In this paper, 33 ChinArray II stations were selected to image the structure in the crust of the study area, and the observed cross-section spreads nearly east‒west, including parts of the orogenic belt at the northern margin of the West Qinling, the Linxia block, the Longzhong block, the Liupanshan tectonic belt, and parts of the southwestern margin of the Ordos block. The 33 broadband seismic stations had an average interstation spacing of ∼30 km, and all stations were equipped with a data acquisition system (Reftek 130) and a three-component broadband seismometer with a natural period of 60 s (Güralp CMG-3EPSC) or 120 s (Güralp CMG-3 T).

P-wave receiver functions (PRFs) have become a relatively well-used method for studying crustal structures (Xu and Zhao, 2008; Chen et al., 2022). The PRF is extracted by separating the time series of the vertical component and the horizontal component from the records of the three-component broadband seismometer, and the vertical component and the horizontal component are deconvoluted to obtain a time series of the PRF that is well documented about the structure beneath the station (Langston, 1979; Ammon et al., 1990). The crustal structure below the station has a certain influence on the P-to-S conversion wave (Ps), so the Ps wave also records the structural information in the crust. Therefore, the PRF can be used to study the Moho structure, crustal structure, crustal thickness, and Poisson’s ratio; the PRF can even be used to image the crust–mantle structure (Langston, 1979; Tian et al., 2005) and can also invert the crustal velocity structure (Liu et al., 1996). The current methods for extracting receiver functions mainly include iterative deconvolution in the time domain (Owens and Zandt, 1997) and deconvolution in the frequency domain (Langston, 1979). On this basis, the structure under the station is obtained through different technical methods.

The extraction of the PRF includes the following steps: first, download the earthquake catalogue combined with the station deployment time, select earthquake events according to the earthquake catalogue, convert the format of the original data, and intercept and select the seismic data at the same time. We select 439 seismic events of Ms ≥ 5.5 and epicentral distances between 30° and 90° that were recorded during the working hours of the array. The waveform data of the 20 s before and 100 s after the arrival of the P wave are intercepted (Figure 2) and then converted into sac format. Preprocessing, such as detrending and demeaning, is performed on the intercepted teleseismic waveform data. Then, we manually select the preprocessed data, and the data selection process is applied to CrazySeismic software (Yu et al., 2017). After the preliminary selection, the three-component event waveforms are bandpass-filtered (two-pole Butterworth filter, with a corner frequency =0.05–2 Hz) and rotated from Z-N-E into Z-R-T components. Finally, based on the theory of equivalent seismic sources (Langston, 1979), the time domain deconvolution algorithm is used for calculation (Ligorria and Ammon, 1999), and the Gaussian filter factor is set to 2.5. Then, the extracted receiver function is further selected, and the data with a significant seismic phase of the P wave, clear seismic phase, and good consistency of Ps and PpPs are selected for the next imaging analysis. The receiver function results of stations 62,350 and 63,056 are shown in Figure 3, and the specific locations are marked in Figure 1. The results of the receiver function show that the quality of each available receiver function is good, the waveform information is clear, the seismic phase characteristics are clear, and the seismic propagation characteristics of the area are thus combined.

The common conversion point stacking method (CCP) of the receiver function is developed based on the high-frequency ray approximate superposition theory. The high-frequency ray superposition method is the earliest-developed array receiver function imaging method (Chen et al., 2022). Dueker and Sheehan (1997) took the lead in introducing the idea of common midpoint superposition (CMP) of reflected waves into converted wave imaging of receiver functions. Through the given velocity model, the receiver function of oblique incidence is dynamically corrected and superimposed according to the relationship between the theoretical travel time and the incident angle, and finally, the interface fluctuation image under the array is obtained. Zhu and Kanamori (2000) made a further improvement on this basis and proposed the receiver function common conversion point superposition method (CCP). According to a given velocity model, the CCP superposition method back-projects each receiver function into the space domain according to the ray path and performs superposition imaging in the space domain (Zhu and Kanamori, 2000; Chen et al., 2022). The CCP superimposed imaging method has the advantages of high computational efficiency and a good horizontal medium imaging effect and has obvious advantages in revealing the interface between the crust and the upper mantle with broadband station data.

We used the Common Conversion Point (CCP) stacking method (Zhu and Kanamori, 2000) to image the crustal architecture beneath profile AA’ (Figure 1B). The amplitudes of the RF waveforms were projected along the ray path of the conversion points based on a reference velocity model. The IASP91 model was selected for the time-depth conversion background velocity, and we fine-tuned the shallow velocity model according to the research results of previous researchers in the study area. In this paper, the three-dimensional P-wave and S-wave velocity models of the crust in the study area obtained by Ding et al. (2017) using the tomography method are selected to replace the velocities shallower than 50 km in the IASP91 model. The volume of the crust beneath the linear array was divided into bins with the sizes being 100 km wide, 35 km long, and 0.1 km thick. All the amplitudes in the same bin were stacked, averaged, and horizontally smoothed, and then the crustal structure was normalized with the amplitude distributions (Figure 6C).

According to the theory of earthquake propagation, we know that the teleseismic P wave propagates from the mantle to the crust below the observation station, part of the P wave is projected there, and part of it is converted into an S wave and continues to propagate. The receiver function method is used to image the structure in the crust by enhancing the signal of this type of converted wave. In addition, the above two types of waves also undergo multiple reflections and propagations between the crust and the Moho interface, forming multiple waves, such as PpPs, PpSs, and PsPs. The expression of earthquake propagation is as follows: t P = H 1 V P 2 − p 2 , t S = H 1 V S 2 − p 2 , (1)

Here, t _P and t _S in Formula 1 represent the travel times of the P wave and S wave in the layer with thickness H, respectively; V _P and V _S represent the velocities of the P wave and S wave, respectively; p represents the ray parameter, which is p = sin ⁡ i V P ; and i represents the incident angle of the P wave on the surface. According to Formula 1, the calculations of the travel times of various converted waves relative to the direct P wave can be obtained. The mathematical expression is as follows: t P s = H 1 V S 2 − p 2 − 1 V P 2 − p 2 , t P p P s = H 1 V S 2 − p 2 + 1 V P 2 − p 2 , t P s P s + P p S s = 2 H 1 V S 2 − p 2 . (2)

Formula 2 shows that the wave velocity ratio κ (Vp/Vs) is more sensitive to the crustal thickness than Vp, so to estimate the crustal thickness as accurately as possible, Zhu and Kanamori (2000) proposed the method of H-κ stacking. This method involves scanning and calculating H and κ within a certain range, bringing the searched H and κ into Formula 2 for calculation, and bringing the calculation results into objective function (3) for the solution: S H , κ = w 1 r t P s + w 2 r t P p P s − w 3 r t P s P s + P p S s , (3)

Through the maximum value of the objective function obtained in Formula 3, the corresponding average crustal thickness and Vp/Vs ratio under the station can be obtained. By knowing the result of the Vp/Vs ratio, we can further obtain Poisson’s ratio according to Formula 4. There is a linear relationship between the wave speed ratio and Poisson’s ratio. The mathematical expression is as follows: δ = 0.5 − 1 2 κ 2 − 1 (4)

In the calculation process, we usually set the speed of the P wave as 6.3 km/s (Chen et al., 2005); the search interval of H is generally set to 20–70 km; the search interval of the κ value is generally set between 1.5 and 2.3; and the weighting coefficients of Ps, PpPs, and PpSs+PsPs multiples are set to 0.7, 0.2, and 0.1. Previous experience tells us that the value of the weighting coefficient has little effect on the calculation results (Tian et al., 2021).

Figure 4 shows the H-κ results under different stations in the study area. The figure shows that the weighted estimation point convergence effect is very good (green pentagonal heart), and the feasibility of the data is high.

In this paper, the receiver functions of 33 stations in the study area were extracted by the above method. We used the iterative deconvolution method in the time domain (Ligorria and Ammon, 1999). This method has the advantages of stable calculation and clear receiver function waveforms. To suppress the influence of noise in the waveform data, the Gaussian filter factor is set to 2.5 in the receiver function calculation. Referring to previous research results (Gao et al., 2006; Li et al., 2006; Tong et al., 2007; Zhang et al., 2015; 2016; Tian et al., 2021), 6.2 km/s was selected as the average crustal velocity for superposition analysis, and the weighting coefficients of PpPs, PpSs, and PsPs multiples were set to 0.7, 0.2, and 0.1. We used the H-κ stacking method to obtain the crustal thickness (H) and the Vp/Vs ratio (κ) below each station. Due to the interference of different factors, such as data quality, venue conditions, and underground structure, the signal of the collection within some periods of individual stations is not good, for example, the data from station 62,363, the data from station 61,057 in 2015, and the data from stations 62,348 and 62,349 in 2014. Because the signal-noise-ratio of these stations is poor, they were discarded directly when the receiving function was selected. Because there are many stations with better PRFs surrounding them, the deep structure of the research area is very good and has a high degree of credibility. Through efficient data selection by CrazySeismic software (Yu et al., 2017), PpPs, PpSs, and PsPs data with clear seismic phases and good theoretical travel times are selected, and the number of receiving functions obtained under each available station is approximately 150–280. The number of effective receiving functions under most stations is large, which ensures the reliability and representativeness of the results. Supplementary Table S1 shows the obtained H-κ results, and Figure 4 shows the H-κ superposition results of the four stations. The green pentagon in Figure 4 represents the best-estimated values of H and κ, the convergence effect is good, and there is no multiextreme value, indicating that the reliability of the data is high, and the station is marked by the numbers in the lower left corner and corresponding H and κ results. The average crustal thickness in the study area is 48.65 km, and the average Poisson’s ratio is 0.25. From the northern margin of the West Qinling Fault to the southwest margin of the Ordos block, the Moho depth gradually changes from 57.6 km to 43.5 km, and the Moho below the Liupanshan fault zone has local depression characteristics. Previous survey results show that the thicknesses of the crust in the Liupanshan structural belt are approximately 45–50 km and has a high Poisson’s ratio, while the thicknesses of the crust inside the Ordos block are approximately 42–43 km (Chen et al., 2005; Tong et al., 2007; Wang et al., 2014; Wang et al., 2017; Tian et al., 2021), which are consistent with our calculation results. The H-κ stacking results (Supplementary Table S1) and CCP stacking profiles (Figure 6) show that the crustal thickness in the study area gradually thins from the West Qinling fault to the southwestern margin of the Ordos block, but there is an obvious deformed depression feature in the Moho below the Liupanshan structural belt (Tong et al., 2007). In summary, our research results are consistent with previous research results in this area.

From Figure 6, we can conclude that the altitude gradually decreases from the West Qinling fault zone to the southwest margin of the Ordos block, and the Moho depth gradually decreases from 57.6 km to 43.6 km. The H-κ results (Supplementary Table S1) also correspond well. The Bouguer gravity anomaly has a good coupling relationship with the basin–mountain structure along the survey line. The red solid line in Figure 6B represents the Bouguer gravity anomaly curve in the study area, the numbers on the left represent the Bouguer gravity anomaly, and the numbers on the right represent the conductivity values. The black dotted line represents the average value of the Bouguer gravity anomaly. In general, the Bouguer gravity anomaly changes from small to large from west to east, and the lower part of the mountain range shows a relatively negative anomaly. There is good correspondence between the Qinling and the lower part of Liupanshan structural belt. The results of magnetotelluric inversion show that there are structural differences in the electrical conductivity between the Linxia block and the Longzhong block (Xin et al., 2021), as shown by the blue line in Figure 6B. The Linxia block has high conductance, while the Longzhong block has low conductance. The rate of conductivity change in the Maxianshan fault is the highest, and the above evidence indicates that the Maxianshan fault is the suture line between the Linxia block and the Longzhong block.

Through the analysis of CCP sections (Figure 6C), we believe that the study area has obvious difference between adjacent blocks. From west to east is the Qinling orogenic belt, Linxia block, Longzhong block, and Ordos block. The layered interface below the Qinling orogenic belt is poor, and the amplitude in the crust is weak (Figures 5, 6C), which may be related to the tectonic movement of the Northwest Qinling Fault (NWQLF) that has undergone multiple periods of strike-slip and thrusting structural deformation (Zhang and Ding, 1996; Cheng et al., 2003; Zhan et al., 2004), and complex tectonic activities destroyed the inner shell structure. The layered interface within the crust of the Linxia block between the Northwest Qinling Fault (NWQLF) and the Maxianshan Fault (MXSF) is visible (Figure 6). Between 102.8° and 103.9°, there are two continuous positive seismic phase boundaries at depths of ∼18 km and ∼32 km, and we speculate that the shallow normal seismic phase boundary is the Conrad boundary (CR). There are two obvious negative seismic phase boundaries above and below the CR, which are located at depths of ∼8 km and ∼25 km, respectively.

The crustal structures of the Longzhong block and the Linxia block are significantly different. There is a continuous negative seismic phase in the upper crust of the Longzhong block approximately 10 km away, it enters the southwest margin of the Ordos block through the Liupanshan Fault and the Xiaoguanshan Fault, and the CR is not developed. A weaker positive seismic phase develops at a depth of approximately 25 km, and a weaker negative seismic phase develops in the lower crust. The lateral interface between the Longzhong block and the Linxia block shows an obvious discontinuous feature, and there is an obvious negative seismic phase transition from west to east that shallows from 20 km to 10 km, which may be the deep manifestation of the Maxianshan fault as a boundary fault in the middle and upper crust. Previous receiver functions and their linear inversion results suggest that there is a dividing line between the Qinling Mountains and the Qilian Mountains at 105°E in the Longxi Basin (Liu et al., 2008b); however, it is evident only from the S-wave velocity, and the deep distribution of this boundary is not proposed. We believe that the continuous west-dipping negative seismic phase (104°–104.5°) below the Maxianshan fault zone in the CCP imaging results is the location of the ancient suture line between the Longzhong block and the Linxia block. This conclusion is consistent with the research results of magnetotelluric (MT) and deep seismic reflection profiles (Guo et al., 2016; Xin et al., 2021). There is a continuous negative seismic phase boundary with a length of ∼250 km at a depth of 10 km between the Maxianshan fault and the Xiaoguanshan fault, which may be deep evidence of the local overall uplift of the Longzhong block and its extension to the Ordos block. Because the surface along the southeast of the Maxianshan Fault in the study area is covered by thick and near-horizontal sedimentary layers and the faults are not developed, the research level of this area is relatively low, but the Cenozoic strata have obvious structural characteristics of overall uplift (Guo et al., 2018a), so our CCP imaging results have a good correspondence with the geological phenomenon of regional uplift.

There is an obvious discontinuity of the transverse velocity interface under the Liupanshan structural belt. The negative seismic phase at ∼10 km shows a tendency to expand eastwards; there is a discontinuous positive seismic phase boundary at ∼15 km depth; and the negative seismic phase at a depth of ∼18 km is relatively continuous in the Ordos block (Figure 6C). The layered structure in the deep part of the Ordos block is better, the crustal deformation is weaker, and it has the characteristics of an ancient cratonic base. There is an obvious positive seismic phase at ∼15 km, which we believe is the Conrad boundary (CR) of the Ordos block. There is a continuous Conrad interface of approximately 18–20 km in the northwest margin of the Ordos block (Chen et al., 2020), which reflects the characteristics of a stable craton with good stratification in the Ordos crust (Chen et al., 2005; Tian et al., 2021; Chen et al., 2022). The velocity discontinuity below the Liupanshan fault zone shows a discontinuous west-dipping trend, which may be controlled by the multiple thrust nappe structures of the Liupanshan fault; the deep structure is relatively broken and is characterized by a west-dipping and eastwards thrust towards the Ordos block.

Poisson’s ratio (σ) is one of the important parameters for studying the crustal structure, its range of variation is related to the rock composition and type, and the continental crustal material composition is related to Poisson’s ratio (Christensen, 1996). Among the crustal material components, as the SiO₂ content increases, Poisson’s ratio decreases. The Poisson’s ratio of acidic rocks with felsic contents as the main component is σ ≤ 0.26, the Poisson’s ratio of neutral rocks is 0.26 < σ ≤ 0.28, and the mafic–basic rocks are 0.28 < σ ≤ 0.3. If the Poisson’s ratio σ > 0.3, the study area may be a tectonic fracture zone, the area may have a fracture undergoing strong serpentinization, or it may be experiencing regional partial melting (Ji et al., 2009; Gong et al., 2015). We calculated the obtained H-k results by formula (4) to obtain the Poisson’s ratio (σ) below each station, and the calculation results are shown in Supplementary Table S1. Through the method of linear interpolation, we obtained the average crustal Poisson’s ratio distribution results in the study area (Figure 7). The Poisson’s ratio results (Supplementary Table S1; Figure 7) show that the overall Poisson’s ratio in the study area is low, and the average Poisson’s ratio in the study area is 0.26, but there are local high Poisson’s ratio anomalies in the Longzhong block and Linxia block (Figure 7). The Poisson’s ratio of the North Qinling fault zone and its adjacent areas is small, ranging from 0.21 to 0.25, but the Poisson’s ratio of station 62,359 is 0.28, showing a local anomaly, which may be related to its deep structural fragmentation. The Poisson’s ratios of the Linxia block range from 0.24 to 0.29, and the Poisson’s ratios of 50% of the stations are less than 0.26. There are small-scale elliptical high Poisson’s ratio anomalies (Figure 7), which may indicate that a small amount of oceanic basalt is preserved in its deep crust. The distribution area of high Poisson’s ratio anomalies in the Longzhong block is relatively large, and Poisson’s ratios range from 0.24 to 0.31, showing an M-shaped east‒west distribution (Figure 7). This distribution may be due to the preservation of the oceanic crust basalt composition after the subduction of the oceanic crust. As a component of the ancient craton, the Ordos block has a lower Poisson’s ratio overall.

The Maxianshan fault is an important boundary fault that divides the Linxia block and the Longzhong block. The Linxia block has an obvious layered structure in the crust, a developed Conrad interface, and low-velocity anomalies in the middle and lower crust, which may contain saline fluid (Jin et al., 2012) and have the characteristics of a Japanese-type island arc (Guo et al., 2016; Xin et al., 2021). The layered structure of the upper crust in the Longzhong block is obvious, the layered structure of the middle and lower crust is weak and presents weak low-velocity characteristics, the Conrad interface is not developed, and it has the characteristics of ocean-island basaltic crust. The Longzhong block may have been formed by Mariana-type island arcs (Guo et al., 2015). The Conrad interface and the Moho lateral variation in the southwestern margin of the Ordos block are weak, showing the characteristics of a stable craton. The Liupanshan tectonic belt is an important boundary fault that divides the Longzhong block and the Ordos block.

A large number of geophysical research results show that the Linxia block and the Longzhong block are important components of the eastern part of the Qilian Mountain orogenic belt (Gao et al., 2011; Jin et al., 2012; Guo et al., 2016). The tectonic evolution of the eastern section of the Qilian Mountain orogenic belt is complex. The late Palaeozoic–early Mesozoic experienced a complex tectonic evolutionary process in which the Qilian Ocean was closed and different blocks merged. During the Late Ordovician–early Silurian, the Linxia block of the Japanese-type island arc and the Longzhong block of the Mariana-type island arc were spliced together due the closure of the Qilian Ocean (Cheng et al., 2016; Xin et al., 2021). The Maxianshan Fault and the Liupanshan Fault, as the dividing line between different blocks, played an important restrictive role in late Mesozoic and Cenozoic tectonic shortening and plateau uplift. The Hezuo–Dajing profile of Jin et al. (2012) resolved a conductive layer in the crust corresponding to partial melt and saline fluids. Xia et al. (2019) resolved a structure consisting of several irregular high-resistivity blocks and low-resistivity layers in the crust, which were thought to be the sources of the earthquakes in this region. The middle and lower crust of the Linxia block is characterized by low-resistivity anomalies, while the upper crust is dominated by high-resistivity anomalies (Xin et al., 2021). The results of the deep seismic reflection profile show that the Linxia block has developed obvious extrusive fold deformation (Guo et al., 2016). Therefore, we speculate that the high-conductivity and low-velocity structure of the lower crust of the Qinling orogenic belt and the Linxia block may be caused by structural fragmentation caused by multistage tectonic deformation and filled with salinized fluids. The study area is affected by the remote effect of the collision between the Indian continent and the Eurasian continent, and the seismic sources of the Linxia block are mostly distributed in the upper crust (Xin et al., 2021), which may be because the rigid upper crust more easily stores and releases energy, thus triggering earthquakes; in addition, the existence of high-conductivity, low-velocity bodies in the middle and lower crust is more likely to cause ductile deformation and vertical thickening. The Japanese-type island arc is a product of oceanic and continental subduction. The oceanic crust below the continental plate subducts to the side of the continental margin. It is often the product of a compressional collisional environment. The dehydration of the subducted oceanic crust generally leads to partial melting or upwelling of the upper mantle material. According to the MT results, there is a high-resistance upper crust in the Linxia block, and the middle and lower crust have the characteristics of high-low conductivity interlayers (Xin et al., 2021), which correspond well to our CCP results (Figure 6C). In the Cenozoic, due to the joint control of the Maxianshan fault and the fault at the northern margin of the West Qinling, the Cenozoic sedimentary strata in the Linxia Block were nearly horizontal. However, due to the shortening of the NE-trending structure on the northeastern margin of the Qinghai–Tibet Plateau, overall uplift occurred, resulting in the loss of Quaternary strata (Guo et al., 2016). In summary, we believe that the crustal layered structure of the Linxia block is obvious, and the low-velocity layer of the lower crust is developed. This low-velocity layer may represent the ductile deformation area of the middle–lower crust caused by the subduction of the Qilian Ocean crust, which contains saline fluids. In the Mesozoic–Cenozoic, this area experienced multiple stages of structural shortening deformation, and the upper crust was dominated by thrust napping and shortening deformation (Darby and Ritts, 2002). This thrust nappe structure corresponds well with the low-velocity layers of the upper crust of the Linxia block and the Longzhong block (Figure 6C). The middle–lower crust is dominated by ductile deformation with lateral shortening and longitudinal thickening.

The P-wave receiver function method is more sensitive to horizontally layered velocity discontinuities in the crust, the layered structure of the whole crust is present in the Linxia block, and the CR is developed. However, the layered structure of the upper crust of the Longzhong block is obvious, the layered interface of the middle and lower crust is not developed, and the CR is not developed. The MT research results indicate that the Longzhong block shows a high-resistance anomaly (Figure 6B; Xin et al., 2021). We believe that the high-resistivity body with no layered structure in the middle and lower crust of the Longzhong block may be related to oceanic island basalts during the formation of Mariana-type island arcs, which are the result of oceanic lithosphere subduction beneath oceanic plates and are characterized by widespread basalts (Hart et al., 1972). In summary, we believe that the Linxia block and the Longzhong block have different crustal and electrical structures, mainly because the Linxia block and the Longzhong block belong to different blocks. The Linxia block may have originally belonged to the continental margin of the Qaidam–Qilian block, which formed a Japanese-type island arc due to the subduction of the Qilian Ocean. The Longzhong block belongs to the Mariana-type island arc that formed during the subduction of the oceanic plate. With the closure of the Qilian Ocean, the Maxianshan fault became the collisional boundary of these two ancient island arcs and now exists in the form of the Maxianshan fault. The mafic dikes found in the Maxianshan Group record this splicing process (He et al., 2008). Our research results show that there are obvious differences in the crustal structures of the Linxia block and the Longzhong block on both sides of the Maxianshan fault (Figure 6). The research results of rock geochemistry (Xiao et al., 2009), MT profiles (Xin et al., 2021) and deep seismic reflection profiles (Guo et al., 2016) all support our conclusion.

In summary, the results of seismic detection and MT research show that there are obvious differences in the crustal structure between the Linxia block, the Longzhong block, and the Ordos block. The layered structure of the upper crust in the Longzhong block is developed, but the layered velocity discontinuity of the middle and lower crust is not developed, and it may be composed of Mariana-type island arcs; the layered structure of the Linxia block is developed, and the Conrad interface is developed. The middle and lower crust have obvious low-velocity and high-conductivity anomalies, which may contain saline fluid and have the characteristics of Japanese-type island arcs. As the boundary fault between the Linxia block and the Longzhong block, the Maxianshan fault has obvious sharp changes in the seismological structure and electrical conductivity. The Haiyuan–Liupanshan fault zone has well-developed thrust nappe structures and frequent seismic activities, which may represent the collisional and splicing boundary between the Ordos block and the Longzhong block (the Mariana-type island arc) on the western margin of the North China block.

The eastern segment of the Qilian orogenic belt is composed of the Linxia block and the Longzhong block, which is now represented by the geomorphic features of the basin. It is located at the forefront of the northeastern margin of the Qinghai–Tibet Plateau, comprising a “rhombic-shaped area” controlled by multiple large, deep boundary faults, where the Qilian–Qaidam, western Qinling, Alxa, and Ordos blocks meet (Song et al., 2006; Figures 1A,B). The Haiyuan and Tianjingshan faults control the contact between the Longzhong block and the Alxa block. The Longzhong block is connected to the Ordos block in the east and divided by the Liupanshan fault. The western Qinling northern margin fault is the boundary between the Linxia block and the western Qinling block (Figure 1A). The Maxianshan fault system, where a pillow-shaped basaltic rock belt is exposed, forms a boundary between the Linxia block and the Longzhong block (Figure 1B). This typical suture mark implies that the Maxianshan fault is an important structural boundary separating the Linxia block and the Longzhong block (Pei et al., 2006; Xin et al., 2021). From the early Palaeozoic to the Mesozoic, the area experienced the collision and splicing of the North China plate and the Qilian–Qaidam block, corresponding to the subduction and closure of the Qilian Ocean’s crust (Wu et al., 1993; Xu et al., 1994; Yang et al., 2002; Song et al., 2006). Evidence from geochemical, chronological, and field geological surveys shows that this area experienced significant crustal shortening and thickening events during the Late Triassic–Early Cretaceous (Lin et a, 2011; Dong and Santosh, 2016; Peng et al., 2018; 2019). Influenced by the uplift and expansion of the Qinghai–Tibet Plateau in the Cenozoic, the study area also experienced differential uplift and orogeny. In summary, it is of great scientific significance to clarify the deep structural and evolutionary processes of the northeastern margin of the Qinghai–Tibet Plateau through our research results and combined with previous geochemical and geochronological results.

The closure and collision of the Qilian Ocean occurred between the Alxa block and the Qaidam–Qilian block (Yin and Harrison, 2000; Gehrels et al., 2003; Song et al., 2013). The Qilian Ocean expanded in the late Palaeozoic and late Cambrian and began to subduct in the early Ordovician, forming island arcs and back-arc basins (corresponding to the Longzhong block and Linxia block). During the Late Ordovician–early Silurian, the Qilian Ocean began to close (Cheng et al., 2016). The North Qilian fold belt formed during the Caledonian orogeny, and many ophiolites, subduction zone-related complexes, and high-pressure metamorphic rocks were exposed in the suture zone (Xu et al., 1994; Song et al., 2013; Andrew et al., 2019). The location of this suture in the western part of the North Qilian orogenic belt is relatively clear (Figure 1B; Xu et al., 1994); however, there are few studies on the closure location, closure method, and suture zone location of the Qilian Ocean in the eastern segment of the Qilian orogenic belt (Figure 1B).

The North Qilian fold belt was formed during the Caledonian orogenic period, and its western section is composed of ophiolites, subduction complexes, and high-pressure metamorphic belts, representing the suture location of the Qaidam–Qilian block and the North China block (Xu et al., 1994; Song et al., 2013; Andrew et al., 2019). Although the MT research results across the Liupanshan Mountains show that there is a trench–arc–basin system in the eastern segment of the Qilian orogenic belt, which may correspond to the subduction and closure of the palaeo-ocean in the study area (Han et al., 2016), the suture location, closure method and geological evidence of the Qilian Ocean in the eastern section of the Qilian Mountain orogenic belt are still unclear (Figure 1). The MT research results show that with the Maxianshan fault as the boundary, there are obvious differences in electrical structure between the Linxia block and the Longzhong block (Xin et al., 2021). Our receiver function research results also show significant differences in the crustal structure between the Linxia block and the Longzhong block (Figures 6, 7). Field geological surveys have found that there are a large number of basic complex rocks in the Maxianshan rock group (He et al., 2008). These basic complex rocks were emplaced in the early Silurian and are considered to be important evidence of tectonic magma in the eastern part of the middle Qilian orogenic belt (Cheng et al., 2016), which was likely in place when the early Silurian Qilian Ocean closed. These basic complexes may have also been formed by the upwelling or overflow of mantle material during the subduction of the oceanic crust below the continental margin of the Qaidam–Qilian block. Its formation and emplacement may be related to the subduction and closure of the Qilian Ocean, or it may be a basic complex formed by the upwelling of mantle materials during the closure of the Qilian Ocean (Figure 9A). Altogether, this evidence suggests that mafic rocks are closely related to the subduction and closure of the Qilian Ocean (Cheng et al., 2016). Therefore, the Maxianshan fault located between the Linxia block and the Longzhong block represents the closure boundary of the Qilian Ocean in the eastern segment of the Qilian orogenic belt. In summary, the Linxia block has the characteristics of Japanese-style island arcs, and the Longzhong block is composed of the original Qilian Ocean Mariana-type island arc. The results of MT and deep seismic reflection profiles have revealed this attribute (Guo et al., 2015; Xin et al., 2021).

Significant crustal thickening and rapid uplift occurred in the eastern segment of the Qilian orogenic belt in the Mesozoic and Middle Jurassic–Early Cretaceous (Darby and Ritts, 2002; Dong and Santosh, 2016; Peng et al., 2021; Cheng et al., 2022). Through the collection of previous geological survey data, we drew a comprehensive map of the contact relationship between the Cretaceous strata and the underlying strata in the southwestern margin of the Ordos Basin and a columnar map of the Liupanshan belt (Figures 8A, C). From Figure 8A, it can be concluded that the western margin of the basin has a high-angle unconformity relationship, and the unconformity angle gradually decreases when entering the interior of the basin to the east, indicating that the structural deformation intensity gradually decreases. On the western margin of the Ordos Basin, the Cretaceous strata cover the Jurassic, Triassic, and Palaeozoic strata; inside the basin, the Cretaceous strata cover the Jurassic strata (Figure 8D; Guo et al., 2018b), which shows that from the interior of the basin to the eastern edge of the Qilian orogenic belt, the denudation intensity of the strata gradually enhances. This deformation and exhumation process occurred in the Middle–Late Jurassic–Early Cretaceous (Figures 8A, B; Guo et al., 2018b; Zhao et al., 2019; Peng et al., 2022a). The results of the two-dimensional seismic section across the Liupanshan structural belt show that two west-dipping thrust faults do not cut through the Cretaceous strata (Figure 8C), which indicates that the Liupanshan fault zone was active before the Cretaceous. Sedimentological and structural geological studies have shown that the western margin of the Ordos block has strong thrust nappe deformation, which was the thrust nappe of the East Qilian orogenic belt in the Middle Jurassic–Early Cretaceous (Darby and Ritts, 2002; Liu et al., 2005; Wang et al., 2019; Cheng et al., 2020). Apatite fission track (AFT) results show that the Longzhong block experienced a significant rapid cooling event in the Middle Jurassic–Early Cretaceous (∼174–120 Ma) (Figure 8B; Peng et al., 2019; Peng et al., 2022a; Duvall et al., 2013; Song et al., 2013). The simulation results show that the temperature of minerals has dropped by 60°C–80°C, and the palaeo geothermal gradient of 40°C/km is used as the estimate (Ren et al., 2005; 2017) since the Middle Jurassic–Early Cretaceous Qilian orogenic belt retreated 1.5–2 km (Peng et al., 2022c) to the east; this retreat was the main stage of crustal shortening in the northeastern margin of the Qinghai–Tibet Plateau, and the tectonic deformation events in this period are generally developed in the Qinghai–Tibet Plateau and its periphery (Kapp et al., 2007; Zhu et al., 2016; Dong et al., 2018). In summary, strong crustal shortening occurred in the study area during the Middle Jurassic–Early Cretaceous (∼174–120 Ma), low-velocity layers in the middle and lower crust mainly developed within the ductile shortening deformation, and the upper crust is dominated by brittle shortening deformation of thrust napping (Liu et al., 2005; Liu et al., 2006; Wang, 2007). This tectonic event led to the differential uplift of the Linxia block and the Longzhong block, and the Longzhong block thrusted and napped on the western edge of the Ordos block (Figure 9B).

The late Cenozoic tectonic evolution of the northeastern margin of the Qinghai–Tibet Plateau is generally believed to have occurred along a series of active faults, such as the Haiyuan fault, the Liupanshan fault, and the Maxianshan fault (Zheng et al., 2010; Duvall et al., 2013; Peng et al., 2022b). The results of low-temperature thermochronology show that Nanhuashan experienced rapid uplift at 12 ± 3 Ma (Lin et al., 2010; Wang et al., 2014). The northern section of Yueliangshan experienced rapid uplift at approximately 7 Ma (Peng et al., 2019). From 7.3 to 8.2 Ma, the eastern section of the Haiyuan fault zone and the hanging wall of the Liupanshan thrust–nappe structural belt experienced significant uplift (Zheng et al., 2006; Zhao et al., 2007). All the above evidence indicates that the Haiyuan–Liupanshan fault zone was gradually activated during 15–7 Ma (Zheng et al., 2006; Zhao et al., 2007; Wang W. et al., 2013). However, the AFT of the Longzhong block showed that no significant structural uplift occurred during 15–7 Ma (Figure 7B; Peng et al., 2022c), which means that the rapid uplift event during 15–7 Ma was limited to the local uplift of the Haiyuan–Liupanshan fault zone. That is, the coastal Yuan-Liupanshan fault zone in the eastern section of the Qilian orogenic belt experienced northeast–nearly east‒west peripheral expansion and plateau growth. Due to the limitation of the North China plate and the limited deformation space, it is speculated that Cenozoic structural deformation mostly occurred in the reactivated Late Jurassic structure, indicating that the preexisting structure controlled the Late Cretaceous–Cenozoic deformation activities. The results of MT research across the Madongshan–Liupanshan structural belt show that there are several low-resistivity striped positive flower structures under the high-resistivity background of the southwestwards-dipping strata under the Longzhong block (Zhan et al., 2017). Seismic reflection sections across the Niushoushan–Luoshan fault zone also reveal evidence of the existence of an upper crustal positive flower structure (Wang Q. et al., 2013). Our imaging results show that the thrust nappe deformation zone of the upper crust of the Longzhong Block (low-velocity layer of the upper crust, Figure 6C) passed through the Liupanshan tectonic belt and entered the southwestern margin of the Ordos Block. Since the late Cenozoic, the Haiyuan–Liupanshan fault zone has been dominated by sinistral strike-slip movement because the deep part of the Liupanshan fault zone was limited by the stable and rigid crystallization base of the Ordos massif (e.g., Wang W. et al., 2013), which has further preserved the evolution of the Zhenghua structure, creating the current geomorphological and structural features, such as Liupanshan and Xiaoguanshan (Figure 9C).

In summary, based on previous geochemical results of rocks and our P-wave receiver function results, we believe that the tectonic evolutionary process of the Linxia block, Longzhong block, and southwestern margin of the Ordos block in the study area was as follows:

1) In the late Palaeozoic–late Cambrian, between the Qaidam–Qilian block and the North China plate, the Qilian Ocean was still an ocean basin, the Qaidam–Qilian block was an active continental margin, and the North China block was a passive continental margin. After the Early Ordovician, the Qilian Ocean began to subduct westwards, the Japanese-style island arc initially formed, and the middle and lower crusts in the eastern margin of the Qaidam–Qilian block began to undergo structural shortening and structural uplift under the stress of passive compression. At the same time, the oceanic lithosphere began to subduct under the Qilian Ocean block, forming a Mariana-style island arc, and the Longzhong block first appeared (Figure 9A). 2) In the Late Ordovician-early Silurian, with the closure of the Qilian Ocean, the oceanic crust continued to subduct, and the Mariana-style island arc and Japanese-style island arc collided. The collisional position of the two island arcs may be the initial position of the Maxianshan fault, accompanied by the subduction of the oceanic crust, and ferromafic mélanges were produced near the Maxianshan fault zone under compressive stress. From the late Palaeozoic to Mesozoic, the eastern part of the Qilian Ocean entered the stage of arc–continent collisional deformation, and the Maxianshan fault and Liupanshan fault zone in the study area developed Triassic magmatic rocks, indicating that the combined postcollisional block zone provided a channel for magma emplacement. During the Middle Jurassic–Early Cretaceous, the Linxia block and the Longzhong block experienced differential and rapid uplift, with horizontal and lateral shortening of 15% (Darby and Ritts, 2002). The tectonic uplift events of this period are recorded in the entire Qilian Mountain orogenic belt (Figure 9B). 3) Since the Cenozoic (∼15–7 Ma), the tectonic evolution of the study area has been affected by the long-distance effect of the continental collision between the Indian plate and the Eurasian plate, and the tectonic activities are mainly carried out along active faults, such as the Yuan–Liupanshan fault, the Maxianshan fault, and the Northwest Qinling fault. Due to the limitation of the North China plate and the limited deformation space, the late Mesozoic boundary faults were reactivated and controlled by the positive flower structure, the Haiyuan–Liupanshan fault zone was partially uplifted, and the Liupanshan structural belt was initially formed. As the northeast edge of the plateau continued to grow and expand to the periphery, the Haiyuan–Liupanshan fault zone was dominated by sinistral strike-slip movement, which promoted the development of the positive flower structure in the eastern margin of the Longzhong block and shaped the current geomorphic features of Madongshan and Liupanshan (Figure 9C).