The experimental sample was collected from the V sand unit of the upper member of the lower Ganchaigou Formation. The sample was washed and pulverized to 100-mesh and then divided into three equal-weight parts (series A‒C). The series A sample comprised source rock containing kerogen, and free and combinative soluble organic matter. The series B sample was extracted with chloroform for 72 h to remove free soluble organic matter; hence, this sample contained only kerogen and combinative organic matter. Series C underwent chloroform extraction and removal of combinative organic matter, so that only kerogen remained. After sample pretreatment, an ELTRA CSI analyzer and a Rock-Eval 6 pyrolysis analyzer were used to measure the total organic carbon (TOC), S₁, S₂, and T_max values. The mineral composition of the source rock sample is shown in Table 1, while the TOC and pyrolysis parameters are shown in Table 2. The high dolomite and low clay mineral contents indicate that the sample is a typical saline lacustrine source rock.

Thermal simulation experiments were conducted using an autoclave‒gold tube system. Details of the thermal simulation experimental process can be found in previous research (Xiao et al., 2005, 2006; Pan et al., 2012; Jin et al., 2013; Li et al., 2013). Briefly, in these experiments, the pressure was set as 50 MPa, and the pressure program controlled its change by ± 0.1 MPa. After the autoclave was rapidly raised from room temperature to 250°C, it was further raised to 450°C at a heating rate of 2°C/h. A sample was taken every 25°C between 250 and 450°C, and pyrolysis data of different temperature points were obtained for each series (A‒C). The pyrolysis experiments were completed simultaneously so that the experimental conditions of the series A‒C samples were utterly consistent.

After each thermal simulation experiment, each gold tube was put into a vacuum glass acquisition system connected to a chromatograph (Tian et al., 2006; Guo et al., 2009), and the gold tube was punctured. Once the gas was balanced, the valve was opened, and the gas was automatically imported into an HP5890 II chromatograph for quantitative analysis. The gas chromatograph was equipped with a fused silica column (Poraplot-Q, 30 m × 0.25 mm × 0.25 µm) with Helium as the carrier gas at a constant flow rate of 1.2 ml/min. The initial gas chromatograph (GC) oven temperature was 60°C for 6 min, followed by an increase of 15°C/min up to 180°C and maintenance for 4 min. The minimum detection limit of the system is 0.01 ml gas, and the relative error is <0.5% (Pan et al., 2012; Jin et al., 2013; Li et al., 2013).

The gas chromatograph-isotope ratio-mass spectrometer (GC-IR-MS) analysis of the gas in this study was performed on a Delta Plus^XL equipped with a GC. The GC was fitted with the same column as described above, and the GC temperature program was as follows: start temperature of 50°C, held for 3 min, followed by an increase gradient of 3°C/min to 200°C (Guo et al., 2009) and a final hold time of 30 min. Individual compounds were oxidized at 930°C for measurement of their carbon isotope ratios.

Liquid hydrocarbon products were extracted with dichloromethane, filtered with a filter membrane (pore size 0.45 μm), and weighed. Hydrocarbons were then identified using a GC/MS system (Agilent 7890, Agilent Technologies, United States) interfaced with a mass spectrometric detector (Xia et al., 2013) (MSD, Agilent 5973N). Helium was used as the carrier gas in constant (Behar et al., 1992) flow mode at 2 ml/min. The injector temperature was 280°C, and the oven temperature program was 3 min at 40°C, followed by an increase of 2°C/min up to 100°C, then 5°C/min up to 280°C, where it was maintained for 20 min. The ionization energy was 70 eV.

Hydrocarbon gas yields and stable carbon isotope compositions at different simulation temperatures are shown in Figure 1 and Table 3. When the simulated temperature was lower than 375°C, the gaseous hydrocarbon yields of the three series were similar, because the samples had not wholly entered the pyrolysis gas production stage. A large number of minerals in series A affected the hydrocarbon generation of organic matter. When the experimental temperature reached 400°C, the hydrocarbon gas yield of the series A sample was considerably higher than those of series B and C. This indicates that the soluble organic matter had begun to produce hydrocarbon gases. As the simulated temperature rose, the role of soluble organic matter became more significant. For example, at 450°C, the total hydrocarbon gas (C_1‒5) yield of series A was 364 mg/gTOC, while the series B and C yields were 262.56 and 270 mg/gTOC, respectively. In addition, the yield data (Figure 1 and Table 3) also show that the methane and (Wang and Cheng, 2000) total hydrocarbon gas yields of the B series sample were similar to those of the C series. In contrast, the liquid hydrocarbon yield (C_6‒14) of the B series sample was higher than that of the C series, indicating that the contribution of bound organic matter to the hydrocarbon products is mainly reflected in the liquid hydrocarbons.

Figure 2 shows the stable carbon isotope compositions of methane (Figure 2A), ethane (Figure 2B), and propane (Figure 2C) in the three series. The isotopic composition of each series became markedly heavier with an increase in temperature. Compared with series C, the carbon isotope compositions of methane, ethane, and propane in series A were markedly enriched in ¹²C. However, for carbon dioxide (Figure 2D), the products in series C were more enriched in ¹²C than those in series A and B.

On the basis of the above experimental results, the differences in gas yield and isotopic characteristics of each series are inferred to be caused by the type and quantity of organic matter in the samples themselves, which leads to different gas production mechanisms and processes. The series A sample contained kerogen and soluble organic matter, and its hydrocarbon production per unit of organic matter was higher than those of series B and C. This is a result of the contribution of soluble organic matter in series A. The differences in carbon isotope signatures are related to the different gas production mechanisms. Gas produced in series C mainly came from kerogen cracking, primarily derived from the demethylation of C₁₄₊ nuclear polycondensation. In the series A sample, in addition to demethylation, the C‒C bond of C₆₊ saturated chains and the C‒C bond of C_2‒5 fatty chains was fractured in the cracking gas (Hill et al., 2003; Guo et al., 2009). The mean δ¹³C value of carbon dioxide in series C was −23.37‰ (Figure 2D), while series A and B samples exhibited positive δ¹³C values (Figure 2D) caused by the thermal cracking of inorganic minerals in the pyrolysis process. We can conclude that the abundant soluble organic matter in the source rocks of this saline lake basin has a significant influence on hydrocarbon generation, both in terms of gas generation potential and controlling the carbon isotope characteristics of hydrocarbon gases.

Liquid hydrocarbons were a fundamental component of the thermal simulation products. When the temperature was lower than 350°C, the liquid hydrocarbon yield increased slowly with an increase in experimental temperature. In comparison, when the temperature was higher than 350°C, the yield decreased gradually as the temperature increased (Figure 3). In the three sample series analyzed herein, the variation trend of liquid hydrocarbon yield with changing temperature was consistent (Figure 3).

Although each series showed a similar change trend, the corresponding yields were markedly different at equivalent temperatures. At a given temperature condition, series A had the largest yield, and series C the smallest. These differences in liquid hydrocarbon yield further demonstrate the contribution of free and combinative soluble organic matter to hydrocarbon generation. The contribution of hydrocarbons generated by pyrolysis of combinative soluble organic matter results in the yield of series B being higher than that of series C. Moreover, concurrent pyrolysis of free and combinative soluble organic matter results in the yield of series A being the highest.

In addition to the yield and carbon isotope analysis of gaseous and liquid hydrocarbons, we also analyzed the characteristics of biomarkers (Figure 4). Because the total hydrocarbon production of series B was low, there was insufficient material to carry out GCMS analysis and biomarker parameter calculation. Therefore, only the differences between the biomarkers of series A and C are discussed below.

Both series A and C displayed abundant steranes at different temperatures, with the αααC₂₇ sterane having the highest concentration (Figures 4G,I). In addition, both samples contained abundant 4-methyl steranes (Figures 4G,I), which are related to a hydrocarbon generation parent material dominated by algae. Moreover, the mean value of gammacerane/αβ-C₃₀ hopane was 0.93 in series A and 0.96 in series C (Figure 4J,L), indicating a high salinity sedimentary environment. The contents of tricyclic terpanes in the two series were relatively consistent, among which the C₂₃ tricyclic terpane/(C₂₃ tricyclic terpane + C₃₀ hopane) of series A was 0.38, and that of series C was 0.26 (Figure 4J,L; Table 4). The above results indicate that the kerogen and soluble organic matter in these samples have the same parent material and source environment.

Although each of the series samples exhibited similar parent material and environment, the isoprenoid contents in the products were markedly different. As shown in Figures 4A,D, series A products contained high contents of Pr and Ph, and Ph was the dominant peak in the total ionization chromatograph. The ratio of (Pr+ Ph)/nC_10-35 in the product reached 40% at 250°C, but the ratio gradually decreased with increasing temperature. By comparison, the contents of Pr and Ph in the products of series C were markedly reduced (Figure 4C), and the ratio of (Pr+Ph)/nC_10-35 was only 1%–3% (Figure 5A; Table 4). Phytane dominance was observed in both series A and C, with Pr/nC₁₇ and Ph/nC₁₈ ratios of 0.89 and 2.46, and 0.35 and 1.38, respectively. However, with the simulated temperature increase, the ratio gradually approached zero owing to a large amount of isoprenoid cracking at high temperatures. This phenomenon is consistent with previous results (Xia et al., 1999).