Insights into the metamorphic history and origin of flake graphite mineralization at the Graphite Creek graphite deposit, Seward Peninsula, Alaska, USA

Graphite Creek is an unusual flake graphite deposit located on the Seward Peninsula, Alaska, USA. We present field observations, uranium-lead (U–Pb) monazite and titanite geochronology, carbon (C) and sulfur (S) stable isotope geochemistry, and graphite Raman spectroscopy data from this deposit that support a new model of flake graphite ore genesis in high-grade metamorphic environments. The Graphite Creek deposit is within the second sillimanite metamorphic zone of the Kigluaik Mountains gneiss dome. Flake graphite, hosted in sillimanite-gneiss and quartz-biotite paragneiss, occurs as disseminations and in sets of very high grade (up to 50 wt.% graphite), semi-massive to massive graphite lenses 0.2 to 1 m wide containing quartz, sillimanite, inclusions of garnet porphyroblasts, K-feldspar, and tourmaline. Restitic garnet, sillimanite, graphite, and biotite accumulations indicate a high degree of anatexis and melt loss. Strong yttrium depletion in monazite, high europium ratios (Eu/Eu*), and excursions of high strontium and thorium concentrations are consistent with biotite dehydration melting. Monazite and titanite U–Pb ages record peak metamorphism from ~ 97 to 92 million years ago (Ma) and a retrograde event at ~ 85 Ma. Raman spectroscopy confirms the presence of carbonaceous material and highly ordered, crystalline graphite. Graphite δ13CVPDB values of − 30 to − 12‰ and pyrrhotite δ34SVCDT values of − 14 to 10‰ are consistent with derivation from organic carbon and sulfur in sedimentary rocks, respectively. These data collectively suggest that formation of massive graphite lenses occurred approximately synchronously with high-temperature metamorphism and anatexis of a highly carbonaceous pelitic protolith. Melt extraction and fluid release associated with anatexis were likely crucial for concentrating graphite. High-temperature, graphitic migmatite sequences within high-strain shear zones may be favorable for the occurrence of high-grade flake graphite deposits.


Introduction
Natural graphite is characterized by chemical inertness, high electrical conductivity, thermal stability, and high lubricity. This unique combination of properties makes it ideal for a wide range of industrial and technological applications. The United States of America (USA) is 100% net import reliant for graphite as of 2020 (U.S. Geological Survey 2021). China supplies 33% of US imports and accounts for over 60% of global production of ~ 1.1 million t. Consequently, the US government has designated graphite as a critical mineral (Fortier et al. 2018). Flake graphite-a highly ordered and crystalline form-is emerging as a particularly important source of spherical graphite used in the anode of lithium-ion batteries that power modern portable electronics, electric Editorial handling: G. Beaudoin. 1 3 vehicles, and renewable energy storage systems (Simandl et al. 2015;Robinson et al. 2017a). Demand for battery graphite in 2040 is expected to be 25 times higher than in 2020 (International Energy Agency 2021). Thus, identifying potential flake graphite resources in North America and elsewhere is crucial for meeting future needs.
Unlike metallic commodities, graphite is an industrial commodity characterized by a range of properties and industrial uses that are dependent on the geologic environment in which it forms. The major commercial classifications of graphite are amorphous (a mixture of graphite and carbonaceous material), lump and chip, and flake graphite; geologic environments in which they form overlap. Lump and chip and flake graphite are considered crystalline graphite. Amorphous graphite deposits typically result from contact metamorphism of coal-bearing sedimentary rocks. Lump and chip deposits generally take the form of large (up to 3 m wide) hydrothermal veins containing coarse graphite crystals together with quartz and other minerals such as tourmaline. The veins are typically hosted in granulite-facies metamorphic terranes and typically formed after peak metamorphism. Examples of these are in central New Hampshire, USA , and Sri Lanka (Katz 1987;Silva 1987). Flake graphite deposits in contrast are formed in regional metamorphic sequences ranging from upper amphibolite to granulite grade, coeval with peak metamorphism, and may also be found in the same districts as vein deposits. Texturally, flake graphite deposits vary from disseminations to high-grade (> 50 wt.%) concentrations in pods or lenses that are typically focused along lithologic contacts and within fold hinges (e.g., Lac Knife and Asbury, Canada; Simandl et al. 2015;Senja, Norway;Gautneb et al. 2020).
Lump and chip vein deposits appear to derive carbon from deep-seated supracrustal or mantle sources during the late stages of high-temperature metamorphic events Katz 1987;Silva 1987;Huizenga and Touret 2012;Luque et al. 2014;Touret et al. 2019). Flake graphite deposits, on the other hand, derive carbon mainly from supracrustal sources and are formed by a range of mechanisms that include graphitization and structural focusing and strain partitioning of organic carbon-rich rocks during metamorphism, decarbonation/devolatilization of marble, or, possibly, from a combination of these processes together with the involvement of H 2 O-CO 2 fluids (e.g., Dissanayake and Chandrajith 1999;Luque et al. 2014;Simandl et al. 2015;Robinson et al. 2017a;Rosing-Schow et al. 2017;Gautneb et al. 2020;Palosaari et al. 2020).
The Graphite Creek graphite deposit, located in the Kigluaik Mountains 60 km north of Nome on the Seward Peninsula, Alaska, is the largest known flake graphite resource in the USA and is among the largest in the world (Table 1; Harrington 1919;Coats 1944;King et al. 2019). The deposit is ~ 5 km in length by 0.2 km in width (strike and dip: ~ 70°/40-60° NW) based on the extent of the inferred resource (King et al. 2019). The region has potential to host other flake graphite deposits. However, a lack of understanding of the geological processes that controlled the genesis of ore-grade flake graphite has inhibited development of a predictive mineral system model for this deposit type. In this contribution, we present and synthesize field observations, U-Pb monazite petrochronology, U-Pb titanite geochronology, carbon and sulfur stable isotope geochemistry, and graphite Raman spectroscopy data that constrain the regional metamorphic history and help to elucidate the origin of high-grade flake graphite enrichment in the Kigluaik Mountains. Exploration implications and mappable criteria suitable for evaluation of flake graphite potential in areas with similar geologic environment are discussed.

Tectonic framework
The Seward Peninsula and Graphite Creek deposit lies within the Arctic Alaska-Chukotka microplate (AACM), which includes Chukotka in Russia, the Brooks Range in northern Alaska, and the Seward Peninsula (Fig. 1). The AACM includes a variety of pre-Mississippian basement types. The North Slope and easternmost Brooks Range are underlain by parts of the Laurentian craton (Strauss et al. 2013), whereas the southwestern Brooks Range, Seward Peninsula, and Chukotka are in part underlain by what is thought to be a Neoproterozoic component of the Baltica craton emplaced onto Baltica in the latest Proterozoic to the earliest Cambrian Gottlieb et al. 2017;Miller et al. 2017, and references therein). The Baltican and Laurentian components of the AACM were juxtaposed after closure of the Iapetus ocean during the Caledonian orogeny that culminated at ca. 430 Ma . Following the Caledonian orogeny, the pre-Devonian passive margin became a subducting plate boundary dipping toward the craton beneath the AACM, and a Late Devonian (ca. 385-365 Ma) volcanic arc formed across both the Alaska and Chukotka parts of the AACM. The Late Devonian arc extended from the present-day eastern Brooks Range (Hitzman et al. 1986;Moore et al. 1994;Newberry et al. 1997) to the Seward Peninsula (Till et al. 2011) and across Chukotka . Rifting in the latest Devonian to Mississippian resulted in development of independent extensional basins in northern Arctic Alaska (Young 2004;Nelson et al. 2013), and a regionalscale ocean to the south (present coordinates) of this extensive magmatic arc (Mayfield et al. 1988;Moore et al. 1994;Miller et al. 2017). Volcanogenic massive sulfide deposits associated with rifting of this arc are widespread (e.g., Ambler district; Hitzman et al. 1986;Newberry et al. 1997). This ocean, referred to as the South Anyui ocean in Russia and the Angayucham ocean south of the Brooks Range, expanded southward through the Triassic. During the Late Devonian to Early Jurassic, the Seward Peninsula-Brooks Range component of the AACM had a passive margin depositional environment (Moore et al. 1994;Nelson et al. 2013;Miller et al. 2017).
In the Early to Middle Jurassic, the Anyui-Angayucham ocean south (present coordinates) of the AACM closed Fig. 1 Map showing the modern tectonic framework of the Arctic Ocean, including the Baltican, Siberian, and Laurentian cratons, the Timanide basement of the Baltican craton, the Caledonian orogenic belt, and belts of tectonostratigraphic terranes that were accreted to the cratons in the Paleozoic, Mesozoic, and Cenozoic (simplified from Miller et al. 2017 andStrauss et al. 2013 (Wirth et al. 1993;Harris 1995) overrode the passive margin, initiating the Brookian orogeny (Patrick 1988). Northvergent imbricated panels of the passive margin deposits of the Brooks Range beneath the arc represent significant horizontal shortening and structural thickening of more than 25 km (Patrick 1988;Patton and Box 1989;Moore et al. 1994, and references therein). Resultant blueschist-facies metamorphism of the deeply buried passive-margin deposits was followed by decompression related to greenschist facies metamorphism; these events are recorded in rocks of the southern Brooks Range and Seward Peninsula (Forbes et al. 1984;Armstrong et al. 1986;Evans and Patrick 1987;Patrick 1988;Till et al. 2011). On the Seward Peninsula, blueschist-facies metamorphism of the Brookian orogeny is overprinted by Barrovian metamorphism related to formation of the Kigluaik gneiss dome Amato and Miller 2004). High-temperature (HT) metamorphic overprint on high-pressure/low-temperature (HP/LT) metamorphism in the Kigluaik Mountains is constrained by intrusion of 110-105 Ma granite and growth of 96-92 Ma monazite in orthogneiss and paragneiss (Amato et al. 1994;Amato and Wright 1998).
Diapiric rise of mantle-derived magma is invoked in part as the cause of exhumation of the Kigluaik gneiss dome (Amato et al. 1994;Calvert et al. 1999). Amato and Miller (2004) suggested that northward subduction of oceanic crust from the Pacific Ocean plate beneath the Seward Peninsula contributed to magma generation and HT metamorphism from 120 to 90 Ma. Northeast-trending alkaline mafic dike swarms are common in high metamorphic-grade exposures on the Seward Peninsula, and in the Kigluaik Mountains, such dikes were dated as young as ca. 83 Ma (biotite 40 Ar/ 39 Ar thermochronology; Amato et al. 2003). This evidence indicates that brittle regional extension followed the formation of ductile extensional structures and associated decompression melting, crustal thinning, and exhumation of the gneiss domes to shallow crustal depths (< 4 km). Apatite fission-track dating constrains the timing of this exhumation from ca. 100 to 70 Ma . Northwest-southeast-directed extension on the Seward Peninsula is attributed to tectonic exhumation of thickened continental crust and southward migration of the Cretaceous arc above a north-dipping subduction zone (Amato et al. 2003Amato and Miller 2004). Jurassic and older arcs that overrode the AACM were intruded by magmas associated with south-facing arcs that migrated southward over oceanic crust of the northward-subducting Pacific plate beginning as early as ca. 135 Ma and continued until ca. 55 Ma Pease et al. 2017;Box et al. 2019).

Geology of the Kigluaik Mountains
The Kigluaik Mountains represent an amphibolite-to granulite-facies gneiss dome that formed in the Late Cretaceous on the Seward Peninsula. The rocks on the Seward Peninsula are divided into two map units that consist of (1) a blueschist-greenschist facies unit and (2) a greenschistamphibolite-granulite facies unit exposed mainly in the Kigluaik, Bendeleben and Darby Mountains. Both units contain Neoproterozoic intrusive rocks, indicating that both units have some Neoproterozoic or older protoliths. Amato and Miller (2004) and Till et al. (2011) assigned the blueschist-greenschist facies rocks to the Nome Complex ( Fig. 2a; unit PzZn). Moffit (1913) designated the higher grade, amphibolite-granulite-facies rocks as part of the Kigluaik Group. Amato et al. (2009) proposed the informal name, Kigluaik metamorphic complex, for the higher-grade rocks (Fig. 2b). Till et al. (2011) assigned these rocks to an unnamed Proterozoic-Paleozoic unit that is mapped in the Kigluaik, Bendeleben, and Darby Mountains (Fig. 2a). For this study, we follow the nomenclature of Amato et al. (2009) and refer to the metamorphic rocks of the Kigluaik Mountains as the Kigluaik metamorphic complex. The timing of metamorphic events from the Neoproterozoic to the early Mesozoic are poorly constrained due to the overprint of Cretaceous HT metamorphism.
The Kigluaik Mountains are composed of a variety of metasedimentary, metaigneous, and metavolcanic rocks. Blueschist-facies rocks, overprinted by greenschist facies mineral assemblages, include pelitic schist, chlorite-white mica-albite schist, marble and impure marble, quartzite, and metabasite and are assigned to the Nome Complex (Till et al. 1986(Till et al. , 2014b. Rocks in the underlying Kigluaik metamorphic complex include (1) dominant biotite-rich pelitic schist that contains quartz, biotite, disseminated graphite, muscovite, garnet, staurolite, sillimanite, and layers of graphitic schist; (2) quartzite, quartz schist, and calc-silicate-bearing psammitic schist composed of quartz, biotite, plagioclase, disseminated graphite, sillimanite, actinolite, diopside, epidote/clinozoisite, and white mica; (3) layered (1-100-m-thick layers), coarse-grained, locally  Amato and Miller (2004). Heavy black line is Kigluaik normal fault, dashed where inferred; ball is on downthrown side. Bright pink lines are isograds: line with one tickmark represents biotite-in isograd; line with two tickmarks is staurolite-in isograd; line with three tickmarks represents sillimanite-in isograd; line with four tickmarks is sillimanite + K-feldspar-in isograd. Star indicates location of the Graphite Creek flake graphite deposit. All samples from this study were collected within ~ 1 km of the star ◂ 1 3 migmatitic pelitic paragneiss that contain quartz, biotite, plagioclase, K-feldspar, sillimanite, garnet, and graphite; and (4) massive gray calcite marble and interlayered calcite marble, dolomitic marble, calc-silicate-bearing marble, and biotite schist and gneiss (Amato and Miller 2004). At least some of these rocks were intruded by ca. 565-555 millionyear-old Neoproterozoic plutons ( Fig. 2b; Amato et al. 1994;Amato and Wright 1998) prior to metamorphism. It is reasonable to assume that some of these units and at least the marbles were once marine deposits, but due to melt loss in much of the migmatitic gneiss in the dome, the nature of the protoliths of the quartz-biotite-plagioclase paragneiss remains undefined.
The Paleozoic and Proterozoic rocks of the Nome Complex, and their possible equivalents in the Kigluaik gneiss dome, were inferred to experience to blueschist-facies metamorphism in the latest Jurassic to Early Cretaceous based on phengite 40 Ar/ 39 Ar plateau minimum ages from the Nome Complex ranging from ca. 125 to 116 Ma (Forbes et al. 1984;Thurston 1985;Hannula and McWilliams 1995;. Evidence for this blueschist-facies event, designated as M 1 by , is not preserved in the high-grade core of the Kigluaik gneiss dome. The M 1 event was followed by deformation, metamorphism, and rise of the gneiss dome that condensed metamorphic mineral isograds during decompression and thermal overprinting. Thermal overprint led to greenschist-, amphibolite-, and granulite-facies conditions, respectively, from the flanks to the core of the dome (Fig. 2) in the early Late Cretaceous (Patrick and Lieberman 1988;Patrick and Evans 1989;Miller et al. 1992;Amato and Miller 2004). This higher-temperature Barrovian event (Miller et al. 1992), assigned M 2 by , caused significant partial melting of the rocks within the sillimanite-K-feldspar zone of the gneiss dome. M 2 metamorphism was broadly coeval with intrusion of gabbro, granodiorite, and granite of the Kigluaik pluton ( Fig. 2b; unit Ki) at ca. 90 Ma (Amato and Miller 2004). Mafic magmatism associated with the Kigluaik pluton may have provided the heat source for HT metamorphism and partial melting (Amato et al. 1994). The progression from greenschist-to granulite-facies conditions preserved in rocks across the Nome Complex to the Kigluaik metamorphic complex during M 2 yields conditions ranging from 0.35-0.45 GPa and 525-575 °C for the staurolite-in isograd to 0.8-1.0 GPa and 800-850 °C in the core of the gneiss dome at Mt. Osborne (Lieberman 1988). Subsequent rapid uplift is reflected in white mica 40 Ar/ 39 Ar plateau ages of ca. 85-82 Ma (Calvert et al. 1999). Asymmetric exhumation was accompanied by extensional normal faulting along the Kigluaik fault on the north flank of the Kigluaik gneiss dome (Hudson and Plafker 1978;Dumitru et al. 1995). Exhumation resulted in the unroofing of amphibolite-to granulite-facies, graphite-quartz-biotite-garnet-sillimanite gneiss, graphitic schist, and anatectic migmatite of the Kigluaik metamorphic complex that host the Graphite Creek deposit at present levels of exposure. This extensional faulting is attributed to regional extension beginning in the early Cenozoic (Miller et al. 1992;Dumitru et al. 1995;Rubin et al. 1995); the Kigluaik fault remains seismically active (Hudson and Plafker 1978).

Methods
This study was based primarily on drill core logging and petrography from samples collected from drill holes across ~ 2 km of the deposit. The drill holes are generally 300 m or closer in spacing. To study Graphite Creek samples for determining their protolith, their depositional and metamorphic history and, to understand the mineralizing processes, we applied a combination of carbon and sulfur stable isotope analyses by magnetic sector-inductively coupled plasma-mass spectrometry (MS-ICP-MS) to graphite and pyrrhotite, respectively, monazite petrochronology by laser ablation split stream (LASS) methods, and graphite Raman spectroscopy techniques. Sample numbers indicate the drill hole number (e.g., 14GCM010) followed by the depth in meters. Sample coordinates and descriptions are in electronic supplementary materials (ESM) Table S1.

Stable isotopes
Samples of disseminated graphite and massive graphite in quartz-biotite gneiss, sillimanite gneiss, and pegmatite were collected for carbon isotope analysis. Samples of disseminated pyrrhotite in quartz-biotite gneiss and pegmatite were collected for sulfur isotope analysis. To identify any spatial variability, samples were collected from drill holes across the deposit and from the top and bottom of individual drill holes. Samples were prepared and analyzed at the U.S. Geological Survey (USGS), Geology, Geophysics, and Geochemistry (G3) Science Center Stable Isotope Laboratory, Denver, Colorado. Graphite was removed from drill core pieces using a dissecting needle under a binocular microscope. Where coarse enough, sulfide minerals were sampled using a carbide-tipped handheld drill. In rocks where the sulfides were too fine to drill, the sulfur was extracted chemically from 0.2 g of powdered rock using the chromium reduction method (Canfield et al. 1986). Isotopic compositions were determined by elemental analyzer-isotope ratio monitoring mass spectrometry using Micromass Optima (carbon) and Thermo Delta Plus (sulfur) mass spectrometers (see Johnson et al. 2018). The instruments were calibrated by analyzing the USGS 24 graphite, NBS 21 graphite, NBS 123 sphalerite, and IAEA-S-3 silver sulfide standards taking accepted values from Brand et al. (2014). Results are expressed in δ notation relative to Vienna Peedee belemnite (VPDB) or Vienna Cañon Diablo troilite (VCDT). Reproducibility was ± 0.1‰ or better for carbon and ± 0.3‰ or better for sulfur.

Titanite U-Pb geochronology
Metamorphic titanite from calc-silicate-rich layers in quartzbiotite gneiss were analyzed in situ in polished thick sections. The calc-silicate layers are sporadic, typically less than 10 cm thick and composed of hornblende ± clinopyroxene, plagioclase, titanite, and calcite. Titanite analysis was conducted at the USGS G3 Science Center. Titanite was ablated with a Teledyne-Photon Machines Excite™ 193 nm ArF excimer laser that was coupled to a Nu Instruments AttoM high-resolution magnetic-sector inductively coupled plasma mass spectrometer in spot mode (150 total bursts) with a repetition rate of 5 Hz, laser energy of ~ 3 mJ, and an energy density of 4.11 J/cm 2 . Pit depths are typically less than 10 µm. The rate of He carrier gas flow from the HelEx cell of the laser was ~ 0.6 L/min. Make-up Ar gas (~ 0.2 L/min) was added to the sample stream prior to its introduction into the plasma. Nitrogen with flow rate of 5.5 mL/min was added to the sample stream to allow for significant reduction in ThO + /Th + (< 0.5%) and improved the ionization of refractory Th (Hu et al. 2008). The laser spot sizes for titanite were ~ 40 µm. With the magnet parked at a constant mass, the flat tops of the isotope peaks of 202 Hg, 204 (Hg + Pb), 206 Pb, 207 Pb, 208 Pb, 232 Th, 235 U, and 238 U were measured by rapidly deflecting the ion beam with a 30 s on-peak background measured prior to each 30-s analysis. Raw data were reduced offline using Iolite™ 2.5 U_Pb_Geo-chronology3 Data Reduction Scheme (Paton et al. 2011) to subtract on-peak background signals, correct for Pb isotope fractionation and U-Pb downhole fractionation, normalize the instrumental mass bias, and determine the content of U, Pb, and Th. The NIST-612 glass standard was analyzed in each session to evaluate U and Th contents. Three analyses each of Titanite reference materials BLR-1 (Aleinikoff et al. 2007) and OLT-1 (Kennedy et al. 2010) and standard reference material NIST-612 were analyzed after each five titanite unknowns. Data reduction for titanite is complicated by the presence of variable amounts of common Pb. We used the Neymark et al. (2018) approach that uses NIST 612 as a primary reference material. Biased U-Pb isotopic dates were calculated for the BLR-1 titanite (matrix-matched secondary reference material) measured in the same analytical sessions as unknown samples by constructing a Tera-Wasserburg (T-W) isochron diagram (Tera and Wasserburg 1972) using Isoplot 3.75 (Ludwig 2012). The difference between the biased U-Pb age and the "true" age of BLR-1 reference material is used to correct for the matrix effect and the instrumental bias. A more detailed discussion of the analytical and data reduction methods can be found in Neymark et al. (2018).

Monazite petrochronology
Four drill core samples were selected for detailed wavelength dispersive spectrometry (WDS) mapping and subsequent in situ LASS monazite petrochronology. Three samples were from sillimanite paragneiss (Unit A), and one was from a strongly tectonized leucogranite sheet.
Polished thin sections were imaged using a Jeol JXA-8530F Electron Microprobe in the Advanced Instrumentation Laboratory at the University of Alaska Fairbanks. Full thin section mapping was undertaken to identify all monazite crystals and obtain a cursory understanding of garnet zonation, fabric relationship with porphyroblasts, and textural distribution of phases relative to graphite layers. Full thin section mapping was performed at 20 keV using a defocused 300 nA beam. Pixel size was set to 30 μm with a dwell time of 25 ms, and spectrometers were set to collect Ca Kα, K Kα, Ce Lα, Y Mα, and Mg Kα X-ray lines. The product images were processed to maximize contrast and overlain to identify the locations of all monazite within the thin section using three maps (Ca, K, and Mg) to provide textural context. Monazite crystals for subsequent work were selected to encapsulate the variability present in the thin section including grain size and textural setting. Submicronscale WDS mapping of individual monazite crystals was performed using 15 keV accelerating voltage and a 200 nA focused electron beam set in beam mapping mode, where the stage is fixed and beam is rastered across a region small enough to maintain Bragg condition. X-ray lines chosen for monazite characterization were Ca Kα, Nd Lα, Th Mα, Si Kα, and Y Lα. This method preserves submicron-scale spatial resolution.
In situ LASS analysis of monazite grains was performed in the Mineral Exploration Research Centre-Isotope Geochemistry Laboratory (MERC-IGL) at Laurentian University, Sudbury, Canada. All grains were analyzed using a 10 μm beam, U-Th-Pb isotopes were collected on a Thermo Neptune Plus multicollector, and geochemical (major and trace elements) analyses were performed simultaneously suing a Thermo iCap-TQ inductively coupled plasma mass spectrometer. Ages calculated were based on textural setting and chemical composition. Weighted averages calculated include all data from each defined textural and compositional population. As a result, large mean squared weighted deviation (MSWD) values are associated with calculated ages, which are left as is and interpreted to indicate a broader age range and perhaps a protracted geologic process. Additionally, recent studies have demonstrated decoupling of U-Pb isotopic record 1 3 and chemical composition of monazite during simple shear twin development and low temperature fluid-mediated dissolution-reprecipitation reactions (Fougerouse et al. 2020;Weinberg et al. 2020). These processes can generate larger spread in calculated U-Pb ages for individual compositional and textural populations.

Raman spectroscopy
Five polished thin sections from drill core samples were analyzed for textural attributes and to characterize the degree of ordering of carbonaceous material using Raman microspectroscopy (Beyssac et al. 2002). Microtextural examination was completed on each sample using optical transmitted and reflected light microscopy at magnifications of 10 × to 50 × . The analyses were undertaken to determine if there is graphite versus carbonaceous material (CM), i.e., highly ordered crystalline carbon versus less ordered carbon, and if these materials have distinct textural attributes.
Raman spectroscopy was performed on a HORIBA LabRAM HR Evolution spectrometer at the Raman Microspectroscopy Laboratory, Department of Geological Sciences, University of Colorado, Boulder. The spectrometer was calibrated with silicon at the beginning of the session. Spectra were collected at 50 × or 100 × magnification with a 532 nm excitation laser. The laser power was reduced to 3 mW at the sample surface. The microscope was autofocused at each point in order to maximize the signal of the G peak (1580 cm −1 ). Two acquisitions lasting 15 s each were collected, screened for cosmic ray spikes, and averaged to produce a final spectrum at representative points in each thin section. Spectra of graphite and CM were collected from locations where they were protected from thin section polishing by an overlying transparent mineral such as quartz, in order to avoid artifacts introduced by polishing (Beyssac et al. 2003).
Data processing was performed using the computer software LabSpec 6 (HORIBA Scientific). A polynomial baseline (degree 8) was fit and subtracted from the spectra. The spectral features attributable to graphite and (or) CM were then fit by a combination of up to four pseudo-Voigt line shapes, including the G peak (1580 cm −1 ), D1 peak (1325-1360 cm −1 ), and D2 peak (1600-1620 cm −1 ). In some spectra, two pseudo-Voigt peaks were required to accurately fit the shape of the D1 peak, and the areas were summed to give a total D1 peak area. D3 peaks were not detected in any spectra. The peak area ratio (R2) was calculated as R2 = D1 / (D1 + D2 + G), and the temperature was determined using the geothermometer of Beyssac et al. (2002). The R2 ratio, which measures the area of the D1 peak near 1350 cm −1 as a fraction of the total area under the D and G peaks, depends on the degree of disorder in the carbonaceous material and is thus a function of thermal maturity.

Paragneiss
The Graphite Creek graphite deposit is hosted in two distinct migmatitic paragneiss units, informally defined here as Unit A and Unit B. Unit A is a strongly foliated, sillimanitequartz-feldspar-biotite-graphite-garnet gneiss (Fig. 3a-d); the relative abundance of these minerals varies throughout and over centimeter scales. Intervals characterized by 10-30 vol.% sillimanite, biotite, and disseminated graphite, and minor quartz, are common throughout. Graphite may form discreet millimeter-scale bands (Fig. 3b). Intervals (10-20 cm thick) comprising garnet porphyroblasts surrounded by sillimanite are less common (Fig. 3c, d). Quartz is less abundant in these mineral concentrations. Unit B is a quartz-feldspar-biotite gneiss with minor (< 3 vol.%) graphite and pyrrhotite, and very little garnet or sillimanite (Fig. 3e). This unit may be massive or banded. Its foliation is less pronounced than Unit A owing to a disseminated rather than segregated distribution of biotite. Unit B contains local, thin (< 0.5 m), green calc-silicate layers composed primarily of albite, calcite, and hornblende ± clinopyroxene with minor titanite and pyrrhotite (Fig. 3f). The contact between the units A and B is typically gradational and characterized by the presence of garnet and sillimanite as minor and to trace amounts, respectively, in Unit B. Sillimanite-and graphite-bearing leucosomes and biotite-rich melanosomes are abundant in both units. Garnet porphyroblasts in Unit A are commonly associated with, and in places cut and surrounded by, quartzo-feldspathic leucosomes.

Igneous rocks
Igneous rocks in the study area are syn-kinematic pegmatitic granite (SKPG), late kinematic granite and pegmatite (LKG), and post-kinematic aphanitic diabase dikes. The SKPGs are generally concordant with the gneissic foliation and are characterized by strongly recrystallized contacts (Fig. 3g). Some SKPGs contain minor pyrrhotite and very coarse (5-10 mm) flake graphite (Fig. 3h). K-feldspar crystals display parallel fractures (Fig. 3h). The pegmatite dikes range in apparent width from ~ 5 to 2 m. Some leucosomes coalesce to form pegmatite dikes. The LKGs are medium grained, equigranular, and are associated with minor pegmatite dikes composed of quartz-feldspar-tourmaline ± graphite (Fig. 3i). LKGs are discordant but typically have diffuse contacts indicating some recrystallization. Late diabase dikes (not shown) are volumetrically minor, aphanitic, and characterized by sharp contacts with all rock types.

Graphite mineralization
Graphite mineralization is characterized by disseminated and semi-massive to massive textures. Disseminated flake graphite is present in every rock type except the basalt dikes. The abundance of disseminated graphite (reported in weight percent in this section) varies from 2 to 8 wt.% in Units A and B. Local extreme enrichments of graphite (> 20-50 wt.%; King et al. 2019) occur as semi-massive to massive lenses confined exclusively to the sillimanite gneiss (Unit A; Fig. 4a). The largest continuous lens exposed in outcrop is at least 7 × 6 × 1 m. Lenses pinch and swell but are generally concordant with the gneissic fabric. There is no evidence of folding or repetition of lenses in drill core or outcrop. Lenses have lesser amounts of sillimanite, biotite, quartz, feldspar, and euhedral tourmaline (Fig. 4b-d;   Fig. 5a). Graphite, biotite, and sillimanite define a strong foliation within the semi-massive intervals (Fig. 4e). Some lenses contain zoned garnet porphyroblasts and inclusions and slivers of adjacent wall rock (Fig. 4f). Pegmatitic granite (SKPG) is associated with some lenses (Fig. 4a, b). Both coarse-and finegrained graphite occur together in the same samples (Fig. 5b). Flake graphite is also intergrown with quartz in leucosomes and in the semi-massive lenses (Fig. 5c, d).

Monazite petrochronology and titanite geochronology
Monazite yttrium (Y) and thorium (Th) WDS maps, thin section magnesium (Mg) WDS maps, and monazite rare earth element (REE) chemistry and age data are shown in   Table 2. Complete monazite trace element chemistry is in ESM Table S2. Monazite features well-annealed microstructures and include asymmetric biotite and graphite pressure shadows developed around garnet porphyroblasts (Fig. 6a, 14GCM010-20.1). Monazite grains are relatively coarse (Table 2) and display systematic compositional zonation and are here defined as comprising inner cores, outer cores, inner rims, and outer rims (Fig. 6b). Some monazite grains (outer cores) contain inclusions of flake graphite oriented parallel to graphite foliation in the host rock (Fig. 6b, sample 14GCM010-28.4).
Sample 12GC002-44.9 is a leucogranite orthogenesis composed of biotite, K-feldspar, quartz, apatite, and sodic plagioclase. Monazite crystals from this sample are 100-200 μm in diameter (Fig. 6b) Sample 14GCM010-20.1 is a flake graphite-bearing migmatitic paragneiss containing coarse garnet porphyroblasts in a strongly deformed matrix of K-feldspar, quartz, biotite, graphite, sillimanite, apatite, monazite, and xenotime. Graphite forms thick mats that are semicontinuous, showing pinch and swell structures together with biotite mica. Garnet porphyroblasts are characterized by an increase in Mg from cores to rims (Fig. 6a). High-Mg garnet is associated with inclusions parallel with gneissic fabric in the matrix, suggesting that garnet growth overlapped with, or outlasted, deformation. Monazite crystals from the sample are relatively fine-grained and have three compositional domains. Inner cores are low in Y and yield a 206 Pb/ 238 U weighted average age of 95.94 ± 1.03 Ma (MSWD = 4.91; n = 16; Fig. 6c; Table 2). Thin zones of high-Y monazite occur as thin outer cores, or make up the cores of other crystals, and yield a 206 Pb/ 238 U weighted average age of 96.53 ± 1.17 Ma (MSWD = 5.86; n = 12). Rim Y compositions return to inner core Y values and yield a 206 Pb/ 238 U weighted average age of 90.50 ± 0.60 Ma (MSWD = 0.93; n = 10).
Graphite in this sample is distributed as thick layers that define the gneissic compositional layering. Mapping shows minimal major element zonation in the garnet porphyroblasts (Fig. 6a). Monazite inner cores contain relatively high-Y concentrations and are surrounded by high Sr and Th outer cores (refer to ESM Table S2). Inner cores yield a 206 Pb/ 238 U weighted average age of 94.79 ± 2.64 Ma (MSWD = 1.76; n = 3) and outer cores 94.07 ± 0.82 Ma (MSWD = 2.45, n = 7; Fig. 6c; Table 2). Monazite inner rims show overprinting textures adjacent to older compositional boundaries (Fig. 6b) and yield a 206 Pb/ 238 U weighted average age of 92.49 ± 1.09 Ma (MSWD = 4.63; n = 18). Lastly, thin (~ 3 μm) outer rims are present along the edges of some monazite crystals. One analysis was performed on an outer rim large enough to fit the 10 μm beam spot and yielded a 206 Pb/ 238 U age of 84.2 ± 1.8 Ma.
Sample 14GCM020-19.3 is a garnet porphyroblastic paragneiss with disseminated graphite and biotite showing a strong foliation defined by K-feldspar, quartz, sillimanite, apatite, and monazite shape preferred orientation. Garnet porphyroblasts are anhedral to subhedral and show an increase in Mg from core to rim similar to that in sample 14GCM010-20.1 (Fig. 6a). Monazite crystals range in size from 25 to 125 μm (Table 2). Inner cores, where present, are defined by high Th and low Y concentrations and yield a 206 Pb/ 238 U weighted average age of 94.33 ± 4.82 Ma (MSWD = 3.9; n = 8; Fig. 6c; Table 2). Low-Y and -Th outer cores, which are generally the cores of monazite crystals that lack high Th inner cores, yield a 206 Pb/ 238 U weighted  Compositional yttrium (Y) and thorium (Th) maps of monazite grains corresponding to each sample in (a). Scale bars are 20 µm. c Age-compositional data from monazite in all samples including REE spider diagrams divided by monazite generation. The spider diagrams start at lanthanum (La) and end at lutetium (Lu); the vertical axis ranges from 10 0 to 10 6 (chondrite normalized; Sun and McDonough 1989). d All U-Pb isotopic data ( 206 Pb/ 238 U ages with 2σ uncertainty) plotted against Eu/Eu* analyzed in monazite for three migmatitic gneiss samples. The orange band encompasses anatexis / peak metamorphism, and the green band retrograde metamorphism. Note spike in Eu/Eu* values between 100 and 90 Ma, interpreted to reflect plagioclase feldspar consumption during biotite-dehydration melting reactions. Eu* is the geometric mean of Sm and Gd abundances average age of 94.25 ± 0.96 Ma (MSWD = 6.52; n = 11). Inner rims are defined by a modest increase in Y content and have a 206 Pb/ 238 U weighted average age of 85.45 ± 0.55 Ma (MSWD = 1.20; n = 17). Most monazite crystals contain a very high-Y outer rim that is developed heterogeneously around the grains that yields a 206 Pb/ 238 U weighted average age of 83.09 ± 1.6 Ma (MSWD = 1.15; n = 5).
Where only carbonaceous material occurs, it tends to occur as discrete angular to clotty individual grains and grain aggregates or else as inclusions in quartz (Fig. 9a) or garnet, as opposed to euhedral blades observed as the primary form of graphite. In one case, graphite was observed as inclusions in garnet. Not only are carbonaceous material and graphite found together in multiple samples but also are found dispersed as inclusions as well as in direct proximity to one another in the same sample (Fig. 9b, c). The samples show two distinct types of graphite textures. The first type includes bladed to fibrous aggregates and well to moderately well foliated, coarse, millimeter-scale euhedral graphite (Fig. 5b). Some graphite blades and fibers show evidence of weak kinking. The second type includes micron-scale, discrete, angular to clotty individual grains, and grain aggregates as separate entities or inclusions most commonly within quartz or garnet (Fig. 9b, c). Some clotty regions of carbon exhibit poikiloblastic textures. Calculated Raman R2 area ratios for all analyzed CM and graphite are shown in Fig. 10. Raw Raman data are in ESM Table S5. Because no D1 peak (1325-1360 cm −1 ) was detected by Raman spectroscopy, 12 spots out of 33 were identified as graphite (R2 = 0). The remaining 21 carbonaceous material points had R2 values ranging from 0 to 0.48, implying a high degree of order (Fig. 10). These R2 values are consistent either with formation during metamorphism at temperatures from 427to > 640 °C (based on the calibration of Beyssac et al. 2002), or with hydrothermal formation at moderate temperatures (~ 500 °C, e.g., Luque et al. 2009). Overlap in ordering and textural attributes of hydrothermal versus metamorphic CMs can lead to non-unique interpretations of Raman-based thermal estimates precluding a quantitative interpretation of Raman  Beyssac et al. 2002;Luque et al. 2009).

Graphite Creek protolith origins and evidence for Late Cretaceous anatexis
The aluminosilicate-rich mineralogy and isotopically light signature of graphite carbon (δ 13 C ≈ − 20 to − 15‰) in the Graphite Creek paragneiss units indicate their protolith rocks were dominantly shale-and sandstone-rich. The presence of calc-silicate layers in the deposit and marble mapped elsewhere in the Kigluaik Mountains indicates some calcareous protoliths as well. The calcareous protoliths suggest a marine depositional environment for the precursor sediments of the Graphite Creek host rocks. Depositional age of the protoliths is broadly constrained by regional igneous and detrital zircon U-Pb ages from the Kigluaik metamorphic complex and Nome Complex that span the Neoproterozoic to Pennsylvanian ( Fig. 11a; Amato et al. 2009;Till et al. 2014a).
Ubiquitous leucosome textures and pegmatitic granites in the paragneiss units at Graphite Creek indicate partial melting took place. The intervals of concentrated sillimanite, garnet, and biotite that contain little quartz are consistent with restite and strongly suggest loss of some of this melt. The monazite petrochronology data presented here build on earlier monazite work by Amato et al. (1994) by providing temporally constrained petrogenetic information about the melting history of such migmatitic paragneisses . Texturally, the well-annealed microstructures in monazite are characteristic of high-temperature dynamic recrystallization during deformation (Passchier and Trouw 2005). The disrupted oscillatory zonation and complex compositional heterogeneity of monazite are indicative of coupled and fluid-mediated, dissolution-reprecipitation reactions . Our data clearly show a strong Y-depletion in monazite associated with high Eu/ Eu* values, high Sr concentrations, and excursions to high Th concentrations, ca. 98 to 93 Ma (Fig. 6d). Similar traits employing similar methods are associated with lower crustal migmatites exposed in the Neoarchean Athabasca granulite terrain (Canada; Dumond et al. 2015;Regan et al. 2017) and the southern part of the Mesoproterozoic Grenville Province (USA; Williams et al. 2019). In both regions, these systematic compositional attributes in monazite from migmatitic paragneisses are attributed to biotite-dehydration melting (Dumond et al. 2015;Williams et al. 2019), as follows: 1. Biotite + plagioclase + sillimanite + quartz = garnet + K-feldspar + melt Here, Y is partitioned into the peritectic garnet during partial melting whereby Eu and Sr from plagioclase feldspar is liberated into the liquid (melt + H 2 O-CO 2 ) phase. Thorium, which is strongly incompatible, will also partition into the liquid. Collectively, we interpret in situ monazite petrochronological data from migmatitic paragneisses to record several-million-years of partial melting, due to biotite-dehydration reactions that corresponded with garnet and K-feldspar growth (Fig. 11b). This is corroborated by a zircon U-Pb igneous crystallization age of ca. 94 Ma from sample 12GC002-44.9 (Case et al. 2020). The monazite data from this sample suggest that deformation persisted until ca. 91 Ma (Fig. 11b). The 96-91 Ma ages are consistent with metamorphic monazite U-Pb ages reported by Amato et al. (1994). Three samples that contain evidence for retrograde Fig. 8 Carbon and sulfur isotope data. a Box and whisker distribution of δ 13 C graphite among rock types. Number of analyses is indicated above each box. SKPG, syn-kinematic pegmatitic granite. b Box and whisker distribution of δ 34 S pyrrhotite among rock types metamorphism have ages previously unrecognized from ca. 85 to 82 Ma (Fig. 6d). This population of monazite, in all three samples, is associated with texturally late, high-Y monazite interpreted to reflect local garnet breakdown during fluid influx that resulted in fluid-mediated, dissolution reprecipitation of monazite. The age spread in metamorphic titanite (ca. 93-85 Ma) also suggests retrograde overprint. The timing for retrograde metamorphism corresponds with ages from existing thermochronology, which consistently yields white mica 40 Ar/ 39 Ar plateau ages ranging from 84 to 80 Ma (Calvert et al. 1999) within the Kigluaik gneiss dome, consistent with garnet breakdown. This overlaps with the age of regional northeast-southwest trending shallow-level dike swarms (Amato et al. 2003). The P-T path inferred for Graphite Creek from the monazite data is consistent with regional P-T path of the Kigluaik metamorphic complex described by .
Migmatitic paragneisses from this study all lack plagioclase feldspar. However, monazite compositional data show a strong feldspar fractionation trend (Fig. 6) defined by Sr and Eu. Williams et al. (2019) presented multiscale compositional mapping and in situ monazite petrochronology on a suite of graphite-bearing migmatitic paragneisses in the Adirondack Mountains in New York (USA), also located in a granulite terrane with a gneiss dome origin . Multiscale compositional mapping showed variations in plagioclase feldspar, which was interpreted to correspond with melt loss . In reaction (1), plagioclase was consumed during progressive melting and in the case of the Adirondack Mountains migmatites appears to have been the limiting phase of the reaction . Samples of graphite-garnet migmatite in the Kigluaik Mountains are similar in both modal mineralogy and texture, except we have not recovered any migmatites that contain plagioclase feldspar associated with the graphite mineralization. Given the strong evidence in monazite for feldspar fractionation, partial melt generation, and complete lack of plagioclase feldspar in samples selected for petrochronology, we interpret the migmatites associated with graphite mineralization at Graphite Creek to have formed by carbonaceous material. Some carbonaceous material is highly angular with bright spots reminiscent of clasts of once-bladed graphite. Raman spectra from representative site acquisitions in each photo A, B, and C, respectively, are shown above the plots. Red dots are acquisition points; blue lines are measured spectra; red lines are fit spectra; G (1580 cm −1 ), D1 (1325-1360 cm −1 ), and D2 (1600-1620 cm −1 ) peaks are also shown partial melting of the shale-and sandstone-rich protoliths and extensive melt loss. Therefore, migmatites associated with flake graphite mineralization in the Kigluaik Mountains are likely restitic. Regional magmatism related to the voluminous ca. 100-90 Ma Kigluaik pluton likely provided the heat source necessary for partial melting here and elsewhere in the Kigluaik gneiss dome (e.g., Miller et al. 1992;Amato and Miller 2004). It follows that graphite-rich rocks in the Kigluaik Mountains have undergone melt loss, which may have been an important process for concentrating graphite in the region, given that graphite is typically most abundant in lenses that contain the peritectic solids (K-feldspar, garnet, and sillimanite).

Origins of flake graphite mineralization at Graphite Creek
Current models of crystalline graphite deposits allow for multiple scenarios to potentially explain aspects of highgrade flake graphite enrichment at Graphite Creek. Incorporation of garnet porphyroblasts within the massive graphite lenses and foliation of semi-massive graphite together with peak graphite temperatures of at least 640 °C suggest that lens development took place during syn-to late-peak metamorphism (Fig. 11). This metamorphism happened between ca. 96 and 92 Ma based on the monazite and titanite data from this study (Fig. 6) and Amato et al. (1994). All scenarios therefore invoke graphite concentration via high-temperature metamorphic processes. Scenario 1 is simple in situ metamorphism of carbonaceous shale. Scenario 2 involves devolatilization of calcareous rocks to remove CO 2 and form Ca-silicate mineral assemblages such as plagioclase, pyroxene, and scapolite. Scenario 3 involves addition of carbon to the rock via hydrothermal fluids, such as for lump and chip vein deposits. Scenario 4 is removal of silicate minerals from carbonaceous shale by anatexis. Our new observations and data help constrain which scenario or combination of scenarios is most likely: Scenario 1-in situ metamorphism of carbonaceous shale-is consistent with the pelitic protolith of Unit A and isotopically light organic signature of the carbon and sedimentary signature of sulfur (Fig. 8;Seal 2006) and can also explain the disseminated flake graphite. On the other hand, graphite lenses that have graphite grades as high as 50 wt.% in the deposit require carbon enrichment by a factor of 4 to 10 depending on the total organic carbon (TOC) content of the protolith, implying substantial solid-state movement of silicate material into meter-scale segregation layers. No such segregation layers are observed to be associated with the graphite lenses. The presence of garnet porphyroblasts in the massive graphite lenses is also difficult to reconcile with simple enrichment. The garnet requires an aluminosilicate groundmass in which to grow, but this groundmass is now absent. Whereas scenario 1 is the most likely process for forming the disseminated graphite, it does not explain many other characteristics of the deposit.
Devolatilization of calcareous rocks during metamorphism (scenario 2) produces isotopically heavy graphite (δ 13 C > − 5‰) and typically forms graphite lenses proximal to marble and calc-silicate rocks from which the carbon was derived. This scenario is inconsistent with both the isotopically light carbon at Graphite Creek (δ 13 C ≈ − 20 to − 15‰) and the general scarcity of calc-silicate rocks in the vicinity of the orebody and is therefore unlikely.
Scenario 3, deposition of graphite from a H 2 O-CO 2 or H 2 O-CH 4 hydrothermal fluid, can explain some of the textures of the lenses, such as euhedral tourmaline within graphite (Fig. 4d) and the abrupt contacts of some lenses with surrounding gneiss. Unlike hydrothermal veins in Sri Lanka and New Hampshire, the lenses at Graphite Creek show clear evidence of plastic deformation during and/or after formation, which may have obscured hydrothermal textures. In the high-temperature, middle-to lower-crustal, metamorphic environment of Graphite Creek, the only possible fluid sources are those formed by prograde metamorphism, separation from hydrous anatectic melt, or the mantle. Graphite would be precipitated through either cooling or mixing with another carbon-bearing fluid (Luque et al. 2014;Simandl et al. 2015). Lower-temperature hydration reactions (Luque et al. 2014) are considered unlikely given the syn-peak metamorphic timing of lens development. The carbon isotopic data allow for hydrothermal graphite derived from an organic source but do not support a mantle origin or mixing with other sources. Although hydrothermal activity cannot be ruled out, we consider it unlikely that sufficient hydrothermal fluid would be present in the metamorphic environment to account for all mineralization.
Scenario 4 involves at least partial concentration of graphite via removal of silicate minerals during partial melting (anatexis) of carbonaceous paragneiss (Fig. 11c). Although this process is not widely recognized in graphite deposits, anatexis is documented in the upper amphiboliteto granulite-facies conditions in which some flake graphite deposits form. In this scenario, biotite dehydration melting removes some silicate phases while graphite, derived locally from protolith organic carbon, remains. Most of the melt is extracted and does not crystallize in place. This process is consistent with observed restitic textures and explains the garnet inclusions within the massive graphite lenses, wherein garnet is part of a residual assemblage after quartz was lost through melt that migrated away (Fig. 11c). On the basis of observed pegmatitic leucosome textures (the SKPGs), monazite chemistry, and the presence of graphite, the anatectic melts were likely H 2 O-and CO 2 /CH 4 -bearing. It is unlikely that these phases would separate in the hightemperature metamorphic environment, but may have upon cooling and uplift to form a hydrothermal fluid. This could explain some of the hydrothermal-like textures described in scenario 3. Assuming regional peak metamorphic conditions of ~ 850 °C and 0.8-1.0 GPa, the amount of partial melt could be in excess of 10 mol.%, depending on the Creek and based on regional age populations from the Kigluaik Mountains (Amato et al. 2009). b High-temperature metamorphism and initiation of partial melting in Middle Cretaceous (ca. 96 Ma), followed by exhumation and retrograde overprint by ca. 85 Ma. P-T paths after  See text for discussion. c Melt extraction and fluid release associated with anatexis facilitated flake graphite enrichment and development of massive/semi-massive lenses of graphite, leaving behind residual layers enriched in garnet and sillimanite. The retrograde event likely caused localized recrystallization of graphite via influx of fluids. Bt biotite, Gr graphite, Grt garnet, Sil sillimanite, Tur tourmaline, MC metamorphic complex composition and water content of the metapelite protolith (Nabelek and Bartlett 2000). However, melt removal may not fully account for graphite enrichment. Small (< 1 m), isolated graphite lenses in migmatitic sillimanite gneiss 6 km south of Graphite Creek suggest that partial melting and/or hydrothermal activity took place over a large area of the gneiss dome. Widespread mafic to felsic magmatism, related to the ca. 90 Ma Kigluaik pluton, is the most likely crustal-scale source of heat necessary for such regionally extensive partial melting in the Kigluaik metamorphic complex (Amato and Miller 2004). Strain related to unroofing of the gneiss dome may have facilitated melt removal.
A combination of scenarios 1, 3, and 4 are considered necessary to explain all the features and high grades of graphite enrichment in the deposit. Flake graphite in the anatectic pegmatitic dikes, together with the carbon isotopic uniformity of graphite across all rock types, strongly suggests some mobility of carbon during partial melting of paragneiss. Coeval shearing and deformation could have provided conduits for melt loss and later fluids. Without addition of graphite from another source, a carbonaceous shale protolith containing high TOC (> 10 wt.%) is likely necessary to produce high-grade enrichments via these processes. The quartz-biotite gneiss, characterized by a spread in both δ 13 C and δ 34 S values for graphite and pyrrhotite, respectively, probably experienced less fluid interaction and partial melting than the sillimanite gneiss, preserving some original isotopic heterogeneity.
Field observations, monazite petrochronology, and Raman data suggest minor disturbance of graphite after peak metamorphism. Younger titanite and monazite ages of ca. 85 Ma in both units likely record subsequent unroofing and retrograde metamorphism. Rare coarse graphite veinlets cutting across the massive lenses suggest local remobilization of graphite by retrograde fluid influx. The occurrence of two distinct carbonaceous material textures, graphite lacking Raman D1 peaks, and the range of significantly lowerordered CMs is probably best explained by such a retrograde fluid.

Comparisons to other crystalline graphite deposits
Graphite Creek is unusual compared to other crystalline graphite deposits. Unlike lump and chip graphite veins (Fig. 12a) in Sri Lanka (Katz 1987) and New Hampshire , graphite lenses at Graphite Creek are broadly coeval with peak metamorphism. Graphite Creek is also different from the few well-described occurrences of synmetamorphic massive flake graphite lenses like Lac Knife, Asbury, and Hartwell, Canada (e.g., Fig. 12b), in that the lenses show no evidence of having formed from marble decarbonation or through structural focusing into fold hinges (Simandl et al. 2015). Whereas "migmatized" is used to describe the gneiss host rock at Lac Knife (Desautels et al. 2014), anatexis has not been suggested as a mineralizing process in that system. The literature available on other flake deposits suggests that most are characterized by disseminated graphite flakes in paragneiss. Graphite Creek appears to be unique in the extent of partial melting of the host rocks as well as in the lack of significant carbonate protoliths adjacent to the lenses. The factor of extremely high heat, indicated by anatectic features, is likely the critical factor in the development of the Graphite Creek large flake graphite deposit. The role of anatectic processes in concentrating and growing very high grade, large flake graphite is not documented elsewhere and warrants further investigation for development of an appropriate deposit model. Fig. 12 Schematic cartoon of major crystalline graphite-forming mineral systems. a Lump and chip graphite is found in hydrothermal vein deposits. The veins typically cut granulite-grade metamorphic fabrics and are focused in fault zones or along lithologic contacts, such as intrusions (red). Examples: Sri Lanka (Katz 1987;Silva 1987), New Hampshire ). b Flake graphite occurs as disseminations in amphibolite-to granulite-grade paragneiss and as concentrations in semi-massive pods and lenses associated with marble/paragneiss contacts. The lenses can be focused in fold hinges or in shear zones. Examples: Lac Knife, Canada; Asbury, Canada (Simandl et al. 2015); Graphite Creek, USA (this study). Not to scale

Exploration implications
Important features of the Graphite Creek deposit suggest critical factors that might serve as exploration tools in the search for large flake graphite deposits. Pelitic protoliths deposited in environments associated with high biological productivity are most favorable for having high primary carbon accumulations needed to form graphite-rich deposits. In the AACM, such depositional environments were prevalent in Late Devonian-Early Mississippian restricted basins and Permian to Late Triassic basins and outer shelf-slope settings in northwestern Alaska (Dumoulin et al. 2004(Dumoulin et al. , 2013.
Voluminous magmatism associated with peak metamorphism, such as the Kigluaik pluton, are a potential heat source for HT metamorphism. The HT metamorphic equivalents of these siliciclastic rocks are sillimanite paragneiss and migmatite characterized by high-temperature restitic textures and assemblages. The compressed Barrovian isograds indicate focused strain uplift of granulite facies rocks from deep in the crust in the northernmost Kigluaik Mountains .
The constraints on flake graphite genesis discussed above can be used to define mappable criteria for the assessment of regional potential for large flake graphite deposits and identification of exploration tools. Exploration should focus on favorable tectonic settings characterized by (1) significant thickness of carbonaceous sedimentary rocks; (2) a magmatic heat source that fostered high-temperature metamorphism, ideally near or above the second sillimanite/granulite grade; and (3) high strain that played a role in the mobilization and of anatectic melts and concentration of restitic minerals during strain partitioning processes, such as is indicated by the compressed isograds and mineral segregation layering on the north flank of the Kigluaik gneiss dome.

Conclusions
The unique flake graphite deposit at Graphite Creek, Alaska, formed due to a combination of factors. Permissive depositional environments for a suitable protolith are biologically productive, restricted, settings. Permissive post-depositional tectonic environments to produce this type of flake graphite mineralization are characterized by high-temperature metamorphism associated with exhumation. Thick intervals of favorable, fine-grained carbonaceous shale protoliths were subjected to at least one episode of high-temperature metamorphism and anatectic melting associated with pluton emplacement, regional extension, and tectonic exhumation on the Seward Peninsula. Strain partitioning and shearing in the northern Kigluaik Mountains, represented by compressed isograds, likely provided conduits for fluid/melt flow and facilitated anatexis, melt loss, and graphite enrichment.
Anatexis as a major mineralizing process is scarcely documented in flake graphite deposits and warrants further study in other districts. Mappable criteria corresponding to these characteristics include the presence of thick sequences of carbonaceous protoliths, migmatitic textures, and mineral assemblages indicative of high-temperature metamorphic grades and partial melting, and narrow Barrovian isograd sequences indicative of high strain and a tectonic setting favorable for anatectic melt loss and graphite concentration.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.