Exhumation of the Qilian Shan and Miocene activity of the Haiyuan fault: insights from apatite (U–Th)/He thermochronology in the Laolongwan basin, northeastern Tibetan Plateau

The Cenozoic uplift of the Tibetan Plateau is the result of ongoing collision between India and Eurasia, which has affected regions more than 2000 km away north of the collision zone (England and Houseman 1986; Dewey et al. 1988; Tapponnier et al. 2001; DeCelles et al. 2002; Molnar and Stock 2009). Understanding the dynamics of the Tibetan Plateau and surrounding areas is essential to comprehending intracontinental deformation (England and Houseman 1986; Tapponnier et al. 2001; Royden et al. 2008), surface uplift mechanisms of orogenic plateaus (Royden et al. 1997; Clark and Royden 2000; Wang et al. 2012; Tian et al. 2013, 2018; Zhao et al. 2023) and their influence on regional and global climate change (Guo et al. 2002; Dupont-Nivet et al. 2007). Two end-member models of how the Tibetan Plateau has deformed during the Cenozoic time have been proposed recently (England and Houseman 1986; Tapponnier et al. 2001; Zhang et al. 2004). One is the continuous deformation model, in which the Tibetan Plateau features distributed shortening and uplift (e.g. England and Houseman 1986). The other is the discrete block model where deformation is largely concentrated along large-scale strike-slip faults that separate rigid blocks, emphasizing the significant role played by these faults (e.g. Tapponnier et al. 2001). Activities of these faults may have controlled the formation of basin-mountain distributions and first-order morphology of the plateau and its surrounding areas (Yuan et al. 2013). It is critical that detailed models regarding the activities of major faults and related intermountain basins are developed in order to reveal the uplift and expansion of the Tibetan Plateau, particularly for the Cenozoic (Burchfiel et al. 1989; Clark et al. 2010; Duvall et al. 2013).

The Qilian Shan region is located along the northeastern plateau margin with an average elevation of ∼4000 m and an area of around 250 000 km² (Fig. 1). Its Cenozoic structural evolution history is key to testing models of the Tibetan Plateau upward and outward growth mechanisms (Wang et al. 2020; Ma et al. 2023). Previous studies inferred the uplift history of the Qilian Shan based on proxy data that record exhumation in ranges by thermochronological methods along the mountain ranges’ steep edges along major thrust systems (e.g. Zheng et al. 2017; Zhuang et al. 2018; Pang et al. 2019) or basin analysis methods, such as detrital zircon U–Pb age distributions, accumulation rate changes and basin environment changes from stratigraphic records within the basins (e.g. Wang et al. 2017; An et al. 2018).

The Haiyuan fault is a major left-lateral strike-slip fault system that runs 1000 km east–west from the west portion of the Qilian Shan in the west to the Liupan Shan region in the east (Fig. 1). The Cenozoic activity of the Haiyuan fault governed the deformation of the Qilian Shan in the northeastern Tibetan Plateau. Since the 1980s, many studies have focused on the distribution of surface rupture created by the 1920 Haiyuan M_w 7.9 earthquake in the east part of the Haiyuan fault (Deng et al. 1986; Zhang et al. 1988) and left-lateral strike-slip rates of the Haiyuan fault zone (Li et al. 2009; Yao et al. 2022). Recently, some thermochronological studies were carried out in the western and eastern segments of the Haiyuan fault to reveal the exhumation/cooling history related to fault activation (Fig. 1; Zheng et al. 2006; Lin et al. 2011; Duvall et al. 2013; Pan et al. 2013; Li et al. 2019; Yu et al. 2019a; Wang et al. 2020). However, there is still a gap of about 400 km in the middle part of the Haiyuan fault without any thermochronological data (Fig. 1). This incomplete knowledge limits our understanding of the Cenozoic tectonic evolution history of the Haiyuan fault zone and Qilian Shan range in the outmost region of the northeastern margin of the Tibetan Plateau.

This study provides detailed logs of sedimentary strata in the Laolongwan basin, which is located in the middle sector of the Haiyuan fault zone. It has been developed in close association with fault activities (Tian et al. 2001; Ding et al. 2004; Liu 2020). New apatite (U–Th)/He data with thermal history modelling results from the Laolongwan basin clearly show that Jurassic to Cretaceous rapid exhumation and middle Miocene accelerated exhumation occurred along the Haiyuan fault zone. These results provide important insights into tectonic deformational histories of the Qilian Shan and the Haiyuan fault zone.

The Qilian Shan is located between the Qaidam block to the south and Alxa block to the north (Fig. 1). The Qilian Shan has been thought to absorb approximately 300–500 km of north–south crustal shortening when the Tibetan Plateau progressively propagated northwards during the Cenozoic (Yin and Harrison 2000). This region has recorded geological history of a Paleozoic arc–continent collision (Hsü et al. 1995), followed by Mesozoic regional extension and intense erosion in the late Mesozoic (Zuza et al. 2018). After a period of tectonic quiescence during the late Cretaceous to the pre-Miocene, the Qilian Shan has experienced contractional deformation and uplift in the course of India–Eurasia collision, marked by active NW–SE-trending thrust faults in the southern and northern range flanks and strike-slip faulting in the central part of the Qilian Shan (Yin et al. 2008b; Yuan et al. 2013; Wang et al. 2017; Zheng et al. 2017).

The Haiyuan fault plays a crucial role in accommodating Cenozoic crustal deformation. According to geological mapping and previous research, the fault has undergone various stages of Cenozoic deformation, including a stage of northward thrusting and associated folding followed by a stage of left-lateral strike-slip faulting (Burchfiel et al. 1991; Zhang et al. 1991; Zheng et al. 2006; Wang et al. 2020). The Haiyuan fault acts as the geomorphic and structural boundary of the northeastern plateau margin (Molnar and Tapponnier 1975; Zhang et al. 1988).

The Longzhong basin in the northeastern portion of the topographic front on the northeastern Tibetan Plateau is a basin bounded by faults. The West Qinling fault marks its southern boundary, while the Riyueshan fault and the Haiyuan fault mark its western and northern boundaries, respectively, and the Liupanshan fault marks its eastern boundary (see Fig. 1). The West Qinling, East Qilian Shan and Liupan Shan define the extent of the Longzhong basin topographically (Fig. 1). Late Cenozoic deformation, characterized by folding and thrusting along the Laji-Jishi Shan, Maxian Shan and Liupan Shan, has produced a series of narrow ranges that divide the Longzhong basin into multiple sub-basins, such as the Xunhua basin, Linxia basin, Xining basin, Lanzhou basin and Laolongwan basin (Fang et al. 2003; Wang et al. 2016a).

Our study focuses on the Laolongwan basin, the largest Cenozoic sedimentary basin developed within the Haiyuan fault in the northeastern Tibetan Plateau (Tian et al. 2001; Ding et al. 2004; Liu 2020). The southern boundary of the Laolongwan basin is defined by the eastward extension of the Laohu Shan frontal fault, while the northern boundary is marked by the Hasi Shan frontal fault. The Zihong Shan fault runs through the basin, dividing the basin into northern and southern parts (Fig. 2). These faults belong to the Haiyuan fault system, which have been active since the Quaternary (Liu et al. 2018; Yao et al. 2022).

During the Cenozoic, the northeastern expansion of the Tibetan Plateau led to the uplift of surrounding mountain ranges and the formation and activation of large boundary faults, forming thousands of metres of sedimentary sequences (Pares et al. 2003; Fang et al. 2005, 2019; Yin et al. 2008a; Craddock et al. 2011; Wang et al. 2011, 2016b, 2017, 2022a; Zhang et al. 2012; Ke et al. 2013). These clastic materials provide valuable information on Cenozoic tectonic processes of the northeastward expansion of the Tibetan Plateau and some large boundary faults, such as the Haiyuan fault. Cenozoic sediments in the study region are separated into six units (Fig. 2; GBGP 1973; Tian et al. 2000; Liu 2020).

Unit1(E₂): In the northern part of the basin, there are thick layers of purple–orange mudstone, argillaceous siltstone and sandstone interbedded with thin-layered and reticular gypsum layers and veins. Mudstone and argillaceous siltstone are mostly massive with a lateral extension of more than 50 m and local horizontal beddings. Grain sizes coarsen upwards in strata. This unit may reflect shallow lake–dry saline lake depositional environments.

Unit2 (E₃): In the northern part of the basin, purple–orange mudstone, argillaceous siltstone and sandstone occur interbedded with medium-thick layers conglomerates. In conglomerate layers, clasts are well sorted, subrounded and distributed in lenses. Grain sizes within conglomerate become smaller moving up-section. This unit may have been deposited in a meandering river environment.

Unit3(N₁): In the southern part and periphery of the basin, there are thick brick-red–orange medium and coarse sandstone interbedded with fine conglomerate containing medium pebbles and thin-layered mudstone. The sandstone develops thick parallel bedding and trough cross-bedding. This unit is consistent with a fluvial facies.

Unit4(N₂^a): In the southern basin, fine-grained conglomerate containing medium-grained gravel is interbedded with thin orange–red coarse sandstone. These fine-grained conglomerates are poorly sorted and subrounded–subangular. It has the characteristics of alluvial fan facies deposition.

Unit5(N₂^b): In the central and eastern basin, rocks are dominantly thick layers of fine sandy siltstone, argillaceous siltstone interbedded with thick layers of medium and coarse sandstone, and fine conglomerate containing medium pebbles, consistent with alluvial plain deposits.

Unit6(Q): Quaternary alluvial deposits are seen in the central and northern basin, deposited in the foothills and along the Yellow River. They are mainly composed of poorly cemented gravel layers with poor layering, intercalated lenticular sand layers and angular and poorly sorted gravels.

We performed field observations in the Laolongwan basin and chose to document a ∼2000 m-long sedimentary log along the Shuiquan section (Figs 2–4). The depositional ages and stratigraphic division were adapted from the GBGP (1973) and geological mapping of Liu (2020). Sedimentological data are documented, including lithology, rock colour and texture, and sedimentary structures and their vertical variations. The lithofacies analysis followed the criteria of Miall (1996) (Table 1).

We collected seven samples along the Shuiquan section from the northern limb of a basin-wide anticline (Fig. 2b). In the stratigraphic column, our samples space approximately ∼200–300 m from each other. For each sample, several kilograms of sandstone were collected and conventional mineral separation was performed to separate apatite.

The apatite (U–Th)/He system has a closure temperature around 70°C and partial retention zone of ∼40–80°C (Wolf et al. 1996; Farley 2000). It could be used to provide constraints on exhumation histories of shallow crustal levels of 2–5 km. Age-elevation profiles of low-temperature thermochronological data have the potential to offer insights into the thermal and exhumation history of a region and are widely applied to understand the regional deformation in the northern Tibetan Plateau (e.g. Clark et al. 2010). In the study of low-temperature thermochronological data from sedimentary basins, an age–depth approach is often used (e.g. Coutand et al. 2006; Zhang et al. 2020).

Standard procedures were employed to separate heavy minerals from rock samples. Apatite grains for (U–Th)/He analysis were handpicked to avoid visible defects and inclusions. Only euhedral apatite grains that are larger than 60 $μ$m in diameter were analysed. Apatite (U–Th)/He analysis was conducted at the Institute of Geology, China Earthquake Administration using standard procedures (Yu et al. 2019b; Farley and Stockli 2002; Hao et al. 2023).

HeFTy (Ketcham 2005) was used to model apatite (U–Th)/He data. This program is based on a simple Monte Carlo approach, making it powerful to generate and evaluate a large number (usually some tens of thousands) of independent time–temperature (t–T) paths (Vermeesch and Tian 2014). The Radiation Damage Accumulation and Annealing Model (Flowers et al. 2009) was used for helium diffusion.

In the modelling exercise, the present-day temperature was set to 10 ± 10°C. A 10% error was attributed to each uncorrected He age and every inverse model was configured to execute until it identified 100 t–T paths with a good fit, enabling comparisons between model runs. The criteria of goodness-of-fit for thermal paths as ‘good’ and ‘acceptable’ were defined as 0.5 and 0.05 (Ketcham 2005). Several constraint boxes in the t–T space were used to restrict the range of t–T histories generated by a non-learning Monte Carlo algorithm in a straightforward and explicit way (Murray et al. 2022). We performed thermal history modelling with HeFTy under the assumption that all grains may have originated from Paleozoic igneous units in the region, which is consistent with Yin et al. (2008a), Zheng et al. (2010), Wang et al. (2016b) and Zuza et al. (2018). The initial temperature range for the first box was set to 200°C/200 Ma–80°C/155 Ma – slightly older than the oldest AHe age. The second box had a restricted range of 40°C/155 Ma–0°C/25 Ma to account for samples passing through the annealing zone during this period. Finally, the last box was configured as 80°C/25 Ma–0°C/0 Ma due to the uplift of the strata to the surface during the Miocene (Fig. 5a–c). Independent geological data suggest that this basin developed during the Miocene (Tian et al. 2000; Ding et al. 2004), implying that test detrital grains must have been exhumed to the surface before the Miocene (∼25 to 5 Ma). The thermal modelling focuses on testing thermal history under a scenario when grains from a modelled sample shared the same thermal history prior to deposition in the sedimentary basin. In the sedimentary basin, it is unlikely that the AHe system would have been reset during the burial as reflected by the dispersed ages, which are older than the depositional age of this basin. We focus on the samples that yielded a positive correlation between ages and eU values (Fig. 5d) as suggested by Flowers et al. (2009).

The logged section could be subdivided into three units from the oldest to youngest (Fig. 2) and 11 lithofacies following Miall (1996) (Figs 3, 4; Table 1).

Lower section: It is mainly thick orange–red-brick lithic quartz medium, coarse sandstone interbedded with conglomerate and purple–red mudstone. The formation develops multiple conglomerate–coarse sandstone–sandstone–mudstone cycles, which gradually becomes finer in the upward direction. The conglomerate is poorly sorted clast-supported, and imbricate structures could be seen, with an average clast size of 2–5 cm and subrounded, and a scour-fill structure at the bottom. The thick sandstone has repeating large parallel bedding and trough cross-bedding, reflecting that the hydrodynamic conditions were strong and the deposition rate was fast during deposition. The mudstone is purplish red and has horizontal bedding. Sand and mudstone lenses can be seen in the section. Based on these observations, it appears that the clastic materials were most likely deposited within a braided river system.

Middle section: It contains pebble, cobble, clast-supported and matrix-supported conglomerates, with a lenticular geometry and disorganized texture. The composition of the conglomerates includes mainly quartz particles, grey, purple–red metamorphic sandstone, green metamorphic rock and black siliceous rock. The maximum clast size of the gravel is 25 cm, and the average clast size is 1–3 cm. This layer exhibits an unconformable relationship with the underlying layer, with the conglomerate's unconformable contact positioned above the sandstone. These lithofacies exhibit lenticular geometry and disorganized texture, and both contain poorly sorted, angular to subrounded clasts. The clast-supported conglomerates show lenticular geometry and erosive bases, indicating that they were most likely deposited as channel fills, while the matrix-supported conglomerates could have been formed due to gravity flows (Miall 1996). Hence, the deposits in the middle section could represent an alluvial fan environment.

Upper section: It mainly contains thick brick-red gravelly sandstone, conglomerate interbedded with medium sandstone, yellow siltstone and purple–red mudstone. The mudstone and sandstone exhibit lenticular geometry. Conglomerates within the upper section are matrix-supported, poorly sorted and sub-angular to angular. It has the characteristics of rapid debris transport and deposition. The upper section was interpreted as deposited in an alluvial plain environment.

A total of 33 apatite grains were analysed (Table 2). For each sample, four to five grains were dated. Our results have modest reproducibility, reflecting the detrital nature of these samples. It was observed that, among the 4–5 apatite grains dated from each sample, 3–4 would yield dates as a cluster, enabling us to provide a sample-average age (Table 2). The relationship between eU and age is also examined (Fig. 5d). Results from SQ-4, 6, 7 show a positive correlation between eU and age, implying that the accumulation of radiation damage might have played a role in controlling the data dispersion (Flowers et al. 2009). Other samples show a wide range of complexities. For example, SQ-3 exhibits a negative correlation, which may be attributed to the presence of microinclusions with high eU, fluid inclusions or other complex factors (Flowers et al. 2009). There is no statistically significant correlation between grain size and ages, which indicates no evidence of the effects of grain size on He retention in the data (Fig. 5e).

Table 2 displays the AHe mean ages and unprocessed data. By eliminating the dispersed ages, the age range of 23 grains was narrowed down from 59.71 to 150.77 Ma (Table 2). The average age for each sample was then calculated as follows: SQ-1 with 78.82 ± 3.10 Ma, SQ-2 with 113.87 ± 12.98 Ma, SQ-3 with 135.65 ± 15.72 Ma, SQ-4 with 105.66 ± 11.61 Ma, SQ-5 with 72.86 ± 11.71 Ma, SQ-6 with 94.57 ± 20.55 Ma and SQ-7 with 121.07 ± 7.75 Ma (Table 2).

To constrain the thermal and exhumation history of the Laolongwan basin and the activity history of Haiyuan fault, t–T paths of these seven samples were modelled (Fig. 5a–c; Fig. S1). We note that SQ-4, 6, 7 yielded ages that are positively correlated with their eU (Fig. 5d). This observation suggests that the accumulation of radiation-induced damage may have exerted an influential role in the modulation of data (Flowers et al. 2009; Stanley and Flowers 2023). Nevertheless, other samples were also modelled with the same constraints as used in SQ-4, 6, 7 in order to further explore the complexities of the dataset (Fig. S1). The Neogene cooling episode inferred from unreset ages requires careful consideration. To address this, we have provided additional information in the Supplementary material (Fig. S1) regarding the boundary conditions used in our HeFTy modelling. We note that this result is modelled from data without much imposed constraints in the t–T space. For samples, we removed age outliers and focused on using the analyses that are close to the mean age for each sample (Table 2). ‘Acceptable’ paths were produced for all samples. However, it is noted that SQ-2, SQ-3 and SQ-5 do not have any ‘good’ paths. SQ-1, SQ-4, SQ-6 and SQ-7 yielded large numbers of ‘good’ paths (Fig. 5a–c; Fig. S1).

All seven samples from our study show a three-stage cooling history: (1) Middle Jurassic to Late Cretaceous (173–86 Ma) rapid cooling; (2) early Cenozoic to middle Miocene burial event; and (3) cooling event at around 13–16 Ma.

Our thermal history modelling was conducted to test key tectonic events in the region during the Mesozoic to Cenozoic.

First, results are consistent with cooling during the Middle Jurassic to Late Cretaceous and provide refined cooling paths (Fig. 5a–c). Relatively fast cooling rates are observed in the early stage during the cooling period from 173 to 86 Ma. Our results are consistent with results from zircon and apatite fission track analysis from Qilian Shan, which observed Jurassic cooling events across a large area in the northern Tibetan Plateau (Jolivet et al. 2001; Li et al. 2019). Our results also confirm recent (U–Th)/He thermochronological results from the middle sector of the Qilian Shan, with rapid cooling during the Late Jurassic–Early Cretaceous (Ma et al. 2023). This cooling phase may represent the far-field influence of the Lhasa block's convergence with the Eurasian plate, occurring along the Bangong–Nujiang suture zone (Allegre et al. 1984; Chengfa et al. 1986; Jolivet et al. 2001; Hu et al. 2022). During the Early Jurassic (∼190–145 Ma), the Bangong–Nujiang Ocean initiated its northward subduction, which was succeeded by the collision between the Qiangtang and Lhasa continental blocks during the Early Cretaceous Period (∼140–120 Ma). By the earliest Late Cretaceous Period, a significant shift occurred as seaways disappeared, giving way to widespread deposition of continental red beds along the Bangong–Nujiang suture zone. This transition marked the commencement of intracontinental convergence, which ultimately led to the initial uplift of the Tibetan Plateau (Hu et al. 2022). Alternatively, this event might reflect the regional extension in the course of the Mesozoic closure of the Meso-Tethys (Vincent and Allen 1999; Zhang et al. 2014; Kapp and DeCelles 2019; Zhao et al. 2020). The onset of magmatism in the northern Lhasa terrane at ∼170 Ma, coinciding with the amalgamation of the Amdo and Qiangtang terranes, may indicate initiation of divergent double subduction (northward and southward) of the Meso-Tethys oceanic lithosphere (e.g. Kapp and DeCelles 2019). This bivergent double subduction is the cause of the extensional stress field in the Middle Jurassic Qaidam area (Zhao et al. 2020). Simultaneously, this might have led to the generation of some far-field extensional basins in the northern Tibetan Plateau, encompassing the Jungar, Tarim, Dunhuang, Tula, western Qaidam, Hexi and Ordos basins (e.g. Vincent and Allen 1999; Dai et al. 2023).

Second, our results imply that the development of a shallow sedimentary basin may have taken place during the Cretaceous to Miocene (Fig. 5a–c). AHe and AFT data imply that the north of the Haiyuan fault and the western Lenglongling region in the eastern Qilian Shan might have undergone a low degree of exhumation during ∼150 and 15 Ma (Wang et al. 2020). Our results are consistent with a scenario in which the broad eastern Qilian Shan was relatively less active during the Cretaceous to Miocene (Zheng et al. 2010; Wang et al. 2016b; Li et al. 2019; Pang et al. 2019; Ma et al. 2023).

Third, our modelling reveals that significant cooling took place around 13–16 Ma. This more recent pulse of exhumation could be explained by the activities of high-angle thrust faulting that occurred locally in the middle Miocene (Li et al. 2019; Ma et al. 2023). We further interpret this observation in the context of regional activities of the Haiyuan fault.

Our data reveal that the Laolongwan basin may have undergone exhumation in the middle Miocene, marking the activity of thrusting along the middle sector of the Haiyuan fault. This scenario is consistent with existing data from the western and eastern sectors of the Haiyuan fault.

In the west, apatite fission track (AFT) data from a right-step restraining bend along the Haiyuan fault suggest the initiation of the Haiyuan fault between 15 and 10 Ma (Li et al. 2019). Thermochronological data from Lenglongling region also indicate that rapid exhumation associated with the fault occurred during the middle Miocene, which further constrains the activity time of the Haiyuan fault to have occurred during this period (Wang et al. 2020). Based on the timing of Dulan-Chakka highland deformation, the segments of the Haiyuan faults were inferred to be active at ∼15 Ma (Duvall et al. 2013). AFT data from the western section of the Haiyuan fault also revealed a rapid cooling event during the Miocene (Pan et al. 2013; Yu et al. 2019a). Although it is still a debate whether the Haiyuan fault was initiated by strike-slip (Lin et al. 2011; Duvall et al. 2013; Li et al. 2019) or thrust faulting (Zheng et al. 2006; Wang et al. 2013, 2020; Pang et al. 2019; Yu et al. 2019a), this evidence indicates that the Haiyuan fault had begun intense tectonic activity during the Miocene.

In the east, AFT data in the Liupanshan region provide a ∼8 Ma rapid cooling event, which was interpreted as a response to the rapid unroofing of the Liupanshan thrust fault in the late Miocene (Zheng et al. 2006). Changes in the stratigraphic architecture and the results of AFT analyses indicate rapid cooling and implicit erosional exhumation in the hanging wall of the Haiyuan-Liupan Shan fault belt (Lin et al. 2011). This provides evidence for a more intense phase of tectonic activity during the late Miocene, consistent with Zheng et al.’s (2006) findings of the most recent shortening phase on the Liupan Shan fault. Miocene thrusting along the Haiyuan fault is also supported by renewed accumulation in the southern Ningxia basin as a result of flexural subsidence in the footwall of the fault (Wang et al. 2013).

Collectively, our data are consistent with a scenario in which exhumation/deformation from the mountain ranges and basins took place along the entire Haiyuan fault, marking its activity, during the late Miocene.

Several lines of evidence suggest deformation and growth of the plateau to its recent northeastern margin at around 10 Ma (e.g. Wang et al. 2022a), joining a growing body of data that reveals the significant deformation broadly located in the northeastern margin of the Tibetan Plateau and pronounced crustal shortening during the Miocene (Yuan et al. 2013).

First, thermochronological data show accelerated mountain exhumation during 18–10 Ma, suggesting the uplift of Qilian Shan. For instance, thermochronological data from the Jinfosi granitic pluton show slow exhumation between the late Mesozoic and ∼18 Ma, followed by rapid exhumation since ∼10 Ma (Zheng et al. 2010). According to AFT age-elevation transects from the Tuolai Shan, there was a rapid exhumation that began at approximately 17–15 Ma (Yu et al. 2019a). AFT samples from Lenglongling also show a significantly accelerated exhumation at ∼15 Ma (Wang et al. 2020). Along the southern margin of the Qilian Shan, AFT data from Zongwulong Shan indicate accelerated exhumation since ∼18–11 Ma, implying the southward propagation of the southern Qilian Shan (Pang et al. 2019).

Second, spatial–temporal patterns of deformation are preserved in sedimentary basins around the Tibetan Plateau (Wang et al. 2017; Nie et al. 2019; Wu et al. 2019). Sedimentary records and associated geochronological data of the Jiuxi basin suggest an enhanced average sedimentation rate and sedimentation recycling during the middle Miocene, which was interpreted as a response to the initial uplift of the North Qilian Shan (Wang et al. 2016b). Detailed magnetostratigraphic studies show that the accumulation of coarse conglomerate began ∼14 Ma ago, indicating that tectonic activities and associated uplift started during the middle Miocene (Sun et al. 2005).

Third, documents on active faulting also provide constraints on the timing of deformation of the northeastern plateau margin. With the assumption of a constant long-term slip rate during the late Cenozoic, the onset of the Elashan and Riyueshan faults is inferred to be ∼10 Ma, highlighting a major stage of deformation in the northeastern margin of the Tibetan Plateau (Yuan et al. 2011).

Extensive tectonic deformation, growth of mountain ranges and increased sedimentation rates have occurred within and around the Tibetan Plateau over the past 10 to 15 Ma (Wang et al. 2022a). Temporal and spatial synchronicities in upward and outward growth are consistent with a scenario in which convective removal of mantle lithosphere took place following the thickening of mantle lithosphere beneath the Tibetan Plateau, leading to plateau uplift and extension (Molnar and Stock 2009). As the Indian and Eurasian plates converge, continuous foundering of convective instabilities beneath the plateau might provide buoyant force to sustain surface uplift whereas compressive stresses applied to periphery of the plateau could promote lateral extension of the plateau (Molnar et al. 1993; Wang et al. 2022a, b).

Apatite (U–Th)/He thermochronological data (AHe) were collected from the middle Haiyuan fault to provide constraints on the deformation of the northeastern margin of the Tibetan Plateau. Seven samples provided 23 clustered single-grain ages ranging from ∼60 to 150 Ma. Single-grain ages show limited correlations with their effective U (eU), except for one sample. Thermal history modelling was carried out with a range of assumptions. Thermal history models support an original exhumation of the potential source rocks of these detrital apatite grains during the Middle Jurassic, followed by basin development during the early Cenozoic. A final exhumation in middle Miocene time was identified. Combined with previous studies, our data support that the Haiyuan fault may have been active in the middle Miocene, which implies that the deformation related to the NE expansion of the plateau was active later in the Cenozoic. Results support plateau growth models that invoke a later, 10 Ma, deformation of the NE Tibetan Plateau following lithospheric thickening and the removal of the lithospheric mantle by convective instabilities.

Thanks to Lei Wu and another anonymous reviewer for their constructive comments. We thank Yuqi Hao and Ying Wang for their analytical support. In addition, Kang Liu wants to thank Xuedong Liu and Ruilan Zhou for their spiritual support.

KL: conceptualization (equal), data curation (equal), writing – original draft (lead), writing – review & editing (equal); WW: conceptualization (equal), data curation (equal), funding acquisition (lead), writing – review & editing (equal); RJ: investigation (supporting); HJ: investigation (supporting); RZ: conceptualization (equal), data curation (equal), writing – review & editing (equal)

This research was supported jointly by the foundation of the Natural Science Foundation of China (Grants 42272248, 42030301) and China Scholarship Council (No. 202206380163).

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

All data generated or analysed during this study are included in this published article and its supporting information files.