Little is known about the mechanisms of magnetic susceptibility (χ) enhancement for the red-clay sequence on the Chinese Loess Plateau (CLP), in comparison to the overlying loess-paleosol sequence. Here we present a rock magnetic study of the red-clay sediments from the central CLP. Our results show that frequency dependence of χ (χfd = χlf − χhf, where χlf and χhf are χ measured at 470 Hz and 4700 Hz, respectively), χlf, and susceptibility of anhysteretic-remanent-magnetization (χARM) are linearly correlated within the red-clay sequence. This linear correlation indicates that the pedogenic magnetic minerals of the red-clay have a rather uniform grain size distribution as in the loess-paleosol sequence, and the grain size is independent of the degree of pedogenesis. Nevertheless, red-clay sediments are slightly more enriched in superparamagnetic magnetic particles than the overlying loess-paleosol sediments as indicated by the higher slope of the regression line between χfd and χlf.
 Global cooling in the Cenozoic, leading to the growth of large continental ice sheets in both hemispheres, may have been caused by the uplift of the Tibetan Plateau and the positive feedbacks initiated by it [Raymo and Ruddiman, 1992]. However, the details remain to be clarified. An ideal place to study detailed sedimentary and climatic processes associated with Tibetan uplift is the Chinese Loess Plateau (CLP), which is immediately adjacent to northeastern Tibet. The wind-blown deposits on the CLP consist of an uppermost loess-paleosol sequence and an underlying red-clay sequence [Liu, 1985]. This sedimentary sequence is one of the best terrestrial climate records for at least the last 8 Ma [An et al., 2001; Guo et al., 2002]. Although the mechanisms for magnetic susceptibility enhancements of Chinese paleosols has been extensively studied [Bloemendal and Liu, 2005; Deng et al., 2005; Liu et al., 2005a, 2005b, 2005c; Verosub et al., 1993; Zhou et al., 1990], few similar studies have been done on the red-clay sequence. This paucity of studies greatly hampers our understanding of the magnetic susceptibility enhancement mechanisms in the red-clay sequence and thus prevents obtaining paleoenvironmental information encoded by magnetic susceptibility variations in Chinese red-clay sediments.
 In this paper, we present a preliminary rock magnetic study of the red-clay sequence and the lower part of the loess-paleosol sequence from the Chaona section (Figure 1), in the central CLP, for the purpose of examining magnetic susceptibility enhancement mechanisms in the Chinese red-clay sediments.
2. Sampling and Experimental Procedure
 Powder samples were collected from the Chaona section (107°12′E, 35°6′N) using a sampling interval of 10–20 cm (Figure 1). The entire section is about 300 m in thickness with the upper 175 m comprising the loess-paleosol sequence and the underlying 125 m comprising the red-clay sequence. The Quaternary loess-paleosol sequences consist of 32 paleosols and 33 loesses. The age model of the Chaona section has been established [Lü et al., 2001] and it is based on control point ages (top and basal ages of astronomically-tuned paleosols from S1 to S32 for the loess-paleosol sequence [Heslop et al., 2000], and paleomagnetic reversal ages [Cande and Kent, 1995] for the red-clay sequence).
 The present study is mainly concerned with the rock magnetic properties of the red-clay sequence, but we also include the lower part of the loess-paleosol sequence for comparison. In this work, powder samples were measured every 20 cm from 90 m to 300 m in depth in the section, for a total of about 1050 samples. The magnetic susceptibility was measured using a Bartington MS2 susceptometer at frequencies of 470 Hz (i.e. χlf) and 4700 Hz (i.e. χhf). ARM was imparted using a 100 mT peak AF and a 0.05 mT constant biasing field. This parameter is also expressed as χARM, after normalizing by the 0.05 mT direct bias field. From these measurements, two measures of frequency-dependant magnetic susceptibility (χfd%, defined as (χlf − χhf)/χlf*100, and χfd, defined as χlf − χhf) were calculated.
3. Results and Discussion
Figure 2 shows the χlf, χfd, χfd%, and χARM results for the interval ∼1.2–∼8.1 Ma from the Chaona section. χlf, χfd, and χARM are highly correlated with each other for the interval studied. These correlations are more obvious in Figures 3 and 4. χfd% is positively but non-linearly correlated with χlf, and reaches a plateau at ∼13% when χlf > ∼10 × 10−7 m3kg−1.
 Thermal magnetic analysis detects magnetite, maghemite, and hematite in the red-clay sequence [X. Liu et al., 2003; Song et al., 2006]. Since hematite is weakly magnetic and relatively hard magnetic mineral, its χlf, χfd, and χARM values are much lower than those of magnetite/maghemite [Thompson and Oldfield, 1986]. Therefore, the signals of χlf, χfd, and χARM are mainly a function of the concentration and the grain size distribution of magnetite/maghemite in the red-clay sequence. Previous workers have revealed that the final phase of pedogenically-produced strong magnetic mineral is maghemite [Chen et al., 2005; Deng et al., 2001, 2000; Heller and Evans, 1995; Q. Liu et al., 2003; Verosub et al., 1993], rather than magnetite, because these pedogenic magnetic particles with high surface to volume ratios are eventually oxidized into maghemite regardless of their initial states (magnetite or maghemite) [Q. Liu et al., 2003; Liu et al., 2004]. Therefore, values of the χlf, χfd, and χARM are primarily a function of the concentration and the grain size distribution of maghemite within both the loess-paleoseol and red-clay sequences. Previous work has shown that the parameter χfd is especially sensitive to grains of ∼20–∼25 nm in size, i.e., larger or viscous superparamagnetic maghemite [Bloemendal et al., 1985; Maher, 1988] and that the parameter χARM is sensitive to grains of ∼25–∼100 nm in size, i.e., larger stable single domain and smaller psudo-single domain maghemite [Banerjee et al., 1981; Dunlop and Özdemir, 1997; King et al., 1982].
 A recent magnetic study [Liu et al., 2005c] of magnetic extracts obtained from paleosol samples from the CLP shows that the dominant grain size of pedogenically-produced particles lies just above the superparamagnetic/single-domain threshold (∼20–25 nm). It also shows that the grain size distribution of pedogenically-produced magnetic particles is almost independent of the degree of pedogenesis in the loess-paleosol sequence on the CLP. Stronger (weaker) pedogenesis produces higher (lower) concentrations of ultrafine magnetite/maghemite of the same size [Liu et al., 2005c]. Although we did not obtain magnetic extracts in our study, the parameters χfd and χARM are primarily sensitive to magnetic grains of <∼100 nm. This size corresponds to the maximum grain size of strong magnetic minerals observed after magnetic extracts [Liu et al., 2005c]. Thus the similar linear relationships that we observe between χfd and χARM in both the lower part of loess-paleosol sequence and in the red-clay sequence strongly suggest that a similar magnetic grain size distribution pattern exists for magnetic particles of pedogenic origin in both sequences. Furthermore, the linear correlations between χlf and χfd, and between χlf and χARM, in the loess-paleosol sequence and in the red-clay sequence, strongly suggest that the enhancement mechansisms of magnetic susceptibility in both sequences are predominantly controlled by pedogenically-produced strongly-magnetic grains.
 Nevertheless, the data do show some subtle differences between the Chinese loess-paleosol sequence and the underlying red-clay sequence (Figures 3 and 4). First, the regression line of χfd to χlf in the red-clay sequence, which represents the χfd% after correcting for the effects of aeolian inputs, has a higher slope than that of the overlying loess-paleosol sequence (Figures 3 and 4). This difference has also been observed by [X. Liu et al., 2003] for the classic Xifeng section (Figure 1). Second, the regression line of χfd to χARM in the red-clay sequence has a higher slope than that of the lower part of loess-paleosol sequence (Figures 3 and 4), indicating that a higher proportion of viscous superparamagnetic maghemite grains to stable single domain maghemite grains are produced in the red-clay sequence than in the loess-paleosol sequence.
 The parameter χfd% is a sensitive parameter for detecting the presence of superparamagnetic grains in soils. Theoretically, χfd% could be ∼100% for samples containing only larger superparamagnetic particles. In the absence of the effects of pseudo-single domain/multi-domain magnetic grains, χfd% is inversely related to the width of pedogenic grain size distribution [Worm, 1998; Worm and Jackson, 1999]. Therefore, the higher values of χfd% observed after correcting for the effects of aeolian inputs in the red-clay sequence indicates that the pedogenically-produced magnetic particles may have a narrower grain size distribution than in the overlying loess-paleosol sequence. Alternately, it could indicate that the most dominant pedogenic magnetic grain size for the red-clay shifts closer superparamagnetic-stable single domain boundary than that in the loess-paleosol sequence. These differences in the specific shape and dominant grain size of the pedogenic grain size distribution probably reflect the effects of different temperature and/or source area, but not precipitation, conditions during the interval of red-clay deposition in comparison to the interval of loess-paleosol deposition. We conclude that precipitation is not significant because it primarily controls the variation in the concentration of pedogenic particles, and has little effect on the grain size distribution of pedogenic particles [Liu et al., 2005c; Maher and Thompson, 1995; Maher et al., 2003].
 1. Pedogenic processes in the red-clay sequence can produce a wide grain size range of magnetic particles, but the overall grain size distribution is relatively invariant as has been previously observed for the overlying loess-paleosol sequence. Therefore, fluctuations in susceptibility and ARM in the red-clay sequence are caused primarily by changes in the concentrations of these fine-grained pedogenic particles as has been previously observed for the overlying loess-paleosol sequence.
 2. Red-clay sediments are slightly more enriched in superparamagnetic magnetic particles than the overlying loess-paleosol sediments. The difference is probably caused by the slight difference in the width of pedogenically-produced grain size distribution and/or the dominant grain size. The differences in grain size characteristics are probably controlled by temperature and/or source materials during the soil-forming processes. Further rock magnetic works are needed to resolve the alternative possibilities.
 We thank L. Lü and Y. Song for providing the age-depth model of the Chaona section, and thank J. Bloemendal and C. Deng for sharing their loess data. Constructive comments on an earlier manuscript were supplied by Q. Liu. We also thank S. Yang, K. Fu, F. Wu, X. Chen, M. Dong, Y. Song, and L. Lü for help in the field work, and two reviewers for their rapid reviews. This work was co-supported by the NSFC (grants 40334038 and 40571171), the CAS Innovation Program (grant kzcx2-yw-104), the Chinese National Key Project on Basic Research (grant 2005CB422001), and the University of Rhode Island.