Microbial nitrification and acidification of lacustrine sediments deduced from the nature of a sedimentary kaolin deposit in central Japan

Kaolin deposits in the Seto-Tono district, central Japan, were formed by intense kaolinization of lacustrine arkose sediments deposited in small and shallow inland lakes in the late Miocene. Based on mineralogical and stable isotopic (Fe, C, N) studies of Motoyama kaolin deposit in the Seto area, we concluded that it was formed by microbial nitrification and acidification of lacustrine sediments underneath an inland lake. Small amounts of Fe–Ti oxides and Fe-hydroxide in the kaolin clay indicated that iron was oxidized and leached during the kaolinization. The field occurrences indicate that leached ferric iron precipitated on the bottom of the kaolin deposit as limonite crusts, and their significantly fractionated Fe isotope compositions suggest the involvement of microbial activity. The C/N ratios of most of the kaolin clay are distinctly higher than those of modern lacustrine sediment. Although, the possibility of a low-temperature hydrothermal origin of the kaolin deposit cannot be completely ruled out, it is more likely that acidification by dilute nitric acid formed from plant-derived ammonia could have caused the kaolinization, Fe oxidation and leaching. The nitrate-dependent microbial Fe oxidation is consistent with dilute nitric acid being the predominant oxidant.

www.nature.com/scientificreports/ The basement of the Seto-Tono district consists of Jurassic accretionary complex and Cretaceous to Paleogene granite batholiths 6 . In the early Miocene, shallow marine sediments of Mizunami Group partly overlayed the basement and, in the late Miocene to Pliocene, non-marine sediments of Tokai Group widely overlayed the former geologic units 7,8 . The kaolin deposits are embedded in the Lower Tokai Group, Seto porcelain clay formation (PCF) and the equivalent Toki PCF. They were deposited 9-11 Ma, based on geochronology of tuff intercalated in the stratigraphy (Ref. 9 ). During Pliocene, the Seto and Toki PCF were overlain by Yadagawa formation of the Upper Tokai Group, and the kaolin deposits survived erosion by the uplift after 3 Ma. This study focused on a Motoyama kaolin deposit which is exploited by Akatsuki, Inzo, and Kasen mines in the Seto area (Fig. 2).

Sedimentary facies of the porcelain clay formations. Sedimentary facies of the Seto and Toki PCF
have been extensively investigated by Refs. 10,11 , respectively. The studies indicate that the kaolin deposition occurred mainly in a lacustrine environment with subordinate fluvial incursions. A pedological study of the lacustrine sediments in the Tono area (Ref. 11 ) revealed that the inland lakes used to be ephemeral bogs with alternating spells of stagnant water and completely dry conditions. The climate would have been tropical to subtropical during the deposition 11 .
The Seto PCF containing the Motoyama kaolin deposit was classified into the following nine lithologies (Ref. 10 ); (1) poorly sorted matrix-supported gravel, (2) poorly sorted clast-supported gravel, (3) poorly sorted sand, (4) sand, medium to very coarse, (5) sandy gravel, (6) poorly sorted clay, (7) sand, mud, (8) well sorted clay, and (9) lignite. On the basis of their field occurrences, the following inferences regarding the depositional processes of the Seto PCF were made: in the early stage, sedimentary basins were separated into east and west parts by a mound of basement granite (see Fig. 2), and two alluvial fans comprising (4) to (6) were separately formed from northeast to southwest. Debris comprising (1) to (3) intermittently flowed into the alluvial fans. (6) was locally deposited in the periphery of the alluvial fans. In the later stage, the early sedimentary basins were reclaimed, and two alluvial fans merged into a large alluvial fan. (6)- (7) were widely deposited in the periphery of the alluvial fan and (8)- (9) were deposited in a bog beyond it along the flows from east to west. The PCF sedimented more slowly and gentle with time. In the latest stage, the inland lake expanded up to the western half of the Seto PCF distribution area.
Motoyama kaolin deposit. The Motoyama deposit is almost the same horizontally as that of the 500 m (EW) by 800 m (NS) oval open pit exploited by the mines. The Motoyama deposit is shaped like a shallow tray, and the strata is mostly flat (Fig. 3a,b). The stratigraphy is largely divided into three formations from the bottom to the top (Fig. 4): a late Cretaceous to Paleogene granite basement, the Seto PCF (15-45 m), and the Yadagawa formation (> 10 m) which is widely distributed in the Seto area. The outer margins of the Motoyama deposit are ambiguously bounded by gentle mounds of basement granite (Fig. 2); the field occurrence suggests that the deposit accumulated in a depression in the basement granite.  www.nature.com/scientificreports/ The Seto PCF comprises five kinds of sediments; kibushi clay, gaerome clay, kaolin clay (white earth), sandy clay, and silica sand. The gaerome clay is an unsorted and compact sediment consisting of coarse-grained quartz 1-5 mm in diameter (Fig. 3c) and fine-grained kaolin with small amounts of organic fragments, originating from arkose sandstone 10 . No carbonate minerals can be seen macroscopically. Gaerome literally means 'frog's eye' in Japanese and is called so because the coarse-grained quartz resembles a frog's eye.
The gaerome clay constitutes the main part of the Seto PCF, having a total thickness of 5-15 m. It varies from light gray to white due to bleaching. The kibushi clay is a dark brown ligneous clayey sediment consisting of finegrained kaolin and quartz with abundant carbonaceous materials. It is in the upper part of the formation, and is 1-6 m thick (Fig. 3a,d). Many in situ plant fossils are found in the gaerome clay, indicating a calm sedimentary environment. The peripheral part of the Motoyama deposit tends to be rich in quartz grains, whereas kaolin clay-rich parts (white earth) occur as lenticular layers and are irregularly distributed on a several meters-scale. The lower part of the Motoyama deposit comprises mainly silica sand, but their lateral continuity is often weak and lacking.
Limonite crusts 5-30 cm in thickness commonly appear along the unconformity plane between the gaerome clay beds and the basement granite (Figs. 3e,f, 4). The crusts are reddish-brown to orange massive aggregates consisting of pebbles cemented by goethite. The limonite crusts also occur as laminated thin beds or networks in gaerome clay beds in places, suggesting their concurrent or subsequent precipitation during the formation of the kaolin deposits. The weathered crusts of the basement granite underlying the kaolin deposit were generally subjected to kaolinization and converted to Fe-kaolinite bearing green saprolite. Incidentally, green saprolite (M07) collected from the neighboring Marufuji mine (Fig. 2) has been described elsewhere 13 .

Results
The gaerome clay (A07) and kibushi clay (A08) specimens were elutriated with a 250-µm sieve to remove coarse crystals, pebbles and carbonaceous plant fossils. Finer fractions, which were not chemically treated, were examined and analyzed. The weight ratios of coarser fractions of A07 and A08 are 27% and 14%, respectively. The coarser fraction of A07 consists mostly of quartz up to 8 mm with subordinate feldspars, carbonaceous materials and chert pebbles, whereas A08 consists mostly of carbonaceous materials up to 15 mm, with subordinate quartz and feldspar grains up to 5 mm in diameter. www.nature.com/scientificreports/ X-ray diffraction analysis. The X-ray diffraction analysis (XRD) patterns showed that the gaerome and kibushi clays consist mostly of kaolinite, quartz with subordinate feldspars, mica, and minor smectite (Fig. 5).
On comparing the peaks of the kaolinite in formamide-treated samples with those of the samples dried in air or at 110 °C, the peak shift of the basal spacing from 7 to 10 Å which indicates the presence of halloysite 14 could not be detected. Although the peaks from mica make the peak shift a little difficult to discern, it can be concluded that the gaerome (A07) and kibushi (A08) clays contain little halloysite.
Scanning electron microscopy of kaolin clay ores. Kaolinite in the gaerome and kibushi clay occurs as irregularly shaped flakes ranging from < 1 to 10 µm (Fig. 6a,b). No thick booklets or hexagonal kaolinite are found in either clay. Quartz and K-feldspar grains ranging from 10 to 50 µm are common in the kibushi and even more common in the gaerome clay (Fig. 6c,d). Plagioclase is rare in both. K-feldspar often exhibits a rough surface as if it had been etched, as would be expected from acid degradation during kaolinization (Fig. 6e,f). Trace amounts of Fe-Ti oxides, anatase, zircon, monazite, xenotime, allanite, gypsum, and framboidal pyrite also occur in both types of clay. is similar to that in Georgian kaolin, whereas the TiO 2 amounts are less than half of that in Georgian kaolin 15 . However, the Georgian kaolin measurements were performed on bulk ores, which contain between 0.6 and 3.3 wt% (mean 2.2 wt%) TiO 2 derived from anatase 16 , rather than on its minerals.
Occurrences of Fe-Ti oxides and Fe-hydroxide in kaolin clay ores. The gaerome and kibushi clays contain trace amounts of Fe-Ti oxides and Fe-hydroxide, which are likely to be decomposed residue of mafic silicates from the source sediments. To investigate redox conditions of kaolin clay ores, Fe-Ti oxides and Fehydroxide were separated from the ores (gaerome clay: A04, A05, A07 and kibushi clay: A06, A08) and investigated with a scanning electron microscope (SEM) and a field-emission electron microprobe analyzer (FE-EMPA). The Fe-Ti oxides are mostly ilmenite or its alteration products. The ilmenite occurs as euhedral to subhedral grains ranging from several to several hundred µm in diameter. The ilmenite grains were often altered to leached ilmenite or pseudorutile along cracks (Fig. 7a,b), and some grains were strongly corroded, leaving a Ti-rich skeleton (Fig. 7c,d). Framboid-like aggregates of ilmenite-pseudorutile are also found in kibushi clay (Fig. 7e,f). Fe-hydroxide is found in only gaerome clay, but the mineral species could not be identified because distinct XRD peaks were not detected due to low crystallinity. The Fe-hydroxide occurs as irregular-shaped, porous grains ranging from several tens to a thousand µm in diameter. The surface of the grains is often coated by rugged crusts and has well-defined rounded lumps 2-3 µm in diameter ( Fig. 8a-d). The body of the grains has a sponge-like texture (Fig. 8e,f).

FE-EMPA analysis of Fe-Ti oxides.
The chemical compositions of ilmenite and magnetite were recalculated as per (Ref. 17 ). However, Fe-Ti oxides whose composition lay between ilmenite and pseudorutile or pseudorutile and rutile (anatase) could not be recalculated using that method. The former has been recognized as leached ilmenite and the latter as pseudorutile or leached pseudorutile 18 . Therefore, the compositions of leached www.nature.com/scientificreports/ ilmenite and pseudorutile (leached pseudorutile) were recalculated by assuming a molecular ratio of 9 oxygen to 3 titanium atoms. The gaerome clay ores contain ilmenite, leached ilmenite, Fe-hydroxide, pseudorutile, leached pseudorutile, anatase and a trace amount of magnetite, whereas Fe-hydroxide and anatase could not be found in the kibushi clay. The recalculated compositions of Fe-Ti oxides and Ti oxide in the kibushi and gaerome clay ores are shown in Fig. 9. In Fig. 9a,c, most of the Fe-Ti oxides in the kibushi and gaerome clay ores fall on the tie lines connecting hematite-ilmenite, ilmenite-pseudorutile and magnetite-ulvöspinel, while the other Fe-Ti oxides lay between the ilmenite-pseudorutile line and rutile (anatase). On the basis of the mineral texture and compositional features of the Fe-Ti oxides, it can be assumed that successive oxidation and Fe-leaching probably occurred in the Motoyama deposits as follows 18 : In Fig. 9b,d, the hematite content in ilmenite varies widely from 0 to 89% in the kibushi clay ores and from 2 to 83% in the gaerome clay ores. It can be inferred that the Fe-Ti oxides are probably not detrital grains derived from the basement granite, because the granite is magnetite-free 19 and our preliminary analysis showed the granitic ilmenite contain less than 1% hematite. www.nature.com/scientificreports/ Geochemistry of kaolin clay ores. To investigate the formation processes of the Motoyama deposit, gaerome, silicic gaerome (gaerome clay rich in quartz) and kibushi clay samples were assayed for C, H, N and S and isotopes of Fe, C and N. Five kibushi clay ores and six gaerome clay ores were quantitatively assayed for C, H, N and S. The results of the analyses are listed in Table 1. There is a positive correlation between C and N concentrations (Fig. 10a). Due to the paucity of carbonate carbon in the samples, the C and N are thought to be derived from organic matter. Correlations between H and N or H and C are not clear, because most of the hydrogen originated from the kaolin clay. The S concentration ranges from 0.002 to 0.08 wt%. The molar ratio of C/N ranges from 1.5 to 55.6 and has a positive correlation with the C concentration (Fig. 10b).
The δ 13 C values of the kibushi and gaerome kaolin clay ores range from − 25.2 to − 28.4‰, and this range is common in C3 land plants and lacustrine algae 21 . Figure 11a shows a negative correlation between δ 13 C and C/N in the kibushi and gaerome clay ores. www.nature.com/scientificreports/ The δ 15 N values of both clays range from 1.9 to 5.5‰, which is a common range for sediments 23 . The δ 15 N values are almost constant except for one value (1.9‰), and no systematic changes between δ 15 N and C/N are seen (Fig. 11b).
Fe isotope compositions. The Fe isotope composition of the same five magnetic separates from the kibushi and gaerome clay ores described in the previous section were determined. In addition, the Fe isotope composition of a green saprolite, seven limonite crust and four fresh basement granite samples, as well as bulk samples, were determined.
The δ 56 Fe values of Fe-bearing minerals in the kibushi and gaerome clay ores, the green saprolite and the basement granite are around 0.2 ‰ including analytical errors, except for one granite sample (− 0.07 ‰). In contrast, δ 56 Fe values in the limonite crust range from − 0.53 to 0.91 ‰, indicating significant isotope fractionation (Fig. 12). www.nature.com/scientificreports/

Discussion
On the basis of field occurrences, mineralogy and geochemical data, the most plausible origin of the Motoyama kaolin deposit was nitrification and acidification of lacustrine sediments due to microbial activity. Dilute nitric acid produced by the nitrification of ammonia could have caused decomposition of feldspars and micas in arkose sediments, resulting in Fe-leaching and intense kaolinization.
Formation fields of kaolin clay. The unsorted nature of gaerome clays indicates that it was basically formed by post-depositional kaolinization rather than being external in origin; i.e., most kaolin clay was not derived from weathering products of basement rocks elsewhere, it was formed in situ. This is because gaerome clay readily separates into silica sand and kaolin clay when suspended in water. Thus, it is unlikely that the gaerome clay would have undergone transportation processes by streams. The transportation processes would occur mainly at the stage of arkose before kaolinization. Although, kibushi clay is well sorted, it would have been a stagnant bog sediment that experienced little transportation process, and it is reasonable to assume that kaolin clay was mostly formed in situ. The scarcity of halloysite in gaerome and kibushi clays also suggests that weathering products of basement rocks is not a main source of the kaolin clay. The formation of clay minerals commenced at the stage of arkose deposited in a lacustrine environment, which has a high permeability under almost neutral conditions, by illitization and smectitization of feldspars and micas. Due to the subsequent acidi- www.nature.com/scientificreports/ fication of sediments, kaolinization progressed, and the permeability gradually decreased. However, since many quartz grains survived kaolinization, the sediments did not form an impermeable layer, and it is presumed that an acidic groundwater environment could be maintained. www.nature.com/scientificreports/ Fe oxidation and leaching in kaolin clay ores. The occurrences of Fe-Ti oxides (Fig. 7a-d) and their compositional trend (Fig. 9) in the gaerome and kibushi clay ores indicates that ilmenite crystallized first, and then Fe in ilmenite was oxidized and leached out. As the gaerome and kibushi clays are largely depleted in Fe, presumably the limonite crusts are precipitates of the Fe that had leached out. Although, the temporal sequence of Fe oxidation/leaching and kaolinization is uncertain, their close field occurrence in the Motoyama deposit suggests that they happened concurrently. Therefore, elucidating the history of the iron is important in understanding the origin of the Motoyama deposit.
Atmospheric oxygen is unlikely to have been the oxidant because the lacustrine sediments would have been in suboxic or anoxic conditions due to the pore water having been mostly stagnant and isolated from the atmosphere. If the oxidant of the iron was hydrosulfuric acid, hydrothermal alteration would be the most probable origin, because hydrothermal solutions are the commonest source of sulfur and often induce kaolinization of rocks. In the case of hydrothermal kaolin deposits, clay minerals such as pyrophyllite and dickite that are often Table 1. Analytical results of C, TOC, H, N, S concentrations, and C, N, Fe stable isotopes. TOC total organic carbon, grm gaerome, kbs kibushi, s-grm silicic gaerome, gsp green saprolite, lmc limonite crust, bgr basement granite, SD standard deviation. -: not analyzed, C/N: molar ratio of carbon and nitrogen.  www.nature.com/scientificreports/ formed at higher temperatures and dissemination/veinlets of sulfides are common (for example Refs. 24,25 ). In addition, a zoning of alteration degrees towards hydrothermal veins is occasionally observed in deposits (for example Ref. 26 ). However, the evidence above has never been found in the Motoyama deposit. Another potential origin of kaolinization is low-temperature hydrothermal and pervasive alteration due to fluids of deep origin 27 . An example is the large-scale kaolin deposits in the Cornubia district of southwestern England 28 . If convective circulation of local groundwater occurred continuously in sedimentary basins of the Seto-Tono district driven by late-stage hydrothermal activities of the basement granite, the iron oxidation and kaolinization might have taken place. In that case, hydrothermal Fe-Si scale (iron hat) similar to limonite crusts might be formed on granites (for example Ref. 29 ). However, the primary (higher temperature) hydrothermal alteration of the basement granites in the Seto-Tono district has never been reported as being similar to that of the Cornubia district. Moreover, the Figure 11. Graphs showing the relationships between δ 13 C and C/N molar ratios and between δ 15 N and C/N molar ratios in kibushi and gaerome clay and Lake Biwa sediments. In the compositions by Ref. 22 , data after 1950 were excluded to avoid anthropogenic effects such as nitrogen fertilizer. www.nature.com/scientificreports/ Seto-Tono district has been located on the trench side of the volcanic front, at least after Miocene 30 , and thus no igneous activity occurred there after Miocene. The tuff beds found in the kaolin deposits are presumably derived from volcanoes on the back-arc side. A possible hydrothermal source might be the fore-arc (Arima-type) hot spring of deep origin 31 , but no Arima-type hot spring is known to have existed in the district. Thus, no evidence of late Miocene hydrothermal activity has been found in the Seto-Tono district, and so iron oxidation by hydrosulfuric acid of hydrothermal origin is unlikely. As described below, the most plausible oxidant of iron was dilute nitric acid in groundwater, which may have been produced by microbial nitrification of plant-derived ammonia.

No
Fe isotope fractionation. The speculation above is consistent with the significant Fe isotope fractionation found in limonite crusts; i.e., Fe isotope fractionation would occur in the processes of oxidation and leaching of iron, and it is likely that the activity of nitrate-dependent Fe oxidizing bacteria, whose activity has been well recognized in experiments (for example Refs. 32,33 ), were involved. Furthermore, it is probable that ferric iron, which is barely soluble in neutral water, readily migrated as dissolved ferric nitrate in acidic conditions and precipitated along the boundary of pH and permeability (= an unconformity plane) as limonite crusts. Moreover, it is geologically improbable that the Fe isotope variation stemmed from the diversity of iron origins, because the lacustrine sediment would have virtually been a closed system with respect to Fe. As the Motoyama deposit was formed ten million years ago, it is unfeasible to obtain direct evidence of microbial activities (for example Ref. 34 ), but the occurrence of framboid-like ilmenite-pseudorutile (Fig. 7e,f) and the regular texture of Fe-hydroxide (Fig. 8) imply the activity of Fe-oxidizing bacteria. Numerous studies on Fe isotopes to date have shown that Fe isotope fractionation occurs in biotic or abiotic 35,36 . The studies on Fe isotope fractionation in soil have reported that Fe isotope compositions produced by the dissolution of silicates tended to be slightly lighter than the original isotope ratios 37,38 , whereas other studies on microbial Fe isotope fractionation with Fe oxidation have shown that Fe isotope compositions of Fe-hydroxide precipitates were generally heavier than that of ferrous iron in aqueous solution 39,40 . In the case of Motoyama deposits, it is difficult to identify at what stage Fe isotope fractionation occurred effectively from the present data, and more detailed examination is required.

C and N concentrations and isotope compositions.
To obtain clues to the origin of the Motoyama kaolin deposits from the C and N concentrations and isotope compositions, it is necessary to know how they have changed since then. Therefore, modern lacustrine sediments from Lake Biwa were compared as an analog of the original lacustrine sediments of the Seto area. Lake Biwa, which is located 80 km west of the Seto area, is larger and deeper (over 41.2 m deep) than any inland lake would have been in Miocene. A positive correlation between C and N contents in the Motoyama deposit means that these elements originally made up plant bodies. The fact that the C/N ratios of gaerome clay are significantly lower than those of kibushi clay is attributable to that the organic matter of gaerome clay was mainly derived from aquatic plants such as lacustrine algae, whereas that of kibushi clay was mainly land plants. The compositions of the Lake Biwa sediments are significantly less carbon-rich than those of the Motoyama deposit (Fig. 10). The C/N ratios are concentrated around 10 20,41 , suggesting that the organic matter is mostly of aquatic plants origin 42 .
In gaerome and kibushi clays, 13 C was depleted with increasing C/N ratios, whereas 13 C was depleted with depth in Lake Biwa sediments 20 . Such variations in carbon isotopes can be caused not only by changes in the source of carbon, but also the selective preservation of 13 C-depleted organic compounds by bacterial and diagenetic degradation 43 . In the case of Lake Biwa sediments, it was concluded that the δ 13 C variation was mainly due to burial diagenesis 20 . However, in the case of Motoyama deposit, the above causes would be acting in combination.
Due to the narrow range of δ 15 N in the kaolin clay ores, little information was obtained regarding the origin of the Motoyama deposit. The C/N ratio of Lake Biwa sediments is almost constant throughout the δ 15 N range (Refs. 22,41 ), because δ 15 N compositions are not sensitive for organic matter degradation 43 .

Nitrification and acidification of lacustrine sediments. The degradation of feldspars and micas into
kaolin clay is the main cause of the formation of the Motoyama kaolin deposit. Acidification (pH < 4.0) is necessary for the degradation of feldspars 44 . In case of hydrothermal argillic alteration, SO 2 and HCl gases are the main cause of acidification 45  . Acidification increases Al 3+ activity in the system, resulting in kaolin clay formation. Of the acidic anions, NO 3 − is the most likely agent for acidification of the lacustrine sediments. HCO 3 − is the major acid anion in soil solution, because H + of H 2 CO 3 is partially consumed by the reactions with bases in soils. Thus, the role of HCO 3 − in acidification should be taken into account. The activity of HCO 3 − in solution is a function of pH and CO 2 partial pressure such that: where the parentheses denote solution activity and PCO 2 refers to atmospheric CO 2 partial pressure (Ref. 46 ). HCO 3 − activity decreases with decrease in pH. Even greater than 3% CO 2 , HCO 3 − activity is extremely low (ca. 5 μeq/L) at pH 4.0 (Refs. 46,47 ). Therefore, HCO 3 − would not be the major anion behind kaolinization, and would play only a secondary role through acidification. On the other hand, SO 4 2− is a strong acid anion that would play a major role in the formation of hydrothermal kaolin deposits 48 . SO 4 2− , however, is not likely to cause kaolinization of the Motoyama deposit due to the above-mentioned geologic circumstances. Organic acids are important agents in mineral weathering near the surface 49,50 . Although, the behavior of organic acids in soil and sediments www.nature.com/scientificreports/ is not yet fully understood, organic acids released from plant roots play major roles in the decomposition of minerals and ingestion of nutrients from soil in the rhizosphere 51 . Therefore, it is reasonable to assume that they contributed a little to the formation of the Motoyama deposit.

Formation processes of the Motoyama kaolin deposit. A shallow inland lake was formed in the
Miocene, and into it flowed weathering products derived mainly from the surrounding granite. The surface of the lake became covered with vegetation, resulting in it becoming like a bog with intermittent dry periods 11 . Anaerobic bacteria decomposed the vegetation, generating an abundance of ammonia. Microbial nitrification of the ammonia occurred in the subsurface under suboxic conditions, and NO 3 − made the sediments acidic (pH < 4.0), resulting in kaolinization. Vegetation flourished whenever the lake refilled with water, but the above events kicked in during dry periods. Some NO 3 − was used for oxidation of iron (denitrification) and, due to dilute nitric acid solutions, resultant ferric iron leached and migrated from the sediments into limonite crusts (Fig. 13).
The amount of nitrogen required for kaolinization under various assumptions can be estimated as follows: where the parentheses denote weight %, E is the efficiency factor of the whole system, and V is the amount of nitric acid solution produced. The amount of nitrogen in solution depends on pH; the required amount of nitrogen increases with decreasing pH. Assuming that nitrogen in plants is 2 wt%, pH of solution = 3.0, temperature is 25 °C, and the efficiency factor is 0.3, 35 kg/m 2 of plant material is necessary for the production of 15 m 3 solution of dilute nitric acid, which is the column volume of the kaolin deposit. The values are reasonable given the biomass of wetlands (for example Ref. 52 ), and kaolinization of lacustrine sediments can occur at least within decades. The alternating spells of stagnant water and dry conditions in the Seto-Tono district were ideal for kaolinization by providing sufficient decomposition of bog plants. This special environment would have been made possible by the slow subsidence of the district that occurred in the late Miocene 53 , and the warmer climate at that time promoting the nitrification reactions 11 . Methods X-ray diffraction analysis. XRD was performed on gaerome and kibushi clays (Rigaku Smart Lab, Geological Survey of Japan: GSJ). The analytical protocol was as follows: CuKα radiation at 40 kV and 100 mA in a range 3-70° (2θ) with a scanning rate of 10° min −1 , using a high-speed one-dimensional detector.
Scanning electron microscopy. Kaolin clay powder, Fe-Ti oxides and Fe-hydroxide extracted from kaolin ores were observed with a scanning electron microscope (JEOL JSM-6610LV at GSJ). In addition, those were embedded in epoxy resin, one side was polished, and its surface was observed with a field-emission electron microprobe analyzer (JEOL JXA-8530F at GSJ).
The extraction of Fe-Ti oxides and Fe-hydroxide was performed using the following method: after elutriation of the ores with a 500-µm sieve, the finer fractions were added to distilled water in a beaker and the precipitates collected. Fe-Ti oxides were extracted from the precipitate with a Franz isodynamic separator. Ti oxide was also extracted from the precipitate by hand-picking.
Field-emission electron microprobe analysis. Fe-Ti oxides and Fe-hydroxide extracted from kaolin ores embedded in the epoxy resin were quantitatively analyzed by FE-EMPA with a 15 kV accelerating voltage and a 20 nA beam current. The raw data were corrected using the ZAF method. C, H, N, and S concentration analysis. The analysis was conducted at Kyuden-Sangyo Co. Inc., Fukuoka, Japan. To analyze total organic carbon (TOC), samples were pretreated with 4 N HCl at 170 °C for two hours to remove carbonate. The C, H and N concentrations were analyzed by an elemental analyzer (CHN628, LECO). Sample (ca. 50 mg) was ground and placed in a tin capsule which was then burned in oxygen, oxidizing C to CO 2 , H to H 2 O, and N to NO x . The combustion gasses were collected in a ballast tank and homogenized. The amounts of H 2 O and CO 2 , and hence C and H, in the gas samples were determined by infrared spectroscopy. The NO x was reduced to N 2 by Cu and the amount of N 2 determined by thermal conductivity. The S concentration was analyzed by ion chromatography (DIONEX ICS-1500, Thermo Scientific). A ground sample (ca. 5 mg) was placed in a crucible and pyrolyzed, converting S into SO x , using a pretreatment device (AQF-100, Mitsubishi Chemical). The combustion gas was absorbed into a solution of tartaric acid (10 ppm) and hydrogen peroxide (900 ppm), forming sulphate ions. The solution was injected into an ion chromatograph to quantify the SO 4 2− and hence the S content. C and N isotope measurements. The samples used for the C, H, N and S analysis were also used for the C and N isotope measurements. For the C isotope measurements, an amount of sample containing 20-50 μg in carbon was placed in a tin capsule. The capsule was burned in the furnace (EA, Eurovector), and the emitted CO 2 gas was separated in a separation column and introduced into a mass spectrometer (IsoPrime at the Kyuden-Sangyo Co., Inc.) with helium gas as the carrier. Each isotope ratio is the mean value of two measurements. The isotopic compositions are reported using the delta notation as the per-mil (‰) deviation of the 13 C/ 12 C ratio of the sample relative to that of PDB (Pee Dee Belemnite) using the following equation: www.nature.com/scientificreports/ Isotopically calibrated acetanilide (δ 13 C = − 28.9 ‰) was used as a working standard. For the N isotope measurement, an amount of sample containing 15 μg of carbon was placed in a tin capsule, and treated in the same way as the C isotope measurements. Each isotope ratio is the mean value of two measurements. The isotopic compositions are reported using the delta notation as the per-mil (‰) deviation of the 15 N/ 14 N ratio of the sample relative to that of the atmospheric N 2 using the following equation: Isotopically calibrated acetanilide (δ 15 N = − 0.9 ‰) was used as a working standard.
Fe isotope measurements. The Fe isotope compositions were measured using a multi-collector ICP-MS (Neptune Plus, Thermo Scientific) at the Research Institute for Humanity and Nature, Kyoto, Japan. Instrumental mass bias was corrected using the standard-sample-standard bracketing method 54 . The analyses were performed in middle-resolution mode (M/ΔM = 8000-9000) in order to eliminate the major isobar interferences of 40 Ar 14 N + , 40  Ni + on 58 Fe + were monitored using 52 Cr + and 60 Ni + , and the contributions were corrected using the isotopic abundance ratios of 54 Cr/ 52 Cr = 0.0282 and 58 Ni/ 60 Ni = 2.616 (Ref. 55 ). The isotopic compositions are reported δ 13 C PDB (‰) = 13 C/ 12 C sample / 13 C/ 12 C PDB − 1 × 1000.  www.nature.com/scientificreports/ using the delta notation as the per-mil (‰) deviation of the 56 Fe/ 54 Fe ratio of the sample relative to that of IRMM-014b (Institute for Reference Materials and Measurements) using the following equation: The reproducibility of the standard during these analyses was ± 0.10 ‰ (2σ).