On the relative motions of long-lived Pacific mantle plumes

Mantle plumes upwelling beneath moving tectonic plates generate age-progressive chains of volcanos (hotspot chains) used to reconstruct plate motion. However, these hotspots appear to move relative to each other, implying that plumes are not laterally fixed. The lack of age constraints on long-lived, coeval hotspot chains hinders attempts to reconstruct plate motion and quantify relative plume motions. Here we provide 40Ar/39Ar ages for a newly identified long-lived mantle plume, which formed the Rurutu hotspot chain. By comparing the inter-hotspot distances between three Pacific hotspots, we show that Hawaii is unique in its strong, rapid southward motion from 60 to 50 Myrs ago, consistent with paleomagnetic observations. Conversely, the Rurutu and Louisville chains show little motion. Current geodynamic plume motion models can reproduce the first-order motions for these plumes, but only when each plume is rooted in the lowermost mantle.


Supplementary 40 Ar/ 39 Ar Age Determinations Notes
In total 31 age dates for separates are reported here, producing 21 reliable plateau dates representing the eruption ages with 40 Ar/ 39 Ar isochron intercepts within error or very close to atmosphere (295.5), 4 samples containing clear and correctable excess Ar patterns still resulting in acceptable eruption ages, and 6 separates with ages deemed undeterminable. Below is a discussion of the age determinations for lava flows from each seamount with all uncertainties reported at the 2σ confidence level. See the accompanying age results supplement for the incremental heating age and K/Ca spectra diagrams along with inverse isochrons, full sample information and plateau justifications.
The first seamount analyzed is Manu Lele Vai (D02), which contained altered aphanitic basalts and ankaramites with calcite vesicle infilling. Sample D02-04 (74.57 ± 0.28 Ma) produced a relatively short plateau (30% 39 Ar) during the low-mid temperature effects wherein the apparent age becomes increasingly younger, likely due to high temperature recoil. The sample is highly radiogenic and thus the isochron points are clustered resulting in a poor intercept value of 445 ± 145. Sample D02-05 was deemed unreliable due to the high scatter (MSWD > 5). Sample D02-17 produced an interesting spectrum with two potentially short plateaus developed at ~67 and ~57 Ma. Neither plateau produced an atmospheric isochron, and due to the potential for two different age interpretations this sample is declared unreliable.
One aphanitic lava flow from the seamount Taring Nui (D03) was analyzed twice with two different leaching procedures. The first split was leached with the standard one hour acid baths described above and the analysis was deemed unreliable having high atmospheric argon, strong recoil effects and generating two potential plateaus with non-atmospheric intercepts. The second split was leached with two hour HCl steps followed by the standard HNO3 and DI H2O steps. The sample was then picked with much greater rigidity producing only 3 mg of material. The resulting plateau is flat and covers 40% of the released 39 Ar, contains an atmospheric intercept, and a MSWD of 0.75. Thus, this flow provides a single age constraint for this seamount at 61.57 ± 0.5 Ma. One sample from Logotau (D04) contained plagioclase phenocrysts and provided an age of 63.66 ± 0.34 Ma along with an atmospheric intercept and MSWD of 1.3.
Multiple splits from two basalts were analyzed for Tefolaha (D07). Sample D07-09 had a plagioclase analysis with an age of 51.01 ± 0.36 Ma that contained an atmospheric intercept (293.5 ± 9.5) and wide plateau (73% 39 Ar) until higher temperature steps wherein distinct increases in apparent age occurred likely due to the excess 40 Ar released from deep seated inclusions. This sample contained very fresh clinopyroxene so a 25 mg separate was analyzed, providing an age of 49. 25 ± 2.48 Ma that meets all the criteria for a reliable age, but is less precise due to its low potassium concentration. The resulting combined plateau age for D07-09 is 51.00 ± 0.31 Ma. A pyroxene and plagioclase phyric clast from the hyaloclastite sample D07-22 was analyzed three times with two splits of groundmass and a plagioclase separate, all from the same irradiation. The first groundmass separate was accidentally pre-cleaned with a higher power CO2 beam resulting in a large loss of gas from the low temperature heating spectrum. The sample still produced a fairly long (51%) plateau containing an age of 54.65 ± 0.12 Ma with an intercept of 419 ± 184. A second split of the groundmass was analyzed with a reliable but more scattered plateau age of 54.95 ± 0.14 Ma. The highly radiogenic nature of the sample caused all the plateau points to cluster tightly and thus the 40 Ar/ 36 Ar intercept was not obtainable. The plagioclase separate from this sample produced a long (94% 39 Ar) plateau with an atmospheric intercept and younger age of 53.64 ± 0.18 Ma. Since the groundmass did not provide a reasonable intercept for both experiments (492 ± 197 and -221 ± 444). These non-atmospheric intercepts thus provide a misleadingly old age for the groundmass samples and the plagioclase values are used in the models and discussions herein.
A groundmass split from sample D10-04, a small and glassy basalt from Nui (D10), was attempted. The experiment produced an apparently young and humped heating spectrum with no reliable age determinable. Trace amounts of hornblende were separated from the basalt D11-10 (Laupapa Seamount) and two small splits (~1 mg each) were analyzed. Both splits produced long (>98%) plateaus with relatively high uncertainties and neither sample degassed until the higher temperature steps. The stacked plateau results in an age of 52.86 ± 0.77 Ma; MSWD of 0.73; intercept of 286 ± 32. For the seamount Tayasa (D13), a plagioclase and groundmass separate from the sample D13-01 was attempted. The groundmass spectrum produced a continuous recoil pattern with the age starting apparently old and becoming increasing younger until the higher temperature steps wherein small amounts of excess 40 Ar increase the age. The plagioclase separate produced a long and reliable plateau with a corresponding age of 50.52 ± 0.20 Ma.
Two samples were analyzed from the seamount Nukufetau (D14). A groundmass separate was analyzed for sample D14-01, which produced a short (33% 39 Ar) plateau with an atmospheric intercept and MSWD of 0.58 resulting in an age of 43.64 ± 0.57 Ma. A plagioclase and groundmass separate from sample D14-08 were analyzed with a resulting age of 49.82 ± 0.18 Ma and 48.53 ± 0.17 Ma, respectively. The plagioclase heating spectrum from D14-08 contained two clear spikes in apparent age, a common sign of releasing excess Ar from melt/fluid inclusions (see 40 Ar/ 39 Ar Age Results), however the resulting plateau intercept was within error of atmosphere (300 ± 11).
The plateau age for the groundmass split was recalculated with an 40 Ar/ 36 Ar intercept value of 342 ± 4 (2σ; n=24). The justifications and methods employed in this correction are discussed in the method section. The recalibrated groundmass age and plagioclase age (calculated assuming the standard 295.5 40 Ar/ 36 Ar trapped argon ratio) are offset by ~1 Ma. These results are of an obvious concern when attempting to understand the age at which a lava flow erupted upon the seamount.
This discrepancy may be due to some minor excess Ar in the plagioclase phase, which causes an apparent age that is too old.
Two samples from dredge 15 (Vaitupu Seamount) were analyzed. A groundmass separate from the basalt D15-02 produced a long (65% 39 Ar) plateau with a slightly above atmospheric intercept of 359 ± 39 and a resulting age of 49.03 ± 0.19 Ma. Sample D15-12 was analyzed twice due to the same pre-heating mistake that afflicted D07-22b. The first analyses did not produce a useable age determination while the second attempt contained a correctable low temperature excess Ar intercept (360.7 ± 5.8; n=19) and was recalculated to provide an age of 49.58 ± 0.18 Ma.
One plagioclase separate from a Telematua Seamount basalt (D16-35) was analyzed providing an age constraint of 46.63 ± 0.49 Ma (MSWD of 0.26; 40 Ar/ 36 Arint of 232.5 ± 164). Two hornblende separates were attempted from Funafuti (D18) producing concordant long age spectrums, low MSWD's and atmospheric intercepts with plateau ages of 48.88 ± 0.12 Ma (D18-07) and 48.93 ± 0.12 Ma (D18-23). Due to the similar lithologic character and age determinations it is possible that these samples represent two chunks broken off of the same lava flow. One groundmass separate was analyzed from the sample D22-29 (Silaga Seamount) producing a short (33% 39 Ar) but useable age constraint of 46.09 ± 0.28 Ma for this seamount.
Dredges 24 sampled Kosciusko Seamount and recovered numerous small (<25 cm) basaltic clasts within Mn nodules. Two small glassy and fairly fresh looking (no thin section was made) aphyric basalt clasts from D24 were analyzed. Groundmass splits from D24-04 (47.37 ± 0.11 Ma) were calculated using a low-temperature non-atmospheric intercept value of 499 ± 6 (n=24). The groundmass separate from D24-11 also contained a consistent low temperature excess argon pattern and thus the plateau age was recalculated using an intercept of 415 ± 51 producing an age of 48.16 ± 0.19 Ma. Multiple samples from D27 were analyzed with results from two samples reported herein. The majority of samples from this dredge are 11-15 Ma in age and are attributed to Samoan hotspot volcanism. Those results along with more samples not related to the Rurutu hotspot from the RR1310 expedition will be presented in an upcoming study. The two samples presented herein are a hornblende separate from D27-35, which produced a long plateau with a very reliable age of 42.24 ± 0.82 Ma. Both a hornblende and plagioclase separate were attempted from D27-64 with the hornblende producing a long, excellent heating spectrum and a resulting age of 45.15 ± 0.12 Ma. The plagioclase separate consistently increases in apparent age until the highest temperature steps wherein a very short (17.48 39 Ar%) plateau is produced. Thus, an age of 45.73 ± 0.14 Ma for the separate is tentatively presented but is less reliable and the hornblende age is accepted as the more accurate eruption age constraint.

The Source of the Tuvalu Seamounts
The mantle source and origin associated with generating the Tuvalu Seamounts hitherto were poorly constrained with arguments either in favor of a Rurutu Hotspot origin 1 or an extinct hotspot 2 . In an effort to better understand the origin of these seamounts we invoke an 'isotopic finger printing' technique 1,3-6 and a 'backtracking' technique 1, [5][6][7]  (focal zone) 9 composition. It is possible that the HIMU Rurutu hotspot could have provided the melts to these seamounts, as ocean island volcanoes (OIV) typically tend to produce an isotopic mixing trend between one of several possible enriched endmember compositions and FOZO 9,10 .
However, due to the uncertainty of Manu Lele Vai and Silaga being fed by an HIMU mantle plume, these two seamounts conservatively are not used in this study's hotspot track reconstruction models.
To investigate the potential deep mantle sources that fed the Tuvalu volcanoes we first compared the along-track age-distance relationship between the Rurutu-aged HIMU seamounts in the Cook-Australs 11-18 , Tuvalu (this study) and Gilbert Ridge 1,6 and a variety of absolute plate motion (APM) models (Supplementary Figure 2) Sakau (75 Ma) found within the Gilbert Ridge that do not correlate to any known modern hotspots 1,6 . Nuilakita is a complex seamount with evidence for both older French Polynesian volcanism (at 42 Ma) and younger Samoan hotspot volcanism at ca. 14 Ma. Therefore, it is uncertain whether the 42 Ma age represents an episode of volcanism from the Rurutu hotspot or volcanism sourced from a different hotspot and is not considered in the presented models. The combination of plate motion based regressions and isotopic 'finger printing' provide strong evidence for the Rurutu Hotspot being the mantle anomaly which sourced the majority of the HIMU-type Tuvalu seamounts. Figure 1: A bathymetric map of the Tuvalu Seamount region with seamounts analyzed in this study labeled. Bathymetric map was generated using the seamount catalog program 26 , with data collected during the RR1310 expedition (see methods) merged with predicted seafloor bathymetry from Smith and Sandwell 27 .

Supplementary Figure 2:
Along track distance from the Rurutu hotspot (Arago Seamount) as a function of age. Shown in blue are the HIMU seamounts that are interpreted to have been source from the Rurutu hotspot (see text for sources). The individual lines represent different APM models for the Pacific including the fixed hotspot models of Duncan and Clague 19 , Koppers, et al. 20 , and Wessel and Kroenke 2 , the moving hotspot models of Steinberger and Gaina 21 and Doubrovine, et al. 22 , along with the plate circuit model of Raymond, et al. 23 . For the moving hotspot models, the misfit with the older age dates indicates a plume motion with a southeast component, consistent with the modelled motion shown in   Locations represent the inferred center of the seamount. Lead isotopic analyses were collected on whole rock separates using a MC-ICP-MS following methods outlined in Konter and Storm 42 . N = number of Pb isotopic analyses per seamount.