Magnetic Component Unmixing of a Lacustrine Sedimentary Drill Core from Heqing Basin

Abstract

Long and continuous lacustrine sediments in Southwest China provide exceptional records of the Indian summer monsoon (ISM) evolution. Rock magnetic and environmental magnetic methods have significant roles in these lacustrine studies. However, lacustrine sedimentary environments are complex and magnetic mineral signatures can be altered by post-depositional processes. This study applies isothermal remanent magnetization (IRM) component unmixing methods to lacustrine sediments from the Heqing core, to identify and quantify magnetic mineral components. We analyzed 104 samples based on lithological variations and magnetic susceptibility (χ) to examine the composition of magnetic minerals and their relative contributions. Three distinct magnetic components were identified in IRM component unmixing results: a low-coercivity detrital component, a medium-coercivity authigenic component, and a hard magnetic component. Based on rock magnetic results, the medium-coercivity component was attributed to greigite. These components exhibit stratigraphic trends that reflect changes in paleoenvironmental conditions. The medium-coercivity component shows an upwards decrease, indicating a significant change in ISM science at about 1.8 Ma. The study highlights the importance of considering post-depositional processes when interpreting magnetic mineral signatures in lacustrine sediments. The CLG model, combined with conventional rock magnetic analyses, provides a rapid approach for characterizing magnetic assemblages in weakly magnetic sediments.

1. Introduction

The Indian monsoon is a pivotal component of the tropical/subtropical climate system [1,2,3], characterized by its seasonal reversal of winds driving intense summer rainfall across South and East Asia. A thorough understanding of its complex dynamics and long-term evolutionary history is essential, not only for regional water resource management and agricultural planning but also for accurate future global climate predictions under anthropogenic forcing scenarios [1,2,4]. Southwest (SW) China acts as the primary conduit for heat and moisture transport via low-level jet streams from the tropical/subtropical Indian Ocean deep into continental China [1,4,5,6,7]. Previous studies have demonstrated that lacustrine sedimentary sequences in this region, due to their high sedimentation rates, continuity, and sensitivity to hydrological changes, provide exceptional, high-resolution records of the ISM variability [1,3,4,5,6,7]. During these studies, rock magnetic and environmental magnetic methodologies have assumed increasingly significant roles due to their rapidity, sensitivity, and cost-effectiveness [4,6,7,8,9]. High magnetic susceptibility was correlated to weak ISM during both “glacial-interglacial” [8] and “stadial-interstadial” [6] climate cycles. Hu et al. [7] combined the carbonate content, ARM/SIRM, and S-ratio into the weathering intensity index and reconstructed the ISM variations of the past 920 ka. Zhang et al. [4] used hematite/(hematite and goethite) from Tengchongqinghai Lake to reconstruct adequate moisture variations over the past 90 ka. However, lacustrine sedimentary environments are highly complex, and the factors influencing the magnetic properties of lake sediments are multifaceted. During deposition, the magnetic properties of sediments are closely linked to the magnetic characteristics of the terrigenous detritus [10,11,12,13,14,15], as well as the input and process of exogenic detrital iron oxides [6,7,8,16,17]. Post-depositionally, reductive diagenetic processes within the lake can dramatically alter detrital magnetic minerals, obscuring their recorded magnetic signatures [13,15,18,19]. Generally speaking, magnetic iron oxides (e.g., magnetite, maghemite and hematite) are primarily clastic in origin [10,12,13,14,15], while magnetic iron sulfides (greigite and pyrrhotite) are authigenic minerals [9,20]. Therefore, the concentration of low coercivity magnetite was used as indicator for hydrodynamic or wind force intensity in environmental magnetic research [6,7,8,14,15]. On the contrary, the magnetic property of magnetic iron sulfides was used to indicate changes in the oxidation–reduction environment [9,13,20]. Nevertheless, detrial originated greigite and authigenic biological magnetite was also reported in marine and lake sediments [9,20,21,22,23]. This leads to ambiguity in the interpretation of magnetic parameters [10,12,13,15,16,17]. Consequently, the separation and quantification of magnetic components have become fundamental tasks in the environmental magnetic study of lake sediments [13,14,24,25,26,27,28,29,30]. In this paper, we used the coercivity spectrum unmixing method to semi-quantitatively characterize variations in the abundance of different magnetic components, and discuss the linkage between paleoenvironmental variation and different magnetic components.

2. Materials and Methods

The Heqing Basin is an N–S-oriented fault-bounded basin at the southeastern margin of the Tibetan Plateau, in southwest China (Figure 1). The local climate is dominated by the Indian Monsoon, which is characterized by warm wet summers and cool dry winters. The catchment bedrock of Heqing Basin consists of Triassic limestone in the west, and Paleogene calcareous conglomerate, sandstone, and silty shale in the east [1,7]. The Heqing drill core (26°33′43.1″ N, 100°10′14.2″ E) was a lacustrine sedimentary core with a bottom depth of 665.83 m [1,5]. Based on tectonic/climatic-sedimentary cycles, we divided the core into three stages. In each cycle, the lithology changes from gravel or clay-sand with gravel at the base to silty clay and clay. Stage 1 (665.8–372.5 m) consists of dark grayish-green calcareous clays and silty clays with thin silt and fine sand layers; Stage 2 (372.5–195.5 m) consists of grayish-green calcareous clays and silty clays with clear bedding, intercalated with light gray silt and fine sand layers; and Stage 3 (195.5–0 m) consists of grayish-green and grayish-yellow calcareous clays and silty clays with clear bedding. The chronology of the HQ drill core was established by magnetic stratigraphy, AMS ¹⁴C dating, and astronomical tuning [1,3,5]. Magnetic stratigraphy employed the thermal demagnetization of 454 U-channels and 1666 discrete samples. Its polarity sequence, correlated with the Geomagnetic Polarity Time Scale, placed the Brunhes/Matuyama (0.78 Ma) and Matuyama/Gauss (2.58 Ma) boundaries at depths of 152.0 m and 614.47 m, respectively [1]. The polarity ending at the first Gauss normal subchron indicated a bottom age of 2.78 Ma [1].

Based on the down-core variations in lithology and magnetic susceptibility (χ), 104 samples were systematically selected from the peaks and troughs of χ across different lithological units for the IRM acquisition curve, magnetic susceptibility versus temperature (χ-T), and FORC diagrams measurements. All samples were clay and silt clay, dried at 25 °C for 48 h in an oven, and then ground into powder. All samples were used for IRM acquisition curves and 62 among them were used for FORC diagram measurements. Approximately 0.15 g dry powder was loaded into a non-magnetic capsule and then measured in a Model MicroMag 3900 vibrating sample magnetometer (VSM) (Princeton Measurements Corporation, Princeton, NJ, USA) for IRM acquisition curve and FORC diagram measurements. Among 104 samples, 57 were selected for χ–T curve measurements. Between 0.1 and 0.2 g of dry powder was loaded into the quartz furnace of the MFK1-FA Kappabridge system (with CS-3 high-temperature furnace). The temperature range was 40–700 °C with a heating rate of 11 °C per minute. The samples were heated and cooled in an argon atmosphere to minimize the oxidation of sedimentary components. Low-temperature measurements were made using a MFK1-FA Kappabridge with a CS-L low-temperature furnace (AGICO, s.r.o., Brno, Czech Republic). The minimum temperature used was −194 °C. Among 104 samples, 10 were selected for Scanning Electron Microscopy (SEM) analysis. Approximately 0.05 g of dry powder was used for the FEI Quanta 400 FEG thermal field emission environmental scanning electron microscope.

For an assemblage of grains comprising a single magnetic mineral with no magnetic interactions, characterization relies on the saturation isothermal remanent magnetization (SIRM), the coercivity at half SIRM, and the coercivity discrete degree (DP), defined as one standard deviation of the logarithmic distribution [24,25,27]. When multiple magnetic minerals are present, the isothermal remanent magnetization (IRM) acquisition curve for each component follows a cumulative log-Gaussian (CLG) function [24,25,26,27,28,29]. Consequently, a measured composite IRM curve can be decomposed into multiple CLG components. The analysis of these curves thus enables the discrimination of magnetic mineralogy and the quantification of relative mineral abundances within a sample [24,25,26,27,28,29].

In this study, component separation using the CLG model was performed with the irmunmix2_2 software [28]. IRM data were imported into the program, and automatic separation was conducted at a 90% confidence level. This yielded two plausible component assemblages (Figure 2): a three-component solution and a four-component solution. As shown in Figure 2, linear acquisition curve plots (LAPs), gradient acquisition curve plots (GAPs), and standardized probability acquisition curve plots (SAPs) were used to assist in curve fitting [24,27,28]. Within this software framework, smaller variance values indicate a superior fitting performance and results that align more closely with the original measurements. The comparative analysis of the variance metrics for representative samples (

Table 1. Residual comparison of the four-component (Fit 1) and three-component (Fit 2) solution.

Fit 1	LAP	GAP	SAP	Fit 2	LAP
N	99	99	99	N	99
S	4.77 × 10−16	1.87 × 10−13	0.163927	S	4.91 × 10−16
CCS	3.08 × 10−28	1.98 × 10−23	8.783283	CCS	3.58 × 10−28
S2	3.14 × 10−30	2.02 × 10−25	0.090549	S2	3.66 × 10−30

N is the number of data. S is the mean square deviation of data. CCS is the sum of squares of the variance and S² is the dispersion degree of the variance.

) revealed that the four-component solution exhibited lower variance and was therefore deemed more robust. Consequently, the four-component solution was adopted for the unmixing of the IRM acquisition curve throughout this investigation. However, the relative contribution of the first component in the four-component assemblage was extremely low. For 104 samples analyzed, excluding a few outliers with noisy acquisition curves, this component consistently represented <1% of the total magnetization. It exhibited high dispersion (DP > 0.25) and the coercivity at half SIRM was uniformly below 1 mT. This component is interpreted not as a physical signal, but as a mathematical artifact arising from insufficient low-field data resolution following logarithmic transformation. Consequently, only the latter three components were utilized in subsequent analyses.

3. Results

3.1. Magnetic Mineralogy

A previous magnetic study of sediments from the Heqing drill core showed that χ exhibits a positive correlation with magnetic mineral concentration parameters such as SIRM and ARM [19]. This indicated that χ was firstly dominated by a concentration of magnetic mineral. Temperature-dependent χ is highly sensitive to mineralogical changes during thermal treatment, and such changes can provide information about magnetic mineral composition [31]. The χ–T curves of representative samples reveal a magnetic susceptibility drop to zero at approximately 580 °C (Figure 3), indicating the Curie temperature of magnetite. Beyond 400 °C, χ increases rapidly, which was commonly observed in marine and lake sediments, and likely due to the thermal decomposition of iron-bearing clay minerals and silicates [11,13,31]. In stages 2 and 3, high χ samples exhibit distinct peaks at ~330 °C followed by gradual decline until 370 °C. This feature primarily reflects the transformation of ferrimagnetic maghemite to antiferromagnetic hematite [31,32,33]. Conversely, low χ samples display sharp increases above 400 °C during heating. This characteristic is commonly observed in samples from reducing environments [9,13,15,20]. In the Ar gas, oxygen could have resulted from the thermal decomposition of iron-bearing clay minerals. The thermal transformation of maghemite into hematite mainly caused the χ decrease. Therefore, the increase in χ was not caused by maghemite, and is closely related to the oxidative decomposition of pyrite [9,13,15,20]. The χ–T curves of samples from stage 1 resemble those of samples with a lower χ in stages 2 and 3.

Samples from stages 2 and 3 of the core with a higher χ (Figure 4a) display strong PSD and SP signals, with peripheral curves resembling MD characteristics [9,11,20,22]. The magnetic domains of these samples are predominantly pseudo-single domains, containing a significant amount of superparamagnetic particles. In contrast, samples from stages 2 and 3 of the core with a lower χ (Figure 4b) exhibit very weak FORC diagram signals. Even with the smoothing factor (SF) increased to 12, only some high-intensity areas along the B_c axis can be observed, with shapes closer to PSD characteristics. This suggests the extremely low content of magnetic minerals in the samples. Samples from stage 1 (Figure 4c) of the core display typical SD FORC diagrams, with contour lines forming concentric circular closures [9,11,20]. The coercivity value at the center is 57–62 mT, indicative of typical SD greigite [20,22].

Saturation magnetization (M_s) ranged from 0.13 to 22.81 × 10⁻⁵ A/m, with an average value of 4.06 × 10⁻⁵ A/m. Saturation isothermal remanent magnetization (M_rs) ranged from 1.27 × 10⁻⁷ to 7.52 × 10⁻⁵ A/m, with an average value of 9.99 × 10⁻⁶ A/m. Coercivity (H_c) ranged from 4.09 to 79.84 mT, with an average value of 20.91 mT. Remanent coercivity (H_cr) ranged from 14.73 to 991.32 mT, with an average value of 43.79 mT. The ratio of saturation isothermal remanent magnetization and saturation magnetization (M_rs/M_s) is plotted against the ratio of remanent coercivity and coercivity (H_cr/H_c) as the ‘Day plot’ (Figure 5) [11]. The samples from stage 1 are distributed within the single domain (SD) region and below the SD and superparamagnetic (SP) admixture line in the PSD region [11]. The samples with a high χ from stages 2 and 3 are clustered in the PSD region and are close to the SD and MD admixture line, while low χ samples are widely scattered in parallel to the SD and MD admixture line.

As shown in the scanning electron microscope (SEM) images (Figure 6), greigite particles can be observed in the samples from stage 1 of the core [9]. However, no such particles are present in the samples from stages 2 and 3 of the core [8]. As depicted in Figure 6, greigite exhibits a cluster distribution within the silicate framework [20,22,34]. The energy dispersive spectrum (EDS) images of a speculated greigite cluster in magnetic extracts (Figure 6c) showed that the atomic percentage of iron (Fe) and sulfur (S) was approximately 3/2, which was a typical characteristic of natural greigite [13,20,22]. Combined with their rock magnetic characteristics, such as the χ–T curves in Figure 3c (HQ6102 and HQ7682) and the FORC diagram in Figure 4c (HQ7682), this evidence confirmed the presence of greigite in lacustrine sediments in stage 1 of the Heqing drill core.

3.2. IRM Component Unmixing Results

Component unmixing was performed on samples located at χ peaks and troughs across the entire core. As illustrated in Figure 7, stage 3 samples display a relatively uniform combination type in the GAP diagram, distinctly dividing into three components with significant differences in coercivity. Component 1 has a coercivity of 35–50 mT and accounts for over 60% of the total content; Component 2 has a coercivity of 70–90 mT and makes up nearly 30% of the total content; Component 3 exhibits a larger variation in coercivity, mostly ranging from 130 to 200 mT, and represents less than 10% of the total content. Stage 2 samples show variations in three components in the GAP diagram. The coercivity of each component is similar to that of stage 3, but the relative abundance varies significantly and can be categorized into two scenarios. In samples with a higher χ, the content of the low-coercivity component is markedly higher than that of the medium-coercivity component. In samples with a lower χ, the low-coercivity component slightly predominates. Stage 1 samples display more complex variations in the GAP diagram, primarily divided into three cases. First, in layers with abundant sand and gravel, samples have a higher χ. Among the three components, the relative abundance of the low-coercivity component is greater than or equal to that of the medium-coercivity component. Second, in samples with a lower χ, the relative abundance of the medium-coercivity component is significantly higher than that of the low-coercivity component.

As illustrated in Figure 8, based on the magnitude of coercivity, component 1 (35–50 mT) exhibits a large degree of dispersion with a DP value exceeding 0.25, corresponding to soft magnetic minerals such as magnetite and maghemite [22,24,30,34,35,36,37]. Component 2 (70–90 mT) aligns with the coercivity of natural greigite [20,24,30] and shows minimal dispersion (DP < 0.2). It suggested that the authigenic magnetic minerals have a regular growth pattern with minimal particle size variation, likely being greigite or biogenetic single-domain magnetite [23,30,34]. Component 3 typically exceeds 150 mT and is attributed to hard magnetic minerals like hematite and goethite [30,37].

4. Discussion

IRM component unmixing identifies three distinct magnetic components: a low-coercivity detrital component, a medium-coercivity authigenic component, and a hard magnetic component. As illustrated in Figure 8, the absolute contents of all three components exhibit parallel stratigraphic trends, indicating their dependence on the total abundance of magnetic minerals [24,26,30]. However, their relative abundance exhibits marked differences (Figure 9). The detrital component was relatively low in stage 1. It gradually increases upwards, showing a significant rise at the onset of stage 2 and subsequently remaining high. The authigenic component was high in stage 1. It decreases upwards, and undergoes a significant decline at the start of stage 2, and then stabilizes at a low level. The hard magnetic component showed only minor variations, likely attributable to its low relative abundance.

Reductive diagenesis is a common process in lacustrine sediments, leading to rapid magnetic mineral dissolution and the consequent degradation of magnetic signals [13,15,18,38]. In stage 1 of the Heqing lacustrine sequence, magnetic susceptibility was extremely low and the TOC was highly stable (Figure 9). Rock magnetism indicated the presence of single-domain greigite, implying intense reductive diagenesis [9,15,20,22]. The Day plot may indicate the distribution of magnetic mineral domains [11]. In the Day plot, the intensification of reductive diagenesis is represented by a gradual shift from the stable PSD region to the stable SD region [9,15,20,22]. Similarly, stage 1 samples in the Heqing core showed a comparable trend in the Day plot (Figure 5). Analyzing the gradient acquisition curves (GAP) of the samples at different evolutionary stages reveals that, in the stable PSD region, the content of the detrital component is higher than or comparable to the authigenic component. Samples approaching the SD region exhibit significantly lower detrital component than the authigenic component. In the stable SD region, component coercivity overlaps, and the medium-coercivity component splits into 2–3 parts. During the initial interval of reductive diagenesis, only fine-grained iron oxides are decomposed [13,15,39,40], akin to samples in the PSD region of the Day plot [8,38], where the detrital component dominates.

As reductive diagenesis intensifies, fine-grained detrital magnetic minerals are exhausted [39,40], and authigenic iron sulfides begin to form [9,15,22]. This shifts the compositional balance, resulting in the higher abundance of the authigenic component relative to the detrital component [15,22]. When reductive diagenesis generates substantial iron sulfides, the magnetic domain state becomes predominantly superparamagnetic to the single-domain (SP-SD), accompanied by enhanced interparticle interactions [15,20]. These interactions not only complicate component separation but likely account for the observed subdivision of medium-coercivity components into 2–3 distinct populations [25]. The Gradient Acquisition Plot (GAP) effectively captures these diagenetically induced variations in component abundances (Figure 9).

In stages 2 and 3, the sediments contain magnetite and maghemite but lack greigite. High magnetic susceptibility (χ) and a low TOC indicate weak diagenesis [13,15]. Under these conditions, reductive dissolution preferentially targets fine-grained iron oxides before coarser fractions [13,15], with maghemite is more soluble than magnetite [13]. The coexistence of maghemite with a high χ implies only the partial dissolution of fine-grained Fe-oxides, while the Day plot positions suggest increased multidomain (MD) contributions [20]. Conversely, low-χ intervals reflect more intense diagenesis. The χ–T curves show no greigite or maghemite, suggesting complete maghemite dissolution. The Heqing core was collected in 2002, but rock magnetic measurements were mainly conducted six years later. Minimal greigite might have been formed and preserved, but was decomposed by oxide in long time conservation.

As discussed, the IRM components in the Heqing core lacustrine sediments exhibited a significant shift at approximately 372.5 m depth, corresponding to ~1.66 Ma. This change is closely linked to past sedimentary environmental conditions, which were strongly influenced by climate. Around 1.6 Ma, the Northern Hemisphere ice sheet reached a critical volume, substantially impacting the intensity of the Asian winter monsoon [3,38]. Concurrently, the forcing mechanisms of the Indian summer monsoon underwent a change around 1.8 Ma [1,3]. These combined shifts profoundly affected the climate of the Heqing Basin. Notably, the amplitude of the Indian summer monsoon index recorded in the Heqing core increased abruptly after 1.8 Ma [1,5], leading to subsequent changes in the sedimentary environment. Prior to 1.66 Ma, the monsoon amplitude was lower [1,3], resulting in a relatively stable sedimentary environment. This stability favored the formation and preservation of greigite. However, after 1.66 Ma, the enhanced monsoon amplitude [1,5] increased the environmental variability. This instability likely disrupted the conditions necessary for greigite preservation.

5. Conclusions

The IRM characteristics of lake sediments can effectively identify and quantify different magnetic mineral components, offering vital insights for paleoenvironmental reconstruction. The CLG model successfully separates three magnetic components corresponding to different magnetic minerals. Their relative abundance variations are closely linked to sedimentary environmental changes. Reductive diagenesis significantly affects the magnetic properties of sediments by reducing their magnetic mineral content and blurring magnetic signals. The medium-coercivity component shows an upwards decrease, indicating a significant change in the Indian summer monsoon science at about 1.8 Ma. The study highlights the importance of considering post-depositional processes when interpreting magnetic mineral signatures in lacustrine sediments. While the CLG model is effective for component identification, it has limitations in quantitative relative abundance analysis, especially in the sediments affected by reductive diagenesis.