palaeosol sequence at Lingtai (Central Chinese Loess Plateau)

Physics of the Earth and Planetary Interiors 140 (2003) 255–275

Detrital and pedogenic magnetic mineral phases in the loess/palaeosol sequence at Lingtai (Central Chinese Loess Plateau) S. Spassov a,∗ , F. Heller a , R. Kretzschmar b , M.E. Evans c , L.P. Yue d , D.K. Nourgaliev e a Institut für Geophysik, ETH Zürich, CH-8093 Zürich, Switzerland Institut für Terrestrische Ökologie, ETH Zürich, CH-8952 Schlieren, Switzerland c Institute for Geophysical Research, University of Alberta, Edmonton, Alberta, Canada T6G 2J1 d Department of Geology, North-West University, Xi’an 710069, China Palaeomagnetic Laboratory, Faculty of Geology, University of Kazan, 18 Lenin st., Kazan 420008, Russia b

e

Received 17 February 2003; received in revised form 5 September 2003; accepted 7 September 2003

Abstract A detailed rock magnetic investigation of loess/palaeosol samples from the section at Lingtai on the central Chinese Loess Plateau (CLP) is presented. Thermal demagnetisation of isothermal remanent magnetisation (IRM) and Curie temperature measurements suggest the presence of magnetite, maghemite and hematite as remanence carrying components. Bulk and grain size fractionated samples have been analysed using coercivity spectra of remanence acquisition/demagnetisation curves, which identify four main remanence carriers in different grain size fractions of loesses and palaeosols. A linear source mixing model quantifies the contribution of the four components which have been experimentally derived as dominating endmembers in specific grain size fractions. Up to two thirds of the total IRM of the palaeosols are due to slightly oxidised pedogenic magnetite. Two detrital components dominate up to 90% of the IRM of the loess samples and are ascribed to maghemite of different oxidation degree. Detrital hematite is present in all samples and contributes up to 10% of the IRM. The iron content of the grain size fractions gives evidence that iron in pedogenically grown remanence carriers does not originate from the detrital iron oxides, but rather from iron-bearing clays and mafic silicates. The contribution of pedogenic magnetite to the bulk IRM increases with the increasing degree of pedogenesis, which depends in turn on climate change. © 2003 Elsevier B.V. All rights reserved. Keywords: Rock magnetism; Component analysis; Loess; Palaeosol; China

1. Introduction ∗

Corresponding author. Present address: Department of Geophysics, School of Geology, Aristotle University of Thessaloniki, P.O. Box 352-1, GR-54124 Thessaloniki, Greece. Tel.: +30-2310-998-485; fax: +30-2310-998-528. E-mail address: [email protected] (S. Spassov).

From deposition to the final stages of diagenesis, sediments are affected by many environmental processes. Transport processes provide detrital input and weathering processes cause dissolution and precipitation of minerals, which includes iron release for

0031-9201/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.pepi.2003.09.003

256

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

the inorganic or biochemical neoformation of magnetic minerals. These processes are mainly driven by climate, as it has been demonstrated for the loess/ palaeosol sequences of the Chinese Loess Plateau (CLP) (e.g. Heller and Evans, 1995). Over the past 2.6 million years, loess has been deposited during cold and arid periods whereas palaeosols developed under warm and humid conditions. Magnetic minerals have been enriched during soil formation and the magnetic susceptibility of palaeosols has been enhanced (cf. Fig. 1). The susceptibility time series of the Chinese loess is dominated by the Earth’s orbital parameters such as precession, obliquity and eccentricity (Heller and Liu, 1986; Wang et al., 1990; Heslop et al., 2000). Magnetic and geochemical sediment parameters have been used for quantitative estimates of palaeoclimate parameters. The magnetic susceptibility signal may be divided into two components, a weak detrital component which depends on the dust accumulation rate and a second component of pedogenic origin which emerges during interglacial periods, when higher precipitation and temperature favour the neoformation of magnetic minerals phases. The enhancement of magnetic susceptibility in palaeosols may not be tied to a single palaeoenvironmental parameter because soils result from a complex interplay between the amount of rainfall, its annual distribution, temperature, orographic position, biological activity and chemistry acting on the loess bedrock during weathering. Magnetic methods offer the great advantage over other methods like X-ray diffraction or colourimetry that extremely low concentrations of ferromagnetic phases can be detected and identified without costly sample preparation and within short measurement time. For instance, concentrations of ∼0.12 ppm (by volume) of pure hematite or goethite and ∼0.79 ppb (by volume) of pure maghemite or magnetite can be identified in a rock matrix using a modern vibrating sample magnetometer. In addition, magnetic parameters give quantitative information about sediment ages and environmental processes of sediment formation from aeolian deposition to anthropogenic pollution. Bulk magnetic measurements, however, represent a smoothed integral over all magnetic mineral components present. Parameters such as low field susceptibility, median destructive field, S-ratio, Day-plot parameters, etc. may not give clear evidence

about the contribution of different magnetic minerals in loess/palaeosol samples. The amount, type and grain size of the individual magnetic mineral phases present in a rock may be derived from coercivity distributions of acquisition or demagnetisation of laboratory produced magnetisations. Robertson and France (1994) were the first to model acquisition curves of isothermal remanent magnetisation (IRM) in order to quantify magnetic mineral populations from different sources—also called endmembers—present in a sample. This unmixing technique was further developed by Stockhausen (1988), Kruiver et al. (2001) and Heslop et al. (2002). Each individual mineral component is characterised by a single theoretical distribution function which is in a first order approximation adapted to be of lognormal nature. It is assumed that the bulk IRM corresponds to the sum of the contributions of all individual components/endmembers (source mixing model). Carter-Stiglitz et al. (2001) could show indeed that the magnetisations of their materials are mixing linearly. In order to model a measured IRM acquisition curve, an unknown possibly infinite number of lognormal functions is necessary. Therefore source mixing or unmixing can result in complicated calculation problems using the techniques available. Egli (2003a) proposes a generalised function for coercivity populations that models endmember distributions without assuming specific (e.g. lognormal) distributions. This approach reduces the number of distributions needed to fit the experimental results. Recently, two theoretical investigations on coercivity distributions showed indeed that these distributions are not necessarily of lognormal nature (Egli, 2003b; Heslop et al., 2003). The method of source mixing can be applied to the loess/palaeosol sediments of the CLP. Evans and Heller (1994) proposed the coexistence of two magnetic mineral components in loess/palaeosol sediments on the CLP and tried to estimate their contributions from the derivatives of IRM acquisition curves. The first component is a detrital population of magnetic particles, which is always present and appears to be uniform across the loess plateau. The second component is a superimposed authigenically grown population varying from layer to layer within a section and from site to site. Eyre (1996) extended this model to a four-component model.

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

257

Fig. 1. Litho- and magnetostratigraphy of the upper part of the Lingtai loess/palaeosol section. From left to right: lithology (white: loess layers, grey: palaeosol layers, lined pattern: carbonate concretion horizons), low-field susceptibility, NRM polarity frequency dependence of susceptibility (F-factor). The depth resolution of the magnetic records is about 0.02 m. Palaeosols are labelled ‘S’ and loesses ‘L’, followed by the consecutive number of the loess or palaeosol layer from top to bottom. The uppermost layer is the Holocene soil labelled ‘S0’, the underlying loess ‘L1’ and so on. The magnetostratigraphy gives evidence of the Matuyama/Brunhes polarity boundary being located in loess L8 at a profile depth of about 62 m (for details see Spassov et al., 2001). The letters J and O denote the Jaramillo and the Olduvai geomagnetic polarity subchrons, respectively. The grey bar at the upper Olduvai boundary indicates that multiple polarity flips occur over a depth interval of 2.88 m. The F-factor (measured at two frequencies 0.47 and 4.7 kHz) is generally between 5 and 12%, with lower values being observed only in fresh and silty loess such as L4, L9 or L15.

258

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

The present study takes a different approach in identifying remanence carriers and their grain size. We measure IRM properties of different bulk and of grain size fractionated loess/palaeosol samples and calculate their coercivity spectra. Some prominent coercivity spectra of the fractionated samples are used to evaluate the endmembers present in the bulk samples. IRM curves of bulk samples are then reconstructed by varying the contribution of the observed endmembers. This approach aims at obtaining more detailed information about the interaction between climate, deposition, alteration and pedogenesis of loess on the central CLP.

2. Sample description The samples have been collected on the central CLP from the loess/palaeosol sequence at Lingtai (34.98◦ N, 107.56◦ E), Gansu Province, China. The upper part of this section consists of 33 pairs of loess/palaeosol layers down to 175 m profile depth. The part below is formed by Mio-/Pliocene red clay with a basal age of about 7.05 Ma (Ding et al., 1998) at a depth of 305 m. In order to cover a wide range of magnetic properties, samples have been selected from the well developed palaeosols S1 and S5 and from the pristine loess L4 (Fig. 1). Weakly developed soils (S7, S8) and loesses of different weathering stage (L7, L8) have been considered, too. Palaeosol S7 is genetically related to loess layer L8 because it formed partly by alteration of loess L8 during warmer climate conditions. An additional loess sample has been chosen from the western CLP where pedogenesis is weaker in a much more arid climate than on the central CLP. This sample (By-L4) originates from loess layer L4 of the Baicaoyuan section (35.7◦ N, 104.9◦ E). Evans and Heller (1994) considered this sample to be not affected by pedogenesis.

3. Experiments and procedures Two types of samples, bulk samples and grain size fractionated samples, were analysed. The samples to be fractionated were disaggregated and decalcified (NaOAc, 1 M, pH 5) using standard laboratory procedures (Klute, 1986). The disaggregated samples were wet-sieved (mesh size 50) to separate the

fraction >50 ␮m. The remaining split ≤50 ␮m was fractionated by sedimentation in water into two grain size classes of 20–50 and <20 ␮m. The split <20 ␮m was fractionated by sedimentation and centrifugation into another four grain size fractions: 5–20, 2–5 and 0.2–2 ␮m (coarse clay) and ≤0.2 ␮m (fine clay), respectively. Parts of all six fractions have been mixed with wax (Hoechst Wachs C) and pressed into pellets, with a constant mass ratio of sample/wax (4.4). The iron and titanium concentrations of the fractionated samples were measured using a X-Lab 2000 energy-dispersive X-ray fluorescence spectrometer (Spectro) equipped with a sequence of secondary targets to generate polarised X-rays. The lower detection limit is 0.5 mg/kg (5 × 10−5 %, Scheinost et al., 2002). Low and high coercivity minerals are expected in the loess/palaeosol samples. In order to determine the low coercivity mineral content of the original bulk samples, anhysteretic remanent magnetisation (ARM) and isothermal remanent magnetisation (IRM) was demagnetised by alternating fields. After initial demagnetisation in an alternating current (ac) peak field of 300 mT, the samples were given an isothermal remanent magnetisation in a direct current (dc) field of 300 mT (IRM300 mT ). An ac peak field of 300 mT with a superimposed dc field of 0.1 mT was utilised for the ARM experiment. Both laboratory remanences were measured and demagnetised using a 2 G cryogenic magnetometer with on-line alternating field (AF) coils. AF demagnetisation along the magnetised axis was started after passing a constant waiting time of 3 min (in order to eliminate viscosity effects) using 45 logarithmically distributed steps between 0 and 300 mT. The high coercivity mineral content of the fractionated samples was tested by stepwise acquisition of IRM using an impulse magnetiser. The pressed pellets were pulverised again and part of this powder (containing sample and wax) was pressed into small gel cups. These gel cups were then magnetised step by step using 38 logarithmically distributed data points ranging from 1 to 4.5 T and always measured after a waiting time of 3 min when the IRM had become stable within measurement time. The magnetisation of the samples has not been corrected for the wax content because of its small mass and remanence contribution (<0.08%). The remanence acquisition and demagnetisation spectra have been calculated following the method of Egli (2003a). The first part of his method concerns

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

IRM and log. IRM gradient [mAm²/kg]

the calculation of coercivity spectra from magnetisation curves: the field of the measured curve is scaled to obtain a symmetric sigmoidal shape of the graph. A scale transformation of the magnetisation linearises the curve. The scaled graph is then fitted with a hyperbolic tangent function, which is assumed to be the (unknown) noise-free magnetisation curve. The residuals between measured and fitted curve are calculated in order to evaluate experimental errors. A Butterworth low-pass filter performs efficient noise reduction, whereby the appropriate choice of filter order and cut-off frequency is eased by the comparison of the filtered and unfiltered Fourier spectra of the residual curve. The subsequent backward transformation results in a fairly noise-free magnetisation curve. Next, the noise-free magnetisation curve is scaled using a logarithmic field scale (base 10). Thus, the field axis becomes unitless and the resulting derivative—also called logarithmic coercivity spectrum (LCS)—has the same units as the magnetisation (see Fig. 2). The maximum error amplitude of the LCS is estimated by comparing the measured and filtered magnetisation curve and is shown as errorband. The IRM acquisition spectra were analysed using grain size fractionated samples in which a specific IRM component, also called endmember, clearly prevails. These experimentally defined components, but not mathematically determined distribution functions

259

as proposed by Egli (2003a), were adapted to model the spectra of the bulk samples. Thermal demagnetisation of the IRM completes the magnetic mineral characterisation. Powder samples have been mixed and shaken with CC High Temperature (OMEGA® ) cement in the mass-ratio 1:6 under dry conditions. After adding a liquid binder the samples were stirred and dried for 24 h in the shielded room of the laboratory. These samples were also AF demagnetised along three orthogonal axes using a peak field of 300 mT. Three IRM coercivity windows have been chosen to analyse different coercivity populations thermally. The samples were first magnetised at a higher dc field value and then demagnetised by a smaller AF field. The field pairs were: (40 and 20 mT); (120 and 70 mT) and (1500 and 280 mT) using one specimen per field pair. The remaining IRM was demagnetised thermally after a 48 h waiting time to reduce the contribution of viscous remanences. The samples were heated for about 40 min after the pre-adjusted temperature within the sample zone of the oven was reached. The susceptibility of the samples was monitored to assess mineralogical changes during the heating process after each of the 32–35 demagnetisation steps. Blank samples have been prepared in the same way for each magnetisation window and the magnetisation of the samples was corrected for the cement magnetisation. This method

IRM acquisition logarithmic IRM gradient with error band 0.6

0.4

0.2

0

0

1

2

3

3.65

Log10 (Field [mT]) Fig. 2. Acquisition curve of isothermal remanent magnetisation and calculated logarithmic IRM gradient (coercivity spectrum) of Lingtai sample Li-L4, fraction <0.2 ␮m. The acquisition curve—diamonds represent the individual measurements—shows two major slopes around 1.41 (∼26 mT) and 2.78 (∼610 mT) and the derivative of the acquisition curve exhibits two separate maxima. The grey band of the gradient curve represents the experimental errors. The maximal applied field amounts to 4500 mT.

260

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

was applied to grain size fractionated samples (field pairs: (40 and 20 mT) and (120 and 70 mT)) and to bulk samples (field pair: (1500 and 280 mT)). The Curie temperature of some grain size fractionated powder samples was measured using a Curie balance of the Kazan State University. Pressed powder samples (70 mm3 ) were heated in air from −196 to 700 ◦ C in fields of 50 mT (palaeosol fractions) and 160 mT (loess fractions). A fast heating rate of 150 ◦ C/min was chosen. Using mineral standards, it has been verified that the thermal lag of the heating curves does not exceed 10 ◦ C. Each sample was heated twice in order to assess mineralogical changes. The low field susceptibility of the original bulk samples and of the grain size fractionated samples was measured using a KLY-2 susceptibility bridge. As the grain size fractionated samples were mixed with wax, the wax susceptibility was subtracted. Each measurement was repeated five times because of the partly very weak signals.

4. Results The magnetic properties of different loess and palaeosol bulk samples have been tested using AF demagnetisation of ARM and IRM. The initial ARM intensity (Fig. 3, upper panel) divides the palaeosols clearly in two categories: weakly (Li-S7, Li-S8) and strongly magnetic palaeosols (Li-S1, Li-S5), respectively. The common property of these palaeosols, independent of their initial intensity, is the consistent LCS amplitude peak at 22 mT. The initial ARM intensities of the loesses are one order of magnitude lower compared to the palaeosols. The maximal LCS amplitude is around 33 mT. Similar to the palaeosols, one magnetic component with a wider coercivity spectrum dominates the ARM demagnetisation spectra of the loesses. AF demagnetisation of IRM yields different results and indicates the coexistence of more than one coercivity population in palaeosols and loesses (Fig. 3f and h). The loess spectra have a major peak around 50–60 mT. The long wide tail towards lower fields indicates a second low coercivity population peaking around 28 mT. The palaeosols in contrast exhibit a well-defined maximum around 20 mT and a long tail towards higher fields indicating another component,

which peaks around 70 mT. In contrast to the ARM, the IRM demagnetisation spectra of both loesses have higher resolution, because all grain sizes of all magnetic mineral phases are better addressed (Fig. 3). Therefore, IRM is used for further experiments. The IRM demagnetisation spectra of the loesses are not much different. In order to observe possible differences, the loess samples with largest magnetic contrast, Li-L4 and Li-L8, have been selected for grain size fractionation and further experiments (see Fig. 3h). Palaeosol Li-S7 has been chosen because it is genetically related to loess Li-L8 (Li-L8 is the parent material of Li-S7) and palaeosol Li-S1 as an example of a strongly magnetic palaeosol (Table 1). The IRM coercivity spectra of the grain size fractionated samples (Fig. 4, Table 2) do not only depend on lithology, but also on grain size. The smallest fraction (<0.2 ␮m) contains a low-coercivity component, which peaks between 21 and 27 mT in all samples. A high-coercivity component between 560–760 mT (H) is also present in this fraction, preferentially in the loesses. This high-coercivity component still contributes to the spectra in the fraction 0.2–2 ␮m. The low-coercivity component peaks here at slightly higher values. In the palaeosols, the latter peak is now at 31 mT (P) and in the loesses at 40 mT to 45 mT. Another component with a peak around 140 mT is also present in the 0.2–2 ␮m fraction, but in the loesses only. This component is strongly developed in the grain size fraction 2–5 ␮m of the loesses but much less apparent in the palaeosols, especially in the magnetically well developed palaeosol Li-S1. The palaeosol spectra of fraction 2–5 ␮m are dominated by a 34 mT peak, which is much less significant in the loesses. The loess grain size fractions 5–20 ␮m exhibit only a single coercivity component at 113 mT with almost identical and skewed spectra. The 5–20 ␮m fraction of Li-S1 is dominated by a low-coercivity component, which peaks at 40 mT, whereas a another component, which peaks at 93 mT, dominates in Li-S7. The spectra of both loesses are again identical and symmetric and peak at 79 mT in grain size fraction 20–50 ␮m. The spectrum of Li-S1 peaks here at 45 mT, that of Li-S7 at 79 mT. The spectra of the largest grain size fraction (>50 ␮m) show some similarities to those of fraction 2–5 ␮m spectrum for both palaeosols, whereas the spectra of the loesses appear as some combination between the 2–5 and 5–20 ␮m fractions.

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

261

AF demagnetisation of anhysteretic remanent magnetisation (ARM) Loesses

Palaeosols ARM [mAm²/kg]

0.8 0.4 0

(a)

log. ARM gradient [mAm²/kg]

0.12

Li-S5 Li-S1 Li-S8 Li-S7

0.5

1 1.5 2 Log 10 (Field [mT])

2.5

1.6 1.2 0.8 0.4 0

0

0.5

(b)

1 1.5 2 Log 10 (Field [mT])

2.5

Li-L8 By-L4 Li-L7 Li-L4

0.08 0.04 0

0

(c)

log. ARM gradient [mAm²/kg]

ARM [mAm²/kg]

1.2

0

0.5

1 1.5 2 Log 10 (Field [mT])

2.5

0

0.5

1 1.5 2 Log 10 (Field [mT])

2.5

0.12 0.08 0.04 0

(d)

AF demagnetisation of isothermal remanent magnetisation (IRM) Loesses

Palaeosols

4 0

log. IRM gradient [mAm²/kg]

IRM [mAm²/kg]

8

(e)

(f)

Li-S5 Li-S1 Li-S8 Li-S7

12

0.5

1 1.5 2 Log 10 (Field [mT])

2.5

16 12 8 4 0

0

0.5

1 1.5 2 Log 10 (Field [mT])

2.5

3 2 1

(g)

(h)

Li-L8 By-L4 Li-L7 Li-L4

4

0 0

log. IRM gradient [mAm²/kg]

IRM [mAm²/kg]

14

0

0.5

1 1.5 2 Log 10 (Field [mT])

2.5

0

0.5

1 1.5 2 Log 10 (Field [mT])

2.5

4 3 2 1 0

Fig. 3. (Upper panel) Demagnetisation curves of anhysteretic remanent magnetisation and their derivatives of palaeosols and loesses (bulk samples) from Lingtai, central CLP (prefix ‘Li’) except By-L4, which originates from Baicaoyuan, western CLP. The palaeosol spectra peak consistently around 22 mT. The loesses also show a single peak only, but at higher coercivities. (Lower panel) Demagnetisation curves of isothermal remanent magnetisation and their derivatives. The palaeosols show a blurred maximum around 20 mT. A further significant component is present around 70 mT. The loess spectra peak around 60 mT and exhibit a minor maximum around 28 mT.

262

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

Table 1 Low field susceptibility, susceptibility contribution, IRM contribution at 4.5 T and total iron and titanium content of four grain size fractionated loess and palaeosol samples from Lingtai Name horizon/depth

Grain size (␮m)

Susceptibility (10−8 m3 /kg)

Susceptibility contribution (10−8 m3 /kg)

IRM4.5 T contribution (mA m2 /kg)

Total iron (wt.%)

Total titanium (wt.%)

Li-S1/9.58 m

<0.2 0.2–2 2–5 5–20 20–50 >50 Stacked Bulk

601.10 565.39 201.17 107.29 70.14 101.01

36.07 96.12 34.20 33.26 18.94 2.02 220.60 229.40

0.52 5.64 2.40 3.30 1.88 0.15 13.88

7.620 7.271 5.034 2.989 1.820 2.451

0.159 0.487 0.645 0.530 0.384 0.347

Li-L4/30.68 m

<0.2 0.2–2 2–5 5–20 20–50 >50 Stacked Bulk

29.98 41.48 43.62 33.45 35.39 30.71

0.60 5.39 4.80 7.86 16.28 1.38 36.31 24.00

0.01 0.33 0.54 1.34 2.14 0.17 4.55

5.793 7.726 6.247 3.674 2.754 3.648

0.239 0.410 0.504 0.511 0.405 0.376

Li-S7/59.60 m

<0.2 0.2–2 2–5 5–20 20–50 >50 Stacked Bulk

236.59 164.03 89.76 47.46 35.81 59.79

14.20 26.24 16.16 14.71 9.67 1.20 82.17 77.41

0.17 1.24 1.17 1.84 1.27 0.09 5.79

8.561 7.669 5.552 2.972 1.755 2.612

0.181 0.469 0.506 0.505 0.376 0.283

Li-L8/60.64 m

<0.2 0.2–2 2–5 5–20 20–50 >50 Stacked Bulk

41.67 32.81 27.64 26.06 45.70 30.19

1.67 3.94 3.46 7.23 18.51 0.98 35.78 33.25

0.03 0.28 0.45 1.36 2.63 0.12 4.87

8.403 7.156 5.840 3.660 2.794 3.149

0.174 0.415 0.467 0.500 0.438 0.354

Susceptibility and IRM contribution are re-calculated taking the mass contribution of the individual grain size fractions into account. The stacked susceptibility agrees rather well with the susceptibility of the bulk samples. The difference in Li-L4 may be due to the decalcification of the grain size fractionated samples. The stacked IRM is also comparable in amplitude with the initial IRM of the bulk samples (cf. Fig. 3f and h).

Fig. 5 shows the IRM acquisition spectra of the fractionated samples on an absolute intensity scale. Amongst the stronger magnetic coarse fractions, fraction 5–20 ␮m has the highest magnetisation in the pristine loess Li-L4, whereas the two smallest fractions (<0.2 and 0.2–2 ␮m) are much weaker. The slightly weathered loess Li-L8 is quite similar, but has the strongest magnetisation in the coarse fraction 20–50 ␮m. Again the smallest fractions are

weakly magnetic like Li-L4. The palaeosols behave completely differently. The fine grain size fraction 0.2–2 ␮m is by far the most magnetic one. Its maximal amplitude in Li-S1 is very prominent and five times higher than in Li-S7. Accordingly, the other fractions are of subordinate importance in Li-S1. The coarser fractions—in particular fraction 5–20 ␮m—gain in importance in Li-S7. Interestingly, the maximal amplitude of this fraction is equal in the genetically

< 0.2 µm

H 560 - 760 mT

0.8 0.6 0.4 0.2 0 0

0.5

1

Normalised log. IRM gradient

2

2.5

3

3.5

Log10 (Field [mT])

(a)

P 31 mT

1 0.2 - 2 µm

0.8

0.4 0.2 0 0

0.5

1

1.5

2

2.5

3

3.5

Log10 (Field [mT])

2 - 5 µm

0.8 0.6 0.4 0.2 0 0

0.5

1

1.5

2

2.5

3

3.5

Log10 (Field [mT])

(c)

Li-L4

D2? 93 mT D 1

113 mT

5 - 20 µm

0.8 0.6 0.4 0.2 0 0

0.5

1

1.5

2

2.5

3

3.5

3

3.5

Log10 (Field [mT]) P? 2 45 mT 79 mT

1 20 - 50 µm

0.8 0.6 0.4 0.2 0 0

0.5

Li-L8

1

1.5

2

2.5

Log10 (Field [mT])

(e)

D1? P 34 mT 128 - 136 mT

1

1

263

D

P? 40 - 45 mT D1? ~ 140 mT

0.6

P? 40 mT

(d)

H 350 700 mT

(b)

Normalised log. IRM gradient

1.5

Normalised log. IRM gradient

1

P 27 mT

Normalised log. IRM gradient

L? 21 mT

Normalised log. IRM gradient

Normalised log. IRM gradient

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

D2? P? 38 mT 100 mT

1

> 50 µm

D1? 128 mT

0.8 0.6 0.4 0.2 0 0

0.5

1

1.5

2

2.5

3

3.5

Log10 (Field [mT])

(f)

Li-S7

Li-S1

Fig. 4. Normalised spectra of IRM acquisition curves of six grain size fractions of two loesses (Li-L4, Li-L8) and two palaeosols (Li-S1, Li-S7). Absolute values are given in Table 2. Four main components (bold letters) can be recognised in all four samples. Component H between 560 and 760 mT can be recognised best in fraction <0.2 ␮m, component P at (31 ± 4) mT in fraction 0.2–2 ␮m, component D1 at (113 ± 13) mT in fraction 5–20 ␮m and component D2 at (79 ± 10) mT in fraction 20–50 ␮m. The designation L, P, D and H is attributed to low, pedogenic, detrital and hematite coercivity components, respectively, and is based on subsequent interpretation.

related loess Li-L8. The IRM acquisition spectra of the different grain size fractions demonstrate, that mainly four coercivity components (bold and italic in Table 2) carry stable remanences in loesses and

palaeosols: P at (31 ± 4) mT, D1 at (113 ± 13) mT, D2 at (79 ± 10) mT and H at (610 ± 150) mT. The different grain size fractions apparently represent different coercivity populations. This is confirmed

264

Grain size (␮m)

Loess

Palaeosol

Li-L4 Peak field (mT)

Li-L8 Maximal amplitude (mA m2 /kg)

Error at maximum amplitude (%)

Peak field (mT)

Li-S7 Maximal amplitude (mA m2 /kg)

Error at maximum amplitude (%)

Peak field (mT)

Li-S1 Maximal amplitude (mA m2 /kg)

Error at maximum amplitude (%)

Peak field (mT)

Maximal amplitude (mA m2 /kg)

Error at maximum amplitude (%)

<0.2

26 610

0.57 0.20

2.99 9.54

30 560–760

0.54 0.34/0.35

5.35 14.2/13.1

23 650

3.50 0.44

2.61 36.50

21 ∼500

11.54 0.37

2.58 32.20

0.2–2

40 ∼125 463

1.92 1.17 0.78

3.24 8.48 12.80

45 ∼125 350–700

1.56 1.17 0.85/0.78

3.35 8.51 17.7/22.6

30 – 714

9.67 – 1.02

1.64 – 20.80

31 – –

46.71 – –

1.25 – –

2–5

∼34 128

2.59 4.99

3.23 4.45

∼40 136

1.70 3.96

4.41 5.98

33 ∼140

5.76 3.57

2.68 8.36

35 ∼125

14.72 7.10

2.26 6.64

5–20

– 113

– 6.24

– 4.45

– 112

– 5.25

– 2.65

∼40 93

4.16 5.31

2.38 2.96

40 ∼100

9.64 8.22

1.40 2.92

20–50

– 82

– 4.45

– 1.66

– 79

– 6.27

– 4.84

∼46 79

3.85 4.07

2.12 2.76

45 ∼80

6.55 5.17

2.44 4.01

∼57 128

2.73 4.35

2.94 3.68

∼50 100

2.62 3.40

3.41 4.42

44 ∼100

3.69 3.02

2.09 4.22

38 ∼100

7.26 4.79

1.76 4.19

>50

Peak field is the field at which a component reaches its maximal amplitude. The bold and italic components have been used as endmembers in the linear mixing model.

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

Table 2 Numerical values of IRM components observed in six grain size fractions of loess and palaeosol samples (cf. Figs. 4 and 5)

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

Loess

Palaeosol P (31 mT)

10

50

D 1 (113 mT)

8

40

6

30

4

Li–L4

H (610 mT)

2

Li–S1

20 10

0

0 1 2 Log 10 (Field [mT])

0 (a)

3

5–20 µm

0

1 2 3 Log 10 (Field [mT])

(c)

< 0.2 µm

0.2–2 µm

2–5 µm

20–50 µm

> 50 µm

10

10

D 2 (79 mT)

8

8

6

max. 5.3 mAm²/kg

6

4

Li–L8

2

max. 5.2 mAm²/kg

4

Li–S7

2

0

0 0

(b)

265

2 1 Log 10 (Field [mT])

3

0 (d)

1 2 3 Log 10 (Field [mT])

Fig. 5. IRM acquisition spectra of the fractionated samples of Fig. 4 plotted on an absolute scale. The largest amplitudes located at 31 mT (P) are found in the grain size fraction 0.2–2 ␮m in both palaeosols but this component plays a minor role in the loesses. The spectrum of grain size class 5–20 ␮m has the largest amplitude in loess Li-L4, whereas grain size class 20–50 ␮m is strongest in loess Li-L8. The 5–20 ␮m fractions of the genetically related loess Li-L8 and palaeosol Li-S7 are characterised by equal development of the D1 component. The high-coercivity component H has a very low intensity and is best observed in the grain size fractions <0.2 ␮m and 0.2–2 ␮m of Li-L4, Li-L8 and Li-S7.

by the thermal demagnetisation of IRM and Curie temperature measurements. Fig. 6 presents thermal demagnetisation results of the IRM given in the coercivity windows as discussed in Section 3. The intensity of the low-coercivity component (equal to remanences from grains with coercivities between 20 and 40 mT) of the palaeosol fractions 0.2–2 ␮m decreases until 400 ◦ C when a constant level is reached. It decreases again at higher temperature and disappears at 620 ◦ C completely (Fig. 6a). The magnetisa-

tion of the loess Li-L4 fraction 0.2–2 ␮m decreases rather monotonously and disappears also at 620 ◦ C. The susceptibility of the three samples increases by 20–30% between 300 and 600 ◦ C, probably due to the formation of new ferromagnetic minerals. The intermediate coercivity component residing in the fractions 5–20 and 20–50 ␮m of the two loess samples shows similar thermal behaviour (Fig. 6b). The remanence being unchanged until 100 ◦ C, decreases then and disappears at 675 ◦ C. The rapid loss up to

266

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

Susceptibility

Low coercivity component: 20 -- 40 mT

Low coercivity component: 20 -- 40 mT

M 0 = 7.42 mAm²/kg M 0 = 2.97 mAm²/kg M 0 = 0.40 mAm²/kg

Li- S1 Li- S5 Li- L4

0.8 0.6 0.2- 2 µm

0.4 0.2 0 0

100

(a)

300

400

500

600

0.6 0.4

0.2 - 2 µm

0.2

Li- S1 Li- S5 Li -L4

0 0

100

200

300

400

500

600

700

Intermediate coercivity component: 70 -- 120 mT

Intermediate coercivity component: 70 -- 120 mT

20 - 50 µm

M 0 = 0.18 mAm²/kg M 0 = 0.22 mAm²/kg

Li- L4 Li- L8

0.8

5 - 20 µm

0.6

Li- L4 Li- L8

0.4

M 0 = 0.27 mAm²/kg M 0 = 0.28 mAm²/kg

0.2 100

200

300

400

500

600

Norm. susceptibility

Temperature [˚C]

0

1 0.8

5 - 20 µm

0.6

Li- L8 Li- L4 20 - 50 µm

0.4

Li L8 Li L4

0.2 0 0

700

100

200

300

400

500

600

700

Temperature [˚C]

Temperature [˚C]

High coercivity component: 280 -- 1500 mT

High coercivity component: 280 -- 1500 mT

(b)

By- L4 Li -S1

1 Normalised IRM

0.8

700

0

0.8

M 0 = 0.46 mAm²/kg M 0 = 0.53 mAm²/kg

0.6 0.4 bulk

0.2 0 0

(c)

1

Temperature [˚C]

1 Normalised IRM

200

100

200

300

400

500

Temperature [˚C]

600

700

Norm. susceptibility

Normalised IRM

1

Norm. susceptibility

Magnetisation

1 0.8 0.6 0.4

bulk

0.2

Li- S1

0 0

100

200

300

400

500

600

700

Temperature [˚C]

Fig. 6. Thermal demagnetisation of IRM and low field susceptibility for the three different coercivity windows. The curves consist of 32–35 demagnetisation steps. (a) The grain size fractions of both palaeosols (Li-S1, Li-S5) which have magnetisations regarded to be of pedogenic origin, behave very similarly. The IRM of the same grain size fraction of loess Li-L4 differs to some extent but its susceptibility change is very similar to that of the palaeosols. Maximum unblocking of IRM occurs at ∼620 ◦ C suggesting the presence of maghemite. (b) Thermal demagnetisation of the two loess grain size fractions is slightly different. The remanence disappears above 665 ◦ C and indicates the presence of hematite. A substantial amount of magnetic minerals is formed during heating (susceptibility increase). (c) Hematite is regarded to be the high coercivity remanence carrier. Its remanence vanishes at 685 ◦ C in both samples. The increase in susceptibility indicates formation of new ferromagnetic minerals, but they do not acquire remanence. The susceptibility of By-L4 has not been measured.

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

1

Li–L4, 1. 2. Li–L8, 1. 2.

0.8 0.6 0.4 5 0.2

Tc = 610 ˚C

0 -200 -100

Normalised magnetisation

(b)

20 µm

0

Li–S1, 1 2

0.8

Tc = 575 ˚C

0.6 0.4 < 0.2 µm

0 -200 -100

0

100 200 300 400 500 600 700 Temperature [˚C]

1

(d)

Li–L4, 1. 2. Li–L8, 1. 2.

0.8 0.6 0.4 0.2 0 -200 -100

100 200 300 400 500 600 700 Temperature [˚C]

1

0.2

(c)

Normalised magnetisation

Normalised magnetisation

(a)

magnetisation disappears at about 610 ◦ C. A second heating cycle shows hardly any ferromagnetic contribution. The thermomagnetic behaviour of both loess fractions 20–50 ␮m (Fig. 7c) shows reduced paramagnetic influence below room temperature. The magnetisation decreases steadily and without inflection between 350 and 500 ◦ C until 610 ◦ C. The considerable loss of magnetisation below this temperature indicates the presence of magnetite. The second heating differs from the first heating indicating destruction of maghemite. The magnetisation curves of the palaeosols do not show paramagnetic behaviour. The magnetisation decreases quasi-linearly and disappears at temperatures of 575 and 625 ◦ C in palaeosol Li-S1, fraction <0.2 ␮m and palaeosol Li-S7, fraction 0.2–2 ␮m, respectively. (The second heating curve of Li-S1 is quite different and indicative for magnetite, which was formed during the first heating

Normalised magnetisation

250 ◦ C is stronger in the finer fraction. Another rapid loss occurs at 650 ◦ C in both fractions. The susceptibility starts to increase by about 60–80% between 400 and 500 ◦ C, indicating the formation of new magnetic minerals. The high-coercivity components of the pristine loess bulk sample By-L4 and the well developed palaeosol bulk sample Li-S1 decay quasi-linearly up to 660 ◦ C (Fig. 6c). Above this temperature the magnetisation decreases very rapidly and disappears at 685 ◦ C. The susceptibility increases by about 30% between 300 and 600 ◦ C. The thermomagnetic curves of the loess samples Li-L4 and Li-L8 (fraction 5–20 ␮m) are identical (Fig. 7a). Paramagnetic behaviour dominates from −196 ◦ C to room temperature. The inflection between 350 and 500 ◦ C may indicate destruction of a thermally unstable phase, possibly maghemite. The

267

20

50 µm

Tc = 610 ˚C 0

100 200 300 400 500 600 700 Temperature [˚C]

1

Li–S7, 1 2

0.8 0.6

Tc = 625 ˚C

0.4 0.2 0 -200 -100

0.2

0

2 µm

100 200 300 400 500 600 700 Temperature [˚C]

Fig. 7. Thermomagnetic curves of loesses (upper panel) measured in a field of 160 mT and of palaeosols (lower panel) measured in a field of 50 mT. The numbers 1 and 2 denote first and second heating runs, respectively. (a) Grain size fraction 5–20 ␮m of pristine loess Li-L4 and weathered loess Li-L8. Both samples behave similarly. The strong inflection between 350 and 500 ◦ C may indicate destruction of a maghemite phase and inversion to hematite. The pronounced paramagnetic behaviour during the second heating confirms the destruction of ferrimagnetic minerals. (c) Loess grain size fraction 20–50 ␮m. A maghemite phase with a Curie point near 610 ◦ C is present in both loess fractions and after both heating runs. (b) The grain size fraction <0.2 ␮m of palaeosol Li-S1 is characterised by magnetite with Curie temperatures of 575 ◦ C. (d) The slightly higher Curie temperature of grain size fraction 0.2–2 ␮m of palaeosol Li-S7 indicates maghemite.

268

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

from previous maghemite contributions. This unusual reaction is due to the admixed wax, which provides a reducing carbon dioxide/monoxide atmosphere.) The susceptibility of the Li-S1 and Li-S7 grain size fractions (Table 1) generally decreases with increasing grain size except for the smallest fractions. Maximal values are observed in the fraction <0.2 ␮m of both palaeosols, minimal values in the fractions 20–50 ␮m. The susceptibility of Li-S1 fractions is always higher than in the fractions of Li-S7. A direct relation between susceptibility and grain size is not recognised in the loesses. The most prominent susceptibilities are found in fraction 2–5 ␮m of Li-L4 and in fraction 20–50 ␮m of Li-L8. In contrast to Li-L4, the susceptibility of the smallest fraction is also enhanced in Li-L8. The masses of the individual grain size fractions, however, do not contribute equally to the mass of the bulk sample. The highest susceptibility contribution is found in fraction 0.2–2 ␮m of both palaeosols whereas fraction 20–50 ␮m dominates in both loesses. The susceptibility contribution in fraction 5–20 ␮m is apparently constant in the two loess samples. Both, susceptibility and susceptibility contribution, are similar to the results of Sartori et al. (1999) for grain size fractionated Hungarian loesses. The total iron content of the different grain size fractions varies between 1.7 and 8.6% (by weight) (Table 1) and is generally higher in the smaller fractions than in the coarser ones. A direct relationship between the iron content of the grain size fractions and their susceptibility does not exist. Regarding palaeosols, increased susceptibility values correlate with increased iron contents only if the two smallest fractions are combined. This is not the case in the loesses where maximal susceptibility anticorrelates with lowest iron content (fraction 20–50 ␮m). In contrast to the iron content, the titanium content varies little from 0.16 to 0.65% (by weight). The highest concentrations are observed in the intermediate grain size fractions (2–5 and 5–20 ␮m) and the lowest values in the fractions <0.2 and >50 ␮m. No correlation with susceptibility is recognised.

5. Linear source mixing model The IRM acquisition spectra of the grain size fractions (Figs. 4 and 5) demonstrate the coexistence of

different coercivity populations as already suggested by the IRM/ARM demagnetisation spectra of the bulk samples (Fig. 3). In order to quantify the magnetisation contributions, a linear source mixing model was developed using principal component analysis. The four different coercivity populations have been characterised in those grain size fractions where they are most prominently developed (see Fig. 4a and 5a–c). Component P has been obtained from the spectra of Li-S1 (0.2–2 ␮m fraction), component D1 from Li-L4 (5–20 ␮m fraction), component D2 from Li-L8 (20–50 ␮m fraction) and component H from Li-L4 (<0.2 ␮m fraction). In order to obtain an analytical expression of the measured characteristic IRM spectra (equal to endmembers), these spectra were fitted by a linear combination of three lognormal functions (Fig. 8). Component H was fitted using only one lognormal function. The best-fitting endmembers (black lines) are nearly congruent to the observed ones (white lines) and are always within the errorband (grey bands) of the observed LCS’s. Although best-fitting and observed endmember spectra of component H are not congruent, the best-fit still runs within the errorband and is considered to be reliable and representative. In a next step, bulk IRM acquisition curves have been modelled by a linear combination of the four best-fitting endmembers (Fig. 9). All characteristics of the best-fitting endmembers (mean coercivity, dispersion, general shape) are determined by the observed spectra of the grain size fractionated samples and have been kept constant at this step. Only the remanence intensity of the endmembers was variable during the modelling procedure. The individual contribution to the total IRM was obtained by integration (Table 3). Small deviations are observed at the low coercivity part of both palaeosols, at the maximal LCS amplitude of palaeosol Li-S1 and between 300 and 600 mT for the loesses. The resulting total IRM’s are comparable in amplitude with the initial IRM’s of the bulk samples presented in Fig. 3. Component P contributes 66% of the total IRM and hence predominates clearly in palaeosol Li-S1, whereas P contributes only about 44% only in palaeosol Li-S7. Component D2 predominates with 58% in loess Li-L8 and component D1 with 52% in loess Li-L4. Component H can be considered as rather constant in all four samples. It varies between 0.35

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

Component D 2

max. amplitude at

Normalised log. IRM gradient

Normalised log. IRM gradient

Component P 1

modelled 32 mT

measured 31 mT

0.8 0.6 0.4 0.2 0 0

0.5

1

1.5

2

2.5

3

3.5

1

max. amplitude at modelled 79 mT measured 79 mT

0.8 0.6 0.4 0.2 0 0

0.5

1

0.8

max. amplitude at modelled 115 mT measured 113 mT

0.6 0.4 0.2 0 0

0.5

1

1.5 Log

10

2.5

2

2

2.5

3

3.5

max. amplitude at modelled 660 mT measured 610 mT

Normalised log. IRM gradient

Normalised log. IRM gradient

Component D 1

1.5

Log 10 (Field [mT])

Log 10 (Field [mT])

1

269

3

3.5

1 0.8 0.6 0.4 Component H

0.2 0 0

0.5

1

1.5 Log

(Field [mT])

10

2

2.5

3

3.5

(Field [mT])

Fig. 8. Four experimentally derived coercivity components have been chosen as endmembers for a linear mixing model for loess and palaeosol samples to model the magnetisation of a bulk sample. The measured endmember spectra (line with error band) have been fitted using a linear combination of three lognormal distribution functions (black line)—except component H, where one function was used—in order to get a best mathematical approximation to the observation. The measured endmember spectra have been truncated (dashed line), because it is assumed that they consist of a single component only. In all cases the best fit, obtained using the Levenberg-Marquardt (Marquardt, 1963) minimisation algorithm of “Mathematica”, is within the error of the measured spectra.

Table 3 Absolute and relative contribution of the four main remanence carriers to the bulk magnetisation of the analysed loesses and palaeosols Contribution

Horizon Li-L4

Li-L8

Absolute (mA m2 /kg) P (31 mT) D2 (79 mT) D1 (113 mT) H (610 mT) Modelled IRM at 4.5 T Stacked IRM at 4.5 T

0.24 1.58 2.36 0.35 4.53 4.55

± ± ± ± ±

0.03 0.02 0.11 0.12 0.26

Relative (%) 5.3 34.8 52.2 7.7

± ± ± ±

0.2 1.8 0.6 2.2

Li-S7

Absolute (mA m2 /kg) 0.10 2.80 1.49 0.47 4.86 4.87

± ± ± ± ±

0.03 0.03 0.18 0.13 0.34

Relative (%) 2.0 57.6 30.7 9.7

± ± ± ±

0.4 3.6 1.5 1.8

Li-S1

Absolute (mA m2 /kg) 2.51 0.75 1.99 0.49 5.74 5.79

± ± ± ± ±

0.13 0.16 0.06 0.10 0.41

Relative (%) 43.7 13.0 34.7 8.6

± ± ± ±

1.1 1.8 1.7 1.0

Absolute (mA m2 /kg) 9.09 3.05 1.28 0.39 13.81

± ± ± ± ±

0.31 0.25 0.21 0.14 0.88

Relative (%) 65.8 22.1 9.3 2.8

± ± ± ±

2.1 0.4 0.9 0.9

13.88

Component P increases with increasing pedogenesis, whereas component D1 and H are rather constant. Component D2 has also variable contributions. The errors are estimated by fitting the upper and lower error boundary (cf. Fig. 9) of the bulk spectra.

Li–L4

4 2 0 0

0.5

1 1.5 2 2.5 Log 10 (Field [mT])

3

3.5

6 Li–L8

4 2 0 0

0.5

1 1.5 2 2.5 Log10 (Field [mT])

3

3.5

Bulk spectra with error band Component P

Component D1

log. IRM gradient [mAm²/kg]

log. IRM gradient [mAm²/kg]

Loesses

6

log. IRM gradient [mAm²/kg]

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

log. IRM gradient [mAm²/kg]

270

16

Palaeosols

12

Li–S1

8 4 0 0

0.5

1 1.5 2 2.5 Log 10 (Field [mT])

3

3.5

6 Li–S7

4 2 0 0

0.5

1 1.5 2 2.5 Log 10 (Field [mT])

3

3.5

Modelled bulk spectrum Component D 2

Component H

Fig. 9. Modelling the bulk IRM spectra using a linear combination of the four best-fitting endmembers (black line) P, D1 , D2 and H discussed in Fig. 8. The white curve with the grey errorband represents the bulk spectrum of the sample. The loess samples Li-L4 and Li-L8 can be well represented by the model and the best fitting linear combination is still within the error band of the measured spectra. The best fit deviates slightly between 300 and 600 mT in the loesses but also in the low coercivity part of the palaeosols, where possibly a fifth component with very low coercivity could be placed. The area below each component represents its individual contribution to the bulk magnetisation of the sample and was obtained by integration (compare Table 3).

and 0.49 mAm2 /kg. Component D1 ranges from 1.28 to 2.36 mAm2 /kg is more constant than component D2 which varies between 0.75 and 3.05 mAm2 /kg. Component P varies strongly from 0.10 to 9.09 mAm2 /kg (Table 3).

6. Interpretation ARM and IRM respond to different magnetic domain stages. In general, ARM focuses mainly on single domain (SD) ferrimagnetic grains (Dunlop and Özdemir, 1997; Egli and Lowrie, 2002), whereas IRM magnetises also pseudosingle domain (PSD) and multidomain (MD) remanence carriers. This explains the blurred appearance of the maxima in the IRM spectra of the low-intensity loesses. The observed LCS max-

ima of ARM and IRM demagnetisation curves are comparable (cf. Fig. 3) as both magnetisations have been demagnetised by ac fields. Component P is best documented in all experiments. It peaks in the ARM demagnetisation curves at 22 mT, in the IRM demagnetisation spectra at 20 mT and in the IRM acquisition spectra at 31 mT (Fig. 10). Our experimental value of 31 mT agrees with the findings of Eyre (1996), who used theoretical coercivity distribution functions for modelling IRM acquisition spectra. He reports a mean coercivity for his theoretical component 2 of (35±8) mT for 41 analysed loess/palaeosol samples. These values are close to Maher’s (1988) results for synthetic single domain magnetite. She found values of 21–29 mT using IRM acquisition. The thermal demagnetisation of IRM and the Curie curves suggest the presence of slightly oxidised

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

D2 P

(79 ± 10) mT altered detrital magnetite

(31 ± 4) mT pedogenic maghemite

Normalised log. IRM gradient

D1 (113 ± 13) mT detrital maghemite

1 0.8

H (610 ± 150) mT detrital hematite

0.6 0.4 0.2 0 0

1

2

3

Log 10 (Field [mT]) measured spectra with error band Fig. 10. A possible mineralogical interpretation of the four main experimentally derived coercivity populations (endmembers) coexisting in loess/palaeosol samples. The high coercivity component H is due to the presence of hematite. Its coercivity of 610 mT suggests a small grain size with an upper limit of 2 ␮m (see text). Because the contribution varies very little between samples (Table 3), its origin is most probably detrital. Component D1 is also considered to be of detrital origin because it varies within fairly narrow limits. The coercivity at 113 mT would correspond to maghemite of higher oxidation stage. Component D2 has a rather variable contribution with a significantly lower coercivity than D2 and is interpreted as altered detrital magnetite. The content of component P with a coercivity of 31 mT varies largely. It is of pedogenic origin and represented by magnetite or maghemite in the single domain state.

magnetite forming component P. Magnetite with a grain size of about 50–60 nm is the main ARM carrier (Dunlop and Özdemir, 1997; Egli and Lowrie, 2002). Chemically grown grains of biogenic or abiotic origin, which may have formed during pedogenesis, have such small grain sizes. A biogenic intracellular origin is excluded because the ARMdc=0.1 mT /SIRM ratio of palaeosol Li-S1 which is predominated by the pedogenic component P is about 0.06. This is much lower than the values reported by Moskowitz et al. (1993) for intracellular magnetite (intact magnetosomes) (ARMdc=0.1 mT /SIRM = 0.15–0.25). Magnetotactic bacteria need specific oxygenation conditions to grow magnetosomes, which probably did not prevail in the “dry” palaeosols on the CLP. Favourable conditions would rather be met in stagnant and/or poorly drained

271

soils with varying water level (Fassbinder et al., 1990) and oxygen-poor water. Redoximorphic features, such as Fe–Mn films, concretions and mottling have not been observed in palaeosol S1 and only rarely in palaeosol S7 (Ding, 1999). Since the spectra of Li-S1 and Li-S7 are almost equal (see Fig. 4b), the formation of magnetosomes has been excluded. Maher and Thompson (1995) show evidence for magnetic particles in palaeosols, which are similar to magnetosomes and suggest that they play a minor role in the magnetic enhancement of palaeosols. This does not imply that biologically mediated iron oxides do not exist generally in loess/palaeosols. Chukhrov et al. (1976) found that Fe3+ -oxide formed by bacterial activity in soils is usually ferrihydrite. According to Schwertmann (1988b), the inorganic formation of magnetite from mixed, possibly complexed Fe2+ /Fe3+ solutions at room temperature under neutral or alkaline pH is possible via (proto-)ferrihydrite or other trivalent Fe oxides such as lepidocrocite (Schwertmann and Taylor, 1973). We suggest production of the pedogenic component in the following way: the weathering of detrital parent material provides a solution of bi- and trivalent iron from which a nonmagnetic protocompound of trivalent iron is built up. A catalyst in the form of a magnetite nucleation germ is needed to force magnetite crystallisation and growth under these conditions. Eventually chemical changes of the soil solution may lead to the development of a diffuse fringe of protocompounds such as ferrihydrites surrounding the growing magnetite grain (Dearing et al., 1996). Pure pedogenic magnetite, however, is difficult to verify in soils (Schwertmann, 1988b), because it easily undergoes partial oxidation. Hence, we assign pedogenic maghemite of low oxidation state as carrier of component P. High unblocking temperatures near 685 ◦ C during thermal demagnetisation of IRM (Fig. 6c), high coercivity and low magnetisation indicate the presence of hematite as carrier of component H (Fig. 10). The grain size is below 5 ␮m according to coercivity data of Dankers (1981). Dunlop’s (1981) values suggest a grain size of 0.2–1.0 ␮m within the single domain range of equidimensional hematite. The remanence contribution of component H is quite constant in all samples including pristine loess Li-L4. Eyre (1996) interpreted his fourth coercivity component in a

272

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

similar manner. His theoretically modelled coercivity is a little bit lower (456 ± 100) mT but fits within the limits of standard deviation. We favour a detrital origin and suggest that component H is due to fine-grained single domain, detrital hematite with an upper grain size limit of about 2 ␮m. The reddish colour of the palaeosols is caused by the in situ growth of hematite grains (Chen et al., 2002). Their size ranges between 0.01 and 0.1 ␮m (Schwertmann, 1988a; Scheinost and Schwertmann, 1999) and is close to superparamagnetic grain size. Trivalent iron may originate from hydrolitic and oxidative decomposition of lithogenic bivalent iron silicates, such as fayalite (Schwertmann, 1988b). Depending on pedoenvironmental conditions such as temperature, rainfall and organic matter, the trivalent iron can either form goethite by crystallisation or ferrihydrite via deprotonisation and hematite via further dehydration and structural rearrangement (Schwertmann, 1988b). The red palaeosol colouration due to pedogenic hematite (Scheinost and Schwertmann, 1999; Chen et al., 2002) does not contribute to IRM and ARM. This is in agreement with the finding of Vandenberghe et al. (1998) using Mössbauer spectra of loess/palaeosols samples from the central CLP. Component D1 is observed in the IRM demagnetisation spectra of palaeosols and loesses, but not in the ARM spectra. The analysis of the IRM acquisition spectra indicates prominent coercivities around 113 mT and is very similar to Eyre’s (1996) theoretical component 3 with 115 ± 5 mT (Fig. 10). This relatively high coercivity, the thermal unblocking of IRM and Curie temperatures around 610 ◦ C (deBoer and Dekkers, 1996) point to the presence of maghemite in the loess grain size fraction 5–20 ␮m, in which component D1 predominates. We interpret component D1 as being due to detrital maghemite because it dominates the bulk magnetisation of the pristine loess Li-L4 with 52% and contributes rather constantly in the four samples investigated. In addition, susceptibility contribution (Table 1) and initial IRM (Fig. 6b) are also constant. This is in agreement with earlier findings (Heller and Liu, 1984; Evans and Heller, 1994) which suggested the presence of detrital maghemite, being resistant against weathering. A similar maghemite phase with Curie temperatures of 610 ◦ C is found in the loess grain size fraction 20–50 ␮m. Magnetite is also present, because the

thermomagnetic curve is not inflected between 300 and 500 ◦ C. Component D2 , with LCS maximum at 79 mT, dominates the grain size fraction 20–50 ␮m in the weathered Li-L8 but occurs in all samples in highly variable amounts. D2 is interpreted as an altered magnetite or as a detrital magnetite with a weathered crust of maghemite (cf. van Velzen and Dekkers, 1999). The linear mixing models of the palaeosols deviate slightly from the measured curves in the very low coercivity part (Fig. 9). A fifth component (L) may be present with a peak between 13 mT (Fig. 3f) and around 21 mT (Fig. 4a). This component could correspond to Eyre’s component 1 peaking at (11 ± 4) mT. Component L would be attributed to magnetite grains (Curie temperature of 575 ◦ C, see Fig. 7b) below the stable single domain size, but large enough to carry stable remanences for the duration of the experiments. It is probably of pedogenic origin because it is much more pronounced in palaeosols than in loesses (Fig. 5a–d). Finally, the remanence contributions of the individual components might be compared cautiously with the total iron content of the grain size fractions. The pedogenic component P, which is strongly developed in the palaeosol grain size fraction 0.2–2 ␮m, goes with high iron content (Fig. 11a). The iron content is also high in this grain size fraction of the loesses, but P has not been developed yet. Hence, iron is exchanged during pedogenesis between detrital non-magnetic ferruginous minerals and new pedogenic iron oxides. The cross iron content remains unchanged. Clay mineral assemblages of Chinese loess sediments are dominated by iron-bearing clays such as illite and chlorite (Liu, 1985). During pedogenesis the clays may release iron so that nanometer-sized ferromagnetic minerals start crystallising between clay minerals and the susceptibility of the clay sized fraction increases. This has been stated recently by Ji et al. (2002) who observed nanometric magnetite, hematite and goethite crystal coatings on clay minerals. Component D1 correlates with the relatively high titanium content of the 5–20 ␮m fraction (Fig. 11b). The chemically immobile titanium is of detrital origin and this supports the arguments for a detrital origin of D1 itself. The titanium may reside in rutile or augite rather than in titanomaghemite for which no experimental evidence was found.

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

IRM contribution [mAm²/kg]

on the observed coercivity spectra of certain grain size fractions was used to model bulk IRM acquisition coercivity spectra of loess/palaeosol samples. The following conclusions can be drawn:

pedogenic

10 8 6 4 2 0 1

2

(a) Fraction 0.2 - 2 µm contributing to component P

3

4 5 6 Total iron [wt%]

Fraction 5 - 20 µm contributing to component D1

7

Li-L4

Li-L4

Li-L8

Li-L8

Li-L8

Li-S7

Li-S7

Li-S7

Li-S1

Li-S1

Li-S1

IRM contribution [mAm²/kg]

8

Fraction 20 - 50 µm contributing to component D2

Li-L4

10 8 6 4

pure detrital

2 0 0.36 0.38 0.4 0.42 0.44 0.46 0.48 0.5 0.52 0.54

(b)

273

Total titanium [wt%]

Fig. 11. Correlation of the iron (a) and titanium (b) content (wt.%) of grain size fractions 0.2–2, 5–20 and 20–50 ␮m with the IRM contribution of components P, D1 and D2 . The iron content is high and constant in fraction 0.2–2 ␮m in all samples, but only the palaeosols (Li-S1, Li-S7) have an enhanced IRM. The slightly increased titanium values of grain size fraction 5–20 ␮m are related to remanence of the detrital component D1 in all samples.

7. Conclusions Different magnetic methods in combination with grain size fractionation identify the coexistence of four coercivity populations in loess/palaeosol samples from Lingtai (Central Chinese Loess Plateau). The two-component model by Evans and Heller (1994) has been extended to four components similar to the model of Eyre (1996). In general, there is a good agreement between our work and Eyre’s (1996) results who uses the lognormal approach. Using the argument of the linear additivity of remanences a mixing model based

• Each loess/palaeosol sample contains at least four different coercivity populations (P, D1 , D2 and H), which are able to carry a stable remanence. The minerals magnetite, maghemite of low and high oxidation state and hematite have been identified as remanence carriers. • The contribution of the different coercivity components varies with lithology and has been quantified. A constant detrital hematite component is present in all samples with contributions up to 10%. Also, the detrital intermediate coercivity component D1 is rather constant, in absolute terms. It is most important in pristine loess contributing 52% of the remanence. The intermediate coercivity component D2 contributes rather variably, but gives a major contribution in weathered loesses (up to 58%). The pedogenic low-coercivity component P dominates the palaeosols with variable contributions of up to 66%. • The detrital remanence carriers H and D1 contribute rather constantly to the IRM in loesses and palaeosols. They have not been destroyed during weathering. Hence, the iron for the new pedogenic (remanence carrying) component P in palaeosols and weathered loesses must derive from clays and other ferruginous silicates. Currently, there is no clear interpretation for component D2 . It is most probably a weathering indicator consisting of partly oxidised detrital magnetite. • The overall iron content of the clay-sized fraction does not change during pedogenesis but the ferromagnetic mineral content increases strongly. The pedogenic iron oxides grow within this grain size fraction on translocated clays. The iron brought into solution by pedogenesis apparently re-precipitates after short migration. • The pedogenic component P increases with the degree of pedogenesis, similarly as the bulk susceptibility (Tables 1 and 3). Hence component P indicates climate variations, both in time and space, which cause variable in situ alteration of the loess material mainly by differing palaeoprecipitation (cf. Heller et al., 1993; Maher and Thompson, 1995). In contrast, variations of component D1 might reflect

274

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275

variable aeolian sedimentation conditions caused by different wind regimes.

Acknowledgements We thank Kurt Barmettler for grain size separation and X-ray diffraction measurements. Special thanks go to Ramon Egli for constructive discussions and for providing his coercivity spectra calculation program (Codica 2.3). This study was financed by Swiss National Science foundation grant No 21-54143.98.

References Carter-Stiglitz, B., Moskowitz, B., Jackson, M., 2001. Unmixing magnetic assemblages and the magnetic behaviour of bimodal mixtures. J. Geophys. Res. 106, 26397–26411. Chen, J., Ji, J.F., Balsam, W., Chen, Y., Liu, L.W., An, Z.S., 2002. Characterisation of the Chinese loess-paleosol stratigraphy by whiteness measurement. Palaeogeogr. Palaeoclimatol. Palaeoecol. 183, 287–297. Chukhrov, F.V., Zvyagin, B.B., Ermilova, L.P., Gorshkov, A.I., 1976. New data on iron oxides in the weathering zone. Mineralog. Deposita 11, 24–32. Dankers, P.H.M., 1981. Relationship between median destructive field and remanent coercive forces for dispersed magnetite, titanomagnetite and hematite. Geophys. J. R. Astron. Soc. 64, 447–461. deBoer, C., Dekkers, M.J., 1996. Grain size dependence of the rock magnetic properties for a natural maghemite. Geophys. Res. Lett. 23, 2815–2818. Dearing, J.A., Hay, K.L., Baban, S.M.J., Huddleston, A.S., Wellington, E.M.H., Loveland, P.J., 1996. Magnetic susceptibility of soil, an evaluation of conflicting theories using a national data set. Geophys. J. Int. 127, 728–734. Ding, Z.L., 1999. Pedostratigraphy and paleomagnetism of a ∼7.0 Ma eolian loess-red clay sequence at Lingtai, loess plateau, north-central China and the implications for paleomonsoon evolution. Palaeogeogr. Palaeoclimatol. Palaeoecol. 152, 49–66. Ding, Z.L., Sun, J.M., Yang, S.L., Liu, T.S., 1998. Preliminary magnetostratigraphy of an aeolian red clay-loess sequence at Lingtai, the Chinese Loess Plateau. Geophys. Res. Lett. 25, 1225–1228. Dunlop, D.J., 1981. The rock magnetism of fine particles. Phys. Earth Planetary Interiors 26, 1–26. Dunlop, J.D., Özdemir, Ö., 1997. Rock Magnetism, Fundamentals and Frontiers. Cambridge University Press, Cambridge, 573 pp. Egli, R., 2003a. Analysis of the field dependence of remanent magnetization curves. J. Geophys. Res. 108 (B2), 2081, doi:10.1029/2002JB002023. Egli, R., 2003b. Characterisation of individual rock magnetic components by analysis of remanence curves. 2. Rock

magnetism of individual components. Phys. Earth Planet. Interiors, submitted for publication. Egli, R., Lowrie, W., 2002. Anhysteretic remanent magnetization of fine magnetic particles. J. Geophys. Res. 107 (B10), 2209, doi:10.1029/2001JB000671. Evans, M.E., Heller, F., 1994. Magnetic enhancement and paleoclimatic: study of a loess/paleosol couplet across the loess plateau of China. Geophys. J. Int. 117, 257–264. Eyre, J.K., 1996. The application of high resolution IRM acquisition to the discrimination of remanence carriers in Chinese loess. Studia geophysica et geodetica 40, 234–242. Fassbinder, J.W.E., Stanjek, H., Vali, H., 1990. Occurrence of magnetic bacteria in soil. Nature 343, 161–163. Heller, F., Liu, T.S., 1984. Magnetism of Chinese loess deposits. Geophys. J. R. Astron. Soc. 77, 125–141. Heller, F., Liu, T.S., 1986. Paleoclimatic and sedimentary history from magnetic susceptibility of loess in China. Geophys. Res. Lett. 13, 1169–1172. Heller, F., Evans, M.E., 1995. Loess magnetism. Rev. Geophys. 33, 211–240. Heller, F., Chen, C.D., Beer, J., Liu, X.M., Liu, T.S., Bronger, A., Suter, M., Bonani, G., 1993. Quantitative estimates of pedogenic ferromagnetic mineral formation in Chinese loess and paleoclimatic implications. Earth Planet. Sci. Lett. 114, 385–390. Heslop, D., Langereis, C.G., Dekkers, M.J., 2000. A new astronomical time scale for the loess deposits of Northern China. Earth Planet. Sci. Lett. 184, 125–139. Heslop, D., Dekkers, M.J., Kruiver, P.P., van Oorschot, I.H.M., 2002. Analysis of isothermal remanent magnetisation acquisition curves using the expectation-maximisation algorithm. Geophys. J. Int. 148, 58–64. Heslop, D., McIntosh, G., Dekkers, M.J., 2003. Using time and temperature dependant Preisach models to investigate the limitations of modelling isothermal remanent magnetisation acquisition curves with cumulative log Gaussian functions. Geophys. J. Int., submitted for publication. Ji, J.F., Chen, J., Xu, H.F., Chen, T.S., 2002. Chemical weathering of chlorite in Chinese loess-paleosol stratigraphy and climate change. Conference Abstracts, GSA, Denver Annual Meeting. Klute, A. (Ed.), 1986. Methods of Soil Analysis: Physical and Mineralogical Methods, Series in Agronomy, Number 9 (Part 1), second ed., American Society of Agronomy, Soil Science Society of America, Madison WI, USA. Kruiver, P.P., Dekkers, M.J., Heslop, D., 2001. Quantification of magnetic coercivity components by analysis of acquisition curves of isothermal remanent magnetisation. Earth Planet. Sci. Lett. 189, 269–276. Liu, T.S. (Ed.), 1985. Loess and the Environment. China Ocean Press, Beijing, 251 pp. Maher, B.A., 1988. Magnetic properties of some synthetic sub-micron magnetites. Geophys. J. 94, 83–96. Maher, B.A., Thompson, R., 1995. Paleorainfall reconstructions from pedogenic magnetic susceptibility variations in the Chinese loess and paleosols. Quat. Res. 44, 383–391. Marquardt, D.W., 1963. An algorithm for least-squares estimation of non-linear parameters. J. Soc. Ind. Appl. Math. 11, 431–441.

S. Spassov et al. / Physics of the Earth and Planetary Interiors 140 (2003) 255–275 Moskowitz, B.M., Frankel, R.B., Bazylinski, D.A., 1993. Rock magnetic criteria for the detection of biogenic magnetite. Earth Planet. Sci. Lett. 120, 283–300. Robertson, D.J., France, D.E., 1994. Discrimination of remanencecarrying minerals in mixtures using isothermal remanent magnetisation acquisition curves. Phys. Earth Planet. Interiors 82, 223–234. Sartori, M., Heller, F., Forster, T., Borkovec, M., Hammann, J., Vincent, E., 1999. Magnetic properties of loess grain size fractions from the section at Paks (Hungary). Phys. Earth Planet. Interiors 116, 53–64. Scheinost, A.C., Schwertmann, U., 1999. Color identification of iron oxides and hydroxysulfates: use and limitations. Soil Sci. Soc. Am. J. 63, 1463–1471. Scheinost, A.C., Kretzschmar, R., Pfister, S., Roberts, D.R., 2002. Combining selective sequential extractions, X-ray absorption spectroscopy, and principal component analysis for quantitative zinc speciation in soil. Environ. Sci. Technol. 36, 5021–5028. Schwertmann, U., 1988a. Some properties of soil and synthetic iron oxides. In: Stucki, J.W., Goodman, B.A., Schwertmann, U. (Eds.), Iron in Soils and Clay Minerals. D. Reidel Publishing, Dordrecht, Boston, Lancaster, Tokyo, pp. 203–250. Schwertmann, U., 1988b. Occurrence and formation of iron oxides in various pedoenvironments. In: Stucki, J.W., Goodman, B.A., Schwertmann, U. (Eds.), Iron in Soils and Clay Minerals.

275

D. Reidel Publishing, Dordrecht, Boston, Lancaster, Tokyo, pp. 267–308. Schwertmann, U., Taylor, R.M., 1973. The in vitro transformation of soil lepidocrocite to goethite, pseudogley and gley. In: Transaction Communications of the Vth and VIth International Society of Soil Sciences, Stuttgart, Hohenheim, pp. 45–54. Spassov, S., Heller, F., Evans, M.E., Yue, L.P., Ding, Z.L., 2001. The Matuyama/Brunhes geomagnetic polarity transition at Lingtai and Baoji, Chinese loess plateau. Phys. Chem. Earth (A) 26, 899–904. Stockhausen, H., 1988. Some new aspects for the modelling of isothermal remanent magnetization acquisition curves by cumulative log Gaussian functions. Geophys. Res. Lett. 25, 2217–2220. Vandenberghe, R.E., Hus, J.J., De Grave, E., 1998. Evidence from Mössbauer spectroscopy of neo-formation of magnetite/ maghemite in the soils of loess/paleosol sequences in China. Hyperfine Interactions 117, 359–369. van Velzen, A., Dekkers, M.J., 1999. Low-temperature oxidation of magnetite in loess-paleosol sequences: a correction of rock magnetic parameters. Studia geophysica et geodaetica 43, 357– 375. Wang, Y., Evans, M.E., Rutter, N., Ding, Z.L., 1990. Magnetic susceptibility of Chinese loess and its bearing on paleoclimate. Geophys. Res. Lett. 17, 2449–2451.