ratios during the release of argon from both interstitial sanidine and plagioclase, and a slight drop in ages associated with lower K/Ca ratios at the highest temperatures due to recoil addition of 39 Ar to the more retentive plagioclase cores and pyroxene. The straightforward weighted mean plateau age (WMPA) (McDougall & Harrison, 1999) calculated from most of the argon reflects the crystallization age of the groundmass-in particular, the age of the fresh interstitial sanidine that holds most of the K. Note that the groundmass in this sample yields a more precise estimate of the age (23.62 ± 0.10 Ma) than the phenocrystic plagioclase (24.07 ± 0.18 Ma-Supporting Information S5 and S6) because of the inherently better precision associated with the higher K content and much lower Ca correction.

Type 2 Spectrum-Progressively Decreasing Ages
Type 2 spectra such as that of DR10A (P84607) groundmass tend to show radiogenic argon release at lower temperatures than most other groundmass samples and display a "descending staircase" type spectrum that starts relatively flat and descends steeply at higher temperatures ( Figure 4h). DR10A is a porphyritic basalt containing ∼25% phenocrysts of plagioclase, clinopyroxene, and completely iddingsitized olivine. The groundmass consists of 28% fresh flow-aligned plagioclase crystals and a few % each of pyroxene, olivine, and Fe-Ti oxide crystals in a fresh glassy matrix (Figure 4e). In addition, there is ∼4% of an unidentified late-stage high K phase that is either sanidine or K-rich secondary clay/zeolite that occurs as tiny interstitial blebs only a few microns in diameter. The plagioclase laths are only 5-20 μm wide and are zoned from calcic cores to sodic rims. More than two-thirds of the total K is held in the groundmass glass, with the remainder mostly split between the late K-rich alteration phase and the thin sodic rims on plagioclase (Table 2, Figure 4f).
Apparent ages of the low temperature gas steps are about 24 Ma. They initially decrease gradually but drop abruptly at higher temperatures to as young as 10 Ma. The age decrease is matched by decreasing K/Ca ratios from 1.6 to 0.06. The relatively flat low to middle T part of the spectrum yields a weighted mean age of 23.2 ± 0.5 Ma that is within error of the much better defined WMPA of 23.60 ± 0.24 Ma obtained from phenocrystic plagioclase. The early analytical steps (600-750°C, K/Ca = 0.2-1.5) mainly represent degassing of fresh groundmass glass and plagioclase rims. Higher temperature steps record an increasing contribution from plagioclase (K/Ca = 0.06-0.2) with apparent ages lowered by excess 39 Ar, possibly due to recoil from adjacent high K glass sites.
The age spectrum is a classic recoil pattern, with ages that step down over the initial 30% of the gas released, flatten through the middle 50%, and descend again at high T. The apparent K/Ca ratios are relatively constant, but climb from 0.35 to 0.65, and then drop abruptly at the high T. Much of the argon related to the 1-2 μm sanidine blebs was likely pumped away during initial degassing (the K/Ca is never high enough to be explained solely by sanidine) and most of the spectrum reflects simultaneous degassing of glass and plagioclase. Older but decreasing ages early in the incremental heating are a consequence of recoil induced 39 Ar loss from outer rims of plagioclase. The central flat portion of the spectrum yields a weighted mean age of 24.78 ± 0.10 (not a statistical plateau) representing 44% of the total gas released from plagioclase rims and cores with perhaps a minor contribution from glass. Decreasing ages in the highest T steps reflect degassing of more retentive plagioclase cores, whose ages have been reduced by the addition of excess 39 Ar. The modest climbing of apparent K/Ca over the middle part of the spectrum likely represents recoil induced redistribution of 37 Ar-especially in plagioclase. Though this sample is very fine-grained with a significant glass component, a reasonably good age was obtained because of the high overall K content, and the fact that K was distributed between the viable reservoirs of relatively high-K plagioclase and fresh glass.
It is likely that much of the argon associated with the relatively high K, but non-retentive, glass and clay were removed during ultrasonic cleaning and initial sample degassing. The analyzed groundmass separate yields a broad hump-shaped spectrum, with a flat "mini-plateau" at the top of the hump, followed by rapidly decreasing ages at the highest temperatures ( Figure 4p). Apparent K/Ca ratios steadily drop from 0.15 at the lowest T step to a central portion where they level at ∼0.045, before dropping to 0.005 at the highest T. The early climbing part of the spectrum records degassing of glass/clays and plagioclase rims, with an increasing contribution from the latter. The young apparent ages from these reservoirs probably reflect both argon loss and recoil, but the former dominates. The best estimate of the sample age (22.5 ± 0.5 Ma) comes from the flat top of the hump reflecting gas derived primarily from both plagioclase rims and cores. The drop in ages at highest temperatures could reflect recoil induced excess 39 Ar added to low K plagioclase cores. Despite extensive recoil and argon loss, a meaningful age was obtained from this sample because a sufficient percentage of the K resides in the groundmass plagioclase, and the incremental experiment was able to effectively isolate and analyze the gas coming from the high K rims of this plagioclase.
The very strongly hump-shaped spectrum climbs from 10 Ma to a maximum of 18.6 Ma and then descends again at higher temperatures, with no two contiguous steps yielding concordant ages (Figure 4t). A low precision, but reasonably well-behaved separate of plagioclase phenocrysts from the same sample yielded a WMPA of 22.2 ± 0.5 Ma, which we take as the best estimate of the age for this sample. This is significantly older than even the oldest apparent age from the top of the hump from the groundmass. The shapes of the groundmass age and K/ Ca spectra suggest that the early climbing part of the spectrum reflects simultaneous degassing of residual clay and glass (both of which experienced significant argon loss) with an ever-increasing contribution from the rims of groundmass plagioclase. The oldest ages were from the central part of the spectrum (43%-81% of the cumulative 39 Ar released, at 750-850°C) and is likely coming from both plagioclase cores and rims given the uniform K/ Ca values of ∼0.06. However, these ages are still too young to be primary because of the continued contribution from the K-rich non-retentive phases that experienced argon loss. At the highest temperatures, ages drop precipitously as do the K/Ca ratios, reflecting the degassing of refractory plagioclase cores and pyroxene, with their attendant excess 39 Ar induced by recoil. At first glance, this sample seems like it would be an excellent candidate for dating (coarse-grained, fresh primary igneous phases) but it is quite problematic because it has a very low overall K content, and much of the K that is present is held in non-retentive fine-grained clays that, despite ultrasound treatment, seemingly were not removed during sample preparation.

Groundmass Age Spectra Interpretations: Key Points
The best groundmass spectra, Type 1, are obtained from fresh, mostly crystalline groundmass where most of the K is situated in groundmass feldspar. It helps if the rock is a reasonably high-K whole rock composition to begin with. The next best, Type 2, spectra are obtained from samples that have a large percentage of fresh groundmass glass and may yield meaningful ages from the early part of the spectrum. Type 3 spectra are obtained from samples with minimal clay alteration and where >30% of the total K resides in groundmass plagioclase. The central flat portion of the spectrum is a good measure of the age and reflects primarily the outgassing of plagioclase high K rims. Type 4 spectra are produced by samples that invariably have some amount of clay alteration which hold a significant fraction of the total K budget. Provided a sufficiently large percentage (>30%) of the total K is held in groundmass glass, and sufficient care is taken to isolate this gas fraction during incremental heating, the central mini-plateau at the top of the hump-shaped Type 4 spectra may provide a reasonably good estimate of the age. Type 5 spectra are produced by samples that are more severely altered and that were low K to begin with, such that nearly all the K resides in very fine-grained aggregates of secondary clays, zeolites, and perhaps some residual glass. These samples are usually quite quenched (i.e., originally had a large fraction of glass), and that glass is now mostly altered. In such samples, severe recoil and argon loss compounded by large Ca corrections conspire to yield strongly hump shaped spectra with no concordant ages, such that the only interpretation that can usually be made is that the top of the hump likely represents a minimum age for the sample.
Taken together, these observations suggest that acceptable ages can be obtained from groundmass separates of reasonably fresh mafic volcanic rocks provided more than at least ∼30% of the total K resides in plagioclase or sanidine. Progressively better data quality and higher precisions are obtained as the overall K-content, crystallinity and grain size of the groundmass increases and the amount of secondary clay/zeolite alteration after glass decreases. The very best results, significantly better than those obtained from associated plagioclase phenocrysts, are obtained from alkaline basalts with fresh holocrystalline groundmasses.

Sample Types
The igneous rocks that were dated in this study range from basalt to granite, but most are mafic to intermediate volcanic rocks (Table 1; Figures 2 and 3; Mortimer & Patriat, 2016). Primary igneous textures are generally well-preserved, and feldspar and pyroxene are generally fresh, but olivine is pervasively iddingsitized and groundmass glass commonly exhibits some degree of secondary alteration (Supporting Information S5). For detailed treatment of the geochemical and isotopic composition, igneous petrogenesis, and additional petrographic details of the VESPA rocks see Agranier et al. (2023) and geochemical data at https://doi.org/10.17882/90050.
Using trace element concentrations and isotopic ratios Agranier et al. (2023) divided the VESPA lavas into three geochemical groups, based on their relative trace element concentrations, and whether they had negative, flat or positive slopes in their chondrite-normalized rare-earth element (REE) patterns: (a) depleted tholeiites (negatively sloping REE patterns); (b) transitional tholeiites (flat REE patterns); (c) enriched lavas (positively sloping REE patterns), including andesites, trachyandesites and trachytes. These three groups also form distinctive Sr, Nd, Hf, and Pb isotopic arrays (Agranier et al., 2023).
The enriched geochemical group includes 14 dated lavas, mostly basaltic trachyandesites and trachyandesites with lesser basalts and trachytes (Table 1). These lavas commonly contain plagioclase and greenish clinopyroxene phenocrysts and are commonly vesicular. Phenocrysts of biotite and/or hornblende are prominent in a few of intermediate composition (e.g., DR41Aii, DR15B, DR26Aiv). The enriched group lavas were mainly dredged from tops of seamounts and ridges on the Loyalty and Three Kings Ridges and from the NW Cook Fracture Zone scarp, in relatively shallow (∼800-2,400 m) water depths (Figures 2 and 3).
The transitional and depleted tholeiite geochemical groups include 19 dated samples. These are aphyric to porphyritic, sometimes vesicular, basalts and basaltic andesites (Table 1). Some samples with doleritic groundmass and devitrified glass mesostasis (e.g., DR25Cii, DR39C) are probably from the centers of lava flows. Others have abundant glass and may be broken pillow rinds (Mortimer & Patriat, 2016). Phenocryst and groundmass phases are primarily plagioclase and colorless clinopyroxene, though sparse olivine is present in at least one sample (DR33Ai). Depleted tholeiites were mostly dredged from 1,400 to 4,100 m water depths on fault scarps and volcanic flanks from along the Cook Fracture Zone, Cagou Trough and/or abyssal plain of the South Fiji Basin. Included in this depleted group is a rhyolite from Bougainville Guyot that was originally recorded as a chert (Collot et al., 1992).
Three dm-sized pieces of altered biotite granite porphyry were dredged from 2,400 m water depth on the western Three Kings Ridge (DR26). Normative calculations of two DR26 analyses confirm that the rock name granite is correct (Streckeisen & Le Maitre, 1979). Other rocks in the dredge included trachyandesite and altered tuffisites (Mortimer & Patriat, 2016). Plagioclase and biotite phenocrysts and a holocrystalline groundmass in the granite suggest a hypabyssal setting.
The threefold geochemical and isotopic subdivision from Agranier et al. (2023) is illustrated in Figure 5. We use this figure to further interpret the dated lavas in terms of their possible tectonic setting of eruption. The fourfold petrotectonic interpretation presented in this paper enables the dated VESPA rocks to be directly compared with earlier southwest Pacific compilations (Mortimer et al., 2007(Mortimer et al., , 2021Mortimer & Scott, 2020). The separate geochemical and petrotectonic classifications of samples are given in Table 1. The same information is presented with geochemical analyses in the main supplemental data file of Agranier et al. (2023). In Figure 5, two geochemically enriched lavas lie on the mid-ocean ridge basalt to ocean island basalt (MORB-OIB) mantle array and the rest lie well above it. The two basalts on the MORB-OIB array are DR08A and DR22A and we interpret them as intraplate alkali basalts of the kind that are found scattered across the southwest Pacific region (Mortimer & Scott, 2020). Most of the VESPA light REE-enriched lavas lie above the mantle array. Their bulk compositions are typical of high-K and shoshonitic suites similar to those found on South Fiji Basin seamounts (Mortimer et al., 1998(Mortimer et al., , 2007(Mortimer et al., , 2022. The DR26 granite plots with the high K-shoshonitic lavas and we regard it as a plutonic member of this same suite. The transitional tholeiites of Agranier et al. (2023) with their flat REE patterns, also scatter along and above the MORB-OIB mantle array. They, too, can be divided into two categories but less clearly than the enriched lavas. We interpret those on or close to the mantle array as being back-arc basin-like and those well above the mantle array as having possible subduction-related arc affinity ( Figure 5). There is an acknowledged large overlap in geochemistry between global arc and back-arc basin lavas (e.g., Shervais, 2022), and this is especially the case in the southwest Pacific when high-Ti Early Arc Tholeiites from Fiji and rear-arc lavas from the Colville Ridge are included (Timm et al., 2019;Todd et al., 2010). Thus, our twofold (red and blue dot symbol) interpretation of these widespread basalts and basaltic andesites is likely to be a simplification. Overall, Figure 5 confirms general subduction-influenced compositions for many VESPA tholeiites including arc, rear-arc, and back-arc basin compositions and (for the Cook Fracture Zone) possibly leaky transform magmatism.
The most depleted back-arc basin-like tholeiites were obtained at VESPA dredge site 39. DR17A is notable because it was recovered from the deepest part of the northwest Cook Fracture Zone (CFZ, 3,500 m) and is a clast from a tectonic breccia in which surfaces have a chloritic polish and slickensides; it is the only sample to contain a structural overprint from the CFZ. Shown for comparison in Figure 5 are other back-arc basin basalts from pre-VESPA sampling in the area of Figures 1-3: the FAUST2-D3 and TAN9912-DR9 dredges, and core from DSDP (Deep Sea Drilling Project) Sites 205 and 285 (Bernardel et al., 2002;Mortimer et al., 2007).

Ar/Ar Ages
Excellent to acceptable Ar/Ar ages were obtained from most VESPA samples. A summary of the preferred Ar/ Ar age for each of the 33 dated samples is presented in Table 1, and associated age spectra are shown in Figure 6. A tabulated summary of results for all 49 dated mineral and groundmass separates is given in Supporting Information S5 and all raw data and spectra in Supporting Information S6. The methods used to obtain reported ages and age uncertainties vary. For samples that yield straightforward plateaus, where well over 50% of the 39 Ar is released in consecutive steps that are concordant (i.e., lie within 2σ analytical uncertainty of each other), we follow the widely accepted convention (e.g., McDougall & Harrison, 1999) of calculating a weighted mean plateau age (WMPA), with each step weighted according to its analytical uncertainty, and calculate an uncertainty in the age (2σ, 95% confidence) from the weighted mean uncertainty of all steps used in the calculation. For spectra where there are analytically concordant adjacent steps that represent less than 50% of the gas released, and for samples where there are no consecutive concordant steps, but 30%-80% of the age spectrum still define a visually flattish segment, we calculate a weighted mean age. For calculating errors for these "non-plateau" samples, we followed Fleck et al. (2014) in multiplying the propagated errors by the square root of the mean square of weighted deviates (MSWD) in cases where MSWD >1.
Having described our interpretive approach to groundmass argon spectra in terms of five types (Figure 4), we now present the VESPA 40 Ar/ 39 Ar results in terms of four quality ranks (A-D) based on a qualitative assessment of our overall confidence in the assigned age and uncertainty for each sample (Table 1, Figure 6). In cases where multiple phases from the same sample were analyzed, the quality rank for samples typically is based on whichever phase yielded the better result. In some cases, reinforcing results from multiple phases (e.g., plagioclase and groundmass) might, together, raise the overall rank of that sample. Quality rank A, B, and C samples all yielded 40 Ar/ 39 Ar ages that are acceptable and useful but have generally increasing age uncertainties. Quality rank D samples typically yielded age spectra that were so disturbed and/or had such vanishingly low K contents and K/Ca ratios that our age assignments and their uncertainties are more akin to educated guesses. A summary of the results from each group is as follows.  QUALITY RANK A samples yielded the most precise Ar/Ar ages (Table 1, top two rows of Figure 6, Supporting Information S5). These are mainly intermediate to silicic volcanic rocks and the assigned age was obtained from either a phenocryst phase (biotite, plagioclase, or hornblende) or a fresh, mostly crystalline, and relatively high-K groundmass, or both. They all belong to the enriched, high K-shoshonitic group identified above and have whole rock K 2 O > 1.2 wt%. Quality Rank A samples are the freshest, least-altered rocks in the sample set. Groundmass samples of this group yield mainly Type 1 spectra. Multiple phases were dated from all but one of the samples, and consistently gave ages that were well within analytical uncertainty of each other and of the integrated total fusion age ( Figure 6, Supporting Information S5 and S6). WMPAs were calculated from 65% to 100% of the gas released and had high (65%-98%) radiogenic yields. Plagioclase phenocrysts have K/Ca = 0.05-0.10 and groundmass separates K/Ca = 0.1-2.0. Even though some bulk groundmass analyses have only moderate K/Ca, most contain late interstitial sanidine which dominates the spectrum behavior (e.g., Figure 4a). All the sample ages of Quality Rank A are closely clustered at approximately 24 Ma with uncertainties of ±0.10-0.26 Ma.
QUALITY RANK B samples also gave precise Ar/Ar ages (Table 1, rows 3-5 of Figure 6), but with somewhat less precise analytical uncertainty. These samples comprise six basalts, two basaltic andesites, a basaltic trachyandesite, and a rhyolite from which we dated seven groundmass separates, six plagioclase separates, and one biotite separate. Geochemically, both enriched and transitional igneous suites are represented. Whole rock K 2 O contents vary from 0.3 to 2.5 wt%. Groundmass/plagioclase pairs were dated for four of these samples, with more precise ages obtained by groundmass for two and plagioclase for two. All age pairs agree except for P84671 the groundmass of which displays extensive argon loss. Plagioclase phenocryst K/Ca ratios were somewhat lower (0.01-0.04) than Quality Rank A samples but yielded reasonably flat spectra with well-defined WMPA from 65% to 100% of the gas released. Groundmass separates produced mainly Type 1 and 3 spectra. Despite having low K 2 O (Agranier et al., 2023), some of the basalts are doleritic and their holocrystalline groundmasses are plagioclase-rich and amenable to dating (Figure 4). Overall, Quality Rank B samples have slightly more secondary alteration than Quality Rank A, especially as clay minerals replace groundmass glass. Six out of 10 samples have plateaus that meet statistical criteria of McDougall and Harrison (1999). The Quality Rank B samples are 26-22 Ma in age except for the geographically distant SEAPSO sample which gave ages of 40.6 ± 0.4 Ma (groundmass) and 41.3 ± 1.2 Ma (plagioclase). This makes it one of the oldest Eocene arc lavas with one of the highest precision ages thus far obtained from the entire southwest Pacific.
QUALITY RANK C samples include nine basalts, one basaltic andesite and one basaltic trachyandesite (Table 1, rows 6-8 of Figure 6). All lack K-rich phenocrysts. Whole rock K 2 O ranges from 0.1 to 1.1 wt% and includes both enriched and depleted suites (Agranier et al., 2023), but no high K-shoshonitic lavas. Interpreted ages are based on five groundmass separates from aphyric lavas, two plagioclase separates from samples with groundmass that was too altered to date, and four samples with plagioclase-groundmass pairs. Groundmass age spectra are mostly of Type 4 and 5, with one each of Type 2 and 3. Most yielded interpretable (but low precision) ages from the flattish central parts of spectra, based on <50% of the gas released. The K/Ca ratios of groundmass separates is highly variable (0.02-0.8), and most of the complexity in the spectra reflects varying degrees of recoil and argon associated with the presence of appreciable glass and/or clay alteration, as discussed in Section 5 above. Plagioclase phenocrysts yielded fairly flat (Type 1) spectra for 80%-100% of the gas released but have large uncertainties due to the small signal sizes, low K contents, very low K/Ca ratios (0.002-0.005) and associated large Ca corrections. Uncertainties in the ages of Quality Rank C samples are significantly higher than Rank A and B, ranging from ±0.3-2.2 Ma, (mostly ± 1.2-1.5 Ma), reflecting their lower K contents and greater degree of alteration.
QUALITY RANK D samples (bottom row of Figure 6) are the most mineralogically altered and have abundant clays and zeolites in groundmass. Three samples are depleted olivine basalts that lack plagioclase phenocrysts, the other is a clay-altered hyaloclastite breccia (DR38A). Whole rock K 2 O of the four samples is 0.2-0.8 wt% and groundmass K/Ca is notably low-generally <0.05. Most results likely do not represent primary crystallization ages but we include them in this paper because they are an instructive end member in the range of Ar/Ar behavior of submarine basalts. At best, only minimum ages can be inferred from most of these strongly hump-shaped Type 5 groundmass spectra, which arise from substantial hydrothermal alteration and recoil issues. An imprecise age of 39 ± 1 Ma from the doleritic groundmass plagioclase of DR25Cii is given by two concordant steps which make up 45% of the gas, but this age is highly uncertain given the overall complexity of the spectrum.
The most notable aspect of our VESPA Ar/Ar dating is that 25 of the 33 samples have estimated ages that are tightly clustered between 25 and 22 Ma that is, they straddle the Oligocene-Miocene boundary (Figure 7). This is irrespective of their geochemical composition, map location, or dredge depth. Four samples gave ages that are slightly younger than this range, but three of them are minima and thus their eruptive ages could also be in the range 25-22 Ma. One DR43 lava-from the Norfolk Ridge-gave an age of 25.9 Ma. The remaining three lavas in our new Ar/Ar data set have ages between 41 and 36 Ma (Middle to Late Eocene); one of these is from Bougainville Guyot which, like DR43, is outside the main VESPA work area (Figures 1-3). Within the 25-22 Ma cluster, the enriched high-K lavas are, overall, slightly older than the more depleted lavas. The geological significance and tectonic implications of the results for the wider southwest Pacific are discussed at more length in Section 7.

U-Pb Zircon Geochronology
Zircons from two samples, biotite granite porphyry DR26Fiii and sandstone DR15Cii, were dated by the U-Pb method (Figure 8). Cathodoluminescence images and raw U-Pb isotope data tables are given in Supporting Information S5. Zircons extracted from VESPA DR26Fiii are up to 600 × 150 μm in size, euhedral, and show oscillatory igneous zoning. Of the 30 zircons dated, 16 were <10% discordant and are plotted in Figure 8b. All but the youngest define a single age population of 23.6 ± 0.3 Ma which we interpret as the intrusion age of the granite porphyry. This is within error of the 23.9 ± 0.1 Ma Ar/Ar age of enriched lava DR26Aiv from the same dredge. Abundant biotite phenocrysts in the DR26Fiii granite porphyry suggests that it is related to the biotite-bearing "enriched" (shoshonitic) group of lavas, as does its whole rock geochemistry (Agranier et al., 2023). Despite targeting zircon cores, no older ages, for example, suggestive of Mesozoic basement, were found. This implies either an absence of Zealandia continental crust under the DR26 dredge site, that granite magmagenesis didn't involve continental crust, or that zircons were fully reset or recrystallized such that evidence of any older inheritance was not preserved.  (Bernardel et al., 2002;Edwards, 1973;Mortimer et al., 1998Mortimer et al., , 2007Mortimer et al., , 2014. These, and the VESPA ages, confirm a younger spreading age range than that based on magnetic anomalies (Davey, 1982;Malahoff et al., 1982;Sdrolias et al., 2003;Watts et al., 1977). In both panels, where circles have no error bars, the 2σ age error lies within the size of the symbol.
VESPA dredge 15 contained c. 600 kg of clastic sedimentary rock pieces. Lithologic contacts between conglomerate, pebbly sandstone, sandstone, siltstone, and mudstone rock types in many of the dredged pieces suggests that we sampled across a well-bedded sedimentary formation (Mortimer & Patriat, 2016). The main rock type is a polymict bouldery to pebbly matrix-rich sandstone with a few thin (<5 cm) medium-grained sandstone to mudstone beds (VESPA DR15Cii). The cobble-and boulder-sized clasts are all volcanic (DR15A, B, Ci; Table 1) and one very-well rounded, 10 cm diameter cobble of silicified limestone (DR15F) was noted. Granule-sized clasts in a thin section of DR15Cii (Figure 8c) are mostly varitextured lava and mudstone, with subordinate clasts of serpentinite, chert, marble, limestone, gabbro with saussuritized feldspars, dolerite, variolitic basalt, amphibole gneiss, and discrete grains of clear monocrystalline quartz, clear (volcanic) feldspar, exsolved (plutonic) clinopyroxene, and pale tremolitic amphibole. Thus, the sandstone appears to have a threefold provenance: chilled volcanic rocks, limestones, and an altered ophiolite sequence. The most obvious candidate for an ophiolitic source area is an offshore correlative of the Peridotite and Poya nappes of the New Caledonia Subduction-Obduction Assemblage (Maurizot, Robineau, et al., 2020). Such rocks have been traced to near the Cook Fracture Zone (Patriat et al., 2018).
Zircons extracted from the sandstone matrix of DR15Cii are up to 400 × 100 μm in size, euhedral, and show oscillatory igneous zoning. None are rounded. Of the 87 zircons dated, 35 were <10% discordant and are plotted in Figure 8d. The majority (28 grains) define a peak at 23.4 ± 0.3 Ma that is very close to the 24.2-24.1 Ma Ar/Ar ages of three dated volcanic clasts from the same sandstone. Older zircon grain ages are 43, 99, 100, 108, 131, 188, and 192 Ma. The Cretaceous and Jurassic ages match those of detrital zircons in New Caledonia and New Zealand Eastern Province basement rocks (e.g., Adams et al., 2009;Campbell et al., 2018). As such they indicate erosion of similar basement terranes, which are expected to lie along the north-eastern edge of Zealandia.

Micropaleontology
A total of 63 limestone and mudstone samples from 23 VESPA dredges were processed and examined for their microfossil content. Of these, 32 yielded no fauna or gave non-determinate ages, and nine gave Plio-Pleistocene ages. The remaining 22 samples from 13 dredge sites yielded pre-Pliocene ages summarized in Table 3 and Figure 7 (these exclude the Norfolk Ridge seamount samples already reported in Mortimer et al. [2021]). Full details of the micropaleontology of the VESPA samples are presented in Supporting Information S8 (Crundwell et al., 2016).
Ten micropaleontologically dated mudstone and limestone samples are in contact with volcanic or volcaniclastic material for example, as crack infills or micrite breccia matrix (Table 3) Paleo-water depths for the VESPA samples (Table 3) are usually minima and poorly constrained due a paucity of calibrated benthic foraminifera depth markers. However, most interpreted paleo-depths broadly match the water depths at which the samples have been dredged. Shelf restricted taxa and faunas in two samples (DR26Gi and DR42Ci) indicate these sample sites have subsided in situ since deposition or the samples have slumped off bathymetric highs and have been redeposited in deeper water.

Best Ar/Ar Practice and Interpretation
In this study, we have improved knowledge of the interpretation of complex Ar/Ar spectra of groundmass separates from mafic volcanic rocks. This has been achieved by integrating detailed petrographic, microprobe, and geochronological data. What has emerged is a much better understanding of the causes of the different argon spectrum shapes and some insights on how best to interpret these disturbed (non-ideal) age spectra to assign a reasonable age. When trying to obtain 40 Ar/ 39 Ar ages from groundmass separates of mafic volcanic rocks, we recommend the following: • If possible, sample holocrystalline interiors of thick lava flows rather than chilled margins. • Conduct careful petrographic examination to identify both phenocryst and groundmass phases and to enable crushing of the sample to a size range that ensures capture of pure groundmass particles and inclusion-free portions of plagioclase phenocrysts, typically 100-250 μm. • Perform several hours of heavy-duty ultrasonic cleaning in repeated baths of deionized water to remove as much secondary clay as possible. • Date both groundmass and phenocrystic plagioclase separates. This allows independent verification of the assessed age. • Conduct quantitative microprobe analyses and create element concentration maps of representative groundmass areas on a polished thin section. This will reveal where potassium is residing in the groundmass and how it is apportioned between different phases. • In degassing experiments of groundmass separates, distribute the gas evenly between at least 9-10 temperature steps during incremental heating. The gas released between the 600-1100°C steps is the most informative. • Interpret age spectra using their shapes, associated K/Ca spectra, degassing behavior, and acquired knowledge of the phases present and K distribution. A reasonable age interpretation can then often be made even for very complicated spectra, where only a small fraction of the gas is used in the age determination.
We recognize that many of these procedures, especially the first three, are now routinely followed by investigators obtaining 40 Ar/ 39 Ar ages from mafic lavas (e.g., Calvert & Lanphere, 2006;Fleck et al., 2014;Leonard et al., 2017), but most of these investigations dated very fresh, relatively young onland basalts where subaerial exposures provide the luxury of selecting the best material possible to date. Here we emphasize the importance of thoroughly investigating the detailed petrography and mineral/glass chemistry of the different groundmass phases for samples that are glassy and/or altered, in order to establish precisely what phases are contributing to the radiogenic argon budget and to better interpret complex age spectra that might otherwise be deemed uninterpretable.

Geology of the Loyalty Ridge-Three Kings Ridge Area
The Loyalty-Three Kings Ridge area was formerly one of the least-known parts of the southwest Pacific. Because of this paper, it is now one of the most densely sampled and dated. Our new dating results are shown on local seismic profile cross sections (Figure 9) and on a regional southwest Pacific map ( Figure 10). The northern and southern cross sections of Figure 9 have the main axes of the Loyalty Ridge and Three Kings Ridge aligned by restoring c. 250 km of Cook Fracture Zone (CFZ) movement between them Herzer et al., 2011). Such a restoration also suggests a pre-CFZ alignment-and therefore correlation-of other features such as South Loyalty Basin = Three Kings Terrace, Félicité Ridge = un-named basement high of the western Three Kings Terrace and, more speculatively, Kwênyii Basin = Philip Trough. The distinctly deep Cagou Trough, which cuts the Philip Trough structure (Figure 3a; DiCaprio et al., 2009;Sdrolias et al., 2004) does not appear to have an obvious counterpart north of the Cook Fracture Zone. Two disadvantages of rock dredging are that sampling is done along a line rather than at a point, and that individual blocks potentially have moved downslope from their in situ locations. Nonetheless, the cross sections of Figure 9 reveal that, with a few (slumped?) exceptions, the high K-shoshonitic lavas are always found on bathymetric highs (800-2,400 m water depth), and most back-arc basin-like lavas are found in bathymetric lows including the Cook Fracture Zone (3,500-4,200 m water depth). The dredging of Eocene as opposed to Oligocene-Miocene arc lavas at DR25, along with the granite at DR26, can be explained by their low structural-stratigraphic position exhumed in the footwall block of a major normal fault on the eastern side of the Cagou Trough.
The most significant result arising from our dating work reported in this paper is recognition of a voluminous and exceptionally widespread pulse of latest Oligocene to earliest Miocene (25-22 Ma) igneous activity within, north, and south of the Cook Fracture Zone between latitudes 25°S and 22°S (Figures 9-11). An episode of Middle to Late Eocene volcanism also seems to be documented but is less well-constrained because of fewer dated rocks both in the VESPA work area and elsewhere. The Middle to Late Eocene DR25 lavas were recovered from the unnamed basement high of the western Three Kings Terrace that is, west of the main Loyalty-Three Kings ridge axis sensu stricto. Early Oligocene DR15C limestones were also recovered in a relatively westerly location from the Félicité Ridge. The DR40 Late Cretaceous mudstone, DR15 detrital zircons of Mesozoic age, and DR15 detrital serpentinite suggest that the ultimate geological basement to the ridges in the VESPA work area could be continental Zealandia. Note. See Supporting Information S8 (Crundwell et al., 2016) for complete micropaleontological report. Stages from Raine et al. (2015). Taxa are foraminifera except for sample DR40F (radiolarians).  Ramsay et al. (1997). TWT = two-way travel time. Patriat et al. (2018) gave interpretations of VESPA seismic lines north of these.
The results greatly add to the picture of southwest Pacific arc volcanism on the Loyalty and Three Kings Ridges.

Age of South Fiji Basin Spreading
For many years the age of spreading of the abyssal South Fiji Basin was cited as 31-25 Ma (Oligocene) based on an interpretation of magnetic anomalies 12-7 (Davey, 1982;Malahoff et al., 1982;Sdrolias et al., 2003;Watts et al., 1977). This age is shown as the gray bar in Figure 7b. In contrast, Ar/Ar dating of five back-arc basin basalts by Bernardel et al. (2002) and Mortimer et al. (2007), including at DSDP Sites 205 and 285 (Figure 7b), suggested a younger age for spreading, and Herzer et al. (2011) showed that the magnetic reversal patterns could also be fitted to Early Miocene anomalies 9-6B. Our VESPA expedition increased the number of direct age determinations on South Fiji and Norfolk back-arc basin basalts from five to 14.
The oldest confirmed age of South Fiji Basin spreading is provided by a high-quality Ar/Ar plagioclase age of 26 ± 1 Ma, from DSDP Site 205 (Mortimer et al., 2007). DSDP Site 205 is situated toward the eastern edge of the basin (Figures 1 and 10) in what are interpreted as older, but not the oldest, spreading anomalies (Herzer et al., 2011). Most of the VESPA back-arc basin basalt ages from near the Cook Fracture Zone (CFZ) for example, DR21, DR30, DR36, TAN9912-DR9 are much younger than 26 Ma (Figure 7a) and, sensibly, lie closer to the fossil Central Minerva-Cook spreading axis (Figures 3a and 10). Five back-arc basin basalts dredged from the Cook Fracture Zone are all c. 23 Ma in age and probably represent leaky transform volcanism between South Fiji and Norfolk Basin spreading centers. Abyssal hill fabric and the negligible change in water depth between the basins indicate crust of similar age on either side of the CFZ (Herzer et al., 2011). A reasonably precise minimum age on CFZ movement and, therefore on South Fiji-Norfolk Basin spreading, is provided by the DR22A seamount dated at 19.7 ± 0.5 Ma. This elongate intraplate volcano protrudes 8 km across the CFZ fault trace Figure 11. Age profile along the >3,000 km-long western volcanic belt between New Zealand and the D'Entrecasteaux Ridge. X-axis position has been calculated as along-strike distance from Auckland, taking into account offset along the Vening Meinesz Fault Zone and Cook Fracture Zone. Ages of individual dredge samples have two sigma error bars; ages for Northland Arc volcanoes are ranges. The VESPA samples, either side of the Cook Fracture Zone, fill a critical data gap. On available information, both the age of inception and the age of cessation of currently interpreted arc volcanism apparently decrease to the south. The ages of the immediately adjacent basins follow a similar trend. Data from same sources as Figure 10. and therefore seals it (Figure 3a). The 14 Ar/Ar ages and their error envelopes thus support a 27-19 Ma age for South Fiji Basin spreading rather than c. 31-25 Ma based on magnetic anomalies (Figure 7b). However, there is space in the northern South Fiji Basin for crust of pre-27 Ma age (Herzer et al., 2011). Late Oligocene-Early Miocene extensional tectonics was not just confined to the South Fiji and Norfolk Basins. Horsts, graben, and tilted terraces are present in adjacent arc and continental crust, notably as South Loyalty and Kwênyii Basins, and Philip and Cagou Troughs (Figures 2, 3, and 9; Patriat et al., 2018).

Western Belt of Remnant Arcs
The broader context of Oligocene-Miocene subduction-related lavas on the southern Loyalty Ridge and northern Three Kings Ridge is illustrated in Figures 10 and 11. Based on age and composition, Mortimer et al. (2007) made the case that the now spatially separate Miocene volcanic arcs of onland Northland, the Northland Plateau and the Three Kings Ridge were formerly contiguous and had been split apart by the dextral Vening Meinesz Fault Zone (Figure 10). There is now sufficient geological evidence for us to postulate the existence of a >3,000 km long Cenozoic volcanic belt between Northland New Zealand and Bougainville Guyot (Figures 1,  10, and 11). Tectonic continuity along similarly long southwest Pacific baselines had been previously postulated in syntheses by Lapouille (1977), Launay et al. (1982), and Recy and Dupont (1982) based on bathymetric and magnetic expression, and by Crawford et al. (2003), Schellart et al. (2006), Whattam et al. (2008), andMcCarthy et al. (2022) using bathymetry and scattered ophiolite occurrences to infer locations of associated subduction zones. Our work is the first to confirm the extent of a belt of Cenozoic volcanic rocks as shown in Figure 10. We are most confident about its continuity between Northland and the Loyalty Ridge and less confident about it on the D'Entrecasteaux Ridge.
As shown in Figures 10 and 11, the age range of subduction-related lavas in the main VESPA work area is 39-22 Ma (this study); that on the D'Entrecasteaux Ridge, including Bougainville Guyot, is 40-31 Ma but with large errors (Mortimer et al., 2014; this study). We have not attempted to document volcanic arc rocks farther west toward Papua New Guinea (but see Whattam et al., 2008). The main gap in our knowledge remains the northern Loyalty Ridge. There is an along-strike distance of ∼1,300 km between our VESPA Loyalty Ridge work area and the next Cenozoic arc-related volcano on the D'Entrecasteaux Ridge with only intraplate basalt sites in-between (Figures 10 and 11).
Most of the >3,000 km Northland-Loyalty-D'Entrecasteaux volcanic belt likely formed where it currently sitsalong the north-eastern continent-ocean boundary of Zealandia. It represents the westernmost manifestation of southwest Pacific Cenozoic subduction-related volcanism. The belt's rearward arc position is supported by the relatively high Nb/Yb relative to Th/Yb of the shoshonites; the VESPA lavas are not volcanic front lavas which would be rich in fluid mobile elements and poor in high field strength elements. Existing ages and our new ages from the western volcanic belt possibly suggest two pulses of subduction-related volcanism c. 40-30 Ma and 25-16 Ma (orange areas in Figure 11), separated by a 30-25 Ma hiatus. In other words, lavas from two episodes of arc volcanism seem to be superimposed in the one belt of volcanic rocks. The small number of Eocene samples so far obtained from the southwest Pacific is a natural consequence of most sampling being by rock dredging. Despite our efforts on the VESPA voyage to sample the lower part of fault scarps, dredge-based sampling will always be biased to the youngest eruptive episodes, particularly given low erosion rates in submarine settings. The c. 40 Ma start of arc volcanism over the >3,000 km length of the western volcanic belt can be correlated with regional Eocene deformation which has been interpreted to record southwest-dipping Pacific subduction initiation (Sutherland et al., 2017Stratford et al., 2022; see also van de Lagemaat et al., 2022). The <40 Ma arc footprint of the western volcanic belt postdates 56-40 Ma events in New Caledonia including ophiolite sole amphibolite formation, dikes with arc chemistry and blueschist facies metamorphism . These older events have been linked to initiation and jamming of a north-east dipping subduction zone culminating in Peridotite Nappe obduction Meffre et al., 2012;Schellart et al., 2006;Whattam, 2009), and which may be unrelated to <40 Ma tectonic geometries.
We see some analogies of the development of our western and eastern volcanic belts split by the South Fiji Basin, with the situation in the Cenozoic western Pacific. There, the remnant western Kyushu-Palau arc was split from the eastern Izu-Bonin-Mariana arc by the Shikoku and Parece Vela Basins; shoshonites are also present in the Parece Vela Basin (Ishizuka et al., 2010(Ishizuka et al., , 2011. Pacific-Australian convergence vectors in north-east Zealandia changed little in the interval 45-0 Ma, apart from a slight acceleration and small change in azimuth at c. 27 Ma (Bache et al., 2012). Possible causes of the apparent pulse of volcanic arc activity at c. 25-22 Ma could include changing nature of the subducting slab, maturation of the subduction zone, onset of intra-arc extension facilitating magmatic flux, backarc basin opening, and/or collisional tectonics.

Eastern Belt of Arcs
The back-arc basins of the southwest Pacific attest to substantial trench rollback (Karig, 1971). Eocene and Miocene arcs were variously split by back-arc, intra-arc, and fore-arc rifting mechanisms ( Figure 6 of Stern [2006]), some rifts developed into wide back-arc basins. The back-arc basin opening shifted segments of Eocene, Oligocene, and/or Miocene arc lavas laterally from the edge of Zealandia to locations in Vanuatu, Fiji, Tonga, and the Tonga forearc ( Figure 10; Bloomer et al., 1994;Duncan et al., 1985;Hathway & Colley, 1994;Macfarlane et al., 1988;Marien et al., 2022;Meffre et al., 2012;Rickard & Williams, 2013;Rodda, 1994;Timm et al., 2019) and, more speculatively, the Vitiaz Trench Lineament and Kermadec forearc (Bassett et al., 2016;Pelletier & Auzende, 1996). Under sampled Eocene to Miocene arc rocks in the eastern belt now form the basement to the more prominent and better-studied Pliocene-Quaternary arcs ( Figure 10). Since the Eocene, active arc volcanism has always lain close to the east-migrating Pacific trench (and, since the Late Miocene subduction polarity flip, the southward migrating Vanuatu Trench).

Southwest Pacific Shoshonitic Volcanism
High K-shoshonitic lavas are not only present on the crests of Loyalty and Three Kings Ridges but are also present elsewhere along and across strike ( Figure 10). Their ages are very well-constrained due to their high K contents. In the VESPA work area, they are consistently 24-23 Ma in age. To the south, 21-20 Ma shoshonite lavas have been dredged from six seamounts in the southern South Fiji Basin and as pebbles on the Three Kings Terrace (Mortimer et al., 1998(Mortimer et al., , 2007(Mortimer et al., , 2022. To the north, two shoshonitic lavas and a pebble, dated at 25-23 Ma, have been dredged from the Félicité Ridge and Pines Ridge south of New Caledonia (Mortimer et al., 2014). By analogy with rift-related Pliocene shoshonites in Fiji (Gill & Whelan, 1989) and arc extension elsewhere (Ishizuka et al., 2010), the Oligocene-Miocene shoshonites in and near the VESPA work area have been related to rapid arc rifting. The scenario is that the shoshonites erupted in and near a waning western volcanic belt above a deep Pacific slab as the South Fiji Basin opened rapidly and the Pacific trench migrated rapidly east (Agranier et al., 2023;Mortimer et al., 2007Mortimer et al., , 2021. This widespread southwest Pacific Early Miocene rift setting fits with the relatively high Nb/Yb of the shoshonites ( Figure 5), subtle indications of Miocene extension in and near New Caledonia (Chardon & Chevillotte, 2006;Lagabrielle et al., 2005), Oligocene-Miocene extension in the Kwênyii and South Loyalty Basins (Patriat et al., 2018), our interpretation of the Cagou Trough as an extensional graben (Figure 9), and Miocene movement on the Vening Meinesz Fracture Zone (Herzer et al., 2009).
Using the arc distance from Auckland as a baseline ( Figure 10), a regression line through the shoshonite ages reveals a southward-younging trend of ∼240 km/m.y. within a relatively short ∼5 m.y. interval ( Figure 11). This may represent a southward-younging trend in the rifting-a long-baseline "unzipping" of the basins (Herzer et al., 2011). Interestingly, the shoshonite regression line in Figure 11 projects toward the age and position of the two granite stocks in New Caledonia (the only known Oligocene igneous rocks there). Mortimer et al. (2014) argued that they were petrogenetically related to the Félicité Ridge and Pines Ridge shoshonites and were related to incipient post-obduction extension (cf., Cluzel et al. [2005] and Sevin et al. [2020] who offered an alternative explanation connecting the granites to east-dipping Oligocene subduction).

Conclusions
This paper has presented the results of detailed Ar/Ar dating of rocks from the 2015 VESPA cruise, supported by U/Pb and micropaleontological dating. The study area of the Three Kings Ridge, Loyalty Ridge, and nearby South Fiji Basin has now gone from one of the least sampled parts of the southwest Pacific to one of the most densely sampled. Challenges in 40 Ar/ 39 Ar dating of submarine basalts can be addressed by appropriate sample preparation techniques, thorough petrographic and microprobe characterization of groundmass phases, carefully designed incremental heating experiments, and by dating groundmass and mineral pairs whenever possible. These protocols allow informed interpretation of "disturbed" Ar/Ar age spectra in terms of varying amounts of argon loss, reactor induced recoil, and excess argon. An understanding of these complexities and especially of