Crystallization of Superfast‐Spreading Oceanic Crust (ODP Hole 1256D, Pacific Ocean): Constraints From Zircon Geochronology

Studies of oceanic crust, which covers a large proportion of the Earth's surface, have provided significant insight into the dynamics of crustal accretion processes at mid‐ocean ridges. It is now recognized that the nature of oceanic crust varies fundamentally as a function of spreading rate. Ocean Drilling Program (ODP) Hole 1256D (eastern Pacific Ocean) was drilled into the crust formed at a superfast spreading rate, and hence represents a crustal end member. Drilling recovered a section through lava and sheeted dykes and into the plutonic sequence, the study of which has yielded abundant insight into magmatic and hydrothermal processes operating at high spreading rates. Here, we present zircon U‐Pb dates for Hole 1256D, which constrain the age of the section, as well as the duration of crustal accretion. We find that the main pulse of zircon crystallization within plutonic rocks occurred at 15.19 Ma, consistent with magnetic anomalies, and lasted tens of thousands of years. During this episode, the main plutonic body intruded, and partial melts of the base of the sheeted dykes crystallized. One sample appears to postdate this episode by up to 0.25 Myr, and may be an off‐axis intrusion. Overall, the duration of crustal accretion was tens to several hundreds of thousands of years, similar to that found at the fast‐spreading East Pacific Rise and the slow‐spreading Mid‐Atlantic Ridge. This indicates that crustal accretion along slow‐ to superfast‐spreading ridges occurs over similar time scales, with substantially longer periods of accretion occurring at ultraslow‐spreading ridges characterized by thick lithosphere.

Zircon geochronology provides an important-yet underutilized-tool to study the dynamics of crustal accretion at mid-ocean ridges.Through U-Pb and U-Th dating, zircon can provide the absolute dates of crystallization of igneous units, enabling reconstructions of a number of important variables and processes.The first variable is the spreading rate (C.J. Lissenberg et al., 2009).The spreading rate can be reconstructed where precise zircon U-Pb dates are obtained in a section across-axis.This approach was first applied to the Vema Lithospheric Section (11°N, Mid-Atlantic Ridge), where a linear age progression away from the spreading axis revealed a spreading rate of ∼16 mm/yr, in agreement with magnetic anomaly derived plate motion estimates (C.J. Lissenberg et al., 2009).
The second variable is the time span of accretion of the plutonic section (Baines et al., 2009;Grimes et al., 2008; C. J. Lissenberg et al., 2009;Rioux et al., 2016;Rioux, Lissenberg, et al., 2012;Schwartz et al., 2005).Previous studies have shown that the duration of crustal accretion is preserved on the grain scale, the sample scale, as well as the crustal scale.Grain-scale information on the duration of crustal accretion occurs in the form of inherited cores overgrown by younger rims, which show age differences as much as ∼1.5 Ma at the slow-to ultraslow-spreading Southwest Indian Ridge (Schwartz et al., 2005).On a sample scale, different zircon grains from the same sample do not always define a uniform age population, but instead may record hundreds of thousands of years of zircon growth (C.J. Lissenberg et al., 2009;Rioux, Lissenberg, et al., 2012).On a lower crustal scale, age differences of the order of hundreds of thousands of years have been documented between different intrusive series in the (I) ODP drill core from both Atlantic (Grimes et al., 2008) and Indian (Rioux et al., 2016) oceanic core complexes.Overall, zircon U-Pb data have demonstrated that crustal accretion is more protracted than previously assumed; even at the East Pacific Rise, where the relatively fast spreading rate would be expected to lead to rapid accretion, significant intra-sample age spread in zircon dates was recorded (Rioux, Lissenberg, et al., 2012).To date, no straightforward correlation appears to exist between the duration of magmatism and spreading rate.
The third variable that U-Pb zircon dating can resolve is the dynamics of oceanic detachment faults.These detachment faults are now recognized to play a significant role in accretion along slow-and ultraslow-spreading ridges (Cannat et al., 2006;Escartin et al., 2008;Smith et al., 2006).Zircon U-Pb dates can enable reconstructions of the rates of slip on the detachments (Baines et al., 2008;Grimes et al., 2008).This has shown that slip along the detachment faults may accommodate the majority of plate separation during periods of detachment faulting, leading to significant asymmetric spreading (Baines et al., 2008;Grimes et al., 2008).The dates can also be used to determine the depth of crystallization of gabbroic rocks in the footwall of the detachments (C.J. Lissenberg et al., 2016).Furthermore, when combined with zircon (U-Th)/He dates, zircon U-Pb dates enable the depth and thermal structure of the detachment faults to be reconstructed (Grimes et al., 2011;Schoolmeesters et al., 2012).
Integrating zircon U-Pb or U-Th dates with dates for systems with lower closure temperatures (e.g., zircon (U-Th)/He and 40 Ar/ 39 Ar of associated phases) has enabled reconstructions of the cooling histories for mid-ocean magmatic systems.This has been applied successfully to both plutonic sections (John et al., 2004;Schwartz et al., 2009) and volcanic rocks (Schmitt et al., 2011).
Finally, U-Pb dating of igneous and hydrothermal zircons have been used to determine the age of both magmatic crystallization and hydrothermal alteration of gabbroic rocks at the Vema Lithospheric Section (Mid-Atlantic Ridge; Rioux et al., 2015) and at Atlantis Bank (Southwest Indian Ridge; Schwartz et al., 2010), providing absolute time constraints on alternating cycles of magmatism and hydrothermal fluid flow at mid-ocean ridges.
Combined, studies using zircon geochronology of the oceanic crust have thus led to considerable new insights into the dynamics of oceanic crustal accretion.Here, we present zircon U-Pb dates of plutonic rocks from superfast-spreading crust recovered in Ocean Drilling Program (ODP) Hole 1256D (eastern Pacific Ocean), with the aim of determining its absolute age as well as the time span and dynamics of its accretion.We show that crustal 10.1029/2023GC010964 3 of 16 accretion was largely completed by 15.19 Ma, but that magmatism may have continued for up to 0.25 Myr.Combined with data from across the spreading rate spectrum, this indicates that the duration of crustal accretion appears to be similar at all spreading rates, with the exception of ultraslow-spreading rates, where accretion may take significantly longer.

ODP Hole 1256D
Hole 1256D was drilled in oceanic crust formed during superfast spreading along the Pacific-Cocos plate boundary in the eastern Pacific (Figure 1).This superfast-spreading episode occurred in the middle Miocene (∼18-11 Ma), with magnetic anomalies indicating spreading rates as high as 220 mm/yr (Wilson, 1996;Wilson et al., 2006).This is the fastest spreading rate on record; hence, the crust formed during this interval represents an end member configuration of oceanic crust.A section of this crust was drilled over four expeditions of the (Integrated) Ocean Drilling Program ((I)ODP): Expeditions 206, 309, 312, and335 (Teagle et al., 2006, 2012;Wilson, Teagle, et al., 2003).The underlying rationale relied on the observation that the depth to the lower crustal magmatic system correlates inversely with the spreading rate (Carbotte et al., 1998;Phipps Morgan & Chen, 1993;Purdy et al., 1992); hence, superfast-spreading crust should have a thin upper crust, enabling drilling through a full upper crustal section and into in situ lower oceanic crust (Wilson et al., 2006).
The uppermost basement rocks recovered from Hole 1256D comprise a 284 m thick off-axis lava sequence of sheet flows and pillows, which overlies the axial volcanic sequence (470 m thick) of predominantly sheet and massive flows.These volcanic rocks have typical N-MORB compositions (Cooper, 2007;Neo et al., 2009;Yamazaki et al., 2009).Below a ∼50 m thick transition zone, a sheeted dyke complex with a thickness of 346 m was recovered.The thin nature of the sheeted dykes relative to the lava has been attributed to the fact that there appears to be no level of neutral buoyancy in the crust, as well as the relatively high magmatic pressures expected at magma-rich spreading centers, both of which favor extrusion over intrusion (Umino et al., 2008).However, the dykes, on average, are more evolved-and hence denser-than the lavas, suggesting that not all dykes erupted (Sano et al., 2011).This is consistent with their predominantly horizontal flow fabrics, as recorded by magnetic fabrics (Veloso et al., 2014).The lower 115 m of the hole comprised the transition from the upper crust (sheeted dykes and lavas) to the lower crust (plutonic rocks): at 1156 m below the top of the basement, the first gabbroic rocks were recovered (Figure 2).These form a heterogeneous gabbroic body of 54 m thick (referred to as Gabbro 1), which is comprised predominantly of medium-grained rocks that range in composition from modally layered olivine gabbros to oxide gabbros (Teagle et al., 2006(Teagle et al., , 2012)).Gabbros are internally heterogeneous down to the thin section scale, with early, relatively primitive subophitic domains forming patches in a matrix with granular textures and evolved compositions (Koepke et al., 2011).Gabbro 1 is intruded near its top by a thin (<1 m) quartzrich oxide diorite (Teagle et al., 2006).Following a dyke screen of 54 m thick, a second gabbroic body (Gabbro 2) was recovered, which appears to be a single intrusion of variably oxide-bearing gabbronorite (Figure 2; Teagle et al., 2006Teagle et al., , 2012)).The hole ended in a second screen of sheeted dykes (Figure 2).
The rocks of both dyke screens, as well as those from the base of the sheeted dyke complex, have granoblastic textures characterized by an assemblage of plagioclase, two pyroxenes and Fe-oxides (Koepke et al., 2008;Teagle et al., 2006Teagle et al., , 2012)).This assemblage is indicative of upper amphibolite to granulite-facies metamorphism, consistent with their positions at the top of the lower crust, the magmatic system of which served as a heat source (Koepke et al., 2008).The gabbroic rocks and granoblastic dykes are intruded by numerous cm-scale patches, veins and dykelets that range in composition from oxide gabbro and diorite to tonalite (Figure 2).These evolved rocks are generally medium-grained with granular textures, and commonly contain accessory phases such as apatite and zircon (Teagle et al., 2012).In contrast to the granoblastic textures of their host rocks, these intrusions retain igneous textures.They do, however, have high-temperature contacts with their host rocks, which indicates that they are late-to post-peak metamorphism (Teagle et al., 2012).The contacts of the patches with the ganoblastic host rocks vary from moderately sharp and comb-textured to gradational; hence, the patches may comprise  (Wilson, 1996).Adapted from Wilson et al. (2006).
both intrusions and segregations (Teagle et al., 2012).The relationships between the granoblastic dykes and the evolved intrusions within them have been the subject of recent interest.A number of observations have led to a model in which both the granoblastic textures and the intrusions are generated by partial melting of hydrothermally altered sheeted dykes during contact metamorphism (Erdmann et al., 2015(Erdmann et al., , 2017;;France et al., 2009France et al., , 2010;  Teagle et al., 2012;Zhang et al., 2017).These observations include a trend of depletion in incompatible element contents of the granoblastic dykes toward their contacts with the gabbroic rocks (Teagle et al., 2012), and the trace element distributions in the evolved intrusions (Zhang et al., 2017).This process has been studied experimentally, with the experiments reproducing both the nature of the granoblastic residue as well as the composition of the partial melts from hydrothermally altered dyke rock starting materials (Erdmann et al., 2015(Erdmann et al., , 2017;;Fischer et al., 2016;France et al., 2010).Hence, the recovered dyke-gabbro transition is a dynamic horizon, characterized by intrusion, metamorphism and partial melting.
We selected six samples recovered from Hole 1256D during IODP Expeditions 312 and 335 that span the igneous history of the section (Figure 2).These data supplement, and significantly expand upon, the geochronological data on three samples (two of Gabbro 1 and one of Gabbro 2) presented in Hayman et al. (2019).Hayman et al. (2019) relied mostly on relatively low-precision SIMS analysis but also provided higher precision ID-TIMS data for one Gabbro 1 sample.The latter data are directly comparable to those that we will present in this paper.
Within our data set, the main episode of crustal magmatism is represented by an oxide-and orthopyroxene bearing gabbro from Gabbro 1 (sample 312-1256D-217R-1, 36-39 cm).This sample is notably less patchy than most of Gabbro 1, suggesting it is predominantly comprised of the earlier, relatively primitive framework rather than any crystallization products of later and more evolved melts; hence, it maximizes the likelihood of dating the crystallization of the main Gabbro 1 magma, rather than any late-stage intrusive in it.The Gabbro 1 sample is supplemented by a sample of the <1 m thick quartz-rich oxide diorite that intrudes Gabbro 1 near its top (sample 312-1256D-214R-1, 37-52 cm).The oxide diorite locally has a comb structure defined by elongate plagioclase and likely magmatic amphibole, characteristic of fairly sharp but high-temperature intrusion into the Gabbro 1 host rock (Teagle et al., 2006).The remaining four samples represent various occurrences of evolved intrusive veins, dykelets and patches in Dyke Screen 2 (Figures 2 and 3).Three samples were taken from rocks that were recovered during cleaning operations in the hole, and derive from the borehole wall rather than drill core (for details, see Teagle et al., 2012).The exact origin and stratigraphic position of these rocks cannot be ascertained; however, the shipboard party considered it most likely that they were derived from near the bottom of the hole (Teagle et al., 2012).The first of these samples is 335-1256D-Run19-RCJB-Rock C, which is a prominent, cm-wide diorite vein with moderately sharp contacts with the granoblastic dykes (Figure 3b).The other two, samples 335-1256D-Run11-EXJB-J6 and 335-1256D-Run13-RCJB-Rock A, represent the evolved patches in Dyke screen 2 (Figure 3a), and are oxide-bearing diorite and oxide quartz diorite, respectively.Both have diffuse margins against their host granoblastic dykes.The last sample, 335-1256D-235R-1, 31-33 cm is a 2 cm wide tonalite dykelet cored from the bottom of the hole.It has fairly diffuse margins against the surrounding granoblastic dykes (Figure 3c), suggesting that it intruded when the dykes were at high temperature.

Methods
Samples were crushed using a steel jaw crusher and disk mill.The <500 μm fraction was processed using a Gemini table in order to extract a heavy mineral concentrate, which was further purified by passing through di-iodomethane with a specific gravity of 3.3.The most magnetic material was extracted using a hand magnet and a Frantz magnetic separator.Zircons, typically in the form of crystal fragments, were hand-picked from the non-magnetic fraction for ID-TIMS analysis.Medial melt inclusions were common.After picking, the zircons were chemically abraded following a modified procedure to remove damaged parts of the crystal that were likely to have experienced open-system behavior.First, the zircons were thermally annealed at 900°C for 60 hr in quartz crucibles before being individually selected, photographed and loaded into FEP Teflon beakers.Zircons were then refluxed in 4 M HNO 3 on a hotplate at 120°C for >2 hr, followed by ultrasonic cleaning for at least 20 min.The zircon crystals were rinsed with acetone and 4 M HNO 3 and loaded into individual 300 μl FEP Teflon microcapsules and leached in 29 M HF inside a Parr vessel (a self-sealing stainless-steel jacket) for 12 hr at 180°C.The zircons were rinsed with 4 M HNO 3 and refluxed in 6 M HCl at 120°C for 2-5 hr, before a final rinsing with 4 M HNO 3 several times.
The leached zircons and all total procedural blanks were spiked with mixed 205 Pb-233 U-235 U (ET535) EARTH-TIME tracer solution and dissolved in ∼150 μl 29 M HF and trace HNO 3 in a Parr vessel at 220°C for ∼60 hr.Complete dissolution was checked by visual inspection of some larger crystals and assumed for smaller grains, following the standard protocol for dissolution at NIGL.The solutions were dried down as fluorides and re-dissolved in 3 M HCl in a Parr vessel overnight at 180°C.U and Pb fractions were isolated by an HCl-based anion exchange procedure using Bio-Rad AG-1 resin in Teflon columns.Lead and U fractions were then recombined and dried down with ∼10 μl of H 3 PO 4 and subsequently loaded onto zone-refined Re filaments in a silica gel matrix to enhance ionization.
Isotope ratios were measured on a Thermo-Electron Triton TIMS.Lead was measured in the dynamic mode on a MassCom secondary electron multiplier; Pb mass bias corrections were made using a fractionation factor of 0.14 ± 0.02% amu-1 (1 sigma) for samples spiked using ET535.Dead-time and linearity of the secondary electron multiplier were monitored using repeated analyses of the standards NBS 982, NBS 981, and U 500.Uranium oxide (UO 2 ) was measured and corrected for isobaric interferences using an 18 O/ 16 O value of 0.00205 (IUPAC value and measured in-house at BGS; Condon et al., 2015;de Laeter et al., 2003).Uranium was measured in dynamic mode and a mass bias fractionation correction was calculated in real-time using the 233 U-235 U ratio of the ET535 tracer solutions.Corrections for the addition of Pb and U during the procedure (i.e., laboratory contamination) were made using the long-term measured isotopic composition and variability of blanks using an amount that is based upon contemporary total procedural blanks.
The U/Pb ratio for each analysis was determined via isotope dilution principles and the ET535 mixed 205 Pb-233 U-235 U tracer (Condon et al., 2015;McLean et al., 2015).A 238 U/ 235 U value of 137.818 was assumed (Hiess et al., 2012) and used in the data reduction algorithm (McLean et al., 2011).Following the approach of Hayman et al. (2019) and to facilitate the comparison of data between these two studies, 238 U-206 Pb dates were corrected for initial 230 Th disequilibrium using a Th/U[magma] value of 2.63.In presenting the data, we use the unit Ma to refer to dates, and Myr and kyr to describe duration.

Results
Zircons from Hole 1256D are dominated by euhedral, prismatic grains and grain fragments (Figure 4).The grains are up to several hundred μm long.In the diorite patches, where zircon is readily apparent in thin section, it is intergrown with amphibole, which forms a primary phase in these rocks, and plagioclase.This indicates that, in these samples, zircon crystallized cogenetically with the rock-forming silicate phases, consistent with the presence of primary Fe-Ti oxides and indicating that they are crystallization products of evolved melts.
The results of the U-Pb analyses are listed in Table S1 and illustrated in Figures 5-7.Overall, the Th-corrected zircon 206 Pb/ 238 U dates range from 14.56 to 15.31 Ma, with a single outlier at 15.82 Ma (Figure 5).All zircons except the 15.82 Ma outlier overlap concordia (Figure 6).2σ errors on individual dates range from 0.02 to 1.03 Myr, with a median of 0.13 Myr, and show a negative power-law correlation with the ratio of radiogenic to common lead (Pb*/Pb c ) (Figure 7a).Calculated zircon Th/U ratio range from 0.9 to 3.9, a range comparable to those reported in Hayman et al. (2019).Pb*/Pb c is highly variable, ranging from 0.15 to 12.59.In addition to precision reducing dramatically at low Pb*/Pb c , the 206 Pb/ 238 U dates of grains with low Pb*/Pb c (<0.5) are younger than those with higher Pb*/Pb c (Figure 7b).This is true even within individual samples, with dates for grains with low Pb*/Pb c deviating from those with higher Pb*/Pbc (Figure 7b).This indicates that those analyses that are dominated by common lead are unlikely to represent the true crystallization age of these grains.We attribute this to inclusions that were observed prior to leaching and not fully removed during chemical abrasion, and that crystallization ages are likely to be best preserved in those grains with high Pb*/Pb c .For this reason, we do not incorporate grains with Pb*/Pb c below 0.5 in the data presentation and discussion below.
The sample representing Gabbro 1 (312-1256D-217R-1, 36-39 cm) yielded single grain 206 Pb/ 238 U dates of 15.25-15.09Ma (n = 9; Figures 5 and 6).All grains are mutually within error, and the population has a mean squared weighted deviation (MSWD) of 0.68.This is consistent with a single population origin for zircon within this sample.These single grain dates, as well as their weighted mean (15.17 ± 0.03 Ma), are in agreement with the ID-TIMS data for Gabbro 1 reported by Hayman et al. (2019) (range 15.24-15.12Ma; weighted mean 15.19 ± 0.04 Ma).They also overlap with the SIMS date for Gabbro 2 (15.23 ± 12; Hayman et al., 2019).Although caution has to be exercised when comparing zircon dates obtained using different methods, this suggests that Gabbro 1 and 2 crystallized in close temporal proximity.
The diorite vein crosscutting Dyke Screen 2 (sample 335-1256D-Run19-RCJB-Rock C), like Gabbro 1 and its crosscutting oxide diorite, define a statistically uniform population at the current level of precision, with a range in dates of 15.22-14.94Ma (n = 8; MSWD = 1.23;Figures 5 and 6).The tonalite dykelet crosscutting Dyke Screen 2 (sample 335-1256D-235R-1, 31-33 cm), on the other hand, does not appear to comprise a single statistical population: although its range in dates is very similar to the other dates rock units (15.12-15.29 Ma), two high-precision dates (2σ errors of 0.04 and 0.05 Ma) are not within error of each other, differing in age by a minimum of 0.02 Myr.This is also reflected by the relatively high MSWD for this sample (2.48; n = 7).
One of the two dated evolved plutonic patches in Dyke Screen 2, sample 335-1256D-Run11-EXJB-J6, shows very similar results to Gabbro 1 and the crosscutting units: its single grain dates range from 15.31 to 15.14 Ma, and all dates are within error, with an MSWD of 1.19 for the population as a whole (n = 8; Figures 5 and 6).The second patch (sample 335-1256D-Run13-RCJB-Rock A), however, deviates from all of the other dated units: its two zircons with Pb*/Pbc>0.5 both have dates of 14.94 Ma (Figures 5 and 6).The significance of this observation will be discussed below.

Significance of Hole 1256D Zircon Dates
To date, significant variations in zircon dates of both individual grains and individual samples have been observed at (ultra)slow-and fast-spreading ridges, suggesting that crustal accretion is a protracted process (Baines et al., 2009;Grimes et al., 2008; C. J. Lissenberg et al., 2009;Rioux, Lissenberg, et al., 2012;Schwartz et al., 2005).However, at the high spreading rates prevalent during the formation of the section drilled in Hole 1256D, the crust is carried away from the spreading center very rapidly.As a result,  crustal accretion is expected to be rapid.Here, we examine the significance of the zircon dates of Hole 1256D, as well as the time span of magmatism that they record for the section.
It is striking that, with the exception of sample 335-1256D-Run13-RCJB-Rock A and the single zircon outlier in the oxide diorite at the top of the section (15.82 Ma), all samples show very similar dates.To test whether there is any statistically significant difference in dates within this population overall, we have calculated the MSWD of the full zircon population from all samples except 335-1256D-Run13-RCJB-Rock A, and excluding the 15.82 Ma outlier.The resulting MSWD is 1.4 (n = 41).This is consistent with the notion that, statistically, all of these samples behave as a single population, the weighted mean date of which is 15.19 ± 0.011 Ma.As pointed out by Hayman et al. (2019), this date fits well with the magnetic anomalies of the area: Hole 1256D was sited just east of the marine magnetic anomaly transition 5Bn-5Br, which corresponds to an age of ∼15.1 Ma using the time scale of Cande and Kent (1995) (Wilson, Hallenborg, et al., 2003; Figure 1).However, it should be noted that the weighted mean of 15.19 Ma is only significant if the population forms a normally distributed dataset defined by analytical error around a single crystallization date.There are a number of indications that this condition is not normally met for zircon in oceanic plutonic rocks.The first, and most direct, is that different zircons from the same sample are not always within error (C.J. Lissenberg et al., 2009;Rioux, Lissenberg, et al., 2012;Schwartz et al., 2005).This is also the case for one of the samples under consideration here: two single-grain zircon dates of the tonalite dykelet crosscutting Dyke Screen 2 (sample 335-1256D-235R-1, 31-33 cm) are not within error of one another, requiring a minimum duration of crystallization of 0.02 Myr.Furthermore, the youngest and oldest grain from the main population, which occur in samples sample 312-1256D-214R-1, 37-52 cm and 335-1256D-235R-1, 31-33 cm, respectively, require a minimum of 0.05 Myr crystallization (taking into account their respective errors).The second indication that oceanic zircons may not represent single crystallization episodes is that they tend to preserve a range of trace element concentrations, even within individual samples (Grimes et al., 2009).This is inconsistent with the crystallization of zircon at a single point in time; instead, it appears to crystallize over an extended part of the liquid line of descent of its parental magma (C. J. et al., 2009).On the basis of these two observations, oceanic zircons are not expected to form uniform populations with a normal distribution related to analytical error.If so, weighted means do not represent statistically valid dates.Nonetheless, at the current level of precision, the large majority of zircons dated in this study appear to define a single population.
In Figure 8, we examine the distribution of dates of this population.Overall, the distribution is not unlike a normal distribution, albeit with a slight skew to older dates.Given the low MSWD of this population, as well as its distribution (Figure 8), we consider 15.19 Ma to be the best estimate of the main episode of zircon crystallization in this section.The minimum spread of 0.05 Myr between different zircons from this population indicates that the main crystallization episode likely spanned some tens of thousands of years.
Both dates (14.94 Ma) for one of the patches in Dyke Screen 2 (sample 335-1256D-Run13-RCJB-Rock A) are on the young end of the spectrum defined by the other samples: they are younger than all of the other zircon grains dated in this study (Figure 8).As a result, when including these two grains in the main population of zircons dated in this study, its MSWD increases from 1.4 to 2.0.This indicates that, statistically, they are unlikely to form part of the main population.Furthermore, the patch is the only sample for which no dates fall in the 15.0-15.30Ma range, and the most precise single-grain date for this patch (14.94 ± 0.17 Ma, 2σ) is outside the error of 39% of all of the other dates presented in this study.The difference in dates between the main population and the patch is illustrated by their probability density distributions, which show only a small amount of overlap (Figure 8).The most likely interpretation, is that this patch is younger than the other units.It is not possible, however, to quantify just how much younger: the 14.94 Ma dates are within error of the younger dates of the main population, and whether the main population can be treated as a single crystallization event with a geologically significant mean date is uncertain (see above).Nonetheless, a maximum age difference is provided by the difference between the 15.19 Ma mean of the main population and the 14.94 Ma dates of the younger zircons.This maximum, which is 0.25 Myr, makes the assumption that the mean date of the main population is geologically significant, and ignores the errors on the dates.In reality, some of the younger zircons of the main population may have crystallized after 15.9 Ma, and the two dates from sample 335-1256D-Run13-RCJB-Rock A overlap within error with the majority of grains from the main population, making it likely that the actual age difference is smaller than 0.25 Myr.
The most remarkable data point is the 15.82 ± 0.11 Ma (2σ) Ma outlier in sample 312-1256D-214R-1, 37-52 cm.Taking into account their respective errors, it is at least 0.51 Myr older than the weighted mean of the main population of zircons (15.19 ± 0.011 Ma).One possible explanation is that the anomalously old zircon is derived from the drilling mud used during drilling of the hole.Sepiolite, one of the types of mud used during scientific ocean drilling, contains abundant zircons with a wide range of ages (2-2889 Ma; Andrews et al., 2016).Of particular note is that the age distribution of this sepiolite mud shows a secondary peak at 15.7 ± 0.2 Ma, which overlaps with the outlier in our data.However, the type of drilling mud used on Expedition 312 was not recorded, so it is not clear whether sepiolite was in fact used.Further, the Th/U of 3.7 of the 15.82 Ma outlier is higher than those of the majority of sepiolite zircon.Hence, although we cannot rule out contamination, it appears unlikely that the outlier is related to drilling mud.Older outliers have been found in oceanic plutonic rocks before: Schwartz et al. (2005) describe zircon cores ∼1.5 Myr older than their surrounding rims in gabbroic rocks from the Atlantis Bank massif in the Indian Ocean.Since the zircon rim dates overlap with the magnetic age of the crust, the cores were interpreted as inherited fragments of zircon formed prior to the main episode of crustal magmatism.At (ultra)slow spreading rates prevalent at the Southwest Indian Ridge this is a plausible scenario since the lithosphere is significantly thicker than the crust (e.g., Schlindwein & Schmid, 2016).This situation leads to the potential for crystallization in the deeper parts of the (mantle) lithosphere, and the subsequent upward transport of the resulting plutons to the shallow part of the lithosphere by corner flow.There, the main axial magmatism has the potential to assimilate zircons from the older plutons that initially crystallized at depth (Schwartz et al., 2005).Anomalously old zircons have also been found in plutonic rocks from the Oman ophiolite (Rioux, Bowring, et al., 2012).These zircons occur as individual grains that are up to 16.8 Myr older than the young cluster of zircons interpreted to reflect the crystallization age of the rocks (Rioux, Bowring, et al., 2012).The origin of these outliers remains enigmatic: they may relate to the propagation of the spreading ridge into older crust, or may represent zircon grains that were derived from subducted material (Rioux, Bowring, et al., 2012).For the 15.82 Ma outlier at Hole 1256D, two of these three possible explanations are unsatisfactory.Unlike Oman, a contribution from subducted material is unlikely as the Hole 1256D section did not form in the vicinity of a subduction zone.Similarly, because of the high rates of heat advection from the upwelling mantle and the large volumes of melt delivered to the crust, superfast-spreading ridge axes should not have thick lithosphere.The third possibility is that the zircon grain is inherited when the main magmatic pulse intrudes into older crust in a propagating rift or overlapping spreading center.Given the half-spreading rate on the Cocos plate during the formation of the Hole 1256D section (∼96 mm/yr; Wilson, 1996), 0.51 Myr of spreading corresponds to 49 km of crustal growth across-axis.Hence, the outlier could represent an inherited relic of crust that was subject to a propagating rift when it was 49 km off-axis.This is within the range of offsets recorded by current overlapping spreading centers at the Pacific, which can be up to 120 km, particularly where spreading rates are very fast (Martínez et al., 1997).However, the southernmost segment of the Pacific-Cocos Ridge had a length of at least 400 km, with no apparent evidence for significant ridge axial discontinuities.Furthermore, the magnetic anomalies around Hole 1256D, which have a spatial resolution of 1 km near the Hole and 5 km around it, show no indications for any offsets (Wilson, Hallenborg, et al., 2003).Hence, there is little supporting evidence for a propagating rift origin of the outlier.We conclude that its origin remains enigmatic.

Implications for Oceanic Crustal Accretion at Hole 1256D
The dates presented in this paper, supplemented by those of Hayman et al. (2019), enable us to place constraints on the history of crustal accretion preserved in Hole 1256D.The main episode of zircon crystallization occurred around 15.19 Ma, and likely lasted at least 0.05 Myr.Given the half-spreading rate of the Cocos plate at the time of accretion of the section in the Miocene (96 mm/yr), this time span translates to 4.8 km of spreading.Although there are currently no mid-ocean ridges on Earth with spreading rates as high as the Pacific-Cocos ridge during the Miocene, seismic data from the fast-spreading East Pacific Rise provide constraints on the width of the magmatic system beneath the ridge axis.Seismic tomography and seafloor compliance data have shown that a low-velocity zone, comprising a region of partial melt, occupies most of the lower crust (Crawford et al., 1999;Crawford & Webb, 2002;Dunn et al., 2000).It is generally ∼8-10 km wide, and is overlain by a narrower lens with higher melt proportions (Sinton & Detrick, 1992).This axial melt lens or AML is observed along the majority of the East Pacific Rise, and is generally <1 km but up to 4 km wide (Detrick et al., 1993;Mutter et al., 1995;Vera et al., 1990).Further constraints on the width of the crustal accretion zones come from the lava: although most lava along the East Pacific Rise erupt in the relatively narrow (hundreds of meters) axial summit through, volcanic rocks continue to erupt up to four km away from the ridge axis (Goldstein et al., 1994;Sims et al., 2003).All of these existing constraints fit well with the inference from the zircon data that the main episode of crystallization occurred over a time span of 0.05 Myr.
During this main episode of crustal accretion, both Gabbro 1 and the oxide quartz diorite that cuts it was formed.Gabbro 1's occurrence as an ∼50 m thick body within the lower part of the sheeted dyke complex suggests that at the time of its intrusion, it formed the type of melt body that is imaged geophysically as the AML.Although its lateral extent is unknown, its thickness, coupled with the fact that it is composite lithologically, is consistent with what would be expected from an AML.If so, Gabbro 1 marks the highest level AML of what would be a series of AMLs over the course of the crustal accretion history of the section, consistent with the notion that the AML horizon is dynamic and may migrate vertically over hundreds of meters (France et al., 2009;Hooft et al., 1997).Given the overlap in dates, it is likely that Gabbro 2 formed around a similar time as Gabbro 1, marking the emplacement of another melt lens within the deep portion of the sheeted dyke complex.
The diorite vein, tonalite dykelet and one of the diorite patches in the granoblastic dykes of Dyke Screen Two crystallized within the same time window as the gabbros.These veins and patches have been attributed to partial melting of the base of the sheeted dykes as a result of contact metamorphism by the main magmatic system below (Erdmann et al., 2015(Erdmann et al., , 2017;;France et al., 2009France et al., , 2010;;Teagle et al., 2012;Zhang et al., 2017).The fact that the zircon dates of the diorite vein, the tonalite dykelet, and one of the diorite patches are indistinguishable from that of Gabbro 1, and that this main population of dates is in keeping with the magnetic age for the crust, is consistent with this scenario: both Gabbro 1 and the partial melt in the metamorphosed granoblastic dykes are formed as part of the main axial sequence.The second diorite patch in the granoblastic dykes, however, appears to postdate this axial sequence.Although it is not possible to quantify the age difference, the offset is <0.25 Myr, which equates to <24 km of spreading.Hence, this diorite may represent an off-axis intrusion.Off-axis sills have been documented in seismic studies, extending up to eight km from the axial low-velocity zone (Canales et al., 2012).Furthermore, sills at the Moho have been found up to 22 km off-axis (Garmany, 1989).The lava stratigraphy of Hole 1256D is capped by a 75 m thick lava pond, which has been interpreted to have erupted off-axis (Crispini et al., 2006;Tartarotti et al., 2009;Tominaga & Umino, 2010;Wilson, Teagle, et al., 2003).There is thus a record of off-axis magmatism at fast-spreading ridges in general, and at 1256D in particular.Hence, it is possible that the diorite patch reflects the late-stage delivery of melts to the crust as it is carried off-axis.In this scenario, it would represent an intrusion of melt that underwent significant amounts of differentiation from off-axis basalt.However, the diffuse margin of the patch with the granoblastic dykes suggests that the host rocks were still at elevated temperatures when the patch intruded, which suggests that either the section remained hot for an extended period of time, or that the age difference between the patch and main population is relatively small.Alternatively, we cannot completely rule out that the patch may be of the same age as the other intrusions in the granoblastic dykes: although an age difference appears to be present (Figures 5 and 8), the associated errors allow for zircon from the patch to have crystallized synchronously to at least some zircons of the main population.

Time Span of Crustal Accretion Across the Spreading Rate Spectrum
With the addition of superfast-spreading crust to the global zircon geochronology dataset for oceanic crust, there are now data on the duration of crustal accretion across nearly the full range of spreading rates of the global mid-ocean ridge system.In Figure 9, we examine the pattern that emerges.As noted before (Grimes et al., 2011;Rioux, Lissenberg, et al., 2012), there appears to be little difference in the duration of crustal accretion between slow-and fast-spreading ridges, with most sites recording periods of ∼100-300 kyr.Our new data suggest that this observation extends to superfast-spreading ridges.This observation likely reflects the competing effects of the longevity of the magmatic system on the one hand and the rate of crustal transport off-axis on the other (Rioux, Lissenberg, et al., 2012).Faster spreading ridges feature more robust-and hence longer-lived-magmatic systems but are carried off-axis rapidly.Magmatic events at lower-spreading ridges may be shorter-lived, but the crust remains in the axial valley for longer periods of time.The net effect is that the total duration recorded by the zircon record is similar.Ultraslow-spreading crust may be an exception to this: the 1.5 Myr span recorded by zircon at the Southwest Indian Ridge far exceeds that observed elsewhere (Figure 9).This likely reflects the fact that the ultraslow-spreading lithosphere is substantially thicker than slow-to fast-spreading ridge lithosphere (Bennett et al., 2019;Schlindwein & Schmid, 2016).This is because mantle upwelling is slow, leading to reduced melt production and heat advection.The cool thermal structure results in thick lithosphere, with earthquake data indicating that the brittle lithosphere may extend to 35 km depth (Schlindwein & Schmid, 2016).This promotes initial crystallization of melts deep in the mantle lithosphere, evidence for which has recently been found in the melt inclusion record (Bennett et al., 2019).The subsequent upward transport of these plutons by corner flow, and the intrusion of melts into them at shallower levels, accounts for the extended time span of zircon crystallization (Schwartz et al., 2005).Although the lithosphere at slow-spreading ridges may locally also be thicker than the  2016); Rioux, Lissenberg, et al. (2012); Schwartz et al. (2005) and this paper.Boundaries between different classes of ocean ridges from Dick et al. (2003).magmatic crust (Cannat, 1996), crystallization depths of plutons are thought to be in the order of ∼6-7 km (Grimes et al., 2011;C. J. Lissenberg et al., 2016;Schoolmeesters et al., 2012), which is insufficient to lead to the extended periods of accretion recorded at the Southwest Indian Ridge.Hence, there appears to be a transition between very long (up to 1.5 Myr) duration of crustal accretion at ultraslow-spreading ridges characterized by thick lithosphere, to ∼100-300 kyr periods at slow-to superfast-spreading ridges.

Conclusions
Zircon U-Pb geochronology of plutonic rocks recovered from ODP Hole 1256D, which formed during an episode of superfast spreading at the Pacific-Cocos plate boundary, has revealed the following: 1.The main episode of crustal accretion occurred around 15.19 Ma, which is agreement with the seafloor age as estimated from magnetic anomalies.This episode likely spanned some tens of thousands of years and formed the main axial sequence.This comprises intrusions of gabbro into the base of the sheeted dyke complex as well as the formation and crystallization of partial melts of the deep part of the sheeted dykes.2. One sample apparently postdates this main episode.Although we cannot quantify the age difference, it was most likely less than 0.25 Myr.This sample may represent an off-axis intrusion, although, based on the uncertainties of both the main population and the dates of the younger sample, we cannot rule out that it formed as part of the axial suite.3. The time span of crustal accretion of oceanic crust is similar (∼100-300 kyr) across large parts of the spreading rate spectrum from slow-to superfast-spreading ridges, but may be substantially longer (up to 1.5 Myr) at ultraslow-spreading ridges characterized by thick lithosphere.enabled significant new insights into oceanic crustal processes.CJL's participation in IODP Expedition 335 was funded by NERC Grant NE/J005592/1.Geochronology was funded by NERC Isotope Geosciences Facility Grant IP-1276-1111 to CJL.This research used samples and/ or data provided by the Integrated Ocean Drilling Program (IODP).Adrian Wood and Nicola Atkinson provided technical assistance with the U-Pb zircon dating.We would like to thank reviewers Axel Schmitt and Joshua Schwartz for their helpful comments and suggestions.This is Cardiff EARTH CRediT Contribution 18.

Figure 1 .
Figure 1.Map showing the location of Ocean Drilling Program Hole 1256D in the eastern Pacific Ocean, along with the seafloor age estimated from magnetic anomaly data(Wilson, 1996).Adapted fromWilson et al. (2006).

Figure 2 .
Figure 2. Igneous stratigraphy of the lower ∼120 m of Hole 1256D, which recovered the transition from the upper crust to the lower crust.Samples studied in this paper are indicated by the white stars.Modified from Teagle et al. (2012).

Figure 5 .
Figure 5. Zircon U-Pb results as a function of stratigraphic depth in Ocean Drilling Program Hole 1256D.The zircon U-Pb TIMS data for the single sample presented by Hayman et al. (2019) are included for comparison.Light gray datapoints are those with high Pb*/Pb c , and are excluded from the interpretations and discussion (see text for details).The stratigraphic log after Teagle et al. (2012).

Figure 6 .
Figure 6.Concordia plots for zircons from Ocean Drilling Program Hole 1256D.

Figure 8 .
Figure 8. Ranked 206 Pb/ 238U dates for all zircons dated in this study, separated into the main population (red) and those of sample 1256D-Run13-RCJB-Rock A (blue), with their respective probability density curves superimposed.Errors on dates are 2σ.