1 Introduction

About 61% of the coalbed methane (CBM) is in depths of 1000 and 2000 m in China (Li et al. 2018). Among China's CBM development areas, the Qinshui Basin is the most important (Qin et al. 2018). In the Qinshui Basin, the critical depth for shallow and deep coal is 650 m and 1000 m, respectively (Zheng et al. 2019). The permeability in Qinshui Basin is critically low, and the permeability of more than half of the layers in south Qinshui Basin is lower than 0.1mD (Zhao et al. 2016). Therefore, permeability enhancement measures are necessary. Liquid nitrogen freeze–thaw (LNFT) has become a hot topic for research in recent years as an environmentally-friendly permeability enhancement technology (Uliasz-Misiak et al. 2020; Liu et al. 2021a, b). However, there are significant differences between shallow and deep coal in terms of coal composition (Ao 2013), pore structure (Sun et al. 2015), and mechanical properties (Xie et al. 2019). Consequently, it is essential to reveal the difference in the impact of LNFT cycling on multistage gas flow between shallow and deep coal.

Scholars have conducted a certain number of researches on multistage gas flow in deep coal and achieved some significant results. Tao found that micropores (size less than 2 nm) play a controlling role in the adsorption/desorption of CBM under high pressure in deep coal seams (Tao 2019). Rong developed a deep coal permeability model considering damage under mining disturbance and verified the model using deep coal samples from the Pingdingshan mining area below 1000 m (Rong 2019). Zhang investigated the relationship between permeability and porosity of deep coal reservoir in the Shizhuang south and Yushe-Wuxiang blocks and found that the relationship between permeability and porosity in the reservoir can be described by an exponential function (Zhang 2019).

The multiscale pore structure is the space for multistage gas flow in coal. The multiscale structure evolution under LNFT has also attracted researchers’ attention. Zhao et al. investigated the impact of LNFT on water-bearing coal by metallographic microscopy and found that there are fracture formations and expansions, and there is a positive relationship between the pore fractures connectivity and water-bearing percentage (Zhao et al. 2018). Qin et al. tested the coal pore structure evolution during the LNFT cycling by mercury intrusion and nitrogen adsorption methods, and found a quadratic relationship between the cumulative pore volume and number of LNFT cycles (Qin et al. 2020). Akhondzadeh et al. studied the coal fractures evolution pre- and post-liquid nitrogen processing by 3D X-ray micro-computed tomography and found the microstructure connectivity is improved by liquid nitrogen treatment (Akhondzadeh et al. 2020).

As for the impact of LNFT on multistage gas flow in coal, some investigations have also been carried out. Wu compared the gas emission rate of coking coal and meager coal pre- and post-LNFT cycling and found that the gas emission rate of these two different rank coals is improved after the LNFT cycling, and the increment of the coking coal is greater than that of the meager coal (Wu 2017). Li et al. found that LNFT cycling can improve the Langmuir volume and has little effect on the adsorption rate (Li et al. 2021). Yan analyzed coal samples with different moisture contents and found that the gas adsorption capacity decreased after LNFT cycling under the same pressure, and the reduction range increased with increasing moisture content. Meanwhile, LNFT also improved the permeability of the samples, where the improvement was more significant with increasing moisture content (Yan 2019). Cai et al. compared the permeability and strength of coal pre- and post-liquid nitrogen cooling and found an increase in permeability of 48.89–93.55% and a decrease in compressive strength of 16.18–33.74% (Cai et al. 2015). Qin et al. investigated the impact of the LNFT on frozen coal mechanical properties and found a decrease in tensile strength of 81.3% and uniaxial compressive strength of 68.9%, respectively (Qin et al. 2022). Yang and Liu investigated the impact of LNFT on Illinois coal sorption and diffusion and found a positive correlation between the Langmuir volume and number of LNFT cycles, while the CH4 diffusion coefficient decreases as LNFT cycles increase (Yang and Liu 2020).

The development of CBM involves multistage gas flow in coal (adsorption/desorption–diffusion-seepage). Moreover, there are significant differences between shallow and deep coal in the composition and stress history, which can induce the differences in multiscale structure (micropores–mesopores–macropores/fractures) and mechanical properties. In literature reviews, it is evident that current studies have not analyzed the multistage gas flow evolution difference between shallow and deep coal due to LNFT. Therefore, this paper investigates the influences of coal composition and stress history on the multiscale structure evolution differences between shallow and deep coal during the LNFT cycling. The differences in the multistage gas flow evolution between shallow and deep coal were revealed from a multiscale structure evolution perspective.

2 Experimental

2.1 Sample preparation and coal composition analyses

Coal samples were obtained from coal seam No. 15 in the Qinshui coalfield in Qinshui Basin. The HL sample was from the Hengling block with a depth of 1436, and the SJZ sample was from the Sijiazhuang coalmine with a 420 m depth. In the Qinshui basin, the critical depth for shallow and deep coal is 650 m and 1000 m, respectively (Zheng et al. 2019). In this paper, the HL sample represents the deep coal, and the SJZ sample represents the shallow coal.

The samples used for porosity and permeability teats were processed into cylindrical sub-samples with a size of Ф25 × H50 mm. The samples were dried at 333 K for 24 h, then placed in a dryer to cool to room temperature prior to gas porosity and permeability measurements. In order to eliminate the impact of cracks on the experimental results of the gas sorption, the samples were crushed and sieved to particle sizes ranging from 60 to 80 mesh (0.18–0.25 mm). Before the gas adsorption and desorption experiment, the samples were vacuumed at 105 °C for 5 h to remove initially adsorbed gas and water.

Proximate analysis was conducted using an SDTGA 5000 proximate analyzer before analyzing the multistage gas flow evolution characteristics due to LNFT cycling. The test results are shown in Table 1. The content of volatiles and ash in the deep coal (HL) was higher than that of the shallow coal (SJZ), while the moisture and fixed carbon contents in the deep coal were lower than those in the shallow coal.

Table 1 Proximate analysis of the coal samples

The vitrinite reflectance of the samples was measured using a DMRT/MSP400/ROT-25 micro-photometer. The results are shown in Table 2. According to the vitrinite reflectance, both samples were anthracite.

Table 2 Vitrinite reflectance of coal samples

The maceral compositions of the samples were measured by a DMLD microscope. The results in Table 3 show that the inertinite content in the shallow coal is lower than that in the deep coal, while the contents of vitrinite and sapropelinite are higher than those in the deep coal.

Table 3 Maceral compositions of coal samples

The mineral compositions of coal samples were also analyzed using a Japanese physics TTR III multifunctional X-ray diffractometer. The results are listed in Table 4, which indicates that the mineral contents of the two coal samples are close.

Table 4 Mineral compositions of coal samples

2.2 Experimental study of the impact of LNFT cycling on the multiscale structure

The porosity of the shallow and deep coal samples was measured pre- and post-LNFT cycling using a KXD-II porosity and permeability joint tester manufactured by Jiangsu Lianyou Scientific Research Instrument Co., Ltd, China, as shown in Fig. 1. The porosity test range is 0.1–40%, and the error for porosity tests are ± 5%.

Fig. 1
figure 1

KXD-II porosity and permeability joint tester

During the porosity test, the coal sample was accommodated in the porosity measuring container. Then, the porosity measuring container was filled with a certain number of standard blocks and the porosity measuring container was installed. The gas supply and gas inlet valves were then opened and the pressure of the reducing valve was adjusted to 0.7 MPa. Once the pressure was stable, the gas inlet valve was closed and the porosity measuring valve was opened, and finally, after the pressure reached equilibrium, the equilibrium pressure was recorded.

The multiscale structure characterization and nuclear magnetic resonance imaging (NMRI) were conducted by the NMRI analyzer. Details about the NMRI analyzer and the experimental procedures are described in the authors’ other paper (Sun et al. 2021).

2.3 Multistage gas flow experiment in coal pre- and post-LNFT cycling

The permeability of the shallow and deep coal samples was measured pre- and post-LNFT cycling using a KXD-II porosity and permeability joint tester in Fig. 1. The permeability test range is 0.01–6000 mD and the error for permeability tests is ± 5%. During the permeability test, the ring pressure of the gripper was increased to 1 MPa after installing the coal sample, and the soap liquid was filled in the soap film flowmeter. Thereafter, the inner wall of the flowmeter was wet with the liquid soap. The air supply valve was then opened and the air supply pressure was slowly increased. Finally, when the soap film in the soap film flowmeter rose at a uniform speed, the rising time of the soap film and the inlet pressure value were recorded.

The gas adsorption–diffusion experiment was carried out using an adsorption analyzer which has been introduced in the authors’ other paper (Sun et al. 2020). The sketch map of the adsorption analyzer is shown in Fig. 2 (Sun et al. 2020). During the experimental process, the pressure was gradually increased to the preset maximum equilibrium pressure (3 MPa for both methane and carbon dioxide) at 303 K. The adsorption constants and diffusion coefficients of methane and carbon dioxide were determined during the adsorption period when gas pressure was gradually improved. Then, the adsorption constants and diffusion coefficients of methane and carbon dioxide were determined during desorption when the gas pressure was gradually reduced (as shown in Fig. 3).

Fig. 2
figure 2

Schematic of H-Sorb 2600 high-pressure volumetric adsorption analyzer (Sun et al. 2020)

Fig. 3
figure 3

Gas sorption amount over time during the adsorption and desorption period

3 Results and discussion

3.1 Evolution of multiscale structure during LNFT cycling

Porosity can not only comprehensively characterize coal structure but also control the permeability. In this study, the porosity evolution was first investigated during LNFT cycling. The porosity was measured by NMR and gas methods. In the NMR method, the linear correlation between the NMR signal and water mass contained in the pore structure was employed to measure the samples’ porosities (Sun et al. 2021). In the gas method, helium was used for porosity measurements.

Figure 4 presents the evolution of porosity measured by two methods during the LNFT cycling. Overall, the porosity tested by gas is larger than that tested by NMR. When the porosity is measured by NMR, due to the surface tension, water cannot enter the small pores, resulting in underestimating porosity. Although the porosity may be underestimated, there are advantages to the NMR method in estimating the evolution of pore structures on different scales (micropores, mesopores, and macropores).

Fig. 4
figure 4

Evolution of porosity measured by different methods during LNFT cycling

As shown in Fig. 5, the initial porosity of deep and shallow coal is 4.25% and 4.18%, respectively. According to the tests on 82 samples from south Qinshui Basin, the porosity of coal ranges from 2.91% to 10.74%, 5.43% on average (Zhao et al. 2016), which is close to the porosity results in this study. Therefore, the porosity results are believable. The larger initial porosity in deep coal can be explained from two aspects. On the other hand, the pores are more developed in the inertinite group than in the vitrinite group (Duan et al. 2009). According to the proximate analysis, there is more inertinite group in the deep coal. On the other hand, before sampling, the deep coal is in high in-situ state induced by the larger burial depth and the stress release caused by deep coal sampling can lead to internal damage of coal samples (Wang 2009).

Fig. 5
figure 5

Evolution of porosity measured by gas during LNFT cycling

As shown in Fig. 5, the porosity of the shallow and deep coal measured by gas increases during LNFT cycling, conforming to a quadratic relationship. Qin et al. also found a quadratic relationship between the cumulative pore volume and the LNFT cycles number (Qin et al. 2020). Chu et al. found the total energy, elastic energy, and dissipated energy of the coal samples decrease with the increase of LNFT cycles (Chu et al. 2022), which can explain the impact of LNFT on coal structure is gradually weakening with the increase of LNFT cycles. The strength of the coal matrix increases with the content of the inertinite group and decreases with the content of the vitrinite group (Zhang 2019). According to the maceral compositions analysis, there is more inertinite group content in the deep coal. As a result, the porosity of the shallow coal is more sensitive to the LNFT cycles, and the porosity of the shallow exceeds the deep coal after four cycles. Figure 6 shows that the pore structure evolution in deep coal is the expansion of fracture and the pore evolution mainly happens around the fracture. The pore structure evolution in shallow coal is mainly the pores connectivity enhancement in pore structure. The difference in the pore structure evolution is also related to the strength difference in coal matrix caused by the vitrinite and inertinite group content. In particular, the lower coal matrix strength in shallow coal induces matrix failure in a large area and enhances the pores connectivity. The higher coal matrix strength in deep coal causes the stress concentration around the initial microfracture and expands the fracture.

Fig. 6
figure 6

Nuclear magnetic resonance imaging pre- and post-LNFT (Sun et al. 2021)

NMR can obtain the pore structure evolution in different scales (micropores, mesopores, and macropores/fractures). As for the initial proportion of micropores and macropores, deep coal is higher than shallow. Previous researches also indicated that the structure of deep coal is denser (Zhang et al. 2019). Li also found the increase in the proportion of macropores with the increase of burial depth (Li et al. 2014). As a kind of sedimentary rock, the influence of the geo-stress caused by the burial depth on the coal is twofold. During the deposition process, the higher stress causes the coal is denser in structure. After diagenesis, it is in a higher stress state (Wang 2009). Stress release caused by deep coal sampling can lead to internal damage of coal samples because of the release of the accumulated energy in the high stress in deep coal (Wang 2009). Consequently, the macropores in deep coal are more developed. From Fig. 7, it can be also found that the volume proportions of macropores and mesopores are increasing with the LNFT cycles, while the volume proportions of micropores are decreasing, indicating more micropores are developed into mesopores and macropores. Other research also found the increase in the macropores and mesopores proportions and the decrease in the micropores proportions of during LNFT cycling (Chu and Zhang 2019). The larger proportion of macropores (fractures) in deep coal creates the condition for fracture expansion. Due to the larger proportion of macropores (fractures) induce by the stress release in deep coal sampling, other researchers found the lower tensile strength in deep coal (Xie et al. 2019; Zhang et al. 2021). The conclusion can be drawn that the fracture expansion is dominated in deep coal because of the higher proportion of macropores (fracture) and the lower tensile strength related to the initial fractures in deep coal. The enhancement of pores connection is dominated in shallow coal because the strength of the shallow coal matrix is lower, and the pore structure can be connected more easily. It can be found from the above discussion that the multiscale structure evolution differences between deep and shallow during LNFT cycling is comprehensively impacted by the coal compositions and stress history. Subsequently, the multiscale structure evolution further affected the multistage gas flow in coal.

Fig. 7
figure 7

Evolution of multiscale structure during LNFT cycling

3.2 Evolution of gas seepage during LNFT cycling

In order to reveal the gas seepage characteristics evolution in shallow and deep coal during LNFT cycling, the permeability of the samples was measured with helium pre-LNFT and after the 2–4–6–8 10 cycles, as shown in Fig. 8. It is evident that the permeability of deep coal (0.10454 mD) is slightly higher than that of shallow coal (0.09183 mD) pre-LNFT cycling, which can be attributed to the higher initial porosity and proportion of macropores in deep coal. According to the tests on 39 coal layers in the south Qinshui Basin, the permeability ranges from 0.01 to 0.92 mD, 0.19 mD on average (Zhao et al. 2016), which verifies the permeability measurement results in this study are believable.

Fig. 8
figure 8

Evolution of permeability during LNFT cycling

As shown in Fig. 8, the permeability of shallow and deep coal increases during LNFT cycling. Su et al. found that the relationship between the permeability and LNFT cycles can be fitted by the Eq. y = a + b ln(x) (Su et al. 2020). In this study, it can be found that the correlation between permeability and LNFT cycles number can also be well fitted by the Eq. y = a + b ln(x). The permeability of shallow coal exceeds that of deep after two cycles, while the porosity of shallow coal exceeds that of deep coal after four cycles. It indicates that permeability is more sensitive to LNFT than the porosity. As found in the pore structure evolution, fracture expansion is dominant in deep coal and pores connectivity enhancement is dominant in shallow coal. It also indicates that the permeability improvement is more significant caused by pore structure connectivity enhancement in soft coal than that caused by fracture expansion in hard coal.

In order to further validate that the pores connectivity enhancement is more significant than fracture expansion in improving the permeability. The permeability of shallow and deep in the same porosity was compared. Moreover, as previously seen in Fig. 4, the differences are evident in the coal porosity measured by different methods during LNFT cycling. It is also necessary to study the relationship between the sample permeability and the porosity measured by different methods. Figure 9 presents that the permeability of shallow and deep coal increases with the increase of porosity and the permeability of deep coal is lower than that of shallow coal for the same porosity. This further validates that the pores connectivity enhancement is more significant than fracture expansion in improving the permeability. Figure 9 also indicates that the relationship between permeability and porosity measured by gas can fit an exponential function well. Zhang found an exponential relationship between permeability and porosity of deep coal gas reservoirs in Shizhuang south block and Yushe-Wuxiang block in #15 coal seam in the Qinshui coalfield (Zhang 2019). Zou et al. also found the exponential relationship between permeability and porosity in coal samples from the Southern Junggar Basin (Zhou et al. 2016). However, the relationship between the permeability of shallow coal and porosity measured by NMR cannot be fitted well by an exponential function. Therefore, attention should be paid to the porosity measurement method, when using it to predict permeability. The relationship between permeability and porosity in Fig. 9 was attempted to be linearly fitted, but it was found that the coefficients of determination in HL curve in Fig. 9a and SJZ Fig. 9b are both lower than 0.8. In this study, only two samples were investigated, and the relationship between the permeability of porosity should be more valuable if more samples are tested in the future study.

Fig. 9
figure 9

Relationship between permeability and porosity measured by different methods

3.3 Evolution of gas adsorption–diffusion during LNFT cycling

The Langmuir isothermal model was used to determine the adsorption isotherm in this study (Langmuir 1918). Figure 10 shows the experimental and Langmuir isotherms of CH4 and CO2 in the adsorption and desorption stages. It can be found that CO2 presents a higher adsorption capacity than CH4.

Fig. 10
figure 10

Experimental and Langmuir isotherms pre- and post-freeze–thaw cycling

The Langmuir volume (VL) and the Langmuir volume (PL) are quantitative parameters to evaluate coal adsorption characteristics. Table 5 shows the Langmuir constants (VL and PL) of CH4 and CO2 in the ad/de-sorption stages of shallow and deep coal pre- and post-LNFT cycling. According to the statistics in the south Qinshui Basin, the Langmuir volume of CH4 in coal ranges from 26.58 to 44.90 ml/g, with an average of 37.02 ml/g (Zhao et al. 2016). In this study, the Langmuir volumes of CH4 for both shallow and deep coal fall in the range, indicating that the Langmuir volume results are appropriate.

Table 5 Langmuir constants

It can be found in Table 5 that the VL of deep coal is smaller than that of shallow coal for both pre- and post-LNFT cycling. The Langmuir volume of both shallow and deep coal increases after cycling. From the previous research, the gas adsorption capacity decreases with the ash content increase (Sun et al. 2020), which can explain that the VL of deep coal is smaller than that of shallow coal for both pre- and post-LNFT cycling. In general, the VL alteration of shallow and deep coal are close, within the range of 4.09–6.49%. The increase in VL is mainly induced by the increase in porosity. The similarity in the porosity and proportion of micropores/mesopores of two coals in the initial and at the end of LNFT cycles can explain why the Langmuir volume variation range is close.

The Langmuir pressure variation of deep coal is more significant than that of shallow coal, varying between − 7.85 and 21.47%, and the Langmuir pressure change of shallow coal is between 0.05 and 5.02%, indicating that the pore surface variation is more significant in deep coal. However, the Langmuir pressure of CH4 in deep coal decreases after LNFT cycles, which can be explained by the larger molecule size of CH4 and the pore collapse in deep coal because of the high ash content. This will be explained in further detail in the diffusion section. The improvements in the Langmuir pressure in both coals indicate that the gas adsorption surface in the coal matrix becomes smoother (Avnir and Jaroniec 1989).

Previous researches generally apply the unipore and bidisperse models to calculate the gas diffusion coefficients in coal (Cervik 1967; Crank 1975; Clarkson and Bustin 1999; Stewart and Hessami 2005; White et al. 2005). The unipore model is simpler for calculation (Cui et al. 2004). Although it does not fit the experimental data perfectly, it is considered sufficient to provide a first-order approximation of the gas diffusion coefficient (Pillalamarry et al. 2011; Liu et al. 2015). The details about the gas diffusion coefficients calculation have been described in the authors’ other paper (Sun et al. 2018). Figure 11 presents the CH4 and CO2 diffusion coefficients in shallow and deep coal during the adsorption and desorption.

Fig. 11
figure 11

Gas diffusion coefficients for CH4 and CO2 pre- and post-freeze–thaw cycling

As shown in Fig. 11, LNFT cycling significantly affects the diffusion coefficient of shallow and deep coal. Although the CH4 diffusion coefficient decreases after LNFT cycles, the CH4 and CO2 diffusion coefficients in shallow and deep coal are improved by LNFT cycles, which can be explained by the development of the porosity and the proportion of mesopores during the LNFT cycling. The gas diffusion coefficients are mainly in the magnitude of 10–12 m2/s, similar with the gas diffusion coefficients reported by other researchers (Busch et al. 2004; Charriere et al. 2010). In addition, CO2 presents a higher diffusion coefficient than CH4, which can be explained by the small kinetic diameters (CO2: 0.33 nm; CH4: 0.38 nm) (Busch et al. 2004). Cui et al. also found the diffusion coefficient of CO2 is one or two order of magnitude greater than those of CH4 and N2 and they also explained the greater diffusion coefficient of CO2 by the smaller kinetic diameter of CO2 (Cui et al. 2004). CH4 and CO2 diffusion coefficients in desorption are higher than those in adsorption, which can be explained by the smoothness of the pore surface built-up by multiple layers of adsorbed gas molecules (Yang and Liu 2020).

The CH4 and CO2 diffusion coefficients in deep coal are smaller than that in shallow coal pre- and post-LNFT cycles, which can be explained by the more complex pore surface in deep coal. As mentioned in the Langmuir pressure section, the more significant variations in Langmuir pressure also indicate that the pore surface in deep coal is more complex. Sun et al. also found that the pore surface is more complex in deep coal than in shallow coal (Sun et al. 2015). The drop of CH4 diffusion coefficient in deep coal is also corresponding to the Langmuir pressure drop after the LNFT cycles. The Langmuir pressure and diffusion coefficient drop indicate the deterioration in pore surface irregularity (Yang and Liu 2020). Yang and Liu also found that there is a positive correlation between the Langmuir volume and the number of LNFT cycles, while the CH4 diffusion coefficient decreases with the increasing LNFT cycles, which was explained by the increase of pore surface roughness induced by the LNFT cycles (Yang and Liu 2020). Liu et al. found that LNFT can cause the fractures and pores to collapse, and the scattered coal scraps may clog the macropores and fractures. Lin et al. also found LNFT can cause the collapse of pore surface (Lin et al. 2021). Besides, the pores and fractures collapse also causes the pores' surface to become rougher (Liu et al. 2021a, b). Coal is a typical composite material. In terms of its structure, it can be simply regarded as composed of mineral impurities and mineral particles. Thermal stress can be induced by the temperature gradient between adjacent mineral particles with heterogeneity features (Hou 2022). When there is a big difference between the hardness of internal mineral impurities and the strength of coal particles, the collapse mainly occurs at the junction of impurity particles and coal particles (Gao 2010). If there is more ash content, pores collapse can happen more easily. The CH4 Langmuir pressure and diffusion coefficient drop in deep coal is related to the higher ash content in deep coal. No decrease in CO2 diffusion coefficient is related to the smaller size of CO2. The smaller size of CO2 causes CO2 to not be sensitive to the collapse of pores. The adsorption-induced deformation may also explain the CH4 diffusion coefficient drop in deep coal. The adsorption-induced deformation is also related to the stress-history, and the high stress in history can promote the adsorption-induced deformation (Bergen et al. 2011). In Qinshui Basin, there is a positive correlation between the in-situ stress (the vertical stress (overburden stress), the minimum horizontal stress and the maximum horizontal stress) and the burial depth (Meng et al. 2011). Authors investigated the pore structure evolution difference between deep and shallow coal during gas adsorption using synchrotron radiation SAXS, and found there is a decrease in average pore diameter in deep coal, as presented in Fig. 12. The decrease in deep coal pore size can explain the CH4 diffusion coefficient drop in deep coal. The deep coal in Fig. 12 were collected from the depth of 1638 m in Wuxiang south block in No.15 seam at Qinshui coalfield located at Qinshui Basin and the shallow coal is same as the shallow coal used in this paper. The gas used is CO2 and the adsorption equilibrium pressure is 1 MPa. In sum, the evolution of gas diffusion in deep and shallow coal is also controlled by the coal compositions and stress history.

Fig. 12
figure 12

Average pore diameter evolution in deep and shallow coal during gas adsorption

4 Conclusion

It is of great significance for CBM development to reveal the difference in the impact of LNFT cycling on the multistage gas flow between shallow and deep coal. There are significant differences in coal composition and stress history between the shallow and deep coal, which can cause the differences in the multiscale structure and mechanical properties. As a result, in this study, the composition differences were first analyzed from the aspects of proximate, maceral, and mineral compositions. The gas and NMR methods characterized the multiscale structure of shallow and deep coal. The permeability, gas adsorption constants and diffusion coefficients were measured pre- and post-the LNFT cycles. Based on the differences between shallow and deep coal in the compositions and stress history, the differences in the impact of LNFT cycling on the multistage gas flow between shallow and deep coal were revealed from the perspective of multiscale structure evolution.

  1. (1)

    The gas-measured porosity of shallow and deep coal increases during the LNFT cycling, following a quadratic relationship. The initial porosity of shallow coal is lower than that of deep coal because of the higher inertinite group content in deep coal and the stress release induced-fractures, but the porosity of deep coal is exceeded by shallow coal after four LNFT cycles. The impact of LNFT cycling on coal multiscale structure is fracture expansion for deep coal and pores connectivity enhancement for shallow coal, which is caused by the lower matrix strength in shallow coal and the stress release induced-fractures in deep coal.

  2. (2)

    The initial permeability of shallow coal is lower than that of deep coal because of the higher initial porosity and proportion of macropores in deep coal. The permeability of both shallow and deep coal increases during the LNFT cycling, and a logarithmic correlation exists between permeability and cycle number. The relationship between the permeability of both shallow and deep coal and porosity measured by gas fits an exponential function. The permeability of deep coal is lower than that of shallow coal with the same porosity, indicating that pore connectivity enhancement is more significant than fracture expansion in permeability improvement.

  3. (3)

    The Langmuir volume of deep coal is smaller than that of shallow coal in both pre- and post-LNFT cycling. After LNFT cycling, the Langmuir volume of both shallow and deep coal increases, and the increasing range is relatively close between 4.09 and 6.49%. The variation range of Langmuir pressure of shallow coal is less significant than that of deep coal.

  4. (4)

    The diffusion coefficient in deep coal is smaller than that in shallow coal in both pre- and post-LNFT cycling. Except for CH4 in deep coal, the Langmuir pressure and gas diffusivity in both coals is improved. the Langmuir pressure and diffusion coefficient of CH4 in deep coal decrease after LNFT cycles, which can be explained by the larger molecule size of CH4, the pores collapse induced by high ash content and the adsorption-induced deformation induced by the high history stress.