The geochemical analysis of the natural gas was conducted at the Experimental and Testing Institute of PetroChina Xinjiang Oilfield Company and Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences. A Hewlett Packard 6890 II gas chromatograph (GC) was used for the analysis of the natural gas components. The hydrocarbon gas component was separated with capillary columns (Plot Al₂O₃ 50 m × 0.53 mm). The furnace temperature of the GC was first set to be 30°C and held for 10 min. Then the temperature was ramped up to 180°C at a rate of 10°C/min. Stable carbon isotope analysis of alkane gas (C₁–C₄) was conducted with a Finnigan Mat Delta S mass spectrometer interfaced with an HP 6890II gas chromatograph. The alkane gas components (C₁–C₃) and CO₂ were separated by using a chromatographic column (Plot Q 30 m × 0.32 mm). The column heating process was as follows: the heating rate was 8°C/min at temperatures of 35–80°C; the temperature was then increased to 260°C at a heating rate of 5°C/min. The final temperature was held for 10 min. Each sample was analyzed three times with an accuracy of ±0.3‰ (VPDB).

The hydrogen isotopic compositions of natural gas were determined on a MAT253 isotopic mass spectrometer (Thermo Fisher Scientific) equipped with a Trace GC Ultra™ using the gas chromatography pyrolysis interface and the water removing device. Helium was used as the carrier gas and a 30 m × 0.32 mm × 20 µm HP–PLOT Q column was used with flow rate of 1.4 ml/min. The inlet temperature was set at 180°C. A split injection mode (split ratio 1:7) was used for the methane hydrogen isotope measurement and a splitless injection mode for the ethane and propane hydrogen isotopes. The initial temperature was 40°C and held for 5 min, then heated from 40 to 80°C at 5°C/min, from 80 to 140°C at 10°C/min and from 140 to 260°C at 30°C/min, respectively. The temperature of the pyrolysis oven was 1,450°C, and gaseous hydrocarbon components were transformed into C and H₂. The H₂ went into mass spectrometer to be measured. The δ²H was calculated relative to VSMOW. The reproducibility and precision of hydrogen isotope value are less than ±3‰.

The GC analysis of oils was performed in an Agilent 7890A gas chromatograph fitted with a 60 m × 0.25 mm × 0.25 µm capillary column with nitrogen (99.999%) as the carrier gas. The GC oven temperature was initially held at 40°C for 10 min, then ramped from 40 to 70°C at 4°C/min and to 300°C at 8°C/min, and finally held at 300°C for 40 min.

The GC–MS analysis was performed in an Agilent 7890–5975C with the same column type as used in the GC analysis, but with helium (99.999%) as the carrier gas. During the GC–MS analysis, the GC oven temperature was initially held at 50°C for 1 min, then ramped to 120°C at 20°C/min, from 120 to 250°C at 4°C/min, and from 250 to 310°C at 3°C/min. Finally, it was held at 310°C for 30 min.

The burial and thermal histories of the source rocks in the study area were reconstructed using PetroMod software. The current heat flow and thermal conductivity values of the source rocks were adopted from previous studies (Wang et al., 2000a, Wang et al., 2000b; Qiu et al., 2000; Qiu et al., 2001; Qin, 2002). The vitrinite reflectance (Ro) values were calculated with the Easy%Ro model proposed by Sweeney and Burnham (1990). This model was proved to be applicable for a Ro range of 0.3%–4.6%.

The analyses were carried out at the China University of Petroleum (Beijing). The 146 rock samples were crushed to powder for total organic carbon (TOC) analysis (Table 1). The powdered samples were split into 200-mg sub-samples and treated with HCl at 60°C to remove the carbonates, then washed with distilled water to remove the HCl. The washed sub-samples were dried overnight at 50°C, and their carbon contents were determined using a LECO CS–230 analyzer.

For rock–eval pyrolysis, 100 g of each crushed rock sample was placed in the vessel of an OGE-II instrument. These samples were heated from 300 to 600°C in a helium atmosphere at a heating rate of 50°C/min, and their Rock–Eval parameters (S₁, S₂, and T_max) were measured. Here S₁ is the amount of free hydrocarbon that can be volatilized from the rock sample (mg HC/g rock), and S₂ is the amount of hydrocarbon produced by the cracking of organic matter (mg HC/g rock). T_max (^oC) is the temperature at which the S₂ yield is maximized, which roughly estimates the thermal maturity of the sediment (Peters, 1986).

Four sets of source rocks were developed in the study area (i.e., J_1–2, P₂w, P₁f, and Carboniferous source rocks) (Cao et al., 2012; Wang et al., 2013). The J_1–2, P₂w, and Carboniferous source rocks are humic, characterized by kerogen types II₂–III with their carbon isotopic composition of kerogen (δ¹³C_kerogen) mainly varying from –26.0‰ to –20.0‰ (Figure 5). These source rocks primarily generate coal-type gases (Dai et al., 1992; Liu G. et al., 2019). The P1f source rock was deposited under a saline lacustrine environment, with algae and lower aquatic organisms as the primary organic input (Cao et al., 2020; Xia et al., 2021; Zhi et al., 2021). It is a set of oil-prone source rock characterized by kerogen type II₁ with their δ¹³C_kerogen mainly varying from –28.0‰ and –24.0‰ (Figure 5). The Mahu super-giant oil field’s petroleum in the Junggar Basin was mainly generated from the P₁f source rock (Wang et al., 2013; Cao et al., 2020). Besides oil, the P₁f source rock can also generate a considerable amount of oil-type gas. As discussed in Section 5.1, Type I gas was an oil-type gas with relatively ¹³C-depleted δ¹³C ratios (Figure 4; Table 1), which was generated from the P₁f source rock, the only sapropelic source rock in the study area.

This conclusion was supported by evidence from the δ²H ratios of natural gas. Identical to δ¹³C ratios, the δ²H ratios of alkane gases also have parent material inheritance (Schoell, 1980, Schoell, 1984; Ni et al., 2011). Under the same or similar thermal evolution conditions, the δ²H ratios of the natural gas generated by marine or saline lacustrine source rocks are usually more enriched in ²H (Liu et al., 2008a; Wang et al., 2015; Liu Q. et al., 2019). Therefore, the δ²H ratios of alkane gas can be used to judge the geneses and sources of natural gas (Schoell, 1980; Wang et al., 2015). Gases from Well PD1 have δ²H-CH₄ and δ¹³C-CH₄ ratios of –164‰ and –38.5‰, respectively, and was located within the distribution area of oil-type gas generated from the saline sapropelic source rock (Wang et al., 2015; Figure 6; Table 2). Two gas samples derived from the P₁f source rock in the Mahu Sag (Dai et al., 1992) were also located in this area (Figure 6; Table 2).

As mentioned in Section 5.1, the Type II gas is a highly mature coal-type gas generated from the humic source rock. Unfortunately, however, three sets of humic source rocks (i.e., J_1–2, P₂w, and Carboniferous) developed in the study area (Figure 5), which brought added complications in identifying the gas source.

In northwestern China, the J_1–2 source rock is a set of high-quality coaly source rock, and a large number of giant gas fields related to the J_1–2 source rock were discovered (Dai et al., 2009, 2014). Some researchers once pointed out that the study area’s natural gas was mainly generated from the J_1–2 source rock (Dai et al., 2009; Li et al., 2009). In this study, the limiting thermal evolution conditions of the four sets of source rocks were simulated (Figure 7A) based on a pseudo well in the deepest part of the Pengyijingxi Sag (for well location see Figure 1C). When the Lower Jurassic Badaowan (J₁b) source rock reached the maximum burial depth, it was still in the immature–low mature stage (Ro = 0.5%–0.7%) (Figure 7A). In other words, this set of source rock in the study area has never entered the main oil generation window, let alone the main gas generation window. Therefore, the contribution of J_1–2 source rock was excluded.

Type II gas from Wells MS1, SD2, and LD1 were produced from the Carboniferous reservoirs (Table 1). The P₂x was interbedded between the Carboniferous and P₂w with a thickness of approximately 200 m in Well MS1. Besides, the gas-producing zone was approximately 400 m lower from the top Carboniferous in Well MS1. Moreover, Permian strata were absent in the area where Well SD2 and Well LD1 were drilled. Therefore, it is inferred from geological conditions that the Type II gas was derived from the Carboniferous source rock rather than from the P₂w source rock. Previous studies have shown that the natural gas of the Kelameili gas field in the eastern Junggar Basin was generated from the Carboniferous source rock (Sun et al., 2016; Gong et al., 2019a, Gong et al., 2019b, Gong et al., 2021). It shared similar δ¹³C ratios with the Type II gas (Figure 4), further supporting our conclusion.

Generally, the δ¹³C_kerogen values of P₂w source rocks were more enriched in ¹³C than Carboniferous source rocks, although the former has a broader distribution range (Figure 5). The δ¹³C_kerogen ratios revealed that the P₂w source rock has more terrigenous organic input and poorer kerogen type (Hunt et al., 1996). When the maturities of natural gas were similar (the C₁/∑C_1–4 ratios were similar) (Table 1), the δ¹³C-CH₄ values of natural gas generated from the P₂w source rocks should theoretically be more enriched in ¹³C than those generated from the Carboniferous source rocks (Liu et al., 2008b; Gai et al., 2018). However, the actual condition was just the opposite (Figure 4). We also noticed that a small number of P₂w source rock samples had relatively ¹³C-depleted δ¹³C_kerogen values (–28‰ to –25‰) (Figure 5). Thus, the contribution to Type III gas by P₂w source rock could not be totally excluded, entirely based on δ¹³C_kerogen values (Figure 5).

However, evidence from the fluid inclusions further precluded the significant contribution from the P₂w source rock. SX16 was a high yield well drilled in 2011 in the study area (Figure 1C). The producing layer was the Carboniferous volcanic rocks, and the daily gas production was 86.27 × 103 m³. Abundant oil and gas inclusions and associated contemporaneous brine inclusions were detected in the calcite-filled fractures in the producing interval (Figure 8). Through microscopic observation, one period of oil charging and one period of gas charging were identified. The oil inclusions mainly emitted blue or blue-green fluorescence (Figures 8A,B). Most natural gas inclusions are circular/oval-shaped and appear black under transmitted light with no fluorescence (Figures 8C,D). The homogenization temperature of the brine inclusions associated with natural gas inclusions ranged from 92.7 to 106.6°C. Combined with the thermal evolution history of Well SX16, the natural gas filling period was mainly the Late Cretaceous (Figure 7B). At that time, the deepest P₂w source rock in the Pengyijingxi Sag was still in the main oil generation window (Ro = 1.0%–1.3%) and had not entered the gas generation stage (Figure 7A). On the contrary, the P₁f and Carboniferous source rocks had entered a high mature–overmature stage in the deep part of the sag (Figure 7A). They could have generated a large amount of natural gas, which accumulated in suitable traps. Therefore, type III gas was most likely a mixture of natural gases generated from the P₁f and Carboniferous source rocks.

In addition, the TOC, S₁ + S₂, and hydrogen index (HI) of P₂w source rock were 0.1%–1.2% (0.5% on average), 0.08–3.16 mg HC/g rock (0.82 mg HC/g rock on average), and 7–103 mg HC/g TOC (53 mg HC/g TOC on average), respectively, indicating poor source rock–non-source rock (Figure 9). The P₂w source rock had a hydrocarbon generation capacity far inferior to P₁f and Carboniferous source rocks (Figure 9). Thus, it did not have the potential to form a large-scale natural gas field (reservoir).

Based on the hypotheses, it was believed that Type III gas was a mixture of natural gas generated from the Carboniferous and P₁f source rocks. The mixing gas (Type III) had δ¹³C-CH₄ values between Type I and Type II gases (Figure 4; Table 1). The δ¹³C ratios of C_2–4 of Type III gas were relatively ¹³C-enriched, similar to those of Type II natural gas (Figure 4; Table 1). Under similar maturity, the oil-type gas usually had a lower content of C_2–4 than the coal-type gas (Dai et al., 1992; Whiticar, 1994; Liu Q et al., 2019), causing the δ¹³C ratios of C_2–4 gas in type III to be closer to those in type II, rather than in Type I gas.

The evidence from hydrogen isotopes further confirmed our conclusion. The δ²H-CH₄ values of three samples from Peng5 gas field were –186‰, –185‰ and –186‰, respectively (Figure 6; Table 2), which was significantly different from those of gases generated from the fresh (humic) source rocks such as Carboniferous or P₂w (Wang et al., 2013). The three samples were located in the transitional area between the coal-type gas (from fresh humic source rock) and oil-type gas (from saline sapropelic source rock) (Figure 6). Considering the number and planar distribution of the samples, the Type III natural gas constituted the main part of the natural gas in the hinterland.

However, recent researches indicated that Well JT 1 in the Shawan Sag has encountered high-quality P₂w source rocks with their TOC, HI and δ¹³C_kerogen being 4.15%, 671 mg HC/g TOC and –28.6‰, respectively (Zhi et al., 2021; Gong et al., 2022). Thus, the locally developed high-quality P₂w source rocks cannot be excluded in the hinter land of Junggar Basin. Since the P₂w source rock has entered condensate-generating period in the Pengyijingxi Sag nowadays (Figure 3A), it may also made some contribution to the Type III gas.

Compared with the δ¹³C ratio of C_2–4 gas, methane is generally more sensitive to the change of maturity (Stahl and Carey, 1975; Dai et al., 1992; Berner and Faber 1996). Thus, the Δδ¹³C-(CH₄ – C₂H₆) ratio is often used to reflect the maturity of natural gas (Prinzhofer and Huc, 1995; Dai et al., 2014; Wu et al., 2019). For primary thermogenic gas, the increase of Δδ¹³C-(CH₄ – C₂H₆) reflected the increase of maturity. However, with the increase of Δδ¹³C-(CH₄ – C₂H₆), the ln(C₁/C₂) of Type IV gas showed a decreasing trend, which was not in line with the characteristics of primary thermogenic gas (Figure 10A).

We believe that Type IV gas was mainly a secondary microbial gas. Natural gas generated via bacterial alteration of organic matter is called primary microbial gas, whose δ¹³C-CH₄ values are generally lower than –55‰ to –60‰ (James and Burns, 1984). The δ¹³C-CH₄ of Type IV natural gas was more enriched in ¹³C than this threshold (Table 1). Secondary microbial gas refers to the natural gas formed by the biodegradation (anaerobic biodegradation in most cases) of crude oil (Head et al., 2003; Etiope et al., 2009; Milkov, 2010; Milkov, 2011). Generally, secondary microbial gas has the following characteristics: 1) δ¹³C-CO₂ > +2‰; 2) Natural gas is associated with biodegraded crude oil; 3) δ¹³C-CH₄ = –55‰ to –35‰; 4) High methane content; 5) Low reservoir temperature (<70–90°C); 6) The reservoirs are mostly sandstone formations with high porosity and permeability; and 7) The reservoirs are usually characterized by normal pressures with poor sealing conditions (Bernard et al., 1992; Milkov, 2010; Milkov, 2011).

The crude oil associated with Type IV gas is generally subjected to biodegradation (Figure 11). The 25-norhopanes are formed by the loss of methyl group at C–10 position of hopanes via the alteration of microorganisms (Seifert and Moldowan, 1979). Their appearance has been widely used as a sign of severe biodegradation to crude oil (Moldowan et al., 1984; Tian et al., 2012). In the study area, abundant C₂₈ and C₂₉ 25-norhopane were detected in the m/z 177 mass chromatogram of crude oil associated with the Type IV gas, indicating that the crude oil had suffered severe biodegradation (Figure 11). In addition, when crude oil was subjected to biodegradation, the compounds with strong resistance to biodegradation formed bulges (i.e., unresolved complex mixtures, UCM) in gas chromatography of saturated hydrocarbons or FID diagram (Killops and Al-Juboori, 1990; Gouch et al., 1992; Hu S. Z. et al., 2020). Noticeable UCM bulges were observed in the FID diagram in the current samples, which also reflected the biodegradation of crude oil (Figure 11).

The δ¹³C-CH₄ and C₁/∑C_1-4 ratios of Type IV gas were –54.8‰ to –43.2‰ (–48.7‰) and 0.93–1.0 (0.97 on average), respectively, which are well in accordance with those of the typical secondary microbial gas (Table 1). The gas-producing zones of Type IV gas had favorable physical properties, with their porosity and permeability being 20.7%–31.0% (24.9% on average) and 313 × 10⁻³ μm²–4,760 × 10⁻³ μm² (1,180.6 × 10⁻³ μm² on average), respectively. All gas-producing zones had normal pressures, with the formation pressure coefficients (obtained by the DST test) being 0.77–1.04 (0.92 on average). The burial depths of the producing zones were 960–2,041.5 m. Assuming the current surface temperature was 20°C and the geothermal gradient was 22.3°C/km (Qiu, 2002; Wang et al., 2013), the current formation temperatures were 41.4–65.5°C (Figure 10B; Table 1). Although δ¹³C-CO₂ data were not obtained in this study, considering other characteristics, the Type IV gas was identified as a secondary microbial gas in this study。

It should be noted that laboratory simulation and field examples showed that the main component of the pure secondary microbial gas was methane, and the content of C_2–4 gas was extremely low (Pallasser, 2000; Milkov and Dzou, 2007; Jones et al., 2008). However, a certain amount of C_2–4 gas components were found in Type IV natural gas (Table 1), indicating the gases were also mixed with a small amount of thermogenic gas.

Bacteria may alter C_1–4 gaseous hydrocarbons in gas reservoirs (Davis, 1967; Whiticar, 1994; Clayton et al., 1997; Pallasser, 2000). In most cases, the alteration on C_2–4 gas components tends to be more prominent (James and Burns, 1984). The selective degradation of a specific gas component by bacteria will reduce its relative content. At the same time, as ¹²C is less stable than ¹³C, it is often degraded first, resulting in a more ¹³C-enriched δ¹³C ratio of natural gas (Dai et al., 1992; Galimov, 2006). The carbon isotopic reversal of ethane and propane were observed in natural gas in Wells L102 and M11, with Δδ¹³C-(C₂H₆ – C₃H₈) values being 2.0 and 0.08‰, respectively (Figure 12A). Meanwhile, the C₂/C₃ ratios of the two samples were relatively low, at 2.25 and 1.76, respectively (Figure 12A). The above characteristics inferred that the ethane was selectively degraded by bacteria.

Similarly, some samples had significant carbon isotopic reversal between propane and butane [Δδ¹³C-(C₄H₁₀ – C₃H₈) < 1‰], accompanied by a decrease in C₄/C₃ ratio (Figure 12B). The propane in this part of the natural gas was selectively degraded by bacteria. This process also resulted in the abnormal decrease of Δδ¹³C-(C₂H₆ – C₃H₈) (<–5‰) and the increase of C₂/C₃ in these natural gas (Figure 13A; Table 1). We noted differences in the degraded gas components in different samples, which may be caused by different types of bacterial strains involved in the degradation processes (Connan et al., 1996; Wilkes et al., 2000; Gong et al., 2017). Moreover, some natural gas samples also displayed δ¹³C-C₃H₈ and δ¹³C-C₄H₁₀ inversion, but the amplitude was relatively small (<1‰), and the C₄/C₃ ratio had not increased significantly (Figure 12B). The carbon isotopic reversal of these samples was more likely caused by the mixing of natural gas with different maturity or different geneses (Dai et al., 2004; Liu et al., 2016).

Gas leakage often occurs in the reservoir when the sealing condition is poor. In this case, the gases in the leakage phase will be rich in methane, and its δ¹³C ratios tend to be depleted in ¹³C (Prinzhofer and Huc, 1995). In contrast, the residual gas in the reservoir tends to show lower C₁/C₂ ratio and higher δ¹³C-CH₄ values [decreasing δ¹³C-(CH₄ – C₂H₆) values] (Prinzhofer and Huc, 1995). As shown in Figure 10A, considerable part of natural gas in the study area deviated from the evolution trend of primary thermogenic gas and showed the characteristics of residual phase gas after leakage. Thus, the leakage process reflected a poor preservation condition of gas reservoirs in the study area, and intense modifications occurred after the formation of these reservoirs. Wells LN6 and SD10 are two typical examples. The extremely low C₁/C₂ values in natural gases (Table 1) indicated that the two gas reservoirs had been completely destroyed after reconstruction. Furthermore, the oil test data showed that LN6 and SD10 only produced more than 100 m³ of natural gas per day, which also confirmed our analysis.