The type locality for the upper Oligocene Nuwok Member of the Sagavanirktok Formation (Carter Creek, North Slope, Alaska, USA) contains an abundant occurrence of glendonite, a pseudomorph after the calcium-carbonate mineral ikaite, which typically forms in the shallow subsurface of cold marine sediments. The region during the time of Nuwok Member deposition was located at a high latitude, similar to today, and the study site is characterized by sands and silty muds interpreted here to have been deposited in coastal and shelfal marine environments. Isotopic (Sr) and biostratigraphic (foraminifera) evidence presented here refine the depositional age of the outcrop to approximately 24 Ma. Glendonites occur in two basic forms: radial clusters, commonly centered around a single larger primary crystal (∼ 10 cm, Type A) and larger single blades generally without accessory crystals (∼ 15–25 cm, Type B).

Microscopic examination reveals a sequence of multiple types of replacive calcite that formed as a direct result of ikaite transformation: Type 1 rhombohedral crystals characterized by microporous and inclusion-rich cores and concentric zones, Type 2A, composed of clear calcite that overgrew and augmented Type 1 crystals, and inclusion-rich, microcrystalline Type 2B, which formed a matrix surrounding the rhombs and commonly dominates the outer rims of glendonite specimens. Type 3 calcite precipitated as fibrous, botryoidal epitaxial cement atop previous phases and is not ikaite-derived. These phases are distributed in similar ways in all examined specimens and are consistent with several previously described glendonite occurrences around the world, despite differing diagenetic and geologic histories. Stable-isotope evidence (δ13C and δ18O) suggests sourcing of glendonite carbon from both organic and methanogenic sources. Glendonites of the Nuwok Member can therefore assist in the determination of a more comprehensive ikaite transformation model, improving our understanding of glendonite formation and the sedimentological and environmental context of their occurrence. Oligocene glendonites are uncommon globally; the well-preserved occurrence described here can allow future studies to better reconstruct Arctic environmental conditions and paleoclimates during this time.

Unique radial aggregations of euhedral calcite crystals, now referred to as glendonite, have been recognized since the 19th century (e.g., Sokolov 1825; Freisleben 1827; Dana 1849; Ko 1896; and others—see reviews of Rogov et al. 2021, 2023, and Schultz et al. 2022), but have only in the past few decades been recognized as pseudomorphs after ikaite (calcium carbonate hexahydrate; CaCO3·6H2O; Kaplan 1978, 1979), a mineral found in nature to form only under frigid conditions (–2 to +7° C; Vickers et al. 2022 and references therein). Modern ikaite has been observed in the shallow subsurface of marine sediments across the globe, but most commonly in polar regions (Suess et al. 1982; Zhou et al. 2015; Rogov et al. 2021, and references therein), where it may develop a distinctive growth habit of dagger- or spear-shaped crystals, commonly radiating outward from a central point and reaching up to decimeter-scale sizes (Schultz et al. 2022). Once ikaite is brought outside of its temperature–pressure stability field, it ultimately transforms to calcite (Marland 1975; Tang et al. 2009; Tollefsen et al. 2018, 2020; Vickers et al. 2022), sometimes in a matter of hours (Kennedy 2022), and commonly retains the original macroscopic crystal morphology (Schultz et al. 2022, 2023, and references therein). The presence of such crystal forms in ancient marine sediments therefore has been used as a proxy for cold aqueous conditions during or shortly after deposition (e.g., Rogov et al. 2017; Grasby et al. 2017; Peckmann 2017), although the laboratory nucleation and growth of ikaite at warmer temperatures has led to uncertainty regarding their use as cold-climate indicators (e.g., Purgstaller et al. 2017; Stockmann et al. 2018; Tollefsen et al. 2020). Nonetheless, glendonites are often found in association with other indicators of near-freezing temperatures (e.g., glacial sediments; Kemper 1987) and several quantitative studies reconstruct low temperatures using ikaite/glendonite isotope data (e.g., Vickers et al. 2020, 2022; Schultz et al. 2023) in sedimentary successions known to have been deposited at high latitudes. These studies highlight the need to examine the geochemical and sedimentological context of glendonites in order to understand the paleoclimatic significance of their occurrence.

On the North Slope of Alaska (USA), large, well-preserved specimens of glendonite are exposed in the silty and sandy mudstones of the Nuwok Member of the Sagavanirktok Formation at Carter Creek. While the exact age of the Nuwok Member has been the subject of debate (see discussion below), previous studies have interpreted outcropping strata to have been deposited in a shallow marine or inner-shelf environment between the late Oligocene and Pliocene, when Arctic Alaska was at a similar latitude to that of today (∼ 70° N). Only a few brief references to these glendonites are present in the literature; this study therefore aims to document in detail the distribution, morphology, mineralogy, and geochemistry of these unique pseudomorphs for the first time, as well as their sedimentological context. Additionally, new biostratigraphic and isotopic analysis provides up-to-date constraints on the depositional environment and age of the section. Such information has implications for understanding the geochemical, oceanographic, and climatic setting(s) required for ikaite/glendonite formation, and may assist with future studies that intend to utilize ancient glendonites for paleoenvironmental reconstruction.

Geologic Setting

Strata exposed at Carter Creek and elsewhere on the Alaska North Slope were deposited in the Colville Foreland Basin, a large depocenter spanning over 1000 km E–W (Bird and Molenaar 1992) and containing as much as 7.5 km of sedimentary fill (Houseknecht 2022). The basin was formed by flexural subsidence during the first phase of the Brookian orogeny during the Jurassic to Early Cretaceous. Subsequent infilling from the Chukotka terrane to the west and the Brooks Range to the south resulted in the eastward and northward progradation of depositional systems that formed large-scale clinothems during much of the Cretaceous and Paleogene (Houseknecht 2019). By Cenomanian time, the western part of the basin was largely filled, forming a continental terrace in which a variety of nonmarine to shallow marine clastic lithologies were deposited. In the eastern part of the basin, including the Carter Creek area, reactivation during the second phase of Brooks Range orogenesis and associated subsidence in the Paleogene led to the development of substantial new accommodation (McMillen and O’Sullivan 1992; Houseknecht 2019). Tectonism, evolution of local tectonic relief, and development of wedge-top basins all contributed to a highly dynamic and locally variable depositional system in this area. Carter Creek outcrops can be placed within this framework through correlations with regional subsurface seismic and well data; these data show that the study area is part of a sequence (T4 of Houseknecht 2019) (Fig. 1A) that is characterized by shallow marine to shoreface topset facies underlain by thick foresets and bottomsets comprising slope and sediment-gravity-flow deposits. Seismic ties to wells with biostratigraphic data (both offshore and in the nearby Point Thompson area) place these topset strata within upper Oligocene strata (Bujak 2005–2019).

Carter Creek strata form the upper part of the regionally extensive and time-transgressive Sagavanirktok Formation (Decker 2010), which comprises four members: the Sagwon, White Hills, Franklin Bluffs, and Nuwok members (Mull et al. 2003) (Fig. 1B). The Nuwok Member, in which glendonite specimens occur, is exposed at only two localities: the type section at Carter Creek on the coastal plain of the Arctic National Wildlife Refuge (ANWR; Detterman et al. 1975) (Bader and Bird 1986) (Fig. 2A) and in the coastal cliffs near Barter Island (Brouwers 1994). The latter coastal outcrop exposes only the uppermost Nuwok Member and the overlying Pliocene to Pleistocene Gubik Formation. The Carter Creek exposure contains a thicker section of the Nuwok Member, although neither the lower contact with the Franklin Bluffs Member nor the upper contact with the Gubik Formation is exposed. The Carter Creek section is exposed at the surface due to uplift associated with the formation of the Marsh Creek Anticline, a passive-roof duplex resulting from Cenozoic thrust faulting of pre-Mississippian and overlying strata (Potter et al. 1999, 2004; Houseknecht 2019) (Fig. 2B). Basement-detached thrusting in the Oligocene to Miocene led to syndepositional and postdepositional deformation of overlying strata (Potter et al. 1999), including those discussed here; exposure of the Nuwok Member is therefore spatially limited on the coastal plain to the modern river incisions at Carter Creek.

Previous Investigations into the Nuwok Member

Although early studies of Arctic Coastal Plain strata identified the beds at Carter Creek as the “Nuwok Formation” as early as 1919 (Dall 1919, 1920; Leffingwell 1919), the Sagavanirktok Formation was not lithostratigraphically formalized until 1951 (Gryc et al. 1951), when the type section was defined at Franklin Bluffs, 150 km to the west of Carter Creek. Detterman et al. (1975) later redefined the Sagavanirktok Formation as encompassing all outcropping North Slope strata between the top of the Cretaceous Prince Creek Formation and the base of the Quaternary Gubik Formation, assigning three members (Sagwon, Franklin Bluffs, and Nuwok). The current status of these units results from a 2003 revision of Colville Basin stratigraphic nomenclature (Mull et al. 2003), which left the Nuwok Member unaffected and added the White Hills Member lower in the Sagavanirktok Formation (Fig. 1B).

These studies all recognized the Cenozoic age of the Nuwok beds on the basis of fossil flora and invertebrates, although the exact epoch has been the subject of discussion. Morris (1957) utilized pelecypods to estimate a range of late Miocene to Pliocene. Ostracods and foraminifera studied by Todd (1957) indicated a similar age, and microfossils studied by Bergquist (unpublished analysis noted in Detterman et al. 1975) resemble late Pliocene to early Pleistocene forms. Numerous species, however, are endemic to the Nuwok Member, resulting in remaining uncertainty as to the exact age of the complete section. Brouwers and Marincovich (1988) reported ostracod and mollusk assemblages from Carter Creek and made an initial description of glendonite horizons. They note that ostracods occur in three distinct zones, with the lowermost containing an ostracod assemblage that is similar morphologically to late Miocene to early Pliocene-age faunas in the eastern USA. The middle zone, where glendonite-bearing horizons occur, was estimated to be at least 3 Ma, but the maximum age is unknown, and the upper zone is at maximum 2.7 Ma. Ostracod zones and the presence of glendonites were also used to correlate uppermost Nuwok Member strata at Carter Creek to coastal exposures at Manning Point and on Barter Island (Brouwers 1994) (Fig. 2A), where they were determined to likely be upper Pliocene. Thus, most of these studies concur that glendonites are Miocene to Pliocene in age.

An older age for the Nuwok Member also has been proposed by several authors and was supported by the most recent examination of the outcrop (Port 2008). Fouch et al. (1990) and McNeil and Miller (1990) argued that the Nuwok beds are Oligocene based on data from strontium-isotope ratios, with 87Sr/86Sr ratios correlated to the magnetostratigraphically calibrated reference curve from DSDP Site 522 yielding ages of 27.0 Ma to 23.8 Ma (Miller et al. 1988, though time scale boundaries have since been modified) and the presence of a late Oligocene index foraminifer, Turrilina alsatica. This age, however, was subsequently questioned on the premise that calcite in the Carter Creek section had been diagenetically altered, potentially rendering ages determined by strontium isotope ratios inaccurate (Marincovich and Powell 1991). Additionally, the presence of a fish otolith (Merlangius arcticus) with an oldest occurrence of early Miocene, was used as further evidence of a Neogene age (Nolf and Marincovich 1994). A plesiomorph sister taxon with a late Oligocene age, however, allows the possibility of an older age estimate, as the Carter Creek otolith was determined to be a new species without a previous Arctic record. Thus, conflicting chemostratigraphic and biostratigraphic evidence has resulted in considerable uncertainty as to the exact age of the outcrop, prompting the new age analyses reported here (biostratigraphy and Sr isotope dating, discussed below).

Field Work

Although glendonite specimens were recognized in talus slopes at the Carter Creek site during numerous previous North Slope field seasons, a systematic investigation of the site’s in situ sedimentology and stratigraphy was begun in 2017 and completed in 2022. Field methods consisted of photography, 0.5- to 1-m sampling of trenched outcrop, and measuring stratigraphic sections at centimeter to decimeter scale in the immediate area of glendonite occurrence. Individual measured sections tens of meters apart in the field were combined to form a single composite stratigraphic column; these were separated by intervals where erosion or thick scree slopes prevented access to the outcrop. A stratigraphically accurate three-dimensional model of the locality was created with Agisoft Metashape photogrammetry software (Nieminski and Graham 2017) using high-resolution GPS-located images taken from a helicopter. A scaled orthomosaic image of the locality was generated from the outcrop model (Fig. 3A) to create a spatially accurate composite stratigraphic section. In situ glendonites were found primarily in the approximate center of the section (red box, Fig. 3B); no specimens were seen in eroded sediments above this interval, but were abundant in eroded material below it.

Biostratigraphy

Initially, palynomorphs were extracted from Carter Creek sediments via acid digestion in 10% HCl, followed by 70% HF for up to four hours. Digested concentrate was centrifuged and decanted multiple times and then immersed in ZnBr2 for heavy-liquid separation. Separated concentrate was again centrifuged and decanted, and then oxidized as needed with Schultze’s solution (a mixture of KClO3 and HNO3) and NH4OH. A second investigation into palynomorphs was conducted independently on a sediment sample immediately adjacent to in situ glendonites. In this instance, the sediment sample was dried, powdered, and treated with HCl before washing at 60°C with citric acid. The residue was subsequently oxidized for 3 minutes in HNO3, neutralized for 1 minute in KOH, and then separated via heavy-liquid (ZnCl2) immersion and stirring. The sample was then sieved on 11 μm mesh. A small fraction of the residue was mounted in gelatin and used to make a slide, which was examined under a transmitted-light microscope. The kerogen slide and the residuum from this analysis are stored at the Geological Survey of Denmark and Greenland (GEUS) for future reference.

Seven samples, ranging in weight from 67.54 g to 83.97 g, were processed for foraminifera. All samples were freeze dried in a Labconco FreeZone1, weighed, then repeatedly submerged in a 5 g/L sodium hexametaphosphate solution for no less than hour, and washed over a 63 μm sieve until clean. Up to 300 foraminifera per sample were picked from washed material by hand to measure planktic/benthic ratios. Select specimens were imaged under a Hitachi SU-5000 Schottky field-emission scanning electron microscope (SEM). Foraminifera from two samples were speciated using references provided in the results below. Biostratigraphic data can be found in Supplementary Data Set 1. Ages referenced here and elsewhere throughout this paper generally follow the GSA geologic time scale (Walker and Geissman 2022), except where noted.

Laboratory Analyses

Thin sections were prepared by both the University of Kansas and National Petrographic Services (NPS). Petrographic work was conducted with a Nikon Eclipse LV100N POL microscope, and photomicrographs were acquired with a Nikon DS-Fi3 camera and NIS-Elements D 5.02.01 software.

Polished mounts of glendonite were examined on a Hitachi SU-5000 Schottky field-emission scanning electron microscope (SEM) fitted with a backscatter electron detector, an Oxford Ultimax 100 mm2 energy-dispersive spectroscopy silicon drift detector (EDS-SDD) and a Delmic SPARC cathodoluminescence (CL) detector (hot cathode). Samples were coated with ∼ 10 nm of carbon, and imaging and analyses were done at a typical accelerating voltage of 15 keV and ∼ 1–4 nA of beam current. Spot EDS analyses (10–30 s count time) were used for qualitative compositional comparisons among the glendonite textural populations. EDS mapping was then used to map the distribution of the compositional variations identified by spot analyses. EDS maps were typically collected for several minutes at an approximately 0.1–0.3 µm spatial resolution, and nominal count rates of 50k counts per second. Red–green–blue (RGB) CL images were obtained using a filter wheel and serial collection of blue (< 505 nm), green (505–575 nm), and red (> 605 nm) channels on a photomultiplier-tube (PMT) detector. Channel mixing was adjusted in Odemis Delmic software to highlight features of interest, and selected images were further processed in GNU Image Manipulation Program (GIMP) to maximize internal contrast.

Glendonites were sampled for isotopic analysis by micro-drilling multiple specimens in areas dominated by certain calcite types, and pulverizing the resulting powder (< 1 gram) with a mortar and pestle. Stable isotopes (δ13C and δ18O) were measured by the University of Arkansas Stable Isotope Laboratory (UASIL) and Oregon State University. For the latter, samples were acidified with 100% phosphoric acid at 70°C using a Kiel IV carbonate device and then analyzed by dual-inlet mass spectrometry using a MAT253 isotope-ratio mass spectrometer. Data were calibrated using an in-house standard (Wiley) that was analyzed ∼ 4× at the beginning and end of each run. The international standard NBS-19 was analyzed at the end of each run as a check standard (i.e., not used to calibrate the run). The standard deviation for δ13C in repeated samples is typically 0.03 per mil or better, and for δ18O it is typically 0.06 per mil or better. UASIL utilizes a similar methodology, with NBS-19 and internal (UASIL 22) standards run on a Delta V Advantage isotope-ratio mass spectrometer with dual Gas Bench II peripherals. This data, along with Raman and other isotopic analyses, can be found in Counts et al. (2023) and Supplementary Data Set 2.

Analysis of δ13Corg in Carter Creek bulk sediments was undertaken by the U.S. Geological Survey (USGS) Petroleum Geochemistry Research Laboratory (PGRL) in Denver, Colorado following methodology described in Révész et al. (2012). Samples were air-dried and then pulverized using a shatterbox. Carbonate material was removed from bulk samples through acid digestion in HCl. Isotope ratios were analyzed using a Thermo Flash 2000 elemental analyzer coupled to a Thermo MAT253 isotope-ratio mass spectrometer (EA-IRMS) via a Thermo Conflo IV. Standards (ASN-02A, GH SugarA, NBS22B, USGS40B) were measured throughout analyses; all were found to be within acceptable tolerances (0.2‰). Measurements of total organic carbon (TOC) were conducted at the same laboratory on the same sample set using a LECO C744 Carbon Analyzer. TOC preparation involved acidification of samples with 6 M HCl to remove inorganic carbon held in carbonate minerals, followed by rinsing in deionized water. Quality-control materials, including reference standards, were analyzed throughout the analytical procedure; precise details on these and other aspects of TOC analysis are outlined in Oliver and Warden (2020).

Micro-Raman spectroscopy on a glendonite sample was carried out at the USGS Geology, Energy & Minerals Science Center using a Horiba Xplora Plus confocal microscope equipped with a 532 nm laser, 50× (0.5 NA) objective, ∼ 5 mW laser power at the sample surface, 50 µm pinhole, 100 µm spectrometer slit, and 2400 groove/mm dispersive grating. Spectra represent the mean of three 10 s acquisitions, collected across a spectral range of 600–1500 cm−1. Spectra were calibrated against the 520.7 cm−1 response of crystalline Si and have a stated Raman shift precision of ± 1 cm−1. Spectra were fitted simultaneously using a linear baseline and a Lorentzian peak centered at ∼ 1087 cm−1 between 1040 and 1140 cm−1 (Meier 2005). White-light photomicrographs were taken pre- and post-laser exposure to check for thermal alteration of the sample. No sample alteration was observed using the stated conditions.

Multiple parts of an in situ glendonite specimen (55 m in Fig. 4) and a bivalve fragment from lower in the section (12 m in Fig. 4) were analyzed for Sr isotope ratios, with the goal of refining the timing of deposition and glendonite formation. A total of five Sr isotope measurements were made, including replicate sets from the white chalky outer layer of the glendonite and from the brown infill and one measurement from the bivalve fragment. The bivalve sample was rinsed in dilute 10% HCl for 20 seconds and transferred and rinsed three times in an ultrasonic bath of deionized water. Duplicate samples of the white chalky outer layer and brown infill were drilled across multiple zones of the glendonite to assess intracrystalline variability. Glendonite powders and the bivalve fragment were leached in 1 mL of 0.2M ammonium acetate for two hours at room temperature to remove any cations in exchangeable sites in detrital or diagenetic silicate material. Samples were then dissolved overnight in 1 mL of 5N acetic acid. Remaining solutions were then subject to clean-column elution chemistry following protocols outlined in Quade et al. (2022). Sr isotopic ratios (87Sr/86Sr) were measured by the multidynamic “triple jump” method on a Thermo-Finnigan Triton thermal-ionization mass spectrometer (TIMS) at the USGS Denver Radiogenic Isotope Lab (DRIL). Raw isotope ratios were corrected for mass bias using an accepted value of 86Sr/88Sr of 0.1194. All reported sample 87Sr/86Sr values were normalized to the value obtained for SRM-987 within the same run (0.710250 ± 0.000011, n = 10, accepted value = 0.710248) and the normalization checked using modern seawater value standard EN-1 (0.709173 ± 0.000011, n = 8, accepted value = 0.709174). Reported SRM987 values in Counts et al. (2023) and Supplementary Data Set 2 are as measured by DRIL during the two analytical sessions measured with the samples. Reported EN-1 values are normalized by the factor required to correct SRM987 from the “as measured” value to the accepted value. Both returned the accepted value within uncertainty.

Powder X-ray diffraction (PXRD) was carried out on dried, powdered bulk sub-samples to characterize the mineralogy of glendonites. PXRD was conducted using a Stoe StadiP transmission (thin-foil) diffractometer with a copper anode at 30 mA, 40 kV and a germanium 111 monochromator to produce Kα1 X-rays (Counts et al. (2023) and Supplementary Data Set 3). The diffracted beam was collected by an 18° 2θ Dectris Mythen1K silicon strip detector. Samples were sandwiched between two thin cellulose acetate discs and mounted in a holder set to spin continuously during data collection. Data sets were scanned from 5 to 65° 2θ stepping at 0.5° and 20 s/step. The resultant raw data have a step of 0.015° 2θ. Machine alignment was monitored using an NBS silicon standard. Phase analysis was done using Bruker’s “Eva” program (Gates-Rector and Blanton 2019) interfaced with the Powder Diffraction File provided by the International Centre for Diffraction Data.

Sedimentology and Isotope Geochemistry of Glendonite-Bearing Strata

Outcrops of the Nuwok Member at the Carter Creek locality are exposed in a series of south-facing erosional gullies that drain to the southwest into Carter Creek, which flows northward into the Beaufort Sea two kilometers to the north. Nuwok Member beds at the site strike approximately 75° and dip to the northwest at approximately 30°, resulting in the exposure of cross sections of Paleogene to Neogene strata on slopes in several ravines (Fig. 3). Strata are generally dominated by clays, muds, silts, and sands, with the overall grain size generally increasing upward throughout the section. The stratigraphically lowest interval exposed (Fig. 4; meters 0–20) consists mainly of clay-rich silty muds, commonly containing thin silt- and very fine-grained sand stringers and irregular patches of orange-yellow staining. Isolated, rounded pebbles, predominantly of chert, are found sporadically in both muds and sands, commonly in hydrodynamically unstable positions where the long axis is subvertical (Fig. 5A). A fine-grained sand-dominated bed with ripple marks, interspersed pebbles, gravel lags, and thin-shelled pelecypods (see Morris 1957 and Port 2008) is present approximately 12 m above the base of the section (Figs. 4F, 5B), and a single in situ glendonite was found in muds at approximately 16 m. No volcanic ash beds were observed in the section. Measured δ13Corg values for the lowermost 12 m of the section were found to be within a narrow range for all samples, between –25.8 and –26.1‰. Stratigraphically above these beds, approximately 20 m of section is covered or inaccessible, though is likely to be mud-dominated but with some sand component based on weathering patterns and lithologies observed on limited exposed surfaces. Talus slopes originating from this covered interval contain large (decimeter-scale) glendonite specimens presumably originating from this section; these are typically of Type B (described below), a habit not observed in situ elsewhere, suggesting that glendonite morphologies may vary stratigraphically throughout the section.

The main interval from which multi-crystal aggregate (Type A) glendonites are found in situ lies 45–65 m above the base of the outcrop. The lower part of this section is dominated by blocky clays and silty muds, which are in turn overlain by an interval where this lithology alternates with rippled silt and sand beds (Figs. 4E, 5C) and thin sand laminae. Ripple paleocurrents indicate a northward flow direction. Patches of orange-yellow staining, possibly incipient iron-oxide concretions, are present throughout. At least one sharp-based scour surface is present, overlain by a meter-scale unit of blocky clays generally lacking in sand. In situ glendonites (Fig. 5D) are embedded in silty muds (Fig. 4D), and the mud fabric wraps around crystal apices, likely reflecting deformation of the sediments during displacive growth of glendonite under shallow burial or compaction of sediments after glendonite formation. Based on these exposures, it is likely that some of the illustrated glendonites found in eroded talus slopes are parts of larger radial aggregates of multiple crystals that are poorly cemented together. Smaller crystals are found a few meters higher in the section, with the continued presence of isolated pebbles in between (Fig. 5E), as well as sand beds containing thin, paired mud laminae (Fig. 5F). Just above the uppermost known glendonite-bearing layer, strata continue to be heterolithically bedded (Fig. 5G, H), with regularly alternating centimeter-scale beds of rippled very fine to fine-grained sand and silty mud (Fig. 4C, including benthic foraminifera). Sand content increases upward in this part of the section before a partially covered interval approximately 64 m from the base. The δ13C values in the lower part of this section also fall within a narrow range, between –25.8 and –26‰, until approximately 55 m, at which point they shift positively to –25.3 to –25.5‰ and continue in that range for another 5 m of section.

Sediments from approximately 75 m to the top of the section contain sand beds that are thicker and coarser than any found below. Just above this point is the only lithified sediment in the section, in the form of an ∼ 20-cm thick cross-stratified sandstone bed that overlies a similar, but unconsolidated, lithology below (Figs. 4B, 5I). This unit in turn underlies an ∼ 5-m thick, gray bed of fine-medium grained, well-sorted, cross-stratified lithic sand (Figs. 4A, 5J) that forms a prominent marker horizon across the upper part of the outcrop. This bed contains granules and is irregularly iron oxide-stained; cross-stratification is generally indistinct. The remaining uppermost portion of the outcrop is poorly exposed, but consists of both sand and silty mud beds, with thicker sand beds containing abundant and possibly paired mud laminae as seen in Figure 5K. The exposure ends at the base of the modern vegetated tundra at the top of the ravine.

Age Investigations

Biostratigraphy

Palynological evaluation of four samples collected from across the stratigraphic range of the outcrop (at 1, 53.5, 54.5, and 95 m) yielded numerous taxa commonly present in early Tertiary strata in Alaska. All four samples contain similar assemblages. All palynomorphs are dominated by species with long temporal ranges. A full list of species can be found in Supplementary Data Set 1. Bisaccate gymnospermous and monosaccate Tsugaepollenites pollen are abundant, as well as fern spores such as Laevigatosporites, Deltoidospora, and Osmundacidites. The lycopodiaceous genus Retitriletes is also common. Angiosperm pollen species commonly found include small triporate forms related to birch and alder. Rare Paleozoic forms and reworked Late Cretaceous taxa are also present. The kerogen content is interpreted to be from land-based sources, consisting of abundant black detrital grains and light yellow-brown plant material. Visual thermal alteration index is estimated at approximately 2.5, consistent with previously reported vitrinite-reflectance values (Bird et al. 1999). Palynomorphs found in Carter Creek strata are not diagnostic of specific Cenozoic epochs, and can only be used to determine the general age of the section. Observed taxa generally are not present in strata older than Eocene, and none of the notable taxa known to indicate Paleocene strata are present, thus bracketing the oldest possible age of the outcrop. All taxa present also continue into the Neogene, and none has a last occurrence that excludes the Oligocene to Miocene age recorded by previous studies. Although a Miocene age is not precluded, the typical Miocene taxa are not present. Growth of the Marsh Creek Anticline to the south (Potter et al. 1999) may have uplifted older strata and provided a source for recycled taxa found in the unit. The second palynological evaluation yielded similarly non-diagnostic results, with an assemblage dominated by pollen, spores, and few poorly preserved dinocysts. These include Odontochitina sp., possibly reworked from Cretaceous sediments.

Analysis of benthic foraminifera in samples from near the middle of the outcrop (meters 44–50, just below the main in situ glendonite zone) proved to be more useful than palynology in assigning an age range to the outcrop. Although numerous species with large, non-diagnostic age ranges are documented (Supplementary Data Set 1), two late Oligocene index species, Turrilina alsatica Andreae (Fig. 6A) and Turrilina andreae Cushman, are present in relatively high abundance; they both have a last occurrence (LO) in the Oligocene. Three other species have a first occurrence (FO) in the Oligocene and are observed in these samples: Buliminella elegantissima d’Orbigny (Fig. 6B, C), Melonis affiis Reuss and Elphidium discoidale d’Orbigny (Fig. 6D, E; Cushman and McGlamery 1937; Bhatia 1955; Batjes 1958), though in relatively low numbers.

Conversely, one species identified here, Chilostomellina fimbriata Cushman (Fig. 6F–H), has an earliest documented occurrence in late Miocene strata from the North Sea (Hoskin and Haskins 1975), and another Miocene species (Globobulimina affinis d’Orbigny) is present, but rare in Carter Creek samples. Chilostomellina fimbriata bears strong resemblance to the “unnamed Chilostomellina” of McNeil (1990) from Carter Creek, which were previously illustrated but misidentified as Nonion labridoricum (Dawson) by Todd (1957). As the latter identification is inconsistent with observed morphological features (e.g., a much-expanded final chamber that extends over the surface of the test, nearly engulfing preceding chambers), here we follow Hoskin and Haskins (1975) in calling the specimens observed in these samples Chilostomellina fimbriata. It should be noted, however, that C. fimbriata has been documented only once in the Miocene (Hoskin and Haskins 1975).

Acknowledging that this taxonomic assemblage is contradictory at times, we favor an age of late Oligocene for the Nuwok Member at Carter Creek, consistent with the late Oligocene age put forth by McNeil and Miller (1990). Miocene or later ages assigned by Todd (1957), Brouwers and Marincovich (1988), and Nolf and Marincovich (1994) are inconsistent with the last occurrence datums of Turrilina alsatica and T. andreae, both of which are present at Carter Creek and do not extend into the Miocene (Revets 1987). While other taxa with Miocene first occurrences also are infrequently present, the high abundance and excellent preservation of the diagnostic Turrilina species makes it unlikely that they are the product of reworking.

Isotope Chemostratigraphy

A late Oligocene age also is consistent with new 87Sr/86Sr data (Fig. 7; Supplementary Data Set 2) taken from both glendonites (around 55 m in Fig. 4) and calcareous fossils (bivalves, collected from around 12 m in Fig. 4) found in the section. The bivalve sample recorded the least radiogenic composition (87Sr/86Sr = 0.708186). Replicate analyses of the exterior (Type 1–2B below) calcite are more radiogenic than the bivalve sample and internally indistinguishable within analytical uncertainty (87Sr/86Sr = 0.708253–0.708269). Assuming these values reflect contemporaneous seawater composition, the most recent global 87Sr/86Sr reference curve (McArthur et al. 2020) would indicate a late Oligocene age of formation (ca. 24 Ma). Internal, brown-colored areas from the same specimens (dominated by later Type 3 calcite overgrowths, discussed below) are the most radiogenic, with replicate 87Sr/86Sr analysis between 0.708402 and 0.708418. These values yield an early Miocene age (late Aquitanian; 21–20 Ma; Fig. 7). These data are similar to 87Sr/86Sr values reported from the section by McNeil and Miller (1990) on bivalves and foraminifera, though changes to the geologic time scale now place those previously published samples close to the Oligocene to Miocene boundary.

Glendonite Morphology

Glendonites display a range of macroscopic morphologies that can be broadly grouped into two types (Fig. 8). Type A glendonites (Fig. 8A–O) are typically 5–15 cm in length, and consist of clusters of numerous individual crystals that commonly radiate outward from the base of a single primary crystal larger than others in the cluster. Type A glendonites show a gradation in the number and size of accessory (subsidiary) crystals present, ranging from small and sparse (Fig. 8A–E), to intermediate forms (Fig. 8F–L), to nearly equal in size to the parent crystal (Fig. 8M–O). In the latter end member, accessory crystals may almost completely obscure the primary crystal through nucleation and continued growth higher along the axis of the primary crystal, leading to more equant forms instead of the “pineapple” shape typical of most specimens. Field observations of specimens in situ show that these may occur as isolated crystals (with the largest crystal pointing upward toward the top of the section), or as part of larger aggregates of crystal clusters that are weakly cemented together. In these aggregates, the largest crystal within a cluster may be oriented bedding-parallel rather than bedding-perpendicular. Thus, despite the wide range of appearances of individual Type A glendonites found in talus slopes, it is likely that they share a similar genetic origin and differ only in the level of secondary crystal development.

Type B glendonites (Fig. 8P–T) differ in that they are much larger overall (15–25 cm), and generally consist of only a single large, thick (typically 3–5 cm) crystal form rather than radial clusters. Crystals are rhombohedral in cross section and may either taper gradually to a single sharp point or display indications of being doubly terminated. Rare smaller subsidiary crystals are present. Most specimens, while large, are only partially complete, and thus may have additional features not observed here. Exterior surfaces of these glendonites are often chalky or pitted, or contain small, translucent euhedral calcite crystals, likely the result of late diagenetic recrystallization and/or extended exposure at the present-day surface. Type B specimens were found only in talus slopes at the outcrop locality, and therefore their growth orientation could not be determined.

Glendonite Mineralogy

Petrographic observations of sectioned glendonites in cross section and thin section (Figs. 911) reveal that they are composed almost entirely of calcite, with varying amounts of Mg and minor components of quartz (likely incorporated from surrounding sediment) and gypsum (potentially from post-exposure precipitation) (Table 2; Supplementary Data Set 3). However, distinct differences exist in the texture, growth habit, relative locations, and geochemistry of the calcite that constitutes glendonite interiors. These calcite phases can be categorized based on their optical and geochemical properties and numbered by their interpreted relative timing (Table 1). Some previous studies have defined three calcite phases in glendonites (e.g., Huggett et al. 2005; Vickers et al. 2018; Schultz et al. 2023): two phases arising during transformation of the glendonite, and a third, late-stage spar. The Carter Creek glendonites show more complex textures than other previously published examples, leading to the division of some calcite types into sub-categories, listed below. Single glendonite samples can be broken into distinct zones (Fig. 9), each of which contains certain combinations of the calcite phases described below.

Type 1

Type 1 calcite is characterized by brown, subhedral hexagonal or rhomboid calcite crystals interspersed throughout the matrix of the glendonite. These rhombs range in size from ∼ 20–200 µm on their long axis, with moderately rounded crystal apices. Crystals often show internal zones of micro-pores and inclusions. Type 1 crystals are ubiquitous throughout glendonites, but are largest and most abundant in macro-porous (mm-scale) internal areas characterized by a granular texture (red areas in Fig. 9). In these areas, Type 1 crystals (with Type 2A overgrowths) are weakly cemented together by Type 3 calcite. The interiors of large Type B glendonite specimens commonly have the majority of their interiors composed of this type of texture, with only a thin rind of Type 2B calcite (Fig. 9E). In many Type A glendonite specimens, interior areas are instead partially to completely occluded by void-filling Type 3 calcite, discussed below (Fig. 9A–D). This calcite type matches the Type 1 calcite phase described by Huggett et al. (2005), and Type I of Vickers et al. (2018).

Type 2A

Type 2A calcite overgrows Type 1 rhombs and is optically distinct from them in thin section (Figs. 10B–E, 11B, C). Optical differences are the result of variations in crystal habit and the abundance of micropores or inclusions; Type 2A is lacking in these micropores compared to Type 1 and is usually characterized by more transparent equant or fibrous calcite. Type 2A calcite is distinct from Type 1 calcite in CL imaging (labeled in Fig. 12) most likely due to minor compositional differences between it and the earlier Type 1 phases. CL and back-scattered electron (BSE) images (Figs. 12 and 13, respectively) reveal additional details that are not clear under visible light, showing that Type 2A calcite commonly has ragged edges (best observable in Figs. 12B–F, 13B, C, E). Not all Type 1 crystals are surrounded by Type 2A calcite; some are in direct contact with Types 2B or Type 3. Type 2A calcite matches the Type II phase described by Vickers et al. (2018), though Huggett et al. (2005) have a wider definition for their “Type 2,” using it to refer to all overgrowths of Type 1 crystals. In Carter Creek glendonites, however, compositional and textural differences between the overgrowths on Type 1 crystals necessitate the separation of overgrowth calcite into Types 2A and Type 3.

Type 2B

Type 2B calcite forms a dark-brown inclusion-rich matrix of irregular crystals, which are much smaller than those of Type 1. Type 2B matrix calcite is commonly found between the Type 1–2A rhombs, and is most common in the outer rims of glendonites (Figs. 9, 10E, F). Boundaries between embedded Type 1–2A and matrix-forming Type 2B calcite are sharp and coincide with crystal boundaries; this is shown well in BSE imaging (Fig. 13B–E). When viewed in CL and BSE images, Type 2B calcite is shown to be composed of a matrix of interlocking crystals with less discrete boundaries than those seen in Types 1 and 2A. These are usually elongate and preferentially oriented in a single direction, often in mm-scale domains where orientations are at differing angles to one another (Fig. 13E). Alternatively, elongate crystals may be at right angles to one another in the same area, forming a lattice-like pattern observable due to the differing extinction orientations and inclusion densities of individual crystals (Fig. 11D, E). This perpendicular fabric also can be seen in CL images through slight color shifts and differences in brightness values (Fig. 12B, C). Elsewhere (most commonly in the interior “cores” of glendonites, as in the opaque areas visible in Fig. 10A), dark inclusions, presumably organic matter (OM), are dense enough to obscure the properties of most individual crystals (Fig. 10F, G). In cut and polished hand samples, areas dominated by Type 2B calcite have an opaque, chalky, white to cream-colored appearance, forming thin rims on the exterior of glendonites and giving the macroscopic crystal its external texture. In thin section under transmitted light, such areas appear dark brown (Figs. 10A, 11A). This calcite sub-phase has not been described in previous studies.

Type 3

In sectioned hand sample, Type 3 calcite forms discrete, translucent, gray-brown patches adjacent to and within areas dominated by lighter-colored Type 2B calcite (Fig. 9). This calcite phase is largely equivalent to some of the outer botryoidal overgrowths called “type 2,” plus the late stage sparry “type 3” described by Huggett et al. (2005) and the “Type III” phase of Vickers et al. (2018), yet here may represent more than one phase of later diagenetic growth. In thin section, Type 3 calcite can be seen to be composed of pore-filling calcite that grows on Type 1–2A rhombs in the form of semi-isopachous, fibrous, radial, sparry cement (Figs. 10B–D, 11B, C). Surficial patterns show a botryoidal habit where macro-scale porosity is not completely occluded. CL imaging reveals prominent margin-parallel zoning in this calcite type (Fig. 12D–F). Although Type 3 calcite nucleates almost exclusively on Type 2A overgrowth crystals, this is not always the case, with rare instances documented where Type 3 calcite grows atop Type 2B (Fig. 11B, C).

The distribution of these calcite phases varies in a consistent way across all observed specimens, and can be observed in sectioned and polished hand samples through their color and texture differences (Fig. 9). Type 1 calcite rhombs with Type 2A overgrowths are ubiquitous in all zones. Areas of weakly cemented Type 1–2A calcite rhombs, commonly in association with porous voids, are present irregularly in the interior of almost all glendonites (Fig. 9). In Type A glendonites, these are usually surrounded by well-cemented zones of predominantly Type 3 calcite, which in turn are surrounded by thin rims dominated by Type 2B calcite (Fig. 9A–D). Type B glendonites generally have smaller proportions of Types 2B and 3 calcite throughout their interiors, instead being composed mainly of granular, porous weakly cemented Type 1–2A rhombs (Fig. 9E). Where Types 2B and 3 calcite are present, however, their pattern of distribution is similar to that seen in Type A glendonites. Such consistency allows for inferences to be made about the nature and sequence of ikaite transformation and replacement.

Geochemical and Isotopic Analysis

EDS maps of Ca, Mg, and Fe element distributions were collected to further characterize the relative geochemical differences among the calcite types (Fig. 14). Types 1 and 3 calcite showed the greatest differences, with Type 1 being relatively enriched in Ca, and Type 3 having significantly higher amounts of Mg and (to a lesser degree) Fe (Fig. 14B–D). Micron-scale banding within Type 3 calcite is visible in maps of both Fe and Mg, but is more prominently expressed by Fe (Fig. 14D). Type 1 calcite also is relatively Ca-enriched compared to Type 2B, and conversely, Type 2B contains higher amounts of Mg relative to Type 1 (Fig. 14C). Type 3 shows higher Fe content than Type 1 (Fig. 14D), but no variation is seen between Types 1 and 2B (Fig. 14H). In both areas mapped, Type 1 crystals show little or no internal variation in element concentration (Fig. 14).

Raman spectroscopy of several polished and sectioned glendonites confirms that CaCO3 forming glendonites is in the form of calcite and not another polymorph (e.g., aragonite; Fig. 15; Supplementary Data Set 2). Raman data were collected from 25 spots across all calcite types from one representative Type A glendonite sample. Spectra from Types 1, 2A, and 3 tend to have a higher signal-to-noise ratios and minimal auto-fluorescence contributions compared to Type 2B. Across the spectral range analyzed, two peaks are observed: the ν4 in-plane bending of the carbonate (CO32–) group located at ∼ 712 cm−1 and the ν1 symmetric stretching mode of the carbonate ion located at 1086 cm−1 (Fig. 15A; Sánchez-Pastor et al. 2016). The full width at half maximum (FWHM) and position of this peak for Types 1, 2A, and 2B form a cluster with lower FWHM and peak positions relative to Type 3, which is significantly offset from Types 1 and 2 (Fig. 15B). These spectral variations are consistently observed between the four types of calcite identified by petrographic observations. Increasing ν1 peak position is indicative of substitution for Ca in calcite (most likely Mg), or increasing MgCO3 content, consistent with the EDS data (Borromeo et al. 2017), and increasing FWHM is usually a result of decreased crystallinity. Types 1, 2A, and 2B ν1 peak positions and FWHMs are suggestive of lower Mg concentrations and higher crystallinities than Type 3.

Although individual calcite types cannot be completely isolated in the volumes needed for isotopic analysis, macroscopic zones dominated by each type were also drilled and analyzed for C and O stable-isotope values. Type 1 and 2A rhombic crystals are therefore present in each sample, either interspersed within Types 2B or 3 or alone as part of poorly cemented, granular zones where Types 2B and 3 are generally absent. Despite this limitation, resolvable differences exist between these calcite zones (Fig. 16). Both granular, weakly cemented Type 1–2A calcite zones and microcrystalline Type 2B calcite zones show consistent δ13C values just below –20‰ VPDB, in both eroded and in situ specimens. Measured δ18O values from areas dominated by Types 1, 2A, and 2B (e.g., Fig. 9 yellow and red areas) range from 0.4 to 1.0‰ VPDB. However, indurated zones dominated by Type 3 calcite (e.g., Fig. 9 blue and green areas) show a different range of both δ13C and δ18O values: δ13C values are largely more positive than other zones by about 5‰, with the exception of a single value measured at –44‰ VPDB (Fig. 16). The δ18O values in Type 3 zones range from ∼ –0.2 to ∼ 1.3. Thus, Type 3 zones can be differentiated from the largely indistinguishable Types 1, 2A, and 2B on the basis of their wider range of values and higher propensity for outliers.

Environmental Interpretations

Sedimentary Depositional Environments

The variety of lithologies and sedimentary structures present at Carter Creek, when combined with biostratigraphic information, allow for process-based interpretations of the depositional environment in which ikaite crystals formed. Marine foraminifera are found throughout the section, with planktic taxa constituting only a small fraction (< 3%) of any given sample, confirming a shallow-marine interpretation for the site. Ratios of infaunal vs. epifaunal benthic foraminifera range from ∼ 46 to 66% in the middle part of the measured section just below the level in which glendonites are found in situ (45–55 m; Supplementary Data Set 1), indicating relatively low bottom-water oxygenation (Jorissen et al. 2007). Epifaunal taxa increase significantly in the uppermost sample analyzed (62 m in Fig. 4), potentially signaling an increase in bottom-water oxygen up-section, consistent with an increasingly proximal depositional environment as inferred from by lithofacies-association interpretations.

The Nuwok Member at Carter Creek is characterized by four sedimentary lithofacies associations interpreted to have been deposited in shallow-marine settings (Table 3). Lithofacies Association 1 dominates the lower parts of the section (0–64 meters in Fig. 4), which consists of silty muds and very fine-grained sands, suggesting that background sedimentation took place in an overall low-energy environment. Vertically oriented chert pebbles in background muds (both lower and higher in the section) likely were deposited as ice-rafted debris (dropstones; Bennett et al. 1996) entrained in seasonal sea ice, an interpretation consistent with known paleoclimate in Arctic Alaska during the early Neogene (Schubert et al. 2017). Pebble lags at the bases of sand beds are indicative of gravitational settling during gravity-flow deposition. These sands are interpreted to be the result of episodic, storm-generated flows that reworked dropstones or transported coarser material farther onto the shelf, though the lack of hummocky cross-stratification and the overall paucity of sand in this part of the section suggests that storms were not a major depositional factor. Additionally, no indicators of shoreface depositional processes were seen (e.g., wave ripples, tidal bedforms). These observations together suggest an environment below fair-weather wave base, potentially within the offshore transition zone or upper shelf.

Lithofacies Association 2, found higher in the section where most glendonites occur, contains features suggestive of a slightly shallower depositional environment. As in other parts of the section, trace fossils are rare or absent. Sand content generally is overall higher, potentially indicating a closer proximity to sources of coarser sediment. Sand beds commonly contain both cm-scale asymmetrical ripples and larger-scale cross-stratified beds (e.g., Fig. 5C) that suggest more sustained high-energy depositional events compared to strata below. Notably, this lithofacies association also contains sedimentary features that are indicative of tidal processes. Heterolithic strata of interbedded mud and sand (Fig. 5G, H) represent bimodal, cyclical changes in depositional flow regime. Double mud drapes (Fig. 5F) also are present, and are typically considered a good indicator of tidal (though also potentially meteorological; see Ainsworth et al. 2012) processes (e.g., Dashtgard et al. 2021). While modern-day tidal range in the Beaufort Sea is relatively low (< 0.3 m; NOAA 2022), tidally generated currents can affect the seafloor far below the low-tide mark, reaching as far as the deepwater environment below the shelf edge (Dykstra 2012). Combined with the lack of definitive, abundant, wave-generated structures, this section is interpreted as having been deposited below fair-weather wave base in the distal part of the lower-shoreface zone.

Lithofacies Association 3 similarly contains rippled and cross-stratified sands, though often in thicker beds or in intervals where sand content is greater than below. Potential tidally generated structures are also present (Fig. 5K), though not as distinct as those seen below, and wave-generated structures are absent. It is therefore likely that Lithofacies Association 3 was deposited in an environment similar to that of the sediments in Lithofacies Association 2, in the lower shoreface zone, but is interpreted to occupy a more proximal position due to the overall increase in sand content observed.

Lithofacies Association 4 is unique in the section in that it consists of decimeter- to meter-scale sand beds, without substantial interstitial mud content. Sands are faintly cross-stratified, and generally lack other sedimentary structures, suggesting that their deposition was the result of a singular or limited suite of processes involving unidirectional flow. Two intervals in the upper part of the section (75–76.5 m and 79–84 m) are assigned to this lithofacies association; both are amongst the thickest beds in the section and likely represent times of extended high-energy conditions where the locality was within the upper shoreface zone, an environment often characterized by multidirectional trough cross-stratification as a result of active wind-, wave-, and storm-generated currents (Pemberton et al. 2012). Abundant lithic fragments in sands, including dark cherty grains, point to a nearby Brooks Range source of sediment and are similar to lithics observed in modern creeks and rivers entering the Beaufort Sea today.

This sequence suggests an overall shallowing-upward succession, with a return to deeper or quieter water conditions around 85 m as Lithofacies Association 4 sharply transitions into Lithofacies Association 2. The stratigraphic distribution of glendonites in the section, with Type B morphologies being found in float originating from Lithofacies Association 1, and Type A occurring in situ and in talus slopes originating mainly from Lithofacies Association 2, suggests an environmental correlation with crystal habit. The lack of glendonite specimens originating from Lithofacies Associations 3 and 4 may be due to initially unfavorable conditions for their nucleation and growth in shallower marine environments, or to warming through time that would bring the locality outside of the ikaite temperature stability window (Bischoff et al. 1993).

Postdepositional and Diagenetic Environments

As glendonites usually form below the sediment–water interface, their postdepositional physical and geochemical environment also must be considered. Studies of the oceanographic setting and pore-water geochemistry of modern sites of ikaite precipitation reveal that these localities have several factors in common, providing insight into the conditions of ancient environments. While low temperatures are the most commonly occurring factor in modern ikaite occurrence (Zhou et al. 2015), other aspects of pore-water chemistry also play critical roles in the initial precipitation of ikaite. Specific aqueous geochemistries differ among sites, but most contain “high alkalinity in combination with chemical inhibitors of the thermodynamically more stable anhydrous carbonate polymorphs, calcite, aragonite, and/or vaterite” (Vickers et al. 2022; p. 202). Such chemical inhibitors may include high concentrations of dissolved Mg2+ and PO43– (suggested by laboratory studies; Purgstaller et al. 2017; Stockmann et al. 2018), or solely PO43– (suggested by studies of glendonites and natural ikaites, e.g., Zhou et al. 2015, and laboratory studies, e.g., Hu et al. 2015, although this may be at odds with some experiments, e.g., Stockmann et al. 2018). High alkalinity and phosphate concentrations in marine pore waters may be the result of microbial interactions with high amounts of OM or methane (Suess et al. 1982; Bischoff et al. 1993; Schubert et al. 1997; Greinert and Derkachev 2004; Zhou et al. 2015), wherein bacteria oxidize organic molecules through sulfate reduction (Peckmann 2017). Early diagenesis of volcanic ash can generate high concentrations of Ca, Mg, and P ions (Jones and Gislason 2008; Olsson et al. 2013), potentially explaining the close association between glendonites and volcanic-ash horizons observed in early Eocene sediments of the Fur Formation in Denmark (e.g., Vickers et al. 2020).

Observations of Nuwok Member sediments shed light on the status of these factors in the Carter Creek section. Fine-grained muds throughout the outcrop contain relatively high amounts of OM (0.5–3.5% TOC; most values between 1 and 2%; Fig. 4), suggesting highly productive surface waters and/or oxygen-depleted bottom waters that favor OM preservation. While programmed pyrolysis (Dreier and Warden 2021) was not conducted for this study, previous work characterizing OM in the Sagavanirktok Formation indicates that it is primarily composed of Type III kerogen (terrestrial material; Magoon et al. 1987), though this does not preclude additional contributions from marine sources. On the modern Beaufort Sea shelf and slope, upwelling is an established process that results in high primary productivity (Pickart et al. 2013). Cold Pacific water (–1.5–2°C) flowing northward into the Chukchi Sea during winter months is redistributed along the northern coast of Arctic Alaska via the Alaska Coastal Current and the Shelfbreak Jet, where it may reverse direction during weather events and result in upwelling of nutrient-rich (i.e., phosphate-rich) basinal water onto the shelf (Pickart et al. 2011; Schulze and Pickart 2012). Importantly, this type of upwelling has been specifically observed in the area of Barter Island (∼ 145° W longitude), where it is highly localized and associated with anomalously high levels of inorganic nutrients in bottom waters (including phosphate), 3–17× greater than areas to the west (Hufford 1974). Given the similarities in continental configurations (Scotese 2016), similar dynamics may have been at work in the late Oligocene. Indeed, glendonites are found in potentially coeval Nuwok Member exposures on Barter Island (Brouwers 1994) in the vicinity of the reported upwelling. Phosphate is a known inhibitor of calcite precipitation and therefore may assist in promoting the precipitation of ikaite over calcite, despite laboratory studies not finding a conclusive link between ikaite precipitation and phosphate concentration (e.g., Stockmann et al. 2018). The presence of glendonites in the Carter Creek–Barter Island area may therefore be related to high phosphate concentrations from upwelling in the area, in combination with near-freezing seawater temperatures and increased OM deposition (and associated bacterial degradation) due to high productivity in nutrient-rich waters. Stable isotopes are consistent with this interpretation: δ13C in glendonites is typically in the range of –15 to –23‰, values that match those of OM in surrounding sediment, suggesting that at least some of the carbon in glendonite calcite is derived from OM. Although these values are much lower than typical dissolved-inorganic-carbon (DIC) values in seawater (indicating some modification of seawater chemistry), the correspondence between 87Sr/86Sr isotope values from glendonites and those of fossils in the section does suggest that waters in which glendonites formed were originally seawater sourced and retained at least the Sr composition of seawater. OM stable-isotope values are fixed well before glendonite formation; therefore, the minor shift in sedimentary δ13Corg at the approximate stratigraphic location of the glendonite interval is unlikely to be directly related.

Numerous other studies have also explored the relationship between ikaite formation and the presence of methane (Schubert et al. 1997; Stein et al. 2003; Greinert and Derkachev 2004; Selleck et al. 2007; Morales et al. 2017; Qu et al. 2017; Hiruta and Matsumoto 2022), as its anaerobic oxidation (AOM) may lead to the formation of bicarbonate (HCO3) ions that increase alkalinity and promote ikaite growth. Whilst these studies do not conclusively prove a direct link between methane and ikaite precipitation (based in part on biomarker and isotopic evidence; Qu et al. 2017; compilation of Rogov et al. 2021, and references therein; Hiruta and Matsumoto 2022), the close proximity of ikaites/glendonites and methane seeps suggest an indirect link or more complex relationship between methane seeps and ikaite than previously thought (e.g., Greinert and Derkachev 2004; Teichert and Luppold 2013; Hiruta and Matsumoto 2022). Methane entrapped within various media (water, ice, and sediments) is frequently found throughout the Beaufort Sea and shelf (Kvenvolden et al. 1993; Lorenson and Kvenvolden 1995), and methane seeps have specifically been reported in Camden Bay, just to the north of the Carter Creek section (Coffin et al. 2013). Additionally, Port (2008) reported evidence for methane seepage specifically at the Carter Creek locality, in the form of a 2–3-m-thick limestone bed near the middle of the outcrop. This bed was interpreted as an authigenic carbonate similar to those found in many localities in modern oceans near methane cold seeps (e.g., Gieskes et al. 2005; Pierre and Fouquet 2007; Pierre et al. 2012; Ruban et al. 2020, and many others). Like many modern examples, carbonate (micrite and isopachous cements) from this bed was characterized by Port (2008) as having very light δ13C values (–73‰) that match most closely with established values for methane and hydrocarbon-derived authigenic carbonates, albeit at the extreme end of the reported ranges (Roberts et al. 2013). The Carter Creek carbonate bed was also reported to contain thiotropic pelecypod and gastropod species known to be associated with chemosynthetic faunal communities on the seafloor, adding to the evidence of a hydrocarbon association. However, during field work conducted for this study, no carbonate beds were found aside from the carbonate-cemented clastic sandstone near the top of the section, and this bed could therefore not be evaluated further. As outcrop conditions may change rapidly on the North Slope due to extreme weather, it is possible that this bed was covered after Port’s (2008) study. As mentioned above, most Carter Creek glendonites have δ13C values in the –15–20‰ range, far heavier than expected if methane had been the carbon source. However, the single value of –44‰ reported in Type 3 calcite is extremely light, and is an indicator of at least partial methanogenic sourcing of glendonite carbon during later stages of mineralization.

Ikaite to Calcite Transformation

Irrespective of the specific conditions leading to ikaite precipitation, all glendonites found at Carter Creek are now composed entirely of calcite. Type 3 calcite contains higher amounts of Mg than other types (Fig. 14), consistent with observations in other glendonites by Vickers et al. (2020) and Schultz et al. (2023). The sequence of mineral growth of all observed calcite types can be deduced through petrographic analysis of polished glendonites, providing an opportunity to evaluate past models of the ikaite-to-calcite transformation sequence in a new setting with an independent geologic history. This transformation can be summarized as having taken place over three primary steps, shown schematically in Figure 17:

  1. Initial precipitation of ikaite takes place within sediment, illustrated as Stage 1 in Figure 17. Zhou et al. (2015) define the ikaite-formation zone (IFZ) by examining concentrations of Ca2+ ions, DIC content, and δ13C values in sediment pore waters, finding that ikaite in the modern deep sea was likely to have precipitated between 2 m and 205 m below the sediment–water interface, with the majority of measurements between 2 m and 10 m. Ikaite initially grew displacively and formed the characteristic morphologies seen in glendonites today (e.g., Schultz et al. 2022). However, the controls on glendonite morphology, including the size of Type B specimens and the number of subsidiary crystals in Type A specimens, remains enigmatic. While Type A glendonites clearly form a continuum in terms of the degree of subsidiary crystal development, whether this represents a sequential growth, or is simply the product of numerous simultaneous crystal nucleation sites, is unknown.

  2. Continued burial and heating resulted in the transformation of ikaite to calcite (Stage 2 in Fig. 17). Vitrinite-reflectance data from nearby wells show that post-Cretaceous strata exposed at the surface in the region experienced only minor postdepositional heating, reaching VRo values of no greater than 0.6% and likely closer to 0.3% or 0.4% (Bird et al. 1999; Houseknecht and Hayba 1999), suggesting a maximum burial temperature of 60–70°C (Barker and Pawlewicz 1986). However, recent thermochronological work utilizing apatite (U-Th)/He to examine the burial and exhumation history of the Marsh Creek Anticline suggests that the Carter Creek locality likely did not even reach these values, and may have been buried only to maximum depths of < 1 km and 20–30° C (W. Craddock, personal communication 2023). Regardless, all such estimates are still above the temperature at which ikaite is considered stable. The current permafrost present across the Alaska North Slope is estimated to have formed between 3 and 2 Ma and may reach depths of up to 550 m in the 1002 area of ANWR (U.S. Geological Survey, Alaska Gas Hydrate Assessment Team 2013); therefore, the transformation of ikaite to glendonite likely took place at some point between ∼ 23 Ma and 3–2 Ma.

    This mineralogical transformation has been studied extensively (Tang et al. 2009; Vickers et al. 2018, 2022; Lázár et al. 2023; Schultz et al. 2023), with some studies finding that ikaite initially decomposes to other carbonate minerals (amorphous calcium carbonate, vaterite, aragonite) as an intermediate step before the complete conversion to calcite (Tang et al. 2009; Purgstaller et al. 2017; Lázár et al. 2023). No indication of these earlier phases is seen in Carter Creek specimens. Several studies note the substantial volume loss associated with ikaite dehydration (c. 2/3; De Lurio and Frakes 1999), leading to porosity creation and a granular texture in the interior of some glendonites that has aided in recognizing them as pseudomorphs (Larsen 1994). This texture is present throughout Nuwok Member specimens, especially Type B glendonites, and is created by weakly cemented subhedral rhombic crystals of Type 1 calcite with Type 2A and small amounts of Type 3 overgrowth. Quantification of volume loss in Nuwok Member specimens (i.e., through point counting) is precluded by: 1) void spaces in some glendonite interiors that is outside of a given thin-section plane, and 2) microporosity in Type 2B calcite that is below the resolution of optical point-counting methods (see Fig. 13C). Textural relationships indicate that Type 1 calcite is the first phase of transformed ikaite; these characteristic rhombs feature an optically dark, homogeneous low-Mg calcite core with zoned high-Mg calcite overgrowths much like those observed elsewhere in ikaite that transformed in air (Schultz et al. 2023). This calcite type is therefore most directly related to the ikaite precursor. Type 2B calcite is interpreted to form simultaneously with the Type 1–2A rhombs, or shortly thereafter, and is also hypothesized to be sourced from an ikaite precursor. Isotopic values of ikaite are interpreted to be retained in Types 1, 2A, and 2B calcite.

    Petrographic analysis reveals rare instances of Type 3 calcite growing atop Type 2B, indicating that it existed by the time Type 2B precipitated (Fig. 11B, C). SEM observations show that the orientation of fabrics constituting Type 3 calcite are largely unaffected by the presence of Type 1–2A rhombs, a characteristic that might be expected if Type 2B had grown at a much later phase than Type 1–2A. Type 2B calcite also contains abundant µm-scale porosity (Fig. 13C–E), a feature that may be related to volume loss from ikaite dehydration as noted above. In this case, the smaller size of both pores and calcite crystals may indicate a more rapid transformation process, in which crystals did not have sufficient time to grow as large as the rhombic crystals of Type 1 calcite. This relationship (between crystal size and transformation speed) has been observed in the past, with “large temperature differences between ikaite and the surrounding environment (producing rapid thermal diffusion) result[ing] in large numbers of very fine crystals rather than granular rhombs” (Larsen 1994). Recent work by Schultz et al. (2023), however, suggests that even with rapid transformation (i.e., at 21° C), a few crystals/rhombs grow large and are surrounded by much smaller blebs that do not show a clear difference between Types 1 and 2A (e.g., Fig. 4C of Schultz et al. 2023). Schultz et al. (2023) also observed a rind of chalky calcite to form when the ikaite transformed at room temperatures, here referred to as Type 2B. Further evidence for Types 1 and 2 forming directly from ikaite lies in their overlapping isotopic and geochemical makeup. Types 1 and 2 show similar Raman spectroscopic characteristics, whereas they show distinct differences from Type 3 (Fig. 15B). EDS elemental mapping shows that Mg content in Type 2 is only slightly higher than Type 1 and has indistinguishable Fe content, in contrast to Type 3, in which stronger differences are present.

  3. Type 3 calcite began to precipitate in remaining pore space epitaxially on (but not in crystallographic continuity with) the surfaces of Type 1 and 2A calcite rhombs. Initially, phases of Type 3 calcite (Stage 3 in Fig. 17) grew with a radial, botryoidal habit leaving substantial pore space unfilled. Continued but incomplete infilling of pore space as sparry crystals that grow atop the previous epitaxial phase completed step 3 (illustration on far right in Fig. 17). These phases are genetically related and may be part of a continuous cementation event, and are thus combined here. Phase 3 was emplaced after the transformation of ikaite, as Type 3 calcite has substantially different geochemistry than Types 1 or 2. Hence, it is unlikely that the elements composing Type 3 calcite were derived from the original ikaite mineral; rather they were introduced as dissolved components in pore fluids. 87Sr/86Sr ratios in areas dominated by this type of calcite are higher than those of other phases, and if porewaters were sourced from seawater, would indicate an early Miocene (Aquitanian) timing for precipitation or porewater isolation. However, isotopic alteration of these porewaters during diagenesis cannot be ruled out—these values may not be reflective of contemporaneous seawater, and any interpretations based on them should be considered hypothetical. Higher Mg content in Type 3 calcite may have resulted from the exclusion of Mg in the earliest phases of transformation and the consequent enrichment in remaining pore waters, which may have subsequently been mixed with later fluids from a new source. As noted above, depletion of 13C in some samples indicates a methane source for carbon in at least some Type 3 calcite samples. Preferential exclusion of magnesium ions from calcite in favor of calcium-ion incorporation during calcite precipitation from a brine is thermodynamically sensible considering the larger enthalpy of hydration possessed by hydrated magnesium ions (Rodriguez-Cruz et al. 1999). In this scenario, Sr values in porewaters would not necessarily be a reliable indicator of timing, as they would not be sourced solely from contemporaneous seawater. Regardless, the precipitation of this calcite phase requires an open system where new pore fluids are introduced, as pore space in glendonites is commonly observed to be completely occluded, which would not be possible with ikaite-derived calcite alone due to the volume loss associated with dehydration.

This sequence appears to be present and well-preserved in all in situ specimens. No further diagenetic phases or changes are seen, aside from some specimens having an eroded or weathered appearance due to dissolution and meteoric-water diagenesis at the surface in recent times.

Global Context and Comparison

A wide variety of glendonites have been identified from other localities throughout the geologic time scale (see Rogov et al. 2021, Schultz et al. 2022, and Rogov et al. 2023 for a review), including many that bear similarities to the types of glendonites described here in both their occurrence and their morphology. Type A glendonites bear particular morphological and mineralogical similarity to other well-documented specimens from Permian strata in Australia (Frank et al. 2008), Cretaceous strata of Norway (Mikhailova et al. 2021), and Paleogene strata of Denmark (Pedersen et al. 1994; Schultz et al. 2020), among others. Type B morphologies also have been found in the Australian Permian (Frank et al. 2008), as well as in Eocene strata of eastern Russia (Rogov et al. 2021), Paleogene strata of Hokkaido, Japan (Shibuya 1977), Jurassic strata of Germany (Teichert and Luppold 2013), and the seafloor of the modern ocean (Bell et al. 2016; Greinert and Derkachev 2004). Other morphologies have also been documented (e.g., “rosette” and “blade with arm” types of Schultz et al. 2022) that were not found at Carter Creek. Glendonites from approximately time-equivalent strata (late Oligocene or early Miocene) have been found in Russia (Gladenkov 1978; Bratzeva et al. 1984; Gladenkov et al. 1987; Serova 2001; Oreshkina 2009), the northwestern USA (Boggs 1972; Nesbitt et al. 2013; Qu et al. 2017), Colorado, USA (Larsen 1994), Japan (Muramiya et al. 2020), and other locations, all of which formed at paleolatitudes over 30° with most over 50° (data from Rogov et al. 2021). In a recent compilation, Rogov et al. (2023) show that these Paleogene to Neogene glendonite occurrences are relatively sparse compared to other intervals in the Mesozoic, and the glendonites reported here add a new data point to this data set. Many of these instances (and others not listed) were also found in lithologies similar to those described at Carter Creek, i.e., some combination of silty muds and sands deposited in a shallow shelfal marine environment (supplemental data of Rogov et al. 2021).

Additionally, ikaite–calcite mineralogical transformation sequences similar to that observed in glendonites of the Nuwok Member have been recognized in other examples around the world. Selected transformation sequences that have been clearly identified in other studies are shown in Table 4, with calcite phases that are generally analogous to one another highlighted by the same color. In particular, Type 1 calcite crystals, commonly with inclusion-rich zoning identical to that described here (but with varying nomenclature), have been identified in many specimens and are often interpreted to be the initial, immediate byproduct of ikaite transformation (Huggett et al. 2005; Frank et al. 2008; Grasby et al. 2017; Morales et al. 2017; Qu et al. 2017; Mikhailova et al. 2021; Rogov et al. 2021; Vasileva et al. 2021; Scheller et al. 2022). Based on published photomicrographs and figures, some authors (Grasby et al. 2017; Vickers et al. 2018; Vasileva et al. 2021) have also included Type 2A calcite in this category, as both are derived from ikaite and may form parts of the same rhombic crystals. Type 2B calcite, with its lattice-like crystal arrangement, however, has not been described in glendonites previously, though Schultz et al. (2023) and Kennedy (2022) describe chalky rims on newly transformed ikaite that likely are similar in origin due to their identical location within the crystal. Many studies of ancient glendonites also recognize equivalents of this study’s Type 3 calcite, especially when it can be seen growing surficially on earlier crystals with a fibrous, radial, or botryoidal habit (highlighted in yellow in Table 4). Huggett et al. (2005), Grasby et al. (2017), and Vickers et al. (2018) all also include a late-stage infilling phase of clear, sparry calcite not considered to be related to ikaite decomposition; this is likely equivalent to the later, sparry Type 3 calcite seen here. Outside of these phases, some models contain other late-stage types of calcite or other minerals that do not have clear analogues with one another (e.g., Mikhailova et al. 2021; Vasileva et al. 2021), as each glendonite occurrence has its own (often unique) set of diagenetic conditions. None of the internal zoning described by Vickers et al. (2018) in glendonites from Svalbard was seen here.

Although specific aspects and nomenclature may differ, these many commonalities indicate that the sequence identified here is, at least in part, widespread across space and time, and may be representative of how the majority of glendonites form. The physical and geochemical conditions responsible for these differences—the specific factors that influence overall calcite type, crystal orientation, size, and the abundance of inclusions—have only recently been investigated in detail (e.g., Schultz et al. 2023). The factors affecting macroscale morphologies (e.g., rosette vs. bladed forms) also remain uncertain, as most experimental work on ikaite growth cannot take place over the time frames required to grow large crystal sizes. Well-preserved examples (at both macroscopic and microscopic scales) in a well-known geologic context are critical for deciphering these controls and interpreting their significance. Additionally, the identification of calcite phases that are most likely to be the direct result of ikaite transformation (here Types 1, 2A, and 2B) has a significant impact on how glendonite analysis may be conducted in the future. These calcite types likely retain many aspects of the parent material, including its isotopic composition, and are the best proxies in the glendonite pseudomorph for original ikaite geochemistry. In studies using glendonites to reconstruct ancient seawater or shallow pore-water properties, including maximum bottom-water paleotemperatures (cf. Vickers et al. 2022), Type 3 calcite and its equivalents should be avoided. Glendonites of the Nuwok Member are therefore important contributors to understanding these and other aspects of glendonite formation, furthering their utility in paleoclimate reconstruction.

Well-preserved glendonites, calcite pseudomorphs after the cold-water mineral ikaite, are abundant in the Nuwok Member of the Sagavanirktok Formation in the northeastern Alaska coastal plain, USA, and are described in detail here for the first time. Glendonites formed in silty and sandy muds that were deposited in a series of shallow shelf to shoreface environments. Biostratigraphic data and isotope stratigraphy place the likely depositional age as late Oligocene, consistent with reflection seismic data tied to biostratigraphic reports in exploration wells within 60 km of the Carter Creek outcrop. Macroscopic glendonite forms include multi-crystal stellate-shaped radial aggregates (termed here Type A glendonites) as well as large singly or doubly terminated blades (Type B), and numerous intermediate forms. Petrographic examination of these reveals a distinct series of microscopic calcite types, the sequence of which can be reconstructed and compared to other ikaite–calcite transformation models. Type 1 calcite, in the form of µm- to mm-scale subhedral, rhombic crystals, is ubiquitous throughout all parts of glendonites and is interpreted to be the initial phase of calcite formed from the dehydration of ikaite. Type 2 calcite, which we subdivide into Types 2A and 2B, also is interpreted to be an early ikaite decomposition product. Type 2A forms zoned overgrowths around Type 1 rhombs, whereas Type 2B consists of smaller inclusion-rich calcite crystals that form the rims and locally the cores of Type A glendonites. The final calcite phase, Type 3, consists of epitaxial, banded cement that grew radially atop Type 1 (and less commonly, Type 2) calcite and filled remaining pore space. Type 3 calcite postdates ikaite transformation and requires open-system behavior during growth. Evidence for open-system behavior is recorded in both isotope and compositional data that show differences between Type 3 and Types 1, 2A, and 2B, as well as in the overall volume of calcite present in glendonites which is greater than expected if only ikaite-derived calcite alone were present. Types 1, 2A, and 3 have analogs with other glendonites described in the literature, though Type 2B appears to be unique to the Nuwok specimens.

Glendonites of the Nuwok Member therefore preserve a wide range of mineralogical and morphological diversity, contributing new information to our understanding of ikaite and glendonite formation and occurrence. This, combined with their lack of diagenetic overprinting or significant thermal alteration, makes them ideal candidates for future studies seeking to utilize these unique low-temperature archives to better understand arctic paleoclimates in the late Paleogene.

This research was funded by the U.S. Geological Survey Energy Resources Program. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. Funding for MLV was also provided by the European Commission, Horizon 2020 (ICECAP; grant no. 10102421), and from the Research Council of Norway through the Centres of Excellence funding scheme, project number 223272. Special thanks to Augusta Warden at the USGS Petroleum Geochemistry Research Laboratory for geochemical and isotopic analysis of sediments, Erik Polluck at the University of Arkansas Stable Isotope Lab for glendonite analysis, Justin Strauss, Chris Connors, and Josh Long for field assistance, and Robert Ravn at The IRF Group, Inc., for palynological analysis. We also extend our thanks to reviewer Matthew Jones. Thanks are also due to the U.S. Fish and Wildlife Services for permission to conduct fieldwork in the Arctic National Wildlife Refuge.

Supplemental files are available from the SEPM Data Archive: https://www.sepm.org/supplemental-materials.

  • Supplementary Data Set 1: Paleontological data included in this study. Taxonomic lists for both pollen and foraminifera are included in separate tabs.

  • Supplementary Data Set 2: Isotopic (13Ccarb, 13Corg, 18O, and 87Sr) and Raman data included in this study.

  • Supplementary Data Set 3: Results of powder X-ray diffraction analysis of glendonites.

Supplementary data