Since the publication of "Petroleum and Natural Gas Generation During Coalification" in 1979^[1], under the guidance of coal-derived gas theory, the geological reserves of natural gas in China have increased rapidly in the past 40 years. Before the occurrence of coal-derived gas theory in 1978, China’s total natural gas reserves were 2284 × 10⁸ m³ (203 × 10⁸ m³ coal-derived gas), and the annual gas production was 137 × 10⁸ m³ (3.43 × 10⁸ m³ coal-derived gas). While in 2016, the natural gas reserves in China were 118 951.20 × 10⁸ m³ (82 889.32 × 10⁸ m³ coal-derived gas), the annual gas production was 1 384 × 10⁸ m³ (742.91 × 10⁸ m³ coal-derived gas), and there were 39 large coal-derived gas fields, which accounted for 66% of the total number of large gas fields (59) in China. The Sulige gas field, which has the largest reserves and the highest annual gas production in China, is a coal-derived gas field^[2]. With the development of economy and the progress of society, the world's demand for oil and gas resources continues to increase, and with the mass discovery of conventional oil and gas, its exploration becomes more and more difficult, and exploration targets continue to expand to other difficult areas, such as unconventional, deep, low mature, high-over mature and abiogenic gases. Since natural gas is mainly composed of a few simple low molecular weight hydrocarbons, its genetic analysis mainly depends on the characteristics of molecular composition, carbon and hydrogen isotopic composition^[2]. Based on the genetic identification of natural gas and gas source correlation, a series of fruitful studies have been carried out by using the carbon and hydrogen isotopic composition of natural gas, which provides an important research method for the identification of natural gas with complex origin in superimposed basins^{[3,4,5,6,7,8,9,10]}.

The Turpan-Hami Basin, an important oil-gas-rich basin in China, has long been regarded as a typical coal-derived oil basin^{[11,12,13,14]}. Its coal-derived gas exploration began in the early 1990s. At present, most of the natural gas exploration work in this basin is concentrated in the Taibei sag. However, there have been many disputes about the origin and source of natural gas in the Taibei sag, and the main points are as follows: Natural gas in the Taibei sag is a low mature coal-derived gas from the Middle and Lower Jurassic source rocks, dominated by the Xishanyao Formation^[15,16] with minor contribution from the Badaowan Formation, or dominated by the Badaowan Formation with minor contribution from the Xishanyao Formation^[17]; Most of the natural gas in the Turpan-Hami Basin belongs to mixed gas with characteristics close to oil-derived gas, which comes from coal-measure mudstone rather than coal seam^[13]; Natural gas in the Taibei sag is mainly coal-derived gas, but some wells in the Baka and Shanshan oil fields belong to biodegraded gas or mixed gas^[18]; The tight sandstone gas in the Qiudong subsag is coal-derived gas and mixed gas from coal-measure mudstone, and in the Kekeya area, the coal-derived gas comes from coal-measure mudstone while the mixed gas mainly comes from the coal-measure source rocks of Xishanyao Formation^[19]. The source and origin of natural gas in the Taibei sag have not been determined yet. According to the geochemical characteristics of natural gas, especially together with the study of gas hydrogen isotopic composition, Ni et al.^[3] proposed that natural gas in the Qiudong and Hongtai gas fields is coal-derived gas from the Middle and Lower Jurassic coal-measure source rocks, but no detailed study about the origin of natural gas in other areas of the Taibei sag has been carried out. According to the molecular composition, stable carbon and hydrogen isotopic compositions, the source and origin of natural gas in the Baka, Shanshan, Qiuling and Wenmi oil fields in the Taibei sag are analyzed in this paper.

The Turpan-Hami Basin, located in eastern Xinjiang and distributed in East-West direction, is one of the three major sedimentary basins in Xinjiang. The basin is 660 km long from east to west and 60 km wide from north to south, with an area about 5.35 × 10⁴ km². The inner structural units of the basin are divided into the eastern Hami depression, the central Liaodun uplift and the western Turpan depression. The Turpan depression is the main depression of the basin, and the Taibei sag is a secondary structural unit of the Turpan depression, with an area of 9 600 m², which is the main oil-gas bearing area of the Jurassic coal measures (Fig. 1). The basin developed the Carboniferous-Quaternary system with a maximum cumulative thickness of more than 9 000 m. The Carboniferous-Permian is a set of marine sedimentary rock-volcanic rock assemblage, the Triassic belongs to semi-deep lacustrine-shallow lacustrine facies, the Middle-Lower Jurassic is dominated by semi-deep lacustrine-fluvial swampy coal-bearing deposit, and the Cretaceous and Tertiary are shallow lacustrine-fluvial facies^[13]. The Jurassic is the most widely distributed in the basin with a thickness of 4 600 m, and the Taibei sag is the one with the most well-developed Jurassic (Fig. 2). Source rocks in the Turpan-Hami Basin mainly include Carboniferous-Lower Permian marine mudstone, Upper Permian and Upper Triassic Huangshanjie Formation semi- deep lacustrine-shallow lacustrine facies mudstone, Middle- Lower Jurassic Badaowan Formation and Xishanyao Formation semi-deep lacustrine-fluvial swampy facies coal-bearing deposits and Middle Jurassic Qiketai Formation black mudstone^[11,14,20]. Among them, the Badaowan Formation and Xishanyao Formation of the Middle-Lower Jurassic are the main source rocks in the basin, and the degree of coalification is relatively low. The R_o value of the superface of the Xishanyao Formation is 0.4%-0.9%, most of which are mature source rocks and some of them are low mature source rocks. The Lower Jurassic has higher maturity, and the Badaowan Formation of the Lower Jurassic is a set of mature-high mature source rocks^[13]. In the Xishanyao Formation, the thickness of dark mudstone is 600 m with a peak value of 200-400 m, the TOC value of dark mudstone is 0.5%-3.6%, and the thickness of coal seam is 40-60 m with maximum of 100 m. In the Badaowan Formation, the thickness of dark mudstone is 50-300 m, the TOC value of dark mudstone is 0.5%-3.0%, and the thickness of coal seam is 40-60 m^[11,13]. In the basin, the thickness is 50-100 m, 50-200 m and 100- 200 m for the dark mudstone of the Middle Sangonghe Formation, the Sanjianfang Formation and the Qiketai Formation, respectively^[13]. The average thickness of coal seam in the basin is 70 m-80 m, and the coal content of Xishanyao Formation is more than that of Badaowan Formation. The Sangonghe Formation, Sanjianfang Formation and Qiketai Frmation have less or almost no coal^[13].

A total of 23 natural gas samples were collected from the Baka, Shanshan, Qiuling and Wenmi oil fields in the Taibei sag, Turpan-Hami Basin. In addition, 12 gas samples in the Hongtai gas field and 11 gas samples in the Qiudong gas field were compared and analyzed^[3]. Gas samples were collected with double valve high pressure cylinders. The molecular composition and carbon/hydrogen isotopic compositions were determined in the PetroChina Research Institute of Petroleum Exploration and Development. The gas composition was measured by an Agilent 7890 gas chromatograph and the carbon isotopic composition was determined by a gas chromatography-isotope ratio mass spectrometry (GC-IRMS). The device is composed of a Thermo Delta V mass spectrometer and a Thermo Trace GC Ultra chromatograph. The hydrogen isotopic composition of natural gas was analyzed on a GC/TC/IRMS, which is connected by a MAT253 mass spectrometer and a Trace GC Ultra chromatograph equipped with a 1 450 °C micro-pyrolysis furnace. Each sample is repeated at least twice. The accuracy is ±0.3‰ for the carbon isotopic analysis and ±3‰ for the hydrogen isotopic analysis. All the values are reported in the δ-notation in per mil (‰) relative to VPDB for the carbon isotope and relative to VSMOW for the hydrogen isotope^[21].

Based on the analysis of gas samples from the Baka, Shanshan, Wenmi and Qiuling oil fields in the Taibei sag (Table 1), and combined with previous studies on natural gas in the Qiudong and Hongtai gas fields (Table 2), the geochemical characteristics of natural gas in this area are further analyzed.

Natural gas in the Baka, Qiuling, Shanshan and Wenmi areas in the Taibei sag is dominated by hydrocarbon gas. The content of methane ranges from 65.84% to 97.94%, with an average of 81.29%, and the content of heavy hydrocarbon gas (C_2-5) varies from 1.55% to 34.98%, with an average of 18.70% (Table 1). The dryness coefficient (C₁/C_1-5) is 0.66- 0.98 with an average value of 0.81, which is slightly lower than that of the gas in Qiudong (average: 0.83) and Hongtai (average: 0.85). However, all of them are wet gas. The content of non-hydrocarbon gas (CO₂, N₂) was very low, only a trace amount of N₂ (less than 0.04%) was found in individual wells, and no CO₂ has been detected.

Natural gas in the Baka, Qiuling, Shanshan and Wenmi oil fields has δ¹³C₁ value of -44.9‰ - -40.4‰ with an average of -41.6‰, δ¹³C₂ value of -28.2‰ - -24.9‰ with an average of -26.9‰, δ¹³C₃ value of -27.1‰ - -18.0‰ with an average of -25.2‰, δ¹³C₄ value of -26.7‰ - -22.1‰ with an average of -24.9‰, and δ¹³C₅ value of -25.9‰ - -22.5‰ with an average of -24.4‰ (Table 1). In addition to wells Ba-23 and Ke-19 in the Baka oil field, methane and its homologues (C₂-C₅) are basically characterized by a normal distribution pattern of carbon isotopic composition (δ¹³C₁<δ¹³C₂<δ¹³C₃< δ¹³C₄<δ¹³C₅), which is consistent with the typical isotopic characteristics of thermogenic alkanes^[25]. The δ¹³C₁ value in the study area is similar to that in the Qiudong gas field (mean δ¹³C₁: -41.2‰), but lower than that in the Hongtai gas field (mean δ¹³C₁: -38.4‰). In general, δ¹³C value of methane in the Baka, Qiuling, Shanshan and Wenmi oil field are similar, all of which are lower than -40‰, indicating that the maturity of the gas is similar to that in the Qiudong gas field but lower than that in the Hongtai gas field^[3]. The δ¹³C value of propane and/or butane from the Ba-23 and Ke-19 wells in the Baka oil field became relatively heavy, which resulted in partial reversal of the carbon isotopic composition among the C₁-C₅ alkanes (Fig. 3), indicating the existence of secondary alteration in the late stage. The δ¹³C value of ethane in the Baka, Qiuling, Shanshan and Wenmi oil fields are also similar, with mean values of -26.5‰, -27.8‰, -27.8‰ and -26.8‰ for the ethane from the Baka, Qiuling, Shanshan and Wenmi oil fields, respectively. The δ¹³C values of ethane in the Wenmi oil field are relatively high and the variation range is very small, ranging from -26.4 ‰ to -27.0 ‰. Except for the wells Ba-23 and Ke-19 (δ¹³C₂ in Well Ba-23: -25.4‰, δ¹³C₂ in Well Ke-19: -24.9‰), the δ¹³C values of ethane in the Baka oil field vary little, -26.7‰ - -27.3‰. The δ¹³C values of propane in wells Ba-23 and Ke-19 are also relatively high, -18.0‰ for Well Ba-23 and -23.8‰ for Well Ke-19, which are higher than those in other wells.

The δD₁ values of natural gas in the Baka, Qiuling, Shanshan and Wenmi oil fields have little change, ranging from -272‰ to -252‰, with an average of -262‰. The variation range of δD₂ is a bit bigger, -236‰ - -200‰ with an average of -225‰, while the range of propane is the biggest, -222‰ - -174‰ with an average of -211‰ (Table 1). The δD value of methane and its homologues (C₂, C₃) are in normal distribution pattern (δD₁<δD₂<δD₃) (Fig. 3).

Methane and its homologues (C₂-C₅) in the Baka, Qiuling, Shanshan and Wenmi oil fields generally have normal distribution pattern of carbon and hydrogen isotopes (δ¹³C₁<δ¹³C₂< δ¹³C₃<δ¹³C₄<δ¹³C₅,δD₁<δD₂<δD₃) (Fig. 3). That is, the carbon isotope composition of alkane gas becomes more enriched in ¹³C with the increase of carbon number, which is consistent with the typical thermogenic alkane gas^[2]. This is the result of kinetic isotope fractionation effect, that is, when an alkyl is separated from its parent organic matter, the ¹²C-¹²C bond is weaker than the ¹²C-¹³C bond, so it breaks first, therefore, the pyrolytic product is more depleted in ¹³C than its precursor^[26]. According to the Whiticar plot^[27] (Fig. 4) and the Bernard plot^[28] (Fig. 5), gas samples in the study area all fall in the area of thermogenic gas. The data are relatively concentrated, and there is no mixture with biogas. It is similar to the sample data of Qiudong and Hongtai, and some of them are almost overlapped. In the Whiticar plot (Fig. 4), the samples mainly fall in the low mature thermogenic gas region, while in the Bernard plot (Fig. 5), samples are mainly inclined to Type Ⅲ kerogen.

According to the different types of parent organic matter, thermogenic gas can be divided into two types: coal-derived gas (mainly from terrestrial humic organic matter) and oil-derived gas (mainly from marine or lacustrine sapropelic and sapropelic-humic organic matter). The kerogen type of coal-derived gas source rock is Type III and Type II₂, which are mainly composed of aromatic structure and short branched chain structure which are relatively enriched in ¹³C. The oil-derived gas is formed by the kerogen of Type I and Type II₁ in the source rocks, mainly composed of long-chain aliphatic structure which is relatively enriched in ¹²C^[29]. During the gas generation process of either humic or sapropelic kerogen at the same or similar thermal maturity, the carbon isotopic composition of the source rocks will be inherited, therefore, the δ¹³C values of coal-derived methane and its homologues are heavier than that of the oil-derived methane and its homologues at similar thermal maturity^[30]. The δ¹³C value of ethane has a strong inheritance of the parent organic matter. Compared to methane, the controlling effects of thermal maturity on the δ¹³C value of ethane is much smaller. Therefore, the δ¹³C value of ethane is often used as an effective indicator to distinguish between coal-derived gas and oil-derived gas^[30]. At present, domestic scholars mainly use -28‰^[31-32] or -29‰^[30,33-34] as a boundary value. According to previous studies, Ni et al.^[3] adopted -28‰ as a boundary value to determine coal-derived and oil-derived gases, and pointed out that the δ¹³C value of ethane in the Qiudong and Hongtai gas fields is not lower than -27.5‰, which belongs to coal-derived gas. The δ¹³C value of ethane in the study area ranges from -28.2‰ to -24.9‰, with an average of -26.9‰. Except for Well Shan13-61C which has δ¹³C₂ value of -28.2‰, the δ¹³C value of ethane in other wells is higher than -28‰ (Fig. 6). According to the δ¹³C value of ethane, natural gas from the Shanshan, Baka, Qiuling and Wenmi oil fields is mainly coal-derived gas by using -28‰ as a distinguishing index.

The δ¹³C value of ethane is mainly influenced by the parent material types, but also by the thermal maturity of source rocks. Generally speaking, the δ¹³C values of methane and ethane increase with increasing thermal maturity of source rocks^[7,29,35]. As shown in Fig. 6, the more mature the gas samples are, the more they fall to the top right of the plot. And samples with much lower maturity will fall in the bottom left of the plot. For the primary gas that has not experienced secondary alteration, if it falls on the same maturity trend line, it may indicate that the gases have similar source but at different thermal evolution stages, and if it falls on different maturity trend lines, it is more likely to reflect that the gases are from different sources or have experienced secondary alteration. Fig. 6 shows different types of δ¹³C₁-δ¹³C₂ relationships of coal-derived gas sourced from Type Ⅲ kerogen^[23,36-38]. The gas samples of the Shanshan, Baka, Qiuling and Wenmi oil fields in the Taibei sag fall on similar maturity trend line as the gases from Type III kerogen in the Sacramento Basin^[36], which indicates that gas in the study area is of coal-derived origin and sourced from Type III kerogen. Compared with the gases from Hongtai, some of the samples in the study area fall in the area more inclined to the bottom right in Fig. 6, that is, at the lower end of the maturity trend line, similar to the gas samples from Qiudong, indicating that the source rock maturity represented by these well samples is lower. This agrees well with the fact that there is more heavy hydrocarbon gas in the composition, that is, the dryness coefficient is lower. Therefore, gases from the four oil fields, Shanshan, Baka, Qiuling and Wenmi, in the Taibei sag, are coal-derived gas with low maturity. The thermal maturity of the Middle and Lower Jurassic coal-measure source rocks in the Xiaocaohu subsag is higher than that in the Qiudong subsag, which is consistent with previous studies^[3,11]. Generally speaking, the gas maturity and carbon isotopes of methane and ethane in the Shanshan, Baka, Qiuling and Wenmi oil fields are relatively low, compared with that of Xujiahe Formation in the Sichuan Basin^[8]. In the δ¹³C₁-δ¹³C₂ plot, it obviously falls in the bottom left (Fig. 6).

By using δ¹³C₁-δ¹³C₂-δ¹³C₃ plot of Dai et al. ^[39] (Fig. 7), gases from the study area mainly fall in the area of coal-derived gas, which is similar to the gases from Hongtai and Qiudong. Gases of wells Ba-23 and Ke-19 deviate from δ¹³C₁- δ¹³C₂ maturity trend obviously in Fig. 6, and Well Ba-23 also falls outside the coal-derived gas zone in Fig.7. The carbon isotopes between propane and butane in the two wells are reversed (δ¹³C₃>δ¹³C₄). Many factors may cause a carbon isotopic reversal of alkane gases, such as mixing, biodegradation^[25]. Well Ba-23 and Well Ke-19 accord with biodegradation, mainly for the following four reasons. (1) The reservoir depth of Well Ba-23 (1 174.0-1 180.0 m, 1 854.6-1 876.2 m, 1 901.8-1 916.0 m and 1 930.0-1 951.2 m) is shallow, less than 2 000 m, and the formation temperature is generally lower than 80 °C, so the bioactivity is strong and the biodegradation is easy to occur. (2) Except for the dry gas of Well Ke-21C, the content of C_2-5 heavy hydrocarbon in wells Ba-23 and Ke-19 is the lowest among the other 22 wells in the study area, in which the content of C_2-5 is 10.92% in Well Ba-23 and 7.16% in Well Ke-19; The C_2-5 gas content of the remaining 20 wells in the study area is 12.67%-34.15%, with an average of 20.50%, which is significantly higher than that of wells Ba-23 and Ke-19. The C_3-5 heavy hydrocarbon content in wells Ba-23 and Ke-19 is even lower. The C_3-5 content is 2.54% in Well Ba-23 and 0.75% in Well Ke-19. The C_3-5 content of the remaining 20 wells in the study area (except for dry gas in Well Ke-21C) is 4.87%-17.31%, with an average of 9.86%, which was significantly higher than that in Well Ba-23 and Well Ke-19. (3) In the 7 wells of Baka oil field, the δ¹³C values of ethane and propane in wells Ba-23 and Ke-19 are obviously high. The δ¹³C values of ethane and propane are -25.4‰ and -18.0‰ in Well Ba-23, and are -24.9‰ and -23.8‰ in Well Ke-19, respectively. While the average δ¹³C values of ethane and propane are -27.0‰ and -25.2‰ for the remaining five wells in the Baka oil field, respectively. (4) The δ¹³C₁ values of wells Ba-23 and Ke-19 are obviously low in the 7 wells of Baka oil field. The δ¹³C₁ values in wells Ba-23 and Ke-19 are -44.9‰ and -42.3‰, respectively, while the mean value of δ¹³C₁ in the other five wells is -41.0‰. During the process of biodegradation, the bacteria will preferentially oxidize the ¹²C-¹²C bond, resulting in the enrichment in ¹³C of the remaining component, thereby making its δ¹³C value heavy. For example, when bacteria oxidizes the propane, propane in natural gas will be degraded and consumed preferentially, resulting in heavier δ¹³C value and less amount of remaining propane^[29]. At the same time, the process of biodegradation can also produce biogas which is dominated by methane and characterized by low δ¹³C value. Therefore, it is inferred that the biodegradation of heavy hydrocarbons occurred in wells Ba-23 and Ke-19, resulting in the decrease of heavy hydrocarbon content, the increase of the carbon isotopes of heavy hydrocarbon, the carbon isotopic reversal between propane and butane, and the low carbon isotope of methane. The Baka oil field is the nearest to the main water supply area in the northern margin of the basin, and the faults are well developed. The burial depth of reservoir is shallow, so groundwater activity and surface water infiltration may destroy the oil/gas reservoirs, resulting in the secondary alteration of the hydrocarbon^[18]. The reservoir burial depth is the shallowest in Well Ba-23, 1 174 m-1 951 m, which is most vulnerable to the biodegradation caused by groundwater activity and surface water infiltration. According to the surface temperature of 20 °C and the geothermal gradient of 2.3 °C/100 m in the Taibei sag ^[40], the reservoir temperature is 47-65 °C in Well Ba-23. Considering the possibility of stratigraphic uplift in geological history, the reservoir temperature is completely suitable for bacterial activity and does not constitute a limiting factor. In addition, the carbon isotopic reversal is most obvious in propane, butane and pentane, and the carbon isotopic composition of methane is also lower than that of other exploration wells (Table 1, Fig. 3), which indicates that the natural gas has indeed been biodegraded. This understanding is consistent with the results of previous studies^[40,41].

The δ¹³C₁ value increases with increasing thermal maturity of source rocks, and there is a logarithmic linear correlation between δ¹³C₁ and R_o. However, the δ¹³C₁-R_o maturity model has a certain range of application, such as maturity range, location, types of the precursor^{[11, 23, 42]}. Although natural gas in the Taibei sag of Turpan-Hami Basin belongs to coal-derived gas, its maturity is relatively low, so this paper adopts the δ¹³C₁-R_o maturity formula by Dai and Qi^[23], Shen et al.^[22], Wang et al.^[11], Liu and Xu^[24] to calculate the gas maturity in the study area. The average values of the calculated R_o are 0.31%, 0.65%, 0.69% and 0.70%, respectively (Table 1). The δ¹³C₁-R_o relationship of Dai and Qi^[23] and Shen et al. ^[22] reflects the characteristics of coal-derived gas under long-term continuous evolution. The former mainly reflects the high thermal evolution stage, while the latter reflects the low thermal evolution stage^[24,43]. In the high thermal evolution stage, the carbon isotopic composition of coal-derived methane is heavier than that of oil-derived methane; However, it is pointed out that the δ¹³C value of coal-derived methane in low thermal evolution stage is not necessarily heavier than that of oil-derived methane^[22], indicating that the gas formation mechanism of coal measures in different thermal evolutionary stages may be different^[24]. Therefore, Liu and Xu^[24] put forward a two-stage carbon isotope fractionation model for coal-derived methane, that is, during the gas formation process of coal measure, the early stage is dominated by the degradation of aliphatic side chain, which results in the relatively low δ¹³C of coal-derived gas. While the late stage is characterized by the condensation of aromatic structure, which results in the relatively high δ¹³C of the coal-derived gas. The δ¹³C₁-R_o maturity formula by Wang et al.^[11] was obtained based on the natural gas in the Turpan-Hami Basin. The δ¹³C₁-R_o relationships by Wang et al.^[11] and Liu and Xu^[24] have similar R_o mean values, which are 0.69% and 0.70%, respectively. The δ¹³C₁-R_o relationship proposed by Wang et al.^[11] is mainly based on the natural gas in the Turpan-Hami Basin. Theoretically, the R_o value calculated by this relationship should be closest to the actual value. However, the accuracy of δ¹³C₁-R_o relationship depends largely on the statistical samples. For the δ¹³C₁-R_o relationship by Wang et al.^[11], the depth of gas wells calculated at that time was mainly shallower than 3 000 m, and gas wells with burial depth more than 3 000 m were few. To some extent, this may cause the R_o value calculated by such δ¹³C₁-R_o maturity formula to be lower than the actual value. Among the four sets of source rocks in the Turpan-Hami Basin, the Xishanyao Formation and Badaowan Formation of the Middle and Lower Jurassic are semi-deep lacustrine-fluvial swampy facies coal-bearing deposits, which are considered to be the main gas source rocks for the coal-derived gas in the Taibei sag^[3,11,13]. The present R_o value is 0.4%-0.9% for the top of Xishanyao Formation, and 0.6%-1.0% for the top of Badaowan Formation, which has reached the hydrocarbon generation threshold and has the condition for the generation of large amount of hydrocarbon^[11,13]. The dark mudstone in Xishanyao Formation is distributed all over the basin. The thickness of dark mudstone is 200-400 m for the Xishanyao Formation and 50-200 m for the Badaowan Formation. The thickness of coal seam is 40-60 m for both Xishanyao and Badaowan formations^[11,13]. The average TOC value and thermal hydrocarbon generation potential (S₁+S₂) of dark mudstone are 1.51% and 1.84 mg/g for the Xishanyao Formation, 2.08% and 3.79 mg/g for the Badaowan Formation, respectively. From the view of the whole basin, in the coal maceral, vitrinite accounts for 60%-80%, exinite accounts for less than 10%, and inertinite accounts for 10%-40%. The average TOC value and thermal hydrocarbon generation potential (S₁+S₂) of coal seams are 62.07% and 154.14 mg/g for the Xishanyao Formation, and 68.35% and 183.43 mg/g for the Badaowan Formation, respectively. On the whole, the TOC value and thermal hydrocarbon generation potential of coal measure source rocks of Xishanyao and Badaowan formations in the Taibei sag are obviously higher than the average value of the whole basin, and have good hydrocarbon generation potential^[13]. The two sets of source rocks are also well distributed in the study area, the hydrocarbon generation potential is large, and the thermal evolution degree is matched, so the two sets of source rocks are considered as the main gas source rocks of natural gas in the study area.

Of all the elements, the relative mass difference between the two stable isotopes of hydrogen (H: 99.985%, D: 0.015%) is the largest, resulting in the largest range of stable isotope ratios^[44,45]. The range of carbon isotope in methane from natural gas lies between -50‰ and -20‰, whereas that of hydrogen isotope in methane from natural gas can vary from -250‰ to -150‰^[46]. Therefore, because of the large variation range of hydrogen isotope composition, the variation increment of hydrogen isotopic composition is bigger than that of carbon isotope composition, and the response to the same environmental geochemical change is also more sensitive than that of carbon isotopic composition. Previous researchers have carried out a series of fruitful studies on the application of hydrogen isotopic composition in oil and gas^{[7-8, 10, 29, 46-50]}. It is pointed out that the hydrogen isotope composition of natural gas is affected not only by the thermal maturity of source rocks, but also the salinity of water medium. The δD value of biogas formed in marine and saline lacustrine environments is generally higher than -190‰^[7] or -200‰^[51]. However, the δD value of biogas formed in terrestrial freshwater environment is lower than -190‰^[7] or -200‰^[51]. The δD value of coal-derived methane has similar characteristics, which mainly depends on the properties of water medium, that is, with the increase of salinity in water medium, the δD value of coal-derived methane becomes heavier^[9]. The δD value of methane in the study area is low, less than -200‰. To some extent, the δD value of methane in the Baka oil field is heavier than that in Shanshan, Qiuling and Wenmi in general, higher than -260‰, which is similar to that in the Hongtai area. While the δD value of methane in Shanshan, Qiuling and Wenmi oil fields is similar to that in the Qiudong gas field, which is less than -260‰ (Fig. 8b). This is likely related to the formation environment of the Middle and Lower Jurassic coal measure source rocks in the Taibei sag, which may exist local salinization of water body^[52], and has little relation with the thermal evolution degree of source rocks, and the correlation between methane carbon and hydrogen isotopic compositions is not strong. Generally speaking, the formation environment of source rocks reflected by the hydrogen isotope composition of natural gas in the study area has the characteristics of terrestrial freshwater environment. For other similar areas in China such as Songliao Basin, hydrogen isotopes of its coal-derived methane are -257‰ - -217‰, indicative of the terrestrial freshwater-brackish water marsh coal-forming environment^[9]. The Middle and Lower Jurassic coal measure source rocks in the Turpan-Hami Basin are mainly freshwater lacustrine-swampy deposits, seawater transgression has not occurred, and local salinization of water bodies may have occurred only in the Baka area^[11,52]. Therefore, the hydrogen isotopic composition of methane (δD₁) in the study area is less than -250‰, which is lower than that of the coal-derived methane in the Songliao Basin.

Differences exist in the carbon and hydrogen isotopic compositions of methane from the Taibei sag. The carbon and hydrogen isotopes of methane in the Hongtai gas field are relatively heavier than those from the Qiudong, Shanshan, Baka, Qiuling and Wenmi oil/gas fields. However, the hydrogen isotopes of methane in the Baka gas field are similar to those in the Hongtai gas field. This is mainly because the thermal maturity of the coal measure source rocks of the Middle and Lower Jurassic in the Xiaocaohu subsag is higher than that in the Qiudong subsag. Therefore, the gas maturity in the Hongtai gas field is relatively higher, and its carbon and hydrogen isotopes are also relatively heavier^[3,11]. The hydrogen isotope of methane in the Baka oil field is similar to that in the Hongtai gas field, which is mainly related to the local salinization of water body during the formation environment of the Middle and Lower Jurassic coal measure source rocks in the Taibei sag^[11,52].

There is good correlation of carbon and hydrogen isotopes between methane and ethane in the study area. For example, excluding wells Ba-23 and Ke-19, the linear correlation coefficient (R²) is 0.7174 for δ¹³C₁-δ¹³C₂ and 0.8165 for δD₁-δD₂ (Fig. 8). This is due to the fact that the carbon and hydrogen isotopic compositions of methane and ethane gradually become heavier with the increase of thermal maturity of source rocks, which shows a linear correlation (maturity trend)^[26,40]. With the gradual increase of thermal maturity of source rocks, the difference of carbon and hydrogen isotopic composition between methane and ethane will gradually decrease, which has a linear correlation with the carbon and hydrogen isotopic composition of methane, respectively. Except wells Ba-23 and Ke-19, there is a good linear correlation between δ¹³C₁ and δ¹³C_2-1 (δ¹³C_2-1 means δ¹³C_C2H6-CH4) in the Shanshan, Qiuling, Baka, Wenmi, Hongtai and Qiudong areas (R² = 0.9126) (Fig. 9a). It shows that the difference of carbon isotope composition between methane and ethane becomes smaller and smaller with the increase of thermal maturity of source rocks. However, with the increasing thermal maturity of source rocks, the difference of hydrogen isotope composition between methane and ethane has not become smaller and smaller, and there is no correlation between them (Fig. 9b). If similar to the carbon isotope, thermal maturity of source rocks will be a major controlling factors on the hydrogen isotopes of natural gas, in this case, with increasing thermal maturity, there should be a linear correlation between δD_2–1 (δD_2–1 means δD_C2H6-CH4) and δD₁. In fact, good linear correlation only exists between δD₁ and δD₂ (R²=0.816 5) (see Fig. 8b), but not between δD_2-1 and δD₁ (R²=0.071) (Fig. 9b). So thermal maturity of source rock is an influencing factor of the hydrogen isotope of natural gas, but not the only one. It is found that the water medium conditions have a strong influence on the hydrogen isotope of natural gas^[7,51,53]. The parent organic matter formed in the marine or saltwater lacustrine environment has much heavier hydrogen isotopes than that formed in the terrestrial freshwater environment. In addition, because of the isotopic exchange reaction during the diagenetic process, the hydrogen isotope composition of the parent organic matter will also be affected by the water medium condition, but such influence is relatively small during the formation process of natural gas^[53]. This may be an important reason why there is no linear correlation between the difference of hydrogen isotope between methane and ethane and the thermal maturity of source rocks.

Although the hydrogen isotopic composition of natural gas can reflect the important characteristics of many geological processes, there may be a series of uncertainties in the interpretation of hydrogen isotopes in field geological samples. For example, isotope exchange with water^[54] and/or clay^[55], thermal maturation^[56,57], biodegradation^[58], washing and migration can seriously change the hydrogen isotope ratio. Based on the study of the natural gas from the Baka, Shanshan, Qiuling and Wenmi oil fields in the Taibei sag, it is proved that the hydrogen isotopes of natural gas in the study area are affected by both the thermal evolution of source rocks and the water medium condition of the formation environment of the source rocks in the study area. Hydrogen isotope of methane, can be used as an index to identify the water medium condition of the formation environment of source rocks. On the whole, methane which was from the source rocks formed in the continental freshwater lacustrine-swampy environment has relatively low hydrogen isotope, less than -250‰ in the study area. While methane from the source rocks which was formed in the marine or marine-terrestrial transitional environment generally has heavy hydrogen isotope. For example, the natural gas from Xujiahe Formation in the Sichuan Basin belongs to coal-derived gas from coal-bearing source rocks. However, its methane has relatively heavy hydrogen isotope (δD₁: -155‰ - -173‰), which may be related to the salinization under the background of marine-terrestrial transitional environment^[8].

Based on the compositional and isotopic analyses of the 23 natural gas samples from the Baka, Shanshan, Qiuling and Wenmi oil fields in the Taibei sag, Turpan-Hami Basin, combined with previous research results and regional geological background, it is pointed out that natural gas in the study area is dominated by alkanes, with a methane content of 65.84%- 92.84% and almost no non-hydrocarbon gases (N₂, CO₂). The mean value of dryness coefficient is 0.81, indicative of wet gas. According to the δ¹³C₁-R_o relationship, the gas maturity (R_o) is averaged at 0.7%. The δ¹³C₂ value of natural gas in the study area ranges from -28.2‰ to -24.9‰, indicating it is coal-derived gas with low maturity and is mainly from the Middle-Lower Jurassic coal-bearing source rocks. Methane and its homologues (C_2-5) basically have normal carbon and hydrogen isotopic distribution pattern (δ¹³C₁<δ¹³C₂<δ¹³C₃< δ¹³C₄<δ¹³C₅, δD₁<δD₂<δD₃), which are consistent with the typical isotopic characteristics of thermogenic gas and indicate that the gas has not undergone secondary alteration. However, the gas in Well Ba-23 and Well Ke-19 in the Baka oil field is biodegraded gas. The δD value of natural gas in the study area is light, less than -250‰, indicating that the source rock is formed in the continental freshwater lacustrine- swampy environment, and there is no transgression.