K-feldspar enrichment in the Pacific pelagic sediments before Miocene

The mineralogy of atmospheric silicate dust controls its interaction with clouds. K-feldspar has a remarkably high ice-nucleating activity, and its distribution may have influenced the global climate throughout Earth's history. However, long-term archives of past atmospheric feldspar are not known. Here, we investigate feldspar mineralogy, content, and grain size in pelagic clay cores. Sediments around Minamitorishima Island contain > 10 wt% of K-feldspar before ~ 35 Ma, which is five times more than the younger sediments. This distribution does not resemble other volcanic minerals or geochemically estimated volcanic input, suggesting that the K-feldspars are not associated with volcanic ash. The K-feldspars are present as isolated grains as well as pseudorhombohedral microcrystals indicative of authigenic overgrowth. On the other hand, they contain some Na, arguing against a purely authigenic origin. Grain size distributions of chemically separated quartz and feldspars show stratigraphic variation analogous to other North Pacific sites, further suggesting a link to eolian materials. Sediments from a South Pacific site also show K-feldspar enrichment over plagioclase before ~ 44 Ma, although the content relative to bulk sediment does not change much. We propose that the K-feldspar may be enriched in the wide area of the Pacific before ~ 30 to 40 Ma.


Introduction
There has been renewed interest in aerosol mineralogy because of the remarkably high ice-nucleating ability of K-feldspar (Atkinson et al. 2013;Harrison et al. 2016).Even though K-feldspar often represents a small fraction (< 5%) of mineral aerosol today, the type, amount, and size of K-feldspar could control the heterogeneous ice nucleation in mixed-phase clouds (e.g.Kanji et al. 2017).This process is critical for cloud microphysics that affects global climate, such as radiative transfer and precipitation.Thus, past changes in ice nucleation activity associated with atmospheric K-feldspar could have affected global paleoclimate, but there is limited evidence for these scenarios.Pankhurst (2017) argued that because granitic upper continental crust contains more K-feldspar than basalt, there should be a Gyr scale secular increase in the dust K-feldspar content following the continental growth.Studying meteorite impact records, Pankhurst et al. (2021) pointed out a positive correlation between the K-feldspar content of the target rock and the associated extinction intensity.They suggested that this correlation reflects global cooling by K-feldspar-rich impact ejecta.However, these hypotheses lack actual relics of eolian K-feldspar.Consequently, the proposed high ice nucleation activity could not be tested.
Pelagic sediments are long-term archives of atmospheric dust including continental materials and volcanic ash (e.g.Rea 1994;Dunlea et al. 2015b).However, little attention has been paid to the detailed feldspar mineralogy in marine sediments.Previous studies of dust mineralogy have often focused on clay to decipher continental climate and provenance (Singer 1984).Only a few comprehensive studies have examined bulk sediments or materials treated by HCl (Heath 1969;Peterson and Goldberg 1962).These studies reported low K-feldspar contents that show little temporal variation.In contrast, a recent study of South Pacific pelagic clay using strong chemical digestion to remove clay minerals showed a marked increase in K-feldspar in the deeper (older) section (Usui et al. 2018).However, the chronology of the sediment and the possible origin of the K-feldspars were not thoroughly discussed.
K-feldspar has also been reported from marine sediments as an authigenic phase, often associated with volcanic materials (Bass 1976;Donnelly and Nalli 1973;Kelts and McKinzie 1976;Lancelot et al. 1972;Matthews 1962;Mellis 1952;Stewart et al. 1973).If one wants to use pelagic sediments as archives of past atmospheric dust minerals, it is necessary to distinguish between eolian and authigenic grains.Empirically, authigenic K-feldspars are chemically pure and contain negligible Na and Ca even formed within marine sediments (e.g.Kastner and Siever 1979).They often constitute only a small fraction of the sediments, except some volcaniclastic sediments contain up to 60% K-feldspar (Kelts and McKinzie 1976).Consequently, stratigraphic variations of authigenic K-feldspar in pelagic sediments have not been studied.In pelagic clay, it has been observed that the occurrence of K-feldspar correlates with that of clinoptilolite (Stonecipher 1976).This leads to the suggestion that K-feldspar may form from the diagenesis of clinoptilolite as in saline alkaline lakes (Kastner 1981).However, there is no direct evidence for this reaction in pelagic clays, and it is unclear if they contain authigenic K-feldspar.
Here we present the feldspar mineralogy of two piston cores (KR13-02 PC05 and MR14-02 PC11) collected around Minamitorishima Island, North Pacific.Together with South Pacific data from International Ocean Drilling Program (IODP) Site U1366, our results indicate that K-feldspar is the dominant feldspar for sediments older than Miocene-Oligocene.We discuss the possible origin of the K-feldspars.

Materials
We studied two piston cores of pelagic red clay, MR14-E02 PC11 (22° 59.02′ N, 154° 0.98′ E, 5647 m water depth) and KR13-02 PC05 (21° 59.03′ N, 153° 56.35′ E, 5735 m water depth), obtained around Minamitorishima Island (Fig. 1).MR14-E02 PC11 was first dated by Os isotope stratigraphy (Nozaki et al. 2019).Subsequent paleomagnetic studies indicate the low 187 Os/ 188 Os signature around 34.4 Ma have been spread out after deposition, resulting in a significant overestimation of sedimentation rate (Usui and Yamazaki 2021).Corroborated with Os data and constraints from realistic sedimentation rates, they assign 35-38 Ma at 12.0 m and 30.6-28.1 Ma at 9.0 m.Chronostratigraphy of KR13-02 PC05 was also estimated using Os isotope ratios (Ohta et al. 2020).It has not been examined if the Os isotope signature is affected by post depositional processes as in MR14-E02 PC11.Nonetheless, microfossil ages are broadly consistent with Os isotope data.Therefore, we will use Os isotope ages for sediments younger than 28 Ma, while we rely on magnetostratigraphy or microfossil age constraints before that time.Bulk chemical compositions of cores obtained in this region established regional chemostratigraphy, which divides sediments into five units; unit I to unit V from the youngest to oldest (Tanaka et al. 2020a, b).The chemostratigraphy indicates that KR13-02 PC05 contains sediment much older than MR14-E02 PC11 (Tanaka et al. 2020a).This is also confirmed by our magnetostratigraphy measurements (Fig. 1b, c).The upper part of unit V sediment from different cores contains biogenic magnetite indicative of high surface production (Usui et al. 2018;Yamazaki et al. 2020).Yamazaki et al. (2020) correlate this signature to a paleo-equatorial high-productivity region, ~ 70 Ma for Minamitorishima.This interpretation indicates that KR13-02 PC05, 9.0 m, may be older than ~ 70 Ma.Regional acoustic survey (Nakamura et al. 2016) and chemostratigraphic and magnetic correlations (Tanaka et al. 2020a;Yamazaki et al. 2020) indicate that the surface sediment in the studied cores may not be of zero age.The Os isotope data (Nozaki et al. 2019;Ohta et al. 2020) suggests core top ages to be ~ 2 Ma at 0.014 m in MR14-E02 PC11 and ~ 11 Ma at 0.11 m in KR13-02 PC05.A hiatus was reported at the boundary between unit I and unit II in MR14-E02 PC11 (Nozaki et al. 2019).In summary, MR14-E02 PC11 covers ~ 40 to 2 Ma, and KR13-02 PC05 ~ 70 to 10 Ma.
In addition to the Minamitorishima cores, we reexamined the mineralogy of International Ocean Discovery Program (IODP) Site U1366 in the South Pacific (Expedition 329 Scientists 2011).The sediments in IODP Site U1366 are divided into two lithostratigraphic Units I and II.Unit I is subdivided into three Subunits Ia, Ib, and Ic.Feldspar in Unit Ia is mostly plagioclase.K-feldspar becomes enriched relative to plagioclase down-core (Usui et al. 2018).An age model based on cobalt flux and Os isotope data (Dunlea et al. 2015a(Dunlea et al. , 2021) ) suggests that the boundary between Unit Ia and Ib may be ~ 44 Ma, and that for Unit Ib and Unit Ic is ~ 62 Ma.

Chemical digestion
We performed digestion by citrate-bicarbonate-dithionite (CBD) (Rea and Janecek 1981) and sodium pyrosulfate fusion to extract feldspar and quartz (Inoue et al. 2021;Kiely and Jackson 1964;Syers et al. 1968;Usui et al. 2018).Dry samples of ~ 1 g were first treated with CBD to remove poorly crystalline Fe-Mn oxyhydroxides.The residues were washed with purified water, freeze-dried, and weighed.The oxyhydroxide-free residues were then subjected to the sodium pyrosulfate fusion as follows.First, the residues were fused with Na 2 S 2 O 7 at 460 °C to decompose sheet silicates.Next, the fusions were treated with 3N HCl and washed with purified water.Then, the residues were heated to 50 °C in 1 M NaOH overnight to remove relict clays.The sodium pyrosulfate fusion was repeated 2 or 3 times to ensure complete digestion.While the CBD and acetic acid treatments have been widely used to extract the eolian component from pelagic sediments, these methods were ineffective in removing biogenic apatite (Usui and Yamazaki 2021) and authigenic zeolite (Ziegler et al. 2007), even though those phases sometimes account for over 30 vol% of the sediment in the studied region (Iijima et al. 2016;Ohta et al. 2020).Strong acid treatment should dissolve biogenic apatite while preserving quartz, but it is not yet clear if the pyrosulfate fusion would remove all clay minerals and zeolite and if it would attack feldspars.While we focus on the pyrosulfate fusion residues, we also compare materials obtained by the CBD treatment and bulk, untreated materials to examine the effect of pyrosulfate fusion.In any case, the materials analyzed in this study should be regarded as operationally defined components.

Mineralogy
We examined the mineralogy of the pyrosulfate fusion residues by X-ray diffraction (XRD).XRD measurements used a diffractometer (MiniFlex II, Rigaku) with Cu Kα radiation.To check the potential effect of chemical digestion on the mineralogy, we also measured selected samples in untreated conditions and after the CBD treatment.We estimate mineral proportions by the Reference Intensity Ratio method with full pattern summation using the software PowdR (Butler and Hillier 2021).We used reference patterns of K-feldspar, plagioclase, quartz, and clinoptilolite from the RockJock reference library (Eberl 2003).We also included glass, tephra, and obsidian reference patterns because some samples revealed a broad peak.We will call them collectively "glass"; however, it is introduced to fit the observed background of the spectra, and not much interpretation should be made on this phase.We also reexamine the XRD data from IODP Site U1366C (Usui et al. 2018) using the same protocol.The fusion residues were also examined under a scanning electron microscope (SEM; JEOL JSM-6010V) equipped with an energy-dispersive spectrometer EDS.

Grain size
Grain size distributions were measured for the fusion residues.The samples were first dispersed in a 0.2% sodium hexametaphosphate and sonicated for 1 h.Then the dispersed sample was introduced into a laser diffraction particle size analyzer (Mastersizer 3000E, Malvern Panalytical) with purified water.

Results
The CBD treatment reduces the sample weight by 10-20%.Sodium pyrosulfate fusion further reduces the sample weight to 10-30% of the original weight.
The XRD spectrograms can be fitted with the selected phases in various proportions (Figs. 2 and 5).The only Fig. 2 XRD diffractograms of fusion residues and their fits of a MR14-E02 PC11, 2.67 m, b MR14-E02 PC11, 8.99 m, and c KR13-02 PC05, 8.01 m.Data are on the modified 2θ angles calculated by the PowdR software.The quartz peak at 26.6° is out of bounds exceptions are KR13-02 PC05, 2.66 m, KR13-02 PC05, 2.79 m, and U1366C-2H-5W, 143 cm, which indicate the presence of small amount of additional phases (Additional file 1).These samples are excluded from further discussion.We cannot completely exclude the presence of additional phases in the other samples, but the XRD spectra suggest that those unspecified phases should be minor components.Large variations in K-feldspar content relative to plagioclase are also evident from spectra for bulk sediments before chemical digestion (Fig. 3).This result confirms that the digestion treatments are not the primary reason for the variations in K-feldspar content.SEM/EDS analyses confirmed the presence of K-feldspar (K-rich Al silicate) as well as plagioclase (Narich Al silicate) and quartz (Fig. 4).K-feldspar appears as isolated grains (Fig. 4b) and grains covered by rhombohedral crystals similar to authigenic K-feldspar (e.g.Worden and Rushton 1992) (Fig. 4c).The resolution of the SEM limits detailed observations to grains over a few microns, which is larger than the median grain size.Both types of K-feldspar appear to contain Na.However, each rhombohedral crystal is too small to be measured separately, and the signature may reflect submicroscopic heterogeneity.The quartz grains are more surrounded compared to the feldspars (Fig. 4e).Otherwise, there is no significant difference in size or morphology between the minerals.The mass fraction of fusion residue and mineral proportions within the residue show systematic stratigraphic variations.In the Minamitorishima cores, the fraction of the residues is the smallest around 30-35 Ma (Fig. 5a,  c).In terms of feldspar mineralogy, all cores are dominated by plagioclase in the shallow part, while K-feldspar is enriched over plagioclase in the deep part.K-feldspar enrichment occurs before ~ 28 to 39 Ma (including late Eocene to early Oligocene in KR13-02 PC05) for Minamitorishima cores, while it occurs before ~ 44 Ma for IODP site U1366 (Fig. 5b, d, f ).Some intervals show signs of increased volcanic input, but this does not cover the entire interval where K-feldspar is enriched.A few samples contain a high proportion of glass within the fusion residue (Fig. 5b, d).These intervals, i.e. ~ 6 m in MR14-E02 PC11 and ~ 3 m in KR13-02 PC05, correspond to the so-called "extremely REY-rich mud" interval (Iijima et al. 2016), where the biogenic apatite content reaches ~ 30 wt%.Samples below 8.1 m in KR13-02 PC05 contain significant clinoptilolite (Fig. 5d), a diagenetic mineral likely derived from volcanic glass (Lisitzina and Butuzova 1982;Petzing and Chester 1979).In addition, it should be noted that some volcanic glass and zeolite are likely to be removed during chemical treatment.In particular, phillipsite has been reported to constitute up to 30% of bulk samples at 1.6-2.0m in KR13-02 PC05 (Ohta et al. 2016(Ohta et al. , 2020)), but it is not detected in our digested samples.However, these variations are not in phase with the consistently high K-feldspar fraction below 4 m in KR13-02 PC05 (Fig. 5b), suggesting that the K-feldspars are not direct volcanic input.
We calculate the mineral content relative to the bulk sediments using the mass fraction of fusion residues and the mineral proportions within the residues.In the Minamitorishima cores, the calculated K-feldspar content reaches 12 wt% of the bulk sediments in the deeper stratigraphic level, while it is about 2 wt% in the shallower level (Fig. 6).At IODP Site U1366, the K-feldspar content does not vary much because the low proportion of fusion residues in Unit Ib (Fig. 5e) (Usui et al. 2018) compensates for the high proportion of K-feldspar in the fusion residues.
Grain size distributions of the fusion residues generally show the mode at 2 µm to 4 µm (Fig. 7).Smaller grain size with the mode at ~ 2 µm is observed in MR14-E02 PC11 below 11.99 m and in KR13-02 PC05 between 3.87 m to 7.05 m.These grain size variations resemble that of the eolian dust in the North Pacific (Janecek 1985;Janecek and Rea 1983), where the grain size is finer between ~ 50 and ~ 30 Ma (Fig. 7e).The depth of the grain size change (11.99 m in MR14-E02 PC11) is different from that of the mineralogical change (7 m).Thus, we interpret these changes reflect original grain size variations, even though it may be possible that the grain sizes are reduced by the chemical digestion.Some samples show bimodal grain size distribution with another maxima at ~ 20 µm, possibly reflecting sorting by strong bottom current (McCave et al. 1995;Ohta et al. 2016).

Discussion
We discovered large variations in K-feldspar content from North Pacific red clay.Together with the South Pacific data (Usui et al. 2018), it appears that Pacific red clay before ~ 30 to 40 Ma is enriched in K-feldspar relative to plagioclase.Moreover, in the North Pacific sites, K-feldspar is a major component (> 10 wt%) of the bulk sediments at that time.The most probable origin of the K-feldspars seems to be either eolian or a replacement of eolian plagioclase.Although K-feldspar may form after volcanic materials (Kelts and McKinzie 1976;Lancelot et al. 1972;Mellis 1952), they are not enough to explain the K-feldspar enrichment.Clinoptilolite content is only significant in the lowermost part of KR13-02 PC05 and does not correlate with K-feldspar content (Fig. 5d).A modeling study using bulk chemistry also suggests the mass fraction of a volcanic component is limited to be < 10 wt% between 4 and 7 m and ~ 20 wt% below 8 m (Tanaka et al. 2022).In contrast, K-feldspar content already reaches ~ 10 wt% between 4 and 7 m in the core (Fig. 6b).These contrasting variations are difficult to explain with K-feldspar formed after volcanic materials.
Eolian transport of the K-feldspars is a natural explanation for their variable abundance.Changes in either provenance or weathering conditions in the source area can result in mineralogical changes.Indeed, mid-and lowlatitude sites in the North Pacific often show provenance changes around 25-45 Ma due to a change in zonal wind patterns or plate migration (Hyeong et al. 2016;Kyte et al. 1993;Pettke et al. 2002;Ziegler and Murray 2007).This timing coincides with the transition from K-feldspar-rich to plagioclase-rich sediments at the Minamitorishima sites.The similarity between grain size variations of chemical digestion residues and previously reported atmospheric dust records in the North Pacific (Janecek 1985;Janecek and Rea 1983) (Fig. 7e) further suggests that the K-feldspars are related to eolian dust.However, further mineral-specific measurements are needed to conclude this point.The variations in the eolian flux in the North Pacific are less than a factor of 2 between 45 and 10 Ma (Janecek 1985;Janecek and Rea 1983).This is much smaller than the observed variations in K-feldspar content, which reach a factor of 5. Thus, if the majority of the K-feldspar is of eolian origin, the atmospheric dust before ~ 28 to 39 Ma would have contained much more K-feldspar than younger times.Because K-feldspar has high ice nucleation potential and dominating the ice formation by silicate dust, our results suggest that clouds over the Pacific before ~ 28 to 39 Ma tends to glaciate more easily.
Alternatively to simple eolian transport of K-feldspar, formation after eolian plagioclase may explain the observations.Although this reaction has not been considered for pelagic clay, the microtexture (Fig. 4c) indicates that at least some of the K-feldspars are formed through the authigenic overgrowth.Formation via dissolution and precipitation could convert plagioclase into isolated K-feldspar grains (e.g.Putnis and John 2010) and explain the grain size variations, even though we do not have positive evidence for authigenic origin for such isolated grains (Fig. 4b).Chemical composition data from isolated K-feldspars (Fig. 4b) argue against a diagenetic origin, as diagenetic K-feldspar has been reported to be chemically pure with minimal Na or Ca substitutions.However, we cannot rule out possible averaging over submicroscopic heterogeneity in authigenic replacement.Detailed microscopic work would clarify their origin (e.g. Lee and Parsons 2003).It should be noted that the studied sites are distributed within narrow paleolatitudes (Fig. 1a), where provenance changes around 25-45 Ma have been reported (Hyeong et al. 2016;Kyte et al. 1993;Pettke et al. 2002;Ziegler and Murray 2007).Examination of samples outside this latitudinal band would test if the K-feldspar enrichment reflect time-dependent authigenic reactions or eolian mineralogy reflecting dust provenance.

Conclusions
Pelagic clays around Minamitorishima Island, western North Pacific, contain varying amounts of K-feldspar.Before ~ 35 Ma it is > 10%, while after ~ 15 Ma it is 2-3% (Fig. 6).IODP Site U1366C in the South Pacific also shows K-feldspar enrichment over plagioclase before ~ 44 Ma (Usui et al. 2018), but its concentration in the bulk sediments is not particularly high.Volcanic glass and zeolite show stratigraphic variations different from K-feldspar.Previous geochemical data suggest that volcanic input is less than the mass fraction of K-feldspar.Thus, we conclude that K-feldspar is not directly associated with volcanic input such as tephra.The K-feldspar appears to contain Na, in contrast to nearly pure authigenic feldspars (Kastner and Siever 1979), although we cannot rule out nanoscale heterogeneity.Grain size variations of chemically separated feldspars and quartz are similar to those of eolian dust in the North Pacific (Fig. 7)  (Janecek 1985;Janecek and Rea 1983).These observations can be explained either by the eolian of the K-feldspars or by authigenic formation replacing eolian plagioclase.If they are eolian dust, then our data indicate that the past atmospheric dust had much higher ice-nucleating activity than modern dust.Alternatively, if the K-feldspar is of authigenic origin, the use of pelagic clays as archives of past eolian dust minerals should be reconsidered.Further study in geographic distribution of K-feldspar enrichment is warranted to constrain past atmospheric dust and potentially widespread redistribution of Na and K within pelagic sediments.e Temporal variation of median grain size from LL44-GPC3 (Janecek and Rea 1983) and DSDP Site 576 (Janecek 1985).The geomagnetic polarity timescale (GPTS) between 40 and 25 Ma (Ogg 2020) is shown on the right.The chemostratigraphic units (Tanaka et al. 2020a) are shown on the right of c and d.For a color cord, see Fig. 1

Fig. 1 a
Fig.1a Map showing the positions and backtracks of the study site (Minamitorishima) and the sites mentioned in the text.Dots represent paleopositions for every 5 Myr back to 70 Ma.Plate motion was calculated using GPlates version 2.3 (https:// www.gplat es.org) with the rotation model of(Müller et al. 2019).Correlation between b KR13-02 PC05 and c MR14-E02 PC11 by magnetic susceptibility (MS) and chemostratigraphy(Tanaka et al. 2020a).Magnetic susceptibility was measured with 7 cc cube specimens using Kappabridge (AGICO, KLY-4).Arrows indicate age constraints obtained by ichthyolith stratigraphy(Ohta et al. 2020), Osmium isotopes stratigraphy(Nozaki et al. 2019;Ohta et al. 2020), and magnetostratigraphy(Usui and Yamazaki 2021).lE stands for late Eocene, and eO for early Oligocene.Orange horizontal line in c indicates a hiatus

Fig. 3
Fig.3The effect of chemical treatments on the XRD diffractogram of a, b MR14-E02 PC11, 2.67 m, c, d MR14-E02 PC11, 8.99 m, and e, f KR13-02 PC05 7.99 m.Blue, green, and red lines show untreated samples, samples after CBD treatment, and samples after sodium pyrosulfate fusion treatment.The spectra was aligned and normalized using the quartz peak at about 26.6°.Black vertical lines are diagnostic peak positions for plagioclase (b), K-feldspar (d), and clinoptilolite (f)

Fig. 7
Fig. 7 a, b Grain size distributions of selected fusion residues of a MR14-E01 PC11 and b KR13-02 PC05.Data are offset according to their depth, and the depth of each sample is indicated on the right.c, d Depth variation of median grain size from c MR14-E02 PC11 and d KR13-02 PC05.Samples with bimodal grain size distributions were not plotted.The paleomagnetic polarity zones of Usui and Yamazaki (2021) are shown on the right of c. Black indicates normal polarity, white indicates reversed polarity, and gray indicates a zone with an uncertain polarity pattern.eTemporal variation of median grain size from LL44-GPC3(Janecek and Rea 1983) and DSDP Site 576(Janecek 1985).The geomagnetic polarity timescale (GPTS) between 40 and 25 Ma(Ogg 2020) is shown on the right.The chemostratigraphic units(Tanaka et al. 2020a) are shown on the right of c and d.For a color cord, see Fig.1