The origin of illite and its effect on geochemical weathering indices

Illite, ranging from 52 to 71% and averaging 60%, dominates clay minerals throughout the NLK section (Fig. 4a). Higher illite contents were found in the two paleosol units compared to the loess units (Fig. 4a). This pattern is similar to that of the Luochun and Longxi sections in the CLP (Zheng et al., Reference Zheng, Gu, Han and Deng1985) and even the Xiashu loess, located in the middle and lower reaches of the Yangtze River (Shi et al., Reference Shi, Zhang, Dai, Song, Yu and Zheng2005), but inconsistent with the Wugong section located on the south of the CLP (Huang et al., Reference Huang, Zhao, Liu, Tan and Koopal2011) and in the Beremend Section, Hungary (Fig. 4b; Varga et al., Reference Varga, Újvári and Raucsik2011). The increase in illite content can be explained in two ways in the CLP: (1) finer eolian dust, carried by weaker winter monsoon winds, contains more illite (Ji et al., Reference Ji, Chen and Lu1999); and (2) detrital minerals in the loess have been transformed into illite by pedogenesis (Zheng et al., Reference Zheng, Gu, Han and Deng1985; Gylesjo and Arnold, Reference Gylesjo and Arnold2006).

Ji et al. (Reference Ji, Chen and Lu1999) summarized three major processes with respect to the illite formation in the weathering zone: (1) neoformation of illite from the weathering of feldspars; (2) illitization of smectite by repeated wetting and drying; and (3) physical and chemical weathering of muscovites to illite. Smectite contents in the NLK loess are too low to have produced significant illite. Illitization of smectite under wetting and drying alternation is always incomplete, and the transformation process can produce significant amounts of mixed-layer illite/smectite (I/S) (Rutter and Ding, Reference Rutter and Ding1993). No mixed layer I/S was identified in the XRD pattern, which excludes the second mechanism. Thus, we choose to focus on the other two processes as an explanation of the high illite contents in the paleosols.

Illite can form during weathering of K-feldspars in continental deposits (Chamley, Reference Chamley1989; Thiry, Reference Thiry2000). However, the main feldspar mineral in the NLK loess is albite (Supplementary Note S1). No lath-shaped crystals were observed by SEM in the two paleosol units (Fig. 5a–b). Consequently, we conclude that the illite in NLK loess has not been produced by the weathering of feldspars. Although muscovite is easily abraded by physical weathering, muscovite is ground into clay-size grains only in the laboratory, not in nature (Srodon and Eberl, Reference Srodon and Eberl1984). Moreover, muscovite is so resistant to chemical weathering that Galovic (Reference Galovic2016) attempted to utilize it to study geological processes experienced by loess deposits in eastern Croatia, based on its edge and surface micro-textures. That, combined with the biotite-like nature of illite in NLK loess (see the following), suggests that illite here has not originated from the weathering of muscovite.

Figure 5

(color online) The scanning electron microscopy (SEM) observations of clay minerals. The photographs (a) and (b) were taken from weak paleosol layers. The rounded outlines (c) of clay minerals suggested a dominantly physical weathering environment and the embayed edges (d) showed intensive dissolution during weathering. The energy dispersive spectrometer (EDS) spectra are also presented to identify clay minerals (e and f).

Therefore, we suggest that the illite in the paleosols has a detrital, not a pedogenic, origin, most likely provided by physical weathering of rocks rich in biotite (similarly to that which the chemical index of illite has exhibited), such as gneisses, granites, and granitoids. Using the geological map of Ren et al. (Reference Ren, Niu, Wang, He, Jin, Xie and Zhao2013; IGMA5000) and a previous report on the lithology of Kyrgyz West Tianshan (Rolland et al., Reference Rolland, Alexeiev, Kröner, Corsini, Loury and Monié2013; Orozbaev et al., Reference Orozbaev, Hirajima, Bakirov, Takasu, Maki, Yoshida and Sakiev2015), it is clear that there are some Ordovician, Silurian, and Devonian granites and small amounts of Permian gneisses in the western parts of the Ili Basin. The illite directly eroded from these bedrock areas by physical weathering may have been transported by fluvial processes to the loess source areas (Gallet et al., Reference Gallet, Jahn, Lanoe, Dia and Rossello1998). Subsequently, it may have been deflated and entrained by wind and transported to the NLK site, giving rise to the broken and rounded outline observed on many of the illite particles (Fig. 5c).

Generally, illite is preserved well in cold and dry environments, and its crystallinity decreases because of hydrolysis and leaching processes in warm and humid environments (mean annual precipitation of 500–1000 mm). The expansible layers become more prevalent in the crystal structure of illite and even lead to the transformation into the expandable 2:1 layer minerals vermiculite or smectite (Chamley, Reference Chamley1989). However, a map of 500–1000 mm isohyets identifies areas having a corresponding sub-humid and humid climate, revealing that such high precipitation levels rarely occur at low altitudes in central Asia. Even in the CLP, illite is not easily weathered in the warm and wet conditions of the East Asian Monsoon (Zhao et al., Reference Zhao, Ji, Chen, Liu, Chen and Balsam2005). Therefore, we conclude that the crystal structure of illite in NLK loess has not changed due to weathering.

Illite crystallinity (IC) and the chemical index of illite (CII) may be more sensitive climate metrics than the absolute amounts of illite because IC and CII are less influenced by sediment dynamic differentiation (Wang and Yang, Reference Wang and Yang2013). The IC value of illite was determined by measuring the half-peak width of the 10 Å peak, and expressed in °Δ2θ (Pandarinath et al., Reference Pandarinath, Prasad and Gupta1999; Ji and Browne, Reference Ji and Browne2000; Ehrmann et al., Reference Ehrmann, Setti and Marinoni2005). High values indicate lower crystallinities, whereas low IC values indicate strong crystallinities. The CII refers to the ratio of the 5 Å and 10 Å peak areas. In this study, the IC values ranged from 0.213 to 0.296, with an average of 0.252; CII values varied from 0.227 to 0.523, with an average of 0.332, suggesting that the illite in the NLK loess has high crystallinity and a more biotite-like structure (Diekmann and Wopfner, Reference Diekmann and Wopfner1996; Ehrmann et al., Reference Ehrmann, Setti and Marinoni2005). Therefore, these data indicate little to no modifications of illite in this section.

In general, both IC and CII decrease in dry and cold climates, where the hydrolysis reaction is slow, but attains higher values in humid and warm climates, where hydrolysis can be rapid (Chamley, Reference Chamley1989; Pandarinath, Reference Pandarinath2009). However, the IC and CII profiles of the NLK section display opposite trends (Fig. 4a), especially at depths of 1200–1800 cm (about 40–50 ka, see Fig. 2c). Kübler (Reference Kübler1967) defined 0.42 and 0.25 as the respective upper and lower bounds of an anchimetamorphic zone or “anchizone.” Therefore, we ascribe the opposing trends in the IC and CII profiles to changes in lithology of the Tianshan bedrock. These bedrock units provide sediments for the potential source areas of NLK loess at ca. 40–50 ka, while tectonic movement may be responsible for the lithological changes (e.g. Yang et al., Reference Yang, Li and Hunag2013); however, the potential source areas show limited changes, because the mean grain size (MGS), i.e., wind strength (see below), does not appear to affect variations in IC and CII values.

The MGS is regarded as an indicator of wind intensity in the CLP (Sun et al., Reference Sun, Wang, Liu and Clemens2010) because stronger winds can transport coarser particles. The grain size distribution in the loess/palaeosol sequence of Wels-Aschet in Upper Austria is also related to the degree of weathering, with pedogenesis decreasing the MGS in the loess (Terhorst et al., Reference Terhorst, Ottner and Wriessnig2012). Quartz grains are more resistant to weathering and remain largely unaltered during post-depositional processes, so their grain size is considered as a more reliable indicator of wind strength (Xiao et al., Reference Xiao, Porter, An, Kumai and Yoshikawa1995). Supplementary Figure 1a shows the grain size distributions of quartz grains isolated from loess (yellow) and paleosol (red) samples. The quartz modal grain size is finer in the paleosol unit than in the primary loess unit. From this we can deduce that wind strength was weaker during periods of pedogenesis, and stronger during periods of loess deposition. The grain size distributions of bulk samples display similar characteristics to those of quartz samples mentioned above (Supplementary Fig. 1b), whereby soil unit modal peaks (red and orange) are smaller than those of primary loess (blue and green). Therefore, we argue that wind strength, rather than post-depositional pedogenesis, has the greatest influence on grain size distribution at NLK. In addition, distance from the loess section to the source area also influences grain size distribution (Ding et al., Reference Ding, Sun, Rutter, Rokosh and Liu1999a, Reference Ding, Derbyshire, Yang, Sun and Liu2005; Yang and Ding, Reference Yang and Ding2008). However, this factor mainly influences the sand content (>63 µm) of loess deposits in loess-desert transition zones (e.g., the northern part of the CLP; Ding et al., Reference Ding, Sun, Rutter, Rokosh and Liu1999a). The NLK is >400 km from the deserts to the west of the Ili Basin (Fig. 1a), and, considering the unmixed end-member of loess grain-size distributions, we found that the fine sand end member is absent in the NLK loess (Li et al., Reference Li, Song, Fitzsimmons, Chang, Orozbaev and Li2017). This suggests that sand fraction may not have had significant impact on the grain size distribution patterns of the NLK loess. Therefore, the influence of distance from the loess profile site to the source area(s) on grain size in NLK loess can be neglected, also implying that MGS is an indicator of wind intensity at the NLK site.

Illite in the NLK section originated directly from physical weathering of parent rocks, and maintained its original structure during transportation and deposition processes. In addition, the MGS variations indicate that wind intensity decreased when weak paleosols developed. Thus, the fine fractions, such as clay minerals, were predominantly transported to the sections, not weathered in-situ, and the enrichment of illite in paleosol units compared to that in loess units was mainly caused by the weaker wind intensity rather than pedogenesis. This interpretation is also supported by a wind tunnel experiment (Wang et al., Reference Wang, Lang, Hua, Zhang and Li2017). The experimental results suggested that there is an inverse correlation between wind velocity and the content of fine fractions with a diameter of <2.5 µm (cf. fig.5 in Wang et al., Reference Wang, Lang, Hua, Zhang and Li2017).

Muscovite contents in bulk samples (Supplementary Note S1) are negatively correlated with MGS (Fig. 4a). Muscovite has a flakey form, implying that it can be readily transported by wind and is not easily deposited even under light winds. This would lead to the sorting of minerals with different crystal morphologies, especially when winds weakened, resulting in the differences in muscovite content between loess and paleosol units (Fig. 4a).

Idealized muscovite gives a chemical index of alteration (CIA) value of 75, and the CIA value of illite ranges between 75 and 85 (Nesbitt and Young, Reference Nesbitt and Young1982). We argue that wind transported more illite and muscovite to the NLK site during the development of the paleosol units, which increased the CIA values of the paleosols. That also indicates the influence of grain size sorting on the major and trace elements contents in loess deposits (Gallet et al., Reference Gallet, Jahn and Torii1996; Bauluz et al., Reference Bauluz, Mayayo, Fernandez-Nieto and Lopez2000; Xiong et al., Reference Xiong, Zhu, Zhou, Lu and Ding2008; Liang et al., Reference Liang, Sun, Beets, Prins, Wu and Vandenberghe2013; Ben Israel et al., Reference Ben Israel, Enzel, Amit and Erel2015; Li et al., Reference Li, Song, Chen, Li, Mamadjanov and Aminov2016b). Therefore, the CIA values of paleosol units in loess sections may be overestimated, especially for those sections in which the paleosols have higher illite contents than the loess. Moreover, illite is enriched in Al when compared to plagioclase (Nesbitt and Young, Reference Nesbitt and Young1982; Mclennan, Reference Mclennan1993). Thus, particular attention must also be paid to the use of Chemical Proxy of Alteration (CPA) to indicate the degree of silicate weathering (Buggle et al., Reference Buggle, Glaser, Hambach, Gerasimenko and Markovic2011).

The origin of chlorite and kaolinite and their abilities to reconstruct paleoclimate

Chlorite accounts for 21−35% of the clay fraction in loess at the NLK section, with an average of 28%, while kaolinite accounted for relatively smaller proportions, with an average of 9% (Fig. 4a). The variations in chlorite and kaolinite contents are independent of loess lithology (Fig. 4a). Kaolinite can form rapidly under a strong chemical weathering regime and its abundance may be an indicator of subtropical to tropical climates (Chamley, Reference Chamley1989; Ruffell et al., Reference Ruffell, McKinley and Worden2002). Given the location of the NLK section in the interior of Eurasia, it can be inferred that the kaolinite in the NLK loess was derived from physical weathering of kaolinite-bearing sedimentary rocks in Tianshan, instead of a weathering product in the source area or in the loess section itself. The occurrence of kaolinite here might not represent wet and warm conditions.

Trioctahedral Fe/Mg-rich clay minerals, e.g., chlorite, are sensitive to chemical weathering, while dioctahedral Al-rich clay minerals such as kaolinite are difficult to weather. Chlorite is the second-most abundant clay component in the section and indicates weak weathering (Chamley, Reference Chamley1998; Terhorst et al., Reference Terhorst, Ottner and Wriessnig2012) in the NLK loess. Chlorite is a common mineral in metamorphic rocks undergoing green schist facies metamorphism and it also exists in sedimentary rocks. Chlorite is the direct product of parent rocks which have undergone little chemical weathering (Robert and Kennett, Reference Robert and Kennett1994); it is not formed pedogenically in the loess sequence.

Kaolinite and chlorite contents have negative relationships in the NLK loess (Fig. 4a). The appearance and preservation of detrital chlorite is connected to the physical weathering of rocks under cold and dry climates (Ehrmann and Mackensen, Reference Ehrmann and Mackensen1992; Ehrmann et al., Reference Ehrmann, Melles, Kuhn and Grobe1992). In contrast, the presence of detrital kaolinite is mainly connected to the chemical weathering of soils in warm and wet climates (Robert and Kennett, Reference Robert and Kennett1994). Therefore, the incompatibility of chlorite and kaolinite is as expected and it seems that both could have been inherited from the same parent rocks, most likely sedimentary rocks. However, the correlation coefficient (R²; depth >200 cm) is 0.31. The low R² value may imply that there are additional sources of chlorite, e.g., metamorphic, but not sedimentary, rocks. Moreover, notably, chlorite contents decrease in the middle of the second paleosol unit, perhaps resulting from either relatively stronger weathering or greater dilution by illite (Fig. 4a and 5d).

The detrital clay minerals in sedimentary rocks from some basins can be used to reliably reflect the paleoenvironmental or tectonic conditions (Liu et al., Reference Liu, Colin, Trentesaux, Siani, Frank, Blamart and Farid2005b; Berrocoso et al., Reference Berrocoso, Zuluaga and Elorza2008; Hong et al., Reference Hong, Zhang and Li2010; Riboulleau et al., Reference Riboulleau, Bout-Roumazeilles and Tribovillard2014). However, loess deposits, as eolian sediments, have more complicated sources. They have experienced at least one cycle of transportation to a source area (Gallet et al., Reference Gallet, Jahn, Lanoe, Dia and Rossello1998) and have been well mixed during transportation by wind. Therefore, the paleoclimatic signals in the detrital clay source minerals in loess deposits can be obscured. Both chlorite and kaolinite in NLK loess were inherited from the bedrock in the Tianshan area, mainly by physical erosion, and mixed during eolian transport. Therefore, it is difficult to extract information about paleoclimate in the provenance region or depositional area. However, the presence of abundant chlorites indicates weak weathering within the NLK loess.

The origin of smectite and its paleoclimate implication

The NLK section contains only a small amount of smectite (possibly including mixed illite-smectite species). The smectite contents in the loess are higher than those in paleosols such as the Wugong section (Huang et al., Reference Huang, Zhao, Liu, Tan and Koopal2011). Volcanic materials can easily be transformed into smectite during early diagenesis or by hydrothermal and chemical weathering processes (Chamley, Reference Chamley1989; Cuadros et al., Reference Cuadros, Caballero, Huertas, De Cisneros, Huertas and Linares1999); and, even in marine sediments, authigenic smectites also can be generated through submarine weathering of basaltic or pyroclastic materials (Biscaye, Reference Biscaye1965; Petschick et al., Reference Petschick, Kuhn and Gingele1996; Ehrmann, Reference Ehrmann1998). However, the volcanogenic smectites are almost always associated with tuff layers in sedimentary sequences (Deconinck et al., Reference Deconinck, Amedro, Baudin, Godet, Pellenard, Robaszynski and Zimmerlin2005; Do Campo et al., Reference Do Campo, del Papa, Nieto, Hongn and Petrinovic2010). No volcanic ash layer was observed in the NLK section: thus, smectite in the NLK loess is unlikely to have this origin. Variations in smectite contents do not correlate with those of kaolinite and chlorite, indicating that the smectite had different origins. Therefore, it can be expected that smectite in NLK loess formed by transformation or neoformation in the source or deposit areas during a weathering process.

Dioctahedral smectite can form as the first weathering product of plagioclase, while trioctahedral smectite generally forms during the weathering of biotites (Pal and Deshpande, Reference Pal and Deshpande1987; Srivastava et al., Reference Srivastava, Parkash and Pal1998). To estimate the crystal structure of smectite, eleven samples were selected from the loess and paleosols to analyze the position (angle) of the (060) diffraction peak. Strong 060 reflection at 1.54 Å indicates that these are trioctahedral smectites in all the selected samples (Supplementary Fig. 2). Furthermore, some of the illite has a biotite-like nature, according to the CII values. Therefore, smectite in the NLK loess probably formed by weathering of biotite or transformation of illite with removal of interlayer K.

Increasing available moisture in soil promotes leaching and, consequently, the release of K from micaceous minerals (Khormali and Kehl, Reference Khormali and Kehl2011). A calcareous environment with high Mg²⁺ and high Si⁴⁺ mobility may provide favorable conditions for the formation of smectite through transformation. The formation of smectite can hence reflect paleohydrology (Chamley, Reference Chamley1989). However, temporal changes in smectite content resemble those of MGS in NLK section (Fig. 4a). MGS can reflect the wind intensity, with weaker winds possibly leading to the relatively low contents of smectite. This characteristic may reflect the tendency of finer particles in loess to become aggregated (Pye, Reference Pye1995; Mason et al., Reference Mason, Jacobs, Greene and Nettleton2003; Qiang et al., Reference Qiang, Lang and Wang2010; Mason et al., Reference Mason, Greene and Joeckel2011; Újvári et al., Reference Újvári, Kok, Varga and Kovács2016), especially for the clay-sized minerals, by electrostatic forces, salts, or organic matter (Pye, Reference Pye1995). This view is also supported by the observations of bulk samples (Fig. S3). Smectite tends to form aggregates during transport due to its high hygroscopic capacity (Singer et al., Reference Singer, Dultz and Argaman2004). The stronger winds can carry larger aggregates, and thus more smectite, to the loess depositional areas. Therefore, smectite was most likely formed in the source area, then transported to the NLK site. Its presence may not indicate moisture conditions, due to the impact of wind intensity. Although the influence of larger calculation errors on the variations in smectite content cannot be excluded, more research on the effect of climate vs. weathering on smectite contents in loess deposits is needed.

The suitability of clay minerals as tracers of dust provenance

In light of the genesis of clay minerals in NLK loess, we believe that clay mineralogy is a potentially useful tracer of dust provenance. The significance of aeolian dust lies in its ability to reconstruct large-scale atmospheric circulations. Dust in ice cores so far has been studied by many researchers (e.g., Kohfeld and Harrison, Reference Kohfeld and Harrison2001, and references therein). Discrimination of potential source region(s) would enable a better understanding of past dust transport paths, assisting the development of model simulations of past atmospheric circulation patterns. The approach of tracing source region(s) involved comparing the specific “fingerprints,” such as mineralogy, chemistry, and isotopes, of potential source areas and dust in the ice cores (Fuhrer et al., Reference Fuhrer, Neftel, Anklin, Staffelbach and Legrand1996; Biscaye et al., Reference Biscaye, Grousset, Revel, VanderGaast, Zielinski, Vaars and Kukla1997; Maggi, Reference Maggi1997; Steffensen, Reference Steffensen1997). Central Asia is also one of the major dust source areas (Muhs, Reference Muhs2013), and the present-day dust flux there accounts for 7.5% of the annual mean global emission flux (Tanaka and Chiba, Reference Tanaka and Chiba2006), or even 16% according to Miller et al. (Reference Miller, Tegen and Perlwitz2004). However, it has not been considered as a potential source region, maybe due to a lack of the data that can serve as the “fingerprint” (Biscaye et al., Reference Biscaye, Grousset, Revel, VanderGaast, Zielinski, Vaars and Kukla1997; Svensson et al., Reference Svensson, Biscaye and Grousset2000; Újvári et al., Reference Újvári, Stevens, Svensson, Klötzli, Manning, Németh and Kovács2015). Here, we presented clay mineralogical compositions of NLK loess in Central Asia, which supplement data on the potential source regions. The kaolinite to chlorite (K/C) ratio is usually a diagnostic employed for the discrimination of source areas (Biscaye et al., Reference Biscaye, Grousset, Revel, VanderGaast, Zielinski, Vaars and Kukla1997; Újvári et al., Reference Újvári, Stevens, Svensson, Klötzli, Manning, Németh and Kovács2015), based on its latitude dependency (Biscaye, Reference Biscaye1965; Griffin et al., Reference Griffin, Windom and Goldberg1968; Scheuvens et al., Reference Scheuvens, Schutz, Kandler, Ebert and Weinbruch2013) and the local lithology and climate and drainage characteristics. Because of the complex behaviors of smectite during emission, transport and deposition, we used a K/C ratio to illite content plot (Újvári et al., Reference Újvári, Stevens, Svensson, Klötzli, Manning, Németh and Kovács2015) to discriminate and identify the source areas from which the dust in Greenland ice cores originated (Fig. 6). These potential source areas can be separated from each other to a large extent. Data points of Greenland ice cores (Biscaye et al., Reference Biscaye, Grousset, Revel, VanderGaast, Zielinski, Vaars and Kukla1997; Svensson et al., Reference Svensson, Biscaye and Grousset2000) are mainly located in the East Asian and central Asias areas, implying that the extensive arid or semi-arid hinterlands of Asia are important dust sources for Greenland ice cores.

Figure 6

(color online) Kaolinite/chlorite (K/C) versus illite contents of Greenland last glacial maximum (LGM) dust and samples. The distribution ranges of the potential source areas in this figure are confined based on the data of loess from East Asia (11 samples); Central Asia (this study, 137 samples); Northeast Asia (1 samples); Alaska (8 samples); and Europe (10 samples). Data from: Biscaye et al., Reference Biscaye, Grousset, Revel, VanderGaast, Zielinski, Vaars and Kukla1997; Svensson et al., Reference Svensson, Biscaye and Grousset2000; Ujvari et al., 2015.